{ "best_metric": null, "best_model_checkpoint": null, "epoch": 4.861601191876421, "global_step": 62000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 9.999921587077551e-05, "loss": 4.5587, "step": 1 }, { "epoch": 0.0, "learning_rate": 9.999843174155101e-05, "loss": 4.5526, "step": 2 }, { "epoch": 0.0, "learning_rate": 9.999764761232652e-05, "loss": 4.0868, "step": 3 }, { "epoch": 0.0, "learning_rate": 9.999686348310202e-05, "loss": 4.0958, "step": 4 }, { "epoch": 0.0, "learning_rate": 9.999607935387752e-05, "loss": 3.8889, "step": 5 }, { "epoch": 0.0, "learning_rate": 9.999529522465303e-05, "loss": 3.8716, "step": 6 }, { "epoch": 0.0, "learning_rate": 9.999451109542853e-05, "loss": 3.858, "step": 7 }, { "epoch": 0.0, "learning_rate": 9.999372696620403e-05, "loss": 3.6864, "step": 8 }, { "epoch": 0.0, "learning_rate": 9.999294283697954e-05, "loss": 3.6482, "step": 9 }, { "epoch": 0.0, "learning_rate": 9.999215870775504e-05, "loss": 3.578, "step": 10 }, { "epoch": 0.0, "learning_rate": 9.999137457853055e-05, "loss": 3.852, "step": 11 }, { "epoch": 0.0, "learning_rate": 9.999059044930605e-05, "loss": 3.6504, "step": 12 }, { "epoch": 0.0, "learning_rate": 9.998980632008155e-05, "loss": 3.5007, "step": 13 }, { "epoch": 0.0, "learning_rate": 9.998902219085706e-05, "loss": 3.6212, "step": 14 }, { "epoch": 0.0, "learning_rate": 9.998823806163256e-05, "loss": 3.6384, "step": 15 }, { "epoch": 0.0, "learning_rate": 9.998745393240808e-05, "loss": 3.621, "step": 16 }, { "epoch": 0.0, "learning_rate": 9.998666980318358e-05, "loss": 3.5207, "step": 17 }, { "epoch": 0.0, "learning_rate": 9.998588567395908e-05, "loss": 3.6126, "step": 18 }, { "epoch": 0.0, "learning_rate": 9.998510154473457e-05, "loss": 3.7293, "step": 19 }, { "epoch": 0.0, "learning_rate": 9.998431741551008e-05, "loss": 3.4815, "step": 20 }, { "epoch": 0.0, "learning_rate": 9.998353328628558e-05, "loss": 3.5952, "step": 21 }, { "epoch": 0.0, "learning_rate": 9.998274915706109e-05, "loss": 4.1334, "step": 22 }, { "epoch": 0.0, "learning_rate": 9.998196502783659e-05, "loss": 3.5209, "step": 23 }, { "epoch": 0.0, "learning_rate": 9.998118089861209e-05, "loss": 3.3834, "step": 24 }, { "epoch": 0.0, "learning_rate": 9.99803967693876e-05, "loss": 3.5079, "step": 25 }, { "epoch": 0.0, "learning_rate": 9.99796126401631e-05, "loss": 4.1679, "step": 26 }, { "epoch": 0.0, "learning_rate": 9.99788285109386e-05, "loss": 3.2866, "step": 27 }, { "epoch": 0.0, "learning_rate": 9.997804438171411e-05, "loss": 3.3633, "step": 28 }, { "epoch": 0.0, "learning_rate": 9.997726025248961e-05, "loss": 4.2479, "step": 29 }, { "epoch": 0.0, "learning_rate": 9.997647612326512e-05, "loss": 3.9563, "step": 30 }, { "epoch": 0.0, "learning_rate": 9.997569199404062e-05, "loss": 4.1227, "step": 31 }, { "epoch": 0.0, "learning_rate": 9.997490786481612e-05, "loss": 3.5163, "step": 32 }, { "epoch": 0.0, "learning_rate": 9.997412373559163e-05, "loss": 3.415, "step": 33 }, { "epoch": 0.0, "learning_rate": 9.997333960636713e-05, "loss": 3.4319, "step": 34 }, { "epoch": 0.0, "learning_rate": 9.997255547714263e-05, "loss": 3.7245, "step": 35 }, { "epoch": 0.0, "learning_rate": 9.997177134791814e-05, "loss": 3.3848, "step": 36 }, { "epoch": 0.0, "learning_rate": 9.997098721869364e-05, "loss": 3.3417, "step": 37 }, { "epoch": 0.0, "learning_rate": 9.997020308946914e-05, "loss": 3.5476, "step": 38 }, { "epoch": 0.0, "learning_rate": 9.996941896024465e-05, "loss": 3.421, "step": 39 }, { "epoch": 0.0, "learning_rate": 9.996863483102015e-05, "loss": 3.6739, "step": 40 }, { "epoch": 0.0, "learning_rate": 9.996785070179566e-05, "loss": 3.5519, "step": 41 }, { "epoch": 0.0, "learning_rate": 9.996706657257117e-05, "loss": 3.6524, "step": 42 }, { "epoch": 0.0, "learning_rate": 9.996628244334668e-05, "loss": 3.3717, "step": 43 }, { "epoch": 0.0, "learning_rate": 9.996549831412218e-05, "loss": 3.7494, "step": 44 }, { "epoch": 0.0, "learning_rate": 9.996471418489768e-05, "loss": 3.3011, "step": 45 }, { "epoch": 0.0, "learning_rate": 9.996393005567319e-05, "loss": 3.558, "step": 46 }, { "epoch": 0.0, "learning_rate": 9.996314592644869e-05, "loss": 3.443, "step": 47 }, { "epoch": 0.0, "learning_rate": 9.99623617972242e-05, "loss": 3.6599, "step": 48 }, { "epoch": 0.0, "learning_rate": 9.99615776679997e-05, "loss": 3.2644, "step": 49 }, { "epoch": 0.0, "learning_rate": 9.99607935387752e-05, "loss": 3.4943, "step": 50 }, { "epoch": 0.0, "learning_rate": 9.99600094095507e-05, "loss": 3.1417, "step": 51 }, { "epoch": 0.0, "learning_rate": 9.995922528032621e-05, "loss": 3.0523, "step": 52 }, { "epoch": 0.0, "learning_rate": 9.995844115110171e-05, "loss": 3.206, "step": 53 }, { "epoch": 0.0, "learning_rate": 9.995765702187722e-05, "loss": 3.3502, "step": 54 }, { "epoch": 0.0, "learning_rate": 9.995687289265271e-05, "loss": 3.3063, "step": 55 }, { "epoch": 0.0, "learning_rate": 9.995608876342821e-05, "loss": 3.3109, "step": 56 }, { "epoch": 0.0, "learning_rate": 9.995530463420371e-05, "loss": 3.8029, "step": 57 }, { "epoch": 0.0, "learning_rate": 9.995452050497922e-05, "loss": 3.4295, "step": 58 }, { "epoch": 0.0, "learning_rate": 9.995373637575472e-05, "loss": 3.8113, "step": 59 }, { "epoch": 0.0, "learning_rate": 9.995295224653023e-05, "loss": 3.4564, "step": 60 }, { "epoch": 0.0, "learning_rate": 9.995216811730573e-05, "loss": 3.6225, "step": 61 }, { "epoch": 0.0, "learning_rate": 9.995138398808123e-05, "loss": 3.5084, "step": 62 }, { "epoch": 0.0, "learning_rate": 9.995059985885674e-05, "loss": 3.2459, "step": 63 }, { "epoch": 0.01, "learning_rate": 9.994981572963224e-05, "loss": 3.3292, "step": 64 }, { "epoch": 0.01, "learning_rate": 9.994903160040774e-05, "loss": 3.2149, "step": 65 }, { "epoch": 0.01, "learning_rate": 9.994824747118325e-05, "loss": 3.4332, "step": 66 }, { "epoch": 0.01, "learning_rate": 9.994746334195876e-05, "loss": 3.5477, "step": 67 }, { "epoch": 0.01, "learning_rate": 9.994667921273427e-05, "loss": 3.0265, "step": 68 }, { "epoch": 0.01, "learning_rate": 9.994589508350977e-05, "loss": 3.1256, "step": 69 }, { "epoch": 0.01, "learning_rate": 9.994511095428528e-05, "loss": 3.2467, "step": 70 }, { "epoch": 0.01, "learning_rate": 9.994432682506078e-05, "loss": 3.2977, "step": 71 }, { "epoch": 0.01, "learning_rate": 9.994354269583628e-05, "loss": 3.2248, "step": 72 }, { "epoch": 0.01, "learning_rate": 9.994275856661179e-05, "loss": 3.5557, "step": 73 }, { "epoch": 0.01, "learning_rate": 9.994197443738729e-05, "loss": 3.4457, "step": 74 }, { "epoch": 0.01, "learning_rate": 9.99411903081628e-05, "loss": 3.2738, "step": 75 }, { "epoch": 0.01, "learning_rate": 9.99404061789383e-05, "loss": 3.48, "step": 76 }, { "epoch": 0.01, "learning_rate": 9.99396220497138e-05, "loss": 3.1733, "step": 77 }, { "epoch": 0.01, "learning_rate": 9.99388379204893e-05, "loss": 3.1949, "step": 78 }, { "epoch": 0.01, "learning_rate": 9.993805379126481e-05, "loss": 3.3129, "step": 79 }, { "epoch": 0.01, "learning_rate": 9.993726966204031e-05, "loss": 3.3222, "step": 80 }, { "epoch": 0.01, "learning_rate": 9.993648553281582e-05, "loss": 3.5299, "step": 81 }, { "epoch": 0.01, "learning_rate": 9.993570140359132e-05, "loss": 3.5908, "step": 82 }, { "epoch": 0.01, "learning_rate": 9.993491727436682e-05, "loss": 3.3547, "step": 83 }, { "epoch": 0.01, "learning_rate": 9.993413314514233e-05, "loss": 3.3336, "step": 84 }, { "epoch": 0.01, "learning_rate": 9.993334901591783e-05, "loss": 3.7887, "step": 85 }, { "epoch": 0.01, "learning_rate": 9.993256488669333e-05, "loss": 3.4899, "step": 86 }, { "epoch": 0.01, "learning_rate": 9.993178075746884e-05, "loss": 3.3542, "step": 87 }, { "epoch": 0.01, "learning_rate": 9.993099662824434e-05, "loss": 3.2486, "step": 88 }, { "epoch": 0.01, "learning_rate": 9.993021249901985e-05, "loss": 3.6982, "step": 89 }, { "epoch": 0.01, "learning_rate": 9.992942836979535e-05, "loss": 3.3127, "step": 90 }, { "epoch": 0.01, "learning_rate": 9.992864424057084e-05, "loss": 3.2811, "step": 91 }, { "epoch": 0.01, "learning_rate": 9.992786011134634e-05, "loss": 3.5652, "step": 92 }, { "epoch": 0.01, "learning_rate": 9.992707598212186e-05, "loss": 3.3731, "step": 93 }, { "epoch": 0.01, "learning_rate": 9.992629185289736e-05, "loss": 3.243, "step": 94 }, { "epoch": 0.01, "learning_rate": 9.992550772367287e-05, "loss": 3.5168, "step": 95 }, { "epoch": 0.01, "learning_rate": 9.992472359444837e-05, "loss": 3.2033, "step": 96 }, { "epoch": 0.01, "learning_rate": 9.992393946522388e-05, "loss": 3.4188, "step": 97 }, { "epoch": 0.01, "learning_rate": 9.992315533599938e-05, "loss": 3.5018, "step": 98 }, { "epoch": 0.01, "learning_rate": 9.992237120677488e-05, "loss": 3.1196, "step": 99 }, { "epoch": 0.01, "learning_rate": 9.992158707755039e-05, "loss": 3.4474, "step": 100 }, { "epoch": 0.01, "learning_rate": 9.992080294832589e-05, "loss": 3.2613, "step": 101 }, { "epoch": 0.01, "learning_rate": 9.99200188191014e-05, "loss": 3.4165, "step": 102 }, { "epoch": 0.01, "learning_rate": 9.99192346898769e-05, "loss": 3.1579, "step": 103 }, { "epoch": 0.01, "learning_rate": 9.99184505606524e-05, "loss": 3.3471, "step": 104 }, { "epoch": 0.01, "learning_rate": 9.99176664314279e-05, "loss": 3.4614, "step": 105 }, { "epoch": 0.01, "learning_rate": 9.991688230220341e-05, "loss": 3.3569, "step": 106 }, { "epoch": 0.01, "learning_rate": 9.991609817297891e-05, "loss": 3.4364, "step": 107 }, { "epoch": 0.01, "learning_rate": 9.991531404375442e-05, "loss": 3.2218, "step": 108 }, { "epoch": 0.01, "learning_rate": 9.991452991452992e-05, "loss": 3.2968, "step": 109 }, { "epoch": 0.01, "learning_rate": 9.991374578530542e-05, "loss": 3.2186, "step": 110 }, { "epoch": 0.01, "learning_rate": 9.991296165608093e-05, "loss": 3.2744, "step": 111 }, { "epoch": 0.01, "learning_rate": 9.991217752685643e-05, "loss": 3.5901, "step": 112 }, { "epoch": 0.01, "learning_rate": 9.991139339763193e-05, "loss": 3.2115, "step": 113 }, { "epoch": 0.01, "learning_rate": 9.991060926840744e-05, "loss": 3.1892, "step": 114 }, { "epoch": 0.01, "learning_rate": 9.990982513918294e-05, "loss": 3.4538, "step": 115 }, { "epoch": 0.01, "learning_rate": 9.990904100995844e-05, "loss": 3.2125, "step": 116 }, { "epoch": 0.01, "learning_rate": 9.990825688073395e-05, "loss": 3.3613, "step": 117 }, { "epoch": 0.01, "learning_rate": 9.990747275150945e-05, "loss": 3.4072, "step": 118 }, { "epoch": 0.01, "learning_rate": 9.990668862228496e-05, "loss": 3.2942, "step": 119 }, { "epoch": 0.01, "learning_rate": 9.990590449306046e-05, "loss": 3.3628, "step": 120 }, { "epoch": 0.01, "learning_rate": 9.990512036383596e-05, "loss": 3.1659, "step": 121 }, { "epoch": 0.01, "learning_rate": 9.990433623461147e-05, "loss": 3.1054, "step": 122 }, { "epoch": 0.01, "learning_rate": 9.990355210538697e-05, "loss": 3.1056, "step": 123 }, { "epoch": 0.01, "learning_rate": 9.990276797616249e-05, "loss": 3.2102, "step": 124 }, { "epoch": 0.01, "learning_rate": 9.990198384693799e-05, "loss": 3.4629, "step": 125 }, { "epoch": 0.01, "learning_rate": 9.99011997177135e-05, "loss": 3.3359, "step": 126 }, { "epoch": 0.01, "learning_rate": 9.990041558848899e-05, "loss": 3.4963, "step": 127 }, { "epoch": 0.01, "learning_rate": 9.989963145926449e-05, "loss": 3.2818, "step": 128 }, { "epoch": 0.01, "learning_rate": 9.989884733003999e-05, "loss": 3.5004, "step": 129 }, { "epoch": 0.01, "learning_rate": 9.98980632008155e-05, "loss": 3.3489, "step": 130 }, { "epoch": 0.01, "learning_rate": 9.9897279071591e-05, "loss": 3.4668, "step": 131 }, { "epoch": 0.01, "learning_rate": 9.98964949423665e-05, "loss": 3.2639, "step": 132 }, { "epoch": 0.01, "learning_rate": 9.989571081314201e-05, "loss": 3.3675, "step": 133 }, { "epoch": 0.01, "learning_rate": 9.989492668391751e-05, "loss": 3.5403, "step": 134 }, { "epoch": 0.01, "learning_rate": 9.989414255469301e-05, "loss": 3.4307, "step": 135 }, { "epoch": 0.01, "learning_rate": 9.989335842546852e-05, "loss": 3.1134, "step": 136 }, { "epoch": 0.01, "learning_rate": 9.989257429624402e-05, "loss": 3.2516, "step": 137 }, { "epoch": 0.01, "learning_rate": 9.989179016701953e-05, "loss": 3.3567, "step": 138 }, { "epoch": 0.01, "learning_rate": 9.989100603779503e-05, "loss": 3.1397, "step": 139 }, { "epoch": 0.01, "learning_rate": 9.989022190857053e-05, "loss": 3.1582, "step": 140 }, { "epoch": 0.01, "learning_rate": 9.988943777934604e-05, "loss": 3.2829, "step": 141 }, { "epoch": 0.01, "learning_rate": 9.988865365012154e-05, "loss": 3.4806, "step": 142 }, { "epoch": 0.01, "learning_rate": 9.988786952089704e-05, "loss": 3.1337, "step": 143 }, { "epoch": 0.01, "learning_rate": 9.988708539167255e-05, "loss": 3.0747, "step": 144 }, { "epoch": 0.01, "learning_rate": 9.988630126244805e-05, "loss": 3.3292, "step": 145 }, { "epoch": 0.01, "learning_rate": 9.988551713322356e-05, "loss": 3.413, "step": 146 }, { "epoch": 0.01, "learning_rate": 9.988473300399906e-05, "loss": 3.3299, "step": 147 }, { "epoch": 0.01, "learning_rate": 9.988394887477456e-05, "loss": 3.2473, "step": 148 }, { "epoch": 0.01, "learning_rate": 9.988316474555007e-05, "loss": 3.8484, "step": 149 }, { "epoch": 0.01, "learning_rate": 9.988238061632558e-05, "loss": 3.2123, "step": 150 }, { "epoch": 0.01, "learning_rate": 9.988159648710109e-05, "loss": 3.3549, "step": 151 }, { "epoch": 0.01, "learning_rate": 9.988081235787659e-05, "loss": 3.3187, "step": 152 }, { "epoch": 0.01, "learning_rate": 9.98800282286521e-05, "loss": 3.2167, "step": 153 }, { "epoch": 0.01, "learning_rate": 9.98792440994276e-05, "loss": 3.4534, "step": 154 }, { "epoch": 0.01, "learning_rate": 9.98784599702031e-05, "loss": 3.4777, "step": 155 }, { "epoch": 0.01, "learning_rate": 9.98776758409786e-05, "loss": 3.3044, "step": 156 }, { "epoch": 0.01, "learning_rate": 9.987689171175411e-05, "loss": 3.5328, "step": 157 }, { "epoch": 0.01, "learning_rate": 9.987610758252961e-05, "loss": 3.11, "step": 158 }, { "epoch": 0.01, "learning_rate": 9.987532345330512e-05, "loss": 2.9752, "step": 159 }, { "epoch": 0.01, "learning_rate": 9.987453932408062e-05, "loss": 3.0961, "step": 160 }, { "epoch": 0.01, "learning_rate": 9.987375519485612e-05, "loss": 3.3432, "step": 161 }, { "epoch": 0.01, "learning_rate": 9.987297106563163e-05, "loss": 3.1348, "step": 162 }, { "epoch": 0.01, "learning_rate": 9.987218693640712e-05, "loss": 3.4401, "step": 163 }, { "epoch": 0.01, "learning_rate": 9.987140280718262e-05, "loss": 3.1561, "step": 164 }, { "epoch": 0.01, "learning_rate": 9.987061867795813e-05, "loss": 3.1357, "step": 165 }, { "epoch": 0.01, "learning_rate": 9.986983454873363e-05, "loss": 3.2044, "step": 166 }, { "epoch": 0.01, "learning_rate": 9.986905041950913e-05, "loss": 3.1009, "step": 167 }, { "epoch": 0.01, "learning_rate": 9.986826629028464e-05, "loss": 3.1925, "step": 168 }, { "epoch": 0.01, "learning_rate": 9.986748216106014e-05, "loss": 3.4686, "step": 169 }, { "epoch": 0.01, "learning_rate": 9.986669803183564e-05, "loss": 3.0748, "step": 170 }, { "epoch": 0.01, "learning_rate": 9.986591390261115e-05, "loss": 3.2755, "step": 171 }, { "epoch": 0.01, "learning_rate": 9.986512977338665e-05, "loss": 3.2225, "step": 172 }, { "epoch": 0.01, "learning_rate": 9.986434564416215e-05, "loss": 3.0934, "step": 173 }, { "epoch": 0.01, "learning_rate": 9.986356151493766e-05, "loss": 3.2184, "step": 174 }, { "epoch": 0.01, "learning_rate": 9.986277738571318e-05, "loss": 3.383, "step": 175 }, { "epoch": 0.01, "learning_rate": 9.986199325648868e-05, "loss": 3.1604, "step": 176 }, { "epoch": 0.01, "learning_rate": 9.986120912726418e-05, "loss": 3.25, "step": 177 }, { "epoch": 0.01, "learning_rate": 9.986042499803969e-05, "loss": 3.1886, "step": 178 }, { "epoch": 0.01, "learning_rate": 9.985964086881519e-05, "loss": 3.165, "step": 179 }, { "epoch": 0.01, "learning_rate": 9.98588567395907e-05, "loss": 3.1203, "step": 180 }, { "epoch": 0.01, "learning_rate": 9.98580726103662e-05, "loss": 3.2279, "step": 181 }, { "epoch": 0.01, "learning_rate": 9.98572884811417e-05, "loss": 3.3475, "step": 182 }, { "epoch": 0.01, "learning_rate": 9.98565043519172e-05, "loss": 3.13, "step": 183 }, { "epoch": 0.01, "learning_rate": 9.985572022269271e-05, "loss": 3.0981, "step": 184 }, { "epoch": 0.01, "learning_rate": 9.985493609346821e-05, "loss": 3.3635, "step": 185 }, { "epoch": 0.01, "learning_rate": 9.985415196424372e-05, "loss": 3.246, "step": 186 }, { "epoch": 0.01, "learning_rate": 9.985336783501922e-05, "loss": 3.1072, "step": 187 }, { "epoch": 0.01, "learning_rate": 9.985258370579472e-05, "loss": 3.2546, "step": 188 }, { "epoch": 0.01, "learning_rate": 9.985179957657023e-05, "loss": 3.2991, "step": 189 }, { "epoch": 0.01, "learning_rate": 9.985101544734573e-05, "loss": 3.1063, "step": 190 }, { "epoch": 0.01, "learning_rate": 9.985023131812123e-05, "loss": 3.3048, "step": 191 }, { "epoch": 0.02, "learning_rate": 9.984944718889674e-05, "loss": 3.2552, "step": 192 }, { "epoch": 0.02, "learning_rate": 9.984866305967224e-05, "loss": 3.4441, "step": 193 }, { "epoch": 0.02, "learning_rate": 9.984787893044775e-05, "loss": 3.1298, "step": 194 }, { "epoch": 0.02, "learning_rate": 9.984709480122325e-05, "loss": 3.2192, "step": 195 }, { "epoch": 0.02, "learning_rate": 9.984631067199875e-05, "loss": 3.0968, "step": 196 }, { "epoch": 0.02, "learning_rate": 9.984552654277426e-05, "loss": 3.3139, "step": 197 }, { "epoch": 0.02, "learning_rate": 9.984474241354976e-05, "loss": 3.4616, "step": 198 }, { "epoch": 0.02, "learning_rate": 9.984395828432525e-05, "loss": 3.281, "step": 199 }, { "epoch": 0.02, "learning_rate": 9.984317415510075e-05, "loss": 3.0801, "step": 200 }, { "epoch": 0.02, "learning_rate": 9.984239002587627e-05, "loss": 3.1771, "step": 201 }, { "epoch": 0.02, "learning_rate": 9.984160589665177e-05, "loss": 3.1777, "step": 202 }, { "epoch": 0.02, "learning_rate": 9.984082176742728e-05, "loss": 3.7293, "step": 203 }, { "epoch": 0.02, "learning_rate": 9.984003763820278e-05, "loss": 3.4512, "step": 204 }, { "epoch": 0.02, "learning_rate": 9.983925350897829e-05, "loss": 3.1495, "step": 205 }, { "epoch": 0.02, "learning_rate": 9.983846937975379e-05, "loss": 3.4093, "step": 206 }, { "epoch": 0.02, "learning_rate": 9.983768525052929e-05, "loss": 3.1642, "step": 207 }, { "epoch": 0.02, "learning_rate": 9.98369011213048e-05, "loss": 3.3563, "step": 208 }, { "epoch": 0.02, "learning_rate": 9.98361169920803e-05, "loss": 3.1526, "step": 209 }, { "epoch": 0.02, "learning_rate": 9.98353328628558e-05, "loss": 3.2614, "step": 210 }, { "epoch": 0.02, "learning_rate": 9.983454873363131e-05, "loss": 3.2499, "step": 211 }, { "epoch": 0.02, "learning_rate": 9.983376460440681e-05, "loss": 3.0497, "step": 212 }, { "epoch": 0.02, "learning_rate": 9.983298047518232e-05, "loss": 3.2463, "step": 213 }, { "epoch": 0.02, "learning_rate": 9.983219634595782e-05, "loss": 2.9783, "step": 214 }, { "epoch": 0.02, "learning_rate": 9.983141221673332e-05, "loss": 3.1121, "step": 215 }, { "epoch": 0.02, "learning_rate": 9.983062808750883e-05, "loss": 3.2371, "step": 216 }, { "epoch": 0.02, "learning_rate": 9.982984395828433e-05, "loss": 3.2329, "step": 217 }, { "epoch": 0.02, "learning_rate": 9.982905982905983e-05, "loss": 3.4202, "step": 218 }, { "epoch": 0.02, "learning_rate": 9.982827569983534e-05, "loss": 3.2389, "step": 219 }, { "epoch": 0.02, "learning_rate": 9.982749157061084e-05, "loss": 3.2535, "step": 220 }, { "epoch": 0.02, "learning_rate": 9.982670744138634e-05, "loss": 3.1621, "step": 221 }, { "epoch": 0.02, "learning_rate": 9.982592331216185e-05, "loss": 3.2331, "step": 222 }, { "epoch": 0.02, "learning_rate": 9.982513918293735e-05, "loss": 3.1246, "step": 223 }, { "epoch": 0.02, "learning_rate": 9.982435505371286e-05, "loss": 3.3528, "step": 224 }, { "epoch": 0.02, "learning_rate": 9.982357092448836e-05, "loss": 3.078, "step": 225 }, { "epoch": 0.02, "learning_rate": 9.982278679526386e-05, "loss": 3.1201, "step": 226 }, { "epoch": 0.02, "learning_rate": 9.982200266603937e-05, "loss": 3.3146, "step": 227 }, { "epoch": 0.02, "learning_rate": 9.982121853681487e-05, "loss": 3.1867, "step": 228 }, { "epoch": 0.02, "learning_rate": 9.982043440759037e-05, "loss": 3.4867, "step": 229 }, { "epoch": 0.02, "learning_rate": 9.981965027836588e-05, "loss": 3.5642, "step": 230 }, { "epoch": 0.02, "learning_rate": 9.981886614914138e-05, "loss": 3.1147, "step": 231 }, { "epoch": 0.02, "learning_rate": 9.98180820199169e-05, "loss": 3.2983, "step": 232 }, { "epoch": 0.02, "learning_rate": 9.98172978906924e-05, "loss": 3.3659, "step": 233 }, { "epoch": 0.02, "learning_rate": 9.98165137614679e-05, "loss": 3.2093, "step": 234 }, { "epoch": 0.02, "learning_rate": 9.98157296322434e-05, "loss": 3.2085, "step": 235 }, { "epoch": 0.02, "learning_rate": 9.98149455030189e-05, "loss": 3.1919, "step": 236 }, { "epoch": 0.02, "learning_rate": 9.98141613737944e-05, "loss": 3.1016, "step": 237 }, { "epoch": 0.02, "learning_rate": 9.981337724456991e-05, "loss": 3.0301, "step": 238 }, { "epoch": 0.02, "learning_rate": 9.981259311534541e-05, "loss": 3.2536, "step": 239 }, { "epoch": 0.02, "learning_rate": 9.981180898612091e-05, "loss": 3.2552, "step": 240 }, { "epoch": 0.02, "learning_rate": 9.981102485689642e-05, "loss": 3.4683, "step": 241 }, { "epoch": 0.02, "learning_rate": 9.981024072767192e-05, "loss": 3.1723, "step": 242 }, { "epoch": 0.02, "learning_rate": 9.980945659844743e-05, "loss": 3.1805, "step": 243 }, { "epoch": 0.02, "learning_rate": 9.980867246922293e-05, "loss": 3.3012, "step": 244 }, { "epoch": 0.02, "learning_rate": 9.980788833999843e-05, "loss": 3.3781, "step": 245 }, { "epoch": 0.02, "learning_rate": 9.980710421077394e-05, "loss": 3.3793, "step": 246 }, { "epoch": 0.02, "learning_rate": 9.980632008154944e-05, "loss": 3.3194, "step": 247 }, { "epoch": 0.02, "learning_rate": 9.980553595232494e-05, "loss": 3.0169, "step": 248 }, { "epoch": 0.02, "learning_rate": 9.980475182310045e-05, "loss": 3.2772, "step": 249 }, { "epoch": 0.02, "learning_rate": 9.980396769387595e-05, "loss": 3.2053, "step": 250 }, { "epoch": 0.02, "learning_rate": 9.980318356465145e-05, "loss": 3.05, "step": 251 }, { "epoch": 0.02, "learning_rate": 9.980239943542696e-05, "loss": 3.2202, "step": 252 }, { "epoch": 0.02, "learning_rate": 9.980161530620246e-05, "loss": 3.1087, "step": 253 }, { "epoch": 0.02, "learning_rate": 9.980083117697797e-05, "loss": 3.2644, "step": 254 }, { "epoch": 0.02, "learning_rate": 9.980004704775347e-05, "loss": 3.4772, "step": 255 }, { "epoch": 0.02, "learning_rate": 9.979926291852897e-05, "loss": 3.0754, "step": 256 }, { "epoch": 0.02, "learning_rate": 9.979847878930448e-05, "loss": 3.1451, "step": 257 }, { "epoch": 0.02, "learning_rate": 9.979769466008e-05, "loss": 3.1574, "step": 258 }, { "epoch": 0.02, "learning_rate": 9.97969105308555e-05, "loss": 3.2629, "step": 259 }, { "epoch": 0.02, "learning_rate": 9.9796126401631e-05, "loss": 3.2131, "step": 260 }, { "epoch": 0.02, "learning_rate": 9.97953422724065e-05, "loss": 3.016, "step": 261 }, { "epoch": 0.02, "learning_rate": 9.979455814318201e-05, "loss": 3.3198, "step": 262 }, { "epoch": 0.02, "learning_rate": 9.979377401395751e-05, "loss": 3.1919, "step": 263 }, { "epoch": 0.02, "learning_rate": 9.979298988473302e-05, "loss": 3.0704, "step": 264 }, { "epoch": 0.02, "learning_rate": 9.979220575550852e-05, "loss": 3.1108, "step": 265 }, { "epoch": 0.02, "learning_rate": 9.979142162628402e-05, "loss": 3.3748, "step": 266 }, { "epoch": 0.02, "learning_rate": 9.979063749705953e-05, "loss": 3.0727, "step": 267 }, { "epoch": 0.02, "learning_rate": 9.978985336783503e-05, "loss": 3.1461, "step": 268 }, { "epoch": 0.02, "learning_rate": 9.978906923861053e-05, "loss": 3.1679, "step": 269 }, { "epoch": 0.02, "learning_rate": 9.978828510938604e-05, "loss": 3.0536, "step": 270 }, { "epoch": 0.02, "learning_rate": 9.978750098016153e-05, "loss": 3.5706, "step": 271 }, { "epoch": 0.02, "learning_rate": 9.978671685093703e-05, "loss": 3.3053, "step": 272 }, { "epoch": 0.02, "learning_rate": 9.978593272171254e-05, "loss": 3.3587, "step": 273 }, { "epoch": 0.02, "learning_rate": 9.978514859248804e-05, "loss": 3.3171, "step": 274 }, { "epoch": 0.02, "learning_rate": 9.978436446326354e-05, "loss": 3.2251, "step": 275 }, { "epoch": 0.02, "learning_rate": 9.978358033403905e-05, "loss": 3.1427, "step": 276 }, { "epoch": 0.02, "learning_rate": 9.978279620481455e-05, "loss": 3.3991, "step": 277 }, { "epoch": 0.02, "learning_rate": 9.978201207559005e-05, "loss": 3.2993, "step": 278 }, { "epoch": 0.02, "learning_rate": 9.978122794636556e-05, "loss": 3.1485, "step": 279 }, { "epoch": 0.02, "learning_rate": 9.978044381714106e-05, "loss": 3.374, "step": 280 }, { "epoch": 0.02, "learning_rate": 9.977965968791657e-05, "loss": 3.2076, "step": 281 }, { "epoch": 0.02, "learning_rate": 9.977887555869207e-05, "loss": 3.23, "step": 282 }, { "epoch": 0.02, "learning_rate": 9.977809142946759e-05, "loss": 3.3611, "step": 283 }, { "epoch": 0.02, "learning_rate": 9.977730730024309e-05, "loss": 3.1782, "step": 284 }, { "epoch": 0.02, "learning_rate": 9.97765231710186e-05, "loss": 3.2457, "step": 285 }, { "epoch": 0.02, "learning_rate": 9.97757390417941e-05, "loss": 3.2661, "step": 286 }, { "epoch": 0.02, "learning_rate": 9.97749549125696e-05, "loss": 3.0601, "step": 287 }, { "epoch": 0.02, "learning_rate": 9.97741707833451e-05, "loss": 3.1283, "step": 288 }, { "epoch": 0.02, "learning_rate": 9.977338665412061e-05, "loss": 3.2584, "step": 289 }, { "epoch": 0.02, "learning_rate": 9.977260252489611e-05, "loss": 3.0931, "step": 290 }, { "epoch": 0.02, "learning_rate": 9.977181839567162e-05, "loss": 3.4919, "step": 291 }, { "epoch": 0.02, "learning_rate": 9.977103426644712e-05, "loss": 3.2826, "step": 292 }, { "epoch": 0.02, "learning_rate": 9.977025013722262e-05, "loss": 3.2102, "step": 293 }, { "epoch": 0.02, "learning_rate": 9.976946600799813e-05, "loss": 3.3619, "step": 294 }, { "epoch": 0.02, "learning_rate": 9.976868187877363e-05, "loss": 3.262, "step": 295 }, { "epoch": 0.02, "learning_rate": 9.976789774954913e-05, "loss": 3.1788, "step": 296 }, { "epoch": 0.02, "learning_rate": 9.976711362032464e-05, "loss": 3.1337, "step": 297 }, { "epoch": 0.02, "learning_rate": 9.976632949110014e-05, "loss": 3.0773, "step": 298 }, { "epoch": 0.02, "learning_rate": 9.976554536187565e-05, "loss": 3.2682, "step": 299 }, { "epoch": 0.02, "learning_rate": 9.976476123265115e-05, "loss": 2.9866, "step": 300 }, { "epoch": 0.02, "learning_rate": 9.976397710342665e-05, "loss": 3.1279, "step": 301 }, { "epoch": 0.02, "learning_rate": 9.976319297420216e-05, "loss": 3.0932, "step": 302 }, { "epoch": 0.02, "learning_rate": 9.976240884497766e-05, "loss": 3.0993, "step": 303 }, { "epoch": 0.02, "learning_rate": 9.976162471575316e-05, "loss": 3.4315, "step": 304 }, { "epoch": 0.02, "learning_rate": 9.976084058652867e-05, "loss": 3.2992, "step": 305 }, { "epoch": 0.02, "learning_rate": 9.976005645730417e-05, "loss": 3.1642, "step": 306 }, { "epoch": 0.02, "learning_rate": 9.975927232807966e-05, "loss": 3.1196, "step": 307 }, { "epoch": 0.02, "learning_rate": 9.975848819885516e-05, "loss": 3.2622, "step": 308 }, { "epoch": 0.02, "learning_rate": 9.975770406963068e-05, "loss": 3.4059, "step": 309 }, { "epoch": 0.02, "learning_rate": 9.975691994040619e-05, "loss": 3.2859, "step": 310 }, { "epoch": 0.02, "learning_rate": 9.975613581118169e-05, "loss": 3.1859, "step": 311 }, { "epoch": 0.02, "learning_rate": 9.975535168195719e-05, "loss": 3.2201, "step": 312 }, { "epoch": 0.02, "learning_rate": 9.97545675527327e-05, "loss": 3.1841, "step": 313 }, { "epoch": 0.02, "learning_rate": 9.97537834235082e-05, "loss": 3.2146, "step": 314 }, { "epoch": 0.02, "learning_rate": 9.97529992942837e-05, "loss": 3.0467, "step": 315 }, { "epoch": 0.02, "learning_rate": 9.975221516505921e-05, "loss": 3.19, "step": 316 }, { "epoch": 0.02, "learning_rate": 9.975143103583471e-05, "loss": 3.4574, "step": 317 }, { "epoch": 0.02, "learning_rate": 9.975064690661021e-05, "loss": 2.8854, "step": 318 }, { "epoch": 0.03, "learning_rate": 9.974986277738572e-05, "loss": 3.2093, "step": 319 }, { "epoch": 0.03, "learning_rate": 9.974907864816122e-05, "loss": 3.2191, "step": 320 }, { "epoch": 0.03, "learning_rate": 9.974829451893673e-05, "loss": 3.0179, "step": 321 }, { "epoch": 0.03, "learning_rate": 9.974751038971223e-05, "loss": 3.1992, "step": 322 }, { "epoch": 0.03, "learning_rate": 9.974672626048773e-05, "loss": 3.0536, "step": 323 }, { "epoch": 0.03, "learning_rate": 9.974594213126324e-05, "loss": 3.289, "step": 324 }, { "epoch": 0.03, "learning_rate": 9.974515800203874e-05, "loss": 3.1943, "step": 325 }, { "epoch": 0.03, "learning_rate": 9.974437387281424e-05, "loss": 3.2183, "step": 326 }, { "epoch": 0.03, "learning_rate": 9.974358974358975e-05, "loss": 3.0698, "step": 327 }, { "epoch": 0.03, "learning_rate": 9.974280561436525e-05, "loss": 3.144, "step": 328 }, { "epoch": 0.03, "learning_rate": 9.974202148514076e-05, "loss": 3.1267, "step": 329 }, { "epoch": 0.03, "learning_rate": 9.974123735591626e-05, "loss": 3.1207, "step": 330 }, { "epoch": 0.03, "learning_rate": 9.974045322669176e-05, "loss": 2.9566, "step": 331 }, { "epoch": 0.03, "learning_rate": 9.973966909746727e-05, "loss": 3.2771, "step": 332 }, { "epoch": 0.03, "learning_rate": 9.973888496824277e-05, "loss": 3.0969, "step": 333 }, { "epoch": 0.03, "learning_rate": 9.973810083901827e-05, "loss": 3.0479, "step": 334 }, { "epoch": 0.03, "learning_rate": 9.973731670979378e-05, "loss": 3.2065, "step": 335 }, { "epoch": 0.03, "learning_rate": 9.973653258056928e-05, "loss": 3.2985, "step": 336 }, { "epoch": 0.03, "learning_rate": 9.973574845134478e-05, "loss": 3.133, "step": 337 }, { "epoch": 0.03, "learning_rate": 9.973496432212029e-05, "loss": 3.111, "step": 338 }, { "epoch": 0.03, "learning_rate": 9.973418019289579e-05, "loss": 3.2693, "step": 339 }, { "epoch": 0.03, "learning_rate": 9.973339606367131e-05, "loss": 3.0016, "step": 340 }, { "epoch": 0.03, "learning_rate": 9.973261193444681e-05, "loss": 3.2326, "step": 341 }, { "epoch": 0.03, "learning_rate": 9.973182780522232e-05, "loss": 3.1093, "step": 342 }, { "epoch": 0.03, "learning_rate": 9.973104367599781e-05, "loss": 3.2998, "step": 343 }, { "epoch": 0.03, "learning_rate": 9.973025954677331e-05, "loss": 3.1511, "step": 344 }, { "epoch": 0.03, "learning_rate": 9.972947541754881e-05, "loss": 3.2061, "step": 345 }, { "epoch": 0.03, "learning_rate": 9.972869128832432e-05, "loss": 3.1103, "step": 346 }, { "epoch": 0.03, "learning_rate": 9.972790715909982e-05, "loss": 3.2466, "step": 347 }, { "epoch": 0.03, "learning_rate": 9.972712302987533e-05, "loss": 3.1636, "step": 348 }, { "epoch": 0.03, "learning_rate": 9.972633890065083e-05, "loss": 3.2079, "step": 349 }, { "epoch": 0.03, "learning_rate": 9.972555477142633e-05, "loss": 3.2777, "step": 350 }, { "epoch": 0.03, "learning_rate": 9.972477064220184e-05, "loss": 3.3092, "step": 351 }, { "epoch": 0.03, "learning_rate": 9.972398651297734e-05, "loss": 3.0783, "step": 352 }, { "epoch": 0.03, "learning_rate": 9.972320238375284e-05, "loss": 3.2828, "step": 353 }, { "epoch": 0.03, "learning_rate": 9.972241825452835e-05, "loss": 3.4592, "step": 354 }, { "epoch": 0.03, "learning_rate": 9.972163412530385e-05, "loss": 3.1111, "step": 355 }, { "epoch": 0.03, "learning_rate": 9.972084999607935e-05, "loss": 3.327, "step": 356 }, { "epoch": 0.03, "learning_rate": 9.972006586685486e-05, "loss": 3.1319, "step": 357 }, { "epoch": 0.03, "learning_rate": 9.971928173763036e-05, "loss": 3.0718, "step": 358 }, { "epoch": 0.03, "learning_rate": 9.971849760840587e-05, "loss": 3.1439, "step": 359 }, { "epoch": 0.03, "learning_rate": 9.971771347918137e-05, "loss": 3.0228, "step": 360 }, { "epoch": 0.03, "learning_rate": 9.971692934995687e-05, "loss": 3.1757, "step": 361 }, { "epoch": 0.03, "learning_rate": 9.971614522073238e-05, "loss": 3.1092, "step": 362 }, { "epoch": 0.03, "learning_rate": 9.971536109150788e-05, "loss": 3.295, "step": 363 }, { "epoch": 0.03, "learning_rate": 9.971457696228338e-05, "loss": 3.2472, "step": 364 }, { "epoch": 0.03, "learning_rate": 9.971379283305889e-05, "loss": 3.1183, "step": 365 }, { "epoch": 0.03, "learning_rate": 9.97130087038344e-05, "loss": 3.3759, "step": 366 }, { "epoch": 0.03, "learning_rate": 9.971222457460991e-05, "loss": 3.2916, "step": 367 }, { "epoch": 0.03, "learning_rate": 9.971144044538541e-05, "loss": 3.2849, "step": 368 }, { "epoch": 0.03, "learning_rate": 9.971065631616092e-05, "loss": 3.3521, "step": 369 }, { "epoch": 0.03, "learning_rate": 9.970987218693642e-05, "loss": 2.9774, "step": 370 }, { "epoch": 0.03, "learning_rate": 9.970908805771192e-05, "loss": 3.3487, "step": 371 }, { "epoch": 0.03, "learning_rate": 9.970830392848743e-05, "loss": 3.3141, "step": 372 }, { "epoch": 0.03, "learning_rate": 9.970751979926293e-05, "loss": 3.1183, "step": 373 }, { "epoch": 0.03, "learning_rate": 9.970673567003843e-05, "loss": 3.1664, "step": 374 }, { "epoch": 0.03, "learning_rate": 9.970595154081394e-05, "loss": 3.1072, "step": 375 }, { "epoch": 0.03, "learning_rate": 9.970516741158944e-05, "loss": 3.1931, "step": 376 }, { "epoch": 0.03, "learning_rate": 9.970438328236495e-05, "loss": 3.363, "step": 377 }, { "epoch": 0.03, "learning_rate": 9.970359915314045e-05, "loss": 3.205, "step": 378 }, { "epoch": 0.03, "learning_rate": 9.970281502391594e-05, "loss": 3.0846, "step": 379 }, { "epoch": 0.03, "learning_rate": 9.970203089469144e-05, "loss": 3.2845, "step": 380 }, { "epoch": 0.03, "learning_rate": 9.970124676546695e-05, "loss": 3.333, "step": 381 }, { "epoch": 0.03, "learning_rate": 9.970046263624245e-05, "loss": 2.9665, "step": 382 }, { "epoch": 0.03, "learning_rate": 9.969967850701795e-05, "loss": 2.9537, "step": 383 }, { "epoch": 0.03, "learning_rate": 9.969889437779346e-05, "loss": 3.0802, "step": 384 }, { "epoch": 0.03, "learning_rate": 9.969811024856896e-05, "loss": 3.3122, "step": 385 }, { "epoch": 0.03, "learning_rate": 9.969732611934447e-05, "loss": 3.1031, "step": 386 }, { "epoch": 0.03, "learning_rate": 9.969654199011997e-05, "loss": 3.2036, "step": 387 }, { "epoch": 0.03, "learning_rate": 9.969575786089547e-05, "loss": 3.094, "step": 388 }, { "epoch": 0.03, "learning_rate": 9.969497373167098e-05, "loss": 3.0654, "step": 389 }, { "epoch": 0.03, "learning_rate": 9.969418960244648e-05, "loss": 3.1923, "step": 390 }, { "epoch": 0.03, "learning_rate": 9.969340547322198e-05, "loss": 3.1463, "step": 391 }, { "epoch": 0.03, "learning_rate": 9.96926213439975e-05, "loss": 3.0191, "step": 392 }, { "epoch": 0.03, "learning_rate": 9.9691837214773e-05, "loss": 3.0037, "step": 393 }, { "epoch": 0.03, "learning_rate": 9.969105308554851e-05, "loss": 3.0894, "step": 394 }, { "epoch": 0.03, "learning_rate": 9.969026895632401e-05, "loss": 3.2564, "step": 395 }, { "epoch": 0.03, "learning_rate": 9.968948482709952e-05, "loss": 3.2662, "step": 396 }, { "epoch": 0.03, "learning_rate": 9.968870069787502e-05, "loss": 3.0901, "step": 397 }, { "epoch": 0.03, "learning_rate": 9.968791656865052e-05, "loss": 3.0989, "step": 398 }, { "epoch": 0.03, "learning_rate": 9.968713243942603e-05, "loss": 3.1298, "step": 399 }, { "epoch": 0.03, "learning_rate": 9.968634831020153e-05, "loss": 3.1096, "step": 400 }, { "epoch": 0.03, "learning_rate": 9.968556418097703e-05, "loss": 3.1652, "step": 401 }, { "epoch": 0.03, "learning_rate": 9.968478005175254e-05, "loss": 3.1363, "step": 402 }, { "epoch": 0.03, "learning_rate": 9.968399592252804e-05, "loss": 3.356, "step": 403 }, { "epoch": 0.03, "learning_rate": 9.968321179330354e-05, "loss": 3.1799, "step": 404 }, { "epoch": 0.03, "learning_rate": 9.968242766407905e-05, "loss": 3.2312, "step": 405 }, { "epoch": 0.03, "learning_rate": 9.968164353485455e-05, "loss": 3.2024, "step": 406 }, { "epoch": 0.03, "learning_rate": 9.968085940563006e-05, "loss": 3.162, "step": 407 }, { "epoch": 0.03, "learning_rate": 9.968007527640556e-05, "loss": 3.0836, "step": 408 }, { "epoch": 0.03, "learning_rate": 9.967929114718106e-05, "loss": 3.5949, "step": 409 }, { "epoch": 0.03, "learning_rate": 9.967850701795657e-05, "loss": 3.1996, "step": 410 }, { "epoch": 0.03, "learning_rate": 9.967772288873207e-05, "loss": 3.3776, "step": 411 }, { "epoch": 0.03, "learning_rate": 9.967693875950757e-05, "loss": 3.0328, "step": 412 }, { "epoch": 0.03, "learning_rate": 9.967615463028308e-05, "loss": 2.9911, "step": 413 }, { "epoch": 0.03, "learning_rate": 9.967537050105858e-05, "loss": 3.1332, "step": 414 }, { "epoch": 0.03, "learning_rate": 9.967458637183407e-05, "loss": 3.2048, "step": 415 }, { "epoch": 0.03, "learning_rate": 9.967380224260958e-05, "loss": 3.0826, "step": 416 }, { "epoch": 0.03, "learning_rate": 9.967301811338509e-05, "loss": 3.0063, "step": 417 }, { "epoch": 0.03, "learning_rate": 9.96722339841606e-05, "loss": 3.3228, "step": 418 }, { "epoch": 0.03, "learning_rate": 9.96714498549361e-05, "loss": 3.2882, "step": 419 }, { "epoch": 0.03, "learning_rate": 9.96706657257116e-05, "loss": 3.1052, "step": 420 }, { "epoch": 0.03, "learning_rate": 9.966988159648711e-05, "loss": 3.1421, "step": 421 }, { "epoch": 0.03, "learning_rate": 9.966909746726261e-05, "loss": 3.2457, "step": 422 }, { "epoch": 0.03, "learning_rate": 9.966831333803811e-05, "loss": 3.3649, "step": 423 }, { "epoch": 0.03, "learning_rate": 9.966752920881362e-05, "loss": 3.436, "step": 424 }, { "epoch": 0.03, "learning_rate": 9.966674507958912e-05, "loss": 3.1526, "step": 425 }, { "epoch": 0.03, "learning_rate": 9.966596095036463e-05, "loss": 3.0683, "step": 426 }, { "epoch": 0.03, "learning_rate": 9.966517682114013e-05, "loss": 3.5385, "step": 427 }, { "epoch": 0.03, "learning_rate": 9.966439269191563e-05, "loss": 3.5821, "step": 428 }, { "epoch": 0.03, "learning_rate": 9.966360856269114e-05, "loss": 3.0507, "step": 429 }, { "epoch": 0.03, "learning_rate": 9.966282443346664e-05, "loss": 3.5349, "step": 430 }, { "epoch": 0.03, "learning_rate": 9.966204030424214e-05, "loss": 3.0532, "step": 431 }, { "epoch": 0.03, "learning_rate": 9.966125617501765e-05, "loss": 3.4426, "step": 432 }, { "epoch": 0.03, "learning_rate": 9.966047204579315e-05, "loss": 3.2097, "step": 433 }, { "epoch": 0.03, "learning_rate": 9.965968791656866e-05, "loss": 3.3481, "step": 434 }, { "epoch": 0.03, "learning_rate": 9.965890378734416e-05, "loss": 3.2209, "step": 435 }, { "epoch": 0.03, "learning_rate": 9.965811965811966e-05, "loss": 3.3886, "step": 436 }, { "epoch": 0.03, "learning_rate": 9.965733552889517e-05, "loss": 3.2022, "step": 437 }, { "epoch": 0.03, "learning_rate": 9.965655139967067e-05, "loss": 3.0826, "step": 438 }, { "epoch": 0.03, "learning_rate": 9.965576727044617e-05, "loss": 3.293, "step": 439 }, { "epoch": 0.03, "learning_rate": 9.965498314122168e-05, "loss": 2.9524, "step": 440 }, { "epoch": 0.03, "learning_rate": 9.965419901199718e-05, "loss": 3.253, "step": 441 }, { "epoch": 0.03, "learning_rate": 9.965341488277268e-05, "loss": 3.2396, "step": 442 }, { "epoch": 0.03, "learning_rate": 9.965263075354819e-05, "loss": 3.3034, "step": 443 }, { "epoch": 0.03, "learning_rate": 9.965184662432369e-05, "loss": 3.144, "step": 444 }, { "epoch": 0.03, "learning_rate": 9.96510624950992e-05, "loss": 3.3926, "step": 445 }, { "epoch": 0.03, "learning_rate": 9.96502783658747e-05, "loss": 2.9694, "step": 446 }, { "epoch": 0.04, "learning_rate": 9.96494942366502e-05, "loss": 3.0221, "step": 447 }, { "epoch": 0.04, "learning_rate": 9.964871010742572e-05, "loss": 3.123, "step": 448 }, { "epoch": 0.04, "learning_rate": 9.964792597820122e-05, "loss": 3.3493, "step": 449 }, { "epoch": 0.04, "learning_rate": 9.964714184897673e-05, "loss": 3.2342, "step": 450 }, { "epoch": 0.04, "learning_rate": 9.964635771975222e-05, "loss": 3.1333, "step": 451 }, { "epoch": 0.04, "learning_rate": 9.964557359052772e-05, "loss": 3.2815, "step": 452 }, { "epoch": 0.04, "learning_rate": 9.964478946130322e-05, "loss": 3.3783, "step": 453 }, { "epoch": 0.04, "learning_rate": 9.964400533207873e-05, "loss": 3.2786, "step": 454 }, { "epoch": 0.04, "learning_rate": 9.964322120285423e-05, "loss": 3.0116, "step": 455 }, { "epoch": 0.04, "learning_rate": 9.964243707362974e-05, "loss": 3.3131, "step": 456 }, { "epoch": 0.04, "learning_rate": 9.964165294440524e-05, "loss": 3.2433, "step": 457 }, { "epoch": 0.04, "learning_rate": 9.964086881518074e-05, "loss": 3.1365, "step": 458 }, { "epoch": 0.04, "learning_rate": 9.964008468595625e-05, "loss": 3.1112, "step": 459 }, { "epoch": 0.04, "learning_rate": 9.963930055673175e-05, "loss": 2.9697, "step": 460 }, { "epoch": 0.04, "learning_rate": 9.963851642750725e-05, "loss": 3.1564, "step": 461 }, { "epoch": 0.04, "learning_rate": 9.963773229828276e-05, "loss": 3.2776, "step": 462 }, { "epoch": 0.04, "learning_rate": 9.963694816905826e-05, "loss": 3.3289, "step": 463 }, { "epoch": 0.04, "learning_rate": 9.963616403983377e-05, "loss": 3.3973, "step": 464 }, { "epoch": 0.04, "learning_rate": 9.963537991060927e-05, "loss": 3.1135, "step": 465 }, { "epoch": 0.04, "learning_rate": 9.963459578138477e-05, "loss": 3.2292, "step": 466 }, { "epoch": 0.04, "learning_rate": 9.963381165216028e-05, "loss": 3.2071, "step": 467 }, { "epoch": 0.04, "learning_rate": 9.963302752293578e-05, "loss": 3.2114, "step": 468 }, { "epoch": 0.04, "learning_rate": 9.963224339371128e-05, "loss": 3.2726, "step": 469 }, { "epoch": 0.04, "learning_rate": 9.963145926448679e-05, "loss": 3.1748, "step": 470 }, { "epoch": 0.04, "learning_rate": 9.963067513526229e-05, "loss": 2.8828, "step": 471 }, { "epoch": 0.04, "learning_rate": 9.96298910060378e-05, "loss": 3.3835, "step": 472 }, { "epoch": 0.04, "learning_rate": 9.96291068768133e-05, "loss": 3.3152, "step": 473 }, { "epoch": 0.04, "learning_rate": 9.962832274758882e-05, "loss": 3.0075, "step": 474 }, { "epoch": 0.04, "learning_rate": 9.962753861836432e-05, "loss": 3.3867, "step": 475 }, { "epoch": 0.04, "learning_rate": 9.962675448913982e-05, "loss": 3.092, "step": 476 }, { "epoch": 0.04, "learning_rate": 9.962597035991533e-05, "loss": 3.2299, "step": 477 }, { "epoch": 0.04, "learning_rate": 9.962518623069083e-05, "loss": 3.2656, "step": 478 }, { "epoch": 0.04, "learning_rate": 9.962440210146633e-05, "loss": 3.1558, "step": 479 }, { "epoch": 0.04, "learning_rate": 9.962361797224184e-05, "loss": 3.2873, "step": 480 }, { "epoch": 0.04, "learning_rate": 9.962283384301734e-05, "loss": 3.0538, "step": 481 }, { "epoch": 0.04, "learning_rate": 9.962204971379285e-05, "loss": 3.0968, "step": 482 }, { "epoch": 0.04, "learning_rate": 9.962126558456835e-05, "loss": 3.2858, "step": 483 }, { "epoch": 0.04, "learning_rate": 9.962048145534385e-05, "loss": 2.8735, "step": 484 }, { "epoch": 0.04, "learning_rate": 9.961969732611936e-05, "loss": 3.2635, "step": 485 }, { "epoch": 0.04, "learning_rate": 9.961891319689486e-05, "loss": 3.3579, "step": 486 }, { "epoch": 0.04, "learning_rate": 9.961812906767035e-05, "loss": 3.1901, "step": 487 }, { "epoch": 0.04, "learning_rate": 9.961734493844585e-05, "loss": 3.0516, "step": 488 }, { "epoch": 0.04, "learning_rate": 9.961656080922136e-05, "loss": 3.2088, "step": 489 }, { "epoch": 0.04, "learning_rate": 9.961577667999686e-05, "loss": 3.2356, "step": 490 }, { "epoch": 0.04, "learning_rate": 9.961499255077236e-05, "loss": 3.2208, "step": 491 }, { "epoch": 0.04, "learning_rate": 9.961420842154787e-05, "loss": 3.1573, "step": 492 }, { "epoch": 0.04, "learning_rate": 9.961342429232337e-05, "loss": 3.0424, "step": 493 }, { "epoch": 0.04, "learning_rate": 9.961264016309888e-05, "loss": 3.094, "step": 494 }, { "epoch": 0.04, "learning_rate": 9.961185603387438e-05, "loss": 3.2563, "step": 495 }, { "epoch": 0.04, "learning_rate": 9.961107190464988e-05, "loss": 3.0432, "step": 496 }, { "epoch": 0.04, "learning_rate": 9.961028777542539e-05, "loss": 3.166, "step": 497 }, { "epoch": 0.04, "learning_rate": 9.960950364620089e-05, "loss": 2.9134, "step": 498 }, { "epoch": 0.04, "learning_rate": 9.96087195169764e-05, "loss": 3.7161, "step": 499 }, { "epoch": 0.04, "learning_rate": 9.960793538775191e-05, "loss": 3.2552, "step": 500 }, { "epoch": 0.04, "learning_rate": 9.960715125852742e-05, "loss": 3.2581, "step": 501 }, { "epoch": 0.04, "learning_rate": 9.960636712930292e-05, "loss": 3.0448, "step": 502 }, { "epoch": 0.04, "learning_rate": 9.960558300007842e-05, "loss": 3.0274, "step": 503 }, { "epoch": 0.04, "learning_rate": 9.960479887085393e-05, "loss": 3.2669, "step": 504 }, { "epoch": 0.04, "learning_rate": 9.960401474162943e-05, "loss": 3.2283, "step": 505 }, { "epoch": 0.04, "learning_rate": 9.960323061240493e-05, "loss": 3.5706, "step": 506 }, { "epoch": 0.04, "learning_rate": 9.960244648318044e-05, "loss": 3.1837, "step": 507 }, { "epoch": 0.04, "learning_rate": 9.960166235395594e-05, "loss": 3.1827, "step": 508 }, { "epoch": 0.04, "learning_rate": 9.960087822473144e-05, "loss": 3.2942, "step": 509 }, { "epoch": 0.04, "learning_rate": 9.960009409550695e-05, "loss": 3.051, "step": 510 }, { "epoch": 0.04, "learning_rate": 9.959930996628245e-05, "loss": 3.2895, "step": 511 }, { "epoch": 0.04, "learning_rate": 9.959852583705796e-05, "loss": 3.2521, "step": 512 }, { "epoch": 0.04, "learning_rate": 9.959774170783346e-05, "loss": 3.3413, "step": 513 }, { "epoch": 0.04, "learning_rate": 9.959695757860896e-05, "loss": 3.2457, "step": 514 }, { "epoch": 0.04, "learning_rate": 9.959617344938447e-05, "loss": 3.1089, "step": 515 }, { "epoch": 0.04, "learning_rate": 9.959538932015997e-05, "loss": 3.0821, "step": 516 }, { "epoch": 0.04, "learning_rate": 9.959460519093547e-05, "loss": 3.063, "step": 517 }, { "epoch": 0.04, "learning_rate": 9.959382106171098e-05, "loss": 3.4875, "step": 518 }, { "epoch": 0.04, "learning_rate": 9.959303693248648e-05, "loss": 3.011, "step": 519 }, { "epoch": 0.04, "learning_rate": 9.959225280326198e-05, "loss": 3.3484, "step": 520 }, { "epoch": 0.04, "learning_rate": 9.959146867403749e-05, "loss": 3.0391, "step": 521 }, { "epoch": 0.04, "learning_rate": 9.959068454481299e-05, "loss": 2.9589, "step": 522 }, { "epoch": 0.04, "learning_rate": 9.958990041558848e-05, "loss": 3.0135, "step": 523 }, { "epoch": 0.04, "learning_rate": 9.958911628636399e-05, "loss": 2.9489, "step": 524 }, { "epoch": 0.04, "learning_rate": 9.95883321571395e-05, "loss": 3.3254, "step": 525 }, { "epoch": 0.04, "learning_rate": 9.958754802791501e-05, "loss": 3.5225, "step": 526 }, { "epoch": 0.04, "learning_rate": 9.958676389869051e-05, "loss": 3.1827, "step": 527 }, { "epoch": 0.04, "learning_rate": 9.958597976946601e-05, "loss": 3.1897, "step": 528 }, { "epoch": 0.04, "learning_rate": 9.958519564024152e-05, "loss": 3.5531, "step": 529 }, { "epoch": 0.04, "learning_rate": 9.958441151101702e-05, "loss": 3.2243, "step": 530 }, { "epoch": 0.04, "learning_rate": 9.958362738179253e-05, "loss": 3.3412, "step": 531 }, { "epoch": 0.04, "learning_rate": 9.958284325256803e-05, "loss": 3.2949, "step": 532 }, { "epoch": 0.04, "learning_rate": 9.958205912334353e-05, "loss": 3.0113, "step": 533 }, { "epoch": 0.04, "learning_rate": 9.958127499411904e-05, "loss": 3.3761, "step": 534 }, { "epoch": 0.04, "learning_rate": 9.958049086489454e-05, "loss": 3.5556, "step": 535 }, { "epoch": 0.04, "learning_rate": 9.957970673567004e-05, "loss": 3.2689, "step": 536 }, { "epoch": 0.04, "learning_rate": 9.957892260644555e-05, "loss": 3.4118, "step": 537 }, { "epoch": 0.04, "learning_rate": 9.957813847722105e-05, "loss": 3.275, "step": 538 }, { "epoch": 0.04, "learning_rate": 9.957735434799655e-05, "loss": 3.1692, "step": 539 }, { "epoch": 0.04, "learning_rate": 9.957657021877206e-05, "loss": 3.4191, "step": 540 }, { "epoch": 0.04, "learning_rate": 9.957578608954756e-05, "loss": 2.9899, "step": 541 }, { "epoch": 0.04, "learning_rate": 9.957500196032307e-05, "loss": 3.0887, "step": 542 }, { "epoch": 0.04, "learning_rate": 9.957421783109857e-05, "loss": 3.4165, "step": 543 }, { "epoch": 0.04, "learning_rate": 9.957343370187407e-05, "loss": 2.9833, "step": 544 }, { "epoch": 0.04, "learning_rate": 9.957264957264958e-05, "loss": 3.2831, "step": 545 }, { "epoch": 0.04, "learning_rate": 9.957186544342508e-05, "loss": 3.2361, "step": 546 }, { "epoch": 0.04, "learning_rate": 9.957108131420058e-05, "loss": 3.0371, "step": 547 }, { "epoch": 0.04, "learning_rate": 9.957029718497609e-05, "loss": 3.315, "step": 548 }, { "epoch": 0.04, "learning_rate": 9.956951305575159e-05, "loss": 3.0886, "step": 549 }, { "epoch": 0.04, "learning_rate": 9.95687289265271e-05, "loss": 3.1341, "step": 550 }, { "epoch": 0.04, "learning_rate": 9.95679447973026e-05, "loss": 3.1766, "step": 551 }, { "epoch": 0.04, "learning_rate": 9.95671606680781e-05, "loss": 3.2957, "step": 552 }, { "epoch": 0.04, "learning_rate": 9.95663765388536e-05, "loss": 3.0901, "step": 553 }, { "epoch": 0.04, "learning_rate": 9.956559240962911e-05, "loss": 3.3093, "step": 554 }, { "epoch": 0.04, "learning_rate": 9.956480828040461e-05, "loss": 3.2271, "step": 555 }, { "epoch": 0.04, "learning_rate": 9.956402415118013e-05, "loss": 3.1091, "step": 556 }, { "epoch": 0.04, "learning_rate": 9.956324002195563e-05, "loss": 3.0004, "step": 557 }, { "epoch": 0.04, "learning_rate": 9.956245589273114e-05, "loss": 3.5609, "step": 558 }, { "epoch": 0.04, "learning_rate": 9.956167176350663e-05, "loss": 2.9411, "step": 559 }, { "epoch": 0.04, "learning_rate": 9.956088763428213e-05, "loss": 3.1096, "step": 560 }, { "epoch": 0.04, "learning_rate": 9.956010350505764e-05, "loss": 3.0921, "step": 561 }, { "epoch": 0.04, "learning_rate": 9.955931937583314e-05, "loss": 3.2584, "step": 562 }, { "epoch": 0.04, "learning_rate": 9.955853524660864e-05, "loss": 3.0958, "step": 563 }, { "epoch": 0.04, "learning_rate": 9.955775111738415e-05, "loss": 2.826, "step": 564 }, { "epoch": 0.04, "learning_rate": 9.955696698815965e-05, "loss": 3.0011, "step": 565 }, { "epoch": 0.04, "learning_rate": 9.955618285893515e-05, "loss": 2.9154, "step": 566 }, { "epoch": 0.04, "learning_rate": 9.955539872971066e-05, "loss": 3.1942, "step": 567 }, { "epoch": 0.04, "learning_rate": 9.955461460048616e-05, "loss": 3.0436, "step": 568 }, { "epoch": 0.04, "learning_rate": 9.955383047126167e-05, "loss": 3.185, "step": 569 }, { "epoch": 0.04, "learning_rate": 9.955304634203717e-05, "loss": 3.1226, "step": 570 }, { "epoch": 0.04, "learning_rate": 9.955226221281267e-05, "loss": 3.3343, "step": 571 }, { "epoch": 0.04, "learning_rate": 9.955147808358818e-05, "loss": 3.3371, "step": 572 }, { "epoch": 0.04, "learning_rate": 9.955069395436368e-05, "loss": 3.158, "step": 573 }, { "epoch": 0.05, "learning_rate": 9.954990982513918e-05, "loss": 3.116, "step": 574 }, { "epoch": 0.05, "learning_rate": 9.954912569591469e-05, "loss": 3.1873, "step": 575 }, { "epoch": 0.05, "learning_rate": 9.954834156669019e-05, "loss": 3.4538, "step": 576 }, { "epoch": 0.05, "learning_rate": 9.95475574374657e-05, "loss": 3.4011, "step": 577 }, { "epoch": 0.05, "learning_rate": 9.95467733082412e-05, "loss": 3.3019, "step": 578 }, { "epoch": 0.05, "learning_rate": 9.95459891790167e-05, "loss": 3.0125, "step": 579 }, { "epoch": 0.05, "learning_rate": 9.95452050497922e-05, "loss": 3.1526, "step": 580 }, { "epoch": 0.05, "learning_rate": 9.954442092056771e-05, "loss": 3.0708, "step": 581 }, { "epoch": 0.05, "learning_rate": 9.954363679134323e-05, "loss": 3.3198, "step": 582 }, { "epoch": 0.05, "learning_rate": 9.954285266211873e-05, "loss": 3.4988, "step": 583 }, { "epoch": 0.05, "learning_rate": 9.954206853289423e-05, "loss": 3.0613, "step": 584 }, { "epoch": 0.05, "learning_rate": 9.954128440366974e-05, "loss": 3.0431, "step": 585 }, { "epoch": 0.05, "learning_rate": 9.954050027444524e-05, "loss": 3.235, "step": 586 }, { "epoch": 0.05, "learning_rate": 9.953971614522074e-05, "loss": 3.3273, "step": 587 }, { "epoch": 0.05, "learning_rate": 9.953893201599625e-05, "loss": 3.5425, "step": 588 }, { "epoch": 0.05, "learning_rate": 9.953814788677175e-05, "loss": 3.1828, "step": 589 }, { "epoch": 0.05, "learning_rate": 9.953736375754726e-05, "loss": 3.0213, "step": 590 }, { "epoch": 0.05, "learning_rate": 9.953657962832276e-05, "loss": 3.0836, "step": 591 }, { "epoch": 0.05, "learning_rate": 9.953579549909826e-05, "loss": 3.1014, "step": 592 }, { "epoch": 0.05, "learning_rate": 9.953501136987377e-05, "loss": 3.295, "step": 593 }, { "epoch": 0.05, "learning_rate": 9.953422724064927e-05, "loss": 3.2491, "step": 594 }, { "epoch": 0.05, "learning_rate": 9.953344311142476e-05, "loss": 3.1932, "step": 595 }, { "epoch": 0.05, "learning_rate": 9.953265898220026e-05, "loss": 3.2711, "step": 596 }, { "epoch": 0.05, "learning_rate": 9.953187485297577e-05, "loss": 3.3648, "step": 597 }, { "epoch": 0.05, "learning_rate": 9.953109072375127e-05, "loss": 3.1584, "step": 598 }, { "epoch": 0.05, "learning_rate": 9.953030659452678e-05, "loss": 3.1974, "step": 599 }, { "epoch": 0.05, "learning_rate": 9.952952246530228e-05, "loss": 3.2777, "step": 600 }, { "epoch": 0.05, "learning_rate": 9.952873833607778e-05, "loss": 3.3113, "step": 601 }, { "epoch": 0.05, "learning_rate": 9.952795420685329e-05, "loss": 3.0523, "step": 602 }, { "epoch": 0.05, "learning_rate": 9.952717007762879e-05, "loss": 3.0858, "step": 603 }, { "epoch": 0.05, "learning_rate": 9.95263859484043e-05, "loss": 3.2163, "step": 604 }, { "epoch": 0.05, "learning_rate": 9.95256018191798e-05, "loss": 2.8933, "step": 605 }, { "epoch": 0.05, "learning_rate": 9.95248176899553e-05, "loss": 2.9385, "step": 606 }, { "epoch": 0.05, "learning_rate": 9.95240335607308e-05, "loss": 3.0143, "step": 607 }, { "epoch": 0.05, "learning_rate": 9.952324943150632e-05, "loss": 3.3591, "step": 608 }, { "epoch": 0.05, "learning_rate": 9.952246530228183e-05, "loss": 3.2456, "step": 609 }, { "epoch": 0.05, "learning_rate": 9.952168117305733e-05, "loss": 3.157, "step": 610 }, { "epoch": 0.05, "learning_rate": 9.952089704383283e-05, "loss": 3.3803, "step": 611 }, { "epoch": 0.05, "learning_rate": 9.952011291460834e-05, "loss": 3.1262, "step": 612 }, { "epoch": 0.05, "learning_rate": 9.951932878538384e-05, "loss": 2.9449, "step": 613 }, { "epoch": 0.05, "learning_rate": 9.951854465615934e-05, "loss": 2.9508, "step": 614 }, { "epoch": 0.05, "learning_rate": 9.951776052693485e-05, "loss": 3.1883, "step": 615 }, { "epoch": 0.05, "learning_rate": 9.951697639771035e-05, "loss": 3.3662, "step": 616 }, { "epoch": 0.05, "learning_rate": 9.951619226848586e-05, "loss": 3.0996, "step": 617 }, { "epoch": 0.05, "learning_rate": 9.951540813926136e-05, "loss": 2.9535, "step": 618 }, { "epoch": 0.05, "learning_rate": 9.951462401003686e-05, "loss": 3.2995, "step": 619 }, { "epoch": 0.05, "learning_rate": 9.951383988081237e-05, "loss": 3.2384, "step": 620 }, { "epoch": 0.05, "learning_rate": 9.951305575158787e-05, "loss": 3.2196, "step": 621 }, { "epoch": 0.05, "learning_rate": 9.951227162236337e-05, "loss": 3.2858, "step": 622 }, { "epoch": 0.05, "learning_rate": 9.951148749313888e-05, "loss": 3.0532, "step": 623 }, { "epoch": 0.05, "learning_rate": 9.951070336391438e-05, "loss": 3.2148, "step": 624 }, { "epoch": 0.05, "learning_rate": 9.950991923468988e-05, "loss": 3.1319, "step": 625 }, { "epoch": 0.05, "learning_rate": 9.950913510546539e-05, "loss": 3.1818, "step": 626 }, { "epoch": 0.05, "learning_rate": 9.950835097624089e-05, "loss": 3.3675, "step": 627 }, { "epoch": 0.05, "learning_rate": 9.95075668470164e-05, "loss": 3.1366, "step": 628 }, { "epoch": 0.05, "learning_rate": 9.95067827177919e-05, "loss": 3.3753, "step": 629 }, { "epoch": 0.05, "learning_rate": 9.95059985885674e-05, "loss": 3.138, "step": 630 }, { "epoch": 0.05, "learning_rate": 9.950521445934289e-05, "loss": 3.0253, "step": 631 }, { "epoch": 0.05, "learning_rate": 9.95044303301184e-05, "loss": 3.1936, "step": 632 }, { "epoch": 0.05, "learning_rate": 9.950364620089391e-05, "loss": 3.0563, "step": 633 }, { "epoch": 0.05, "learning_rate": 9.950286207166942e-05, "loss": 3.2123, "step": 634 }, { "epoch": 0.05, "learning_rate": 9.950207794244492e-05, "loss": 3.0471, "step": 635 }, { "epoch": 0.05, "learning_rate": 9.950129381322043e-05, "loss": 3.1302, "step": 636 }, { "epoch": 0.05, "learning_rate": 9.950050968399593e-05, "loss": 3.3349, "step": 637 }, { "epoch": 0.05, "learning_rate": 9.949972555477143e-05, "loss": 3.0905, "step": 638 }, { "epoch": 0.05, "learning_rate": 9.949894142554694e-05, "loss": 3.3398, "step": 639 }, { "epoch": 0.05, "learning_rate": 9.949815729632244e-05, "loss": 3.2953, "step": 640 }, { "epoch": 0.05, "learning_rate": 9.949737316709794e-05, "loss": 3.1994, "step": 641 }, { "epoch": 0.05, "learning_rate": 9.949658903787345e-05, "loss": 2.9685, "step": 642 }, { "epoch": 0.05, "learning_rate": 9.949580490864895e-05, "loss": 3.2503, "step": 643 }, { "epoch": 0.05, "learning_rate": 9.949502077942445e-05, "loss": 3.1077, "step": 644 }, { "epoch": 0.05, "learning_rate": 9.949423665019996e-05, "loss": 3.1775, "step": 645 }, { "epoch": 0.05, "learning_rate": 9.949345252097546e-05, "loss": 3.2973, "step": 646 }, { "epoch": 0.05, "learning_rate": 9.949266839175097e-05, "loss": 3.2192, "step": 647 }, { "epoch": 0.05, "learning_rate": 9.949188426252647e-05, "loss": 3.1487, "step": 648 }, { "epoch": 0.05, "learning_rate": 9.949110013330197e-05, "loss": 3.4639, "step": 649 }, { "epoch": 0.05, "learning_rate": 9.949031600407748e-05, "loss": 3.0705, "step": 650 }, { "epoch": 0.05, "learning_rate": 9.948953187485298e-05, "loss": 3.3007, "step": 651 }, { "epoch": 0.05, "learning_rate": 9.948874774562848e-05, "loss": 3.2024, "step": 652 }, { "epoch": 0.05, "learning_rate": 9.948796361640399e-05, "loss": 3.2049, "step": 653 }, { "epoch": 0.05, "learning_rate": 9.948717948717949e-05, "loss": 3.1166, "step": 654 }, { "epoch": 0.05, "learning_rate": 9.9486395357955e-05, "loss": 3.2324, "step": 655 }, { "epoch": 0.05, "learning_rate": 9.94856112287305e-05, "loss": 3.1736, "step": 656 }, { "epoch": 0.05, "learning_rate": 9.9484827099506e-05, "loss": 2.9636, "step": 657 }, { "epoch": 0.05, "learning_rate": 9.94840429702815e-05, "loss": 3.0394, "step": 658 }, { "epoch": 0.05, "learning_rate": 9.948325884105701e-05, "loss": 3.2536, "step": 659 }, { "epoch": 0.05, "learning_rate": 9.948247471183251e-05, "loss": 3.1607, "step": 660 }, { "epoch": 0.05, "learning_rate": 9.948169058260802e-05, "loss": 3.1347, "step": 661 }, { "epoch": 0.05, "learning_rate": 9.948090645338352e-05, "loss": 3.1609, "step": 662 }, { "epoch": 0.05, "learning_rate": 9.948012232415902e-05, "loss": 3.1127, "step": 663 }, { "epoch": 0.05, "learning_rate": 9.947933819493454e-05, "loss": 3.0744, "step": 664 }, { "epoch": 0.05, "learning_rate": 9.947855406571005e-05, "loss": 3.298, "step": 665 }, { "epoch": 0.05, "learning_rate": 9.947776993648555e-05, "loss": 3.3367, "step": 666 }, { "epoch": 0.05, "learning_rate": 9.947698580726104e-05, "loss": 3.3293, "step": 667 }, { "epoch": 0.05, "learning_rate": 9.947620167803654e-05, "loss": 3.1962, "step": 668 }, { "epoch": 0.05, "learning_rate": 9.947541754881205e-05, "loss": 3.2238, "step": 669 }, { "epoch": 0.05, "learning_rate": 9.947463341958755e-05, "loss": 3.0184, "step": 670 }, { "epoch": 0.05, "learning_rate": 9.947384929036305e-05, "loss": 3.1292, "step": 671 }, { "epoch": 0.05, "learning_rate": 9.947306516113856e-05, "loss": 3.1501, "step": 672 }, { "epoch": 0.05, "learning_rate": 9.947228103191406e-05, "loss": 3.2156, "step": 673 }, { "epoch": 0.05, "learning_rate": 9.947149690268956e-05, "loss": 3.1766, "step": 674 }, { "epoch": 0.05, "learning_rate": 9.947071277346507e-05, "loss": 3.0738, "step": 675 }, { "epoch": 0.05, "learning_rate": 9.946992864424057e-05, "loss": 3.242, "step": 676 }, { "epoch": 0.05, "learning_rate": 9.946914451501608e-05, "loss": 2.891, "step": 677 }, { "epoch": 0.05, "learning_rate": 9.946836038579158e-05, "loss": 2.8727, "step": 678 }, { "epoch": 0.05, "learning_rate": 9.946757625656708e-05, "loss": 3.4422, "step": 679 }, { "epoch": 0.05, "learning_rate": 9.946679212734259e-05, "loss": 3.1058, "step": 680 }, { "epoch": 0.05, "learning_rate": 9.946600799811809e-05, "loss": 3.2169, "step": 681 }, { "epoch": 0.05, "learning_rate": 9.94652238688936e-05, "loss": 3.0404, "step": 682 }, { "epoch": 0.05, "learning_rate": 9.94644397396691e-05, "loss": 3.2115, "step": 683 }, { "epoch": 0.05, "learning_rate": 9.94636556104446e-05, "loss": 3.1309, "step": 684 }, { "epoch": 0.05, "learning_rate": 9.94628714812201e-05, "loss": 3.1406, "step": 685 }, { "epoch": 0.05, "learning_rate": 9.946208735199561e-05, "loss": 3.1197, "step": 686 }, { "epoch": 0.05, "learning_rate": 9.946130322277111e-05, "loss": 3.4001, "step": 687 }, { "epoch": 0.05, "learning_rate": 9.946051909354662e-05, "loss": 3.2419, "step": 688 }, { "epoch": 0.05, "learning_rate": 9.945973496432212e-05, "loss": 2.9975, "step": 689 }, { "epoch": 0.05, "learning_rate": 9.945895083509764e-05, "loss": 3.057, "step": 690 }, { "epoch": 0.05, "learning_rate": 9.945816670587314e-05, "loss": 3.2027, "step": 691 }, { "epoch": 0.05, "learning_rate": 9.945738257664864e-05, "loss": 3.2587, "step": 692 }, { "epoch": 0.05, "learning_rate": 9.945659844742415e-05, "loss": 3.1001, "step": 693 }, { "epoch": 0.05, "learning_rate": 9.945581431819965e-05, "loss": 3.009, "step": 694 }, { "epoch": 0.05, "learning_rate": 9.945503018897516e-05, "loss": 2.9887, "step": 695 }, { "epoch": 0.05, "learning_rate": 9.945424605975066e-05, "loss": 3.2833, "step": 696 }, { "epoch": 0.05, "learning_rate": 9.945346193052616e-05, "loss": 2.9899, "step": 697 }, { "epoch": 0.05, "learning_rate": 9.945267780130167e-05, "loss": 3.076, "step": 698 }, { "epoch": 0.05, "learning_rate": 9.945189367207717e-05, "loss": 3.1712, "step": 699 }, { "epoch": 0.05, "learning_rate": 9.945110954285267e-05, "loss": 3.2354, "step": 700 }, { "epoch": 0.05, "learning_rate": 9.945032541362818e-05, "loss": 3.0761, "step": 701 }, { "epoch": 0.06, "learning_rate": 9.944954128440368e-05, "loss": 2.9758, "step": 702 }, { "epoch": 0.06, "learning_rate": 9.944875715517917e-05, "loss": 3.2593, "step": 703 }, { "epoch": 0.06, "learning_rate": 9.944797302595468e-05, "loss": 3.1387, "step": 704 }, { "epoch": 0.06, "learning_rate": 9.944718889673018e-05, "loss": 3.0544, "step": 705 }, { "epoch": 0.06, "learning_rate": 9.944640476750568e-05, "loss": 2.9724, "step": 706 }, { "epoch": 0.06, "learning_rate": 9.944562063828119e-05, "loss": 3.0661, "step": 707 }, { "epoch": 0.06, "learning_rate": 9.944483650905669e-05, "loss": 3.1199, "step": 708 }, { "epoch": 0.06, "learning_rate": 9.94440523798322e-05, "loss": 3.172, "step": 709 }, { "epoch": 0.06, "learning_rate": 9.94432682506077e-05, "loss": 3.1771, "step": 710 }, { "epoch": 0.06, "learning_rate": 9.94424841213832e-05, "loss": 3.1386, "step": 711 }, { "epoch": 0.06, "learning_rate": 9.94416999921587e-05, "loss": 3.189, "step": 712 }, { "epoch": 0.06, "learning_rate": 9.944091586293421e-05, "loss": 3.0567, "step": 713 }, { "epoch": 0.06, "learning_rate": 9.944013173370971e-05, "loss": 3.2727, "step": 714 }, { "epoch": 0.06, "learning_rate": 9.943934760448522e-05, "loss": 3.2786, "step": 715 }, { "epoch": 0.06, "learning_rate": 9.943856347526073e-05, "loss": 3.0603, "step": 716 }, { "epoch": 0.06, "learning_rate": 9.943777934603624e-05, "loss": 3.2428, "step": 717 }, { "epoch": 0.06, "learning_rate": 9.943699521681174e-05, "loss": 3.1154, "step": 718 }, { "epoch": 0.06, "learning_rate": 9.943621108758724e-05, "loss": 3.1642, "step": 719 }, { "epoch": 0.06, "learning_rate": 9.943542695836275e-05, "loss": 3.0629, "step": 720 }, { "epoch": 0.06, "learning_rate": 9.943464282913825e-05, "loss": 3.2497, "step": 721 }, { "epoch": 0.06, "learning_rate": 9.943385869991375e-05, "loss": 3.1812, "step": 722 }, { "epoch": 0.06, "learning_rate": 9.943307457068926e-05, "loss": 2.9726, "step": 723 }, { "epoch": 0.06, "learning_rate": 9.943229044146476e-05, "loss": 3.006, "step": 724 }, { "epoch": 0.06, "learning_rate": 9.943150631224027e-05, "loss": 3.0614, "step": 725 }, { "epoch": 0.06, "learning_rate": 9.943072218301577e-05, "loss": 3.2081, "step": 726 }, { "epoch": 0.06, "learning_rate": 9.942993805379127e-05, "loss": 3.0069, "step": 727 }, { "epoch": 0.06, "learning_rate": 9.942915392456678e-05, "loss": 3.0493, "step": 728 }, { "epoch": 0.06, "learning_rate": 9.942836979534228e-05, "loss": 3.2553, "step": 729 }, { "epoch": 0.06, "learning_rate": 9.942758566611778e-05, "loss": 3.3998, "step": 730 }, { "epoch": 0.06, "learning_rate": 9.942680153689329e-05, "loss": 3.2799, "step": 731 }, { "epoch": 0.06, "learning_rate": 9.942601740766879e-05, "loss": 3.2542, "step": 732 }, { "epoch": 0.06, "learning_rate": 9.94252332784443e-05, "loss": 3.2314, "step": 733 }, { "epoch": 0.06, "learning_rate": 9.94244491492198e-05, "loss": 3.2102, "step": 734 }, { "epoch": 0.06, "learning_rate": 9.94236650199953e-05, "loss": 3.2298, "step": 735 }, { "epoch": 0.06, "learning_rate": 9.94228808907708e-05, "loss": 3.2027, "step": 736 }, { "epoch": 0.06, "learning_rate": 9.942209676154631e-05, "loss": 3.1494, "step": 737 }, { "epoch": 0.06, "learning_rate": 9.942131263232181e-05, "loss": 3.2959, "step": 738 }, { "epoch": 0.06, "learning_rate": 9.94205285030973e-05, "loss": 3.253, "step": 739 }, { "epoch": 0.06, "learning_rate": 9.941974437387281e-05, "loss": 3.2032, "step": 740 }, { "epoch": 0.06, "learning_rate": 9.941896024464832e-05, "loss": 3.4396, "step": 741 }, { "epoch": 0.06, "learning_rate": 9.941817611542383e-05, "loss": 3.0249, "step": 742 }, { "epoch": 0.06, "learning_rate": 9.941739198619933e-05, "loss": 3.3358, "step": 743 }, { "epoch": 0.06, "learning_rate": 9.941660785697484e-05, "loss": 3.2645, "step": 744 }, { "epoch": 0.06, "learning_rate": 9.941582372775034e-05, "loss": 3.1934, "step": 745 }, { "epoch": 0.06, "learning_rate": 9.941503959852584e-05, "loss": 3.0798, "step": 746 }, { "epoch": 0.06, "learning_rate": 9.941425546930135e-05, "loss": 3.3134, "step": 747 }, { "epoch": 0.06, "learning_rate": 9.941347134007685e-05, "loss": 2.9742, "step": 748 }, { "epoch": 0.06, "learning_rate": 9.941268721085235e-05, "loss": 3.1543, "step": 749 }, { "epoch": 0.06, "learning_rate": 9.941190308162786e-05, "loss": 3.0738, "step": 750 }, { "epoch": 0.06, "learning_rate": 9.941111895240336e-05, "loss": 2.9643, "step": 751 }, { "epoch": 0.06, "learning_rate": 9.941033482317887e-05, "loss": 3.2609, "step": 752 }, { "epoch": 0.06, "learning_rate": 9.940955069395437e-05, "loss": 3.2771, "step": 753 }, { "epoch": 0.06, "learning_rate": 9.940876656472987e-05, "loss": 3.1296, "step": 754 }, { "epoch": 0.06, "learning_rate": 9.940798243550538e-05, "loss": 3.1022, "step": 755 }, { "epoch": 0.06, "learning_rate": 9.940719830628088e-05, "loss": 3.0556, "step": 756 }, { "epoch": 0.06, "learning_rate": 9.940641417705638e-05, "loss": 3.3328, "step": 757 }, { "epoch": 0.06, "learning_rate": 9.940563004783189e-05, "loss": 2.9277, "step": 758 }, { "epoch": 0.06, "learning_rate": 9.940484591860739e-05, "loss": 3.2285, "step": 759 }, { "epoch": 0.06, "learning_rate": 9.94040617893829e-05, "loss": 3.0349, "step": 760 }, { "epoch": 0.06, "learning_rate": 9.94032776601584e-05, "loss": 3.0904, "step": 761 }, { "epoch": 0.06, "learning_rate": 9.94024935309339e-05, "loss": 3.2172, "step": 762 }, { "epoch": 0.06, "learning_rate": 9.94017094017094e-05, "loss": 2.9515, "step": 763 }, { "epoch": 0.06, "learning_rate": 9.940092527248491e-05, "loss": 3.1761, "step": 764 }, { "epoch": 0.06, "learning_rate": 9.940014114326041e-05, "loss": 3.1005, "step": 765 }, { "epoch": 0.06, "learning_rate": 9.939935701403592e-05, "loss": 3.4855, "step": 766 }, { "epoch": 0.06, "learning_rate": 9.939857288481142e-05, "loss": 3.2689, "step": 767 }, { "epoch": 0.06, "learning_rate": 9.939778875558692e-05, "loss": 3.2558, "step": 768 }, { "epoch": 0.06, "learning_rate": 9.939700462636243e-05, "loss": 3.216, "step": 769 }, { "epoch": 0.06, "learning_rate": 9.939622049713793e-05, "loss": 3.1597, "step": 770 }, { "epoch": 0.06, "learning_rate": 9.939543636791344e-05, "loss": 3.1292, "step": 771 }, { "epoch": 0.06, "learning_rate": 9.939465223868894e-05, "loss": 3.1844, "step": 772 }, { "epoch": 0.06, "learning_rate": 9.939386810946446e-05, "loss": 3.2508, "step": 773 }, { "epoch": 0.06, "learning_rate": 9.939308398023996e-05, "loss": 3.0789, "step": 774 }, { "epoch": 0.06, "learning_rate": 9.939229985101545e-05, "loss": 3.3029, "step": 775 }, { "epoch": 0.06, "learning_rate": 9.939151572179095e-05, "loss": 3.1544, "step": 776 }, { "epoch": 0.06, "learning_rate": 9.939073159256646e-05, "loss": 3.1199, "step": 777 }, { "epoch": 0.06, "learning_rate": 9.938994746334196e-05, "loss": 3.1914, "step": 778 }, { "epoch": 0.06, "learning_rate": 9.938916333411746e-05, "loss": 3.0112, "step": 779 }, { "epoch": 0.06, "learning_rate": 9.938837920489297e-05, "loss": 3.1239, "step": 780 }, { "epoch": 0.06, "learning_rate": 9.938759507566847e-05, "loss": 3.029, "step": 781 }, { "epoch": 0.06, "learning_rate": 9.938681094644398e-05, "loss": 3.1998, "step": 782 }, { "epoch": 0.06, "learning_rate": 9.938602681721948e-05, "loss": 3.1218, "step": 783 }, { "epoch": 0.06, "learning_rate": 9.938524268799498e-05, "loss": 3.0598, "step": 784 }, { "epoch": 0.06, "learning_rate": 9.938445855877049e-05, "loss": 3.0199, "step": 785 }, { "epoch": 0.06, "learning_rate": 9.938367442954599e-05, "loss": 3.154, "step": 786 }, { "epoch": 0.06, "learning_rate": 9.93828903003215e-05, "loss": 3.11, "step": 787 }, { "epoch": 0.06, "learning_rate": 9.9382106171097e-05, "loss": 3.4102, "step": 788 }, { "epoch": 0.06, "learning_rate": 9.93813220418725e-05, "loss": 3.1712, "step": 789 }, { "epoch": 0.06, "learning_rate": 9.9380537912648e-05, "loss": 2.8982, "step": 790 }, { "epoch": 0.06, "learning_rate": 9.937975378342351e-05, "loss": 2.9467, "step": 791 }, { "epoch": 0.06, "learning_rate": 9.937896965419901e-05, "loss": 3.2789, "step": 792 }, { "epoch": 0.06, "learning_rate": 9.937818552497452e-05, "loss": 3.4862, "step": 793 }, { "epoch": 0.06, "learning_rate": 9.937740139575002e-05, "loss": 3.0116, "step": 794 }, { "epoch": 0.06, "learning_rate": 9.937661726652552e-05, "loss": 3.1692, "step": 795 }, { "epoch": 0.06, "learning_rate": 9.937583313730103e-05, "loss": 3.2302, "step": 796 }, { "epoch": 0.06, "learning_rate": 9.937504900807653e-05, "loss": 3.0446, "step": 797 }, { "epoch": 0.06, "learning_rate": 9.937426487885205e-05, "loss": 3.4153, "step": 798 }, { "epoch": 0.06, "learning_rate": 9.937348074962755e-05, "loss": 3.1311, "step": 799 }, { "epoch": 0.06, "learning_rate": 9.937269662040306e-05, "loss": 3.1825, "step": 800 }, { "epoch": 0.06, "learning_rate": 9.937191249117856e-05, "loss": 3.0285, "step": 801 }, { "epoch": 0.06, "learning_rate": 9.937112836195406e-05, "loss": 2.9952, "step": 802 }, { "epoch": 0.06, "learning_rate": 9.937034423272957e-05, "loss": 3.0925, "step": 803 }, { "epoch": 0.06, "learning_rate": 9.936956010350507e-05, "loss": 2.9601, "step": 804 }, { "epoch": 0.06, "learning_rate": 9.936877597428057e-05, "loss": 2.9146, "step": 805 }, { "epoch": 0.06, "learning_rate": 9.936799184505608e-05, "loss": 3.2762, "step": 806 }, { "epoch": 0.06, "learning_rate": 9.936720771583158e-05, "loss": 3.1326, "step": 807 }, { "epoch": 0.06, "learning_rate": 9.936642358660708e-05, "loss": 3.1682, "step": 808 }, { "epoch": 0.06, "learning_rate": 9.936563945738259e-05, "loss": 3.4135, "step": 809 }, { "epoch": 0.06, "learning_rate": 9.936485532815809e-05, "loss": 3.0339, "step": 810 }, { "epoch": 0.06, "learning_rate": 9.936407119893358e-05, "loss": 3.0255, "step": 811 }, { "epoch": 0.06, "learning_rate": 9.936328706970909e-05, "loss": 3.3962, "step": 812 }, { "epoch": 0.06, "learning_rate": 9.936250294048459e-05, "loss": 3.221, "step": 813 }, { "epoch": 0.06, "learning_rate": 9.93617188112601e-05, "loss": 3.1168, "step": 814 }, { "epoch": 0.06, "learning_rate": 9.93609346820356e-05, "loss": 2.9844, "step": 815 }, { "epoch": 0.06, "learning_rate": 9.93601505528111e-05, "loss": 3.0504, "step": 816 }, { "epoch": 0.06, "learning_rate": 9.93593664235866e-05, "loss": 3.5533, "step": 817 }, { "epoch": 0.06, "learning_rate": 9.935858229436211e-05, "loss": 3.1357, "step": 818 }, { "epoch": 0.06, "learning_rate": 9.935779816513761e-05, "loss": 3.1304, "step": 819 }, { "epoch": 0.06, "learning_rate": 9.935701403591312e-05, "loss": 3.1395, "step": 820 }, { "epoch": 0.06, "learning_rate": 9.935622990668862e-05, "loss": 3.4991, "step": 821 }, { "epoch": 0.06, "learning_rate": 9.935544577746412e-05, "loss": 2.9579, "step": 822 }, { "epoch": 0.06, "learning_rate": 9.935466164823963e-05, "loss": 2.9992, "step": 823 }, { "epoch": 0.06, "learning_rate": 9.935387751901514e-05, "loss": 3.0199, "step": 824 }, { "epoch": 0.06, "learning_rate": 9.935309338979065e-05, "loss": 3.2823, "step": 825 }, { "epoch": 0.06, "learning_rate": 9.935230926056615e-05, "loss": 3.0768, "step": 826 }, { "epoch": 0.06, "learning_rate": 9.935152513134165e-05, "loss": 3.2494, "step": 827 }, { "epoch": 0.06, "learning_rate": 9.935074100211716e-05, "loss": 3.1048, "step": 828 }, { "epoch": 0.07, "learning_rate": 9.934995687289266e-05, "loss": 3.3804, "step": 829 }, { "epoch": 0.07, "learning_rate": 9.934917274366817e-05, "loss": 3.2666, "step": 830 }, { "epoch": 0.07, "learning_rate": 9.934838861444367e-05, "loss": 3.4271, "step": 831 }, { "epoch": 0.07, "learning_rate": 9.934760448521917e-05, "loss": 3.1553, "step": 832 }, { "epoch": 0.07, "learning_rate": 9.934682035599468e-05, "loss": 3.5189, "step": 833 }, { "epoch": 0.07, "learning_rate": 9.934603622677018e-05, "loss": 3.2785, "step": 834 }, { "epoch": 0.07, "learning_rate": 9.934525209754568e-05, "loss": 2.9601, "step": 835 }, { "epoch": 0.07, "learning_rate": 9.934446796832119e-05, "loss": 3.1148, "step": 836 }, { "epoch": 0.07, "learning_rate": 9.934368383909669e-05, "loss": 3.4372, "step": 837 }, { "epoch": 0.07, "learning_rate": 9.93428997098722e-05, "loss": 3.2099, "step": 838 }, { "epoch": 0.07, "learning_rate": 9.93421155806477e-05, "loss": 3.0602, "step": 839 }, { "epoch": 0.07, "learning_rate": 9.93413314514232e-05, "loss": 3.0891, "step": 840 }, { "epoch": 0.07, "learning_rate": 9.93405473221987e-05, "loss": 3.0219, "step": 841 }, { "epoch": 0.07, "learning_rate": 9.933976319297421e-05, "loss": 3.2251, "step": 842 }, { "epoch": 0.07, "learning_rate": 9.933897906374971e-05, "loss": 3.0633, "step": 843 }, { "epoch": 0.07, "learning_rate": 9.933819493452522e-05, "loss": 2.9953, "step": 844 }, { "epoch": 0.07, "learning_rate": 9.933741080530072e-05, "loss": 3.1255, "step": 845 }, { "epoch": 0.07, "learning_rate": 9.933662667607622e-05, "loss": 3.1666, "step": 846 }, { "epoch": 0.07, "learning_rate": 9.933584254685171e-05, "loss": 3.3603, "step": 847 }, { "epoch": 0.07, "learning_rate": 9.933505841762722e-05, "loss": 3.2183, "step": 848 }, { "epoch": 0.07, "learning_rate": 9.933427428840274e-05, "loss": 3.0996, "step": 849 }, { "epoch": 0.07, "learning_rate": 9.933349015917824e-05, "loss": 3.0944, "step": 850 }, { "epoch": 0.07, "learning_rate": 9.933270602995374e-05, "loss": 3.2057, "step": 851 }, { "epoch": 0.07, "learning_rate": 9.933192190072925e-05, "loss": 3.2211, "step": 852 }, { "epoch": 0.07, "learning_rate": 9.933113777150475e-05, "loss": 3.1333, "step": 853 }, { "epoch": 0.07, "learning_rate": 9.933035364228025e-05, "loss": 3.1046, "step": 854 }, { "epoch": 0.07, "learning_rate": 9.932956951305576e-05, "loss": 3.2272, "step": 855 }, { "epoch": 0.07, "learning_rate": 9.932878538383126e-05, "loss": 3.2972, "step": 856 }, { "epoch": 0.07, "learning_rate": 9.932800125460676e-05, "loss": 2.8363, "step": 857 }, { "epoch": 0.07, "learning_rate": 9.932721712538227e-05, "loss": 3.0727, "step": 858 }, { "epoch": 0.07, "learning_rate": 9.932643299615777e-05, "loss": 3.1189, "step": 859 }, { "epoch": 0.07, "learning_rate": 9.932564886693328e-05, "loss": 3.3167, "step": 860 }, { "epoch": 0.07, "learning_rate": 9.932486473770878e-05, "loss": 3.0787, "step": 861 }, { "epoch": 0.07, "learning_rate": 9.932408060848428e-05, "loss": 3.3439, "step": 862 }, { "epoch": 0.07, "learning_rate": 9.932329647925979e-05, "loss": 3.0556, "step": 863 }, { "epoch": 0.07, "learning_rate": 9.932251235003529e-05, "loss": 3.0321, "step": 864 }, { "epoch": 0.07, "learning_rate": 9.93217282208108e-05, "loss": 3.2196, "step": 865 }, { "epoch": 0.07, "learning_rate": 9.93209440915863e-05, "loss": 3.4268, "step": 866 }, { "epoch": 0.07, "learning_rate": 9.93201599623618e-05, "loss": 3.1419, "step": 867 }, { "epoch": 0.07, "learning_rate": 9.93193758331373e-05, "loss": 3.0178, "step": 868 }, { "epoch": 0.07, "learning_rate": 9.931859170391281e-05, "loss": 3.1989, "step": 869 }, { "epoch": 0.07, "learning_rate": 9.931780757468831e-05, "loss": 3.2595, "step": 870 }, { "epoch": 0.07, "learning_rate": 9.931702344546382e-05, "loss": 3.2932, "step": 871 }, { "epoch": 0.07, "learning_rate": 9.931623931623932e-05, "loss": 3.0702, "step": 872 }, { "epoch": 0.07, "learning_rate": 9.931545518701482e-05, "loss": 3.0661, "step": 873 }, { "epoch": 0.07, "learning_rate": 9.931467105779033e-05, "loss": 3.0366, "step": 874 }, { "epoch": 0.07, "learning_rate": 9.931388692856583e-05, "loss": 3.3101, "step": 875 }, { "epoch": 0.07, "learning_rate": 9.931310279934133e-05, "loss": 3.636, "step": 876 }, { "epoch": 0.07, "learning_rate": 9.931231867011684e-05, "loss": 3.2578, "step": 877 }, { "epoch": 0.07, "learning_rate": 9.931153454089234e-05, "loss": 3.3904, "step": 878 }, { "epoch": 0.07, "learning_rate": 9.931075041166785e-05, "loss": 3.1025, "step": 879 }, { "epoch": 0.07, "learning_rate": 9.930996628244335e-05, "loss": 3.0579, "step": 880 }, { "epoch": 0.07, "learning_rate": 9.930918215321887e-05, "loss": 2.9384, "step": 881 }, { "epoch": 0.07, "learning_rate": 9.930839802399437e-05, "loss": 3.3605, "step": 882 }, { "epoch": 0.07, "learning_rate": 9.930761389476986e-05, "loss": 2.8693, "step": 883 }, { "epoch": 0.07, "learning_rate": 9.930682976554536e-05, "loss": 3.1319, "step": 884 }, { "epoch": 0.07, "learning_rate": 9.930604563632087e-05, "loss": 3.3206, "step": 885 }, { "epoch": 0.07, "learning_rate": 9.930526150709637e-05, "loss": 3.2885, "step": 886 }, { "epoch": 0.07, "learning_rate": 9.930447737787188e-05, "loss": 3.2446, "step": 887 }, { "epoch": 0.07, "learning_rate": 9.930369324864738e-05, "loss": 3.1065, "step": 888 }, { "epoch": 0.07, "learning_rate": 9.930290911942288e-05, "loss": 3.3146, "step": 889 }, { "epoch": 0.07, "learning_rate": 9.930212499019839e-05, "loss": 3.2227, "step": 890 }, { "epoch": 0.07, "learning_rate": 9.930134086097389e-05, "loss": 2.9824, "step": 891 }, { "epoch": 0.07, "learning_rate": 9.93005567317494e-05, "loss": 2.94, "step": 892 }, { "epoch": 0.07, "learning_rate": 9.92997726025249e-05, "loss": 3.2887, "step": 893 }, { "epoch": 0.07, "learning_rate": 9.92989884733004e-05, "loss": 3.019, "step": 894 }, { "epoch": 0.07, "learning_rate": 9.92982043440759e-05, "loss": 3.1536, "step": 895 }, { "epoch": 0.07, "learning_rate": 9.929742021485141e-05, "loss": 3.0792, "step": 896 }, { "epoch": 0.07, "learning_rate": 9.929663608562691e-05, "loss": 2.9859, "step": 897 }, { "epoch": 0.07, "learning_rate": 9.929585195640242e-05, "loss": 3.2677, "step": 898 }, { "epoch": 0.07, "learning_rate": 9.929506782717792e-05, "loss": 3.0683, "step": 899 }, { "epoch": 0.07, "learning_rate": 9.929428369795342e-05, "loss": 3.0071, "step": 900 }, { "epoch": 0.07, "learning_rate": 9.929349956872893e-05, "loss": 3.1736, "step": 901 }, { "epoch": 0.07, "learning_rate": 9.929271543950443e-05, "loss": 3.3013, "step": 902 }, { "epoch": 0.07, "learning_rate": 9.929193131027993e-05, "loss": 3.1547, "step": 903 }, { "epoch": 0.07, "learning_rate": 9.929114718105544e-05, "loss": 2.9915, "step": 904 }, { "epoch": 0.07, "learning_rate": 9.929036305183094e-05, "loss": 3.2034, "step": 905 }, { "epoch": 0.07, "learning_rate": 9.928957892260646e-05, "loss": 3.384, "step": 906 }, { "epoch": 0.07, "learning_rate": 9.928879479338196e-05, "loss": 3.0034, "step": 907 }, { "epoch": 0.07, "learning_rate": 9.928801066415747e-05, "loss": 3.2853, "step": 908 }, { "epoch": 0.07, "learning_rate": 9.928722653493297e-05, "loss": 3.0974, "step": 909 }, { "epoch": 0.07, "learning_rate": 9.928644240570847e-05, "loss": 3.0104, "step": 910 }, { "epoch": 0.07, "learning_rate": 9.928565827648398e-05, "loss": 3.2462, "step": 911 }, { "epoch": 0.07, "learning_rate": 9.928487414725948e-05, "loss": 3.1186, "step": 912 }, { "epoch": 0.07, "learning_rate": 9.928409001803498e-05, "loss": 3.188, "step": 913 }, { "epoch": 0.07, "learning_rate": 9.928330588881049e-05, "loss": 3.0224, "step": 914 }, { "epoch": 0.07, "learning_rate": 9.928252175958599e-05, "loss": 3.4843, "step": 915 }, { "epoch": 0.07, "learning_rate": 9.92817376303615e-05, "loss": 3.1032, "step": 916 }, { "epoch": 0.07, "learning_rate": 9.9280953501137e-05, "loss": 3.2291, "step": 917 }, { "epoch": 0.07, "learning_rate": 9.92801693719125e-05, "loss": 3.2696, "step": 918 }, { "epoch": 0.07, "learning_rate": 9.927938524268799e-05, "loss": 3.0309, "step": 919 }, { "epoch": 0.07, "learning_rate": 9.92786011134635e-05, "loss": 3.2214, "step": 920 }, { "epoch": 0.07, "learning_rate": 9.9277816984239e-05, "loss": 3.0251, "step": 921 }, { "epoch": 0.07, "learning_rate": 9.92770328550145e-05, "loss": 