{ "best_metric": null, "best_model_checkpoint": null, "epoch": 0.9999611936823315, "eval_steps": 500, "global_step": 12884, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 0.0, "loss": 1.8329, "step": 1 }, { "epoch": 0.0, "learning_rate": 2.3266122046345402e-07, "loss": 1.7418, "step": 2 }, { "epoch": 0.0, "learning_rate": 3.687593098065924e-07, "loss": 1.7542, "step": 3 }, { "epoch": 0.0, "learning_rate": 4.6532244092690805e-07, "loss": 1.7463, "step": 4 }, { "epoch": 0.0, "learning_rate": 5.402226243848765e-07, "loss": 1.2061, "step": 5 }, { "epoch": 0.0, "learning_rate": 6.014205302700464e-07, "loss": 1.7438, "step": 6 }, { "epoch": 0.0, "learning_rate": 6.531626224400069e-07, "loss": 1.7852, "step": 7 }, { "epoch": 0.0, "learning_rate": 6.979836613903621e-07, "loss": 1.6267, "step": 8 }, { "epoch": 0.0, "learning_rate": 7.375186196131847e-07, "loss": 1.7093, "step": 9 }, { "epoch": 0.0, "learning_rate": 7.728838448483306e-07, "loss": 1.5677, "step": 10 }, { "epoch": 0.0, "learning_rate": 8.048755825020159e-07, "loss": 1.6714, "step": 11 }, { "epoch": 0.0, "learning_rate": 8.340817507335004e-07, "loss": 1.2275, "step": 12 }, { "epoch": 0.0, "learning_rate": 8.609488210741464e-07, "loss": 1.6229, "step": 13 }, { "epoch": 0.0, "learning_rate": 8.85823842903461e-07, "loss": 1.6321, "step": 14 }, { "epoch": 0.0, "learning_rate": 9.089819341914687e-07, "loss": 1.6408, "step": 15 }, { "epoch": 0.0, "learning_rate": 9.306448818538161e-07, "loss": 1.5972, "step": 16 }, { "epoch": 0.0, "learning_rate": 9.509940932443214e-07, "loss": 1.58, "step": 17 }, { "epoch": 0.0, "learning_rate": 9.701798400766387e-07, "loss": 1.54, "step": 18 }, { "epoch": 0.0, "learning_rate": 9.883279997180702e-07, "loss": 1.5979, "step": 19 }, { "epoch": 0.0, "learning_rate": 1.0055450653117846e-06, "loss": 1.1854, "step": 20 }, { "epoch": 0.0, "learning_rate": 1.0219219322465993e-06, "loss": 1.5965, "step": 21 }, { "epoch": 0.0, "learning_rate": 1.0375368029654699e-06, "loss": 1.2017, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.052457445538274e-06, "loss": 1.5139, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.0667429711969544e-06, "loss": 1.1292, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.080445248769753e-06, "loss": 1.5651, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.0936100415376005e-06, "loss": 1.514, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.106277929419777e-06, "loss": 1.5673, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.118485063366915e-06, "loss": 1.4855, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.1302637873146459e-06, "loss": 1.556, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.1416431546549228e-06, "loss": 1.5329, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.1526493599901512e-06, "loss": 1.4612, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.16330610231727e-06, "loss": 1.5798, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.1736348923086082e-06, "loss": 1.6281, "step": 33 }, { "epoch": 0.0, "learning_rate": 1.1836553137077756e-06, "loss": 1.5225, "step": 34 }, { "epoch": 0.0, "learning_rate": 1.1933852468248836e-06, "loss": 1.5462, "step": 35 }, { "epoch": 0.0, "learning_rate": 1.2028410605400928e-06, "loss": 1.5172, "step": 36 }, { "epoch": 0.0, "learning_rate": 1.2120377779946798e-06, "loss": 1.513, "step": 37 }, { "epoch": 0.0, "learning_rate": 1.2209892201815241e-06, "loss": 1.4609, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.2297081308807386e-06, "loss": 1.0969, "step": 39 }, { "epoch": 0.0, "learning_rate": 1.2382062857752385e-06, "loss": 1.5036, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.2464945880908681e-06, "loss": 1.5319, "step": 41 }, { "epoch": 0.0, "learning_rate": 1.2545831527100534e-06, "loss": 1.5279, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.2624813803868152e-06, "loss": 1.4456, "step": 43 }, { "epoch": 0.0, "learning_rate": 1.2701980234289238e-06, "loss": 1.5206, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.2777412439980611e-06, "loss": 1.4725, "step": 45 }, { "epoch": 0.0, "learning_rate": 1.285118666001728e-06, "loss": 1.4483, "step": 46 }, { "epoch": 0.0, "learning_rate": 1.2923374214040148e-06, "loss": 1.5063, "step": 47 }, { "epoch": 0.0, "learning_rate": 1.2994041916604086e-06, "loss": 1.453, "step": 48 }, { "epoch": 0.0, "learning_rate": 1.3063252448800138e-06, "loss": 1.4816, "step": 49 }, { "epoch": 0.0, "learning_rate": 1.3131064692332068e-06, "loss": 1.5217, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.3197534030509136e-06, "loss": 1.4677, "step": 51 }, { "epoch": 0.0, "learning_rate": 1.3262712620010546e-06, "loss": 1.4816, "step": 52 }, { "epoch": 0.0, "learning_rate": 1.3326649636762563e-06, "loss": 1.4437, "step": 53 }, { "epoch": 0.0, "learning_rate": 1.3389391498832311e-06, "loss": 1.4649, "step": 54 }, { "epoch": 0.0, "learning_rate": 1.3450982068868923e-06, "loss": 1.4659, "step": 55 }, { "epoch": 0.0, "learning_rate": 1.3511462838303694e-06, "loss": 1.4372, "step": 56 }, { "epoch": 0.0, "learning_rate": 1.3570873095246625e-06, "loss": 1.4831, "step": 57 }, { "epoch": 0.0, "learning_rate": 1.3629250077780998e-06, "loss": 1.4668, "step": 58 }, { "epoch": 0.0, "learning_rate": 1.3686629114153809e-06, "loss": 1.4278, "step": 59 }, { "epoch": 0.0, "learning_rate": 1.3743043751183767e-06, "loss": 1.4226, "step": 60 }, { "epoch": 0.0, "learning_rate": 1.379852587205557e-06, "loss": 1.5118, "step": 61 }, { "epoch": 0.0, "learning_rate": 1.3853105804536054e-06, "loss": 1.4505, "step": 62 }, { "epoch": 0.0, "learning_rate": 1.3906812420531918e-06, "loss": 1.427, "step": 63 }, { "epoch": 0.0, "learning_rate": 1.3959673227807241e-06, "loss": 1.4638, "step": 64 }, { "epoch": 0.01, "learning_rate": 1.4011714454590227e-06, "loss": 1.4513, "step": 65 }, { "epoch": 0.01, "learning_rate": 1.4062961127720621e-06, "loss": 1.4479, "step": 66 }, { "epoch": 0.01, "learning_rate": 1.411343714492071e-06, "loss": 1.4922, "step": 67 }, { "epoch": 0.01, "learning_rate": 1.4163165341712296e-06, "loss": 1.4485, "step": 68 }, { "epoch": 0.01, "learning_rate": 1.4212167553448664e-06, "loss": 1.4685, "step": 69 }, { "epoch": 0.01, "learning_rate": 1.4260464672883377e-06, "loss": 1.4605, "step": 70 }, { "epoch": 0.01, "learning_rate": 1.4308076703655703e-06, "loss": 1.0756, "step": 71 }, { "epoch": 0.01, "learning_rate": 1.435502281003547e-06, "loss": 1.4623, "step": 72 }, { "epoch": 0.01, "learning_rate": 1.440132136323686e-06, "loss": 1.3797, "step": 73 }, { "epoch": 0.01, "learning_rate": 1.444698998458134e-06, "loss": 1.4052, "step": 74 }, { "epoch": 0.01, "learning_rate": 1.449204558576345e-06, "loss": 1.4475, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.453650440644978e-06, "loss": 1.4695, "step": 76 }, { "epoch": 0.01, "learning_rate": 1.458038204942023e-06, "loss": 1.4379, "step": 77 }, { "epoch": 0.01, "learning_rate": 1.4623693513441925e-06, "loss": 1.4372, "step": 78 }, { "epoch": 0.01, "learning_rate": 1.4666453224049103e-06, "loss": 1.3829, "step": 79 }, { "epoch": 0.01, "learning_rate": 1.4708675062386924e-06, "loss": 1.4743, "step": 80 }, { "epoch": 0.01, "learning_rate": 1.4750372392263695e-06, "loss": 1.4475, "step": 81 }, { "epoch": 0.01, "learning_rate": 1.479155808554322e-06, "loss": 1.4521, "step": 82 }, { "epoch": 0.01, "learning_rate": 1.4832244545998196e-06, "loss": 1.4647, "step": 83 }, { "epoch": 0.01, "learning_rate": 1.4872443731735073e-06, "loss": 1.4119, "step": 84 }, { "epoch": 0.01, "learning_rate": 1.4912167176291979e-06, "loss": 1.3972, "step": 85 }, { "epoch": 0.01, "learning_rate": 1.495142600850269e-06, "loss": 1.4409, "step": 86 }, { "epoch": 0.01, "learning_rate": 1.4990230971212384e-06, "loss": 1.4703, "step": 87 }, { "epoch": 0.01, "learning_rate": 1.5028592438923781e-06, "loss": 1.4225, "step": 88 }, { "epoch": 0.01, "learning_rate": 1.5066520434446327e-06, "loss": 1.3966, "step": 89 }, { "epoch": 0.01, "learning_rate": 1.5104024644615152e-06, "loss": 1.3796, "step": 90 }, { "epoch": 0.01, "learning_rate": 1.5141114435141534e-06, "loss": 1.3908, "step": 91 }, { "epoch": 0.01, "learning_rate": 1.517779886465182e-06, "loss": 1.4497, "step": 92 }, { "epoch": 0.01, "learning_rate": 1.521408669796744e-06, "loss": 1.376, "step": 93 }, { "epoch": 0.01, "learning_rate": 1.5249986418674687e-06, "loss": 1.4351, "step": 94 }, { "epoch": 0.01, "learning_rate": 1.5285506241029464e-06, "loss": 1.4001, "step": 95 }, { "epoch": 0.01, "learning_rate": 1.5320654121238625e-06, "loss": 1.4388, "step": 96 }, { "epoch": 0.01, "learning_rate": 1.535543776815681e-06, "loss": 1.3803, "step": 97 }, { "epoch": 0.01, "learning_rate": 1.5389864653434684e-06, "loss": 1.4036, "step": 98 }, { "epoch": 0.01, "learning_rate": 1.5423942021152005e-06, "loss": 1.0499, "step": 99 }, { "epoch": 0.01, "learning_rate": 1.5457676896966612e-06, "loss": 1.3961, "step": 100 }, { "epoch": 0.01, "learning_rate": 1.5491076096808166e-06, "loss": 0.967, "step": 101 }, { "epoch": 0.01, "learning_rate": 1.5524146235143675e-06, "loss": 1.3717, "step": 102 }, { "epoch": 0.01, "learning_rate": 1.5556893732839763e-06, "loss": 1.3962, "step": 103 }, { "epoch": 0.01, "learning_rate": 1.5589324824645085e-06, "loss": 1.4114, "step": 104 }, { "epoch": 0.01, "learning_rate": 1.5621445566314757e-06, "loss": 1.4528, "step": 105 }, { "epoch": 0.01, "learning_rate": 1.5653261841397102e-06, "loss": 1.4008, "step": 106 }, { "epoch": 0.01, "learning_rate": 1.5684779367701741e-06, "loss": 1.3541, "step": 107 }, { "epoch": 0.01, "learning_rate": 1.571600370346685e-06, "loss": 1.3759, "step": 108 }, { "epoch": 0.01, "learning_rate": 1.5746940253242183e-06, "loss": 0.9856, "step": 109 }, { "epoch": 0.01, "learning_rate": 1.5777594273503464e-06, "loss": 1.4227, "step": 110 }, { "epoch": 0.01, "learning_rate": 1.5807970878012719e-06, "loss": 1.3908, "step": 111 }, { "epoch": 0.01, "learning_rate": 1.5838075042938231e-06, "loss": 0.9702, "step": 112 }, { "epoch": 0.01, "learning_rate": 1.5867911611746912e-06, "loss": 1.4589, "step": 113 }, { "epoch": 0.01, "learning_rate": 1.5897485299881166e-06, "loss": 1.4113, "step": 114 }, { "epoch": 0.01, "learning_rate": 1.5926800699231503e-06, "loss": 1.4017, "step": 115 }, { "epoch": 0.01, "learning_rate": 1.595586228241554e-06, "loss": 1.4127, "step": 116 }, { "epoch": 0.01, "learning_rate": 1.598467440687331e-06, "loss": 1.4194, "step": 117 }, { "epoch": 0.01, "learning_rate": 1.601324131878835e-06, "loss": 1.3879, "step": 118 }, { "epoch": 0.01, "learning_rate": 1.6041567156843285e-06, "loss": 1.3792, "step": 119 }, { "epoch": 0.01, "learning_rate": 1.6069655955818308e-06, "loss": 1.4461, "step": 120 }, { "epoch": 0.01, "learning_rate": 1.6097511650040319e-06, "loss": 0.974, "step": 121 }, { "epoch": 0.01, "learning_rate": 1.6125138076690112e-06, "loss": 1.3654, "step": 122 }, { "epoch": 0.01, "learning_rate": 1.6152538978974604e-06, "loss": 1.4293, "step": 123 }, { "epoch": 0.01, "learning_rate": 1.6179718009170595e-06, "loss": 1.3615, "step": 124 }, { "epoch": 0.01, "learning_rate": 1.6206678731546295e-06, "loss": 1.4113, "step": 125 }, { "epoch": 0.01, "learning_rate": 1.6233424625166457e-06, "loss": 1.3984, "step": 126 }, { "epoch": 0.01, "learning_rate": 1.625995908658664e-06, "loss": 0.9458, "step": 127 }, { "epoch": 0.01, "learning_rate": 1.6286285432441783e-06, "loss": 1.3966, "step": 128 }, { "epoch": 0.01, "learning_rate": 1.6312406901934074e-06, "loss": 1.4058, "step": 129 }, { "epoch": 0.01, "learning_rate": 1.6338326659224768e-06, "loss": 1.4127, "step": 130 }, { "epoch": 0.01, "learning_rate": 1.6364047795734332e-06, "loss": 0.9742, "step": 131 }, { "epoch": 0.01, "learning_rate": 1.6389573332355163e-06, "loss": 1.3918, "step": 132 }, { "epoch": 0.01, "learning_rate": 1.641490622158077e-06, "loss": 0.9462, "step": 133 }, { "epoch": 0.01, "learning_rate": 1.6440049349555251e-06, "loss": 1.3678, "step": 134 }, { "epoch": 0.01, "learning_rate": 1.6465005538046534e-06, "loss": 1.3412, "step": 135 }, { "epoch": 0.01, "learning_rate": 1.6489777546346837e-06, "loss": 1.3386, "step": 136 }, { "epoch": 0.01, "learning_rate": 1.651436807310349e-06, "loss": 1.3901, "step": 137 }, { "epoch": 0.01, "learning_rate": 1.6538779758083206e-06, "loss": 1.3969, "step": 138 }, { "epoch": 0.01, "learning_rate": 1.6563015183872617e-06, "loss": 1.3249, "step": 139 }, { "epoch": 0.01, "learning_rate": 1.6587076877517914e-06, "loss": 1.398, "step": 140 }, { "epoch": 0.01, "learning_rate": 1.6610967312106073e-06, "loss": 1.4239, "step": 141 }, { "epoch": 0.01, "learning_rate": 1.6634688908290242e-06, "loss": 1.3837, "step": 142 }, { "epoch": 0.01, "learning_rate": 1.665824403576162e-06, "loss": 1.435, "step": 143 }, { "epoch": 0.01, "learning_rate": 1.6681635014670008e-06, "loss": 1.4146, "step": 144 }, { "epoch": 0.01, "learning_rate": 1.6704864116995222e-06, "loss": 1.3785, "step": 145 }, { "epoch": 0.01, "learning_rate": 1.6727933567871398e-06, "loss": 1.3283, "step": 146 }, { "epoch": 0.01, "learning_rate": 1.6750845546866063e-06, "loss": 1.3706, "step": 147 }, { "epoch": 0.01, "learning_rate": 1.6773602189215877e-06, "loss": 1.3414, "step": 148 }, { "epoch": 0.01, "learning_rate": 1.6796205587020743e-06, "loss": 0.9969, "step": 149 }, { "epoch": 0.01, "learning_rate": 1.6818657790397991e-06, "loss": 1.4461, "step": 150 }, { "epoch": 0.01, "learning_rate": 1.6840960808598226e-06, "loss": 1.3549, "step": 151 }, { "epoch": 0.01, "learning_rate": 1.6863116611084322e-06, "loss": 1.4025, "step": 152 }, { "epoch": 0.01, "learning_rate": 1.688512712857506e-06, "loss": 1.3711, "step": 153 }, { "epoch": 0.01, "learning_rate": 1.6906994254054769e-06, "loss": 1.4257, "step": 154 }, { "epoch": 0.01, "learning_rate": 1.6928719843750278e-06, "loss": 1.4001, "step": 155 }, { "epoch": 0.01, "learning_rate": 1.6950305718076467e-06, "loss": 1.4185, "step": 156 }, { "epoch": 0.01, "learning_rate": 1.6971753662551609e-06, "loss": 1.4642, "step": 157 }, { "epoch": 0.01, "learning_rate": 1.6993065428683642e-06, "loss": 1.3635, "step": 158 }, { "epoch": 0.01, "learning_rate": 1.7014242734828486e-06, "loss": 1.4398, "step": 159 }, { "epoch": 0.01, "learning_rate": 1.7035287267021466e-06, "loss": 1.3399, "step": 160 }, { "epoch": 0.01, "learning_rate": 1.705620067978281e-06, "loss": 1.384, "step": 161 }, { "epoch": 0.01, "learning_rate": 1.7076984596898232e-06, "loss": 1.4493, "step": 162 }, { "epoch": 0.01, "learning_rate": 1.7097640612175484e-06, "loss": 1.3889, "step": 163 }, { "epoch": 0.01, "learning_rate": 1.7118170290177762e-06, "loss": 1.3969, "step": 164 }, { "epoch": 0.01, "learning_rate": 1.7138575166934846e-06, "loss": 1.3598, "step": 165 }, { "epoch": 0.01, "learning_rate": 1.7158856750632738e-06, "loss": 1.4027, "step": 166 }, { "epoch": 0.01, "learning_rate": 1.7179016522282575e-06, "loss": 1.4031, "step": 167 }, { "epoch": 0.01, "learning_rate": 1.7199055936369615e-06, "loss": 0.9415, "step": 168 }, { "epoch": 0.01, "learning_rate": 1.7218976421482927e-06, "loss": 1.4139, "step": 169 }, { "epoch": 0.01, "learning_rate": 1.723877938092652e-06, "loss": 0.9286, "step": 170 }, { "epoch": 0.01, "learning_rate": 1.7258466193312548e-06, "loss": 1.3945, "step": 171 }, { "epoch": 0.01, "learning_rate": 1.7278038213137232e-06, "loss": 0.9339, "step": 172 }, { "epoch": 0.01, "learning_rate": 1.7297496771340066e-06, "loss": 1.3969, "step": 173 }, { "epoch": 0.01, "learning_rate": 1.7316843175846923e-06, "loss": 1.3829, "step": 174 }, { "epoch": 0.01, "learning_rate": 1.73360787120976e-06, "loss": 1.4253, "step": 175 }, { "epoch": 0.01, "learning_rate": 1.7355204643558318e-06, "loss": 1.3696, "step": 176 }, { "epoch": 0.01, "learning_rate": 1.7374222212219731e-06, "loss": 1.3406, "step": 177 }, { "epoch": 0.01, "learning_rate": 1.7393132639080868e-06, "loss": 1.3578, "step": 178 }, { "epoch": 0.01, "learning_rate": 1.7411937124619546e-06, "loss": 1.386, "step": 179 }, { "epoch": 0.01, "learning_rate": 1.7430636849249692e-06, "loss": 1.3663, "step": 180 }, { "epoch": 0.01, "learning_rate": 1.7449232973765948e-06, "loss": 1.3832, "step": 181 }, { "epoch": 0.01, "learning_rate": 1.7467726639776073e-06, "loss": 1.3055, "step": 182 }, { "epoch": 0.01, "learning_rate": 1.7486118970121496e-06, "loss": 0.9751, "step": 183 }, { "epoch": 0.01, "learning_rate": 1.7504411069286361e-06, "loss": 1.322, "step": 184 }, { "epoch": 0.01, "learning_rate": 1.752260402379556e-06, "loss": 0.9425, "step": 185 }, { "epoch": 0.01, "learning_rate": 1.7540698902601978e-06, "loss": 1.2915, "step": 186 }, { "epoch": 0.01, "learning_rate": 1.7558696757463375e-06, "loss": 1.3391, "step": 187 }, { "epoch": 0.01, "learning_rate": 1.7576598623309229e-06, "loss": 1.4092, "step": 188 }, { "epoch": 0.01, "learning_rate": 1.759440551859784e-06, "loss": 1.3364, "step": 189 }, { "epoch": 0.01, "learning_rate": 1.7612118445664005e-06, "loss": 1.3481, "step": 190 }, { "epoch": 0.01, "learning_rate": 1.7629738391057573e-06, "loss": 1.4378, "step": 191 }, { "epoch": 0.01, "learning_rate": 1.7647266325873164e-06, "loss": 1.4115, "step": 192 }, { "epoch": 0.01, "learning_rate": 1.7664703206071319e-06, "loss": 1.4175, "step": 193 }, { "epoch": 0.02, "learning_rate": 1.768204997279135e-06, "loss": 1.3051, "step": 194 }, { "epoch": 0.02, "learning_rate": 1.769930755265615e-06, "loss": 1.3283, "step": 195 }, { "epoch": 0.02, "learning_rate": 1.771647685806922e-06, "loss": 1.3571, "step": 196 }, { "epoch": 0.02, "learning_rate": 1.773355878750411e-06, "loss": 1.4211, "step": 197 }, { "epoch": 0.02, "learning_rate": 1.7750554225786544e-06, "loss": 1.3772, "step": 198 }, { "epoch": 0.02, "learning_rate": 1.7767464044369469e-06, "loss": 1.3887, "step": 199 }, { "epoch": 0.02, "learning_rate": 1.7784289101601149e-06, "loss": 1.354, "step": 200 }, { "epoch": 0.02, "learning_rate": 1.7801030242986635e-06, "loss": 0.9364, "step": 201 }, { "epoch": 0.02, "learning_rate": 1.7817688301442705e-06, "loss": 1.3927, "step": 202 }, { "epoch": 0.02, "learning_rate": 1.7834264097546527e-06, "loss": 1.3777, "step": 203 }, { "epoch": 0.02, "learning_rate": 1.7850758439778218e-06, "loss": 1.3317, "step": 204 }, { "epoch": 0.02, "learning_rate": 1.7867172124757447e-06, "loss": 1.3767, "step": 205 }, { "epoch": 0.02, "learning_rate": 1.7883505937474305e-06, "loss": 1.3476, "step": 206 }, { "epoch": 0.02, "learning_rate": 1.7899760651514585e-06, "loss": 1.3153, "step": 207 }, { "epoch": 0.02, "learning_rate": 1.7915937029279624e-06, "loss": 1.319, "step": 208 }, { "epoch": 0.02, "learning_rate": 1.793203582220086e-06, "loss": 1.3241, "step": 209 }, { "epoch": 0.02, "learning_rate": 1.7948057770949298e-06, "loss": 1.3414, "step": 210 }, { "epoch": 0.02, "learning_rate": 1.7964003605639985e-06, "loss": 1.375, "step": 211 }, { "epoch": 0.02, "learning_rate": 1.7979874046031644e-06, "loss": 0.9251, "step": 212 }, { "epoch": 0.02, "learning_rate": 1.7995669801721626e-06, "loss": 1.3664, "step": 213 }, { "epoch": 0.02, "learning_rate": 1.8011391572336283e-06, "loss": 1.3533, "step": 214 }, { "epoch": 0.02, "learning_rate": 1.8027040047716915e-06, "loss": 1.3455, "step": 215 }, { "epoch": 0.02, "learning_rate": 1.8042615908101392e-06, "loss": 1.3764, "step": 216 }, { "epoch": 0.02, "learning_rate": 1.8058119824301583e-06, "loss": 1.3617, "step": 217 }, { "epoch": 0.02, "learning_rate": 1.8073552457876723e-06, "loss": 1.34, "step": 218 }, { "epoch": 0.02, "learning_rate": 1.8088914461302782e-06, "loss": 1.3381, "step": 219 }, { "epoch": 0.02, "learning_rate": 1.8104206478138003e-06, "loss": 1.4155, "step": 220 }, { "epoch": 0.02, "learning_rate": 1.8119429143184677e-06, "loss": 1.3458, "step": 221 }, { "epoch": 0.02, "learning_rate": 1.8134583082647262e-06, "loss": 1.3181, "step": 222 }, { "epoch": 0.02, "learning_rate": 1.8149668914286947e-06, "loss": 1.3279, "step": 223 }, { "epoch": 0.02, "learning_rate": 1.8164687247572772e-06, "loss": 1.3351, "step": 224 }, { "epoch": 0.02, "learning_rate": 1.8179638683829375e-06, "loss": 1.3376, "step": 225 }, { "epoch": 0.02, "learning_rate": 1.8194523816381454e-06, "loss": 1.3651, "step": 226 }, { "epoch": 0.02, "learning_rate": 1.8209343230695042e-06, "loss": 1.3892, "step": 227 }, { "epoch": 0.02, "learning_rate": 1.8224097504515708e-06, "loss": 1.3193, "step": 228 }, { "epoch": 0.02, "learning_rate": 1.8238787208003673e-06, "loss": 1.3657, "step": 229 }, { "epoch": 0.02, "learning_rate": 1.8253412903866047e-06, "loss": 1.4316, "step": 230 }, { "epoch": 0.02, "learning_rate": 1.8267975147486152e-06, "loss": 1.3862, "step": 231 }, { "epoch": 0.02, "learning_rate": 1.8282474487050078e-06, "loss": 1.3845, "step": 232 }, { "epoch": 0.02, "learning_rate": 1.82969114636705e-06, "loss": 1.3829, "step": 233 }, { "epoch": 0.02, "learning_rate": 1.8311286611507852e-06, "loss": 1.3523, "step": 234 }, { "epoch": 0.02, "learning_rate": 1.8325600457888912e-06, "loss": 0.8746, "step": 235 }, { "epoch": 0.02, "learning_rate": 1.833985352342289e-06, "loss": 1.387, "step": 236 }, { "epoch": 0.02, "learning_rate": 1.8354046322115028e-06, "loss": 1.3598, "step": 237 }, { "epoch": 0.02, "learning_rate": 1.8368179361477827e-06, "loss": 0.9115, "step": 238 }, { "epoch": 0.02, "learning_rate": 1.8382253142639954e-06, "loss": 1.3583, "step": 239 }, { "epoch": 0.02, "learning_rate": 1.8396268160452847e-06, "loss": 1.3845, "step": 240 }, { "epoch": 0.02, "learning_rate": 1.8410224903595135e-06, "loss": 1.3479, "step": 241 }, { "epoch": 0.02, "learning_rate": 1.8424123854674858e-06, "loss": 1.38, "step": 242 }, { "epoch": 0.02, "learning_rate": 1.8437965490329618e-06, "loss": 1.3944, "step": 243 }, { "epoch": 0.02, "learning_rate": 1.8451750281324651e-06, "loss": 1.3689, "step": 244 }, { "epoch": 0.02, "learning_rate": 1.8465478692648906e-06, "loss": 1.35, "step": 245 }, { "epoch": 0.02, "learning_rate": 1.8479151183609143e-06, "loss": 1.3699, "step": 246 }, { "epoch": 0.02, "learning_rate": 1.8492768207922166e-06, "loss": 1.2805, "step": 247 }, { "epoch": 0.02, "learning_rate": 1.8506330213805134e-06, "loss": 1.3054, "step": 248 }, { "epoch": 0.02, "learning_rate": 1.851983764406412e-06, "loss": 1.4013, "step": 249 }, { "epoch": 0.02, "learning_rate": 1.8533290936180832e-06, "loss": 1.3459, "step": 250 }, { "epoch": 0.02, "learning_rate": 1.8546690522397664e-06, "loss": 1.3402, "step": 251 }, { "epoch": 0.02, "learning_rate": 1.8560036829800998e-06, "loss": 1.3482, "step": 252 }, { "epoch": 0.02, "learning_rate": 1.85733302804029e-06, "loss": 1.315, "step": 253 }, { "epoch": 0.02, "learning_rate": 1.8586571291221181e-06, "loss": 0.959, "step": 254 }, { "epoch": 0.02, "learning_rate": 1.8599760274357901e-06, "loss": 1.4186, "step": 255 }, { "epoch": 0.02, "learning_rate": 1.8612897637076322e-06, "loss": 1.3573, "step": 256 }, { "epoch": 0.02, "learning_rate": 1.8625983781876369e-06, "loss": 1.3329, "step": 257 }, { "epoch": 0.02, "learning_rate": 1.8639019106568613e-06, "loss": 1.3558, "step": 258 }, { "epoch": 0.02, "learning_rate": 1.8652004004346866e-06, "loss": 1.3078, "step": 259 }, { "epoch": 0.02, "learning_rate": 1.8664938863859308e-06, "loss": 1.362, "step": 260 }, { "epoch": 0.02, "learning_rate": 1.8677824069278306e-06, "loss": 1.2855, "step": 261 }, { "epoch": 0.02, "learning_rate": 1.8690660000368872e-06, "loss": 1.3481, "step": 262 }, { "epoch": 0.02, "learning_rate": 1.8703447032555835e-06, "loss": 1.3419, "step": 263 }, { "epoch": 0.02, "learning_rate": 1.8716185536989702e-06, "loss": 1.3271, "step": 264 }, { "epoch": 0.02, "learning_rate": 1.8728875880611329e-06, "loss": 1.2865, "step": 265 }, { "epoch": 0.02, "learning_rate": 1.8741518426215312e-06, "loss": 0.8885, "step": 266 }, { "epoch": 0.02, "learning_rate": 1.875411353251225e-06, "loss": 1.3601, "step": 267 }, { "epoch": 0.02, "learning_rate": 1.876666155418979e-06, "loss": 0.913, "step": 268 }, { "epoch": 0.02, "learning_rate": 1.8779162841972581e-06, "loss": 0.9086, "step": 269 }, { "epoch": 0.02, "learning_rate": 1.8791617742681077e-06, "loss": 1.4015, "step": 270 }, { "epoch": 0.02, "learning_rate": 1.8804026599289269e-06, "loss": 0.9742, "step": 271 }, { "epoch": 0.02, "learning_rate": 1.8816389750981376e-06, "loss": 1.3545, "step": 272 }, { "epoch": 0.02, "learning_rate": 1.8828707533207457e-06, "loss": 1.3445, "step": 273 }, { "epoch": 0.02, "learning_rate": 1.884098027773803e-06, "loss": 1.3181, "step": 274 }, { "epoch": 0.02, "learning_rate": 1.8853208312717689e-06, "loss": 1.3077, "step": 275 }, { "epoch": 0.02, "learning_rate": 1.8865391962717743e-06, "loss": 1.3396, "step": 276 }, { "epoch": 0.02, "learning_rate": 1.8877531548787934e-06, "loss": 1.3286, "step": 277 }, { "epoch": 0.02, "learning_rate": 1.8889627388507158e-06, "loss": 1.3406, "step": 278 }, { "epoch": 0.02, "learning_rate": 1.8901679796033358e-06, "loss": 1.3613, "step": 279 }, { "epoch": 0.02, "learning_rate": 1.8913689082152456e-06, "loss": 1.3562, "step": 280 }, { "epoch": 0.02, "learning_rate": 1.8925655554326439e-06, "loss": 1.4209, "step": 281 }, { "epoch": 0.02, "learning_rate": 1.8937579516740612e-06, "loss": 1.4182, "step": 282 }, { "epoch": 0.02, "learning_rate": 1.8949461270349966e-06, "loss": 1.3881, "step": 283 }, { "epoch": 0.02, "learning_rate": 1.8961301112924784e-06, "loss": 1.3838, "step": 284 }, { "epoch": 0.02, "learning_rate": 1.897309933909539e-06, "loss": 1.3463, "step": 285 }, { "epoch": 0.02, "learning_rate": 1.898485624039616e-06, "loss": 1.3604, "step": 286 }, { "epoch": 0.02, "learning_rate": 1.8996572105308751e-06, "loss": 1.3503, "step": 287 }, { "epoch": 0.02, "learning_rate": 1.900824721930455e-06, "loss": 1.3652, "step": 288 }, { "epoch": 0.02, "learning_rate": 1.9019881864886428e-06, "loss": 1.2915, "step": 289 }, { "epoch": 0.02, "learning_rate": 1.9031476321629762e-06, "loss": 1.3539, "step": 290 }, { "epoch": 0.02, "learning_rate": 1.9043030866222733e-06, "loss": 0.933, "step": 291 }, { "epoch": 0.02, "learning_rate": 1.905454577250594e-06, "loss": 1.3623, "step": 292 }, { "epoch": 0.02, "learning_rate": 1.9066021311511358e-06, "loss": 1.3633, "step": 293 }, { "epoch": 0.02, "learning_rate": 1.9077457751500605e-06, "loss": 1.3358, "step": 294 }, { "epoch": 0.02, "learning_rate": 1.9088855358002575e-06, "loss": 1.345, "step": 295 }, { "epoch": 0.02, "learning_rate": 1.910021439385042e-06, "loss": 1.4061, "step": 296 }, { "epoch": 0.02, "learning_rate": 1.911153511921793e-06, "loss": 1.3382, "step": 297 }, { "epoch": 0.02, "learning_rate": 1.9122817791655284e-06, "loss": 1.3342, "step": 298 }, { "epoch": 0.02, "learning_rate": 1.9134062666124203e-06, "loss": 1.2955, "step": 299 }, { "epoch": 0.02, "learning_rate": 1.914526999503253e-06, "loss": 1.351, "step": 300 }, { "epoch": 0.02, "learning_rate": 1.915644002826822e-06, "loss": 1.3264, "step": 301 }, { "epoch": 0.02, "learning_rate": 1.916757301323277e-06, "loss": 1.3663, "step": 302 }, { "epoch": 0.02, "learning_rate": 1.9178669194874084e-06, "loss": 0.9333, "step": 303 }, { "epoch": 0.02, "learning_rate": 1.9189728815718865e-06, "loss": 1.382, "step": 304 }, { "epoch": 0.02, "learning_rate": 1.9200752115904337e-06, "loss": 1.4231, "step": 305 }, { "epoch": 0.02, "learning_rate": 1.92117393332096e-06, "loss": 1.3285, "step": 306 }, { "epoch": 0.02, "learning_rate": 1.9222690703086385e-06, "loss": 1.3446, "step": 307 }, { "epoch": 0.02, "learning_rate": 1.923360645868931e-06, "loss": 1.3454, "step": 308 }, { "epoch": 0.02, "learning_rate": 1.924448683090569e-06, "loss": 1.286, "step": 309 }, { "epoch": 0.02, "learning_rate": 1.925533204838482e-06, "loss": 1.4012, "step": 310 }, { "epoch": 0.02, "learning_rate": 1.9266142337566822e-06, "loss": 1.2971, "step": 311 }, { "epoch": 0.02, "learning_rate": 1.927691792271101e-06, "loss": 1.3211, "step": 312 }, { "epoch": 0.02, "learning_rate": 1.9287659025923787e-06, "loss": 0.894, "step": 313 }, { "epoch": 0.02, "learning_rate": 1.929836586718615e-06, "loss": 0.9369, "step": 314 }, { "epoch": 0.02, "learning_rate": 1.930903866438068e-06, "loss": 1.3443, "step": 315 }, { "epoch": 0.02, "learning_rate": 1.9319677633318183e-06, "loss": 1.3793, "step": 316 }, { "epoch": 0.02, "learning_rate": 1.9330282987763844e-06, "loss": 1.3463, "step": 317 }, { "epoch": 0.02, "learning_rate": 1.9340854939463027e-06, "loss": 1.323, "step": 318 }, { "epoch": 0.02, "learning_rate": 1.9351393698166616e-06, "loss": 1.3553, "step": 319 }, { "epoch": 0.02, "learning_rate": 1.9361899471656003e-06, "loss": 1.347, "step": 320 }, { "epoch": 0.02, "learning_rate": 1.9372372465767664e-06, "loss": 1.3477, "step": 321 }, { "epoch": 0.02, "learning_rate": 1.9382812884417353e-06, "loss": 1.3633, "step": 322 }, { "epoch": 0.03, "learning_rate": 1.9393220929623915e-06, "loss": 1.3788, "step": 323 }, { "epoch": 0.03, "learning_rate": 1.9403596801532773e-06, "loss": 0.9247, "step": 324 }, { "epoch": 0.03, "learning_rate": 1.941394069843899e-06, "loss": 0.9102, "step": 325 }, { "epoch": 0.03, "learning_rate": 1.9424252816810026e-06, "loss": 1.3235, "step": 326 }, { "epoch": 0.03, "learning_rate": 1.9434533351308106e-06, "loss": 1.3702, "step": 327 }, { "epoch": 0.03, "learning_rate": 1.9444782494812303e-06, "loss": 1.3147, "step": 328 }, { "epoch": 0.03, "learning_rate": 1.945500043844022e-06, "loss": 1.3611, "step": 329 }, { "epoch": 0.03, "learning_rate": 1.9465187371569383e-06, "loss": 1.3071, "step": 330 }, { "epoch": 0.03, "learning_rate": 1.9475343481858323e-06, "loss": 1.3195, "step": 331 }, { "epoch": 0.03, "learning_rate": 1.9485468955267275e-06, "loss": 1.3844, "step": 332 }, { "epoch": 0.03, "learning_rate": 1.9495563976078644e-06, "loss": 1.3889, "step": 333 }, { "epoch": 0.03, "learning_rate": 1.9505628726917117e-06, "loss": 1.3262, "step": 334 }, { "epoch": 0.03, "learning_rate": 1.9515663388769473e-06, "loss": 1.3634, "step": 335 }, { "epoch": 0.03, "learning_rate": 1.9525668141004156e-06, "loss": 1.3741, "step": 336 }, { "epoch": 0.03, "learning_rate": 1.9535643161390447e-06, "loss": 1.3645, "step": 337 }, { "epoch": 0.03, "learning_rate": 1.954558862611747e-06, "loss": 1.2848, "step": 338 }, { "epoch": 0.03, "learning_rate": 1.9555504709812835e-06, "loss": 1.3474, "step": 339 }, { "epoch": 0.03, "learning_rate": 1.956539158556106e-06, "loss": 1.3278, "step": 340 }, { "epoch": 0.03, "learning_rate": 1.957524942492167e-06, "loss": 1.365, "step": 341 }, { "epoch": 0.03, "learning_rate": 1.958507839794709e-06, "loss": 1.3333, "step": 342 }, { "epoch": 0.03, "learning_rate": 1.959487867320021e-06, "loss": 1.3732, "step": 343 }, { "epoch": 0.03, "learning_rate": 1.960465041777177e-06, "loss": 1.3545, "step": 344 }, { "epoch": 0.03, "learning_rate": 1.9614393797297432e-06, "loss": 1.3838, "step": 345 }, { "epoch": 0.03, "learning_rate": 1.9624108975974607e-06, "loss": 1.3289, "step": 346 }, { "epoch": 0.03, "learning_rate": 1.9633796116579104e-06, "loss": 1.3199, "step": 347 }, { "epoch": 0.03, "learning_rate": 1.9643455380481464e-06, "loss": 1.3379, "step": 348 }, { "epoch": 0.03, "learning_rate": 1.9653086927663084e-06, "loss": 1.3261, "step": 349 }, { "epoch": 0.03, "learning_rate": 1.9662690916732137e-06, "loss": 1.3196, "step": 350 }, { "epoch": 0.03, "learning_rate": 1.9672267504939234e-06, "loss": 1.3143, "step": 351 }, { "epoch": 0.03, "learning_rate": 1.968181684819286e-06, "loss": 1.3391, "step": 352 }, { "epoch": 0.03, "learning_rate": 1.9691339101074625e-06, "loss": 1.3281, "step": 353 }, { "epoch": 0.03, "learning_rate": 1.970083441685427e-06, "loss": 0.8824, "step": 354 }, { "epoch": 0.03, "learning_rate": 1.971030294750447e-06, "loss": 1.2953, "step": 355 }, { "epoch": 0.03, "learning_rate": 1.9719744843715405e-06, "loss": 1.2779, "step": 356 }, { "epoch": 0.03, "learning_rate": 1.9729160254909206e-06, "loss": 1.3201, "step": 357 }, { "epoch": 0.03, "learning_rate": 1.9738549329254088e-06, "loss": 1.3823, "step": 358 }, { "epoch": 0.03, "learning_rate": 1.974791221367836e-06, "loss": 1.3081, "step": 359 }, { "epoch": 0.03, "learning_rate": 1.9757249053884235e-06, "loss": 1.3883, "step": 360 }, { "epoch": 0.03, "learning_rate": 1.9766559994361405e-06, "loss": 0.8353, "step": 361 }, { "epoch": 0.03, "learning_rate": 1.9775845178400487e-06, "loss": 1.3557, "step": 362 }, { "epoch": 0.03, "learning_rate": 1.978510474810624e-06, "loss": 1.2812, "step": 363 }, { "epoch": 0.03, "learning_rate": 1.9794338844410612e-06, "loss": 1.34, "step": 364 }, { "epoch": 0.03, "learning_rate": 1.980354760708562e-06, "loss": 1.3696, "step": 365 }, { "epoch": 0.03, "learning_rate": 1.9812731174756033e-06, "loss": 1.3407, "step": 366 }, { "epoch": 0.03, "learning_rate": 1.9821889684911902e-06, "loss": 1.338, "step": 367 }, { "epoch": 0.03, "learning_rate": 1.9831023273920905e-06, "loss": 1.3413, "step": 368 }, { "epoch": 0.03, "learning_rate": 1.9840132077040527e-06, "loss": 1.3107, "step": 369 }, { "epoch": 0.03, "learning_rate": 1.98492162284301e-06, "loss": 1.3182, "step": 370 }, { "epoch": 0.03, "learning_rate": 1.9858275861162633e-06, "loss": 1.3965, "step": 371 }, { "epoch": 0.03, "learning_rate": 1.9867311107236515e-06, "loss": 1.3071, "step": 372 }, { "epoch": 0.03, "learning_rate": 1.987632209758707e-06, "loss": 1.3272, "step": 373 }, { "epoch": 0.03, "learning_rate": 1.9885308962097914e-06, "loss": 1.3546, "step": 374 }, { "epoch": 0.03, "learning_rate": 1.9894271829612216e-06, "loss": 1.327, "step": 375 }, { "epoch": 0.03, "learning_rate": 1.990321082794377e-06, "loss": 1.3211, "step": 376 }, { "epoch": 0.03, "learning_rate": 1.9912126083887924e-06, "loss": 1.3561, "step": 377 }, { "epoch": 0.03, "learning_rate": 1.9921017723232384e-06, "loss": 1.3462, "step": 378 }, { "epoch": 0.03, "learning_rate": 1.9929885870767852e-06, "loss": 1.298, "step": 379 }, { "epoch": 0.03, "learning_rate": 1.9938730650298546e-06, "loss": 1.321, "step": 380 }, { "epoch": 0.03, "learning_rate": 1.9947552184652565e-06, "loss": 1.3147, "step": 381 }, { "epoch": 0.03, "learning_rate": 1.9956350595692115e-06, "loss": 1.2968, "step": 382 }, { "epoch": 0.03, "learning_rate": 1.9965126004323623e-06, "loss": 1.2835, "step": 383 }, { "epoch": 0.03, "learning_rate": 1.9973878530507707e-06, "loss": 1.3552, "step": 384 }, { "epoch": 0.03, "learning_rate": 1.9982608293268993e-06, "loss": 1.339, "step": 385 }, { "epoch": 0.03, "learning_rate": 1.999131541070586e-06, "loss": 1.3013, "step": 386 }, { "epoch": 0.03, "learning_rate": 2e-06, "loss": 1.3007, "step": 387 }, { "epoch": 0.03, "learning_rate": 1.999999968402101e-06, "loss": 1.3585, "step": 388 }, { "epoch": 0.03, "learning_rate": 1.9999998736084053e-06, "loss": 1.328, "step": 389 }, { "epoch": 0.03, "learning_rate": 1.99999971561892e-06, "loss": 1.2637, "step": 390 }, { "epoch": 0.03, "learning_rate": 1.9999994944336546e-06, "loss": 1.3495, "step": 391 }, { "epoch": 0.03, "learning_rate": 1.9999992100526226e-06, "loss": 1.3601, "step": 392 }, { "epoch": 0.03, "learning_rate": 1.9999988624758426e-06, "loss": 0.8613, "step": 393 }, { "epoch": 0.03, "learning_rate": 1.999998451703336e-06, "loss": 1.3005, "step": 394 }, { "epoch": 0.03, "learning_rate": 1.9999979777351294e-06, "loss": 1.3292, "step": 395 }, { "epoch": 0.03, "learning_rate": 1.999997440571252e-06, "loss": 1.297, "step": 396 }, { "epoch": 0.03, "learning_rate": 1.9999968402117382e-06, "loss": 1.3336, "step": 397 }, { "epoch": 0.03, "learning_rate": 1.9999961766566263e-06, "loss": 1.3175, "step": 398 }, { "epoch": 0.03, "learning_rate": 1.999995449905958e-06, "loss": 1.3293, "step": 399 }, { "epoch": 0.03, "learning_rate": 1.9999946599597782e-06, "loss": 1.3277, "step": 400 }, { "epoch": 0.03, "learning_rate": 1.999993806818138e-06, "loss": 1.3374, "step": 401 }, { "epoch": 0.03, "learning_rate": 1.9999928904810916e-06, "loss": 1.3237, "step": 402 }, { "epoch": 0.03, "learning_rate": 1.999991910948696e-06, "loss": 1.3201, "step": 403 }, { "epoch": 0.03, "learning_rate": 1.9999908682210136e-06, "loss": 1.369, "step": 404 }, { "epoch": 0.03, "learning_rate": 1.9999897622981097e-06, "loss": 1.3604, "step": 405 }, { "epoch": 0.03, "learning_rate": 1.999988593180055e-06, "loss": 1.3545, "step": 406 }, { "epoch": 0.03, "learning_rate": 1.9999873608669223e-06, "loss": 1.3192, "step": 407 }, { "epoch": 0.03, "learning_rate": 1.999986065358791e-06, "loss": 1.2951, "step": 408 }, { "epoch": 0.03, "learning_rate": 1.9999847066557417e-06, "loss": 1.3222, "step": 409 }, { "epoch": 0.03, "learning_rate": 1.9999832847578605e-06, "loss": 1.3419, "step": 410 }, { "epoch": 0.03, "learning_rate": 1.999981799665238e-06, "loss": 1.3713, "step": 411 }, { "epoch": 0.03, "learning_rate": 1.9999802513779672e-06, "loss": 1.2969, "step": 412 }, { "epoch": 0.03, "learning_rate": 1.999978639896146e-06, "loss": 1.3713, "step": 413 }, { "epoch": 0.03, "learning_rate": 1.999976965219877e-06, "loss": 1.3439, "step": 414 }, { "epoch": 0.03, "learning_rate": 1.999975227349265e-06, "loss": 0.9319, "step": 415 }, { "epoch": 0.03, "learning_rate": 1.999973426284421e-06, "loss": 1.2975, "step": 416 }, { "epoch": 0.03, "learning_rate": 1.999971562025458e-06, "loss": 1.3241, "step": 417 }, { "epoch": 0.03, "learning_rate": 1.9999696345724936e-06, "loss": 1.3307, "step": 418 }, { "epoch": 0.03, "learning_rate": 1.9999676439256503e-06, "loss": 1.308, "step": 419 }, { "epoch": 0.03, "learning_rate": 1.9999655900850534e-06, "loss": 1.3259, "step": 420 }, { "epoch": 0.03, "learning_rate": 1.999963473050833e-06, "loss": 1.3157, "step": 421 }, { "epoch": 0.03, "learning_rate": 1.9999612928231226e-06, "loss": 1.2975, "step": 422 }, { "epoch": 0.03, "learning_rate": 1.99995904940206e-06, "loss": 1.2367, "step": 423 }, { "epoch": 0.03, "learning_rate": 1.9999567427877876e-06, "loss": 1.3229, "step": 424 }, { "epoch": 0.03, "learning_rate": 1.999954372980451e-06, "loss": 1.2949, "step": 425 }, { "epoch": 0.03, "learning_rate": 1.9999519399801985e-06, "loss": 1.2987, "step": 426 }, { "epoch": 0.03, "learning_rate": 1.999949443787186e-06, "loss": 0.9513, "step": 427 }, { "epoch": 0.03, "learning_rate": 1.9999468844015698e-06, "loss": 1.3478, "step": 428 }, { "epoch": 0.03, "learning_rate": 1.999944261823512e-06, "loss": 1.3922, "step": 429 }, { "epoch": 0.03, "learning_rate": 1.999941576053179e-06, "loss": 1.332, "step": 430 }, { "epoch": 0.03, "learning_rate": 1.9999388270907397e-06, "loss": 1.2696, "step": 431 }, { "epoch": 0.03, "learning_rate": 1.999936014936368e-06, "loss": 1.3642, "step": 432 }, { "epoch": 0.03, "learning_rate": 1.999933139590242e-06, "loss": 1.3663, "step": 433 }, { "epoch": 0.03, "learning_rate": 1.9999302010525426e-06, "loss": 0.8789, "step": 434 }, { "epoch": 0.03, "learning_rate": 1.9999271993234563e-06, "loss": 1.3202, "step": 435 }, { "epoch": 0.03, "learning_rate": 1.9999241344031726e-06, "loss": 1.3244, "step": 436 }, { "epoch": 0.03, "learning_rate": 1.9999210062918853e-06, "loss": 1.3594, "step": 437 }, { "epoch": 0.03, "learning_rate": 1.9999178149897916e-06, "loss": 0.9351, "step": 438 }, { "epoch": 0.03, "learning_rate": 1.999914560497093e-06, "loss": 1.3042, "step": 439 }, { "epoch": 0.03, "learning_rate": 1.9999112428139966e-06, "loss": 1.3221, "step": 440 }, { "epoch": 0.03, "learning_rate": 1.9999078619407107e-06, "loss": 1.3152, "step": 441 }, { "epoch": 0.03, "learning_rate": 1.999904417877449e-06, "loss": 0.9301, "step": 442 }, { "epoch": 0.03, "learning_rate": 1.99990091062443e-06, "loss": 1.2941, "step": 443 }, { "epoch": 0.03, "learning_rate": 1.9998973401818747e-06, "loss": 1.3241, "step": 444 }, { "epoch": 0.03, "learning_rate": 1.999893706550009e-06, "loss": 1.3537, "step": 445 }, { "epoch": 0.03, "learning_rate": 1.999890009729062e-06, "loss": 1.2958, "step": 446 }, { "epoch": 0.03, "learning_rate": 1.9998862497192683e-06, "loss": 1.3497, "step": 447 }, { "epoch": 0.03, "learning_rate": 1.9998824265208644e-06, "loss": 1.3156, "step": 448 }, { "epoch": 0.03, "learning_rate": 1.9998785401340927e-06, "loss": 1.3499, "step": 449 }, { "epoch": 0.03, "learning_rate": 1.999874590559199e-06, "loss": 1.3709, "step": 450 }, { "epoch": 0.04, "learning_rate": 1.999870577796432e-06, "loss": 1.37, "step": 451 }, { "epoch": 0.04, "learning_rate": 1.9998665018460458e-06, "loss": 1.3228, "step": 452 }, { "epoch": 0.04, "learning_rate": 1.999862362708298e-06, "loss": 1.3587, "step": 453 }, { "epoch": 0.04, "learning_rate": 1.99985816038345e-06, "loss": 1.3178, "step": 454 }, { "epoch": 0.04, "learning_rate": 1.9998538948717675e-06, "loss": 1.3174, "step": 455 }, { "epoch": 0.04, "learning_rate": 1.9998495661735205e-06, "loss": 1.3399, "step": 456 }, { "epoch": 0.04, "learning_rate": 1.999845174288982e-06, "loss": 1.3124, "step": 457 }, { "epoch": 0.04, "learning_rate": 1.9998407192184294e-06, "loss": 1.3579, "step": 458 }, { "epoch": 0.04, "learning_rate": 1.9998362009621442e-06, "loss": 1.2892, "step": 459 }, { "epoch": 0.04, "learning_rate": 1.9998316195204127e-06, "loss": 1.3669, "step": 460 }, { "epoch": 0.04, "learning_rate": 1.9998269748935237e-06, "loss": 1.3354, "step": 461 }, { "epoch": 0.04, "learning_rate": 1.999822267081771e-06, "loss": 1.2935, "step": 462 }, { "epoch": 0.04, "learning_rate": 1.9998174960854523e-06, "loss": 1.3451, "step": 463 }, { "epoch": 0.04, "learning_rate": 1.999812661904869e-06, "loss": 1.318, "step": 464 }, { "epoch": 0.04, "learning_rate": 1.9998077645403262e-06, "loss": 1.3878, "step": 465 }, { "epoch": 0.04, "learning_rate": 1.999802803992134e-06, "loss": 1.3463, "step": 466 }, { "epoch": 0.04, "learning_rate": 1.9997977802606052e-06, "loss": 0.9409, "step": 467 }, { "epoch": 0.04, "learning_rate": 1.9997926933460582e-06, "loss": 1.2724, "step": 468 }, { "epoch": 0.04, "learning_rate": 1.9997875432488135e-06, "loss": 1.287, "step": 469 }, { "epoch": 0.04, "learning_rate": 1.9997823299691968e-06, "loss": 1.2995, "step": 470 }, { "epoch": 0.04, "learning_rate": 1.9997770535075383e-06, "loss": 1.2887, "step": 471 }, { "epoch": 0.04, "learning_rate": 1.9997717138641706e-06, "loss": 1.2982, "step": 472 }, { "epoch": 0.04, "learning_rate": 1.9997663110394317e-06, "loss": 1.2954, "step": 473 }, { "epoch": 0.04, "learning_rate": 1.9997608450336624e-06, "loss": 1.3689, "step": 474 }, { "epoch": 0.04, "learning_rate": 1.999755315847209e-06, "loss": 0.8786, "step": 475 }, { "epoch": 0.04, "learning_rate": 1.99974972348042e-06, "loss": 1.3287, "step": 476 }, { "epoch": 0.04, "learning_rate": 1.9997440679336492e-06, "loss": 0.8904, "step": 477 }, { "epoch": 0.04, "learning_rate": 1.9997383492072543e-06, "loss": 1.337, "step": 478 }, { "epoch": 0.04, "learning_rate": 1.9997325673015964e-06, "loss": 0.9378, "step": 479 }, { "epoch": 0.04, "learning_rate": 1.999726722217041e-06, "loss": 1.3055, "step": 480 }, { "epoch": 0.04, "learning_rate": 1.999720813953957e-06, "loss": 1.3397, "step": 481 }, { "epoch": 0.04, "learning_rate": 1.9997148425127187e-06, "loss": 1.3012, "step": 482 }, { "epoch": 0.04, "learning_rate": 1.999708807893703e-06, "loss": 1.2855, "step": 483 }, { "epoch": 0.04, "learning_rate": 1.999702710097291e-06, "loss": 1.3615, "step": 484 }, { "epoch": 0.04, "learning_rate": 1.9996965491238686e-06, "loss": 1.3211, "step": 485 }, { "epoch": 0.04, "learning_rate": 1.999690324973824e-06, "loss": 1.3599, "step": 486 }, { "epoch": 0.04, "learning_rate": 1.999684037647552e-06, "loss": 1.3093, "step": 487 }, { "epoch": 0.04, "learning_rate": 1.9996776871454497e-06, "loss": 1.336, "step": 488 }, { "epoch": 0.04, "learning_rate": 1.999671273467918e-06, "loss": 1.2758, "step": 489 }, { "epoch": 0.04, "learning_rate": 1.9996647966153617e-06, "loss": 1.3114, "step": 490 }, { "epoch": 0.04, "learning_rate": 1.9996582565881906e-06, "loss": 0.9006, "step": 491 }, { "epoch": 0.04, "learning_rate": 1.9996516533868187e-06, "loss": 0.9057, "step": 492 }, { "epoch": 0.04, "learning_rate": 1.999644987011662e-06, "loss": 1.3306, "step": 493 }, { "epoch": 0.04, "learning_rate": 1.999638257463143e-06, "loss": 1.3181, "step": 494 }, { "epoch": 0.04, "learning_rate": 1.9996314647416863e-06, "loss": 1.3147, "step": 495 }, { "epoch": 0.04, "learning_rate": 1.999624608847721e-06, "loss": 1.2981, "step": 496 }, { "epoch": 0.04, "learning_rate": 1.999617689781681e-06, "loss": 1.3151, "step": 497 }, { "epoch": 0.04, "learning_rate": 1.999610707544003e-06, "loss": 1.2886, "step": 498 }, { "epoch": 0.04, "learning_rate": 1.9996036621351287e-06, "loss": 1.3669, "step": 499 }, { "epoch": 0.04, "learning_rate": 1.9995965535555032e-06, "loss": 1.3267, "step": 500 }, { "epoch": 0.04, "learning_rate": 1.999589381805575e-06, "loss": 1.3384, "step": 501 }, { "epoch": 0.04, "learning_rate": 1.9995821468857984e-06, "loss": 1.2921, "step": 502 }, { "epoch": 0.04, "learning_rate": 1.9995748487966304e-06, "loss": 1.2947, "step": 503 }, { "epoch": 0.04, "learning_rate": 1.9995674875385315e-06, "loss": 1.3097, "step": 504 }, { "epoch": 0.04, "learning_rate": 1.999560063111968e-06, "loss": 1.2693, "step": 505 }, { "epoch": 0.04, "learning_rate": 1.999552575517408e-06, "loss": 1.3388, "step": 506 }, { "epoch": 0.04, "learning_rate": 1.999545024755325e-06, "loss": 1.3426, "step": 507 }, { "epoch": 0.04, "learning_rate": 1.9995374108261973e-06, "loss": 1.3075, "step": 508 }, { "epoch": 0.04, "learning_rate": 1.9995297337305043e-06, "loss": 1.2834, "step": 509 }, { "epoch": 0.04, "learning_rate": 1.999521993468732e-06, "loss": 1.3054, "step": 510 }, { "epoch": 0.04, "learning_rate": 1.9995141900413696e-06, "loss": 1.3303, "step": 511 }, { "epoch": 0.04, "learning_rate": 1.9995063234489104e-06, "loss": 0.9256, "step": 512 }, { "epoch": 0.04, "learning_rate": 1.999498393691851e-06, "loss": 1.2864, "step": 513 }, { "epoch": 0.04, "learning_rate": 1.999490400770693e-06, "loss": 1.3035, "step": 514 }, { "epoch": 0.04, "learning_rate": 1.9994823446859417e-06, "loss": 1.3108, "step": 515 }, { "epoch": 0.04, "learning_rate": 1.999474225438105e-06, "loss": 1.3223, "step": 516 }, { "epoch": 0.04, "learning_rate": 1.9994660430276974e-06, "loss": 0.877, "step": 517 }, { "epoch": 0.04, "learning_rate": 1.9994577974552355e-06, "loss": 1.3121, "step": 518 }, { "epoch": 0.04, "learning_rate": 1.9994494887212403e-06, "loss": 1.2901, "step": 519 }, { "epoch": 0.04, "learning_rate": 1.999441116826237e-06, "loss": 1.2855, "step": 520 }, { "epoch": 0.04, "learning_rate": 1.9994326817707544e-06, "loss": 1.2736, "step": 521 }, { "epoch": 0.04, "learning_rate": 1.999424183555326e-06, "loss": 1.3513, "step": 522 }, { "epoch": 0.04, "learning_rate": 1.9994156221804883e-06, "loss": 1.3843, "step": 523 }, { "epoch": 0.04, "learning_rate": 1.999406997646783e-06, "loss": 1.3414, "step": 524 }, { "epoch": 0.04, "learning_rate": 1.9993983099547543e-06, "loss": 1.3233, "step": 525 }, { "epoch": 0.04, "learning_rate": 1.999389559104952e-06, "loss": 0.9318, "step": 526 }, { "epoch": 0.04, "learning_rate": 1.9993807450979287e-06, "loss": 1.2801, "step": 527 }, { "epoch": 0.04, "learning_rate": 1.9993718679342418e-06, "loss": 1.2854, "step": 528 }, { "epoch": 0.04, "learning_rate": 1.999362927614452e-06, "loss": 1.3423, "step": 529 }, { "epoch": 0.04, "learning_rate": 1.999353924139124e-06, "loss": 1.3175, "step": 530 }, { "epoch": 0.04, "learning_rate": 1.999344857508827e-06, "loss": 1.2543, "step": 531 }, { "epoch": 0.04, "learning_rate": 1.9993357277241346e-06, "loss": 1.2836, "step": 532 }, { "epoch": 0.04, "learning_rate": 1.999326534785623e-06, "loss": 1.3509, "step": 533 }, { "epoch": 0.04, "learning_rate": 1.999317278693873e-06, "loss": 1.345, "step": 534 }, { "epoch": 0.04, "learning_rate": 1.9993079594494707e-06, "loss": 1.3124, "step": 535 }, { "epoch": 0.04, "learning_rate": 1.999298577053004e-06, "loss": 1.2848, "step": 536 }, { "epoch": 0.04, "learning_rate": 1.999289131505066e-06, "loss": 1.3232, "step": 537 }, { "epoch": 0.04, "learning_rate": 1.9992796228062534e-06, "loss": 1.3528, "step": 538 }, { "epoch": 0.04, "learning_rate": 1.999270050957168e-06, "loss": 1.3209, "step": 539 }, { "epoch": 0.04, "learning_rate": 1.9992604159584137e-06, "loss": 1.2837, "step": 540 }, { "epoch": 0.04, "learning_rate": 1.9992507178106003e-06, "loss": 1.3038, "step": 541 }, { "epoch": 0.04, "learning_rate": 1.9992409565143398e-06, "loss": 1.3255, "step": 542 }, { "epoch": 0.04, "learning_rate": 1.9992311320702496e-06, "loss": 1.3283, "step": 543 }, { "epoch": 0.04, "learning_rate": 1.9992212444789504e-06, "loss": 1.3025, "step": 544 }, { "epoch": 0.04, "learning_rate": 1.999211293741067e-06, "loss": 1.2789, "step": 545 }, { "epoch": 0.04, "learning_rate": 1.9992012798572286e-06, "loss": 1.3389, "step": 546 }, { "epoch": 0.04, "learning_rate": 1.999191202828068e-06, "loss": 1.3167, "step": 547 }, { "epoch": 0.04, "learning_rate": 1.9991810626542215e-06, "loss": 1.3667, "step": 548 }, { "epoch": 0.04, "learning_rate": 1.99917085933633e-06, "loss": 1.3037, "step": 549 }, { "epoch": 0.04, "learning_rate": 1.999160592875039e-06, "loss": 1.2912, "step": 550 }, { "epoch": 0.04, "learning_rate": 1.9991502632709963e-06, "loss": 1.3679, "step": 551 }, { "epoch": 0.04, "learning_rate": 1.9991398705248555e-06, "loss": 0.8531, "step": 552 }, { "epoch": 0.04, "learning_rate": 1.999129414637273e-06, "loss": 1.301, "step": 553 }, { "epoch": 0.04, "learning_rate": 1.9991188956089096e-06, "loss": 1.3328, "step": 554 }, { "epoch": 0.04, "learning_rate": 1.9991083134404305e-06, "loss": 1.3562, "step": 555 }, { "epoch": 0.04, "learning_rate": 1.999097668132504e-06, "loss": 1.3201, "step": 556 }, { "epoch": 0.04, "learning_rate": 1.9990869596858026e-06, "loss": 1.2764, "step": 557 }, { "epoch": 0.04, "learning_rate": 1.9990761881010032e-06, "loss": 1.3491, "step": 558 }, { "epoch": 0.04, "learning_rate": 1.9990653533787873e-06, "loss": 1.3218, "step": 559 }, { "epoch": 0.04, "learning_rate": 1.9990544555198382e-06, "loss": 1.3307, "step": 560 }, { "epoch": 0.04, "learning_rate": 1.9990434945248456e-06, "loss": 1.3583, "step": 561 }, { "epoch": 0.04, "learning_rate": 1.9990324703945027e-06, "loss": 1.2896, "step": 562 }, { "epoch": 0.04, "learning_rate": 1.9990213831295046e-06, "loss": 1.246, "step": 563 }, { "epoch": 0.04, "learning_rate": 1.999010232730553e-06, "loss": 1.3027, "step": 564 }, { "epoch": 0.04, "learning_rate": 1.998999019198352e-06, "loss": 1.2679, "step": 565 }, { "epoch": 0.04, "learning_rate": 1.9989877425336113e-06, "loss": 1.3371, "step": 566 }, { "epoch": 0.04, "learning_rate": 1.9989764027370423e-06, "loss": 1.3075, "step": 567 }, { "epoch": 0.04, "learning_rate": 1.9989649998093625e-06, "loss": 1.3327, "step": 568 }, { "epoch": 0.04, "learning_rate": 1.998953533751292e-06, "loss": 1.2673, "step": 569 }, { "epoch": 0.04, "learning_rate": 1.9989420045635555e-06, "loss": 1.3386, "step": 570 }, { "epoch": 0.04, "learning_rate": 1.998930412246882e-06, "loss": 1.3416, "step": 571 }, { "epoch": 0.04, "learning_rate": 1.9989187568020037e-06, "loss": 1.3485, "step": 572 }, { "epoch": 0.04, "learning_rate": 1.9989070382296573e-06, "loss": 0.8793, "step": 573 }, { "epoch": 0.04, "learning_rate": 1.998895256530583e-06, "loss": 1.2563, "step": 574 }, { "epoch": 0.04, "learning_rate": 1.9988834117055256e-06, "loss": 1.3046, "step": 575 }, { "epoch": 0.04, "learning_rate": 1.998871503755234e-06, "loss": 1.333, "step": 576 }, { "epoch": 0.04, "learning_rate": 1.9988595326804603e-06, "loss": 1.2683, "step": 577 }, { "epoch": 0.04, "learning_rate": 1.9988474984819613e-06, "loss": 1.3286, "step": 578 }, { "epoch": 0.04, "learning_rate": 1.9988354011604977e-06, "loss": 0.8912, "step": 579 }, { "epoch": 0.05, "learning_rate": 1.9988232407168332e-06, "loss": 1.2952, "step": 580 }, { "epoch": 0.05, "learning_rate": 1.998811017151737e-06, "loss": 1.3306, "step": 581 }, { "epoch": 0.05, "learning_rate": 1.9987987304659813e-06, "loss": 1.3507, "step": 582 }, { "epoch": 0.05, "learning_rate": 1.998786380660342e-06, "loss": 1.3516, "step": 583 }, { "epoch": 0.05, "learning_rate": 1.998773967735601e-06, "loss": 1.2768, "step": 584 }, { "epoch": 0.05, "learning_rate": 1.9987614916925418e-06, "loss": 1.2946, "step": 585 }, { "epoch": 0.05, "learning_rate": 1.998748952531953e-06, "loss": 1.3377, "step": 586 }, { "epoch": 0.05, "learning_rate": 1.9987363502546265e-06, "loss": 1.2661, "step": 587 }, { "epoch": 0.05, "learning_rate": 1.998723684861359e-06, "loss": 1.343, "step": 588 }, { "epoch": 0.05, "learning_rate": 1.998710956352952e-06, "loss": 1.2928, "step": 589 }, { "epoch": 0.05, "learning_rate": 1.998698164730208e-06, "loss": 1.3226, "step": 590 }, { "epoch": 0.05, "learning_rate": 1.998685309993937e-06, "loss": 1.3217, "step": 591 }, { "epoch": 0.05, "learning_rate": 1.9986723921449505e-06, "loss": 1.2851, "step": 592 }, { "epoch": 0.05, "learning_rate": 1.9986594111840648e-06, "loss": 1.3289, "step": 593 }, { "epoch": 0.05, "learning_rate": 1.998646367112101e-06, "loss": 1.336, "step": 594 }, { "epoch": 0.05, "learning_rate": 1.998633259929883e-06, "loss": 1.3199, "step": 595 }, { "epoch": 0.05, "learning_rate": 1.998620089638238e-06, "loss": 1.2855, "step": 596 }, { "epoch": 0.05, "learning_rate": 1.9986068562380007e-06, "loss": 1.2946, "step": 597 }, { "epoch": 0.05, "learning_rate": 1.998593559730005e-06, "loss": 1.2981, "step": 598 }, { "epoch": 0.05, "learning_rate": 1.9985802001150927e-06, "loss": 1.3018, "step": 599 }, { "epoch": 0.05, "learning_rate": 1.998566777394108e-06, "loss": 1.3184, "step": 600 }, { "epoch": 0.05, "learning_rate": 1.998553291567898e-06, "loss": 1.3208, "step": 601 }, { "epoch": 0.05, "learning_rate": 1.9985397426373164e-06, "loss": 1.3464, "step": 602 }, { "epoch": 0.05, "learning_rate": 1.9985261306032185e-06, "loss": 1.2993, "step": 603 }, { "epoch": 0.05, "learning_rate": 1.9985124554664644e-06, "loss": 1.2172, "step": 604 }, { "epoch": 0.05, "learning_rate": 1.998498717227919e-06, "loss": 1.3292, "step": 605 }, { "epoch": 0.05, "learning_rate": 1.9984849158884503e-06, "loss": 1.2769, "step": 606 }, { "epoch": 0.05, "learning_rate": 1.99847105144893e-06, "loss": 1.3156, "step": 607 }, { "epoch": 0.05, "learning_rate": 1.998457123910235e-06, "loss": 1.2957, "step": 608 }, { "epoch": 0.05, "learning_rate": 1.998443133273245e-06, "loss": 1.2779, "step": 609 }, { "epoch": 0.05, "learning_rate": 1.9984290795388446e-06, "loss": 1.3131, "step": 610 }, { "epoch": 0.05, "learning_rate": 1.9984149627079214e-06, "loss": 1.3117, "step": 611 }, { "epoch": 0.05, "learning_rate": 1.9984007827813674e-06, "loss": 1.285, "step": 612 }, { "epoch": 0.05, "learning_rate": 1.9983865397600793e-06, "loss": 1.2465, "step": 613 }, { "epoch": 0.05, "learning_rate": 1.9983722336449565e-06, "loss": 1.4006, "step": 614 }, { "epoch": 0.05, "learning_rate": 1.998357864436904e-06, "loss": 1.2593, "step": 615 }, { "epoch": 0.05, "learning_rate": 1.9983434321368293e-06, "loss": 1.3076, "step": 616 }, { "epoch": 0.05, "learning_rate": 1.9983289367456443e-06, "loss": 1.2898, "step": 617 }, { "epoch": 0.05, "learning_rate": 1.9983143782642657e-06, "loss": 1.362, "step": 618 }, { "epoch": 0.05, "learning_rate": 1.998299756693613e-06, "loss": 1.3157, "step": 619 }, { "epoch": 0.05, "learning_rate": 1.9982850720346102e-06, "loss": 0.8839, "step": 620 }, { "epoch": 0.05, "learning_rate": 1.9982703242881855e-06, "loss": 1.2772, "step": 621 }, { "epoch": 0.05, "learning_rate": 1.998255513455271e-06, "loss": 1.3084, "step": 622 }, { "epoch": 0.05, "learning_rate": 1.9982406395368024e-06, "loss": 1.2993, "step": 623 }, { "epoch": 0.05, "learning_rate": 1.9982257025337197e-06, "loss": 1.3483, "step": 624 }, { "epoch": 0.05, "learning_rate": 1.9982107024469673e-06, "loss": 1.2408, "step": 625 }, { "epoch": 0.05, "learning_rate": 1.9981956392774924e-06, "loss": 1.3409, "step": 626 }, { "epoch": 0.05, "learning_rate": 1.998180513026248e-06, "loss": 1.3133, "step": 627 }, { "epoch": 0.05, "learning_rate": 1.9981653236941887e-06, "loss": 1.3126, "step": 628 }, { "epoch": 0.05, "learning_rate": 1.9981500712822755e-06, "loss": 1.3453, "step": 629 }, { "epoch": 0.05, "learning_rate": 1.9981347557914713e-06, "loss": 1.2694, "step": 630 }, { "epoch": 0.05, "learning_rate": 1.998119377222745e-06, "loss": 1.2652, "step": 631 }, { "epoch": 0.05, "learning_rate": 1.998103935577068e-06, "loss": 1.2792, "step": 632 }, { "epoch": 0.05, "learning_rate": 1.9980884308554158e-06, "loss": 0.9073, "step": 633 }, { "epoch": 0.05, "learning_rate": 1.9980728630587687e-06, "loss": 1.2586, "step": 634 }, { "epoch": 0.05, "learning_rate": 1.9980572321881106e-06, "loss": 1.2605, "step": 635 }, { "epoch": 0.05, "learning_rate": 1.998041538244429e-06, "loss": 1.2964, "step": 636 }, { "epoch": 0.05, "learning_rate": 1.9980257812287154e-06, "loss": 1.3103, "step": 637 }, { "epoch": 0.05, "learning_rate": 1.9980099611419663e-06, "loss": 1.3016, "step": 638 }, { "epoch": 0.05, "learning_rate": 1.9979940779851812e-06, "loss": 1.3406, "step": 639 }, { "epoch": 0.05, "learning_rate": 1.9979781317593635e-06, "loss": 1.3191, "step": 640 }, { "epoch": 0.05, "learning_rate": 1.9979621224655215e-06, "loss": 1.2481, "step": 641 }, { "epoch": 0.05, "learning_rate": 1.9979460501046665e-06, "loss": 1.2859, "step": 642 }, { "epoch": 0.05, "learning_rate": 1.997929914677814e-06, "loss": 1.3318, "step": 643 }, { "epoch": 0.05, "learning_rate": 1.9979137161859845e-06, "loss": 1.3151, "step": 644 }, { "epoch": 0.05, "learning_rate": 1.9978974546302004e-06, "loss": 1.2894, "step": 645 }, { "epoch": 0.05, "learning_rate": 1.997881130011491e-06, "loss": 0.9352, "step": 646 }, { "epoch": 0.05, "learning_rate": 1.997864742330887e-06, "loss": 1.3489, "step": 647 }, { "epoch": 0.05, "learning_rate": 1.997848291589424e-06, "loss": 1.2799, "step": 648 }, { "epoch": 0.05, "learning_rate": 1.9978317777881414e-06, "loss": 1.3323, "step": 649 }, { "epoch": 0.05, "learning_rate": 1.9978152009280837e-06, "loss": 0.924, "step": 650 }, { "epoch": 0.05, "learning_rate": 1.9977985610102977e-06, "loss": 1.329, "step": 651 }, { "epoch": 0.05, "learning_rate": 1.997781858035835e-06, "loss": 0.9226, "step": 652 }, { "epoch": 0.05, "learning_rate": 1.9977650920057515e-06, "loss": 1.3456, "step": 653 }, { "epoch": 0.05, "learning_rate": 1.9977482629211067e-06, "loss": 1.311, "step": 654 }, { "epoch": 0.05, "learning_rate": 1.9977313707829643e-06, "loss": 1.326, "step": 655 }, { "epoch": 0.05, "learning_rate": 1.9977144155923914e-06, "loss": 0.9034, "step": 656 }, { "epoch": 0.05, "learning_rate": 1.9976973973504596e-06, "loss": 1.265, "step": 657 }, { "epoch": 0.05, "learning_rate": 1.9976803160582442e-06, "loss": 1.326, "step": 658 }, { "epoch": 0.05, "learning_rate": 1.9976631717168252e-06, "loss": 1.3641, "step": 659 }, { "epoch": 0.05, "learning_rate": 1.9976459643272855e-06, "loss": 1.2956, "step": 660 }, { "epoch": 0.05, "learning_rate": 1.997628693890713e-06, "loss": 0.8696, "step": 661 }, { "epoch": 0.05, "learning_rate": 1.9976113604081986e-06, "loss": 1.2527, "step": 662 }, { "epoch": 0.05, "learning_rate": 1.9975939638808384e-06, "loss": 1.3052, "step": 663 }, { "epoch": 0.05, "learning_rate": 1.9975765043097316e-06, "loss": 1.3069, "step": 664 }, { "epoch": 0.05, "learning_rate": 1.9975589816959806e-06, "loss": 1.3177, "step": 665 }, { "epoch": 0.05, "learning_rate": 1.997541396040694e-06, "loss": 1.3155, "step": 666 }, { "epoch": 0.05, "learning_rate": 1.9975237473449828e-06, "loss": 1.2519, "step": 667 }, { "epoch": 0.05, "learning_rate": 1.997506035609962e-06, "loss": 1.288, "step": 668 }, { "epoch": 0.05, "learning_rate": 1.997488260836751e-06, "loss": 1.3296, "step": 669 }, { "epoch": 0.05, "learning_rate": 1.997470423026473e-06, "loss": 1.3303, "step": 670 }, { "epoch": 0.05, "learning_rate": 1.997452522180256e-06, "loss": 1.2762, "step": 671 }, { "epoch": 0.05, "learning_rate": 1.9974345582992304e-06, "loss": 1.2979, "step": 672 }, { "epoch": 0.05, "learning_rate": 1.997416531384532e-06, "loss": 1.3435, "step": 673 }, { "epoch": 0.05, "learning_rate": 1.9973984414372997e-06, "loss": 1.2998, "step": 674 }, { "epoch": 0.05, "learning_rate": 1.9973802884586766e-06, "loss": 1.3293, "step": 675 }, { "epoch": 0.05, "learning_rate": 1.99736207244981e-06, "loss": 1.2513, "step": 676 }, { "epoch": 0.05, "learning_rate": 1.9973437934118514e-06, "loss": 1.2505, "step": 677 }, { "epoch": 0.05, "learning_rate": 1.997325451345956e-06, "loss": 0.881, "step": 678 }, { "epoch": 0.05, "learning_rate": 1.9973070462532818e-06, "loss": 1.2639, "step": 679 }, { "epoch": 0.05, "learning_rate": 1.997288578134993e-06, "loss": 1.2799, "step": 680 }, { "epoch": 0.05, "learning_rate": 1.997270046992257e-06, "loss": 1.3031, "step": 681 }, { "epoch": 0.05, "learning_rate": 1.997251452826244e-06, "loss": 1.3289, "step": 682 }, { "epoch": 0.05, "learning_rate": 1.99723279563813e-06, "loss": 1.2874, "step": 683 }, { "epoch": 0.05, "learning_rate": 1.997214075429093e-06, "loss": 1.3518, "step": 684 }, { "epoch": 0.05, "learning_rate": 1.997195292200317e-06, "loss": 1.2894, "step": 685 }, { "epoch": 0.05, "learning_rate": 1.997176445952988e-06, "loss": 1.3106, "step": 686 }, { "epoch": 0.05, "learning_rate": 1.997157536688298e-06, "loss": 1.2491, "step": 687 }, { "epoch": 0.05, "learning_rate": 1.997138564407441e-06, "loss": 1.2851, "step": 688 }, { "epoch": 0.05, "learning_rate": 1.9971195291116173e-06, "loss": 1.296, "step": 689 }, { "epoch": 0.05, "learning_rate": 1.9971004308020284e-06, "loss": 1.2928, "step": 690 }, { "epoch": 0.05, "learning_rate": 1.9970812694798823e-06, "loss": 1.264, "step": 691 }, { "epoch": 0.05, "learning_rate": 1.9970620451463895e-06, "loss": 1.3428, "step": 692 }, { "epoch": 0.05, "learning_rate": 1.997042757802765e-06, "loss": 1.2731, "step": 693 }, { "epoch": 0.05, "learning_rate": 1.9970234074502272e-06, "loss": 1.2916, "step": 694 }, { "epoch": 0.05, "learning_rate": 1.9970039940899998e-06, "loss": 1.2975, "step": 695 }, { "epoch": 0.05, "learning_rate": 1.9969845177233087e-06, "loss": 1.3554, "step": 696 }, { "epoch": 0.05, "learning_rate": 1.9969649783513856e-06, "loss": 0.9001, "step": 697 }, { "epoch": 0.05, "learning_rate": 1.996945375975465e-06, "loss": 1.2986, "step": 698 }, { "epoch": 0.05, "learning_rate": 1.9969257105967853e-06, "loss": 1.3393, "step": 699 }, { "epoch": 0.05, "learning_rate": 1.99690598221659e-06, "loss": 1.2524, "step": 700 }, { "epoch": 0.05, "learning_rate": 1.9968861908361253e-06, "loss": 1.3285, "step": 701 }, { "epoch": 0.05, "learning_rate": 1.996866336456642e-06, "loss": 1.3136, "step": 702 }, { "epoch": 0.05, "learning_rate": 1.996846419079395e-06, "loss": 1.2919, "step": 703 }, { "epoch": 0.05, "learning_rate": 1.996826438705643e-06, "loss": 1.3375, "step": 704 }, { "epoch": 0.05, "learning_rate": 1.9968063953366483e-06, "loss": 0.9315, "step": 705 }, { "epoch": 0.05, "learning_rate": 1.996786288973678e-06, "loss": 0.865, "step": 706 }, { "epoch": 0.05, "learning_rate": 1.996766119618002e-06, "loss": 0.9023, "step": 707 }, { "epoch": 0.05, "learning_rate": 1.996745887270896e-06, "loss": 1.279, "step": 708 }, { "epoch": 0.06, "learning_rate": 1.996725591933638e-06, "loss": 1.2991, "step": 709 }, { "epoch": 0.06, "learning_rate": 1.9967052336075108e-06, "loss": 1.1962, "step": 710 }, { "epoch": 0.06, "learning_rate": 1.996684812293801e-06, "loss": 1.2965, "step": 711 }, { "epoch": 0.06, "learning_rate": 1.996664327993798e-06, "loss": 1.2941, "step": 712 }, { "epoch": 0.06, "learning_rate": 1.996643780708798e-06, "loss": 1.3256, "step": 713 }, { "epoch": 0.06, "learning_rate": 1.996623170440099e-06, "loss": 0.8904, "step": 714 }, { "epoch": 0.06, "learning_rate": 1.996602497189003e-06, "loss": 1.2773, "step": 715 }, { "epoch": 0.06, "learning_rate": 1.9965817609568164e-06, "loss": 1.3236, "step": 716 }, { "epoch": 0.06, "learning_rate": 1.9965609617448505e-06, "loss": 1.3199, "step": 717 }, { "epoch": 0.06, "learning_rate": 1.996540099554419e-06, "loss": 1.2849, "step": 718 }, { "epoch": 0.06, "learning_rate": 1.99651917438684e-06, "loss": 1.2723, "step": 719 }, { "epoch": 0.06, "learning_rate": 1.996498186243437e-06, "loss": 1.3236, "step": 720 }, { "epoch": 0.06, "learning_rate": 1.996477135125536e-06, "loss": 1.2873, "step": 721 }, { "epoch": 0.06, "learning_rate": 1.9964560210344665e-06, "loss": 1.2486, "step": 722 }, { "epoch": 0.06, "learning_rate": 1.9964348439715632e-06, "loss": 1.3039, "step": 723 }, { "epoch": 0.06, "learning_rate": 1.9964136039381648e-06, "loss": 1.3542, "step": 724 }, { "epoch": 0.06, "learning_rate": 1.9963923009356136e-06, "loss": 0.8452, "step": 725 }, { "epoch": 0.06, "learning_rate": 1.9963709349652557e-06, "loss": 1.3065, "step": 726 }, { "epoch": 0.06, "learning_rate": 1.996349506028441e-06, "loss": 1.3537, "step": 727 }, { "epoch": 0.06, "learning_rate": 1.9963280141265243e-06, "loss": 1.2967, "step": 728 }, { "epoch": 0.06, "learning_rate": 1.9963064592608634e-06, "loss": 1.3003, "step": 729 }, { "epoch": 0.06, "learning_rate": 1.9962848414328205e-06, "loss": 1.2725, "step": 730 }, { "epoch": 0.06, "learning_rate": 1.9962631606437616e-06, "loss": 1.2785, "step": 731 }, { "epoch": 0.06, "learning_rate": 1.9962414168950575e-06, "loss": 1.3304, "step": 732 }, { "epoch": 0.06, "learning_rate": 1.996219610188082e-06, "loss": 1.3271, "step": 733 }, { "epoch": 0.06, "learning_rate": 1.9961977405242125e-06, "loss": 1.3006, "step": 734 }, { "epoch": 0.06, "learning_rate": 1.996175807904832e-06, "loss": 1.3037, "step": 735 }, { "epoch": 0.06, "learning_rate": 1.9961538123313256e-06, "loss": 1.3035, "step": 736 }, { "epoch": 0.06, "learning_rate": 1.9961317538050846e-06, "loss": 1.294, "step": 737 }, { "epoch": 0.06, "learning_rate": 1.9961096323275024e-06, "loss": 1.2929, "step": 738 }, { "epoch": 0.06, "learning_rate": 1.9960874478999765e-06, "loss": 1.3145, "step": 739 }, { "epoch": 0.06, "learning_rate": 1.9960652005239096e-06, "loss": 1.3082, "step": 740 }, { "epoch": 0.06, "learning_rate": 1.996042890200707e-06, "loss": 1.3213, "step": 741 }, { "epoch": 0.06, "learning_rate": 1.9960205169317794e-06, "loss": 1.3832, "step": 742 }, { "epoch": 0.06, "learning_rate": 1.9959980807185396e-06, "loss": 1.2895, "step": 743 }, { "epoch": 0.06, "learning_rate": 1.9959755815624064e-06, "loss": 1.2467, "step": 744 }, { "epoch": 0.06, "learning_rate": 1.9959530194648017e-06, "loss": 1.2765, "step": 745 }, { "epoch": 0.06, "learning_rate": 1.9959303944271504e-06, "loss": 1.2645, "step": 746 }, { "epoch": 0.06, "learning_rate": 1.9959077064508833e-06, "loss": 1.2874, "step": 747 }, { "epoch": 0.06, "learning_rate": 1.995884955537434e-06, "loss": 1.222, "step": 748 }, { "epoch": 0.06, "learning_rate": 1.9958621416882397e-06, "loss": 1.3126, "step": 749 }, { "epoch": 0.06, "learning_rate": 1.9958392649047423e-06, "loss": 1.2989, "step": 750 }, { "epoch": 0.06, "learning_rate": 1.9958163251883876e-06, "loss": 1.2952, "step": 751 }, { "epoch": 0.06, "learning_rate": 1.9957933225406257e-06, "loss": 1.2999, "step": 752 }, { "epoch": 0.06, "learning_rate": 1.99577025696291e-06, "loss": 1.3156, "step": 753 }, { "epoch": 0.06, "learning_rate": 1.9957471284566982e-06, "loss": 1.2598, "step": 754 }, { "epoch": 0.06, "learning_rate": 1.9957239370234515e-06, "loss": 1.3231, "step": 755 }, { "epoch": 0.06, "learning_rate": 1.995700682664636e-06, "loss": 1.2835, "step": 756 }, { "epoch": 0.06, "learning_rate": 1.9956773653817215e-06, "loss": 1.2678, "step": 757 }, { "epoch": 0.06, "learning_rate": 1.9956539851761804e-06, "loss": 0.9154, "step": 758 }, { "epoch": 0.06, "learning_rate": 1.9956305420494918e-06, "loss": 0.8767, "step": 759 }, { "epoch": 0.06, "learning_rate": 1.995607036003136e-06, "loss": 1.3625, "step": 760 }, { "epoch": 0.06, "learning_rate": 1.9955834670385987e-06, "loss": 1.2875, "step": 761 }, { "epoch": 0.06, "learning_rate": 1.9955598351573695e-06, "loss": 0.8733, "step": 762 }, { "epoch": 0.06, "learning_rate": 1.9955361403609424e-06, "loss": 1.3085, "step": 763 }, { "epoch": 0.06, "learning_rate": 1.995512382650814e-06, "loss": 1.279, "step": 764 }, { "epoch": 0.06, "learning_rate": 1.9954885620284857e-06, "loss": 1.2765, "step": 765 }, { "epoch": 0.06, "learning_rate": 1.9954646784954636e-06, "loss": 0.8931, "step": 766 }, { "epoch": 0.06, "learning_rate": 1.9954407320532563e-06, "loss": 1.2915, "step": 767 }, { "epoch": 0.06, "learning_rate": 1.9954167227033777e-06, "loss": 1.2577, "step": 768 }, { "epoch": 0.06, "learning_rate": 1.9953926504473446e-06, "loss": 1.2433, "step": 769 }, { "epoch": 0.06, "learning_rate": 1.9953685152866785e-06, "loss": 1.3036, "step": 770 }, { "epoch": 0.06, "learning_rate": 1.9953443172229047e-06, "loss": 1.2503, "step": 771 }, { "epoch": 0.06, "learning_rate": 1.995320056257552e-06, "loss": 1.2773, "step": 772 }, { "epoch": 0.06, "learning_rate": 1.995295732392154e-06, "loss": 1.3089, "step": 773 }, { "epoch": 0.06, "learning_rate": 1.995271345628248e-06, "loss": 0.8692, "step": 774 }, { "epoch": 0.06, "learning_rate": 1.9952468959673747e-06, "loss": 1.3279, "step": 775 }, { "epoch": 0.06, "learning_rate": 1.99522238341108e-06, "loss": 1.2985, "step": 776 }, { "epoch": 0.06, "learning_rate": 1.9951978079609116e-06, "loss": 1.3472, "step": 777 }, { "epoch": 0.06, "learning_rate": 1.9951731696184234e-06, "loss": 1.2837, "step": 778 }, { "epoch": 0.06, "learning_rate": 1.995148468385173e-06, "loss": 0.8659, "step": 779 }, { "epoch": 0.06, "learning_rate": 1.9951237042627206e-06, "loss": 1.3141, "step": 780 }, { "epoch": 0.06, "learning_rate": 1.995098877252631e-06, "loss": 1.2944, "step": 781 }, { "epoch": 0.06, "learning_rate": 1.995073987356474e-06, "loss": 0.8855, "step": 782 }, { "epoch": 0.06, "learning_rate": 1.9950490345758225e-06, "loss": 1.3305, "step": 783 }, { "epoch": 0.06, "learning_rate": 1.9950240189122524e-06, "loss": 1.2995, "step": 784 }, { "epoch": 0.06, "learning_rate": 1.9949989403673454e-06, "loss": 1.3264, "step": 785 }, { "epoch": 0.06, "learning_rate": 1.9949737989426862e-06, "loss": 1.2855, "step": 786 }, { "epoch": 0.06, "learning_rate": 1.9949485946398636e-06, "loss": 1.3535, "step": 787 }, { "epoch": 0.06, "learning_rate": 1.9949233274604706e-06, "loss": 1.3004, "step": 788 }, { "epoch": 0.06, "learning_rate": 1.9948979974061036e-06, "loss": 0.8624, "step": 789 }, { "epoch": 0.06, "learning_rate": 1.9948726044783634e-06, "loss": 1.2797, "step": 790 }, { "epoch": 0.06, "learning_rate": 1.9948471486788553e-06, "loss": 1.2632, "step": 791 }, { "epoch": 0.06, "learning_rate": 1.994821630009187e-06, "loss": 1.2946, "step": 792 }, { "epoch": 0.06, "learning_rate": 1.9947960484709723e-06, "loss": 1.2989, "step": 793 }, { "epoch": 0.06, "learning_rate": 1.994770404065827e-06, "loss": 1.2819, "step": 794 }, { "epoch": 0.06, "learning_rate": 1.9947446967953725e-06, "loss": 1.2721, "step": 795 }, { "epoch": 0.06, "learning_rate": 1.9947189266612324e-06, "loss": 1.2393, "step": 796 }, { "epoch": 0.06, "learning_rate": 1.9946930936650356e-06, "loss": 1.2902, "step": 797 }, { "epoch": 0.06, "learning_rate": 1.994667197808415e-06, "loss": 1.3824, "step": 798 }, { "epoch": 0.06, "learning_rate": 1.9946412390930073e-06, "loss": 1.2854, "step": 799 }, { "epoch": 0.06, "learning_rate": 1.994615217520452e-06, "loss": 1.3114, "step": 800 }, { "epoch": 0.06, "learning_rate": 1.994589133092394e-06, "loss": 1.3194, "step": 801 }, { "epoch": 0.06, "learning_rate": 1.994562985810483e-06, "loss": 1.3563, "step": 802 }, { "epoch": 0.06, "learning_rate": 1.994536775676369e-06, "loss": 1.284, "step": 803 }, { "epoch": 0.06, "learning_rate": 1.9945105026917103e-06, "loss": 1.3303, "step": 804 }, { "epoch": 0.06, "learning_rate": 1.994484166858167e-06, "loss": 0.8672, "step": 805 }, { "epoch": 0.06, "learning_rate": 1.994457768177402e-06, "loss": 1.2942, "step": 806 }, { "epoch": 0.06, "learning_rate": 1.994431306651085e-06, "loss": 1.3249, "step": 807 }, { "epoch": 0.06, "learning_rate": 1.994404782280888e-06, "loss": 1.3062, "step": 808 }, { "epoch": 0.06, "learning_rate": 1.9943781950684874e-06, "loss": 1.3036, "step": 809 }, { "epoch": 0.06, "learning_rate": 1.9943515450155623e-06, "loss": 1.3314, "step": 810 }, { "epoch": 0.06, "learning_rate": 1.994324832123798e-06, "loss": 0.8557, "step": 811 }, { "epoch": 0.06, "learning_rate": 1.9942980563948824e-06, "loss": 1.3246, "step": 812 }, { "epoch": 0.06, "learning_rate": 1.9942712178305075e-06, "loss": 1.3605, "step": 813 }, { "epoch": 0.06, "learning_rate": 1.9942443164323694e-06, "loss": 1.3088, "step": 814 }, { "epoch": 0.06, "learning_rate": 1.9942173522021683e-06, "loss": 1.3057, "step": 815 }, { "epoch": 0.06, "learning_rate": 1.9941903251416077e-06, "loss": 1.3089, "step": 816 }, { "epoch": 0.06, "learning_rate": 1.994163235252396e-06, "loss": 1.3275, "step": 817 }, { "epoch": 0.06, "learning_rate": 1.994136082536245e-06, "loss": 1.3158, "step": 818 }, { "epoch": 0.06, "learning_rate": 1.994108866994871e-06, "loss": 1.3151, "step": 819 }, { "epoch": 0.06, "learning_rate": 1.9940815886299937e-06, "loss": 1.3211, "step": 820 }, { "epoch": 0.06, "learning_rate": 1.994054247443337e-06, "loss": 1.2989, "step": 821 }, { "epoch": 0.06, "learning_rate": 1.9940268434366285e-06, "loss": 1.3461, "step": 822 }, { "epoch": 0.06, "learning_rate": 1.9939993766116003e-06, "loss": 1.2355, "step": 823 }, { "epoch": 0.06, "learning_rate": 1.993971846969988e-06, "loss": 1.2871, "step": 824 }, { "epoch": 0.06, "learning_rate": 1.9939442545135314e-06, "loss": 1.2661, "step": 825 }, { "epoch": 0.06, "learning_rate": 1.9939165992439746e-06, "loss": 1.3391, "step": 826 }, { "epoch": 0.06, "learning_rate": 1.9938888811630647e-06, "loss": 1.3056, "step": 827 }, { "epoch": 0.06, "learning_rate": 1.993861100272554e-06, "loss": 1.2872, "step": 828 }, { "epoch": 0.06, "learning_rate": 1.993833256574198e-06, "loss": 1.2929, "step": 829 }, { "epoch": 0.06, "learning_rate": 1.9938053500697553e-06, "loss": 1.3109, "step": 830 }, { "epoch": 0.06, "learning_rate": 1.993777380760991e-06, "loss": 1.3229, "step": 831 }, { "epoch": 0.06, "learning_rate": 1.9937493486496712e-06, "loss": 1.3034, "step": 832 }, { "epoch": 0.06, "learning_rate": 1.9937212537375683e-06, "loss": 1.2984, "step": 833 }, { "epoch": 0.06, "learning_rate": 1.9936930960264584e-06, "loss": 1.2851, "step": 834 }, { "epoch": 0.06, "learning_rate": 1.9936648755181193e-06, "loss": 1.3035, "step": 835 }, { "epoch": 0.06, "learning_rate": 1.9936365922143354e-06, "loss": 1.2774, "step": 836 }, { "epoch": 0.06, "learning_rate": 1.9936082461168943e-06, "loss": 1.3096, "step": 837 }, { "epoch": 0.07, "learning_rate": 1.9935798372275866e-06, "loss": 1.3198, "step": 838 }, { "epoch": 0.07, "learning_rate": 1.993551365548208e-06, "loss": 0.8915, "step": 839 }, { "epoch": 0.07, "learning_rate": 1.9935228310805583e-06, "loss": 0.8778, "step": 840 }, { "epoch": 0.07, "learning_rate": 1.99349423382644e-06, "loss": 1.3142, "step": 841 }, { "epoch": 0.07, "learning_rate": 1.993465573787661e-06, "loss": 1.276, "step": 842 }, { "epoch": 0.07, "learning_rate": 1.9934368509660316e-06, "loss": 1.3338, "step": 843 }, { "epoch": 0.07, "learning_rate": 1.9934080653633676e-06, "loss": 1.3105, "step": 844 }, { "epoch": 0.07, "learning_rate": 1.993379216981488e-06, "loss": 1.2966, "step": 845 }, { "epoch": 0.07, "learning_rate": 1.993350305822216e-06, "loss": 1.3911, "step": 846 }, { "epoch": 0.07, "learning_rate": 1.9933213318873784e-06, "loss": 1.2709, "step": 847 }, { "epoch": 0.07, "learning_rate": 1.9932922951788063e-06, "loss": 1.28, "step": 848 }, { "epoch": 0.07, "learning_rate": 1.993263195698335e-06, "loss": 0.8934, "step": 849 }, { "epoch": 0.07, "learning_rate": 1.993234033447803e-06, "loss": 0.8802, "step": 850 }, { "epoch": 0.07, "learning_rate": 1.993204808429054e-06, "loss": 1.3002, "step": 851 }, { "epoch": 0.07, "learning_rate": 1.993175520643934e-06, "loss": 1.2849, "step": 852 }, { "epoch": 0.07, "learning_rate": 1.993146170094294e-06, "loss": 1.2455, "step": 853 }, { "epoch": 0.07, "learning_rate": 1.9931167567819895e-06, "loss": 1.2607, "step": 854 }, { "epoch": 0.07, "learning_rate": 1.9930872807088785e-06, "loss": 1.2663, "step": 855 }, { "epoch": 0.07, "learning_rate": 1.9930577418768243e-06, "loss": 1.3155, "step": 856 }, { "epoch": 0.07, "learning_rate": 1.9930281402876937e-06, "loss": 1.3108, "step": 857 }, { "epoch": 0.07, "learning_rate": 1.992998475943357e-06, "loss": 1.3391, "step": 858 }, { "epoch": 0.07, "learning_rate": 1.9929687488456886e-06, "loss": 1.3144, "step": 859 }, { "epoch": 0.07, "learning_rate": 1.9929389589965682e-06, "loss": 1.3207, "step": 860 }, { "epoch": 0.07, "learning_rate": 1.9929091063978774e-06, "loss": 1.2905, "step": 861 }, { "epoch": 0.07, "learning_rate": 1.9928791910515034e-06, "loss": 1.2698, "step": 862 }, { "epoch": 0.07, "learning_rate": 1.992849212959336e-06, "loss": 1.3015, "step": 863 }, { "epoch": 0.07, "learning_rate": 1.9928191721232707e-06, "loss": 0.8971, "step": 864 }, { "epoch": 0.07, "learning_rate": 1.992789068545205e-06, "loss": 1.3323, "step": 865 }, { "epoch": 0.07, "learning_rate": 1.9927589022270417e-06, "loss": 1.3159, "step": 866 }, { "epoch": 0.07, "learning_rate": 1.992728673170687e-06, "loss": 1.2944, "step": 867 }, { "epoch": 0.07, "learning_rate": 1.9926983813780515e-06, "loss": 1.2521, "step": 868 }, { "epoch": 0.07, "learning_rate": 1.99266802685105e-06, "loss": 1.3023, "step": 869 }, { "epoch": 0.07, "learning_rate": 1.9926376095915997e-06, "loss": 0.8926, "step": 870 }, { "epoch": 0.07, "learning_rate": 1.992607129601623e-06, "loss": 1.3001, "step": 871 }, { "epoch": 0.07, "learning_rate": 1.9925765868830473e-06, "loss": 1.2623, "step": 872 }, { "epoch": 0.07, "learning_rate": 1.9925459814378015e-06, "loss": 1.2963, "step": 873 }, { "epoch": 0.07, "learning_rate": 1.99251531326782e-06, "loss": 1.2938, "step": 874 }, { "epoch": 0.07, "learning_rate": 1.9924845823750414e-06, "loss": 1.2973, "step": 875 }, { "epoch": 0.07, "learning_rate": 1.9924537887614073e-06, "loss": 1.2931, "step": 876 }, { "epoch": 0.07, "learning_rate": 1.992422932428864e-06, "loss": 1.2787, "step": 877 }, { "epoch": 0.07, "learning_rate": 1.9923920133793614e-06, "loss": 0.8702, "step": 878 }, { "epoch": 0.07, "learning_rate": 1.992361031614853e-06, "loss": 1.3219, "step": 879 }, { "epoch": 0.07, "learning_rate": 1.9923299871372973e-06, "loss": 1.3305, "step": 880 }, { "epoch": 0.07, "learning_rate": 1.9922988799486564e-06, "loss": 1.3063, "step": 881 }, { "epoch": 0.07, "learning_rate": 1.992267710050895e-06, "loss": 1.2785, "step": 882 }, { "epoch": 0.07, "learning_rate": 1.9922364774459843e-06, "loss": 1.3318, "step": 883 }, { "epoch": 0.07, "learning_rate": 1.9922051821358972e-06, "loss": 1.2767, "step": 884 }, { "epoch": 0.07, "learning_rate": 1.9921738241226115e-06, "loss": 1.2531, "step": 885 }, { "epoch": 0.07, "learning_rate": 1.9921424034081095e-06, "loss": 1.324, "step": 886 }, { "epoch": 0.07, "learning_rate": 1.992110919994376e-06, "loss": 0.8865, "step": 887 }, { "epoch": 0.07, "learning_rate": 1.9920793738834016e-06, "loss": 1.2696, "step": 888 }, { "epoch": 0.07, "learning_rate": 1.9920477650771788e-06, "loss": 1.2754, "step": 889 }, { "epoch": 0.07, "learning_rate": 1.9920160935777057e-06, "loss": 1.258, "step": 890 }, { "epoch": 0.07, "learning_rate": 1.991984359386984e-06, "loss": 1.3045, "step": 891 }, { "epoch": 0.07, "learning_rate": 1.9919525625070185e-06, "loss": 1.3257, "step": 892 }, { "epoch": 0.07, "learning_rate": 1.991920702939819e-06, "loss": 1.3366, "step": 893 }, { "epoch": 0.07, "learning_rate": 1.9918887806873993e-06, "loss": 1.3058, "step": 894 }, { "epoch": 0.07, "learning_rate": 1.9918567957517765e-06, "loss": 1.2821, "step": 895 }, { "epoch": 0.07, "learning_rate": 1.991824748134972e-06, "loss": 0.8725, "step": 896 }, { "epoch": 0.07, "learning_rate": 1.9917926378390103e-06, "loss": 1.275, "step": 897 }, { "epoch": 0.07, "learning_rate": 1.9917604648659212e-06, "loss": 1.2965, "step": 898 }, { "epoch": 0.07, "learning_rate": 1.9917282292177384e-06, "loss": 1.3177, "step": 899 }, { "epoch": 0.07, "learning_rate": 1.9916959308964986e-06, "loss": 1.2333, "step": 900 }, { "epoch": 0.07, "learning_rate": 1.991663569904242e-06, "loss": 1.2581, "step": 901 }, { "epoch": 0.07, "learning_rate": 1.9916311462430157e-06, "loss": 1.3357, "step": 902 }, { "epoch": 0.07, "learning_rate": 1.991598659914867e-06, "loss": 1.3092, "step": 903 }, { "epoch": 0.07, "learning_rate": 1.99156611092185e-06, "loss": 1.3395, "step": 904 }, { "epoch": 0.07, "learning_rate": 1.991533499266021e-06, "loss": 1.309, "step": 905 }, { "epoch": 0.07, "learning_rate": 1.991500824949441e-06, "loss": 1.374, "step": 906 }, { "epoch": 0.07, "learning_rate": 1.9914680879741746e-06, "loss": 0.8908, "step": 907 }, { "epoch": 0.07, "learning_rate": 1.9914352883422913e-06, "loss": 1.2713, "step": 908 }, { "epoch": 0.07, "learning_rate": 1.9914024260558637e-06, "loss": 0.9248, "step": 909 }, { "epoch": 0.07, "learning_rate": 1.9913695011169684e-06, "loss": 1.2943, "step": 910 }, { "epoch": 0.07, "learning_rate": 1.9913365135276864e-06, "loss": 1.2805, "step": 911 }, { "epoch": 0.07, "learning_rate": 1.991303463290102e-06, "loss": 1.2137, "step": 912 }, { "epoch": 0.07, "learning_rate": 1.9912703504063036e-06, "loss": 1.3099, "step": 913 }, { "epoch": 0.07, "learning_rate": 1.991237174878385e-06, "loss": 1.3507, "step": 914 }, { "epoch": 0.07, "learning_rate": 1.9912039367084414e-06, "loss": 1.2989, "step": 915 }, { "epoch": 0.07, "learning_rate": 1.991170635898574e-06, "loss": 1.3059, "step": 916 }, { "epoch": 0.07, "learning_rate": 1.991137272450887e-06, "loss": 1.2581, "step": 917 }, { "epoch": 0.07, "learning_rate": 1.991103846367489e-06, "loss": 1.2907, "step": 918 }, { "epoch": 0.07, "learning_rate": 1.9910703576504927e-06, "loss": 0.9189, "step": 919 }, { "epoch": 0.07, "learning_rate": 1.991036806302014e-06, "loss": 0.8889, "step": 920 }, { "epoch": 0.07, "learning_rate": 1.991003192324173e-06, "loss": 1.2364, "step": 921 }, { "epoch": 0.07, "learning_rate": 1.9909695157190943e-06, "loss": 1.2619, "step": 922 }, { "epoch": 0.07, "learning_rate": 1.990935776488906e-06, "loss": 1.2924, "step": 923 }, { "epoch": 0.07, "learning_rate": 1.9909019746357406e-06, "loss": 1.2534, "step": 924 }, { "epoch": 0.07, "learning_rate": 1.990868110161734e-06, "loss": 1.2197, "step": 925 }, { "epoch": 0.07, "learning_rate": 1.9908341830690265e-06, "loss": 1.2747, "step": 926 }, { "epoch": 0.07, "learning_rate": 1.9908001933597614e-06, "loss": 1.2877, "step": 927 }, { "epoch": 0.07, "learning_rate": 1.9907661410360874e-06, "loss": 1.2664, "step": 928 }, { "epoch": 0.07, "learning_rate": 1.9907320261001568e-06, "loss": 1.3102, "step": 929 }, { "epoch": 0.07, "learning_rate": 1.9906978485541244e-06, "loss": 1.2947, "step": 930 }, { "epoch": 0.07, "learning_rate": 1.990663608400151e-06, "loss": 1.3239, "step": 931 }, { "epoch": 0.07, "learning_rate": 1.9906293056404e-06, "loss": 1.2589, "step": 932 }, { "epoch": 0.07, "learning_rate": 1.9905949402770394e-06, "loss": 1.2488, "step": 933 }, { "epoch": 0.07, "learning_rate": 1.9905605123122413e-06, "loss": 1.3114, "step": 934 }, { "epoch": 0.07, "learning_rate": 1.9905260217481805e-06, "loss": 0.8647, "step": 935 }, { "epoch": 0.07, "learning_rate": 1.9904914685870373e-06, "loss": 1.3126, "step": 936 }, { "epoch": 0.07, "learning_rate": 1.9904568528309952e-06, "loss": 1.3384, "step": 937 }, { "epoch": 0.07, "learning_rate": 1.9904221744822415e-06, "loss": 1.2768, "step": 938 }, { "epoch": 0.07, "learning_rate": 1.990387433542968e-06, "loss": 1.3255, "step": 939 }, { "epoch": 0.07, "learning_rate": 1.99035263001537e-06, "loss": 1.3383, "step": 940 }, { "epoch": 0.07, "learning_rate": 1.9903177639016475e-06, "loss": 1.2718, "step": 941 }, { "epoch": 0.07, "learning_rate": 1.990282835204003e-06, "loss": 1.3151, "step": 942 }, { "epoch": 0.07, "learning_rate": 1.990247843924645e-06, "loss": 1.2793, "step": 943 }, { "epoch": 0.07, "learning_rate": 1.9902127900657836e-06, "loss": 0.873, "step": 944 }, { "epoch": 0.07, "learning_rate": 1.9901776736296346e-06, "loss": 1.328, "step": 945 }, { "epoch": 0.07, "learning_rate": 1.990142494618417e-06, "loss": 1.2559, "step": 946 }, { "epoch": 0.07, "learning_rate": 1.990107253034354e-06, "loss": 1.2842, "step": 947 }, { "epoch": 0.07, "learning_rate": 1.990071948879674e-06, "loss": 1.2837, "step": 948 }, { "epoch": 0.07, "learning_rate": 1.9900365821566057e-06, "loss": 1.2541, "step": 949 }, { "epoch": 0.07, "learning_rate": 1.990001152867386e-06, "loss": 1.2688, "step": 950 }, { "epoch": 0.07, "learning_rate": 1.989965661014253e-06, "loss": 1.2763, "step": 951 }, { "epoch": 0.07, "learning_rate": 1.98993010659945e-06, "loss": 1.276, "step": 952 }, { "epoch": 0.07, "learning_rate": 1.9898944896252234e-06, "loss": 1.3314, "step": 953 }, { "epoch": 0.07, "learning_rate": 1.989858810093825e-06, "loss": 1.289, "step": 954 }, { "epoch": 0.07, "learning_rate": 1.9898230680075088e-06, "loss": 1.3232, "step": 955 }, { "epoch": 0.07, "learning_rate": 1.989787263368534e-06, "loss": 0.8663, "step": 956 }, { "epoch": 0.07, "learning_rate": 1.989751396179162e-06, "loss": 1.3086, "step": 957 }, { "epoch": 0.07, "learning_rate": 1.9897154664416614e-06, "loss": 1.2892, "step": 958 }, { "epoch": 0.07, "learning_rate": 1.989679474158302e-06, "loss": 1.2887, "step": 959 }, { "epoch": 0.07, "learning_rate": 1.989643419331358e-06, "loss": 1.3086, "step": 960 }, { "epoch": 0.07, "learning_rate": 1.9896073019631087e-06, "loss": 1.2668, "step": 961 }, { "epoch": 0.07, "learning_rate": 1.9895711220558356e-06, "loss": 1.2585, "step": 962 }, { "epoch": 0.07, "learning_rate": 1.9895348796118256e-06, "loss": 1.2644, "step": 963 }, { "epoch": 0.07, "learning_rate": 1.989498574633369e-06, "loss": 1.1971, "step": 964 }, { "epoch": 0.07, "learning_rate": 1.9894622071227605e-06, "loss": 1.272, "step": 965 }, { "epoch": 0.07, "learning_rate": 1.989425777082298e-06, "loss": 1.2645, "step": 966 }, { "epoch": 0.08, "learning_rate": 1.9893892845142837e-06, "loss": 1.2715, "step": 967 }, { "epoch": 0.08, "learning_rate": 1.9893527294210236e-06, "loss": 1.2759, "step": 968 }, { "epoch": 0.08, "learning_rate": 1.989316111804828e-06, "loss": 1.2732, "step": 969 }, { "epoch": 0.08, "learning_rate": 1.989279431668012e-06, "loss": 1.3313, "step": 970 }, { "epoch": 0.08, "learning_rate": 1.9892426890128915e-06, "loss": 1.2366, "step": 971 }, { "epoch": 0.08, "learning_rate": 1.9892058838417905e-06, "loss": 1.3219, "step": 972 }, { "epoch": 0.08, "learning_rate": 1.9891690161570338e-06, "loss": 1.3001, "step": 973 }, { "epoch": 0.08, "learning_rate": 1.9891320859609516e-06, "loss": 1.2918, "step": 974 }, { "epoch": 0.08, "learning_rate": 1.9890950932558777e-06, "loss": 1.281, "step": 975 }, { "epoch": 0.08, "learning_rate": 1.98905803804415e-06, "loss": 1.3136, "step": 976 }, { "epoch": 0.08, "learning_rate": 1.98902092032811e-06, "loss": 1.331, "step": 977 }, { "epoch": 0.08, "learning_rate": 1.988983740110104e-06, "loss": 1.2832, "step": 978 }, { "epoch": 0.08, "learning_rate": 1.988946497392481e-06, "loss": 0.9025, "step": 979 }, { "epoch": 0.08, "learning_rate": 1.988909192177594e-06, "loss": 1.3327, "step": 980 }, { "epoch": 0.08, "learning_rate": 1.988871824467802e-06, "loss": 1.2681, "step": 981 }, { "epoch": 0.08, "learning_rate": 1.9888343942654656e-06, "loss": 1.2923, "step": 982 }, { "epoch": 0.08, "learning_rate": 1.98879690157295e-06, "loss": 0.8785, "step": 983 }, { "epoch": 0.08, "learning_rate": 1.9887593463926256e-06, "loss": 1.3133, "step": 984 }, { "epoch": 0.08, "learning_rate": 1.988721728726865e-06, "loss": 0.9004, "step": 985 }, { "epoch": 0.08, "learning_rate": 1.988684048578045e-06, "loss": 1.2603, "step": 986 }, { "epoch": 0.08, "learning_rate": 1.988646305948548e-06, "loss": 1.3407, "step": 987 }, { "epoch": 0.08, "learning_rate": 1.9886085008407584e-06, "loss": 1.293, "step": 988 }, { "epoch": 0.08, "learning_rate": 1.9885706332570656e-06, "loss": 1.2729, "step": 989 }, { "epoch": 0.08, "learning_rate": 1.9885327031998623e-06, "loss": 0.8749, "step": 990 }, { "epoch": 0.08, "learning_rate": 1.988494710671546e-06, "loss": 1.3001, "step": 991 }, { "epoch": 0.08, "learning_rate": 1.9884566556745174e-06, "loss": 1.2748, "step": 992 }, { "epoch": 0.08, "learning_rate": 1.9884185382111814e-06, "loss": 1.3322, "step": 993 }, { "epoch": 0.08, "learning_rate": 1.9883803582839475e-06, "loss": 1.2895, "step": 994 }, { "epoch": 0.08, "learning_rate": 1.9883421158952275e-06, "loss": 1.3058, "step": 995 }, { "epoch": 0.08, "learning_rate": 1.9883038110474385e-06, "loss": 0.8699, "step": 996 }, { "epoch": 0.08, "learning_rate": 1.9882654437430014e-06, "loss": 1.2845, "step": 997 }, { "epoch": 0.08, "learning_rate": 1.9882270139843407e-06, "loss": 0.8582, "step": 998 }, { "epoch": 0.08, "learning_rate": 1.9881885217738854e-06, "loss": 1.2018, "step": 999 }, { "epoch": 0.08, "learning_rate": 1.9881499671140676e-06, "loss": 1.302, "step": 1000 }, { "epoch": 0.08, "learning_rate": 1.988111350007324e-06, "loss": 1.2466, "step": 1001 }, { "epoch": 0.08, "learning_rate": 1.9880726704560945e-06, "loss": 1.3104, "step": 1002 }, { "epoch": 0.08, "learning_rate": 1.988033928462824e-06, "loss": 1.271, "step": 1003 }, { "epoch": 0.08, "learning_rate": 1.987995124029961e-06, "loss": 1.2956, "step": 1004 }, { "epoch": 0.08, "learning_rate": 1.987956257159958e-06, "loss": 1.275, "step": 1005 }, { "epoch": 0.08, "learning_rate": 1.98791732785527e-06, "loss": 1.3235, "step": 1006 }, { "epoch": 0.08, "learning_rate": 1.9878783361183583e-06, "loss": 1.3276, "step": 1007 }, { "epoch": 0.08, "learning_rate": 1.9878392819516863e-06, "loss": 1.2808, "step": 1008 }, { "epoch": 0.08, "learning_rate": 1.9878001653577227e-06, "loss": 0.9148, "step": 1009 }, { "epoch": 0.08, "learning_rate": 1.987760986338939e-06, "loss": 1.2992, "step": 1010 }, { "epoch": 0.08, "learning_rate": 1.987721744897812e-06, "loss": 1.3166, "step": 1011 }, { "epoch": 0.08, "learning_rate": 1.9876824410368203e-06, "loss": 1.2489, "step": 1012 }, { "epoch": 0.08, "learning_rate": 1.9876430747584486e-06, "loss": 1.292, "step": 1013 }, { "epoch": 0.08, "learning_rate": 1.9876036460651844e-06, "loss": 1.2665, "step": 1014 }, { "epoch": 0.08, "learning_rate": 1.9875641549595194e-06, "loss": 1.325, "step": 1015 }, { "epoch": 0.08, "learning_rate": 1.9875246014439497e-06, "loss": 1.3235, "step": 1016 }, { "epoch": 0.08, "learning_rate": 1.9874849855209745e-06, "loss": 1.2681, "step": 1017 }, { "epoch": 0.08, "learning_rate": 1.987445307193097e-06, "loss": 1.3062, "step": 1018 }, { "epoch": 0.08, "learning_rate": 1.987405566462826e-06, "loss": 1.2697, "step": 1019 }, { "epoch": 0.08, "learning_rate": 1.987365763332671e-06, "loss": 0.843, "step": 1020 }, { "epoch": 0.08, "learning_rate": 1.9873258978051495e-06, "loss": 1.2524, "step": 1021 }, { "epoch": 0.08, "learning_rate": 1.9872859698827793e-06, "loss": 1.345, "step": 1022 }, { "epoch": 0.08, "learning_rate": 1.9872459795680845e-06, "loss": 1.2795, "step": 1023 }, { "epoch": 0.08, "learning_rate": 1.987205926863592e-06, "loss": 1.3247, "step": 1024 }, { "epoch": 0.08, "learning_rate": 1.987165811771833e-06, "loss": 1.2663, "step": 1025 }, { "epoch": 0.08, "learning_rate": 1.987125634295342e-06, "loss": 1.323, "step": 1026 }, { "epoch": 0.08, "learning_rate": 1.987085394436659e-06, "loss": 1.3467, "step": 1027 }, { "epoch": 0.08, "learning_rate": 1.9870450921983267e-06, "loss": 1.3013, "step": 1028 }, { "epoch": 0.08, "learning_rate": 1.987004727582892e-06, "loss": 1.3073, "step": 1029 }, { "epoch": 0.08, "learning_rate": 1.986964300592906e-06, "loss": 1.2821, "step": 1030 }, { "epoch": 0.08, "learning_rate": 1.986923811230923e-06, "loss": 1.2968, "step": 1031 }, { "epoch": 0.08, "learning_rate": 1.986883259499502e-06, "loss": 1.2776, "step": 1032 }, { "epoch": 0.08, "learning_rate": 1.9868426454012054e-06, "loss": 1.2684, "step": 1033 }, { "epoch": 0.08, "learning_rate": 1.9868019689386007e-06, "loss": 1.3089, "step": 1034 }, { "epoch": 0.08, "learning_rate": 1.986761230114258e-06, "loss": 1.2653, "step": 1035 }, { "epoch": 0.08, "learning_rate": 1.986720428930751e-06, "loss": 1.3417, "step": 1036 }, { "epoch": 0.08, "learning_rate": 1.9866795653906595e-06, "loss": 1.3056, "step": 1037 }, { "epoch": 0.08, "learning_rate": 1.9866386394965653e-06, "loss": 0.8478, "step": 1038 }, { "epoch": 0.08, "learning_rate": 1.9865976512510544e-06, "loss": 1.2502, "step": 1039 }, { "epoch": 0.08, "learning_rate": 1.9865566006567176e-06, "loss": 1.3009, "step": 1040 }, { "epoch": 0.08, "learning_rate": 1.986515487716149e-06, "loss": 1.3273, "step": 1041 }, { "epoch": 0.08, "learning_rate": 1.986474312431947e-06, "loss": 0.8885, "step": 1042 }, { "epoch": 0.08, "learning_rate": 1.986433074806713e-06, "loss": 1.2377, "step": 1043 }, { "epoch": 0.08, "learning_rate": 1.9863917748430538e-06, "loss": 1.2687, "step": 1044 }, { "epoch": 0.08, "learning_rate": 1.986350412543579e-06, "loss": 1.2854, "step": 1045 }, { "epoch": 0.08, "learning_rate": 1.9863089879109027e-06, "loss": 1.3561, "step": 1046 }, { "epoch": 0.08, "learning_rate": 1.9862675009476425e-06, "loss": 1.2948, "step": 1047 }, { "epoch": 0.08, "learning_rate": 1.9862259516564204e-06, "loss": 0.8811, "step": 1048 }, { "epoch": 0.08, "learning_rate": 1.9861843400398623e-06, "loss": 1.2792, "step": 1049 }, { "epoch": 0.08, "learning_rate": 1.9861426661005973e-06, "loss": 1.3413, "step": 1050 }, { "epoch": 0.08, "learning_rate": 1.9861009298412594e-06, "loss": 1.2795, "step": 1051 }, { "epoch": 0.08, "learning_rate": 1.9860591312644865e-06, "loss": 1.2938, "step": 1052 }, { "epoch": 0.08, "learning_rate": 1.9860172703729197e-06, "loss": 1.2853, "step": 1053 }, { "epoch": 0.08, "learning_rate": 1.9859753471692043e-06, "loss": 1.2917, "step": 1054 }, { "epoch": 0.08, "learning_rate": 1.98593336165599e-06, "loss": 1.2597, "step": 1055 }, { "epoch": 0.08, "learning_rate": 1.9858913138359297e-06, "loss": 1.2886, "step": 1056 }, { "epoch": 0.08, "learning_rate": 1.9858492037116813e-06, "loss": 1.2757, "step": 1057 }, { "epoch": 0.08, "learning_rate": 1.9858070312859053e-06, "loss": 1.2689, "step": 1058 }, { "epoch": 0.08, "learning_rate": 1.9857647965612672e-06, "loss": 0.847, "step": 1059 }, { "epoch": 0.08, "learning_rate": 1.985722499540436e-06, "loss": 1.3263, "step": 1060 }, { "epoch": 0.08, "learning_rate": 1.9856801402260844e-06, "loss": 0.8686, "step": 1061 }, { "epoch": 0.08, "learning_rate": 1.98563771862089e-06, "loss": 1.3175, "step": 1062 }, { "epoch": 0.08, "learning_rate": 1.985595234727533e-06, "loss": 1.274, "step": 1063 }, { "epoch": 0.08, "learning_rate": 1.9855526885486986e-06, "loss": 1.2806, "step": 1064 }, { "epoch": 0.08, "learning_rate": 1.985510080087075e-06, "loss": 1.2886, "step": 1065 }, { "epoch": 0.08, "learning_rate": 1.9854674093453562e-06, "loss": 1.28, "step": 1066 }, { "epoch": 0.08, "learning_rate": 1.985424676326237e-06, "loss": 0.8632, "step": 1067 }, { "epoch": 0.08, "learning_rate": 1.9853818810324192e-06, "loss": 1.2414, "step": 1068 }, { "epoch": 0.08, "learning_rate": 1.985339023466607e-06, "loss": 1.2898, "step": 1069 }, { "epoch": 0.08, "learning_rate": 1.9852961036315086e-06, "loss": 1.2616, "step": 1070 }, { "epoch": 0.08, "learning_rate": 1.9852531215298364e-06, "loss": 1.2486, "step": 1071 }, { "epoch": 0.08, "learning_rate": 1.9852100771643067e-06, "loss": 1.2725, "step": 1072 }, { "epoch": 0.08, "learning_rate": 1.98516697053764e-06, "loss": 1.242, "step": 1073 }, { "epoch": 0.08, "learning_rate": 1.9851238016525602e-06, "loss": 1.3039, "step": 1074 }, { "epoch": 0.08, "learning_rate": 1.9850805705117954e-06, "loss": 1.3308, "step": 1075 }, { "epoch": 0.08, "learning_rate": 1.9850372771180777e-06, "loss": 1.3167, "step": 1076 }, { "epoch": 0.08, "learning_rate": 1.9849939214741436e-06, "loss": 0.9227, "step": 1077 }, { "epoch": 0.08, "learning_rate": 1.9849505035827313e-06, "loss": 1.3148, "step": 1078 }, { "epoch": 0.08, "learning_rate": 1.9849070234465867e-06, "loss": 0.8625, "step": 1079 }, { "epoch": 0.08, "learning_rate": 1.984863481068456e-06, "loss": 1.2802, "step": 1080 }, { "epoch": 0.08, "learning_rate": 1.9848198764510918e-06, "loss": 1.3068, "step": 1081 }, { "epoch": 0.08, "learning_rate": 1.9847762095972497e-06, "loss": 1.2435, "step": 1082 }, { "epoch": 0.08, "learning_rate": 1.984732480509689e-06, "loss": 0.8656, "step": 1083 }, { "epoch": 0.08, "learning_rate": 1.9846886891911726e-06, "loss": 1.3585, "step": 1084 }, { "epoch": 0.08, "learning_rate": 1.984644835644469e-06, "loss": 1.2397, "step": 1085 }, { "epoch": 0.08, "learning_rate": 1.984600919872349e-06, "loss": 1.2522, "step": 1086 }, { "epoch": 0.08, "learning_rate": 1.984556941877588e-06, "loss": 1.2516, "step": 1087 }, { "epoch": 0.08, "learning_rate": 1.984512901662965e-06, "loss": 1.3059, "step": 1088 }, { "epoch": 0.08, "learning_rate": 1.984468799231264e-06, "loss": 1.2789, "step": 1089 }, { "epoch": 0.08, "learning_rate": 1.9844246345852708e-06, "loss": 0.8941, "step": 1090 }, { "epoch": 0.08, "learning_rate": 1.9843804077277773e-06, "loss": 1.2528, "step": 1091 }, { "epoch": 0.08, "learning_rate": 1.984336118661578e-06, "loss": 0.8681, "step": 1092 }, { "epoch": 0.08, "learning_rate": 1.984291767389472e-06, "loss": 1.307, "step": 1093 }, { "epoch": 0.08, "learning_rate": 1.9842473539142625e-06, "loss": 1.2714, "step": 1094 }, { "epoch": 0.08, "learning_rate": 1.9842028782387558e-06, "loss": 1.306, "step": 1095 }, { "epoch": 0.09, "learning_rate": 1.984158340365763e-06, "loss": 1.272, "step": 1096 }, { "epoch": 0.09, "learning_rate": 1.9841137402980976e-06, "loss": 1.2959, "step": 1097 }, { "epoch": 0.09, "learning_rate": 1.984069078038579e-06, "loss": 1.285, "step": 1098 }, { "epoch": 0.09, "learning_rate": 1.9840243535900297e-06, "loss": 1.2535, "step": 1099 }, { "epoch": 0.09, "learning_rate": 1.9839795669552763e-06, "loss": 1.3134, "step": 1100 }, { "epoch": 0.09, "learning_rate": 1.9839347181371483e-06, "loss": 1.3013, "step": 1101 }, { "epoch": 0.09, "learning_rate": 1.9838898071384803e-06, "loss": 1.3202, "step": 1102 }, { "epoch": 0.09, "learning_rate": 1.983844833962111e-06, "loss": 1.3317, "step": 1103 }, { "epoch": 0.09, "learning_rate": 1.983799798610882e-06, "loss": 1.3043, "step": 1104 }, { "epoch": 0.09, "learning_rate": 1.98375470108764e-06, "loss": 1.2877, "step": 1105 }, { "epoch": 0.09, "learning_rate": 1.9837095413952336e-06, "loss": 1.2835, "step": 1106 }, { "epoch": 0.09, "learning_rate": 1.983664319536518e-06, "loss": 1.3019, "step": 1107 }, { "epoch": 0.09, "learning_rate": 1.9836190355143507e-06, "loss": 1.2703, "step": 1108 }, { "epoch": 0.09, "learning_rate": 1.9835736893315928e-06, "loss": 1.2561, "step": 1109 }, { "epoch": 0.09, "learning_rate": 1.983528280991111e-06, "loss": 1.2829, "step": 1110 }, { "epoch": 0.09, "learning_rate": 1.9834828104957746e-06, "loss": 1.3274, "step": 1111 }, { "epoch": 0.09, "learning_rate": 1.9834372778484563e-06, "loss": 1.2675, "step": 1112 }, { "epoch": 0.09, "learning_rate": 1.983391683052035e-06, "loss": 0.8373, "step": 1113 }, { "epoch": 0.09, "learning_rate": 1.983346026109391e-06, "loss": 1.2511, "step": 1114 }, { "epoch": 0.09, "learning_rate": 1.98330030702341e-06, "loss": 1.3504, "step": 1115 }, { "epoch": 0.09, "learning_rate": 1.983254525796981e-06, "loss": 1.3203, "step": 1116 }, { "epoch": 0.09, "learning_rate": 1.983208682432998e-06, "loss": 1.3046, "step": 1117 }, { "epoch": 0.09, "learning_rate": 1.9831627769343574e-06, "loss": 1.2737, "step": 1118 }, { "epoch": 0.09, "learning_rate": 1.98311680930396e-06, "loss": 0.8835, "step": 1119 }, { "epoch": 0.09, "learning_rate": 1.9830707795447113e-06, "loss": 1.3335, "step": 1120 }, { "epoch": 0.09, "learning_rate": 1.9830246876595204e-06, "loss": 0.9135, "step": 1121 }, { "epoch": 0.09, "learning_rate": 1.9829785336512996e-06, "loss": 1.2995, "step": 1122 }, { "epoch": 0.09, "learning_rate": 1.9829323175229657e-06, "loss": 1.2647, "step": 1123 }, { "epoch": 0.09, "learning_rate": 1.9828860392774393e-06, "loss": 1.2239, "step": 1124 }, { "epoch": 0.09, "learning_rate": 1.982839698917645e-06, "loss": 1.3352, "step": 1125 }, { "epoch": 0.09, "learning_rate": 1.982793296446512e-06, "loss": 1.2966, "step": 1126 }, { "epoch": 0.09, "learning_rate": 1.9827468318669715e-06, "loss": 1.318, "step": 1127 }, { "epoch": 0.09, "learning_rate": 1.9827003051819607e-06, "loss": 1.3088, "step": 1128 }, { "epoch": 0.09, "learning_rate": 1.98265371639442e-06, "loss": 1.2154, "step": 1129 }, { "epoch": 0.09, "learning_rate": 1.9826070655072935e-06, "loss": 1.2294, "step": 1130 }, { "epoch": 0.09, "learning_rate": 1.982560352523529e-06, "loss": 1.2566, "step": 1131 }, { "epoch": 0.09, "learning_rate": 1.982513577446079e-06, "loss": 1.3015, "step": 1132 }, { "epoch": 0.09, "learning_rate": 1.982466740277899e-06, "loss": 1.3022, "step": 1133 }, { "epoch": 0.09, "learning_rate": 1.982419841021949e-06, "loss": 1.3426, "step": 1134 }, { "epoch": 0.09, "learning_rate": 1.982372879681193e-06, "loss": 1.2952, "step": 1135 }, { "epoch": 0.09, "learning_rate": 1.982325856258599e-06, "loss": 1.2945, "step": 1136 }, { "epoch": 0.09, "learning_rate": 1.9822787707571385e-06, "loss": 0.9345, "step": 1137 }, { "epoch": 0.09, "learning_rate": 1.982231623179787e-06, "loss": 1.258, "step": 1138 }, { "epoch": 0.09, "learning_rate": 1.982184413529524e-06, "loss": 1.2694, "step": 1139 }, { "epoch": 0.09, "learning_rate": 1.9821371418093328e-06, "loss": 1.3245, "step": 1140 }, { "epoch": 0.09, "learning_rate": 1.9820898080222015e-06, "loss": 1.313, "step": 1141 }, { "epoch": 0.09, "learning_rate": 1.9820424121711205e-06, "loss": 1.1979, "step": 1142 }, { "epoch": 0.09, "learning_rate": 1.9819949542590857e-06, "loss": 1.2671, "step": 1143 }, { "epoch": 0.09, "learning_rate": 1.9819474342890957e-06, "loss": 1.2952, "step": 1144 }, { "epoch": 0.09, "learning_rate": 1.981899852264154e-06, "loss": 1.2195, "step": 1145 }, { "epoch": 0.09, "learning_rate": 1.9818522081872674e-06, "loss": 1.3568, "step": 1146 }, { "epoch": 0.09, "learning_rate": 1.981804502061447e-06, "loss": 1.3161, "step": 1147 }, { "epoch": 0.09, "learning_rate": 1.981756733889707e-06, "loss": 1.2599, "step": 1148 }, { "epoch": 0.09, "learning_rate": 1.981708903675067e-06, "loss": 1.3262, "step": 1149 }, { "epoch": 0.09, "learning_rate": 1.9816610114205486e-06, "loss": 0.8293, "step": 1150 }, { "epoch": 0.09, "learning_rate": 1.9816130571291798e-06, "loss": 1.2312, "step": 1151 }, { "epoch": 0.09, "learning_rate": 1.9815650408039897e-06, "loss": 1.2684, "step": 1152 }, { "epoch": 0.09, "learning_rate": 1.981516962448014e-06, "loss": 1.3312, "step": 1153 }, { "epoch": 0.09, "learning_rate": 1.98146882206429e-06, "loss": 1.2621, "step": 1154 }, { "epoch": 0.09, "learning_rate": 1.9814206196558607e-06, "loss": 1.2947, "step": 1155 }, { "epoch": 0.09, "learning_rate": 1.9813723552257715e-06, "loss": 1.2703, "step": 1156 }, { "epoch": 0.09, "learning_rate": 1.9813240287770734e-06, "loss": 1.239, "step": 1157 }, { "epoch": 0.09, "learning_rate": 1.98127564031282e-06, "loss": 1.2994, "step": 1158 }, { "epoch": 0.09, "learning_rate": 1.9812271898360694e-06, "loss": 1.3587, "step": 1159 }, { "epoch": 0.09, "learning_rate": 1.9811786773498833e-06, "loss": 1.2712, "step": 1160 }, { "epoch": 0.09, "learning_rate": 1.9811301028573275e-06, "loss": 1.2667, "step": 1161 }, { "epoch": 0.09, "learning_rate": 1.9810814663614716e-06, "loss": 1.2518, "step": 1162 }, { "epoch": 0.09, "learning_rate": 1.981032767865389e-06, "loss": 1.2841, "step": 1163 }, { "epoch": 0.09, "learning_rate": 1.9809840073721586e-06, "loss": 1.2584, "step": 1164 }, { "epoch": 0.09, "learning_rate": 1.9809351848848603e-06, "loss": 1.244, "step": 1165 }, { "epoch": 0.09, "learning_rate": 1.98088630040658e-06, "loss": 1.2425, "step": 1166 }, { "epoch": 0.09, "learning_rate": 1.980837353940407e-06, "loss": 1.2622, "step": 1167 }, { "epoch": 0.09, "learning_rate": 1.9807883454894344e-06, "loss": 1.3287, "step": 1168 }, { "epoch": 0.09, "learning_rate": 1.98073927505676e-06, "loss": 1.3233, "step": 1169 }, { "epoch": 0.09, "learning_rate": 1.980690142645484e-06, "loss": 1.2853, "step": 1170 }, { "epoch": 0.09, "learning_rate": 1.980640948258711e-06, "loss": 1.2794, "step": 1171 }, { "epoch": 0.09, "learning_rate": 1.9805916918995513e-06, "loss": 1.3066, "step": 1172 }, { "epoch": 0.09, "learning_rate": 1.980542373571117e-06, "loss": 0.8887, "step": 1173 }, { "epoch": 0.09, "learning_rate": 1.9804929932765243e-06, "loss": 1.3178, "step": 1174 }, { "epoch": 0.09, "learning_rate": 1.980443551018895e-06, "loss": 1.2549, "step": 1175 }, { "epoch": 0.09, "learning_rate": 1.9803940468013524e-06, "loss": 1.3033, "step": 1176 }, { "epoch": 0.09, "learning_rate": 1.9803444806270254e-06, "loss": 1.2584, "step": 1177 }, { "epoch": 0.09, "learning_rate": 1.9802948524990467e-06, "loss": 1.2337, "step": 1178 }, { "epoch": 0.09, "learning_rate": 1.980245162420552e-06, "loss": 1.2807, "step": 1179 }, { "epoch": 0.09, "learning_rate": 1.9801954103946824e-06, "loss": 1.237, "step": 1180 }, { "epoch": 0.09, "learning_rate": 1.980145596424581e-06, "loss": 1.2759, "step": 1181 }, { "epoch": 0.09, "learning_rate": 1.9800957205133965e-06, "loss": 1.2829, "step": 1182 }, { "epoch": 0.09, "learning_rate": 1.980045782664281e-06, "loss": 1.3203, "step": 1183 }, { "epoch": 0.09, "learning_rate": 1.9799957828803894e-06, "loss": 1.2893, "step": 1184 }, { "epoch": 0.09, "learning_rate": 1.979945721164882e-06, "loss": 1.2639, "step": 1185 }, { "epoch": 0.09, "learning_rate": 1.9798955975209235e-06, "loss": 1.2508, "step": 1186 }, { "epoch": 0.09, "learning_rate": 1.97984541195168e-06, "loss": 1.2429, "step": 1187 }, { "epoch": 0.09, "learning_rate": 1.9797951644603234e-06, "loss": 1.2922, "step": 1188 }, { "epoch": 0.09, "learning_rate": 1.9797448550500296e-06, "loss": 1.2486, "step": 1189 }, { "epoch": 0.09, "learning_rate": 1.9796944837239776e-06, "loss": 1.2157, "step": 1190 }, { "epoch": 0.09, "learning_rate": 1.9796440504853507e-06, "loss": 1.2825, "step": 1191 }, { "epoch": 0.09, "learning_rate": 1.979593555337336e-06, "loss": 0.8467, "step": 1192 }, { "epoch": 0.09, "learning_rate": 1.9795429982831248e-06, "loss": 1.3177, "step": 1193 }, { "epoch": 0.09, "learning_rate": 1.9794923793259123e-06, "loss": 1.2248, "step": 1194 }, { "epoch": 0.09, "learning_rate": 1.9794416984688966e-06, "loss": 1.2616, "step": 1195 }, { "epoch": 0.09, "learning_rate": 1.979390955715281e-06, "loss": 1.3012, "step": 1196 }, { "epoch": 0.09, "learning_rate": 1.9793401510682724e-06, "loss": 1.3148, "step": 1197 }, { "epoch": 0.09, "learning_rate": 1.9792892845310815e-06, "loss": 1.2859, "step": 1198 }, { "epoch": 0.09, "learning_rate": 1.9792383561069225e-06, "loss": 1.3039, "step": 1199 }, { "epoch": 0.09, "learning_rate": 1.9791873657990136e-06, "loss": 1.2583, "step": 1200 }, { "epoch": 0.09, "learning_rate": 1.979136313610578e-06, "loss": 1.2716, "step": 1201 }, { "epoch": 0.09, "learning_rate": 1.9790851995448413e-06, "loss": 1.2645, "step": 1202 }, { "epoch": 0.09, "learning_rate": 1.979034023605034e-06, "loss": 1.3465, "step": 1203 }, { "epoch": 0.09, "learning_rate": 1.97898278579439e-06, "loss": 1.2666, "step": 1204 }, { "epoch": 0.09, "learning_rate": 1.978931486116148e-06, "loss": 1.2997, "step": 1205 }, { "epoch": 0.09, "learning_rate": 1.9788801245735486e-06, "loss": 1.2157, "step": 1206 }, { "epoch": 0.09, "learning_rate": 1.978828701169839e-06, "loss": 0.871, "step": 1207 }, { "epoch": 0.09, "learning_rate": 1.978777215908268e-06, "loss": 1.2357, "step": 1208 }, { "epoch": 0.09, "learning_rate": 1.97872566879209e-06, "loss": 1.2302, "step": 1209 }, { "epoch": 0.09, "learning_rate": 1.9786740598245614e-06, "loss": 1.281, "step": 1210 }, { "epoch": 0.09, "learning_rate": 1.978622389008945e-06, "loss": 1.23, "step": 1211 }, { "epoch": 0.09, "learning_rate": 1.9785706563485056e-06, "loss": 1.2623, "step": 1212 }, { "epoch": 0.09, "learning_rate": 1.9785188618465123e-06, "loss": 1.2175, "step": 1213 }, { "epoch": 0.09, "learning_rate": 1.978467005506239e-06, "loss": 1.2779, "step": 1214 }, { "epoch": 0.09, "learning_rate": 1.9784150873309614e-06, "loss": 1.2838, "step": 1215 }, { "epoch": 0.09, "learning_rate": 1.9783631073239622e-06, "loss": 1.3069, "step": 1216 }, { "epoch": 0.09, "learning_rate": 1.9783110654885254e-06, "loss": 1.2791, "step": 1217 }, { "epoch": 0.09, "learning_rate": 1.97825896182794e-06, "loss": 1.2835, "step": 1218 }, { "epoch": 0.09, "learning_rate": 1.978206796345498e-06, "loss": 1.3009, "step": 1219 }, { "epoch": 0.09, "learning_rate": 1.978154569044497e-06, "loss": 1.2798, "step": 1220 }, { "epoch": 0.09, "learning_rate": 1.9781022799282376e-06, "loss": 1.2808, "step": 1221 }, { "epoch": 0.09, "learning_rate": 1.978049929000024e-06, "loss": 1.3, "step": 1222 }, { "epoch": 0.09, "learning_rate": 1.9779975162631645e-06, "loss": 1.2815, "step": 1223 }, { "epoch": 0.09, "learning_rate": 1.9779450417209707e-06, "loss": 1.301, "step": 1224 }, { "epoch": 0.1, "learning_rate": 1.9778925053767602e-06, "loss": 1.2806, "step": 1225 }, { "epoch": 0.1, "learning_rate": 1.9778399072338518e-06, "loss": 1.3239, "step": 1226 }, { "epoch": 0.1, "learning_rate": 1.97778724729557e-06, "loss": 1.3116, "step": 1227 }, { "epoch": 0.1, "learning_rate": 1.9777345255652428e-06, "loss": 0.8321, "step": 1228 }, { "epoch": 0.1, "learning_rate": 1.9776817420462016e-06, "loss": 1.2622, "step": 1229 }, { "epoch": 0.1, "learning_rate": 1.977628896741783e-06, "loss": 1.2633, "step": 1230 }, { "epoch": 0.1, "learning_rate": 1.977575989655325e-06, "loss": 1.2448, "step": 1231 }, { "epoch": 0.1, "learning_rate": 1.9775230207901727e-06, "loss": 1.3059, "step": 1232 }, { "epoch": 0.1, "learning_rate": 1.9774699901496727e-06, "loss": 1.2558, "step": 1233 }, { "epoch": 0.1, "learning_rate": 1.9774168977371767e-06, "loss": 1.2362, "step": 1234 }, { "epoch": 0.1, "learning_rate": 1.9773637435560392e-06, "loss": 1.2151, "step": 1235 }, { "epoch": 0.1, "learning_rate": 1.97731052760962e-06, "loss": 1.2776, "step": 1236 }, { "epoch": 0.1, "learning_rate": 1.9772572499012824e-06, "loss": 0.8769, "step": 1237 }, { "epoch": 0.1, "learning_rate": 1.9772039104343927e-06, "loss": 1.2314, "step": 1238 }, { "epoch": 0.1, "learning_rate": 1.9771505092123213e-06, "loss": 1.3129, "step": 1239 }, { "epoch": 0.1, "learning_rate": 1.977097046238444e-06, "loss": 1.2578, "step": 1240 }, { "epoch": 0.1, "learning_rate": 1.9770435215161388e-06, "loss": 1.3022, "step": 1241 }, { "epoch": 0.1, "learning_rate": 1.9769899350487886e-06, "loss": 1.3084, "step": 1242 }, { "epoch": 0.1, "learning_rate": 1.9769362868397796e-06, "loss": 1.2782, "step": 1243 }, { "epoch": 0.1, "learning_rate": 1.9768825768925025e-06, "loss": 1.2856, "step": 1244 }, { "epoch": 0.1, "learning_rate": 1.9768288052103505e-06, "loss": 1.2112, "step": 1245 }, { "epoch": 0.1, "learning_rate": 1.976774971796723e-06, "loss": 1.3793, "step": 1246 }, { "epoch": 0.1, "learning_rate": 1.9767210766550215e-06, "loss": 1.2432, "step": 1247 }, { "epoch": 0.1, "learning_rate": 1.976667119788652e-06, "loss": 0.8317, "step": 1248 }, { "epoch": 0.1, "learning_rate": 1.976613101201024e-06, "loss": 1.2512, "step": 1249 }, { "epoch": 0.1, "learning_rate": 1.9765590208955513e-06, "loss": 1.298, "step": 1250 }, { "epoch": 0.1, "learning_rate": 1.9765048788756522e-06, "loss": 1.2858, "step": 1251 }, { "epoch": 0.1, "learning_rate": 1.976450675144748e-06, "loss": 1.3085, "step": 1252 }, { "epoch": 0.1, "learning_rate": 1.976396409706264e-06, "loss": 1.2737, "step": 1253 }, { "epoch": 0.1, "learning_rate": 1.9763420825636293e-06, "loss": 1.2996, "step": 1254 }, { "epoch": 0.1, "learning_rate": 1.9762876937202776e-06, "loss": 1.2584, "step": 1255 }, { "epoch": 0.1, "learning_rate": 1.9762332431796458e-06, "loss": 1.2846, "step": 1256 }, { "epoch": 0.1, "learning_rate": 1.976178730945175e-06, "loss": 1.3227, "step": 1257 }, { "epoch": 0.1, "learning_rate": 1.97612415702031e-06, "loss": 1.2818, "step": 1258 }, { "epoch": 0.1, "learning_rate": 1.9760695214084997e-06, "loss": 1.2556, "step": 1259 }, { "epoch": 0.1, "learning_rate": 1.9760148241131973e-06, "loss": 1.2765, "step": 1260 }, { "epoch": 0.1, "learning_rate": 1.975960065137859e-06, "loss": 1.3295, "step": 1261 }, { "epoch": 0.1, "learning_rate": 1.975905244485945e-06, "loss": 1.3358, "step": 1262 }, { "epoch": 0.1, "learning_rate": 1.9758503621609203e-06, "loss": 1.2768, "step": 1263 }, { "epoch": 0.1, "learning_rate": 1.975795418166253e-06, "loss": 1.2539, "step": 1264 }, { "epoch": 0.1, "learning_rate": 1.9757404125054157e-06, "loss": 1.2928, "step": 1265 }, { "epoch": 0.1, "learning_rate": 1.975685345181884e-06, "loss": 1.2646, "step": 1266 }, { "epoch": 0.1, "learning_rate": 1.975630216199138e-06, "loss": 1.3204, "step": 1267 }, { "epoch": 0.1, "learning_rate": 1.975575025560662e-06, "loss": 1.3493, "step": 1268 }, { "epoch": 0.1, "learning_rate": 1.9755197732699435e-06, "loss": 1.2674, "step": 1269 }, { "epoch": 0.1, "learning_rate": 1.9754644593304745e-06, "loss": 1.3242, "step": 1270 }, { "epoch": 0.1, "learning_rate": 1.97540908374575e-06, "loss": 1.3179, "step": 1271 }, { "epoch": 0.1, "learning_rate": 1.97535364651927e-06, "loss": 1.2891, "step": 1272 }, { "epoch": 0.1, "learning_rate": 1.975298147654538e-06, "loss": 1.3244, "step": 1273 }, { "epoch": 0.1, "learning_rate": 1.975242587155061e-06, "loss": 1.2784, "step": 1274 }, { "epoch": 0.1, "learning_rate": 1.9751869650243503e-06, "loss": 1.2043, "step": 1275 }, { "epoch": 0.1, "learning_rate": 1.9751312812659207e-06, "loss": 1.2957, "step": 1276 }, { "epoch": 0.1, "learning_rate": 1.975075535883292e-06, "loss": 1.2731, "step": 1277 }, { "epoch": 0.1, "learning_rate": 1.975019728879986e-06, "loss": 1.264, "step": 1278 }, { "epoch": 0.1, "learning_rate": 1.9749638602595304e-06, "loss": 1.2737, "step": 1279 }, { "epoch": 0.1, "learning_rate": 1.9749079300254556e-06, "loss": 1.2744, "step": 1280 }, { "epoch": 0.1, "learning_rate": 1.9748519381812954e-06, "loss": 1.2049, "step": 1281 }, { "epoch": 0.1, "learning_rate": 1.9747958847305892e-06, "loss": 1.2557, "step": 1282 }, { "epoch": 0.1, "learning_rate": 1.974739769676879e-06, "loss": 1.3007, "step": 1283 }, { "epoch": 0.1, "learning_rate": 1.974683593023711e-06, "loss": 1.2985, "step": 1284 }, { "epoch": 0.1, "learning_rate": 1.9746273547746354e-06, "loss": 1.2644, "step": 1285 }, { "epoch": 0.1, "learning_rate": 1.974571054933206e-06, "loss": 1.2636, "step": 1286 }, { "epoch": 0.1, "learning_rate": 1.9745146935029814e-06, "loss": 1.2517, "step": 1287 }, { "epoch": 0.1, "learning_rate": 1.9744582704875224e-06, "loss": 1.263, "step": 1288 }, { "epoch": 0.1, "learning_rate": 1.9744017858903952e-06, "loss": 1.3133, "step": 1289 }, { "epoch": 0.1, "learning_rate": 1.97434523971517e-06, "loss": 1.3086, "step": 1290 }, { "epoch": 0.1, "learning_rate": 1.974288631965419e-06, "loss": 1.2628, "step": 1291 }, { "epoch": 0.1, "learning_rate": 1.9742319626447206e-06, "loss": 1.3238, "step": 1292 }, { "epoch": 0.1, "learning_rate": 1.9741752317566555e-06, "loss": 1.2915, "step": 1293 }, { "epoch": 0.1, "learning_rate": 1.974118439304809e-06, "loss": 1.2907, "step": 1294 }, { "epoch": 0.1, "learning_rate": 1.9740615852927708e-06, "loss": 1.315, "step": 1295 }, { "epoch": 0.1, "learning_rate": 1.9740046697241327e-06, "loss": 1.3041, "step": 1296 }, { "epoch": 0.1, "learning_rate": 1.973947692602492e-06, "loss": 0.915, "step": 1297 }, { "epoch": 0.1, "learning_rate": 1.97389065393145e-06, "loss": 1.2982, "step": 1298 }, { "epoch": 0.1, "learning_rate": 1.9738335537146105e-06, "loss": 1.2563, "step": 1299 }, { "epoch": 0.1, "learning_rate": 1.973776391955582e-06, "loss": 1.2645, "step": 1300 }, { "epoch": 0.1, "learning_rate": 1.9737191686579776e-06, "loss": 1.276, "step": 1301 }, { "epoch": 0.1, "learning_rate": 1.973661883825413e-06, "loss": 1.321, "step": 1302 }, { "epoch": 0.1, "learning_rate": 1.9736045374615083e-06, "loss": 1.283, "step": 1303 }, { "epoch": 0.1, "learning_rate": 1.973547129569888e-06, "loss": 1.2925, "step": 1304 }, { "epoch": 0.1, "learning_rate": 1.9734896601541795e-06, "loss": 1.2502, "step": 1305 }, { "epoch": 0.1, "learning_rate": 1.9734321292180153e-06, "loss": 1.3344, "step": 1306 }, { "epoch": 0.1, "learning_rate": 1.9733745367650304e-06, "loss": 1.308, "step": 1307 }, { "epoch": 0.1, "learning_rate": 1.973316882798865e-06, "loss": 1.2612, "step": 1308 }, { "epoch": 0.1, "learning_rate": 1.973259167323162e-06, "loss": 1.2747, "step": 1309 }, { "epoch": 0.1, "learning_rate": 1.973201390341569e-06, "loss": 0.8794, "step": 1310 }, { "epoch": 0.1, "learning_rate": 1.973143551857738e-06, "loss": 1.2912, "step": 1311 }, { "epoch": 0.1, "learning_rate": 1.973085651875323e-06, "loss": 1.2969, "step": 1312 }, { "epoch": 0.1, "learning_rate": 1.973027690397984e-06, "loss": 1.267, "step": 1313 }, { "epoch": 0.1, "learning_rate": 1.9729696674293828e-06, "loss": 1.2151, "step": 1314 }, { "epoch": 0.1, "learning_rate": 1.9729115829731867e-06, "loss": 1.2612, "step": 1315 }, { "epoch": 0.1, "learning_rate": 1.972853437033067e-06, "loss": 0.826, "step": 1316 }, { "epoch": 0.1, "learning_rate": 1.9727952296126983e-06, "loss": 1.2846, "step": 1317 }, { "epoch": 0.1, "learning_rate": 1.9727369607157577e-06, "loss": 1.2548, "step": 1318 }, { "epoch": 0.1, "learning_rate": 1.972678630345929e-06, "loss": 1.2966, "step": 1319 }, { "epoch": 0.1, "learning_rate": 1.972620238506898e-06, "loss": 0.8734, "step": 1320 }, { "epoch": 0.1, "learning_rate": 1.972561785202354e-06, "loss": 1.2744, "step": 1321 }, { "epoch": 0.1, "learning_rate": 1.9725032704359924e-06, "loss": 0.89, "step": 1322 }, { "epoch": 0.1, "learning_rate": 1.97244469421151e-06, "loss": 1.2396, "step": 1323 }, { "epoch": 0.1, "learning_rate": 1.9723860565326094e-06, "loss": 1.286, "step": 1324 }, { "epoch": 0.1, "learning_rate": 1.9723273574029956e-06, "loss": 1.2853, "step": 1325 }, { "epoch": 0.1, "learning_rate": 1.9722685968263782e-06, "loss": 1.269, "step": 1326 }, { "epoch": 0.1, "learning_rate": 1.972209774806471e-06, "loss": 0.8529, "step": 1327 }, { "epoch": 0.1, "learning_rate": 1.972150891346991e-06, "loss": 1.2501, "step": 1328 }, { "epoch": 0.1, "learning_rate": 1.9720919464516595e-06, "loss": 1.2843, "step": 1329 }, { "epoch": 0.1, "learning_rate": 1.9720329401242017e-06, "loss": 1.3282, "step": 1330 }, { "epoch": 0.1, "learning_rate": 1.9719738723683463e-06, "loss": 1.3373, "step": 1331 }, { "epoch": 0.1, "learning_rate": 1.9719147431878257e-06, "loss": 1.272, "step": 1332 }, { "epoch": 0.1, "learning_rate": 1.9718555525863774e-06, "loss": 1.2926, "step": 1333 }, { "epoch": 0.1, "learning_rate": 1.971796300567742e-06, "loss": 1.2321, "step": 1334 }, { "epoch": 0.1, "learning_rate": 1.971736987135664e-06, "loss": 1.2448, "step": 1335 }, { "epoch": 0.1, "learning_rate": 1.971677612293891e-06, "loss": 1.2862, "step": 1336 }, { "epoch": 0.1, "learning_rate": 1.9716181760461758e-06, "loss": 1.2843, "step": 1337 }, { "epoch": 0.1, "learning_rate": 1.9715586783962743e-06, "loss": 0.9134, "step": 1338 }, { "epoch": 0.1, "learning_rate": 1.971499119347947e-06, "loss": 1.306, "step": 1339 }, { "epoch": 0.1, "learning_rate": 1.971439498904957e-06, "loss": 1.2681, "step": 1340 }, { "epoch": 0.1, "learning_rate": 1.971379817071073e-06, "loss": 1.2644, "step": 1341 }, { "epoch": 0.1, "learning_rate": 1.971320073850066e-06, "loss": 0.8803, "step": 1342 }, { "epoch": 0.1, "learning_rate": 1.9712602692457114e-06, "loss": 0.9511, "step": 1343 }, { "epoch": 0.1, "learning_rate": 1.971200403261789e-06, "loss": 1.2809, "step": 1344 }, { "epoch": 0.1, "learning_rate": 1.9711404759020817e-06, "loss": 1.2935, "step": 1345 }, { "epoch": 0.1, "learning_rate": 1.971080487170377e-06, "loss": 1.2299, "step": 1346 }, { "epoch": 0.1, "learning_rate": 1.971020437070466e-06, "loss": 1.3316, "step": 1347 }, { "epoch": 0.1, "learning_rate": 1.9709603256061434e-06, "loss": 1.3276, "step": 1348 }, { "epoch": 0.1, "learning_rate": 1.970900152781208e-06, "loss": 1.2707, "step": 1349 }, { "epoch": 0.1, "learning_rate": 1.9708399185994626e-06, "loss": 1.287, "step": 1350 }, { "epoch": 0.1, "learning_rate": 1.970779623064713e-06, "loss": 1.2772, "step": 1351 }, { "epoch": 0.1, "learning_rate": 1.970719266180771e-06, "loss": 1.3425, "step": 1352 }, { "epoch": 0.11, "learning_rate": 1.9706588479514496e-06, "loss": 1.2865, "step": 1353 }, { "epoch": 0.11, "learning_rate": 1.970598368380568e-06, "loss": 1.2319, "step": 1354 }, { "epoch": 0.11, "learning_rate": 1.9705378274719474e-06, "loss": 1.3049, "step": 1355 }, { "epoch": 0.11, "learning_rate": 1.9704772252294145e-06, "loss": 0.8674, "step": 1356 }, { "epoch": 0.11, "learning_rate": 1.9704165616567987e-06, "loss": 1.3021, "step": 1357 }, { "epoch": 0.11, "learning_rate": 1.9703558367579337e-06, "loss": 0.8512, "step": 1358 }, { "epoch": 0.11, "learning_rate": 1.9702950505366573e-06, "loss": 1.2593, "step": 1359 }, { "epoch": 0.11, "learning_rate": 1.9702342029968104e-06, "loss": 1.3026, "step": 1360 }, { "epoch": 0.11, "learning_rate": 1.9701732941422385e-06, "loss": 1.3281, "step": 1361 }, { "epoch": 0.11, "learning_rate": 1.970112323976791e-06, "loss": 1.247, "step": 1362 }, { "epoch": 0.11, "learning_rate": 1.9700512925043208e-06, "loss": 1.3034, "step": 1363 }, { "epoch": 0.11, "learning_rate": 1.969990199728685e-06, "loss": 1.2637, "step": 1364 }, { "epoch": 0.11, "learning_rate": 1.9699290456537445e-06, "loss": 1.3466, "step": 1365 }, { "epoch": 0.11, "learning_rate": 1.9698678302833634e-06, "loss": 1.346, "step": 1366 }, { "epoch": 0.11, "learning_rate": 1.969806553621411e-06, "loss": 1.2221, "step": 1367 }, { "epoch": 0.11, "learning_rate": 1.969745215671759e-06, "loss": 1.2947, "step": 1368 }, { "epoch": 0.11, "learning_rate": 1.9696838164382846e-06, "loss": 1.3437, "step": 1369 }, { "epoch": 0.11, "learning_rate": 1.969622355924867e-06, "loss": 1.284, "step": 1370 }, { "epoch": 0.11, "learning_rate": 1.9695608341353908e-06, "loss": 0.9391, "step": 1371 }, { "epoch": 0.11, "learning_rate": 1.969499251073744e-06, "loss": 1.3274, "step": 1372 }, { "epoch": 0.11, "learning_rate": 1.9694376067438177e-06, "loss": 1.2798, "step": 1373 }, { "epoch": 0.11, "learning_rate": 1.9693759011495084e-06, "loss": 1.3171, "step": 1374 }, { "epoch": 0.11, "learning_rate": 1.9693141342947155e-06, "loss": 1.2852, "step": 1375 }, { "epoch": 0.11, "learning_rate": 1.969252306183342e-06, "loss": 1.2753, "step": 1376 }, { "epoch": 0.11, "learning_rate": 1.9691904168192953e-06, "loss": 1.2483, "step": 1377 }, { "epoch": 0.11, "learning_rate": 1.9691284662064868e-06, "loss": 0.8457, "step": 1378 }, { "epoch": 0.11, "learning_rate": 1.969066454348831e-06, "loss": 0.8592, "step": 1379 }, { "epoch": 0.11, "learning_rate": 1.9690043812502472e-06, "loss": 1.2644, "step": 1380 }, { "epoch": 0.11, "learning_rate": 1.9689422469146584e-06, "loss": 1.2549, "step": 1381 }, { "epoch": 0.11, "learning_rate": 1.9688800513459906e-06, "loss": 1.3011, "step": 1382 }, { "epoch": 0.11, "learning_rate": 1.968817794548175e-06, "loss": 1.2409, "step": 1383 }, { "epoch": 0.11, "learning_rate": 1.968755476525145e-06, "loss": 1.2567, "step": 1384 }, { "epoch": 0.11, "learning_rate": 1.96869309728084e-06, "loss": 1.2862, "step": 1385 }, { "epoch": 0.11, "learning_rate": 1.9686306568192015e-06, "loss": 1.2579, "step": 1386 }, { "epoch": 0.11, "learning_rate": 1.968568155144175e-06, "loss": 0.8925, "step": 1387 }, { "epoch": 0.11, "learning_rate": 1.9685055922597116e-06, "loss": 1.2996, "step": 1388 }, { "epoch": 0.11, "learning_rate": 1.9684429681697633e-06, "loss": 1.1838, "step": 1389 }, { "epoch": 0.11, "learning_rate": 1.9683802828782892e-06, "loss": 1.2469, "step": 1390 }, { "epoch": 0.11, "learning_rate": 1.96831753638925e-06, "loss": 1.2707, "step": 1391 }, { "epoch": 0.11, "learning_rate": 1.9682547287066114e-06, "loss": 1.2063, "step": 1392 }, { "epoch": 0.11, "learning_rate": 1.968191859834342e-06, "loss": 1.263, "step": 1393 }, { "epoch": 0.11, "learning_rate": 1.968128929776416e-06, "loss": 1.2437, "step": 1394 }, { "epoch": 0.11, "learning_rate": 1.968065938536809e-06, "loss": 0.8769, "step": 1395 }, { "epoch": 0.11, "learning_rate": 1.9680028861195025e-06, "loss": 1.2812, "step": 1396 }, { "epoch": 0.11, "learning_rate": 1.967939772528481e-06, "loss": 1.2642, "step": 1397 }, { "epoch": 0.11, "learning_rate": 1.9678765977677324e-06, "loss": 1.3259, "step": 1398 }, { "epoch": 0.11, "learning_rate": 1.9678133618412504e-06, "loss": 1.3037, "step": 1399 }, { "epoch": 0.11, "learning_rate": 1.9677500647530302e-06, "loss": 1.2523, "step": 1400 }, { "epoch": 0.11, "learning_rate": 1.9676867065070725e-06, "loss": 1.2975, "step": 1401 }, { "epoch": 0.11, "learning_rate": 1.9676232871073804e-06, "loss": 0.825, "step": 1402 }, { "epoch": 0.11, "learning_rate": 1.9675598065579627e-06, "loss": 0.8736, "step": 1403 }, { "epoch": 0.11, "learning_rate": 1.9674962648628308e-06, "loss": 1.2779, "step": 1404 }, { "epoch": 0.11, "learning_rate": 1.967432662026e-06, "loss": 1.2771, "step": 1405 }, { "epoch": 0.11, "learning_rate": 1.9673689980514903e-06, "loss": 1.316, "step": 1406 }, { "epoch": 0.11, "learning_rate": 1.9673052729433243e-06, "loss": 1.3556, "step": 1407 }, { "epoch": 0.11, "learning_rate": 1.9672414867055297e-06, "loss": 1.2631, "step": 1408 }, { "epoch": 0.11, "learning_rate": 1.9671776393421372e-06, "loss": 1.3093, "step": 1409 }, { "epoch": 0.11, "learning_rate": 1.967113730857182e-06, "loss": 1.2748, "step": 1410 }, { "epoch": 0.11, "learning_rate": 1.9670497612547028e-06, "loss": 1.2718, "step": 1411 }, { "epoch": 0.11, "learning_rate": 1.9669857305387414e-06, "loss": 1.2586, "step": 1412 }, { "epoch": 0.11, "learning_rate": 1.9669216387133454e-06, "loss": 1.2573, "step": 1413 }, { "epoch": 0.11, "learning_rate": 1.9668574857825644e-06, "loss": 1.3114, "step": 1414 }, { "epoch": 0.11, "learning_rate": 1.966793271750453e-06, "loss": 0.8322, "step": 1415 }, { "epoch": 0.11, "learning_rate": 1.966728996621069e-06, "loss": 1.3232, "step": 1416 }, { "epoch": 0.11, "learning_rate": 1.9666646603984747e-06, "loss": 1.2408, "step": 1417 }, { "epoch": 0.11, "learning_rate": 1.9666002630867355e-06, "loss": 1.3302, "step": 1418 }, { "epoch": 0.11, "learning_rate": 1.966535804689921e-06, "loss": 1.2853, "step": 1419 }, { "epoch": 0.11, "learning_rate": 1.966471285212105e-06, "loss": 1.2412, "step": 1420 }, { "epoch": 0.11, "learning_rate": 1.966406704657365e-06, "loss": 1.2645, "step": 1421 }, { "epoch": 0.11, "learning_rate": 1.966342063029781e-06, "loss": 1.2679, "step": 1422 }, { "epoch": 0.11, "learning_rate": 1.96627736033344e-06, "loss": 1.2915, "step": 1423 }, { "epoch": 0.11, "learning_rate": 1.96621259657243e-06, "loss": 1.2485, "step": 1424 }, { "epoch": 0.11, "learning_rate": 1.966147771750843e-06, "loss": 1.2605, "step": 1425 }, { "epoch": 0.11, "learning_rate": 1.966082885872777e-06, "loss": 1.2644, "step": 1426 }, { "epoch": 0.11, "learning_rate": 1.9660179389423317e-06, "loss": 1.2541, "step": 1427 }, { "epoch": 0.11, "learning_rate": 1.9659529309636114e-06, "loss": 1.3064, "step": 1428 }, { "epoch": 0.11, "learning_rate": 1.965887861940725e-06, "loss": 0.8402, "step": 1429 }, { "epoch": 0.11, "learning_rate": 1.9658227318777843e-06, "loss": 1.2508, "step": 1430 }, { "epoch": 0.11, "learning_rate": 1.965757540778905e-06, "loss": 1.2723, "step": 1431 }, { "epoch": 0.11, "learning_rate": 1.965692288648207e-06, "loss": 1.218, "step": 1432 }, { "epoch": 0.11, "learning_rate": 1.965626975489814e-06, "loss": 1.27, "step": 1433 }, { "epoch": 0.11, "learning_rate": 1.965561601307854e-06, "loss": 1.2505, "step": 1434 }, { "epoch": 0.11, "learning_rate": 1.965496166106457e-06, "loss": 1.2843, "step": 1435 }, { "epoch": 0.11, "learning_rate": 1.96543066988976e-06, "loss": 1.2837, "step": 1436 }, { "epoch": 0.11, "learning_rate": 1.9653651126619007e-06, "loss": 1.2545, "step": 1437 }, { "epoch": 0.11, "learning_rate": 1.965299494427023e-06, "loss": 1.3046, "step": 1438 }, { "epoch": 0.11, "learning_rate": 1.9652338151892727e-06, "loss": 1.2975, "step": 1439 }, { "epoch": 0.11, "learning_rate": 1.9651680749528014e-06, "loss": 1.2897, "step": 1440 }, { "epoch": 0.11, "learning_rate": 1.9651022737217633e-06, "loss": 0.8974, "step": 1441 }, { "epoch": 0.11, "learning_rate": 1.965036411500317e-06, "loss": 1.2848, "step": 1442 }, { "epoch": 0.11, "learning_rate": 1.9649704882926235e-06, "loss": 1.2863, "step": 1443 }, { "epoch": 0.11, "learning_rate": 1.9649045041028504e-06, "loss": 1.2878, "step": 1444 }, { "epoch": 0.11, "learning_rate": 1.964838458935167e-06, "loss": 0.8603, "step": 1445 }, { "epoch": 0.11, "learning_rate": 1.964772352793747e-06, "loss": 1.228, "step": 1446 }, { "epoch": 0.11, "learning_rate": 1.9647061856827677e-06, "loss": 0.8175, "step": 1447 }, { "epoch": 0.11, "learning_rate": 1.9646399576064114e-06, "loss": 0.8911, "step": 1448 }, { "epoch": 0.11, "learning_rate": 1.964573668568863e-06, "loss": 1.2792, "step": 1449 }, { "epoch": 0.11, "learning_rate": 1.964507318574312e-06, "loss": 1.308, "step": 1450 }, { "epoch": 0.11, "learning_rate": 1.9644409076269505e-06, "loss": 1.2641, "step": 1451 }, { "epoch": 0.11, "learning_rate": 1.964374435730976e-06, "loss": 1.259, "step": 1452 }, { "epoch": 0.11, "learning_rate": 1.9643079028905904e-06, "loss": 1.2236, "step": 1453 }, { "epoch": 0.11, "learning_rate": 1.964241309109996e-06, "loss": 1.2619, "step": 1454 }, { "epoch": 0.11, "learning_rate": 1.964174654393403e-06, "loss": 1.2888, "step": 1455 }, { "epoch": 0.11, "learning_rate": 1.9641079387450232e-06, "loss": 1.2716, "step": 1456 }, { "epoch": 0.11, "learning_rate": 1.9640411621690726e-06, "loss": 1.2294, "step": 1457 }, { "epoch": 0.11, "learning_rate": 1.963974324669771e-06, "loss": 1.3197, "step": 1458 }, { "epoch": 0.11, "learning_rate": 1.963907426251343e-06, "loss": 1.3295, "step": 1459 }, { "epoch": 0.11, "learning_rate": 1.9638404669180154e-06, "loss": 1.1899, "step": 1460 }, { "epoch": 0.11, "learning_rate": 1.9637734466740203e-06, "loss": 1.2508, "step": 1461 }, { "epoch": 0.11, "learning_rate": 1.963706365523593e-06, "loss": 1.2823, "step": 1462 }, { "epoch": 0.11, "learning_rate": 1.963639223470973e-06, "loss": 1.2645, "step": 1463 }, { "epoch": 0.11, "learning_rate": 1.9635720205204025e-06, "loss": 1.2815, "step": 1464 }, { "epoch": 0.11, "learning_rate": 1.9635047566761295e-06, "loss": 1.2881, "step": 1465 }, { "epoch": 0.11, "learning_rate": 1.9634374319424043e-06, "loss": 1.2412, "step": 1466 }, { "epoch": 0.11, "learning_rate": 1.9633700463234816e-06, "loss": 1.3035, "step": 1467 }, { "epoch": 0.11, "learning_rate": 1.96330259982362e-06, "loss": 1.317, "step": 1468 }, { "epoch": 0.11, "learning_rate": 1.9632350924470813e-06, "loss": 1.3077, "step": 1469 }, { "epoch": 0.11, "learning_rate": 1.963167524198132e-06, "loss": 1.2777, "step": 1470 }, { "epoch": 0.11, "learning_rate": 1.9630998950810433e-06, "loss": 1.3216, "step": 1471 }, { "epoch": 0.11, "learning_rate": 1.963032205100087e-06, "loss": 1.3123, "step": 1472 }, { "epoch": 0.11, "learning_rate": 1.9629644542595424e-06, "loss": 0.8498, "step": 1473 }, { "epoch": 0.11, "learning_rate": 1.96289664256369e-06, "loss": 0.8898, "step": 1474 }, { "epoch": 0.11, "learning_rate": 1.962828770016816e-06, "loss": 1.3054, "step": 1475 }, { "epoch": 0.11, "learning_rate": 1.962760836623209e-06, "loss": 1.2396, "step": 1476 }, { "epoch": 0.11, "learning_rate": 1.962692842387163e-06, "loss": 1.219, "step": 1477 }, { "epoch": 0.11, "learning_rate": 1.9626247873129743e-06, "loss": 1.2601, "step": 1478 }, { "epoch": 0.11, "learning_rate": 1.9625566714049437e-06, "loss": 1.3365, "step": 1479 }, { "epoch": 0.11, "learning_rate": 1.962488494667376e-06, "loss": 1.2864, "step": 1480 }, { "epoch": 0.11, "learning_rate": 1.9624202571045795e-06, "loss": 1.2492, "step": 1481 }, { "epoch": 0.12, "learning_rate": 1.9623519587208666e-06, "loss": 1.3054, "step": 1482 }, { "epoch": 0.12, "learning_rate": 1.9622835995205536e-06, "loss": 1.229, "step": 1483 }, { "epoch": 0.12, "learning_rate": 1.9622151795079607e-06, "loss": 1.2634, "step": 1484 }, { "epoch": 0.12, "learning_rate": 1.9621466986874117e-06, "loss": 1.2732, "step": 1485 }, { "epoch": 0.12, "learning_rate": 1.9620781570632336e-06, "loss": 1.2903, "step": 1486 }, { "epoch": 0.12, "learning_rate": 1.9620095546397586e-06, "loss": 1.2482, "step": 1487 }, { "epoch": 0.12, "learning_rate": 1.9619408914213222e-06, "loss": 1.2818, "step": 1488 }, { "epoch": 0.12, "learning_rate": 1.961872167412263e-06, "loss": 1.2644, "step": 1489 }, { "epoch": 0.12, "learning_rate": 1.961803382616925e-06, "loss": 1.2796, "step": 1490 }, { "epoch": 0.12, "learning_rate": 1.961734537039654e-06, "loss": 0.8964, "step": 1491 }, { "epoch": 0.12, "learning_rate": 1.9616656306848014e-06, "loss": 1.2878, "step": 1492 }, { "epoch": 0.12, "learning_rate": 1.9615966635567216e-06, "loss": 1.3389, "step": 1493 }, { "epoch": 0.12, "learning_rate": 1.9615276356597736e-06, "loss": 0.8647, "step": 1494 }, { "epoch": 0.12, "learning_rate": 1.9614585469983185e-06, "loss": 1.2767, "step": 1495 }, { "epoch": 0.12, "learning_rate": 1.9613893975767235e-06, "loss": 1.276, "step": 1496 }, { "epoch": 0.12, "learning_rate": 1.9613201873993585e-06, "loss": 1.2122, "step": 1497 }, { "epoch": 0.12, "learning_rate": 1.9612509164705963e-06, "loss": 1.2949, "step": 1498 }, { "epoch": 0.12, "learning_rate": 1.9611815847948156e-06, "loss": 1.2943, "step": 1499 }, { "epoch": 0.12, "learning_rate": 1.9611121923763972e-06, "loss": 1.23, "step": 1500 }, { "epoch": 0.12, "learning_rate": 1.9610427392197268e-06, "loss": 1.2943, "step": 1501 }, { "epoch": 0.12, "learning_rate": 1.9609732253291935e-06, "loss": 1.2895, "step": 1502 }, { "epoch": 0.12, "learning_rate": 1.9609036507091902e-06, "loss": 1.2733, "step": 1503 }, { "epoch": 0.12, "learning_rate": 1.960834015364114e-06, "loss": 1.2698, "step": 1504 }, { "epoch": 0.12, "learning_rate": 1.960764319298364e-06, "loss": 1.2872, "step": 1505 }, { "epoch": 0.12, "learning_rate": 1.960694562516347e-06, "loss": 1.2597, "step": 1506 }, { "epoch": 0.12, "learning_rate": 1.9606247450224703e-06, "loss": 1.3222, "step": 1507 }, { "epoch": 0.12, "learning_rate": 1.960554866821146e-06, "loss": 0.9035, "step": 1508 }, { "epoch": 0.12, "learning_rate": 1.9604849279167905e-06, "loss": 1.3099, "step": 1509 }, { "epoch": 0.12, "learning_rate": 1.960414928313823e-06, "loss": 1.2859, "step": 1510 }, { "epoch": 0.12, "learning_rate": 1.9603448680166673e-06, "loss": 0.8611, "step": 1511 }, { "epoch": 0.12, "learning_rate": 1.9602747470297517e-06, "loss": 1.2935, "step": 1512 }, { "epoch": 0.12, "learning_rate": 1.9602045653575066e-06, "loss": 1.3184, "step": 1513 }, { "epoch": 0.12, "learning_rate": 1.9601343230043675e-06, "loss": 1.1647, "step": 1514 }, { "epoch": 0.12, "learning_rate": 1.9600640199747737e-06, "loss": 1.328, "step": 1515 }, { "epoch": 0.12, "learning_rate": 1.959993656273168e-06, "loss": 1.2685, "step": 1516 }, { "epoch": 0.12, "learning_rate": 1.9599232319039965e-06, "loss": 1.2936, "step": 1517 }, { "epoch": 0.12, "learning_rate": 1.959852746871711e-06, "loss": 1.308, "step": 1518 }, { "epoch": 0.12, "learning_rate": 1.959782201180764e-06, "loss": 1.3448, "step": 1519 }, { "epoch": 0.12, "learning_rate": 1.9597115948356153e-06, "loss": 1.3418, "step": 1520 }, { "epoch": 0.12, "learning_rate": 1.959640927840726e-06, "loss": 1.2219, "step": 1521 }, { "epoch": 0.12, "learning_rate": 1.9595702002005627e-06, "loss": 1.2581, "step": 1522 }, { "epoch": 0.12, "learning_rate": 1.9594994119195946e-06, "loss": 1.2898, "step": 1523 }, { "epoch": 0.12, "learning_rate": 1.9594285630022952e-06, "loss": 1.3092, "step": 1524 }, { "epoch": 0.12, "learning_rate": 1.959357653453142e-06, "loss": 1.2623, "step": 1525 }, { "epoch": 0.12, "learning_rate": 1.959286683276616e-06, "loss": 1.2238, "step": 1526 }, { "epoch": 0.12, "learning_rate": 1.959215652477203e-06, "loss": 1.2144, "step": 1527 }, { "epoch": 0.12, "learning_rate": 1.9591445610593908e-06, "loss": 1.3131, "step": 1528 }, { "epoch": 0.12, "learning_rate": 1.959073409027672e-06, "loss": 1.2182, "step": 1529 }, { "epoch": 0.12, "learning_rate": 1.959002196386544e-06, "loss": 1.2356, "step": 1530 }, { "epoch": 0.12, "learning_rate": 1.958930923140507e-06, "loss": 1.2885, "step": 1531 }, { "epoch": 0.12, "learning_rate": 1.9588595892940647e-06, "loss": 1.2713, "step": 1532 }, { "epoch": 0.12, "learning_rate": 1.9587881948517254e-06, "loss": 1.2911, "step": 1533 }, { "epoch": 0.12, "learning_rate": 1.9587167398180007e-06, "loss": 1.2651, "step": 1534 }, { "epoch": 0.12, "learning_rate": 1.9586452241974064e-06, "loss": 1.3212, "step": 1535 }, { "epoch": 0.12, "learning_rate": 1.958573647994462e-06, "loss": 1.2189, "step": 1536 }, { "epoch": 0.12, "learning_rate": 1.958502011213691e-06, "loss": 1.2889, "step": 1537 }, { "epoch": 0.12, "learning_rate": 1.95843031385962e-06, "loss": 1.2698, "step": 1538 }, { "epoch": 0.12, "learning_rate": 1.9583585559367806e-06, "loss": 1.2587, "step": 1539 }, { "epoch": 0.12, "learning_rate": 1.9582867374497076e-06, "loss": 1.2751, "step": 1540 }, { "epoch": 0.12, "learning_rate": 1.9582148584029387e-06, "loss": 0.8684, "step": 1541 }, { "epoch": 0.12, "learning_rate": 1.9581429188010174e-06, "loss": 1.2072, "step": 1542 }, { "epoch": 0.12, "learning_rate": 1.9580709186484894e-06, "loss": 1.2477, "step": 1543 }, { "epoch": 0.12, "learning_rate": 1.9579988579499052e-06, "loss": 1.2113, "step": 1544 }, { "epoch": 0.12, "learning_rate": 1.957926736709818e-06, "loss": 1.291, "step": 1545 }, { "epoch": 0.12, "learning_rate": 1.9578545549327866e-06, "loss": 1.2911, "step": 1546 }, { "epoch": 0.12, "learning_rate": 1.9577823126233717e-06, "loss": 1.2673, "step": 1547 }, { "epoch": 0.12, "learning_rate": 1.9577100097861394e-06, "loss": 1.2715, "step": 1548 }, { "epoch": 0.12, "learning_rate": 1.9576376464256587e-06, "loss": 1.3088, "step": 1549 }, { "epoch": 0.12, "learning_rate": 1.9575652225465022e-06, "loss": 0.8977, "step": 1550 }, { "epoch": 0.12, "learning_rate": 1.9574927381532478e-06, "loss": 1.3121, "step": 1551 }, { "epoch": 0.12, "learning_rate": 1.9574201932504744e-06, "loss": 0.885, "step": 1552 }, { "epoch": 0.12, "learning_rate": 1.9573475878427684e-06, "loss": 1.2638, "step": 1553 }, { "epoch": 0.12, "learning_rate": 1.9572749219347177e-06, "loss": 1.2495, "step": 1554 }, { "epoch": 0.12, "learning_rate": 1.957202195530914e-06, "loss": 1.3474, "step": 1555 }, { "epoch": 0.12, "learning_rate": 1.9571294086359534e-06, "loss": 1.3197, "step": 1556 }, { "epoch": 0.12, "learning_rate": 1.957056561254436e-06, "loss": 1.2462, "step": 1557 }, { "epoch": 0.12, "learning_rate": 1.9569836533909654e-06, "loss": 1.2947, "step": 1558 }, { "epoch": 0.12, "learning_rate": 1.9569106850501485e-06, "loss": 1.2452, "step": 1559 }, { "epoch": 0.12, "learning_rate": 1.9568376562365978e-06, "loss": 1.2217, "step": 1560 }, { "epoch": 0.12, "learning_rate": 1.956764566954927e-06, "loss": 1.2919, "step": 1561 }, { "epoch": 0.12, "learning_rate": 1.956691417209756e-06, "loss": 1.2625, "step": 1562 }, { "epoch": 0.12, "learning_rate": 1.9566182070057074e-06, "loss": 1.2699, "step": 1563 }, { "epoch": 0.12, "learning_rate": 1.956544936347408e-06, "loss": 1.2656, "step": 1564 }, { "epoch": 0.12, "learning_rate": 1.956471605239487e-06, "loss": 1.2208, "step": 1565 }, { "epoch": 0.12, "learning_rate": 1.95639821368658e-06, "loss": 1.3065, "step": 1566 }, { "epoch": 0.12, "learning_rate": 1.9563247616933246e-06, "loss": 1.3148, "step": 1567 }, { "epoch": 0.12, "learning_rate": 1.9562512492643623e-06, "loss": 1.2581, "step": 1568 }, { "epoch": 0.12, "learning_rate": 1.956177676404339e-06, "loss": 0.8551, "step": 1569 }, { "epoch": 0.12, "learning_rate": 1.9561040431179046e-06, "loss": 1.2952, "step": 1570 }, { "epoch": 0.12, "learning_rate": 1.956030349409712e-06, "loss": 1.2739, "step": 1571 }, { "epoch": 0.12, "learning_rate": 1.9559565952844178e-06, "loss": 1.2882, "step": 1572 }, { "epoch": 0.12, "learning_rate": 1.955882780746684e-06, "loss": 0.8459, "step": 1573 }, { "epoch": 0.12, "learning_rate": 1.955808905801175e-06, "loss": 1.3008, "step": 1574 }, { "epoch": 0.12, "learning_rate": 1.9557349704525593e-06, "loss": 1.2547, "step": 1575 }, { "epoch": 0.12, "learning_rate": 1.9556609747055093e-06, "loss": 1.3142, "step": 1576 }, { "epoch": 0.12, "learning_rate": 1.9555869185647005e-06, "loss": 1.2794, "step": 1577 }, { "epoch": 0.12, "learning_rate": 1.9555128020348146e-06, "loss": 1.2419, "step": 1578 }, { "epoch": 0.12, "learning_rate": 1.955438625120534e-06, "loss": 1.2862, "step": 1579 }, { "epoch": 0.12, "learning_rate": 1.955364387826547e-06, "loss": 1.2822, "step": 1580 }, { "epoch": 0.12, "learning_rate": 1.9552900901575447e-06, "loss": 1.2323, "step": 1581 }, { "epoch": 0.12, "learning_rate": 1.955215732118223e-06, "loss": 0.9116, "step": 1582 }, { "epoch": 0.12, "learning_rate": 1.9551413137132805e-06, "loss": 1.2918, "step": 1583 }, { "epoch": 0.12, "learning_rate": 1.9550668349474202e-06, "loss": 1.2283, "step": 1584 }, { "epoch": 0.12, "learning_rate": 1.9549922958253494e-06, "loss": 1.2722, "step": 1585 }, { "epoch": 0.12, "learning_rate": 1.954917696351778e-06, "loss": 1.2963, "step": 1586 }, { "epoch": 0.12, "learning_rate": 1.954843036531421e-06, "loss": 1.2296, "step": 1587 }, { "epoch": 0.12, "learning_rate": 1.9547683163689952e-06, "loss": 1.2806, "step": 1588 }, { "epoch": 0.12, "learning_rate": 1.9546935358692244e-06, "loss": 1.2981, "step": 1589 }, { "epoch": 0.12, "learning_rate": 1.9546186950368333e-06, "loss": 1.2096, "step": 1590 }, { "epoch": 0.12, "learning_rate": 1.9545437938765516e-06, "loss": 1.2976, "step": 1591 }, { "epoch": 0.12, "learning_rate": 1.954468832393113e-06, "loss": 0.8842, "step": 1592 }, { "epoch": 0.12, "learning_rate": 1.954393810591255e-06, "loss": 1.3299, "step": 1593 }, { "epoch": 0.12, "learning_rate": 1.954318728475718e-06, "loss": 0.8464, "step": 1594 }, { "epoch": 0.12, "learning_rate": 1.954243586051248e-06, "loss": 1.2623, "step": 1595 }, { "epoch": 0.12, "learning_rate": 1.954168383322592e-06, "loss": 1.2484, "step": 1596 }, { "epoch": 0.12, "learning_rate": 1.9540931202945037e-06, "loss": 1.2377, "step": 1597 }, { "epoch": 0.12, "learning_rate": 1.9540177969717395e-06, "loss": 1.3007, "step": 1598 }, { "epoch": 0.12, "learning_rate": 1.9539424133590586e-06, "loss": 1.2795, "step": 1599 }, { "epoch": 0.12, "learning_rate": 1.953866969461226e-06, "loss": 1.272, "step": 1600 }, { "epoch": 0.12, "learning_rate": 1.9537914652830086e-06, "loss": 1.234, "step": 1601 }, { "epoch": 0.12, "learning_rate": 1.953715900829178e-06, "loss": 1.2529, "step": 1602 }, { "epoch": 0.12, "learning_rate": 1.95364027610451e-06, "loss": 0.8568, "step": 1603 }, { "epoch": 0.12, "learning_rate": 1.9535645911137843e-06, "loss": 1.2226, "step": 1604 }, { "epoch": 0.12, "learning_rate": 1.9534888458617824e-06, "loss": 1.2723, "step": 1605 }, { "epoch": 0.12, "learning_rate": 1.953413040353292e-06, "loss": 0.8432, "step": 1606 }, { "epoch": 0.12, "learning_rate": 1.9533371745931034e-06, "loss": 1.2765, "step": 1607 }, { "epoch": 0.12, "learning_rate": 1.9532612485860113e-06, "loss": 1.2806, "step": 1608 }, { "epoch": 0.12, "learning_rate": 1.9531852623368137e-06, "loss": 0.8754, "step": 1609 }, { "epoch": 0.12, "learning_rate": 1.9531092158503123e-06, "loss": 1.2982, "step": 1610 }, { "epoch": 0.13, "learning_rate": 1.9530331091313137e-06, "loss": 1.2886, "step": 1611 }, { "epoch": 0.13, "learning_rate": 1.952956942184627e-06, "loss": 1.2912, "step": 1612 }, { "epoch": 0.13, "learning_rate": 1.9528807150150658e-06, "loss": 1.3305, "step": 1613 }, { "epoch": 0.13, "learning_rate": 1.9528044276274467e-06, "loss": 1.2268, "step": 1614 }, { "epoch": 0.13, "learning_rate": 1.952728080026592e-06, "loss": 1.2844, "step": 1615 }, { "epoch": 0.13, "learning_rate": 1.9526516722173253e-06, "loss": 1.2812, "step": 1616 }, { "epoch": 0.13, "learning_rate": 1.952575204204476e-06, "loss": 1.264, "step": 1617 }, { "epoch": 0.13, "learning_rate": 1.9524986759928762e-06, "loss": 1.2669, "step": 1618 }, { "epoch": 0.13, "learning_rate": 1.9524220875873626e-06, "loss": 1.2758, "step": 1619 }, { "epoch": 0.13, "learning_rate": 1.9523454389927745e-06, "loss": 1.2616, "step": 1620 }, { "epoch": 0.13, "learning_rate": 1.952268730213957e-06, "loss": 1.2887, "step": 1621 }, { "epoch": 0.13, "learning_rate": 1.952191961255756e-06, "loss": 1.2614, "step": 1622 }, { "epoch": 0.13, "learning_rate": 1.952115132123025e-06, "loss": 