3.14, "step": 922 }, { "epoch": 0.07, "learning_rate": 9.927624872579001e-05, "loss": 3.3153, "step": 923 }, { "epoch": 0.07, "learning_rate": 9.927546459656551e-05, "loss": 3.2242, "step": 924 }, { "epoch": 0.07, "learning_rate": 9.927468046734102e-05, "loss": 3.1054, "step": 925 }, { "epoch": 0.07, "learning_rate": 9.927389633811652e-05, "loss": 3.033, "step": 926 }, { "epoch": 0.07, "learning_rate": 9.927311220889202e-05, "loss": 3.1239, "step": 927 }, { "epoch": 0.07, "learning_rate": 9.927232807966753e-05, "loss": 3.6545, "step": 928 }, { "epoch": 0.07, "learning_rate": 9.927154395044303e-05, "loss": 3.1246, "step": 929 }, { "epoch": 0.07, "learning_rate": 9.927075982121853e-05, "loss": 3.1513, "step": 930 }, { "epoch": 0.07, "learning_rate": 9.926997569199404e-05, "loss": 3.0133, "step": 931 }, { "epoch": 0.07, "learning_rate": 9.926919156276955e-05, "loss": 3.151, "step": 932 }, { "epoch": 0.07, "learning_rate": 9.926840743354506e-05, "loss": 3.0827, "step": 933 }, { "epoch": 0.07, "learning_rate": 9.926762330432056e-05, "loss": 3.0602, "step": 934 }, { "epoch": 0.07, "learning_rate": 9.926683917509607e-05, "loss": 3.0564, "step": 935 }, { "epoch": 0.07, "learning_rate": 9.926605504587157e-05, "loss": 3.0217, "step": 936 }, { "epoch": 0.07, "learning_rate": 9.926527091664707e-05, "loss": 3.2422, "step": 937 }, { "epoch": 0.07, "learning_rate": 9.926448678742258e-05, "loss": 3.0216, "step": 938 }, { "epoch": 0.07, "learning_rate": 9.926370265819808e-05, "loss": 3.2622, "step": 939 }, { "epoch": 0.07, "learning_rate": 9.926291852897358e-05, "loss": 3.1299, "step": 940 }, { "epoch": 0.07, "learning_rate": 9.926213439974909e-05, "loss": 3.2178, "step": 941 }, { "epoch": 0.07, "learning_rate": 9.926135027052459e-05, "loss": 3.1465, "step": 942 }, { "epoch": 0.07, "learning_rate": 9.92605661413001e-05, "loss": 3.1921, "step": 943 }, { "epoch": 0.07, "learning_rate": 9.92597820120756e-05, "loss": 2.9515, "step": 944 }, { "epoch": 0.07, "learning_rate": 9.92589978828511e-05, "loss": 3.1412, "step": 945 }, { "epoch": 0.07, "learning_rate": 9.92582137536266e-05, "loss": 3.1832, "step": 946 }, { "epoch": 0.07, "learning_rate": 9.925742962440211e-05, "loss": 3.1132, "step": 947 }, { "epoch": 0.07, "learning_rate": 9.925664549517761e-05, "loss": 3.2382, "step": 948 }, { "epoch": 0.07, "learning_rate": 9.925586136595312e-05, "loss": 3.306, "step": 949 }, { "epoch": 0.07, "learning_rate": 9.925507723672862e-05, "loss": 3.1241, "step": 950 }, { "epoch": 0.07, "learning_rate": 9.925429310750412e-05, "loss": 3.1953, "step": 951 }, { "epoch": 0.07, "learning_rate": 9.925350897827963e-05, "loss": 2.9679, "step": 952 }, { "epoch": 0.07, "learning_rate": 9.925272484905513e-05, "loss": 3.0339, "step": 953 }, { "epoch": 0.07, "learning_rate": 9.925194071983064e-05, "loss": 3.2348, "step": 954 }, { "epoch": 0.07, "learning_rate": 9.925115659060613e-05, "loss": 2.9753, "step": 955 }, { "epoch": 0.07, "learning_rate": 9.925037246138163e-05, "loss": 3.1005, "step": 956 }, { "epoch": 0.08, "learning_rate": 9.924958833215715e-05, "loss": 2.9713, "step": 957 }, { "epoch": 0.08, "learning_rate": 9.924880420293265e-05, "loss": 3.0893, "step": 958 }, { "epoch": 0.08, "learning_rate": 9.924802007370815e-05, "loss": 3.242, "step": 959 }, { "epoch": 0.08, "learning_rate": 9.924723594448366e-05, "loss": 3.0999, "step": 960 }, { "epoch": 0.08, "learning_rate": 9.924645181525916e-05, "loss": 3.0007, "step": 961 }, { "epoch": 0.08, "learning_rate": 9.924566768603466e-05, "loss": 3.0273, "step": 962 }, { "epoch": 0.08, "learning_rate": 9.924488355681017e-05, "loss": 3.135, "step": 963 }, { "epoch": 0.08, "learning_rate": 9.924409942758567e-05, "loss": 3.2212, "step": 964 }, { "epoch": 0.08, "learning_rate": 9.924331529836118e-05, "loss": 3.3647, "step": 965 }, { "epoch": 0.08, "learning_rate": 9.924253116913668e-05, "loss": 3.2161, "step": 966 }, { "epoch": 0.08, "learning_rate": 9.924174703991218e-05, "loss": 2.9059, "step": 967 }, { "epoch": 0.08, "learning_rate": 9.924096291068769e-05, "loss": 3.2692, "step": 968 }, { "epoch": 0.08, "learning_rate": 9.924017878146319e-05, "loss": 2.9946, "step": 969 }, { "epoch": 0.08, "learning_rate": 9.92393946522387e-05, "loss": 3.1492, "step": 970 }, { "epoch": 0.08, "learning_rate": 9.92386105230142e-05, "loss": 3.1379, "step": 971 }, { "epoch": 0.08, "learning_rate": 9.92378263937897e-05, "loss": 3.3149, "step": 972 }, { "epoch": 0.08, "learning_rate": 9.92370422645652e-05, "loss": 3.2849, "step": 973 }, { "epoch": 0.08, "learning_rate": 9.923625813534071e-05, "loss": 3.398, "step": 974 }, { "epoch": 0.08, "learning_rate": 9.923547400611621e-05, "loss": 3.2507, "step": 975 }, { "epoch": 0.08, "learning_rate": 9.923468987689172e-05, "loss": 3.2178, "step": 976 }, { "epoch": 0.08, "learning_rate": 9.923390574766722e-05, "loss": 3.1213, "step": 977 }, { "epoch": 0.08, "learning_rate": 9.923312161844272e-05, "loss": 3.7974, "step": 978 }, { "epoch": 0.08, "learning_rate": 9.923233748921823e-05, "loss": 3.4064, "step": 979 }, { "epoch": 0.08, "learning_rate": 9.923155335999373e-05, "loss": 3.1552, "step": 980 }, { "epoch": 0.08, "learning_rate": 9.923076923076923e-05, "loss": 3.2644, "step": 981 }, { "epoch": 0.08, "learning_rate": 9.922998510154474e-05, "loss": 3.0139, "step": 982 }, { "epoch": 0.08, "learning_rate": 9.922920097232024e-05, "loss": 2.9899, "step": 983 }, { "epoch": 0.08, "learning_rate": 9.922841684309575e-05, "loss": 3.3665, "step": 984 }, { "epoch": 0.08, "learning_rate": 9.922763271387125e-05, "loss": 3.3634, "step": 985 }, { "epoch": 0.08, "learning_rate": 9.922684858464675e-05, "loss": 3.2983, "step": 986 }, { "epoch": 0.08, "learning_rate": 9.922606445542226e-05, "loss": 3.1539, "step": 987 }, { "epoch": 0.08, "learning_rate": 9.922528032619776e-05, "loss": 3.296, "step": 988 }, { "epoch": 0.08, "learning_rate": 9.922449619697328e-05, "loss": 3.4118, "step": 989 }, { "epoch": 0.08, "learning_rate": 9.922371206774878e-05, "loss": 3.4181, "step": 990 }, { "epoch": 0.08, "learning_rate": 9.922292793852427e-05, "loss": 3.5677, "step": 991 }, { "epoch": 0.08, "learning_rate": 9.922214380929978e-05, "loss": 3.0084, "step": 992 }, { "epoch": 0.08, "learning_rate": 9.922135968007528e-05, "loss": 3.1574, "step": 993 }, { "epoch": 0.08, "learning_rate": 9.922057555085078e-05, "loss": 3.2769, "step": 994 }, { "epoch": 0.08, "learning_rate": 9.921979142162629e-05, "loss": 3.4353, "step": 995 }, { "epoch": 0.08, "learning_rate": 9.921900729240179e-05, "loss": 3.0839, "step": 996 }, { "epoch": 0.08, "learning_rate": 9.92182231631773e-05, "loss": 3.0128, "step": 997 }, { "epoch": 0.08, "learning_rate": 9.92174390339528e-05, "loss": 2.9902, "step": 998 }, { "epoch": 0.08, "learning_rate": 9.92166549047283e-05, "loss": 2.9693, "step": 999 }, { "epoch": 0.08, "learning_rate": 9.92158707755038e-05, "loss": 3.046, "step": 1000 }, { "epoch": 0.08, "eval_loss": 3.048098087310791, "eval_runtime": 218.4479, "eval_samples_per_second": 51.875, "eval_steps_per_second": 3.246, "step": 1000 }, { "epoch": 0.08, "learning_rate": 9.921508664627931e-05, "loss": 3.1005, "step": 1001 }, { "epoch": 0.08, "learning_rate": 9.921430251705481e-05, "loss": 3.3615, "step": 1002 }, { "epoch": 0.08, "learning_rate": 9.921351838783032e-05, "loss": 3.0515, "step": 1003 }, { "epoch": 0.08, "learning_rate": 9.921273425860582e-05, "loss": 3.0693, "step": 1004 }, { "epoch": 0.08, "learning_rate": 9.921195012938132e-05, "loss": 2.996, "step": 1005 }, { "epoch": 0.08, "learning_rate": 9.921116600015683e-05, "loss": 3.0938, "step": 1006 }, { "epoch": 0.08, "learning_rate": 9.921038187093233e-05, "loss": 3.0525, "step": 1007 }, { "epoch": 0.08, "learning_rate": 9.920959774170783e-05, "loss": 3.2096, "step": 1008 }, { "epoch": 0.08, "learning_rate": 9.920881361248334e-05, "loss": 2.9976, "step": 1009 }, { "epoch": 0.08, "learning_rate": 9.920802948325884e-05, "loss": 3.2065, "step": 1010 }, { "epoch": 0.08, "learning_rate": 9.920724535403434e-05, "loss": 3.1199, "step": 1011 }, { "epoch": 0.08, "learning_rate": 9.920646122480985e-05, "loss": 2.9808, "step": 1012 }, { "epoch": 0.08, "learning_rate": 9.920567709558535e-05, "loss": 2.9599, "step": 1013 }, { "epoch": 0.08, "learning_rate": 9.920489296636087e-05, "loss": 3.0812, "step": 1014 }, { "epoch": 0.08, "learning_rate": 9.920410883713637e-05, "loss": 3.087, "step": 1015 }, { "epoch": 0.08, "learning_rate": 9.920332470791188e-05, "loss": 2.864, "step": 1016 }, { "epoch": 0.08, "learning_rate": 9.920254057868738e-05, "loss": 3.0776, "step": 1017 }, { "epoch": 0.08, "learning_rate": 9.920175644946288e-05, "loss": 3.4207, "step": 1018 }, { "epoch": 0.08, "learning_rate": 9.920097232023839e-05, "loss": 3.1728, "step": 1019 }, { "epoch": 0.08, "learning_rate": 9.920018819101389e-05, "loss": 2.8753, "step": 1020 }, { "epoch": 0.08, "learning_rate": 9.91994040617894e-05, "loss": 3.2256, "step": 1021 }, { "epoch": 0.08, "learning_rate": 9.91986199325649e-05, "loss": 3.1326, "step": 1022 }, { "epoch": 0.08, "learning_rate": 9.91978358033404e-05, "loss": 2.9921, "step": 1023 }, { "epoch": 0.08, "learning_rate": 9.91970516741159e-05, "loss": 3.0245, "step": 1024 }, { "epoch": 0.08, "learning_rate": 9.919626754489141e-05, "loss": 3.0288, "step": 1025 }, { "epoch": 0.08, "learning_rate": 9.919548341566691e-05, "loss": 3.084, "step": 1026 }, { "epoch": 0.08, "learning_rate": 9.91946992864424e-05, "loss": 3.2312, "step": 1027 }, { "epoch": 0.08, "learning_rate": 9.919391515721791e-05, "loss": 3.3009, "step": 1028 }, { "epoch": 0.08, "learning_rate": 9.919313102799341e-05, "loss": 3.0136, "step": 1029 }, { "epoch": 0.08, "learning_rate": 9.919234689876891e-05, "loss": 3.1758, "step": 1030 }, { "epoch": 0.08, "learning_rate": 9.919156276954442e-05, "loss": 3.0394, "step": 1031 }, { "epoch": 0.08, "learning_rate": 9.919077864031992e-05, "loss": 3.074, "step": 1032 }, { "epoch": 0.08, "learning_rate": 9.918999451109543e-05, "loss": 3.1628, "step": 1033 }, { "epoch": 0.08, "learning_rate": 9.918921038187093e-05, "loss": 2.9588, "step": 1034 }, { "epoch": 0.08, "learning_rate": 9.918842625264643e-05, "loss": 3.1513, "step": 1035 }, { "epoch": 0.08, "learning_rate": 9.918764212342194e-05, "loss": 2.959, "step": 1036 }, { "epoch": 0.08, "learning_rate": 9.918685799419744e-05, "loss": 3.0392, "step": 1037 }, { "epoch": 0.08, "learning_rate": 9.918607386497294e-05, "loss": 3.4355, "step": 1038 }, { "epoch": 0.08, "learning_rate": 9.918528973574845e-05, "loss": 3.2951, "step": 1039 }, { "epoch": 0.08, "learning_rate": 9.918450560652397e-05, "loss": 3.1312, "step": 1040 }, { "epoch": 0.08, "learning_rate": 9.918372147729947e-05, "loss": 3.0272, "step": 1041 }, { "epoch": 0.08, "learning_rate": 9.918293734807497e-05, "loss": 3.1779, "step": 1042 }, { "epoch": 0.08, "learning_rate": 9.918215321885048e-05, "loss": 3.1614, "step": 1043 }, { "epoch": 0.08, "learning_rate": 9.918136908962598e-05, "loss": 3.1943, "step": 1044 }, { "epoch": 0.08, "learning_rate": 9.918058496040148e-05, "loss": 3.0234, "step": 1045 }, { "epoch": 0.08, "learning_rate": 9.917980083117699e-05, "loss": 3.0621, "step": 1046 }, { "epoch": 0.08, "learning_rate": 9.917901670195249e-05, "loss": 3.179, "step": 1047 }, { "epoch": 0.08, "learning_rate": 9.9178232572728e-05, "loss": 3.2246, "step": 1048 }, { "epoch": 0.08, "learning_rate": 9.91774484435035e-05, "loss": 3.0992, "step": 1049 }, { "epoch": 0.08, "learning_rate": 9.9176664314279e-05, "loss": 3.0399, "step": 1050 }, { "epoch": 0.08, "learning_rate": 9.91758801850545e-05, "loss": 3.0214, "step": 1051 }, { "epoch": 0.08, "learning_rate": 9.917509605583001e-05, "loss": 3.0477, "step": 1052 }, { "epoch": 0.08, "learning_rate": 9.917431192660551e-05, "loss": 3.1692, "step": 1053 }, { "epoch": 0.08, "learning_rate": 9.917352779738102e-05, "loss": 3.2354, "step": 1054 }, { "epoch": 0.08, "learning_rate": 9.917274366815652e-05, "loss": 3.0913, "step": 1055 }, { "epoch": 0.08, "learning_rate": 9.917195953893202e-05, "loss": 2.8882, "step": 1056 }, { "epoch": 0.08, "learning_rate": 9.917117540970753e-05, "loss": 3.3529, "step": 1057 }, { "epoch": 0.08, "learning_rate": 9.917039128048303e-05, "loss": 2.8832, "step": 1058 }, { "epoch": 0.08, "learning_rate": 9.916960715125853e-05, "loss": 3.2336, "step": 1059 }, { "epoch": 0.08, "learning_rate": 9.916882302203404e-05, "loss": 2.8813, "step": 1060 }, { "epoch": 0.08, "learning_rate": 9.916803889280954e-05, "loss": 3.0355, "step": 1061 }, { "epoch": 0.08, "learning_rate": 9.916725476358505e-05, "loss": 3.3341, "step": 1062 }, { "epoch": 0.08, "learning_rate": 9.916647063436054e-05, "loss": 3.2534, "step": 1063 }, { "epoch": 0.08, "learning_rate": 9.916568650513604e-05, "loss": 3.3719, "step": 1064 }, { "epoch": 0.08, "learning_rate": 9.916490237591156e-05, "loss": 3.242, "step": 1065 }, { "epoch": 0.08, "learning_rate": 9.916411824668706e-05, "loss": 3.2343, "step": 1066 }, { "epoch": 0.08, "learning_rate": 9.916333411746256e-05, "loss": 3.4548, "step": 1067 }, { "epoch": 0.08, "learning_rate": 9.916254998823807e-05, "loss": 3.1417, "step": 1068 }, { "epoch": 0.08, "learning_rate": 9.916176585901357e-05, "loss": 3.3453, "step": 1069 }, { "epoch": 0.08, "learning_rate": 9.916098172978908e-05, "loss": 3.1477, "step": 1070 }, { "epoch": 0.08, "learning_rate": 9.916019760056458e-05, "loss": 3.1447, "step": 1071 }, { "epoch": 0.08, "learning_rate": 9.915941347134008e-05, "loss": 3.1492, "step": 1072 }, { "epoch": 0.08, "learning_rate": 9.915862934211559e-05, "loss": 3.1066, "step": 1073 }, { "epoch": 0.08, "learning_rate": 9.915784521289109e-05, "loss": 2.9798, "step": 1074 }, { "epoch": 0.08, "learning_rate": 9.91570610836666e-05, "loss": 3.0399, "step": 1075 }, { "epoch": 0.08, "learning_rate": 9.91562769544421e-05, "loss": 2.8458, "step": 1076 }, { "epoch": 0.08, "learning_rate": 9.91554928252176e-05, "loss": 3.1453, "step": 1077 }, { "epoch": 0.08, "learning_rate": 9.91547086959931e-05, "loss": 3.2591, "step": 1078 }, { "epoch": 0.08, "learning_rate": 9.915392456676861e-05, "loss": 3.2787, "step": 1079 }, { "epoch": 0.08, "learning_rate": 9.915314043754411e-05, "loss": 3.2179, "step": 1080 }, { "epoch": 0.08, "learning_rate": 9.915235630831962e-05, "loss": 3.1829, "step": 1081 }, { "epoch": 0.08, "learning_rate": 9.915157217909512e-05, "loss": 3.25, "step": 1082 }, { "epoch": 0.08, "learning_rate": 9.915078804987062e-05, "loss": 3.0841, "step": 1083 }, { "epoch": 0.08, "learning_rate": 9.915000392064613e-05, "loss": 3.2853, "step": 1084 }, { "epoch": 0.09, "learning_rate": 9.914921979142163e-05, "loss": 3.2651, "step": 1085 }, { "epoch": 0.09, "learning_rate": 9.914843566219713e-05, "loss": 3.0143, "step": 1086 }, { "epoch": 0.09, "learning_rate": 9.914765153297264e-05, "loss": 3.0046, "step": 1087 }, { "epoch": 0.09, "learning_rate": 9.914686740374814e-05, "loss": 3.078, "step": 1088 }, { "epoch": 0.09, "learning_rate": 9.914608327452365e-05, "loss": 3.0602, "step": 1089 }, { "epoch": 0.09, "learning_rate": 9.914529914529915e-05, "loss": 3.2391, "step": 1090 }, { "epoch": 0.09, "learning_rate": 9.914451501607465e-05, "loss": 3.1273, "step": 1091 }, { "epoch": 0.09, "learning_rate": 9.914373088685016e-05, "loss": 3.1678, "step": 1092 }, { "epoch": 0.09, "learning_rate": 9.914294675762566e-05, "loss": 3.3928, "step": 1093 }, { "epoch": 0.09, "learning_rate": 9.914216262840116e-05, "loss": 3.2411, "step": 1094 }, { "epoch": 0.09, "learning_rate": 9.914137849917667e-05, "loss": 3.0639, "step": 1095 }, { "epoch": 0.09, "learning_rate": 9.914059436995217e-05, "loss": 3.0436, "step": 1096 }, { "epoch": 0.09, "learning_rate": 9.913981024072769e-05, "loss": 2.8827, "step": 1097 }, { "epoch": 0.09, "learning_rate": 9.913902611150319e-05, "loss": 3.1258, "step": 1098 }, { "epoch": 0.09, "learning_rate": 9.913824198227868e-05, "loss": 3.3534, "step": 1099 }, { "epoch": 0.09, "learning_rate": 9.913745785305419e-05, "loss": 3.2099, "step": 1100 }, { "epoch": 0.09, "learning_rate": 9.913667372382969e-05, "loss": 3.1454, "step": 1101 }, { "epoch": 0.09, "learning_rate": 9.913588959460519e-05, "loss": 3.3792, "step": 1102 }, { "epoch": 0.09, "learning_rate": 9.91351054653807e-05, "loss": 3.2823, "step": 1103 }, { "epoch": 0.09, "learning_rate": 9.91343213361562e-05, "loss": 3.3135, "step": 1104 }, { "epoch": 0.09, "learning_rate": 9.91335372069317e-05, "loss": 3.0765, "step": 1105 }, { "epoch": 0.09, "learning_rate": 9.913275307770721e-05, "loss": 2.9361, "step": 1106 }, { "epoch": 0.09, "learning_rate": 9.913196894848271e-05, "loss": 3.3237, "step": 1107 }, { "epoch": 0.09, "learning_rate": 9.913118481925822e-05, "loss": 3.136, "step": 1108 }, { "epoch": 0.09, "learning_rate": 9.913040069003372e-05, "loss": 3.1083, "step": 1109 }, { "epoch": 0.09, "learning_rate": 9.912961656080922e-05, "loss": 3.2438, "step": 1110 }, { "epoch": 0.09, "learning_rate": 9.912883243158473e-05, "loss": 3.1338, "step": 1111 }, { "epoch": 0.09, "learning_rate": 9.912804830236023e-05, "loss": 3.1264, "step": 1112 }, { "epoch": 0.09, "learning_rate": 9.912726417313573e-05, "loss": 3.348, "step": 1113 }, { "epoch": 0.09, "learning_rate": 9.912648004391124e-05, "loss": 3.2594, "step": 1114 }, { "epoch": 0.09, "learning_rate": 9.912569591468674e-05, "loss": 3.1072, "step": 1115 }, { "epoch": 0.09, "learning_rate": 9.912491178546224e-05, "loss": 3.195, "step": 1116 }, { "epoch": 0.09, "learning_rate": 9.912412765623775e-05, "loss": 2.9228, "step": 1117 }, { "epoch": 0.09, "learning_rate": 9.912334352701325e-05, "loss": 3.1718, "step": 1118 }, { "epoch": 0.09, "learning_rate": 9.912255939778876e-05, "loss": 3.0529, "step": 1119 }, { "epoch": 0.09, "learning_rate": 9.912177526856426e-05, "loss": 3.1118, "step": 1120 }, { "epoch": 0.09, "learning_rate": 9.912099113933976e-05, "loss": 3.1525, "step": 1121 }, { "epoch": 0.09, "learning_rate": 9.912020701011528e-05, "loss": 3.1969, "step": 1122 }, { "epoch": 0.09, "learning_rate": 9.911942288089078e-05, "loss": 3.3079, "step": 1123 }, { "epoch": 0.09, "learning_rate": 9.911863875166629e-05, "loss": 3.072, "step": 1124 }, { "epoch": 0.09, "learning_rate": 9.911785462244179e-05, "loss": 3.4167, "step": 1125 }, { "epoch": 0.09, "learning_rate": 9.91170704932173e-05, "loss": 3.1173, "step": 1126 }, { "epoch": 0.09, "learning_rate": 9.91162863639928e-05, "loss": 3.1789, "step": 1127 }, { "epoch": 0.09, "learning_rate": 9.91155022347683e-05, "loss": 3.0432, "step": 1128 }, { "epoch": 0.09, "learning_rate": 9.91147181055438e-05, "loss": 2.9734, "step": 1129 }, { "epoch": 0.09, "learning_rate": 9.911393397631931e-05, "loss": 3.3223, "step": 1130 }, { "epoch": 0.09, "learning_rate": 9.911314984709481e-05, "loss": 3.3611, "step": 1131 }, { "epoch": 0.09, "learning_rate": 9.911236571787032e-05, "loss": 3.0859, "step": 1132 }, { "epoch": 0.09, "learning_rate": 9.911158158864582e-05, "loss": 3.2421, "step": 1133 }, { "epoch": 0.09, "learning_rate": 9.911079745942132e-05, "loss": 3.1917, "step": 1134 }, { "epoch": 0.09, "learning_rate": 9.911001333019681e-05, "loss": 3.1164, "step": 1135 }, { "epoch": 0.09, "learning_rate": 9.910922920097232e-05, "loss": 3.2174, "step": 1136 }, { "epoch": 0.09, "learning_rate": 9.910844507174782e-05, "loss": 2.9587, "step": 1137 }, { "epoch": 0.09, "learning_rate": 9.910766094252333e-05, "loss": 3.2715, "step": 1138 }, { "epoch": 0.09, "learning_rate": 9.910687681329883e-05, "loss": 3.4484, "step": 1139 }, { "epoch": 0.09, "learning_rate": 9.910609268407433e-05, "loss": 2.8751, "step": 1140 }, { "epoch": 0.09, "learning_rate": 9.910530855484984e-05, "loss": 2.9412, "step": 1141 }, { "epoch": 0.09, "learning_rate": 9.910452442562534e-05, "loss": 3.1147, "step": 1142 }, { "epoch": 0.09, "learning_rate": 9.910374029640084e-05, "loss": 3.0409, "step": 1143 }, { "epoch": 0.09, "learning_rate": 9.910295616717635e-05, "loss": 3.0636, "step": 1144 }, { "epoch": 0.09, "learning_rate": 9.910217203795185e-05, "loss": 3.2887, "step": 1145 }, { "epoch": 0.09, "learning_rate": 9.910138790872735e-05, "loss": 3.2072, "step": 1146 }, { "epoch": 0.09, "learning_rate": 9.910060377950286e-05, "loss": 3.316, "step": 1147 }, { "epoch": 0.09, "learning_rate": 9.909981965027838e-05, "loss": 3.0733, "step": 1148 }, { "epoch": 0.09, "learning_rate": 9.909903552105388e-05, "loss": 2.9727, "step": 1149 }, { "epoch": 0.09, "learning_rate": 9.909825139182938e-05, "loss": 2.9355, "step": 1150 }, { "epoch": 0.09, "learning_rate": 9.909746726260489e-05, "loss": 3.3606, "step": 1151 }, { "epoch": 0.09, "learning_rate": 9.909668313338039e-05, "loss": 3.2749, "step": 1152 }, { "epoch": 0.09, "learning_rate": 9.90958990041559e-05, "loss": 3.2376, "step": 1153 }, { "epoch": 0.09, "learning_rate": 9.90951148749314e-05, "loss": 3.469, "step": 1154 }, { "epoch": 0.09, "learning_rate": 9.90943307457069e-05, "loss": 3.1385, "step": 1155 }, { "epoch": 0.09, "learning_rate": 9.90935466164824e-05, "loss": 3.3018, "step": 1156 }, { "epoch": 0.09, "learning_rate": 9.909276248725791e-05, "loss": 3.1213, "step": 1157 }, { "epoch": 0.09, "learning_rate": 9.909197835803341e-05, "loss": 3.0108, "step": 1158 }, { "epoch": 0.09, "learning_rate": 9.909119422880892e-05, "loss": 3.2823, "step": 1159 }, { "epoch": 0.09, "learning_rate": 9.909041009958442e-05, "loss": 3.1965, "step": 1160 }, { "epoch": 0.09, "learning_rate": 9.908962597035992e-05, "loss": 3.0568, "step": 1161 }, { "epoch": 0.09, "learning_rate": 9.908884184113543e-05, "loss": 3.3142, "step": 1162 }, { "epoch": 0.09, "learning_rate": 9.908805771191093e-05, "loss": 3.0464, "step": 1163 }, { "epoch": 0.09, "learning_rate": 9.908727358268643e-05, "loss": 3.2571, "step": 1164 }, { "epoch": 0.09, "learning_rate": 9.908648945346194e-05, "loss": 2.9744, "step": 1165 }, { "epoch": 0.09, "learning_rate": 9.908570532423744e-05, "loss": 3.115, "step": 1166 }, { "epoch": 0.09, "learning_rate": 9.908492119501295e-05, "loss": 3.1051, "step": 1167 }, { "epoch": 0.09, "learning_rate": 9.908413706578845e-05, "loss": 3.0886, "step": 1168 }, { "epoch": 0.09, "learning_rate": 9.908335293656395e-05, "loss": 3.1187, "step": 1169 }, { "epoch": 0.09, "learning_rate": 9.908256880733946e-05, "loss": 2.9958, "step": 1170 }, { "epoch": 0.09, "learning_rate": 9.908178467811495e-05, "loss": 3.1953, "step": 1171 }, { "epoch": 0.09, "learning_rate": 9.908100054889045e-05, "loss": 3.3605, "step": 1172 }, { "epoch": 0.09, "learning_rate": 9.908021641966595e-05, "loss": 3.3837, "step": 1173 }, { "epoch": 0.09, "learning_rate": 9.907943229044147e-05, "loss": 3.2705, "step": 1174 }, { "epoch": 0.09, "learning_rate": 9.907864816121698e-05, "loss": 3.1538, "step": 1175 }, { "epoch": 0.09, "learning_rate": 9.907786403199248e-05, "loss": 3.3109, "step": 1176 }, { "epoch": 0.09, "learning_rate": 9.907707990276798e-05, "loss": 3.0772, "step": 1177 }, { "epoch": 0.09, "learning_rate": 9.907629577354349e-05, "loss": 3.3141, "step": 1178 }, { "epoch": 0.09, "learning_rate": 9.907551164431899e-05, "loss": 3.219, "step": 1179 }, { "epoch": 0.09, "learning_rate": 9.90747275150945e-05, "loss": 3.2049, "step": 1180 }, { "epoch": 0.09, "learning_rate": 9.907394338587e-05, "loss": 3.2614, "step": 1181 }, { "epoch": 0.09, "learning_rate": 9.90731592566455e-05, "loss": 3.2138, "step": 1182 }, { "epoch": 0.09, "learning_rate": 9.9072375127421e-05, "loss": 3.0189, "step": 1183 }, { "epoch": 0.09, "learning_rate": 9.907159099819651e-05, "loss": 3.0247, "step": 1184 }, { "epoch": 0.09, "learning_rate": 9.907080686897201e-05, "loss": 3.1606, "step": 1185 }, { "epoch": 0.09, "learning_rate": 9.907002273974752e-05, "loss": 3.1703, "step": 1186 }, { "epoch": 0.09, "learning_rate": 9.906923861052302e-05, "loss": 3.0909, "step": 1187 }, { "epoch": 0.09, "learning_rate": 9.906845448129852e-05, "loss": 3.1489, "step": 1188 }, { "epoch": 0.09, "learning_rate": 9.906767035207403e-05, "loss": 3.1396, "step": 1189 }, { "epoch": 0.09, "learning_rate": 9.906688622284953e-05, "loss": 3.0427, "step": 1190 }, { "epoch": 0.09, "learning_rate": 9.906610209362503e-05, "loss": 3.0138, "step": 1191 }, { "epoch": 0.09, "learning_rate": 9.906531796440054e-05, "loss": 3.2434, "step": 1192 }, { "epoch": 0.09, "learning_rate": 9.906453383517604e-05, "loss": 3.4006, "step": 1193 }, { "epoch": 0.09, "learning_rate": 9.906374970595155e-05, "loss": 3.0379, "step": 1194 }, { "epoch": 0.09, "learning_rate": 9.906296557672705e-05, "loss": 3.1682, "step": 1195 }, { "epoch": 0.09, "learning_rate": 9.906218144750255e-05, "loss": 3.1925, "step": 1196 }, { "epoch": 0.09, "learning_rate": 9.906139731827806e-05, "loss": 3.1834, "step": 1197 }, { "epoch": 0.09, "learning_rate": 9.906061318905356e-05, "loss": 3.2467, "step": 1198 }, { "epoch": 0.09, "learning_rate": 9.905982905982906e-05, "loss": 2.9653, "step": 1199 }, { "epoch": 0.09, "learning_rate": 9.905904493060457e-05, "loss": 3.2994, "step": 1200 }, { "epoch": 0.09, "learning_rate": 9.905826080138007e-05, "loss": 2.9396, "step": 1201 }, { "epoch": 0.09, "learning_rate": 9.905747667215557e-05, "loss": 3.2233, "step": 1202 }, { "epoch": 0.09, "learning_rate": 9.905669254293108e-05, "loss": 3.0326, "step": 1203 }, { "epoch": 0.09, "learning_rate": 9.905590841370658e-05, "loss": 3.0154, "step": 1204 }, { "epoch": 0.09, "learning_rate": 9.90551242844821e-05, "loss": 3.1832, "step": 1205 }, { "epoch": 0.09, "learning_rate": 9.90543401552576e-05, "loss": 2.8659, "step": 1206 }, { "epoch": 0.09, "learning_rate": 9.905355602603309e-05, "loss": 3.2133, "step": 1207 }, { "epoch": 0.09, "learning_rate": 9.90527718968086e-05, "loss": 3.2397, "step": 1208 }, { "epoch": 0.09, "learning_rate": 9.90519877675841e-05, "loss": 3.0715, "step": 1209 }, { "epoch": 0.09, "learning_rate": 9.90512036383596e-05, "loss": 3.1774, "step": 1210 }, { "epoch": 0.09, "learning_rate": 9.905041950913511e-05, "loss": 3.1205, "step": 1211 }, { "epoch": 0.1, "learning_rate": 9.904963537991061e-05, "loss": 3.2301, "step": 1212 }, { "epoch": 0.1, "learning_rate": 9.904885125068611e-05, "loss": 3.1773, "step": 1213 }, { "epoch": 0.1, "learning_rate": 9.904806712146162e-05, "loss": 3.0197, "step": 1214 }, { "epoch": 0.1, "learning_rate": 9.904728299223712e-05, "loss": 3.0759, "step": 1215 }, { "epoch": 0.1, "learning_rate": 9.904649886301263e-05, "loss": 3.1581, "step": 1216 }, { "epoch": 0.1, "learning_rate": 9.904571473378813e-05, "loss": 3.3695, "step": 1217 }, { "epoch": 0.1, "learning_rate": 9.904493060456363e-05, "loss": 3.1095, "step": 1218 }, { "epoch": 0.1, "learning_rate": 9.904414647533914e-05, "loss": 3.1601, "step": 1219 }, { "epoch": 0.1, "learning_rate": 9.904336234611464e-05, "loss": 3.0742, "step": 1220 }, { "epoch": 0.1, "learning_rate": 9.904257821689014e-05, "loss": 3.0648, "step": 1221 }, { "epoch": 0.1, "learning_rate": 9.904179408766565e-05, "loss": 3.4052, "step": 1222 }, { "epoch": 0.1, "learning_rate": 9.904100995844115e-05, "loss": 3.1799, "step": 1223 }, { "epoch": 0.1, "learning_rate": 9.904022582921666e-05, "loss": 3.2003, "step": 1224 }, { "epoch": 0.1, "learning_rate": 9.903944169999216e-05, "loss": 3.0793, "step": 1225 }, { "epoch": 0.1, "learning_rate": 9.903865757076766e-05, "loss": 3.1907, "step": 1226 }, { "epoch": 0.1, "learning_rate": 9.903787344154317e-05, "loss": 2.9451, "step": 1227 }, { "epoch": 0.1, "learning_rate": 9.903708931231867e-05, "loss": 3.0639, "step": 1228 }, { "epoch": 0.1, "learning_rate": 9.903630518309417e-05, "loss": 3.092, "step": 1229 }, { "epoch": 0.1, "learning_rate": 9.903552105386969e-05, "loss": 3.0776, "step": 1230 }, { "epoch": 0.1, "learning_rate": 9.90347369246452e-05, "loss": 3.1549, "step": 1231 }, { "epoch": 0.1, "learning_rate": 9.90339527954207e-05, "loss": 2.9567, "step": 1232 }, { "epoch": 0.1, "learning_rate": 9.90331686661962e-05, "loss": 3.1087, "step": 1233 }, { "epoch": 0.1, "learning_rate": 9.90323845369717e-05, "loss": 3.2541, "step": 1234 }, { "epoch": 0.1, "learning_rate": 9.903160040774721e-05, "loss": 3.1976, "step": 1235 }, { "epoch": 0.1, "learning_rate": 9.903081627852271e-05, "loss": 3.2078, "step": 1236 }, { "epoch": 0.1, "learning_rate": 9.903003214929822e-05, "loss": 3.2802, "step": 1237 }, { "epoch": 0.1, "learning_rate": 9.902924802007372e-05, "loss": 3.065, "step": 1238 }, { "epoch": 0.1, "learning_rate": 9.902846389084922e-05, "loss": 3.0794, "step": 1239 }, { "epoch": 0.1, "learning_rate": 9.902767976162473e-05, "loss": 3.0657, "step": 1240 }, { "epoch": 0.1, "learning_rate": 9.902689563240023e-05, "loss": 3.2023, "step": 1241 }, { "epoch": 0.1, "learning_rate": 9.902611150317574e-05, "loss": 3.1142, "step": 1242 }, { "epoch": 0.1, "learning_rate": 9.902532737395123e-05, "loss": 3.0913, "step": 1243 }, { "epoch": 0.1, "learning_rate": 9.902454324472673e-05, "loss": 3.37, "step": 1244 }, { "epoch": 0.1, "learning_rate": 9.902375911550223e-05, "loss": 3.1897, "step": 1245 }, { "epoch": 0.1, "learning_rate": 9.902297498627774e-05, "loss": 3.0261, "step": 1246 }, { "epoch": 0.1, "learning_rate": 9.902219085705324e-05, "loss": 3.1379, "step": 1247 }, { "epoch": 0.1, "learning_rate": 9.902140672782874e-05, "loss": 3.0622, "step": 1248 }, { "epoch": 0.1, "learning_rate": 9.902062259860425e-05, "loss": 3.2412, "step": 1249 }, { "epoch": 0.1, "learning_rate": 9.901983846937975e-05, "loss": 3.0885, "step": 1250 }, { "epoch": 0.1, "learning_rate": 9.901905434015525e-05, "loss": 3.1694, "step": 1251 }, { "epoch": 0.1, "learning_rate": 9.901827021093076e-05, "loss": 3.1827, "step": 1252 }, { "epoch": 0.1, "learning_rate": 9.901748608170626e-05, "loss": 3.1189, "step": 1253 }, { "epoch": 0.1, "learning_rate": 9.901670195248177e-05, "loss": 3.1126, "step": 1254 }, { "epoch": 0.1, "learning_rate": 9.901591782325727e-05, "loss": 2.9537, "step": 1255 }, { "epoch": 0.1, "learning_rate": 9.901513369403279e-05, "loss": 3.2766, "step": 1256 }, { "epoch": 0.1, "learning_rate": 9.901434956480829e-05, "loss": 3.0692, "step": 1257 }, { "epoch": 0.1, "learning_rate": 9.90135654355838e-05, "loss": 3.1903, "step": 1258 }, { "epoch": 0.1, "learning_rate": 9.90127813063593e-05, "loss": 2.9878, "step": 1259 }, { "epoch": 0.1, "learning_rate": 9.90119971771348e-05, "loss": 3.1295, "step": 1260 }, { "epoch": 0.1, "learning_rate": 9.90112130479103e-05, "loss": 3.2907, "step": 1261 }, { "epoch": 0.1, "learning_rate": 9.901042891868581e-05, "loss": 3.2072, "step": 1262 }, { "epoch": 0.1, "learning_rate": 9.900964478946131e-05, "loss": 3.0621, "step": 1263 }, { "epoch": 0.1, "learning_rate": 9.900886066023682e-05, "loss": 3.4918, "step": 1264 }, { "epoch": 0.1, "learning_rate": 9.900807653101232e-05, "loss": 3.2853, "step": 1265 }, { "epoch": 0.1, "learning_rate": 9.900729240178782e-05, "loss": 3.2459, "step": 1266 }, { "epoch": 0.1, "learning_rate": 9.900650827256333e-05, "loss": 3.5678, "step": 1267 }, { "epoch": 0.1, "learning_rate": 9.900572414333883e-05, "loss": 3.1929, "step": 1268 }, { "epoch": 0.1, "learning_rate": 9.900494001411433e-05, "loss": 3.1237, "step": 1269 }, { "epoch": 0.1, "learning_rate": 9.900415588488984e-05, "loss": 3.1503, "step": 1270 }, { "epoch": 0.1, "learning_rate": 9.900337175566534e-05, "loss": 3.0212, "step": 1271 }, { "epoch": 0.1, "learning_rate": 9.900258762644085e-05, "loss": 3.4976, "step": 1272 }, { "epoch": 0.1, "learning_rate": 9.900180349721635e-05, "loss": 3.1736, "step": 1273 }, { "epoch": 0.1, "learning_rate": 9.900101936799185e-05, "loss": 2.9712, "step": 1274 }, { "epoch": 0.1, "learning_rate": 9.900023523876736e-05, "loss": 3.2315, "step": 1275 }, { "epoch": 0.1, "learning_rate": 9.899945110954286e-05, "loss": 3.2497, "step": 1276 }, { "epoch": 0.1, "learning_rate": 9.899866698031836e-05, "loss": 3.0933, "step": 1277 }, { "epoch": 0.1, "learning_rate": 9.899788285109387e-05, "loss": 3.3819, "step": 1278 }, { "epoch": 0.1, "learning_rate": 9.899709872186936e-05, "loss": 2.9531, "step": 1279 }, { "epoch": 0.1, "learning_rate": 9.899631459264486e-05, "loss": 3.1937, "step": 1280 }, { "epoch": 0.1, "learning_rate": 9.899553046342037e-05, "loss": 3.1368, "step": 1281 }, { "epoch": 0.1, "learning_rate": 9.899474633419588e-05, "loss": 3.0722, "step": 1282 }, { "epoch": 0.1, "learning_rate": 9.899396220497139e-05, "loss": 2.9836, "step": 1283 }, { "epoch": 0.1, "learning_rate": 9.899317807574689e-05, "loss": 3.2394, "step": 1284 }, { "epoch": 0.1, "learning_rate": 9.89923939465224e-05, "loss": 3.4877, "step": 1285 }, { "epoch": 0.1, "learning_rate": 9.89916098172979e-05, "loss": 3.0969, "step": 1286 }, { "epoch": 0.1, "learning_rate": 9.89908256880734e-05, "loss": 2.9977, "step": 1287 }, { "epoch": 0.1, "learning_rate": 9.89900415588489e-05, "loss": 3.0137, "step": 1288 }, { "epoch": 0.1, "learning_rate": 9.898925742962441e-05, "loss": 3.1333, "step": 1289 }, { "epoch": 0.1, "learning_rate": 9.898847330039991e-05, "loss": 3.6581, "step": 1290 }, { "epoch": 0.1, "learning_rate": 9.898768917117542e-05, "loss": 2.9642, "step": 1291 }, { "epoch": 0.1, "learning_rate": 9.898690504195092e-05, "loss": 3.1593, "step": 1292 }, { "epoch": 0.1, "learning_rate": 9.898612091272642e-05, "loss": 3.0208, "step": 1293 }, { "epoch": 0.1, "learning_rate": 9.898533678350193e-05, "loss": 3.1094, "step": 1294 }, { "epoch": 0.1, "learning_rate": 9.898455265427743e-05, "loss": 3.0611, "step": 1295 }, { "epoch": 0.1, "learning_rate": 9.898376852505293e-05, "loss": 3.1328, "step": 1296 }, { "epoch": 0.1, "learning_rate": 9.898298439582844e-05, "loss": 3.217, "step": 1297 }, { "epoch": 0.1, "learning_rate": 9.898220026660394e-05, "loss": 3.1269, "step": 1298 }, { "epoch": 0.1, "learning_rate": 9.898141613737944e-05, "loss": 2.8505, "step": 1299 }, { "epoch": 0.1, "learning_rate": 9.898063200815495e-05, "loss": 3.1151, "step": 1300 }, { "epoch": 0.1, "learning_rate": 9.897984787893045e-05, "loss": 3.0838, "step": 1301 }, { "epoch": 0.1, "learning_rate": 9.897906374970596e-05, "loss": 3.1487, "step": 1302 }, { "epoch": 0.1, "learning_rate": 9.897827962048146e-05, "loss": 3.0878, "step": 1303 }, { "epoch": 0.1, "learning_rate": 9.897749549125696e-05, "loss": 3.3351, "step": 1304 }, { "epoch": 0.1, "learning_rate": 9.897671136203247e-05, "loss": 3.2417, "step": 1305 }, { "epoch": 0.1, "learning_rate": 9.897592723280797e-05, "loss": 3.4475, "step": 1306 }, { "epoch": 0.1, "learning_rate": 9.897514310358347e-05, "loss": 3.0883, "step": 1307 }, { "epoch": 0.1, "learning_rate": 9.897435897435898e-05, "loss": 3.3431, "step": 1308 }, { "epoch": 0.1, "learning_rate": 9.897357484513448e-05, "loss": 3.0734, "step": 1309 }, { "epoch": 0.1, "learning_rate": 9.897279071590999e-05, "loss": 2.9405, "step": 1310 }, { "epoch": 0.1, "learning_rate": 9.897200658668549e-05, "loss": 3.2093, "step": 1311 }, { "epoch": 0.1, "learning_rate": 9.897122245746099e-05, "loss": 3.09, "step": 1312 }, { "epoch": 0.1, "learning_rate": 9.897043832823651e-05, "loss": 2.873, "step": 1313 }, { "epoch": 0.1, "learning_rate": 9.896965419901201e-05, "loss": 3.1915, "step": 1314 }, { "epoch": 0.1, "learning_rate": 9.89688700697875e-05, "loss": 3.3421, "step": 1315 }, { "epoch": 0.1, "learning_rate": 9.896808594056301e-05, "loss": 3.0354, "step": 1316 }, { "epoch": 0.1, "learning_rate": 9.896730181133851e-05, "loss": 3.2733, "step": 1317 }, { "epoch": 0.1, "learning_rate": 9.896651768211401e-05, "loss": 3.1576, "step": 1318 }, { "epoch": 0.1, "learning_rate": 9.896573355288952e-05, "loss": 3.0265, "step": 1319 }, { "epoch": 0.1, "learning_rate": 9.896494942366502e-05, "loss": 3.0631, "step": 1320 }, { "epoch": 0.1, "learning_rate": 9.896416529444053e-05, "loss": 3.2107, "step": 1321 }, { "epoch": 0.1, "learning_rate": 9.896338116521603e-05, "loss": 3.2969, "step": 1322 }, { "epoch": 0.1, "learning_rate": 9.896259703599153e-05, "loss": 3.0479, "step": 1323 }, { "epoch": 0.1, "learning_rate": 9.896181290676704e-05, "loss": 3.0023, "step": 1324 }, { "epoch": 0.1, "learning_rate": 9.896102877754254e-05, "loss": 3.0309, "step": 1325 }, { "epoch": 0.1, "learning_rate": 9.896024464831804e-05, "loss": 3.3326, "step": 1326 }, { "epoch": 0.1, "learning_rate": 9.895946051909355e-05, "loss": 3.0228, "step": 1327 }, { "epoch": 0.1, "learning_rate": 9.895867638986905e-05, "loss": 2.914, "step": 1328 }, { "epoch": 0.1, "learning_rate": 9.895789226064456e-05, "loss": 3.3144, "step": 1329 }, { "epoch": 0.1, "learning_rate": 9.895710813142006e-05, "loss": 3.205, "step": 1330 }, { "epoch": 0.1, "learning_rate": 9.895632400219556e-05, "loss": 3.042, "step": 1331 }, { "epoch": 0.1, "learning_rate": 9.895553987297107e-05, "loss": 3.3372, "step": 1332 }, { "epoch": 0.1, "learning_rate": 9.895475574374657e-05, "loss": 3.1162, "step": 1333 }, { "epoch": 0.1, "learning_rate": 9.895397161452207e-05, "loss": 3.0597, "step": 1334 }, { "epoch": 0.1, "learning_rate": 9.895318748529758e-05, "loss": 3.0133, "step": 1335 }, { "epoch": 0.1, "learning_rate": 9.895240335607308e-05, "loss": 3.1947, "step": 1336 }, { "epoch": 0.1, "learning_rate": 9.895161922684858e-05, "loss": 3.2083, "step": 1337 }, { "epoch": 0.1, "learning_rate": 9.89508350976241e-05, "loss": 3.2457, "step": 1338 }, { "epoch": 0.1, "learning_rate": 9.89500509683996e-05, "loss": 3.2167, "step": 1339 }, { "epoch": 0.11, "learning_rate": 9.894926683917511e-05, "loss": 3.0909, "step": 1340 }, { "epoch": 0.11, "learning_rate": 9.894848270995061e-05, "loss": 2.9283, "step": 1341 }, { "epoch": 0.11, "learning_rate": 9.894769858072612e-05, "loss": 3.194, "step": 1342 }, { "epoch": 0.11, "learning_rate": 9.894691445150162e-05, "loss": 3.1066, "step": 1343 }, { "epoch": 0.11, "learning_rate": 9.894613032227712e-05, "loss": 3.3003, "step": 1344 }, { "epoch": 0.11, "learning_rate": 9.894534619305263e-05, "loss": 3.162, "step": 1345 }, { "epoch": 0.11, "learning_rate": 9.894456206382813e-05, "loss": 3.0064, "step": 1346 }, { "epoch": 0.11, "learning_rate": 9.894377793460363e-05, "loss": 3.0135, "step": 1347 }, { "epoch": 0.11, "learning_rate": 9.894299380537914e-05, "loss": 3.1971, "step": 1348 }, { "epoch": 0.11, "learning_rate": 9.894220967615464e-05, "loss": 3.1649, "step": 1349 }, { "epoch": 0.11, "learning_rate": 9.894142554693015e-05, "loss": 3.1192, "step": 1350 }, { "epoch": 0.11, "learning_rate": 9.894064141770564e-05, "loss": 3.2881, "step": 1351 }, { "epoch": 0.11, "learning_rate": 9.893985728848114e-05, "loss": 3.1851, "step": 1352 }, { "epoch": 0.11, "learning_rate": 9.893907315925664e-05, "loss": 3.0879, "step": 1353 }, { "epoch": 0.11, "learning_rate": 9.893828903003215e-05, "loss": 3.4533, "step": 1354 }, { "epoch": 0.11, "learning_rate": 9.893750490080765e-05, "loss": 2.9808, "step": 1355 }, { "epoch": 0.11, "learning_rate": 9.893672077158315e-05, "loss": 3.1694, "step": 1356 }, { "epoch": 0.11, "learning_rate": 9.893593664235866e-05, "loss": 3.137, "step": 1357 }, { "epoch": 0.11, "learning_rate": 9.893515251313416e-05, "loss": 3.0198, "step": 1358 }, { "epoch": 0.11, "learning_rate": 9.893436838390967e-05, "loss": 3.0707, "step": 1359 }, { "epoch": 0.11, "learning_rate": 9.893358425468517e-05, "loss": 3.2343, "step": 1360 }, { "epoch": 0.11, "learning_rate": 9.893280012546067e-05, "loss": 2.9515, "step": 1361 }, { "epoch": 0.11, "learning_rate": 9.893201599623618e-05, "loss": 3.2129, "step": 1362 }, { "epoch": 0.11, "learning_rate": 9.893123186701168e-05, "loss": 3.1308, "step": 1363 }, { "epoch": 0.11, "learning_rate": 9.89304477377872e-05, "loss": 3.1078, "step": 1364 }, { "epoch": 0.11, "learning_rate": 9.89296636085627e-05, "loss": 2.8625, "step": 1365 }, { "epoch": 0.11, "learning_rate": 9.89288794793382e-05, "loss": 2.9222, "step": 1366 }, { "epoch": 0.11, "learning_rate": 9.892809535011371e-05, "loss": 3.1194, "step": 1367 }, { "epoch": 0.11, "learning_rate": 9.892731122088921e-05, "loss": 3.1381, "step": 1368 }, { "epoch": 0.11, "learning_rate": 9.892652709166472e-05, "loss": 2.9733, "step": 1369 }, { "epoch": 0.11, "learning_rate": 9.892574296244022e-05, "loss": 2.9569, "step": 1370 }, { "epoch": 0.11, "learning_rate": 9.892495883321572e-05, "loss": 3.2632, "step": 1371 }, { "epoch": 0.11, "learning_rate": 9.892417470399123e-05, "loss": 2.9434, "step": 1372 }, { "epoch": 0.11, "learning_rate": 9.892339057476673e-05, "loss": 2.9913, "step": 1373 }, { "epoch": 0.11, "learning_rate": 9.892260644554223e-05, "loss": 2.8915, "step": 1374 }, { "epoch": 0.11, "learning_rate": 9.892182231631774e-05, "loss": 3.3077, "step": 1375 }, { "epoch": 0.11, "learning_rate": 9.892103818709324e-05, "loss": 3.15, "step": 1376 }, { "epoch": 0.11, "learning_rate": 9.892025405786875e-05, "loss": 3.0502, "step": 1377 }, { "epoch": 0.11, "learning_rate": 9.891946992864425e-05, "loss": 3.1927, "step": 1378 }, { "epoch": 0.11, "learning_rate": 9.891868579941975e-05, "loss": 3.0811, "step": 1379 }, { "epoch": 0.11, "learning_rate": 9.891790167019526e-05, "loss": 3.2068, "step": 1380 }, { "epoch": 0.11, "learning_rate": 9.891711754097076e-05, "loss": 2.8836, "step": 1381 }, { "epoch": 0.11, "learning_rate": 9.891633341174626e-05, "loss": 3.2578, "step": 1382 }, { "epoch": 0.11, "learning_rate": 9.891554928252177e-05, "loss": 3.1244, "step": 1383 }, { "epoch": 0.11, "learning_rate": 9.891476515329727e-05, "loss": 3.3884, "step": 1384 }, { "epoch": 0.11, "learning_rate": 9.891398102407277e-05, "loss": 3.0663, "step": 1385 }, { "epoch": 0.11, "learning_rate": 9.891319689484828e-05, "loss": 3.2721, "step": 1386 }, { "epoch": 0.11, "learning_rate": 9.891241276562377e-05, "loss": 3.4986, "step": 1387 }, { "epoch": 0.11, "learning_rate": 9.891162863639927e-05, "loss": 2.9032, "step": 1388 }, { "epoch": 0.11, "learning_rate": 9.891084450717478e-05, "loss": 2.9366, "step": 1389 }, { "epoch": 0.11, "learning_rate": 9.891006037795029e-05, "loss": 3.3108, "step": 1390 }, { "epoch": 0.11, "learning_rate": 9.89092762487258e-05, "loss": 3.5184, "step": 1391 }, { "epoch": 0.11, "learning_rate": 9.89084921195013e-05, "loss": 3.1931, "step": 1392 }, { "epoch": 0.11, "learning_rate": 9.89077079902768e-05, "loss": 3.2166, "step": 1393 }, { "epoch": 0.11, "learning_rate": 9.890692386105231e-05, "loss": 3.2743, "step": 1394 }, { "epoch": 0.11, "learning_rate": 9.890613973182781e-05, "loss": 3.0411, "step": 1395 }, { "epoch": 0.11, "learning_rate": 9.890535560260332e-05, "loss": 3.0895, "step": 1396 }, { "epoch": 0.11, "learning_rate": 9.890457147337882e-05, "loss": 3.011, "step": 1397 }, { "epoch": 0.11, "learning_rate": 9.890378734415432e-05, "loss": 3.4778, "step": 1398 }, { "epoch": 0.11, "learning_rate": 9.890300321492983e-05, "loss": 3.0649, "step": 1399 }, { "epoch": 0.11, "learning_rate": 9.890221908570533e-05, "loss": 3.3379, "step": 1400 }, { "epoch": 0.11, "learning_rate": 9.890143495648083e-05, "loss": 2.9481, "step": 1401 }, { "epoch": 0.11, "learning_rate": 9.890065082725634e-05, "loss": 3.3191, "step": 1402 }, { "epoch": 0.11, "learning_rate": 9.889986669803184e-05, "loss": 3.1337, "step": 1403 }, { "epoch": 0.11, "learning_rate": 9.889908256880734e-05, "loss": 3.1249, "step": 1404 }, { "epoch": 0.11, "learning_rate": 9.889829843958285e-05, "loss": 2.8565, "step": 1405 }, { "epoch": 0.11, "learning_rate": 9.889751431035835e-05, "loss": 3.228, "step": 1406 }, { "epoch": 0.11, "learning_rate": 9.889673018113386e-05, "loss": 3.3677, "step": 1407 }, { "epoch": 0.11, "learning_rate": 9.889594605190936e-05, "loss": 3.1984, "step": 1408 }, { "epoch": 0.11, "learning_rate": 9.889516192268486e-05, "loss": 3.0906, "step": 1409 }, { "epoch": 0.11, "learning_rate": 9.889437779346037e-05, "loss": 2.869, "step": 1410 }, { "epoch": 0.11, "learning_rate": 9.889359366423587e-05, "loss": 3.0972, "step": 1411 }, { "epoch": 0.11, "learning_rate": 9.889280953501137e-05, "loss": 3.0667, "step": 1412 }, { "epoch": 0.11, "learning_rate": 9.889202540578688e-05, "loss": 3.2319, "step": 1413 }, { "epoch": 0.11, "learning_rate": 9.889124127656238e-05, "loss": 3.101, "step": 1414 }, { "epoch": 0.11, "learning_rate": 9.889045714733788e-05, "loss": 3.1895, "step": 1415 }, { "epoch": 0.11, "learning_rate": 9.888967301811339e-05, "loss": 3.0036, "step": 1416 }, { "epoch": 0.11, "learning_rate": 9.888888888888889e-05, "loss": 3.1265, "step": 1417 }, { "epoch": 0.11, "learning_rate": 9.88881047596644e-05, "loss": 3.3617, "step": 1418 }, { "epoch": 0.11, "learning_rate": 9.88873206304399e-05, "loss": 3.2536, "step": 1419 }, { "epoch": 0.11, "learning_rate": 9.88865365012154e-05, "loss": 3.1853, "step": 1420 }, { "epoch": 0.11, "learning_rate": 9.888575237199092e-05, "loss": 2.8626, "step": 1421 }, { "epoch": 0.11, "learning_rate": 9.888496824276642e-05, "loss": 3.189, "step": 1422 }, { "epoch": 0.11, "learning_rate": 9.888418411354191e-05, "loss": 3.0177, "step": 1423 }, { "epoch": 0.11, "learning_rate": 9.888339998431742e-05, "loss": 3.3467, "step": 1424 }, { "epoch": 0.11, "learning_rate": 9.888261585509292e-05, "loss": 3.4226, "step": 1425 }, { "epoch": 0.11, "learning_rate": 9.888183172586843e-05, "loss": 3.2105, "step": 1426 }, { "epoch": 0.11, "learning_rate": 9.888104759664393e-05, "loss": 3.0746, "step": 1427 }, { "epoch": 0.11, "learning_rate": 9.888026346741943e-05, "loss": 3.0987, "step": 1428 }, { "epoch": 0.11, "learning_rate": 9.887947933819494e-05, "loss": 3.0716, "step": 1429 }, { "epoch": 0.11, "learning_rate": 9.887869520897044e-05, "loss": 2.9689, "step": 1430 }, { "epoch": 0.11, "learning_rate": 9.887791107974594e-05, "loss": 3.3007, "step": 1431 }, { "epoch": 0.11, "learning_rate": 9.887712695052145e-05, "loss": 3.0168, "step": 1432 }, { "epoch": 0.11, "learning_rate": 9.887634282129695e-05, "loss": 3.2591, "step": 1433 }, { "epoch": 0.11, "learning_rate": 9.887555869207245e-05, "loss": 3.1083, "step": 1434 }, { "epoch": 0.11, "learning_rate": 9.887477456284796e-05, "loss": 2.915, "step": 1435 }, { "epoch": 0.11, "learning_rate": 9.887399043362346e-05, "loss": 2.9879, "step": 1436 }, { "epoch": 0.11, "learning_rate": 9.887320630439897e-05, "loss": 3.2625, "step": 1437 }, { "epoch": 0.11, "learning_rate": 9.887242217517447e-05, "loss": 3.1101, "step": 1438 }, { "epoch": 0.11, "learning_rate": 9.887163804594997e-05, "loss": 3.0484, "step": 1439 }, { "epoch": 0.11, "learning_rate": 9.887085391672548e-05, "loss": 3.4116, "step": 1440 }, { "epoch": 0.11, "learning_rate": 9.887006978750098e-05, "loss": 3.1763, "step": 1441 }, { "epoch": 0.11, "learning_rate": 9.886928565827648e-05, "loss": 3.3682, "step": 1442 }, { "epoch": 0.11, "learning_rate": 9.886850152905199e-05, "loss": 3.1819, "step": 1443 }, { "epoch": 0.11, "learning_rate": 9.886771739982749e-05, "loss": 2.8793, "step": 1444 }, { "epoch": 0.11, "learning_rate": 9.8866933270603e-05, "loss": 2.9949, "step": 1445 }, { "epoch": 0.11, "learning_rate": 9.88661491413785e-05, "loss": 3.1021, "step": 1446 }, { "epoch": 0.11, "learning_rate": 9.886536501215402e-05, "loss": 3.0546, "step": 1447 }, { "epoch": 0.11, "learning_rate": 9.886458088292952e-05, "loss": 3.041, "step": 1448 }, { "epoch": 0.11, "learning_rate": 9.886379675370502e-05, "loss": 2.9372, "step": 1449 }, { "epoch": 0.11, "learning_rate": 9.886301262448053e-05, "loss": 3.1265, "step": 1450 }, { "epoch": 0.11, "learning_rate": 9.886222849525603e-05, "loss": 3.36, "step": 1451 }, { "epoch": 0.11, "learning_rate": 9.886144436603153e-05, "loss": 3.4156, "step": 1452 }, { "epoch": 0.11, "learning_rate": 9.886066023680704e-05, "loss": 2.9527, "step": 1453 }, { "epoch": 0.11, "learning_rate": 9.885987610758254e-05, "loss": 3.2284, "step": 1454 }, { "epoch": 0.11, "learning_rate": 9.885909197835805e-05, "loss": 3.1903, "step": 1455 }, { "epoch": 0.11, "learning_rate": 9.885830784913355e-05, "loss": 2.9403, "step": 1456 }, { "epoch": 0.11, "learning_rate": 9.885752371990905e-05, "loss": 2.9086, "step": 1457 }, { "epoch": 0.11, "learning_rate": 9.885673959068456e-05, "loss": 2.9385, "step": 1458 }, { "epoch": 0.11, "learning_rate": 9.885595546146005e-05, "loss": 3.083, "step": 1459 }, { "epoch": 0.11, "learning_rate": 9.885517133223555e-05, "loss": 3.0467, "step": 1460 }, { "epoch": 0.11, "learning_rate": 9.885438720301105e-05, "loss": 3.3424, "step": 1461 }, { "epoch": 0.11, "learning_rate": 9.885360307378656e-05, "loss": 3.4081, "step": 1462 }, { "epoch": 0.11, "learning_rate": 9.885281894456206e-05, "loss": 3.0715, "step": 1463 }, { "epoch": 0.11, "learning_rate": 9.885203481533757e-05, "loss": 3.1657, "step": 1464 }, { "epoch": 0.11, "learning_rate": 9.885125068611307e-05, "loss": 3.3453, "step": 1465 }, { "epoch": 0.11, "learning_rate": 9.885046655688857e-05, "loss": 3.1332, "step": 1466 }, { "epoch": 0.12, "learning_rate": 9.884968242766408e-05, "loss": 2.897, "step": 1467 }, { "epoch": 0.12, "learning_rate": 9.884889829843958e-05, "loss": 3.2023, "step": 1468 }, { "epoch": 0.12, "learning_rate": 9.884811416921508e-05, "loss": 2.9796, "step": 1469 }, { "epoch": 0.12, "learning_rate": 9.884733003999059e-05, "loss": 3.0951, "step": 1470 }, { "epoch": 0.12, "learning_rate": 9.884654591076609e-05, "loss": 3.3819, "step": 1471 }, { "epoch": 0.12, "learning_rate": 9.884576178154161e-05, "loss": 2.7976, "step": 1472 }, { "epoch": 0.12, "learning_rate": 9.884497765231711e-05, "loss": 3.0944, "step": 1473 }, { "epoch": 0.12, "learning_rate": 9.884419352309262e-05, "loss": 3.004, "step": 1474 }, { "epoch": 0.12, "learning_rate": 9.884340939386812e-05, "loss": 3.2898, "step": 1475 }, { "epoch": 0.12, "learning_rate": 9.884262526464362e-05, "loss": 3.1823, "step": 1476 }, { "epoch": 0.12, "learning_rate": 9.884184113541913e-05, "loss": 3.3849, "step": 1477 }, { "epoch": 0.12, "learning_rate": 9.884105700619463e-05, "loss": 3.1991, "step": 1478 }, { "epoch": 0.12, "learning_rate": 9.884027287697013e-05, "loss": 3.2408, "step": 1479 }, { "epoch": 0.12, "learning_rate": 9.883948874774564e-05, "loss": 3.2297, "step": 1480 }, { "epoch": 0.12, "learning_rate": 9.883870461852114e-05, "loss": 3.0318, "step": 1481 }, { "epoch": 0.12, "learning_rate": 9.883792048929664e-05, "loss": 2.9963, "step": 1482 }, { "epoch": 0.12, "learning_rate": 9.883713636007215e-05, "loss": 3.0563, "step": 1483 }, { "epoch": 0.12, "learning_rate": 9.883635223084765e-05, "loss": 2.9725, "step": 1484 }, { "epoch": 0.12, "learning_rate": 9.883556810162316e-05, "loss": 3.12, "step": 1485 }, { "epoch": 0.12, "learning_rate": 9.883478397239866e-05, "loss": 3.0518, "step": 1486 }, { "epoch": 0.12, "learning_rate": 9.883399984317416e-05, "loss": 3.1775, "step": 1487 }, { "epoch": 0.12, "learning_rate": 9.883321571394967e-05, "loss": 3.1603, "step": 1488 }, { "epoch": 0.12, "learning_rate": 9.883243158472517e-05, "loss": 3.1325, "step": 1489 }, { "epoch": 0.12, "learning_rate": 9.883164745550067e-05, "loss": 3.4543, "step": 1490 }, { "epoch": 0.12, "learning_rate": 9.883086332627618e-05, "loss": 3.4831, "step": 1491 }, { "epoch": 0.12, "learning_rate": 9.883007919705168e-05, "loss": 3.0658, "step": 1492 }, { "epoch": 0.12, "learning_rate": 9.882929506782719e-05, "loss": 3.1945, "step": 1493 }, { "epoch": 0.12, "learning_rate": 9.882851093860269e-05, "loss": 3.1204, "step": 1494 }, { "epoch": 0.12, "learning_rate": 9.882772680937818e-05, "loss": 3.1509, "step": 1495 }, { "epoch": 0.12, "learning_rate": 9.882694268015368e-05, "loss": 3.0549, "step": 1496 }, { "epoch": 0.12, "learning_rate": 9.882615855092919e-05, "loss": 3.1781, "step": 1497 }, { "epoch": 0.12, "learning_rate": 9.88253744217047e-05, "loss": 2.946, "step": 1498 }, { "epoch": 0.12, "learning_rate": 9.882459029248021e-05, "loss": 3.1465, "step": 1499 }, { "epoch": 0.12, "learning_rate": 9.882380616325571e-05, "loss": 3.4231, "step": 1500 }, { "epoch": 0.12, "learning_rate": 9.882302203403121e-05, "loss": 3.2081, "step": 1501 }, { "epoch": 0.12, "learning_rate": 9.882223790480672e-05, "loss": 3.2578, "step": 1502 }, { "epoch": 0.12, "learning_rate": 9.882145377558222e-05, "loss": 3.3521, "step": 1503 }, { "epoch": 0.12, "learning_rate": 9.882066964635773e-05, "loss": 3.2004, "step": 1504 }, { "epoch": 0.12, "learning_rate": 9.881988551713323e-05, "loss": 3.1028, "step": 1505 }, { "epoch": 0.12, "learning_rate": 9.881910138790873e-05, "loss": 3.1501, "step": 1506 }, { "epoch": 0.12, "learning_rate": 9.881831725868424e-05, "loss": 3.0736, "step": 1507 }, { "epoch": 0.12, "learning_rate": 9.881753312945974e-05, "loss": 3.0903, "step": 1508 }, { "epoch": 0.12, "learning_rate": 9.881674900023524e-05, "loss": 3.2174, "step": 1509 }, { "epoch": 0.12, "learning_rate": 9.881596487101075e-05, "loss": 3.2778, "step": 1510 }, { "epoch": 0.12, "learning_rate": 9.881518074178625e-05, "loss": 3.3049, "step": 1511 }, { "epoch": 0.12, "learning_rate": 9.881439661256176e-05, "loss": 3.1288, "step": 1512 }, { "epoch": 0.12, "learning_rate": 9.881361248333726e-05, "loss": 2.8825, "step": 1513 }, { "epoch": 0.12, "learning_rate": 9.881282835411276e-05, "loss": 3.2353, "step": 1514 }, { "epoch": 0.12, "learning_rate": 9.881204422488827e-05, "loss": 3.561, "step": 1515 }, { "epoch": 0.12, "learning_rate": 9.881126009566377e-05, "loss": 3.0242, "step": 1516 }, { "epoch": 0.12, "learning_rate": 9.881047596643927e-05, "loss": 3.0771, "step": 1517 }, { "epoch": 0.12, "learning_rate": 9.880969183721478e-05, "loss": 3.0246, "step": 1518 }, { "epoch": 0.12, "learning_rate": 9.880890770799028e-05, "loss": 3.1155, "step": 1519 }, { "epoch": 0.12, "learning_rate": 9.880812357876578e-05, "loss": 2.9836, "step": 1520 }, { "epoch": 0.12, "learning_rate": 9.880733944954129e-05, "loss": 3.165, "step": 1521 }, { "epoch": 0.12, "learning_rate": 9.880655532031679e-05, "loss": 2.9024, "step": 1522 }, { "epoch": 0.12, "learning_rate": 9.88057711910923e-05, "loss": 3.1135, "step": 1523 }, { "epoch": 0.12, "learning_rate": 9.88049870618678e-05, "loss": 3.0308, "step": 1524 }, { "epoch": 0.12, "learning_rate": 9.88042029326433e-05, "loss": 3.2208, "step": 1525 }, { "epoch": 0.12, "learning_rate": 9.88034188034188e-05, "loss": 3.2278, "step": 1526 }, { "epoch": 0.12, "learning_rate": 9.880263467419431e-05, "loss": 2.8736, "step": 1527 }, { "epoch": 0.12, "learning_rate": 9.880185054496981e-05, "loss": 3.0572, "step": 1528 }, { "epoch": 0.12, "learning_rate": 9.880106641574533e-05, "loss": 2.989, "step": 1529 }, { "epoch": 0.12, "learning_rate": 9.880028228652083e-05, "loss": 3.0879, "step": 1530 }, { "epoch": 0.12, "learning_rate": 9.879949815729633e-05, "loss": 3.2913, "step": 1531 }, { "epoch": 0.12, "learning_rate": 9.879871402807183e-05, "loss": 3.256, "step": 1532 }, { "epoch": 0.12, "learning_rate": 9.879792989884733e-05, "loss": 3.027, "step": 1533 }, { "epoch": 0.12, "learning_rate": 9.879714576962284e-05, "loss": 2.9233, "step": 1534 }, { "epoch": 0.12, "learning_rate": 9.879636164039834e-05, "loss": 3.1868, "step": 1535 }, { "epoch": 0.12, "learning_rate": 9.879557751117384e-05, "loss": 3.102, "step": 1536 }, { "epoch": 0.12, "learning_rate": 9.879479338194935e-05, "loss": 3.1081, "step": 1537 }, { "epoch": 0.12, "learning_rate": 9.879400925272485e-05, "loss": 3.18, "step": 1538 }, { "epoch": 0.12, "learning_rate": 9.879322512350035e-05, "loss": 3.0042, "step": 1539 }, { "epoch": 0.12, "learning_rate": 9.879244099427586e-05, "loss": 2.9332, "step": 1540 }, { "epoch": 0.12, "learning_rate": 9.879165686505136e-05, "loss": 3.3857, "step": 1541 }, { "epoch": 0.12, "learning_rate": 9.879087273582687e-05, "loss": 3.089, "step": 1542 }, { "epoch": 0.12, "learning_rate": 9.879008860660237e-05, "loss": 2.9869, "step": 1543 }, { "epoch": 0.12, "learning_rate": 9.878930447737787e-05, "loss": 2.8975, "step": 1544 }, { "epoch": 0.12, "learning_rate": 9.878852034815338e-05, "loss": 3.0524, "step": 1545 }, { "epoch": 0.12, "learning_rate": 9.878773621892888e-05, "loss": 3.4173, "step": 1546 }, { "epoch": 0.12, "learning_rate": 9.878695208970438e-05, "loss": 3.0511, "step": 1547 }, { "epoch": 0.12, "learning_rate": 9.878616796047989e-05, "loss": 2.9519, "step": 1548 }, { "epoch": 0.12, "learning_rate": 9.878538383125539e-05, "loss": 2.9593, "step": 1549 }, { "epoch": 0.12, "learning_rate": 9.87845997020309e-05, "loss": 3.166, "step": 1550 }, { "epoch": 0.12, "learning_rate": 9.87838155728064e-05, "loss": 3.4108, "step": 1551 }, { "epoch": 0.12, "learning_rate": 9.87830314435819e-05, "loss": 3.0507, "step": 1552 }, { "epoch": 0.12, "learning_rate": 9.87822473143574e-05, "loss": 2.8844, "step": 1553 }, { "epoch": 0.12, "learning_rate": 9.878146318513291e-05, "loss": 3.0364, "step": 1554 }, { "epoch": 0.12, "learning_rate": 9.878067905590843e-05, "loss": 3.1249, "step": 1555 }, { "epoch": 0.12, "learning_rate": 9.877989492668393e-05, "loss": 2.9421, "step": 1556 }, { "epoch": 0.12, "learning_rate": 9.877911079745943e-05, "loss": 3.1567, "step": 1557 }, { "epoch": 0.12, "learning_rate": 9.877832666823494e-05, "loss": 3.2403, "step": 1558 }, { "epoch": 0.12, "learning_rate": 9.877754253901044e-05, "loss": 3.0531, "step": 1559 }, { "epoch": 0.12, "learning_rate": 9.877675840978595e-05, "loss": 3.0929, "step": 1560 }, { "epoch": 0.12, "learning_rate": 9.877597428056145e-05, "loss": 3.0215, "step": 1561 }, { "epoch": 0.12, "learning_rate": 9.877519015133695e-05, "loss": 3.6596, "step": 1562 }, { "epoch": 0.12, "learning_rate": 9.877440602211246e-05, "loss": 3.323, "step": 1563 }, { "epoch": 0.12, "learning_rate": 9.877362189288796e-05, "loss": 3.0994, "step": 1564 }, { "epoch": 0.12, "learning_rate": 9.877283776366346e-05, "loss": 3.143, "step": 1565 }, { "epoch": 0.12, "learning_rate": 9.877205363443897e-05, "loss": 3.2979, "step": 1566 }, { "epoch": 0.12, "learning_rate": 9.877126950521446e-05, "loss": 3.2728, "step": 1567 }, { "epoch": 0.12, "learning_rate": 9.877048537598996e-05, "loss": 2.9154, "step": 1568 }, { "epoch": 0.12, "learning_rate": 9.876970124676546e-05, "loss": 2.9975, "step": 1569 }, { "epoch": 0.12, "learning_rate": 9.876891711754097e-05, "loss": 2.8476, "step": 1570 }, { "epoch": 0.12, "learning_rate": 9.876813298831647e-05, "loss": 3.2544, "step": 1571 }, { "epoch": 0.12, "learning_rate": 9.876734885909198e-05, "loss": 3.0921, "step": 1572 }, { "epoch": 0.12, "learning_rate": 9.876656472986748e-05, "loss": 3.0583, "step": 1573 }, { "epoch": 0.12, "learning_rate": 9.876578060064298e-05, "loss": 3.2358, "step": 1574 }, { "epoch": 0.12, "learning_rate": 9.876499647141849e-05, "loss": 3.0365, "step": 1575 }, { "epoch": 0.12, "learning_rate": 9.876421234219399e-05, "loss": 3.0353, "step": 1576 }, { "epoch": 0.12, "learning_rate": 9.87634282129695e-05, "loss": 3.191, "step": 1577 }, { "epoch": 0.12, "learning_rate": 9.8762644083745e-05, "loss": 3.2316, "step": 1578 }, { "epoch": 0.12, "learning_rate": 9.87618599545205e-05, "loss": 3.0811, "step": 1579 }, { "epoch": 0.12, "learning_rate": 9.876107582529602e-05, "loss": 3.2373, "step": 1580 }, { "epoch": 0.12, "learning_rate": 9.876029169607152e-05, "loss": 3.1227, "step": 1581 }, { "epoch": 0.12, "learning_rate": 9.875950756684703e-05, "loss": 3.3183, "step": 1582 }, { "epoch": 0.12, "learning_rate": 9.875872343762253e-05, "loss": 3.3168, "step": 1583 }, { "epoch": 0.12, "learning_rate": 9.875793930839803e-05, "loss": 3.38, "step": 1584 }, { "epoch": 0.12, "learning_rate": 9.875715517917354e-05, "loss": 3.1215, "step": 1585 }, { "epoch": 0.12, "learning_rate": 9.875637104994904e-05, "loss": 3.1232, "step": 1586 }, { "epoch": 0.12, "learning_rate": 9.875558692072454e-05, "loss": 3.0302, "step": 1587 }, { "epoch": 0.12, "learning_rate": 9.875480279150005e-05, "loss": 2.8684, "step": 1588 }, { "epoch": 0.12, "learning_rate": 9.875401866227555e-05, "loss": 3.1792, "step": 1589 }, { "epoch": 0.12, "learning_rate": 9.875323453305106e-05, "loss": 2.9412, "step": 1590 }, { "epoch": 0.12, "learning_rate": 9.875245040382656e-05, "loss": 3.3178, "step": 1591 }, { "epoch": 0.12, "learning_rate": 9.875166627460206e-05, "loss": 3.0521, "step": 1592 }, { "epoch": 0.12, "learning_rate": 9.875088214537757e-05, "loss": 3.2311, "step": 1593 }, { "epoch": 0.12, "learning_rate": 9.875009801615307e-05, "loss": 3.2305, "step": 1594 }, { "epoch": 0.13, "learning_rate": 9.874931388692857e-05, "loss": 3.2568, "step": 1595 }, { "epoch": 0.13, "learning_rate": 9.874852975770408e-05, "loss": 2.9288, "step": 1596 }, { "epoch": 0.13, "learning_rate": 9.874774562847958e-05, "loss": 3.1433, "step": 1597 }, { "epoch": 0.13, "learning_rate": 9.874696149925509e-05, "loss": 3.0877, "step": 1598 }, { "epoch": 0.13, "learning_rate": 9.874617737003059e-05, "loss": 3.1066, "step": 1599 }, { "epoch": 0.13, "learning_rate": 9.874539324080609e-05, "loss": 2.9007, "step": 1600 }, { "epoch": 0.13, "learning_rate": 9.87446091115816e-05, "loss": 3.2084, "step": 1601 }, { "epoch": 0.13, "learning_rate": 9.87438249823571e-05, "loss": 3.1208, "step": 1602 }, { "epoch": 0.13, "learning_rate": 9.874304085313259e-05, "loss": 3.4457, "step": 1603 }, { "epoch": 0.13, "learning_rate": 9.87422567239081e-05, "loss": 3.3043, "step": 1604 }, { "epoch": 0.13, "learning_rate": 9.87414725946836e-05, "loss": 3.3506, "step": 1605 }, { "epoch": 0.13, "learning_rate": 9.874068846545911e-05, "loss": 3.2062, "step": 1606 }, { "epoch": 0.13, "learning_rate": 9.873990433623462e-05, "loss": 3.1382, "step": 1607 }, { "epoch": 0.13, "learning_rate": 9.873912020701012e-05, "loss": 3.1339, "step": 1608 }, { "epoch": 0.13, "learning_rate": 9.873833607778563e-05, "loss": 3.4782, "step": 1609 }, { "epoch": 0.13, "learning_rate": 9.873755194856113e-05, "loss": 3.0585, "step": 1610 }, { "epoch": 0.13, "learning_rate": 9.873676781933663e-05, "loss": 3.0147, "step": 1611 }, { "epoch": 0.13, "learning_rate": 9.873598369011214e-05, "loss": 2.9302, "step": 1612 }, { "epoch": 0.13, "learning_rate": 9.873519956088764e-05, "loss": 2.8518, "step": 1613 }, { "epoch": 0.13, "learning_rate": 9.873441543166314e-05, "loss": 2.9819, "step": 1614 }, { "epoch": 0.13, "learning_rate": 9.873363130243865e-05, "loss": 2.9833, "step": 1615 }, { "epoch": 0.13, "learning_rate": 9.873284717321415e-05, "loss": 3.1634, "step": 1616 }, { "epoch": 0.13, "learning_rate": 9.873206304398965e-05, "loss": 2.9929, "step": 1617 }, { "epoch": 0.13, "learning_rate": 9.873127891476516e-05, "loss": 3.0746, "step": 1618 }, { "epoch": 0.13, "learning_rate": 9.873049478554066e-05, "loss": 3.1802, "step": 1619 }, { "epoch": 0.13, "learning_rate": 9.872971065631617e-05, "loss": 3.1186, "step": 1620 }, { "epoch": 0.13, "learning_rate": 9.872892652709167e-05, "loss": 3.0627, "step": 1621 }, { "epoch": 0.13, "learning_rate": 9.872814239786717e-05, "loss": 3.0987, "step": 1622 }, { "epoch": 0.13, "learning_rate": 9.872735826864268e-05, "loss": 2.953, "step": 1623 }, { "epoch": 0.13, "learning_rate": 9.872657413941818e-05, "loss": 2.9641, "step": 1624 }, { "epoch": 0.13, "learning_rate": 9.872579001019368e-05, "loss": 3.075, "step": 1625 }, { "epoch": 0.13, "learning_rate": 9.872500588096919e-05, "loss": 2.8192, "step": 1626 }, { "epoch": 0.13, "learning_rate": 9.872422175174469e-05, "loss": 3.0644, "step": 1627 }, { "epoch": 0.13, "learning_rate": 9.87234376225202e-05, "loss": 3.3025, "step": 1628 }, { "epoch": 0.13, "learning_rate": 9.87226534932957e-05, "loss": 3.1314, "step": 1629 }, { "epoch": 0.13, "learning_rate": 9.87218693640712e-05, "loss": 3.1013, "step": 1630 }, { "epoch": 0.13, "learning_rate": 9.87210852348467e-05, "loss": 3.1027, "step": 1631 }, { "epoch": 0.13, "learning_rate": 9.872030110562221e-05, "loss": 3.033, "step": 1632 }, { "epoch": 0.13, "learning_rate": 9.871951697639771e-05, "loss": 2.955, "step": 1633 }, { "epoch": 0.13, "learning_rate": 9.871873284717322e-05, "loss": 3.0398, "step": 1634 }, { "epoch": 0.13, "learning_rate": 9.871794871794872e-05, "loss": 3.3188, "step": 1635 }, { "epoch": 0.13, "learning_rate": 9.871716458872422e-05, "loss": 3.1236, "step": 1636 }, { "epoch": 0.13, "learning_rate": 9.871638045949974e-05, "loss": 2.9513, "step": 1637 }, { "epoch": 0.13, "learning_rate": 9.871559633027525e-05, "loss": 3.183, "step": 1638 }, { "epoch": 0.13, "learning_rate": 9.871481220105074e-05, "loss": 3.2814, "step": 1639 }, { "epoch": 0.13, "learning_rate": 9.871402807182624e-05, "loss": 3.1469, "step": 1640 }, { "epoch": 0.13, "learning_rate": 9.871324394260174e-05, "loss": 2.9464, "step": 1641 }, { "epoch": 0.13, "learning_rate": 9.871245981337725e-05, "loss": 3.1823, "step": 1642 }, { "epoch": 0.13, "learning_rate": 9.871167568415275e-05, "loss": 3.2961, "step": 1643 }, { "epoch": 0.13, "learning_rate": 9.871089155492825e-05, "loss": 3.0695, "step": 1644 }, { "epoch": 0.13, "learning_rate": 9.871010742570376e-05, "loss": 3.0607, "step": 1645 }, { "epoch": 0.13, "learning_rate": 9.870932329647926e-05, "loss": 3.191, "step": 1646 }, { "epoch": 0.13, "learning_rate": 9.870853916725477e-05, "loss": 3.0085, "step": 1647 }, { "epoch": 0.13, "learning_rate": 9.870775503803027e-05, "loss": 3.4441, "step": 1648 }, { "epoch": 0.13, "learning_rate": 9.870697090880577e-05, "loss": 2.9375, "step": 1649 }, { "epoch": 0.13, "learning_rate": 9.870618677958128e-05, "loss": 2.9699, "step": 1650 }, { "epoch": 0.13, "learning_rate": 9.870540265035678e-05, "loss": 3.3121, "step": 1651 }, { "epoch": 0.13, "learning_rate": 9.870461852113228e-05, "loss": 3.2112, "step": 1652 }, { "epoch": 0.13, "learning_rate": 9.870383439190779e-05, "loss": 3.0275, "step": 1653 }, { "epoch": 0.13, "learning_rate": 9.870305026268329e-05, "loss": 3.1143, "step": 1654 }, { "epoch": 0.13, "learning_rate": 9.87022661334588e-05, "loss": 3.2317, "step": 1655 }, { "epoch": 0.13, "learning_rate": 9.87014820042343e-05, "loss": 3.1274, "step": 1656 }, { "epoch": 0.13, "learning_rate": 9.87006978750098e-05, "loss": 3.224, "step": 1657 }, { "epoch": 0.13, "learning_rate": 9.86999137457853e-05, "loss": 2.9665, "step": 1658 }, { "epoch": 0.13, "learning_rate": 9.869912961656081e-05, "loss": 3.1158, "step": 1659 }, { "epoch": 0.13, "learning_rate": 9.869834548733631e-05, "loss": 3.078, "step": 1660 }, { "epoch": 0.13, "learning_rate": 9.869756135811182e-05, "loss": 3.014, "step": 1661 }, { "epoch": 0.13, "learning_rate": 9.869677722888732e-05, "loss": 3.1889, "step": 1662 }, { "epoch": 0.13, "learning_rate": 9.869599309966284e-05, "loss": 3.0888, "step": 1663 }, { "epoch": 0.13, "learning_rate": 9.869520897043834e-05, "loss": 3.0197, "step": 1664 }, { "epoch": 0.13, "learning_rate": 9.869442484121384e-05, "loss": 2.981, "step": 1665 }, { "epoch": 0.13, "learning_rate": 9.869364071198935e-05, "loss": 3.0441, "step": 1666 }, { "epoch": 0.13, "learning_rate": 9.869285658276485e-05, "loss": 3.274, "step": 1667 }, { "epoch": 0.13, "learning_rate": 9.869207245354036e-05, "loss": 2.9847, "step": 1668 }, { "epoch": 0.13, "learning_rate": 9.869128832431586e-05, "loss": 3.199, "step": 1669 }, { "epoch": 0.13, "learning_rate": 9.869050419509136e-05, "loss": 3.1346, "step": 1670 }, { "epoch": 0.13, "learning_rate": 9.868972006586687e-05, "loss": 3.1825, "step": 1671 }, { "epoch": 0.13, "learning_rate": 9.868893593664237e-05, "loss": 3.1018, "step": 1672 }, { "epoch": 0.13, "learning_rate": 9.868815180741787e-05, "loss": 3.2892, "step": 1673 }, { "epoch": 0.13, "learning_rate": 9.868736767819338e-05, "loss": 3.1792, "step": 1674 }, { "epoch": 0.13, "learning_rate": 9.868658354896887e-05, "loss": 3.6144, "step": 1675 }, { "epoch": 0.13, "learning_rate": 9.868579941974437e-05, "loss": 3.0998, "step": 1676 }, { "epoch": 0.13, "learning_rate": 9.868501529051988e-05, "loss": 3.154, "step": 1677 }, { "epoch": 0.13, "learning_rate": 9.868423116129538e-05, "loss": 3.1148, "step": 1678 }, { "epoch": 0.13, "learning_rate": 9.868344703207088e-05, "loss": 2.7949, "step": 1679 }, { "epoch": 0.13, "learning_rate": 9.868266290284639e-05, "loss": 3.0633, "step": 1680 }, { "epoch": 0.13, "learning_rate": 9.868187877362189e-05, "loss": 2.9215, "step": 1681 }, { "epoch": 0.13, "learning_rate": 9.86810946443974e-05, "loss": 3.0641, "step": 1682 }, { "epoch": 0.13, "learning_rate": 9.86803105151729e-05, "loss": 2.9662, "step": 1683 }, { "epoch": 0.13, "learning_rate": 9.86795263859484e-05, "loss": 2.8715, "step": 1684 }, { "epoch": 0.13, "learning_rate": 9.86787422567239e-05, "loss": 3.3588, "step": 1685 }, { "epoch": 0.13, "learning_rate": 9.867795812749941e-05, "loss": 3.1151, "step": 1686 }, { "epoch": 0.13, "learning_rate": 9.867717399827491e-05, "loss": 3.0969, "step": 1687 }, { "epoch": 0.13, "learning_rate": 9.867638986905043e-05, "loss": 3.2061, "step": 1688 }, { "epoch": 0.13, "learning_rate": 9.867560573982593e-05, "loss": 3.0098, "step": 1689 }, { "epoch": 0.13, "learning_rate": 9.867482161060144e-05, "loss": 2.8891, "step": 1690 }, { "epoch": 0.13, "learning_rate": 9.867403748137694e-05, "loss": 3.1221, "step": 1691 }, { "epoch": 0.13, "learning_rate": 9.867325335215244e-05, "loss": 3.0914, "step": 1692 }, { "epoch": 0.13, "learning_rate": 9.867246922292795e-05, "loss": 3.1775, "step": 1693 }, { "epoch": 0.13, "learning_rate": 9.867168509370345e-05, "loss": 3.0623, "step": 1694 }, { "epoch": 0.13, "learning_rate": 9.867090096447896e-05, "loss": 3.0686, "step": 1695 }, { "epoch": 0.13, "learning_rate": 9.867011683525446e-05, "loss": 3.0962, "step": 1696 }, { "epoch": 0.13, "learning_rate": 9.866933270602996e-05, "loss": 3.0195, "step": 1697 }, { "epoch": 0.13, "learning_rate": 9.866854857680547e-05, "loss": 3.1435, "step": 1698 }, { "epoch": 0.13, "learning_rate": 9.866776444758097e-05, "loss": 3.2533, "step": 1699 }, { "epoch": 0.13, "learning_rate": 9.866698031835647e-05, "loss": 3.1533, "step": 1700 }, { "epoch": 0.13, "learning_rate": 9.866619618913198e-05, "loss": 2.9515, "step": 1701 }, { "epoch": 0.13, "learning_rate": 9.866541205990748e-05, "loss": 3.509, "step": 1702 }, { "epoch": 0.13, "learning_rate": 9.866462793068298e-05, "loss": 2.9913, "step": 1703 }, { "epoch": 0.13, "learning_rate": 9.866384380145849e-05, "loss": 2.9144, "step": 1704 }, { "epoch": 0.13, "learning_rate": 9.866305967223399e-05, "loss": 3.1397, "step": 1705 }, { "epoch": 0.13, "learning_rate": 9.86622755430095e-05, "loss": 3.2701, "step": 1706 }, { "epoch": 0.13, "learning_rate": 9.8661491413785e-05, "loss": 3.024, "step": 1707 }, { "epoch": 0.13, "learning_rate": 9.86607072845605e-05, "loss": 3.1526, "step": 1708 }, { "epoch": 0.13, "learning_rate": 9.865992315533601e-05, "loss": 3.0827, "step": 1709 }, { "epoch": 0.13, "learning_rate": 9.865913902611151e-05, "loss": 3.1225, "step": 1710 }, { "epoch": 0.13, "learning_rate": 9.8658354896887e-05, "loss": 2.9939, "step": 1711 }, { "epoch": 0.13, "learning_rate": 9.86575707676625e-05, "loss": 3.2159, "step": 1712 }, { "epoch": 0.13, "learning_rate": 9.865678663843801e-05, "loss": 3.1727, "step": 1713 }, { "epoch": 0.13, "learning_rate": 9.865600250921353e-05, "loss": 3.1115, "step": 1714 }, { "epoch": 0.13, "learning_rate": 9.865521837998903e-05, "loss": 3.1166, "step": 1715 }, { "epoch": 0.13, "learning_rate": 9.865443425076453e-05, "loss": 3.15, "step": 1716 }, { "epoch": 0.13, "learning_rate": 9.865365012154004e-05, "loss": 3.1844, "step": 1717 }, { "epoch": 0.13, "learning_rate": 9.865286599231554e-05, "loss": 3.1332, "step": 1718 }, { "epoch": 0.13, "learning_rate": 9.865208186309104e-05, "loss": 3.3661, "step": 1719 }, { "epoch": 0.13, "learning_rate": 9.865129773386655e-05, "loss": 3.1772, "step": 1720 }, { "epoch": 0.13, "learning_rate": 9.865051360464205e-05, "loss": 2.9612, "step": 1721 }, { "epoch": 0.14, "learning_rate": 9.864972947541755e-05, "loss": 3.003, "step": 1722 }, { "epoch": 0.14, "learning_rate": 9.864894534619306e-05, "loss": 3.0157, "step": 1723 }, { "epoch": 0.14, "learning_rate": 9.864816121696856e-05, "loss": 3.2348, "step": 1724 }, { "epoch": 0.14, "learning_rate": 9.864737708774407e-05, "loss": 3.267, "step": 1725 }, { "epoch": 0.14, "learning_rate": 9.864659295851957e-05, "loss": 3.0689, "step": 1726 }, { "epoch": 0.14, "learning_rate": 9.864580882929507e-05, "loss": 2.887, "step": 1727 }, { "epoch": 0.14, "learning_rate": 9.864502470007058e-05, "loss": 3.0673, "step": 1728 }, { "epoch": 0.14, "learning_rate": 9.864424057084608e-05, "loss": 2.9243, "step": 1729 }, { "epoch": 0.14, "learning_rate": 9.864345644162158e-05, "loss": 3.0423, "step": 1730 }, { "epoch": 0.14, "learning_rate": 9.864267231239709e-05, "loss": 3.1184, "step": 1731 }, { "epoch": 0.14, "learning_rate": 9.864188818317259e-05, "loss": 3.16, "step": 1732 }, { "epoch": 0.14, "learning_rate": 9.86411040539481e-05, "loss": 3.0382, "step": 1733 }, { "epoch": 0.14, "learning_rate": 9.86403199247236e-05, "loss": 2.9683, "step": 1734 }, { "epoch": 0.14, "learning_rate": 9.86395357954991e-05, "loss": 3.0435, "step": 1735 }, { "epoch": 0.14, "learning_rate": 9.86387516662746e-05, "loss": 3.0288, "step": 1736 }, { "epoch": 0.14, "learning_rate": 9.863796753705011e-05, "loss": 2.9108, "step": 1737 }, { "epoch": 0.14, "learning_rate": 9.863718340782561e-05, "loss": 2.8989, "step": 1738 }, { "epoch": 0.14, "learning_rate": 9.863639927860112e-05, "loss": 3.3966, "step": 1739 }, { "epoch": 0.14, "learning_rate": 9.863561514937662e-05, "loss": 2.948, "step": 1740 }, { "epoch": 0.14, "learning_rate": 9.863483102015212e-05, "loss": 3.1626, "step": 1741 }, { "epoch": 0.14, "learning_rate": 9.863404689092763e-05, "loss": 3.1109, "step": 1742 }, { "epoch": 0.14, "learning_rate": 9.863326276170313e-05, "loss": 3.1583, "step": 1743 }, { "epoch": 0.14, "learning_rate": 9.863247863247864e-05, "loss": 3.1306, "step": 1744 }, { "epoch": 0.14, "learning_rate": 9.863169450325415e-05, "loss": 3.219, "step": 1745 }, { "epoch": 0.14, "learning_rate": 9.863091037402966e-05, "loss": 2.9298, "step": 1746 }, { "epoch": 0.14, "learning_rate": 9.863012624480515e-05, "loss": 3.0511, "step": 1747 }, { "epoch": 0.14, "learning_rate": 9.862934211558065e-05, "loss": 3.0498, "step": 1748 }, { "epoch": 0.14, "learning_rate": 9.862855798635615e-05, "loss": 3.1406, "step": 1749 }, { "epoch": 0.14, "learning_rate": 9.862777385713166e-05, "loss": 3.1204, "step": 1750 }, { "epoch": 0.14, "learning_rate": 9.862698972790716e-05, "loss": 2.854, "step": 1751 }, { "epoch": 0.14, "learning_rate": 9.862620559868266e-05, "loss": 3.0535, "step": 1752 }, { "epoch": 0.14, "learning_rate": 9.862542146945817e-05, "loss": 2.999, "step": 1753 }, { "epoch": 0.14, "learning_rate": 9.862463734023367e-05, "loss": 3.0427, "step": 1754 }, { "epoch": 0.14, "learning_rate": 9.862385321100918e-05, "loss": 3.2687, "step": 1755 }, { "epoch": 0.14, "learning_rate": 9.862306908178468e-05, "loss": 2.7969, "step": 1756 }, { "epoch": 0.14, "learning_rate": 9.862228495256018e-05, "loss": 3.2113, "step": 1757 }, { "epoch": 0.14, "learning_rate": 9.862150082333569e-05, "loss": 3.2107, "step": 1758 }, { "epoch": 0.14, "learning_rate": 9.862071669411119e-05, "loss": 2.8957, "step": 1759 }, { "epoch": 0.14, "learning_rate": 9.86199325648867e-05, "loss": 3.1804, "step": 1760 }, { "epoch": 0.14, "learning_rate": 9.86191484356622e-05, "loss": 3.3019, "step": 1761 }, { "epoch": 0.14, "learning_rate": 9.86183643064377e-05, "loss": 2.9691, "step": 1762 }, { "epoch": 0.14, "learning_rate": 9.86175801772132e-05, "loss": 3.1995, "step": 1763 }, { "epoch": 0.14, "learning_rate": 9.861679604798871e-05, "loss": 3.1415, "step": 1764 }, { "epoch": 0.14, "learning_rate": 9.861601191876421e-05, "loss": 3.003, "step": 1765 }, { "epoch": 0.14, "learning_rate": 9.861522778953972e-05, "loss": 2.991, "step": 1766 }, { "epoch": 0.14, "learning_rate": 9.861444366031522e-05, "loss": 3.1624, "step": 1767 }, { "epoch": 0.14, "learning_rate": 9.861365953109072e-05, "loss": 2.984, "step": 1768 }, { "epoch": 0.14, "learning_rate": 9.861287540186623e-05, "loss": 2.8279, "step": 1769 }, { "epoch": 0.14, "learning_rate": 9.861209127264173e-05, "loss": 2.9157, "step": 1770 }, { "epoch": 0.14, "learning_rate": 9.861130714341725e-05, "loss": 3.136, "step": 1771 }, { "epoch": 0.14, "learning_rate": 9.861052301419275e-05, "loss": 3.0817, "step": 1772 }, { "epoch": 0.14, "learning_rate": 9.860973888496826e-05, "loss": 3.0845, "step": 1773 }, { "epoch": 0.14, "learning_rate": 9.860895475574376e-05, "loss": 3.0158, "step": 1774 }, { "epoch": 0.14, "learning_rate": 9.860817062651926e-05, "loss": 2.9337, "step": 1775 }, { "epoch": 0.14, "learning_rate": 9.860738649729477e-05, "loss": 3.0417, "step": 1776 }, { "epoch": 0.14, "learning_rate": 9.860660236807027e-05, "loss": 2.9325, "step": 1777 }, { "epoch": 0.14, "learning_rate": 9.860581823884577e-05, "loss": 3.1975, "step": 1778 }, { "epoch": 0.14, "learning_rate": 9.860503410962128e-05, "loss": 3.0423, "step": 1779 }, { "epoch": 0.14, "learning_rate": 9.860424998039678e-05, "loss": 3.0633, "step": 1780 }, { "epoch": 0.14, "learning_rate": 9.860346585117229e-05, "loss": 3.2249, "step": 1781 }, { "epoch": 0.14, "learning_rate": 9.860268172194779e-05, "loss": 2.9464, "step": 1782 }, { "epoch": 0.14, "learning_rate": 9.860189759272328e-05, "loss": 3.1771, "step": 1783 }, { "epoch": 0.14, "learning_rate": 9.860111346349878e-05, "loss": 3.5157, "step": 1784 }, { "epoch": 0.14, "learning_rate": 9.860032933427429e-05, "loss": 3.1123, "step": 1785 }, { "epoch": 0.14, "learning_rate": 9.859954520504979e-05, "loss": 3.3181, "step": 1786 }, { "epoch": 0.14, "learning_rate": 9.85987610758253e-05, "loss": 3.01, "step": 1787 }, { "epoch": 0.14, "learning_rate": 9.85979769466008e-05, "loss": 3.0572, "step": 1788 }, { "epoch": 0.14, "learning_rate": 9.85971928173763e-05, "loss": 3.3106, "step": 1789 }, { "epoch": 0.14, "learning_rate": 9.85964086881518e-05, "loss": 3.0484, "step": 1790 }, { "epoch": 0.14, "learning_rate": 9.859562455892731e-05, "loss": 3.0936, "step": 1791 }, { "epoch": 0.14, "learning_rate": 9.859484042970281e-05, "loss": 3.2658, "step": 1792 }, { "epoch": 0.14, "learning_rate": 9.859405630047832e-05, "loss": 3.2446, "step": 1793 }, { "epoch": 0.14, "learning_rate": 9.859327217125382e-05, "loss": 3.1841, "step": 1794 }, { "epoch": 0.14, "learning_rate": 9.859248804202932e-05, "loss": 3.4841, "step": 1795 }, { "epoch": 0.14, "learning_rate": 9.859170391280484e-05, "loss": 3.0766, "step": 1796 }, { "epoch": 0.14, "learning_rate": 9.859091978358034e-05, "loss": 3.0807, "step": 1797 }, { "epoch": 0.14, "learning_rate": 9.859013565435585e-05, "loss": 3.2227, "step": 1798 }, { "epoch": 0.14, "learning_rate": 9.858935152513135e-05, "loss": 3.1127, "step": 1799 }, { "epoch": 0.14, "learning_rate": 9.858856739590686e-05, "loss": 3.1294, "step": 1800 }, { "epoch": 0.14, "learning_rate": 9.858778326668236e-05, "loss": 3.005, "step": 1801 }, { "epoch": 0.14, "learning_rate": 9.858699913745786e-05, "loss": 3.0717, "step": 1802 }, { "epoch": 0.14, "learning_rate": 9.858621500823337e-05, "loss": 3.1771, "step": 1803 }, { "epoch": 0.14, "learning_rate": 9.858543087900887e-05, "loss": 3.2311, "step": 1804 }, { "epoch": 0.14, "learning_rate": 9.858464674978437e-05, "loss": 3.0794, "step": 1805 }, { "epoch": 0.14, "learning_rate": 9.858386262055988e-05, "loss": 3.1487, "step": 1806 }, { "epoch": 0.14, "learning_rate": 9.858307849133538e-05, "loss": 3.2194, "step": 1807 }, { "epoch": 0.14, "learning_rate": 9.858229436211088e-05, "loss": 3.1861, "step": 1808 }, { "epoch": 0.14, "learning_rate": 9.858151023288639e-05, "loss": 3.2736, "step": 1809 }, { "epoch": 0.14, "learning_rate": 9.858072610366189e-05, "loss": 3.378, "step": 1810 }, { "epoch": 0.14, "learning_rate": 9.85799419744374e-05, "loss": 3.1065, "step": 1811 }, { "epoch": 0.14, "learning_rate": 9.85791578452129e-05, "loss": 3.1068, "step": 1812 }, { "epoch": 0.14, "learning_rate": 9.85783737159884e-05, "loss": 3.0687, "step": 1813 }, { "epoch": 0.14, "learning_rate": 9.85775895867639e-05, "loss": 3.1219, "step": 1814 }, { "epoch": 0.14, "learning_rate": 9.857680545753941e-05, "loss": 3.1873, "step": 1815 }, { "epoch": 0.14, "learning_rate": 9.857602132831491e-05, "loss": 3.3185, "step": 1816 }, { "epoch": 0.14, "learning_rate": 9.857523719909042e-05, "loss": 3.1464, "step": 1817 }, { "epoch": 0.14, "learning_rate": 9.857445306986592e-05, "loss": 3.039, "step": 1818 }, { "epoch": 0.14, "learning_rate": 9.857366894064141e-05, "loss": 3.0383, "step": 1819 }, { "epoch": 0.14, "learning_rate": 9.857288481141692e-05, "loss": 3.1491, "step": 1820 }, { "epoch": 0.14, "learning_rate": 9.857210068219242e-05, "loss": 3.3387, "step": 1821 }, { "epoch": 0.14, "learning_rate": 9.857131655296794e-05, "loss": 3.3689, "step": 1822 }, { "epoch": 0.14, "learning_rate": 9.857053242374344e-05, "loss": 2.7612, "step": 1823 }, { "epoch": 0.14, "learning_rate": 9.856974829451894e-05, "loss": 3.1589, "step": 1824 }, { "epoch": 0.14, "learning_rate": 9.856896416529445e-05, "loss": 2.8784, "step": 1825 }, { "epoch": 0.14, "learning_rate": 9.856818003606995e-05, "loss": 3.2088, "step": 1826 }, { "epoch": 0.14, "learning_rate": 9.856739590684545e-05, "loss": 3.2262, "step": 1827 }, { "epoch": 0.14, "learning_rate": 9.856661177762096e-05, "loss": 3.2531, "step": 1828 }, { "epoch": 0.14, "learning_rate": 9.856582764839646e-05, "loss": 2.8091, "step": 1829 }, { "epoch": 0.14, "learning_rate": 9.856504351917197e-05, "loss": 3.1396, "step": 1830 }, { "epoch": 0.14, "learning_rate": 9.856425938994747e-05, "loss": 3.151, "step": 1831 }, { "epoch": 0.14, "learning_rate": 9.856347526072297e-05, "loss": 3.3492, "step": 1832 }, { "epoch": 0.14, "learning_rate": 9.856269113149848e-05, "loss": 2.9948, "step": 1833 }, { "epoch": 0.14, "learning_rate": 9.856190700227398e-05, "loss": 2.9689, "step": 1834 }, { "epoch": 0.14, "learning_rate": 9.856112287304948e-05, "loss": 2.8761, "step": 1835 }, { "epoch": 0.14, "learning_rate": 9.856033874382499e-05, "loss": 2.9617, "step": 1836 }, { "epoch": 0.14, "learning_rate": 9.855955461460049e-05, "loss": 3.2169, "step": 1837 }, { "epoch": 0.14, "learning_rate": 9.8558770485376e-05, "loss": 3.0646, "step": 1838 }, { "epoch": 0.14, "learning_rate": 9.85579863561515e-05, "loss": 3.0134, "step": 1839 }, { "epoch": 0.14, "learning_rate": 9.8557202226927e-05, "loss": 3.1673, "step": 1840 }, { "epoch": 0.14, "learning_rate": 9.85564180977025e-05, "loss": 2.9478, "step": 1841 }, { "epoch": 0.14, "learning_rate": 9.855563396847801e-05, "loss": 3.2463, "step": 1842 }, { "epoch": 0.14, "learning_rate": 9.855484983925351e-05, "loss": 3.3139, "step": 1843 }, { "epoch": 0.14, "learning_rate": 9.855406571002902e-05, "loss": 3.3282, "step": 1844 }, { "epoch": 0.14, "learning_rate": 9.855328158080452e-05, "loss": 3.0208, "step": 1845 }, { "epoch": 0.14, "learning_rate": 9.855249745158002e-05, "loss": 3.1134, "step": 1846 }, { "epoch": 0.14, "learning_rate": 9.855171332235553e-05, "loss": 3.0192, "step": 1847 }, { "epoch": 0.14, "learning_rate": 9.855092919313103e-05, "loss": 2.9708, "step": 1848 }, { "epoch": 0.14, "learning_rate": 9.855014506390654e-05, "loss": 3.2949, "step": 1849 }, { "epoch": 0.15, "learning_rate": 9.854936093468204e-05, "loss": 3.1856, "step": 1850 }, { "epoch": 0.15, "learning_rate": 9.854857680545754e-05, "loss": 3.4477, "step": 1851 }, { "epoch": 0.15, "learning_rate": 9.854779267623305e-05, "loss": 2.8677, "step": 1852 }, { "epoch": 0.15, "learning_rate": 9.854700854700856e-05, "loss": 3.0563, "step": 1853 }, { "epoch": 0.15, "learning_rate": 9.854622441778407e-05, "loss": 3.2458, "step": 1854 }, { "epoch": 0.15, "learning_rate": 9.854544028855956e-05, "loss": 3.1924, "step": 1855 }, { "epoch": 0.15, "learning_rate": 9.854465615933506e-05, "loss": 3.1865, "step": 1856 }, { "epoch": 0.15, "learning_rate": 9.854387203011056e-05, "loss": 3.0992, "step": 1857 }, { "epoch": 0.15, "learning_rate": 9.854308790088607e-05, "loss": 3.0038, "step": 1858 }, { "epoch": 0.15, "learning_rate": 9.854230377166157e-05, "loss": 3.6503, "step": 1859 }, { "epoch": 0.15, "learning_rate": 9.854151964243708e-05, "loss": 3.3183, "step": 1860 }, { "epoch": 0.15, "learning_rate": 9.854073551321258e-05, "loss": 2.9805, "step": 1861 }, { "epoch": 0.15, "learning_rate": 9.853995138398808e-05, "loss": 3.0016, "step": 1862 }, { "epoch": 0.15, "learning_rate": 9.853916725476359e-05, "loss": 3.1822, "step": 1863 }, { "epoch": 0.15, "learning_rate": 9.853838312553909e-05, "loss": 2.907, "step": 1864 }, { "epoch": 0.15, "learning_rate": 9.85375989963146e-05, "loss": 2.9702, "step": 1865 }, { "epoch": 0.15, "learning_rate": 9.85368148670901e-05, "loss": 3.1059, "step": 1866 }, { "epoch": 0.15, "learning_rate": 9.85360307378656e-05, "loss": 3.1771, "step": 1867 }, { "epoch": 0.15, "learning_rate": 9.85352466086411e-05, "loss": 3.1244, "step": 1868 }, { "epoch": 0.15, "learning_rate": 9.853446247941661e-05, "loss": 3.5171, "step": 1869 }, { "epoch": 0.15, "learning_rate": 9.853367835019211e-05, "loss": 3.1766, "step": 1870 }, { "epoch": 0.15, "learning_rate": 9.853289422096762e-05, "loss": 3.3925, "step": 1871 }, { "epoch": 0.15, "learning_rate": 9.853211009174312e-05, "loss": 3.0522, "step": 1872 }, { "epoch": 0.15, "learning_rate": 9.853132596251862e-05, "loss": 3.0404, "step": 1873 }, { "epoch": 0.15, "learning_rate": 9.853054183329413e-05, "loss": 3.099, "step": 1874 }, { "epoch": 0.15, "learning_rate": 9.852975770406963e-05, "loss": 3.0827, "step": 1875 }, { "epoch": 0.15, "learning_rate": 9.852897357484513e-05, "loss": 3.0194, "step": 1876 }, { "epoch": 0.15, "learning_rate": 9.852818944562064e-05, "loss": 2.9007, "step": 1877 }, { "epoch": 0.15, "learning_rate": 9.852740531639614e-05, "loss": 2.8445, "step": 1878 }, { "epoch": 0.15, "learning_rate": 9.852662118717166e-05, "loss": 3.0973, "step": 1879 }, { "epoch": 0.15, "learning_rate": 9.852583705794716e-05, "loss": 3.0704, "step": 1880 }, { "epoch": 0.15, "learning_rate": 9.852505292872267e-05, "loss": 3.0758, "step": 1881 }, { "epoch": 0.15, "learning_rate": 9.852426879949817e-05, "loss": 3.4079, "step": 1882 }, { "epoch": 0.15, "learning_rate": 9.852348467027367e-05, "loss": 3.1029, "step": 1883 }, { "epoch": 0.15, "learning_rate": 9.852270054104918e-05, "loss": 3.3763, "step": 1884 }, { "epoch": 0.15, "learning_rate": 9.852191641182468e-05, "loss": 3.0616, "step": 1885 }, { "epoch": 0.15, "learning_rate": 9.852113228260018e-05, "loss": 3.0155, "step": 1886 }, { "epoch": 0.15, "learning_rate": 9.852034815337569e-05, "loss": 3.303, "step": 1887 }, { "epoch": 0.15, "learning_rate": 9.851956402415119e-05, "loss": 2.9723, "step": 1888 }, { "epoch": 0.15, "learning_rate": 9.85187798949267e-05, "loss": 3.0548, "step": 1889 }, { "epoch": 0.15, "learning_rate": 9.85179957657022e-05, "loss": 3.0925, "step": 1890 }, { "epoch": 0.15, "learning_rate": 9.851721163647769e-05, "loss": 3.5174, "step": 1891 }, { "epoch": 0.15, "learning_rate": 9.85164275072532e-05, "loss": 3.1855, "step": 1892 }, { "epoch": 0.15, "learning_rate": 9.85156433780287e-05, "loss": 3.0914, "step": 1893 }, { "epoch": 0.15, "learning_rate": 9.85148592488042e-05, "loss": 3.1353, "step": 1894 }, { "epoch": 0.15, "learning_rate": 9.85140751195797e-05, "loss": 3.1545, "step": 1895 }, { "epoch": 0.15, "learning_rate": 9.851329099035521e-05, "loss": 3.1677, "step": 1896 }, { "epoch": 0.15, "learning_rate": 9.851250686113071e-05, "loss": 3.0106, "step": 1897 }, { "epoch": 0.15, "learning_rate": 9.851172273190622e-05, "loss": 3.0652, "step": 1898 }, { "epoch": 0.15, "learning_rate": 9.851093860268172e-05, "loss": 3.1327, "step": 1899 }, { "epoch": 0.15, "learning_rate": 9.851015447345722e-05, "loss": 3.2451, "step": 1900 }, { "epoch": 0.15, "learning_rate": 9.850937034423273e-05, "loss": 2.9106, "step": 1901 }, { "epoch": 0.15, "learning_rate": 9.850858621500823e-05, "loss": 3.0882, "step": 1902 }, { "epoch": 0.15, "learning_rate": 9.850780208578373e-05, "loss": 2.8566, "step": 1903 }, { "epoch": 0.15, "learning_rate": 9.850701795655925e-05, "loss": 3.248, "step": 1904 }, { "epoch": 0.15, "learning_rate": 9.850623382733475e-05, "loss": 3.4263, "step": 1905 }, { "epoch": 0.15, "learning_rate": 9.850544969811026e-05, "loss": 3.1128, "step": 1906 }, { "epoch": 0.15, "learning_rate": 9.850466556888576e-05, "loss": 3.0939, "step": 1907 }, { "epoch": 0.15, "learning_rate": 9.850388143966127e-05, "loss": 3.0397, "step": 1908 }, { "epoch": 0.15, "learning_rate": 9.850309731043677e-05, "loss": 3.2601, "step": 1909 }, { "epoch": 0.15, "learning_rate": 9.850231318121227e-05, "loss": 3.2137, "step": 1910 }, { "epoch": 0.15, "learning_rate": 9.850152905198778e-05, "loss": 3.2465, "step": 1911 }, { "epoch": 0.15, "learning_rate": 9.850074492276328e-05, "loss": 3.327, "step": 1912 }, { "epoch": 0.15, "learning_rate": 9.849996079353878e-05, "loss": 3.0796, "step": 1913 }, { "epoch": 0.15, "learning_rate": 9.849917666431429e-05, "loss": 3.1992, "step": 1914 }, { "epoch": 0.15, "learning_rate": 9.849839253508979e-05, "loss": 3.0308, "step": 1915 }, { "epoch": 0.15, "learning_rate": 9.84976084058653e-05, "loss": 3.136, "step": 1916 }, { "epoch": 0.15, "learning_rate": 9.84968242766408e-05, "loss": 2.9067, "step": 1917 }, { "epoch": 0.15, "learning_rate": 9.84960401474163e-05, "loss": 2.8958, "step": 1918 }, { "epoch": 0.15, "learning_rate": 9.84952560181918e-05, "loss": 3.9009, "step": 1919 }, { "epoch": 0.15, "learning_rate": 9.849447188896731e-05, "loss": 3.2443, "step": 1920 }, { "epoch": 0.15, "learning_rate": 9.849368775974281e-05, "loss": 3.3812, "step": 1921 }, { "epoch": 0.15, "learning_rate": 9.849290363051832e-05, "loss": 3.0186, "step": 1922 }, { "epoch": 0.15, "learning_rate": 9.849211950129382e-05, "loss": 3.4192, "step": 1923 }, { "epoch": 0.15, "learning_rate": 9.849133537206932e-05, "loss": 3.1083, "step": 1924 }, { "epoch": 0.15, "learning_rate": 9.849055124284483e-05, "loss": 2.7828, "step": 1925 }, { "epoch": 0.15, "learning_rate": 9.848976711362033e-05, "loss": 2.9214, "step": 1926 }, { "epoch": 0.15, "learning_rate": 9.848898298439582e-05, "loss": 2.804, "step": 1927 }, { "epoch": 0.15, "learning_rate": 9.848819885517133e-05, "loss": 3.0976, "step": 1928 }, { "epoch": 0.15, "learning_rate": 9.848741472594683e-05, "loss": 3.2357, "step": 1929 }, { "epoch": 0.15, "learning_rate": 9.848663059672235e-05, "loss": 2.9242, "step": 1930 }, { "epoch": 0.15, "learning_rate": 9.848584646749785e-05, "loss": 2.9877, "step": 1931 }, { "epoch": 0.15, "learning_rate": 9.848506233827335e-05, "loss": 2.8793, "step": 1932 }, { "epoch": 0.15, "learning_rate": 9.848427820904886e-05, "loss": 3.0089, "step": 1933 }, { "epoch": 0.15, "learning_rate": 9.848349407982436e-05, "loss": 3.1792, "step": 1934 }, { "epoch": 0.15, "learning_rate": 9.848270995059987e-05, "loss": 2.9735, "step": 1935 }, { "epoch": 0.15, "learning_rate": 9.848192582137537e-05, "loss": 2.9568, "step": 1936 }, { "epoch": 0.15, "learning_rate": 9.848114169215087e-05, "loss": 3.0189, "step": 1937 }, { "epoch": 0.15, "learning_rate": 9.848035756292638e-05, "loss": 3.0685, "step": 1938 }, { "epoch": 0.15, "learning_rate": 9.847957343370188e-05, "loss": 3.1597, "step": 1939 }, { "epoch": 0.15, "learning_rate": 9.847878930447738e-05, "loss": 3.0914, "step": 1940 }, { "epoch": 0.15, "learning_rate": 9.847800517525289e-05, "loss": 2.962, "step": 1941 }, { "epoch": 0.15, "learning_rate": 9.847722104602839e-05, "loss": 3.4818, "step": 1942 }, { "epoch": 0.15, "learning_rate": 9.84764369168039e-05, "loss": 3.0487, "step": 1943 }, { "epoch": 0.15, "learning_rate": 9.84756527875794e-05, "loss": 3.0625, "step": 1944 }, { "epoch": 0.15, "learning_rate": 9.84748686583549e-05, "loss": 3.1229, "step": 1945 }, { "epoch": 0.15, "learning_rate": 9.84740845291304e-05, "loss": 2.9937, "step": 1946 }, { "epoch": 0.15, "learning_rate": 9.847330039990591e-05, "loss": 3.3653, "step": 1947 }, { "epoch": 0.15, "learning_rate": 9.847251627068141e-05, "loss": 3.0257, "step": 1948 }, { "epoch": 0.15, "learning_rate": 9.847173214145692e-05, "loss": 3.2434, "step": 1949 }, { "epoch": 0.15, "learning_rate": 9.847094801223242e-05, "loss": 3.0059, "step": 1950 }, { "epoch": 0.15, "learning_rate": 9.847016388300792e-05, "loss": 3.0139, "step": 1951 }, { "epoch": 0.15, "learning_rate": 9.846937975378343e-05, "loss": 3.1085, "step": 1952 }, { "epoch": 0.15, "learning_rate": 9.846859562455893e-05, "loss": 3.0486, "step": 1953 }, { "epoch": 0.15, "learning_rate": 9.846781149533443e-05, "loss": 3.2745, "step": 1954 }, { "epoch": 0.15, "learning_rate": 9.846702736610994e-05, "loss": 3.2095, "step": 1955 }, { "epoch": 0.15, "learning_rate": 9.846624323688544e-05, "loss": 2.913, "step": 1956 }, { "epoch": 0.15, "learning_rate": 9.846545910766095e-05, "loss": 2.9902, "step": 1957 }, { "epoch": 0.15, "learning_rate": 9.846467497843645e-05, "loss": 2.8432, "step": 1958 }, { "epoch": 0.15, "learning_rate": 9.846389084921195e-05, "loss": 3.3665, "step": 1959 }, { "epoch": 0.15, "learning_rate": 9.846310671998746e-05, "loss": 3.0227, "step": 1960 }, { "epoch": 0.15, "learning_rate": 9.846232259076297e-05, "loss": 3.0628, "step": 1961 }, { "epoch": 0.15, "learning_rate": 9.846153846153848e-05, "loss": 3.0113, "step": 1962 }, { "epoch": 0.15, "learning_rate": 9.846075433231397e-05, "loss": 3.0134, "step": 1963 }, { "epoch": 0.15, "learning_rate": 9.845997020308947e-05, "loss": 3.4042, "step": 1964 }, { "epoch": 0.15, "learning_rate": 9.845918607386498e-05, "loss": 3.1729, "step": 1965 }, { "epoch": 0.15, "learning_rate": 9.845840194464048e-05, "loss": 3.1096, "step": 1966 }, { "epoch": 0.15, "learning_rate": 9.845761781541598e-05, "loss": 3.2339, "step": 1967 }, { "epoch": 0.15, "learning_rate": 9.845683368619149e-05, "loss": 3.0908, "step": 1968 }, { "epoch": 0.15, "learning_rate": 9.845604955696699e-05, "loss": 3.0152, "step": 1969 }, { "epoch": 0.15, "learning_rate": 9.84552654277425e-05, "loss": 3.1414, "step": 1970 }, { "epoch": 0.15, "learning_rate": 9.8454481298518e-05, "loss": 3.1245, "step": 1971 }, { "epoch": 0.15, "learning_rate": 9.84536971692935e-05, "loss": 3.2282, "step": 1972 }, { "epoch": 0.15, "learning_rate": 9.8452913040069e-05, "loss": 3.2797, "step": 1973 }, { "epoch": 0.15, "learning_rate": 9.845212891084451e-05, "loss": 3.3251, "step": 1974 }, { "epoch": 0.15, "learning_rate": 9.845134478162001e-05, "loss": 3.1307, "step": 1975 }, { "epoch": 0.15, "learning_rate": 9.845056065239552e-05, "loss": 3.0692, "step": 1976 }, { "epoch": 0.16, "learning_rate": 9.844977652317102e-05, "loss": 3.0752, "step": 1977 }, { "epoch": 0.16, "learning_rate": 9.844899239394652e-05, "loss": 2.995, "step": 1978 }, { "epoch": 0.16, "learning_rate": 9.844820826472203e-05, "loss": 3.236, "step": 1979 }, { "epoch": 0.16, "learning_rate": 9.844742413549753e-05, "loss": 3.2715, "step": 1980 }, { "epoch": 0.16, "learning_rate": 9.844664000627303e-05, "loss": 3.1507, "step": 1981 }, { "epoch": 0.16, "learning_rate": 9.844585587704854e-05, "loss": 3.2261, "step": 1982 }, { "epoch": 0.16, "learning_rate": 9.844507174782404e-05, "loss": 3.4254, "step": 1983 }, { "epoch": 0.16, "learning_rate": 9.844428761859955e-05, "loss": 3.0573, "step": 1984 }, { "epoch": 0.16, "learning_rate": 9.844350348937505e-05, "loss": 2.8852, "step": 1985 }, { "epoch": 0.16, "learning_rate": 9.844271936015055e-05, "loss": 3.1068, "step": 1986 }, { "epoch": 0.16, "learning_rate": 9.844193523092607e-05, "loss": 3.1927, "step": 1987 }, { "epoch": 0.16, "learning_rate": 9.844115110170157e-05, "loss": 3.5253, "step": 1988 }, { "epoch": 0.16, "learning_rate": 9.844036697247708e-05, "loss": 3.366, "step": 1989 }, { "epoch": 0.16, "learning_rate": 9.843958284325258e-05, "loss": 3.2111, "step": 1990 }, { "epoch": 0.16, "learning_rate": 9.843879871402808e-05, "loss": 2.9358, "step": 1991 }, { "epoch": 0.16, "learning_rate": 9.843801458480359e-05, "loss": 2.9343, "step": 1992 }, { "epoch": 0.16, "learning_rate": 9.843723045557909e-05, "loss": 3.0403, "step": 1993 }, { "epoch": 0.16, "learning_rate": 9.84364463263546e-05, "loss": 2.8905, "step": 1994 }, { "epoch": 0.16, "learning_rate": 9.84356621971301e-05, "loss": 3.0383, "step": 1995 }, { "epoch": 0.16, "learning_rate": 9.84348780679056e-05, "loss": 3.0011, "step": 1996 }, { "epoch": 0.16, "learning_rate": 9.84340939386811e-05, "loss": 3.0942, "step": 1997 }, { "epoch": 0.16, "learning_rate": 9.843330980945661e-05, "loss": 3.1853, "step": 1998 }, { "epoch": 0.16, "learning_rate": 9.84325256802321e-05, "loss": 3.3515, "step": 1999 }, { "epoch": 0.16, "learning_rate": 9.84317415510076e-05, "loss": 3.3083, "step": 2000 }, { "epoch": 0.16, "eval_loss": 3.011477470397949, "eval_runtime": 218.2787, "eval_samples_per_second": 51.915, "eval_steps_per_second": 3.248, "step": 2000 }, { "epoch": 0.16, "learning_rate": 9.843095742178311e-05, "loss": 2.9555, "step": 2001 }, { "epoch": 0.16, "learning_rate": 9.843017329255861e-05, "loss": 2.9177, "step": 2002 }, { "epoch": 0.16, "learning_rate": 9.842938916333412e-05, "loss": 3.3575, "step": 2003 }, { "epoch": 0.16, "learning_rate": 9.842860503410962e-05, "loss": 2.8193, "step": 2004 }, { "epoch": 0.16, "learning_rate": 9.842782090488512e-05, "loss": 3.0255, "step": 2005 }, { "epoch": 0.16, "learning_rate": 9.842703677566063e-05, "loss": 2.9728, "step": 2006 }, { "epoch": 0.16, "learning_rate": 9.842625264643613e-05, "loss": 3.0044, "step": 2007 }, { "epoch": 0.16, "learning_rate": 9.842546851721163e-05, "loss": 3.1261, "step": 2008 }, { "epoch": 0.16, "learning_rate": 9.842468438798714e-05, "loss": 2.8733, "step": 2009 }, { "epoch": 0.16, "learning_rate": 9.842390025876264e-05, "loss": 3.2745, "step": 2010 }, { "epoch": 0.16, "learning_rate": 9.842311612953814e-05, "loss": 2.9864, "step": 2011 }, { "epoch": 0.16, "learning_rate": 9.842233200031366e-05, "loss": 3.1904, "step": 2012 }, { "epoch": 0.16, "learning_rate": 9.842154787108917e-05, "loss": 3.1887, "step": 2013 }, { "epoch": 0.16, "learning_rate": 9.842076374186467e-05, "loss": 3.1569, "step": 2014 }, { "epoch": 0.16, "learning_rate": 9.841997961264017e-05, "loss": 3.0843, "step": 2015 }, { "epoch": 0.16, "learning_rate": 9.841919548341568e-05, "loss": 3.378, "step": 2016 }, { "epoch": 0.16, "learning_rate": 9.841841135419118e-05, "loss": 2.9012, "step": 2017 }, { "epoch": 0.16, "learning_rate": 9.841762722496668e-05, "loss": 3.1074, "step": 2018 }, { "epoch": 0.16, "learning_rate": 9.841684309574219e-05, "loss": 2.972, "step": 2019 }, { "epoch": 0.16, "learning_rate": 9.841605896651769e-05, "loss": 3.073, "step": 2020 }, { "epoch": 0.16, "learning_rate": 9.84152748372932e-05, "loss": 3.1458, "step": 2021 }, { "epoch": 0.16, "learning_rate": 9.84144907080687e-05, "loss": 3.1506, "step": 2022 }, { "epoch": 0.16, "learning_rate": 9.84137065788442e-05, "loss": 3.084, "step": 2023 }, { "epoch": 0.16, "learning_rate": 9.84129224496197e-05, "loss": 3.06, "step": 2024 }, { "epoch": 0.16, "learning_rate": 9.841213832039521e-05, "loss": 3.1582, "step": 2025 }, { "epoch": 0.16, "learning_rate": 9.841135419117071e-05, "loss": 3.0994, "step": 2026 }, { "epoch": 0.16, "learning_rate": 9.841057006194622e-05, "loss": 3.3384, "step": 2027 }, { "epoch": 0.16, "learning_rate": 9.840978593272172e-05, "loss": 2.9624, "step": 2028 }, { "epoch": 0.16, "learning_rate": 9.840900180349722e-05, "loss": 3.0735, "step": 2029 }, { "epoch": 0.16, "learning_rate": 9.840821767427273e-05, "loss": 3.0399, "step": 2030 }, { "epoch": 0.16, "learning_rate": 9.840743354504823e-05, "loss": 2.9396, "step": 2031 }, { "epoch": 0.16, "learning_rate": 9.840664941582374e-05, "loss": 3.2, "step": 2032 }, { "epoch": 0.16, "learning_rate": 9.840586528659924e-05, "loss": 3.0073, "step": 2033 }, { "epoch": 0.16, "learning_rate": 9.840508115737474e-05, "loss": 2.8726, "step": 2034 }, { "epoch": 0.16, "learning_rate": 9.840429702815023e-05, "loss": 3.0208, "step": 2035 }, { "epoch": 0.16, "learning_rate": 9.840351289892574e-05, "loss": 3.1177, "step": 2036 }, { "epoch": 0.16, "learning_rate": 9.840272876970124e-05, "loss": 3.2624, "step": 2037 }, { "epoch": 0.16, "learning_rate": 9.840194464047676e-05, "loss": 3.0934, "step": 2038 }, { "epoch": 0.16, "learning_rate": 9.840116051125226e-05, "loss": 3.0357, "step": 2039 }, { "epoch": 0.16, "learning_rate": 9.840037638202776e-05, "loss": 3.1882, "step": 2040 }, { "epoch": 0.16, "learning_rate": 9.839959225280327e-05, "loss": 2.9084, "step": 2041 }, { "epoch": 0.16, "learning_rate": 9.839880812357877e-05, "loss": 3.1087, "step": 2042 }, { "epoch": 0.16, "learning_rate": 9.839802399435428e-05, "loss": 3.1831, "step": 2043 }, { "epoch": 0.16, "learning_rate": 9.839723986512978e-05, "loss": 3.0212, "step": 2044 }, { "epoch": 0.16, "learning_rate": 9.839645573590528e-05, "loss": 3.2572, "step": 2045 }, { "epoch": 0.16, "learning_rate": 9.839567160668079e-05, "loss": 3.0279, "step": 2046 }, { "epoch": 0.16, "learning_rate": 9.839488747745629e-05, "loss": 3.0198, "step": 2047 }, { "epoch": 0.16, "learning_rate": 9.83941033482318e-05, "loss": 2.8282, "step": 2048 }, { "epoch": 0.16, "learning_rate": 9.83933192190073e-05, "loss": 3.0135, "step": 2049 }, { "epoch": 0.16, "learning_rate": 9.83925350897828e-05, "loss": 3.1292, "step": 2050 }, { "epoch": 0.16, "learning_rate": 9.83917509605583e-05, "loss": 3.203, "step": 2051 }, { "epoch": 0.16, "learning_rate": 9.839096683133381e-05, "loss": 3.1743, "step": 2052 }, { "epoch": 0.16, "learning_rate": 9.839018270210931e-05, "loss": 3.1239, "step": 2053 }, { "epoch": 0.16, "learning_rate": 9.838939857288482e-05, "loss": 3.0378, "step": 2054 }, { "epoch": 0.16, "learning_rate": 9.838861444366032e-05, "loss": 3.0319, "step": 2055 }, { "epoch": 0.16, "learning_rate": 9.838783031443582e-05, "loss": 3.0349, "step": 2056 }, { "epoch": 0.16, "learning_rate": 9.838704618521133e-05, "loss": 3.0503, "step": 2057 }, { "epoch": 0.16, "learning_rate": 9.838626205598683e-05, "loss": 2.8861, "step": 2058 }, { "epoch": 0.16, "learning_rate": 9.838547792676233e-05, "loss": 2.9436, "step": 2059 }, { "epoch": 0.16, "learning_rate": 9.838469379753784e-05, "loss": 3.2964, "step": 2060 }, { "epoch": 0.16, "learning_rate": 9.838390966831334e-05, "loss": 3.3665, "step": 2061 }, { "epoch": 0.16, "learning_rate": 9.838312553908885e-05, "loss": 3.0664, "step": 2062 }, { "epoch": 0.16, "learning_rate": 9.838234140986435e-05, "loss": 3.0699, "step": 2063 }, { "epoch": 0.16, "learning_rate": 9.838155728063985e-05, "loss": 2.9023, "step": 2064 }, { "epoch": 0.16, "learning_rate": 9.838077315141536e-05, "loss": 3.0811, "step": 2065 }, { "epoch": 0.16, "learning_rate": 9.837998902219086e-05, "loss": 3.1035, "step": 2066 }, { "epoch": 0.16, "learning_rate": 9.837920489296636e-05, "loss": 3.3457, "step": 2067 }, { "epoch": 0.16, "learning_rate": 9.837842076374187e-05, "loss": 3.4628, "step": 2068 }, { "epoch": 0.16, "learning_rate": 9.837763663451738e-05, "loss": 3.2268, "step": 2069 }, { "epoch": 0.16, "learning_rate": 9.837685250529289e-05, "loss": 3.2523, "step": 2070 }, { "epoch": 0.16, "learning_rate": 9.837606837606838e-05, "loss": 2.9634, "step": 2071 }, { "epoch": 0.16, "learning_rate": 9.837528424684388e-05, "loss": 3.0922, "step": 2072 }, { "epoch": 0.16, "learning_rate": 9.837450011761939e-05, "loss": 3.0814, "step": 2073 }, { "epoch": 0.16, "learning_rate": 9.837371598839489e-05, "loss": 3.3619, "step": 2074 }, { "epoch": 0.16, "learning_rate": 9.83729318591704e-05, "loss": 3.48, "step": 2075 }, { "epoch": 0.16, "learning_rate": 9.83721477299459e-05, "loss": 3.4349, "step": 2076 }, { "epoch": 0.16, "learning_rate": 9.83713636007214e-05, "loss": 3.0537, "step": 2077 }, { "epoch": 0.16, "learning_rate": 9.83705794714969e-05, "loss": 3.2636, "step": 2078 }, { "epoch": 0.16, "learning_rate": 9.836979534227241e-05, "loss": 3.2567, "step": 2079 }, { "epoch": 0.16, "learning_rate": 9.836901121304791e-05, "loss": 3.4791, "step": 2080 }, { "epoch": 0.16, "learning_rate": 9.836822708382342e-05, "loss": 3.4451, "step": 2081 }, { "epoch": 0.16, "learning_rate": 9.836744295459892e-05, "loss": 2.8824, "step": 2082 }, { "epoch": 0.16, "learning_rate": 9.836665882537442e-05, "loss": 3.3786, "step": 2083 }, { "epoch": 0.16, "learning_rate": 9.836587469614993e-05, "loss": 3.2116, "step": 2084 }, { "epoch": 0.16, "learning_rate": 9.836509056692543e-05, "loss": 3.3205, "step": 2085 }, { "epoch": 0.16, "learning_rate": 9.836430643770093e-05, "loss": 3.3194, "step": 2086 }, { "epoch": 0.16, "learning_rate": 9.836352230847644e-05, "loss": 3.1649, "step": 2087 }, { "epoch": 0.16, "learning_rate": 9.836273817925194e-05, "loss": 3.0789, "step": 2088 }, { "epoch": 0.16, "learning_rate": 9.836195405002745e-05, "loss": 2.8328, "step": 2089 }, { "epoch": 0.16, "learning_rate": 9.836116992080295e-05, "loss": 3.1033, "step": 2090 }, { "epoch": 0.16, "learning_rate": 9.836038579157845e-05, "loss": 3.0025, "step": 2091 }, { "epoch": 0.16, "learning_rate": 9.835960166235396e-05, "loss": 3.3175, "step": 2092 }, { "epoch": 0.16, "learning_rate": 9.835881753312946e-05, "loss": 2.9928, "step": 2093 }, { "epoch": 0.16, "learning_rate": 9.835803340390496e-05, "loss": 3.0152, "step": 2094 }, { "epoch": 0.16, "learning_rate": 9.835724927468048e-05, "loss": 2.8943, "step": 2095 }, { "epoch": 0.16, "learning_rate": 9.835646514545598e-05, "loss": 3.3058, "step": 2096 }, { "epoch": 0.16, "learning_rate": 9.835568101623149e-05, "loss": 3.325, "step": 2097 }, { "epoch": 0.16, "learning_rate": 9.835489688700699e-05, "loss": 3.0701, "step": 2098 }, { "epoch": 0.16, "learning_rate": 9.83541127577825e-05, "loss": 2.9556, "step": 2099 }, { "epoch": 0.16, "learning_rate": 9.8353328628558e-05, "loss": 3.0182, "step": 2100 }, { "epoch": 0.16, "learning_rate": 9.83525444993335e-05, "loss": 3.2172, "step": 2101 }, { "epoch": 0.16, "learning_rate": 9.8351760370109e-05, "loss": 3.2256, "step": 2102 }, { "epoch": 0.16, "learning_rate": 9.835097624088451e-05, "loss": 3.1559, "step": 2103 }, { "epoch": 0.16, "learning_rate": 9.835019211166001e-05, "loss": 3.378, "step": 2104 }, { "epoch": 0.17, "learning_rate": 9.834940798243552e-05, "loss": 2.9946, "step": 2105 }, { "epoch": 0.17, "learning_rate": 9.834862385321102e-05, "loss": 2.8826, "step": 2106 }, { "epoch": 0.17, "learning_rate": 9.834783972398651e-05, "loss": 3.2827, "step": 2107 }, { "epoch": 0.17, "learning_rate": 9.834705559476201e-05, "loss": 3.0231, "step": 2108 }, { "epoch": 0.17, "learning_rate": 9.834627146553752e-05, "loss": 3.0821, "step": 2109 }, { "epoch": 0.17, "learning_rate": 9.834548733631302e-05, "loss": 3.1198, "step": 2110 }, { "epoch": 0.17, "learning_rate": 9.834470320708853e-05, "loss": 3.2491, "step": 2111 }, { "epoch": 0.17, "learning_rate": 9.834391907786403e-05, "loss": 2.9674, "step": 2112 }, { "epoch": 0.17, "learning_rate": 9.834313494863953e-05, "loss": 2.9006, "step": 2113 }, { "epoch": 0.17, "learning_rate": 9.834235081941504e-05, "loss": 3.1887, "step": 2114 }, { "epoch": 0.17, "learning_rate": 9.834156669019054e-05, "loss": 3.027, "step": 2115 }, { "epoch": 0.17, "learning_rate": 9.834078256096604e-05, "loss": 3.373, "step": 2116 }, { "epoch": 0.17, "learning_rate": 9.833999843174155e-05, "loss": 2.9841, "step": 2117 }, { "epoch": 0.17, "learning_rate": 9.833921430251705e-05, "loss": 3.2248, "step": 2118 }, { "epoch": 0.17, "learning_rate": 9.833843017329256e-05, "loss": 3.0488, "step": 2119 }, { "epoch": 0.17, "learning_rate": 9.833764604406807e-05, "loss": 2.8245, "step": 2120 }, { "epoch": 0.17, "learning_rate": 9.833686191484358e-05, "loss": 3.0899, "step": 2121 }, { "epoch": 0.17, "learning_rate": 9.833607778561908e-05, "loss": 2.9273, "step": 2122 }, { "epoch": 0.17, "learning_rate": 9.833529365639458e-05, "loss": 3.4119, "step": 2123 }, { "epoch": 0.17, "learning_rate": 9.833450952717009e-05, "loss": 3.0801, "step": 2124 }, { "epoch": 0.17, "learning_rate": 9.833372539794559e-05, "loss": 2.9374, "step": 2125 }, { "epoch": 0.17, "learning_rate": 9.83329412687211e-05, "loss": 3.0202, "step": 2126 }, { "epoch": 0.17, "learning_rate": 9.83321571394966e-05, "loss": 2.9047, "step": 2127 }, { "epoch": 0.17, "learning_rate": 9.83313730102721e-05, "loss": 3.1377, "step": 2128 }, { "epoch": 0.17, "learning_rate": 9.83305888810476e-05, "loss": 2.9644, "step": 2129 }, { "epoch": 0.17, "learning_rate": 9.832980475182311e-05, "loss": 3.4691, "step": 2130 }, { "epoch": 0.17, "learning_rate": 9.832902062259861e-05, "loss": 3.0146, "step": 2131 }, { "epoch": 0.17, "learning_rate": 9.832823649337412e-05, "loss": 2.8934, "step": 2132 }, { "epoch": 0.17, "learning_rate": 9.832745236414962e-05, "loss": 3.4879, "step": 2133 }, { "epoch": 0.17, "learning_rate": 9.832666823492512e-05, "loss": 3.0744, "step": 2134 }, { "epoch": 0.17, "learning_rate": 9.832588410570063e-05, "loss": 2.9799, "step": 2135 }, { "epoch": 0.17, "learning_rate": 9.832509997647613e-05, "loss": 3.0143, "step": 2136 }, { "epoch": 0.17, "learning_rate": 9.832431584725164e-05, "loss": 3.2109, "step": 2137 }, { "epoch": 0.17, "learning_rate": 9.832353171802714e-05, "loss": 3.0843, "step": 2138 }, { "epoch": 0.17, "learning_rate": 9.832274758880264e-05, "loss": 3.1361, "step": 2139 }, { "epoch": 0.17, "learning_rate": 9.832196345957815e-05, "loss": 3.253, "step": 2140 }, { "epoch": 0.17, "learning_rate": 9.832117933035365e-05, "loss": 2.7297, "step": 2141 }, { "epoch": 0.17, "learning_rate": 9.832039520112915e-05, "loss": 2.9769, "step": 2142 }, { "epoch": 0.17, "learning_rate": 9.831961107190464e-05, "loss": 2.8427, "step": 2143 }, { "epoch": 0.17, "learning_rate": 9.831882694268015e-05, "loss": 2.9059, "step": 2144 }, { "epoch": 0.17, "learning_rate": 9.831804281345565e-05, "loss": 3.4742, "step": 2145 }, { "epoch": 0.17, "learning_rate": 9.831725868423117e-05, "loss": 3.12, "step": 2146 }, { "epoch": 0.17, "learning_rate": 9.831647455500667e-05, "loss": 3.0143, "step": 2147 }, { "epoch": 0.17, "learning_rate": 9.831569042578218e-05, "loss": 3.1035, "step": 2148 }, { "epoch": 0.17, "learning_rate": 9.831490629655768e-05, "loss": 3.0938, "step": 2149 }, { "epoch": 0.17, "learning_rate": 9.831412216733318e-05, "loss": 3.1101, "step": 2150 }, { "epoch": 0.17, "learning_rate": 9.831333803810869e-05, "loss": 3.0384, "step": 2151 }, { "epoch": 0.17, "learning_rate": 9.831255390888419e-05, "loss": 3.1315, "step": 2152 }, { "epoch": 0.17, "learning_rate": 9.83117697796597e-05, "loss": 3.074, "step": 2153 }, { "epoch": 0.17, "learning_rate": 9.83109856504352e-05, "loss": 3.2676, "step": 2154 }, { "epoch": 0.17, "learning_rate": 9.83102015212107e-05, "loss": 3.1245, "step": 2155 }, { "epoch": 0.17, "learning_rate": 9.83094173919862e-05, "loss": 3.1556, "step": 2156 }, { "epoch": 0.17, "learning_rate": 9.830863326276171e-05, "loss": 3.2886, "step": 2157 }, { "epoch": 0.17, "learning_rate": 9.830784913353721e-05, "loss": 2.9979, "step": 2158 }, { "epoch": 0.17, "learning_rate": 9.830706500431272e-05, "loss": 3.2754, "step": 2159 }, { "epoch": 0.17, "learning_rate": 9.830628087508822e-05, "loss": 3.141, "step": 2160 }, { "epoch": 0.17, "learning_rate": 9.830549674586372e-05, "loss": 3.2098, "step": 2161 }, { "epoch": 0.17, "learning_rate": 9.830471261663923e-05, "loss": 3.0858, "step": 2162 }, { "epoch": 0.17, "learning_rate": 9.830392848741473e-05, "loss": 2.9725, "step": 2163 }, { "epoch": 0.17, "learning_rate": 9.830314435819023e-05, "loss": 3.324, "step": 2164 }, { "epoch": 0.17, "learning_rate": 9.830236022896574e-05, "loss": 3.0928, "step": 2165 }, { "epoch": 0.17, "learning_rate": 9.830157609974124e-05, "loss": 3.0639, "step": 2166 }, { "epoch": 0.17, "learning_rate": 9.830079197051675e-05, "loss": 3.0407, "step": 2167 }, { "epoch": 0.17, "learning_rate": 9.830000784129225e-05, "loss": 3.2383, "step": 2168 }, { "epoch": 0.17, "learning_rate": 9.829922371206775e-05, "loss": 3.0251, "step": 2169 }, { "epoch": 0.17, "learning_rate": 9.829843958284326e-05, "loss": 3.2353, "step": 2170 }, { "epoch": 0.17, "learning_rate": 9.829765545361876e-05, "loss": 3.3938, "step": 2171 }, { "epoch": 0.17, "learning_rate": 9.829687132439426e-05, "loss": 3.3046, "step": 2172 }, { "epoch": 0.17, "learning_rate": 9.829608719516977e-05, "loss": 3.2257, "step": 2173 }, { "epoch": 0.17, "learning_rate": 9.829530306594527e-05, "loss": 3.203, "step": 2174 }, { "epoch": 0.17, "learning_rate": 9.829451893672077e-05, "loss": 2.9598, "step": 2175 }, { "epoch": 0.17, "learning_rate": 9.829373480749628e-05, "loss": 2.8838, "step": 2176 }, { "epoch": 0.17, "learning_rate": 9.82929506782718e-05, "loss": 2.9839, "step": 2177 }, { "epoch": 0.17, "learning_rate": 9.82921665490473e-05, "loss": 2.9982, "step": 2178 }, { "epoch": 0.17, "learning_rate": 9.829138241982279e-05, "loss": 3.0523, "step": 2179 }, { "epoch": 0.17, "learning_rate": 9.829059829059829e-05, "loss": 2.9694, "step": 2180 }, { "epoch": 0.17, "learning_rate": 9.82898141613738e-05, "loss": 3.1182, "step": 2181 }, { "epoch": 0.17, "learning_rate": 9.82890300321493e-05, "loss": 2.8598, "step": 2182 }, { "epoch": 0.17, "learning_rate": 9.82882459029248e-05, "loss": 2.8364, "step": 2183 }, { "epoch": 0.17, "learning_rate": 9.828746177370031e-05, "loss": 3.0926, "step": 2184 }, { "epoch": 0.17, "learning_rate": 9.828667764447581e-05, "loss": 3.1492, "step": 2185 }, { "epoch": 0.17, "learning_rate": 9.828589351525132e-05, "loss": 3.1187, "step": 2186 }, { "epoch": 0.17, "learning_rate": 9.828510938602682e-05, "loss": 3.2128, "step": 2187 }, { "epoch": 0.17, "learning_rate": 9.828432525680232e-05, "loss": 3.3513, "step": 2188 }, { "epoch": 0.17, "learning_rate": 9.828354112757783e-05, "loss": 2.9962, "step": 2189 }, { "epoch": 0.17, "learning_rate": 9.828275699835333e-05, "loss": 3.0076, "step": 2190 }, { "epoch": 0.17, "learning_rate": 9.828197286912883e-05, "loss": 3.4033, "step": 2191 }, { "epoch": 0.17, "learning_rate": 9.828118873990434e-05, "loss": 2.9684, "step": 2192 }, { "epoch": 0.17, "learning_rate": 9.828040461067984e-05, "loss": 2.9362, "step": 2193 }, { "epoch": 0.17, "learning_rate": 9.827962048145534e-05, "loss": 3.01, "step": 2194 }, { "epoch": 0.17, "learning_rate": 9.827883635223085e-05, "loss": 3.0849, "step": 2195 }, { "epoch": 0.17, "learning_rate": 9.827805222300635e-05, "loss": 2.9615, "step": 2196 }, { "epoch": 0.17, "learning_rate": 9.827726809378186e-05, "loss": 3.1153, "step": 2197 }, { "epoch": 0.17, "learning_rate": 9.827648396455736e-05, "loss": 3.1516, "step": 2198 }, { "epoch": 0.17, "learning_rate": 9.827569983533286e-05, "loss": 3.2334, "step": 2199 }, { "epoch": 0.17, "learning_rate": 9.827491570610837e-05, "loss": 3.3184, "step": 2200 }, { "epoch": 0.17, "learning_rate": 9.827413157688387e-05, "loss": 3.0758, "step": 2201 }, { "epoch": 0.17, "learning_rate": 9.827334744765937e-05, "loss": 3.0789, "step": 2202 }, { "epoch": 0.17, "learning_rate": 9.827256331843489e-05, "loss": 3.1817, "step": 2203 }, { "epoch": 0.17, "learning_rate": 9.82717791892104e-05, "loss": 3.3449, "step": 2204 }, { "epoch": 0.17, "learning_rate": 9.82709950599859e-05, "loss": 2.9832, "step": 2205 }, { "epoch": 0.17, "learning_rate": 9.82702109307614e-05, "loss": 3.0679, "step": 2206 }, { "epoch": 0.17, "learning_rate": 9.82694268015369e-05, "loss": 3.3506, "step": 2207 }, { "epoch": 0.17, "learning_rate": 9.826864267231241e-05, "loss": 3.1003, "step": 2208 }, { "epoch": 0.17, "learning_rate": 9.826785854308791e-05, "loss": 3.1034, "step": 2209 }, { "epoch": 0.17, "learning_rate": 9.826707441386342e-05, "loss": 3.1754, "step": 2210 }, { "epoch": 0.17, "learning_rate": 9.826629028463892e-05, "loss": 2.9961, "step": 2211 }, { "epoch": 0.17, "learning_rate": 9.826550615541442e-05, "loss": 3.1612, "step": 2212 }, { "epoch": 0.17, "learning_rate": 9.826472202618993e-05, "loss": 3.0874, "step": 2213 }, { "epoch": 0.17, "learning_rate": 9.826393789696543e-05, "loss": 3.2744, "step": 2214 }, { "epoch": 0.17, "learning_rate": 9.826315376774092e-05, "loss": 3.2152, "step": 2215 }, { "epoch": 0.17, "learning_rate": 9.826236963851643e-05, "loss": 3.0536, "step": 2216 }, { "epoch": 0.17, "learning_rate": 9.826158550929193e-05, "loss": 3.1146, "step": 2217 }, { "epoch": 0.17, "learning_rate": 9.826080138006743e-05, "loss": 3.0069, "step": 2218 }, { "epoch": 0.17, "learning_rate": 9.826001725084294e-05, "loss": 3.0925, "step": 2219 }, { "epoch": 0.17, "learning_rate": 9.825923312161844e-05, "loss": 2.9104, "step": 2220 }, { "epoch": 0.17, "learning_rate": 9.825844899239394e-05, "loss": 3.0253, "step": 2221 }, { "epoch": 0.17, "learning_rate": 9.825766486316945e-05, "loss": 2.9858, "step": 2222 }, { "epoch": 0.17, "learning_rate": 9.825688073394495e-05, "loss": 3.2542, "step": 2223 }, { "epoch": 0.17, "learning_rate": 9.825609660472046e-05, "loss": 3.4353, "step": 2224 }, { "epoch": 0.17, "learning_rate": 9.825531247549596e-05, "loss": 3.0763, "step": 2225 }, { "epoch": 0.17, "learning_rate": 9.825452834627146e-05, "loss": 3.1562, "step": 2226 }, { "epoch": 0.17, "learning_rate": 9.825374421704697e-05, "loss": 2.9341, "step": 2227 }, { "epoch": 0.17, "learning_rate": 9.825296008782247e-05, "loss": 2.9908, "step": 2228 }, { "epoch": 0.17, "learning_rate": 9.825217595859799e-05, "loss": 2.9937, "step": 2229 }, { "epoch": 0.17, "learning_rate": 9.825139182937349e-05, "loss": 3.0366, "step": 2230 }, { "epoch": 0.17, "learning_rate": 9.8250607700149e-05, "loss": 3.0454, "step": 2231 }, { "epoch": 0.18, "learning_rate": 9.82498235709245e-05, "loss": 3.0761, "step": 2232 }, { "epoch": 0.18, "learning_rate": 9.82490394417e-05, "loss": 3.1229, "step": 2233 }, { "epoch": 0.18, "learning_rate": 9.82482553124755e-05, "loss": 2.9429, "step": 2234 }, { "epoch": 0.18, "learning_rate": 9.824747118325101e-05, "loss": 3.0489, "step": 2235 }, { "epoch": 0.18, "learning_rate": 9.824668705402651e-05, "loss": 3.2631, "step": 2236 }, { "epoch": 0.18, "learning_rate": 9.824590292480202e-05, "loss": 3.2208, "step": 2237 }, { "epoch": 0.18, "learning_rate": 9.824511879557752e-05, "loss": 3.4695, "step": 2238 }, { "epoch": 0.18, "learning_rate": 9.824433466635302e-05, "loss": 3.1277, "step": 2239 }, { "epoch": 0.18, "learning_rate": 9.824355053712853e-05, "loss": 3.2278, "step": 2240 }, { "epoch": 0.18, "learning_rate": 9.824276640790403e-05, "loss": 2.8215, "step": 2241 }, { "epoch": 0.18, "learning_rate": 9.824198227867953e-05, "loss": 3.1653, "step": 2242 }, { "epoch": 0.18, "learning_rate": 9.824119814945504e-05, "loss": 3.2767, "step": 2243 }, { "epoch": 0.18, "learning_rate": 9.824041402023054e-05, "loss": 3.1379, "step": 2244 }, { "epoch": 0.18, "learning_rate": 9.823962989100605e-05, "loss": 3.0959, "step": 2245 }, { "epoch": 0.18, "learning_rate": 9.823884576178155e-05, "loss": 2.9805, "step": 2246 }, { "epoch": 0.18, "learning_rate": 9.823806163255705e-05, "loss": 3.2688, "step": 2247 }, { "epoch": 0.18, "learning_rate": 9.823727750333256e-05, "loss": 3.0483, "step": 2248 }, { "epoch": 0.18, "learning_rate": 9.823649337410806e-05, "loss": 3.3104, "step": 2249 }, { "epoch": 0.18, "learning_rate": 9.823570924488356e-05, "loss": 3.0653, "step": 2250 }, { "epoch": 0.18, "learning_rate": 9.823492511565905e-05, "loss": 3.1612, "step": 2251 }, { "epoch": 0.18, "learning_rate": 9.823414098643456e-05, "loss": 3.2119, "step": 2252 }, { "epoch": 0.18, "learning_rate": 9.823335685721006e-05, "loss": 3.2205, "step": 2253 }, { "epoch": 0.18, "learning_rate": 9.823257272798558e-05, "loss": 3.1711, "step": 2254 }, { "epoch": 0.18, "learning_rate": 9.823178859876108e-05, "loss": 3.0817, "step": 2255 }, { "epoch": 0.18, "learning_rate": 9.823100446953659e-05, "loss": 3.1748, "step": 2256 }, { "epoch": 0.18, "learning_rate": 9.823022034031209e-05, "loss": 3.1893, "step": 2257 }, { "epoch": 0.18, "learning_rate": 9.82294362110876e-05, "loss": 3.112, "step": 2258 }, { "epoch": 0.18, "learning_rate": 9.82286520818631e-05, "loss": 3.0286, "step": 2259 }, { "epoch": 0.18, "learning_rate": 9.82278679526386e-05, "loss": 3.2695, "step": 2260 }, { "epoch": 0.18, "learning_rate": 9.82270838234141e-05, "loss": 3.1013, "step": 2261 }, { "epoch": 0.18, "learning_rate": 9.822629969418961e-05, "loss": 3.0454, "step": 2262 }, { "epoch": 0.18, "learning_rate": 9.822551556496511e-05, "loss": 3.1477, "step": 2263 }, { "epoch": 0.18, "learning_rate": 9.822473143574062e-05, "loss": 3.0171, "step": 2264 }, { "epoch": 0.18, "learning_rate": 9.822394730651612e-05, "loss": 3.0585, "step": 2265 }, { "epoch": 0.18, "learning_rate": 9.822316317729162e-05, "loss": 3.1789, "step": 2266 }, { "epoch": 0.18, "learning_rate": 9.822237904806713e-05, "loss": 2.9331, "step": 2267 }, { "epoch": 0.18, "learning_rate": 9.822159491884263e-05, "loss": 2.9861, "step": 2268 }, { "epoch": 0.18, "learning_rate": 9.822081078961813e-05, "loss": 2.7833, "step": 2269 }, { "epoch": 0.18, "learning_rate": 9.822002666039364e-05, "loss": 3.0427, "step": 2270 }, { "epoch": 0.18, "learning_rate": 9.821924253116914e-05, "loss": 3.2618, "step": 2271 }, { "epoch": 0.18, "learning_rate": 9.821845840194465e-05, "loss": 3.2892, "step": 2272 }, { "epoch": 0.18, "learning_rate": 9.821767427272015e-05, "loss": 3.1442, "step": 2273 }, { "epoch": 0.18, "learning_rate": 9.821689014349565e-05, "loss": 3.1387, "step": 2274 }, { "epoch": 0.18, "learning_rate": 9.821610601427116e-05, "loss": 3.0647, "step": 2275 }, { "epoch": 0.18, "learning_rate": 9.821532188504666e-05, "loss": 3.3126, "step": 2276 }, { "epoch": 0.18, "learning_rate": 9.821453775582216e-05, "loss": 3.007, "step": 2277 }, { "epoch": 0.18, "learning_rate": 9.821375362659767e-05, "loss": 3.2139, "step": 2278 }, { "epoch": 0.18, "learning_rate": 9.821296949737317e-05, "loss": 3.0236, "step": 2279 }, { "epoch": 0.18, "learning_rate": 9.821218536814867e-05, "loss": 3.153, "step": 2280 }, { "epoch": 0.18, "learning_rate": 9.821140123892418e-05, "loss": 3.213, "step": 2281 }, { "epoch": 0.18, "learning_rate": 9.821061710969968e-05, "loss": 3.0033, "step": 2282 }, { "epoch": 0.18, "learning_rate": 9.820983298047519e-05, "loss": 3.2776, "step": 2283 }, { "epoch": 0.18, "learning_rate": 9.820904885125069e-05, "loss": 3.1796, "step": 2284 }, { "epoch": 0.18, "learning_rate": 9.82082647220262e-05, "loss": 3.1393, "step": 2285 }, { "epoch": 0.18, "learning_rate": 9.820748059280171e-05, "loss": 3.2074, "step": 2286 }, { "epoch": 0.18, "learning_rate": 9.82066964635772e-05, "loss": 3.2275, "step": 2287 }, { "epoch": 0.18, "learning_rate": 9.82059123343527e-05, "loss": 2.9571, "step": 2288 }, { "epoch": 0.18, "learning_rate": 9.820512820512821e-05, "loss": 3.027, "step": 2289 }, { "epoch": 0.18, "learning_rate": 9.820434407590371e-05, "loss": 2.8584, "step": 2290 }, { "epoch": 0.18, "learning_rate": 9.820355994667922e-05, "loss": 3.1164, "step": 2291 }, { "epoch": 0.18, "learning_rate": 9.820277581745472e-05, "loss": 3.0151, "step": 2292 }, { "epoch": 0.18, "learning_rate": 9.820199168823022e-05, "loss": 2.9262, "step": 2293 }, { "epoch": 0.18, "learning_rate": 9.820120755900573e-05, "loss": 2.9683, "step": 2294 }, { "epoch": 0.18, "learning_rate": 9.820042342978123e-05, "loss": 3.2365, "step": 2295 }, { "epoch": 0.18, "learning_rate": 9.819963930055673e-05, "loss": 3.266, "step": 2296 }, { "epoch": 0.18, "learning_rate": 9.819885517133224e-05, "loss": 2.9583, "step": 2297 }, { "epoch": 0.18, "learning_rate": 9.819807104210774e-05, "loss": 3.0837, "step": 2298 }, { "epoch": 0.18, "learning_rate": 9.819728691288324e-05, "loss": 3.1597, "step": 2299 }, { "epoch": 0.18, "learning_rate": 9.819650278365875e-05, "loss": 3.4857, "step": 2300 }, { "epoch": 0.18, "learning_rate": 9.819571865443425e-05, "loss": 3.1851, "step": 2301 }, { "epoch": 0.18, "learning_rate": 9.819493452520976e-05, "loss": 3.1176, "step": 2302 }, { "epoch": 0.18, "learning_rate": 9.819415039598526e-05, "loss": 2.9032, "step": 2303 }, { "epoch": 0.18, "learning_rate": 9.819336626676076e-05, "loss": 2.9363, "step": 2304 }, { "epoch": 0.18, "learning_rate": 9.819258213753627e-05, "loss": 3.0243, "step": 2305 }, { "epoch": 0.18, "learning_rate": 9.819179800831177e-05, "loss": 2.9431, "step": 2306 }, { "epoch": 0.18, "learning_rate": 9.819101387908727e-05, "loss": 3.2319, "step": 2307 }, { "epoch": 0.18, "learning_rate": 9.819022974986278e-05, "loss": 3.1894, "step": 2308 }, { "epoch": 0.18, "learning_rate": 9.818944562063828e-05, "loss": 3.2393, "step": 2309 }, { "epoch": 0.18, "learning_rate": 9.818866149141378e-05, "loss": 3.2716, "step": 2310 }, { "epoch": 0.18, "learning_rate": 9.81878773621893e-05, "loss": 3.4703, "step": 2311 }, { "epoch": 0.18, "learning_rate": 9.81870932329648e-05, "loss": 3.3697, "step": 2312 }, { "epoch": 0.18, "learning_rate": 9.818630910374031e-05, "loss": 3.0776, "step": 2313 }, { "epoch": 0.18, "learning_rate": 9.818552497451581e-05, "loss": 3.0804, "step": 2314 }, { "epoch": 0.18, "learning_rate": 9.818474084529132e-05, "loss": 3.0557, "step": 2315 }, { "epoch": 0.18, "learning_rate": 9.818395671606682e-05, "loss": 3.0183, "step": 2316 }, { "epoch": 0.18, "learning_rate": 9.818317258684232e-05, "loss": 3.4702, "step": 2317 }, { "epoch": 0.18, "learning_rate": 9.818238845761783e-05, "loss": 3.0577, "step": 2318 }, { "epoch": 0.18, "learning_rate": 9.818160432839333e-05, "loss": 3.127, "step": 2319 }, { "epoch": 0.18, "learning_rate": 9.818082019916884e-05, "loss": 3.0976, "step": 2320 }, { "epoch": 0.18, "learning_rate": 9.818003606994434e-05, "loss": 2.936, "step": 2321 }, { "epoch": 0.18, "learning_rate": 9.817925194071984e-05, "loss": 3.0401, "step": 2322 }, { "epoch": 0.18, "learning_rate": 9.817846781149533e-05, "loss": 2.8595, "step": 2323 }, { "epoch": 0.18, "learning_rate": 9.817768368227084e-05, "loss": 2.9202, "step": 2324 }, { "epoch": 0.18, "learning_rate": 9.817689955304634e-05, "loss": 3.2472, "step": 2325 }, { "epoch": 0.18, "learning_rate": 9.817611542382184e-05, "loss": 3.2703, "step": 2326 }, { "epoch": 0.18, "learning_rate": 9.817533129459735e-05, "loss": 3.1616, "step": 2327 }, { "epoch": 0.18, "learning_rate": 9.817454716537285e-05, "loss": 3.1688, "step": 2328 }, { "epoch": 0.18, "learning_rate": 9.817376303614835e-05, "loss": 2.9198, "step": 2329 }, { "epoch": 0.18, "learning_rate": 9.817297890692386e-05, "loss": 2.9219, "step": 2330 }, { "epoch": 0.18, "learning_rate": 9.817219477769936e-05, "loss": 3.1313, "step": 2331 }, { "epoch": 0.18, "learning_rate": 9.817141064847487e-05, "loss": 3.387, "step": 2332 }, { "epoch": 0.18, "learning_rate": 9.817062651925037e-05, "loss": 2.9916, "step": 2333 }, { "epoch": 0.18, "learning_rate": 9.816984239002587e-05, "loss": 2.9408, "step": 2334 }, { "epoch": 0.18, "learning_rate": 9.816905826080138e-05, "loss": 3.0786, "step": 2335 }, { "epoch": 0.18, "learning_rate": 9.816827413157688e-05, "loss": 3.5478, "step": 2336 }, { "epoch": 0.18, "learning_rate": 9.81674900023524e-05, "loss": 3.284, "step": 2337 }, { "epoch": 0.18, "learning_rate": 9.81667058731279e-05, "loss": 3.4303, "step": 2338 }, { "epoch": 0.18, "learning_rate": 9.81659217439034e-05, "loss": 2.9725, "step": 2339 }, { "epoch": 0.18, "learning_rate": 9.816513761467891e-05, "loss": 3.06, "step": 2340 }, { "epoch": 0.18, "learning_rate": 9.816435348545441e-05, "loss": 2.9725, "step": 2341 }, { "epoch": 0.18, "learning_rate": 9.816356935622992e-05, "loss": 3.1334, "step": 2342 }, { "epoch": 0.18, "learning_rate": 9.816278522700542e-05, "loss": 3.4486, "step": 2343 }, { "epoch": 0.18, "learning_rate": 9.816200109778092e-05, "loss": 3.3412, "step": 2344 }, { "epoch": 0.18, "learning_rate": 9.816121696855643e-05, "loss": 3.1556, "step": 2345 }, { "epoch": 0.18, "learning_rate": 9.816043283933193e-05, "loss": 2.834, "step": 2346 }, { "epoch": 0.18, "learning_rate": 9.815964871010743e-05, "loss": 3.1126, "step": 2347 }, { "epoch": 0.18, "learning_rate": 9.815886458088294e-05, "loss": 3.1698, "step": 2348 }, { "epoch": 0.18, "learning_rate": 9.815808045165844e-05, "loss": 3.0203, "step": 2349 }, { "epoch": 0.18, "learning_rate": 9.815729632243395e-05, "loss": 3.1294, "step": 2350 }, { "epoch": 0.18, "learning_rate": 9.815651219320945e-05, "loss": 2.9242, "step": 2351 }, { "epoch": 0.18, "learning_rate": 9.815572806398495e-05, "loss": 3.3402, "step": 2352 }, { "epoch": 0.18, "learning_rate": 9.815494393476046e-05, "loss": 3.1822, "step": 2353 }, { "epoch": 0.18, "learning_rate": 9.815415980553596e-05, "loss": 3.1527, "step": 2354 }, { "epoch": 0.18, "learning_rate": 9.815337567631146e-05, "loss": 3.1361, "step": 2355 }, { "epoch": 0.18, "learning_rate": 9.815259154708697e-05, "loss": 3.1039, "step": 2356 }, { "epoch": 0.18, "learning_rate": 9.815180741786247e-05, "loss": 2.8433, "step": 2357 }, { "epoch": 0.18, "learning_rate": 9.815102328863797e-05, "loss": 3.0802, "step": 2358 }, { "epoch": 0.18, "learning_rate": 9.815023915941347e-05, "loss": 2.813, "step": 2359 }, { "epoch": 0.19, "learning_rate": 9.814945503018897e-05, "loss": 3.2319, "step": 2360 }, { "epoch": 0.19, "learning_rate": 9.814867090096447e-05, "loss": 3.0924, "step": 2361 }, { "epoch": 0.19, "learning_rate": 9.814788677173999e-05, "loss": 3.1709, "step": 2362 }, { "epoch": 0.19, "learning_rate": 9.81471026425155e-05, "loss": 3.1869, "step": 2363 }, { "epoch": 0.19, "learning_rate": 9.8146318513291e-05, "loss": 2.9533, "step": 2364 }, { "epoch": 0.19, "learning_rate": 9.81455343840665e-05, "loss": 3.1146, "step": 2365 }, { "epoch": 0.19, "learning_rate": 9.8144750254842e-05, "loss": 3.4951, "step": 2366 }, { "epoch": 0.19, "learning_rate": 9.814396612561751e-05, "loss": 3.23, "step": 2367 }, { "epoch": 0.19, "learning_rate": 9.814318199639301e-05, "loss": 3.0757, "step": 2368 }, { "epoch": 0.19, "learning_rate": 9.814239786716852e-05, "loss": 2.9853, "step": 2369 }, { "epoch": 0.19, "learning_rate": 9.814161373794402e-05, "loss": 3.0153, "step": 2370 }, { "epoch": 0.19, "learning_rate": 9.814082960871952e-05, "loss": 2.953, "step": 2371 }, { "epoch": 0.19, "learning_rate": 9.814004547949503e-05, "loss": 3.2864, "step": 2372 }, { "epoch": 0.19, "learning_rate": 9.813926135027053e-05, "loss": 3.0528, "step": 2373 }, { "epoch": 0.19, "learning_rate": 9.813847722104603e-05, "loss": 2.9572, "step": 2374 }, { "epoch": 0.19, "learning_rate": 9.813769309182154e-05, "loss": 3.1396, "step": 2375 }, { "epoch": 0.19, "learning_rate": 9.813690896259704e-05, "loss": 3.0074, "step": 2376 }, { "epoch": 0.19, "learning_rate": 9.813612483337254e-05, "loss": 3.1266, "step": 2377 }, { "epoch": 0.19, "learning_rate": 9.813534070414805e-05, "loss": 2.9561, "step": 2378 }, { "epoch": 0.19, "learning_rate": 9.813455657492355e-05, "loss": 3.1993, "step": 2379 }, { "epoch": 0.19, "learning_rate": 9.813377244569906e-05, "loss": 2.8439, "step": 2380 }, { "epoch": 0.19, "learning_rate": 9.813298831647456e-05, "loss": 3.0241, "step": 2381 }, { "epoch": 0.19, "learning_rate": 9.813220418725006e-05, "loss": 3.3528, "step": 2382 }, { "epoch": 0.19, "learning_rate": 9.813142005802557e-05, "loss": 3.1499, "step": 2383 }, { "epoch": 0.19, "learning_rate": 9.813063592880107e-05, "loss": 3.396, "step": 2384 }, { "epoch": 0.19, "learning_rate": 9.812985179957657e-05, "loss": 3.2382, "step": 2385 }, { "epoch": 0.19, "learning_rate": 9.812906767035208e-05, "loss": 2.8594, "step": 2386 }, { "epoch": 0.19, "learning_rate": 9.812828354112758e-05, "loss": 2.9376, "step": 2387 }, { "epoch": 0.19, "learning_rate": 9.812749941190309e-05, "loss": 3.0579, "step": 2388 }, { "epoch": 0.19, "learning_rate": 9.812671528267859e-05, "loss": 3.1359, "step": 2389 }, { "epoch": 0.19, "learning_rate": 9.812593115345409e-05, "loss": 3.047, "step": 2390 }, { "epoch": 0.19, "learning_rate": 9.81251470242296e-05, "loss": 3.0049, "step": 2391 }, { "epoch": 0.19, "learning_rate": 9.81243628950051e-05, "loss": 3.1895, "step": 2392 }, { "epoch": 0.19, "learning_rate": 9.812357876578062e-05, "loss": 2.9995, "step": 2393 }, { "epoch": 0.19, "learning_rate": 9.812279463655612e-05, "loss": 3.1565, "step": 2394 }, { "epoch": 0.19, "learning_rate": 9.812201050733161e-05, "loss": 2.941, "step": 2395 }, { "epoch": 0.19, "learning_rate": 9.812122637810711e-05, "loss": 3.0957, "step": 2396 }, { "epoch": 0.19, "learning_rate": 9.812044224888262e-05, "loss": 3.1058, "step": 2397 }, { "epoch": 0.19, "learning_rate": 9.811965811965812e-05, "loss": 3.1728, "step": 2398 }, { "epoch": 0.19, "learning_rate": 9.811887399043363e-05, "loss": 3.333, "step": 2399 }, { "epoch": 0.19, "learning_rate": 9.811808986120913e-05, "loss": 2.9845, "step": 2400 }, { "epoch": 0.19, "learning_rate": 9.811730573198463e-05, "loss": 3.2833, "step": 2401 }, { "epoch": 0.19, "learning_rate": 9.811652160276014e-05, "loss": 2.959, "step": 2402 }, { "epoch": 0.19, "learning_rate": 9.811573747353564e-05, "loss": 3.1673, "step": 2403 }, { "epoch": 0.19, "learning_rate": 9.811495334431114e-05, "loss": 3.079, "step": 2404 }, { "epoch": 0.19, "learning_rate": 9.811416921508665e-05, "loss": 3.2803, "step": 2405 }, { "epoch": 0.19, "learning_rate": 9.811338508586215e-05, "loss": 3.0782, "step": 2406 }, { "epoch": 0.19, "learning_rate": 9.811260095663766e-05, "loss": 3.2479, "step": 2407 }, { "epoch": 0.19, "learning_rate": 9.811181682741316e-05, "loss": 3.2001, "step": 2408 }, { "epoch": 0.19, "learning_rate": 9.811103269818866e-05, "loss": 2.9564, "step": 2409 }, { "epoch": 0.19, "learning_rate": 9.811024856896417e-05, "loss": 3.2408, "step": 2410 }, { "epoch": 0.19, "learning_rate": 9.810946443973967e-05, "loss": 3.0624, "step": 2411 }, { "epoch": 0.19, "learning_rate": 9.810868031051517e-05, "loss": 3.3029, "step": 2412 }, { "epoch": 0.19, "learning_rate": 9.810789618129068e-05, "loss": 3.0769, "step": 2413 }, { "epoch": 0.19, "learning_rate": 9.810711205206618e-05, "loss": 3.1378, "step": 2414 }, { "epoch": 0.19, "learning_rate": 9.810632792284168e-05, "loss": 3.0677, "step": 2415 }, { "epoch": 0.19, "learning_rate": 9.810554379361719e-05, "loss": 3.0142, "step": 2416 }, { "epoch": 0.19, "learning_rate": 9.810475966439269e-05, "loss": 3.1007, "step": 2417 }, { "epoch": 0.19, "learning_rate": 9.81039755351682e-05, "loss": 3.2971, "step": 2418 }, { "epoch": 0.19, "learning_rate": 9.810319140594371e-05, "loss": 3.012, "step": 2419 }, { "epoch": 0.19, "learning_rate": 9.810240727671922e-05, "loss": 3.0939, "step": 2420 }, { "epoch": 0.19, "learning_rate": 9.810162314749472e-05, "loss": 2.9181, "step": 2421 }, { "epoch": 0.19, "learning_rate": 9.810083901827022e-05, "loss": 3.3589, "step": 2422 }, { "epoch": 0.19, "learning_rate": 9.810005488904573e-05, "loss": 3.0079, "step": 2423 }, { "epoch": 0.19, "learning_rate": 9.809927075982123e-05, "loss": 3.0806, "step": 2424 }, { "epoch": 0.19, "learning_rate": 9.809848663059673e-05, "loss": 3.2975, "step": 2425 }, { "epoch": 0.19, "learning_rate": 9.809770250137224e-05, "loss": 3.2321, "step": 2426 }, { "epoch": 0.19, "learning_rate": 9.809691837214774e-05, "loss": 3.3448, "step": 2427 }, { "epoch": 0.19, "learning_rate": 9.809613424292325e-05, "loss": 2.859, "step": 2428 }, { "epoch": 0.19, "learning_rate": 9.809535011369875e-05, "loss": 3.1623, "step": 2429 }, { "epoch": 0.19, "learning_rate": 9.809456598447425e-05, "loss": 2.9823, "step": 2430 }, { "epoch": 0.19, "learning_rate": 9.809378185524974e-05, "loss": 3.247, "step": 2431 }, { "epoch": 0.19, "learning_rate": 9.809299772602525e-05, "loss": 2.9553, "step": 2432 }, { "epoch": 0.19, "learning_rate": 9.809221359680075e-05, "loss": 3.0825, "step": 2433 }, { "epoch": 0.19, "learning_rate": 9.809142946757625e-05, "loss": 3.5421, "step": 2434 }, { "epoch": 0.19, "learning_rate": 9.809064533835176e-05, "loss": 2.8492, "step": 2435 }, { "epoch": 0.19, "learning_rate": 9.808986120912726e-05, "loss": 2.9866, "step": 2436 }, { "epoch": 0.19, "learning_rate": 9.808907707990277e-05, "loss": 2.9698, "step": 2437 }, { "epoch": 0.19, "learning_rate": 9.808829295067827e-05, "loss": 2.9021, "step": 2438 }, { "epoch": 0.19, "learning_rate": 9.808750882145377e-05, "loss": 3.0794, "step": 2439 }, { "epoch": 0.19, "learning_rate": 9.808672469222928e-05, "loss": 3.1743, "step": 2440 }, { "epoch": 0.19, "learning_rate": 9.808594056300478e-05, "loss": 3.0525, "step": 2441 }, { "epoch": 0.19, "learning_rate": 9.808515643378028e-05, "loss": 2.983, "step": 2442 }, { "epoch": 0.19, "learning_rate": 9.808437230455579e-05, "loss": 3.3634, "step": 2443 }, { "epoch": 0.19, "learning_rate": 9.808358817533129e-05, "loss": 3.0537, "step": 2444 }, { "epoch": 0.19, "learning_rate": 9.808280404610681e-05, "loss": 3.0775, "step": 2445 }, { "epoch": 0.19, "learning_rate": 9.808201991688231e-05, "loss": 3.3916, "step": 2446 }, { "epoch": 0.19, "learning_rate": 9.808123578765782e-05, "loss": 3.0095, "step": 2447 }, { "epoch": 0.19, "learning_rate": 9.808045165843332e-05, "loss": 2.9759, "step": 2448 }, { "epoch": 0.19, "learning_rate": 9.807966752920882e-05, "loss": 2.9242, "step": 2449 }, { "epoch": 0.19, "learning_rate": 9.807888339998433e-05, "loss": 3.2217, "step": 2450 }, { "epoch": 0.19, "learning_rate": 9.807809927075983e-05, "loss": 3.1535, "step": 2451 }, { "epoch": 0.19, "learning_rate": 9.807731514153533e-05, "loss": 3.0855, "step": 2452 }, { "epoch": 0.19, "learning_rate": 9.807653101231084e-05, "loss": 3.0176, "step": 2453 }, { "epoch": 0.19, "learning_rate": 9.807574688308634e-05, "loss": 3.0201, "step": 2454 }, { "epoch": 0.19, "learning_rate": 9.807496275386185e-05, "loss": 3.169, "step": 2455 }, { "epoch": 0.19, "learning_rate": 9.807417862463735e-05, "loss": 3.2493, "step": 2456 }, { "epoch": 0.19, "learning_rate": 9.807339449541285e-05, "loss": 3.2654, "step": 2457 }, { "epoch": 0.19, "learning_rate": 9.807261036618836e-05, "loss": 2.978, "step": 2458 }, { "epoch": 0.19, "learning_rate": 9.807182623696386e-05, "loss": 3.0836, "step": 2459 }, { "epoch": 0.19, "learning_rate": 9.807104210773936e-05, "loss": 2.9953, "step": 2460 }, { "epoch": 0.19, "learning_rate": 9.807025797851487e-05, "loss": 2.9524, "step": 2461 }, { "epoch": 0.19, "learning_rate": 9.806947384929037e-05, "loss": 2.8817, "step": 2462 }, { "epoch": 0.19, "learning_rate": 9.806868972006587e-05, "loss": 3.5706, "step": 2463 }, { "epoch": 0.19, "learning_rate": 9.806790559084138e-05, "loss": 3.2594, "step": 2464 }, { "epoch": 0.19, "learning_rate": 9.806712146161688e-05, "loss": 3.2162, "step": 2465 }, { "epoch": 0.19, "learning_rate": 9.806633733239239e-05, "loss": 2.8328, "step": 2466 }, { "epoch": 0.19, "learning_rate": 9.806555320316788e-05, "loss": 3.1214, "step": 2467 }, { "epoch": 0.19, "learning_rate": 9.806476907394338e-05, "loss": 3.2266, "step": 2468 }, { "epoch": 0.19, "learning_rate": 9.806398494471888e-05, "loss": 3.2083, "step": 2469 }, { "epoch": 0.19, "learning_rate": 9.80632008154944e-05, "loss": 3.0905, "step": 2470 }, { "epoch": 0.19, "learning_rate": 9.80624166862699e-05, "loss": 3.2371, "step": 2471 }, { "epoch": 0.19, "learning_rate": 9.806163255704541e-05, "loss": 3.1612, "step": 2472 }, { "epoch": 0.19, "learning_rate": 9.806084842782091e-05, "loss": 3.1173, "step": 2473 }, { "epoch": 0.19, "learning_rate": 9.806006429859642e-05, "loss": 2.9307, "step": 2474 }, { "epoch": 0.19, "learning_rate": 9.805928016937192e-05, "loss": 3.2978, "step": 2475 }, { "epoch": 0.19, "learning_rate": 9.805849604014742e-05, "loss": 3.219, "step": 2476 }, { "epoch": 0.19, "learning_rate": 9.805771191092293e-05, "loss": 3.0872, "step": 2477 }, { "epoch": 0.19, "learning_rate": 9.805692778169843e-05, "loss": 3.0056, "step": 2478 }, { "epoch": 0.19, "learning_rate": 9.805614365247393e-05, "loss": 3.2059, "step": 2479 }, { "epoch": 0.19, "learning_rate": 9.805535952324944e-05, "loss": 3.0094, "step": 2480 }, { "epoch": 0.19, "learning_rate": 9.805457539402494e-05, "loss": 3.2685, "step": 2481 }, { "epoch": 0.19, "learning_rate": 9.805379126480044e-05, "loss": 3.0171, "step": 2482 }, { "epoch": 0.19, "learning_rate": 9.805300713557595e-05, "loss": 3.2134, "step": 2483 }, { "epoch": 0.19, "learning_rate": 9.805222300635145e-05, "loss": 3.0536, "step": 2484 }, { "epoch": 0.19, "learning_rate": 9.805143887712696e-05, "loss": 3.0313, "step": 2485 }, { "epoch": 0.19, "learning_rate": 9.805065474790246e-05, "loss": 3.1048, "step": 2486 }, { "epoch": 0.2, "learning_rate": 9.804987061867796e-05, "loss": 3.63, "step": 2487 }, { "epoch": 0.2, "learning_rate": 9.804908648945347e-05, "loss": 3.1464, "step": 2488 }, { "epoch": 0.2, "learning_rate": 9.804830236022897e-05, "loss": 3.0968, "step": 2489 }, { "epoch": 0.2, "learning_rate": 9.804751823100447e-05, "loss": 3.0572, "step": 2490 }, { "epoch": 0.2, "learning_rate": 9.804673410177998e-05, "loss": 3.0649, "step": 2491 }, { "epoch": 0.2, "learning_rate": 9.804594997255548e-05, "loss": 3.0351, "step": 2492 }, { "epoch": 0.2, "learning_rate": 9.804516584333099e-05, "loss": 2.8998, "step": 2493 }, { "epoch": 0.2, "learning_rate": 9.804438171410649e-05, "loss": 3.2569, "step": 2494 }, { "epoch": 0.2, "learning_rate": 9.804359758488199e-05, "loss": 3.1363, "step": 2495 }, { "epoch": 0.2, "learning_rate": 9.80428134556575e-05, "loss": 3.4042, "step": 2496 }, { "epoch": 0.2, "learning_rate": 9.8042029326433e-05, "loss": 3.2061, "step": 2497 }, { "epoch": 0.2, "learning_rate": 9.80412451972085e-05, "loss": 3.3805, "step": 2498 }, { "epoch": 0.2, "learning_rate": 9.804046106798401e-05, "loss": 2.7502, "step": 2499 }, { "epoch": 0.2, "learning_rate": 9.803967693875951e-05, "loss": 3.009, "step": 2500 }, { "epoch": 0.2, "learning_rate": 9.803889280953503e-05, "loss": 3.189, "step": 2501 }, { "epoch": 0.2, "learning_rate": 9.803810868031053e-05, "loss": 3.1422, "step": 2502 }, { "epoch": 0.2, "learning_rate": 9.803732455108602e-05, "loss": 3.3541, "step": 2503 }, { "epoch": 0.2, "learning_rate": 9.803654042186153e-05, "loss": 2.989, "step": 2504 }, { "epoch": 0.2, "learning_rate": 9.803575629263703e-05, "loss": 3.0145, "step": 2505 }, { "epoch": 0.2, "learning_rate": 9.803497216341253e-05, "loss": 2.9857, "step": 2506 }, { "epoch": 0.2, "learning_rate": 9.803418803418804e-05, "loss": 2.9217, "step": 2507 }, { "epoch": 0.2, "learning_rate": 9.803340390496354e-05, "loss": 3.2447, "step": 2508 }, { "epoch": 0.2, "learning_rate": 9.803261977573904e-05, "loss": 3.2841, "step": 2509 }, { "epoch": 0.2, "learning_rate": 9.803183564651455e-05, "loss": 2.975, "step": 2510 }, { "epoch": 0.2, "learning_rate": 9.803105151729005e-05, "loss": 2.9722, "step": 2511 }, { "epoch": 0.2, "learning_rate": 9.803026738806555e-05, "loss": 3.0569, "step": 2512 }, { "epoch": 0.2, "learning_rate": 9.802948325884106e-05, "loss": 2.8543, "step": 2513 }, { "epoch": 0.2, "learning_rate": 9.802869912961656e-05, "loss": 2.8965, "step": 2514 }, { "epoch": 0.2, "learning_rate": 9.802791500039207e-05, "loss": 3.0713, "step": 2515 }, { "epoch": 0.2, "learning_rate": 9.802713087116757e-05, "loss": 3.2658, "step": 2516 }, { "epoch": 0.2, "learning_rate": 9.802634674194307e-05, "loss": 2.9961, "step": 2517 }, { "epoch": 0.2, "learning_rate": 9.802556261271858e-05, "loss": 2.9568, "step": 2518 }, { "epoch": 0.2, "learning_rate": 9.802477848349408e-05, "loss": 3.2161, "step": 2519 }, { "epoch": 0.2, "learning_rate": 9.802399435426958e-05, "loss": 3.1522, "step": 2520 }, { "epoch": 0.2, "learning_rate": 9.802321022504509e-05, "loss": 2.9335, "step": 2521 }, { "epoch": 0.2, "learning_rate": 9.802242609582059e-05, "loss": 3.1782, "step": 2522 }, { "epoch": 0.2, "learning_rate": 9.80216419665961e-05, "loss": 3.4779, "step": 2523 }, { "epoch": 0.2, "learning_rate": 9.80208578373716e-05, "loss": 2.8224, "step": 2524 }, { "epoch": 0.2, "learning_rate": 9.80200737081471e-05, "loss": 3.3293, "step": 2525 }, { "epoch": 0.2, "learning_rate": 9.80192895789226e-05, "loss": 3.1786, "step": 2526 }, { "epoch": 0.2, "learning_rate": 9.801850544969812e-05, "loss": 3.2758, "step": 2527 }, { "epoch": 0.2, "learning_rate": 9.801772132047363e-05, "loss": 3.1579, "step": 2528 }, { "epoch": 0.2, "learning_rate": 9.801693719124913e-05, "loss": 3.0901, "step": 2529 }, { "epoch": 0.2, "learning_rate": 9.801615306202463e-05, "loss": 3.2468, "step": 2530 }, { "epoch": 0.2, "learning_rate": 9.801536893280014e-05, "loss": 2.9713, "step": 2531 }, { "epoch": 0.2, "learning_rate": 9.801458480357564e-05, "loss": 3.0941, "step": 2532 }, { "epoch": 0.2, "learning_rate": 9.801380067435115e-05, "loss": 3.1184, "step": 2533 }, { "epoch": 0.2, "learning_rate": 9.801301654512665e-05, "loss": 3.17, "step": 2534 }, { "epoch": 0.2, "learning_rate": 9.801223241590215e-05, "loss": 3.1609, "step": 2535 }, { "epoch": 0.2, "learning_rate": 9.801144828667766e-05, "loss": 3.0288, "step": 2536 }, { "epoch": 0.2, "learning_rate": 9.801066415745316e-05, "loss": 3.115, "step": 2537 }, { "epoch": 0.2, "learning_rate": 9.800988002822866e-05, "loss": 3.1485, "step": 2538 }, { "epoch": 0.2, "learning_rate": 9.800909589900415e-05, "loss": 3.0019, "step": 2539 }, { "epoch": 0.2, "learning_rate": 9.800831176977966e-05, "loss": 3.0255, "step": 2540 }, { "epoch": 0.2, "learning_rate": 9.800752764055516e-05, "loss": 3.1607, "step": 2541 }, { "epoch": 0.2, "learning_rate": 9.800674351133067e-05, "loss": 3.1483, "step": 2542 }, { "epoch": 0.2, "learning_rate": 9.800595938210617e-05, "loss": 3.0072, "step": 2543 }, { "epoch": 0.2, "learning_rate": 9.800517525288167e-05, "loss": 3.1881, "step": 2544 }, { "epoch": 0.2, "learning_rate": 9.800439112365718e-05, "loss": 3.237, "step": 2545 }, { "epoch": 0.2, "learning_rate": 9.800360699443268e-05, "loss": 3.1207, "step": 2546 }, { "epoch": 0.2, "learning_rate": 9.800282286520818e-05, "loss": 3.2337, "step": 2547 }, { "epoch": 0.2, "learning_rate": 9.800203873598369e-05, "loss": 2.9525, "step": 2548 }, { "epoch": 0.2, "learning_rate": 9.800125460675919e-05, "loss": 2.8195, "step": 2549 }, { "epoch": 0.2, "learning_rate": 9.80004704775347e-05, "loss": 3.1201, "step": 2550 }, { "epoch": 0.2, "learning_rate": 9.79996863483102e-05, "loss": 3.0888, "step": 2551 }, { "epoch": 0.2, "learning_rate": 9.79989022190857e-05, "loss": 2.9223, "step": 2552 }, { "epoch": 0.2, "learning_rate": 9.799811808986122e-05, "loss": 3.0781, "step": 2553 }, { "epoch": 0.2, "learning_rate": 9.799733396063672e-05, "loss": 3.2304, "step": 2554 }, { "epoch": 0.2, "learning_rate": 9.799654983141223e-05, "loss": 3.1958, "step": 2555 }, { "epoch": 0.2, "learning_rate": 9.799576570218773e-05, "loss": 3.2312, "step": 2556 }, { "epoch": 0.2, "learning_rate": 9.799498157296323e-05, "loss": 3.0509, "step": 2557 }, { "epoch": 0.2, "learning_rate": 9.799419744373874e-05, "loss": 3.32, "step": 2558 }, { "epoch": 0.2, "learning_rate": 9.799341331451424e-05, "loss": 3.1892, "step": 2559 }, { "epoch": 0.2, "learning_rate": 9.799262918528974e-05, "loss": 3.0124, "step": 2560 }, { "epoch": 0.2, "learning_rate": 9.799184505606525e-05, "loss": 3.2977, "step": 2561 }, { "epoch": 0.2, "learning_rate": 9.799106092684075e-05, "loss": 3.1706, "step": 2562 }, { "epoch": 0.2, "learning_rate": 9.799027679761626e-05, "loss": 3.3214, "step": 2563 }, { "epoch": 0.2, "learning_rate": 9.798949266839176e-05, "loss": 3.2169, "step": 2564 }, { "epoch": 0.2, "learning_rate": 9.798870853916726e-05, "loss": 2.9699, "step": 2565 }, { "epoch": 0.2, "learning_rate": 9.798792440994277e-05, "loss": 3.1436, "step": 2566 }, { "epoch": 0.2, "learning_rate": 9.798714028071827e-05, "loss": 3.2345, "step": 2567 }, { "epoch": 0.2, "learning_rate": 9.798635615149377e-05, "loss": 3.0342, "step": 2568 }, { "epoch": 0.2, "learning_rate": 9.798557202226928e-05, "loss": 3.0686, "step": 2569 }, { "epoch": 0.2, "learning_rate": 9.798478789304478e-05, "loss": 3.0414, "step": 2570 }, { "epoch": 0.2, "learning_rate": 9.798400376382029e-05, "loss": 3.0694, "step": 2571 }, { "epoch": 0.2, "learning_rate": 9.798321963459579e-05, "loss": 3.1466, "step": 2572 }, { "epoch": 0.2, "learning_rate": 9.798243550537129e-05, "loss": 3.01, "step": 2573 }, { "epoch": 0.2, "learning_rate": 9.79816513761468e-05, "loss": 3.1144, "step": 2574 }, { "epoch": 0.2, "learning_rate": 9.798086724692229e-05, "loss": 2.9624, "step": 2575 }, { "epoch": 0.2, "learning_rate": 9.798008311769779e-05, "loss": 2.8154, "step": 2576 }, { "epoch": 0.2, "learning_rate": 9.79792989884733e-05, "loss": 3.0705, "step": 2577 }, { "epoch": 0.2, "learning_rate": 9.797851485924881e-05, "loss": 3.1903, "step": 2578 }, { "epoch": 0.2, "learning_rate": 9.797773073002431e-05, "loss": 3.3632, "step": 2579 }, { "epoch": 0.2, "learning_rate": 9.797694660079982e-05, "loss": 3.1217, "step": 2580 }, { "epoch": 0.2, "learning_rate": 9.797616247157532e-05, "loss": 3.1393, "step": 2581 }, { "epoch": 0.2, "learning_rate": 9.797537834235083e-05, "loss": 2.9621, "step": 2582 }, { "epoch": 0.2, "learning_rate": 9.797459421312633e-05, "loss": 3.1161, "step": 2583 }, { "epoch": 0.2, "learning_rate": 9.797381008390183e-05, "loss": 3.1868, "step": 2584 }, { "epoch": 0.2, "learning_rate": 9.797302595467734e-05, "loss": 3.0278, "step": 2585 }, { "epoch": 0.2, "learning_rate": 9.797224182545284e-05, "loss": 3.0236, "step": 2586 }, { "epoch": 0.2, "learning_rate": 9.797145769622834e-05, "loss": 2.9021, "step": 2587 }, { "epoch": 0.2, "learning_rate": 9.797067356700385e-05, "loss": 3.042, "step": 2588 }, { "epoch": 0.2, "learning_rate": 9.796988943777935e-05, "loss": 3.2184, "step": 2589 }, { "epoch": 0.2, "learning_rate": 9.796910530855486e-05, "loss": 2.9992, "step": 2590 }, { "epoch": 0.2, "learning_rate": 9.796832117933036e-05, "loss": 3.281, "step": 2591 }, { "epoch": 0.2, "learning_rate": 9.796753705010586e-05, "loss": 2.9095, "step": 2592 }, { "epoch": 0.2, "learning_rate": 9.796675292088137e-05, "loss": 3.3042, "step": 2593 }, { "epoch": 0.2, "learning_rate": 9.796596879165687e-05, "loss": 3.0472, "step": 2594 }, { "epoch": 0.2, "learning_rate": 9.796518466243237e-05, "loss": 3.0229, "step": 2595 }, { "epoch": 0.2, "learning_rate": 9.796440053320788e-05, "loss": 3.0937, "step": 2596 }, { "epoch": 0.2, "learning_rate": 9.796361640398338e-05, "loss": 3.0228, "step": 2597 }, { "epoch": 0.2, "learning_rate": 9.796283227475888e-05, "loss": 3.148, "step": 2598 }, { "epoch": 0.2, "learning_rate": 9.796204814553439e-05, "loss": 3.0878, "step": 2599 }, { "epoch": 0.2, "learning_rate": 9.796126401630989e-05, "loss": 3.01, "step": 2600 }, { "epoch": 0.2, "learning_rate": 9.79604798870854e-05, "loss": 3.1354, "step": 2601 }, { "epoch": 0.2, "learning_rate": 9.79596957578609e-05, "loss": 3.0885, "step": 2602 }, { "epoch": 0.2, "learning_rate": 9.79589116286364e-05, "loss": 3.2305, "step": 2603 }, { "epoch": 0.2, "learning_rate": 9.795812749941191e-05, "loss": 3.372, "step": 2604 }, { "epoch": 0.2, "learning_rate": 9.795734337018741e-05, "loss": 2.786, "step": 2605 }, { "epoch": 0.2, "learning_rate": 9.795655924096291e-05, "loss": 3.528, "step": 2606 }, { "epoch": 0.2, "learning_rate": 9.795577511173842e-05, "loss": 3.0968, "step": 2607 }, { "epoch": 0.2, "learning_rate": 9.795499098251392e-05, "loss": 3.07, "step": 2608 }, { "epoch": 0.2, "learning_rate": 9.795420685328943e-05, "loss": 2.7288, "step": 2609 }, { "epoch": 0.2, "learning_rate": 9.795342272406494e-05, "loss": 3.0088, "step": 2610 }, { "epoch": 0.2, "learning_rate": 9.795263859484043e-05, "loss": 3.2046, "step": 2611 }, { "epoch": 0.2, "learning_rate": 9.795185446561594e-05, "loss": 2.9197, "step": 2612 }, { "epoch": 0.2, "learning_rate": 9.795107033639144e-05, "loss": 3.0686, "step": 2613 }, { "epoch": 0.2, "learning_rate": 9.795028620716694e-05, "loss": 3.4896, "step": 2614 }, { "epoch": 0.21, "learning_rate": 9.794950207794245e-05, "loss": 3.1332, "step": 2615 }, { "epoch": 0.21, "learning_rate": 9.794871794871795e-05, "loss": 3.0575, "step": 2616 }, { "epoch": 0.21, "learning_rate": 9.794793381949345e-05, "loss": 3.284, "step": 2617 }, { "epoch": 0.21, "learning_rate": 9.794714969026896e-05, "loss": 3.2094, "step": 2618 }, { "epoch": 0.21, "learning_rate": 9.794636556104446e-05, "loss": 3.1662, "step": 2619 }, { "epoch": 0.21, "learning_rate": 9.794558143181997e-05, "loss": 3.031, "step": 2620 }, { "epoch": 0.21, "learning_rate": 9.794479730259547e-05, "loss": 3.0346, "step": 2621 }, { "epoch": 0.21, "learning_rate": 9.794401317337097e-05, "loss": 3.0467, "step": 2622 }, { "epoch": 0.21, "learning_rate": 9.794322904414648e-05, "loss": 2.9646, "step": 2623 }, { "epoch": 0.21, "learning_rate": 9.794244491492198e-05, "loss": 3.0462, "step": 2624 }, { "epoch": 0.21, "learning_rate": 9.794166078569748e-05, "loss": 3.2873, "step": 2625 }, { "epoch": 0.21, "learning_rate": 9.794087665647299e-05, "loss": 3.3376, "step": 2626 }, { "epoch": 0.21, "learning_rate": 9.794009252724849e-05, "loss": 3.0384, "step": 2627 }, { "epoch": 0.21, "learning_rate": 9.7939308398024e-05, "loss": 3.0736, "step": 2628 }, { "epoch": 0.21, "learning_rate": 9.79385242687995e-05, "loss": 3.0182, "step": 2629 }, { "epoch": 0.21, "learning_rate": 9.7937740139575e-05, "loss": 3.1195, "step": 2630 }, { "epoch": 0.21, "learning_rate": 9.79369560103505e-05, "loss": 2.9416, "step": 2631 }, { "epoch": 0.21, "learning_rate": 9.793617188112601e-05, "loss": 3.0569, "step": 2632 }, { "epoch": 0.21, "learning_rate": 9.793538775190151e-05, "loss": 2.9746, "step": 2633 }, { "epoch": 0.21, "learning_rate": 9.793460362267702e-05, "loss": 3.1726, "step": 2634 }, { "epoch": 0.21, "learning_rate": 9.793381949345253e-05, "loss": 3.0947, "step": 2635 }, { "epoch": 0.21, "learning_rate": 9.793303536422804e-05, "loss": 3.1439, "step": 2636 }, { "epoch": 0.21, "learning_rate": 9.793225123500354e-05, "loss": 2.9139, "step": 2637 }, { "epoch": 0.21, "learning_rate": 9.793146710577905e-05, "loss": 3.3328, "step": 2638 }, { "epoch": 0.21, "learning_rate": 9.793068297655455e-05, "loss": 3.0296, "step": 2639 }, { "epoch": 0.21, "learning_rate": 9.792989884733005e-05, "loss": 3.0614, "step": 2640 }, { "epoch": 0.21, "learning_rate": 9.792911471810556e-05, "loss": 3.0197, "step": 2641 }, { "epoch": 0.21, "learning_rate": 9.792833058888106e-05, "loss": 3.5951, "step": 2642 }, { "epoch": 0.21, "learning_rate": 9.792754645965656e-05, "loss": 2.7807, "step": 2643 }, { "epoch": 0.21, "learning_rate": 9.792676233043207e-05, "loss": 2.975, "step": 2644 }, { "epoch": 0.21, "learning_rate": 9.792597820120757e-05, "loss": 2.9845, "step": 2645 }, { "epoch": 0.21, "learning_rate": 9.792519407198307e-05, "loss": 3.0258, "step": 2646 }, { "epoch": 0.21, "learning_rate": 9.792440994275856e-05, "loss": 3.291, "step": 2647 }, { "epoch": 0.21, "learning_rate": 9.792362581353407e-05, "loss": 3.2085, "step": 2648 }, { "epoch": 0.21, "learning_rate": 9.792284168430957e-05, "loss": 3.4183, "step": 2649 }, { "epoch": 0.21, "learning_rate": 9.792205755508508e-05, "loss": 3.0864, "step": 2650 }, { "epoch": 0.21, "learning_rate": 9.792127342586058e-05, "loss": 2.9018, "step": 2651 }, { "epoch": 0.21, "learning_rate": 9.792048929663608e-05, "loss": 2.9581, "step": 2652 }, { "epoch": 0.21, "learning_rate": 9.791970516741159e-05, "loss": 2.9817, "step": 2653 }, { "epoch": 0.21, "learning_rate": 9.791892103818709e-05, "loss": 3.1559, "step": 2654 }, { "epoch": 0.21, "learning_rate": 9.79181369089626e-05, "loss": 2.8911, "step": 2655 }, { "epoch": 0.21, "learning_rate": 9.79173527797381e-05, "loss": 2.9521, "step": 2656 }, { "epoch": 0.21, "learning_rate": 9.79165686505136e-05, "loss": 2.9377, "step": 2657 }, { "epoch": 0.21, "learning_rate": 9.79157845212891e-05, "loss": 3.0394, "step": 2658 }, { "epoch": 0.21, "learning_rate": 9.791500039206461e-05, "loss": 2.8624, "step": 2659 }, { "epoch": 0.21, "learning_rate": 9.791421626284011e-05, "loss": 3.3915, "step": 2660 }, { "epoch": 0.21, "learning_rate": 9.791343213361563e-05, "loss": 3.2079, "step": 2661 }, { "epoch": 0.21, "learning_rate": 9.791264800439113e-05, "loss": 3.0897, "step": 2662 }, { "epoch": 0.21, "learning_rate": 9.791186387516664e-05, "loss": 3.1514, "step": 2663 }, { "epoch": 0.21, "learning_rate": 9.791107974594214e-05, "loss": 3.1488, "step": 2664 }, { "epoch": 0.21, "learning_rate": 9.791029561671764e-05, "loss": 3.0116, "step": 2665 }, { "epoch": 0.21, "learning_rate": 9.790951148749315e-05, "loss": 3.1326, "step": 2666 }, { "epoch": 0.21, "learning_rate": 9.790872735826865e-05, "loss": 3.1762, "step": 2667 }, { "epoch": 0.21, "learning_rate": 9.790794322904416e-05, "loss": 2.8535, "step": 2668 }, { "epoch": 0.21, "learning_rate": 9.790715909981966e-05, "loss": 3.3251, "step": 2669 }, { "epoch": 0.21, "learning_rate": 9.790637497059516e-05, "loss": 3.1281, "step": 2670 }, { "epoch": 0.21, "learning_rate": 9.790559084137067e-05, "loss": 2.9569, "step": 2671 }, { "epoch": 0.21, "learning_rate": 9.790480671214617e-05, "loss": 3.3593, "step": 2672 }, { "epoch": 0.21, "learning_rate": 9.790402258292167e-05, "loss": 3.3203, "step": 2673 }, { "epoch": 0.21, "learning_rate": 9.790323845369718e-05, "loss": 3.2145, "step": 2674 }, { "epoch": 0.21, "learning_rate": 9.790245432447268e-05, "loss": 3.0886, "step": 2675 }, { "epoch": 0.21, "learning_rate": 9.790167019524819e-05, "loss": 3.3121, "step": 2676 }, { "epoch": 0.21, "learning_rate": 9.790088606602369e-05, "loss": 2.9799, "step": 2677 }, { "epoch": 0.21, "learning_rate": 9.790010193679919e-05, "loss": 3.2186, "step": 2678 }, { "epoch": 0.21, "learning_rate": 9.78993178075747e-05, "loss": 3.1622, "step": 2679 }, { "epoch": 0.21, "learning_rate": 9.78985336783502e-05, "loss": 2.9339, "step": 2680 }, { "epoch": 0.21, "learning_rate": 9.78977495491257e-05, "loss": 3.1037, "step": 2681 }, { "epoch": 0.21, "learning_rate": 9.789696541990121e-05, "loss": 2.831, "step": 2682 }, { "epoch": 0.21, "learning_rate": 9.78961812906767e-05, "loss": 3.1092, "step": 2683 }, { "epoch": 0.21, "learning_rate": 9.78953971614522e-05, "loss": 3.1899, "step": 2684 }, { "epoch": 0.21, "learning_rate": 9.78946130322277e-05, "loss": 3.1223, "step": 2685 }, { "epoch": 0.21, "learning_rate": 9.789382890300322e-05, "loss": 3.3285, "step": 2686 }, { "epoch": 0.21, "learning_rate": 9.789304477377873e-05, "loss": 3.0859, "step": 2687 }, { "epoch": 0.21, "learning_rate": 9.789226064455423e-05, "loss": 2.9284, "step": 2688 }, { "epoch": 0.21, "learning_rate": 9.789147651532973e-05, "loss": 3.2802, "step": 2689 }, { "epoch": 0.21, "learning_rate": 9.789069238610524e-05, "loss": 3.0767, "step": 2690 }, { "epoch": 0.21, "learning_rate": 9.788990825688074e-05, "loss": 3.1495, "step": 2691 }, { "epoch": 0.21, "learning_rate": 9.788912412765624e-05, "loss": 2.9845, "step": 2692 }, { "epoch": 0.21, "learning_rate": 9.788833999843175e-05, "loss": 3.238, "step": 2693 }, { "epoch": 0.21, "learning_rate": 9.788755586920725e-05, "loss": 3.1551, "step": 2694 }, { "epoch": 0.21, "learning_rate": 9.788677173998276e-05, "loss": 3.3109, "step": 2695 }, { "epoch": 0.21, "learning_rate": 9.788598761075826e-05, "loss": 3.3058, "step": 2696 }, { "epoch": 0.21, "learning_rate": 9.788520348153376e-05, "loss": 2.997, "step": 2697 }, { "epoch": 0.21, "learning_rate": 9.788441935230927e-05, "loss": 2.8952, "step": 2698 }, { "epoch": 0.21, "learning_rate": 9.788363522308477e-05, "loss": 3.2435, "step": 2699 }, { "epoch": 0.21, "learning_rate": 9.788285109386027e-05, "loss": 3.1304, "step": 2700 }, { "epoch": 0.21, "learning_rate": 9.788206696463578e-05, "loss": 3.3418, "step": 2701 }, { "epoch": 0.21, "learning_rate": 9.788128283541128e-05, "loss": 2.9984, "step": 2702 }, { "epoch": 0.21, "learning_rate": 9.788049870618678e-05, "loss": 2.7814, "step": 2703 }, { "epoch": 0.21, "learning_rate": 9.787971457696229e-05, "loss": 2.916, "step": 2704 }, { "epoch": 0.21, "learning_rate": 9.787893044773779e-05, "loss": 2.9904, "step": 2705 }, { "epoch": 0.21, "learning_rate": 9.78781463185133e-05, "loss": 2.9386, "step": 2706 }, { "epoch": 0.21, "learning_rate": 9.78773621892888e-05, "loss": 3.0616, "step": 2707 }, { "epoch": 0.21, "learning_rate": 9.78765780600643e-05, "loss": 3.0627, "step": 2708 }, { "epoch": 0.21, "learning_rate": 9.78757939308398e-05, "loss": 3.0327, "step": 2709 }, { "epoch": 0.21, "learning_rate": 9.787500980161531e-05, "loss": 2.8651, "step": 2710 }, { "epoch": 0.21, "learning_rate": 9.787422567239081e-05, "loss": 3.07, "step": 2711 }, { "epoch": 0.21, "learning_rate": 9.787344154316632e-05, "loss": 3.4313, "step": 2712 }, { "epoch": 0.21, "learning_rate": 9.787265741394182e-05, "loss": 2.9433, "step": 2713 }, { "epoch": 0.21, "learning_rate": 9.787187328471732e-05, "loss": 3.2949, "step": 2714 }, { "epoch": 0.21, "learning_rate": 9.787108915549283e-05, "loss": 2.9367, "step": 2715 }, { "epoch": 0.21, "learning_rate": 9.787030502626833e-05, "loss": 3.1843, "step": 2716 }, { "epoch": 0.21, "learning_rate": 9.786952089704384e-05, "loss": 3.0057, "step": 2717 }, { "epoch": 0.21, "learning_rate": 9.786873676781935e-05, "loss": 3.0989, "step": 2718 }, { "epoch": 0.21, "learning_rate": 9.786795263859484e-05, "loss": 3.17, "step": 2719 }, { "epoch": 0.21, "learning_rate": 9.786716850937035e-05, "loss": 3.0007, "step": 2720 }, { "epoch": 0.21, "learning_rate": 9.786638438014585e-05, "loss": 3.1749, "step": 2721 }, { "epoch": 0.21, "learning_rate": 9.786560025092135e-05, "loss": 3.3354, "step": 2722 }, { "epoch": 0.21, "learning_rate": 9.786481612169686e-05, "loss": 2.9759, "step": 2723 }, { "epoch": 0.21, "learning_rate": 9.786403199247236e-05, "loss": 3.0421, "step": 2724 }, { "epoch": 0.21, "learning_rate": 9.786324786324787e-05, "loss": 3.123, "step": 2725 }, { "epoch": 0.21, "learning_rate": 9.786246373402337e-05, "loss": 3.5905, "step": 2726 }, { "epoch": 0.21, "learning_rate": 9.786167960479887e-05, "loss": 3.0286, "step": 2727 }, { "epoch": 0.21, "learning_rate": 9.786089547557438e-05, "loss": 3.2645, "step": 2728 }, { "epoch": 0.21, "learning_rate": 9.786011134634988e-05, "loss": 3.1113, "step": 2729 }, { "epoch": 0.21, "learning_rate": 9.785932721712538e-05, "loss": 2.8709, "step": 2730 }, { "epoch": 0.21, "learning_rate": 9.785854308790089e-05, "loss": 2.9259, "step": 2731 }, { "epoch": 0.21, "learning_rate": 9.785775895867639e-05, "loss": 3.4158, "step": 2732 }, { "epoch": 0.21, "learning_rate": 9.78569748294519e-05, "loss": 3.0721, "step": 2733 }, { "epoch": 0.21, "learning_rate": 9.78561907002274e-05, "loss": 3.2022, "step": 2734 }, { "epoch": 0.21, "learning_rate": 9.78554065710029e-05, "loss": 3.2397, "step": 2735 }, { "epoch": 0.21, "learning_rate": 9.78546224417784e-05, "loss": 3.203, "step": 2736 }, { "epoch": 0.21, "learning_rate": 9.785383831255391e-05, "loss": 2.9212, "step": 2737 }, { "epoch": 0.21, "learning_rate": 9.785305418332941e-05, "loss": 3.1472, "step": 2738 }, { "epoch": 0.21, "learning_rate": 9.785227005410492e-05, "loss": 3.2495, "step": 2739 }, { "epoch": 0.21, "learning_rate": 9.785148592488042e-05, "loss": 3.445, "step": 2740 }, { "epoch": 0.21, "learning_rate": 9.785070179565592e-05, "loss": 3.0285, "step": 2741 }, { "epoch": 0.22, "learning_rate": 9.784991766643143e-05, "loss": 3.0009, "step": 2742 }, { "epoch": 0.22, "learning_rate": 9.784913353720695e-05, "loss": 3.1024, "step": 2743 }, { "epoch": 0.22, "learning_rate": 9.784834940798245e-05, "loss": 3.1411, "step": 2744 }, { "epoch": 