1.2885, "step": 1623 }, { "epoch": 0.13, "learning_rate": 1.9520382428206175e-06, "loss": 1.2758, "step": 1624 }, { "epoch": 0.13, "learning_rate": 1.9519612933533938e-06, "loss": 0.8721, "step": 1625 }, { "epoch": 0.13, "learning_rate": 1.9518842837262166e-06, "loss": 1.2468, "step": 1626 }, { "epoch": 0.13, "learning_rate": 1.9518072139439516e-06, "loss": 1.2744, "step": 1627 }, { "epoch": 0.13, "learning_rate": 1.9517300840114707e-06, "loss": 1.2733, "step": 1628 }, { "epoch": 0.13, "learning_rate": 1.9516528939336467e-06, "loss": 1.2586, "step": 1629 }, { "epoch": 0.13, "learning_rate": 1.951575643715359e-06, "loss": 1.2769, "step": 1630 }, { "epoch": 0.13, "learning_rate": 1.9514983333614887e-06, "loss": 1.2417, "step": 1631 }, { "epoch": 0.13, "learning_rate": 1.9514209628769223e-06, "loss": 1.2788, "step": 1632 }, { "epoch": 0.13, "learning_rate": 1.951343532266548e-06, "loss": 1.3189, "step": 1633 }, { "epoch": 0.13, "learning_rate": 1.9512660415352603e-06, "loss": 1.2513, "step": 1634 }, { "epoch": 0.13, "learning_rate": 1.951188490687956e-06, "loss": 1.2985, "step": 1635 }, { "epoch": 0.13, "learning_rate": 1.9511108797295355e-06, "loss": 1.2889, "step": 1636 }, { "epoch": 0.13, "learning_rate": 1.9510332086649037e-06, "loss": 1.1861, "step": 1637 }, { "epoch": 0.13, "learning_rate": 1.9509554774989695e-06, "loss": 1.2497, "step": 1638 }, { "epoch": 0.13, "learning_rate": 1.9508776862366445e-06, "loss": 1.2314, "step": 1639 }, { "epoch": 0.13, "learning_rate": 1.9507998348828456e-06, "loss": 1.2004, "step": 1640 }, { "epoch": 0.13, "learning_rate": 1.950721923442492e-06, "loss": 1.2951, "step": 1641 }, { "epoch": 0.13, "learning_rate": 1.9506439519205074e-06, "loss": 1.3222, "step": 1642 }, { "epoch": 0.13, "learning_rate": 1.9505659203218196e-06, "loss": 1.291, "step": 1643 }, { "epoch": 0.13, "learning_rate": 1.95048782865136e-06, "loss": 1.2848, "step": 1644 }, { "epoch": 0.13, "learning_rate": 1.9504096769140633e-06, "loss": 1.2363, "step": 1645 }, { "epoch": 0.13, "learning_rate": 1.950331465114868e-06, "loss": 1.3146, "step": 1646 }, { "epoch": 0.13, "learning_rate": 1.950253193258718e-06, "loss": 1.2854, "step": 1647 }, { "epoch": 0.13, "learning_rate": 1.9501748613505583e-06, "loss": 1.2675, "step": 1648 }, { "epoch": 0.13, "learning_rate": 1.9500964693953403e-06, "loss": 1.276, "step": 1649 }, { "epoch": 0.13, "learning_rate": 1.950018017398018e-06, "loss": 1.2758, "step": 1650 }, { "epoch": 0.13, "learning_rate": 1.9499395053635476e-06, "loss": 1.1766, "step": 1651 }, { "epoch": 0.13, "learning_rate": 1.949860933296893e-06, "loss": 1.2911, "step": 1652 }, { "epoch": 0.13, "learning_rate": 1.9497823012030177e-06, "loss": 1.2726, "step": 1653 }, { "epoch": 0.13, "learning_rate": 1.9497036090868925e-06, "loss": 1.2739, "step": 1654 }, { "epoch": 0.13, "learning_rate": 1.9496248569534888e-06, "loss": 1.3081, "step": 1655 }, { "epoch": 0.13, "learning_rate": 1.949546044807785e-06, "loss": 1.2722, "step": 1656 }, { "epoch": 0.13, "learning_rate": 1.949467172654761e-06, "loss": 1.2926, "step": 1657 }, { "epoch": 0.13, "learning_rate": 1.949388240499401e-06, "loss": 1.256, "step": 1658 }, { "epoch": 0.13, "learning_rate": 1.9493092483466933e-06, "loss": 1.3269, "step": 1659 }, { "epoch": 0.13, "learning_rate": 1.9492301962016304e-06, "loss": 1.2636, "step": 1660 }, { "epoch": 0.13, "learning_rate": 1.949151084069207e-06, "loss": 1.3284, "step": 1661 }, { "epoch": 0.13, "learning_rate": 1.9490719119544236e-06, "loss": 0.8737, "step": 1662 }, { "epoch": 0.13, "learning_rate": 1.9489926798622835e-06, "loss": 1.296, "step": 1663 }, { "epoch": 0.13, "learning_rate": 1.948913387797793e-06, "loss": 1.2083, "step": 1664 }, { "epoch": 0.13, "learning_rate": 1.9488340357659636e-06, "loss": 0.8475, "step": 1665 }, { "epoch": 0.13, "learning_rate": 1.94875462377181e-06, "loss": 1.2315, "step": 1666 }, { "epoch": 0.13, "learning_rate": 1.9486751518203504e-06, "loss": 1.3124, "step": 1667 }, { "epoch": 0.13, "learning_rate": 1.948595619916608e-06, "loss": 1.2922, "step": 1668 }, { "epoch": 0.13, "learning_rate": 1.948516028065608e-06, "loss": 1.2361, "step": 1669 }, { "epoch": 0.13, "learning_rate": 1.9484363762723805e-06, "loss": 1.3225, "step": 1670 }, { "epoch": 0.13, "learning_rate": 1.948356664541959e-06, "loss": 1.3005, "step": 1671 }, { "epoch": 0.13, "learning_rate": 1.9482768928793812e-06, "loss": 1.2881, "step": 1672 }, { "epoch": 0.13, "learning_rate": 1.9481970612896887e-06, "loss": 1.2342, "step": 1673 }, { "epoch": 0.13, "learning_rate": 1.9481171697779256e-06, "loss": 1.2801, "step": 1674 }, { "epoch": 0.13, "learning_rate": 1.948037218349141e-06, "loss": 1.2655, "step": 1675 }, { "epoch": 0.13, "learning_rate": 1.9479572070083883e-06, "loss": 1.2793, "step": 1676 }, { "epoch": 0.13, "learning_rate": 1.947877135760723e-06, "loss": 1.2281, "step": 1677 }, { "epoch": 0.13, "learning_rate": 1.9477970046112054e-06, "loss": 1.2927, "step": 1678 }, { "epoch": 0.13, "learning_rate": 1.9477168135648997e-06, "loss": 1.2562, "step": 1679 }, { "epoch": 0.13, "learning_rate": 1.9476365626268735e-06, "loss": 1.2563, "step": 1680 }, { "epoch": 0.13, "learning_rate": 1.9475562518021982e-06, "loss": 0.8364, "step": 1681 }, { "epoch": 0.13, "learning_rate": 1.9474758810959494e-06, "loss": 1.2847, "step": 1682 }, { "epoch": 0.13, "learning_rate": 1.9473954505132062e-06, "loss": 1.3153, "step": 1683 }, { "epoch": 0.13, "learning_rate": 1.9473149600590514e-06, "loss": 1.3263, "step": 1684 }, { "epoch": 0.13, "learning_rate": 1.947234409738571e-06, "loss": 1.2029, "step": 1685 }, { "epoch": 0.13, "learning_rate": 1.9471537995568564e-06, "loss": 1.2415, "step": 1686 }, { "epoch": 0.13, "learning_rate": 1.9470731295190014e-06, "loss": 1.2627, "step": 1687 }, { "epoch": 0.13, "learning_rate": 1.946992399630104e-06, "loss": 1.2259, "step": 1688 }, { "epoch": 0.13, "learning_rate": 1.946911609895266e-06, "loss": 1.2636, "step": 1689 }, { "epoch": 0.13, "learning_rate": 1.946830760319593e-06, "loss": 0.8593, "step": 1690 }, { "epoch": 0.13, "learning_rate": 1.9467498509081944e-06, "loss": 1.2328, "step": 1691 }, { "epoch": 0.13, "learning_rate": 1.9466688816661834e-06, "loss": 1.2653, "step": 1692 }, { "epoch": 0.13, "learning_rate": 1.946587852598677e-06, "loss": 1.2041, "step": 1693 }, { "epoch": 0.13, "learning_rate": 1.9465067637107953e-06, "loss": 1.2748, "step": 1694 }, { "epoch": 0.13, "learning_rate": 1.9464256150076635e-06, "loss": 1.2435, "step": 1695 }, { "epoch": 0.13, "learning_rate": 1.9463444064944096e-06, "loss": 1.2523, "step": 1696 }, { "epoch": 0.13, "learning_rate": 1.9462631381761653e-06, "loss": 1.2404, "step": 1697 }, { "epoch": 0.13, "learning_rate": 1.9461818100580666e-06, "loss": 1.3065, "step": 1698 }, { "epoch": 0.13, "learning_rate": 1.9461004221452534e-06, "loss": 1.2511, "step": 1699 }, { "epoch": 0.13, "learning_rate": 1.9460189744428687e-06, "loss": 1.2628, "step": 1700 }, { "epoch": 0.13, "learning_rate": 1.94593746695606e-06, "loss": 1.261, "step": 1701 }, { "epoch": 0.13, "learning_rate": 1.945855899689978e-06, "loss": 1.2604, "step": 1702 }, { "epoch": 0.13, "learning_rate": 1.9457742726497775e-06, "loss": 1.2899, "step": 1703 }, { "epoch": 0.13, "learning_rate": 1.9456925858406166e-06, "loss": 1.2773, "step": 1704 }, { "epoch": 0.13, "learning_rate": 1.9456108392676583e-06, "loss": 1.2978, "step": 1705 }, { "epoch": 0.13, "learning_rate": 1.945529032936068e-06, "loss": 1.2531, "step": 1706 }, { "epoch": 0.13, "learning_rate": 1.9454471668510164e-06, "loss": 1.3154, "step": 1707 }, { "epoch": 0.13, "learning_rate": 1.945365241017676e-06, "loss": 1.2633, "step": 1708 }, { "epoch": 0.13, "learning_rate": 1.9452832554412243e-06, "loss": 1.3376, "step": 1709 }, { "epoch": 0.13, "learning_rate": 1.945201210126843e-06, "loss": 1.2677, "step": 1710 }, { "epoch": 0.13, "learning_rate": 1.9451191050797168e-06, "loss": 1.2028, "step": 1711 }, { "epoch": 0.13, "learning_rate": 1.9450369403050345e-06, "loss": 1.2318, "step": 1712 }, { "epoch": 0.13, "learning_rate": 1.9449547158079883e-06, "loss": 1.2967, "step": 1713 }, { "epoch": 0.13, "learning_rate": 1.944872431593775e-06, "loss": 1.2423, "step": 1714 }, { "epoch": 0.13, "learning_rate": 1.944790087667594e-06, "loss": 0.9043, "step": 1715 }, { "epoch": 0.13, "learning_rate": 1.944707684034649e-06, "loss": 1.3034, "step": 1716 }, { "epoch": 0.13, "learning_rate": 1.944625220700148e-06, "loss": 1.2889, "step": 1717 }, { "epoch": 0.13, "learning_rate": 1.9445426976693025e-06, "loss": 1.2335, "step": 1718 }, { "epoch": 0.13, "learning_rate": 1.944460114947327e-06, "loss": 1.2662, "step": 1719 }, { "epoch": 0.13, "learning_rate": 1.9443774725394405e-06, "loss": 1.2984, "step": 1720 }, { "epoch": 0.13, "learning_rate": 1.9442947704508664e-06, "loss": 1.2312, "step": 1721 }, { "epoch": 0.13, "learning_rate": 1.94421200868683e-06, "loss": 1.2377, "step": 1722 }, { "epoch": 0.13, "learning_rate": 1.9441291872525626e-06, "loss": 1.2444, "step": 1723 }, { "epoch": 0.13, "learning_rate": 1.944046306153297e-06, "loss": 1.272, "step": 1724 }, { "epoch": 0.13, "learning_rate": 1.943963365394272e-06, "loss": 1.2463, "step": 1725 }, { "epoch": 0.13, "learning_rate": 1.9438803649807287e-06, "loss": 0.8801, "step": 1726 }, { "epoch": 0.13, "learning_rate": 1.943797304917912e-06, "loss": 0.8518, "step": 1727 }, { "epoch": 0.13, "learning_rate": 1.9437141852110715e-06, "loss": 1.3212, "step": 1728 }, { "epoch": 0.13, "learning_rate": 1.9436310058654598e-06, "loss": 1.2706, "step": 1729 }, { "epoch": 0.13, "learning_rate": 1.943547766886334e-06, "loss": 1.3023, "step": 1730 }, { "epoch": 0.13, "learning_rate": 1.9434644682789533e-06, "loss": 0.877, "step": 1731 }, { "epoch": 0.13, "learning_rate": 1.9433811100485826e-06, "loss": 1.3065, "step": 1732 }, { "epoch": 0.13, "learning_rate": 1.94329769220049e-06, "loss": 1.2995, "step": 1733 }, { "epoch": 0.13, "learning_rate": 1.9432142147399464e-06, "loss": 1.3053, "step": 1734 }, { "epoch": 0.13, "learning_rate": 1.9431306776722277e-06, "loss": 1.3138, "step": 1735 }, { "epoch": 0.13, "learning_rate": 1.9430470810026134e-06, "loss": 1.2832, "step": 1736 }, { "epoch": 0.13, "learning_rate": 1.9429634247363856e-06, "loss": 0.8967, "step": 1737 }, { "epoch": 0.13, "learning_rate": 1.942879708878832e-06, "loss": 1.2937, "step": 1738 }, { "epoch": 0.13, "learning_rate": 1.942795933435242e-06, "loss": 1.2431, "step": 1739 }, { "epoch": 0.14, "learning_rate": 1.942712098410911e-06, "loss": 1.2184, "step": 1740 }, { "epoch": 0.14, "learning_rate": 1.9426282038111363e-06, "loss": 1.2694, "step": 1741 }, { "epoch": 0.14, "learning_rate": 1.94254424964122e-06, "loss": 1.2793, "step": 1742 }, { "epoch": 0.14, "learning_rate": 1.942460235906468e-06, "loss": 0.864, "step": 1743 }, { "epoch": 0.14, "learning_rate": 1.9423761626121884e-06, "loss": 1.2515, "step": 1744 }, { "epoch": 0.14, "learning_rate": 1.9422920297636953e-06, "loss": 1.2839, "step": 1745 }, { "epoch": 0.14, "learning_rate": 1.942207837366305e-06, "loss": 1.2941, "step": 1746 }, { "epoch": 0.14, "learning_rate": 1.9421235854253383e-06, "loss": 1.2546, "step": 1747 }, { "epoch": 0.14, "learning_rate": 1.94203927394612e-06, "loss": 1.2932, "step": 1748 }, { "epoch": 0.14, "learning_rate": 1.941954902933978e-06, "loss": 1.2886, "step": 1749 }, { "epoch": 0.14, "learning_rate": 1.9418704723942435e-06, "loss": 1.247, "step": 1750 }, { "epoch": 0.14, "learning_rate": 1.9417859823322533e-06, "loss": 1.2811, "step": 1751 }, { "epoch": 0.14, "learning_rate": 1.9417014327533457e-06, "loss": 1.2897, "step": 1752 }, { "epoch": 0.14, "learning_rate": 1.9416168236628646e-06, "loss": 1.249, "step": 1753 }, { "epoch": 0.14, "learning_rate": 1.941532155066157e-06, "loss": 1.245, "step": 1754 }, { "epoch": 0.14, "learning_rate": 1.941447426968573e-06, "loss": 1.2878, "step": 1755 }, { "epoch": 0.14, "learning_rate": 1.9413626393754675e-06, "loss": 1.2749, "step": 1756 }, { "epoch": 0.14, "learning_rate": 1.9412777922921987e-06, "loss": 1.2059, "step": 1757 }, { "epoch": 0.14, "learning_rate": 1.9411928857241285e-06, "loss": 1.2845, "step": 1758 }, { "epoch": 0.14, "learning_rate": 1.941107919676623e-06, "loss": 1.2723, "step": 1759 }, { "epoch": 0.14, "learning_rate": 1.941022894155051e-06, "loss": 1.2667, "step": 1760 }, { "epoch": 0.14, "learning_rate": 1.940937809164786e-06, "loss": 1.2934, "step": 1761 }, { "epoch": 0.14, "learning_rate": 1.940852664711205e-06, "loss": 1.2244, "step": 1762 }, { "epoch": 0.14, "learning_rate": 1.940767460799689e-06, "loss": 1.233, "step": 1763 }, { "epoch": 0.14, "learning_rate": 1.9406821974356227e-06, "loss": 0.8701, "step": 1764 }, { "epoch": 0.14, "learning_rate": 1.940596874624394e-06, "loss": 1.2069, "step": 1765 }, { "epoch": 0.14, "learning_rate": 1.940511492371395e-06, "loss": 1.287, "step": 1766 }, { "epoch": 0.14, "learning_rate": 1.9404260506820215e-06, "loss": 1.2465, "step": 1767 }, { "epoch": 0.14, "learning_rate": 1.9403405495616734e-06, "loss": 1.2408, "step": 1768 }, { "epoch": 0.14, "learning_rate": 1.940254989015753e-06, "loss": 1.229, "step": 1769 }, { "epoch": 0.14, "learning_rate": 1.9401693690496686e-06, "loss": 1.2555, "step": 1770 }, { "epoch": 0.14, "learning_rate": 1.9400836896688306e-06, "loss": 1.2971, "step": 1771 }, { "epoch": 0.14, "learning_rate": 1.939997950878653e-06, "loss": 1.2419, "step": 1772 }, { "epoch": 0.14, "learning_rate": 1.9399121526845554e-06, "loss": 1.3058, "step": 1773 }, { "epoch": 0.14, "learning_rate": 1.9398262950919584e-06, "loss": 1.3223, "step": 1774 }, { "epoch": 0.14, "learning_rate": 1.939740378106289e-06, "loss": 1.2536, "step": 1775 }, { "epoch": 0.14, "learning_rate": 1.9396544017329756e-06, "loss": 1.2743, "step": 1776 }, { "epoch": 0.14, "learning_rate": 1.9395683659774527e-06, "loss": 1.2831, "step": 1777 }, { "epoch": 0.14, "learning_rate": 1.9394822708451573e-06, "loss": 1.2723, "step": 1778 }, { "epoch": 0.14, "learning_rate": 1.9393961163415295e-06, "loss": 1.2664, "step": 1779 }, { "epoch": 0.14, "learning_rate": 1.9393099024720145e-06, "loss": 0.859, "step": 1780 }, { "epoch": 0.14, "learning_rate": 1.9392236292420605e-06, "loss": 1.3044, "step": 1781 }, { "epoch": 0.14, "learning_rate": 1.9391372966571192e-06, "loss": 1.2819, "step": 1782 }, { "epoch": 0.14, "learning_rate": 1.9390509047226474e-06, "loss": 0.8423, "step": 1783 }, { "epoch": 0.14, "learning_rate": 1.9389644534441037e-06, "loss": 1.2796, "step": 1784 }, { "epoch": 0.14, "learning_rate": 1.9388779428269523e-06, "loss": 1.2657, "step": 1785 }, { "epoch": 0.14, "learning_rate": 1.9387913728766596e-06, "loss": 1.2449, "step": 1786 }, { "epoch": 0.14, "learning_rate": 1.938704743598697e-06, "loss": 0.8408, "step": 1787 }, { "epoch": 0.14, "learning_rate": 1.9386180549985388e-06, "loss": 0.8069, "step": 1788 }, { "epoch": 0.14, "learning_rate": 1.938531307081663e-06, "loss": 1.2919, "step": 1789 }, { "epoch": 0.14, "learning_rate": 1.9384444998535527e-06, "loss": 1.3099, "step": 1790 }, { "epoch": 0.14, "learning_rate": 1.938357633319693e-06, "loss": 1.2562, "step": 1791 }, { "epoch": 0.14, "learning_rate": 1.938270707485574e-06, "loss": 1.2362, "step": 1792 }, { "epoch": 0.14, "learning_rate": 1.938183722356688e-06, "loss": 1.2916, "step": 1793 }, { "epoch": 0.14, "learning_rate": 1.9380966779385335e-06, "loss": 1.1876, "step": 1794 }, { "epoch": 0.14, "learning_rate": 1.9380095742366103e-06, "loss": 1.2941, "step": 1795 }, { "epoch": 0.14, "learning_rate": 1.9379224112564233e-06, "loss": 1.2254, "step": 1796 }, { "epoch": 0.14, "learning_rate": 1.9378351890034813e-06, "loss": 1.3112, "step": 1797 }, { "epoch": 0.14, "learning_rate": 1.9377479074832957e-06, "loss": 0.8737, "step": 1798 }, { "epoch": 0.14, "learning_rate": 1.9376605667013827e-06, "loss": 1.2503, "step": 1799 }, { "epoch": 0.14, "learning_rate": 1.9375731666632617e-06, "loss": 1.2443, "step": 1800 }, { "epoch": 0.14, "learning_rate": 1.9374857073744563e-06, "loss": 1.2862, "step": 1801 }, { "epoch": 0.14, "learning_rate": 1.937398188840493e-06, "loss": 1.268, "step": 1802 }, { "epoch": 0.14, "learning_rate": 1.937310611066903e-06, "loss": 1.2606, "step": 1803 }, { "epoch": 0.14, "learning_rate": 1.937222974059221e-06, "loss": 1.3101, "step": 1804 }, { "epoch": 0.14, "learning_rate": 1.937135277822985e-06, "loss": 1.2717, "step": 1805 }, { "epoch": 0.14, "learning_rate": 1.9370475223637368e-06, "loss": 1.2644, "step": 1806 }, { "epoch": 0.14, "learning_rate": 1.936959707687023e-06, "loss": 1.3137, "step": 1807 }, { "epoch": 0.14, "learning_rate": 1.9368718337983923e-06, "loss": 1.2554, "step": 1808 }, { "epoch": 0.14, "learning_rate": 1.9367839007033985e-06, "loss": 1.2771, "step": 1809 }, { "epoch": 0.14, "learning_rate": 1.936695908407598e-06, "loss": 1.2356, "step": 1810 }, { "epoch": 0.14, "learning_rate": 1.9366078569165523e-06, "loss": 1.2662, "step": 1811 }, { "epoch": 0.14, "learning_rate": 1.9365197462358253e-06, "loss": 1.309, "step": 1812 }, { "epoch": 0.14, "learning_rate": 1.936431576370985e-06, "loss": 1.2279, "step": 1813 }, { "epoch": 0.14, "learning_rate": 1.9363433473276046e-06, "loss": 1.2248, "step": 1814 }, { "epoch": 0.14, "learning_rate": 1.936255059111259e-06, "loss": 0.8957, "step": 1815 }, { "epoch": 0.14, "learning_rate": 1.936166711727527e-06, "loss": 1.2794, "step": 1816 }, { "epoch": 0.14, "learning_rate": 1.936078305181993e-06, "loss": 1.2385, "step": 1817 }, { "epoch": 0.14, "learning_rate": 1.935989839480243e-06, "loss": 1.2661, "step": 1818 }, { "epoch": 0.14, "learning_rate": 1.935901314627868e-06, "loss": 1.3165, "step": 1819 }, { "epoch": 0.14, "learning_rate": 1.9358127306304627e-06, "loss": 1.2046, "step": 1820 }, { "epoch": 0.14, "learning_rate": 1.9357240874936247e-06, "loss": 1.3025, "step": 1821 }, { "epoch": 0.14, "learning_rate": 1.935635385222956e-06, "loss": 1.25, "step": 1822 }, { "epoch": 0.14, "learning_rate": 1.9355466238240624e-06, "loss": 1.2872, "step": 1823 }, { "epoch": 0.14, "learning_rate": 1.9354578033025533e-06, "loss": 1.3044, "step": 1824 }, { "epoch": 0.14, "learning_rate": 1.9353689236640414e-06, "loss": 1.2761, "step": 1825 }, { "epoch": 0.14, "learning_rate": 1.9352799849141444e-06, "loss": 1.2856, "step": 1826 }, { "epoch": 0.14, "learning_rate": 1.935190987058482e-06, "loss": 1.2572, "step": 1827 }, { "epoch": 0.14, "learning_rate": 1.935101930102678e-06, "loss": 0.8872, "step": 1828 }, { "epoch": 0.14, "learning_rate": 1.935012814052362e-06, "loss": 1.2665, "step": 1829 }, { "epoch": 0.14, "learning_rate": 1.9349236389131645e-06, "loss": 1.2607, "step": 1830 }, { "epoch": 0.14, "learning_rate": 1.934834404690722e-06, "loss": 1.2454, "step": 1831 }, { "epoch": 0.14, "learning_rate": 1.9347451113906725e-06, "loss": 1.31, "step": 1832 }, { "epoch": 0.14, "learning_rate": 1.9346557590186597e-06, "loss": 1.264, "step": 1833 }, { "epoch": 0.14, "learning_rate": 1.93456634758033e-06, "loss": 1.2576, "step": 1834 }, { "epoch": 0.14, "learning_rate": 1.9344768770813344e-06, "loss": 1.2991, "step": 1835 }, { "epoch": 0.14, "learning_rate": 1.9343873475273263e-06, "loss": 1.2653, "step": 1836 }, { "epoch": 0.14, "learning_rate": 1.9342977589239643e-06, "loss": 1.2348, "step": 1837 }, { "epoch": 0.14, "learning_rate": 1.93420811127691e-06, "loss": 1.2889, "step": 1838 }, { "epoch": 0.14, "learning_rate": 1.9341184045918275e-06, "loss": 1.3178, "step": 1839 }, { "epoch": 0.14, "learning_rate": 1.9340286388743874e-06, "loss": 1.2961, "step": 1840 }, { "epoch": 0.14, "learning_rate": 1.933938814130262e-06, "loss": 1.2507, "step": 1841 }, { "epoch": 0.14, "learning_rate": 1.933848930365128e-06, "loss": 1.288, "step": 1842 }, { "epoch": 0.14, "learning_rate": 1.933758987584665e-06, "loss": 1.1972, "step": 1843 }, { "epoch": 0.14, "learning_rate": 1.933668985794558e-06, "loss": 1.2563, "step": 1844 }, { "epoch": 0.14, "learning_rate": 1.9335789250004935e-06, "loss": 1.2942, "step": 1845 }, { "epoch": 0.14, "learning_rate": 1.933488805208164e-06, "loss": 1.2404, "step": 1846 }, { "epoch": 0.14, "learning_rate": 1.9333986264232644e-06, "loss": 1.2474, "step": 1847 }, { "epoch": 0.14, "learning_rate": 1.9333083886514934e-06, "loss": 1.2832, "step": 1848 }, { "epoch": 0.14, "learning_rate": 1.933218091898554e-06, "loss": 1.2633, "step": 1849 }, { "epoch": 0.14, "learning_rate": 1.9331277361701524e-06, "loss": 0.8597, "step": 1850 }, { "epoch": 0.14, "learning_rate": 1.9330373214719987e-06, "loss": 1.234, "step": 1851 }, { "epoch": 0.14, "learning_rate": 1.932946847809807e-06, "loss": 1.3041, "step": 1852 }, { "epoch": 0.14, "learning_rate": 1.9328563151892943e-06, "loss": 1.3105, "step": 1853 }, { "epoch": 0.14, "learning_rate": 1.9327657236161824e-06, "loss": 1.2949, "step": 1854 }, { "epoch": 0.14, "learning_rate": 1.9326750730961957e-06, "loss": 1.2168, "step": 1855 }, { "epoch": 0.14, "learning_rate": 1.9325843636350634e-06, "loss": 1.3075, "step": 1856 }, { "epoch": 0.14, "learning_rate": 1.932493595238518e-06, "loss": 1.2823, "step": 1857 }, { "epoch": 0.14, "learning_rate": 1.9324027679122955e-06, "loss": 0.8398, "step": 1858 }, { "epoch": 0.14, "learning_rate": 1.932311881662136e-06, "loss": 1.2529, "step": 1859 }, { "epoch": 0.14, "learning_rate": 1.932220936493783e-06, "loss": 1.2634, "step": 1860 }, { "epoch": 0.14, "learning_rate": 1.932129932412984e-06, "loss": 1.2601, "step": 1861 }, { "epoch": 0.14, "learning_rate": 1.93203886942549e-06, "loss": 1.2484, "step": 1862 }, { "epoch": 0.14, "learning_rate": 1.931947747537055e-06, "loss": 0.8841, "step": 1863 }, { "epoch": 0.14, "learning_rate": 1.9318565667534386e-06, "loss": 1.2926, "step": 1864 }, { "epoch": 0.14, "learning_rate": 1.9317653270804025e-06, "loss": 1.2414, "step": 1865 }, { "epoch": 0.14, "learning_rate": 1.931674028523713e-06, "loss": 1.2774, "step": 1866 }, { "epoch": 0.14, "learning_rate": 1.93158267108914e-06, "loss": 1.2675, "step": 1867 }, { "epoch": 0.14, "learning_rate": 1.931491254782456e-06, "loss": 0.8745, "step": 1868 }, { "epoch": 0.15, "learning_rate": 1.931399779609439e-06, "loss": 1.2793, "step": 1869 }, { "epoch": 0.15, "learning_rate": 1.9313082455758695e-06, "loss": 1.224, "step": 1870 }, { "epoch": 0.15, "learning_rate": 1.9312166526875316e-06, "loss": 1.2683, "step": 1871 }, { "epoch": 0.15, "learning_rate": 1.9311250009502143e-06, "loss": 1.258, "step": 1872 }, { "epoch": 0.15, "learning_rate": 1.9310332903697096e-06, "loss": 0.8761, "step": 1873 }, { "epoch": 0.15, "learning_rate": 1.930941520951813e-06, "loss": 1.2543, "step": 1874 }, { "epoch": 0.15, "learning_rate": 1.9308496927023238e-06, "loss": 1.2633, "step": 1875 }, { "epoch": 0.15, "learning_rate": 1.930757805627045e-06, "loss": 1.2451, "step": 1876 }, { "epoch": 0.15, "learning_rate": 1.930665859731784e-06, "loss": 1.2647, "step": 1877 }, { "epoch": 0.15, "learning_rate": 1.930573855022351e-06, "loss": 1.3025, "step": 1878 }, { "epoch": 0.15, "learning_rate": 1.9304817915045607e-06, "loss": 1.2891, "step": 1879 }, { "epoch": 0.15, "learning_rate": 1.930389669184231e-06, "loss": 1.2492, "step": 1880 }, { "epoch": 0.15, "learning_rate": 1.9302974880671834e-06, "loss": 1.3179, "step": 1881 }, { "epoch": 0.15, "learning_rate": 1.930205248159243e-06, "loss": 1.2668, "step": 1882 }, { "epoch": 0.15, "learning_rate": 1.9301129494662406e-06, "loss": 1.2844, "step": 1883 }, { "epoch": 0.15, "learning_rate": 1.9300205919940067e-06, "loss": 1.2506, "step": 1884 }, { "epoch": 0.15, "learning_rate": 1.92992817574838e-06, "loss": 1.2778, "step": 1885 }, { "epoch": 0.15, "learning_rate": 1.9298357007351997e-06, "loss": 1.251, "step": 1886 }, { "epoch": 0.15, "learning_rate": 1.9297431669603105e-06, "loss": 0.8975, "step": 1887 }, { "epoch": 0.15, "learning_rate": 1.9296505744295596e-06, "loss": 1.2768, "step": 1888 }, { "epoch": 0.15, "learning_rate": 1.9295579231487983e-06, "loss": 0.9285, "step": 1889 }, { "epoch": 0.15, "learning_rate": 1.929465213123882e-06, "loss": 1.2393, "step": 1890 }, { "epoch": 0.15, "learning_rate": 1.92937244436067e-06, "loss": 1.2444, "step": 1891 }, { "epoch": 0.15, "learning_rate": 1.9292796168650244e-06, "loss": 1.277, "step": 1892 }, { "epoch": 0.15, "learning_rate": 1.929186730642812e-06, "loss": 1.2583, "step": 1893 }, { "epoch": 0.15, "learning_rate": 1.929093785699902e-06, "loss": 1.2602, "step": 1894 }, { "epoch": 0.15, "learning_rate": 1.9290007820421693e-06, "loss": 1.2834, "step": 1895 }, { "epoch": 0.15, "learning_rate": 1.9289077196754906e-06, "loss": 1.291, "step": 1896 }, { "epoch": 0.15, "learning_rate": 1.9288145986057466e-06, "loss": 1.257, "step": 1897 }, { "epoch": 0.15, "learning_rate": 1.928721418838823e-06, "loss": 1.2539, "step": 1898 }, { "epoch": 0.15, "learning_rate": 1.9286281803806077e-06, "loss": 1.2736, "step": 1899 }, { "epoch": 0.15, "learning_rate": 1.9285348832369937e-06, "loss": 1.3169, "step": 1900 }, { "epoch": 0.15, "learning_rate": 1.9284415274138764e-06, "loss": 0.8746, "step": 1901 }, { "epoch": 0.15, "learning_rate": 1.9283481129171557e-06, "loss": 1.2569, "step": 1902 }, { "epoch": 0.15, "learning_rate": 1.9282546397527354e-06, "loss": 0.8743, "step": 1903 }, { "epoch": 0.15, "learning_rate": 1.928161107926522e-06, "loss": 1.2757, "step": 1904 }, { "epoch": 0.15, "learning_rate": 1.9280675174444264e-06, "loss": 1.2508, "step": 1905 }, { "epoch": 0.15, "learning_rate": 1.9279738683123634e-06, "loss": 1.3002, "step": 1906 }, { "epoch": 0.15, "learning_rate": 1.927880160536251e-06, "loss": 1.2824, "step": 1907 }, { "epoch": 0.15, "learning_rate": 1.9277863941220113e-06, "loss": 1.2981, "step": 1908 }, { "epoch": 0.15, "learning_rate": 1.92769256907557e-06, "loss": 1.2739, "step": 1909 }, { "epoch": 0.15, "learning_rate": 1.9275986854028563e-06, "loss": 1.243, "step": 1910 }, { "epoch": 0.15, "learning_rate": 1.927504743109803e-06, "loss": 1.2885, "step": 1911 }, { "epoch": 0.15, "learning_rate": 1.9274107422023475e-06, "loss": 1.2367, "step": 1912 }, { "epoch": 0.15, "learning_rate": 1.92731668268643e-06, "loss": 1.29, "step": 1913 }, { "epoch": 0.15, "learning_rate": 1.927222564567994e-06, "loss": 0.8289, "step": 1914 }, { "epoch": 0.15, "learning_rate": 1.9271283878529883e-06, "loss": 1.225, "step": 1915 }, { "epoch": 0.15, "learning_rate": 1.9270341525473643e-06, "loss": 0.8552, "step": 1916 }, { "epoch": 0.15, "learning_rate": 1.926939858657077e-06, "loss": 1.2565, "step": 1917 }, { "epoch": 0.15, "learning_rate": 1.9268455061880854e-06, "loss": 1.2263, "step": 1918 }, { "epoch": 0.15, "learning_rate": 1.9267510951463525e-06, "loss": 1.3185, "step": 1919 }, { "epoch": 0.15, "learning_rate": 1.926656625537844e-06, "loss": 1.288, "step": 1920 }, { "epoch": 0.15, "learning_rate": 1.9265620973685304e-06, "loss": 1.253, "step": 1921 }, { "epoch": 0.15, "learning_rate": 1.926467510644386e-06, "loss": 1.3032, "step": 1922 }, { "epoch": 0.15, "learning_rate": 1.926372865371387e-06, "loss": 1.2559, "step": 1923 }, { "epoch": 0.15, "learning_rate": 1.926278161555516e-06, "loss": 1.2714, "step": 1924 }, { "epoch": 0.15, "learning_rate": 1.926183399202757e-06, "loss": 1.2685, "step": 1925 }, { "epoch": 0.15, "learning_rate": 1.9260885783190993e-06, "loss": 0.8877, "step": 1926 }, { "epoch": 0.15, "learning_rate": 1.9259936989105342e-06, "loss": 1.2571, "step": 1927 }, { "epoch": 0.15, "learning_rate": 1.925898760983058e-06, "loss": 1.2531, "step": 1928 }, { "epoch": 0.15, "learning_rate": 1.925803764542671e-06, "loss": 0.8549, "step": 1929 }, { "epoch": 0.15, "learning_rate": 1.925708709595376e-06, "loss": 1.2634, "step": 1930 }, { "epoch": 0.15, "learning_rate": 1.92561359614718e-06, "loss": 1.2398, "step": 1931 }, { "epoch": 0.15, "learning_rate": 1.9255184242040947e-06, "loss": 1.332, "step": 1932 }, { "epoch": 0.15, "learning_rate": 1.925423193772133e-06, "loss": 1.2023, "step": 1933 }, { "epoch": 0.15, "learning_rate": 1.9253279048573147e-06, "loss": 1.1602, "step": 1934 }, { "epoch": 0.15, "learning_rate": 1.9252325574656602e-06, "loss": 0.8461, "step": 1935 }, { "epoch": 0.15, "learning_rate": 1.925137151603196e-06, "loss": 1.259, "step": 1936 }, { "epoch": 0.15, "learning_rate": 1.925041687275951e-06, "loss": 1.2693, "step": 1937 }, { "epoch": 0.15, "learning_rate": 1.9249461644899583e-06, "loss": 1.2476, "step": 1938 }, { "epoch": 0.15, "learning_rate": 1.9248505832512545e-06, "loss": 1.2801, "step": 1939 }, { "epoch": 0.15, "learning_rate": 1.9247549435658797e-06, "loss": 1.1833, "step": 1940 }, { "epoch": 0.15, "learning_rate": 1.9246592454398784e-06, "loss": 1.3407, "step": 1941 }, { "epoch": 0.15, "learning_rate": 1.924563488879298e-06, "loss": 1.2744, "step": 1942 }, { "epoch": 0.15, "learning_rate": 1.92446767389019e-06, "loss": 1.2882, "step": 1943 }, { "epoch": 0.15, "learning_rate": 1.9243718004786095e-06, "loss": 1.227, "step": 1944 }, { "epoch": 0.15, "learning_rate": 1.924275868650615e-06, "loss": 0.8499, "step": 1945 }, { "epoch": 0.15, "learning_rate": 1.924179878412269e-06, "loss": 1.2691, "step": 1946 }, { "epoch": 0.15, "learning_rate": 1.9240838297696383e-06, "loss": 1.2649, "step": 1947 }, { "epoch": 0.15, "learning_rate": 1.9239877227287925e-06, "loss": 1.2377, "step": 1948 }, { "epoch": 0.15, "learning_rate": 1.9238915572958048e-06, "loss": 1.2563, "step": 1949 }, { "epoch": 0.15, "learning_rate": 1.923795333476753e-06, "loss": 1.2796, "step": 1950 }, { "epoch": 0.15, "learning_rate": 1.9236990512777175e-06, "loss": 0.8387, "step": 1951 }, { "epoch": 0.15, "learning_rate": 1.923602710704783e-06, "loss": 1.2343, "step": 1952 }, { "epoch": 0.15, "learning_rate": 1.9235063117640386e-06, "loss": 1.3281, "step": 1953 }, { "epoch": 0.15, "learning_rate": 1.9234098544615754e-06, "loss": 1.2797, "step": 1954 }, { "epoch": 0.15, "learning_rate": 1.9233133388034893e-06, "loss": 1.3069, "step": 1955 }, { "epoch": 0.15, "learning_rate": 1.92321676479588e-06, "loss": 1.2233, "step": 1956 }, { "epoch": 0.15, "learning_rate": 1.92312013244485e-06, "loss": 1.2091, "step": 1957 }, { "epoch": 0.15, "learning_rate": 1.9230234417565066e-06, "loss": 1.2527, "step": 1958 }, { "epoch": 0.15, "learning_rate": 1.9229266927369603e-06, "loss": 1.3075, "step": 1959 }, { "epoch": 0.15, "learning_rate": 1.9228298853923244e-06, "loss": 0.8681, "step": 1960 }, { "epoch": 0.15, "learning_rate": 1.922733019728718e-06, "loss": 1.2424, "step": 1961 }, { "epoch": 0.15, "learning_rate": 1.922636095752261e-06, "loss": 1.2346, "step": 1962 }, { "epoch": 0.15, "learning_rate": 1.9225391134690806e-06, "loss": 1.2519, "step": 1963 }, { "epoch": 0.15, "learning_rate": 1.922442072885304e-06, "loss": 1.2752, "step": 1964 }, { "epoch": 0.15, "learning_rate": 1.9223449740070643e-06, "loss": 1.2551, "step": 1965 }, { "epoch": 0.15, "learning_rate": 1.922247816840498e-06, "loss": 1.2636, "step": 1966 }, { "epoch": 0.15, "learning_rate": 1.922150601391745e-06, "loss": 1.2209, "step": 1967 }, { "epoch": 0.15, "learning_rate": 1.9220533276669482e-06, "loss": 1.2814, "step": 1968 }, { "epoch": 0.15, "learning_rate": 1.9219559956722554e-06, "loss": 1.2958, "step": 1969 }, { "epoch": 0.15, "learning_rate": 1.9218586054138178e-06, "loss": 1.2523, "step": 1970 }, { "epoch": 0.15, "learning_rate": 1.92176115689779e-06, "loss": 1.2707, "step": 1971 }, { "epoch": 0.15, "learning_rate": 1.92166365013033e-06, "loss": 1.2596, "step": 1972 }, { "epoch": 0.15, "learning_rate": 1.9215660851175996e-06, "loss": 1.2496, "step": 1973 }, { "epoch": 0.15, "learning_rate": 1.921468461865765e-06, "loss": 1.3025, "step": 1974 }, { "epoch": 0.15, "learning_rate": 1.9213707803809957e-06, "loss": 1.2554, "step": 1975 }, { "epoch": 0.15, "learning_rate": 1.921273040669465e-06, "loss": 1.231, "step": 1976 }, { "epoch": 0.15, "learning_rate": 1.9211752427373482e-06, "loss": 1.2883, "step": 1977 }, { "epoch": 0.15, "learning_rate": 1.921077386590827e-06, "loss": 1.2233, "step": 1978 }, { "epoch": 0.15, "learning_rate": 1.9209794722360854e-06, "loss": 1.233, "step": 1979 }, { "epoch": 0.15, "learning_rate": 1.9208814996793107e-06, "loss": 1.2449, "step": 1980 }, { "epoch": 0.15, "learning_rate": 1.9207834689266948e-06, "loss": 1.2272, "step": 1981 }, { "epoch": 0.15, "learning_rate": 1.9206853799844324e-06, "loss": 1.2647, "step": 1982 }, { "epoch": 0.15, "learning_rate": 1.920587232858723e-06, "loss": 1.2807, "step": 1983 }, { "epoch": 0.15, "learning_rate": 1.920489027555768e-06, "loss": 1.2655, "step": 1984 }, { "epoch": 0.15, "learning_rate": 1.9203907640817745e-06, "loss": 1.3032, "step": 1985 }, { "epoch": 0.15, "learning_rate": 1.9202924424429525e-06, "loss": 1.2232, "step": 1986 }, { "epoch": 0.15, "learning_rate": 1.9201940626455146e-06, "loss": 1.3462, "step": 1987 }, { "epoch": 0.15, "learning_rate": 1.920095624695678e-06, "loss": 1.3151, "step": 1988 }, { "epoch": 0.15, "learning_rate": 1.9199971285996646e-06, "loss": 1.283, "step": 1989 }, { "epoch": 0.15, "learning_rate": 1.9198985743636983e-06, "loss": 1.2783, "step": 1990 }, { "epoch": 0.15, "learning_rate": 1.9197999619940067e-06, "loss": 1.2487, "step": 1991 }, { "epoch": 0.15, "learning_rate": 1.919701291496823e-06, "loss": 1.2442, "step": 1992 }, { "epoch": 0.15, "learning_rate": 1.919602562878382e-06, "loss": 1.2019, "step": 1993 }, { "epoch": 0.15, "learning_rate": 1.9195037761449232e-06, "loss": 1.221, "step": 1994 }, { "epoch": 0.15, "learning_rate": 1.919404931302689e-06, "loss": 0.8975, "step": 1995 }, { "epoch": 0.15, "learning_rate": 1.9193060283579263e-06, "loss": 0.8621, "step": 1996 }, { "epoch": 0.15, "learning_rate": 1.9192070673168853e-06, "loss": 1.2505, "step": 1997 }, { "epoch": 0.16, "learning_rate": 1.9191080481858203e-06, "loss": 1.2815, "step": 1998 }, { "epoch": 0.16, "learning_rate": 1.9190089709709885e-06, "loss": 1.299, "step": 1999 }, { "epoch": 0.16, "learning_rate": 1.9189098356786514e-06, "loss": 1.2639, "step": 2000 }, { "epoch": 0.16, "learning_rate": 1.918810642315074e-06, "loss": 1.2151, "step": 2001 }, { "epoch": 0.16, "learning_rate": 1.9187113908865237e-06, "loss": 1.2403, "step": 2002 }, { "epoch": 0.16, "learning_rate": 1.918612081399274e-06, "loss": 1.262, "step": 2003 }, { "epoch": 0.16, "learning_rate": 1.9185127138596014e-06, "loss": 1.2714, "step": 2004 }, { "epoch": 0.16, "learning_rate": 1.9184132882737844e-06, "loss": 1.2644, "step": 2005 }, { "epoch": 0.16, "learning_rate": 1.9183138046481064e-06, "loss": 1.2241, "step": 2006 }, { "epoch": 0.16, "learning_rate": 1.9182142629888546e-06, "loss": 1.2317, "step": 2007 }, { "epoch": 0.16, "learning_rate": 1.9181146633023194e-06, "loss": 1.2385, "step": 2008 }, { "epoch": 0.16, "learning_rate": 1.9180150055947952e-06, "loss": 1.2131, "step": 2009 }, { "epoch": 0.16, "learning_rate": 1.9179152898725805e-06, "loss": 1.2585, "step": 2010 }, { "epoch": 0.16, "learning_rate": 1.917815516141976e-06, "loss": 1.2861, "step": 2011 }, { "epoch": 0.16, "learning_rate": 1.917715684409287e-06, "loss": 1.2284, "step": 2012 }, { "epoch": 0.16, "learning_rate": 1.9176157946808235e-06, "loss": 1.2189, "step": 2013 }, { "epoch": 0.16, "learning_rate": 1.9175158469628973e-06, "loss": 1.2645, "step": 2014 }, { "epoch": 0.16, "learning_rate": 1.9174158412618245e-06, "loss": 1.2513, "step": 2015 }, { "epoch": 0.16, "learning_rate": 1.9173157775839257e-06, "loss": 1.1915, "step": 2016 }, { "epoch": 0.16, "learning_rate": 1.917215655935524e-06, "loss": 1.2312, "step": 2017 }, { "epoch": 0.16, "learning_rate": 1.9171154763229466e-06, "loss": 1.2686, "step": 2018 }, { "epoch": 0.16, "learning_rate": 1.917015238752525e-06, "loss": 1.2582, "step": 2019 }, { "epoch": 0.16, "learning_rate": 1.916914943230593e-06, "loss": 1.2263, "step": 2020 }, { "epoch": 0.16, "learning_rate": 1.91681458976349e-06, "loss": 1.2687, "step": 2021 }, { "epoch": 0.16, "learning_rate": 1.916714178357557e-06, "loss": 1.2904, "step": 2022 }, { "epoch": 0.16, "learning_rate": 1.9166137090191396e-06, "loss": 1.2114, "step": 2023 }, { "epoch": 0.16, "learning_rate": 1.9165131817545874e-06, "loss": 1.2826, "step": 2024 }, { "epoch": 0.16, "learning_rate": 1.916412596570253e-06, "loss": 1.2332, "step": 2025 }, { "epoch": 0.16, "learning_rate": 1.9163119534724926e-06, "loss": 1.2444, "step": 2026 }, { "epoch": 0.16, "learning_rate": 1.9162112524676678e-06, "loss": 1.2215, "step": 2027 }, { "epoch": 0.16, "learning_rate": 1.916110493562141e-06, "loss": 1.2718, "step": 2028 }, { "epoch": 0.16, "learning_rate": 1.916009676762281e-06, "loss": 1.2854, "step": 2029 }, { "epoch": 0.16, "learning_rate": 1.9159088020744577e-06, "loss": 1.274, "step": 2030 }, { "epoch": 0.16, "learning_rate": 1.9158078695050468e-06, "loss": 1.2753, "step": 2031 }, { "epoch": 0.16, "learning_rate": 1.9157068790604267e-06, "loss": 0.8458, "step": 2032 }, { "epoch": 0.16, "learning_rate": 1.9156058307469794e-06, "loss": 1.2091, "step": 2033 }, { "epoch": 0.16, "learning_rate": 1.9155047245710907e-06, "loss": 1.258, "step": 2034 }, { "epoch": 0.16, "learning_rate": 1.91540356053915e-06, "loss": 1.2642, "step": 2035 }, { "epoch": 0.16, "learning_rate": 1.915302338657551e-06, "loss": 1.266, "step": 2036 }, { "epoch": 0.16, "learning_rate": 1.9152010589326906e-06, "loss": 1.2875, "step": 2037 }, { "epoch": 0.16, "learning_rate": 1.915099721370968e-06, "loss": 1.2603, "step": 2038 }, { "epoch": 0.16, "learning_rate": 1.9149983259787885e-06, "loss": 1.268, "step": 2039 }, { "epoch": 0.16, "learning_rate": 1.9148968727625594e-06, "loss": 1.2262, "step": 2040 }, { "epoch": 0.16, "learning_rate": 1.914795361728692e-06, "loss": 1.2514, "step": 2041 }, { "epoch": 0.16, "learning_rate": 1.9146937928836015e-06, "loss": 0.8804, "step": 2042 }, { "epoch": 0.16, "learning_rate": 1.914592166233707e-06, "loss": 1.2308, "step": 2043 }, { "epoch": 0.16, "learning_rate": 1.9144904817854305e-06, "loss": 1.2696, "step": 2044 }, { "epoch": 0.16, "learning_rate": 1.9143887395451976e-06, "loss": 1.248, "step": 2045 }, { "epoch": 0.16, "learning_rate": 1.914286939519439e-06, "loss": 1.2729, "step": 2046 }, { "epoch": 0.16, "learning_rate": 1.9141850817145873e-06, "loss": 1.281, "step": 2047 }, { "epoch": 0.16, "learning_rate": 1.9140831661370797e-06, "loss": 1.2716, "step": 2048 }, { "epoch": 0.16, "learning_rate": 1.9139811927933567e-06, "loss": 1.2228, "step": 2049 }, { "epoch": 0.16, "learning_rate": 1.9138791616898627e-06, "loss": 1.2195, "step": 2050 }, { "epoch": 0.16, "learning_rate": 1.913777072833046e-06, "loss": 1.26, "step": 2051 }, { "epoch": 0.16, "learning_rate": 1.9136749262293573e-06, "loss": 1.3103, "step": 2052 }, { "epoch": 0.16, "learning_rate": 1.9135727218852526e-06, "loss": 1.2416, "step": 2053 }, { "epoch": 0.16, "learning_rate": 1.9134704598071907e-06, "loss": 1.2453, "step": 2054 }, { "epoch": 0.16, "learning_rate": 1.9133681400016337e-06, "loss": 0.8208, "step": 2055 }, { "epoch": 0.16, "learning_rate": 1.9132657624750483e-06, "loss": 1.2512, "step": 2056 }, { "epoch": 0.16, "learning_rate": 1.913163327233904e-06, "loss": 1.2226, "step": 2057 }, { "epoch": 0.16, "learning_rate": 1.913060834284674e-06, "loss": 1.2732, "step": 2058 }, { "epoch": 0.16, "learning_rate": 1.9129582836338366e-06, "loss": 1.3009, "step": 2059 }, { "epoch": 0.16, "learning_rate": 1.912855675287871e-06, "loss": 1.2038, "step": 2060 }, { "epoch": 0.16, "learning_rate": 1.9127530092532627e-06, "loss": 1.286, "step": 2061 }, { "epoch": 0.16, "learning_rate": 1.9126502855365e-06, "loss": 1.2164, "step": 2062 }, { "epoch": 0.16, "learning_rate": 1.9125475041440734e-06, "loss": 1.2456, "step": 2063 }, { "epoch": 0.16, "learning_rate": 1.9124446650824792e-06, "loss": 1.2689, "step": 2064 }, { "epoch": 0.16, "learning_rate": 1.9123417683582155e-06, "loss": 1.2863, "step": 2065 }, { "epoch": 0.16, "learning_rate": 1.912238813977786e-06, "loss": 1.2733, "step": 2066 }, { "epoch": 0.16, "learning_rate": 1.9121358019476965e-06, "loss": 1.2412, "step": 2067 }, { "epoch": 0.16, "learning_rate": 1.912032732274457e-06, "loss": 1.2424, "step": 2068 }, { "epoch": 0.16, "learning_rate": 1.911929604964581e-06, "loss": 1.2821, "step": 2069 }, { "epoch": 0.16, "learning_rate": 1.9118264200245853e-06, "loss": 1.2339, "step": 2070 }, { "epoch": 0.16, "learning_rate": 1.9117231774609917e-06, "loss": 1.2861, "step": 2071 }, { "epoch": 0.16, "learning_rate": 1.911619877280324e-06, "loss": 1.2635, "step": 2072 }, { "epoch": 0.16, "learning_rate": 1.9115165194891104e-06, "loss": 1.2826, "step": 2073 }, { "epoch": 0.16, "learning_rate": 1.911413104093883e-06, "loss": 1.2822, "step": 2074 }, { "epoch": 0.16, "learning_rate": 1.9113096311011774e-06, "loss": 1.253, "step": 2075 }, { "epoch": 0.16, "learning_rate": 1.9112061005175313e-06, "loss": 1.2348, "step": 2076 }, { "epoch": 0.16, "learning_rate": 1.911102512349489e-06, "loss": 0.8222, "step": 2077 }, { "epoch": 0.16, "learning_rate": 1.9109988666035964e-06, "loss": 1.2682, "step": 2078 }, { "epoch": 0.16, "learning_rate": 1.910895163286403e-06, "loss": 1.262, "step": 2079 }, { "epoch": 0.16, "learning_rate": 1.910791402404463e-06, "loss": 1.2345, "step": 2080 }, { "epoch": 0.16, "learning_rate": 1.9106875839643328e-06, "loss": 1.2522, "step": 2081 }, { "epoch": 0.16, "learning_rate": 1.910583707972574e-06, "loss": 1.2673, "step": 2082 }, { "epoch": 0.16, "learning_rate": 1.910479774435751e-06, "loss": 1.2613, "step": 2083 }, { "epoch": 0.16, "learning_rate": 1.9103757833604323e-06, "loss": 0.9091, "step": 2084 }, { "epoch": 0.16, "learning_rate": 1.910271734753189e-06, "loss": 0.8478, "step": 2085 }, { "epoch": 0.16, "learning_rate": 1.9101676286205967e-06, "loss": 1.2166, "step": 2086 }, { "epoch": 0.16, "learning_rate": 1.910063464969235e-06, "loss": 1.2426, "step": 2087 }, { "epoch": 0.16, "learning_rate": 1.909959243805686e-06, "loss": 1.2116, "step": 2088 }, { "epoch": 0.16, "learning_rate": 1.9098549651365363e-06, "loss": 0.8753, "step": 2089 }, { "epoch": 0.16, "learning_rate": 1.909750628968376e-06, "loss": 1.1612, "step": 2090 }, { "epoch": 0.16, "learning_rate": 1.9096462353077986e-06, "loss": 0.8425, "step": 2091 }, { "epoch": 0.16, "learning_rate": 1.909541784161401e-06, "loss": 1.256, "step": 2092 }, { "epoch": 0.16, "learning_rate": 1.9094372755357844e-06, "loss": 1.213, "step": 2093 }, { "epoch": 0.16, "learning_rate": 1.9093327094375535e-06, "loss": 1.2632, "step": 2094 }, { "epoch": 0.16, "learning_rate": 1.909228085873316e-06, "loss": 1.2495, "step": 2095 }, { "epoch": 0.16, "learning_rate": 1.9091234048496844e-06, "loss": 1.2654, "step": 2096 }, { "epoch": 0.16, "learning_rate": 1.909018666373273e-06, "loss": 1.2666, "step": 2097 }, { "epoch": 0.16, "learning_rate": 1.9089138704507018e-06, "loss": 1.2316, "step": 2098 }, { "epoch": 0.16, "learning_rate": 1.9088090170885927e-06, "loss": 1.2307, "step": 2099 }, { "epoch": 0.16, "learning_rate": 1.9087041062935726e-06, "loss": 1.2201, "step": 2100 }, { "epoch": 0.16, "learning_rate": 1.9085991380722713e-06, "loss": 1.289, "step": 2101 }, { "epoch": 0.16, "learning_rate": 1.908494112431322e-06, "loss": 0.8785, "step": 2102 }, { "epoch": 0.16, "learning_rate": 1.908389029377362e-06, "loss": 1.2742, "step": 2103 }, { "epoch": 0.16, "learning_rate": 1.908283888917033e-06, "loss": 1.2296, "step": 2104 }, { "epoch": 0.16, "learning_rate": 1.9081786910569783e-06, "loss": 1.2233, "step": 2105 }, { "epoch": 0.16, "learning_rate": 1.908073435803846e-06, "loss": 1.2449, "step": 2106 }, { "epoch": 0.16, "learning_rate": 1.9079681231642883e-06, "loss": 1.2836, "step": 2107 }, { "epoch": 0.16, "learning_rate": 1.9078627531449602e-06, "loss": 1.2707, "step": 2108 }, { "epoch": 0.16, "learning_rate": 1.907757325752521e-06, "loss": 1.2363, "step": 2109 }, { "epoch": 0.16, "learning_rate": 1.907651840993633e-06, "loss": 1.2838, "step": 2110 }, { "epoch": 0.16, "learning_rate": 1.9075462988749626e-06, "loss": 1.2806, "step": 2111 }, { "epoch": 0.16, "learning_rate": 1.9074406994031793e-06, "loss": 0.8507, "step": 2112 }, { "epoch": 0.16, "learning_rate": 1.9073350425849565e-06, "loss": 1.2667, "step": 2113 }, { "epoch": 0.16, "learning_rate": 1.9072293284269717e-06, "loss": 1.217, "step": 2114 }, { "epoch": 0.16, "learning_rate": 1.9071235569359052e-06, "loss": 1.2014, "step": 2115 }, { "epoch": 0.16, "learning_rate": 1.9070177281184417e-06, "loss": 0.8582, "step": 2116 }, { "epoch": 0.16, "learning_rate": 1.9069118419812686e-06, "loss": 1.2634, "step": 2117 }, { "epoch": 0.16, "learning_rate": 1.9068058985310781e-06, "loss": 0.8398, "step": 2118 }, { "epoch": 0.16, "learning_rate": 1.9066998977745649e-06, "loss": 1.2369, "step": 2119 }, { "epoch": 0.16, "learning_rate": 1.906593839718428e-06, "loss": 1.2361, "step": 2120 }, { "epoch": 0.16, "learning_rate": 1.9064877243693697e-06, "loss": 1.2696, "step": 2121 }, { "epoch": 0.16, "learning_rate": 1.906381551734096e-06, "loss": 1.2361, "step": 2122 }, { "epoch": 0.16, "learning_rate": 1.9062753218193172e-06, "loss": 0.8762, "step": 2123 }, { "epoch": 0.16, "learning_rate": 1.9061690346317455e-06, "loss": 1.2402, "step": 2124 }, { "epoch": 0.16, "learning_rate": 1.9060626901780987e-06, "loss": 1.3018, "step": 2125 }, { "epoch": 0.17, "learning_rate": 1.9059562884650968e-06, "loss": 1.242, "step": 2126 }, { "epoch": 0.17, "learning_rate": 1.9058498294994643e-06, "loss": 1.2906, "step": 2127 }, { "epoch": 0.17, "learning_rate": 1.9057433132879286e-06, "loss": 1.2631, "step": 2128 }, { "epoch": 0.17, "learning_rate": 1.9056367398372215e-06, "loss": 1.2709, "step": 2129 }, { "epoch": 0.17, "learning_rate": 1.9055301091540777e-06, "loss": 0.8722, "step": 2130 }, { "epoch": 0.17, "learning_rate": 1.9054234212452359e-06, "loss": 1.282, "step": 2131 }, { "epoch": 0.17, "learning_rate": 1.905316676117438e-06, "loss": 1.2137, "step": 2132 }, { "epoch": 0.17, "learning_rate": 1.9052098737774305e-06, "loss": 1.2249, "step": 2133 }, { "epoch": 0.17, "learning_rate": 1.9051030142319626e-06, "loss": 1.277, "step": 2134 }, { "epoch": 0.17, "learning_rate": 1.904996097487787e-06, "loss": 1.2522, "step": 2135 }, { "epoch": 0.17, "learning_rate": 1.9048891235516607e-06, "loss": 1.3335, "step": 2136 }, { "epoch": 0.17, "learning_rate": 1.9047820924303442e-06, "loss": 1.1655, "step": 2137 }, { "epoch": 0.17, "learning_rate": 1.9046750041306007e-06, "loss": 1.2475, "step": 2138 }, { "epoch": 0.17, "learning_rate": 1.9045678586591988e-06, "loss": 1.2782, "step": 2139 }, { "epoch": 0.17, "learning_rate": 1.9044606560229086e-06, "loss": 1.2911, "step": 2140 }, { "epoch": 0.17, "learning_rate": 1.9043533962285056e-06, "loss": 1.2667, "step": 2141 }, { "epoch": 0.17, "learning_rate": 1.9042460792827679e-06, "loss": 1.2493, "step": 2142 }, { "epoch": 0.17, "learning_rate": 1.9041387051924772e-06, "loss": 1.2584, "step": 2143 }, { "epoch": 0.17, "learning_rate": 1.9040312739644198e-06, "loss": 1.2779, "step": 2144 }, { "epoch": 0.17, "learning_rate": 1.9039237856053843e-06, "loss": 1.1978, "step": 2145 }, { "epoch": 0.17, "learning_rate": 1.9038162401221633e-06, "loss": 1.1724, "step": 2146 }, { "epoch": 0.17, "learning_rate": 1.9037086375215542e-06, "loss": 1.21, "step": 2147 }, { "epoch": 0.17, "learning_rate": 1.9036009778103561e-06, "loss": 1.2652, "step": 2148 }, { "epoch": 0.17, "learning_rate": 1.903493260995373e-06, "loss": 1.2362, "step": 2149 }, { "epoch": 0.17, "learning_rate": 1.903385487083412e-06, "loss": 1.3186, "step": 2150 }, { "epoch": 0.17, "learning_rate": 1.9032776560812844e-06, "loss": 1.2223, "step": 2151 }, { "epoch": 0.17, "learning_rate": 1.903169767995804e-06, "loss": 1.258, "step": 2152 }, { "epoch": 0.17, "learning_rate": 1.9030618228337896e-06, "loss": 1.2744, "step": 2153 }, { "epoch": 0.17, "learning_rate": 1.9029538206020621e-06, "loss": 1.212, "step": 2154 }, { "epoch": 0.17, "learning_rate": 1.9028457613074474e-06, "loss": 0.8441, "step": 2155 }, { "epoch": 0.17, "learning_rate": 1.9027376449567743e-06, "loss": 0.8875, "step": 2156 }, { "epoch": 0.17, "learning_rate": 1.902629471556875e-06, "loss": 1.3457, "step": 2157 }, { "epoch": 0.17, "learning_rate": 1.9025212411145856e-06, "loss": 1.3193, "step": 2158 }, { "epoch": 0.17, "learning_rate": 1.9024129536367463e-06, "loss": 1.2966, "step": 2159 }, { "epoch": 0.17, "learning_rate": 1.9023046091302e-06, "loss": 1.2147, "step": 2160 }, { "epoch": 0.17, "learning_rate": 1.902196207601794e-06, "loss": 1.1981, "step": 2161 }, { "epoch": 0.17, "learning_rate": 1.9020877490583782e-06, "loss": 1.2355, "step": 2162 }, { "epoch": 0.17, "learning_rate": 1.901979233506807e-06, "loss": 1.2625, "step": 2163 }, { "epoch": 0.17, "learning_rate": 1.9018706609539385e-06, "loss": 1.2683, "step": 2164 }, { "epoch": 0.17, "learning_rate": 1.9017620314066336e-06, "loss": 1.228, "step": 2165 }, { "epoch": 0.17, "learning_rate": 1.9016533448717573e-06, "loss": 1.2164, "step": 2166 }, { "epoch": 0.17, "learning_rate": 1.9015446013561784e-06, "loss": 1.2547, "step": 2167 }, { "epoch": 0.17, "learning_rate": 1.9014358008667686e-06, "loss": 1.2875, "step": 2168 }, { "epoch": 0.17, "learning_rate": 1.9013269434104041e-06, "loss": 1.3044, "step": 2169 }, { "epoch": 0.17, "learning_rate": 1.9012180289939636e-06, "loss": 0.8259, "step": 2170 }, { "epoch": 0.17, "learning_rate": 1.9011090576243309e-06, "loss": 0.8641, "step": 2171 }, { "epoch": 0.17, "learning_rate": 1.9010000293083917e-06, "loss": 1.2732, "step": 2172 }, { "epoch": 0.17, "learning_rate": 1.9008909440530366e-06, "loss": 0.856, "step": 2173 }, { "epoch": 0.17, "learning_rate": 1.9007818018651593e-06, "loss": 1.2895, "step": 2174 }, { "epoch": 0.17, "learning_rate": 1.900672602751657e-06, "loss": 1.2583, "step": 2175 }, { "epoch": 0.17, "learning_rate": 1.9005633467194305e-06, "loss": 1.2517, "step": 2176 }, { "epoch": 0.17, "learning_rate": 1.9004540337753846e-06, "loss": 1.2302, "step": 2177 }, { "epoch": 0.17, "learning_rate": 1.9003446639264276e-06, "loss": 1.2683, "step": 2178 }, { "epoch": 0.17, "learning_rate": 1.9002352371794705e-06, "loss": 0.8593, "step": 2179 }, { "epoch": 0.17, "learning_rate": 1.9001257535414291e-06, "loss": 0.8482, "step": 2180 }, { "epoch": 0.17, "learning_rate": 1.9000162130192226e-06, "loss": 1.2069, "step": 2181 }, { "epoch": 0.17, "learning_rate": 1.899906615619773e-06, "loss": 1.2499, "step": 2182 }, { "epoch": 0.17, "learning_rate": 1.8997969613500064e-06, "loss": 1.2286, "step": 2183 }, { "epoch": 0.17, "learning_rate": 1.8996872502168526e-06, "loss": 1.2703, "step": 2184 }, { "epoch": 0.17, "learning_rate": 1.899577482227245e-06, "loss": 1.2924, "step": 2185 }, { "epoch": 0.17, "learning_rate": 1.8994676573881203e-06, "loss": 1.2902, "step": 2186 }, { "epoch": 0.17, "learning_rate": 1.8993577757064191e-06, "loss": 1.2167, "step": 2187 }, { "epoch": 0.17, "learning_rate": 1.8992478371890854e-06, "loss": 1.2309, "step": 2188 }, { "epoch": 0.17, "learning_rate": 1.8991378418430672e-06, "loss": 1.245, "step": 2189 }, { "epoch": 0.17, "learning_rate": 1.899027789675315e-06, "loss": 1.1618, "step": 2190 }, { "epoch": 0.17, "learning_rate": 1.8989176806927843e-06, "loss": 1.2402, "step": 2191 }, { "epoch": 0.17, "learning_rate": 1.8988075149024332e-06, "loss": 1.2486, "step": 2192 }, { "epoch": 0.17, "learning_rate": 1.8986972923112237e-06, "loss": 1.2254, "step": 2193 }, { "epoch": 0.17, "learning_rate": 1.8985870129261215e-06, "loss": 1.2979, "step": 2194 }, { "epoch": 0.17, "learning_rate": 1.8984766767540959e-06, "loss": 1.2919, "step": 2195 }, { "epoch": 0.17, "learning_rate": 1.8983662838021195e-06, "loss": 1.2514, "step": 2196 }, { "epoch": 0.17, "learning_rate": 1.898255834077169e-06, "loss": 1.2932, "step": 2197 }, { "epoch": 0.17, "learning_rate": 1.8981453275862233e-06, "loss": 1.2674, "step": 2198 }, { "epoch": 0.17, "learning_rate": 1.8980347643362676e-06, "loss": 0.8717, "step": 2199 }, { "epoch": 0.17, "learning_rate": 1.8979241443342877e-06, "loss": 1.2626, "step": 2200 }, { "epoch": 0.17, "learning_rate": 1.897813467587275e-06, "loss": 0.8622, "step": 2201 }, { "epoch": 0.17, "learning_rate": 1.8977027341022236e-06, "loss": 1.2226, "step": 2202 }, { "epoch": 0.17, "learning_rate": 1.8975919438861313e-06, "loss": 1.2287, "step": 2203 }, { "epoch": 0.17, "learning_rate": 1.8974810969459997e-06, "loss": 1.2831, "step": 2204 }, { "epoch": 0.17, "learning_rate": 1.8973701932888338e-06, "loss": 1.2761, "step": 2205 }, { "epoch": 0.17, "learning_rate": 1.8972592329216423e-06, "loss": 1.2953, "step": 2206 }, { "epoch": 0.17, "learning_rate": 1.8971482158514376e-06, "loss": 0.8201, "step": 2207 }, { "epoch": 0.17, "learning_rate": 1.897037142085235e-06, "loss": 1.3128, "step": 2208 }, { "epoch": 0.17, "learning_rate": 1.8969260116300542e-06, "loss": 1.274, "step": 2209 }, { "epoch": 0.17, "learning_rate": 1.8968148244929185e-06, "loss": 0.8692, "step": 2210 }, { "epoch": 0.17, "learning_rate": 1.896703580680854e-06, "loss": 1.272, "step": 2211 }, { "epoch": 0.17, "learning_rate": 1.896592280200891e-06, "loss": 1.2558, "step": 2212 }, { "epoch": 0.17, "learning_rate": 1.8964809230600634e-06, "loss": 0.8562, "step": 2213 }, { "epoch": 0.17, "learning_rate": 1.896369509265408e-06, "loss": 1.2199, "step": 2214 }, { "epoch": 0.17, "learning_rate": 1.8962580388239661e-06, "loss": 1.2846, "step": 2215 }, { "epoch": 0.17, "learning_rate": 1.8961465117427822e-06, "loss": 1.2117, "step": 2216 }, { "epoch": 0.17, "learning_rate": 1.896034928028904e-06, "loss": 1.2685, "step": 2217 }, { "epoch": 0.17, "learning_rate": 1.8959232876893832e-06, "loss": 1.278, "step": 2218 }, { "epoch": 0.17, "learning_rate": 1.8958115907312755e-06, "loss": 1.2516, "step": 2219 }, { "epoch": 0.17, "learning_rate": 1.8956998371616391e-06, "loss": 1.1862, "step": 2220 }, { "epoch": 0.17, "learning_rate": 1.8955880269875365e-06, "loss": 0.8442, "step": 2221 }, { "epoch": 0.17, "learning_rate": 1.8954761602160339e-06, "loss": 1.2417, "step": 2222 }, { "epoch": 0.17, "learning_rate": 1.8953642368542005e-06, "loss": 1.2921, "step": 2223 }, { "epoch": 0.17, "learning_rate": 1.8952522569091096e-06, "loss": 1.2393, "step": 2224 }, { "epoch": 0.17, "learning_rate": 1.8951402203878378e-06, "loss": 1.2543, "step": 2225 }, { "epoch": 0.17, "learning_rate": 1.8950281272974648e-06, "loss": 1.2925, "step": 2226 }, { "epoch": 0.17, "learning_rate": 1.8949159776450754e-06, "loss": 1.2829, "step": 2227 }, { "epoch": 0.17, "learning_rate": 1.8948037714377562e-06, "loss": 0.8687, "step": 2228 }, { "epoch": 0.17, "learning_rate": 1.8946915086825986e-06, "loss": 1.2388, "step": 2229 }, { "epoch": 0.17, "learning_rate": 1.894579189386697e-06, "loss": 1.222, "step": 2230 }, { "epoch": 0.17, "learning_rate": 1.8944668135571495e-06, "loss": 1.1972, "step": 2231 }, { "epoch": 0.17, "learning_rate": 1.8943543812010577e-06, "loss": 1.2406, "step": 2232 }, { "epoch": 0.17, "learning_rate": 1.8942418923255272e-06, "loss": 1.2789, "step": 2233 }, { "epoch": 0.17, "learning_rate": 1.8941293469376663e-06, "loss": 1.2161, "step": 2234 }, { "epoch": 0.17, "learning_rate": 1.8940167450445875e-06, "loss": 1.2158, "step": 2235 }, { "epoch": 0.17, "learning_rate": 1.8939040866534071e-06, "loss": 1.2854, "step": 2236 }, { "epoch": 0.17, "learning_rate": 1.8937913717712446e-06, "loss": 1.2704, "step": 2237 }, { "epoch": 0.17, "learning_rate": 1.8936786004052227e-06, "loss": 1.2974, "step": 2238 }, { "epoch": 0.17, "learning_rate": 1.8935657725624683e-06, "loss": 1.2365, "step": 2239 }, { "epoch": 0.17, "learning_rate": 1.893452888250112e-06, "loss": 0.8648, "step": 2240 }, { "epoch": 0.17, "learning_rate": 1.8933399474752872e-06, "loss": 1.2463, "step": 2241 }, { "epoch": 0.17, "learning_rate": 1.8932269502451312e-06, "loss": 1.2652, "step": 2242 }, { "epoch": 0.17, "learning_rate": 1.8931138965667853e-06, "loss": 1.1855, "step": 2243 }, { "epoch": 0.17, "learning_rate": 1.8930007864473935e-06, "loss": 1.3048, "step": 2244 }, { "epoch": 0.17, "learning_rate": 1.8928876198941045e-06, "loss": 1.2988, "step": 2245 }, { "epoch": 0.17, "learning_rate": 1.8927743969140697e-06, "loss": 1.2223, "step": 2246 }, { "epoch": 0.17, "learning_rate": 1.8926611175144442e-06, "loss": 1.2977, "step": 2247 }, { "epoch": 0.17, "learning_rate": 1.892547781702387e-06, "loss": 1.2713, "step": 2248 }, { "epoch": 0.17, "learning_rate": 1.8924343894850602e-06, "loss": 1.2012, "step": 2249 }, { "epoch": 0.17, "learning_rate": 1.89232094086963e-06, "loss": 1.2536, "step": 2250 }, { "epoch": 0.17, "learning_rate": 1.8922074358632653e-06, "loss": 1.2504, "step": 2251 }, { "epoch": 0.17, "learning_rate": 1.89209387447314e-06, "loss": 1.2832, "step": 2252 }, { "epoch": 0.17, "learning_rate": 1.89198025670643e-06, "loss": 1.2201, "step": 2253 }, { "epoch": 0.17, "learning_rate": 1.8918665825703157e-06, "loss": 1.3118, "step": 2254 }, { "epoch": 0.18, "learning_rate": 1.8917528520719812e-06, "loss": 0.8794, "step": 2255 }, { "epoch": 0.18, "learning_rate": 1.8916390652186132e-06, "loss": 1.2409, "step": 2256 }, { "epoch": 0.18, "learning_rate": 1.8915252220174027e-06, "loss": 1.2445, "step": 2257 }, { "epoch": 0.18, "learning_rate": 1.8914113224755444e-06, "loss": 1.2956, "step": 2258 }, { "epoch": 0.18, "learning_rate": 1.891297366600236e-06, "loss": 1.2991, "step": 2259 }, { "epoch": 0.18, "learning_rate": 1.891183354398679e-06, "loss": 1.2435, "step": 2260 }, { "epoch": 0.18, "learning_rate": 1.8910692858780787e-06, "loss": 1.2488, "step": 2261 }, { "epoch": 0.18, "learning_rate": 1.8909551610456433e-06, "loss": 1.2303, "step": 2262 }, { "epoch": 0.18, "learning_rate": 1.890840979908586e-06, "loss": 1.2757, "step": 2263 }, { "epoch": 0.18, "learning_rate": 1.8907267424741216e-06, "loss": 1.2161, "step": 2264 }, { "epoch": 0.18, "learning_rate": 1.8906124487494694e-06, "loss": 0.8686, "step": 2265 }, { "epoch": 0.18, "learning_rate": 1.8904980987418533e-06, "loss": 1.2572, "step": 2266 }, { "epoch": 0.18, "learning_rate": 1.8903836924584986e-06, "loss": 1.2339, "step": 2267 }, { "epoch": 0.18, "learning_rate": 1.890269229906636e-06, "loss": 0.8533, "step": 2268 }, { "epoch": 0.18, "learning_rate": 1.8901547110934985e-06, "loss": 1.2285, "step": 2269 }, { "epoch": 0.18, "learning_rate": 1.890040136026324e-06, "loss": 0.8933, "step": 2270 }, { "epoch": 0.18, "learning_rate": 1.8899255047123523e-06, "loss": 1.2723, "step": 2271 }, { "epoch": 0.18, "learning_rate": 1.8898108171588283e-06, "loss": 1.2864, "step": 2272 }, { "epoch": 0.18, "learning_rate": 1.8896960733729994e-06, "loss": 1.2904, "step": 2273 }, { "epoch": 0.18, "learning_rate": 1.889581273362117e-06, "loss": 1.226, "step": 2274 }, { "epoch": 0.18, "learning_rate": 1.8894664171334358e-06, "loss": 1.262, "step": 2275 }, { "epoch": 0.18, "learning_rate": 1.8893515046942148e-06, "loss": 1.2576, "step": 2276 }, { "epoch": 0.18, "learning_rate": 1.8892365360517152e-06, "loss": 1.2736, "step": 2277 }, { "epoch": 0.18, "learning_rate": 1.8891215112132031e-06, "loss": 1.2027, "step": 2278 }, { "epoch": 0.18, "learning_rate": 1.8890064301859473e-06, "loss": 1.2475, "step": 2279 }, { "epoch": 0.18, "learning_rate": 1.8888912929772206e-06, "loss": 1.2807, "step": 2280 }, { "epoch": 0.18, "learning_rate": 1.8887760995942993e-06, "loss": 1.282, "step": 2281 }, { "epoch": 0.18, "learning_rate": 1.888660850044463e-06, "loss": 1.2666, "step": 2282 }, { "epoch": 0.18, "learning_rate": 1.888545544334995e-06, "loss": 0.871, "step": 2283 }, { "epoch": 0.18, "learning_rate": 1.8884301824731818e-06, "loss": 1.2719, "step": 2284 }, { "epoch": 0.18, "learning_rate": 1.8883147644663145e-06, "loss": 1.1834, "step": 2285 }, { "epoch": 0.18, "learning_rate": 1.8881992903216863e-06, "loss": 1.2526, "step": 2286 }, { "epoch": 0.18, "learning_rate": 1.8880837600465952e-06, "loss": 1.2948, "step": 2287 }, { "epoch": 0.18, "learning_rate": 1.8879681736483421e-06, "loss": 1.2934, "step": 2288 }, { "epoch": 0.18, "learning_rate": 1.8878525311342312e-06, "loss": 1.2975, "step": 2289 }, { "epoch": 0.18, "learning_rate": 1.8877368325115712e-06, "loss": 0.8297, "step": 2290 }, { "epoch": 0.18, "learning_rate": 1.8876210777876734e-06, "loss": 1.2237, "step": 2291 }, { "epoch": 0.18, "learning_rate": 1.8875052669698533e-06, "loss": 1.2894, "step": 2292 }, { "epoch": 0.18, "learning_rate": 1.887389400065429e-06, "loss": 1.2875, "step": 2293 }, { "epoch": 0.18, "learning_rate": 1.8872734770817238e-06, "loss": 1.2637, "step": 2294 }, { "epoch": 0.18, "learning_rate": 1.8871574980260627e-06, "loss": 1.3073, "step": 2295 }, { "epoch": 0.18, "learning_rate": 1.8870414629057755e-06, "loss": 1.2202, "step": 2296 }, { "epoch": 0.18, "learning_rate": 1.886925371728195e-06, "loss": 1.261, "step": 2297 }, { "epoch": 0.18, "learning_rate": 1.8868092245006575e-06, "loss": 1.2782, "step": 2298 }, { "epoch": 0.18, "learning_rate": 1.8866930212305036e-06, "loss": 1.2709, "step": 2299 }, { "epoch": 0.18, "learning_rate": 1.8865767619250764e-06, "loss": 1.2461, "step": 2300 }, { "epoch": 0.18, "learning_rate": 1.886460446591723e-06, "loss": 1.2451, "step": 2301 }, { "epoch": 0.18, "learning_rate": 1.8863440752377939e-06, "loss": 1.2616, "step": 2302 }, { "epoch": 0.18, "learning_rate": 1.886227647870644e-06, "loss": 1.2163, "step": 2303 }, { "epoch": 0.18, "learning_rate": 1.8861111644976302e-06, "loss": 1.1949, "step": 2304 }, { "epoch": 0.18, "learning_rate": 1.885994625126114e-06, "loss": 1.2589, "step": 2305 }, { "epoch": 0.18, "learning_rate": 1.8858780297634605e-06, "loss": 1.2475, "step": 2306 }, { "epoch": 0.18, "learning_rate": 1.8857613784170378e-06, "loss": 1.2602, "step": 2307 }, { "epoch": 0.18, "learning_rate": 1.8856446710942177e-06, "loss": 1.325, "step": 2308 }, { "epoch": 0.18, "learning_rate": 1.8855279078023758e-06, "loss": 1.2872, "step": 2309 }, { "epoch": 0.18, "learning_rate": 1.8854110885488909e-06, "loss": 1.1843, "step": 2310 }, { "epoch": 0.18, "learning_rate": 1.8852942133411455e-06, "loss": 0.8474, "step": 2311 }, { "epoch": 0.18, "learning_rate": 1.8851772821865257e-06, "loss": 1.2815, "step": 2312 }, { "epoch": 0.18, "learning_rate": 1.8850602950924213e-06, "loss": 0.8403, "step": 2313 }, { "epoch": 0.18, "learning_rate": 1.884943252066225e-06, "loss": 1.2714, "step": 2314 }, { "epoch": 0.18, "learning_rate": 1.8848261531153331e-06, "loss": 1.2114, "step": 2315 }, { "epoch": 0.18, "learning_rate": 1.8847089982471466e-06, "loss": 1.2519, "step": 2316 }, { "epoch": 0.18, "learning_rate": 1.884591787469069e-06, "loss": 1.2789, "step": 2317 }, { "epoch": 0.18, "learning_rate": 1.8844745207885068e-06, "loss": 0.8859, "step": 2318 }, { "epoch": 0.18, "learning_rate": 1.8843571982128715e-06, "loss": 1.27, "step": 2319 }, { "epoch": 0.18, "learning_rate": 1.8842398197495772e-06, "loss": 1.1917, "step": 2320 }, { "epoch": 0.18, "learning_rate": 1.8841223854060416e-06, "loss": 1.283, "step": 2321 }, { "epoch": 0.18, "learning_rate": 1.8840048951896866e-06, "loss": 1.1977, "step": 2322 }, { "epoch": 0.18, "learning_rate": 1.8838873491079364e-06, "loss": 1.2412, "step": 2323 }, { "epoch": 0.18, "learning_rate": 1.8837697471682196e-06, "loss": 1.2485, "step": 2324 }, { "epoch": 0.18, "learning_rate": 1.8836520893779681e-06, "loss": 1.2531, "step": 2325 }, { "epoch": 0.18, "learning_rate": 1.883534375744618e-06, "loss": 1.2342, "step": 2326 }, { "epoch": 0.18, "learning_rate": 1.8834166062756073e-06, "loss": 1.2733, "step": 2327 }, { "epoch": 0.18, "learning_rate": 1.8832987809783792e-06, "loss": 1.2915, "step": 2328 }, { "epoch": 0.18, "learning_rate": 1.8831808998603795e-06, "loss": 1.264, "step": 2329 }, { "epoch": 0.18, "learning_rate": 1.883062962929058e-06, "loss": 1.2602, "step": 2330 }, { "epoch": 0.18, "learning_rate": 1.8829449701918679e-06, "loss": 1.1803, "step": 2331 }, { "epoch": 0.18, "learning_rate": 1.8828269216562656e-06, "loss": 1.2898, "step": 2332 }, { "epoch": 0.18, "learning_rate": 1.882708817329711e-06, "loss": 1.2647, "step": 2333 }, { "epoch": 0.18, "learning_rate": 1.8825906572196683e-06, "loss": 1.2388, "step": 2334 }, { "epoch": 0.18, "learning_rate": 1.8824724413336046e-06, "loss": 1.2296, "step": 2335 }, { "epoch": 0.18, "learning_rate": 1.8823541696789906e-06, "loss": 1.2398, "step": 2336 }, { "epoch": 0.18, "learning_rate": 1.8822358422633008e-06, "loss": 0.839, "step": 2337 }, { "epoch": 0.18, "learning_rate": 1.8821174590940125e-06, "loss": 1.267, "step": 2338 }, { "epoch": 0.18, "learning_rate": 1.8819990201786073e-06, "loss": 1.2831, "step": 2339 }, { "epoch": 0.18, "learning_rate": 1.8818805255245702e-06, "loss": 1.2966, "step": 2340 }, { "epoch": 0.18, "learning_rate": 1.8817619751393892e-06, "loss": 1.2934, "step": 2341 }, { "epoch": 0.18, "learning_rate": 1.8816433690305565e-06, "loss": 0.8548, "step": 2342 }, { "epoch": 0.18, "learning_rate": 1.8815247072055671e-06, "loss": 1.1998, "step": 2343 }, { "epoch": 0.18, "learning_rate": 1.8814059896719204e-06, "loss": 1.2279, "step": 2344 }, { "epoch": 0.18, "learning_rate": 1.8812872164371187e-06, "loss": 1.2815, "step": 2345 }, { "epoch": 0.18, "learning_rate": 1.8811683875086681e-06, "loss": 1.1995, "step": 2346 }, { "epoch": 0.18, "learning_rate": 1.8810495028940779e-06, "loss": 1.2342, "step": 2347 }, { "epoch": 0.18, "learning_rate": 1.880930562600861e-06, "loss": 1.2847, "step": 2348 }, { "epoch": 0.18, "learning_rate": 1.8808115666365341e-06, "loss": 1.1437, "step": 2349 }, { "epoch": 0.18, "learning_rate": 1.880692515008617e-06, "loss": 1.3002, "step": 2350 }, { "epoch": 0.18, "learning_rate": 1.8805734077246335e-06, "loss": 1.2097, "step": 2351 }, { "epoch": 0.18, "learning_rate": 1.8804542447921108e-06, "loss": 1.2823, "step": 2352 }, { "epoch": 0.18, "learning_rate": 1.8803350262185794e-06, "loss": 1.2883, "step": 2353 }, { "epoch": 0.18, "learning_rate": 1.8802157520115732e-06, "loss": 1.2243, "step": 2354 }, { "epoch": 0.18, "learning_rate": 1.88009642217863e-06, "loss": 1.2547, "step": 2355 }, { "epoch": 0.18, "learning_rate": 1.8799770367272912e-06, "loss": 1.2879, "step": 2356 }, { "epoch": 0.18, "learning_rate": 1.8798575956651009e-06, "loss": 1.2863, "step": 2357 }, { "epoch": 0.18, "learning_rate": 1.8797380989996076e-06, "loss": 1.2326, "step": 2358 }, { "epoch": 0.18, "learning_rate": 1.8796185467383628e-06, "loss": 1.2621, "step": 2359 }, { "epoch": 0.18, "learning_rate": 1.879498938888922e-06, "loss": 1.2289, "step": 2360 }, { "epoch": 0.18, "learning_rate": 1.8793792754588437e-06, "loss": 1.2646, "step": 2361 }, { "epoch": 0.18, "learning_rate": 1.8792595564556906e-06, "loss": 1.2791, "step": 2362 }, { "epoch": 0.18, "learning_rate": 1.8791397818870274e-06, "loss": 1.2505, "step": 2363 }, { "epoch": 0.18, "learning_rate": 1.8790199517604244e-06, "loss": 1.2414, "step": 2364 }, { "epoch": 0.18, "learning_rate": 1.8789000660834537e-06, "loss": 1.2273, "step": 2365 }, { "epoch": 0.18, "learning_rate": 1.878780124863692e-06, "loss": 1.268, "step": 2366 }, { "epoch": 0.18, "learning_rate": 1.8786601281087187e-06, "loss": 1.2441, "step": 2367 }, { "epoch": 0.18, "learning_rate": 1.8785400758261175e-06, "loss": 1.2103, "step": 2368 }, { "epoch": 0.18, "learning_rate": 1.8784199680234748e-06, "loss": 1.2651, "step": 2369 }, { "epoch": 0.18, "learning_rate": 1.8782998047083809e-06, "loss": 1.2876, "step": 2370 }, { "epoch": 0.18, "learning_rate": 1.8781795858884302e-06, "loss": 1.246, "step": 2371 }, { "epoch": 0.18, "learning_rate": 1.8780593115712193e-06, "loss": 1.2059, "step": 2372 }, { "epoch": 0.18, "learning_rate": 1.8779389817643497e-06, "loss": 1.2785, "step": 2373 }, { "epoch": 0.18, "learning_rate": 1.8778185964754251e-06, "loss": 1.2802, "step": 2374 }, { "epoch": 0.18, "learning_rate": 1.8776981557120538e-06, "loss": 1.1966, "step": 2375 }, { "epoch": 0.18, "learning_rate": 1.877577659481847e-06, "loss": 1.2794, "step": 2376 }, { "epoch": 0.18, "learning_rate": 1.8774571077924194e-06, "loss": 1.2454, "step": 2377 }, { "epoch": 0.18, "learning_rate": 1.8773365006513898e-06, "loss": 0.8389, "step": 2378 }, { "epoch": 0.18, "learning_rate": 1.8772158380663795e-06, "loss": 1.2031, "step": 2379 }, { "epoch": 0.18, "learning_rate": 1.877095120045014e-06, "loss": 1.2896, "step": 2380 }, { "epoch": 0.18, "learning_rate": 1.8769743465949226e-06, "loss": 1.2494, "step": 2381 }, { "epoch": 0.18, "learning_rate": 1.8768535177237372e-06, "loss": 1.2258, "step": 2382 }, { "epoch": 0.18, "learning_rate": 1.8767326334390941e-06, "loss": 1.2639, "step": 2383 }, { "epoch": 0.19, "learning_rate": 1.8766116937486322e-06, "loss": 1.2829, "step": 2384 }, { "epoch": 0.19, "learning_rate": 1.8764906986599946e-06, "loss": 1.2406, "step": 2385 }, { "epoch": 0.19, "learning_rate": 1.876369648180828e-06, "loss": 1.1916, "step": 2386 }, { "epoch": 0.19, "learning_rate": 1.8762485423187817e-06, "loss": 1.2763, "step": 2387 }, { "epoch": 0.19, "learning_rate": 1.8761273810815096e-06, "loss": 1.2772, "step": 2388 }, { "epoch": 0.19, "learning_rate": 1.8760061644766683e-06, "loss": 1.2188, "step": 2389 }, { "epoch": 0.19, "learning_rate": 1.875884892511918e-06, "loss": 0.8327, "step": 2390 }, { "epoch": 0.19, "learning_rate": 1.875763565194923e-06, "loss": 1.2771, "step": 2391 }, { "epoch": 0.19, "learning_rate": 1.8756421825333506e-06, "loss": 1.2353, "step": 2392 }, { "epoch": 0.19, "learning_rate": 1.8755207445348714e-06, "loss": 1.2982, "step": 2393 }, { "epoch": 0.19, "learning_rate": 1.8753992512071597e-06, "loss": 0.8366, "step": 2394 }, { "epoch": 0.19, "learning_rate": 1.875277702557894e-06, "loss": 1.2286, "step": 2395 }, { "epoch": 0.19, "learning_rate": 1.875156098594755e-06, "loss": 1.3072, "step": 2396 }, { "epoch": 0.19, "learning_rate": 1.875034439325428e-06, "loss": 1.2593, "step": 2397 }, { "epoch": 0.19, "learning_rate": 1.8749127247576011e-06, "loss": 1.2507, "step": 2398 }, { "epoch": 0.19, "learning_rate": 1.8747909548989663e-06, "loss": 1.2291, "step": 2399 }, { "epoch": 0.19, "learning_rate": 1.8746691297572186e-06, "loss": 1.2195, "step": 2400 }, { "epoch": 0.19, "learning_rate": 1.8745472493400575e-06, "loss": 1.2823, "step": 2401 }, { "epoch": 0.19, "learning_rate": 1.8744253136551845e-06, "loss": 1.3144, "step": 2402 }, { "epoch": 0.19, "learning_rate": 1.8743033227103061e-06, "loss": 1.1425, "step": 2403 }, { "epoch": 0.19, "learning_rate": 1.8741812765131313e-06, "loss": 0.8762, "step": 2404 }, { "epoch": 0.19, "learning_rate": 1.874059175071373e-06, "loss": 1.2406, "step": 2405 }, { "epoch": 0.19, "learning_rate": 1.8739370183927474e-06, "loss": 0.8724, "step": 2406 }, { "epoch": 0.19, "learning_rate": 1.8738148064849745e-06, "loss": 0.8442, "step": 2407 }, { "epoch": 0.19, "learning_rate": 1.8736925393557774e-06, "loss": 1.2661, "step": 2408 }, { "epoch": 0.19, "learning_rate": 1.8735702170128832e-06, "loss": 1.231, "step": 2409 }, { "epoch": 0.19, "learning_rate": 1.8734478394640213e-06, "loss": 0.9003, "step": 2410 }, { "epoch": 0.19, "learning_rate": 1.8733254067169263e-06, "loss": 1.2556, "step": 2411 }, { "epoch": 0.19, "learning_rate": 1.8732029187793352e-06, "loss": 1.2117, "step": 2412 }, { "epoch": 0.19, "learning_rate": 1.8730803756589886e-06, "loss": 1.2388, "step": 2413 }, { "epoch": 0.19, "learning_rate": 1.872957777363631e-06, "loss": 0.828, "step": 2414 }, { "epoch": 0.19, "learning_rate": 1.8728351239010096e-06, "loss": 1.1724, "step": 2415 }, { "epoch": 0.19, "learning_rate": 1.872712415278876e-06, "loss": 1.2714, "step": 2416 }, { "epoch": 0.19, "learning_rate": 1.8725896515049847e-06, "loss": 1.1769, "step": 2417 }, { "epoch": 0.19, "learning_rate": 1.872466832587094e-06, "loss": 1.2145, "step": 2418 }, { "epoch": 0.19, "learning_rate": 1.8723439585329653e-06, "loss": 1.2478, "step": 2419 }, { "epoch": 0.19, "learning_rate": 1.872221029350364e-06, "loss": 1.2676, "step": 2420 }, { "epoch": 0.19, "learning_rate": 1.8720980450470582e-06, "loss": 1.2843, "step": 2421 }, { "epoch": 0.19, "learning_rate": 1.8719750056308207e-06, "loss": 0.8363, "step": 2422 }, { "epoch": 0.19, "learning_rate": 1.8718519111094269e-06, "loss": 1.2747, "step": 2423 }, { "epoch": 0.19, "learning_rate": 1.8717287614906553e-06, "loss": 1.2525, "step": 2424 }, { "epoch": 0.19, "learning_rate": 1.8716055567822888e-06, "loss": 1.2468, "step": 2425 }, { "epoch": 0.19, "learning_rate": 1.8714822969921137e-06, "loss": 1.248, "step": 2426 }, { "epoch": 0.19, "learning_rate": 1.8713589821279192e-06, "loss": 1.2579, "step": 2427 }, { "epoch": 0.19, "learning_rate": 1.8712356121974982e-06, "loss": 1.2021, "step": 2428 }, { "epoch": 0.19, "learning_rate": 1.8711121872086473e-06, "loss": 1.2294, "step": 2429 }, { "epoch": 0.19, "learning_rate": 1.8709887071691664e-06, "loss": 1.2069, "step": 2430 }, { "epoch": 0.19, "learning_rate": 1.870865172086859e-06, "loss": 1.218, "step": 2431 }, { "epoch": 0.19, "learning_rate": 1.8707415819695322e-06, "loss": 1.1968, "step": 2432 }, { "epoch": 0.19, "learning_rate": 1.8706179368249959e-06, "loss": 1.2364, "step": 2433 }, { "epoch": 0.19, "learning_rate": 1.870494236661064e-06, "loss": 0.8301, "step": 2434 }, { "epoch": 0.19, "learning_rate": 1.8703704814855543e-06, "loss": 1.2182, "step": 2435 }, { "epoch": 0.19, "learning_rate": 1.8702466713062872e-06, "loss": 1.1941, "step": 2436 }, { "epoch": 0.19, "learning_rate": 1.870122806131087e-06, "loss": 0.8451, "step": 2437 }, { "epoch": 0.19, "learning_rate": 1.8699988859677815e-06, "loss": 1.2059, "step": 2438 }, { "epoch": 0.19, "learning_rate": 1.8698749108242022e-06, "loss": 1.2351, "step": 2439 }, { "epoch": 0.19, "learning_rate": 1.8697508807081835e-06, "loss": 1.2925, "step": 2440 }, { "epoch": 0.19, "learning_rate": 1.8696267956275637e-06, "loss": 1.2622, "step": 2441 }, { "epoch": 0.19, "learning_rate": 1.8695026555901844e-06, "loss": 1.2736, "step": 2442 }, { "epoch": 0.19, "learning_rate": 1.869378460603891e-06, "loss": 1.2317, "step": 2443 }, { "epoch": 0.19, "learning_rate": 1.8692542106765315e-06, "loss": 1.2933, "step": 2444 }, { "epoch": 0.19, "learning_rate": 1.8691299058159587e-06, "loss": 1.2764, "step": 2445 }, { "epoch": 0.19, "learning_rate": 1.8690055460300273e-06, "loss": 1.2808, "step": 2446 }, { "epoch": 0.19, "learning_rate": 1.8688811313265971e-06, "loss": 1.2285, "step": 2447 }, { "epoch": 0.19, "learning_rate": 1.8687566617135304e-06, "loss": 1.2432, "step": 2448 }, { "epoch": 0.19, "learning_rate": 1.8686321371986927e-06, "loss": 0.887, "step": 2449 }, { "epoch": 0.19, "learning_rate": 1.8685075577899541e-06, "loss": 1.2088, "step": 2450 }, { "epoch": 0.19, "learning_rate": 1.8683829234951869e-06, "loss": 1.2547, "step": 2451 }, { "epoch": 0.19, "learning_rate": 1.8682582343222682e-06, "loss": 1.203, "step": 2452 }, { "epoch": 0.19, "learning_rate": 1.868133490279077e-06, "loss": 0.8588, "step": 2453 }, { "epoch": 0.19, "learning_rate": 1.8680086913734967e-06, "loss": 1.2735, "step": 2454 }, { "epoch": 0.19, "learning_rate": 1.8678838376134146e-06, "loss": 1.2245, "step": 2455 }, { "epoch": 0.19, "learning_rate": 1.8677589290067207e-06, "loss": 1.209, "step": 2456 }, { "epoch": 0.19, "learning_rate": 1.8676339655613086e-06, "loss": 1.2626, "step": 2457 }, { "epoch": 0.19, "learning_rate": 1.8675089472850756e-06, "loss": 1.2325, "step": 2458 }, { "epoch": 0.19, "learning_rate": 1.867383874185922e-06, "loss": 1.2409, "step": 2459 }, { "epoch": 0.19, "learning_rate": 1.8672587462717525e-06, "loss": 1.2162, "step": 2460 }, { "epoch": 0.19, "learning_rate": 1.8671335635504742e-06, "loss": 1.2678, "step": 2461 }, { "epoch": 0.19, "learning_rate": 1.867008326029998e-06, "loss": 1.2021, "step": 2462 }, { "epoch": 0.19, "learning_rate": 1.8668830337182389e-06, "loss": 1.2219, "step": 2463 }, { "epoch": 0.19, "learning_rate": 1.8667576866231144e-06, "loss": 1.2496, "step": 2464 }, { "epoch": 0.19, "learning_rate": 1.866632284752546e-06, "loss": 1.2689, "step": 2465 }, { "epoch": 0.19, "learning_rate": 1.8665068281144588e-06, "loss": 1.2686, "step": 2466 }, { "epoch": 0.19, "learning_rate": 1.8663813167167807e-06, "loss": 1.2255, "step": 2467 }, { "epoch": 0.19, "learning_rate": 1.866255750567444e-06, "loss": 1.293, "step": 2468 }, { "epoch": 0.19, "learning_rate": 1.8661301296743836e-06, "loss": 1.242, "step": 2469 }, { "epoch": 0.19, "learning_rate": 1.8660044540455387e-06, "loss": 1.3054, "step": 2470 }, { "epoch": 0.19, "learning_rate": 1.8658787236888506e-06, "loss": 1.1731, "step": 2471 }, { "epoch": 0.19, "learning_rate": 1.8657529386122656e-06, "loss": 1.2609, "step": 2472 }, { "epoch": 0.19, "learning_rate": 1.8656270988237328e-06, "loss": 1.238, "step": 2473 }, { "epoch": 0.19, "learning_rate": 1.8655012043312044e-06, "loss": 1.2091, "step": 2474 }, { "epoch": 0.19, "learning_rate": 1.8653752551426364e-06, "loss": 1.3234, "step": 2475 }, { "epoch": 0.19, "learning_rate": 1.8652492512659887e-06, "loss": 1.1694, "step": 2476 }, { "epoch": 0.19, "learning_rate": 1.8651231927092239e-06, "loss": 1.2366, "step": 2477 }, { "epoch": 0.19, "learning_rate": 1.8649970794803084e-06, "loss": 1.2872, "step": 2478 }, { "epoch": 0.19, "learning_rate": 1.8648709115872121e-06, "loss": 1.2397, "step": 2479 }, { "epoch": 0.19, "learning_rate": 1.8647446890379082e-06, "loss": 0.8745, "step": 2480 }, { "epoch": 0.19, "learning_rate": 1.8646184118403735e-06, "loss": 1.2504, "step": 2481 }, { "epoch": 0.19, "learning_rate": 1.8644920800025879e-06, "loss": 0.8189, "step": 2482 }, { "epoch": 0.19, "learning_rate": 1.8643656935325354e-06, "loss": 0.8852, "step": 2483 }, { "epoch": 0.19, "learning_rate": 1.8642392524382031e-06, "loss": 1.2284, "step": 2484 }, { "epoch": 0.19, "learning_rate": 1.8641127567275812e-06, "loss": 1.2281, "step": 2485 }, { "epoch": 0.19, "learning_rate": 1.863986206408664e-06, "loss": 1.2751, "step": 2486 }, { "epoch": 0.19, "learning_rate": 1.863859601489449e-06, "loss": 1.2389, "step": 2487 }, { "epoch": 0.19, "learning_rate": 1.863732941977937e-06, "loss": 0.9157, "step": 2488 }, { "epoch": 0.19, "learning_rate": 1.8636062278821322e-06, "loss": 1.2645, "step": 2489 }, { "epoch": 0.19, "learning_rate": 1.8634794592100425e-06, "loss": 1.2622, "step": 2490 }, { "epoch": 0.19, "learning_rate": 1.8633526359696791e-06, "loss": 0.8645, "step": 2491 }, { "epoch": 0.19, "learning_rate": 1.863225758169057e-06, "loss": 1.2873, "step": 2492 }, { "epoch": 0.19, "learning_rate": 1.8630988258161938e-06, "loss": 1.238, "step": 2493 }, { "epoch": 0.19, "learning_rate": 1.8629718389191118e-06, "loss": 1.2161, "step": 2494 }, { "epoch": 0.19, "learning_rate": 1.8628447974858354e-06, "loss": 1.2345, "step": 2495 }, { "epoch": 0.19, "learning_rate": 1.8627177015243931e-06, "loss": 1.2522, "step": 2496 }, { "epoch": 0.19, "learning_rate": 1.8625905510428174e-06, "loss": 1.1995, "step": 2497 }, { "epoch": 0.19, "learning_rate": 1.8624633460491431e-06, "loss": 1.2699, "step": 2498 }, { "epoch": 0.19, "learning_rate": 1.8623360865514095e-06, "loss": 1.2285, "step": 2499 }, { "epoch": 0.19, "learning_rate": 1.8622087725576585e-06, "loss": 1.2718, "step": 2500 }, { "epoch": 0.19, "learning_rate": 1.8620814040759359e-06, "loss": 1.2285, "step": 2501 }, { "epoch": 0.19, "learning_rate": 1.8619539811142907e-06, "loss": 0.8572, "step": 2502 }, { "epoch": 0.19, "learning_rate": 1.8618265036807758e-06, "loss": 1.2228, "step": 2503 }, { "epoch": 0.19, "learning_rate": 1.8616989717834473e-06, "loss": 0.858, "step": 2504 }, { "epoch": 0.19, "learning_rate": 1.8615713854303643e-06, "loss": 1.2708, "step": 2505 }, { "epoch": 0.19, "learning_rate": 1.8614437446295898e-06, "loss": 1.2707, "step": 2506 }, { "epoch": 0.19, "learning_rate": 1.8613160493891904e-06, "loss": 1.2426, "step": 2507 }, { "epoch": 0.19, "learning_rate": 1.8611882997172358e-06, "loss": 1.2299, "step": 2508 }, { "epoch": 0.19, "learning_rate": 1.8610604956217993e-06, "loss": 1.2517, "step": 2509 }, { "epoch": 0.19, "learning_rate": 1.8609326371109573e-06, "loss": 1.2682, "step": 2510 }, { "epoch": 0.19, "learning_rate": 1.8608047241927904e-06, "loss": 1.2753, "step": 2511 }, { "epoch": 0.19, "learning_rate": 1.8606767568753817e-06, "loss": 1.233, "step": 2512 }, { "epoch": 0.2, "learning_rate": 1.8605487351668184e-06, "loss": 1.2125, "step": 2513 }, { "epoch": 0.2, "learning_rate": 1.8604206590751908e-06, "loss": 0.8804, "step": 2514 }, { "epoch": 0.2, "learning_rate": 1.8602925286085932e-06, "loss": 1.2743, "step": 2515 }, { "epoch": 0.2, "learning_rate": 1.8601643437751223e-06, "loss": 1.2317, "step": 2516 }, { "epoch": 0.2, "learning_rate": 1.8600361045828792e-06, "loss": 1.2451, "step": 2517 }, { "epoch": 0.2, "learning_rate": 1.859907811039968e-06, "loss": 1.2635, "step": 2518 }, { "epoch": 0.2, "learning_rate": 1.8597794631544964e-06, "loss": 1.2231, "step": 2519 }, { "epoch": 0.2, "learning_rate": 1.859651060934575e-06, "loss": 1.2275, "step": 2520 }, { "epoch": 0.2, "learning_rate": 1.859522604388319e-06, "loss": 1.2816, "step": 2521 }, { "epoch": 0.2, "learning_rate": 1.8593940935238458e-06, "loss": 1.2494, "step": 2522 }, { "epoch": 0.2, "learning_rate": 1.8592655283492772e-06, "loss": 1.2338, "step": 2523 }, { "epoch": 0.2, "learning_rate": 1.8591369088727376e-06, "loss": 1.266, "step": 2524 }, { "epoch": 0.2, "learning_rate": 1.859008235102355e-06, "loss": 1.2735, "step": 2525 }, { "epoch": 0.2, "learning_rate": 1.8588795070462618e-06, "loss": 1.2083, "step": 2526 }, { "epoch": 0.2, "learning_rate": 1.8587507247125922e-06, "loss": 1.2285, "step": 2527 }, { "epoch": 0.2, "learning_rate": 1.8586218881094852e-06, "loss": 1.2579, "step": 2528 }, { "epoch": 0.2, "learning_rate": 1.8584929972450828e-06, "loss": 1.2502, "step": 2529 }, { "epoch": 0.2, "learning_rate": 1.8583640521275302e-06, "loss": 1.2975, "step": 2530 }, { "epoch": 0.2, "learning_rate": 1.8582350527649761e-06, "loss": 1.1716, "step": 2531 }, { "epoch": 0.2, "learning_rate": 1.858105999165573e-06, "loss": 1.2486, "step": 2532 }, { "epoch": 0.2, "learning_rate": 1.857976891337476e-06, "loss": 1.2841, "step": 2533 }, { "epoch": 0.2, "learning_rate": 1.8578477292888448e-06, "loss": 1.2261, "step": 2534 }, { "epoch": 0.2, "learning_rate": 1.8577185130278417e-06, "loss": 1.247, "step": 2535 }, { "epoch": 0.2, "learning_rate": 1.8575892425626325e-06, "loss": 1.2314, "step": 2536 }, { "epoch": 0.2, "learning_rate": 1.8574599179013866e-06, "loss": 0.874, "step": 2537 }, { "epoch": 0.2, "learning_rate": 1.8573305390522767e-06, "loss": 1.2784, "step": 2538 }, { "epoch": 0.2, "learning_rate": 1.857201106023479e-06, "loss": 1.2437, "step": 2539 }, { "epoch": 0.2, "learning_rate": 1.8570716188231734e-06, "loss": 1.2389, "step": 2540 }, { "epoch": 0.2, "learning_rate": 1.8569420774595427e-06, "loss": 1.2109, "step": 2541 }, { "epoch": 0.2, "learning_rate": 1.8568124819407735e-06, "loss": 1.2144, "step": 2542 }, { "epoch": 0.2, "learning_rate": 1.8566828322750554e-06, "loss": 0.9052, "step": 2543 }, { "epoch": 0.2, "learning_rate": 1.856553128470582e-06, "loss": 1.2591, "step": 2544 }, { "epoch": 0.2, "learning_rate": 1.8564233705355501e-06, "loss": 1.2587, "step": 2545 }, { "epoch": 0.2, "learning_rate": 1.8562935584781597e-06, "loss": 1.2757, "step": 2546 }, { "epoch": 0.2, "learning_rate": 1.8561636923066142e-06, "loss": 1.2594, "step": 2547 }, { "epoch": 0.2, "learning_rate": 1.856033772029121e-06, "loss": 1.2251, "step": 2548 }, { "epoch": 0.2, "learning_rate": 1.85590379765389e-06, "loss": 1.2518, "step": 2549 }, { "epoch": 0.2, "learning_rate": 1.8557737691891358e-06, "loss": 1.193, "step": 2550 }, { "epoch": 0.2, "learning_rate": 1.855643686643075e-06, "loss": 1.2579, "step": 2551 }, { "epoch": 0.2, "learning_rate": 1.8555135500239283e-06, "loss": 1.251, "step": 2552 }, { "epoch": 0.2, "learning_rate": 1.8553833593399202e-06, "loss": 0.8625, "step": 2553 }, { "epoch": 0.2, "learning_rate": 1.8552531145992777e-06, "loss": 1.2358, "step": 2554 }, { "epoch": 0.2, "learning_rate": 1.8551228158102324e-06, "loss": 1.2033, "step": 2555 }, { "epoch": 0.2, "learning_rate": 1.8549924629810179e-06, "loss": 1.2438, "step": 2556 }, { "epoch": 0.2, "learning_rate": 1.8548620561198723e-06, "loss": 1.2676, "step": 2557 }, { "epoch": 0.2, "learning_rate": 1.8547315952350368e-06, "loss": 1.2599, "step": 2558 }, { "epoch": 0.2, "learning_rate": 1.8546010803347558e-06, "loss": 1.2734, "step": 2559 }, { "epoch": 0.2, "learning_rate": 1.8544705114272774e-06, "loss": 1.3108, "step": 2560 }, { "epoch": 0.2, "learning_rate": 1.8543398885208533e-06, "loss": 1.1941, "step": 2561 }, { "epoch": 0.2, "learning_rate": 1.8542092116237375e-06, "loss": 1.2256, "step": 2562 }, { "epoch": 0.2, "learning_rate": 1.8540784807441892e-06, "loss": 1.2833, "step": 2563 }, { "epoch": 0.2, "learning_rate": 1.8539476958904695e-06, "loss": 1.2576, "step": 2564 }, { "epoch": 0.2, "learning_rate": 1.8538168570708437e-06, "loss": 1.2516, "step": 2565 }, { "epoch": 0.2, "learning_rate": 1.8536859642935798e-06, "loss": 1.2399, "step": 2566 }, { "epoch": 0.2, "learning_rate": 1.8535550175669502e-06, "loss": 1.2333, "step": 2567 }, { "epoch": 0.2, "learning_rate": 1.8534240168992299e-06, "loss": 1.3017, "step": 2568 }, { "epoch": 0.2, "learning_rate": 1.8532929622986977e-06, "loss": 1.2158, "step": 2569 }, { "epoch": 0.2, "learning_rate": 1.8531618537736357e-06, "loss": 1.328, "step": 2570 }, { "epoch": 0.2, "learning_rate": 1.8530306913323293e-06, "loss": 0.8668, "step": 2571 }, { "epoch": 0.2, "learning_rate": 1.8528994749830674e-06, "loss": 1.2651, "step": 2572 }, { "epoch": 0.2, "learning_rate": 1.8527682047341427e-06, "loss": 1.2914, "step": 2573 }, { "epoch": 0.2, "learning_rate": 1.8526368805938504e-06, "loss": 1.244, "step": 2574 }, { "epoch": 0.2, "learning_rate": 1.85250550257049e-06, "loss": 1.2308, "step": 2575 }, { "epoch": 0.2, "learning_rate": 1.8523740706723635e-06, "loss": 1.2427, "step": 2576 }, { "epoch": 0.2, "learning_rate": 1.852242584907778e-06, "loss": 1.252, "step": 2577 }, { "epoch": 0.2, "learning_rate": 1.8521110452850418e-06, "loss": 1.2218, "step": 2578 }, { "epoch": 0.2, "learning_rate": 1.8519794518124678e-06, "loss": 1.2411, "step": 2579 }, { "epoch": 0.2, "learning_rate": 1.8518478044983726e-06, "loss": 0.8266, "step": 2580 }, { "epoch": 0.2, "learning_rate": 1.851716103351075e-06, "loss": 1.2773, "step": 2581 }, { "epoch": 0.2, "learning_rate": 1.851584348378899e-06, "loss": 1.2644, "step": 2582 }, { "epoch": 0.2, "learning_rate": 1.8514525395901699e-06, "loss": 1.2241, "step": 2583 }, { "epoch": 0.2, "learning_rate": 1.8513206769932181e-06, "loss": 1.2096, "step": 2584 }, { "epoch": 0.2, "learning_rate": 1.8511887605963765e-06, "loss": 1.2925, "step": 2585 }, { "epoch": 0.2, "learning_rate": 1.851056790407982e-06, "loss": 1.2612, "step": 2586 }, { "epoch": 0.2, "learning_rate": 1.8509247664363742e-06, "loss": 0.8639, "step": 2587 }, { "epoch": 0.2, "learning_rate": 1.8507926886898968e-06, "loss": 1.2638, "step": 2588 }, { "epoch": 0.2, "learning_rate": 1.8506605571768961e-06, "loss": 1.2689, "step": 2589 }, { "epoch": 0.2, "learning_rate": 1.8505283719057228e-06, "loss": 1.2259, "step": 2590 }, { "epoch": 0.2, "learning_rate": 1.85039613288473e-06, "loss": 1.2714, "step": 2591 }, { "epoch": 0.2, "learning_rate": 1.8502638401222747e-06, "loss": 1.2321, "step": 2592 }, { "epoch": 0.2, "learning_rate": 1.8501314936267174e-06, "loss": 1.1849, "step": 2593 }, { "epoch": 0.2, "learning_rate": 1.849999093406422e-06, "loss": 1.2714, "step": 2594 }, { "epoch": 0.2, "learning_rate": 1.8498666394697552e-06, "loss": 1.2624, "step": 2595 }, { "epoch": 0.2, "learning_rate": 1.849734131825088e-06, "loss": 1.2261, "step": 2596 }, { "epoch": 0.2, "learning_rate": 1.849601570480794e-06, "loss": 1.2795, "step": 2597 }, { "epoch": 0.2, "learning_rate": 1.8494689554452505e-06, "loss": 1.2444, "step": 2598 }, { "epoch": 0.2, "learning_rate": 1.8493362867268385e-06, "loss": 1.3041, "step": 2599 }, { "epoch": 0.2, "learning_rate": 1.849203564333942e-06, "loss": 1.1922, "step": 2600 }, { "epoch": 0.2, "learning_rate": 1.849070788274948e-06, "loss": 1.2192, "step": 2601 }, { "epoch": 0.2, "learning_rate": 1.8489379585582485e-06, "loss": 0.868, "step": 2602 }, { "epoch": 0.2, "learning_rate": 1.8488050751922365e-06, "loss": 1.2465, "step": 2603 }, { "epoch": 0.2, "learning_rate": 1.8486721381853108e-06, "loss": 1.2562, "step": 2604 }, { "epoch": 0.2, "learning_rate": 1.8485391475458716e-06, "loss": 0.8473, "step": 2605 }, { "epoch": 0.2, "learning_rate": 1.8484061032823237e-06, "loss": 1.257, "step": 2606 }, { "epoch": 0.2, "learning_rate": 1.848273005403075e-06, "loss": 1.2512, "step": 2607 }, { "epoch": 0.2, "learning_rate": 1.8481398539165366e-06, "loss": 1.2669, "step": 2608 }, { "epoch": 0.2, "learning_rate": 1.8480066488311232e-06, "loss": 0.8355, "step": 2609 }, { "epoch": 0.2, "learning_rate": 1.8478733901552528e-06, "loss": 1.2766, "step": 2610 }, { "epoch": 0.2, "learning_rate": 1.8477400778973466e-06, "loss": 1.2273, "step": 2611 }, { "epoch": 0.2, "learning_rate": 1.8476067120658296e-06, "loss": 1.2504, "step": 2612 }, { "epoch": 0.2, "learning_rate": 1.8474732926691302e-06, "loss": 1.175, "step": 2613 }, { "epoch": 0.2, "learning_rate": 1.847339819715679e-06, "loss": 1.2092, "step": 2614 }, { "epoch": 0.2, "learning_rate": 1.8472062932139122e-06, "loss": 1.2264, "step": 2615 }, { "epoch": 0.2, "learning_rate": 1.847072713172267e-06, "loss": 1.2257, "step": 2616 }, { "epoch": 0.2, "learning_rate": 1.8469390795991859e-06, "loss": 0.8547, "step": 2617 }, { "epoch": 0.2, "learning_rate": 1.8468053925031136e-06, "loss": 1.2503, "step": 2618 }, { "epoch": 0.2, "learning_rate": 1.8466716518924985e-06, "loss": 1.2602, "step": 2619 }, { "epoch": 0.2, "learning_rate": 1.8465378577757926e-06, "loss": 1.2279, "step": 2620 }, { "epoch": 0.2, "learning_rate": 1.8464040101614512e-06, "loss": 1.2786, "step": 2621 }, { "epoch": 0.2, "learning_rate": 1.846270109057933e-06, "loss": 1.2202, "step": 2622 }, { "epoch": 0.2, "learning_rate": 1.846136154473699e-06, "loss": 1.2474, "step": 2623 }, { "epoch": 0.2, "learning_rate": 1.846002146417216e-06, "loss": 1.2563, "step": 2624 }, { "epoch": 0.2, "learning_rate": 1.8458680848969518e-06, "loss": 1.248, "step": 2625 }, { "epoch": 0.2, "learning_rate": 1.8457339699213792e-06, "loss": 1.2533, "step": 2626 }, { "epoch": 0.2, "learning_rate": 1.8455998014989731e-06, "loss": 1.2531, "step": 2627 }, { "epoch": 0.2, "learning_rate": 1.8454655796382126e-06, "loss": 1.2432, "step": 2628 }, { "epoch": 0.2, "learning_rate": 1.84533130434758e-06, "loss": 1.2129, "step": 2629 }, { "epoch": 0.2, "learning_rate": 1.8451969756355607e-06, "loss": 1.2391, "step": 2630 }, { "epoch": 0.2, "learning_rate": 1.845062593510644e-06, "loss": 1.2397, "step": 2631 }, { "epoch": 0.2, "learning_rate": 1.8449281579813225e-06, "loss": 1.2615, "step": 2632 }, { "epoch": 0.2, "learning_rate": 1.8447936690560915e-06, "loss": 1.2862, "step": 2633 }, { "epoch": 0.2, "learning_rate": 1.8446591267434502e-06, "loss": 1.2164, "step": 2634 }, { "epoch": 0.2, "learning_rate": 1.8445245310519013e-06, "loss": 1.2985, "step": 2635 }, { "epoch": 0.2, "learning_rate": 1.8443898819899504e-06, "loss": 1.2059, "step": 2636 }, { "epoch": 0.2, "learning_rate": 1.844255179566107e-06, "loss": 1.2342, "step": 2637 }, { "epoch": 0.2, "learning_rate": 1.8441204237888836e-06, "loss": 1.277, "step": 2638 }, { "epoch": 0.2, "learning_rate": 1.8439856146667966e-06, "loss": 1.2687, "step": 2639 }, { "epoch": 0.2, "learning_rate": 1.8438507522083647e-06, "loss": 1.2353, "step": 2640 }, { "epoch": 0.2, "learning_rate": 1.8437158364221111e-06, "loss": 1.2063, "step": 2641 }, { "epoch": 0.21, "learning_rate": 1.8435808673165616e-06, "loss": 1.235, "step": 2642 }, { "epoch": 0.21, "learning_rate": 1.8434458449002462e-06, "loss": 1.2914, "step": 2643 }, { "epoch": 0.21, "learning_rate": 1.8433107691816972e-06, "loss": 0.8578, "step": 2644 }, { "epoch": 0.21, "learning_rate": 1.8431756401694508e-06, "loss": 1.2414, "step": 2645 }, { "epoch": 0.21, "learning_rate": 1.8430404578720472e-06, "loss": 1.2298, "step": 2646 }, { "epoch": 0.21, "learning_rate": 1.8429052222980288e-06, "loss": 1.245, "step": 2647 }, { "epoch": 0.21, "learning_rate": 1.8427699334559418e-06, "loss": 0.848, "step": 2648 }, { "epoch": 0.21, "learning_rate": 1.842634591354337e-06, "loss": 0.8424, "step": 2649 }, { "epoch": 0.21, "learning_rate": 1.842499196001766e-06, "loss": 1.252, "step": 2650 }, { "epoch": 0.21, "learning_rate": 1.842363747406786e-06, "loss": 1.3076, "step": 2651 }, { "epoch": 0.21, "learning_rate": 1.8422282455779564e-06, "loss": 1.2352, "step": 2652 }, { "epoch": 0.21, "learning_rate": 1.8420926905238408e-06, "loss": 1.245, "step": 2653 }, { "epoch": 0.21, "learning_rate": 1.8419570822530054e-06, "loss": 1.236, "step": 2654 }, { "epoch": 0.21, "learning_rate": 1.84182142077402e-06, "loss": 1.2365, "step": 2655 }, { "epoch": 0.21, "learning_rate": 1.8416857060954584e-06, "loss": 1.2334, "step": 2656 }, { "epoch": 0.21, "learning_rate": 1.8415499382258964e-06, "loss": 0.812, "step": 2657 }, { "epoch": 0.21, "learning_rate": 1.8414141171739148e-06, "loss": 1.2128, "step": 2658 }, { "epoch": 0.21, "learning_rate": 1.8412782429480962e-06, "loss": 1.2495, "step": 2659 }, { "epoch": 0.21, "learning_rate": 1.8411423155570276e-06, "loss": 1.297, "step": 2660 }, { "epoch": 0.21, "learning_rate": 1.8410063350092986e-06, "loss": 1.209, "step": 2661 }, { "epoch": 0.21, "learning_rate": 1.8408703013135034e-06, "loss": 1.1641, "step": 2662 }, { "epoch": 0.21, "learning_rate": 1.8407342144782382e-06, "loss": 1.2636, "step": 2663 }, { "epoch": 0.21, "learning_rate": 1.8405980745121034e-06, "loss": 1.2404, "step": 2664 }, { "epoch": 0.21, "learning_rate": 1.8404618814237023e-06, "loss": 1.2727, "step": 2665 }, { "epoch": 0.21, "learning_rate": 1.840325635221642e-06, "loss": 1.2406, "step": 2666 }, { "epoch": 0.21, "learning_rate": 1.8401893359145317e-06, "loss": 1.2123, "step": 2667 }, { "epoch": 0.21, "learning_rate": 1.8400529835109864e-06, "loss": 0.8192, "step": 2668 }, { "epoch": 0.21, "learning_rate": 1.8399165780196219e-06, "loss": 1.2525, "step": 2669 }, { "epoch": 0.21, "learning_rate": 1.839780119449059e-06, "loss": 1.296, "step": 2670 }, { "epoch": 0.21, "learning_rate": 1.8396436078079214e-06, "loss": 1.2518, "step": 2671 }, { "epoch": 0.21, "learning_rate": 1.8395070431048353e-06, "loss": 1.2139, "step": 2672 }, { "epoch": 0.21, "learning_rate": 1.8393704253484322e-06, "loss": 0.8184, "step": 2673 }, { "epoch": 0.21, "learning_rate": 1.8392337545473446e-06, "loss": 1.2417, "step": 2674 }, { "epoch": 0.21, "learning_rate": 1.8390970307102102e-06, "loss": 1.2035, "step": 2675 }, { "epoch": 0.21, "learning_rate": 1.838960253845669e-06, "loss": 0.844, "step": 2676 }, { "epoch": 0.21, "learning_rate": 1.8388234239623652e-06, "loss": 1.2396, "step": 2677 }, { "epoch": 0.21, "learning_rate": 1.8386865410689454e-06, "loss": 1.2179, "step": 2678 }, { "epoch": 0.21, "learning_rate": 1.8385496051740603e-06, "loss": 1.2051, "step": 2679 }, { "epoch": 0.21, "learning_rate": 1.8384126162863635e-06, "loss": 1.2316, "step": 2680 }, { "epoch": 0.21, "learning_rate": 1.8382755744145125e-06, "loss": 1.2415, "step": 2681 }, { "epoch": 0.21, "learning_rate": 1.838138479567167e-06, "loss": 1.2487, "step": 2682 }, { "epoch": 0.21, "learning_rate": 1.8380013317529918e-06, "loss": 1.1995, "step": 2683 }, { "epoch": 0.21, "learning_rate": 1.8378641309806532e-06, "loss": 1.2199, "step": 2684 }, { "epoch": 0.21, "learning_rate": 1.8377268772588223e-06, "loss": 1.2677, "step": 2685 }, { "epoch": 0.21, "learning_rate": 1.8375895705961726e-06, "loss": 1.328, "step": 2686 }, { "epoch": 0.21, "learning_rate": 1.8374522110013815e-06, "loss": 1.2214, "step": 2687 }, { "epoch": 0.21, "learning_rate": 1.8373147984831295e-06, "loss": 1.2156, "step": 2688 }, { "epoch": 0.21, "learning_rate": 1.8371773330501003e-06, "loss": 1.2506, "step": 2689 }, { "epoch": 0.21, "learning_rate": 1.8370398147109816e-06, "loss": 1.1879, "step": 2690 }, { "epoch": 0.21, "learning_rate": 1.8369022434744637e-06, "loss": 1.268, "step": 2691 }, { "epoch": 0.21, "learning_rate": 1.8367646193492404e-06, "loss": 1.1984, "step": 2692 }, { "epoch": 0.21, "learning_rate": 1.8366269423440091e-06, "loss": 1.2615, "step": 2693 }, { "epoch": 0.21, "learning_rate": 1.8364892124674704e-06, "loss": 1.2945, "step": 2694 }, { "epoch": 0.21, "learning_rate": 1.8363514297283283e-06, "loss": 1.2319, "step": 2695 }, { "epoch": 0.21, "learning_rate": 1.83621359413529e-06, "loss": 1.2338, "step": 2696 }, { "epoch": 0.21, "learning_rate": 1.8360757056970661e-06, "loss": 1.2108, "step": 2697 }, { "epoch": 0.21, "learning_rate": 1.8359377644223707e-06, "loss": 1.2533, "step": 2698 }, { "epoch": 0.21, "learning_rate": 1.835799770319921e-06, "loss": 1.2414, "step": 2699 }, { "epoch": 0.21, "learning_rate": 1.835661723398438e-06, "loss": 0.8149, "step": 2700 }, { "epoch": 0.21, "learning_rate": 1.835523623666645e-06, "loss": 1.2763, "step": 2701 }, { "epoch": 0.21, "learning_rate": 1.83538547113327e-06, "loss": 1.2233, "step": 2702 }, { "epoch": 0.21, "learning_rate": 1.835247265807043e-06, "loss": 1.218, "step": 2703 }, { "epoch": 0.21, "learning_rate": 1.8351090076966988e-06, "loss": 1.2283, "step": 2704 }, { "epoch": 0.21, "learning_rate": 1.834970696810974e-06, "loss": 1.1809, "step": 2705 }, { "epoch": 0.21, "learning_rate": 1.8348323331586095e-06, "loss": 1.2732, "step": 2706 }, { "epoch": 0.21, "learning_rate": 1.8346939167483497e-06, "loss": 1.2474, "step": 2707 }, { "epoch": 0.21, "learning_rate": 1.8345554475889414e-06, "loss": 1.1598, "step": 2708 }, { "epoch": 0.21, "learning_rate": 1.8344169256891355e-06, "loss": 1.2241, "step": 2709 }, { "epoch": 0.21, "learning_rate": 1.8342783510576858e-06, "loss": 1.2203, "step": 2710 }, { "epoch": 0.21, "learning_rate": 1.83413972370335e-06, "loss": 1.2688, "step": 2711 }, { "epoch": 0.21, "learning_rate": 1.8340010436348885e-06, "loss": 1.3093, "step": 2712 }, { "epoch": 0.21, "learning_rate": 1.8338623108610653e-06, "loss": 1.2471, "step": 2713 }, { "epoch": 0.21, "learning_rate": 1.8337235253906478e-06, "loss": 1.2393, "step": 2714 }, { "epoch": 0.21, "learning_rate": 1.833584687232407e-06, "loss": 1.2681, "step": 2715 }, { "epoch": 0.21, "learning_rate": 1.8334457963951163e-06, "loss": 1.2233, "step": 2716 }, { "epoch": 0.21, "learning_rate": 1.833306852887553e-06, "loss": 1.224, "step": 2717 }, { "epoch": 0.21, "learning_rate": 1.8331678567184982e-06, "loss": 0.8528, "step": 2718 }, { "epoch": 0.21, "learning_rate": 1.8330288078967358e-06, "loss": 1.2062, "step": 2719 }, { "epoch": 0.21, "learning_rate": 1.832889706431053e-06, "loss": 1.2275, "step": 2720 }, { "epoch": 0.21, "learning_rate": 1.83275055233024e-06, "loss": 0.889, "step": 2721 }, { "epoch": 0.21, "learning_rate": 1.8326113456030915e-06, "loss": 1.2489, "step": 2722 }, { "epoch": 0.21, "learning_rate": 1.8324720862584043e-06, "loss": 1.2054, "step": 2723 }, { "epoch": 0.21, "learning_rate": 1.8323327743049794e-06, "loss": 1.2763, "step": 2724 }, { "epoch": 0.21, "learning_rate": 1.8321934097516203e-06, "loss": 1.2002, "step": 2725 }, { "epoch": 0.21, "learning_rate": 1.8320539926071346e-06, "loss": 1.2013, "step": 2726 }, { "epoch": 0.21, "learning_rate": 1.8319145228803325e-06, "loss": 1.2799, "step": 2727 }, { "epoch": 0.21, "learning_rate": 1.8317750005800284e-06, "loss": 1.2999, "step": 2728 }, { "epoch": 0.21, "learning_rate": 1.8316354257150388e-06, "loss": 1.2535, "step": 2729 }, { "epoch": 0.21, "learning_rate": 1.8314957982941847e-06, "loss": 1.2409, "step": 2730 }, { "epoch": 0.21, "learning_rate": 1.8313561183262904e-06, "loss": 1.2691, "step": 2731 }, { "epoch": 0.21, "learning_rate": 1.8312163858201824e-06, "loss": 1.2427, "step": 2732 }, { "epoch": 0.21, "learning_rate": 1.831076600784691e-06, "loss": 1.2044, "step": 2733 }, { "epoch": 0.21, "learning_rate": 1.8309367632286512e-06, "loss": 1.2399, "step": 2734 }, { "epoch": 0.21, "learning_rate": 1.8307968731608989e-06, "loss": 1.2227, "step": 2735 }, { "epoch": 0.21, "learning_rate": 1.8306569305902754e-06, "loss": 1.1916, "step": 2736 }, { "epoch": 0.21, "learning_rate": 1.8305169355256242e-06, "loss": 1.2847, "step": 2737 }, { "epoch": 0.21, "learning_rate": 1.8303768879757921e-06, "loss": 1.1918, "step": 2738 }, { "epoch": 0.21, "learning_rate": 1.8302367879496298e-06, "loss": 0.8522, "step": 2739 }, { "epoch": 0.21, "learning_rate": 1.8300966354559911e-06, "loss": 1.2519, "step": 2740 }, { "epoch": 0.21, "learning_rate": 1.829956430503733e-06, "loss": 1.2133, "step": 2741 }, { "epoch": 0.21, "learning_rate": 1.8298161731017157e-06, "loss": 1.2595, "step": 2742 }, { "epoch": 0.21, "learning_rate": 1.829675863258803e-06, "loss": 1.2486, "step": 2743 }, { "epoch": 0.21, "learning_rate": 1.829535500983862e-06, "loss": 1.2729, "step": 2744 }, { "epoch": 0.21, "learning_rate": 1.8293950862857628e-06, "loss": 1.2365, "step": 2745 }, { "epoch": 0.21, "learning_rate": 1.8292546191733794e-06, "loss": 1.2409, "step": 2746 }, { "epoch": 0.21, "learning_rate": 1.829114099655588e-06, "loss": 1.2291, "step": 2747 }, { "epoch": 0.21, "learning_rate": 1.8289735277412696e-06, "loss": 1.2253, "step": 2748 }, { "epoch": 0.21, "learning_rate": 1.8288329034393073e-06, "loss": 1.2362, "step": 2749 }, { "epoch": 0.21, "learning_rate": 1.828692226758588e-06, "loss": 1.2447, "step": 2750 }, { "epoch": 0.21, "learning_rate": 1.8285514977080023e-06, "loss": 1.252, "step": 2751 }, { "epoch": 0.21, "learning_rate": 1.8284107162964434e-06, "loss": 1.2215, "step": 2752 }, { "epoch": 0.21, "learning_rate": 1.8282698825328078e-06, "loss": 1.1928, "step": 2753 }, { "epoch": 0.21, "learning_rate": 1.8281289964259958e-06, "loss": 1.1925, "step": 2754 }, { "epoch": 0.21, "learning_rate": 1.8279880579849112e-06, "loss": 1.2623, "step": 2755 }, { "epoch": 0.21, "learning_rate": 1.8278470672184602e-06, "loss": 1.2696, "step": 2756 }, { "epoch": 0.21, "learning_rate": 1.8277060241355531e-06, "loss": 1.235, "step": 2757 }, { "epoch": 0.21, "learning_rate": 1.8275649287451032e-06, "loss": 1.2147, "step": 2758 }, { "epoch": 0.21, "learning_rate": 1.8274237810560268e-06, "loss": 1.2399, "step": 2759 }, { "epoch": 0.21, "learning_rate": 1.8272825810772444e-06, "loss": 0.8504, "step": 2760 }, { "epoch": 0.21, "learning_rate": 1.8271413288176788e-06, "loss": 1.1751, "step": 2761 }, { "epoch": 0.21, "learning_rate": 1.8270000242862568e-06, "loss": 1.2584, "step": 2762 }, { "epoch": 0.21, "learning_rate": 1.826858667491908e-06, "loss": 1.282, "step": 2763 }, { "epoch": 0.21, "learning_rate": 1.826717258443566e-06, "loss": 1.2356, "step": 2764 }, { "epoch": 0.21, "learning_rate": 1.8265757971501669e-06, "loss": 1.2557, "step": 2765 }, { "epoch": 0.21, "learning_rate": 1.8264342836206506e-06, "loss": 1.2818, "step": 2766 }, { "epoch": 0.21, "learning_rate": 1.82629271786396e-06, "loss": 1.2015, "step": 2767 }, { "epoch": 0.21, "learning_rate": 1.8261510998890417e-06, "loss": 1.1873, "step": 2768 }, { "epoch": 0.21, "learning_rate": 1.8260094297048449e-06, "loss": 1.2725, "step": 2769 }, { "epoch": 0.21, "learning_rate": 1.8258677073203233e-06, "loss": 1.2665, "step": 2770 }, { "epoch": 0.22, "learning_rate": 1.8257259327444325e-06, "loss": 1.2405, "step": 2771 }, { "epoch": 0.22, "learning_rate": 1.8255841059861323e-06, "loss": 1.2503, "step": 2772 }, { "epoch": 0.22, "learning_rate": 1.8254422270543858e-06, "loss": 1.3094, "step": 2773 }, { "epoch": 0.22, "learning_rate": 1.8253002959581586e-06, "loss": 1.2609, "step": 2774 }, { "epoch": 0.22, "learning_rate": 1.8251583127064207e-06, "loss": 0.8699, "step": 2775 }, { "epoch": 0.22, "learning_rate": 1.8250162773081444e-06, "loss": 1.2861, "step": 2776 }, { "epoch": 0.22, "learning_rate": 1.8248741897723062e-06, "loss": 1.2306, "step": 2777 }, { "epoch": 0.22, "learning_rate": 1.824732050107885e-06, "loss": 1.208, "step": 2778 }, { "epoch": 0.22, "learning_rate": 1.8245898583238634e-06, "loss": 1.2621, "step": 2779 }, { "epoch": 0.22, "learning_rate": 1.824447614429228e-06, "loss": 1.1734, "step": 2780 }, { "epoch": 0.22, "learning_rate": 1.8243053184329673e-06, "loss": 1.2806, "step": 2781 }, { "epoch": 0.22, "learning_rate": 1.8241629703440739e-06, "loss": 1.2503, "step": 2782 }, { "epoch": 0.22, "learning_rate": 1.8240205701715442e-06, "loss": 1.2234, "step": 2783 }, { "epoch": 0.22, "learning_rate": 1.8238781179243765e-06, "loss": 0.8655, "step": 2784 }, { "epoch": 0.22, "learning_rate": 1.8237356136115736e-06, "loss": 0.8384, "step": 2785 }, { "epoch": 0.22, "learning_rate": 1.8235930572421412e-06, "loss": 1.3195, "step": 2786 }, { "epoch": 0.22, "learning_rate": 1.8234504488250878e-06, "loss": 1.2898, "step": 2787 }, { "epoch": 0.22, "learning_rate": 1.8233077883694265e-06, "loss": 1.2083, "step": 2788 }, { "epoch": 0.22, "learning_rate": 1.8231650758841724e-06, "loss": 1.2438, "step": 2789 }, { "epoch": 0.22, "learning_rate": 1.8230223113783439e-06, "loss": 1.2347, "step": 2790 }, { "epoch": 0.22, "learning_rate": 1.8228794948609638e-06, "loss": 1.2608, "step": 2791 }, { "epoch": 0.22, "learning_rate": 1.822736626341057e-06, "loss": 1.2402, "step": 2792 }, { "epoch": 0.22, "learning_rate": 1.8225937058276527e-06, "loss": 1.1826, "step": 2793 }, { "epoch": 0.22, "learning_rate": 1.8224507333297826e-06, "loss": 1.2104, "step": 2794 }, { "epoch": 0.22, "learning_rate": 1.8223077088564817e-06, "loss": 0.8269, "step": 2795 }, { "epoch": 0.22, "learning_rate": 1.822164632416789e-06, "loss": 1.3078, "step": 2796 }, { "epoch": 0.22, "learning_rate": 1.8220215040197458e-06, "loss": 1.2184, "step": 2797 }, { "epoch": 0.22, "learning_rate": 1.8218783236743978e-06, "loss": 1.2465, "step": 2798 }, { "epoch": 0.22, "learning_rate": 1.821735091389793e-06, "loss": 1.2417, "step": 2799 }, { "epoch": 0.22, "learning_rate": 1.8215918071749833e-06, "loss": 1.2094, "step": 2800 }, { "epoch": 0.22, "learning_rate": 1.8214484710390235e-06, "loss": 1.2526, "step": 2801 }, { "epoch": 0.22, "learning_rate": 1.821305082990972e-06, "loss": 1.2027, "step": 2802 }, { "epoch": 0.22, "learning_rate": 1.82116164303989e-06, "loss": 1.2221, "step": 2803 }, { "epoch": 0.22, "learning_rate": 1.8210181511948429e-06, "loss": 1.2967, "step": 2804 }, { "epoch": 0.22, "learning_rate": 1.8208746074648982e-06, "loss": 1.2356, "step": 2805 }, { "epoch": 0.22, "learning_rate": 1.8207310118591277e-06, "loss": 1.2504, "step": 2806 }, { "epoch": 0.22, "learning_rate": 1.8205873643866054e-06, "loss": 1.2595, "step": 2807 }, { "epoch": 0.22, "learning_rate": 1.82044366505641e-06, "loss": 1.2892, "step": 2808 }, { "epoch": 0.22, "learning_rate": 1.8202999138776222e-06, "loss": 1.2634, "step": 2809 }, { "epoch": 0.22, "learning_rate": 1.8201561108593263e-06, "loss": 1.2706, "step": 2810 }, { "epoch": 0.22, "learning_rate": 1.8200122560106111e-06, "loss": 1.2358, "step": 2811 }, { "epoch": 0.22, "learning_rate": 1.8198683493405661e-06, "loss": 1.2901, "step": 2812 }, { "epoch": 0.22, "learning_rate": 1.819724390858287e-06, "loss": 0.9029, "step": 2813 }, { "epoch": 0.22, "learning_rate": 1.8195803805728703e-06, "loss": 1.2219, "step": 2814 }, { "epoch": 0.22, "learning_rate": 1.8194363184934174e-06, "loss": 1.2206, "step": 2815 }, { "epoch": 0.22, "learning_rate": 1.8192922046290323e-06, "loss": 1.2693, "step": 2816 }, { "epoch": 0.22, "learning_rate": 1.819148038988822e-06, "loss": 1.1873, "step": 2817 }, { "epoch": 0.22, "learning_rate": 1.8190038215818981e-06, "loss": 1.2951, "step": 2818 }, { "epoch": 0.22, "learning_rate": 1.8188595524173737e-06, "loss": 1.2253, "step": 2819 }, { "epoch": 0.22, "learning_rate": 1.8187152315043665e-06, "loss": 1.2292, "step": 2820 }, { "epoch": 0.22, "learning_rate": 1.8185708588519968e-06, "loss": 1.1903, "step": 2821 }, { "epoch": 0.22, "learning_rate": 1.818426434469388e-06, "loss": 1.2809, "step": 2822 }, { "epoch": 0.22, "learning_rate": 1.8182819583656677e-06, "loss": 1.2749, "step": 2823 }, { "epoch": 0.22, "learning_rate": 1.8181374305499656e-06, "loss": 1.2313, "step": 2824 }, { "epoch": 0.22, "learning_rate": 1.8179928510314158e-06, "loss": 1.2947, "step": 2825 }, { "epoch": 0.22, "learning_rate": 1.8178482198191549e-06, "loss": 1.1602, "step": 2826 }, { "epoch": 0.22, "learning_rate": 1.8177035369223228e-06, "loss": 1.2299, "step": 2827 }, { "epoch": 0.22, "learning_rate": 1.8175588023500633e-06, "loss": 1.284, "step": 2828 }, { "epoch": 0.22, "learning_rate": 1.8174140161115227e-06, "loss": 1.2022, "step": 2829 }, { "epoch": 0.22, "learning_rate": 1.8172691782158504e-06, "loss": 1.2336, "step": 2830 }, { "epoch": 0.22, "learning_rate": 1.8171242886722004e-06, "loss": 1.2608, "step": 2831 }, { "epoch": 0.22, "learning_rate": 1.816979347489729e-06, "loss": 1.2682, "step": 2832 }, { "epoch": 0.22, "learning_rate": 1.8168343546775952e-06, "loss": 1.2363, "step": 2833 }, { "epoch": 0.22, "learning_rate": 1.8166893102449627e-06, "loss": 1.2579, "step": 2834 }, { "epoch": 0.22, "learning_rate": 1.8165442142009973e-06, "loss": 1.2168, "step": 2835 }, { "epoch": 0.22, "learning_rate": 1.8163990665548683e-06, "loss": 1.2484, "step": 2836 }, { "epoch": 0.22, "learning_rate": 1.816253867315749e-06, "loss": 1.2317, "step": 2837 }, { "epoch": 0.22, "learning_rate": 1.8161086164928148e-06, "loss": 1.2453, "step": 2838 }, { "epoch": 0.22, "learning_rate": 1.8159633140952452e-06, "loss": 1.2431, "step": 2839 }, { "epoch": 0.22, "learning_rate": 1.815817960132223e-06, "loss": 1.231, "step": 2840 }, { "epoch": 0.22, "learning_rate": 1.8156725546129332e-06, "loss": 1.199, "step": 2841 }, { "epoch": 0.22, "learning_rate": 1.8155270975465655e-06, "loss": 1.2617, "step": 2842 }, { "epoch": 0.22, "learning_rate": 1.8153815889423118e-06, "loss": 1.2551, "step": 2843 }, { "epoch": 0.22, "learning_rate": 1.815236028809368e-06, "loss": 1.2448, "step": 2844 }, { "epoch": 0.22, "learning_rate": 1.8150904171569326e-06, "loss": 1.2478, "step": 2845 }, { "epoch": 0.22, "learning_rate": 1.8149447539942075e-06, "loss": 1.2496, "step": 2846 }, { "epoch": 0.22, "learning_rate": 1.8147990393303983e-06, "loss": 1.264, "step": 2847 }, { "epoch": 0.22, "learning_rate": 1.8146532731747134e-06, "loss": 1.1594, "step": 2848 }, { "epoch": 0.22, "learning_rate": 1.8145074555363648e-06, "loss": 1.2561, "step": 2849 }, { "epoch": 0.22, "learning_rate": 1.8143615864245674e-06, "loss": 1.2332, "step": 2850 }, { "epoch": 0.22, "learning_rate": 1.8142156658485397e-06, "loss": 1.2317, "step": 2851 }, { "epoch": 0.22, "learning_rate": 1.8140696938175028e-06, "loss": 1.276, "step": 2852 }, { "epoch": 0.22, "learning_rate": 1.813923670340682e-06, "loss": 1.2362, "step": 2853 }, { "epoch": 0.22, "learning_rate": 1.813777595427305e-06, "loss": 1.2567, "step": 2854 }, { "epoch": 0.22, "learning_rate": 1.8136314690866034e-06, "loss": 1.2034, "step": 2855 }, { "epoch": 0.22, "learning_rate": 1.813485291327812e-06, "loss": 1.2982, "step": 2856 }, { "epoch": 0.22, "learning_rate": 1.8133390621601679e-06, "loss": 1.2607, "step": 2857 }, { "epoch": 0.22, "learning_rate": 1.813192781592913e-06, "loss": 1.2205, "step": 2858 }, { "epoch": 0.22, "learning_rate": 1.8130464496352907e-06, "loss": 1.2955, "step": 2859 }, { "epoch": 0.22, "learning_rate": 1.8129000662965493e-06, "loss": 1.2779, "step": 2860 }, { "epoch": 0.22, "learning_rate": 1.8127536315859393e-06, "loss": 1.2657, "step": 2861 }, { "epoch": 0.22, "learning_rate": 1.812607145512715e-06, "loss": 1.253, "step": 2862 }, { "epoch": 0.22, "learning_rate": 1.8124606080861334e-06, "loss": 1.2379, "step": 2863 }, { "epoch": 0.22, "learning_rate": 1.8123140193154551e-06, "loss": 1.1624, "step": 2864 }, { "epoch": 0.22, "learning_rate": 1.812167379209944e-06, "loss": 0.8853, "step": 2865 }, { "epoch": 0.22, "learning_rate": 1.8120206877788673e-06, "loss": 1.2645, "step": 2866 }, { "epoch": 0.22, "learning_rate": 1.811873945031495e-06, "loss": 1.2398, "step": 2867 }, { "epoch": 0.22, "learning_rate": 1.8117271509771008e-06, "loss": 1.2376, "step": 2868 }, { "epoch": 0.22, "learning_rate": 1.8115803056249614e-06, "loss": 1.2468, "step": 2869 }, { "epoch": 0.22, "learning_rate": 1.8114334089843567e-06, "loss": 0.8552, "step": 2870 }, { "epoch": 0.22, "learning_rate": 1.81128646106457e-06, "loss": 1.2612, "step": 2871 }, { "epoch": 0.22, "learning_rate": 1.8111394618748878e-06, "loss": 1.2489, "step": 2872 }, { "epoch": 0.22, "learning_rate": 1.8109924114246e-06, "loss": 1.2215, "step": 2873 }, { "epoch": 0.22, "learning_rate": 1.8108453097229996e-06, "loss": 1.1952, "step": 2874 }, { "epoch": 0.22, "learning_rate": 1.8106981567793825e-06, "loss": 1.2679, "step": 2875 }, { "epoch": 0.22, "learning_rate": 1.8105509526030483e-06, "loss": 1.2662, "step": 2876 }, { "epoch": 0.22, "learning_rate": 1.8104036972032994e-06, "loss": 1.2757, "step": 2877 }, { "epoch": 0.22, "learning_rate": 1.8102563905894424e-06, "loss": 1.2215, "step": 2878 }, { "epoch": 0.22, "learning_rate": 1.810109032770786e-06, "loss": 1.2909, "step": 2879 }, { "epoch": 0.22, "learning_rate": 1.8099616237566427e-06, "loss": 1.2564, "step": 2880 }, { "epoch": 0.22, "learning_rate": 1.8098141635563277e-06, "loss": 1.2433, "step": 2881 }, { "epoch": 0.22, "learning_rate": 1.8096666521791608e-06, "loss": 1.2419, "step": 2882 }, { "epoch": 0.22, "learning_rate": 1.8095190896344631e-06, "loss": 1.3218, "step": 2883 }, { "epoch": 0.22, "learning_rate": 1.8093714759315609e-06, "loss": 1.2053, "step": 2884 }, { "epoch": 0.22, "learning_rate": 1.8092238110797819e-06, "loss": 1.2073, "step": 2885 }, { "epoch": 0.22, "learning_rate": 1.8090760950884586e-06, "loss": 1.2876, "step": 2886 }, { "epoch": 0.22, "learning_rate": 1.8089283279669253e-06, "loss": 1.1857, "step": 2887 }, { "epoch": 0.22, "learning_rate": 1.8087805097245207e-06, "loss": 1.2628, "step": 2888 }, { "epoch": 0.22, "learning_rate": 1.8086326403705862e-06, "loss": 1.25, "step": 2889 }, { "epoch": 0.22, "learning_rate": 1.8084847199144666e-06, "loss": 1.2627, "step": 2890 }, { "epoch": 0.22, "learning_rate": 1.8083367483655102e-06, "loss": 1.2728, "step": 2891 }, { "epoch": 0.22, "learning_rate": 1.8081887257330672e-06, "loss": 1.2928, "step": 2892 }, { "epoch": 0.22, "learning_rate": 1.8080406520264928e-06, "loss": 1.2426, "step": 2893 }, { "epoch": 0.22, "learning_rate": 1.8078925272551447e-06, "loss": 1.2832, "step": 2894 }, { "epoch": 0.22, "learning_rate": 1.8077443514283832e-06, "loss": 1.2358, "step": 2895 }, { "epoch": 0.22, "learning_rate": 1.8075961245555727e-06, "loss": 1.2348, "step": 2896 }, { "epoch": 0.22, "learning_rate": 1.8074478466460805e-06, "loss": 1.2397, "step": 2897 }, { "epoch": 0.22, "learning_rate": 1.8072995177092773e-06, "loss": 1.2367, "step": 2898 }, { "epoch": 0.22, "learning_rate": 1.8071511377545367e-06, "loss": 0.8696, "step": 2899 }, { "epoch": 0.23, "learning_rate": 1.8070027067912354e-06, "loss": 1.2343, "step": 2900 }, { "epoch": 0.23, "learning_rate": 1.8068542248287544e-06, "loss": 1.2286, "step": 2901 }, { "epoch": 0.23, "learning_rate": 1.8067056918764764e-06, "loss": 1.2737, "step": 2902 }, { "epoch": 0.23, "learning_rate": 1.8065571079437882e-06, "loss": 1.2685, "step": 2903 }, { "epoch": 0.23, "learning_rate": 1.80640847304008e-06, "loss": 0.8966, "step": 2904 }, { "epoch": 0.23, "learning_rate": 1.8062597871747447e-06, "loss": 1.211, "step": 2905 }, { "epoch": 0.23, "learning_rate": 1.8061110503571784e-06, "loss": 1.2218, "step": 2906 }, { "epoch": 0.23, "learning_rate": 1.8059622625967807e-06, "loss": 1.2001, "step": 2907 }, { "epoch": 0.23, "learning_rate": 1.805813423902955e-06, "loss": 1.213, "step": 2908 }, { "epoch": 0.23, "learning_rate": 1.8056645342851064e-06, "loss": 1.2647, "step": 2909 }, { "epoch": 0.23, "learning_rate": 1.805515593752645e-06, "loss": 1.1815, "step": 2910 }, { "epoch": 0.23, "learning_rate": 1.805366602314982e-06, "loss": 1.2825, "step": 2911 }, { "epoch": 0.23, "learning_rate": 1.8052175599815343e-06, "loss": 1.2217, "step": 2912 }, { "epoch": 0.23, "learning_rate": 1.80506846676172e-06, "loss": 1.2182, "step": 2913 }, { "epoch": 0.23, "learning_rate": 1.8049193226649612e-06, "loss": 1.2531, "step": 2914 }, { "epoch": 0.23, "learning_rate": 1.8047701277006835e-06, "loss": 0.8381, "step": 2915 }, { "epoch": 0.23, "learning_rate": 1.8046208818783152e-06, "loss": 1.2408, "step": 2916 }, { "epoch": 0.23, "learning_rate": 1.8044715852072882e-06, "loss": 1.2418, "step": 2917 }, { "epoch": 0.23, "learning_rate": 1.804322237697037e-06, "loss": 1.2306, "step": 2918 }, { "epoch": 0.23, "learning_rate": 1.804172839357e-06, "loss": 1.2595, "step": 2919 }, { "epoch": 0.23, "learning_rate": 1.804023390196619e-06, "loss": 1.2537, "step": 2920 }, { "epoch": 0.23, "learning_rate": 1.8038738902253376e-06, "loss": 0.8384, "step": 2921 }, { "epoch": 0.23, "learning_rate": 1.803724339452604e-06, "loss": 1.1937, "step": 2922 }, { "epoch": 0.23, "learning_rate": 1.8035747378878696e-06, "loss": 1.2642, "step": 2923 }, { "epoch": 0.23, "learning_rate": 1.803425085540588e-06, "loss": 1.2276, "step": 2924 }, { "epoch": 0.23, "learning_rate": 1.8032753824202173e-06, "loss": 1.2665, "step": 2925 }, { "epoch": 0.23, "learning_rate": 1.8031256285362173e-06, "loss": 1.2568, "step": 2926 }, { "epoch": 0.23, "learning_rate": 1.802975823898052e-06, "loss": 1.3072, "step": 2927 }, { "epoch": 0.23, "learning_rate": 1.8028259685151888e-06, "loss": 1.2952, "step": 2928 }, { "epoch": 0.23, "learning_rate": 1.8026760623970978e-06, "loss": 1.2058, "step": 2929 }, { "epoch": 0.23, "learning_rate": 1.802526105553252e-06, "loss": 0.8424, "step": 2930 }, { "epoch": 0.23, "learning_rate": 1.8023760979931289e-06, "loss": 1.2289, "step": 2931 }, { "epoch": 0.23, "learning_rate": 1.8022260397262077e-06, "loss": 0.8758, "step": 2932 }, { "epoch": 0.23, "learning_rate": 1.8020759307619716e-06, "loss": 1.2334, "step": 2933 }, { "epoch": 0.23, "learning_rate": 1.8019257711099068e-06, "loss": 1.2817, "step": 2934 }, { "epoch": 0.23, "learning_rate": 1.8017755607795027e-06, "loss": 1.2399, "step": 2935 }, { "epoch": 0.23, "learning_rate": 1.801625299780252e-06, "loss": 1.2465, "step": 2936 }, { "epoch": 0.23, "learning_rate": 1.8014749881216508e-06, "loss": 1.2471, "step": 2937 }, { "epoch": 0.23, "learning_rate": 1.8013246258131983e-06, "loss": 1.2827, "step": 2938 }, { "epoch": 0.23, "learning_rate": 1.801174212864396e-06, "loss": 1.2628, "step": 2939 }, { "epoch": 0.23, "learning_rate": 1.8010237492847499e-06, "loss": 1.2315, "step": 2940 }, { "epoch": 0.23, "learning_rate": 1.800873235083769e-06, "loss": 1.2511, "step": 2941 }, { "epoch": 0.23, "learning_rate": 1.8007226702709644e-06, "loss": 1.2499, "step": 2942 }, { "epoch": 0.23, "learning_rate": 1.8005720548558516e-06, "loss": 1.2448, "step": 2943 }, { "epoch": 0.23, "learning_rate": 1.8004213888479488e-06, "loss": 1.2791, "step": 2944 }, { "epoch": 0.23, "learning_rate": 1.8002706722567775e-06, "loss": 1.2616, "step": 2945 }, { "epoch": 0.23, "learning_rate": 1.8001199050918622e-06, "loss": 1.2363, "step": 2946 }, { "epoch": 0.23, "learning_rate": 1.7999690873627312e-06, "loss": 0.8686, "step": 2947 }, { "epoch": 0.23, "learning_rate": 1.799818219078915e-06, "loss": 1.2587, "step": 2948 }, { "epoch": 0.23, "learning_rate": 1.7996673002499478e-06, "loss": 1.2166, "step": 2949 }, { "epoch": 0.23, "learning_rate": 1.7995163308853676e-06, "loss": 1.2228, "step": 2950 }, { "epoch": 0.23, "learning_rate": 1.7993653109947144e-06, "loss": 1.2577, "step": 2951 }, { "epoch": 0.23, "learning_rate": 1.7992142405875326e-06, "loss": 1.2383, "step": 2952 }, { "epoch": 0.23, "learning_rate": 1.7990631196733687e-06, "loss": 1.2576, "step": 2953 }, { "epoch": 0.23, "learning_rate": 1.7989119482617736e-06, "loss": 0.8608, "step": 2954 }, { "epoch": 0.23, "learning_rate": 1.7987607263623002e-06, "loss": 1.254, "step": 2955 }, { "epoch": 0.23, "learning_rate": 1.7986094539845048e-06, "loss": 0.8626, "step": 2956 }, { "epoch": 0.23, "learning_rate": 1.7984581311379476e-06, "loss": 1.208, "step": 2957 }, { "epoch": 0.23, "learning_rate": 1.7983067578321917e-06, "loss": 1.2601, "step": 2958 }, { "epoch": 0.23, "learning_rate": 1.798155334076803e-06, "loss": 1.2531, "step": 2959 }, { "epoch": 0.23, "learning_rate": 1.7980038598813506e-06, "loss": 1.2409, "step": 2960 }, { "epoch": 0.23, "learning_rate": 1.7978523352554075e-06, "loss": 1.2743, "step": 2961 }, { "epoch": 0.23, "learning_rate": 1.7977007602085492e-06, "loss": 0.8792, "step": 2962 }, { "epoch": 0.23, "learning_rate": 1.7975491347503548e-06, "loss": 1.2178, "step": 2963 }, { "epoch": 0.23, "learning_rate": 1.7973974588904061e-06, "loss": 1.2773, "step": 2964 }, { "epoch": 0.23, "learning_rate": 1.7972457326382889e-06, "loss": 1.1954, "step": 2965 }, { "epoch": 0.23, "learning_rate": 1.7970939560035908e-06, "loss": 1.2238, "step": 2966 }, { "epoch": 0.23, "learning_rate": 1.7969421289959041e-06, "loss": 1.2405, "step": 2967 }, { "epoch": 0.23, "learning_rate": 1.7967902516248236e-06, "loss": 0.8304, "step": 2968 }, { "epoch": 0.23, "learning_rate": 1.796638323899947e-06, "loss": 1.2266, "step": 2969 }, { "epoch": 0.23, "learning_rate": 1.7964863458308756e-06, "loss": 1.2193, "step": 2970 }, { "epoch": 0.23, "learning_rate": 1.7963343174272141e-06, "loss": 1.2824, "step": 2971 }, { "epoch": 0.23, "learning_rate": 1.7961822386985695e-06, "loss": 1.2594, "step": 2972 }, { "epoch": 0.23, "learning_rate": 1.7960301096545529e-06, "loss": 1.2369, "step": 2973 }, { "epoch": 0.23, "learning_rate": 1.7958779303047784e-06, "loss": 1.2833, "step": 2974 }, { "epoch": 0.23, "learning_rate": 1.7957257006588623e-06, "loss": 1.1985, "step": 2975 }, { "epoch": 0.23, "learning_rate": 1.7955734207264262e-06, "loss": 1.3156, "step": 2976 }, { "epoch": 0.23, "learning_rate": 1.795421090517092e-06, "loss": 1.2987, "step": 2977 }, { "epoch": 0.23, "learning_rate": 1.7952687100404875e-06, "loss": 1.2922, "step": 2978 }, { "epoch": 0.23, "learning_rate": 1.7951162793062423e-06, "loss": 1.1512, "step": 2979 }, { "epoch": 0.23, "learning_rate": 1.7949637983239888e-06, "loss": 0.9055, "step": 2980 }, { "epoch": 0.23, "learning_rate": 1.7948112671033635e-06, "loss": 1.258, "step": 2981 }, { "epoch": 0.23, "learning_rate": 1.794658685654006e-06, "loss": 1.2678, "step": 2982 }, { "epoch": 0.23, "learning_rate": 1.7945060539855586e-06, "loss": 1.2291, "step": 2983 }, { "epoch": 0.23, "learning_rate": 1.794353372107667e-06, "loss": 1.2066, "step": 2984 }, { "epoch": 0.23, "learning_rate": 1.7942006400299796e-06, "loss": 1.2399, "step": 2985 }, { "epoch": 0.23, "learning_rate": 1.7940478577621492e-06, "loss": 1.2199, "step": 2986 }, { "epoch": 0.23, "learning_rate": 1.7938950253138305e-06, "loss": 1.2027, "step": 2987 }, { "epoch": 0.23, "learning_rate": 1.793742142694682e-06, "loss": 1.1945, "step": 2988 }, { "epoch": 0.23, "learning_rate": 1.7935892099143655e-06, "loss": 1.2705, "step": 2989 }, { "epoch": 0.23, "learning_rate": 1.7934362269825455e-06, "loss": 1.2162, "step": 2990 }, { "epoch": 0.23, "learning_rate": 1.7932831939088894e-06, "loss": 1.2749, "step": 2991 }, { "epoch": 0.23, "learning_rate": 1.7931301107030689e-06, "loss": 1.2504, "step": 2992 }, { "epoch": 0.23, "learning_rate": 1.792976977374758e-06, "loss": 1.252, "step": 2993 }, { "epoch": 0.23, "learning_rate": 1.792823793933634e-06, "loss": 1.2726, "step": 2994 }, { "epoch": 0.23, "learning_rate": 1.7926705603893776e-06, "loss": 1.2834, "step": 2995 }, { "epoch": 0.23, "learning_rate": 1.7925172767516725e-06, "loss": 1.221, "step": 2996 }, { "epoch": 0.23, "learning_rate": 1.7923639430302056e-06, "loss": 1.239, "step": 2997 }, { "epoch": 0.23, "learning_rate": 1.7922105592346665e-06, "loss": 1.2759, "step": 2998 }, { "epoch": 0.23, "learning_rate": 1.792057125374749e-06, "loss": 1.196, "step": 2999 }, { "epoch": 0.23, "learning_rate": 1.7919036414601493e-06, "loss": 1.2468, "step": 3000 }, { "epoch": 0.23, "learning_rate": 1.791750107500567e-06, "loss": 0.8204, "step": 3001 }, { "epoch": 0.23, "learning_rate": 1.7915965235057046e-06, "loss": 1.2314, "step": 3002 }, { "epoch": 0.23, "learning_rate": 1.791442889485268e-06, "loss": 1.1751, "step": 3003 }, { "epoch": 0.23, "learning_rate": 1.7912892054489662e-06, "loss": 1.2385, "step": 3004 }, { "epoch": 0.23, "learning_rate": 1.791135471406512e-06, "loss": 1.2822, "step": 3005 }, { "epoch": 0.23, "learning_rate": 1.7909816873676195e-06, "loss": 1.2475, "step": 3006 }, { "epoch": 0.23, "learning_rate": 1.7908278533420082e-06, "loss": 1.2566, "step": 3007 }, { "epoch": 0.23, "learning_rate": 1.7906739693393997e-06, "loss": 1.2213, "step": 3008 }, { "epoch": 0.23, "learning_rate": 1.7905200353695185e-06, "loss": 0.8528, "step": 3009 }, { "epoch": 0.23, "learning_rate": 1.7903660514420924e-06, "loss": 1.2907, "step": 3010 }, { "epoch": 0.23, "learning_rate": 1.7902120175668532e-06, "loss": 1.2293, "step": 3011 }, { "epoch": 0.23, "learning_rate": 1.7900579337535347e-06, "loss": 0.8165, "step": 3012 }, { "epoch": 0.23, "learning_rate": 1.7899038000118744e-06, "loss": 1.2398, "step": 3013 }, { "epoch": 0.23, "learning_rate": 1.789749616351613e-06, "loss": 1.1809, "step": 3014 }, { "epoch": 0.23, "learning_rate": 1.7895953827824943e-06, "loss": 1.1751, "step": 3015 }, { "epoch": 0.23, "learning_rate": 1.7894410993142652e-06, "loss": 1.1951, "step": 3016 }, { "epoch": 0.23, "learning_rate": 1.7892867659566756e-06, "loss": 1.2652, "step": 3017 }, { "epoch": 0.23, "learning_rate": 1.7891323827194788e-06, "loss": 0.868, "step": 3018 }, { "epoch": 0.23, "learning_rate": 1.7889779496124314e-06, "loss": 1.3163, "step": 3019 }, { "epoch": 0.23, "learning_rate": 1.7888234666452923e-06, "loss": 1.2935, "step": 3020 }, { "epoch": 0.23, "learning_rate": 1.7886689338278252e-06, "loss": 1.2132, "step": 3021 }, { "epoch": 0.23, "learning_rate": 1.7885143511697951e-06, "loss": 1.2717, "step": 3022 }, { "epoch": 0.23, "learning_rate": 1.788359718680971e-06, "loss": 1.2627, "step": 3023 }, { "epoch": 0.23, "learning_rate": 1.788205036371125e-06, "loss": 1.2458, "step": 3024 }, { "epoch": 0.23, "learning_rate": 1.7880503042500331e-06, "loss": 1.2282, "step": 3025 }, { "epoch": 0.23, "learning_rate": 1.7878955223274732e-06, "loss": 0.849, "step": 3026 }, { "epoch": 0.23, "learning_rate": 1.787740690613227e-06, "loss": 1.2493, "step": 3027 }, { "epoch": 0.24, "learning_rate": 1.7875858091170785e-06, "loss": 0.8727, "step": 3028 }, { "epoch": 0.24, "learning_rate": 1.7874308778488167e-06, "loss": 1.2182, "step": 3029 }, { "epoch": 0.24, "learning_rate": 1.7872758968182318e-06, "loss": 1.2733, "step": 3030 }, { "epoch": 0.24, "learning_rate": 1.7871208660351185e-06, "loss": 1.261, "step": 3031 }, { "epoch": 0.24, "learning_rate": 1.7869657855092736e-06, "loss": 1.2767, "step": 3032 }, { "epoch": 0.24, "learning_rate": 1.7868106552504975e-06, "loss": 1.2619, "step": 3033 }, { "epoch": 0.24, "learning_rate": 1.7866554752685943e-06, "loss": 1.2284, "step": 3034 }, { "epoch": 0.24, "learning_rate": 1.7865002455733707e-06, "loss": 1.2146, "step": 3035 }, { "epoch": 0.24, "learning_rate": 1.786344966174636e-06, "loss": 1.2493, "step": 3036 }, { "epoch": 0.24, "learning_rate": 1.7861896370822034e-06, "loss": 1.2835, "step": 3037 }, { "epoch": 0.24, "learning_rate": 1.7860342583058893e-06, "loss": 1.2544, "step": 3038 }, { "epoch": 0.24, "learning_rate": 1.785878829855513e-06, "loss": 1.2719, "step": 3039 }, { "epoch": 0.24, "learning_rate": 1.7857233517408968e-06, "loss": 0.8426, "step": 3040 }, { "epoch": 0.24, "learning_rate": 1.7855678239718659e-06, "loss": 1.2319, "step": 3041 }, { "epoch": 0.24, "learning_rate": 1.7854122465582496e-06, "loss": 1.2553, "step": 3042 }, { "epoch": 0.24, "learning_rate": 1.7852566195098793e-06, "loss": 1.2147, "step": 3043 }, { "epoch": 0.24, "learning_rate": 1.7851009428365904e-06, "loss": 1.2423, "step": 3044 }, { "epoch": 0.24, "learning_rate": 1.7849452165482207e-06, "loss": 1.2762, "step": 3045 }, { "epoch": 0.24, "learning_rate": 1.7847894406546113e-06, "loss": 1.2418, "step": 3046 }, { "epoch": 0.24, "learning_rate": 1.784633615165607e-06, "loss": 1.2744, "step": 3047 }, { "epoch": 0.24, "learning_rate": 1.7844777400910552e-06, "loss": 1.2465, "step": 3048 }, { "epoch": 0.24, "learning_rate": 1.7843218154408062e-06, "loss": 1.2578, "step": 3049 }, { "epoch": 0.24, "learning_rate": 1.7841658412247143e-06, "loss": 0.805, "step": 3050 }, { "epoch": 0.24, "learning_rate": 1.784009817452636e-06, "loss": 1.2179, "step": 3051 }, { "epoch": 0.24, "learning_rate": 1.7838537441344314e-06, "loss": 1.2195, "step": 3052 }, { "epoch": 0.24, "learning_rate": 1.783697621279964e-06, "loss": 1.2007, "step": 3053 }, { "epoch": 0.24, "learning_rate": 1.7835414488990998e-06, "loss": 1.2613, "step": 3054 }, { "epoch": 0.24, "learning_rate": 1.7833852270017085e-06, "loss": 0.8544, "step": 3055 }, { "epoch": 0.24, "learning_rate": 1.7832289555976622e-06, "loss": 1.2259, "step": 3056 }, { "epoch": 0.24, "learning_rate": 1.783072634696837e-06, "loss": 1.2323, "step": 3057 }, { "epoch": 0.24, "learning_rate": 1.7829162643091118e-06, "loss": 1.2676, "step": 3058 }, { "epoch": 0.24, "learning_rate": 1.782759844444368e-06, "loss": 0.8604, "step": 3059 }, { "epoch": 0.24, "learning_rate": 1.7826033751124916e-06, "loss": 1.2178, "step": 3060 }, { "epoch": 0.24, "learning_rate": 1.7824468563233699e-06, "loss": 1.2937, "step": 3061 }, { "epoch": 0.24, "learning_rate": 1.7822902880868946e-06, "loss": 1.192, "step": 3062 }, { "epoch": 0.24, "learning_rate": 1.7821336704129601e-06, "loss": 1.277, "step": 3063 }, { "epoch": 0.24, "learning_rate": 1.7819770033114641e-06, "loss": 1.2301, "step": 3064 }, { "epoch": 0.24, "learning_rate": 1.7818202867923073e-06, "loss": 1.3084, "step": 3065 }, { "epoch": 0.24, "learning_rate": 1.7816635208653935e-06, "loss": 0.8424, "step": 3066 }, { "epoch": 0.24, "learning_rate": 1.7815067055406292e-06, "loss": 1.2607, "step": 3067 }, { "epoch": 0.24, "learning_rate": 1.781349840827925e-06, "loss": 1.2659, "step": 3068 }, { "epoch": 0.24, "learning_rate": 1.7811929267371941e-06, "loss": 0.8378, "step": 3069 }, { "epoch": 0.24, "learning_rate": 1.7810359632783527e-06, "loss": 0.7917, "step": 3070 }, { "epoch": 0.24, "learning_rate": 1.7808789504613202e-06, "loss": 1.2458, "step": 3071 }, { "epoch": 0.24, "learning_rate": 1.7807218882960191e-06, "loss": 1.1711, "step": 3072 }, { "epoch": 0.24, "learning_rate": 1.7805647767923748e-06, "loss": 1.2639, "step": 3073 }, { "epoch": 0.24, "learning_rate": 1.7804076159603167e-06, "loss": 1.2536, "step": 3074 }, { "epoch": 0.24, "learning_rate": 1.7802504058097765e-06, "loss": 1.2584, "step": 3075 }, { "epoch": 0.24, "learning_rate": 1.780093146350689e-06, "loss": 1.2661, "step": 3076 }, { "epoch": 0.24, "learning_rate": 1.7799358375929927e-06, "loss": 1.2377, "step": 3077 }, { "epoch": 0.24, "learning_rate": 1.7797784795466281e-06, "loss": 1.2196, "step": 3078 }, { "epoch": 0.24, "learning_rate": 1.7796210722215403e-06, "loss": 1.2564, "step": 3079 }, { "epoch": 0.24, "learning_rate": 1.7794636156276767e-06, "loss": 1.2566, "step": 3080 }, { "epoch": 0.24, "learning_rate": 1.7793061097749878e-06, "loss": 1.2325, "step": 3081 }, { "epoch": 0.24, "learning_rate": 1.7791485546734272e-06, "loss": 1.1872, "step": 3082 }, { "epoch": 0.24, "learning_rate": 1.7789909503329517e-06, "loss": 1.2705, "step": 3083 }, { "epoch": 0.24, "learning_rate": 1.7788332967635213e-06, "loss": 1.2101, "step": 3084 }, { "epoch": 0.24, "learning_rate": 1.7786755939750991e-06, "loss": 1.2511, "step": 3085 }, { "epoch": 0.24, "learning_rate": 1.7785178419776514e-06, "loss": 1.262, "step": 3086 }, { "epoch": 0.24, "learning_rate": 1.778360040781147e-06, "loss": 0.8485, "step": 3087 }, { "epoch": 0.24, "learning_rate": 1.7782021903955585e-06, "loss": 1.3066, "step": 3088 }, { "epoch": 0.24, "learning_rate": 1.778044290830862e-06, "loss": 1.263, "step": 3089 }, { "epoch": 0.24, "learning_rate": 1.7778863420970352e-06, "loss": 1.2438, "step": 3090 }, { "epoch": 0.24, "learning_rate": 1.77772834420406e-06, "loss": 1.2628, "step": 3091 }, { "epoch": 0.24, "learning_rate": 1.7775702971619212e-06, "loss": 1.2266, "step": 3092 }, { "epoch": 0.24, "learning_rate": 1.7774122009806072e-06, "loss": 1.231, "step": 3093 }, { "epoch": 0.24, "learning_rate": 1.7772540556701085e-06, "loss": 1.2041, "step": 3094 }, { "epoch": 0.24, "learning_rate": 1.7770958612404193e-06, "loss": 1.2589, "step": 3095 }, { "epoch": 0.24, "learning_rate": 1.7769376177015367e-06, "loss": 1.2692, "step": 3096 }, { "epoch": 0.24, "learning_rate": 1.7767793250634618e-06, "loss": 1.2159, "step": 3097 }, { "epoch": 0.24, "learning_rate": 1.7766209833361967e-06, "loss": 1.2549, "step": 3098 }, { "epoch": 0.24, "learning_rate": 1.7764625925297491e-06, "loss": 1.2627, "step": 3099 }, { "epoch": 0.24, "learning_rate": 1.7763041526541285e-06, "loss": 1.2504, "step": 3100 }, { "epoch": 0.24, "learning_rate": 1.7761456637193471e-06, "loss": 1.2058, "step": 3101 }, { "epoch": 0.24, "learning_rate": 1.7759871257354207e-06, "loss": 1.2264, "step": 3102 }, { "epoch": 0.24, "learning_rate": 1.7758285387123687e-06, "loss": 1.1833, "step": 3103 }, { "epoch": 0.24, "learning_rate": 1.775669902660213e-06, "loss": 1.1963, "step": 3104 }, { "epoch": 0.24, "learning_rate": 1.7755112175889786e-06, "loss": 1.2495, "step": 3105 }, { "epoch": 0.24, "learning_rate": 1.7753524835086938e-06, "loss": 1.2125, "step": 3106 }, { "epoch": 0.24, "learning_rate": 1.77519370042939e-06, "loss": 1.2301, "step": 3107 }, { "epoch": 0.24, "learning_rate": 1.7750348683611014e-06, "loss": 1.2707, "step": 3108 }, { "epoch": 0.24, "learning_rate": 1.774875987313866e-06, "loss": 1.2362, "step": 3109 }, { "epoch": 0.24, "learning_rate": 1.7747170572977239e-06, "loss": 1.2183, "step": 3110 }, { "epoch": 0.24, "learning_rate": 1.774558078322719e-06, "loss": 1.2603, "step": 3111 }, { "epoch": 0.24, "learning_rate": 1.7743990503988982e-06, "loss": 1.2714, "step": 3112 }, { "epoch": 0.24, "learning_rate": 1.7742399735363111e-06, "loss": 1.1989, "step": 3113 }, { "epoch": 0.24, "learning_rate": 1.7740808477450112e-06, "loss": 1.2946, "step": 3114 }, { "epoch": 0.24, "learning_rate": 1.7739216730350538e-06, "loss": 1.2182, "step": 3115 }, { "epoch": 0.24, "learning_rate": 1.773762449416499e-06, "loss": 1.2606, "step": 3116 }, { "epoch": 0.24, "learning_rate": 1.7736031768994085e-06, "loss": 1.2633, "step": 3117 }, { "epoch": 0.24, "learning_rate": 1.7734438554938475e-06, "loss": 1.273, "step": 3118 }, { "epoch": 0.24, "learning_rate": 1.773284485209885e-06, "loss": 1.2741, "step": 3119 }, { "epoch": 0.24, "learning_rate": 1.773125066057592e-06, "loss": 1.1995, "step": 3120 }, { "epoch": 0.24, "learning_rate": 1.7729655980470437e-06, "loss": 1.2061, "step": 3121 }, { "epoch": 0.24, "learning_rate": 1.7728060811883174e-06, "loss": 1.2446, "step": 3122 }, { "epoch": 0.24, "learning_rate": 1.7726465154914939e-06, "loss": 1.2232, "step": 3123 }, { "epoch": 0.24, "learning_rate": 1.772486900966657e-06, "loss": 1.2307, "step": 3124 }, { "epoch": 0.24, "learning_rate": 1.7723272376238935e-06, "loss": 1.2003, "step": 3125 }, { "epoch": 0.24, "learning_rate": 1.7721675254732945e-06, "loss": 1.2158, "step": 3126 }, { "epoch": 0.24, "learning_rate": 1.772007764524952e-06, "loss": 1.2149, "step": 3127 }, { "epoch": 0.24, "learning_rate": 1.7718479547889627e-06, "loss": 1.2201, "step": 3128 }, { "epoch": 0.24, "learning_rate": 1.7716880962754256e-06, "loss": 1.266, "step": 3129 }, { "epoch": 0.24, "learning_rate": 1.7715281889944437e-06, "loss": 1.2183, "step": 3130 }, { "epoch": 0.24, "learning_rate": 1.7713682329561217e-06, "loss": 1.215, "step": 3131 }, { "epoch": 0.24, "learning_rate": 1.771208228170569e-06, "loss": 1.2434, "step": 3132 }, { "epoch": 0.24, "learning_rate": 1.7710481746478965e-06, "loss": 1.2267, "step": 3133 }, { "epoch": 0.24, "learning_rate": 1.7708880723982192e-06, "loss": 1.2717, "step": 3134 }, { "epoch": 0.24, "learning_rate": 1.7707279214316548e-06, "loss": 1.2533, "step": 3135 }, { "epoch": 0.24, "learning_rate": 1.770567721758324e-06, "loss": 1.2716, "step": 3136 }, { "epoch": 0.24, "learning_rate": 1.7704074733883514e-06, "loss": 1.2447, "step": 3137 }, { "epoch": 0.24, "learning_rate": 1.7702471763318633e-06, "loss": 1.2596, "step": 3138 }, { "epoch": 0.24, "learning_rate": 1.7700868305989904e-06, "loss": 1.2126, "step": 3139 }, { "epoch": 0.24, "learning_rate": 1.7699264361998654e-06, "loss": 1.2406, "step": 3140 }, { "epoch": 0.24, "learning_rate": 1.7697659931446247e-06, "loss": 0.8782, "step": 3141 }, { "epoch": 0.24, "learning_rate": 1.7696055014434075e-06, "loss": 1.2071, "step": 3142 }, { "epoch": 0.24, "learning_rate": 1.7694449611063567e-06, "loss": 1.1772, "step": 3143 }, { "epoch": 0.24, "learning_rate": 1.769284372143617e-06, "loss": 1.3087, "step": 3144 }, { "epoch": 0.24, "learning_rate": 1.7691237345653378e-06, "loss": 1.2454, "step": 3145 }, { "epoch": 0.24, "learning_rate": 1.76896304838167e-06, "loss": 1.2416, "step": 3146 }, { "epoch": 0.24, "learning_rate": 1.7688023136027687e-06, "loss": 1.2627, "step": 3147 }, { "epoch": 0.24, "learning_rate": 1.7686415302387917e-06, "loss": 1.245, "step": 3148 }, { "epoch": 0.24, "learning_rate": 1.7684806982998994e-06, "loss": 1.2897, "step": 3149 }, { "epoch": 0.24, "learning_rate": 1.7683198177962564e-06, "loss": 1.2624, "step": 3150 }, { "epoch": 0.24, "learning_rate": 1.768158888738029e-06, "loss": 1.2504, "step": 3151 }, { "epoch": 0.24, "learning_rate": 1.7679979111353874e-06, "loss": 1.1818, "step": 3152 }, { "epoch": 0.24, "learning_rate": 1.7678368849985049e-06, "loss": 1.2177, "step": 3153 }, { "epoch": 0.24, "learning_rate": 1.7676758103375576e-06, "loss": 1.236, "step": 3154 }, { "epoch": 0.24, "learning_rate": 1.7675146871627246e-06, "loss": 1.2766, "step": 3155 }, { "epoch": 0.24, "learning_rate": 1.7673535154841884e-06, "loss": 1.2612, "step": 3156 }, { "epoch": 0.25, "learning_rate": 1.7671922953121343e-06, "loss": 1.2312, "step": 3157 }, { "epoch": 0.25, "learning_rate": 1.7670310266567509e-06, "loss": 1.2383, "step": 3158 }, { "epoch": 0.25, "learning_rate": 1.7668697095282298e-06, "loss": 1.2104, "step": 3159 }, { "epoch": 0.25, "learning_rate": 1.7667083439367647e-06, "loss": 1.1991, "step": 3160 }, { "epoch": 0.25, "learning_rate": 1.766546929892554e-06, "loss": 1.2632, "step": 3161 }, { "epoch": 0.25, "learning_rate": 1.7663854674057983e-06, "loss": 1.2541, "step": 3162 }, { "epoch": 0.25, "learning_rate": 1.7662239564867011e-06, "loss": 1.2164, "step": 3163 }, { "epoch": 0.25, "learning_rate": 1.7660623971454696e-06, "loss": 1.2314, "step": 3164 }, { "epoch": 0.25, "learning_rate": 1.7659007893923134e-06, "loss": 1.2384, "step": 3165 }, { "epoch": 0.25, "learning_rate": 1.7657391332374458e-06, "loss": 0.8432, "step": 3166 }, { "epoch": 0.25, "learning_rate": 1.765577428691082e-06, "loss": 1.24, "step": 3167 }, { "epoch": 0.25, "learning_rate": 1.7654156757634419e-06, "loss": 1.2242, "step": 3168 }, { "epoch": 0.25, "learning_rate": 1.7652538744647469e-06, "loss": 1.2575, "step": 3169 }, { "epoch": 0.25, "learning_rate": 1.7650920248052229e-06, "loss": 1.2294, "step": 3170 }, { "epoch": 0.25, "learning_rate": 1.7649301267950974e-06, "loss": 0.8887, "step": 3171 }, { "epoch": 0.25, "learning_rate": 1.7647681804446019e-06, "loss": 1.2032, "step": 3172 }, { "epoch": 0.25, "learning_rate": 1.7646061857639709e-06, "loss": 1.2423, "step": 3173 }, { "epoch": 0.25, "learning_rate": 1.7644441427634418e-06, "loss": 1.1703, "step": 3174 }, { "epoch": 0.25, "learning_rate": 1.764282051453255e-06, "loss": 0.8811, "step": 3175 }, { "epoch": 0.25, "learning_rate": 1.7641199118436538e-06, "loss": 1.2914, "step": 3176 }, { "epoch": 0.25, "learning_rate": 1.7639577239448848e-06, "loss": 1.2186, "step": 3177 }, { "epoch": 0.25, "learning_rate": 1.763795487767198e-06, "loss": 1.2661, "step": 3178 }, { "epoch": 0.25, "learning_rate": 1.7636332033208455e-06, "loss": 1.2214, "step": 3179 }, { "epoch": 0.25, "learning_rate": 1.763470870616083e-06, "loss": 1.2079, "step": 3180 }, { "epoch": 0.25, "learning_rate": 1.7633084896631695e-06, "loss": 1.2471, "step": 3181 }, { "epoch": 0.25, "learning_rate": 1.7631460604723669e-06, "loss": 1.266, "step": 3182 }, { "epoch": 0.25, "learning_rate": 1.7629835830539399e-06, "loss": 1.2194, "step": 3183 }, { "epoch": 0.25, "learning_rate": 1.7628210574181562e-06, "loss": 0.8433, "step": 3184 }, { "epoch": 0.25, "learning_rate": 1.762658483575287e-06, "loss": 1.3028, "step": 3185 }, { "epoch": 0.25, "learning_rate": 1.7624958615356062e-06, "loss": 1.2407, "step": 3186 }, { "epoch": 0.25, "learning_rate": 1.762333191309391e-06, "loss": 1.262, "step": 3187 }, { "epoch": 0.25, "learning_rate": 1.7621704729069208e-06, "loss": 1.2068, "step": 3188 }, { "epoch": 0.25, "learning_rate": 1.7620077063384797e-06, "loss": 1.2824, "step": 3189 }, { "epoch": 0.25, "learning_rate": 1.7618448916143533e-06, "loss": 1.2589, "step": 3190 }, { "epoch": 0.25, "learning_rate": 1.7616820287448308e-06, "loss": 1.218, "step": 3191 }, { "epoch": 0.25, "learning_rate": 1.7615191177402047e-06, "loss": 1.2304, "step": 3192 }, { "epoch": 0.25, "learning_rate": 1.7613561586107698e-06, "loss": 1.2462, "step": 3193 }, { "epoch": 0.25, "learning_rate": 1.7611931513668252e-06, "loss": 1.2377, "step": 3194 }, { "epoch": 0.25, "learning_rate": 1.7610300960186714e-06, "loss": 1.2461, "step": 3195 }, { "epoch": 0.25, "learning_rate": 1.7608669925766133e-06, "loss": 1.1818, "step": 3196 }, { "epoch": 0.25, "learning_rate": 1.7607038410509586e-06, "loss": 1.1838, "step": 3197 }, { "epoch": 0.25, "learning_rate": 1.7605406414520175e-06, "loss": 1.2518, "step": 3198 }, { "epoch": 0.25, "learning_rate": 1.7603773937901032e-06, "loss": 1.2288, "step": 3199 }, { "epoch": 0.25, "learning_rate": 1.7602140980755325e-06, "loss": 1.1972, "step": 3200 }, { "epoch": 0.25, "learning_rate": 1.7600507543186254e-06, "loss": 1.233, "step": 3201 }, { "epoch": 0.25, "learning_rate": 1.7598873625297039e-06, "loss": 0.8249, "step": 3202 }, { "epoch": 0.25, "learning_rate": 1.7597239227190946e-06, "loss": 1.2151, "step": 3203 }, { "epoch": 0.25, "learning_rate": 1.7595604348971249e-06, "loss": 0.8196, "step": 3204 }, { "epoch": 0.25, "learning_rate": 1.7593968990741278e-06, "loss": 1.2432, "step": 3205 }, { "epoch": 0.25, "learning_rate": 1.7592333152604373e-06, "loss": 1.2603, "step": 3206 }, { "epoch": 0.25, "learning_rate": 1.7590696834663912e-06, "loss": 1.2791, "step": 3207 }, { "epoch": 0.25, "learning_rate": 1.7589060037023308e-06, "loss": 1.2589, "step": 3208 }, { "epoch": 0.25, "learning_rate": 1.7587422759786e-06, "loss": 1.2179, "step": 3209 }, { "epoch": 0.25, "learning_rate": 1.7585785003055452e-06, "loss": 1.2716, "step": 3210 }, { "epoch": 0.25, "learning_rate": 1.7584146766935165e-06, "loss": 1.2678, "step": 3211 }, { "epoch": 0.25, "learning_rate": 1.7582508051528673e-06, "loss": 1.2314, "step": 3212 }, { "epoch": 0.25, "learning_rate": 1.7580868856939528e-06, "loss": 1.2715, "step": 3213 }, { "epoch": 0.25, "learning_rate": 1.7579229183271327e-06, "loss": 1.2311, "step": 3214 }, { "epoch": 0.25, "learning_rate": 1.7577589030627688e-06, "loss": 1.2586, "step": 3215 }, { "epoch": 0.25, "learning_rate": 1.7575948399112263e-06, "loss": 1.2372, "step": 3216 }, { "epoch": 0.25, "learning_rate": 1.7574307288828734e-06, "loss": 1.1723, "step": 3217 }, { "epoch": 0.25, "learning_rate": 1.7572665699880807e-06, "loss": 1.1624, "step": 3218 }, { "epoch": 0.25, "learning_rate": 1.7571023632372227e-06, "loss": 1.2531, "step": 3219 }, { "epoch": 0.25, "learning_rate": 1.7569381086406765e-06, "loss": 0.8814, "step": 3220 }, { "epoch": 0.25, "learning_rate": 1.7567738062088223e-06, "loss": 1.241, "step": 3221 }, { "epoch": 0.25, "learning_rate": 1.7566094559520436e-06, "loss": 1.248, "step": 3222 }, { "epoch": 0.25, "learning_rate": 1.7564450578807261e-06, "loss": 1.2278, "step": 3223 }, { "epoch": 0.25, "learning_rate": 1.7562806120052599e-06, "loss": 1.2171, "step": 3224 }, { "epoch": 0.25, "learning_rate": 1.7561161183360364e-06, "loss": 1.2196, "step": 3225 }, { "epoch": 0.25, "learning_rate": 1.7559515768834514e-06, "loss": 1.227, "step": 3226 }, { "epoch": 0.25, "learning_rate": 1.755786987657903e-06, "loss": 0.8365, "step": 3227 }, { "epoch": 0.25, "learning_rate": 1.755622350669793e-06, "loss": 1.2356, "step": 3228 }, { "epoch": 0.25, "learning_rate": 1.755457665929525e-06, "loss": 1.2397, "step": 3229 }, { "epoch": 0.25, "learning_rate": 1.755292933447507e-06, "loss": 1.1686, "step": 3230 }, { "epoch": 0.25, "learning_rate": 1.7551281532341492e-06, "loss": 1.2371, "step": 3231 }, { "epoch": 0.25, "learning_rate": 1.7549633252998653e-06, "loss": 1.2273, "step": 3232 }, { "epoch": 0.25, "learning_rate": 1.754798449655071e-06, "loss": 1.1966, "step": 3233 }, { "epoch": 0.25, "learning_rate": 1.7546335263101866e-06, "loss": 1.3016, "step": 3234 }, { "epoch": 0.25, "learning_rate": 1.754468555275634e-06, "loss": 1.2564, "step": 3235 }, { "epoch": 0.25, "learning_rate": 1.7543035365618386e-06, "loss": 1.2257, "step": 3236 }, { "epoch": 0.25, "learning_rate": 1.7541384701792295e-06, "loss": 1.1889, "step": 3237 }, { "epoch": 0.25, "learning_rate": 1.7539733561382377e-06, "loss": 1.2058, "step": 3238 }, { "epoch": 0.25, "learning_rate": 1.7538081944492977e-06, "loss": 1.3009, "step": 3239 }, { "epoch": 0.25, "learning_rate": 1.7536429851228473e-06, "loss": 1.2288, "step": 3240 }, { "epoch": 0.25, "learning_rate": 1.753477728169327e-06, "loss": 1.3302, "step": 3241 }, { "epoch": 0.25, "learning_rate": 1.7533124235991804e-06, "loss": 1.2356, "step": 3242 }, { "epoch": 0.25, "learning_rate": 1.7531470714228538e-06, "loss": 1.2057, "step": 3243 }, { "epoch": 0.25, "learning_rate": 1.752981671650797e-06, "loss": 1.1809, "step": 3244 }, { "epoch": 0.25, "learning_rate": 1.7528162242934621e-06, "loss": 1.2867, "step": 3245 }, { "epoch": 0.25, "learning_rate": 1.7526507293613051e-06, "loss": 1.2118, "step": 3246 }, { "epoch": 0.25, "learning_rate": 1.7524851868647844e-06, "loss": 1.2496, "step": 3247 }, { "epoch": 0.25, "learning_rate": 1.7523195968143623e-06, "loss": 1.2305, "step": 3248 }, { "epoch": 0.25, "learning_rate": 1.7521539592205023e-06, "loss": 1.1804, "step": 3249 }, { "epoch": 0.25, "learning_rate": 1.7519882740936725e-06, "loss": 1.1985, "step": 3250 }, { "epoch": 0.25, "learning_rate": 1.7518225414443438e-06, "loss": 1.2247, "step": 3251 }, { "epoch": 0.25, "learning_rate": 1.7516567612829891e-06, "loss": 1.2411, "step": 3252 }, { "epoch": 0.25, "learning_rate": 1.7514909336200856e-06, "loss": 1.2758, "step": 3253 }, { "epoch": 0.25, "learning_rate": 1.7513250584661128e-06, "loss": 1.2141, "step": 3254 }, { "epoch": 0.25, "learning_rate": 1.7511591358315532e-06, "loss": 1.2007, "step": 3255 }, { "epoch": 0.25, "learning_rate": 1.7509931657268925e-06, "loss": 0.8384, "step": 3256 }, { "epoch": 0.25, "learning_rate": 1.7508271481626191e-06, "loss": 1.2562, "step": 3257 }, { "epoch": 0.25, "learning_rate": 1.7506610831492248e-06, "loss": 1.1657, "step": 3258 }, { "epoch": 0.25, "learning_rate": 1.7504949706972042e-06, "loss": 1.211, "step": 3259 }, { "epoch": 0.25, "learning_rate": 1.7503288108170549e-06, "loss": 0.8583, "step": 3260 }, { "epoch": 0.25, "learning_rate": 1.7501626035192775e-06, "loss": 1.1819, "step": 3261 }, { "epoch": 0.25, "learning_rate": 1.7499963488143759e-06, "loss": 1.2721, "step": 3262 }, { "epoch": 0.25, "learning_rate": 1.749830046712856e-06, "loss": 1.2416, "step": 3263 }, { "epoch": 0.25, "learning_rate": 1.749663697225228e-06, "loss": 1.2259, "step": 3264 }, { "epoch": 0.25, "learning_rate": 1.7494973003620044e-06, "loss": 1.3066, "step": 3265 }, { "epoch": 0.25, "learning_rate": 1.7493308561337006e-06, "loss": 1.1754, "step": 3266 }, { "epoch": 0.25, "learning_rate": 1.749164364550835e-06, "loss": 1.2012, "step": 3267 }, { "epoch": 0.25, "learning_rate": 1.7489978256239297e-06, "loss": 1.2431, "step": 3268 }, { "epoch": 0.25, "learning_rate": 1.7488312393635089e-06, "loss": 1.2333, "step": 3269 }, { "epoch": 0.25, "learning_rate": 1.7486646057801005e-06, "loss": 1.2507, "step": 3270 }, { "epoch": 0.25, "learning_rate": 1.7484979248842342e-06, "loss": 1.2474, "step": 3271 }, { "epoch": 0.25, "learning_rate": 1.7483311966864445e-06, "loss": 1.2642, "step": 3272 }, { "epoch": 0.25, "learning_rate": 1.7481644211972673e-06, "loss": 1.2385, "step": 3273 }, { "epoch": 0.25, "learning_rate": 1.7479975984272424e-06, "loss": 1.2312, "step": 3274 }, { "epoch": 0.25, "learning_rate": 1.7478307283869125e-06, "loss": 1.2282, "step": 3275 }, { "epoch": 0.25, "learning_rate": 1.7476638110868225e-06, "loss": 1.258, "step": 3276 }, { "epoch": 0.25, "learning_rate": 1.7474968465375212e-06, "loss": 1.2181, "step": 3277 }, { "epoch": 0.25, "learning_rate": 1.7473298347495602e-06, "loss": 0.8311, "step": 3278 }, { "epoch": 0.25, "learning_rate": 1.7471627757334936e-06, "loss": 1.2705, "step": 3279 }, { "epoch": 0.25, "learning_rate": 1.746995669499879e-06, "loss": 0.816, "step": 3280 }, { "epoch": 0.25, "learning_rate": 1.7468285160592773e-06, "loss": 1.2475, "step": 3281 }, { "epoch": 0.25, "learning_rate": 1.746661315422251e-06, "loss": 1.2653, "step": 3282 }, { "epoch": 0.25, "learning_rate": 1.7464940675993672e-06, "loss": 1.2454, "step": 3283 }, { "epoch": 0.25, "learning_rate": 1.7463267726011946e-06, "loss": 1.2505, "step": 3284 }, { "epoch": 0.25, "learning_rate": 1.746159430438306e-06, "loss": 1.2204, "step": 3285 }, { "epoch": 0.26, "learning_rate": 1.7459920411212768e-06, "loss": 1.2263, "step": 3286 }, { "epoch": 0.26, "learning_rate": 1.745824604660685e-06, "loss": 1.2493, "step": 3287 }, { "epoch": 0.26, "learning_rate": 1.7456571210671118e-06, "loss": 1.2169, "step": 3288 }, { "epoch": 0.26, "learning_rate": 1.7454895903511422e-06, "loss": 1.2357, "step": 3289 }, { "epoch": 0.26, "learning_rate": 1.7453220125233624e-06, "loss": 1.2296, "step": 3290 }, { "epoch": 0.26, "learning_rate": 1.7451543875943635e-06, "loss": 1.2165, "step": 3291 }, { "epoch": 0.26, "learning_rate": 1.744986715574738e-06, "loss": 1.202, "step": 3292 }, { "epoch": 0.26, "learning_rate": 1.7448189964750827e-06, "loss": 0.8507, "step": 3293 }, { "epoch": 0.26, "learning_rate": 1.744651230305996e-06, "loss": 1.3051, "step": 3294 }, { "epoch": 0.26, "learning_rate": 1.7444834170780806e-06, "loss": 1.167, "step": 3295 }, { "epoch": 0.26, "learning_rate": 1.7443155568019416e-06, "loss": 0.8761, "step": 3296 }, { "epoch": 0.26, "learning_rate": 1.7441476494881871e-06, "loss": 1.2722, "step": 3297 }, { "epoch": 0.26, "learning_rate": 1.7439796951474274e-06, "loss": 0.8361, "step": 3298 }, { "epoch": 0.26, "learning_rate": 1.7438116937902772e-06, "loss": 1.2847, "step": 3299 }, { "epoch": 0.26, "learning_rate": 1.7436436454273533e-06, "loss": 1.2047, "step": 3300 }, { "epoch": 0.26, "learning_rate": 1.743475550069276e-06, "loss": 1.2319, "step": 3301 }, { "epoch": 0.26, "learning_rate": 1.7433074077266674e-06, "loss": 1.2227, "step": 3302 }, { "epoch": 0.26, "learning_rate": 1.7431392184101542e-06, "loss": 1.2696, "step": 3303 }, { "epoch": 0.26, "learning_rate": 1.7429709821303648e-06, "loss": 1.1413, "step": 3304 }, { "epoch": 0.26, "learning_rate": 1.7428026988979313e-06, "loss": 0.8646, "step": 3305 }, { "epoch": 0.26, "learning_rate": 1.7426343687234881e-06, "loss": 1.2509, "step": 3306 }, { "epoch": 0.26, "learning_rate": 1.7424659916176734e-06, "loss": 1.1762, "step": 3307 }, { "epoch": 0.26, "learning_rate": 1.742297567591128e-06, "loss": 1.2318, "step": 3308 }, { "epoch": 0.26, "learning_rate": 1.7421290966544948e-06, "loss": 1.1979, "step": 3309 }, { "epoch": 0.26, "learning_rate": 1.7419605788184214e-06, "loss": 1.1928, "step": 3310 }, { "epoch": 0.26, "learning_rate": 1.7417920140935567e-06, "loss": 0.8265, "step": 3311 }, { "epoch": 0.26, "learning_rate": 1.7416234024905536e-06, "loss": 1.2523, "step": 3312 }, { "epoch": 0.26, "learning_rate": 1.7414547440200678e-06, "loss": 0.865, "step": 3313 }, { "epoch": 0.26, "learning_rate": 1.7412860386927577e-06, "loss": 1.1949, "step": 3314 }, { "epoch": 0.26, "learning_rate": 1.7411172865192847e-06, "loss": 1.2189, "step": 3315 }, { "epoch": 0.26, "learning_rate": 1.740948487510313e-06, "loss": 1.2266, "step": 3316 }, { "epoch": 0.26, "learning_rate": 1.7407796416765105e-06, "loss": 1.2181, "step": 3317 }, { "epoch": 0.26, "learning_rate": 1.740610749028547e-06, "loss": 1.2099, "step": 3318 }, { "epoch": 0.26, "learning_rate": 1.740441809577096e-06, "loss": 1.2343, "step": 3319 }, { "epoch": 0.26, "learning_rate": 1.7402728233328343e-06, "loss": 1.1972, "step": 3320 }, { "epoch": 0.26, "learning_rate": 1.7401037903064405e-06, "loss": 1.1885, "step": 3321 }, { "epoch": 0.26, "learning_rate": 1.7399347105085967e-06, "loss": 1.2373, "step": 3322 }, { "epoch": 0.26, "learning_rate": 1.7397655839499886e-06, "loss": 1.2462, "step": 3323 }, { "epoch": 0.26, "learning_rate": 1.7395964106413036e-06, "loss": 1.2874, "step": 3324 }, { "epoch": 0.26, "learning_rate": 1.7394271905932333e-06, "loss": 1.2038, "step": 3325 }, { "epoch": 0.26, "learning_rate": 1.7392579238164715e-06, "loss": 1.236, "step": 3326 }, { "epoch": 0.26, "learning_rate": 1.739088610321715e-06, "loss": 1.2148, "step": 3327 }, { "epoch": 0.26, "learning_rate": 1.738919250119664e-06, "loss": 1.2306, "step": 3328 }, { "epoch": 0.26, "learning_rate": 1.7387498432210212e-06, "loss": 1.2118, "step": 3329 }, { "epoch": 0.26, "learning_rate": 1.7385803896364924e-06, "loss": 1.2843, "step": 3330 }, { "epoch": 0.26, "learning_rate": 1.7384108893767865e-06, "loss": 1.2369, "step": 3331 }, { "epoch": 0.26, "learning_rate": 1.7382413424526148e-06, "loss": 1.2091, "step": 3332 }, { "epoch": 0.26, "learning_rate": 1.7380717488746921e-06, "loss": 1.2426, "step": 3333 }, { "epoch": 0.26, "learning_rate": 1.7379021086537364e-06, "loss": 1.2234, "step": 3334 }, { "epoch": 0.26, "learning_rate": 1.7377324218004678e-06, "loss": 1.1982, "step": 3335 }, { "epoch": 0.26, "learning_rate": 1.7375626883256103e-06, "loss": 1.2546, "step": 3336 }, { "epoch": 0.26, "learning_rate": 1.7373929082398897e-06, "loss": 1.2291, "step": 3337 }, { "epoch": 0.26, "learning_rate": 1.7372230815540356e-06, "loss": 0.8886, "step": 3338 }, { "epoch": 0.26, "learning_rate": 1.7370532082787805e-06, "loss": 1.2361, "step": 3339 }, { "epoch": 0.26, "learning_rate": 1.7368832884248596e-06, "loss": 0.8569, "step": 3340 }, { "epoch": 0.26, "learning_rate": 1.7367133220030115e-06, "loss": 1.2486, "step": 3341 }, { "epoch": 0.26, "learning_rate": 1.7365433090239766e-06, "loss": 1.204, "step": 3342 }, { "epoch": 0.26, "learning_rate": 1.7363732494984995e-06, "loss": 1.2007, "step": 3343 }, { "epoch": 0.26, "learning_rate": 1.736203143437327e-06, "loss": 1.218, "step": 3344 }, { "epoch": 0.26, "learning_rate": 1.7360329908512095e-06, "loss": 1.2177, "step": 3345 }, { "epoch": 0.26, "learning_rate": 1.7358627917508994e-06, "loss": 1.2124, "step": 3346 }, { "epoch": 0.26, "learning_rate": 1.735692546147153e-06, "loss": 1.2486, "step": 3347 }, { "epoch": 0.26, "learning_rate": 1.7355222540507288e-06, "loss": 1.2163, "step": 3348 }, { "epoch": 0.26, "learning_rate": 1.7353519154723891e-06, "loss": 1.2135, "step": 3349 }, { "epoch": 0.26, "learning_rate": 1.7351815304228976e-06, "loss": 1.2126, "step": 3350 }, { "epoch": 0.26, "learning_rate": 1.7350110989130227e-06, "loss": 1.2316, "step": 3351 }, { "epoch": 0.26, "learning_rate": 1.734840620953535e-06, "loss": 1.2793, "step": 3352 }, { "epoch": 0.26, "learning_rate": 1.7346700965552073e-06, "loss": 1.2458, "step": 3353 }, { "epoch": 0.26, "learning_rate": 1.7344995257288163e-06, "loss": 1.2084, "step": 3354 }, { "epoch": 0.26, "learning_rate": 1.734328908485142e-06, "loss": 1.1883, "step": 3355 }, { "epoch": 0.26, "learning_rate": 1.734158244834966e-06, "loss": 1.2371, "step": 3356 }, { "epoch": 0.26, "learning_rate": 1.7339875347890738e-06, "loss": 1.2543, "step": 3357 }, { "epoch": 0.26, "learning_rate": 1.7338167783582533e-06, "loss": 0.8575, "step": 3358 }, { "epoch": 0.26, "learning_rate": 1.733645975553296e-06, "loss": 1.2677, "step": 3359 }, { "epoch": 0.26, "learning_rate": 1.7334751263849955e-06, "loss": 1.2305, "step": 3360 }, { "epoch": 0.26, "learning_rate": 1.7333042308641492e-06, "loss": 1.2226, "step": 3361 }, { "epoch": 0.26, "learning_rate": 1.7331332890015566e-06, "loss": 0.834, "step": 3362 }, { "epoch": 0.26, "learning_rate": 1.7329623008080207e-06, "loss": 1.2308, "step": 3363 }, { "epoch": 0.26, "learning_rate": 1.7327912662943472e-06, "loss": 1.2569, "step": 3364 }, { "epoch": 0.26, "learning_rate": 1.7326201854713442e-06, "loss": 1.2022, "step": 3365 }, { "epoch": 0.26, "learning_rate": 1.7324490583498244e-06, "loss": 1.1945, "step": 3366 }, { "epoch": 0.26, "learning_rate": 1.7322778849406014e-06, "loss": 1.2035, "step": 3367 }, { "epoch": 0.26, "learning_rate": 1.7321066652544932e-06, "loss": 1.2061, "step": 3368 }, { "epoch": 0.26, "learning_rate": 1.7319353993023201e-06, "loss": 1.2828, "step": 3369 }, { "epoch": 0.26, "learning_rate": 1.7317640870949051e-06, "loss": 1.2354, "step": 3370 }, { "epoch": 0.26, "learning_rate": 1.7315927286430744e-06, "loss": 1.3139, "step": 3371 }, { "epoch": 0.26, "learning_rate": 1.7314213239576572e-06, "loss": 1.2088, "step": 3372 }, { "epoch": 0.26, "learning_rate": 1.7312498730494856e-06, "loss": 1.1923, "step": 3373 }, { "epoch": 0.26, "learning_rate": 1.7310783759293947e-06, "loss": 1.2498, "step": 3374 }, { "epoch": 0.26, "learning_rate": 1.7309068326082224e-06, "loss": 1.2192, "step": 3375 }, { "epoch": 0.26, "learning_rate": 1.7307352430968094e-06, "loss": 1.2802, "step": 3376 }, { "epoch": 0.26, "learning_rate": 1.7305636074059992e-06, "loss": 1.1202, "step": 3377 }, { "epoch": 0.26, "learning_rate": 1.7303919255466393e-06, "loss": 1.2437, "step": 3378 }, { "epoch": 0.26, "learning_rate": 1.730220197529578e-06, "loss": 1.269, "step": 3379 }, { "epoch": 0.26, "learning_rate": 1.7300484233656692e-06, "loss": 1.2779, "step": 3380 }, { "epoch": 0.26, "learning_rate": 1.729876603065767e-06, "loss": 1.2322, "step": 3381 }, { "epoch": 0.26, "learning_rate": 1.7297047366407304e-06, "loss": 0.8346, "step": 3382 }, { "epoch": 0.26, "learning_rate": 1.7295328241014207e-06, "loss": 1.1785, "step": 3383 }, { "epoch": 0.26, "learning_rate": 1.7293608654587017e-06, "loss": 1.2449, "step": 3384 }, { "epoch": 0.26, "learning_rate": 1.7291888607234406e-06, "loss": 1.2233, "step": 3385 }, { "epoch": 0.26, "learning_rate": 1.7290168099065078e-06, "loss": 1.2893, "step": 3386 }, { "epoch": 0.26, "learning_rate": 1.7288447130187755e-06, "loss": 1.2232, "step": 3387 }, { "epoch": 0.26, "learning_rate": 1.7286725700711197e-06, "loss": 1.2736, "step": 3388 }, { "epoch": 0.26, "learning_rate": 1.7285003810744195e-06, "loss": 1.2074, "step": 3389 }, { "epoch": 0.26, "learning_rate": 1.7283281460395563e-06, "loss": 0.8618, "step": 3390 }, { "epoch": 0.26, "learning_rate": 1.7281558649774142e-06, "loss": 1.2661, "step": 3391 }, { "epoch": 0.26, "learning_rate": 1.7279835378988814e-06, "loss": 1.2224, "step": 3392 }, { "epoch": 0.26, "learning_rate": 1.7278111648148474e-06, "loss": 1.2036, "step": 3393 }, { "epoch": 0.26, "learning_rate": 1.7276387457362063e-06, "loss": 0.8433, "step": 3394 }, { "epoch": 0.26, "learning_rate": 1.7274662806738537e-06, "loss": 1.2173, "step": 3395 }, { "epoch": 0.26, "learning_rate": 1.727293769638689e-06, "loss": 1.2258, "step": 3396 }, { "epoch": 0.26, "learning_rate": 1.7271212126416141e-06, "loss": 0.827, "step": 3397 }, { "epoch": 0.26, "learning_rate": 1.7269486096935336e-06, "loss": 1.2272, "step": 3398 }, { "epoch": 0.26, "learning_rate": 1.7267759608053555e-06, "loss": 1.1802, "step": 3399 }, { "epoch": 0.26, "learning_rate": 1.7266032659879905e-06, "loss": 1.2389, "step": 3400 }, { "epoch": 0.26, "learning_rate": 1.726430525252352e-06, "loss": 1.2042, "step": 3401 }, { "epoch": 0.26, "learning_rate": 1.726257738609357e-06, "loss": 1.2822, "step": 3402 }, { "epoch": 0.26, "learning_rate": 1.7260849060699242e-06, "loss": 0.828, "step": 3403 }, { "epoch": 0.26, "learning_rate": 1.7259120276449766e-06, "loss": 1.307, "step": 3404 }, { "epoch": 0.26, "learning_rate": 1.7257391033454386e-06, "loss": 0.8347, "step": 3405 }, { "epoch": 0.26, "learning_rate": 1.7255661331822386e-06, "loss": 1.2646, "step": 3406 }, { "epoch": 0.26, "learning_rate": 1.725393117166308e-06, "loss": 1.2397, "step": 3407 }, { "epoch": 0.26, "learning_rate": 1.7252200553085805e-06, "loss": 1.2314, "step": 3408 }, { "epoch": 0.26, "learning_rate": 1.7250469476199925e-06, "loss": 1.2186, "step": 3409 }, { "epoch": 0.26, "learning_rate": 1.7248737941114837e-06, "loss": 1.2703, "step": 3410 }, { "epoch": 0.26, "learning_rate": 1.7247005947939975e-06, "loss": 1.2873, "step": 3411 }, { "epoch": 0.26, "learning_rate": 1.7245273496784782e-06, "loss": 1.2697, "step": 3412 }, { "epoch": 0.26, "learning_rate": 1.7243540587758751e-06, "loss": 1.2129, "step": 3413 }, { "epoch": 0.26, "learning_rate": 1.724180722097139e-06, "loss": 0.8356, "step": 3414 }, { "epoch": 0.27, "learning_rate": 1.724007339653224e-06, "loss": 1.2201, "step": 3415 }, { "epoch": 0.27, "learning_rate": 1.7238339114550873e-06, "loss": 1.1671, "step": 3416 }, { "epoch": 0.27, "learning_rate": 1.723660437513689e-06, "loss": 1.2016, "step": 3417 }, { "epoch": 0.27, "learning_rate": 1.7234869178399917e-06, "loss": 1.2574, "step": 3418 }, { "epoch": 0.27, "learning_rate": 1.723313352444961e-06, "loss": 1.1708, "step": 3419 }, { "epoch": 0.27, "learning_rate": 1.723139741339566e-06, "loss": 1.2554, "step": 3420 }, { "epoch": 0.27, "learning_rate": 1.7229660845347773e-06, "loss": 1.2754, "step": 3421 }, { "epoch": 0.27, "learning_rate": 1.7227923820415702e-06, "loss": 1.2237, "step": 3422 }, { "epoch": 0.27, "learning_rate": 1.7226186338709214e-06, "loss": 1.2338, "step": 3423 }, { "epoch": 0.27, "learning_rate": 1.7224448400338117e-06, "loss": 1.1949, "step": 3424 }, { "epoch": 0.27, "learning_rate": 1.7222710005412234e-06, "loss": 1.269, "step": 3425 }, { "epoch": 0.27, "learning_rate": 1.7220971154041423e-06, "loss": 1.2407, "step": 3426 }, { "epoch": 0.27, "learning_rate": 1.7219231846335582e-06, "loss": 1.2387, "step": 3427 }, { "epoch": 0.27, "learning_rate": 1.721749208240462e-06, "loss": 1.2184, "step": 3428 }, { "epoch": 0.27, "learning_rate": 1.721575186235849e-06, "loss": 1.2118, "step": 3429 }, { "epoch": 0.27, "learning_rate": 1.7214011186307155e-06, "loss": 1.2498, "step": 3430 }, { "epoch": 0.27, "learning_rate": 1.7212270054360627e-06, "loss": 0.8738, "step": 3431 }, { "epoch": 0.27, "learning_rate": 1.7210528466628937e-06, "loss": 0.8551, "step": 3432 }, { "epoch": 0.27, "learning_rate": 1.7208786423222144e-06, "loss": 0.8625, "step": 3433 }, { "epoch": 0.27, "learning_rate": 1.7207043924250339e-06, "loss": 1.2983, "step": 3434 }, { "epoch": 0.27, "learning_rate": 1.7205300969823645e-06, "loss": 1.2776, "step": 3435 }, { "epoch": 0.27, "learning_rate": 1.72035575600522e-06, "loss": 0.8517, "step": 3436 }, { "epoch": 0.27, "learning_rate": 1.7201813695046185e-06, "loss": 1.2699, "step": 3437 }, { "epoch": 0.27, "learning_rate": 1.7200069374915808e-06, "loss": 1.276, "step": 3438 }, { "epoch": 0.27, "learning_rate": 1.71983245997713e-06, "loss": 0.8939, "step": 3439 }, { "epoch": 0.27, "learning_rate": 1.7196579369722923e-06, "loss": 1.249, "step": 3440 }, { "epoch": 0.27, "learning_rate": 1.7194833684880966e-06, "loss": 1.1797, "step": 3441 }, { "epoch": 0.27, "learning_rate": 1.7193087545355755e-06, "loss": 1.2399, "step": 3442 }, { "epoch": 0.27, "learning_rate": 1.7191340951257635e-06, "loss": 0.8538, "step": 3443 }, { "epoch": 0.27, "learning_rate": 1.718959390269698e-06, "loss": 1.2647, "step": 3444 }, { "epoch": 0.27, "learning_rate": 1.7187846399784203e-06, "loss": 1.1991, "step": 3445 }, { "epoch": 0.27, "learning_rate": 1.7186098442629737e-06, "loss": 1.2897, "step": 3446 }, { "epoch": 0.27, "learning_rate": 1.7184350031344042e-06, "loss": 1.2347, "step": 3447 }, { "epoch": 0.27, "learning_rate": 1.7182601166037611e-06, "loss": 0.8468, "step": 3448 }, { "epoch": 0.27, "learning_rate": 1.718085184682097e-06, "loss": 1.2399, "step": 3449 }, { "epoch": 0.27, "learning_rate": 1.7179102073804658e-06, "loss": 1.2897, "step": 3450 }, { "epoch": 0.27, "learning_rate": 1.717735184709927e-06, "loss": 1.2403, "step": 3451 }, { "epoch": 0.27, "learning_rate": 1.7175601166815397e-06, "loss": 1.2304, "step": 3452 }, { "epoch": 0.27, "learning_rate": 1.7173850033063682e-06, "loss": 1.2471, "step": 3453 }, { "epoch": 0.27, "learning_rate": 1.7172098445954787e-06, "loss": 1.2491, "step": 3454 }, { "epoch": 0.27, "learning_rate": 1.7170346405599408e-06, "loss": 1.2268, "step": 3455 }, { "epoch": 0.27, "learning_rate": 1.7168593912108263e-06, "loss": 1.2743, "step": 3456 }, { "epoch": 0.27, "learning_rate": 1.7166840965592104e-06, "loss": 1.2092, "step": 3457 }, { "epoch": 0.27, "learning_rate": 1.7165087566161711e-06, "loss": 1.1962, "step": 3458 }, { "epoch": 0.27, "learning_rate": 1.716333371392789e-06, "loss": 1.2615, "step": 3459 }, { "epoch": 0.27, "learning_rate": 1.7161579409001477e-06, "loss": 1.2484, "step": 3460 }, { "epoch": 0.27, "learning_rate": 1.7159824651493338e-06, "loss": 1.236, "step": 3461 }, { "epoch": 0.27, "learning_rate": 1.7158069441514365e-06, "loss": 0.8288, "step": 3462 }, { "epoch": 0.27, "learning_rate": 1.715631377917548e-06, "loss": 1.217, "step": 3463 }, { "epoch": 0.27, "learning_rate": 1.7154557664587632e-06, "loss": 1.1934, "step": 3464 }, { "epoch": 0.27, "learning_rate": 1.7152801097861805e-06, "loss": 1.2566, "step": 3465 }, { "epoch": 0.27, "learning_rate": 1.7151044079109e-06, "loss": 1.2155, "step": 3466 }, { "epoch": 0.27, "learning_rate": 1.7149286608440256e-06, "loss": 1.217, "step": 3467 }, { "epoch": 0.27, "learning_rate": 1.7147528685966644e-06, "loss": 1.1977, "step": 3468 }, { "epoch": 0.27, "learning_rate": 1.714577031179925e-06, "loss": 1.2255, "step": 3469 }, { "epoch": 0.27, "learning_rate": 1.7144011486049196e-06, "loss": 1.2819, "step": 3470 }, { "epoch": 0.27, "learning_rate": 1.714225220882763e-06, "loss": 1.2629, "step": 3471 }, { "epoch": 0.27, "learning_rate": 1.7140492480245739e-06, "loss": 1.2549, "step": 3472 }, { "epoch": 0.27, "learning_rate": 1.713873230041473e-06, "loss": 1.2716, "step": 3473 }, { "epoch": 0.27, "learning_rate": 1.7136971669445829e-06, "loss": 1.2647, "step": 3474 }, { "epoch": 0.27, "learning_rate": 1.713521058745031e-06, "loss": 1.2569, "step": 3475 }, { "epoch": 0.27, "learning_rate": 1.7133449054539463e-06, "loss": 1.2724, "step": 3476 }, { "epoch": 0.27, "learning_rate": 1.7131687070824606e-06, "loss": 1.2753, "step": 3477 }, { "epoch": 0.27, "learning_rate": 1.7129924636417092e-06, "loss": 1.2705, "step": 3478 }, { "epoch": 0.27, "learning_rate": 1.7128161751428306e-06, "loss": 1.2703, "step": 3479 }, { "epoch": 0.27, "learning_rate": 1.7126398415969644e-06, "loss": 1.2572, "step": 3480 }, { "epoch": 0.27, "learning_rate": 1.7124634630152547e-06, "loss": 1.2426, "step": 3481 }, { "epoch": 0.27, "learning_rate": 1.7122870394088476e-06, "loss": 1.2409, "step": 3482 }, { "epoch": 0.27, "learning_rate": 1.7121105707888925e-06, "loss": 1.2468, "step": 3483 }, { "epoch": 0.27, "learning_rate": 1.711934057166542e-06, "loss": 1.2252, "step": 3484 }, { "epoch": 0.27, "learning_rate": 1.71175749855295e-06, "loss": 1.2614, "step": 3485 }, { "epoch": 0.27, "learning_rate": 1.711580894959275e-06, "loss": 1.2624, "step": 3486 }, { "epoch": 0.27, "learning_rate": 1.7114042463966772e-06, "loss": 0.8628, "step": 3487 }, { "epoch": 0.27, "learning_rate": 1.7112275528763202e-06, "loss": 1.2128, "step": 3488 }, { "epoch": 0.27, "learning_rate": 1.7110508144093705e-06, "loss": 1.2524, "step": 3489 }, { "epoch": 0.27, "learning_rate": 1.710874031006997e-06, "loss": 1.2322, "step": 3490 }, { "epoch": 0.27, "learning_rate": 1.7106972026803716e-06, "loss": 1.2774, "step": 3491 }, { "epoch": 0.27, "learning_rate": 1.7105203294406694e-06, "loss": 1.2143, "step": 3492 }, { "epoch": 0.27, "learning_rate": 1.7103434112990678e-06, "loss": 1.2065, "step": 3493 }, { "epoch": 0.27, "learning_rate": 1.7101664482667472e-06, "loss": 1.2111, "step": 3494 }, { "epoch": 0.27, "learning_rate": 1.7099894403548913e-06, "loss": 1.281, "step": 3495 }, { "epoch": 0.27, "learning_rate": 1.7098123875746859e-06, "loss": 0.8565, "step": 3496 }, { "epoch": 0.27, "learning_rate": 1.7096352899373198e-06, "loss": 1.2084, "step": 3497 }, { "epoch": 0.27, "learning_rate": 1.7094581474539855e-06, "loss": 1.2311, "step": 3498 }, { "epoch": 0.27, "learning_rate": 1.7092809601358773e-06, "loss": 1.2045, "step": 3499 }, { "epoch": 0.27, "learning_rate": 1.7091037279941926e-06, "loss": 1.2477, "step": 3500 }, { "epoch": 0.27, "learning_rate": 1.7089264510401318e-06, "loss": 1.2282, "step": 3501 }, { "epoch": 0.27, "learning_rate": 1.7087491292848981e-06, "loss": 1.245, "step": 3502 }, { "epoch": 0.27, "learning_rate": 1.7085717627396977e-06, "loss": 1.2428, "step": 3503 }, { "epoch": 0.27, "learning_rate": 1.7083943514157388e-06, "loss": 1.191, "step": 3504 }, { "epoch": 0.27, "learning_rate": 1.7082168953242336e-06, "loss": 1.2514, "step": 3505 }, { "epoch": 0.27, "learning_rate": 1.7080393944763965e-06, "loss": 1.2137, "step": 3506 }, { "epoch": 0.27, "learning_rate": 1.7078618488834446e-06, "loss": 1.2802, "step": 3507 }, { "epoch": 0.27, "learning_rate": 1.7076842585565983e-06, "loss": 1.202, "step": 3508 }, { "epoch": 0.27, "learning_rate": 1.70750662350708e-06, "loss": 1.1923, "step": 3509 }, { "epoch": 0.27, "learning_rate": 1.7073289437461164e-06, "loss": 1.2625, "step": 3510 }, { "epoch": 0.27, "learning_rate": 1.7071512192849356e-06, "loss": 1.2638, "step": 3511 }, { "epoch": 0.27, "learning_rate": 1.706973450134769e-06, "loss": 1.1931, "step": 3512 }, { "epoch": 0.27, "learning_rate": 1.7067956363068508e-06, "loss": 1.2601, "step": 3513 }, { "epoch": 0.27, "learning_rate": 1.7066177778124182e-06, "loss": 1.2688, "step": 3514 }, { "epoch": 0.27, "learning_rate": 1.706439874662711e-06, "loss": 1.221, "step": 3515 }, { "epoch": 0.27, "learning_rate": 1.7062619268689724e-06, "loss": 1.2704, "step": 3516 }, { "epoch": 0.27, "learning_rate": 1.7060839344424477e-06, "loss": 1.2579, "step": 3517 }, { "epoch": 0.27, "learning_rate": 1.705905897394385e-06, "loss": 1.2362, "step": 3518 }, { "epoch": 0.27, "learning_rate": 1.7057278157360355e-06, "loss": 0.8437, "step": 3519 }, { "epoch": 0.27, "learning_rate": 1.7055496894786534e-06, "loss": 1.2156, "step": 3520 }, { "epoch": 0.27, "learning_rate": 1.7053715186334955e-06, "loss": 1.2584, "step": 3521 }, { "epoch": 0.27, "learning_rate": 1.7051933032118217e-06, "loss": 1.2583, "step": 3522 }, { "epoch": 0.27, "learning_rate": 1.7050150432248943e-06, "loss": 1.2893, "step": 3523 }, { "epoch": 0.27, "learning_rate": 1.704836738683978e-06, "loss": 0.8891, "step": 3524 }, { "epoch": 0.27, "learning_rate": 1.704658389600342e-06, "loss": 1.2111, "step": 3525 }, { "epoch": 0.27, "learning_rate": 1.7044799959852566e-06, "loss": 1.2631, "step": 3526 }, { "epoch": 0.27, "learning_rate": 1.704301557849995e-06, "loss": 1.1757, "step": 3527 }, { "epoch": 0.27, "learning_rate": 1.7041230752058348e-06, "loss": 1.2509, "step": 3528 }, { "epoch": 0.27, "learning_rate": 1.7039445480640547e-06, "loss": 0.8709, "step": 3529 }, { "epoch": 0.27, "learning_rate": 1.703765976435937e-06, "loss": 1.2691, "step": 3530 }, { "epoch": 0.27, "learning_rate": 1.7035873603327666e-06, "loss": 1.2612, "step": 3531 }, { "epoch": 0.27, "learning_rate": 1.7034086997658315e-06, "loss": 1.2079, "step": 3532 }, { "epoch": 0.27, "learning_rate": 1.7032299947464222e-06, "loss": 1.2066, "step": 3533 }, { "epoch": 0.27, "learning_rate": 1.7030512452858318e-06, "loss": 1.3011, "step": 3534 }, { "epoch": 0.27, "learning_rate": 1.7028724513953572e-06, "loss": 1.2004, "step": 3535 }, { "epoch": 0.27, "learning_rate": 1.7026936130862968e-06, "loss": 1.2066, "step": 3536 }, { "epoch": 0.27, "learning_rate": 1.7025147303699527e-06, "loss": 1.2526, "step": 3537 }, { "epoch": 0.27, "learning_rate": 1.7023358032576295e-06, "loss": 1.2391, "step": 3538 }, { "epoch": 0.27, "learning_rate": 1.7021568317606345e-06, "loss": 0.8115, "step": 3539 }, { "epoch": 0.27, "learning_rate": 1.7019778158902783e-06, "loss": 1.2306, "step": 3540 }, { "epoch": 0.27, "learning_rate": 1.7017987556578735e-06, "loss": 1.2553, "step": 3541 }, { "epoch": 0.27, "learning_rate": 1.7016196510747362e-06, "loss": 1.1997, "step": 3542 }, { "epoch": 0.27, "learning_rate": 1.701440502152185e-06, "loss": 1.2627, "step": 3543 }, { "epoch": 0.28, "learning_rate": 1.7012613089015418e-06, "loss": 1.2392, "step": 3544 }, { "epoch": 0.28, "learning_rate": 1.7010820713341301e-06, "loss": 1.2349, "step": 3545 }, { "epoch": 0.28, "learning_rate": 1.7009027894612774e-06, "loss": 1.2021, "step": 3546 }, { "epoch": 0.28, "learning_rate": 1.700723463294313e-06, "loss": 1.2462, "step": 3547 }, { "epoch": 0.28, "learning_rate": 1.7005440928445708e-06, "loss": 0.855, "step": 3548 }, { "epoch": 0.28, "learning_rate": 1.700364678123385e-06, "loss": 0.8295, "step": 3549 }, { "epoch": 0.28, "learning_rate": 1.7001852191420945e-06, "loss": 1.1992, "step": 3550 }, { "epoch": 0.28, "learning_rate": 1.7000057159120403e-06, "loss": 1.2725, "step": 3551 }, { "epoch": 0.28, "learning_rate": 1.6998261684445662e-06, "loss": 1.2607, "step": 3552 }, { "epoch": 0.28, "learning_rate": 1.6996465767510183e-06, "loss": 1.2169, "step": 3553 }, { "epoch": 0.28, "learning_rate": 1.6994669408427469e-06, "loss": 0.8509, "step": 3554 }, { "epoch": 0.28, "learning_rate": 1.6992872607311041e-06, "loss": 1.2409, "step": 3555 }, { "epoch": 0.28, "learning_rate": 1.6991075364274446e-06, "loss": 0.8302, "step": 3556 }, { "epoch": 0.28, "learning_rate": 1.698927767943126e-06, "loss": 1.2401, "step": 3557 }, { "epoch": 0.28, "learning_rate": 1.6987479552895094e-06, "loss": 1.2926, "step": 3558 }, { "epoch": 0.28, "learning_rate": 1.698568098477958e-06, "loss": 1.2785, "step": 3559 }, { "epoch": 0.28, "learning_rate": 1.6983881975198383e-06, "loss": 1.1894, "step": 3560 }, { "epoch": 0.28, "learning_rate": 1.698208252426519e-06, "loss": 1.2666, "step": 3561 }, { "epoch": 0.28, "learning_rate": 1.6980282632093715e-06, "loss": 1.2468, "step": 3562 }, { "epoch": 0.28, "learning_rate": 1.6978482298797712e-06, "loss": 1.2432, "step": 3563 }, { "epoch": 0.28, "learning_rate": 1.6976681524490945e-06, "loss": 1.2449, "step": 3564 }, { "epoch": 0.28, "learning_rate": 1.6974880309287225e-06, "loss": 1.2315, "step": 3565 }, { "epoch": 0.28, "learning_rate": 1.6973078653300372e-06, "loss": 1.2937, "step": 3566 }, { "epoch": 0.28, "learning_rate": 1.6971276556644254e-06, "loss": 1.2319, "step": 3567 }, { "epoch": 0.28, "learning_rate": 1.6969474019432746e-06, "loss": 1.2442, "step": 3568 }, { "epoch": 0.28, "learning_rate": 1.6967671041779765e-06, "loss": 1.1856, "step": 3569 }, { "epoch": 0.28, "learning_rate": 1.696586762379925e-06, "loss": 1.1909, "step": 3570 }, { "epoch": 0.28, "learning_rate": 1.6964063765605173e-06, "loss": 1.2434, "step": 3571 }, { "epoch": 0.28, "learning_rate": 1.6962259467311525e-06, "loss": 1.1952, "step": 3572 }, { "epoch": 0.28, "learning_rate": 1.6960454729032337e-06, "loss": 1.2199, "step": 3573 }, { "epoch": 0.28, "learning_rate": 1.695864955088165e-06, "loss": 0.8556, "step": 3574 }, { "epoch": 0.28, "learning_rate": 1.6956843932973555e-06, "loss": 1.2774, "step": 3575 }, { "epoch": 0.28, "learning_rate": 1.6955037875422154e-06, "loss": 1.2699, "step": 3576 }, { "epoch": 0.28, "learning_rate": 1.6953231378341583e-06, "loss": 1.2304, "step": 3577 }, { "epoch": 0.28, "learning_rate": 1.6951424441846008e-06, "loss": 1.2779, "step": 3578 }, { "epoch": 0.28, "learning_rate": 1.6949617066049614e-06, "loss": 1.2516, "step": 3579 }, { "epoch": 0.28, "learning_rate": 1.694780925106662e-06, "loss": 1.2342, "step": 3580 }, { "epoch": 0.28, "learning_rate": 1.6946000997011278e-06, "loss": 1.2205, "step": 3581 }, { "epoch": 0.28, "learning_rate": 1.694419230399786e-06, "loss": 1.1926, "step": 3582 }, { "epoch": 0.28, "learning_rate": 1.6942383172140664e-06, "loss": 0.8588, "step": 3583 }, { "epoch": 0.28, "learning_rate": 1.6940573601554023e-06, "loss": 0.8595, "step": 3584 }, { "epoch": 0.28, "learning_rate": 1.6938763592352294e-06, "loss": 1.2273, "step": 3585 }, { "epoch": 0.28, "learning_rate": 1.693695314464986e-06, "loss": 1.2329, "step": 3586 }, { "epoch": 0.28, "learning_rate": 1.6935142258561138e-06, "loss": 1.276, "step": 3587 }, { "epoch": 0.28, "learning_rate": 1.6933330934200562e-06, "loss": 1.1845, "step": 3588 }, { "epoch": 0.28, "learning_rate": 1.6931519171682607e-06, "loss": 1.2505, "step": 3589 }, { "epoch": 0.28, "learning_rate": 1.6929706971121764e-06, "loss": 1.2314, "step": 3590 }, { "epoch": 0.28, "learning_rate": 1.6927894332632557e-06, "loss": 1.273, "step": 3591 }, { "epoch": 0.28, "learning_rate": 1.6926081256329539e-06, "loss": 1.2706, "step": 3592 }, { "epoch": 0.28, "learning_rate": 1.6924267742327287e-06, "loss": 1.266, "step": 3593 }, { "epoch": 0.28, "learning_rate": 1.6922453790740407e-06, "loss": 1.2539, "step": 3594 }, { "epoch": 0.28, "learning_rate": 1.6920639401683539e-06, "loss": 1.2323, "step": 3595 }, { "epoch": 0.28, "learning_rate": 1.6918824575271336e-06, "loss": 1.2418, "step": 3596 }, { "epoch": 0.28, "learning_rate": 1.691700931161849e-06, "loss": 1.19, "step": 3597 }, { "epoch": 0.28, "learning_rate": 1.6915193610839724e-06, "loss": 1.2511, "step": 3598 }, { "epoch": 0.28, "learning_rate": 1.6913377473049774e-06, "loss": 1.2698, "step": 3599 }, { "epoch": 0.28, "learning_rate": 1.6911560898363422e-06, "loss": 1.2132, "step": 3600 }, { "epoch": 0.28, "learning_rate": 1.6909743886895458e-06, "loss": 1.2507, "step": 3601 }, { "epoch": 0.28, "learning_rate": 1.6907926438760714e-06, "loss": 0.8604, "step": 3602 }, { "epoch": 0.28, "learning_rate": 1.6906108554074045e-06, "loss": 1.2508, "step": 3603 }, { "epoch": 0.28, "learning_rate": 1.6904290232950333e-06, "loss": 1.2746, "step": 3604 }, { "epoch": 0.28, "learning_rate": 1.6902471475504494e-06, "loss": 1.2085, "step": 3605 }, { "epoch": 0.28, "learning_rate": 1.6900652281851455e-06, "loss": 1.2909, "step": 3606 }, { "epoch": 0.28, "learning_rate": 1.6898832652106193e-06, "loss": 0.8263, "step": 3607 }, { "epoch": 0.28, "learning_rate": 1.6897012586383687e-06, "loss": 1.2285, "step": 3608 }, { "epoch": 0.28, "learning_rate": 1.689519208479897e-06, "loss": 1.2276, "step": 3609 }, { "epoch": 0.28, "learning_rate": 1.6893371147467087e-06, "loss": 0.833, "step": 3610 }, { "epoch": 0.28, "learning_rate": 1.6891549774503112e-06, "loss": 1.2015, "step": 3611 }, { "epoch": 0.28, "learning_rate": 1.6889727966022144e-06, "loss": 1.2592, "step": 3612 }, { "epoch": 0.28, "learning_rate": 1.688790572213932e-06, "loss": 1.1977, "step": 3613 }, { "epoch": 0.28, "learning_rate": 1.6886083042969798e-06, "loss": 0.8419, "step": 3614 }, { "epoch": 0.28, "learning_rate": 1.6884259928628761e-06, "loss": 1.2561, "step": 3615 }, { "epoch": 0.28, "learning_rate": 1.6882436379231425e-06, "loss": 1.259, "step": 3616 }, { "epoch": 0.28, "learning_rate": 1.6880612394893026e-06, "loss": 1.2743, "step": 3617 }, { "epoch": 0.28, "learning_rate": 1.6878787975728833e-06, "loss": 1.2905, "step": 3618 }, { "epoch": 0.28, "learning_rate": 1.6876963121854146e-06, "loss": 1.2461, "step": 3619 }, { "epoch": 0.28, "learning_rate": 1.6875137833384285e-06, "loss": 1.2387, "step": 3620 }, { "epoch": 0.28, "learning_rate": 1.6873312110434602e-06, "loss": 1.2143, "step": 3621 }, { "epoch": 0.28, "learning_rate": 1.6871485953120473e-06, "loss": 1.2605, "step": 3622 }, { "epoch": 0.28, "learning_rate": 1.6869659361557304e-06, "loss": 1.2745, "step": 3623 }, { "epoch": 0.28, "learning_rate": 1.6867832335860527e-06, "loss": 1.2282, "step": 3624 }, { "epoch": 0.28, "learning_rate": 1.6866004876145608e-06, "loss": 1.2432, "step": 3625 }, { "epoch": 0.28, "learning_rate": 1.6864176982528026e-06, "loss": 1.2197, "step": 3626 }, { "epoch": 0.28, "learning_rate": 1.6862348655123306e-06, "loss": 1.2416, "step": 3627 }, { "epoch": 0.28, "learning_rate": 1.6860519894046982e-06, "loss": 1.212, "step": 3628 }, { "epoch": 0.28, "learning_rate": 1.6858690699414629e-06, "loss": 1.2319, "step": 3629 }, { "epoch": 0.28, "learning_rate": 1.6856861071341841e-06, "loss": 1.2252, "step": 3630 }, { "epoch": 0.28, "learning_rate": 1.6855031009944245e-06, "loss": 1.2325, "step": 3631 }, { "epoch": 0.28, "learning_rate": 1.6853200515337491e-06, "loss": 1.2869, "step": 3632 }, { "epoch": 0.28, "learning_rate": 1.6851369587637266e-06, "loss": 1.2376, "step": 3633 }, { "epoch": 0.28, "learning_rate": 1.6849538226959267e-06, "loss": 1.2518, "step": 3634 }, { "epoch": 0.28, "learning_rate": 1.6847706433419233e-06, "loss": 1.2227, "step": 3635 }, { "epoch": 0.28, "learning_rate": 1.6845874207132926e-06, "loss": 1.1789, "step": 3636 }, { "epoch": 0.28, "learning_rate": 1.6844041548216134e-06, "loss": 1.2533, "step": 3637 }, { "epoch": 0.28, "learning_rate": 1.6842208456784675e-06, "loss": 1.2085, "step": 3638 }, { "epoch": 0.28, "learning_rate": 1.6840374932954388e-06, "loss": 1.2428, "step": 3639 }, { "epoch": 0.28, "learning_rate": 1.683854097684115e-06, "loss": 1.2336, "step": 3640 }, { "epoch": 0.28, "learning_rate": 1.6836706588560853e-06, "loss": 1.1731, "step": 3641 }, { "epoch": 0.28, "learning_rate": 1.6834871768229428e-06, "loss": 1.2154, "step": 3642 }, { "epoch": 0.28, "learning_rate": 1.6833036515962829e-06, "loss": 1.2383, "step": 3643 }, { "epoch": 0.28, "learning_rate": 1.683120083187703e-06, "loss": 1.2709, "step": 3644 }, { "epoch": 0.28, "learning_rate": 1.6829364716088046e-06, "loss": 1.2033, "step": 3645 }, { "epoch": 0.28, "learning_rate": 1.6827528168711905e-06, "loss": 0.8561, "step": 3646 }, { "epoch": 0.28, "learning_rate": 1.6825691189864673e-06, "loss": 1.2798, "step": 3647 }, { "epoch": 0.28, "learning_rate": 1.6823853779662435e-06, "loss": 1.227, "step": 3648 }, { "epoch": 0.28, "learning_rate": 1.6822015938221314e-06, "loss": 1.2162, "step": 3649 }, { "epoch": 0.28, "learning_rate": 1.6820177665657448e-06, "loss": 1.2847, "step": 3650 }, { "epoch": 0.28, "learning_rate": 1.6818338962087012e-06, "loss": 1.2401, "step": 3651 }, { "epoch": 0.28, "learning_rate": 1.6816499827626204e-06, "loss": 1.2393, "step": 3652 }, { "epoch": 0.28, "learning_rate": 1.6814660262391247e-06, "loss": 1.2183, "step": 3653 }, { "epoch": 0.28, "learning_rate": 1.6812820266498396e-06, "loss": 1.1952, "step": 3654 }, { "epoch": 0.28, "learning_rate": 1.681097984006393e-06, "loss": 1.2084, "step": 3655 }, { "epoch": 0.28, "learning_rate": 1.680913898320416e-06, "loss": 1.2129, "step": 3656 }, { "epoch": 0.28, "learning_rate": 1.6807297696035412e-06, "loss": 1.2682, "step": 3657 }, { "epoch": 0.28, "learning_rate": 1.6805455978674057e-06, "loss": 1.2248, "step": 3658 }, { "epoch": 0.28, "learning_rate": 1.6803613831236476e-06, "loss": 1.24, "step": 3659 }, { "epoch": 0.28, "learning_rate": 1.680177125383909e-06, "loss": 1.2415, "step": 3660 }, { "epoch": 0.28, "learning_rate": 1.679992824659834e-06, "loss": 1.2292, "step": 3661 }, { "epoch": 0.28, "learning_rate": 1.6798084809630697e-06, "loss": 1.2689, "step": 3662 }, { "epoch": 0.28, "learning_rate": 1.6796240943052657e-06, "loss": 1.1754, "step": 3663 }, { "epoch": 0.28, "learning_rate": 1.679439664698075e-06, "loss": 1.22, "step": 3664 }, { "epoch": 0.28, "learning_rate": 1.679255192153152e-06, "loss": 1.235, "step": 3665 }, { "epoch": 0.28, "learning_rate": 1.6790706766821553e-06, "loss": 1.2381, "step": 3666 }, { "epoch": 0.28, "learning_rate": 1.678886118296745e-06, "loss": 1.2061, "step": 3667 }, { "epoch": 0.28, "learning_rate": 1.6787015170085845e-06, "loss": 1.2239, "step": 3668 }, { "epoch": 0.28, "learning_rate": 1.67851687282934e-06, "loss": 1.2632, "step": 3669 }, { "epoch": 0.28, "learning_rate": 1.67833218577068e-06, "loss": 1.2394, "step": 3670 }, { "epoch": 0.28, "learning_rate": 1.6781474558442762e-06, "loss": 1.1887, "step": 3671 }, { "epoch": 0.28, "learning_rate": 1.6779626830618028e-06, "loss": 1.2537, "step": 3672 }, { "epoch": 0.29, "learning_rate": 1.6777778674349365e-06, "loss": 0.8561, "step": 3673 }, { "epoch": 0.29, "learning_rate": 1.6775930089753566e-06, "loss": 1.1858, "step": 3674 }, { "epoch": 0.29, "learning_rate": 1.6774081076947456e-06, "loss": 1.2282, "step": 3675 }, { "epoch": 0.29, "learning_rate": 1.6772231636047888e-06, "loss": 1.2717, "step": 3676 }, { "epoch": 0.29, "learning_rate": 1.6770381767171737e-06, "loss": 1.1894, "step": 3677 }, { "epoch": 0.29, "learning_rate": 1.6768531470435904e-06, "loss": 1.1903, "step": 3678 }, { "epoch": 0.29, "learning_rate": 1.6766680745957325e-06, "loss": 1.2644, "step": 3679 }, { "epoch": 0.29, "learning_rate": 1.6764829593852955e-06, "loss": 1.2024, "step": 3680 }, { "epoch": 0.29, "learning_rate": 1.6762978014239778e-06, "loss": 1.223, "step": 3681 }, { "epoch": 0.29, "learning_rate": 1.6761126007234804e-06, "loss": 0.8766, "step": 3682 }, { "epoch": 0.29, "learning_rate": 1.675927357295508e-06, "loss": 0.816, "step": 3683 }, { "epoch": 0.29, "learning_rate": 1.6757420711517666e-06, "loss": 1.2352, "step": 3684 }, { "epoch": 0.29, "learning_rate": 1.6755567423039655e-06, "loss": 1.2382, "step": 3685 }, { "epoch": 0.29, "learning_rate": 1.675371370763817e-06, "loss": 1.268, "step": 3686 }, { "epoch": 0.29, "learning_rate": 1.6751859565430357e-06, "loss": 1.1995, "step": 3687 }, { "epoch": 0.29, "learning_rate": 1.6750004996533388e-06, "loss": 1.216, "step": 3688 }, { "epoch": 0.29, "learning_rate": 1.6748150001064469e-06, "loss": 1.2315, "step": 3689 }, { "epoch": 0.29, "learning_rate": 1.6746294579140822e-06, "loss": 1.2474, "step": 3690 }, { "epoch": 0.29, "learning_rate": 1.6744438730879703e-06, "loss": 1.2337, "step": 3691 }, { "epoch": 0.29, "learning_rate": 1.67425824563984e-06, "loss": 1.2209, "step": 3692 }, { "epoch": 0.29, "learning_rate": 1.6740725755814213e-06, "loss": 1.2713, "step": 3693 }, { "epoch": 0.29, "learning_rate": 1.6738868629244483e-06, "loss": 1.199, "step": 3694 }, { "epoch": 0.29, "learning_rate": 1.6737011076806569e-06, "loss": 1.2091, "step": 3695 }, { "epoch": 0.29, "learning_rate": 1.6735153098617865e-06, "loss": 1.2832, "step": 3696 }, { "epoch": 0.29, "learning_rate": 1.6733294694795784e-06, "loss": 1.248, "step": 3697 }, { "epoch": 0.29, "learning_rate": 1.6731435865457768e-06, "loss": 0.8409, "step": 3698 }, { "epoch": 0.29, "learning_rate": 1.6729576610721293e-06, "loss": 1.2767, "step": 3699 }, { "epoch": 0.29, "learning_rate": 1.6727716930703853e-06, "loss": 1.2311, "step": 3700 }, { "epoch": 0.29, "learning_rate": 1.672585682552297e-06, "loss": 1.2529, "step": 3701 }, { "epoch": 0.29, "learning_rate": 1.6723996295296197e-06, "loss": 1.2162, "step": 3702 }, { "epoch": 0.29, "learning_rate": 1.672213534014111e-06, "loss": 1.1452, "step": 3703 }, { "epoch": 0.29, "learning_rate": 1.6720273960175318e-06, "loss": 1.2632, "step": 3704 }, { "epoch": 0.29, "learning_rate": 1.6718412155516445e-06, "loss": 1.246, "step": 3705 }, { "epoch": 0.29, "learning_rate": 1.6716549926282158e-06, "loss": 1.2531, "step": 3706 }, { "epoch": 0.29, "learning_rate": 1.6714687272590133e-06, "loss": 0.8297, "step": 3707 }, { "epoch": 0.29, "learning_rate": 1.6712824194558088e-06, "loss": 1.2564, "step": 3708 }, { "epoch": 0.29, "learning_rate": 1.671096069230376e-06, "loss": 1.2604, "step": 3709 }, { "epoch": 0.29, "learning_rate": 1.6709096765944915e-06, "loss": 1.2578, "step": 3710 }, { "epoch": 0.29, "learning_rate": 1.6707232415599344e-06, "loss": 1.1989, "step": 3711 }, { "epoch": 0.29, "learning_rate": 1.6705367641384866e-06, "loss": 1.2538, "step": 3712 }, { "epoch": 0.29, "learning_rate": 1.6703502443419329e-06, "loss": 1.2079, "step": 3713 }, { "epoch": 0.29, "learning_rate": 1.6701636821820605e-06, "loss": 1.2065, "step": 3714 }, { "epoch": 0.29, "learning_rate": 1.6699770776706592e-06, "loss": 1.2388, "step": 3715 }, { "epoch": 0.29, "learning_rate": 1.6697904308195218e-06, "loss": 1.1694, "step": 3716 }, { "epoch": 0.29, "learning_rate": 1.6696037416404435e-06, "loss": 1.3081, "step": 3717 }, { "epoch": 0.29, "learning_rate": 1.6694170101452222e-06, "loss": 1.2626, "step": 3718 }, { "epoch": 0.29, "learning_rate": 1.6692302363456587e-06, "loss": 1.2392, "step": 3719 }, { "epoch": 0.29, "learning_rate": 1.6690434202535563e-06, "loss": 1.1837, "step": 3720 }, { "epoch": 0.29, "learning_rate": 1.6688565618807208e-06, "loss": 1.2545, "step": 3721 }, { "epoch": 0.29, "learning_rate": 1.6686696612389613e-06, "loss": 0.8618, "step": 3722 }, { "epoch": 0.29, "learning_rate": 1.6684827183400888e-06, "loss": 1.2319, "step": 3723 }, { "epoch": 0.29, "learning_rate": 1.6682957331959169e-06, "loss": 1.2348, "step": 3724 }, { "epoch": 0.29, "learning_rate": 1.668108705818263e-06, "loss": 1.2248, "step": 3725 }, { "epoch": 0.29, "learning_rate": 1.667921636218946e-06, "loss": 1.2638, "step": 3726 }, { "epoch": 0.29, "learning_rate": 1.6677345244097886e-06, "loss": 1.206, "step": 3727 }, { "epoch": 0.29, "learning_rate": 1.6675473704026147e-06, "loss": 1.1499, "step": 3728 }, { "epoch": 0.29, "learning_rate": 1.6673601742092515e-06, "loss": 1.2115, "step": 3729 }, { "epoch": 0.29, "learning_rate": 1.6671729358415299e-06, "loss": 0.8979, "step": 3730 }, { "epoch": 0.29, "learning_rate": 1.666985655311282e-06, "loss": 1.2537, "step": 3731 }, { "epoch": 0.29, "learning_rate": 1.6667983326303432e-06, "loss": 1.2463, "step": 3732 }, { "epoch": 0.29, "learning_rate": 1.6666109678105516e-06, "loss": 1.2248, "step": 3733 }, { "epoch": 0.29, "learning_rate": 1.6664235608637476e-06, "loss": 1.2549, "step": 3734 }, { "epoch": 0.29, "learning_rate": 1.666236111801775e-06, "loss": 1.2515, "step": 3735 }, { "epoch": 0.29, "learning_rate": 1.6660486206364792e-06, "loss": 1.2348, "step": 3736 }, { "epoch": 0.29, "learning_rate": 1.6658610873797094e-06, "loss": 1.238, "step": 3737 }, { "epoch": 0.29, "learning_rate": 1.6656735120433169e-06, "loss": 0.8357, "step": 3738 }, { "epoch": 0.29, "learning_rate": 1.665485894639155e-06, "loss": 1.2512, "step": 3739 }, { "epoch": 0.29, "learning_rate": 1.6652982351790814e-06, "loss": 1.2089, "step": 3740 }, { "epoch": 0.29, "learning_rate": 1.6651105336749545e-06, "loss": 1.2019, "step": 3741 }, { "epoch": 0.29, "learning_rate": 1.6649227901386364e-06, "loss": 1.2208, "step": 3742 }, { "epoch": 0.29, "learning_rate": 1.6647350045819922e-06, "loss": 1.2681, "step": 3743 }, { "epoch": 0.29, "learning_rate": 1.6645471770168884e-06, "loss": 1.2272, "step": 3744 }, { "epoch": 0.29, "learning_rate": 1.6643593074551955e-06, "loss": 1.2808, "step": 3745 }, { "epoch": 0.29, "learning_rate": 1.6641713959087863e-06, "loss": 1.1879, "step": 3746 }, { "epoch": 0.29, "learning_rate": 1.6639834423895351e-06, "loss": 1.188, "step": 3747 }, { "epoch": 0.29, "learning_rate": 1.6637954469093204e-06, "loss": 1.227, "step": 3748 }, { "epoch": 0.29, "learning_rate": 1.6636074094800228e-06, "loss": 1.2584, "step": 3749 }, { "epoch": 0.29, "learning_rate": 1.6634193301135251e-06, "loss": 1.2838, "step": 3750 }, { "epoch": 0.29, "learning_rate": 1.6632312088217134e-06, "loss": 1.1633, "step": 3751 }, { "epoch": 0.29, "learning_rate": 1.6630430456164758e-06, "loss": 1.2063, "step": 3752 }, { "epoch": 0.29, "learning_rate": 1.6628548405097038e-06, "loss": 1.2405, "step": 3753 }, { "epoch": 0.29, "learning_rate": 1.6626665935132914e-06, "loss": 1.1552, "step": 3754 }, { "epoch": 0.29, "learning_rate": 1.6624783046391345e-06, "loss": 1.223, "step": 3755 }, { "epoch": 0.29, "learning_rate": 1.6622899738991324e-06, "loss": 1.2069, "step": 3756 }, { "epoch": 0.29, "learning_rate": 1.6621016013051865e-06, "loss": 1.1998, "step": 3757 }, { "epoch": 0.29, "learning_rate": 1.6619131868692016e-06, "loss": 1.2281, "step": 3758 }, { "epoch": 0.29, "learning_rate": 1.6617247306030845e-06, "loss": 0.8236, "step": 3759 }, { "epoch": 0.29, "learning_rate": 1.6615362325187452e-06, "loss": 1.2564, "step": 3760 }, { "epoch": 0.29, "learning_rate": 1.6613476926280953e-06, "loss": 1.2348, "step": 3761 }, { "epoch": 0.29, "learning_rate": 1.6611591109430503e-06, "loss": 1.2472, "step": 3762 }, { "epoch": 0.29, "learning_rate": 1.6609704874755275e-06, "loss": 1.267, "step": 3763 }, { "epoch": 0.29, "learning_rate": 1.660781822237447e-06, "loss": 1.2496, "step": 3764 }, { "epoch": 0.29, "learning_rate": 1.6605931152407322e-06, "loss": 1.2255, "step": 3765 }, { "epoch": 0.29, "learning_rate": 1.660404366497308e-06, "loss": 1.2097, "step": 3766 }, { "epoch": 0.29, "learning_rate": 1.6602155760191025e-06, "loss": 1.2291, "step": 3767 }, { "epoch": 0.29, "learning_rate": 1.6600267438180472e-06, "loss": 1.2697, "step": 3768 }, { "epoch": 0.29, "learning_rate": 1.6598378699060748e-06, "loss": 1.2415, "step": 3769 }, { "epoch": 0.29, "learning_rate": 1.6596489542951214e-06, "loss": 1.2211, "step": 3770 }, { "epoch": 0.29, "learning_rate": 1.6594599969971256e-06, "loss": 1.1778, "step": 3771 }, { "epoch": 0.29, "learning_rate": 1.6592709980240295e-06, "loss": 1.2252, "step": 3772 }, { "epoch": 0.29, "learning_rate": 1.6590819573877761e-06, "loss": 1.2139, "step": 3773 }, { "epoch": 0.29, "learning_rate": 1.6588928751003124e-06, "loss": 1.2705, "step": 3774 }, { "epoch": 0.29, "learning_rate": 1.6587037511735876e-06, "loss": 1.1443, "step": 3775 }, { "epoch": 0.29, "learning_rate": 1.6585145856195535e-06, "loss": 1.2768, "step": 3776 }, { "epoch": 0.29, "learning_rate": 1.6583253784501646e-06, "loss": 1.1718, "step": 3777 }, { "epoch": 0.29, "learning_rate": 1.6581361296773777e-06, "loss": 1.2707, "step": 3778 }, { "epoch": 0.29, "learning_rate": 1.657946839313153e-06, "loss": 1.2225, "step": 3779 }, { "epoch": 0.29, "learning_rate": 1.6577575073694522e-06, "loss": 1.189, "step": 3780 }, { "epoch": 0.29, "learning_rate": 1.6575681338582412e-06, "loss": 1.2027, "step": 3781 }, { "epoch": 0.29, "learning_rate": 1.6573787187914872e-06, "loss": 0.8705, "step": 3782 }, { "epoch": 0.29, "learning_rate": 1.65718926218116e-06, "loss": 0.8645, "step": 3783 }, { "epoch": 0.29, "learning_rate": 1.656999764039233e-06, "loss": 1.2362, "step": 3784 }, { "epoch": 0.29, "learning_rate": 1.6568102243776816e-06, "loss": 1.238, "step": 3785 }, { "epoch": 0.29, "learning_rate": 1.6566206432084838e-06, "loss": 1.2287, "step": 3786 }, { "epoch": 0.29, "learning_rate": 1.6564310205436203e-06, "loss": 1.2893, "step": 3787 }, { "epoch": 0.29, "learning_rate": 1.6562413563950744e-06, "loss": 1.2389, "step": 3788 }, { "epoch": 0.29, "learning_rate": 1.6560516507748323e-06, "loss": 1.206, "step": 3789 }, { "epoch": 0.29, "learning_rate": 1.6558619036948822e-06, "loss": 1.2108, "step": 3790 }, { "epoch": 0.29, "learning_rate": 1.655672115167216e-06, "loss": 1.2397, "step": 3791 }, { "epoch": 0.29, "learning_rate": 1.655482285203827e-06, "loss": 1.2709, "step": 3792 }, { "epoch": 0.29, "learning_rate": 1.6552924138167117e-06, "loss": 1.2624, "step": 3793 }, { "epoch": 0.29, "learning_rate": 1.6551025010178693e-06, "loss": 1.2748, "step": 3794 }, { "epoch": 0.29, "learning_rate": 1.6549125468193017e-06, "loss": 1.2596, "step": 3795 }, { "epoch": 0.29, "learning_rate": 1.6547225512330126e-06, "loss": 1.2377, "step": 3796 }, { "epoch": 0.29, "learning_rate": 1.6545325142710095e-06, "loss": 0.8401, "step": 3797 }, { "epoch": 0.29, "learning_rate": 1.6543424359453018e-06, "loss": 1.2142, "step": 3798 }, { "epoch": 0.29, "learning_rate": 1.6541523162679015e-06, "loss": 1.2243, "step": 3799 }, { "epoch": 0.29, "learning_rate": 1.6539621552508232e-06, "loss": 0.8779, "step": 3800 }, { "epoch": 0.3, "learning_rate": 1.6537719529060848e-06, "loss": 1.2503, "step": 3801 }, { "epoch": 0.3, "learning_rate": 1.653581709245706e-06, "loss": 1.2311, "step": 3802 }, { "epoch": 0.3, "learning_rate": 1.6533914242817093e-06, "loss": 1.1739, "step": 3803 }, { "epoch": 0.3, "learning_rate": 1.6532010980261202e-06, "loss": 1.2365, "step": 3804 }, { "epoch": 0.3, "learning_rate": 1.653010730490966e-06, "loss": 1.2371, "step": 3805 }, { "epoch": 0.3, "learning_rate": 1.6528203216882778e-06, "loss": 1.2383, "step": 3806 }, { "epoch": 0.3, "learning_rate": 1.6526298716300882e-06, "loss": 1.2124, "step": 3807 }, { "epoch": 0.3, "learning_rate": 1.652439380328433e-06, "loss": 1.2523, "step": 3808 }, { "epoch": 0.3, "learning_rate": 1.6522488477953503e-06, "loss": 1.2098, "step": 3809 }, { "epoch": 0.3, "learning_rate": 1.652058274042881e-06, "loss": 1.2812, "step": 3810 }, { "epoch": 0.3, "learning_rate": 1.6518676590830686e-06, "loss": 1.2064, "step": 3811 }, { "epoch": 0.3, "learning_rate": 1.6516770029279593e-06, "loss": 1.2008, "step": 3812 }, { "epoch": 0.3, "learning_rate": 1.6514863055896018e-06, "loss": 1.2865, "step": 3813 }, { "epoch": 0.3, "learning_rate": 1.651295567080047e-06, "loss": 1.1969, "step": 3814 }, { "epoch": 0.3, "learning_rate": 1.6511047874113491e-06, "loss": 1.2518, "step": 3815 }, { "epoch": 0.3, "learning_rate": 1.6509139665955643e-06, "loss": 0.878, "step": 3816 }, { "epoch": 0.3, "learning_rate": 1.650723104644752e-06, "loss": 1.1916, "step": 3817 }, { "epoch": 0.3, "learning_rate": 1.6505322015709734e-06, "loss": 1.2037, "step": 3818 }, { "epoch": 0.3, "learning_rate": 1.6503412573862933e-06, "loss": 1.1654, "step": 3819 }, { "epoch": 0.3, "learning_rate": 1.6501502721027785e-06, "loss": 1.2721, "step": 3820 }, { "epoch": 0.3, "learning_rate": 1.6499592457324982e-06, "loss": 1.3317, "step": 3821 }, { "epoch": 0.3, "learning_rate": 1.6497681782875245e-06, "loss": 1.237, "step": 3822 }, { "epoch": 0.3, "learning_rate": 1.6495770697799323e-06, "loss": 1.2578, "step": 3823 }, { "epoch": 0.3, "learning_rate": 1.6493859202217988e-06, "loss": 1.1798, "step": 3824 }, { "epoch": 0.3, "learning_rate": 1.6491947296252033e-06, "loss": 0.853, "step": 3825 }, { "epoch": 0.3, "learning_rate": 1.6490034980022292e-06, "loss": 1.2845, "step": 3826 }, { "epoch": 0.3, "learning_rate": 1.6488122253649609e-06, "loss": 1.2426, "step": 3827 }, { "epoch": 0.3, "learning_rate": 1.648620911725486e-06, "loss": 1.2512, "step": 3828 }, { "epoch": 0.3, "learning_rate": 1.6484295570958946e-06, "loss": 1.2019, "step": 3829 }, { "epoch": 0.3, "learning_rate": 1.6482381614882802e-06, "loss": 1.2181, "step": 3830 }, { "epoch": 0.3, "learning_rate": 1.6480467249147376e-06, "loss": 1.1668, "step": 3831 }, { "epoch": 0.3, "learning_rate": 1.647855247387365e-06, "loss": 1.2521, "step": 3832 }, { "epoch": 0.3, "learning_rate": 1.647663728918263e-06, "loss": 0.8559, "step": 3833 }, { "epoch": 0.3, "learning_rate": 1.6474721695195347e-06, "loss": 1.2264, "step": 3834 }, { "epoch": 0.3, "learning_rate": 1.6472805692032856e-06, "loss": 1.2306, "step": 3835 }, { "epoch": 0.3, "learning_rate": 1.6470889279816244e-06, "loss": 1.2678, "step": 3836 }, { "epoch": 0.3, "learning_rate": 1.6468972458666622e-06, "loss": 1.2235, "step": 3837 }, { "epoch": 0.3, "learning_rate": 1.646705522870512e-06, "loss": 1.2672, "step": 3838 }, { "epoch": 0.3, "learning_rate": 1.64651375900529e-06, "loss": 1.2008, "step": 3839 }, { "epoch": 0.3, "learning_rate": 1.646321954283115e-06, "loss": 1.197, "step": 3840 }, { "epoch": 0.3, "learning_rate": 1.6461301087161083e-06, "loss": 1.217, "step": 3841 }, { "epoch": 0.3, "learning_rate": 1.6459382223163938e-06, "loss": 1.233, "step": 3842 }, { "epoch": 0.3, "learning_rate": 1.6457462950960977e-06, "loss": 1.1674, "step": 3843 }, { "epoch": 0.3, "learning_rate": 1.6455543270673488e-06, "loss": 0.8227, "step": 3844 }, { "epoch": 0.3, "learning_rate": 1.6453623182422792e-06, "loss": 1.2471, "step": 3845 }, { "epoch": 0.3, "learning_rate": 1.6451702686330225e-06, "loss": 1.2162, "step": 3846 }, { "epoch": 0.3, "learning_rate": 1.6449781782517161e-06, "loss": 1.2424, "step": 3847 }, { "epoch": 0.3, "learning_rate": 1.6447860471104986e-06, "loss": 1.2511, "step": 3848 }, { "epoch": 0.3, "learning_rate": 1.6445938752215123e-06, "loss": 1.214, "step": 3849 }, { "epoch": 0.3, "learning_rate": 1.6444016625969016e-06, "loss": 1.2198, "step": 3850 }, { "epoch": 0.3, "learning_rate": 1.6442094092488134e-06, "loss": 1.2461, "step": 3851 }, { "epoch": 0.3, "learning_rate": 1.6440171151893974e-06, "loss": 0.8221, "step": 3852 }, { "epoch": 0.3, "learning_rate": 1.6438247804308058e-06, "loss": 1.1922, "step": 3853 }, { "epoch": 0.3, "learning_rate": 1.6436324049851936e-06, "loss": 1.2474, "step": 3854 }, { "epoch": 0.3, "learning_rate": 1.6434399888647175e-06, "loss": 1.2246, "step": 3855 }, { "epoch": 0.3, "learning_rate": 1.6432475320815376e-06, "loss": 1.2231, "step": 3856 }, { "epoch": 0.3, "learning_rate": 1.6430550346478164e-06, "loss": 1.2749, "step": 3857 }, { "epoch": 0.3, "learning_rate": 1.6428624965757197e-06, "loss": 1.2782, "step": 3858 }, { "epoch": 0.3, "learning_rate": 1.6426699178774137e-06, "loss": 1.2025, "step": 3859 }, { "epoch": 0.3, "learning_rate": 1.6424772985650696e-06, "loss": 1.2541, "step": 3860 }, { "epoch": 0.3, "learning_rate": 1.6422846386508598e-06, "loss": 1.2437, "step": 3861 }, { "epoch": 0.3, "learning_rate": 1.6420919381469595e-06, "loss": 1.204, "step": 3862 }, { "epoch": 0.3, "learning_rate": 1.6418991970655469e-06, "loss": 1.228, "step": 3863 }, { "epoch": 0.3, "learning_rate": 1.641706415418802e-06, "loss": 1.2763, "step": 3864 }, { "epoch": 0.3, "learning_rate": 1.641513593218908e-06, "loss": 1.2212, "step": 3865 }, { "epoch": 0.3, "learning_rate": 1.6413207304780502e-06, "loss": 1.2455, "step": 3866 }, { "epoch": 0.3, "learning_rate": 1.6411278272084172e-06, "loss": 1.2754, "step": 3867 }, { "epoch": 0.3, "learning_rate": 1.6409348834221996e-06, "loss": 1.2723, "step": 3868 }, { "epoch": 0.3, "learning_rate": 1.6407418991315902e-06, "loss": 1.2418, "step": 3869 }, { "epoch": 0.3, "learning_rate": 1.6405488743487851e-06, "loss": 1.215, "step": 3870 }, { "epoch": 0.3, "learning_rate": 1.6403558090859828e-06, "loss": 1.1657, "step": 3871 }, { "epoch": 0.3, "learning_rate": 1.6401627033553838e-06, "loss": 1.2066, "step": 3872 }, { "epoch": 0.3, "learning_rate": 1.6399695571691918e-06, "loss": 1.1429, "step": 3873 }, { "epoch": 0.3, "learning_rate": 1.6397763705396131e-06, "loss": 1.2256, "step": 3874 }, { "epoch": 0.3, "learning_rate": 1.6395831434788558e-06, "loss": 1.2143, "step": 3875 }, { "epoch": 0.3, "learning_rate": 1.6393898759991314e-06, "loss": 1.1576, "step": 3876 }, { "epoch": 0.3, "learning_rate": 1.6391965681126532e-06, "loss": 0.8606, "step": 3877 }, { "epoch": 0.3, "learning_rate": 1.639003219831638e-06, "loss": 1.2209, "step": 3878 }, { "epoch": 0.3, "learning_rate": 1.6388098311683038e-06, "loss": 1.2637, "step": 3879 }, { "epoch": 0.3, "learning_rate": 1.6386164021348727e-06, "loss": 1.2302, "step": 3880 }, { "epoch": 0.3, "learning_rate": 1.6384229327435683e-06, "loss": 1.2421, "step": 3881 }, { "epoch": 0.3, "learning_rate": 1.6382294230066172e-06, "loss": 1.2387, "step": 3882 }, { "epoch": 0.3, "learning_rate": 1.6380358729362485e-06, "loss": 1.2248, "step": 3883 }, { "epoch": 0.3, "learning_rate": 1.6378422825446929e-06, "loss": 1.2093, "step": 3884 }, { "epoch": 0.3, "learning_rate": 1.6376486518441856e-06, "loss": 1.2042, "step": 3885 }, { "epoch": 0.3, "learning_rate": 1.6374549808469628e-06, "loss": 1.2206, "step": 3886 }, { "epoch": 0.3, "learning_rate": 1.6372612695652636e-06, "loss": 1.2244, "step": 3887 }, { "epoch": 0.3, "learning_rate": 1.6370675180113298e-06, "loss": 1.2096, "step": 3888 }, { "epoch": 0.3, "learning_rate": 1.6368737261974056e-06, "loss": 0.8221, "step": 3889 }, { "epoch": 0.3, "learning_rate": 1.6366798941357384e-06, "loss": 0.8115, "step": 3890 }, { "epoch": 0.3, "learning_rate": 1.6364860218385763e-06, "loss": 1.2354, "step": 3891 }, { "epoch": 0.3, "learning_rate": 1.6362921093181727e-06, "loss": 1.1572, "step": 3892 }, { "epoch": 0.3, "learning_rate": 1.6360981565867812e-06, "loss": 1.2395, "step": 3893 }, { "epoch": 0.3, "learning_rate": 1.6359041636566587e-06, "loss": 1.2258, "step": 3894 }, { "epoch": 0.3, "learning_rate": 1.6357101305400654e-06, "loss": 1.2568, "step": 3895 }, { "epoch": 0.3, "learning_rate": 1.635516057249263e-06, "loss": 1.2126, "step": 3896 }, { "epoch": 0.3, "learning_rate": 1.6353219437965158e-06, "loss": 1.2715, "step": 3897 }, { "epoch": 0.3, "learning_rate": 1.6351277901940915e-06, "loss": 1.2731, "step": 3898 }, { "epoch": 0.3, "learning_rate": 1.6349335964542594e-06, "loss": 1.2114, "step": 3899 }, { "epoch": 0.3, "learning_rate": 1.634739362589292e-06, "loss": 0.8534, "step": 3900 }, { "epoch": 0.3, "learning_rate": 1.6345450886114638e-06, "loss": 1.2506, "step": 3901 }, { "epoch": 0.3, "learning_rate": 1.6343507745330524e-06, "loss": 1.2291, "step": 3902 }, { "epoch": 0.3, "learning_rate": 1.6341564203663376e-06, "loss": 1.2514, "step": 3903 }, { "epoch": 0.3, "learning_rate": 1.6339620261236014e-06, "loss": 1.2432, "step": 3904 }, { "epoch": 0.3, "learning_rate": 1.6337675918171287e-06, "loss": 1.2175, "step": 3905 }, { "epoch": 0.3, "learning_rate": 1.6335731174592078e-06, "loss": 1.2248, "step": 3906 }, { "epoch": 0.3, "learning_rate": 1.6333786030621276e-06, "loss": 1.2191, "step": 3907 }, { "epoch": 0.3, "learning_rate": 1.6331840486381813e-06, "loss": 1.253, "step": 3908 }, { "epoch": 0.3, "learning_rate": 1.6329894541996634e-06, "loss": 1.2071, "step": 3909 }, { "epoch": 0.3, "learning_rate": 1.6327948197588719e-06, "loss": 1.2117, "step": 3910 }, { "epoch": 0.3, "learning_rate": 1.6326001453281065e-06, "loss": 1.2404, "step": 3911 }, { "epoch": 0.3, "learning_rate": 1.63240543091967e-06, "loss": 1.2246, "step": 3912 }, { "epoch": 0.3, "learning_rate": 1.6322106765458676e-06, "loss": 1.1938, "step": 3913 }, { "epoch": 0.3, "learning_rate": 1.6320158822190072e-06, "loss": 1.269, "step": 3914 }, { "epoch": 0.3, "learning_rate": 1.631821047951398e-06, "loss": 1.2423, "step": 3915 }, { "epoch": 0.3, "learning_rate": 1.6316261737553539e-06, "loss": 1.2315, "step": 3916 }, { "epoch": 0.3, "learning_rate": 1.6314312596431891e-06, "loss": 1.2607, "step": 3917 }, { "epoch": 0.3, "learning_rate": 1.6312363056272224e-06, "loss": 0.8096, "step": 3918 }, { "epoch": 0.3, "learning_rate": 1.6310413117197732e-06, "loss": 1.2389, "step": 3919 }, { "epoch": 0.3, "learning_rate": 1.630846277933165e-06, "loss": 1.2394, "step": 3920 }, { "epoch": 0.3, "learning_rate": 1.6306512042797222e-06, "loss": 1.1978, "step": 3921 }, { "epoch": 0.3, "learning_rate": 1.6304560907717735e-06, "loss": 1.1973, "step": 3922 }, { "epoch": 0.3, "learning_rate": 1.630260937421649e-06, "loss": 1.2267, "step": 3923 }, { "epoch": 0.3, "learning_rate": 1.6300657442416813e-06, "loss": 1.1917, "step": 3924 }, { "epoch": 0.3, "learning_rate": 1.6298705112442059e-06, "loss": 1.2231, "step": 3925 }, { "epoch": 0.3, "learning_rate": 1.6296752384415608e-06, "loss": 1.2212, "step": 3926 }, { "epoch": 0.3, "learning_rate": 1.6294799258460868e-06, "loss": 1.2486, "step": 3927 }, { "epoch": 0.3, "learning_rate": 1.6292845734701262e-06, "loss": 1.2563, "step": 3928 }, { "epoch": 0.3, "learning_rate": 1.6290891813260245e-06, "loss": 1.2386, "step": 3929 }, { "epoch": 0.31, "learning_rate": 1.62889374942613e-06, "loss": 1.2586, "step": 3930 }, { "epoch": 0.31, "learning_rate": 1.6286982777827933e-06, "loss": 1.2577, "step": 3931 }, { "epoch": 0.31, "learning_rate": 1.6285027664083671e-06, "loss": 1.2408, "step": 3932 }, { "epoch": 0.31, "learning_rate": 1.6283072153152063e-06, "loss": 1.2182, "step": 3933 }, { "epoch": 0.31, "learning_rate": 1.6281116245156702e-06, "loss": 1.2534, "step": 3934 }, { "epoch": 0.31, "learning_rate": 1.6279159940221185e-06, "loss": 1.1747, "step": 3935 }, { "epoch": 0.31, "learning_rate": 1.6277203238469142e-06, "loss": 1.2015, "step": 3936 }, { "epoch": 0.31, "learning_rate": 1.6275246140024232e-06, "loss": 1.1789, "step": 3937 }, { "epoch": 0.31, "learning_rate": 1.6273288645010132e-06, "loss": 1.2242, "step": 3938 }, { "epoch": 0.31, "learning_rate": 1.6271330753550547e-06, "loss": 1.2336, "step": 3939 }, { "epoch": 0.31, "learning_rate": 1.6269372465769214e-06, "loss": 1.2839, "step": 3940 }, { "epoch": 0.31, "learning_rate": 1.626741378178988e-06, "loss": 1.2097, "step": 3941 }, { "epoch": 0.31, "learning_rate": 1.6265454701736335e-06, "loss": 1.2095, "step": 3942 }, { "epoch": 0.31, "learning_rate": 1.6263495225732372e-06, "loss": 0.868, "step": 3943 }, { "epoch": 0.31, "learning_rate": 1.6261535353901833e-06, "loss": 0.8521, "step": 3944 }, { "epoch": 0.31, "learning_rate": 1.625957508636857e-06, "loss": 1.2007, "step": 3945 }, { "epoch": 0.31, "learning_rate": 1.6257614423256461e-06, "loss": 1.2263, "step": 3946 }, { "epoch": 0.31, "learning_rate": 1.6255653364689414e-06, "loss": 1.2391, "step": 3947 }, { "epoch": 0.31, "learning_rate": 1.625369191079136e-06, "loss": 1.2115, "step": 3948 }, { "epoch": 0.31, "learning_rate": 1.6251730061686252e-06, "loss": 1.1503, "step": 3949 }, { "epoch": 0.31, "learning_rate": 1.6249767817498076e-06, "loss": 1.2304, "step": 3950 }, { "epoch": 0.31, "learning_rate": 1.6247805178350835e-06, "loss": 1.226, "step": 3951 }, { "epoch": 0.31, "learning_rate": 1.6245842144368558e-06, "loss": 1.2606, "step": 3952 }, { "epoch": 0.31, "learning_rate": 1.6243878715675299e-06, "loss": 1.185, "step": 3953 }, { "epoch": 0.31, "learning_rate": 1.6241914892395142e-06, "loss": 1.2397, "step": 3954 }, { "epoch": 0.31, "learning_rate": 1.6239950674652193e-06, "loss": 1.2222, "step": 3955 }, { "epoch": 0.31, "learning_rate": 1.6237986062570576e-06, "loss": 1.2331, "step": 3956 }, { "epoch": 0.31, "learning_rate": 1.6236021056274455e-06, "loss": 1.2248, "step": 3957 }, { "epoch": 0.31, "learning_rate": 1.6234055655888005e-06, "loss": 1.2451, "step": 3958 }, { "epoch": 0.31, "learning_rate": 1.6232089861535433e-06, "loss": 1.2054, "step": 3959 }, { "epoch": 0.31, "learning_rate": 1.6230123673340965e-06, "loss": 1.236, "step": 3960 }, { "epoch": 0.31, "learning_rate": 1.6228157091428863e-06, "loss": 1.2906, "step": 3961 }, { "epoch": 0.31, "learning_rate": 1.6226190115923398e-06, "loss": 1.2418, "step": 3962 }, { "epoch": 0.31, "learning_rate": 1.6224222746948883e-06, "loss": 1.1856, "step": 3963 }, { "epoch": 0.31, "learning_rate": 1.622225498462964e-06, "loss": 1.2029, "step": 3964 }, { "epoch": 0.31, "learning_rate": 1.6220286829090029e-06, "loss": 1.2659, "step": 3965 }, { "epoch": 0.31, "learning_rate": 1.6218318280454426e-06, "loss": 1.1856, "step": 3966 }, { "epoch": 0.31, "learning_rate": 1.6216349338847234e-06, "loss": 1.241, "step": 3967 }, { "epoch": 0.31, "learning_rate": 1.6214380004392887e-06, "loss": 1.2463, "step": 3968 }, { "epoch": 0.31, "learning_rate": 1.6212410277215837e-06, "loss": 1.2233, "step": 3969 }, { "epoch": 0.31, "learning_rate": 1.6210440157440557e-06, "loss": 1.2831, "step": 3970 }, { "epoch": 0.31, "learning_rate": 1.6208469645191553e-06, "loss": 1.2299, "step": 3971 }, { "epoch": 0.31, "learning_rate": 1.6206498740593358e-06, "loss": 1.2437, "step": 3972 }, { "epoch": 0.31, "learning_rate": 1.620452744377052e-06, "loss": 1.1935, "step": 3973 }, { "epoch": 0.31, "learning_rate": 1.6202555754847617e-06, "loss": 1.2465, "step": 3974 }, { "epoch": 0.31, "learning_rate": 1.6200583673949253e-06, "loss": 1.2113, "step": 3975 }, { "epoch": 0.31, "learning_rate": 1.6198611201200057e-06, "loss": 1.2849, "step": 3976 }, { "epoch": 0.31, "learning_rate": 1.6196638336724674e-06, "loss": 1.2471, "step": 3977 }, { "epoch": 0.31, "learning_rate": 1.6194665080647787e-06, "loss": 1.1904, "step": 3978 }, { "epoch": 0.31, "learning_rate": 1.6192691433094096e-06, "loss": 0.8625, "step": 3979 }, { "epoch": 0.31, "learning_rate": 1.6190717394188332e-06, "loss": 1.2724, "step": 3980 }, { "epoch": 0.31, "learning_rate": 1.6188742964055234e-06, "loss": 1.2465, "step": 3981 }, { "epoch": 0.31, "learning_rate": 1.6186768142819587e-06, "loss": 1.2396, "step": 3982 }, { "epoch": 0.31, "learning_rate": 1.6184792930606188e-06, "loss": 1.2505, "step": 3983 }, { "epoch": 0.31, "learning_rate": 1.6182817327539864e-06, "loss": 1.2826, "step": 3984 }, { "epoch": 0.31, "learning_rate": 1.6180841333745465e-06, "loss": 1.2515, "step": 3985 }, { "epoch": 0.31, "learning_rate": 1.6178864949347863e-06, "loss": 0.8661, "step": 3986 }, { "epoch": 0.31, "learning_rate": 1.617688817447196e-06, "loss": 1.2302, "step": 3987 }, { "epoch": 0.31, "learning_rate": 1.6174911009242678e-06, "loss": 1.2745, "step": 3988 }, { "epoch": 0.31, "learning_rate": 1.6172933453784965e-06, "loss": 1.2348, "step": 3989 }, { "epoch": 0.31, "learning_rate": 1.6170955508223795e-06, "loss": 1.1725, "step": 3990 }, { "epoch": 0.31, "learning_rate": 1.6168977172684171e-06, "loss": 1.2558, "step": 3991 }, { "epoch": 0.31, "learning_rate": 1.6166998447291107e-06, "loss": 1.2424, "step": 3992 }, { "epoch": 0.31, "learning_rate": 1.6165019332169654e-06, "loss": 1.233, "step": 3993 }, { "epoch": 0.31, "learning_rate": 1.616303982744488e-06, "loss": 0.8624, "step": 3994 }, { "epoch": 0.31, "learning_rate": 1.6161059933241893e-06, "loss": 1.231, "step": 3995 }, { "epoch": 0.31, "learning_rate": 1.61590796496858e-06, "loss": 1.2007, "step": 3996 }, { "epoch": 0.31, "learning_rate": 1.6157098976901752e-06, "loss": 1.3182, "step": 3997 }, { "epoch": 0.31, "learning_rate": 1.6155117915014923e-06, "loss": 1.2541, "step": 3998 }, { "epoch": 0.31, "learning_rate": 1.6153136464150503e-06, "loss": 1.2453, "step": 3999 }, { "epoch": 0.31, "learning_rate": 1.6151154624433713e-06, "loss": 1.2767, "step": 4000 }, { "epoch": 0.31, "learning_rate": 1.6149172395989795e-06, "loss": 1.2276, "step": 4001 }, { "epoch": 0.31, "learning_rate": 1.614718977894402e-06, "loss": 1.2227, "step": 4002 }, { "epoch": 0.31, "learning_rate": 1.6145206773421682e-06, "loss": 1.2655, "step": 4003 }, { "epoch": 0.31, "learning_rate": 1.6143223379548098e-06, "loss": 1.266, "step": 4004 }, { "epoch": 0.31, "learning_rate": 1.6141239597448606e-06, "loss": 1.2339, "step": 4005 }, { "epoch": 0.31, "learning_rate": 1.613925542724858e-06, "loss": 1.216, "step": 4006 }, { "epoch": 0.31, "learning_rate": 1.6137270869073403e-06, "loss": 0.8447, "step": 4007 }, { "epoch": 0.31, "learning_rate": 1.6135285923048495e-06, "loss": 1.1991, "step": 4008 }, { "epoch": 0.31, "learning_rate": 1.6133300589299296e-06, "loss": 1.2599, "step": 4009 }, { "epoch": 0.31, "learning_rate": 1.6131314867951273e-06, "loss": 1.2004, "step": 4010 }, { "epoch": 0.31, "learning_rate": 1.6129328759129912e-06, "loss": 0.8094, "step": 4011 }, { "epoch": 0.31, "learning_rate": 1.6127342262960728e-06, "loss": 1.2235, "step": 4012 }, { "epoch": 0.31, "learning_rate": 1.6125355379569258e-06, "loss": 1.2298, "step": 4013 }, { "epoch": 0.31, "learning_rate": 1.612336810908107e-06, "loss": 1.2415, "step": 4014 }, { "epoch": 0.31, "learning_rate": 1.6121380451621742e-06, "loss": 1.1946, "step": 4015 }, { "epoch": 0.31, "learning_rate": 1.611939240731689e-06, "loss": 1.2412, "step": 4016 }, { "epoch": 0.31, "learning_rate": 1.6117403976292155e-06, "loss": 1.2127, "step": 4017 }, { "epoch": 0.31, "learning_rate": 1.6115415158673188e-06, "loss": 1.2995, "step": 4018 }, { "epoch": 0.31, "learning_rate": 1.6113425954585681e-06, "loss": 1.2588, "step": 4019 }, { "epoch": 0.31, "learning_rate": 1.6111436364155343e-06, "loss": 1.2191, "step": 4020 }, { "epoch": 0.31, "learning_rate": 1.6109446387507904e-06, "loss": 1.2465, "step": 4021 }, { "epoch": 0.31, "learning_rate": 1.610745602476912e-06, "loss": 1.2345, "step": 4022 }, { "epoch": 0.31, "learning_rate": 1.6105465276064783e-06, "loss": 1.1859, "step": 4023 }, { "epoch": 0.31, "learning_rate": 1.6103474141520694e-06, "loss": 1.2821, "step": 4024 }, { "epoch": 0.31, "learning_rate": 1.6101482621262685e-06, "loss": 1.2333, "step": 4025 }, { "epoch": 0.31, "learning_rate": 1.6099490715416613e-06, "loss": 1.2562, "step": 4026 }, { "epoch": 0.31, "learning_rate": 1.6097498424108353e-06, "loss": 1.2333, "step": 4027 }, { "epoch": 0.31, "learning_rate": 1.6095505747463816e-06, "loss": 1.2032, "step": 4028 }, { "epoch": 0.31, "learning_rate": 1.6093512685608928e-06, "loss": 1.2482, "step": 4029 }, { "epoch": 0.31, "learning_rate": 1.609151923866964e-06, "loss": 1.2471, "step": 4030 }, { "epoch": 0.31, "learning_rate": 1.6089525406771933e-06, "loss": 1.2555, "step": 4031 }, { "epoch": 0.31, "learning_rate": 1.6087531190041808e-06, "loss": 1.2056, "step": 4032 }, { "epoch": 0.31, "learning_rate": 1.608553658860529e-06, "loss": 1.2523, "step": 4033 }, { "epoch": 0.31, "learning_rate": 1.6083541602588428e-06, "loss": 1.2406, "step": 4034 }, { "epoch": 0.31, "learning_rate": 1.60815462321173e-06, "loss": 0.8411, "step": 4035 }, { "epoch": 0.31, "learning_rate": 1.6079550477318005e-06, "loss": 1.1844, "step": 4036 }, { "epoch": 0.31, "learning_rate": 1.6077554338316666e-06, "loss": 1.1864, "step": 4037 }, { "epoch": 0.31, "learning_rate": 1.6075557815239425e-06, "loss": 1.2869, "step": 4038 }, { "epoch": 0.31, "learning_rate": 1.6073560908212463e-06, "loss": 1.2801, "step": 4039 }, { "epoch": 0.31, "learning_rate": 1.607156361736197e-06, "loss": 1.2115, "step": 4040 }, { "epoch": 0.31, "learning_rate": 1.6069565942814167e-06, "loss": 1.2294, "step": 4041 }, { "epoch": 0.31, "learning_rate": 1.60675678846953e-06, "loss": 1.1314, "step": 4042 }, { "epoch": 0.31, "learning_rate": 1.6065569443131638e-06, "loss": 1.2227, "step": 4043 }, { "epoch": 0.31, "learning_rate": 1.6063570618249474e-06, "loss": 1.2494, "step": 4044 }, { "epoch": 0.31, "learning_rate": 1.6061571410175124e-06, "loss": 1.2129, "step": 4045 }, { "epoch": 0.31, "learning_rate": 1.6059571819034932e-06, "loss": 1.2274, "step": 4046 }, { "epoch": 0.31, "learning_rate": 1.6057571844955262e-06, "loss": 0.8336, "step": 4047 }, { "epoch": 0.31, "learning_rate": 1.6055571488062506e-06, "loss": 1.2299, "step": 4048 }, { "epoch": 0.31, "learning_rate": 1.6053570748483071e-06, "loss": 1.2441, "step": 4049 }, { "epoch": 0.31, "learning_rate": 1.6051569626343405e-06, "loss": 1.2263, "step": 4050 }, { "epoch": 0.31, "learning_rate": 1.6049568121769966e-06, "loss": 1.1578, "step": 4051 }, { "epoch": 0.31, "learning_rate": 1.604756623488924e-06, "loss": 1.2044, "step": 4052 }, { "epoch": 0.31, "learning_rate": 1.6045563965827738e-06, "loss": 1.2126, "step": 4053 }, { "epoch": 0.31, "learning_rate": 1.6043561314711998e-06, "loss": 0.8552, "step": 4054 }, { "epoch": 0.31, "learning_rate": 1.6041558281668573e-06, "loss": 1.2569, "step": 4055 }, { "epoch": 0.31, "learning_rate": 1.6039554866824052e-06, "loss": 1.1759, "step": 4056 }, { "epoch": 0.31, "learning_rate": 1.6037551070305043e-06, "loss": 1.2186, "step": 4057 }, { "epoch": 0.31, "learning_rate": 1.6035546892238173e-06, "loss": 1.2849, "step": 4058 }, { "epoch": 0.32, "learning_rate": 1.6033542332750099e-06, "loss": 1.2181, "step": 4059 }, { "epoch": 0.32, "learning_rate": 1.6031537391967503e-06, "loss": 1.1915, "step": 4060 }, { "epoch": 0.32, "learning_rate": 1.6029532070017084e-06, "loss": 1.2544, "step": 4061 }, { "epoch": 0.32, "learning_rate": 1.602752636702558e-06, "loss": 1.253, "step": 4062 }, { "epoch": 0.32, "learning_rate": 1.6025520283119732e-06, "loss": 1.1866, "step": 4063 }, { "epoch": 0.32, "learning_rate": 1.6023513818426318e-06, "loss": 1.1944, "step": 4064 }, { "epoch": 0.32, "learning_rate": 1.6021506973072143e-06, "loss": 1.2541, "step": 4065 }, { "epoch": 0.32, "learning_rate": 1.6019499747184026e-06, "loss": 1.217, "step": 4066 }, { "epoch": 0.32, "learning_rate": 1.6017492140888822e-06, "loss": 1.2286, "step": 4067 }, { "epoch": 0.32, "learning_rate": 1.6015484154313397e-06, "loss": 1.2271, "step": 4068 }, { "epoch": 0.32, "learning_rate": 1.601347578758465e-06, "loss": 1.2598, "step": 4069 }, { "epoch": 0.32, "learning_rate": 1.60114670408295e-06, "loss": 1.3061, "step": 4070 }, { "epoch": 0.32, "learning_rate": 1.6009457914174894e-06, "loss": 1.2137, "step": 4071 }, { "epoch": 0.32, "learning_rate": 1.6007448407747798e-06, "loss": 1.1915, "step": 4072 }, { "epoch": 0.32, "learning_rate": 1.6005438521675205e-06, "loss": 1.2077, "step": 4073 }, { "epoch": 0.32, "learning_rate": 1.6003428256084127e-06, "loss": 1.2495, "step": 4074 }, { "epoch": 0.32, "learning_rate": 1.6001417611101612e-06, "loss": 1.2685, "step": 4075 }, { "epoch": 0.32, "learning_rate": 1.5999406586854722e-06, "loss": 1.2215, "step": 4076 }, { "epoch": 0.32, "learning_rate": 1.599739518347054e-06, "loss": 1.2294, "step": 4077 }, { "epoch": 0.32, "learning_rate": 1.5995383401076188e-06, "loss": 1.2554, "step": 4078 }, { "epoch": 0.32, "learning_rate": 1.5993371239798794e-06, "loss": 0.8797, "step": 4079 }, { "epoch": 0.32, "learning_rate": 1.599135869976552e-06, "loss": 1.2231, "step": 4080 }, { "epoch": 0.32, "learning_rate": 1.5989345781103552e-06, "loss": 1.278, "step": 4081 }, { "epoch": 0.32, "learning_rate": 1.5987332483940094e-06, "loss": 1.2194, "step": 4082 }, { "epoch": 0.32, "learning_rate": 1.5985318808402383e-06, "loss": 1.2305, "step": 4083 }, { "epoch": 0.32, "learning_rate": 1.5983304754617676e-06, "loss": 1.2157, "step": 4084 }, { "epoch": 0.32, "learning_rate": 1.5981290322713244e-06, "loss": 1.2668, "step": 4085 }, { "epoch": 0.32, "learning_rate": 1.5979275512816396e-06, "loss": 1.2372, "step": 4086 }, { "epoch": 0.32, "learning_rate": 1.5977260325054461e-06, "loss": 1.309, "step": 4087 }, { "epoch": 0.32, "learning_rate": 1.597524475955479e-06, "loss": 0.836, "step": 4088 }, { "epoch": 0.32, "learning_rate": 1.5973228816444756e-06, "loss": 0.8623, "step": 4089 }, { "epoch": 0.32, "learning_rate": 1.5971212495851761e-06, "loss": 1.219, "step": 4090 }, { "epoch": 0.32, "learning_rate": 1.5969195797903223e-06, "loss": 1.2324, "step": 4091 }, { "epoch": 0.32, "learning_rate": 1.5967178722726594e-06, "loss": 1.2236, "step": 4092 }, { "epoch": 0.32, "learning_rate": 1.5965161270449343e-06, "loss": 1.2357, "step": 4093 }, { "epoch": 0.32, "learning_rate": 1.596314344119896e-06, "loss": 1.2246, "step": 4094 }, { "epoch": 0.32, "learning_rate": 1.5961125235102974e-06, "loss": 0.8677, "step": 4095 }, { "epoch": 0.32, "learning_rate": 1.5959106652288919e-06, "loss": 1.2194, "step": 4096 }, { "epoch": 0.32, "learning_rate": 1.595708769288436e-06, "loss": 1.2231, "step": 4097 }, { "epoch": 0.32, "learning_rate": 1.595506835701689e-06, "loss": 1.1751, "step": 4098 }, { "epoch": 0.32, "learning_rate": 1.5953048644814125e-06, "loss": 1.1608, "step": 4099 }, { "epoch": 0.32, "learning_rate": 1.5951028556403697e-06, "loss": 1.1723, "step": 4100 }, { "epoch": 0.32, "learning_rate": 1.5949008091913272e-06, "loss": 1.2616, "step": 4101 }, { "epoch": 0.32, "learning_rate": 1.5946987251470526e-06, "loss": 1.2161, "step": 4102 }, { "epoch": 0.32, "learning_rate": 1.594496603520318e-06, "loss": 1.2622, "step": 4103 }, { "epoch": 0.32, "learning_rate": 1.5942944443238958e-06, "loss": 1.1895, "step": 4104 }, { "epoch": 0.32, "learning_rate": 1.5940922475705617e-06, "loss": 1.2445, "step": 4105 }, { "epoch": 0.32, "learning_rate": 1.5938900132730942e-06, "loss": 1.1895, "step": 4106 }, { "epoch": 0.32, "learning_rate": 1.593687741444273e-06, "loss": 1.2209, "step": 4107 }, { "epoch": 0.32, "learning_rate": 1.593485432096881e-06, "loss": 1.2451, "step": 4108 }, { "epoch": 0.32, "learning_rate": 1.5932830852437037e-06, "loss": 1.2247, "step": 4109 }, { "epoch": 0.32, "learning_rate": 1.5930807008975283e-06, "loss": 1.3261, "step": 4110 }, { "epoch": 0.32, "learning_rate": 1.5928782790711441e-06, "loss": 1.238, "step": 4111 }, { "epoch": 0.32, "learning_rate": 1.5926758197773444e-06, "loss": 1.2496, "step": 4112 }, { "epoch": 0.32, "learning_rate": 1.5924733230289231e-06, "loss": 1.281, "step": 4113 }, { "epoch": 0.32, "learning_rate": 1.5922707888386773e-06, "loss": 1.2404, "step": 4114 }, { "epoch": 0.32, "learning_rate": 1.592068217219406e-06, "loss": 1.2207, "step": 4115 }, { "epoch": 0.32, "learning_rate": 1.5918656081839113e-06, "loss": 0.8547, "step": 4116 }, { "epoch": 0.32, "learning_rate": 1.5916629617449972e-06, "loss": 1.2607, "step": 4117 }, { "epoch": 0.32, "learning_rate": 1.5914602779154698e-06, "loss": 0.8194, "step": 4118 }, { "epoch": 0.32, "learning_rate": 1.5912575567081379e-06, "loss": 1.2025, "step": 4119 }, { "epoch": 0.32, "learning_rate": 1.591054798135813e-06, "loss": 1.1924, "step": 4120 }, { "epoch": 0.32, "learning_rate": 1.5908520022113083e-06, "loss": 1.2324, "step": 4121 }, { "epoch": 0.32, "learning_rate": 1.5906491689474397e-06, "loss": 1.2039, "step": 4122 }, { "epoch": 0.32, "learning_rate": 1.5904462983570257e-06, "loss": 0.837, "step": 4123 }, { "epoch": 0.32, "learning_rate": 1.5902433904528864e-06, "loss": 1.2222, "step": 4124 }, { "epoch": 0.32, "learning_rate": 1.590040445247845e-06, "loss": 1.1972, "step": 4125 }, { "epoch": 0.32, "learning_rate": 1.5898374627547262e-06, "loss": 0.8919, "step": 4126 }, { "epoch": 0.32, "learning_rate": 1.5896344429863586e-06, "loss": 1.2875, "step": 4127 }, { "epoch": 0.32, "learning_rate": 1.5894313859555718e-06, "loss": 1.215, "step": 4128 }, { "epoch": 0.32, "learning_rate": 1.589228291675198e-06, "loss": 1.2577, "step": 4129 }, { "epoch": 0.32, "learning_rate": 1.589025160158072e-06, "loss": 1.2533, "step": 4130 }, { "epoch": 0.32, "learning_rate": 1.588821991417031e-06, "loss": 1.2551, "step": 4131 }, { "epoch": 0.32, "learning_rate": 1.5886187854649137e-06, "loss": 1.2636, "step": 4132 }, { "epoch": 0.32, "learning_rate": 1.5884155423145628e-06, "loss": 1.2238, "step": 4133 }, { "epoch": 0.32, "learning_rate": 1.5882122619788224e-06, "loss": 0.8469, "step": 4134 }, { "epoch": 0.32, "learning_rate": 1.588008944470538e-06, "loss": 1.231, "step": 4135 }, { "epoch": 0.32, "learning_rate": 1.5878055898025596e-06, "loss": 1.227, "step": 4136 }, { "epoch": 0.32, "learning_rate": 1.5876021979877373e-06, "loss": 1.2092, "step": 4137 }, { "epoch": 0.32, "learning_rate": 1.5873987690389253e-06, "loss": 1.1726, "step": 4138 }, { "epoch": 0.32, "learning_rate": 1.5871953029689794e-06, "loss": 0.8299, "step": 4139 }, { "epoch": 0.32, "learning_rate": 1.5869917997907574e-06, "loss": 0.8799, "step": 4140 }, { "epoch": 0.32, "learning_rate": 1.58678825951712e-06, "loss": 1.1831, "step": 4141 }, { "epoch": 0.32, "learning_rate": 1.58658468216093e-06, "loss": 1.2374, "step": 4142 }, { "epoch": 0.32, "learning_rate": 1.5863810677350533e-06, "loss": 1.2415, "step": 4143 }, { "epoch": 0.32, "learning_rate": 1.586177416252357e-06, "loss": 1.3324, "step": 4144 }, { "epoch": 0.32, "learning_rate": 1.5859737277257108e-06, "loss": 1.2586, "step": 4145 }, { "epoch": 0.32, "learning_rate": 1.5857700021679871e-06, "loss": 1.2592, "step": 4146 }, { "epoch": 0.32, "learning_rate": 1.5855662395920607e-06, "loss": 1.2526, "step": 4147 }, { "epoch": 0.32, "learning_rate": 1.5853624400108084e-06, "loss": 0.8437, "step": 4148 }, { "epoch": 0.32, "learning_rate": 1.5851586034371094e-06, "loss": 1.1662, "step": 4149 }, { "epoch": 0.32, "learning_rate": 1.5849547298838455e-06, "loss": 1.1953, "step": 4150 }, { "epoch": 0.32, "learning_rate": 1.5847508193639002e-06, "loss": 1.2295, "step": 4151 }, { "epoch": 0.32, "learning_rate": 1.5845468718901606e-06, "loss": 1.2171, "step": 4152 }, { "epoch": 0.32, "learning_rate": 1.5843428874755146e-06, "loss": 1.2136, "step": 4153 }, { "epoch": 0.32, "learning_rate": 1.5841388661328537e-06, "loss": 1.2788, "step": 4154 }, { "epoch": 0.32, "learning_rate": 1.5839348078750705e-06, "loss": 1.2293, "step": 4155 }, { "epoch": 0.32, "learning_rate": 1.5837307127150612e-06, "loss": 1.2384, "step": 4156 }, { "epoch": 0.32, "learning_rate": 1.5835265806657237e-06, "loss": 1.186, "step": 4157 }, { "epoch": 0.32, "learning_rate": 1.5833224117399578e-06, "loss": 1.219, "step": 4158 }, { "epoch": 0.32, "learning_rate": 1.5831182059506672e-06, "loss": 1.2413, "step": 4159 }, { "epoch": 0.32, "learning_rate": 1.5829139633107555e-06, "loss": 1.2098, "step": 4160 }, { "epoch": 0.32, "learning_rate": 1.5827096838331308e-06, "loss": 0.8439, "step": 4161 }, { "epoch": 0.32, "learning_rate": 1.582505367530703e-06, "loss": 1.16, "step": 4162 }, { "epoch": 0.32, "learning_rate": 1.582301014416383e-06, "loss": 0.8331, "step": 4163 }, { "epoch": 0.32, "learning_rate": 1.5820966245030856e-06, "loss": 1.2296, "step": 4164 }, { "epoch": 0.32, "learning_rate": 1.5818921978037276e-06, "loss": 1.22, "step": 4165 }, { "epoch": 0.32, "learning_rate": 1.5816877343312276e-06, "loss": 0.8605, "step": 4166 }, { "epoch": 0.32, "learning_rate": 1.5814832340985072e-06, "loss": 1.216, "step": 4167 }, { "epoch": 0.32, "learning_rate": 1.5812786971184892e-06, "loss": 1.2424, "step": 4168 }, { "epoch": 0.32, "learning_rate": 1.5810741234041002e-06, "loss": 1.2593, "step": 4169 }, { "epoch": 0.32, "learning_rate": 1.5808695129682685e-06, "loss": 0.8184, "step": 4170 }, { "epoch": 0.32, "learning_rate": 1.5806648658239239e-06, "loss": 1.2182, "step": 4171 }, { "epoch": 0.32, "learning_rate": 1.5804601819839997e-06, "loss": 1.2307, "step": 4172 }, { "epoch": 0.32, "learning_rate": 1.5802554614614312e-06, "loss": 1.2065, "step": 4173 }, { "epoch": 0.32, "learning_rate": 1.5800507042691553e-06, "loss": 1.2309, "step": 4174 }, { "epoch": 0.32, "learning_rate": 1.5798459104201124e-06, "loss": 1.2283, "step": 4175 }, { "epoch": 0.32, "learning_rate": 1.5796410799272443e-06, "loss": 1.2317, "step": 4176 }, { "epoch": 0.32, "learning_rate": 1.5794362128034954e-06, "loss": 1.2728, "step": 4177 }, { "epoch": 0.32, "learning_rate": 1.579231309061813e-06, "loss": 1.2222, "step": 4178 }, { "epoch": 0.32, "learning_rate": 1.5790263687151452e-06, "loss": 0.883, "step": 4179 }, { "epoch": 0.32, "learning_rate": 1.5788213917764442e-06, "loss": 0.8028, "step": 4180 }, { "epoch": 0.32, "learning_rate": 1.5786163782586632e-06, "loss": 1.2316, "step": 4181 }, { "epoch": 0.32, "learning_rate": 1.5784113281747584e-06, "loss": 1.1995, "step": 4182 }, { "epoch": 0.32, "learning_rate": 1.5782062415376877e-06, "loss": 1.2709, "step": 4183 }, { "epoch": 0.32, "learning_rate": 1.5780011183604125e-06, "loss": 1.2412, "step": 4184 }, { "epoch": 0.32, "learning_rate": 1.577795958655895e-06, "loss": 1.2481, "step": 4185 }, { "epoch": 0.32, "learning_rate": 1.5775907624371009e-06, "loss": 1.2087, "step": 4186 }, { "epoch": 0.32, "learning_rate": 1.5773855297169973e-06, "loss": 0.8388, "step": 4187 }, { "epoch": 0.33, "learning_rate": 1.5771802605085544e-06, "loss": 1.2312, "step": 4188 }, { "epoch": 0.33, "learning_rate": 1.5769749548247444e-06, "loss": 1.2352, "step": 4189 }, { "epoch": 0.33, "learning_rate": 1.5767696126785418e-06, "loss": 1.1985, "step": 4190 }, { "epoch": 0.33, "learning_rate": 1.5765642340829226e-06, "loss": 1.2042, "step": 4191 }, { "epoch": 0.33, "learning_rate": 1.5763588190508665e-06, "loss": 1.2114, "step": 4192 }, { "epoch": 0.33, "learning_rate": 1.576153367595355e-06, "loss": 1.1985, "step": 4193 }, { "epoch": 0.33, "learning_rate": 1.5759478797293718e-06, "loss": 1.2204, "step": 4194 }, { "epoch": 0.33, "learning_rate": 1.5757423554659024e-06, "loss": 1.2167, "step": 4195 }, { "epoch": 0.33, "learning_rate": 1.575536794817935e-06, "loss": 1.2011, "step": 4196 }, { "epoch": 0.33, "learning_rate": 1.5753311977984609e-06, "loss": 1.2538, "step": 4197 }, { "epoch": 0.33, "learning_rate": 1.5751255644204724e-06, "loss": 1.2066, "step": 4198 }, { "epoch": 0.33, "learning_rate": 1.5749198946969647e-06, "loss": 1.2719, "step": 4199 }, { "epoch": 0.33, "learning_rate": 1.5747141886409354e-06, "loss": 1.2664, "step": 4200 }, { "epoch": 0.33, "learning_rate": 1.5745084462653844e-06, "loss": 1.215, "step": 4201 }, { "epoch": 0.33, "learning_rate": 1.5743026675833134e-06, "loss": 0.8345, "step": 4202 }, { "epoch": 0.33, "learning_rate": 1.574096852607727e-06, "loss": 1.2397, "step": 4203 }, { "epoch": 0.33, "learning_rate": 1.5738910013516315e-06, "loss": 1.2492, "step": 4204 }, { "epoch": 0.33, "learning_rate": 1.5736851138280365e-06, "loss": 1.2692, "step": 4205 }, { "epoch": 0.33, "learning_rate": 1.5734791900499529e-06, "loss": 1.2358, "step": 4206 }, { "epoch": 0.33, "learning_rate": 1.5732732300303938e-06, "loss": 1.2431, "step": 4207 }, { "epoch": 0.33, "learning_rate": 1.5730672337823754e-06, "loss": 1.1943, "step": 4208 }, { "epoch": 0.33, "learning_rate": 1.5728612013189157e-06, "loss": 1.2485, "step": 4209 }, { "epoch": 0.33, "learning_rate": 1.5726551326530354e-06, "loss": 1.3119, "step": 4210 }, { "epoch": 0.33, "learning_rate": 1.5724490277977569e-06, "loss": 1.2567, "step": 4211 }, { "epoch": 0.33, "learning_rate": 1.572242886766105e-06, "loss": 1.2546, "step": 4212 }, { "epoch": 0.33, "learning_rate": 1.5720367095711071e-06, "loss": 1.1974, "step": 4213 }, { "epoch": 0.33, "learning_rate": 1.571830496225793e-06, "loss": 1.2204, "step": 4214 }, { "epoch": 0.33, "learning_rate": 1.5716242467431937e-06, "loss": 1.2627, "step": 4215 }, { "epoch": 0.33, "learning_rate": 1.5714179611363445e-06, "loss": 1.2301, "step": 4216 }, { "epoch": 0.33, "learning_rate": 1.5712116394182808e-06, "loss": 1.2495, "step": 4217 }, { "epoch": 0.33, "learning_rate": 1.5710052816020419e-06, "loss": 1.2283, "step": 4218 }, { "epoch": 0.33, "learning_rate": 1.5707988877006684e-06, "loss": 1.1701, "step": 4219 }, { "epoch": 0.33, "learning_rate": 1.570592457727203e-06, "loss": 0.8571, "step": 4220 }, { "epoch": 0.33, "learning_rate": 1.5703859916946925e-06, "loss": 1.1885, "step": 4221 }, { "epoch": 0.33, "learning_rate": 1.570179489616184e-06, "loss": 1.2479, "step": 4222 }, { "epoch": 0.33, "learning_rate": 1.5699729515047273e-06, "loss": 0.8239, "step": 4223 }, { "epoch": 0.33, "learning_rate": 1.5697663773733752e-06, "loss": 1.2073, "step": 4224 }, { "epoch": 0.33, "learning_rate": 1.5695597672351815e-06, "loss": 1.2013, "step": 4225 }, { "epoch": 0.33, "learning_rate": 1.5693531211032042e-06, "loss": 1.2577, "step": 4226 }, { "epoch": 0.33, "learning_rate": 1.569146438990502e-06, "loss": 1.1993, "step": 4227 }, { "epoch": 0.33, "learning_rate": 1.5689397209101363e-06, "loss": 1.1878, "step": 4228 }, { "epoch": 0.33, "learning_rate": 1.5687329668751708e-06, "loss": 1.2182, "step": 4229 }, { "epoch": 0.33, "learning_rate": 1.5685261768986712e-06, "loss": 1.2569, "step": 4230 }, { "epoch": 0.33, "learning_rate": 1.5683193509937062e-06, "loss": 1.2, "step": 4231 }, { "epoch": 0.33, "learning_rate": 1.5681124891733463e-06, "loss": 1.2721, "step": 4232 }, { "epoch": 0.33, "learning_rate": 1.5679055914506644e-06, "loss": 1.2429, "step": 4233 }, { "epoch": 0.33, "learning_rate": 1.567698657838735e-06, "loss": 1.2769, "step": 4234 }, { "epoch": 0.33, "learning_rate": 1.567491688350636e-06, "loss": 1.2413, "step": 4235 }, { "epoch": 0.33, "learning_rate": 1.5672846829994465e-06, "loss": 1.2134, "step": 4236 }, { "epoch": 0.33, "learning_rate": 1.567077641798249e-06, "loss": 1.2422, "step": 4237 }, { "epoch": 0.33, "learning_rate": 1.5668705647601273e-06, "loss": 1.2248, "step": 4238 }, { "epoch": 0.33, "learning_rate": 1.5666634518981677e-06, "loss": 0.8436, "step": 4239 }, { "epoch": 0.33, "learning_rate": 1.566456303225459e-06, "loss": 1.2252, "step": 4240 }, { "epoch": 0.33, "learning_rate": 1.566249118755092e-06, "loss": 1.2027, "step": 4241 }, { "epoch": 0.33, "learning_rate": 1.56604189850016e-06, "loss": 1.2569, "step": 4242 }, { "epoch": 0.33, "learning_rate": 1.5658346424737582e-06, "loss": 1.2113, "step": 4243 }, { "epoch": 0.33, "learning_rate": 1.565627350688985e-06, "loss": 1.2087, "step": 4244 }, { "epoch": 0.33, "learning_rate": 1.5654200231589396e-06, "loss": 1.233, "step": 4245 }, { "epoch": 0.33, "learning_rate": 1.5652126598967245e-06, "loss": 1.2419, "step": 4246 }, { "epoch": 0.33, "learning_rate": 1.5650052609154445e-06, "loss": 1.1823, "step": 4247 }, { "epoch": 0.33, "learning_rate": 1.5647978262282057e-06, "loss": 1.1917, "step": 4248 }, { "epoch": 0.33, "learning_rate": 1.5645903558481177e-06, "loss": 0.8206, "step": 4249 }, { "epoch": 0.33, "learning_rate": 1.5643828497882916e-06, "loss": 1.1948, "step": 4250 }, { "epoch": 0.33, "learning_rate": 1.5641753080618408e-06, "loss": 1.1676, "step": 4251 }, { "epoch": 0.33, "learning_rate": 1.563967730681881e-06, "loss": 1.2625, "step": 4252 }, { "epoch": 0.33, "learning_rate": 1.56376011766153e-06, "loss": 1.2569, "step": 4253 }, { "epoch": 0.33, "learning_rate": 1.5635524690139087e-06, "loss": 1.2381, "step": 4254 }, { "epoch": 0.33, "learning_rate": 1.5633447847521396e-06, "loss": 1.2276, "step": 4255 }, { "epoch": 0.33, "learning_rate": 1.5631370648893466e-06, "loss": 1.2414, "step": 4256 }, { "epoch": 0.33, "learning_rate": 1.5629293094386578e-06, "loss": 1.1821, "step": 4257 }, { "epoch": 0.33, "learning_rate": 1.5627215184132015e-06, "loss": 1.2318, "step": 4258 }, { "epoch": 0.33, "learning_rate": 1.5625136918261098e-06, "loss": 1.2205, "step": 4259 }, { "epoch": 0.33, "learning_rate": 1.5623058296905164e-06, "loss": 1.2699, "step": 4260 }, { "epoch": 0.33, "learning_rate": 1.5620979320195575e-06, "loss": 1.232, "step": 4261 }, { "epoch": 0.33, "learning_rate": 1.5618899988263712e-06, "loss": 0.8427, "step": 4262 }, { "epoch": 0.33, "learning_rate": 1.5616820301240976e-06, "loss": 1.2211, "step": 4263 }, { "epoch": 0.33, "learning_rate": 1.56147402592588e-06, "loss": 1.1907, "step": 4264 }, { "epoch": 0.33, "learning_rate": 1.561265986244863e-06, "loss": 1.2776, "step": 4265 }, { "epoch": 0.33, "learning_rate": 1.5610579110941942e-06, "loss": 1.2311, "step": 4266 }, { "epoch": 0.33, "learning_rate": 1.5608498004870227e-06, "loss": 1.2099, "step": 4267 }, { "epoch": 0.33, "learning_rate": 1.5606416544365005e-06, "loss": 1.2567, "step": 4268 }, { "epoch": 0.33, "learning_rate": 1.5604334729557816e-06, "loss": 1.2517, "step": 4269 }, { "epoch": 0.33, "learning_rate": 1.5602252560580218e-06, "loss": 1.2066, "step": 4270 }, { "epoch": 0.33, "learning_rate": 1.5600170037563797e-06, "loss": 1.2591, "step": 4271 }, { "epoch": 0.33, "learning_rate": 1.5598087160640164e-06, "loss": 1.2416, "step": 4272 }, { "epoch": 0.33, "learning_rate": 1.5596003929940944e-06, "loss": 1.2718, "step": 4273 }, { "epoch": 0.33, "learning_rate": 1.5593920345597786e-06, "loss": 1.2216, "step": 4274 }, { "epoch": 0.33, "learning_rate": 1.559183640774237e-06, "loss": 1.1984, "step": 4275 }, { "epoch": 0.33, "learning_rate": 1.5589752116506388e-06, "loss": 1.23, "step": 4276 }, { "epoch": 0.33, "learning_rate": 1.558766747202156e-06, "loss": 1.2466, "step": 4277 }, { "epoch": 0.33, "learning_rate": 1.5585582474419626e-06, "loss": 0.8294, "step": 4278 }, { "epoch": 0.33, "learning_rate": 1.5583497123832349e-06, "loss": 1.1891, "step": 4279 }, { "epoch": 0.33, "learning_rate": 1.558141142039151e-06, "loss": 1.2387, "step": 4280 }, { "epoch": 0.33, "learning_rate": 1.5579325364228929e-06, "loss": 1.2147, "step": 4281 }, { "epoch": 0.33, "learning_rate": 1.5577238955476425e-06, "loss": 1.2279, "step": 4282 }, { "epoch": 0.33, "learning_rate": 1.5575152194265852e-06, "loss": 1.2199, "step": 4283 }, { "epoch": 0.33, "learning_rate": 1.5573065080729086e-06, "loss": 1.2223, "step": 4284 }, { "epoch": 0.33, "learning_rate": 1.5570977614998027e-06, "loss": 1.2272, "step": 4285 }, { "epoch": 0.33, "learning_rate": 1.5568889797204587e-06, "loss": 1.1584, "step": 4286 }, { "epoch": 0.33, "learning_rate": 1.5566801627480714e-06, "loss": 1.2595, "step": 4287 }, { "epoch": 0.33, "learning_rate": 1.5564713105958372e-06, "loss": 1.2263, "step": 4288 }, { "epoch": 0.33, "learning_rate": 1.556262423276954e-06, "loss": 1.1897, "step": 4289 }, { "epoch": 0.33, "learning_rate": 1.556053500804623e-06, "loss": 1.1978, "step": 4290 }, { "epoch": 0.33, "learning_rate": 1.5558445431920468e-06, "loss": 1.277, "step": 4291 }, { "epoch": 0.33, "learning_rate": 1.5556355504524316e-06, "loss": 1.2682, "step": 4292 }, { "epoch": 0.33, "learning_rate": 1.5554265225989843e-06, "loss": 1.2405, "step": 4293 }, { "epoch": 0.33, "learning_rate": 1.5552174596449144e-06, "loss": 1.2313, "step": 4294 }, { "epoch": 0.33, "learning_rate": 1.5550083616034338e-06, "loss": 1.1675, "step": 4295 }, { "epoch": 0.33, "learning_rate": 1.5547992284877568e-06, "loss": 1.2023, "step": 4296 }, { "epoch": 0.33, "learning_rate": 1.5545900603110998e-06, "loss": 1.2268, "step": 4297 }, { "epoch": 0.33, "learning_rate": 1.5543808570866813e-06, "loss": 0.8727, "step": 4298 }, { "epoch": 0.33, "learning_rate": 1.5541716188277223e-06, "loss": 1.2114, "step": 4299 }, { "epoch": 0.33, "learning_rate": 1.5539623455474454e-06, "loss": 1.2604, "step": 4300 }, { "epoch": 0.33, "learning_rate": 1.5537530372590757e-06, "loss": 1.2307, "step": 4301 }, { "epoch": 0.33, "learning_rate": 1.553543693975841e-06, "loss": 1.2317, "step": 4302 }, { "epoch": 0.33, "learning_rate": 1.5533343157109705e-06, "loss": 1.3203, "step": 4303 }, { "epoch": 0.33, "learning_rate": 1.5531249024776964e-06, "loss": 1.1904, "step": 4304 }, { "epoch": 0.33, "learning_rate": 1.5529154542892528e-06, "loss": 1.2211, "step": 4305 }, { "epoch": 0.33, "learning_rate": 1.5527059711588756e-06, "loss": 1.2393, "step": 4306 }, { "epoch": 0.33, "learning_rate": 1.5524964530998035e-06, "loss": 1.248, "step": 4307 }, { "epoch": 0.33, "learning_rate": 1.5522869001252767e-06, "loss": 1.2968, "step": 4308 }, { "epoch": 0.33, "learning_rate": 1.5520773122485387e-06, "loss": 1.252, "step": 4309 }, { "epoch": 0.33, "learning_rate": 1.5518676894828343e-06, "loss": 1.2415, "step": 4310 }, { "epoch": 0.33, "learning_rate": 1.5516580318414108e-06, "loss": 1.2883, "step": 4311 }, { "epoch": 0.33, "learning_rate": 1.5514483393375174e-06, "loss": 1.2491, "step": 4312 }, { "epoch": 0.33, "learning_rate": 1.5512386119844063e-06, "loss": 1.2428, "step": 4313 }, { "epoch": 0.33, "learning_rate": 1.551028849795331e-06, "loss": 1.1415, "step": 4314 }, { "epoch": 0.33, "learning_rate": 1.550819052783548e-06, "loss": 1.2103, "step": 4315 }, { "epoch": 0.33, "learning_rate": 1.550609220962315e-06, "loss": 1.2297, "step": 4316 }, { "epoch": 0.34, "learning_rate": 1.5503993543448928e-06, "loss": 1.2481, "step": 4317 }, { "epoch": 0.34, "learning_rate": 1.550189452944544e-06, "loss": 1.2209, "step": 4318 }, { "epoch": 0.34, "learning_rate": 1.5499795167745337e-06, "loss": 0.898, "step": 4319 }, { "epoch": 0.34, "learning_rate": 1.549769545848129e-06, "loss": 1.2197, "step": 4320 }, { "epoch": 0.34, "learning_rate": 1.549559540178599e-06, "loss": 1.1875, "step": 4321 }, { "epoch": 0.34, "learning_rate": 1.549349499779215e-06, "loss": 1.2069, "step": 4322 }, { "epoch": 0.34, "learning_rate": 1.549139424663251e-06, "loss": 1.2311, "step": 4323 }, { "epoch": 0.34, "learning_rate": 1.5489293148439831e-06, "loss": 1.2395, "step": 4324 }, { "epoch": 0.34, "learning_rate": 1.5487191703346885e-06, "loss": 1.2111, "step": 4325 }, { "epoch": 0.34, "learning_rate": 1.548508991148648e-06, "loss": 1.192, "step": 4326 }, { "epoch": 0.34, "learning_rate": 1.5482987772991443e-06, "loss": 1.2397, "step": 4327 }, { "epoch": 0.34, "learning_rate": 1.5480885287994614e-06, "loss": 1.2765, "step": 4328 }, { "epoch": 0.34, "learning_rate": 1.5478782456628867e-06, "loss": 1.1719, "step": 4329 }, { "epoch": 0.34, "learning_rate": 1.5476679279027089e-06, "loss": 1.2284, "step": 4330 }, { "epoch": 0.34, "learning_rate": 1.5474575755322191e-06, "loss": 1.1716, "step": 4331 }, { "epoch": 0.34, "learning_rate": 1.5472471885647112e-06, "loss": 1.2293, "step": 4332 }, { "epoch": 0.34, "learning_rate": 1.5470367670134801e-06, "loss": 1.2128, "step": 4333 }, { "epoch": 0.34, "learning_rate": 1.546826310891824e-06, "loss": 1.2103, "step": 4334 }, { "epoch": 0.34, "learning_rate": 1.5466158202130423e-06, "loss": 1.2833, "step": 4335 }, { "epoch": 0.34, "learning_rate": 1.5464052949904382e-06, "loss": 1.1918, "step": 4336 }, { "epoch": 0.34, "learning_rate": 1.546194735237315e-06, "loss": 1.2019, "step": 4337 }, { "epoch": 0.34, "learning_rate": 1.5459841409669798e-06, "loss": 1.206, "step": 4338 }, { "epoch": 0.34, "learning_rate": 1.5457735121927406e-06, "loss": 1.1916, "step": 4339 }, { "epoch": 0.34, "learning_rate": 1.5455628489279091e-06, "loss": 1.2237, "step": 4340 }, { "epoch": 0.34, "learning_rate": 1.545352151185798e-06, "loss": 1.202, "step": 4341 }, { "epoch": 0.34, "learning_rate": 1.5451414189797219e-06, "loss": 1.2838, "step": 4342 }, { "epoch": 0.34, "learning_rate": 1.5449306523229991e-06, "loss": 1.1869, "step": 4343 }, { "epoch": 0.34, "learning_rate": 1.5447198512289485e-06, "loss": 0.8451, "step": 4344 }, { "epoch": 0.34, "learning_rate": 1.5445090157108923e-06, "loss": 1.2396, "step": 4345 }, { "epoch": 0.34, "learning_rate": 1.5442981457821542e-06, "loss": 1.1551, "step": 4346 }, { "epoch": 0.34, "learning_rate": 1.5440872414560603e-06, "loss": 1.2377, "step": 4347 }, { "epoch": 0.34, "learning_rate": 1.5438763027459389e-06, "loss": 1.2267, "step": 4348 }, { "epoch": 0.34, "learning_rate": 1.5436653296651206e-06, "loss": 1.2178, "step": 4349 }, { "epoch": 0.34, "learning_rate": 1.5434543222269375e-06, "loss": 1.2065, "step": 4350 }, { "epoch": 0.34, "learning_rate": 1.543243280444725e-06, "loss": 1.2166, "step": 4351 }, { "epoch": 0.34, "learning_rate": 1.5430322043318197e-06, "loss": 1.2499, "step": 4352 }, { "epoch": 0.34, "learning_rate": 1.5428210939015607e-06, "loss": 1.2128, "step": 4353 }, { "epoch": 0.34, "learning_rate": 1.5426099491672897e-06, "loss": 1.2378, "step": 4354 }, { "epoch": 0.34, "learning_rate": 1.5423987701423498e-06, "loss": 1.199, "step": 4355 }, { "epoch": 0.34, "learning_rate": 1.5421875568400864e-06, "loss": 1.2469, "step": 4356 }, { "epoch": 0.34, "learning_rate": 1.5419763092738478e-06, "loss": 1.2369, "step": 4357 }, { "epoch": 0.34, "learning_rate": 1.5417650274569837e-06, "loss": 1.2217, "step": 4358 }, { "epoch": 0.34, "learning_rate": 1.5415537114028462e-06, "loss": 1.2339, "step": 4359 }, { "epoch": 0.34, "learning_rate": 1.5413423611247898e-06, "loss": 1.222, "step": 4360 }, { "epoch": 0.34, "learning_rate": 1.541130976636171e-06, "loss": 1.2695, "step": 4361 }, { "epoch": 0.34, "learning_rate": 1.5409195579503477e-06, "loss": 1.1929, "step": 4362 }, { "epoch": 0.34, "learning_rate": 1.5407081050806814e-06, "loss": 1.2461, "step": 4363 }, { "epoch": 0.34, "learning_rate": 1.5404966180405347e-06, "loss": 1.195, "step": 4364 }, { "epoch": 0.34, "learning_rate": 1.5402850968432732e-06, "loss": 1.2402, "step": 4365 }, { "epoch": 0.34, "learning_rate": 1.5400735415022634e-06, "loss": 1.1909, "step": 4366 }, { "epoch": 0.34, "learning_rate": 1.5398619520308753e-06, "loss": 1.1572, "step": 4367 }, { "epoch": 0.34, "learning_rate": 1.5396503284424801e-06, "loss": 1.2004, "step": 4368 }, { "epoch": 0.34, "learning_rate": 1.5394386707504516e-06, "loss": 1.2554, "step": 4369 }, { "epoch": 0.34, "learning_rate": 1.539226978968166e-06, "loss": 1.2237, "step": 4370 }, { "epoch": 0.34, "learning_rate": 1.5390152531090008e-06, "loss": 1.1863, "step": 4371 }, { "epoch": 0.34, "learning_rate": 1.5388034931863367e-06, "loss": 1.205, "step": 4372 }, { "epoch": 0.34, "learning_rate": 1.5385916992135557e-06, "loss": 0.8583, "step": 4373 }, { "epoch": 0.34, "learning_rate": 1.538379871204042e-06, "loss": 1.2077, "step": 4374 }, { "epoch": 0.34, "learning_rate": 1.538168009171183e-06, "loss": 1.1554, "step": 4375 }, { "epoch": 0.34, "learning_rate": 1.5379561131283673e-06, "loss": 1.1924, "step": 4376 }, { "epoch": 0.34, "learning_rate": 1.5377441830889852e-06, "loss": 0.8293, "step": 4377 }, { "epoch": 0.34, "learning_rate": 1.5375322190664304e-06, "loss": 1.2056, "step": 4378 }, { "epoch": 0.34, "learning_rate": 1.5373202210740979e-06, "loss": 1.2644, "step": 4379 }, { "epoch": 0.34, "learning_rate": 1.5371081891253855e-06, "loss": 1.2343, "step": 4380 }, { "epoch": 0.34, "learning_rate": 1.5368961232336918e-06, "loss": 0.8627, "step": 4381 }, { "epoch": 0.34, "learning_rate": 1.5366840234124195e-06, "loss": 1.2452, "step": 4382 }, { "epoch": 0.34, "learning_rate": 1.5364718896749718e-06, "loss": 1.2329, "step": 4383 }, { "epoch": 0.34, "learning_rate": 1.5362597220347548e-06, "loss": 1.2312, "step": 4384 }, { "epoch": 0.34, "learning_rate": 1.5360475205051768e-06, "loss": 1.1793, "step": 4385 }, { "epoch": 0.34, "learning_rate": 1.5358352850996479e-06, "loss": 1.2525, "step": 4386 }, { "epoch": 0.34, "learning_rate": 1.5356230158315802e-06, "loss": 1.2727, "step": 4387 }, { "epoch": 0.34, "learning_rate": 1.5354107127143887e-06, "loss": 1.2104, "step": 4388 }, { "epoch": 0.34, "learning_rate": 1.5351983757614897e-06, "loss": 1.1705, "step": 4389 }, { "epoch": 0.34, "learning_rate": 1.5349860049863023e-06, "loss": 1.1872, "step": 4390 }, { "epoch": 0.34, "learning_rate": 1.5347736004022473e-06, "loss": 0.8155, "step": 4391 }, { "epoch": 0.34, "learning_rate": 1.5345611620227477e-06, "loss": 1.2177, "step": 4392 }, { "epoch": 0.34, "learning_rate": 1.534348689861229e-06, "loss": 1.1768, "step": 4393 }, { "epoch": 0.34, "learning_rate": 1.5341361839311182e-06, "loss": 1.2134, "step": 4394 }, { "epoch": 0.34, "learning_rate": 1.5339236442458448e-06, "loss": 1.2393, "step": 4395 }, { "epoch": 0.34, "learning_rate": 1.5337110708188405e-06, "loss": 0.8265, "step": 4396 }, { "epoch": 0.34, "learning_rate": 1.5334984636635394e-06, "loss": 1.2019, "step": 4397 }, { "epoch": 0.34, "learning_rate": 1.5332858227933768e-06, "loss": 1.2098, "step": 4398 }, { "epoch": 0.34, "learning_rate": 1.5330731482217911e-06, "loss": 1.2346, "step": 4399 }, { "epoch": 0.34, "learning_rate": 1.532860439962222e-06, "loss": 1.2425, "step": 4400 }, { "epoch": 0.34, "learning_rate": 1.5326476980281124e-06, "loss": 1.1629, "step": 4401 }, { "epoch": 0.34, "learning_rate": 1.5324349224329064e-06, "loss": 1.2317, "step": 4402 }, { "epoch": 0.34, "learning_rate": 1.5322221131900507e-06, "loss": 1.2242, "step": 4403 }, { "epoch": 0.34, "learning_rate": 1.5320092703129931e-06, "loss": 1.2121, "step": 4404 }, { "epoch": 0.34, "learning_rate": 1.5317963938151856e-06, "loss": 1.2173, "step": 4405 }, { "epoch": 0.34, "learning_rate": 1.5315834837100802e-06, "loss": 1.2255, "step": 4406 }, { "epoch": 0.34, "learning_rate": 1.5313705400111325e-06, "loss": 1.2662, "step": 4407 }, { "epoch": 0.34, "learning_rate": 1.5311575627317993e-06, "loss": 1.2348, "step": 4408 }, { "epoch": 0.34, "learning_rate": 1.5309445518855397e-06, "loss": 1.2178, "step": 4409 }, { "epoch": 0.34, "learning_rate": 1.5307315074858158e-06, "loss": 1.246, "step": 4410 }, { "epoch": 0.34, "learning_rate": 1.5305184295460904e-06, "loss": 1.2442, "step": 4411 }, { "epoch": 0.34, "learning_rate": 1.5303053180798293e-06, "loss": 1.2484, "step": 4412 }, { "epoch": 0.34, "learning_rate": 1.5300921731005003e-06, "loss": 1.219, "step": 4413 }, { "epoch": 0.34, "learning_rate": 1.5298789946215735e-06, "loss": 1.1808, "step": 4414 }, { "epoch": 0.34, "learning_rate": 1.5296657826565212e-06, "loss": 1.2169, "step": 4415 }, { "epoch": 0.34, "learning_rate": 1.5294525372188166e-06, "loss": 1.2275, "step": 4416 }, { "epoch": 0.34, "learning_rate": 1.5292392583219358e-06, "loss": 0.8431, "step": 4417 }, { "epoch": 0.34, "learning_rate": 1.5290259459793582e-06, "loss": 1.2143, "step": 4418 }, { "epoch": 0.34, "learning_rate": 1.5288126002045636e-06, "loss": 0.8529, "step": 4419 }, { "epoch": 0.34, "learning_rate": 1.5285992210110346e-06, "loss": 1.2022, "step": 4420 }, { "epoch": 0.34, "learning_rate": 1.5283858084122558e-06, "loss": 0.8294, "step": 4421 }, { "epoch": 0.34, "learning_rate": 1.528172362421714e-06, "loss": 1.1752, "step": 4422 }, { "epoch": 0.34, "learning_rate": 1.5279588830528987e-06, "loss": 1.1968, "step": 4423 }, { "epoch": 0.34, "learning_rate": 1.5277453703193e-06, "loss": 1.2544, "step": 4424 }, { "epoch": 0.34, "learning_rate": 1.5275318242344113e-06, "loss": 1.1255, "step": 4425 }, { "epoch": 0.34, "learning_rate": 1.527318244811728e-06, "loss": 0.8574, "step": 4426 }, { "epoch": 0.34, "learning_rate": 1.5271046320647477e-06, "loss": 1.243, "step": 4427 }, { "epoch": 0.34, "learning_rate": 1.5268909860069688e-06, "loss": 0.8487, "step": 4428 }, { "epoch": 0.34, "learning_rate": 1.526677306651894e-06, "loss": 1.2034, "step": 4429 }, { "epoch": 0.34, "learning_rate": 1.5264635940130259e-06, "loss": 1.2985, "step": 4430 }, { "epoch": 0.34, "learning_rate": 1.5262498481038715e-06, "loss": 1.271, "step": 4431 }, { "epoch": 0.34, "learning_rate": 1.5260360689379374e-06, "loss": 1.2971, "step": 4432 }, { "epoch": 0.34, "learning_rate": 1.5258222565287342e-06, "loss": 1.232, "step": 4433 }, { "epoch": 0.34, "learning_rate": 1.5256084108897735e-06, "loss": 1.2338, "step": 4434 }, { "epoch": 0.34, "learning_rate": 1.52539453203457e-06, "loss": 1.2239, "step": 4435 }, { "epoch": 0.34, "learning_rate": 1.52518061997664e-06, "loss": 1.2372, "step": 4436 }, { "epoch": 0.34, "learning_rate": 1.5249666747295012e-06, "loss": 1.2758, "step": 4437 }, { "epoch": 0.34, "learning_rate": 1.5247526963066743e-06, "loss": 1.2485, "step": 4438 }, { "epoch": 0.34, "learning_rate": 1.5245386847216821e-06, "loss": 1.1977, "step": 4439 }, { "epoch": 0.34, "learning_rate": 1.5243246399880488e-06, "loss": 1.207, "step": 4440 }, { "epoch": 0.34, "learning_rate": 1.5241105621193016e-06, "loss": 1.193, "step": 4441 }, { "epoch": 0.34, "learning_rate": 1.5238964511289689e-06, "loss": 0.851, "step": 4442 }, { "epoch": 0.34, "learning_rate": 1.5236823070305821e-06, "loss": 1.2104, "step": 4443 }, { "epoch": 0.34, "learning_rate": 1.5234681298376737e-06, "loss": 1.2039, "step": 4444 }, { "epoch": 0.34, "learning_rate": 1.5232539195637793e-06, "loss": 1.2497, "step": 4445 }, { "epoch": 0.35, "learning_rate": 1.5230396762224353e-06, "loss": 1.2349, "step": 4446 }, { "epoch": 0.35, "learning_rate": 1.5228253998271816e-06, "loss": 1.228, "step": 4447 }, { "epoch": 0.35, "learning_rate": 1.5226110903915599e-06, "loss": 0.847, "step": 4448 }, { "epoch": 0.35, "learning_rate": 1.522396747929113e-06, "loss": 1.2321, "step": 4449 }, { "epoch": 0.35, "learning_rate": 1.5221823724533865e-06, "loss": 1.2735, "step": 4450 }, { "epoch": 0.35, "learning_rate": 1.5219679639779278e-06, "loss": 1.2569, "step": 4451 }, { "epoch": 0.35, "learning_rate": 1.5217535225162877e-06, "loss": 1.1811, "step": 4452 }, { "epoch": 0.35, "learning_rate": 1.521539048082017e-06, "loss": 1.2538, "step": 4453 }, { "epoch": 0.35, "learning_rate": 1.5213245406886702e-06, "loss": 1.2308, "step": 4454 }, { "epoch": 0.35, "learning_rate": 1.5211100003498024e-06, "loss": 1.1753, "step": 4455 }, { "epoch": 0.35, "learning_rate": 1.5208954270789724e-06, "loss": 1.2654, "step": 4456 }, { "epoch": 0.35, "learning_rate": 1.5206808208897401e-06, "loss": 1.1671, "step": 4457 }, { "epoch": 0.35, "learning_rate": 1.5204661817956677e-06, "loss": 1.2458, "step": 4458 }, { "epoch": 0.35, "learning_rate": 1.5202515098103196e-06, "loss": 1.1646, "step": 4459 }, { "epoch": 0.35, "learning_rate": 1.5200368049472623e-06, "loss": 1.26, "step": 4460 }, { "epoch": 0.35, "learning_rate": 1.5198220672200638e-06, "loss": 1.1621, "step": 4461 }, { "epoch": 0.35, "learning_rate": 1.5196072966422947e-06, "loss": 1.2694, "step": 4462 }, { "epoch": 0.35, "learning_rate": 1.5193924932275279e-06, "loss": 0.8069, "step": 4463 }, { "epoch": 0.35, "learning_rate": 1.519177656989338e-06, "loss": 1.2325, "step": 4464 }, { "epoch": 0.35, "learning_rate": 1.5189627879413016e-06, "loss": 1.2722, "step": 4465 }, { "epoch": 0.35, "learning_rate": 1.5187478860969978e-06, "loss": 1.2161, "step": 4466 }, { "epoch": 0.35, "learning_rate": 1.5185329514700069e-06, "loss": 1.2301, "step": 4467 }, { "epoch": 0.35, "learning_rate": 1.5183179840739123e-06, "loss": 1.2509, "step": 4468 }, { "epoch": 0.35, "learning_rate": 1.518102983922299e-06, "loss": 1.2496, "step": 4469 }, { "epoch": 0.35, "learning_rate": 1.5178879510287544e-06, "loss": 1.226, "step": 4470 }, { "epoch": 0.35, "learning_rate": 1.5176728854068675e-06, "loss": 1.1973, "step": 4471 }, { "epoch": 0.35, "learning_rate": 1.5174577870702287e-06, "loss": 1.2453, "step": 4472 }, { "epoch": 0.35, "learning_rate": 1.5172426560324323e-06, "loss": 1.28, "step": 4473 }, { "epoch": 0.35, "learning_rate": 1.5170274923070736e-06, "loss": 1.2479, "step": 4474 }, { "epoch": 0.35, "learning_rate": 1.51681229590775e-06, "loss": 1.2245, "step": 4475 }, { "epoch": 0.35, "learning_rate": 1.5165970668480604e-06, "loss": 1.2565, "step": 4476 }, { "epoch": 0.35, "learning_rate": 1.5163818051416071e-06, "loss": 1.2328, "step": 4477 }, { "epoch": 0.35, "learning_rate": 1.5161665108019933e-06, "loss": 1.2765, "step": 4478 }, { "epoch": 0.35, "learning_rate": 1.5159511838428251e-06, "loss": 1.2168, "step": 4479 }, { "epoch": 0.35, "learning_rate": 1.5157358242777098e-06, "loss": 1.1993, "step": 4480 }, { "epoch": 0.35, "learning_rate": 1.5155204321202574e-06, "loss": 1.2468, "step": 4481 }, { "epoch": 0.35, "learning_rate": 1.51530500738408e-06, "loss": 1.205, "step": 4482 }, { "epoch": 0.35, "learning_rate": 1.5150895500827914e-06, "loss": 1.1932, "step": 4483 }, { "epoch": 0.35, "learning_rate": 1.5148740602300074e-06, "loss": 0.8436, "step": 4484 }, { "epoch": 0.35, "learning_rate": 1.5146585378393463e-06, "loss": 1.1845, "step": 4485 }, { "epoch": 0.35, "learning_rate": 1.5144429829244277e-06, "loss": 1.2099, "step": 4486 }, { "epoch": 0.35, "learning_rate": 1.5142273954988749e-06, "loss": 1.2687, "step": 4487 }, { "epoch": 0.35, "learning_rate": 1.514011775576311e-06, "loss": 1.3001, "step": 4488 }, { "epoch": 0.35, "learning_rate": 1.5137961231703625e-06, "loss": 1.2422, "step": 4489 }, { "epoch": 0.35, "learning_rate": 1.513580438294658e-06, "loss": 1.2257, "step": 4490 }, { "epoch": 0.35, "learning_rate": 1.5133647209628278e-06, "loss": 1.2414, "step": 4491 }, { "epoch": 0.35, "learning_rate": 1.5131489711885047e-06, "loss": 1.1786, "step": 4492 }, { "epoch": 0.35, "learning_rate": 1.5129331889853224e-06, "loss": 1.1979, "step": 4493 }, { "epoch": 0.35, "learning_rate": 1.5127173743669178e-06, "loss": 1.1822, "step": 4494 }, { "epoch": 0.35, "learning_rate": 1.5125015273469295e-06, "loss": 1.264, "step": 4495 }, { "epoch": 0.35, "learning_rate": 1.512285647938998e-06, "loss": 1.2038, "step": 4496 }, { "epoch": 0.35, "learning_rate": 1.5120697361567663e-06, "loss": 1.1453, "step": 4497 }, { "epoch": 0.35, "learning_rate": 1.5118537920138788e-06, "loss": 1.2612, "step": 4498 }, { "epoch": 0.35, "learning_rate": 1.5116378155239823e-06, "loss": 0.8218, "step": 4499 }, { "epoch": 0.35, "learning_rate": 1.5114218067007257e-06, "loss": 1.2219, "step": 4500 }, { "epoch": 0.35, "learning_rate": 1.51120576555776e-06, "loss": 1.2557, "step": 4501 }, { "epoch": 0.35, "learning_rate": 1.5109896921087377e-06, "loss": 1.2149, "step": 4502 }, { "epoch": 0.35, "learning_rate": 1.510773586367314e-06, "loss": 0.8508, "step": 4503 }, { "epoch": 0.35, "learning_rate": 1.5105574483471459e-06, "loss": 1.2393, "step": 4504 }, { "epoch": 0.35, "learning_rate": 1.5103412780618925e-06, "loss": 1.2419, "step": 4505 }, { "epoch": 0.35, "learning_rate": 1.5101250755252144e-06, "loss": 1.1753, "step": 4506 }, { "epoch": 0.35, "learning_rate": 1.5099088407507749e-06, "loss": 1.1638, "step": 4507 }, { "epoch": 0.35, "learning_rate": 1.5096925737522398e-06, "loss": 1.2632, "step": 4508 }, { "epoch": 0.35, "learning_rate": 1.509476274543275e-06, "loss": 1.2707, "step": 4509 }, { "epoch": 0.35, "learning_rate": 1.5092599431375507e-06, "loss": 1.258, "step": 4510 }, { "epoch": 0.35, "learning_rate": 1.5090435795487377e-06, "loss": 1.2449, "step": 4511 }, { "epoch": 0.35, "learning_rate": 1.5088271837905095e-06, "loss": 1.2648, "step": 4512 }, { "epoch": 0.35, "learning_rate": 1.5086107558765412e-06, "loss": 1.2493, "step": 4513 }, { "epoch": 0.35, "learning_rate": 1.5083942958205102e-06, "loss": 0.8501, "step": 4514 }, { "epoch": 0.35, "learning_rate": 1.508177803636096e-06, "loss": 0.8405, "step": 4515 }, { "epoch": 0.35, "learning_rate": 1.5079612793369798e-06, "loss": 1.2667, "step": 4516 }, { "epoch": 0.35, "learning_rate": 1.5077447229368451e-06, "loss": 1.3105, "step": 4517 }, { "epoch": 0.35, "learning_rate": 1.5075281344493776e-06, "loss": 1.1855, "step": 4518 }, { "epoch": 0.35, "learning_rate": 1.507311513888264e-06, "loss": 1.1966, "step": 4519 }, { "epoch": 0.35, "learning_rate": 1.5070948612671947e-06, "loss": 1.1883, "step": 4520 }, { "epoch": 0.35, "learning_rate": 1.506878176599861e-06, "loss": 1.2055, "step": 4521 }, { "epoch": 0.35, "learning_rate": 1.5066614598999564e-06, "loss": 1.2087, "step": 4522 }, { "epoch": 0.35, "learning_rate": 1.506444711181176e-06, "loss": 1.2316, "step": 4523 }, { "epoch": 0.35, "learning_rate": 1.506227930457218e-06, "loss": 0.8358, "step": 4524 }, { "epoch": 0.35, "learning_rate": 1.5060111177417822e-06, "loss": 1.2446, "step": 4525 }, { "epoch": 0.35, "learning_rate": 1.5057942730485698e-06, "loss": 1.244, "step": 4526 }, { "epoch": 0.35, "learning_rate": 1.5055773963912843e-06, "loss": 1.2486, "step": 4527 }, { "epoch": 0.35, "learning_rate": 1.5053604877836318e-06, "loss": 1.2561, "step": 4528 }, { "epoch": 0.35, "learning_rate": 1.5051435472393202e-06, "loss": 1.2488, "step": 4529 }, { "epoch": 0.35, "learning_rate": 1.504926574772059e-06, "loss": 1.2765, "step": 4530 }, { "epoch": 0.35, "learning_rate": 1.5047095703955593e-06, "loss": 1.2467, "step": 4531 }, { "epoch": 0.35, "learning_rate": 1.504492534123536e-06, "loss": 0.8292, "step": 4532 }, { "epoch": 0.35, "learning_rate": 1.5042754659697038e-06, "loss": 1.2302, "step": 4533 }, { "epoch": 0.35, "learning_rate": 1.5040583659477814e-06, "loss": 1.2462, "step": 4534 }, { "epoch": 0.35, "learning_rate": 1.5038412340714883e-06, "loss": 1.2113, "step": 4535 }, { "epoch": 0.35, "learning_rate": 1.5036240703545461e-06, "loss": 1.1957, "step": 4536 }, { "epoch": 0.35, "learning_rate": 1.503406874810679e-06, "loss": 0.8778, "step": 4537 }, { "epoch": 0.35, "learning_rate": 1.5031896474536123e-06, "loss": 1.2236, "step": 4538 }, { "epoch": 0.35, "learning_rate": 1.5029723882970746e-06, "loss": 1.2407, "step": 4539 }, { "epoch": 0.35, "learning_rate": 1.5027550973547949e-06, "loss": 1.1951, "step": 4540 }, { "epoch": 0.35, "learning_rate": 1.5025377746405058e-06, "loss": 1.2444, "step": 4541 }, { "epoch": 0.35, "learning_rate": 1.502320420167941e-06, "loss": 1.2414, "step": 4542 }, { "epoch": 0.35, "learning_rate": 1.5021030339508363e-06, "loss": 1.2181, "step": 4543 }, { "epoch": 0.35, "learning_rate": 1.5018856160029291e-06, "loss": 1.2322, "step": 4544 }, { "epoch": 0.35, "learning_rate": 1.5016681663379603e-06, "loss": 1.2223, "step": 4545 }, { "epoch": 0.35, "learning_rate": 1.5014506849696712e-06, "loss": 1.1424, "step": 4546 }, { "epoch": 0.35, "learning_rate": 1.501233171911806e-06, "loss": 1.227, "step": 4547 }, { "epoch": 0.35, "learning_rate": 1.50101562717811e-06, "loss": 0.884, "step": 4548 }, { "epoch": 0.35, "learning_rate": 1.500798050782332e-06, "loss": 0.8399, "step": 4549 }, { "epoch": 0.35, "learning_rate": 1.500580442738221e-06, "loss": 1.1956, "step": 4550 }, { "epoch": 0.35, "learning_rate": 1.5003628030595298e-06, "loss": 1.2588, "step": 4551 }, { "epoch": 0.35, "learning_rate": 1.5001451317600115e-06, "loss": 1.2825, "step": 4552 }, { "epoch": 0.35, "learning_rate": 1.4999274288534228e-06, "loss": 1.2368, "step": 4553 }, { "epoch": 0.35, "learning_rate": 1.499709694353521e-06, "loss": 1.2109, "step": 4554 }, { "epoch": 0.35, "learning_rate": 1.4994919282740661e-06, "loss": 1.2661, "step": 4555 }, { "epoch": 0.35, "learning_rate": 1.4992741306288203e-06, "loss": 1.2163, "step": 4556 }, { "epoch": 0.35, "learning_rate": 1.4990563014315471e-06, "loss": 1.2652, "step": 4557 }, { "epoch": 0.35, "learning_rate": 1.4988384406960127e-06, "loss": 1.2159, "step": 4558 }, { "epoch": 0.35, "learning_rate": 1.4986205484359852e-06, "loss": 1.2439, "step": 4559 }, { "epoch": 0.35, "learning_rate": 1.4984026246652337e-06, "loss": 1.2131, "step": 4560 }, { "epoch": 0.35, "learning_rate": 1.4981846693975304e-06, "loss": 1.2384, "step": 4561 }, { "epoch": 0.35, "learning_rate": 1.4979666826466497e-06, "loss": 1.1957, "step": 4562 }, { "epoch": 0.35, "learning_rate": 1.497748664426367e-06, "loss": 1.2715, "step": 4563 }, { "epoch": 0.35, "learning_rate": 1.4975306147504601e-06, "loss": 1.2375, "step": 4564 }, { "epoch": 0.35, "learning_rate": 1.4973125336327085e-06, "loss": 1.2098, "step": 4565 }, { "epoch": 0.35, "learning_rate": 1.4970944210868948e-06, "loss": 1.2083, "step": 4566 }, { "epoch": 0.35, "learning_rate": 1.4968762771268021e-06, "loss": 1.2478, "step": 4567 }, { "epoch": 0.35, "learning_rate": 1.4966581017662168e-06, "loss": 1.2633, "step": 4568 }, { "epoch": 0.35, "learning_rate": 1.4964398950189263e-06, "loss": 1.2265, "step": 4569 }, { "epoch": 0.35, "learning_rate": 1.4962216568987203e-06, "loss": 1.2206, "step": 4570 }, { "epoch": 0.35, "learning_rate": 1.4960033874193906e-06, "loss": 0.8232, "step": 4571 }, { "epoch": 0.35, "learning_rate": 1.495785086594731e-06, "loss": 1.2368, "step": 4572 }, { "epoch": 0.35, "learning_rate": 1.4955667544385371e-06, "loss": 1.1767, "step": 4573 }, { "epoch": 0.36, "learning_rate": 1.4953483909646065e-06, "loss": 1.2153, "step": 4574 }, { "epoch": 0.36, "learning_rate": 1.495129996186739e-06, "loss": 1.2006, "step": 4575 }, { "epoch": 0.36, "learning_rate": 1.4949115701187364e-06, "loss": 1.2536, "step": 4576 }, { "epoch": 0.36, "learning_rate": 1.4946931127744018e-06, "loss": 1.2061, "step": 4577 }, { "epoch": 0.36, "learning_rate": 1.494474624167541e-06, "loss": 1.2182, "step": 4578 }, { "epoch": 0.36, "learning_rate": 1.4942561043119619e-06, "loss": 1.2107, "step": 4579 }, { "epoch": 0.36, "learning_rate": 1.4940375532214739e-06, "loss": 0.8726, "step": 4580 }, { "epoch": 0.36, "learning_rate": 1.4938189709098883e-06, "loss": 0.8149, "step": 4581 }, { "epoch": 0.36, "learning_rate": 1.4936003573910183e-06, "loss": 1.2407, "step": 4582 }, { "epoch": 0.36, "learning_rate": 1.49338171267868e-06, "loss": 1.2273, "step": 4583 }, { "epoch": 0.36, "learning_rate": 1.4931630367866907e-06, "loss": 1.2172, "step": 4584 }, { "epoch": 0.36, "learning_rate": 1.4929443297288698e-06, "loss": 1.2083, "step": 4585 }, { "epoch": 0.36, "learning_rate": 1.4927255915190382e-06, "loss": 1.3112, "step": 4586 }, { "epoch": 0.36, "learning_rate": 1.4925068221710196e-06, "loss": 0.8856, "step": 4587 }, { "epoch": 0.36, "learning_rate": 1.4922880216986393e-06, "loss": 1.2552, "step": 4588 }, { "epoch": 0.36, "learning_rate": 1.4920691901157245e-06, "loss": 1.2415, "step": 4589 }, { "epoch": 0.36, "learning_rate": 1.4918503274361044e-06, "loss": 1.2073, "step": 4590 }, { "epoch": 0.36, "learning_rate": 1.4916314336736106e-06, "loss": 1.2507, "step": 4591 }, { "epoch": 0.36, "learning_rate": 1.4914125088420758e-06, "loss": 1.1672, "step": 4592 }, { "epoch": 0.36, "learning_rate": 1.4911935529553351e-06, "loss": 1.2347, "step": 4593 }, { "epoch": 0.36, "learning_rate": 1.490974566027226e-06, "loss": 0.8205, "step": 4594 }, { "epoch": 0.36, "learning_rate": 1.4907555480715869e-06, "loss": 1.2593, "step": 4595 }, { "epoch": 0.36, "learning_rate": 1.4905364991022599e-06, "loss": 0.8109, "step": 4596 }, { "epoch": 0.36, "learning_rate": 1.4903174191330868e-06, "loss": 1.2026, "step": 4597 }, { "epoch": 0.36, "learning_rate": 1.490098308177913e-06, "loss": 0.8246, "step": 4598 }, { "epoch": 0.36, "learning_rate": 1.4898791662505853e-06, "loss": 1.1534, "step": 4599 }, { "epoch": 0.36, "learning_rate": 1.4896599933649532e-06, "loss": 1.2023, "step": 4600 }, { "epoch": 0.36, "learning_rate": 1.489440789534867e-06, "loss": 0.8648, "step": 4601 }, { "epoch": 0.36, "learning_rate": 1.4892215547741791e-06, "loss": 0.8291, "step": 4602 }, { "epoch": 0.36, "learning_rate": 1.4890022890967445e-06, "loss": 1.262, "step": 4603 }, { "epoch": 0.36, "learning_rate": 1.48878299251642e-06, "loss": 1.1826, "step": 4604 }, { "epoch": 0.36, "learning_rate": 1.4885636650470644e-06, "loss": 1.2361, "step": 4605 }, { "epoch": 0.36, "learning_rate": 1.4883443067025376e-06, "loss": 1.214, "step": 4606 }, { "epoch": 0.36, "learning_rate": 1.488124917496703e-06, "loss": 1.2628, "step": 4607 }, { "epoch": 0.36, "learning_rate": 1.4879054974434244e-06, "loss": 1.2363, "step": 4608 }, { "epoch": 0.36, "learning_rate": 1.4876860465565686e-06, "loss": 1.245, "step": 4609 }, { "epoch": 0.36, "learning_rate": 1.4874665648500035e-06, "loss": 1.2441, "step": 4610 }, { "epoch": 0.36, "learning_rate": 1.4872470523376002e-06, "loss": 1.1831, "step": 4611 }, { "epoch": 0.36, "learning_rate": 1.48702750903323e-06, "loss": 1.2554, "step": 4612 }, { "epoch": 0.36, "learning_rate": 1.486807934950768e-06, "loss": 1.2183, "step": 4613 }, { "epoch": 0.36, "learning_rate": 1.48658833010409e-06, "loss": 1.22, "step": 4614 }, { "epoch": 0.36, "learning_rate": 1.4863686945070742e-06, "loss": 1.2195, "step": 4615 }, { "epoch": 0.36, "learning_rate": 1.4861490281736e-06, "loss": 1.1887, "step": 4616 }, { "epoch": 0.36, "learning_rate": 1.4859293311175502e-06, "loss": 1.262, "step": 4617 }, { "epoch": 0.36, "learning_rate": 1.485709603352809e-06, "loss": 0.8598, "step": 4618 }, { "epoch": 0.36, "learning_rate": 1.4854898448932613e-06, "loss": 1.1759, "step": 4619 }, { "epoch": 0.36, "learning_rate": 1.4852700557527951e-06, "loss": 1.1822, "step": 4620 }, { "epoch": 0.36, "learning_rate": 1.4850502359453006e-06, "loss": 1.2127, "step": 4621 }, { "epoch": 0.36, "learning_rate": 1.4848303854846692e-06, "loss": 1.2298, "step": 4622 }, { "epoch": 0.36, "learning_rate": 1.4846105043847947e-06, "loss": 1.2655, "step": 4623 }, { "epoch": 0.36, "learning_rate": 1.4843905926595728e-06, "loss": 0.8629, "step": 4624 }, { "epoch": 0.36, "learning_rate": 1.4841706503229006e-06, "loss": 1.1826, "step": 4625 }, { "epoch": 0.36, "learning_rate": 1.4839506773886776e-06, "loss": 1.205, "step": 4626 }, { "epoch": 0.36, "learning_rate": 1.4837306738708052e-06, "loss": 1.2258, "step": 4627 }, { "epoch": 0.36, "learning_rate": 1.483510639783187e-06, "loss": 1.2246, "step": 4628 }, { "epoch": 0.36, "learning_rate": 1.4832905751397277e-06, "loss": 0.8595, "step": 4629 }, { "epoch": 0.36, "learning_rate": 1.483070479954335e-06, "loss": 1.2136, "step": 4630 }, { "epoch": 0.36, "learning_rate": 1.4828503542409177e-06, "loss": 1.2435, "step": 4631 }, { "epoch": 0.36, "learning_rate": 1.4826301980133868e-06, "loss": 1.2297, "step": 4632 }, { "epoch": 0.36, "learning_rate": 1.4824100112856553e-06, "loss": 1.2271, "step": 4633 }, { "epoch": 0.36, "learning_rate": 1.4821897940716378e-06, "loss": 1.2472, "step": 4634 }, { "epoch": 0.36, "learning_rate": 1.4819695463852522e-06, "loss": 1.178, "step": 4635 }, { "epoch": 0.36, "learning_rate": 1.4817492682404158e-06, "loss": 1.2457, "step": 4636 }, { "epoch": 0.36, "learning_rate": 1.4815289596510499e-06, "loss": 1.2129, "step": 4637 }, { "epoch": 0.36, "learning_rate": 1.4813086206310773e-06, "loss": 1.2448, "step": 4638 }, { "epoch": 0.36, "learning_rate": 1.4810882511944217e-06, "loss": 1.1921, "step": 4639 }, { "epoch": 0.36, "learning_rate": 1.4808678513550108e-06, "loss": 0.8283, "step": 4640 }, { "epoch": 0.36, "learning_rate": 1.4806474211267718e-06, "loss": 1.2002, "step": 4641 }, { "epoch": 0.36, "learning_rate": 1.4804269605236356e-06, "loss": 1.2384, "step": 4642 }, { "epoch": 0.36, "learning_rate": 1.480206469559534e-06, "loss": 1.2701, "step": 4643 }, { "epoch": 0.36, "learning_rate": 1.4799859482484012e-06, "loss": 1.2291, "step": 4644 }, { "epoch": 0.36, "learning_rate": 1.4797653966041734e-06, "loss": 1.1998, "step": 4645 }, { "epoch": 0.36, "learning_rate": 1.4795448146407886e-06, "loss": 1.2519, "step": 4646 }, { "epoch": 0.36, "learning_rate": 1.4793242023721862e-06, "loss": 1.2354, "step": 4647 }, { "epoch": 0.36, "learning_rate": 1.4791035598123082e-06, "loss": 1.1568, "step": 4648 }, { "epoch": 0.36, "learning_rate": 1.4788828869750987e-06, "loss": 1.2099, "step": 4649 }, { "epoch": 0.36, "learning_rate": 1.4786621838745025e-06, "loss": 1.1863, "step": 4650 }, { "epoch": 0.36, "learning_rate": 1.4784414505244679e-06, "loss": 1.2273, "step": 4651 }, { "epoch": 0.36, "learning_rate": 1.4782206869389438e-06, "loss": 1.2181, "step": 4652 }, { "epoch": 0.36, "learning_rate": 1.4779998931318816e-06, "loss": 1.2348, "step": 4653 }, { "epoch": 0.36, "learning_rate": 1.4777790691172345e-06, "loss": 0.8556, "step": 4654 }, { "epoch": 0.36, "learning_rate": 1.477558214908958e-06, "loss": 1.2175, "step": 4655 }, { "epoch": 0.36, "learning_rate": 1.4773373305210088e-06, "loss": 1.1899, "step": 4656 }, { "epoch": 0.36, "learning_rate": 1.4771164159673463e-06, "loss": 1.2245, "step": 4657 }, { "epoch": 0.36, "learning_rate": 1.4768954712619305e-06, "loss": 1.1839, "step": 4658 }, { "epoch": 0.36, "learning_rate": 1.4766744964187252e-06, "loss": 1.2321, "step": 4659 }, { "epoch": 0.36, "learning_rate": 1.4764534914516946e-06, "loss": 1.228, "step": 4660 }, { "epoch": 0.36, "learning_rate": 1.4762324563748052e-06, "loss": 1.2591, "step": 4661 }, { "epoch": 0.36, "learning_rate": 1.4760113912020257e-06, "loss": 1.1581, "step": 4662 }, { "epoch": 0.36, "learning_rate": 1.4757902959473265e-06, "loss": 0.8943, "step": 4663 }, { "epoch": 0.36, "learning_rate": 1.4755691706246795e-06, "loss": 1.2293, "step": 4664 }, { "epoch": 0.36, "learning_rate": 1.4753480152480592e-06, "loss": 1.2454, "step": 4665 }, { "epoch": 0.36, "learning_rate": 1.475126829831442e-06, "loss": 1.2205, "step": 4666 }, { "epoch": 0.36, "learning_rate": 1.4749056143888052e-06, "loss": 0.819, "step": 4667 }, { "epoch": 0.36, "learning_rate": 1.4746843689341292e-06, "loss": 1.2238, "step": 4668 }, { "epoch": 0.36, "learning_rate": 1.4744630934813956e-06, "loss": 1.2531, "step": 4669 }, { "epoch": 0.36, "learning_rate": 1.474241788044588e-06, "loss": 1.2064, "step": 4670 }, { "epoch": 0.36, "learning_rate": 1.474020452637692e-06, "loss": 1.2806, "step": 4671 }, { "epoch": 0.36, "learning_rate": 1.4737990872746953e-06, "loss": 1.2414, "step": 4672 }, { "epoch": 0.36, "learning_rate": 1.4735776919695872e-06, "loss": 1.2655, "step": 4673 }, { "epoch": 0.36, "learning_rate": 1.473356266736359e-06, "loss": 1.2262, "step": 4674 }, { "epoch": 0.36, "learning_rate": 1.473134811589003e-06, "loss": 1.2276, "step": 4675 }, { "epoch": 0.36, "learning_rate": 1.4729133265415155e-06, "loss": 1.2693, "step": 4676 }, { "epoch": 0.36, "learning_rate": 1.4726918116078928e-06, "loss": 1.2131, "step": 4677 }, { "epoch": 0.36, "learning_rate": 1.4724702668021337e-06, "loss": 1.1894, "step": 4678 }, { "epoch": 0.36, "learning_rate": 1.472248692138239e-06, "loss": 1.1739, "step": 4679 }, { "epoch": 0.36, "learning_rate": 1.4720270876302113e-06, "loss": 1.2686, "step": 4680 }, { "epoch": 0.36, "learning_rate": 1.4718054532920549e-06, "loss": 1.2207, "step": 4681 }, { "epoch": 0.36, "learning_rate": 1.4715837891377765e-06, "loss": 1.2483, "step": 4682 }, { "epoch": 0.36, "learning_rate": 1.471362095181384e-06, "loss": 1.2083, "step": 4683 }, { "epoch": 0.36, "learning_rate": 1.4711403714368877e-06, "loss": 1.2403, "step": 4684 }, { "epoch": 0.36, "learning_rate": 1.4709186179182998e-06, "loss": 0.826, "step": 4685 }, { "epoch": 0.36, "learning_rate": 1.4706968346396337e-06, "loss": 1.2376, "step": 4686 }, { "epoch": 0.36, "learning_rate": 1.4704750216149053e-06, "loss": 0.8501, "step": 4687 }, { "epoch": 0.36, "learning_rate": 1.4702531788581328e-06, "loss": 1.2294, "step": 4688 }, { "epoch": 0.36, "learning_rate": 1.4700313063833347e-06, "loss": 1.2426, "step": 4689 }, { "epoch": 0.36, "learning_rate": 1.4698094042045334e-06, "loss": 1.2172, "step": 4690 }, { "epoch": 0.36, "learning_rate": 1.469587472335752e-06, "loss": 1.1831, "step": 4691 }, { "epoch": 0.36, "learning_rate": 1.469365510791015e-06, "loss": 1.2236, "step": 4692 }, { "epoch": 0.36, "learning_rate": 1.4691435195843503e-06, "loss": 1.2233, "step": 4693 }, { "epoch": 0.36, "learning_rate": 1.4689214987297863e-06, "loss": 0.808, "step": 4694 }, { "epoch": 0.36, "learning_rate": 1.4686994482413541e-06, "loss": 1.2293, "step": 4695 }, { "epoch": 0.36, "learning_rate": 1.4684773681330857e-06, "loss": 1.2493, "step": 4696 }, { "epoch": 0.36, "learning_rate": 1.4682552584190165e-06, "loss": 1.2781, "step": 4697 }, { "epoch": 0.36, "learning_rate": 1.4680331191131826e-06, "loss": 1.2637, "step": 4698 }, { "epoch": 0.36, "learning_rate": 1.4678109502296216e-06, "loss": 1.219, "step": 4699 }, { "epoch": 0.36, "learning_rate": 1.4675887517823746e-06, "loss": 1.2204, "step": 4700 }, { "epoch": 0.36, "learning_rate": 1.467366523785483e-06, "loss": 1.2878, "step": 4701 }, { "epoch": 0.36, "learning_rate": 1.4671442662529913e-06, "loss": 1.2245, "step": 4702 }, { "epoch": 0.37, "learning_rate": 1.4669219791989444e-06, "loss": 1.2331, "step": 4703 }, { "epoch": 0.37, "learning_rate": 1.4666996626373907e-06, "loss": 1.1697, "step": 4704 }, { "epoch": 0.37, "learning_rate": 1.4664773165823791e-06, "loss": 1.2459, "step": 4705 }, { "epoch": 0.37, "learning_rate": 1.466254941047961e-06, "loss": 1.2452, "step": 4706 }, { "epoch": 0.37, "learning_rate": 1.46603253604819e-06, "loss": 0.8347, "step": 4707 }, { "epoch": 0.37, "learning_rate": 1.4658101015971207e-06, "loss": 1.1949, "step": 4708 }, { "epoch": 0.37, "learning_rate": 1.4655876377088102e-06, "loss": 1.2358, "step": 4709 }, { "epoch": 0.37, "learning_rate": 1.4653651443973177e-06, "loss": 1.234, "step": 4710 }, { "epoch": 0.37, "learning_rate": 1.4651426216767032e-06, "loss": 1.1757, "step": 4711 }, { "epoch": 0.37, "learning_rate": 1.4649200695610295e-06, "loss": 1.2086, "step": 4712 }, { "epoch": 0.37, "learning_rate": 1.4646974880643609e-06, "loss": 1.1629, "step": 4713 }, { "epoch": 0.37, "learning_rate": 1.4644748772007635e-06, "loss": 1.2144, "step": 4714 }, { "epoch": 0.37, "learning_rate": 1.4642522369843055e-06, "loss": 1.1787, "step": 4715 }, { "epoch": 0.37, "learning_rate": 1.464029567429057e-06, "loss": 1.2247, "step": 4716 }, { "epoch": 0.37, "learning_rate": 1.4638068685490897e-06, "loss": 1.2013, "step": 4717 }, { "epoch": 0.37, "learning_rate": 1.4635841403584767e-06, "loss": 1.2722, "step": 4718 }, { "epoch": 0.37, "learning_rate": 1.4633613828712944e-06, "loss": 0.9148, "step": 4719 }, { "epoch": 0.37, "learning_rate": 1.4631385961016194e-06, "loss": 1.1909, "step": 4720 }, { "epoch": 0.37, "learning_rate": 1.4629157800635311e-06, "loss": 1.1909, "step": 4721 }, { "epoch": 0.37, "learning_rate": 1.462692934771111e-06, "loss": 1.2499, "step": 4722 }, { "epoch": 0.37, "learning_rate": 1.4624700602384415e-06, "loss": 1.2746, "step": 4723 }, { "epoch": 0.37, "learning_rate": 1.4622471564796071e-06, "loss": 1.212, "step": 4724 }, { "epoch": 0.37, "learning_rate": 1.4620242235086948e-06, "loss": 1.1702, "step": 4725 }, { "epoch": 0.37, "learning_rate": 1.4618012613397927e-06, "loss": 1.2306, "step": 4726 }, { "epoch": 0.37, "learning_rate": 1.4615782699869915e-06, "loss": 1.1425, "step": 4727 }, { "epoch": 0.37, "learning_rate": 1.4613552494643834e-06, "loss": 1.2515, "step": 4728 }, { "epoch": 0.37, "learning_rate": 1.4611321997860617e-06, "loss": 1.1941, "step": 4729 }, { "epoch": 0.37, "learning_rate": 1.4609091209661225e-06, "loss": 1.2286, "step": 4730 }, { "epoch": 0.37, "learning_rate": 1.4606860130186638e-06, "loss": 1.2304, "step": 4731 }, { "epoch": 0.37, "learning_rate": 1.4604628759577846e-06, "loss": 1.2273, "step": 4732 }, { "epoch": 0.37, "learning_rate": 1.4602397097975865e-06, "loss": 1.2132, "step": 4733 }, { "epoch": 0.37, "learning_rate": 1.4600165145521725e-06, "loss": 1.2248, "step": 4734 }, { "epoch": 0.37, "learning_rate": 1.4597932902356477e-06, "loss": 1.2034, "step": 4735 }, { "epoch": 0.37, "learning_rate": 1.4595700368621191e-06, "loss": 1.2253, "step": 4736 }, { "epoch": 0.37, "learning_rate": 1.4593467544456951e-06, "loss": 1.1787, "step": 4737 }, { "epoch": 0.37, "learning_rate": 1.4591234430004864e-06, "loss": 1.2884, "step": 4738 }, { "epoch": 0.37, "learning_rate": 1.458900102540605e-06, "loss": 1.2991, "step": 4739 }, { "epoch": 0.37, "learning_rate": 1.4586767330801654e-06, "loss": 1.2001, "step": 4740 }, { "epoch": 0.37, "learning_rate": 1.458453334633284e-06, "loss": 1.2238, "step": 4741 }, { "epoch": 0.37, "learning_rate": 1.4582299072140779e-06, "loss": 1.2392, "step": 4742 }, { "epoch": 0.37, "learning_rate": 1.4580064508366667e-06, "loss": 1.2466, "step": 4743 }, { "epoch": 0.37, "learning_rate": 1.4577829655151726e-06, "loss": 1.2407, "step": 4744 }, { "epoch": 0.37, "learning_rate": 1.4575594512637186e-06, "loss": 1.2746, "step": 4745 }, { "epoch": 0.37, "learning_rate": 1.45733590809643e-06, "loss": 1.1625, "step": 4746 }, { "epoch": 0.37, "learning_rate": 1.4571123360274333e-06, "loss": 1.2356, "step": 4747 }, { "epoch": 0.37, "learning_rate": 1.4568887350708578e-06, "loss": 1.1758, "step": 4748 }, { "epoch": 0.37, "learning_rate": 1.456665105240834e-06, "loss": 1.1927, "step": 4749 }, { "epoch": 0.37, "learning_rate": 1.4564414465514944e-06, "loss": 1.2151, "step": 4750 }, { "epoch": 0.37, "learning_rate": 1.4562177590169736e-06, "loss": 1.2199, "step": 4751 }, { "epoch": 0.37, "learning_rate": 1.455994042651407e-06, "loss": 1.2017, "step": 4752 }, { "epoch": 0.37, "learning_rate": 1.455770297468933e-06, "loss": 1.2046, "step": 4753 }, { "epoch": 0.37, "learning_rate": 1.4555465234836911e-06, "loss": 1.2458, "step": 4754 }, { "epoch": 0.37, "learning_rate": 1.4553227207098232e-06, "loss": 0.8522, "step": 4755 }, { "epoch": 0.37, "learning_rate": 1.4550988891614726e-06, "loss": 1.2694, "step": 4756 }, { "epoch": 0.37, "learning_rate": 1.4548750288527844e-06, "loss": 0.8551, "step": 4757 }, { "epoch": 0.37, "learning_rate": 1.4546511397979058e-06, "loss": 1.1882, "step": 4758 }, { "epoch": 0.37, "learning_rate": 1.4544272220109855e-06, "loss": 1.2028, "step": 4759 }, { "epoch": 0.37, "learning_rate": 1.4542032755061739e-06, "loss": 1.1659, "step": 4760 }, { "epoch": 0.37, "learning_rate": 1.453979300297624e-06, "loss": 1.2406, "step": 4761 }, { "epoch": 0.37, "learning_rate": 1.45375529639949e-06, "loss": 1.2587, "step": 4762 }, { "epoch": 0.37, "learning_rate": 1.4535312638259276e-06, "loss": 0.8585, "step": 4763 }, { "epoch": 0.37, "learning_rate": 1.4533072025910952e-06, "loss": 1.2281, "step": 4764 }, { "epoch": 0.37, "learning_rate": 1.453083112709152e-06, "loss": 1.1329, "step": 4765 }, { "epoch": 0.37, "learning_rate": 1.45285899419426e-06, "loss": 1.1964, "step": 4766 }, { "epoch": 0.37, "learning_rate": 1.4526348470605825e-06, "loss": 1.1931, "step": 4767 }, { "epoch": 0.37, "learning_rate": 1.4524106713222843e-06, "loss": 1.1846, "step": 4768 }, { "epoch": 0.37, "learning_rate": 1.4521864669935329e-06, "loss": 1.2068, "step": 4769 }, { "epoch": 0.37, "learning_rate": 1.4519622340884968e-06, "loss": 1.1902, "step": 4770 }, { "epoch": 0.37, "learning_rate": 1.4517379726213464e-06, "loss": 1.1873, "step": 4771 }, { "epoch": 0.37, "learning_rate": 1.4515136826062544e-06, "loss": 0.8675, "step": 4772 }, { "epoch": 0.37, "learning_rate": 1.4512893640573947e-06, "loss": 1.1732, "step": 4773 }, { "epoch": 0.37, "learning_rate": 1.4510650169889433e-06, "loss": 1.193, "step": 4774 }, { "epoch": 0.37, "learning_rate": 1.4508406414150784e-06, "loss": 1.1991, "step": 4775 }, { "epoch": 0.37, "learning_rate": 1.4506162373499793e-06, "loss": 1.202, "step": 4776 }, { "epoch": 0.37, "learning_rate": 1.4503918048078274e-06, "loss": 1.1709, "step": 4777 }, { "epoch": 0.37, "learning_rate": 1.4501673438028059e-06, "loss": 1.187, "step": 4778 }, { "epoch": 0.37, "learning_rate": 1.4499428543490997e-06, "loss": 0.8619, "step": 4779 }, { "epoch": 0.37, "learning_rate": 1.4497183364608957e-06, "loss": 1.1889, "step": 4780 }, { "epoch": 0.37, "learning_rate": 1.4494937901523825e-06, "loss": 1.1728, "step": 4781 }, { "epoch": 0.37, "learning_rate": 1.4492692154377503e-06, "loss": 1.2236, "step": 4782 }, { "epoch": 0.37, "learning_rate": 1.4490446123311916e-06, "loss": 1.2437, "step": 4783 }, { "epoch": 0.37, "learning_rate": 1.4488199808469004e-06, "loss": 0.8162, "step": 4784 }, { "epoch": 0.37, "learning_rate": 1.448595320999072e-06, "loss": 0.8233, "step": 4785 }, { "epoch": 0.37, "learning_rate": 1.4483706328019043e-06, "loss": 1.2394, "step": 4786 }, { "epoch": 0.37, "learning_rate": 1.4481459162695967e-06, "loss": 1.187, "step": 4787 }, { "epoch": 0.37, "learning_rate": 1.44792117141635e-06, "loss": 1.2496, "step": 4788 }, { "epoch": 0.37, "learning_rate": 1.4476963982563674e-06, "loss": 1.226, "step": 4789 }, { "epoch": 0.37, "learning_rate": 1.4474715968038535e-06, "loss": 1.206, "step": 4790 }, { "epoch": 0.37, "learning_rate": 1.447246767073015e-06, "loss": 1.2636, "step": 4791 }, { "epoch": 0.37, "learning_rate": 1.4470219090780603e-06, "loss": 1.2273, "step": 4792 }, { "epoch": 0.37, "learning_rate": 1.4467970228331988e-06, "loss": 1.2502, "step": 4793 }, { "epoch": 0.37, "learning_rate": 1.4465721083526432e-06, "loss": 1.2328, "step": 4794 }, { "epoch": 0.37, "learning_rate": 1.4463471656506068e-06, "loss": 1.2218, "step": 4795 }, { "epoch": 0.37, "learning_rate": 1.4461221947413047e-06, "loss": 1.2287, "step": 4796 }, { "epoch": 0.37, "learning_rate": 1.4458971956389545e-06, "loss": 1.1822, "step": 4797 }, { "epoch": 0.37, "learning_rate": 1.4456721683577753e-06, "loss": 1.2535, "step": 4798 }, { "epoch": 0.37, "learning_rate": 1.4454471129119872e-06, "loss": 0.8703, "step": 4799 }, { "epoch": 0.37, "learning_rate": 1.445222029315814e-06, "loss": 1.2173, "step": 4800 }, { "epoch": 0.37, "learning_rate": 1.4449969175834787e-06, "loss": 1.1991, "step": 4801 }, { "epoch": 0.37, "learning_rate": 1.4447717777292083e-06, "loss": 0.8705, "step": 4802 }, { "epoch": 0.37, "learning_rate": 1.44454660976723e-06, "loss": 1.1939, "step": 4803 }, { "epoch": 0.37, "learning_rate": 1.4443214137117742e-06, "loss": 1.2492, "step": 4804 }, { "epoch": 0.37, "learning_rate": 1.444096189577072e-06, "loss": 1.2247, "step": 4805 }, { "epoch": 0.37, "learning_rate": 1.4438709373773565e-06, "loss": 1.2228, "step": 4806 }, { "epoch": 0.37, "learning_rate": 1.443645657126863e-06, "loss": 1.2489, "step": 4807 }, { "epoch": 0.37, "learning_rate": 1.4434203488398282e-06, "loss": 0.8462, "step": 4808 }, { "epoch": 0.37, "learning_rate": 1.4431950125304903e-06, "loss": 1.2535, "step": 4809 }, { "epoch": 0.37, "learning_rate": 1.44296964821309e-06, "loss": 1.2334, "step": 4810 }, { "epoch": 0.37, "learning_rate": 1.442744255901869e-06, "loss": 1.2102, "step": 4811 }, { "epoch": 0.37, "learning_rate": 1.4425188356110716e-06, "loss": 1.1932, "step": 4812 }, { "epoch": 0.37, "learning_rate": 1.442293387354943e-06, "loss": 1.2638, "step": 4813 }, { "epoch": 0.37, "learning_rate": 1.4420679111477307e-06, "loss": 1.2059, "step": 4814 }, { "epoch": 0.37, "learning_rate": 1.441842407003684e-06, "loss": 1.1867, "step": 4815 }, { "epoch": 0.37, "learning_rate": 1.4416168749370536e-06, "loss": 1.2448, "step": 4816 }, { "epoch": 0.37, "learning_rate": 1.4413913149620928e-06, "loss": 0.8272, "step": 4817 }, { "epoch": 0.37, "learning_rate": 1.441165727093055e-06, "loss": 0.8244, "step": 4818 }, { "epoch": 0.37, "learning_rate": 1.4409401113441973e-06, "loss": 1.2574, "step": 4819 }, { "epoch": 0.37, "learning_rate": 1.4407144677297768e-06, "loss": 1.2237, "step": 4820 }, { "epoch": 0.37, "learning_rate": 1.4404887962640539e-06, "loss": 1.247, "step": 4821 }, { "epoch": 0.37, "learning_rate": 1.4402630969612905e-06, "loss": 0.8758, "step": 4822 }, { "epoch": 0.37, "learning_rate": 1.4400373698357487e-06, "loss": 1.238, "step": 4823 }, { "epoch": 0.37, "learning_rate": 1.4398116149016943e-06, "loss": 1.2267, "step": 4824 }, { "epoch": 0.37, "learning_rate": 1.4395858321733938e-06, "loss": 0.8455, "step": 4825 }, { "epoch": 0.37, "learning_rate": 1.4393600216651157e-06, "loss": 1.2619, "step": 4826 }, { "epoch": 0.37, "learning_rate": 1.4391341833911305e-06, "loss": 1.1863, "step": 4827 }, { "epoch": 0.37, "learning_rate": 1.4389083173657102e-06, "loss": 1.2566, "step": 4828 }, { "epoch": 0.37, "learning_rate": 1.438682423603128e-06, "loss": 1.2109, "step": 4829 }, { "epoch": 0.37, "learning_rate": 1.4384565021176603e-06, "loss": 1.2055, "step": 4830 }, { "epoch": 0.37, "learning_rate": 1.4382305529235839e-06, "loss": 1.2641, "step": 4831 }, { "epoch": 0.38, "learning_rate": 1.4380045760351778e-06, "loss": 1.18, "step": 4832 }, { "epoch": 0.38, "learning_rate": 1.4377785714667229e-06, "loss": 1.2071, "step": 4833 }, { "epoch": 0.38, "learning_rate": 1.4375525392325018e-06, "loss": 1.244, "step": 4834 }, { "epoch": 0.38, "learning_rate": 1.437326479346799e-06, "loss": 1.231, "step": 4835 }, { "epoch": 0.38, "learning_rate": 1.4371003918239e-06, "loss": 1.2653, "step": 4836 }, { "epoch": 0.38, "learning_rate": 1.4368742766780927e-06, "loss": 1.171, "step": 4837 }, { "epoch": 0.38, "learning_rate": 1.436648133923667e-06, "loss": 0.8063, "step": 4838 }, { "epoch": 0.38, "learning_rate": 1.436421963574914e-06, "loss": 1.2554, "step": 4839 }, { "epoch": 0.38, "learning_rate": 1.4361957656461266e-06, "loss": 