0.22, "learning_rate": 9.784756527875795e-05, "loss": 2.9934, "step": 2745 }, { "epoch": 0.22, "learning_rate": 9.784678114953346e-05, "loss": 3.1544, "step": 2746 }, { "epoch": 0.22, "learning_rate": 9.784599702030896e-05, "loss": 3.3984, "step": 2747 }, { "epoch": 0.22, "learning_rate": 9.784521289108446e-05, "loss": 3.1044, "step": 2748 }, { "epoch": 0.22, "learning_rate": 9.784442876185997e-05, "loss": 3.0093, "step": 2749 }, { "epoch": 0.22, "learning_rate": 9.784364463263547e-05, "loss": 3.0026, "step": 2750 }, { "epoch": 0.22, "learning_rate": 9.784286050341097e-05, "loss": 3.2542, "step": 2751 }, { "epoch": 0.22, "learning_rate": 9.784207637418648e-05, "loss": 3.0306, "step": 2752 }, { "epoch": 0.22, "learning_rate": 9.784129224496198e-05, "loss": 2.7968, "step": 2753 }, { "epoch": 0.22, "learning_rate": 9.784050811573749e-05, "loss": 3.1006, "step": 2754 }, { "epoch": 0.22, "learning_rate": 9.783972398651298e-05, "loss": 3.1303, "step": 2755 }, { "epoch": 0.22, "learning_rate": 9.783893985728848e-05, "loss": 3.0749, "step": 2756 }, { "epoch": 0.22, "learning_rate": 9.783815572806398e-05, "loss": 2.9466, "step": 2757 }, { "epoch": 0.22, "learning_rate": 9.783737159883949e-05, "loss": 3.0408, "step": 2758 }, { "epoch": 0.22, "learning_rate": 9.783658746961499e-05, "loss": 3.1587, "step": 2759 }, { "epoch": 0.22, "learning_rate": 9.78358033403905e-05, "loss": 3.2141, "step": 2760 }, { "epoch": 0.22, "learning_rate": 9.7835019211166e-05, "loss": 3.1215, "step": 2761 }, { "epoch": 0.22, "learning_rate": 9.78342350819415e-05, "loss": 3.2476, "step": 2762 }, { "epoch": 0.22, "learning_rate": 9.7833450952717e-05, "loss": 2.9687, "step": 2763 }, { "epoch": 0.22, "learning_rate": 9.783266682349251e-05, "loss": 2.9249, "step": 2764 }, { "epoch": 0.22, "learning_rate": 9.783188269426801e-05, "loss": 3.0923, "step": 2765 }, { "epoch": 0.22, "learning_rate": 9.783109856504352e-05, "loss": 2.9119, "step": 2766 }, { "epoch": 0.22, "learning_rate": 9.783031443581902e-05, "loss": 3.0103, "step": 2767 }, { "epoch": 0.22, "learning_rate": 9.782953030659452e-05, "loss": 3.2893, "step": 2768 }, { "epoch": 0.22, "learning_rate": 9.782874617737004e-05, "loss": 3.0645, "step": 2769 }, { "epoch": 0.22, "learning_rate": 9.782796204814554e-05, "loss": 3.067, "step": 2770 }, { "epoch": 0.22, "learning_rate": 9.782717791892105e-05, "loss": 2.9482, "step": 2771 }, { "epoch": 0.22, "learning_rate": 9.782639378969655e-05, "loss": 3.1536, "step": 2772 }, { "epoch": 0.22, "learning_rate": 9.782560966047206e-05, "loss": 3.0258, "step": 2773 }, { "epoch": 0.22, "learning_rate": 9.782482553124756e-05, "loss": 3.0854, "step": 2774 }, { "epoch": 0.22, "learning_rate": 9.782404140202306e-05, "loss": 3.0893, "step": 2775 }, { "epoch": 0.22, "learning_rate": 9.782325727279857e-05, "loss": 3.1122, "step": 2776 }, { "epoch": 0.22, "learning_rate": 9.782247314357407e-05, "loss": 3.0304, "step": 2777 }, { "epoch": 0.22, "learning_rate": 9.782168901434957e-05, "loss": 3.0582, "step": 2778 }, { "epoch": 0.22, "learning_rate": 9.782090488512508e-05, "loss": 3.1544, "step": 2779 }, { "epoch": 0.22, "learning_rate": 9.782012075590058e-05, "loss": 3.0873, "step": 2780 }, { "epoch": 0.22, "learning_rate": 9.781933662667608e-05, "loss": 3.1861, "step": 2781 }, { "epoch": 0.22, "learning_rate": 9.781855249745159e-05, "loss": 3.0822, "step": 2782 }, { "epoch": 0.22, "learning_rate": 9.781776836822709e-05, "loss": 3.1993, "step": 2783 }, { "epoch": 0.22, "learning_rate": 9.78169842390026e-05, "loss": 3.1669, "step": 2784 }, { "epoch": 0.22, "learning_rate": 9.78162001097781e-05, "loss": 3.3644, "step": 2785 }, { "epoch": 0.22, "learning_rate": 9.78154159805536e-05, "loss": 3.0982, "step": 2786 }, { "epoch": 0.22, "learning_rate": 9.781463185132911e-05, "loss": 3.0591, "step": 2787 }, { "epoch": 0.22, "learning_rate": 9.781384772210461e-05, "loss": 2.9359, "step": 2788 }, { "epoch": 0.22, "learning_rate": 9.781306359288011e-05, "loss": 2.8917, "step": 2789 }, { "epoch": 0.22, "learning_rate": 9.781227946365562e-05, "loss": 3.2505, "step": 2790 }, { "epoch": 0.22, "learning_rate": 9.781149533443111e-05, "loss": 3.1941, "step": 2791 }, { "epoch": 0.22, "learning_rate": 9.781071120520661e-05, "loss": 3.2418, "step": 2792 }, { "epoch": 0.22, "learning_rate": 9.780992707598212e-05, "loss": 3.2214, "step": 2793 }, { "epoch": 0.22, "learning_rate": 9.780914294675763e-05, "loss": 3.0543, "step": 2794 }, { "epoch": 0.22, "learning_rate": 9.780835881753314e-05, "loss": 2.8766, "step": 2795 }, { "epoch": 0.22, "learning_rate": 9.780757468830864e-05, "loss": 3.071, "step": 2796 }, { "epoch": 0.22, "learning_rate": 9.780679055908414e-05, "loss": 2.9372, "step": 2797 }, { "epoch": 0.22, "learning_rate": 9.780600642985965e-05, "loss": 3.132, "step": 2798 }, { "epoch": 0.22, "learning_rate": 9.780522230063515e-05, "loss": 3.1289, "step": 2799 }, { "epoch": 0.22, "learning_rate": 9.780443817141065e-05, "loss": 3.0896, "step": 2800 }, { "epoch": 0.22, "learning_rate": 9.780365404218616e-05, "loss": 3.0797, "step": 2801 }, { "epoch": 0.22, "learning_rate": 9.780286991296166e-05, "loss": 3.0125, "step": 2802 }, { "epoch": 0.22, "learning_rate": 9.780208578373717e-05, "loss": 3.0973, "step": 2803 }, { "epoch": 0.22, "learning_rate": 9.780130165451267e-05, "loss": 2.9509, "step": 2804 }, { "epoch": 0.22, "learning_rate": 9.780051752528817e-05, "loss": 3.1357, "step": 2805 }, { "epoch": 0.22, "learning_rate": 9.779973339606368e-05, "loss": 3.282, "step": 2806 }, { "epoch": 0.22, "learning_rate": 9.779894926683918e-05, "loss": 3.0587, "step": 2807 }, { "epoch": 0.22, "learning_rate": 9.779816513761468e-05, "loss": 3.2699, "step": 2808 }, { "epoch": 0.22, "learning_rate": 9.779738100839019e-05, "loss": 3.1986, "step": 2809 }, { "epoch": 0.22, "learning_rate": 9.779659687916569e-05, "loss": 3.2144, "step": 2810 }, { "epoch": 0.22, "learning_rate": 9.77958127499412e-05, "loss": 3.2032, "step": 2811 }, { "epoch": 0.22, "learning_rate": 9.77950286207167e-05, "loss": 3.5221, "step": 2812 }, { "epoch": 0.22, "learning_rate": 9.77942444914922e-05, "loss": 3.0578, "step": 2813 }, { "epoch": 0.22, "learning_rate": 9.77934603622677e-05, "loss": 3.0597, "step": 2814 }, { "epoch": 0.22, "learning_rate": 9.779267623304321e-05, "loss": 3.0382, "step": 2815 }, { "epoch": 0.22, "learning_rate": 9.779189210381871e-05, "loss": 3.0327, "step": 2816 }, { "epoch": 0.22, "learning_rate": 9.779110797459422e-05, "loss": 3.1778, "step": 2817 }, { "epoch": 0.22, "learning_rate": 9.779032384536972e-05, "loss": 3.0248, "step": 2818 }, { "epoch": 0.22, "learning_rate": 9.778953971614522e-05, "loss": 3.1724, "step": 2819 }, { "epoch": 0.22, "learning_rate": 9.778875558692073e-05, "loss": 3.1376, "step": 2820 }, { "epoch": 0.22, "learning_rate": 9.778797145769623e-05, "loss": 2.9112, "step": 2821 }, { "epoch": 0.22, "learning_rate": 9.778718732847174e-05, "loss": 3.1051, "step": 2822 }, { "epoch": 0.22, "learning_rate": 9.778640319924724e-05, "loss": 3.2248, "step": 2823 }, { "epoch": 0.22, "learning_rate": 9.778561907002274e-05, "loss": 3.2019, "step": 2824 }, { "epoch": 0.22, "learning_rate": 9.778483494079825e-05, "loss": 3.3325, "step": 2825 }, { "epoch": 0.22, "learning_rate": 9.778405081157376e-05, "loss": 3.1202, "step": 2826 }, { "epoch": 0.22, "learning_rate": 9.778326668234925e-05, "loss": 3.0347, "step": 2827 }, { "epoch": 0.22, "learning_rate": 9.778248255312476e-05, "loss": 3.1338, "step": 2828 }, { "epoch": 0.22, "learning_rate": 9.778169842390026e-05, "loss": 3.0322, "step": 2829 }, { "epoch": 0.22, "learning_rate": 9.778091429467577e-05, "loss": 3.0345, "step": 2830 }, { "epoch": 0.22, "learning_rate": 9.778013016545127e-05, "loss": 3.3084, "step": 2831 }, { "epoch": 0.22, "learning_rate": 9.777934603622677e-05, "loss": 3.1846, "step": 2832 }, { "epoch": 0.22, "learning_rate": 9.777856190700228e-05, "loss": 2.9826, "step": 2833 }, { "epoch": 0.22, "learning_rate": 9.777777777777778e-05, "loss": 3.1778, "step": 2834 }, { "epoch": 0.22, "learning_rate": 9.777699364855328e-05, "loss": 2.8239, "step": 2835 }, { "epoch": 0.22, "learning_rate": 9.777620951932879e-05, "loss": 3.2657, "step": 2836 }, { "epoch": 0.22, "learning_rate": 9.777542539010429e-05, "loss": 2.9577, "step": 2837 }, { "epoch": 0.22, "learning_rate": 9.77746412608798e-05, "loss": 3.183, "step": 2838 }, { "epoch": 0.22, "learning_rate": 9.77738571316553e-05, "loss": 2.9305, "step": 2839 }, { "epoch": 0.22, "learning_rate": 9.77730730024308e-05, "loss": 3.0038, "step": 2840 }, { "epoch": 0.22, "learning_rate": 9.77722888732063e-05, "loss": 3.0617, "step": 2841 }, { "epoch": 0.22, "learning_rate": 9.777150474398181e-05, "loss": 3.396, "step": 2842 }, { "epoch": 0.22, "learning_rate": 9.777072061475731e-05, "loss": 3.1112, "step": 2843 }, { "epoch": 0.22, "learning_rate": 9.776993648553282e-05, "loss": 3.3344, "step": 2844 }, { "epoch": 0.22, "learning_rate": 9.776915235630832e-05, "loss": 3.2225, "step": 2845 }, { "epoch": 0.22, "learning_rate": 9.776836822708382e-05, "loss": 3.0273, "step": 2846 }, { "epoch": 0.22, "learning_rate": 9.776758409785933e-05, "loss": 3.2641, "step": 2847 }, { "epoch": 0.22, "learning_rate": 9.776679996863483e-05, "loss": 2.8535, "step": 2848 }, { "epoch": 0.22, "learning_rate": 9.776601583941033e-05, "loss": 3.0266, "step": 2849 }, { "epoch": 0.22, "learning_rate": 9.776523171018584e-05, "loss": 2.9092, "step": 2850 }, { "epoch": 0.22, "learning_rate": 9.776444758096136e-05, "loss": 3.2877, "step": 2851 }, { "epoch": 0.22, "learning_rate": 9.776366345173686e-05, "loss": 3.2175, "step": 2852 }, { "epoch": 0.22, "learning_rate": 9.776287932251236e-05, "loss": 3.0864, "step": 2853 }, { "epoch": 0.22, "learning_rate": 9.776209519328787e-05, "loss": 3.0778, "step": 2854 }, { "epoch": 0.22, "learning_rate": 9.776131106406337e-05, "loss": 3.044, "step": 2855 }, { "epoch": 0.22, "learning_rate": 9.776052693483887e-05, "loss": 3.0164, "step": 2856 }, { "epoch": 0.22, "learning_rate": 9.775974280561438e-05, "loss": 2.9259, "step": 2857 }, { "epoch": 0.22, "learning_rate": 9.775895867638988e-05, "loss": 3.3453, "step": 2858 }, { "epoch": 0.22, "learning_rate": 9.775817454716539e-05, "loss": 3.0417, "step": 2859 }, { "epoch": 0.22, "learning_rate": 9.775739041794089e-05, "loss": 3.1863, "step": 2860 }, { "epoch": 0.22, "learning_rate": 9.775660628871639e-05, "loss": 2.7631, "step": 2861 }, { "epoch": 0.22, "learning_rate": 9.77558221594919e-05, "loss": 2.8799, "step": 2862 }, { "epoch": 0.22, "learning_rate": 9.775503803026739e-05, "loss": 3.0975, "step": 2863 }, { "epoch": 0.22, "learning_rate": 9.775425390104289e-05, "loss": 2.8891, "step": 2864 }, { "epoch": 0.22, "learning_rate": 9.77534697718184e-05, "loss": 3.2887, "step": 2865 }, { "epoch": 0.22, "learning_rate": 9.77526856425939e-05, "loss": 3.2533, "step": 2866 }, { "epoch": 0.22, "learning_rate": 9.77519015133694e-05, "loss": 3.2942, "step": 2867 }, { "epoch": 0.22, "learning_rate": 9.77511173841449e-05, "loss": 2.8555, "step": 2868 }, { "epoch": 0.22, "learning_rate": 9.775033325492041e-05, "loss": 3.0924, "step": 2869 }, { "epoch": 0.23, "learning_rate": 9.774954912569591e-05, "loss": 2.9244, "step": 2870 }, { "epoch": 0.23, "learning_rate": 9.774876499647142e-05, "loss": 3.241, "step": 2871 }, { "epoch": 0.23, "learning_rate": 9.774798086724692e-05, "loss": 2.997, "step": 2872 }, { "epoch": 0.23, "learning_rate": 9.774719673802242e-05, "loss": 3.0043, "step": 2873 }, { "epoch": 0.23, "learning_rate": 9.774641260879793e-05, "loss": 3.205, "step": 2874 }, { "epoch": 0.23, "learning_rate": 9.774562847957343e-05, "loss": 2.9134, "step": 2875 }, { "epoch": 0.23, "learning_rate": 9.774484435034893e-05, "loss": 2.9827, "step": 2876 }, { "epoch": 0.23, "learning_rate": 9.774406022112445e-05, "loss": 3.0045, "step": 2877 }, { "epoch": 0.23, "learning_rate": 9.774327609189996e-05, "loss": 3.2701, "step": 2878 }, { "epoch": 0.23, "learning_rate": 9.774249196267546e-05, "loss": 3.1616, "step": 2879 }, { "epoch": 0.23, "learning_rate": 9.774170783345096e-05, "loss": 3.1314, "step": 2880 }, { "epoch": 0.23, "learning_rate": 9.774092370422647e-05, "loss": 3.1417, "step": 2881 }, { "epoch": 0.23, "learning_rate": 9.774013957500197e-05, "loss": 3.1708, "step": 2882 }, { "epoch": 0.23, "learning_rate": 9.773935544577747e-05, "loss": 3.0569, "step": 2883 }, { "epoch": 0.23, "learning_rate": 9.773857131655298e-05, "loss": 3.1704, "step": 2884 }, { "epoch": 0.23, "learning_rate": 9.773778718732848e-05, "loss": 3.2441, "step": 2885 }, { "epoch": 0.23, "learning_rate": 9.773700305810398e-05, "loss": 3.3869, "step": 2886 }, { "epoch": 0.23, "learning_rate": 9.773621892887949e-05, "loss": 3.1614, "step": 2887 }, { "epoch": 0.23, "learning_rate": 9.773543479965499e-05, "loss": 2.9326, "step": 2888 }, { "epoch": 0.23, "learning_rate": 9.77346506704305e-05, "loss": 3.0657, "step": 2889 }, { "epoch": 0.23, "learning_rate": 9.7733866541206e-05, "loss": 3.0368, "step": 2890 }, { "epoch": 0.23, "learning_rate": 9.77330824119815e-05, "loss": 3.0234, "step": 2891 }, { "epoch": 0.23, "learning_rate": 9.7732298282757e-05, "loss": 2.9334, "step": 2892 }, { "epoch": 0.23, "learning_rate": 9.773151415353251e-05, "loss": 2.9255, "step": 2893 }, { "epoch": 0.23, "learning_rate": 9.773073002430801e-05, "loss": 3.0418, "step": 2894 }, { "epoch": 0.23, "learning_rate": 9.772994589508352e-05, "loss": 3.2256, "step": 2895 }, { "epoch": 0.23, "learning_rate": 9.772916176585902e-05, "loss": 3.1838, "step": 2896 }, { "epoch": 0.23, "learning_rate": 9.772837763663453e-05, "loss": 3.1068, "step": 2897 }, { "epoch": 0.23, "learning_rate": 9.772759350741003e-05, "loss": 3.0296, "step": 2898 }, { "epoch": 0.23, "learning_rate": 9.772680937818552e-05, "loss": 3.0959, "step": 2899 }, { "epoch": 0.23, "learning_rate": 9.772602524896102e-05, "loss": 3.1738, "step": 2900 }, { "epoch": 0.23, "learning_rate": 9.772524111973653e-05, "loss": 2.8848, "step": 2901 }, { "epoch": 0.23, "learning_rate": 9.772445699051204e-05, "loss": 3.1487, "step": 2902 }, { "epoch": 0.23, "learning_rate": 9.772367286128755e-05, "loss": 3.3004, "step": 2903 }, { "epoch": 0.23, "learning_rate": 9.772288873206305e-05, "loss": 2.9547, "step": 2904 }, { "epoch": 0.23, "learning_rate": 9.772210460283855e-05, "loss": 3.246, "step": 2905 }, { "epoch": 0.23, "learning_rate": 9.772132047361406e-05, "loss": 3.1089, "step": 2906 }, { "epoch": 0.23, "learning_rate": 9.772053634438956e-05, "loss": 3.1301, "step": 2907 }, { "epoch": 0.23, "learning_rate": 9.771975221516507e-05, "loss": 3.044, "step": 2908 }, { "epoch": 0.23, "learning_rate": 9.771896808594057e-05, "loss": 3.3673, "step": 2909 }, { "epoch": 0.23, "learning_rate": 9.771818395671607e-05, "loss": 2.975, "step": 2910 }, { "epoch": 0.23, "learning_rate": 9.771739982749158e-05, "loss": 2.9239, "step": 2911 }, { "epoch": 0.23, "learning_rate": 9.771661569826708e-05, "loss": 3.1914, "step": 2912 }, { "epoch": 0.23, "learning_rate": 9.771583156904258e-05, "loss": 3.3305, "step": 2913 }, { "epoch": 0.23, "learning_rate": 9.771504743981809e-05, "loss": 3.0924, "step": 2914 }, { "epoch": 0.23, "learning_rate": 9.771426331059359e-05, "loss": 3.2523, "step": 2915 }, { "epoch": 0.23, "learning_rate": 9.77134791813691e-05, "loss": 3.047, "step": 2916 }, { "epoch": 0.23, "learning_rate": 9.77126950521446e-05, "loss": 3.0201, "step": 2917 }, { "epoch": 0.23, "learning_rate": 9.77119109229201e-05, "loss": 2.9884, "step": 2918 }, { "epoch": 0.23, "learning_rate": 9.77111267936956e-05, "loss": 2.9163, "step": 2919 }, { "epoch": 0.23, "learning_rate": 9.771034266447111e-05, "loss": 3.0924, "step": 2920 }, { "epoch": 0.23, "learning_rate": 9.770955853524661e-05, "loss": 3.0845, "step": 2921 }, { "epoch": 0.23, "learning_rate": 9.770877440602212e-05, "loss": 3.1507, "step": 2922 }, { "epoch": 0.23, "learning_rate": 9.770799027679762e-05, "loss": 2.957, "step": 2923 }, { "epoch": 0.23, "learning_rate": 9.770720614757312e-05, "loss": 3.2055, "step": 2924 }, { "epoch": 0.23, "learning_rate": 9.770642201834863e-05, "loss": 3.2149, "step": 2925 }, { "epoch": 0.23, "learning_rate": 9.770563788912413e-05, "loss": 3.1363, "step": 2926 }, { "epoch": 0.23, "learning_rate": 9.770485375989964e-05, "loss": 3.0726, "step": 2927 }, { "epoch": 0.23, "learning_rate": 9.770406963067514e-05, "loss": 3.5107, "step": 2928 }, { "epoch": 0.23, "learning_rate": 9.770328550145064e-05, "loss": 3.196, "step": 2929 }, { "epoch": 0.23, "learning_rate": 9.770250137222615e-05, "loss": 2.9364, "step": 2930 }, { "epoch": 0.23, "learning_rate": 9.770171724300165e-05, "loss": 3.0854, "step": 2931 }, { "epoch": 0.23, "learning_rate": 9.770093311377715e-05, "loss": 3.0603, "step": 2932 }, { "epoch": 0.23, "learning_rate": 9.770014898455266e-05, "loss": 3.1006, "step": 2933 }, { "epoch": 0.23, "learning_rate": 9.769936485532817e-05, "loss": 3.0755, "step": 2934 }, { "epoch": 0.23, "learning_rate": 9.769858072610366e-05, "loss": 3.2424, "step": 2935 }, { "epoch": 0.23, "learning_rate": 9.769779659687917e-05, "loss": 2.8051, "step": 2936 }, { "epoch": 0.23, "learning_rate": 9.769701246765467e-05, "loss": 3.0393, "step": 2937 }, { "epoch": 0.23, "learning_rate": 9.769622833843018e-05, "loss": 3.2846, "step": 2938 }, { "epoch": 0.23, "learning_rate": 9.769544420920568e-05, "loss": 3.1236, "step": 2939 }, { "epoch": 0.23, "learning_rate": 9.769466007998118e-05, "loss": 3.2403, "step": 2940 }, { "epoch": 0.23, "learning_rate": 9.769387595075669e-05, "loss": 2.9898, "step": 2941 }, { "epoch": 0.23, "learning_rate": 9.769309182153219e-05, "loss": 2.8476, "step": 2942 }, { "epoch": 0.23, "learning_rate": 9.76923076923077e-05, "loss": 3.0822, "step": 2943 }, { "epoch": 0.23, "learning_rate": 9.76915235630832e-05, "loss": 3.1251, "step": 2944 }, { "epoch": 0.23, "learning_rate": 9.76907394338587e-05, "loss": 2.9367, "step": 2945 }, { "epoch": 0.23, "learning_rate": 9.76899553046342e-05, "loss": 3.0625, "step": 2946 }, { "epoch": 0.23, "learning_rate": 9.768917117540971e-05, "loss": 3.126, "step": 2947 }, { "epoch": 0.23, "learning_rate": 9.768838704618521e-05, "loss": 3.0276, "step": 2948 }, { "epoch": 0.23, "learning_rate": 9.768760291696072e-05, "loss": 3.0785, "step": 2949 }, { "epoch": 0.23, "learning_rate": 9.768681878773622e-05, "loss": 3.0184, "step": 2950 }, { "epoch": 0.23, "learning_rate": 9.768603465851172e-05, "loss": 2.8738, "step": 2951 }, { "epoch": 0.23, "learning_rate": 9.768525052928723e-05, "loss": 3.1118, "step": 2952 }, { "epoch": 0.23, "learning_rate": 9.768446640006273e-05, "loss": 3.2241, "step": 2953 }, { "epoch": 0.23, "learning_rate": 9.768368227083823e-05, "loss": 3.0163, "step": 2954 }, { "epoch": 0.23, "learning_rate": 9.768289814161374e-05, "loss": 3.1195, "step": 2955 }, { "epoch": 0.23, "learning_rate": 9.768211401238924e-05, "loss": 3.1102, "step": 2956 }, { "epoch": 0.23, "learning_rate": 9.768132988316475e-05, "loss": 2.9229, "step": 2957 }, { "epoch": 0.23, "learning_rate": 9.768054575394025e-05, "loss": 3.4397, "step": 2958 }, { "epoch": 0.23, "learning_rate": 9.767976162471577e-05, "loss": 3.0937, "step": 2959 }, { "epoch": 0.23, "learning_rate": 9.767897749549127e-05, "loss": 3.1359, "step": 2960 }, { "epoch": 0.23, "learning_rate": 9.767819336626677e-05, "loss": 3.181, "step": 2961 }, { "epoch": 0.23, "learning_rate": 9.767740923704228e-05, "loss": 3.0535, "step": 2962 }, { "epoch": 0.23, "learning_rate": 9.767662510781778e-05, "loss": 3.2358, "step": 2963 }, { "epoch": 0.23, "learning_rate": 9.767584097859328e-05, "loss": 2.9228, "step": 2964 }, { "epoch": 0.23, "learning_rate": 9.767505684936879e-05, "loss": 3.1526, "step": 2965 }, { "epoch": 0.23, "learning_rate": 9.767427272014429e-05, "loss": 2.8127, "step": 2966 }, { "epoch": 0.23, "learning_rate": 9.76734885909198e-05, "loss": 3.147, "step": 2967 }, { "epoch": 0.23, "learning_rate": 9.76727044616953e-05, "loss": 3.0786, "step": 2968 }, { "epoch": 0.23, "learning_rate": 9.76719203324708e-05, "loss": 3.0973, "step": 2969 }, { "epoch": 0.23, "learning_rate": 9.767113620324631e-05, "loss": 3.2798, "step": 2970 }, { "epoch": 0.23, "learning_rate": 9.76703520740218e-05, "loss": 3.1659, "step": 2971 }, { "epoch": 0.23, "learning_rate": 9.76695679447973e-05, "loss": 3.1545, "step": 2972 }, { "epoch": 0.23, "learning_rate": 9.76687838155728e-05, "loss": 3.1572, "step": 2973 }, { "epoch": 0.23, "learning_rate": 9.766799968634831e-05, "loss": 2.9521, "step": 2974 }, { "epoch": 0.23, "learning_rate": 9.766721555712381e-05, "loss": 2.9313, "step": 2975 }, { "epoch": 0.23, "learning_rate": 9.766643142789932e-05, "loss": 3.1598, "step": 2976 }, { "epoch": 0.23, "learning_rate": 9.766564729867482e-05, "loss": 3.1254, "step": 2977 }, { "epoch": 0.23, "learning_rate": 9.766486316945032e-05, "loss": 3.179, "step": 2978 }, { "epoch": 0.23, "learning_rate": 9.766407904022583e-05, "loss": 3.3412, "step": 2979 }, { "epoch": 0.23, "learning_rate": 9.766329491100133e-05, "loss": 3.0116, "step": 2980 }, { "epoch": 0.23, "learning_rate": 9.766251078177683e-05, "loss": 2.9625, "step": 2981 }, { "epoch": 0.23, "learning_rate": 9.766172665255234e-05, "loss": 3.0876, "step": 2982 }, { "epoch": 0.23, "learning_rate": 9.766094252332784e-05, "loss": 2.9555, "step": 2983 }, { "epoch": 0.23, "learning_rate": 9.766015839410334e-05, "loss": 3.1755, "step": 2984 }, { "epoch": 0.23, "learning_rate": 9.765937426487886e-05, "loss": 3.0041, "step": 2985 }, { "epoch": 0.23, "learning_rate": 9.765859013565437e-05, "loss": 3.0916, "step": 2986 }, { "epoch": 0.23, "learning_rate": 9.765780600642987e-05, "loss": 2.9697, "step": 2987 }, { "epoch": 0.23, "learning_rate": 9.765702187720537e-05, "loss": 3.0599, "step": 2988 }, { "epoch": 0.23, "learning_rate": 9.765623774798088e-05, "loss": 3.4115, "step": 2989 }, { "epoch": 0.23, "learning_rate": 9.765545361875638e-05, "loss": 3.2693, "step": 2990 }, { "epoch": 0.23, "learning_rate": 9.765466948953188e-05, "loss": 3.1627, "step": 2991 }, { "epoch": 0.23, "learning_rate": 9.765388536030739e-05, "loss": 2.9566, "step": 2992 }, { "epoch": 0.23, "learning_rate": 9.765310123108289e-05, "loss": 2.9388, "step": 2993 }, { "epoch": 0.23, "learning_rate": 9.76523171018584e-05, "loss": 3.1419, "step": 2994 }, { "epoch": 0.23, "learning_rate": 9.76515329726339e-05, "loss": 3.0379, "step": 2995 }, { "epoch": 0.23, "learning_rate": 9.76507488434094e-05, "loss": 3.0196, "step": 2996 }, { "epoch": 0.24, "learning_rate": 9.76499647141849e-05, "loss": 3.3589, "step": 2997 }, { "epoch": 0.24, "learning_rate": 9.764918058496041e-05, "loss": 2.8666, "step": 2998 }, { "epoch": 0.24, "learning_rate": 9.764839645573591e-05, "loss": 2.9707, "step": 2999 }, { "epoch": 0.24, "learning_rate": 9.764761232651142e-05, "loss": 3.2642, "step": 3000 }, { "epoch": 0.24, "eval_loss": 2.994882345199585, "eval_runtime": 218.0675, "eval_samples_per_second": 51.966, "eval_steps_per_second": 3.251, "step": 3000 }, { "epoch": 0.24, "learning_rate": 9.764682819728692e-05, "loss": 3.1871, "step": 3001 }, { "epoch": 0.24, "learning_rate": 9.764604406806242e-05, "loss": 3.0605, "step": 3002 }, { "epoch": 0.24, "learning_rate": 9.764525993883793e-05, "loss": 2.9863, "step": 3003 }, { "epoch": 0.24, "learning_rate": 9.764447580961343e-05, "loss": 2.911, "step": 3004 }, { "epoch": 0.24, "learning_rate": 9.764369168038894e-05, "loss": 3.019, "step": 3005 }, { "epoch": 0.24, "learning_rate": 9.764290755116444e-05, "loss": 2.8943, "step": 3006 }, { "epoch": 0.24, "learning_rate": 9.764212342193993e-05, "loss": 3.044, "step": 3007 }, { "epoch": 0.24, "learning_rate": 9.764133929271543e-05, "loss": 3.0295, "step": 3008 }, { "epoch": 0.24, "learning_rate": 9.764055516349094e-05, "loss": 3.1791, "step": 3009 }, { "epoch": 0.24, "learning_rate": 9.763977103426644e-05, "loss": 3.2548, "step": 3010 }, { "epoch": 0.24, "learning_rate": 9.763898690504196e-05, "loss": 3.288, "step": 3011 }, { "epoch": 0.24, "learning_rate": 9.763820277581746e-05, "loss": 3.1007, "step": 3012 }, { "epoch": 0.24, "learning_rate": 9.763741864659297e-05, "loss": 2.9996, "step": 3013 }, { "epoch": 0.24, "learning_rate": 9.763663451736847e-05, "loss": 3.0791, "step": 3014 }, { "epoch": 0.24, "learning_rate": 9.763585038814397e-05, "loss": 3.4243, "step": 3015 }, { "epoch": 0.24, "learning_rate": 9.763506625891948e-05, "loss": 3.2501, "step": 3016 }, { "epoch": 0.24, "learning_rate": 9.763428212969498e-05, "loss": 3.1661, "step": 3017 }, { "epoch": 0.24, "learning_rate": 9.763349800047048e-05, "loss": 3.2846, "step": 3018 }, { "epoch": 0.24, "learning_rate": 9.763271387124599e-05, "loss": 3.1065, "step": 3019 }, { "epoch": 0.24, "learning_rate": 9.763192974202149e-05, "loss": 3.1199, "step": 3020 }, { "epoch": 0.24, "learning_rate": 9.7631145612797e-05, "loss": 3.0715, "step": 3021 }, { "epoch": 0.24, "learning_rate": 9.76303614835725e-05, "loss": 3.0766, "step": 3022 }, { "epoch": 0.24, "learning_rate": 9.7629577354348e-05, "loss": 2.9833, "step": 3023 }, { "epoch": 0.24, "learning_rate": 9.76287932251235e-05, "loss": 2.9394, "step": 3024 }, { "epoch": 0.24, "learning_rate": 9.762800909589901e-05, "loss": 2.8803, "step": 3025 }, { "epoch": 0.24, "learning_rate": 9.762722496667451e-05, "loss": 2.9943, "step": 3026 }, { "epoch": 0.24, "learning_rate": 9.762644083745002e-05, "loss": 3.2743, "step": 3027 }, { "epoch": 0.24, "learning_rate": 9.762565670822552e-05, "loss": 3.1194, "step": 3028 }, { "epoch": 0.24, "learning_rate": 9.762487257900102e-05, "loss": 2.957, "step": 3029 }, { "epoch": 0.24, "learning_rate": 9.762408844977653e-05, "loss": 3.0137, "step": 3030 }, { "epoch": 0.24, "learning_rate": 9.762330432055203e-05, "loss": 3.2359, "step": 3031 }, { "epoch": 0.24, "learning_rate": 9.762252019132754e-05, "loss": 3.1102, "step": 3032 }, { "epoch": 0.24, "learning_rate": 9.762173606210304e-05, "loss": 3.0073, "step": 3033 }, { "epoch": 0.24, "learning_rate": 9.762095193287854e-05, "loss": 3.2329, "step": 3034 }, { "epoch": 0.24, "learning_rate": 9.762016780365405e-05, "loss": 3.111, "step": 3035 }, { "epoch": 0.24, "learning_rate": 9.761938367442955e-05, "loss": 3.0926, "step": 3036 }, { "epoch": 0.24, "learning_rate": 9.761859954520505e-05, "loss": 3.0307, "step": 3037 }, { "epoch": 0.24, "learning_rate": 9.761781541598056e-05, "loss": 3.3259, "step": 3038 }, { "epoch": 0.24, "learning_rate": 9.761703128675606e-05, "loss": 3.2134, "step": 3039 }, { "epoch": 0.24, "learning_rate": 9.761624715753156e-05, "loss": 3.0981, "step": 3040 }, { "epoch": 0.24, "learning_rate": 9.761546302830707e-05, "loss": 3.1429, "step": 3041 }, { "epoch": 0.24, "learning_rate": 9.761467889908259e-05, "loss": 3.1936, "step": 3042 }, { "epoch": 0.24, "learning_rate": 9.761389476985808e-05, "loss": 2.9811, "step": 3043 }, { "epoch": 0.24, "learning_rate": 9.761311064063358e-05, "loss": 3.1014, "step": 3044 }, { "epoch": 0.24, "learning_rate": 9.761232651140908e-05, "loss": 3.2384, "step": 3045 }, { "epoch": 0.24, "learning_rate": 9.761154238218459e-05, "loss": 3.0976, "step": 3046 }, { "epoch": 0.24, "learning_rate": 9.761075825296009e-05, "loss": 3.0065, "step": 3047 }, { "epoch": 0.24, "learning_rate": 9.76099741237356e-05, "loss": 3.0517, "step": 3048 }, { "epoch": 0.24, "learning_rate": 9.76091899945111e-05, "loss": 3.04, "step": 3049 }, { "epoch": 0.24, "learning_rate": 9.76084058652866e-05, "loss": 2.9475, "step": 3050 }, { "epoch": 0.24, "learning_rate": 9.76076217360621e-05, "loss": 3.2121, "step": 3051 }, { "epoch": 0.24, "learning_rate": 9.760683760683761e-05, "loss": 2.9873, "step": 3052 }, { "epoch": 0.24, "learning_rate": 9.760605347761311e-05, "loss": 3.1106, "step": 3053 }, { "epoch": 0.24, "learning_rate": 9.760526934838862e-05, "loss": 2.9669, "step": 3054 }, { "epoch": 0.24, "learning_rate": 9.760448521916412e-05, "loss": 3.0124, "step": 3055 }, { "epoch": 0.24, "learning_rate": 9.760370108993962e-05, "loss": 3.3808, "step": 3056 }, { "epoch": 0.24, "learning_rate": 9.760291696071513e-05, "loss": 3.0082, "step": 3057 }, { "epoch": 0.24, "learning_rate": 9.760213283149063e-05, "loss": 2.9825, "step": 3058 }, { "epoch": 0.24, "learning_rate": 9.760134870226613e-05, "loss": 3.0195, "step": 3059 }, { "epoch": 0.24, "learning_rate": 9.760056457304164e-05, "loss": 2.9639, "step": 3060 }, { "epoch": 0.24, "learning_rate": 9.759978044381714e-05, "loss": 3.0087, "step": 3061 }, { "epoch": 0.24, "learning_rate": 9.759899631459265e-05, "loss": 3.1056, "step": 3062 }, { "epoch": 0.24, "learning_rate": 9.759821218536815e-05, "loss": 3.0746, "step": 3063 }, { "epoch": 0.24, "learning_rate": 9.759742805614365e-05, "loss": 3.4503, "step": 3064 }, { "epoch": 0.24, "learning_rate": 9.759664392691916e-05, "loss": 3.1218, "step": 3065 }, { "epoch": 0.24, "learning_rate": 9.759585979769466e-05, "loss": 3.015, "step": 3066 }, { "epoch": 0.24, "learning_rate": 9.759507566847018e-05, "loss": 3.2321, "step": 3067 }, { "epoch": 0.24, "learning_rate": 9.759429153924568e-05, "loss": 2.9461, "step": 3068 }, { "epoch": 0.24, "learning_rate": 9.759350741002118e-05, "loss": 3.0608, "step": 3069 }, { "epoch": 0.24, "learning_rate": 9.759272328079669e-05, "loss": 2.9834, "step": 3070 }, { "epoch": 0.24, "learning_rate": 9.759193915157219e-05, "loss": 3.0373, "step": 3071 }, { "epoch": 0.24, "learning_rate": 9.75911550223477e-05, "loss": 3.1057, "step": 3072 }, { "epoch": 0.24, "learning_rate": 9.75903708931232e-05, "loss": 3.0628, "step": 3073 }, { "epoch": 0.24, "learning_rate": 9.75895867638987e-05, "loss": 3.1487, "step": 3074 }, { "epoch": 0.24, "learning_rate": 9.75888026346742e-05, "loss": 3.2012, "step": 3075 }, { "epoch": 0.24, "learning_rate": 9.758801850544971e-05, "loss": 3.1261, "step": 3076 }, { "epoch": 0.24, "learning_rate": 9.758723437622521e-05, "loss": 3.2431, "step": 3077 }, { "epoch": 0.24, "learning_rate": 9.758645024700072e-05, "loss": 3.3319, "step": 3078 }, { "epoch": 0.24, "learning_rate": 9.758566611777621e-05, "loss": 3.046, "step": 3079 }, { "epoch": 0.24, "learning_rate": 9.758488198855171e-05, "loss": 3.0546, "step": 3080 }, { "epoch": 0.24, "learning_rate": 9.758409785932722e-05, "loss": 3.0628, "step": 3081 }, { "epoch": 0.24, "learning_rate": 9.758331373010272e-05, "loss": 3.1124, "step": 3082 }, { "epoch": 0.24, "learning_rate": 9.758252960087822e-05, "loss": 3.2276, "step": 3083 }, { "epoch": 0.24, "learning_rate": 9.758174547165373e-05, "loss": 3.2106, "step": 3084 }, { "epoch": 0.24, "learning_rate": 9.758096134242923e-05, "loss": 3.2959, "step": 3085 }, { "epoch": 0.24, "learning_rate": 9.758017721320473e-05, "loss": 2.7672, "step": 3086 }, { "epoch": 0.24, "learning_rate": 9.757939308398024e-05, "loss": 2.9263, "step": 3087 }, { "epoch": 0.24, "learning_rate": 9.757860895475574e-05, "loss": 3.3471, "step": 3088 }, { "epoch": 0.24, "learning_rate": 9.757782482553124e-05, "loss": 3.1822, "step": 3089 }, { "epoch": 0.24, "learning_rate": 9.757704069630675e-05, "loss": 3.2534, "step": 3090 }, { "epoch": 0.24, "learning_rate": 9.757625656708225e-05, "loss": 3.0936, "step": 3091 }, { "epoch": 0.24, "learning_rate": 9.757547243785776e-05, "loss": 2.8921, "step": 3092 }, { "epoch": 0.24, "learning_rate": 9.757468830863327e-05, "loss": 3.0411, "step": 3093 }, { "epoch": 0.24, "learning_rate": 9.757390417940878e-05, "loss": 3.0024, "step": 3094 }, { "epoch": 0.24, "learning_rate": 9.757312005018428e-05, "loss": 2.9798, "step": 3095 }, { "epoch": 0.24, "learning_rate": 9.757233592095978e-05, "loss": 2.9622, "step": 3096 }, { "epoch": 0.24, "learning_rate": 9.757155179173529e-05, "loss": 3.3022, "step": 3097 }, { "epoch": 0.24, "learning_rate": 9.757076766251079e-05, "loss": 3.2888, "step": 3098 }, { "epoch": 0.24, "learning_rate": 9.75699835332863e-05, "loss": 2.8932, "step": 3099 }, { "epoch": 0.24, "learning_rate": 9.75691994040618e-05, "loss": 2.9228, "step": 3100 }, { "epoch": 0.24, "learning_rate": 9.75684152748373e-05, "loss": 3.0259, "step": 3101 }, { "epoch": 0.24, "learning_rate": 9.75676311456128e-05, "loss": 2.9217, "step": 3102 }, { "epoch": 0.24, "learning_rate": 9.756684701638831e-05, "loss": 3.3304, "step": 3103 }, { "epoch": 0.24, "learning_rate": 9.756606288716381e-05, "loss": 3.0316, "step": 3104 }, { "epoch": 0.24, "learning_rate": 9.756527875793932e-05, "loss": 3.0345, "step": 3105 }, { "epoch": 0.24, "learning_rate": 9.756449462871482e-05, "loss": 3.1857, "step": 3106 }, { "epoch": 0.24, "learning_rate": 9.756371049949032e-05, "loss": 3.1908, "step": 3107 }, { "epoch": 0.24, "learning_rate": 9.756292637026583e-05, "loss": 2.9767, "step": 3108 }, { "epoch": 0.24, "learning_rate": 9.756214224104133e-05, "loss": 3.0407, "step": 3109 }, { "epoch": 0.24, "learning_rate": 9.756135811181684e-05, "loss": 3.0178, "step": 3110 }, { "epoch": 0.24, "learning_rate": 9.756057398259234e-05, "loss": 3.0219, "step": 3111 }, { "epoch": 0.24, "learning_rate": 9.755978985336784e-05, "loss": 3.1144, "step": 3112 }, { "epoch": 0.24, "learning_rate": 9.755900572414335e-05, "loss": 3.024, "step": 3113 }, { "epoch": 0.24, "learning_rate": 9.755822159491885e-05, "loss": 3.2947, "step": 3114 }, { "epoch": 0.24, "learning_rate": 9.755743746569434e-05, "loss": 3.006, "step": 3115 }, { "epoch": 0.24, "learning_rate": 9.755665333646984e-05, "loss": 3.0201, "step": 3116 }, { "epoch": 0.24, "learning_rate": 9.755586920724535e-05, "loss": 3.1769, "step": 3117 }, { "epoch": 0.24, "learning_rate": 9.755508507802085e-05, "loss": 2.9196, "step": 3118 }, { "epoch": 0.24, "learning_rate": 9.755430094879637e-05, "loss": 3.256, "step": 3119 }, { "epoch": 0.24, "learning_rate": 9.755351681957187e-05, "loss": 2.9198, "step": 3120 }, { "epoch": 0.24, "learning_rate": 9.755273269034738e-05, "loss": 3.2519, "step": 3121 }, { "epoch": 0.24, "learning_rate": 9.755194856112288e-05, "loss": 3.2042, "step": 3122 }, { "epoch": 0.24, "learning_rate": 9.755116443189838e-05, "loss": 3.4028, "step": 3123 }, { "epoch": 0.24, "learning_rate": 9.755038030267389e-05, "loss": 3.104, "step": 3124 }, { "epoch": 0.25, "learning_rate": 9.754959617344939e-05, "loss": 3.3684, "step": 3125 }, { "epoch": 0.25, "learning_rate": 9.75488120442249e-05, "loss": 2.9244, "step": 3126 }, { "epoch": 0.25, "learning_rate": 9.75480279150004e-05, "loss": 3.3866, "step": 3127 }, { "epoch": 0.25, "learning_rate": 9.75472437857759e-05, "loss": 3.0843, "step": 3128 }, { "epoch": 0.25, "learning_rate": 9.75464596565514e-05, "loss": 2.9233, "step": 3129 }, { "epoch": 0.25, "learning_rate": 9.754567552732691e-05, "loss": 3.1496, "step": 3130 }, { "epoch": 0.25, "learning_rate": 9.754489139810241e-05, "loss": 3.031, "step": 3131 }, { "epoch": 0.25, "learning_rate": 9.754410726887792e-05, "loss": 3.0714, "step": 3132 }, { "epoch": 0.25, "learning_rate": 9.754332313965342e-05, "loss": 3.1672, "step": 3133 }, { "epoch": 0.25, "learning_rate": 9.754253901042892e-05, "loss": 2.9466, "step": 3134 }, { "epoch": 0.25, "learning_rate": 9.754175488120443e-05, "loss": 2.7957, "step": 3135 }, { "epoch": 0.25, "learning_rate": 9.754097075197993e-05, "loss": 3.0904, "step": 3136 }, { "epoch": 0.25, "learning_rate": 9.754018662275543e-05, "loss": 2.9162, "step": 3137 }, { "epoch": 0.25, "learning_rate": 9.753940249353094e-05, "loss": 3.0914, "step": 3138 }, { "epoch": 0.25, "learning_rate": 9.753861836430644e-05, "loss": 3.3235, "step": 3139 }, { "epoch": 0.25, "learning_rate": 9.753783423508195e-05, "loss": 3.2268, "step": 3140 }, { "epoch": 0.25, "learning_rate": 9.753705010585745e-05, "loss": 3.2046, "step": 3141 }, { "epoch": 0.25, "learning_rate": 9.753626597663295e-05, "loss": 3.3311, "step": 3142 }, { "epoch": 0.25, "learning_rate": 9.753548184740846e-05, "loss": 3.1378, "step": 3143 }, { "epoch": 0.25, "learning_rate": 9.753469771818396e-05, "loss": 2.959, "step": 3144 }, { "epoch": 0.25, "learning_rate": 9.753391358895946e-05, "loss": 3.0255, "step": 3145 }, { "epoch": 0.25, "learning_rate": 9.753312945973497e-05, "loss": 3.1401, "step": 3146 }, { "epoch": 0.25, "learning_rate": 9.753234533051047e-05, "loss": 2.9379, "step": 3147 }, { "epoch": 0.25, "learning_rate": 9.753156120128598e-05, "loss": 2.8525, "step": 3148 }, { "epoch": 0.25, "learning_rate": 9.753077707206148e-05, "loss": 3.118, "step": 3149 }, { "epoch": 0.25, "learning_rate": 9.7529992942837e-05, "loss": 2.7992, "step": 3150 }, { "epoch": 0.25, "learning_rate": 9.752920881361249e-05, "loss": 3.2782, "step": 3151 }, { "epoch": 0.25, "learning_rate": 9.752842468438799e-05, "loss": 2.9455, "step": 3152 }, { "epoch": 0.25, "learning_rate": 9.75276405551635e-05, "loss": 3.0195, "step": 3153 }, { "epoch": 0.25, "learning_rate": 9.7526856425939e-05, "loss": 2.9624, "step": 3154 }, { "epoch": 0.25, "learning_rate": 9.75260722967145e-05, "loss": 3.3109, "step": 3155 }, { "epoch": 0.25, "learning_rate": 9.752528816749e-05, "loss": 2.9779, "step": 3156 }, { "epoch": 0.25, "learning_rate": 9.752450403826551e-05, "loss": 3.1752, "step": 3157 }, { "epoch": 0.25, "learning_rate": 9.752371990904101e-05, "loss": 3.0691, "step": 3158 }, { "epoch": 0.25, "learning_rate": 9.752293577981652e-05, "loss": 3.005, "step": 3159 }, { "epoch": 0.25, "learning_rate": 9.752215165059202e-05, "loss": 3.2364, "step": 3160 }, { "epoch": 0.25, "learning_rate": 9.752136752136752e-05, "loss": 3.1318, "step": 3161 }, { "epoch": 0.25, "learning_rate": 9.752058339214303e-05, "loss": 3.1341, "step": 3162 }, { "epoch": 0.25, "learning_rate": 9.751979926291853e-05, "loss": 3.1904, "step": 3163 }, { "epoch": 0.25, "learning_rate": 9.751901513369403e-05, "loss": 2.9556, "step": 3164 }, { "epoch": 0.25, "learning_rate": 9.751823100446954e-05, "loss": 3.1869, "step": 3165 }, { "epoch": 0.25, "learning_rate": 9.751744687524504e-05, "loss": 2.9493, "step": 3166 }, { "epoch": 0.25, "learning_rate": 9.751666274602055e-05, "loss": 2.9889, "step": 3167 }, { "epoch": 0.25, "learning_rate": 9.751587861679605e-05, "loss": 3.4439, "step": 3168 }, { "epoch": 0.25, "learning_rate": 9.751509448757155e-05, "loss": 3.1166, "step": 3169 }, { "epoch": 0.25, "learning_rate": 9.751431035834706e-05, "loss": 2.9756, "step": 3170 }, { "epoch": 0.25, "learning_rate": 9.751352622912256e-05, "loss": 3.2556, "step": 3171 }, { "epoch": 0.25, "learning_rate": 9.751274209989806e-05, "loss": 3.5275, "step": 3172 }, { "epoch": 0.25, "learning_rate": 9.751195797067357e-05, "loss": 3.0883, "step": 3173 }, { "epoch": 0.25, "learning_rate": 9.751117384144907e-05, "loss": 3.1836, "step": 3174 }, { "epoch": 0.25, "learning_rate": 9.751038971222459e-05, "loss": 3.0742, "step": 3175 }, { "epoch": 0.25, "learning_rate": 9.750960558300009e-05, "loss": 3.0094, "step": 3176 }, { "epoch": 0.25, "learning_rate": 9.75088214537756e-05, "loss": 3.1858, "step": 3177 }, { "epoch": 0.25, "learning_rate": 9.75080373245511e-05, "loss": 2.9849, "step": 3178 }, { "epoch": 0.25, "learning_rate": 9.75072531953266e-05, "loss": 2.8766, "step": 3179 }, { "epoch": 0.25, "learning_rate": 9.75064690661021e-05, "loss": 3.3818, "step": 3180 }, { "epoch": 0.25, "learning_rate": 9.750568493687761e-05, "loss": 3.0081, "step": 3181 }, { "epoch": 0.25, "learning_rate": 9.750490080765311e-05, "loss": 3.1805, "step": 3182 }, { "epoch": 0.25, "learning_rate": 9.750411667842862e-05, "loss": 3.0883, "step": 3183 }, { "epoch": 0.25, "learning_rate": 9.750333254920412e-05, "loss": 3.0593, "step": 3184 }, { "epoch": 0.25, "learning_rate": 9.750254841997962e-05, "loss": 2.9215, "step": 3185 }, { "epoch": 0.25, "learning_rate": 9.750176429075513e-05, "loss": 3.0824, "step": 3186 }, { "epoch": 0.25, "learning_rate": 9.750098016153062e-05, "loss": 3.176, "step": 3187 }, { "epoch": 0.25, "learning_rate": 9.750019603230612e-05, "loss": 3.1879, "step": 3188 }, { "epoch": 0.25, "learning_rate": 9.749941190308163e-05, "loss": 3.0164, "step": 3189 }, { "epoch": 0.25, "learning_rate": 9.749862777385713e-05, "loss": 2.94, "step": 3190 }, { "epoch": 0.25, "learning_rate": 9.749784364463263e-05, "loss": 3.0602, "step": 3191 }, { "epoch": 0.25, "learning_rate": 9.749705951540814e-05, "loss": 2.9877, "step": 3192 }, { "epoch": 0.25, "learning_rate": 9.749627538618364e-05, "loss": 3.266, "step": 3193 }, { "epoch": 0.25, "learning_rate": 9.749549125695914e-05, "loss": 3.2068, "step": 3194 }, { "epoch": 0.25, "learning_rate": 9.749470712773465e-05, "loss": 3.0203, "step": 3195 }, { "epoch": 0.25, "learning_rate": 9.749392299851015e-05, "loss": 2.7346, "step": 3196 }, { "epoch": 0.25, "learning_rate": 9.749313886928566e-05, "loss": 2.918, "step": 3197 }, { "epoch": 0.25, "learning_rate": 9.749235474006116e-05, "loss": 3.0598, "step": 3198 }, { "epoch": 0.25, "learning_rate": 9.749157061083666e-05, "loss": 3.167, "step": 3199 }, { "epoch": 0.25, "learning_rate": 9.749078648161217e-05, "loss": 3.1555, "step": 3200 }, { "epoch": 0.25, "learning_rate": 9.749000235238768e-05, "loss": 3.1885, "step": 3201 }, { "epoch": 0.25, "learning_rate": 9.748921822316319e-05, "loss": 3.0878, "step": 3202 }, { "epoch": 0.25, "learning_rate": 9.748843409393869e-05, "loss": 3.1896, "step": 3203 }, { "epoch": 0.25, "learning_rate": 9.74876499647142e-05, "loss": 3.0554, "step": 3204 }, { "epoch": 0.25, "learning_rate": 9.74868658354897e-05, "loss": 2.9461, "step": 3205 }, { "epoch": 0.25, "learning_rate": 9.74860817062652e-05, "loss": 3.0515, "step": 3206 }, { "epoch": 0.25, "learning_rate": 9.74852975770407e-05, "loss": 3.0569, "step": 3207 }, { "epoch": 0.25, "learning_rate": 9.748451344781621e-05, "loss": 3.0008, "step": 3208 }, { "epoch": 0.25, "learning_rate": 9.748372931859171e-05, "loss": 3.2681, "step": 3209 }, { "epoch": 0.25, "learning_rate": 9.748294518936722e-05, "loss": 3.0869, "step": 3210 }, { "epoch": 0.25, "learning_rate": 9.748216106014272e-05, "loss": 3.0412, "step": 3211 }, { "epoch": 0.25, "learning_rate": 9.748137693091822e-05, "loss": 3.1393, "step": 3212 }, { "epoch": 0.25, "learning_rate": 9.748059280169373e-05, "loss": 3.012, "step": 3213 }, { "epoch": 0.25, "learning_rate": 9.747980867246923e-05, "loss": 3.1067, "step": 3214 }, { "epoch": 0.25, "learning_rate": 9.747902454324474e-05, "loss": 3.2719, "step": 3215 }, { "epoch": 0.25, "learning_rate": 9.747824041402024e-05, "loss": 3.1193, "step": 3216 }, { "epoch": 0.25, "learning_rate": 9.747745628479574e-05, "loss": 3.3726, "step": 3217 }, { "epoch": 0.25, "learning_rate": 9.747667215557125e-05, "loss": 3.3065, "step": 3218 }, { "epoch": 0.25, "learning_rate": 9.747588802634675e-05, "loss": 3.3481, "step": 3219 }, { "epoch": 0.25, "learning_rate": 9.747510389712225e-05, "loss": 3.2311, "step": 3220 }, { "epoch": 0.25, "learning_rate": 9.747431976789776e-05, "loss": 3.0, "step": 3221 }, { "epoch": 0.25, "learning_rate": 9.747353563867326e-05, "loss": 3.1112, "step": 3222 }, { "epoch": 0.25, "learning_rate": 9.747275150944875e-05, "loss": 2.7043, "step": 3223 }, { "epoch": 0.25, "learning_rate": 9.747196738022425e-05, "loss": 3.0156, "step": 3224 }, { "epoch": 0.25, "learning_rate": 9.747118325099976e-05, "loss": 2.9931, "step": 3225 }, { "epoch": 0.25, "learning_rate": 9.747039912177526e-05, "loss": 3.1303, "step": 3226 }, { "epoch": 0.25, "learning_rate": 9.746961499255078e-05, "loss": 3.1557, "step": 3227 }, { "epoch": 0.25, "learning_rate": 9.746883086332628e-05, "loss": 3.139, "step": 3228 }, { "epoch": 0.25, "learning_rate": 9.746804673410179e-05, "loss": 2.8636, "step": 3229 }, { "epoch": 0.25, "learning_rate": 9.746726260487729e-05, "loss": 2.9187, "step": 3230 }, { "epoch": 0.25, "learning_rate": 9.74664784756528e-05, "loss": 3.0412, "step": 3231 }, { "epoch": 0.25, "learning_rate": 9.74656943464283e-05, "loss": 3.3275, "step": 3232 }, { "epoch": 0.25, "learning_rate": 9.74649102172038e-05, "loss": 3.2734, "step": 3233 }, { "epoch": 0.25, "learning_rate": 9.74641260879793e-05, "loss": 3.1724, "step": 3234 }, { "epoch": 0.25, "learning_rate": 9.746334195875481e-05, "loss": 2.9547, "step": 3235 }, { "epoch": 0.25, "learning_rate": 9.746255782953031e-05, "loss": 3.2408, "step": 3236 }, { "epoch": 0.25, "learning_rate": 9.746177370030582e-05, "loss": 3.0159, "step": 3237 }, { "epoch": 0.25, "learning_rate": 9.746098957108132e-05, "loss": 3.1044, "step": 3238 }, { "epoch": 0.25, "learning_rate": 9.746020544185682e-05, "loss": 3.0995, "step": 3239 }, { "epoch": 0.25, "learning_rate": 9.745942131263233e-05, "loss": 3.1581, "step": 3240 }, { "epoch": 0.25, "learning_rate": 9.745863718340783e-05, "loss": 3.0703, "step": 3241 }, { "epoch": 0.25, "learning_rate": 9.745785305418333e-05, "loss": 3.0524, "step": 3242 }, { "epoch": 0.25, "learning_rate": 9.745706892495884e-05, "loss": 3.0868, "step": 3243 }, { "epoch": 0.25, "learning_rate": 9.745628479573434e-05, "loss": 3.0083, "step": 3244 }, { "epoch": 0.25, "learning_rate": 9.745550066650985e-05, "loss": 3.0418, "step": 3245 }, { "epoch": 0.25, "learning_rate": 9.745471653728535e-05, "loss": 2.9055, "step": 3246 }, { "epoch": 0.25, "learning_rate": 9.745393240806085e-05, "loss": 3.3526, "step": 3247 }, { "epoch": 0.25, "learning_rate": 9.745314827883636e-05, "loss": 3.318, "step": 3248 }, { "epoch": 0.25, "learning_rate": 9.745236414961186e-05, "loss": 2.8897, "step": 3249 }, { "epoch": 0.25, "learning_rate": 9.745158002038736e-05, "loss": 2.9627, "step": 3250 }, { "epoch": 0.25, "learning_rate": 9.745079589116287e-05, "loss": 3.0532, "step": 3251 }, { "epoch": 0.25, "learning_rate": 9.745001176193837e-05, "loss": 3.0251, "step": 3252 }, { "epoch": 0.26, "learning_rate": 9.744922763271387e-05, "loss": 2.8321, "step": 3253 }, { "epoch": 0.26, "learning_rate": 9.744844350348938e-05, "loss": 3.1479, "step": 3254 }, { "epoch": 0.26, "learning_rate": 9.744765937426488e-05, "loss": 2.9461, "step": 3255 }, { "epoch": 0.26, "learning_rate": 9.744687524504039e-05, "loss": 2.9035, "step": 3256 }, { "epoch": 0.26, "learning_rate": 9.744609111581589e-05, "loss": 3.103, "step": 3257 }, { "epoch": 0.26, "learning_rate": 9.744530698659141e-05, "loss": 3.1007, "step": 3258 }, { "epoch": 0.26, "learning_rate": 9.74445228573669e-05, "loss": 3.0756, "step": 3259 }, { "epoch": 0.26, "learning_rate": 9.74437387281424e-05, "loss": 2.9046, "step": 3260 }, { "epoch": 0.26, "learning_rate": 9.74429545989179e-05, "loss": 2.8669, "step": 3261 }, { "epoch": 0.26, "learning_rate": 9.744217046969341e-05, "loss": 3.1885, "step": 3262 }, { "epoch": 0.26, "learning_rate": 9.744138634046891e-05, "loss": 3.259, "step": 3263 }, { "epoch": 0.26, "learning_rate": 9.744060221124442e-05, "loss": 2.9044, "step": 3264 }, { "epoch": 0.26, "learning_rate": 9.743981808201992e-05, "loss": 3.3291, "step": 3265 }, { "epoch": 0.26, "learning_rate": 9.743903395279542e-05, "loss": 3.1744, "step": 3266 }, { "epoch": 0.26, "learning_rate": 9.743824982357093e-05, "loss": 3.1577, "step": 3267 }, { "epoch": 0.26, "learning_rate": 9.743746569434643e-05, "loss": 3.0698, "step": 3268 }, { "epoch": 0.26, "learning_rate": 9.743668156512193e-05, "loss": 2.9376, "step": 3269 }, { "epoch": 0.26, "learning_rate": 9.743589743589744e-05, "loss": 3.116, "step": 3270 }, { "epoch": 0.26, "learning_rate": 9.743511330667294e-05, "loss": 2.8982, "step": 3271 }, { "epoch": 0.26, "learning_rate": 9.743432917744844e-05, "loss": 3.1054, "step": 3272 }, { "epoch": 0.26, "learning_rate": 9.743354504822395e-05, "loss": 3.0506, "step": 3273 }, { "epoch": 0.26, "learning_rate": 9.743276091899945e-05, "loss": 3.0386, "step": 3274 }, { "epoch": 0.26, "learning_rate": 9.743197678977496e-05, "loss": 3.1433, "step": 3275 }, { "epoch": 0.26, "learning_rate": 9.743119266055046e-05, "loss": 3.3275, "step": 3276 }, { "epoch": 0.26, "learning_rate": 9.743040853132596e-05, "loss": 3.0237, "step": 3277 }, { "epoch": 0.26, "learning_rate": 9.742962440210147e-05, "loss": 3.1479, "step": 3278 }, { "epoch": 0.26, "learning_rate": 9.742884027287697e-05, "loss": 2.8435, "step": 3279 }, { "epoch": 0.26, "learning_rate": 9.742805614365247e-05, "loss": 3.17, "step": 3280 }, { "epoch": 0.26, "learning_rate": 9.742727201442798e-05, "loss": 2.955, "step": 3281 }, { "epoch": 0.26, "learning_rate": 9.742648788520348e-05, "loss": 2.7181, "step": 3282 }, { "epoch": 0.26, "learning_rate": 9.742570375597899e-05, "loss": 2.9961, "step": 3283 }, { "epoch": 0.26, "learning_rate": 9.74249196267545e-05, "loss": 3.4442, "step": 3284 }, { "epoch": 0.26, "learning_rate": 9.742413549753e-05, "loss": 2.9372, "step": 3285 }, { "epoch": 0.26, "learning_rate": 9.742335136830551e-05, "loss": 3.2237, "step": 3286 }, { "epoch": 0.26, "learning_rate": 9.742256723908101e-05, "loss": 3.0431, "step": 3287 }, { "epoch": 0.26, "learning_rate": 9.742178310985652e-05, "loss": 3.0761, "step": 3288 }, { "epoch": 0.26, "learning_rate": 9.742099898063202e-05, "loss": 2.9479, "step": 3289 }, { "epoch": 0.26, "learning_rate": 9.742021485140752e-05, "loss": 2.9939, "step": 3290 }, { "epoch": 0.26, "learning_rate": 9.741943072218303e-05, "loss": 3.2952, "step": 3291 }, { "epoch": 0.26, "learning_rate": 9.741864659295853e-05, "loss": 3.0816, "step": 3292 }, { "epoch": 0.26, "learning_rate": 9.741786246373404e-05, "loss": 3.0898, "step": 3293 }, { "epoch": 0.26, "learning_rate": 9.741707833450954e-05, "loss": 3.2708, "step": 3294 }, { "epoch": 0.26, "learning_rate": 9.741629420528503e-05, "loss": 3.1365, "step": 3295 }, { "epoch": 0.26, "learning_rate": 9.741551007606053e-05, "loss": 3.1143, "step": 3296 }, { "epoch": 0.26, "learning_rate": 9.741472594683604e-05, "loss": 3.0207, "step": 3297 }, { "epoch": 0.26, "learning_rate": 9.741394181761154e-05, "loss": 3.339, "step": 3298 }, { "epoch": 0.26, "learning_rate": 9.741315768838704e-05, "loss": 2.9333, "step": 3299 }, { "epoch": 0.26, "learning_rate": 9.741237355916255e-05, "loss": 3.1869, "step": 3300 }, { "epoch": 0.26, "learning_rate": 9.741158942993805e-05, "loss": 3.0892, "step": 3301 }, { "epoch": 0.26, "learning_rate": 9.741080530071356e-05, "loss": 3.164, "step": 3302 }, { "epoch": 0.26, "learning_rate": 9.741002117148906e-05, "loss": 3.1603, "step": 3303 }, { "epoch": 0.26, "learning_rate": 9.740923704226456e-05, "loss": 3.0297, "step": 3304 }, { "epoch": 0.26, "learning_rate": 9.740845291304007e-05, "loss": 3.1872, "step": 3305 }, { "epoch": 0.26, "learning_rate": 9.740766878381557e-05, "loss": 2.9429, "step": 3306 }, { "epoch": 0.26, "learning_rate": 9.740688465459107e-05, "loss": 3.0654, "step": 3307 }, { "epoch": 0.26, "learning_rate": 9.740610052536658e-05, "loss": 3.2921, "step": 3308 }, { "epoch": 0.26, "learning_rate": 9.74053163961421e-05, "loss": 3.1024, "step": 3309 }, { "epoch": 0.26, "learning_rate": 9.74045322669176e-05, "loss": 3.1026, "step": 3310 }, { "epoch": 0.26, "learning_rate": 9.74037481376931e-05, "loss": 2.9514, "step": 3311 }, { "epoch": 0.26, "learning_rate": 9.74029640084686e-05, "loss": 2.8031, "step": 3312 }, { "epoch": 0.26, "learning_rate": 9.740217987924411e-05, "loss": 2.994, "step": 3313 }, { "epoch": 0.26, "learning_rate": 9.740139575001961e-05, "loss": 3.1543, "step": 3314 }, { "epoch": 0.26, "learning_rate": 9.740061162079512e-05, "loss": 3.0358, "step": 3315 }, { "epoch": 0.26, "learning_rate": 9.739982749157062e-05, "loss": 3.0372, "step": 3316 }, { "epoch": 0.26, "learning_rate": 9.739904336234612e-05, "loss": 3.331, "step": 3317 }, { "epoch": 0.26, "learning_rate": 9.739825923312163e-05, "loss": 3.193, "step": 3318 }, { "epoch": 0.26, "learning_rate": 9.739747510389713e-05, "loss": 3.2949, "step": 3319 }, { "epoch": 0.26, "learning_rate": 9.739669097467263e-05, "loss": 3.1736, "step": 3320 }, { "epoch": 0.26, "learning_rate": 9.739590684544814e-05, "loss": 2.9386, "step": 3321 }, { "epoch": 0.26, "learning_rate": 9.739512271622364e-05, "loss": 3.3703, "step": 3322 }, { "epoch": 0.26, "learning_rate": 9.739433858699915e-05, "loss": 3.6071, "step": 3323 }, { "epoch": 0.26, "learning_rate": 9.739355445777465e-05, "loss": 3.0418, "step": 3324 }, { "epoch": 0.26, "learning_rate": 9.739277032855015e-05, "loss": 3.1007, "step": 3325 }, { "epoch": 0.26, "learning_rate": 9.739198619932566e-05, "loss": 2.8506, "step": 3326 }, { "epoch": 0.26, "learning_rate": 9.739120207010116e-05, "loss": 3.0994, "step": 3327 }, { "epoch": 0.26, "learning_rate": 9.739041794087666e-05, "loss": 3.1407, "step": 3328 }, { "epoch": 0.26, "learning_rate": 9.738963381165217e-05, "loss": 3.1882, "step": 3329 }, { "epoch": 0.26, "learning_rate": 9.738884968242767e-05, "loss": 3.2877, "step": 3330 }, { "epoch": 0.26, "learning_rate": 9.738806555320316e-05, "loss": 3.1501, "step": 3331 }, { "epoch": 0.26, "learning_rate": 9.738728142397867e-05, "loss": 3.1761, "step": 3332 }, { "epoch": 0.26, "learning_rate": 9.738649729475417e-05, "loss": 2.9, "step": 3333 }, { "epoch": 0.26, "learning_rate": 9.738571316552967e-05, "loss": 2.918, "step": 3334 }, { "epoch": 0.26, "learning_rate": 9.738492903630519e-05, "loss": 3.015, "step": 3335 }, { "epoch": 0.26, "learning_rate": 9.73841449070807e-05, "loss": 3.0322, "step": 3336 }, { "epoch": 0.26, "learning_rate": 9.73833607778562e-05, "loss": 3.0798, "step": 3337 }, { "epoch": 0.26, "learning_rate": 9.73825766486317e-05, "loss": 3.4616, "step": 3338 }, { "epoch": 0.26, "learning_rate": 9.73817925194072e-05, "loss": 3.3323, "step": 3339 }, { "epoch": 0.26, "learning_rate": 9.738100839018271e-05, "loss": 2.8876, "step": 3340 }, { "epoch": 0.26, "learning_rate": 9.738022426095821e-05, "loss": 3.0768, "step": 3341 }, { "epoch": 0.26, "learning_rate": 9.737944013173372e-05, "loss": 2.9237, "step": 3342 }, { "epoch": 0.26, "learning_rate": 9.737865600250922e-05, "loss": 3.1752, "step": 3343 }, { "epoch": 0.26, "learning_rate": 9.737787187328472e-05, "loss": 3.1151, "step": 3344 }, { "epoch": 0.26, "learning_rate": 9.737708774406023e-05, "loss": 2.9744, "step": 3345 }, { "epoch": 0.26, "learning_rate": 9.737630361483573e-05, "loss": 2.9569, "step": 3346 }, { "epoch": 0.26, "learning_rate": 9.737551948561123e-05, "loss": 3.0979, "step": 3347 }, { "epoch": 0.26, "learning_rate": 9.737473535638674e-05, "loss": 3.3937, "step": 3348 }, { "epoch": 0.26, "learning_rate": 9.737395122716224e-05, "loss": 2.8891, "step": 3349 }, { "epoch": 0.26, "learning_rate": 9.737316709793775e-05, "loss": 3.1427, "step": 3350 }, { "epoch": 0.26, "learning_rate": 9.737238296871325e-05, "loss": 3.3034, "step": 3351 }, { "epoch": 0.26, "learning_rate": 9.737159883948875e-05, "loss": 3.2232, "step": 3352 }, { "epoch": 0.26, "learning_rate": 9.737081471026426e-05, "loss": 3.0869, "step": 3353 }, { "epoch": 0.26, "learning_rate": 9.737003058103976e-05, "loss": 3.5297, "step": 3354 }, { "epoch": 0.26, "learning_rate": 9.736924645181526e-05, "loss": 3.063, "step": 3355 }, { "epoch": 0.26, "learning_rate": 9.736846232259077e-05, "loss": 2.8877, "step": 3356 }, { "epoch": 0.26, "learning_rate": 9.736767819336627e-05, "loss": 3.3358, "step": 3357 }, { "epoch": 0.26, "learning_rate": 9.736689406414177e-05, "loss": 3.0281, "step": 3358 }, { "epoch": 0.26, "learning_rate": 9.736610993491728e-05, "loss": 3.1502, "step": 3359 }, { "epoch": 0.26, "learning_rate": 9.736532580569278e-05, "loss": 3.3911, "step": 3360 }, { "epoch": 0.26, "learning_rate": 9.736454167646829e-05, "loss": 3.048, "step": 3361 }, { "epoch": 0.26, "learning_rate": 9.736375754724379e-05, "loss": 2.9448, "step": 3362 }, { "epoch": 0.26, "learning_rate": 9.736297341801929e-05, "loss": 3.1058, "step": 3363 }, { "epoch": 0.26, "learning_rate": 9.73621892887948e-05, "loss": 2.9977, "step": 3364 }, { "epoch": 0.26, "learning_rate": 9.73614051595703e-05, "loss": 3.1401, "step": 3365 }, { "epoch": 0.26, "learning_rate": 9.736062103034582e-05, "loss": 3.1388, "step": 3366 }, { "epoch": 0.26, "learning_rate": 9.735983690112131e-05, "loss": 3.2343, "step": 3367 }, { "epoch": 0.26, "learning_rate": 9.735905277189681e-05, "loss": 2.9629, "step": 3368 }, { "epoch": 0.26, "learning_rate": 9.735826864267232e-05, "loss": 2.8522, "step": 3369 }, { "epoch": 0.26, "learning_rate": 9.735748451344782e-05, "loss": 3.2306, "step": 3370 }, { "epoch": 0.26, "learning_rate": 9.735670038422332e-05, "loss": 3.1402, "step": 3371 }, { "epoch": 0.26, "learning_rate": 9.735591625499883e-05, "loss": 3.1767, "step": 3372 }, { "epoch": 0.26, "learning_rate": 9.735513212577433e-05, "loss": 3.1061, "step": 3373 }, { "epoch": 0.26, "learning_rate": 9.735434799654983e-05, "loss": 3.19, "step": 3374 }, { "epoch": 0.26, "learning_rate": 9.735356386732534e-05, "loss": 3.0348, "step": 3375 }, { "epoch": 0.26, "learning_rate": 9.735277973810084e-05, "loss": 3.1484, "step": 3376 }, { "epoch": 0.26, "learning_rate": 9.735199560887634e-05, "loss": 2.9615, "step": 3377 }, { "epoch": 0.26, "learning_rate": 9.735121147965185e-05, "loss": 3.1757, "step": 3378 }, { "epoch": 0.26, "learning_rate": 9.735042735042735e-05, "loss": 2.8924, "step": 3379 }, { "epoch": 0.27, "learning_rate": 9.734964322120286e-05, "loss": 2.9161, "step": 3380 }, { "epoch": 0.27, "learning_rate": 9.734885909197836e-05, "loss": 3.0296, "step": 3381 }, { "epoch": 0.27, "learning_rate": 9.734807496275386e-05, "loss": 3.0077, "step": 3382 }, { "epoch": 0.27, "learning_rate": 9.734729083352937e-05, "loss": 2.9689, "step": 3383 }, { "epoch": 0.27, "learning_rate": 9.734650670430487e-05, "loss": 2.9053, "step": 3384 }, { "epoch": 0.27, "learning_rate": 9.734572257508037e-05, "loss": 2.9711, "step": 3385 }, { "epoch": 0.27, "learning_rate": 9.734493844585588e-05, "loss": 3.0825, "step": 3386 }, { "epoch": 0.27, "learning_rate": 9.734415431663138e-05, "loss": 3.1472, "step": 3387 }, { "epoch": 0.27, "learning_rate": 9.734337018740688e-05, "loss": 2.8802, "step": 3388 }, { "epoch": 0.27, "learning_rate": 9.734258605818239e-05, "loss": 2.9877, "step": 3389 }, { "epoch": 0.27, "learning_rate": 9.734180192895789e-05, "loss": 2.8573, "step": 3390 }, { "epoch": 0.27, "learning_rate": 9.73410177997334e-05, "loss": 3.2961, "step": 3391 }, { "epoch": 0.27, "learning_rate": 9.734023367050891e-05, "loss": 3.3145, "step": 3392 }, { "epoch": 0.27, "learning_rate": 9.733944954128442e-05, "loss": 3.0686, "step": 3393 }, { "epoch": 0.27, "learning_rate": 9.733866541205992e-05, "loss": 3.107, "step": 3394 }, { "epoch": 0.27, "learning_rate": 9.733788128283542e-05, "loss": 3.0175, "step": 3395 }, { "epoch": 0.27, "learning_rate": 9.733709715361093e-05, "loss": 3.1173, "step": 3396 }, { "epoch": 0.27, "learning_rate": 9.733631302438643e-05, "loss": 3.0575, "step": 3397 }, { "epoch": 0.27, "learning_rate": 9.733552889516194e-05, "loss": 2.9987, "step": 3398 }, { "epoch": 0.27, "learning_rate": 9.733474476593744e-05, "loss": 2.9738, "step": 3399 }, { "epoch": 0.27, "learning_rate": 9.733396063671294e-05, "loss": 3.0958, "step": 3400 }, { "epoch": 0.27, "learning_rate": 9.733317650748845e-05, "loss": 3.4143, "step": 3401 }, { "epoch": 0.27, "learning_rate": 9.733239237826395e-05, "loss": 3.114, "step": 3402 }, { "epoch": 0.27, "learning_rate": 9.733160824903944e-05, "loss": 3.0736, "step": 3403 }, { "epoch": 0.27, "learning_rate": 9.733082411981494e-05, "loss": 3.1249, "step": 3404 }, { "epoch": 0.27, "learning_rate": 9.733003999059045e-05, "loss": 3.0523, "step": 3405 }, { "epoch": 0.27, "learning_rate": 9.732925586136595e-05, "loss": 3.1113, "step": 3406 }, { "epoch": 0.27, "learning_rate": 9.732847173214145e-05, "loss": 2.9611, "step": 3407 }, { "epoch": 0.27, "learning_rate": 9.732768760291696e-05, "loss": 3.2999, "step": 3408 }, { "epoch": 0.27, "learning_rate": 9.732690347369246e-05, "loss": 3.043, "step": 3409 }, { "epoch": 0.27, "learning_rate": 9.732611934446797e-05, "loss": 3.1347, "step": 3410 }, { "epoch": 0.27, "learning_rate": 9.732533521524347e-05, "loss": 3.0119, "step": 3411 }, { "epoch": 0.27, "learning_rate": 9.732455108601897e-05, "loss": 3.0102, "step": 3412 }, { "epoch": 0.27, "learning_rate": 9.732376695679448e-05, "loss": 3.5343, "step": 3413 }, { "epoch": 0.27, "learning_rate": 9.732298282756998e-05, "loss": 3.1728, "step": 3414 }, { "epoch": 0.27, "learning_rate": 9.732219869834548e-05, "loss": 3.126, "step": 3415 }, { "epoch": 0.27, "learning_rate": 9.732141456912099e-05, "loss": 3.1751, "step": 3416 }, { "epoch": 0.27, "learning_rate": 9.73206304398965e-05, "loss": 3.0023, "step": 3417 }, { "epoch": 0.27, "learning_rate": 9.731984631067201e-05, "loss": 3.0345, "step": 3418 }, { "epoch": 0.27, "learning_rate": 9.731906218144751e-05, "loss": 2.9128, "step": 3419 }, { "epoch": 0.27, "learning_rate": 9.731827805222302e-05, "loss": 3.1728, "step": 3420 }, { "epoch": 0.27, "learning_rate": 9.731749392299852e-05, "loss": 2.9221, "step": 3421 }, { "epoch": 0.27, "learning_rate": 9.731670979377402e-05, "loss": 3.0158, "step": 3422 }, { "epoch": 0.27, "learning_rate": 9.731592566454953e-05, "loss": 3.268, "step": 3423 }, { "epoch": 0.27, "learning_rate": 9.731514153532503e-05, "loss": 2.8138, "step": 3424 }, { "epoch": 0.27, "learning_rate": 9.731435740610053e-05, "loss": 3.0109, "step": 3425 }, { "epoch": 0.27, "learning_rate": 9.731357327687604e-05, "loss": 3.2212, "step": 3426 }, { "epoch": 0.27, "learning_rate": 9.731278914765154e-05, "loss": 3.2457, "step": 3427 }, { "epoch": 0.27, "learning_rate": 9.731200501842705e-05, "loss": 2.9443, "step": 3428 }, { "epoch": 0.27, "learning_rate": 9.731122088920255e-05, "loss": 2.8529, "step": 3429 }, { "epoch": 0.27, "learning_rate": 9.731043675997805e-05, "loss": 3.1778, "step": 3430 }, { "epoch": 0.27, "learning_rate": 9.730965263075356e-05, "loss": 3.2117, "step": 3431 }, { "epoch": 0.27, "learning_rate": 9.730886850152906e-05, "loss": 3.1056, "step": 3432 }, { "epoch": 0.27, "learning_rate": 9.730808437230456e-05, "loss": 3.1105, "step": 3433 }, { "epoch": 0.27, "learning_rate": 9.730730024308007e-05, "loss": 3.2013, "step": 3434 }, { "epoch": 0.27, "learning_rate": 9.730651611385557e-05, "loss": 3.2215, "step": 3435 }, { "epoch": 0.27, "learning_rate": 9.730573198463108e-05, "loss": 2.9056, "step": 3436 }, { "epoch": 0.27, "learning_rate": 9.730494785540658e-05, "loss": 3.2214, "step": 3437 }, { "epoch": 0.27, "learning_rate": 9.730416372618208e-05, "loss": 3.0053, "step": 3438 }, { "epoch": 0.27, "learning_rate": 9.730337959695757e-05, "loss": 2.8908, "step": 3439 }, { "epoch": 0.27, "learning_rate": 9.730259546773308e-05, "loss": 2.9961, "step": 3440 }, { "epoch": 0.27, "learning_rate": 9.730181133850858e-05, "loss": 2.9656, "step": 3441 }, { "epoch": 0.27, "learning_rate": 9.730102720928408e-05, "loss": 3.1832, "step": 3442 }, { "epoch": 0.27, "learning_rate": 9.73002430800596e-05, "loss": 3.3923, "step": 3443 }, { "epoch": 0.27, "learning_rate": 9.72994589508351e-05, "loss": 3.0431, "step": 3444 }, { "epoch": 0.27, "learning_rate": 9.729867482161061e-05, "loss": 3.0687, "step": 3445 }, { "epoch": 0.27, "learning_rate": 9.729789069238611e-05, "loss": 3.1596, "step": 3446 }, { "epoch": 0.27, "learning_rate": 9.729710656316162e-05, "loss": 3.1426, "step": 3447 }, { "epoch": 0.27, "learning_rate": 9.729632243393712e-05, "loss": 2.8968, "step": 3448 }, { "epoch": 0.27, "learning_rate": 9.729553830471262e-05, "loss": 2.94, "step": 3449 }, { "epoch": 0.27, "learning_rate": 9.729475417548813e-05, "loss": 2.8483, "step": 3450 }, { "epoch": 0.27, "learning_rate": 9.729397004626363e-05, "loss": 3.0196, "step": 3451 }, { "epoch": 0.27, "learning_rate": 9.729318591703913e-05, "loss": 3.1374, "step": 3452 }, { "epoch": 0.27, "learning_rate": 9.729240178781464e-05, "loss": 2.892, "step": 3453 }, { "epoch": 0.27, "learning_rate": 9.729161765859014e-05, "loss": 3.1346, "step": 3454 }, { "epoch": 0.27, "learning_rate": 9.729083352936564e-05, "loss": 2.9131, "step": 3455 }, { "epoch": 0.27, "learning_rate": 9.729004940014115e-05, "loss": 3.0428, "step": 3456 }, { "epoch": 0.27, "learning_rate": 9.728926527091665e-05, "loss": 3.3022, "step": 3457 }, { "epoch": 0.27, "learning_rate": 9.728848114169216e-05, "loss": 2.9629, "step": 3458 }, { "epoch": 0.27, "learning_rate": 9.728769701246766e-05, "loss": 3.2927, "step": 3459 }, { "epoch": 0.27, "learning_rate": 9.728691288324316e-05, "loss": 3.1276, "step": 3460 }, { "epoch": 0.27, "learning_rate": 9.728612875401867e-05, "loss": 3.0416, "step": 3461 }, { "epoch": 0.27, "learning_rate": 9.728534462479417e-05, "loss": 2.7848, "step": 3462 }, { "epoch": 0.27, "learning_rate": 9.728456049556967e-05, "loss": 3.0687, "step": 3463 }, { "epoch": 0.27, "learning_rate": 9.728377636634518e-05, "loss": 2.9647, "step": 3464 }, { "epoch": 0.27, "learning_rate": 9.728299223712068e-05, "loss": 3.1759, "step": 3465 }, { "epoch": 0.27, "learning_rate": 9.728220810789619e-05, "loss": 3.3867, "step": 3466 }, { "epoch": 0.27, "learning_rate": 9.728142397867169e-05, "loss": 3.4461, "step": 3467 }, { "epoch": 0.27, "learning_rate": 9.728063984944719e-05, "loss": 3.0009, "step": 3468 }, { "epoch": 0.27, "learning_rate": 9.72798557202227e-05, "loss": 2.9413, "step": 3469 }, { "epoch": 0.27, "learning_rate": 9.72790715909982e-05, "loss": 3.2106, "step": 3470 }, { "epoch": 0.27, "learning_rate": 9.72782874617737e-05, "loss": 3.071, "step": 3471 }, { "epoch": 0.27, "learning_rate": 9.727750333254921e-05, "loss": 3.0792, "step": 3472 }, { "epoch": 0.27, "learning_rate": 9.727671920332471e-05, "loss": 3.1722, "step": 3473 }, { "epoch": 0.27, "learning_rate": 9.727593507410023e-05, "loss": 2.8419, "step": 3474 }, { "epoch": 0.27, "learning_rate": 9.727515094487572e-05, "loss": 3.1331, "step": 3475 }, { "epoch": 0.27, "learning_rate": 9.727436681565122e-05, "loss": 2.9154, "step": 3476 }, { "epoch": 0.27, "learning_rate": 9.727358268642673e-05, "loss": 2.8443, "step": 3477 }, { "epoch": 0.27, "learning_rate": 9.727279855720223e-05, "loss": 2.9549, "step": 3478 }, { "epoch": 0.27, "learning_rate": 9.727201442797773e-05, "loss": 3.0803, "step": 3479 }, { "epoch": 0.27, "learning_rate": 9.727123029875324e-05, "loss": 2.9046, "step": 3480 }, { "epoch": 0.27, "learning_rate": 9.727044616952874e-05, "loss": 3.2452, "step": 3481 }, { "epoch": 0.27, "learning_rate": 9.726966204030424e-05, "loss": 3.1289, "step": 3482 }, { "epoch": 0.27, "learning_rate": 9.726887791107975e-05, "loss": 2.9864, "step": 3483 }, { "epoch": 0.27, "learning_rate": 9.726809378185525e-05, "loss": 3.0237, "step": 3484 }, { "epoch": 0.27, "learning_rate": 9.726730965263076e-05, "loss": 2.9319, "step": 3485 }, { "epoch": 0.27, "learning_rate": 9.726652552340626e-05, "loss": 2.8904, "step": 3486 }, { "epoch": 0.27, "learning_rate": 9.726574139418176e-05, "loss": 3.245, "step": 3487 }, { "epoch": 0.27, "learning_rate": 9.726495726495727e-05, "loss": 3.048, "step": 3488 }, { "epoch": 0.27, "learning_rate": 9.726417313573277e-05, "loss": 3.213, "step": 3489 }, { "epoch": 0.27, "learning_rate": 9.726338900650827e-05, "loss": 2.9176, "step": 3490 }, { "epoch": 0.27, "learning_rate": 9.726260487728378e-05, "loss": 3.1979, "step": 3491 }, { "epoch": 0.27, "learning_rate": 9.726182074805928e-05, "loss": 3.5258, "step": 3492 }, { "epoch": 0.27, "learning_rate": 9.726103661883478e-05, "loss": 3.354, "step": 3493 }, { "epoch": 0.27, "learning_rate": 9.726025248961029e-05, "loss": 3.1902, "step": 3494 }, { "epoch": 0.27, "learning_rate": 9.725946836038579e-05, "loss": 3.1503, "step": 3495 }, { "epoch": 0.27, "learning_rate": 9.72586842311613e-05, "loss": 3.0789, "step": 3496 }, { "epoch": 0.27, "learning_rate": 9.72579001019368e-05, "loss": 2.9557, "step": 3497 }, { "epoch": 0.27, "learning_rate": 9.72571159727123e-05, "loss": 3.1596, "step": 3498 }, { "epoch": 0.27, "learning_rate": 9.72563318434878e-05, "loss": 2.9159, "step": 3499 }, { "epoch": 0.27, "learning_rate": 9.725554771426332e-05, "loss": 3.1021, "step": 3500 }, { "epoch": 0.27, "learning_rate": 9.725476358503883e-05, "loss": 3.2842, "step": 3501 }, { "epoch": 0.27, "learning_rate": 9.725397945581433e-05, "loss": 3.1265, "step": 3502 }, { "epoch": 0.27, "learning_rate": 9.725319532658984e-05, "loss": 2.9722, "step": 3503 }, { "epoch": 0.27, "learning_rate": 9.725241119736534e-05, "loss": 2.9903, "step": 3504 }, { "epoch": 0.27, "learning_rate": 9.725162706814084e-05, "loss": 3.0984, "step": 3505 }, { "epoch": 0.27, "learning_rate": 9.725084293891635e-05, "loss": 3.1766, "step": 3506 }, { "epoch": 0.27, "learning_rate": 9.725005880969185e-05, "loss": 2.9987, "step": 3507 }, { "epoch": 0.28, "learning_rate": 9.724927468046735e-05, "loss": 3.0624, "step": 3508 }, { "epoch": 0.28, "learning_rate": 9.724849055124286e-05, "loss": 2.9429, "step": 3509 }, { "epoch": 0.28, "learning_rate": 9.724770642201836e-05, "loss": 2.8659, "step": 3510 }, { "epoch": 0.28, "learning_rate": 9.724692229279385e-05, "loss": 3.0621, "step": 3511 }, { "epoch": 0.28, "learning_rate": 9.724613816356935e-05, "loss": 3.0456, "step": 3512 }, { "epoch": 0.28, "learning_rate": 9.724535403434486e-05, "loss": 3.222, "step": 3513 }, { "epoch": 0.28, "learning_rate": 9.724456990512036e-05, "loss": 3.2128, "step": 3514 }, { "epoch": 0.28, "learning_rate": 9.724378577589587e-05, "loss": 3.452, "step": 3515 }, { "epoch": 0.28, "learning_rate": 9.724300164667137e-05, "loss": 3.0343, "step": 3516 }, { "epoch": 0.28, "learning_rate": 9.724221751744687e-05, "loss": 3.2325, "step": 3517 }, { "epoch": 0.28, "learning_rate": 9.724143338822238e-05, "loss": 2.9768, "step": 3518 }, { "epoch": 0.28, "learning_rate": 9.724064925899788e-05, "loss": 3.0464, "step": 3519 }, { "epoch": 0.28, "learning_rate": 9.723986512977338e-05, "loss": 3.0838, "step": 3520 }, { "epoch": 0.28, "learning_rate": 9.723908100054889e-05, "loss": 3.1324, "step": 3521 }, { "epoch": 0.28, "learning_rate": 9.723829687132439e-05, "loss": 3.4038, "step": 3522 }, { "epoch": 0.28, "learning_rate": 9.72375127420999e-05, "loss": 3.0914, "step": 3523 }, { "epoch": 0.28, "learning_rate": 9.72367286128754e-05, "loss": 3.2248, "step": 3524 }, { "epoch": 0.28, "learning_rate": 9.723594448365092e-05, "loss": 3.0188, "step": 3525 }, { "epoch": 0.28, "learning_rate": 9.723516035442642e-05, "loss": 3.0841, "step": 3526 }, { "epoch": 0.28, "learning_rate": 9.723437622520192e-05, "loss": 3.1843, "step": 3527 }, { "epoch": 0.28, "learning_rate": 9.723359209597743e-05, "loss": 3.291, "step": 3528 }, { "epoch": 0.28, "learning_rate": 9.723280796675293e-05, "loss": 3.263, "step": 3529 }, { "epoch": 0.28, "learning_rate": 9.723202383752843e-05, "loss": 3.2491, "step": 3530 }, { "epoch": 0.28, "learning_rate": 9.723123970830394e-05, "loss": 3.0124, "step": 3531 }, { "epoch": 0.28, "learning_rate": 9.723045557907944e-05, "loss": 3.1232, "step": 3532 }, { "epoch": 0.28, "learning_rate": 9.722967144985495e-05, "loss": 2.9201, "step": 3533 }, { "epoch": 0.28, "learning_rate": 9.722888732063045e-05, "loss": 2.8074, "step": 3534 }, { "epoch": 0.28, "learning_rate": 9.722810319140595e-05, "loss": 3.1184, "step": 3535 }, { "epoch": 0.28, "learning_rate": 9.722731906218146e-05, "loss": 3.0605, "step": 3536 }, { "epoch": 0.28, "learning_rate": 9.722653493295696e-05, "loss": 3.0601, "step": 3537 }, { "epoch": 0.28, "learning_rate": 9.722575080373246e-05, "loss": 3.1211, "step": 3538 }, { "epoch": 0.28, "learning_rate": 9.722496667450797e-05, "loss": 3.0289, "step": 3539 }, { "epoch": 0.28, "learning_rate": 9.722418254528347e-05, "loss": 3.2705, "step": 3540 }, { "epoch": 0.28, "learning_rate": 9.722339841605897e-05, "loss": 2.9562, "step": 3541 }, { "epoch": 0.28, "learning_rate": 9.722261428683448e-05, "loss": 3.0534, "step": 3542 }, { "epoch": 0.28, "learning_rate": 9.722183015760998e-05, "loss": 2.9848, "step": 3543 }, { "epoch": 0.28, "learning_rate": 9.722104602838549e-05, "loss": 3.0468, "step": 3544 }, { "epoch": 0.28, "learning_rate": 9.722026189916099e-05, "loss": 3.0574, "step": 3545 }, { "epoch": 0.28, "learning_rate": 9.721947776993649e-05, "loss": 3.1146, "step": 3546 }, { "epoch": 0.28, "learning_rate": 9.721869364071198e-05, "loss": 3.0209, "step": 3547 }, { "epoch": 0.28, "learning_rate": 9.721790951148749e-05, "loss": 3.1696, "step": 3548 }, { "epoch": 0.28, "learning_rate": 9.721712538226299e-05, "loss": 2.9351, "step": 3549 }, { "epoch": 0.28, "learning_rate": 9.72163412530385e-05, "loss": 3.2792, "step": 3550 }, { "epoch": 0.28, "learning_rate": 9.721555712381401e-05, "loss": 3.1932, "step": 3551 }, { "epoch": 0.28, "learning_rate": 9.721477299458952e-05, "loss": 3.0003, "step": 3552 }, { "epoch": 0.28, "learning_rate": 9.721398886536502e-05, "loss": 3.3177, "step": 3553 }, { "epoch": 0.28, "learning_rate": 9.721320473614052e-05, "loss": 3.0108, "step": 3554 }, { "epoch": 0.28, "learning_rate": 9.721242060691603e-05, "loss": 2.92, "step": 3555 }, { "epoch": 0.28, "learning_rate": 9.721163647769153e-05, "loss": 3.0905, "step": 3556 }, { "epoch": 0.28, "learning_rate": 9.721085234846703e-05, "loss": 3.0348, "step": 3557 }, { "epoch": 0.28, "learning_rate": 9.721006821924254e-05, "loss": 3.0129, "step": 3558 }, { "epoch": 0.28, "learning_rate": 9.720928409001804e-05, "loss": 3.1731, "step": 3559 }, { "epoch": 0.28, "learning_rate": 9.720849996079354e-05, "loss": 3.4989, "step": 3560 }, { "epoch": 0.28, "learning_rate": 9.720771583156905e-05, "loss": 3.0649, "step": 3561 }, { "epoch": 0.28, "learning_rate": 9.720693170234455e-05, "loss": 3.1979, "step": 3562 }, { "epoch": 0.28, "learning_rate": 9.720614757312006e-05, "loss": 3.2668, "step": 3563 }, { "epoch": 0.28, "learning_rate": 9.720536344389556e-05, "loss": 3.1801, "step": 3564 }, { "epoch": 0.28, "learning_rate": 9.720457931467106e-05, "loss": 2.9422, "step": 3565 }, { "epoch": 0.28, "learning_rate": 9.720379518544657e-05, "loss": 2.9856, "step": 3566 }, { "epoch": 0.28, "learning_rate": 9.720301105622207e-05, "loss": 3.0322, "step": 3567 }, { "epoch": 0.28, "learning_rate": 9.720222692699757e-05, "loss": 2.9736, "step": 3568 }, { "epoch": 0.28, "learning_rate": 9.720144279777308e-05, "loss": 3.1757, "step": 3569 }, { "epoch": 0.28, "learning_rate": 9.720065866854858e-05, "loss": 2.9854, "step": 3570 }, { "epoch": 0.28, "learning_rate": 9.719987453932409e-05, "loss": 3.1353, "step": 3571 }, { "epoch": 0.28, "learning_rate": 9.719909041009959e-05, "loss": 2.8898, "step": 3572 }, { "epoch": 0.28, "learning_rate": 9.719830628087509e-05, "loss": 2.9841, "step": 3573 }, { "epoch": 0.28, "learning_rate": 9.71975221516506e-05, "loss": 3.2831, "step": 3574 }, { "epoch": 0.28, "learning_rate": 9.71967380224261e-05, "loss": 3.1767, "step": 3575 }, { "epoch": 0.28, "learning_rate": 9.71959538932016e-05, "loss": 2.9383, "step": 3576 }, { "epoch": 0.28, "learning_rate": 9.719516976397711e-05, "loss": 3.1046, "step": 3577 }, { "epoch": 0.28, "learning_rate": 9.719438563475261e-05, "loss": 3.0604, "step": 3578 }, { "epoch": 0.28, "learning_rate": 9.719360150552811e-05, "loss": 2.9605, "step": 3579 }, { "epoch": 0.28, "learning_rate": 9.719281737630362e-05, "loss": 2.9988, "step": 3580 }, { "epoch": 0.28, "learning_rate": 9.719203324707912e-05, "loss": 3.0394, "step": 3581 }, { "epoch": 0.28, "learning_rate": 9.719124911785464e-05, "loss": 3.119, "step": 3582 }, { "epoch": 0.28, "learning_rate": 9.719046498863013e-05, "loss": 2.968, "step": 3583 }, { "epoch": 0.28, "learning_rate": 9.718968085940563e-05, "loss": 3.2433, "step": 3584 }, { "epoch": 0.28, "learning_rate": 9.718889673018114e-05, "loss": 3.0881, "step": 3585 }, { "epoch": 0.28, "learning_rate": 9.718811260095664e-05, "loss": 2.996, "step": 3586 }, { "epoch": 0.28, "learning_rate": 9.718732847173214e-05, "loss": 3.0765, "step": 3587 }, { "epoch": 0.28, "learning_rate": 9.718654434250765e-05, "loss": 3.111, "step": 3588 }, { "epoch": 0.28, "learning_rate": 9.718576021328315e-05, "loss": 3.0767, "step": 3589 }, { "epoch": 0.28, "learning_rate": 9.718497608405865e-05, "loss": 3.1669, "step": 3590 }, { "epoch": 0.28, "learning_rate": 9.718419195483416e-05, "loss": 3.4216, "step": 3591 }, { "epoch": 0.28, "learning_rate": 9.718340782560966e-05, "loss": 3.1069, "step": 3592 }, { "epoch": 0.28, "learning_rate": 9.718262369638517e-05, "loss": 2.9131, "step": 3593 }, { "epoch": 0.28, "learning_rate": 9.718183956716067e-05, "loss": 2.9593, "step": 3594 }, { "epoch": 0.28, "learning_rate": 9.718105543793617e-05, "loss": 3.2652, "step": 3595 }, { "epoch": 0.28, "learning_rate": 9.718027130871168e-05, "loss": 3.0242, "step": 3596 }, { "epoch": 0.28, "learning_rate": 9.717948717948718e-05, "loss": 3.0575, "step": 3597 }, { "epoch": 0.28, "learning_rate": 9.717870305026268e-05, "loss": 3.0038, "step": 3598 }, { "epoch": 0.28, "learning_rate": 9.717791892103819e-05, "loss": 2.9711, "step": 3599 }, { "epoch": 0.28, "learning_rate": 9.717713479181369e-05, "loss": 2.9991, "step": 3600 }, { "epoch": 0.28, "learning_rate": 9.71763506625892e-05, "loss": 2.9589, "step": 3601 }, { "epoch": 0.28, "learning_rate": 9.71755665333647e-05, "loss": 3.47, "step": 3602 }, { "epoch": 0.28, "learning_rate": 9.71747824041402e-05, "loss": 2.9126, "step": 3603 }, { "epoch": 0.28, "learning_rate": 9.71739982749157e-05, "loss": 3.1668, "step": 3604 }, { "epoch": 0.28, "learning_rate": 9.717321414569121e-05, "loss": 3.0082, "step": 3605 }, { "epoch": 0.28, "learning_rate": 9.717243001646671e-05, "loss": 3.28, "step": 3606 }, { "epoch": 0.28, "learning_rate": 9.717164588724222e-05, "loss": 2.9926, "step": 3607 }, { "epoch": 0.28, "learning_rate": 9.717086175801773e-05, "loss": 3.2233, "step": 3608 }, { "epoch": 0.28, "learning_rate": 9.717007762879324e-05, "loss": 3.045, "step": 3609 }, { "epoch": 0.28, "learning_rate": 9.716929349956874e-05, "loss": 3.0178, "step": 3610 }, { "epoch": 0.28, "learning_rate": 9.716850937034425e-05, "loss": 2.9868, "step": 3611 }, { "epoch": 0.28, "learning_rate": 9.716772524111975e-05, "loss": 3.1625, "step": 3612 }, { "epoch": 0.28, "learning_rate": 9.716694111189525e-05, "loss": 3.0552, "step": 3613 }, { "epoch": 0.28, "learning_rate": 9.716615698267076e-05, "loss": 3.1419, "step": 3614 }, { "epoch": 0.28, "learning_rate": 9.716537285344626e-05, "loss": 3.0069, "step": 3615 }, { "epoch": 0.28, "learning_rate": 9.716458872422176e-05, "loss": 3.1639, "step": 3616 }, { "epoch": 0.28, "learning_rate": 9.716380459499727e-05, "loss": 3.3474, "step": 3617 }, { "epoch": 0.28, "learning_rate": 9.716302046577277e-05, "loss": 3.0765, "step": 3618 }, { "epoch": 0.28, "learning_rate": 9.716223633654826e-05, "loss": 2.996, "step": 3619 }, { "epoch": 0.28, "learning_rate": 9.716145220732377e-05, "loss": 2.9368, "step": 3620 }, { "epoch": 0.28, "learning_rate": 9.716066807809927e-05, "loss": 3.2165, "step": 3621 }, { "epoch": 0.28, "learning_rate": 9.715988394887477e-05, "loss": 2.9779, "step": 3622 }, { "epoch": 0.28, "learning_rate": 9.715909981965028e-05, "loss": 3.1563, "step": 3623 }, { "epoch": 0.28, "learning_rate": 9.715831569042578e-05, "loss": 3.2642, "step": 3624 }, { "epoch": 0.28, "learning_rate": 9.715753156120128e-05, "loss": 2.9419, "step": 3625 }, { "epoch": 0.28, "learning_rate": 9.715674743197679e-05, "loss": 3.0093, "step": 3626 }, { "epoch": 0.28, "learning_rate": 9.715596330275229e-05, "loss": 3.1391, "step": 3627 }, { "epoch": 0.28, "learning_rate": 9.71551791735278e-05, "loss": 3.2854, "step": 3628 }, { "epoch": 0.28, "learning_rate": 9.71543950443033e-05, "loss": 3.0818, "step": 3629 }, { "epoch": 0.28, "learning_rate": 9.71536109150788e-05, "loss": 3.037, "step": 3630 }, { "epoch": 0.28, "learning_rate": 9.71528267858543e-05, "loss": 2.9658, "step": 3631 }, { "epoch": 0.28, "learning_rate": 9.715204265662981e-05, "loss": 3.0594, "step": 3632 }, { "epoch": 0.28, "learning_rate": 9.715125852740533e-05, "loss": 3.3715, "step": 3633 }, { "epoch": 0.28, "learning_rate": 9.715047439818083e-05, "loss": 2.9836, "step": 3634 }, { "epoch": 0.29, "learning_rate": 9.714969026895633e-05, "loss": 3.343, "step": 3635 }, { "epoch": 0.29, "learning_rate": 9.714890613973184e-05, "loss": 3.1084, "step": 3636 }, { "epoch": 0.29, "learning_rate": 9.714812201050734e-05, "loss": 3.1831, "step": 3637 }, { "epoch": 0.29, "learning_rate": 9.714733788128285e-05, "loss": 3.3912, "step": 3638 }, { "epoch": 0.29, "learning_rate": 9.714655375205835e-05, "loss": 2.9761, "step": 3639 }, { "epoch": 0.29, "learning_rate": 9.714576962283385e-05, "loss": 3.1379, "step": 3640 }, { "epoch": 0.29, "learning_rate": 9.714498549360936e-05, "loss": 2.992, "step": 3641 }, { "epoch": 0.29, "learning_rate": 9.714420136438486e-05, "loss": 3.0088, "step": 3642 }, { "epoch": 0.29, "learning_rate": 9.714341723516036e-05, "loss": 3.0559, "step": 3643 }, { "epoch": 0.29, "learning_rate": 9.714263310593587e-05, "loss": 2.877, "step": 3644 }, { "epoch": 0.29, "learning_rate": 9.714184897671137e-05, "loss": 3.2157, "step": 3645 }, { "epoch": 0.29, "learning_rate": 9.714106484748687e-05, "loss": 2.9847, "step": 3646 }, { "epoch": 0.29, "learning_rate": 9.714028071826238e-05, "loss": 2.9999, "step": 3647 }, { "epoch": 0.29, "learning_rate": 9.713949658903788e-05, "loss": 3.1657, "step": 3648 }, { "epoch": 0.29, "learning_rate": 9.713871245981339e-05, "loss": 3.1787, "step": 3649 }, { "epoch": 0.29, "learning_rate": 9.713792833058889e-05, "loss": 3.1069, "step": 3650 }, { "epoch": 0.29, "learning_rate": 9.713714420136439e-05, "loss": 2.9191, "step": 3651 }, { "epoch": 0.29, "learning_rate": 9.71363600721399e-05, "loss": 3.2185, "step": 3652 }, { "epoch": 0.29, "learning_rate": 9.71355759429154e-05, "loss": 3.0532, "step": 3653 }, { "epoch": 0.29, "learning_rate": 9.71347918136909e-05, "loss": 3.2163, "step": 3654 }, { "epoch": 0.29, "learning_rate": 9.71340076844664e-05, "loss": 2.8741, "step": 3655 }, { "epoch": 0.29, "learning_rate": 9.71332235552419e-05, "loss": 2.752, "step": 3656 }, { "epoch": 0.29, "learning_rate": 9.71324394260174e-05, "loss": 3.0314, "step": 3657 }, { "epoch": 0.29, "learning_rate": 9.71316552967929e-05, "loss": 2.8926, "step": 3658 }, { "epoch": 0.29, "learning_rate": 9.713087116756842e-05, "loss": 3.1478, "step": 3659 }, { "epoch": 0.29, "learning_rate": 9.713008703834393e-05, "loss": 3.1605, "step": 3660 }, { "epoch": 0.29, "learning_rate": 9.712930290911943e-05, "loss": 2.9217, "step": 3661 }, { "epoch": 0.29, "learning_rate": 9.712851877989493e-05, "loss": 2.8266, "step": 3662 }, { "epoch": 0.29, "learning_rate": 9.712773465067044e-05, "loss": 3.1326, "step": 3663 }, { "epoch": 0.29, "learning_rate": 9.712695052144594e-05, "loss": 3.0299, "step": 3664 }, { "epoch": 0.29, "learning_rate": 9.712616639222144e-05, "loss": 3.0, "step": 3665 }, { "epoch": 0.29, "learning_rate": 9.712538226299695e-05, "loss": 3.2389, "step": 3666 }, { "epoch": 0.29, "learning_rate": 9.712459813377245e-05, "loss": 2.9543, "step": 3667 }, { "epoch": 0.29, "learning_rate": 9.712381400454796e-05, "loss": 3.1448, "step": 3668 }, { "epoch": 0.29, "learning_rate": 9.712302987532346e-05, "loss": 3.4548, "step": 3669 }, { "epoch": 0.29, "learning_rate": 9.712224574609896e-05, "loss": 3.2004, "step": 3670 }, { "epoch": 0.29, "learning_rate": 9.712146161687447e-05, "loss": 3.1709, "step": 3671 }, { "epoch": 0.29, "learning_rate": 9.712067748764997e-05, "loss": 3.0624, "step": 3672 }, { "epoch": 0.29, "learning_rate": 9.711989335842547e-05, "loss": 3.1188, "step": 3673 }, { "epoch": 0.29, "learning_rate": 9.711910922920098e-05, "loss": 2.8309, "step": 3674 }, { "epoch": 0.29, "learning_rate": 9.711832509997648e-05, "loss": 3.2688, "step": 3675 }, { "epoch": 0.29, "learning_rate": 9.711754097075198e-05, "loss": 3.1195, "step": 3676 }, { "epoch": 0.29, "learning_rate": 9.711675684152749e-05, "loss": 3.163, "step": 3677 }, { "epoch": 0.29, "learning_rate": 9.711597271230299e-05, "loss": 2.9521, "step": 3678 }, { "epoch": 0.29, "learning_rate": 9.71151885830785e-05, "loss": 2.865, "step": 3679 }, { "epoch": 0.29, "learning_rate": 9.7114404453854e-05, "loss": 3.0173, "step": 3680 }, { "epoch": 0.29, "learning_rate": 9.71136203246295e-05, "loss": 2.8999, "step": 3681 }, { "epoch": 0.29, "learning_rate": 9.711283619540501e-05, "loss": 3.0403, "step": 3682 }, { "epoch": 0.29, "learning_rate": 9.711205206618051e-05, "loss": 3.0291, "step": 3683 }, { "epoch": 0.29, "learning_rate": 9.711126793695601e-05, "loss": 3.0235, "step": 3684 }, { "epoch": 0.29, "learning_rate": 9.711048380773152e-05, "loss": 3.0531, "step": 3685 }, { "epoch": 0.29, "learning_rate": 9.710969967850702e-05, "loss": 3.209, "step": 3686 }, { "epoch": 0.29, "learning_rate": 9.710891554928253e-05, "loss": 2.9522, "step": 3687 }, { "epoch": 0.29, "learning_rate": 9.710813142005803e-05, "loss": 2.867, "step": 3688 }, { "epoch": 0.29, "learning_rate": 9.710734729083353e-05, "loss": 3.1913, "step": 3689 }, { "epoch": 0.29, "learning_rate": 9.710656316160905e-05, "loss": 2.9752, "step": 3690 }, { "epoch": 0.29, "learning_rate": 9.710577903238454e-05, "loss": 3.1072, "step": 3691 }, { "epoch": 0.29, "learning_rate": 9.710499490316004e-05, "loss": 3.2441, "step": 3692 }, { "epoch": 0.29, "learning_rate": 9.710421077393555e-05, "loss": 3.2715, "step": 3693 }, { "epoch": 0.29, "learning_rate": 9.710342664471105e-05, "loss": 2.8991, "step": 3694 }, { "epoch": 0.29, "learning_rate": 9.710264251548655e-05, "loss": 2.9391, "step": 3695 }, { "epoch": 0.29, "learning_rate": 9.710185838626206e-05, "loss": 3.1418, "step": 3696 }, { "epoch": 0.29, "learning_rate": 9.710107425703756e-05, "loss": 3.0201, "step": 3697 }, { "epoch": 0.29, "learning_rate": 9.710029012781307e-05, "loss": 3.0172, "step": 3698 }, { "epoch": 0.29, "learning_rate": 9.709950599858857e-05, "loss": 3.0, "step": 3699 }, { "epoch": 0.29, "learning_rate": 9.709872186936407e-05, "loss": 3.6181, "step": 3700 }, { "epoch": 0.29, "learning_rate": 9.709793774013958e-05, "loss": 3.1591, "step": 3701 }, { "epoch": 0.29, "learning_rate": 9.709715361091508e-05, "loss": 3.0592, "step": 3702 }, { "epoch": 0.29, "learning_rate": 9.709636948169058e-05, "loss": 3.183, "step": 3703 }, { "epoch": 0.29, "learning_rate": 9.709558535246609e-05, "loss": 3.1496, "step": 3704 }, { "epoch": 0.29, "learning_rate": 9.709480122324159e-05, "loss": 3.1723, "step": 3705 }, { "epoch": 0.29, "learning_rate": 9.70940170940171e-05, "loss": 3.0524, "step": 3706 }, { "epoch": 0.29, "learning_rate": 9.70932329647926e-05, "loss": 3.2351, "step": 3707 }, { "epoch": 0.29, "learning_rate": 9.70924488355681e-05, "loss": 3.1317, "step": 3708 }, { "epoch": 0.29, "learning_rate": 9.70916647063436e-05, "loss": 2.8777, "step": 3709 }, { "epoch": 0.29, "learning_rate": 9.709088057711911e-05, "loss": 2.8698, "step": 3710 }, { "epoch": 0.29, "learning_rate": 9.709009644789461e-05, "loss": 3.2469, "step": 3711 }, { "epoch": 0.29, "learning_rate": 9.708931231867012e-05, "loss": 3.0992, "step": 3712 }, { "epoch": 0.29, "learning_rate": 9.708852818944562e-05, "loss": 3.2132, "step": 3713 }, { "epoch": 0.29, "learning_rate": 9.708774406022112e-05, "loss": 2.9929, "step": 3714 }, { "epoch": 0.29, "learning_rate": 9.708695993099663e-05, "loss": 3.3622, "step": 3715 }, { "epoch": 0.29, "learning_rate": 9.708617580177215e-05, "loss": 3.3861, "step": 3716 }, { "epoch": 0.29, "learning_rate": 9.708539167254765e-05, "loss": 3.2785, "step": 3717 }, { "epoch": 0.29, "learning_rate": 9.708460754332315e-05, "loss": 2.965, "step": 3718 }, { "epoch": 0.29, "learning_rate": 9.708382341409866e-05, "loss": 3.0081, "step": 3719 }, { "epoch": 0.29, "learning_rate": 9.708303928487416e-05, "loss": 3.0898, "step": 3720 }, { "epoch": 0.29, "learning_rate": 9.708225515564966e-05, "loss": 3.3732, "step": 3721 }, { "epoch": 0.29, "learning_rate": 9.708147102642517e-05, "loss": 3.3338, "step": 3722 }, { "epoch": 0.29, "learning_rate": 9.708068689720067e-05, "loss": 3.2563, "step": 3723 }, { "epoch": 0.29, "learning_rate": 9.707990276797617e-05, "loss": 2.9776, "step": 3724 }, { "epoch": 0.29, "learning_rate": 9.707911863875168e-05, "loss": 3.0443, "step": 3725 }, { "epoch": 0.29, "learning_rate": 9.707833450952718e-05, "loss": 2.9939, "step": 3726 }, { "epoch": 0.29, "learning_rate": 9.707755038030267e-05, "loss": 2.9727, "step": 3727 }, { "epoch": 0.29, "learning_rate": 9.707676625107818e-05, "loss": 2.9982, "step": 3728 }, { "epoch": 0.29, "learning_rate": 9.707598212185368e-05, "loss": 3.1977, "step": 3729 }, { "epoch": 0.29, "learning_rate": 9.707519799262918e-05, "loss": 3.4363, "step": 3730 }, { "epoch": 0.29, "learning_rate": 9.707441386340469e-05, "loss": 3.5455, "step": 3731 }, { "epoch": 0.29, "learning_rate": 9.707362973418019e-05, "loss": 3.1376, "step": 3732 }, { "epoch": 0.29, "learning_rate": 9.70728456049557e-05, "loss": 3.1202, "step": 3733 }, { "epoch": 0.29, "learning_rate": 9.70720614757312e-05, "loss": 2.9301, "step": 3734 }, { "epoch": 0.29, "learning_rate": 9.70712773465067e-05, "loss": 3.1093, "step": 3735 }, { "epoch": 0.29, "learning_rate": 9.70704932172822e-05, "loss": 3.2499, "step": 3736 }, { "epoch": 0.29, "learning_rate": 9.706970908805771e-05, "loss": 2.9747, "step": 3737 }, { "epoch": 0.29, "learning_rate": 9.706892495883321e-05, "loss": 3.1313, "step": 3738 }, { "epoch": 0.29, "learning_rate": 9.706814082960872e-05, "loss": 3.3504, "step": 3739 }, { "epoch": 0.29, "learning_rate": 9.706735670038422e-05, "loss": 3.0875, "step": 3740 }, { "epoch": 0.29, "learning_rate": 9.706657257115974e-05, "loss": 3.0394, "step": 3741 }, { "epoch": 0.29, "learning_rate": 9.706578844193524e-05, "loss": 3.0511, "step": 3742 }, { "epoch": 0.29, "learning_rate": 9.706500431271074e-05, "loss": 2.7769, "step": 3743 }, { "epoch": 0.29, "learning_rate": 9.706422018348625e-05, "loss": 3.1254, "step": 3744 }, { "epoch": 0.29, "learning_rate": 9.706343605426175e-05, "loss": 3.0856, "step": 3745 }, { "epoch": 0.29, "learning_rate": 9.706265192503726e-05, "loss": 3.029, "step": 3746 }, { "epoch": 0.29, "learning_rate": 9.706186779581276e-05, "loss": 3.1622, "step": 3747 }, { "epoch": 0.29, "learning_rate": 9.706108366658826e-05, "loss": 3.2699, "step": 3748 }, { "epoch": 0.29, "learning_rate": 9.706029953736377e-05, "loss": 3.1963, "step": 3749 }, { "epoch": 0.29, "learning_rate": 9.705951540813927e-05, "loss": 3.0346, "step": 3750 }, { "epoch": 0.29, "learning_rate": 9.705873127891477e-05, "loss": 3.0327, "step": 3751 }, { "epoch": 0.29, "learning_rate": 9.705794714969028e-05, "loss": 3.2109, "step": 3752 }, { "epoch": 0.29, "learning_rate": 9.705716302046578e-05, "loss": 3.0609, "step": 3753 }, { "epoch": 0.29, "learning_rate": 9.705637889124129e-05, "loss": 3.0576, "step": 3754 }, { "epoch": 0.29, "learning_rate": 9.705559476201679e-05, "loss": 2.8643, "step": 3755 }, { "epoch": 0.29, "learning_rate": 9.705481063279229e-05, "loss": 3.1235, "step": 3756 }, { "epoch": 0.29, "learning_rate": 9.70540265035678e-05, "loss": 3.3888, "step": 3757 }, { "epoch": 0.29, "learning_rate": 9.70532423743433e-05, "loss": 3.3301, "step": 3758 }, { "epoch": 0.29, "learning_rate": 9.70524582451188e-05, "loss": 3.1774, "step": 3759 }, { "epoch": 0.29, "learning_rate": 9.705167411589431e-05, "loss": 3.1721, "step": 3760 }, { "epoch": 0.29, "learning_rate": 9.705088998666981e-05, "loss": 2.9422, "step": 3761 }, { "epoch": 0.29, "learning_rate": 9.705010585744531e-05, "loss": 2.8353, "step": 3762 }, { "epoch": 0.3, "learning_rate": 9.70493217282208e-05, "loss": 2.88, "step": 3763 }, { "epoch": 0.3, "learning_rate": 9.704853759899631e-05, "loss": 2.8512, "step": 3764 }, { "epoch": 0.3, "learning_rate": 9.704775346977181e-05, "loss": 3.1592, "step": 3765 }, { "epoch": 0.3, "learning_rate": 9.704696934054732e-05, "loss": 3.2029, "step": 3766 }, { "epoch": 0.3, "learning_rate": 9.704618521132283e-05, "loss": 3.2064, "step": 3767 }, { "epoch": 0.3, "learning_rate": 9.704540108209834e-05, "loss": 3.282, "step": 3768 }, { "epoch": 0.3, "learning_rate": 9.704461695287384e-05, "loss": 3.093, "step": 3769 }, { "epoch": 0.3, "learning_rate": 9.704383282364934e-05, "loss": 3.0127, "step": 3770 }, { "epoch": 0.3, "learning_rate": 9.704304869442485e-05, "loss": 3.2935, "step": 3771 }, { "epoch": 0.3, "learning_rate": 9.704226456520035e-05, "loss": 3.0861, "step": 3772 }, { "epoch": 0.3, "learning_rate": 9.704148043597586e-05, "loss": 3.4577, "step": 3773 }, { "epoch": 0.3, "learning_rate": 9.704069630675136e-05, "loss": 3.1297, "step": 3774 }, { "epoch": 0.3, "learning_rate": 9.703991217752686e-05, "loss": 3.3689, "step": 3775 }, { "epoch": 0.3, "learning_rate": 9.703912804830237e-05, "loss": 3.1767, "step": 3776 }, { "epoch": 0.3, "learning_rate": 9.703834391907787e-05, "loss": 3.1731, "step": 3777 }, { "epoch": 0.3, "learning_rate": 9.703755978985337e-05, "loss": 3.0971, "step": 3778 }, { "epoch": 0.3, "learning_rate": 9.703677566062888e-05, "loss": 3.26, "step": 3779 }, { "epoch": 0.3, "learning_rate": 9.703599153140438e-05, "loss": 3.2149, "step": 3780 }, { "epoch": 0.3, "learning_rate": 9.703520740217988e-05, "loss": 2.9551, "step": 3781 }, { "epoch": 0.3, "learning_rate": 9.703442327295539e-05, "loss": 3.0587, "step": 3782 }, { "epoch": 0.3, "learning_rate": 9.703363914373089e-05, "loss": 2.8949, "step": 3783 }, { "epoch": 0.3, "learning_rate": 9.70328550145064e-05, "loss": 3.1495, "step": 3784 }, { "epoch": 0.3, "learning_rate": 9.70320708852819e-05, "loss": 2.927, "step": 3785 }, { "epoch": 0.3, "learning_rate": 9.70312867560574e-05, "loss": 3.2958, "step": 3786 }, { "epoch": 0.3, "learning_rate": 9.70305026268329e-05, "loss": 2.9721, "step": 3787 }, { "epoch": 0.3, "learning_rate": 9.702971849760841e-05, "loss": 2.8812, "step": 3788 }, { "epoch": 0.3, "learning_rate": 9.702893436838391e-05, "loss": 2.9965, "step": 3789 }, { "epoch": 0.3, "learning_rate": 9.702815023915942e-05, "loss": 3.5372, "step": 3790 }, { "epoch": 0.3, "learning_rate": 9.702736610993492e-05, "loss": 3.2521, "step": 3791 }, { "epoch": 0.3, "learning_rate": 9.702658198071042e-05, "loss": 3.0411, "step": 3792 }, { "epoch": 0.3, "learning_rate": 9.702579785148593e-05, "loss": 3.1966, "step": 3793 }, { "epoch": 0.3, "learning_rate": 9.702501372226143e-05, "loss": 3.1464, "step": 3794 }, { "epoch": 0.3, "learning_rate": 9.702422959303694e-05, "loss": 3.2195, "step": 3795 }, { "epoch": 0.3, "learning_rate": 9.702344546381244e-05, "loss": 2.9429, "step": 3796 }, { "epoch": 0.3, "learning_rate": 9.702266133458794e-05, "loss": 3.0289, "step": 3797 }, { "epoch": 0.3, "learning_rate": 9.702187720536346e-05, "loss": 2.8898, "step": 3798 }, { "epoch": 0.3, "learning_rate": 9.702109307613895e-05, "loss": 2.895, "step": 3799 }, { "epoch": 0.3, "learning_rate": 9.702030894691445e-05, "loss": 3.3318, "step": 3800 }, { "epoch": 0.3, "learning_rate": 9.701952481768996e-05, "loss": 3.0519, "step": 3801 }, { "epoch": 0.3, "learning_rate": 9.701874068846546e-05, "loss": 3.2106, "step": 3802 }, { "epoch": 0.3, "learning_rate": 9.701795655924097e-05, "loss": 2.7584, "step": 3803 }, { "epoch": 0.3, "learning_rate": 9.701717243001647e-05, "loss": 3.0032, "step": 3804 }, { "epoch": 0.3, "learning_rate": 9.701638830079197e-05, "loss": 2.9237, "step": 3805 }, { "epoch": 0.3, "learning_rate": 9.701560417156748e-05, "loss": 3.0714, "step": 3806 }, { "epoch": 0.3, "learning_rate": 9.701482004234298e-05, "loss": 2.9753, "step": 3807 }, { "epoch": 0.3, "learning_rate": 9.701403591311848e-05, "loss": 3.2293, "step": 3808 }, { "epoch": 0.3, "learning_rate": 9.701325178389399e-05, "loss": 3.0482, "step": 3809 }, { "epoch": 0.3, "learning_rate": 9.701246765466949e-05, "loss": 2.9832, "step": 3810 }, { "epoch": 0.3, "learning_rate": 9.7011683525445e-05, "loss": 2.9675, "step": 3811 }, { "epoch": 0.3, "learning_rate": 9.70108993962205e-05, "loss": 3.3128, "step": 3812 }, { "epoch": 0.3, "learning_rate": 9.7010115266996e-05, "loss": 2.993, "step": 3813 }, { "epoch": 0.3, "learning_rate": 9.70093311377715e-05, "loss": 2.9517, "step": 3814 }, { "epoch": 0.3, "learning_rate": 9.700854700854701e-05, "loss": 3.0268, "step": 3815 }, { "epoch": 0.3, "learning_rate": 9.700776287932251e-05, "loss": 3.0368, "step": 3816 }, { "epoch": 0.3, "learning_rate": 9.700697875009802e-05, "loss": 2.9574, "step": 3817 }, { "epoch": 0.3, "learning_rate": 9.700619462087352e-05, "loss": 3.1233, "step": 3818 }, { "epoch": 0.3, "learning_rate": 9.700541049164902e-05, "loss": 3.2789, "step": 3819 }, { "epoch": 0.3, "learning_rate": 9.700462636242453e-05, "loss": 3.0693, "step": 3820 }, { "epoch": 0.3, "learning_rate": 9.700384223320003e-05, "loss": 2.9903, "step": 3821 }, { "epoch": 0.3, "learning_rate": 9.700305810397554e-05, "loss": 3.0706, "step": 3822 }, { "epoch": 0.3, "learning_rate": 9.700227397475104e-05, "loss": 3.1692, "step": 3823 }, { "epoch": 0.3, "learning_rate": 9.700148984552656e-05, "loss": 3.0684, "step": 3824 }, { "epoch": 0.3, "learning_rate": 9.700070571630206e-05, "loss": 3.3632, "step": 3825 }, { "epoch": 0.3, "learning_rate": 9.699992158707756e-05, "loss": 3.3485, "step": 3826 }, { "epoch": 0.3, "learning_rate": 9.699913745785307e-05, "loss": 2.9073, "step": 3827 }, { "epoch": 0.3, "learning_rate": 9.699835332862857e-05, "loss": 3.002, "step": 3828 }, { "epoch": 0.3, "learning_rate": 9.699756919940407e-05, "loss": 2.7725, "step": 3829 }, { "epoch": 0.3, "learning_rate": 9.699678507017958e-05, "loss": 3.0183, "step": 3830 }, { "epoch": 0.3, "learning_rate": 9.699600094095508e-05, "loss": 2.9536, "step": 3831 }, { "epoch": 0.3, "learning_rate": 9.699521681173059e-05, "loss": 3.2501, "step": 3832 }, { "epoch": 0.3, "learning_rate": 9.699443268250609e-05, "loss": 3.1041, "step": 3833 }, { "epoch": 0.3, "learning_rate": 9.699364855328159e-05, "loss": 3.1199, "step": 3834 }, { "epoch": 0.3, "learning_rate": 9.699286442405708e-05, "loss": 2.8905, "step": 3835 }, { "epoch": 0.3, "learning_rate": 9.699208029483259e-05, "loss": 3.157, "step": 3836 }, { "epoch": 0.3, "learning_rate": 9.699129616560809e-05, "loss": 3.1501, "step": 3837 }, { "epoch": 0.3, "learning_rate": 9.69905120363836e-05, "loss": 3.3637, "step": 3838 }, { "epoch": 0.3, "learning_rate": 9.69897279071591e-05, "loss": 2.921, "step": 3839 }, { "epoch": 0.3, "learning_rate": 9.69889437779346e-05, "loss": 3.3839, "step": 3840 }, { "epoch": 0.3, "learning_rate": 9.69881596487101e-05, "loss": 2.9633, "step": 3841 }, { "epoch": 0.3, "learning_rate": 9.698737551948561e-05, "loss": 3.0274, "step": 3842 }, { "epoch": 0.3, "learning_rate": 9.698659139026111e-05, "loss": 3.0182, "step": 3843 }, { "epoch": 0.3, "learning_rate": 9.698580726103662e-05, "loss": 2.9618, "step": 3844 }, { "epoch": 0.3, "learning_rate": 9.698502313181212e-05, "loss": 3.0276, "step": 3845 }, { "epoch": 0.3, "learning_rate": 9.698423900258762e-05, "loss": 3.1458, "step": 3846 }, { "epoch": 0.3, "learning_rate": 9.698345487336313e-05, "loss": 3.08, "step": 3847 }, { "epoch": 0.3, "learning_rate": 9.698267074413863e-05, "loss": 3.0655, "step": 3848 }, { "epoch": 0.3, "learning_rate": 9.698188661491415e-05, "loss": 3.0063, "step": 3849 }, { "epoch": 0.3, "learning_rate": 9.698110248568965e-05, "loss": 3.2463, "step": 3850 }, { "epoch": 0.3, "learning_rate": 9.698031835646516e-05, "loss": 3.263, "step": 3851 }, { "epoch": 0.3, "learning_rate": 9.697953422724066e-05, "loss": 2.9438, "step": 3852 }, { "epoch": 0.3, "learning_rate": 9.697875009801616e-05, "loss": 3.08, "step": 3853 }, { "epoch": 0.3, "learning_rate": 9.697796596879167e-05, "loss": 3.0545, "step": 3854 }, { "epoch": 0.3, "learning_rate": 9.697718183956717e-05, "loss": 3.2941, "step": 3855 }, { "epoch": 0.3, "learning_rate": 9.697639771034267e-05, "loss": 3.2513, "step": 3856 }, { "epoch": 0.3, "learning_rate": 9.697561358111818e-05, "loss": 3.0882, "step": 3857 }, { "epoch": 0.3, "learning_rate": 9.697482945189368e-05, "loss": 2.9613, "step": 3858 }, { "epoch": 0.3, "learning_rate": 9.697404532266918e-05, "loss": 2.9968, "step": 3859 }, { "epoch": 0.3, "learning_rate": 9.697326119344469e-05, "loss": 3.0727, "step": 3860 }, { "epoch": 0.3, "learning_rate": 9.697247706422019e-05, "loss": 3.1322, "step": 3861 }, { "epoch": 0.3, "learning_rate": 9.69716929349957e-05, "loss": 3.043, "step": 3862 }, { "epoch": 0.3, "learning_rate": 9.69709088057712e-05, "loss": 3.0213, "step": 3863 }, { "epoch": 0.3, "learning_rate": 9.69701246765467e-05, "loss": 2.8114, "step": 3864 }, { "epoch": 0.3, "learning_rate": 9.696934054732221e-05, "loss": 3.2556, "step": 3865 }, { "epoch": 0.3, "learning_rate": 9.696855641809771e-05, "loss": 2.8671, "step": 3866 }, { "epoch": 0.3, "learning_rate": 9.696777228887321e-05, "loss": 2.9684, "step": 3867 }, { "epoch": 0.3, "learning_rate": 9.696698815964872e-05, "loss": 2.9276, "step": 3868 }, { "epoch": 0.3, "learning_rate": 9.696620403042422e-05, "loss": 3.2555, "step": 3869 }, { "epoch": 0.3, "learning_rate": 9.696541990119973e-05, "loss": 3.3511, "step": 3870 }, { "epoch": 0.3, "learning_rate": 9.696463577197522e-05, "loss": 3.1244, "step": 3871 }, { "epoch": 0.3, "learning_rate": 9.696385164275072e-05, "loss": 3.0889, "step": 3872 }, { "epoch": 0.3, "learning_rate": 9.696306751352622e-05, "loss": 3.0457, "step": 3873 }, { "epoch": 0.3, "learning_rate": 9.696228338430173e-05, "loss": 3.2294, "step": 3874 }, { "epoch": 0.3, "learning_rate": 9.696149925507724e-05, "loss": 3.0185, "step": 3875 }, { "epoch": 0.3, "learning_rate": 9.696071512585275e-05, "loss": 2.8666, "step": 3876 }, { "epoch": 0.3, "learning_rate": 9.695993099662825e-05, "loss": 3.2605, "step": 3877 }, { "epoch": 0.3, "learning_rate": 9.695914686740375e-05, "loss": 2.8847, "step": 3878 }, { "epoch": 0.3, "learning_rate": 9.695836273817926e-05, "loss": 3.1012, "step": 3879 }, { "epoch": 0.3, "learning_rate": 9.695757860895476e-05, "loss": 3.2383, "step": 3880 }, { "epoch": 0.3, "learning_rate": 9.695679447973027e-05, "loss": 3.3464, "step": 3881 }, { "epoch": 0.3, "learning_rate": 9.695601035050577e-05, "loss": 2.9572, "step": 3882 }, { "epoch": 0.3, "learning_rate": 9.695522622128127e-05, "loss": 3.0532, "step": 3883 }, { "epoch": 0.3, "learning_rate": 9.695444209205678e-05, "loss": 2.9478, "step": 3884 }, { "epoch": 0.3, "learning_rate": 9.695365796283228e-05, "loss": 3.1661, "step": 3885 }, { "epoch": 0.3, "learning_rate": 9.695287383360778e-05, "loss": 2.9857, "step": 3886 }, { "epoch": 0.3, "learning_rate": 9.695208970438329e-05, "loss": 3.004, "step": 3887 }, { "epoch": 0.3, "learning_rate": 9.695130557515879e-05, "loss": 2.9627, "step": 3888 }, { "epoch": 0.3, "learning_rate": 9.69505214459343e-05, "loss": 3.46, "step": 3889 }, { "epoch": 0.31, "learning_rate": 9.69497373167098e-05, "loss": 3.1624, "step": 3890 }, { "epoch": 0.31, "learning_rate": 9.69489531874853e-05, "loss": 3.1368, "step": 3891 }, { "epoch": 0.31, "learning_rate": 9.69481690582608e-05, "loss": 3.0671, "step": 3892 }, { "epoch": 0.31, "learning_rate": 9.694738492903631e-05, "loss": 3.2628, "step": 3893 }, { "epoch": 0.31, "learning_rate": 9.694660079981181e-05, "loss": 2.8684, "step": 3894 }, { "epoch": 0.31, "learning_rate": 9.694581667058732e-05, "loss": 3.2647, "step": 3895 }, { "epoch": 0.31, "learning_rate": 9.694503254136282e-05, "loss": 3.1005, "step": 3896 }, { "epoch": 0.31, "learning_rate": 9.694424841213832e-05, "loss": 3.2468, "step": 3897 }, { "epoch": 0.31, "learning_rate": 9.694346428291383e-05, "loss": 3.0066, "step": 3898 }, { "epoch": 0.31, "learning_rate": 9.694268015368933e-05, "loss": 3.0444, "step": 3899 }, { "epoch": 0.31, "learning_rate": 9.694189602446484e-05, "loss": 3.2471, "step": 3900 }, { "epoch": 0.31, "learning_rate": 9.694111189524034e-05, "loss": 2.9774, "step": 3901 }, { "epoch": 0.31, "learning_rate": 9.694032776601584e-05, "loss": 3.0183, "step": 3902 }, { "epoch": 0.31, "learning_rate": 9.693954363679135e-05, "loss": 2.9109, "step": 3903 }, { "epoch": 0.31, "learning_rate": 9.693875950756685e-05, "loss": 3.1362, "step": 3904 }, { "epoch": 0.31, "learning_rate": 9.693797537834235e-05, "loss": 2.9179, "step": 3905 }, { "epoch": 0.31, "learning_rate": 9.693719124911787e-05, "loss": 3.1907, "step": 3906 }, { "epoch": 0.31, "learning_rate": 9.693640711989336e-05, "loss": 2.8797, "step": 3907 }, { "epoch": 0.31, "learning_rate": 9.693562299066887e-05, "loss": 3.2507, "step": 3908 }, { "epoch": 0.31, "learning_rate": 9.693483886144437e-05, "loss": 2.6582, "step": 3909 }, { "epoch": 0.31, "learning_rate": 9.693405473221987e-05, "loss": 3.0758, "step": 3910 }, { "epoch": 0.31, "learning_rate": 9.693327060299538e-05, "loss": 2.9222, "step": 3911 }, { "epoch": 0.31, "learning_rate": 9.693248647377088e-05, "loss": 3.1962, "step": 3912 }, { "epoch": 0.31, "learning_rate": 9.693170234454638e-05, "loss": 3.2072, "step": 3913 }, { "epoch": 0.31, "learning_rate": 9.693091821532189e-05, "loss": 3.1841, "step": 3914 }, { "epoch": 0.31, "learning_rate": 9.693013408609739e-05, "loss": 3.4418, "step": 3915 }, { "epoch": 0.31, "learning_rate": 9.69293499568729e-05, "loss": 2.9997, "step": 3916 }, { "epoch": 0.31, "learning_rate": 9.69285658276484e-05, "loss": 2.9267, "step": 3917 }, { "epoch": 0.31, "learning_rate": 9.69277816984239e-05, "loss": 3.3379, "step": 3918 }, { "epoch": 0.31, "learning_rate": 9.69269975691994e-05, "loss": 3.0369, "step": 3919 }, { "epoch": 0.31, "learning_rate": 9.692621343997491e-05, "loss": 2.7614, "step": 3920 }, { "epoch": 0.31, "learning_rate": 9.692542931075041e-05, "loss": 3.0418, "step": 3921 }, { "epoch": 0.31, "learning_rate": 9.692464518152592e-05, "loss": 3.2516, "step": 3922 }, { "epoch": 0.31, "learning_rate": 9.692386105230142e-05, "loss": 3.0693, "step": 3923 }, { "epoch": 0.31, "learning_rate": 9.692307692307692e-05, "loss": 3.0124, "step": 3924 }, { "epoch": 0.31, "learning_rate": 9.692229279385243e-05, "loss": 3.0486, "step": 3925 }, { "epoch": 0.31, "learning_rate": 9.692150866462793e-05, "loss": 3.216, "step": 3926 }, { "epoch": 0.31, "learning_rate": 9.692072453540344e-05, "loss": 3.0056, "step": 3927 }, { "epoch": 0.31, "learning_rate": 9.691994040617894e-05, "loss": 3.0208, "step": 3928 }, { "epoch": 0.31, "learning_rate": 9.691915627695444e-05, "loss": 2.7631, "step": 3929 }, { "epoch": 0.31, "learning_rate": 9.691837214772995e-05, "loss": 3.0426, "step": 3930 }, { "epoch": 0.31, "learning_rate": 9.691758801850545e-05, "loss": 3.1735, "step": 3931 }, { "epoch": 0.31, "learning_rate": 9.691680388928097e-05, "loss": 3.2269, "step": 3932 }, { "epoch": 0.31, "learning_rate": 9.691601976005647e-05, "loss": 3.3108, "step": 3933 }, { "epoch": 0.31, "learning_rate": 9.691523563083197e-05, "loss": 3.1285, "step": 3934 }, { "epoch": 0.31, "learning_rate": 9.691445150160748e-05, "loss": 3.0887, "step": 3935 }, { "epoch": 0.31, "learning_rate": 9.691366737238298e-05, "loss": 3.2736, "step": 3936 }, { "epoch": 0.31, "learning_rate": 9.691288324315849e-05, "loss": 3.0504, "step": 3937 }, { "epoch": 0.31, "learning_rate": 9.691209911393399e-05, "loss": 3.0764, "step": 3938 }, { "epoch": 0.31, "learning_rate": 9.691131498470949e-05, "loss": 2.9719, "step": 3939 }, { "epoch": 0.31, "learning_rate": 9.6910530855485e-05, "loss": 2.8768, "step": 3940 }, { "epoch": 0.31, "learning_rate": 9.69097467262605e-05, "loss": 3.2315, "step": 3941 }, { "epoch": 0.31, "learning_rate": 9.6908962597036e-05, "loss": 3.1132, "step": 3942 }, { "epoch": 0.31, "learning_rate": 9.69081784678115e-05, "loss": 3.051, "step": 3943 }, { "epoch": 0.31, "learning_rate": 9.6907394338587e-05, "loss": 2.8903, "step": 3944 }, { "epoch": 0.31, "learning_rate": 9.69066102093625e-05, "loss": 3.0482, "step": 3945 }, { "epoch": 0.31, "learning_rate": 9.6905826080138e-05, "loss": 3.1533, "step": 3946 }, { "epoch": 0.31, "learning_rate": 9.690504195091351e-05, "loss": 2.9788, "step": 3947 }, { "epoch": 0.31, "learning_rate": 9.690425782168901e-05, "loss": 3.1319, "step": 3948 }, { "epoch": 0.31, "learning_rate": 9.690347369246452e-05, "loss": 3.3732, "step": 3949 }, { "epoch": 0.31, "learning_rate": 9.690268956324002e-05, "loss": 3.0261, "step": 3950 }, { "epoch": 0.31, "learning_rate": 9.690190543401552e-05, "loss": 3.1318, "step": 3951 }, { "epoch": 0.31, "learning_rate": 9.690112130479103e-05, "loss": 2.8911, "step": 3952 }, { "epoch": 0.31, "learning_rate": 9.690033717556653e-05, "loss": 3.0335, "step": 3953 }, { "epoch": 0.31, "learning_rate": 9.689955304634203e-05, "loss": 3.2479, "step": 3954 }, { "epoch": 0.31, "learning_rate": 9.689876891711754e-05, "loss": 3.1859, "step": 3955 }, { "epoch": 0.31, "learning_rate": 9.689798478789304e-05, "loss": 3.2194, "step": 3956 }, { "epoch": 0.31, "learning_rate": 9.689720065866856e-05, "loss": 3.0853, "step": 3957 }, { "epoch": 0.31, "learning_rate": 9.689641652944406e-05, "loss": 3.0663, "step": 3958 }, { "epoch": 0.31, "learning_rate": 9.689563240021957e-05, "loss": 2.9592, "step": 3959 }, { "epoch": 0.31, "learning_rate": 9.689484827099507e-05, "loss": 3.0229, "step": 3960 }, { "epoch": 0.31, "learning_rate": 9.689406414177057e-05, "loss": 2.8756, "step": 3961 }, { "epoch": 0.31, "learning_rate": 9.689328001254608e-05, "loss": 3.0624, "step": 3962 }, { "epoch": 0.31, "learning_rate": 9.689249588332158e-05, "loss": 2.9235, "step": 3963 }, { "epoch": 0.31, "learning_rate": 9.689171175409708e-05, "loss": 3.1022, "step": 3964 }, { "epoch": 0.31, "learning_rate": 9.689092762487259e-05, "loss": 3.1158, "step": 3965 }, { "epoch": 0.31, "learning_rate": 9.689014349564809e-05, "loss": 3.1474, "step": 3966 }, { "epoch": 0.31, "learning_rate": 9.68893593664236e-05, "loss": 3.3162, "step": 3967 }, { "epoch": 0.31, "learning_rate": 9.68885752371991e-05, "loss": 2.9163, "step": 3968 }, { "epoch": 0.31, "learning_rate": 9.68877911079746e-05, "loss": 3.1709, "step": 3969 }, { "epoch": 0.31, "learning_rate": 9.68870069787501e-05, "loss": 3.3089, "step": 3970 }, { "epoch": 0.31, "learning_rate": 9.688622284952561e-05, "loss": 3.1769, "step": 3971 }, { "epoch": 0.31, "learning_rate": 9.688543872030111e-05, "loss": 2.8593, "step": 3972 }, { "epoch": 0.31, "learning_rate": 9.688465459107662e-05, "loss": 3.0118, "step": 3973 }, { "epoch": 0.31, "learning_rate": 9.688387046185212e-05, "loss": 3.1925, "step": 3974 }, { "epoch": 0.31, "learning_rate": 9.688308633262763e-05, "loss": 3.1044, "step": 3975 }, { "epoch": 0.31, "learning_rate": 9.688230220340313e-05, "loss": 3.0158, "step": 3976 }, { "epoch": 0.31, "learning_rate": 9.688151807417863e-05, "loss": 3.4558, "step": 3977 }, { "epoch": 0.31, "learning_rate": 9.688073394495414e-05, "loss": 3.0843, "step": 3978 }, { "epoch": 0.31, "learning_rate": 9.687994981572963e-05, "loss": 2.9265, "step": 3979 }, { "epoch": 0.31, "learning_rate": 9.687916568650513e-05, "loss": 2.9932, "step": 3980 }, { "epoch": 0.31, "learning_rate": 9.687838155728063e-05, "loss": 2.9756, "step": 3981 }, { "epoch": 0.31, "learning_rate": 9.687759742805614e-05, "loss": 3.2058, "step": 3982 }, { "epoch": 0.31, "learning_rate": 9.687681329883165e-05, "loss": 2.9226, "step": 3983 }, { "epoch": 0.31, "learning_rate": 9.687602916960716e-05, "loss": 2.9605, "step": 3984 }, { "epoch": 0.31, "learning_rate": 9.687524504038266e-05, "loss": 2.898, "step": 3985 }, { "epoch": 0.31, "learning_rate": 9.687446091115817e-05, "loss": 2.9467, "step": 3986 }, { "epoch": 0.31, "learning_rate": 9.687367678193367e-05, "loss": 3.1249, "step": 3987 }, { "epoch": 0.31, "learning_rate": 9.687289265270917e-05, "loss": 2.9844, "step": 3988 }, { "epoch": 0.31, "learning_rate": 9.687210852348468e-05, "loss": 2.791, "step": 3989 }, { "epoch": 0.31, "learning_rate": 9.687132439426018e-05, "loss": 3.232, "step": 3990 }, { "epoch": 0.31, "learning_rate": 9.687054026503568e-05, "loss": 2.8182, "step": 3991 }, { "epoch": 0.31, "learning_rate": 9.686975613581119e-05, "loss": 3.2804, "step": 3992 }, { "epoch": 0.31, "learning_rate": 9.686897200658669e-05, "loss": 3.3393, "step": 3993 }, { "epoch": 0.31, "learning_rate": 9.68681878773622e-05, "loss": 3.2309, "step": 3994 }, { "epoch": 0.31, "learning_rate": 9.68674037481377e-05, "loss": 3.2253, "step": 3995 }, { "epoch": 0.31, "learning_rate": 9.68666196189132e-05, "loss": 2.9768, "step": 3996 }, { "epoch": 0.31, "learning_rate": 9.68658354896887e-05, "loss": 2.8912, "step": 3997 }, { "epoch": 0.31, "learning_rate": 9.686505136046421e-05, "loss": 2.941, "step": 3998 }, { "epoch": 0.31, "learning_rate": 9.686426723123971e-05, "loss": 2.9073, "step": 3999 }, { "epoch": 0.31, "learning_rate": 9.686348310201522e-05, "loss": 3.1587, "step": 4000 }, { "epoch": 0.31, "eval_loss": 2.982861280441284, "eval_runtime": 218.0222, "eval_samples_per_second": 51.976, "eval_steps_per_second": 3.252, "step": 4000 }, { "epoch": 0.31, "learning_rate": 9.686269897279072e-05, "loss": 3.0675, "step": 4001 }, { "epoch": 0.31, "learning_rate": 9.686191484356622e-05, "loss": 2.918, "step": 4002 }, { "epoch": 0.31, "learning_rate": 9.686113071434173e-05, "loss": 3.0371, "step": 4003 }, { "epoch": 0.31, "learning_rate": 9.686034658511723e-05, "loss": 3.0458, "step": 4004 }, { "epoch": 0.31, "learning_rate": 9.685956245589274e-05, "loss": 3.1068, "step": 4005 }, { "epoch": 0.31, "learning_rate": 9.685877832666824e-05, "loss": 3.0913, "step": 4006 }, { "epoch": 0.31, "learning_rate": 9.685799419744374e-05, "loss": 2.9711, "step": 4007 }, { "epoch": 0.31, "learning_rate": 9.685721006821925e-05, "loss": 3.1215, "step": 4008 }, { "epoch": 0.31, "learning_rate": 9.685642593899475e-05, "loss": 3.3922, "step": 4009 }, { "epoch": 0.31, "learning_rate": 9.685564180977025e-05, "loss": 3.1925, "step": 4010 }, { "epoch": 0.31, "learning_rate": 9.685485768054576e-05, "loss": 3.2307, "step": 4011 }, { "epoch": 0.31, "learning_rate": 9.685407355132126e-05, "loss": 2.9652, "step": 4012 }, { "epoch": 0.31, "learning_rate": 9.685328942209676e-05, "loss": 2.8924, "step": 4013 }, { "epoch": 0.31, "learning_rate": 9.685250529287228e-05, "loss": 2.8685, "step": 4014 }, { "epoch": 0.31, "learning_rate": 9.685172116364777e-05, "loss": 3.0679, "step": 4015 }, { "epoch": 0.31, "learning_rate": 9.685093703442328e-05, "loss": 3.1164, "step": 4016 }, { "epoch": 0.31, "learning_rate": 9.685015290519878e-05, "loss": 3.4, "step": 4017 }, { "epoch": 0.32, "learning_rate": 9.684936877597428e-05, "loss": 2.977, "step": 4018 }, { "epoch": 0.32, "learning_rate": 9.684858464674979e-05, "loss": 3.0607, "step": 4019 }, { "epoch": 0.32, "learning_rate": 9.684780051752529e-05, "loss": 2.8459, "step": 4020 }, { "epoch": 0.32, "learning_rate": 9.68470163883008e-05, "loss": 3.1132, "step": 4021 }, { "epoch": 0.32, "learning_rate": 9.68462322590763e-05, "loss": 3.0576, "step": 4022 }, { "epoch": 0.32, "learning_rate": 9.68454481298518e-05, "loss": 3.2596, "step": 4023 }, { "epoch": 0.32, "learning_rate": 9.68446640006273e-05, "loss": 3.2339, "step": 4024 }, { "epoch": 0.32, "learning_rate": 9.684387987140281e-05, "loss": 3.321, "step": 4025 }, { "epoch": 0.32, "learning_rate": 9.684309574217831e-05, "loss": 3.1627, "step": 4026 }, { "epoch": 0.32, "learning_rate": 9.684231161295382e-05, "loss": 3.2863, "step": 4027 }, { "epoch": 0.32, "learning_rate": 9.684152748372932e-05, "loss": 3.1731, "step": 4028 }, { "epoch": 0.32, "learning_rate": 9.684074335450482e-05, "loss": 2.9901, "step": 4029 }, { "epoch": 0.32, "learning_rate": 9.683995922528033e-05, "loss": 3.1382, "step": 4030 }, { "epoch": 0.32, "learning_rate": 9.683917509605583e-05, "loss": 3.3485, "step": 4031 }, { "epoch": 0.32, "learning_rate": 9.683839096683133e-05, "loss": 3.1574, "step": 4032 }, { "epoch": 0.32, "learning_rate": 9.683760683760684e-05, "loss": 2.973, "step": 4033 }, { "epoch": 0.32, "learning_rate": 9.683682270838234e-05, "loss": 2.9367, "step": 4034 }, { "epoch": 0.32, "learning_rate": 9.683603857915785e-05, "loss": 2.9975, "step": 4035 }, { "epoch": 0.32, "learning_rate": 9.683525444993335e-05, "loss": 3.1032, "step": 4036 }, { "epoch": 0.32, "learning_rate": 9.683447032070885e-05, "loss": 3.0904, "step": 4037 }, { "epoch": 0.32, "learning_rate": 9.683368619148436e-05, "loss": 2.8347, "step": 4038 }, { "epoch": 0.32, "learning_rate": 9.683290206225986e-05, "loss": 3.1362, "step": 4039 }, { "epoch": 0.32, "learning_rate": 9.683211793303538e-05, "loss": 3.4685, "step": 4040 }, { "epoch": 0.32, "learning_rate": 9.683133380381088e-05, "loss": 2.9128, "step": 4041 }, { "epoch": 0.32, "learning_rate": 9.683054967458639e-05, "loss": 3.0796, "step": 4042 }, { "epoch": 0.32, "learning_rate": 9.682976554536189e-05, "loss": 3.1586, "step": 4043 }, { "epoch": 0.32, "learning_rate": 9.682898141613739e-05, "loss": 3.3512, "step": 4044 }, { "epoch": 0.32, "learning_rate": 9.68281972869129e-05, "loss": 3.0127, "step": 4045 }, { "epoch": 0.32, "learning_rate": 9.68274131576884e-05, "loss": 3.2654, "step": 4046 }, { "epoch": 0.32, "learning_rate": 9.68266290284639e-05, "loss": 3.0313, "step": 4047 }, { "epoch": 0.32, "learning_rate": 9.682584489923941e-05, "loss": 3.0836, "step": 4048 }, { "epoch": 0.32, "learning_rate": 9.682506077001491e-05, "loss": 3.2863, "step": 4049 }, { "epoch": 0.32, "learning_rate": 9.682427664079041e-05, "loss": 3.046, "step": 4050 }, { "epoch": 0.32, "learning_rate": 9.68234925115659e-05, "loss": 2.8681, "step": 4051 }, { "epoch": 0.32, "learning_rate": 9.682270838234141e-05, "loss": 2.9665, "step": 4052 }, { "epoch": 0.32, "learning_rate": 9.682192425311691e-05, "loss": 3.0093, "step": 4053 }, { "epoch": 0.32, "learning_rate": 9.682114012389242e-05, "loss": 3.1235, "step": 4054 }, { "epoch": 0.32, "learning_rate": 9.682035599466792e-05, "loss": 3.1173, "step": 4055 }, { "epoch": 0.32, "learning_rate": 9.681957186544342e-05, "loss": 3.1539, "step": 4056 }, { "epoch": 0.32, "learning_rate": 9.681878773621893e-05, "loss": 2.9871, "step": 4057 }, { "epoch": 0.32, "learning_rate": 9.681800360699443e-05, "loss": 3.3049, "step": 4058 }, { "epoch": 0.32, "learning_rate": 9.681721947776993e-05, "loss": 3.0964, "step": 4059 }, { "epoch": 0.32, "learning_rate": 9.681643534854544e-05, "loss": 3.021, "step": 4060 }, { "epoch": 0.32, "learning_rate": 9.681565121932094e-05, "loss": 3.0405, "step": 4061 }, { "epoch": 0.32, "learning_rate": 9.681486709009645e-05, "loss": 3.101, "step": 4062 }, { "epoch": 0.32, "learning_rate": 9.681408296087195e-05, "loss": 2.8254, "step": 4063 }, { "epoch": 0.32, "learning_rate": 9.681329883164745e-05, "loss": 2.9493, "step": 4064 }, { "epoch": 0.32, "learning_rate": 9.681251470242296e-05, "loss": 3.1549, "step": 4065 }, { "epoch": 0.32, "learning_rate": 9.681173057319847e-05, "loss": 2.9182, "step": 4066 }, { "epoch": 0.32, "learning_rate": 9.681094644397398e-05, "loss": 3.0846, "step": 4067 }, { "epoch": 0.32, "learning_rate": 9.681016231474948e-05, "loss": 3.2502, "step": 4068 }, { "epoch": 0.32, "learning_rate": 9.680937818552498e-05, "loss": 3.0497, "step": 4069 }, { "epoch": 0.32, "learning_rate": 9.680859405630049e-05, "loss": 3.0645, "step": 4070 }, { "epoch": 0.32, "learning_rate": 9.680780992707599e-05, "loss": 3.3103, "step": 4071 }, { "epoch": 0.32, "learning_rate": 9.68070257978515e-05, "loss": 3.1863, "step": 4072 }, { "epoch": 0.32, "learning_rate": 9.6806241668627e-05, "loss": 2.8751, "step": 4073 }, { "epoch": 0.32, "learning_rate": 9.68054575394025e-05, "loss": 2.9726, "step": 4074 }, { "epoch": 0.32, "learning_rate": 9.6804673410178e-05, "loss": 3.5093, "step": 4075 }, { "epoch": 0.32, "learning_rate": 9.680388928095351e-05, "loss": 3.0559, "step": 4076 }, { "epoch": 0.32, "learning_rate": 9.680310515172901e-05, "loss": 3.0646, "step": 4077 }, { "epoch": 0.32, "learning_rate": 9.680232102250452e-05, "loss": 3.1991, "step": 4078 }, { "epoch": 0.32, "learning_rate": 9.680153689328002e-05, "loss": 3.22, "step": 4079 }, { "epoch": 0.32, "learning_rate": 9.680075276405552e-05, "loss": 3.0406, "step": 4080 }, { "epoch": 0.32, "learning_rate": 9.679996863483103e-05, "loss": 2.9599, "step": 4081 }, { "epoch": 0.32, "learning_rate": 9.679918450560653e-05, "loss": 3.1137, "step": 4082 }, { "epoch": 0.32, "learning_rate": 9.679840037638204e-05, "loss": 3.1963, "step": 4083 }, { "epoch": 0.32, "learning_rate": 9.679761624715754e-05, "loss": 3.2615, "step": 4084 }, { "epoch": 0.32, "learning_rate": 9.679683211793304e-05, "loss": 3.1348, "step": 4085 }, { "epoch": 0.32, "learning_rate": 9.679604798870855e-05, "loss": 3.1111, "step": 4086 }, { "epoch": 0.32, "learning_rate": 9.679526385948404e-05, "loss": 3.0946, "step": 4087 }, { "epoch": 0.32, "learning_rate": 9.679447973025954e-05, "loss": 3.0364, "step": 4088 }, { "epoch": 0.32, "learning_rate": 9.679369560103504e-05, "loss": 3.0914, "step": 4089 }, { "epoch": 0.32, "learning_rate": 9.679291147181055e-05, "loss": 3.1649, "step": 4090 }, { "epoch": 0.32, "learning_rate": 9.679212734258607e-05, "loss": 3.2874, "step": 4091 }, { "epoch": 0.32, "learning_rate": 9.679134321336157e-05, "loss": 3.0628, "step": 4092 }, { "epoch": 0.32, "learning_rate": 9.679055908413707e-05, "loss": 2.8337, "step": 4093 }, { "epoch": 0.32, "learning_rate": 9.678977495491258e-05, "loss": 3.181, "step": 4094 }, { "epoch": 0.32, "learning_rate": 9.678899082568808e-05, "loss": 3.2914, "step": 4095 }, { "epoch": 0.32, "learning_rate": 9.678820669646358e-05, "loss": 2.9743, "step": 4096 }, { "epoch": 0.32, "learning_rate": 9.678742256723909e-05, "loss": 3.1741, "step": 4097 }, { "epoch": 0.32, "learning_rate": 9.678663843801459e-05, "loss": 3.2515, "step": 4098 }, { "epoch": 0.32, "learning_rate": 9.67858543087901e-05, "loss": 3.0902, "step": 4099 }, { "epoch": 0.32, "learning_rate": 9.67850701795656e-05, "loss": 3.0561, "step": 4100 }, { "epoch": 0.32, "learning_rate": 9.67842860503411e-05, "loss": 3.0715, "step": 4101 }, { "epoch": 0.32, "learning_rate": 9.67835019211166e-05, "loss": 2.9697, "step": 4102 }, { "epoch": 0.32, "learning_rate": 9.678271779189211e-05, "loss": 3.0587, "step": 4103 }, { "epoch": 0.32, "learning_rate": 9.678193366266761e-05, "loss": 3.0237, "step": 4104 }, { "epoch": 0.32, "learning_rate": 9.678114953344312e-05, "loss": 3.0508, "step": 4105 }, { "epoch": 0.32, "learning_rate": 9.678036540421862e-05, "loss": 3.0293, "step": 4106 }, { "epoch": 0.32, "learning_rate": 9.677958127499412e-05, "loss": 2.8616, "step": 4107 }, { "epoch": 0.32, "learning_rate": 9.677879714576963e-05, "loss": 3.1639, "step": 4108 }, { "epoch": 0.32, "learning_rate": 9.677801301654513e-05, "loss": 3.3756, "step": 4109 }, { "epoch": 0.32, "learning_rate": 9.677722888732064e-05, "loss": 3.2161, "step": 4110 }, { "epoch": 0.32, "learning_rate": 9.677644475809614e-05, "loss": 3.2398, "step": 4111 }, { "epoch": 0.32, "learning_rate": 9.677566062887164e-05, "loss": 2.9261, "step": 4112 }, { "epoch": 0.32, "learning_rate": 9.677487649964715e-05, "loss": 3.0514, "step": 4113 }, { "epoch": 0.32, "learning_rate": 9.677409237042265e-05, "loss": 3.0701, "step": 4114 }, { "epoch": 0.32, "learning_rate": 9.677330824119815e-05, "loss": 2.9905, "step": 4115 }, { "epoch": 0.32, "learning_rate": 9.677252411197366e-05, "loss": 3.0815, "step": 4116 }, { "epoch": 0.32, "learning_rate": 9.677173998274916e-05, "loss": 3.0553, "step": 4117 }, { "epoch": 0.32, "learning_rate": 9.677095585352466e-05, "loss": 3.0496, "step": 4118 }, { "epoch": 0.32, "learning_rate": 9.677017172430017e-05, "loss": 3.3421, "step": 4119 }, { "epoch": 0.32, "learning_rate": 9.676938759507567e-05, "loss": 2.9296, "step": 4120 }, { "epoch": 0.32, "learning_rate": 9.676860346585118e-05, "loss": 3.2418, "step": 4121 }, { "epoch": 0.32, "learning_rate": 9.676781933662669e-05, "loss": 3.1139, "step": 4122 }, { "epoch": 0.32, "learning_rate": 9.676703520740218e-05, "loss": 3.0071, "step": 4123 }, { "epoch": 0.32, "learning_rate": 9.676625107817769e-05, "loss": 3.0302, "step": 4124 }, { "epoch": 0.32, "learning_rate": 9.676546694895319e-05, "loss": 3.1173, "step": 4125 }, { "epoch": 0.32, "learning_rate": 9.67646828197287e-05, "loss": 3.3184, "step": 4126 }, { "epoch": 0.32, "learning_rate": 9.67638986905042e-05, "loss": 3.1677, "step": 4127 }, { "epoch": 0.32, "learning_rate": 9.67631145612797e-05, "loss": 3.3169, "step": 4128 }, { "epoch": 0.32, "learning_rate": 9.67623304320552e-05, "loss": 3.1032, "step": 4129 }, { "epoch": 0.32, "learning_rate": 9.676154630283071e-05, "loss": 3.1339, "step": 4130 }, { "epoch": 0.32, "learning_rate": 9.676076217360621e-05, "loss": 2.8565, "step": 4131 }, { "epoch": 0.32, "learning_rate": 9.675997804438172e-05, "loss": 2.8549, "step": 4132 }, { "epoch": 0.32, "learning_rate": 9.675919391515722e-05, "loss": 3.2476, "step": 4133 }, { "epoch": 0.32, "learning_rate": 9.675840978593272e-05, "loss": 3.1009, "step": 4134 }, { "epoch": 0.32, "learning_rate": 9.675762565670823e-05, "loss": 2.8527, "step": 4135 }, { "epoch": 0.32, "learning_rate": 9.675684152748373e-05, "loss": 3.2394, "step": 4136 }, { "epoch": 0.32, "learning_rate": 9.675605739825923e-05, "loss": 2.9487, "step": 4137 }, { "epoch": 0.32, "learning_rate": 9.675527326903474e-05, "loss": 3.1809, "step": 4138 }, { "epoch": 0.32, "learning_rate": 9.675448913981024e-05, "loss": 3.0729, "step": 4139 }, { "epoch": 0.32, "learning_rate": 9.675370501058575e-05, "loss": 3.2905, "step": 4140 }, { "epoch": 0.32, "learning_rate": 9.675292088136125e-05, "loss": 3.0408, "step": 4141 }, { "epoch": 0.32, "learning_rate": 9.675213675213675e-05, "loss": 3.1057, "step": 4142 }, { "epoch": 0.32, "learning_rate": 9.675135262291226e-05, "loss": 2.9767, "step": 4143 }, { "epoch": 0.32, "learning_rate": 9.675056849368776e-05, "loss": 3.0832, "step": 4144 }, { "epoch": 0.33, "learning_rate": 9.674978436446326e-05, "loss": 3.1189, "step": 4145 }, { "epoch": 0.33, "learning_rate": 9.674900023523877e-05, "loss": 2.9078, "step": 4146 }, { "epoch": 0.33, "learning_rate": 9.674821610601427e-05, "loss": 3.3212, "step": 4147 }, { "epoch": 0.33, "learning_rate": 9.674743197678979e-05, "loss": 3.1591, "step": 4148 }, { "epoch": 0.33, "learning_rate": 9.674664784756529e-05, "loss": 3.0518, "step": 4149 }, { "epoch": 0.33, "learning_rate": 9.67458637183408e-05, "loss": 3.0832, "step": 4150 }, { "epoch": 0.33, "learning_rate": 9.67450795891163e-05, "loss": 3.0293, "step": 4151 }, { "epoch": 0.33, "learning_rate": 9.67442954598918e-05, "loss": 3.2801, "step": 4152 }, { "epoch": 0.33, "learning_rate": 9.674351133066731e-05, "loss": 2.9858, "step": 4153 }, { "epoch": 0.33, "learning_rate": 9.674272720144281e-05, "loss": 3.0661, "step": 4154 }, { "epoch": 0.33, "learning_rate": 9.674194307221831e-05, "loss": 2.9381, "step": 4155 }, { "epoch": 0.33, "learning_rate": 9.674115894299382e-05, "loss": 2.8091, "step": 4156 }, { "epoch": 0.33, "learning_rate": 9.674037481376932e-05, "loss": 2.8968, "step": 4157 }, { "epoch": 0.33, "learning_rate": 9.673959068454483e-05, "loss": 3.1936, "step": 4158 }, { "epoch": 0.33, "learning_rate": 9.673880655532032e-05, "loss": 2.9726, "step": 4159 }, { "epoch": 0.33, "learning_rate": 9.673802242609582e-05, "loss": 3.1743, "step": 4160 }, { "epoch": 0.33, "learning_rate": 9.673723829687132e-05, "loss": 3.1594, "step": 4161 }, { "epoch": 0.33, "learning_rate": 9.673645416764683e-05, "loss": 2.8032, "step": 4162 }, { "epoch": 0.33, "learning_rate": 9.673567003842233e-05, "loss": 3.461, "step": 4163 }, { "epoch": 0.33, "learning_rate": 9.673488590919783e-05, "loss": 2.8991, "step": 4164 }, { "epoch": 0.33, "learning_rate": 9.673410177997334e-05, "loss": 3.0616, "step": 4165 }, { "epoch": 0.33, "learning_rate": 9.673331765074884e-05, "loss": 3.1899, "step": 4166 }, { "epoch": 0.33, "learning_rate": 9.673253352152434e-05, "loss": 3.3315, "step": 4167 }, { "epoch": 0.33, "learning_rate": 9.673174939229985e-05, "loss": 2.9661, "step": 4168 }, { "epoch": 0.33, "learning_rate": 9.673096526307535e-05, "loss": 3.3782, "step": 4169 }, { "epoch": 0.33, "learning_rate": 9.673018113385086e-05, "loss": 3.1747, "step": 4170 }, { "epoch": 0.33, "learning_rate": 9.672939700462636e-05, "loss": 3.142, "step": 4171 }, { "epoch": 0.33, "learning_rate": 9.672861287540186e-05, "loss": 3.49, "step": 4172 }, { "epoch": 0.33, "learning_rate": 9.672782874617737e-05, "loss": 3.1864, "step": 4173 }, { "epoch": 0.33, "learning_rate": 9.672704461695288e-05, "loss": 3.2844, "step": 4174 }, { "epoch": 0.33, "learning_rate": 9.672626048772839e-05, "loss": 3.176, "step": 4175 }, { "epoch": 0.33, "learning_rate": 9.672547635850389e-05, "loss": 3.1322, "step": 4176 }, { "epoch": 0.33, "learning_rate": 9.67246922292794e-05, "loss": 3.056, "step": 4177 }, { "epoch": 0.33, "learning_rate": 9.67239081000549e-05, "loss": 3.0685, "step": 4178 }, { "epoch": 0.33, "learning_rate": 9.67231239708304e-05, "loss": 3.1423, "step": 4179 }, { "epoch": 0.33, "learning_rate": 9.67223398416059e-05, "loss": 3.1047, "step": 4180 }, { "epoch": 0.33, "learning_rate": 9.672155571238141e-05, "loss": 3.3343, "step": 4181 }, { "epoch": 0.33, "learning_rate": 9.672077158315691e-05, "loss": 2.9965, "step": 4182 }, { "epoch": 0.33, "learning_rate": 9.671998745393242e-05, "loss": 3.0121, "step": 4183 }, { "epoch": 0.33, "learning_rate": 9.671920332470792e-05, "loss": 3.0322, "step": 4184 }, { "epoch": 0.33, "learning_rate": 9.671841919548342e-05, "loss": 3.0043, "step": 4185 }, { "epoch": 0.33, "learning_rate": 9.671763506625893e-05, "loss": 3.0781, "step": 4186 }, { "epoch": 0.33, "learning_rate": 9.671685093703443e-05, "loss": 3.034, "step": 4187 }, { "epoch": 0.33, "learning_rate": 9.671606680780994e-05, "loss": 3.046, "step": 4188 }, { "epoch": 0.33, "learning_rate": 9.671528267858544e-05, "loss": 3.2058, "step": 4189 }, { "epoch": 0.33, "learning_rate": 9.671449854936094e-05, "loss": 3.1956, "step": 4190 }, { "epoch": 0.33, "learning_rate": 9.671371442013645e-05, "loss": 3.094, "step": 4191 }, { "epoch": 0.33, "learning_rate": 9.671293029091195e-05, "loss": 3.1474, "step": 4192 }, { "epoch": 0.33, "learning_rate": 9.671214616168745e-05, "loss": 2.8941, "step": 4193 }, { "epoch": 0.33, "learning_rate": 9.671136203246296e-05, "loss": 2.8155, "step": 4194 }, { "epoch": 0.33, "learning_rate": 9.671057790323845e-05, "loss": 2.8634, "step": 4195 }, { "epoch": 0.33, "learning_rate": 9.670979377401395e-05, "loss": 3.5202, "step": 4196 }, { "epoch": 0.33, "learning_rate": 9.670900964478946e-05, "loss": 3.2487, "step": 4197 }, { "epoch": 0.33, "learning_rate": 9.670822551556496e-05, "loss": 3.1205, "step": 4198 }, { "epoch": 0.33, "learning_rate": 9.670744138634048e-05, "loss": 3.0342, "step": 4199 }, { "epoch": 0.33, "learning_rate": 9.670665725711598e-05, "loss": 3.3386, "step": 4200 }, { "epoch": 0.33, "learning_rate": 9.670587312789148e-05, "loss": 3.1515, "step": 4201 }, { "epoch": 0.33, "learning_rate": 9.670508899866699e-05, "loss": 3.3067, "step": 4202 }, { "epoch": 0.33, "learning_rate": 9.670430486944249e-05, "loss": 3.0627, "step": 4203 }, { "epoch": 0.33, "learning_rate": 9.6703520740218e-05, "loss": 3.0926, "step": 4204 }, { "epoch": 0.33, "learning_rate": 9.67027366109935e-05, "loss": 2.9982, "step": 4205 }, { "epoch": 0.33, "learning_rate": 9.6701952481769e-05, "loss": 3.1569, "step": 4206 }, { "epoch": 0.33, "learning_rate": 9.67011683525445e-05, "loss": 3.1435, "step": 4207 }, { "epoch": 0.33, "learning_rate": 9.670038422332001e-05, "loss": 3.424, "step": 4208 }, { "epoch": 0.33, "learning_rate": 9.669960009409551e-05, "loss": 3.331, "step": 4209 }, { "epoch": 0.33, "learning_rate": 9.669881596487102e-05, "loss": 2.9709, "step": 4210 }, { "epoch": 0.33, "learning_rate": 9.669803183564652e-05, "loss": 3.1413, "step": 4211 }, { "epoch": 0.33, "learning_rate": 9.669724770642202e-05, "loss": 3.0812, "step": 4212 }, { "epoch": 0.33, "learning_rate": 9.669646357719753e-05, "loss": 3.1509, "step": 4213 }, { "epoch": 0.33, "learning_rate": 9.669567944797303e-05, "loss": 3.03, "step": 4214 }, { "epoch": 0.33, "learning_rate": 9.669489531874853e-05, "loss": 2.9186, "step": 4215 }, { "epoch": 0.33, "learning_rate": 9.669411118952404e-05, "loss": 3.202, "step": 4216 }, { "epoch": 0.33, "learning_rate": 9.669332706029954e-05, "loss": 3.1288, "step": 4217 }, { "epoch": 0.33, "learning_rate": 9.669254293107505e-05, "loss": 3.0814, "step": 4218 }, { "epoch": 0.33, "learning_rate": 9.669175880185055e-05, "loss": 2.931, "step": 4219 }, { "epoch": 0.33, "learning_rate": 9.669097467262605e-05, "loss": 3.0033, "step": 4220 }, { "epoch": 0.33, "learning_rate": 9.669019054340156e-05, "loss": 2.9809, "step": 4221 }, { "epoch": 0.33, "learning_rate": 9.668940641417706e-05, "loss": 3.0425, "step": 4222 }, { "epoch": 0.33, "learning_rate": 9.668862228495256e-05, "loss": 3.1452, "step": 4223 }, { "epoch": 0.33, "learning_rate": 9.668783815572807e-05, "loss": 3.331, "step": 4224 }, { "epoch": 0.33, "learning_rate": 9.668705402650357e-05, "loss": 3.0283, "step": 4225 }, { "epoch": 0.33, "learning_rate": 9.668626989727908e-05, "loss": 2.9408, "step": 4226 }, { "epoch": 0.33, "learning_rate": 9.668548576805458e-05, "loss": 3.1648, "step": 4227 }, { "epoch": 0.33, "learning_rate": 9.668470163883008e-05, "loss": 3.184, "step": 4228 }, { "epoch": 0.33, "learning_rate": 9.668391750960559e-05, "loss": 3.0855, "step": 4229 }, { "epoch": 0.33, "learning_rate": 9.66831333803811e-05, "loss": 3.2106, "step": 4230 }, { "epoch": 0.33, "learning_rate": 9.66823492511566e-05, "loss": 2.8448, "step": 4231 }, { "epoch": 0.33, "learning_rate": 9.66815651219321e-05, "loss": 2.7763, "step": 4232 }, { "epoch": 0.33, "learning_rate": 9.66807809927076e-05, "loss": 3.0121, "step": 4233 }, { "epoch": 0.33, "learning_rate": 9.66799968634831e-05, "loss": 2.9007, "step": 4234 }, { "epoch": 0.33, "learning_rate": 9.667921273425861e-05, "loss": 3.2752, "step": 4235 }, { "epoch": 0.33, "learning_rate": 9.667842860503411e-05, "loss": 3.04, "step": 4236 }, { "epoch": 0.33, "learning_rate": 9.667764447580962e-05, "loss": 3.1616, "step": 4237 }, { "epoch": 0.33, "learning_rate": 9.667686034658512e-05, "loss": 2.9274, "step": 4238 }, { "epoch": 0.33, "learning_rate": 9.667607621736062e-05, "loss": 3.1498, "step": 4239 }, { "epoch": 0.33, "learning_rate": 9.667529208813613e-05, "loss": 2.9089, "step": 4240 }, { "epoch": 0.33, "learning_rate": 9.667450795891163e-05, "loss": 3.006, "step": 4241 }, { "epoch": 0.33, "learning_rate": 9.667372382968713e-05, "loss": 3.0808, "step": 4242 }, { "epoch": 0.33, "learning_rate": 9.667293970046264e-05, "loss": 3.1886, "step": 4243 }, { "epoch": 0.33, "learning_rate": 9.667215557123814e-05, "loss": 2.9286, "step": 4244 }, { "epoch": 0.33, "learning_rate": 9.667137144201365e-05, "loss": 2.996, "step": 4245 }, { "epoch": 0.33, "learning_rate": 9.667058731278915e-05, "loss": 2.9945, "step": 4246 }, { "epoch": 0.33, "learning_rate": 9.666980318356465e-05, "loss": 2.9422, "step": 4247 }, { "epoch": 0.33, "learning_rate": 9.666901905434016e-05, "loss": 2.988, "step": 4248 }, { "epoch": 0.33, "learning_rate": 9.666823492511566e-05, "loss": 3.0896, "step": 4249 }, { "epoch": 0.33, "learning_rate": 9.666745079589116e-05, "loss": 3.0067, "step": 4250 }, { "epoch": 0.33, "learning_rate": 9.666666666666667e-05, "loss": 2.8335, "step": 4251 }, { "epoch": 0.33, "learning_rate": 9.666588253744217e-05, "loss": 3.1244, "step": 4252 }, { "epoch": 0.33, "learning_rate": 9.666509840821767e-05, "loss": 3.1477, "step": 4253 }, { "epoch": 0.33, "learning_rate": 9.666431427899318e-05, "loss": 2.9443, "step": 4254 }, { "epoch": 0.33, "learning_rate": 9.666353014976868e-05, "loss": 3.3087, "step": 4255 }, { "epoch": 0.33, "learning_rate": 9.66627460205442e-05, "loss": 3.2057, "step": 4256 }, { "epoch": 0.33, "learning_rate": 9.66619618913197e-05, "loss": 3.2001, "step": 4257 }, { "epoch": 0.33, "learning_rate": 9.66611777620952e-05, "loss": 3.3544, "step": 4258 }, { "epoch": 0.33, "learning_rate": 9.666039363287071e-05, "loss": 3.1796, "step": 4259 }, { "epoch": 0.33, "learning_rate": 9.665960950364621e-05, "loss": 2.9843, "step": 4260 }, { "epoch": 0.33, "learning_rate": 9.665882537442172e-05, "loss": 2.9969, "step": 4261 }, { "epoch": 0.33, "learning_rate": 9.665804124519722e-05, "loss": 3.462, "step": 4262 }, { "epoch": 0.33, "learning_rate": 9.665725711597272e-05, "loss": 2.9668, "step": 4263 }, { "epoch": 0.33, "learning_rate": 9.665647298674823e-05, "loss": 3.4073, "step": 4264 }, { "epoch": 0.33, "learning_rate": 9.665568885752373e-05, "loss": 2.8424, "step": 4265 }, { "epoch": 0.33, "learning_rate": 9.665490472829924e-05, "loss": 3.187, "step": 4266 }, { "epoch": 0.33, "learning_rate": 9.665412059907473e-05, "loss": 3.1407, "step": 4267 }, { "epoch": 0.33, "learning_rate": 9.665333646985023e-05, "loss": 3.0438, "step": 4268 }, { "epoch": 0.33, "learning_rate": 9.665255234062573e-05, "loss": 2.8681, "step": 4269 }, { "epoch": 0.33, "learning_rate": 9.665176821140124e-05, "loss": 3.0212, "step": 4270 }, { "epoch": 0.33, "learning_rate": 9.665098408217674e-05, "loss": 3.265, "step": 4271 }, { "epoch": 0.33, "learning_rate": 9.665019995295224e-05, "loss": 3.1615, "step": 4272 }, { "epoch": 0.34, "learning_rate": 9.664941582372775e-05, "loss": 2.9126, "step": 4273 }, { "epoch": 0.34, "learning_rate": 9.664863169450325e-05, "loss": 3.1814, "step": 4274 }, { "epoch": 0.34, "learning_rate": 9.664784756527876e-05, "loss": 3.0264, "step": 4275 }, { "epoch": 0.34, "learning_rate": 9.664706343605426e-05, "loss": 3.1443, "step": 4276 }, { "epoch": 0.34, "learning_rate": 9.664627930682976e-05, "loss": 3.1449, "step": 4277 }, { "epoch": 0.34, "learning_rate": 9.664549517760527e-05, "loss": 3.0321, "step": 4278 }, { "epoch": 0.34, "learning_rate": 9.664471104838077e-05, "loss": 3.2793, "step": 4279 }, { "epoch": 0.34, "learning_rate": 9.664392691915627e-05, "loss": 2.982, "step": 4280 }, { "epoch": 0.34, "learning_rate": 9.664314278993178e-05, "loss": 3.3235, "step": 4281 }, { "epoch": 0.34, "learning_rate": 9.66423586607073e-05, "loss": 2.9201, "step": 4282 }, { "epoch": 0.34, "learning_rate": 9.66415745314828e-05, "loss": 3.1737, "step": 4283 }, { "epoch": 0.34, "learning_rate": 9.66407904022583e-05, "loss": 3.0965, "step": 4284 }, { "epoch": 0.34, "learning_rate": 9.66400062730338e-05, "loss": 3.0179, "step": 4285 }, { "epoch": 0.34, "learning_rate": 9.663922214380931e-05, "loss": 2.9595, "step": 4286 }, { "epoch": 0.34, "learning_rate": 9.663843801458481e-05, "loss": 2.9371, "step": 4287 }, { "epoch": 0.34, "learning_rate": 9.663765388536032e-05, "loss": 2.9227, "step": 4288 }, { "epoch": 0.34, "learning_rate": 9.663686975613582e-05, "loss": 3.0001, "step": 4289 }, { "epoch": 0.34, "learning_rate": 9.663608562691132e-05, "loss": 3.0262, "step": 4290 }, { "epoch": 0.34, "learning_rate": 9.663530149768683e-05, "loss": 3.0043, "step": 4291 }, { "epoch": 0.34, "learning_rate": 9.663451736846233e-05, "loss": 3.2149, "step": 4292 }, { "epoch": 0.34, "learning_rate": 9.663373323923784e-05, "loss": 3.0222, "step": 4293 }, { "epoch": 0.34, "learning_rate": 9.663294911001334e-05, "loss": 3.055, "step": 4294 }, { "epoch": 0.34, "learning_rate": 9.663216498078884e-05, "loss": 3.1833, "step": 4295 }, { "epoch": 0.34, "learning_rate": 9.663138085156435e-05, "loss": 2.8882, "step": 4296 }, { "epoch": 0.34, "learning_rate": 9.663059672233985e-05, "loss": 2.871, "step": 4297 }, { "epoch": 0.34, "learning_rate": 9.662981259311535e-05, "loss": 3.1467, "step": 4298 }, { "epoch": 0.34, "learning_rate": 9.662902846389086e-05, "loss": 3.2375, "step": 4299 }, { "epoch": 0.34, "learning_rate": 9.662824433466636e-05, "loss": 2.925, "step": 4300 }, { "epoch": 0.34, "learning_rate": 9.662746020544186e-05, "loss": 3.1067, "step": 4301 }, { "epoch": 0.34, "learning_rate": 9.662667607621737e-05, "loss": 3.2708, "step": 4302 }, { "epoch": 0.34, "learning_rate": 9.662589194699286e-05, "loss": 3.0102, "step": 4303 }, { "epoch": 0.34, "learning_rate": 9.662510781776836e-05, "loss": 2.9534, "step": 4304 }, { "epoch": 0.34, "learning_rate": 9.662432368854387e-05, "loss": 3.0904, "step": 4305 }, { "epoch": 0.34, "learning_rate": 9.662353955931937e-05, "loss": 2.8865, "step": 4306 }, { "epoch": 0.34, "learning_rate": 9.662275543009489e-05, "loss": 2.7677, "step": 4307 }, { "epoch": 0.34, "learning_rate": 9.662197130087039e-05, "loss": 3.2044, "step": 4308 }, { "epoch": 0.34, "learning_rate": 9.66211871716459e-05, "loss": 3.0481, "step": 4309 }, { "epoch": 0.34, "learning_rate": 9.66204030424214e-05, "loss": 3.0793, "step": 4310 }, { "epoch": 0.34, "learning_rate": 9.66196189131969e-05, "loss": 3.2265, "step": 4311 }, { "epoch": 0.34, "learning_rate": 9.66188347839724e-05, "loss": 3.4106, "step": 4312 }, { "epoch": 0.34, "learning_rate": 9.661805065474791e-05, "loss": 2.9138, "step": 4313 }, { "epoch": 0.34, "learning_rate": 9.661726652552341e-05, "loss": 3.0776, "step": 4314 }, { "epoch": 0.34, "learning_rate": 9.661648239629892e-05, "loss": 3.1514, "step": 4315 }, { "epoch": 0.34, "learning_rate": 9.661569826707442e-05, "loss": 3.1472, "step": 4316 }, { "epoch": 0.34, "learning_rate": 9.661491413784992e-05, "loss": 3.0158, "step": 4317 }, { "epoch": 0.34, "learning_rate": 9.661413000862543e-05, "loss": 2.8868, "step": 4318 }, { "epoch": 0.34, "learning_rate": 9.661334587940093e-05, "loss": 3.0172, "step": 4319 }, { "epoch": 0.34, "learning_rate": 9.661256175017643e-05, "loss": 3.0328, "step": 4320 }, { "epoch": 0.34, "learning_rate": 9.661177762095194e-05, "loss": 3.2302, "step": 4321 }, { "epoch": 0.34, "learning_rate": 9.661099349172744e-05, "loss": 3.1303, "step": 4322 }, { "epoch": 0.34, "learning_rate": 9.661020936250295e-05, "loss": 3.0973, "step": 4323 }, { "epoch": 0.34, "learning_rate": 9.660942523327845e-05, "loss": 2.9979, "step": 4324 }, { "epoch": 0.34, "learning_rate": 9.660864110405395e-05, "loss": 2.9549, "step": 4325 }, { "epoch": 0.34, "learning_rate": 9.660785697482946e-05, "loss": 3.1811, "step": 4326 }, { "epoch": 0.34, "learning_rate": 9.660707284560496e-05, "loss": 2.9096, "step": 4327 }, { "epoch": 0.34, "learning_rate": 9.660628871638046e-05, "loss": 3.1555, "step": 4328 }, { "epoch": 0.34, "learning_rate": 9.660550458715597e-05, "loss": 2.9052, "step": 4329 }, { "epoch": 0.34, "learning_rate": 9.660472045793147e-05, "loss": 3.4615, "step": 4330 }, { "epoch": 0.34, "learning_rate": 9.660393632870698e-05, "loss": 3.189, "step": 4331 }, { "epoch": 0.34, "learning_rate": 9.660315219948248e-05, "loss": 3.4313, "step": 4332 }, { "epoch": 0.34, "learning_rate": 9.660236807025798e-05, "loss": 3.2448, "step": 4333 }, { "epoch": 0.34, "learning_rate": 9.660158394103349e-05, "loss": 3.0053, "step": 4334 }, { "epoch": 0.34, "learning_rate": 9.660079981180899e-05, "loss": 3.1436, "step": 4335 }, { "epoch": 0.34, "learning_rate": 9.66000156825845e-05, "loss": 3.2268, "step": 4336 }, { "epoch": 0.34, "learning_rate": 9.659923155336e-05, "loss": 3.3249, "step": 4337 }, { "epoch": 0.34, "learning_rate": 9.65984474241355e-05, "loss": 2.9442, "step": 4338 }, { "epoch": 0.34, "learning_rate": 9.6597663294911e-05, "loss": 3.1081, "step": 4339 }, { "epoch": 0.34, "learning_rate": 9.659687916568651e-05, "loss": 3.3602, "step": 4340 }, { "epoch": 0.34, "learning_rate": 9.659609503646201e-05, "loss": 3.3213, "step": 4341 }, { "epoch": 0.34, "learning_rate": 9.659531090723752e-05, "loss": 2.9341, "step": 4342 }, { "epoch": 0.34, "learning_rate": 9.659452677801302e-05, "loss": 3.0633, "step": 4343 }, { "epoch": 0.34, "learning_rate": 9.659374264878852e-05, "loss": 2.9662, "step": 4344 }, { "epoch": 0.34, "learning_rate": 9.659295851956403e-05, "loss": 3.3639, "step": 4345 }, { "epoch": 0.34, "learning_rate": 9.659217439033953e-05, "loss": 2.9401, "step": 4346 }, { "epoch": 0.34, "learning_rate": 9.659139026111503e-05, "loss": 3.3126, "step": 4347 }, { "epoch": 0.34, "learning_rate": 9.659060613189054e-05, "loss": 2.9773, "step": 4348 }, { "epoch": 0.34, "learning_rate": 9.658982200266604e-05, "loss": 3.0763, "step": 4349 }, { "epoch": 0.34, "learning_rate": 9.658903787344154e-05, "loss": 3.163, "step": 4350 }, { "epoch": 0.34, "learning_rate": 9.658825374421705e-05, "loss": 3.6644, "step": 4351 }, { "epoch": 0.34, "learning_rate": 9.658746961499255e-05, "loss": 2.864, "step": 4352 }, { "epoch": 0.34, "learning_rate": 9.658668548576806e-05, "loss": 3.5502, "step": 4353 }, { "epoch": 0.34, "learning_rate": 9.658590135654356e-05, "loss": 3.2726, "step": 4354 }, { "epoch": 0.34, "learning_rate": 9.658511722731906e-05, "loss": 3.0075, "step": 4355 }, { "epoch": 0.34, "learning_rate": 9.658433309809457e-05, "loss": 3.0634, "step": 4356 }, { "epoch": 0.34, "learning_rate": 9.658354896887007e-05, "loss": 3.3762, "step": 4357 }, { "epoch": 0.34, "learning_rate": 9.658276483964557e-05, "loss": 3.3509, "step": 4358 }, { "epoch": 0.34, "learning_rate": 9.658198071042108e-05, "loss": 2.9303, "step": 4359 }, { "epoch": 0.34, "learning_rate": 9.658119658119658e-05, "loss": 3.1114, "step": 4360 }, { "epoch": 0.34, "learning_rate": 9.658041245197209e-05, "loss": 3.0705, "step": 4361 }, { "epoch": 0.34, "learning_rate": 9.657962832274759e-05, "loss": 2.9885, "step": 4362 }, { "epoch": 0.34, "learning_rate": 9.657884419352309e-05, "loss": 3.3335, "step": 4363 }, { "epoch": 0.34, "learning_rate": 9.657806006429861e-05, "loss": 3.0338, "step": 4364 }, { "epoch": 0.34, "learning_rate": 9.657727593507411e-05, "loss": 2.9507, "step": 4365 }, { "epoch": 0.34, "learning_rate": 9.657649180584962e-05, "loss": 3.0758, "step": 4366 }, { "epoch": 0.34, "learning_rate": 9.657570767662512e-05, "loss": 2.7548, "step": 4367 }, { "epoch": 0.34, "learning_rate": 9.657492354740062e-05, "loss": 3.1419, "step": 4368 }, { "epoch": 0.34, "learning_rate": 9.657413941817613e-05, "loss": 3.122, "step": 4369 }, { "epoch": 0.34, "learning_rate": 9.657335528895163e-05, "loss": 2.9644, "step": 4370 }, { "epoch": 0.34, "learning_rate": 9.657257115972714e-05, "loss": 2.9894, "step": 4371 }, { "epoch": 0.34, "learning_rate": 9.657178703050264e-05, "loss": 3.1957, "step": 4372 }, { "epoch": 0.34, "learning_rate": 9.657100290127814e-05, "loss": 3.0496, "step": 4373 }, { "epoch": 0.34, "learning_rate": 9.657021877205365e-05, "loss": 3.0859, "step": 4374 }, { "epoch": 0.34, "learning_rate": 9.656943464282914e-05, "loss": 3.3527, "step": 4375 }, { "epoch": 0.34, "learning_rate": 9.656865051360464e-05, "loss": 3.2396, "step": 4376 }, { "epoch": 0.34, "learning_rate": 9.656786638438014e-05, "loss": 2.9852, "step": 4377 }, { "epoch": 0.34, "learning_rate": 9.656708225515565e-05, "loss": 3.4962, "step": 4378 }, { "epoch": 0.34, "learning_rate": 9.656629812593115e-05, "loss": 3.2545, "step": 4379 }, { "epoch": 0.34, "learning_rate": 9.656551399670666e-05, "loss": 2.9585, "step": 4380 }, { "epoch": 0.34, "learning_rate": 9.656472986748216e-05, "loss": 3.297, "step": 4381 }, { "epoch": 0.34, "learning_rate": 9.656394573825766e-05, "loss": 3.3524, "step": 4382 }, { "epoch": 0.34, "learning_rate": 9.656316160903317e-05, "loss": 3.1501, "step": 4383 }, { "epoch": 0.34, "learning_rate": 9.656237747980867e-05, "loss": 3.0256, "step": 4384 }, { "epoch": 0.34, "learning_rate": 9.656159335058417e-05, "loss": 3.0048, "step": 4385 }, { "epoch": 0.34, "learning_rate": 9.656080922135968e-05, "loss": 3.3647, "step": 4386 }, { "epoch": 0.34, "learning_rate": 9.656002509213518e-05, "loss": 2.9468, "step": 4387 }, { "epoch": 0.34, "learning_rate": 9.655924096291068e-05, "loss": 3.0926, "step": 4388 }, { "epoch": 0.34, "learning_rate": 9.655845683368619e-05, "loss": 2.9486, "step": 4389 }, { "epoch": 0.34, "learning_rate": 9.65576727044617e-05, "loss": 2.8666, "step": 4390 }, { "epoch": 0.34, "learning_rate": 9.655688857523721e-05, "loss": 3.0912, "step": 4391 }, { "epoch": 0.34, "learning_rate": 9.655610444601271e-05, "loss": 2.9321, "step": 4392 }, { "epoch": 0.34, "learning_rate": 9.655532031678822e-05, "loss": 2.9753, "step": 4393 }, { "epoch": 0.34, "learning_rate": 9.655453618756372e-05, "loss": 2.9325, "step": 4394 }, { "epoch": 0.34, "learning_rate": 9.655375205833922e-05, "loss": 2.9233, "step": 4395 }, { "epoch": 0.34, "learning_rate": 9.655296792911473e-05, "loss": 3.1597, "step": 4396 }, { "epoch": 0.34, "learning_rate": 9.655218379989023e-05, "loss": 2.9654, "step": 4397 }, { "epoch": 0.34, "learning_rate": 9.655139967066573e-05, "loss": 2.8859, "step": 4398 }, { "epoch": 0.34, "learning_rate": 9.655061554144124e-05, "loss": 3.2351, "step": 4399 }, { "epoch": 0.35, "learning_rate": 9.654983141221674e-05, "loss": 3.2267, "step": 4400 }, { "epoch": 0.35, "learning_rate": 9.654904728299225e-05, "loss": 2.7978, "step": 4401 }, { "epoch": 0.35, "learning_rate": 9.654826315376775e-05, "loss": 2.9517, "step": 4402 }, { "epoch": 0.35, "learning_rate": 9.654747902454325e-05, "loss": 3.1628, "step": 4403 }, { "epoch": 0.35, "learning_rate": 9.654669489531876e-05, "loss": 3.048, "step": 4404 }, { "epoch": 0.35, "learning_rate": 9.654591076609426e-05, "loss": 2.9493, "step": 4405 }, { "epoch": 0.35, "learning_rate": 9.654512663686976e-05, "loss": 3.1069, "step": 4406 }, { "epoch": 0.35, "learning_rate": 9.654434250764527e-05, "loss": 2.9798, "step": 4407 }, { "epoch": 0.35, "learning_rate": 9.654355837842077e-05, "loss": 2.8752, "step": 4408 }, { "epoch": 0.35, "learning_rate": 9.654277424919628e-05, "loss": 3.0949, "step": 4409 }, { "epoch": 0.35, "learning_rate": 9.654199011997178e-05, "loss": 2.9088, "step": 4410 }, { "epoch": 0.35, "learning_rate": 9.654120599074727e-05, "loss": 2.8269, "step": 4411 }, { "epoch": 0.35, "learning_rate": 9.654042186152277e-05, "loss": 3.1666, "step": 4412 }, { "epoch": 0.35, "learning_rate": 9.653963773229828e-05, "loss": 3.3309, "step": 4413 }, { "epoch": 0.35, "learning_rate": 9.653885360307378e-05, "loss": 3.0868, "step": 4414 }, { "epoch": 0.35, "learning_rate": 9.65380694738493e-05, "loss": 3.111, "step": 4415 }, { "epoch": 0.35, "learning_rate": 9.65372853446248e-05, "loss": 2.9267, "step": 4416 }, { "epoch": 0.35, "learning_rate": 9.65365012154003e-05, "loss": 3.0214, "step": 4417 }, { "epoch": 0.35, "learning_rate": 9.653571708617581e-05, "loss": 3.0872, "step": 4418 }, { "epoch": 0.35, "learning_rate": 9.653493295695131e-05, "loss": 3.0547, "step": 4419 }, { "epoch": 0.35, "learning_rate": 9.653414882772682e-05, "loss": 3.4851, "step": 4420 }, { "epoch": 0.35, "learning_rate": 9.653336469850232e-05, "loss": 3.2329, "step": 4421 }, { "epoch": 0.35, "learning_rate": 9.653258056927782e-05, "loss": 2.91, "step": 4422 }, { "epoch": 0.35, "learning_rate": 9.653179644005333e-05, "loss": 2.9364, "step": 4423 }, { "epoch": 0.35, "learning_rate": 9.653101231082883e-05, "loss": 3.1838, "step": 4424 }, { "epoch": 0.35, "learning_rate": 9.653022818160433e-05, "loss": 2.949, "step": 4425 }, { "epoch": 0.35, "learning_rate": 9.652944405237984e-05, "loss": 2.9568, "step": 4426 }, { "epoch": 0.35, "learning_rate": 9.652865992315534e-05, "loss": 3.1015, "step": 4427 }, { "epoch": 0.35, "learning_rate": 9.652787579393085e-05, "loss": 3.0639, "step": 4428 }, { "epoch": 0.35, "learning_rate": 9.652709166470635e-05, "loss": 2.9897, "step": 4429 }, { "epoch": 0.35, "learning_rate": 9.652630753548185e-05, "loss": 3.0596, "step": 4430 }, { "epoch": 0.35, "learning_rate": 9.652552340625736e-05, "loss": 3.0571, "step": 4431 }, { "epoch": 0.35, "learning_rate": 9.652473927703286e-05, "loss": 3.1654, "step": 4432 }, { "epoch": 0.35, "learning_rate": 9.652395514780836e-05, "loss": 2.8545, "step": 4433 }, { "epoch": 0.35, "learning_rate": 9.652317101858387e-05, "loss": 2.8401, "step": 4434 }, { "epoch": 0.35, "learning_rate": 9.652238688935937e-05, "loss": 2.8573, "step": 4435 }, { "epoch": 0.35, "learning_rate": 9.652160276013487e-05, "loss": 3.2596, "step": 4436 }, { "epoch": 0.35, "learning_rate": 9.652081863091038e-05, "loss": 3.1873, "step": 4437 }, { "epoch": 0.35, "learning_rate": 9.652003450168588e-05, "loss": 2.9775, "step": 4438 }, { "epoch": 0.35, "learning_rate": 9.651925037246139e-05, "loss": 2.9903, "step": 4439 }, { "epoch": 0.35, "learning_rate": 9.651846624323689e-05, "loss": 2.9962, "step": 4440 }, { "epoch": 0.35, "learning_rate": 9.651768211401239e-05, "loss": 3.5175, "step": 4441 }, { "epoch": 0.35, "learning_rate": 9.65168979847879e-05, "loss": 2.96, "step": 4442 }, { "epoch": 0.35, "learning_rate": 9.65161138555634e-05, "loss": 3.2179, "step": 4443 }, { "epoch": 0.35, "learning_rate": 9.65153297263389e-05, "loss": 2.9704, "step": 4444 }, { "epoch": 0.35, "learning_rate": 9.651454559711441e-05, "loss": 3.0333, "step": 4445 }, { "epoch": 0.35, "learning_rate": 9.651376146788991e-05, "loss": 2.8094, "step": 4446 }, { "epoch": 0.35, "learning_rate": 9.651297733866542e-05, "loss": 3.0622, "step": 4447 }, { "epoch": 0.35, "learning_rate": 9.651219320944092e-05, "loss": 3.0609, "step": 4448 }, { "epoch": 0.35, "learning_rate": 9.651140908021642e-05, "loss": 2.9505, "step": 4449 }, { "epoch": 0.35, "learning_rate": 9.651062495099193e-05, "loss": 2.9764, "step": 4450 }, { "epoch": 0.35, "learning_rate": 9.650984082176743e-05, "loss": 3.1251, "step": 4451 }, { "epoch": 0.35, "learning_rate": 9.650905669254293e-05, "loss": 3.2531, "step": 4452 }, { "epoch": 0.35, "learning_rate": 9.650827256331844e-05, "loss": 3.1491, "step": 4453 }, { "epoch": 0.35, "learning_rate": 9.650748843409394e-05, "loss": 2.826, "step": 4454 }, { "epoch": 0.35, "learning_rate": 9.650670430486944e-05, "loss": 3.1264, "step": 4455 }, { "epoch": 0.35, "learning_rate": 9.650592017564495e-05, "loss": 2.7872, "step": 4456 }, { "epoch": 0.35, "learning_rate": 9.650513604642045e-05, "loss": 3.0823, "step": 4457 }, { "epoch": 0.35, "learning_rate": 9.650435191719596e-05, "loss": 2.8663, "step": 4458 }, { "epoch": 0.35, "learning_rate": 9.650356778797146e-05, "loss": 3.1832, "step": 4459 }, { "epoch": 0.35, "learning_rate": 9.650278365874696e-05, "loss": 3.2609, "step": 4460 }, { "epoch": 0.35, "learning_rate": 9.650199952952247e-05, "loss": 2.8995, "step": 4461 }, { "epoch": 0.35, "learning_rate": 9.650121540029797e-05, "loss": 3.0889, "step": 4462 }, { "epoch": 0.35, "learning_rate": 9.650043127107347e-05, "loss": 3.1413, "step": 4463 }, { "epoch": 0.35, "learning_rate": 9.649964714184898e-05, "loss": 3.2848, "step": 4464 }, { "epoch": 0.35, "learning_rate": 9.649886301262448e-05, "loss": 3.1274, "step": 4465 }, { "epoch": 0.35, "learning_rate": 9.649807888339999e-05, "loss": 2.8845, "step": 4466 }, { "epoch": 0.35, "learning_rate": 9.649729475417549e-05, "loss": 3.0738, "step": 4467 }, { "epoch": 0.35, "learning_rate": 9.649651062495099e-05, "loss": 3.2355, "step": 4468 }, { "epoch": 0.35, "learning_rate": 9.64957264957265e-05, "loss": 2.8455, "step": 4469 }, { "epoch": 0.35, "learning_rate": 9.6494942366502e-05, "loss": 2.9035, "step": 4470 }, { "epoch": 0.35, "learning_rate": 9.64941582372775e-05, "loss": 3.061, "step": 4471 }, { "epoch": 0.35, "learning_rate": 9.649337410805302e-05, "loss": 2.9152, "step": 4472 }, { "epoch": 0.35, "learning_rate": 9.649258997882852e-05, "loss": 3.0675, "step": 4473 }, { "epoch": 0.35, "learning_rate": 9.649180584960403e-05, "loss": 2.9605, "step": 4474 }, { "epoch": 0.35, "learning_rate": 9.649102172037953e-05, "loss": 3.3621, "step": 4475 }, { "epoch": 0.35, "learning_rate": 9.649023759115504e-05, "loss": 2.9327, "step": 4476 }, { "epoch": 0.35, "learning_rate": 9.648945346193054e-05, "loss": 3.2089, "step": 4477 }, { "epoch": 0.35, "learning_rate": 9.648866933270604e-05, "loss": 3.1233, "step": 4478 }, { "epoch": 0.35, "learning_rate": 9.648788520348155e-05, "loss": 3.0779, "step": 4479 }, { "epoch": 0.35, "learning_rate": 9.648710107425705e-05, "loss": 3.4546, "step": 4480 }, { "epoch": 0.35, "learning_rate": 9.648631694503255e-05, "loss": 2.975, "step": 4481 }, { "epoch": 0.35, "learning_rate": 9.648553281580806e-05, "loss": 2.8629, "step": 4482 }, { "epoch": 0.35, "learning_rate": 9.648474868658355e-05, "loss": 3.3121, "step": 4483 }, { "epoch": 0.35, "learning_rate": 9.648396455735905e-05, "loss": 3.0242, "step": 4484 }, { "epoch": 0.35, "learning_rate": 9.648318042813455e-05, "loss": 3.2521, "step": 4485 }, { "epoch": 0.35, "learning_rate": 9.648239629891006e-05, "loss": 2.8978, "step": 4486 }, { "epoch": 0.35, "learning_rate": 9.648161216968556e-05, "loss": 3.2778, "step": 4487 }, { "epoch": 0.35, "learning_rate": 9.648082804046107e-05, "loss": 3.0047, "step": 4488 }, { "epoch": 0.35, "learning_rate": 9.648004391123657e-05, "loss": 3.1243, "step": 4489 }, { "epoch": 0.35, "learning_rate": 9.647925978201207e-05, "loss": 3.1083, "step": 4490 }, { "epoch": 0.35, "learning_rate": 9.647847565278758e-05, "loss": 3.2164, "step": 4491 }, { "epoch": 0.35, "learning_rate": 9.647769152356308e-05, "loss": 3.0609, "step": 4492 }, { "epoch": 0.35, "learning_rate": 9.647690739433858e-05, "loss": 3.2522, "step": 4493 }, { "epoch": 0.35, "learning_rate": 9.647612326511409e-05, "loss": 3.3344, "step": 4494 }, { "epoch": 0.35, "learning_rate": 9.647533913588959e-05, "loss": 2.9764, "step": 4495 }, { "epoch": 0.35, "learning_rate": 9.64745550066651e-05, "loss": 3.0885, "step": 4496 }, { "epoch": 0.35, "learning_rate": 9.64737708774406e-05, "loss": 3.4256, "step": 4497 }, { "epoch": 0.35, "learning_rate": 9.647298674821612e-05, "loss": 3.033, "step": 4498 }, { "epoch": 0.35, "learning_rate": 9.647220261899162e-05, "loss": 3.0292, "step": 4499 }, { "epoch": 0.35, "learning_rate": 9.647141848976712e-05, "loss": 3.0286, "step": 4500 }, { "epoch": 0.35, "learning_rate": 9.647063436054263e-05, "loss": 3.1414, "step": 4501 }, { "epoch": 0.35, "learning_rate": 9.646985023131813e-05, "loss": 3.0596, "step": 4502 }, { "epoch": 0.35, "learning_rate": 9.646906610209363e-05, "loss": 3.2814, "step": 4503 }, { "epoch": 0.35, "learning_rate": 9.646828197286914e-05, "loss": 2.8652, "step": 4504 }, { "epoch": 0.35, "learning_rate": 9.646749784364464e-05, "loss": 3.0261, "step": 4505 }, { "epoch": 0.35, "learning_rate": 9.646671371442015e-05, "loss": 3.0859, "step": 4506 }, { "epoch": 0.35, "learning_rate": 9.646592958519565e-05, "loss": 2.8924, "step": 4507 }, { "epoch": 0.35, "learning_rate": 9.646514545597115e-05, "loss": 3.0006, "step": 4508 }, { "epoch": 0.35, "learning_rate": 9.646436132674666e-05, "loss": 2.8368, "step": 4509 }, { "epoch": 0.35, "learning_rate": 9.646357719752216e-05, "loss": 3.0712, "step": 4510 }, { "epoch": 0.35, "learning_rate": 9.646279306829766e-05, "loss": 3.0988, "step": 4511 }, { "epoch": 0.35, "learning_rate": 9.646200893907317e-05, "loss": 2.9647, "step": 4512 }, { "epoch": 0.35, "learning_rate": 9.646122480984867e-05, "loss": 3.1298, "step": 4513 }, { "epoch": 0.35, "learning_rate": 9.646044068062418e-05, "loss": 3.0139, "step": 4514 }, { "epoch": 0.35, "learning_rate": 9.645965655139968e-05, "loss": 3.19, "step": 4515 }, { "epoch": 0.35, "learning_rate": 9.645887242217518e-05, "loss": 2.9725, "step": 4516 }, { "epoch": 0.35, "learning_rate": 9.645808829295069e-05, "loss": 3.1776, "step": 4517 }, { "epoch": 0.35, "learning_rate": 9.645730416372619e-05, "loss": 2.9708, "step": 4518 }, { "epoch": 0.35, "learning_rate": 9.645652003450168e-05, "loss": 3.1278, "step": 4519 }, { "epoch": 0.35, "learning_rate": 9.645573590527718e-05, "loss": 3.0863, "step": 4520 }, { "epoch": 0.35, "learning_rate": 9.645495177605269e-05, "loss": 3.4612, "step": 4521 }, { "epoch": 0.35, "learning_rate": 9.645416764682819e-05, "loss": 2.8573, "step": 4522 }, { "epoch": 0.35, "learning_rate": 9.645338351760371e-05, "loss": 3.0917, "step": 4523 }, { "epoch": 0.35, "learning_rate": 9.645259938837921e-05, "loss": 2.9401, "step": 4524 }, { "epoch": 0.35, "learning_rate": 9.645181525915472e-05, "loss": 3.1072, "step": 4525 }, { "epoch": 0.35, "learning_rate": 9.645103112993022e-05, "loss": 3.1157, "step": 4526 }, { "epoch": 0.35, "learning_rate": 9.645024700070572e-05, "loss": 2.9095, "step": 4527 }, { "epoch": 0.36, "learning_rate": 9.644946287148123e-05, "loss": 2.9701, "step": 4528 }, { "epoch": 0.36, "learning_rate": 9.644867874225673e-05, "loss": 2.9435, "step": 4529 }, { "epoch": 0.36, "learning_rate": 9.644789461303223e-05, "loss": 3.387, "step": 4530 }, { "epoch": 0.36, "learning_rate": 9.644711048380774e-05, "loss": 3.0524, "step": 4531 }, { "epoch": 0.36, "learning_rate": 9.644632635458324e-05, "loss": 3.0402, "step": 4532 }, { "epoch": 0.36, "learning_rate": 9.644554222535875e-05, "loss": 2.7947, "step": 4533 }, { "epoch": 0.36, "learning_rate": 9.644475809613425e-05, "loss": 3.2092, "step": 4534 }, { "epoch": 0.36, "learning_rate": 9.644397396690975e-05, "loss": 2.9746, "step": 4535 }, { "epoch": 0.36, "learning_rate": 9.644318983768526e-05, "loss": 2.9597, "step": 4536 }, { "epoch": 0.36, "learning_rate": 9.644240570846076e-05, "loss": 3.0956, "step": 4537 }, { "epoch": 0.36, "learning_rate": 9.644162157923626e-05, "loss": 2.94, "step": 4538 }, { "epoch": 0.36, "learning_rate": 9.644083745001177e-05, "loss": 3.0467, "step": 4539 }, { "epoch": 0.36, "learning_rate": 9.644005332078727e-05, "loss": 3.0165, "step": 4540 }, { "epoch": 0.36, "learning_rate": 9.643926919156277e-05, "loss": 3.2951, "step": 4541 }, { "epoch": 0.36, "learning_rate": 9.643848506233828e-05, "loss": 2.9395, "step": 4542 }, { "epoch": 0.36, "learning_rate": 9.643770093311378e-05, "loss": 3.2038, "step": 4543 }, { "epoch": 0.36, "learning_rate": 9.643691680388929e-05, "loss": 2.9971, "step": 4544 }, { "epoch": 0.36, "learning_rate": 9.643613267466479e-05, "loss": 2.918, "step": 4545 }, { "epoch": 0.36, "learning_rate": 9.643534854544029e-05, "loss": 3.2542, "step": 4546 }, { "epoch": 0.36, "learning_rate": 9.64345644162158e-05, "loss": 3.1341, "step": 4547 }, { "epoch": 0.36, "learning_rate": 9.64337802869913e-05, "loss": 3.1715, "step": 4548 }, { "epoch": 0.36, "learning_rate": 9.64329961577668e-05, "loss": 2.9882, "step": 4549 }, { "epoch": 0.36, "learning_rate": 9.643221202854231e-05, "loss": 3.0071, "step": 4550 }, { "epoch": 0.36, "learning_rate": 9.643142789931781e-05, "loss": 3.3986, "step": 4551 }, { "epoch": 0.36, "learning_rate": 9.643064377009331e-05, "loss": 3.0561, "step": 4552 }, { "epoch": 0.36, "learning_rate": 9.642985964086882e-05, "loss": 3.2995, "step": 4553 }, { "epoch": 0.36, "learning_rate": 9.642907551164432e-05, "loss": 2.986, "step": 4554 }, { "epoch": 0.36, "learning_rate": 9.642829138241983e-05, "loss": 3.2338, "step": 4555 }, { "epoch": 0.36, "learning_rate": 9.642750725319533e-05, "loss": 2.9402, "step": 4556 }, { "epoch": 0.36, "learning_rate": 9.642672312397083e-05, "loss": 2.9178, "step": 4557 }, { "epoch": 0.36, "learning_rate": 9.642593899474634e-05, "loss": 3.3257, "step": 4558 }, { "epoch": 0.36, "learning_rate": 9.642515486552184e-05, "loss": 3.1016, "step": 4559 }, { "epoch": 0.36, "learning_rate": 9.642437073629734e-05, "loss": 3.1165, "step": 4560 }, { "epoch": 0.36, "learning_rate": 9.642358660707285e-05, "loss": 3.1735, "step": 4561 }, { "epoch": 0.36, "learning_rate": 9.642280247784835e-05, "loss": 3.4275, "step": 4562 }, { "epoch": 0.36, "learning_rate": 9.642201834862386e-05, "loss": 2.9231, "step": 4563 }, { "epoch": 0.36, "learning_rate": 9.642123421939936e-05, "loss": 2.9124, "step": 4564 }, { "epoch": 