0.8664, "step": 4840 }, { "epoch": 0.38, "learning_rate": 1.4359695401515993e-06, "loss": 1.2232, "step": 4841 }, { "epoch": 0.38, "learning_rate": 1.4357432871056293e-06, "loss": 1.2204, "step": 4842 }, { "epoch": 0.38, "learning_rate": 1.4355170065225146e-06, "loss": 1.2045, "step": 4843 }, { "epoch": 0.38, "learning_rate": 1.435290698416555e-06, "loss": 1.1713, "step": 4844 }, { "epoch": 0.38, "learning_rate": 1.435064362802052e-06, "loss": 1.2303, "step": 4845 }, { "epoch": 0.38, "learning_rate": 1.4348379996933095e-06, "loss": 1.2369, "step": 4846 }, { "epoch": 0.38, "learning_rate": 1.4346116091046326e-06, "loss": 0.8747, "step": 4847 }, { "epoch": 0.38, "learning_rate": 1.434385191050328e-06, "loss": 1.2239, "step": 4848 }, { "epoch": 0.38, "learning_rate": 1.4341587455447046e-06, "loss": 1.247, "step": 4849 }, { "epoch": 0.38, "learning_rate": 1.4339322726020728e-06, "loss": 1.2481, "step": 4850 }, { "epoch": 0.38, "learning_rate": 1.4337057722367448e-06, "loss": 1.2652, "step": 4851 }, { "epoch": 0.38, "learning_rate": 1.4334792444630342e-06, "loss": 1.2096, "step": 4852 }, { "epoch": 0.38, "learning_rate": 1.4332526892952565e-06, "loss": 1.2552, "step": 4853 }, { "epoch": 0.38, "learning_rate": 1.4330261067477295e-06, "loss": 1.2676, "step": 4854 }, { "epoch": 0.38, "learning_rate": 1.4327994968347723e-06, "loss": 1.2117, "step": 4855 }, { "epoch": 0.38, "learning_rate": 1.4325728595707051e-06, "loss": 1.2244, "step": 4856 }, { "epoch": 0.38, "learning_rate": 1.432346194969851e-06, "loss": 1.1726, "step": 4857 }, { "epoch": 0.38, "learning_rate": 1.4321195030465338e-06, "loss": 1.2504, "step": 4858 }, { "epoch": 0.38, "learning_rate": 1.4318927838150799e-06, "loss": 1.1668, "step": 4859 }, { "epoch": 0.38, "learning_rate": 1.4316660372898167e-06, "loss": 1.1993, "step": 4860 }, { "epoch": 0.38, "learning_rate": 1.4314392634850735e-06, "loss": 1.21, "step": 4861 }, { "epoch": 0.38, "learning_rate": 1.431212462415182e-06, "loss": 1.1949, "step": 4862 }, { "epoch": 0.38, "learning_rate": 1.4309856340944748e-06, "loss": 1.2532, "step": 4863 }, { "epoch": 0.38, "learning_rate": 1.4307587785372862e-06, "loss": 1.2022, "step": 4864 }, { "epoch": 0.38, "learning_rate": 1.430531895757953e-06, "loss": 1.2325, "step": 4865 }, { "epoch": 0.38, "learning_rate": 1.430304985770813e-06, "loss": 1.1657, "step": 4866 }, { "epoch": 0.38, "learning_rate": 1.4300780485902059e-06, "loss": 1.2216, "step": 4867 }, { "epoch": 0.38, "learning_rate": 1.4298510842304729e-06, "loss": 1.1364, "step": 4868 }, { "epoch": 0.38, "learning_rate": 1.429624092705958e-06, "loss": 1.2258, "step": 4869 }, { "epoch": 0.38, "learning_rate": 1.4293970740310054e-06, "loss": 1.2232, "step": 4870 }, { "epoch": 0.38, "learning_rate": 1.4291700282199619e-06, "loss": 1.2201, "step": 4871 }, { "epoch": 0.38, "learning_rate": 1.4289429552871764e-06, "loss": 1.1781, "step": 4872 }, { "epoch": 0.38, "learning_rate": 1.428715855246998e-06, "loss": 1.2331, "step": 4873 }, { "epoch": 0.38, "learning_rate": 1.4284887281137791e-06, "loss": 1.2655, "step": 4874 }, { "epoch": 0.38, "learning_rate": 1.4282615739018728e-06, "loss": 1.2257, "step": 4875 }, { "epoch": 0.38, "learning_rate": 1.4280343926256348e-06, "loss": 1.2367, "step": 4876 }, { "epoch": 0.38, "learning_rate": 1.4278071842994216e-06, "loss": 1.2616, "step": 4877 }, { "epoch": 0.38, "learning_rate": 1.4275799489375922e-06, "loss": 1.1945, "step": 4878 }, { "epoch": 0.38, "learning_rate": 1.427352686554506e-06, "loss": 1.2016, "step": 4879 }, { "epoch": 0.38, "learning_rate": 1.4271253971645262e-06, "loss": 0.827, "step": 4880 }, { "epoch": 0.38, "learning_rate": 1.4268980807820156e-06, "loss": 1.242, "step": 4881 }, { "epoch": 0.38, "learning_rate": 1.4266707374213405e-06, "loss": 1.2561, "step": 4882 }, { "epoch": 0.38, "learning_rate": 1.4264433670968674e-06, "loss": 1.2622, "step": 4883 }, { "epoch": 0.38, "learning_rate": 1.4262159698229654e-06, "loss": 1.2475, "step": 4884 }, { "epoch": 0.38, "learning_rate": 1.425988545614005e-06, "loss": 0.8357, "step": 4885 }, { "epoch": 0.38, "learning_rate": 1.425761094484358e-06, "loss": 1.2563, "step": 4886 }, { "epoch": 0.38, "learning_rate": 1.4255336164483993e-06, "loss": 1.2784, "step": 4887 }, { "epoch": 0.38, "learning_rate": 1.4253061115205038e-06, "loss": 1.2193, "step": 4888 }, { "epoch": 0.38, "learning_rate": 1.4250785797150494e-06, "loss": 1.2047, "step": 4889 }, { "epoch": 0.38, "learning_rate": 1.4248510210464147e-06, "loss": 1.1983, "step": 4890 }, { "epoch": 0.38, "learning_rate": 1.4246234355289806e-06, "loss": 0.8225, "step": 4891 }, { "epoch": 0.38, "learning_rate": 1.4243958231771293e-06, "loss": 1.1896, "step": 4892 }, { "epoch": 0.38, "learning_rate": 1.4241681840052455e-06, "loss": 1.223, "step": 4893 }, { "epoch": 0.38, "learning_rate": 1.4239405180277152e-06, "loss": 1.2048, "step": 4894 }, { "epoch": 0.38, "learning_rate": 1.4237128252589247e-06, "loss": 0.8338, "step": 4895 }, { "epoch": 0.38, "learning_rate": 1.4234851057132645e-06, "loss": 1.2355, "step": 4896 }, { "epoch": 0.38, "learning_rate": 1.4232573594051246e-06, "loss": 1.1731, "step": 4897 }, { "epoch": 0.38, "learning_rate": 1.4230295863488983e-06, "loss": 1.1997, "step": 4898 }, { "epoch": 0.38, "learning_rate": 1.4228017865589798e-06, "loss": 1.2025, "step": 4899 }, { "epoch": 0.38, "learning_rate": 1.4225739600497648e-06, "loss": 1.2318, "step": 4900 }, { "epoch": 0.38, "learning_rate": 1.422346106835651e-06, "loss": 1.226, "step": 4901 }, { "epoch": 0.38, "learning_rate": 1.422118226931038e-06, "loss": 1.2049, "step": 4902 }, { "epoch": 0.38, "learning_rate": 1.4218903203503268e-06, "loss": 1.2042, "step": 4903 }, { "epoch": 0.38, "learning_rate": 1.42166238710792e-06, "loss": 1.2307, "step": 4904 }, { "epoch": 0.38, "learning_rate": 1.4214344272182216e-06, "loss": 1.2328, "step": 4905 }, { "epoch": 0.38, "learning_rate": 1.4212064406956388e-06, "loss": 1.2233, "step": 4906 }, { "epoch": 0.38, "learning_rate": 1.4209784275545785e-06, "loss": 1.2552, "step": 4907 }, { "epoch": 0.38, "learning_rate": 1.4207503878094507e-06, "loss": 0.9008, "step": 4908 }, { "epoch": 0.38, "learning_rate": 1.420522321474666e-06, "loss": 1.1828, "step": 4909 }, { "epoch": 0.38, "learning_rate": 1.4202942285646379e-06, "loss": 1.2229, "step": 4910 }, { "epoch": 0.38, "learning_rate": 1.4200661090937805e-06, "loss": 1.1758, "step": 4911 }, { "epoch": 0.38, "learning_rate": 1.4198379630765098e-06, "loss": 0.8382, "step": 4912 }, { "epoch": 0.38, "learning_rate": 1.4196097905272438e-06, "loss": 1.1779, "step": 4913 }, { "epoch": 0.38, "learning_rate": 1.4193815914604024e-06, "loss": 0.8595, "step": 4914 }, { "epoch": 0.38, "learning_rate": 1.4191533658904064e-06, "loss": 1.1378, "step": 4915 }, { "epoch": 0.38, "learning_rate": 1.4189251138316794e-06, "loss": 1.2139, "step": 4916 }, { "epoch": 0.38, "learning_rate": 1.418696835298645e-06, "loss": 1.2321, "step": 4917 }, { "epoch": 0.38, "learning_rate": 1.41846853030573e-06, "loss": 1.2787, "step": 4918 }, { "epoch": 0.38, "learning_rate": 1.4182401988673623e-06, "loss": 1.2485, "step": 4919 }, { "epoch": 0.38, "learning_rate": 1.418011840997971e-06, "loss": 0.8329, "step": 4920 }, { "epoch": 0.38, "learning_rate": 1.4177834567119883e-06, "loss": 1.1762, "step": 4921 }, { "epoch": 0.38, "learning_rate": 1.4175550460238463e-06, "loss": 1.2009, "step": 4922 }, { "epoch": 0.38, "learning_rate": 1.4173266089479801e-06, "loss": 0.8118, "step": 4923 }, { "epoch": 0.38, "learning_rate": 1.4170981454988256e-06, "loss": 1.1765, "step": 4924 }, { "epoch": 0.38, "learning_rate": 1.4168696556908209e-06, "loss": 1.2339, "step": 4925 }, { "epoch": 0.38, "learning_rate": 1.4166411395384054e-06, "loss": 1.2644, "step": 4926 }, { "epoch": 0.38, "learning_rate": 1.4164125970560208e-06, "loss": 1.2104, "step": 4927 }, { "epoch": 0.38, "learning_rate": 1.4161840282581098e-06, "loss": 0.8502, "step": 4928 }, { "epoch": 0.38, "learning_rate": 1.4159554331591167e-06, "loss": 1.1978, "step": 4929 }, { "epoch": 0.38, "learning_rate": 1.415726811773488e-06, "loss": 1.2361, "step": 4930 }, { "epoch": 0.38, "learning_rate": 1.4154981641156716e-06, "loss": 1.2104, "step": 4931 }, { "epoch": 0.38, "learning_rate": 1.4152694902001174e-06, "loss": 1.2314, "step": 4932 }, { "epoch": 0.38, "learning_rate": 1.4150407900412763e-06, "loss": 1.2627, "step": 4933 }, { "epoch": 0.38, "learning_rate": 1.4148120636536005e-06, "loss": 1.2146, "step": 4934 }, { "epoch": 0.38, "learning_rate": 1.4145833110515457e-06, "loss": 1.2744, "step": 4935 }, { "epoch": 0.38, "learning_rate": 1.4143545322495679e-06, "loss": 1.2467, "step": 4936 }, { "epoch": 0.38, "learning_rate": 1.4141257272621243e-06, "loss": 1.2376, "step": 4937 }, { "epoch": 0.38, "learning_rate": 1.413896896103675e-06, "loss": 1.2966, "step": 4938 }, { "epoch": 0.38, "learning_rate": 1.413668038788681e-06, "loss": 1.1934, "step": 4939 }, { "epoch": 0.38, "learning_rate": 1.4134391553316053e-06, "loss": 1.2167, "step": 4940 }, { "epoch": 0.38, "learning_rate": 1.413210245746912e-06, "loss": 0.8436, "step": 4941 }, { "epoch": 0.38, "learning_rate": 1.4129813100490672e-06, "loss": 1.1954, "step": 4942 }, { "epoch": 0.38, "learning_rate": 1.412752348252539e-06, "loss": 1.257, "step": 4943 }, { "epoch": 0.38, "learning_rate": 1.4125233603717967e-06, "loss": 1.2332, "step": 4944 }, { "epoch": 0.38, "learning_rate": 1.4122943464213115e-06, "loss": 1.2378, "step": 4945 }, { "epoch": 0.38, "learning_rate": 1.4120653064155562e-06, "loss": 1.2097, "step": 4946 }, { "epoch": 0.38, "learning_rate": 1.4118362403690043e-06, "loss": 0.8118, "step": 4947 }, { "epoch": 0.38, "learning_rate": 1.411607148296133e-06, "loss": 1.2154, "step": 4948 }, { "epoch": 0.38, "learning_rate": 1.4113780302114193e-06, "loss": 1.1982, "step": 4949 }, { "epoch": 0.38, "learning_rate": 1.4111488861293426e-06, "loss": 1.278, "step": 4950 }, { "epoch": 0.38, "learning_rate": 1.4109197160643837e-06, "loss": 1.2034, "step": 4951 }, { "epoch": 0.38, "learning_rate": 1.4106905200310256e-06, "loss": 1.2536, "step": 4952 }, { "epoch": 0.38, "learning_rate": 1.4104612980437523e-06, "loss": 1.1815, "step": 4953 }, { "epoch": 0.38, "learning_rate": 1.4102320501170497e-06, "loss": 1.1934, "step": 4954 }, { "epoch": 0.38, "learning_rate": 1.410002776265405e-06, "loss": 1.198, "step": 4955 }, { "epoch": 0.38, "learning_rate": 1.4097734765033078e-06, "loss": 1.2698, "step": 4956 }, { "epoch": 0.38, "learning_rate": 1.4095441508452485e-06, "loss": 1.2944, "step": 4957 }, { "epoch": 0.38, "learning_rate": 1.40931479930572e-06, "loss": 1.2254, "step": 4958 }, { "epoch": 0.38, "learning_rate": 1.4090854218992156e-06, "loss": 1.1721, "step": 4959 }, { "epoch": 0.38, "learning_rate": 1.4088560186402319e-06, "loss": 1.2112, "step": 4960 }, { "epoch": 0.39, "learning_rate": 1.4086265895432655e-06, "loss": 0.8565, "step": 4961 }, { "epoch": 0.39, "learning_rate": 1.4083971346228156e-06, "loss": 1.2406, "step": 4962 }, { "epoch": 0.39, "learning_rate": 1.4081676538933828e-06, "loss": 1.2217, "step": 4963 }, { "epoch": 0.39, "learning_rate": 1.4079381473694693e-06, "loss": 1.2212, "step": 4964 }, { "epoch": 0.39, "learning_rate": 1.4077086150655793e-06, "loss": 1.1923, "step": 4965 }, { "epoch": 0.39, "learning_rate": 1.4074790569962178e-06, "loss": 1.261, "step": 4966 }, { "epoch": 0.39, "learning_rate": 1.407249473175892e-06, "loss": 1.1872, "step": 4967 }, { "epoch": 0.39, "learning_rate": 1.4070198636191103e-06, "loss": 1.2448, "step": 4968 }, { "epoch": 0.39, "learning_rate": 1.406790228340384e-06, "loss": 1.2258, "step": 4969 }, { "epoch": 0.39, "learning_rate": 1.4065605673542243e-06, "loss": 1.1963, "step": 4970 }, { "epoch": 0.39, "learning_rate": 1.4063308806751452e-06, "loss": 0.8333, "step": 4971 }, { "epoch": 0.39, "learning_rate": 1.4061011683176615e-06, "loss": 1.2338, "step": 4972 }, { "epoch": 0.39, "learning_rate": 1.4058714302962905e-06, "loss": 1.1783, "step": 4973 }, { "epoch": 0.39, "learning_rate": 1.4056416666255504e-06, "loss": 1.2232, "step": 4974 }, { "epoch": 0.39, "learning_rate": 1.4054118773199615e-06, "loss": 1.2101, "step": 4975 }, { "epoch": 0.39, "learning_rate": 1.4051820623940453e-06, "loss": 1.1717, "step": 4976 }, { "epoch": 0.39, "learning_rate": 1.4049522218623256e-06, "loss": 1.1618, "step": 4977 }, { "epoch": 0.39, "learning_rate": 1.404722355739327e-06, "loss": 1.2896, "step": 4978 }, { "epoch": 0.39, "learning_rate": 1.4044924640395758e-06, "loss": 1.2163, "step": 4979 }, { "epoch": 0.39, "learning_rate": 1.4042625467776003e-06, "loss": 1.2602, "step": 4980 }, { "epoch": 0.39, "learning_rate": 1.4040326039679306e-06, "loss": 1.2067, "step": 4981 }, { "epoch": 0.39, "learning_rate": 1.4038026356250982e-06, "loss": 1.2299, "step": 4982 }, { "epoch": 0.39, "learning_rate": 1.403572641763636e-06, "loss": 1.2495, "step": 4983 }, { "epoch": 0.39, "learning_rate": 1.4033426223980784e-06, "loss": 1.2369, "step": 4984 }, { "epoch": 0.39, "learning_rate": 1.4031125775429618e-06, "loss": 1.2433, "step": 4985 }, { "epoch": 0.39, "learning_rate": 1.4028825072128241e-06, "loss": 1.1804, "step": 4986 }, { "epoch": 0.39, "learning_rate": 1.402652411422205e-06, "loss": 1.201, "step": 4987 }, { "epoch": 0.39, "learning_rate": 1.4024222901856454e-06, "loss": 1.1969, "step": 4988 }, { "epoch": 0.39, "learning_rate": 1.4021921435176876e-06, "loss": 1.1843, "step": 4989 }, { "epoch": 0.39, "learning_rate": 1.4019619714328765e-06, "loss": 1.2607, "step": 4990 }, { "epoch": 0.39, "learning_rate": 1.4017317739457578e-06, "loss": 1.1854, "step": 4991 }, { "epoch": 0.39, "learning_rate": 1.401501551070879e-06, "loss": 1.1903, "step": 4992 }, { "epoch": 0.39, "learning_rate": 1.401271302822789e-06, "loss": 1.2207, "step": 4993 }, { "epoch": 0.39, "learning_rate": 1.401041029216039e-06, "loss": 1.2349, "step": 4994 }, { "epoch": 0.39, "learning_rate": 1.400810730265181e-06, "loss": 1.2434, "step": 4995 }, { "epoch": 0.39, "learning_rate": 1.400580405984769e-06, "loss": 0.8484, "step": 4996 }, { "epoch": 0.39, "learning_rate": 1.4003500563893584e-06, "loss": 1.2359, "step": 4997 }, { "epoch": 0.39, "learning_rate": 1.4001196814935065e-06, "loss": 0.8476, "step": 4998 }, { "epoch": 0.39, "learning_rate": 1.399889281311772e-06, "loss": 1.2434, "step": 4999 }, { "epoch": 0.39, "learning_rate": 1.399658855858715e-06, "loss": 1.2142, "step": 5000 }, { "epoch": 0.39, "learning_rate": 1.3994284051488977e-06, "loss": 1.1924, "step": 5001 }, { "epoch": 0.39, "learning_rate": 1.3991979291968834e-06, "loss": 1.244, "step": 5002 }, { "epoch": 0.39, "learning_rate": 1.3989674280172376e-06, "loss": 1.194, "step": 5003 }, { "epoch": 0.39, "learning_rate": 1.3987369016245269e-06, "loss": 1.2076, "step": 5004 }, { "epoch": 0.39, "learning_rate": 1.398506350033319e-06, "loss": 0.8292, "step": 5005 }, { "epoch": 0.39, "learning_rate": 1.3982757732581844e-06, "loss": 1.2071, "step": 5006 }, { "epoch": 0.39, "learning_rate": 1.3980451713136945e-06, "loss": 1.2517, "step": 5007 }, { "epoch": 0.39, "learning_rate": 1.397814544214422e-06, "loss": 0.8692, "step": 5008 }, { "epoch": 0.39, "learning_rate": 1.3975838919749422e-06, "loss": 1.1787, "step": 5009 }, { "epoch": 0.39, "learning_rate": 1.3973532146098308e-06, "loss": 1.2259, "step": 5010 }, { "epoch": 0.39, "learning_rate": 1.397122512133666e-06, "loss": 1.2329, "step": 5011 }, { "epoch": 0.39, "learning_rate": 1.396891784561027e-06, "loss": 1.213, "step": 5012 }, { "epoch": 0.39, "learning_rate": 1.396661031906495e-06, "loss": 1.2251, "step": 5013 }, { "epoch": 0.39, "learning_rate": 1.3964302541846523e-06, "loss": 1.1989, "step": 5014 }, { "epoch": 0.39, "learning_rate": 1.3961994514100833e-06, "loss": 0.8577, "step": 5015 }, { "epoch": 0.39, "learning_rate": 1.3959686235973736e-06, "loss": 1.2226, "step": 5016 }, { "epoch": 0.39, "learning_rate": 1.3957377707611108e-06, "loss": 0.837, "step": 5017 }, { "epoch": 0.39, "learning_rate": 1.3955068929158838e-06, "loss": 1.1958, "step": 5018 }, { "epoch": 0.39, "learning_rate": 1.3952759900762827e-06, "loss": 1.1538, "step": 5019 }, { "epoch": 0.39, "learning_rate": 1.3950450622569004e-06, "loss": 1.1839, "step": 5020 }, { "epoch": 0.39, "learning_rate": 1.3948141094723297e-06, "loss": 1.2379, "step": 5021 }, { "epoch": 0.39, "learning_rate": 1.3945831317371663e-06, "loss": 1.2674, "step": 5022 }, { "epoch": 0.39, "learning_rate": 1.394352129066007e-06, "loss": 1.233, "step": 5023 }, { "epoch": 0.39, "learning_rate": 1.3941211014734502e-06, "loss": 1.2195, "step": 5024 }, { "epoch": 0.39, "learning_rate": 1.3938900489740957e-06, "loss": 1.2733, "step": 5025 }, { "epoch": 0.39, "learning_rate": 1.3936589715825452e-06, "loss": 1.2454, "step": 5026 }, { "epoch": 0.39, "learning_rate": 1.3934278693134018e-06, "loss": 1.1645, "step": 5027 }, { "epoch": 0.39, "learning_rate": 1.3931967421812702e-06, "loss": 1.2161, "step": 5028 }, { "epoch": 0.39, "learning_rate": 1.3929655902007565e-06, "loss": 1.1966, "step": 5029 }, { "epoch": 0.39, "learning_rate": 1.392734413386469e-06, "loss": 1.1792, "step": 5030 }, { "epoch": 0.39, "learning_rate": 1.3925032117530164e-06, "loss": 1.2344, "step": 5031 }, { "epoch": 0.39, "learning_rate": 1.3922719853150102e-06, "loss": 0.8876, "step": 5032 }, { "epoch": 0.39, "learning_rate": 1.3920407340870628e-06, "loss": 1.2189, "step": 5033 }, { "epoch": 0.39, "learning_rate": 1.3918094580837882e-06, "loss": 1.1839, "step": 5034 }, { "epoch": 0.39, "learning_rate": 1.3915781573198022e-06, "loss": 0.8375, "step": 5035 }, { "epoch": 0.39, "learning_rate": 1.3913468318097219e-06, "loss": 1.2347, "step": 5036 }, { "epoch": 0.39, "learning_rate": 1.3911154815681665e-06, "loss": 0.8074, "step": 5037 }, { "epoch": 0.39, "learning_rate": 1.3908841066097558e-06, "loss": 1.2681, "step": 5038 }, { "epoch": 0.39, "learning_rate": 1.3906527069491123e-06, "loss": 1.214, "step": 5039 }, { "epoch": 0.39, "learning_rate": 1.3904212826008585e-06, "loss": 1.2102, "step": 5040 }, { "epoch": 0.39, "learning_rate": 1.3901898335796207e-06, "loss": 1.2308, "step": 5041 }, { "epoch": 0.39, "learning_rate": 1.389958359900025e-06, "loss": 1.2508, "step": 5042 }, { "epoch": 0.39, "learning_rate": 1.3897268615766994e-06, "loss": 1.265, "step": 5043 }, { "epoch": 0.39, "learning_rate": 1.3894953386242735e-06, "loss": 1.2782, "step": 5044 }, { "epoch": 0.39, "learning_rate": 1.3892637910573789e-06, "loss": 0.8141, "step": 5045 }, { "epoch": 0.39, "learning_rate": 1.3890322188906483e-06, "loss": 1.2006, "step": 5046 }, { "epoch": 0.39, "learning_rate": 1.3888006221387161e-06, "loss": 1.2012, "step": 5047 }, { "epoch": 0.39, "learning_rate": 1.388569000816218e-06, "loss": 0.8469, "step": 5048 }, { "epoch": 0.39, "learning_rate": 1.3883373549377922e-06, "loss": 1.2154, "step": 5049 }, { "epoch": 0.39, "learning_rate": 1.3881056845180773e-06, "loss": 1.2976, "step": 5050 }, { "epoch": 0.39, "learning_rate": 1.3878739895717134e-06, "loss": 1.2438, "step": 5051 }, { "epoch": 0.39, "learning_rate": 1.3876422701133437e-06, "loss": 1.232, "step": 5052 }, { "epoch": 0.39, "learning_rate": 1.387410526157611e-06, "loss": 0.8015, "step": 5053 }, { "epoch": 0.39, "learning_rate": 1.387178757719161e-06, "loss": 1.1994, "step": 5054 }, { "epoch": 0.39, "learning_rate": 1.3869469648126404e-06, "loss": 1.1993, "step": 5055 }, { "epoch": 0.39, "learning_rate": 1.3867151474526972e-06, "loss": 0.8642, "step": 5056 }, { "epoch": 0.39, "learning_rate": 1.3864833056539817e-06, "loss": 1.2165, "step": 5057 }, { "epoch": 0.39, "learning_rate": 1.3862514394311454e-06, "loss": 1.2579, "step": 5058 }, { "epoch": 0.39, "learning_rate": 1.3860195487988411e-06, "loss": 1.2294, "step": 5059 }, { "epoch": 0.39, "learning_rate": 1.3857876337717233e-06, "loss": 1.2038, "step": 5060 }, { "epoch": 0.39, "learning_rate": 1.3855556943644477e-06, "loss": 1.1713, "step": 5061 }, { "epoch": 0.39, "learning_rate": 1.3853237305916724e-06, "loss": 1.0838, "step": 5062 }, { "epoch": 0.39, "learning_rate": 1.3850917424680568e-06, "loss": 1.2593, "step": 5063 }, { "epoch": 0.39, "learning_rate": 1.384859730008261e-06, "loss": 1.1789, "step": 5064 }, { "epoch": 0.39, "learning_rate": 1.3846276932269473e-06, "loss": 1.1533, "step": 5065 }, { "epoch": 0.39, "learning_rate": 1.3843956321387792e-06, "loss": 1.2073, "step": 5066 }, { "epoch": 0.39, "learning_rate": 1.3841635467584226e-06, "loss": 1.2029, "step": 5067 }, { "epoch": 0.39, "learning_rate": 1.383931437100544e-06, "loss": 0.866, "step": 5068 }, { "epoch": 0.39, "learning_rate": 1.3836993031798119e-06, "loss": 1.1901, "step": 5069 }, { "epoch": 0.39, "learning_rate": 1.3834671450108961e-06, "loss": 0.8702, "step": 5070 }, { "epoch": 0.39, "learning_rate": 1.3832349626084677e-06, "loss": 1.2446, "step": 5071 }, { "epoch": 0.39, "learning_rate": 1.3830027559872002e-06, "loss": 0.8382, "step": 5072 }, { "epoch": 0.39, "learning_rate": 1.3827705251617679e-06, "loss": 1.1857, "step": 5073 }, { "epoch": 0.39, "learning_rate": 1.3825382701468465e-06, "loss": 1.2305, "step": 5074 }, { "epoch": 0.39, "learning_rate": 1.382305990957114e-06, "loss": 1.2454, "step": 5075 }, { "epoch": 0.39, "learning_rate": 1.3820736876072492e-06, "loss": 1.2423, "step": 5076 }, { "epoch": 0.39, "learning_rate": 1.3818413601119328e-06, "loss": 1.2458, "step": 5077 }, { "epoch": 0.39, "learning_rate": 1.3816090084858466e-06, "loss": 1.1835, "step": 5078 }, { "epoch": 0.39, "learning_rate": 1.3813766327436743e-06, "loss": 1.2393, "step": 5079 }, { "epoch": 0.39, "learning_rate": 1.381144232900102e-06, "loss": 1.2362, "step": 5080 }, { "epoch": 0.39, "learning_rate": 1.3809118089698155e-06, "loss": 0.8762, "step": 5081 }, { "epoch": 0.39, "learning_rate": 1.3806793609675027e-06, "loss": 1.2136, "step": 5082 }, { "epoch": 0.39, "learning_rate": 1.3804468889078544e-06, "loss": 1.2122, "step": 5083 }, { "epoch": 0.39, "learning_rate": 1.380214392805561e-06, "loss": 1.2071, "step": 5084 }, { "epoch": 0.39, "learning_rate": 1.3799818726753157e-06, "loss": 1.2204, "step": 5085 }, { "epoch": 0.39, "learning_rate": 1.3797493285318125e-06, "loss": 1.2496, "step": 5086 }, { "epoch": 0.39, "learning_rate": 1.3795167603897475e-06, "loss": 1.1983, "step": 5087 }, { "epoch": 0.39, "learning_rate": 1.379284168263818e-06, "loss": 1.2257, "step": 5088 }, { "epoch": 0.39, "learning_rate": 1.3790515521687227e-06, "loss": 1.1855, "step": 5089 }, { "epoch": 0.4, "learning_rate": 1.378818912119162e-06, "loss": 1.2423, "step": 5090 }, { "epoch": 0.4, "learning_rate": 1.3785862481298376e-06, "loss": 1.22, "step": 5091 }, { "epoch": 0.4, "learning_rate": 1.3783535602154533e-06, "loss": 1.1697, "step": 5092 }, { "epoch": 0.4, "learning_rate": 1.3781208483907137e-06, "loss": 1.2267, "step": 5093 }, { "epoch": 0.4, "learning_rate": 1.377888112670325e-06, "loss": 1.2657, "step": 5094 }, { "epoch": 0.4, "learning_rate": 1.3776553530689956e-06, "loss": 1.2391, "step": 5095 }, { "epoch": 0.4, "learning_rate": 1.3774225696014349e-06, "loss": 1.2071, "step": 5096 }, { "epoch": 0.4, "learning_rate": 1.3771897622823536e-06, "loss": 1.2257, "step": 5097 }, { "epoch": 0.4, "learning_rate": 1.3769569311264642e-06, "loss": 1.2071, "step": 5098 }, { "epoch": 0.4, "learning_rate": 1.3767240761484804e-06, "loss": 1.2206, "step": 5099 }, { "epoch": 0.4, "learning_rate": 1.3764911973631182e-06, "loss": 1.2077, "step": 5100 }, { "epoch": 0.4, "learning_rate": 1.376258294785094e-06, "loss": 1.2747, "step": 5101 }, { "epoch": 0.4, "learning_rate": 1.376025368429127e-06, "loss": 1.2718, "step": 5102 }, { "epoch": 0.4, "learning_rate": 1.3757924183099364e-06, "loss": 1.2461, "step": 5103 }, { "epoch": 0.4, "learning_rate": 1.3755594444422438e-06, "loss": 1.267, "step": 5104 }, { "epoch": 0.4, "learning_rate": 1.3753264468407725e-06, "loss": 1.2856, "step": 5105 }, { "epoch": 0.4, "learning_rate": 1.3750934255202472e-06, "loss": 1.1633, "step": 5106 }, { "epoch": 0.4, "learning_rate": 1.3748603804953932e-06, "loss": 0.8625, "step": 5107 }, { "epoch": 0.4, "learning_rate": 1.3746273117809381e-06, "loss": 1.2192, "step": 5108 }, { "epoch": 0.4, "learning_rate": 1.374394219391611e-06, "loss": 1.2395, "step": 5109 }, { "epoch": 0.4, "learning_rate": 1.3741611033421428e-06, "loss": 1.1964, "step": 5110 }, { "epoch": 0.4, "learning_rate": 1.3739279636472645e-06, "loss": 1.2003, "step": 5111 }, { "epoch": 0.4, "learning_rate": 1.3736948003217102e-06, "loss": 1.2583, "step": 5112 }, { "epoch": 0.4, "learning_rate": 1.373461613380215e-06, "loss": 1.2086, "step": 5113 }, { "epoch": 0.4, "learning_rate": 1.373228402837515e-06, "loss": 1.2291, "step": 5114 }, { "epoch": 0.4, "learning_rate": 1.372995168708348e-06, "loss": 1.2577, "step": 5115 }, { "epoch": 0.4, "learning_rate": 1.3727619110074535e-06, "loss": 1.2088, "step": 5116 }, { "epoch": 0.4, "learning_rate": 1.3725286297495727e-06, "loss": 1.2231, "step": 5117 }, { "epoch": 0.4, "learning_rate": 1.3722953249494477e-06, "loss": 1.1846, "step": 5118 }, { "epoch": 0.4, "learning_rate": 1.3720619966218227e-06, "loss": 1.1621, "step": 5119 }, { "epoch": 0.4, "learning_rate": 1.3718286447814424e-06, "loss": 1.2174, "step": 5120 }, { "epoch": 0.4, "learning_rate": 1.3715952694430546e-06, "loss": 1.2242, "step": 5121 }, { "epoch": 0.4, "learning_rate": 1.3713618706214067e-06, "loss": 1.2397, "step": 5122 }, { "epoch": 0.4, "learning_rate": 1.3711284483312493e-06, "loss": 0.8493, "step": 5123 }, { "epoch": 0.4, "learning_rate": 1.3708950025873334e-06, "loss": 1.24, "step": 5124 }, { "epoch": 0.4, "learning_rate": 1.3706615334044116e-06, "loss": 1.266, "step": 5125 }, { "epoch": 0.4, "learning_rate": 1.3704280407972383e-06, "loss": 1.1848, "step": 5126 }, { "epoch": 0.4, "learning_rate": 1.3701945247805697e-06, "loss": 1.2422, "step": 5127 }, { "epoch": 0.4, "learning_rate": 1.3699609853691622e-06, "loss": 0.8194, "step": 5128 }, { "epoch": 0.4, "learning_rate": 1.369727422577775e-06, "loss": 1.2132, "step": 5129 }, { "epoch": 0.4, "learning_rate": 1.3694938364211686e-06, "loss": 1.1741, "step": 5130 }, { "epoch": 0.4, "learning_rate": 1.3692602269141039e-06, "loss": 1.1792, "step": 5131 }, { "epoch": 0.4, "learning_rate": 1.3690265940713444e-06, "loss": 1.2424, "step": 5132 }, { "epoch": 0.4, "learning_rate": 1.3687929379076547e-06, "loss": 0.8303, "step": 5133 }, { "epoch": 0.4, "learning_rate": 1.3685592584378012e-06, "loss": 1.2278, "step": 5134 }, { "epoch": 0.4, "learning_rate": 1.368325555676551e-06, "loss": 1.1881, "step": 5135 }, { "epoch": 0.4, "learning_rate": 1.3680918296386734e-06, "loss": 1.2662, "step": 5136 }, { "epoch": 0.4, "learning_rate": 1.3678580803389388e-06, "loss": 1.2137, "step": 5137 }, { "epoch": 0.4, "learning_rate": 1.367624307792119e-06, "loss": 0.8409, "step": 5138 }, { "epoch": 0.4, "learning_rate": 1.367390512012988e-06, "loss": 1.1739, "step": 5139 }, { "epoch": 0.4, "learning_rate": 1.3671566930163201e-06, "loss": 1.1928, "step": 5140 }, { "epoch": 0.4, "learning_rate": 1.3669228508168922e-06, "loss": 1.1968, "step": 5141 }, { "epoch": 0.4, "learning_rate": 1.3666889854294816e-06, "loss": 1.2424, "step": 5142 }, { "epoch": 0.4, "learning_rate": 1.3664550968688679e-06, "loss": 1.2113, "step": 5143 }, { "epoch": 0.4, "learning_rate": 1.366221185149832e-06, "loss": 1.2091, "step": 5144 }, { "epoch": 0.4, "learning_rate": 1.3659872502871559e-06, "loss": 0.8394, "step": 5145 }, { "epoch": 0.4, "learning_rate": 1.3657532922956235e-06, "loss": 1.2611, "step": 5146 }, { "epoch": 0.4, "learning_rate": 1.36551931119002e-06, "loss": 1.2382, "step": 5147 }, { "epoch": 0.4, "learning_rate": 1.3652853069851314e-06, "loss": 0.8644, "step": 5148 }, { "epoch": 0.4, "learning_rate": 1.3650512796957466e-06, "loss": 1.1945, "step": 5149 }, { "epoch": 0.4, "learning_rate": 1.3648172293366545e-06, "loss": 1.2015, "step": 5150 }, { "epoch": 0.4, "learning_rate": 1.3645831559226465e-06, "loss": 0.8008, "step": 5151 }, { "epoch": 0.4, "learning_rate": 1.3643490594685155e-06, "loss": 1.2026, "step": 5152 }, { "epoch": 0.4, "learning_rate": 1.3641149399890544e-06, "loss": 0.863, "step": 5153 }, { "epoch": 0.4, "learning_rate": 1.363880797499059e-06, "loss": 1.2152, "step": 5154 }, { "epoch": 0.4, "learning_rate": 1.3636466320133263e-06, "loss": 1.2341, "step": 5155 }, { "epoch": 0.4, "learning_rate": 1.3634124435466544e-06, "loss": 1.2585, "step": 5156 }, { "epoch": 0.4, "learning_rate": 1.3631782321138429e-06, "loss": 1.2269, "step": 5157 }, { "epoch": 0.4, "learning_rate": 1.3629439977296932e-06, "loss": 1.1762, "step": 5158 }, { "epoch": 0.4, "learning_rate": 1.362709740409008e-06, "loss": 0.8504, "step": 5159 }, { "epoch": 0.4, "learning_rate": 1.3624754601665913e-06, "loss": 0.7982, "step": 5160 }, { "epoch": 0.4, "learning_rate": 1.3622411570172485e-06, "loss": 1.1992, "step": 5161 }, { "epoch": 0.4, "learning_rate": 1.3620068309757867e-06, "loss": 1.2313, "step": 5162 }, { "epoch": 0.4, "learning_rate": 1.3617724820570142e-06, "loss": 1.2326, "step": 5163 }, { "epoch": 0.4, "learning_rate": 1.3615381102757408e-06, "loss": 0.8412, "step": 5164 }, { "epoch": 0.4, "learning_rate": 1.361303715646778e-06, "loss": 1.2688, "step": 5165 }, { "epoch": 0.4, "learning_rate": 1.3610692981849388e-06, "loss": 1.1671, "step": 5166 }, { "epoch": 0.4, "learning_rate": 1.3608348579050368e-06, "loss": 1.2235, "step": 5167 }, { "epoch": 0.4, "learning_rate": 1.360600394821888e-06, "loss": 0.848, "step": 5168 }, { "epoch": 0.4, "learning_rate": 1.3603659089503097e-06, "loss": 1.2309, "step": 5169 }, { "epoch": 0.4, "learning_rate": 1.3601314003051198e-06, "loss": 1.1688, "step": 5170 }, { "epoch": 0.4, "learning_rate": 1.3598968689011386e-06, "loss": 0.7951, "step": 5171 }, { "epoch": 0.4, "learning_rate": 1.3596623147531876e-06, "loss": 1.2406, "step": 5172 }, { "epoch": 0.4, "learning_rate": 1.3594277378760898e-06, "loss": 1.2288, "step": 5173 }, { "epoch": 0.4, "learning_rate": 1.359193138284669e-06, "loss": 1.2364, "step": 5174 }, { "epoch": 0.4, "learning_rate": 1.3589585159937511e-06, "loss": 1.2147, "step": 5175 }, { "epoch": 0.4, "learning_rate": 1.3587238710181632e-06, "loss": 1.2427, "step": 5176 }, { "epoch": 0.4, "learning_rate": 1.3584892033727342e-06, "loss": 1.2245, "step": 5177 }, { "epoch": 0.4, "learning_rate": 1.3582545130722935e-06, "loss": 1.2346, "step": 5178 }, { "epoch": 0.4, "learning_rate": 1.3580198001316732e-06, "loss": 1.2653, "step": 5179 }, { "epoch": 0.4, "learning_rate": 1.3577850645657057e-06, "loss": 0.8371, "step": 5180 }, { "epoch": 0.4, "learning_rate": 1.3575503063892257e-06, "loss": 1.2674, "step": 5181 }, { "epoch": 0.4, "learning_rate": 1.3573155256170686e-06, "loss": 1.2136, "step": 5182 }, { "epoch": 0.4, "learning_rate": 1.3570807222640715e-06, "loss": 1.151, "step": 5183 }, { "epoch": 0.4, "learning_rate": 1.3568458963450733e-06, "loss": 1.2009, "step": 5184 }, { "epoch": 0.4, "learning_rate": 1.3566110478749137e-06, "loss": 1.1877, "step": 5185 }, { "epoch": 0.4, "learning_rate": 1.3563761768684345e-06, "loss": 0.8442, "step": 5186 }, { "epoch": 0.4, "learning_rate": 1.3561412833404782e-06, "loss": 1.2605, "step": 5187 }, { "epoch": 0.4, "learning_rate": 1.355906367305889e-06, "loss": 1.2087, "step": 5188 }, { "epoch": 0.4, "learning_rate": 1.3556714287795133e-06, "loss": 1.1979, "step": 5189 }, { "epoch": 0.4, "learning_rate": 1.3554364677761973e-06, "loss": 1.197, "step": 5190 }, { "epoch": 0.4, "learning_rate": 1.3552014843107906e-06, "loss": 1.1918, "step": 5191 }, { "epoch": 0.4, "learning_rate": 1.354966478398142e-06, "loss": 1.2416, "step": 5192 }, { "epoch": 0.4, "learning_rate": 1.3547314500531038e-06, "loss": 1.2377, "step": 5193 }, { "epoch": 0.4, "learning_rate": 1.3544963992905286e-06, "loss": 1.1686, "step": 5194 }, { "epoch": 0.4, "learning_rate": 1.3542613261252702e-06, "loss": 1.2486, "step": 5195 }, { "epoch": 0.4, "learning_rate": 1.3540262305721847e-06, "loss": 0.8463, "step": 5196 }, { "epoch": 0.4, "learning_rate": 1.353791112646129e-06, "loss": 1.2252, "step": 5197 }, { "epoch": 0.4, "learning_rate": 1.3535559723619613e-06, "loss": 1.2116, "step": 5198 }, { "epoch": 0.4, "learning_rate": 1.3533208097345421e-06, "loss": 1.2028, "step": 5199 }, { "epoch": 0.4, "learning_rate": 1.3530856247787322e-06, "loss": 1.1942, "step": 5200 }, { "epoch": 0.4, "learning_rate": 1.3528504175093942e-06, "loss": 1.2386, "step": 5201 }, { "epoch": 0.4, "learning_rate": 1.3526151879413927e-06, "loss": 1.217, "step": 5202 }, { "epoch": 0.4, "learning_rate": 1.3523799360895926e-06, "loss": 0.8193, "step": 5203 }, { "epoch": 0.4, "learning_rate": 1.3521446619688615e-06, "loss": 1.2204, "step": 5204 }, { "epoch": 0.4, "learning_rate": 1.3519093655940675e-06, "loss": 1.2367, "step": 5205 }, { "epoch": 0.4, "learning_rate": 1.3516740469800798e-06, "loss": 1.2224, "step": 5206 }, { "epoch": 0.4, "learning_rate": 1.3514387061417706e-06, "loss": 1.1345, "step": 5207 }, { "epoch": 0.4, "learning_rate": 1.3512033430940114e-06, "loss": 1.2167, "step": 5208 }, { "epoch": 0.4, "learning_rate": 1.3509679578516764e-06, "loss": 1.263, "step": 5209 }, { "epoch": 0.4, "learning_rate": 1.3507325504296415e-06, "loss": 1.2691, "step": 5210 }, { "epoch": 0.4, "learning_rate": 1.3504971208427831e-06, "loss": 0.8041, "step": 5211 }, { "epoch": 0.4, "learning_rate": 1.3502616691059797e-06, "loss": 1.1987, "step": 5212 }, { "epoch": 0.4, "learning_rate": 1.35002619523411e-06, "loss": 1.2397, "step": 5213 }, { "epoch": 0.4, "learning_rate": 1.3497906992420558e-06, "loss": 1.2107, "step": 5214 }, { "epoch": 0.4, "learning_rate": 1.3495551811446992e-06, "loss": 1.2487, "step": 5215 }, { "epoch": 0.4, "learning_rate": 1.3493196409569239e-06, "loss": 1.1667, "step": 5216 }, { "epoch": 0.4, "learning_rate": 1.3490840786936151e-06, "loss": 1.2298, "step": 5217 }, { "epoch": 0.4, "learning_rate": 1.3488484943696592e-06, "loss": 1.2096, "step": 5218 }, { "epoch": 0.41, "learning_rate": 1.3486128879999444e-06, "loss": 1.2417, "step": 5219 }, { "epoch": 0.41, "learning_rate": 1.34837725959936e-06, "loss": 1.2121, "step": 5220 }, { "epoch": 0.41, "learning_rate": 1.3481416091827962e-06, "loss": 1.1804, "step": 5221 }, { "epoch": 0.41, "learning_rate": 1.3479059367651458e-06, "loss": 1.2525, "step": 5222 }, { "epoch": 0.41, "learning_rate": 1.3476702423613018e-06, "loss": 0.8164, "step": 5223 }, { "epoch": 0.41, "learning_rate": 1.3474345259861598e-06, "loss": 1.2678, "step": 5224 }, { "epoch": 0.41, "learning_rate": 1.3471987876546152e-06, "loss": 1.2281, "step": 5225 }, { "epoch": 0.41, "learning_rate": 1.346963027381566e-06, "loss": 1.2174, "step": 5226 }, { "epoch": 0.41, "learning_rate": 1.3467272451819116e-06, "loss": 0.8436, "step": 5227 }, { "epoch": 0.41, "learning_rate": 1.346491441070552e-06, "loss": 1.2181, "step": 5228 }, { "epoch": 0.41, "learning_rate": 1.3462556150623897e-06, "loss": 1.1882, "step": 5229 }, { "epoch": 0.41, "learning_rate": 1.346019767172327e-06, "loss": 1.2512, "step": 5230 }, { "epoch": 0.41, "learning_rate": 1.345783897415269e-06, "loss": 1.2138, "step": 5231 }, { "epoch": 0.41, "learning_rate": 1.3455480058061215e-06, "loss": 1.248, "step": 5232 }, { "epoch": 0.41, "learning_rate": 1.345312092359792e-06, "loss": 1.275, "step": 5233 }, { "epoch": 0.41, "learning_rate": 1.3450761570911895e-06, "loss": 1.2703, "step": 5234 }, { "epoch": 0.41, "learning_rate": 1.3448402000152237e-06, "loss": 1.2055, "step": 5235 }, { "epoch": 0.41, "learning_rate": 1.344604221146806e-06, "loss": 1.1738, "step": 5236 }, { "epoch": 0.41, "learning_rate": 1.3443682205008496e-06, "loss": 1.1795, "step": 5237 }, { "epoch": 0.41, "learning_rate": 1.3441321980922689e-06, "loss": 1.2265, "step": 5238 }, { "epoch": 0.41, "learning_rate": 1.343896153935979e-06, "loss": 0.8722, "step": 5239 }, { "epoch": 0.41, "learning_rate": 1.343660088046897e-06, "loss": 0.8235, "step": 5240 }, { "epoch": 0.41, "learning_rate": 1.3434240004399417e-06, "loss": 1.231, "step": 5241 }, { "epoch": 0.41, "learning_rate": 1.3431878911300326e-06, "loss": 1.2012, "step": 5242 }, { "epoch": 0.41, "learning_rate": 1.3429517601320908e-06, "loss": 1.2444, "step": 5243 }, { "epoch": 0.41, "learning_rate": 1.3427156074610384e-06, "loss": 0.8444, "step": 5244 }, { "epoch": 0.41, "learning_rate": 1.3424794331318e-06, "loss": 1.2349, "step": 5245 }, { "epoch": 0.41, "learning_rate": 1.3422432371593005e-06, "loss": 1.2265, "step": 5246 }, { "epoch": 0.41, "learning_rate": 1.3420070195584662e-06, "loss": 1.1995, "step": 5247 }, { "epoch": 0.41, "learning_rate": 1.3417707803442254e-06, "loss": 0.858, "step": 5248 }, { "epoch": 0.41, "learning_rate": 1.3415345195315074e-06, "loss": 1.1937, "step": 5249 }, { "epoch": 0.41, "learning_rate": 1.3412982371352428e-06, "loss": 1.2326, "step": 5250 }, { "epoch": 0.41, "learning_rate": 1.3410619331703638e-06, "loss": 1.2178, "step": 5251 }, { "epoch": 0.41, "learning_rate": 1.3408256076518037e-06, "loss": 1.2106, "step": 5252 }, { "epoch": 0.41, "learning_rate": 1.340589260594497e-06, "loss": 1.2684, "step": 5253 }, { "epoch": 0.41, "learning_rate": 1.3403528920133805e-06, "loss": 1.215, "step": 5254 }, { "epoch": 0.41, "learning_rate": 1.3401165019233912e-06, "loss": 1.2914, "step": 5255 }, { "epoch": 0.41, "learning_rate": 1.339880090339468e-06, "loss": 1.1713, "step": 5256 }, { "epoch": 0.41, "learning_rate": 1.3396436572765513e-06, "loss": 1.1699, "step": 5257 }, { "epoch": 0.41, "learning_rate": 1.3394072027495824e-06, "loss": 1.1901, "step": 5258 }, { "epoch": 0.41, "learning_rate": 1.3391707267735047e-06, "loss": 1.2468, "step": 5259 }, { "epoch": 0.41, "learning_rate": 1.3389342293632621e-06, "loss": 1.2284, "step": 5260 }, { "epoch": 0.41, "learning_rate": 1.3386977105337999e-06, "loss": 1.2343, "step": 5261 }, { "epoch": 0.41, "learning_rate": 1.338461170300066e-06, "loss": 1.1559, "step": 5262 }, { "epoch": 0.41, "learning_rate": 1.3382246086770085e-06, "loss": 1.1874, "step": 5263 }, { "epoch": 0.41, "learning_rate": 1.3379880256795767e-06, "loss": 1.228, "step": 5264 }, { "epoch": 0.41, "learning_rate": 1.3377514213227214e-06, "loss": 1.2315, "step": 5265 }, { "epoch": 0.41, "learning_rate": 1.337514795621396e-06, "loss": 1.2448, "step": 5266 }, { "epoch": 0.41, "learning_rate": 1.3372781485905535e-06, "loss": 1.215, "step": 5267 }, { "epoch": 0.41, "learning_rate": 1.3370414802451493e-06, "loss": 1.2424, "step": 5268 }, { "epoch": 0.41, "learning_rate": 1.3368047906001396e-06, "loss": 1.2348, "step": 5269 }, { "epoch": 0.41, "learning_rate": 1.3365680796704822e-06, "loss": 1.205, "step": 5270 }, { "epoch": 0.41, "learning_rate": 1.3363313474711366e-06, "loss": 1.2664, "step": 5271 }, { "epoch": 0.41, "learning_rate": 1.3360945940170631e-06, "loss": 1.1937, "step": 5272 }, { "epoch": 0.41, "learning_rate": 1.3358578193232233e-06, "loss": 1.2023, "step": 5273 }, { "epoch": 0.41, "learning_rate": 1.3356210234045805e-06, "loss": 1.2087, "step": 5274 }, { "epoch": 0.41, "learning_rate": 1.3353842062760993e-06, "loss": 1.2219, "step": 5275 }, { "epoch": 0.41, "learning_rate": 1.3351473679527455e-06, "loss": 1.2161, "step": 5276 }, { "epoch": 0.41, "learning_rate": 1.3349105084494864e-06, "loss": 1.2768, "step": 5277 }, { "epoch": 0.41, "learning_rate": 1.3346736277812903e-06, "loss": 1.2046, "step": 5278 }, { "epoch": 0.41, "learning_rate": 1.3344367259631273e-06, "loss": 1.2392, "step": 5279 }, { "epoch": 0.41, "learning_rate": 1.3341998030099682e-06, "loss": 1.2376, "step": 5280 }, { "epoch": 0.41, "learning_rate": 1.333962858936786e-06, "loss": 1.1585, "step": 5281 }, { "epoch": 0.41, "learning_rate": 1.333725893758554e-06, "loss": 0.8664, "step": 5282 }, { "epoch": 0.41, "learning_rate": 1.333488907490248e-06, "loss": 1.2093, "step": 5283 }, { "epoch": 0.41, "learning_rate": 1.3332519001468446e-06, "loss": 1.1842, "step": 5284 }, { "epoch": 0.41, "learning_rate": 1.333014871743321e-06, "loss": 1.252, "step": 5285 }, { "epoch": 0.41, "learning_rate": 1.332777822294657e-06, "loss": 1.2255, "step": 5286 }, { "epoch": 0.41, "learning_rate": 1.3325407518158328e-06, "loss": 1.2291, "step": 5287 }, { "epoch": 0.41, "learning_rate": 1.3323036603218302e-06, "loss": 0.8401, "step": 5288 }, { "epoch": 0.41, "learning_rate": 1.3320665478276327e-06, "loss": 1.1709, "step": 5289 }, { "epoch": 0.41, "learning_rate": 1.3318294143482249e-06, "loss": 1.178, "step": 5290 }, { "epoch": 0.41, "learning_rate": 1.3315922598985921e-06, "loss": 1.1971, "step": 5291 }, { "epoch": 0.41, "learning_rate": 1.3313550844937217e-06, "loss": 0.7968, "step": 5292 }, { "epoch": 0.41, "learning_rate": 1.3311178881486025e-06, "loss": 1.221, "step": 5293 }, { "epoch": 0.41, "learning_rate": 1.330880670878224e-06, "loss": 0.8422, "step": 5294 }, { "epoch": 0.41, "learning_rate": 1.3306434326975774e-06, "loss": 1.2098, "step": 5295 }, { "epoch": 0.41, "learning_rate": 1.3304061736216551e-06, "loss": 1.2031, "step": 5296 }, { "epoch": 0.41, "learning_rate": 1.3301688936654512e-06, "loss": 0.8599, "step": 5297 }, { "epoch": 0.41, "learning_rate": 1.3299315928439603e-06, "loss": 1.2532, "step": 5298 }, { "epoch": 0.41, "learning_rate": 1.3296942711721787e-06, "loss": 0.8935, "step": 5299 }, { "epoch": 0.41, "learning_rate": 1.3294569286651048e-06, "loss": 0.8013, "step": 5300 }, { "epoch": 0.41, "learning_rate": 1.3292195653377375e-06, "loss": 1.2139, "step": 5301 }, { "epoch": 0.41, "learning_rate": 1.328982181205077e-06, "loss": 1.1937, "step": 5302 }, { "epoch": 0.41, "learning_rate": 1.3287447762821247e-06, "loss": 1.2209, "step": 5303 }, { "epoch": 0.41, "learning_rate": 1.328507350583884e-06, "loss": 1.209, "step": 5304 }, { "epoch": 0.41, "learning_rate": 1.3282699041253591e-06, "loss": 1.207, "step": 5305 }, { "epoch": 0.41, "learning_rate": 1.3280324369215555e-06, "loss": 1.211, "step": 5306 }, { "epoch": 0.41, "learning_rate": 1.3277949489874802e-06, "loss": 1.164, "step": 5307 }, { "epoch": 0.41, "learning_rate": 1.3275574403381415e-06, "loss": 1.1768, "step": 5308 }, { "epoch": 0.41, "learning_rate": 1.327319910988549e-06, "loss": 1.2274, "step": 5309 }, { "epoch": 0.41, "learning_rate": 1.3270823609537135e-06, "loss": 1.2246, "step": 5310 }, { "epoch": 0.41, "learning_rate": 1.326844790248647e-06, "loss": 1.2414, "step": 5311 }, { "epoch": 0.41, "learning_rate": 1.326607198888363e-06, "loss": 1.2308, "step": 5312 }, { "epoch": 0.41, "learning_rate": 1.3263695868878766e-06, "loss": 1.165, "step": 5313 }, { "epoch": 0.41, "learning_rate": 1.3261319542622037e-06, "loss": 1.22, "step": 5314 }, { "epoch": 0.41, "learning_rate": 1.3258943010263614e-06, "loss": 1.2944, "step": 5315 }, { "epoch": 0.41, "learning_rate": 1.3256566271953685e-06, "loss": 1.1909, "step": 5316 }, { "epoch": 0.41, "learning_rate": 1.3254189327842453e-06, "loss": 1.2693, "step": 5317 }, { "epoch": 0.41, "learning_rate": 1.3251812178080133e-06, "loss": 1.1899, "step": 5318 }, { "epoch": 0.41, "learning_rate": 1.3249434822816943e-06, "loss": 1.2003, "step": 5319 }, { "epoch": 0.41, "learning_rate": 1.3247057262203124e-06, "loss": 1.1928, "step": 5320 }, { "epoch": 0.41, "learning_rate": 1.324467949638893e-06, "loss": 1.2189, "step": 5321 }, { "epoch": 0.41, "learning_rate": 1.3242301525524631e-06, "loss": 1.2276, "step": 5322 }, { "epoch": 0.41, "learning_rate": 1.3239923349760493e-06, "loss": 1.2485, "step": 5323 }, { "epoch": 0.41, "learning_rate": 1.3237544969246813e-06, "loss": 1.1995, "step": 5324 }, { "epoch": 0.41, "learning_rate": 1.32351663841339e-06, "loss": 1.2023, "step": 5325 }, { "epoch": 0.41, "learning_rate": 1.3232787594572062e-06, "loss": 1.1752, "step": 5326 }, { "epoch": 0.41, "learning_rate": 1.3230408600711631e-06, "loss": 1.2118, "step": 5327 }, { "epoch": 0.41, "learning_rate": 1.3228029402702951e-06, "loss": 1.191, "step": 5328 }, { "epoch": 0.41, "learning_rate": 1.3225650000696375e-06, "loss": 0.8438, "step": 5329 }, { "epoch": 0.41, "learning_rate": 1.3223270394842273e-06, "loss": 0.8866, "step": 5330 }, { "epoch": 0.41, "learning_rate": 1.3220890585291027e-06, "loss": 1.2038, "step": 5331 }, { "epoch": 0.41, "learning_rate": 1.321851057219303e-06, "loss": 1.2323, "step": 5332 }, { "epoch": 0.41, "learning_rate": 1.3216130355698685e-06, "loss": 1.227, "step": 5333 }, { "epoch": 0.41, "learning_rate": 1.3213749935958417e-06, "loss": 1.2496, "step": 5334 }, { "epoch": 0.41, "learning_rate": 1.3211369313122659e-06, "loss": 1.2438, "step": 5335 }, { "epoch": 0.41, "learning_rate": 1.320898848734185e-06, "loss": 1.2195, "step": 5336 }, { "epoch": 0.41, "learning_rate": 1.3206607458766451e-06, "loss": 1.2325, "step": 5337 }, { "epoch": 0.41, "learning_rate": 1.3204226227546939e-06, "loss": 1.2238, "step": 5338 }, { "epoch": 0.41, "learning_rate": 1.3201844793833792e-06, "loss": 1.2144, "step": 5339 }, { "epoch": 0.41, "learning_rate": 1.3199463157777508e-06, "loss": 1.1612, "step": 5340 }, { "epoch": 0.41, "learning_rate": 1.3197081319528593e-06, "loss": 1.1862, "step": 5341 }, { "epoch": 0.41, "learning_rate": 1.3194699279237575e-06, "loss": 1.1967, "step": 5342 }, { "epoch": 0.41, "learning_rate": 1.3192317037054985e-06, "loss": 1.1872, "step": 5343 }, { "epoch": 0.41, "learning_rate": 1.3189934593131372e-06, "loss": 1.2138, "step": 5344 }, { "epoch": 0.41, "learning_rate": 1.3187551947617296e-06, "loss": 1.2352, "step": 5345 }, { "epoch": 0.41, "learning_rate": 1.3185169100663332e-06, "loss": 1.2447, "step": 5346 }, { "epoch": 0.41, "learning_rate": 1.3182786052420064e-06, "loss": 1.2484, "step": 5347 }, { "epoch": 0.42, "learning_rate": 1.3180402803038088e-06, "loss": 1.1503, "step": 5348 }, { "epoch": 0.42, "learning_rate": 1.3178019352668019e-06, "loss": 1.2515, "step": 5349 }, { "epoch": 0.42, "learning_rate": 1.317563570146048e-06, "loss": 1.1924, "step": 5350 }, { "epoch": 0.42, "learning_rate": 1.3173251849566114e-06, "loss": 1.1799, "step": 5351 }, { "epoch": 0.42, "learning_rate": 1.317086779713556e-06, "loss": 0.8494, "step": 5352 }, { "epoch": 0.42, "learning_rate": 1.3168483544319484e-06, "loss": 1.1849, "step": 5353 }, { "epoch": 0.42, "learning_rate": 1.3166099091268562e-06, "loss": 1.1552, "step": 5354 }, { "epoch": 0.42, "learning_rate": 1.3163714438133482e-06, "loss": 0.8336, "step": 5355 }, { "epoch": 0.42, "learning_rate": 1.3161329585064945e-06, "loss": 1.2118, "step": 5356 }, { "epoch": 0.42, "learning_rate": 1.3158944532213658e-06, "loss": 1.1901, "step": 5357 }, { "epoch": 0.42, "learning_rate": 1.3156559279730349e-06, "loss": 1.2284, "step": 5358 }, { "epoch": 0.42, "learning_rate": 1.3154173827765758e-06, "loss": 1.2785, "step": 5359 }, { "epoch": 0.42, "learning_rate": 1.3151788176470636e-06, "loss": 1.1904, "step": 5360 }, { "epoch": 0.42, "learning_rate": 1.3149402325995741e-06, "loss": 1.2367, "step": 5361 }, { "epoch": 0.42, "learning_rate": 1.3147016276491855e-06, "loss": 1.2075, "step": 5362 }, { "epoch": 0.42, "learning_rate": 1.3144630028109764e-06, "loss": 1.2572, "step": 5363 }, { "epoch": 0.42, "learning_rate": 1.3142243581000267e-06, "loss": 0.8819, "step": 5364 }, { "epoch": 0.42, "learning_rate": 1.313985693531418e-06, "loss": 1.1898, "step": 5365 }, { "epoch": 0.42, "learning_rate": 1.3137470091202327e-06, "loss": 1.2225, "step": 5366 }, { "epoch": 0.42, "learning_rate": 1.3135083048815547e-06, "loss": 1.2051, "step": 5367 }, { "epoch": 0.42, "learning_rate": 1.3132695808304692e-06, "loss": 1.208, "step": 5368 }, { "epoch": 0.42, "learning_rate": 1.3130308369820627e-06, "loss": 1.2237, "step": 5369 }, { "epoch": 0.42, "learning_rate": 1.3127920733514222e-06, "loss": 1.2161, "step": 5370 }, { "epoch": 0.42, "learning_rate": 1.312553289953637e-06, "loss": 1.2335, "step": 5371 }, { "epoch": 0.42, "learning_rate": 1.3123144868037972e-06, "loss": 1.2072, "step": 5372 }, { "epoch": 0.42, "learning_rate": 1.3120756639169945e-06, "loss": 0.8383, "step": 5373 }, { "epoch": 0.42, "learning_rate": 1.3118368213083208e-06, "loss": 1.1852, "step": 5374 }, { "epoch": 0.42, "learning_rate": 1.3115979589928703e-06, "loss": 1.2672, "step": 5375 }, { "epoch": 0.42, "learning_rate": 1.311359076985738e-06, "loss": 1.2459, "step": 5376 }, { "epoch": 0.42, "learning_rate": 1.3111201753020205e-06, "loss": 1.2344, "step": 5377 }, { "epoch": 0.42, "learning_rate": 1.3108812539568153e-06, "loss": 1.2015, "step": 5378 }, { "epoch": 0.42, "learning_rate": 1.3106423129652207e-06, "loss": 1.2732, "step": 5379 }, { "epoch": 0.42, "learning_rate": 1.3104033523423375e-06, "loss": 1.2333, "step": 5380 }, { "epoch": 0.42, "learning_rate": 1.3101643721032666e-06, "loss": 1.2418, "step": 5381 }, { "epoch": 0.42, "learning_rate": 1.3099253722631107e-06, "loss": 0.8712, "step": 5382 }, { "epoch": 0.42, "learning_rate": 1.3096863528369736e-06, "loss": 1.1877, "step": 5383 }, { "epoch": 0.42, "learning_rate": 1.3094473138399597e-06, "loss": 1.2236, "step": 5384 }, { "epoch": 0.42, "learning_rate": 1.3092082552871764e-06, "loss": 1.1955, "step": 5385 }, { "epoch": 0.42, "learning_rate": 1.3089691771937306e-06, "loss": 1.252, "step": 5386 }, { "epoch": 0.42, "learning_rate": 1.3087300795747307e-06, "loss": 1.183, "step": 5387 }, { "epoch": 0.42, "learning_rate": 1.3084909624452868e-06, "loss": 1.2092, "step": 5388 }, { "epoch": 0.42, "learning_rate": 1.3082518258205112e-06, "loss": 1.1801, "step": 5389 }, { "epoch": 0.42, "learning_rate": 1.3080126697155147e-06, "loss": 1.1777, "step": 5390 }, { "epoch": 0.42, "learning_rate": 1.307773494145412e-06, "loss": 1.228, "step": 5391 }, { "epoch": 0.42, "learning_rate": 1.3075342991253172e-06, "loss": 0.8223, "step": 5392 }, { "epoch": 0.42, "learning_rate": 1.3072950846703475e-06, "loss": 1.197, "step": 5393 }, { "epoch": 0.42, "learning_rate": 1.3070558507956194e-06, "loss": 1.1933, "step": 5394 }, { "epoch": 0.42, "learning_rate": 1.3068165975162519e-06, "loss": 1.2379, "step": 5395 }, { "epoch": 0.42, "learning_rate": 1.3065773248473642e-06, "loss": 1.267, "step": 5396 }, { "epoch": 0.42, "learning_rate": 1.306338032804078e-06, "loss": 1.2366, "step": 5397 }, { "epoch": 0.42, "learning_rate": 1.3060987214015154e-06, "loss": 1.1978, "step": 5398 }, { "epoch": 0.42, "learning_rate": 1.3058593906547996e-06, "loss": 1.2276, "step": 5399 }, { "epoch": 0.42, "learning_rate": 1.3056200405790555e-06, "loss": 1.1581, "step": 5400 }, { "epoch": 0.42, "learning_rate": 1.3053806711894094e-06, "loss": 1.2058, "step": 5401 }, { "epoch": 0.42, "learning_rate": 1.3051412825009876e-06, "loss": 1.2664, "step": 5402 }, { "epoch": 0.42, "learning_rate": 1.304901874528919e-06, "loss": 1.1785, "step": 5403 }, { "epoch": 0.42, "learning_rate": 1.304662447288333e-06, "loss": 1.2115, "step": 5404 }, { "epoch": 0.42, "learning_rate": 1.3044230007943604e-06, "loss": 1.168, "step": 5405 }, { "epoch": 0.42, "learning_rate": 1.3041835350621335e-06, "loss": 1.1466, "step": 5406 }, { "epoch": 0.42, "learning_rate": 1.3039440501067853e-06, "loss": 1.217, "step": 5407 }, { "epoch": 0.42, "learning_rate": 1.30370454594345e-06, "loss": 1.2544, "step": 5408 }, { "epoch": 0.42, "learning_rate": 1.3034650225872637e-06, "loss": 0.8394, "step": 5409 }, { "epoch": 0.42, "learning_rate": 1.303225480053363e-06, "loss": 1.2529, "step": 5410 }, { "epoch": 0.42, "learning_rate": 1.3029859183568865e-06, "loss": 1.2569, "step": 5411 }, { "epoch": 0.42, "learning_rate": 1.3027463375129726e-06, "loss": 0.8173, "step": 5412 }, { "epoch": 0.42, "learning_rate": 1.3025067375367622e-06, "loss": 1.2261, "step": 5413 }, { "epoch": 0.42, "learning_rate": 1.3022671184433971e-06, "loss": 1.2153, "step": 5414 }, { "epoch": 0.42, "learning_rate": 1.3020274802480203e-06, "loss": 1.2073, "step": 5415 }, { "epoch": 0.42, "learning_rate": 1.301787822965776e-06, "loss": 1.1697, "step": 5416 }, { "epoch": 0.42, "learning_rate": 1.301548146611809e-06, "loss": 0.8918, "step": 5417 }, { "epoch": 0.42, "learning_rate": 1.3013084512012665e-06, "loss": 1.1961, "step": 5418 }, { "epoch": 0.42, "learning_rate": 1.3010687367492955e-06, "loss": 1.1935, "step": 5419 }, { "epoch": 0.42, "learning_rate": 1.3008290032710459e-06, "loss": 1.232, "step": 5420 }, { "epoch": 0.42, "learning_rate": 1.3005892507816668e-06, "loss": 0.8511, "step": 5421 }, { "epoch": 0.42, "learning_rate": 1.3003494792963106e-06, "loss": 1.2286, "step": 5422 }, { "epoch": 0.42, "learning_rate": 1.300109688830129e-06, "loss": 1.2418, "step": 5423 }, { "epoch": 0.42, "learning_rate": 1.299869879398276e-06, "loss": 1.2325, "step": 5424 }, { "epoch": 0.42, "learning_rate": 1.2996300510159071e-06, "loss": 1.262, "step": 5425 }, { "epoch": 0.42, "learning_rate": 1.2993902036981772e-06, "loss": 1.2182, "step": 5426 }, { "epoch": 0.42, "learning_rate": 1.299150337460245e-06, "loss": 1.2483, "step": 5427 }, { "epoch": 0.42, "learning_rate": 1.2989104523172684e-06, "loss": 1.1926, "step": 5428 }, { "epoch": 0.42, "learning_rate": 1.2986705482844072e-06, "loss": 1.2222, "step": 5429 }, { "epoch": 0.42, "learning_rate": 1.298430625376822e-06, "loss": 1.194, "step": 5430 }, { "epoch": 0.42, "learning_rate": 1.2981906836096754e-06, "loss": 1.1733, "step": 5431 }, { "epoch": 0.42, "learning_rate": 1.2979507229981304e-06, "loss": 1.2464, "step": 5432 }, { "epoch": 0.42, "learning_rate": 1.297710743557352e-06, "loss": 1.2651, "step": 5433 }, { "epoch": 0.42, "learning_rate": 1.297470745302505e-06, "loss": 1.2199, "step": 5434 }, { "epoch": 0.42, "learning_rate": 1.2972307282487574e-06, "loss": 1.2083, "step": 5435 }, { "epoch": 0.42, "learning_rate": 1.2969906924112764e-06, "loss": 0.8256, "step": 5436 }, { "epoch": 0.42, "learning_rate": 1.2967506378052314e-06, "loss": 1.2308, "step": 5437 }, { "epoch": 0.42, "learning_rate": 1.2965105644457934e-06, "loss": 1.2538, "step": 5438 }, { "epoch": 0.42, "learning_rate": 1.2962704723481333e-06, "loss": 1.224, "step": 5439 }, { "epoch": 0.42, "learning_rate": 1.2960303615274244e-06, "loss": 1.1936, "step": 5440 }, { "epoch": 0.42, "learning_rate": 1.2957902319988403e-06, "loss": 1.2515, "step": 5441 }, { "epoch": 0.42, "learning_rate": 1.2955500837775566e-06, "loss": 1.1116, "step": 5442 }, { "epoch": 0.42, "learning_rate": 1.2953099168787491e-06, "loss": 1.2073, "step": 5443 }, { "epoch": 0.42, "learning_rate": 1.2950697313175962e-06, "loss": 1.2321, "step": 5444 }, { "epoch": 0.42, "learning_rate": 1.2948295271092757e-06, "loss": 1.1966, "step": 5445 }, { "epoch": 0.42, "learning_rate": 1.2945893042689681e-06, "loss": 1.1184, "step": 5446 }, { "epoch": 0.42, "learning_rate": 1.2943490628118543e-06, "loss": 1.1961, "step": 5447 }, { "epoch": 0.42, "learning_rate": 1.2941088027531165e-06, "loss": 1.2061, "step": 5448 }, { "epoch": 0.42, "learning_rate": 1.293868524107938e-06, "loss": 1.1866, "step": 5449 }, { "epoch": 0.42, "learning_rate": 1.293628226891504e-06, "loss": 1.2209, "step": 5450 }, { "epoch": 0.42, "learning_rate": 1.2933879111189993e-06, "loss": 1.226, "step": 5451 }, { "epoch": 0.42, "learning_rate": 1.2931475768056117e-06, "loss": 0.8391, "step": 5452 }, { "epoch": 0.42, "learning_rate": 1.292907223966529e-06, "loss": 1.2324, "step": 5453 }, { "epoch": 0.42, "learning_rate": 1.2926668526169402e-06, "loss": 1.2023, "step": 5454 }, { "epoch": 0.42, "learning_rate": 1.2924264627720364e-06, "loss": 1.2014, "step": 5455 }, { "epoch": 0.42, "learning_rate": 1.2921860544470086e-06, "loss": 1.1996, "step": 5456 }, { "epoch": 0.42, "learning_rate": 1.2919456276570501e-06, "loss": 1.16, "step": 5457 }, { "epoch": 0.42, "learning_rate": 1.2917051824173545e-06, "loss": 1.2341, "step": 5458 }, { "epoch": 0.42, "learning_rate": 1.2914647187431172e-06, "loss": 1.2566, "step": 5459 }, { "epoch": 0.42, "learning_rate": 1.2912242366495341e-06, "loss": 0.8698, "step": 5460 }, { "epoch": 0.42, "learning_rate": 1.2909837361518032e-06, "loss": 1.1932, "step": 5461 }, { "epoch": 0.42, "learning_rate": 1.2907432172651227e-06, "loss": 1.217, "step": 5462 }, { "epoch": 0.42, "learning_rate": 1.2905026800046925e-06, "loss": 0.9052, "step": 5463 }, { "epoch": 0.42, "learning_rate": 1.2902621243857134e-06, "loss": 1.2634, "step": 5464 }, { "epoch": 0.42, "learning_rate": 1.290021550423388e-06, "loss": 1.2324, "step": 5465 }, { "epoch": 0.42, "learning_rate": 1.2897809581329192e-06, "loss": 1.2455, "step": 5466 }, { "epoch": 0.42, "learning_rate": 1.2895403475295115e-06, "loss": 1.2144, "step": 5467 }, { "epoch": 0.42, "learning_rate": 1.28929971862837e-06, "loss": 1.252, "step": 5468 }, { "epoch": 0.42, "learning_rate": 1.2890590714447022e-06, "loss": 1.2138, "step": 5469 }, { "epoch": 0.42, "learning_rate": 1.2888184059937158e-06, "loss": 1.2356, "step": 5470 }, { "epoch": 0.42, "learning_rate": 1.2885777222906192e-06, "loss": 1.1854, "step": 5471 }, { "epoch": 0.42, "learning_rate": 1.2883370203506236e-06, "loss": 1.2647, "step": 5472 }, { "epoch": 0.42, "learning_rate": 1.2880963001889397e-06, "loss": 1.2552, "step": 5473 }, { "epoch": 0.42, "learning_rate": 1.2878555618207803e-06, "loss": 0.8205, "step": 5474 }, { "epoch": 0.42, "learning_rate": 1.2876148052613588e-06, "loss": 1.228, "step": 5475 }, { "epoch": 0.43, "learning_rate": 1.28737403052589e-06, "loss": 1.2252, "step": 5476 }, { "epoch": 0.43, "learning_rate": 1.28713323762959e-06, "loss": 1.2377, "step": 5477 }, { "epoch": 0.43, "learning_rate": 1.2868924265876759e-06, "loss": 0.8549, "step": 5478 }, { "epoch": 0.43, "learning_rate": 1.286651597415366e-06, "loss": 1.1885, "step": 5479 }, { "epoch": 0.43, "learning_rate": 1.2864107501278794e-06, "loss": 1.2044, "step": 5480 }, { "epoch": 0.43, "learning_rate": 1.286169884740437e-06, "loss": 0.8472, "step": 5481 }, { "epoch": 0.43, "learning_rate": 1.28592900126826e-06, "loss": 1.2411, "step": 5482 }, { "epoch": 0.43, "learning_rate": 1.2856880997265723e-06, "loss": 1.1606, "step": 5483 }, { "epoch": 0.43, "learning_rate": 1.2854471801305967e-06, "loss": 0.8344, "step": 5484 }, { "epoch": 0.43, "learning_rate": 1.2852062424955583e-06, "loss": 1.2189, "step": 5485 }, { "epoch": 0.43, "learning_rate": 1.284965286836684e-06, "loss": 1.2187, "step": 5486 }, { "epoch": 0.43, "learning_rate": 1.2847243131692011e-06, "loss": 1.2041, "step": 5487 }, { "epoch": 0.43, "learning_rate": 1.2844833215083382e-06, "loss": 1.2029, "step": 5488 }, { "epoch": 0.43, "learning_rate": 1.2842423118693243e-06, "loss": 1.291, "step": 5489 }, { "epoch": 0.43, "learning_rate": 1.2840012842673906e-06, "loss": 1.2845, "step": 5490 }, { "epoch": 0.43, "learning_rate": 1.2837602387177692e-06, "loss": 1.2186, "step": 5491 }, { "epoch": 0.43, "learning_rate": 1.2835191752356931e-06, "loss": 1.2288, "step": 5492 }, { "epoch": 0.43, "learning_rate": 1.2832780938363964e-06, "loss": 1.2261, "step": 5493 }, { "epoch": 0.43, "learning_rate": 1.2830369945351145e-06, "loss": 1.1952, "step": 5494 }, { "epoch": 0.43, "learning_rate": 1.2827958773470838e-06, "loss": 1.222, "step": 5495 }, { "epoch": 0.43, "learning_rate": 1.2825547422875418e-06, "loss": 1.2066, "step": 5496 }, { "epoch": 0.43, "learning_rate": 1.2823135893717274e-06, "loss": 1.2435, "step": 5497 }, { "epoch": 0.43, "learning_rate": 1.2820724186148801e-06, "loss": 1.2108, "step": 5498 }, { "epoch": 0.43, "learning_rate": 1.2818312300322416e-06, "loss": 1.2595, "step": 5499 }, { "epoch": 0.43, "learning_rate": 1.2815900236390534e-06, "loss": 1.2288, "step": 5500 }, { "epoch": 0.43, "learning_rate": 1.2813487994505589e-06, "loss": 1.2308, "step": 5501 }, { "epoch": 0.43, "learning_rate": 1.2811075574820023e-06, "loss": 1.2024, "step": 5502 }, { "epoch": 0.43, "learning_rate": 1.2808662977486292e-06, "loss": 1.2131, "step": 5503 }, { "epoch": 0.43, "learning_rate": 1.2806250202656865e-06, "loss": 1.1909, "step": 5504 }, { "epoch": 0.43, "learning_rate": 1.2803837250484217e-06, "loss": 1.2634, "step": 5505 }, { "epoch": 0.43, "learning_rate": 1.2801424121120834e-06, "loss": 1.2323, "step": 5506 }, { "epoch": 0.43, "learning_rate": 1.2799010814719217e-06, "loss": 1.1971, "step": 5507 }, { "epoch": 0.43, "learning_rate": 1.279659733143188e-06, "loss": 1.2065, "step": 5508 }, { "epoch": 0.43, "learning_rate": 1.2794183671411341e-06, "loss": 1.1893, "step": 5509 }, { "epoch": 0.43, "learning_rate": 1.2791769834810135e-06, "loss": 1.2259, "step": 5510 }, { "epoch": 0.43, "learning_rate": 1.2789355821780805e-06, "loss": 1.2289, "step": 5511 }, { "epoch": 0.43, "learning_rate": 1.278694163247591e-06, "loss": 0.8328, "step": 5512 }, { "epoch": 0.43, "learning_rate": 1.2784527267048012e-06, "loss": 1.2143, "step": 5513 }, { "epoch": 0.43, "learning_rate": 1.2782112725649692e-06, "loss": 1.1736, "step": 5514 }, { "epoch": 0.43, "learning_rate": 1.2779698008433535e-06, "loss": 1.2486, "step": 5515 }, { "epoch": 0.43, "learning_rate": 1.2777283115552146e-06, "loss": 1.2044, "step": 5516 }, { "epoch": 0.43, "learning_rate": 1.2774868047158134e-06, "loss": 1.2444, "step": 5517 }, { "epoch": 0.43, "learning_rate": 1.277245280340412e-06, "loss": 1.2328, "step": 5518 }, { "epoch": 0.43, "learning_rate": 1.2770037384442735e-06, "loss": 1.2621, "step": 5519 }, { "epoch": 0.43, "learning_rate": 1.276762179042663e-06, "loss": 1.274, "step": 5520 }, { "epoch": 0.43, "learning_rate": 1.2765206021508462e-06, "loss": 1.2432, "step": 5521 }, { "epoch": 0.43, "learning_rate": 1.2762790077840886e-06, "loss": 1.1902, "step": 5522 }, { "epoch": 0.43, "learning_rate": 1.2760373959576584e-06, "loss": 1.2701, "step": 5523 }, { "epoch": 0.43, "learning_rate": 1.275795766686825e-06, "loss": 1.2615, "step": 5524 }, { "epoch": 0.43, "learning_rate": 1.2755541199868578e-06, "loss": 1.2356, "step": 5525 }, { "epoch": 0.43, "learning_rate": 1.2753124558730282e-06, "loss": 1.2014, "step": 5526 }, { "epoch": 0.43, "learning_rate": 1.2750707743606081e-06, "loss": 0.8155, "step": 5527 }, { "epoch": 0.43, "learning_rate": 1.274829075464871e-06, "loss": 1.2177, "step": 5528 }, { "epoch": 0.43, "learning_rate": 1.274587359201091e-06, "loss": 1.1513, "step": 5529 }, { "epoch": 0.43, "learning_rate": 1.2743456255845438e-06, "loss": 0.8529, "step": 5530 }, { "epoch": 0.43, "learning_rate": 1.2741038746305058e-06, "loss": 1.2458, "step": 5531 }, { "epoch": 0.43, "learning_rate": 1.2738621063542546e-06, "loss": 1.2271, "step": 5532 }, { "epoch": 0.43, "learning_rate": 1.273620320771069e-06, "loss": 1.2244, "step": 5533 }, { "epoch": 0.43, "learning_rate": 1.2733785178962287e-06, "loss": 1.2305, "step": 5534 }, { "epoch": 0.43, "learning_rate": 1.2731366977450148e-06, "loss": 1.1978, "step": 5535 }, { "epoch": 0.43, "learning_rate": 1.2728948603327092e-06, "loss": 0.8131, "step": 5536 }, { "epoch": 0.43, "learning_rate": 1.2726530056745952e-06, "loss": 1.2248, "step": 5537 }, { "epoch": 0.43, "learning_rate": 1.2724111337859571e-06, "loss": 1.2599, "step": 5538 }, { "epoch": 0.43, "learning_rate": 1.2721692446820797e-06, "loss": 0.8256, "step": 5539 }, { "epoch": 0.43, "learning_rate": 1.2719273383782497e-06, "loss": 1.2262, "step": 5540 }, { "epoch": 0.43, "learning_rate": 1.2716854148897544e-06, "loss": 1.1943, "step": 5541 }, { "epoch": 0.43, "learning_rate": 1.2714434742318826e-06, "loss": 1.1936, "step": 5542 }, { "epoch": 0.43, "learning_rate": 1.2712015164199242e-06, "loss": 0.8596, "step": 5543 }, { "epoch": 0.43, "learning_rate": 1.2709595414691687e-06, "loss": 1.2139, "step": 5544 }, { "epoch": 0.43, "learning_rate": 1.270717549394909e-06, "loss": 0.8406, "step": 5545 }, { "epoch": 0.43, "learning_rate": 1.270475540212438e-06, "loss": 1.2382, "step": 5546 }, { "epoch": 0.43, "learning_rate": 1.2702335139370489e-06, "loss": 1.1758, "step": 5547 }, { "epoch": 0.43, "learning_rate": 1.2699914705840374e-06, "loss": 1.2484, "step": 5548 }, { "epoch": 0.43, "learning_rate": 1.2697494101686995e-06, "loss": 1.2366, "step": 5549 }, { "epoch": 0.43, "learning_rate": 1.269507332706332e-06, "loss": 1.1939, "step": 5550 }, { "epoch": 0.43, "learning_rate": 1.2692652382122338e-06, "loss": 1.2733, "step": 5551 }, { "epoch": 0.43, "learning_rate": 1.269023126701704e-06, "loss": 1.234, "step": 5552 }, { "epoch": 0.43, "learning_rate": 1.2687809981900426e-06, "loss": 1.2139, "step": 5553 }, { "epoch": 0.43, "learning_rate": 1.268538852692552e-06, "loss": 1.2182, "step": 5554 }, { "epoch": 0.43, "learning_rate": 1.2682966902245339e-06, "loss": 1.2377, "step": 5555 }, { "epoch": 0.43, "learning_rate": 1.2680545108012925e-06, "loss": 1.2239, "step": 5556 }, { "epoch": 0.43, "learning_rate": 1.2678123144381322e-06, "loss": 1.2004, "step": 5557 }, { "epoch": 0.43, "learning_rate": 1.2675701011503587e-06, "loss": 1.1529, "step": 5558 }, { "epoch": 0.43, "learning_rate": 1.2673278709532795e-06, "loss": 1.1931, "step": 5559 }, { "epoch": 0.43, "learning_rate": 1.267085623862202e-06, "loss": 1.2207, "step": 5560 }, { "epoch": 0.43, "learning_rate": 1.2668433598924351e-06, "loss": 1.2139, "step": 5561 }, { "epoch": 0.43, "learning_rate": 1.2666010790592894e-06, "loss": 1.2058, "step": 5562 }, { "epoch": 0.43, "learning_rate": 1.2663587813780756e-06, "loss": 1.2345, "step": 5563 }, { "epoch": 0.43, "learning_rate": 1.266116466864106e-06, "loss": 0.8496, "step": 5564 }, { "epoch": 0.43, "learning_rate": 1.2658741355326938e-06, "loss": 1.192, "step": 5565 }, { "epoch": 0.43, "learning_rate": 1.2656317873991535e-06, "loss": 1.2443, "step": 5566 }, { "epoch": 0.43, "learning_rate": 1.2653894224788004e-06, "loss": 1.2209, "step": 5567 }, { "epoch": 0.43, "learning_rate": 1.265147040786951e-06, "loss": 0.8509, "step": 5568 }, { "epoch": 0.43, "learning_rate": 1.2649046423389225e-06, "loss": 1.2334, "step": 5569 }, { "epoch": 0.43, "learning_rate": 1.2646622271500337e-06, "loss": 0.8577, "step": 5570 }, { "epoch": 0.43, "learning_rate": 1.2644197952356042e-06, "loss": 1.2404, "step": 5571 }, { "epoch": 0.43, "learning_rate": 1.2641773466109548e-06, "loss": 1.2366, "step": 5572 }, { "epoch": 0.43, "learning_rate": 1.2639348812914072e-06, "loss": 1.2173, "step": 5573 }, { "epoch": 0.43, "learning_rate": 1.2636923992922837e-06, "loss": 1.2336, "step": 5574 }, { "epoch": 0.43, "learning_rate": 1.2634499006289088e-06, "loss": 1.1829, "step": 5575 }, { "epoch": 0.43, "learning_rate": 1.2632073853166074e-06, "loss": 0.8488, "step": 5576 }, { "epoch": 0.43, "learning_rate": 1.262964853370705e-06, "loss": 1.2236, "step": 5577 }, { "epoch": 0.43, "learning_rate": 1.2627223048065285e-06, "loss": 1.1949, "step": 5578 }, { "epoch": 0.43, "learning_rate": 1.2624797396394066e-06, "loss": 1.2285, "step": 5579 }, { "epoch": 0.43, "learning_rate": 1.262237157884668e-06, "loss": 1.2205, "step": 5580 }, { "epoch": 0.43, "learning_rate": 1.2619945595576429e-06, "loss": 1.1832, "step": 5581 }, { "epoch": 0.43, "learning_rate": 1.2617519446736626e-06, "loss": 1.2432, "step": 5582 }, { "epoch": 0.43, "learning_rate": 1.261509313248059e-06, "loss": 1.2383, "step": 5583 }, { "epoch": 0.43, "learning_rate": 1.2612666652961657e-06, "loss": 1.1691, "step": 5584 }, { "epoch": 0.43, "learning_rate": 1.261024000833317e-06, "loss": 1.1678, "step": 5585 }, { "epoch": 0.43, "learning_rate": 1.2607813198748482e-06, "loss": 1.221, "step": 5586 }, { "epoch": 0.43, "learning_rate": 1.260538622436096e-06, "loss": 0.803, "step": 5587 }, { "epoch": 0.43, "learning_rate": 1.2602959085323973e-06, "loss": 1.1966, "step": 5588 }, { "epoch": 0.43, "learning_rate": 1.260053178179091e-06, "loss": 0.8476, "step": 5589 }, { "epoch": 0.43, "learning_rate": 1.2598104313915162e-06, "loss": 1.2319, "step": 5590 }, { "epoch": 0.43, "learning_rate": 1.259567668185014e-06, "loss": 1.2539, "step": 5591 }, { "epoch": 0.43, "learning_rate": 1.259324888574926e-06, "loss": 1.2306, "step": 5592 }, { "epoch": 0.43, "learning_rate": 1.2590820925765946e-06, "loss": 1.2482, "step": 5593 }, { "epoch": 0.43, "learning_rate": 1.2588392802053637e-06, "loss": 1.1638, "step": 5594 }, { "epoch": 0.43, "learning_rate": 1.2585964514765774e-06, "loss": 1.1643, "step": 5595 }, { "epoch": 0.43, "learning_rate": 1.2583536064055824e-06, "loss": 1.2198, "step": 5596 }, { "epoch": 0.43, "learning_rate": 1.258110745007725e-06, "loss": 1.2269, "step": 5597 }, { "epoch": 0.43, "learning_rate": 1.2578678672983532e-06, "loss": 1.205, "step": 5598 }, { "epoch": 0.43, "learning_rate": 1.257624973292815e-06, "loss": 0.8905, "step": 5599 }, { "epoch": 0.43, "learning_rate": 1.2573820630064615e-06, "loss": 1.2136, "step": 5600 }, { "epoch": 0.43, "learning_rate": 1.2571391364546431e-06, "loss": 1.2894, "step": 5601 }, { "epoch": 0.43, "learning_rate": 1.2568961936527118e-06, "loss": 1.2346, "step": 5602 }, { "epoch": 0.43, "learning_rate": 1.2566532346160203e-06, "loss": 0.8447, "step": 5603 }, { "epoch": 0.43, "learning_rate": 1.256410259359923e-06, "loss": 1.1988, "step": 5604 }, { "epoch": 0.44, "learning_rate": 1.2561672678997744e-06, "loss": 1.1991, "step": 5605 }, { "epoch": 0.44, "learning_rate": 1.2559242602509308e-06, "loss": 1.1728, "step": 5606 }, { "epoch": 0.44, "learning_rate": 1.2556812364287497e-06, "loss": 1.2076, "step": 5607 }, { "epoch": 0.44, "learning_rate": 1.2554381964485886e-06, "loss": 1.1903, "step": 5608 }, { "epoch": 0.44, "learning_rate": 1.2551951403258066e-06, "loss": 1.2344, "step": 5609 }, { "epoch": 0.44, "learning_rate": 1.2549520680757644e-06, "loss": 0.8251, "step": 5610 }, { "epoch": 0.44, "learning_rate": 1.2547089797138223e-06, "loss": 1.1861, "step": 5611 }, { "epoch": 0.44, "learning_rate": 1.2544658752553432e-06, "loss": 1.1553, "step": 5612 }, { "epoch": 0.44, "learning_rate": 1.2542227547156903e-06, "loss": 1.1955, "step": 5613 }, { "epoch": 0.44, "learning_rate": 1.2539796181102267e-06, "loss": 1.255, "step": 5614 }, { "epoch": 0.44, "learning_rate": 1.2537364654543192e-06, "loss": 0.8104, "step": 5615 }, { "epoch": 0.44, "learning_rate": 1.2534932967633327e-06, "loss": 1.2253, "step": 5616 }, { "epoch": 0.44, "learning_rate": 1.2532501120526353e-06, "loss": 1.2415, "step": 5617 }, { "epoch": 0.44, "learning_rate": 1.253006911337595e-06, "loss": 1.2047, "step": 5618 }, { "epoch": 0.44, "learning_rate": 1.2527636946335808e-06, "loss": 1.2383, "step": 5619 }, { "epoch": 0.44, "learning_rate": 1.2525204619559632e-06, "loss": 1.2095, "step": 5620 }, { "epoch": 0.44, "learning_rate": 1.2522772133201133e-06, "loss": 1.1941, "step": 5621 }, { "epoch": 0.44, "learning_rate": 1.2520339487414036e-06, "loss": 1.2363, "step": 5622 }, { "epoch": 0.44, "learning_rate": 1.2517906682352078e-06, "loss": 1.2362, "step": 5623 }, { "epoch": 0.44, "learning_rate": 1.251547371816899e-06, "loss": 1.1902, "step": 5624 }, { "epoch": 0.44, "learning_rate": 1.2513040595018539e-06, "loss": 1.19, "step": 5625 }, { "epoch": 0.44, "learning_rate": 1.2510607313054476e-06, "loss": 1.2329, "step": 5626 }, { "epoch": 0.44, "learning_rate": 1.2508173872430585e-06, "loss": 1.223, "step": 5627 }, { "epoch": 0.44, "learning_rate": 1.250574027330064e-06, "loss": 1.2542, "step": 5628 }, { "epoch": 0.44, "learning_rate": 1.2503306515818441e-06, "loss": 1.197, "step": 5629 }, { "epoch": 0.44, "learning_rate": 1.2500872600137792e-06, "loss": 1.2526, "step": 5630 }, { "epoch": 0.44, "learning_rate": 1.2498438526412494e-06, "loss": 1.1871, "step": 5631 }, { "epoch": 0.44, "learning_rate": 1.2496004294796388e-06, "loss": 1.2454, "step": 5632 }, { "epoch": 0.44, "learning_rate": 1.249356990544329e-06, "loss": 1.2118, "step": 5633 }, { "epoch": 0.44, "learning_rate": 1.2491135358507056e-06, "loss": 0.8703, "step": 5634 }, { "epoch": 0.44, "learning_rate": 1.2488700654141538e-06, "loss": 1.2534, "step": 5635 }, { "epoch": 0.44, "learning_rate": 1.2486265792500591e-06, "loss": 1.1791, "step": 5636 }, { "epoch": 0.44, "learning_rate": 1.2483830773738095e-06, "loss": 1.1876, "step": 5637 }, { "epoch": 0.44, "learning_rate": 1.248139559800793e-06, "loss": 1.1825, "step": 5638 }, { "epoch": 0.44, "learning_rate": 1.2478960265463989e-06, "loss": 1.2762, "step": 5639 }, { "epoch": 0.44, "learning_rate": 1.2476524776260177e-06, "loss": 1.1781, "step": 5640 }, { "epoch": 0.44, "learning_rate": 1.2474089130550404e-06, "loss": 1.2285, "step": 5641 }, { "epoch": 0.44, "learning_rate": 1.2471653328488593e-06, "loss": 1.1596, "step": 5642 }, { "epoch": 0.44, "learning_rate": 1.2469217370228676e-06, "loss": 1.2072, "step": 5643 }, { "epoch": 0.44, "learning_rate": 1.24667812559246e-06, "loss": 0.8238, "step": 5644 }, { "epoch": 0.44, "learning_rate": 1.2464344985730312e-06, "loss": 1.216, "step": 5645 }, { "epoch": 0.44, "learning_rate": 1.2461908559799775e-06, "loss": 1.2362, "step": 5646 }, { "epoch": 0.44, "learning_rate": 1.2459471978286962e-06, "loss": 1.1643, "step": 5647 }, { "epoch": 0.44, "learning_rate": 1.2457035241345851e-06, "loss": 1.2065, "step": 5648 }, { "epoch": 0.44, "learning_rate": 1.2454598349130443e-06, "loss": 1.2058, "step": 5649 }, { "epoch": 0.44, "learning_rate": 1.2452161301794727e-06, "loss": 1.2291, "step": 5650 }, { "epoch": 0.44, "learning_rate": 1.2449724099492725e-06, "loss": 1.2107, "step": 5651 }, { "epoch": 0.44, "learning_rate": 1.2447286742378453e-06, "loss": 0.8346, "step": 5652 }, { "epoch": 0.44, "learning_rate": 1.244484923060594e-06, "loss": 1.2056, "step": 5653 }, { "epoch": 0.44, "learning_rate": 1.2442411564329227e-06, "loss": 1.2044, "step": 5654 }, { "epoch": 0.44, "learning_rate": 1.2439973743702367e-06, "loss": 1.2364, "step": 5655 }, { "epoch": 0.44, "learning_rate": 1.243753576887942e-06, "loss": 1.1918, "step": 5656 }, { "epoch": 0.44, "learning_rate": 1.2435097640014456e-06, "loss": 1.2311, "step": 5657 }, { "epoch": 0.44, "learning_rate": 1.243265935726155e-06, "loss": 1.1927, "step": 5658 }, { "epoch": 0.44, "learning_rate": 1.2430220920774795e-06, "loss": 1.2171, "step": 5659 }, { "epoch": 0.44, "learning_rate": 1.2427782330708288e-06, "loss": 1.2055, "step": 5660 }, { "epoch": 0.44, "learning_rate": 1.2425343587216141e-06, "loss": 1.2179, "step": 5661 }, { "epoch": 0.44, "learning_rate": 1.242290469045247e-06, "loss": 1.2622, "step": 5662 }, { "epoch": 0.44, "learning_rate": 1.2420465640571403e-06, "loss": 1.2272, "step": 5663 }, { "epoch": 0.44, "learning_rate": 1.2418026437727076e-06, "loss": 1.2079, "step": 5664 }, { "epoch": 0.44, "learning_rate": 1.2415587082073637e-06, "loss": 1.194, "step": 5665 }, { "epoch": 0.44, "learning_rate": 1.241314757376525e-06, "loss": 1.1916, "step": 5666 }, { "epoch": 0.44, "learning_rate": 1.2410707912956074e-06, "loss": 1.224, "step": 5667 }, { "epoch": 0.44, "learning_rate": 1.2408268099800283e-06, "loss": 1.2426, "step": 5668 }, { "epoch": 0.44, "learning_rate": 1.2405828134452071e-06, "loss": 1.2006, "step": 5669 }, { "epoch": 0.44, "learning_rate": 1.2403388017065627e-06, "loss": 1.1482, "step": 5670 }, { "epoch": 0.44, "learning_rate": 1.240094774779516e-06, "loss": 1.2353, "step": 5671 }, { "epoch": 0.44, "learning_rate": 1.2398507326794888e-06, "loss": 1.2472, "step": 5672 }, { "epoch": 0.44, "learning_rate": 1.2396066754219026e-06, "loss": 1.2499, "step": 5673 }, { "epoch": 0.44, "learning_rate": 1.2393626030221818e-06, "loss": 1.2546, "step": 5674 }, { "epoch": 0.44, "learning_rate": 1.2391185154957496e-06, "loss": 1.1892, "step": 5675 }, { "epoch": 0.44, "learning_rate": 1.2388744128580324e-06, "loss": 1.1892, "step": 5676 }, { "epoch": 0.44, "learning_rate": 1.2386302951244556e-06, "loss": 0.8995, "step": 5677 }, { "epoch": 0.44, "learning_rate": 1.2383861623104472e-06, "loss": 1.1741, "step": 5678 }, { "epoch": 0.44, "learning_rate": 1.238142014431435e-06, "loss": 1.2114, "step": 5679 }, { "epoch": 0.44, "learning_rate": 1.237897851502848e-06, "loss": 1.2536, "step": 5680 }, { "epoch": 0.44, "learning_rate": 1.2376536735401165e-06, "loss": 1.1963, "step": 5681 }, { "epoch": 0.44, "learning_rate": 1.2374094805586713e-06, "loss": 0.8343, "step": 5682 }, { "epoch": 0.44, "learning_rate": 1.2371652725739445e-06, "loss": 1.2807, "step": 5683 }, { "epoch": 0.44, "learning_rate": 1.236921049601369e-06, "loss": 1.2254, "step": 5684 }, { "epoch": 0.44, "learning_rate": 1.2366768116563785e-06, "loss": 1.1874, "step": 5685 }, { "epoch": 0.44, "learning_rate": 1.2364325587544085e-06, "loss": 1.2352, "step": 5686 }, { "epoch": 0.44, "learning_rate": 1.2361882909108935e-06, "loss": 1.1516, "step": 5687 }, { "epoch": 0.44, "learning_rate": 1.2359440081412716e-06, "loss": 1.234, "step": 5688 }, { "epoch": 0.44, "learning_rate": 1.2356997104609796e-06, "loss": 1.2374, "step": 5689 }, { "epoch": 0.44, "learning_rate": 1.235455397885456e-06, "loss": 1.1843, "step": 5690 }, { "epoch": 0.44, "learning_rate": 1.2352110704301409e-06, "loss": 1.1996, "step": 5691 }, { "epoch": 0.44, "learning_rate": 1.234966728110474e-06, "loss": 1.2092, "step": 5692 }, { "epoch": 0.44, "learning_rate": 1.2347223709418977e-06, "loss": 1.1958, "step": 5693 }, { "epoch": 0.44, "learning_rate": 1.2344779989398536e-06, "loss": 1.2562, "step": 5694 }, { "epoch": 0.44, "learning_rate": 1.2342336121197852e-06, "loss": 1.2533, "step": 5695 }, { "epoch": 0.44, "learning_rate": 1.2339892104971368e-06, "loss": 1.2385, "step": 5696 }, { "epoch": 0.44, "learning_rate": 1.2337447940873532e-06, "loss": 1.1852, "step": 5697 }, { "epoch": 0.44, "learning_rate": 1.2335003629058812e-06, "loss": 0.8114, "step": 5698 }, { "epoch": 0.44, "learning_rate": 1.2332559169681668e-06, "loss": 1.2032, "step": 5699 }, { "epoch": 0.44, "learning_rate": 1.2330114562896594e-06, "loss": 1.2317, "step": 5700 }, { "epoch": 0.44, "learning_rate": 1.2327669808858065e-06, "loss": 1.2063, "step": 5701 }, { "epoch": 0.44, "learning_rate": 1.2325224907720584e-06, "loss": 1.2232, "step": 5702 }, { "epoch": 0.44, "learning_rate": 1.232277985963866e-06, "loss": 1.2003, "step": 5703 }, { "epoch": 0.44, "learning_rate": 1.232033466476681e-06, "loss": 0.8019, "step": 5704 }, { "epoch": 0.44, "learning_rate": 1.2317889323259557e-06, "loss": 1.1745, "step": 5705 }, { "epoch": 0.44, "learning_rate": 1.2315443835271441e-06, "loss": 1.1794, "step": 5706 }, { "epoch": 0.44, "learning_rate": 1.2312998200956999e-06, "loss": 1.2431, "step": 5707 }, { "epoch": 0.44, "learning_rate": 1.2310552420470794e-06, "loss": 1.1397, "step": 5708 }, { "epoch": 0.44, "learning_rate": 1.230810649396738e-06, "loss": 1.2111, "step": 5709 }, { "epoch": 0.44, "learning_rate": 1.2305660421601339e-06, "loss": 1.2299, "step": 5710 }, { "epoch": 0.44, "learning_rate": 1.2303214203527245e-06, "loss": 1.1904, "step": 5711 }, { "epoch": 0.44, "learning_rate": 1.230076783989969e-06, "loss": 1.141, "step": 5712 }, { "epoch": 0.44, "learning_rate": 1.2298321330873277e-06, "loss": 1.1685, "step": 5713 }, { "epoch": 0.44, "learning_rate": 1.2295874676602611e-06, "loss": 1.2415, "step": 5714 }, { "epoch": 0.44, "learning_rate": 1.2293427877242313e-06, "loss": 1.2283, "step": 5715 }, { "epoch": 0.44, "learning_rate": 1.2290980932947007e-06, "loss": 1.2595, "step": 5716 }, { "epoch": 0.44, "learning_rate": 1.2288533843871338e-06, "loss": 1.2137, "step": 5717 }, { "epoch": 0.44, "learning_rate": 1.2286086610169944e-06, "loss": 1.2292, "step": 5718 }, { "epoch": 0.44, "learning_rate": 1.2283639231997478e-06, "loss": 1.2135, "step": 5719 }, { "epoch": 0.44, "learning_rate": 1.2281191709508614e-06, "loss": 1.199, "step": 5720 }, { "epoch": 0.44, "learning_rate": 1.2278744042858017e-06, "loss": 1.2256, "step": 5721 }, { "epoch": 0.44, "learning_rate": 1.2276296232200372e-06, "loss": 0.8321, "step": 5722 }, { "epoch": 0.44, "learning_rate": 1.2273848277690367e-06, "loss": 1.2481, "step": 5723 }, { "epoch": 0.44, "learning_rate": 1.2271400179482704e-06, "loss": 1.2238, "step": 5724 }, { "epoch": 0.44, "learning_rate": 1.2268951937732099e-06, "loss": 1.1915, "step": 5725 }, { "epoch": 0.44, "learning_rate": 1.2266503552593262e-06, "loss": 0.8709, "step": 5726 }, { "epoch": 0.44, "learning_rate": 1.2264055024220923e-06, "loss": 1.1516, "step": 5727 }, { "epoch": 0.44, "learning_rate": 1.2261606352769824e-06, "loss": 1.1718, "step": 5728 }, { "epoch": 0.44, "learning_rate": 1.2259157538394703e-06, "loss": 1.2143, "step": 5729 }, { "epoch": 0.44, "learning_rate": 1.2256708581250317e-06, "loss": 1.2297, "step": 5730 }, { "epoch": 0.44, "learning_rate": 1.2254259481491433e-06, "loss": 1.2719, "step": 5731 }, { "epoch": 0.44, "learning_rate": 1.2251810239272821e-06, "loss": 0.8588, "step": 5732 }, { "epoch": 0.44, "learning_rate": 1.2249360854749268e-06, "loss": 1.2026, "step": 5733 }, { "epoch": 0.45, "learning_rate": 1.2246911328075554e-06, "loss": 1.2097, "step": 5734 }, { "epoch": 0.45, "learning_rate": 1.2244461659406487e-06, "loss": 1.2162, "step": 5735 }, { "epoch": 0.45, "learning_rate": 1.2242011848896871e-06, "loss": 1.2261, "step": 5736 }, { "epoch": 0.45, "learning_rate": 1.2239561896701533e-06, "loss": 1.2831, "step": 5737 }, { "epoch": 0.45, "learning_rate": 1.223711180297529e-06, "loss": 1.1867, "step": 5738 }, { "epoch": 0.45, "learning_rate": 1.2234661567872981e-06, "loss": 1.2415, "step": 5739 }, { "epoch": 0.45, "learning_rate": 1.2232211191549452e-06, "loss": 1.1597, "step": 5740 }, { "epoch": 0.45, "learning_rate": 1.2229760674159553e-06, "loss": 1.25, "step": 5741 }, { "epoch": 0.45, "learning_rate": 1.2227310015858153e-06, "loss": 1.1471, "step": 5742 }, { "epoch": 0.45, "learning_rate": 1.2224859216800112e-06, "loss": 1.2169, "step": 5743 }, { "epoch": 0.45, "learning_rate": 1.222240827714032e-06, "loss": 1.232, "step": 5744 }, { "epoch": 0.45, "learning_rate": 1.2219957197033666e-06, "loss": 1.1775, "step": 5745 }, { "epoch": 0.45, "learning_rate": 1.2217505976635041e-06, "loss": 1.2579, "step": 5746 }, { "epoch": 0.45, "learning_rate": 1.2215054616099356e-06, "loss": 1.2466, "step": 5747 }, { "epoch": 0.45, "learning_rate": 1.2212603115581528e-06, "loss": 1.263, "step": 5748 }, { "epoch": 0.45, "learning_rate": 1.221015147523648e-06, "loss": 1.244, "step": 5749 }, { "epoch": 0.45, "learning_rate": 1.2207699695219146e-06, "loss": 1.2168, "step": 5750 }, { "epoch": 0.45, "learning_rate": 1.2205247775684466e-06, "loss": 1.2329, "step": 5751 }, { "epoch": 0.45, "learning_rate": 1.2202795716787395e-06, "loss": 1.1296, "step": 5752 }, { "epoch": 0.45, "learning_rate": 1.2200343518682886e-06, "loss": 1.2416, "step": 5753 }, { "epoch": 0.45, "learning_rate": 1.2197891181525916e-06, "loss": 1.2089, "step": 5754 }, { "epoch": 0.45, "learning_rate": 1.2195438705471458e-06, "loss": 1.2588, "step": 5755 }, { "epoch": 0.45, "learning_rate": 1.21929860906745e-06, "loss": 1.2208, "step": 5756 }, { "epoch": 0.45, "learning_rate": 1.2190533337290033e-06, "loss": 1.1659, "step": 5757 }, { "epoch": 0.45, "learning_rate": 1.2188080445473061e-06, "loss": 1.2509, "step": 5758 }, { "epoch": 0.45, "learning_rate": 1.2185627415378605e-06, "loss": 0.8724, "step": 5759 }, { "epoch": 0.45, "learning_rate": 1.2183174247161674e-06, "loss": 1.2041, "step": 5760 }, { "epoch": 0.45, "learning_rate": 1.2180720940977307e-06, "loss": 1.2243, "step": 5761 }, { "epoch": 0.45, "learning_rate": 1.2178267496980544e-06, "loss": 1.2228, "step": 5762 }, { "epoch": 0.45, "learning_rate": 1.2175813915326418e-06, "loss": 1.2054, "step": 5763 }, { "epoch": 0.45, "learning_rate": 1.217336019617e-06, "loss": 1.2012, "step": 5764 }, { "epoch": 0.45, "learning_rate": 1.2170906339666352e-06, "loss": 1.2454, "step": 5765 }, { "epoch": 0.45, "learning_rate": 1.2168452345970544e-06, "loss": 0.8468, "step": 5766 }, { "epoch": 0.45, "learning_rate": 1.2165998215237658e-06, "loss": 1.1648, "step": 5767 }, { "epoch": 0.45, "learning_rate": 1.2163543947622785e-06, "loss": 1.1726, "step": 5768 }, { "epoch": 0.45, "learning_rate": 1.2161089543281026e-06, "loss": 1.2168, "step": 5769 }, { "epoch": 0.45, "learning_rate": 1.2158635002367485e-06, "loss": 1.2354, "step": 5770 }, { "epoch": 0.45, "learning_rate": 1.215618032503729e-06, "loss": 1.1936, "step": 5771 }, { "epoch": 0.45, "learning_rate": 1.2153725511445554e-06, "loss": 1.2174, "step": 5772 }, { "epoch": 0.45, "learning_rate": 1.2151270561747415e-06, "loss": 1.1907, "step": 5773 }, { "epoch": 0.45, "learning_rate": 1.2148815476098016e-06, "loss": 0.8363, "step": 5774 }, { "epoch": 0.45, "learning_rate": 1.2146360254652505e-06, "loss": 1.1679, "step": 5775 }, { "epoch": 0.45, "learning_rate": 1.214390489756605e-06, "loss": 0.8286, "step": 5776 }, { "epoch": 0.45, "learning_rate": 1.2141449404993813e-06, "loss": 1.1958, "step": 5777 }, { "epoch": 0.45, "learning_rate": 1.2138993777090967e-06, "loss": 1.2305, "step": 5778 }, { "epoch": 0.45, "learning_rate": 1.2136538014012708e-06, "loss": 0.8818, "step": 5779 }, { "epoch": 0.45, "learning_rate": 1.2134082115914218e-06, "loss": 1.1838, "step": 5780 }, { "epoch": 0.45, "learning_rate": 1.213162608295071e-06, "loss": 1.2305, "step": 5781 }, { "epoch": 0.45, "learning_rate": 1.2129169915277386e-06, "loss": 1.2184, "step": 5782 }, { "epoch": 0.45, "learning_rate": 1.2126713613049474e-06, "loss": 1.2572, "step": 5783 }, { "epoch": 0.45, "learning_rate": 1.2124257176422198e-06, "loss": 1.2209, "step": 5784 }, { "epoch": 0.45, "learning_rate": 1.212180060555079e-06, "loss": 1.2255, "step": 5785 }, { "epoch": 0.45, "learning_rate": 1.2119343900590499e-06, "loss": 1.1638, "step": 5786 }, { "epoch": 0.45, "learning_rate": 1.2116887061696578e-06, "loss": 1.2135, "step": 5787 }, { "epoch": 0.45, "learning_rate": 1.2114430089024294e-06, "loss": 1.1795, "step": 5788 }, { "epoch": 0.45, "learning_rate": 1.2111972982728912e-06, "loss": 1.2247, "step": 5789 }, { "epoch": 0.45, "learning_rate": 1.2109515742965712e-06, "loss": 0.828, "step": 5790 }, { "epoch": 0.45, "learning_rate": 1.210705836988998e-06, "loss": 1.295, "step": 5791 }, { "epoch": 0.45, "learning_rate": 1.210460086365701e-06, "loss": 1.2289, "step": 5792 }, { "epoch": 0.45, "learning_rate": 1.2102143224422113e-06, "loss": 0.8735, "step": 5793 }, { "epoch": 0.45, "learning_rate": 1.2099685452340593e-06, "loss": 1.2018, "step": 5794 }, { "epoch": 0.45, "learning_rate": 1.2097227547567773e-06, "loss": 0.8579, "step": 5795 }, { "epoch": 0.45, "learning_rate": 1.2094769510258993e-06, "loss": 1.2183, "step": 5796 }, { "epoch": 0.45, "learning_rate": 1.2092311340569573e-06, "loss": 1.2347, "step": 5797 }, { "epoch": 0.45, "learning_rate": 1.2089853038654874e-06, "loss": 1.1914, "step": 5798 }, { "epoch": 0.45, "learning_rate": 1.2087394604670243e-06, "loss": 1.2246, "step": 5799 }, { "epoch": 0.45, "learning_rate": 1.2084936038771038e-06, "loss": 1.2353, "step": 5800 }, { "epoch": 0.45, "learning_rate": 1.2082477341112642e-06, "loss": 1.2299, "step": 5801 }, { "epoch": 0.45, "learning_rate": 1.2080018511850426e-06, "loss": 1.1808, "step": 5802 }, { "epoch": 0.45, "learning_rate": 1.207755955113978e-06, "loss": 1.2234, "step": 5803 }, { "epoch": 0.45, "learning_rate": 1.2075100459136097e-06, "loss": 0.8726, "step": 5804 }, { "epoch": 0.45, "learning_rate": 1.2072641235994792e-06, "loss": 1.1692, "step": 5805 }, { "epoch": 0.45, "learning_rate": 1.2070181881871265e-06, "loss": 0.8546, "step": 5806 }, { "epoch": 0.45, "learning_rate": 1.2067722396920941e-06, "loss": 1.1832, "step": 5807 }, { "epoch": 0.45, "learning_rate": 1.206526278129925e-06, "loss": 1.163, "step": 5808 }, { "epoch": 0.45, "learning_rate": 1.2062803035161628e-06, "loss": 1.1692, "step": 5809 }, { "epoch": 0.45, "learning_rate": 1.2060343158663525e-06, "loss": 0.8779, "step": 5810 }, { "epoch": 0.45, "learning_rate": 1.205788315196039e-06, "loss": 1.1734, "step": 5811 }, { "epoch": 0.45, "learning_rate": 1.2055423015207687e-06, "loss": 1.189, "step": 5812 }, { "epoch": 0.45, "learning_rate": 1.2052962748560887e-06, "loss": 1.2054, "step": 5813 }, { "epoch": 0.45, "learning_rate": 1.2050502352175461e-06, "loss": 1.2235, "step": 5814 }, { "epoch": 0.45, "learning_rate": 1.2048041826206912e-06, "loss": 1.1859, "step": 5815 }, { "epoch": 0.45, "learning_rate": 1.204558117081072e-06, "loss": 1.2008, "step": 5816 }, { "epoch": 0.45, "learning_rate": 1.2043120386142392e-06, "loss": 1.1977, "step": 5817 }, { "epoch": 0.45, "learning_rate": 1.2040659472357444e-06, "loss": 1.1756, "step": 5818 }, { "epoch": 0.45, "learning_rate": 1.2038198429611389e-06, "loss": 1.2189, "step": 5819 }, { "epoch": 0.45, "learning_rate": 1.2035737258059757e-06, "loss": 1.1984, "step": 5820 }, { "epoch": 0.45, "learning_rate": 1.2033275957858086e-06, "loss": 0.889, "step": 5821 }, { "epoch": 0.45, "learning_rate": 1.2030814529161917e-06, "loss": 1.2466, "step": 5822 }, { "epoch": 0.45, "learning_rate": 1.2028352972126802e-06, "loss": 1.2408, "step": 5823 }, { "epoch": 0.45, "learning_rate": 1.2025891286908302e-06, "loss": 1.248, "step": 5824 }, { "epoch": 0.45, "learning_rate": 1.2023429473661988e-06, "loss": 1.1978, "step": 5825 }, { "epoch": 0.45, "learning_rate": 1.202096753254343e-06, "loss": 1.2361, "step": 5826 }, { "epoch": 0.45, "learning_rate": 1.2018505463708218e-06, "loss": 1.219, "step": 5827 }, { "epoch": 0.45, "learning_rate": 1.201604326731194e-06, "loss": 1.2294, "step": 5828 }, { "epoch": 0.45, "learning_rate": 1.20135809435102e-06, "loss": 1.2194, "step": 5829 }, { "epoch": 0.45, "learning_rate": 1.2011118492458608e-06, "loss": 1.176, "step": 5830 }, { "epoch": 0.45, "learning_rate": 1.200865591431277e-06, "loss": 1.1396, "step": 5831 }, { "epoch": 0.45, "learning_rate": 1.2006193209228326e-06, "loss": 1.2093, "step": 5832 }, { "epoch": 0.45, "learning_rate": 1.2003730377360897e-06, "loss": 1.2259, "step": 5833 }, { "epoch": 0.45, "learning_rate": 1.2001267418866125e-06, "loss": 1.1981, "step": 5834 }, { "epoch": 0.45, "learning_rate": 1.1998804333899668e-06, "loss": 1.1285, "step": 5835 }, { "epoch": 0.45, "learning_rate": 1.1996341122617169e-06, "loss": 1.2413, "step": 5836 }, { "epoch": 0.45, "learning_rate": 1.1993877785174302e-06, "loss": 1.196, "step": 5837 }, { "epoch": 0.45, "learning_rate": 1.199141432172674e-06, "loss": 1.2723, "step": 5838 }, { "epoch": 0.45, "learning_rate": 1.1988950732430153e-06, "loss": 1.2174, "step": 5839 }, { "epoch": 0.45, "learning_rate": 1.1986487017440244e-06, "loss": 0.8644, "step": 5840 }, { "epoch": 0.45, "learning_rate": 1.1984023176912698e-06, "loss": 1.2026, "step": 5841 }, { "epoch": 0.45, "learning_rate": 1.1981559211003227e-06, "loss": 0.824, "step": 5842 }, { "epoch": 0.45, "learning_rate": 1.197909511986754e-06, "loss": 1.2048, "step": 5843 }, { "epoch": 0.45, "learning_rate": 1.1976630903661356e-06, "loss": 1.1587, "step": 5844 }, { "epoch": 0.45, "learning_rate": 1.1974166562540405e-06, "loss": 1.1996, "step": 5845 }, { "epoch": 0.45, "learning_rate": 1.1971702096660417e-06, "loss": 1.2004, "step": 5846 }, { "epoch": 0.45, "learning_rate": 1.1969237506177148e-06, "loss": 1.222, "step": 5847 }, { "epoch": 0.45, "learning_rate": 1.1966772791246339e-06, "loss": 1.1797, "step": 5848 }, { "epoch": 0.45, "learning_rate": 1.1964307952023758e-06, "loss": 1.1893, "step": 5849 }, { "epoch": 0.45, "learning_rate": 1.1961842988665165e-06, "loss": 1.2425, "step": 5850 }, { "epoch": 0.45, "learning_rate": 1.1959377901326336e-06, "loss": 0.8508, "step": 5851 }, { "epoch": 0.45, "learning_rate": 1.1956912690163063e-06, "loss": 1.2311, "step": 5852 }, { "epoch": 0.45, "learning_rate": 1.1954447355331126e-06, "loss": 1.2238, "step": 5853 }, { "epoch": 0.45, "learning_rate": 1.1951981896986327e-06, "loss": 0.8486, "step": 5854 }, { "epoch": 0.45, "learning_rate": 1.194951631528448e-06, "loss": 1.2223, "step": 5855 }, { "epoch": 0.45, "learning_rate": 1.194705061038139e-06, "loss": 1.1503, "step": 5856 }, { "epoch": 0.45, "learning_rate": 1.1944584782432883e-06, "loss": 1.1709, "step": 5857 }, { "epoch": 0.45, "learning_rate": 1.1942118831594785e-06, "loss": 1.2309, "step": 5858 }, { "epoch": 0.45, "learning_rate": 1.193965275802294e-06, "loss": 0.8239, "step": 5859 }, { "epoch": 0.45, "learning_rate": 1.1937186561873195e-06, "loss": 1.2653, "step": 5860 }, { "epoch": 0.45, "learning_rate": 1.193472024330139e-06, "loss": 1.2131, "step": 5861 }, { "epoch": 0.45, "learning_rate": 1.19322538024634e-06, "loss": 1.233, "step": 5862 }, { "epoch": 0.46, "learning_rate": 1.1929787239515088e-06, "loss": 1.2315, "step": 5863 }, { "epoch": 0.46, "learning_rate": 1.1927320554612335e-06, "loss": 1.179, "step": 5864 }, { "epoch": 0.46, "learning_rate": 1.1924853747911016e-06, "loss": 1.2104, "step": 5865 }, { "epoch": 0.46, "learning_rate": 1.1922386819567032e-06, "loss": 1.1832, "step": 5866 }, { "epoch": 0.46, "learning_rate": 1.1919919769736276e-06, "loss": 1.2469, "step": 5867 }, { "epoch": 0.46, "learning_rate": 1.1917452598574656e-06, "loss": 0.8405, "step": 5868 }, { "epoch": 0.46, "learning_rate": 1.1914985306238092e-06, "loss": 1.1886, "step": 5869 }, { "epoch": 0.46, "learning_rate": 1.19125178928825e-06, "loss": 1.1754, "step": 5870 }, { "epoch": 0.46, "learning_rate": 1.1910050358663818e-06, "loss": 0.8735, "step": 5871 }, { "epoch": 0.46, "learning_rate": 1.1907582703737977e-06, "loss": 1.2611, "step": 5872 }, { "epoch": 0.46, "learning_rate": 1.1905114928260922e-06, "loss": 1.1798, "step": 5873 }, { "epoch": 0.46, "learning_rate": 1.190264703238861e-06, "loss": 0.8591, "step": 5874 }, { "epoch": 0.46, "learning_rate": 1.1900179016277e-06, "loss": 1.226, "step": 5875 }, { "epoch": 0.46, "learning_rate": 1.1897710880082058e-06, "loss": 1.1826, "step": 5876 }, { "epoch": 0.46, "learning_rate": 1.1895242623959768e-06, "loss": 1.1792, "step": 5877 }, { "epoch": 0.46, "learning_rate": 1.1892774248066105e-06, "loss": 1.1907, "step": 5878 }, { "epoch": 0.46, "learning_rate": 1.189030575255706e-06, "loss": 1.2419, "step": 5879 }, { "epoch": 0.46, "learning_rate": 1.1887837137588634e-06, "loss": 1.2474, "step": 5880 }, { "epoch": 0.46, "learning_rate": 1.1885368403316839e-06, "loss": 1.1743, "step": 5881 }, { "epoch": 0.46, "learning_rate": 1.1882899549897678e-06, "loss": 1.1627, "step": 5882 }, { "epoch": 0.46, "learning_rate": 1.188043057748718e-06, "loss": 1.1906, "step": 5883 }, { "epoch": 0.46, "learning_rate": 1.1877961486241367e-06, "loss": 1.2425, "step": 5884 }, { "epoch": 0.46, "learning_rate": 1.187549227631628e-06, "loss": 1.2198, "step": 5885 }, { "epoch": 0.46, "learning_rate": 1.1873022947867965e-06, "loss": 1.2237, "step": 5886 }, { "epoch": 0.46, "learning_rate": 1.1870553501052469e-06, "loss": 1.2147, "step": 5887 }, { "epoch": 0.46, "learning_rate": 1.1868083936025848e-06, "loss": 1.243, "step": 5888 }, { "epoch": 0.46, "learning_rate": 1.1865614252944176e-06, "loss": 1.2786, "step": 5889 }, { "epoch": 0.46, "learning_rate": 1.1863144451963515e-06, "loss": 1.1549, "step": 5890 }, { "epoch": 0.46, "learning_rate": 1.1860674533239962e-06, "loss": 1.2326, "step": 5891 }, { "epoch": 0.46, "learning_rate": 1.185820449692959e-06, "loss": 1.2394, "step": 5892 }, { "epoch": 0.46, "learning_rate": 1.1855734343188506e-06, "loss": 1.2051, "step": 5893 }, { "epoch": 0.46, "learning_rate": 1.1853264072172808e-06, "loss": 1.2202, "step": 5894 }, { "epoch": 0.46, "learning_rate": 1.1850793684038607e-06, "loss": 1.1942, "step": 5895 }, { "epoch": 0.46, "learning_rate": 1.1848323178942022e-06, "loss": 1.1692, "step": 5896 }, { "epoch": 0.46, "learning_rate": 1.1845852557039177e-06, "loss": 1.2031, "step": 5897 }, { "epoch": 0.46, "learning_rate": 1.1843381818486209e-06, "loss": 1.1597, "step": 5898 }, { "epoch": 0.46, "learning_rate": 1.1840910963439254e-06, "loss": 1.2199, "step": 5899 }, { "epoch": 0.46, "learning_rate": 1.183843999205446e-06, "loss": 1.2325, "step": 5900 }, { "epoch": 0.46, "learning_rate": 1.1835968904487986e-06, "loss": 1.2159, "step": 5901 }, { "epoch": 0.46, "learning_rate": 1.1833497700895988e-06, "loss": 1.2003, "step": 5902 }, { "epoch": 0.46, "learning_rate": 1.1831026381434645e-06, "loss": 0.8209, "step": 5903 }, { "epoch": 0.46, "learning_rate": 1.1828554946260125e-06, "loss": 1.1786, "step": 5904 }, { "epoch": 0.46, "learning_rate": 1.1826083395528614e-06, "loss": 1.2662, "step": 5905 }, { "epoch": 0.46, "learning_rate": 1.182361172939631e-06, "loss": 1.2409, "step": 5906 }, { "epoch": 0.46, "learning_rate": 1.1821139948019403e-06, "loss": 1.2295, "step": 5907 }, { "epoch": 0.46, "learning_rate": 1.1818668051554107e-06, "loss": 0.8463, "step": 5908 }, { "epoch": 0.46, "learning_rate": 1.181619604015663e-06, "loss": 1.2151, "step": 5909 }, { "epoch": 0.46, "learning_rate": 1.181372391398319e-06, "loss": 1.1978, "step": 5910 }, { "epoch": 0.46, "learning_rate": 1.1811251673190027e-06, "loss": 1.2031, "step": 5911 }, { "epoch": 0.46, "learning_rate": 1.1808779317933364e-06, "loss": 1.2023, "step": 5912 }, { "epoch": 0.46, "learning_rate": 1.1806306848369449e-06, "loss": 1.1957, "step": 5913 }, { "epoch": 0.46, "learning_rate": 1.1803834264654533e-06, "loss": 1.2007, "step": 5914 }, { "epoch": 0.46, "learning_rate": 1.180136156694487e-06, "loss": 1.2365, "step": 5915 }, { "epoch": 0.46, "learning_rate": 1.1798888755396725e-06, "loss": 1.2081, "step": 5916 }, { "epoch": 0.46, "learning_rate": 1.1796415830166369e-06, "loss": 1.1799, "step": 5917 }, { "epoch": 0.46, "learning_rate": 1.179394279141008e-06, "loss": 1.2576, "step": 5918 }, { "epoch": 0.46, "learning_rate": 1.1791469639284145e-06, "loss": 1.1242, "step": 5919 }, { "epoch": 0.46, "learning_rate": 1.1788996373944858e-06, "loss": 1.1926, "step": 5920 }, { "epoch": 0.46, "learning_rate": 1.1786522995548516e-06, "loss": 1.2061, "step": 5921 }, { "epoch": 0.46, "learning_rate": 1.1784049504251425e-06, "loss": 1.2127, "step": 5922 }, { "epoch": 0.46, "learning_rate": 1.1781575900209906e-06, "loss": 1.1227, "step": 5923 }, { "epoch": 0.46, "learning_rate": 1.177910218358027e-06, "loss": 1.2328, "step": 5924 }, { "epoch": 0.46, "learning_rate": 1.1776628354518859e-06, "loss": 1.2035, "step": 5925 }, { "epoch": 0.46, "learning_rate": 1.1774154413181995e-06, "loss": 1.219, "step": 5926 }, { "epoch": 0.46, "learning_rate": 1.1771680359726027e-06, "loss": 1.2546, "step": 5927 }, { "epoch": 0.46, "learning_rate": 1.1769206194307311e-06, "loss": 1.1986, "step": 5928 }, { "epoch": 0.46, "learning_rate": 1.1766731917082188e-06, "loss": 0.8484, "step": 5929 }, { "epoch": 0.46, "learning_rate": 1.1764257528207037e-06, "loss": 1.2461, "step": 5930 }, { "epoch": 0.46, "learning_rate": 1.1761783027838226e-06, "loss": 1.2098, "step": 5931 }, { "epoch": 0.46, "learning_rate": 1.1759308416132123e-06, "loss": 1.2501, "step": 5932 }, { "epoch": 0.46, "learning_rate": 1.1756833693245127e-06, "loss": 1.2459, "step": 5933 }, { "epoch": 0.46, "learning_rate": 1.175435885933362e-06, "loss": 1.1942, "step": 5934 }, { "epoch": 0.46, "learning_rate": 1.1751883914554004e-06, "loss": 1.2133, "step": 5935 }, { "epoch": 0.46, "learning_rate": 1.1749408859062687e-06, "loss": 1.173, "step": 5936 }, { "epoch": 0.46, "learning_rate": 1.174693369301608e-06, "loss": 1.2173, "step": 5937 }, { "epoch": 0.46, "learning_rate": 1.1744458416570607e-06, "loss": 1.1903, "step": 5938 }, { "epoch": 0.46, "learning_rate": 1.1741983029882688e-06, "loss": 0.8233, "step": 5939 }, { "epoch": 0.46, "learning_rate": 1.1739507533108764e-06, "loss": 1.2709, "step": 5940 }, { "epoch": 0.46, "learning_rate": 1.173703192640527e-06, "loss": 1.2011, "step": 5941 }, { "epoch": 0.46, "learning_rate": 1.1734556209928662e-06, "loss": 1.2356, "step": 5942 }, { "epoch": 0.46, "learning_rate": 1.1732080383835386e-06, "loss": 1.185, "step": 5943 }, { "epoch": 0.46, "learning_rate": 1.1729604448281904e-06, "loss": 0.8298, "step": 5944 }, { "epoch": 0.46, "learning_rate": 1.1727128403424696e-06, "loss": 1.2547, "step": 5945 }, { "epoch": 0.46, "learning_rate": 1.1724652249420225e-06, "loss": 1.202, "step": 5946 }, { "epoch": 0.46, "learning_rate": 1.172217598642498e-06, "loss": 1.1904, "step": 5947 }, { "epoch": 0.46, "learning_rate": 1.1719699614595452e-06, "loss": 1.2355, "step": 5948 }, { "epoch": 0.46, "learning_rate": 1.1717223134088127e-06, "loss": 1.2095, "step": 5949 }, { "epoch": 0.46, "learning_rate": 1.1714746545059521e-06, "loss": 1.2135, "step": 5950 }, { "epoch": 0.46, "learning_rate": 1.1712269847666137e-06, "loss": 1.1889, "step": 5951 }, { "epoch": 0.46, "learning_rate": 1.1709793042064492e-06, "loss": 1.2166, "step": 5952 }, { "epoch": 0.46, "learning_rate": 1.1707316128411115e-06, "loss": 1.2072, "step": 5953 }, { "epoch": 0.46, "learning_rate": 1.1704839106862527e-06, "loss": 1.1705, "step": 5954 }, { "epoch": 0.46, "learning_rate": 1.1702361977575276e-06, "loss": 1.2186, "step": 5955 }, { "epoch": 0.46, "learning_rate": 1.1699884740705897e-06, "loss": 1.2262, "step": 5956 }, { "epoch": 0.46, "learning_rate": 1.1697407396410948e-06, "loss": 1.1972, "step": 5957 }, { "epoch": 0.46, "learning_rate": 1.1694929944846979e-06, "loss": 1.2278, "step": 5958 }, { "epoch": 0.46, "learning_rate": 1.1692452386170567e-06, "loss": 1.2226, "step": 5959 }, { "epoch": 0.46, "learning_rate": 1.168997472053827e-06, "loss": 1.2405, "step": 5960 }, { "epoch": 0.46, "learning_rate": 1.168749694810667e-06, "loss": 1.1964, "step": 5961 }, { "epoch": 0.46, "learning_rate": 1.168501906903236e-06, "loss": 1.2303, "step": 5962 }, { "epoch": 0.46, "learning_rate": 1.1682541083471917e-06, "loss": 1.1871, "step": 5963 }, { "epoch": 0.46, "learning_rate": 1.1680062991581954e-06, "loss": 1.2055, "step": 5964 }, { "epoch": 0.46, "learning_rate": 1.167758479351907e-06, "loss": 1.2366, "step": 5965 }, { "epoch": 0.46, "learning_rate": 1.167510648943987e-06, "loss": 1.2536, "step": 5966 }, { "epoch": 0.46, "learning_rate": 1.1672628079500983e-06, "loss": 1.2112, "step": 5967 }, { "epoch": 0.46, "learning_rate": 1.1670149563859023e-06, "loss": 1.188, "step": 5968 }, { "epoch": 0.46, "learning_rate": 1.1667670942670636e-06, "loss": 1.2261, "step": 5969 }, { "epoch": 0.46, "learning_rate": 1.1665192216092453e-06, "loss": 1.2041, "step": 5970 }, { "epoch": 0.46, "learning_rate": 1.1662713384281112e-06, "loss": 1.2244, "step": 5971 }, { "epoch": 0.46, "learning_rate": 1.1660234447393273e-06, "loss": 1.2196, "step": 5972 }, { "epoch": 0.46, "learning_rate": 1.1657755405585597e-06, "loss": 1.2059, "step": 5973 }, { "epoch": 0.46, "learning_rate": 1.1655276259014741e-06, "loss": 1.2227, "step": 5974 }, { "epoch": 0.46, "learning_rate": 1.1652797007837383e-06, "loss": 1.1895, "step": 5975 }, { "epoch": 0.46, "learning_rate": 1.16503176522102e-06, "loss": 1.1706, "step": 5976 }, { "epoch": 0.46, "learning_rate": 1.1647838192289874e-06, "loss": 1.2508, "step": 5977 }, { "epoch": 0.46, "learning_rate": 1.1645358628233096e-06, "loss": 0.8779, "step": 5978 }, { "epoch": 0.46, "learning_rate": 1.164287896019657e-06, "loss": 1.2425, "step": 5979 }, { "epoch": 0.46, "learning_rate": 1.1640399188336993e-06, "loss": 0.8243, "step": 5980 }, { "epoch": 0.46, "learning_rate": 1.1637919312811082e-06, "loss": 0.8573, "step": 5981 }, { "epoch": 0.46, "learning_rate": 1.1635439333775554e-06, "loss": 1.2298, "step": 5982 }, { "epoch": 0.46, "learning_rate": 1.163295925138713e-06, "loss": 1.182, "step": 5983 }, { "epoch": 0.46, "learning_rate": 1.1630479065802545e-06, "loss": 1.2179, "step": 5984 }, { "epoch": 0.46, "learning_rate": 1.1627998777178532e-06, "loss": 0.8501, "step": 5985 }, { "epoch": 0.46, "learning_rate": 1.1625518385671837e-06, "loss": 1.1974, "step": 5986 }, { "epoch": 0.46, "learning_rate": 1.1623037891439214e-06, "loss": 1.2507, "step": 5987 }, { "epoch": 0.46, "learning_rate": 1.1620557294637409e-06, "loss": 1.2119, "step": 5988 }, { "epoch": 0.46, "learning_rate": 1.1618076595423196e-06, "loss": 1.1564, "step": 5989 }, { "epoch": 0.46, "learning_rate": 1.161559579395334e-06, "loss": 1.18, "step": 5990 }, { "epoch": 0.46, "learning_rate": 1.161311489038462e-06, "loss": 1.2399, "step": 5991 }, { "epoch": 0.47, "learning_rate": 1.1610633884873816e-06, "loss": 1.1612, "step": 5992 }, { "epoch": 0.47, "learning_rate": 1.1608152777577718e-06, "loss": 1.2338, "step": 5993 }, { "epoch": 0.47, "learning_rate": 1.1605671568653122e-06, "loss": 1.2701, "step": 5994 }, { "epoch": 0.47, "learning_rate": 1.1603190258256826e-06, "loss": 0.8727, "step": 5995 }, { "epoch": 0.47, "learning_rate": 1.1600708846545648e-06, "loss": 1.2252, "step": 5996 }, { "epoch": 0.47, "learning_rate": 1.1598227333676395e-06, "loss": 1.2167, "step": 5997 }, { "epoch": 0.47, "learning_rate": 1.1595745719805887e-06, "loss": 1.1698, "step": 5998 }, { "epoch": 0.47, "learning_rate": 1.159326400509096e-06, "loss": 1.2101, "step": 5999 }, { "epoch": 0.47, "learning_rate": 1.1590782189688435e-06, "loss": 1.2436, "step": 6000 }, { "epoch": 0.47, "learning_rate": 1.1588300273755168e-06, "loss": 1.1845, "step": 6001 }, { "epoch": 0.47, "learning_rate": 1.158581825744799e-06, "loss": 1.2088, "step": 6002 }, { "epoch": 0.47, "learning_rate": 1.1583336140923765e-06, "loss": 1.2156, "step": 6003 }, { "epoch": 0.47, "learning_rate": 1.1580853924339351e-06, "loss": 0.8421, "step": 6004 }, { "epoch": 0.47, "learning_rate": 1.1578371607851606e-06, "loss": 1.1848, "step": 6005 }, { "epoch": 0.47, "learning_rate": 1.1575889191617412e-06, "loss": 0.8783, "step": 6006 }, { "epoch": 0.47, "learning_rate": 1.1573406675793641e-06, "loss": 1.2026, "step": 6007 }, { "epoch": 0.47, "learning_rate": 1.1570924060537179e-06, "loss": 1.2553, "step": 6008 }, { "epoch": 0.47, "learning_rate": 1.156844134600492e-06, "loss": 1.2001, "step": 6009 }, { "epoch": 0.47, "learning_rate": 1.1565958532353754e-06, "loss": 1.1704, "step": 6010 }, { "epoch": 0.47, "learning_rate": 1.156347561974059e-06, "loss": 1.2189, "step": 6011 }, { "epoch": 0.47, "learning_rate": 1.1560992608322335e-06, "loss": 1.2216, "step": 6012 }, { "epoch": 0.47, "learning_rate": 1.155850949825591e-06, "loss": 1.2614, "step": 6013 }, { "epoch": 0.47, "learning_rate": 1.1556026289698233e-06, "loss": 1.2102, "step": 6014 }, { "epoch": 0.47, "learning_rate": 1.1553542982806229e-06, "loss": 1.2345, "step": 6015 }, { "epoch": 0.47, "learning_rate": 1.155105957773684e-06, "loss": 1.213, "step": 6016 }, { "epoch": 0.47, "learning_rate": 1.1548576074646999e-06, "loss": 1.2244, "step": 6017 }, { "epoch": 0.47, "learning_rate": 1.154609247369366e-06, "loss": 0.8589, "step": 6018 }, { "epoch": 0.47, "learning_rate": 1.1543608775033773e-06, "loss": 1.2327, "step": 6019 }, { "epoch": 0.47, "learning_rate": 1.1541124978824294e-06, "loss": 1.2384, "step": 6020 }, { "epoch": 0.47, "learning_rate": 1.1538641085222196e-06, "loss": 1.1792, "step": 6021 }, { "epoch": 0.47, "learning_rate": 1.1536157094384441e-06, "loss": 1.2497, "step": 6022 }, { "epoch": 0.47, "learning_rate": 1.1533673006468018e-06, "loss": 1.1769, "step": 6023 }, { "epoch": 0.47, "learning_rate": 1.1531188821629902e-06, "loss": 0.8093, "step": 6024 }, { "epoch": 0.47, "learning_rate": 1.152870454002709e-06, "loss": 1.2277, "step": 6025 }, { "epoch": 0.47, "learning_rate": 1.1526220161816572e-06, "loss": 1.2781, "step": 6026 }, { "epoch": 0.47, "learning_rate": 1.1523735687155351e-06, "loss": 1.2486, "step": 6027 }, { "epoch": 0.47, "learning_rate": 1.1521251116200439e-06, "loss": 0.8205, "step": 6028 }, { "epoch": 0.47, "learning_rate": 1.1518766449108847e-06, "loss": 1.233, "step": 6029 }, { "epoch": 0.47, "learning_rate": 1.1516281686037598e-06, "loss": 1.189, "step": 6030 }, { "epoch": 0.47, "learning_rate": 1.151379682714372e-06, "loss": 0.8458, "step": 6031 }, { "epoch": 0.47, "learning_rate": 1.1511311872584243e-06, "loss": 1.2265, "step": 6032 }, { "epoch": 0.47, "learning_rate": 1.1508826822516202e-06, "loss": 0.8447, "step": 6033 }, { "epoch": 0.47, "learning_rate": 1.1506341677096645e-06, "loss": 1.205, "step": 6034 }, { "epoch": 0.47, "learning_rate": 1.1503856436482628e-06, "loss": 0.8427, "step": 6035 }, { "epoch": 0.47, "learning_rate": 1.1501371100831201e-06, "loss": 1.2068, "step": 6036 }, { "epoch": 0.47, "learning_rate": 1.1498885670299427e-06, "loss": 1.2385, "step": 6037 }, { "epoch": 0.47, "learning_rate": 1.1496400145044383e-06, "loss": 1.2099, "step": 6038 }, { "epoch": 0.47, "learning_rate": 1.149391452522313e-06, "loss": 1.2733, "step": 6039 }, { "epoch": 0.47, "learning_rate": 1.1491428810992758e-06, "loss": 1.2456, "step": 6040 }, { "epoch": 0.47, "learning_rate": 1.1488943002510354e-06, "loss": 1.2858, "step": 6041 }, { "epoch": 0.47, "learning_rate": 1.148645709993301e-06, "loss": 1.1684, "step": 6042 }, { "epoch": 0.47, "learning_rate": 1.1483971103417822e-06, "loss": 1.2182, "step": 6043 }, { "epoch": 0.47, "learning_rate": 1.1481485013121892e-06, "loss": 1.2568, "step": 6044 }, { "epoch": 0.47, "learning_rate": 1.1478998829202336e-06, "loss": 1.2154, "step": 6045 }, { "epoch": 0.47, "learning_rate": 1.1476512551816267e-06, "loss": 1.2479, "step": 6046 }, { "epoch": 0.47, "learning_rate": 1.1474026181120813e-06, "loss": 1.2474, "step": 6047 }, { "epoch": 0.47, "learning_rate": 1.1471539717273095e-06, "loss": 1.2004, "step": 6048 }, { "epoch": 0.47, "learning_rate": 1.146905316043025e-06, "loss": 1.2125, "step": 6049 }, { "epoch": 0.47, "learning_rate": 1.146656651074942e-06, "loss": 1.2389, "step": 6050 }, { "epoch": 0.47, "learning_rate": 1.1464079768387747e-06, "loss": 1.2004, "step": 6051 }, { "epoch": 0.47, "learning_rate": 1.1461592933502386e-06, "loss": 1.2563, "step": 6052 }, { "epoch": 0.47, "learning_rate": 1.1459106006250491e-06, "loss": 1.2091, "step": 6053 }, { "epoch": 0.47, "learning_rate": 1.145661898678923e-06, "loss": 1.2077, "step": 6054 }, { "epoch": 0.47, "learning_rate": 1.1454131875275768e-06, "loss": 1.2658, "step": 6055 }, { "epoch": 0.47, "learning_rate": 1.1451644671867278e-06, "loss": 1.2254, "step": 6056 }, { "epoch": 0.47, "learning_rate": 1.1449157376720949e-06, "loss": 1.2322, "step": 6057 }, { "epoch": 0.47, "learning_rate": 1.1446669989993961e-06, "loss": 1.2173, "step": 6058 }, { "epoch": 0.47, "learning_rate": 1.1444182511843508e-06, "loss": 1.1699, "step": 6059 }, { "epoch": 0.47, "learning_rate": 1.1441694942426793e-06, "loss": 1.2022, "step": 6060 }, { "epoch": 0.47, "learning_rate": 1.1439207281901009e-06, "loss": 1.1945, "step": 6061 }, { "epoch": 0.47, "learning_rate": 1.1436719530423374e-06, "loss": 1.1678, "step": 6062 }, { "epoch": 0.47, "learning_rate": 1.1434231688151099e-06, "loss": 1.2176, "step": 6063 }, { "epoch": 0.47, "learning_rate": 1.1431743755241415e-06, "loss": 1.1706, "step": 6064 }, { "epoch": 0.47, "learning_rate": 1.1429255731851538e-06, "loss": 1.2107, "step": 6065 }, { "epoch": 0.47, "learning_rate": 1.14267676181387e-06, "loss": 1.2145, "step": 6066 }, { "epoch": 0.47, "learning_rate": 1.1424279414260149e-06, "loss": 1.2479, "step": 6067 }, { "epoch": 0.47, "learning_rate": 1.1421791120373119e-06, "loss": 1.2045, "step": 6068 }, { "epoch": 0.47, "learning_rate": 1.141930273663487e-06, "loss": 1.1736, "step": 6069 }, { "epoch": 0.47, "learning_rate": 1.141681426320265e-06, "loss": 1.203, "step": 6070 }, { "epoch": 0.47, "learning_rate": 1.1414325700233717e-06, "loss": 1.2074, "step": 6071 }, { "epoch": 0.47, "learning_rate": 1.1411837047885349e-06, "loss": 1.1555, "step": 6072 }, { "epoch": 0.47, "learning_rate": 1.1409348306314805e-06, "loss": 1.1814, "step": 6073 }, { "epoch": 0.47, "learning_rate": 1.1406859475679376e-06, "loss": 1.2306, "step": 6074 }, { "epoch": 0.47, "learning_rate": 1.140437055613634e-06, "loss": 0.7768, "step": 6075 }, { "epoch": 0.47, "learning_rate": 1.140188154784298e-06, "loss": 0.7889, "step": 6076 }, { "epoch": 0.47, "learning_rate": 1.1399392450956601e-06, "loss": 1.2113, "step": 6077 }, { "epoch": 0.47, "learning_rate": 1.1396903265634496e-06, "loss": 1.173, "step": 6078 }, { "epoch": 0.47, "learning_rate": 1.1394413992033978e-06, "loss": 1.2048, "step": 6079 }, { "epoch": 0.47, "learning_rate": 1.1391924630312353e-06, "loss": 1.1996, "step": 6080 }, { "epoch": 0.47, "learning_rate": 1.138943518062694e-06, "loss": 0.7871, "step": 6081 }, { "epoch": 0.47, "learning_rate": 1.138694564313506e-06, "loss": 1.1295, "step": 6082 }, { "epoch": 0.47, "learning_rate": 1.1384456017994044e-06, "loss": 1.222, "step": 6083 }, { "epoch": 0.47, "learning_rate": 1.1381966305361225e-06, "loss": 1.2061, "step": 6084 }, { "epoch": 0.47, "learning_rate": 1.1379476505393942e-06, "loss": 1.2059, "step": 6085 }, { "epoch": 0.47, "learning_rate": 1.1376986618249542e-06, "loss": 1.2176, "step": 6086 }, { "epoch": 0.47, "learning_rate": 1.1374496644085374e-06, "loss": 1.1708, "step": 6087 }, { "epoch": 0.47, "learning_rate": 1.1372006583058788e-06, "loss": 0.8632, "step": 6088 }, { "epoch": 0.47, "learning_rate": 1.1369516435327159e-06, "loss": 1.2653, "step": 6089 }, { "epoch": 0.47, "learning_rate": 1.1367026201047837e-06, "loss": 1.1949, "step": 6090 }, { "epoch": 0.47, "learning_rate": 1.136453588037821e-06, "loss": 1.2144, "step": 6091 }, { "epoch": 0.47, "learning_rate": 1.1362045473475648e-06, "loss": 1.1773, "step": 6092 }, { "epoch": 0.47, "learning_rate": 1.1359554980497532e-06, "loss": 1.2714, "step": 6093 }, { "epoch": 0.47, "learning_rate": 1.1357064401601257e-06, "loss": 1.2135, "step": 6094 }, { "epoch": 0.47, "learning_rate": 1.1354573736944211e-06, "loss": 1.1899, "step": 6095 }, { "epoch": 0.47, "learning_rate": 1.1352082986683797e-06, "loss": 1.2321, "step": 6096 }, { "epoch": 0.47, "learning_rate": 1.1349592150977422e-06, "loss": 1.2136, "step": 6097 }, { "epoch": 0.47, "learning_rate": 1.1347101229982493e-06, "loss": 1.2371, "step": 6098 }, { "epoch": 0.47, "learning_rate": 1.1344610223856426e-06, "loss": 1.198, "step": 6099 }, { "epoch": 0.47, "learning_rate": 1.1342119132756641e-06, "loss": 1.196, "step": 6100 }, { "epoch": 0.47, "learning_rate": 1.1339627956840567e-06, "loss": 1.1766, "step": 6101 }, { "epoch": 0.47, "learning_rate": 1.1337136696265638e-06, "loss": 0.855, "step": 6102 }, { "epoch": 0.47, "learning_rate": 1.1334645351189285e-06, "loss": 1.2172, "step": 6103 }, { "epoch": 0.47, "learning_rate": 1.1332153921768954e-06, "loss": 1.1681, "step": 6104 }, { "epoch": 0.47, "learning_rate": 1.1329662408162091e-06, "loss": 1.2456, "step": 6105 }, { "epoch": 0.47, "learning_rate": 1.1327170810526156e-06, "loss": 1.2031, "step": 6106 }, { "epoch": 0.47, "learning_rate": 1.1324679129018597e-06, "loss": 1.2071, "step": 6107 }, { "epoch": 0.47, "learning_rate": 1.1322187363796886e-06, "loss": 1.1736, "step": 6108 }, { "epoch": 0.47, "learning_rate": 1.1319695515018491e-06, "loss": 1.2331, "step": 6109 }, { "epoch": 0.47, "learning_rate": 1.131720358284088e-06, "loss": 1.2152, "step": 6110 }, { "epoch": 0.47, "learning_rate": 1.1314711567421542e-06, "loss": 1.1957, "step": 6111 }, { "epoch": 0.47, "learning_rate": 1.131221946891795e-06, "loss": 1.1543, "step": 6112 }, { "epoch": 0.47, "learning_rate": 1.1309727287487606e-06, "loss": 1.1752, "step": 6113 }, { "epoch": 0.47, "learning_rate": 1.1307235023288004e-06, "loss": 1.2055, "step": 6114 }, { "epoch": 0.47, "learning_rate": 1.1304742676476636e-06, "loss": 1.2351, "step": 6115 }, { "epoch": 0.47, "learning_rate": 1.1302250247211015e-06, "loss": 0.8038, "step": 6116 }, { "epoch": 0.47, "learning_rate": 1.129975773564865e-06, "loss": 1.1711, "step": 6117 }, { "epoch": 0.47, "learning_rate": 1.129726514194706e-06, "loss": 1.2557, "step": 6118 }, { "epoch": 0.47, "learning_rate": 1.1294772466263762e-06, "loss": 1.1947, "step": 6119 }, { "epoch": 0.47, "learning_rate": 1.1292279708756283e-06, "loss": 1.2225, "step": 6120 }, { "epoch": 0.48, "learning_rate": 1.1289786869582156e-06, "loss": 1.181, "step": 6121 }, { "epoch": 0.48, "learning_rate": 1.1287293948898918e-06, "loss": 1.2392, "step": 6122 }, { "epoch": 0.48, "learning_rate": 1.1284800946864114e-06, "loss": 0.8708, "step": 6123 }, { "epoch": 0.48, "learning_rate": 1.1282307863635289e-06, "loss": 1.2579, "step": 6124 }, { "epoch": 0.48, "learning_rate": 1.1279814699369993e-06, "loss": 1.2587, "step": 6125 }, { "epoch": 0.48, "learning_rate": 1.1277321454225789e-06, "loss": 1.2038, "step": 6126 }, { "epoch": 0.48, "learning_rate": 1.127482812836023e-06, "loss": 1.1883, "step": 6127 }, { "epoch": 0.48, "learning_rate": 1.1272334721930897e-06, "loss": 1.1886, "step": 6128 }, { "epoch": 0.48, "learning_rate": 1.1269841235095347e-06, "loss": 1.2083, "step": 6129 }, { "epoch": 0.48, "learning_rate": 1.1267347668011173e-06, "loss": 0.844, "step": 6130 }, { "epoch": 0.48, "learning_rate": 1.1264854020835955e-06, "loss": 1.2696, "step": 6131 }, { "epoch": 0.48, "learning_rate": 1.126236029372727e-06, "loss": 1.182, "step": 6132 }, { "epoch": 0.48, "learning_rate": 1.125986648684272e-06, "loss": 1.2389, "step": 6133 }, { "epoch": 0.48, "learning_rate": 1.1257372600339903e-06, "loss": 1.2084, "step": 6134 }, { "epoch": 0.48, "learning_rate": 1.1254878634376422e-06, "loss": 1.2257, "step": 6135 }, { "epoch": 0.48, "learning_rate": 1.1252384589109886e-06, "loss": 1.2301, "step": 6136 }, { "epoch": 0.48, "learning_rate": 1.1249890464697902e-06, "loss": 1.2007, "step": 6137 }, { "epoch": 0.48, "learning_rate": 1.1247396261298092e-06, "loss": 1.1777, "step": 6138 }, { "epoch": 0.48, "learning_rate": 1.124490197906808e-06, "loss": 1.1996, "step": 6139 }, { "epoch": 0.48, "learning_rate": 1.1242407618165498e-06, "loss": 1.1508, "step": 6140 }, { "epoch": 0.48, "learning_rate": 1.1239913178747973e-06, "loss": 0.8162, "step": 6141 }, { "epoch": 0.48, "learning_rate": 1.1237418660973143e-06, "loss": 1.2723, "step": 6142 }, { "epoch": 0.48, "learning_rate": 1.1234924064998654e-06, "loss": 1.2067, "step": 6143 }, { "epoch": 0.48, "learning_rate": 1.1232429390982153e-06, "loss": 1.247, "step": 6144 }, { "epoch": 0.48, "learning_rate": 1.1229934639081296e-06, "loss": 1.2659, "step": 6145 }, { "epoch": 0.48, "learning_rate": 1.1227439809453734e-06, "loss": 1.1797, "step": 6146 }, { "epoch": 0.48, "learning_rate": 1.1224944902257135e-06, "loss": 1.242, "step": 6147 }, { "epoch": 0.48, "learning_rate": 1.1222449917649166e-06, "loss": 1.2209, "step": 6148 }, { "epoch": 0.48, "learning_rate": 1.1219954855787495e-06, "loss": 1.2281, "step": 6149 }, { "epoch": 0.48, "learning_rate": 1.1217459716829804e-06, "loss": 1.2112, "step": 6150 }, { "epoch": 0.48, "learning_rate": 1.1214964500933778e-06, "loss": 1.2064, "step": 6151 }, { "epoch": 0.48, "learning_rate": 1.1212469208257099e-06, "loss": 1.2115, "step": 6152 }, { "epoch": 0.48, "learning_rate": 1.1209973838957463e-06, "loss": 1.1704, "step": 6153 }, { "epoch": 0.48, "learning_rate": 1.1207478393192557e-06, "loss": 1.2365, "step": 6154 }, { "epoch": 0.48, "learning_rate": 1.1204982871120097e-06, "loss": 1.2563, "step": 6155 }, { "epoch": 0.48, "learning_rate": 1.1202487272897777e-06, "loss": 1.2397, "step": 6156 }, { "epoch": 0.48, "learning_rate": 1.1199991598683321e-06, "loss": 1.2134, "step": 6157 }, { "epoch": 0.48, "learning_rate": 1.1197495848634435e-06, "loss": 1.1864, "step": 6158 }, { "epoch": 0.48, "learning_rate": 1.1195000022908843e-06, "loss": 1.2083, "step": 6159 }, { "epoch": 0.48, "learning_rate": 1.119250412166427e-06, "loss": 1.2016, "step": 6160 }, { "epoch": 0.48, "learning_rate": 1.1190008145058445e-06, "loss": 1.1908, "step": 6161 }, { "epoch": 0.48, "learning_rate": 1.1187512093249112e-06, "loss": 1.1752, "step": 6162 }, { "epoch": 0.48, "learning_rate": 1.1185015966394e-06, "loss": 0.8216, "step": 6163 }, { "epoch": 0.48, "learning_rate": 1.1182519764650857e-06, "loss": 1.1941, "step": 6164 }, { "epoch": 0.48, "learning_rate": 1.1180023488177438e-06, "loss": 1.2061, "step": 6165 }, { "epoch": 0.48, "learning_rate": 1.1177527137131486e-06, "loss": 1.218, "step": 6166 }, { "epoch": 0.48, "learning_rate": 1.1175030711670772e-06, "loss": 1.1208, "step": 6167 }, { "epoch": 0.48, "learning_rate": 1.1172534211953055e-06, "loss": 1.2193, "step": 6168 }, { "epoch": 0.48, "learning_rate": 1.11700376381361e-06, "loss": 1.2251, "step": 6169 }, { "epoch": 0.48, "learning_rate": 1.1167540990377684e-06, "loss": 1.1061, "step": 6170 }, { "epoch": 0.48, "learning_rate": 1.1165044268835583e-06, "loss": 1.1751, "step": 6171 }, { "epoch": 0.48, "learning_rate": 1.1162547473667577e-06, "loss": 1.2043, "step": 6172 }, { "epoch": 0.48, "learning_rate": 1.1160050605031455e-06, "loss": 0.843, "step": 6173 }, { "epoch": 0.48, "learning_rate": 1.1157553663085013e-06, "loss": 1.2223, "step": 6174 }, { "epoch": 0.48, "learning_rate": 1.1155056647986042e-06, "loss": 1.237, "step": 6175 }, { "epoch": 0.48, "learning_rate": 1.1152559559892345e-06, "loss": 1.2313, "step": 6176 }, { "epoch": 0.48, "learning_rate": 1.1150062398961725e-06, "loss": 1.2473, "step": 6177 }, { "epoch": 0.48, "learning_rate": 1.1147565165351994e-06, "loss": 1.2369, "step": 6178 }, { "epoch": 0.48, "learning_rate": 1.1145067859220968e-06, "loss": 0.843, "step": 6179 }, { "epoch": 0.48, "learning_rate": 1.1142570480726463e-06, "loss": 1.2453, "step": 6180 }, { "epoch": 0.48, "learning_rate": 1.1140073030026304e-06, "loss": 1.2597, "step": 6181 }, { "epoch": 0.48, "learning_rate": 1.1137575507278323e-06, "loss": 1.161, "step": 6182 }, { "epoch": 0.48, "learning_rate": 1.1135077912640346e-06, "loss": 1.1945, "step": 6183 }, { "epoch": 0.48, "learning_rate": 1.1132580246270218e-06, "loss": 1.246, "step": 6184 }, { "epoch": 0.48, "learning_rate": 1.1130082508325775e-06, "loss": 1.2821, "step": 6185 }, { "epoch": 0.48, "learning_rate": 1.1127584698964865e-06, "loss": 1.159, "step": 6186 }, { "epoch": 0.48, "learning_rate": 1.1125086818345343e-06, "loss": 1.1926, "step": 6187 }, { "epoch": 0.48, "learning_rate": 1.1122588866625058e-06, "loss": 1.1676, "step": 6188 }, { "epoch": 0.48, "learning_rate": 1.1120090843961876e-06, "loss": 1.1328, "step": 6189 }, { "epoch": 0.48, "learning_rate": 1.111759275051366e-06, "loss": 1.2075, "step": 6190 }, { "epoch": 0.48, "learning_rate": 1.1115094586438275e-06, "loss": 0.8431, "step": 6191 }, { "epoch": 0.48, "learning_rate": 1.11125963518936e-06, "loss": 1.1656, "step": 6192 }, { "epoch": 0.48, "learning_rate": 1.1110098047037508e-06, "loss": 1.2374, "step": 6193 }, { "epoch": 0.48, "learning_rate": 1.1107599672027885e-06, "loss": 1.1836, "step": 6194 }, { "epoch": 0.48, "learning_rate": 1.1105101227022615e-06, "loss": 1.2289, "step": 6195 }, { "epoch": 0.48, "learning_rate": 1.1102602712179595e-06, "loss": 1.2407, "step": 6196 }, { "epoch": 0.48, "learning_rate": 1.1100104127656714e-06, "loss": 1.2158, "step": 6197 }, { "epoch": 0.48, "learning_rate": 1.1097605473611874e-06, "loss": 1.2058, "step": 6198 }, { "epoch": 0.48, "learning_rate": 1.1095106750202983e-06, "loss": 0.8606, "step": 6199 }, { "epoch": 0.48, "learning_rate": 1.1092607957587942e-06, "loss": 1.2066, "step": 6200 }, { "epoch": 0.48, "learning_rate": 1.1090109095924675e-06, "loss": 1.256, "step": 6201 }, { "epoch": 0.48, "learning_rate": 1.108761016537109e-06, "loss": 1.1921, "step": 6202 }, { "epoch": 0.48, "learning_rate": 1.1085111166085114e-06, "loss": 1.2026, "step": 6203 }, { "epoch": 0.48, "learning_rate": 1.1082612098224671e-06, "loss": 1.2101, "step": 6204 }, { "epoch": 0.48, "learning_rate": 1.108011296194769e-06, "loss": 1.1973, "step": 6205 }, { "epoch": 0.48, "learning_rate": 1.107761375741211e-06, "loss": 1.2491, "step": 6206 }, { "epoch": 0.48, "learning_rate": 1.1075114484775873e-06, "loss": 1.2675, "step": 6207 }, { "epoch": 0.48, "learning_rate": 1.107261514419691e-06, "loss": 1.2027, "step": 6208 }, { "epoch": 0.48, "learning_rate": 1.107011573583318e-06, "loss": 1.2343, "step": 6209 }, { "epoch": 0.48, "learning_rate": 1.1067616259842635e-06, "loss": 1.2197, "step": 6210 }, { "epoch": 0.48, "learning_rate": 1.1065116716383225e-06, "loss": 1.2491, "step": 6211 }, { "epoch": 0.48, "learning_rate": 1.1062617105612918e-06, "loss": 1.2043, "step": 6212 }, { "epoch": 0.48, "learning_rate": 1.106011742768967e-06, "loss": 0.8573, "step": 6213 }, { "epoch": 0.48, "learning_rate": 1.1057617682771455e-06, "loss": 1.2167, "step": 6214 }, { "epoch": 0.48, "learning_rate": 1.1055117871016246e-06, "loss": 1.2468, "step": 6215 }, { "epoch": 0.48, "learning_rate": 1.1052617992582026e-06, "loss": 1.2218, "step": 6216 }, { "epoch": 0.48, "learning_rate": 1.1050118047626766e-06, "loss": 1.2, "step": 6217 }, { "epoch": 0.48, "learning_rate": 1.1047618036308465e-06, "loss": 1.2496, "step": 6218 }, { "epoch": 0.48, "learning_rate": 1.10451179587851e-06, "loss": 1.234, "step": 6219 }, { "epoch": 0.48, "learning_rate": 1.104261781521467e-06, "loss": 1.1596, "step": 6220 }, { "epoch": 0.48, "learning_rate": 1.104011760575518e-06, "loss": 1.2261, "step": 6221 }, { "epoch": 0.48, "learning_rate": 1.1037617330564623e-06, "loss": 1.2488, "step": 6222 }, { "epoch": 0.48, "learning_rate": 1.1035116989801014e-06, "loss": 1.1818, "step": 6223 }, { "epoch": 0.48, "learning_rate": 1.1032616583622363e-06, "loss": 1.1905, "step": 6224 }, { "epoch": 0.48, "learning_rate": 1.1030116112186679e-06, "loss": 1.2247, "step": 6225 }, { "epoch": 0.48, "learning_rate": 1.1027615575651984e-06, "loss": 1.2268, "step": 6226 }, { "epoch": 0.48, "learning_rate": 1.1025114974176304e-06, "loss": 1.2175, "step": 6227 }, { "epoch": 0.48, "learning_rate": 1.1022614307917667e-06, "loss": 1.1748, "step": 6228 }, { "epoch": 0.48, "learning_rate": 1.1020113577034103e-06, "loss": 1.2329, "step": 6229 }, { "epoch": 0.48, "learning_rate": 1.1017612781683642e-06, "loss": 1.2592, "step": 6230 }, { "epoch": 0.48, "learning_rate": 1.1015111922024331e-06, "loss": 1.2142, "step": 6231 }, { "epoch": 0.48, "learning_rate": 1.1012610998214212e-06, "loss": 1.2319, "step": 6232 }, { "epoch": 0.48, "learning_rate": 1.1010110010411335e-06, "loss": 1.2104, "step": 6233 }, { "epoch": 0.48, "learning_rate": 1.1007608958773748e-06, "loss": 1.2475, "step": 6234 }, { "epoch": 0.48, "learning_rate": 1.1005107843459512e-06, "loss": 0.8199, "step": 6235 }, { "epoch": 0.48, "learning_rate": 1.100260666462668e-06, "loss": 1.1706, "step": 6236 }, { "epoch": 0.48, "learning_rate": 1.1000105422433316e-06, "loss": 1.2381, "step": 6237 }, { "epoch": 0.48, "learning_rate": 1.09976041170375e-06, "loss": 1.254, "step": 6238 }, { "epoch": 0.48, "learning_rate": 1.0995102748597287e-06, "loss": 1.2374, "step": 6239 }, { "epoch": 0.48, "learning_rate": 1.0992601317270767e-06, "loss": 1.2018, "step": 6240 }, { "epoch": 0.48, "learning_rate": 1.0990099823216014e-06, "loss": 0.8495, "step": 6241 }, { "epoch": 0.48, "learning_rate": 1.0987598266591112e-06, "loss": 1.1804, "step": 6242 }, { "epoch": 0.48, "learning_rate": 1.0985096647554147e-06, "loss": 1.191, "step": 6243 }, { "epoch": 0.48, "learning_rate": 1.0982594966263217e-06, "loss": 0.8427, "step": 6244 }, { "epoch": 0.48, "learning_rate": 1.098009322287641e-06, "loss": 1.221, "step": 6245 }, { "epoch": 0.48, "learning_rate": 1.0977591417551835e-06, "loss": 1.2211, "step": 6246 }, { "epoch": 0.48, "learning_rate": 1.0975089550447583e-06, "loss": 0.8437, "step": 6247 }, { "epoch": 0.48, "learning_rate": 1.0972587621721772e-06, "loss": 1.249, "step": 6248 }, { "epoch": 0.49, "learning_rate": 1.0970085631532508e-06, "loss": 1.1828, "step": 6249 }, { "epoch": 0.49, "learning_rate": 1.0967583580037913e-06, "loss": 1.209, "step": 6250 }, { "epoch": 0.49, "learning_rate": 1.0965081467396098e-06, "loss": 1.1832, "step": 6251 }, { "epoch": 0.49, "learning_rate": 1.096257929376519e-06, "loss": 1.2466, "step": 6252 }, { "epoch": 0.49, "learning_rate": 1.0960077059303315e-06, "loss": 1.228, "step": 6253 }, { "epoch": 0.49, "learning_rate": 1.0957574764168601e-06, "loss": 1.2389, "step": 6254 }, { "epoch": 0.49, "learning_rate": 1.0955072408519192e-06, "loss": 1.2059, "step": 6255 }, { "epoch": 0.49, "learning_rate": 1.0952569992513216e-06, "loss": 1.2017, "step": 6256 }, { "epoch": 0.49, "learning_rate": 1.0950067516308817e-06, "loss": 1.2053, "step": 6257 }, { "epoch": 0.49, "learning_rate": 1.0947564980064148e-06, "loss": 1.1461, "step": 6258 }, { "epoch": 0.49, "learning_rate": 1.0945062383937348e-06, "loss": 1.1931, "step": 6259 }, { "epoch": 0.49, "learning_rate": 1.0942559728086581e-06, "loss": 1.251, "step": 6260 }, { "epoch": 0.49, "learning_rate": 1.0940057012669997e-06, "loss": 1.2255, "step": 6261 }, { "epoch": 0.49, "learning_rate": 1.0937554237845761e-06, "loss": 1.1915, "step": 6262 }, { "epoch": 0.49, "learning_rate": 1.0935051403772037e-06, "loss": 1.1957, "step": 6263 }, { "epoch": 0.49, "learning_rate": 1.0932548510606992e-06, "loss": 1.227, "step": 6264 }, { "epoch": 0.49, "learning_rate": 1.09300455585088e-06, "loss": 1.2707, "step": 6265 }, { "epoch": 0.49, "learning_rate": 1.0927542547635634e-06, "loss": 1.2241, "step": 6266 }, { "epoch": 0.49, "learning_rate": 1.092503947814568e-06, "loss": 1.1859, "step": 6267 }, { "epoch": 0.49, "learning_rate": 1.0922536350197119e-06, "loss": 1.2218, "step": 6268 }, { "epoch": 0.49, "learning_rate": 1.0920033163948135e-06, "loss": 1.1906, "step": 6269 }, { "epoch": 0.49, "learning_rate": 1.091752991955692e-06, "loss": 1.2015, "step": 6270 }, { "epoch": 0.49, "learning_rate": 1.091502661718167e-06, "loss": 1.2091, "step": 6271 }, { "epoch": 0.49, "learning_rate": 1.0912523256980583e-06, "loss": 1.2325, "step": 6272 }, { "epoch": 0.49, "learning_rate": 1.0910019839111859e-06, "loss": 1.2326, "step": 6273 }, { "epoch": 0.49, "learning_rate": 1.0907516363733703e-06, "loss": 1.1588, "step": 6274 }, { "epoch": 0.49, "learning_rate": 1.0905012831004334e-06, "loss": 1.2402, "step": 6275 }, { "epoch": 0.49, "learning_rate": 1.090250924108195e-06, "loss": 0.8539, "step": 6276 }, { "epoch": 0.49, "learning_rate": 1.0900005594124777e-06, "loss": 1.227, "step": 6277 }, { "epoch": 0.49, "learning_rate": 1.0897501890291032e-06, "loss": 1.1545, "step": 6278 }, { "epoch": 0.49, "learning_rate": 1.0894998129738936e-06, "loss": 1.2305, "step": 6279 }, { "epoch": 0.49, "learning_rate": 1.089249431262672e-06, "loss": 1.2355, "step": 6280 }, { "epoch": 0.49, "learning_rate": 1.0889990439112612e-06, "loss": 1.2131, "step": 6281 }, { "epoch": 0.49, "learning_rate": 1.0887486509354851e-06, "loss": 1.2087, "step": 6282 }, { "epoch": 0.49, "learning_rate": 1.088498252351167e-06, "loss": 1.2434, "step": 6283 }, { "epoch": 0.49, "learning_rate": 1.0882478481741315e-06, "loss": 0.8061, "step": 6284 }, { "epoch": 0.49, "learning_rate": 1.0879974384202025e-06, "loss": 1.2132, "step": 6285 }, { "epoch": 0.49, "learning_rate": 1.0877470231052053e-06, "loss": 1.2067, "step": 6286 }, { "epoch": 0.49, "learning_rate": 1.0874966022449648e-06, "loss": 1.1924, "step": 6287 }, { "epoch": 0.49, "learning_rate": 1.0872461758553064e-06, "loss": 1.1197, "step": 6288 }, { "epoch": 0.49, "learning_rate": 1.086995743952057e-06, "loss": 1.1769, "step": 6289 }, { "epoch": 0.49, "learning_rate": 1.0867453065510417e-06, "loss": 1.2121, "step": 6290 }, { "epoch": 0.49, "learning_rate": 1.0864948636680873e-06, "loss": 1.173, "step": 6291 }, { "epoch": 0.49, "learning_rate": 1.0862444153190213e-06, "loss": 1.1834, "step": 6292 }, { "epoch": 0.49, "learning_rate": 1.0859939615196705e-06, "loss": 1.1975, "step": 6293 }, { "epoch": 0.49, "learning_rate": 1.0857435022858627e-06, "loss": 1.1748, "step": 6294 }, { "epoch": 0.49, "learning_rate": 1.085493037633426e-06, "loss": 1.1594, "step": 6295 }, { "epoch": 0.49, "learning_rate": 1.085242567578188e-06, "loss": 1.2352, "step": 6296 }, { "epoch": 0.49, "learning_rate": 1.0849920921359785e-06, "loss": 1.2711, "step": 6297 }, { "epoch": 0.49, "learning_rate": 1.0847416113226256e-06, "loss": 0.8199, "step": 6298 }, { "epoch": 0.49, "learning_rate": 1.084491125153959e-06, "loss": 1.2234, "step": 6299 }, { "epoch": 0.49, "learning_rate": 1.0842406336458085e-06, "loss": 1.2194, "step": 6300 }, { "epoch": 0.49, "learning_rate": 1.0839901368140037e-06, "loss": 1.2392, "step": 6301 }, { "epoch": 0.49, "learning_rate": 1.0837396346743751e-06, "loss": 1.1698, "step": 6302 }, { "epoch": 0.49, "learning_rate": 1.0834891272427536e-06, "loss": 0.8605, "step": 6303 }, { "epoch": 0.49, "learning_rate": 1.0832386145349699e-06, "loss": 1.185, "step": 6304 }, { "epoch": 0.49, "learning_rate": 1.0829880965668555e-06, "loss": 1.1868, "step": 6305 }, { "epoch": 0.49, "learning_rate": 1.0827375733542424e-06, "loss": 1.2006, "step": 6306 }, { "epoch": 0.49, "learning_rate": 1.0824870449129622e-06, "loss": 1.1852, "step": 6307 }, { "epoch": 0.49, "learning_rate": 1.0822365112588469e-06, "loss": 1.217, "step": 6308 }, { "epoch": 0.49, "learning_rate": 1.0819859724077304e-06, "loss": 1.1854, "step": 6309 }, { "epoch": 0.49, "learning_rate": 1.0817354283754443e-06, "loss": 1.203, "step": 6310 }, { "epoch": 0.49, "learning_rate": 1.081484879177823e-06, "loss": 1.218, "step": 6311 }, { "epoch": 0.49, "learning_rate": 1.0812343248306997e-06, "loss": 1.2084, "step": 6312 }, { "epoch": 0.49, "learning_rate": 1.080983765349908e-06, "loss": 1.1908, "step": 6313 }, { "epoch": 0.49, "learning_rate": 1.080733200751283e-06, "loss": 1.2038, "step": 6314 }, { "epoch": 0.49, "learning_rate": 1.0804826310506588e-06, "loss": 1.1757, "step": 6315 }, { "epoch": 0.49, "learning_rate": 1.0802320562638703e-06, "loss": 1.2278, "step": 6316 }, { "epoch": 0.49, "learning_rate": 1.0799814764067534e-06, "loss": 1.2501, "step": 6317 }, { "epoch": 0.49, "learning_rate": 1.0797308914951428e-06, "loss": 1.205, "step": 6318 }, { "epoch": 0.49, "learning_rate": 1.0794803015448756e-06, "loss": 1.2393, "step": 6319 }, { "epoch": 0.49, "learning_rate": 1.0792297065717864e-06, "loss": 1.2324, "step": 6320 }, { "epoch": 0.49, "learning_rate": 1.078979106591713e-06, "loss": 1.2352, "step": 6321 }, { "epoch": 0.49, "learning_rate": 1.0787285016204922e-06, "loss": 1.2043, "step": 6322 }, { "epoch": 0.49, "learning_rate": 1.0784778916739605e-06, "loss": 1.2201, "step": 6323 }, { "epoch": 0.49, "learning_rate": 1.0782272767679558e-06, "loss": 1.1955, "step": 6324 }, { "epoch": 0.49, "learning_rate": 1.077976656918316e-06, "loss": 0.8073, "step": 6325 }, { "epoch": 0.49, "learning_rate": 1.0777260321408794e-06, "loss": 1.1518, "step": 6326 }, { "epoch": 0.49, "learning_rate": 1.0774754024514838e-06, "loss": 1.1985, "step": 6327 }, { "epoch": 0.49, "learning_rate": 1.0772247678659688e-06, "loss": 1.1848, "step": 6328 }, { "epoch": 0.49, "learning_rate": 1.0769741284001724e-06, "loss": 0.8371, "step": 6329 }, { "epoch": 0.49, "learning_rate": 1.0767234840699347e-06, "loss": 0.8442, "step": 6330 }, { "epoch": 0.49, "learning_rate": 1.0764728348910953e-06, "loss": 1.2171, "step": 6331 }, { "epoch": 0.49, "learning_rate": 1.0762221808794937e-06, "loss": 1.2134, "step": 6332 }, { "epoch": 0.49, "learning_rate": 1.0759715220509708e-06, "loss": 1.2125, "step": 6333 }, { "epoch": 0.49, "learning_rate": 1.0757208584213673e-06, "loss": 1.1889, "step": 6334 }, { "epoch": 0.49, "learning_rate": 1.075470190006523e-06, "loss": 1.2197, "step": 6335 }, { "epoch": 0.49, "learning_rate": 1.0752195168222805e-06, "loss": 1.2504, "step": 6336 }, { "epoch": 0.49, "learning_rate": 1.07496883888448e-06, "loss": 1.2163, "step": 6337 }, { "epoch": 0.49, "learning_rate": 1.0747181562089642e-06, "loss": 1.2469, "step": 6338 }, { "epoch": 0.49, "learning_rate": 1.074467468811575e-06, "loss": 1.1958, "step": 6339 }, { "epoch": 0.49, "learning_rate": 1.0742167767081542e-06, "loss": 1.1949, "step": 6340 }, { "epoch": 0.49, "learning_rate": 1.0739660799145454e-06, "loss": 1.2177, "step": 6341 }, { "epoch": 0.49, "learning_rate": 1.0737153784465906e-06, "loss": 1.2143, "step": 6342 }, { "epoch": 0.49, "learning_rate": 1.0734646723201343e-06, "loss": 1.2567, "step": 6343 }, { "epoch": 0.49, "learning_rate": 1.0732139615510192e-06, "loss": 1.2425, "step": 6344 }, { "epoch": 0.49, "learning_rate": 1.0729632461550892e-06, "loss": 0.8742, "step": 6345 }, { "epoch": 0.49, "learning_rate": 1.0727125261481889e-06, "loss": 1.2089, "step": 6346 }, { "epoch": 0.49, "learning_rate": 1.072461801546162e-06, "loss": 1.154, "step": 6347 }, { "epoch": 0.49, "learning_rate": 1.0722110723648542e-06, "loss": 1.1996, "step": 6348 }, { "epoch": 0.49, "learning_rate": 1.0719603386201094e-06, "loss": 1.1856, "step": 6349 }, { "epoch": 0.49, "learning_rate": 1.0717096003277742e-06, "loss": 1.1825, "step": 6350 }, { "epoch": 0.49, "learning_rate": 1.0714588575036936e-06, "loss": 1.1796, "step": 6351 }, { "epoch": 0.49, "learning_rate": 1.0712081101637128e-06, "loss": 1.2035, "step": 6352 }, { "epoch": 0.49, "learning_rate": 1.0709573583236796e-06, "loss": 1.2334, "step": 6353 }, { "epoch": 0.49, "learning_rate": 1.0707066019994386e-06, "loss": 0.8305, "step": 6354 }, { "epoch": 0.49, "learning_rate": 1.0704558412068378e-06, "loss": 0.8386, "step": 6355 }, { "epoch": 0.49, "learning_rate": 1.070205075961724e-06, "loss": 1.249, "step": 6356 }, { "epoch": 0.49, "learning_rate": 1.0699543062799441e-06, "loss": 1.2863, "step": 6357 }, { "epoch": 0.49, "learning_rate": 1.0697035321773463e-06, "loss": 1.2539, "step": 6358 }, { "epoch": 0.49, "learning_rate": 1.0694527536697779e-06, "loss": 1.2104, "step": 6359 }, { "epoch": 0.49, "learning_rate": 1.0692019707730874e-06, "loss": 1.2173, "step": 6360 }, { "epoch": 0.49, "learning_rate": 1.0689511835031233e-06, "loss": 0.8217, "step": 6361 }, { "epoch": 0.49, "learning_rate": 1.0687003918757339e-06, "loss": 1.226, "step": 6362 }, { "epoch": 0.49, "learning_rate": 1.0684495959067685e-06, "loss": 1.1917, "step": 6363 }, { "epoch": 0.49, "learning_rate": 1.068198795612076e-06, "loss": 1.1994, "step": 6364 }, { "epoch": 0.49, "learning_rate": 1.0679479910075065e-06, "loss": 1.2411, "step": 6365 }, { "epoch": 0.49, "learning_rate": 1.0676971821089093e-06, "loss": 1.2085, "step": 6366 }, { "epoch": 0.49, "learning_rate": 1.0674463689321346e-06, "loss": 1.2576, "step": 6367 }, { "epoch": 0.49, "learning_rate": 1.067195551493033e-06, "loss": 1.1779, "step": 6368 }, { "epoch": 0.49, "learning_rate": 1.0669447298074545e-06, "loss": 1.2021, "step": 6369 }, { "epoch": 0.49, "learning_rate": 1.066693903891251e-06, "loss": 1.1632, "step": 6370 }, { "epoch": 0.49, "learning_rate": 1.0664430737602724e-06, "loss": 1.1433, "step": 6371 }, { "epoch": 0.49, "learning_rate": 1.066192239430371e-06, "loss": 1.2019, "step": 6372 }, { "epoch": 0.49, "learning_rate": 1.0659414009173984e-06, "loss": 1.2415, "step": 6373 }, { "epoch": 0.49, "learning_rate": 1.0656905582372057e-06, "loss": 1.2029, "step": 6374 }, { "epoch": 0.49, "learning_rate": 1.0654397114056464e-06, "loss": 1.2243, "step": 6375 }, { "epoch": 0.49, "learning_rate": 1.0651888604385719e-06, "loss": 1.2427, "step": 6376 }, { "epoch": 0.49, "learning_rate": 1.0649380053518355e-06, "loss": 1.1552, "step": 6377 }, { "epoch": 0.5, "learning_rate": 1.06468714616129e-06, "loss": 1.236, "step": 6378 }, { "epoch": 0.5, "learning_rate": 1.0644362828827886e-06, "loss": 1.1717, "step": 6379 }, { "epoch": 0.5, "learning_rate": 1.064185415532185e-06, "loss": 1.2333, "step": 6380 }, { "epoch": 0.5, "learning_rate": 1.0639345441253325e-06, "loss": 1.2588, "step": 6381 }, { "epoch": 0.5, "learning_rate": 1.0636836686780862e-06, "loss": 1.2183, "step": 6382 }, { "epoch": 0.5, "learning_rate": 1.063432789206299e-06, "loss": 1.2234, "step": 6383 }, { "epoch": 0.5, "learning_rate": 1.063181905725826e-06, "loss": 1.1825, "step": 6384 }, { "epoch": 0.5, "learning_rate": 1.0629310182525224e-06, "loss": 0.8565, "step": 6385 }, { "epoch": 0.5, "learning_rate": 1.0626801268022426e-06, "loss": 1.2165, "step": 6386 }, { "epoch": 0.5, "learning_rate": 1.0624292313908424e-06, "loss": 1.1685, "step": 6387 }, { "epoch": 0.5, "learning_rate": 1.062178332034177e-06, "loss": 1.1665, "step": 6388 }, { "epoch": 0.5, "learning_rate": 1.061927428748102e-06, "loss": 0.8835, "step": 6389 }, { "epoch": 0.5, "learning_rate": 1.0616765215484742e-06, "loss": 0.824, "step": 6390 }, { "epoch": 0.5, "learning_rate": 1.061425610451149e-06, "loss": 1.2376, "step": 6391 }, { "epoch": 0.5, "learning_rate": 1.0611746954719834e-06, "loss": 0.8647, "step": 6392 }, { "epoch": 0.5, "learning_rate": 1.0609237766268342e-06, "loss": 1.23, "step": 6393 }, { "epoch": 0.5, "learning_rate": 1.0606728539315581e-06, "loss": 1.28, "step": 6394 }, { "epoch": 0.5, "learning_rate": 1.060421927402013e-06, "loss": 1.1271, "step": 6395 }, { "epoch": 0.5, "learning_rate": 1.0601709970540555e-06, "loss": 1.2389, "step": 6396 }, { "epoch": 0.5, "learning_rate": 1.0599200629035442e-06, "loss": 0.8337, "step": 6397 }, { "epoch": 0.5, "learning_rate": 1.059669124966336e-06, "loss": 1.2074, "step": 6398 }, { "epoch": 0.5, "learning_rate": 1.0594181832582906e-06, "loss": 1.2249, "step": 6399 }, { "epoch": 0.5, "learning_rate": 1.0591672377952655e-06, "loss": 1.197, "step": 6400 }, { "epoch": 0.5, "learning_rate": 1.0589162885931193e-06, "loss": 1.2073, "step": 6401 }, { "epoch": 0.5, "learning_rate": 1.0586653356677115e-06, "loss": 1.1866, "step": 6402 }, { "epoch": 0.5, "learning_rate": 1.0584143790349008e-06, "loss": 1.2207, "step": 6403 }, { "epoch": 0.5, "learning_rate": 1.058163418710547e-06, "loss": 1.1857, "step": 6404 }, { "epoch": 0.5, "learning_rate": 1.0579124547105096e-06, "loss": 1.2521, "step": 6405 }, { "epoch": 0.5, "learning_rate": 1.057661487050648e-06, "loss": 1.1806, "step": 6406 }, { "epoch": 0.5, "learning_rate": 1.0574105157468232e-06, "loss": 0.8045, "step": 6407 }, { "epoch": 0.5, "learning_rate": 1.0571595408148944e-06, "loss": 1.1884, "step": 6408 }, { "epoch": 0.5, "learning_rate": 1.0569085622707233e-06, "loss": 0.8535, "step": 6409 }, { "epoch": 0.5, "learning_rate": 1.0566575801301702e-06, "loss": 1.2406, "step": 6410 }, { "epoch": 0.5, "learning_rate": 1.0564065944090957e-06, "loss": 1.2548, "step": 6411 }, { "epoch": 0.5, "learning_rate": 1.056155605123362e-06, "loss": 0.8528, "step": 6412 }, { "epoch": 0.5, "learning_rate": 1.0559046122888293e-06, "loss": 1.1781, "step": 6413 }, { "epoch": 0.5, "learning_rate": 1.0556536159213604e-06, "loss": 1.1267, "step": 6414 }, { "epoch": 0.5, "learning_rate": 1.0554026160368166e-06, "loss": 1.2616, "step": 6415 }, { "epoch": 0.5, "learning_rate": 1.0551516126510608e-06, "loss": 1.1819, "step": 6416 }, { "epoch": 0.5, "learning_rate": 1.0549006057799545e-06, "loss": 1.1746, "step": 6417 }, { "epoch": 0.5, "learning_rate": 1.0546495954393605e-06, "loss": 1.1888, "step": 6418 }, { "epoch": 0.5, "learning_rate": 1.0543985816451419e-06, "loss": 1.2282, "step": 6419 }, { "epoch": 0.5, "learning_rate": 1.0541475644131612e-06, "loss": 1.2034, "step": 6420 }, { "epoch": 0.5, "learning_rate": 1.0538965437592825e-06, "loss": 1.1704, "step": 6421 }, { "epoch": 0.5, "learning_rate": 1.053645519699368e-06, "loss": 1.2287, "step": 6422 }, { "epoch": 0.5, "learning_rate": 1.0533944922492823e-06, "loss": 1.2129, "step": 6423 }, { "epoch": 0.5, "learning_rate": 1.0531434614248894e-06, "loss": 1.1978, "step": 6424 }, { "epoch": 0.5, "learning_rate": 1.0528924272420524e-06, "loss": 1.2455, "step": 6425 }, { "epoch": 0.5, "learning_rate": 1.0526413897166369e-06, "loss": 1.2007, "step": 6426 }, { "epoch": 0.5, "learning_rate": 1.0523903488645068e-06, "loss": 0.8351, "step": 6427 }, { "epoch": 0.5, "learning_rate": 1.0521393047015261e-06, "loss": 1.2545, "step": 6428 }, { "epoch": 0.5, "learning_rate": 1.051888257243561e-06, "loss": 1.2098, "step": 6429 }, { "epoch": 0.5, "learning_rate": 1.0516372065064758e-06, "loss": 1.2225, "step": 6430 }, { "epoch": 0.5, "learning_rate": 1.051386152506136e-06, "loss": 1.2626, "step": 6431 }, { "epoch": 0.5, "learning_rate": 1.0511350952584076e-06, "loss": 1.2141, "step": 6432 }, { "epoch": 0.5, "learning_rate": 1.0508840347791558e-06, "loss": 1.2197, "step": 6433 }, { "epoch": 0.5, "learning_rate": 1.050632971084247e-06, "loss": 1.1944, "step": 6434 }, { "epoch": 0.5, "learning_rate": 1.0503819041895473e-06, "loss": 0.8669, "step": 6435 }, { "epoch": 0.5, "learning_rate": 1.0501308341109225e-06, "loss": 1.2298, "step": 6436 }, { "epoch": 0.5, "learning_rate": 1.0498797608642397e-06, "loss": 0.827, "step": 6437 }, { "epoch": 0.5, "learning_rate": 1.049628684465366e-06, "loss": 1.1917, "step": 6438 }, { "epoch": 0.5, "learning_rate": 1.0493776049301677e-06, "loss": 1.1784, "step": 6439 }, { "epoch": 0.5, "learning_rate": 1.049126522274512e-06, "loss": 1.2396, "step": 6440 }, { "epoch": 0.5, "learning_rate": 1.0488754365142672e-06, "loss": 1.2293, "step": 6441 }, { "epoch": 0.5, "learning_rate": 1.0486243476652996e-06, "loss": 1.2238, "step": 6442 }, { "epoch": 0.5, "learning_rate": 1.0483732557434777e-06, "loss": 1.2048, "step": 6443 }, { "epoch": 0.5, "learning_rate": 1.0481221607646695e-06, "loss": 1.2009, "step": 6444 }, { "epoch": 0.5, "learning_rate": 1.0478710627447425e-06, "loss": 1.21, "step": 6445 }, { "epoch": 0.5, "learning_rate": 1.047619961699566e-06, "loss": 1.19, "step": 6446 }, { "epoch": 0.5, "learning_rate": 1.0473688576450075e-06, "loss": 1.216, "step": 6447 }, { "epoch": 0.5, "learning_rate": 1.0471177505969364e-06, "loss": 1.2201, "step": 6448 }, { "epoch": 0.5, "learning_rate": 1.0468666405712219e-06, "loss": 