0.36, "learning_rate": 9.642045009017486e-05, "loss": 2.9103, "step": 4565 }, { "epoch": 0.36, "learning_rate": 9.641966596095037e-05, "loss": 3.2165, "step": 4566 }, { "epoch": 0.36, "learning_rate": 9.641888183172587e-05, "loss": 3.1633, "step": 4567 }, { "epoch": 0.36, "learning_rate": 9.641809770250137e-05, "loss": 3.1326, "step": 4568 }, { "epoch": 0.36, "learning_rate": 9.641731357327688e-05, "loss": 2.9445, "step": 4569 }, { "epoch": 0.36, "learning_rate": 9.641652944405238e-05, "loss": 3.0297, "step": 4570 }, { "epoch": 0.36, "learning_rate": 9.641574531482788e-05, "loss": 2.9471, "step": 4571 }, { "epoch": 0.36, "learning_rate": 9.641496118560339e-05, "loss": 3.4149, "step": 4572 }, { "epoch": 0.36, "learning_rate": 9.641417705637889e-05, "loss": 3.1906, "step": 4573 }, { "epoch": 0.36, "learning_rate": 9.64133929271544e-05, "loss": 3.113, "step": 4574 }, { "epoch": 0.36, "learning_rate": 9.64126087979299e-05, "loss": 3.0367, "step": 4575 }, { "epoch": 0.36, "learning_rate": 9.64118246687054e-05, "loss": 3.0137, "step": 4576 }, { "epoch": 0.36, "learning_rate": 9.641104053948091e-05, "loss": 3.1221, "step": 4577 }, { "epoch": 0.36, "learning_rate": 9.641025641025641e-05, "loss": 3.0644, "step": 4578 }, { "epoch": 0.36, "learning_rate": 9.640947228103191e-05, "loss": 3.2233, "step": 4579 }, { "epoch": 0.36, "learning_rate": 9.640868815180743e-05, "loss": 3.1048, "step": 4580 }, { "epoch": 0.36, "learning_rate": 9.640790402258294e-05, "loss": 3.1252, "step": 4581 }, { "epoch": 0.36, "learning_rate": 9.640711989335844e-05, "loss": 3.1845, "step": 4582 }, { "epoch": 0.36, "learning_rate": 9.640633576413394e-05, "loss": 3.0356, "step": 4583 }, { "epoch": 0.36, "learning_rate": 9.640555163490945e-05, "loss": 3.1723, "step": 4584 }, { "epoch": 0.36, "learning_rate": 9.640476750568495e-05, "loss": 2.9918, "step": 4585 }, { "epoch": 0.36, "learning_rate": 9.640398337646045e-05, "loss": 2.8542, "step": 4586 }, { "epoch": 0.36, "learning_rate": 9.640319924723596e-05, "loss": 3.1843, "step": 4587 }, { "epoch": 0.36, "learning_rate": 9.640241511801146e-05, "loss": 3.1414, "step": 4588 }, { "epoch": 0.36, "learning_rate": 9.640163098878696e-05, "loss": 2.934, "step": 4589 }, { "epoch": 0.36, "learning_rate": 9.640084685956247e-05, "loss": 3.1808, "step": 4590 }, { "epoch": 0.36, "learning_rate": 9.640006273033796e-05, "loss": 3.0619, "step": 4591 }, { "epoch": 0.36, "learning_rate": 9.639927860111346e-05, "loss": 2.9919, "step": 4592 }, { "epoch": 0.36, "learning_rate": 9.639849447188897e-05, "loss": 3.0578, "step": 4593 }, { "epoch": 0.36, "learning_rate": 9.639771034266447e-05, "loss": 3.227, "step": 4594 }, { "epoch": 0.36, "learning_rate": 9.639692621343997e-05, "loss": 3.1445, "step": 4595 }, { "epoch": 0.36, "learning_rate": 9.639614208421548e-05, "loss": 3.0019, "step": 4596 }, { "epoch": 0.36, "learning_rate": 9.639535795499098e-05, "loss": 2.8601, "step": 4597 }, { "epoch": 0.36, "learning_rate": 9.639457382576648e-05, "loss": 3.2162, "step": 4598 }, { "epoch": 0.36, "learning_rate": 9.639378969654199e-05, "loss": 3.3611, "step": 4599 }, { "epoch": 0.36, "learning_rate": 9.639300556731749e-05, "loss": 3.0073, "step": 4600 }, { "epoch": 0.36, "learning_rate": 9.6392221438093e-05, "loss": 2.8401, "step": 4601 }, { "epoch": 0.36, "learning_rate": 9.63914373088685e-05, "loss": 3.2336, "step": 4602 }, { "epoch": 0.36, "learning_rate": 9.6390653179644e-05, "loss": 2.8741, "step": 4603 }, { "epoch": 0.36, "learning_rate": 9.63898690504195e-05, "loss": 2.7793, "step": 4604 }, { "epoch": 0.36, "learning_rate": 9.638908492119501e-05, "loss": 2.8531, "step": 4605 }, { "epoch": 0.36, "learning_rate": 9.638830079197053e-05, "loss": 3.1873, "step": 4606 }, { "epoch": 0.36, "learning_rate": 9.638751666274603e-05, "loss": 3.0704, "step": 4607 }, { "epoch": 0.36, "learning_rate": 9.638673253352153e-05, "loss": 2.9537, "step": 4608 }, { "epoch": 0.36, "learning_rate": 9.638594840429704e-05, "loss": 2.936, "step": 4609 }, { "epoch": 0.36, "learning_rate": 9.638516427507254e-05, "loss": 3.085, "step": 4610 }, { "epoch": 0.36, "learning_rate": 9.638438014584805e-05, "loss": 3.0489, "step": 4611 }, { "epoch": 0.36, "learning_rate": 9.638359601662355e-05, "loss": 2.9041, "step": 4612 }, { "epoch": 0.36, "learning_rate": 9.638281188739905e-05, "loss": 3.0821, "step": 4613 }, { "epoch": 0.36, "learning_rate": 9.638202775817456e-05, "loss": 3.1295, "step": 4614 }, { "epoch": 0.36, "learning_rate": 9.638124362895006e-05, "loss": 3.082, "step": 4615 }, { "epoch": 0.36, "learning_rate": 9.638045949972556e-05, "loss": 3.132, "step": 4616 }, { "epoch": 0.36, "learning_rate": 9.637967537050107e-05, "loss": 2.7692, "step": 4617 }, { "epoch": 0.36, "learning_rate": 9.637889124127657e-05, "loss": 3.0123, "step": 4618 }, { "epoch": 0.36, "learning_rate": 9.637810711205207e-05, "loss": 3.0409, "step": 4619 }, { "epoch": 0.36, "learning_rate": 9.637732298282758e-05, "loss": 3.0595, "step": 4620 }, { "epoch": 0.36, "learning_rate": 9.637653885360308e-05, "loss": 2.8866, "step": 4621 }, { "epoch": 0.36, "learning_rate": 9.637575472437859e-05, "loss": 3.1013, "step": 4622 }, { "epoch": 0.36, "learning_rate": 9.637497059515409e-05, "loss": 3.027, "step": 4623 }, { "epoch": 0.36, "learning_rate": 9.63741864659296e-05, "loss": 2.8886, "step": 4624 }, { "epoch": 0.36, "learning_rate": 9.63734023367051e-05, "loss": 2.9269, "step": 4625 }, { "epoch": 0.36, "learning_rate": 9.63726182074806e-05, "loss": 3.0812, "step": 4626 }, { "epoch": 0.36, "learning_rate": 9.637183407825609e-05, "loss": 3.1613, "step": 4627 }, { "epoch": 0.36, "learning_rate": 9.63710499490316e-05, "loss": 3.3127, "step": 4628 }, { "epoch": 0.36, "learning_rate": 9.63702658198071e-05, "loss": 2.9646, "step": 4629 }, { "epoch": 0.36, "learning_rate": 9.63694816905826e-05, "loss": 3.2011, "step": 4630 }, { "epoch": 0.36, "learning_rate": 9.636869756135812e-05, "loss": 3.0702, "step": 4631 }, { "epoch": 0.36, "learning_rate": 9.636791343213362e-05, "loss": 3.1433, "step": 4632 }, { "epoch": 0.36, "learning_rate": 9.636712930290913e-05, "loss": 2.9997, "step": 4633 }, { "epoch": 0.36, "learning_rate": 9.636634517368463e-05, "loss": 3.161, "step": 4634 }, { "epoch": 0.36, "learning_rate": 9.636556104446013e-05, "loss": 3.4307, "step": 4635 }, { "epoch": 0.36, "learning_rate": 9.636477691523564e-05, "loss": 3.0632, "step": 4636 }, { "epoch": 0.36, "learning_rate": 9.636399278601114e-05, "loss": 3.4959, "step": 4637 }, { "epoch": 0.36, "learning_rate": 9.636320865678664e-05, "loss": 3.0354, "step": 4638 }, { "epoch": 0.36, "learning_rate": 9.636242452756215e-05, "loss": 3.1089, "step": 4639 }, { "epoch": 0.36, "learning_rate": 9.636164039833765e-05, "loss": 3.1464, "step": 4640 }, { "epoch": 0.36, "learning_rate": 9.636085626911316e-05, "loss": 3.0657, "step": 4641 }, { "epoch": 0.36, "learning_rate": 9.636007213988866e-05, "loss": 3.05, "step": 4642 }, { "epoch": 0.36, "learning_rate": 9.635928801066416e-05, "loss": 2.9995, "step": 4643 }, { "epoch": 0.36, "learning_rate": 9.635850388143967e-05, "loss": 2.9069, "step": 4644 }, { "epoch": 0.36, "learning_rate": 9.635771975221517e-05, "loss": 3.2093, "step": 4645 }, { "epoch": 0.36, "learning_rate": 9.635693562299067e-05, "loss": 3.2217, "step": 4646 }, { "epoch": 0.36, "learning_rate": 9.635615149376618e-05, "loss": 3.127, "step": 4647 }, { "epoch": 0.36, "learning_rate": 9.635536736454168e-05, "loss": 2.8849, "step": 4648 }, { "epoch": 0.36, "learning_rate": 9.635458323531719e-05, "loss": 2.8597, "step": 4649 }, { "epoch": 0.36, "learning_rate": 9.635379910609269e-05, "loss": 3.2184, "step": 4650 }, { "epoch": 0.36, "learning_rate": 9.635301497686819e-05, "loss": 3.3112, "step": 4651 }, { "epoch": 0.36, "learning_rate": 9.63522308476437e-05, "loss": 3.0768, "step": 4652 }, { "epoch": 0.36, "learning_rate": 9.63514467184192e-05, "loss": 3.0087, "step": 4653 }, { "epoch": 0.36, "learning_rate": 9.63506625891947e-05, "loss": 3.0239, "step": 4654 }, { "epoch": 0.37, "learning_rate": 9.634987845997021e-05, "loss": 2.9974, "step": 4655 }, { "epoch": 0.37, "learning_rate": 9.634909433074571e-05, "loss": 3.1939, "step": 4656 }, { "epoch": 0.37, "learning_rate": 9.634831020152121e-05, "loss": 2.9612, "step": 4657 }, { "epoch": 0.37, "learning_rate": 9.634752607229672e-05, "loss": 3.1363, "step": 4658 }, { "epoch": 0.37, "learning_rate": 9.634674194307222e-05, "loss": 3.1641, "step": 4659 }, { "epoch": 0.37, "learning_rate": 9.634595781384773e-05, "loss": 3.019, "step": 4660 }, { "epoch": 0.37, "learning_rate": 9.634517368462323e-05, "loss": 2.8625, "step": 4661 }, { "epoch": 0.37, "learning_rate": 9.634438955539873e-05, "loss": 3.0801, "step": 4662 }, { "epoch": 0.37, "learning_rate": 9.634360542617424e-05, "loss": 3.0963, "step": 4663 }, { "epoch": 0.37, "learning_rate": 9.634282129694974e-05, "loss": 3.1835, "step": 4664 }, { "epoch": 0.37, "learning_rate": 9.634203716772524e-05, "loss": 3.116, "step": 4665 }, { "epoch": 0.37, "learning_rate": 9.634125303850075e-05, "loss": 2.8832, "step": 4666 }, { "epoch": 0.37, "learning_rate": 9.634046890927625e-05, "loss": 3.215, "step": 4667 }, { "epoch": 0.37, "learning_rate": 9.633968478005176e-05, "loss": 3.0351, "step": 4668 }, { "epoch": 0.37, "learning_rate": 9.633890065082726e-05, "loss": 3.3137, "step": 4669 }, { "epoch": 0.37, "learning_rate": 9.633811652160276e-05, "loss": 3.3489, "step": 4670 }, { "epoch": 0.37, "learning_rate": 9.633733239237827e-05, "loss": 3.062, "step": 4671 }, { "epoch": 0.37, "learning_rate": 9.633654826315377e-05, "loss": 3.1826, "step": 4672 }, { "epoch": 0.37, "learning_rate": 9.633576413392927e-05, "loss": 3.1645, "step": 4673 }, { "epoch": 0.37, "learning_rate": 9.633498000470478e-05, "loss": 3.0576, "step": 4674 }, { "epoch": 0.37, "learning_rate": 9.633419587548028e-05, "loss": 3.0175, "step": 4675 }, { "epoch": 0.37, "learning_rate": 9.633341174625578e-05, "loss": 3.1267, "step": 4676 }, { "epoch": 0.37, "learning_rate": 9.633262761703129e-05, "loss": 2.9136, "step": 4677 }, { "epoch": 0.37, "learning_rate": 9.633184348780679e-05, "loss": 3.0587, "step": 4678 }, { "epoch": 0.37, "learning_rate": 9.63310593585823e-05, "loss": 3.1311, "step": 4679 }, { "epoch": 0.37, "learning_rate": 9.63302752293578e-05, "loss": 3.0161, "step": 4680 }, { "epoch": 0.37, "learning_rate": 9.63294911001333e-05, "loss": 3.1078, "step": 4681 }, { "epoch": 0.37, "learning_rate": 9.63287069709088e-05, "loss": 3.0157, "step": 4682 }, { "epoch": 0.37, "learning_rate": 9.632792284168431e-05, "loss": 3.3438, "step": 4683 }, { "epoch": 0.37, "learning_rate": 9.632713871245981e-05, "loss": 3.2136, "step": 4684 }, { "epoch": 0.37, "learning_rate": 9.632635458323532e-05, "loss": 3.179, "step": 4685 }, { "epoch": 0.37, "learning_rate": 9.632557045401082e-05, "loss": 3.2262, "step": 4686 }, { "epoch": 0.37, "learning_rate": 9.632478632478632e-05, "loss": 2.9165, "step": 4687 }, { "epoch": 0.37, "learning_rate": 9.632400219556184e-05, "loss": 2.9962, "step": 4688 }, { "epoch": 0.37, "learning_rate": 9.632321806633735e-05, "loss": 2.8619, "step": 4689 }, { "epoch": 0.37, "learning_rate": 9.632243393711285e-05, "loss": 3.1942, "step": 4690 }, { "epoch": 0.37, "learning_rate": 9.632164980788835e-05, "loss": 2.9808, "step": 4691 }, { "epoch": 0.37, "learning_rate": 9.632086567866386e-05, "loss": 3.0254, "step": 4692 }, { "epoch": 0.37, "learning_rate": 9.632008154943936e-05, "loss": 2.9672, "step": 4693 }, { "epoch": 0.37, "learning_rate": 9.631929742021486e-05, "loss": 3.2879, "step": 4694 }, { "epoch": 0.37, "learning_rate": 9.631851329099037e-05, "loss": 3.1468, "step": 4695 }, { "epoch": 0.37, "learning_rate": 9.631772916176587e-05, "loss": 3.0162, "step": 4696 }, { "epoch": 0.37, "learning_rate": 9.631694503254138e-05, "loss": 3.0131, "step": 4697 }, { "epoch": 0.37, "learning_rate": 9.631616090331688e-05, "loss": 3.1869, "step": 4698 }, { "epoch": 0.37, "learning_rate": 9.631537677409237e-05, "loss": 3.1812, "step": 4699 }, { "epoch": 0.37, "learning_rate": 9.631459264486787e-05, "loss": 3.1333, "step": 4700 }, { "epoch": 0.37, "learning_rate": 9.631380851564338e-05, "loss": 2.8368, "step": 4701 }, { "epoch": 0.37, "learning_rate": 9.631302438641888e-05, "loss": 3.029, "step": 4702 }, { "epoch": 0.37, "learning_rate": 9.631224025719438e-05, "loss": 3.0705, "step": 4703 }, { "epoch": 0.37, "learning_rate": 9.631145612796989e-05, "loss": 3.1157, "step": 4704 }, { "epoch": 0.37, "learning_rate": 9.631067199874539e-05, "loss": 2.9134, "step": 4705 }, { "epoch": 0.37, "learning_rate": 9.63098878695209e-05, "loss": 3.1645, "step": 4706 }, { "epoch": 0.37, "learning_rate": 9.63091037402964e-05, "loss": 3.2968, "step": 4707 }, { "epoch": 0.37, "learning_rate": 9.63083196110719e-05, "loss": 3.0827, "step": 4708 }, { "epoch": 0.37, "learning_rate": 9.63075354818474e-05, "loss": 2.9966, "step": 4709 }, { "epoch": 0.37, "learning_rate": 9.630675135262291e-05, "loss": 3.2048, "step": 4710 }, { "epoch": 0.37, "learning_rate": 9.630596722339841e-05, "loss": 3.3919, "step": 4711 }, { "epoch": 0.37, "learning_rate": 9.630518309417392e-05, "loss": 2.9028, "step": 4712 }, { "epoch": 0.37, "learning_rate": 9.630439896494942e-05, "loss": 2.8664, "step": 4713 }, { "epoch": 0.37, "learning_rate": 9.630361483572494e-05, "loss": 2.9214, "step": 4714 }, { "epoch": 0.37, "learning_rate": 9.630283070650044e-05, "loss": 3.0786, "step": 4715 }, { "epoch": 0.37, "learning_rate": 9.630204657727595e-05, "loss": 2.8043, "step": 4716 }, { "epoch": 0.37, "learning_rate": 9.630126244805145e-05, "loss": 3.0621, "step": 4717 }, { "epoch": 0.37, "learning_rate": 9.630047831882695e-05, "loss": 3.3536, "step": 4718 }, { "epoch": 0.37, "learning_rate": 9.629969418960246e-05, "loss": 2.9948, "step": 4719 }, { "epoch": 0.37, "learning_rate": 9.629891006037796e-05, "loss": 2.9575, "step": 4720 }, { "epoch": 0.37, "learning_rate": 9.629812593115346e-05, "loss": 3.0337, "step": 4721 }, { "epoch": 0.37, "learning_rate": 9.629734180192897e-05, "loss": 3.0727, "step": 4722 }, { "epoch": 0.37, "learning_rate": 9.629655767270447e-05, "loss": 3.1511, "step": 4723 }, { "epoch": 0.37, "learning_rate": 9.629577354347997e-05, "loss": 2.6564, "step": 4724 }, { "epoch": 0.37, "learning_rate": 9.629498941425548e-05, "loss": 2.9327, "step": 4725 }, { "epoch": 0.37, "learning_rate": 9.629420528503098e-05, "loss": 2.923, "step": 4726 }, { "epoch": 0.37, "learning_rate": 9.629342115580649e-05, "loss": 3.0405, "step": 4727 }, { "epoch": 0.37, "learning_rate": 9.629263702658199e-05, "loss": 3.0969, "step": 4728 }, { "epoch": 0.37, "learning_rate": 9.629185289735749e-05, "loss": 3.0656, "step": 4729 }, { "epoch": 0.37, "learning_rate": 9.6291068768133e-05, "loss": 3.1149, "step": 4730 }, { "epoch": 0.37, "learning_rate": 9.62902846389085e-05, "loss": 3.5613, "step": 4731 }, { "epoch": 0.37, "learning_rate": 9.6289500509684e-05, "loss": 2.9264, "step": 4732 }, { "epoch": 0.37, "learning_rate": 9.628871638045951e-05, "loss": 3.0952, "step": 4733 }, { "epoch": 0.37, "learning_rate": 9.628793225123501e-05, "loss": 3.1538, "step": 4734 }, { "epoch": 0.37, "learning_rate": 9.62871481220105e-05, "loss": 3.111, "step": 4735 }, { "epoch": 0.37, "learning_rate": 9.6286363992786e-05, "loss": 3.0279, "step": 4736 }, { "epoch": 0.37, "learning_rate": 9.628557986356151e-05, "loss": 3.0167, "step": 4737 }, { "epoch": 0.37, "learning_rate": 9.628479573433701e-05, "loss": 2.7718, "step": 4738 }, { "epoch": 0.37, "learning_rate": 9.628401160511253e-05, "loss": 2.8469, "step": 4739 }, { "epoch": 0.37, "learning_rate": 9.628322747588803e-05, "loss": 3.0102, "step": 4740 }, { "epoch": 0.37, "learning_rate": 9.628244334666354e-05, "loss": 3.2224, "step": 4741 }, { "epoch": 0.37, "learning_rate": 9.628165921743904e-05, "loss": 3.0872, "step": 4742 }, { "epoch": 0.37, "learning_rate": 9.628087508821454e-05, "loss": 3.2833, "step": 4743 }, { "epoch": 0.37, "learning_rate": 9.628009095899005e-05, "loss": 3.0789, "step": 4744 }, { "epoch": 0.37, "learning_rate": 9.627930682976555e-05, "loss": 2.9604, "step": 4745 }, { "epoch": 0.37, "learning_rate": 9.627852270054106e-05, "loss": 3.0925, "step": 4746 }, { "epoch": 0.37, "learning_rate": 9.627773857131656e-05, "loss": 3.1659, "step": 4747 }, { "epoch": 0.37, "learning_rate": 9.627695444209206e-05, "loss": 3.3557, "step": 4748 }, { "epoch": 0.37, "learning_rate": 9.627617031286757e-05, "loss": 3.1322, "step": 4749 }, { "epoch": 0.37, "learning_rate": 9.627538618364307e-05, "loss": 3.0545, "step": 4750 }, { "epoch": 0.37, "learning_rate": 9.627460205441857e-05, "loss": 3.1279, "step": 4751 }, { "epoch": 0.37, "learning_rate": 9.627381792519408e-05, "loss": 3.1415, "step": 4752 }, { "epoch": 0.37, "learning_rate": 9.627303379596958e-05, "loss": 2.9242, "step": 4753 }, { "epoch": 0.37, "learning_rate": 9.627224966674508e-05, "loss": 3.0445, "step": 4754 }, { "epoch": 0.37, "learning_rate": 9.627146553752059e-05, "loss": 3.317, "step": 4755 }, { "epoch": 0.37, "learning_rate": 9.627068140829609e-05, "loss": 2.9391, "step": 4756 }, { "epoch": 0.37, "learning_rate": 9.62698972790716e-05, "loss": 3.2194, "step": 4757 }, { "epoch": 0.37, "learning_rate": 9.62691131498471e-05, "loss": 3.0861, "step": 4758 }, { "epoch": 0.37, "learning_rate": 9.62683290206226e-05, "loss": 3.0829, "step": 4759 }, { "epoch": 0.37, "learning_rate": 9.626754489139811e-05, "loss": 2.891, "step": 4760 }, { "epoch": 0.37, "learning_rate": 9.626676076217361e-05, "loss": 3.1063, "step": 4761 }, { "epoch": 0.37, "learning_rate": 9.626597663294911e-05, "loss": 3.0279, "step": 4762 }, { "epoch": 0.37, "learning_rate": 9.626519250372462e-05, "loss": 3.0081, "step": 4763 }, { "epoch": 0.37, "learning_rate": 9.626440837450012e-05, "loss": 2.9913, "step": 4764 }, { "epoch": 0.37, "learning_rate": 9.626362424527563e-05, "loss": 3.1751, "step": 4765 }, { "epoch": 0.37, "learning_rate": 9.626284011605113e-05, "loss": 3.1729, "step": 4766 }, { "epoch": 0.37, "learning_rate": 9.626205598682663e-05, "loss": 3.2658, "step": 4767 }, { "epoch": 0.37, "learning_rate": 9.626127185760214e-05, "loss": 3.1206, "step": 4768 }, { "epoch": 0.37, "learning_rate": 9.626048772837764e-05, "loss": 3.1615, "step": 4769 }, { "epoch": 0.37, "learning_rate": 9.625970359915314e-05, "loss": 2.9181, "step": 4770 }, { "epoch": 0.37, "learning_rate": 9.625891946992865e-05, "loss": 3.2385, "step": 4771 }, { "epoch": 0.37, "learning_rate": 9.625813534070415e-05, "loss": 2.8695, "step": 4772 }, { "epoch": 0.37, "learning_rate": 9.625735121147965e-05, "loss": 3.2883, "step": 4773 }, { "epoch": 0.37, "learning_rate": 9.625656708225516e-05, "loss": 3.1223, "step": 4774 }, { "epoch": 0.37, "learning_rate": 9.625578295303066e-05, "loss": 2.9624, "step": 4775 }, { "epoch": 0.37, "learning_rate": 9.625499882380617e-05, "loss": 3.1007, "step": 4776 }, { "epoch": 0.37, "learning_rate": 9.625421469458167e-05, "loss": 2.9715, "step": 4777 }, { "epoch": 0.37, "learning_rate": 9.625343056535717e-05, "loss": 2.9002, "step": 4778 }, { "epoch": 0.37, "learning_rate": 9.625264643613268e-05, "loss": 3.2807, "step": 4779 }, { "epoch": 0.37, "learning_rate": 9.625186230690818e-05, "loss": 3.0588, "step": 4780 }, { "epoch": 0.37, "learning_rate": 9.625107817768368e-05, "loss": 3.0395, "step": 4781 }, { "epoch": 0.37, "learning_rate": 9.625029404845919e-05, "loss": 2.8681, "step": 4782 }, { "epoch": 0.38, "learning_rate": 9.624950991923469e-05, "loss": 3.2045, "step": 4783 }, { "epoch": 0.38, "learning_rate": 9.62487257900102e-05, "loss": 3.1944, "step": 4784 }, { "epoch": 0.38, "learning_rate": 9.62479416607857e-05, "loss": 3.0626, "step": 4785 }, { "epoch": 0.38, "learning_rate": 9.62471575315612e-05, "loss": 2.9427, "step": 4786 }, { "epoch": 0.38, "learning_rate": 9.62463734023367e-05, "loss": 3.0218, "step": 4787 }, { "epoch": 0.38, "learning_rate": 9.624558927311221e-05, "loss": 3.1012, "step": 4788 }, { "epoch": 0.38, "learning_rate": 9.624480514388771e-05, "loss": 3.1535, "step": 4789 }, { "epoch": 0.38, "learning_rate": 9.624402101466322e-05, "loss": 3.1222, "step": 4790 }, { "epoch": 0.38, "learning_rate": 9.624323688543872e-05, "loss": 3.1789, "step": 4791 }, { "epoch": 0.38, "learning_rate": 9.624245275621422e-05, "loss": 3.0142, "step": 4792 }, { "epoch": 0.38, "learning_rate": 9.624166862698973e-05, "loss": 3.1683, "step": 4793 }, { "epoch": 0.38, "learning_rate": 9.624088449776523e-05, "loss": 3.0977, "step": 4794 }, { "epoch": 0.38, "learning_rate": 9.624010036854074e-05, "loss": 3.0834, "step": 4795 }, { "epoch": 0.38, "learning_rate": 9.623931623931625e-05, "loss": 2.9633, "step": 4796 }, { "epoch": 0.38, "learning_rate": 9.623853211009176e-05, "loss": 2.9782, "step": 4797 }, { "epoch": 0.38, "learning_rate": 9.623774798086726e-05, "loss": 3.087, "step": 4798 }, { "epoch": 0.38, "learning_rate": 9.623696385164276e-05, "loss": 2.9654, "step": 4799 }, { "epoch": 0.38, "learning_rate": 9.623617972241827e-05, "loss": 3.0795, "step": 4800 }, { "epoch": 0.38, "learning_rate": 9.623539559319377e-05, "loss": 3.1119, "step": 4801 }, { "epoch": 0.38, "learning_rate": 9.623461146396927e-05, "loss": 3.1467, "step": 4802 }, { "epoch": 0.38, "learning_rate": 9.623382733474478e-05, "loss": 3.0838, "step": 4803 }, { "epoch": 0.38, "learning_rate": 9.623304320552028e-05, "loss": 2.9526, "step": 4804 }, { "epoch": 0.38, "learning_rate": 9.623225907629579e-05, "loss": 2.9941, "step": 4805 }, { "epoch": 0.38, "learning_rate": 9.623147494707129e-05, "loss": 3.1671, "step": 4806 }, { "epoch": 0.38, "learning_rate": 9.623069081784678e-05, "loss": 2.8584, "step": 4807 }, { "epoch": 0.38, "learning_rate": 9.622990668862228e-05, "loss": 3.1142, "step": 4808 }, { "epoch": 0.38, "learning_rate": 9.622912255939779e-05, "loss": 2.9631, "step": 4809 }, { "epoch": 0.38, "learning_rate": 9.622833843017329e-05, "loss": 3.043, "step": 4810 }, { "epoch": 0.38, "learning_rate": 9.62275543009488e-05, "loss": 2.9753, "step": 4811 }, { "epoch": 0.38, "learning_rate": 9.62267701717243e-05, "loss": 2.9644, "step": 4812 }, { "epoch": 0.38, "learning_rate": 9.62259860424998e-05, "loss": 2.9963, "step": 4813 }, { "epoch": 0.38, "learning_rate": 9.62252019132753e-05, "loss": 3.3092, "step": 4814 }, { "epoch": 0.38, "learning_rate": 9.622441778405081e-05, "loss": 2.9072, "step": 4815 }, { "epoch": 0.38, "learning_rate": 9.622363365482631e-05, "loss": 2.8429, "step": 4816 }, { "epoch": 0.38, "learning_rate": 9.622284952560182e-05, "loss": 3.0892, "step": 4817 }, { "epoch": 0.38, "learning_rate": 9.622206539637732e-05, "loss": 3.1073, "step": 4818 }, { "epoch": 0.38, "learning_rate": 9.622128126715282e-05, "loss": 3.2084, "step": 4819 }, { "epoch": 0.38, "learning_rate": 9.622049713792833e-05, "loss": 3.1424, "step": 4820 }, { "epoch": 0.38, "learning_rate": 9.621971300870383e-05, "loss": 3.1802, "step": 4821 }, { "epoch": 0.38, "learning_rate": 9.621892887947935e-05, "loss": 2.9713, "step": 4822 }, { "epoch": 0.38, "learning_rate": 9.621814475025485e-05, "loss": 2.9074, "step": 4823 }, { "epoch": 0.38, "learning_rate": 9.621736062103036e-05, "loss": 2.8772, "step": 4824 }, { "epoch": 0.38, "learning_rate": 9.621657649180586e-05, "loss": 3.3471, "step": 4825 }, { "epoch": 0.38, "learning_rate": 9.621579236258136e-05, "loss": 3.0946, "step": 4826 }, { "epoch": 0.38, "learning_rate": 9.621500823335687e-05, "loss": 3.3003, "step": 4827 }, { "epoch": 0.38, "learning_rate": 9.621422410413237e-05, "loss": 3.029, "step": 4828 }, { "epoch": 0.38, "learning_rate": 9.621343997490787e-05, "loss": 2.8498, "step": 4829 }, { "epoch": 0.38, "learning_rate": 9.621265584568338e-05, "loss": 3.0388, "step": 4830 }, { "epoch": 0.38, "learning_rate": 9.621187171645888e-05, "loss": 3.5475, "step": 4831 }, { "epoch": 0.38, "learning_rate": 9.621108758723439e-05, "loss": 3.0025, "step": 4832 }, { "epoch": 0.38, "learning_rate": 9.621030345800989e-05, "loss": 2.989, "step": 4833 }, { "epoch": 0.38, "learning_rate": 9.620951932878539e-05, "loss": 3.1327, "step": 4834 }, { "epoch": 0.38, "learning_rate": 9.62087351995609e-05, "loss": 3.0579, "step": 4835 }, { "epoch": 0.38, "learning_rate": 9.62079510703364e-05, "loss": 2.9827, "step": 4836 }, { "epoch": 0.38, "learning_rate": 9.62071669411119e-05, "loss": 3.0737, "step": 4837 }, { "epoch": 0.38, "learning_rate": 9.620638281188741e-05, "loss": 3.1781, "step": 4838 }, { "epoch": 0.38, "learning_rate": 9.620559868266291e-05, "loss": 3.0338, "step": 4839 }, { "epoch": 0.38, "learning_rate": 9.620481455343841e-05, "loss": 2.8532, "step": 4840 }, { "epoch": 0.38, "learning_rate": 9.620403042421392e-05, "loss": 2.7537, "step": 4841 }, { "epoch": 0.38, "learning_rate": 9.620324629498942e-05, "loss": 2.9627, "step": 4842 }, { "epoch": 0.38, "learning_rate": 9.620246216576491e-05, "loss": 2.9645, "step": 4843 }, { "epoch": 0.38, "learning_rate": 9.620167803654042e-05, "loss": 3.0117, "step": 4844 }, { "epoch": 0.38, "learning_rate": 9.620089390731592e-05, "loss": 2.9386, "step": 4845 }, { "epoch": 0.38, "learning_rate": 9.620010977809142e-05, "loss": 3.0353, "step": 4846 }, { "epoch": 0.38, "learning_rate": 9.619932564886693e-05, "loss": 3.1044, "step": 4847 }, { "epoch": 0.38, "learning_rate": 9.619854151964244e-05, "loss": 2.9046, "step": 4848 }, { "epoch": 0.38, "learning_rate": 9.619775739041795e-05, "loss": 3.1812, "step": 4849 }, { "epoch": 0.38, "learning_rate": 9.619697326119345e-05, "loss": 3.269, "step": 4850 }, { "epoch": 0.38, "learning_rate": 9.619618913196896e-05, "loss": 3.1467, "step": 4851 }, { "epoch": 0.38, "learning_rate": 9.619540500274446e-05, "loss": 3.1806, "step": 4852 }, { "epoch": 0.38, "learning_rate": 9.619462087351996e-05, "loss": 3.2817, "step": 4853 }, { "epoch": 0.38, "learning_rate": 9.619383674429547e-05, "loss": 2.9518, "step": 4854 }, { "epoch": 0.38, "learning_rate": 9.619305261507097e-05, "loss": 3.231, "step": 4855 }, { "epoch": 0.38, "learning_rate": 9.619226848584647e-05, "loss": 2.8974, "step": 4856 }, { "epoch": 0.38, "learning_rate": 9.619148435662198e-05, "loss": 3.238, "step": 4857 }, { "epoch": 0.38, "learning_rate": 9.619070022739748e-05, "loss": 3.0237, "step": 4858 }, { "epoch": 0.38, "learning_rate": 9.618991609817298e-05, "loss": 2.9478, "step": 4859 }, { "epoch": 0.38, "learning_rate": 9.618913196894849e-05, "loss": 2.9714, "step": 4860 }, { "epoch": 0.38, "learning_rate": 9.618834783972399e-05, "loss": 3.1611, "step": 4861 }, { "epoch": 0.38, "learning_rate": 9.61875637104995e-05, "loss": 3.1356, "step": 4862 }, { "epoch": 0.38, "learning_rate": 9.6186779581275e-05, "loss": 3.0575, "step": 4863 }, { "epoch": 0.38, "learning_rate": 9.61859954520505e-05, "loss": 2.9932, "step": 4864 }, { "epoch": 0.38, "learning_rate": 9.6185211322826e-05, "loss": 3.118, "step": 4865 }, { "epoch": 0.38, "learning_rate": 9.618442719360151e-05, "loss": 2.9794, "step": 4866 }, { "epoch": 0.38, "learning_rate": 9.618364306437701e-05, "loss": 3.0978, "step": 4867 }, { "epoch": 0.38, "learning_rate": 9.618285893515252e-05, "loss": 3.4096, "step": 4868 }, { "epoch": 0.38, "learning_rate": 9.618207480592802e-05, "loss": 2.9602, "step": 4869 }, { "epoch": 0.38, "learning_rate": 9.618129067670353e-05, "loss": 2.8082, "step": 4870 }, { "epoch": 0.38, "learning_rate": 9.618050654747903e-05, "loss": 2.895, "step": 4871 }, { "epoch": 0.38, "learning_rate": 9.617972241825453e-05, "loss": 3.0106, "step": 4872 }, { "epoch": 0.38, "learning_rate": 9.617893828903004e-05, "loss": 3.234, "step": 4873 }, { "epoch": 0.38, "learning_rate": 9.617815415980554e-05, "loss": 2.8927, "step": 4874 }, { "epoch": 0.38, "learning_rate": 9.617737003058104e-05, "loss": 3.1386, "step": 4875 }, { "epoch": 0.38, "learning_rate": 9.617658590135655e-05, "loss": 3.0023, "step": 4876 }, { "epoch": 0.38, "learning_rate": 9.617580177213205e-05, "loss": 3.0343, "step": 4877 }, { "epoch": 0.38, "learning_rate": 9.617501764290755e-05, "loss": 2.8967, "step": 4878 }, { "epoch": 0.38, "learning_rate": 9.617423351368306e-05, "loss": 3.0018, "step": 4879 }, { "epoch": 0.38, "learning_rate": 9.617344938445856e-05, "loss": 3.0533, "step": 4880 }, { "epoch": 0.38, "learning_rate": 9.617266525523407e-05, "loss": 3.0849, "step": 4881 }, { "epoch": 0.38, "learning_rate": 9.617188112600957e-05, "loss": 3.3045, "step": 4882 }, { "epoch": 0.38, "learning_rate": 9.617109699678507e-05, "loss": 3.1184, "step": 4883 }, { "epoch": 0.38, "learning_rate": 9.617031286756058e-05, "loss": 2.8019, "step": 4884 }, { "epoch": 0.38, "learning_rate": 9.616952873833608e-05, "loss": 3.0595, "step": 4885 }, { "epoch": 0.38, "learning_rate": 9.616874460911158e-05, "loss": 3.167, "step": 4886 }, { "epoch": 0.38, "learning_rate": 9.616796047988709e-05, "loss": 2.9799, "step": 4887 }, { "epoch": 0.38, "learning_rate": 9.616717635066259e-05, "loss": 2.9536, "step": 4888 }, { "epoch": 0.38, "learning_rate": 9.61663922214381e-05, "loss": 3.0194, "step": 4889 }, { "epoch": 0.38, "learning_rate": 9.61656080922136e-05, "loss": 3.0645, "step": 4890 }, { "epoch": 0.38, "learning_rate": 9.61648239629891e-05, "loss": 2.9794, "step": 4891 }, { "epoch": 0.38, "learning_rate": 9.61640398337646e-05, "loss": 3.112, "step": 4892 }, { "epoch": 0.38, "learning_rate": 9.616325570454011e-05, "loss": 3.0694, "step": 4893 }, { "epoch": 0.38, "learning_rate": 9.616247157531561e-05, "loss": 3.0659, "step": 4894 }, { "epoch": 0.38, "learning_rate": 9.616168744609112e-05, "loss": 3.227, "step": 4895 }, { "epoch": 0.38, "learning_rate": 9.616090331686662e-05, "loss": 3.001, "step": 4896 }, { "epoch": 0.38, "learning_rate": 9.616011918764212e-05, "loss": 3.0304, "step": 4897 }, { "epoch": 0.38, "learning_rate": 9.615933505841763e-05, "loss": 2.9937, "step": 4898 }, { "epoch": 0.38, "learning_rate": 9.615855092919313e-05, "loss": 3.2246, "step": 4899 }, { "epoch": 0.38, "learning_rate": 9.615776679996864e-05, "loss": 2.9888, "step": 4900 }, { "epoch": 0.38, "learning_rate": 9.615698267074414e-05, "loss": 2.9835, "step": 4901 }, { "epoch": 0.38, "learning_rate": 9.615619854151964e-05, "loss": 3.1368, "step": 4902 }, { "epoch": 0.38, "learning_rate": 9.615541441229515e-05, "loss": 3.1164, "step": 4903 }, { "epoch": 0.38, "learning_rate": 9.615463028307066e-05, "loss": 3.089, "step": 4904 }, { "epoch": 0.38, "learning_rate": 9.615384615384617e-05, "loss": 2.9491, "step": 4905 }, { "epoch": 0.38, "learning_rate": 9.615306202462167e-05, "loss": 3.2746, "step": 4906 }, { "epoch": 0.38, "learning_rate": 9.615227789539717e-05, "loss": 3.1086, "step": 4907 }, { "epoch": 0.38, "learning_rate": 9.615149376617268e-05, "loss": 3.0236, "step": 4908 }, { "epoch": 0.38, "learning_rate": 9.615070963694818e-05, "loss": 3.4233, "step": 4909 }, { "epoch": 0.39, "learning_rate": 9.614992550772369e-05, "loss": 3.1058, "step": 4910 }, { "epoch": 0.39, "learning_rate": 9.614914137849919e-05, "loss": 3.5039, "step": 4911 }, { "epoch": 0.39, "learning_rate": 9.614835724927469e-05, "loss": 2.9607, "step": 4912 }, { "epoch": 0.39, "learning_rate": 9.61475731200502e-05, "loss": 3.0528, "step": 4913 }, { "epoch": 0.39, "learning_rate": 9.61467889908257e-05, "loss": 3.3905, "step": 4914 }, { "epoch": 0.39, "learning_rate": 9.614600486160119e-05, "loss": 3.0334, "step": 4915 }, { "epoch": 0.39, "learning_rate": 9.61452207323767e-05, "loss": 3.1667, "step": 4916 }, { "epoch": 0.39, "learning_rate": 9.61444366031522e-05, "loss": 3.111, "step": 4917 }, { "epoch": 0.39, "learning_rate": 9.61436524739277e-05, "loss": 3.1902, "step": 4918 }, { "epoch": 0.39, "learning_rate": 9.61428683447032e-05, "loss": 3.4121, "step": 4919 }, { "epoch": 0.39, "learning_rate": 9.614208421547871e-05, "loss": 3.0363, "step": 4920 }, { "epoch": 0.39, "learning_rate": 9.614130008625421e-05, "loss": 3.1904, "step": 4921 }, { "epoch": 0.39, "learning_rate": 9.614051595702972e-05, "loss": 2.9319, "step": 4922 }, { "epoch": 0.39, "learning_rate": 9.613973182780522e-05, "loss": 3.0569, "step": 4923 }, { "epoch": 0.39, "learning_rate": 9.613894769858072e-05, "loss": 3.1252, "step": 4924 }, { "epoch": 0.39, "learning_rate": 9.613816356935623e-05, "loss": 2.9518, "step": 4925 }, { "epoch": 0.39, "learning_rate": 9.613737944013173e-05, "loss": 2.9947, "step": 4926 }, { "epoch": 0.39, "learning_rate": 9.613659531090723e-05, "loss": 3.0162, "step": 4927 }, { "epoch": 0.39, "learning_rate": 9.613581118168274e-05, "loss": 3.1575, "step": 4928 }, { "epoch": 0.39, "learning_rate": 9.613502705245824e-05, "loss": 3.0169, "step": 4929 }, { "epoch": 0.39, "learning_rate": 9.613424292323376e-05, "loss": 3.0897, "step": 4930 }, { "epoch": 0.39, "learning_rate": 9.613345879400926e-05, "loss": 3.1385, "step": 4931 }, { "epoch": 0.39, "learning_rate": 9.613267466478477e-05, "loss": 2.9982, "step": 4932 }, { "epoch": 0.39, "learning_rate": 9.613189053556027e-05, "loss": 2.8791, "step": 4933 }, { "epoch": 0.39, "learning_rate": 9.613110640633577e-05, "loss": 3.2015, "step": 4934 }, { "epoch": 0.39, "learning_rate": 9.613032227711128e-05, "loss": 3.5098, "step": 4935 }, { "epoch": 0.39, "learning_rate": 9.612953814788678e-05, "loss": 3.0065, "step": 4936 }, { "epoch": 0.39, "learning_rate": 9.612875401866229e-05, "loss": 2.8272, "step": 4937 }, { "epoch": 0.39, "learning_rate": 9.612796988943779e-05, "loss": 3.4349, "step": 4938 }, { "epoch": 0.39, "learning_rate": 9.612718576021329e-05, "loss": 3.1832, "step": 4939 }, { "epoch": 0.39, "learning_rate": 9.61264016309888e-05, "loss": 2.9361, "step": 4940 }, { "epoch": 0.39, "learning_rate": 9.61256175017643e-05, "loss": 3.1542, "step": 4941 }, { "epoch": 0.39, "learning_rate": 9.61248333725398e-05, "loss": 3.1491, "step": 4942 }, { "epoch": 0.39, "learning_rate": 9.612404924331531e-05, "loss": 3.1365, "step": 4943 }, { "epoch": 0.39, "learning_rate": 9.612326511409081e-05, "loss": 3.313, "step": 4944 }, { "epoch": 0.39, "learning_rate": 9.612248098486631e-05, "loss": 2.8785, "step": 4945 }, { "epoch": 0.39, "learning_rate": 9.612169685564182e-05, "loss": 3.0112, "step": 4946 }, { "epoch": 0.39, "learning_rate": 9.612091272641732e-05, "loss": 3.0096, "step": 4947 }, { "epoch": 0.39, "learning_rate": 9.612012859719283e-05, "loss": 3.1432, "step": 4948 }, { "epoch": 0.39, "learning_rate": 9.611934446796833e-05, "loss": 3.0477, "step": 4949 }, { "epoch": 0.39, "learning_rate": 9.611856033874383e-05, "loss": 2.9165, "step": 4950 }, { "epoch": 0.39, "learning_rate": 9.611777620951932e-05, "loss": 2.852, "step": 4951 }, { "epoch": 0.39, "learning_rate": 9.611699208029483e-05, "loss": 3.0887, "step": 4952 }, { "epoch": 0.39, "learning_rate": 9.611620795107033e-05, "loss": 3.2216, "step": 4953 }, { "epoch": 0.39, "learning_rate": 9.611542382184583e-05, "loss": 3.1463, "step": 4954 }, { "epoch": 0.39, "learning_rate": 9.611463969262134e-05, "loss": 2.8537, "step": 4955 }, { "epoch": 0.39, "learning_rate": 9.611385556339685e-05, "loss": 2.8838, "step": 4956 }, { "epoch": 0.39, "learning_rate": 9.611307143417236e-05, "loss": 3.0396, "step": 4957 }, { "epoch": 0.39, "learning_rate": 9.611228730494786e-05, "loss": 3.1176, "step": 4958 }, { "epoch": 0.39, "learning_rate": 9.611150317572337e-05, "loss": 3.4162, "step": 4959 }, { "epoch": 0.39, "learning_rate": 9.611071904649887e-05, "loss": 2.932, "step": 4960 }, { "epoch": 0.39, "learning_rate": 9.610993491727437e-05, "loss": 3.0734, "step": 4961 }, { "epoch": 0.39, "learning_rate": 9.610915078804988e-05, "loss": 3.7068, "step": 4962 }, { "epoch": 0.39, "learning_rate": 9.610836665882538e-05, "loss": 3.2223, "step": 4963 }, { "epoch": 0.39, "learning_rate": 9.610758252960088e-05, "loss": 3.1146, "step": 4964 }, { "epoch": 0.39, "learning_rate": 9.610679840037639e-05, "loss": 3.1574, "step": 4965 }, { "epoch": 0.39, "learning_rate": 9.610601427115189e-05, "loss": 3.1357, "step": 4966 }, { "epoch": 0.39, "learning_rate": 9.61052301419274e-05, "loss": 2.9338, "step": 4967 }, { "epoch": 0.39, "learning_rate": 9.61044460127029e-05, "loss": 3.0274, "step": 4968 }, { "epoch": 0.39, "learning_rate": 9.61036618834784e-05, "loss": 3.2608, "step": 4969 }, { "epoch": 0.39, "learning_rate": 9.61028777542539e-05, "loss": 2.9883, "step": 4970 }, { "epoch": 0.39, "learning_rate": 9.610209362502941e-05, "loss": 3.0192, "step": 4971 }, { "epoch": 0.39, "learning_rate": 9.610130949580491e-05, "loss": 3.0152, "step": 4972 }, { "epoch": 0.39, "learning_rate": 9.610052536658042e-05, "loss": 3.1988, "step": 4973 }, { "epoch": 0.39, "learning_rate": 9.609974123735592e-05, "loss": 3.2376, "step": 4974 }, { "epoch": 0.39, "learning_rate": 9.609895710813142e-05, "loss": 3.0248, "step": 4975 }, { "epoch": 0.39, "learning_rate": 9.609817297890693e-05, "loss": 3.0974, "step": 4976 }, { "epoch": 0.39, "learning_rate": 9.609738884968243e-05, "loss": 2.9388, "step": 4977 }, { "epoch": 0.39, "learning_rate": 9.609660472045794e-05, "loss": 3.0087, "step": 4978 }, { "epoch": 0.39, "learning_rate": 9.609582059123344e-05, "loss": 3.2679, "step": 4979 }, { "epoch": 0.39, "learning_rate": 9.609503646200894e-05, "loss": 3.0003, "step": 4980 }, { "epoch": 0.39, "learning_rate": 9.609425233278445e-05, "loss": 2.9335, "step": 4981 }, { "epoch": 0.39, "learning_rate": 9.609346820355995e-05, "loss": 2.9902, "step": 4982 }, { "epoch": 0.39, "learning_rate": 9.609268407433545e-05, "loss": 2.8761, "step": 4983 }, { "epoch": 0.39, "learning_rate": 9.609189994511096e-05, "loss": 3.2514, "step": 4984 }, { "epoch": 0.39, "learning_rate": 9.609111581588646e-05, "loss": 3.1841, "step": 4985 }, { "epoch": 0.39, "learning_rate": 9.609033168666197e-05, "loss": 3.093, "step": 4986 }, { "epoch": 0.39, "learning_rate": 9.608954755743747e-05, "loss": 2.9099, "step": 4987 }, { "epoch": 0.39, "learning_rate": 9.608876342821297e-05, "loss": 3.2072, "step": 4988 }, { "epoch": 0.39, "learning_rate": 9.608797929898848e-05, "loss": 2.814, "step": 4989 }, { "epoch": 0.39, "learning_rate": 9.608719516976398e-05, "loss": 3.0668, "step": 4990 }, { "epoch": 0.39, "learning_rate": 9.608641104053948e-05, "loss": 2.9901, "step": 4991 }, { "epoch": 0.39, "learning_rate": 9.608562691131499e-05, "loss": 3.3681, "step": 4992 }, { "epoch": 0.39, "learning_rate": 9.608484278209049e-05, "loss": 3.0688, "step": 4993 }, { "epoch": 0.39, "learning_rate": 9.6084058652866e-05, "loss": 2.9379, "step": 4994 }, { "epoch": 0.39, "learning_rate": 9.60832745236415e-05, "loss": 3.1657, "step": 4995 }, { "epoch": 0.39, "learning_rate": 9.6082490394417e-05, "loss": 3.0819, "step": 4996 }, { "epoch": 0.39, "learning_rate": 9.60817062651925e-05, "loss": 3.1882, "step": 4997 }, { "epoch": 0.39, "learning_rate": 9.608092213596801e-05, "loss": 2.7852, "step": 4998 }, { "epoch": 0.39, "learning_rate": 9.608013800674351e-05, "loss": 3.1488, "step": 4999 }, { "epoch": 0.39, "learning_rate": 9.607935387751902e-05, "loss": 3.275, "step": 5000 }, { "epoch": 0.39, "eval_loss": 2.973233222961426, "eval_runtime": 218.0187, "eval_samples_per_second": 51.977, "eval_steps_per_second": 3.252, "step": 5000 }, { "epoch": 0.39, "learning_rate": 9.607856974829452e-05, "loss": 3.1193, "step": 5001 }, { "epoch": 0.39, "learning_rate": 9.607778561907002e-05, "loss": 2.71, "step": 5002 }, { "epoch": 0.39, "learning_rate": 9.607700148984553e-05, "loss": 2.8833, "step": 5003 }, { "epoch": 0.39, "learning_rate": 9.607621736062103e-05, "loss": 3.1822, "step": 5004 }, { "epoch": 0.39, "learning_rate": 9.607543323139654e-05, "loss": 3.1581, "step": 5005 }, { "epoch": 0.39, "learning_rate": 9.607464910217204e-05, "loss": 3.0151, "step": 5006 }, { "epoch": 0.39, "learning_rate": 9.607386497294754e-05, "loss": 3.0502, "step": 5007 }, { "epoch": 0.39, "learning_rate": 9.607308084372305e-05, "loss": 3.0899, "step": 5008 }, { "epoch": 0.39, "learning_rate": 9.607229671449855e-05, "loss": 3.0961, "step": 5009 }, { "epoch": 0.39, "learning_rate": 9.607151258527405e-05, "loss": 3.0854, "step": 5010 }, { "epoch": 0.39, "learning_rate": 9.607072845604956e-05, "loss": 3.14, "step": 5011 }, { "epoch": 0.39, "learning_rate": 9.606994432682507e-05, "loss": 3.2829, "step": 5012 }, { "epoch": 0.39, "learning_rate": 9.606916019760058e-05, "loss": 2.913, "step": 5013 }, { "epoch": 0.39, "learning_rate": 9.606837606837608e-05, "loss": 3.186, "step": 5014 }, { "epoch": 0.39, "learning_rate": 9.606759193915159e-05, "loss": 3.2282, "step": 5015 }, { "epoch": 0.39, "learning_rate": 9.606680780992709e-05, "loss": 3.1549, "step": 5016 }, { "epoch": 0.39, "learning_rate": 9.606602368070259e-05, "loss": 2.9768, "step": 5017 }, { "epoch": 0.39, "learning_rate": 9.60652395514781e-05, "loss": 3.0791, "step": 5018 }, { "epoch": 0.39, "learning_rate": 9.60644554222536e-05, "loss": 3.0557, "step": 5019 }, { "epoch": 0.39, "learning_rate": 9.60636712930291e-05, "loss": 2.9908, "step": 5020 }, { "epoch": 0.39, "learning_rate": 9.606288716380461e-05, "loss": 3.08, "step": 5021 }, { "epoch": 0.39, "learning_rate": 9.606210303458011e-05, "loss": 2.803, "step": 5022 }, { "epoch": 0.39, "learning_rate": 9.60613189053556e-05, "loss": 3.0129, "step": 5023 }, { "epoch": 0.39, "learning_rate": 9.60605347761311e-05, "loss": 3.2611, "step": 5024 }, { "epoch": 0.39, "learning_rate": 9.605975064690661e-05, "loss": 3.0016, "step": 5025 }, { "epoch": 0.39, "learning_rate": 9.605896651768211e-05, "loss": 3.0524, "step": 5026 }, { "epoch": 0.39, "learning_rate": 9.605818238845762e-05, "loss": 3.1367, "step": 5027 }, { "epoch": 0.39, "learning_rate": 9.605739825923312e-05, "loss": 3.279, "step": 5028 }, { "epoch": 0.39, "learning_rate": 9.605661413000862e-05, "loss": 3.1529, "step": 5029 }, { "epoch": 0.39, "learning_rate": 9.605583000078413e-05, "loss": 3.1031, "step": 5030 }, { "epoch": 0.39, "learning_rate": 9.605504587155963e-05, "loss": 3.113, "step": 5031 }, { "epoch": 0.39, "learning_rate": 9.605426174233513e-05, "loss": 2.8232, "step": 5032 }, { "epoch": 0.39, "learning_rate": 9.605347761311064e-05, "loss": 3.1594, "step": 5033 }, { "epoch": 0.39, "learning_rate": 9.605269348388614e-05, "loss": 3.0033, "step": 5034 }, { "epoch": 0.39, "learning_rate": 9.605190935466165e-05, "loss": 3.0747, "step": 5035 }, { "epoch": 0.39, "learning_rate": 9.605112522543715e-05, "loss": 3.2261, "step": 5036 }, { "epoch": 0.39, "learning_rate": 9.605034109621265e-05, "loss": 3.2339, "step": 5037 }, { "epoch": 0.4, "learning_rate": 9.604955696698817e-05, "loss": 3.0216, "step": 5038 }, { "epoch": 0.4, "learning_rate": 9.604877283776367e-05, "loss": 3.0775, "step": 5039 }, { "epoch": 0.4, "learning_rate": 9.604798870853918e-05, "loss": 2.9282, "step": 5040 }, { "epoch": 0.4, "learning_rate": 9.604720457931468e-05, "loss": 3.209, "step": 5041 }, { "epoch": 0.4, "learning_rate": 9.604642045009018e-05, "loss": 3.0554, "step": 5042 }, { "epoch": 0.4, "learning_rate": 9.604563632086569e-05, "loss": 3.1025, "step": 5043 }, { "epoch": 0.4, "learning_rate": 9.604485219164119e-05, "loss": 2.8979, "step": 5044 }, { "epoch": 0.4, "learning_rate": 9.60440680624167e-05, "loss": 3.4478, "step": 5045 }, { "epoch": 0.4, "learning_rate": 9.60432839331922e-05, "loss": 3.1144, "step": 5046 }, { "epoch": 0.4, "learning_rate": 9.60424998039677e-05, "loss": 3.252, "step": 5047 }, { "epoch": 0.4, "learning_rate": 9.604171567474321e-05, "loss": 3.0797, "step": 5048 }, { "epoch": 0.4, "learning_rate": 9.604093154551871e-05, "loss": 2.9311, "step": 5049 }, { "epoch": 0.4, "learning_rate": 9.604014741629421e-05, "loss": 3.1177, "step": 5050 }, { "epoch": 0.4, "learning_rate": 9.603936328706972e-05, "loss": 3.0072, "step": 5051 }, { "epoch": 0.4, "learning_rate": 9.603857915784522e-05, "loss": 2.8686, "step": 5052 }, { "epoch": 0.4, "learning_rate": 9.603779502862073e-05, "loss": 3.2619, "step": 5053 }, { "epoch": 0.4, "learning_rate": 9.603701089939623e-05, "loss": 3.0539, "step": 5054 }, { "epoch": 0.4, "learning_rate": 9.603622677017173e-05, "loss": 2.9313, "step": 5055 }, { "epoch": 0.4, "learning_rate": 9.603544264094724e-05, "loss": 2.9903, "step": 5056 }, { "epoch": 0.4, "learning_rate": 9.603465851172274e-05, "loss": 3.2297, "step": 5057 }, { "epoch": 0.4, "learning_rate": 9.603387438249824e-05, "loss": 3.1941, "step": 5058 }, { "epoch": 0.4, "learning_rate": 9.603309025327373e-05, "loss": 2.8259, "step": 5059 }, { "epoch": 0.4, "learning_rate": 9.603230612404924e-05, "loss": 3.1724, "step": 5060 }, { "epoch": 0.4, "learning_rate": 9.603152199482474e-05, "loss": 3.2361, "step": 5061 }, { "epoch": 0.4, "learning_rate": 9.603073786560024e-05, "loss": 2.9428, "step": 5062 }, { "epoch": 0.4, "learning_rate": 9.602995373637575e-05, "loss": 3.025, "step": 5063 }, { "epoch": 0.4, "learning_rate": 9.602916960715127e-05, "loss": 2.9298, "step": 5064 }, { "epoch": 0.4, "learning_rate": 9.602838547792677e-05, "loss": 2.9121, "step": 5065 }, { "epoch": 0.4, "learning_rate": 9.602760134870227e-05, "loss": 3.0896, "step": 5066 }, { "epoch": 0.4, "learning_rate": 9.602681721947778e-05, "loss": 3.3682, "step": 5067 }, { "epoch": 0.4, "learning_rate": 9.602603309025328e-05, "loss": 2.8216, "step": 5068 }, { "epoch": 0.4, "learning_rate": 9.602524896102878e-05, "loss": 2.897, "step": 5069 }, { "epoch": 0.4, "learning_rate": 9.602446483180429e-05, "loss": 3.0307, "step": 5070 }, { "epoch": 0.4, "learning_rate": 9.602368070257979e-05, "loss": 2.9415, "step": 5071 }, { "epoch": 0.4, "learning_rate": 9.60228965733553e-05, "loss": 2.9771, "step": 5072 }, { "epoch": 0.4, "learning_rate": 9.60221124441308e-05, "loss": 3.1659, "step": 5073 }, { "epoch": 0.4, "learning_rate": 9.60213283149063e-05, "loss": 3.1639, "step": 5074 }, { "epoch": 0.4, "learning_rate": 9.60205441856818e-05, "loss": 3.4434, "step": 5075 }, { "epoch": 0.4, "learning_rate": 9.601976005645731e-05, "loss": 2.9054, "step": 5076 }, { "epoch": 0.4, "learning_rate": 9.601897592723281e-05, "loss": 2.9269, "step": 5077 }, { "epoch": 0.4, "learning_rate": 9.601819179800832e-05, "loss": 3.2162, "step": 5078 }, { "epoch": 0.4, "learning_rate": 9.601740766878382e-05, "loss": 3.0783, "step": 5079 }, { "epoch": 0.4, "learning_rate": 9.601662353955932e-05, "loss": 2.8247, "step": 5080 }, { "epoch": 0.4, "learning_rate": 9.601583941033483e-05, "loss": 2.7445, "step": 5081 }, { "epoch": 0.4, "learning_rate": 9.601505528111033e-05, "loss": 3.0472, "step": 5082 }, { "epoch": 0.4, "learning_rate": 9.601427115188584e-05, "loss": 2.974, "step": 5083 }, { "epoch": 0.4, "learning_rate": 9.601348702266134e-05, "loss": 3.1202, "step": 5084 }, { "epoch": 0.4, "learning_rate": 9.601270289343684e-05, "loss": 3.1567, "step": 5085 }, { "epoch": 0.4, "learning_rate": 9.601191876421235e-05, "loss": 3.1276, "step": 5086 }, { "epoch": 0.4, "learning_rate": 9.601113463498785e-05, "loss": 2.9458, "step": 5087 }, { "epoch": 0.4, "learning_rate": 9.601035050576335e-05, "loss": 3.1269, "step": 5088 }, { "epoch": 0.4, "learning_rate": 9.600956637653886e-05, "loss": 3.0943, "step": 5089 }, { "epoch": 0.4, "learning_rate": 9.600878224731436e-05, "loss": 2.7759, "step": 5090 }, { "epoch": 0.4, "learning_rate": 9.600799811808986e-05, "loss": 3.0868, "step": 5091 }, { "epoch": 0.4, "learning_rate": 9.600721398886537e-05, "loss": 2.8547, "step": 5092 }, { "epoch": 0.4, "learning_rate": 9.600642985964087e-05, "loss": 3.4053, "step": 5093 }, { "epoch": 0.4, "learning_rate": 9.600564573041638e-05, "loss": 3.383, "step": 5094 }, { "epoch": 0.4, "learning_rate": 9.600486160119188e-05, "loss": 2.9879, "step": 5095 }, { "epoch": 0.4, "learning_rate": 9.600407747196738e-05, "loss": 2.8347, "step": 5096 }, { "epoch": 0.4, "learning_rate": 9.600329334274289e-05, "loss": 3.0535, "step": 5097 }, { "epoch": 0.4, "learning_rate": 9.600250921351839e-05, "loss": 3.0204, "step": 5098 }, { "epoch": 0.4, "learning_rate": 9.60017250842939e-05, "loss": 2.9781, "step": 5099 }, { "epoch": 0.4, "learning_rate": 9.60009409550694e-05, "loss": 2.9376, "step": 5100 }, { "epoch": 0.4, "learning_rate": 9.60001568258449e-05, "loss": 2.945, "step": 5101 }, { "epoch": 0.4, "learning_rate": 9.59993726966204e-05, "loss": 3.1287, "step": 5102 }, { "epoch": 0.4, "learning_rate": 9.599858856739591e-05, "loss": 3.082, "step": 5103 }, { "epoch": 0.4, "learning_rate": 9.599780443817141e-05, "loss": 3.1282, "step": 5104 }, { "epoch": 0.4, "learning_rate": 9.599702030894692e-05, "loss": 2.8642, "step": 5105 }, { "epoch": 0.4, "learning_rate": 9.599623617972242e-05, "loss": 2.9381, "step": 5106 }, { "epoch": 0.4, "learning_rate": 9.599545205049792e-05, "loss": 3.1479, "step": 5107 }, { "epoch": 0.4, "learning_rate": 9.599466792127343e-05, "loss": 3.1615, "step": 5108 }, { "epoch": 0.4, "learning_rate": 9.599388379204893e-05, "loss": 2.8164, "step": 5109 }, { "epoch": 0.4, "learning_rate": 9.599309966282443e-05, "loss": 3.0741, "step": 5110 }, { "epoch": 0.4, "learning_rate": 9.599231553359994e-05, "loss": 3.361, "step": 5111 }, { "epoch": 0.4, "learning_rate": 9.599153140437544e-05, "loss": 3.1316, "step": 5112 }, { "epoch": 0.4, "learning_rate": 9.599074727515095e-05, "loss": 3.047, "step": 5113 }, { "epoch": 0.4, "learning_rate": 9.598996314592645e-05, "loss": 3.0963, "step": 5114 }, { "epoch": 0.4, "learning_rate": 9.598917901670195e-05, "loss": 3.0061, "step": 5115 }, { "epoch": 0.4, "learning_rate": 9.598839488747746e-05, "loss": 3.0226, "step": 5116 }, { "epoch": 0.4, "learning_rate": 9.598761075825296e-05, "loss": 3.3681, "step": 5117 }, { "epoch": 0.4, "learning_rate": 9.598682662902846e-05, "loss": 2.9183, "step": 5118 }, { "epoch": 0.4, "learning_rate": 9.598604249980397e-05, "loss": 3.3408, "step": 5119 }, { "epoch": 0.4, "learning_rate": 9.598525837057947e-05, "loss": 3.2365, "step": 5120 }, { "epoch": 0.4, "learning_rate": 9.598447424135499e-05, "loss": 2.8901, "step": 5121 }, { "epoch": 0.4, "learning_rate": 9.598369011213049e-05, "loss": 2.9112, "step": 5122 }, { "epoch": 0.4, "learning_rate": 9.5982905982906e-05, "loss": 2.9781, "step": 5123 }, { "epoch": 0.4, "learning_rate": 9.59821218536815e-05, "loss": 3.1, "step": 5124 }, { "epoch": 0.4, "learning_rate": 9.5981337724457e-05, "loss": 3.2045, "step": 5125 }, { "epoch": 0.4, "learning_rate": 9.598055359523251e-05, "loss": 3.1753, "step": 5126 }, { "epoch": 0.4, "learning_rate": 9.597976946600801e-05, "loss": 3.1287, "step": 5127 }, { "epoch": 0.4, "learning_rate": 9.597898533678351e-05, "loss": 2.9801, "step": 5128 }, { "epoch": 0.4, "learning_rate": 9.597820120755902e-05, "loss": 3.1631, "step": 5129 }, { "epoch": 0.4, "learning_rate": 9.597741707833452e-05, "loss": 2.7597, "step": 5130 }, { "epoch": 0.4, "learning_rate": 9.597663294911001e-05, "loss": 3.3156, "step": 5131 }, { "epoch": 0.4, "learning_rate": 9.597584881988552e-05, "loss": 3.2329, "step": 5132 }, { "epoch": 0.4, "learning_rate": 9.597506469066102e-05, "loss": 3.1067, "step": 5133 }, { "epoch": 0.4, "learning_rate": 9.597428056143652e-05, "loss": 2.9706, "step": 5134 }, { "epoch": 0.4, "learning_rate": 9.597349643221203e-05, "loss": 2.986, "step": 5135 }, { "epoch": 0.4, "learning_rate": 9.597271230298753e-05, "loss": 2.8983, "step": 5136 }, { "epoch": 0.4, "learning_rate": 9.597192817376303e-05, "loss": 2.8041, "step": 5137 }, { "epoch": 0.4, "learning_rate": 9.597114404453854e-05, "loss": 2.9533, "step": 5138 }, { "epoch": 0.4, "learning_rate": 9.597035991531404e-05, "loss": 3.0495, "step": 5139 }, { "epoch": 0.4, "learning_rate": 9.596957578608955e-05, "loss": 2.8785, "step": 5140 }, { "epoch": 0.4, "learning_rate": 9.596879165686505e-05, "loss": 2.9584, "step": 5141 }, { "epoch": 0.4, "learning_rate": 9.596800752764055e-05, "loss": 3.0697, "step": 5142 }, { "epoch": 0.4, "learning_rate": 9.596722339841606e-05, "loss": 3.1513, "step": 5143 }, { "epoch": 0.4, "learning_rate": 9.596643926919156e-05, "loss": 3.2924, "step": 5144 }, { "epoch": 0.4, "learning_rate": 9.596565513996706e-05, "loss": 3.3287, "step": 5145 }, { "epoch": 0.4, "learning_rate": 9.596487101074258e-05, "loss": 2.7887, "step": 5146 }, { "epoch": 0.4, "learning_rate": 9.596408688151808e-05, "loss": 2.9167, "step": 5147 }, { "epoch": 0.4, "learning_rate": 9.596330275229359e-05, "loss": 3.0337, "step": 5148 }, { "epoch": 0.4, "learning_rate": 9.596251862306909e-05, "loss": 3.1065, "step": 5149 }, { "epoch": 0.4, "learning_rate": 9.59617344938446e-05, "loss": 2.9816, "step": 5150 }, { "epoch": 0.4, "learning_rate": 9.59609503646201e-05, "loss": 3.0258, "step": 5151 }, { "epoch": 0.4, "learning_rate": 9.59601662353956e-05, "loss": 3.2749, "step": 5152 }, { "epoch": 0.4, "learning_rate": 9.59593821061711e-05, "loss": 3.1398, "step": 5153 }, { "epoch": 0.4, "learning_rate": 9.595859797694661e-05, "loss": 3.4612, "step": 5154 }, { "epoch": 0.4, "learning_rate": 9.595781384772211e-05, "loss": 2.9308, "step": 5155 }, { "epoch": 0.4, "learning_rate": 9.595702971849762e-05, "loss": 3.2733, "step": 5156 }, { "epoch": 0.4, "learning_rate": 9.595624558927312e-05, "loss": 3.0336, "step": 5157 }, { "epoch": 0.4, "learning_rate": 9.595546146004862e-05, "loss": 3.0723, "step": 5158 }, { "epoch": 0.4, "learning_rate": 9.595467733082413e-05, "loss": 3.1843, "step": 5159 }, { "epoch": 0.4, "learning_rate": 9.595389320159963e-05, "loss": 2.8776, "step": 5160 }, { "epoch": 0.4, "learning_rate": 9.595310907237514e-05, "loss": 3.0315, "step": 5161 }, { "epoch": 0.4, "learning_rate": 9.595232494315064e-05, "loss": 2.887, "step": 5162 }, { "epoch": 0.4, "learning_rate": 9.595154081392614e-05, "loss": 2.8132, "step": 5163 }, { "epoch": 0.4, "learning_rate": 9.595075668470165e-05, "loss": 3.2354, "step": 5164 }, { "epoch": 0.41, "learning_rate": 9.594997255547715e-05, "loss": 3.4305, "step": 5165 }, { "epoch": 0.41, "learning_rate": 9.594918842625265e-05, "loss": 3.0658, "step": 5166 }, { "epoch": 0.41, "learning_rate": 9.594840429702814e-05, "loss": 3.0944, "step": 5167 }, { "epoch": 0.41, "learning_rate": 9.594762016780365e-05, "loss": 3.0721, "step": 5168 }, { "epoch": 0.41, "learning_rate": 9.594683603857915e-05, "loss": 2.7802, "step": 5169 }, { "epoch": 0.41, "learning_rate": 9.594605190935466e-05, "loss": 3.0331, "step": 5170 }, { "epoch": 0.41, "learning_rate": 9.594526778013016e-05, "loss": 2.8725, "step": 5171 }, { "epoch": 0.41, "learning_rate": 9.594448365090568e-05, "loss": 3.0026, "step": 5172 }, { "epoch": 0.41, "learning_rate": 9.594369952168118e-05, "loss": 3.0493, "step": 5173 }, { "epoch": 0.41, "learning_rate": 9.594291539245668e-05, "loss": 3.1623, "step": 5174 }, { "epoch": 0.41, "learning_rate": 9.594213126323219e-05, "loss": 2.9925, "step": 5175 }, { "epoch": 0.41, "learning_rate": 9.594134713400769e-05, "loss": 2.8688, "step": 5176 }, { "epoch": 0.41, "learning_rate": 9.59405630047832e-05, "loss": 3.1644, "step": 5177 }, { "epoch": 0.41, "learning_rate": 9.59397788755587e-05, "loss": 2.9291, "step": 5178 }, { "epoch": 0.41, "learning_rate": 9.59389947463342e-05, "loss": 3.2357, "step": 5179 }, { "epoch": 0.41, "learning_rate": 9.59382106171097e-05, "loss": 3.2784, "step": 5180 }, { "epoch": 0.41, "learning_rate": 9.593742648788521e-05, "loss": 2.8757, "step": 5181 }, { "epoch": 0.41, "learning_rate": 9.593664235866071e-05, "loss": 2.9527, "step": 5182 }, { "epoch": 0.41, "learning_rate": 9.593585822943622e-05, "loss": 3.0334, "step": 5183 }, { "epoch": 0.41, "learning_rate": 9.593507410021172e-05, "loss": 3.0325, "step": 5184 }, { "epoch": 0.41, "learning_rate": 9.593428997098722e-05, "loss": 2.9374, "step": 5185 }, { "epoch": 0.41, "learning_rate": 9.593350584176273e-05, "loss": 3.1178, "step": 5186 }, { "epoch": 0.41, "learning_rate": 9.593272171253823e-05, "loss": 2.9472, "step": 5187 }, { "epoch": 0.41, "learning_rate": 9.593193758331374e-05, "loss": 3.1534, "step": 5188 }, { "epoch": 0.41, "learning_rate": 9.593115345408924e-05, "loss": 3.1031, "step": 5189 }, { "epoch": 0.41, "learning_rate": 9.593036932486474e-05, "loss": 3.1957, "step": 5190 }, { "epoch": 0.41, "learning_rate": 9.592958519564025e-05, "loss": 3.2039, "step": 5191 }, { "epoch": 0.41, "learning_rate": 9.592880106641575e-05, "loss": 2.907, "step": 5192 }, { "epoch": 0.41, "learning_rate": 9.592801693719125e-05, "loss": 3.2118, "step": 5193 }, { "epoch": 0.41, "learning_rate": 9.592723280796676e-05, "loss": 2.9798, "step": 5194 }, { "epoch": 0.41, "learning_rate": 9.592644867874226e-05, "loss": 2.9241, "step": 5195 }, { "epoch": 0.41, "learning_rate": 9.592566454951776e-05, "loss": 2.9411, "step": 5196 }, { "epoch": 0.41, "learning_rate": 9.592488042029327e-05, "loss": 3.1408, "step": 5197 }, { "epoch": 0.41, "learning_rate": 9.592409629106877e-05, "loss": 2.9101, "step": 5198 }, { "epoch": 0.41, "learning_rate": 9.592331216184428e-05, "loss": 2.9887, "step": 5199 }, { "epoch": 0.41, "learning_rate": 9.592252803261978e-05, "loss": 3.3016, "step": 5200 }, { "epoch": 0.41, "learning_rate": 9.592174390339528e-05, "loss": 2.869, "step": 5201 }, { "epoch": 0.41, "learning_rate": 9.592095977417079e-05, "loss": 3.0211, "step": 5202 }, { "epoch": 0.41, "learning_rate": 9.592017564494629e-05, "loss": 3.1016, "step": 5203 }, { "epoch": 0.41, "learning_rate": 9.59193915157218e-05, "loss": 2.9104, "step": 5204 }, { "epoch": 0.41, "learning_rate": 9.59186073864973e-05, "loss": 3.2006, "step": 5205 }, { "epoch": 0.41, "learning_rate": 9.59178232572728e-05, "loss": 3.0032, "step": 5206 }, { "epoch": 0.41, "learning_rate": 9.59170391280483e-05, "loss": 3.3262, "step": 5207 }, { "epoch": 0.41, "learning_rate": 9.591625499882381e-05, "loss": 3.005, "step": 5208 }, { "epoch": 0.41, "learning_rate": 9.591547086959931e-05, "loss": 2.7868, "step": 5209 }, { "epoch": 0.41, "learning_rate": 9.591468674037482e-05, "loss": 2.8163, "step": 5210 }, { "epoch": 0.41, "learning_rate": 9.591390261115032e-05, "loss": 3.0867, "step": 5211 }, { "epoch": 0.41, "learning_rate": 9.591311848192582e-05, "loss": 2.9728, "step": 5212 }, { "epoch": 0.41, "learning_rate": 9.591233435270133e-05, "loss": 3.1328, "step": 5213 }, { "epoch": 0.41, "learning_rate": 9.591155022347683e-05, "loss": 2.9631, "step": 5214 }, { "epoch": 0.41, "learning_rate": 9.591076609425233e-05, "loss": 3.2864, "step": 5215 }, { "epoch": 0.41, "learning_rate": 9.590998196502784e-05, "loss": 2.9666, "step": 5216 }, { "epoch": 0.41, "learning_rate": 9.590919783580334e-05, "loss": 3.309, "step": 5217 }, { "epoch": 0.41, "learning_rate": 9.590841370657885e-05, "loss": 3.1207, "step": 5218 }, { "epoch": 0.41, "learning_rate": 9.590762957735435e-05, "loss": 2.8641, "step": 5219 }, { "epoch": 0.41, "learning_rate": 9.590684544812985e-05, "loss": 3.0383, "step": 5220 }, { "epoch": 0.41, "learning_rate": 9.590606131890536e-05, "loss": 3.0234, "step": 5221 }, { "epoch": 0.41, "learning_rate": 9.590527718968086e-05, "loss": 2.8712, "step": 5222 }, { "epoch": 0.41, "learning_rate": 9.590449306045636e-05, "loss": 3.3452, "step": 5223 }, { "epoch": 0.41, "learning_rate": 9.590370893123187e-05, "loss": 3.0527, "step": 5224 }, { "epoch": 0.41, "learning_rate": 9.590292480200737e-05, "loss": 2.7306, "step": 5225 }, { "epoch": 0.41, "learning_rate": 9.590214067278287e-05, "loss": 2.7602, "step": 5226 }, { "epoch": 0.41, "learning_rate": 9.590135654355838e-05, "loss": 3.037, "step": 5227 }, { "epoch": 0.41, "learning_rate": 9.590057241433388e-05, "loss": 2.8089, "step": 5228 }, { "epoch": 0.41, "learning_rate": 9.58997882851094e-05, "loss": 3.2188, "step": 5229 }, { "epoch": 0.41, "learning_rate": 9.58990041558849e-05, "loss": 3.2384, "step": 5230 }, { "epoch": 0.41, "learning_rate": 9.589822002666041e-05, "loss": 2.9348, "step": 5231 }, { "epoch": 0.41, "learning_rate": 9.589743589743591e-05, "loss": 3.3017, "step": 5232 }, { "epoch": 0.41, "learning_rate": 9.589665176821141e-05, "loss": 3.0867, "step": 5233 }, { "epoch": 0.41, "learning_rate": 9.589586763898692e-05, "loss": 3.0088, "step": 5234 }, { "epoch": 0.41, "learning_rate": 9.589508350976242e-05, "loss": 3.0196, "step": 5235 }, { "epoch": 0.41, "learning_rate": 9.589429938053793e-05, "loss": 3.1286, "step": 5236 }, { "epoch": 0.41, "learning_rate": 9.589351525131343e-05, "loss": 3.2577, "step": 5237 }, { "epoch": 0.41, "learning_rate": 9.589273112208893e-05, "loss": 3.0144, "step": 5238 }, { "epoch": 0.41, "learning_rate": 9.589194699286442e-05, "loss": 2.9958, "step": 5239 }, { "epoch": 0.41, "learning_rate": 9.589116286363993e-05, "loss": 2.8047, "step": 5240 }, { "epoch": 0.41, "learning_rate": 9.589037873441543e-05, "loss": 3.0976, "step": 5241 }, { "epoch": 0.41, "learning_rate": 9.588959460519093e-05, "loss": 2.9233, "step": 5242 }, { "epoch": 0.41, "learning_rate": 9.588881047596644e-05, "loss": 3.1471, "step": 5243 }, { "epoch": 0.41, "learning_rate": 9.588802634674194e-05, "loss": 2.8696, "step": 5244 }, { "epoch": 0.41, "learning_rate": 9.588724221751744e-05, "loss": 2.9837, "step": 5245 }, { "epoch": 0.41, "learning_rate": 9.588645808829295e-05, "loss": 2.8987, "step": 5246 }, { "epoch": 0.41, "learning_rate": 9.588567395906845e-05, "loss": 3.0737, "step": 5247 }, { "epoch": 0.41, "learning_rate": 9.588488982984396e-05, "loss": 2.9269, "step": 5248 }, { "epoch": 0.41, "learning_rate": 9.588410570061946e-05, "loss": 2.9487, "step": 5249 }, { "epoch": 0.41, "learning_rate": 9.588332157139496e-05, "loss": 3.2019, "step": 5250 }, { "epoch": 0.41, "learning_rate": 9.588253744217047e-05, "loss": 2.8894, "step": 5251 }, { "epoch": 0.41, "learning_rate": 9.588175331294597e-05, "loss": 3.446, "step": 5252 }, { "epoch": 0.41, "learning_rate": 9.588096918372147e-05, "loss": 3.0324, "step": 5253 }, { "epoch": 0.41, "learning_rate": 9.588018505449699e-05, "loss": 3.2628, "step": 5254 }, { "epoch": 0.41, "learning_rate": 9.58794009252725e-05, "loss": 3.0328, "step": 5255 }, { "epoch": 0.41, "learning_rate": 9.5878616796048e-05, "loss": 3.0121, "step": 5256 }, { "epoch": 0.41, "learning_rate": 9.58778326668235e-05, "loss": 3.2933, "step": 5257 }, { "epoch": 0.41, "learning_rate": 9.5877048537599e-05, "loss": 2.9056, "step": 5258 }, { "epoch": 0.41, "learning_rate": 9.587626440837451e-05, "loss": 2.9296, "step": 5259 }, { "epoch": 0.41, "learning_rate": 9.587548027915001e-05, "loss": 3.002, "step": 5260 }, { "epoch": 0.41, "learning_rate": 9.587469614992552e-05, "loss": 3.0777, "step": 5261 }, { "epoch": 0.41, "learning_rate": 9.587391202070102e-05, "loss": 3.0576, "step": 5262 }, { "epoch": 0.41, "learning_rate": 9.587312789147652e-05, "loss": 2.9627, "step": 5263 }, { "epoch": 0.41, "learning_rate": 9.587234376225203e-05, "loss": 2.941, "step": 5264 }, { "epoch": 0.41, "learning_rate": 9.587155963302753e-05, "loss": 3.1194, "step": 5265 }, { "epoch": 0.41, "learning_rate": 9.587077550380304e-05, "loss": 3.2858, "step": 5266 }, { "epoch": 0.41, "learning_rate": 9.586999137457854e-05, "loss": 3.1299, "step": 5267 }, { "epoch": 0.41, "learning_rate": 9.586920724535404e-05, "loss": 2.9696, "step": 5268 }, { "epoch": 0.41, "learning_rate": 9.586842311612955e-05, "loss": 3.0955, "step": 5269 }, { "epoch": 0.41, "learning_rate": 9.586763898690505e-05, "loss": 2.9971, "step": 5270 }, { "epoch": 0.41, "learning_rate": 9.586685485768055e-05, "loss": 3.3155, "step": 5271 }, { "epoch": 0.41, "learning_rate": 9.586607072845606e-05, "loss": 3.2201, "step": 5272 }, { "epoch": 0.41, "learning_rate": 9.586528659923156e-05, "loss": 2.8533, "step": 5273 }, { "epoch": 0.41, "learning_rate": 9.586450247000707e-05, "loss": 3.0853, "step": 5274 }, { "epoch": 0.41, "learning_rate": 9.586371834078256e-05, "loss": 3.4219, "step": 5275 }, { "epoch": 0.41, "learning_rate": 9.586293421155806e-05, "loss": 3.0922, "step": 5276 }, { "epoch": 0.41, "learning_rate": 9.586215008233356e-05, "loss": 3.0781, "step": 5277 }, { "epoch": 0.41, "learning_rate": 9.586136595310907e-05, "loss": 2.9837, "step": 5278 }, { "epoch": 0.41, "learning_rate": 9.586058182388457e-05, "loss": 3.1669, "step": 5279 }, { "epoch": 0.41, "learning_rate": 9.585979769466009e-05, "loss": 2.8833, "step": 5280 }, { "epoch": 0.41, "learning_rate": 9.585901356543559e-05, "loss": 2.9748, "step": 5281 }, { "epoch": 0.41, "learning_rate": 9.58582294362111e-05, "loss": 2.9857, "step": 5282 }, { "epoch": 0.41, "learning_rate": 9.58574453069866e-05, "loss": 3.1836, "step": 5283 }, { "epoch": 0.41, "learning_rate": 9.58566611777621e-05, "loss": 3.0226, "step": 5284 }, { "epoch": 0.41, "learning_rate": 9.58558770485376e-05, "loss": 3.2492, "step": 5285 }, { "epoch": 0.41, "learning_rate": 9.585509291931311e-05, "loss": 3.2015, "step": 5286 }, { "epoch": 0.41, "learning_rate": 9.585430879008861e-05, "loss": 2.8794, "step": 5287 }, { "epoch": 0.41, "learning_rate": 9.585352466086412e-05, "loss": 3.0193, "step": 5288 }, { "epoch": 0.41, "learning_rate": 9.585274053163962e-05, "loss": 2.8565, "step": 5289 }, { "epoch": 0.41, "learning_rate": 9.585195640241512e-05, "loss": 3.0683, "step": 5290 }, { "epoch": 0.41, "learning_rate": 9.585117227319063e-05, "loss": 3.3859, "step": 5291 }, { "epoch": 0.41, "learning_rate": 9.585038814396613e-05, "loss": 2.8928, "step": 5292 }, { "epoch": 0.42, "learning_rate": 9.584960401474163e-05, "loss": 3.0969, "step": 5293 }, { "epoch": 0.42, "learning_rate": 9.584881988551714e-05, "loss": 3.5926, "step": 5294 }, { "epoch": 0.42, "learning_rate": 9.584803575629264e-05, "loss": 3.2393, "step": 5295 }, { "epoch": 0.42, "learning_rate": 9.584725162706815e-05, "loss": 2.9999, "step": 5296 }, { "epoch": 0.42, "learning_rate": 9.584646749784365e-05, "loss": 3.136, "step": 5297 }, { "epoch": 0.42, "learning_rate": 9.584568336861915e-05, "loss": 2.9494, "step": 5298 }, { "epoch": 0.42, "learning_rate": 9.584489923939466e-05, "loss": 3.1678, "step": 5299 }, { "epoch": 0.42, "learning_rate": 9.584411511017016e-05, "loss": 2.9924, "step": 5300 }, { "epoch": 0.42, "learning_rate": 9.584333098094566e-05, "loss": 3.0044, "step": 5301 }, { "epoch": 0.42, "learning_rate": 9.584254685172117e-05, "loss": 3.2467, "step": 5302 }, { "epoch": 0.42, "learning_rate": 9.584176272249667e-05, "loss": 3.1675, "step": 5303 }, { "epoch": 0.42, "learning_rate": 9.584097859327218e-05, "loss": 3.1807, "step": 5304 }, { "epoch": 0.42, "learning_rate": 9.584019446404768e-05, "loss": 3.0194, "step": 5305 }, { "epoch": 0.42, "learning_rate": 9.583941033482318e-05, "loss": 2.8368, "step": 5306 }, { "epoch": 0.42, "learning_rate": 9.583862620559869e-05, "loss": 3.0768, "step": 5307 }, { "epoch": 0.42, "learning_rate": 9.583784207637419e-05, "loss": 2.8677, "step": 5308 }, { "epoch": 0.42, "learning_rate": 9.58370579471497e-05, "loss": 3.0626, "step": 5309 }, { "epoch": 0.42, "learning_rate": 9.58362738179252e-05, "loss": 2.9115, "step": 5310 }, { "epoch": 0.42, "learning_rate": 9.58354896887007e-05, "loss": 3.2027, "step": 5311 }, { "epoch": 0.42, "learning_rate": 9.58347055594762e-05, "loss": 3.0644, "step": 5312 }, { "epoch": 0.42, "learning_rate": 9.583392143025171e-05, "loss": 3.2997, "step": 5313 }, { "epoch": 0.42, "learning_rate": 9.583313730102721e-05, "loss": 3.3963, "step": 5314 }, { "epoch": 0.42, "learning_rate": 9.583235317180272e-05, "loss": 3.1787, "step": 5315 }, { "epoch": 0.42, "learning_rate": 9.583156904257822e-05, "loss": 3.2667, "step": 5316 }, { "epoch": 0.42, "learning_rate": 9.583078491335372e-05, "loss": 3.1495, "step": 5317 }, { "epoch": 0.42, "learning_rate": 9.583000078412923e-05, "loss": 3.2016, "step": 5318 }, { "epoch": 0.42, "learning_rate": 9.582921665490473e-05, "loss": 3.1057, "step": 5319 }, { "epoch": 0.42, "learning_rate": 9.582843252568023e-05, "loss": 3.0703, "step": 5320 }, { "epoch": 0.42, "learning_rate": 9.582764839645574e-05, "loss": 3.1179, "step": 5321 }, { "epoch": 0.42, "learning_rate": 9.582686426723124e-05, "loss": 3.0807, "step": 5322 }, { "epoch": 0.42, "learning_rate": 9.582608013800675e-05, "loss": 3.0393, "step": 5323 }, { "epoch": 0.42, "learning_rate": 9.582529600878225e-05, "loss": 3.3012, "step": 5324 }, { "epoch": 0.42, "learning_rate": 9.582451187955775e-05, "loss": 2.9807, "step": 5325 }, { "epoch": 0.42, "learning_rate": 9.582372775033326e-05, "loss": 3.0631, "step": 5326 }, { "epoch": 0.42, "learning_rate": 9.582294362110876e-05, "loss": 3.0052, "step": 5327 }, { "epoch": 0.42, "learning_rate": 9.582215949188426e-05, "loss": 3.5159, "step": 5328 }, { "epoch": 0.42, "learning_rate": 9.582137536265977e-05, "loss": 3.1248, "step": 5329 }, { "epoch": 0.42, "learning_rate": 9.582059123343527e-05, "loss": 3.1168, "step": 5330 }, { "epoch": 0.42, "learning_rate": 9.581980710421077e-05, "loss": 2.9382, "step": 5331 }, { "epoch": 0.42, "learning_rate": 9.581902297498628e-05, "loss": 2.9301, "step": 5332 }, { "epoch": 0.42, "learning_rate": 9.581823884576178e-05, "loss": 2.9574, "step": 5333 }, { "epoch": 0.42, "learning_rate": 9.581745471653729e-05, "loss": 3.0682, "step": 5334 }, { "epoch": 0.42, "learning_rate": 9.581667058731279e-05, "loss": 3.1701, "step": 5335 }, { "epoch": 0.42, "learning_rate": 9.581588645808829e-05, "loss": 2.9804, "step": 5336 }, { "epoch": 0.42, "learning_rate": 9.581510232886381e-05, "loss": 3.1133, "step": 5337 }, { "epoch": 0.42, "learning_rate": 9.581431819963931e-05, "loss": 2.947, "step": 5338 }, { "epoch": 0.42, "learning_rate": 9.581353407041482e-05, "loss": 3.1432, "step": 5339 }, { "epoch": 0.42, "learning_rate": 9.581274994119032e-05, "loss": 3.1664, "step": 5340 }, { "epoch": 0.42, "learning_rate": 9.581196581196583e-05, "loss": 3.078, "step": 5341 }, { "epoch": 0.42, "learning_rate": 9.581118168274133e-05, "loss": 3.28, "step": 5342 }, { "epoch": 0.42, "learning_rate": 9.581039755351683e-05, "loss": 3.1173, "step": 5343 }, { "epoch": 0.42, "learning_rate": 9.580961342429234e-05, "loss": 2.8845, "step": 5344 }, { "epoch": 0.42, "learning_rate": 9.580882929506784e-05, "loss": 2.9426, "step": 5345 }, { "epoch": 0.42, "learning_rate": 9.580804516584334e-05, "loss": 3.1019, "step": 5346 }, { "epoch": 0.42, "learning_rate": 9.580726103661883e-05, "loss": 2.869, "step": 5347 }, { "epoch": 0.42, "learning_rate": 9.580647690739434e-05, "loss": 2.901, "step": 5348 }, { "epoch": 0.42, "learning_rate": 9.580569277816984e-05, "loss": 3.2738, "step": 5349 }, { "epoch": 0.42, "learning_rate": 9.580490864894534e-05, "loss": 3.0041, "step": 5350 }, { "epoch": 0.42, "learning_rate": 9.580412451972085e-05, "loss": 2.8156, "step": 5351 }, { "epoch": 0.42, "learning_rate": 9.580334039049635e-05, "loss": 3.0601, "step": 5352 }, { "epoch": 0.42, "learning_rate": 9.580255626127186e-05, "loss": 3.0883, "step": 5353 }, { "epoch": 0.42, "learning_rate": 9.580177213204736e-05, "loss": 3.0956, "step": 5354 }, { "epoch": 0.42, "learning_rate": 9.580098800282286e-05, "loss": 3.0381, "step": 5355 }, { "epoch": 0.42, "learning_rate": 9.580020387359837e-05, "loss": 3.0514, "step": 5356 }, { "epoch": 0.42, "learning_rate": 9.579941974437387e-05, "loss": 3.1271, "step": 5357 }, { "epoch": 0.42, "learning_rate": 9.579863561514937e-05, "loss": 3.036, "step": 5358 }, { "epoch": 0.42, "learning_rate": 9.579785148592488e-05, "loss": 2.9618, "step": 5359 }, { "epoch": 0.42, "learning_rate": 9.579706735670038e-05, "loss": 3.377, "step": 5360 }, { "epoch": 0.42, "learning_rate": 9.579628322747589e-05, "loss": 2.9142, "step": 5361 }, { "epoch": 0.42, "learning_rate": 9.57954990982514e-05, "loss": 2.9081, "step": 5362 }, { "epoch": 0.42, "learning_rate": 9.57947149690269e-05, "loss": 2.9618, "step": 5363 }, { "epoch": 0.42, "learning_rate": 9.579393083980241e-05, "loss": 3.0921, "step": 5364 }, { "epoch": 0.42, "learning_rate": 9.579314671057791e-05, "loss": 3.0859, "step": 5365 }, { "epoch": 0.42, "learning_rate": 9.579236258135342e-05, "loss": 2.915, "step": 5366 }, { "epoch": 0.42, "learning_rate": 9.579157845212892e-05, "loss": 2.9667, "step": 5367 }, { "epoch": 0.42, "learning_rate": 9.579079432290442e-05, "loss": 3.0461, "step": 5368 }, { "epoch": 0.42, "learning_rate": 9.579001019367993e-05, "loss": 3.0747, "step": 5369 }, { "epoch": 0.42, "learning_rate": 9.578922606445543e-05, "loss": 2.9253, "step": 5370 }, { "epoch": 0.42, "learning_rate": 9.578844193523094e-05, "loss": 2.8445, "step": 5371 }, { "epoch": 0.42, "learning_rate": 9.578765780600644e-05, "loss": 3.0896, "step": 5372 }, { "epoch": 0.42, "learning_rate": 9.578687367678194e-05, "loss": 3.1493, "step": 5373 }, { "epoch": 0.42, "learning_rate": 9.578608954755745e-05, "loss": 2.8528, "step": 5374 }, { "epoch": 0.42, "learning_rate": 9.578530541833295e-05, "loss": 3.0725, "step": 5375 }, { "epoch": 0.42, "learning_rate": 9.578452128910845e-05, "loss": 3.0769, "step": 5376 }, { "epoch": 0.42, "learning_rate": 9.578373715988396e-05, "loss": 3.0775, "step": 5377 }, { "epoch": 0.42, "learning_rate": 9.578295303065946e-05, "loss": 3.0953, "step": 5378 }, { "epoch": 0.42, "learning_rate": 9.578216890143496e-05, "loss": 2.9652, "step": 5379 }, { "epoch": 0.42, "learning_rate": 9.578138477221047e-05, "loss": 3.2616, "step": 5380 }, { "epoch": 0.42, "learning_rate": 9.578060064298597e-05, "loss": 2.6841, "step": 5381 }, { "epoch": 0.42, "learning_rate": 9.577981651376148e-05, "loss": 3.2427, "step": 5382 }, { "epoch": 0.42, "learning_rate": 9.577903238453697e-05, "loss": 3.1284, "step": 5383 }, { "epoch": 0.42, "learning_rate": 9.577824825531247e-05, "loss": 3.0523, "step": 5384 }, { "epoch": 0.42, "learning_rate": 9.577746412608797e-05, "loss": 2.9182, "step": 5385 }, { "epoch": 0.42, "learning_rate": 9.577667999686348e-05, "loss": 2.9413, "step": 5386 }, { "epoch": 0.42, "learning_rate": 9.577589586763898e-05, "loss": 3.1101, "step": 5387 }, { "epoch": 0.42, "learning_rate": 9.57751117384145e-05, "loss": 2.8773, "step": 5388 }, { "epoch": 0.42, "learning_rate": 9.577432760919e-05, "loss": 3.0947, "step": 5389 }, { "epoch": 0.42, "learning_rate": 9.57735434799655e-05, "loss": 3.4225, "step": 5390 }, { "epoch": 0.42, "learning_rate": 9.577275935074101e-05, "loss": 3.2784, "step": 5391 }, { "epoch": 0.42, "learning_rate": 9.577197522151651e-05, "loss": 3.1746, "step": 5392 }, { "epoch": 0.42, "learning_rate": 9.577119109229202e-05, "loss": 2.98, "step": 5393 }, { "epoch": 0.42, "learning_rate": 9.577040696306752e-05, "loss": 2.9278, "step": 5394 }, { "epoch": 0.42, "learning_rate": 9.576962283384302e-05, "loss": 2.9783, "step": 5395 }, { "epoch": 0.42, "learning_rate": 9.576883870461853e-05, "loss": 3.1327, "step": 5396 }, { "epoch": 0.42, "learning_rate": 9.576805457539403e-05, "loss": 3.4306, "step": 5397 }, { "epoch": 0.42, "learning_rate": 9.576727044616953e-05, "loss": 3.1013, "step": 5398 }, { "epoch": 0.42, "learning_rate": 9.576648631694504e-05, "loss": 3.166, "step": 5399 }, { "epoch": 0.42, "learning_rate": 9.576570218772054e-05, "loss": 2.8559, "step": 5400 }, { "epoch": 0.42, "learning_rate": 9.576491805849605e-05, "loss": 2.95, "step": 5401 }, { "epoch": 0.42, "learning_rate": 9.576413392927155e-05, "loss": 3.0394, "step": 5402 }, { "epoch": 0.42, "learning_rate": 9.576334980004705e-05, "loss": 3.2049, "step": 5403 }, { "epoch": 0.42, "learning_rate": 9.576256567082256e-05, "loss": 3.1171, "step": 5404 }, { "epoch": 0.42, "learning_rate": 9.576178154159806e-05, "loss": 3.1053, "step": 5405 }, { "epoch": 0.42, "learning_rate": 9.576099741237356e-05, "loss": 3.036, "step": 5406 }, { "epoch": 0.42, "learning_rate": 9.576021328314907e-05, "loss": 3.0609, "step": 5407 }, { "epoch": 0.42, "learning_rate": 9.575942915392457e-05, "loss": 3.0355, "step": 5408 }, { "epoch": 0.42, "learning_rate": 9.575864502470008e-05, "loss": 2.9977, "step": 5409 }, { "epoch": 0.42, "learning_rate": 9.575786089547558e-05, "loss": 3.279, "step": 5410 }, { "epoch": 0.42, "learning_rate": 9.575707676625108e-05, "loss": 3.1274, "step": 5411 }, { "epoch": 0.42, "learning_rate": 9.575629263702659e-05, "loss": 2.8837, "step": 5412 }, { "epoch": 0.42, "learning_rate": 9.575550850780209e-05, "loss": 3.1156, "step": 5413 }, { "epoch": 0.42, "learning_rate": 9.57547243785776e-05, "loss": 2.821, "step": 5414 }, { "epoch": 0.42, "learning_rate": 9.57539402493531e-05, "loss": 2.9348, "step": 5415 }, { "epoch": 0.42, "learning_rate": 9.57531561201286e-05, "loss": 3.2241, "step": 5416 }, { "epoch": 0.42, "learning_rate": 9.57523719909041e-05, "loss": 2.8888, "step": 5417 }, { "epoch": 0.42, "learning_rate": 9.575158786167961e-05, "loss": 3.2865, "step": 5418 }, { "epoch": 0.42, "learning_rate": 9.575080373245511e-05, "loss": 2.8943, "step": 5419 }, { "epoch": 0.42, "learning_rate": 9.575001960323062e-05, "loss": 3.0517, "step": 5420 }, { "epoch": 0.43, "learning_rate": 9.574923547400612e-05, "loss": 2.8569, "step": 5421 }, { "epoch": 0.43, "learning_rate": 9.574845134478162e-05, "loss": 3.3237, "step": 5422 }, { "epoch": 0.43, "learning_rate": 9.574766721555713e-05, "loss": 3.1219, "step": 5423 }, { "epoch": 0.43, "learning_rate": 9.574688308633263e-05, "loss": 2.9073, "step": 5424 }, { "epoch": 0.43, "learning_rate": 9.574609895710813e-05, "loss": 3.1052, "step": 5425 }, { "epoch": 0.43, "learning_rate": 9.574531482788364e-05, "loss": 2.8412, "step": 5426 }, { "epoch": 0.43, "learning_rate": 9.574453069865914e-05, "loss": 2.9481, "step": 5427 }, { "epoch": 0.43, "learning_rate": 9.574374656943464e-05, "loss": 3.1571, "step": 5428 }, { "epoch": 0.43, "learning_rate": 9.574296244021015e-05, "loss": 3.0127, "step": 5429 }, { "epoch": 0.43, "learning_rate": 9.574217831098565e-05, "loss": 2.9274, "step": 5430 }, { "epoch": 0.43, "learning_rate": 9.574139418176116e-05, "loss": 2.9059, "step": 5431 }, { "epoch": 0.43, "learning_rate": 9.574061005253666e-05, "loss": 2.9094, "step": 5432 }, { "epoch": 0.43, "learning_rate": 9.573982592331216e-05, "loss": 3.3059, "step": 5433 }, { "epoch": 0.43, "learning_rate": 9.573904179408767e-05, "loss": 3.1236, "step": 5434 }, { "epoch": 0.43, "learning_rate": 9.573825766486317e-05, "loss": 3.3704, "step": 5435 }, { "epoch": 0.43, "learning_rate": 9.573747353563867e-05, "loss": 3.1541, "step": 5436 }, { "epoch": 0.43, "learning_rate": 9.573668940641418e-05, "loss": 3.1985, "step": 5437 }, { "epoch": 0.43, "learning_rate": 9.573590527718968e-05, "loss": 2.8257, "step": 5438 }, { "epoch": 0.43, "learning_rate": 9.573512114796519e-05, "loss": 3.0578, "step": 5439 }, { "epoch": 0.43, "learning_rate": 9.573433701874069e-05, "loss": 3.0457, "step": 5440 }, { "epoch": 0.43, "learning_rate": 9.573355288951619e-05, "loss": 2.7689, "step": 5441 }, { "epoch": 0.43, "learning_rate": 9.57327687602917e-05, "loss": 3.0632, "step": 5442 }, { "epoch": 0.43, "learning_rate": 9.57319846310672e-05, "loss": 2.9116, "step": 5443 }, { "epoch": 0.43, "learning_rate": 9.57312005018427e-05, "loss": 3.0704, "step": 5444 }, { "epoch": 0.43, "learning_rate": 9.573041637261822e-05, "loss": 3.1146, "step": 5445 }, { "epoch": 0.43, "learning_rate": 9.572963224339372e-05, "loss": 3.4189, "step": 5446 }, { "epoch": 0.43, "learning_rate": 9.572884811416923e-05, "loss": 3.0262, "step": 5447 }, { "epoch": 0.43, "learning_rate": 9.572806398494473e-05, "loss": 2.9899, "step": 5448 }, { "epoch": 0.43, "learning_rate": 9.572727985572024e-05, "loss": 3.1574, "step": 5449 }, { "epoch": 0.43, "learning_rate": 9.572649572649574e-05, "loss": 3.3298, "step": 5450 }, { "epoch": 0.43, "learning_rate": 9.572571159727124e-05, "loss": 2.9013, "step": 5451 }, { "epoch": 0.43, "learning_rate": 9.572492746804675e-05, "loss": 3.1385, "step": 5452 }, { "epoch": 0.43, "learning_rate": 9.572414333882225e-05, "loss": 3.1469, "step": 5453 }, { "epoch": 0.43, "learning_rate": 9.572335920959775e-05, "loss": 2.9621, "step": 5454 }, { "epoch": 0.43, "learning_rate": 9.572257508037324e-05, "loss": 2.9553, "step": 5455 }, { "epoch": 0.43, "learning_rate": 9.572179095114875e-05, "loss": 3.4143, "step": 5456 }, { "epoch": 0.43, "learning_rate": 9.572100682192425e-05, "loss": 3.1398, "step": 5457 }, { "epoch": 0.43, "learning_rate": 9.572022269269976e-05, "loss": 3.1445, "step": 5458 }, { "epoch": 0.43, "learning_rate": 9.571943856347526e-05, "loss": 2.9224, "step": 5459 }, { "epoch": 0.43, "learning_rate": 9.571865443425076e-05, "loss": 2.9942, "step": 5460 }, { "epoch": 0.43, "learning_rate": 9.571787030502627e-05, "loss": 3.0514, "step": 5461 }, { "epoch": 0.43, "learning_rate": 9.571708617580177e-05, "loss": 3.1726, "step": 5462 }, { "epoch": 0.43, "learning_rate": 9.571630204657727e-05, "loss": 3.2977, "step": 5463 }, { "epoch": 0.43, "learning_rate": 9.571551791735278e-05, "loss": 3.095, "step": 5464 }, { "epoch": 0.43, "learning_rate": 9.571473378812828e-05, "loss": 3.2159, "step": 5465 }, { "epoch": 0.43, "learning_rate": 9.571394965890378e-05, "loss": 2.9276, "step": 5466 }, { "epoch": 0.43, "learning_rate": 9.571316552967929e-05, "loss": 2.978, "step": 5467 }, { "epoch": 0.43, "learning_rate": 9.571238140045479e-05, "loss": 3.3141, "step": 5468 }, { "epoch": 0.43, "learning_rate": 9.57115972712303e-05, "loss": 3.1743, "step": 5469 }, { "epoch": 0.43, "learning_rate": 9.571081314200581e-05, "loss": 3.1258, "step": 5470 }, { "epoch": 0.43, "learning_rate": 9.571002901278132e-05, "loss": 3.0794, "step": 5471 }, { "epoch": 0.43, "learning_rate": 9.570924488355682e-05, "loss": 3.109, "step": 5472 }, { "epoch": 0.43, "learning_rate": 9.570846075433232e-05, "loss": 2.8862, "step": 5473 }, { "epoch": 0.43, "learning_rate": 9.570767662510783e-05, "loss": 3.1666, "step": 5474 }, { "epoch": 0.43, "learning_rate": 9.570689249588333e-05, "loss": 3.1714, "step": 