1.1419, "step": 6449 }, { "epoch": 0.5, "learning_rate": 1.0466155275837319e-06, "loss": 1.1606, "step": 6450 }, { "epoch": 0.5, "learning_rate": 1.0463644116503367e-06, "loss": 1.1329, "step": 6451 }, { "epoch": 0.5, "learning_rate": 1.0461132927869056e-06, "loss": 1.2175, "step": 6452 }, { "epoch": 0.5, "learning_rate": 1.0458621710093084e-06, "loss": 0.8379, "step": 6453 }, { "epoch": 0.5, "learning_rate": 1.0456110463334142e-06, "loss": 1.1855, "step": 6454 }, { "epoch": 0.5, "learning_rate": 1.0453599187750939e-06, "loss": 1.1818, "step": 6455 }, { "epoch": 0.5, "learning_rate": 1.0451087883502173e-06, "loss": 1.2154, "step": 6456 }, { "epoch": 0.5, "learning_rate": 1.0448576550746545e-06, "loss": 1.2128, "step": 6457 }, { "epoch": 0.5, "learning_rate": 1.0446065189642766e-06, "loss": 1.2422, "step": 6458 }, { "epoch": 0.5, "learning_rate": 1.0443553800349541e-06, "loss": 1.2023, "step": 6459 }, { "epoch": 0.5, "learning_rate": 1.0441042383025578e-06, "loss": 0.8145, "step": 6460 }, { "epoch": 0.5, "learning_rate": 1.0438530937829592e-06, "loss": 1.2103, "step": 6461 }, { "epoch": 0.5, "learning_rate": 1.043601946492029e-06, "loss": 1.2017, "step": 6462 }, { "epoch": 0.5, "learning_rate": 1.0433507964456396e-06, "loss": 1.214, "step": 6463 }, { "epoch": 0.5, "learning_rate": 1.0430996436596615e-06, "loss": 1.2254, "step": 6464 }, { "epoch": 0.5, "learning_rate": 1.0428484881499673e-06, "loss": 1.2357, "step": 6465 }, { "epoch": 0.5, "learning_rate": 1.0425973299324286e-06, "loss": 1.195, "step": 6466 }, { "epoch": 0.5, "learning_rate": 1.0423461690229176e-06, "loss": 1.2107, "step": 6467 }, { "epoch": 0.5, "learning_rate": 1.0420950054373065e-06, "loss": 1.2458, "step": 6468 }, { "epoch": 0.5, "learning_rate": 1.0418438391914682e-06, "loss": 1.2778, "step": 6469 }, { "epoch": 0.5, "learning_rate": 1.041592670301275e-06, "loss": 1.1879, "step": 6470 }, { "epoch": 0.5, "learning_rate": 1.0413414987825999e-06, "loss": 1.2677, "step": 6471 }, { "epoch": 0.5, "learning_rate": 1.0410903246513153e-06, "loss": 0.81, "step": 6472 }, { "epoch": 0.5, "learning_rate": 1.0408391479232954e-06, "loss": 1.2353, "step": 6473 }, { "epoch": 0.5, "learning_rate": 1.0405879686144125e-06, "loss": 1.1609, "step": 6474 }, { "epoch": 0.5, "learning_rate": 1.040336786740541e-06, "loss": 1.202, "step": 6475 }, { "epoch": 0.5, "learning_rate": 1.040085602317554e-06, "loss": 1.2167, "step": 6476 }, { "epoch": 0.5, "learning_rate": 1.039834415361325e-06, "loss": 1.2218, "step": 6477 }, { "epoch": 0.5, "learning_rate": 1.0395832258877286e-06, "loss": 1.1925, "step": 6478 }, { "epoch": 0.5, "learning_rate": 1.0393320339126384e-06, "loss": 1.2402, "step": 6479 }, { "epoch": 0.5, "learning_rate": 1.0390808394519293e-06, "loss": 0.8468, "step": 6480 }, { "epoch": 0.5, "learning_rate": 1.038829642521475e-06, "loss": 1.1736, "step": 6481 }, { "epoch": 0.5, "learning_rate": 1.0385784431371507e-06, "loss": 1.2263, "step": 6482 }, { "epoch": 0.5, "learning_rate": 1.038327241314831e-06, "loss": 1.2189, "step": 6483 }, { "epoch": 0.5, "learning_rate": 1.0380760370703902e-06, "loss": 0.8382, "step": 6484 }, { "epoch": 0.5, "learning_rate": 1.037824830419704e-06, "loss": 1.1507, "step": 6485 }, { "epoch": 0.5, "learning_rate": 1.0375736213786476e-06, "loss": 1.2403, "step": 6486 }, { "epoch": 0.5, "learning_rate": 1.0373224099630964e-06, "loss": 1.2142, "step": 6487 }, { "epoch": 0.5, "learning_rate": 1.0370711961889257e-06, "loss": 1.1754, "step": 6488 }, { "epoch": 0.5, "learning_rate": 1.0368199800720107e-06, "loss": 0.8196, "step": 6489 }, { "epoch": 0.5, "learning_rate": 1.0365687616282278e-06, "loss": 0.8406, "step": 6490 }, { "epoch": 0.5, "learning_rate": 1.036317540873453e-06, "loss": 1.205, "step": 6491 }, { "epoch": 0.5, "learning_rate": 1.0360663178235625e-06, "loss": 1.2399, "step": 6492 }, { "epoch": 0.5, "learning_rate": 1.0358150924944318e-06, "loss": 1.1873, "step": 6493 }, { "epoch": 0.5, "learning_rate": 1.0355638649019377e-06, "loss": 1.1607, "step": 6494 }, { "epoch": 0.5, "learning_rate": 1.0353126350619571e-06, "loss": 0.7718, "step": 6495 }, { "epoch": 0.5, "learning_rate": 1.0350614029903662e-06, "loss": 1.1816, "step": 6496 }, { "epoch": 0.5, "learning_rate": 1.034810168703042e-06, "loss": 1.1426, "step": 6497 }, { "epoch": 0.5, "learning_rate": 1.0345589322158614e-06, "loss": 1.2133, "step": 6498 }, { "epoch": 0.5, "learning_rate": 1.0343076935447013e-06, "loss": 1.1865, "step": 6499 }, { "epoch": 0.5, "learning_rate": 1.0340564527054393e-06, "loss": 1.1499, "step": 6500 }, { "epoch": 0.5, "learning_rate": 1.0338052097139525e-06, "loss": 1.2199, "step": 6501 }, { "epoch": 0.5, "learning_rate": 1.0335539645861187e-06, "loss": 1.1964, "step": 6502 }, { "epoch": 0.5, "learning_rate": 1.033302717337815e-06, "loss": 1.2334, "step": 6503 }, { "epoch": 0.5, "learning_rate": 1.0330514679849199e-06, "loss": 1.1822, "step": 6504 }, { "epoch": 0.5, "learning_rate": 1.0328002165433107e-06, "loss": 1.2185, "step": 6505 }, { "epoch": 0.5, "learning_rate": 1.0325489630288655e-06, "loss": 1.2519, "step": 6506 }, { "epoch": 0.51, "learning_rate": 1.0322977074574627e-06, "loss": 1.184, "step": 6507 }, { "epoch": 0.51, "learning_rate": 1.0320464498449802e-06, "loss": 1.2302, "step": 6508 }, { "epoch": 0.51, "learning_rate": 1.0317951902072973e-06, "loss": 1.2199, "step": 6509 }, { "epoch": 0.51, "learning_rate": 1.0315439285602919e-06, "loss": 1.2035, "step": 6510 }, { "epoch": 0.51, "learning_rate": 1.0312926649198423e-06, "loss": 1.1682, "step": 6511 }, { "epoch": 0.51, "learning_rate": 1.031041399301828e-06, "loss": 1.2144, "step": 6512 }, { "epoch": 0.51, "learning_rate": 1.0307901317221276e-06, "loss": 1.1746, "step": 6513 }, { "epoch": 0.51, "learning_rate": 1.0305388621966208e-06, "loss": 1.2377, "step": 6514 }, { "epoch": 0.51, "learning_rate": 1.0302875907411856e-06, "loss": 1.2473, "step": 6515 }, { "epoch": 0.51, "learning_rate": 1.030036317371702e-06, "loss": 1.2193, "step": 6516 }, { "epoch": 0.51, "learning_rate": 1.0297850421040498e-06, "loss": 0.8289, "step": 6517 }, { "epoch": 0.51, "learning_rate": 1.0295337649541074e-06, "loss": 1.208, "step": 6518 }, { "epoch": 0.51, "learning_rate": 1.0292824859377556e-06, "loss": 1.1606, "step": 6519 }, { "epoch": 0.51, "learning_rate": 1.0290312050708738e-06, "loss": 1.2341, "step": 6520 }, { "epoch": 0.51, "learning_rate": 1.0287799223693413e-06, "loss": 1.1806, "step": 6521 }, { "epoch": 0.51, "learning_rate": 1.0285286378490393e-06, "loss": 1.1992, "step": 6522 }, { "epoch": 0.51, "learning_rate": 1.028277351525847e-06, "loss": 1.1752, "step": 6523 }, { "epoch": 0.51, "learning_rate": 1.0280260634156447e-06, "loss": 1.1966, "step": 6524 }, { "epoch": 0.51, "learning_rate": 1.027774773534313e-06, "loss": 1.2396, "step": 6525 }, { "epoch": 0.51, "learning_rate": 1.0275234818977326e-06, "loss": 1.2049, "step": 6526 }, { "epoch": 0.51, "learning_rate": 1.0272721885217837e-06, "loss": 1.1833, "step": 6527 }, { "epoch": 0.51, "learning_rate": 1.0270208934223469e-06, "loss": 1.1931, "step": 6528 }, { "epoch": 0.51, "learning_rate": 1.0267695966153035e-06, "loss": 1.244, "step": 6529 }, { "epoch": 0.51, "learning_rate": 1.0265182981165334e-06, "loss": 1.2322, "step": 6530 }, { "epoch": 0.51, "learning_rate": 1.026266997941919e-06, "loss": 1.1857, "step": 6531 }, { "epoch": 0.51, "learning_rate": 1.0260156961073404e-06, "loss": 1.1728, "step": 6532 }, { "epoch": 0.51, "learning_rate": 1.025764392628679e-06, "loss": 1.2353, "step": 6533 }, { "epoch": 0.51, "learning_rate": 1.0255130875218165e-06, "loss": 1.238, "step": 6534 }, { "epoch": 0.51, "learning_rate": 1.0252617808026338e-06, "loss": 1.2078, "step": 6535 }, { "epoch": 0.51, "learning_rate": 1.025010472487013e-06, "loss": 1.1657, "step": 6536 }, { "epoch": 0.51, "learning_rate": 1.0247591625908353e-06, "loss": 1.1994, "step": 6537 }, { "epoch": 0.51, "learning_rate": 1.0245078511299826e-06, "loss": 1.2407, "step": 6538 }, { "epoch": 0.51, "learning_rate": 1.0242565381203367e-06, "loss": 1.2174, "step": 6539 }, { "epoch": 0.51, "learning_rate": 1.0240052235777794e-06, "loss": 0.8323, "step": 6540 }, { "epoch": 0.51, "learning_rate": 1.023753907518193e-06, "loss": 1.1701, "step": 6541 }, { "epoch": 0.51, "learning_rate": 1.023502589957459e-06, "loss": 1.1734, "step": 6542 }, { "epoch": 0.51, "learning_rate": 1.0232512709114607e-06, "loss": 1.2343, "step": 6543 }, { "epoch": 0.51, "learning_rate": 1.0229999503960794e-06, "loss": 1.2582, "step": 6544 }, { "epoch": 0.51, "learning_rate": 1.022748628427198e-06, "loss": 1.1926, "step": 6545 }, { "epoch": 0.51, "learning_rate": 1.022497305020699e-06, "loss": 0.8238, "step": 6546 }, { "epoch": 0.51, "learning_rate": 1.0222459801924644e-06, "loss": 1.146, "step": 6547 }, { "epoch": 0.51, "learning_rate": 1.0219946539583777e-06, "loss": 1.2273, "step": 6548 }, { "epoch": 0.51, "learning_rate": 1.021743326334321e-06, "loss": 1.1787, "step": 6549 }, { "epoch": 0.51, "learning_rate": 1.0214919973361778e-06, "loss": 1.2073, "step": 6550 }, { "epoch": 0.51, "learning_rate": 1.0212406669798305e-06, "loss": 1.2065, "step": 6551 }, { "epoch": 0.51, "learning_rate": 1.0209893352811622e-06, "loss": 1.2246, "step": 6552 }, { "epoch": 0.51, "learning_rate": 1.0207380022560564e-06, "loss": 1.201, "step": 6553 }, { "epoch": 0.51, "learning_rate": 1.0204866679203958e-06, "loss": 1.1703, "step": 6554 }, { "epoch": 0.51, "learning_rate": 1.0202353322900637e-06, "loss": 1.2009, "step": 6555 }, { "epoch": 0.51, "learning_rate": 1.0199839953809442e-06, "loss": 1.1627, "step": 6556 }, { "epoch": 0.51, "learning_rate": 1.0197326572089193e-06, "loss": 1.2412, "step": 6557 }, { "epoch": 0.51, "learning_rate": 1.019481317789874e-06, "loss": 1.2187, "step": 6558 }, { "epoch": 0.51, "learning_rate": 1.0192299771396913e-06, "loss": 1.2139, "step": 6559 }, { "epoch": 0.51, "learning_rate": 1.0189786352742545e-06, "loss": 1.1606, "step": 6560 }, { "epoch": 0.51, "learning_rate": 1.018727292209448e-06, "loss": 1.1743, "step": 6561 }, { "epoch": 0.51, "learning_rate": 1.018475947961155e-06, "loss": 1.2278, "step": 6562 }, { "epoch": 0.51, "learning_rate": 1.0182246025452598e-06, "loss": 1.206, "step": 6563 }, { "epoch": 0.51, "learning_rate": 1.0179732559776462e-06, "loss": 1.2495, "step": 6564 }, { "epoch": 0.51, "learning_rate": 1.0177219082741988e-06, "loss": 1.2289, "step": 6565 }, { "epoch": 0.51, "learning_rate": 1.0174705594508011e-06, "loss": 1.1886, "step": 6566 }, { "epoch": 0.51, "learning_rate": 1.0172192095233369e-06, "loss": 1.1553, "step": 6567 }, { "epoch": 0.51, "learning_rate": 1.0169678585076917e-06, "loss": 1.2466, "step": 6568 }, { "epoch": 0.51, "learning_rate": 1.0167165064197488e-06, "loss": 1.1897, "step": 6569 }, { "epoch": 0.51, "learning_rate": 1.016465153275393e-06, "loss": 1.1865, "step": 6570 }, { "epoch": 0.51, "learning_rate": 1.016213799090509e-06, "loss": 1.1518, "step": 6571 }, { "epoch": 0.51, "learning_rate": 1.0159624438809805e-06, "loss": 1.2576, "step": 6572 }, { "epoch": 0.51, "learning_rate": 1.0157110876626931e-06, "loss": 1.1754, "step": 6573 }, { "epoch": 0.51, "learning_rate": 1.0154597304515305e-06, "loss": 1.2202, "step": 6574 }, { "epoch": 0.51, "learning_rate": 1.0152083722633782e-06, "loss": 1.2231, "step": 6575 }, { "epoch": 0.51, "learning_rate": 1.0149570131141208e-06, "loss": 1.2219, "step": 6576 }, { "epoch": 0.51, "learning_rate": 1.0147056530196428e-06, "loss": 1.2222, "step": 6577 }, { "epoch": 0.51, "learning_rate": 1.0144542919958292e-06, "loss": 1.228, "step": 6578 }, { "epoch": 0.51, "learning_rate": 1.0142029300585653e-06, "loss": 1.1352, "step": 6579 }, { "epoch": 0.51, "learning_rate": 1.0139515672237358e-06, "loss": 1.1976, "step": 6580 }, { "epoch": 0.51, "learning_rate": 1.0137002035072261e-06, "loss": 1.202, "step": 6581 }, { "epoch": 0.51, "learning_rate": 1.0134488389249206e-06, "loss": 1.1845, "step": 6582 }, { "epoch": 0.51, "learning_rate": 1.0131974734927054e-06, "loss": 1.1959, "step": 6583 }, { "epoch": 0.51, "learning_rate": 1.012946107226465e-06, "loss": 1.1586, "step": 6584 }, { "epoch": 0.51, "learning_rate": 1.0126947401420856e-06, "loss": 1.2565, "step": 6585 }, { "epoch": 0.51, "learning_rate": 1.0124433722554512e-06, "loss": 1.1962, "step": 6586 }, { "epoch": 0.51, "learning_rate": 1.0121920035824488e-06, "loss": 1.2063, "step": 6587 }, { "epoch": 0.51, "learning_rate": 1.0119406341389625e-06, "loss": 1.2297, "step": 6588 }, { "epoch": 0.51, "learning_rate": 1.011689263940878e-06, "loss": 1.2319, "step": 6589 }, { "epoch": 0.51, "learning_rate": 1.011437893004082e-06, "loss": 1.1978, "step": 6590 }, { "epoch": 0.51, "learning_rate": 1.0111865213444583e-06, "loss": 1.196, "step": 6591 }, { "epoch": 0.51, "learning_rate": 1.010935148977894e-06, "loss": 1.2215, "step": 6592 }, { "epoch": 0.51, "learning_rate": 1.0106837759202742e-06, "loss": 0.834, "step": 6593 }, { "epoch": 0.51, "learning_rate": 1.0104324021874843e-06, "loss": 1.217, "step": 6594 }, { "epoch": 0.51, "learning_rate": 1.0101810277954107e-06, "loss": 1.1856, "step": 6595 }, { "epoch": 0.51, "learning_rate": 1.009929652759939e-06, "loss": 1.2139, "step": 6596 }, { "epoch": 0.51, "learning_rate": 1.009678277096955e-06, "loss": 1.1639, "step": 6597 }, { "epoch": 0.51, "learning_rate": 1.0094269008223447e-06, "loss": 1.2326, "step": 6598 }, { "epoch": 0.51, "learning_rate": 1.0091755239519933e-06, "loss": 1.1715, "step": 6599 }, { "epoch": 0.51, "learning_rate": 1.0089241465017875e-06, "loss": 1.2324, "step": 6600 }, { "epoch": 0.51, "learning_rate": 1.008672768487613e-06, "loss": 0.838, "step": 6601 }, { "epoch": 0.51, "learning_rate": 1.0084213899253565e-06, "loss": 1.1865, "step": 6602 }, { "epoch": 0.51, "learning_rate": 1.0081700108309035e-06, "loss": 1.2078, "step": 6603 }, { "epoch": 0.51, "learning_rate": 1.0079186312201397e-06, "loss": 1.2199, "step": 6604 }, { "epoch": 0.51, "learning_rate": 1.0076672511089522e-06, "loss": 1.2237, "step": 6605 }, { "epoch": 0.51, "learning_rate": 1.007415870513226e-06, "loss": 1.2164, "step": 6606 }, { "epoch": 0.51, "learning_rate": 1.0071644894488484e-06, "loss": 1.2524, "step": 6607 }, { "epoch": 0.51, "learning_rate": 1.006913107931705e-06, "loss": 1.225, "step": 6608 }, { "epoch": 0.51, "learning_rate": 1.0066617259776825e-06, "loss": 1.2465, "step": 6609 }, { "epoch": 0.51, "learning_rate": 1.0064103436026666e-06, "loss": 1.1773, "step": 6610 }, { "epoch": 0.51, "learning_rate": 1.006158960822544e-06, "loss": 1.1886, "step": 6611 }, { "epoch": 0.51, "learning_rate": 1.0059075776532011e-06, "loss": 1.2297, "step": 6612 }, { "epoch": 0.51, "learning_rate": 1.005656194110524e-06, "loss": 1.2218, "step": 6613 }, { "epoch": 0.51, "learning_rate": 1.0054048102103992e-06, "loss": 1.2047, "step": 6614 }, { "epoch": 0.51, "learning_rate": 1.0051534259687133e-06, "loss": 1.2037, "step": 6615 }, { "epoch": 0.51, "learning_rate": 1.004902041401352e-06, "loss": 1.215, "step": 6616 }, { "epoch": 0.51, "learning_rate": 1.0046506565242027e-06, "loss": 1.2147, "step": 6617 }, { "epoch": 0.51, "learning_rate": 1.0043992713531512e-06, "loss": 1.1952, "step": 6618 }, { "epoch": 0.51, "learning_rate": 1.0041478859040846e-06, "loss": 0.7907, "step": 6619 }, { "epoch": 0.51, "learning_rate": 1.003896500192889e-06, "loss": 1.1722, "step": 6620 }, { "epoch": 0.51, "learning_rate": 1.0036451142354508e-06, "loss": 1.1811, "step": 6621 }, { "epoch": 0.51, "learning_rate": 1.0033937280476565e-06, "loss": 1.2289, "step": 6622 }, { "epoch": 0.51, "learning_rate": 1.003142341645393e-06, "loss": 1.2499, "step": 6623 }, { "epoch": 0.51, "learning_rate": 1.002890955044547e-06, "loss": 1.2195, "step": 6624 }, { "epoch": 0.51, "learning_rate": 1.0026395682610044e-06, "loss": 1.2062, "step": 6625 }, { "epoch": 0.51, "learning_rate": 1.0023881813106522e-06, "loss": 1.1832, "step": 6626 }, { "epoch": 0.51, "learning_rate": 1.0021367942093772e-06, "loss": 1.2378, "step": 6627 }, { "epoch": 0.51, "learning_rate": 1.0018854069730656e-06, "loss": 1.221, "step": 6628 }, { "epoch": 0.51, "learning_rate": 1.0016340196176043e-06, "loss": 1.156, "step": 6629 }, { "epoch": 0.51, "learning_rate": 1.0013826321588798e-06, "loss": 1.2078, "step": 6630 }, { "epoch": 0.51, "learning_rate": 1.0011312446127789e-06, "loss": 1.2254, "step": 6631 }, { "epoch": 0.51, "learning_rate": 1.0008798569951881e-06, "loss": 1.2499, "step": 6632 }, { "epoch": 0.51, "learning_rate": 1.0006284693219937e-06, "loss": 1.2057, "step": 6633 }, { "epoch": 0.51, "learning_rate": 1.000377081609083e-06, "loss": 0.8388, "step": 6634 }, { "epoch": 0.51, "learning_rate": 1.000125693872342e-06, "loss": 1.226, "step": 6635 }, { "epoch": 0.52, "learning_rate": 9.99874306127658e-07, "loss": 1.1796, "step": 6636 }, { "epoch": 0.52, "learning_rate": 9.996229183909172e-07, "loss": 1.2122, "step": 6637 }, { "epoch": 0.52, "learning_rate": 9.993715306780064e-07, "loss": 1.1542, "step": 6638 }, { "epoch": 0.52, "learning_rate": 9.991201430048122e-07, "loss": 0.8701, "step": 6639 }, { "epoch": 0.52, "learning_rate": 9.98868755387221e-07, "loss": 1.2135, "step": 6640 }, { "epoch": 0.52, "learning_rate": 9.986173678411203e-07, "loss": 1.2185, "step": 6641 }, { "epoch": 0.52, "learning_rate": 9.983659803823957e-07, "loss": 1.2002, "step": 6642 }, { "epoch": 0.52, "learning_rate": 9.981145930269343e-07, "loss": 1.2109, "step": 6643 }, { "epoch": 0.52, "learning_rate": 9.97863205790623e-07, "loss": 0.8042, "step": 6644 }, { "epoch": 0.52, "learning_rate": 9.976118186893477e-07, "loss": 1.1835, "step": 6645 }, { "epoch": 0.52, "learning_rate": 9.973604317389957e-07, "loss": 1.2114, "step": 6646 }, { "epoch": 0.52, "learning_rate": 9.971090449554533e-07, "loss": 1.2079, "step": 6647 }, { "epoch": 0.52, "learning_rate": 9.968576583546068e-07, "loss": 1.2316, "step": 6648 }, { "epoch": 0.52, "learning_rate": 9.966062719523436e-07, "loss": 1.1553, "step": 6649 }, { "epoch": 0.52, "learning_rate": 9.963548857645493e-07, "loss": 1.1744, "step": 6650 }, { "epoch": 0.52, "learning_rate": 9.961034998071112e-07, "loss": 1.1801, "step": 6651 }, { "epoch": 0.52, "learning_rate": 9.958521140959153e-07, "loss": 1.2403, "step": 6652 }, { "epoch": 0.52, "learning_rate": 9.956007286468487e-07, "loss": 1.2083, "step": 6653 }, { "epoch": 0.52, "learning_rate": 9.953493434757974e-07, "loss": 0.9055, "step": 6654 }, { "epoch": 0.52, "learning_rate": 9.95097958598648e-07, "loss": 1.2588, "step": 6655 }, { "epoch": 0.52, "learning_rate": 9.94846574031287e-07, "loss": 1.2119, "step": 6656 }, { "epoch": 0.52, "learning_rate": 9.945951897896007e-07, "loss": 1.2368, "step": 6657 }, { "epoch": 0.52, "learning_rate": 9.94343805889476e-07, "loss": 1.2365, "step": 6658 }, { "epoch": 0.52, "learning_rate": 9.94092422346799e-07, "loss": 0.8607, "step": 6659 }, { "epoch": 0.52, "learning_rate": 9.938410391774558e-07, "loss": 1.2554, "step": 6660 }, { "epoch": 0.52, "learning_rate": 9.935896563973335e-07, "loss": 1.2136, "step": 6661 }, { "epoch": 0.52, "learning_rate": 9.933382740223174e-07, "loss": 1.2049, "step": 6662 }, { "epoch": 0.52, "learning_rate": 9.93086892068295e-07, "loss": 0.8012, "step": 6663 }, { "epoch": 0.52, "learning_rate": 9.928355105511517e-07, "loss": 0.8481, "step": 6664 }, { "epoch": 0.52, "learning_rate": 9.92584129486774e-07, "loss": 1.2387, "step": 6665 }, { "epoch": 0.52, "learning_rate": 9.923327488910482e-07, "loss": 1.1978, "step": 6666 }, { "epoch": 0.52, "learning_rate": 9.920813687798602e-07, "loss": 1.2126, "step": 6667 }, { "epoch": 0.52, "learning_rate": 9.918299891690966e-07, "loss": 1.1934, "step": 6668 }, { "epoch": 0.52, "learning_rate": 9.915786100746436e-07, "loss": 1.2082, "step": 6669 }, { "epoch": 0.52, "learning_rate": 9.913272315123867e-07, "loss": 1.2169, "step": 6670 }, { "epoch": 0.52, "learning_rate": 9.910758534982124e-07, "loss": 1.2267, "step": 6671 }, { "epoch": 0.52, "learning_rate": 9.908244760480068e-07, "loss": 1.231, "step": 6672 }, { "epoch": 0.52, "learning_rate": 9.905730991776557e-07, "loss": 1.2051, "step": 6673 }, { "epoch": 0.52, "learning_rate": 9.90321722903045e-07, "loss": 1.1637, "step": 6674 }, { "epoch": 0.52, "learning_rate": 9.90070347240061e-07, "loss": 1.2462, "step": 6675 }, { "epoch": 0.52, "learning_rate": 9.898189722045892e-07, "loss": 1.2424, "step": 6676 }, { "epoch": 0.52, "learning_rate": 9.895675978125154e-07, "loss": 1.2101, "step": 6677 }, { "epoch": 0.52, "learning_rate": 9.893162240797261e-07, "loss": 1.1593, "step": 6678 }, { "epoch": 0.52, "learning_rate": 9.89064851022106e-07, "loss": 1.1903, "step": 6679 }, { "epoch": 0.52, "learning_rate": 9.888134786555416e-07, "loss": 1.2155, "step": 6680 }, { "epoch": 0.52, "learning_rate": 9.885621069959184e-07, "loss": 1.2106, "step": 6681 }, { "epoch": 0.52, "learning_rate": 9.883107360591217e-07, "loss": 0.8446, "step": 6682 }, { "epoch": 0.52, "learning_rate": 9.880593658610378e-07, "loss": 1.2271, "step": 6683 }, { "epoch": 0.52, "learning_rate": 9.878079964175513e-07, "loss": 1.1905, "step": 6684 }, { "epoch": 0.52, "learning_rate": 9.875566277445485e-07, "loss": 1.1946, "step": 6685 }, { "epoch": 0.52, "learning_rate": 9.873052598579148e-07, "loss": 1.2352, "step": 6686 }, { "epoch": 0.52, "learning_rate": 9.870538927735347e-07, "loss": 1.27, "step": 6687 }, { "epoch": 0.52, "learning_rate": 9.868025265072947e-07, "loss": 1.1993, "step": 6688 }, { "epoch": 0.52, "learning_rate": 9.865511610750793e-07, "loss": 1.2435, "step": 6689 }, { "epoch": 0.52, "learning_rate": 9.862997964927742e-07, "loss": 1.1582, "step": 6690 }, { "epoch": 0.52, "learning_rate": 9.860484327762645e-07, "loss": 1.1957, "step": 6691 }, { "epoch": 0.52, "learning_rate": 9.857970699414348e-07, "loss": 1.2065, "step": 6692 }, { "epoch": 0.52, "learning_rate": 9.855457080041707e-07, "loss": 1.2169, "step": 6693 }, { "epoch": 0.52, "learning_rate": 9.852943469803571e-07, "loss": 1.1999, "step": 6694 }, { "epoch": 0.52, "learning_rate": 9.850429868858793e-07, "loss": 1.1411, "step": 6695 }, { "epoch": 0.52, "learning_rate": 9.847916277366217e-07, "loss": 1.1664, "step": 6696 }, { "epoch": 0.52, "learning_rate": 9.845402695484694e-07, "loss": 1.2303, "step": 6697 }, { "epoch": 0.52, "learning_rate": 9.842889123373072e-07, "loss": 1.2024, "step": 6698 }, { "epoch": 0.52, "learning_rate": 9.840375561190192e-07, "loss": 1.2174, "step": 6699 }, { "epoch": 0.52, "learning_rate": 9.837862009094912e-07, "loss": 1.2336, "step": 6700 }, { "epoch": 0.52, "learning_rate": 9.835348467246068e-07, "loss": 1.2136, "step": 6701 }, { "epoch": 0.52, "learning_rate": 9.83283493580251e-07, "loss": 1.1395, "step": 6702 }, { "epoch": 0.52, "learning_rate": 9.830321414923084e-07, "loss": 1.2412, "step": 6703 }, { "epoch": 0.52, "learning_rate": 9.827807904766628e-07, "loss": 1.194, "step": 6704 }, { "epoch": 0.52, "learning_rate": 9.82529440549199e-07, "loss": 1.1778, "step": 6705 }, { "epoch": 0.52, "learning_rate": 9.822780917258013e-07, "loss": 1.2376, "step": 6706 }, { "epoch": 0.52, "learning_rate": 9.820267440223535e-07, "loss": 1.1395, "step": 6707 }, { "epoch": 0.52, "learning_rate": 9.817753974547403e-07, "loss": 1.2545, "step": 6708 }, { "epoch": 0.52, "learning_rate": 9.81524052038845e-07, "loss": 1.1893, "step": 6709 }, { "epoch": 0.52, "learning_rate": 9.812727077905522e-07, "loss": 1.2431, "step": 6710 }, { "epoch": 0.52, "learning_rate": 9.810213647257455e-07, "loss": 1.218, "step": 6711 }, { "epoch": 0.52, "learning_rate": 9.80770022860309e-07, "loss": 1.1955, "step": 6712 }, { "epoch": 0.52, "learning_rate": 9.805186822101262e-07, "loss": 1.1653, "step": 6713 }, { "epoch": 0.52, "learning_rate": 9.802673427910808e-07, "loss": 1.1601, "step": 6714 }, { "epoch": 0.52, "learning_rate": 9.800160046190561e-07, "loss": 1.1766, "step": 6715 }, { "epoch": 0.52, "learning_rate": 9.79764667709936e-07, "loss": 1.2389, "step": 6716 }, { "epoch": 0.52, "learning_rate": 9.795133320796043e-07, "loss": 1.2269, "step": 6717 }, { "epoch": 0.52, "learning_rate": 9.792619977439435e-07, "loss": 0.8469, "step": 6718 }, { "epoch": 0.52, "learning_rate": 9.790106647188377e-07, "loss": 1.1733, "step": 6719 }, { "epoch": 0.52, "learning_rate": 9.787593330201696e-07, "loss": 1.1767, "step": 6720 }, { "epoch": 0.52, "learning_rate": 9.78508002663822e-07, "loss": 1.1829, "step": 6721 }, { "epoch": 0.52, "learning_rate": 9.78256673665679e-07, "loss": 1.2098, "step": 6722 }, { "epoch": 0.52, "learning_rate": 9.780053460416223e-07, "loss": 1.203, "step": 6723 }, { "epoch": 0.52, "learning_rate": 9.777540198075355e-07, "loss": 1.1945, "step": 6724 }, { "epoch": 0.52, "learning_rate": 9.775026949793014e-07, "loss": 0.8521, "step": 6725 }, { "epoch": 0.52, "learning_rate": 9.77251371572802e-07, "loss": 1.2094, "step": 6726 }, { "epoch": 0.52, "learning_rate": 9.770000496039205e-07, "loss": 1.2039, "step": 6727 }, { "epoch": 0.52, "learning_rate": 9.767487290885392e-07, "loss": 1.2283, "step": 6728 }, { "epoch": 0.52, "learning_rate": 9.764974100425408e-07, "loss": 1.2181, "step": 6729 }, { "epoch": 0.52, "learning_rate": 9.762460924818071e-07, "loss": 1.2317, "step": 6730 }, { "epoch": 0.52, "learning_rate": 9.759947764222207e-07, "loss": 1.2071, "step": 6731 }, { "epoch": 0.52, "learning_rate": 9.757434618796634e-07, "loss": 1.217, "step": 6732 }, { "epoch": 0.52, "learning_rate": 9.754921488700173e-07, "loss": 1.1908, "step": 6733 }, { "epoch": 0.52, "learning_rate": 9.752408374091648e-07, "loss": 1.1554, "step": 6734 }, { "epoch": 0.52, "learning_rate": 9.74989527512987e-07, "loss": 1.1774, "step": 6735 }, { "epoch": 0.52, "learning_rate": 9.747382191973659e-07, "loss": 1.2183, "step": 6736 }, { "epoch": 0.52, "learning_rate": 9.744869124781837e-07, "loss": 1.1784, "step": 6737 }, { "epoch": 0.52, "learning_rate": 9.742356073713208e-07, "loss": 1.2426, "step": 6738 }, { "epoch": 0.52, "learning_rate": 9.7398430389266e-07, "loss": 1.236, "step": 6739 }, { "epoch": 0.52, "learning_rate": 9.73733002058081e-07, "loss": 1.197, "step": 6740 }, { "epoch": 0.52, "learning_rate": 9.734817018834663e-07, "loss": 1.2435, "step": 6741 }, { "epoch": 0.52, "learning_rate": 9.73230403384697e-07, "loss": 1.2403, "step": 6742 }, { "epoch": 0.52, "learning_rate": 9.729791065776532e-07, "loss": 0.8047, "step": 6743 }, { "epoch": 0.52, "learning_rate": 9.727278114782164e-07, "loss": 1.2464, "step": 6744 }, { "epoch": 0.52, "learning_rate": 9.724765181022673e-07, "loss": 1.2528, "step": 6745 }, { "epoch": 0.52, "learning_rate": 9.722252264656869e-07, "loss": 1.216, "step": 6746 }, { "epoch": 0.52, "learning_rate": 9.719739365843554e-07, "loss": 1.2271, "step": 6747 }, { "epoch": 0.52, "learning_rate": 9.71722648474153e-07, "loss": 1.2185, "step": 6748 }, { "epoch": 0.52, "learning_rate": 9.714713621509608e-07, "loss": 1.1966, "step": 6749 }, { "epoch": 0.52, "learning_rate": 9.712200776306584e-07, "loss": 1.2266, "step": 6750 }, { "epoch": 0.52, "learning_rate": 9.709687949291265e-07, "loss": 1.1901, "step": 6751 }, { "epoch": 0.52, "learning_rate": 9.707175140622445e-07, "loss": 1.1635, "step": 6752 }, { "epoch": 0.52, "learning_rate": 9.704662350458923e-07, "loss": 1.1583, "step": 6753 }, { "epoch": 0.52, "learning_rate": 9.702149578959506e-07, "loss": 1.2158, "step": 6754 }, { "epoch": 0.52, "learning_rate": 9.69963682628298e-07, "loss": 0.8853, "step": 6755 }, { "epoch": 0.52, "learning_rate": 9.697124092588145e-07, "loss": 1.1908, "step": 6756 }, { "epoch": 0.52, "learning_rate": 9.694611378033798e-07, "loss": 1.1688, "step": 6757 }, { "epoch": 0.52, "learning_rate": 9.692098682778723e-07, "loss": 1.1986, "step": 6758 }, { "epoch": 0.52, "learning_rate": 9.689586006981721e-07, "loss": 1.2423, "step": 6759 }, { "epoch": 0.52, "learning_rate": 9.687073350801577e-07, "loss": 1.2492, "step": 6760 }, { "epoch": 0.52, "learning_rate": 9.684560714397083e-07, "loss": 1.2226, "step": 6761 }, { "epoch": 0.52, "learning_rate": 9.682048097927026e-07, "loss": 1.2115, "step": 6762 }, { "epoch": 0.52, "learning_rate": 9.679535501550195e-07, "loss": 1.2197, "step": 6763 }, { "epoch": 0.52, "learning_rate": 9.677022925425374e-07, "loss": 1.154, "step": 6764 }, { "epoch": 0.53, "learning_rate": 9.674510369711346e-07, "loss": 1.1689, "step": 6765 }, { "epoch": 0.53, "learning_rate": 9.671997834566894e-07, "loss": 1.1595, "step": 6766 }, { "epoch": 0.53, "learning_rate": 9.6694853201508e-07, "loss": 1.1927, "step": 6767 }, { "epoch": 0.53, "learning_rate": 9.66697282662185e-07, "loss": 1.2289, "step": 6768 }, { "epoch": 0.53, "learning_rate": 9.664460354138815e-07, "loss": 0.8071, "step": 6769 }, { "epoch": 0.53, "learning_rate": 9.661947902860474e-07, "loss": 1.1879, "step": 6770 }, { "epoch": 0.53, "learning_rate": 9.659435472945608e-07, "loss": 1.2751, "step": 6771 }, { "epoch": 0.53, "learning_rate": 9.656923064552986e-07, "loss": 1.1938, "step": 6772 }, { "epoch": 0.53, "learning_rate": 9.654410677841387e-07, "loss": 1.2408, "step": 6773 }, { "epoch": 0.53, "learning_rate": 9.651898312969583e-07, "loss": 1.1816, "step": 6774 }, { "epoch": 0.53, "learning_rate": 9.649385970096337e-07, "loss": 1.2367, "step": 6775 }, { "epoch": 0.53, "learning_rate": 9.64687364938043e-07, "loss": 1.2262, "step": 6776 }, { "epoch": 0.53, "learning_rate": 9.644361350980622e-07, "loss": 0.7959, "step": 6777 }, { "epoch": 0.53, "learning_rate": 9.641849075055681e-07, "loss": 1.1694, "step": 6778 }, { "epoch": 0.53, "learning_rate": 9.639336821764378e-07, "loss": 1.2177, "step": 6779 }, { "epoch": 0.53, "learning_rate": 9.636824591265467e-07, "loss": 0.8384, "step": 6780 }, { "epoch": 0.53, "learning_rate": 9.63431238371772e-07, "loss": 1.1929, "step": 6781 }, { "epoch": 0.53, "learning_rate": 9.631800199279892e-07, "loss": 1.1942, "step": 6782 }, { "epoch": 0.53, "learning_rate": 9.629288038110746e-07, "loss": 1.1599, "step": 6783 }, { "epoch": 0.53, "learning_rate": 9.626775900369035e-07, "loss": 1.1716, "step": 6784 }, { "epoch": 0.53, "learning_rate": 9.624263786213523e-07, "loss": 1.1782, "step": 6785 }, { "epoch": 0.53, "learning_rate": 9.621751695802958e-07, "loss": 0.853, "step": 6786 }, { "epoch": 0.53, "learning_rate": 9.619239629296095e-07, "loss": 1.219, "step": 6787 }, { "epoch": 0.53, "learning_rate": 9.616727586851693e-07, "loss": 1.1803, "step": 6788 }, { "epoch": 0.53, "learning_rate": 9.614215568628492e-07, "loss": 1.2112, "step": 6789 }, { "epoch": 0.53, "learning_rate": 9.61170357478525e-07, "loss": 1.243, "step": 6790 }, { "epoch": 0.53, "learning_rate": 9.609191605480708e-07, "loss": 1.2538, "step": 6791 }, { "epoch": 0.53, "learning_rate": 9.606679660873613e-07, "loss": 1.234, "step": 6792 }, { "epoch": 0.53, "learning_rate": 9.604167741122717e-07, "loss": 1.2404, "step": 6793 }, { "epoch": 0.53, "learning_rate": 9.60165584638675e-07, "loss": 1.1671, "step": 6794 }, { "epoch": 0.53, "learning_rate": 9.599143976824461e-07, "loss": 1.2819, "step": 6795 }, { "epoch": 0.53, "learning_rate": 9.596632132594593e-07, "loss": 1.2142, "step": 6796 }, { "epoch": 0.53, "learning_rate": 9.594120313855872e-07, "loss": 1.2288, "step": 6797 }, { "epoch": 0.53, "learning_rate": 9.59160852076705e-07, "loss": 1.2413, "step": 6798 }, { "epoch": 0.53, "learning_rate": 9.589096753486846e-07, "loss": 1.2085, "step": 6799 }, { "epoch": 0.53, "learning_rate": 9.586585012174003e-07, "loss": 1.1743, "step": 6800 }, { "epoch": 0.53, "learning_rate": 9.584073296987253e-07, "loss": 1.1722, "step": 6801 }, { "epoch": 0.53, "learning_rate": 9.58156160808532e-07, "loss": 1.1919, "step": 6802 }, { "epoch": 0.53, "learning_rate": 9.579049945626934e-07, "loss": 1.1595, "step": 6803 }, { "epoch": 0.53, "learning_rate": 9.576538309770823e-07, "loss": 1.1582, "step": 6804 }, { "epoch": 0.53, "learning_rate": 9.574026700675715e-07, "loss": 0.8374, "step": 6805 }, { "epoch": 0.53, "learning_rate": 9.571515118500326e-07, "loss": 1.1897, "step": 6806 }, { "epoch": 0.53, "learning_rate": 9.569003563403386e-07, "loss": 1.2359, "step": 6807 }, { "epoch": 0.53, "learning_rate": 9.566492035543606e-07, "loss": 1.1712, "step": 6808 }, { "epoch": 0.53, "learning_rate": 9.563980535079706e-07, "loss": 1.1788, "step": 6809 }, { "epoch": 0.53, "learning_rate": 9.56146906217041e-07, "loss": 1.2309, "step": 6810 }, { "epoch": 0.53, "learning_rate": 9.55895761697442e-07, "loss": 1.2225, "step": 6811 }, { "epoch": 0.53, "learning_rate": 9.556446199650458e-07, "loss": 1.2241, "step": 6812 }, { "epoch": 0.53, "learning_rate": 9.553934810357236e-07, "loss": 0.8276, "step": 6813 }, { "epoch": 0.53, "learning_rate": 9.551423449253454e-07, "loss": 1.2207, "step": 6814 }, { "epoch": 0.53, "learning_rate": 9.54891211649783e-07, "loss": 1.1937, "step": 6815 }, { "epoch": 0.53, "learning_rate": 9.54640081224906e-07, "loss": 1.2651, "step": 6816 }, { "epoch": 0.53, "learning_rate": 9.543889536665857e-07, "loss": 1.2223, "step": 6817 }, { "epoch": 0.53, "learning_rate": 9.54137828990692e-07, "loss": 1.1847, "step": 6818 }, { "epoch": 0.53, "learning_rate": 9.538867072130943e-07, "loss": 1.1612, "step": 6819 }, { "epoch": 0.53, "learning_rate": 9.536355883496631e-07, "loss": 0.8799, "step": 6820 }, { "epoch": 0.53, "learning_rate": 9.533844724162678e-07, "loss": 1.2589, "step": 6821 }, { "epoch": 0.53, "learning_rate": 9.531333594287785e-07, "loss": 1.1986, "step": 6822 }, { "epoch": 0.53, "learning_rate": 9.528822494030636e-07, "loss": 1.1982, "step": 6823 }, { "epoch": 0.53, "learning_rate": 9.526311423549926e-07, "loss": 1.2107, "step": 6824 }, { "epoch": 0.53, "learning_rate": 9.523800383004343e-07, "loss": 1.208, "step": 6825 }, { "epoch": 0.53, "learning_rate": 9.521289372552573e-07, "loss": 1.2221, "step": 6826 }, { "epoch": 0.53, "learning_rate": 9.518778392353309e-07, "loss": 1.1919, "step": 6827 }, { "epoch": 0.53, "learning_rate": 9.516267442565222e-07, "loss": 1.248, "step": 6828 }, { "epoch": 0.53, "learning_rate": 9.513756523347003e-07, "loss": 1.2293, "step": 6829 }, { "epoch": 0.53, "learning_rate": 9.511245634857331e-07, "loss": 1.1528, "step": 6830 }, { "epoch": 0.53, "learning_rate": 9.508734777254876e-07, "loss": 0.8328, "step": 6831 }, { "epoch": 0.53, "learning_rate": 9.506223950698324e-07, "loss": 1.1633, "step": 6832 }, { "epoch": 0.53, "learning_rate": 9.50371315534634e-07, "loss": 0.859, "step": 6833 }, { "epoch": 0.53, "learning_rate": 9.501202391357601e-07, "loss": 1.2209, "step": 6834 }, { "epoch": 0.53, "learning_rate": 9.498691658890777e-07, "loss": 1.2, "step": 6835 }, { "epoch": 0.53, "learning_rate": 9.49618095810453e-07, "loss": 0.8588, "step": 6836 }, { "epoch": 0.53, "learning_rate": 9.49367028915753e-07, "loss": 1.2491, "step": 6837 }, { "epoch": 0.53, "learning_rate": 9.491159652208441e-07, "loss": 0.8443, "step": 6838 }, { "epoch": 0.53, "learning_rate": 9.488649047415924e-07, "loss": 1.2297, "step": 6839 }, { "epoch": 0.53, "learning_rate": 9.48613847493864e-07, "loss": 1.2001, "step": 6840 }, { "epoch": 0.53, "learning_rate": 9.483627934935243e-07, "loss": 1.201, "step": 6841 }, { "epoch": 0.53, "learning_rate": 9.481117427564393e-07, "loss": 1.2206, "step": 6842 }, { "epoch": 0.53, "learning_rate": 9.478606952984737e-07, "loss": 1.1963, "step": 6843 }, { "epoch": 0.53, "learning_rate": 9.476096511354936e-07, "loss": 1.1644, "step": 6844 }, { "epoch": 0.53, "learning_rate": 9.473586102833631e-07, "loss": 1.2126, "step": 6845 }, { "epoch": 0.53, "learning_rate": 9.471075727579472e-07, "loss": 1.1832, "step": 6846 }, { "epoch": 0.53, "learning_rate": 9.468565385751108e-07, "loss": 1.2501, "step": 6847 }, { "epoch": 0.53, "learning_rate": 9.466055077507175e-07, "loss": 1.2323, "step": 6848 }, { "epoch": 0.53, "learning_rate": 9.463544803006321e-07, "loss": 1.2647, "step": 6849 }, { "epoch": 0.53, "learning_rate": 9.461034562407177e-07, "loss": 1.2422, "step": 6850 }, { "epoch": 0.53, "learning_rate": 9.458524355868387e-07, "loss": 1.2193, "step": 6851 }, { "epoch": 0.53, "learning_rate": 9.456014183548584e-07, "loss": 1.1572, "step": 6852 }, { "epoch": 0.53, "learning_rate": 9.453504045606395e-07, "loss": 1.2309, "step": 6853 }, { "epoch": 0.53, "learning_rate": 9.450993942200456e-07, "loss": 1.253, "step": 6854 }, { "epoch": 0.53, "learning_rate": 9.448483873489393e-07, "loss": 1.2295, "step": 6855 }, { "epoch": 0.53, "learning_rate": 9.445973839631831e-07, "loss": 1.1726, "step": 6856 }, { "epoch": 0.53, "learning_rate": 9.443463840786395e-07, "loss": 1.2184, "step": 6857 }, { "epoch": 0.53, "learning_rate": 9.440953877111706e-07, "loss": 1.2364, "step": 6858 }, { "epoch": 0.53, "learning_rate": 9.438443948766384e-07, "loss": 1.2297, "step": 6859 }, { "epoch": 0.53, "learning_rate": 9.435934055909041e-07, "loss": 1.2122, "step": 6860 }, { "epoch": 0.53, "learning_rate": 9.433424198698301e-07, "loss": 1.2183, "step": 6861 }, { "epoch": 0.53, "learning_rate": 9.430914377292768e-07, "loss": 1.2454, "step": 6862 }, { "epoch": 0.53, "learning_rate": 9.428404591851053e-07, "loss": 1.2004, "step": 6863 }, { "epoch": 0.53, "learning_rate": 9.425894842531771e-07, "loss": 1.1713, "step": 6864 }, { "epoch": 0.53, "learning_rate": 9.423385129493519e-07, "loss": 1.1546, "step": 6865 }, { "epoch": 0.53, "learning_rate": 9.420875452894907e-07, "loss": 1.239, "step": 6866 }, { "epoch": 0.53, "learning_rate": 9.418365812894531e-07, "loss": 1.2173, "step": 6867 }, { "epoch": 0.53, "learning_rate": 9.41585620965099e-07, "loss": 1.1795, "step": 6868 }, { "epoch": 0.53, "learning_rate": 9.413346643322885e-07, "loss": 1.1769, "step": 6869 }, { "epoch": 0.53, "learning_rate": 9.410837114068806e-07, "loss": 1.197, "step": 6870 }, { "epoch": 0.53, "learning_rate": 9.408327622047345e-07, "loss": 0.8182, "step": 6871 }, { "epoch": 0.53, "learning_rate": 9.405818167417093e-07, "loss": 1.19, "step": 6872 }, { "epoch": 0.53, "learning_rate": 9.403308750336637e-07, "loss": 1.237, "step": 6873 }, { "epoch": 0.53, "learning_rate": 9.400799370964563e-07, "loss": 0.8284, "step": 6874 }, { "epoch": 0.53, "learning_rate": 9.398290029459444e-07, "loss": 1.2301, "step": 6875 }, { "epoch": 0.53, "learning_rate": 9.395780725979873e-07, "loss": 1.2185, "step": 6876 }, { "epoch": 0.53, "learning_rate": 9.393271460684417e-07, "loss": 1.1859, "step": 6877 }, { "epoch": 0.53, "learning_rate": 9.39076223373166e-07, "loss": 1.2037, "step": 6878 }, { "epoch": 0.53, "learning_rate": 9.388253045280166e-07, "loss": 1.1953, "step": 6879 }, { "epoch": 0.53, "learning_rate": 9.385743895488509e-07, "loss": 1.2035, "step": 6880 }, { "epoch": 0.53, "learning_rate": 9.383234784515261e-07, "loss": 1.2248, "step": 6881 }, { "epoch": 0.53, "learning_rate": 9.380725712518977e-07, "loss": 1.1637, "step": 6882 }, { "epoch": 0.53, "learning_rate": 9.378216679658233e-07, "loss": 1.2681, "step": 6883 }, { "epoch": 0.53, "learning_rate": 9.375707686091578e-07, "loss": 1.1848, "step": 6884 }, { "epoch": 0.53, "learning_rate": 9.373198731977573e-07, "loss": 1.2129, "step": 6885 }, { "epoch": 0.53, "learning_rate": 9.370689817474778e-07, "loss": 1.1374, "step": 6886 }, { "epoch": 0.53, "learning_rate": 9.368180942741739e-07, "loss": 1.2398, "step": 6887 }, { "epoch": 0.53, "learning_rate": 9.36567210793701e-07, "loss": 1.2202, "step": 6888 }, { "epoch": 0.53, "learning_rate": 9.363163313219142e-07, "loss": 1.1761, "step": 6889 }, { "epoch": 0.53, "learning_rate": 9.360654558746673e-07, "loss": 1.2485, "step": 6890 }, { "epoch": 0.53, "learning_rate": 9.358145844678152e-07, "loss": 1.1819, "step": 6891 }, { "epoch": 0.53, "learning_rate": 9.355637171172113e-07, "loss": 1.2469, "step": 6892 }, { "epoch": 0.53, "learning_rate": 9.353128538387101e-07, "loss": 1.2329, "step": 6893 }, { "epoch": 0.54, "learning_rate": 9.350619946481643e-07, "loss": 1.2056, "step": 6894 }, { "epoch": 0.54, "learning_rate": 9.348111395614281e-07, "loss": 0.8636, "step": 6895 }, { "epoch": 0.54, "learning_rate": 9.345602885943538e-07, "loss": 1.1968, "step": 6896 }, { "epoch": 0.54, "learning_rate": 9.343094417627939e-07, "loss": 1.2087, "step": 6897 }, { "epoch": 0.54, "learning_rate": 9.340585990826018e-07, "loss": 1.2004, "step": 6898 }, { "epoch": 0.54, "learning_rate": 9.338077605696289e-07, "loss": 1.1722, "step": 6899 }, { "epoch": 0.54, "learning_rate": 9.335569262397276e-07, "loss": 0.8422, "step": 6900 }, { "epoch": 0.54, "learning_rate": 9.333060961087493e-07, "loss": 1.1978, "step": 6901 }, { "epoch": 0.54, "learning_rate": 9.330552701925452e-07, "loss": 1.2102, "step": 6902 }, { "epoch": 0.54, "learning_rate": 9.328044485069671e-07, "loss": 1.1899, "step": 6903 }, { "epoch": 0.54, "learning_rate": 9.325536310678652e-07, "loss": 1.1242, "step": 6904 }, { "epoch": 0.54, "learning_rate": 9.323028178910906e-07, "loss": 1.2191, "step": 6905 }, { "epoch": 0.54, "learning_rate": 9.320520089924937e-07, "loss": 0.8286, "step": 6906 }, { "epoch": 0.54, "learning_rate": 9.318012043879239e-07, "loss": 1.2117, "step": 6907 }, { "epoch": 0.54, "learning_rate": 9.315504040932318e-07, "loss": 1.195, "step": 6908 }, { "epoch": 0.54, "learning_rate": 9.312996081242661e-07, "loss": 1.2215, "step": 6909 }, { "epoch": 0.54, "learning_rate": 9.310488164968768e-07, "loss": 1.2599, "step": 6910 }, { "epoch": 0.54, "learning_rate": 9.307980292269127e-07, "loss": 1.2352, "step": 6911 }, { "epoch": 0.54, "learning_rate": 9.30547246330222e-07, "loss": 1.1522, "step": 6912 }, { "epoch": 0.54, "learning_rate": 9.302964678226538e-07, "loss": 1.1792, "step": 6913 }, { "epoch": 0.54, "learning_rate": 9.300456937200558e-07, "loss": 1.1691, "step": 6914 }, { "epoch": 0.54, "learning_rate": 9.297949240382762e-07, "loss": 1.1992, "step": 6915 }, { "epoch": 0.54, "learning_rate": 9.29544158793162e-07, "loss": 1.2077, "step": 6916 }, { "epoch": 0.54, "learning_rate": 9.292933980005615e-07, "loss": 1.1554, "step": 6917 }, { "epoch": 0.54, "learning_rate": 9.290426416763209e-07, "loss": 1.211, "step": 6918 }, { "epoch": 0.54, "learning_rate": 9.287918898362869e-07, "loss": 1.2001, "step": 6919 }, { "epoch": 0.54, "learning_rate": 9.285411424963068e-07, "loss": 1.1978, "step": 6920 }, { "epoch": 0.54, "learning_rate": 9.282903996722257e-07, "loss": 1.2731, "step": 6921 }, { "epoch": 0.54, "learning_rate": 9.280396613798903e-07, "loss": 1.2342, "step": 6922 }, { "epoch": 0.54, "learning_rate": 9.277889276351462e-07, "loss": 1.1359, "step": 6923 }, { "epoch": 0.54, "learning_rate": 9.275381984538378e-07, "loss": 1.2005, "step": 6924 }, { "epoch": 0.54, "learning_rate": 9.272874738518115e-07, "loss": 1.1884, "step": 6925 }, { "epoch": 0.54, "learning_rate": 9.270367538449108e-07, "loss": 1.2338, "step": 6926 }, { "epoch": 0.54, "learning_rate": 9.267860384489809e-07, "loss": 1.2359, "step": 6927 }, { "epoch": 0.54, "learning_rate": 9.265353276798659e-07, "loss": 1.2547, "step": 6928 }, { "epoch": 0.54, "learning_rate": 9.26284621553409e-07, "loss": 1.1874, "step": 6929 }, { "epoch": 0.54, "learning_rate": 9.260339200854546e-07, "loss": 1.2338, "step": 6930 }, { "epoch": 0.54, "learning_rate": 9.257832232918457e-07, "loss": 1.1922, "step": 6931 }, { "epoch": 0.54, "learning_rate": 9.255325311884253e-07, "loss": 1.1441, "step": 6932 }, { "epoch": 0.54, "learning_rate": 9.252818437910362e-07, "loss": 1.2292, "step": 6933 }, { "epoch": 0.54, "learning_rate": 9.2503116111552e-07, "loss": 0.8534, "step": 6934 }, { "epoch": 0.54, "learning_rate": 9.247804831777197e-07, "loss": 1.1926, "step": 6935 }, { "epoch": 0.54, "learning_rate": 9.245298099934767e-07, "loss": 1.1778, "step": 6936 }, { "epoch": 0.54, "learning_rate": 9.24279141578633e-07, "loss": 1.1736, "step": 6937 }, { "epoch": 0.54, "learning_rate": 9.24028477949029e-07, "loss": 1.2054, "step": 6938 }, { "epoch": 0.54, "learning_rate": 9.23777819120506e-07, "loss": 1.1779, "step": 6939 }, { "epoch": 0.54, "learning_rate": 9.235271651089049e-07, "loss": 1.1904, "step": 6940 }, { "epoch": 0.54, "learning_rate": 9.232765159300653e-07, "loss": 0.7972, "step": 6941 }, { "epoch": 0.54, "learning_rate": 9.230258715998278e-07, "loss": 1.2096, "step": 6942 }, { "epoch": 0.54, "learning_rate": 9.227752321340314e-07, "loss": 1.2227, "step": 6943 }, { "epoch": 0.54, "learning_rate": 9.225245975485161e-07, "loss": 0.8443, "step": 6944 }, { "epoch": 0.54, "learning_rate": 9.222739678591208e-07, "loss": 0.8247, "step": 6945 }, { "epoch": 0.54, "learning_rate": 9.220233430816838e-07, "loss": 1.2378, "step": 6946 }, { "epoch": 0.54, "learning_rate": 9.217727232320441e-07, "loss": 1.184, "step": 6947 }, { "epoch": 0.54, "learning_rate": 9.215221083260396e-07, "loss": 1.2129, "step": 6948 }, { "epoch": 0.54, "learning_rate": 9.212714983795081e-07, "loss": 1.2207, "step": 6949 }, { "epoch": 0.54, "learning_rate": 9.210208934082873e-07, "loss": 1.1929, "step": 6950 }, { "epoch": 0.54, "learning_rate": 9.207702934282137e-07, "loss": 1.1806, "step": 6951 }, { "epoch": 0.54, "learning_rate": 9.205196984551249e-07, "loss": 1.2717, "step": 6952 }, { "epoch": 0.54, "learning_rate": 9.202691085048569e-07, "loss": 0.804, "step": 6953 }, { "epoch": 0.54, "learning_rate": 9.200185235932468e-07, "loss": 1.2274, "step": 6954 }, { "epoch": 0.54, "learning_rate": 9.197679437361297e-07, "loss": 1.2184, "step": 6955 }, { "epoch": 0.54, "learning_rate": 9.195173689493411e-07, "loss": 1.1108, "step": 6956 }, { "epoch": 0.54, "learning_rate": 9.192667992487172e-07, "loss": 1.2221, "step": 6957 }, { "epoch": 0.54, "learning_rate": 9.190162346500918e-07, "loss": 1.2068, "step": 6958 }, { "epoch": 0.54, "learning_rate": 9.187656751693006e-07, "loss": 1.2327, "step": 6959 }, { "epoch": 0.54, "learning_rate": 9.18515120822177e-07, "loss": 1.1922, "step": 6960 }, { "epoch": 0.54, "learning_rate": 9.182645716245555e-07, "loss": 1.2201, "step": 6961 }, { "epoch": 0.54, "learning_rate": 9.180140275922699e-07, "loss": 1.1772, "step": 6962 }, { "epoch": 0.54, "learning_rate": 9.177634887411527e-07, "loss": 1.1953, "step": 6963 }, { "epoch": 0.54, "learning_rate": 9.175129550870378e-07, "loss": 1.2129, "step": 6964 }, { "epoch": 0.54, "learning_rate": 9.172624266457575e-07, "loss": 1.2001, "step": 6965 }, { "epoch": 0.54, "learning_rate": 9.170119034331444e-07, "loss": 1.1933, "step": 6966 }, { "epoch": 0.54, "learning_rate": 9.167613854650304e-07, "loss": 1.2124, "step": 6967 }, { "epoch": 0.54, "learning_rate": 9.165108727572465e-07, "loss": 0.8207, "step": 6968 }, { "epoch": 0.54, "learning_rate": 9.16260365325625e-07, "loss": 0.868, "step": 6969 }, { "epoch": 0.54, "learning_rate": 9.160098631859962e-07, "loss": 1.1913, "step": 6970 }, { "epoch": 0.54, "learning_rate": 9.157593663541917e-07, "loss": 1.2402, "step": 6971 }, { "epoch": 0.54, "learning_rate": 9.15508874846041e-07, "loss": 1.1794, "step": 6972 }, { "epoch": 0.54, "learning_rate": 9.152583886773745e-07, "loss": 1.1952, "step": 6973 }, { "epoch": 0.54, "learning_rate": 9.150079078640216e-07, "loss": 1.2055, "step": 6974 }, { "epoch": 0.54, "learning_rate": 9.147574324218117e-07, "loss": 1.2502, "step": 6975 }, { "epoch": 0.54, "learning_rate": 9.145069623665743e-07, "loss": 1.1943, "step": 6976 }, { "epoch": 0.54, "learning_rate": 9.142564977141375e-07, "loss": 0.8627, "step": 6977 }, { "epoch": 0.54, "learning_rate": 9.140060384803295e-07, "loss": 1.2057, "step": 6978 }, { "epoch": 0.54, "learning_rate": 9.137555846809789e-07, "loss": 1.1909, "step": 6979 }, { "epoch": 0.54, "learning_rate": 9.135051363319126e-07, "loss": 1.204, "step": 6980 }, { "epoch": 0.54, "learning_rate": 9.132546934489585e-07, "loss": 1.2229, "step": 6981 }, { "epoch": 0.54, "learning_rate": 9.130042560479432e-07, "loss": 1.2198, "step": 6982 }, { "epoch": 0.54, "learning_rate": 9.127538241446934e-07, "loss": 1.1936, "step": 6983 }, { "epoch": 0.54, "learning_rate": 9.125033977550355e-07, "loss": 1.2461, "step": 6984 }, { "epoch": 0.54, "learning_rate": 9.122529768947948e-07, "loss": 1.2084, "step": 6985 }, { "epoch": 0.54, "learning_rate": 9.120025615797975e-07, "loss": 1.21, "step": 6986 }, { "epoch": 0.54, "learning_rate": 9.117521518258684e-07, "loss": 1.2089, "step": 6987 }, { "epoch": 0.54, "learning_rate": 9.115017476488329e-07, "loss": 1.2023, "step": 6988 }, { "epoch": 0.54, "learning_rate": 9.112513490645149e-07, "loss": 1.2203, "step": 6989 }, { "epoch": 0.54, "learning_rate": 9.110009560887386e-07, "loss": 1.1916, "step": 6990 }, { "epoch": 0.54, "learning_rate": 9.107505687373281e-07, "loss": 1.169, "step": 6991 }, { "epoch": 0.54, "learning_rate": 9.105001870261063e-07, "loss": 1.2424, "step": 6992 }, { "epoch": 0.54, "learning_rate": 9.102498109708972e-07, "loss": 1.1898, "step": 6993 }, { "epoch": 0.54, "learning_rate": 9.099994405875225e-07, "loss": 1.2269, "step": 6994 }, { "epoch": 0.54, "learning_rate": 9.09749075891805e-07, "loss": 1.2215, "step": 6995 }, { "epoch": 0.54, "learning_rate": 9.094987168995669e-07, "loss": 1.2425, "step": 6996 }, { "epoch": 0.54, "learning_rate": 9.092483636266293e-07, "loss": 0.8885, "step": 6997 }, { "epoch": 0.54, "learning_rate": 9.08998016088814e-07, "loss": 1.1874, "step": 6998 }, { "epoch": 0.54, "learning_rate": 9.08747674301942e-07, "loss": 1.1868, "step": 6999 }, { "epoch": 0.54, "learning_rate": 9.084973382818331e-07, "loss": 1.2537, "step": 7000 }, { "epoch": 0.54, "learning_rate": 9.082470080443082e-07, "loss": 1.2028, "step": 7001 }, { "epoch": 0.54, "learning_rate": 9.079966836051866e-07, "loss": 1.1884, "step": 7002 }, { "epoch": 0.54, "learning_rate": 9.077463649802882e-07, "loss": 1.2433, "step": 7003 }, { "epoch": 0.54, "learning_rate": 9.074960521854317e-07, "loss": 1.2319, "step": 7004 }, { "epoch": 0.54, "learning_rate": 9.072457452364363e-07, "loss": 1.2016, "step": 7005 }, { "epoch": 0.54, "learning_rate": 9.069954441491201e-07, "loss": 1.2191, "step": 7006 }, { "epoch": 0.54, "learning_rate": 9.067451489393009e-07, "loss": 1.2026, "step": 7007 }, { "epoch": 0.54, "learning_rate": 9.064948596227963e-07, "loss": 1.1595, "step": 7008 }, { "epoch": 0.54, "learning_rate": 9.062445762154237e-07, "loss": 1.2478, "step": 7009 }, { "epoch": 0.54, "learning_rate": 9.059942987330003e-07, "loss": 1.1961, "step": 7010 }, { "epoch": 0.54, "learning_rate": 9.05744027191342e-07, "loss": 1.2294, "step": 7011 }, { "epoch": 0.54, "learning_rate": 9.05493761606265e-07, "loss": 1.186, "step": 7012 }, { "epoch": 0.54, "learning_rate": 9.052435019935855e-07, "loss": 0.8795, "step": 7013 }, { "epoch": 0.54, "learning_rate": 9.049932483691181e-07, "loss": 0.8552, "step": 7014 }, { "epoch": 0.54, "learning_rate": 9.047430007486785e-07, "loss": 1.2239, "step": 7015 }, { "epoch": 0.54, "learning_rate": 9.044927591480811e-07, "loss": 0.8026, "step": 7016 }, { "epoch": 0.54, "learning_rate": 9.042425235831396e-07, "loss": 1.2126, "step": 7017 }, { "epoch": 0.54, "learning_rate": 9.039922940696687e-07, "loss": 1.1625, "step": 7018 }, { "epoch": 0.54, "learning_rate": 9.037420706234811e-07, "loss": 1.1936, "step": 7019 }, { "epoch": 0.54, "learning_rate": 9.034918532603902e-07, "loss": 1.1822, "step": 7020 }, { "epoch": 0.54, "learning_rate": 9.03241641996209e-07, "loss": 1.1705, "step": 7021 }, { "epoch": 0.54, "learning_rate": 9.029914368467489e-07, "loss": 1.2086, "step": 7022 }, { "epoch": 0.55, "learning_rate": 9.027412378278227e-07, "loss": 1.2166, "step": 7023 }, { "epoch": 0.55, "learning_rate": 9.024910449552417e-07, "loss": 1.231, "step": 7024 }, { "epoch": 0.55, "learning_rate": 9.022408582448169e-07, "loss": 1.2105, "step": 7025 }, { "epoch": 0.55, "learning_rate": 9.019906777123587e-07, "loss": 1.2465, "step": 7026 }, { "epoch": 0.55, "learning_rate": 9.017405033736785e-07, "loss": 0.8261, "step": 7027 }, { "epoch": 0.55, "learning_rate": 9.014903352445853e-07, "loss": 1.2097, "step": 7028 }, { "epoch": 0.55, "learning_rate": 9.012401733408887e-07, "loss": 1.2647, "step": 7029 }, { "epoch": 0.55, "learning_rate": 9.009900176783987e-07, "loss": 1.2002, "step": 7030 }, { "epoch": 0.55, "learning_rate": 9.007398682729231e-07, "loss": 0.8527, "step": 7031 }, { "epoch": 0.55, "learning_rate": 9.00489725140271e-07, "loss": 1.1709, "step": 7032 }, { "epoch": 0.55, "learning_rate": 9.002395882962503e-07, "loss": 1.1702, "step": 7033 }, { "epoch": 0.55, "learning_rate": 8.999894577566681e-07, "loss": 1.1982, "step": 7034 }, { "epoch": 0.55, "learning_rate": 8.997393335373323e-07, "loss": 1.1957, "step": 7035 }, { "epoch": 0.55, "learning_rate": 8.99489215654049e-07, "loss": 1.2301, "step": 7036 }, { "epoch": 0.55, "learning_rate": 8.99239104122625e-07, "loss": 1.2093, "step": 7037 }, { "epoch": 0.55, "learning_rate": 8.989889989588666e-07, "loss": 1.2296, "step": 7038 }, { "epoch": 0.55, "learning_rate": 8.987389001785785e-07, "loss": 1.1585, "step": 7039 }, { "epoch": 0.55, "learning_rate": 8.984888077975668e-07, "loss": 1.1986, "step": 7040 }, { "epoch": 0.55, "learning_rate": 8.982387218316359e-07, "loss": 1.1525, "step": 7041 }, { "epoch": 0.55, "learning_rate": 8.979886422965901e-07, "loss": 1.2121, "step": 7042 }, { "epoch": 0.55, "learning_rate": 8.977385692082332e-07, "loss": 1.2601, "step": 7043 }, { "epoch": 0.55, "learning_rate": 8.974885025823696e-07, "loss": 1.2213, "step": 7044 }, { "epoch": 0.55, "learning_rate": 8.972384424348017e-07, "loss": 1.2027, "step": 7045 }, { "epoch": 0.55, "learning_rate": 8.969883887813321e-07, "loss": 1.1561, "step": 7046 }, { "epoch": 0.55, "learning_rate": 8.967383416377639e-07, "loss": 1.2565, "step": 7047 }, { "epoch": 0.55, "learning_rate": 8.964883010198983e-07, "loss": 1.2142, "step": 7048 }, { "epoch": 0.55, "learning_rate": 8.962382669435375e-07, "loss": 1.1086, "step": 7049 }, { "epoch": 0.55, "learning_rate": 8.959882394244822e-07, "loss": 1.2273, "step": 7050 }, { "epoch": 0.55, "learning_rate": 8.957382184785328e-07, "loss": 0.8362, "step": 7051 }, { "epoch": 0.55, "learning_rate": 8.954882041214903e-07, "loss": 1.2046, "step": 7052 }, { "epoch": 0.55, "learning_rate": 8.952381963691538e-07, "loss": 1.1893, "step": 7053 }, { "epoch": 0.55, "learning_rate": 8.949881952373232e-07, "loss": 1.1825, "step": 7054 }, { "epoch": 0.55, "learning_rate": 8.947382007417977e-07, "loss": 1.2097, "step": 7055 }, { "epoch": 0.55, "learning_rate": 8.944882128983752e-07, "loss": 1.201, "step": 7056 }, { "epoch": 0.55, "learning_rate": 8.942382317228545e-07, "loss": 1.192, "step": 7057 }, { "epoch": 0.55, "learning_rate": 8.939882572310332e-07, "loss": 1.1933, "step": 7058 }, { "epoch": 0.55, "learning_rate": 8.937382894387085e-07, "loss": 0.8382, "step": 7059 }, { "epoch": 0.55, "learning_rate": 8.934883283616778e-07, "loss": 0.788, "step": 7060 }, { "epoch": 0.55, "learning_rate": 8.932383740157367e-07, "loss": 1.194, "step": 7061 }, { "epoch": 0.55, "learning_rate": 8.929884264166819e-07, "loss": 1.1811, "step": 7062 }, { "epoch": 0.55, "learning_rate": 8.927384855803088e-07, "loss": 1.219, "step": 7063 }, { "epoch": 0.55, "learning_rate": 8.924885515224131e-07, "loss": 1.2442, "step": 7064 }, { "epoch": 0.55, "learning_rate": 8.922386242587887e-07, "loss": 1.2459, "step": 7065 }, { "epoch": 0.55, "learning_rate": 8.91988703805231e-07, "loss": 1.2088, "step": 7066 }, { "epoch": 0.55, "learning_rate": 8.917387901775332e-07, "loss": 1.2459, "step": 7067 }, { "epoch": 0.55, "learning_rate": 8.914888833914886e-07, "loss": 1.2285, "step": 7068 }, { "epoch": 0.55, "learning_rate": 8.912389834628912e-07, "loss": 1.2752, "step": 7069 }, { "epoch": 0.55, "learning_rate": 8.909890904075325e-07, "loss": 1.1548, "step": 7070 }, { "epoch": 0.55, "learning_rate": 8.907392042412055e-07, "loss": 0.8037, "step": 7071 }, { "epoch": 0.55, "learning_rate": 8.904893249797018e-07, "loss": 1.1972, "step": 7072 }, { "epoch": 0.55, "learning_rate": 8.902394526388124e-07, "loss": 0.8699, "step": 7073 }, { "epoch": 0.55, "learning_rate": 8.899895872343285e-07, "loss": 1.2181, "step": 7074 }, { "epoch": 0.55, "learning_rate": 8.897397287820405e-07, "loss": 1.2241, "step": 7075 }, { "epoch": 0.55, "learning_rate": 8.894898772977383e-07, "loss": 1.2587, "step": 7076 }, { "epoch": 0.55, "learning_rate": 8.892400327972117e-07, "loss": 1.2106, "step": 7077 }, { "epoch": 0.55, "learning_rate": 8.889901952962492e-07, "loss": 1.1756, "step": 7078 }, { "epoch": 0.55, "learning_rate": 8.887403648106402e-07, "loss": 1.2018, "step": 7079 }, { "epoch": 0.55, "learning_rate": 8.884905413561724e-07, "loss": 1.2026, "step": 7080 }, { "epoch": 0.55, "learning_rate": 8.882407249486342e-07, "loss": 1.2096, "step": 7081 }, { "epoch": 0.55, "learning_rate": 8.879909156038125e-07, "loss": 1.184, "step": 7082 }, { "epoch": 0.55, "learning_rate": 8.877411133374943e-07, "loss": 1.2041, "step": 7083 }, { "epoch": 0.55, "learning_rate": 8.87491318165466e-07, "loss": 0.8292, "step": 7084 }, { "epoch": 0.55, "learning_rate": 8.872415301035133e-07, "loss": 1.2069, "step": 7085 }, { "epoch": 0.55, "learning_rate": 8.869917491674227e-07, "loss": 1.1674, "step": 7086 }, { "epoch": 0.55, "learning_rate": 8.867419753729783e-07, "loss": 1.2267, "step": 7087 }, { "epoch": 0.55, "learning_rate": 8.864922087359652e-07, "loss": 1.1805, "step": 7088 }, { "epoch": 0.55, "learning_rate": 8.86242449272168e-07, "loss": 1.1793, "step": 7089 }, { "epoch": 0.55, "learning_rate": 8.859926969973695e-07, "loss": 1.163, "step": 7090 }, { "epoch": 0.55, "learning_rate": 8.857429519273536e-07, "loss": 1.2423, "step": 7091 }, { "epoch": 0.55, "learning_rate": 8.854932140779032e-07, "loss": 1.1715, "step": 7092 }, { "epoch": 0.55, "learning_rate": 8.852434834648005e-07, "loss": 1.2064, "step": 7093 }, { "epoch": 0.55, "learning_rate": 8.849937601038276e-07, "loss": 1.1985, "step": 7094 }, { "epoch": 0.55, "learning_rate": 8.847440440107655e-07, "loss": 1.2313, "step": 7095 }, { "epoch": 0.55, "learning_rate": 8.844943352013957e-07, "loss": 1.1571, "step": 7096 }, { "epoch": 0.55, "learning_rate": 8.842446336914984e-07, "loss": 1.2235, "step": 7097 }, { "epoch": 0.55, "learning_rate": 8.839949394968544e-07, "loss": 1.2222, "step": 7098 }, { "epoch": 0.55, "learning_rate": 8.837452526332424e-07, "loss": 1.2327, "step": 7099 }, { "epoch": 0.55, "learning_rate": 8.83495573116442e-07, "loss": 1.1437, "step": 7100 }, { "epoch": 0.55, "learning_rate": 8.832459009622318e-07, "loss": 1.1448, "step": 7101 }, { "epoch": 0.55, "learning_rate": 8.8299623618639e-07, "loss": 1.1958, "step": 7102 }, { "epoch": 0.55, "learning_rate": 8.827465788046948e-07, "loss": 1.1162, "step": 7103 }, { "epoch": 0.55, "learning_rate": 8.824969288329229e-07, "loss": 0.8166, "step": 7104 }, { "epoch": 0.55, "learning_rate": 8.822472862868511e-07, "loss": 1.1665, "step": 7105 }, { "epoch": 0.55, "learning_rate": 8.819976511822565e-07, "loss": 1.1919, "step": 7106 }, { "epoch": 0.55, "learning_rate": 8.817480235349142e-07, "loss": 1.2473, "step": 7107 }, { "epoch": 0.55, "learning_rate": 8.814984033606001e-07, "loss": 1.1922, "step": 7108 }, { "epoch": 0.55, "learning_rate": 8.81248790675089e-07, "loss": 1.1866, "step": 7109 }, { "epoch": 0.55, "learning_rate": 8.809991854941552e-07, "loss": 1.184, "step": 7110 }, { "epoch": 0.55, "learning_rate": 8.807495878335733e-07, "loss": 1.2234, "step": 7111 }, { "epoch": 0.55, "learning_rate": 8.804999977091157e-07, "loss": 1.1868, "step": 7112 }, { "epoch": 0.55, "learning_rate": 8.802504151365565e-07, "loss": 1.1858, "step": 7113 }, { "epoch": 0.55, "learning_rate": 8.800008401316678e-07, "loss": 1.17, "step": 7114 }, { "epoch": 0.55, "learning_rate": 8.79751272710222e-07, "loss": 0.8283, "step": 7115 }, { "epoch": 0.55, "learning_rate": 8.795017128879903e-07, "loss": 1.1806, "step": 7116 }, { "epoch": 0.55, "learning_rate": 8.792521606807442e-07, "loss": 1.2053, "step": 7117 }, { "epoch": 0.55, "learning_rate": 8.790026161042541e-07, "loss": 1.2195, "step": 7118 }, { "epoch": 0.55, "learning_rate": 8.7875307917429e-07, "loss": 1.1641, "step": 7119 }, { "epoch": 0.55, "learning_rate": 8.785035499066223e-07, "loss": 1.2467, "step": 7120 }, { "epoch": 0.55, "learning_rate": 8.782540283170196e-07, "loss": 0.8657, "step": 7121 }, { "epoch": 0.55, "learning_rate": 8.780045144212503e-07, "loss": 1.1077, "step": 7122 }, { "epoch": 0.55, "learning_rate": 8.777550082350838e-07, "loss": 1.1414, "step": 7123 }, { "epoch": 0.55, "learning_rate": 8.775055097742865e-07, "loss": 1.2111, "step": 7124 }, { "epoch": 0.55, "learning_rate": 8.772560190546267e-07, "loss": 1.2466, "step": 7125 }, { "epoch": 0.55, "learning_rate": 8.770065360918707e-07, "loss": 1.1753, "step": 7126 }, { "epoch": 0.55, "learning_rate": 8.767570609017846e-07, "loss": 1.1968, "step": 7127 }, { "epoch": 0.55, "learning_rate": 8.765075935001347e-07, "loss": 1.186, "step": 7128 }, { "epoch": 0.55, "learning_rate": 8.762581339026857e-07, "loss": 1.195, "step": 7129 }, { "epoch": 0.55, "learning_rate": 8.760086821252027e-07, "loss": 1.2252, "step": 7130 }, { "epoch": 0.55, "learning_rate": 8.7575923818345e-07, "loss": 1.2265, "step": 7131 }, { "epoch": 0.55, "learning_rate": 8.755098020931917e-07, "loss": 0.7936, "step": 7132 }, { "epoch": 0.55, "learning_rate": 8.752603738701907e-07, "loss": 1.2193, "step": 7133 }, { "epoch": 0.55, "learning_rate": 8.750109535302099e-07, "loss": 1.182, "step": 7134 }, { "epoch": 0.55, "learning_rate": 8.747615410890116e-07, "loss": 1.2109, "step": 7135 }, { "epoch": 0.55, "learning_rate": 8.745121365623576e-07, "loss": 1.2201, "step": 7136 }, { "epoch": 0.55, "learning_rate": 8.742627399660096e-07, "loss": 1.1543, "step": 7137 }, { "epoch": 0.55, "learning_rate": 8.74013351315728e-07, "loss": 1.25, "step": 7138 }, { "epoch": 0.55, "learning_rate": 8.73763970627273e-07, "loss": 1.1847, "step": 7139 }, { "epoch": 0.55, "learning_rate": 8.735145979164049e-07, "loss": 1.181, "step": 7140 }, { "epoch": 0.55, "learning_rate": 8.732652331988824e-07, "loss": 1.2584, "step": 7141 }, { "epoch": 0.55, "learning_rate": 8.730158764904652e-07, "loss": 1.2226, "step": 7142 }, { "epoch": 0.55, "learning_rate": 8.727665278069107e-07, "loss": 1.2291, "step": 7143 }, { "epoch": 0.55, "learning_rate": 8.725171871639768e-07, "loss": 1.1646, "step": 7144 }, { "epoch": 0.55, "learning_rate": 8.722678545774215e-07, "loss": 1.2302, "step": 7145 }, { "epoch": 0.55, "learning_rate": 8.720185300630007e-07, "loss": 1.1973, "step": 7146 }, { "epoch": 0.55, "learning_rate": 8.717692136364711e-07, "loss": 0.8311, "step": 7147 }, { "epoch": 0.55, "learning_rate": 8.715199053135887e-07, "loss": 0.8389, "step": 7148 }, { "epoch": 0.55, "learning_rate": 8.712706051101079e-07, "loss": 1.208, "step": 7149 }, { "epoch": 0.55, "learning_rate": 8.710213130417844e-07, "loss": 1.2017, "step": 7150 }, { "epoch": 0.56, "learning_rate": 8.707720291243718e-07, "loss": 1.2043, "step": 7151 }, { "epoch": 0.56, "learning_rate": 8.70522753373624e-07, "loss": 1.2232, "step": 7152 }, { "epoch": 0.56, "learning_rate": 8.70273485805294e-07, "loss": 1.2274, "step": 7153 }, { "epoch": 0.56, "learning_rate": 8.700242264351349e-07, "loss": 1.1703, "step": 7154 }, { "epoch": 0.56, "learning_rate": 8.697749752788985e-07, "loss": 0.9102, "step": 7155 }, { "epoch": 0.56, "learning_rate": 8.695257323523361e-07, "loss": 1.2345, "step": 7156 }, { "epoch": 0.56, "learning_rate": 8.692764976711998e-07, "loss": 0.8511, "step": 7157 }, { "epoch": 0.56, "learning_rate": 8.690272712512391e-07, "loss": 1.1971, "step": 7158 }, { "epoch": 0.56, "learning_rate": 8.687780531082048e-07, "loss": 1.195, "step": 7159 }, { "epoch": 0.56, "learning_rate": 8.685288432578462e-07, "loss": 1.2132, "step": 7160 }, { "epoch": 0.56, "learning_rate": 8.682796417159119e-07, "loss": 1.2182, "step": 7161 }, { "epoch": 0.56, "learning_rate": 8.680304484981512e-07, "loss": 1.2015, "step": 7162 }, { "epoch": 0.56, "learning_rate": 8.677812636203112e-07, "loss": 1.1763, "step": 7163 }, { "epoch": 0.56, "learning_rate": 8.675320870981402e-07, "loss": 1.2387, "step": 7164 }, { "epoch": 0.56, "learning_rate": 8.672829189473847e-07, "loss": 1.2122, "step": 7165 }, { "epoch": 0.56, "learning_rate": 8.670337591837907e-07, "loss": 0.8681, "step": 7166 }, { "epoch": 0.56, "learning_rate": 8.667846078231046e-07, "loss": 1.1982, "step": 7167 }, { "epoch": 0.56, "learning_rate": 8.665354648810716e-07, "loss": 1.1957, "step": 7168 }, { "epoch": 0.56, "learning_rate": 8.662863303734363e-07, "loss": 1.1557, "step": 7169 }, { "epoch": 0.56, "learning_rate": 8.660372043159434e-07, "loss": 1.1877, "step": 7170 }, { "epoch": 0.56, "learning_rate": 8.657880867243359e-07, "loss": 1.2166, "step": 7171 }, { "epoch": 0.56, "learning_rate": 8.655389776143576e-07, "loss": 1.1944, "step": 7172 }, { "epoch": 0.56, "learning_rate": 8.652898770017506e-07, "loss": 1.2075, "step": 7173 }, { "epoch": 0.56, "learning_rate": 8.650407849022578e-07, "loss": 1.2959, "step": 7174 }, { "epoch": 0.56, "learning_rate": 8.647917013316201e-07, "loss": 1.1924, "step": 7175 }, { "epoch": 0.56, "learning_rate": 8.64542626305579e-07, "loss": 1.1809, "step": 7176 }, { "epoch": 0.56, "learning_rate": 8.642935598398745e-07, "loss": 1.1754, "step": 7177 }, { "epoch": 0.56, "learning_rate": 8.640445019502467e-07, "loss": 1.2026, "step": 7178 }, { "epoch": 0.56, "learning_rate": 8.637954526524355e-07, "loss": 1.1959, "step": 7179 }, { "epoch": 0.56, "learning_rate": 8.635464119621789e-07, "loss": 1.1454, "step": 7180 }, { "epoch": 0.56, "learning_rate": 8.632973798952159e-07, "loss": 0.8571, "step": 7181 }, { "epoch": 0.56, "learning_rate": 8.630483564672845e-07, "loss": 1.2011, "step": 7182 }, { "epoch": 0.56, "learning_rate": 8.627993416941209e-07, "loss": 1.2389, "step": 7183 }, { "epoch": 0.56, "learning_rate": 8.625503355914627e-07, "loss": 1.1612, "step": 7184 }, { "epoch": 0.56, "learning_rate": 8.623013381750458e-07, "loss": 1.1751, "step": 7185 }, { "epoch": 0.56, "learning_rate": 8.620523494606056e-07, "loss": 1.1945, "step": 7186 }, { "epoch": 0.56, "learning_rate": 8.618033694638777e-07, "loss": 1.166, "step": 7187 }, { "epoch": 0.56, "learning_rate": 8.615543982005956e-07, "loss": 1.1701, "step": 7188 }, { "epoch": 0.56, "learning_rate": 8.613054356864941e-07, "loss": 1.2202, "step": 7189 }, { "epoch": 0.56, "learning_rate": 8.610564819373059e-07, "loss": 1.2326, "step": 7190 }, { "epoch": 0.56, "learning_rate": 8.608075369687648e-07, "loss": 0.8962, "step": 7191 }, { "epoch": 0.56, "learning_rate": 8.605586007966023e-07, "loss": 1.2245, "step": 7192 }, { "epoch": 0.56, "learning_rate": 8.603096734365504e-07, "loss": 1.2038, "step": 7193 }, { "epoch": 0.56, "learning_rate": 8.600607549043401e-07, "loss": 1.2139, "step": 7194 }, { "epoch": 0.56, "learning_rate": 8.598118452157018e-07, "loss": 1.245, "step": 7195 }, { "epoch": 0.56, "learning_rate": 8.595629443863664e-07, "loss": 1.2705, "step": 7196 }, { "epoch": 0.56, "learning_rate": 8.593140524320622e-07, "loss": 1.1719, "step": 7197 }, { "epoch": 0.56, "learning_rate": 8.590651693685192e-07, "loss": 1.1673, "step": 7198 }, { "epoch": 0.56, "learning_rate": 8.588162952114653e-07, "loss": 1.1938, "step": 7199 }, { "epoch": 0.56, "learning_rate": 8.58567429976628e-07, "loss": 1.1843, "step": 7200 }, { "epoch": 0.56, "learning_rate": 8.583185736797354e-07, "loss": 1.2081, "step": 7201 }, { "epoch": 0.56, "learning_rate": 8.58069726336513e-07, "loss": 1.2091, "step": 7202 }, { "epoch": 0.56, "learning_rate": 8.578208879626878e-07, "loss": 1.2454, "step": 7203 }, { "epoch": 0.56, "learning_rate": 8.575720585739853e-07, "loss": 1.1721, "step": 7204 }, { "epoch": 0.56, "learning_rate": 8.573232381861298e-07, "loss": 1.2414, "step": 7205 }, { "epoch": 0.56, "learning_rate": 8.570744268148464e-07, "loss": 1.2532, "step": 7206 }, { "epoch": 0.56, "learning_rate": 8.568256244758584e-07, "loss": 1.2852, "step": 7207 }, { "epoch": 0.56, "learning_rate": 8.565768311848897e-07, "loss": 1.2323, "step": 7208 }, { "epoch": 0.56, "learning_rate": 8.563280469576626e-07, "loss": 1.2184, "step": 7209 }, { "epoch": 0.56, "learning_rate": 8.560792718098993e-07, "loss": 1.2105, "step": 7210 }, { "epoch": 0.56, "learning_rate": 8.558305057573211e-07, "loss": 1.2026, "step": 7211 }, { "epoch": 0.56, "learning_rate": 8.55581748815649e-07, "loss": 1.2135, "step": 7212 }, { "epoch": 0.56, "learning_rate": 8.553330010006039e-07, "loss": 1.2117, "step": 7213 }, { "epoch": 0.56, "learning_rate": 8.550842623279053e-07, "loss": 1.2225, "step": 7214 }, { "epoch": 0.56, "learning_rate": 8.548355328132719e-07, "loss": 1.2049, "step": 7215 }, { "epoch": 0.56, "learning_rate": 8.545868124724235e-07, "loss": 1.2144, "step": 7216 }, { "epoch": 0.56, "learning_rate": 8.543381013210771e-07, "loss": 1.2012, "step": 7217 }, { "epoch": 0.56, "learning_rate": 8.54089399374951e-07, "loss": 1.22, "step": 7218 }, { "epoch": 0.56, "learning_rate": 8.538407066497614e-07, "loss": 1.1969, "step": 7219 }, { "epoch": 0.56, "learning_rate": 8.535920231612252e-07, "loss": 1.2067, "step": 7220 }, { "epoch": 0.56, "learning_rate": 8.533433489250582e-07, "loss": 1.2345, "step": 7221 }, { "epoch": 0.56, "learning_rate": 8.530946839569748e-07, "loss": 1.242, "step": 7222 }, { "epoch": 0.56, "learning_rate": 8.528460282726904e-07, "loss": 1.2357, "step": 7223 }, { "epoch": 0.56, "learning_rate": 8.525973818879185e-07, "loss": 1.2295, "step": 7224 }, { "epoch": 0.56, "learning_rate": 8.52348744818373e-07, "loss": 0.8509, "step": 7225 }, { "epoch": 0.56, "learning_rate": 8.521001170797663e-07, "loss": 0.8764, "step": 7226 }, { "epoch": 0.56, "learning_rate": 8.51851498687811e-07, "loss": 1.2155, "step": 7227 }, { "epoch": 0.56, "learning_rate": 8.516028896582181e-07, "loss": 1.1827, "step": 7228 }, { "epoch": 0.56, "learning_rate": 8.51354290006699e-07, "loss": 0.8831, "step": 7229 }, { "epoch": 0.56, "learning_rate": 8.511056997489647e-07, "loss": 1.2001, "step": 7230 }, { "epoch": 0.56, "learning_rate": 8.508571189007241e-07, "loss": 1.2328, "step": 7231 }, { "epoch": 0.56, "learning_rate": 8.506085474776868e-07, "loss": 0.8621, "step": 7232 }, { "epoch": 0.56, "learning_rate": 8.503599854955621e-07, "loss": 0.8096, "step": 7233 }, { "epoch": 0.56, "learning_rate": 8.501114329700571e-07, "loss": 1.2429, "step": 7234 }, { "epoch": 0.56, "learning_rate": 8.498628899168801e-07, "loss": 1.1929, "step": 7235 }, { "epoch": 0.56, "learning_rate": 8.496143563517374e-07, "loss": 1.2484, "step": 7236 }, { "epoch": 0.56, "learning_rate": 8.493658322903352e-07, "loss": 1.2019, "step": 7237 }, { "epoch": 0.56, "learning_rate": 8.491173177483801e-07, "loss": 1.2365, "step": 7238 }, { "epoch": 0.56, "learning_rate": 8.48868812741576e-07, "loss": 0.7743, "step": 7239 }, { "epoch": 0.56, "learning_rate": 8.486203172856281e-07, "loss": 1.2286, "step": 7240 }, { "epoch": 0.56, "learning_rate": 8.4837183139624e-07, "loss": 1.2253, "step": 7241 }, { "epoch": 0.56, "learning_rate": 8.481233550891152e-07, "loss": 1.1306, "step": 7242 }, { "epoch": 0.56, "learning_rate": 8.47874888379956e-07, "loss": 1.2274, "step": 7243 }, { "epoch": 0.56, "learning_rate": 8.476264312844649e-07, "loss": 1.189, "step": 7244 }, { "epoch": 0.56, "learning_rate": 8.473779838183429e-07, "loss": 1.2426, "step": 7245 }, { "epoch": 0.56, "learning_rate": 8.471295459972909e-07, "loss": 1.2001, "step": 7246 }, { "epoch": 0.56, "learning_rate": 8.468811178370097e-07, "loss": 1.1753, "step": 7247 }, { "epoch": 0.56, "learning_rate": 8.466326993531982e-07, "loss": 1.2573, "step": 7248 }, { "epoch": 0.56, "learning_rate": 8.463842905615555e-07, "loss": 1.199, "step": 7249 }, { "epoch": 0.56, "learning_rate": 8.461358914777807e-07, "loss": 1.2275, "step": 7250 }, { "epoch": 0.56, "learning_rate": 8.458875021175704e-07, "loss": 0.8364, "step": 7251 }, { "epoch": 0.56, "learning_rate": 8.456391224966231e-07, "loss": 1.1469, "step": 7252 }, { "epoch": 0.56, "learning_rate": 8.453907526306342e-07, "loss": 0.8529, "step": 7253 }, { "epoch": 0.56, "learning_rate": 8.451423925353e-07, "loss": 0.8159, "step": 7254 }, { "epoch": 0.56, "learning_rate": 8.448940422263163e-07, "loss": 1.2064, "step": 7255 }, { "epoch": 0.56, "learning_rate": 8.446457017193771e-07, "loss": 1.2122, "step": 7256 }, { "epoch": 0.56, "learning_rate": 8.443973710301768e-07, "loss": 1.2049, "step": 7257 }, { "epoch": 0.56, "learning_rate": 8.441490501744092e-07, "loss": 1.1843, "step": 7258 }, { "epoch": 0.56, "learning_rate": 8.439007391677663e-07, "loss": 1.2067, "step": 7259 }, { "epoch": 0.56, "learning_rate": 8.436524380259411e-07, "loss": 1.1578, "step": 7260 }, { "epoch": 0.56, "learning_rate": 8.434041467646245e-07, "loss": 0.8488, "step": 7261 }, { "epoch": 0.56, "learning_rate": 8.431558653995082e-07, "loss": 1.2446, "step": 7262 }, { "epoch": 0.56, "learning_rate": 8.429075939462818e-07, "loss": 1.1776, "step": 7263 }, { "epoch": 0.56, "learning_rate": 8.426593324206359e-07, "loss": 1.2243, "step": 7264 }, { "epoch": 0.56, "learning_rate": 8.424110808382589e-07, "loss": 1.1885, "step": 7265 }, { "epoch": 0.56, "learning_rate": 8.421628392148392e-07, "loss": 1.2026, "step": 7266 }, { "epoch": 0.56, "learning_rate": 8.419146075660651e-07, "loss": 1.141, "step": 7267 }, { "epoch": 0.56, "learning_rate": 8.416663859076233e-07, "loss": 1.2518, "step": 7268 }, { "epoch": 0.56, "learning_rate": 8.41418174255201e-07, "loss": 1.1873, "step": 7269 }, { "epoch": 0.56, "learning_rate": 8.411699726244835e-07, "loss": 1.1376, "step": 7270 }, { "epoch": 0.56, "learning_rate": 8.409217810311562e-07, "loss": 1.1635, "step": 7271 }, { "epoch": 0.56, "learning_rate": 8.406735994909044e-07, "loss": 1.2265, "step": 7272 }, { "epoch": 0.56, "learning_rate": 8.404254280194112e-07, "loss": 1.2345, "step": 7273 }, { "epoch": 0.56, "learning_rate": 8.401772666323605e-07, "loss": 1.2357, "step": 7274 }, { "epoch": 0.56, "learning_rate": 8.399291153454354e-07, "loss": 1.2087, "step": 7275 }, { "epoch": 0.56, "learning_rate": 8.396809741743172e-07, "loss": 1.2429, "step": 7276 }, { "epoch": 0.56, "learning_rate": 8.394328431346881e-07, "loss": 1.2421, "step": 7277 }, { "epoch": 0.56, "learning_rate": 8.391847222422282e-07, "loss": 1.2059, "step": 7278 }, { "epoch": 0.56, "learning_rate": 8.389366115126184e-07, "loss": 1.2327, "step": 7279 }, { "epoch": 0.57, "learning_rate": 8.386885109615383e-07, "loss": 1.2188, "step": 7280 }, { "epoch": 0.57, "learning_rate": 8.38440420604666e-07, "loss": 1.2232, "step": 7281 }, { "epoch": 0.57, "learning_rate": 8.381923404576805e-07, "loss": 1.1762, "step": 7282 }, { "epoch": 0.57, "learning_rate": 8.379442705362589e-07, "loss": 1.2043, "step": 7283 }, { "epoch": 0.57, "learning_rate": 8.37696210856079e-07, "loss": 1.1772, "step": 7284 }, { "epoch": 0.57, "learning_rate": 8.374481614328162e-07, "loss": 1.2266, "step": 7285 }, { "epoch": 0.57, "learning_rate": 8.37200122282147e-07, "loss": 0.8357, "step": 7286 }, { "epoch": 0.57, "learning_rate": 8.369520934197458e-07, "loss": 0.7957, "step": 7287 }, { "epoch": 0.57, "learning_rate": 8.367040748612868e-07, "loss": 1.1743, "step": 7288 }, { "epoch": 0.57, "learning_rate": 8.364560666224447e-07, "loss": 0.8116, "step": 7289 }, { "epoch": 0.57, "learning_rate": 8.362080687188916e-07, "loss": 0.8519, "step": 7290 }, { "epoch": 0.57, "learning_rate": 8.359600811663006e-07, "loss": 1.2766, "step": 7291 }, { "epoch": 0.57, "learning_rate": 8.357121039803432e-07, "loss": 0.8582, "step": 7292 }, { "epoch": 0.57, "learning_rate": 8.354641371766902e-07, "loss": 0.8554, "step": 7293 }, { "epoch": 0.57, "learning_rate": 8.35216180771013e-07, "loss": 1.1678, "step": 7294 }, { "epoch": 0.57, "learning_rate": 8.349682347789801e-07, "loss": 1.2204, "step": 7295 }, { "epoch": 0.57, "learning_rate": 8.347202992162617e-07, "loss": 1.1824, "step": 7296 }, { "epoch": 0.57, "learning_rate": 8.34472374098526e-07, "loss": 1.2862, "step": 7297 }, { "epoch": 0.57, "learning_rate": 8.342244594414404e-07, "loss": 1.1754, "step": 7298 }, { "epoch": 0.57, "learning_rate": 8.339765552606725e-07, "loss": 1.2284, "step": 7299 }, { "epoch": 0.57, "learning_rate": 8.33728661571889e-07, "loss": 1.1893, "step": 7300 }, { "epoch": 0.57, "learning_rate": 8.334807783907551e-07, "loss": 1.1323, "step": 7301 }, { "epoch": 0.57, "learning_rate": 8.332329057329366e-07, "loss": 1.2186, "step": 7302 }, { "epoch": 0.57, "learning_rate": 8.329850436140975e-07, "loss": 1.1891, "step": 7303 }, { "epoch": 0.57, "learning_rate": 8.32737192049902e-07, "loss": 1.2248, "step": 7304 }, { "epoch": 0.57, "learning_rate": 8.324893510560129e-07, "loss": 1.1681, "step": 7305 }, { "epoch": 0.57, "learning_rate": 8.322415206480934e-07, "loss": 1.2283, "step": 7306 }, { "epoch": 0.57, "learning_rate": 8.319937008418046e-07, "loss": 1.1811, "step": 7307 }, { "epoch": 0.57, "learning_rate": 8.317458916528079e-07, "loss": 1.2499, "step": 7308 }, { "epoch": 0.57, "learning_rate": 8.314980930967644e-07, "loss": 1.196, "step": 7309 }, { "epoch": 0.57, "learning_rate": 8.312503051893328e-07, "loss": 1.2267, "step": 7310 }, { "epoch": 0.57, "learning_rate": 8.310025279461733e-07, "loss": 1.1553, "step": 7311 }, { "epoch": 0.57, "learning_rate": 8.307547613829436e-07, "loss": 1.2132, "step": 7312 }, { "epoch": 0.57, "learning_rate": 8.305070055153018e-07, "loss": 1.2095, "step": 7313 }, { "epoch": 0.57, "learning_rate": 8.302592603589056e-07, "loss": 1.1927, "step": 7314 }, { "epoch": 0.57, "learning_rate": 8.300115259294103e-07, "loss": 0.8186, "step": 7315 }, { "epoch": 0.57, "learning_rate": 8.297638022424727e-07, "loss": 0.8605, "step": 7316 }, { "epoch": 0.57, "learning_rate": 8.295160893137472e-07, "loss": 1.2201, "step": 7317 }, { "epoch": 0.57, "learning_rate": 8.292683871588888e-07, "loss": 1.2676, "step": 7318 }, { "epoch": 0.57, "learning_rate": 8.290206957935508e-07, "loss": 1.1404, "step": 7319 }, { "epoch": 0.57, "learning_rate": 8.287730152333864e-07, "loss": 1.2395, "step": 7320 }, { "epoch": 0.57, "learning_rate": 8.285253454940481e-07, "loss": 0.8512, "step": 7321 }, { "epoch": 0.57, "learning_rate": 8.282776865911871e-07, "loss": 1.1705, "step": 7322 }, { "epoch": 0.57, "learning_rate": 8.280300385404553e-07, "loss": 1.1714, "step": 7323 }, { "epoch": 0.57, "learning_rate": 8.277824013575021e-07, "loss": 1.2731, "step": 7324 }, { "epoch": 0.57, "learning_rate": 8.275347750579775e-07, "loss": 1.1952, "step": 7325 }, { "epoch": 0.57, "learning_rate": 8.272871596575307e-07, "loss": 1.1787, "step": 7326 }, { "epoch": 0.57, "learning_rate": 8.270395551718093e-07, "loss": 1.2169, "step": 7327 }, { "epoch": 0.57, "learning_rate": 8.267919616164618e-07, "loss": 1.1592, "step": 7328 }, { "epoch": 0.57, "learning_rate": 8.26544379007134e-07, "loss": 1.2098, "step": 7329 }, { "epoch": 0.57, "learning_rate": 8.262968073594729e-07, "loss": 1.1734, "step": 7330 }, { "epoch": 0.57, "learning_rate": 8.260492466891239e-07, "loss": 0.8444, "step": 7331 }, { "epoch": 0.57, "learning_rate": 8.258016970117312e-07, "loss": 1.1574, "step": 7332 }, { "epoch": 0.57, "learning_rate": 8.255541583429394e-07, "loss": 1.1617, "step": 7333 }, { "epoch": 0.57, "learning_rate": 8.253066306983919e-07, "loss": 0.8495, "step": 7334 }, { "epoch": 0.57, "learning_rate": 8.250591140937313e-07, "loss": 1.1735, "step": 7335 }, { "epoch": 0.57, "learning_rate": 8.248116085445998e-07, "loss": 1.1825, "step": 7336 }, { "epoch": 0.57, "learning_rate": 8.245641140666382e-07, "loss": 1.2255, "step": 7337 }, { "epoch": 0.57, "learning_rate": 8.243166306754876e-07, "loss": 1.1308, "step": 7338 }, { "epoch": 0.57, "learning_rate": 8.240691583867873e-07, "loss": 1.1835, "step": 7339 }, { "epoch": 0.57, "learning_rate": 8.238216972161778e-07, "loss": 1.1975, "step": 7340 }, { "epoch": 0.57, "learning_rate": 8.235742471792963e-07, "loss": 1.2226, "step": 7341 }, { "epoch": 0.57, "learning_rate": 8.233268082917809e-07, "loss": 1.1715, "step": 7342 }, { "epoch": 0.57, "learning_rate": 8.230793805692693e-07, "loss": 1.1801, "step": 7343 }, { "epoch": 0.57, "learning_rate": 8.228319640273971e-07, "loss": 0.8718, "step": 7344 }, { "epoch": 0.57, "learning_rate": 8.225845586818006e-07, "loss": 1.2202, "step": 7345 }, { "epoch": 0.57, "learning_rate": 8.223371645481146e-07, "loss": 1.2046, "step": 7346 }, { "epoch": 0.57, "learning_rate": 8.220897816419728e-07, "loss": 1.1977, "step": 7347 }, { "epoch": 0.57, "learning_rate": 8.218424099790098e-07, "loss": 1.1935, "step": 7348 }, { "epoch": 0.57, "learning_rate": 8.215950495748574e-07, "loss": 1.2294, "step": 7349 }, { "epoch": 0.57, "learning_rate": 8.213477004451484e-07, "loss": 0.8432, "step": 7350 }, { "epoch": 0.57, "learning_rate": 8.211003626055143e-07, "loss": 0.817, "step": 7351 }, { "epoch": 0.57, "learning_rate": 8.208530360715855e-07, "loss": 1.2227, "step": 7352 }, { "epoch": 0.57, "learning_rate": 8.206057208589921e-07, "loss": 0.877, "step": 7353 }, { "epoch": 0.57, "learning_rate": 8.20358416983363e-07, "loss": 1.2266, "step": 7354 }, { "epoch": 0.57, "learning_rate": 8.201111244603276e-07, "loss": 1.235, "step": 7355 }, { "epoch": 0.57, "learning_rate": 8.198638433055128e-07, "loss": 0.8739, "step": 7356 }, { "epoch": 0.57, "learning_rate": 8.196165735345468e-07, "loss": 1.2089, "step": 7357 }, { "epoch": 0.57, "learning_rate": 8.193693151630549e-07, "loss": 0.8301, "step": 7358 }, { "epoch": 0.57, "learning_rate": 8.191220682066634e-07, "loss": 1.1432, "step": 7359 }, { "epoch": 0.57, "learning_rate": 8.188748326809975e-07, "loss": 1.2249, "step": 7360 }, { "epoch": 0.57, "learning_rate": 8.186276086016807e-07, "loss": 1.19, "step": 7361 }, { "epoch": 0.57, "learning_rate": 8.183803959843373e-07, "loss": 1.2138, "step": 7362 }, { "epoch": 0.57, "learning_rate": 8.181331948445895e-07, "loss": 1.154, "step": 7363 }, { "epoch": 0.57, "learning_rate": 8.178860051980596e-07, "loss": 1.2229, "step": 7364 }, { "epoch": 0.57, "learning_rate": 8.176388270603692e-07, "loss": 0.8265, "step": 7365 }, { "epoch": 0.57, "learning_rate": 8.173916604471383e-07, "loss": 1.217, "step": 7366 }, { "epoch": 0.57, "learning_rate": 8.171445053739874e-07, "loss": 0.8438, "step": 7367 }, { "epoch": 0.57, "learning_rate": 8.168973618565357e-07, "loss": 1.1568, "step": 7368 }, { "epoch": 0.57, "learning_rate": 8.166502299104009e-07, "loss": 1.1896, "step": 7369 }, { "epoch": 0.57, "learning_rate": 8.164031095512016e-07, "loss": 1.251, "step": 7370 }, { "epoch": 0.57, "learning_rate": 8.161560007945538e-07, "loss": 1.1859, "step": 7371 }, { "epoch": 0.57, "learning_rate": 8.159089036560747e-07, "loss": 1.2363, "step": 7372 }, { "epoch": 0.57, "learning_rate": 8.15661818151379e-07, "loss": 1.1619, "step": 7373 }, { "epoch": 0.57, "learning_rate": 8.154147442960822e-07, "loss": 1.2482, "step": 7374 }, { "epoch": 0.57, "learning_rate": 8.151676821057979e-07, "loss": 1.1736, "step": 7375 }, { "epoch": 0.57, "learning_rate": 8.149206315961395e-07, "loss": 1.169, "step": 7376 }, { "epoch": 0.57, "learning_rate": 8.146735927827193e-07, "loss": 1.1623, "step": 7377 }, { "epoch": 0.57, "learning_rate": 8.144265656811493e-07, "loss": 1.1825, "step": 7378 }, { "epoch": 0.57, "learning_rate": 8.14179550307041e-07, "loss": 1.2672, "step": 7379 }, { "epoch": 0.57, "learning_rate": 8.13932546676004e-07, "loss": 1.2286, "step": 7380 }, { "epoch": 0.57, "learning_rate": 8.136855548036481e-07, "loss": 1.2287, "step": 7381 }, { "epoch": 0.57, "learning_rate": 8.134385747055828e-07, "loss": 1.2009, "step": 7382 }, { "epoch": 0.57, "learning_rate": 8.131916063974151e-07, "loss": 1.2167, "step": 7383 }, { "epoch": 0.57, "learning_rate": 8.129446498947533e-07, "loss": 1.2018, "step": 7384 }, { "epoch": 0.57, "learning_rate": 8.126977052132037e-07, "loss": 1.2218, "step": 7385 }, { "epoch": 0.57, "learning_rate": 8.124507723683718e-07, "loss": 0.8316, "step": 7386 }, { "epoch": 0.57, "learning_rate": 8.122038513758634e-07, "loss": 1.1804, "step": 7387 }, { "epoch": 0.57, "learning_rate": 8.119569422512821e-07, "loss": 1.2283, "step": 7388 }, { "epoch": 0.57, "learning_rate": 8.117100450102321e-07, "loss": 0.8277, "step": 7389 }, { "epoch": 0.57, "learning_rate": 8.114631596683164e-07, "loss": 1.2168, "step": 7390 }, { "epoch": 0.57, "learning_rate": 8.112162862411364e-07, "loss": 1.1984, "step": 7391 }, { "epoch": 0.57, "learning_rate": 8.10969424744294e-07, "loss": 1.2135, "step": 7392 }, { "epoch": 0.57, "learning_rate": 8.107225751933898e-07, "loss": 1.2027, "step": 7393 }, { "epoch": 0.57, "learning_rate": 8.104757376040233e-07, "loss": 1.1564, "step": 7394 }, { "epoch": 0.57, "learning_rate": 8.102289119917938e-07, "loss": 1.1603, "step": 7395 }, { "epoch": 0.57, "learning_rate": 8.099820983723001e-07, "loss": 1.1925, "step": 7396 }, { "epoch": 0.57, "learning_rate": 8.097352967611391e-07, "loss": 1.2113, "step": 7397 }, { "epoch": 0.57, "learning_rate": 8.094885071739077e-07, "loss": 1.1863, "step": 7398 }, { "epoch": 0.57, "learning_rate": 8.092417296262027e-07, "loss": 1.2304, "step": 7399 }, { "epoch": 0.57, "learning_rate": 8.089949641336182e-07, "loss": 1.2507, "step": 7400 }, { "epoch": 0.57, "learning_rate": 8.087482107117497e-07, "loss": 0.8019, "step": 7401 }, { "epoch": 0.57, "learning_rate": 8.085014693761909e-07, "loss": 1.1807, "step": 7402 }, { "epoch": 0.57, "learning_rate": 8.082547401425342e-07, "loss": 1.1548, "step": 7403 }, { "epoch": 0.57, "learning_rate": 8.080080230263726e-07, "loss": 1.2161, "step": 7404 }, { "epoch": 0.57, "learning_rate": 8.077613180432969e-07, "loss": 0.8321, "step": 7405 }, { "epoch": 0.57, "learning_rate": 8.075146252088983e-07, "loss": 1.2098, "step": 7406 }, { "epoch": 0.57, "learning_rate": 8.072679445387669e-07, "loss": 1.1658, "step": 7407 }, { "epoch": 0.57, "learning_rate": 8.07021276048491e-07, "loss": 1.1977, "step": 7408 }, { "epoch": 0.58, "learning_rate": 8.067746197536599e-07, "loss": 1.1987, "step": 7409 }, { "epoch": 0.58, "learning_rate": 8.065279756698608e-07, "loss": 1.1872, "step": 7410 }, { "epoch": 0.58, "learning_rate": 8.062813438126809e-07, "loss": 1.1679, "step": 7411 }, { "epoch": 0.58, "learning_rate": 8.060347241977062e-07, "loss": 1.2163, "step": 7412 }, { "epoch": 0.58, "learning_rate": 8.057881168405216e-07, "loss": 1.185, "step": 7413 }, { "epoch": 0.58, "learning_rate": 8.05541521756712e-07, "loss": 1.1942, "step": 7414 }, { "epoch": 0.58, "learning_rate": 8.052949389618611e-07, "loss": 1.1987, "step": 7415 }, { "epoch": 0.58, "learning_rate": 8.050483684715524e-07, "loss": 1.2085, "step": 7416 }, { "epoch": 0.58, "learning_rate": 8.048018103013669e-07, "loss": 1.2171, "step": 7417 }, { "epoch": 0.58, "learning_rate": 8.045552644668873e-07, "loss": 0.867, "step": 7418 }, { "epoch": 0.58, "learning_rate": 8.04308730983694e-07, "loss": 0.8241, "step": 7419 }, { "epoch": 0.58, "learning_rate": 8.040622098673662e-07, "loss": 1.1624, "step": 7420 }, { "epoch": 0.58, "learning_rate": 8.038157011334837e-07, "loss": 1.1694, "step": 7421 }, { "epoch": 0.58, "learning_rate": 8.035692047976243e-07, "loss": 1.1434, "step": 7422 }, { "epoch": 0.58, "learning_rate": 8.033227208753659e-07, "loss": 1.2617, "step": 7423 }, { "epoch": 0.58, "learning_rate": 8.030762493822854e-07, "loss": 0.8382, "step": 7424 }, { "epoch": 0.58, "learning_rate": 8.02829790333958e-07, "loss": 0.8603, "step": 7425 }, { "epoch": 0.58, "learning_rate": 8.025833437459597e-07, "loss": 1.2487, "step": 7426 }, { "epoch": 0.58, "learning_rate": 8.023369096338646e-07, "loss": 1.1839, "step": 7427 }, { "epoch": 0.58, "learning_rate": 8.020904880132461e-07, "loss": 1.2237, "step": 7428 }, { "epoch": 0.58, "learning_rate": 8.018440788996775e-07, "loss": 1.2343, "step": 7429 }, { "epoch": 0.58, "learning_rate": 8.0159768230873e-07, "loss": 1.2123, "step": 7430 }, { "epoch": 0.58, "learning_rate": 8.013512982559757e-07, "loss": 1.18, "step": 7431 }, { "epoch": 0.58, "learning_rate": 8.011049267569843e-07, "loss": 1.1874, "step": 7432 }, { "epoch": 0.58, "learning_rate": 8.008585678273263e-07, "loss": 1.1795, "step": 7433 }, { "epoch": 0.58, "learning_rate": 8.006122214825698e-07, "loss": 1.2329, "step": 7434 }, { "epoch": 0.58, "learning_rate": 8.003658877382828e-07, "loss": 0.8574, "step": 7435 }, { "epoch": 0.58, "learning_rate": 8.001195666100336e-07, "loss": 1.1971, "step": 7436 }, { "epoch": 0.58, "learning_rate": 7.998732581133873e-07, "loss": 0.8079, "step": 7437 }, { "epoch": 0.58, "learning_rate": 7.996269622639105e-07, "loss": 0.8308, "step": 7438 }, { "epoch": 0.58, "learning_rate": 7.993806790771675e-07, "loss": 1.2154, "step": 7439 }, { "epoch": 0.58, "learning_rate": 7.991344085687228e-07, "loss": 1.1933, "step": 7440 }, { "epoch": 0.58, "learning_rate": 7.988881507541396e-07, "loss": 1.1772, "step": 7441 }, { "epoch": 0.58, "learning_rate": 7.9864190564898e-07, "loss": 1.2084, "step": 7442 }, { "epoch": 0.58, "learning_rate": 7.983956732688058e-07, "loss": 1.2655, "step": 7443 }, { "epoch": 0.58, "learning_rate": 7.981494536291781e-07, "loss": 1.2161, "step": 7444 }, { "epoch": 0.58, "learning_rate": 7.97903246745657e-07, "loss": 1.2209, "step": 7445 }, { "epoch": 0.58, "learning_rate": 7.976570526338016e-07, "loss": 1.2465, "step": 7446 }, { "epoch": 0.58, "learning_rate": 7.974108713091697e-07, "loss": 0.8403, "step": 7447 }, { "epoch": 0.58, "learning_rate": 7.971647027873198e-07, "loss": 1.1976, "step": 7448 }, { "epoch": 0.58, "learning_rate": 7.969185470838083e-07, "loss": 1.1705, "step": 7449 }, { "epoch": 0.58, "learning_rate": 7.966724042141915e-07, "loss": 1.213, "step": 7450 }, { "epoch": 0.58, "learning_rate": 7.964262741940245e-07, "loss": 1.1939, "step": 7451 }, { "epoch": 0.58, "learning_rate": 7.961801570388614e-07, "loss": 1.1383, "step": 7452 }, { "epoch": 0.58, "learning_rate": 7.959340527642558e-07, "loss": 1.1812, "step": 7453 }, { "epoch": 0.58, "learning_rate": 7.956879613857606e-07, "loss": 1.2258, "step": 7454 }, { "epoch": 0.58, "learning_rate": 7.954418829189282e-07, "loss": 1.182, "step": 7455 }, { "epoch": 0.58, "learning_rate": 7.951958173793091e-07, "loss": 1.1829, "step": 7456 }, { "epoch": 0.58, "learning_rate": 7.949497647824535e-07, "loss": 1.1868, "step": 7457 }, { "epoch": 0.58, "learning_rate": 7.947037251439116e-07, "loss": 1.191, "step": 7458 }, { "epoch": 0.58, "learning_rate": 7.944576984792313e-07, "loss": 1.2373, "step": 7459 }, { "epoch": 0.58, "learning_rate": 7.94211684803961e-07, "loss": 1.2355, "step": 7460 }, { "epoch": 0.58, "learning_rate": 7.939656841336474e-07, "loss": 1.1971, "step": 7461 }, { "epoch": 0.58, "learning_rate": 7.93719696483837e-07, "loss": 1.1937, "step": 7462 }, { "epoch": 0.58, "learning_rate": 7.934737218700751e-07, "loss": 1.2221, "step": 7463 }, { "epoch": 0.58, "learning_rate": 7.932277603079058e-07, "loss": 1.1248, "step": 7464 }, { "epoch": 0.58, "learning_rate": 7.929818118128736e-07, "loss": 1.2768, "step": 7465 }, { "epoch": 0.58, "learning_rate": 7.927358764005208e-07, "loss": 1.1626, "step": 7466 }, { "epoch": 0.58, "learning_rate": 7.9248995408639e-07, "loss": 1.2398, "step": 7467 }, { "epoch": 0.58, "learning_rate": 7.92244044886022e-07, "loss": 1.2401, "step": 7468 }, { "epoch": 0.58, "learning_rate": 7.919981488149575e-07, "loss": 1.1845, "step": 7469 }, { "epoch": 0.58, "learning_rate": 7.917522658887359e-07, "loss": 1.216, "step": 7470 }, { "epoch": 0.58, "learning_rate": 7.91506396122896e-07, "loss": 1.2035, "step": 7471 }, { "epoch": 0.58, "learning_rate": 7.912605395329762e-07, "loss": 1.1885, "step": 7472 }, { "epoch": 0.58, "learning_rate": 7.910146961345127e-07, "loss": 1.2454, "step": 7473 }, { "epoch": 0.58, "learning_rate": 7.907688659430425e-07, "loss": 1.1899, "step": 7474 }, { "epoch": 0.58, "learning_rate": 7.905230489741011e-07, "loss": 1.1842, "step": 7475 }, { "epoch": 0.58, "learning_rate": 7.902772452432222e-07, "loss": 1.2408, "step": 7476 }, { "epoch": 0.58, "learning_rate": 7.900314547659407e-07, "loss": 1.1836, "step": 7477 }, { "epoch": 0.58, "learning_rate": 7.897856775577891e-07, "loss": 0.8291, "step": 7478 }, { "epoch": 0.58, "learning_rate": 7.895399136342988e-07, "loss": 1.1919, "step": 7479 }, { "epoch": 0.58, "learning_rate": 7.892941630110023e-07, "loss": 1.2389, "step": 7480 }, { "epoch": 0.58, "learning_rate": 7.890484257034289e-07, "loss": 1.1721, "step": 7481 }, { "epoch": 0.58, "learning_rate": 7.888027017271087e-07, "loss": 1.2106, "step": 7482 }, { "epoch": 0.58, "learning_rate": 7.885569910975703e-07, "loss": 1.1987, "step": 7483 }, { "epoch": 0.58, "learning_rate": 7.883112938303419e-07, "loss": 1.2213, "step": 7484 }, { "epoch": 0.58, "learning_rate": 7.8806560994095e-07, "loss": 1.1807, "step": 7485 }, { "epoch": 0.58, "learning_rate": 7.878199394449212e-07, "loss": 1.2191, "step": 7486 }, { "epoch": 0.58, "learning_rate": 7.875742823577805e-07, "loss": 1.1996, "step": 7487 }, { "epoch": 0.58, "learning_rate": 7.873286386950527e-07, "loss": 1.2252, "step": 7488 }, { "epoch": 0.58, "learning_rate": 7.870830084722614e-07, "loss": 0.8047, "step": 7489 }, { "epoch": 0.58, "learning_rate": 7.868373917049292e-07, "loss": 1.1887, "step": 7490 }, { "epoch": 0.58, "learning_rate": 7.86591788408578e-07, "loss": 1.2014, "step": 7491 }, { "epoch": 0.58, "learning_rate": 7.863461985987295e-07, "loss": 1.2258, "step": 7492 }, { "epoch": 0.58, "learning_rate": 7.86100622290903e-07, "loss": 1.2176, "step": 7493 }, { "epoch": 0.58, "learning_rate": 7.858550595006188e-07, "loss": 1.2093, "step": 7494 }, { "epoch": 0.58, "learning_rate": 7.856095102433952e-07, "loss": 1.1584, "step": 7495 }, { "epoch": 0.58, "learning_rate": 7.853639745347492e-07, "loss": 1.1711, "step": 7496 }, { "epoch": 0.58, "learning_rate": 7.851184523901986e-07, "loss": 1.196, "step": 7497 }, { "epoch": 0.58, "learning_rate": 7.848729438252585e-07, "loss": 1.1933, "step": 7498 }, { "epoch": 0.58, "learning_rate": 7.846274488554446e-07, "loss": 1.1818, "step": 7499 }, { "epoch": 0.58, "learning_rate": 7.843819674962713e-07, "loss": 1.1983, "step": 7500 }, { "epoch": 0.58, "learning_rate": 7.841364997632511e-07, "loss": 1.2382, "step": 7501 }, { "epoch": 0.58, "learning_rate": 7.838910456718975e-07, "loss": 1.2505, "step": 7502 }, { "epoch": 0.58, "learning_rate": 7.836456052377216e-07, "loss": 1.1664, "step": 7503 }, { "epoch": 0.58, "learning_rate": 7.834001784762344e-07, "loss": 1.2426, "step": 7504 }, { "epoch": 0.58, "learning_rate": 7.831547654029455e-07, "loss": 1.1686, "step": 7505 }, { "epoch": 0.58, "learning_rate": 7.829093660333649e-07, "loss": 1.1845, "step": 7506 }, { "epoch": 0.58, "learning_rate": 7.826639803829999e-07, "loss": 1.2047, "step": 7507 }, { "epoch": 0.58, "learning_rate": 7.824186084673578e-07, "loss": 1.2161, "step": 7508 }, { "epoch": 0.58, "learning_rate": 7.82173250301946e-07, "loss": 1.1649, "step": 7509 }, { "epoch": 0.58, "learning_rate": 7.81927905902269e-07, "loss": 1.1579, "step": 7510 }, { "epoch": 0.58, "learning_rate": 7.816825752838323e-07, "loss": 1.1521, "step": 7511 }, { "epoch": 0.58, "learning_rate": 7.814372584621397e-07, "loss": 1.1586, "step": 7512 }, { "epoch": 0.58, "learning_rate": 7.811919554526935e-07, "loss": 1.2296, "step": 7513 }, { "epoch": 0.58, "learning_rate": 7.80946666270997e-07, "loss": 1.2296, "step": 7514 }, { "epoch": 0.58, "learning_rate": 7.807013909325502e-07, "loss": 1.1388, "step": 7515 }, { "epoch": 0.58, "learning_rate": 7.804561294528541e-07, "loss": 1.2087, "step": 7516 }, { "epoch": 0.58, "learning_rate": 7.802108818474084e-07, "loss": 1.1961, "step": 7517 }, { "epoch": 0.58, "learning_rate": 7.799656481317111e-07, "loss": 1.2007, "step": 7518 }, { "epoch": 0.58, "learning_rate": 7.797204283212607e-07, "loss": 1.2024, "step": 7519 }, { "epoch": 0.58, "learning_rate": 7.794752224315534e-07, "loss": 1.2408, "step": 7520 }, { "epoch": 0.58, "learning_rate": 7.792300304780855e-07, "loss": 1.2076, "step": 7521 }, { "epoch": 0.58, "learning_rate": 7.789848524763522e-07, "loss": 1.1979, "step": 7522 }, { "epoch": 0.58, "learning_rate": 7.787396884418472e-07, "loss": 1.2056, "step": 7523 }, { "epoch": 0.58, "learning_rate": 7.784945383900644e-07, "loss": 1.1754, "step": 7524 }, { "epoch": 0.58, "learning_rate": 7.782494023364959e-07, "loss": 1.2202, "step": 7525 }, { "epoch": 0.58, "learning_rate": 7.780042802966337e-07, "loss": 1.1412, "step": 7526 }, { "epoch": 0.58, "learning_rate": 7.777591722859678e-07, "loss": 1.2125, "step": 7527 }, { "epoch": 0.58, "learning_rate": 7.775140783199888e-07, "loss": 1.2372, "step": 7528 }, { "epoch": 0.58, "learning_rate": 7.772689984141852e-07, "loss": 1.1929, "step": 7529 }, { "epoch": 0.58, "learning_rate": 7.770239325840445e-07, "loss": 1.2415, "step": 7530 }, { "epoch": 0.58, "learning_rate": 7.767788808450551e-07, "loss": 1.1769, "step": 7531 }, { "epoch": 0.58, "learning_rate": 7.765338432127018e-07, "loss": 1.2107, "step": 7532 }, { "epoch": 0.58, "learning_rate": 7.76288819702471e-07, "loss": 1.2355, "step": 7533 }, { "epoch": 0.58, "learning_rate": 7.76043810329847e-07, "loss": 1.2032, "step": 7534 }, { "epoch": 0.58, "learning_rate": 7.757988151103127e-07, "loss": 1.175, "step": 7535 }, { "epoch": 0.58, "learning_rate": 7.755538340593514e-07, "loss": 1.2116, "step": 7536 }, { "epoch": 0.58, "learning_rate": 7.753088671924448e-07, "loss": 1.2391, "step": 7537 }, { "epoch": 0.59, "learning_rate": 7.750639145250736e-07, "loss": 1.1439, "step": 7538 }, { "epoch": 0.59, "learning_rate": 7.748189760727182e-07, "loss": 0.8379, "step": 7539 }, { "epoch": 0.59, "learning_rate": 7.745740518508567e-07, "loss": 1.1798, "step": 7540 }, { "epoch": 0.59, "learning_rate": 7.743291418749684e-07, "loss": 1.1884, "step": 7541 }, { "epoch": 0.59, "learning_rate": 7.740842461605297e-07, "loss": 1.2183, "step": 7542 }, { "epoch": 0.59, "learning_rate": 7.738393647230179e-07, "loss": 1.2056, "step": 7543 }, { "epoch": 0.59, "learning_rate": 7.735944975779076e-07, "loss": 1.229, "step": 7544 }, { "epoch": 0.59, "learning_rate": 7.733496447406739e-07, "loss": 1.2383, "step": 7545 }, { "epoch": 0.59, "learning_rate": 7.731048062267903e-07, "loss": 1.1686, "step": 7546 }, { "epoch": 0.59, "learning_rate": 7.728599820517293e-07, "loss": 1.2332, "step": 7547 }, { "epoch": 0.59, "learning_rate": 7.726151722309635e-07, "loss": 1.1971, "step": 7548 }, { "epoch": 0.59, "learning_rate": 7.72370376779963e-07, "loss": 1.2465, "step": 7549 }, { "epoch": 0.59, "learning_rate": 7.721255957141984e-07, "loss": 1.2053, "step": 7550 }, { "epoch": 0.59, "learning_rate": 7.718808290491388e-07, "loss": 1.1756, "step": 7551 }, { "epoch": 0.59, "learning_rate": 7.716360768002519e-07, "loss": 1.2092, "step": 7552 }, { "epoch": 0.59, "learning_rate": 7.713913389830057e-07, "loss": 1.1717, "step": 7553 }, { "epoch": 0.59, "learning_rate": 7.711466156128662e-07, "loss": 1.2058, "step": 7554 }, { "epoch": 0.59, "learning_rate": 7.709019067052991e-07, "loss": 1.189, "step": 7555 }, { "epoch": 0.59, "learning_rate": 7.706572122757689e-07, "loss": 1.1903, "step": 7556 }, { "epoch": 0.59, "learning_rate": 7.70412532339739e-07, "loss": 1.1685, "step": 7557 }, { "epoch": 0.59, "learning_rate": 7.701678669126725e-07, "loss": 1.2239, "step": 7558 }, { "epoch": 0.59, "learning_rate": 7.699232160100308e-07, "loss": 1.2065, "step": 7559 }, { "epoch": 0.59, "learning_rate": 7.696785796472757e-07, "loss": 1.2292, "step": 7560 }, { "epoch": 0.59, "learning_rate": 7.694339578398662e-07, "loss": 1.1579, "step": 7561 }, { "epoch": 0.59, "learning_rate": 7.691893506032618e-07, "loss": 1.2213, "step": 7562 }, { "epoch": 0.59, "learning_rate": 7.689447579529207e-07, "loss": 1.2036, "step": 7563 }, { "epoch": 0.59, "learning_rate": 7.687001799043e-07, "loss": 1.2209, "step": 7564 }, { "epoch": 0.59, "learning_rate": 7.684556164728562e-07, "loss": 1.1945, "step": 7565 }, { "epoch": 0.59, "learning_rate": 7.682110676740441e-07, "loss": 1.2336, "step": 7566 }, { "epoch": 0.59, "learning_rate": 7.67966533523319e-07, "loss": 1.2016, "step": 7567 }, { "epoch": 0.59, "learning_rate": 7.67722014036134e-07, "loss": 1.1904, "step": 7568 }, { "epoch": 0.59, "learning_rate": 7.674775092279415e-07, "loss": 1.2054, "step": 7569 }, { "epoch": 0.59, "learning_rate": 7.672330191141934e-07, "loss": 0.8426, "step": 7570 }, { "epoch": 0.59, "learning_rate": 7.669885437103408e-07, "loss": 1.1947, "step": 7571 }, { "epoch": 0.59, "learning_rate": 7.667440830318328e-07, "loss": 1.1681, "step": 7572 }, { "epoch": 0.59, "learning_rate": 7.664996370941191e-07, "loss": 1.191, "step": 7573 }, { "epoch": 0.59, "learning_rate": 7.662552059126466e-07, "loss": 1.1637, "step": 7574 }, { "epoch": 0.59, "learning_rate": 7.660107895028633e-07, "loss": 1.2068, "step": 7575 }, { "epoch": 0.59, "learning_rate": 7.657663878802145e-07, "loss": 1.2439, "step": 7576 }, { "epoch": 0.59, "learning_rate": 7.655220010601464e-07, "loss": 1.2091, "step": 7577 }, { "epoch": 0.59, "learning_rate": 7.652776290581023e-07, "loss": 1.223, "step": 7578 }, { "epoch": 0.59, "learning_rate": 7.650332718895258e-07, "loss": 1.1971, "step": 7579 }, { "epoch": 0.59, "learning_rate": 7.647889295698593e-07, "loss": 1.1697, "step": 7580 }, { "epoch": 0.59, "learning_rate": 7.645446021145439e-07, "loss": 1.1928, "step": 7581 }, { "epoch": 0.59, "learning_rate": 7.643002895390207e-07, "loss": 0.8549, "step": 7582 }, { "epoch": 0.59, "learning_rate": 7.640559918587286e-07, "loss": 1.2063, "step": 7583 }, { "epoch": 0.59, "learning_rate": 7.638117090891062e-07, "loss": 1.2289, "step": 7584 }, { "epoch": 0.59, "learning_rate": 7.635674412455918e-07, "loss": 1.1815, "step": 7585 }, { "epoch": 0.59, "learning_rate": 7.633231883436213e-07, "loss": 1.2158, "step": 7586 }, { "epoch": 0.59, "learning_rate": 7.630789503986311e-07, "loss": 1.173, "step": 7587 }, { "epoch": 0.59, "learning_rate": 7.628347274260555e-07, "loss": 1.1837, "step": 7588 }, { "epoch": 0.59, "learning_rate": 7.625905194413287e-07, "loss": 1.1995, "step": 7589 }, { "epoch": 0.59, "learning_rate": 7.623463264598837e-07, "loss": 0.8128, "step": 7590 }, { "epoch": 0.59, "learning_rate": 7.621021484971519e-07, "loss": 1.1672, "step": 7591 }, { "epoch": 0.59, "learning_rate": 7.61857985568565e-07, "loss": 1.2409, "step": 7592 }, { "epoch": 0.59, "learning_rate": 7.616138376895524e-07, "loss": 1.1816, "step": 7593 }, { "epoch": 0.59, "learning_rate": 7.613697048755442e-07, "loss": 1.1612, "step": 7594 }, { "epoch": 0.59, "learning_rate": 7.611255871419678e-07, "loss": 1.2241, "step": 7595 }, { "epoch": 0.59, "learning_rate": 7.608814845042504e-07, "loss": 1.2277, "step": 7596 }, { "epoch": 0.59, "learning_rate": 7.606373969778185e-07, "loss": 1.1755, "step": 7597 }, { "epoch": 0.59, "learning_rate": 7.603933245780971e-07, "loss": 1.2181, "step": 7598 }, { "epoch": 0.59, "learning_rate": 7.601492673205115e-07, "loss": 1.2866, "step": 7599 }, { "epoch": 0.59, "learning_rate": 7.599052252204839e-07, "loss": 1.178, "step": 7600 }, { "epoch": 0.59, "learning_rate": 7.59661198293437e-07, "loss": 1.2427, "step": 7601 }, { "epoch": 0.59, "learning_rate": 7.59417186554793e-07, "loss": 1.2005, "step": 7602 }, { "epoch": 0.59, "learning_rate": 7.591731900199716e-07, "loss": 1.1857, "step": 7603 }, { "epoch": 0.59, "learning_rate": 7.58929208704393e-07, "loss": 1.1616, "step": 7604 }, { "epoch": 0.59, "learning_rate": 7.586852426234753e-07, "loss": 1.2321, "step": 7605 }, { "epoch": 0.59, "learning_rate": 7.58441291792636e-07, "loss": 1.199, "step": 7606 }, { "epoch": 0.59, "learning_rate": 7.581973562272927e-07, "loss": 1.2197, "step": 7607 }, { "epoch": 0.59, "learning_rate": 7.579534359428598e-07, "loss": 0.8163, "step": 7608 }, { "epoch": 0.59, "learning_rate": 7.577095309547531e-07, "loss": 1.2429, "step": 7609 }, { "epoch": 0.59, "learning_rate": 7.574656412783858e-07, "loss": 1.2058, "step": 7610 }, { "epoch": 0.59, "learning_rate": 7.572217669291712e-07, "loss": 1.1676, "step": 7611 }, { "epoch": 0.59, "learning_rate": 7.569779079225205e-07, "loss": 1.1731, "step": 7612 }, { "epoch": 0.59, "learning_rate": 7.567340642738451e-07, "loss": 1.1725, "step": 7613 }, { "epoch": 0.59, "learning_rate": 7.564902359985546e-07, "loss": 1.154, "step": 7614 }, { "epoch": 0.59, "learning_rate": 7.562464231120578e-07, "loss": 0.8172, "step": 7615 }, { "epoch": 0.59, "learning_rate": 7.560026256297632e-07, "loss": 1.2168, "step": 7616 }, { "epoch": 0.59, "learning_rate": 7.557588435670774e-07, "loss": 1.1906, "step": 7617 }, { "epoch": 0.59, "learning_rate": 7.55515076939406e-07, "loss": 1.2123, "step": 7618 }, { "epoch": 0.59, "learning_rate": 7.552713257621549e-07, "loss": 1.1997, "step": 7619 }, { "epoch": 0.59, "learning_rate": 7.550275900507273e-07, "loss": 1.1863, "step": 7620 }, { "epoch": 0.59, "learning_rate": 7.547838698205273e-07, "loss": 1.2718, "step": 7621 }, { "epoch": 0.59, "learning_rate": 7.545401650869559e-07, "loss": 1.2143, "step": 7622 }, { "epoch": 0.59, "learning_rate": 7.542964758654147e-07, "loss": 1.2273, "step": 7623 }, { "epoch": 0.59, "learning_rate": 7.540528021713041e-07, "loss": 1.1931, "step": 7624 }, { "epoch": 0.59, "learning_rate": 7.538091440200225e-07, "loss": 1.2063, "step": 7625 }, { "epoch": 0.59, "learning_rate": 7.535655014269689e-07, "loss": 1.2157, "step": 7626 }, { "epoch": 0.59, "learning_rate": 7.533218744075403e-07, "loss": 1.1512, "step": 7627 }, { "epoch": 0.59, "learning_rate": 7.530782629771323e-07, "loss": 1.2086, "step": 7628 }, { "epoch": 0.59, "learning_rate": 7.528346671511407e-07, "loss": 1.2472, "step": 7629 }, { "epoch": 0.59, "learning_rate": 7.525910869449598e-07, "loss": 1.2206, "step": 7630 }, { "epoch": 0.59, "learning_rate": 7.523475223739825e-07, "loss": 1.2316, "step": 7631 }, { "epoch": 0.59, "learning_rate": 7.521039734536008e-07, "loss": 1.1916, "step": 7632 }, { "epoch": 0.59, "learning_rate": 7.51860440199207e-07, "loss": 1.2256, "step": 7633 }, { "epoch": 0.59, "learning_rate": 7.516169226261906e-07, "loss": 1.2034, "step": 7634 }, { "epoch": 0.59, "learning_rate": 7.513734207499407e-07, "loss": 1.1934, "step": 7635 }, { "epoch": 0.59, "learning_rate": 7.511299345858464e-07, "loss": 1.2668, "step": 7636 }, { "epoch": 0.59, "learning_rate": 7.50886464149294e-07, "loss": 1.2094, "step": 7637 }, { "epoch": 0.59, "learning_rate": 7.506430094556708e-07, "loss": 0.8111, "step": 7638 }, { "epoch": 0.59, "learning_rate": 7.503995705203615e-07, "loss": 0.8145, "step": 7639 }, { "epoch": 0.59, "learning_rate": 7.501561473587503e-07, "loss": 0.8442, "step": 7640 }, { "epoch": 0.59, "learning_rate": 7.499127399862213e-07, "loss": 1.2136, "step": 7641 }, { "epoch": 0.59, "learning_rate": 7.496693484181558e-07, "loss": 1.242, "step": 7642 }, { "epoch": 0.59, "learning_rate": 7.494259726699359e-07, "loss": 1.1834, "step": 7643 }, { "epoch": 0.59, "learning_rate": 7.491826127569419e-07, "loss": 1.1923, "step": 7644 }, { "epoch": 0.59, "learning_rate": 7.489392686945524e-07, "loss": 1.2192, "step": 7645 }, { "epoch": 0.59, "learning_rate": 7.486959404981463e-07, "loss": 1.2094, "step": 7646 }, { "epoch": 0.59, "learning_rate": 7.484526281831009e-07, "loss": 1.2057, "step": 7647 }, { "epoch": 0.59, "learning_rate": 7.482093317647926e-07, "loss": 1.1498, "step": 7648 }, { "epoch": 0.59, "learning_rate": 7.479660512585966e-07, "loss": 0.8831, "step": 7649 }, { "epoch": 0.59, "learning_rate": 7.477227866798868e-07, "loss": 1.214, "step": 7650 }, { "epoch": 0.59, "learning_rate": 7.474795380440371e-07, "loss": 0.8086, "step": 7651 }, { "epoch": 0.59, "learning_rate": 7.472363053664193e-07, "loss": 1.199, "step": 7652 }, { "epoch": 0.59, "learning_rate": 7.469930886624052e-07, "loss": 1.1795, "step": 7653 }, { "epoch": 0.59, "learning_rate": 7.467498879473646e-07, "loss": 1.2028, "step": 7654 }, { "epoch": 0.59, "learning_rate": 7.465067032366673e-07, "loss": 1.209, "step": 7655 }, { "epoch": 0.59, "learning_rate": 7.46263534545681e-07, "loss": 1.1422, "step": 7656 }, { "epoch": 0.59, "learning_rate": 7.46020381889773e-07, "loss": 1.2046, "step": 7657 }, { "epoch": 0.59, "learning_rate": 7.457772452843101e-07, "loss": 1.206, "step": 7658 }, { "epoch": 0.59, "learning_rate": 7.455341247446567e-07, "loss": 1.2149, "step": 7659 }, { "epoch": 0.59, "learning_rate": 7.452910202861775e-07, "loss": 1.2238, "step": 7660 }, { "epoch": 0.59, "learning_rate": 7.450479319242358e-07, "loss": 1.2328, "step": 7661 }, { "epoch": 0.59, "learning_rate": 7.448048596741932e-07, "loss": 1.189, "step": 7662 }, { "epoch": 0.59, "learning_rate": 7.445618035514116e-07, "loss": 1.1794, "step": 7663 }, { "epoch": 0.59, "learning_rate": 7.443187635712503e-07, "loss": 1.1869, "step": 7664 }, { "epoch": 0.59, "learning_rate": 7.44075739749069e-07, "loss": 1.2221, "step": 7665 }, { "epoch": 0.59, "learning_rate": 7.438327321002258e-07, "loss": 1.1952, "step": 7666 }, { "epoch": 0.6, "learning_rate": 7.435897406400772e-07, "loss": 1.1781, "step": 7667 }, { "epoch": 0.6, "learning_rate": 7.433467653839798e-07, "loss": 1.2147, "step": 7668 }, { "epoch": 0.6, "learning_rate": 7.431038063472881e-07, "loss": 1.1674, "step": 7669 }, { "epoch": 0.6, "learning_rate": 7.428608635453568e-07, "loss": 1.2173, "step": 7670 }, { "epoch": 0.6, "learning_rate": 7.426179369935384e-07, "loss": 1.2246, "step": 7671 }, { "epoch": 0.6, "learning_rate": 7.423750267071849e-07, "loss": 1.1389, "step": 7672 }, { "epoch": 0.6, "learning_rate": 7.421321327016472e-07, "loss": 1.2095, "step": 7673 }, { "epoch": 0.6, "learning_rate": 7.41889254992275e-07, "loss": 1.1953, "step": 7674 }, { "epoch": 0.6, "learning_rate": 7.416463935944177e-07, "loss": 1.2492, "step": 7675 }, { "epoch": 0.6, "learning_rate": 7.414035485234224e-07, "loss": 1.2351, "step": 7676 }, { "epoch": 0.6, "learning_rate": 7.411607197946364e-07, "loss": 1.2046, "step": 7677 }, { "epoch": 0.6, "learning_rate": 7.409179074234055e-07, "loss": 1.1824, "step": 7678 }, { "epoch": 0.6, "learning_rate": 7.40675111425074e-07, "loss": 1.2109, "step": 7679 }, { "epoch": 0.6, "learning_rate": 7.404323318149861e-07, "loss": 1.162, "step": 7680 }, { "epoch": 0.6, "learning_rate": 7.401895686084838e-07, "loss": 1.2464, "step": 7681 }, { "epoch": 0.6, "learning_rate": 7.399468218209093e-07, "loss": 1.1669, "step": 7682 }, { "epoch": 0.6, "learning_rate": 7.397040914676032e-07, "loss": 0.8314, "step": 7683 }, { "epoch": 0.6, "learning_rate": 7.394613775639043e-07, "loss": 1.1988, "step": 7684 }, { "epoch": 0.6, "learning_rate": 7.392186801251517e-07, "loss": 1.2342, "step": 7685 }, { "epoch": 0.6, "learning_rate": 7.389759991666829e-07, "loss": 1.2361, "step": 7686 }, { "epoch": 0.6, "learning_rate": 7.387333347038342e-07, "loss": 1.2059, "step": 7687 }, { "epoch": 0.6, "learning_rate": 7.384906867519409e-07, "loss": 1.2482, "step": 7688 }, { "epoch": 0.6, "learning_rate": 7.382480553263375e-07, "loss": 0.8612, "step": 7689 }, { "epoch": 0.6, "learning_rate": 7.38005440442357e-07, "loss": 1.2184, "step": 7690 }, { "epoch": 0.6, "learning_rate": 7.377628421153317e-07, "loss": 1.2743, "step": 7691 }, { "epoch": 0.6, "learning_rate": 7.375202603605933e-07, "loss": 1.1248, "step": 7692 }, { "epoch": 0.6, "learning_rate": 7.372776951934714e-07, "loss": 1.2162, "step": 7693 }, { "epoch": 0.6, "learning_rate": 7.37035146629295e-07, "loss": 1.2381, "step": 7694 }, { "epoch": 0.6, "learning_rate": 7.367926146833928e-07, "loss": 1.2171, "step": 7695 }, { "epoch": 0.6, "learning_rate": 7.36550099371091e-07, "loss": 0.8343, "step": 7696 }, { "epoch": 0.6, "learning_rate": 7.363076007077163e-07, "loss": 1.1965, "step": 7697 }, { "epoch": 0.6, "learning_rate": 7.360651187085928e-07, "loss": 0.8539, "step": 7698 }, { "epoch": 0.6, "learning_rate": 7.358226533890452e-07, "loss": 1.1878, "step": 7699 }, { "epoch": 0.6, "learning_rate": 7.35580204764396e-07, "loss": 1.2054, "step": 7700 }, { "epoch": 0.6, "learning_rate": 7.353377728499663e-07, "loss": 1.2073, "step": 7701 }, { "epoch": 0.6, "learning_rate": 7.350953576610776e-07, "loss": 0.8645, "step": 7702 }, { "epoch": 0.6, "learning_rate": 7.348529592130492e-07, "loss": 0.8098, "step": 7703 }, { "epoch": 0.6, "learning_rate": 7.346105775211997e-07, "loss": 1.2196, "step": 7704 }, { "epoch": 0.6, "learning_rate": 7.343682126008465e-07, "loss": 1.1645, "step": 7705 }, { "epoch": 0.6, "learning_rate": 7.341258644673062e-07, "loss": 1.1839, "step": 7706 }, { "epoch": 0.6, "learning_rate": 7.338835331358941e-07, "loss": 1.2107, "step": 7707 }, { "epoch": 0.6, "learning_rate": 7.336412186219243e-07, "loss": 1.1966, "step": 7708 }, { "epoch": 0.6, "learning_rate": 7.333989209407107e-07, "loss": 1.166, "step": 7709 }, { "epoch": 0.6, "learning_rate": 7.331566401075648e-07, "loss": 1.2007, "step": 7710 }, { "epoch": 0.6, "learning_rate": 7.329143761377979e-07, "loss": 1.1913, "step": 7711 }, { "epoch": 0.6, "learning_rate": 7.326721290467206e-07, "loss": 1.2045, "step": 7712 }, { "epoch": 0.6, "learning_rate": 7.324298988496412e-07, "loss": 1.1903, "step": 7713 }, { "epoch": 0.6, "learning_rate": 7.32187685561868e-07, "loss": 1.2203, "step": 7714 }, { "epoch": 0.6, "learning_rate": 7.319454891987078e-07, "loss": 0.809, "step": 7715 }, { "epoch": 0.6, "learning_rate": 7.317033097754661e-07, "loss": 1.1686, "step": 7716 }, { "epoch": 0.6, "learning_rate": 7.314611473074484e-07, "loss": 1.203, "step": 7717 }, { "epoch": 0.6, "learning_rate": 7.312190018099573e-07, "loss": 0.8498, "step": 7718 }, { "epoch": 0.6, "learning_rate": 7.309768732982961e-07, "loss": 1.1692, "step": 7719 }, { "epoch": 0.6, "learning_rate": 7.307347617877662e-07, "loss": 1.1847, "step": 7720 }, { "epoch": 0.6, "learning_rate": 7.30492667293668e-07, "loss": 0.8467, "step": 7721 }, { "epoch": 0.6, "learning_rate": 7.302505898313006e-07, "loss": 0.8207, "step": 7722 }, { "epoch": 0.6, "learning_rate": 7.300085294159625e-07, "loss": 1.2248, "step": 7723 }, { "epoch": 0.6, "learning_rate": 7.297664860629511e-07, "loss": 1.2014, "step": 7724 }, { "epoch": 0.6, "learning_rate": 7.295244597875621e-07, "loss": 1.171, "step": 7725 }, { "epoch": 0.6, "learning_rate": 7.292824506050909e-07, "loss": 0.8597, "step": 7726 }, { "epoch": 0.6, "learning_rate": 7.290404585308314e-07, "loss": 0.8181, "step": 7727 }, { "epoch": 0.6, "learning_rate": 7.28798483580076e-07, "loss": 1.1071, "step": 7728 }, { "epoch": 0.6, "learning_rate": 7.285565257681175e-07, "loss": 1.2575, "step": 7729 }, { "epoch": 0.6, "learning_rate": 7.283145851102453e-07, "loss": 0.864, "step": 7730 }, { "epoch": 0.6, "learning_rate": 7.280726616217505e-07, "loss": 1.1729, "step": 7731 }, { "epoch": 0.6, "learning_rate": 7.278307553179204e-07, "loss": 1.2223, "step": 7732 }, { "epoch": 0.6, "learning_rate": 7.275888662140428e-07, "loss": 1.1927, "step": 7733 }, { "epoch": 0.6, "learning_rate": 7.273469943254048e-07, "loss": 1.2463, "step": 7734 }, { "epoch": 0.6, "learning_rate": 7.271051396672906e-07, "loss": 1.1848, "step": 7735 }, { "epoch": 0.6, "learning_rate": 7.268633022549851e-07, "loss": 1.1447, "step": 7736 }, { "epoch": 0.6, "learning_rate": 7.266214821037715e-07, "loss": 1.1669, "step": 7737 }, { "epoch": 0.6, "learning_rate": 7.263796792289312e-07, "loss": 0.8409, "step": 7738 }, { "epoch": 0.6, "learning_rate": 7.261378936457458e-07, "loss": 1.2019, "step": 7739 }, { "epoch": 0.6, "learning_rate": 7.258961253694943e-07, "loss": 1.2234, "step": 7740 }, { "epoch": 0.6, "learning_rate": 7.256543744154562e-07, "loss": 1.1969, "step": 7741 }, { "epoch": 0.6, "learning_rate": 7.254126407989087e-07, "loss": 1.2189, "step": 7742 }, { "epoch": 0.6, "learning_rate": 7.25170924535129e-07, "loss": 1.2196, "step": 7743 }, { "epoch": 0.6, "learning_rate": 7.249292256393919e-07, "loss": 1.1907, "step": 7744 }, { "epoch": 0.6, "learning_rate": 7.246875441269716e-07, "loss": 1.1924, "step": 7745 }, { "epoch": 0.6, "learning_rate": 7.244458800131421e-07, "loss": 1.215, "step": 7746 }, { "epoch": 0.6, "learning_rate": 7.242042333131749e-07, "loss": 1.2157, "step": 7747 }, { "epoch": 0.6, "learning_rate": 7.239626040423416e-07, "loss": 1.2474, "step": 7748 }, { "epoch": 0.6, "learning_rate": 7.237209922159117e-07, "loss": 1.2125, "step": 7749 }, { "epoch": 0.6, "learning_rate": 7.23479397849154e-07, "loss": 1.1687, "step": 7750 }, { "epoch": 0.6, "learning_rate": 7.232378209573368e-07, "loss": 1.184, "step": 7751 }, { "epoch": 0.6, "learning_rate": 7.229962615557261e-07, "loss": 1.1961, "step": 7752 }, { "epoch": 0.6, "learning_rate": 7.22754719659588e-07, "loss": 1.1686, "step": 7753 }, { "epoch": 0.6, "learning_rate": 7.225131952841868e-07, "loss": 1.1591, "step": 7754 }, { "epoch": 0.6, "learning_rate": 7.222716884447853e-07, "loss": 1.1846, "step": 7755 }, { "epoch": 0.6, "learning_rate": 7.220301991566467e-07, "loss": 1.2499, "step": 7756 }, { "epoch": 0.6, "learning_rate": 7.217887274350309e-07, "loss": 1.2093, "step": 7757 }, { "epoch": 0.6, "learning_rate": 7.215472732951988e-07, "loss": 1.2404, "step": 7758 }, { "epoch": 0.6, "learning_rate": 7.213058367524092e-07, "loss": 0.8273, "step": 7759 }, { "epoch": 0.6, "learning_rate": 7.210644178219194e-07, "loss": 1.2213, "step": 7760 }, { "epoch": 0.6, "learning_rate": 7.208230165189865e-07, "loss": 1.1503, "step": 7761 }, { "epoch": 0.6, "learning_rate": 7.205816328588657e-07, "loss": 1.1861, "step": 7762 }, { "epoch": 0.6, "learning_rate": 7.203402668568119e-07, "loss": 1.1972, "step": 7763 }, { "epoch": 0.6, "learning_rate": 7.200989185280779e-07, "loss": 1.2386, "step": 7764 }, { "epoch": 0.6, "learning_rate": 7.198575878879167e-07, "loss": 1.2531, "step": 7765 }, { "epoch": 0.6, "learning_rate": 7.196162749515784e-07, "loss": 1.1953, "step": 7766 }, { "epoch": 0.6, "learning_rate": 7.193749797343131e-07, "loss": 1.2121, "step": 7767 }, { "epoch": 0.6, "learning_rate": 7.191337022513706e-07, "loss": 1.2244, "step": 7768 }, { "epoch": 0.6, "learning_rate": 7.188924425179975e-07, "loss": 1.1727, "step": 7769 }, { "epoch": 0.6, "learning_rate": 7.186512005494411e-07, "loss": 1.2055, "step": 7770 }, { "epoch": 0.6, "learning_rate": 7.184099763609468e-07, "loss": 1.1921, "step": 7771 }, { "epoch": 0.6, "learning_rate": 7.181687699677583e-07, "loss": 1.182, "step": 7772 }, { "epoch": 0.6, "learning_rate": 7.1792758138512e-07, "loss": 1.198, "step": 7773 }, { "epoch": 0.6, "learning_rate": 7.176864106282728e-07, "loss": 1.2368, "step": 7774 }, { "epoch": 0.6, "learning_rate": 7.174452577124583e-07, "loss": 1.1684, "step": 7775 }, { "epoch": 0.6, "learning_rate": 7.172041226529166e-07, "loss": 1.2161, "step": 7776 }, { "epoch": 0.6, "learning_rate": 7.169630054648856e-07, "loss": 1.1666, "step": 7777 }, { "epoch": 0.6, "learning_rate": 7.167219061636036e-07, "loss": 0.8199, "step": 7778 }, { "epoch": 0.6, "learning_rate": 7.164808247643069e-07, "loss": 1.2029, "step": 7779 }, { "epoch": 0.6, "learning_rate": 7.162397612822308e-07, "loss": 1.2159, "step": 7780 }, { "epoch": 0.6, "learning_rate": 7.159987157326094e-07, "loss": 1.1654, "step": 7781 }, { "epoch": 0.6, "learning_rate": 7.157576881306759e-07, "loss": 0.8638, "step": 7782 }, { "epoch": 0.6, "learning_rate": 7.155166784916622e-07, "loss": 1.1581, "step": 7783 }, { "epoch": 0.6, "learning_rate": 7.152756868307988e-07, "loss": 1.2019, "step": 7784 }, { "epoch": 0.6, "learning_rate": 7.15034713163316e-07, "loss": 1.2062, "step": 7785 }, { "epoch": 0.6, "learning_rate": 7.147937575044416e-07, "loss": 1.1846, "step": 7786 }, { "epoch": 0.6, "learning_rate": 7.145528198694036e-07, "loss": 1.2281, "step": 7787 }, { "epoch": 0.6, "learning_rate": 7.143119002734281e-07, "loss": 1.176, "step": 7788 }, { "epoch": 0.6, "learning_rate": 7.140709987317397e-07, "loss": 1.1082, "step": 7789 }, { "epoch": 0.6, "learning_rate": 7.138301152595631e-07, "loss": 1.1823, "step": 7790 }, { "epoch": 0.6, "learning_rate": 7.135892498721206e-07, "loss": 1.1789, "step": 7791 }, { "epoch": 0.6, "learning_rate": 7.13348402584634e-07, "loss": 0.8362, "step": 7792 }, { "epoch": 0.6, "learning_rate": 7.131075734123243e-07, "loss": 1.2247, "step": 7793 }, { "epoch": 0.6, "learning_rate": 7.1286676237041e-07, "loss": 1.1494, "step": 7794 }, { "epoch": 0.6, "learning_rate": 7.1262596947411e-07, "loss": 1.2134, "step": 7795 }, { "epoch": 0.61, "learning_rate": 7.123851947386412e-07, "loss": 1.1775, "step": 7796 }, { "epoch": 0.61, "learning_rate": 7.121444381792198e-07, "loss": 1.1892, "step": 7797 }, { "epoch": 0.61, "learning_rate": 7.119036998110603e-07, "loss": 1.1892, "step": 7798 }, { "epoch": 0.61, "learning_rate": 7.116629796493763e-07, "loss": 1.1859, "step": 7799 }, { "epoch": 0.61, "learning_rate": 7.114222777093806e-07, "loss": 1.2059, "step": 7800 }, { "epoch": 0.61, "learning_rate": 7.11181594006284e-07, "loss": 1.1342, "step": 7801 }, { "epoch": 0.61, "learning_rate": 7.109409285552977e-07, "loss": 1.205, "step": 7802 }, { "epoch": 0.61, "learning_rate": 7.107002813716299e-07, "loss": 1.1865, "step": 7803 }, { "epoch": 0.61, "learning_rate": 7.104596524704885e-07, "loss": 1.2222, "step": 7804 }, { "epoch": 0.61, "learning_rate": 7.102190418670808e-07, "loss": 1.2482, "step": 7805 }, { "epoch": 0.61, "learning_rate": 7.099784495766118e-07, "loss": 1.2028, "step": 7806 }, { "epoch": 0.61, "learning_rate": 7.097378756142864e-07, "loss": 1.253, "step": 7807 }, { "epoch": 0.61, "learning_rate": 7.094973199953074e-07, "loss": 1.1822, "step": 7808 }, { "epoch": 0.61, "learning_rate": 7.092567827348772e-07, "loss": 1.2002, "step": 7809 }, { "epoch": 0.61, "learning_rate": 7.090162638481969e-07, "loss": 1.2103, "step": 7810 }, { "epoch": 0.61, "learning_rate": 7.087757633504658e-07, "loss": 1.2528, "step": 7811 }, { "epoch": 0.61, "learning_rate": 7.085352812568828e-07, "loss": 1.2139, "step": 7812 }, { "epoch": 0.61, "learning_rate": 7.082948175826455e-07, "loss": 1.2775, "step": 7813 }, { "epoch": 0.61, "learning_rate": 7.0805437234295e-07, "loss": 1.2244, "step": 7814 }, { "epoch": 0.61, "learning_rate": 7.078139455529916e-07, "loss": 1.1807, "step": 7815 }, { "epoch": 0.61, "learning_rate": 7.075735372279636e-07, "loss": 1.1854, "step": 7816 }, { "epoch": 0.61, "learning_rate": 7.073331473830597e-07, "loss": 1.2201, "step": 7817 }, { "epoch": 0.61, "learning_rate": 7.07092776033471e-07, "loss": 1.2067, "step": 7818 }, { "epoch": 0.61, "learning_rate": 7.068524231943884e-07, "loss": 1.2143, "step": 7819 }, { "epoch": 0.61, "learning_rate": 7.066120888810007e-07, "loss": 1.1851, "step": 7820 }, { "epoch": 0.61, "learning_rate": 7.06371773108496e-07, "loss": 1.1705, "step": 7821 }, { "epoch": 0.61, "learning_rate": 7.06131475892062e-07, "loss": 1.1556, "step": 7822 }, { "epoch": 0.61, "learning_rate": 7.058911972468836e-07, "loss": 0.8486, "step": 7823 }, { "epoch": 0.61, "learning_rate": 7.056509371881458e-07, "loss": 1.1752, "step": 7824 }, { "epoch": 0.61, "learning_rate": 7.05410695731032e-07, "loss": 1.1778, "step": 7825 }, { "epoch": 0.61, "learning_rate": 7.051704728907241e-07, "loss": 1.2168, "step": 7826 }, { "epoch": 0.61, "learning_rate": 7.049302686824042e-07, "loss": 1.2352, "step": 7827 }, { "epoch": 0.61, "learning_rate": 7.046900831212508e-07, "loss": 1.2298, "step": 7828 }, { "epoch": 0.61, "learning_rate": 7.044499162224436e-07, "loss": 1.1932, "step": 7829 }, { "epoch": 0.61, "learning_rate": 7.042097680011596e-07, "loss": 1.1908, "step": 7830 }, { "epoch": 0.61, "learning_rate": 7.039696384725757e-07, "loss": 1.2329, "step": 7831 }, { "epoch": 0.61, "learning_rate": 7.037295276518669e-07, "loss": 1.1538, "step": 7832 }, { "epoch": 0.61, "learning_rate": 7.034894355542067e-07, "loss": 1.1977, "step": 7833 }, { "epoch": 0.61, "learning_rate": 7.032493621947684e-07, "loss": 1.2145, "step": 7834 }, { "epoch": 0.61, "learning_rate": 7.030093075887233e-07, "loss": 0.8699, "step": 7835 }, { "epoch": 0.61, "learning_rate": 7.027692717512427e-07, "loss": 1.1874, "step": 7836 }, { "epoch": 0.61, "learning_rate": 7.025292546974948e-07, "loss": 0.8532, "step": 7837 }, { "epoch": 0.61, "learning_rate": 7.022892564426479e-07, "loss": 1.1683, "step": 7838 }, { "epoch": 0.61, "learning_rate": 7.020492770018695e-07, "loss": 1.2332, "step": 7839 }, { "epoch": 0.61, "learning_rate": 7.018093163903244e-07, "loss": 1.203, "step": 7840 }, { "epoch": 0.61, "learning_rate": 7.015693746231781e-07, "loss": 1.2043, "step": 7841 }, { "epoch": 0.61, "learning_rate": 7.013294517155931e-07, "loss": 1.2312, "step": 7842 }, { "epoch": 0.61, "learning_rate": 7.010895476827317e-07, "loss": 1.2389, "step": 7843 }, { "epoch": 0.61, "learning_rate": 7.008496625397551e-07, "loss": 1.2048, "step": 7844 }, { "epoch": 0.61, "learning_rate": 7.006097963018225e-07, "loss": 1.2098, "step": 7845 }, { "epoch": 0.61, "learning_rate": 7.003699489840931e-07, "loss": 1.1182, "step": 7846 }, { "epoch": 0.61, "learning_rate": 7.00130120601724e-07, "loss": 1.2345, "step": 7847 }, { "epoch": 0.61, "learning_rate": 6.99890311169871e-07, "loss": 1.2495, "step": 7848 }, { "epoch": 0.61, "learning_rate": 6.996505207036897e-07, "loss": 1.1956, "step": 7849 }, { "epoch": 0.61, "learning_rate": 6.99410749218333e-07, "loss": 1.2208, "step": 7850 }, { "epoch": 0.61, "learning_rate": 6.991709967289542e-07, "loss": 1.1978, "step": 7851 }, { "epoch": 0.61, "learning_rate": 6.989312632507041e-07, "loss": 1.1832, "step": 7852 }, { "epoch": 0.61, "learning_rate": 6.986915487987337e-07, "loss": 1.2446, "step": 7853 }, { "epoch": 0.61, "learning_rate": 6.984518533881909e-07, "loss": 1.1816, "step": 7854 }, { "epoch": 0.61, "learning_rate": 6.982121770342242e-07, "loss": 0.8075, "step": 7855 }, { "epoch": 0.61, "learning_rate": 6.979725197519797e-07, "loss": 1.1939, "step": 7856 }, { "epoch": 0.61, "learning_rate": 6.977328815566026e-07, "loss": 1.2059, "step": 7857 }, { "epoch": 0.61, "learning_rate": 6.974932624632379e-07, "loss": 1.1785, "step": 7858 }, { "epoch": 0.61, "learning_rate": 6.972536624870276e-07, "loss": 0.8251, "step": 7859 }, { "epoch": 0.61, "learning_rate": 6.970140816431136e-07, "loss": 1.2236, "step": 7860 }, { "epoch": 0.61, "learning_rate": 6.96774519946637e-07, "loss": 1.205, "step": 7861 }, { "epoch": 0.61, "learning_rate": 6.965349774127361e-07, "loss": 1.1914, "step": 7862 }, { "epoch": 0.61, "learning_rate": 6.962954540565498e-07, "loss": 1.1897, "step": 7863 }, { "epoch": 0.61, "learning_rate": 6.96055949893215e-07, "loss": 1.1975, "step": 7864 }, { "epoch": 0.61, "learning_rate": 6.958164649378664e-07, "loss": 1.2504, "step": 7865 }, { "epoch": 0.61, "learning_rate": 6.955769992056398e-07, "loss": 1.2196, "step": 7866 }, { "epoch": 0.61, "learning_rate": 6.95337552711667e-07, "loss": 1.1906, "step": 7867 }, { "epoch": 0.61, "learning_rate": 6.950981254710811e-07, "loss": 1.218, "step": 7868 }, { "epoch": 0.61, "learning_rate": 6.948587174990128e-07, "loss": 1.1923, "step": 7869 }, { "epoch": 0.61, "learning_rate": 6.946193288105908e-07, "loss": 1.2501, "step": 7870 }, { "epoch": 0.61, "learning_rate": 6.943799594209444e-07, "loss": 0.8573, "step": 7871 }, { "epoch": 0.61, "learning_rate": 6.941406093452004e-07, "loss": 1.1939, "step": 7872 }, { "epoch": 0.61, "learning_rate": 6.939012785984847e-07, "loss": 1.2161, "step": 7873 }, { "epoch": 0.61, "learning_rate": 6.936619671959218e-07, "loss": 1.219, "step": 7874 }, { "epoch": 0.61, "learning_rate": 6.934226751526359e-07, "loss": 0.827, "step": 7875 }, { "epoch": 0.61, "learning_rate": 6.931834024837484e-07, "loss": 1.176, "step": 7876 }, { "epoch": 0.61, "learning_rate": 6.929441492043806e-07, "loss": 1.2508, "step": 7877 }, { "epoch": 0.61, "learning_rate": 6.927049153296527e-07, "loss": 1.2012, "step": 7878 }, { "epoch": 0.61, "learning_rate": 6.924657008746825e-07, "loss": 1.1656, "step": 7879 }, { "epoch": 0.61, "learning_rate": 6.922265058545881e-07, "loss": 0.8739, "step": 7880 }, { "epoch": 0.61, "learning_rate": 6.919873302844854e-07, "loss": 1.1961, "step": 7881 }, { "epoch": 0.61, "learning_rate": 6.91748174179489e-07, "loss": 1.1745, "step": 7882 }, { "epoch": 0.61, "learning_rate": 6.91509037554713e-07, "loss": 1.2017, "step": 7883 }, { "epoch": 0.61, "learning_rate": 6.912699204252692e-07, "loss": 1.1637, "step": 7884 }, { "epoch": 0.61, "learning_rate": 6.910308228062695e-07, "loss": 1.189, "step": 7885 }, { "epoch": 0.61, "learning_rate": 6.907917447128238e-07, "loss": 1.1798, "step": 7886 }, { "epoch": 0.61, "learning_rate": 6.905526861600401e-07, "loss": 1.2583, "step": 7887 }, { "epoch": 0.61, "learning_rate": 6.903136471630267e-07, "loss": 1.2225, "step": 7888 }, { "epoch": 0.61, "learning_rate": 6.900746277368894e-07, "loss": 1.1899, "step": 7889 }, { "epoch": 0.61, "learning_rate": 6.898356278967335e-07, "loss": 1.1533, "step": 7890 }, { "epoch": 0.61, "learning_rate": 6.895966476576628e-07, "loss": 1.195, "step": 7891 }, { "epoch": 0.61, "learning_rate": 6.893576870347792e-07, "loss": 1.2102, "step": 7892 }, { "epoch": 0.61, "learning_rate": 6.891187460431849e-07, "loss": 1.2304, "step": 7893 }, { "epoch": 0.61, "learning_rate": 6.888798246979794e-07, "loss": 1.2096, "step": 7894 }, { "epoch": 0.61, "learning_rate": 6.88640923014262e-07, "loss": 1.1767, "step": 7895 }, { "epoch": 0.61, "learning_rate": 6.884020410071296e-07, "loss": 1.1945, "step": 7896 }, { "epoch": 0.61, "learning_rate": 6.881631786916791e-07, "loss": 0.8259, "step": 7897 }, { "epoch": 0.61, "learning_rate": 6.879243360830057e-07, "loss": 1.2039, "step": 7898 }, { "epoch": 0.61, "learning_rate": 6.876855131962025e-07, "loss": 1.1748, "step": 7899 }, { "epoch": 0.61, "learning_rate": 6.874467100463631e-07, "loss": 0.8176, "step": 7900 }, { "epoch": 0.61, "learning_rate": 6.872079266485778e-07, "loss": 1.1927, "step": 7901 }, { "epoch": 0.61, "learning_rate": 6.869691630179375e-07, "loss": 0.8239, "step": 7902 }, { "epoch": 0.61, "learning_rate": 6.86730419169531e-07, "loss": 1.2238, "step": 7903 }, { "epoch": 0.61, "learning_rate": 6.864916951184453e-07, "loss": 1.1945, "step": 7904 }, { "epoch": 0.61, "learning_rate": 6.862529908797674e-07, "loss": 1.1545, "step": 7905 }, { "epoch": 0.61, "learning_rate": 6.86014306468582e-07, "loss": 1.1728, "step": 7906 }, { "epoch": 0.61, "learning_rate": 6.857756418999734e-07, "loss": 1.1964, "step": 7907 }, { "epoch": 0.61, "learning_rate": 6.855369971890239e-07, "loss": 0.8228, "step": 7908 }, { "epoch": 0.61, "learning_rate": 6.852983723508145e-07, "loss": 1.1969, "step": 7909 }, { "epoch": 0.61, "learning_rate": 6.850597674004259e-07, "loss": 1.2462, "step": 7910 }, { "epoch": 0.61, "learning_rate": 6.848211823529364e-07, "loss": 1.2241, "step": 7911 }, { "epoch": 0.61, "learning_rate": 6.845826172234242e-07, "loss": 0.877, "step": 7912 }, { "epoch": 0.61, "learning_rate": 6.843440720269653e-07, "loss": 1.2243, "step": 7913 }, { "epoch": 0.61, "learning_rate": 6.841055467786346e-07, "loss": 1.1663, "step": 7914 }, { "epoch": 0.61, "learning_rate": 6.83867041493506e-07, "loss": 1.1594, "step": 7915 }, { "epoch": 0.61, "learning_rate": 6.836285561866517e-07, "loss": 1.1716, "step": 7916 }, { "epoch": 0.61, "learning_rate": 6.833900908731438e-07, "loss": 1.2199, "step": 7917 }, { "epoch": 0.61, "learning_rate": 6.831516455680515e-07, "loss": 0.8632, "step": 7918 }, { "epoch": 0.61, "learning_rate": 6.82913220286444e-07, "loss": 1.1605, "step": 7919 }, { "epoch": 0.61, "learning_rate": 6.826748150433889e-07, "loss": 0.8637, "step": 7920 }, { "epoch": 0.61, "learning_rate": 6.824364298539516e-07, "loss": 0.8534, "step": 7921 }, { "epoch": 0.61, "learning_rate": 6.821980647331979e-07, "loss": 1.1799, "step": 7922 }, { "epoch": 0.61, "learning_rate": 6.819597196961913e-07, "loss": 1.2266, "step": 7923 }, { "epoch": 0.62, "learning_rate": 6.817213947579939e-07, "loss": 1.2123, "step": 7924 }, { "epoch": 0.62, "learning_rate": 6.814830899336671e-07, "loss": 1.22, "step": 7925 }, { "epoch": 0.62, "learning_rate": 6.812448052382704e-07, "loss": 1.232, "step": 7926 }, { "epoch": 0.62, "learning_rate": 6.810065406868629e-07, "loss": 1.2028, "step": 7927 }, { "epoch": 0.62, "learning_rate": 6.807682962945014e-07, "loss": 1.2126, "step": 7928 }, { "epoch": 0.62, "learning_rate": 6.805300720762426e-07, "loss": 0.8141, "step": 7929 }, { "epoch": 0.62, "learning_rate": 6.802918680471407e-07, "loss": 1.2112, "step": 7930 }, { "epoch": 0.62, "learning_rate": 6.800536842222492e-07, "loss": 0.8302, "step": 7931 }, { "epoch": 0.62, "learning_rate": 6.798155206166209e-07, "loss": 0.8428, "step": 7932 }, { "epoch": 0.62, "learning_rate": 6.795773772453059e-07, "loss": 1.189, "step": 7933 }, { "epoch": 0.62, "learning_rate": 6.793392541233547e-07, "loss": 1.1666, "step": 7934 }, { "epoch": 0.62, "learning_rate": 6.791011512658152e-07, "loss": 1.1091, "step": 7935 }, { "epoch": 0.62, "learning_rate": 6.788630686877343e-07, "loss": 1.2104, "step": 7936 }, { "epoch": 0.62, "learning_rate": 6.786250064041584e-07, "loss": 1.2237, "step": 7937 }, { "epoch": 0.62, "learning_rate": 6.783869644301315e-07, "loss": 1.214, "step": 7938 }, { "epoch": 0.62, "learning_rate": 6.781489427806972e-07, "loss": 1.1646, "step": 7939 }, { "epoch": 0.62, "learning_rate": 6.779109414708974e-07, "loss": 1.1613, "step": 7940 }, { "epoch": 0.62, "learning_rate": 6.776729605157727e-07, "loss": 1.2419, "step": 7941 }, { "epoch": 0.62, "learning_rate": 6.774349999303628e-07, "loss": 1.2021, "step": 7942 }, { "epoch": 0.62, "learning_rate": 6.77197059729705e-07, "loss": 1.2129, "step": 7943 }, { "epoch": 0.62, "learning_rate": 6.76959139928837e-07, "loss": 1.1805, "step": 7944 }, { "epoch": 0.62, "learning_rate": 6.767212405427936e-07, "loss": 0.8304, "step": 7945 }, { "epoch": 0.62, "learning_rate": 6.764833615866101e-07, "loss": 1.1909, "step": 7946 }, { "epoch": 0.62, "learning_rate": 6.762455030753185e-07, "loss": 1.1748, "step": 7947 }, { "epoch": 0.62, "learning_rate": 6.760076650239506e-07, "loss": 1.1886, "step": 7948 }, { "epoch": 0.62, "learning_rate": 6.757698474475371e-07, "loss": 1.2473, "step": 7949 }, { "epoch": 0.62, "learning_rate": 6.755320503611066e-07, "loss": 1.2172, "step": 7950 }, { "epoch": 0.62, "learning_rate": 6.752942737796876e-07, "loss": 1.1915, "step": 7951 }, { "epoch": 0.62, "learning_rate": 6.750565177183059e-07, "loss": 1.2426, "step": 7952 }, { "epoch": 0.62, "learning_rate": 6.748187821919867e-07, "loss": 1.1767, "step": 7953 }, { "epoch": 0.62, "learning_rate": 6.745810672157545e-07, "loss": 0.8121, "step": 7954 }, { "epoch": 0.62, "learning_rate": 6.743433728046313e-07, "loss": 1.2068, "step": 7955 }, { "epoch": 0.62, "learning_rate": 6.741056989736387e-07, "loss": 1.2044, "step": 7956 }, { "epoch": 0.62, "learning_rate": 6.738680457377965e-07, "loss": 1.1619, "step": 7957 }, { "epoch": 0.62, "learning_rate": 6.736304131121233e-07, "loss": 1.2132, "step": 7958 }, { "epoch": 0.62, "learning_rate": 6.733928011116371e-07, "loss": 1.1925, "step": 7959 }, { "epoch": 0.62, "learning_rate": 6.731552097513531e-07, "loss": 1.1588, "step": 7960 }, { "epoch": 0.62, "learning_rate": 6.729176390462865e-07, "loss": 1.1935, "step": 7961 }, { "epoch": 0.62, "learning_rate": 6.726800890114507e-07, "loss": 0.8351, "step": 7962 }, { "epoch": 0.62, "learning_rate": 6.724425596618584e-07, "loss": 1.2004, "step": 7963 }, { "epoch": 0.62, "learning_rate": 6.722050510125198e-07, "loss": 0.7896, "step": 7964 }, { "epoch": 0.62, "learning_rate": 6.719675630784446e-07, "loss": 1.2167, "step": 7965 }, { "epoch": 0.62, "learning_rate": 6.71730095874641e-07, "loss": 1.1981, "step": 7966 }, { "epoch": 0.62, "learning_rate": 6.714926494161159e-07, "loss": 1.2176, "step": 7967 }, { "epoch": 0.62, "learning_rate": 6.712552237178755e-07, "loss": 1.2279, "step": 7968 }, { "epoch": 0.62, "learning_rate": 6.710178187949232e-07, "loss": 1.2254, "step": 7969 }, { "epoch": 0.62, "learning_rate": 6.707804346622623e-07, "loss": 1.1746, "step": 7970 }, { "epoch": 0.62, "learning_rate": 6.705430713348952e-07, "loss": 0.8292, "step": 7971 }, { "epoch": 0.62, "learning_rate": 6.703057288278211e-07, "loss": 1.1789, "step": 7972 }, { "epoch": 0.62, "learning_rate": 6.700684071560399e-07, "loss": 1.191, "step": 7973 }, { "epoch": 0.62, "learning_rate": 6.698311063345492e-07, "loss": 1.249, "step": 7974 }, { "epoch": 0.62, "learning_rate": 6.695938263783448e-07, "loss": 1.194, "step": 7975 }, { "epoch": 0.62, "learning_rate": 6.693565673024228e-07, "loss": 0.8067, "step": 7976 }, { "epoch": 0.62, "learning_rate": 6.691193291217759e-07, "loss": 1.2033, "step": 7977 }, { "epoch": 0.62, "learning_rate": 6.688821118513975e-07, "loss": 0.8885, "step": 7978 }, { "epoch": 0.62, "learning_rate": 6.686449155062783e-07, "loss": 1.2056, "step": 7979 }, { "epoch": 0.62, "learning_rate": 6.684077401014079e-07, "loss": 1.1945, "step": 7980 }, { "epoch": 0.62, "learning_rate": 6.681705856517752e-07, "loss": 1.2262, "step": 7981 }, { "epoch": 0.62, "learning_rate": 6.679334521723672e-07, "loss": 1.2081, "step": 7982 }, { "epoch": 0.62, "learning_rate": 6.676963396781698e-07, "loss": 1.1939, "step": 7983 }, { "epoch": 0.62, "learning_rate": 6.674592481841671e-07, "loss": 1.1994, "step": 7984 }, { "epoch": 0.62, "learning_rate": 6.672221777053431e-07, "loss": 1.1822, "step": 7985 }, { "epoch": 0.62, "learning_rate": 6.66985128256679e-07, "loss": 1.1962, "step": 7986 }, { "epoch": 0.62, "learning_rate": 6.667480998531554e-07, "loss": 1.2234, "step": 7987 }, { "epoch": 0.62, "learning_rate": 6.665110925097519e-07, "loss": 1.2186, "step": 7988 }, { "epoch": 0.62, "learning_rate": 6.662741062414459e-07, "loss": 1.1706, "step": 7989 }, { "epoch": 0.62, "learning_rate": 6.660371410632143e-07, "loss": 1.1885, "step": 7990 }, { "epoch": 0.62, "learning_rate": 6.65800196990032e-07, "loss": 1.2116, "step": 7991 }, { "epoch": 0.62, "learning_rate": 6.655632740368728e-07, "loss": 0.8331, "step": 7992 }, { "epoch": 0.62, "learning_rate": 6.653263722187099e-07, "loss": 1.2043, "step": 7993 }, { "epoch": 0.62, "learning_rate": 6.650894915505136e-07, "loss": 1.2545, "step": 7994 }, { "epoch": 0.62, "learning_rate": 6.648526320472543e-07, "loss": 1.1269, "step": 7995 }, { "epoch": 0.62, "learning_rate": 6.646157937239008e-07, "loss": 1.184, "step": 7996 }, { "epoch": 0.62, "learning_rate": 6.643789765954194e-07, "loss": 1.185, "step": 7997 }, { "epoch": 0.62, "learning_rate": 6.641421806767766e-07, "loss": 1.1151, "step": 7998 }, { "epoch": 0.62, "learning_rate": 6.63905405982937e-07, "loss": 1.2074, "step": 7999 }, { "epoch": 0.62, "learning_rate": 6.636686525288633e-07, "loss": 1.1775, "step": 8000 }, { "epoch": 0.62, "learning_rate": 6.634319203295179e-07, "loss": 1.1589, "step": 8001 }, { "epoch": 0.62, "learning_rate": 6.631952093998605e-07, "loss": 1.2543, "step": 8002 }, { "epoch": 0.62, "learning_rate": 6.629585197548508e-07, "loss": 1.2236, "step": 8003 }, { "epoch": 0.62, "learning_rate": 6.627218514094464e-07, "loss": 1.2213, "step": 8004 }, { "epoch": 0.62, "learning_rate": 6.62485204378604e-07, "loss": 1.2707, "step": 8005 }, { "epoch": 0.62, "learning_rate": 6.622485786772782e-07, "loss": 1.2395, "step": 8006 }, { "epoch": 0.62, "learning_rate": 6.620119743204236e-07, "loss": 1.2274, "step": 8007 }, { "epoch": 0.62, "learning_rate": 6.617753913229916e-07, "loss": 1.1891, "step": 8008 }, { "epoch": 0.62, "learning_rate": 6.615388296999337e-07, "loss": 1.1838, "step": 8009 }, { "epoch": 0.62, "learning_rate": 6.613022894661999e-07, "loss": 1.2178, "step": 8010 }, { "epoch": 0.62, "learning_rate": 6.61065770636738e-07, "loss": 1.2239, "step": 8011 }, { "epoch": 0.62, "learning_rate": 6.608292732264953e-07, "loss": 1.234, "step": 8012 }, { "epoch": 0.62, "learning_rate": 6.605927972504178e-07, "loss": 1.2208, "step": 8013 }, { "epoch": 0.62, "learning_rate": 6.603563427234487e-07, "loss": 0.8591, "step": 8014 }, { "epoch": 0.62, "learning_rate": 6.60119909660532e-07, "loss": 1.1679, "step": 8015 }, { "epoch": 0.62, "learning_rate": 6.598834980766088e-07, "loss": 1.1763, "step": 8016 }, { "epoch": 0.62, "learning_rate": 6.596471079866195e-07, "loss": 1.1256, "step": 8017 }, { "epoch": 0.62, "learning_rate": 6.59410739405503e-07, "loss": 1.1896, "step": 8018 }, { "epoch": 0.62, "learning_rate": 6.591743923481965e-07, "loss": 1.1847, "step": 8019 }, { "epoch": 0.62, "learning_rate": 6.589380668296362e-07, "loss": 1.2155, "step": 8020 }, { "epoch": 0.62, "learning_rate": 6.58701762864757e-07, "loss": 1.2064, "step": 8021 }, { "epoch": 0.62, "learning_rate": 6.584654804684926e-07, "loss": 1.1955, "step": 8022 }, { "epoch": 0.62, "learning_rate": 6.582292196557746e-07, "loss": 1.1696, "step": 8023 }, { "epoch": 0.62, "learning_rate": 6.579929804415339e-07, "loss": 1.1406, "step": 8024 }, { "epoch": 0.62, "learning_rate": 6.577567628406997e-07, "loss": 1.1627, "step": 8025 }, { "epoch": 0.62, "learning_rate": 6.575205668682e-07, "loss": 1.1958, "step": 8026 }, { "epoch": 0.62, "learning_rate": 6.572843925389617e-07, "loss": 1.2341, "step": 8027 }, { "epoch": 0.62, "learning_rate": 6.570482398679093e-07, "loss": 1.2373, "step": 8028 }, { "epoch": 0.62, "learning_rate": 6.568121088699674e-07, "loss": 1.197, "step": 8029 }, { "epoch": 0.62, "learning_rate": 6.565759995600585e-07, "loss": 0.8178, "step": 8030 }, { "epoch": 0.62, "learning_rate": 6.563399119531029e-07, "loss": 1.2091, "step": 8031 }, { "epoch": 0.62, "learning_rate": 6.561038460640211e-07, "loss": 1.242, "step": 8032 }, { "epoch": 0.62, "learning_rate": 6.558678019077312e-07, "loss": 1.2146, "step": 8033 }, { "epoch": 0.62, "learning_rate": 6.556317794991504e-07, "loss": 0.8112, "step": 8034 }, { "epoch": 0.62, "learning_rate": 6.553957788531942e-07, "loss": 1.1789, "step": 8035 }, { "epoch": 0.62, "learning_rate": 6.551597999847764e-07, "loss": 1.2775, "step": 8036 }, { "epoch": 0.62, "learning_rate": 6.549238429088105e-07, "loss": 1.2212, "step": 8037 }, { "epoch": 0.62, "learning_rate": 6.546879076402076e-07, "loss": 1.1967, "step": 8038 }, { "epoch": 0.62, "learning_rate": 6.544519941938784e-07, "loss": 1.2199, "step": 8039 }, { "epoch": 0.62, "learning_rate": 6.542161025847311e-07, "loss": 1.1397, "step": 8040 }, { "epoch": 0.62, "learning_rate": 6.539802328276731e-07, "loss": 1.1771, "step": 8041 }, { "epoch": 0.62, "learning_rate": 6.537443849376106e-07, "loss": 1.2306, "step": 8042 }, { "epoch": 0.62, "learning_rate": 6.535085589294478e-07, "loss": 1.2118, "step": 8043 }, { "epoch": 0.62, "learning_rate": 6.532727548180886e-07, "loss": 1.1941, "step": 8044 }, { "epoch": 0.62, "learning_rate": 6.53036972618434e-07, "loss": 1.2494, "step": 8045 }, { "epoch": 0.62, "learning_rate": 6.528012123453847e-07, "loss": 1.1677, "step": 8046 }, { "epoch": 0.62, "learning_rate": 6.525654740138406e-07, "loss": 1.2109, "step": 8047 }, { "epoch": 0.62, "learning_rate": 6.523297576386981e-07, "loss": 1.1964, "step": 8048 }, { "epoch": 0.62, "learning_rate": 6.520940632348543e-07, "loss": 0.8246, "step": 8049 }, { "epoch": 0.62, "learning_rate": 6.518583908172038e-07, "loss": 1.1882, "step": 8050 }, { "epoch": 0.62, "learning_rate": 6.516227404006403e-07, "loss": 1.2326, "step": 8051 }, { "epoch": 0.62, "learning_rate": 6.513871120000558e-07, "loss": 1.196, "step": 8052 }, { "epoch": 0.63, "learning_rate": 6.511515056303407e-07, "loss": 1.2125, "step": 8053 }, { "epoch": 0.63, "learning_rate": 6.50915921306385e-07, "loss": 1.2041, "step": 8054 }, { "epoch": 0.63, "learning_rate": 6.506803590430758e-07, "loss": 1.2177, "step": 8055 }, { "epoch": 0.63, "learning_rate": 6.504448188553006e-07, "loss": 1.1815, "step": 8056 }, { "epoch": 0.63, "learning_rate": 6.502093007579441e-07, "loss": 0.8867, "step": 8057 }, { "epoch": 0.63, "learning_rate": 6.499738047658898e-07, "loss": 1.1989, "step": 8058 }, { "epoch": 0.63, "learning_rate": 6.497383308940205e-07, "loss": 0.858, "step": 8059 }, { "epoch": 0.63, "learning_rate": 6.495028791572166e-07, "loss": 1.1607, "step": 8060 }, { "epoch": 0.63, "learning_rate": 6.492674495703585e-07, "loss": 1.1983, "step": 8061 }, { "epoch": 0.63, "learning_rate": 6.490320421483235e-07, "loss": 1.2181, "step": 8062 }, { "epoch": 0.63, "learning_rate": 6.487966569059886e-07, "loss": 1.2576, "step": 8063 }, { "epoch": 0.63, "learning_rate": 6.485612938582299e-07, "loss": 1.1556, "step": 8064 }, { "epoch": 0.63, "learning_rate": 6.483259530199201e-07, "loss": 1.1426, "step": 8065 }, { "epoch": 0.63, "learning_rate": 6.480906344059329e-07, "loss": 1.208, "step": 8066 }, { "epoch": 0.63, "learning_rate": 6.478553380311387e-07, "loss": 1.2386, "step": 8067 }, { "epoch": 0.63, "learning_rate": 6.476200639104072e-07, "loss": 1.2289, "step": 8068 }, { "epoch": 0.63, "learning_rate": 6.473848120586076e-07, "loss": 1.2708, "step": 8069 }, { "epoch": 0.63, "learning_rate": 6.471495824906058e-07, "loss": 1.1896, "step": 8070 }, { "epoch": 0.63, "learning_rate": 6.46914375221268e-07, "loss": 1.1996, "step": 8071 }, { "epoch": 0.63, "learning_rate": 6.466791902654578e-07, "loss": 1.1896, "step": 8072 }, { "epoch": 0.63, "learning_rate": 6.464440276380386e-07, "loss": 1.1824, "step": 8073 }, { "epoch": 0.63, "learning_rate": 6.462088873538711e-07, "loss": 1.1903, "step": 8074 }, { "epoch": 0.63, "learning_rate": 6.459737694278153e-07, "loss": 1.1862, "step": 8075 }, { "epoch": 0.63, "learning_rate": 6.457386738747298e-07, "loss": 1.183, "step": 8076 }, { "epoch": 0.63, "learning_rate": 6.455036007094714e-07, "loss": 1.2209, "step": 8077 }, { "epoch": 0.63, "learning_rate": 6.452685499468962e-07, "loss": 1.182, "step": 8078 }, { "epoch": 0.63, "learning_rate": 6.450335216018579e-07, "loss": 1.2313, "step": 8079 }, { "epoch": 0.63, "learning_rate": 6.447985156892094e-07, "loss": 1.254, "step": 8080 }, { "epoch": 0.63, "learning_rate": 6.445635322238025e-07, "loss": 1.2201, "step": 8081 }, { "epoch": 0.63, "learning_rate": 6.443285712204867e-07, "loss": 1.2164, "step": 8082 }, { "epoch": 0.63, "learning_rate": 6.44093632694111e-07, "loss": 1.2169, "step": 8083 }, { "epoch": 0.63, "learning_rate": 6.43858716659522e-07, "loss": 1.2418, "step": 8084 }, { "epoch": 0.63, "learning_rate": 6.436238231315656e-07, "loss": 1.1867, "step": 8085 }, { "epoch": 0.63, "learning_rate": 6.433889521250864e-07, "loss": 1.1846, "step": 8086 }, { "epoch": 0.63, "learning_rate": 6.431541036549268e-07, "loss": 1.1889, "step": 8087 }, { "epoch": 0.63, "learning_rate": 6.429192777359284e-07, "loss": 1.2423, "step": 8088 }, { "epoch": 0.63, "learning_rate": 6.426844743829314e-07, "loss": 1.1934, "step": 8089 }, { "epoch": 0.63, "learning_rate": 6.424496936107744e-07, "loss": 1.171, "step": 8090 }, { "epoch": 0.63, "learning_rate": 6.422149354342943e-07, "loss": 1.1511, "step": 8091 }, { "epoch": 0.63, "learning_rate": 6.419801998683267e-07, "loss": 1.2198, "step": 8092 }, { "epoch": 0.63, "learning_rate": 6.417454869277063e-07, "loss": 1.1514, "step": 8093 }, { "epoch": 0.63, "learning_rate": 6.415107966272657e-07, "loss": 1.2033, "step": 8094 }, { "epoch": 0.63, "learning_rate": 6.412761289818367e-07, "loss": 0.8223, "step": 8095 }, { "epoch": 0.63, "learning_rate": 6.41041484006249e-07, "loss": 1.1945, "step": 8096 }, { "epoch": 0.63, "learning_rate": 6.408068617153308e-07, "loss": 1.2187, "step": 8097 }, { "epoch": 0.63, "learning_rate": 6.405722621239103e-07, "loss": 1.2196, "step": 8098 }, { "epoch": 0.63, "learning_rate": 6.403376852468121e-07, "loss": 1.1353, "step": 8099 }, { "epoch": 0.63, "learning_rate": 6.401031310988614e-07, "loss": 1.1702, "step": 8100 }, { "epoch": 0.63, "learning_rate": 6.398685996948802e-07, "loss": 1.1447, "step": 8101 }, { "epoch": 0.63, "learning_rate": 6.396340910496903e-07, "loss": 1.1484, "step": 8102 }, { "epoch": 0.63, "learning_rate": 6.393996051781121e-07, "loss": 1.2212, "step": 8103 }, { "epoch": 0.63, "learning_rate": 6.391651420949631e-07, "loss": 1.2043, "step": 8104 }, { "epoch": 0.63, "learning_rate": 6.389307018150612e-07, "loss": 1.1804, "step": 8105 }, { "epoch": 0.63, "learning_rate": 6.386962843532219e-07, "loss": 1.1796, "step": 8106 }, { "epoch": 0.63, "learning_rate": 6.38461889724259e-07, "loss": 0.8651, "step": 8107 }, { "epoch": 0.63, "learning_rate": 6.382275179429859e-07, "loss": 1.2073, "step": 8108 }, { "epoch": 0.63, "learning_rate": 6.379931690242134e-07, "loss": 1.1806, "step": 8109 }, { "epoch": 0.63, "learning_rate": 6.377588429827515e-07, "loss": 1.2493, "step": 8110 }, { "epoch": 0.63, "learning_rate": 6.375245398334086e-07, "loss": 1.1996, "step": 8111 }, { "epoch": 0.63, "learning_rate": 6.372902595909919e-07, "loss": 1.175, "step": 8112 }, { "epoch": 0.63, "learning_rate": 6.370560022703066e-07, "loss": 1.1798, "step": 8113 }, { "epoch": 0.63, "learning_rate": 6.368217678861569e-07, "loss": 1.2026, "step": 8114 }, { "epoch": 0.63, "learning_rate": 6.365875564533458e-07, "loss": 1.2047, "step": 8115 }, { "epoch": 0.63, "learning_rate": 6.363533679866737e-07, "loss": 1.2135, "step": 8116 }, { "epoch": 0.63, "learning_rate": 6.361192025009412e-07, "loss": 1.1721, "step": 8117 }, { "epoch": 0.63, "learning_rate": 6.358850600109459e-07, "loss": 1.1434, "step": 8118 }, { "epoch": 0.63, "learning_rate": 6.356509405314846e-07, "loss": 1.1619, "step": 8119 }, { "epoch": 0.63, "learning_rate": 6.354168440773534e-07, "loss": 1.2306, "step": 8120 }, { "epoch": 0.63, "learning_rate": 6.351827706633453e-07, "loss": 1.1814, "step": 8121 }, { "epoch": 0.63, "learning_rate": 6.349487203042535e-07, "loss": 1.229, "step": 8122 }, { "epoch": 0.63, "learning_rate": 6.347146930148688e-07, "loss": 1.2026, "step": 8123 }, { "epoch": 0.63, "learning_rate": 6.344806888099803e-07, "loss": 1.2176, "step": 8124 }, { "epoch": 0.63, "learning_rate": 6.342467077043765e-07, "loss": 0.8358, "step": 8125 }, { "epoch": 0.63, "learning_rate": 6.34012749712844e-07, "loss": 1.2487, "step": 8126 }, { "epoch": 0.63, "learning_rate": 6.33778814850168e-07, "loss": 0.8591, "step": 8127 }, { "epoch": 0.63, "learning_rate": 6.335449031311323e-07, "loss": 1.191, "step": 8128 }, { "epoch": 0.63, "learning_rate": 6.333110145705185e-07, "loss": 1.2288, "step": 8129 }, { "epoch": 0.63, "learning_rate": 6.330771491831079e-07, "loss": 1.2026, "step": 8130 }, { "epoch": 0.63, "learning_rate": 6.328433069836796e-07, "loss": 1.2235, "step": 8131 }, { "epoch": 0.63, "learning_rate": 6.326094879870121e-07, "loss": 1.1704, "step": 8132 }, { "epoch": 0.63, "learning_rate": 6.323756922078807e-07, "loss": 1.2107, "step": 8133 }, { "epoch": 0.63, "learning_rate": 6.321419196610613e-07, "loss": 1.2083, "step": 8134 }, { "epoch": 0.63, "learning_rate": 6.319081703613267e-07, "loss": 1.1905, "step": 8135 }, { "epoch": 0.63, "learning_rate": 6.316744443234489e-07, "loss": 1.204, "step": 8136 }, { "epoch": 0.63, "learning_rate": 6.314407415621989e-07, "loss": 1.1884, "step": 8137 }, { "epoch": 0.63, "learning_rate": 6.312070620923451e-07, "loss": 0.8149, "step": 8138 }, { "epoch": 0.63, "learning_rate": 6.309734059286555e-07, "loss": 1.1928, "step": 8139 }, { "epoch": 0.63, "learning_rate": 6.307397730858965e-07, "loss": 1.231, "step": 8140 }, { "epoch": 0.63, "learning_rate": 6.305061635788316e-07, "loss": 1.199, "step": 8141 }, { "epoch": 0.63, "learning_rate": 6.30272577422225e-07, "loss": 0.8599, "step": 8142 }, { "epoch": 0.63, "learning_rate": 6.300390146308377e-07, "loss": 1.1741, "step": 8143 }, { "epoch": 0.63, "learning_rate": 6.298054752194304e-07, "loss": 1.1551, "step": 8144 }, { "epoch": 0.63, "learning_rate": 6.295719592027617e-07, "loss": 1.1872, "step": 8145 }, { "epoch": 0.63, "learning_rate": 6.293384665955883e-07, "loss": 1.2391, "step": 8146 }, { "epoch": 0.63, "learning_rate": 6.291049974126667e-07, "loss": 1.1659, "step": 8147 }, { "epoch": 0.63, "learning_rate": 6.288715516687504e-07, "loss": 1.1738, "step": 8148 }, { "epoch": 0.63, "learning_rate": 6.286381293785932e-07, "loss": 1.257, "step": 8149 }, { "epoch": 0.63, "learning_rate": 6.284047305569455e-07, "loss": 1.1984, "step": 8150 }, { "epoch": 0.63, "learning_rate": 6.281713552185575e-07, "loss": 1.188, "step": 8151 }, { "epoch": 0.63, "learning_rate": 6.279380033781776e-07, "loss": 1.1728, "step": 8152 }, { "epoch": 0.63, "learning_rate": 6.277046750505522e-07, "loss": 1.2099, "step": 8153 }, { "epoch": 0.63, "learning_rate": 6.274713702504276e-07, "loss": 1.216, "step": 8154 }, { "epoch": 0.63, "learning_rate": 6.272380889925464e-07, "loss": 1.1687, "step": 8155 }, { "epoch": 0.63, "learning_rate": 6.27004831291652e-07, "loss": 1.205, "step": 8156 }, { "epoch": 0.63, "learning_rate": 6.267715971624854e-07, "loss": 1.1458, "step": 8157 }, { "epoch": 0.63, "learning_rate": 6.265383866197851e-07, "loss": 1.1721, "step": 8158 }, { "epoch": 0.63, "learning_rate": 6.263051996782899e-07, "loss": 1.2153, "step": 8159 }, { "epoch": 0.63, "learning_rate": 6.260720363527355e-07, "loss": 1.1899, "step": 8160 }, { "epoch": 0.63, "learning_rate": 6.258388966578574e-07, "loss": 1.1825, "step": 8161 }, { "epoch": 0.63, "learning_rate": 6.256057806083891e-07, "loss": 0.8154, "step": 8162 }, { "epoch": 0.63, "learning_rate": 6.25372688219062e-07, "loss": 1.1905, "step": 8163 }, { "epoch": 0.63, "learning_rate": 6.251396195046071e-07, "loss": 0.8158, "step": 8164 }, { "epoch": 0.63, "learning_rate": 6.249065744797528e-07, "loss": 1.1738, "step": 8165 }, { "epoch": 0.63, "learning_rate": 6.246735531592273e-07, "loss": 1.2135, "step": 8166 }, { "epoch": 0.63, "learning_rate": 6.244405555577561e-07, "loss": 1.2085, "step": 8167 }, { "epoch": 0.63, "learning_rate": 6.242075816900637e-07, "loss": 1.1961, "step": 8168 }, { "epoch": 0.63, "learning_rate": 6.239746315708732e-07, "loss": 1.2381, "step": 8169 }, { "epoch": 0.63, "learning_rate": 6.237417052149056e-07, "loss": 1.1983, "step": 8170 }, { "epoch": 0.63, "learning_rate": 6.23508802636882e-07, "loss": 1.201, "step": 8171 }, { "epoch": 0.63, "learning_rate": 6.232759238515197e-07, "loss": 1.1887, "step": 8172 }, { "epoch": 0.63, "learning_rate": 6.230430688735358e-07, "loss": 1.207, "step": 8173 }, { "epoch": 0.63, "learning_rate": 6.228102377176466e-07, "loss": 1.1652, "step": 8174 }, { "epoch": 0.63, "learning_rate": 6.225774303985652e-07, "loss": 1.1591, "step": 8175 }, { "epoch": 0.63, "learning_rate": 6.223446469310045e-07, "loss": 0.8242, "step": 8176 }, { "epoch": 0.63, "learning_rate": 6.221118873296749e-07, "loss": 1.2038, "step": 8177 }, { "epoch": 0.63, "learning_rate": 6.218791516092865e-07, "loss": 1.1976, "step": 8178 }, { "epoch": 0.63, "learning_rate": 6.216464397845471e-07, "loss": 1.2175, "step": 8179 }, { "epoch": 0.63, "learning_rate": 6.214137518701625e-07, "loss": 0.796, "step": 8180 }, { "epoch": 0.63, "learning_rate": 6.211810878808381e-07, "loss": 0.8336, "step": 8181 }, { "epoch": 0.64, "learning_rate": 6.209484478312774e-07, "loss": 1.1864, "step": 8182 }, { "epoch": 0.64, "learning_rate": 6.20715831736182e-07, "loss": 0.8396, "step": 8183 }, { "epoch": 0.64, "learning_rate": 6.204832396102525e-07, "loss": 1.2382, "step": 8184 }, { "epoch": 0.64, "learning_rate": 6.202506714681874e-07, "loss": 0.8478, "step": 8185 }, { "epoch": 0.64, "learning_rate": 6.200181273246843e-07, "loss": 1.1923, "step": 8186 }, { "epoch": 0.64, "learning_rate": 6.197856071944388e-07, "loss": 1.2351, "step": 8187 }, { "epoch": 0.64, "learning_rate": 6.195531110921457e-07, "loss": 1.2147, "step": 8188 }, { "epoch": 0.64, "learning_rate": 6.193206390324972e-07, "loss": 1.2292, "step": 8189 }, { "epoch": 0.64, "learning_rate": 6.190881910301846e-07, "loss": 1.2046, "step": 8190 }, { "epoch": 0.64, "learning_rate": 6.188557670998982e-07, "loss": 1.2168, "step": 8191 }, { "epoch": 0.64, "learning_rate": 6.186233672563252e-07, "loss": 1.1634, "step": 8192 }, { "epoch": 0.64, "learning_rate": 6.183909915141537e-07, "loss": 1.1366, "step": 8193 }, { "epoch": 0.64, "learning_rate": 6.181586398880676e-07, "loss": 1.1982, "step": 8194 }, { "epoch": 0.64, "learning_rate": 6.17926312392751e-07, "loss": 1.2031, "step": 8195 }, { "epoch": 0.64, "learning_rate": 6.176940090428862e-07, "loss": 1.2028, "step": 8196 }, { "epoch": 0.64, "learning_rate": 6.174617298531535e-07, "loss": 1.2118, "step": 8197 }, { "epoch": 0.64, "learning_rate": 6.172294748382321e-07, "loss": 1.2081, "step": 8198 }, { "epoch": 0.64, "learning_rate": 6.169972440127998e-07, "loss": 1.1953, "step": 8199 }, { "epoch": 0.64, "learning_rate": 6.167650373915322e-07, "loss": 1.1432, "step": 8200 }, { "epoch": 0.64, "learning_rate": 6.16532854989104e-07, "loss": 1.1927, "step": 8201 }, { "epoch": 0.64, "learning_rate": 6.163006968201881e-07, "loss": 1.1627, "step": 8202 }, { "epoch": 0.64, "learning_rate": 6.160685628994558e-07, "loss": 1.164, "step": 8203 }, { "epoch": 0.64, "learning_rate": 6.158364532415772e-07, "loss": 1.2062, "step": 8204 }, { "epoch": 0.64, "learning_rate": 6.156043678612207e-07, "loss": 1.2271, "step": 8205 }, { "epoch": 0.64, "learning_rate": 6.15372306773053e-07, "loss": 1.2259, "step": 8206 }, { "epoch": 0.64, "learning_rate": 6.15140269991739e-07, "loss": 0.8519, "step": 8207 }, { "epoch": 0.64, "learning_rate": 6.149082575319433e-07, "loss": 1.2061, "step": 8208 }, { "epoch": 0.64, "learning_rate": 6.146762694083272e-07, "loss": 1.173, "step": 8209 }, { "epoch": 0.64, "learning_rate": 6.144443056355522e-07, "loss": 1.1626, "step": 8210 }, { "epoch": 0.64, "learning_rate": 6.14212366228277e-07, "loss": 1.2308, "step": 8211 }, { "epoch": 0.64, "learning_rate": 6.139804512011588e-07, "loss": 1.1848, "step": 8212 }, { "epoch": 0.64, "learning_rate": 6.137485605688546e-07, "loss": 1.2004, "step": 8213 }, { "epoch": 0.64, "learning_rate": 6.135166943460182e-07, "loss": 1.159, "step": 8214 }, { "epoch": 0.64, "learning_rate": 6.132848525473026e-07, "loss": 1.1745, "step": 8215 }, { "epoch": 0.64, "learning_rate": 6.1305303518736e-07, "loss": 1.2012, "step": 8216 }, { "epoch": 0.64, "learning_rate": 6.128212422808391e-07, "loss": 1.1447, "step": 8217 }, { "epoch": 0.64, "learning_rate": 6.125894738423892e-07, "loss": 1.1748, "step": 8218 }, { "epoch": 0.64, "learning_rate": 6.123577298866564e-07, "loss": 0.8353, "step": 8219 }, { "epoch": 0.64, "learning_rate": 6.121260104282865e-07, "loss": 1.1912, "step": 8220 }, { "epoch": 0.64, "learning_rate": 6.118943154819227e-07, "loss": 1.203, "step": 8221 }, { "epoch": 0.64, "learning_rate": 6.116626450622078e-07, "loss": 1.1893, "step": 8222 }, { "epoch": 0.64, "learning_rate": 6.114309991837818e-07, "loss": 1.2011, "step": 8223 }, { "epoch": 0.64, "learning_rate": 6.111993778612839e-07, "loss": 1.2381, "step": 8224 }, { "epoch": 0.64, "learning_rate": 6.109677811093519e-07, "loss": 1.1485, "step": 8225 }, { "epoch": 0.64, "learning_rate": 6.107362089426211e-07, "loss": 0.8698, "step": 8226 }, { "epoch": 0.64, "learning_rate": 6.105046613757267e-07, "loss": 1.1829, "step": 8227 }, { "epoch": 0.64, "learning_rate": 6.102731384233009e-07, "loss": 1.2317, "step": 8228 }, { "epoch": 0.64, "learning_rate": 6.100416400999749e-07, "loss": 0.8257, "step": 8229 }, { "epoch": 0.64, "learning_rate": 6.098101664203793e-07, "loss": 1.205, "step": 8230 }, { "epoch": 0.64, "learning_rate": 6.095787173991411e-07, "loss": 1.1848, "step": 8231 }, { "epoch": 0.64, "learning_rate": 6.093472930508878e-07, "loss": 1.2116, "step": 8232 }, { "epoch": 0.64, "learning_rate": 6.091158933902442e-07, "loss": 1.2735, "step": 8233 }, { "epoch": 0.64, "learning_rate": 6.088845184318335e-07, "loss": 1.22, "step": 8234 }, { "epoch": 0.64, "learning_rate": 6.086531681902782e-07, "loss": 0.8759, "step": 8235 }, { "epoch": 0.64, "learning_rate": 6.084218426801977e-07, "loss": 1.1253, "step": 8236 }, { "epoch": 0.64, "learning_rate": 6.081905419162118e-07, "loss": 1.2644, "step": 8237 }, { "epoch": 0.64, "learning_rate": 6.079592659129376e-07, "loss": 0.8218, "step": 8238 }, { "epoch": 0.64, "learning_rate": 6.077280146849899e-07, "loss": 1.2341, "step": 8239 }, { "epoch": 0.64, "learning_rate": 6.074967882469837e-07, "loss": 1.2107, "step": 8240 }, { "epoch": 0.64, "learning_rate": 6.07265586613531e-07, "loss": 1.2174, "step": 8241 }, { "epoch": 0.64, "learning_rate": 6.070344097992435e-07, "loss": 1.1989, "step": 8242 }, { "epoch": 0.64, "learning_rate": 6.068032578187297e-07, "loss": 1.1783, "step": 8243 }, { "epoch": 0.64, "learning_rate": 6.065721306865983e-07, "loss": 1.2235, "step": 8244 }, { "epoch": 0.64, "learning_rate": 6.063410284174548e-07, "loss": 1.2766, "step": 8245 }, { "epoch": 0.64, "learning_rate": 6.061099510259041e-07, "loss": 1.1967, "step": 8246 }, { "epoch": 0.64, "learning_rate": 6.0587889852655e-07, "loss": 1.1984, "step": 8247 }, { "epoch": 0.64, "learning_rate": 6.056478709339928e-07, "loss": 1.1534, "step": 8248 }, { "epoch": 0.64, "learning_rate": 6.054168682628335e-07, "loss": 0.8464, "step": 8249 }, { "epoch": 0.64, "learning_rate": 6.051858905276703e-07, "loss": 1.1461, "step": 8250 }, { "epoch": 0.64, "learning_rate": 6.049549377430996e-07, "loss": 0.8177, "step": 8251 }, { "epoch": 0.64, "learning_rate": 6.047240099237173e-07, "loss": 1.151, "step": 8252 }, { "epoch": 0.64, "learning_rate": 6.044931070841162e-07, "loss": 1.203, "step": 8253 }, { "epoch": 0.64, "learning_rate": 6.042622292388891e-07, "loss": 1.2111, "step": 8254 }, { "epoch": 0.64, "learning_rate": 6.040313764026266e-07, "loss": 1.2053, "step": 8255 }, { "epoch": 0.64, "learning_rate": 6.03800548589917e-07, "loss": 1.2224, "step": 8256 }, { "epoch": 0.64, "learning_rate": 6.03569745815348e-07, "loss": 1.2008, "step": 8257 }, { "epoch": 0.64, "learning_rate": 6.033389680935053e-07, "loss": 1.1711, "step": 8258 }, { "epoch": 0.64, "learning_rate": 6.031082154389731e-07, "loss": 1.2457, "step": 8259 }, { "epoch": 0.64, "learning_rate": 6.028774878663341e-07, "loss": 1.1987, "step": 8260 }, { "epoch": 0.64, "learning_rate": 6.026467853901692e-07, "loss": 1.2304, "step": 8261 }, { "epoch": 0.64, "learning_rate": 6.024161080250578e-07, "loss": 1.175, "step": 8262 }, { "epoch": 0.64, "learning_rate": 6.021854557855776e-07, "loss": 1.1902, "step": 8263 }, { "epoch": 0.64, "learning_rate": 6.019548286863056e-07, "loss": 1.2034, "step": 8264 }, { "epoch": 0.64, "learning_rate": 6.017242267418155e-07, "loss": 1.1561, "step": 8265 }, { "epoch": 0.64, "learning_rate": 6.014936499666809e-07, "loss": 1.187, "step": 8266 }, { "epoch": 0.64, "learning_rate": 6.012630983754734e-07, "loss": 1.1908, "step": 8267 }, { "epoch": 0.64, "learning_rate": 6.010325719827623e-07, "loss": 1.1903, "step": 8268 }, { "epoch": 0.64, "learning_rate": 6.008020708031165e-07, "loss": 1.2349, "step": 8269 }, { "epoch": 0.64, "learning_rate": 6.005715948511021e-07, "loss": 1.2044, "step": 8270 }, { "epoch": 0.64, "learning_rate": 6.003411441412849e-07, "loss": 1.1819, "step": 8271 }, { "epoch": 0.64, "learning_rate": 6.001107186882282e-07, "loss": 0.8303, "step": 8272 }, { "epoch": 0.64, "learning_rate": 5.998803185064936e-07, "loss": 1.2442, "step": 8273 }, { "epoch": 0.64, "learning_rate": 5.996499436106416e-07, "loss": 1.2367, "step": 8274 }, { "epoch": 0.64, "learning_rate": 5.994195940152311e-07, "loss": 0.8511, "step": 8275 }, { "epoch": 0.64, "learning_rate": 5.99189269734819e-07, "loss": 1.1978, "step": 8276 }, { "epoch": 0.64, "learning_rate": 5.989589707839612e-07, "loss": 1.2229, "step": 8277 }, { "epoch": 0.64, "learning_rate": 5.987286971772108e-07, "loss": 0.8397, "step": 8278 }, { "epoch": 0.64, "learning_rate": 5.984984489291211e-07, "loss": 1.2184, "step": 8279 }, { "epoch": 0.64, "learning_rate": 5.98268226054242e-07, "loss": 1.1935, "step": 8280 }, { "epoch": 0.64, "learning_rate": 5.980380285671235e-07, "loss": 1.1932, "step": 8281 }, { "epoch": 0.64, "learning_rate": 5.978078564823123e-07, "loss": 1.161, "step": 8282 }, { "epoch": 0.64, "learning_rate": 5.975777098143547e-07, "loss": 1.1584, "step": 8283 }, { "epoch": 0.64, "learning_rate": 5.97347588577795e-07, "loss": 1.193, "step": 8284 }, { "epoch": 0.64, "learning_rate": 5.971174927871756e-07, "loss": 1.191, "step": 8285 }, { "epoch": 0.64, "learning_rate": 5.968874224570383e-07, "loss": 1.2265, "step": 8286 }, { "epoch": 0.64, "learning_rate": 5.966573776019214e-07, "loss": 1.2657, "step": 8287 }, { "epoch": 0.64, "learning_rate": 5.964273582363641e-07, "loss": 0.8295, "step": 8288 }, { "epoch": 0.64, "learning_rate": 5.961973643749018e-07, "loss": 1.2398, "step": 8289 }, { "epoch": 0.64, "learning_rate": 5.959673960320693e-07, "loss": 0.862, "step": 8290 }, { "epoch": 0.64, "learning_rate": 5.957374532223996e-07, "loss": 1.1905, "step": 8291 }, { "epoch": 0.64, "learning_rate": 5.955075359604245e-07, "loss": 1.2291, "step": 8292 }, { "epoch": 0.64, "learning_rate": 5.952776442606733e-07, "loss": 1.1295, "step": 8293 }, { "epoch": 0.64, "learning_rate": 5.950477781376746e-07, "loss": 1.2185, "step": 8294 }, { "epoch": 0.64, "learning_rate": 5.948179376059545e-07, "loss": 1.1789, "step": 8295 }, { "epoch": 0.64, "learning_rate": 5.945881226800385e-07, "loss": 1.1837, "step": 8296 }, { "epoch": 0.64, "learning_rate": 5.943583333744493e-07, "loss": 0.8517, "step": 8297 }, { "epoch": 0.64, "learning_rate": 5.941285697037095e-07, "loss": 1.2147, "step": 8298 }, { "epoch": 0.64, "learning_rate": 5.938988316823385e-07, "loss": 1.1926, "step": 8299 }, { "epoch": 0.64, "learning_rate": 5.936691193248547e-07, "loss": 1.1986, "step": 8300 }, { "epoch": 0.64, "learning_rate": 5.934394326457757e-07, "loss": 1.1872, "step": 8301 }, { "epoch": 0.64, "learning_rate": 5.932097716596159e-07, "loss": 1.1508, "step": 8302 }, { "epoch": 0.64, "learning_rate": 5.929801363808897e-07, "loss": 0.7892, "step": 8303 }, { "epoch": 0.64, "learning_rate": 5.927505268241082e-07, "loss": 1.2072, "step": 8304 }, { "epoch": 0.64, "learning_rate": 5.925209430037823e-07, "loss": 1.2174, "step": 8305 }, { "epoch": 0.64, "learning_rate": 5.922913849344209e-07, "loss": 1.2243, "step": 8306 }, { "epoch": 0.64, "learning_rate": 5.920618526305304e-07, "loss": 1.1713, "step": 8307 }, { "epoch": 0.64, "learning_rate": 5.91832346106617e-07, "loss": 0.855, "step": 8308 }, { "epoch": 0.64, "learning_rate": 5.916028653771843e-07, "loss": 0.8285, "step": 8309 }, { "epoch": 0.64, "learning_rate": 5.913734104567345e-07, "loss": 1.2299, "step": 8310 }, { "epoch": 0.65, "learning_rate": 5.911439813597684e-07, "loss": 1.2101, "step": 8311 }, { "epoch": 0.65, "learning_rate": 5.909145781007843e-07, "loss": 0.8124, "step": 8312 }, { "epoch": 0.65, "learning_rate": 5.906852006942802e-07, "loss": 1.2243, "step": 8313 }, { "epoch": 0.65, "learning_rate": 5.904558491547514e-07, "loss": 1.2172, "step": 8314 }, { "epoch": 0.65, "learning_rate": 5.902265234966923e-07, "loss": 0.771, "step": 8315 }, { "epoch": 0.65, "learning_rate": 5.89997223734595e-07, "loss": 1.157, "step": 8316 }, { "epoch": 0.65, "learning_rate": 5.897679498829505e-07, "loss": 1.1979, "step": 8317 }, { "epoch": 0.65, "learning_rate": 5.895387019562478e-07, "loss": 1.1678, "step": 8318 }, { "epoch": 0.65, "learning_rate": 5.893094799689741e-07, "loss": 1.2199, "step": 8319 }, { "epoch": 0.65, "learning_rate": 5.890802839356162e-07, "loss": 1.2223, "step": 8320 }, { "epoch": 0.65, "learning_rate": 5.888511138706575e-07, "loss": 1.2295, "step": 8321 }, { "epoch": 0.65, "learning_rate": 5.886219697885807e-07, "loss": 1.2489, "step": 8322 }, { "epoch": 0.65, "learning_rate": 5.883928517038672e-07, "loss": 1.1896, "step": 8323 }, { "epoch": 0.65, "learning_rate": 5.881637596309955e-07, "loss": 0.7805, "step": 8324 }, { "epoch": 0.65, "learning_rate": 5.87934693584444e-07, "loss": 1.1029, "step": 8325 }, { "epoch": 0.65, "learning_rate": 5.877056535786886e-07, "loss": 1.1612, "step": 8326 }, { "epoch": 0.65, "learning_rate": 5.874766396282032e-07, "loss": 1.2299, "step": 8327 }, { "epoch": 0.65, "learning_rate": 5.872476517474612e-07, "loss": 1.1903, "step": 8328 }, { "epoch": 0.65, "learning_rate": 5.870186899509328e-07, "loss": 1.1544, "step": 8329 }, { "epoch": 0.65, "learning_rate": 5.867897542530882e-07, "loss": 1.185, "step": 8330 }, { "epoch": 0.65, "learning_rate": 5.865608446683947e-07, "loss": 1.152, "step": 8331 }, { "epoch": 0.65, "learning_rate": 5.863319612113191e-07, "loss": 1.247, "step": 8332 }, { "epoch": 0.65, "learning_rate": 5.86103103896325e-07, "loss": 0.8282, "step": 8333 }, { "epoch": 0.65, "learning_rate": 5.858742727378757e-07, "loss": 1.2096, "step": 8334 }, { "epoch": 0.65, "learning_rate": 5.856454677504324e-07, "loss": 1.2449, "step": 8335 }, { "epoch": 0.65, "learning_rate": 5.854166889484539e-07, "loss": 1.2555, "step": 8336 }, { "epoch": 0.65, "learning_rate": 5.851879363463994e-07, "loss": 1.2126, "step": 8337 }, { "epoch": 0.65, "learning_rate": 5.849592099587241e-07, "loss": 1.2351, "step": 8338 }, { "epoch": 0.65, "learning_rate": 5.847305097998827e-07, "loss": 1.2044, "step": 8339 }, { "epoch": 0.65, "learning_rate": 5.845018358843284e-07, "loss": 1.2015, "step": 8340 }, { "epoch": 0.65, "learning_rate": 5.84273188226512e-07, "loss": 1.2172, "step": 8341 }, { "epoch": 0.65, "learning_rate": 5.840445668408833e-07, "loss": 1.1741, "step": 8342 }, { "epoch": 0.65, "learning_rate": 5.838159717418904e-07, "loss": 0.7931, "step": 8343 }, { "epoch": 0.65, "learning_rate": 5.835874029439791e-07, "loss": 1.1861, "step": 8344 }, { "epoch": 0.65, "learning_rate": 5.833588604615945e-07, "loss": 1.2346, "step": 8345 }, { "epoch": 0.65, "learning_rate": 5.83130344309179e-07, "loss": 1.1548, "step": 8346 }, { "epoch": 0.65, "learning_rate": 5.829018545011745e-07, "loss": 1.1335, "step": 8347 }, { "epoch": 0.65, "learning_rate": 5.8267339105202e-07, "loss": 1.1457, "step": 8348 }, { "epoch": 0.65, "learning_rate": 5.824449539761535e-07, "loss": 1.1806, "step": 8349 }, { "epoch": 0.65, "learning_rate": 5.822165432880117e-07, "loss": 1.225, "step": 8350 }, { "epoch": 0.65, "learning_rate": 5.819881590020288e-07, "loss": 1.1736, "step": 8351 }, { "epoch": 0.65, "learning_rate": 5.817598011326379e-07, "loss": 1.2593, "step": 8352 }, { "epoch": 0.65, "learning_rate": 5.815314696942701e-07, "loss": 1.2131, "step": 8353 }, { "epoch": 0.65, "learning_rate": 5.813031647013551e-07, "loss": 1.1874, "step": 8354 }, { "epoch": 0.65, "learning_rate": 5.810748861683208e-07, "loss": 1.2162, "step": 8355 }, { "epoch": 0.65, "learning_rate": 5.808466341095934e-07, "loss": 1.2296, "step": 8356 }, { "epoch": 0.65, "learning_rate": 5.806184085395976e-07, "loss": 1.1911, "step": 8357 }, { "epoch": 0.65, "learning_rate": 5.803902094727558e-07, "loss": 1.2603, "step": 8358 }, { "epoch": 0.65, "learning_rate": 5.801620369234905e-07, "loss": 1.1781, "step": 8359 }, { "epoch": 0.65, "learning_rate": 5.7993389090622e-07, "loss": 1.1536, "step": 8360 }, { "epoch": 0.65, "learning_rate": 5.797057714353623e-07, "loss": 1.2403, "step": 8361 }, { "epoch": 0.65, "learning_rate": 5.794776785253341e-07, "loss": 1.1707, "step": 8362 }, { "epoch": 0.65, "learning_rate": 5.792496121905491e-07, "loss": 1.2543, "step": 8363 }, { "epoch": 0.65, "learning_rate": 5.790215724454216e-07, "loss": 1.2128, "step": 8364 }, { "epoch": 0.65, "learning_rate": 5.787935593043614e-07, "loss": 0.8475, "step": 8365 }, { "epoch": 0.65, "learning_rate": 5.78565572781778e-07, "loss": 1.2193, "step": 8366 }, { "epoch": 0.65, "learning_rate": 5.783376128920805e-07, "loss": 1.2112, "step": 8367 }, { "epoch": 0.65, "learning_rate": 5.781096796496732e-07, "loss": 0.8362, "step": 8368 }, { "epoch": 0.65, "learning_rate": 5.778817730689622e-07, "loss": 1.1817, "step": 8369 }, { "epoch": 0.65, "learning_rate": 5.776538931643492e-07, "loss": 1.1487, "step": 8370 }, { "epoch": 0.65, "learning_rate": 5.774260399502351e-07, "loss": 1.195, "step": 8371 }, { "epoch": 0.65, "learning_rate": 5.771982134410204e-07, "loss": 1.1372, "step": 8372 }, { "epoch": 0.65, "learning_rate": 5.769704136511013e-07, "loss": 1.1438, "step": 8373 }, { "epoch": 0.65, "learning_rate": 5.767426405948751e-07, "loss": 1.2374, "step": 8374 }, { "epoch": 0.65, "learning_rate": 5.765148942867355e-07, "loss": 0.8274, "step": 8375 }, { "epoch": 0.65, "learning_rate": 5.762871747410751e-07, "loss": 1.1505, "step": 8376 }, { "epoch": 0.65, "learning_rate": 5.760594819722849e-07, "loss": 1.237, "step": 8377 }, { "epoch": 0.65, "learning_rate": 5.758318159947542e-07, "loss": 1.1734, "step": 8378 }, { "epoch": 0.65, "learning_rate": 5.756041768228703e-07, "loss": 1.1944, "step": 8379 }, { "epoch": 0.65, "learning_rate": 5.753765644710194e-07, "loss": 0.847, "step": 8380 }, { "epoch": 0.65, "learning_rate": 5.751489789535853e-07, "loss": 1.1713, "step": 8381 }, { "epoch": 0.65, "learning_rate": 5.749214202849507e-07, "loss": 1.1812, "step": 8382 }, { "epoch": 0.65, "learning_rate": 5.746938884794961e-07, "loss": 1.1888, "step": 8383 }, { "epoch": 0.65, "learning_rate": 5.744663835516006e-07, "loss": 1.1816, "step": 8384 }, { "epoch": 0.65, "learning_rate": 5.742389055156418e-07, "loss": 1.1871, "step": 8385 }, { "epoch": 0.65, "learning_rate": 5.740114543859952e-07, "loss": 1.2411, "step": 8386 }, { "epoch": 0.65, "learning_rate": 5.737840301770347e-07, "loss": 1.1464, "step": 8387 }, { "epoch": 0.65, "learning_rate": 5.735566329031327e-07, "loss": 0.8255, "step": 8388 }, { "epoch": 0.65, "learning_rate": 5.733292625786596e-07, "loss": 1.2014, "step": 8389 }, { "epoch": 0.65, "learning_rate": 5.731019192179842e-07, "loss": 1.2015, "step": 8390 }, { "epoch": 0.65, "learning_rate": 5.728746028354738e-07, "loss": 1.2203, "step": 8391 }, { "epoch": 0.65, "learning_rate": 5.726473134454939e-07, "loss": 0.8354, "step": 8392 }, { "epoch": 0.65, "learning_rate": 5.724200510624082e-07, "loss": 1.2095, "step": 8393 }, { "epoch": 0.65, "learning_rate": 5.721928157005784e-07, "loss": 1.2185, "step": 8394 }, { "epoch": 0.65, "learning_rate": 5.719656073743653e-07, "loss": 1.2148, "step": 8395 }, { "epoch": 0.65, "learning_rate": 5.717384260981272e-07, "loss": 0.8057, "step": 8396 }, { "epoch": 0.65, "learning_rate": 5.715112718862207e-07, "loss": 1.2224, "step": 8397 }, { "epoch": 0.65, "learning_rate": 5.712841447530021e-07, "loss": 1.1989, "step": 8398 }, { "epoch": 0.65, "learning_rate": 5.710570447128239e-07, "loss": 1.2237, "step": 8399 }, { "epoch": 0.65, "learning_rate": 5.708299717800377e-07, "loss": 1.1595, "step": 8400 }, { "epoch": 0.65, "learning_rate": 5.706029259689947e-07, "loss": 1.1923, "step": 8401 }, { "epoch": 0.65, "learning_rate": 5.703759072940418e-07, "loss": 1.1814, "step": 8402 }, { "epoch": 0.65, "learning_rate": 5.701489157695271e-07, "loss": 1.2051, "step": 8403 }, { "epoch": 0.65, "learning_rate": 5.699219514097945e-07, "loss": 1.1711, "step": 8404 }, { "epoch": 0.65, "learning_rate": 5.696950142291869e-07, "loss": 1.1527, "step": 8405 }, { "epoch": 0.65, "learning_rate": 5.694681042420473e-07, "loss": 1.2006, "step": 8406 }, { "epoch": 0.65, "learning_rate": 5.692412214627136e-07, "loss": 0.8535, "step": 8407 }, { "epoch": 0.65, "learning_rate": 5.690143659055251e-07, "loss": 1.1704, "step": 8408 }, { "epoch": 0.65, "learning_rate": 5.687875375848181e-07, "loss": 1.2289, "step": 8409 }, { "epoch": 0.65, "learning_rate": 5.685607365149262e-07, "loss": 1.1918, "step": 8410 }, { "epoch": 0.65, "learning_rate": 5.683339627101837e-07, "loss": 1.246, "step": 8411 }, { "epoch": 0.65, "learning_rate": 5.6810721618492e-07, "loss": 1.2181, "step": 8412 }, { "epoch": 0.65, "learning_rate": 5.678804969534661e-07, "loss": 1.1459, "step": 8413 }, { "epoch": 0.65, "learning_rate": 5.676538050301494e-07, "loss": 1.2313, "step": 8414 }, { "epoch": 0.65, "learning_rate": 5.674271404292949e-07, "loss": 1.1923, "step": 8415 }, { "epoch": 0.65, "learning_rate": 5.672005031652278e-07, "loss": 1.2329, "step": 8416 }, { "epoch": 0.65, "learning_rate": 5.669738932522703e-07, "loss": 1.1812, "step": 8417 }, { "epoch": 0.65, "learning_rate": 5.667473107047434e-07, "loss": 0.8259, "step": 8418 }, { "epoch": 0.65, "learning_rate": 5.66520755536966e-07, "loss": 1.1675, "step": 8419 }, { "epoch": 0.65, "learning_rate": 5.662942277632555e-07, "loss": 1.206, "step": 8420 }, { "epoch": 0.65, "learning_rate": 5.660677273979272e-07, "loss": 1.2123, "step": 8421 }, { "epoch": 0.65, "learning_rate": 5.658412544552953e-07, "loss": 1.2438, "step": 8422 }, { "epoch": 0.65, "learning_rate": 5.656148089496719e-07, "loss": 0.8228, "step": 8423 }, { "epoch": 0.65, "learning_rate": 5.653883908953674e-07, "loss": 0.7955, "step": 8424 }, { "epoch": 0.65, "learning_rate": 5.651620003066905e-07, "loss": 1.1749, "step": 8425 }, { "epoch": 0.65, "learning_rate": 5.64935637197948e-07, "loss": 1.1915, "step": 8426 }, { "epoch": 0.65, "learning_rate": 5.647093015834452e-07, "loss": 1.223, "step": 8427 }, { "epoch": 0.65, "learning_rate": 5.644829934774855e-07, "loss": 1.2604, "step": 8428 }, { "epoch": 0.65, "learning_rate": 5.642567128943706e-07, "loss": 1.266, "step": 8429 }, { "epoch": 0.65, "learning_rate": 5.640304598484005e-07, "loss": 1.1932, "step": 8430 }, { "epoch": 0.65, "learning_rate": 5.638042343538735e-07, "loss": 0.813, "step": 8431 }, { "epoch": 0.65, "learning_rate": 5.635780364250862e-07, "loss": 1.2021, "step": 8432 }, { "epoch": 0.65, "learning_rate": 5.633518660763331e-07, "loss": 0.8585, "step": 8433 }, { "epoch": 0.65, "learning_rate": 5.63125723321907e-07, "loss": 1.2417, "step": 8434 }, { "epoch": 0.65, "learning_rate": 5.628996081761003e-07, "loss": 1.1745, "step": 8435 }, { "epoch": 0.65, "learning_rate": 5.626735206532014e-07, "loss": 1.1986, "step": 8436 }, { "epoch": 0.65, "learning_rate": 5.624474607674983e-07, "loss": 1.2072, "step": 8437 }, { "epoch": 0.65, "learning_rate": 5.622214285332772e-07, "loss": 1.2018, "step": 8438 }, { "epoch": 0.65, "learning_rate": 5.61995423964822e-07, "loss": 1.2082, "step": 8439 }, { "epoch": 0.66, "learning_rate": 5.617694470764163e-07, "loss": 0.8175, "step": 8440 }, { "epoch": 0.66, "learning_rate": 5.615434978823396e-07, "loss": 1.2468, "step": 8441 }, { "epoch": 0.66, "learning_rate": 5.613175763968717e-07, "loss": 1.2055, "step": 8442 }, { "epoch": 0.66, "learning_rate": 5.610916826342902e-07, "loss": 1.1727, "step": 8443 }, { "epoch": 0.66, "learning_rate": 5.608658166088693e-07, "loss": 1.1415, "step": 8444 }, { "epoch": 0.66, "learning_rate": 5.606399783348844e-07, "loss": 0.81, "step": 8445 }, { "epoch": 0.66, "learning_rate": 5.604141678266061e-07, "loss": 1.2149, "step": 8446 }, { "epoch": 0.66, "learning_rate": 5.601883850983055e-07, "loss": 1.1765, "step": 8447 }, { "epoch": 0.66, "learning_rate": 5.599626301642514e-07, "loss": 1.1334, "step": 8448 }, { "epoch": 0.66, "learning_rate": 5.597369030387096e-07, "loss": 1.2609, "step": 8449 }, { "epoch": 0.66, "learning_rate": 5.595112037359457e-07, "loss": 1.2105, "step": 8450 }, { "epoch": 0.66, "learning_rate": 5.59285532270223e-07, "loss": 1.1753, "step": 8451 }, { "epoch": 0.66, "learning_rate": 5.590598886558029e-07, "loss": 1.2345, "step": 8452 }, { "epoch": 0.66, "learning_rate": 5.58834272906945e-07, "loss": 0.8567, "step": 8453 }, { "epoch": 0.66, "learning_rate": 5.586086850379074e-07, "loss": 1.2388, "step": 8454 }, { "epoch": 0.66, "learning_rate": 5.583831250629463e-07, "loss": 1.2457, "step": 8455 }, { "epoch": 0.66, "learning_rate": 5.58157592996316e-07, "loss": 1.1891, "step": 8456 }, { "epoch": 0.66, "learning_rate": 5.579320888522691e-07, "loss": 1.2032, "step": 8457 }, { "epoch": 0.66, "learning_rate": 5.57706612645057e-07, "loss": 1.2078, "step": 8458 }, { "epoch": 0.66, "learning_rate": 5.574811643889285e-07, "loss": 1.2007, "step": 8459 }, { "epoch": 0.66, "learning_rate": 5.572557440981309e-07, "loss": 1.2479, "step": 8460 }, { "epoch": 0.66, "learning_rate": 5.570303517869101e-07, "loss": 1.192, "step": 8461 }, { "epoch": 0.66, "learning_rate": 5.568049874695096e-07, "loss": 1.1622, "step": 8462 }, { "epoch": 0.66, "learning_rate": 5.565796511601719e-07, "loss": 1.2369, "step": 8463 }, { "epoch": 0.66, "learning_rate": 5.563543428731369e-07, "loss": 1.2263, "step": 8464 }, { "epoch": 0.66, "learning_rate": 5.561290626226434e-07, "loss": 0.7941, "step": 8465 }, { "epoch": 0.66, "learning_rate": 5.559038104229279e-07, "loss": 1.2242, "step": 8466 }, { "epoch": 0.66, "learning_rate": 5.556785862882258e-07, "loss": 1.2122, "step": 8467 }, { "epoch": 0.66, "learning_rate": 5.5545339023277e-07, "loss": 1.1688, "step": 8468 }, { "epoch": 0.66, "learning_rate": 5.55228222270792e-07, "loss": 1.2457, "step": 8469 }, { "epoch": 0.66, "learning_rate": 5.550030824165215e-07, "loss": 1.2315, "step": 8470 }, { "epoch": 0.66, "learning_rate": 5.547779706841864e-07, "loss": 1.1848, "step": 8471 }, { "epoch": 0.66, "learning_rate": 5.545528870880127e-07, "loss": 1.2238, "step": 8472 }, { "epoch": 0.66, "learning_rate": 5.543278316422246e-07, "loss": 1.2036, "step": 8473 }, { "epoch": 0.66, "learning_rate": 5.541028043610456e-07, "loss": 0.8166, "step": 8474 }, { "epoch": 0.66, "learning_rate": 5.538778052586956e-07, "loss": 1.1937, "step": 8475 }, { "epoch": 0.66, "learning_rate": 5.536528343493931e-07, "loss": 1.1256, "step": 8476 }, { "epoch": 0.66, "learning_rate": 5.534278916473569e-07, "loss": 1.1957, "step": 8477 }, { "epoch": 0.66, "learning_rate": 5.532029771668009e-07, "loss": 1.1998, "step": 8478 }, { "epoch": 0.66, "learning_rate": 5.5297809092194e-07, "loss": 0.8287, "step": 8479 }, { "epoch": 0.66, "learning_rate": 5.52753232926985e-07, "loss": 1.1897, "step": 8480 }, { "epoch": 0.66, "learning_rate": 5.525284031961462e-07, "loss": 1.181, "step": 8481 }, { "epoch": 0.66, "learning_rate": 5.523036017436329e-07, "loss": 0.8305, "step": 8482 }, { "epoch": 0.66, "learning_rate": 5.5207882858365e-07, "loss": 1.1876, "step": 8483 }, { "epoch": 0.66, "learning_rate": 5.518540837304033e-07, "loss": 1.2168, "step": 8484 }, { "epoch": 0.66, "learning_rate": 5.516293671980957e-07, "loss": 1.2131, "step": 8485 }, { "epoch": 0.66, "learning_rate": 5.514046790009279e-07, "loss": 0.823, "step": 8486 }, { "epoch": 0.66, "learning_rate": 5.511800191531e-07, "loss": 1.1818, "step": 8487 }, { "epoch": 0.66, "learning_rate": 5.509553876688083e-07, "loss": 1.1633, "step": 8488 }, { "epoch": 0.66, "learning_rate": 5.507307845622495e-07, "loss": 1.2244, "step": 8489 }, { "epoch": 0.66, "learning_rate": 5.505062098476174e-07, "loss": 0.8357, "step": 8490 }, { "epoch": 0.66, "learning_rate": 5.502816635391043e-07, "loss": 1.2012, "step": 8491 }, { "epoch": 0.66, "learning_rate": 5.500571456509003e-07, "loss": 1.1686, "step": 8492 }, { "epoch": 0.66, "learning_rate": 5.498326561971942e-07, "loss": 1.1353, "step": 8493 }, { "epoch": 0.66, "learning_rate": 5.496081951921726e-07, "loss": 1.1959, "step": 8494 }, { "epoch": 0.66, "learning_rate": 5.493837626500207e-07, "loss": 1.1844, "step": 8495 }, { "epoch": 0.66, "learning_rate": 5.491593585849215e-07, "loss": 1.1865, "step": 8496 }, { "epoch": 0.66, "learning_rate": 5.489349830110565e-07, "loss": 1.2226, "step": 8497 }, { "epoch": 0.66, "learning_rate": 5.487106359426053e-07, "loss": 1.1735, "step": 8498 }, { "epoch": 0.66, "learning_rate": 5.484863173937457e-07, "loss": 1.1741, "step": 8499 }, { "epoch": 0.66, "learning_rate": 5.482620273786537e-07, "loss": 1.1962, "step": 8500 }, { "epoch": 0.66, "learning_rate": 5.480377659115035e-07, "loss": 0.8432, "step": 8501 }, { "epoch": 0.66, "learning_rate": 5.478135330064671e-07, "loss": 1.2073, "step": 8502 }, { "epoch": 0.66, "learning_rate": 5.475893286777157e-07, "loss": 1.1989, "step": 8503 }, { "epoch": 0.66, "learning_rate": 5.473651529394176e-07, "loss": 1.2394, "step": 8504 }, { "epoch": 0.66, "learning_rate": 5.471410058057402e-07, "loss": 1.155, "step": 8505 }, { "epoch": 0.66, "learning_rate": 5.469168872908482e-07, "loss": 1.1987, "step": 8506 }, { "epoch": 0.66, "learning_rate": 5.466927974089049e-07, "loss": 1.1389, "step": 8507 }, { "epoch": 0.66, "learning_rate": 5.464687361740726e-07, "loss": 0.8456, "step": 8508 }, { "epoch": 0.66, "learning_rate": 5.462447036005104e-07, "loss": 1.1724, "step": 8509 }, { "epoch": 0.66, "learning_rate": 5.460206997023758e-07, "loss": 1.1922, "step": 8510 }, { "epoch": 0.66, "learning_rate": 5.457967244938261e-07, "loss": 0.7723, "step": 8511 }, { "epoch": 0.66, "learning_rate": 5.455727779890145e-07, "loss": 1.1968, "step": 8512 }, { "epoch": 0.66, "learning_rate": 5.453488602020944e-07, "loss": 0.8335, "step": 8513 }, { "epoch": 0.66, "learning_rate": 5.451249711472156e-07, "loss": 1.2172, "step": 8514 }, { "epoch": 0.66, "learning_rate": 5.449011108385271e-07, "loss": 1.1836, "step": 8515 }, { "epoch": 0.66, "learning_rate": 5.446772792901768e-07, "loss": 1.1516, "step": 8516 }, { "epoch": 0.66, "learning_rate": 5.444534765163086e-07, "loss": 1.183, "step": 8517 }, { "epoch": 0.66, "learning_rate": 5.44229702531067e-07, "loss": 1.2204, "step": 8518 }, { "epoch": 0.66, "learning_rate": 5.440059573485933e-07, "loss": 1.2051, "step": 8519 }, { "epoch": 0.66, "learning_rate": 5.437822409830264e-07, "loss": 1.2384, "step": 8520 }, { "epoch": 0.66, "learning_rate": 5.435585534485056e-07, "loss": 1.1955, "step": 8521 }, { "epoch": 0.66, "learning_rate": 5.433348947591657e-07, "loss": 0.8597, "step": 8522 }, { "epoch": 0.66, "learning_rate": 5.43111264929142e-07, "loss": 1.2521, "step": 8523 }, { "epoch": 0.66, "learning_rate": 5.42887663972567e-07, "loss": 1.245, "step": 8524 }, { "epoch": 0.66, "learning_rate": 5.4266409190357e-07, "loss": 1.2183, "step": 8525 }, { "epoch": 0.66, "learning_rate": 5.424405487362813e-07, "loss": 1.2134, "step": 8526 }, { "epoch": 0.66, "learning_rate": 5.422170344848274e-07, "loss": 1.2169, "step": 8527 }, { "epoch": 0.66, "learning_rate": 5.419935491633332e-07, "loss": 1.2259, "step": 8528 }, { "epoch": 0.66, "learning_rate": 5.417700927859224e-07, "loss": 1.1921, "step": 8529 }, { "epoch": 0.66, "learning_rate": 5.415466653667162e-07, "loss": 1.2461, "step": 8530 }, { "epoch": 0.66, "learning_rate": 5.413232669198344e-07, "loss": 1.1666, "step": 8531 }, { "epoch": 0.66, "learning_rate": 5.41099897459395e-07, "loss": 1.2295, "step": 8532 }, { "epoch": 0.66, "learning_rate": 5.408765569995138e-07, "loss": 1.1772, "step": 8533 }, { "epoch": 0.66, "learning_rate": 5.406532455543049e-07, "loss": 1.2136, "step": 8534 }, { "epoch": 0.66, "learning_rate": 5.404299631378809e-07, "loss": 1.1213, "step": 8535 }, { "epoch": 0.66, "learning_rate": 5.402067097643523e-07, "loss": 1.1832, "step": 8536 }, { "epoch": 0.66, "learning_rate": 5.399834854478274e-07, "loss": 0.8254, "step": 8537 }, { "epoch": 0.66, "learning_rate": 5.397602902024136e-07, "loss": 1.2067, "step": 8538 }, { "epoch": 0.66, "learning_rate": 5.395371240422156e-07, "loss": 1.1573, "step": 8539 }, { "epoch": 0.66, "learning_rate": 5.393139869813364e-07, "loss": 1.2258, "step": 8540 }, { "epoch": 0.66, "learning_rate": 5.390908790338775e-07, "loss": 1.2265, "step": 8541 }, { "epoch": 0.66, "learning_rate": 5.388678002139385e-07, "loss": 1.181, "step": 8542 }, { "epoch": 0.66, "learning_rate": 5.386447505356169e-07, "loss": 1.1511, "step": 8543 }, { "epoch": 0.66, "learning_rate": 5.384217300130085e-07, "loss": 1.1731, "step": 8544 }, { "epoch": 0.66, "learning_rate": 5.381987386602074e-07, "loss": 1.219, "step": 8545 }, { "epoch": 0.66, "learning_rate": 5.379757764913056e-07, "loss": 1.2106, "step": 8546 }, { "epoch": 0.66, "learning_rate": 5.377528435203932e-07, "loss": 1.2042, "step": 8547 }, { "epoch": 0.66, "learning_rate": 5.375299397615589e-07, "loss": 1.2321, "step": 8548 }, { "epoch": 0.66, "learning_rate": 5.373070652288889e-07, "loss": 0.8618, "step": 8549 }, { "epoch": 0.66, "learning_rate": 5.370842199364689e-07, "loss": 1.2136, "step": 8550 }, { "epoch": 0.66, "learning_rate": 5.368614038983803e-07, "loss": 1.2334, "step": 8551 }, { "epoch": 0.66, "learning_rate": 5.366386171287054e-07, "loss": 1.2062, "step": 8552 }, { "epoch": 0.66, "learning_rate": 5.364158596415234e-07, "loss": 1.2171, "step": 8553 }, { "epoch": 0.66, "learning_rate": 5.361931314509103e-07, "loss": 1.1554, "step": 8554 }, { "epoch": 0.66, "learning_rate": 5.359704325709432e-07, "loss": 1.2557, "step": 8555 }, { "epoch": 0.66, "learning_rate": 5.357477630156942e-07, "loss": 1.1014, "step": 8556 }, { "epoch": 0.66, "learning_rate": 5.355251227992363e-07, "loss": 1.2221, "step": 8557 }, { "epoch": 0.66, "learning_rate": 5.353025119356395e-07, "loss": 1.1788, "step": 8558 }, { "epoch": 0.66, "learning_rate": 5.350799304389705e-07, "loss": 0.8307, "step": 8559 }, { "epoch": 0.66, "learning_rate": 5.348573783232968e-07, "loss": 1.1569, "step": 8560 }, { "epoch": 0.66, "learning_rate": 5.346348556026823e-07, "loss": 1.1702, "step": 8561 }, { "epoch": 0.66, "learning_rate": 5.344123622911896e-07, "loss": 1.2073, "step": 8562 }, { "epoch": 0.66, "learning_rate": 5.341898984028797e-07, "loss": 1.185, "step": 8563 }, { "epoch": 0.66, "learning_rate": 5.3396746395181e-07, "loss": 1.2121, "step": 8564 }, { "epoch": 0.66, "learning_rate": 5.337450589520389e-07, "loss": 1.1909, "step": 8565 }, { "epoch": 0.66, "learning_rate": 5.33522683417621e-07, "loss": 1.1882, "step": 8566 }, { "epoch": 0.66, "learning_rate": 5.333003373626095e-07, "loss": 1.1649, "step": 8567 }, { "epoch": 0.66, "learning_rate": 5.330780208010555e-07, "loss": 1.2398, "step": 8568 }, { "epoch": 0.67, "learning_rate": 5.328557337470088e-07, "loss": 1.1663, "step": 8569 }, { "epoch": 0.67, "learning_rate": 5.326334762145168e-07, "loss": 1.1915, "step": 8570 }, { "epoch": 0.67, "learning_rate": 5.324112482176254e-07, "loss": 1.1859, "step": 8571 }, { "epoch": 0.67, "learning_rate": 5.321890497703783e-07, "loss": 1.215, "step": 8572 }, { "epoch": 0.67, "learning_rate": 5.319668808868176e-07, "loss": 1.2373, "step": 8573 }, { "epoch": 0.67, "learning_rate": 5.317447415809835e-07, "loss": 1.1583, "step": 8574 }, { "epoch": 0.67, "learning_rate": 5.315226318669141e-07, "loss": 1.197, "step": 8575 }, { "epoch": 0.67, "learning_rate": 5.313005517586461e-07, "loss": 1.2378, "step": 8576 }, { "epoch": 0.67, "learning_rate": 5.310785012702137e-07, "loss": 1.2325, "step": 8577 }, { "epoch": 0.67, "learning_rate": 5.308564804156497e-07, "loss": 1.201, "step": 8578 }, { "epoch": 0.67, "learning_rate": 5.306344892089849e-07, "loss": 1.1895, "step": 8579 }, { "epoch": 0.67, "learning_rate": 5.304125276642482e-07, "loss": 1.2486, "step": 8580 }, { "epoch": 0.67, "learning_rate": 5.301905957954666e-07, "loss": 1.1815, "step": 8581 }, { "epoch": 0.67, "learning_rate": 5.299686936166653e-07, "loss": 1.1912, "step": 8582 }, { "epoch": 0.67, "learning_rate": 5.297468211418672e-07, "loss": 1.1915, "step": 8583 }, { "epoch": 0.67, "learning_rate": 5.295249783850948e-07, "loss": 1.2165, "step": 8584 }, { "epoch": 0.67, "learning_rate": 5.293031653603666e-07, "loss": 0.8834, "step": 8585 }, { "epoch": 0.67, "learning_rate": 5.290813820817006e-07, "loss": 1.1421, "step": 8586 }, { "epoch": 0.67, "learning_rate": 5.288596285631124e-07, "loss": 1.1912, "step": 8587 }, { "epoch": 0.67, "learning_rate": 5.286379048186158e-07, "loss": 1.2297, "step": 8588 }, { "epoch": 0.67, "learning_rate": 5.284162108622237e-07, "loss": 1.1992, "step": 8589 }, { "epoch": 0.67, "learning_rate": 5.281945467079448e-07, "loss": 1.1992, "step": 8590 }, { "epoch": 0.67, "learning_rate": 5.279729123697885e-07, "loss": 1.1642, "step": 8591 }, { "epoch": 0.67, "learning_rate": 5.277513078617613e-07, "loss": 1.1737, "step": 8592 }, { "epoch": 0.67, "learning_rate": 5.27529733197866e-07, "loss": 1.1869, "step": 8593 }, { "epoch": 0.67, "learning_rate": 5.273081883921071e-07, "loss": 1.2042, "step": 8594 }, { "epoch": 0.67, "learning_rate": 5.270866734584843e-07, "loss": 1.2076, "step": 8595 }, { "epoch": 0.67, "learning_rate": 5.268651884109967e-07, "loss": 1.1735, "step": 8596 }, { "epoch": 0.67, "learning_rate": 5.266437332636415e-07, "loss": 1.1976, "step": 8597 }, { "epoch": 0.67, "learning_rate": 5.264223080304127e-07, "loss": 1.2083, "step": 8598 }, { "epoch": 0.67, "learning_rate": 5.262009127253045e-07, "loss": 0.831, "step": 8599 }, { "epoch": 0.67, "learning_rate": 5.259795473623079e-07, "loss": 1.1816, "step": 8600 }, { "epoch": 0.67, "learning_rate": 5.25758211955412e-07, "loss": 0.857, "step": 8601 }, { "epoch": 0.67, "learning_rate": 5.255369065186045e-07, "loss": 1.1686, "step": 8602 }, { "epoch": 0.67, "learning_rate": 5.253156310658708e-07, "loss": 0.8283, "step": 8603 }, { "epoch": 0.67, "learning_rate": 5.250943856111949e-07, "loss": 1.2253, "step": 8604 }, { "epoch": 0.67, "learning_rate": 5.248731701685581e-07, "loss": 1.188, "step": 8605 }, { "epoch": 0.67, "learning_rate": 5.246519847519407e-07, "loss": 1.2195, "step": 8606 }, { "epoch": 0.67, "learning_rate": 5.244308293753205e-07, "loss": 1.1456, "step": 8607 }, { "epoch": 0.67, "learning_rate": 5.242097040526736e-07, "loss": 1.1629, "step": 8608 }, { "epoch": 0.67, "learning_rate": 5.239886087979743e-07, "loss": 1.1822, "step": 8609 }, { "epoch": 0.67, "learning_rate": 5.237675436251947e-07, "loss": 1.2473, "step": 8610 }, { "epoch": 0.67, "learning_rate": 5.235465085483054e-07, "loss": 1.2662, "step": 8611 }, { "epoch": 0.67, "learning_rate": 5.233255035812747e-07, "loss": 1.1922, "step": 8612 }, { "epoch": 0.67, "learning_rate": 5.231045287380693e-07, "loss": 1.191, "step": 8613 }, { "epoch": 0.67, "learning_rate": 5.22883584032654e-07, "loss": 1.2236, "step": 8614 }, { "epoch": 0.67, "learning_rate": 5.226626694789912e-07, "loss": 1.1868, "step": 8615 }, { "epoch": 0.67, "learning_rate": 5.224417850910421e-07, "loss": 0.8164, "step": 8616 }, { "epoch": 0.67, "learning_rate": 5.222209308827652e-07, "loss": 1.1686, "step": 8617 }, { "epoch": 0.67, "learning_rate": 5.220001068681186e-07, "loss": 1.1769, "step": 8618 }, { "epoch": 0.67, "learning_rate": 5.217793130610565e-07, "loss": 0.83, "step": 8619 }, { "epoch": 0.67, "learning_rate": 5.215585494755323e-07, "loss": 0.7966, "step": 8620 }, { "epoch": 0.67, "learning_rate": 5.213378161254975e-07, "loss": 1.1697, "step": 8621 }, { "epoch": 0.67, "learning_rate": 5.211171130249013e-07, "loss": 1.1968, "step": 8622 }, { "epoch": 0.67, "learning_rate": 5.208964401876919e-07, "loss": 1.1562, "step": 8623 }, { "epoch": 0.67, "learning_rate": 5.20675797627814e-07, "loss": 0.8605, "step": 8624 }, { "epoch": 0.67, "learning_rate": 5.204551853592114e-07, "loss": 1.1705, "step": 8625 }, { "epoch": 0.67, "learning_rate": 5.202346033958267e-07, "loss": 1.2155, "step": 8626 }, { "epoch": 0.67, "learning_rate": 5.200140517515985e-07, "loss": 1.1562, "step": 8627 }, { "epoch": 0.67, "learning_rate": 5.197935304404659e-07, "loss": 1.2189, "step": 8628 }, { "epoch": 0.67, "learning_rate": 5.195730394763647e-07, "loss": 0.8588, "step": 8629 }, { "epoch": 0.67, "learning_rate": 5.193525788732281e-07, "loss": 1.1988, "step": 8630 }, { "epoch": 0.67, "learning_rate": 5.191321486449895e-07, "loss": 0.845, "step": 8631 }, { "epoch": 0.67, "learning_rate": 5.189117488055779e-07, "loss": 1.1677, "step": 8632 }, { "epoch": 0.67, "learning_rate": 5.186913793689227e-07, "loss": 1.1814, "step": 8633 }, { "epoch": 0.67, "learning_rate": 5.184710403489499e-07, "loss": 1.226, "step": 8634 }, { "epoch": 0.67, "learning_rate": 5.182507317595841e-07, "loss": 1.155, "step": 8635 }, { "epoch": 0.67, "learning_rate": 5.180304536147479e-07, "loss": 1.1917, "step": 8636 }, { "epoch": 0.67, "learning_rate": 5.178102059283618e-07, "loss": 1.2171, "step": 8637 }, { "epoch": 0.67, "learning_rate": 5.175899887143446e-07, "loss": 1.2, "step": 8638 }, { "epoch": 0.67, "learning_rate": 5.173698019866133e-07, "loss": 1.1682, "step": 8639 }, { "epoch": 0.67, "learning_rate": 5.171496457590822e-07, "loss": 1.1744, "step": 8640 }, { "epoch": 0.67, "learning_rate": 5.16929520045665e-07, "loss": 1.2186, "step": 8641 }, { "epoch": 0.67, "learning_rate": 5.167094248602722e-07, "loss": 0.8576, "step": 8642 }, { "epoch": 0.67, "learning_rate": 5.164893602168131e-07, "loss": 0.8523, "step": 8643 }, { "epoch": 0.67, "learning_rate": 5.162693261291948e-07, "loss": 1.2446, "step": 8644 }, { "epoch": 0.67, "learning_rate": 5.160493226113225e-07, "loss": 1.2071, "step": 8645 }, { "epoch": 0.67, "learning_rate": 5.158293496770996e-07, "loss": 1.2318, "step": 8646 }, { "epoch": 0.67, "learning_rate": 5.156094073404273e-07, "loss": 1.1366, "step": 8647 }, { "epoch": 0.67, "learning_rate": 5.153894956152052e-07, "loss": 1.1816, "step": 8648 }, { "epoch": 0.67, "learning_rate": 5.151696145153308e-07, "loss": 1.1888, "step": 8649 }, { "epoch": 0.67, "learning_rate": 5.149497640546994e-07, "loss": 1.2629, "step": 8650 }, { "epoch": 0.67, "learning_rate": 5.14729944247205e-07, "loss": 0.8296, "step": 8651 }, { "epoch": 0.67, "learning_rate": 5.14510155106739e-07, "loss": 1.2331, "step": 8652 }, { "epoch": 0.67, "learning_rate": 5.142903966471913e-07, "loss": 1.2254, "step": 8653 }, { "epoch": 0.67, "learning_rate": 5.140706688824496e-07, "loss": 1.2259, "step": 8654 }, { "epoch": 0.67, "learning_rate": 5.138509718264e-07, "loss": 0.8183, "step": 8655 }, { "epoch": 0.67, "learning_rate": 5.136313054929257e-07, "loss": 0.8029, "step": 8656 }, { "epoch": 0.67, "learning_rate": 5.134116698959101e-07, "loss": 1.233, "step": 8657 }, { "epoch": 0.67, "learning_rate": 5.13192065049232e-07, "loss": 1.1826, "step": 8658 }, { "epoch": 0.67, "learning_rate": 5.129724909667697e-07, "loss": 1.1932, "step": 8659 }, { "epoch": 0.67, "learning_rate": 5.127529476624001e-07, "loss": 1.1294, "step": 8660 }, { "epoch": 0.67, "learning_rate": 5.125334351499961e-07, "loss": 1.2067, "step": 8661 }, { "epoch": 0.67, "learning_rate": 5.123139534434317e-07, "loss": 1.1955, "step": 8662 }, { "epoch": 0.67, "learning_rate": 5.120945025565757e-07, "loss": 1.217, "step": 8663 }, { "epoch": 0.67, "learning_rate": 5.118750825032969e-07, "loss": 1.2173, "step": 8664 }, { "epoch": 0.67, "learning_rate": 5.116556932974625e-07, "loss": 0.8264, "step": 8665 }, { "epoch": 0.67, "learning_rate": 5.114363349529355e-07, "loss": 1.2525, "step": 8666 }, { "epoch": 0.67, "learning_rate": 5.112170074835797e-07, "loss": 1.1469, "step": 8667 }, { "epoch": 0.67, "learning_rate": 5.109977109032557e-07, "loss": 1.1935, "step": 8668 }, { "epoch": 0.67, "learning_rate": 5.107784452258209e-07, "loss": 1.2098, "step": 8669 }, { "epoch": 0.67, "learning_rate": 5.105592104651331e-07, "loss": 1.1713, "step": 8670 }, { "epoch": 0.67, "learning_rate": 5.103400066350467e-07, "loss": 1.2389, "step": 8671 }, { "epoch": 0.67, "learning_rate": 5.101208337494143e-07, "loss": 0.823, "step": 8672 }, { "epoch": 0.67, "learning_rate": 5.099016918220873e-07, "loss": 1.2054, "step": 8673 }, { "epoch": 0.67, "learning_rate": 5.096825808669133e-07, "loss": 1.1535, "step": 8674 }, { "epoch": 0.67, "learning_rate": 5.094635008977404e-07, "loss": 1.1813, "step": 8675 }, { "epoch": 0.67, "learning_rate": 5.092444519284129e-07, "loss": 1.1688, "step": 8676 }, { "epoch": 0.67, "learning_rate": 5.090254339727742e-07, "loss": 1.2115, "step": 8677 }, { "epoch": 0.67, "learning_rate": 5.088064470446649e-07, "loss": 0.862, "step": 8678 }, { "epoch": 0.67, "learning_rate": 5.085874911579244e-07, "loss": 1.1808, "step": 8679 }, { "epoch": 0.67, "learning_rate": 5.083685663263895e-07, "loss": 1.2055, "step": 8680 }, { "epoch": 0.67, "learning_rate": 5.081496725638954e-07, "loss": 1.2249, "step": 8681 }, { "epoch": 0.67, "learning_rate": 5.079308098842755e-07, "loss": 0.8308, "step": 8682 }, { "epoch": 0.67, "learning_rate": 5.077119783013608e-07, "loss": 1.2073, "step": 8683 }, { "epoch": 0.67, "learning_rate": 5.074931778289805e-07, "loss": 1.1985, "step": 8684 }, { "epoch": 0.67, "learning_rate": 5.07274408480962e-07, "loss": 1.1867, "step": 8685 }, { "epoch": 0.67, "learning_rate": 5.070556702711304e-07, "loss": 1.1437, "step": 8686 }, { "epoch": 0.67, "learning_rate": 5.068369632133094e-07, "loss": 0.822, "step": 8687 }, { "epoch": 0.67, "learning_rate": 5.066182873213198e-07, "loss": 0.8154, "step": 8688 }, { "epoch": 0.67, "learning_rate": 5.063996426089816e-07, "loss": 1.183, "step": 8689 }, { "epoch": 0.67, "learning_rate": 5.061810290901119e-07, "loss": 1.1636, "step": 8690 }, { "epoch": 0.67, "learning_rate": 5.059624467785262e-07, "loss": 1.1841, "step": 8691 }, { "epoch": 0.67, "learning_rate": 5.057438956880382e-07, "loss": 1.1721, "step": 8692 }, { "epoch": 0.67, "learning_rate": 5.055253758324587e-07, "loss": 0.8227, "step": 8693 }, { "epoch": 0.67, "learning_rate": 5.053068872255985e-07, "loss": 1.211, "step": 8694 }, { "epoch": 0.67, "learning_rate": 5.050884298812639e-07, "loss": 1.1882, "step": 8695 }, { "epoch": 0.67, "learning_rate": 5.048700038132612e-07, "loss": 0.7999, "step": 8696 }, { "epoch": 0.67, "learning_rate": 5.046516090353937e-07, "loss": 1.1893, "step": 8697 }, { "epoch": 0.68, "learning_rate": 5.044332455614629e-07, "loss": 1.1509, "step": 8698 }, { "epoch": 0.68, "learning_rate": 5.042149134052693e-07, "loss": 1.1981, "step": 8699 }, { "epoch": 0.68, "learning_rate": 5.039966125806092e-07, "loss": 1.1453, "step": 8700 }, { "epoch": 0.68, "learning_rate": 5.037783431012796e-07, "loss": 1.1986, "step": 8701 }, { "epoch": 0.68, "learning_rate": 5.03560104981074e-07, "loss": 1.2407, "step": 8702 }, { "epoch": 0.68, "learning_rate": 5.03341898233783e-07, "loss": 1.1903, "step": 8703 }, { "epoch": 0.68, "learning_rate": 5.031237228731974e-07, "loss": 0.821, "step": 8704 }, { "epoch": 0.68, "learning_rate": 5.02905578913105e-07, "loss": 1.2038, "step": 8705 }, { "epoch": 0.68, "learning_rate": 5.026874663672913e-07, "loss": 1.2121, "step": 8706 }, { "epoch": 0.68, "learning_rate": 5.024693852495403e-07, "loss": 1.2384, "step": 8707 }, { "epoch": 0.68, "learning_rate": 5.022513355736331e-07, "loss": 1.2655, "step": 8708 }, { "epoch": 0.68, "learning_rate": 5.020333173533502e-07, "loss": 1.2269, "step": 8709 }, { "epoch": 0.68, "learning_rate": 5.018153306024693e-07, "loss": 1.1859, "step": 8710 }, { "epoch": 0.68, "learning_rate": 5.015973753347663e-07, "loss": 1.1992, "step": 8711 }, { "epoch": 0.68, "learning_rate": 5.01379451564015e-07, "loss": 1.2362, "step": 8712 }, { "epoch": 0.68, "learning_rate": 5.011615593039872e-07, "loss": 0.8583, "step": 8713 }, { "epoch": 0.68, "learning_rate": 5.009436985684528e-07, "loss": 1.2072, "step": 8714 }, { "epoch": 0.68, "learning_rate": 5.007258693711797e-07, "loss": 1.1969, "step": 8715 }, { "epoch": 0.68, "learning_rate": 5.005080717259338e-07, "loss": 1.1672, "step": 8716 }, { "epoch": 0.68, "learning_rate": 5.00290305646479e-07, "loss": 1.222, "step": 8717 }, { "epoch": 0.68, "learning_rate": 5.000725711465772e-07, "loss": 1.1843, "step": 8718 }, { "epoch": 0.68, "learning_rate": 4.998548682399883e-07, "loss": 1.2018, "step": 8719 }, { "epoch": 0.68, "learning_rate": 4.996371969404702e-07, "loss": 1.2547, "step": 8720 }, { "epoch": 0.68, "learning_rate": 4.994195572617789e-07, "loss": 1.1905, "step": 8721 }, { "epoch": 0.68, "learning_rate": 4.992019492176681e-07, "loss": 1.1129, "step": 8722 }, { "epoch": 0.68, "learning_rate": 4.989843728218899e-07, "loss": 1.1668, "step": 8723 }, { "epoch": 0.68, "learning_rate": 4.987668280881942e-07, "loss": 1.2202, "step": 8724 }, { "epoch": 0.68, "learning_rate": 4.985493150303288e-07, "loss": 1.2104, "step": 8725 }, { "epoch": 0.68, "learning_rate": 4.983318336620397e-07, "loss": 1.1248, "step": 8726 }, { "epoch": 0.68, "learning_rate": 4.981143839970704e-07, "loss": 1.2442, "step": 8727 }, { "epoch": 0.68, "learning_rate": 4.978969660491639e-07, "loss": 1.2494, "step": 8728 }, { "epoch": 0.68, "learning_rate": 4.976795798320593e-07, "loss": 1.2253, "step": 8729 }, { "epoch": 0.68, "learning_rate": 4.974622253594943e-07, "loss": 1.1911, "step": 8730 }, { "epoch": 0.68, "learning_rate": 4.972449026452051e-07, "loss": 1.0957, "step": 8731 }, { "epoch": 0.68, "learning_rate": 4.970276117029253e-07, "loss": 1.1542, "step": 8732 }, { "epoch": 0.68, "learning_rate": 4.968103525463878e-07, "loss": 1.1499, "step": 8733 }, { "epoch": 0.68, "learning_rate": 4.965931251893213e-07, "loss": 1.1982, "step": 8734 }, { "epoch": 0.68, "learning_rate": 4.963759296454536e-07, "loss": 1.1693, "step": 8735 }, { "epoch": 0.68, "learning_rate": 4.961587659285118e-07, "loss": 1.2302, "step": 8736 }, { "epoch": 0.68, "learning_rate": 4.959416340522182e-07, "loss": 1.2947, "step": 8737 }, { "epoch": 0.68, "learning_rate": 4.957245340302963e-07, "loss": 1.2285, "step": 8738 }, { "epoch": 0.68, "learning_rate": 4.955074658764643e-07, "loss": 1.1696, "step": 8739 }, { "epoch": 0.68, "learning_rate": 4.952904296044404e-07, "loss": 1.2563, "step": 8740 }, { "epoch": 0.68, "learning_rate": 4.950734252279415e-07, "loss": 1.2472, "step": 8741 }, { "epoch": 0.68, "learning_rate": 4.948564527606797e-07, "loss": 1.193, "step": 8742 }, { "epoch": 0.68, "learning_rate": 4.946395122163679e-07, "loss": 1.1851, "step": 8743 }, { "epoch": 0.68, "learning_rate": 4.944226036087155e-07, "loss": 1.2061, "step": 8744 }, { "epoch": 0.68, "learning_rate": 4.942057269514303e-07, "loss": 1.1991, "step": 8745 }, { "epoch": 0.68, "learning_rate": 4.939888822582177e-07, "loss": 1.2386, "step": 8746 }, { "epoch": 0.68, "learning_rate": 4.937720695427817e-07, "loss": 0.8513, "step": 8747 }, { "epoch": 0.68, "learning_rate": 4.935552888188239e-07, "loss": 1.2078, "step": 8748 }, { "epoch": 0.68, "learning_rate": 4.933385401000437e-07, "loss": 1.1839, "step": 8749 }, { "epoch": 0.68, "learning_rate": 4.93121823400139e-07, "loss": 1.1942, "step": 8750 }, { "epoch": 0.68, "learning_rate": 4.929051387328051e-07, "loss": 1.1807, "step": 8751 }, { "epoch": 0.68, "learning_rate": 4.926884861117358e-07, "loss": 1.1356, "step": 8752 }, { "epoch": 0.68, "learning_rate": 4.924718655506226e-07, "loss": 1.1855, "step": 8753 }, { "epoch": 0.68, "learning_rate": 4.92255277063155e-07, "loss": 1.1813, "step": 8754 }, { "epoch": 0.68, "learning_rate": 4.920387206630202e-07, "loss": 1.2512, "step": 8755 }, { "epoch": 0.68, "learning_rate": 4.918221963639041e-07, "loss": 1.1985, "step": 8756 }, { "epoch": 0.68, "learning_rate": 4.916057041794898e-07, "loss": 1.2341, "step": 8757 }, { "epoch": 0.68, "learning_rate": 4.913892441234589e-07, "loss": 1.1921, "step": 8758 }, { "epoch": 0.68, "learning_rate": 4.911728162094905e-07, "loss": 1.145, "step": 8759 }, { "epoch": 0.68, "learning_rate": 4.909564204512622e-07, "loss": 1.2098, "step": 8760 }, { "epoch": 0.68, "learning_rate": 4.907400568624493e-07, "loss": 1.2238, "step": 8761 }, { "epoch": 0.68, "learning_rate": 4.90523725456725e-07, "loss": 0.8381, "step": 8762 }, { "epoch": 0.68, "learning_rate": 4.903074262477606e-07, "loss": 1.1542, "step": 8763 }, { "epoch": 0.68, "learning_rate": 4.90091159249225e-07, "loss": 1.1626, "step": 8764 }, { "epoch": 0.68, "learning_rate": 4.898749244747857e-07, "loss": 1.1696, "step": 8765 }, { "epoch": 0.68, "learning_rate": 4.896587219381073e-07, "loss": 1.1953, "step": 8766 }, { "epoch": 0.68, "learning_rate": 4.89442551652854e-07, "loss": 1.1906, "step": 8767 }, { "epoch": 0.68, "learning_rate": 4.89226413632686e-07, "loss": 1.1537, "step": 8768 }, { "epoch": 0.68, "learning_rate": 4.89010307891262e-07, "loss": 1.1502, "step": 8769 }, { "epoch": 0.68, "learning_rate": 4.887942344422401e-07, "loss": 1.2126, "step": 8770 }, { "epoch": 0.68, "learning_rate": 4.88578193299274e-07, "loss": 1.1844, "step": 8771 }, { "epoch": 0.68, "learning_rate": 4.883621844760178e-07, "loss": 1.2137, "step": 8772 }, { "epoch": 0.68, "learning_rate": 4.881462079861214e-07, "loss": 1.1948, "step": 8773 }, { "epoch": 0.68, "learning_rate": 4.879302638432336e-07, "loss": 1.1987, "step": 8774 }, { "epoch": 0.68, "learning_rate": 4.877143520610022e-07, "loss": 1.246, "step": 8775 }, { "epoch": 0.68, "learning_rate": 4.874984726530704e-07, "loss": 1.1898, "step": 8776 }, { "epoch": 0.68, "learning_rate": 4.872826256330822e-07, "loss": 1.1559, "step": 8777 }, { "epoch": 0.68, "learning_rate": 4.870668110146781e-07, "loss": 0.8471, "step": 8778 }, { "epoch": 0.68, "learning_rate": 4.868510288114954e-07, "loss": 1.1825, "step": 8779 }, { "epoch": 0.68, "learning_rate": 4.866352790371719e-07, "loss": 1.2417, "step": 8780 }, { "epoch": 0.68, "learning_rate": 4.864195617053418e-07, "loss": 1.2014, "step": 8781 }, { "epoch": 0.68, "learning_rate": 4.862038768296375e-07, "loss": 1.2209, "step": 8782 }, { "epoch": 0.68, "learning_rate": 4.859882244236895e-07, "loss": 1.2448, "step": 8783 }, { "epoch": 0.68, "learning_rate": 4.857726045011252e-07, "loss": 0.8447, "step": 8784 }, { "epoch": 0.68, "learning_rate": 4.855570170755721e-07, "loss": 1.1771, "step": 8785 }, { "epoch": 0.68, "learning_rate": 4.853414621606539e-07, "loss": 1.2457, "step": 8786 }, { "epoch": 0.68, "learning_rate": 4.851259397699926e-07, "loss": 1.1846, "step": 8787 }, { "epoch": 0.68, "learning_rate": 4.849104499172087e-07, "loss": 1.1773, "step": 8788 }, { "epoch": 0.68, "learning_rate": 4.846949926159199e-07, "loss": 1.1544, "step": 8789 }, { "epoch": 0.68, "learning_rate": 4.844795678797426e-07, "loss": 1.1901, "step": 8790 }, { "epoch": 0.68, "learning_rate": 4.842641757222902e-07, "loss": 1.1655, "step": 8791 }, { "epoch": 0.68, "learning_rate": 4.84048816157175e-07, "loss": 1.177, "step": 8792 }, { "epoch": 0.68, "learning_rate": 4.838334891980066e-07, "loss": 1.1605, "step": 8793 }, { "epoch": 0.68, "learning_rate": 4.83618194858393e-07, "loss": 1.1581, "step": 8794 }, { "epoch": 0.68, "learning_rate": 4.834029331519396e-07, "loss": 1.212, "step": 8795 }, { "epoch": 0.68, "learning_rate": 4.831877040922502e-07, "loss": 1.2508, "step": 8796 }, { "epoch": 0.68, "learning_rate": 4.829725076929264e-07, "loss": 1.244, "step": 8797 }, { "epoch": 0.68, "learning_rate": 4.827573439675677e-07, "loss": 1.1728, "step": 8798 }, { "epoch": 0.68, "learning_rate": 4.825422129297714e-07, "loss": 1.1997, "step": 8799 }, { "epoch": 0.68, "learning_rate": 4.823271145931329e-07, "loss": 1.2051, "step": 8800 }, { "epoch": 0.68, "learning_rate": 4.821120489712457e-07, "loss": 1.1841, "step": 8801 }, { "epoch": 0.68, "learning_rate": 4.81897016077701e-07, "loss": 1.2421, "step": 8802 }, { "epoch": 0.68, "learning_rate": 4.816820159260873e-07, "loss": 1.1556, "step": 8803 }, { "epoch": 0.68, "learning_rate": 4.814670485299934e-07, "loss": 1.2087, "step": 8804 }, { "epoch": 0.68, "learning_rate": 4.812521139030025e-07, "loss": 1.1877, "step": 8805 }, { "epoch": 0.68, "learning_rate": 4.810372120586987e-07, "loss": 1.2005, "step": 8806 }, { "epoch": 0.68, "learning_rate": 4.808223430106622e-07, "loss": 1.1669, "step": 8807 }, { "epoch": 0.68, "learning_rate": 4.806075067724718e-07, "loss": 1.1521, "step": 8808 }, { "epoch": 0.68, "learning_rate": 4.803927033577054e-07, "loss": 1.233, "step": 8809 }, { "epoch": 0.68, "learning_rate": 4.801779327799362e-07, "loss": 1.2014, "step": 8810 }, { "epoch": 0.68, "learning_rate": 4.799631950527376e-07, "loss": 1.2498, "step": 8811 }, { "epoch": 0.68, "learning_rate": 4.797484901896806e-07, "loss": 1.1969, "step": 8812 }, { "epoch": 0.68, "learning_rate": 4.795338182043321e-07, "loss": 1.211, "step": 8813 }, { "epoch": 0.68, "learning_rate": 4.793191791102601e-07, "loss": 1.1713, "step": 8814 }, { "epoch": 0.68, "learning_rate": 4.791045729210275e-07, "loss": 1.203, "step": 8815 }, { "epoch": 0.68, "learning_rate": 4.788899996501975e-07, "loss": 1.1642, "step": 8816 }, { "epoch": 0.68, "learning_rate": 4.786754593113303e-07, "loss": 1.2094, "step": 8817 }, { "epoch": 0.68, "learning_rate": 4.784609519179828e-07, "loss": 1.1854, "step": 8818 }, { "epoch": 0.68, "learning_rate": 4.782464774837123e-07, "loss": 1.236, "step": 8819 }, { "epoch": 0.68, "learning_rate": 4.780320360220719e-07, "loss": 1.1616, "step": 8820 }, { "epoch": 0.68, "learning_rate": 4.778176275466135e-07, "loss": 1.1538, "step": 8821 }, { "epoch": 0.68, "learning_rate": 4.776032520708872e-07, "loss": 1.1495, "step": 8822 }, { "epoch": 0.68, "learning_rate": 4.773889096084402e-07, "loss": 0.8783, "step": 8823 }, { "epoch": 0.68, "learning_rate": 4.771746001728182e-07, "loss": 1.219, "step": 8824 }, { "epoch": 0.68, "learning_rate": 4.769603237775647e-07, "loss": 1.1951, "step": 8825 }, { "epoch": 0.69, "learning_rate": 4.767460804362209e-07, "loss": 1.1451, "step": 8826 }, { "epoch": 0.69, "learning_rate": 4.7653187016232635e-07, "loss": 1.2285, "step": 8827 }, { "epoch": 0.69, "learning_rate": 4.7631769296941794e-07, "loss": 1.2356, "step": 8828 }, { "epoch": 0.69, "learning_rate": 4.7610354887103097e-07, "loss": 1.2387, "step": 8829 }, { "epoch": 0.69, "learning_rate": 4.7588943788069846e-07, "loss": 0.7883, "step": 8830 }, { "epoch": 0.69, "learning_rate": 4.756753600119512e-07, "loss": 1.1883, "step": 8831 }, { "epoch": 0.69, "learning_rate": 4.7546131527831804e-07, "loss": 1.2268, "step": 8832 }, { "epoch": 0.69, "learning_rate": 4.752473036933258e-07, "loss": 1.1638, "step": 8833 }, { "epoch": 0.69, "learning_rate": 4.75033325270499e-07, "loss": 1.1852, "step": 8834 }, { "epoch": 0.69, "learning_rate": 4.748193800233602e-07, "loss": 1.211, "step": 8835 }, { "epoch": 0.69, "learning_rate": 4.746054679654299e-07, "loss": 1.2269, "step": 8836 }, { "epoch": 0.69, "learning_rate": 4.7439158911022603e-07, "loss": 1.1927, "step": 8837 }, { "epoch": 0.69, "learning_rate": 4.7417774347126604e-07, "loss": 1.221, "step": 8838 }, { "epoch": 0.69, "learning_rate": 4.739639310620628e-07, "loss": 1.1994, "step": 8839 }, { "epoch": 0.69, "learning_rate": 4.7375015189612876e-07, "loss": 1.2024, "step": 8840 }, { "epoch": 0.69, "learning_rate": 4.73536405986974e-07, "loss": 0.799, "step": 8841 }, { "epoch": 0.69, "learning_rate": 4.733226933481059e-07, "loss": 1.2233, "step": 8842 }, { "epoch": 0.69, "learning_rate": 4.7310901399303125e-07, "loss": 1.1786, "step": 8843 }, { "epoch": 0.69, "learning_rate": 4.7289536793525273e-07, "loss": 1.172, "step": 8844 }, { "epoch": 0.69, "learning_rate": 4.726817551882717e-07, "loss": 1.1773, "step": 8845 }, { "epoch": 0.69, "learning_rate": 4.724681757655888e-07, "loss": 1.2522, "step": 8846 }, { "epoch": 0.69, "learning_rate": 4.7225462968069994e-07, "loss": 1.2176, "step": 8847 }, { "epoch": 0.69, "learning_rate": 4.720411169471017e-07, "loss": 1.2031, "step": 8848 }, { "epoch": 0.69, "learning_rate": 4.7182763757828603e-07, "loss": 1.223, "step": 8849 }, { "epoch": 0.69, "learning_rate": 4.716141915877441e-07, "loss": 1.1547, "step": 8850 }, { "epoch": 0.69, "learning_rate": 4.714007789889658e-07, "loss": 1.1521, "step": 8851 }, { "epoch": 0.69, "learning_rate": 4.711873997954364e-07, "loss": 1.2224, "step": 8852 }, { "epoch": 0.69, "learning_rate": 4.7097405402064175e-07, "loss": 1.1979, "step": 8853 }, { "epoch": 0.69, "learning_rate": 4.70760741678064e-07, "loss": 0.8556, "step": 8854 }, { "epoch": 0.69, "learning_rate": 4.7054746278118365e-07, "loss": 0.8144, "step": 8855 }, { "epoch": 0.69, "learning_rate": 4.703342173434789e-07, "loss": 1.2341, "step": 8856 }, { "epoch": 0.69, "learning_rate": 4.7012100537842613e-07, "loss": 1.2195, "step": 8857 }, { "epoch": 0.69, "learning_rate": 4.699078268994994e-07, "loss": 0.8472, "step": 8858 }, { "epoch": 0.69, "learning_rate": 4.696946819201706e-07, "loss": 1.1632, "step": 8859 }, { "epoch": 0.69, "learning_rate": 4.694815704539096e-07, "loss": 1.2239, "step": 8860 }, { "epoch": 0.69, "learning_rate": 4.692684925141842e-07, "loss": 1.2317, "step": 8861 }, { "epoch": 0.69, "learning_rate": 4.690554481144602e-07, "loss": 1.2186, "step": 8862 }, { "epoch": 0.69, "learning_rate": 4.688424372682008e-07, "loss": 1.2265, "step": 8863 }, { "epoch": 0.69, "learning_rate": 4.686294599888676e-07, "loss": 1.2056, "step": 8864 }, { "epoch": 0.69, "learning_rate": 4.6841651628991976e-07, "loss": 1.1857, "step": 8865 }, { "epoch": 0.69, "learning_rate": 4.6820360618481446e-07, "loss": 1.1878, "step": 8866 }, { "epoch": 0.69, "learning_rate": 4.679907296870067e-07, "loss": 1.1265, "step": 8867 }, { "epoch": 0.69, "learning_rate": 4.6777788680994956e-07, "loss": 0.8726, "step": 8868 }, { "epoch": 0.69, "learning_rate": 4.675650775670935e-07, "loss": 1.2085, "step": 8869 }, { "epoch": 0.69, "learning_rate": 4.6735230197188745e-07, "loss": 1.1908, "step": 8870 }, { "epoch": 0.69, "learning_rate": 4.6713956003777786e-07, "loss": 1.2163, "step": 8871 }, { "epoch": 0.69, "learning_rate": 4.66926851778209e-07, "loss": 0.7941, "step": 8872 }, { "epoch": 0.69, "learning_rate": 4.667141772066232e-07, "loss": 1.2312, "step": 8873 }, { "epoch": 0.69, "learning_rate": 4.665015363364607e-07, "loss": 0.8532, "step": 8874 }, { "epoch": 0.69, "learning_rate": 4.662889291811595e-07, "loss": 1.1624, "step": 8875 }, { "epoch": 0.69, "learning_rate": 4.6607635575415495e-07, "loss": 1.1625, "step": 8876 }, { "epoch": 0.69, "learning_rate": 4.6586381606888203e-07, "loss": 1.2369, "step": 8877 }, { "epoch": 0.69, "learning_rate": 4.6565131013877124e-07, "loss": 0.8979, "step": 8878 }, { "epoch": 0.69, "learning_rate": 4.654388379772519e-07, "loss": 0.8492, "step": 8879 }, { "epoch": 0.69, "learning_rate": 4.652263995977528e-07, "loss": 1.1787, "step": 8880 }, { "epoch": 0.69, "learning_rate": 4.650139950136974e-07, "loss": 1.1033, "step": 8881 }, { "epoch": 0.69, "learning_rate": 4.6480162423851034e-07, "loss": 0.8516, "step": 8882 }, { "epoch": 0.69, "learning_rate": 4.645892872856114e-07, "loss": 1.2192, "step": 8883 }, { "epoch": 0.69, "learning_rate": 4.6437698416841953e-07, "loss": 1.1507, "step": 8884 }, { "epoch": 0.69, "learning_rate": 4.6416471490035234e-07, "loss": 1.1613, "step": 8885 }, { "epoch": 0.69, "learning_rate": 4.6395247949482295e-07, "loss": 1.1629, "step": 8886 }, { "epoch": 0.69, "learning_rate": 4.6374027796524497e-07, "loss": 0.7997, "step": 8887 }, { "epoch": 0.69, "learning_rate": 4.6352811032502835e-07, "loss": 1.1943, "step": 8888 }, { "epoch": 0.69, "learning_rate": 4.6331597658758036e-07, "loss": 1.2066, "step": 8889 }, { "epoch": 0.69, "learning_rate": 4.6310387676630837e-07, "loss": 0.812, "step": 8890 }, { "epoch": 0.69, "learning_rate": 4.628918108746146e-07, "loss": 1.2177, "step": 8891 }, { "epoch": 0.69, "learning_rate": 4.6267977892590193e-07, "loss": 1.1892, "step": 8892 }, { "epoch": 0.69, "learning_rate": 4.624677809335699e-07, "loss": 0.8183, "step": 8893 }, { "epoch": 0.69, "learning_rate": 4.6225581691101467e-07, "loss": 1.2206, "step": 8894 }, { "epoch": 0.69, "learning_rate": 4.620438868716328e-07, "loss": 1.1782, "step": 8895 }, { "epoch": 0.69, "learning_rate": 4.6183199082881683e-07, "loss": 1.1502, "step": 8896 }, { "epoch": 0.69, "learning_rate": 4.616201287959578e-07, "loss": 1.2122, "step": 8897 }, { "epoch": 0.69, "learning_rate": 4.6140830078644444e-07, "loss": 0.837, "step": 8898 }, { "epoch": 0.69, "learning_rate": 4.611965068136633e-07, "loss": 1.146, "step": 8899 }, { "epoch": 0.69, "learning_rate": 4.609847468909991e-07, "loss": 1.2073, "step": 8900 }, { "epoch": 0.69, "learning_rate": 4.60773021031834e-07, "loss": 1.183, "step": 8901 }, { "epoch": 0.69, "learning_rate": 4.605613292495482e-07, "loss": 1.1896, "step": 8902 }, { "epoch": 0.69, "learning_rate": 4.603496715575198e-07, "loss": 1.151, "step": 8903 }, { "epoch": 0.69, "learning_rate": 4.601380479691247e-07, "loss": 1.2154, "step": 8904 }, { "epoch": 0.69, "learning_rate": 4.599264584977365e-07, "loss": 1.2044, "step": 8905 }, { "epoch": 0.69, "learning_rate": 4.597149031567268e-07, "loss": 1.2575, "step": 8906 }, { "epoch": 0.69, "learning_rate": 4.595033819594651e-07, "loss": 1.2143, "step": 8907 }, { "epoch": 0.69, "learning_rate": 4.5929189491931855e-07, "loss": 1.158, "step": 8908 }, { "epoch": 0.69, "learning_rate": 4.5908044204965235e-07, "loss": 1.211, "step": 8909 }, { "epoch": 0.69, "learning_rate": 4.5886902336382927e-07, "loss": 0.8523, "step": 8910 }, { "epoch": 0.69, "learning_rate": 4.5865763887521025e-07, "loss": 1.2083, "step": 8911 }, { "epoch": 0.69, "learning_rate": 4.5844628859715374e-07, "loss": 1.1814, "step": 8912 }, { "epoch": 0.69, "learning_rate": 4.5823497254301634e-07, "loss": 1.2275, "step": 8913 }, { "epoch": 0.69, "learning_rate": 4.580236907261523e-07, "loss": 0.8575, "step": 8914 }, { "epoch": 0.69, "learning_rate": 4.5781244315991364e-07, "loss": 1.2016, "step": 8915 }, { "epoch": 0.69, "learning_rate": 4.576012298576505e-07, "loss": 1.1899, "step": 8916 }, { "epoch": 0.69, "learning_rate": 4.5739005083271054e-07, "loss": 0.8828, "step": 8917 }, { "epoch": 0.69, "learning_rate": 4.57178906098439e-07, "loss": 1.1107, "step": 8918 }, { "epoch": 0.69, "learning_rate": 4.5696779566818045e-07, "loss": 1.1764, "step": 8919 }, { "epoch": 0.69, "learning_rate": 4.5675671955527486e-07, "loss": 1.2069, "step": 8920 }, { "epoch": 0.69, "learning_rate": 4.565456777730623e-07, "loss": 1.1455, "step": 8921 }, { "epoch": 0.69, "learning_rate": 4.5633467033487983e-07, "loss": 1.2116, "step": 8922 }, { "epoch": 0.69, "learning_rate": 4.5612369725406097e-07, "loss": 1.1557, "step": 8923 }, { "epoch": 0.69, "learning_rate": 4.5591275854393995e-07, "loss": 1.1928, "step": 8924 }, { "epoch": 0.69, "learning_rate": 4.557018542178458e-07, "loss": 1.1714, "step": 8925 }, { "epoch": 0.69, "learning_rate": 4.5549098428910757e-07, "loss": 1.1976, "step": 8926 }, { "epoch": 0.69, "learning_rate": 4.552801487710518e-07, "loss": 1.1833, "step": 8927 }, { "epoch": 0.69, "learning_rate": 4.550693476770009e-07, "loss": 1.2153, "step": 8928 }, { "epoch": 0.69, "learning_rate": 4.5485858102027794e-07, "loss": 1.1916, "step": 8929 }, { "epoch": 0.69, "learning_rate": 4.5464784881420215e-07, "loss": 1.246, "step": 8930 }, { "epoch": 0.69, "learning_rate": 4.544371510720908e-07, "loss": 1.1627, "step": 8931 }, { "epoch": 0.69, "learning_rate": 4.5422648780725917e-07, "loss": 1.181, "step": 8932 }, { "epoch": 0.69, "learning_rate": 4.540158590330202e-07, "loss": 0.8335, "step": 8933 }, { "epoch": 0.69, "learning_rate": 4.5380526476268486e-07, "loss": 1.2372, "step": 8934 }, { "epoch": 0.69, "learning_rate": 4.535947050095618e-07, "loss": 1.1787, "step": 8935 }, { "epoch": 0.69, "learning_rate": 4.533841797869574e-07, "loss": 1.244, "step": 8936 }, { "epoch": 0.69, "learning_rate": 4.5317368910817613e-07, "loss": 1.2117, "step": 8937 }, { "epoch": 0.69, "learning_rate": 4.5296323298652e-07, "loss": 1.1594, "step": 8938 }, { "epoch": 0.69, "learning_rate": 4.5275281143528897e-07, "loss": 1.1497, "step": 8939 }, { "epoch": 0.69, "learning_rate": 4.525424244677808e-07, "loss": 0.8399, "step": 8940 }, { "epoch": 0.69, "learning_rate": 4.523320720972912e-07, "loss": 1.2136, "step": 8941 }, { "epoch": 0.69, "learning_rate": 4.521217543371133e-07, "loss": 1.2439, "step": 8942 }, { "epoch": 0.69, "learning_rate": 4.519114712005385e-07, "loss": 1.2733, "step": 8943 }, { "epoch": 0.69, "learning_rate": 4.5170122270085585e-07, "loss": 1.2137, "step": 8944 }, { "epoch": 0.69, "learning_rate": 4.5149100885135195e-07, "loss": 0.8565, "step": 8945 }, { "epoch": 0.69, "learning_rate": 4.5128082966531155e-07, "loss": 1.163, "step": 8946 }, { "epoch": 0.69, "learning_rate": 4.510706851560172e-07, "loss": 1.1739, "step": 8947 }, { "epoch": 0.69, "learning_rate": 4.5086057533674893e-07, "loss": 1.1584, "step": 8948 }, { "epoch": 0.69, "learning_rate": 4.50650500220785e-07, "loss": 1.1857, "step": 8949 }, { "epoch": 0.69, "learning_rate": 4.504404598214011e-07, "loss": 1.1963, "step": 8950 }, { "epoch": 0.69, "learning_rate": 4.502304541518711e-07, "loss": 1.1952, "step": 8951 }, { "epoch": 0.69, "learning_rate": 4.5002048322546594e-07, "loss": 1.2385, "step": 8952 }, { "epoch": 0.69, "learning_rate": 4.4981054705545594e-07, "loss": 1.192, "step": 8953 }, { "epoch": 0.69, "learning_rate": 4.4960064565510736e-07, "loss": 1.1982, "step": 8954 }, { "epoch": 0.7, "learning_rate": 4.4939077903768486e-07, "loss": 1.197, "step": 8955 }, { "epoch": 0.7, "learning_rate": 4.491809472164524e-07, "loss": 0.8459, "step": 8956 }, { "epoch": 0.7, "learning_rate": 4.489711502046688e-07, "loss": 1.1853, "step": 8957 }, { "epoch": 0.7, "learning_rate": 4.4876138801559394e-07, "loss": 1.2032, "step": 8958 }, { "epoch": 0.7, "learning_rate": 4.4855166066248275e-07, "loss": 1.1998, "step": 8959 }, { "epoch": 0.7, "learning_rate": 4.483419681585892e-07, "loss": 1.2075, "step": 8960 }, { "epoch": 0.7, "learning_rate": 4.481323105171659e-07, "loss": 1.1976, "step": 8961 }, { "epoch": 0.7, "learning_rate": 4.479226877514611e-07, "loss": 1.1956, "step": 8962 }, { "epoch": 0.7, "learning_rate": 4.4771309987472303e-07, "loss": 1.1854, "step": 8963 }, { "epoch": 0.7, "learning_rate": 4.4750354690019655e-07, "loss": 1.2065, "step": 8964 }, { "epoch": 0.7, "learning_rate": 4.4729402884112434e-07, "loss": 1.283, "step": 8965 }, { "epoch": 0.7, "learning_rate": 4.470845457107475e-07, "loss": 0.8192, "step": 8966 }, { "epoch": 0.7, "learning_rate": 4.468750975223033e-07, "loss": 1.1476, "step": 8967 }, { "epoch": 0.7, "learning_rate": 4.4666568428902927e-07, "loss": 1.1723, "step": 8968 }, { "epoch": 0.7, "learning_rate": 4.4645630602415895e-07, "loss": 0.882, "step": 8969 }, { "epoch": 0.7, "learning_rate": 4.4624696274092424e-07, "loss": 1.1791, "step": 8970 }, { "epoch": 0.7, "learning_rate": 4.4603765445255457e-07, "loss": 1.1575, "step": 8971 }, { "epoch": 0.7, "learning_rate": 4.4582838117227763e-07, "loss": 1.1967, "step": 8972 }, { "epoch": 0.7, "learning_rate": 4.4561914291331846e-07, "loss": 1.1732, "step": 8973 }, { "epoch": 0.7, "learning_rate": 4.4540993968890006e-07, "loss": 1.1854, "step": 8974 }, { "epoch": 0.7, "learning_rate": 4.452007715122431e-07, "loss": 1.2379, "step": 8975 }, { "epoch": 0.7, "learning_rate": 4.449916383965663e-07, "loss": 1.2045, "step": 8976 }, { "epoch": 0.7, "learning_rate": 4.447825403550858e-07, "loss": 1.199, "step": 8977 }, { "epoch": 0.7, "learning_rate": 4.445734774010159e-07, "loss": 0.8224, "step": 8978 }, { "epoch": 0.7, "learning_rate": 4.4436444954756845e-07, "loss": 1.199, "step": 8979 }, { "epoch": 0.7, "learning_rate": 4.441554568079531e-07, "loss": 1.1843, "step": 8980 }, { "epoch": 0.7, "learning_rate": 4.439464991953772e-07, "loss": 1.2084, "step": 8981 }, { "epoch": 0.7, "learning_rate": 4.4373757672304625e-07, "loss": 1.1127, "step": 8982 }, { "epoch": 0.7, "learning_rate": 4.435286894041631e-07, "loss": 1.227, "step": 8983 }, { "epoch": 0.7, "learning_rate": 4.433198372519286e-07, "loss": 1.2017, "step": 8984 }, { "epoch": 0.7, "learning_rate": 4.4311102027954114e-07, "loss": 1.2287, "step": 8985 }, { "epoch": 0.7, "learning_rate": 4.4290223850019714e-07, "loss": 0.8297, "step": 8986 }, { "epoch": 0.7, "learning_rate": 4.426934919270914e-07, "loss": 0.8235, "step": 8987 }, { "epoch": 0.7, "learning_rate": 4.42484780573415e-07, "loss": 1.2306, "step": 8988 }, { "epoch": 0.7, "learning_rate": 4.4227610445235786e-07, "loss": 1.1671, "step": 8989 }, { "epoch": 0.7, "learning_rate": 4.4206746357710747e-07, "loss": 1.1791, "step": 8990 }, { "epoch": 0.7, "learning_rate": 4.4185885796084866e-07, "loss": 1.1836, "step": 8991 }, { "epoch": 0.7, "learning_rate": 4.416502876167655e-07, "loss": 1.2196, "step": 8992 }, { "epoch": 0.7, "learning_rate": 4.414417525580377e-07, "loss": 1.1651, "step": 8993 }, { "epoch": 0.7, "learning_rate": 4.41233252797844e-07, "loss": 0.8427, "step": 8994 }, { "epoch": 0.7, "learning_rate": 4.410247883493614e-07, "loss": 1.2246, "step": 8995 }, { "epoch": 0.7, "learning_rate": 4.408163592257629e-07, "loss": 1.1542, "step": 8996 }, { "epoch": 0.7, "learning_rate": 4.406079654402212e-07, "loss": 1.2258, "step": 8997 }, { "epoch": 0.7, "learning_rate": 4.4039960700590594e-07, "loss": 1.1739, "step": 8998 }, { "epoch": 0.7, "learning_rate": 4.401912839359835e-07, "loss": 1.2056, "step": 8999 }, { "epoch": 0.7, "learning_rate": 4.3998299624362035e-07, "loss": 1.17, "step": 9000 }, { "epoch": 0.7, "learning_rate": 4.397747439419781e-07, "loss": 1.2196, "step": 9001 }, { "epoch": 0.7, "learning_rate": 4.3956652704421836e-07, "loss": 1.2141, "step": 9002 }, { "epoch": 0.7, "learning_rate": 4.393583455634997e-07, "loss": 1.1962, "step": 9003 }, { "epoch": 0.7, "learning_rate": 4.391501995129772e-07, "loss": 0.82, "step": 9004 }, { "epoch": 0.7, "learning_rate": 4.3894208890580575e-07, "loss": 1.2083, "step": 9005 }, { "epoch": 0.7, "learning_rate": 4.3873401375513687e-07, "loss": 1.2626, "step": 9006 }, { "epoch": 0.7, "learning_rate": 4.3852597407411994e-07, "loss": 1.1701, "step": 9007 }, { "epoch": 0.7, "learning_rate": 4.383179698759023e-07, "loss": 1.1995, "step": 9008 }, { "epoch": 0.7, "learning_rate": 4.3811000117362884e-07, "loss": 1.1612, "step": 9009 }, { "epoch": 0.7, "learning_rate": 4.379020679804423e-07, "loss": 1.2257, "step": 9010 }, { "epoch": 0.7, "learning_rate": 4.376941703094833e-07, "loss": 1.1765, "step": 9011 }, { "epoch": 0.7, "learning_rate": 4.3748630817388997e-07, "loss": 1.1973, "step": 9012 }, { "epoch": 0.7, "learning_rate": 4.372784815867985e-07, "loss": 1.1708, "step": 9013 }, { "epoch": 0.7, "learning_rate": 4.370706905613424e-07, "loss": 1.2082, "step": 9014 }, { "epoch": 0.7, "learning_rate": 4.3686293511065333e-07, "loss": 1.2135, "step": 9015 }, { "epoch": 0.7, "learning_rate": 4.3665521524786063e-07, "loss": 0.852, "step": 9016 }, { "epoch": 0.7, "learning_rate": 4.364475309860912e-07, "loss": 1.1828, "step": 9017 }, { "epoch": 0.7, "learning_rate": 4.3623988233846996e-07, "loss": 1.1504, "step": 9018 }, { "epoch": 0.7, "learning_rate": 4.360322693181192e-07, "loss": 1.2512, "step": 9019 }, { "epoch": 0.7, "learning_rate": 4.358246919381594e-07, "loss": 1.2523, "step": 9020 }, { "epoch": 0.7, "learning_rate": 4.3561715021170854e-07, "loss": 1.1884, "step": 9021 }, { "epoch": 0.7, "learning_rate": 4.354096441518823e-07, "loss": 1.2201, "step": 9022 }, { "epoch": 0.7, "learning_rate": 4.3520217377179427e-07, "loss": 1.2166, "step": 9023 }, { "epoch": 0.7, "learning_rate": 4.3499473908455564e-07, "loss": 1.1623, "step": 9024 }, { "epoch": 0.7, "learning_rate": 4.347873401032754e-07, "loss": 1.2047, "step": 9025 }, { "epoch": 0.7, "learning_rate": 4.345799768410605e-07, "loss": 0.819, "step": 9026 }, { "epoch": 0.7, "learning_rate": 4.3437264931101516e-07, "loss": 1.2232, "step": 9027 }, { "epoch": 0.7, "learning_rate": 4.3416535752624137e-07, "loss": 1.1748, "step": 9028 }, { "epoch": 0.7, "learning_rate": 4.339581014998401e-07, "loss": 1.1745, "step": 9029 }, { "epoch": 0.7, "learning_rate": 4.3375088124490777e-07, "loss": 1.2199, "step": 9030 }, { "epoch": 0.7, "learning_rate": 4.335436967745408e-07, "loss": 1.2096, "step": 9031 }, { "epoch": 0.7, "learning_rate": 4.3333654810183254e-07, "loss": 1.1926, "step": 9032 }, { "epoch": 0.7, "learning_rate": 4.3312943523987256e-07, "loss": 1.1965, "step": 9033 }, { "epoch": 0.7, "learning_rate": 4.3292235820175104e-07, "loss": 1.1815, "step": 9034 }, { "epoch": 0.7, "learning_rate": 4.327153170005531e-07, "loss": 1.238, "step": 9035 }, { "epoch": 0.7, "learning_rate": 4.3250831164936385e-07, "loss": 1.149, "step": 9036 }, { "epoch": 0.7, "learning_rate": 4.3230134216126514e-07, "loss": 1.191, "step": 9037 }, { "epoch": 0.7, "learning_rate": 4.3209440854933554e-07, "loss": 1.1954, "step": 9038 }, { "epoch": 0.7, "learning_rate": 4.3188751082665345e-07, "loss": 1.1596, "step": 9039 }, { "epoch": 0.7, "learning_rate": 4.3168064900629354e-07, "loss": 1.1833, "step": 9040 }, { "epoch": 0.7, "learning_rate": 4.3147382310132863e-07, "loss": 1.1796, "step": 9041 }, { "epoch": 0.7, "learning_rate": 4.312670331248297e-07, "loss": 1.213, "step": 9042 }, { "epoch": 0.7, "learning_rate": 4.3106027908986374e-07, "loss": 1.1856, "step": 9043 }, { "epoch": 0.7, "learning_rate": 4.30853561009498e-07, "loss": 1.1657, "step": 9044 }, { "epoch": 0.7, "learning_rate": 4.306468788967956e-07, "loss": 1.0885, "step": 9045 }, { "epoch": 0.7, "learning_rate": 4.304402327648183e-07, "loss": 1.2077, "step": 9046 }, { "epoch": 0.7, "learning_rate": 4.30233622626625e-07, "loss": 1.1844, "step": 9047 }, { "epoch": 0.7, "learning_rate": 4.300270484952727e-07, "loss": 1.1756, "step": 9048 }, { "epoch": 0.7, "learning_rate": 4.2982051038381606e-07, "loss": 0.8293, "step": 9049 }, { "epoch": 0.7, "learning_rate": 4.296140083053074e-07, "loss": 1.2079, "step": 9050 }, { "epoch": 0.7, "learning_rate": 4.294075422727966e-07, "loss": 1.1855, "step": 9051 }, { "epoch": 0.7, "learning_rate": 4.2920111229933177e-07, "loss": 1.1682, "step": 9052 }, { "epoch": 0.7, "learning_rate": 4.289947183979582e-07, "loss": 1.1481, "step": 9053 }, { "epoch": 0.7, "learning_rate": 4.287883605817192e-07, "loss": 0.8454, "step": 9054 }, { "epoch": 0.7, "learning_rate": 4.285820388636555e-07, "loss": 1.2347, "step": 9055 }, { "epoch": 0.7, "learning_rate": 4.283757532568062e-07, "loss": 1.1586, "step": 9056 }, { "epoch": 0.7, "learning_rate": 4.281695037742072e-07, "loss": 1.1813, "step": 9057 }, { "epoch": 0.7, "learning_rate": 4.27963290428893e-07, "loss": 1.2174, "step": 9058 }, { "epoch": 0.7, "learning_rate": 4.277571132338952e-07, "loss": 1.155, "step": 9059 }, { "epoch": 0.7, "learning_rate": 4.275509722022433e-07, "loss": 1.1661, "step": 9060 }, { "epoch": 0.7, "learning_rate": 4.2734486734696475e-07, "loss": 1.25, "step": 9061 }, { "epoch": 0.7, "learning_rate": 4.271387986810839e-07, "loss": 0.7957, "step": 9062 }, { "epoch": 0.7, "learning_rate": 4.2693276621762473e-07, "loss": 0.8757, "step": 9063 }, { "epoch": 0.7, "learning_rate": 4.2672676996960644e-07, "loss": 1.1559, "step": 9064 }, { "epoch": 0.7, "learning_rate": 4.2652080995004746e-07, "loss": 1.2045, "step": 9065 }, { "epoch": 0.7, "learning_rate": 4.263148861719637e-07, "loss": 1.2341, "step": 9066 }, { "epoch": 0.7, "learning_rate": 4.2610899864836816e-07, "loss": 1.1806, "step": 9067 }, { "epoch": 0.7, "learning_rate": 4.259031473922733e-07, "loss": 1.1842, "step": 9068 }, { "epoch": 0.7, "learning_rate": 4.256973324166868e-07, "loss": 1.1878, "step": 9069 }, { "epoch": 0.7, "learning_rate": 4.2549155373461553e-07, "loss": 1.154, "step": 9070 }, { "epoch": 0.7, "learning_rate": 4.2528581135906485e-07, "loss": 0.8178, "step": 9071 }, { "epoch": 0.7, "learning_rate": 4.250801053030352e-07, "loss": 0.8358, "step": 9072 }, { "epoch": 0.7, "learning_rate": 4.248744355795276e-07, "loss": 0.8679, "step": 9073 }, { "epoch": 0.7, "learning_rate": 4.246688022015391e-07, "loss": 1.199, "step": 9074 }, { "epoch": 0.7, "learning_rate": 4.244632051820648e-07, "loss": 1.1925, "step": 9075 }, { "epoch": 0.7, "learning_rate": 4.2425764453409805e-07, "loss": 1.2758, "step": 9076 }, { "epoch": 0.7, "learning_rate": 4.240521202706283e-07, "loss": 1.157, "step": 9077 }, { "epoch": 0.7, "learning_rate": 4.238466324046448e-07, "loss": 1.2094, "step": 9078 }, { "epoch": 0.7, "learning_rate": 4.2364118094913323e-07, "loss": 1.1567, "step": 9079 }, { "epoch": 0.7, "learning_rate": 4.234357659170774e-07, "loss": 1.2353, "step": 9080 }, { "epoch": 0.7, "learning_rate": 4.232303873214584e-07, "loss": 1.2033, "step": 9081 }, { "epoch": 0.7, "learning_rate": 4.230250451752555e-07, "loss": 1.1887, "step": 9082 }, { "epoch": 0.7, "learning_rate": 4.228197394914454e-07, "loss": 1.229, "step": 9083 }, { "epoch": 0.71, "learning_rate": 4.2261447028300256e-07, "loss": 1.1783, "step": 9084 }, { "epoch": 0.71, "learning_rate": 4.2240923756289915e-07, "loss": 1.2599, "step": 9085 }, { "epoch": 0.71, "learning_rate": 4.222040413441049e-07, "loss": 1.1835, "step": 9086 }, { "epoch": 0.71, "learning_rate": 4.2199888163958755e-07, "loss": 0.8453, "step": 9087 }, { "epoch": 0.71, "learning_rate": 4.217937584623121e-07, "loss": 1.1651, "step": 9088 }, { "epoch": 0.71, "learning_rate": 4.215886718252417e-07, "loss": 1.2215, "step": 9089 }, { "epoch": 0.71, "learning_rate": 4.213836217413369e-07, "loss": 1.1701, "step": 9090 }, { "epoch": 0.71, "learning_rate": 4.2117860822355587e-07, "loss": 1.2115, "step": 9091 }, { "epoch": 0.71, "learning_rate": 4.209736312848547e-07, "loss": 0.812, "step": 9092 }, { "epoch": 0.71, "learning_rate": 4.2076869093818713e-07, "loss": 1.2492, "step": 9093 }, { "epoch": 0.71, "learning_rate": 4.205637871965044e-07, "loss": 1.1935, "step": 9094 }, { "epoch": 0.71, "learning_rate": 4.203589200727556e-07, "loss": 1.1563, "step": 9095 }, { "epoch": 0.71, "learning_rate": 4.201540895798872e-07, "loss": 1.2173, "step": 9096 }, { "epoch": 0.71, "learning_rate": 4.199492957308447e-07, "loss": 1.1823, "step": 9097 }, { "epoch": 0.71, "learning_rate": 4.1974453853856893e-07, "loss": 1.1629, "step": 9098 }, { "epoch": 0.71, "learning_rate": 4.1953981801600033e-07, "loss": 1.1335, "step": 9099 }, { "epoch": 0.71, "learning_rate": 4.1933513417607626e-07, "loss": 1.1929, "step": 9100 }, { "epoch": 0.71, "learning_rate": 4.191304870317314e-07, "loss": 1.201, "step": 9101 }, { "epoch": 0.71, "learning_rate": 4.189258765958997e-07, "loss": 1.2535, "step": 9102 }, { "epoch": 0.71, "learning_rate": 4.187213028815109e-07, "loss": 1.1724, "step": 9103 }, { "epoch": 0.71, "learning_rate": 4.1851676590149275e-07, "loss": 0.8332, "step": 9104 }, { "epoch": 0.71, "learning_rate": 4.1831226566877254e-07, "loss": 1.1776, "step": 9105 }, { "epoch": 0.71, "learning_rate": 4.181078021962723e-07, "loss": 1.213, "step": 9106 }, { "epoch": 0.71, "learning_rate": 4.179033754969142e-07, "loss": 1.2208, "step": 9107 }, { "epoch": 0.71, "learning_rate": 4.1769898558361736e-07, "loss": 0.8472, "step": 9108 }, { "epoch": 0.71, "learning_rate": 4.1749463246929705e-07, "loss": 1.2495, "step": 9109 }, { "epoch": 0.71, "learning_rate": 4.172903161668693e-07, "loss": 1.212, "step": 9110 }, { "epoch": 0.71, "learning_rate": 4.1708603668924427e-07, "loss": 1.2087, "step": 9111 }, { "epoch": 0.71, "learning_rate": 4.168817940493328e-07, "loss": 1.1706, "step": 9112 }, { "epoch": 0.71, "learning_rate": 4.1667758826004185e-07, "loss": 1.2166, "step": 9113 }, { "epoch": 0.71, "learning_rate": 4.1647341933427624e-07, "loss": 1.1315, "step": 9114 }, { "epoch": 0.71, "learning_rate": 4.162692872849387e-07, "loss": 1.2, "step": 9115 }, { "epoch": 0.71, "learning_rate": 4.160651921249294e-07, "loss": 1.2251, "step": 9116 }, { "epoch": 0.71, "learning_rate": 4.158611338671464e-07, "loss": 1.215, "step": 9117 }, { "epoch": 0.71, "learning_rate": 4.156571125244853e-07, "loss": 1.1951, "step": 9118 }, { "epoch": 0.71, "learning_rate": 4.1545312810983936e-07, "loss": 1.1675, "step": 9119 }, { "epoch": 0.71, "learning_rate": 4.152491806360996e-07, "loss": 1.191, "step": 9120 }, { "epoch": 0.71, "learning_rate": 4.1504527011615463e-07, "loss": 1.1988, "step": 9121 }, { "epoch": 0.71, "learning_rate": 4.148413965628906e-07, "loss": 1.2336, "step": 9122 }, { "epoch": 0.71, "learning_rate": 4.146375599891917e-07, "loss": 1.2202, "step": 9123 }, { "epoch": 0.71, "learning_rate": 4.144337604079393e-07, "loss": 1.1808, "step": 9124 }, { "epoch": 0.71, "learning_rate": 4.1422999783201295e-07, "loss": 1.1805, "step": 9125 }, { "epoch": 0.71, "learning_rate": 4.140262722742893e-07, "loss": 1.2043, "step": 9126 }, { "epoch": 0.71, "learning_rate": 4.1382258374764314e-07, "loss": 1.2605, "step": 9127 }, { "epoch": 0.71, "learning_rate": 4.1361893226494673e-07, "loss": 0.8435, "step": 9128 }, { "epoch": 0.71, "learning_rate": 4.134153178390698e-07, "loss": 0.819, "step": 9129 }, { "epoch": 0.71, "learning_rate": 4.132117404828801e-07, "loss": 1.2232, "step": 9130 }, { "epoch": 0.71, "learning_rate": 4.1300820020924287e-07, "loss": 1.1548, "step": 9131 }, { "epoch": 0.71, "learning_rate": 4.128046970310209e-07, "loss": 1.2331, "step": 9132 }, { "epoch": 0.71, "learning_rate": 4.126012309610748e-07, "loss": 1.227, "step": 9133 }, { "epoch": 0.71, "learning_rate": 4.123978020122628e-07, "loss": 0.8238, "step": 9134 }, { "epoch": 0.71, "learning_rate": 4.121944101974404e-07, "loss": 1.2093, "step": 9135 }, { "epoch": 0.71, "learning_rate": 4.11991055529462e-07, "loss": 1.1939, "step": 9136 }, { "epoch": 0.71, "learning_rate": 4.1178773802117795e-07, "loss": 0.86, "step": 9137 }, { "epoch": 0.71, "learning_rate": 4.115844576854368e-07, "loss": 1.2041, "step": 9138 }, { "epoch": 0.71, "learning_rate": 4.1138121453508636e-07, "loss": 0.8265, "step": 9139 }, { "epoch": 0.71, "learning_rate": 4.1117800858296903e-07, "loss": 1.2082, "step": 9140 }, { "epoch": 0.71, "learning_rate": 4.1097483984192826e-07, "loss": 1.2061, "step": 9141 }, { "epoch": 0.71, "learning_rate": 4.107717083248022e-07, "loss": 1.2284, "step": 9142 }, { "epoch": 0.71, "learning_rate": 4.105686140444281e-07, "loss": 1.1835, "step": 9143 }, { "epoch": 0.71, "learning_rate": 4.103655570136415e-07, "loss": 1.2096, "step": 9144 }, { "epoch": 0.71, "learning_rate": 4.101625372452735e-07, "loss": 1.1858, "step": 9145 }, { "epoch": 0.71, "learning_rate": 4.0995955475215517e-07, "loss": 1.1939, "step": 9146 }, { "epoch": 0.71, "learning_rate": 4.09756609547114e-07, "loss": 1.1064, "step": 9147 }, { "epoch": 0.71, "learning_rate": 4.0955370164297434e-07, "loss": 1.2237, "step": 9148 }, { "epoch": 0.71, "learning_rate": 4.093508310525601e-07, "loss": 1.1742, "step": 9149 }, { "epoch": 0.71, "learning_rate": 4.0914799778869146e-07, "loss": 1.1909, "step": 9150 }, { "epoch": 0.71, "learning_rate": 4.0894520186418677e-07, "loss": 1.2281, "step": 9151 }, { "epoch": 0.71, "learning_rate": 4.0874244329186226e-07, "loss": 1.1848, "step": 9152 }, { "epoch": 0.71, "learning_rate": 4.0853972208453026e-07, "loss": 1.157, "step": 9153 }, { "epoch": 0.71, "learning_rate": 4.083370382550029e-07, "loss": 1.1981, "step": 9154 }, { "epoch": 0.71, "learning_rate": 4.0813439181608865e-07, "loss": 1.1257, "step": 9155 }, { "epoch": 0.71, "learning_rate": 4.0793178278059394e-07, "loss": 1.2109, "step": 9156 }, { "epoch": 0.71, "learning_rate": 4.0772921116132273e-07, "loss": 1.2462, "step": 9157 }, { "epoch": 0.71, "learning_rate": 4.075266769710769e-07, "loss": 1.1975, "step": 9158 }, { "epoch": 0.71, "learning_rate": 4.0732418022265547e-07, "loss": 0.8583, "step": 9159 }, { "epoch": 0.71, "learning_rate": 4.0712172092885564e-07, "loss": 0.8355, "step": 9160 }, { "epoch": 0.71, "learning_rate": 4.069192991024718e-07, "loss": 1.2315, "step": 9161 }, { "epoch": 0.71, "learning_rate": 4.0671691475629635e-07, "loss": 1.225, "step": 9162 }, { "epoch": 0.71, "learning_rate": 4.0651456790311887e-07, "loss": 1.2374, "step": 9163 }, { "epoch": 0.71, "learning_rate": 4.06312258555727e-07, "loss": 1.1981, "step": 9164 }, { "epoch": 0.71, "learning_rate": 4.0610998672690597e-07, "loss": 1.2215, "step": 9165 }, { "epoch": 0.71, "learning_rate": 4.059077524294382e-07, "loss": 1.1592, "step": 9166 }, { "epoch": 0.71, "learning_rate": 4.057055556761043e-07, "loss": 1.2297, "step": 9167 }, { "epoch": 0.71, "learning_rate": 4.055033964796821e-07, "loss": 1.2388, "step": 9168 }, { "epoch": 0.71, "learning_rate": 4.053012748529473e-07, "loss": 1.1998, "step": 9169 }, { "epoch": 0.71, "learning_rate": 4.050991908086732e-07, "loss": 1.1733, "step": 9170 }, { "epoch": 0.71, "learning_rate": 4.048971443596304e-07, "loss": 1.193, "step": 9171 }, { "epoch": 0.71, "learning_rate": 4.0469513551858734e-07, "loss": 1.1976, "step": 9172 }, { "epoch": 0.71, "learning_rate": 4.0449316429831106e-07, "loss": 1.1581, "step": 9173 }, { "epoch": 0.71, "learning_rate": 4.042912307115641e-07, "loss": 1.2537, "step": 9174 }, { "epoch": 0.71, "learning_rate": 4.040893347711084e-07, "loss": 1.1951, "step": 9175 }, { "epoch": 0.71, "learning_rate": 4.038874764897028e-07, "loss": 1.2067, "step": 9176 }, { "epoch": 0.71, "learning_rate": 4.036856558801035e-07, "loss": 1.2432, "step": 9177 }, { "epoch": 0.71, "learning_rate": 4.0348387295506604e-07, "loss": 0.8592, "step": 9178 }, { "epoch": 0.71, "learning_rate": 4.032821277273405e-07, "loss": 1.1982, "step": 9179 }, { "epoch": 0.71, "learning_rate": 4.030804202096776e-07, "loss": 1.2005, "step": 9180 }, { "epoch": 0.71, "learning_rate": 4.028787504148242e-07, "loss": 1.2039, "step": 9181 }, { "epoch": 0.71, "learning_rate": 4.026771183555242e-07, "loss": 1.1538, "step": 9182 }, { "epoch": 0.71, "learning_rate": 4.024755240445209e-07, "loss": 1.198, "step": 9183 }, { "epoch": 0.71, "learning_rate": 4.022739674945537e-07, "loss": 1.17, "step": 9184 }, { "epoch": 0.71, "learning_rate": 4.020724487183602e-07, "loss": 1.2133, "step": 9185 }, { "epoch": 0.71, "learning_rate": 4.018709677286759e-07, "loss": 1.1935, "step": 9186 }, { "epoch": 0.71, "learning_rate": 4.0166952453823257e-07, "loss": 1.17, "step": 9187 }, { "epoch": 0.71, "learning_rate": 4.0146811915976154e-07, "loss": 1.2077, "step": 9188 }, { "epoch": 0.71, "learning_rate": 4.012667516059903e-07, "loss": 0.8679, "step": 9189 }, { "epoch": 0.71, "learning_rate": 4.010654218896448e-07, "loss": 1.1749, "step": 9190 }, { "epoch": 0.71, "learning_rate": 4.00864130023448e-07, "loss": 1.1956, "step": 9191 }, { "epoch": 0.71, "learning_rate": 4.006628760201206e-07, "loss": 0.8517, "step": 9192 }, { "epoch": 0.71, "learning_rate": 4.0046165989238125e-07, "loss": 1.17, "step": 9193 }, { "epoch": 0.71, "learning_rate": 4.002604816529457e-07, "loss": 1.1747, "step": 9194 }, { "epoch": 0.71, "learning_rate": 4.000593413145278e-07, "loss": 1.1797, "step": 9195 }, { "epoch": 0.71, "learning_rate": 3.998582388898387e-07, "loss": 1.224, "step": 9196 }, { "epoch": 0.71, "learning_rate": 3.996571743915872e-07, "loss": 0.8335, "step": 9197 }, { "epoch": 0.71, "learning_rate": 3.994561478324797e-07, "loss": 1.2017, "step": 9198 }, { "epoch": 0.71, "learning_rate": 3.9925515922522035e-07, "loss": 1.2717, "step": 9199 }, { "epoch": 0.71, "learning_rate": 3.9905420858251073e-07, "loss": 1.213, "step": 9200 }, { "epoch": 0.71, "learning_rate": 3.9885329591704995e-07, "loss": 1.1839, "step": 9201 }, { "epoch": 0.71, "learning_rate": 3.9865242124153497e-07, "loss": 1.1506, "step": 9202 }, { "epoch": 0.71, "learning_rate": 3.9845158456866023e-07, "loss": 1.1657, "step": 9203 }, { "epoch": 0.71, "learning_rate": 3.982507859111178e-07, "loss": 1.1959, "step": 9204 }, { "epoch": 0.71, "learning_rate": 3.9805002528159725e-07, "loss": 0.8145, "step": 9205 }, { "epoch": 0.71, "learning_rate": 3.978493026927854e-07, "loss": 1.1751, "step": 9206 }, { "epoch": 0.71, "learning_rate": 3.9764861815736827e-07, "loss": 1.1648, "step": 9207 }, { "epoch": 0.71, "learning_rate": 3.9744797168802713e-07, "loss": 1.2224, "step": 9208 }, { "epoch": 0.71, "learning_rate": 3.9724736329744235e-07, "loss": 1.1642, "step": 9209 }, { "epoch": 0.71, "learning_rate": 3.9704679299829147e-07, "loss": 1.172, "step": 9210 }, { "epoch": 0.71, "learning_rate": 3.9684626080324955e-07, "loss": 1.1824, "step": 9211 }, { "epoch": 0.71, "learning_rate": 3.9664576672499026e-07, "loss": 1.2084, "step": 9212 }, { "epoch": 0.72, "learning_rate": 3.9644531077618294e-07, "loss": 1.2189, "step": 9213 }, { "epoch": 0.72, "learning_rate": 3.962448929694956e-07, "loss": 1.1955, "step": 9214 }, { "epoch": 0.72, "learning_rate": 3.960445133175949e-07, "loss": 1.1948, "step": 9215 }, { "epoch": 0.72, "learning_rate": 3.9584417183314247e-07, "loss": 1.1579, "step": 9216 }, { "epoch": 0.72, "learning_rate": 3.956438685288006e-07, "loss": 1.2326, "step": 9217 }, { "epoch": 0.72, "learning_rate": 3.954436034172265e-07, "loss": 0.8633, "step": 9218 }, { "epoch": 0.72, "learning_rate": 3.9524337651107597e-07, "loss": 1.2305, "step": 9219 }, { "epoch": 0.72, "learning_rate": 3.950431878230037e-07, "loss": 1.2382, "step": 9220 }, { "epoch": 0.72, "learning_rate": 3.9484303736565937e-07, "loss": 1.209, "step": 9221 }, { "epoch": 0.72, "learning_rate": 3.946429251516927e-07, "loss": 1.189, "step": 9222 }, { "epoch": 0.72, "learning_rate": 3.944428511937495e-07, "loss": 1.1181, "step": 9223 }, { "epoch": 0.72, "learning_rate": 3.9424281550447367e-07, "loss": 1.1621, "step": 9224 }, { "epoch": 0.72, "learning_rate": 3.9404281809650676e-07, "loss": 0.8157, "step": 9225 }, { "epoch": 0.72, "learning_rate": 3.9384285898248736e-07, "loss": 1.2144, "step": 9226 }, { "epoch": 0.72, "learning_rate": 3.936429381750525e-07, "loss": 1.2117, "step": 9227 }, { "epoch": 0.72, "learning_rate": 3.934430556868361e-07, "loss": 0.8218, "step": 9228 }, { "epoch": 0.72, "learning_rate": 3.932432115304699e-07, "loss": 1.1904, "step": 9229 }, { "epoch": 0.72, "learning_rate": 3.930434057185833e-07, "loss": 1.2077, "step": 9230 }, { "epoch": 0.72, "learning_rate": 3.9284363826380304e-07, "loss": 1.1936, "step": 9231 }, { "epoch": 0.72, "learning_rate": 3.926439091787538e-07, "loss": 1.1733, "step": 9232 }, { "epoch": 0.72, "learning_rate": 3.9244421847605733e-07, "loss": 1.2286, "step": 9233 }, { "epoch": 0.72, "learning_rate": 3.922445661683336e-07, "loss": 1.1819, "step": 9234 }, { "epoch": 0.72, "learning_rate": 3.9204495226819956e-07, "loss": 1.2285, "step": 9235 }, { "epoch": 0.72, "learning_rate": 3.918453767882699e-07, "loss": 1.1959, "step": 9236 }, { "epoch": 0.72, "learning_rate": 3.9164583974115715e-07, "loss": 1.2096, "step": 9237 }, { "epoch": 0.72, "learning_rate": 3.9144634113947116e-07, "loss": 1.1401, "step": 9238 }, { "epoch": 0.72, "learning_rate": 3.912468809958194e-07, "loss": 1.1838, "step": 9239 }, { "epoch": 0.72, "learning_rate": 3.910474593228068e-07, "loss": 1.1945, "step": 9240 }, { "epoch": 0.72, "learning_rate": 3.9084807613303604e-07, "loss": 1.175, "step": 9241 }, { "epoch": 0.72, "learning_rate": 3.9064873143910736e-07, "loss": 1.2417, "step": 9242 }, { "epoch": 0.72, "learning_rate": 3.904494252536186e-07, "loss": 1.2163, "step": 9243 }, { "epoch": 0.72, "learning_rate": 3.902501575891648e-07, "loss": 1.214, "step": 9244 }, { "epoch": 0.72, "learning_rate": 3.9005092845833864e-07, "loss": 1.2133, "step": 9245 }, { "epoch": 0.72, "learning_rate": 3.898517378737316e-07, "loss": 1.1771, "step": 9246 }, { "epoch": 0.72, "learning_rate": 3.896525858479307e-07, "loss": 1.1834, "step": 9247 }, { "epoch": 0.72, "learning_rate": 3.894534723935213e-07, "loss": 1.1534, "step": 9248 }, { "epoch": 0.72, "learning_rate": 3.89254397523088e-07, "loss": 1.2014, "step": 9249 }, { "epoch": 0.72, "learning_rate": 3.8905536124920956e-07, "loss": 1.2349, "step": 9250 }, { "epoch": 0.72, "learning_rate": 3.8885636358446607e-07, "loss": 1.1465, "step": 9251 }, { "epoch": 0.72, "learning_rate": 3.8865740454143214e-07, "loss": 1.2679, "step": 9252 }, { "epoch": 0.72, "learning_rate": 3.88458484132681e-07, "loss": 1.1674, "step": 9253 }, { "epoch": 0.72, "learning_rate": 3.882596023707849e-07, "loss": 1.1872, "step": 9254 }, { "epoch": 0.72, "learning_rate": 3.880607592683108e-07, "loss": 1.2375, "step": 9255 }, { "epoch": 0.72, "learning_rate": 3.878619548378258e-07, "loss": 0.8448, "step": 9256 }, { "epoch": 0.72, "learning_rate": 3.8766318909189343e-07, "loss": 1.1989, "step": 9257 }, { "epoch": 0.72, "learning_rate": 3.8746446204307394e-07, "loss": 1.2163, "step": 9258 }, { "epoch": 0.72, "learning_rate": 3.87265773703927e-07, "loss": 1.2347, "step": 9259 }, { "epoch": 0.72, "learning_rate": 3.870671240870086e-07, "loss": 1.2288, "step": 9260 }, { "epoch": 0.72, "learning_rate": 3.868685132048726e-07, "loss": 1.2325, "step": 9261 }, { "epoch": 0.72, "learning_rate": 3.8666994107007044e-07, "loss": 1.2196, "step": 9262 }, { "epoch": 0.72, "learning_rate": 3.864714076951504e-07, "loss": 1.1782, "step": 9263 }, { "epoch": 0.72, "learning_rate": 3.8627291309265964e-07, "loss": 0.796, "step": 9264 }, { "epoch": 0.72, "learning_rate": 3.860744572751421e-07, "loss": 1.1819, "step": 9265 }, { "epoch": 0.72, "learning_rate": 3.858760402551393e-07, "loss": 1.1931, "step": 9266 }, { "epoch": 0.72, "learning_rate": 3.8567766204519025e-07, "loss": 1.1166, "step": 9267 }, { "epoch": 0.72, "learning_rate": 3.854793226578317e-07, "loss": 1.1854, "step": 9268 }, { "epoch": 0.72, "learning_rate": 3.8528102210559777e-07, "loss": 0.815, "step": 9269 }, { "epoch": 0.72, "learning_rate": 3.850827604010204e-07, "loss": 1.1796, "step": 9270 }, { "epoch": 0.72, "learning_rate": 3.8488453755662874e-07, "loss": 1.1851, "step": 9271 }, { "epoch": 0.72, "learning_rate": 3.846863535849497e-07, "loss": 1.231, "step": 9272 }, { "epoch": 0.72, "learning_rate": 3.844882084985077e-07, "loss": 1.2325, "step": 9273 }, { "epoch": 0.72, "learning_rate": 3.842901023098247e-07, "loss": 1.2133, "step": 9274 }, { "epoch": 0.72, "learning_rate": 3.8409203503142007e-07, "loss": 1.1943, "step": 9275 }, { "epoch": 0.72, "learning_rate": 3.8389400667581096e-07, "loss": 1.2277, "step": 9276 }, { "epoch": 0.72, "learning_rate": 3.836960172555118e-07, "loss": 1.212, "step": 9277 }, { "epoch": 0.72, "learning_rate": 3.834980667830348e-07, "loss": 1.2209, "step": 9278 }, { "epoch": 0.72, "learning_rate": 3.833001552708894e-07, "loss": 1.1904, "step": 9279 }, { "epoch": 0.72, "learning_rate": 3.8310228273158305e-07, "loss": 1.1878, "step": 9280 }, { "epoch": 0.72, "learning_rate": 3.8290444917762033e-07, "loss": 1.2222, "step": 9281 }, { "epoch": 0.72, "learning_rate": 3.8270665462150307e-07, "loss": 1.1872, "step": 9282 }, { "epoch": 0.72, "learning_rate": 3.8250889907573224e-07, "loss": 1.1839, "step": 9283 }, { "epoch": 0.72, "learning_rate": 3.8231118255280413e-07, "loss": 1.1995, "step": 9284 }, { "epoch": 0.72, "learning_rate": 3.821135050652138e-07, "loss": 1.2747, "step": 9285 }, { "epoch": 0.72, "learning_rate": 3.8191586662545364e-07, "loss": 1.1289, "step": 9286 }, { "epoch": 0.72, "learning_rate": 3.8171826724601333e-07, "loss": 1.2085, "step": 9287 }, { "epoch": 0.72, "learning_rate": 3.815207069393813e-07, "loss": 1.1832, "step": 9288 }, { "epoch": 0.72, "learning_rate": 3.8132318571804125e-07, "loss": 1.2, "step": 9289 }, { "epoch": 0.72, "learning_rate": 3.8112570359447647e-07, "loss": 1.209, "step": 9290 }, { "epoch": 0.72, "learning_rate": 3.8092826058116736e-07, "loss": 1.1762, "step": 9291 }, { "epoch": 0.72, "learning_rate": 3.807308566905901e-07, "loss": 1.1338, "step": 9292 }, { "epoch": 0.72, "learning_rate": 3.8053349193522135e-07, "loss": 1.1743, "step": 9293 }, { "epoch": 0.72, "learning_rate": 3.803361663275324e-07, "loss": 1.2076, "step": 9294 }, { "epoch": 0.72, "learning_rate": 3.8013887987999425e-07, "loss": 1.1403, "step": 9295 }, { "epoch": 0.72, "learning_rate": 3.7994163260507485e-07, "loss": 1.2235, "step": 9296 }, { "epoch": 0.72, "learning_rate": 3.797444245152381e-07, "loss": 1.2208, "step": 9297 }, { "epoch": 0.72, "learning_rate": 3.7954725562294787e-07, "loss": 1.1561, "step": 9298 }, { "epoch": 0.72, "learning_rate": 3.7935012594066407e-07, "loss": 1.2284, "step": 9299 }, { "epoch": 0.72, "learning_rate": 3.7915303548084457e-07, "loss": 0.7794, "step": 9300 }, { "epoch": 0.72, "learning_rate": 3.7895598425594434e-07, "loss": 1.2201, "step": 9301 }, { "epoch": 0.72, "learning_rate": 3.7875897227841646e-07, "loss": 1.1868, "step": 9302 }, { "epoch": 0.72, "learning_rate": 3.785619995607112e-07, "loss": 0.87, "step": 9303 }, { "epoch": 0.72, "learning_rate": 3.7836506611527643e-07, "loss": 0.8268, "step": 9304 }, { "epoch": 0.72, "learning_rate": 3.781681719545574e-07, "loss": 1.2034, "step": 9305 }, { "epoch": 0.72, "learning_rate": 3.779713170909972e-07, "loss": 1.2285, "step": 9306 }, { "epoch": 0.72, "learning_rate": 3.7777450153703593e-07, "loss": 0.7948, "step": 9307 }, { "epoch": 0.72, "learning_rate": 3.775777253051118e-07, "loss": 1.1118, "step": 9308 }, { "epoch": 0.72, "learning_rate": 3.7738098840766006e-07, "loss": 1.1772, "step": 9309 }, { "epoch": 0.72, "learning_rate": 3.771842908571138e-07, "loss": 1.2278, "step": 9310 }, { "epoch": 0.72, "learning_rate": 3.769876326659033e-07, "loss": 1.1685, "step": 9311 }, { "epoch": 0.72, "learning_rate": 3.7679101384645675e-07, "loss": 1.1772, "step": 9312 }, { "epoch": 0.72, "learning_rate": 3.765944344111994e-07, "loss": 1.1917, "step": 9313 }, { "epoch": 0.72, "learning_rate": 3.763978943725544e-07, "loss": 1.1821, "step": 9314 }, { "epoch": 0.72, "learning_rate": 3.7620139374294225e-07, "loss": 1.1301, "step": 9315 }, { "epoch": 0.72, "learning_rate": 3.7600493253478084e-07, "loss": 0.8519, "step": 9316 }, { "epoch": 0.72, "learning_rate": 3.758085107604858e-07, "loss": 1.2166, "step": 9317 }, { "epoch": 0.72, "learning_rate": 3.7561212843247014e-07, "loss": 1.1879, "step": 9318 }, { "epoch": 0.72, "learning_rate": 3.754157855631445e-07, "loss": 1.1764, "step": 9319 }, { "epoch": 0.72, "learning_rate": 3.752194821649167e-07, "loss": 1.1758, "step": 9320 }, { "epoch": 0.72, "learning_rate": 3.750232182501921e-07, "loss": 1.1544, "step": 9321 }, { "epoch": 0.72, "learning_rate": 3.748269938313747e-07, "loss": 1.2481, "step": 9322 }, { "epoch": 0.72, "learning_rate": 3.746308089208642e-07, "loss": 0.7961, "step": 9323 }, { "epoch": 0.72, "learning_rate": 3.7443466353105845e-07, "loss": 1.1886, "step": 9324 }, { "epoch": 0.72, "learning_rate": 3.742385576743542e-07, "loss": 1.1984, "step": 9325 }, { "epoch": 0.72, "learning_rate": 3.7404249136314305e-07, "loss": 1.2143, "step": 9326 }, { "epoch": 0.72, "learning_rate": 3.7384646460981694e-07, "loss": 0.8731, "step": 9327 }, { "epoch": 0.72, "learning_rate": 3.7365047742676294e-07, "loss": 1.183, "step": 9328 }, { "epoch": 0.72, "learning_rate": 3.7345452982636663e-07, "loss": 1.1748, "step": 9329 }, { "epoch": 0.72, "learning_rate": 3.7325862182101206e-07, "loss": 1.1915, "step": 9330 }, { "epoch": 0.72, "learning_rate": 3.7306275342307847e-07, "loss": 1.2823, "step": 9331 }, { "epoch": 0.72, "learning_rate": 3.7286692464494496e-07, "loss": 1.2306, "step": 9332 }, { "epoch": 0.72, "learning_rate": 3.7267113549898664e-07, "loss": 1.1776, "step": 9333 }, { "epoch": 0.72, "learning_rate": 3.724753859975768e-07, "loss": 1.2147, "step": 9334 }, { "epoch": 0.72, "learning_rate": 3.722796761530856e-07, "loss": 1.2213, "step": 9335 }, { "epoch": 0.72, "learning_rate": 3.720840059778814e-07, "loss": 1.2124, "step": 9336 }, { "epoch": 0.72, "learning_rate": 3.718883754843296e-07, "loss": 1.1593, "step": 9337 }, { "epoch": 0.72, "learning_rate": 3.7169278468479337e-07, "loss": 1.2119, "step": 9338 }, { "epoch": 0.72, "learning_rate": 3.71497233591633e-07, "loss": 1.1853, "step": 9339 }, { "epoch": 0.72, "learning_rate": 3.7130172221720656e-07, "loss": 1.1546, "step": 9340 }, { "epoch": 0.72, "learning_rate": 3.711062505738697e-07, "loss": 1.1614, "step": 9341 }, { "epoch": 0.73, "learning_rate": 3.709108186739753e-07, "loss": 1.1566, "step": 9342 }, { "epoch": 0.73, "learning_rate": 3.7071542652987385e-07, "loss": 1.1748, "step": 9343 }, { "epoch": 0.73, "learning_rate": 3.7052007415391315e-07, "loss": 1.2131, "step": 9344 }, { "epoch": 0.73, "learning_rate": 3.7032476155843893e-07, "loss": 1.1823, "step": 9345 }, { "epoch": 0.73, "learning_rate": 3.701294887557941e-07, "loss": 1.1539, "step": 9346 }, { "epoch": 0.73, "learning_rate": 3.699342557583188e-07, "loss": 1.1291, "step": 9347 }, { "epoch": 0.73, "learning_rate": 3.697390625783512e-07, "loss": 1.1464, "step": 9348 }, { "epoch": 0.73, "learning_rate": 3.695439092282265e-07, "loss": 1.1811, "step": 9349 }, { "epoch": 0.73, "learning_rate": 3.693487957202778e-07, "loss": 1.1993, "step": 9350 }, { "epoch": 0.73, "learning_rate": 3.6915372206683524e-07, "loss": 1.1763, "step": 9351 }, { "epoch": 0.73, "learning_rate": 3.689586882802268e-07, "loss": 1.2266, "step": 9352 }, { "epoch": 0.73, "learning_rate": 3.6876369437277765e-07, "loss": 1.2108, "step": 9353 }, { "epoch": 0.73, "learning_rate": 3.6856874035681073e-07, "loss": 1.1667, "step": 9354 }, { "epoch": 0.73, "learning_rate": 3.6837382624464595e-07, "loss": 1.2336, "step": 9355 }, { "epoch": 0.73, "learning_rate": 3.681789520486021e-07, "loss": 1.2356, "step": 9356 }, { "epoch": 0.73, "learning_rate": 3.679841177809931e-07, "loss": 1.1869, "step": 9357 }, { "epoch": 0.73, "learning_rate": 3.677893234541322e-07, "loss": 1.1887, "step": 9358 }, { "epoch": 0.73, "learning_rate": 3.6759456908033004e-07, "loss": 0.8309, "step": 9359 }, { "epoch": 0.73, "learning_rate": 3.673998546718933e-07, "loss": 1.1763, "step": 9360 }, { "epoch": 0.73, "learning_rate": 3.6720518024112837e-07, "loss": 1.1813, "step": 9361 }, { "epoch": 0.73, "learning_rate": 3.6701054580033673e-07, "loss": 1.1593, "step": 9362 }, { "epoch": 0.73, "learning_rate": 3.668159513618186e-07, "loss": 1.1406, "step": 9363 }, { "epoch": 0.73, "learning_rate": 3.6662139693787254e-07, "loss": 0.7993, "step": 9364 }, { "epoch": 0.73, "learning_rate": 3.664268825407921e-07, "loss": 1.1987, "step": 9365 }, { "epoch": 0.73, "learning_rate": 3.6623240818287094e-07, "loss": 1.2114, "step": 9366 }, { "epoch": 0.73, "learning_rate": 3.660379738763989e-07, "loss": 1.2145, "step": 9367 }, { "epoch": 0.73, "learning_rate": 3.658435796336624e-07, "loss": 0.8442, "step": 9368 }, { "epoch": 0.73, "learning_rate": 3.6564922546694776e-07, "loss": 1.175, "step": 9369 }, { "epoch": 0.73, "learning_rate": 3.6545491138853603e-07, "loss": 1.2537, "step": 9370 }, { "epoch": 0.73, "learning_rate": 3.652606374107079e-07, "loss": 1.1895, "step": 9371 }, { "epoch": 0.73, "learning_rate": 3.650664035457408e-07, "loss": 1.1797, "step": 9372 }, { "epoch": 0.73, "learning_rate": 3.6487220980590847e-07, "loss": 1.1644, "step": 9373 }, { "epoch": 0.73, "learning_rate": 3.6467805620348415e-07, "loss": 0.8506, "step": 9374 }, { "epoch": 0.73, "learning_rate": 3.64483942750737e-07, "loss": 1.1971, "step": 9375 }, { "epoch": 0.73, "learning_rate": 3.6428986945993443e-07, "loss": 1.1965, "step": 9376 }, { "epoch": 0.73, "learning_rate": 3.6409583634334107e-07, "loss": 1.1891, "step": 9377 }, { "epoch": 0.73, "learning_rate": 3.639018434132188e-07, "loss": 1.2012, "step": 9378 }, { "epoch": 0.73, "learning_rate": 3.6370789068182726e-07, "loss": 1.2132, "step": 9379 }, { "epoch": 0.73, "learning_rate": 3.635139781614235e-07, "loss": 1.177, "step": 9380 }, { "epoch": 0.73, "learning_rate": 3.6332010586426187e-07, "loss": 1.1418, "step": 9381 }, { "epoch": 0.73, "learning_rate": 3.631262738025943e-07, "loss": 1.1847, "step": 9382 }, { "epoch": 0.73, "learning_rate": 3.6293248198867023e-07, "loss": 1.1664, "step": 9383 }, { "epoch": 0.73, "learning_rate": 3.6273873043473646e-07, "loss": 1.1538, "step": 9384 }, { "epoch": 0.73, "learning_rate": 3.6254501915303727e-07, "loss": 1.1611, "step": 9385 }, { "epoch": 0.73, "learning_rate": 3.6235134815581434e-07, "loss": 0.8695, "step": 9386 }, { "epoch": 0.73, "learning_rate": 3.62157717455307e-07, "loss": 1.2099, "step": 9387 }, { "epoch": 0.73, "learning_rate": 3.619641270637518e-07, "loss": 1.1476, "step": 9388 }, { "epoch": 0.73, "learning_rate": 3.617705769933829e-07, "loss": 1.2092, "step": 9389 }, { "epoch": 0.73, "learning_rate": 3.6157706725643164e-07, "loss": 1.1635, "step": 9390 }, { "epoch": 0.73, "learning_rate": 3.613835978651273e-07, "loss": 0.8642, "step": 9391 }, { "epoch": 0.73, "learning_rate": 3.611901688316963e-07, "loss": 1.1919, "step": 9392 }, { "epoch": 0.73, "learning_rate": 3.6099678016836234e-07, "loss": 1.1549, "step": 9393 }, { "epoch": 0.73, "learning_rate": 3.6080343188734696e-07, "loss": 1.1799, "step": 9394 }, { "epoch": 0.73, "learning_rate": 3.6061012400086887e-07, "loss": 1.1925, "step": 9395 }, { "epoch": 0.73, "learning_rate": 3.604168565211443e-07, "loss": 0.8275, "step": 9396 }, { "epoch": 0.73, "learning_rate": 3.6022362946038667e-07, "loss": 1.1584, "step": 9397 }, { "epoch": 0.73, "learning_rate": 3.600304428308082e-07, "loss": 1.1376, "step": 9398 }, { "epoch": 0.73, "learning_rate": 3.598372966446159e-07, "loss": 1.2021, "step": 9399 }, { "epoch": 0.73, "learning_rate": 3.5964419091401696e-07, "loss": 1.1372, "step": 9400 }, { "epoch": 0.73, "learning_rate": 3.5945112565121493e-07, "loss": 1.1911, "step": 9401 }, { "epoch": 0.73, "learning_rate": 3.5925810086840956e-07, "loss": 1.1847, "step": 9402 }, { "epoch": 0.73, "learning_rate": 3.590651165778006e-07, "loss": 1.1742, "step": 9403 }, { "epoch": 0.73, "learning_rate": 3.5887217279158244e-07, "loss": 1.2173, "step": 9404 }, { "epoch": 0.73, "learning_rate": 3.586792695219495e-07, "loss": 0.836, "step": 9405 }, { "epoch": 0.73, "learning_rate": 3.584864067810922e-07, "loss": 1.2086, "step": 9406 }, { "epoch": 0.73, "learning_rate": 3.5829358458119795e-07, "loss": 1.1829, "step": 9407 }, { "epoch": 0.73, "learning_rate": 3.581008029344531e-07, "loss": 1.2916, "step": 9408 }, { "epoch": 0.73, "learning_rate": 3.5790806185304033e-07, "loss": 1.2412, "step": 9409 }, { "epoch": 0.73, "learning_rate": 3.5771536134914016e-07, "loss": 1.168, "step": 9410 }, { "epoch": 0.73, "learning_rate": 3.5752270143493023e-07, "loss": 1.2266, "step": 9411 }, { "epoch": 0.73, "learning_rate": 3.5733008212258617e-07, "loss": 1.192, "step": 9412 }, { "epoch": 0.73, "learning_rate": 3.571375034242804e-07, "loss": 1.2097, "step": 9413 }, { "epoch": 0.73, "learning_rate": 3.569449653521833e-07, "loss": 1.1926, "step": 9414 }, { "epoch": 0.73, "learning_rate": 3.5675246791846235e-07, "loss": 1.1949, "step": 9415 }, { "epoch": 0.73, "learning_rate": 3.5656001113528265e-07, "loss": 1.1911, "step": 9416 }, { "epoch": 0.73, "learning_rate": 3.5636759501480663e-07, "loss": 1.1725, "step": 9417 }, { "epoch": 0.73, "learning_rate": 3.5617521956919415e-07, "loss": 1.2076, "step": 9418 }, { "epoch": 0.73, "learning_rate": 3.5598288481060256e-07, "loss": 1.2248, "step": 9419 }, { "epoch": 0.73, "learning_rate": 3.557905907511867e-07, "loss": 1.1856, "step": 9420 }, { "epoch": 0.73, "learning_rate": 3.555983374030985e-07, "loss": 1.21, "step": 9421 }, { "epoch": 0.73, "learning_rate": 3.5540612477848776e-07, "loss": 1.2481, "step": 9422 }, { "epoch": 0.73, "learning_rate": 3.552139528895015e-07, "loss": 1.1548, "step": 9423 }, { "epoch": 0.73, "learning_rate": 3.550218217482841e-07, "loss": 1.2561, "step": 9424 }, { "epoch": 0.73, "learning_rate": 3.548297313669776e-07, "loss": 1.1822, "step": 9425 }, { "epoch": 0.73, "learning_rate": 3.54637681757721e-07, "loss": 1.1947, "step": 9426 }, { "epoch": 0.73, "learning_rate": 3.544456729326514e-07, "loss": 1.212, "step": 9427 }, { "epoch": 0.73, "learning_rate": 3.542537049039026e-07, "loss": 1.2083, "step": 9428 }, { "epoch": 0.73, "learning_rate": 3.5406177768360643e-07, "loss": 0.7883, "step": 9429 }, { "epoch": 0.73, "learning_rate": 3.538698912838918e-07, "loss": 1.1947, "step": 9430 }, { "epoch": 0.73, "learning_rate": 3.536780457168846e-07, "loss": 1.2147, "step": 9431 }, { "epoch": 0.73, "learning_rate": 3.534862409947101e-07, "loss": 1.2356, "step": 9432 }, { "epoch": 0.73, "learning_rate": 3.5329447712948825e-07, "loss": 1.2435, "step": 9433 }, { "epoch": 0.73, "learning_rate": 3.531027541333377e-07, "loss": 1.2226, "step": 9434 }, { "epoch": 0.73, "learning_rate": 3.5291107201837556e-07, "loss": 1.1985, "step": 9435 }, { "epoch": 0.73, "learning_rate": 3.5271943079671416e-07, "loss": 1.2173, "step": 9436 }, { "epoch": 0.73, "learning_rate": 3.5252783048046563e-07, "loss": 1.1877, "step": 9437 }, { "epoch": 0.73, "learning_rate": 3.5233627108173727e-07, "loss": 1.2062, "step": 9438 }, { "epoch": 0.73, "learning_rate": 3.5214475261263485e-07, "loss": 1.1686, "step": 9439 }, { "epoch": 0.73, "learning_rate": 3.5195327508526265e-07, "loss": 1.2015, "step": 9440 }, { "epoch": 0.73, "learning_rate": 3.517618385117197e-07, "loss": 1.2024, "step": 9441 }, { "epoch": 0.73, "learning_rate": 3.515704429041052e-07, "loss": 1.2107, "step": 9442 }, { "epoch": 0.73, "learning_rate": 3.513790882745141e-07, "loss": 1.2231, "step": 9443 }, { "epoch": 0.73, "learning_rate": 3.511877746350392e-07, "loss": 1.1932, "step": 9444 }, { "epoch": 0.73, "learning_rate": 3.509965019977712e-07, "loss": 1.2187, "step": 9445 }, { "epoch": 0.73, "learning_rate": 3.5080527037479646e-07, "loss": 1.18, "step": 9446 }, { "epoch": 0.73, "learning_rate": 3.506140797782012e-07, "loss": 1.1851, "step": 9447 }, { "epoch": 0.73, "learning_rate": 3.5042293022006765e-07, "loss": 1.2316, "step": 9448 }, { "epoch": 0.73, "learning_rate": 3.502318217124753e-07, "loss": 1.2184, "step": 9449 }, { "epoch": 0.73, "learning_rate": 3.500407542675017e-07, "loss": 0.7998, "step": 9450 }, { "epoch": 0.73, "learning_rate": 3.4984972789722144e-07, "loss": 1.2363, "step": 9451 }, { "epoch": 0.73, "learning_rate": 3.496587426137065e-07, "loss": 1.1587, "step": 9452 }, { "epoch": 0.73, "learning_rate": 3.494677984290264e-07, "loss": 1.2043, "step": 9453 }, { "epoch": 0.73, "learning_rate": 3.492768953552481e-07, "loss": 1.2186, "step": 9454 }, { "epoch": 0.73, "learning_rate": 3.4908603340443575e-07, "loss": 1.1581, "step": 9455 }, { "epoch": 0.73, "learning_rate": 3.4889521258865106e-07, "loss": 1.231, "step": 9456 }, { "epoch": 0.73, "learning_rate": 3.487044329199531e-07, "loss": 1.2227, "step": 9457 }, { "epoch": 0.73, "learning_rate": 3.4851369441039836e-07, "loss": 1.1865, "step": 9458 }, { "epoch": 0.73, "learning_rate": 3.483229970720406e-07, "loss": 0.8696, "step": 9459 }, { "epoch": 0.73, "learning_rate": 3.481323409169313e-07, "loss": 1.1668, "step": 9460 }, { "epoch": 0.73, "learning_rate": 3.47941725957119e-07, "loss": 1.1622, "step": 9461 }, { "epoch": 0.73, "learning_rate": 3.477511522046498e-07, "loss": 1.1908, "step": 9462 }, { "epoch": 0.73, "learning_rate": 3.4756061967156715e-07, "loss": 1.2321, "step": 9463 }, { "epoch": 0.73, "learning_rate": 3.4737012836991195e-07, "loss": 1.1375, "step": 9464 }, { "epoch": 0.73, "learning_rate": 3.47179678311722e-07, "loss": 1.1607, "step": 9465 }, { "epoch": 0.73, "learning_rate": 3.4698926950903393e-07, "loss": 1.1846, "step": 9466 }, { "epoch": 0.73, "learning_rate": 3.4679890197388004e-07, "loss": 1.1934, "step": 9467 }, { "epoch": 0.73, "learning_rate": 3.4660857571829085e-07, "loss": 1.1981, "step": 9468 }, { "epoch": 0.73, "learning_rate": 3.4641829075429416e-07, "loss": 1.2097, "step": 9469 }, { "epoch": 0.73, "learning_rate": 3.4622804709391494e-07, "loss": 1.2365, "step": 9470 }, { "epoch": 0.74, "learning_rate": 3.460378447491769e-07, "loss": 1.1678, "step": 9471 }, { "epoch": 0.74, "learning_rate": 3.4584768373209884e-07, "loss": 1.1836, "step": 9472 }, { "epoch": 0.74, "learning_rate": 3.4565756405469815e-07, "loss": 1.1394, "step": 9473 }, { "epoch": 0.74, "learning_rate": 3.4546748572899067e-07, "loss": 1.2184, "step": 9474 }, { "epoch": 0.74, "learning_rate": 3.4527744876698716e-07, "loss": 1.232, "step": 9475 }, { "epoch": 0.74, "learning_rate": 3.4508745318069833e-07, "loss": 0.8623, "step": 9476 }, { "epoch": 0.74, "learning_rate": 3.4489749898213084e-07, "loss": 1.219, "step": 9477 }, { "epoch": 0.74, "learning_rate": 3.447075861832882e-07, "loss": 1.145, "step": 9478 }, { "epoch": 0.74, "learning_rate": 3.445177147961733e-07, "loss": 1.2246, "step": 9479 }, { "epoch": 0.74, "learning_rate": 3.4432788483278396e-07, "loss": 1.187, "step": 9480 }, { "epoch": 0.74, "learning_rate": 3.441380963051176e-07, "loss": 1.2369, "step": 9481 }, { "epoch": 0.74, "learning_rate": 3.4394834922516804e-07, "loss": 1.1582, "step": 9482 }, { "epoch": 0.74, "learning_rate": 3.437586436049256e-07, "loss": 1.2212, "step": 9483 }, { "epoch": 0.74, "learning_rate": 3.4356897945637976e-07, "loss": 1.1578, "step": 9484 }, { "epoch": 0.74, "learning_rate": 3.4337935679151633e-07, "loss": 1.1918, "step": 9485 }, { "epoch": 0.74, "learning_rate": 3.431897756223183e-07, "loss": 0.8275, "step": 9486 }, { "epoch": 0.74, "learning_rate": 3.4300023596076686e-07, "loss": 1.2043, "step": 9487 }, { "epoch": 0.74, "learning_rate": 3.4281073781883984e-07, "loss": 1.1972, "step": 9488 }, { "epoch": 0.74, "learning_rate": 3.4262128120851285e-07, "loss": 1.2044, "step": 9489 }, { "epoch": 0.74, "learning_rate": 3.4243186614175866e-07, "loss": 1.1759, "step": 9490 }, { "epoch": 0.74, "learning_rate": 3.422424926305475e-07, "loss": 1.1579, "step": 9491 }, { "epoch": 0.74, "learning_rate": 3.420531606868471e-07, "loss": 1.1664, "step": 9492 }, { "epoch": 0.74, "learning_rate": 3.4186387032262233e-07, "loss": 1.1966, "step": 9493 }, { "epoch": 0.74, "learning_rate": 3.416746215498356e-07, "loss": 1.2236, "step": 9494 }, { "epoch": 0.74, "learning_rate": 3.4148541438044655e-07, "loss": 1.1871, "step": 9495 }, { "epoch": 0.74, "learning_rate": 3.4129624882641237e-07, "loss": 1.1858, "step": 9496 }, { "epoch": 0.74, "learning_rate": 3.4110712489968765e-07, "loss": 1.2076, "step": 9497 }, { "epoch": 0.74, "learning_rate": 3.409180426122239e-07, "loss": 1.2221, "step": 9498 }, { "epoch": 0.74, "learning_rate": 3.4072900197597067e-07, "loss": 1.2003, "step": 9499 }, { "epoch": 0.74, "learning_rate": 3.405400030028743e-07, "loss": 0.8473, "step": 9500 }, { "epoch": 0.74, "learning_rate": 3.4035104570487883e-07, "loss": 1.2094, "step": 9501 }, { "epoch": 0.74, "learning_rate": 3.4016213009392547e-07, "loss": 1.1672, "step": 9502 }, { "epoch": 0.74, "learning_rate": 3.3997325618195305e-07, "loss": 1.1939, "step": 9503 }, { "epoch": 0.74, "learning_rate": 3.3978442398089744e-07, "loss": 1.2019, "step": 9504 }, { "epoch": 0.74, "learning_rate": 3.395956335026922e-07, "loss": 0.8713, "step": 9505 }, { "epoch": 0.74, "learning_rate": 3.39406884759268e-07, "loss": 1.1901, "step": 9506 }, { "epoch": 0.74, "learning_rate": 3.392181777625527e-07, "loss": 1.1886, "step": 9507 }, { "epoch": 0.74, "learning_rate": 3.3902951252447265e-07, "loss": 1.1756, "step": 9508 }, { "epoch": 0.74, "learning_rate": 3.388408890569495e-07, "loss": 1.1925, "step": 9509 }, { "epoch": 0.74, "learning_rate": 3.386523073719044e-07, "loss": 1.2101, "step": 9510 }, { "epoch": 0.74, "learning_rate": 3.3846376748125504e-07, "loss": 1.2184, "step": 9511 }, { "epoch": 0.74, "learning_rate": 3.382752693969151e-07, "loss": 1.2165, "step": 9512 }, { "epoch": 0.74, "learning_rate": 3.3808681313079847e-07, "loss": 1.1818, "step": 9513 }, { "epoch": 0.74, "learning_rate": 3.378983986948133e-07, "loss": 1.1818, "step": 9514 }, { "epoch": 0.74, "learning_rate": 3.377100261008676e-07, "loss": 1.1455, "step": 9515 }, { "epoch": 0.74, "learning_rate": 3.375216953608658e-07, "loss": 1.21, "step": 9516 }, { "epoch": 0.74, "learning_rate": 3.373334064867086e-07, "loss": 1.1762, "step": 9517 }, { "epoch": 0.74, "learning_rate": 3.37145159490296e-07, "loss": 1.1524, "step": 9518 }, { "epoch": 0.74, "learning_rate": 3.369569543835241e-07, "loss": 1.1881, "step": 9519 }, { "epoch": 0.74, "learning_rate": 3.367687911782867e-07, "loss": 1.1689, "step": 9520 }, { "epoch": 0.74, "learning_rate": 3.3658066988647525e-07, "loss": 1.1898, "step": 9521 }, { "epoch": 0.74, "learning_rate": 3.3639259051997726e-07, "loss": 1.1722, "step": 9522 }, { "epoch": 0.74, "learning_rate": 3.3620455309067954e-07, "loss": 1.2036, "step": 9523 }, { "epoch": 0.74, "learning_rate": 3.3601655761046487e-07, "loss": 0.8358, "step": 9524 }, { "epoch": 0.74, "learning_rate": 3.3582860409121383e-07, "loss": 1.1945, "step": 9525 }, { "epoch": 0.74, "learning_rate": 3.356406925448043e-07, "loss": 1.1789, "step": 9526 }, { "epoch": 0.74, "learning_rate": 3.354528229831114e-07, "loss": 1.188, "step": 9527 }, { "epoch": 0.74, "learning_rate": 3.352649954180079e-07, "loss": 1.1333, "step": 9528 }, { "epoch": 0.74, "learning_rate": 3.3507720986136347e-07, "loss": 1.2056, "step": 9529 }, { "epoch": 0.74, "learning_rate": 3.3488946632504553e-07, "loss": 1.2166, "step": 9530 }, { "epoch": 0.74, "learning_rate": 3.347017648209187e-07, "loss": 1.2125, "step": 9531 }, { "epoch": 0.74, "learning_rate": 3.345141053608448e-07, "loss": 0.8822, "step": 9532 }, { "epoch": 0.74, "learning_rate": 3.343264879566833e-07, "loss": 1.1603, "step": 9533 }, { "epoch": 0.74, "learning_rate": 3.341389126202906e-07, "loss": 1.2056, "step": 9534 }, { "epoch": 0.74, "learning_rate": 3.339513793635208e-07, "loss": 1.1824, "step": 9535 }, { "epoch": 0.74, "learning_rate": 3.337638881982252e-07, "loss": 1.1921, "step": 9536 }, { "epoch": 0.74, "learning_rate": 3.3357643913625255e-07, "loss": 1.1664, "step": 9537 }, { "epoch": 0.74, "learning_rate": 3.333890321894487e-07, "loss": 1.1767, "step": 9538 }, { "epoch": 0.74, "learning_rate": 3.3320166736965704e-07, "loss": 1.1339, "step": 9539 }, { "epoch": 0.74, "learning_rate": 3.330143446887181e-07, "loss": 1.2072, "step": 9540 }, { "epoch": 0.74, "learning_rate": 3.3282706415846995e-07, "loss": 1.2081, "step": 9541 }, { "epoch": 0.74, "learning_rate": 3.3263982579074845e-07, "loss": 0.8214, "step": 9542 }, { "epoch": 0.74, "learning_rate": 3.324526295973856e-07, "loss": 1.1847, "step": 9543 }, { "epoch": 0.74, "learning_rate": 3.3226547559021165e-07, "loss": 1.2437, "step": 9544 }, { "epoch": 0.74, "learning_rate": 3.3207836378105394e-07, "loss": 1.1832, "step": 9545 }, { "epoch": 0.74, "learning_rate": 3.3189129418173677e-07, "loss": 1.1928, "step": 9546 }, { "epoch": 0.74, "learning_rate": 3.3170426680408314e-07, "loss": 1.2053, "step": 9547 }, { "epoch": 0.74, "learning_rate": 3.3151728165991154e-07, "loss": 1.2097, "step": 9548 }, { "epoch": 0.74, "learning_rate": 3.3133033876103856e-07, "loss": 1.2311, "step": 9549 }, { "epoch": 0.74, "learning_rate": 3.3114343811927925e-07, "loss": 1.1739, "step": 9550 }, { "epoch": 0.74, "learning_rate": 3.309565797464435e-07, "loss": 1.1943, "step": 9551 }, { "epoch": 0.74, "learning_rate": 3.307697636543411e-07, "loss": 1.2262, "step": 9552 }, { "epoch": 0.74, "learning_rate": 3.305829898547776e-07, "loss": 1.206, "step": 9553 }, { "epoch": 0.74, "learning_rate": 3.303962583595563e-07, "loss": 1.2103, "step": 9554 }, { "epoch": 0.74, "learning_rate": 3.302095691804784e-07, "loss": 1.1901, "step": 9555 }, { "epoch": 0.74, "learning_rate": 3.300229223293406e-07, "loss": 1.1901, "step": 9556 }, { "epoch": 0.74, "learning_rate": 3.298363178179394e-07, "loss": 1.2471, "step": 9557 }, { "epoch": 0.74, "learning_rate": 3.2964975565806694e-07, "loss": 1.1904, "step": 9558 }, { "epoch": 0.74, "learning_rate": 3.2946323586151326e-07, "loss": 1.1477, "step": 9559 }, { "epoch": 0.74, "learning_rate": 3.2927675844006563e-07, "loss": 0.775, "step": 9560 }, { "epoch": 0.74, "learning_rate": 3.2909032340550855e-07, "loss": 1.2311, "step": 9561 }, { "epoch": 0.74, "learning_rate": 3.28903930769624e-07, "loss": 0.7843, "step": 9562 }, { "epoch": 0.74, "learning_rate": 3.287175805441912e-07, "loss": 1.2036, "step": 9563 }, { "epoch": 0.74, "learning_rate": 3.285312727409867e-07, "loss": 1.2067, "step": 9564 }, { "epoch": 0.74, "learning_rate": 3.283450073717844e-07, "loss": 0.8577, "step": 9565 }, { "epoch": 0.74, "learning_rate": 3.2815878444835543e-07, "loss": 1.1906, "step": 9566 }, { "epoch": 0.74, "learning_rate": 3.279726039824684e-07, "loss": 1.2019, "step": 9567 }, { "epoch": 0.74, "learning_rate": 3.27786465985889e-07, "loss": 1.2103, "step": 9568 }, { "epoch": 0.74, "learning_rate": 3.276003704703804e-07, "loss": 1.2313, "step": 9569 }, { "epoch": 0.74, "learning_rate": 3.274143174477031e-07, "loss": 1.1559, "step": 9570 }, { "epoch": 0.74, "learning_rate": 3.272283069296148e-07, "loss": 1.1764, "step": 9571 }, { "epoch": 0.74, "learning_rate": 3.270423389278707e-07, "loss": 0.8416, "step": 9572 }, { "epoch": 0.74, "learning_rate": 3.2685641345422313e-07, "loss": 0.8096, "step": 9573 }, { "epoch": 0.74, "learning_rate": 3.266705305204218e-07, "loss": 1.2631, "step": 9574 }, { "epoch": 0.74, "learning_rate": 3.2648469013821335e-07, "loss": 1.2277, "step": 9575 }, { "epoch": 0.74, "learning_rate": 3.262988923193433e-07, "loss": 1.1909, "step": 9576 }, { "epoch": 0.74, "learning_rate": 3.26113137075552e-07, "loss": 1.2088, "step": 9577 }, { "epoch": 0.74, "learning_rate": 3.259274244185789e-07, "loss": 1.1729, "step": 9578 }, { "epoch": 0.74, "learning_rate": 3.257417543601603e-07, "loss": 1.2137, "step": 9579 }, { "epoch": 0.74, "learning_rate": 3.2555612691202927e-07, "loss": 1.1719, "step": 9580 }, { "epoch": 0.74, "learning_rate": 3.2537054208591796e-07, "loss": 1.1662, "step": 9581 }, { "epoch": 0.74, "learning_rate": 3.251849998935532e-07, "loss": 1.2141, "step": 9582 }, { "epoch": 0.74, "learning_rate": 3.249995003466608e-07, "loss": 1.2298, "step": 9583 }, { "epoch": 0.74, "learning_rate": 3.248140434569644e-07, "loss": 1.2217, "step": 9584 }, { "epoch": 0.74, "learning_rate": 3.246286292361827e-07, "loss": 0.8445, "step": 9585 }, { "epoch": 0.74, "learning_rate": 3.244432576960342e-07, "loss": 1.2308, "step": 9586 }, { "epoch": 0.74, "learning_rate": 3.242579288482337e-07, "loss": 0.8329, "step": 9587 }, { "epoch": 0.74, "learning_rate": 3.240726427044919e-07, "loss": 1.2074, "step": 9588 }, { "epoch": 0.74, "learning_rate": 3.2388739927651963e-07, "loss": 1.1654, "step": 9589 }, { "epoch": 0.74, "learning_rate": 3.237021985760223e-07, "loss": 1.1714, "step": 9590 }, { "epoch": 0.74, "learning_rate": 3.2351704061470455e-07, "loss": 0.8477, "step": 9591 }, { "epoch": 0.74, "learning_rate": 3.233319254042678e-07, "loss": 1.2079, "step": 9592 }, { "epoch": 0.74, "learning_rate": 3.231468529564094e-07, "loss": 1.1793, "step": 9593 }, { "epoch": 0.74, "learning_rate": 3.229618232828262e-07, "loss": 1.2045, "step": 9594 }, { "epoch": 0.74, "learning_rate": 3.22776836395211e-07, "loss": 1.2137, "step": 9595 }, { "epoch": 0.74, "learning_rate": 3.2259189230525417e-07, "loss": 1.2176, "step": 9596 }, { "epoch": 0.74, "learning_rate": 3.224069910246433e-07, "loss": 1.2137, "step": 9597 }, { "epoch": 0.74, "learning_rate": 3.222221325650636e-07, "loss": 1.1958, "step": 9598 }, { "epoch": 0.75, "learning_rate": 3.2203731693819714e-07, "loss": 1.2124, "step": 9599 }, { "epoch": 0.75, "learning_rate": 3.218525441557236e-07, "loss": 1.1635, "step": 9600 }, { "epoch": 0.75, "learning_rate": 3.216678142293199e-07, "loss": 1.1903, "step": 9601 }, { "epoch": 0.75, "learning_rate": 3.2148312717065994e-07, "loss": 1.2025, "step": 9602 }, { "epoch": 0.75, "learning_rate": 3.212984829914155e-07, "loss": 1.1997, "step": 9603 }, { "epoch": 0.75, "learning_rate": 3.21113881703255e-07, "loss": 1.171, "step": 9604 }, { "epoch": 0.75, "learning_rate": 3.209293233178447e-07, "loss": 1.1805, "step": 9605 }, { "epoch": 0.75, "learning_rate": 3.207448078468479e-07, "loss": 1.1819, "step": 9606 }, { "epoch": 0.75, "learning_rate": 3.2056033530192506e-07, "loss": 0.7996, "step": 9607 }, { "epoch": 0.75, "learning_rate": 3.203759056947342e-07, "loss": 1.1952, "step": 9608 }, { "epoch": 0.75, "learning_rate": 3.2019151903693033e-07, "loss": 1.2028, "step": 9609 }, { "epoch": 0.75, "learning_rate": 3.200071753401661e-07, "loss": 1.2183, "step": 9610 }, { "epoch": 0.75, "learning_rate": 3.198228746160911e-07, "loss": 1.1435, "step": 9611 }, { "epoch": 0.75, "learning_rate": 3.196386168763525e-07, "loss": 1.2113, "step": 9612 }, { "epoch": 0.75, "learning_rate": 3.1945440213259454e-07, "loss": 1.1348, "step": 9613 }, { "epoch": 0.75, "learning_rate": 3.192702303964585e-07, "loss": 1.2014, "step": 9614 }, { "epoch": 0.75, "learning_rate": 3.190861016795843e-07, "loss": 1.1754, "step": 9615 }, { "epoch": 0.75, "learning_rate": 3.18902015993607e-07, "loss": 1.1939, "step": 9616 }, { "epoch": 0.75, "learning_rate": 3.1871797335016013e-07, "loss": 1.2075, "step": 9617 }, { "epoch": 0.75, "learning_rate": 3.1853397376087545e-07, "loss": 1.1531, "step": 9618 }, { "epoch": 0.75, "learning_rate": 3.1835001723737945e-07, "loss": 1.1703, "step": 9619 }, { "epoch": 0.75, "learning_rate": 3.181661037912989e-07, "loss": 1.2078, "step": 9620 }, { "epoch": 0.75, "learning_rate": 3.1798223343425536e-07, "loss": 1.1864, "step": 9621 }, { "epoch": 0.75, "learning_rate": 3.177984061778686e-07, "loss": 1.2029, "step": 9622 }, { "epoch": 0.75, "learning_rate": 3.176146220337567e-07, "loss": 1.1974, "step": 9623 }, { "epoch": 0.75, "learning_rate": 3.1743088101353275e-07, "loss": 1.1931, "step": 9624 }, { "epoch": 0.75, "learning_rate": 3.172471831288095e-07, "loss": 0.8147, "step": 9625 }, { "epoch": 0.75, "learning_rate": 3.1706352839119565e-07, "loss": 1.1975, "step": 9626 }, { "epoch": 0.75, "learning_rate": 3.1687991681229674e-07, "loss": 1.1908, "step": 9627 }, { "epoch": 0.75, "learning_rate": 3.1669634840371693e-07, "loss": 1.1577, "step": 9628 }, { "epoch": 0.75, "learning_rate": 3.165128231770569e-07, "loss": 1.2211, "step": 9629 }, { "epoch": 0.75, "learning_rate": 3.163293411439145e-07, "loss": 1.1993, "step": 9630 }, { "epoch": 0.75, "learning_rate": 3.161459023158853e-07, "loss": 1.2423, "step": 9631 }, { "epoch": 0.75, "learning_rate": 3.159625067045612e-07, "loss": 1.164, "step": 9632 }, { "epoch": 0.75, "learning_rate": 3.157791543215327e-07, "loss": 1.1946, "step": 9633 }, { "epoch": 0.75, "learning_rate": 3.155958451783866e-07, "loss": 1.1706, "step": 9634 }, { "epoch": 0.75, "learning_rate": 3.154125792867074e-07, "loss": 1.217, "step": 9635 }, { "epoch": 0.75, "learning_rate": 3.152293566580767e-07, "loss": 1.1881, "step": 9636 }, { "epoch": 0.75, "learning_rate": 3.1504617730407335e-07, "loss": 1.2198, "step": 9637 }, { "epoch": 0.75, "learning_rate": 3.148630412362735e-07, "loss": 1.2155, "step": 9638 }, { "epoch": 0.75, "learning_rate": 3.146799484662507e-07, "loss": 0.7827, "step": 9639 }, { "epoch": 0.75, "learning_rate": 3.1449689900557553e-07, "loss": 1.2169, "step": 9640 }, { "epoch": 0.75, "learning_rate": 3.143138928658159e-07, "loss": 1.1847, "step": 9641 }, { "epoch": 0.75, "learning_rate": 3.141309300585372e-07, "loss": 1.1708, "step": 9642 }, { "epoch": 0.75, "learning_rate": 3.1394801059530186e-07, "loss": 1.1453, "step": 9643 }, { "epoch": 0.75, "learning_rate": 3.137651344876695e-07, "loss": 1.1751, "step": 9644 }, { "epoch": 0.75, "learning_rate": 3.1358230174719727e-07, "loss": 1.1841, "step": 9645 }, { "epoch": 0.75, "learning_rate": 3.133995123854393e-07, "loss": 1.2173, "step": 9646 }, { "epoch": 0.75, "learning_rate": 3.1321676641394714e-07, "loss": 1.1981, "step": 9647 }, { "epoch": 0.75, "learning_rate": 3.130340638442697e-07, "loss": 1.1976, "step": 9648 }, { "epoch": 0.75, "learning_rate": 3.128514046879528e-07, "loss": 1.2302, "step": 9649 }, { "epoch": 0.75, "learning_rate": 3.126687889565399e-07, "loss": 1.2062, "step": 9650 }, { "epoch": 0.75, "learning_rate": 3.1248621666157127e-07, "loss": 0.8429, "step": 9651 }, { "epoch": 0.75, "learning_rate": 3.123036878145855e-07, "loss": 1.1995, "step": 9652 }, { "epoch": 0.75, "learning_rate": 3.121212024271167e-07, "loss": 1.1439, "step": 9653 }, { "epoch": 0.75, "learning_rate": 3.119387605106977e-07, "loss": 1.1906, "step": 9654 }, { "epoch": 0.75, "learning_rate": 3.117563620768577e-07, "loss": 1.2084, "step": 9655 }, { "epoch": 0.75, "learning_rate": 3.115740071371237e-07, "loss": 1.2098, "step": 9656 }, { "epoch": 0.75, "learning_rate": 3.1139169570302024e-07, "loss": 1.2272, "step": 9657 }, { "epoch": 0.75, "learning_rate": 3.112094277860676e-07, "loss": 0.7894, "step": 9658 }, { "epoch": 0.75, "learning_rate": 3.110272033977853e-07, "loss": 0.8383, "step": 9659 }, { "epoch": 0.75, "learning_rate": 3.1084502254968913e-07, "loss": 1.2741, "step": 9660 }, { "epoch": 0.75, "learning_rate": 3.106628852532912e-07, "loss": 1.1837, "step": 9661 }, { "epoch": 0.75, "learning_rate": 3.1048079152010276e-07, "loss": 1.1497, "step": 9662 }, { "epoch": 0.75, "learning_rate": 3.1029874136163103e-07, "loss": 1.1828, "step": 9663 }, { "epoch": 0.75, "learning_rate": 3.1011673478938096e-07, "loss": 1.2203, "step": 9664 }, { "epoch": 0.75, "learning_rate": 3.099347718148546e-07, "loss": 1.2449, "step": 9665 }, { "epoch": 0.75, "learning_rate": 3.097528524495506e-07, "loss": 1.2005, "step": 9666 }, { "epoch": 0.75, "learning_rate": 3.0957097670496634e-07, "loss": 1.2136, "step": 9667 }, { "epoch": 0.75, "learning_rate": 3.093891445925954e-07, "loss": 1.1944, "step": 9668 }, { "epoch": 0.75, "learning_rate": 3.092073561239286e-07, "loss": 1.1563, "step": 9669 }, { "epoch": 0.75, "learning_rate": 3.090256113104542e-07, "loss": 0.8625, "step": 9670 }, { "epoch": 0.75, "learning_rate": 3.0884391016365794e-07, "loss": 1.1986, "step": 9671 }, { "epoch": 0.75, "learning_rate": 3.0866225269502243e-07, "loss": 1.2305, "step": 9672 }, { "epoch": 0.75, "learning_rate": 3.084806389160276e-07, "loss": 1.185, "step": 9673 }, { "epoch": 0.75, "learning_rate": 3.0829906883815084e-07, "loss": 1.2265, "step": 9674 }, { "epoch": 0.75, "learning_rate": 3.081175424728665e-07, "loss": 1.2444, "step": 9675 }, { "epoch": 0.75, "learning_rate": 3.0793605983164626e-07, "loss": 1.1822, "step": 9676 }, { "epoch": 0.75, "learning_rate": 3.077546209259592e-07, "loss": 1.1776, "step": 9677 }, { "epoch": 0.75, "learning_rate": 3.075732257672713e-07, "loss": 1.2069, "step": 9678 }, { "epoch": 0.75, "learning_rate": 3.073918743670462e-07, "loss": 0.8452, "step": 9679 }, { "epoch": 0.75, "learning_rate": 3.072105667367443e-07, "loss": 1.2058, "step": 9680 }, { "epoch": 0.75, "learning_rate": 3.070293028878237e-07, "loss": 1.177, "step": 9681 }, { "epoch": 0.75, "learning_rate": 3.0684808283173935e-07, "loss": 1.2147, "step": 9682 }, { "epoch": 0.75, "learning_rate": 3.0666690657994364e-07, "loss": 1.1717, "step": 9683 }, { "epoch": 0.75, "learning_rate": 3.0648577414388634e-07, "loss": 1.1838, "step": 9684 }, { "epoch": 0.75, "learning_rate": 3.063046855350135e-07, "loss": 1.1508, "step": 9685 }, { "epoch": 0.75, "learning_rate": 3.061236407647706e-07, "loss": 1.2242, "step": 9686 }, { "epoch": 0.75, "learning_rate": 3.0594263984459766e-07, "loss": 1.1702, "step": 9687 }, { "epoch": 0.75, "learning_rate": 3.057616827859336e-07, "loss": 1.1544, "step": 9688 }, { "epoch": 0.75, "learning_rate": 3.0558076960021416e-07, "loss": 1.2178, "step": 9689 }, { "epoch": 0.75, "learning_rate": 3.0539990029887185e-07, "loss": 1.1931, "step": 9690 }, { "epoch": 0.75, "learning_rate": 3.0521907489333797e-07, "loss": 1.1606, "step": 9691 }, { "epoch": 0.75, "learning_rate": 3.0503829339503884e-07, "loss": 1.1604, "step": 9692 }, { "epoch": 0.75, "learning_rate": 3.048575558153992e-07, "loss": 1.2031, "step": 9693 }, { "epoch": 0.75, "learning_rate": 3.0467686216584174e-07, "loss": 1.1357, "step": 9694 }, { "epoch": 0.75, "learning_rate": 3.0449621245778433e-07, "loss": 1.1769, "step": 9695 }, { "epoch": 0.75, "learning_rate": 3.0431560670264454e-07, "loss": 1.219, "step": 9696 }, { "epoch": 0.75, "learning_rate": 3.0413504491183503e-07, "loss": 1.1743, "step": 9697 }, { "epoch": 0.75, "learning_rate": 3.039545270967664e-07, "loss": 1.1962, "step": 9698 }, { "epoch": 0.75, "learning_rate": 3.0377405326884763e-07, "loss": 1.2408, "step": 9699 }, { "epoch": 0.75, "learning_rate": 3.0359362343948267e-07, "loss": 1.131, "step": 9700 }, { "epoch": 0.75, "learning_rate": 3.034132376200749e-07, "loss": 1.2248, "step": 9701 }, { "epoch": 0.75, "learning_rate": 3.0323289582202336e-07, "loss": 1.198, "step": 9702 }, { "epoch": 0.75, "learning_rate": 3.030525980567253e-07, "loss": 1.1797, "step": 9703 }, { "epoch": 0.75, "learning_rate": 3.0287234433557465e-07, "loss": 1.1999, "step": 9704 }, { "epoch": 0.75, "learning_rate": 3.026921346699625e-07, "loss": 1.2106, "step": 9705 }, { "epoch": 0.75, "learning_rate": 3.0251196907127753e-07, "loss": 1.2152, "step": 9706 }, { "epoch": 0.75, "learning_rate": 3.0233184755090537e-07, "loss": 1.1673, "step": 9707 }, { "epoch": 0.75, "learning_rate": 3.0215177012022897e-07, "loss": 1.2092, "step": 9708 }, { "epoch": 0.75, "learning_rate": 3.019717367906284e-07, "loss": 1.1738, "step": 9709 }, { "epoch": 0.75, "learning_rate": 3.0179174757348116e-07, "loss": 1.2384, "step": 9710 }, { "epoch": 0.75, "learning_rate": 3.0161180248016173e-07, "loss": 1.1799, "step": 9711 }, { "epoch": 0.75, "learning_rate": 3.014319015220418e-07, "loss": 1.1961, "step": 9712 }, { "epoch": 0.75, "learning_rate": 3.0125204471049047e-07, "loss": 1.2154, "step": 9713 }, { "epoch": 0.75, "learning_rate": 3.0107223205687395e-07, "loss": 1.2002, "step": 9714 }, { "epoch": 0.75, "learning_rate": 3.008924635725556e-07, "loss": 1.1851, "step": 9715 }, { "epoch": 0.75, "learning_rate": 3.0071273926889594e-07, "loss": 1.2289, "step": 9716 }, { "epoch": 0.75, "learning_rate": 3.005330591572529e-07, "loss": 1.1849, "step": 9717 }, { "epoch": 0.75, "learning_rate": 3.003534232489816e-07, "loss": 1.1988, "step": 9718 }, { "epoch": 0.75, "learning_rate": 3.0017383155543407e-07, "loss": 0.8284, "step": 9719 }, { "epoch": 0.75, "learning_rate": 2.9999428408795977e-07, "loss": 1.2184, "step": 9720 }, { "epoch": 0.75, "learning_rate": 2.998147808579056e-07, "loss": 1.2554, "step": 9721 }, { "epoch": 0.75, "learning_rate": 2.99635321876615e-07, "loss": 1.1771, "step": 9722 }, { "epoch": 0.75, "learning_rate": 2.9945590715542947e-07, "loss": 1.1664, "step": 9723 }, { "epoch": 0.75, "learning_rate": 2.9927653670568655e-07, "loss": 1.2389, "step": 9724 }, { "epoch": 0.75, "learning_rate": 2.990972105387229e-07, "loss": 1.1557, "step": 9725 }, { "epoch": 0.75, "learning_rate": 2.9891792866587016e-07, "loss": 1.1396, "step": 9726 }, { "epoch": 0.75, "learning_rate": 2.9873869109845816e-07, "loss": 1.2509, "step": 9727 }, { "epoch": 0.76, "learning_rate": 2.98559497847815e-07, "loss": 1.1743, "step": 9728 }, { "epoch": 0.76, "learning_rate": 2.9838034892526353e-07, "loss": 1.1913, "step": 9729 }, { "epoch": 0.76, "learning_rate": 2.982012443421267e-07, "loss": 1.1762, "step": 9730 }, { "epoch": 0.76, "learning_rate": 2.980221841097219e-07, "loss": 1.2164, "step": 9731 }, { "epoch": 0.76, "learning_rate": 2.978431682393653e-07, "loss": 1.225, "step": 9732 }, { "epoch": 0.76, "learning_rate": 2.9766419674237076e-07, "loss": 1.2083, "step": 9733 }, { "epoch": 0.76, "learning_rate": 2.974852696300472e-07, "loss": 0.8539, "step": 9734 }, { "epoch": 0.76, "learning_rate": 2.9730638691370304e-07, "loss": 1.1469, "step": 9735 }, { "epoch": 0.76, "learning_rate": 2.9712754860464294e-07, "loss": 1.2228, "step": 9736 }, { "epoch": 0.76, "learning_rate": 2.969487547141679e-07, "loss": 1.1835, "step": 9737 }, { "epoch": 0.76, "learning_rate": 2.967700052535778e-07, "loss": 1.2233, "step": 9738 }, { "epoch": 0.76, "learning_rate": 2.9659130023416833e-07, "loss": 1.182, "step": 9739 }, { "epoch": 0.76, "learning_rate": 2.964126396672333e-07, "loss": 1.1739, "step": 9740 }, { "epoch": 0.76, "learning_rate": 2.9623402356406326e-07, "loss": 1.1898, "step": 9741 }, { "epoch": 0.76, "learning_rate": 2.960554519359453e-07, "loss": 1.227, "step": 9742 }, { "epoch": 0.76, "learning_rate": 2.9587692479416516e-07, "loss": 0.8433, "step": 9743 }, { "epoch": 0.76, "learning_rate": 2.956984421500048e-07, "loss": 1.185, "step": 9744 }, { "epoch": 0.76, "learning_rate": 2.955200040147435e-07, "loss": 1.0933, "step": 9745 }, { "epoch": 0.76, "learning_rate": 2.95341610399658e-07, "loss": 1.187, "step": 9746 }, { "epoch": 0.76, "learning_rate": 2.9516326131602177e-07, "loss": 1.1847, "step": 9747 }, { "epoch": 0.76, "learning_rate": 2.949849567751058e-07, "loss": 1.1942, "step": 9748 }, { "epoch": 0.76, "learning_rate": 2.9480669678817825e-07, "loss": 1.1672, "step": 9749 }, { "epoch": 0.76, "learning_rate": 2.946284813665043e-07, "loss": 1.1675, "step": 9750 }, { "epoch": 0.76, "learning_rate": 2.944503105213466e-07, "loss": 1.1616, "step": 9751 }, { "epoch": 0.76, "learning_rate": 2.942721842639646e-07, "loss": 1.1945, "step": 9752 }, { "epoch": 0.76, "learning_rate": 2.940941026056153e-07, "loss": 1.1832, "step": 9753 }, { "epoch": 0.76, "learning_rate": 2.9391606555755255e-07, "loss": 1.1659, "step": 9754 }, { "epoch": 0.76, "learning_rate": 2.937380731310276e-07, "loss": 1.1833, "step": 9755 }, { "epoch": 0.76, "learning_rate": 2.935601253372889e-07, "loss": 1.1987, "step": 9756 }, { "epoch": 0.76, "learning_rate": 2.933822221875819e-07, "loss": 1.2168, "step": 9757 }, { "epoch": 0.76, "learning_rate": 2.932043636931494e-07, "loss": 1.1666, "step": 9758 }, { "epoch": 0.76, "learning_rate": 2.9302654986523124e-07, "loss": 1.251, "step": 9759 }, { "epoch": 0.76, "learning_rate": 2.928487807150646e-07, "loss": 0.8027, "step": 9760 }, { "epoch": 0.76, "learning_rate": 2.926710562538833e-07, "loss": 1.1963, "step": 9761 }, { "epoch": 0.76, "learning_rate": 2.9249337649291993e-07, "loss": 1.1874, "step": 9762 }, { "epoch": 0.76, "learning_rate": 2.9231574144340186e-07, "loss": 1.1938, "step": 9763 }, { "epoch": 0.76, "learning_rate": 2.921381511165555e-07, "loss": 1.1859, "step": 9764 }, { "epoch": 0.76, "learning_rate": 2.919606055236037e-07, "loss": 1.2325, "step": 9765 }, { "epoch": 0.76, "learning_rate": 2.917831046757662e-07, "loss": 1.1851, "step": 9766 }, { "epoch": 0.76, "learning_rate": 2.916056485842613e-07, "loss": 1.1653, "step": 9767 }, { "epoch": 0.76, "learning_rate": 2.914282372603023e-07, "loss": 1.2048, "step": 9768 }, { "epoch": 0.76, "learning_rate": 2.912508707151017e-07, "loss": 1.1529, "step": 9769 }, { "epoch": 0.76, "learning_rate": 2.9107354895986835e-07, "loss": 1.1684, "step": 9770 }, { "epoch": 0.76, "learning_rate": 2.908962720058072e-07, "loss": 1.196, "step": 9771 }, { "epoch": 0.76, "learning_rate": 2.9071903986412296e-07, "loss": 0.872, "step": 9772 }, { "epoch": 0.76, "learning_rate": 2.905418525460143e-07, "loss": 1.1881, "step": 9773 }, { "epoch": 0.76, "learning_rate": 2.903647100626799e-07, "loss": 1.2051, "step": 9774 }, { "epoch": 0.76, "learning_rate": 2.901876124253144e-07, "loss": 1.1619, "step": 9775 }, { "epoch": 0.76, "learning_rate": 2.9001055964510866e-07, "loss": 1.1305, "step": 9776 }, { "epoch": 0.76, "learning_rate": 2.898335517332526e-07, "loss": 1.187, "step": 9777 }, { "epoch": 0.76, "learning_rate": 2.896565887009321e-07, "loss": 1.2306, "step": 9778 }, { "epoch": 0.76, "learning_rate": 2.894796705593305e-07, "loss": 0.8406, "step": 9779 }, { "epoch": 0.76, "learning_rate": 2.8930279731962826e-07, "loss": 1.216, "step": 9780 }, { "epoch": 0.76, "learning_rate": 2.891259689930029e-07, "loss": 1.1609, "step": 9781 }, { "epoch": 0.76, "learning_rate": 2.889491855906294e-07, "loss": 1.194, "step": 9782 }, { "epoch": 0.76, "learning_rate": 2.8877244712367967e-07, "loss": 1.1695, "step": 9783 }, { "epoch": 0.76, "learning_rate": 2.885957536033228e-07, "loss": 1.204, "step": 9784 }, { "epoch": 0.76, "learning_rate": 2.884191050407251e-07, "loss": 1.187, "step": 9785 }, { "epoch": 0.76, "learning_rate": 2.8824250144704997e-07, "loss": 1.1662, "step": 9786 }, { "epoch": 0.76, "learning_rate": 2.880659428334582e-07, "loss": 1.1837, "step": 9787 }, { "epoch": 0.76, "learning_rate": 2.8788942921110736e-07, "loss": 1.2193, "step": 9788 }, { "epoch": 0.76, "learning_rate": 2.8771296059115236e-07, "loss": 0.8062, "step": 9789 }, { "epoch": 0.76, "learning_rate": 2.8753653698474544e-07, "loss": 1.2126, "step": 9790 }, { "epoch": 0.76, "learning_rate": 2.8736015840303574e-07, "loss": 1.2051, "step": 9791 }, { "epoch": 0.76, "learning_rate": 2.871838248571696e-07, "loss": 0.837, "step": 9792 }, { "epoch": 0.76, "learning_rate": 2.870075363582906e-07, "loss": 1.2265, "step": 9793 }, { "epoch": 0.76, "learning_rate": 2.868312929175395e-07, "loss": 0.8616, "step": 9794 }, { "epoch": 0.76, "learning_rate": 2.8665509454605397e-07, "loss": 1.157, "step": 9795 }, { "epoch": 0.76, "learning_rate": 2.864789412549692e-07, "loss": 0.8014, "step": 9796 }, { "epoch": 0.76, "learning_rate": 2.8630283305541726e-07, "loss": 1.1761, "step": 9797 }, { "epoch": 0.76, "learning_rate": 2.861267699585274e-07, "loss": 1.168, "step": 9798 }, { "epoch": 0.76, "learning_rate": 2.8595075197542606e-07, "loss": 1.2093, "step": 9799 }, { "epoch": 0.76, "learning_rate": 2.857747791172367e-07, "loss": 1.1861, "step": 9800 }, { "epoch": 0.76, "learning_rate": 2.8559885139508076e-07, "loss": 1.1844, "step": 9801 }, { "epoch": 0.76, "learning_rate": 2.854229688200753e-07, "loss": 1.1986, "step": 9802 }, { "epoch": 0.76, "learning_rate": 2.8524713140333545e-07, "loss": 0.8304, "step": 9803 }, { "epoch": 0.76, "learning_rate": 2.8507133915597427e-07, "loss": 1.1765, "step": 9804 }, { "epoch": 0.76, "learning_rate": 2.8489559208909983e-07, "loss": 1.1838, "step": 9805 }, { "epoch": 0.76, "learning_rate": 2.847198902138198e-07, "loss": 0.8452, "step": 9806 }, { "epoch": 0.76, "learning_rate": 2.8454423354123694e-07, "loss": 0.8621, "step": 9807 }, { "epoch": 0.76, "learning_rate": 2.8436862208245196e-07, "loss": 0.8047, "step": 9808 }, { "epoch": 0.76, "learning_rate": 2.8419305584856377e-07, "loss": 1.1693, "step": 9809 }, { "epoch": 0.76, "learning_rate": 2.8401753485066616e-07, "loss": 0.798, "step": 9810 }, { "epoch": 0.76, "learning_rate": 2.8384205909985215e-07, "loss": 1.218, "step": 9811 }, { "epoch": 0.76, "learning_rate": 2.836666286072109e-07, "loss": 1.1863, "step": 9812 }, { "epoch": 0.76, "learning_rate": 2.834912433838288e-07, "loss": 1.1591, "step": 9813 }, { "epoch": 0.76, "learning_rate": 2.8331590344078947e-07, "loss": 0.8281, "step": 9814 }, { "epoch": 0.76, "learning_rate": 2.831406087891737e-07, "loss": 1.2262, "step": 9815 }, { "epoch": 0.76, "learning_rate": 2.829653594400593e-07, "loss": 0.8277, "step": 9816 }, { "epoch": 0.76, "learning_rate": 2.827901554045212e-07, "loss": 1.1431, "step": 9817 }, { "epoch": 0.76, "learning_rate": 2.8261499669363187e-07, "loss": 1.181, "step": 9818 }, { "epoch": 0.76, "learning_rate": 2.8243988331846033e-07, "loss": 1.2349, "step": 9819 }, { "epoch": 0.76, "learning_rate": 2.822648152900732e-07, "loss": 1.2507, "step": 9820 }, { "epoch": 0.76, "learning_rate": 2.820897926195339e-07, "loss": 1.206, "step": 9821 }, { "epoch": 0.76, "learning_rate": 2.819148153179032e-07, "loss": 1.2045, "step": 9822 }, { "epoch": 0.76, "learning_rate": 2.8173988339623887e-07, "loss": 0.8017, "step": 9823 }, { "epoch": 0.76, "learning_rate": 2.815649968655959e-07, "loss": 1.1569, "step": 9824 }, { "epoch": 0.76, "learning_rate": 2.813901557370264e-07, "loss": 1.1971, "step": 9825 }, { "epoch": 0.76, "learning_rate": 2.812153600215796e-07, "loss": 1.1924, "step": 9826 }, { "epoch": 0.76, "learning_rate": 2.810406097303019e-07, "loss": 1.1812, "step": 9827 }, { "epoch": 0.76, "learning_rate": 2.8086590487423654e-07, "loss": 1.184, "step": 9828 }, { "epoch": 0.76, "learning_rate": 2.806912454644245e-07, "loss": 1.1783, "step": 9829 }, { "epoch": 0.76, "learning_rate": 2.805166315119033e-07, "loss": 1.1654, "step": 9830 }, { "epoch": 0.76, "learning_rate": 2.803420630277078e-07, "loss": 1.1362, "step": 9831 }, { "epoch": 0.76, "learning_rate": 2.8016754002287e-07, "loss": 0.7968, "step": 9832 }, { "epoch": 0.76, "learning_rate": 2.7999306250841914e-07, "loss": 1.2138, "step": 9833 }, { "epoch": 0.76, "learning_rate": 2.798186304953811e-07, "loss": 1.1916, "step": 9834 }, { "epoch": 0.76, "learning_rate": 2.7964424399478013e-07, "loss": 1.1588, "step": 9835 }, { "epoch": 0.76, "learning_rate": 2.7946990301763583e-07, "loss": 0.8756, "step": 9836 }, { "epoch": 0.76, "learning_rate": 2.792956075749657e-07, "loss": 1.1616, "step": 9837 }, { "epoch": 0.76, "learning_rate": 2.791213576777857e-07, "loss": 1.1834, "step": 9838 }, { "epoch": 0.76, "learning_rate": 2.789471533371062e-07, "loss": 1.1431, "step": 9839 }, { "epoch": 0.76, "learning_rate": 2.7877299456393745e-07, "loss": 1.2017, "step": 9840 }, { "epoch": 0.76, "learning_rate": 2.7859888136928464e-07, "loss": 1.2146, "step": 9841 }, { "epoch": 0.76, "learning_rate": 2.7842481376415105e-07, "loss": 1.226, "step": 9842 }, { "epoch": 0.76, "learning_rate": 2.7825079175953805e-07, "loss": 0.851, "step": 9843 }, { "epoch": 0.76, "learning_rate": 2.7807681536644154e-07, "loss": 1.2059, "step": 9844 }, { "epoch": 0.76, "learning_rate": 2.779028845958573e-07, "loss": 1.1216, "step": 9845 }, { "epoch": 0.76, "learning_rate": 2.7772899945877693e-07, "loss": 1.252, "step": 9846 }, { "epoch": 0.76, "learning_rate": 2.775551599661884e-07, "loss": 1.208, "step": 9847 }, { "epoch": 0.76, "learning_rate": 2.773813661290787e-07, "loss": 1.177, "step": 9848 }, { "epoch": 0.76, "learning_rate": 2.772076179584297e-07, "loss": 1.2067, "step": 9849 }, { "epoch": 0.76, "learning_rate": 2.770339154652226e-07, "loss": 1.1917, "step": 9850 }, { "epoch": 0.76, "learning_rate": 2.768602586604345e-07, "loss": 0.8439, "step": 9851 }, { "epoch": 0.76, "learning_rate": 2.766866475550389e-07, "loss": 0.8235, "step": 9852 }, { "epoch": 0.76, "learning_rate": 2.765130821600083e-07, "loss": 0.8259, "step": 9853 }, { "epoch": 0.76, "learning_rate": 2.7633956248631094e-07, "loss": 1.1579, "step": 9854 }, { "epoch": 0.76, "learning_rate": 2.761660885449126e-07, "loss": 1.1749, "step": 9855 }, { "epoch": 0.76, "learning_rate": 2.75992660346776e-07, "loss": 1.2081, "step": 9856 }, { "epoch": 0.77, "learning_rate": 2.7581927790286106e-07, "loss": 1.2618, "step": 9857 }, { "epoch": 0.77, "learning_rate": 2.75645941224125e-07, "loss": 0.8528, "step": 9858 }, { "epoch": 0.77, "learning_rate": 2.754726503215218e-07, "loss": 1.1862, "step": 9859 }, { "epoch": 0.77, "learning_rate": 2.752994052060027e-07, "loss": 1.176, "step": 9860 }, { "epoch": 0.77, "learning_rate": 2.7512620588851623e-07, "loss": 1.2393, "step": 9861 }, { "epoch": 0.77, "learning_rate": 2.7495305238000764e-07, "loss": 1.1728, "step": 9862 }, { "epoch": 0.77, "learning_rate": 2.7477994469141964e-07, "loss": 0.7832, "step": 9863 }, { "epoch": 0.77, "learning_rate": 2.7460688283369193e-07, "loss": 1.1881, "step": 9864 }, { "epoch": 0.77, "learning_rate": 2.744338668177613e-07, "loss": 0.8029, "step": 9865 }, { "epoch": 0.77, "learning_rate": 2.7426089665456143e-07, "loss": 1.1879, "step": 9866 }, { "epoch": 0.77, "learning_rate": 2.740879723550236e-07, "loss": 1.1869, "step": 9867 }, { "epoch": 0.77, "learning_rate": 2.7391509393007573e-07, "loss": 0.8282, "step": 9868 }, { "epoch": 0.77, "learning_rate": 2.7374226139064303e-07, "loss": 1.1739, "step": 9869 }, { "epoch": 0.77, "learning_rate": 2.735694747476478e-07, "loss": 1.151, "step": 9870 }, { "epoch": 0.77, "learning_rate": 2.7339673401200946e-07, "loss": 1.2645, "step": 9871 }, { "epoch": 0.77, "learning_rate": 2.7322403919464457e-07, "loss": 1.2156, "step": 9872 }, { "epoch": 0.77, "learning_rate": 2.730513903064664e-07, "loss": 0.7897, "step": 9873 }, { "epoch": 0.77, "learning_rate": 2.72878787358386e-07, "loss": 1.1735, "step": 9874 }, { "epoch": 0.77, "learning_rate": 2.72706230361311e-07, "loss": 1.1833, "step": 9875 }, { "epoch": 0.77, "learning_rate": 2.7253371932614586e-07, "loss": 1.1874, "step": 9876 }, { "epoch": 0.77, "learning_rate": 2.7236125426379373e-07, "loss": 1.1649, "step": 9877 }, { "epoch": 0.77, "learning_rate": 2.7218883518515226e-07, "loss": 1.1675, "step": 9878 }, { "epoch": 0.77, "learning_rate": 2.7201646210111853e-07, "loss": 1.1592, "step": 9879 }, { "epoch": 0.77, "learning_rate": 2.7184413502258586e-07, "loss": 0.86, "step": 9880 }, { "epoch": 0.77, "learning_rate": 2.7167185396044377e-07, "loss": 1.149, "step": 9881 }, { "epoch": 0.77, "learning_rate": 2.714996189255807e-07, "loss": 1.2124, "step": 9882 }, { "epoch": 0.77, "learning_rate": 2.713274299288801e-07, "loss": 1.2061, "step": 9883 }, { "epoch": 0.77, "learning_rate": 2.7115528698122446e-07, "loss": 1.219, "step": 9884 }, { "epoch": 0.77, "learning_rate": 2.7098319009349257e-07, "loss": 1.2359, "step": 9885 }, { "epoch": 0.77, "learning_rate": 2.708111392765592e-07, "loss": 1.1537, "step": 9886 }, { "epoch": 0.77, "learning_rate": 2.706391345412983e-07, "loss": 1.235, "step": 9887 }, { "epoch": 0.77, "learning_rate": 2.704671758985794e-07, "loss": 1.237, "step": 9888 }, { "epoch": 0.77, "learning_rate": 2.7029526335926955e-07, "loss": 0.8764, "step": 9889 }, { "epoch": 0.77, "learning_rate": 2.7012339693423337e-07, "loss": 0.8604, "step": 9890 }, { "epoch": 0.77, "learning_rate": 2.69951576634331e-07, "loss": 1.0922, "step": 9891 }, { "epoch": 0.77, "learning_rate": 2.6977980247042176e-07, "loss": 1.1678, "step": 9892 }, { "epoch": 0.77, "learning_rate": 2.6960807445336077e-07, "loss": 1.2012, "step": 9893 }, { "epoch": 0.77, "learning_rate": 2.6943639259400054e-07, "loss": 1.2298, "step": 9894 }, { "epoch": 0.77, "learning_rate": 2.692647569031906e-07, "loss": 1.1724, "step": 9895 }, { "epoch": 0.77, "learning_rate": 2.6909316739177755e-07, "loss": 1.2337, "step": 9896 }, { "epoch": 0.77, "learning_rate": 2.6892162407060514e-07, "loss": 1.2337, "step": 9897 }, { "epoch": 0.77, "learning_rate": 2.687501269505144e-07, "loss": 1.2477, "step": 9898 }, { "epoch": 0.77, "learning_rate": 2.6857867604234286e-07, "loss": 1.2075, "step": 9899 }, { "epoch": 0.77, "learning_rate": 2.6840727135692577e-07, "loss": 1.139, "step": 9900 }, { "epoch": 0.77, "learning_rate": 2.6823591290509517e-07, "loss": 1.2809, "step": 9901 }, { "epoch": 0.77, "learning_rate": 2.6806460069768e-07, "loss": 1.2046, "step": 9902 }, { "epoch": 0.77, "learning_rate": 2.678933347455067e-07, "loss": 0.8774, "step": 9903 }, { "epoch": 0.77, "learning_rate": 2.6772211505939846e-07, "loss": 0.8236, "step": 9904 }, { "epoch": 0.77, "learning_rate": 2.6755094165017567e-07, "loss": 0.8437, "step": 9905 }, { "epoch": 0.77, "learning_rate": 2.6737981452865567e-07, "loss": 1.175, "step": 9906 }, { "epoch": 0.77, "learning_rate": 2.672087337056531e-07, "loss": 1.139, "step": 9907 }, { "epoch": 0.77, "learning_rate": 2.670376991919795e-07, "loss": 1.1865, "step": 9908 }, { "epoch": 0.77, "learning_rate": 2.668667109984435e-07, "loss": 1.1711, "step": 9909 }, { "epoch": 0.77, "learning_rate": 2.6669576913585056e-07, "loss": 1.1725, "step": 9910 }, { "epoch": 0.77, "learning_rate": 2.665248736150044e-07, "loss": 1.1957, "step": 9911 }, { "epoch": 0.77, "learning_rate": 2.663540244467041e-07, "loss": 1.1937, "step": 9912 }, { "epoch": 0.77, "learning_rate": 2.661832216417463e-07, "loss": 1.2385, "step": 9913 }, { "epoch": 0.77, "learning_rate": 2.660124652109264e-07, "loss": 1.2432, "step": 9914 }, { "epoch": 0.77, "learning_rate": 2.658417551650337e-07, "loss": 0.8525, "step": 9915 }, { "epoch": 0.77, "learning_rate": 2.6567109151485813e-07, "loss": 1.1852, "step": 9916 }, { "epoch": 0.77, "learning_rate": 2.655004742711836e-07, "loss": 1.1556, "step": 9917 }, { "epoch": 0.77, "learning_rate": 2.6532990344479264e-07, "loss": 1.2128, "step": 9918 }, { "epoch": 0.77, "learning_rate": 2.6515937904646537e-07, "loss": 1.207, "step": 9919 }, { "epoch": 0.77, "learning_rate": 2.6498890108697713e-07, "loss": 1.1605, "step": 9920 }, { "epoch": 0.77, "learning_rate": 2.648184695771023e-07, "loss": 1.19, "step": 9921 }, { "epoch": 0.77, "learning_rate": 2.64648084527611e-07, "loss": 1.1885, "step": 9922 }, { "epoch": 0.77, "learning_rate": 2.6447774594927096e-07, "loss": 1.1847, "step": 9923 }, { "epoch": 0.77, "learning_rate": 2.6430745385284713e-07, "loss": 1.2169, "step": 9924 }, { "epoch": 0.77, "learning_rate": 2.6413720824910034e-07, "loss": 1.2056, "step": 9925 }, { "epoch": 0.77, "learning_rate": 2.6396700914879045e-07, "loss": 1.1666, "step": 9926 }, { "epoch": 0.77, "learning_rate": 2.637968565626728e-07, "loss": 1.1763, "step": 9927 }, { "epoch": 0.77, "learning_rate": 2.636267505015004e-07, "loss": 0.8278, "step": 9928 }, { "epoch": 0.77, "learning_rate": 2.634566909760233e-07, "loss": 1.199, "step": 9929 }, { "epoch": 0.77, "learning_rate": 2.632866779969886e-07, "loss": 0.8296, "step": 9930 }, { "epoch": 0.77, "learning_rate": 2.6311671157514016e-07, "loss": 1.1988, "step": 9931 }, { "epoch": 0.77, "learning_rate": 2.6294679172121935e-07, "loss": 1.2109, "step": 9932 }, { "epoch": 0.77, "learning_rate": 2.627769184459643e-07, "loss": 1.1783, "step": 9933 }, { "epoch": 0.77, "learning_rate": 2.626070917601104e-07, "loss": 1.1472, "step": 9934 }, { "epoch": 0.77, "learning_rate": 2.624373116743899e-07, "loss": 1.2303, "step": 9935 }, { "epoch": 0.77, "learning_rate": 2.6226757819953204e-07, "loss": 1.207, "step": 9936 }, { "epoch": 0.77, "learning_rate": 2.620978913462636e-07, "loss": 1.1764, "step": 9937 }, { "epoch": 0.77, "learning_rate": 2.619282511253078e-07, "loss": 1.1809, "step": 9938 }, { "epoch": 0.77, "learning_rate": 2.6175865754738534e-07, "loss": 1.2188, "step": 9939 }, { "epoch": 0.77, "learning_rate": 2.6158911062321376e-07, "loss": 1.235, "step": 9940 }, { "epoch": 0.77, "learning_rate": 2.6141961036350767e-07, "loss": 1.2236, "step": 9941 }, { "epoch": 0.77, "learning_rate": 2.612501567789789e-07, "loss": 1.2053, "step": 9942 }, { "epoch": 0.77, "learning_rate": 2.6108074988033603e-07, "loss": 1.1699, "step": 9943 }, { "epoch": 0.77, "learning_rate": 2.6091138967828475e-07, "loss": 1.1799, "step": 9944 }, { "epoch": 0.77, "learning_rate": 2.607420761835286e-07, "loss": 1.2092, "step": 9945 }, { "epoch": 0.77, "learning_rate": 2.6057280940676686e-07, "loss": 1.189, "step": 9946 }, { "epoch": 0.77, "learning_rate": 2.6040358935869646e-07, "loss": 1.185, "step": 9947 }, { "epoch": 0.77, "learning_rate": 2.602344160500116e-07, "loss": 1.2237, "step": 9948 }, { "epoch": 0.77, "learning_rate": 2.600652894914031e-07, "loss": 1.1832, "step": 9949 }, { "epoch": 0.77, "learning_rate": 2.5989620969355974e-07, "loss": 1.1656, "step": 9950 }, { "epoch": 0.77, "learning_rate": 2.597271766671658e-07, "loss": 1.2449, "step": 9951 }, { "epoch": 0.77, "learning_rate": 2.595581904229036e-07, "loss": 1.1999, "step": 9952 }, { "epoch": 0.77, "learning_rate": 2.5938925097145317e-07, "loss": 0.8401, "step": 9953 }, { "epoch": 0.77, "learning_rate": 2.5922035832348944e-07, "loss": 0.8517, "step": 9954 }, { "epoch": 0.77, "learning_rate": 2.590515124896868e-07, "loss": 1.1729, "step": 9955 }, { "epoch": 0.77, "learning_rate": 2.5888271348071554e-07, "loss": 1.2123, "step": 9956 }, { "epoch": 0.77, "learning_rate": 2.5871396130724224e-07, "loss": 1.1996, "step": 9957 }, { "epoch": 0.77, "learning_rate": 2.5854525597993225e-07, "loss": 1.2072, "step": 9958 }, { "epoch": 0.77, "learning_rate": 2.5837659750944617e-07, "loss": 0.8229, "step": 9959 }, { "epoch": 0.77, "learning_rate": 2.5820798590644323e-07, "loss": 1.1812, "step": 9960 }, { "epoch": 0.77, "learning_rate": 2.5803942118157895e-07, "loss": 1.2069, "step": 9961 }, { "epoch": 0.77, "learning_rate": 2.578709033455051e-07, "loss": 1.2446, "step": 9962 }, { "epoch": 0.77, "learning_rate": 2.5770243240887214e-07, "loss": 0.834, "step": 9963 }, { "epoch": 0.77, "learning_rate": 2.575340083823264e-07, "loss": 1.2117, "step": 9964 }, { "epoch": 0.77, "learning_rate": 2.573656312765118e-07, "loss": 1.2488, "step": 9965 }, { "epoch": 0.77, "learning_rate": 2.5719730110206874e-07, "loss": 1.2331, "step": 9966 }, { "epoch": 0.77, "learning_rate": 2.570290178696352e-07, "loss": 1.1686, "step": 9967 }, { "epoch": 0.77, "learning_rate": 2.568607815898458e-07, "loss": 1.1283, "step": 9968 }, { "epoch": 0.77, "learning_rate": 2.566925922733325e-07, "loss": 1.1817, "step": 9969 }, { "epoch": 0.77, "learning_rate": 2.565244499307241e-07, "loss": 1.1836, "step": 9970 }, { "epoch": 0.77, "learning_rate": 2.563563545726466e-07, "loss": 1.1973, "step": 9971 }, { "epoch": 0.77, "learning_rate": 2.561883062097228e-07, "loss": 1.2153, "step": 9972 }, { "epoch": 0.77, "learning_rate": 2.560203048525726e-07, "loss": 1.2079, "step": 9973 }, { "epoch": 0.77, "learning_rate": 2.558523505118132e-07, "loss": 0.8407, "step": 9974 }, { "epoch": 0.77, "learning_rate": 2.5568444319805837e-07, "loss": 1.2004, "step": 9975 }, { "epoch": 0.77, "learning_rate": 2.555165829219192e-07, "loss": 1.1479, "step": 9976 }, { "epoch": 0.77, "learning_rate": 2.55348769694004e-07, "loss": 1.1956, "step": 9977 }, { "epoch": 0.77, "learning_rate": 2.5518100352491746e-07, "loss": 1.1943, "step": 9978 }, { "epoch": 0.77, "learning_rate": 2.5501328442526206e-07, "loss": 1.199, "step": 9979 }, { "epoch": 0.77, "learning_rate": 2.548456124056366e-07, "loss": 1.2113, "step": 9980 }, { "epoch": 0.77, "learning_rate": 2.546779874766376e-07, "loss": 1.1374, "step": 9981 }, { "epoch": 0.77, "learning_rate": 2.5451040964885796e-07, "loss": 1.214, "step": 9982 }, { "epoch": 0.77, "learning_rate": 2.543428789328881e-07, "loss": 1.2135, "step": 9983 }, { "epoch": 0.77, "learning_rate": 2.541753953393151e-07, "loss": 1.1492, "step": 9984 }, { "epoch": 0.77, "learning_rate": 2.540079588787234e-07, "loss": 1.1928, "step": 9985 }, { "epoch": 0.78, "learning_rate": 2.5384056956169375e-07, "loss": 1.2481, "step": 9986 }, { "epoch": 0.78, "learning_rate": 2.536732273988056e-07, "loss": 1.1731, "step": 9987 }, { "epoch": 0.78, "learning_rate": 2.535059324006328e-07, "loss": 1.2168, "step": 9988 }, { "epoch": 0.78, "learning_rate": 2.533386845777489e-07, "loss": 1.1439, "step": 9989 }, { "epoch": 0.78, "learning_rate": 2.5317148394072296e-07, "loss": 1.1874, "step": 9990 }, { "epoch": 0.78, "learning_rate": 2.530043305001206e-07, "loss": 1.2056, "step": 9991 }, { "epoch": 0.78, "learning_rate": 2.5283722426650644e-07, "loss": 1.1876, "step": 9992 }, { "epoch": 0.78, "learning_rate": 2.526701652504397e-07, "loss": 1.168, "step": 9993 }, { "epoch": 0.78, "learning_rate": 2.5250315346247863e-07, "loss": 1.183, "step": 9994 }, { "epoch": 0.78, "learning_rate": 2.523361889131778e-07, "loss": 1.2216, "step": 9995 }, { "epoch": 0.78, "learning_rate": 2.5216927161308755e-07, "loss": 0.8042, "step": 9996 }, { "epoch": 0.78, "learning_rate": 2.5200240157275744e-07, "loss": 0.8356, "step": 9997 }, { "epoch": 0.78, "learning_rate": 2.5183557880273255e-07, "loss": 0.7923, "step": 9998 }, { "epoch": 0.78, "learning_rate": 2.5166880331355553e-07, "loss": 0.7959, "step": 9999 }, { "epoch": 0.78, "learning_rate": 2.5150207511576606e-07, "loss": 1.2045, "step": 10000 }, { "epoch": 0.78, "learning_rate": 2.513353942198997e-07, "loss": 0.8056, "step": 10001 }, { "epoch": 0.78, "learning_rate": 2.5116876063649104e-07, "loss": 1.1594, "step": 10002 }, { "epoch": 0.78, "learning_rate": 2.510021743760702e-07, "loss": 0.811, "step": 10003 }, { "epoch": 0.78, "learning_rate": 2.508356354491649e-07, "loss": 1.2087, "step": 10004 }, { "epoch": 0.78, "learning_rate": 2.506691438662995e-07, "loss": 1.1503, "step": 10005 }, { "epoch": 0.78, "learning_rate": 2.505026996379955e-07, "loss": 1.1438, "step": 10006 }, { "epoch": 0.78, "learning_rate": 2.503363027747718e-07, "loss": 1.2472, "step": 10007 }, { "epoch": 0.78, "learning_rate": 2.5016995328714386e-07, "loss": 1.1931, "step": 10008 }, { "epoch": 0.78, "learning_rate": 2.5000365118562414e-07, "loss": 1.1643, "step": 10009 }, { "epoch": 0.78, "learning_rate": 2.498373964807223e-07, "loss": 1.2057, "step": 10010 }, { "epoch": 0.78, "learning_rate": 2.4967118918294494e-07, "loss": 1.2269, "step": 10011 }, { "epoch": 0.78, "learning_rate": 2.4950502930279573e-07, "loss": 1.1376, "step": 10012 }, { "epoch": 0.78, "learning_rate": 2.4933891685077526e-07, "loss": 1.1859, "step": 10013 }, { "epoch": 0.78, "learning_rate": 2.49172851837381e-07, "loss": 1.138, "step": 10014 }, { "epoch": 0.78, "learning_rate": 2.4900683427310776e-07, "loss": 1.2353, "step": 10015 }, { "epoch": 0.78, "learning_rate": 2.4884086416844696e-07, "loss": 1.1868, "step": 10016 }, { "epoch": 0.78, "learning_rate": 2.4867494153388735e-07, "loss": 1.2044, "step": 10017 }, { "epoch": 0.78, "learning_rate": 2.4850906637991444e-07, "loss": 1.1945, "step": 10018 }, { "epoch": 0.78, "learning_rate": 2.48343238717011e-07, "loss": 1.1582, "step": 10019 }, { "epoch": 0.78, "learning_rate": 2.481774585556562e-07, "loss": 1.209, "step": 10020 }, { "epoch": 0.78, "learning_rate": 2.4801172590632755e-07, "loss": 1.2062, "step": 10021 }, { "epoch": 0.78, "learning_rate": 2.4784604077949787e-07, "loss": 1.1452, "step": 10022 }, { "epoch": 0.78, "learning_rate": 2.47680403185638e-07, "loss": 1.1804, "step": 10023 }, { "epoch": 0.78, "learning_rate": 2.475148131352155e-07, "loss": 1.1941, "step": 10024 }, { "epoch": 0.78, "learning_rate": 2.473492706386947e-07, "loss": 1.1904, "step": 10025 }, { "epoch": 0.78, "learning_rate": 2.4718377570653814e-07, "loss": 1.1811, "step": 10026 }, { "epoch": 0.78, "learning_rate": 2.4701832834920333e-07, "loss": 1.1802, "step": 10027 }, { "epoch": 0.78, "learning_rate": 2.468529285771461e-07, "loss": 1.1597, "step": 10028 }, { "epoch": 0.78, "learning_rate": 2.4668757640081974e-07, "loss": 1.2107, "step": 10029 }, { "epoch": 0.78, "learning_rate": 2.4652227183067266e-07, "loss": 1.1799, "step": 10030 }, { "epoch": 0.78, "learning_rate": 2.4635701487715234e-07, "loss": 1.2006, "step": 10031 }, { "epoch": 0.78, "learning_rate": 2.4619180555070206e-07, "loss": 1.1654, "step": 10032 }, { "epoch": 0.78, "learning_rate": 2.460266438617622e-07, "loss": 0.7986, "step": 10033 }, { "epoch": 0.78, "learning_rate": 2.458615298207708e-07, "loss": 1.2183, "step": 10034 }, { "epoch": 0.78, "learning_rate": 2.456964634381612e-07, "loss": 1.2294, "step": 10035 }, { "epoch": 0.78, "learning_rate": 2.4553144472436606e-07, "loss": 0.8664, "step": 10036 }, { "epoch": 0.78, "learning_rate": 2.4536647368981346e-07, "loss": 1.206, "step": 10037 }, { "epoch": 0.78, "learning_rate": 2.45201550344929e-07, "loss": 1.1798, "step": 10038 }, { "epoch": 0.78, "learning_rate": 2.450366747001349e-07, "loss": 0.8488, "step": 10039 }, { "epoch": 0.78, "learning_rate": 2.448718467658507e-07, "loss": 1.1908, "step": 10040 }, { "epoch": 0.78, "learning_rate": 2.4470706655249293e-07, "loss": 1.1893, "step": 10041 }, { "epoch": 0.78, "learning_rate": 2.44542334070475e-07, "loss": 1.2569, "step": 10042 }, { "epoch": 0.78, "learning_rate": 2.4437764933020713e-07, "loss": 1.1936, "step": 10043 }, { "epoch": 0.78, "learning_rate": 2.442130123420969e-07, "loss": 1.21, "step": 10044 }, { "epoch": 0.78, "learning_rate": 2.440484231165486e-07, "loss": 1.1532, "step": 10045 }, { "epoch": 0.78, "learning_rate": 2.4388388166396354e-07, "loss": 1.1742, "step": 10046 }, { "epoch": 0.78, "learning_rate": 2.437193879947402e-07, "loss": 1.2055, "step": 10047 }, { "epoch": 0.78, "learning_rate": 2.4355494211927374e-07, "loss": 1.164, "step": 10048 }, { "epoch": 0.78, "learning_rate": 2.4339054404795644e-07, "loss": 1.2439, "step": 10049 }, { "epoch": 0.78, "learning_rate": 2.432261937911777e-07, "loss": 1.2005, "step": 10050 }, { "epoch": 0.78, "learning_rate": 2.4306189135932364e-07, "loss": 1.1578, "step": 10051 }, { "epoch": 0.78, "learning_rate": 2.428976367627775e-07, "loss": 0.7747, "step": 10052 }, { "epoch": 0.78, "learning_rate": 2.427334300119195e-07, "loss": 1.2002, "step": 10053 }, { "epoch": 0.78, "learning_rate": 2.4256927111712653e-07, "loss": 1.1913, "step": 10054 }, { "epoch": 0.78, "learning_rate": 2.4240516008877364e-07, "loss": 1.2243, "step": 10055 }, { "epoch": 0.78, "learning_rate": 2.4224109693723114e-07, "loss": 1.2422, "step": 10056 }, { "epoch": 0.78, "learning_rate": 2.420770816728672e-07, "loss": 1.1779, "step": 10057 }, { "epoch": 0.78, "learning_rate": 2.4191311430604724e-07, "loss": 1.1972, "step": 10058 }, { "epoch": 0.78, "learning_rate": 2.417491948471326e-07, "loss": 0.8604, "step": 10059 }, { "epoch": 0.78, "learning_rate": 2.4158532330648353e-07, "loss": 1.2271, "step": 10060 }, { "epoch": 0.78, "learning_rate": 2.4142149969445493e-07, "loss": 1.1979, "step": 10061 }, { "epoch": 0.78, "learning_rate": 2.412577240213999e-07, "loss": 1.2513, "step": 10062 }, { "epoch": 0.78, "learning_rate": 2.4109399629766923e-07, "loss": 0.8114, "step": 10063 }, { "epoch": 0.78, "learning_rate": 2.4093031653360853e-07, "loss": 1.1577, "step": 10064 }, { "epoch": 0.78, "learning_rate": 2.407666847395627e-07, "loss": 1.159, "step": 10065 }, { "epoch": 0.78, "learning_rate": 2.4060310092587244e-07, "loss": 1.1987, "step": 10066 }, { "epoch": 0.78, "learning_rate": 2.4043956510287486e-07, "loss": 1.19, "step": 10067 }, { "epoch": 0.78, "learning_rate": 2.402760772809058e-07, "loss": 1.255, "step": 10068 }, { "epoch": 0.78, "learning_rate": 2.4011263747029586e-07, "loss": 1.1642, "step": 10069 }, { "epoch": 0.78, "learning_rate": 2.399492456813745e-07, "loss": 1.205, "step": 10070 }, { "epoch": 0.78, "learning_rate": 2.397859019244677e-07, "loss": 1.1845, "step": 10071 }, { "epoch": 0.78, "learning_rate": 2.3962260620989683e-07, "loss": 1.135, "step": 10072 }, { "epoch": 0.78, "learning_rate": 2.394593585479827e-07, "loss": 1.1914, "step": 10073 }, { "epoch": 0.78, "learning_rate": 2.392961589490414e-07, "loss": 0.8225, "step": 10074 }, { "epoch": 0.78, "learning_rate": 2.391330074233865e-07, "loss": 1.2407, "step": 10075 }, { "epoch": 0.78, "learning_rate": 2.3896990398132853e-07, "loss": 1.1262, "step": 10076 }, { "epoch": 0.78, "learning_rate": 2.388068486331749e-07, "loss": 1.2236, "step": 10077 }, { "epoch": 0.78, "learning_rate": 2.386438413892301e-07, "loss": 1.2079, "step": 10078 }, { "epoch": 0.78, "learning_rate": 2.3848088225979535e-07, "loss": 0.8212, "step": 10079 }, { "epoch": 0.78, "learning_rate": 2.3831797125516918e-07, "loss": 0.8209, "step": 10080 }, { "epoch": 0.78, "learning_rate": 2.3815510838564666e-07, "loss": 0.7934, "step": 10081 }, { "epoch": 0.78, "learning_rate": 2.379922936615202e-07, "loss": 1.2212, "step": 10082 }, { "epoch": 0.78, "learning_rate": 2.378295270930789e-07, "loss": 1.1903, "step": 10083 }, { "epoch": 0.78, "learning_rate": 2.3766680869060907e-07, "loss": 1.2321, "step": 10084 }, { "epoch": 0.78, "learning_rate": 2.3750413846439376e-07, "loss": 1.2085, "step": 10085 }, { "epoch": 0.78, "learning_rate": 2.373415164247129e-07, "loss": 1.2151, "step": 10086 }, { "epoch": 0.78, "learning_rate": 2.3717894258184378e-07, "loss": 1.1901, "step": 10087 }, { "epoch": 0.78, "learning_rate": 2.370164169460601e-07, "loss": 0.8381, "step": 10088 }, { "epoch": 0.78, "learning_rate": 2.368539395276331e-07, "loss": 1.2097, "step": 10089 }, { "epoch": 0.78, "learning_rate": 2.3669151033683044e-07, "loss": 1.2039, "step": 10090 }, { "epoch": 0.78, "learning_rate": 2.365291293839171e-07, "loss": 1.2083, "step": 10091 }, { "epoch": 0.78, "learning_rate": 2.363667966791547e-07, "loss": 1.2136, "step": 10092 }, { "epoch": 0.78, "learning_rate": 2.3620451223280223e-07, "loss": 1.247, "step": 10093 }, { "epoch": 0.78, "learning_rate": 2.3604227605511518e-07, "loss": 1.1848, "step": 10094 }, { "epoch": 0.78, "learning_rate": 2.3588008815634631e-07, "loss": 1.1607, "step": 10095 }, { "epoch": 0.78, "learning_rate": 2.3571794854674488e-07, "loss": 1.2461, "step": 10096 }, { "epoch": 0.78, "learning_rate": 2.3555585723655823e-07, "loss": 1.2321, "step": 10097 }, { "epoch": 0.78, "learning_rate": 2.3539381423602879e-07, "loss": 1.2209, "step": 10098 }, { "epoch": 0.78, "learning_rate": 2.3523181955539816e-07, "loss": 0.8415, "step": 10099 }, { "epoch": 0.78, "learning_rate": 2.350698732049028e-07, "loss": 1.1909, "step": 10100 }, { "epoch": 0.78, "learning_rate": 2.3490797519477712e-07, "loss": 1.1992, "step": 10101 }, { "epoch": 0.78, "learning_rate": 2.347461255352532e-07, "loss": 1.2515, "step": 10102 }, { "epoch": 0.78, "learning_rate": 2.3458432423655805e-07, "loss": 1.2014, "step": 10103 }, { "epoch": 0.78, "learning_rate": 2.3442257130891775e-07, "loss": 1.2037, "step": 10104 }, { "epoch": 0.78, "learning_rate": 2.342608667625544e-07, "loss": 0.8298, "step": 10105 }, { "epoch": 0.78, "learning_rate": 2.3409921060768635e-07, "loss": 1.1777, "step": 10106 }, { "epoch": 0.78, "learning_rate": 2.3393760285453011e-07, "loss": 1.2238, "step": 10107 }, { "epoch": 0.78, "learning_rate": 2.3377604351329862e-07, "loss": 1.2715, "step": 10108 }, { "epoch": 0.78, "learning_rate": 2.3361453259420162e-07, "loss": 1.2067, "step": 10109 }, { "epoch": 0.78, "learning_rate": 2.334530701074462e-07, "loss": 1.1396, "step": 10110 }, { "epoch": 0.78, "learning_rate": 2.3329165606323532e-07, "loss": 0.8527, "step": 10111 }, { "epoch": 0.78, "learning_rate": 2.3313029047177048e-07, "loss": 0.8282, "step": 10112 }, { "epoch": 0.78, "learning_rate": 2.3296897334324894e-07, "loss": 1.193, "step": 10113 }, { "epoch": 0.78, "learning_rate": 2.328077046878655e-07, "loss": 1.1959, "step": 10114 }, { "epoch": 0.79, "learning_rate": 2.3264648451581147e-07, "loss": 1.2276, "step": 10115 }, { "epoch": 0.79, "learning_rate": 2.3248531283727523e-07, "loss": 1.2285, "step": 10116 }, { "epoch": 0.79, "learning_rate": 2.3232418966244237e-07, "loss": 1.1877, "step": 10117 }, { "epoch": 0.79, "learning_rate": 2.321631150014951e-07, "loss": 1.1444, "step": 10118 }, { "epoch": 0.79, "learning_rate": 2.3200208886461258e-07, "loss": 1.2451, "step": 10119 }, { "epoch": 0.79, "learning_rate": 2.3184111126197114e-07, "loss": 1.1939, "step": 10120 }, { "epoch": 0.79, "learning_rate": 2.3168018220374374e-07, "loss": 0.8198, "step": 10121 }, { "epoch": 0.79, "learning_rate": 2.3151930170010047e-07, "loss": 1.174, "step": 10122 }, { "epoch": 0.79, "learning_rate": 2.3135846976120832e-07, "loss": 1.1928, "step": 10123 }, { "epoch": 0.79, "learning_rate": 2.3119768639723124e-07, "loss": 1.156, "step": 10124 }, { "epoch": 0.79, "learning_rate": 2.3103695161832994e-07, "loss": 1.1597, "step": 10125 }, { "epoch": 0.79, "learning_rate": 2.3087626543466233e-07, "loss": 1.2261, "step": 10126 }, { "epoch": 0.79, "learning_rate": 2.3071562785638288e-07, "loss": 1.1912, "step": 10127 }, { "epoch": 0.79, "learning_rate": 2.3055503889364347e-07, "loss": 1.2264, "step": 10128 }, { "epoch": 0.79, "learning_rate": 2.3039449855659255e-07, "loss": 1.1995, "step": 10129 }, { "epoch": 0.79, "learning_rate": 2.302340068553752e-07, "loss": 1.184, "step": 10130 }, { "epoch": 0.79, "learning_rate": 2.3007356380013477e-07, "loss": 1.1761, "step": 10131 }, { "epoch": 0.79, "learning_rate": 2.299131694010098e-07, "loss": 1.2176, "step": 10132 }, { "epoch": 0.79, "learning_rate": 2.2975282366813676e-07, "loss": 0.8272, "step": 10133 }, { "epoch": 0.79, "learning_rate": 2.2959252661164875e-07, "loss": 1.148, "step": 10134 }, { "epoch": 0.79, "learning_rate": 2.2943227824167567e-07, "loss": 1.2073, "step": 10135 }, { "epoch": 0.79, "learning_rate": 2.292720785683454e-07, "loss": 0.8484, "step": 10136 }, { "epoch": 0.79, "learning_rate": 2.2911192760178077e-07, "loss": 1.2277, "step": 10137 }, { "epoch": 0.79, "learning_rate": 2.2895182535210344e-07, "loss": 1.1804, "step": 10138 }, { "epoch": 0.79, "learning_rate": 2.2879177182943122e-07, "loss": 1.2206, "step": 10139 }, { "epoch": 0.79, "learning_rate": 2.2863176704387798e-07, "loss": 1.2095, "step": 10140 }, { "epoch": 0.79, "learning_rate": 2.2847181100555623e-07, "loss": 1.2164, "step": 10141 }, { "epoch": 0.79, "learning_rate": 2.283119037245741e-07, "loss": 1.1937, "step": 10142 }, { "epoch": 0.79, "learning_rate": 2.2815204521103725e-07, "loss": 1.2063, "step": 10143 }, { "epoch": 0.79, "learning_rate": 2.279922354750482e-07, "loss": 1.2323, "step": 10144 }, { "epoch": 0.79, "learning_rate": 2.2783247452670552e-07, "loss": 1.2006, "step": 10145 }, { "epoch": 0.79, "learning_rate": 2.2767276237610623e-07, "loss": 1.1731, "step": 10146 }, { "epoch": 0.79, "learning_rate": 2.2751309903334302e-07, "loss": 1.255, "step": 10147 }, { "epoch": 0.79, "learning_rate": 2.2735348450850632e-07, "loss": 1.2283, "step": 10148 }, { "epoch": 0.79, "learning_rate": 2.271939188116827e-07, "loss": 1.1944, "step": 10149 }, { "epoch": 0.79, "learning_rate": 2.2703440195295632e-07, "loss": 1.2271, "step": 10150 }, { "epoch": 0.79, "learning_rate": 2.2687493394240787e-07, "loss": 1.2334, "step": 10151 }, { "epoch": 0.79, "learning_rate": 2.2671551479011497e-07, "loss": 0.8706, "step": 10152 }, { "epoch": 0.79, "learning_rate": 2.2655614450615245e-07, "loss": 1.152, "step": 10153 }, { "epoch": 0.79, "learning_rate": 2.263968231005916e-07, "loss": 1.2102, "step": 10154 }, { "epoch": 0.79, "learning_rate": 2.26237550583501e-07, "loss": 1.1776, "step": 10155 }, { "epoch": 0.79, "learning_rate": 2.2607832696494611e-07, "loss": 1.1505, "step": 10156 }, { "epoch": 0.79, "learning_rate": 2.2591915225498902e-07, "loss": 1.2129, "step": 10157 }, { "epoch": 0.79, "learning_rate": 2.2576002646368896e-07, "loss": 1.2401, "step": 10158 }, { "epoch": 0.79, "learning_rate": 2.25600949601102e-07, "loss": 1.1869, "step": 10159 }, { "epoch": 0.79, "learning_rate": 2.2544192167728105e-07, "loss": 1.2122, "step": 10160 }, { "epoch": 0.79, "learning_rate": 2.252829427022762e-07, "loss": 1.195, "step": 10161 }, { "epoch": 0.79, "learning_rate": 2.2512401268613413e-07, "loss": 1.1423, "step": 10162 }, { "epoch": 0.79, "learning_rate": 2.2496513163889853e-07, "loss": 1.2173, "step": 10163 }, { "epoch": 0.79, "learning_rate": 2.2480629957060981e-07, "loss": 1.2385, "step": 10164 }, { "epoch": 0.79, "learning_rate": 2.246475164913063e-07, "loss": 1.2042, "step": 10165 }, { "epoch": 0.79, "learning_rate": 2.2448878241102154e-07, "loss": 1.1898, "step": 10166 }, { "epoch": 0.79, "learning_rate": 2.2433009733978714e-07, "loss": 1.2316, "step": 10167 }, { "epoch": 0.79, "learning_rate": 2.2417146128763143e-07, "loss": 1.2026, "step": 10168 }, { "epoch": 0.79, "learning_rate": 2.2401287426457916e-07, "loss": 1.1439, "step": 10169 }, { "epoch": 0.79, "learning_rate": 2.238543362806532e-07, "loss": 1.1996, "step": 10170 }, { "epoch": 0.79, "learning_rate": 2.2369584734587176e-07, "loss": 1.1508, "step": 10171 }, { "epoch": 0.79, "learning_rate": 2.235374074702505e-07, "loss": 1.1365, "step": 10172 }, { "epoch": 0.79, "learning_rate": 2.2337901666380321e-07, "loss": 1.1822, "step": 10173 }, { "epoch": 0.79, "learning_rate": 2.2322067493653817e-07, "loss": 1.2188, "step": 10174 }, { "epoch": 0.79, "learning_rate": 2.230623822984633e-07, "loss": 1.2054, "step": 10175 }, { "epoch": 0.79, "learning_rate": 2.2290413875958092e-07, "loss": 1.1947, "step": 10176 }, { "epoch": 0.79, "learning_rate": 2.2274594432989147e-07, "loss": 1.2037, "step": 10177 }, { "epoch": 0.79, "learning_rate": 2.2258779901939294e-07, "loss": 1.2472, "step": 10178 }, { "epoch": 0.79, "learning_rate": 2.224297028380786e-07, "loss": 1.1984, "step": 10179 }, { "epoch": 0.79, "learning_rate": 2.2227165579593999e-07, "loss": 1.2025, "step": 10180 }, { "epoch": 0.79, "learning_rate": 2.221136579029649e-07, "loss": 1.1701, "step": 10181 }, { "epoch": 0.79, "learning_rate": 2.2195570916913798e-07, "loss": 1.165, "step": 10182 }, { "epoch": 0.79, "learning_rate": 2.2179780960444116e-07, "loss": 1.2052, "step": 10183 }, { "epoch": 0.79, "learning_rate": 2.216399592188528e-07, "loss": 1.141, "step": 10184 }, { "epoch": 0.79, "learning_rate": 2.214821580223486e-07, "loss": 1.2046, "step": 10185 }, { "epoch": 0.79, "learning_rate": 2.2132440602490078e-07, "loss": 1.1447, "step": 10186 }, { "epoch": 0.79, "learning_rate": 2.2116670323647857e-07, "loss": 0.8532, "step": 10187 }, { "epoch": 0.79, "learning_rate": 2.2100904966704825e-07, "loss": 1.1816, "step": 10188 }, { "epoch": 0.79, "learning_rate": 2.2085144532657286e-07, "loss": 0.82, "step": 10189 }, { "epoch": 0.79, "learning_rate": 2.2069389022501216e-07, "loss": 1.202, "step": 10190 }, { "epoch": 0.79, "learning_rate": 2.2053638437232326e-07, "loss": 1.2125, "step": 10191 }, { "epoch": 0.79, "learning_rate": 2.2037892777845956e-07, "loss": 1.2126, "step": 10192 }, { "epoch": 0.79, "learning_rate": 2.2022152045337182e-07, "loss": 0.8178, "step": 10193 }, { "epoch": 0.79, "learning_rate": 2.2006416240700755e-07, "loss": 1.2228, "step": 10194 }, { "epoch": 0.79, "learning_rate": 2.1990685364931095e-07, "loss": 1.2707, "step": 10195 }, { "epoch": 0.79, "learning_rate": 2.1974959419022354e-07, "loss": 0.8453, "step": 10196 }, { "epoch": 0.79, "learning_rate": 2.1959238403968316e-07, "loss": 0.8026, "step": 10197 }, { "epoch": 0.79, "learning_rate": 2.1943522320762508e-07, "loss": 1.223, "step": 10198 }, { "epoch": 0.79, "learning_rate": 2.1927811170398104e-07, "loss": 1.1999, "step": 10199 }, { "epoch": 0.79, "learning_rate": 2.1912104953867993e-07, "loss": 1.2092, "step": 10200 }, { "epoch": 0.79, "learning_rate": 2.1896403672164732e-07, "loss": 1.1674, "step": 10201 }, { "epoch": 0.79, "learning_rate": 2.188070732628059e-07, "loss": 1.2184, "step": 10202 }, { "epoch": 0.79, "learning_rate": 2.1865015917207462e-07, "loss": 1.2096, "step": 10203 }, { "epoch": 0.79, "learning_rate": 2.1849329445937083e-07, "loss": 1.1773, "step": 10204 }, { "epoch": 0.79, "learning_rate": 2.1833647913460684e-07, "loss": 1.1791, "step": 10205 }, { "epoch": 0.79, "learning_rate": 2.1817971320769256e-07, "loss": 1.2389, "step": 10206 }, { "epoch": 0.79, "learning_rate": 2.1802299668853596e-07, "loss": 1.2118, "step": 10207 }, { "epoch": 0.79, "learning_rate": 2.1786632958703965e-07, "loss": 1.1831, "step": 10208 }, { "epoch": 0.79, "learning_rate": 2.177097119131055e-07, "loss": 1.1674, "step": 10209 }, { "epoch": 0.79, "learning_rate": 2.1755314367663025e-07, "loss": 0.8393, "step": 10210 }, { "epoch": 0.79, "learning_rate": 2.1739662488750832e-07, "loss": 1.1615, "step": 10211 }, { "epoch": 0.79, "learning_rate": 2.1724015555563192e-07, "loss": 0.8362, "step": 10212 }, { "epoch": 0.79, "learning_rate": 2.1708373569088812e-07, "loss": 1.187, "step": 10213 }, { "epoch": 0.79, "learning_rate": 2.169273653031628e-07, "loss": 1.2129, "step": 10214 }, { "epoch": 0.79, "learning_rate": 2.1677104440233785e-07, "loss": 1.2355, "step": 10215 }, { "epoch": 0.79, "learning_rate": 2.1661477299829146e-07, "loss": 1.1802, "step": 10216 }, { "epoch": 0.79, "learning_rate": 2.164585511009003e-07, "loss": 1.2242, "step": 10217 }, { "epoch": 0.79, "learning_rate": 2.1630237872003586e-07, "loss": 1.1283, "step": 10218 }, { "epoch": 0.79, "learning_rate": 2.1614625586556844e-07, "loss": 0.816, "step": 10219 }, { "epoch": 0.79, "learning_rate": 2.1599018254736422e-07, "loss": 1.1898, "step": 10220 }, { "epoch": 0.79, "learning_rate": 2.1583415877528565e-07, "loss": 1.1405, "step": 10221 }, { "epoch": 0.79, "learning_rate": 2.1567818455919372e-07, "loss": 1.2142, "step": 10222 }, { "epoch": 0.79, "learning_rate": 2.1552225990894478e-07, "loss": 1.1873, "step": 10223 }, { "epoch": 0.79, "learning_rate": 2.1536638483439295e-07, "loss": 1.1899, "step": 10224 }, { "epoch": 0.79, "learning_rate": 2.1521055934538856e-07, "loss": 1.2336, "step": 10225 }, { "epoch": 0.79, "learning_rate": 2.1505478345177929e-07, "loss": 1.1562, "step": 10226 }, { "epoch": 0.79, "learning_rate": 2.1489905716340962e-07, "loss": 1.2273, "step": 10227 }, { "epoch": 0.79, "learning_rate": 2.1474338049012064e-07, "loss": 1.1111, "step": 10228 }, { "epoch": 0.79, "learning_rate": 2.1458775344175039e-07, "loss": 0.8382, "step": 10229 }, { "epoch": 0.79, "learning_rate": 2.1443217602813401e-07, "loss": 1.155, "step": 10230 }, { "epoch": 0.79, "learning_rate": 2.142766482591034e-07, "loss": 1.1861, "step": 10231 }, { "epoch": 0.79, "learning_rate": 2.1412117014448706e-07, "loss": 1.1401, "step": 10232 }, { "epoch": 0.79, "learning_rate": 2.1396574169411064e-07, "loss": 0.8435, "step": 10233 }, { "epoch": 0.79, "learning_rate": 2.1381036291779654e-07, "loss": 1.2204, "step": 10234 }, { "epoch": 0.79, "learning_rate": 2.136550338253641e-07, "loss": 1.1483, "step": 10235 }, { "epoch": 0.79, "learning_rate": 2.1349975442662949e-07, "loss": 0.8438, "step": 10236 }, { "epoch": 0.79, "learning_rate": 2.1334452473140562e-07, "loss": 1.2088, "step": 10237 }, { "epoch": 0.79, "learning_rate": 2.1318934474950244e-07, "loss": 1.2158, "step": 10238 }, { "epoch": 0.79, "learning_rate": 2.1303421449072656e-07, "loss": 0.8686, "step": 10239 }, { "epoch": 0.79, "learning_rate": 2.1287913396488145e-07, "loss": 1.1825, "step": 10240 }, { "epoch": 0.79, "learning_rate": 2.1272410318176826e-07, "loss": 1.1701, "step": 10241 }, { "epoch": 0.79, "learning_rate": 2.1256912215118339e-07, "loss": 1.1975, "step": 10242 }, { "epoch": 0.79, "learning_rate": 2.1241419088292146e-07, "loss": 1.1862, "step": 10243 }, { "epoch": 0.8, "learning_rate": 2.1225930938677327e-07, "loss": 1.2275, "step": 10244 }, { "epoch": 0.8, "learning_rate": 2.121044776725266e-07, "loss": 1.2011, "step": 10245 }, { "epoch": 0.8, "learning_rate": 2.119496957499669e-07, "loss": 1.2156, "step": 10246 }, { "epoch": 0.8, "learning_rate": 2.117949636288745e-07, "loss": 1.211, "step": 10247 }, { "epoch": 0.8, "learning_rate": 2.116402813190289e-07, "loss": 1.1814, "step": 10248 }, { "epoch": 0.8, "learning_rate": 2.1148564883020525e-07, "loss": 1.1699, "step": 10249 }, { "epoch": 0.8, "learning_rate": 2.1133106617217478e-07, "loss": 1.1913, "step": 10250 }, { "epoch": 0.8, "learning_rate": 2.1117653335470763e-07, "loss": 1.1868, "step": 10251 }, { "epoch": 0.8, "learning_rate": 2.110220503875686e-07, "loss": 1.1839, "step": 10252 }, { "epoch": 0.8, "learning_rate": 2.1086761728052105e-07, "loss": 1.1758, "step": 10253 }, { "epoch": 0.8, "learning_rate": 2.1071323404332464e-07, "loss": 1.2019, "step": 10254 }, { "epoch": 0.8, "learning_rate": 2.1055890068573477e-07, "loss": 1.2102, "step": 10255 }, { "epoch": 0.8, "learning_rate": 2.1040461721750556e-07, "loss": 0.8016, "step": 10256 }, { "epoch": 0.8, "learning_rate": 2.1025038364838687e-07, "loss": 1.1552, "step": 10257 }, { "epoch": 0.8, "learning_rate": 2.1009619998812543e-07, "loss": 1.1618, "step": 10258 }, { "epoch": 0.8, "learning_rate": 2.0994206624646527e-07, "loss": 1.1565, "step": 10259 }, { "epoch": 0.8, "learning_rate": 2.0978798243314667e-07, "loss": 1.2373, "step": 10260 }, { "epoch": 0.8, "learning_rate": 2.096339485579074e-07, "loss": 1.18, "step": 10261 }, { "epoch": 0.8, "learning_rate": 2.0947996463048156e-07, "loss": 1.2047, "step": 10262 }, { "epoch": 0.8, "learning_rate": 2.093260306606003e-07, "loss": 1.2289, "step": 10263 }, { "epoch": 0.8, "learning_rate": 2.0917214665799165e-07, "loss": 1.2099, "step": 10264 }, { "epoch": 0.8, "learning_rate": 2.0901831263238034e-07, "loss": 1.1362, "step": 10265 }, { "epoch": 0.8, "learning_rate": 2.0886452859348824e-07, "loss": 1.2483, "step": 10266 }, { "epoch": 0.8, "learning_rate": 2.0871079455103357e-07, "loss": 1.2159, "step": 10267 }, { "epoch": 0.8, "learning_rate": 2.085571105147319e-07, "loss": 1.2374, "step": 10268 }, { "epoch": 0.8, "learning_rate": 2.0840347649429534e-07, "loss": 1.2393, "step": 10269 }, { "epoch": 0.8, "learning_rate": 2.0824989249943303e-07, "loss": 1.2237, "step": 10270 }, { "epoch": 0.8, "learning_rate": 2.0809635853985053e-07, "loss": 1.1592, "step": 10271 }, { "epoch": 0.8, "learning_rate": 2.0794287462525083e-07, "loss": 0.8344, "step": 10272 }, { "epoch": 0.8, "learning_rate": 2.0778944076533343e-07, "loss": 1.1816, "step": 10273 }, { "epoch": 0.8, "learning_rate": 2.0763605696979458e-07, "loss": 0.7969, "step": 10274 }, { "epoch": 0.8, "learning_rate": 2.074827232483275e-07, "loss": 1.1959, "step": 10275 }, { "epoch": 0.8, "learning_rate": 2.0732943961062231e-07, "loss": 1.1798, "step": 10276 }, { "epoch": 0.8, "learning_rate": 2.07176206066366e-07, "loss": 1.1929, "step": 10277 }, { "epoch": 0.8, "learning_rate": 2.0702302262524207e-07, "loss": 1.2168, "step": 10278 }, { "epoch": 0.8, "learning_rate": 2.0686988929693083e-07, "loss": 1.2185, "step": 10279 }, { "epoch": 0.8, "learning_rate": 2.0671680609111064e-07, "loss": 1.1579, "step": 10280 }, { "epoch": 0.8, "learning_rate": 2.0656377301745475e-07, "loss": 1.1774, "step": 10281 }, { "epoch": 0.8, "learning_rate": 2.0641079008563433e-07, "loss": 1.1662, "step": 10282 }, { "epoch": 0.8, "learning_rate": 2.0625785730531787e-07, "loss": 1.2046, "step": 10283 }, { "epoch": 0.8, "learning_rate": 2.0610497468616915e-07, "loss": 1.1902, "step": 10284 }, { "epoch": 0.8, "learning_rate": 2.0595214223785085e-07, "loss": 1.2214, "step": 10285 }, { "epoch": 0.8, "learning_rate": 2.057993599700204e-07, "loss": 1.2437, "step": 10286 }, { "epoch": 0.8, "learning_rate": 2.05646627892333e-07, "loss": 1.1723, "step": 10287 }, { "epoch": 0.8, "learning_rate": 2.0549394601444158e-07, "loss": 1.1681, "step": 10288 }, { "epoch": 0.8, "learning_rate": 2.053413143459939e-07, "loss": 1.1926, "step": 10289 }, { "epoch": 0.8, "learning_rate": 2.051887328966363e-07, "loss": 1.2074, "step": 10290 }, { "epoch": 0.8, "learning_rate": 2.0503620167601122e-07, "loss": 1.1268, "step": 10291 }, { "epoch": 0.8, "learning_rate": 2.0488372069375782e-07, "loss": 1.1709, "step": 10292 }, { "epoch": 0.8, "learning_rate": 2.0473128995951262e-07, "loss": 1.1512, "step": 10293 }, { "epoch": 0.8, "learning_rate": 2.0457890948290779e-07, "loss": 1.1898, "step": 10294 }, { "epoch": 0.8, "learning_rate": 2.0442657927357387e-07, "loss": 1.1878, "step": 10295 }, { "epoch": 0.8, "learning_rate": 2.0427429934113738e-07, "loss": 1.178, "step": 10296 }, { "epoch": 0.8, "learning_rate": 2.041220696952216e-07, "loss": 1.2214, "step": 10297 }, { "epoch": 0.8, "learning_rate": 2.0396989034544697e-07, "loss": 1.2517, "step": 10298 }, { "epoch": 0.8, "learning_rate": 2.0381776130143048e-07, "loss": 1.1809, "step": 10299 }, { "epoch": 0.8, "learning_rate": 2.0366568257278594e-07, "loss": 1.1832, "step": 10300 }, { "epoch": 0.8, "learning_rate": 2.0351365416912436e-07, "loss": 1.2196, "step": 10301 }, { "epoch": 0.8, "learning_rate": 2.0336167610005305e-07, "loss": 0.8068, "step": 10302 }, { "epoch": 0.8, "learning_rate": 2.0320974837517645e-07, "loss": 1.2161, "step": 10303 }, { "epoch": 0.8, "learning_rate": 2.0305787100409577e-07, "loss": 0.7936, "step": 10304 }, { "epoch": 0.8, "learning_rate": 2.029060439964091e-07, "loss": 1.1624, "step": 10305 }, { "epoch": 0.8, "learning_rate": 2.0275426736171119e-07, "loss": 1.2095, "step": 10306 }, { "epoch": 0.8, "learning_rate": 2.0260254110959374e-07, "loss": 1.2167, "step": 10307 }, { "epoch": 0.8, "learning_rate": 2.0245086524964515e-07, "loss": 1.2545, "step": 10308 }, { "epoch": 0.8, "learning_rate": 2.022992397914507e-07, "loss": 1.1756, "step": 10309 }, { "epoch": 0.8, "learning_rate": 2.0214766474459243e-07, "loss": 0.7866, "step": 10310 }, { "epoch": 0.8, "learning_rate": 2.019961401186494e-07, "loss": 1.2194, "step": 10311 }, { "epoch": 0.8, "learning_rate": 2.0184466592319716e-07, "loss": 1.1747, "step": 10312 }, { "epoch": 0.8, "learning_rate": 2.0169324216780814e-07, "loss": 1.2139, "step": 10313 }, { "epoch": 0.8, "learning_rate": 2.0154186886205237e-07, "loss": 1.1798, "step": 10314 }, { "epoch": 0.8, "learning_rate": 2.0139054601549533e-07, "loss": 1.1815, "step": 10315 }, { "epoch": 0.8, "learning_rate": 2.0123927363769977e-07, "loss": 0.8354, "step": 10316 }, { "epoch": 0.8, "learning_rate": 2.0108805173822652e-07, "loss": 1.2383, "step": 10317 }, { "epoch": 0.8, "learning_rate": 2.0093688032663082e-07, "loss": 1.176, "step": 10318 }, { "epoch": 0.8, "learning_rate": 2.0078575941246744e-07, "loss": 1.1977, "step": 10319 }, { "epoch": 0.8, "learning_rate": 2.0063468900528557e-07, "loss": 1.2153, "step": 10320 }, { "epoch": 0.8, "learning_rate": 2.0048366911463233e-07, "loss": 1.2498, "step": 10321 }, { "epoch": 0.8, "learning_rate": 2.003326997500523e-07, "loss": 0.7856, "step": 10322 }, { "epoch": 0.8, "learning_rate": 2.0018178092108508e-07, "loss": 1.1617, "step": 10323 }, { "epoch": 0.8, "learning_rate": 2.0003091263726878e-07, "loss": 1.2098, "step": 10324 }, { "epoch": 0.8, "learning_rate": 1.9988009490813784e-07, "loss": 0.8433, "step": 10325 }, { "epoch": 0.8, "learning_rate": 1.9972932774322239e-07, "loss": 1.2392, "step": 10326 }, { "epoch": 0.8, "learning_rate": 1.9957861115205132e-07, "loss": 1.183, "step": 10327 }, { "epoch": 0.8, "learning_rate": 1.994279451441483e-07, "loss": 1.2072, "step": 10328 }, { "epoch": 0.8, "learning_rate": 1.9927732972903556e-07, "loss": 1.1932, "step": 10329 }, { "epoch": 0.8, "learning_rate": 1.9912676491623125e-07, "loss": 1.1855, "step": 10330 }, { "epoch": 0.8, "learning_rate": 1.9897625071524992e-07, "loss": 1.2398, "step": 10331 }, { "epoch": 0.8, "learning_rate": 1.9882578713560394e-07, "loss": 1.2152, "step": 10332 }, { "epoch": 0.8, "learning_rate": 1.986753741868018e-07, "loss": 1.2107, "step": 10333 }, { "epoch": 0.8, "learning_rate": 1.9852501187834903e-07, "loss": 1.2404, "step": 10334 }, { "epoch": 0.8, "learning_rate": 1.983747002197479e-07, "loss": 1.2426, "step": 10335 }, { "epoch": 0.8, "learning_rate": 1.9822443922049736e-07, "loss": 1.1898, "step": 10336 }, { "epoch": 0.8, "learning_rate": 1.9807422889009328e-07, "loss": 0.8503, "step": 10337 }, { "epoch": 0.8, "learning_rate": 1.9792406923802852e-07, "loss": 1.1623, "step": 10338 }, { "epoch": 0.8, "learning_rate": 1.9777396027379233e-07, "loss": 1.1682, "step": 10339 }, { "epoch": 0.8, "learning_rate": 1.9762390200687106e-07, "loss": 1.1781, "step": 10340 }, { "epoch": 0.8, "learning_rate": 1.9747389444674767e-07, "loss": 1.149, "step": 10341 }, { "epoch": 0.8, "learning_rate": 1.9732393760290223e-07, "loss": 1.1945, "step": 10342 }, { "epoch": 0.8, "learning_rate": 1.9717403148481116e-07, "loss": 1.1872, "step": 10343 }, { "epoch": 0.8, "learning_rate": 1.970241761019479e-07, "loss": 1.154, "step": 10344 }, { "epoch": 0.8, "learning_rate": 1.9687437146378282e-07, "loss": 1.2023, "step": 10345 }, { "epoch": 0.8, "learning_rate": 1.9672461757978288e-07, "loss": 1.2071, "step": 10346 }, { "epoch": 0.8, "learning_rate": 1.9657491445941178e-07, "loss": 1.2056, "step": 10347 }, { "epoch": 0.8, "learning_rate": 1.964252621121304e-07, "loss": 1.1583, "step": 10348 }, { "epoch": 0.8, "learning_rate": 1.9627566054739585e-07, "loss": 1.1588, "step": 10349 }, { "epoch": 0.8, "learning_rate": 1.961261097746625e-07, "loss": 1.2405, "step": 10350 }, { "epoch": 0.8, "learning_rate": 1.9597660980338126e-07, "loss": 1.1466, "step": 10351 }, { "epoch": 0.8, "learning_rate": 1.9582716064299986e-07, "loss": 1.2281, "step": 10352 }, { "epoch": 0.8, "learning_rate": 1.9567776230296306e-07, "loss": 1.2302, "step": 10353 }, { "epoch": 0.8, "learning_rate": 1.9552841479271198e-07, "loss": 1.1575, "step": 10354 }, { "epoch": 0.8, "learning_rate": 1.953791181216845e-07, "loss": 0.8222, "step": 10355 }, { "epoch": 0.8, "learning_rate": 1.9522987229931653e-07, "loss": 1.2499, "step": 10356 }, { "epoch": 0.8, "learning_rate": 1.9508067733503853e-07, "loss": 0.8289, "step": 10357 }, { "epoch": 0.8, "learning_rate": 1.9493153323827994e-07, "loss": 0.8346, "step": 10358 }, { "epoch": 0.8, "learning_rate": 1.9478244001846589e-07, "loss": 1.2336, "step": 10359 }, { "epoch": 0.8, "learning_rate": 1.9463339768501773e-07, "loss": 1.2013, "step": 10360 }, { "epoch": 0.8, "learning_rate": 1.944844062473554e-07, "loss": 1.1754, "step": 10361 }, { "epoch": 0.8, "learning_rate": 1.9433546571489334e-07, "loss": 1.2302, "step": 10362 }, { "epoch": 0.8, "learning_rate": 1.9418657609704492e-07, "loss": 1.1532, "step": 10363 }, { "epoch": 0.8, "learning_rate": 1.9403773740321927e-07, "loss": 1.1412, "step": 10364 }, { "epoch": 0.8, "learning_rate": 1.9388894964282153e-07, "loss": 1.1788, "step": 10365 }, { "epoch": 0.8, "learning_rate": 1.9374021282525532e-07, "loss": 1.1961, "step": 10366 }, { "epoch": 0.8, "learning_rate": 1.9359152695991997e-07, "loss": 1.258, "step": 10367 }, { "epoch": 0.8, "learning_rate": 1.9344289205621167e-07, "loss": 1.1698, "step": 10368 }, { "epoch": 0.8, "learning_rate": 1.9329430812352376e-07, "loss": 1.1901, "step": 10369 }, { "epoch": 0.8, "learning_rate": 1.9314577517124553e-07, "loss": 1.1977, "step": 10370 }, { "epoch": 0.8, "learning_rate": 1.9299729320876434e-07, "loss": 1.1851, "step": 10371 }, { "epoch": 0.8, "learning_rate": 1.9284886224546325e-07, "loss": 1.1757, "step": 10372 }, { "epoch": 0.81, "learning_rate": 1.9270048229072267e-07, "loss": 1.1902, "step": 10373 }, { "epoch": 0.81, "learning_rate": 1.925521533539194e-07, "loss": 1.2065, "step": 10374 }, { "epoch": 0.81, "learning_rate": 1.9240387544442726e-07, "loss": 1.1833, "step": 10375 }, { "epoch": 0.81, "learning_rate": 1.9225564857161681e-07, "loss": 1.1897, "step": 10376 }, { "epoch": 0.81, "learning_rate": 1.921074727448554e-07, "loss": 1.1323, "step": 10377 }, { "epoch": 0.81, "learning_rate": 1.919593479735071e-07, "loss": 1.1833, "step": 10378 }, { "epoch": 0.81, "learning_rate": 1.9181127426693278e-07, "loss": 1.1935, "step": 10379 }, { "epoch": 0.81, "learning_rate": 1.9166325163449004e-07, "loss": 1.1798, "step": 10380 }, { "epoch": 0.81, "learning_rate": 1.915152800855333e-07, "loss": 0.8004, "step": 10381 }, { "epoch": 0.81, "learning_rate": 1.9136735962941375e-07, "loss": 1.1189, "step": 10382 }, { "epoch": 0.81, "learning_rate": 1.9121949027547946e-07, "loss": 1.2287, "step": 10383 }, { "epoch": 0.81, "learning_rate": 1.9107167203307484e-07, "loss": 1.232, "step": 10384 }, { "epoch": 0.81, "learning_rate": 1.9092390491154165e-07, "loss": 1.1811, "step": 10385 }, { "epoch": 0.81, "learning_rate": 1.9077618892021808e-07, "loss": 1.1608, "step": 10386 }, { "epoch": 0.81, "learning_rate": 1.906285240684392e-07, "loss": 1.19, "step": 10387 }, { "epoch": 0.81, "learning_rate": 1.904809103655367e-07, "loss": 1.2099, "step": 10388 }, { "epoch": 0.81, "learning_rate": 1.9033334782083899e-07, "loss": 1.17, "step": 10389 }, { "epoch": 0.81, "learning_rate": 1.9018583644367215e-07, "loss": 1.2097, "step": 10390 }, { "epoch": 0.81, "learning_rate": 1.900383762433575e-07, "loss": 1.1652, "step": 10391 }, { "epoch": 0.81, "learning_rate": 1.8989096722921381e-07, "loss": 1.1686, "step": 10392 }, { "epoch": 0.81, "learning_rate": 1.8974360941055768e-07, "loss": 1.1832, "step": 10393 }, { "epoch": 0.81, "learning_rate": 1.8959630279670025e-07, "loss": 1.2098, "step": 10394 }, { "epoch": 0.81, "learning_rate": 1.8944904739695188e-07, "loss": 1.1819, "step": 10395 }, { "epoch": 0.81, "learning_rate": 1.893018432206177e-07, "loss": 1.1804, "step": 10396 }, { "epoch": 0.81, "learning_rate": 1.8915469027700038e-07, "loss": 1.2009, "step": 10397 }, { "epoch": 0.81, "learning_rate": 1.890075885754e-07, "loss": 1.2014, "step": 10398 }, { "epoch": 0.81, "learning_rate": 1.8886053812511204e-07, "loss": 0.8322, "step": 10399 }, { "epoch": 0.81, "learning_rate": 1.8871353893542996e-07, "loss": 1.1978, "step": 10400 }, { "epoch": 0.81, "learning_rate": 1.8856659101564332e-07, "loss": 1.2353, "step": 10401 }, { "epoch": 0.81, "learning_rate": 1.8841969437503857e-07, "loss": 1.1958, "step": 10402 }, { "epoch": 0.81, "learning_rate": 1.8827284902289942e-07, "loss": 0.8256, "step": 10403 }, { "epoch": 0.81, "learning_rate": 1.8812605496850487e-07, "loss": 1.2188, "step": 10404 }, { "epoch": 0.81, "learning_rate": 1.879793122211326e-07, "loss": 1.1859, "step": 10405 }, { "epoch": 0.81, "learning_rate": 1.878326207900558e-07, "loss": 1.2364, "step": 10406 }, { "epoch": 0.81, "learning_rate": 1.8768598068454477e-07, "loss": 1.2364, "step": 10407 }, { "epoch": 0.81, "learning_rate": 1.8753939191386658e-07, "loss": 1.1437, "step": 10408 }, { "epoch": 0.81, "learning_rate": 1.8739285448728505e-07, "loss": 1.2061, "step": 10409 }, { "epoch": 0.81, "learning_rate": 1.8724636841406061e-07, "loss": 1.1743, "step": 10410 }, { "epoch": 0.81, "learning_rate": 1.8709993370345067e-07, "loss": 1.1753, "step": 10411 }, { "epoch": 0.81, "learning_rate": 1.8695355036470928e-07, "loss": 1.2101, "step": 10412 }, { "epoch": 0.81, "learning_rate": 1.8680721840708725e-07, "loss": 1.2063, "step": 10413 }, { "epoch": 0.81, "learning_rate": 1.8666093783983205e-07, "loss": 1.212, "step": 10414 }, { "epoch": 0.81, "learning_rate": 1.8651470867218811e-07, "loss": 1.2081, "step": 10415 }, { "epoch": 0.81, "learning_rate": 1.8636853091339656e-07, "loss": 1.1834, "step": 10416 }, { "epoch": 0.81, "learning_rate": 1.86222404572695e-07, "loss": 1.2538, "step": 10417 }, { "epoch": 0.81, "learning_rate": 1.8607632965931807e-07, "loss": 1.152, "step": 10418 }, { "epoch": 0.81, "learning_rate": 1.8593030618249727e-07, "loss": 1.1815, "step": 10419 }, { "epoch": 0.81, "learning_rate": 1.8578433415146043e-07, "loss": 1.1149, "step": 10420 }, { "epoch": 0.81, "learning_rate": 1.8563841357543263e-07, "loss": 1.1527, "step": 10421 }, { "epoch": 0.81, "learning_rate": 1.8549254446363515e-07, "loss": 1.2152, "step": 10422 }, { "epoch": 0.81, "learning_rate": 1.853467268252863e-07, "loss": 1.1367, "step": 10423 }, { "epoch": 0.81, "learning_rate": 1.8520096066960168e-07, "loss": 1.1927, "step": 10424 }, { "epoch": 0.81, "learning_rate": 1.850552460057926e-07, "loss": 0.8535, "step": 10425 }, { "epoch": 0.81, "learning_rate": 1.8490958284306756e-07, "loss": 1.1712, "step": 10426 }, { "epoch": 0.81, "learning_rate": 1.8476397119063215e-07, "loss": 1.1909, "step": 10427 }, { "epoch": 0.81, "learning_rate": 1.8461841105768795e-07, "loss": 1.1628, "step": 10428 }, { "epoch": 0.81, "learning_rate": 1.844729024534346e-07, "loss": 1.2179, "step": 10429 }, { "epoch": 0.81, "learning_rate": 1.843274453870668e-07, "loss": 1.198, "step": 10430 }, { "epoch": 0.81, "learning_rate": 1.8418203986777703e-07, "loss": 0.856, "step": 10431 }, { "epoch": 0.81, "learning_rate": 1.8403668590475485e-07, "loss": 1.2463, "step": 10432 }, { "epoch": 0.81, "learning_rate": 1.8389138350718504e-07, "loss": 1.1607, "step": 10433 }, { "epoch": 0.81, "learning_rate": 1.8374613268425088e-07, "loss": 1.2363, "step": 10434 }, { "epoch": 0.81, "learning_rate": 1.8360093344513172e-07, "loss": 1.1825, "step": 10435 }, { "epoch": 0.81, "learning_rate": 1.8345578579900267e-07, "loss": 1.1668, "step": 10436 }, { "epoch": 0.81, "learning_rate": 1.833106897550375e-07, "loss": 1.2252, "step": 10437 }, { "epoch": 0.81, "learning_rate": 1.8316564532240464e-07, "loss": 1.2465, "step": 10438 }, { "epoch": 0.81, "learning_rate": 1.830206525102711e-07, "loss": 1.1545, "step": 10439 }, { "epoch": 0.81, "learning_rate": 1.8287571132779967e-07, "loss": 0.8501, "step": 10440 }, { "epoch": 0.81, "learning_rate": 1.8273082178414944e-07, "loss": 0.8542, "step": 10441 }, { "epoch": 0.81, "learning_rate": 1.8258598388847746e-07, "loss": 1.2508, "step": 10442 }, { "epoch": 0.81, "learning_rate": 1.824411976499367e-07, "loss": 1.1869, "step": 10443 }, { "epoch": 0.81, "learning_rate": 1.82296463077677e-07, "loss": 1.2263, "step": 10444 }, { "epoch": 0.81, "learning_rate": 1.8215178018084497e-07, "loss": 1.2453, "step": 10445 }, { "epoch": 0.81, "learning_rate": 1.8200714896858406e-07, "loss": 1.1618, "step": 10446 }, { "epoch": 0.81, "learning_rate": 1.818625694500342e-07, "loss": 1.1835, "step": 10447 }, { "epoch": 0.81, "learning_rate": 1.8171804163433236e-07, "loss": 1.1979, "step": 10448 }, { "epoch": 0.81, "learning_rate": 1.8157356553061198e-07, "loss": 0.842, "step": 10449 }, { "epoch": 0.81, "learning_rate": 1.8142914114800335e-07, "loss": 1.2522, "step": 10450 }, { "epoch": 0.81, "learning_rate": 1.8128476849563344e-07, "loss": 1.2196, "step": 10451 }, { "epoch": 0.81, "learning_rate": 1.8114044758262613e-07, "loss": 0.8785, "step": 10452 }, { "epoch": 0.81, "learning_rate": 1.809961784181019e-07, "loss": 1.2285, "step": 10453 }, { "epoch": 0.81, "learning_rate": 1.8085196101117785e-07, "loss": 1.2001, "step": 10454 }, { "epoch": 0.81, "learning_rate": 1.807077953709678e-07, "loss": 1.0853, "step": 10455 }, { "epoch": 0.81, "learning_rate": 1.8056368150658275e-07, "loss": 1.1515, "step": 10456 }, { "epoch": 0.81, "learning_rate": 1.804196194271298e-07, "loss": 0.8401, "step": 10457 }, { "epoch": 0.81, "learning_rate": 1.8027560914171325e-07, "loss": 1.1399, "step": 10458 }, { "epoch": 0.81, "learning_rate": 1.801316506594338e-07, "loss": 1.1544, "step": 10459 }, { "epoch": 0.81, "learning_rate": 1.799877439893891e-07, "loss": 1.1634, "step": 10460 }, { "epoch": 0.81, "learning_rate": 1.798438891406735e-07, "loss": 1.1862, "step": 10461 }, { "epoch": 0.81, "learning_rate": 1.7970008612237786e-07, "loss": 1.2043, "step": 10462 }, { "epoch": 0.81, "learning_rate": 1.7955633494359e-07, "loss": 1.194, "step": 10463 }, { "epoch": 0.81, "learning_rate": 1.794126356133946e-07, "loss": 1.2418, "step": 10464 }, { "epoch": 0.81, "learning_rate": 1.7926898814087222e-07, "loss": 0.7884, "step": 10465 }, { "epoch": 0.81, "learning_rate": 1.7912539253510184e-07, "loss": 1.1721, "step": 10466 }, { "epoch": 0.81, "learning_rate": 1.7898184880515687e-07, "loss": 1.1562, "step": 10467 }, { "epoch": 0.81, "learning_rate": 1.7883835696010962e-07, "loss": 1.189, "step": 10468 }, { "epoch": 0.81, "learning_rate": 1.7869491700902806e-07, "loss": 1.1572, "step": 10469 }, { "epoch": 0.81, "learning_rate": 1.7855152896097625e-07, "loss": 1.2026, "step": 10470 }, { "epoch": 0.81, "learning_rate": 1.7840819282501684e-07, "loss": 1.1511, "step": 10471 }, { "epoch": 0.81, "learning_rate": 1.7826490861020682e-07, "loss": 1.1959, "step": 10472 }, { "epoch": 0.81, "learning_rate": 1.7812167632560203e-07, "loss": 1.2323, "step": 10473 }, { "epoch": 0.81, "learning_rate": 1.7797849598025427e-07, "loss": 1.2052, "step": 10474 }, { "epoch": 0.81, "learning_rate": 1.7783536758321105e-07, "loss": 0.7922, "step": 10475 }, { "epoch": 0.81, "learning_rate": 1.7769229114351824e-07, "loss": 1.1909, "step": 10476 }, { "epoch": 0.81, "learning_rate": 1.7754926667021742e-07, "loss": 1.236, "step": 10477 }, { "epoch": 0.81, "learning_rate": 1.774062941723472e-07, "loss": 1.2137, "step": 10478 }, { "epoch": 0.81, "learning_rate": 1.7726337365894295e-07, "loss": 1.1703, "step": 10479 }, { "epoch": 0.81, "learning_rate": 1.7712050513903608e-07, "loss": 1.2127, "step": 10480 }, { "epoch": 0.81, "learning_rate": 1.7697768862165597e-07, "loss": 0.8525, "step": 10481 }, { "epoch": 0.81, "learning_rate": 1.7683492411582768e-07, "loss": 1.2086, "step": 10482 }, { "epoch": 0.81, "learning_rate": 1.7669221163057336e-07, "loss": 1.2139, "step": 10483 }, { "epoch": 0.81, "learning_rate": 1.7654955117491187e-07, "loss": 1.1974, "step": 10484 }, { "epoch": 0.81, "learning_rate": 1.7640694275785883e-07, "loss": 1.1479, "step": 10485 }, { "epoch": 0.81, "learning_rate": 1.7626438638842643e-07, "loss": 1.112, "step": 10486 }, { "epoch": 0.81, "learning_rate": 1.7612188207562351e-07, "loss": 1.1415, "step": 10487 }, { "epoch": 0.81, "learning_rate": 1.7597942982845594e-07, "loss": 1.1914, "step": 10488 }, { "epoch": 0.81, "learning_rate": 1.7583702965592594e-07, "loss": 0.8395, "step": 10489 }, { "epoch": 0.81, "learning_rate": 1.7569468156703283e-07, "loss": 1.2479, "step": 10490 }, { "epoch": 0.81, "learning_rate": 1.7555238557077212e-07, "loss": 1.1733, "step": 10491 }, { "epoch": 0.81, "learning_rate": 1.7541014167613655e-07, "loss": 1.2436, "step": 10492 }, { "epoch": 0.81, "learning_rate": 1.7526794989211514e-07, "loss": 1.1687, "step": 10493 }, { "epoch": 0.81, "learning_rate": 1.7512581022769403e-07, "loss": 1.2026, "step": 10494 }, { "epoch": 0.81, "learning_rate": 1.749837226918557e-07, "loss": 1.2222, "step": 10495 }, { "epoch": 0.81, "learning_rate": 1.748416872935795e-07, "loss": 1.1603, "step": 10496 }, { "epoch": 0.81, "learning_rate": 1.7469970404184142e-07, "loss": 0.8481, "step": 10497 }, { "epoch": 0.81, "learning_rate": 1.7455777294561436e-07, "loss": 1.1929, "step": 10498 }, { "epoch": 0.81, "learning_rate": 1.7441589401386747e-07, "loss": 1.1797, "step": 10499 }, { "epoch": 0.81, "learning_rate": 1.742740672555676e-07, "loss": 1.154, "step": 10500 }, { "epoch": 0.82, "learning_rate": 1.7413229267967688e-07, "loss": 1.214, "step": 10501 }, { "epoch": 0.82, "learning_rate": 1.7399057029515508e-07, "loss": 1.2228, "step": 10502 }, { "epoch": 0.82, "learning_rate": 1.7384890011095853e-07, "loss": 1.2259, "step": 10503 }, { "epoch": 0.82, "learning_rate": 1.7370728213603992e-07, "loss": 0.8413, "step": 10504 }, { "epoch": 0.82, "learning_rate": 1.735657163793496e-07, "loss": 0.8615, "step": 10505 }, { "epoch": 0.82, "learning_rate": 1.7342420284983316e-07, "loss": 1.1434, "step": 10506 }, { "epoch": 0.82, "learning_rate": 1.7328274155643385e-07, "loss": 1.211, "step": 10507 }, { "epoch": 0.82, "learning_rate": 1.7314133250809193e-07, "loss": 0.843, "step": 10508 }, { "epoch": 0.82, "learning_rate": 1.7299997571374314e-07, "loss": 1.2453, "step": 10509 }, { "epoch": 0.82, "learning_rate": 1.728586711823211e-07, "loss": 0.8574, "step": 10510 }, { "epoch": 0.82, "learning_rate": 1.727174189227556e-07, "loss": 1.203, "step": 10511 }, { "epoch": 0.82, "learning_rate": 1.725762189439731e-07, "loss": 1.1842, "step": 10512 }, { "epoch": 0.82, "learning_rate": 1.7243507125489708e-07, "loss": 1.173, "step": 10513 }, { "epoch": 0.82, "learning_rate": 1.7229397586444695e-07, "loss": 1.2351, "step": 10514 }, { "epoch": 0.82, "learning_rate": 1.7215293278153976e-07, "loss": 1.212, "step": 10515 }, { "epoch": 0.82, "learning_rate": 1.7201194201508872e-07, "loss": 1.2028, "step": 10516 }, { "epoch": 0.82, "learning_rate": 1.7187100357400398e-07, "loss": 1.1928, "step": 10517 }, { "epoch": 0.82, "learning_rate": 1.7173011746719213e-07, "loss": 1.206, "step": 10518 }, { "epoch": 0.82, "learning_rate": 1.7158928370355664e-07, "loss": 1.2002, "step": 10519 }, { "epoch": 0.82, "learning_rate": 1.714485022919976e-07, "loss": 1.2459, "step": 10520 }, { "epoch": 0.82, "learning_rate": 1.7130777324141176e-07, "loss": 1.2304, "step": 10521 }, { "epoch": 0.82, "learning_rate": 1.711670965606926e-07, "loss": 1.1614, "step": 10522 }, { "epoch": 0.82, "learning_rate": 1.7102647225873035e-07, "loss": 1.1626, "step": 10523 }, { "epoch": 0.82, "learning_rate": 1.7088590034441185e-07, "loss": 1.2125, "step": 10524 }, { "epoch": 0.82, "learning_rate": 1.707453808266207e-07, "loss": 1.2214, "step": 10525 }, { "epoch": 0.82, "learning_rate": 1.7060491371423712e-07, "loss": 1.2118, "step": 10526 }, { "epoch": 0.82, "learning_rate": 1.704644990161379e-07, "loss": 1.1887, "step": 10527 }, { "epoch": 0.82, "learning_rate": 1.7032413674119695e-07, "loss": 1.2251, "step": 10528 }, { "epoch": 0.82, "learning_rate": 1.7018382689828425e-07, "loss": 1.1979, "step": 10529 }, { "epoch": 0.82, "learning_rate": 1.7004356949626708e-07, "loss": 1.2147, "step": 10530 }, { "epoch": 0.82, "learning_rate": 1.6990336454400888e-07, "loss": 1.154, "step": 10531 }, { "epoch": 0.82, "learning_rate": 1.6976321205037015e-07, "loss": 1.1328, "step": 10532 }, { "epoch": 0.82, "learning_rate": 1.6962311202420764e-07, "loss": 1.2012, "step": 10533 }, { "epoch": 0.82, "learning_rate": 1.69483064474376e-07, "loss": 0.8411, "step": 10534 }, { "epoch": 0.82, "learning_rate": 1.6934306940972455e-07, "loss": 1.1697, "step": 10535 }, { "epoch": 0.82, "learning_rate": 1.6920312683910098e-07, "loss": 1.2591, "step": 10536 }, { "epoch": 0.82, "learning_rate": 1.6906323677134883e-07, "loss": 1.2018, "step": 10537 }, { "epoch": 0.82, "learning_rate": 1.6892339921530852e-07, "loss": 1.1916, "step": 10538 }, { "epoch": 0.82, "learning_rate": 1.687836141798178e-07, "loss": 1.1321, "step": 10539 }, { "epoch": 0.82, "learning_rate": 1.6864388167370979e-07, "loss": 1.1711, "step": 10540 }, { "epoch": 0.82, "learning_rate": 1.6850420170581493e-07, "loss": 1.1995, "step": 10541 }, { "epoch": 0.82, "learning_rate": 1.6836457428496131e-07, "loss": 1.2198, "step": 10542 }, { "epoch": 0.82, "learning_rate": 1.682249994199717e-07, "loss": 1.2304, "step": 10543 }, { "epoch": 0.82, "learning_rate": 1.680854771196676e-07, "loss": 1.1571, "step": 10544 }, { "epoch": 0.82, "learning_rate": 1.6794600739286557e-07, "loss": 1.2465, "step": 10545 }, { "epoch": 0.82, "learning_rate": 1.678065902483795e-07, "loss": 1.2047, "step": 10546 }, { "epoch": 0.82, "learning_rate": 1.6766722569502067e-07, "loss": 1.2236, "step": 10547 }, { "epoch": 0.82, "learning_rate": 1.6752791374159536e-07, "loss": 1.2158, "step": 10548 }, { "epoch": 0.82, "learning_rate": 1.673886543969083e-07, "loss": 1.123, "step": 10549 }, { "epoch": 0.82, "learning_rate": 1.6724944766976001e-07, "loss": 1.1653, "step": 10550 }, { "epoch": 0.82, "learning_rate": 1.6711029356894713e-07, "loss": 1.1884, "step": 10551 }, { "epoch": 0.82, "learning_rate": 1.6697119210326417e-07, "loss": 1.2398, "step": 10552 }, { "epoch": 0.82, "learning_rate": 1.6683214328150164e-07, "loss": 1.1668, "step": 10553 }, { "epoch": 0.82, "learning_rate": 1.6669314711244685e-07, "loss": 1.1734, "step": 10554 }, { "epoch": 0.82, "learning_rate": 1.6655420360488382e-07, "loss": 1.161, "step": 10555 }, { "epoch": 0.82, "learning_rate": 1.6641531276759313e-07, "loss": 1.2162, "step": 10556 }, { "epoch": 0.82, "learning_rate": 1.66276474609352e-07, "loss": 1.1981, "step": 10557 }, { "epoch": 0.82, "learning_rate": 1.6613768913893467e-07, "loss": 1.1899, "step": 10558 }, { "epoch": 0.82, "learning_rate": 1.6599895636511152e-07, "loss": 1.1845, "step": 10559 }, { "epoch": 0.82, "learning_rate": 1.6586027629665e-07, "loss": 1.1638, "step": 10560 }, { "epoch": 0.82, "learning_rate": 1.657216489423141e-07, "loss": 1.2284, "step": 10561 }, { "epoch": 0.82, "learning_rate": 1.655830743108646e-07, "loss": 1.1951, "step": 10562 }, { "epoch": 0.82, "learning_rate": 1.654445524110587e-07, "loss": 1.2381, "step": 10563 }, { "epoch": 0.82, "learning_rate": 1.653060832516504e-07, "loss": 1.1836, "step": 10564 }, { "epoch": 0.82, "learning_rate": 1.6516766684139038e-07, "loss": 0.822, "step": 10565 }, { "epoch": 0.82, "learning_rate": 1.6502930318902607e-07, "loss": 1.2027, "step": 10566 }, { "epoch": 0.82, "learning_rate": 1.6489099230330128e-07, "loss": 1.2235, "step": 10567 }, { "epoch": 0.82, "learning_rate": 1.6475273419295687e-07, "loss": 1.2013, "step": 10568 }, { "epoch": 0.82, "learning_rate": 1.6461452886673012e-07, "loss": 1.2201, "step": 10569 }, { "epoch": 0.82, "learning_rate": 1.64476376333355e-07, "loss": 1.1886, "step": 10570 }, { "epoch": 0.82, "learning_rate": 1.6433827660156218e-07, "loss": 1.2345, "step": 10571 }, { "epoch": 0.82, "learning_rate": 1.6420022968007895e-07, "loss": 1.2127, "step": 10572 }, { "epoch": 0.82, "learning_rate": 1.6406223557762934e-07, "loss": 1.1936, "step": 10573 }, { "epoch": 0.82, "learning_rate": 1.6392429430293396e-07, "loss": 1.1769, "step": 10574 }, { "epoch": 0.82, "learning_rate": 1.6378640586470994e-07, "loss": 1.2506, "step": 10575 }, { "epoch": 0.82, "learning_rate": 1.6364857027167178e-07, "loss": 1.1561, "step": 10576 }, { "epoch": 0.82, "learning_rate": 1.6351078753252944e-07, "loss": 1.1849, "step": 10577 }, { "epoch": 0.82, "learning_rate": 1.6337305765599108e-07, "loss": 1.1983, "step": 10578 }, { "epoch": 0.82, "learning_rate": 1.632353806507597e-07, "loss": 1.1522, "step": 10579 }, { "epoch": 0.82, "learning_rate": 1.6309775652553626e-07, "loss": 1.1859, "step": 10580 }, { "epoch": 0.82, "learning_rate": 1.6296018528901845e-07, "loss": 1.1995, "step": 10581 }, { "epoch": 0.82, "learning_rate": 1.6282266694989944e-07, "loss": 1.1911, "step": 10582 }, { "epoch": 0.82, "learning_rate": 1.6268520151687038e-07, "loss": 1.1715, "step": 10583 }, { "epoch": 0.82, "learning_rate": 1.6254778899861866e-07, "loss": 1.1691, "step": 10584 }, { "epoch": 0.82, "learning_rate": 1.6241042940382733e-07, "loss": 1.2192, "step": 10585 }, { "epoch": 0.82, "learning_rate": 1.6227312274117766e-07, "loss": 1.1841, "step": 10586 }, { "epoch": 0.82, "learning_rate": 1.6213586901934662e-07, "loss": 1.2226, "step": 10587 }, { "epoch": 0.82, "learning_rate": 1.6199866824700813e-07, "loss": 1.1634, "step": 10588 }, { "epoch": 0.82, "learning_rate": 1.6186152043283297e-07, "loss": 1.2177, "step": 10589 }, { "epoch": 0.82, "learning_rate": 1.617244255854875e-07, "loss": 1.2461, "step": 10590 }, { "epoch": 0.82, "learning_rate": 1.6158738371363634e-07, "loss": 1.1634, "step": 10591 }, { "epoch": 0.82, "learning_rate": 1.6145039482593958e-07, "loss": 1.1077, "step": 10592 }, { "epoch": 0.82, "learning_rate": 1.6131345893105452e-07, "loss": 1.1756, "step": 10593 }, { "epoch": 0.82, "learning_rate": 1.611765760376348e-07, "loss": 1.1845, "step": 10594 }, { "epoch": 0.82, "learning_rate": 1.610397461543308e-07, "loss": 1.135, "step": 10595 }, { "epoch": 0.82, "learning_rate": 1.6090296928978986e-07, "loss": 1.1499, "step": 10596 }, { "epoch": 0.82, "learning_rate": 1.6076624545265537e-07, "loss": 0.8733, "step": 10597 }, { "epoch": 0.82, "learning_rate": 1.606295746515679e-07, "loss": 1.1999, "step": 10598 }, { "epoch": 0.82, "learning_rate": 1.6049295689516451e-07, "loss": 1.1748, "step": 10599 }, { "epoch": 0.82, "learning_rate": 1.6035639219207865e-07, "loss": 0.8033, "step": 10600 }, { "epoch": 0.82, "learning_rate": 1.6021988055094083e-07, "loss": 1.2096, "step": 10601 }, { "epoch": 0.82, "learning_rate": 1.6008342198037794e-07, "loss": 1.1792, "step": 10602 }, { "epoch": 0.82, "learning_rate": 1.5994701648901364e-07, "loss": 1.189, "step": 10603 }, { "epoch": 0.82, "learning_rate": 1.5981066408546817e-07, "loss": 1.148, "step": 10604 }, { "epoch": 0.82, "learning_rate": 1.5967436477835827e-07, "loss": 1.1911, "step": 10605 }, { "epoch": 0.82, "learning_rate": 1.5953811857629773e-07, "loss": 1.1905, "step": 10606 }, { "epoch": 0.82, "learning_rate": 1.5940192548789654e-07, "loss": 1.1015, "step": 10607 }, { "epoch": 0.82, "learning_rate": 1.5926578552176174e-07, "loss": 1.1986, "step": 10608 }, { "epoch": 0.82, "learning_rate": 1.591296986864964e-07, "loss": 1.1689, "step": 10609 }, { "epoch": 0.82, "learning_rate": 1.5899366499070133e-07, "loss": 1.2164, "step": 10610 }, { "epoch": 0.82, "learning_rate": 1.5885768444297266e-07, "loss": 1.19, "step": 10611 }, { "epoch": 0.82, "learning_rate": 1.5872175705190404e-07, "loss": 1.2039, "step": 10612 }, { "epoch": 0.82, "learning_rate": 1.5858588282608544e-07, "loss": 0.8798, "step": 10613 }, { "epoch": 0.82, "learning_rate": 1.5845006177410325e-07, "loss": 1.2346, "step": 10614 }, { "epoch": 0.82, "learning_rate": 1.5831429390454177e-07, "loss": 1.2161, "step": 10615 }, { "epoch": 0.82, "learning_rate": 1.5817857922597988e-07, "loss": 0.8564, "step": 10616 }, { "epoch": 0.82, "learning_rate": 1.5804291774699452e-07, "loss": 1.2157, "step": 10617 }, { "epoch": 0.82, "learning_rate": 1.5790730947615938e-07, "loss": 1.2194, "step": 10618 }, { "epoch": 0.82, "learning_rate": 1.5777175442204348e-07, "loss": 1.1748, "step": 10619 }, { "epoch": 0.82, "learning_rate": 1.5763625259321435e-07, "loss": 1.2154, "step": 10620 }, { "epoch": 0.82, "learning_rate": 1.5750080399823407e-07, "loss": 0.8001, "step": 10621 }, { "epoch": 0.82, "learning_rate": 1.5736540864566318e-07, "loss": 0.8441, "step": 10622 }, { "epoch": 0.82, "learning_rate": 1.5723006654405812e-07, "loss": 1.1772, "step": 10623 }, { "epoch": 0.82, "learning_rate": 1.5709477770197123e-07, "loss": 1.2381, "step": 10624 }, { "epoch": 0.82, "learning_rate": 1.5695954212795283e-07, "loss": 1.1842, "step": 10625 }, { "epoch": 0.82, "learning_rate": 1.5682435983054908e-07, "loss": 1.1907, "step": 10626 }, { "epoch": 0.82, "learning_rate": 1.56689230818303e-07, "loss": 1.2098, "step": 10627 }, { "epoch": 0.82, "learning_rate": 1.565541550997539e-07, "loss": 1.2028, "step": 10628 }, { "epoch": 0.82, "learning_rate": 1.5641913268343832e-07, "loss": 0.8209, "step": 10629 }, { "epoch": 0.83, "learning_rate": 1.5628416357788898e-07, "loss": 0.8466, "step": 10630 }, { "epoch": 0.83, "learning_rate": 1.561492477916353e-07, "loss": 1.1869, "step": 10631 }, { "epoch": 0.83, "learning_rate": 1.5601438533320355e-07, "loss": 1.209, "step": 10632 }, { "epoch": 0.83, "learning_rate": 1.5587957621111625e-07, "loss": 1.1541, "step": 10633 }, { "epoch": 0.83, "learning_rate": 1.5574482043389303e-07, "loss": 0.8802, "step": 10634 }, { "epoch": 0.83, "learning_rate": 1.5561011801004963e-07, "loss": 1.1702, "step": 10635 }, { "epoch": 0.83, "learning_rate": 1.554754689480988e-07, "loss": 1.2638, "step": 10636 }, { "epoch": 0.83, "learning_rate": 1.5534087325654987e-07, "loss": 1.2394, "step": 10637 }, { "epoch": 0.83, "learning_rate": 1.5520633094390855e-07, "loss": 1.1866, "step": 10638 }, { "epoch": 0.83, "learning_rate": 1.550718420186775e-07, "loss": 1.1634, "step": 10639 }, { "epoch": 0.83, "learning_rate": 1.549374064893557e-07, "loss": 1.2102, "step": 10640 }, { "epoch": 0.83, "learning_rate": 1.5480302436443915e-07, "loss": 1.2167, "step": 10641 }, { "epoch": 0.83, "learning_rate": 1.5466869565242003e-07, "loss": 1.2319, "step": 10642 }, { "epoch": 0.83, "learning_rate": 1.5453442036178721e-07, "loss": 1.1418, "step": 10643 }, { "epoch": 0.83, "learning_rate": 1.5440019850102693e-07, "loss": 1.2091, "step": 10644 }, { "epoch": 0.83, "learning_rate": 1.5426603007862082e-07, "loss": 1.1468, "step": 10645 }, { "epoch": 0.83, "learning_rate": 1.5413191510304802e-07, "loss": 1.1324, "step": 10646 }, { "epoch": 0.83, "learning_rate": 1.5399785358278394e-07, "loss": 1.2078, "step": 10647 }, { "epoch": 0.83, "learning_rate": 1.5386384552630061e-07, "loss": 1.2192, "step": 10648 }, { "epoch": 0.83, "learning_rate": 1.5372989094206733e-07, "loss": 1.1849, "step": 10649 }, { "epoch": 0.83, "learning_rate": 1.5359598983854883e-07, "loss": 1.1789, "step": 10650 }, { "epoch": 0.83, "learning_rate": 1.5346214222420717e-07, "loss": 1.1855, "step": 10651 }, { "epoch": 0.83, "learning_rate": 1.533283481075015e-07, "loss": 1.1906, "step": 10652 }, { "epoch": 0.83, "learning_rate": 1.5319460749688628e-07, "loss": 1.216, "step": 10653 }, { "epoch": 0.83, "learning_rate": 1.530609204008142e-07, "loss": 1.2126, "step": 10654 }, { "epoch": 0.83, "learning_rate": 1.5292728682773305e-07, "loss": 1.1741, "step": 10655 }, { "epoch": 0.83, "learning_rate": 1.527937067860877e-07, "loss": 0.8369, "step": 10656 }, { "epoch": 0.83, "learning_rate": 1.5266018028432094e-07, "loss": 1.176, "step": 10657 }, { "epoch": 0.83, "learning_rate": 1.525267073308699e-07, "loss": 1.222, "step": 10658 }, { "epoch": 0.83, "learning_rate": 1.5239328793417029e-07, "loss": 0.8122, "step": 10659 }, { "epoch": 0.83, "learning_rate": 1.5225992210265327e-07, "loss": 1.2126, "step": 10660 }, { "epoch": 0.83, "learning_rate": 1.5212660984474724e-07, "loss": 1.2102, "step": 10661 }, { "epoch": 0.83, "learning_rate": 1.519933511688768e-07, "loss": 0.8161, "step": 10662 }, { "epoch": 0.83, "learning_rate": 1.5186014608346344e-07, "loss": 1.2117, "step": 10663 }, { "epoch": 0.83, "learning_rate": 1.5172699459692505e-07, "loss": 1.1873, "step": 10664 }, { "epoch": 0.83, "learning_rate": 1.515938967176763e-07, "loss": 1.1968, "step": 10665 }, { "epoch": 0.83, "learning_rate": 1.5146085245412843e-07, "loss": 1.1627, "step": 10666 }, { "epoch": 0.83, "learning_rate": 1.5132786181468927e-07, "loss": 1.1908, "step": 10667 }, { "epoch": 0.83, "learning_rate": 1.511949248077633e-07, "loss": 1.1739, "step": 10668 }, { "epoch": 0.83, "learning_rate": 1.510620414417516e-07, "loss": 1.2204, "step": 10669 }, { "epoch": 0.83, "learning_rate": 1.509292117250518e-07, "loss": 0.8485, "step": 10670 }, { "epoch": 0.83, "learning_rate": 1.5079643566605804e-07, "loss": 1.1733, "step": 10671 }, { "epoch": 0.83, "learning_rate": 1.5066371327316153e-07, "loss": 1.2043, "step": 10672 }, { "epoch": 0.83, "learning_rate": 1.5053104455474942e-07, "loss": 1.2095, "step": 10673 }, { "epoch": 0.83, "learning_rate": 1.50398429519206e-07, "loss": 1.1648, "step": 10674 }, { "epoch": 0.83, "learning_rate": 1.5026586817491204e-07, "loss": 1.1523, "step": 10675 }, { "epoch": 0.83, "learning_rate": 1.5013336053024474e-07, "loss": 1.1801, "step": 10676 }, { "epoch": 0.83, "learning_rate": 1.50000906593578e-07, "loss": 1.1736, "step": 10677 }, { "epoch": 0.83, "learning_rate": 1.4986850637328253e-07, "loss": 1.1418, "step": 10678 }, { "epoch": 0.83, "learning_rate": 1.4973615987772524e-07, "loss": 1.1671, "step": 10679 }, { "epoch": 0.83, "learning_rate": 1.4960386711527005e-07, "loss": 1.1682, "step": 10680 }, { "epoch": 0.83, "learning_rate": 1.4947162809427726e-07, "loss": 1.219, "step": 10681 }, { "epoch": 0.83, "learning_rate": 1.4933944282310363e-07, "loss": 1.1692, "step": 10682 }, { "epoch": 0.83, "learning_rate": 1.4920731131010322e-07, "loss": 0.8223, "step": 10683 }, { "epoch": 0.83, "learning_rate": 1.4907523356362573e-07, "loss": 1.2341, "step": 10684 }, { "epoch": 0.83, "learning_rate": 1.4894320959201777e-07, "loss": 1.1827, "step": 10685 }, { "epoch": 0.83, "learning_rate": 1.488112394036234e-07, "loss": 1.1779, "step": 10686 }, { "epoch": 0.83, "learning_rate": 1.4867932300678175e-07, "loss": 1.2113, "step": 10687 }, { "epoch": 0.83, "learning_rate": 1.4854746040983024e-07, "loss": 1.2431, "step": 10688 }, { "epoch": 0.83, "learning_rate": 1.4841565162110136e-07, "loss": 1.2018, "step": 10689 }, { "epoch": 0.83, "learning_rate": 1.4828389664892483e-07, "loss": 1.2476, "step": 10690 }, { "epoch": 0.83, "learning_rate": 1.4815219550162772e-07, "loss": 1.2525, "step": 10691 }, { "epoch": 0.83, "learning_rate": 1.4802054818753206e-07, "loss": 1.2636, "step": 10692 }, { "epoch": 0.83, "learning_rate": 1.4788895471495822e-07, "loss": 1.2045, "step": 10693 }, { "epoch": 0.83, "learning_rate": 1.4775741509222218e-07, "loss": 1.17, "step": 10694 }, { "epoch": 0.83, "learning_rate": 1.4762592932763606e-07, "loss": 1.2351, "step": 10695 }, { "epoch": 0.83, "learning_rate": 1.4749449742951014e-07, "loss": 0.832, "step": 10696 }, { "epoch": 0.83, "learning_rate": 1.4736311940614955e-07, "loss": 1.199, "step": 10697 }, { "epoch": 0.83, "learning_rate": 1.4723179526585728e-07, "loss": 1.242, "step": 10698 }, { "epoch": 0.83, "learning_rate": 1.471005250169327e-07, "loss": 1.2238, "step": 10699 }, { "epoch": 0.83, "learning_rate": 1.4696930866767065e-07, "loss": 1.1904, "step": 10700 }, { "epoch": 0.83, "learning_rate": 1.4683814622636437e-07, "loss": 1.1784, "step": 10701 }, { "epoch": 0.83, "learning_rate": 1.4670703770130232e-07, "loss": 1.2231, "step": 10702 }, { "epoch": 0.83, "learning_rate": 1.4657598310077013e-07, "loss": 1.2549, "step": 10703 }, { "epoch": 0.83, "learning_rate": 1.464449824330498e-07, "loss": 1.1635, "step": 10704 }, { "epoch": 0.83, "learning_rate": 1.463140357064202e-07, "loss": 1.1782, "step": 10705 }, { "epoch": 0.83, "learning_rate": 1.4618314292915646e-07, "loss": 1.1628, "step": 10706 }, { "epoch": 0.83, "learning_rate": 1.460523041095304e-07, "loss": 1.1897, "step": 10707 }, { "epoch": 0.83, "learning_rate": 1.4592151925581076e-07, "loss": 1.2625, "step": 10708 }, { "epoch": 0.83, "learning_rate": 1.4579078837626234e-07, "loss": 1.1714, "step": 10709 }, { "epoch": 0.83, "learning_rate": 1.4566011147914682e-07, "loss": 1.1896, "step": 10710 }, { "epoch": 0.83, "learning_rate": 1.4552948857272252e-07, "loss": 0.8219, "step": 10711 }, { "epoch": 0.83, "learning_rate": 1.453989196652442e-07, "loss": 1.1275, "step": 10712 }, { "epoch": 0.83, "learning_rate": 1.4526840476496327e-07, "loss": 1.2487, "step": 10713 }, { "epoch": 0.83, "learning_rate": 1.4513794388012768e-07, "loss": 1.1854, "step": 10714 }, { "epoch": 0.83, "learning_rate": 1.4500753701898215e-07, "loss": 1.1821, "step": 10715 }, { "epoch": 0.83, "learning_rate": 1.4487718418976769e-07, "loss": 1.1705, "step": 10716 }, { "epoch": 0.83, "learning_rate": 1.447468854007221e-07, "loss": 0.8393, "step": 10717 }, { "epoch": 0.83, "learning_rate": 1.4461664066007983e-07, "loss": 1.1799, "step": 10718 }, { "epoch": 0.83, "learning_rate": 1.444864499760714e-07, "loss": 1.173, "step": 10719 }, { "epoch": 0.83, "learning_rate": 1.443563133569251e-07, "loss": 1.2121, "step": 10720 }, { "epoch": 0.83, "learning_rate": 1.4422623081086427e-07, "loss": 0.818, "step": 10721 }, { "epoch": 0.83, "learning_rate": 1.4409620234610985e-07, "loss": 1.2121, "step": 10722 }, { "epoch": 0.83, "learning_rate": 1.439662279708791e-07, "loss": 1.1622, "step": 10723 }, { "epoch": 0.83, "learning_rate": 1.438363076933856e-07, "loss": 1.2296, "step": 10724 }, { "epoch": 0.83, "learning_rate": 1.4370644152184042e-07, "loss": 0.8655, "step": 10725 }, { "epoch": 0.83, "learning_rate": 1.435766294644497e-07, "loss": 1.2365, "step": 10726 }, { "epoch": 0.83, "learning_rate": 1.434468715294178e-07, "loss": 1.1579, "step": 10727 }, { "epoch": 0.83, "learning_rate": 1.4331716772494462e-07, "loss": 1.2173, "step": 10728 }, { "epoch": 0.83, "learning_rate": 1.4318751805922647e-07, "loss": 1.1935, "step": 10729 }, { "epoch": 0.83, "learning_rate": 1.4305792254045746e-07, "loss": 1.1815, "step": 10730 }, { "epoch": 0.83, "learning_rate": 1.429283811768265e-07, "loss": 0.8286, "step": 10731 }, { "epoch": 0.83, "learning_rate": 1.427988939765208e-07, "loss": 0.8359, "step": 10732 }, { "epoch": 0.83, "learning_rate": 1.4266946094772348e-07, "loss": 1.1416, "step": 10733 }, { "epoch": 0.83, "learning_rate": 1.4254008209861333e-07, "loss": 1.2009, "step": 10734 }, { "epoch": 0.83, "learning_rate": 1.424107574373674e-07, "loss": 1.1986, "step": 10735 }, { "epoch": 0.83, "learning_rate": 1.4228148697215813e-07, "loss": 1.1854, "step": 10736 }, { "epoch": 0.83, "learning_rate": 1.4215227071115498e-07, "loss": 1.179, "step": 10737 }, { "epoch": 0.83, "learning_rate": 1.4202310866252376e-07, "loss": 1.2157, "step": 10738 }, { "epoch": 0.83, "learning_rate": 1.4189400083442694e-07, "loss": 1.1785, "step": 10739 }, { "epoch": 0.83, "learning_rate": 1.4176494723502374e-07, "loss": 1.2052, "step": 10740 }, { "epoch": 0.83, "learning_rate": 1.416359478724697e-07, "loss": 0.8312, "step": 10741 }, { "epoch": 0.83, "learning_rate": 1.41507002754917e-07, "loss": 1.2316, "step": 10742 }, { "epoch": 0.83, "learning_rate": 1.4137811189051463e-07, "loss": 1.1815, "step": 10743 }, { "epoch": 0.83, "learning_rate": 1.4124927528740772e-07, "loss": 1.2155, "step": 10744 }, { "epoch": 0.83, "learning_rate": 1.411204929537383e-07, "loss": 1.1314, "step": 10745 }, { "epoch": 0.83, "learning_rate": 1.4099176489764487e-07, "loss": 1.1924, "step": 10746 }, { "epoch": 0.83, "learning_rate": 1.4086309112726246e-07, "loss": 1.1693, "step": 10747 }, { "epoch": 0.83, "learning_rate": 1.4073447165072272e-07, "loss": 1.1835, "step": 10748 }, { "epoch": 0.83, "learning_rate": 1.40605906476154e-07, "loss": 1.2108, "step": 10749 }, { "epoch": 0.83, "learning_rate": 1.4047739561168092e-07, "loss": 1.2195, "step": 10750 }, { "epoch": 0.83, "learning_rate": 1.4034893906542488e-07, "loss": 1.1867, "step": 10751 }, { "epoch": 0.83, "learning_rate": 1.4022053684550372e-07, "loss": 1.2527, "step": 10752 }, { "epoch": 0.83, "learning_rate": 1.4009218896003204e-07, "loss": 1.1993, "step": 10753 }, { "epoch": 0.83, "learning_rate": 1.3996389541712094e-07, "loss": 1.2026, "step": 10754 }, { "epoch": 0.83, "learning_rate": 1.398356562248778e-07, "loss": 0.8472, "step": 10755 }, { "epoch": 0.83, "learning_rate": 1.3970747139140705e-07, "loss": 1.1952, "step": 10756 }, { "epoch": 0.83, "learning_rate": 1.3957934092480915e-07, "loss": 1.2002, "step": 10757 }, { "epoch": 0.83, "learning_rate": 1.3945126483318147e-07, "loss": 1.1329, "step": 10758 }, { "epoch": 0.84, "learning_rate": 1.3932324312461842e-07, "loss": 1.2224, "step": 10759 }, { "epoch": 0.84, "learning_rate": 1.3919527580720968e-07, "loss": 1.1492, "step": 10760 }, { "epoch": 0.84, "learning_rate": 1.3906736288904242e-07, "loss": 1.1797, "step": 10761 }, { "epoch": 0.84, "learning_rate": 1.389395043782008e-07, "loss": 1.2038, "step": 10762 }, { "epoch": 0.84, "learning_rate": 1.388117002827639e-07, "loss": 1.2196, "step": 10763 }, { "epoch": 0.84, "learning_rate": 1.3868395061080952e-07, "loss": 1.2057, "step": 10764 }, { "epoch": 0.84, "learning_rate": 1.3855625537041017e-07, "loss": 0.8618, "step": 10765 }, { "epoch": 0.84, "learning_rate": 1.3842861456963562e-07, "loss": 1.1531, "step": 10766 }, { "epoch": 0.84, "learning_rate": 1.3830102821655288e-07, "loss": 1.2147, "step": 10767 }, { "epoch": 0.84, "learning_rate": 1.3817349631922404e-07, "loss": 1.2087, "step": 10768 }, { "epoch": 0.84, "learning_rate": 1.380460188857091e-07, "loss": 1.1679, "step": 10769 }, { "epoch": 0.84, "learning_rate": 1.3791859592406407e-07, "loss": 1.1561, "step": 10770 }, { "epoch": 0.84, "learning_rate": 1.377912274423415e-07, "loss": 1.1853, "step": 10771 }, { "epoch": 0.84, "learning_rate": 1.3766391344859062e-07, "loss": 1.1913, "step": 10772 }, { "epoch": 0.84, "learning_rate": 1.3753665395085667e-07, "loss": 1.2237, "step": 10773 }, { "epoch": 0.84, "learning_rate": 1.3740944895718253e-07, "loss": 0.8567, "step": 10774 }, { "epoch": 0.84, "learning_rate": 1.372822984756067e-07, "loss": 0.8439, "step": 10775 }, { "epoch": 0.84, "learning_rate": 1.3715520251416457e-07, "loss": 1.1845, "step": 10776 }, { "epoch": 0.84, "learning_rate": 1.370281610808882e-07, "loss": 1.2426, "step": 10777 }, { "epoch": 0.84, "learning_rate": 1.3690117418380598e-07, "loss": 1.219, "step": 10778 }, { "epoch": 0.84, "learning_rate": 1.36774241830943e-07, "loss": 1.1932, "step": 10779 }, { "epoch": 0.84, "learning_rate": 1.3664736403032073e-07, "loss": 1.1793, "step": 10780 }, { "epoch": 0.84, "learning_rate": 1.3652054078995744e-07, "loss": 1.1464, "step": 10781 }, { "epoch": 0.84, "learning_rate": 1.3639377211786785e-07, "loss": 0.8598, "step": 10782 }, { "epoch": 0.84, "learning_rate": 1.3626705802206307e-07, "loss": 1.1387, "step": 10783 }, { "epoch": 0.84, "learning_rate": 1.3614039851055092e-07, "loss": 1.2318, "step": 10784 }, { "epoch": 0.84, "learning_rate": 1.3601379359133591e-07, "loss": 1.1879, "step": 10785 }, { "epoch": 0.84, "learning_rate": 1.3588724327241875e-07, "loss": 1.2546, "step": 10786 }, { "epoch": 0.84, "learning_rate": 1.3576074756179701e-07, "loss": 1.1753, "step": 10787 }, { "epoch": 0.84, "learning_rate": 1.3563430646746465e-07, "loss": 1.1816, "step": 10788 }, { "epoch": 0.84, "learning_rate": 1.3550791999741218e-07, "loss": 1.123, "step": 10789 }, { "epoch": 0.84, "learning_rate": 1.3538158815962675e-07, "loss": 1.2585, "step": 10790 }, { "epoch": 0.84, "learning_rate": 1.3525531096209199e-07, "loss": 1.2008, "step": 10791 }, { "epoch": 0.84, "learning_rate": 1.3512908841278768e-07, "loss": 1.1699, "step": 10792 }, { "epoch": 0.84, "learning_rate": 1.3500292051969165e-07, "loss": 1.1556, "step": 10793 }, { "epoch": 0.84, "learning_rate": 1.348768072907761e-07, "loss": 1.225, "step": 10794 }, { "epoch": 0.84, "learning_rate": 1.3475074873401104e-07, "loss": 1.1256, "step": 10795 }, { "epoch": 0.84, "learning_rate": 1.3462474485736352e-07, "loss": 1.1998, "step": 10796 }, { "epoch": 0.84, "learning_rate": 1.344987956687954e-07, "loss": 1.1308, "step": 10797 }, { "epoch": 0.84, "learning_rate": 1.343729011762673e-07, "loss": 1.1492, "step": 10798 }, { "epoch": 0.84, "learning_rate": 1.3424706138773434e-07, "loss": 0.8163, "step": 10799 }, { "epoch": 0.84, "learning_rate": 1.3412127631114922e-07, "loss": 1.1877, "step": 10800 }, { "epoch": 0.84, "learning_rate": 1.339955459544615e-07, "loss": 1.2185, "step": 10801 }, { "epoch": 0.84, "learning_rate": 1.3386987032561614e-07, "loss": 1.214, "step": 10802 }, { "epoch": 0.84, "learning_rate": 1.337442494325558e-07, "loss": 1.2168, "step": 10803 }, { "epoch": 0.84, "learning_rate": 1.3361868328321935e-07, "loss": 1.1818, "step": 10804 }, { "epoch": 0.84, "learning_rate": 1.334931718855412e-07, "loss": 1.1732, "step": 10805 }, { "epoch": 0.84, "learning_rate": 1.3336771524745417e-07, "loss": 1.1749, "step": 10806 }, { "epoch": 0.84, "learning_rate": 1.3324231337688564e-07, "loss": 1.2039, "step": 10807 }, { "epoch": 0.84, "learning_rate": 1.3311696628176116e-07, "loss": 1.1512, "step": 10808 }, { "epoch": 0.84, "learning_rate": 1.329916739700021e-07, "loss": 1.1966, "step": 10809 }, { "epoch": 0.84, "learning_rate": 1.3286643644952588e-07, "loss": 1.1569, "step": 10810 }, { "epoch": 0.84, "learning_rate": 1.327412537282475e-07, "loss": 1.2295, "step": 10811 }, { "epoch": 0.84, "learning_rate": 1.3261612581407779e-07, "loss": 1.1698, "step": 10812 }, { "epoch": 0.84, "learning_rate": 1.3249105271492434e-07, "loss": 1.2021, "step": 10813 }, { "epoch": 0.84, "learning_rate": 1.3236603443869133e-07, "loss": 1.1804, "step": 10814 }, { "epoch": 0.84, "learning_rate": 1.3224107099327918e-07, "loss": 1.1655, "step": 10815 }, { "epoch": 0.84, "learning_rate": 1.321161623865853e-07, "loss": 1.1776, "step": 10816 }, { "epoch": 0.84, "learning_rate": 1.319913086265032e-07, "loss": 1.1689, "step": 10817 }, { "epoch": 0.84, "learning_rate": 1.3186650972092294e-07, "loss": 1.21, "step": 10818 }, { "epoch": 0.84, "learning_rate": 1.3174176567773198e-07, "loss": 1.1541, "step": 10819 }, { "epoch": 0.84, "learning_rate": 1.3161707650481291e-07, "loss": 1.2184, "step": 10820 }, { "epoch": 0.84, "learning_rate": 1.3149244221004584e-07, "loss": 1.2441, "step": 10821 }, { "epoch": 0.84, "learning_rate": 1.313678628013072e-07, "loss": 1.2159, "step": 10822 }, { "epoch": 0.84, "learning_rate": 1.312433382864696e-07, "loss": 1.1992, "step": 10823 }, { "epoch": 0.84, "learning_rate": 1.3111886867340284e-07, "loss": 1.1383, "step": 10824 }, { "epoch": 0.84, "learning_rate": 1.3099445396997233e-07, "loss": 0.8695, "step": 10825 }, { "epoch": 0.84, "learning_rate": 1.308700941840415e-07, "loss": 0.8508, "step": 10826 }, { "epoch": 0.84, "learning_rate": 1.3074578932346846e-07, "loss": 1.2546, "step": 10827 }, { "epoch": 0.84, "learning_rate": 1.3062153939610888e-07, "loss": 1.2183, "step": 10828 }, { "epoch": 0.84, "learning_rate": 1.3049734440981551e-07, "loss": 1.2122, "step": 10829 }, { "epoch": 0.84, "learning_rate": 1.3037320437243626e-07, "loss": 1.1972, "step": 10830 }, { "epoch": 0.84, "learning_rate": 1.3024911929181627e-07, "loss": 1.2326, "step": 10831 }, { "epoch": 0.84, "learning_rate": 1.3012508917579778e-07, "loss": 1.2281, "step": 10832 }, { "epoch": 0.84, "learning_rate": 1.3000111403221837e-07, "loss": 1.1844, "step": 10833 }, { "epoch": 0.84, "learning_rate": 1.2987719386891282e-07, "loss": 1.2288, "step": 10834 }, { "epoch": 0.84, "learning_rate": 1.297533286937127e-07, "loss": 1.2143, "step": 10835 }, { "epoch": 0.84, "learning_rate": 1.296295185144458e-07, "loss": 1.1627, "step": 10836 }, { "epoch": 0.84, "learning_rate": 1.2950576333893582e-07, "loss": 1.1889, "step": 10837 }, { "epoch": 0.84, "learning_rate": 1.29382063175004e-07, "loss": 1.1846, "step": 10838 }, { "epoch": 0.84, "learning_rate": 1.29258418030468e-07, "loss": 1.1864, "step": 10839 }, { "epoch": 0.84, "learning_rate": 1.291348279131409e-07, "loss": 1.1607, "step": 10840 }, { "epoch": 0.84, "learning_rate": 1.2901129283083335e-07, "loss": 1.2505, "step": 10841 }, { "epoch": 0.84, "learning_rate": 1.2888781279135275e-07, "loss": 1.1888, "step": 10842 }, { "epoch": 0.84, "learning_rate": 1.2876438780250186e-07, "loss": 1.2232, "step": 10843 }, { "epoch": 0.84, "learning_rate": 1.286410178720807e-07, "loss": 1.2123, "step": 10844 }, { "epoch": 0.84, "learning_rate": 1.2851770300788622e-07, "loss": 1.2, "step": 10845 }, { "epoch": 0.84, "learning_rate": 1.283944432177112e-07, "loss": 1.1762, "step": 10846 }, { "epoch": 0.84, "learning_rate": 1.2827123850934462e-07, "loss": 1.2187, "step": 10847 }, { "epoch": 0.84, "learning_rate": 1.2814808889057316e-07, "loss": 1.244, "step": 10848 }, { "epoch": 0.84, "learning_rate": 1.2802499436917935e-07, "loss": 1.1669, "step": 10849 }, { "epoch": 0.84, "learning_rate": 1.2790195495294155e-07, "loss": 0.8136, "step": 10850 }, { "epoch": 0.84, "learning_rate": 1.2777897064963604e-07, "loss": 1.141, "step": 10851 }, { "epoch": 0.84, "learning_rate": 1.2765604146703457e-07, "loss": 1.2253, "step": 10852 }, { "epoch": 0.84, "learning_rate": 1.27533167412906e-07, "loss": 1.1624, "step": 10853 }, { "epoch": 0.84, "learning_rate": 1.274103484950152e-07, "loss": 0.8467, "step": 10854 }, { "epoch": 0.84, "learning_rate": 1.2728758472112388e-07, "loss": 1.2192, "step": 10855 }, { "epoch": 0.84, "learning_rate": 1.271648760989903e-07, "loss": 1.1774, "step": 10856 }, { "epoch": 0.84, "learning_rate": 1.27042222636369e-07, "loss": 1.1902, "step": 10857 }, { "epoch": 0.84, "learning_rate": 1.2691962434101122e-07, "loss": 1.1897, "step": 10858 }, { "epoch": 0.84, "learning_rate": 1.267970812206649e-07, "loss": 1.2249, "step": 10859 }, { "epoch": 0.84, "learning_rate": 1.2667459328307361e-07, "loss": 1.1479, "step": 10860 }, { "epoch": 0.84, "learning_rate": 1.265521605359786e-07, "loss": 1.1797, "step": 10861 }, { "epoch": 0.84, "learning_rate": 1.2642978298711672e-07, "loss": 1.195, "step": 10862 }, { "epoch": 0.84, "learning_rate": 1.263074606442225e-07, "loss": 1.1567, "step": 10863 }, { "epoch": 0.84, "learning_rate": 1.2618519351502544e-07, "loss": 0.8662, "step": 10864 }, { "epoch": 0.84, "learning_rate": 1.2606298160725226e-07, "loss": 1.1681, "step": 10865 }, { "epoch": 0.84, "learning_rate": 1.2594082492862702e-07, "loss": 0.8674, "step": 10866 }, { "epoch": 0.84, "learning_rate": 1.2581872348686874e-07, "loss": 1.1626, "step": 10867 }, { "epoch": 0.84, "learning_rate": 1.256966772896937e-07, "loss": 1.2231, "step": 10868 }, { "epoch": 0.84, "learning_rate": 1.2557468634481526e-07, "loss": 1.2077, "step": 10869 }, { "epoch": 0.84, "learning_rate": 1.2545275065994277e-07, "loss": 1.2237, "step": 10870 }, { "epoch": 0.84, "learning_rate": 1.2533087024278143e-07, "loss": 1.2037, "step": 10871 }, { "epoch": 0.84, "learning_rate": 1.252090451010337e-07, "loss": 1.1925, "step": 10872 }, { "epoch": 0.84, "learning_rate": 1.2508727524239905e-07, "loss": 1.2274, "step": 10873 }, { "epoch": 0.84, "learning_rate": 1.249655606745721e-07, "loss": 1.2013, "step": 10874 }, { "epoch": 0.84, "learning_rate": 1.2484390140524482e-07, "loss": 1.205, "step": 10875 }, { "epoch": 0.84, "learning_rate": 1.2472229744210606e-07, "loss": 1.1736, "step": 10876 }, { "epoch": 0.84, "learning_rate": 1.246007487928402e-07, "loss": 1.2017, "step": 10877 }, { "epoch": 0.84, "learning_rate": 1.244792554651285e-07, "loss": 1.1679, "step": 10878 }, { "epoch": 0.84, "learning_rate": 1.243578174666493e-07, "loss": 1.2549, "step": 10879 }, { "epoch": 0.84, "learning_rate": 1.2423643480507706e-07, "loss": 1.217, "step": 10880 }, { "epoch": 0.84, "learning_rate": 1.2411510748808173e-07, "loss": 1.1821, "step": 10881 }, { "epoch": 0.84, "learning_rate": 1.2399383552333166e-07, "loss": 1.1959, "step": 10882 }, { "epoch": 0.84, "learning_rate": 1.2387261891849056e-07, "loss": 1.2192, "step": 10883 }, { "epoch": 0.84, "learning_rate": 1.237514576812181e-07, "loss": 1.2222, "step": 10884 }, { "epoch": 0.84, "learning_rate": 1.2363035181917192e-07, "loss": 1.2379, "step": 10885 }, { "epoch": 0.84, "learning_rate": 1.2350930134000536e-07, "loss": 1.2427, "step": 10886 }, { "epoch": 0.84, "learning_rate": 1.2338830625136766e-07, "loss": 1.2228, "step": 10887 }, { "epoch": 0.85, "learning_rate": 1.2326736656090587e-07, "loss": 1.1533, "step": 10888 }, { "epoch": 0.85, "learning_rate": 1.2314648227626267e-07, "loss": 1.2484, "step": 10889 }, { "epoch": 0.85, "learning_rate": 1.2302565340507733e-07, "loss": 1.1927, "step": 10890 }, { "epoch": 0.85, "learning_rate": 1.2290487995498588e-07, "loss": 1.1696, "step": 10891 }, { "epoch": 0.85, "learning_rate": 1.2278416193362062e-07, "loss": 1.1518, "step": 10892 }, { "epoch": 0.85, "learning_rate": 1.2266349934861055e-07, "loss": 1.1535, "step": 10893 }, { "epoch": 0.85, "learning_rate": 1.225428922075805e-07, "loss": 1.1875, "step": 10894 }, { "epoch": 0.85, "learning_rate": 1.2242234051815303e-07, "loss": 1.1748, "step": 10895 }, { "epoch": 0.85, "learning_rate": 1.22301844287946e-07, "loss": 0.8507, "step": 10896 }, { "epoch": 0.85, "learning_rate": 1.2218140352457484e-07, "loss": 0.853, "step": 10897 }, { "epoch": 0.85, "learning_rate": 1.220610182356504e-07, "loss": 1.2011, "step": 10898 }, { "epoch": 0.85, "learning_rate": 1.2194068842878036e-07, "loss": 1.1659, "step": 10899 }, { "epoch": 0.85, "learning_rate": 1.2182041411156984e-07, "loss": 1.1685, "step": 10900 }, { "epoch": 0.85, "learning_rate": 1.2170019529161902e-07, "loss": 1.2254, "step": 10901 }, { "epoch": 0.85, "learning_rate": 1.215800319765251e-07, "loss": 1.1624, "step": 10902 }, { "epoch": 0.85, "learning_rate": 1.2145992417388262e-07, "loss": 1.2129, "step": 10903 }, { "epoch": 0.85, "learning_rate": 1.2133987189128124e-07, "loss": 0.8403, "step": 10904 }, { "epoch": 0.85, "learning_rate": 1.2121987513630806e-07, "loss": 1.1486, "step": 10905 }, { "epoch": 0.85, "learning_rate": 1.2109993391654604e-07, "loss": 1.2093, "step": 10906 }, { "epoch": 0.85, "learning_rate": 1.2098004823957565e-07, "loss": 0.8216, "step": 10907 }, { "epoch": 0.85, "learning_rate": 1.2086021811297252e-07, "loss": 1.2117, "step": 10908 }, { "epoch": 0.85, "learning_rate": 1.2074044354430946e-07, "loss": 1.1713, "step": 10909 }, { "epoch": 0.85, "learning_rate": 1.2062072454115623e-07, "loss": 1.1847, "step": 10910 }, { "epoch": 0.85, "learning_rate": 1.2050106111107793e-07, "loss": 1.1543, "step": 10911 }, { "epoch": 0.85, "learning_rate": 1.2038145326163696e-07, "loss": 1.1715, "step": 10912 }, { "epoch": 0.85, "learning_rate": 1.2026190100039234e-07, "loss": 1.1691, "step": 10913 }, { "epoch": 0.85, "learning_rate": 1.201424043348993e-07, "loss": 1.1999, "step": 10914 }, { "epoch": 0.85, "learning_rate": 1.200229632727089e-07, "loss": 1.1649, "step": 10915 }, { "epoch": 0.85, "learning_rate": 1.1990357782136985e-07, "loss": 1.1963, "step": 10916 }, { "epoch": 0.85, "learning_rate": 1.197842479884269e-07, "loss": 0.825, "step": 10917 }, { "epoch": 0.85, "learning_rate": 1.196649737814205e-07, "loss": 1.0983, "step": 10918 }, { "epoch": 0.85, "learning_rate": 1.19545755207889e-07, "loss": 1.276, "step": 10919 }, { "epoch": 0.85, "learning_rate": 1.1942659227536655e-07, "loss": 1.1824, "step": 10920 }, { "epoch": 0.85, "learning_rate": 1.1930748499138298e-07, "loss": 1.2489, "step": 10921 }, { "epoch": 0.85, "learning_rate": 1.1918843336346596e-07, "loss": 1.2137, "step": 10922 }, { "epoch": 0.85, "learning_rate": 1.190694373991391e-07, "loss": 1.156, "step": 10923 }, { "epoch": 0.85, "learning_rate": 1.1895049710592231e-07, "loss": 0.8432, "step": 10924 }, { "epoch": 0.85, "learning_rate": 1.1883161249133178e-07, "loss": 1.226, "step": 10925 }, { "epoch": 0.85, "learning_rate": 1.1871278356288106e-07, "loss": 1.1826, "step": 10926 }, { "epoch": 0.85, "learning_rate": 1.1859401032807958e-07, "loss": 1.1725, "step": 10927 }, { "epoch": 0.85, "learning_rate": 1.1847529279443269e-07, "loss": 1.2445, "step": 10928 }, { "epoch": 0.85, "learning_rate": 1.183566309694436e-07, "loss": 1.239, "step": 10929 }, { "epoch": 0.85, "learning_rate": 1.1823802486061096e-07, "loss": 1.164, "step": 10930 }, { "epoch": 0.85, "learning_rate": 1.1811947447542992e-07, "loss": 1.1912, "step": 10931 }, { "epoch": 0.85, "learning_rate": 1.1800097982139268e-07, "loss": 1.2205, "step": 10932 }, { "epoch": 0.85, "learning_rate": 1.1788254090598726e-07, "loss": 1.2535, "step": 10933 }, { "epoch": 0.85, "learning_rate": 1.1776415773669924e-07, "loss": 1.1774, "step": 10934 }, { "epoch": 0.85, "learning_rate": 1.1764583032100927e-07, "loss": 1.2133, "step": 10935 }, { "epoch": 0.85, "learning_rate": 1.1752755866639507e-07, "loss": 0.8167, "step": 10936 }, { "epoch": 0.85, "learning_rate": 1.1740934278033155e-07, "loss": 1.1427, "step": 10937 }, { "epoch": 0.85, "learning_rate": 1.1729118267028892e-07, "loss": 1.1891, "step": 10938 }, { "epoch": 0.85, "learning_rate": 1.1717307834373458e-07, "loss": 1.1904, "step": 10939 }, { "epoch": 0.85, "learning_rate": 1.1705502980813198e-07, "loss": 1.2026, "step": 10940 }, { "epoch": 0.85, "learning_rate": 1.1693703707094193e-07, "loss": 1.2004, "step": 10941 }, { "epoch": 0.85, "learning_rate": 1.1681910013962049e-07, "loss": 0.8416, "step": 10942 }, { "epoch": 0.85, "learning_rate": 1.1670121902162067e-07, "loss": 1.2363, "step": 10943 }, { "epoch": 0.85, "learning_rate": 1.1658339372439285e-07, "loss": 1.1839, "step": 10944 }, { "epoch": 0.85, "learning_rate": 1.164656242553823e-07, "loss": 1.1881, "step": 10945 }, { "epoch": 0.85, "learning_rate": 1.1634791062203164e-07, "loss": 1.1961, "step": 10946 }, { "epoch": 0.85, "learning_rate": 1.1623025283178056e-07, "loss": 1.1118, "step": 10947 }, { "epoch": 0.85, "learning_rate": 1.1611265089206379e-07, "loss": 1.1782, "step": 10948 }, { "epoch": 0.85, "learning_rate": 1.1599510481031327e-07, "loss": 1.2105, "step": 10949 }, { "epoch": 0.85, "learning_rate": 1.1587761459395806e-07, "loss": 0.8362, "step": 10950 }, { "epoch": 0.85, "learning_rate": 1.1576018025042278e-07, "loss": 1.1599, "step": 10951 }, { "epoch": 0.85, "learning_rate": 1.1564280178712827e-07, "loss": 1.1917, "step": 10952 }, { "epoch": 0.85, "learning_rate": 1.1552547921149302e-07, "loss": 1.241, "step": 10953 }, { "epoch": 0.85, "learning_rate": 1.1540821253093135e-07, "loss": 1.2151, "step": 10954 }, { "epoch": 0.85, "learning_rate": 1.1529100175285322e-07, "loss": 1.2114, "step": 10955 }, { "epoch": 0.85, "learning_rate": 1.151738468846667e-07, "loss": 1.1679, "step": 10956 }, { "epoch": 0.85, "learning_rate": 1.1505674793377507e-07, "loss": 1.1787, "step": 10957 }, { "epoch": 0.85, "learning_rate": 1.1493970490757887e-07, "loss": 1.1848, "step": 10958 }, { "epoch": 0.85, "learning_rate": 1.1482271781347409e-07, "loss": 1.2082, "step": 10959 }, { "epoch": 0.85, "learning_rate": 1.1470578665885433e-07, "loss": 1.1849, "step": 10960 }, { "epoch": 0.85, "learning_rate": 1.1458891145110916e-07, "loss": 1.22, "step": 10961 }, { "epoch": 0.85, "learning_rate": 1.144720921976241e-07, "loss": 1.2011, "step": 10962 }, { "epoch": 0.85, "learning_rate": 1.1435532890578214e-07, "loss": 1.2404, "step": 10963 }, { "epoch": 0.85, "learning_rate": 1.1423862158296238e-07, "loss": 1.1375, "step": 10964 }, { "epoch": 0.85, "learning_rate": 1.1412197023653936e-07, "loss": 1.2067, "step": 10965 }, { "epoch": 0.85, "learning_rate": 1.1400537487388584e-07, "loss": 1.1926, "step": 10966 }, { "epoch": 0.85, "learning_rate": 1.1388883550236961e-07, "loss": 1.1935, "step": 10967 }, { "epoch": 0.85, "learning_rate": 1.1377235212935609e-07, "loss": 1.1738, "step": 10968 }, { "epoch": 0.85, "learning_rate": 1.1365592476220586e-07, "loss": 0.8104, "step": 10969 }, { "epoch": 0.85, "learning_rate": 1.135395534082768e-07, "loss": 1.2139, "step": 10970 }, { "epoch": 0.85, "learning_rate": 1.1342323807492372e-07, "loss": 1.1628, "step": 10971 }, { "epoch": 0.85, "learning_rate": 1.1330697876949646e-07, "loss": 1.2439, "step": 10972 }, { "epoch": 0.85, "learning_rate": 1.1319077549934231e-07, "loss": 1.1728, "step": 10973 }, { "epoch": 0.85, "learning_rate": 1.1307462827180502e-07, "loss": 1.1638, "step": 10974 }, { "epoch": 0.85, "learning_rate": 1.1295853709422454e-07, "loss": 1.2364, "step": 10975 }, { "epoch": 0.85, "learning_rate": 1.1284250197393741e-07, "loss": 1.2167, "step": 10976 }, { "epoch": 0.85, "learning_rate": 1.1272652291827611e-07, "loss": 1.2057, "step": 10977 }, { "epoch": 0.85, "learning_rate": 1.126105999345709e-07, "loss": 1.2119, "step": 10978 }, { "epoch": 0.85, "learning_rate": 1.124947330301469e-07, "loss": 1.1726, "step": 10979 }, { "epoch": 0.85, "learning_rate": 1.1237892221232637e-07, "loss": 1.2348, "step": 10980 }, { "epoch": 0.85, "learning_rate": 1.1226316748842879e-07, "loss": 1.1179, "step": 10981 }, { "epoch": 0.85, "learning_rate": 1.1214746886576876e-07, "loss": 1.2098, "step": 10982 }, { "epoch": 0.85, "learning_rate": 1.1203182635165786e-07, "loss": 1.1959, "step": 10983 }, { "epoch": 0.85, "learning_rate": 1.119162399534046e-07, "loss": 0.8038, "step": 10984 }, { "epoch": 0.85, "learning_rate": 1.1180070967831367e-07, "loss": 1.167, "step": 10985 }, { "epoch": 0.85, "learning_rate": 1.1168523553368537e-07, "loss": 1.2163, "step": 10986 }, { "epoch": 0.85, "learning_rate": 1.1156981752681782e-07, "loss": 1.2343, "step": 10987 }, { "epoch": 0.85, "learning_rate": 1.1145445566500511e-07, "loss": 1.1641, "step": 10988 }, { "epoch": 0.85, "learning_rate": 1.1133914995553673e-07, "loss": 1.2661, "step": 10989 }, { "epoch": 0.85, "learning_rate": 1.112239004057004e-07, "loss": 1.2306, "step": 10990 }, { "epoch": 0.85, "learning_rate": 1.111087070227793e-07, "loss": 1.1636, "step": 10991 }, { "epoch": 0.85, "learning_rate": 1.1099356981405261e-07, "loss": 1.1554, "step": 10992 }, { "epoch": 0.85, "learning_rate": 1.1087848878679684e-07, "loss": 0.8648, "step": 10993 }, { "epoch": 0.85, "learning_rate": 1.1076346394828473e-07, "loss": 1.2585, "step": 10994 }, { "epoch": 0.85, "learning_rate": 1.1064849530578547e-07, "loss": 1.223, "step": 10995 }, { "epoch": 0.85, "learning_rate": 1.1053358286656412e-07, "loss": 1.1873, "step": 10996 }, { "epoch": 0.85, "learning_rate": 1.1041872663788299e-07, "loss": 1.164, "step": 10997 }, { "epoch": 0.85, "learning_rate": 1.1030392662700083e-07, "loss": 1.1908, "step": 10998 }, { "epoch": 0.85, "learning_rate": 1.1018918284117162e-07, "loss": 1.156, "step": 10999 }, { "epoch": 0.85, "learning_rate": 1.1007449528764756e-07, "loss": 0.8346, "step": 11000 }, { "epoch": 0.85, "learning_rate": 1.0995986397367585e-07, "loss": 1.197, "step": 11001 }, { "epoch": 0.85, "learning_rate": 1.0984528890650124e-07, "loss": 1.2135, "step": 11002 }, { "epoch": 0.85, "learning_rate": 1.0973077009336407e-07, "loss": 0.8307, "step": 11003 }, { "epoch": 0.85, "learning_rate": 1.0961630754150142e-07, "loss": 1.185, "step": 11004 }, { "epoch": 0.85, "learning_rate": 1.0950190125814685e-07, "loss": 0.7967, "step": 11005 }, { "epoch": 0.85, "learning_rate": 1.0938755125053045e-07, "loss": 1.2256, "step": 11006 }, { "epoch": 0.85, "learning_rate": 1.0927325752587857e-07, "loss": 1.1892, "step": 11007 }, { "epoch": 0.85, "learning_rate": 1.0915902009141408e-07, "loss": 1.1983, "step": 11008 }, { "epoch": 0.85, "learning_rate": 1.0904483895435646e-07, "loss": 1.16, "step": 11009 }, { "epoch": 0.85, "learning_rate": 1.0893071412192145e-07, "loss": 1.1375, "step": 11010 }, { "epoch": 0.85, "learning_rate": 1.0881664560132086e-07, "loss": 1.1815, "step": 11011 }, { "epoch": 0.85, "learning_rate": 1.0870263339976416e-07, "loss": 1.2053, "step": 11012 }, { "epoch": 0.85, "learning_rate": 1.0858867752445567e-07, "loss": 1.2097, "step": 11013 }, { "epoch": 0.85, "learning_rate": 1.0847477798259707e-07, "loss": 1.2059, "step": 11014 }, { "epoch": 0.85, "learning_rate": 1.0836093478138686e-07, "loss": 1.201, "step": 11015 }, { "epoch": 0.85, "learning_rate": 1.0824714792801893e-07, "loss": 1.2219, "step": 11016 }, { "epoch": 0.86, "learning_rate": 1.0813341742968396e-07, "loss": 1.1824, "step": 11017 }, { "epoch": 0.86, "learning_rate": 1.0801974329357e-07, "loss": 1.178, "step": 11018 }, { "epoch": 0.86, "learning_rate": 1.0790612552686007e-07, "loss": 1.1853, "step": 11019 }, { "epoch": 0.86, "learning_rate": 1.0779256413673444e-07, "loss": 1.1913, "step": 11020 }, { "epoch": 0.86, "learning_rate": 1.0767905913037001e-07, "loss": 1.2133, "step": 11021 }, { "epoch": 0.86, "learning_rate": 1.0756561051493996e-07, "loss": 1.2078, "step": 11022 }, { "epoch": 0.86, "learning_rate": 1.0745221829761319e-07, "loss": 1.1947, "step": 11023 }, { "epoch": 0.86, "learning_rate": 1.0733888248555567e-07, "loss": 1.1903, "step": 11024 }, { "epoch": 0.86, "learning_rate": 1.0722560308593043e-07, "loss": 1.2136, "step": 11025 }, { "epoch": 0.86, "learning_rate": 1.0711238010589562e-07, "loss": 1.2371, "step": 11026 }, { "epoch": 0.86, "learning_rate": 1.0699921355260633e-07, "loss": 1.2154, "step": 11027 }, { "epoch": 0.86, "learning_rate": 1.068861034332148e-07, "loss": 1.1687, "step": 11028 }, { "epoch": 0.86, "learning_rate": 1.06773049754869e-07, "loss": 1.2098, "step": 11029 }, { "epoch": 0.86, "learning_rate": 1.0666005252471288e-07, "loss": 1.1416, "step": 11030 }, { "epoch": 0.86, "learning_rate": 1.0654711174988796e-07, "loss": 1.1649, "step": 11031 }, { "epoch": 0.86, "learning_rate": 1.0643422743753172e-07, "loss": 1.1906, "step": 11032 }, { "epoch": 0.86, "learning_rate": 1.0632139959477727e-07, "loss": 1.1698, "step": 11033 }, { "epoch": 0.86, "learning_rate": 1.0620862822875543e-07, "loss": 1.1958, "step": 11034 }, { "epoch": 0.86, "learning_rate": 1.0609591334659295e-07, "loss": 1.1911, "step": 11035 }, { "epoch": 0.86, "learning_rate": 1.0598325495541239e-07, "loss": 1.2304, "step": 11036 }, { "epoch": 0.86, "learning_rate": 1.0587065306233378e-07, "loss": 1.1756, "step": 11037 }, { "epoch": 0.86, "learning_rate": 1.0575810767447291e-07, "loss": 1.1548, "step": 11038 }, { "epoch": 0.86, "learning_rate": 1.0564561879894218e-07, "loss": 1.2157, "step": 11039 }, { "epoch": 0.86, "learning_rate": 1.0553318644285037e-07, "loss": 1.2081, "step": 11040 }, { "epoch": 0.86, "learning_rate": 1.0542081061330287e-07, "loss": 1.2384, "step": 11041 }, { "epoch": 0.86, "learning_rate": 1.0530849131740127e-07, "loss": 1.1835, "step": 11042 }, { "epoch": 0.86, "learning_rate": 1.0519622856224364e-07, "loss": 1.2338, "step": 11043 }, { "epoch": 0.86, "learning_rate": 1.0508402235492464e-07, "loss": 1.2184, "step": 11044 }, { "epoch": 0.86, "learning_rate": 1.0497187270253493e-07, "loss": 1.2034, "step": 11045 }, { "epoch": 0.86, "learning_rate": 1.0485977961216242e-07, "loss": 1.1824, "step": 11046 }, { "epoch": 0.86, "learning_rate": 1.0474774309089051e-07, "loss": 0.7847, "step": 11047 }, { "epoch": 0.86, "learning_rate": 1.0463576314579936e-07, "loss": 1.1837, "step": 11048 }, { "epoch": 0.86, "learning_rate": 1.0452383978396617e-07, "loss": 1.1842, "step": 11049 }, { "epoch": 0.86, "learning_rate": 1.044119730124634e-07, "loss": 1.2135, "step": 11050 }, { "epoch": 0.86, "learning_rate": 1.0430016283836074e-07, "loss": 1.2178, "step": 11051 }, { "epoch": 0.86, "learning_rate": 1.0418840926872462e-07, "loss": 1.1662, "step": 11052 }, { "epoch": 0.86, "learning_rate": 1.0407671231061676e-07, "loss": 1.245, "step": 11053 }, { "epoch": 0.86, "learning_rate": 1.0396507197109594e-07, "loss": 0.8195, "step": 11054 }, { "epoch": 0.86, "learning_rate": 1.0385348825721785e-07, "loss": 1.1736, "step": 11055 }, { "epoch": 0.86, "learning_rate": 1.0374196117603395e-07, "loss": 1.2017, "step": 11056 }, { "epoch": 0.86, "learning_rate": 1.0363049073459206e-07, "loss": 0.8065, "step": 11057 }, { "epoch": 0.86, "learning_rate": 1.0351907693993667e-07, "loss": 1.1764, "step": 11058 }, { "epoch": 0.86, "learning_rate": 1.0340771979910901e-07, "loss": 1.2035, "step": 11059 }, { "epoch": 0.86, "learning_rate": 1.0329641931914601e-07, "loss": 1.1866, "step": 11060 }, { "epoch": 0.86, "learning_rate": 1.0318517550708128e-07, "loss": 1.2049, "step": 11061 }, { "epoch": 0.86, "learning_rate": 1.0307398836994563e-07, "loss": 1.1766, "step": 11062 }, { "epoch": 0.86, "learning_rate": 1.02962857914765e-07, "loss": 0.8318, "step": 11063 }, { "epoch": 0.86, "learning_rate": 1.0285178414856232e-07, "loss": 1.1815, "step": 11064 }, { "epoch": 0.86, "learning_rate": 1.0274076707835744e-07, "loss": 1.1921, "step": 11065 }, { "epoch": 0.86, "learning_rate": 1.0262980671116617e-07, "loss": 1.1763, "step": 11066 }, { "epoch": 0.86, "learning_rate": 1.0251890305400012e-07, "loss": 1.143, "step": 11067 }, { "epoch": 0.86, "learning_rate": 1.024080561138686e-07, "loss": 1.2024, "step": 11068 }, { "epoch": 0.86, "learning_rate": 1.0229726589777654e-07, "loss": 1.2046, "step": 11069 }, { "epoch": 0.86, "learning_rate": 1.021865324127249e-07, "loss": 0.8093, "step": 11070 }, { "epoch": 0.86, "learning_rate": 1.020758556657122e-07, "loss": 1.1975, "step": 11071 }, { "epoch": 0.86, "learning_rate": 1.0196523566373239e-07, "loss": 1.1683, "step": 11072 }, { "epoch": 0.86, "learning_rate": 1.0185467241377644e-07, "loss": 1.2264, "step": 11073 }, { "epoch": 0.86, "learning_rate": 1.0174416592283131e-07, "loss": 1.2175, "step": 11074 }, { "epoch": 0.86, "learning_rate": 1.0163371619788053e-07, "loss": 1.1925, "step": 11075 }, { "epoch": 0.86, "learning_rate": 1.0152332324590429e-07, "loss": 1.2238, "step": 11076 }, { "epoch": 0.86, "learning_rate": 1.0141298707387847e-07, "loss": 1.1949, "step": 11077 }, { "epoch": 0.86, "learning_rate": 1.0130270768877625e-07, "loss": 1.1892, "step": 11078 }, { "epoch": 0.86, "learning_rate": 1.0119248509756706e-07, "loss": 1.1521, "step": 11079 }, { "epoch": 0.86, "learning_rate": 1.0108231930721567e-07, "loss": 1.2103, "step": 11080 }, { "epoch": 0.86, "learning_rate": 1.0097221032468495e-07, "loss": 1.2345, "step": 11081 }, { "epoch": 0.86, "learning_rate": 1.0086215815693266e-07, "loss": 1.2401, "step": 11082 }, { "epoch": 0.86, "learning_rate": 1.0075216281091437e-07, "loss": 1.2062, "step": 11083 }, { "epoch": 0.86, "learning_rate": 1.0064222429358071e-07, "loss": 1.1966, "step": 11084 }, { "epoch": 0.86, "learning_rate": 1.0053234261187937e-07, "loss": 1.1315, "step": 11085 }, { "epoch": 0.86, "learning_rate": 1.0042251777275501e-07, "loss": 1.2274, "step": 11086 }, { "epoch": 0.86, "learning_rate": 1.0031274978314742e-07, "loss": 1.2291, "step": 11087 }, { "epoch": 0.86, "learning_rate": 1.0020303864999346e-07, "loss": 1.1847, "step": 11088 }, { "epoch": 0.86, "learning_rate": 1.0009338438022696e-07, "loss": 1.2031, "step": 11089 }, { "epoch": 0.86, "learning_rate": 9.998378698077747e-08, "loss": 1.2256, "step": 11090 }, { "epoch": 0.86, "learning_rate": 9.987424645857079e-08, "loss": 1.1503, "step": 11091 }, { "epoch": 0.86, "learning_rate": 9.976476282052926e-08, "loss": 1.1556, "step": 11092 }, { "epoch": 0.86, "learning_rate": 9.965533607357257e-08, "loss": 1.1582, "step": 11093 }, { "epoch": 0.86, "learning_rate": 9.95459662246153e-08, "loss": 1.2448, "step": 11094 }, { "epoch": 0.86, "learning_rate": 9.943665328056927e-08, "loss": 1.2129, "step": 11095 }, { "epoch": 0.86, "learning_rate": 9.932739724834316e-08, "loss": 1.1665, "step": 11096 }, { "epoch": 0.86, "learning_rate": 9.921819813484078e-08, "loss": 0.8215, "step": 11097 }, { "epoch": 0.86, "learning_rate": 9.910905594696328e-08, "loss": 1.222, "step": 11098 }, { "epoch": 0.86, "learning_rate": 9.899997069160814e-08, "loss": 0.8244, "step": 11099 }, { "epoch": 0.86, "learning_rate": 9.88909423756693e-08, "loss": 1.1468, "step": 11100 }, { "epoch": 0.86, "learning_rate": 9.878197100603613e-08, "loss": 0.8339, "step": 11101 }, { "epoch": 0.86, "learning_rate": 9.867305658959591e-08, "loss": 1.187, "step": 11102 }, { "epoch": 0.86, "learning_rate": 9.856419913323144e-08, "loss": 0.8607, "step": 11103 }, { "epoch": 0.86, "learning_rate": 9.845539864382157e-08, "loss": 1.1941, "step": 11104 }, { "epoch": 0.86, "learning_rate": 9.834665512824248e-08, "loss": 1.2226, "step": 11105 }, { "epoch": 0.86, "learning_rate": 9.823796859336642e-08, "loss": 1.2009, "step": 11106 }, { "epoch": 0.86, "learning_rate": 9.812933904606135e-08, "loss": 1.1894, "step": 11107 }, { "epoch": 0.86, "learning_rate": 9.80207664931928e-08, "loss": 1.1876, "step": 11108 }, { "epoch": 0.86, "learning_rate": 9.791225094162181e-08, "loss": 1.2161, "step": 11109 }, { "epoch": 0.86, "learning_rate": 9.780379239820624e-08, "loss": 1.1766, "step": 11110 }, { "epoch": 0.86, "learning_rate": 9.769539086979983e-08, "loss": 1.2204, "step": 11111 }, { "epoch": 0.86, "learning_rate": 9.758704636325354e-08, "loss": 1.195, "step": 11112 }, { "epoch": 0.86, "learning_rate": 9.747875888541434e-08, "loss": 0.8438, "step": 11113 }, { "epoch": 0.86, "learning_rate": 9.737052844312498e-08, "loss": 1.1504, "step": 11114 }, { "epoch": 0.86, "learning_rate": 9.726235504322578e-08, "loss": 1.1876, "step": 11115 }, { "epoch": 0.86, "learning_rate": 9.715423869255235e-08, "loss": 1.198, "step": 11116 }, { "epoch": 0.86, "learning_rate": 9.70461793979378e-08, "loss": 1.1277, "step": 11117 }, { "epoch": 0.86, "learning_rate": 9.693817716621056e-08, "loss": 1.2271, "step": 11118 }, { "epoch": 0.86, "learning_rate": 9.683023200419571e-08, "loss": 1.2575, "step": 11119 }, { "epoch": 0.86, "learning_rate": 9.67223439187157e-08, "loss": 1.1337, "step": 11120 }, { "epoch": 0.86, "learning_rate": 9.661451291658796e-08, "loss": 1.1821, "step": 11121 }, { "epoch": 0.86, "learning_rate": 9.65067390046268e-08, "loss": 1.182, "step": 11122 }, { "epoch": 0.86, "learning_rate": 9.63990221896439e-08, "loss": 1.2034, "step": 11123 }, { "epoch": 0.86, "learning_rate": 9.629136247844594e-08, "loss": 1.1495, "step": 11124 }, { "epoch": 0.86, "learning_rate": 9.618375987783655e-08, "loss": 1.1567, "step": 11125 }, { "epoch": 0.86, "learning_rate": 9.607621439461566e-08, "loss": 1.2212, "step": 11126 }, { "epoch": 0.86, "learning_rate": 9.596872603558026e-08, "loss": 1.181, "step": 11127 }, { "epoch": 0.86, "learning_rate": 9.586129480752259e-08, "loss": 1.1347, "step": 11128 }, { "epoch": 0.86, "learning_rate": 9.575392071723187e-08, "loss": 1.2133, "step": 11129 }, { "epoch": 0.86, "learning_rate": 9.564660377149436e-08, "loss": 1.1884, "step": 11130 }, { "epoch": 0.86, "learning_rate": 9.553934397709129e-08, "loss": 1.1483, "step": 11131 }, { "epoch": 0.86, "learning_rate": 9.543214134080112e-08, "loss": 1.2098, "step": 11132 }, { "epoch": 0.86, "learning_rate": 9.532499586939902e-08, "loss": 1.2178, "step": 11133 }, { "epoch": 0.86, "learning_rate": 9.521790756965597e-08, "loss": 1.1935, "step": 11134 }, { "epoch": 0.86, "learning_rate": 9.511087644833916e-08, "loss": 0.8094, "step": 11135 }, { "epoch": 0.86, "learning_rate": 9.50039025122129e-08, "loss": 1.18, "step": 11136 }, { "epoch": 0.86, "learning_rate": 9.489698576803761e-08, "loss": 1.1719, "step": 11137 }, { "epoch": 0.86, "learning_rate": 9.479012622256933e-08, "loss": 1.1598, "step": 11138 }, { "epoch": 0.86, "learning_rate": 9.468332388256173e-08, "loss": 1.183, "step": 11139 }, { "epoch": 0.86, "learning_rate": 9.457657875476421e-08, "loss": 0.849, "step": 11140 }, { "epoch": 0.86, "learning_rate": 9.446989084592217e-08, "loss": 0.8428, "step": 11141 }, { "epoch": 0.86, "learning_rate": 9.43632601627784e-08, "loss": 1.1565, "step": 11142 }, { "epoch": 0.86, "learning_rate": 9.425668671207121e-08, "loss": 1.2327, "step": 11143 }, { "epoch": 0.86, "learning_rate": 9.415017050053575e-08, "loss": 1.1656, "step": 11144 }, { "epoch": 0.86, "learning_rate": 9.404371153490298e-08, "loss": 1.2274, "step": 11145 }, { "epoch": 0.87, "learning_rate": 9.393730982190117e-08, "loss": 1.1565, "step": 11146 }, { "epoch": 0.87, "learning_rate": 9.383096536825452e-08, "loss": 1.2155, "step": 11147 }, { "epoch": 0.87, "learning_rate": 9.372467818068286e-08, "loss": 1.16, "step": 11148 }, { "epoch": 0.87, "learning_rate": 9.36184482659037e-08, "loss": 1.176, "step": 11149 }, { "epoch": 0.87, "learning_rate": 9.351227563063002e-08, "loss": 1.2148, "step": 11150 }, { "epoch": 0.87, "learning_rate": 9.340616028157189e-08, "loss": 1.1746, "step": 11151 }, { "epoch": 0.87, "learning_rate": 9.330010222543505e-08, "loss": 1.219, "step": 11152 }, { "epoch": 0.87, "learning_rate": 9.31941014689216e-08, "loss": 0.8078, "step": 11153 }, { "epoch": 0.87, "learning_rate": 9.308815801873115e-08, "loss": 1.2, "step": 11154 }, { "epoch": 0.87, "learning_rate": 9.298227188155827e-08, "loss": 1.2364, "step": 11155 }, { "epoch": 0.87, "learning_rate": 9.28764430640947e-08, "loss": 1.2127, "step": 11156 }, { "epoch": 0.87, "learning_rate": 9.277067157302831e-08, "loss": 1.2328, "step": 11157 }, { "epoch": 0.87, "learning_rate": 9.266495741504344e-08, "loss": 1.2248, "step": 11158 }, { "epoch": 0.87, "learning_rate": 9.255930059682082e-08, "loss": 1.1569, "step": 11159 }, { "epoch": 0.87, "learning_rate": 9.245370112503737e-08, "loss": 0.8169, "step": 11160 }, { "epoch": 0.87, "learning_rate": 9.234815900636694e-08, "loss": 0.8482, "step": 11161 }, { "epoch": 0.87, "learning_rate": 9.22426742474789e-08, "loss": 1.1923, "step": 11162 }, { "epoch": 0.87, "learning_rate": 9.213724685503954e-08, "loss": 1.1608, "step": 11163 }, { "epoch": 0.87, "learning_rate": 9.203187683571178e-08, "loss": 1.1812, "step": 11164 }, { "epoch": 0.87, "learning_rate": 9.192656419615408e-08, "loss": 1.2117, "step": 11165 }, { "epoch": 0.87, "learning_rate": 9.182130894302176e-08, "loss": 1.1352, "step": 11166 }, { "epoch": 0.87, "learning_rate": 9.171611108296717e-08, "loss": 1.2034, "step": 11167 }, { "epoch": 0.87, "learning_rate": 9.161097062263768e-08, "loss": 1.2323, "step": 11168 }, { "epoch": 0.87, "learning_rate": 9.150588756867772e-08, "loss": 1.2115, "step": 11169 }, { "epoch": 0.87, "learning_rate": 9.140086192772856e-08, "loss": 1.2073, "step": 11170 }, { "epoch": 0.87, "learning_rate": 9.129589370642731e-08, "loss": 1.1982, "step": 11171 }, { "epoch": 0.87, "learning_rate": 9.119098291140703e-08, "loss": 0.873, "step": 11172 }, { "epoch": 0.87, "learning_rate": 9.108612954929817e-08, "loss": 1.1706, "step": 11173 }, { "epoch": 0.87, "learning_rate": 9.0981333626727e-08, "loss": 1.1373, "step": 11174 }, { "epoch": 0.87, "learning_rate": 9.087659515031576e-08, "loss": 1.1629, "step": 11175 }, { "epoch": 0.87, "learning_rate": 9.077191412668362e-08, "loss": 1.2078, "step": 11176 }, { "epoch": 0.87, "learning_rate": 9.066729056244627e-08, "loss": 0.8022, "step": 11177 }, { "epoch": 0.87, "learning_rate": 9.056272446421554e-08, "loss": 1.2203, "step": 11178 }, { "epoch": 0.87, "learning_rate": 9.045821583859892e-08, "loss": 1.1894, "step": 11179 }, { "epoch": 0.87, "learning_rate": 9.035376469220146e-08, "loss": 0.8712, "step": 11180 }, { "epoch": 0.87, "learning_rate": 9.02493710316241e-08, "loss": 1.2003, "step": 11181 }, { "epoch": 0.87, "learning_rate": 9.014503486346358e-08, "loss": 1.1672, "step": 11182 }, { "epoch": 0.87, "learning_rate": 9.004075619431394e-08, "loss": 1.1481, "step": 11183 }, { "epoch": 0.87, "learning_rate": 8.993653503076515e-08, "loss": 1.1972, "step": 11184 }, { "epoch": 0.87, "learning_rate": 8.983237137940313e-08, "loss": 1.205, "step": 11185 }, { "epoch": 0.87, "learning_rate": 8.972826524681099e-08, "loss": 1.1992, "step": 11186 }, { "epoch": 0.87, "learning_rate": 8.962421663956754e-08, "loss": 1.145, "step": 11187 }, { "epoch": 0.87, "learning_rate": 8.952022556424877e-08, "loss": 0.8275, "step": 11188 }, { "epoch": 0.87, "learning_rate": 8.941629202742584e-08, "loss": 1.1897, "step": 11189 }, { "epoch": 0.87, "learning_rate": 8.93124160356672e-08, "loss": 1.1979, "step": 11190 }, { "epoch": 0.87, "learning_rate": 8.920859759553723e-08, "loss": 1.1972, "step": 11191 }, { "epoch": 0.87, "learning_rate": 8.910483671359703e-08, "loss": 0.8541, "step": 11192 }, { "epoch": 0.87, "learning_rate": 8.900113339640358e-08, "loss": 1.2723, "step": 11193 }, { "epoch": 0.87, "learning_rate": 8.889748765051053e-08, "loss": 1.1756, "step": 11194 }, { "epoch": 0.87, "learning_rate": 8.87938994824684e-08, "loss": 1.1874, "step": 11195 }, { "epoch": 0.87, "learning_rate": 8.869036889882286e-08, "loss": 1.1948, "step": 11196 }, { "epoch": 0.87, "learning_rate": 8.858689590611657e-08, "loss": 1.2096, "step": 11197 }, { "epoch": 0.87, "learning_rate": 8.848348051088938e-08, "loss": 1.1547, "step": 11198 }, { "epoch": 0.87, "learning_rate": 8.838012271967587e-08, "loss": 1.1696, "step": 11199 }, { "epoch": 0.87, "learning_rate": 8.827682253900803e-08, "loss": 1.2213, "step": 11200 }, { "epoch": 0.87, "learning_rate": 8.817357997541453e-08, "loss": 1.1508, "step": 11201 }, { "epoch": 0.87, "learning_rate": 8.807039503541913e-08, "loss": 1.2122, "step": 11202 }, { "epoch": 0.87, "learning_rate": 8.796726772554297e-08, "loss": 1.1671, "step": 11203 }, { "epoch": 0.87, "learning_rate": 8.786419805230339e-08, "loss": 0.8456, "step": 11204 }, { "epoch": 0.87, "learning_rate": 8.776118602221405e-08, "loss": 1.2057, "step": 11205 }, { "epoch": 0.87, "learning_rate": 8.76582316417842e-08, "loss": 1.2643, "step": 11206 }, { "epoch": 0.87, "learning_rate": 8.755533491752099e-08, "loss": 1.1486, "step": 11207 }, { "epoch": 0.87, "learning_rate": 8.745249585592673e-08, "loss": 1.1869, "step": 11208 }, { "epoch": 0.87, "learning_rate": 8.734971446350026e-08, "loss": 1.2116, "step": 11209 }, { "epoch": 0.87, "learning_rate": 8.72469907467368e-08, "loss": 1.2055, "step": 11210 }, { "epoch": 0.87, "learning_rate": 8.714432471212874e-08, "loss": 1.2073, "step": 11211 }, { "epoch": 0.87, "learning_rate": 8.704171636616354e-08, "loss": 0.8383, "step": 11212 }, { "epoch": 0.87, "learning_rate": 8.693916571532545e-08, "loss": 1.255, "step": 11213 }, { "epoch": 0.87, "learning_rate": 8.683667276609585e-08, "loss": 1.1941, "step": 11214 }, { "epoch": 0.87, "learning_rate": 8.673423752495179e-08, "loss": 1.213, "step": 11215 }, { "epoch": 0.87, "learning_rate": 8.663185999836609e-08, "loss": 1.1818, "step": 11216 }, { "epoch": 0.87, "learning_rate": 8.652954019280923e-08, "loss": 0.845, "step": 11217 }, { "epoch": 0.87, "learning_rate": 8.642727811474737e-08, "loss": 1.216, "step": 11218 }, { "epoch": 0.87, "learning_rate": 8.632507377064258e-08, "loss": 1.1789, "step": 11219 }, { "epoch": 0.87, "learning_rate": 8.622292716695412e-08, "loss": 1.1993, "step": 11220 }, { "epoch": 0.87, "learning_rate": 8.612083831013695e-08, "loss": 1.1924, "step": 11221 }, { "epoch": 0.87, "learning_rate": 8.601880720664323e-08, "loss": 1.1332, "step": 11222 }, { "epoch": 0.87, "learning_rate": 8.591683386292036e-08, "loss": 0.8607, "step": 11223 }, { "epoch": 0.87, "learning_rate": 8.581491828541277e-08, "loss": 1.1968, "step": 11224 }, { "epoch": 0.87, "learning_rate": 8.571306048056103e-08, "loss": 1.173, "step": 11225 }, { "epoch": 0.87, "learning_rate": 8.561126045480227e-08, "loss": 1.294, "step": 11226 }, { "epoch": 0.87, "learning_rate": 8.550951821456975e-08, "loss": 1.2009, "step": 11227 }, { "epoch": 0.87, "learning_rate": 8.540783376629323e-08, "loss": 1.1919, "step": 11228 }, { "epoch": 0.87, "learning_rate": 8.530620711639836e-08, "loss": 1.1546, "step": 11229 }, { "epoch": 0.87, "learning_rate": 8.520463827130808e-08, "loss": 1.1976, "step": 11230 }, { "epoch": 0.87, "learning_rate": 8.510312723744062e-08, "loss": 0.7982, "step": 11231 }, { "epoch": 0.87, "learning_rate": 8.500167402121162e-08, "loss": 1.2019, "step": 11232 }, { "epoch": 0.87, "learning_rate": 8.490027862903193e-08, "loss": 1.19, "step": 11233 }, { "epoch": 0.87, "learning_rate": 8.479894106730945e-08, "loss": 1.1925, "step": 11234 }, { "epoch": 0.87, "learning_rate": 8.469766134244871e-08, "loss": 0.852, "step": 11235 }, { "epoch": 0.87, "learning_rate": 8.45964394608497e-08, "loss": 1.2097, "step": 11236 }, { "epoch": 0.87, "learning_rate": 8.44952754289091e-08, "loss": 0.8435, "step": 11237 }, { "epoch": 0.87, "learning_rate": 8.439416925302056e-08, "loss": 1.1467, "step": 11238 }, { "epoch": 0.87, "learning_rate": 8.429312093957341e-08, "loss": 1.1929, "step": 11239 }, { "epoch": 0.87, "learning_rate": 8.41921304949531e-08, "loss": 1.1794, "step": 11240 }, { "epoch": 0.87, "learning_rate": 8.40911979255422e-08, "loss": 1.2237, "step": 11241 }, { "epoch": 0.87, "learning_rate": 8.399032323771926e-08, "loss": 1.1752, "step": 11242 }, { "epoch": 0.87, "learning_rate": 8.388950643785886e-08, "loss": 1.2067, "step": 11243 }, { "epoch": 0.87, "learning_rate": 8.378874753233212e-08, "loss": 0.8135, "step": 11244 }, { "epoch": 0.87, "learning_rate": 8.368804652750715e-08, "loss": 1.2069, "step": 11245 }, { "epoch": 0.87, "learning_rate": 8.358740342974735e-08, "loss": 1.1989, "step": 11246 }, { "epoch": 0.87, "learning_rate": 8.348681824541271e-08, "loss": 1.1948, "step": 11247 }, { "epoch": 0.87, "learning_rate": 8.338629098086047e-08, "loss": 1.2239, "step": 11248 }, { "epoch": 0.87, "learning_rate": 8.328582164244324e-08, "loss": 0.8279, "step": 11249 }, { "epoch": 0.87, "learning_rate": 8.318541023650993e-08, "loss": 1.1765, "step": 11250 }, { "epoch": 0.87, "learning_rate": 8.308505676940658e-08, "loss": 1.2058, "step": 11251 }, { "epoch": 0.87, "learning_rate": 8.2984761247475e-08, "loss": 1.2087, "step": 11252 }, { "epoch": 0.87, "learning_rate": 8.288452367705312e-08, "loss": 1.2014, "step": 11253 }, { "epoch": 0.87, "learning_rate": 8.278434406447599e-08, "loss": 1.2186, "step": 11254 }, { "epoch": 0.87, "learning_rate": 8.268422241607443e-08, "loss": 1.2041, "step": 11255 }, { "epoch": 0.87, "learning_rate": 8.258415873817537e-08, "loss": 1.1799, "step": 11256 }, { "epoch": 0.87, "learning_rate": 8.248415303710277e-08, "loss": 1.2331, "step": 11257 }, { "epoch": 0.87, "learning_rate": 8.238420531917645e-08, "loss": 1.1354, "step": 11258 }, { "epoch": 0.87, "learning_rate": 8.228431559071269e-08, "loss": 1.1336, "step": 11259 }, { "epoch": 0.87, "learning_rate": 8.218448385802412e-08, "loss": 1.1111, "step": 11260 }, { "epoch": 0.87, "learning_rate": 8.208471012741969e-08, "loss": 1.1977, "step": 11261 }, { "epoch": 0.87, "learning_rate": 8.198499440520479e-08, "loss": 1.2043, "step": 11262 }, { "epoch": 0.87, "learning_rate": 8.188533669768061e-08, "loss": 1.161, "step": 11263 }, { "epoch": 0.87, "learning_rate": 8.178573701114555e-08, "loss": 1.2319, "step": 11264 }, { "epoch": 0.87, "learning_rate": 8.168619535189347e-08, "loss": 1.1769, "step": 11265 }, { "epoch": 0.87, "learning_rate": 8.158671172621578e-08, "loss": 1.1543, "step": 11266 }, { "epoch": 0.87, "learning_rate": 8.148728614039857e-08, "loss": 0.8099, "step": 11267 }, { "epoch": 0.87, "learning_rate": 8.138791860072536e-08, "loss": 1.1801, "step": 11268 }, { "epoch": 0.87, "learning_rate": 8.128860911347624e-08, "loss": 1.2291, "step": 11269 }, { "epoch": 0.87, "learning_rate": 8.118935768492652e-08, "loss": 1.2151, "step": 11270 }, { "epoch": 0.87, "learning_rate": 8.109016432134852e-08, "loss": 1.1967, "step": 11271 }, { "epoch": 0.87, "learning_rate": 8.099102902901145e-08, "loss": 1.1848, "step": 11272 }, { "epoch": 0.87, "learning_rate": 8.089195181417952e-08, "loss": 1.2225, "step": 11273 }, { "epoch": 0.88, "learning_rate": 8.079293268311449e-08, "loss": 1.1995, "step": 11274 }, { "epoch": 0.88, "learning_rate": 8.069397164207348e-08, "loss": 1.1959, "step": 11275 }, { "epoch": 0.88, "learning_rate": 8.059506869731103e-08, "loss": 1.2109, "step": 11276 }, { "epoch": 0.88, "learning_rate": 8.049622385507704e-08, "loss": 1.1898, "step": 11277 }, { "epoch": 0.88, "learning_rate": 8.039743712161784e-08, "loss": 1.189, "step": 11278 }, { "epoch": 0.88, "learning_rate": 8.029870850317689e-08, "loss": 1.1869, "step": 11279 }, { "epoch": 0.88, "learning_rate": 8.020003800599306e-08, "loss": 1.1618, "step": 11280 }, { "epoch": 0.88, "learning_rate": 8.010142563630173e-08, "loss": 1.1855, "step": 11281 }, { "epoch": 0.88, "learning_rate": 8.000287140033535e-08, "loss": 1.143, "step": 11282 }, { "epoch": 0.88, "learning_rate": 7.990437530432193e-08, "loss": 1.1693, "step": 11283 }, { "epoch": 0.88, "learning_rate": 7.98059373544856e-08, "loss": 1.1756, "step": 11284 }, { "epoch": 0.88, "learning_rate": 7.970755755704761e-08, "loss": 1.1574, "step": 11285 }, { "epoch": 0.88, "learning_rate": 7.960923591822533e-08, "loss": 1.1954, "step": 11286 }, { "epoch": 0.88, "learning_rate": 7.951097244423166e-08, "loss": 1.1739, "step": 11287 }, { "epoch": 0.88, "learning_rate": 7.941276714127698e-08, "loss": 1.1137, "step": 11288 }, { "epoch": 0.88, "learning_rate": 7.931462001556744e-08, "loss": 0.8303, "step": 11289 }, { "epoch": 0.88, "learning_rate": 7.921653107330506e-08, "loss": 1.2103, "step": 11290 }, { "epoch": 0.88, "learning_rate": 7.911850032068901e-08, "loss": 1.1857, "step": 11291 }, { "epoch": 0.88, "learning_rate": 7.902052776391443e-08, "loss": 0.8129, "step": 11292 }, { "epoch": 0.88, "learning_rate": 7.892261340917272e-08, "loss": 1.1421, "step": 11293 }, { "epoch": 0.88, "learning_rate": 7.882475726265169e-08, "loss": 0.8231, "step": 11294 }, { "epoch": 0.88, "learning_rate": 7.872695933053531e-08, "loss": 0.8381, "step": 11295 }, { "epoch": 0.88, "learning_rate": 7.862921961900426e-08, "loss": 1.1475, "step": 11296 }, { "epoch": 0.88, "learning_rate": 7.853153813423475e-08, "loss": 1.2337, "step": 11297 }, { "epoch": 0.88, "learning_rate": 7.843391488240026e-08, "loss": 1.2095, "step": 11298 }, { "epoch": 0.88, "learning_rate": 7.833634986967041e-08, "loss": 1.1867, "step": 11299 }, { "epoch": 0.88, "learning_rate": 7.823884310221018e-08, "loss": 1.2078, "step": 11300 }, { "epoch": 0.88, "learning_rate": 7.814139458618207e-08, "loss": 1.2559, "step": 11301 }, { "epoch": 0.88, "learning_rate": 7.804400432774427e-08, "loss": 1.2362, "step": 11302 }, { "epoch": 0.88, "learning_rate": 7.794667233305185e-08, "loss": 1.2598, "step": 11303 }, { "epoch": 0.88, "learning_rate": 7.784939860825523e-08, "loss": 0.8164, "step": 11304 }, { "epoch": 0.88, "learning_rate": 7.775218315950171e-08, "loss": 1.2042, "step": 11305 }, { "epoch": 0.88, "learning_rate": 7.76550259929355e-08, "loss": 1.1559, "step": 11306 }, { "epoch": 0.88, "learning_rate": 7.755792711469588e-08, "loss": 1.2394, "step": 11307 }, { "epoch": 0.88, "learning_rate": 7.746088653091942e-08, "loss": 1.1448, "step": 11308 }, { "epoch": 0.88, "learning_rate": 7.73639042477383e-08, "loss": 1.2182, "step": 11309 }, { "epoch": 0.88, "learning_rate": 7.72669802712821e-08, "loss": 1.2037, "step": 11310 }, { "epoch": 0.88, "learning_rate": 7.717011460767542e-08, "loss": 1.2058, "step": 11311 }, { "epoch": 0.88, "learning_rate": 7.707330726303974e-08, "loss": 1.1172, "step": 11312 }, { "epoch": 0.88, "learning_rate": 7.697655824349336e-08, "loss": 1.1715, "step": 11313 }, { "epoch": 0.88, "learning_rate": 7.687986755514996e-08, "loss": 1.1649, "step": 11314 }, { "epoch": 0.88, "learning_rate": 7.678323520412e-08, "loss": 1.1788, "step": 11315 }, { "epoch": 0.88, "learning_rate": 7.668666119651079e-08, "loss": 1.2088, "step": 11316 }, { "epoch": 0.88, "learning_rate": 7.65901455384247e-08, "loss": 1.1518, "step": 11317 }, { "epoch": 0.88, "learning_rate": 7.649368823596136e-08, "loss": 1.1802, "step": 11318 }, { "epoch": 0.88, "learning_rate": 7.63972892952166e-08, "loss": 1.1992, "step": 11319 }, { "epoch": 0.88, "learning_rate": 7.630094872228254e-08, "loss": 1.2112, "step": 11320 }, { "epoch": 0.88, "learning_rate": 7.620466652324687e-08, "loss": 1.2648, "step": 11321 }, { "epoch": 0.88, "learning_rate": 7.610844270419492e-08, "loss": 1.1849, "step": 11322 }, { "epoch": 0.88, "learning_rate": 7.601227727120751e-08, "loss": 1.2009, "step": 11323 }, { "epoch": 0.88, "learning_rate": 7.591617023036146e-08, "loss": 0.8667, "step": 11324 }, { "epoch": 0.88, "learning_rate": 7.582012158773066e-08, "loss": 0.8111, "step": 11325 }, { "epoch": 0.88, "learning_rate": 7.572413134938505e-08, "loss": 1.1892, "step": 11326 }, { "epoch": 0.88, "learning_rate": 7.562819952139088e-08, "loss": 1.203, "step": 11327 }, { "epoch": 0.88, "learning_rate": 7.553232610980997e-08, "loss": 1.1873, "step": 11328 }, { "epoch": 0.88, "learning_rate": 7.543651112070193e-08, "loss": 0.8617, "step": 11329 }, { "epoch": 0.88, "learning_rate": 7.534075456012156e-08, "loss": 1.2313, "step": 11330 }, { "epoch": 0.88, "learning_rate": 7.524505643412004e-08, "loss": 1.182, "step": 11331 }, { "epoch": 0.88, "learning_rate": 7.514941674874542e-08, "loss": 1.1802, "step": 11332 }, { "epoch": 0.88, "learning_rate": 7.505383551004174e-08, "loss": 1.2263, "step": 11333 }, { "epoch": 0.88, "learning_rate": 7.495831272404884e-08, "loss": 1.1941, "step": 11334 }, { "epoch": 0.88, "learning_rate": 7.48628483968039e-08, "loss": 1.1984, "step": 11335 }, { "epoch": 0.88, "learning_rate": 7.476744253433954e-08, "loss": 1.1919, "step": 11336 }, { "epoch": 0.88, "learning_rate": 7.467209514268546e-08, "loss": 1.2008, "step": 11337 }, { "epoch": 0.88, "learning_rate": 7.457680622786677e-08, "loss": 1.1366, "step": 11338 }, { "epoch": 0.88, "learning_rate": 7.448157579590531e-08, "loss": 1.204, "step": 11339 }, { "epoch": 0.88, "learning_rate": 7.438640385281969e-08, "loss": 0.8421, "step": 11340 }, { "epoch": 0.88, "learning_rate": 7.42912904046239e-08, "loss": 1.2338, "step": 11341 }, { "epoch": 0.88, "learning_rate": 7.4196235457329e-08, "loss": 1.2, "step": 11342 }, { "epoch": 0.88, "learning_rate": 7.410123901694188e-08, "loss": 1.2155, "step": 11343 }, { "epoch": 0.88, "learning_rate": 7.400630108946592e-08, "loss": 1.1815, "step": 11344 }, { "epoch": 0.88, "learning_rate": 7.391142168090092e-08, "loss": 1.1562, "step": 11345 }, { "epoch": 0.88, "learning_rate": 7.381660079724272e-08, "loss": 0.8346, "step": 11346 }, { "epoch": 0.88, "learning_rate": 7.372183844448387e-08, "loss": 0.8518, "step": 11347 }, { "epoch": 0.88, "learning_rate": 7.362713462861281e-08, "loss": 1.2318, "step": 11348 }, { "epoch": 0.88, "learning_rate": 7.353248935561406e-08, "loss": 1.2023, "step": 11349 }, { "epoch": 0.88, "learning_rate": 7.343790263146953e-08, "loss": 1.1692, "step": 11350 }, { "epoch": 0.88, "learning_rate": 7.334337446215611e-08, "loss": 1.1881, "step": 11351 }, { "epoch": 0.88, "learning_rate": 7.324890485364766e-08, "loss": 0.8394, "step": 11352 }, { "epoch": 0.88, "learning_rate": 7.315449381191441e-08, "loss": 1.2193, "step": 11353 }, { "epoch": 0.88, "learning_rate": 7.306014134292304e-08, "loss": 1.178, "step": 11354 }, { "epoch": 0.88, "learning_rate": 7.296584745263545e-08, "loss": 0.8681, "step": 11355 }, { "epoch": 0.88, "learning_rate": 7.287161214701132e-08, "loss": 1.2249, "step": 11356 }, { "epoch": 0.88, "learning_rate": 7.277743543200576e-08, "loss": 1.2087, "step": 11357 }, { "epoch": 0.88, "learning_rate": 7.268331731357002e-08, "loss": 1.1819, "step": 11358 }, { "epoch": 0.88, "learning_rate": 7.258925779765235e-08, "loss": 1.2073, "step": 11359 }, { "epoch": 0.88, "learning_rate": 7.249525689019686e-08, "loss": 1.2065, "step": 11360 }, { "epoch": 0.88, "learning_rate": 7.240131459714383e-08, "loss": 1.2083, "step": 11361 }, { "epoch": 0.88, "learning_rate": 7.230743092442992e-08, "loss": 0.7982, "step": 11362 }, { "epoch": 0.88, "learning_rate": 7.221360587798853e-08, "loss": 1.1624, "step": 11363 }, { "epoch": 0.88, "learning_rate": 7.2119839463749e-08, "loss": 1.218, "step": 11364 }, { "epoch": 0.88, "learning_rate": 7.20261316876366e-08, "loss": 1.1614, "step": 11365 }, { "epoch": 0.88, "learning_rate": 7.193248255557361e-08, "loss": 0.8271, "step": 11366 }, { "epoch": 0.88, "learning_rate": 7.183889207347826e-08, "loss": 1.2057, "step": 11367 }, { "epoch": 0.88, "learning_rate": 7.174536024726463e-08, "loss": 1.1763, "step": 11368 }, { "epoch": 0.88, "learning_rate": 7.165188708284408e-08, "loss": 1.1861, "step": 11369 }, { "epoch": 0.88, "learning_rate": 7.155847258612335e-08, "loss": 1.1835, "step": 11370 }, { "epoch": 0.88, "learning_rate": 7.146511676300637e-08, "loss": 1.2059, "step": 11371 }, { "epoch": 0.88, "learning_rate": 7.13718196193922e-08, "loss": 1.2128, "step": 11372 }, { "epoch": 0.88, "learning_rate": 7.127858116117702e-08, "loss": 1.1688, "step": 11373 }, { "epoch": 0.88, "learning_rate": 7.118540139425355e-08, "loss": 1.1919, "step": 11374 }, { "epoch": 0.88, "learning_rate": 7.109228032450975e-08, "loss": 1.1853, "step": 11375 }, { "epoch": 0.88, "learning_rate": 7.09992179578307e-08, "loss": 1.1889, "step": 11376 }, { "epoch": 0.88, "learning_rate": 7.09062143000977e-08, "loss": 1.1709, "step": 11377 }, { "epoch": 0.88, "learning_rate": 7.081326935718801e-08, "loss": 0.8122, "step": 11378 }, { "epoch": 0.88, "learning_rate": 7.072038313497542e-08, "loss": 0.8456, "step": 11379 }, { "epoch": 0.88, "learning_rate": 7.062755563932976e-08, "loss": 1.2019, "step": 11380 }, { "epoch": 0.88, "learning_rate": 7.05347868761178e-08, "loss": 1.2378, "step": 11381 }, { "epoch": 0.88, "learning_rate": 7.044207685120185e-08, "loss": 1.1803, "step": 11382 }, { "epoch": 0.88, "learning_rate": 7.034942557044054e-08, "loss": 0.804, "step": 11383 }, { "epoch": 0.88, "learning_rate": 7.025683303968966e-08, "loss": 1.1758, "step": 11384 }, { "epoch": 0.88, "learning_rate": 7.016429926480016e-08, "loss": 1.1842, "step": 11385 }, { "epoch": 0.88, "learning_rate": 7.007182425161973e-08, "loss": 1.1896, "step": 11386 }, { "epoch": 0.88, "learning_rate": 6.9979408005993e-08, "loss": 1.1743, "step": 11387 }, { "epoch": 0.88, "learning_rate": 6.988705053375966e-08, "loss": 0.8623, "step": 11388 }, { "epoch": 0.88, "learning_rate": 6.979475184075645e-08, "loss": 1.1785, "step": 11389 }, { "epoch": 0.88, "learning_rate": 6.970251193281651e-08, "loss": 1.2, "step": 11390 }, { "epoch": 0.88, "learning_rate": 6.961033081576906e-08, "loss": 0.8488, "step": 11391 }, { "epoch": 0.88, "learning_rate": 6.95182084954391e-08, "loss": 1.189, "step": 11392 }, { "epoch": 0.88, "learning_rate": 6.942614497764865e-08, "loss": 1.269, "step": 11393 }, { "epoch": 0.88, "learning_rate": 6.933414026821605e-08, "loss": 1.1478, "step": 11394 }, { "epoch": 0.88, "learning_rate": 6.924219437295509e-08, "loss": 1.1949, "step": 11395 }, { "epoch": 0.88, "learning_rate": 6.915030729767634e-08, "loss": 1.1125, "step": 11396 }, { "epoch": 0.88, "learning_rate": 6.905847904818707e-08, "loss": 1.2371, "step": 11397 }, { "epoch": 0.88, "learning_rate": 6.896670963029049e-08, "loss": 1.2154, "step": 11398 }, { "epoch": 0.88, "learning_rate": 6.887499904978544e-08, "loss": 0.823, "step": 11399 }, { "epoch": 0.88, "learning_rate": 6.878334731246826e-08, "loss": 1.2163, "step": 11400 }, { "epoch": 0.88, "learning_rate": 6.869175442413089e-08, "loss": 1.1286, "step": 11401 }, { "epoch": 0.88, "learning_rate": 6.860022039056113e-08, "loss": 1.1743, "step": 11402 }, { "epoch": 0.89, "learning_rate": 6.85087452175439e-08, "loss": 1.1707, "step": 11403 }, { "epoch": 0.89, "learning_rate": 6.841732891086028e-08, "loss": 1.1977, "step": 11404 }, { "epoch": 0.89, "learning_rate": 6.832597147628682e-08, "loss": 1.1794, "step": 11405 }, { "epoch": 0.89, "learning_rate": 6.823467291959739e-08, "loss": 1.1742, "step": 11406 }, { "epoch": 0.89, "learning_rate": 6.814343324656147e-08, "loss": 1.1984, "step": 11407 }, { "epoch": 0.89, "learning_rate": 6.80522524629451e-08, "loss": 0.8419, "step": 11408 }, { "epoch": 0.89, "learning_rate": 6.796113057451048e-08, "loss": 1.2174, "step": 11409 }, { "epoch": 0.89, "learning_rate": 6.787006758701608e-08, "loss": 1.2166, "step": 11410 }, { "epoch": 0.89, "learning_rate": 6.777906350621687e-08, "loss": 1.2095, "step": 11411 }, { "epoch": 0.89, "learning_rate": 6.768811833786381e-08, "loss": 1.1494, "step": 11412 }, { "epoch": 0.89, "learning_rate": 6.759723208770419e-08, "loss": 1.2035, "step": 11413 }, { "epoch": 0.89, "learning_rate": 6.750640476148162e-08, "loss": 0.8009, "step": 11414 }, { "epoch": 0.89, "learning_rate": 6.741563636493642e-08, "loss": 1.1645, "step": 11415 }, { "epoch": 0.89, "learning_rate": 6.73249269038042e-08, "loss": 1.2045, "step": 11416 }, { "epoch": 0.89, "learning_rate": 6.723427638381763e-08, "loss": 1.2341, "step": 11417 }, { "epoch": 0.89, "learning_rate": 6.714368481070565e-08, "loss": 1.2054, "step": 11418 }, { "epoch": 0.89, "learning_rate": 6.705315219019302e-08, "loss": 1.1854, "step": 11419 }, { "epoch": 0.89, "learning_rate": 6.696267852800098e-08, "loss": 1.1444, "step": 11420 }, { "epoch": 0.89, "learning_rate": 6.687226382984745e-08, "loss": 1.1674, "step": 11421 }, { "epoch": 0.89, "learning_rate": 6.67819081014459e-08, "loss": 1.1311, "step": 11422 }, { "epoch": 0.89, "learning_rate": 6.669161134850631e-08, "loss": 1.2012, "step": 11423 }, { "epoch": 0.89, "learning_rate": 6.660137357673545e-08, "loss": 1.2011, "step": 11424 }, { "epoch": 0.89, "learning_rate": 6.651119479183609e-08, "loss": 1.2325, "step": 11425 }, { "epoch": 0.89, "learning_rate": 6.642107499950655e-08, "loss": 1.1617, "step": 11426 }, { "epoch": 0.89, "learning_rate": 6.63310142054423e-08, "loss": 1.2036, "step": 11427 }, { "epoch": 0.89, "learning_rate": 6.624101241533508e-08, "loss": 1.1112, "step": 11428 }, { "epoch": 0.89, "learning_rate": 6.615106963487227e-08, "loss": 1.2038, "step": 11429 }, { "epoch": 0.89, "learning_rate": 6.606118586973774e-08, "loss": 1.146, "step": 11430 }, { "epoch": 0.89, "learning_rate": 6.597136112561241e-08, "loss": 1.2141, "step": 11431 }, { "epoch": 0.89, "learning_rate": 6.588159540817228e-08, "loss": 0.8724, "step": 11432 }, { "epoch": 0.89, "learning_rate": 6.579188872309016e-08, "loss": 1.1548, "step": 11433 }, { "epoch": 0.89, "learning_rate": 6.57022410760355e-08, "loss": 0.8284, "step": 11434 }, { "epoch": 0.89, "learning_rate": 6.561265247267356e-08, "loss": 1.1948, "step": 11435 }, { "epoch": 0.89, "learning_rate": 6.552312291866547e-08, "loss": 0.8369, "step": 11436 }, { "epoch": 0.89, "learning_rate": 6.54336524196698e-08, "loss": 1.1474, "step": 11437 }, { "epoch": 0.89, "learning_rate": 6.53442409813405e-08, "loss": 1.204, "step": 11438 }, { "epoch": 0.89, "learning_rate": 6.525488860932759e-08, "loss": 1.2365, "step": 11439 }, { "epoch": 0.89, "learning_rate": 6.516559530927823e-08, "loss": 1.2155, "step": 11440 }, { "epoch": 0.89, "learning_rate": 6.507636108683523e-08, "loss": 1.16, "step": 11441 }, { "epoch": 0.89, "learning_rate": 6.498718594763785e-08, "loss": 1.2142, "step": 11442 }, { "epoch": 0.89, "learning_rate": 6.48980698973216e-08, "loss": 1.1616, "step": 11443 }, { "epoch": 0.89, "learning_rate": 6.480901294151819e-08, "loss": 1.1684, "step": 11444 }, { "epoch": 0.89, "learning_rate": 6.472001508585557e-08, "loss": 1.192, "step": 11445 }, { "epoch": 0.89, "learning_rate": 6.463107633595822e-08, "loss": 1.2151, "step": 11446 }, { "epoch": 0.89, "learning_rate": 6.454219669744654e-08, "loss": 1.1719, "step": 11447 }, { "epoch": 0.89, "learning_rate": 6.44533761759376e-08, "loss": 1.1562, "step": 11448 }, { "epoch": 0.89, "learning_rate": 6.436461477704391e-08, "loss": 1.1898, "step": 11449 }, { "epoch": 0.89, "learning_rate": 6.427591250637531e-08, "loss": 1.1557, "step": 11450 }, { "epoch": 0.89, "learning_rate": 6.418726936953722e-08, "loss": 1.2076, "step": 11451 }, { "epoch": 0.89, "learning_rate": 6.409868537213193e-08, "loss": 1.2015, "step": 11452 }, { "epoch": 0.89, "learning_rate": 6.401016051975705e-08, "loss": 1.188, "step": 11453 }, { "epoch": 0.89, "learning_rate": 6.392169481800691e-08, "loss": 1.1526, "step": 11454 }, { "epoch": 0.89, "learning_rate": 6.38332882724728e-08, "loss": 1.1937, "step": 11455 }, { "epoch": 0.89, "learning_rate": 6.374494088874117e-08, "loss": 1.1712, "step": 11456 }, { "epoch": 0.89, "learning_rate": 6.365665267239506e-08, "loss": 1.1818, "step": 11457 }, { "epoch": 0.89, "learning_rate": 6.356842362901437e-08, "loss": 1.1821, "step": 11458 }, { "epoch": 0.89, "learning_rate": 6.348025376417476e-08, "loss": 1.1755, "step": 11459 }, { "epoch": 0.89, "learning_rate": 6.339214308344776e-08, "loss": 1.2047, "step": 11460 }, { "epoch": 0.89, "learning_rate": 6.33040915924018e-08, "loss": 1.2064, "step": 11461 }, { "epoch": 0.89, "learning_rate": 6.321609929660165e-08, "loss": 1.1722, "step": 11462 }, { "epoch": 0.89, "learning_rate": 6.312816620160777e-08, "loss": 1.1571, "step": 11463 }, { "epoch": 0.89, "learning_rate": 6.304029231297691e-08, "loss": 1.1676, "step": 11464 }, { "epoch": 0.89, "learning_rate": 6.29524776362631e-08, "loss": 1.1722, "step": 11465 }, { "epoch": 0.89, "learning_rate": 6.286472217701522e-08, "loss": 1.1898, "step": 11466 }, { "epoch": 0.89, "learning_rate": 6.277702594077893e-08, "loss": 1.2017, "step": 11467 }, { "epoch": 0.89, "learning_rate": 6.268938893309683e-08, "loss": 1.1812, "step": 11468 }, { "epoch": 0.89, "learning_rate": 6.260181115950714e-08, "loss": 1.1567, "step": 11469 }, { "epoch": 0.89, "learning_rate": 6.251429262554376e-08, "loss": 1.1891, "step": 11470 }, { "epoch": 0.89, "learning_rate": 6.242683333673814e-08, "loss": 1.1737, "step": 11471 }, { "epoch": 0.89, "learning_rate": 6.233943329861735e-08, "loss": 0.8339, "step": 11472 }, { "epoch": 0.89, "learning_rate": 6.225209251670416e-08, "loss": 1.249, "step": 11473 }, { "epoch": 0.89, "learning_rate": 6.21648109965186e-08, "loss": 1.1664, "step": 11474 }, { "epoch": 0.89, "learning_rate": 6.20775887435765e-08, "loss": 1.2451, "step": 11475 }, { "epoch": 0.89, "learning_rate": 6.199042576338954e-08, "loss": 1.2025, "step": 11476 }, { "epoch": 0.89, "learning_rate": 6.190332206146653e-08, "loss": 0.8522, "step": 11477 }, { "epoch": 0.89, "learning_rate": 6.181627764331177e-08, "loss": 1.2228, "step": 11478 }, { "epoch": 0.89, "learning_rate": 6.172929251442627e-08, "loss": 1.2082, "step": 11479 }, { "epoch": 0.89, "learning_rate": 6.164236668030687e-08, "loss": 0.8421, "step": 11480 }, { "epoch": 0.89, "learning_rate": 6.155550014644716e-08, "loss": 1.1702, "step": 11481 }, { "epoch": 0.89, "learning_rate": 6.14686929183369e-08, "loss": 1.2221, "step": 11482 }, { "epoch": 0.89, "learning_rate": 6.138194500146132e-08, "loss": 0.8516, "step": 11483 }, { "epoch": 0.89, "learning_rate": 6.129525640130306e-08, "loss": 1.1762, "step": 11484 }, { "epoch": 0.89, "learning_rate": 6.120862712334019e-08, "loss": 1.1788, "step": 11485 }, { "epoch": 0.89, "learning_rate": 6.112205717304775e-08, "loss": 1.1996, "step": 11486 }, { "epoch": 0.89, "learning_rate": 6.103554655589615e-08, "loss": 1.1848, "step": 11487 }, { "epoch": 0.89, "learning_rate": 6.094909527735248e-08, "loss": 1.2418, "step": 11488 }, { "epoch": 0.89, "learning_rate": 6.086270334288057e-08, "loss": 1.2144, "step": 11489 }, { "epoch": 0.89, "learning_rate": 6.077637075793962e-08, "loss": 1.2267, "step": 11490 }, { "epoch": 0.89, "learning_rate": 6.069009752798538e-08, "loss": 1.2206, "step": 11491 }, { "epoch": 0.89, "learning_rate": 6.060388365847047e-08, "loss": 1.1297, "step": 11492 }, { "epoch": 0.89, "learning_rate": 6.051772915484277e-08, "loss": 1.1714, "step": 11493 }, { "epoch": 0.89, "learning_rate": 6.043163402254703e-08, "loss": 1.2049, "step": 11494 }, { "epoch": 0.89, "learning_rate": 6.034559826702401e-08, "loss": 1.218, "step": 11495 }, { "epoch": 0.89, "learning_rate": 6.025962189371125e-08, "loss": 1.2246, "step": 11496 }, { "epoch": 0.89, "learning_rate": 6.017370490804164e-08, "loss": 1.2204, "step": 11497 }, { "epoch": 0.89, "learning_rate": 6.00878473154447e-08, "loss": 1.1261, "step": 11498 }, { "epoch": 0.89, "learning_rate": 6.000204912134676e-08, "loss": 1.108, "step": 11499 }, { "epoch": 0.89, "learning_rate": 5.99163103311694e-08, "loss": 0.8533, "step": 11500 }, { "epoch": 0.89, "learning_rate": 5.983063095033103e-08, "loss": 1.2129, "step": 11501 }, { "epoch": 0.89, "learning_rate": 5.974501098424656e-08, "loss": 1.1541, "step": 11502 }, { "epoch": 0.89, "learning_rate": 5.965945043832676e-08, "loss": 0.7821, "step": 11503 }, { "epoch": 0.89, "learning_rate": 5.957394931797832e-08, "loss": 1.1611, "step": 11504 }, { "epoch": 0.89, "learning_rate": 5.94885076286048e-08, "loss": 1.1764, "step": 11505 }, { "epoch": 0.89, "learning_rate": 5.940312537560599e-08, "loss": 1.1728, "step": 11506 }, { "epoch": 0.89, "learning_rate": 5.931780256437702e-08, "loss": 1.2078, "step": 11507 }, { "epoch": 0.89, "learning_rate": 5.9232539200310574e-08, "loss": 1.2113, "step": 11508 }, { "epoch": 0.89, "learning_rate": 5.914733528879479e-08, "loss": 1.167, "step": 11509 }, { "epoch": 0.89, "learning_rate": 5.906219083521391e-08, "loss": 1.1992, "step": 11510 }, { "epoch": 0.89, "learning_rate": 5.8977105844949084e-08, "loss": 0.8228, "step": 11511 }, { "epoch": 0.89, "learning_rate": 5.8892080323377116e-08, "loss": 1.1543, "step": 11512 }, { "epoch": 0.89, "learning_rate": 5.880711427587137e-08, "loss": 1.2121, "step": 11513 }, { "epoch": 0.89, "learning_rate": 5.872220770780112e-08, "loss": 1.1837, "step": 11514 }, { "epoch": 0.89, "learning_rate": 5.863736062453228e-08, "loss": 1.2303, "step": 11515 }, { "epoch": 0.89, "learning_rate": 5.855257303142702e-08, "loss": 1.239, "step": 11516 }, { "epoch": 0.89, "learning_rate": 5.8467844933843046e-08, "loss": 1.227, "step": 11517 }, { "epoch": 0.89, "learning_rate": 5.8383176337135297e-08, "loss": 1.2029, "step": 11518 }, { "epoch": 0.89, "learning_rate": 5.829856724665438e-08, "loss": 1.1736, "step": 11519 }, { "epoch": 0.89, "learning_rate": 5.821401766774692e-08, "loss": 1.1582, "step": 11520 }, { "epoch": 0.89, "learning_rate": 5.81295276057564e-08, "loss": 1.1641, "step": 11521 }, { "epoch": 0.89, "learning_rate": 5.804509706602212e-08, "loss": 1.2348, "step": 11522 }, { "epoch": 0.89, "learning_rate": 5.796072605388003e-08, "loss": 1.2265, "step": 11523 }, { "epoch": 0.89, "learning_rate": 5.7876414574661636e-08, "loss": 1.1831, "step": 11524 }, { "epoch": 0.89, "learning_rate": 5.779216263369513e-08, "loss": 1.1924, "step": 11525 }, { "epoch": 0.89, "learning_rate": 5.7707970236305027e-08, "loss": 1.2596, "step": 11526 }, { "epoch": 0.89, "learning_rate": 5.762383738781185e-08, "loss": 0.8323, "step": 11527 }, { "epoch": 0.89, "learning_rate": 5.753976409353245e-08, "loss": 1.2482, "step": 11528 }, { "epoch": 0.89, "learning_rate": 5.74557503587797e-08, "loss": 1.1359, "step": 11529 }, { "epoch": 0.89, "learning_rate": 5.737179618886345e-08, "loss": 1.1958, "step": 11530 }, { "epoch": 0.89, "learning_rate": 5.7287901589088804e-08, "loss": 0.8622, "step": 11531 }, { "epoch": 0.9, "learning_rate": 5.720406656475751e-08, "loss": 1.2011, "step": 11532 }, { "epoch": 0.9, "learning_rate": 5.712029112116812e-08, "loss": 1.1987, "step": 11533 }, { "epoch": 0.9, "learning_rate": 5.703657526361427e-08, "loss": 1.2073, "step": 11534 }, { "epoch": 0.9, "learning_rate": 5.6952918997386524e-08, "loss": 1.1756, "step": 11535 }, { "epoch": 0.9, "learning_rate": 5.68693223277722e-08, "loss": 1.1949, "step": 11536 }, { "epoch": 0.9, "learning_rate": 5.678578526005362e-08, "loss": 1.2421, "step": 11537 }, { "epoch": 0.9, "learning_rate": 5.670230779951002e-08, "loss": 1.1952, "step": 11538 }, { "epoch": 0.9, "learning_rate": 5.661888995141717e-08, "loss": 1.1639, "step": 11539 }, { "epoch": 0.9, "learning_rate": 5.6535531721046746e-08, "loss": 1.1258, "step": 11540 }, { "epoch": 0.9, "learning_rate": 5.645223311366609e-08, "loss": 1.2473, "step": 11541 }, { "epoch": 0.9, "learning_rate": 5.6368994134539773e-08, "loss": 1.1459, "step": 11542 }, { "epoch": 0.9, "learning_rate": 5.628581478892835e-08, "loss": 1.2278, "step": 11543 }, { "epoch": 0.9, "learning_rate": 5.6202695082087745e-08, "loss": 1.1959, "step": 11544 }, { "epoch": 0.9, "learning_rate": 5.6119635019271305e-08, "loss": 1.1572, "step": 11545 }, { "epoch": 0.9, "learning_rate": 5.603663460572783e-08, "loss": 1.2035, "step": 11546 }, { "epoch": 0.9, "learning_rate": 5.59536938467029e-08, "loss": 1.182, "step": 11547 }, { "epoch": 0.9, "learning_rate": 5.587081274743755e-08, "loss": 1.1655, "step": 11548 }, { "epoch": 0.9, "learning_rate": 5.5787991313169915e-08, "loss": 1.1984, "step": 11549 }, { "epoch": 0.9, "learning_rate": 5.5705229549133926e-08, "loss": 1.1707, "step": 11550 }, { "epoch": 0.9, "learning_rate": 5.562252746055929e-08, "loss": 1.2276, "step": 11551 }, { "epoch": 0.9, "learning_rate": 5.553988505267304e-08, "loss": 1.1821, "step": 11552 }, { "epoch": 0.9, "learning_rate": 5.5457302330697785e-08, "loss": 1.1773, "step": 11553 }, { "epoch": 0.9, "learning_rate": 5.5374779299851904e-08, "loss": 1.2027, "step": 11554 }, { "epoch": 0.9, "learning_rate": 5.5292315965351e-08, "loss": 1.1714, "step": 11555 }, { "epoch": 0.9, "learning_rate": 5.520991233240602e-08, "loss": 1.1716, "step": 11556 }, { "epoch": 0.9, "learning_rate": 5.5127568406225014e-08, "loss": 1.2629, "step": 11557 }, { "epoch": 0.9, "learning_rate": 5.5045284192011376e-08, "loss": 1.1714, "step": 11558 }, { "epoch": 0.9, "learning_rate": 5.4963059694965284e-08, "loss": 1.1777, "step": 11559 }, { "epoch": 0.9, "learning_rate": 5.4880894920282913e-08, "loss": 1.2237, "step": 11560 }, { "epoch": 0.9, "learning_rate": 5.479878987315678e-08, "loss": 0.8615, "step": 11561 }, { "epoch": 0.9, "learning_rate": 5.4716744558775615e-08, "loss": 1.205, "step": 11562 }, { "epoch": 0.9, "learning_rate": 5.463475898232417e-08, "loss": 1.2338, "step": 11563 }, { "epoch": 0.9, "learning_rate": 5.4552833148983734e-08, "loss": 1.231, "step": 11564 }, { "epoch": 0.9, "learning_rate": 5.447096706393173e-08, "loss": 1.2554, "step": 11565 }, { "epoch": 0.9, "learning_rate": 5.438916073234145e-08, "loss": 1.1351, "step": 11566 }, { "epoch": 0.9, "learning_rate": 5.4307414159383117e-08, "loss": 1.1691, "step": 11567 }, { "epoch": 0.9, "learning_rate": 5.4225727350222575e-08, "loss": 1.1785, "step": 11568 }, { "epoch": 0.9, "learning_rate": 5.4144100310021924e-08, "loss": 0.8346, "step": 11569 }, { "epoch": 0.9, "learning_rate": 5.406253304394004e-08, "loss": 1.2073, "step": 11570 }, { "epoch": 0.9, "learning_rate": 5.3981025557131246e-08, "loss": 1.1646, "step": 11571 }, { "epoch": 0.9, "learning_rate": 5.3899577854746526e-08, "loss": 1.1755, "step": 11572 }, { "epoch": 0.9, "learning_rate": 5.3818189941933324e-08, "loss": 1.2194, "step": 11573 }, { "epoch": 0.9, "learning_rate": 5.373686182383497e-08, "loss": 0.8199, "step": 11574 }, { "epoch": 0.9, "learning_rate": 5.365559350559057e-08, "loss": 1.211, "step": 11575 }, { "epoch": 0.9, "learning_rate": 5.3574384992336464e-08, "loss": 1.2277, "step": 11576 }, { "epoch": 0.9, "learning_rate": 5.349323628920466e-08, "loss": 1.2118, "step": 11577 }, { "epoch": 0.9, "learning_rate": 5.341214740132316e-08, "loss": 1.2076, "step": 11578 }, { "epoch": 0.9, "learning_rate": 5.3331118333816314e-08, "loss": 1.2056, "step": 11579 }, { "epoch": 0.9, "learning_rate": 5.3250149091805476e-08, "loss": 1.2465, "step": 11580 }, { "epoch": 0.9, "learning_rate": 5.3169239680406874e-08, "loss": 0.8374, "step": 11581 }, { "epoch": 0.9, "learning_rate": 5.308839010473376e-08, "loss": 1.1833, "step": 11582 }, { "epoch": 0.9, "learning_rate": 5.300760036989593e-08, "loss": 0.7675, "step": 11583 }, { "epoch": 0.9, "learning_rate": 5.2926870480998644e-08, "loss": 1.183, "step": 11584 }, { "epoch": 0.9, "learning_rate": 5.284620044314347e-08, "loss": 1.2088, "step": 11585 }, { "epoch": 0.9, "learning_rate": 5.27655902614288e-08, "loss": 1.1962, "step": 11586 }, { "epoch": 0.9, "learning_rate": 5.268503994094897e-08, "loss": 1.1909, "step": 11587 }, { "epoch": 0.9, "learning_rate": 5.260454948679371e-08, "loss": 1.2169, "step": 11588 }, { "epoch": 0.9, "learning_rate": 5.252411890405039e-08, "loss": 1.2345, "step": 11589 }, { "epoch": 0.9, "learning_rate": 5.24437481978014e-08, "loss": 1.1639, "step": 11590 }, { "epoch": 0.9, "learning_rate": 5.236343737312643e-08, "loss": 0.8464, "step": 11591 }, { "epoch": 0.9, "learning_rate": 5.228318643510021e-08, "loss": 1.2125, "step": 11592 }, { "epoch": 0.9, "learning_rate": 5.220299538879447e-08, "loss": 1.2514, "step": 11593 }, { "epoch": 0.9, "learning_rate": 5.212286423927703e-08, "loss": 1.1688, "step": 11594 }, { "epoch": 0.9, "learning_rate": 5.204279299161174e-08, "loss": 1.2399, "step": 11595 }, { "epoch": 0.9, "learning_rate": 5.1962781650858654e-08, "loss": 1.2143, "step": 11596 }, { "epoch": 0.9, "learning_rate": 5.188283022207462e-08, "loss": 0.8366, "step": 11597 }, { "epoch": 0.9, "learning_rate": 5.18029387103115e-08, "loss": 1.2115, "step": 11598 }, { "epoch": 0.9, "learning_rate": 5.172310712061856e-08, "loss": 1.1692, "step": 11599 }, { "epoch": 0.9, "learning_rate": 5.164333545804078e-08, "loss": 1.2288, "step": 11600 }, { "epoch": 0.9, "learning_rate": 5.156362372761958e-08, "loss": 1.2168, "step": 11601 }, { "epoch": 0.9, "learning_rate": 5.148397193439202e-08, "loss": 0.7836, "step": 11602 }, { "epoch": 0.9, "learning_rate": 5.140438008339187e-08, "loss": 1.209, "step": 11603 }, { "epoch": 0.9, "learning_rate": 5.1324848179649325e-08, "loss": 1.2013, "step": 11604 }, { "epoch": 0.9, "learning_rate": 5.124537622819003e-08, "loss": 0.8448, "step": 11605 }, { "epoch": 0.9, "learning_rate": 5.1165964234036406e-08, "loss": 1.0791, "step": 11606 }, { "epoch": 0.9, "learning_rate": 5.108661220220723e-08, "loss": 1.2499, "step": 11607 }, { "epoch": 0.9, "learning_rate": 5.10073201377168e-08, "loss": 1.1989, "step": 11608 }, { "epoch": 0.9, "learning_rate": 5.0928088045576136e-08, "loss": 1.1759, "step": 11609 }, { "epoch": 0.9, "learning_rate": 5.084891593079266e-08, "loss": 0.8043, "step": 11610 }, { "epoch": 0.9, "learning_rate": 5.0769803798369704e-08, "loss": 1.1916, "step": 11611 }, { "epoch": 0.9, "learning_rate": 5.0690751653306385e-08, "loss": 1.1891, "step": 11612 }, { "epoch": 0.9, "learning_rate": 5.0611759500598705e-08, "loss": 1.2019, "step": 11613 }, { "epoch": 0.9, "learning_rate": 5.053282734523889e-08, "loss": 1.184, "step": 11614 }, { "epoch": 0.9, "learning_rate": 5.045395519221485e-08, "loss": 1.178, "step": 11615 }, { "epoch": 0.9, "learning_rate": 5.0375143046510806e-08, "loss": 1.1541, "step": 11616 }, { "epoch": 0.9, "learning_rate": 5.0296390913107666e-08, "loss": 1.2115, "step": 11617 }, { "epoch": 0.9, "learning_rate": 5.021769879698223e-08, "loss": 1.1833, "step": 11618 }, { "epoch": 0.9, "learning_rate": 5.0139066703107304e-08, "loss": 1.1949, "step": 11619 }, { "epoch": 0.9, "learning_rate": 5.006049463645223e-08, "loss": 1.2527, "step": 11620 }, { "epoch": 0.9, "learning_rate": 4.998198260198261e-08, "loss": 1.1566, "step": 11621 }, { "epoch": 0.9, "learning_rate": 4.990353060465957e-08, "loss": 1.1529, "step": 11622 }, { "epoch": 0.9, "learning_rate": 4.982513864944138e-08, "loss": 1.1234, "step": 11623 }, { "epoch": 0.9, "learning_rate": 4.9746806741282174e-08, "loss": 0.8278, "step": 11624 }, { "epoch": 0.9, "learning_rate": 4.966853488513167e-08, "loss": 1.2125, "step": 11625 }, { "epoch": 0.9, "learning_rate": 4.959032308593669e-08, "loss": 1.2233, "step": 11626 }, { "epoch": 0.9, "learning_rate": 4.9512171348639944e-08, "loss": 1.2187, "step": 11627 }, { "epoch": 0.9, "learning_rate": 4.943407967818014e-08, "loss": 1.1688, "step": 11628 }, { "epoch": 0.9, "learning_rate": 4.935604807949245e-08, "loss": 1.1368, "step": 11629 }, { "epoch": 0.9, "learning_rate": 4.9278076557508044e-08, "loss": 0.8409, "step": 11630 }, { "epoch": 0.9, "learning_rate": 4.9200165117154523e-08, "loss": 1.1727, "step": 11631 }, { "epoch": 0.9, "learning_rate": 4.9122313763355296e-08, "loss": 1.182, "step": 11632 }, { "epoch": 0.9, "learning_rate": 4.904452250103053e-08, "loss": 1.2418, "step": 11633 }, { "epoch": 0.9, "learning_rate": 4.896679133509607e-08, "loss": 1.186, "step": 11634 }, { "epoch": 0.9, "learning_rate": 4.888912027046454e-08, "loss": 1.175, "step": 11635 }, { "epoch": 0.9, "learning_rate": 4.881150931204403e-08, "loss": 1.1702, "step": 11636 }, { "epoch": 0.9, "learning_rate": 4.8733958464739375e-08, "loss": 1.2047, "step": 11637 }, { "epoch": 0.9, "learning_rate": 4.865646773345178e-08, "loss": 1.2477, "step": 11638 }, { "epoch": 0.9, "learning_rate": 4.8579037123077874e-08, "loss": 1.2062, "step": 11639 }, { "epoch": 0.9, "learning_rate": 4.8501666638510964e-08, "loss": 1.2159, "step": 11640 }, { "epoch": 0.9, "learning_rate": 4.842435628464092e-08, "loss": 1.2319, "step": 11641 }, { "epoch": 0.9, "learning_rate": 4.834710606635317e-08, "loss": 1.2267, "step": 11642 }, { "epoch": 0.9, "learning_rate": 4.8269915988529474e-08, "loss": 1.1743, "step": 11643 }, { "epoch": 0.9, "learning_rate": 4.819278605604815e-08, "loss": 1.1585, "step": 11644 }, { "epoch": 0.9, "learning_rate": 4.8115716273783635e-08, "loss": 1.2597, "step": 11645 }, { "epoch": 0.9, "learning_rate": 4.803870664660603e-08, "loss": 1.1493, "step": 11646 }, { "epoch": 0.9, "learning_rate": 4.796175717938211e-08, "loss": 1.1779, "step": 11647 }, { "epoch": 0.9, "learning_rate": 4.78848678769751e-08, "loss": 0.8683, "step": 11648 }, { "epoch": 0.9, "learning_rate": 4.780803874424366e-08, "loss": 1.2023, "step": 11649 }, { "epoch": 0.9, "learning_rate": 4.7731269786043136e-08, "loss": 0.8931, "step": 11650 }, { "epoch": 0.9, "learning_rate": 4.765456100722531e-08, "loss": 1.1621, "step": 11651 }, { "epoch": 0.9, "learning_rate": 4.7577912412637536e-08, "loss": 1.2127, "step": 11652 }, { "epoch": 0.9, "learning_rate": 4.750132400712359e-08, "loss": 1.1579, "step": 11653 }, { "epoch": 0.9, "learning_rate": 4.742479579552394e-08, "loss": 0.8625, "step": 11654 }, { "epoch": 0.9, "learning_rate": 4.7348327782674725e-08, "loss": 1.1898, "step": 11655 }, { "epoch": 0.9, "learning_rate": 4.7271919973408184e-08, "loss": 1.1997, "step": 11656 }, { "epoch": 0.9, "learning_rate": 4.719557237255312e-08, "loss": 1.2327, "step": 11657 }, { "epoch": 0.9, "learning_rate": 4.711928498493456e-08, "loss": 1.2015, "step": 11658 }, { "epoch": 0.9, "learning_rate": 4.704305781537299e-08, "loss": 1.2055, "step": 11659 }, { "epoch": 0.9, "learning_rate": 4.696689086868622e-08, "loss": 1.1533, "step": 11660 }, { "epoch": 0.91, "learning_rate": 4.68907841496875e-08, "loss": 1.1783, "step": 11661 }, { "epoch": 0.91, "learning_rate": 4.681473766318633e-08, "loss": 1.1591, "step": 11662 }, { "epoch": 0.91, "learning_rate": 4.673875141398875e-08, "loss": 1.1811, "step": 11663 }, { "epoch": 0.91, "learning_rate": 4.6662825406896476e-08, "loss": 1.1726, "step": 11664 }, { "epoch": 0.91, "learning_rate": 4.658695964670822e-08, "loss": 1.2121, "step": 11665 }, { "epoch": 0.91, "learning_rate": 4.65111541382176e-08, "loss": 1.1632, "step": 11666 }, { "epoch": 0.91, "learning_rate": 4.643540888621589e-08, "loss": 1.1939, "step": 11667 }, { "epoch": 0.91, "learning_rate": 4.63597238954897e-08, "loss": 1.2163, "step": 11668 }, { "epoch": 0.91, "learning_rate": 4.628409917082166e-08, "loss": 1.1782, "step": 11669 }, { "epoch": 0.91, "learning_rate": 4.620853471699138e-08, "loss": 1.1745, "step": 11670 }, { "epoch": 0.91, "learning_rate": 4.613303053877393e-08, "loss": 1.1978, "step": 11671 }, { "epoch": 0.91, "learning_rate": 4.605758664094117e-08, "loss": 1.1797, "step": 11672 }, { "epoch": 0.91, "learning_rate": 4.5982203028260615e-08, "loss": 1.1954, "step": 11673 }, { "epoch": 0.91, "learning_rate": 4.5906879705496005e-08, "loss": 1.1875, "step": 11674 }, { "epoch": 0.91, "learning_rate": 4.583161667740787e-08, "loss": 1.1867, "step": 11675 }, { "epoch": 0.91, "learning_rate": 4.575641394875229e-08, "loss": 0.7988, "step": 11676 }, { "epoch": 0.91, "learning_rate": 4.56812715242817e-08, "loss": 1.2251, "step": 11677 }, { "epoch": 0.91, "learning_rate": 4.560618940874483e-08, "loss": 1.1504, "step": 11678 }, { "epoch": 0.91, "learning_rate": 4.55311676068868e-08, "loss": 1.146, "step": 11679 }, { "epoch": 0.91, "learning_rate": 4.545620612344836e-08, "loss": 1.1707, "step": 11680 }, { "epoch": 0.91, "learning_rate": 4.538130496316672e-08, "loss": 1.174, "step": 11681 }, { "epoch": 0.91, "learning_rate": 4.530646413077577e-08, "loss": 1.176, "step": 11682 }, { "epoch": 0.91, "learning_rate": 4.5231683631004605e-08, "loss": 1.1091, "step": 11683 }, { "epoch": 0.91, "learning_rate": 4.5156963468579225e-08, "loss": 1.1899, "step": 11684 }, { "epoch": 0.91, "learning_rate": 4.5082303648221967e-08, "loss": 1.1962, "step": 11685 }, { "epoch": 0.91, "learning_rate": 4.50077041746505e-08, "loss": 1.2225, "step": 11686 }, { "epoch": 0.91, "learning_rate": 4.493316505257927e-08, "loss": 1.2058, "step": 11687 }, { "epoch": 0.91, "learning_rate": 4.485868628671918e-08, "loss": 1.2105, "step": 11688 }, { "epoch": 0.91, "learning_rate": 4.478426788177692e-08, "loss": 1.1992, "step": 11689 }, { "epoch": 0.91, "learning_rate": 4.470990984245504e-08, "loss": 1.1955, "step": 11690 }, { "epoch": 0.91, "learning_rate": 4.463561217345302e-08, "loss": 1.1937, "step": 11691 }, { "epoch": 0.91, "learning_rate": 4.4561374879466095e-08, "loss": 1.2085, "step": 11692 }, { "epoch": 0.91, "learning_rate": 4.448719796518552e-08, "loss": 1.2065, "step": 11693 }, { "epoch": 0.91, "learning_rate": 4.44130814352992e-08, "loss": 1.2276, "step": 11694 }, { "epoch": 0.91, "learning_rate": 4.4339025294490965e-08, "loss": 1.1698, "step": 11695 }, { "epoch": 0.91, "learning_rate": 4.426502954744083e-08, "loss": 1.1911, "step": 11696 }, { "epoch": 0.91, "learning_rate": 4.4191094198825055e-08, "loss": 0.8144, "step": 11697 }, { "epoch": 0.91, "learning_rate": 4.411721925331591e-08, "loss": 1.114, "step": 11698 }, { "epoch": 0.91, "learning_rate": 4.40434047155821e-08, "loss": 0.8343, "step": 11699 }, { "epoch": 0.91, "learning_rate": 4.396965059028823e-08, "loss": 1.2157, "step": 11700 }, { "epoch": 0.91, "learning_rate": 4.389595688209546e-08, "loss": 0.8424, "step": 11701 }, { "epoch": 0.91, "learning_rate": 4.382232359566096e-08, "loss": 1.2373, "step": 11702 }, { "epoch": 0.91, "learning_rate": 4.374875073563766e-08, "loss": 1.2277, "step": 11703 }, { "epoch": 0.91, "learning_rate": 4.367523830667541e-08, "loss": 1.1533, "step": 11704 }, { "epoch": 0.91, "learning_rate": 4.360178631341971e-08, "loss": 1.1413, "step": 11705 }, { "epoch": 0.91, "learning_rate": 4.352839476051273e-08, "loss": 1.1858, "step": 11706 }, { "epoch": 0.91, "learning_rate": 4.345506365259222e-08, "loss": 1.1522, "step": 11707 }, { "epoch": 0.91, "learning_rate": 4.338179299429223e-08, "loss": 1.1916, "step": 11708 }, { "epoch": 0.91, "learning_rate": 4.330858279024374e-08, "loss": 1.2325, "step": 11709 }, { "epoch": 0.91, "learning_rate": 4.323543304507271e-08, "loss": 1.1711, "step": 11710 }, { "epoch": 0.91, "learning_rate": 4.316234376340222e-08, "loss": 1.2304, "step": 11711 }, { "epoch": 0.91, "learning_rate": 4.308931494985124e-08, "loss": 1.1932, "step": 11712 }, { "epoch": 0.91, "learning_rate": 4.301634660903464e-08, "loss": 1.1731, "step": 11713 }, { "epoch": 0.91, "learning_rate": 4.294343874556394e-08, "loss": 0.8537, "step": 11714 }, { "epoch": 0.91, "learning_rate": 4.287059136404636e-08, "loss": 1.1768, "step": 11715 }, { "epoch": 0.91, "learning_rate": 4.279780446908599e-08, "loss": 1.1797, "step": 11716 }, { "epoch": 0.91, "learning_rate": 4.272507806528225e-08, "loss": 0.8462, "step": 11717 }, { "epoch": 0.91, "learning_rate": 4.265241215723114e-08, "loss": 1.1815, "step": 11718 }, { "epoch": 0.91, "learning_rate": 4.2579806749525213e-08, "loss": 1.2278, "step": 11719 }, { "epoch": 0.91, "learning_rate": 4.250726184675257e-08, "loss": 0.8114, "step": 11720 }, { "epoch": 0.91, "learning_rate": 4.243477745349755e-08, "loss": 1.1846, "step": 11721 }, { "epoch": 0.91, "learning_rate": 4.236235357434126e-08, "loss": 1.1991, "step": 11722 }, { "epoch": 0.91, "learning_rate": 4.228999021386048e-08, "loss": 1.2135, "step": 11723 }, { "epoch": 0.91, "learning_rate": 4.22176873766279e-08, "loss": 1.207, "step": 11724 }, { "epoch": 0.91, "learning_rate": 4.214544506721318e-08, "loss": 1.2322, "step": 11725 }, { "epoch": 0.91, "learning_rate": 4.207326329018179e-08, "loss": 1.2035, "step": 11726 }, { "epoch": 0.91, "learning_rate": 4.2001142050094864e-08, "loss": 1.1579, "step": 11727 }, { "epoch": 0.91, "learning_rate": 4.192908135151052e-08, "loss": 1.1609, "step": 11728 }, { "epoch": 0.91, "learning_rate": 4.185708119898268e-08, "loss": 1.1962, "step": 11729 }, { "epoch": 0.91, "learning_rate": 4.178514159706126e-08, "loss": 0.8262, "step": 11730 }, { "epoch": 0.91, "learning_rate": 4.1713262550292505e-08, "loss": 1.2033, "step": 11731 }, { "epoch": 0.91, "learning_rate": 4.1641444063219235e-08, "loss": 1.1706, "step": 11732 }, { "epoch": 0.91, "learning_rate": 4.156968614037981e-08, "loss": 0.8399, "step": 11733 }, { "epoch": 0.91, "learning_rate": 4.1497988786308945e-08, "loss": 1.2606, "step": 11734 }, { "epoch": 0.91, "learning_rate": 4.142635200553779e-08, "loss": 0.8046, "step": 11735 }, { "epoch": 0.91, "learning_rate": 4.135477580259361e-08, "loss": 1.1026, "step": 11736 }, { "epoch": 0.91, "learning_rate": 4.128326018199923e-08, "loss": 1.2034, "step": 11737 }, { "epoch": 0.91, "learning_rate": 4.1211805148274606e-08, "loss": 1.1731, "step": 11738 }, { "epoch": 0.91, "learning_rate": 4.11404107059351e-08, "loss": 1.2122, "step": 11739 }, { "epoch": 0.91, "learning_rate": 4.10690768594929e-08, "loss": 1.1824, "step": 11740 }, { "epoch": 0.91, "learning_rate": 4.0997803613455725e-08, "loss": 1.1609, "step": 11741 }, { "epoch": 0.91, "learning_rate": 4.0926590972327644e-08, "loss": 1.2543, "step": 11742 }, { "epoch": 0.91, "learning_rate": 4.085543894060939e-08, "loss": 1.1904, "step": 11743 }, { "epoch": 0.91, "learning_rate": 4.078434752279725e-08, "loss": 1.2579, "step": 11744 }, { "epoch": 0.91, "learning_rate": 4.0713316723383736e-08, "loss": 1.1904, "step": 11745 }, { "epoch": 0.91, "learning_rate": 4.064234654685794e-08, "loss": 1.1841, "step": 11746 }, { "epoch": 0.91, "learning_rate": 4.057143699770471e-08, "loss": 1.2072, "step": 11747 }, { "epoch": 0.91, "learning_rate": 4.050058808040535e-08, "loss": 1.2052, "step": 11748 }, { "epoch": 0.91, "learning_rate": 4.042979979943706e-08, "loss": 1.1809, "step": 11749 }, { "epoch": 0.91, "learning_rate": 4.03590721592737e-08, "loss": 0.8604, "step": 11750 }, { "epoch": 0.91, "learning_rate": 4.02884051643847e-08, "loss": 1.1509, "step": 11751 }, { "epoch": 0.91, "learning_rate": 4.0217798819235707e-08, "loss": 1.2059, "step": 11752 }, { "epoch": 0.91, "learning_rate": 4.0147253128289374e-08, "loss": 1.2056, "step": 11753 }, { "epoch": 0.91, "learning_rate": 4.0076768096003245e-08, "loss": 1.2122, "step": 11754 }, { "epoch": 0.91, "learning_rate": 4.000634372683198e-08, "loss": 1.1427, "step": 11755 }, { "epoch": 0.91, "learning_rate": 3.9935980025226244e-08, "loss": 1.2055, "step": 11756 }, { "epoch": 0.91, "learning_rate": 3.986567699563237e-08, "loss": 1.216, "step": 11757 }, { "epoch": 0.91, "learning_rate": 3.9795434642493356e-08, "loss": 1.2092, "step": 11758 }, { "epoch": 0.91, "learning_rate": 3.972525297024831e-08, "loss": 1.1457, "step": 11759 }, { "epoch": 0.91, "learning_rate": 3.965513198333259e-08, "loss": 1.2353, "step": 11760 }, { "epoch": 0.91, "learning_rate": 3.9585071686176976e-08, "loss": 1.1551, "step": 11761 }, { "epoch": 0.91, "learning_rate": 3.9515072083209476e-08, "loss": 1.1911, "step": 11762 }, { "epoch": 0.91, "learning_rate": 3.9445133178853786e-08, "loss": 0.8357, "step": 11763 }, { "epoch": 0.91, "learning_rate": 3.9375254977529584e-08, "loss": 1.2023, "step": 11764 }, { "epoch": 0.91, "learning_rate": 3.9305437483652673e-08, "loss": 0.831, "step": 11765 }, { "epoch": 0.91, "learning_rate": 3.923568070163552e-08, "loss": 1.22, "step": 11766 }, { "epoch": 0.91, "learning_rate": 3.9165984635886496e-08, "loss": 1.1262, "step": 11767 }, { "epoch": 0.91, "learning_rate": 3.909634929080985e-08, "loss": 1.1731, "step": 11768 }, { "epoch": 0.91, "learning_rate": 3.902677467080651e-08, "loss": 1.1718, "step": 11769 }, { "epoch": 0.91, "learning_rate": 3.895726078027317e-08, "loss": 1.1882, "step": 11770 }, { "epoch": 0.91, "learning_rate": 3.888780762360266e-08, "loss": 1.1972, "step": 11771 }, { "epoch": 0.91, "learning_rate": 3.881841520518436e-08, "loss": 1.2116, "step": 11772 }, { "epoch": 0.91, "learning_rate": 3.874908352940365e-08, "loss": 0.84, "step": 11773 }, { "epoch": 0.91, "learning_rate": 3.867981260064168e-08, "loss": 0.8404, "step": 11774 }, { "epoch": 0.91, "learning_rate": 3.861060242327629e-08, "loss": 1.2347, "step": 11775 }, { "epoch": 0.91, "learning_rate": 3.85414530016811e-08, "loss": 1.1373, "step": 11776 }, { "epoch": 0.91, "learning_rate": 3.847236434022649e-08, "loss": 1.2435, "step": 11777 }, { "epoch": 0.91, "learning_rate": 3.84033364432782e-08, "loss": 1.1754, "step": 11778 }, { "epoch": 0.91, "learning_rate": 3.833436931519851e-08, "loss": 1.2184, "step": 11779 }, { "epoch": 0.91, "learning_rate": 3.826546296034605e-08, "loss": 1.2168, "step": 11780 }, { "epoch": 0.91, "learning_rate": 3.8196617383075225e-08, "loss": 0.8143, "step": 11781 }, { "epoch": 0.91, "learning_rate": 3.8127832587736775e-08, "loss": 1.2136, "step": 11782 }, { "epoch": 0.91, "learning_rate": 3.805910857867767e-08, "loss": 1.2228, "step": 11783 }, { "epoch": 0.91, "learning_rate": 3.799044536024121e-08, "loss": 1.168, "step": 11784 }, { "epoch": 0.91, "learning_rate": 3.792184293676626e-08, "loss": 1.2506, "step": 11785 }, { "epoch": 0.91, "learning_rate": 3.785330131258824e-08, "loss": 1.1627, "step": 11786 }, { "epoch": 0.91, "learning_rate": 3.778482049203913e-08, "loss": 1.2075, "step": 11787 }, { "epoch": 0.91, "learning_rate": 3.771640047944613e-08, "loss": 1.1765, "step": 11788 }, { "epoch": 0.91, "learning_rate": 3.764804127913313e-08, "loss": 1.2295, "step": 11789 }, { "epoch": 0.92, "learning_rate": 3.757974289542054e-08, "loss": 1.1933, "step": 11790 }, { "epoch": 0.92, "learning_rate": 3.751150533262415e-08, "loss": 1.2066, "step": 11791 }, { "epoch": 0.92, "learning_rate": 3.744332859505628e-08, "loss": 1.212, "step": 11792 }, { "epoch": 0.92, "learning_rate": 3.73752126870257e-08, "loss": 1.1448, "step": 11793 }, { "epoch": 0.92, "learning_rate": 3.730715761283709e-08, "loss": 1.2356, "step": 11794 }, { "epoch": 0.92, "learning_rate": 3.723916337679067e-08, "loss": 1.2152, "step": 11795 }, { "epoch": 0.92, "learning_rate": 3.7171229983184003e-08, "loss": 1.1484, "step": 11796 }, { "epoch": 0.92, "learning_rate": 3.71033574363101e-08, "loss": 1.2293, "step": 11797 }, { "epoch": 0.92, "learning_rate": 3.7035545740457864e-08, "loss": 1.2219, "step": 11798 }, { "epoch": 0.92, "learning_rate": 3.696779489991286e-08, "loss": 0.7879, "step": 11799 }, { "epoch": 0.92, "learning_rate": 3.690010491895701e-08, "loss": 1.1795, "step": 11800 }, { "epoch": 0.92, "learning_rate": 3.683247580186766e-08, "loss": 0.8284, "step": 11801 }, { "epoch": 0.92, "learning_rate": 3.67649075529185e-08, "loss": 1.1503, "step": 11802 }, { "epoch": 0.92, "learning_rate": 3.669740017638012e-08, "loss": 0.7953, "step": 11803 }, { "epoch": 0.92, "learning_rate": 3.6629953676518445e-08, "loss": 1.1963, "step": 11804 }, { "epoch": 0.92, "learning_rate": 3.656256805759561e-08, "loss": 0.8644, "step": 11805 }, { "epoch": 0.92, "learning_rate": 3.649524332387033e-08, "loss": 1.1874, "step": 11806 }, { "epoch": 0.92, "learning_rate": 3.642797947959742e-08, "loss": 1.1591, "step": 11807 }, { "epoch": 0.92, "learning_rate": 3.636077652902714e-08, "loss": 1.2272, "step": 11808 }, { "epoch": 0.92, "learning_rate": 3.6293634476406875e-08, "loss": 1.2053, "step": 11809 }, { "epoch": 0.92, "learning_rate": 3.622655332597957e-08, "loss": 1.2526, "step": 11810 }, { "epoch": 0.92, "learning_rate": 3.615953308198449e-08, "loss": 1.1963, "step": 11811 }, { "epoch": 0.92, "learning_rate": 3.609257374865704e-08, "loss": 1.1693, "step": 11812 }, { "epoch": 0.92, "learning_rate": 3.6025675330228824e-08, "loss": 1.1929, "step": 11813 }, { "epoch": 0.92, "learning_rate": 3.5958837830927346e-08, "loss": 1.1842, "step": 11814 }, { "epoch": 0.92, "learning_rate": 3.589206125497668e-08, "loss": 1.1909, "step": 11815 }, { "epoch": 0.92, "learning_rate": 3.582534560659678e-08, "loss": 1.1563, "step": 11816 }, { "epoch": 0.92, "learning_rate": 3.575869089000383e-08, "loss": 1.1478, "step": 11817 }, { "epoch": 0.92, "learning_rate": 3.569209710940968e-08, "loss": 1.1488, "step": 11818 }, { "epoch": 0.92, "learning_rate": 3.562556426902341e-08, "loss": 1.2258, "step": 11819 }, { "epoch": 0.92, "learning_rate": 3.5559092373049216e-08, "loss": 1.2213, "step": 11820 }, { "epoch": 0.92, "learning_rate": 3.549268142568817e-08, "loss": 1.1258, "step": 11821 }, { "epoch": 0.92, "learning_rate": 3.5426331431136824e-08, "loss": 1.1175, "step": 11822 }, { "epoch": 0.92, "learning_rate": 3.536004239358825e-08, "loss": 1.1874, "step": 11823 }, { "epoch": 0.92, "learning_rate": 3.5293814317231996e-08, "loss": 1.1812, "step": 11824 }, { "epoch": 0.92, "learning_rate": 3.5227647206253045e-08, "loss": 1.1829, "step": 11825 }, { "epoch": 0.92, "learning_rate": 3.516154106483293e-08, "loss": 0.8362, "step": 11826 }, { "epoch": 0.92, "learning_rate": 3.509549589714944e-08, "loss": 1.1938, "step": 11827 }, { "epoch": 0.92, "learning_rate": 3.5029511707376335e-08, "loss": 1.1717, "step": 11828 }, { "epoch": 0.92, "learning_rate": 3.496358849968328e-08, "loss": 1.156, "step": 11829 }, { "epoch": 0.92, "learning_rate": 3.4897726278236504e-08, "loss": 1.1566, "step": 11830 }, { "epoch": 0.92, "learning_rate": 3.483192504719845e-08, "loss": 1.1874, "step": 11831 }, { "epoch": 0.92, "learning_rate": 3.476618481072713e-08, "loss": 1.1715, "step": 11832 }, { "epoch": 0.92, "learning_rate": 3.470050557297699e-08, "loss": 1.2711, "step": 11833 }, { "epoch": 0.92, "learning_rate": 3.4634887338099275e-08, "loss": 1.1759, "step": 11834 }, { "epoch": 0.92, "learning_rate": 3.45693301102401e-08, "loss": 1.2431, "step": 11835 }, { "epoch": 0.92, "learning_rate": 3.4503833893542725e-08, "loss": 1.2062, "step": 11836 }, { "epoch": 0.92, "learning_rate": 3.4438398692146264e-08, "loss": 1.2401, "step": 11837 }, { "epoch": 0.92, "learning_rate": 3.437302451018598e-08, "loss": 1.1523, "step": 11838 }, { "epoch": 0.92, "learning_rate": 3.4307711351793e-08, "loss": 1.21, "step": 11839 }, { "epoch": 0.92, "learning_rate": 3.4242459221095035e-08, "loss": 1.1491, "step": 11840 }, { "epoch": 0.92, "learning_rate": 3.417726812221577e-08, "loss": 0.8725, "step": 11841 }, { "epoch": 0.92, "learning_rate": 3.411213805927482e-08, "loss": 1.2331, "step": 11842 }, { "epoch": 0.92, "learning_rate": 3.404706903638832e-08, "loss": 1.1954, "step": 11843 }, { "epoch": 0.92, "learning_rate": 3.3982061057668434e-08, "loss": 1.1938, "step": 11844 }, { "epoch": 0.92, "learning_rate": 3.391711412722298e-08, "loss": 1.1672, "step": 11845 }, { "epoch": 0.92, "learning_rate": 3.385222824915679e-08, "loss": 1.1946, "step": 11846 }, { "epoch": 0.92, "learning_rate": 3.378740342757014e-08, "loss": 1.0939, "step": 11847 }, { "epoch": 0.92, "learning_rate": 3.372263966655986e-08, "loss": 1.1894, "step": 11848 }, { "epoch": 0.92, "learning_rate": 3.365793697021846e-08, "loss": 1.178, "step": 11849 }, { "epoch": 0.92, "learning_rate": 3.359329534263511e-08, "loss": 1.2125, "step": 11850 }, { "epoch": 0.92, "learning_rate": 3.352871478789488e-08, "loss": 1.1723, "step": 11851 }, { "epoch": 0.92, "learning_rate": 3.3464195310078824e-08, "loss": 0.7904, "step": 11852 }, { "epoch": 0.92, "learning_rate": 3.3399736913264475e-08, "loss": 1.1682, "step": 11853 }, { "epoch": 0.92, "learning_rate": 3.3335339601525125e-08, "loss": 0.8723, "step": 11854 }, { "epoch": 0.92, "learning_rate": 3.327100337893074e-08, "loss": 1.151, "step": 11855 }, { "epoch": 0.92, "learning_rate": 3.320672824954696e-08, "loss": 1.2424, "step": 11856 }, { "epoch": 0.92, "learning_rate": 3.3142514217435414e-08, "loss": 0.7872, "step": 11857 }, { "epoch": 0.92, "learning_rate": 3.307836128665464e-08, "loss": 1.2027, "step": 11858 }, { "epoch": 0.92, "learning_rate": 3.3014269461258624e-08, "loss": 1.186, "step": 11859 }, { "epoch": 0.92, "learning_rate": 3.295023874529745e-08, "loss": 1.2424, "step": 11860 }, { "epoch": 0.92, "learning_rate": 3.2886269142818004e-08, "loss": 1.2689, "step": 11861 }, { "epoch": 0.92, "learning_rate": 3.28223606578627e-08, "loss": 1.2576, "step": 11862 }, { "epoch": 0.92, "learning_rate": 3.275851329447021e-08, "loss": 1.1802, "step": 11863 }, { "epoch": 0.92, "learning_rate": 3.269472705667542e-08, "loss": 1.2342, "step": 11864 }, { "epoch": 0.92, "learning_rate": 3.263100194850965e-08, "loss": 1.159, "step": 11865 }, { "epoch": 0.92, "learning_rate": 3.2567337973999687e-08, "loss": 1.1893, "step": 11866 }, { "epoch": 0.92, "learning_rate": 3.2503735137168975e-08, "loss": 0.8296, "step": 11867 }, { "epoch": 0.92, "learning_rate": 3.244019344203708e-08, "loss": 1.1748, "step": 11868 }, { "epoch": 0.92, "learning_rate": 3.2376712892619453e-08, "loss": 1.2254, "step": 11869 }, { "epoch": 0.92, "learning_rate": 3.231329349292755e-08, "loss": 1.1566, "step": 11870 }, { "epoch": 0.92, "learning_rate": 3.224993524696962e-08, "loss": 1.1886, "step": 11871 }, { "epoch": 0.92, "learning_rate": 3.218663815874956e-08, "loss": 1.208, "step": 11872 }, { "epoch": 0.92, "learning_rate": 3.212340223226717e-08, "loss": 1.2031, "step": 11873 }, { "epoch": 0.92, "learning_rate": 3.206022747151904e-08, "loss": 1.2289, "step": 11874 }, { "epoch": 0.92, "learning_rate": 3.199711388049753e-08, "loss": 1.2002, "step": 11875 }, { "epoch": 0.92, "learning_rate": 3.1934061463190884e-08, "loss": 1.2452, "step": 11876 }, { "epoch": 0.92, "learning_rate": 3.187107022358393e-08, "loss": 0.8176, "step": 11877 }, { "epoch": 0.92, "learning_rate": 3.180814016565758e-08, "loss": 1.196, "step": 11878 }, { "epoch": 0.92, "learning_rate": 3.1745271293388444e-08, "loss": 1.1945, "step": 11879 }, { "epoch": 0.92, "learning_rate": 3.1682463610749775e-08, "loss": 1.2025, "step": 11880 }, { "epoch": 0.92, "learning_rate": 3.1619717121710634e-08, "loss": 1.2164, "step": 11881 }, { "epoch": 0.92, "learning_rate": 3.155703183023661e-08, "loss": 1.2298, "step": 11882 }, { "epoch": 0.92, "learning_rate": 3.149440774028866e-08, "loss": 1.184, "step": 11883 }, { "epoch": 0.92, "learning_rate": 3.143184485582484e-08, "loss": 1.2509, "step": 11884 }, { "epoch": 0.92, "learning_rate": 3.136934318079876e-08, "loss": 0.8411, "step": 11885 }, { "epoch": 0.92, "learning_rate": 3.130690271915992e-08, "loss": 1.1833, "step": 11886 }, { "epoch": 0.92, "learning_rate": 3.124452347485473e-08, "loss": 1.1855, "step": 11887 }, { "epoch": 0.92, "learning_rate": 3.118220545182515e-08, "loss": 1.2732, "step": 11888 }, { "epoch": 0.92, "learning_rate": 3.111994865400924e-08, "loss": 1.1979, "step": 11889 }, { "epoch": 0.92, "learning_rate": 3.105775308534153e-08, "loss": 1.1661, "step": 11890 }, { "epoch": 0.92, "learning_rate": 3.0995618749752536e-08, "loss": 1.1859, "step": 11891 }, { "epoch": 0.92, "learning_rate": 3.09335456511689e-08, "loss": 0.8565, "step": 11892 }, { "epoch": 0.92, "learning_rate": 3.087153379351337e-08, "loss": 1.1835, "step": 11893 }, { "epoch": 0.92, "learning_rate": 3.080958318070459e-08, "loss": 1.1703, "step": 11894 }, { "epoch": 0.92, "learning_rate": 3.074769381665809e-08, "loss": 1.2213, "step": 11895 }, { "epoch": 0.92, "learning_rate": 3.068586570528453e-08, "loss": 1.2009, "step": 11896 }, { "epoch": 0.92, "learning_rate": 3.0624098850491444e-08, "loss": 0.7964, "step": 11897 }, { "epoch": 0.92, "learning_rate": 3.056239325618193e-08, "loss": 1.216, "step": 11898 }, { "epoch": 0.92, "learning_rate": 3.050074892625609e-08, "loss": 1.2193, "step": 11899 }, { "epoch": 0.92, "learning_rate": 3.0439165864609154e-08, "loss": 1.2086, "step": 11900 }, { "epoch": 0.92, "learning_rate": 3.037764407513277e-08, "loss": 1.1584, "step": 11901 }, { "epoch": 0.92, "learning_rate": 3.0316183561715504e-08, "loss": 1.1628, "step": 11902 }, { "epoch": 0.92, "learning_rate": 3.0254784328240804e-08, "loss": 1.2095, "step": 11903 }, { "epoch": 0.92, "learning_rate": 3.0193446378588894e-08, "loss": 0.833, "step": 11904 }, { "epoch": 0.92, "learning_rate": 3.013216971663646e-08, "loss": 1.2248, "step": 11905 }, { "epoch": 0.92, "learning_rate": 3.0070954346255505e-08, "loss": 1.2407, "step": 11906 }, { "epoch": 0.92, "learning_rate": 3.000980027131472e-08, "loss": 0.8538, "step": 11907 }, { "epoch": 0.92, "learning_rate": 2.994870749567902e-08, "loss": 1.2295, "step": 11908 }, { "epoch": 0.92, "learning_rate": 2.988767602320907e-08, "loss": 1.2194, "step": 11909 }, { "epoch": 0.92, "learning_rate": 2.982670585776148e-08, "loss": 1.2146, "step": 11910 }, { "epoch": 0.92, "learning_rate": 2.976579700318971e-08, "loss": 1.1272, "step": 11911 }, { "epoch": 0.92, "learning_rate": 2.970494946334301e-08, "loss": 1.1403, "step": 11912 }, { "epoch": 0.92, "learning_rate": 2.96441632420662e-08, "loss": 1.1625, "step": 11913 }, { "epoch": 0.92, "learning_rate": 2.9583438343201205e-08, "loss": 1.1647, "step": 11914 }, { "epoch": 0.92, "learning_rate": 2.952277477058529e-08, "loss": 1.1944, "step": 11915 }, { "epoch": 0.92, "learning_rate": 2.9462172528052388e-08, "loss": 1.1706, "step": 11916 }, { "epoch": 0.92, "learning_rate": 2.9401631619431878e-08, "loss": 1.2186, "step": 11917 }, { "epoch": 0.92, "learning_rate": 2.9341152048550145e-08, "loss": 1.2275, "step": 11918 }, { "epoch": 0.93, "learning_rate": 2.928073381922913e-08, "loss": 1.2025, "step": 11919 }, { "epoch": 0.93, "learning_rate": 2.9220376935286783e-08, "loss": 1.237, "step": 11920 }, { "epoch": 0.93, "learning_rate": 2.9160081400537718e-08, "loss": 1.2008, "step": 11921 }, { "epoch": 0.93, "learning_rate": 2.9099847218792217e-08, "loss": 1.181, "step": 11922 }, { "epoch": 0.93, "learning_rate": 2.9039674393856685e-08, "loss": 1.167, "step": 11923 }, { "epoch": 0.93, "learning_rate": 2.8979562929533964e-08, "loss": 1.1932, "step": 11924 }, { "epoch": 0.93, "learning_rate": 2.891951282962268e-08, "loss": 1.2152, "step": 11925 }, { "epoch": 0.93, "learning_rate": 2.8859524097918142e-08, "loss": 1.1857, "step": 11926 }, { "epoch": 0.93, "learning_rate": 2.8799596738211084e-08, "loss": 1.2515, "step": 11927 }, { "epoch": 0.93, "learning_rate": 2.8739730754288595e-08, "loss": 1.22, "step": 11928 }, { "epoch": 0.93, "learning_rate": 2.8679926149934085e-08, "loss": 0.8288, "step": 11929 }, { "epoch": 0.93, "learning_rate": 2.8620182928926984e-08, "loss": 1.1705, "step": 11930 }, { "epoch": 0.93, "learning_rate": 2.856050109504271e-08, "loss": 0.8366, "step": 11931 }, { "epoch": 0.93, "learning_rate": 2.8500880652052917e-08, "loss": 1.1528, "step": 11932 }, { "epoch": 0.93, "learning_rate": 2.8441321603725476e-08, "loss": 1.1421, "step": 11933 }, { "epoch": 0.93, "learning_rate": 2.8381823953824157e-08, "loss": 1.159, "step": 11934 }, { "epoch": 0.93, "learning_rate": 2.832238770610884e-08, "loss": 1.1943, "step": 11935 }, { "epoch": 0.93, "learning_rate": 2.8263012864336077e-08, "loss": 1.1782, "step": 11936 }, { "epoch": 0.93, "learning_rate": 2.8203699432257753e-08, "loss": 1.1701, "step": 11937 }, { "epoch": 0.93, "learning_rate": 2.8144447413622097e-08, "loss": 1.1746, "step": 11938 }, { "epoch": 0.93, "learning_rate": 2.808525681217411e-08, "loss": 0.8419, "step": 11939 }, { "epoch": 0.93, "learning_rate": 2.8026127631653906e-08, "loss": 1.1925, "step": 11940 }, { "epoch": 0.93, "learning_rate": 2.7967059875798392e-08, "loss": 1.1764, "step": 11941 }, { "epoch": 0.93, "learning_rate": 2.790805354834036e-08, "loss": 1.2036, "step": 11942 }, { "epoch": 0.93, "learning_rate": 2.7849108653008935e-08, "loss": 1.2122, "step": 11943 }, { "epoch": 0.93, "learning_rate": 2.7790225193528806e-08, "loss": 1.1927, "step": 11944 }, { "epoch": 0.93, "learning_rate": 2.773140317362155e-08, "loss": 1.1845, "step": 11945 }, { "epoch": 0.93, "learning_rate": 2.7672642597004413e-08, "loss": 1.1881, "step": 11946 }, { "epoch": 0.93, "learning_rate": 2.761394346739043e-08, "loss": 0.7962, "step": 11947 }, { "epoch": 0.93, "learning_rate": 2.7555305788489635e-08, "loss": 1.1975, "step": 11948 }, { "epoch": 0.93, "learning_rate": 2.7496729564007503e-08, "loss": 1.1868, "step": 11949 }, { "epoch": 0.93, "learning_rate": 2.7438214797645632e-08, "loss": 1.1897, "step": 11950 }, { "epoch": 0.93, "learning_rate": 2.737976149310206e-08, "loss": 1.2463, "step": 11951 }, { "epoch": 0.93, "learning_rate": 2.7321369654070837e-08, "loss": 1.1924, "step": 11952 }, { "epoch": 0.93, "learning_rate": 2.7263039284242118e-08, "loss": 1.1674, "step": 11953 }, { "epoch": 0.93, "learning_rate": 2.7204770387301845e-08, "loss": 1.1778, "step": 11954 }, { "epoch": 0.93, "learning_rate": 2.714656296693263e-08, "loss": 1.1801, "step": 11955 }, { "epoch": 0.93, "learning_rate": 2.708841702681308e-08, "loss": 1.1572, "step": 11956 }, { "epoch": 0.93, "learning_rate": 2.7030332570617375e-08, "loss": 1.1929, "step": 11957 }, { "epoch": 0.93, "learning_rate": 2.6972309602016353e-08, "loss": 1.1988, "step": 11958 }, { "epoch": 0.93, "learning_rate": 2.6914348124676967e-08, "loss": 1.1785, "step": 11959 }, { "epoch": 0.93, "learning_rate": 2.6856448142262068e-08, "loss": 1.1712, "step": 11960 }, { "epoch": 0.93, "learning_rate": 2.6798609658430727e-08, "loss": 1.2183, "step": 11961 }, { "epoch": 0.93, "learning_rate": 2.674083267683791e-08, "loss": 1.1947, "step": 11962 }, { "epoch": 0.93, "learning_rate": 2.6683117201135142e-08, "loss": 1.2387, "step": 11963 }, { "epoch": 0.93, "learning_rate": 2.6625463234969504e-08, "loss": 1.2113, "step": 11964 }, { "epoch": 0.93, "learning_rate": 2.6567870781984746e-08, "loss": 1.218, "step": 11965 }, { "epoch": 0.93, "learning_rate": 2.6510339845820297e-08, "loss": 1.2001, "step": 11966 }, { "epoch": 0.93, "learning_rate": 2.645287043011202e-08, "loss": 1.1652, "step": 11967 }, { "epoch": 0.93, "learning_rate": 2.6395462538491565e-08, "loss": 1.2448, "step": 11968 }, { "epoch": 0.93, "learning_rate": 2.6338116174586922e-08, "loss": 1.251, "step": 11969 }, { "epoch": 0.93, "learning_rate": 2.6280831342022413e-08, "loss": 1.1822, "step": 11970 }, { "epoch": 0.93, "learning_rate": 2.622360804441781e-08, "loss": 1.1746, "step": 11971 }, { "epoch": 0.93, "learning_rate": 2.6166446285389554e-08, "loss": 1.1624, "step": 11972 }, { "epoch": 0.93, "learning_rate": 2.6109346068550085e-08, "loss": 1.1588, "step": 11973 }, { "epoch": 0.93, "learning_rate": 2.6052307397507744e-08, "loss": 1.1779, "step": 11974 }, { "epoch": 0.93, "learning_rate": 2.59953302758672e-08, "loss": 1.1995, "step": 11975 }, { "epoch": 0.93, "learning_rate": 2.593841470722935e-08, "loss": 1.1854, "step": 11976 }, { "epoch": 0.93, "learning_rate": 2.5881560695190762e-08, "loss": 1.2094, "step": 11977 }, { "epoch": 0.93, "learning_rate": 2.582476824334434e-08, "loss": 1.189, "step": 11978 }, { "epoch": 0.93, "learning_rate": 2.5768037355279325e-08, "loss": 1.2537, "step": 11979 }, { "epoch": 0.93, "learning_rate": 2.571136803458096e-08, "loss": 1.2323, "step": 11980 }, { "epoch": 0.93, "learning_rate": 2.5654760284830158e-08, "loss": 0.8279, "step": 11981 }, { "epoch": 0.93, "learning_rate": 2.55982141096045e-08, "loss": 1.1902, "step": 11982 }, { "epoch": 0.93, "learning_rate": 2.5541729512477572e-08, "loss": 1.2299, "step": 11983 }, { "epoch": 0.93, "learning_rate": 2.5485306497018745e-08, "loss": 1.1762, "step": 11984 }, { "epoch": 0.93, "learning_rate": 2.5428945066793717e-08, "loss": 1.1795, "step": 11985 }, { "epoch": 0.93, "learning_rate": 2.5372645225364532e-08, "loss": 1.167, "step": 11986 }, { "epoch": 0.93, "learning_rate": 2.5316406976289007e-08, "loss": 0.7767, "step": 11987 }, { "epoch": 0.93, "learning_rate": 2.5260230323120858e-08, "loss": 1.1464, "step": 11988 }, { "epoch": 0.93, "learning_rate": 2.5204115269410686e-08, "loss": 1.2224, "step": 11989 }, { "epoch": 0.93, "learning_rate": 2.5148061818704547e-08, "loss": 1.1434, "step": 11990 }, { "epoch": 0.93, "learning_rate": 2.5092069974544492e-08, "loss": 0.8661, "step": 11991 }, { "epoch": 0.93, "learning_rate": 2.5036139740469474e-08, "loss": 1.1681, "step": 11992 }, { "epoch": 0.93, "learning_rate": 2.4980271120013775e-08, "loss": 1.2437, "step": 11993 }, { "epoch": 0.93, "learning_rate": 2.49244641167079e-08, "loss": 1.2128, "step": 11994 }, { "epoch": 0.93, "learning_rate": 2.486871873407892e-08, "loss": 1.1879, "step": 11995 }, { "epoch": 0.93, "learning_rate": 2.4813034975649573e-08, "loss": 1.2162, "step": 11996 }, { "epoch": 0.93, "learning_rate": 2.475741284493893e-08, "loss": 1.1846, "step": 11997 }, { "epoch": 0.93, "learning_rate": 2.4701852345461958e-08, "loss": 1.194, "step": 11998 }, { "epoch": 0.93, "learning_rate": 2.4646353480729855e-08, "loss": 1.158, "step": 11999 }, { "epoch": 0.93, "learning_rate": 2.4590916254250026e-08, "loss": 1.1877, "step": 12000 }, { "epoch": 0.93, "learning_rate": 2.4535540669525567e-08, "loss": 1.2116, "step": 12001 }, { "epoch": 0.93, "learning_rate": 2.4480226730056454e-08, "loss": 1.128, "step": 12002 }, { "epoch": 0.93, "learning_rate": 2.4424974439337888e-08, "loss": 1.2003, "step": 12003 }, { "epoch": 0.93, "learning_rate": 2.4369783800861852e-08, "loss": 0.8751, "step": 12004 }, { "epoch": 0.93, "learning_rate": 2.431465481811601e-08, "loss": 1.1951, "step": 12005 }, { "epoch": 0.93, "learning_rate": 2.4259587494584233e-08, "loss": 1.2451, "step": 12006 }, { "epoch": 0.93, "learning_rate": 2.420458183374685e-08, "loss": 1.2648, "step": 12007 }, { "epoch": 0.93, "learning_rate": 2.414963783907975e-08, "loss": 1.1708, "step": 12008 }, { "epoch": 0.93, "learning_rate": 2.4094755514054932e-08, "loss": 1.1834, "step": 12009 }, { "epoch": 0.93, "learning_rate": 2.4039934862141286e-08, "loss": 1.2222, "step": 12010 }, { "epoch": 0.93, "learning_rate": 2.398517588680282e-08, "loss": 0.8361, "step": 12011 }, { "epoch": 0.93, "learning_rate": 2.3930478591500104e-08, "loss": 1.1973, "step": 12012 }, { "epoch": 0.93, "learning_rate": 2.387584297968992e-08, "loss": 1.2027, "step": 12013 }, { "epoch": 0.93, "learning_rate": 2.382126905482518e-08, "loss": 1.1891, "step": 12014 }, { "epoch": 0.93, "learning_rate": 2.376675682035434e-08, "loss": 1.1647, "step": 12015 }, { "epoch": 0.93, "learning_rate": 2.3712306279722317e-08, "loss": 1.1557, "step": 12016 }, { "epoch": 0.93, "learning_rate": 2.3657917436370687e-08, "loss": 0.8262, "step": 12017 }, { "epoch": 0.93, "learning_rate": 2.3603590293736043e-08, "loss": 1.1935, "step": 12018 }, { "epoch": 0.93, "learning_rate": 2.3549324855251852e-08, "loss": 1.1521, "step": 12019 }, { "epoch": 0.93, "learning_rate": 2.3495121124347595e-08, "loss": 1.1071, "step": 12020 }, { "epoch": 0.93, "learning_rate": 2.3440979104448534e-08, "loss": 1.2068, "step": 12021 }, { "epoch": 0.93, "learning_rate": 2.3386898798976153e-08, "loss": 1.15, "step": 12022 }, { "epoch": 0.93, "learning_rate": 2.3332880211348272e-08, "loss": 1.1691, "step": 12023 }, { "epoch": 0.93, "learning_rate": 2.3278923344978717e-08, "loss": 0.8335, "step": 12024 }, { "epoch": 0.93, "learning_rate": 2.3225028203276985e-08, "loss": 1.2628, "step": 12025 }, { "epoch": 0.93, "learning_rate": 2.3171194789649352e-08, "loss": 1.1996, "step": 12026 }, { "epoch": 0.93, "learning_rate": 2.311742310749776e-08, "loss": 1.2441, "step": 12027 }, { "epoch": 0.93, "learning_rate": 2.3063713160220267e-08, "loss": 0.849, "step": 12028 }, { "epoch": 0.93, "learning_rate": 2.301006495121116e-08, "loss": 1.2002, "step": 12029 }, { "epoch": 0.93, "learning_rate": 2.295647848386095e-08, "loss": 0.8425, "step": 12030 }, { "epoch": 0.93, "learning_rate": 2.290295376155582e-08, "loss": 1.2072, "step": 12031 }, { "epoch": 0.93, "learning_rate": 2.2849490787678505e-08, "loss": 1.1439, "step": 12032 }, { "epoch": 0.93, "learning_rate": 2.2796089565607524e-08, "loss": 1.209, "step": 12033 }, { "epoch": 0.93, "learning_rate": 2.2742750098717733e-08, "loss": 0.7846, "step": 12034 }, { "epoch": 0.93, "learning_rate": 2.2689472390379662e-08, "loss": 1.1767, "step": 12035 }, { "epoch": 0.93, "learning_rate": 2.263625644396061e-08, "loss": 1.1086, "step": 12036 }, { "epoch": 0.93, "learning_rate": 2.258310226282345e-08, "loss": 1.186, "step": 12037 }, { "epoch": 0.93, "learning_rate": 2.2530009850327158e-08, "loss": 1.1713, "step": 12038 }, { "epoch": 0.93, "learning_rate": 2.2476979209827164e-08, "loss": 1.1809, "step": 12039 }, { "epoch": 0.93, "learning_rate": 2.242401034467467e-08, "loss": 1.2052, "step": 12040 }, { "epoch": 0.93, "learning_rate": 2.237110325821723e-08, "loss": 1.1637, "step": 12041 }, { "epoch": 0.93, "learning_rate": 2.2318257953798268e-08, "loss": 1.16, "step": 12042 }, { "epoch": 0.93, "learning_rate": 2.2265474434757127e-08, "loss": 1.2628, "step": 12043 }, { "epoch": 0.93, "learning_rate": 2.2212752704430016e-08, "loss": 1.222, "step": 12044 }, { "epoch": 0.93, "learning_rate": 2.2160092766148276e-08, "loss": 1.1609, "step": 12045 }, { "epoch": 0.93, "learning_rate": 2.2107494623239907e-08, "loss": 0.7969, "step": 12046 }, { "epoch": 0.93, "learning_rate": 2.2054958279029035e-08, "loss": 1.1896, "step": 12047 }, { "epoch": 0.94, "learning_rate": 2.2002483736835777e-08, "loss": 0.8118, "step": 12048 }, { "epoch": 0.94, "learning_rate": 2.1950070999976032e-08, "loss": 1.2027, "step": 12049 }, { "epoch": 0.94, "learning_rate": 2.189772007176216e-08, "loss": 1.2122, "step": 12050 }, { "epoch": 0.94, "learning_rate": 2.1845430955502732e-08, "loss": 1.1891, "step": 12051 }, { "epoch": 0.94, "learning_rate": 2.179320365450188e-08, "loss": 1.2146, "step": 12052 }, { "epoch": 0.94, "learning_rate": 2.17410381720603e-08, "loss": 1.226, "step": 12053 }, { "epoch": 0.94, "learning_rate": 2.16889345114748e-08, "loss": 0.8539, "step": 12054 }, { "epoch": 0.94, "learning_rate": 2.1636892676037856e-08, "loss": 1.197, "step": 12055 }, { "epoch": 0.94, "learning_rate": 2.158491266903828e-08, "loss": 1.1679, "step": 12056 }, { "epoch": 0.94, "learning_rate": 2.1532994493761114e-08, "loss": 1.2265, "step": 12057 }, { "epoch": 0.94, "learning_rate": 2.1481138153487512e-08, "loss": 1.2117, "step": 12058 }, { "epoch": 0.94, "learning_rate": 2.1429343651494293e-08, "loss": 0.8442, "step": 12059 }, { "epoch": 0.94, "learning_rate": 2.137761099105484e-08, "loss": 1.1874, "step": 12060 }, { "epoch": 0.94, "learning_rate": 2.132594017543843e-08, "loss": 1.1786, "step": 12061 }, { "epoch": 0.94, "learning_rate": 2.1274331207910224e-08, "loss": 1.2111, "step": 12062 }, { "epoch": 0.94, "learning_rate": 2.122278409173195e-08, "loss": 1.1749, "step": 12063 }, { "epoch": 0.94, "learning_rate": 2.117129883016111e-08, "loss": 1.2192, "step": 12064 }, { "epoch": 0.94, "learning_rate": 2.1119875426451218e-08, "loss": 1.2115, "step": 12065 }, { "epoch": 0.94, "learning_rate": 2.106851388385211e-08, "loss": 1.1803, "step": 12066 }, { "epoch": 0.94, "learning_rate": 2.1017214205609758e-08, "loss": 1.1887, "step": 12067 }, { "epoch": 0.94, "learning_rate": 2.0965976394966e-08, "loss": 1.1727, "step": 12068 }, { "epoch": 0.94, "learning_rate": 2.091480045515859e-08, "loss": 1.2132, "step": 12069 }, { "epoch": 0.94, "learning_rate": 2.086368638942193e-08, "loss": 1.215, "step": 12070 }, { "epoch": 0.94, "learning_rate": 2.0812634200986223e-08, "loss": 0.833, "step": 12071 }, { "epoch": 0.94, "learning_rate": 2.0761643893077552e-08, "loss": 1.2372, "step": 12072 }, { "epoch": 0.94, "learning_rate": 2.071071546891845e-08, "loss": 1.2128, "step": 12073 }, { "epoch": 0.94, "learning_rate": 2.065984893172723e-08, "loss": 1.1923, "step": 12074 }, { "epoch": 0.94, "learning_rate": 2.0609044284718657e-08, "loss": 0.8042, "step": 12075 }, { "epoch": 0.94, "learning_rate": 2.055830153110327e-08, "loss": 1.2651, "step": 12076 }, { "epoch": 0.94, "learning_rate": 2.0507620674087734e-08, "loss": 1.1657, "step": 12077 }, { "epoch": 0.94, "learning_rate": 2.0457001716874923e-08, "loss": 1.2189, "step": 12078 }, { "epoch": 0.94, "learning_rate": 2.0406444662663835e-08, "loss": 1.2577, "step": 12079 }, { "epoch": 0.94, "learning_rate": 2.0355949514649252e-08, "loss": 0.8278, "step": 12080 }, { "epoch": 0.94, "learning_rate": 2.0305516276022393e-08, "loss": 0.8274, "step": 12081 }, { "epoch": 0.94, "learning_rate": 2.025514494997038e-08, "loss": 1.1275, "step": 12082 }, { "epoch": 0.94, "learning_rate": 2.0204835539676555e-08, "loss": 1.1715, "step": 12083 }, { "epoch": 0.94, "learning_rate": 2.0154588048320154e-08, "loss": 1.1754, "step": 12084 }, { "epoch": 0.94, "learning_rate": 2.0104402479076744e-08, "loss": 0.8487, "step": 12085 }, { "epoch": 0.94, "learning_rate": 2.0054278835117677e-08, "loss": 1.1816, "step": 12086 }, { "epoch": 0.94, "learning_rate": 2.0004217119610533e-08, "loss": 1.1464, "step": 12087 }, { "epoch": 0.94, "learning_rate": 1.9954217335719227e-08, "loss": 1.2185, "step": 12088 }, { "epoch": 0.94, "learning_rate": 1.9904279486603336e-08, "loss": 0.7854, "step": 12089 }, { "epoch": 0.94, "learning_rate": 1.9854403575418786e-08, "loss": 1.1308, "step": 12090 }, { "epoch": 0.94, "learning_rate": 1.9804589605317612e-08, "loss": 1.1907, "step": 12091 }, { "epoch": 0.94, "learning_rate": 1.9754837579447737e-08, "loss": 1.2173, "step": 12092 }, { "epoch": 0.94, "learning_rate": 1.9705147500953202e-08, "loss": 0.8657, "step": 12093 }, { "epoch": 0.94, "learning_rate": 1.9655519372974494e-08, "loss": 1.1904, "step": 12094 }, { "epoch": 0.94, "learning_rate": 1.960595319864766e-08, "loss": 1.2079, "step": 12095 }, { "epoch": 0.94, "learning_rate": 1.9556448981105088e-08, "loss": 1.1894, "step": 12096 }, { "epoch": 0.94, "learning_rate": 1.9507006723475384e-08, "loss": 1.1795, "step": 12097 }, { "epoch": 0.94, "learning_rate": 1.9457626428883044e-08, "loss": 1.1291, "step": 12098 }, { "epoch": 0.94, "learning_rate": 1.9408308100448466e-08, "loss": 1.162, "step": 12099 }, { "epoch": 0.94, "learning_rate": 1.93590517412886e-08, "loss": 0.8247, "step": 12100 }, { "epoch": 0.94, "learning_rate": 1.930985735451618e-08, "loss": 1.2066, "step": 12101 }, { "epoch": 0.94, "learning_rate": 1.926072494324027e-08, "loss": 1.1729, "step": 12102 }, { "epoch": 0.94, "learning_rate": 1.921165451056539e-08, "loss": 1.1635, "step": 12103 }, { "epoch": 0.94, "learning_rate": 1.9162646059592947e-08, "loss": 1.2196, "step": 12104 }, { "epoch": 0.94, "learning_rate": 1.911369959342013e-08, "loss": 0.7873, "step": 12105 }, { "epoch": 0.94, "learning_rate": 1.90648151151398e-08, "loss": 1.2077, "step": 12106 }, { "epoch": 0.94, "learning_rate": 1.901599262784148e-08, "loss": 1.1896, "step": 12107 }, { "epoch": 0.94, "learning_rate": 1.896723213461049e-08, "loss": 1.2161, "step": 12108 }, { "epoch": 0.94, "learning_rate": 1.8918533638528246e-08, "loss": 1.1922, "step": 12109 }, { "epoch": 0.94, "learning_rate": 1.886989714267251e-08, "loss": 1.1386, "step": 12110 }, { "epoch": 0.94, "learning_rate": 1.8821322650116488e-08, "loss": 1.1471, "step": 12111 }, { "epoch": 0.94, "learning_rate": 1.8772810163930397e-08, "loss": 1.1606, "step": 12112 }, { "epoch": 0.94, "learning_rate": 1.872435968717967e-08, "loss": 1.2267, "step": 12113 }, { "epoch": 0.94, "learning_rate": 1.8675971222926302e-08, "loss": 1.1658, "step": 12114 }, { "epoch": 0.94, "learning_rate": 1.8627644774228178e-08, "loss": 1.1944, "step": 12115 }, { "epoch": 0.94, "learning_rate": 1.8579380344139305e-08, "loss": 1.2047, "step": 12116 }, { "epoch": 0.94, "learning_rate": 1.8531177935709908e-08, "loss": 1.2107, "step": 12117 }, { "epoch": 0.94, "learning_rate": 1.848303755198588e-08, "loss": 1.2316, "step": 12118 }, { "epoch": 0.94, "learning_rate": 1.8434959196010014e-08, "loss": 1.1696, "step": 12119 }, { "epoch": 0.94, "learning_rate": 1.838694287082032e-08, "loss": 1.2063, "step": 12120 }, { "epoch": 0.94, "learning_rate": 1.8338988579451043e-08, "loss": 1.1481, "step": 12121 }, { "epoch": 0.94, "learning_rate": 1.8291096324933196e-08, "loss": 1.1671, "step": 12122 }, { "epoch": 0.94, "learning_rate": 1.8243266110293032e-08, "loss": 0.7939, "step": 12123 }, { "epoch": 0.94, "learning_rate": 1.8195497938553128e-08, "loss": 1.1753, "step": 12124 }, { "epoch": 0.94, "learning_rate": 1.8147791812732627e-08, "loss": 1.2528, "step": 12125 }, { "epoch": 0.94, "learning_rate": 1.8100147735845894e-08, "loss": 1.1887, "step": 12126 }, { "epoch": 0.94, "learning_rate": 1.805256571090408e-08, "loss": 1.2091, "step": 12127 }, { "epoch": 0.94, "learning_rate": 1.8005045740914217e-08, "loss": 1.1069, "step": 12128 }, { "epoch": 0.94, "learning_rate": 1.795758782887935e-08, "loss": 1.2291, "step": 12129 }, { "epoch": 0.94, "learning_rate": 1.7910191977798417e-08, "loss": 1.1553, "step": 12130 }, { "epoch": 0.94, "learning_rate": 1.7862858190666908e-08, "loss": 1.1963, "step": 12131 }, { "epoch": 0.94, "learning_rate": 1.7815586470475986e-08, "loss": 1.184, "step": 12132 }, { "epoch": 0.94, "learning_rate": 1.7768376820213037e-08, "loss": 1.2149, "step": 12133 }, { "epoch": 0.94, "learning_rate": 1.7721229242861347e-08, "loss": 1.1416, "step": 12134 }, { "epoch": 0.94, "learning_rate": 1.7674143741400748e-08, "loss": 1.1591, "step": 12135 }, { "epoch": 0.94, "learning_rate": 1.7627120318806864e-08, "loss": 1.203, "step": 12136 }, { "epoch": 0.94, "learning_rate": 1.7580158978050875e-08, "loss": 1.1906, "step": 12137 }, { "epoch": 0.94, "learning_rate": 1.753325972210107e-08, "loss": 1.1565, "step": 12138 }, { "epoch": 0.94, "learning_rate": 1.7486422553921187e-08, "loss": 1.2131, "step": 12139 }, { "epoch": 0.94, "learning_rate": 1.7439647476470865e-08, "loss": 1.171, "step": 12140 }, { "epoch": 0.94, "learning_rate": 1.73929344927064e-08, "loss": 1.1862, "step": 12141 }, { "epoch": 0.94, "learning_rate": 1.734628360557977e-08, "loss": 1.2684, "step": 12142 }, { "epoch": 0.94, "learning_rate": 1.7299694818038944e-08, "loss": 1.2052, "step": 12143 }, { "epoch": 0.94, "learning_rate": 1.725316813302835e-08, "loss": 1.1971, "step": 12144 }, { "epoch": 0.94, "learning_rate": 1.720670355348808e-08, "loss": 1.2119, "step": 12145 }, { "epoch": 0.94, "learning_rate": 1.7160301082354776e-08, "loss": 1.2327, "step": 12146 }, { "epoch": 0.94, "learning_rate": 1.7113960722560662e-08, "loss": 1.1227, "step": 12147 }, { "epoch": 0.94, "learning_rate": 1.7067682477034383e-08, "loss": 1.2055, "step": 12148 }, { "epoch": 0.94, "learning_rate": 1.7021466348700386e-08, "loss": 1.1102, "step": 12149 }, { "epoch": 0.94, "learning_rate": 1.6975312340479443e-08, "loss": 1.174, "step": 12150 }, { "epoch": 0.94, "learning_rate": 1.6929220455288328e-08, "loss": 1.2127, "step": 12151 }, { "epoch": 0.94, "learning_rate": 1.688319069603983e-08, "loss": 1.2314, "step": 12152 }, { "epoch": 0.94, "learning_rate": 1.6837223065642614e-08, "loss": 1.2048, "step": 12153 }, { "epoch": 0.94, "learning_rate": 1.6791317567001916e-08, "loss": 1.1677, "step": 12154 }, { "epoch": 0.94, "learning_rate": 1.674547420301853e-08, "loss": 1.2402, "step": 12155 }, { "epoch": 0.94, "learning_rate": 1.669969297659002e-08, "loss": 1.1901, "step": 12156 }, { "epoch": 0.94, "learning_rate": 1.6653973890609073e-08, "loss": 1.2303, "step": 12157 }, { "epoch": 0.94, "learning_rate": 1.660831694796505e-08, "loss": 1.1741, "step": 12158 }, { "epoch": 0.94, "learning_rate": 1.656272215154353e-08, "loss": 1.2014, "step": 12159 }, { "epoch": 0.94, "learning_rate": 1.651718950422565e-08, "loss": 1.1843, "step": 12160 }, { "epoch": 0.94, "learning_rate": 1.6471719008888995e-08, "loss": 1.2012, "step": 12161 }, { "epoch": 0.94, "learning_rate": 1.642631066840705e-08, "loss": 1.1904, "step": 12162 }, { "epoch": 0.94, "learning_rate": 1.6380964485649627e-08, "loss": 0.8378, "step": 12163 }, { "epoch": 0.94, "learning_rate": 1.6335680463481995e-08, "loss": 1.1457, "step": 12164 }, { "epoch": 0.94, "learning_rate": 1.62904586047663e-08, "loss": 1.2094, "step": 12165 }, { "epoch": 0.94, "learning_rate": 1.6245298912360373e-08, "loss": 1.1676, "step": 12166 }, { "epoch": 0.94, "learning_rate": 1.620020138911793e-08, "loss": 1.2294, "step": 12167 }, { "epoch": 0.94, "learning_rate": 1.61551660378888e-08, "loss": 1.1914, "step": 12168 }, { "epoch": 0.94, "learning_rate": 1.611019286151949e-08, "loss": 1.1414, "step": 12169 }, { "epoch": 0.94, "learning_rate": 1.6065281862851832e-08, "loss": 1.1908, "step": 12170 }, { "epoch": 0.94, "learning_rate": 1.6020433044723892e-08, "loss": 1.227, "step": 12171 }, { "epoch": 0.94, "learning_rate": 1.5975646409970068e-08, "loss": 1.1965, "step": 12172 }, { "epoch": 0.94, "learning_rate": 1.5930921961420872e-08, "loss": 1.2205, "step": 12173 }, { "epoch": 0.94, "learning_rate": 1.5886259701902382e-08, "loss": 1.1744, "step": 12174 }, { "epoch": 0.94, "learning_rate": 1.5841659634237225e-08, "loss": 1.2413, "step": 12175 }, { "epoch": 0.95, "learning_rate": 1.5797121761244036e-08, "loss": 1.1932, "step": 12176 }, { "epoch": 0.95, "learning_rate": 1.5752646085737233e-08, "loss": 1.186, "step": 12177 }, { "epoch": 0.95, "learning_rate": 1.570823261052756e-08, "loss": 1.1827, "step": 12178 }, { "epoch": 0.95, "learning_rate": 1.566388133842167e-08, "loss": 1.1875, "step": 12179 }, { "epoch": 0.95, "learning_rate": 1.5619592272222647e-08, "loss": 1.1617, "step": 12180 }, { "epoch": 0.95, "learning_rate": 1.5575365414729148e-08, "loss": 1.1586, "step": 12181 }, { "epoch": 0.95, "learning_rate": 1.5531200768736153e-08, "loss": 1.1951, "step": 12182 }, { "epoch": 0.95, "learning_rate": 1.5487098337034654e-08, "loss": 0.8101, "step": 12183 }, { "epoch": 0.95, "learning_rate": 1.5443058122411868e-08, "loss": 1.2326, "step": 12184 }, { "epoch": 0.95, "learning_rate": 1.53990801276509e-08, "loss": 1.1787, "step": 12185 }, { "epoch": 0.95, "learning_rate": 1.535516435553097e-08, "loss": 1.1942, "step": 12186 }, { "epoch": 0.95, "learning_rate": 1.5311310808827192e-08, "loss": 1.1828, "step": 12187 }, { "epoch": 0.95, "learning_rate": 1.5267519490311133e-08, "loss": 1.1849, "step": 12188 }, { "epoch": 0.95, "learning_rate": 1.5223790402750125e-08, "loss": 0.8546, "step": 12189 }, { "epoch": 0.95, "learning_rate": 1.5180123548907853e-08, "loss": 0.8348, "step": 12190 }, { "epoch": 0.95, "learning_rate": 1.513651893154366e-08, "loss": 1.2438, "step": 12191 }, { "epoch": 0.95, "learning_rate": 1.5092976553413238e-08, "loss": 1.1792, "step": 12192 }, { "epoch": 0.95, "learning_rate": 1.5049496417268382e-08, "loss": 1.2314, "step": 12193 }, { "epoch": 0.95, "learning_rate": 1.500607852585667e-08, "loss": 1.2388, "step": 12194 }, { "epoch": 0.95, "learning_rate": 1.496272288192202e-08, "loss": 1.1892, "step": 12195 }, { "epoch": 0.95, "learning_rate": 1.491942948820435e-08, "loss": 0.8612, "step": 12196 }, { "epoch": 0.95, "learning_rate": 1.4876198347439694e-08, "loss": 1.1954, "step": 12197 }, { "epoch": 0.95, "learning_rate": 1.483302946235987e-08, "loss": 1.2113, "step": 12198 }, { "epoch": 0.95, "learning_rate": 1.4789922835693135e-08, "loss": 1.1938, "step": 12199 }, { "epoch": 0.95, "learning_rate": 1.4746878470163649e-08, "loss": 1.1878, "step": 12200 }, { "epoch": 0.95, "learning_rate": 1.4703896368491564e-08, "loss": 1.21, "step": 12201 }, { "epoch": 0.95, "learning_rate": 1.4660976533393044e-08, "loss": 1.2035, "step": 12202 }, { "epoch": 0.95, "learning_rate": 1.4618118967580805e-08, "loss": 1.2263, "step": 12203 }, { "epoch": 0.95, "learning_rate": 1.4575323673762907e-08, "loss": 1.2102, "step": 12204 }, { "epoch": 0.95, "learning_rate": 1.4532590654643962e-08, "loss": 1.1807, "step": 12205 }, { "epoch": 0.95, "learning_rate": 1.4489919912924587e-08, "loss": 0.8304, "step": 12206 }, { "epoch": 0.95, "learning_rate": 1.4447311451301403e-08, "loss": 1.1941, "step": 12207 }, { "epoch": 0.95, "learning_rate": 1.4404765272466812e-08, "loss": 1.1417, "step": 12208 }, { "epoch": 0.95, "learning_rate": 1.4362281379109886e-08, "loss": 1.1716, "step": 12209 }, { "epoch": 0.95, "learning_rate": 1.4319859773915366e-08, "loss": 1.217, "step": 12210 }, { "epoch": 0.95, "learning_rate": 1.4277500459563885e-08, "loss": 1.1537, "step": 12211 }, { "epoch": 0.95, "learning_rate": 1.4235203438732745e-08, "loss": 0.8289, "step": 12212 }, { "epoch": 0.95, "learning_rate": 1.41929687140947e-08, "loss": 1.1638, "step": 12213 }, { "epoch": 0.95, "learning_rate": 1.4150796288318723e-08, "loss": 1.1903, "step": 12214 }, { "epoch": 0.95, "learning_rate": 1.4108686164070126e-08, "loss": 1.1787, "step": 12215 }, { "epoch": 0.95, "learning_rate": 1.4066638344010007e-08, "loss": 1.2091, "step": 12216 }, { "epoch": 0.95, "learning_rate": 1.4024652830795569e-08, "loss": 1.1741, "step": 12217 }, { "epoch": 0.95, "learning_rate": 1.3982729627080247e-08, "loss": 1.2096, "step": 12218 }, { "epoch": 0.95, "learning_rate": 1.3940868735513367e-08, "loss": 1.1729, "step": 12219 }, { "epoch": 0.95, "learning_rate": 1.3899070158740367e-08, "loss": 1.2051, "step": 12220 }, { "epoch": 0.95, "learning_rate": 1.385733389940269e-08, "loss": 1.1928, "step": 12221 }, { "epoch": 0.95, "learning_rate": 1.3815659960137783e-08, "loss": 1.1768, "step": 12222 }, { "epoch": 0.95, "learning_rate": 1.3774048343579425e-08, "loss": 1.1344, "step": 12223 }, { "epoch": 0.95, "learning_rate": 1.3732499052357405e-08, "loss": 1.2015, "step": 12224 }, { "epoch": 0.95, "learning_rate": 1.3691012089097287e-08, "loss": 1.1648, "step": 12225 }, { "epoch": 0.95, "learning_rate": 1.3649587456420864e-08, "loss": 1.1508, "step": 12226 }, { "epoch": 0.95, "learning_rate": 1.360822515694604e-08, "loss": 0.8233, "step": 12227 }, { "epoch": 0.95, "learning_rate": 1.3566925193286838e-08, "loss": 1.093, "step": 12228 }, { "epoch": 0.95, "learning_rate": 1.3525687568052946e-08, "loss": 1.1921, "step": 12229 }, { "epoch": 0.95, "learning_rate": 1.3484512283850835e-08, "loss": 1.2023, "step": 12230 }, { "epoch": 0.95, "learning_rate": 1.3443399343282313e-08, "loss": 1.2157, "step": 12231 }, { "epoch": 0.95, "learning_rate": 1.3402348748945524e-08, "loss": 1.2054, "step": 12232 }, { "epoch": 0.95, "learning_rate": 1.3361360503434838e-08, "loss": 1.2284, "step": 12233 }, { "epoch": 0.95, "learning_rate": 1.3320434609340514e-08, "loss": 1.2213, "step": 12234 }, { "epoch": 0.95, "learning_rate": 1.3279571069248819e-08, "loss": 1.151, "step": 12235 }, { "epoch": 0.95, "learning_rate": 1.323876988574224e-08, "loss": 1.2067, "step": 12236 }, { "epoch": 0.95, "learning_rate": 1.3198031061399273e-08, "loss": 1.2144, "step": 12237 }, { "epoch": 0.95, "learning_rate": 1.3157354598794302e-08, "loss": 0.8017, "step": 12238 }, { "epoch": 0.95, "learning_rate": 1.3116740500497936e-08, "loss": 1.1583, "step": 12239 }, { "epoch": 0.95, "learning_rate": 1.3076188769076901e-08, "loss": 0.8141, "step": 12240 }, { "epoch": 0.95, "learning_rate": 1.3035699407093926e-08, "loss": 1.2155, "step": 12241 }, { "epoch": 0.95, "learning_rate": 1.2995272417107628e-08, "loss": 1.1799, "step": 12242 }, { "epoch": 0.95, "learning_rate": 1.2954907801672855e-08, "loss": 1.1869, "step": 12243 }, { "epoch": 0.95, "learning_rate": 1.2914605563340674e-08, "loss": 1.1641, "step": 12244 }, { "epoch": 0.95, "learning_rate": 1.2874365704657719e-08, "loss": 0.8586, "step": 12245 }, { "epoch": 0.95, "learning_rate": 1.2834188228167175e-08, "loss": 1.1959, "step": 12246 }, { "epoch": 0.95, "learning_rate": 1.2794073136408123e-08, "loss": 1.2108, "step": 12247 }, { "epoch": 0.95, "learning_rate": 1.2754020431915425e-08, "loss": 1.2177, "step": 12248 }, { "epoch": 0.95, "learning_rate": 1.2714030117220498e-08, "loss": 1.1917, "step": 12249 }, { "epoch": 0.95, "learning_rate": 1.2674102194850433e-08, "loss": 1.1744, "step": 12250 }, { "epoch": 0.95, "learning_rate": 1.2634236667328657e-08, "loss": 0.8386, "step": 12251 }, { "epoch": 0.95, "learning_rate": 1.2594433537174265e-08, "loss": 1.1719, "step": 12252 }, { "epoch": 0.95, "learning_rate": 1.2554692806902688e-08, "loss": 1.1879, "step": 12253 }, { "epoch": 0.95, "learning_rate": 1.2515014479025699e-08, "loss": 1.2178, "step": 12254 }, { "epoch": 0.95, "learning_rate": 1.2475398556050287e-08, "loss": 1.1385, "step": 12255 }, { "epoch": 0.95, "learning_rate": 1.2435845040480452e-08, "loss": 1.2729, "step": 12256 }, { "epoch": 0.95, "learning_rate": 1.2396353934815752e-08, "loss": 1.1167, "step": 12257 }, { "epoch": 0.95, "learning_rate": 1.235692524155152e-08, "loss": 1.1858, "step": 12258 }, { "epoch": 0.95, "learning_rate": 1.2317558963179764e-08, "loss": 1.1928, "step": 12259 }, { "epoch": 0.95, "learning_rate": 1.2278255102188162e-08, "loss": 1.1594, "step": 12260 }, { "epoch": 0.95, "learning_rate": 1.2239013661060726e-08, "loss": 1.1565, "step": 12261 }, { "epoch": 0.95, "learning_rate": 1.2199834642277251e-08, "loss": 1.2275, "step": 12262 }, { "epoch": 0.95, "learning_rate": 1.2160718048313533e-08, "loss": 1.2011, "step": 12263 }, { "epoch": 0.95, "learning_rate": 1.212166388164182e-08, "loss": 1.1644, "step": 12264 }, { "epoch": 0.95, "learning_rate": 1.2082672144730021e-08, "loss": 1.1581, "step": 12265 }, { "epoch": 0.95, "learning_rate": 1.204374284004228e-08, "loss": 0.8375, "step": 12266 }, { "epoch": 0.95, "learning_rate": 1.2004875970038741e-08, "loss": 1.1844, "step": 12267 }, { "epoch": 0.95, "learning_rate": 1.196607153717577e-08, "loss": 0.8171, "step": 12268 }, { "epoch": 0.95, "learning_rate": 1.1927329543905518e-08, "loss": 1.1536, "step": 12269 }, { "epoch": 0.95, "learning_rate": 1.1888649992676137e-08, "loss": 0.8376, "step": 12270 }, { "epoch": 0.95, "learning_rate": 1.1850032885932448e-08, "loss": 1.2382, "step": 12271 }, { "epoch": 0.95, "learning_rate": 1.1811478226114612e-08, "loss": 1.1647, "step": 12272 }, { "epoch": 0.95, "learning_rate": 1.1772986015659015e-08, "loss": 0.8826, "step": 12273 }, { "epoch": 0.95, "learning_rate": 1.1734556256998485e-08, "loss": 1.2, "step": 12274 }, { "epoch": 0.95, "learning_rate": 1.1696188952561414e-08, "loss": 1.2174, "step": 12275 }, { "epoch": 0.95, "learning_rate": 1.1657884104772531e-08, "loss": 1.1979, "step": 12276 }, { "epoch": 0.95, "learning_rate": 1.1619641716052564e-08, "loss": 1.2042, "step": 12277 }, { "epoch": 0.95, "learning_rate": 1.1581461788818248e-08, "loss": 0.7828, "step": 12278 }, { "epoch": 0.95, "learning_rate": 1.1543344325482318e-08, "loss": 1.1618, "step": 12279 }, { "epoch": 0.95, "learning_rate": 1.1505289328453738e-08, "loss": 1.17, "step": 12280 }, { "epoch": 0.95, "learning_rate": 1.1467296800137472e-08, "loss": 1.2005, "step": 12281 }, { "epoch": 0.95, "learning_rate": 1.1429366742934266e-08, "loss": 1.1658, "step": 12282 }, { "epoch": 0.95, "learning_rate": 1.1391499159241424e-08, "loss": 1.1705, "step": 12283 }, { "epoch": 0.95, "learning_rate": 1.1353694051451812e-08, "loss": 1.1848, "step": 12284 }, { "epoch": 0.95, "learning_rate": 1.131595142195474e-08, "loss": 1.1689, "step": 12285 }, { "epoch": 0.95, "learning_rate": 1.1278271273135075e-08, "loss": 1.1637, "step": 12286 }, { "epoch": 0.95, "learning_rate": 1.124065360737425e-08, "loss": 0.799, "step": 12287 }, { "epoch": 0.95, "learning_rate": 1.1203098427049695e-08, "loss": 1.1818, "step": 12288 }, { "epoch": 0.95, "learning_rate": 1.11656057345344e-08, "loss": 1.1578, "step": 12289 }, { "epoch": 0.95, "learning_rate": 1.1128175532197914e-08, "loss": 1.1964, "step": 12290 }, { "epoch": 0.95, "learning_rate": 1.109080782240579e-08, "loss": 0.8181, "step": 12291 }, { "epoch": 0.95, "learning_rate": 1.105350260751925e-08, "loss": 1.2064, "step": 12292 }, { "epoch": 0.95, "learning_rate": 1.1016259889896074e-08, "loss": 0.862, "step": 12293 }, { "epoch": 0.95, "learning_rate": 1.0979079671889601e-08, "loss": 1.2146, "step": 12294 }, { "epoch": 0.95, "learning_rate": 1.0941961955849843e-08, "loss": 1.1795, "step": 12295 }, { "epoch": 0.95, "learning_rate": 1.0904906744122144e-08, "loss": 1.1201, "step": 12296 }, { "epoch": 0.95, "learning_rate": 1.0867914039048187e-08, "loss": 1.2169, "step": 12297 }, { "epoch": 0.95, "learning_rate": 1.0830983842966101e-08, "loss": 1.181, "step": 12298 }, { "epoch": 0.95, "learning_rate": 1.0794116158209465e-08, "loss": 1.2145, "step": 12299 }, { "epoch": 0.95, "learning_rate": 1.0757310987108193e-08, "loss": 0.7926, "step": 12300 }, { "epoch": 0.95, "learning_rate": 1.07205683319882e-08, "loss": 1.2086, "step": 12301 }, { "epoch": 0.95, "learning_rate": 1.0683888195171632e-08, "loss": 1.196, "step": 12302 }, { "epoch": 0.95, "learning_rate": 1.0647270578976298e-08, "loss": 1.2236, "step": 12303 }, { "epoch": 0.95, "learning_rate": 1.0610715485716348e-08, "loss": 0.8091, "step": 12304 }, { "epoch": 0.96, "learning_rate": 1.0574222917702046e-08, "loss": 1.1628, "step": 12305 }, { "epoch": 0.96, "learning_rate": 1.0537792877239437e-08, "loss": 1.2155, "step": 12306 }, { "epoch": 0.96, "learning_rate": 1.0501425366630678e-08, "loss": 1.1626, "step": 12307 }, { "epoch": 0.96, "learning_rate": 1.0465120388174376e-08, "loss": 1.2081, "step": 12308 }, { "epoch": 0.96, "learning_rate": 1.042887794416436e-08, "loss": 1.1156, "step": 12309 }, { "epoch": 0.96, "learning_rate": 1.0392698036891356e-08, "loss": 1.1651, "step": 12310 }, { "epoch": 0.96, "learning_rate": 1.0356580668641646e-08, "loss": 1.2121, "step": 12311 }, { "epoch": 0.96, "learning_rate": 1.0320525841697847e-08, "loss": 1.2125, "step": 12312 }, { "epoch": 0.96, "learning_rate": 1.0284533558338249e-08, "loss": 1.211, "step": 12313 }, { "epoch": 0.96, "learning_rate": 1.0248603820837476e-08, "loss": 1.1291, "step": 12314 }, { "epoch": 0.96, "learning_rate": 1.021273663146638e-08, "loss": 1.168, "step": 12315 }, { "epoch": 0.96, "learning_rate": 1.0176931992491255e-08, "loss": 1.219, "step": 12316 }, { "epoch": 0.96, "learning_rate": 1.0141189906174963e-08, "loss": 1.1726, "step": 12317 }, { "epoch": 0.96, "learning_rate": 1.010551037477636e-08, "loss": 1.217, "step": 12318 }, { "epoch": 0.96, "learning_rate": 1.006989340055009e-08, "loss": 1.1638, "step": 12319 }, { "epoch": 0.96, "learning_rate": 1.0034338985747016e-08, "loss": 1.2227, "step": 12320 }, { "epoch": 0.96, "learning_rate": 9.998847132614008e-09, "loss": 1.207, "step": 12321 }, { "epoch": 0.96, "learning_rate": 9.963417843394162e-09, "loss": 1.2084, "step": 12322 }, { "epoch": 0.96, "learning_rate": 9.928051120326242e-09, "loss": 1.1995, "step": 12323 }, { "epoch": 0.96, "learning_rate": 9.89274696564546e-09, "loss": 1.2154, "step": 12324 }, { "epoch": 0.96, "learning_rate": 9.85750538158292e-09, "loss": 1.1904, "step": 12325 }, { "epoch": 0.96, "learning_rate": 9.822326370365508e-09, "loss": 1.1958, "step": 12326 }, { "epoch": 0.96, "learning_rate": 9.787209934216444e-09, "loss": 1.2055, "step": 12327 }, { "epoch": 0.96, "learning_rate": 9.752156075355067e-09, "loss": 1.1812, "step": 12328 }, { "epoch": 0.96, "learning_rate": 9.717164795996714e-09, "loss": 0.8307, "step": 12329 }, { "epoch": 0.96, "learning_rate": 9.682236098352391e-09, "loss": 1.1719, "step": 12330 }, { "epoch": 0.96, "learning_rate": 9.647369984629671e-09, "loss": 1.1894, "step": 12331 }, { "epoch": 0.96, "learning_rate": 9.6125664570319e-09, "loss": 1.1356, "step": 12332 }, { "epoch": 0.96, "learning_rate": 9.577825517758542e-09, "loss": 1.1733, "step": 12333 }, { "epoch": 0.96, "learning_rate": 9.54314716900495e-09, "loss": 1.1781, "step": 12334 }, { "epoch": 0.96, "learning_rate": 9.508531412962707e-09, "loss": 1.1749, "step": 12335 }, { "epoch": 0.96, "learning_rate": 9.473978251819503e-09, "loss": 1.19, "step": 12336 }, { "epoch": 0.96, "learning_rate": 9.439487687758818e-09, "loss": 1.2163, "step": 12337 }, { "epoch": 0.96, "learning_rate": 9.40505972296024e-09, "loss": 1.1938, "step": 12338 }, { "epoch": 0.96, "learning_rate": 9.370694359599806e-09, "loss": 1.1296, "step": 12339 }, { "epoch": 0.96, "learning_rate": 9.336391599848891e-09, "loss": 1.2069, "step": 12340 }, { "epoch": 0.96, "learning_rate": 9.302151445875428e-09, "loss": 1.1804, "step": 12341 }, { "epoch": 0.96, "learning_rate": 9.26797389984335e-09, "loss": 1.216, "step": 12342 }, { "epoch": 0.96, "learning_rate": 9.233858963912378e-09, "loss": 1.1908, "step": 12343 }, { "epoch": 0.96, "learning_rate": 9.19980664023845e-09, "loss": 1.2013, "step": 12344 }, { "epoch": 0.96, "learning_rate": 9.165816930973735e-09, "loss": 1.1744, "step": 12345 }, { "epoch": 0.96, "learning_rate": 9.13188983826596e-09, "loss": 1.2151, "step": 12346 }, { "epoch": 0.96, "learning_rate": 9.098025364259188e-09, "loss": 1.126, "step": 12347 }, { "epoch": 0.96, "learning_rate": 9.064223511093817e-09, "loss": 1.2216, "step": 12348 }, { "epoch": 0.96, "learning_rate": 9.030484280905693e-09, "loss": 1.2056, "step": 12349 }, { "epoch": 0.96, "learning_rate": 8.996807675827112e-09, "loss": 1.2415, "step": 12350 }, { "epoch": 0.96, "learning_rate": 8.96319369798626e-09, "loss": 1.2649, "step": 12351 }, { "epoch": 0.96, "learning_rate": 8.929642349507438e-09, "loss": 1.1989, "step": 12352 }, { "epoch": 0.96, "learning_rate": 8.896153632510838e-09, "loss": 1.229, "step": 12353 }, { "epoch": 0.96, "learning_rate": 8.86272754911288e-09, "loss": 1.2091, "step": 12354 }, { "epoch": 0.96, "learning_rate": 8.829364101425984e-09, "loss": 1.2137, "step": 12355 }, { "epoch": 0.96, "learning_rate": 8.796063291558687e-09, "loss": 1.1475, "step": 12356 }, { "epoch": 0.96, "learning_rate": 8.76282512161508e-09, "loss": 1.2167, "step": 12357 }, { "epoch": 0.96, "learning_rate": 8.729649593696153e-09, "loss": 0.828, "step": 12358 }, { "epoch": 0.96, "learning_rate": 8.696536709898117e-09, "loss": 1.1688, "step": 12359 }, { "epoch": 0.96, "learning_rate": 8.66348647231363e-09, "loss": 1.1542, "step": 12360 }, { "epoch": 0.96, "learning_rate": 8.630498883031467e-09, "loss": 1.13, "step": 12361 }, { "epoch": 0.96, "learning_rate": 8.597573944136293e-09, "loss": 1.195, "step": 12362 }, { "epoch": 0.96, "learning_rate": 8.564711657708557e-09, "loss": 1.1988, "step": 12363 }, { "epoch": 0.96, "learning_rate": 8.53191202582526e-09, "loss": 1.2417, "step": 12364 }, { "epoch": 0.96, "learning_rate": 8.499175050559193e-09, "loss": 1.2367, "step": 12365 }, { "epoch": 0.96, "learning_rate": 8.466500733979142e-09, "loss": 1.1586, "step": 12366 }, { "epoch": 0.96, "learning_rate": 8.433889078150013e-09, "loss": 1.1665, "step": 12367 }, { "epoch": 0.96, "learning_rate": 8.401340085132713e-09, "loss": 0.8062, "step": 12368 }, { "epoch": 0.96, "learning_rate": 8.368853756984262e-09, "loss": 1.1732, "step": 12369 }, { "epoch": 0.96, "learning_rate": 8.336430095757463e-09, "loss": 1.179, "step": 12370 }, { "epoch": 0.96, "learning_rate": 8.304069103501565e-09, "loss": 1.1567, "step": 12371 }, { "epoch": 0.96, "learning_rate": 8.271770782261488e-09, "loss": 1.2178, "step": 12372 }, { "epoch": 0.96, "learning_rate": 8.23953513407849e-09, "loss": 1.2364, "step": 12373 }, { "epoch": 0.96, "learning_rate": 8.207362160989717e-09, "loss": 1.1851, "step": 12374 }, { "epoch": 0.96, "learning_rate": 8.17525186502821e-09, "loss": 1.1343, "step": 12375 }, { "epoch": 0.96, "learning_rate": 8.143204248223456e-09, "loss": 0.7854, "step": 12376 }, { "epoch": 0.96, "learning_rate": 8.111219312600503e-09, "loss": 1.1338, "step": 12377 }, { "epoch": 0.96, "learning_rate": 8.079297060180623e-09, "loss": 1.2187, "step": 12378 }, { "epoch": 0.96, "learning_rate": 8.047437492981534e-09, "loss": 1.158, "step": 12379 }, { "epoch": 0.96, "learning_rate": 8.015640613016294e-09, "loss": 1.1677, "step": 12380 }, { "epoch": 0.96, "learning_rate": 7.983906422294295e-09, "loss": 1.196, "step": 12381 }, { "epoch": 0.96, "learning_rate": 7.952234922821266e-09, "loss": 1.1473, "step": 12382 }, { "epoch": 0.96, "learning_rate": 7.920626116598605e-09, "loss": 1.1918, "step": 12383 }, { "epoch": 0.96, "learning_rate": 7.889080005623827e-09, "loss": 1.214, "step": 12384 }, { "epoch": 0.96, "learning_rate": 7.857596591890447e-09, "loss": 1.2026, "step": 12385 }, { "epoch": 0.96, "learning_rate": 7.826175877388209e-09, "loss": 1.1402, "step": 12386 }, { "epoch": 0.96, "learning_rate": 7.794817864102743e-09, "loss": 1.1815, "step": 12387 }, { "epoch": 0.96, "learning_rate": 7.763522554015578e-09, "loss": 1.1804, "step": 12388 }, { "epoch": 0.96, "learning_rate": 7.732289949104686e-09, "loss": 0.8278, "step": 12389 }, { "epoch": 0.96, "learning_rate": 7.701120051343823e-09, "loss": 1.1406, "step": 12390 }, { "epoch": 0.96, "learning_rate": 7.670012862702525e-09, "loss": 1.1598, "step": 12391 }, { "epoch": 0.96, "learning_rate": 7.638968385146882e-09, "loss": 1.2268, "step": 12392 }, { "epoch": 0.96, "learning_rate": 7.607986620638773e-09, "loss": 1.2271, "step": 12393 }, { "epoch": 0.96, "learning_rate": 7.577067571135965e-09, "loss": 1.2133, "step": 12394 }, { "epoch": 0.96, "learning_rate": 7.54621123859256e-09, "loss": 1.2339, "step": 12395 }, { "epoch": 0.96, "learning_rate": 7.515417624958553e-09, "loss": 1.1518, "step": 12396 }, { "epoch": 0.96, "learning_rate": 7.484686732179835e-09, "loss": 1.2036, "step": 12397 }, { "epoch": 0.96, "learning_rate": 7.454018562198517e-09, "loss": 1.1696, "step": 12398 }, { "epoch": 0.96, "learning_rate": 7.423413116952715e-09, "loss": 0.8327, "step": 12399 }, { "epoch": 0.96, "learning_rate": 7.39287039837666e-09, "loss": 1.2102, "step": 12400 }, { "epoch": 0.96, "learning_rate": 7.362390408400365e-09, "loss": 1.1811, "step": 12401 }, { "epoch": 0.96, "learning_rate": 7.331973148950177e-09, "loss": 1.1835, "step": 12402 }, { "epoch": 0.96, "learning_rate": 7.301618621948225e-09, "loss": 1.1924, "step": 12403 }, { "epoch": 0.96, "learning_rate": 7.271326829312862e-09, "loss": 1.1945, "step": 12404 }, { "epoch": 0.96, "learning_rate": 7.241097772958338e-09, "loss": 1.1708, "step": 12405 }, { "epoch": 0.96, "learning_rate": 7.210931454795122e-09, "loss": 1.1935, "step": 12406 }, { "epoch": 0.96, "learning_rate": 7.180827876729356e-09, "loss": 1.2247, "step": 12407 }, { "epoch": 0.96, "learning_rate": 7.150787040663741e-09, "loss": 1.2076, "step": 12408 }, { "epoch": 0.96, "learning_rate": 7.120808948496537e-09, "loss": 1.1838, "step": 12409 }, { "epoch": 0.96, "learning_rate": 7.090893602122338e-09, "loss": 1.161, "step": 12410 }, { "epoch": 0.96, "learning_rate": 7.061041003431633e-09, "loss": 1.1986, "step": 12411 }, { "epoch": 0.96, "learning_rate": 7.031251154310913e-09, "loss": 1.2188, "step": 12412 }, { "epoch": 0.96, "learning_rate": 7.001524056643005e-09, "loss": 0.835, "step": 12413 }, { "epoch": 0.96, "learning_rate": 6.971859712306294e-09, "loss": 1.2187, "step": 12414 }, { "epoch": 0.96, "learning_rate": 6.942258123175393e-09, "loss": 1.2094, "step": 12415 }, { "epoch": 0.96, "learning_rate": 6.9127192911212496e-09, "loss": 1.1892, "step": 12416 }, { "epoch": 0.96, "learning_rate": 6.883243218010592e-09, "loss": 1.1695, "step": 12417 }, { "epoch": 0.96, "learning_rate": 6.8538299057059325e-09, "loss": 1.1666, "step": 12418 }, { "epoch": 0.96, "learning_rate": 6.824479356066115e-09, "loss": 1.1781, "step": 12419 }, { "epoch": 0.96, "learning_rate": 6.795191570946213e-09, "loss": 1.1796, "step": 12420 }, { "epoch": 0.96, "learning_rate": 6.765966552196855e-09, "loss": 1.223, "step": 12421 }, { "epoch": 0.96, "learning_rate": 6.736804301665011e-09, "loss": 0.8308, "step": 12422 }, { "epoch": 0.96, "learning_rate": 6.707704821193649e-09, "loss": 1.1113, "step": 12423 }, { "epoch": 0.96, "learning_rate": 6.678668112621744e-09, "loss": 1.206, "step": 12424 }, { "epoch": 0.96, "learning_rate": 6.649694177784049e-09, "loss": 1.1841, "step": 12425 }, { "epoch": 0.96, "learning_rate": 6.620783018511988e-09, "loss": 1.1234, "step": 12426 }, { "epoch": 0.96, "learning_rate": 6.591934636632435e-09, "loss": 1.2235, "step": 12427 }, { "epoch": 0.96, "learning_rate": 6.563149033968374e-09, "loss": 0.8712, "step": 12428 }, { "epoch": 0.96, "learning_rate": 6.534426212339128e-09, "loss": 1.2134, "step": 12429 }, { "epoch": 0.96, "learning_rate": 6.5057661735599125e-09, "loss": 1.1805, "step": 12430 }, { "epoch": 0.96, "learning_rate": 6.477168919441611e-09, "loss": 1.2003, "step": 12431 }, { "epoch": 0.96, "learning_rate": 6.448634451791668e-09, "loss": 1.2165, "step": 12432 }, { "epoch": 0.96, "learning_rate": 6.420162772413307e-09, "loss": 1.2144, "step": 12433 }, { "epoch": 0.97, "learning_rate": 6.3917538831057546e-09, "loss": 1.1497, "step": 12434 }, { "epoch": 0.97, "learning_rate": 6.363407785664465e-09, "loss": 1.1603, "step": 12435 }, { "epoch": 0.97, "learning_rate": 6.335124481880672e-09, "loss": 1.1763, "step": 12436 }, { "epoch": 0.97, "learning_rate": 6.306903973541944e-09, "loss": 1.1871, "step": 12437 }, { "epoch": 0.97, "learning_rate": 6.278746262431412e-09, "loss": 1.278, "step": 12438 }, { "epoch": 0.97, "learning_rate": 6.250651350328651e-09, "loss": 0.8417, "step": 12439 }, { "epoch": 0.97, "learning_rate": 6.222619239009352e-09, "loss": 1.2289, "step": 12440 }, { "epoch": 0.97, "learning_rate": 6.194649930244655e-09, "loss": 0.81, "step": 12441 }, { "epoch": 0.97, "learning_rate": 6.166743425802367e-09, "loss": 1.2235, "step": 12442 }, { "epoch": 0.97, "learning_rate": 6.138899727445857e-09, "loss": 1.239, "step": 12443 }, { "epoch": 0.97, "learning_rate": 6.111118836935048e-09, "loss": 1.2165, "step": 12444 }, { "epoch": 0.97, "learning_rate": 6.0834007560253144e-09, "loss": 1.1358, "step": 12445 }, { "epoch": 0.97, "learning_rate": 6.055745486468366e-09, "loss": 1.1961, "step": 12446 }, { "epoch": 0.97, "learning_rate": 6.028153030011918e-09, "loss": 0.8655, "step": 12447 }, { "epoch": 0.97, "learning_rate": 6.000623388399795e-09, "loss": 1.2153, "step": 12448 }, { "epoch": 0.97, "learning_rate": 5.973156563371496e-09, "loss": 1.1616, "step": 12449 }, { "epoch": 0.97, "learning_rate": 5.945752556663186e-09, "loss": 0.8383, "step": 12450 }, { "epoch": 0.97, "learning_rate": 5.918411370006371e-09, "loss": 0.8357, "step": 12451 }, { "epoch": 0.97, "learning_rate": 5.891133005129001e-09, "loss": 1.1898, "step": 12452 }, { "epoch": 0.97, "learning_rate": 5.863917463754919e-09, "loss": 1.1453, "step": 12453 }, { "epoch": 0.97, "learning_rate": 5.836764747604084e-09, "loss": 1.1666, "step": 12454 }, { "epoch": 0.97, "learning_rate": 5.809674858392455e-09, "loss": 1.2321, "step": 12455 }, { "epoch": 0.97, "learning_rate": 5.782647797831886e-09, "loss": 1.1788, "step": 12456 }, { "epoch": 0.97, "learning_rate": 5.755683567630565e-09, "loss": 1.2197, "step": 12457 }, { "epoch": 0.97, "learning_rate": 5.728782169492352e-09, "loss": 1.2232, "step": 12458 }, { "epoch": 0.97, "learning_rate": 5.701943605117332e-09, "loss": 1.1814, "step": 12459 }, { "epoch": 0.97, "learning_rate": 5.675167876201703e-09, "loss": 1.2195, "step": 12460 }, { "epoch": 0.97, "learning_rate": 5.648454984437445e-09, "loss": 1.1686, "step": 12461 }, { "epoch": 0.97, "learning_rate": 5.621804931512653e-09, "loss": 1.1872, "step": 12462 }, { "epoch": 0.97, "learning_rate": 5.595217719111645e-09, "loss": 1.2179, "step": 12463 }, { "epoch": 0.97, "learning_rate": 5.568693348914632e-09, "loss": 1.1752, "step": 12464 }, { "epoch": 0.97, "learning_rate": 5.54223182259772e-09, "loss": 1.1716, "step": 12465 }, { "epoch": 0.97, "learning_rate": 5.515833141833237e-09, "loss": 1.21, "step": 12466 }, { "epoch": 0.97, "learning_rate": 5.489497308289404e-09, "loss": 1.2394, "step": 12467 }, { "epoch": 0.97, "learning_rate": 5.463224323630555e-09, "loss": 1.1984, "step": 12468 }, { "epoch": 0.97, "learning_rate": 5.4370141895171415e-09, "loss": 1.2189, "step": 12469 }, { "epoch": 0.97, "learning_rate": 5.410866907605505e-09, "loss": 1.2316, "step": 12470 }, { "epoch": 0.97, "learning_rate": 5.384782479547989e-09, "loss": 1.1527, "step": 12471 }, { "epoch": 0.97, "learning_rate": 5.358760906992832e-09, "loss": 1.149, "step": 12472 }, { "epoch": 0.97, "learning_rate": 5.332802191584829e-09, "loss": 1.1614, "step": 12473 }, { "epoch": 0.97, "learning_rate": 5.306906334964334e-09, "loss": 1.2337, "step": 12474 }, { "epoch": 0.97, "learning_rate": 5.281073338767817e-09, "loss": 1.1796, "step": 12475 }, { "epoch": 0.97, "learning_rate": 5.2553032046277476e-09, "loss": 1.1827, "step": 12476 }, { "epoch": 0.97, "learning_rate": 5.229595934172937e-09, "loss": 1.2, "step": 12477 }, { "epoch": 0.97, "learning_rate": 5.203951529027639e-09, "loss": 1.1783, "step": 12478 }, { "epoch": 0.97, "learning_rate": 5.17836999081267e-09, "loss": 1.1453, "step": 12479 }, { "epoch": 0.97, "learning_rate": 5.152851321144735e-09, "loss": 1.1623, "step": 12480 }, { "epoch": 0.97, "learning_rate": 5.127395521636435e-09, "loss": 1.1628, "step": 12481 }, { "epoch": 0.97, "learning_rate": 5.102002593896371e-09, "loss": 1.1694, "step": 12482 }, { "epoch": 0.97, "learning_rate": 5.0766725395293695e-09, "loss": 1.1668, "step": 12483 }, { "epoch": 0.97, "learning_rate": 5.051405360136262e-09, "loss": 1.139, "step": 12484 }, { "epoch": 0.97, "learning_rate": 5.0262010573136615e-09, "loss": 0.8385, "step": 12485 }, { "epoch": 0.97, "learning_rate": 5.001059632654514e-09, "loss": 1.2029, "step": 12486 }, { "epoch": 0.97, "learning_rate": 4.9759810877474384e-09, "loss": 1.1686, "step": 12487 }, { "epoch": 0.97, "learning_rate": 4.950965424177611e-09, "loss": 1.184, "step": 12488 }, { "epoch": 0.97, "learning_rate": 4.926012643525768e-09, "loss": 1.1716, "step": 12489 }, { "epoch": 0.97, "learning_rate": 4.9011227473686464e-09, "loss": 1.262, "step": 12490 }, { "epoch": 0.97, "learning_rate": 4.876295737279435e-09, "loss": 1.132, "step": 12491 }, { "epoch": 0.97, "learning_rate": 4.851531614826987e-09, "loss": 1.1855, "step": 12492 }, { "epoch": 0.97, "learning_rate": 4.8268303815762746e-09, "loss": 1.2213, "step": 12493 }, { "epoch": 0.97, "learning_rate": 4.8021920390883815e-09, "loss": 1.1463, "step": 12494 }, { "epoch": 0.97, "learning_rate": 4.7776165889202855e-09, "loss": 1.2372, "step": 12495 }, { "epoch": 0.97, "learning_rate": 4.753104032625077e-09, "loss": 1.1546, "step": 12496 }, { "epoch": 0.97, "learning_rate": 4.72865437175185e-09, "loss": 1.1666, "step": 12497 }, { "epoch": 0.97, "learning_rate": 4.704267607845702e-09, "loss": 1.1529, "step": 12498 }, { "epoch": 0.97, "learning_rate": 4.679943742447845e-09, "loss": 1.1898, "step": 12499 }, { "epoch": 0.97, "learning_rate": 4.6556827770952715e-09, "loss": 1.1679, "step": 12500 }, { "epoch": 0.97, "learning_rate": 4.63148471332142e-09, "loss": 1.2026, "step": 12501 }, { "epoch": 0.97, "learning_rate": 4.607349552655293e-09, "loss": 1.1752, "step": 12502 }, { "epoch": 0.97, "learning_rate": 4.583277296622112e-09, "loss": 1.2022, "step": 12503 }, { "epoch": 0.97, "learning_rate": 4.5592679467434394e-09, "loss": 1.2276, "step": 12504 }, { "epoch": 0.97, "learning_rate": 4.535321504536282e-09, "loss": 1.1619, "step": 12505 }, { "epoch": 0.97, "learning_rate": 4.511437971513987e-09, "loss": 1.1883, "step": 12506 }, { "epoch": 0.97, "learning_rate": 4.487617349186013e-09, "loss": 1.1848, "step": 12507 }, { "epoch": 0.97, "learning_rate": 4.4638596390575985e-09, "loss": 1.2235, "step": 12508 }, { "epoch": 0.97, "learning_rate": 4.440164842630212e-09, "loss": 1.1517, "step": 12509 }, { "epoch": 0.97, "learning_rate": 4.416532961401209e-09, "loss": 1.2115, "step": 12510 }, { "epoch": 0.97, "learning_rate": 4.392963996864174e-09, "loss": 1.1974, "step": 12511 }, { "epoch": 0.97, "learning_rate": 4.369457950508359e-09, "loss": 1.1753, "step": 12512 }, { "epoch": 0.97, "learning_rate": 4.346014823819355e-09, "loss": 1.166, "step": 12513 }, { "epoch": 0.97, "learning_rate": 4.322634618278642e-09, "loss": 1.2088, "step": 12514 }, { "epoch": 0.97, "learning_rate": 4.299317335363817e-09, "loss": 1.1459, "step": 12515 }, { "epoch": 0.97, "learning_rate": 4.276062976548256e-09, "loss": 0.8707, "step": 12516 }, { "epoch": 0.97, "learning_rate": 4.252871543301784e-09, "loss": 1.103, "step": 12517 }, { "epoch": 0.97, "learning_rate": 4.229743037089895e-09, "loss": 0.8347, "step": 12518 }, { "epoch": 0.97, "learning_rate": 4.206677459374086e-09, "loss": 1.1964, "step": 12519 }, { "epoch": 0.97, "learning_rate": 4.183674811612192e-09, "loss": 1.1665, "step": 12520 }, { "epoch": 0.97, "learning_rate": 4.160735095257717e-09, "loss": 1.1967, "step": 12521 }, { "epoch": 0.97, "learning_rate": 4.137858311760501e-09, "loss": 1.2009, "step": 12522 }, { "epoch": 0.97, "learning_rate": 4.115044462566275e-09, "loss": 1.1605, "step": 12523 }, { "epoch": 0.97, "learning_rate": 4.092293549116554e-09, "loss": 1.1875, "step": 12524 }, { "epoch": 0.97, "learning_rate": 4.0696055728494104e-09, "loss": 1.2111, "step": 12525 }, { "epoch": 0.97, "learning_rate": 4.046980535198363e-09, "loss": 1.2199, "step": 12526 }, { "epoch": 0.97, "learning_rate": 4.02441843759338e-09, "loss": 1.1762, "step": 12527 }, { "epoch": 0.97, "learning_rate": 4.001919281460209e-09, "loss": 1.1409, "step": 12528 }, { "epoch": 0.97, "learning_rate": 3.979483068220712e-09, "loss": 1.1479, "step": 12529 }, { "epoch": 0.97, "learning_rate": 3.957109799292868e-09, "loss": 1.2251, "step": 12530 }, { "epoch": 0.97, "learning_rate": 3.934799476090434e-09, "loss": 1.132, "step": 12531 }, { "epoch": 0.97, "learning_rate": 3.912552100023392e-09, "loss": 0.8604, "step": 12532 }, { "epoch": 0.97, "learning_rate": 3.890367672497619e-09, "loss": 0.8569, "step": 12533 }, { "epoch": 0.97, "learning_rate": 3.868246194915214e-09, "loss": 1.188, "step": 12534 }, { "epoch": 0.97, "learning_rate": 3.846187668674061e-09, "loss": 1.1888, "step": 12535 }, { "epoch": 0.97, "learning_rate": 3.824192095168155e-09, "loss": 0.8657, "step": 12536 }, { "epoch": 0.97, "learning_rate": 3.802259475787606e-09, "loss": 1.1971, "step": 12537 }, { "epoch": 0.97, "learning_rate": 3.780389811918416e-09, "loss": 1.2443, "step": 12538 }, { "epoch": 0.97, "learning_rate": 3.758583104942592e-09, "loss": 1.1904, "step": 12539 }, { "epoch": 0.97, "learning_rate": 3.736839356238253e-09, "loss": 1.1837, "step": 12540 }, { "epoch": 0.97, "learning_rate": 3.7151585671796324e-09, "loss": 1.2025, "step": 12541 }, { "epoch": 0.97, "learning_rate": 3.693540739136747e-09, "loss": 1.2217, "step": 12542 }, { "epoch": 0.97, "learning_rate": 3.6719858734757247e-09, "loss": 1.2108, "step": 12543 }, { "epoch": 0.97, "learning_rate": 3.650493971558921e-09, "loss": 1.1834, "step": 12544 }, { "epoch": 0.97, "learning_rate": 3.6290650347443607e-09, "loss": 1.125, "step": 12545 }, { "epoch": 0.97, "learning_rate": 3.6076990643862937e-09, "loss": 1.1603, "step": 12546 }, { "epoch": 0.97, "learning_rate": 3.586396061834973e-09, "loss": 1.1797, "step": 12547 }, { "epoch": 0.97, "learning_rate": 3.5651560284366555e-09, "loss": 1.16, "step": 12548 }, { "epoch": 0.97, "learning_rate": 3.5439789655337117e-09, "loss": 1.1731, "step": 12549 }, { "epoch": 0.97, "learning_rate": 3.522864874464404e-09, "loss": 1.2103, "step": 12550 }, { "epoch": 0.97, "learning_rate": 3.501813756562888e-09, "loss": 1.1841, "step": 12551 }, { "epoch": 0.97, "learning_rate": 3.4808256131596545e-09, "loss": 1.1709, "step": 12552 }, { "epoch": 0.97, "learning_rate": 3.4599004455810876e-09, "loss": 1.2505, "step": 12553 }, { "epoch": 0.97, "learning_rate": 3.439038255149573e-09, "loss": 1.202, "step": 12554 }, { "epoch": 0.97, "learning_rate": 3.4182390431833908e-09, "loss": 1.1387, "step": 12555 }, { "epoch": 0.97, "learning_rate": 3.397502810997044e-09, "loss": 1.1975, "step": 12556 }, { "epoch": 0.97, "learning_rate": 3.3768295599010397e-09, "loss": 1.231, "step": 12557 }, { "epoch": 0.97, "learning_rate": 3.3562192912016674e-09, "loss": 0.859, "step": 12558 }, { "epoch": 0.97, "learning_rate": 3.335672006201662e-09, "loss": 1.2092, "step": 12559 }, { "epoch": 0.97, "learning_rate": 3.315187706199207e-09, "loss": 1.1753, "step": 12560 }, { "epoch": 0.97, "learning_rate": 3.2947663924890456e-09, "loss": 1.2053, "step": 12561 }, { "epoch": 0.97, "learning_rate": 3.274408066361811e-09, "loss": 1.2639, "step": 12562 }, { "epoch": 0.98, "learning_rate": 3.2541127291038085e-09, "loss": 1.1904, "step": 12563 }, { "epoch": 0.98, "learning_rate": 3.233880381997678e-09, "loss": 0.8168, "step": 12564 }, { "epoch": 0.98, "learning_rate": 3.2137110263221746e-09, "loss": 1.1865, "step": 12565 }, { "epoch": 0.98, "learning_rate": 3.1936046633517233e-09, "loss": 1.2036, "step": 12566 }, { "epoch": 0.98, "learning_rate": 3.1735612943570855e-09, "loss": 1.1348, "step": 12567 }, { "epoch": 0.98, "learning_rate": 3.153580920604915e-09, "loss": 1.1678, "step": 12568 }, { "epoch": 0.98, "learning_rate": 3.1336635433578675e-09, "loss": 1.2032, "step": 12569 }, { "epoch": 0.98, "learning_rate": 3.1138091638746033e-09, "loss": 1.1908, "step": 12570 }, { "epoch": 0.98, "learning_rate": 3.0940177834098966e-09, "loss": 1.1986, "step": 12571 }, { "epoch": 0.98, "learning_rate": 3.074289403214414e-09, "loss": 1.1586, "step": 12572 }, { "epoch": 0.98, "learning_rate": 3.0546240245349355e-09, "loss": 1.2364, "step": 12573 }, { "epoch": 0.98, "learning_rate": 3.0350216486142444e-09, "loss": 1.1743, "step": 12574 }, { "epoch": 0.98, "learning_rate": 3.015482276691128e-09, "loss": 0.8328, "step": 12575 }, { "epoch": 0.98, "learning_rate": 2.9960059100003764e-09, "loss": 1.1783, "step": 12576 }, { "epoch": 0.98, "learning_rate": 2.976592549772783e-09, "loss": 1.1955, "step": 12577 }, { "epoch": 0.98, "learning_rate": 2.9572421972351437e-09, "loss": 1.21, "step": 12578 }, { "epoch": 0.98, "learning_rate": 2.9379548536105913e-09, "loss": 1.1295, "step": 12579 }, { "epoch": 0.98, "learning_rate": 2.918730520117707e-09, "loss": 1.2246, "step": 12580 }, { "epoch": 0.98, "learning_rate": 2.8995691979715185e-09, "loss": 0.8377, "step": 12581 }, { "epoch": 0.98, "learning_rate": 2.8804708883828356e-09, "loss": 1.1837, "step": 12582 }, { "epoch": 0.98, "learning_rate": 2.861435592558803e-09, "loss": 1.2294, "step": 12583 }, { "epoch": 0.98, "learning_rate": 2.8424633117021257e-09, "loss": 1.1412, "step": 12584 }, { "epoch": 0.98, "learning_rate": 2.8235540470119555e-09, "loss": 1.219, "step": 12585 }, { "epoch": 0.98, "learning_rate": 2.804707799683226e-09, "loss": 1.2125, "step": 12586 }, { "epoch": 0.98, "learning_rate": 2.7859245709068726e-09, "loss": 1.1875, "step": 12587 }, { "epoch": 0.98, "learning_rate": 2.767204361870057e-09, "loss": 1.1828, "step": 12588 }, { "epoch": 0.98, "learning_rate": 2.748547173755611e-09, "loss": 1.1928, "step": 12589 }, { "epoch": 0.98, "learning_rate": 2.7299530077428136e-09, "loss": 1.1392, "step": 12590 }, { "epoch": 0.98, "learning_rate": 2.7114218650065025e-09, "loss": 0.8475, "step": 12591 }, { "epoch": 0.98, "learning_rate": 2.692953746717963e-09, "loss": 1.1746, "step": 12592 }, { "epoch": 0.98, "learning_rate": 2.674548654044373e-09, "loss": 1.1697, "step": 12593 }, { "epoch": 0.98, "learning_rate": 2.6562065881485797e-09, "loss": 1.1911, "step": 12594 }, { "epoch": 0.98, "learning_rate": 2.637927550189878e-09, "loss": 1.1982, "step": 12595 }, { "epoch": 0.98, "learning_rate": 2.619711541323455e-09, "loss": 1.1981, "step": 12596 }, { "epoch": 0.98, "learning_rate": 2.60155856270039e-09, "loss": 1.1978, "step": 12597 }, { "epoch": 0.98, "learning_rate": 2.5834686154679875e-09, "loss": 1.2227, "step": 12598 }, { "epoch": 0.98, "learning_rate": 2.565441700769444e-09, "loss": 1.1525, "step": 12599 }, { "epoch": 0.98, "learning_rate": 2.5474778197438484e-09, "loss": 0.8053, "step": 12600 }, { "epoch": 0.98, "learning_rate": 2.5295769735265148e-09, "loss": 1.1673, "step": 12601 }, { "epoch": 0.98, "learning_rate": 2.5117391632488716e-09, "loss": 1.1711, "step": 12602 }, { "epoch": 0.98, "learning_rate": 2.493964390037906e-09, "loss": 1.1971, "step": 12603 }, { "epoch": 0.98, "learning_rate": 2.4762526550171635e-09, "loss": 1.1626, "step": 12604 }, { "epoch": 0.98, "learning_rate": 2.4586039593057494e-09, "loss": 1.207, "step": 12605 }, { "epoch": 0.98, "learning_rate": 2.4410183040191045e-09, "loss": 1.2157, "step": 12606 }, { "epoch": 0.98, "learning_rate": 2.4234956902684513e-09, "loss": 1.2281, "step": 12607 }, { "epoch": 0.98, "learning_rate": 2.406036119161348e-09, "loss": 1.1822, "step": 12608 }, { "epoch": 0.98, "learning_rate": 2.3886395918010226e-09, "loss": 1.1991, "step": 12609 }, { "epoch": 0.98, "learning_rate": 2.3713061092868193e-09, "loss": 1.1873, "step": 12610 }, { "epoch": 0.98, "learning_rate": 2.3540356727143053e-09, "loss": 1.1639, "step": 12611 }, { "epoch": 0.98, "learning_rate": 2.3368282831747186e-09, "loss": 1.2048, "step": 12612 }, { "epoch": 0.98, "learning_rate": 2.3196839417556348e-09, "loss": 1.213, "step": 12613 }, { "epoch": 0.98, "learning_rate": 2.3026026495404084e-09, "loss": 1.2257, "step": 12614 }, { "epoch": 0.98, "learning_rate": 2.2855844076086207e-09, "loss": 1.2337, "step": 12615 }, { "epoch": 0.98, "learning_rate": 2.2686292170356338e-09, "loss": 1.1812, "step": 12616 }, { "epoch": 0.98, "learning_rate": 2.2517370788930344e-09, "loss": 1.1949, "step": 12617 }, { "epoch": 0.98, "learning_rate": 2.234907994248192e-09, "loss": 1.1664, "step": 12618 }, { "epoch": 0.98, "learning_rate": 2.21814196416481e-09, "loss": 1.2061, "step": 12619 }, { "epoch": 0.98, "learning_rate": 2.2014389897023755e-09, "loss": 1.2391, "step": 12620 }, { "epoch": 0.98, "learning_rate": 2.184799071916377e-09, "loss": 1.1744, "step": 12621 }, { "epoch": 0.98, "learning_rate": 2.168222211858417e-09, "loss": 1.1925, "step": 12622 }, { "epoch": 0.98, "learning_rate": 2.151708410576103e-09, "loss": 1.1806, "step": 12623 }, { "epoch": 0.98, "learning_rate": 2.135257669113044e-09, "loss": 1.1666, "step": 12624 }, { "epoch": 0.98, "learning_rate": 2.1188699885088534e-09, "loss": 1.2379, "step": 12625 }, { "epoch": 0.98, "learning_rate": 2.102545369799258e-09, "loss": 1.217, "step": 12626 }, { "epoch": 0.98, "learning_rate": 2.0862838140156548e-09, "loss": 1.1636, "step": 12627 }, { "epoch": 0.98, "learning_rate": 2.0700853221859994e-09, "loss": 1.195, "step": 12628 }, { "epoch": 0.98, "learning_rate": 2.0539498953336954e-09, "loss": 1.1827, "step": 12629 }, { "epoch": 0.98, "learning_rate": 2.0378775344785936e-09, "loss": 1.12, "step": 12630 }, { "epoch": 0.98, "learning_rate": 2.0218682406364375e-09, "loss": 1.158, "step": 12631 }, { "epoch": 0.98, "learning_rate": 2.0059220148188615e-09, "loss": 0.8449, "step": 12632 }, { "epoch": 0.98, "learning_rate": 1.9900388580336157e-09, "loss": 1.1684, "step": 12633 }, { "epoch": 0.98, "learning_rate": 1.9742187712844526e-09, "loss": 1.1142, "step": 12634 }, { "epoch": 0.98, "learning_rate": 1.958461755571128e-09, "loss": 0.8274, "step": 12635 }, { "epoch": 0.98, "learning_rate": 1.94276781188929e-09, "loss": 0.8235, "step": 12636 }, { "epoch": 0.98, "learning_rate": 1.9271369412310334e-09, "loss": 1.175, "step": 12637 }, { "epoch": 0.98, "learning_rate": 1.911569144584013e-09, "loss": 1.2231, "step": 12638 }, { "epoch": 0.98, "learning_rate": 1.896064422931998e-09, "loss": 1.1709, "step": 12639 }, { "epoch": 0.98, "learning_rate": 1.880622777254759e-09, "loss": 1.1543, "step": 12640 }, { "epoch": 0.98, "learning_rate": 1.8652442085284047e-09, "loss": 1.2451, "step": 12641 }, { "epoch": 0.98, "learning_rate": 1.8499287177244916e-09, "loss": 1.1558, "step": 12642 }, { "epoch": 0.98, "learning_rate": 1.8346763058111336e-09, "loss": 1.211, "step": 12643 }, { "epoch": 0.98, "learning_rate": 1.8194869737520046e-09, "loss": 1.1976, "step": 12644 }, { "epoch": 0.98, "learning_rate": 1.8043607225072254e-09, "loss": 1.2452, "step": 12645 }, { "epoch": 0.98, "learning_rate": 1.789297553032587e-09, "loss": 1.1755, "step": 12646 }, { "epoch": 0.98, "learning_rate": 1.7742974662799947e-09, "loss": 1.2012, "step": 12647 }, { "epoch": 0.98, "learning_rate": 1.7593604631974679e-09, "loss": 1.1579, "step": 12648 }, { "epoch": 0.98, "learning_rate": 1.744486544728918e-09, "loss": 1.1443, "step": 12649 }, { "epoch": 0.98, "learning_rate": 1.7296757118143711e-09, "loss": 1.1408, "step": 12650 }, { "epoch": 0.98, "learning_rate": 1.7149279653897453e-09, "loss": 1.1577, "step": 12651 }, { "epoch": 0.98, "learning_rate": 1.7002433063869615e-09, "loss": 1.1731, "step": 12652 }, { "epoch": 0.98, "learning_rate": 1.6856217357342773e-09, "loss": 0.8465, "step": 12653 }, { "epoch": 0.98, "learning_rate": 1.6710632543555092e-09, "loss": 1.1824, "step": 12654 }, { "epoch": 0.98, "learning_rate": 1.6565678631706993e-09, "loss": 1.2442, "step": 12655 }, { "epoch": 0.98, "learning_rate": 1.6421355630958921e-09, "loss": 1.123, "step": 12656 }, { "epoch": 0.98, "learning_rate": 1.6277663550432475e-09, "loss": 1.1739, "step": 12657 }, { "epoch": 0.98, "learning_rate": 1.6134602399208164e-09, "loss": 1.2051, "step": 12658 }, { "epoch": 0.98, "learning_rate": 1.5992172186326535e-09, "loss": 1.1701, "step": 12659 }, { "epoch": 0.98, "learning_rate": 1.5850372920788168e-09, "loss": 0.8276, "step": 12660 }, { "epoch": 0.98, "learning_rate": 1.5709204611554782e-09, "loss": 1.1727, "step": 12661 }, { "epoch": 0.98, "learning_rate": 1.556866726754702e-09, "loss": 1.2426, "step": 12662 }, { "epoch": 0.98, "learning_rate": 1.5428760897647774e-09, "loss": 1.1747, "step": 12663 }, { "epoch": 0.98, "learning_rate": 1.5289485510696642e-09, "loss": 1.1671, "step": 12664 }, { "epoch": 0.98, "learning_rate": 1.5150841115496582e-09, "loss": 1.1751, "step": 12665 }, { "epoch": 0.98, "learning_rate": 1.5012827720808363e-09, "loss": 0.835, "step": 12666 }, { "epoch": 0.98, "learning_rate": 1.4875445335353897e-09, "loss": 1.2051, "step": 12667 }, { "epoch": 0.98, "learning_rate": 1.473869396781624e-09, "loss": 1.2417, "step": 12668 }, { "epoch": 0.98, "learning_rate": 1.4602573626836257e-09, "loss": 1.2591, "step": 12669 }, { "epoch": 0.98, "learning_rate": 1.4467084321017064e-09, "loss": 1.2356, "step": 12670 }, { "epoch": 0.98, "learning_rate": 1.4332226058920705e-09, "loss": 1.1831, "step": 12671 }, { "epoch": 0.98, "learning_rate": 1.4197998849070359e-09, "loss": 1.2089, "step": 12672 }, { "epoch": 0.98, "learning_rate": 1.406440269994702e-09, "loss": 1.1841, "step": 12673 }, { "epoch": 0.98, "learning_rate": 1.3931437619995046e-09, "loss": 1.1793, "step": 12674 }, { "epoch": 0.98, "learning_rate": 1.3799103617615494e-09, "loss": 1.2169, "step": 12675 }, { "epoch": 0.98, "learning_rate": 1.3667400701172782e-09, "loss": 1.1847, "step": 12676 }, { "epoch": 0.98, "learning_rate": 1.3536328878990256e-09, "loss": 1.1584, "step": 12677 }, { "epoch": 0.98, "learning_rate": 1.3405888159349066e-09, "loss": 1.2214, "step": 12678 }, { "epoch": 0.98, "learning_rate": 1.327607855049484e-09, "loss": 1.2134, "step": 12679 }, { "epoch": 0.98, "learning_rate": 1.3146900060629906e-09, "loss": 1.2136, "step": 12680 }, { "epoch": 0.98, "learning_rate": 1.3018352697917734e-09, "loss": 1.1765, "step": 12681 }, { "epoch": 0.98, "learning_rate": 1.2890436470481825e-09, "loss": 1.2281, "step": 12682 }, { "epoch": 0.98, "learning_rate": 1.2763151386406824e-09, "loss": 1.2102, "step": 12683 }, { "epoch": 0.98, "learning_rate": 1.2636497453735184e-09, "loss": 1.1919, "step": 12684 }, { "epoch": 0.98, "learning_rate": 1.2510474680472727e-09, "loss": 0.8123, "step": 12685 }, { "epoch": 0.98, "learning_rate": 1.238508307458197e-09, "loss": 1.2577, "step": 12686 }, { "epoch": 0.98, "learning_rate": 1.2260322643988796e-09, "loss": 1.227, "step": 12687 }, { "epoch": 0.98, "learning_rate": 1.2136193396575788e-09, "loss": 1.159, "step": 12688 }, { "epoch": 0.98, "learning_rate": 1.2012695340187783e-09, "loss": 1.1614, "step": 12689 }, { "epoch": 0.98, "learning_rate": 1.188982848262965e-09, "loss": 1.2591, "step": 12690 }, { "epoch": 0.98, "learning_rate": 1.1767592831667395e-09, "loss": 1.1737, "step": 12691 }, { "epoch": 0.99, "learning_rate": 1.1645988395022622e-09, "loss": 0.8161, "step": 12692 }, { "epoch": 0.99, "learning_rate": 1.1525015180383623e-09, "loss": 1.2009, "step": 12693 }, { "epoch": 0.99, "learning_rate": 1.1404673195393178e-09, "loss": 1.1822, "step": 12694 }, { "epoch": 0.99, "learning_rate": 1.128496244765742e-09, "loss": 1.2117, "step": 12695 }, { "epoch": 0.99, "learning_rate": 1.11658829447403e-09, "loss": 1.1704, "step": 12696 }, { "epoch": 0.99, "learning_rate": 1.1047434694169134e-09, "loss": 1.1788, "step": 12697 }, { "epoch": 0.99, "learning_rate": 1.0929617703427929e-09, "loss": 1.1492, "step": 12698 }, { "epoch": 0.99, "learning_rate": 1.0812431979962954e-09, "loss": 1.1485, "step": 12699 }, { "epoch": 0.99, "learning_rate": 1.0695877531178287e-09, "loss": 1.207, "step": 12700 }, { "epoch": 0.99, "learning_rate": 1.0579954364442477e-09, "loss": 1.166, "step": 12701 }, { "epoch": 0.99, "learning_rate": 1.0464662487078556e-09, "loss": 1.237, "step": 12702 }, { "epoch": 0.99, "learning_rate": 1.035000190637403e-09, "loss": 1.1359, "step": 12703 }, { "epoch": 0.99, "learning_rate": 1.0235972629575318e-09, "loss": 1.2177, "step": 12704 }, { "epoch": 0.99, "learning_rate": 1.0122574663887773e-09, "loss": 1.1739, "step": 12705 }, { "epoch": 0.99, "learning_rate": 1.0009808016476772e-09, "loss": 1.2107, "step": 12706 }, { "epoch": 0.99, "learning_rate": 9.89767269446995e-10, "loss": 1.1644, "step": 12707 }, { "epoch": 0.99, "learning_rate": 9.786168704954967e-10, "loss": 1.1685, "step": 12708 }, { "epoch": 0.99, "learning_rate": 9.675296054976189e-10, "loss": 1.1893, "step": 12709 }, { "epoch": 0.99, "learning_rate": 9.565054751540236e-10, "loss": 1.1671, "step": 12710 }, { "epoch": 0.99, "learning_rate": 9.455444801615975e-10, "loss": 1.1541, "step": 12711 }, { "epoch": 0.99, "learning_rate": 9.346466212128977e-10, "loss": 0.8123, "step": 12712 }, { "epoch": 0.99, "learning_rate": 9.23811898996485e-10, "loss": 1.208, "step": 12713 }, { "epoch": 0.99, "learning_rate": 9.130403141973664e-10, "loss": 0.8423, "step": 12714 }, { "epoch": 0.99, "learning_rate": 9.023318674961089e-10, "loss": 1.2271, "step": 12715 }, { "epoch": 0.99, "learning_rate": 8.916865595693934e-10, "loss": 1.2173, "step": 12716 }, { "epoch": 0.99, "learning_rate": 8.811043910901261e-10, "loss": 1.179, "step": 12717 }, { "epoch": 0.99, "learning_rate": 8.705853627268833e-10, "loss": 1.2403, "step": 12718 }, { "epoch": 0.99, "learning_rate": 8.601294751443555e-10, "loss": 1.1703, "step": 12719 }, { "epoch": 0.99, "learning_rate": 8.497367290035696e-10, "loss": 1.236, "step": 12720 }, { "epoch": 0.99, "learning_rate": 8.394071249611113e-10, "loss": 1.2336, "step": 12721 }, { "epoch": 0.99, "learning_rate": 8.291406636699027e-10, "loss": 0.8413, "step": 12722 }, { "epoch": 0.99, "learning_rate": 8.189373457786475e-10, "loss": 1.2016, "step": 12723 }, { "epoch": 0.99, "learning_rate": 8.087971719320519e-10, "loss": 1.201, "step": 12724 }, { "epoch": 0.99, "learning_rate": 7.987201427711588e-10, "loss": 1.1654, "step": 12725 }, { "epoch": 0.99, "learning_rate": 7.887062589326809e-10, "loss": 1.1586, "step": 12726 }, { "epoch": 0.99, "learning_rate": 7.787555210494456e-10, "loss": 1.2305, "step": 12727 }, { "epoch": 0.99, "learning_rate": 7.688679297502831e-10, "loss": 1.2048, "step": 12728 }, { "epoch": 0.99, "learning_rate": 7.59043485660138e-10, "loss": 1.2278, "step": 12729 }, { "epoch": 0.99, "learning_rate": 7.492821893997359e-10, "loss": 1.1706, "step": 12730 }, { "epoch": 0.99, "learning_rate": 7.395840415861387e-10, "loss": 1.1574, "step": 12731 }, { "epoch": 0.99, "learning_rate": 7.299490428319677e-10, "loss": 1.177, "step": 12732 }, { "epoch": 0.99, "learning_rate": 7.203771937464021e-10, "loss": 1.1608, "step": 12733 }, { "epoch": 0.99, "learning_rate": 7.108684949340693e-10, "loss": 1.2465, "step": 12734 }, { "epoch": 0.99, "learning_rate": 7.014229469960442e-10, "loss": 1.2269, "step": 12735 }, { "epoch": 0.99, "learning_rate": 6.920405505292936e-10, "loss": 1.1878, "step": 12736 }, { "epoch": 0.99, "learning_rate": 6.827213061265657e-10, "loss": 1.221, "step": 12737 }, { "epoch": 0.99, "learning_rate": 6.734652143769449e-10, "loss": 1.1079, "step": 12738 }, { "epoch": 0.99, "learning_rate": 6.642722758654073e-10, "loss": 1.2113, "step": 12739 }, { "epoch": 0.99, "learning_rate": 6.55142491172711e-10, "loss": 1.1936, "step": 12740 }, { "epoch": 0.99, "learning_rate": 6.460758608759498e-10, "loss": 0.8219, "step": 12741 }, { "epoch": 0.99, "learning_rate": 6.370723855482207e-10, "loss": 1.2115, "step": 12742 }, { "epoch": 0.99, "learning_rate": 6.281320657581801e-10, "loss": 1.2075, "step": 12743 }, { "epoch": 0.99, "learning_rate": 6.192549020711535e-10, "loss": 1.2008, "step": 12744 }, { "epoch": 0.99, "learning_rate": 6.104408950479145e-10, "loss": 1.1852, "step": 12745 }, { "epoch": 0.99, "learning_rate": 6.016900452455731e-10, "loss": 0.8489, "step": 12746 }, { "epoch": 0.99, "learning_rate": 5.930023532171314e-10, "loss": 1.137, "step": 12747 }, { "epoch": 0.99, "learning_rate": 5.843778195117055e-10, "loss": 1.1547, "step": 12748 }, { "epoch": 0.99, "learning_rate": 5.758164446740821e-10, "loss": 1.2051, "step": 12749 }, { "epoch": 0.99, "learning_rate": 5.673182292454948e-10, "loss": 0.8474, "step": 12750 }, { "epoch": 0.99, "learning_rate": 5.588831737629584e-10, "loss": 1.1853, "step": 12751 }, { "epoch": 0.99, "learning_rate": 5.505112787596023e-10, "loss": 1.2522, "step": 12752 }, { "epoch": 0.99, "learning_rate": 5.422025447643364e-10, "loss": 1.1853, "step": 12753 }, { "epoch": 0.99, "learning_rate": 5.339569723024073e-10, "loss": 0.8512, "step": 12754 }, { "epoch": 0.99, "learning_rate": 5.257745618947318e-10, "loss": 0.8584, "step": 12755 }, { "epoch": 0.99, "learning_rate": 5.176553140585627e-10, "loss": 1.1589, "step": 12756 }, { "epoch": 0.99, "learning_rate": 5.09599229306934e-10, "loss": 1.1724, "step": 12757 }, { "epoch": 0.99, "learning_rate": 5.01606308148883e-10, "loss": 1.2397, "step": 12758 }, { "epoch": 0.99, "learning_rate": 4.936765510896724e-10, "loss": 1.2341, "step": 12759 }, { "epoch": 0.99, "learning_rate": 4.858099586302344e-10, "loss": 1.2249, "step": 12760 }, { "epoch": 0.99, "learning_rate": 4.780065312679493e-10, "loss": 1.1282, "step": 12761 }, { "epoch": 0.99, "learning_rate": 4.70266269495867e-10, "loss": 1.1952, "step": 12762 }, { "epoch": 0.99, "learning_rate": 4.625891738029297e-10, "loss": 0.8526, "step": 12763 }, { "epoch": 0.99, "learning_rate": 4.54975244674638e-10, "loss": 1.2068, "step": 12764 }, { "epoch": 0.99, "learning_rate": 4.4742448259194046e-10, "loss": 1.1995, "step": 12765 }, { "epoch": 0.99, "learning_rate": 4.399368880321219e-10, "loss": 1.1884, "step": 12766 }, { "epoch": 0.99, "learning_rate": 4.3251246146824847e-10, "loss": 1.1868, "step": 12767 }, { "epoch": 0.99, "learning_rate": 4.251512033696114e-10, "loss": 1.1682, "step": 12768 }, { "epoch": 0.99, "learning_rate": 4.178531142013941e-10, "loss": 1.1686, "step": 12769 }, { "epoch": 0.99, "learning_rate": 4.106181944247833e-10, "loss": 0.8237, "step": 12770 }, { "epoch": 0.99, "learning_rate": 4.0344644449696874e-10, "loss": 1.2304, "step": 12771 }, { "epoch": 0.99, "learning_rate": 3.9633786487125456e-10, "loss": 1.2116, "step": 12772 }, { "epoch": 0.99, "learning_rate": 3.892924559968369e-10, "loss": 1.191, "step": 12773 }, { "epoch": 0.99, "learning_rate": 3.823102183189153e-10, "loss": 1.2122, "step": 12774 }, { "epoch": 0.99, "learning_rate": 3.7539115227880337e-10, "loss": 1.1289, "step": 12775 }, { "epoch": 0.99, "learning_rate": 3.685352583137069e-10, "loss": 1.1601, "step": 12776 }, { "epoch": 0.99, "learning_rate": 3.61742536856946e-10, "loss": 1.1439, "step": 12777 }, { "epoch": 0.99, "learning_rate": 3.5501298833773287e-10, "loss": 0.8362, "step": 12778 }, { "epoch": 0.99, "learning_rate": 3.483466131812829e-10, "loss": 1.1764, "step": 12779 }, { "epoch": 0.99, "learning_rate": 3.4174341180903675e-10, "loss": 1.1784, "step": 12780 }, { "epoch": 0.99, "learning_rate": 3.352033846382163e-10, "loss": 1.1827, "step": 12781 }, { "epoch": 0.99, "learning_rate": 3.287265320821575e-10, "loss": 1.2186, "step": 12782 }, { "epoch": 0.99, "learning_rate": 3.2231285455019965e-10, "loss": 1.1517, "step": 12783 }, { "epoch": 0.99, "learning_rate": 3.1596235244746307e-10, "loss": 1.2227, "step": 12784 }, { "epoch": 0.99, "learning_rate": 3.0967502617551544e-10, "loss": 1.1395, "step": 12785 }, { "epoch": 0.99, "learning_rate": 3.0345087613148355e-10, "loss": 1.212, "step": 12786 }, { "epoch": 0.99, "learning_rate": 2.972899027088305e-10, "loss": 1.2174, "step": 12787 }, { "epoch": 0.99, "learning_rate": 2.911921062969114e-10, "loss": 1.2074, "step": 12788 }, { "epoch": 0.99, "learning_rate": 2.851574872810847e-10, "loss": 1.1616, "step": 12789 }, { "epoch": 0.99, "learning_rate": 2.79186046042712e-10, "loss": 1.1757, "step": 12790 }, { "epoch": 0.99, "learning_rate": 2.7327778295904715e-10, "loss": 1.2041, "step": 12791 }, { "epoch": 0.99, "learning_rate": 2.674326984035691e-10, "loss": 1.1794, "step": 12792 }, { "epoch": 0.99, "learning_rate": 2.6165079274564905e-10, "loss": 1.1993, "step": 12793 }, { "epoch": 0.99, "learning_rate": 2.5593206635066146e-10, "loss": 1.2347, "step": 12794 }, { "epoch": 0.99, "learning_rate": 2.502765195800949e-10, "loss": 1.1971, "step": 12795 }, { "epoch": 0.99, "learning_rate": 2.4468415279121914e-10, "loss": 1.2224, "step": 12796 }, { "epoch": 0.99, "learning_rate": 2.3915496633741817e-10, "loss": 1.2487, "step": 12797 }, { "epoch": 0.99, "learning_rate": 2.336889605683012e-10, "loss": 1.1966, "step": 12798 }, { "epoch": 0.99, "learning_rate": 2.2828613582925872e-10, "loss": 0.8348, "step": 12799 }, { "epoch": 0.99, "learning_rate": 2.2294649246157315e-10, "loss": 1.1666, "step": 12800 }, { "epoch": 0.99, "learning_rate": 2.176700308028634e-10, "loss": 1.1692, "step": 12801 }, { "epoch": 0.99, "learning_rate": 2.1245675118641836e-10, "loss": 1.1415, "step": 12802 }, { "epoch": 0.99, "learning_rate": 2.0730665394186332e-10, "loss": 1.152, "step": 12803 }, { "epoch": 0.99, "learning_rate": 2.0221973939449356e-10, "loss": 1.206, "step": 12804 }, { "epoch": 0.99, "learning_rate": 1.971960078659407e-10, "loss": 1.1406, "step": 12805 }, { "epoch": 0.99, "learning_rate": 1.9223545967361755e-10, "loss": 1.1434, "step": 12806 }, { "epoch": 0.99, "learning_rate": 1.87338095130829e-10, "loss": 1.1793, "step": 12807 }, { "epoch": 0.99, "learning_rate": 1.825039145474383e-10, "loss": 1.1951, "step": 12808 }, { "epoch": 0.99, "learning_rate": 1.7773291822864578e-10, "loss": 1.2108, "step": 12809 }, { "epoch": 0.99, "learning_rate": 1.7302510647609903e-10, "loss": 1.2038, "step": 12810 }, { "epoch": 0.99, "learning_rate": 1.6838047958722678e-10, "loss": 1.1723, "step": 12811 }, { "epoch": 0.99, "learning_rate": 1.6379903785557204e-10, "loss": 1.2044, "step": 12812 }, { "epoch": 0.99, "learning_rate": 1.5928078157068092e-10, "loss": 1.2065, "step": 12813 }, { "epoch": 0.99, "learning_rate": 1.5482571101821385e-10, "loss": 1.2268, "step": 12814 }, { "epoch": 0.99, "learning_rate": 1.5043382647950131e-10, "loss": 1.1769, "step": 12815 }, { "epoch": 0.99, "learning_rate": 1.461051282322101e-10, "loss": 1.2075, "step": 12816 }, { "epoch": 0.99, "learning_rate": 1.4183961654978816e-10, "loss": 1.1898, "step": 12817 }, { "epoch": 0.99, "learning_rate": 1.3763729170190863e-10, "loss": 1.157, "step": 12818 }, { "epoch": 0.99, "learning_rate": 1.3349815395413688e-10, "loss": 0.8174, "step": 12819 }, { "epoch": 0.99, "learning_rate": 1.2942220356804146e-10, "loss": 0.818, "step": 12820 }, { "epoch": 1.0, "learning_rate": 1.2540944080119408e-10, "loss": 1.1753, "step": 12821 }, { "epoch": 1.0, "learning_rate": 1.214598659070587e-10, "loss": 1.2074, "step": 12822 }, { "epoch": 1.0, "learning_rate": 1.1757347913543548e-10, "loss": 1.2007, "step": 12823 }, { "epoch": 1.0, "learning_rate": 1.1375028073179471e-10, "loss": 1.2146, "step": 12824 }, { "epoch": 1.0, "learning_rate": 1.0999027093783197e-10, "loss": 1.2008, "step": 12825 }, { "epoch": 1.0, "learning_rate": 1.0629344999113499e-10, "loss": 1.2071, "step": 12826 }, { "epoch": 1.0, "learning_rate": 1.0265981812529467e-10, "loss": 1.2061, "step": 12827 }, { "epoch": 1.0, "learning_rate": 9.908937557001618e-11, "loss": 1.186, "step": 12828 }, { "epoch": 1.0, "learning_rate": 9.558212255078579e-11, "loss": 1.1528, "step": 12829 }, { "epoch": 1.0, "learning_rate": 9.213805928942608e-11, "loss": 1.2121, "step": 12830 }, { "epoch": 1.0, "learning_rate": 8.875718600342974e-11, "loss": 1.2143, "step": 12831 }, { "epoch": 1.0, "learning_rate": 8.543950290662572e-11, "loss": 1.1725, "step": 12832 }, { "epoch": 1.0, "learning_rate": 8.21850102085131e-11, "loss": 1.1669, "step": 12833 }, { "epoch": 1.0, "learning_rate": 7.899370811481621e-11, "loss": 1.2046, "step": 12834 }, { "epoch": 1.0, "learning_rate": 7.586559682726256e-11, "loss": 1.1987, "step": 12835 }, { "epoch": 1.0, "learning_rate": 7.280067654347188e-11, "loss": 1.2148, "step": 12836 }, { "epoch": 1.0, "learning_rate": 6.979894745717807e-11, "loss": 1.1664, "step": 12837 }, { "epoch": 1.0, "learning_rate": 6.686040975811824e-11, "loss": 1.1981, "step": 12838 }, { "epoch": 1.0, "learning_rate": 6.398506363181066e-11, "loss": 1.2335, "step": 12839 }, { "epoch": 1.0, "learning_rate": 6.11729092602209e-11, "loss": 1.1152, "step": 12840 }, { "epoch": 1.0, "learning_rate": 5.842394682087359e-11, "loss": 1.21, "step": 12841 }, { "epoch": 1.0, "learning_rate": 5.573817648751866e-11, "loss": 1.1839, "step": 12842 }, { "epoch": 1.0, "learning_rate": 5.3115598430020225e-11, "loss": 1.2049, "step": 12843 }, { "epoch": 1.0, "learning_rate": 5.0556212813912534e-11, "loss": 1.1762, "step": 12844 }, { "epoch": 1.0, "learning_rate": 4.8060019801177135e-11, "loss": 1.2302, "step": 12845 }, { "epoch": 1.0, "learning_rate": 4.562701954935466e-11, "loss": 1.1925, "step": 12846 }, { "epoch": 1.0, "learning_rate": 4.3257212212211016e-11, "loss": 1.2404, "step": 12847 }, { "epoch": 1.0, "learning_rate": 4.09505979396263e-11, "loss": 1.1662, "step": 12848 }, { "epoch": 1.0, "learning_rate": 3.870717687726177e-11, "loss": 1.2157, "step": 12849 }, { "epoch": 1.0, "learning_rate": 3.652694916700394e-11, "loss": 1.1808, "step": 12850 }, { "epoch": 1.0, "learning_rate": 3.440991494663148e-11, "loss": 1.1945, "step": 12851 }, { "epoch": 1.0, "learning_rate": 3.2356074349815244e-11, "loss": 1.2338, "step": 12852 }, { "epoch": 1.0, "learning_rate": 3.03654275063403e-11, "loss": 0.84, "step": 12853 }, { "epoch": 1.0, "learning_rate": 2.8437974542105947e-11, "loss": 1.1537, "step": 12854 }, { "epoch": 1.0, "learning_rate": 2.6573715579014665e-11, "loss": 1.1873, "step": 12855 }, { "epoch": 1.0, "learning_rate": 2.4772650734639077e-11, "loss": 1.2369, "step": 12856 }, { "epoch": 1.0, "learning_rate": 2.3034780122888065e-11, "loss": 1.16, "step": 12857 }, { "epoch": 1.0, "learning_rate": 2.1360103853673704e-11, "loss": 1.1632, "step": 12858 }, { "epoch": 1.0, "learning_rate": 1.9748622032689233e-11, "loss": 0.8478, "step": 12859 }, { "epoch": 1.0, "learning_rate": 1.8200334761964143e-11, "loss": 1.1679, "step": 12860 }, { "epoch": 1.0, "learning_rate": 1.671524213919806e-11, "loss": 1.2476, "step": 12861 }, { "epoch": 1.0, "learning_rate": 1.5293344258204833e-11, "loss": 1.1188, "step": 12862 }, { "epoch": 1.0, "learning_rate": 1.3934641209023545e-11, "loss": 1.2289, "step": 12863 }, { "epoch": 1.0, "learning_rate": 1.2639133077363417e-11, "loss": 1.1885, "step": 12864 }, { "epoch": 1.0, "learning_rate": 1.1406819945047885e-11, "loss": 1.2419, "step": 12865 }, { "epoch": 1.0, "learning_rate": 1.0237701890125628e-11, "loss": 1.1821, "step": 12866 }, { "epoch": 1.0, "learning_rate": 9.131778986426475e-12, "loss": 1.2116, "step": 12867 }, { "epoch": 1.0, "learning_rate": 8.089051303783456e-12, "loss": 1.1866, "step": 12868 }, { "epoch": 1.0, "learning_rate": 7.109518908143819e-12, "loss": 1.1833, "step": 12869 }, { "epoch": 1.0, "learning_rate": 6.193181861458008e-12, "loss": 1.2215, "step": 12870 }, { "epoch": 1.0, "learning_rate": 5.34004022145762e-12, "loss": 1.2036, "step": 12871 }, { "epoch": 1.0, "learning_rate": 4.550094042210517e-12, "loss": 1.2003, "step": 12872 }, { "epoch": 1.0, "learning_rate": 3.823343373565713e-12, "loss": 1.1742, "step": 12873 }, { "epoch": 1.0, "learning_rate": 3.15978826148644e-12, "loss": 1.1624, "step": 12874 }, { "epoch": 1.0, "learning_rate": 2.5594287478281075e-12, "loss": 1.1877, "step": 12875 }, { "epoch": 1.0, "learning_rate": 2.0222648707823865e-12, "loss": 1.1734, "step": 12876 }, { "epoch": 1.0, "learning_rate": 1.5482966639890348e-12, "loss": 1.2328, "step": 12877 }, { "epoch": 1.0, "learning_rate": 1.137524157535097e-12, "loss": 1.2421, "step": 12878 }, { "epoch": 1.0, "learning_rate": 7.899473773997911e-13, "loss": 0.8294, "step": 12879 }, { "epoch": 1.0, "learning_rate": 5.055663454545111e-13, "loss": 0.8049, "step": 12880 }, { "epoch": 1.0, "learning_rate": 2.8438107979589233e-13, "loss": 1.1592, "step": 12881 }, { "epoch": 1.0, "learning_rate": 1.263915944127447e-13, "loss": 1.1691, "step": 12882 }, { "epoch": 1.0, "learning_rate": 3.1597899075030964e-14, "loss": 1.1507, "step": 12883 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 1.2258, "step": 12884 }, { "epoch": 1.0, "step": 12884, "total_flos": 2.37126329333239e+18, "train_loss": 0.34703066483375933, "train_runtime": 117668.8243, "train_samples_per_second": 14.016, "train_steps_per_second": 0.109 } ], "logging_steps": 1.0, "max_steps": 12884, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 1000, "total_flos": 2.37126329333239e+18, "train_batch_size": 2, "trial_name": null, "trial_params": null }