5475 }, { "epoch": 0.43, "learning_rate": 9.570610836665884e-05, "loss": 2.9881, "step": 5476 }, { "epoch": 0.43, "learning_rate": 9.570532423743434e-05, "loss": 3.1763, "step": 5477 }, { "epoch": 0.43, "learning_rate": 9.570454010820984e-05, "loss": 2.9983, "step": 5478 }, { "epoch": 0.43, "learning_rate": 9.570375597898535e-05, "loss": 3.105, "step": 5479 }, { "epoch": 0.43, "learning_rate": 9.570297184976085e-05, "loss": 2.9051, "step": 5480 }, { "epoch": 0.43, "learning_rate": 9.570218772053635e-05, "loss": 3.0056, "step": 5481 }, { "epoch": 0.43, "learning_rate": 9.570140359131186e-05, "loss": 3.1933, "step": 5482 }, { "epoch": 0.43, "learning_rate": 9.570061946208736e-05, "loss": 3.0664, "step": 5483 }, { "epoch": 0.43, "learning_rate": 9.569983533286286e-05, "loss": 2.9923, "step": 5484 }, { "epoch": 0.43, "learning_rate": 9.569905120363837e-05, "loss": 2.8304, "step": 5485 }, { "epoch": 0.43, "learning_rate": 9.569826707441387e-05, "loss": 2.8881, "step": 5486 }, { "epoch": 0.43, "learning_rate": 9.569748294518938e-05, "loss": 3.0714, "step": 5487 }, { "epoch": 0.43, "learning_rate": 9.569669881596488e-05, "loss": 3.205, "step": 5488 }, { "epoch": 0.43, "learning_rate": 9.569591468674038e-05, "loss": 3.1316, "step": 5489 }, { "epoch": 0.43, "learning_rate": 9.569513055751589e-05, "loss": 2.9794, "step": 5490 }, { "epoch": 0.43, "learning_rate": 9.569434642829138e-05, "loss": 3.0743, "step": 5491 }, { "epoch": 0.43, "learning_rate": 9.569356229906688e-05, "loss": 3.0218, "step": 5492 }, { "epoch": 0.43, "learning_rate": 9.569277816984238e-05, "loss": 3.1008, "step": 5493 }, { "epoch": 0.43, "learning_rate": 9.569199404061789e-05, "loss": 3.1525, "step": 5494 }, { "epoch": 0.43, "learning_rate": 9.569120991139339e-05, "loss": 3.0156, "step": 5495 }, { "epoch": 0.43, "learning_rate": 9.569042578216891e-05, "loss": 2.9397, "step": 5496 }, { "epoch": 0.43, "learning_rate": 9.568964165294441e-05, "loss": 3.164, "step": 5497 }, { "epoch": 0.43, "learning_rate": 9.568885752371992e-05, "loss": 2.9511, "step": 5498 }, { "epoch": 0.43, "learning_rate": 9.568807339449542e-05, "loss": 2.9906, "step": 5499 }, { "epoch": 0.43, "learning_rate": 9.568728926527092e-05, "loss": 2.8266, "step": 5500 }, { "epoch": 0.43, "learning_rate": 9.568650513604643e-05, "loss": 2.9161, "step": 5501 }, { "epoch": 0.43, "learning_rate": 9.568572100682193e-05, "loss": 3.2804, "step": 5502 }, { "epoch": 0.43, "learning_rate": 9.568493687759743e-05, "loss": 3.3728, "step": 5503 }, { "epoch": 0.43, "learning_rate": 9.568415274837294e-05, "loss": 3.0488, "step": 5504 }, { "epoch": 0.43, "learning_rate": 9.568336861914844e-05, "loss": 3.2644, "step": 5505 }, { "epoch": 0.43, "learning_rate": 9.568258448992395e-05, "loss": 2.9189, "step": 5506 }, { "epoch": 0.43, "learning_rate": 9.568180036069945e-05, "loss": 3.1864, "step": 5507 }, { "epoch": 0.43, "learning_rate": 9.568101623147495e-05, "loss": 2.8636, "step": 5508 }, { "epoch": 0.43, "learning_rate": 9.568023210225046e-05, "loss": 3.2473, "step": 5509 }, { "epoch": 0.43, "learning_rate": 9.567944797302596e-05, "loss": 2.8916, "step": 5510 }, { "epoch": 0.43, "learning_rate": 9.567866384380146e-05, "loss": 3.0656, "step": 5511 }, { "epoch": 0.43, "learning_rate": 9.567787971457697e-05, "loss": 2.8248, "step": 5512 }, { "epoch": 0.43, "learning_rate": 9.567709558535247e-05, "loss": 3.1417, "step": 5513 }, { "epoch": 0.43, "learning_rate": 9.567631145612797e-05, "loss": 3.0421, "step": 5514 }, { "epoch": 0.43, "learning_rate": 9.567552732690348e-05, "loss": 3.1346, "step": 5515 }, { "epoch": 0.43, "learning_rate": 9.567474319767898e-05, "loss": 3.0541, "step": 5516 }, { "epoch": 0.43, "learning_rate": 9.567395906845449e-05, "loss": 3.2149, "step": 5517 }, { "epoch": 0.43, "learning_rate": 9.567317493922999e-05, "loss": 2.8716, "step": 5518 }, { "epoch": 0.43, "learning_rate": 9.567239081000549e-05, "loss": 3.0758, "step": 5519 }, { "epoch": 0.43, "learning_rate": 9.5671606680781e-05, "loss": 3.0102, "step": 5520 }, { "epoch": 0.43, "learning_rate": 9.56708225515565e-05, "loss": 3.0069, "step": 5521 }, { "epoch": 0.43, "learning_rate": 9.5670038422332e-05, "loss": 3.2785, "step": 5522 }, { "epoch": 0.43, "learning_rate": 9.566925429310751e-05, "loss": 3.0298, "step": 5523 }, { "epoch": 0.43, "learning_rate": 9.566847016388301e-05, "loss": 3.3891, "step": 5524 }, { "epoch": 0.43, "learning_rate": 9.566768603465852e-05, "loss": 3.3198, "step": 5525 }, { "epoch": 0.43, "learning_rate": 9.566690190543402e-05, "loss": 2.8175, "step": 5526 }, { "epoch": 0.43, "learning_rate": 9.566611777620952e-05, "loss": 3.1307, "step": 5527 }, { "epoch": 0.43, "learning_rate": 9.566533364698503e-05, "loss": 3.4286, "step": 5528 }, { "epoch": 0.43, "learning_rate": 9.566454951776053e-05, "loss": 2.9437, "step": 5529 }, { "epoch": 0.43, "learning_rate": 9.566376538853603e-05, "loss": 2.9584, "step": 5530 }, { "epoch": 0.43, "learning_rate": 9.566298125931154e-05, "loss": 3.0454, "step": 5531 }, { "epoch": 0.43, "learning_rate": 9.566219713008704e-05, "loss": 2.883, "step": 5532 }, { "epoch": 0.43, "learning_rate": 9.566141300086254e-05, "loss": 3.1345, "step": 5533 }, { "epoch": 0.43, "learning_rate": 9.566062887163805e-05, "loss": 3.3627, "step": 5534 }, { "epoch": 0.43, "learning_rate": 9.565984474241355e-05, "loss": 3.0011, "step": 5535 }, { "epoch": 0.43, "learning_rate": 9.565906061318906e-05, "loss": 2.9224, "step": 5536 }, { "epoch": 0.43, "learning_rate": 9.565827648396456e-05, "loss": 2.859, "step": 5537 }, { "epoch": 0.43, "learning_rate": 9.565749235474006e-05, "loss": 3.3042, "step": 5538 }, { "epoch": 0.43, "learning_rate": 9.565670822551557e-05, "loss": 3.0085, "step": 5539 }, { "epoch": 0.43, "learning_rate": 9.565592409629107e-05, "loss": 3.2192, "step": 5540 }, { "epoch": 0.43, "learning_rate": 9.565513996706657e-05, "loss": 3.0728, "step": 5541 }, { "epoch": 0.43, "learning_rate": 9.565435583784208e-05, "loss": 3.0294, "step": 5542 }, { "epoch": 0.43, "learning_rate": 9.565357170861758e-05, "loss": 3.2411, "step": 5543 }, { "epoch": 0.43, "learning_rate": 9.565278757939309e-05, "loss": 3.0661, "step": 5544 }, { "epoch": 0.43, "learning_rate": 9.565200345016859e-05, "loss": 3.0229, "step": 5545 }, { "epoch": 0.43, "learning_rate": 9.565121932094409e-05, "loss": 2.6494, "step": 5546 }, { "epoch": 0.43, "learning_rate": 9.56504351917196e-05, "loss": 3.0851, "step": 5547 }, { "epoch": 0.44, "learning_rate": 9.56496510624951e-05, "loss": 2.9666, "step": 5548 }, { "epoch": 0.44, "learning_rate": 9.56488669332706e-05, "loss": 2.933, "step": 5549 }, { "epoch": 0.44, "learning_rate": 9.564808280404611e-05, "loss": 3.3648, "step": 5550 }, { "epoch": 0.44, "learning_rate": 9.564729867482161e-05, "loss": 2.9363, "step": 5551 }, { "epoch": 0.44, "learning_rate": 9.564651454559711e-05, "loss": 2.9367, "step": 5552 }, { "epoch": 0.44, "learning_rate": 9.564573041637263e-05, "loss": 3.0579, "step": 5553 }, { "epoch": 0.44, "learning_rate": 9.564494628714814e-05, "loss": 3.2457, "step": 5554 }, { "epoch": 0.44, "learning_rate": 9.564416215792364e-05, "loss": 3.1735, "step": 5555 }, { "epoch": 0.44, "learning_rate": 9.564337802869914e-05, "loss": 3.1035, "step": 5556 }, { "epoch": 0.44, "learning_rate": 9.564259389947465e-05, "loss": 3.201, "step": 5557 }, { "epoch": 0.44, "learning_rate": 9.564180977025015e-05, "loss": 3.1226, "step": 5558 }, { "epoch": 0.44, "learning_rate": 9.564102564102565e-05, "loss": 2.9457, "step": 5559 }, { "epoch": 0.44, "learning_rate": 9.564024151180116e-05, "loss": 2.8778, "step": 5560 }, { "epoch": 0.44, "learning_rate": 9.563945738257666e-05, "loss": 2.8772, "step": 5561 }, { "epoch": 0.44, "learning_rate": 9.563867325335216e-05, "loss": 3.1324, "step": 5562 }, { "epoch": 0.44, "learning_rate": 9.563788912412766e-05, "loss": 2.9842, "step": 5563 }, { "epoch": 0.44, "learning_rate": 9.563710499490316e-05, "loss": 3.2515, "step": 5564 }, { "epoch": 0.44, "learning_rate": 9.563632086567866e-05, "loss": 3.0129, "step": 5565 }, { "epoch": 0.44, "learning_rate": 9.563553673645417e-05, "loss": 2.892, "step": 5566 }, { "epoch": 0.44, "learning_rate": 9.563475260722967e-05, "loss": 2.8553, "step": 5567 }, { "epoch": 0.44, "learning_rate": 9.563396847800517e-05, "loss": 3.1003, "step": 5568 }, { "epoch": 0.44, "learning_rate": 9.563318434878068e-05, "loss": 2.9481, "step": 5569 }, { "epoch": 0.44, "learning_rate": 9.563240021955618e-05, "loss": 3.0614, "step": 5570 }, { "epoch": 0.44, "learning_rate": 9.563161609033168e-05, "loss": 2.9111, "step": 5571 }, { "epoch": 0.44, "learning_rate": 9.563083196110719e-05, "loss": 3.2656, "step": 5572 }, { "epoch": 0.44, "learning_rate": 9.563004783188269e-05, "loss": 2.9665, "step": 5573 }, { "epoch": 0.44, "learning_rate": 9.56292637026582e-05, "loss": 3.1211, "step": 5574 }, { "epoch": 0.44, "learning_rate": 9.56284795734337e-05, "loss": 3.0972, "step": 5575 }, { "epoch": 0.44, "learning_rate": 9.56276954442092e-05, "loss": 2.9636, "step": 5576 }, { "epoch": 0.44, "learning_rate": 9.56269113149847e-05, "loss": 2.8894, "step": 5577 }, { "epoch": 0.44, "learning_rate": 9.562612718576022e-05, "loss": 3.0303, "step": 5578 }, { "epoch": 0.44, "learning_rate": 9.562534305653573e-05, "loss": 2.8136, "step": 5579 }, { "epoch": 0.44, "learning_rate": 9.562455892731123e-05, "loss": 3.2288, "step": 5580 }, { "epoch": 0.44, "learning_rate": 9.562377479808673e-05, "loss": 3.2057, "step": 5581 }, { "epoch": 0.44, "learning_rate": 9.562299066886224e-05, "loss": 3.136, "step": 5582 }, { "epoch": 0.44, "learning_rate": 9.562220653963774e-05, "loss": 2.9079, "step": 5583 }, { "epoch": 0.44, "learning_rate": 9.562142241041325e-05, "loss": 3.1624, "step": 5584 }, { "epoch": 0.44, "learning_rate": 9.562063828118875e-05, "loss": 2.9412, "step": 5585 }, { "epoch": 0.44, "learning_rate": 9.561985415196425e-05, "loss": 2.904, "step": 5586 }, { "epoch": 0.44, "learning_rate": 9.561907002273976e-05, "loss": 3.0781, "step": 5587 }, { "epoch": 0.44, "learning_rate": 9.561828589351526e-05, "loss": 2.8995, "step": 5588 }, { "epoch": 0.44, "learning_rate": 9.561750176429076e-05, "loss": 3.3028, "step": 5589 }, { "epoch": 0.44, "learning_rate": 9.561671763506627e-05, "loss": 3.0718, "step": 5590 }, { "epoch": 0.44, "learning_rate": 9.561593350584177e-05, "loss": 3.126, "step": 5591 }, { "epoch": 0.44, "learning_rate": 9.561514937661728e-05, "loss": 3.1176, "step": 5592 }, { "epoch": 0.44, "learning_rate": 9.561436524739278e-05, "loss": 3.219, "step": 5593 }, { "epoch": 0.44, "learning_rate": 9.561358111816828e-05, "loss": 3.1014, "step": 5594 }, { "epoch": 0.44, "learning_rate": 9.561279698894379e-05, "loss": 3.097, "step": 5595 }, { "epoch": 0.44, "learning_rate": 9.561201285971929e-05, "loss": 3.2682, "step": 5596 }, { "epoch": 0.44, "learning_rate": 9.56112287304948e-05, "loss": 3.1689, "step": 5597 }, { "epoch": 0.44, "learning_rate": 9.56104446012703e-05, "loss": 3.0286, "step": 5598 }, { "epoch": 0.44, "learning_rate": 9.560966047204579e-05, "loss": 3.1764, "step": 5599 }, { "epoch": 0.44, "learning_rate": 9.560887634282129e-05, "loss": 2.9913, "step": 5600 }, { "epoch": 0.44, "learning_rate": 9.56080922135968e-05, "loss": 3.0817, "step": 5601 }, { "epoch": 0.44, "learning_rate": 9.56073080843723e-05, "loss": 3.0941, "step": 5602 }, { "epoch": 0.44, "learning_rate": 9.56065239551478e-05, "loss": 3.5315, "step": 5603 }, { "epoch": 0.44, "learning_rate": 9.560573982592332e-05, "loss": 3.0849, "step": 5604 }, { "epoch": 0.44, "learning_rate": 9.560495569669882e-05, "loss": 2.937, "step": 5605 }, { "epoch": 0.44, "learning_rate": 9.560417156747433e-05, "loss": 2.8784, "step": 5606 }, { "epoch": 0.44, "learning_rate": 9.560338743824983e-05, "loss": 3.0691, "step": 5607 }, { "epoch": 0.44, "learning_rate": 9.560260330902533e-05, "loss": 3.5703, "step": 5608 }, { "epoch": 0.44, "learning_rate": 9.560181917980084e-05, "loss": 2.8481, "step": 5609 }, { "epoch": 0.44, "learning_rate": 9.560103505057634e-05, "loss": 2.8981, "step": 5610 }, { "epoch": 0.44, "learning_rate": 9.560025092135185e-05, "loss": 3.3065, "step": 5611 }, { "epoch": 0.44, "learning_rate": 9.559946679212735e-05, "loss": 2.926, "step": 5612 }, { "epoch": 0.44, "learning_rate": 9.559868266290285e-05, "loss": 3.0467, "step": 5613 }, { "epoch": 0.44, "learning_rate": 9.559789853367836e-05, "loss": 2.9212, "step": 5614 }, { "epoch": 0.44, "learning_rate": 9.559711440445386e-05, "loss": 3.0376, "step": 5615 }, { "epoch": 0.44, "learning_rate": 9.559633027522936e-05, "loss": 2.9948, "step": 5616 }, { "epoch": 0.44, "learning_rate": 9.559554614600487e-05, "loss": 3.0068, "step": 5617 }, { "epoch": 0.44, "learning_rate": 9.559476201678037e-05, "loss": 3.0804, "step": 5618 }, { "epoch": 0.44, "learning_rate": 9.559397788755587e-05, "loss": 3.0594, "step": 5619 }, { "epoch": 0.44, "learning_rate": 9.559319375833138e-05, "loss": 3.0009, "step": 5620 }, { "epoch": 0.44, "learning_rate": 9.559240962910688e-05, "loss": 2.9836, "step": 5621 }, { "epoch": 0.44, "learning_rate": 9.559162549988239e-05, "loss": 3.1462, "step": 5622 }, { "epoch": 0.44, "learning_rate": 9.559084137065789e-05, "loss": 2.9242, "step": 5623 }, { "epoch": 0.44, "learning_rate": 9.559005724143339e-05, "loss": 3.1268, "step": 5624 }, { "epoch": 0.44, "learning_rate": 9.55892731122089e-05, "loss": 3.1697, "step": 5625 }, { "epoch": 0.44, "learning_rate": 9.55884889829844e-05, "loss": 3.1488, "step": 5626 }, { "epoch": 0.44, "learning_rate": 9.55877048537599e-05, "loss": 2.921, "step": 5627 }, { "epoch": 0.44, "learning_rate": 9.558692072453541e-05, "loss": 2.9205, "step": 5628 }, { "epoch": 0.44, "learning_rate": 9.558613659531091e-05, "loss": 2.7555, "step": 5629 }, { "epoch": 0.44, "learning_rate": 9.558535246608641e-05, "loss": 3.1048, "step": 5630 }, { "epoch": 0.44, "learning_rate": 9.558456833686192e-05, "loss": 3.1579, "step": 5631 }, { "epoch": 0.44, "learning_rate": 9.558378420763742e-05, "loss": 3.126, "step": 5632 }, { "epoch": 0.44, "learning_rate": 9.558300007841293e-05, "loss": 2.914, "step": 5633 }, { "epoch": 0.44, "learning_rate": 9.558221594918843e-05, "loss": 2.8698, "step": 5634 }, { "epoch": 0.44, "learning_rate": 9.558143181996393e-05, "loss": 3.151, "step": 5635 }, { "epoch": 0.44, "learning_rate": 9.558064769073944e-05, "loss": 3.0516, "step": 5636 }, { "epoch": 0.44, "learning_rate": 9.557986356151494e-05, "loss": 3.217, "step": 5637 }, { "epoch": 0.44, "learning_rate": 9.557907943229044e-05, "loss": 3.0287, "step": 5638 }, { "epoch": 0.44, "learning_rate": 9.557829530306595e-05, "loss": 3.1496, "step": 5639 }, { "epoch": 0.44, "learning_rate": 9.557751117384145e-05, "loss": 3.0322, "step": 5640 }, { "epoch": 0.44, "learning_rate": 9.557672704461696e-05, "loss": 2.9063, "step": 5641 }, { "epoch": 0.44, "learning_rate": 9.557594291539246e-05, "loss": 3.3519, "step": 5642 }, { "epoch": 0.44, "learning_rate": 9.557515878616796e-05, "loss": 2.9665, "step": 5643 }, { "epoch": 0.44, "learning_rate": 9.557437465694347e-05, "loss": 3.0752, "step": 5644 }, { "epoch": 0.44, "learning_rate": 9.557359052771897e-05, "loss": 3.073, "step": 5645 }, { "epoch": 0.44, "learning_rate": 9.557280639849447e-05, "loss": 3.0895, "step": 5646 }, { "epoch": 0.44, "learning_rate": 9.557202226926998e-05, "loss": 3.3252, "step": 5647 }, { "epoch": 0.44, "learning_rate": 9.557123814004548e-05, "loss": 2.858, "step": 5648 }, { "epoch": 0.44, "learning_rate": 9.557045401082098e-05, "loss": 3.0621, "step": 5649 }, { "epoch": 0.44, "learning_rate": 9.556966988159649e-05, "loss": 2.7779, "step": 5650 }, { "epoch": 0.44, "learning_rate": 9.556888575237199e-05, "loss": 3.0271, "step": 5651 }, { "epoch": 0.44, "learning_rate": 9.55681016231475e-05, "loss": 2.9128, "step": 5652 }, { "epoch": 0.44, "learning_rate": 9.5567317493923e-05, "loss": 3.0428, "step": 5653 }, { "epoch": 0.44, "learning_rate": 9.55665333646985e-05, "loss": 3.0364, "step": 5654 }, { "epoch": 0.44, "learning_rate": 9.556574923547401e-05, "loss": 2.9558, "step": 5655 }, { "epoch": 0.44, "learning_rate": 9.556496510624951e-05, "loss": 3.0895, "step": 5656 }, { "epoch": 0.44, "learning_rate": 9.556418097702501e-05, "loss": 3.0555, "step": 5657 }, { "epoch": 0.44, "learning_rate": 9.556339684780052e-05, "loss": 3.0445, "step": 5658 }, { "epoch": 0.44, "learning_rate": 9.556261271857602e-05, "loss": 2.9531, "step": 5659 }, { "epoch": 0.44, "learning_rate": 9.556182858935153e-05, "loss": 2.9819, "step": 5660 }, { "epoch": 0.44, "learning_rate": 9.556104446012704e-05, "loss": 2.8899, "step": 5661 }, { "epoch": 0.44, "learning_rate": 9.556026033090255e-05, "loss": 2.9969, "step": 5662 }, { "epoch": 0.44, "learning_rate": 9.555947620167805e-05, "loss": 2.9721, "step": 5663 }, { "epoch": 0.44, "learning_rate": 9.555869207245355e-05, "loss": 2.8769, "step": 5664 }, { "epoch": 0.44, "learning_rate": 9.555790794322906e-05, "loss": 3.2057, "step": 5665 }, { "epoch": 0.44, "learning_rate": 9.555712381400456e-05, "loss": 2.9544, "step": 5666 }, { "epoch": 0.44, "learning_rate": 9.555633968478006e-05, "loss": 3.1062, "step": 5667 }, { "epoch": 0.44, "learning_rate": 9.555555555555557e-05, "loss": 3.2791, "step": 5668 }, { "epoch": 0.44, "learning_rate": 9.555477142633107e-05, "loss": 3.39, "step": 5669 }, { "epoch": 0.44, "learning_rate": 9.555398729710658e-05, "loss": 3.2228, "step": 5670 }, { "epoch": 0.44, "learning_rate": 9.555320316788207e-05, "loss": 3.0029, "step": 5671 }, { "epoch": 0.44, "learning_rate": 9.555241903865757e-05, "loss": 2.9412, "step": 5672 }, { "epoch": 0.44, "learning_rate": 9.555163490943307e-05, "loss": 3.0367, "step": 5673 }, { "epoch": 0.44, "learning_rate": 9.555085078020858e-05, "loss": 3.1606, "step": 5674 }, { "epoch": 0.44, "learning_rate": 9.555006665098408e-05, "loss": 3.2274, "step": 5675 }, { "epoch": 0.45, "learning_rate": 9.554928252175958e-05, "loss": 2.9217, "step": 5676 }, { "epoch": 0.45, "learning_rate": 9.554849839253509e-05, "loss": 3.2922, "step": 5677 }, { "epoch": 0.45, "learning_rate": 9.554771426331059e-05, "loss": 3.47, "step": 5678 }, { "epoch": 0.45, "learning_rate": 9.55469301340861e-05, "loss": 2.868, "step": 5679 }, { "epoch": 0.45, "learning_rate": 9.55461460048616e-05, "loss": 2.9583, "step": 5680 }, { "epoch": 0.45, "learning_rate": 9.55453618756371e-05, "loss": 3.0819, "step": 5681 }, { "epoch": 0.45, "learning_rate": 9.55445777464126e-05, "loss": 3.1298, "step": 5682 }, { "epoch": 0.45, "learning_rate": 9.554379361718811e-05, "loss": 2.9794, "step": 5683 }, { "epoch": 0.45, "learning_rate": 9.554300948796361e-05, "loss": 3.3282, "step": 5684 }, { "epoch": 0.45, "learning_rate": 9.554222535873912e-05, "loss": 2.8687, "step": 5685 }, { "epoch": 0.45, "learning_rate": 9.554144122951463e-05, "loss": 3.0258, "step": 5686 }, { "epoch": 0.45, "learning_rate": 9.554065710029014e-05, "loss": 3.0993, "step": 5687 }, { "epoch": 0.45, "learning_rate": 9.553987297106564e-05, "loss": 2.9159, "step": 5688 }, { "epoch": 0.45, "learning_rate": 9.553908884184115e-05, "loss": 2.9275, "step": 5689 }, { "epoch": 0.45, "learning_rate": 9.553830471261665e-05, "loss": 3.0622, "step": 5690 }, { "epoch": 0.45, "learning_rate": 9.553752058339215e-05, "loss": 3.0614, "step": 5691 }, { "epoch": 0.45, "learning_rate": 9.553673645416766e-05, "loss": 3.0886, "step": 5692 }, { "epoch": 0.45, "learning_rate": 9.553595232494316e-05, "loss": 2.8782, "step": 5693 }, { "epoch": 0.45, "learning_rate": 9.553516819571866e-05, "loss": 2.9567, "step": 5694 }, { "epoch": 0.45, "learning_rate": 9.553438406649417e-05, "loss": 3.2527, "step": 5695 }, { "epoch": 0.45, "learning_rate": 9.553359993726967e-05, "loss": 3.0391, "step": 5696 }, { "epoch": 0.45, "learning_rate": 9.553281580804517e-05, "loss": 2.9494, "step": 5697 }, { "epoch": 0.45, "learning_rate": 9.553203167882068e-05, "loss": 2.873, "step": 5698 }, { "epoch": 0.45, "learning_rate": 9.553124754959618e-05, "loss": 3.0472, "step": 5699 }, { "epoch": 0.45, "learning_rate": 9.553046342037169e-05, "loss": 3.1095, "step": 5700 }, { "epoch": 0.45, "learning_rate": 9.552967929114719e-05, "loss": 2.9553, "step": 5701 }, { "epoch": 0.45, "learning_rate": 9.55288951619227e-05, "loss": 2.8548, "step": 5702 }, { "epoch": 0.45, "learning_rate": 9.55281110326982e-05, "loss": 3.2963, "step": 5703 }, { "epoch": 0.45, "learning_rate": 9.55273269034737e-05, "loss": 3.164, "step": 5704 }, { "epoch": 0.45, "learning_rate": 9.55265427742492e-05, "loss": 3.6482, "step": 5705 }, { "epoch": 0.45, "learning_rate": 9.552575864502471e-05, "loss": 3.1983, "step": 5706 }, { "epoch": 0.45, "learning_rate": 9.55249745158002e-05, "loss": 2.8529, "step": 5707 }, { "epoch": 0.45, "learning_rate": 9.55241903865757e-05, "loss": 3.1457, "step": 5708 }, { "epoch": 0.45, "learning_rate": 9.55234062573512e-05, "loss": 3.181, "step": 5709 }, { "epoch": 0.45, "learning_rate": 9.552262212812671e-05, "loss": 3.3688, "step": 5710 }, { "epoch": 0.45, "learning_rate": 9.552183799890221e-05, "loss": 3.2433, "step": 5711 }, { "epoch": 0.45, "learning_rate": 9.552105386967773e-05, "loss": 2.903, "step": 5712 }, { "epoch": 0.45, "learning_rate": 9.552026974045323e-05, "loss": 3.136, "step": 5713 }, { "epoch": 0.45, "learning_rate": 9.551948561122874e-05, "loss": 3.3113, "step": 5714 }, { "epoch": 0.45, "learning_rate": 9.551870148200424e-05, "loss": 2.8959, "step": 5715 }, { "epoch": 0.45, "learning_rate": 9.551791735277974e-05, "loss": 2.9354, "step": 5716 }, { "epoch": 0.45, "learning_rate": 9.551713322355525e-05, "loss": 2.9948, "step": 5717 }, { "epoch": 0.45, "learning_rate": 9.551634909433075e-05, "loss": 2.9846, "step": 5718 }, { "epoch": 0.45, "learning_rate": 9.551556496510626e-05, "loss": 3.05, "step": 5719 }, { "epoch": 0.45, "learning_rate": 9.551478083588176e-05, "loss": 2.9428, "step": 5720 }, { "epoch": 0.45, "learning_rate": 9.551399670665726e-05, "loss": 3.1493, "step": 5721 }, { "epoch": 0.45, "learning_rate": 9.551321257743277e-05, "loss": 2.9185, "step": 5722 }, { "epoch": 0.45, "learning_rate": 9.551242844820827e-05, "loss": 3.0129, "step": 5723 }, { "epoch": 0.45, "learning_rate": 9.551164431898377e-05, "loss": 3.0141, "step": 5724 }, { "epoch": 0.45, "learning_rate": 9.551086018975928e-05, "loss": 2.966, "step": 5725 }, { "epoch": 0.45, "learning_rate": 9.551007606053478e-05, "loss": 2.8233, "step": 5726 }, { "epoch": 0.45, "learning_rate": 9.550929193131029e-05, "loss": 2.8918, "step": 5727 }, { "epoch": 0.45, "learning_rate": 9.550850780208579e-05, "loss": 2.9669, "step": 5728 }, { "epoch": 0.45, "learning_rate": 9.550772367286129e-05, "loss": 3.2602, "step": 5729 }, { "epoch": 0.45, "learning_rate": 9.55069395436368e-05, "loss": 3.3215, "step": 5730 }, { "epoch": 0.45, "learning_rate": 9.55061554144123e-05, "loss": 3.1944, "step": 5731 }, { "epoch": 0.45, "learning_rate": 9.55053712851878e-05, "loss": 2.9268, "step": 5732 }, { "epoch": 0.45, "learning_rate": 9.550458715596331e-05, "loss": 2.9688, "step": 5733 }, { "epoch": 0.45, "learning_rate": 9.550380302673881e-05, "loss": 3.2409, "step": 5734 }, { "epoch": 0.45, "learning_rate": 9.550301889751431e-05, "loss": 3.1103, "step": 5735 }, { "epoch": 0.45, "learning_rate": 9.550223476828982e-05, "loss": 2.9514, "step": 5736 }, { "epoch": 0.45, "learning_rate": 9.550145063906532e-05, "loss": 3.1068, "step": 5737 }, { "epoch": 0.45, "learning_rate": 9.550066650984083e-05, "loss": 3.0329, "step": 5738 }, { "epoch": 0.45, "learning_rate": 9.549988238061633e-05, "loss": 3.1818, "step": 5739 }, { "epoch": 0.45, "learning_rate": 9.549909825139183e-05, "loss": 3.1985, "step": 5740 }, { "epoch": 0.45, "learning_rate": 9.549831412216734e-05, "loss": 3.3909, "step": 5741 }, { "epoch": 0.45, "learning_rate": 9.549752999294284e-05, "loss": 3.007, "step": 5742 }, { "epoch": 0.45, "learning_rate": 9.549674586371834e-05, "loss": 3.032, "step": 5743 }, { "epoch": 0.45, "learning_rate": 9.549596173449385e-05, "loss": 3.1707, "step": 5744 }, { "epoch": 0.45, "learning_rate": 9.549517760526935e-05, "loss": 3.0041, "step": 5745 }, { "epoch": 0.45, "learning_rate": 9.549439347604486e-05, "loss": 2.9467, "step": 5746 }, { "epoch": 0.45, "learning_rate": 9.549360934682036e-05, "loss": 3.1032, "step": 5747 }, { "epoch": 0.45, "learning_rate": 9.549282521759586e-05, "loss": 3.3305, "step": 5748 }, { "epoch": 0.45, "learning_rate": 9.549204108837137e-05, "loss": 3.0316, "step": 5749 }, { "epoch": 0.45, "learning_rate": 9.549125695914687e-05, "loss": 3.1599, "step": 5750 }, { "epoch": 0.45, "learning_rate": 9.549047282992237e-05, "loss": 3.1318, "step": 5751 }, { "epoch": 0.45, "learning_rate": 9.548968870069788e-05, "loss": 3.0116, "step": 5752 }, { "epoch": 0.45, "learning_rate": 9.548890457147338e-05, "loss": 3.0039, "step": 5753 }, { "epoch": 0.45, "learning_rate": 9.548812044224888e-05, "loss": 3.0433, "step": 5754 }, { "epoch": 0.45, "learning_rate": 9.548733631302439e-05, "loss": 3.1354, "step": 5755 }, { "epoch": 0.45, "learning_rate": 9.548655218379989e-05, "loss": 3.0747, "step": 5756 }, { "epoch": 0.45, "learning_rate": 9.54857680545754e-05, "loss": 2.8235, "step": 5757 }, { "epoch": 0.45, "learning_rate": 9.54849839253509e-05, "loss": 2.991, "step": 5758 }, { "epoch": 0.45, "learning_rate": 9.54841997961264e-05, "loss": 3.1455, "step": 5759 }, { "epoch": 0.45, "learning_rate": 9.54834156669019e-05, "loss": 3.1002, "step": 5760 }, { "epoch": 0.45, "learning_rate": 9.548263153767741e-05, "loss": 3.138, "step": 5761 }, { "epoch": 0.45, "learning_rate": 9.548184740845291e-05, "loss": 3.011, "step": 5762 }, { "epoch": 0.45, "learning_rate": 9.548106327922842e-05, "loss": 3.1562, "step": 5763 }, { "epoch": 0.45, "learning_rate": 9.548027915000392e-05, "loss": 3.216, "step": 5764 }, { "epoch": 0.45, "learning_rate": 9.547949502077943e-05, "loss": 2.8331, "step": 5765 }, { "epoch": 0.45, "learning_rate": 9.547871089155493e-05, "loss": 3.1708, "step": 5766 }, { "epoch": 0.45, "learning_rate": 9.547792676233043e-05, "loss": 3.1351, "step": 5767 }, { "epoch": 0.45, "learning_rate": 9.547714263310594e-05, "loss": 2.9295, "step": 5768 }, { "epoch": 0.45, "learning_rate": 9.547635850388145e-05, "loss": 3.1415, "step": 5769 }, { "epoch": 0.45, "learning_rate": 9.547557437465696e-05, "loss": 3.177, "step": 5770 }, { "epoch": 0.45, "learning_rate": 9.547479024543246e-05, "loss": 2.8552, "step": 5771 }, { "epoch": 0.45, "learning_rate": 9.547400611620796e-05, "loss": 3.0339, "step": 5772 }, { "epoch": 0.45, "learning_rate": 9.547322198698347e-05, "loss": 2.9708, "step": 5773 }, { "epoch": 0.45, "learning_rate": 9.547243785775897e-05, "loss": 3.148, "step": 5774 }, { "epoch": 0.45, "learning_rate": 9.547165372853448e-05, "loss": 3.2709, "step": 5775 }, { "epoch": 0.45, "learning_rate": 9.547086959930998e-05, "loss": 3.201, "step": 5776 }, { "epoch": 0.45, "learning_rate": 9.547008547008548e-05, "loss": 3.3462, "step": 5777 }, { "epoch": 0.45, "learning_rate": 9.546930134086099e-05, "loss": 2.7831, "step": 5778 }, { "epoch": 0.45, "learning_rate": 9.546851721163648e-05, "loss": 2.9074, "step": 5779 }, { "epoch": 0.45, "learning_rate": 9.546773308241198e-05, "loss": 3.251, "step": 5780 }, { "epoch": 0.45, "learning_rate": 9.546694895318748e-05, "loss": 3.1739, "step": 5781 }, { "epoch": 0.45, "learning_rate": 9.546616482396299e-05, "loss": 2.9333, "step": 5782 }, { "epoch": 0.45, "learning_rate": 9.546538069473849e-05, "loss": 3.0737, "step": 5783 }, { "epoch": 0.45, "learning_rate": 9.5464596565514e-05, "loss": 3.0407, "step": 5784 }, { "epoch": 0.45, "learning_rate": 9.54638124362895e-05, "loss": 3.15, "step": 5785 }, { "epoch": 0.45, "learning_rate": 9.5463028307065e-05, "loss": 2.9662, "step": 5786 }, { "epoch": 0.45, "learning_rate": 9.54622441778405e-05, "loss": 3.1392, "step": 5787 }, { "epoch": 0.45, "learning_rate": 9.546146004861601e-05, "loss": 3.0557, "step": 5788 }, { "epoch": 0.45, "learning_rate": 9.546067591939151e-05, "loss": 3.3779, "step": 5789 }, { "epoch": 0.45, "learning_rate": 9.545989179016702e-05, "loss": 2.713, "step": 5790 }, { "epoch": 0.45, "learning_rate": 9.545910766094252e-05, "loss": 3.0718, "step": 5791 }, { "epoch": 0.45, "learning_rate": 9.545832353171802e-05, "loss": 2.9959, "step": 5792 }, { "epoch": 0.45, "learning_rate": 9.545753940249353e-05, "loss": 2.6544, "step": 5793 }, { "epoch": 0.45, "learning_rate": 9.545675527326905e-05, "loss": 2.9651, "step": 5794 }, { "epoch": 0.45, "learning_rate": 9.545597114404455e-05, "loss": 2.9508, "step": 5795 }, { "epoch": 0.45, "learning_rate": 9.545518701482005e-05, "loss": 3.0976, "step": 5796 }, { "epoch": 0.45, "learning_rate": 9.545440288559556e-05, "loss": 3.1709, "step": 5797 }, { "epoch": 0.45, "learning_rate": 9.545361875637106e-05, "loss": 2.9471, "step": 5798 }, { "epoch": 0.45, "learning_rate": 9.545283462714656e-05, "loss": 3.2134, "step": 5799 }, { "epoch": 0.45, "learning_rate": 9.545205049792207e-05, "loss": 2.9342, "step": 5800 }, { "epoch": 0.45, "learning_rate": 9.545126636869757e-05, "loss": 3.1818, "step": 5801 }, { "epoch": 0.45, "learning_rate": 9.545048223947307e-05, "loss": 2.8405, "step": 5802 }, { "epoch": 0.46, "learning_rate": 9.544969811024858e-05, "loss": 2.9397, "step": 5803 }, { "epoch": 0.46, "learning_rate": 9.544891398102408e-05, "loss": 3.2263, "step": 5804 }, { "epoch": 0.46, "learning_rate": 9.544812985179959e-05, "loss": 2.8473, "step": 5805 }, { "epoch": 0.46, "learning_rate": 9.544734572257509e-05, "loss": 3.0666, "step": 5806 }, { "epoch": 0.46, "learning_rate": 9.544656159335059e-05, "loss": 2.7729, "step": 5807 }, { "epoch": 0.46, "learning_rate": 9.54457774641261e-05, "loss": 2.8152, "step": 5808 }, { "epoch": 0.46, "learning_rate": 9.54449933349016e-05, "loss": 2.9161, "step": 5809 }, { "epoch": 0.46, "learning_rate": 9.54442092056771e-05, "loss": 3.3018, "step": 5810 }, { "epoch": 0.46, "learning_rate": 9.544342507645261e-05, "loss": 3.1176, "step": 5811 }, { "epoch": 0.46, "learning_rate": 9.544264094722811e-05, "loss": 3.0982, "step": 5812 }, { "epoch": 0.46, "learning_rate": 9.544185681800362e-05, "loss": 3.1885, "step": 5813 }, { "epoch": 0.46, "learning_rate": 9.544107268877912e-05, "loss": 2.7945, "step": 5814 }, { "epoch": 0.46, "learning_rate": 9.544028855955461e-05, "loss": 3.2171, "step": 5815 }, { "epoch": 0.46, "learning_rate": 9.543950443033011e-05, "loss": 2.9263, "step": 5816 }, { "epoch": 0.46, "learning_rate": 9.543872030110562e-05, "loss": 3.1144, "step": 5817 }, { "epoch": 0.46, "learning_rate": 9.543793617188112e-05, "loss": 3.1877, "step": 5818 }, { "epoch": 0.46, "learning_rate": 9.543715204265662e-05, "loss": 3.1823, "step": 5819 }, { "epoch": 0.46, "learning_rate": 9.543636791343214e-05, "loss": 3.0289, "step": 5820 }, { "epoch": 0.46, "learning_rate": 9.543558378420764e-05, "loss": 2.8998, "step": 5821 }, { "epoch": 0.46, "learning_rate": 9.543479965498315e-05, "loss": 3.0368, "step": 5822 }, { "epoch": 0.46, "learning_rate": 9.543401552575865e-05, "loss": 3.0446, "step": 5823 }, { "epoch": 0.46, "learning_rate": 9.543323139653416e-05, "loss": 2.9575, "step": 5824 }, { "epoch": 0.46, "learning_rate": 9.543244726730966e-05, "loss": 3.0551, "step": 5825 }, { "epoch": 0.46, "learning_rate": 9.543166313808516e-05, "loss": 3.1128, "step": 5826 }, { "epoch": 0.46, "learning_rate": 9.543087900886067e-05, "loss": 2.7555, "step": 5827 }, { "epoch": 0.46, "learning_rate": 9.543009487963617e-05, "loss": 3.0626, "step": 5828 }, { "epoch": 0.46, "learning_rate": 9.542931075041167e-05, "loss": 2.9856, "step": 5829 }, { "epoch": 0.46, "learning_rate": 9.542852662118718e-05, "loss": 3.3036, "step": 5830 }, { "epoch": 0.46, "learning_rate": 9.542774249196268e-05, "loss": 2.9007, "step": 5831 }, { "epoch": 0.46, "learning_rate": 9.542695836273818e-05, "loss": 2.9542, "step": 5832 }, { "epoch": 0.46, "learning_rate": 9.542617423351369e-05, "loss": 2.9353, "step": 5833 }, { "epoch": 0.46, "learning_rate": 9.542539010428919e-05, "loss": 3.0643, "step": 5834 }, { "epoch": 0.46, "learning_rate": 9.54246059750647e-05, "loss": 3.2172, "step": 5835 }, { "epoch": 0.46, "learning_rate": 9.54238218458402e-05, "loss": 2.7624, "step": 5836 }, { "epoch": 0.46, "learning_rate": 9.54230377166157e-05, "loss": 3.0058, "step": 5837 }, { "epoch": 0.46, "learning_rate": 9.542225358739121e-05, "loss": 3.2025, "step": 5838 }, { "epoch": 0.46, "learning_rate": 9.542146945816671e-05, "loss": 3.1909, "step": 5839 }, { "epoch": 0.46, "learning_rate": 9.542068532894221e-05, "loss": 2.9661, "step": 5840 }, { "epoch": 0.46, "learning_rate": 9.541990119971772e-05, "loss": 2.9818, "step": 5841 }, { "epoch": 0.46, "learning_rate": 9.541911707049322e-05, "loss": 3.0452, "step": 5842 }, { "epoch": 0.46, "learning_rate": 9.541833294126873e-05, "loss": 2.8559, "step": 5843 }, { "epoch": 0.46, "learning_rate": 9.541754881204423e-05, "loss": 2.9815, "step": 5844 }, { "epoch": 0.46, "learning_rate": 9.541676468281973e-05, "loss": 3.0202, "step": 5845 }, { "epoch": 0.46, "learning_rate": 9.541598055359524e-05, "loss": 2.9495, "step": 5846 }, { "epoch": 0.46, "learning_rate": 9.541519642437074e-05, "loss": 3.451, "step": 5847 }, { "epoch": 0.46, "learning_rate": 9.541441229514624e-05, "loss": 3.0428, "step": 5848 }, { "epoch": 0.46, "learning_rate": 9.541362816592175e-05, "loss": 2.9677, "step": 5849 }, { "epoch": 0.46, "learning_rate": 9.541284403669725e-05, "loss": 3.1006, "step": 5850 }, { "epoch": 0.46, "learning_rate": 9.541205990747275e-05, "loss": 2.9708, "step": 5851 }, { "epoch": 0.46, "learning_rate": 9.541127577824826e-05, "loss": 3.1964, "step": 5852 }, { "epoch": 0.46, "learning_rate": 9.541049164902376e-05, "loss": 3.0821, "step": 5853 }, { "epoch": 0.46, "learning_rate": 9.540970751979927e-05, "loss": 2.7723, "step": 5854 }, { "epoch": 0.46, "learning_rate": 9.540892339057477e-05, "loss": 3.0722, "step": 5855 }, { "epoch": 0.46, "learning_rate": 9.540813926135027e-05, "loss": 3.2072, "step": 5856 }, { "epoch": 0.46, "learning_rate": 9.540735513212578e-05, "loss": 3.18, "step": 5857 }, { "epoch": 0.46, "learning_rate": 9.540657100290128e-05, "loss": 2.8683, "step": 5858 }, { "epoch": 0.46, "learning_rate": 9.540578687367678e-05, "loss": 3.1569, "step": 5859 }, { "epoch": 0.46, "learning_rate": 9.540500274445229e-05, "loss": 3.089, "step": 5860 }, { "epoch": 0.46, "learning_rate": 9.540421861522779e-05, "loss": 3.2498, "step": 5861 }, { "epoch": 0.46, "learning_rate": 9.54034344860033e-05, "loss": 3.0707, "step": 5862 }, { "epoch": 0.46, "learning_rate": 9.54026503567788e-05, "loss": 3.1682, "step": 5863 }, { "epoch": 0.46, "learning_rate": 9.54018662275543e-05, "loss": 3.1354, "step": 5864 }, { "epoch": 0.46, "learning_rate": 9.54010820983298e-05, "loss": 3.0681, "step": 5865 }, { "epoch": 0.46, "learning_rate": 9.540029796910531e-05, "loss": 2.8986, "step": 5866 }, { "epoch": 0.46, "learning_rate": 9.539951383988081e-05, "loss": 2.996, "step": 5867 }, { "epoch": 0.46, "learning_rate": 9.539872971065632e-05, "loss": 2.8464, "step": 5868 }, { "epoch": 0.46, "learning_rate": 9.539794558143182e-05, "loss": 3.0577, "step": 5869 }, { "epoch": 0.46, "learning_rate": 9.539716145220732e-05, "loss": 3.0487, "step": 5870 }, { "epoch": 0.46, "learning_rate": 9.539637732298283e-05, "loss": 3.2215, "step": 5871 }, { "epoch": 0.46, "learning_rate": 9.539559319375833e-05, "loss": 3.0956, "step": 5872 }, { "epoch": 0.46, "learning_rate": 9.539480906453384e-05, "loss": 2.7989, "step": 5873 }, { "epoch": 0.46, "learning_rate": 9.539402493530934e-05, "loss": 3.2472, "step": 5874 }, { "epoch": 0.46, "learning_rate": 9.539324080608484e-05, "loss": 2.9577, "step": 5875 }, { "epoch": 0.46, "learning_rate": 9.539245667686035e-05, "loss": 2.8996, "step": 5876 }, { "epoch": 0.46, "learning_rate": 9.539167254763586e-05, "loss": 3.0679, "step": 5877 }, { "epoch": 0.46, "learning_rate": 9.539088841841137e-05, "loss": 2.9858, "step": 5878 }, { "epoch": 0.46, "learning_rate": 9.539010428918687e-05, "loss": 3.1401, "step": 5879 }, { "epoch": 0.46, "learning_rate": 9.538932015996238e-05, "loss": 2.9635, "step": 5880 }, { "epoch": 0.46, "learning_rate": 9.538853603073788e-05, "loss": 3.1992, "step": 5881 }, { "epoch": 0.46, "learning_rate": 9.538775190151338e-05, "loss": 3.0512, "step": 5882 }, { "epoch": 0.46, "learning_rate": 9.538696777228889e-05, "loss": 3.0923, "step": 5883 }, { "epoch": 0.46, "learning_rate": 9.538618364306439e-05, "loss": 3.0913, "step": 5884 }, { "epoch": 0.46, "learning_rate": 9.53853995138399e-05, "loss": 3.0552, "step": 5885 }, { "epoch": 0.46, "learning_rate": 9.53846153846154e-05, "loss": 3.0997, "step": 5886 }, { "epoch": 0.46, "learning_rate": 9.538383125539089e-05, "loss": 3.0546, "step": 5887 }, { "epoch": 0.46, "learning_rate": 9.538304712616639e-05, "loss": 2.9432, "step": 5888 }, { "epoch": 0.46, "learning_rate": 9.53822629969419e-05, "loss": 3.0412, "step": 5889 }, { "epoch": 0.46, "learning_rate": 9.53814788677174e-05, "loss": 3.0692, "step": 5890 }, { "epoch": 0.46, "learning_rate": 9.53806947384929e-05, "loss": 3.1105, "step": 5891 }, { "epoch": 0.46, "learning_rate": 9.53799106092684e-05, "loss": 3.073, "step": 5892 }, { "epoch": 0.46, "learning_rate": 9.537912648004391e-05, "loss": 2.8756, "step": 5893 }, { "epoch": 0.46, "learning_rate": 9.537834235081941e-05, "loss": 2.8938, "step": 5894 }, { "epoch": 0.46, "learning_rate": 9.537755822159492e-05, "loss": 2.9831, "step": 5895 }, { "epoch": 0.46, "learning_rate": 9.537677409237042e-05, "loss": 2.8677, "step": 5896 }, { "epoch": 0.46, "learning_rate": 9.537598996314592e-05, "loss": 3.1679, "step": 5897 }, { "epoch": 0.46, "learning_rate": 9.537520583392143e-05, "loss": 2.9832, "step": 5898 }, { "epoch": 0.46, "learning_rate": 9.537442170469693e-05, "loss": 3.1175, "step": 5899 }, { "epoch": 0.46, "learning_rate": 9.537363757547244e-05, "loss": 2.939, "step": 5900 }, { "epoch": 0.46, "learning_rate": 9.537285344624794e-05, "loss": 2.9597, "step": 5901 }, { "epoch": 0.46, "learning_rate": 9.537206931702344e-05, "loss": 3.2293, "step": 5902 }, { "epoch": 0.46, "learning_rate": 9.537128518779896e-05, "loss": 3.2107, "step": 5903 }, { "epoch": 0.46, "learning_rate": 9.537050105857446e-05, "loss": 2.909, "step": 5904 }, { "epoch": 0.46, "learning_rate": 9.536971692934997e-05, "loss": 3.2273, "step": 5905 }, { "epoch": 0.46, "learning_rate": 9.536893280012547e-05, "loss": 2.9805, "step": 5906 }, { "epoch": 0.46, "learning_rate": 9.536814867090097e-05, "loss": 2.899, "step": 5907 }, { "epoch": 0.46, "learning_rate": 9.536736454167648e-05, "loss": 2.9812, "step": 5908 }, { "epoch": 0.46, "learning_rate": 9.536658041245198e-05, "loss": 2.969, "step": 5909 }, { "epoch": 0.46, "learning_rate": 9.536579628322749e-05, "loss": 3.17, "step": 5910 }, { "epoch": 0.46, "learning_rate": 9.536501215400299e-05, "loss": 2.8079, "step": 5911 }, { "epoch": 0.46, "learning_rate": 9.536422802477849e-05, "loss": 3.1963, "step": 5912 }, { "epoch": 0.46, "learning_rate": 9.5363443895554e-05, "loss": 3.0039, "step": 5913 }, { "epoch": 0.46, "learning_rate": 9.53626597663295e-05, "loss": 3.3428, "step": 5914 }, { "epoch": 0.46, "learning_rate": 9.5361875637105e-05, "loss": 2.9926, "step": 5915 }, { "epoch": 0.46, "learning_rate": 9.536109150788051e-05, "loss": 3.2595, "step": 5916 }, { "epoch": 0.46, "learning_rate": 9.536030737865601e-05, "loss": 2.907, "step": 5917 }, { "epoch": 0.46, "learning_rate": 9.535952324943151e-05, "loss": 3.0857, "step": 5918 }, { "epoch": 0.46, "learning_rate": 9.535873912020702e-05, "loss": 3.0974, "step": 5919 }, { "epoch": 0.46, "learning_rate": 9.535795499098252e-05, "loss": 3.254, "step": 5920 }, { "epoch": 0.46, "learning_rate": 9.535717086175803e-05, "loss": 3.1101, "step": 5921 }, { "epoch": 0.46, "learning_rate": 9.535638673253353e-05, "loss": 2.9642, "step": 5922 }, { "epoch": 0.46, "learning_rate": 9.535560260330902e-05, "loss": 2.8694, "step": 5923 }, { "epoch": 0.46, "learning_rate": 9.535481847408452e-05, "loss": 3.1083, "step": 5924 }, { "epoch": 0.46, "learning_rate": 9.535403434486003e-05, "loss": 3.068, "step": 5925 }, { "epoch": 0.46, "learning_rate": 9.535325021563553e-05, "loss": 3.0518, "step": 5926 }, { "epoch": 0.46, "learning_rate": 9.535246608641103e-05, "loss": 2.9912, "step": 5927 }, { "epoch": 0.46, "learning_rate": 9.535168195718655e-05, "loss": 2.9679, "step": 5928 }, { "epoch": 0.46, "learning_rate": 9.535089782796206e-05, "loss": 3.2566, "step": 5929 }, { "epoch": 0.46, "learning_rate": 9.535011369873756e-05, "loss": 2.9812, "step": 5930 }, { "epoch": 0.47, "learning_rate": 9.534932956951306e-05, "loss": 3.1738, "step": 5931 }, { "epoch": 0.47, "learning_rate": 9.534854544028857e-05, "loss": 3.1463, "step": 5932 }, { "epoch": 0.47, "learning_rate": 9.534776131106407e-05, "loss": 2.9674, "step": 5933 }, { "epoch": 0.47, "learning_rate": 9.534697718183957e-05, "loss": 2.9587, "step": 5934 }, { "epoch": 0.47, "learning_rate": 9.534619305261508e-05, "loss": 2.8615, "step": 5935 }, { "epoch": 0.47, "learning_rate": 9.534540892339058e-05, "loss": 3.2097, "step": 5936 }, { "epoch": 0.47, "learning_rate": 9.534462479416608e-05, "loss": 3.0841, "step": 5937 }, { "epoch": 0.47, "learning_rate": 9.534384066494159e-05, "loss": 3.1649, "step": 5938 }, { "epoch": 0.47, "learning_rate": 9.534305653571709e-05, "loss": 3.1089, "step": 5939 }, { "epoch": 0.47, "learning_rate": 9.53422724064926e-05, "loss": 3.1575, "step": 5940 }, { "epoch": 0.47, "learning_rate": 9.53414882772681e-05, "loss": 2.9615, "step": 5941 }, { "epoch": 0.47, "learning_rate": 9.53407041480436e-05, "loss": 2.9649, "step": 5942 }, { "epoch": 0.47, "learning_rate": 9.53399200188191e-05, "loss": 3.074, "step": 5943 }, { "epoch": 0.47, "learning_rate": 9.533913588959461e-05, "loss": 3.2082, "step": 5944 }, { "epoch": 0.47, "learning_rate": 9.533835176037011e-05, "loss": 2.9071, "step": 5945 }, { "epoch": 0.47, "learning_rate": 9.533756763114562e-05, "loss": 3.306, "step": 5946 }, { "epoch": 0.47, "learning_rate": 9.533678350192112e-05, "loss": 3.0907, "step": 5947 }, { "epoch": 0.47, "learning_rate": 9.533599937269663e-05, "loss": 3.0301, "step": 5948 }, { "epoch": 0.47, "learning_rate": 9.533521524347213e-05, "loss": 3.1329, "step": 5949 }, { "epoch": 0.47, "learning_rate": 9.533443111424763e-05, "loss": 3.0378, "step": 5950 }, { "epoch": 0.47, "learning_rate": 9.533364698502314e-05, "loss": 3.1678, "step": 5951 }, { "epoch": 0.47, "learning_rate": 9.533286285579864e-05, "loss": 2.9474, "step": 5952 }, { "epoch": 0.47, "learning_rate": 9.533207872657414e-05, "loss": 2.9799, "step": 5953 }, { "epoch": 0.47, "learning_rate": 9.533129459734965e-05, "loss": 2.8435, "step": 5954 }, { "epoch": 0.47, "learning_rate": 9.533051046812515e-05, "loss": 2.9263, "step": 5955 }, { "epoch": 0.47, "learning_rate": 9.532972633890065e-05, "loss": 3.1627, "step": 5956 }, { "epoch": 0.47, "learning_rate": 9.532894220967616e-05, "loss": 2.9052, "step": 5957 }, { "epoch": 0.47, "learning_rate": 9.532815808045166e-05, "loss": 3.0847, "step": 5958 }, { "epoch": 0.47, "learning_rate": 9.532737395122717e-05, "loss": 3.1287, "step": 5959 }, { "epoch": 0.47, "learning_rate": 9.532658982200267e-05, "loss": 3.0299, "step": 5960 }, { "epoch": 0.47, "learning_rate": 9.532580569277817e-05, "loss": 3.0696, "step": 5961 }, { "epoch": 0.47, "learning_rate": 9.532502156355368e-05, "loss": 2.8827, "step": 5962 }, { "epoch": 0.47, "learning_rate": 9.532423743432918e-05, "loss": 3.0111, "step": 5963 }, { "epoch": 0.47, "learning_rate": 9.532345330510468e-05, "loss": 3.0807, "step": 5964 }, { "epoch": 0.47, "learning_rate": 9.532266917588019e-05, "loss": 3.011, "step": 5965 }, { "epoch": 0.47, "learning_rate": 9.532188504665569e-05, "loss": 3.104, "step": 5966 }, { "epoch": 0.47, "learning_rate": 9.53211009174312e-05, "loss": 3.3034, "step": 5967 }, { "epoch": 0.47, "learning_rate": 9.53203167882067e-05, "loss": 3.0224, "step": 5968 }, { "epoch": 0.47, "learning_rate": 9.53195326589822e-05, "loss": 2.9741, "step": 5969 }, { "epoch": 0.47, "learning_rate": 9.53187485297577e-05, "loss": 3.2911, "step": 5970 }, { "epoch": 0.47, "learning_rate": 9.531796440053321e-05, "loss": 2.9929, "step": 5971 }, { "epoch": 0.47, "learning_rate": 9.531718027130871e-05, "loss": 3.1106, "step": 5972 }, { "epoch": 0.47, "learning_rate": 9.531639614208422e-05, "loss": 3.0827, "step": 5973 }, { "epoch": 0.47, "learning_rate": 9.531561201285972e-05, "loss": 3.2876, "step": 5974 }, { "epoch": 0.47, "learning_rate": 9.531482788363522e-05, "loss": 3.0587, "step": 5975 }, { "epoch": 0.47, "learning_rate": 9.531404375441073e-05, "loss": 2.924, "step": 5976 }, { "epoch": 0.47, "learning_rate": 9.531325962518623e-05, "loss": 2.9477, "step": 5977 }, { "epoch": 0.47, "learning_rate": 9.531247549596174e-05, "loss": 3.1183, "step": 5978 }, { "epoch": 0.47, "learning_rate": 9.531169136673724e-05, "loss": 2.9101, "step": 5979 }, { "epoch": 0.47, "learning_rate": 9.531090723751274e-05, "loss": 2.9172, "step": 5980 }, { "epoch": 0.47, "learning_rate": 9.531012310828825e-05, "loss": 3.0365, "step": 5981 }, { "epoch": 0.47, "learning_rate": 9.530933897906375e-05, "loss": 3.0303, "step": 5982 }, { "epoch": 0.47, "learning_rate": 9.530855484983925e-05, "loss": 3.1979, "step": 5983 }, { "epoch": 0.47, "learning_rate": 9.530777072061476e-05, "loss": 2.992, "step": 5984 }, { "epoch": 0.47, "learning_rate": 9.530698659139027e-05, "loss": 3.0202, "step": 5985 }, { "epoch": 0.47, "learning_rate": 9.530620246216578e-05, "loss": 2.9322, "step": 5986 }, { "epoch": 0.47, "learning_rate": 9.530541833294128e-05, "loss": 3.0201, "step": 5987 }, { "epoch": 0.47, "learning_rate": 9.530463420371679e-05, "loss": 3.1639, "step": 5988 }, { "epoch": 0.47, "learning_rate": 9.530385007449229e-05, "loss": 3.3819, "step": 5989 }, { "epoch": 0.47, "learning_rate": 9.530306594526779e-05, "loss": 2.8857, "step": 5990 }, { "epoch": 0.47, "learning_rate": 9.53022818160433e-05, "loss": 3.2054, "step": 5991 }, { "epoch": 0.47, "learning_rate": 9.53014976868188e-05, "loss": 3.3705, "step": 5992 }, { "epoch": 0.47, "learning_rate": 9.53007135575943e-05, "loss": 3.1003, "step": 5993 }, { "epoch": 0.47, "learning_rate": 9.529992942836981e-05, "loss": 3.0771, "step": 5994 }, { "epoch": 0.47, "learning_rate": 9.52991452991453e-05, "loss": 3.1241, "step": 5995 }, { "epoch": 0.47, "learning_rate": 9.52983611699208e-05, "loss": 3.0486, "step": 5996 }, { "epoch": 0.47, "learning_rate": 9.52975770406963e-05, "loss": 3.0335, "step": 5997 }, { "epoch": 0.47, "learning_rate": 9.529679291147181e-05, "loss": 3.1542, "step": 5998 }, { "epoch": 0.47, "learning_rate": 9.529600878224731e-05, "loss": 2.9108, "step": 5999 }, { "epoch": 0.47, "learning_rate": 9.529522465302282e-05, "loss": 3.227, "step": 6000 }, { "epoch": 0.47, "eval_loss": 2.9627065658569336, "eval_runtime": 218.1032, "eval_samples_per_second": 51.957, "eval_steps_per_second": 3.251, "step": 6000 }, { "epoch": 0.47, "learning_rate": 9.529444052379832e-05, "loss": 3.1093, "step": 6001 }, { "epoch": 0.47, "learning_rate": 9.529365639457382e-05, "loss": 2.9088, "step": 6002 }, { "epoch": 0.47, "learning_rate": 9.529287226534933e-05, "loss": 2.8573, "step": 6003 }, { "epoch": 0.47, "learning_rate": 9.529208813612483e-05, "loss": 2.8712, "step": 6004 }, { "epoch": 0.47, "learning_rate": 9.529130400690033e-05, "loss": 3.2473, "step": 6005 }, { "epoch": 0.47, "learning_rate": 9.529051987767584e-05, "loss": 3.5506, "step": 6006 }, { "epoch": 0.47, "learning_rate": 9.528973574845134e-05, "loss": 3.0818, "step": 6007 }, { "epoch": 0.47, "learning_rate": 9.528895161922685e-05, "loss": 3.2307, "step": 6008 }, { "epoch": 0.47, "learning_rate": 9.528816749000235e-05, "loss": 3.7507, "step": 6009 }, { "epoch": 0.47, "learning_rate": 9.528738336077785e-05, "loss": 3.2684, "step": 6010 }, { "epoch": 0.47, "learning_rate": 9.528659923155337e-05, "loss": 2.9399, "step": 6011 }, { "epoch": 0.47, "learning_rate": 9.528581510232887e-05, "loss": 2.984, "step": 6012 }, { "epoch": 0.47, "learning_rate": 9.528503097310438e-05, "loss": 2.9557, "step": 6013 }, { "epoch": 0.47, "learning_rate": 9.528424684387988e-05, "loss": 3.3019, "step": 6014 }, { "epoch": 0.47, "learning_rate": 9.528346271465539e-05, "loss": 3.1156, "step": 6015 }, { "epoch": 0.47, "learning_rate": 9.528267858543089e-05, "loss": 3.0467, "step": 6016 }, { "epoch": 0.47, "learning_rate": 9.528189445620639e-05, "loss": 3.0381, "step": 6017 }, { "epoch": 0.47, "learning_rate": 9.52811103269819e-05, "loss": 2.9587, "step": 6018 }, { "epoch": 0.47, "learning_rate": 9.52803261977574e-05, "loss": 2.9152, "step": 6019 }, { "epoch": 0.47, "learning_rate": 9.52795420685329e-05, "loss": 3.0571, "step": 6020 }, { "epoch": 0.47, "learning_rate": 9.527875793930841e-05, "loss": 3.0672, "step": 6021 }, { "epoch": 0.47, "learning_rate": 9.527797381008391e-05, "loss": 3.2205, "step": 6022 }, { "epoch": 0.47, "learning_rate": 9.527718968085941e-05, "loss": 2.9731, "step": 6023 }, { "epoch": 0.47, "learning_rate": 9.527640555163492e-05, "loss": 3.2185, "step": 6024 }, { "epoch": 0.47, "learning_rate": 9.527562142241042e-05, "loss": 3.0172, "step": 6025 }, { "epoch": 0.47, "learning_rate": 9.527483729318593e-05, "loss": 2.9975, "step": 6026 }, { "epoch": 0.47, "learning_rate": 9.527405316396143e-05, "loss": 2.9213, "step": 6027 }, { "epoch": 0.47, "learning_rate": 9.527326903473693e-05, "loss": 3.0601, "step": 6028 }, { "epoch": 0.47, "learning_rate": 9.527248490551244e-05, "loss": 2.8813, "step": 6029 }, { "epoch": 0.47, "learning_rate": 9.527170077628794e-05, "loss": 2.7725, "step": 6030 }, { "epoch": 0.47, "learning_rate": 9.527091664706343e-05, "loss": 3.2574, "step": 6031 }, { "epoch": 0.47, "learning_rate": 9.527013251783893e-05, "loss": 3.0587, "step": 6032 }, { "epoch": 0.47, "learning_rate": 9.526934838861444e-05, "loss": 3.0485, "step": 6033 }, { "epoch": 0.47, "learning_rate": 9.526856425938994e-05, "loss": 3.1378, "step": 6034 }, { "epoch": 0.47, "learning_rate": 9.526778013016545e-05, "loss": 2.9034, "step": 6035 }, { "epoch": 0.47, "learning_rate": 9.526699600094096e-05, "loss": 2.8611, "step": 6036 }, { "epoch": 0.47, "learning_rate": 9.526621187171647e-05, "loss": 3.0203, "step": 6037 }, { "epoch": 0.47, "learning_rate": 9.526542774249197e-05, "loss": 3.2341, "step": 6038 }, { "epoch": 0.47, "learning_rate": 9.526464361326747e-05, "loss": 3.3295, "step": 6039 }, { "epoch": 0.47, "learning_rate": 9.526385948404298e-05, "loss": 3.3171, "step": 6040 }, { "epoch": 0.47, "learning_rate": 9.526307535481848e-05, "loss": 3.1103, "step": 6041 }, { "epoch": 0.47, "learning_rate": 9.526229122559398e-05, "loss": 3.0512, "step": 6042 }, { "epoch": 0.47, "learning_rate": 9.526150709636949e-05, "loss": 3.0767, "step": 6043 }, { "epoch": 0.47, "learning_rate": 9.526072296714499e-05, "loss": 3.0291, "step": 6044 }, { "epoch": 0.47, "learning_rate": 9.52599388379205e-05, "loss": 3.0995, "step": 6045 }, { "epoch": 0.47, "learning_rate": 9.5259154708696e-05, "loss": 2.9587, "step": 6046 }, { "epoch": 0.47, "learning_rate": 9.52583705794715e-05, "loss": 3.1229, "step": 6047 }, { "epoch": 0.47, "learning_rate": 9.5257586450247e-05, "loss": 3.0906, "step": 6048 }, { "epoch": 0.47, "learning_rate": 9.525680232102251e-05, "loss": 3.0157, "step": 6049 }, { "epoch": 0.47, "learning_rate": 9.525601819179801e-05, "loss": 3.0605, "step": 6050 }, { "epoch": 0.47, "learning_rate": 9.525523406257352e-05, "loss": 3.0868, "step": 6051 }, { "epoch": 0.47, "learning_rate": 9.525444993334902e-05, "loss": 3.0157, "step": 6052 }, { "epoch": 0.47, "learning_rate": 9.525366580412452e-05, "loss": 2.9625, "step": 6053 }, { "epoch": 0.47, "learning_rate": 9.525288167490003e-05, "loss": 2.8233, "step": 6054 }, { "epoch": 0.47, "learning_rate": 9.525209754567553e-05, "loss": 3.347, "step": 6055 }, { "epoch": 0.47, "learning_rate": 9.525131341645104e-05, "loss": 3.2086, "step": 6056 }, { "epoch": 0.47, "learning_rate": 9.525052928722654e-05, "loss": 3.2257, "step": 6057 }, { "epoch": 0.48, "learning_rate": 9.524974515800204e-05, "loss": 2.9205, "step": 6058 }, { "epoch": 0.48, "learning_rate": 9.524896102877755e-05, "loss": 3.1663, "step": 6059 }, { "epoch": 0.48, "learning_rate": 9.524817689955305e-05, "loss": 3.1693, "step": 6060 }, { "epoch": 0.48, "learning_rate": 9.524739277032855e-05, "loss": 3.0836, "step": 6061 }, { "epoch": 0.48, "learning_rate": 9.524660864110406e-05, "loss": 3.115, "step": 6062 }, { "epoch": 0.48, "learning_rate": 9.524582451187956e-05, "loss": 2.8768, "step": 6063 }, { "epoch": 0.48, "learning_rate": 9.524504038265507e-05, "loss": 2.9467, "step": 6064 }, { "epoch": 0.48, "learning_rate": 9.524425625343057e-05, "loss": 3.0299, "step": 6065 }, { "epoch": 0.48, "learning_rate": 9.524347212420607e-05, "loss": 3.1663, "step": 6066 }, { "epoch": 0.48, "learning_rate": 9.524268799498158e-05, "loss": 3.2701, "step": 6067 }, { "epoch": 0.48, "learning_rate": 9.524190386575708e-05, "loss": 3.0189, "step": 6068 }, { "epoch": 0.48, "learning_rate": 9.524111973653258e-05, "loss": 3.1152, "step": 6069 }, { "epoch": 0.48, "learning_rate": 9.524033560730809e-05, "loss": 3.0736, "step": 6070 }, { "epoch": 0.48, "learning_rate": 9.523955147808359e-05, "loss": 3.04, "step": 6071 }, { "epoch": 0.48, "learning_rate": 9.52387673488591e-05, "loss": 2.8516, "step": 6072 }, { "epoch": 0.48, "learning_rate": 9.52379832196346e-05, "loss": 3.1121, "step": 6073 }, { "epoch": 0.48, "learning_rate": 9.52371990904101e-05, "loss": 2.8898, "step": 6074 }, { "epoch": 0.48, "learning_rate": 9.52364149611856e-05, "loss": 3.2755, "step": 6075 }, { "epoch": 0.48, "learning_rate": 9.523563083196111e-05, "loss": 2.9302, "step": 6076 }, { "epoch": 0.48, "learning_rate": 9.523484670273661e-05, "loss": 3.0849, "step": 6077 }, { "epoch": 0.48, "learning_rate": 9.523406257351212e-05, "loss": 2.9848, "step": 6078 }, { "epoch": 0.48, "learning_rate": 9.523327844428762e-05, "loss": 2.9092, "step": 6079 }, { "epoch": 0.48, "learning_rate": 9.523249431506312e-05, "loss": 3.0617, "step": 6080 }, { "epoch": 0.48, "learning_rate": 9.523171018583863e-05, "loss": 2.9356, "step": 6081 }, { "epoch": 0.48, "learning_rate": 9.523092605661413e-05, "loss": 3.0233, "step": 6082 }, { "epoch": 0.48, "learning_rate": 9.523014192738964e-05, "loss": 3.0378, "step": 6083 }, { "epoch": 0.48, "learning_rate": 9.522935779816514e-05, "loss": 3.043, "step": 6084 }, { "epoch": 0.48, "learning_rate": 9.522857366894064e-05, "loss": 3.0377, "step": 6085 }, { "epoch": 0.48, "learning_rate": 9.522778953971615e-05, "loss": 3.2686, "step": 6086 }, { "epoch": 0.48, "learning_rate": 9.522700541049165e-05, "loss": 2.8556, "step": 6087 }, { "epoch": 0.48, "learning_rate": 9.522622128126715e-05, "loss": 3.1217, "step": 6088 }, { "epoch": 0.48, "learning_rate": 9.522543715204266e-05, "loss": 2.968, "step": 6089 }, { "epoch": 0.48, "learning_rate": 9.522465302281816e-05, "loss": 3.2243, "step": 6090 }, { "epoch": 0.48, "learning_rate": 9.522386889359366e-05, "loss": 2.8338, "step": 6091 }, { "epoch": 0.48, "learning_rate": 9.522308476436917e-05, "loss": 3.0398, "step": 6092 }, { "epoch": 0.48, "learning_rate": 9.522230063514469e-05, "loss": 3.1185, "step": 6093 }, { "epoch": 0.48, "learning_rate": 9.522151650592019e-05, "loss": 3.0644, "step": 6094 }, { "epoch": 0.48, "learning_rate": 9.522073237669569e-05, "loss": 3.2683, "step": 6095 }, { "epoch": 0.48, "learning_rate": 9.52199482474712e-05, "loss": 2.9285, "step": 6096 }, { "epoch": 0.48, "learning_rate": 9.52191641182467e-05, "loss": 3.1383, "step": 6097 }, { "epoch": 0.48, "learning_rate": 9.52183799890222e-05, "loss": 3.2004, "step": 6098 }, { "epoch": 0.48, "learning_rate": 9.521759585979771e-05, "loss": 3.02, "step": 6099 }, { "epoch": 0.48, "learning_rate": 9.521681173057321e-05, "loss": 3.0611, "step": 6100 }, { "epoch": 0.48, "learning_rate": 9.521602760134871e-05, "loss": 3.2132, "step": 6101 }, { "epoch": 0.48, "learning_rate": 9.521524347212422e-05, "loss": 3.2275, "step": 6102 }, { "epoch": 0.48, "learning_rate": 9.521445934289971e-05, "loss": 2.899, "step": 6103 }, { "epoch": 0.48, "learning_rate": 9.521367521367521e-05, "loss": 2.8158, "step": 6104 }, { "epoch": 0.48, "learning_rate": 9.521289108445072e-05, "loss": 2.9477, "step": 6105 }, { "epoch": 0.48, "learning_rate": 9.521210695522622e-05, "loss": 3.1289, "step": 6106 }, { "epoch": 0.48, "learning_rate": 9.521132282600172e-05, "loss": 3.0717, "step": 6107 }, { "epoch": 0.48, "learning_rate": 9.521053869677723e-05, "loss": 2.8354, "step": 6108 }, { "epoch": 0.48, "learning_rate": 9.520975456755273e-05, "loss": 3.1422, "step": 6109 }, { "epoch": 0.48, "learning_rate": 9.520897043832823e-05, "loss": 2.9449, "step": 6110 }, { "epoch": 0.48, "learning_rate": 9.520818630910374e-05, "loss": 3.239, "step": 6111 }, { "epoch": 0.48, "learning_rate": 9.520740217987924e-05, "loss": 3.032, "step": 6112 }, { "epoch": 0.48, "learning_rate": 9.520661805065475e-05, "loss": 2.9226, "step": 6113 }, { "epoch": 0.48, "learning_rate": 9.520583392143025e-05, "loss": 3.2126, "step": 6114 }, { "epoch": 0.48, "learning_rate": 9.520504979220575e-05, "loss": 2.8298, "step": 6115 }, { "epoch": 0.48, "learning_rate": 9.520426566298126e-05, "loss": 3.3783, "step": 6116 }, { "epoch": 0.48, "learning_rate": 9.520348153375676e-05, "loss": 3.0789, "step": 6117 }, { "epoch": 0.48, "learning_rate": 9.520269740453226e-05, "loss": 3.0047, "step": 6118 }, { "epoch": 0.48, "learning_rate": 9.520191327530778e-05, "loss": 3.022, "step": 6119 }, { "epoch": 0.48, "learning_rate": 9.520112914608328e-05, "loss": 2.9632, "step": 6120 }, { "epoch": 0.48, "learning_rate": 9.520034501685879e-05, "loss": 3.0998, "step": 6121 }, { "epoch": 0.48, "learning_rate": 9.519956088763429e-05, "loss": 3.3225, "step": 6122 }, { "epoch": 0.48, "learning_rate": 9.51987767584098e-05, "loss": 2.9256, "step": 6123 }, { "epoch": 0.48, "learning_rate": 9.51979926291853e-05, "loss": 3.224, "step": 6124 }, { "epoch": 0.48, "learning_rate": 9.51972084999608e-05, "loss": 3.435, "step": 6125 }, { "epoch": 0.48, "learning_rate": 9.519642437073631e-05, "loss": 2.7632, "step": 6126 }, { "epoch": 0.48, "learning_rate": 9.519564024151181e-05, "loss": 3.0976, "step": 6127 }, { "epoch": 0.48, "learning_rate": 9.519485611228731e-05, "loss": 3.0804, "step": 6128 }, { "epoch": 0.48, "learning_rate": 9.519407198306282e-05, "loss": 2.8359, "step": 6129 }, { "epoch": 0.48, "learning_rate": 9.519328785383832e-05, "loss": 2.8956, "step": 6130 }, { "epoch": 0.48, "learning_rate": 9.519250372461383e-05, "loss": 2.9133, "step": 6131 }, { "epoch": 0.48, "learning_rate": 9.519171959538933e-05, "loss": 2.9438, "step": 6132 }, { "epoch": 0.48, "learning_rate": 9.519093546616483e-05, "loss": 3.0331, "step": 6133 }, { "epoch": 0.48, "learning_rate": 9.519015133694034e-05, "loss": 2.9451, "step": 6134 }, { "epoch": 0.48, "learning_rate": 9.518936720771584e-05, "loss": 3.0806, "step": 6135 }, { "epoch": 0.48, "learning_rate": 9.518858307849134e-05, "loss": 3.1704, "step": 6136 }, { "epoch": 0.48, "learning_rate": 9.518779894926685e-05, "loss": 2.8511, "step": 6137 }, { "epoch": 0.48, "learning_rate": 9.518701482004235e-05, "loss": 3.3813, "step": 6138 }, { "epoch": 0.48, "learning_rate": 9.518623069081784e-05, "loss": 2.7999, "step": 6139 }, { "epoch": 0.48, "learning_rate": 9.518544656159334e-05, "loss": 3.0191, "step": 6140 }, { "epoch": 0.48, "learning_rate": 9.518466243236885e-05, "loss": 3.1307, "step": 6141 }, { "epoch": 0.48, "learning_rate": 9.518387830314435e-05, "loss": 3.0352, "step": 6142 }, { "epoch": 0.48, "learning_rate": 9.518309417391986e-05, "loss": 3.105, "step": 6143 }, { "epoch": 0.48, "learning_rate": 9.518231004469537e-05, "loss": 2.9369, "step": 6144 }, { "epoch": 0.48, "learning_rate": 9.518152591547088e-05, "loss": 3.23, "step": 6145 }, { "epoch": 0.48, "learning_rate": 9.518074178624638e-05, "loss": 3.0399, "step": 6146 }, { "epoch": 0.48, "learning_rate": 9.517995765702188e-05, "loss": 3.1946, "step": 6147 }, { "epoch": 0.48, "learning_rate": 9.517917352779739e-05, "loss": 3.0692, "step": 6148 }, { "epoch": 0.48, "learning_rate": 9.517838939857289e-05, "loss": 2.912, "step": 6149 }, { "epoch": 0.48, "learning_rate": 9.51776052693484e-05, "loss": 2.8672, "step": 6150 }, { "epoch": 0.48, "learning_rate": 9.51768211401239e-05, "loss": 3.363, "step": 6151 }, { "epoch": 0.48, "learning_rate": 9.51760370108994e-05, "loss": 2.8667, "step": 6152 }, { "epoch": 0.48, "learning_rate": 9.51752528816749e-05, "loss": 2.9289, "step": 6153 }, { "epoch": 0.48, "learning_rate": 9.517446875245041e-05, "loss": 2.8833, "step": 6154 }, { "epoch": 0.48, "learning_rate": 9.517368462322591e-05, "loss": 3.007, "step": 6155 }, { "epoch": 0.48, "learning_rate": 9.517290049400142e-05, "loss": 2.9158, "step": 6156 }, { "epoch": 0.48, "learning_rate": 9.517211636477692e-05, "loss": 3.0623, "step": 6157 }, { "epoch": 0.48, "learning_rate": 9.517133223555242e-05, "loss": 2.8437, "step": 6158 }, { "epoch": 0.48, "learning_rate": 9.517054810632793e-05, "loss": 2.928, "step": 6159 }, { "epoch": 0.48, "learning_rate": 9.516976397710343e-05, "loss": 2.8447, "step": 6160 }, { "epoch": 0.48, "learning_rate": 9.516897984787894e-05, "loss": 3.0245, "step": 6161 }, { "epoch": 0.48, "learning_rate": 9.516819571865444e-05, "loss": 3.1301, "step": 6162 }, { "epoch": 0.48, "learning_rate": 9.516741158942994e-05, "loss": 3.0328, "step": 6163 }, { "epoch": 0.48, "learning_rate": 9.516662746020545e-05, "loss": 2.9702, "step": 6164 }, { "epoch": 0.48, "learning_rate": 9.516584333098095e-05, "loss": 2.9629, "step": 6165 }, { "epoch": 0.48, "learning_rate": 9.516505920175645e-05, "loss": 2.9652, "step": 6166 }, { "epoch": 0.48, "learning_rate": 9.516427507253196e-05, "loss": 3.095, "step": 6167 }, { "epoch": 0.48, "learning_rate": 9.516349094330746e-05, "loss": 2.9528, "step": 6168 }, { "epoch": 0.48, "learning_rate": 9.516270681408297e-05, "loss": 3.0115, "step": 6169 }, { "epoch": 0.48, "learning_rate": 9.516192268485847e-05, "loss": 2.8879, "step": 6170 }, { "epoch": 0.48, "learning_rate": 9.516113855563397e-05, "loss": 3.136, "step": 6171 }, { "epoch": 0.48, "learning_rate": 9.516035442640948e-05, "loss": 3.0924, "step": 6172 }, { "epoch": 0.48, "learning_rate": 9.515957029718498e-05, "loss": 3.1891, "step": 6173 }, { "epoch": 0.48, "learning_rate": 9.515878616796048e-05, "loss": 2.7953, "step": 6174 }, { "epoch": 0.48, "learning_rate": 9.515800203873599e-05, "loss": 2.8304, "step": 6175 }, { "epoch": 0.48, "learning_rate": 9.515721790951149e-05, "loss": 3.0296, "step": 6176 }, { "epoch": 0.48, "learning_rate": 9.5156433780287e-05, "loss": 2.9311, "step": 6177 }, { "epoch": 0.48, "learning_rate": 9.51556496510625e-05, "loss": 3.1556, "step": 6178 }, { "epoch": 0.48, "learning_rate": 9.5154865521838e-05, "loss": 2.8593, "step": 6179 }, { "epoch": 0.48, "learning_rate": 9.51540813926135e-05, "loss": 2.9882, "step": 6180 }, { "epoch": 0.48, "learning_rate": 9.515329726338901e-05, "loss": 3.156, "step": 6181 }, { "epoch": 0.48, "learning_rate": 9.515251313416451e-05, "loss": 3.0251, "step": 6182 }, { "epoch": 0.48, "learning_rate": 9.515172900494002e-05, "loss": 2.9705, "step": 6183 }, { "epoch": 0.48, "learning_rate": 9.515094487571552e-05, "loss": 3.1872, "step": 6184 }, { "epoch": 0.48, "learning_rate": 9.515016074649102e-05, "loss": 2.8252, "step": 6185 }, { "epoch": 0.49, "learning_rate": 9.514937661726653e-05, "loss": 3.0696, "step": 6186 }, { "epoch": 0.49, "learning_rate": 9.514859248804203e-05, "loss": 3.0817, "step": 6187 }, { "epoch": 0.49, "learning_rate": 9.514780835881753e-05, "loss": 2.9249, "step": 6188 }, { "epoch": 0.49, "learning_rate": 9.514702422959304e-05, "loss": 3.1339, "step": 6189 }, { "epoch": 0.49, "learning_rate": 9.514624010036854e-05, "loss": 3.1592, "step": 6190 }, { "epoch": 0.49, "learning_rate": 9.514545597114405e-05, "loss": 2.881, "step": 6191 }, { "epoch": 0.49, "learning_rate": 9.514467184191955e-05, "loss": 3.2616, "step": 6192 }, { "epoch": 0.49, "learning_rate": 9.514388771269505e-05, "loss": 2.9693, "step": 6193 }, { "epoch": 0.49, "learning_rate": 9.514310358347056e-05, "loss": 3.2482, "step": 6194 }, { "epoch": 0.49, "learning_rate": 9.514231945424606e-05, "loss": 2.8107, "step": 6195 }, { "epoch": 0.49, "learning_rate": 9.514153532502156e-05, "loss": 3.1291, "step": 6196 }, { "epoch": 0.49, "learning_rate": 9.514075119579707e-05, "loss": 3.2051, "step": 6197 }, { "epoch": 0.49, "learning_rate": 9.513996706657257e-05, "loss": 3.0568, "step": 6198 }, { "epoch": 0.49, "learning_rate": 9.513918293734808e-05, "loss": 3.0896, "step": 6199 }, { "epoch": 0.49, "learning_rate": 9.513839880812358e-05, "loss": 2.984, "step": 6200 }, { "epoch": 0.49, "learning_rate": 9.51376146788991e-05, "loss": 3.1798, "step": 6201 }, { "epoch": 0.49, "learning_rate": 9.51368305496746e-05, "loss": 3.128, "step": 6202 }, { "epoch": 0.49, "learning_rate": 9.51360464204501e-05, "loss": 3.259, "step": 6203 }, { "epoch": 0.49, "learning_rate": 9.513526229122561e-05, "loss": 3.0115, "step": 6204 }, { "epoch": 0.49, "learning_rate": 9.513447816200111e-05, "loss": 3.1497, "step": 6205 }, { "epoch": 0.49, "learning_rate": 9.513369403277661e-05, "loss": 2.9449, "step": 6206 }, { "epoch": 0.49, "learning_rate": 9.513290990355212e-05, "loss": 2.9026, "step": 6207 }, { "epoch": 0.49, "learning_rate": 9.513212577432762e-05, "loss": 3.1241, "step": 6208 }, { "epoch": 0.49, "learning_rate": 9.513134164510313e-05, "loss": 2.9687, "step": 6209 }, { "epoch": 0.49, "learning_rate": 9.513055751587863e-05, "loss": 3.0473, "step": 6210 }, { "epoch": 0.49, "learning_rate": 9.512977338665412e-05, "loss": 3.0658, "step": 6211 }, { "epoch": 0.49, "learning_rate": 9.512898925742962e-05, "loss": 2.957, "step": 6212 }, { "epoch": 0.49, "learning_rate": 9.512820512820513e-05, "loss": 3.0468, "step": 6213 }, { "epoch": 0.49, "learning_rate": 9.512742099898063e-05, "loss": 2.9481, "step": 6214 }, { "epoch": 0.49, "learning_rate": 9.512663686975613e-05, "loss": 2.9623, "step": 6215 }, { "epoch": 0.49, "learning_rate": 9.512585274053164e-05, "loss": 3.0145, "step": 6216 }, { "epoch": 0.49, "learning_rate": 9.512506861130714e-05, "loss": 3.0267, "step": 6217 }, { "epoch": 0.49, "learning_rate": 9.512428448208265e-05, "loss": 3.362, "step": 6218 }, { "epoch": 0.49, "learning_rate": 9.512350035285815e-05, "loss": 2.9465, "step": 6219 }, { "epoch": 0.49, "learning_rate": 9.512271622363365e-05, "loss": 3.0602, "step": 6220 }, { "epoch": 0.49, "learning_rate": 9.512193209440916e-05, "loss": 2.9006, "step": 6221 }, { "epoch": 0.49, "learning_rate": 9.512114796518466e-05, "loss": 3.1976, "step": 6222 }, { "epoch": 0.49, "learning_rate": 9.512036383596016e-05, "loss": 3.0725, "step": 6223 }, { "epoch": 0.49, "learning_rate": 9.511957970673567e-05, "loss": 3.0335, "step": 6224 }, { "epoch": 0.49, "learning_rate": 9.511879557751117e-05, "loss": 3.0425, "step": 6225 }, { "epoch": 0.49, "learning_rate": 9.511801144828667e-05, "loss": 3.096, "step": 6226 }, { "epoch": 0.49, "learning_rate": 9.511722731906219e-05, "loss": 3.0232, "step": 6227 }, { "epoch": 0.49, "learning_rate": 9.51164431898377e-05, "loss": 2.7638, "step": 6228 }, { "epoch": 0.49, "learning_rate": 9.51156590606132e-05, "loss": 3.0884, "step": 6229 }, { "epoch": 0.49, "learning_rate": 9.51148749313887e-05, "loss": 2.9153, "step": 6230 }, { "epoch": 0.49, "learning_rate": 9.51140908021642e-05, "loss": 3.0732, "step": 6231 }, { "epoch": 0.49, "learning_rate": 9.511330667293971e-05, "loss": 3.0532, "step": 6232 }, { "epoch": 0.49, "learning_rate": 9.511252254371521e-05, "loss": 2.9313, "step": 6233 }, { "epoch": 0.49, "learning_rate": 9.511173841449072e-05, "loss": 3.3531, "step": 6234 }, { "epoch": 0.49, "learning_rate": 9.511095428526622e-05, "loss": 3.164, "step": 6235 }, { "epoch": 0.49, "learning_rate": 9.511017015604172e-05, "loss": 3.1884, "step": 6236 }, { "epoch": 0.49, "learning_rate": 9.510938602681723e-05, "loss": 3.0103, "step": 6237 }, { "epoch": 0.49, "learning_rate": 9.510860189759273e-05, "loss": 2.9125, "step": 6238 }, { "epoch": 0.49, "learning_rate": 9.510781776836824e-05, "loss": 2.9555, "step": 6239 }, { "epoch": 0.49, "learning_rate": 9.510703363914374e-05, "loss": 3.1458, "step": 6240 }, { "epoch": 0.49, "learning_rate": 9.510624950991924e-05, "loss": 3.0713, "step": 6241 }, { "epoch": 0.49, "learning_rate": 9.510546538069475e-05, "loss": 3.2097, "step": 6242 }, { "epoch": 0.49, "learning_rate": 9.510468125147025e-05, "loss": 2.9132, "step": 6243 }, { "epoch": 0.49, "learning_rate": 9.510389712224575e-05, "loss": 2.8704, "step": 6244 }, { "epoch": 0.49, "learning_rate": 9.510311299302126e-05, "loss": 3.0784, "step": 6245 }, { "epoch": 0.49, "learning_rate": 9.510232886379676e-05, "loss": 3.0304, "step": 6246 }, { "epoch": 0.49, "learning_rate": 9.510154473457225e-05, "loss": 3.2448, "step": 6247 }, { "epoch": 0.49, "learning_rate": 9.510076060534776e-05, "loss": 2.8049, "step": 6248 }, { "epoch": 0.49, "learning_rate": 9.509997647612326e-05, "loss": 3.1456, "step": 6249 }, { "epoch": 0.49, "learning_rate": 9.509919234689876e-05, "loss": 3.0282, "step": 6250 }, { "epoch": 0.49, "learning_rate": 9.509840821767427e-05, "loss": 2.7963, "step": 6251 }, { "epoch": 0.49, "learning_rate": 9.509762408844978e-05, "loss": 3.0891, "step": 6252 }, { "epoch": 0.49, "learning_rate": 9.509683995922529e-05, "loss": 3.1499, "step": 6253 }, { "epoch": 0.49, "learning_rate": 9.509605583000079e-05, "loss": 3.1146, "step": 6254 }, { "epoch": 0.49, "learning_rate": 9.50952717007763e-05, "loss": 2.9733, "step": 6255 }, { "epoch": 0.49, "learning_rate": 9.50944875715518e-05, "loss": 2.7951, "step": 6256 }, { "epoch": 0.49, "learning_rate": 9.50937034423273e-05, "loss": 2.9353, "step": 6257 }, { "epoch": 0.49, "learning_rate": 9.50929193131028e-05, "loss": 3.0683, "step": 6258 }, { "epoch": 0.49, "learning_rate": 9.509213518387831e-05, "loss": 2.7833, "step": 6259 }, { "epoch": 0.49, "learning_rate": 9.509135105465381e-05, "loss": 3.1155, "step": 6260 }, { "epoch": 0.49, "learning_rate": 9.509056692542932e-05, "loss": 3.2104, "step": 6261 }, { "epoch": 0.49, "learning_rate": 9.508978279620482e-05, "loss": 3.0532, "step": 6262 }, { "epoch": 0.49, "learning_rate": 9.508899866698032e-05, "loss": 2.687, "step": 6263 }, { "epoch": 0.49, "learning_rate": 9.508821453775583e-05, "loss": 2.993, "step": 6264 }, { "epoch": 0.49, "learning_rate": 9.508743040853133e-05, "loss": 2.9583, "step": 6265 }, { "epoch": 0.49, "learning_rate": 9.508664627930684e-05, "loss": 2.9038, "step": 6266 }, { "epoch": 0.49, "learning_rate": 9.508586215008234e-05, "loss": 2.9184, "step": 6267 }, { "epoch": 0.49, "learning_rate": 9.508507802085784e-05, "loss": 3.0422, "step": 6268 }, { "epoch": 0.49, "learning_rate": 9.508429389163335e-05, "loss": 3.3747, "step": 6269 }, { "epoch": 0.49, "learning_rate": 9.508350976240885e-05, "loss": 2.9545, "step": 6270 }, { "epoch": 0.49, "learning_rate": 9.508272563318435e-05, "loss": 3.0932, "step": 6271 }, { "epoch": 0.49, "learning_rate": 9.508194150395986e-05, "loss": 3.0389, "step": 6272 }, { "epoch": 0.49, "learning_rate": 9.508115737473536e-05, "loss": 3.1889, "step": 6273 }, { "epoch": 0.49, "learning_rate": 9.508037324551086e-05, "loss": 3.1536, "step": 6274 }, { "epoch": 0.49, "learning_rate": 9.507958911628637e-05, "loss": 2.9319, "step": 6275 }, { "epoch": 0.49, "learning_rate": 9.507880498706187e-05, "loss": 3.0298, "step": 6276 }, { "epoch": 0.49, "learning_rate": 9.507802085783738e-05, "loss": 2.7399, "step": 6277 }, { "epoch": 0.49, "learning_rate": 9.507723672861288e-05, "loss": 3.0399, "step": 6278 }, { "epoch": 0.49, "learning_rate": 9.507645259938838e-05, "loss": 2.9536, "step": 6279 }, { "epoch": 0.49, "learning_rate": 9.507566847016389e-05, "loss": 3.1693, "step": 6280 }, { "epoch": 0.49, "learning_rate": 9.507488434093939e-05, "loss": 3.3325, "step": 6281 }, { "epoch": 0.49, "learning_rate": 9.50741002117149e-05, "loss": 2.8531, "step": 6282 }, { "epoch": 0.49, "learning_rate": 9.50733160824904e-05, "loss": 2.9994, "step": 6283 }, { "epoch": 0.49, "learning_rate": 9.50725319532659e-05, "loss": 3.0448, "step": 6284 }, { "epoch": 0.49, "learning_rate": 9.50717478240414e-05, "loss": 3.2632, "step": 6285 }, { "epoch": 0.49, "learning_rate": 9.507096369481691e-05, "loss": 2.9842, "step": 6286 }, { "epoch": 0.49, "learning_rate": 9.507017956559241e-05, "loss": 2.8464, "step": 6287 }, { "epoch": 0.49, "learning_rate": 9.506939543636792e-05, "loss": 3.0078, "step": 6288 }, { "epoch": 0.49, "learning_rate": 9.506861130714342e-05, "loss": 2.7487, "step": 6289 }, { "epoch": 0.49, "learning_rate": 9.506782717791892e-05, "loss": 3.1418, "step": 6290 }, { "epoch": 0.49, "learning_rate": 9.506704304869443e-05, "loss": 2.8107, "step": 6291 }, { "epoch": 0.49, "learning_rate": 9.506625891946993e-05, "loss": 2.8409, "step": 6292 }, { "epoch": 0.49, "learning_rate": 9.506547479024543e-05, "loss": 3.2755, "step": 6293 }, { "epoch": 0.49, "learning_rate": 9.506469066102094e-05, "loss": 3.1395, "step": 6294 }, { "epoch": 0.49, "learning_rate": 9.506390653179644e-05, "loss": 3.253, "step": 6295 }, { "epoch": 0.49, "learning_rate": 9.506312240257195e-05, "loss": 2.9664, "step": 6296 }, { "epoch": 0.49, "learning_rate": 9.506233827334745e-05, "loss": 3.1072, "step": 6297 }, { "epoch": 0.49, "learning_rate": 9.506155414412295e-05, "loss": 3.2836, "step": 6298 }, { "epoch": 0.49, "learning_rate": 9.506077001489846e-05, "loss": 2.8894, "step": 6299 }, { "epoch": 0.49, "learning_rate": 9.505998588567396e-05, "loss": 3.0171, "step": 6300 }, { "epoch": 0.49, "learning_rate": 9.505920175644946e-05, "loss": 3.1805, "step": 6301 }, { "epoch": 0.49, "learning_rate": 9.505841762722497e-05, "loss": 2.8845, "step": 6302 }, { "epoch": 0.49, "learning_rate": 9.505763349800047e-05, "loss": 2.8784, "step": 6303 }, { "epoch": 0.49, "learning_rate": 9.505684936877598e-05, "loss": 3.2407, "step": 6304 }, { "epoch": 0.49, "learning_rate": 9.505606523955148e-05, "loss": 2.8563, "step": 6305 }, { "epoch": 0.49, "learning_rate": 9.505528111032698e-05, "loss": 3.0008, "step": 6306 }, { "epoch": 0.49, "learning_rate": 9.505449698110249e-05, "loss": 3.0667, "step": 6307 }, { "epoch": 0.49, "learning_rate": 9.505371285187799e-05, "loss": 3.0348, "step": 6308 }, { "epoch": 0.49, "learning_rate": 9.505292872265351e-05, "loss": 3.2463, "step": 6309 }, { "epoch": 0.49, "learning_rate": 9.505214459342901e-05, "loss": 2.9002, "step": 6310 }, { "epoch": 0.49, "learning_rate": 9.505136046420451e-05, "loss": 3.0143, "step": 6311 }, { "epoch": 0.49, "learning_rate": 9.505057633498002e-05, "loss": 2.9742, "step": 6312 }, { "epoch": 0.5, "learning_rate": 9.504979220575552e-05, "loss": 3.2817, "step": 6313 }, { "epoch": 0.5, "learning_rate": 9.504900807653103e-05, "loss": 3.1863, "step": 6314 }, { "epoch": 0.5, "learning_rate": 9.504822394730653e-05, "loss": 3.1995, "step": 6315 }, { "epoch": 0.5, "learning_rate": 9.504743981808203e-05, "loss": 3.2594, "step": 6316 }, { "epoch": 0.5, "learning_rate": 9.504665568885754e-05, "loss": 3.1501, "step": 6317 }, { "epoch": 0.5, "learning_rate": 9.504587155963304e-05, "loss": 2.8648, "step": 6318 }, { "epoch": 0.5, "learning_rate": 9.504508743040853e-05, "loss": 3.2731, "step": 6319 }, { "epoch": 0.5, "learning_rate": 9.504430330118403e-05, "loss": 3.0547, "step": 6320 }, { "epoch": 0.5, "learning_rate": 9.504351917195954e-05, "loss": 3.0093, "step": 6321 }, { "epoch": 0.5, "learning_rate": 9.504273504273504e-05, "loss": 3.0354, "step": 6322 }, { "epoch": 0.5, "learning_rate": 9.504195091351054e-05, "loss": 3.1915, "step": 6323 }, { "epoch": 0.5, "learning_rate": 9.504116678428605e-05, "loss": 3.1839, "step": 6324 }, { "epoch": 0.5, "learning_rate": 9.504038265506155e-05, "loss": 3.2846, "step": 6325 }, { "epoch": 0.5, "learning_rate": 9.503959852583706e-05, "loss": 2.9913, "step": 6326 }, { "epoch": 0.5, "learning_rate": 9.503881439661256e-05, "loss": 2.8767, "step": 6327 }, { "epoch": 0.5, "learning_rate": 9.503803026738806e-05, "loss": 3.0102, "step": 6328 }, { "epoch": 0.5, "learning_rate": 9.503724613816357e-05, "loss": 3.0792, "step": 6329 }, { "epoch": 0.5, "learning_rate": 9.503646200893907e-05, "loss": 3.0743, "step": 6330 }, { "epoch": 0.5, "learning_rate": 9.503567787971457e-05, "loss": 3.1418, "step": 6331 }, { "epoch": 0.5, "learning_rate": 9.503489375049008e-05, "loss": 3.0077, "step": 6332 }, { "epoch": 0.5, "learning_rate": 9.503410962126558e-05, "loss": 3.1555, "step": 6333 }, { "epoch": 0.5, "learning_rate": 9.503332549204109e-05, "loss": 2.8452, "step": 6334 }, { "epoch": 0.5, "learning_rate": 9.50325413628166e-05, "loss": 2.9236, "step": 6335 }, { "epoch": 0.5, "learning_rate": 9.50317572335921e-05, "loss": 3.0112, "step": 6336 }, { "epoch": 0.5, "learning_rate": 9.503097310436761e-05, "loss": 3.0339, "step": 6337 }, { "epoch": 0.5, "learning_rate": 9.503018897514311e-05, "loss": 3.0239, "step": 6338 }, { "epoch": 0.5, "learning_rate": 9.502940484591862e-05, "loss": 2.8312, "step": 6339 }, { "epoch": 0.5, "learning_rate": 9.502862071669412e-05, "loss": 2.937, "step": 6340 }, { "epoch": 0.5, "learning_rate": 9.502783658746962e-05, "loss": 3.0747, "step": 6341 }, { "epoch": 0.5, "learning_rate": 9.502705245824513e-05, "loss": 3.2917, "step": 6342 }, { "epoch": 0.5, "learning_rate": 9.502626832902063e-05, "loss": 3.0302, "step": 6343 }, { "epoch": 0.5, "learning_rate": 9.502548419979614e-05, "loss": 2.8691, "step": 6344 }, { "epoch": 0.5, "learning_rate": 9.502470007057164e-05, "loss": 3.115, "step": 6345 }, { "epoch": 0.5, "learning_rate": 9.502391594134714e-05, "loss": 3.0512, "step": 6346 }, { "epoch": 0.5, "learning_rate": 9.502313181212265e-05, "loss": 2.9127, "step": 6347 }, { "epoch": 0.5, "learning_rate": 9.502234768289815e-05, "loss": 2.8989, "step": 6348 }, { "epoch": 0.5, "learning_rate": 9.502156355367365e-05, "loss": 3.1107, "step": 6349 }, { "epoch": 0.5, "learning_rate": 9.502077942444916e-05, "loss": 3.1463, "step": 6350 }, { "epoch": 0.5, "learning_rate": 9.501999529522466e-05, "loss": 3.3447, "step": 6351 }, { "epoch": 0.5, "learning_rate": 9.501921116600017e-05, "loss": 2.8816, "step": 6352 }, { "epoch": 0.5, "learning_rate": 9.501842703677567e-05, "loss": 3.1621, "step": 6353 }, { "epoch": 0.5, "learning_rate": 9.501764290755117e-05, "loss": 2.9744, "step": 6354 }, { "epoch": 0.5, "learning_rate": 9.501685877832666e-05, "loss": 3.3058, "step": 6355 }, { "epoch": 0.5, "learning_rate": 9.501607464910217e-05, "loss": 3.2753, "step": 6356 }, { "epoch": 0.5, "learning_rate": 9.501529051987767e-05, "loss": 3.3757, "step": 6357 }, { "epoch": 0.5, "learning_rate": 9.501450639065317e-05, "loss": 2.9274, "step": 6358 }, { "epoch": 0.5, "learning_rate": 9.501372226142868e-05, "loss": 3.1545, "step": 6359 }, { "epoch": 0.5, "learning_rate": 9.50129381322042e-05, "loss": 3.3056, "step": 6360 }, { "epoch": 0.5, "learning_rate": 9.50121540029797e-05, "loss": 3.0271, "step": 6361 }, { "epoch": 0.5, "learning_rate": 9.50113698737552e-05, "loss": 3.0509, "step": 6362 }, { "epoch": 0.5, "learning_rate": 9.50105857445307e-05, "loss": 3.026, "step": 6363 }, { "epoch": 0.5, "learning_rate": 9.500980161530621e-05, "loss": 2.8507, "step": 6364 }, { "epoch": 0.5, "learning_rate": 9.500901748608171e-05, "loss": 2.868, "step": 6365 }, { "epoch": 0.5, "learning_rate": 9.500823335685722e-05, "loss": 3.1319, "step": 6366 }, { "epoch": 0.5, "learning_rate": 9.500744922763272e-05, "loss": 3.0533, "step": 6367 }, { "epoch": 0.5, "learning_rate": 9.500666509840822e-05, "loss": 3.1453, "step": 6368 }, { "epoch": 0.5, "learning_rate": 9.500588096918373e-05, "loss": 2.8556, "step": 6369 }, { "epoch": 0.5, "learning_rate": 9.500509683995923e-05, "loss": 2.9554, "step": 6370 }, { "epoch": 0.5, "learning_rate": 9.500431271073474e-05, "loss": 2.9682, "step": 6371 }, { "epoch": 0.5, "learning_rate": 9.500352858151024e-05, "loss": 3.0599, "step": 6372 }, { "epoch": 0.5, "learning_rate": 9.500274445228574e-05, "loss": 3.2885, "step": 6373 }, { "epoch": 0.5, "learning_rate": 9.500196032306125e-05, "loss": 2.9299, "step": 6374 }, { "epoch": 0.5, "learning_rate": 9.500117619383675e-05, "loss": 3.0286, "step": 6375 }, { "epoch": 0.5, "learning_rate": 9.500039206461225e-05, "loss": 2.9538, "step": 6376 }, { "epoch": 0.5, "learning_rate": 9.499960793538776e-05, "loss": 3.2501, "step": 6377 }, { "epoch": 0.5, "learning_rate": 9.499882380616326e-05, "loss": 2.7805, "step": 6378 }, { "epoch": 0.5, "learning_rate": 9.499803967693876e-05, "loss": 3.1045, "step": 6379 }, { "epoch": 0.5, "learning_rate": 9.499725554771427e-05, "loss": 3.0304, "step": 6380 }, { "epoch": 0.5, "learning_rate": 9.499647141848977e-05, "loss": 2.9358, "step": 6381 }, { "epoch": 0.5, "learning_rate": 9.499568728926528e-05, "loss": 3.2871, "step": 6382 }, { "epoch": 0.5, "learning_rate": 9.499490316004078e-05, "loss": 3.0068, "step": 6383 }, { "epoch": 0.5, "learning_rate": 9.499411903081628e-05, "loss": 3.2452, "step": 6384 }, { "epoch": 0.5, "learning_rate": 9.499333490159179e-05, "loss": 3.1774, "step": 6385 }, { "epoch": 0.5, "learning_rate": 9.499255077236729e-05, "loss": 3.0206, "step": 6386 }, { "epoch": 0.5, "learning_rate": 9.49917666431428e-05, "loss": 3.0714, "step": 6387 }, { "epoch": 0.5, "learning_rate": 9.49909825139183e-05, "loss": 3.0927, "step": 6388 }, { "epoch": 0.5, "learning_rate": 9.49901983846938e-05, "loss": 3.0226, "step": 6389 }, { "epoch": 0.5, "learning_rate": 9.49894142554693e-05, "loss": 2.8524, "step": 6390 }, { "epoch": 0.5, "learning_rate": 9.498863012624481e-05, "loss": 2.9713, "step": 6391 }, { "epoch": 0.5, "learning_rate": 9.498784599702031e-05, "loss": 2.832, "step": 6392 }, { "epoch": 0.5, "learning_rate": 9.498706186779582e-05, "loss": 3.2372, "step": 6393 }, { "epoch": 0.5, "learning_rate": 9.498627773857132e-05, "loss": 2.8197, "step": 6394 }, { "epoch": 0.5, "learning_rate": 9.498549360934682e-05, "loss": 3.0183, "step": 6395 }, { "epoch": 0.5, "learning_rate": 9.498470948012233e-05, "loss": 3.1677, "step": 6396 }, { "epoch": 0.5, "learning_rate": 9.498392535089783e-05, "loss": 3.0848, "step": 6397 }, { "epoch": 0.5, "learning_rate": 9.498314122167333e-05, "loss": 3.0789, "step": 6398 }, { "epoch": 0.5, "learning_rate": 9.498235709244884e-05, "loss": 2.9836, "step": 6399 }, { "epoch": 0.5, "learning_rate": 9.498157296322434e-05, "loss": 3.0177, "step": 6400 }, { "epoch": 0.5, "learning_rate": 9.498078883399985e-05, "loss": 2.9788, "step": 6401 }, { "epoch": 0.5, "learning_rate": 9.498000470477535e-05, "loss": 2.9598, "step": 6402 }, { "epoch": 0.5, "learning_rate": 9.497922057555085e-05, "loss": 2.9781, "step": 6403 }, { "epoch": 0.5, "learning_rate": 9.497843644632636e-05, "loss": 3.074, "step": 6404 }, { "epoch": 0.5, "learning_rate": 9.497765231710186e-05, "loss": 2.9387, "step": 6405 }, { "epoch": 0.5, "learning_rate": 9.497686818787736e-05, "loss": 2.9656, "step": 6406 }, { "epoch": 0.5, "learning_rate": 9.497608405865287e-05, "loss": 2.7085, "step": 6407 }, { "epoch": 0.5, "learning_rate": 9.497529992942837e-05, "loss": 3.3774, "step": 6408 }, { "epoch": 0.5, "learning_rate": 9.497451580020387e-05, "loss": 3.0497, "step": 6409 }, { "epoch": 0.5, "learning_rate": 9.497373167097938e-05, "loss": 2.7708, "step": 6410 }, { "epoch": 0.5, "learning_rate": 9.497294754175488e-05, "loss": 3.0451, "step": 6411 }, { "epoch": 0.5, "learning_rate": 9.497216341253039e-05, "loss": 2.8974, "step": 6412 }, { "epoch": 0.5, "learning_rate": 9.497137928330589e-05, "loss": 2.9558, "step": 6413 }, { "epoch": 0.5, "learning_rate": 9.497059515408139e-05, "loss": 3.0039, "step": 6414 }, { "epoch": 0.5, "learning_rate": 9.49698110248569e-05, "loss": 3.3249, "step": 6415 }, { "epoch": 0.5, "learning_rate": 9.49690268956324e-05, "loss": 2.8475, "step": 6416 }, { "epoch": 0.5, "learning_rate": 9.496824276640792e-05, "loss": 3.212, "step": 6417 }, { "epoch": 0.5, "learning_rate": 9.496745863718342e-05, "loss": 2.8928, "step": 6418 }, { "epoch": 0.5, "learning_rate": 9.496667450795893e-05, "loss": 3.0328, "step": 6419 }, { "epoch": 0.5, "learning_rate": 9.496589037873443e-05, "loss": 3.1599, "step": 6420 }, { "epoch": 0.5, "learning_rate": 9.496510624950993e-05, "loss": 3.1008, "step": 6421 }, { "epoch": 0.5, "learning_rate": 9.496432212028544e-05, "loss": 3.2142, "step": 6422 }, { "epoch": 0.5, "learning_rate": 9.496353799106094e-05, "loss": 3.1945, "step": 6423 }, { "epoch": 0.5, "learning_rate": 9.496275386183644e-05, "loss": 2.8744, "step": 6424 }, { "epoch": 0.5, "learning_rate": 9.496196973261195e-05, "loss": 2.7707, "step": 6425 }, { "epoch": 0.5, "learning_rate": 9.496118560338745e-05, "loss": 3.3194, "step": 6426 }, { "epoch": 0.5, "learning_rate": 9.496040147416294e-05, "loss": 2.94, "step": 6427 }, { "epoch": 0.5, "learning_rate": 9.495961734493844e-05, "loss": 2.9853, "step": 6428 }, { "epoch": 0.5, "learning_rate": 9.495883321571395e-05, "loss": 2.8738, "step": 6429 }, { "epoch": 0.5, "learning_rate": 9.495804908648945e-05, "loss": 3.2043, "step": 6430 }, { "epoch": 0.5, "learning_rate": 9.495726495726496e-05, "loss": 2.998, "step": 6431 }, { "epoch": 0.5, "learning_rate": 9.495648082804046e-05, "loss": 2.9441, "step": 6432 }, { "epoch": 0.5, "learning_rate": 9.495569669881596e-05, "loss": 3.029, "step": 6433 }, { "epoch": 0.5, "learning_rate": 9.495491256959147e-05, "loss": 3.1804, "step": 6434 }, { "epoch": 0.5, "learning_rate": 9.495412844036697e-05, "loss": 2.9591, "step": 6435 }, { "epoch": 0.5, "learning_rate": 9.495334431114247e-05, "loss": 3.239, "step": 6436 }, { "epoch": 0.5, "learning_rate": 9.495256018191798e-05, "loss": 3.0369, "step": 6437 }, { "epoch": 0.5, "learning_rate": 9.495177605269348e-05, "loss": 2.8472, "step": 6438 }, { "epoch": 0.5, "learning_rate": 9.495099192346899e-05, "loss": 2.973, "step": 6439 }, { "epoch": 0.5, "learning_rate": 9.495020779424449e-05, "loss": 2.9222, "step": 6440 }, { "epoch": 0.51, "learning_rate": 9.494942366501999e-05, "loss": 3.1973, "step": 6441 }, { "epoch": 0.51, "learning_rate": 9.49486395357955e-05, "loss": 3.0684, "step": 6442 }, { "epoch": 0.51, "learning_rate": 9.494785540657101e-05, "loss": 2.8435, "step": 6443 }, { "epoch": 0.51, "learning_rate": 9.494707127734652e-05, "loss": 2.8185, "step": 6444 }, { "epoch": 0.51, "learning_rate": 9.494628714812202e-05, "loss": 2.9353, "step": 6445 }, { "epoch": 0.51, "learning_rate": 9.494550301889752e-05, "loss": 3.0919, "step": 6446 }, { "epoch": 0.51, "learning_rate": 9.494471888967303e-05, "loss": 3.3854, "step": 6447 }, { "epoch": 0.51, "learning_rate": 9.494393476044853e-05, "loss": 3.1314, "step": 6448 }, { "epoch": 0.51, "learning_rate": 9.494315063122404e-05, "loss": 3.3308, "step": 6449 }, { "epoch": 0.51, "learning_rate": 9.494236650199954e-05, "loss": 3.0402, "step": 6450 }, { "epoch": 0.51, "learning_rate": 9.494158237277504e-05, "loss": 2.8627, "step": 6451 }, { "epoch": 0.51, "learning_rate": 9.494079824355055e-05, "loss": 2.9121, "step": 6452 }, { "epoch": 0.51, "learning_rate": 9.494001411432605e-05, "loss": 3.1941, "step": 6453 }, { "epoch": 0.51, "learning_rate": 9.493922998510155e-05, "loss": 3.0795, "step": 6454 }, { "epoch": 0.51, "learning_rate": 9.493844585587706e-05, "loss": 2.9224, "step": 6455 }, { "epoch": 0.51, "learning_rate": 9.493766172665256e-05, "loss": 2.9711, "step": 6456 }, { "epoch": 0.51, "learning_rate": 9.493687759742806e-05, "loss": 3.021, "step": 6457 }, { "epoch": 0.51, "learning_rate": 9.493609346820357e-05, "loss": 3.0236, "step": 6458 }, { "epoch": 0.51, "learning_rate": 9.493530933897907e-05, "loss": 3.0167, "step": 6459 }, { "epoch": 0.51, "learning_rate": 9.493452520975458e-05, "loss": 2.9835, "step": 6460 }, { "epoch": 0.51, "learning_rate": 9.493374108053008e-05, "loss": 3.2034, "step": 6461 }, { "epoch": 0.51, "learning_rate": 9.493295695130558e-05, "loss": 2.885, "step": 6462 }, { "epoch": 0.51, "learning_rate": 9.493217282208107e-05, "loss": 2.943, "step": 6463 }, { "epoch": 0.51, "learning_rate": 9.493138869285658e-05, "loss": 2.9324, "step": 6464 }, { "epoch": 0.51, "learning_rate": 9.493060456363208e-05, "loss": 3.0273, "step": 6465 }, { "epoch": 0.51, "learning_rate": 9.492982043440758e-05, "loss": 3.02, "step": 6466 }, { "epoch": 0.51, "learning_rate": 9.492903630518309e-05, "loss": 3.4044, "step": 6467 }, { "epoch": 0.51, "learning_rate": 9.49282521759586e-05, "loss": 2.8829, "step": 6468 }, { "epoch": 0.51, "learning_rate": 9.492746804673411e-05, "loss": 2.9522, "step": 6469 }, { "epoch": 0.51, "learning_rate": 9.492668391750961e-05, "loss": 2.9824, "step": 6470 }, { "epoch": 0.51, "learning_rate": 9.492589978828512e-05, "loss": 3.0606, "step": 6471 }, { "epoch": 0.51, "learning_rate": 9.492511565906062e-05, "loss": 3.0466, "step": 6472 }, { "epoch": 0.51, "learning_rate": 9.492433152983612e-05, "loss": 2.9087, "step": 6473 }, { "epoch": 0.51, "learning_rate": 9.492354740061163e-05, "loss": 3.2125, "step": 6474 }, { "epoch": 0.51, "learning_rate": 9.492276327138713e-05, "loss": 3.3227, "step": 6475 }, { "epoch": 0.51, "learning_rate": 9.492197914216263e-05, "loss": 3.4202, "step": 6476 }, { "epoch": 0.51, "learning_rate": 9.492119501293814e-05, "loss": 3.1638, "step": 6477 }, { "epoch": 0.51, "learning_rate": 9.492041088371364e-05, "loss": 3.1213, "step": 6478 }, { "epoch": 0.51, "learning_rate": 9.491962675448915e-05, "loss": 2.9752, "step": 6479 }, { "epoch": 0.51, "learning_rate": 9.491884262526465e-05, "loss": 3.0668, "step": 6480 }, { "epoch": 0.51, "learning_rate": 9.491805849604015e-05, "loss": 3.1523, "step": 6481 }, { "epoch": 0.51, "learning_rate": 9.491727436681566e-05, "loss": 3.374, "step": 6482 }, { "epoch": 0.51, "learning_rate": 9.491649023759116e-05, "loss": 3.0813, "step": 6483 }, { "epoch": 0.51, "learning_rate": 9.491570610836666e-05, "loss": 3.2765, "step": 6484 }, { "epoch": 0.51, "learning_rate": 9.491492197914217e-05, "loss": 2.8253, "step": 6485 }, { "epoch": 0.51, "learning_rate": 9.491413784991767e-05, "loss": 3.1287, "step": 6486 }, { "epoch": 0.51, "learning_rate": 9.491335372069318e-05, "loss": 2.9404, "step": 6487 }, { "epoch": 0.51, "learning_rate": 9.491256959146868e-05, "loss": 2.9879, "step": 6488 }, { "epoch": 0.51, "learning_rate": 9.491178546224418e-05, "loss": 3.0147, "step": 6489 }, { "epoch": 0.51, "learning_rate": 9.491100133301969e-05, "loss": 2.9728, "step": 6490 }, { "epoch": 0.51, "learning_rate": 9.491021720379519e-05, "loss": 3.005, "step": 6491 }, { "epoch": 0.51, "learning_rate": 9.49094330745707e-05, "loss": 2.8574, "step": 6492 }, { "epoch": 0.51, "learning_rate": 9.49086489453462e-05, "loss": 3.0102, "step": 6493 }, { "epoch": 0.51, "learning_rate": 9.49078648161217e-05, "loss": 2.9153, "step": 6494 }, { "epoch": 0.51, "learning_rate": 9.49070806868972e-05, "loss": 2.8267, "step": 6495 }, { "epoch": 0.51, "learning_rate": 9.490629655767271e-05, "loss": 2.8376, "step": 6496 }, { "epoch": 0.51, "learning_rate": 9.490551242844821e-05, "loss": 2.8814, "step": 6497 }, { "epoch": 0.51, "learning_rate": 9.490472829922372e-05, "loss": 3.2641, "step": 6498 }, { "epoch": 0.51, "learning_rate": 9.490394416999922e-05, "loss": 3.2022, "step": 6499 }, { "epoch": 0.51, "learning_rate": 9.490316004077472e-05, "loss": 3.1524, "step": 6500 }, { "epoch": 0.51, "learning_rate": 9.490237591155023e-05, "loss": 2.9997, "step": 6501 }, { "epoch": 0.51, "learning_rate": 9.490159178232573e-05, "loss": 3.1859, "step": 6502 }, { "epoch": 0.51, "learning_rate": 9.490080765310123e-05, "loss": 3.089, "step": 6503 }, { "epoch": 0.51, "learning_rate": 9.490002352387674e-05, "loss": 3.0369, "step": 6504 }, { "epoch": 0.51, "learning_rate": 9.489923939465224e-05, "loss": 2.8689, "step": 6505 }, { "epoch": 0.51, "learning_rate": 9.489845526542775e-05, "loss": 2.9959, "step": 6506 }, { "epoch": 0.51, "learning_rate": 9.489767113620325e-05, "loss": 2.8794, "step": 6507 }, { "epoch": 0.51, "learning_rate": 9.489688700697875e-05, "loss": 3.0083, "step": 6508 }, { "epoch": 0.51, "learning_rate": 9.489610287775426e-05, "loss": 3.1526, "step": 6509 }, { "epoch": 0.51, "learning_rate": 9.489531874852976e-05, "loss": 2.9172, "step": 6510 }, { "epoch": 0.51, "learning_rate": 9.489453461930526e-05, "loss": 3.0877, "step": 6511 }, { "epoch": 0.51, "learning_rate": 9.489375049008077e-05, "loss": 2.9136, "step": 6512 }, { "epoch": 0.51, "learning_rate": 9.489296636085627e-05, "loss": 3.0953, "step": 6513 }, { "epoch": 0.51, "learning_rate": 9.489218223163177e-05, "loss": 2.9589, "step": 6514 }, { "epoch": 0.51, "learning_rate": 9.489139810240728e-05, "loss": 2.9525, "step": 6515 }, { "epoch": 0.51, "learning_rate": 9.489061397318278e-05, "loss": 2.7398, "step": 6516 }, { "epoch": 0.51, "learning_rate": 9.488982984395829e-05, "loss": 3.0759, "step": 6517 }, { "epoch": 0.51, "learning_rate": 9.488904571473379e-05, "loss": 2.9686, "step": 6518 }, { "epoch": 0.51, "learning_rate": 9.488826158550929e-05, "loss": 2.9522, "step": 6519 }, { "epoch": 0.51, "learning_rate": 9.48874774562848e-05, "loss": 3.0594, "step": 6520 }, { "epoch": 0.51, "learning_rate": 9.48866933270603e-05, "loss": 2.8651, "step": 6521 }, { "epoch": 0.51, "learning_rate": 9.48859091978358e-05, "loss": 2.8642, "step": 6522 }, { "epoch": 0.51, "learning_rate": 9.488512506861131e-05, "loss": 2.873, "step": 6523 }, { "epoch": 0.51, "learning_rate": 9.488434093938681e-05, "loss": 2.7041, "step": 6524 }, { "epoch": 0.51, "learning_rate": 9.488355681016233e-05, "loss": 3.1346, "step": 6525 }, { "epoch": 0.51, "learning_rate": 9.488277268093783e-05, "loss": 3.0457, "step": 6526 }, { "epoch": 0.51, "learning_rate": 9.488198855171334e-05, "loss": 3.1443, "step": 6527 }, { "epoch": 0.51, "learning_rate": 9.488120442248884e-05, "loss": 2.9306, "step": 6528 }, { "epoch": 0.51, "learning_rate": 9.488042029326434e-05, "loss": 3.15, "step": 6529 }, { "epoch": 0.51, "learning_rate": 9.487963616403985e-05, "loss": 2.9601, "step": 6530 }, { "epoch": 0.51, "learning_rate": 9.487885203481535e-05, "loss": 3.0266, "step": 6531 }, { "epoch": 0.51, "learning_rate": 9.487806790559085e-05, "loss": 3.0966, "step": 6532 }, { "epoch": 0.51, "learning_rate": 9.487728377636636e-05, "loss": 3.117, "step": 6533 }, { "epoch": 0.51, "learning_rate": 9.487649964714186e-05, "loss": 2.9469, "step": 6534 }, { "epoch": 0.51, "learning_rate": 9.487571551791735e-05, "loss": 3.1091, "step": 6535 }, { "epoch": 0.51, "learning_rate": 9.487493138869286e-05, "loss": 3.0107, "step": 6536 }, { "epoch": 0.51, "learning_rate": 9.487414725946836e-05, "loss": 3.0263, "step": 6537 }, { "epoch": 0.51, "learning_rate": 9.487336313024386e-05, "loss": 2.8235, "step": 6538 }, { "epoch": 0.51, "learning_rate": 9.487257900101937e-05, "loss": 3.266, "step": 6539 }, { "epoch": 0.51, "learning_rate": 9.487179487179487e-05, "loss": 3.0847, "step": 6540 }, { "epoch": 0.51, "learning_rate": 9.487101074257037e-05, "loss": 3.1349, "step": 6541 }, { "epoch": 0.51, "learning_rate": 9.487022661334588e-05, "loss": 2.821, "step": 6542 }, { "epoch": 0.51, "learning_rate": 9.486944248412138e-05, "loss": 3.0406, "step": 6543 }, { "epoch": 0.51, "learning_rate": 9.486865835489688e-05, "loss": 3.462, "step": 6544 }, { "epoch": 0.51, "learning_rate": 9.486787422567239e-05, "loss": 3.1116, "step": 6545 }, { "epoch": 0.51, "learning_rate": 9.486709009644789e-05, "loss": 3.1913, "step": 6546 }, { "epoch": 0.51, "learning_rate": 9.48663059672234e-05, "loss": 2.8806, "step": 6547 }, { "epoch": 0.51, "learning_rate": 9.48655218379989e-05, "loss": 3.208, "step": 6548 }, { "epoch": 0.51, "learning_rate": 9.48647377087744e-05, "loss": 2.9654, "step": 6549 }, { "epoch": 0.51, "learning_rate": 9.486395357954991e-05, "loss": 2.9887, "step": 6550 }, { "epoch": 0.51, "learning_rate": 9.486316945032542e-05, "loss": 3.0882, "step": 6551 }, { "epoch": 0.51, "learning_rate": 9.486238532110093e-05, "loss": 3.1072, "step": 6552 }, { "epoch": 0.51, "learning_rate": 9.486160119187643e-05, "loss": 2.7707, "step": 6553 }, { "epoch": 0.51, "learning_rate": 9.486081706265194e-05, "loss": 3.2358, "step": 6554 }, { "epoch": 0.51, "learning_rate": 9.486003293342744e-05, "loss": 3.0511, "step": 6555 }, { "epoch": 0.51, "learning_rate": 9.485924880420294e-05, "loss": 2.9353, "step": 6556 }, { "epoch": 0.51, "learning_rate": 9.485846467497845e-05, "loss": 3.2747, "step": 6557 }, { "epoch": 0.51, "learning_rate": 9.485768054575395e-05, "loss": 3.0911, "step": 6558 }, { "epoch": 0.51, "learning_rate": 9.485689641652945e-05, "loss": 3.0281, "step": 6559 }, { "epoch": 0.51, "learning_rate": 9.485611228730496e-05, "loss": 2.9921, "step": 6560 }, { "epoch": 0.51, "learning_rate": 9.485532815808046e-05, "loss": 2.8874, "step": 6561 }, { "epoch": 0.51, "learning_rate": 9.485454402885596e-05, "loss": 3.1054, "step": 6562 }, { "epoch": 0.51, "learning_rate": 9.485375989963147e-05, "loss": 3.1364, "step": 6563 }, { "epoch": 0.51, "learning_rate": 9.485297577040697e-05, "loss": 3.1427, "step": 6564 }, { "epoch": 0.51, "learning_rate": 9.485219164118248e-05, "loss": 3.1408, "step": 6565 }, { "epoch": 0.51, "learning_rate": 9.485140751195798e-05, "loss": 3.1091, "step": 6566 }, { "epoch": 0.51, "learning_rate": 9.485062338273348e-05, "loss": 3.1631, "step": 6567 }, { "epoch": 0.52, "learning_rate": 9.484983925350899e-05, "loss": 3.1203, "step": 6568 }, { "epoch": 0.52, "learning_rate": 9.484905512428449e-05, "loss": 2.9481, "step": 6569 }, { "epoch": 0.52, "learning_rate": 9.484827099506e-05, "loss": 3.1061, "step": 6570 }, { "epoch": 0.52, "learning_rate": 9.484748686583548e-05, "loss": 3.2672, "step": 6571 }, { "epoch": 0.52, "learning_rate": 9.484670273661099e-05, "loss": 3.2457, "step": 6572 }, { "epoch": 0.52, "learning_rate": 9.484591860738649e-05, "loss": 3.132, "step": 6573 }, { "epoch": 0.52, "learning_rate": 9.4845134478162e-05, "loss": 3.2258, "step": 6574 }, { "epoch": 0.52, "learning_rate": 9.48443503489375e-05, "loss": 3.0268, "step": 6575 }, { "epoch": 0.52, "learning_rate": 9.484356621971302e-05, "loss": 3.391, "step": 6576 }, { "epoch": 0.52, "learning_rate": 9.484278209048852e-05, "loss": 3.419, "step": 6577 }, { "epoch": 0.52, "learning_rate": 9.484199796126402e-05, "loss": 3.0409, "step": 6578 }, { "epoch": 0.52, "learning_rate": 9.484121383203953e-05, "loss": 2.9453, "step": 6579 }, { "epoch": 0.52, "learning_rate": 9.484042970281503e-05, "loss": 3.075, "step": 6580 }, { "epoch": 0.52, "learning_rate": 9.483964557359053e-05, "loss": 3.063, "step": 6581 }, { "epoch": 0.52, "learning_rate": 9.483886144436604e-05, "loss": 3.073, "step": 6582 }, { "epoch": 0.52, "learning_rate": 9.483807731514154e-05, "loss": 3.229, "step": 6583 }, { "epoch": 0.52, "learning_rate": 9.483729318591705e-05, "loss": 3.2354, "step": 6584 }, { "epoch": 0.52, "learning_rate": 9.483650905669255e-05, "loss": 3.1575, "step": 6585 }, { "epoch": 0.52, "learning_rate": 9.483572492746805e-05, "loss": 2.9233, "step": 6586 }, { "epoch": 0.52, "learning_rate": 9.483494079824356e-05, "loss": 3.2475, "step": 6587 }, { "epoch": 0.52, "learning_rate": 9.483415666901906e-05, "loss": 2.9322, "step": 6588 }, { "epoch": 0.52, "learning_rate": 9.483337253979456e-05, "loss": 2.9685, "step": 6589 }, { "epoch": 0.52, "learning_rate": 9.483258841057007e-05, "loss": 3.192, "step": 6590 }, { "epoch": 0.52, "learning_rate": 9.483180428134557e-05, "loss": 3.033, "step": 6591 }, { "epoch": 0.52, "learning_rate": 9.483102015212107e-05, "loss": 3.0442, "step": 6592 }, { "epoch": 0.52, "learning_rate": 9.483023602289658e-05, "loss": 3.0494, "step": 6593 }, { "epoch": 0.52, "learning_rate": 9.482945189367208e-05, "loss": 2.9344, "step": 6594 }, { "epoch": 0.52, "learning_rate": 9.482866776444759e-05, "loss": 2.9509, "step": 6595 }, { "epoch": 0.52, "learning_rate": 9.482788363522309e-05, "loss": 2.9245, "step": 6596 }, { "epoch": 0.52, "learning_rate": 9.48270995059986e-05, "loss": 3.2041, "step": 6597 }, { "epoch": 0.52, "learning_rate": 9.48263153767741e-05, "loss": 2.9715, "step": 6598 }, { "epoch": 0.52, "learning_rate": 9.48255312475496e-05, "loss": 3.1331, "step": 6599 }, { "epoch": 0.52, "learning_rate": 9.48247471183251e-05, "loss": 3.0189, "step": 6600 }, { "epoch": 0.52, "learning_rate": 9.482396298910061e-05, "loss": 3.2955, "step": 6601 }, { "epoch": 0.52, "learning_rate": 9.482317885987611e-05, "loss": 2.9536, "step": 6602 }, { "epoch": 0.52, "learning_rate": 9.482239473065162e-05, "loss": 3.1788, "step": 6603 }, { "epoch": 0.52, "learning_rate": 9.482161060142712e-05, "loss": 3.4481, "step": 6604 }, { "epoch": 0.52, "learning_rate": 9.482082647220262e-05, "loss": 3.0769, "step": 6605 }, { "epoch": 0.52, "learning_rate": 9.482004234297813e-05, "loss": 2.9689, "step": 6606 }, { "epoch": 0.52, "learning_rate": 9.481925821375363e-05, "loss": 2.9832, "step": 6607 }, { "epoch": 0.52, "learning_rate": 9.481847408452913e-05, "loss": 3.098, "step": 6608 }, { "epoch": 0.52, "learning_rate": 9.481768995530464e-05, "loss": 3.1413, "step": 6609 }, { "epoch": 0.52, "learning_rate": 9.481690582608014e-05, "loss": 3.0735, "step": 6610 }, { "epoch": 0.52, "learning_rate": 9.481612169685564e-05, "loss": 3.1596, "step": 6611 }, { "epoch": 0.52, "learning_rate": 9.481533756763115e-05, "loss": 3.0494, "step": 6612 }, { "epoch": 0.52, "learning_rate": 9.481455343840665e-05, "loss": 3.0322, "step": 6613 }, { "epoch": 0.52, "learning_rate": 9.481376930918216e-05, "loss": 3.0178, "step": 6614 }, { "epoch": 0.52, "learning_rate": 9.481298517995766e-05, "loss": 3.1702, "step": 6615 }, { "epoch": 0.52, "learning_rate": 9.481220105073316e-05, "loss": 2.9114, "step": 6616 }, { "epoch": 0.52, "learning_rate": 9.481141692150867e-05, "loss": 3.0268, "step": 6617 }, { "epoch": 0.52, "learning_rate": 9.481063279228417e-05, "loss": 3.1546, "step": 6618 }, { "epoch": 0.52, "learning_rate": 9.480984866305967e-05, "loss": 3.3458, "step": 6619 }, { "epoch": 0.52, "learning_rate": 9.480906453383518e-05, "loss": 3.0553, "step": 6620 }, { "epoch": 0.52, "learning_rate": 9.480828040461068e-05, "loss": 3.3167, "step": 6621 }, { "epoch": 0.52, "learning_rate": 9.480749627538619e-05, "loss": 3.1932, "step": 6622 }, { "epoch": 0.52, "learning_rate": 9.480671214616169e-05, "loss": 3.0608, "step": 6623 }, { "epoch": 0.52, "learning_rate": 9.480592801693719e-05, "loss": 3.2474, "step": 6624 }, { "epoch": 0.52, "learning_rate": 9.48051438877127e-05, "loss": 2.9767, "step": 6625 }, { "epoch": 0.52, "learning_rate": 9.48043597584882e-05, "loss": 3.0102, "step": 6626 }, { "epoch": 0.52, "learning_rate": 9.48035756292637e-05, "loss": 3.1769, "step": 6627 }, { "epoch": 0.52, "learning_rate": 9.480279150003921e-05, "loss": 2.9621, "step": 6628 }, { "epoch": 0.52, "learning_rate": 9.480200737081471e-05, "loss": 2.849, "step": 6629 }, { "epoch": 0.52, "learning_rate": 9.480122324159021e-05, "loss": 2.7336, "step": 6630 }, { "epoch": 0.52, "learning_rate": 9.480043911236572e-05, "loss": 3.1712, "step": 6631 }, { "epoch": 0.52, "learning_rate": 9.479965498314122e-05, "loss": 3.1777, "step": 6632 }, { "epoch": 0.52, "learning_rate": 9.479887085391674e-05, "loss": 2.9843, "step": 6633 }, { "epoch": 0.52, "learning_rate": 9.479808672469224e-05, "loss": 3.2555, "step": 6634 }, { "epoch": 0.52, "learning_rate": 9.479730259546775e-05, "loss": 2.884, "step": 6635 }, { "epoch": 0.52, "learning_rate": 9.479651846624325e-05, "loss": 3.1954, "step": 6636 }, { "epoch": 0.52, "learning_rate": 9.479573433701875e-05, "loss": 3.0334, "step": 6637 }, { "epoch": 0.52, "learning_rate": 9.479495020779426e-05, "loss": 3.0558, "step": 6638 }, { "epoch": 0.52, "learning_rate": 9.479416607856976e-05, "loss": 2.7061, "step": 6639 }, { "epoch": 0.52, "learning_rate": 9.479338194934526e-05, "loss": 3.025, "step": 6640 }, { "epoch": 0.52, "learning_rate": 9.479259782012077e-05, "loss": 2.8171, "step": 6641 }, { "epoch": 0.52, "learning_rate": 9.479181369089627e-05, "loss": 3.3028, "step": 6642 }, { "epoch": 0.52, "learning_rate": 9.479102956167176e-05, "loss": 3.0181, "step": 6643 }, { "epoch": 0.52, "learning_rate": 9.479024543244727e-05, "loss": 3.0146, "step": 6644 }, { "epoch": 0.52, "learning_rate": 9.478946130322277e-05, "loss": 2.8345, "step": 6645 }, { "epoch": 0.52, "learning_rate": 9.478867717399827e-05, "loss": 3.156, "step": 6646 }, { "epoch": 0.52, "learning_rate": 9.478789304477378e-05, "loss": 3.0086, "step": 6647 }, { "epoch": 0.52, "learning_rate": 9.478710891554928e-05, "loss": 3.0904, "step": 6648 }, { "epoch": 0.52, "learning_rate": 9.478632478632478e-05, "loss": 3.1138, "step": 6649 }, { "epoch": 0.52, "learning_rate": 9.478554065710029e-05, "loss": 3.0336, "step": 6650 }, { "epoch": 0.52, "learning_rate": 9.478475652787579e-05, "loss": 3.1698, "step": 6651 }, { "epoch": 0.52, "learning_rate": 9.47839723986513e-05, "loss": 3.0018, "step": 6652 }, { "epoch": 0.52, "learning_rate": 9.47831882694268e-05, "loss": 2.9936, "step": 6653 }, { "epoch": 0.52, "learning_rate": 9.47824041402023e-05, "loss": 2.9253, "step": 6654 }, { "epoch": 0.52, "learning_rate": 9.47816200109778e-05, "loss": 2.9258, "step": 6655 }, { "epoch": 0.52, "learning_rate": 9.478083588175331e-05, "loss": 3.0958, "step": 6656 }, { "epoch": 0.52, "learning_rate": 9.478005175252881e-05, "loss": 2.9516, "step": 6657 }, { "epoch": 0.52, "learning_rate": 9.477926762330432e-05, "loss": 3.013, "step": 6658 }, { "epoch": 0.52, "learning_rate": 9.477848349407983e-05, "loss": 3.1808, "step": 6659 }, { "epoch": 0.52, "learning_rate": 9.477769936485534e-05, "loss": 2.9412, "step": 6660 }, { "epoch": 0.52, "learning_rate": 9.477691523563084e-05, "loss": 3.0754, "step": 6661 }, { "epoch": 0.52, "learning_rate": 9.477613110640635e-05, "loss": 2.8971, "step": 6662 }, { "epoch": 0.52, "learning_rate": 9.477534697718185e-05, "loss": 2.808, "step": 6663 }, { "epoch": 0.52, "learning_rate": 9.477456284795735e-05, "loss": 2.8077, "step": 6664 }, { "epoch": 0.52, "learning_rate": 9.477377871873286e-05, "loss": 2.8252, "step": 6665 }, { "epoch": 0.52, "learning_rate": 9.477299458950836e-05, "loss": 3.0526, "step": 6666 }, { "epoch": 0.52, "learning_rate": 9.477221046028386e-05, "loss": 2.9956, "step": 6667 }, { "epoch": 0.52, "learning_rate": 9.477142633105937e-05, "loss": 2.6986, "step": 6668 }, { "epoch": 0.52, "learning_rate": 9.477064220183487e-05, "loss": 2.888, "step": 6669 }, { "epoch": 0.52, "learning_rate": 9.476985807261038e-05, "loss": 3.0136, "step": 6670 }, { "epoch": 0.52, "learning_rate": 9.476907394338588e-05, "loss": 3.0691, "step": 6671 }, { "epoch": 0.52, "learning_rate": 9.476828981416138e-05, "loss": 3.1777, "step": 6672 }, { "epoch": 0.52, "learning_rate": 9.476750568493689e-05, "loss": 3.0484, "step": 6673 }, { "epoch": 0.52, "learning_rate": 9.476672155571239e-05, "loss": 3.0334, "step": 6674 }, { "epoch": 0.52, "learning_rate": 9.47659374264879e-05, "loss": 2.9906, "step": 6675 }, { "epoch": 0.52, "learning_rate": 9.47651532972634e-05, "loss": 3.2212, "step": 6676 }, { "epoch": 0.52, "learning_rate": 9.47643691680389e-05, "loss": 2.9349, "step": 6677 }, { "epoch": 0.52, "learning_rate": 9.47635850388144e-05, "loss": 3.0326, "step": 6678 }, { "epoch": 0.52, "learning_rate": 9.47628009095899e-05, "loss": 2.9881, "step": 6679 }, { "epoch": 0.52, "learning_rate": 9.47620167803654e-05, "loss": 3.3451, "step": 6680 }, { "epoch": 0.52, "learning_rate": 9.47612326511409e-05, "loss": 2.7578, "step": 6681 }, { "epoch": 0.52, "learning_rate": 9.47604485219164e-05, "loss": 3.0588, "step": 6682 }, { "epoch": 0.52, "learning_rate": 9.475966439269191e-05, "loss": 3.147, "step": 6683 }, { "epoch": 0.52, "learning_rate": 9.475888026346741e-05, "loss": 2.9112, "step": 6684 }, { "epoch": 0.52, "learning_rate": 9.475809613424293e-05, "loss": 3.1005, "step": 6685 }, { "epoch": 0.52, "learning_rate": 9.475731200501843e-05, "loss": 3.1241, "step": 6686 }, { "epoch": 0.52, "learning_rate": 9.475652787579394e-05, "loss": 2.867, "step": 6687 }, { "epoch": 0.52, "learning_rate": 9.475574374656944e-05, "loss": 3.3743, "step": 6688 }, { "epoch": 0.52, "learning_rate": 9.475495961734495e-05, "loss": 3.3479, "step": 6689 }, { "epoch": 0.52, "learning_rate": 9.475417548812045e-05, "loss": 3.2826, "step": 6690 }, { "epoch": 0.52, "learning_rate": 9.475339135889595e-05, "loss": 2.7922, "step": 6691 }, { "epoch": 0.52, "learning_rate": 9.475260722967146e-05, "loss": 2.8708, "step": 6692 }, { "epoch": 0.52, "learning_rate": 9.475182310044696e-05, "loss": 2.8038, "step": 6693 }, { "epoch": 0.52, "learning_rate": 9.475103897122246e-05, "loss": 3.1573, "step": 6694 }, { "epoch": 0.52, "learning_rate": 9.475025484199797e-05, "loss": 3.2146, "step": 6695 }, { "epoch": 0.53, "learning_rate": 9.474947071277347e-05, "loss": 3.1582, "step": 6696 }, { "epoch": 0.53, "learning_rate": 9.474868658354897e-05, "loss": 3.132, "step": 6697 }, { "epoch": 0.53, "learning_rate": 9.474790245432448e-05, "loss": 3.2093, "step": 6698 }, { "epoch": 0.53, "learning_rate": 9.474711832509998e-05, "loss": 3.1315, "step": 6699 }, { "epoch": 0.53, "learning_rate": 9.474633419587549e-05, "loss": 2.8634, "step": 6700 }, { "epoch": 0.53, "learning_rate": 9.474555006665099e-05, "loss": 3.1743, "step": 6701 }, { "epoch": 0.53, "learning_rate": 9.474476593742649e-05, "loss": 3.1835, "step": 6702 }, { "epoch": 0.53, "learning_rate": 9.4743981808202e-05, "loss": 3.0443, "step": 6703 }, { "epoch": 0.53, "learning_rate": 9.47431976789775e-05, "loss": 3.2309, "step": 6704 }, { "epoch": 0.53, "learning_rate": 9.4742413549753e-05, "loss": 3.08, "step": 6705 }, { "epoch": 0.53, "learning_rate": 9.474162942052851e-05, "loss": 3.0192, "step": 6706 }, { "epoch": 0.53, "learning_rate": 9.474084529130401e-05, "loss": 2.9376, "step": 6707 }, { "epoch": 0.53, "learning_rate": 9.474006116207952e-05, "loss": 3.1925, "step": 6708 }, { "epoch": 0.53, "learning_rate": 9.473927703285502e-05, "loss": 2.9655, "step": 6709 }, { "epoch": 0.53, "learning_rate": 9.473849290363052e-05, "loss": 3.2388, "step": 6710 }, { "epoch": 0.53, "learning_rate": 9.473770877440603e-05, "loss": 3.1847, "step": 6711 }, { "epoch": 0.53, "learning_rate": 9.473692464518153e-05, "loss": 2.8895, "step": 6712 }, { "epoch": 0.53, "learning_rate": 9.473614051595703e-05, "loss": 3.1139, "step": 6713 }, { "epoch": 0.53, "learning_rate": 9.473535638673254e-05, "loss": 3.259, "step": 6714 }, { "epoch": 0.53, "learning_rate": 9.473457225750804e-05, "loss": 2.9267, "step": 6715 }, { "epoch": 0.53, "learning_rate": 9.473378812828354e-05, "loss": 3.2244, "step": 6716 }, { "epoch": 0.53, "learning_rate": 9.473300399905905e-05, "loss": 2.8889, "step": 6717 }, { "epoch": 0.53, "learning_rate": 9.473221986983455e-05, "loss": 3.0147, "step": 6718 }, { "epoch": 0.53, "learning_rate": 9.473143574061006e-05, "loss": 3.1724, "step": 6719 }, { "epoch": 0.53, "learning_rate": 9.473065161138556e-05, "loss": 3.018, "step": 6720 }, { "epoch": 0.53, "learning_rate": 9.472986748216106e-05, "loss": 3.0416, "step": 6721 }, { "epoch": 0.53, "learning_rate": 9.472908335293657e-05, "loss": 2.8996, "step": 6722 }, { "epoch": 0.53, "learning_rate": 9.472829922371207e-05, "loss": 2.8688, "step": 6723 }, { "epoch": 0.53, "learning_rate": 9.472751509448757e-05, "loss": 3.0051, "step": 6724 }, { "epoch": 0.53, "learning_rate": 9.472673096526308e-05, "loss": 3.0515, "step": 6725 }, { "epoch": 0.53, "learning_rate": 9.472594683603858e-05, "loss": 3.1413, "step": 6726 }, { "epoch": 0.53, "learning_rate": 9.472516270681408e-05, "loss": 3.0871, "step": 6727 }, { "epoch": 0.53, "learning_rate": 9.472437857758959e-05, "loss": 3.0751, "step": 6728 }, { "epoch": 0.53, "learning_rate": 9.472359444836509e-05, "loss": 3.1603, "step": 6729 }, { "epoch": 0.53, "learning_rate": 9.47228103191406e-05, "loss": 2.8142, "step": 6730 }, { "epoch": 0.53, "learning_rate": 9.47220261899161e-05, "loss": 3.2201, "step": 6731 }, { "epoch": 0.53, "learning_rate": 9.47212420606916e-05, "loss": 2.9184, "step": 6732 }, { "epoch": 0.53, "learning_rate": 9.472045793146711e-05, "loss": 2.9142, "step": 6733 }, { "epoch": 0.53, "learning_rate": 9.471967380224261e-05, "loss": 3.0756, "step": 6734 }, { "epoch": 0.53, "learning_rate": 9.471888967301811e-05, "loss": 3.0945, "step": 6735 }, { "epoch": 0.53, "learning_rate": 9.471810554379362e-05, "loss": 2.9717, "step": 6736 }, { "epoch": 0.53, "learning_rate": 9.471732141456912e-05, "loss": 3.2034, "step": 6737 }, { "epoch": 0.53, "learning_rate": 9.471653728534463e-05, "loss": 2.9218, "step": 6738 }, { "epoch": 0.53, "learning_rate": 9.471575315612013e-05, "loss": 3.1941, "step": 6739 }, { "epoch": 0.53, "learning_rate": 9.471496902689563e-05, "loss": 3.1794, "step": 6740 }, { "epoch": 0.53, "learning_rate": 9.471418489767115e-05, "loss": 3.1791, "step": 6741 }, { "epoch": 0.53, "learning_rate": 9.471340076844665e-05, "loss": 2.9681, "step": 6742 }, { "epoch": 0.53, "learning_rate": 9.471261663922216e-05, "loss": 3.0894, "step": 6743 }, { "epoch": 0.53, "learning_rate": 9.471183250999766e-05, "loss": 2.9202, "step": 6744 }, { "epoch": 0.53, "learning_rate": 9.471104838077316e-05, "loss": 3.0267, "step": 6745 }, { "epoch": 0.53, "learning_rate": 9.471026425154867e-05, "loss": 3.0616, "step": 6746 }, { "epoch": 0.53, "learning_rate": 9.470948012232417e-05, "loss": 2.988, "step": 6747 }, { "epoch": 0.53, "learning_rate": 9.470869599309968e-05, "loss": 3.1793, "step": 6748 }, { "epoch": 0.53, "learning_rate": 9.470791186387518e-05, "loss": 2.7966, "step": 6749 }, { "epoch": 0.53, "learning_rate": 9.470712773465068e-05, "loss": 2.9192, "step": 6750 }, { "epoch": 0.53, "learning_rate": 9.470634360542617e-05, "loss": 3.0904, "step": 6751 }, { "epoch": 0.53, "learning_rate": 9.470555947620168e-05, "loss": 3.356, "step": 6752 }, { "epoch": 0.53, "learning_rate": 9.470477534697718e-05, "loss": 3.0977, "step": 6753 }, { "epoch": 0.53, "learning_rate": 9.470399121775268e-05, "loss": 3.316, "step": 6754 }, { "epoch": 0.53, "learning_rate": 9.470320708852819e-05, "loss": 3.1787, "step": 6755 }, { "epoch": 0.53, "learning_rate": 9.470242295930369e-05, "loss": 2.8336, "step": 6756 }, { "epoch": 0.53, "learning_rate": 9.47016388300792e-05, "loss": 3.0851, "step": 6757 }, { "epoch": 0.53, "learning_rate": 9.47008547008547e-05, "loss": 2.8861, "step": 6758 }, { "epoch": 0.53, "learning_rate": 9.47000705716302e-05, "loss": 3.0556, "step": 6759 }, { "epoch": 0.53, "learning_rate": 9.46992864424057e-05, "loss": 3.0832, "step": 6760 }, { "epoch": 0.53, "learning_rate": 9.469850231318121e-05, "loss": 3.1531, "step": 6761 }, { "epoch": 0.53, "learning_rate": 9.469771818395671e-05, "loss": 3.2286, "step": 6762 }, { "epoch": 0.53, "learning_rate": 9.469693405473222e-05, "loss": 3.1641, "step": 6763 }, { "epoch": 0.53, "learning_rate": 9.469614992550772e-05, "loss": 2.8775, "step": 6764 }, { "epoch": 0.53, "learning_rate": 9.469536579628322e-05, "loss": 3.0844, "step": 6765 }, { "epoch": 0.53, "learning_rate": 9.469458166705873e-05, "loss": 3.0726, "step": 6766 }, { "epoch": 0.53, "learning_rate": 9.469379753783425e-05, "loss": 3.0072, "step": 6767 }, { "epoch": 0.53, "learning_rate": 9.469301340860975e-05, "loss": 3.2343, "step": 6768 }, { "epoch": 0.53, "learning_rate": 9.469222927938525e-05, "loss": 3.1549, "step": 6769 }, { "epoch": 0.53, "learning_rate": 9.469144515016076e-05, "loss": 2.9886, "step": 6770 }, { "epoch": 0.53, "learning_rate": 9.469066102093626e-05, "loss": 3.1659, "step": 6771 }, { "epoch": 0.53, "learning_rate": 9.468987689171176e-05, "loss": 3.0882, "step": 6772 }, { "epoch": 0.53, "learning_rate": 9.468909276248727e-05, "loss": 2.9735, "step": 6773 }, { "epoch": 0.53, "learning_rate": 9.468830863326277e-05, "loss": 3.1015, "step": 6774 }, { "epoch": 0.53, "learning_rate": 9.468752450403828e-05, "loss": 3.2352, "step": 6775 }, { "epoch": 0.53, "learning_rate": 9.468674037481378e-05, "loss": 3.0812, "step": 6776 }, { "epoch": 0.53, "learning_rate": 9.468595624558928e-05, "loss": 3.0426, "step": 6777 }, { "epoch": 0.53, "learning_rate": 9.468517211636479e-05, "loss": 3.043, "step": 6778 }, { "epoch": 0.53, "learning_rate": 9.468438798714029e-05, "loss": 3.0301, "step": 6779 }, { "epoch": 0.53, "learning_rate": 9.46836038579158e-05, "loss": 3.1123, "step": 6780 }, { "epoch": 0.53, "learning_rate": 9.46828197286913e-05, "loss": 2.924, "step": 6781 }, { "epoch": 0.53, "learning_rate": 9.46820355994668e-05, "loss": 2.7151, "step": 6782 }, { "epoch": 0.53, "learning_rate": 9.46812514702423e-05, "loss": 3.1129, "step": 6783 }, { "epoch": 0.53, "learning_rate": 9.468046734101781e-05, "loss": 3.11, "step": 6784 }, { "epoch": 0.53, "learning_rate": 9.467968321179331e-05, "loss": 3.0384, "step": 6785 }, { "epoch": 0.53, "learning_rate": 9.467889908256882e-05, "loss": 2.881, "step": 6786 }, { "epoch": 0.53, "learning_rate": 9.46781149533443e-05, "loss": 3.0738, "step": 6787 }, { "epoch": 0.53, "learning_rate": 9.467733082411981e-05, "loss": 3.1055, "step": 6788 }, { "epoch": 0.53, "learning_rate": 9.467654669489531e-05, "loss": 2.824, "step": 6789 }, { "epoch": 0.53, "learning_rate": 9.467576256567082e-05, "loss": 2.9527, "step": 6790 }, { "epoch": 0.53, "learning_rate": 9.467497843644632e-05, "loss": 3.1018, "step": 6791 }, { "epoch": 0.53, "learning_rate": 9.467419430722182e-05, "loss": 2.9647, "step": 6792 }, { "epoch": 0.53, "learning_rate": 9.467341017799734e-05, "loss": 3.2544, "step": 6793 }, { "epoch": 0.53, "learning_rate": 9.467262604877284e-05, "loss": 2.9541, "step": 6794 }, { "epoch": 0.53, "learning_rate": 9.467184191954835e-05, "loss": 2.8686, "step": 6795 }, { "epoch": 0.53, "learning_rate": 9.467105779032385e-05, "loss": 2.8688, "step": 6796 }, { "epoch": 0.53, "learning_rate": 9.467027366109936e-05, "loss": 2.936, "step": 6797 }, { "epoch": 0.53, "learning_rate": 9.466948953187486e-05, "loss": 2.8087, "step": 6798 }, { "epoch": 0.53, "learning_rate": 9.466870540265036e-05, "loss": 2.9425, "step": 6799 }, { "epoch": 0.53, "learning_rate": 9.466792127342587e-05, "loss": 3.1744, "step": 6800 }, { "epoch": 0.53, "learning_rate": 9.466713714420137e-05, "loss": 2.9928, "step": 6801 }, { "epoch": 0.53, "learning_rate": 9.466635301497687e-05, "loss": 2.7824, "step": 6802 }, { "epoch": 0.53, "learning_rate": 9.466556888575238e-05, "loss": 3.0789, "step": 6803 }, { "epoch": 0.53, "learning_rate": 9.466478475652788e-05, "loss": 3.2275, "step": 6804 }, { "epoch": 0.53, "learning_rate": 9.466400062730339e-05, "loss": 3.0194, "step": 6805 }, { "epoch": 0.53, "learning_rate": 9.466321649807889e-05, "loss": 2.948, "step": 6806 }, { "epoch": 0.53, "learning_rate": 9.466243236885439e-05, "loss": 3.1289, "step": 6807 }, { "epoch": 0.53, "learning_rate": 9.46616482396299e-05, "loss": 3.0696, "step": 6808 }, { "epoch": 0.53, "learning_rate": 9.46608641104054e-05, "loss": 3.1689, "step": 6809 }, { "epoch": 0.53, "learning_rate": 9.46600799811809e-05, "loss": 3.3812, "step": 6810 }, { "epoch": 0.53, "learning_rate": 9.465929585195641e-05, "loss": 3.2188, "step": 6811 }, { "epoch": 0.53, "learning_rate": 9.465851172273191e-05, "loss": 2.9955, "step": 6812 }, { "epoch": 0.53, "learning_rate": 9.465772759350741e-05, "loss": 3.0579, "step": 6813 }, { "epoch": 0.53, "learning_rate": 9.465694346428292e-05, "loss": 3.1637, "step": 6814 }, { "epoch": 0.53, "learning_rate": 9.465615933505842e-05, "loss": 2.982, "step": 6815 }, { "epoch": 0.53, "learning_rate": 9.465537520583393e-05, "loss": 3.1529, "step": 6816 }, { "epoch": 0.53, "learning_rate": 9.465459107660943e-05, "loss": 3.3816, "step": 6817 }, { "epoch": 0.53, "learning_rate": 9.465380694738493e-05, "loss": 2.769, "step": 6818 }, { "epoch": 0.53, "learning_rate": 9.465302281816044e-05, "loss": 3.2591, "step": 6819 }, { "epoch": 0.53, "learning_rate": 9.465223868893594e-05, "loss": 2.8505, "step": 6820 }, { "epoch": 0.53, "learning_rate": 9.465145455971144e-05, "loss": 3.242, "step": 6821 }, { "epoch": 0.53, "learning_rate": 9.465067043048695e-05, "loss": 3.0779, "step": 6822 }, { "epoch": 0.54, "learning_rate": 9.464988630126245e-05, "loss": 3.0415, "step": 6823 }, { "epoch": 0.54, "learning_rate": 9.464910217203796e-05, "loss": 3.0236, "step": 6824 }, { "epoch": 0.54, "learning_rate": 9.464831804281346e-05, "loss": 2.9467, "step": 6825 }, { "epoch": 0.54, "learning_rate": 9.464753391358896e-05, "loss": 3.0168, "step": 6826 }, { "epoch": 0.54, "learning_rate": 9.464674978436447e-05, "loss": 3.1166, "step": 6827 }, { "epoch": 0.54, "learning_rate": 9.464596565513997e-05, "loss": 2.8988, "step": 6828 }, { "epoch": 0.54, "learning_rate": 9.464518152591547e-05, "loss": 3.0817, "step": 6829 }, { "epoch": 0.54, "learning_rate": 9.464439739669098e-05, "loss": 2.8991, "step": 6830 }, { "epoch": 0.54, "learning_rate": 9.464361326746648e-05, "loss": 3.2589, "step": 6831 }, { "epoch": 0.54, "learning_rate": 9.464282913824198e-05, "loss": 3.1458, "step": 6832 }, { "epoch": 0.54, "learning_rate": 9.464204500901749e-05, "loss": 2.8699, "step": 6833 }, { "epoch": 0.54, "learning_rate": 9.464126087979299e-05, "loss": 3.0917, "step": 6834 }, { "epoch": 0.54, "learning_rate": 9.46404767505685e-05, "loss": 3.1665, "step": 6835 }, { "epoch": 0.54, "learning_rate": 9.4639692621344e-05, "loss": 2.9125, "step": 6836 }, { "epoch": 0.54, "learning_rate": 9.46389084921195e-05, "loss": 3.2672, "step": 6837 }, { "epoch": 0.54, "learning_rate": 9.4638124362895e-05, "loss": 3.1846, "step": 6838 }, { "epoch": 0.54, "learning_rate": 9.463734023367051e-05, "loss": 2.8934, "step": 6839 }, { "epoch": 0.54, "learning_rate": 9.463655610444601e-05, "loss": 3.1774, "step": 6840 }, { "epoch": 0.54, "learning_rate": 9.463577197522152e-05, "loss": 2.8984, "step": 6841 }, { "epoch": 0.54, "learning_rate": 9.463498784599702e-05, "loss": 3.1049, "step": 6842 }, { "epoch": 0.54, "learning_rate": 9.463420371677253e-05, "loss": 3.2301, "step": 6843 }, { "epoch": 0.54, "learning_rate": 9.463341958754803e-05, "loss": 3.0102, "step": 6844 }, { "epoch": 0.54, "learning_rate": 9.463263545832353e-05, "loss": 2.9141, "step": 6845 }, { "epoch": 0.54, "learning_rate": 9.463185132909904e-05, "loss": 3.1571, "step": 6846 }, { "epoch": 0.54, "learning_rate": 9.463106719987454e-05, "loss": 2.7324, "step": 6847 }, { "epoch": 0.54, "learning_rate": 9.463028307065004e-05, "loss": 2.9593, "step": 6848 }, { "epoch": 0.54, "learning_rate": 9.462949894142556e-05, "loss": 2.9049, "step": 6849 }, { "epoch": 0.54, "learning_rate": 9.462871481220106e-05, "loss": 3.069, "step": 6850 }, { "epoch": 0.54, "learning_rate": 9.462793068297657e-05, "loss": 3.0519, "step": 6851 }, { "epoch": 0.54, "learning_rate": 9.462714655375207e-05, "loss": 3.3799, "step": 6852 }, { "epoch": 0.54, "learning_rate": 9.462636242452758e-05, "loss": 2.9637, "step": 6853 }, { "epoch": 0.54, "learning_rate": 9.462557829530308e-05, "loss": 2.8842, "step": 6854 }, { "epoch": 0.54, "learning_rate": 9.462479416607858e-05, "loss": 3.0171, "step": 6855 }, { "epoch": 0.54, "learning_rate": 9.462401003685409e-05, "loss": 3.0561, "step": 6856 }, { "epoch": 0.54, "learning_rate": 9.462322590762959e-05, "loss": 3.1378, "step": 6857 }, { "epoch": 0.54, "learning_rate": 9.46224417784051e-05, "loss": 3.157, "step": 6858 }, { "epoch": 0.54, "learning_rate": 9.462165764918058e-05, "loss": 3.0955, "step": 6859 }, { "epoch": 0.54, "learning_rate": 9.462087351995609e-05, "loss": 3.1266, "step": 6860 }, { "epoch": 0.54, "learning_rate": 9.462008939073159e-05, "loss": 3.0096, "step": 6861 }, { "epoch": 0.54, "learning_rate": 9.46193052615071e-05, "loss": 2.7782, "step": 6862 }, { "epoch": 0.54, "learning_rate": 9.46185211322826e-05, "loss": 2.8284, "step": 6863 }, { "epoch": 0.54, "learning_rate": 9.46177370030581e-05, "loss": 3.0158, "step": 6864 }, { "epoch": 0.54, "learning_rate": 9.46169528738336e-05, "loss": 3.1423, "step": 6865 }, { "epoch": 0.54, "learning_rate": 9.461616874460911e-05, "loss": 3.1383, "step": 6866 }, { "epoch": 0.54, "learning_rate": 9.461538461538461e-05, "loss": 2.7748, "step": 6867 }, { "epoch": 0.54, "learning_rate": 9.461460048616012e-05, "loss": 3.1132, "step": 6868 }, { "epoch": 0.54, "learning_rate": 9.461381635693562e-05, "loss": 3.1166, "step": 6869 }, { "epoch": 0.54, "learning_rate": 9.461303222771112e-05, "loss": 2.9164, "step": 6870 }, { "epoch": 0.54, "learning_rate": 9.461224809848663e-05, "loss": 2.9357, "step": 6871 }, { "epoch": 0.54, "learning_rate": 9.461146396926213e-05, "loss": 2.9317, "step": 6872 }, { "epoch": 0.54, "learning_rate": 9.461067984003764e-05, "loss": 3.0648, "step": 6873 }, { "epoch": 0.54, "learning_rate": 9.460989571081314e-05, "loss": 2.9629, "step": 6874 }, { "epoch": 0.54, "learning_rate": 9.460911158158866e-05, "loss": 3.2496, "step": 6875 }, { "epoch": 0.54, "learning_rate": 9.460832745236416e-05, "loss": 2.9377, "step": 6876 }, { "epoch": 0.54, "learning_rate": 9.460754332313966e-05, "loss": 3.0708, "step": 6877 }, { "epoch": 0.54, "learning_rate": 9.460675919391517e-05, "loss": 2.9952, "step": 6878 }, { "epoch": 0.54, "learning_rate": 9.460597506469067e-05, "loss": 2.8283, "step": 6879 }, { "epoch": 0.54, "learning_rate": 9.460519093546617e-05, "loss": 2.8442, "step": 6880 }, { "epoch": 0.54, "learning_rate": 9.460440680624168e-05, "loss": 3.1239, "step": 6881 }, { "epoch": 0.54, "learning_rate": 9.460362267701718e-05, "loss": 3.1749, "step": 6882 }, { "epoch": 0.54, "learning_rate": 9.460283854779269e-05, "loss": 3.1116, "step": 6883 }, { "epoch": 0.54, "learning_rate": 9.460205441856819e-05, "loss": 2.9236, "step": 6884 }, { "epoch": 0.54, "learning_rate": 9.460127028934369e-05, "loss": 3.4337, "step": 6885 }, { "epoch": 0.54, "learning_rate": 9.46004861601192e-05, "loss": 2.9743, "step": 6886 }, { "epoch": 0.54, "learning_rate": 9.45997020308947e-05, "loss": 2.8534, "step": 6887 }, { "epoch": 0.54, "learning_rate": 9.45989179016702e-05, "loss": 3.1151, "step": 6888 }, { "epoch": 0.54, "learning_rate": 9.459813377244571e-05, "loss": 2.9927, "step": 6889 }, { "epoch": 0.54, "learning_rate": 9.459734964322121e-05, "loss": 2.7203, "step": 6890 }, { "epoch": 0.54, "learning_rate": 9.459656551399672e-05, "loss": 2.7952, "step": 6891 }, { "epoch": 0.54, "learning_rate": 9.459578138477222e-05, "loss": 3.0106, "step": 6892 }, { "epoch": 0.54, "learning_rate": 9.459499725554772e-05, "loss": 3.1497, "step": 6893 }, { "epoch": 0.54, "learning_rate": 9.459421312632323e-05, "loss": 3.0133, "step": 6894 }, { "epoch": 0.54, "learning_rate": 9.459342899709872e-05, "loss": 2.8459, "step": 6895 }, { "epoch": 0.54, "learning_rate": 9.459264486787422e-05, "loss": 3.0507, "step": 6896 }, { "epoch": 0.54, "learning_rate": 9.459186073864972e-05, "loss": 2.936, "step": 6897 }, { "epoch": 0.54, "learning_rate": 9.459107660942523e-05, "loss": 3.1763, "step": 6898 }, { "epoch": 0.54, "learning_rate": 9.459029248020073e-05, "loss": 3.1728, "step": 6899 }, { "epoch": 0.54, "learning_rate": 9.458950835097623e-05, "loss": 3.0748, "step": 6900 }, { "epoch": 0.54, "learning_rate": 9.458872422175175e-05, "loss": 2.957, "step": 6901 }, { "epoch": 0.54, "learning_rate": 9.458794009252726e-05, "loss": 3.0372, "step": 6902 }, { "epoch": 0.54, "learning_rate": 9.458715596330276e-05, "loss": 2.9085, "step": 6903 }, { "epoch": 0.54, "learning_rate": 9.458637183407826e-05, "loss": 2.9069, "step": 6904 }, { "epoch": 0.54, "learning_rate": 9.458558770485377e-05, "loss": 3.246, "step": 6905 }, { "epoch": 0.54, "learning_rate": 9.458480357562927e-05, "loss": 3.3727, "step": 6906 }, { "epoch": 0.54, "learning_rate": 9.458401944640477e-05, "loss": 2.8627, "step": 6907 }, { "epoch": 0.54, "learning_rate": 9.458323531718028e-05, "loss": 2.9555, "step": 6908 }, { "epoch": 0.54, "learning_rate": 9.458245118795578e-05, "loss": 3.0517, "step": 6909 }, { "epoch": 0.54, "learning_rate": 9.458166705873129e-05, "loss": 3.2566, "step": 6910 }, { "epoch": 0.54, "learning_rate": 9.458088292950679e-05, "loss": 3.1734, "step": 6911 }, { "epoch": 0.54, "learning_rate": 9.458009880028229e-05, "loss": 3.0643, "step": 6912 }, { "epoch": 0.54, "learning_rate": 9.45793146710578e-05, "loss": 3.1629, "step": 6913 }, { "epoch": 0.54, "learning_rate": 9.45785305418333e-05, "loss": 3.3538, "step": 6914 }, { "epoch": 0.54, "learning_rate": 9.45777464126088e-05, "loss": 3.1445, "step": 6915 }, { "epoch": 0.54, "learning_rate": 9.457696228338431e-05, "loss": 2.8855, "step": 6916 }, { "epoch": 0.54, "learning_rate": 9.457617815415981e-05, "loss": 2.9375, "step": 6917 }, { "epoch": 0.54, "learning_rate": 9.457539402493531e-05, "loss": 3.0413, "step": 6918 }, { "epoch": 0.54, "learning_rate": 9.457460989571082e-05, "loss": 2.9757, "step": 6919 }, { "epoch": 0.54, "learning_rate": 9.457382576648632e-05, "loss": 2.9986, "step": 6920 }, { "epoch": 0.54, "learning_rate": 9.457304163726183e-05, "loss": 3.2327, "step": 6921 }, { "epoch": 0.54, "learning_rate": 9.457225750803733e-05, "loss": 3.0276, "step": 6922 }, { "epoch": 0.54, "learning_rate": 9.457147337881283e-05, "loss": 2.9008, "step": 6923 }, { "epoch": 0.54, "learning_rate": 9.457068924958834e-05, "loss": 3.0691, "step": 6924 }, { "epoch": 0.54, "learning_rate": 9.456990512036384e-05, "loss": 3.0108, "step": 6925 }, { "epoch": 0.54, "learning_rate": 9.456912099113934e-05, "loss": 2.8858, "step": 6926 }, { "epoch": 0.54, "learning_rate": 9.456833686191485e-05, "loss": 2.979, "step": 6927 }, { "epoch": 0.54, "learning_rate": 9.456755273269035e-05, "loss": 3.1269, "step": 6928 }, { "epoch": 0.54, "learning_rate": 9.456676860346585e-05, "loss": 3.1238, "step": 6929 }, { "epoch": 0.54, "learning_rate": 9.456598447424136e-05, "loss": 2.9733, "step": 6930 }, { "epoch": 0.54, "learning_rate": 9.456520034501686e-05, "loss": 2.9184, "step": 6931 }, { "epoch": 0.54, "learning_rate": 9.456441621579237e-05, "loss": 3.2019, "step": 6932 }, { "epoch": 0.54, "learning_rate": 9.456363208656787e-05, "loss": 3.0405, "step": 6933 }, { "epoch": 0.54, "learning_rate": 9.456284795734337e-05, "loss": 2.9892, "step": 6934 }, { "epoch": 0.54, "learning_rate": 9.456206382811888e-05, "loss": 2.9897, "step": 6935 }, { "epoch": 0.54, "learning_rate": 9.456127969889438e-05, "loss": 3.1353, "step": 6936 }, { "epoch": 0.54, "learning_rate": 9.456049556966988e-05, "loss": 3.022, "step": 6937 }, { "epoch": 0.54, "learning_rate": 9.455971144044539e-05, "loss": 2.9107, "step": 6938 }, { "epoch": 0.54, "learning_rate": 9.455892731122089e-05, "loss": 3.1653, "step": 6939 }, { "epoch": 0.54, "learning_rate": 9.45581431819964e-05, "loss": 3.012, "step": 6940 }, { "epoch": 0.54, "learning_rate": 9.45573590527719e-05, "loss": 3.0433, "step": 6941 }, { "epoch": 0.54, "learning_rate": 9.45565749235474e-05, "loss": 2.9265, "step": 6942 }, { "epoch": 0.54, "learning_rate": 9.45557907943229e-05, "loss": 3.2601, "step": 6943 }, { "epoch": 0.54, "learning_rate": 9.455500666509841e-05, "loss": 2.778, "step": 6944 }, { "epoch": 0.54, "learning_rate": 9.455422253587391e-05, "loss": 3.1694, "step": 6945 }, { "epoch": 0.54, "learning_rate": 9.455343840664942e-05, "loss": 3.0237, "step": 6946 }, { "epoch": 0.54, "learning_rate": 9.455265427742492e-05, "loss": 3.1079, "step": 6947 }, { "epoch": 0.54, "learning_rate": 9.455187014820042e-05, "loss": 3.0848, "step": 6948 }, { "epoch": 0.54, "learning_rate": 9.455108601897593e-05, "loss": 2.9832, "step": 6949 }, { "epoch": 0.54, "learning_rate": 9.455030188975143e-05, "loss": 2.9661, "step": 6950 }, { "epoch": 0.55, "learning_rate": 9.454951776052694e-05, "loss": 2.9202, "step": 6951 }, { "epoch": 0.55, "learning_rate": 9.454873363130244e-05, "loss": 3.4117, "step": 6952 }, { "epoch": 0.55, "learning_rate": 9.454794950207794e-05, "loss": 3.2292, "step": 6953 }, { "epoch": 0.55, "learning_rate": 9.454716537285345e-05, "loss": 2.8699, "step": 6954 }, { "epoch": 0.55, "learning_rate": 9.454638124362895e-05, "loss": 3.2675, "step": 6955 }, { "epoch": 0.55, "learning_rate": 9.454559711440445e-05, "loss": 3.1679, "step": 6956 }, { "epoch": 0.55, "learning_rate": 9.454481298517996e-05, "loss": 2.9906, "step": 6957 }, { "epoch": 0.55, "learning_rate": 9.454402885595548e-05, "loss": 3.1245, "step": 6958 }, { "epoch": 0.55, "learning_rate": 9.454324472673098e-05, "loss": 2.874, "step": 6959 }, { "epoch": 0.55, "learning_rate": 9.454246059750648e-05, "loss": 3.0224, "step": 6960 }, { "epoch": 0.55, "learning_rate": 9.454167646828199e-05, "loss": 2.9316, "step": 6961 }, { "epoch": 0.55, "learning_rate": 9.454089233905749e-05, "loss": 3.219, "step": 6962 }, { "epoch": 0.55, "learning_rate": 9.4540108209833e-05, "loss": 3.1451, "step": 6963 }, { "epoch": 0.55, "learning_rate": 9.45393240806085e-05, "loss": 3.2007, "step": 6964 }, { "epoch": 0.55, "learning_rate": 9.4538539951384e-05, "loss": 3.1685, "step": 6965 }, { "epoch": 0.55, "learning_rate": 9.45377558221595e-05, "loss": 3.0271, "step": 6966 }, { "epoch": 0.55, "learning_rate": 9.4536971692935e-05, "loss": 3.0715, "step": 6967 }, { "epoch": 0.55, "learning_rate": 9.45361875637105e-05, "loss": 3.179, "step": 6968 }, { "epoch": 0.55, "learning_rate": 9.4535403434486e-05, "loss": 3.0439, "step": 6969 }, { "epoch": 0.55, "learning_rate": 9.45346193052615e-05, "loss": 3.2903, "step": 6970 }, { "epoch": 0.55, "learning_rate": 9.453383517603701e-05, "loss": 2.9043, "step": 6971 }, { "epoch": 0.55, "learning_rate": 9.453305104681251e-05, "loss": 3.1285, "step": 6972 }, { "epoch": 0.55, "learning_rate": 9.453226691758802e-05, "loss": 3.0351, "step": 6973 }, { "epoch": 0.55, "learning_rate": 9.453148278836352e-05, "loss": 3.1241, "step": 6974 }, { "epoch": 0.55, "learning_rate": 9.453069865913902e-05, "loss": 3.1724, "step": 6975 }, { "epoch": 0.55, "learning_rate": 9.452991452991453e-05, "loss": 3.2876, "step": 6976 }, { "epoch": 0.55, "learning_rate": 9.452913040069003e-05, "loss": 3.0085, "step": 6977 }, { "epoch": 0.55, "learning_rate": 9.452834627146554e-05, "loss": 3.1416, "step": 6978 }, { "epoch": 0.55, "learning_rate": 9.452756214224104e-05, "loss": 3.147, "step": 6979 }, { "epoch": 0.55, "learning_rate": 9.452677801301654e-05, "loss": 3.0571, "step": 6980 }, { "epoch": 0.55, "learning_rate": 9.452599388379205e-05, "loss": 3.1928, "step": 6981 }, { "epoch": 0.55, "learning_rate": 9.452520975456755e-05, "loss": 3.1871, "step": 6982 }, { "epoch": 0.55, "learning_rate": 9.452442562534307e-05, "loss": 3.1706, "step": 6983 }, { "epoch": 0.55, "learning_rate": 9.452364149611857e-05, "loss": 2.9407, "step": 6984 }, { "epoch": 0.55, "learning_rate": 9.452285736689407e-05, "loss": 2.8636, "step": 6985 }, { "epoch": 0.55, "learning_rate": 9.452207323766958e-05, "loss": 2.9446, "step": 6986 }, { "epoch": 0.55, "learning_rate": 9.452128910844508e-05, "loss": 3.0626, "step": 6987 }, { "epoch": 0.55, "learning_rate": 9.452050497922059e-05, "loss": 2.9753, "step": 6988 }, { "epoch": 0.55, "learning_rate": 9.451972084999609e-05, "loss": 2.8124, "step": 6989 }, { "epoch": 0.55, "learning_rate": 9.451893672077159e-05, "loss": 2.9435, "step": 6990 }, { "epoch": 0.55, "learning_rate": 9.45181525915471e-05, "loss": 3.0087, "step": 6991 }, { "epoch": 0.55, "learning_rate": 9.45173684623226e-05, "loss": 2.9973, "step": 6992 }, { "epoch": 0.55, "learning_rate": 9.45165843330981e-05, "loss": 3.3015, "step": 6993 }, { "epoch": 0.55, "learning_rate": 9.451580020387361e-05, "loss": 3.2367, "step": 6994 }, { "epoch": 0.55, "learning_rate": 9.451501607464911e-05, "loss": 3.0313, "step": 6995 }, { "epoch": 0.55, "learning_rate": 9.451423194542461e-05, "loss": 3.1167, "step": 6996 }, { "epoch": 0.55, "learning_rate": 9.451344781620012e-05, "loss": 3.2645, "step": 6997 }, { "epoch": 0.55, "learning_rate": 9.451266368697562e-05, "loss": 2.9585, "step": 6998 }, { "epoch": 0.55, "learning_rate": 9.451187955775113e-05, "loss": 3.3176, "step": 6999 }, { "epoch": 0.55, "learning_rate": 9.451109542852663e-05, "loss": 2.9347, "step": 7000 }, { "epoch": 0.55, "eval_loss": 2.955150604248047, "eval_runtime": 218.1815, "eval_samples_per_second": 51.938, "eval_steps_per_second": 3.25, "step": 7000 }, { "epoch": 0.55, "learning_rate": 9.451031129930213e-05, "loss": 2.9411, "step": 7001 }, { "epoch": 0.55, "learning_rate": 9.450952717007764e-05, "loss": 2.9442, "step": 7002 }, { "epoch": 0.55, "learning_rate": 9.450874304085313e-05, "loss": 3.1547, "step": 7003 }, { "epoch": 0.55, "learning_rate": 9.450795891162863e-05, "loss": 3.7114, "step": 7004 }, { "epoch": 0.55, "learning_rate": 9.450717478240413e-05, "loss": 3.0663, "step": 7005 }, { "epoch": 0.55, "learning_rate": 9.450639065317964e-05, "loss": 2.8808, "step": 7006 }, { "epoch": 0.55, "learning_rate": 9.450560652395514e-05, "loss": 3.3667, "step": 7007 }, { "epoch": 0.55, "learning_rate": 9.450482239473065e-05, "loss": 2.8979, "step": 7008 }, { "epoch": 0.55, "learning_rate": 9.450403826550616e-05, "loss": 2.8783, "step": 7009 }, { "epoch": 0.55, "learning_rate": 9.450325413628167e-05, "loss": 3.0649, "step": 7010 }, { "epoch": 0.55, "learning_rate": 9.450247000705717e-05, "loss": 3.1186, "step": 7011 }, { "epoch": 0.55, "learning_rate": 9.450168587783267e-05, "loss": 2.9753, "step": 7012 }, { "epoch": 0.55, "learning_rate": 9.450090174860818e-05, "loss": 3.3421, "step": 7013 }, { "epoch": 0.55, "learning_rate": 9.450011761938368e-05, "loss": 3.0847, "step": 7014 }, { "epoch": 0.55, "learning_rate": 9.449933349015918e-05, "loss": 3.1354, "step": 7015 }, { "epoch": 0.55, "learning_rate": 9.449854936093469e-05, "loss": 2.8981, "step": 7016 }, { "epoch": 0.55, "learning_rate": 9.449776523171019e-05, "loss": 3.0251, "step": 7017 }, { "epoch": 0.55, "learning_rate": 9.44969811024857e-05, "loss": 3.4027, "step": 7018 }, { "epoch": 0.55, "learning_rate": 9.44961969732612e-05, "loss": 3.0814, "step": 7019 }, { "epoch": 0.55, "learning_rate": 9.44954128440367e-05, "loss": 2.9215, "step": 7020 }, { "epoch": 0.55, "learning_rate": 9.449462871481221e-05, "loss": 3.6235, "step": 7021 }, { "epoch": 0.55, "learning_rate": 9.449384458558771e-05, "loss": 2.8476, "step": 7022 }, { "epoch": 0.55, "learning_rate": 9.449306045636321e-05, "loss": 2.8891, "step": 7023 }, { "epoch": 0.55, "learning_rate": 9.449227632713872e-05, "loss": 3.1012, "step": 7024 }, { "epoch": 0.55, "learning_rate": 9.449149219791422e-05, "loss": 3.0812, "step": 7025 }, { "epoch": 0.55, "learning_rate": 9.449070806868973e-05, "loss": 3.2271, "step": 7026 }, { "epoch": 0.55, "learning_rate": 9.448992393946523e-05, "loss": 2.8671, "step": 7027 }, { "epoch": 0.55, "learning_rate": 9.448913981024073e-05, "loss": 2.9693, "step": 7028 }, { "epoch": 0.55, "learning_rate": 9.448835568101624e-05, "loss": 2.8931, "step": 7029 }, { "epoch": 0.55, "learning_rate": 9.448757155179174e-05, "loss": 2.9903, "step": 7030 }, { "epoch": 0.55, "learning_rate": 9.448678742256724e-05, "loss": 3.0799, "step": 7031 }, { "epoch": 0.55, "learning_rate": 9.448600329334275e-05, "loss": 2.9839, "step": 7032 }, { "epoch": 0.55, "learning_rate": 9.448521916411825e-05, "loss": 3.1315, "step": 7033 }, { "epoch": 0.55, "learning_rate": 9.448443503489375e-05, "loss": 3.0308, "step": 7034 }, { "epoch": 0.55, "learning_rate": 9.448365090566926e-05, "loss": 3.1009, "step": 7035 }, { "epoch": 0.55, "learning_rate": 9.448286677644476e-05, "loss": 3.0341, "step": 7036 }, { "epoch": 0.55, "learning_rate": 9.448208264722027e-05, "loss": 3.0978, "step": 7037 }, { "epoch": 0.55, "learning_rate": 9.448129851799577e-05, "loss": 2.781, "step": 7038 }, { "epoch": 0.55, "learning_rate": 9.448051438877127e-05, "loss": 3.034, "step": 7039 }, { "epoch": 0.55, "learning_rate": 9.447973025954678e-05, "loss": 2.9395, "step": 7040 }, { "epoch": 0.55, "learning_rate": 9.447894613032228e-05, "loss": 2.9506, "step": 7041 }, { "epoch": 0.55, "learning_rate": 9.447816200109778e-05, "loss": 2.9452, "step": 7042 }, { "epoch": 0.55, "learning_rate": 9.447737787187329e-05, "loss": 2.8543, "step": 7043 }, { "epoch": 0.55, "learning_rate": 9.447659374264879e-05, "loss": 3.065, "step": 7044 }, { "epoch": 0.55, "learning_rate": 9.44758096134243e-05, "loss": 2.8657, "step": 7045 }, { "epoch": 0.55, "learning_rate": 9.44750254841998e-05, "loss": 2.8808, "step": 7046 }, { "epoch": 0.55, "learning_rate": 9.44742413549753e-05, "loss": 3.1644, "step": 7047 }, { "epoch": 0.55, "learning_rate": 9.44734572257508e-05, "loss": 3.0525, "step": 7048 }, { "epoch": 0.55, "learning_rate": 9.447267309652631e-05, "loss": 2.8321, "step": 7049 }, { "epoch": 0.55, "learning_rate": 9.447188896730181e-05, "loss": 3.1853, "step": 7050 }, { "epoch": 0.55, "learning_rate": 9.447110483807732e-05, "loss": 3.3633, "step": 7051 }, { "epoch": 0.55, "learning_rate": 9.447032070885282e-05, "loss": 2.7514, "step": 7052 }, { "epoch": 0.55, "learning_rate": 9.446953657962832e-05, "loss": 3.1143, "step": 7053 }, { "epoch": 0.55, "learning_rate": 9.446875245040383e-05, "loss": 2.857, "step": 7054 }, { "epoch": 0.55, "learning_rate": 9.446796832117933e-05, "loss": 3.4429, "step": 7055 }, { "epoch": 0.55, "learning_rate": 9.446718419195484e-05, "loss": 3.1844, "step": 7056 }, { "epoch": 0.55, "learning_rate": 9.446640006273034e-05, "loss": 2.9086, "step": 7057 }, { "epoch": 0.55, "learning_rate": 9.446561593350584e-05, "loss": 3.3941, "step": 7058 }, { "epoch": 0.55, "learning_rate": 9.446483180428135e-05, "loss": 3.1233, "step": 7059 }, { "epoch": 0.55, "learning_rate": 9.446404767505685e-05, "loss": 3.2233, "step": 7060 }, { "epoch": 0.55, "learning_rate": 9.446326354583235e-05, "loss": 3.1887, "step": 7061 }, { "epoch": 0.55, "learning_rate": 9.446247941660786e-05, "loss": 3.1619, "step": 7062 }, { "epoch": 0.55, "learning_rate": 9.446169528738336e-05, "loss": 3.0969, "step": 7063 }, { "epoch": 0.55, "learning_rate": 9.446091115815886e-05, "loss": 2.899, "step": 7064 }, { "epoch": 0.55, "learning_rate": 9.446012702893437e-05, "loss": 3.1408, "step": 7065 }, { "epoch": 0.55, "learning_rate": 9.445934289970989e-05, "loss": 2.8924, "step": 7066 }, { "epoch": 0.55, "learning_rate": 9.445855877048539e-05, "loss": 2.7481, "step": 7067 }, { "epoch": 0.55, "learning_rate": 9.44577746412609e-05, "loss": 3.0011, "step": 7068 }, { "epoch": 0.55, "learning_rate": 9.44569905120364e-05, "loss": 3.0778, "step": 7069 }, { "epoch": 0.55, "learning_rate": 9.44562063828119e-05, "loss": 3.1327, "step": 7070 }, { "epoch": 0.55, "learning_rate": 9.44554222535874e-05, "loss": 3.1417, "step": 7071 }, { "epoch": 0.55, "learning_rate": 9.445463812436291e-05, "loss": 3.1896, "step": 7072 }, { "epoch": 0.55, "learning_rate": 9.445385399513841e-05, "loss": 3.1745, "step": 7073 }, { "epoch": 0.55, "learning_rate": 9.445306986591392e-05, "loss": 3.1609, "step": 7074 }, { "epoch": 0.55, "learning_rate": 9.44522857366894e-05, "loss": 3.2143, "step": 7075 }, { "epoch": 0.55, "learning_rate": 9.445150160746491e-05, "loss": 2.9975, "step": 7076 }, { "epoch": 0.55, "learning_rate": 9.445071747824041e-05, "loss": 3.1606, "step": 7077 }, { "epoch": 0.56, "learning_rate": 9.444993334901592e-05, "loss": 3.0821, "step": 7078 }, { "epoch": 0.56, "learning_rate": 9.444914921979142e-05, "loss": 2.7317, "step": 7079 }, { "epoch": 0.56, "learning_rate": 9.444836509056692e-05, "loss": 3.136, "step": 7080 }, { "epoch": 0.56, "learning_rate": 9.444758096134243e-05, "loss": 3.3627, "step": 7081 }, { "epoch": 0.56, "learning_rate": 9.444679683211793e-05, "loss": 3.2768, "step": 7082 }, { "epoch": 0.56, "learning_rate": 9.444601270289343e-05, "loss": 2.6848, "step": 7083 }, { "epoch": 0.56, "learning_rate": 9.444522857366894e-05, "loss": 3.0223, "step": 7084 }, { "epoch": 0.56, "learning_rate": 9.444444444444444e-05, "loss": 3.379, "step": 7085 }, { "epoch": 0.56, "learning_rate": 9.444366031521995e-05, "loss": 2.9832, "step": 7086 }, { "epoch": 0.56, "learning_rate": 9.444287618599545e-05, "loss": 3.0539, "step": 7087 }, { "epoch": 0.56, "learning_rate": 9.444209205677095e-05, "loss": 3.0829, "step": 7088 }, { "epoch": 0.56, "learning_rate": 9.444130792754646e-05, "loss": 3.0178, "step": 7089 }, { "epoch": 0.56, "learning_rate": 9.444052379832196e-05, "loss": 3.0502, "step": 7090 }, { "epoch": 0.56, "learning_rate": 9.443973966909748e-05, "loss": 3.2286, "step": 7091 }, { "epoch": 0.56, "learning_rate": 9.443895553987298e-05, "loss": 3.0443, "step": 7092 }, { "epoch": 0.56, "learning_rate": 9.443817141064849e-05, "loss": 2.8342, "step": 7093 }, { "epoch": 0.56, "learning_rate": 9.443738728142399e-05, "loss": 3.0402, "step": 7094 }, { "epoch": 0.56, "learning_rate": 9.443660315219949e-05, "loss": 3.4218, "step": 7095 }, { "epoch": 0.56, "learning_rate": 9.4435819022975e-05, "loss": 3.0349, "step": 7096 }, { "epoch": 0.56, "learning_rate": 9.44350348937505e-05, "loss": 3.0304, "step": 7097 }, { "epoch": 0.56, "learning_rate": 9.4434250764526e-05, "loss": 2.9892, "step": 7098 }, { "epoch": 0.56, "learning_rate": 9.443346663530151e-05, "loss": 2.9281, "step": 7099 }, { "epoch": 0.56, "learning_rate": 9.443268250607701e-05, "loss": 3.1977, "step": 7100 }, { "epoch": 0.56, "learning_rate": 9.443189837685251e-05, "loss": 3.0797, "step": 7101 }, { "epoch": 0.56, "learning_rate": 9.443111424762802e-05, "loss": 3.2295, "step": 7102 }, { "epoch": 0.56, "learning_rate": 9.443033011840352e-05, "loss": 3.1151, "step": 7103 }, { "epoch": 0.56, "learning_rate": 9.442954598917903e-05, "loss": 3.0692, "step": 7104 }, { "epoch": 0.56, "learning_rate": 9.442876185995453e-05, "loss": 3.0393, "step": 7105 }, { "epoch": 0.56, "learning_rate": 9.442797773073003e-05, "loss": 2.937, "step": 7106 }, { "epoch": 0.56, "learning_rate": 9.442719360150554e-05, "loss": 3.0138, "step": 7107 }, { "epoch": 0.56, "learning_rate": 9.442640947228104e-05, "loss": 3.1055, "step": 7108 }, { "epoch": 0.56, "learning_rate": 9.442562534305654e-05, "loss": 3.5674, "step": 7109 }, { "epoch": 0.56, "learning_rate": 9.442484121383205e-05, "loss": 3.245, "step": 7110 }, { "epoch": 0.56, "learning_rate": 9.442405708460755e-05, "loss": 2.9332, "step": 7111 }, { "epoch": 0.56, "learning_rate": 9.442327295538304e-05, "loss": 3.0108, "step": 7112 }, { "epoch": 0.56, "learning_rate": 9.442248882615855e-05, "loss": 3.0161, "step": 7113 }, { "epoch": 0.56, "learning_rate": 9.442170469693405e-05, "loss": 2.8846, "step": 7114 }, { "epoch": 0.56, "learning_rate": 9.442092056770955e-05, "loss": 3.0797, "step": 7115 }, { "epoch": 0.56, "learning_rate": 9.442013643848506e-05, "loss": 3.1434, "step": 7116 }, { "epoch": 0.56, "learning_rate": 9.441935230926057e-05, "loss": 3.1529, "step": 7117 }, { "epoch": 0.56, "learning_rate": 9.441856818003608e-05, "loss": 3.1058, "step": 7118 }, { "epoch": 0.56, "learning_rate": 9.441778405081158e-05, "loss": 2.9194, "step": 7119 }, { "epoch": 0.56, "learning_rate": 9.441699992158708e-05, "loss": 3.2636, "step": 7120 }, { "epoch": 0.56, "learning_rate": 9.441621579236259e-05, "loss": 3.2232, "step": 7121 }, { "epoch": 0.56, "learning_rate": 9.441543166313809e-05, "loss": 3.1582, "step": 7122 }, { "epoch": 0.56, "learning_rate": 9.44146475339136e-05, "loss": 2.8087, "step": 7123 }, { "epoch": 0.56, "learning_rate": 9.44138634046891e-05, "loss": 3.2311, "step": 7124 }, { "epoch": 0.56, "learning_rate": 9.44130792754646e-05, "loss": 2.9704, "step": 7125 }, { "epoch": 0.56, "learning_rate": 9.44122951462401e-05, "loss": 3.2136, "step": 7126 }, { "epoch": 0.56, "learning_rate": 9.441151101701561e-05, "loss": 2.8708, "step": 7127 }, { "epoch": 0.56, "learning_rate": 9.441072688779111e-05, "loss": 3.1095, "step": 7128 }, { "epoch": 0.56, "learning_rate": 9.440994275856662e-05, "loss": 2.9904, "step": 7129 }, { "epoch": 0.56, "learning_rate": 9.440915862934212e-05, "loss": 3.1928, "step": 7130 }, { "epoch": 0.56, "learning_rate": 9.440837450011762e-05, "loss": 3.2236, "step": 7131 }, { "epoch": 0.56, "learning_rate": 9.440759037089313e-05, "loss": 3.1591, "step": 7132 }, { "epoch": 0.56, "learning_rate": 9.440680624166863e-05, "loss": 2.9663, "step": 7133 }, { "epoch": 0.56, "learning_rate": 9.440602211244414e-05, "loss": 2.9781, "step": 7134 }, { "epoch": 0.56, "learning_rate": 9.440523798321964e-05, "loss": 3.0204, "step": 7135 }, { "epoch": 0.56, "learning_rate": 9.440445385399514e-05, "loss": 3.075, "step": 7136 }, { "epoch": 0.56, "learning_rate": 9.440366972477065e-05, "loss": 2.9527, "step": 7137 }, { "epoch": 0.56, "learning_rate": 9.440288559554615e-05, "loss": 3.0599, "step": 7138 }, { "epoch": 0.56, "learning_rate": 9.440210146632165e-05, "loss": 3.1541, "step": 7139 }, { "epoch": 0.56, "learning_rate": 9.440131733709716e-05, "loss": 3.249, "step": 7140 }, { "epoch": 0.56, "learning_rate": 9.440053320787266e-05, "loss": 3.1962, "step": 7141 }, { "epoch": 0.56, "learning_rate": 9.439974907864817e-05, "loss": 2.8724, "step": 7142 }, { "epoch": 0.56, "learning_rate": 9.439896494942367e-05, "loss": 3.3297, "step": 7143 }, { "epoch": 0.56, "learning_rate": 9.439818082019917e-05, "loss": 3.1734, "step": 7144 }, { "epoch": 0.56, "learning_rate": 9.439739669097468e-05, "loss": 2.9091, "step": 7145 }, { "epoch": 0.56, "learning_rate": 9.439661256175018e-05, "loss": 2.8113, "step": 7146 }, { "epoch": 0.56, "learning_rate": 9.439582843252568e-05, "loss": 2.8568, "step": 7147 }, { "epoch": 0.56, "learning_rate": 9.439504430330119e-05, "loss": 3.2731, "step": 7148 }, { "epoch": 0.56, "learning_rate": 9.439426017407669e-05, "loss": 2.9901, "step": 7149 }, { "epoch": 0.56, "learning_rate": 9.43934760448522e-05, "loss": 3.3389, "step": 7150 }, { "epoch": 0.56, "learning_rate": 9.43926919156277e-05, "loss": 3.3555, "step": 7151 }, { "epoch": 0.56, "learning_rate": 9.43919077864032e-05, "loss": 3.0196, "step": 7152 }, { "epoch": 0.56, "learning_rate": 9.43911236571787e-05, "loss": 3.0023, "step": 7153 }, { "epoch": 0.56, "learning_rate": 9.439033952795421e-05, "loss": 2.8609, "step": 7154 }, { "epoch": 0.56, "learning_rate": 9.438955539872971e-05, "loss": 3.004, "step": 7155 }, { "epoch": 0.56, "learning_rate": 9.438877126950522e-05, "loss": 3.0122, "step": 7156 }, { "epoch": 0.56, "learning_rate": 9.438798714028072e-05, "loss": 3.1077, "step": 7157 }, { "epoch": 0.56, "learning_rate": 9.438720301105622e-05, "loss": 2.9053, "step": 7158 }, { "epoch": 0.56, "learning_rate": 9.438641888183173e-05, "loss": 3.0621, "step": 7159 }, { "epoch": 0.56, "learning_rate": 9.438563475260723e-05, "loss": 3.1292, "step": 7160 }, { "epoch": 0.56, "learning_rate": 9.438485062338274e-05, "loss": 3.0659, "step": 7161 }, { "epoch": 0.56, "learning_rate": 9.438406649415824e-05, "loss": 3.0135, "step": 7162 }, { "epoch": 0.56, "learning_rate": 9.438328236493374e-05, "loss": 3.1568, "step": 7163 }, { "epoch": 0.56, "learning_rate": 9.438249823570925e-05, "loss": 3.2687, "step": 7164 }, { "epoch": 0.56, "learning_rate": 9.438171410648475e-05, "loss": 3.2103, "step": 7165 }, { "epoch": 0.56, "learning_rate": 9.438092997726025e-05, "loss": 2.8581, "step": 7166 }, { "epoch": 0.56, "learning_rate": 9.438014584803576e-05, "loss": 3.1135, "step": 7167 }, { "epoch": 0.56, "learning_rate": 9.437936171881126e-05, "loss": 2.9841, "step": 7168 }, { "epoch": 0.56, "learning_rate": 9.437857758958676e-05, "loss": 3.1401, "step": 7169 }, { "epoch": 0.56, "learning_rate": 9.437779346036227e-05, "loss": 3.1676, "step": 7170 }, { "epoch": 0.56, "learning_rate": 9.437700933113777e-05, "loss": 3.0525, "step": 7171 }, { "epoch": 0.56, "learning_rate": 9.437622520191328e-05, "loss": 3.1396, "step": 7172 }, { "epoch": 0.56, "learning_rate": 9.437544107268878e-05, "loss": 3.0865, "step": 7173 }, { "epoch": 0.56, "learning_rate": 9.43746569434643e-05, "loss": 3.0271, "step": 7174 }, { "epoch": 0.56, "learning_rate": 9.43738728142398e-05, "loss": 3.0284, "step": 7175 }, { "epoch": 0.56, "learning_rate": 9.43730886850153e-05, "loss": 3.0352, "step": 7176 }, { "epoch": 0.56, "learning_rate": 9.437230455579081e-05, "loss": 3.0117, "step": 7177 }, { "epoch": 0.56, "learning_rate": 9.437152042656631e-05, "loss": 2.855, "step": 7178 }, { "epoch": 0.56, "learning_rate": 9.437073629734182e-05, "loss": 3.0979, "step": 7179 }, { "epoch": 0.56, "learning_rate": 9.436995216811732e-05, "loss": 3.1618, "step": 7180 }, { "epoch": 0.56, "learning_rate": 9.436916803889282e-05, "loss": 2.9345, "step": 7181 }, { "epoch": 0.56, "learning_rate": 9.436838390966833e-05, "loss": 3.2405, "step": 7182 }, { "epoch": 0.56, "learning_rate": 9.436759978044383e-05, "loss": 2.9414, "step": 7183 }, { "epoch": 0.56, "learning_rate": 9.436681565121932e-05, "loss": 2.9399, "step": 7184 }, { "epoch": 0.56, "learning_rate": 9.436603152199482e-05, "loss": 3.0478, "step": 7185 }, { "epoch": 0.56, "learning_rate": 9.436524739277033e-05, "loss": 2.956, "step": 7186 }, { "epoch": 0.56, "learning_rate": 9.436446326354583e-05, "loss": 2.9058, "step": 7187 }, { "epoch": 0.56, "learning_rate": 9.436367913432133e-05, "loss": 3.0957, "step": 7188 }, { "epoch": 0.56, "learning_rate": 9.436289500509684e-05, "loss": 3.1031, "step": 7189 }, { "epoch": 0.56, "learning_rate": 9.436211087587234e-05, "loss": 2.8537, "step": 7190 }, { "epoch": 0.56, "learning_rate": 9.436132674664785e-05, "loss": 3.017, "step": 7191 }, { "epoch": 0.56, "learning_rate": 9.436054261742335e-05, "loss": 3.031, "step": 7192 }, { "epoch": 0.56, "learning_rate": 9.435975848819885e-05, "loss": 2.8174, "step": 7193 }, { "epoch": 0.56, "learning_rate": 9.435897435897436e-05, "loss": 2.9313, "step": 7194 }, { "epoch": 0.56, "learning_rate": 9.435819022974986e-05, "loss": 2.9707, "step": 7195 }, { "epoch": 0.56, "learning_rate": 9.435740610052536e-05, "loss": 3.0363, "step": 7196 }, { "epoch": 0.56, "learning_rate": 9.435662197130087e-05, "loss": 3.086, "step": 7197 }, { "epoch": 0.56, "learning_rate": 9.435583784207637e-05, "loss": 2.9491, "step": 7198 }, { "epoch": 0.56, "learning_rate": 9.435505371285189e-05, "loss": 3.1427, "step": 7199 }, { "epoch": 0.56, "learning_rate": 9.435426958362739e-05, "loss": 2.9854, "step": 7200 }, { "epoch": 0.56, "learning_rate": 9.43534854544029e-05, "loss": 2.867, "step": 7201 }, { "epoch": 0.56, "learning_rate": 9.43527013251784e-05, "loss": 3.0731, "step": 7202 }, { "epoch": 0.56, "learning_rate": 9.43519171959539e-05, "loss": 3.0794, "step": 7203 }, { "epoch": 0.56, "learning_rate": 9.435113306672941e-05, "loss": 3.0568, "step": 7204 }, { "epoch": 0.56, "learning_rate": 9.435034893750491e-05, "loss": 2.923, "step": 7205 }, { "epoch": 0.57, "learning_rate": 9.434956480828041e-05, "loss": 2.7448, "step": 7206 }, { "epoch": 0.57, "learning_rate": 9.434878067905592e-05, "loss": 3.0938, "step": 7207 }, { "epoch": 0.57, "learning_rate": 9.434799654983142e-05, "loss": 3.1339, "step": 7208 }, { "epoch": 0.57, "learning_rate": 9.434721242060693e-05, "loss": 3.208, "step": 7209 }, { "epoch": 0.57, "learning_rate": 9.434642829138243e-05, "loss": 2.8986, "step": 7210 }, { "epoch": 0.57, "learning_rate": 9.434564416215793e-05, "loss": 3.0713, "step": 7211 }, { "epoch": 0.57, "learning_rate": 9.434486003293344e-05, "loss": 3.0401, "step": 7212 }, { "epoch": 0.57, "learning_rate": 9.434407590370894e-05, "loss": 3.3286, "step": 7213 }, { "epoch": 0.57, "learning_rate": 9.434329177448444e-05, "loss": 3.0412, "step": 7214 }, { "epoch": 0.57, "learning_rate": 9.434250764525995e-05, "loss": 3.3226, "step": 7215 }, { "epoch": 0.57, "learning_rate": 9.434172351603545e-05, "loss": 3.2268, "step": 7216 }, { "epoch": 0.57, "learning_rate": 9.434093938681095e-05, "loss": 3.0019, "step": 7217 }, { "epoch": 0.57, "learning_rate": 9.434015525758646e-05, "loss": 3.2945, "step": 7218 }, { "epoch": 0.57, "learning_rate": 9.433937112836196e-05, "loss": 3.3083, "step": 7219 }, { "epoch": 0.57, "learning_rate": 9.433858699913745e-05, "loss": 2.6855, "step": 7220 }, { "epoch": 0.57, "learning_rate": 9.433780286991296e-05, "loss": 3.3305, "step": 7221 }, { "epoch": 0.57, "learning_rate": 9.433701874068846e-05, "loss": 3.205, "step": 7222 }, { "epoch": 0.57, "learning_rate": 9.433623461146396e-05, "loss": 3.0328, "step": 7223 }, { "epoch": 0.57, "learning_rate": 9.433545048223947e-05, "loss": 3.0769, "step": 7224 }, { "epoch": 0.57, "learning_rate": 9.433466635301498e-05, "loss": 3.4884, "step": 7225 }, { "epoch": 0.57, "learning_rate": 9.433388222379049e-05, "loss": 3.0417, "step": 7226 }, { "epoch": 0.57, "learning_rate": 9.433309809456599e-05, "loss": 2.8317, "step": 7227 }, { "epoch": 0.57, "learning_rate": 9.43323139653415e-05, "loss": 3.0107, "step": 7228 }, { "epoch": 0.57, "learning_rate": 9.4331529836117e-05, "loss": 2.9983, "step": 7229 }, { "epoch": 0.57, "learning_rate": 9.43307457068925e-05, "loss": 2.5897, "step": 7230 }, { "epoch": 0.57, "learning_rate": 9.4329961577668e-05, "loss": 3.0853, "step": 7231 }, { "epoch": 0.57, "learning_rate": 9.432917744844351e-05, "loss": 2.9208, "step": 7232 }, { "epoch": 0.57, "learning_rate": 9.432839331921901e-05, "loss": 2.9779, "step": 7233 }, { "epoch": 0.57, "learning_rate": 9.432760918999452e-05, "loss": 2.8142, "step": 7234 }, { "epoch": 0.57, "learning_rate": 9.432682506077002e-05, "loss": 3.1504, "step": 7235 }, { "epoch": 0.57, "learning_rate": 9.432604093154552e-05, "loss": 2.9481, "step": 7236 }, { "epoch": 0.57, "learning_rate": 9.432525680232103e-05, "loss": 2.9678, "step": 7237 }, { "epoch": 0.57, "learning_rate": 9.432447267309653e-05, "loss": 2.9732, "step": 7238 }, { "epoch": 0.57, "learning_rate": 9.432368854387204e-05, "loss": 3.2491, "step": 7239 }, { "epoch": 0.57, "learning_rate": 9.432290441464754e-05, "loss": 3.0328, "step": 7240 }, { "epoch": 0.57, "learning_rate": 9.432212028542304e-05, "loss": 3.1392, "step": 7241 }, { "epoch": 0.57, "learning_rate": 9.432133615619855e-05, "loss": 3.3038, "step": 7242 }, { "epoch": 0.57, "learning_rate": 9.432055202697405e-05, "loss": 3.2417, "step": 7243 }, { "epoch": 0.57, "learning_rate": 9.431976789774955e-05, "loss": 3.0138, "step": 7244 }, { "epoch": 0.57, "learning_rate": 9.431898376852506e-05, "loss": 3.0752, "step": 7245 }, { "epoch": 0.57, "learning_rate": 9.431819963930056e-05, "loss": 2.6666, "step": 7246 }, { "epoch": 0.57, "learning_rate": 9.431741551007607e-05, "loss": 3.2334, "step": 7247 }, { "epoch": 0.57, "learning_rate": 9.431663138085157e-05, "loss": 2.9173, "step": 7248 }, { "epoch": 0.57, "learning_rate": 9.431584725162707e-05, "loss": 2.9965, "step": 7249 }, { "epoch": 0.57, "learning_rate": 9.431506312240258e-05, "loss": 3.3679, "step": 7250 }, { "epoch": 0.57, "learning_rate": 9.431427899317808e-05, "loss": 2.8411, "step": 7251 }, { "epoch": 0.57, "learning_rate": 9.431349486395358e-05, "loss": 3.0826, "step": 7252 }, { "epoch": 0.57, "learning_rate": 9.431271073472909e-05, "loss": 3.3194, "step": 7253 }, { "epoch": 0.57, "learning_rate": 9.431192660550459e-05, "loss": 3.5496, "step": 7254 }, { "epoch": 0.57, "learning_rate": 9.43111424762801e-05, "loss": 3.199, "step": 7255 }, { "epoch": 0.57, "learning_rate": 9.43103583470556e-05, "loss": 2.9475, "step": 7256 }, { "epoch": 0.57, "learning_rate": 9.43095742178311e-05, "loss": 3.0312, "step": 7257 }, { "epoch": 0.57, "learning_rate": 9.43087900886066e-05, "loss": 2.9687, "step": 7258 }, { "epoch": 0.57, "learning_rate": 9.430800595938211e-05, "loss": 3.0884, "step": 7259 }, { "epoch": 0.57, "learning_rate": 9.430722183015761e-05, "loss": 3.0692, "step": 7260 }, { "epoch": 0.57, "learning_rate": 9.430643770093312e-05, "loss": 2.9723, "step": 7261 }, { "epoch": 0.57, "learning_rate": 9.430565357170862e-05, "loss": 3.0232, "step": 7262 }, { "epoch": 0.57, "learning_rate": 9.430486944248412e-05, "loss": 2.9914, "step": 7263 }, { "epoch": 0.57, "learning_rate": 9.430408531325963e-05, "loss": 2.9656, "step": 7264 }, { "epoch": 0.57, "learning_rate": 9.430330118403513e-05, "loss": 3.112, "step": 7265 }, { "epoch": 0.57, "learning_rate": 9.430251705481063e-05, "loss": 3.4083, "step": 7266 }, { "epoch": 0.57, "learning_rate": 9.430173292558614e-05, "loss": 3.1378, "step": 7267 }, { "epoch": 0.57, "learning_rate": 9.430094879636164e-05, "loss": 3.118, "step": 7268 }, { "epoch": 0.57, "learning_rate": 9.430016466713715e-05, "loss": 3.398, "step": 7269 }, { "epoch": 0.57, "learning_rate": 9.429938053791265e-05, "loss": 2.9144, "step": 7270 }, { "epoch": 0.57, "learning_rate": 9.429859640868815e-05, "loss": 3.1396, "step": 7271 }, { "epoch": 0.57, "learning_rate": 9.429781227946366e-05, "loss": 2.9888, "step": 7272 }, { "epoch": 0.57, "learning_rate": 9.429702815023916e-05, "loss": 2.9662, "step": 7273 }, { "epoch": 0.57, "learning_rate": 9.429624402101466e-05, "loss": 3.0559, "step": 7274 }, { "epoch": 0.57, "learning_rate": 9.429545989179017e-05, "loss": 2.9568, "step": 7275 }, { "epoch": 0.57, "learning_rate": 9.429467576256567e-05, "loss": 2.9954, "step": 7276 }, { "epoch": 0.57, "learning_rate": 9.429389163334118e-05, "loss": 3.1816, "step": 7277 }, { "epoch": 0.57, "learning_rate": 9.429310750411668e-05, "loss": 2.9085, "step": 7278 }, { "epoch": 0.57, "learning_rate": 9.429232337489218e-05, "loss": 3.0248, "step": 7279 }, { "epoch": 0.57, "learning_rate": 9.429153924566769e-05, "loss": 2.9494, "step": 7280 }, { "epoch": 0.57, "learning_rate": 9.429075511644319e-05, "loss": 3.2584, "step": 7281 }, { "epoch": 0.57, "learning_rate": 9.428997098721871e-05, "loss": 3.0198, "step": 7282 }, { "epoch": 0.57, "learning_rate": 9.428918685799421e-05, "loss": 2.8055, "step": 7283 }, { "epoch": 0.57, "learning_rate": 9.428840272876971e-05, "loss": 2.9225, "step": 7284 }, { "epoch": 0.57, "learning_rate": 9.428761859954522e-05, "loss": 3.0056, "step": 7285 }, { "epoch": 0.57, "learning_rate": 9.428683447032072e-05, "loss": 2.9309, "step": 7286 }, { "epoch": 0.57, "learning_rate": 9.428605034109623e-05, "loss": 3.0172, "step": 7287 }, { "epoch": 0.57, "learning_rate": 9.428526621187173e-05, "loss": 3.0556, "step": 7288 }, { "epoch": 0.57, "learning_rate": 9.428448208264723e-05, "loss": 3.2671, "step": 7289 }, { "epoch": 0.57, "learning_rate": 9.428369795342274e-05, "loss": 2.9823, "step": 7290 }, { "epoch": 0.57, "learning_rate": 9.428291382419824e-05, "loss": 3.0943, "step": 7291 }, { "epoch": 0.57, "learning_rate": 9.428212969497373e-05, "loss": 2.9914, "step": 7292 }, { "epoch": 0.57, "learning_rate": 9.428134556574923e-05, "loss": 3.0297, "step": 7293 }, { "epoch": 0.57, "learning_rate": 9.428056143652474e-05, "loss": 3.0028, "step": 7294 }, { "epoch": 0.57, "learning_rate": 9.427977730730024e-05, "loss": 3.1687, "step": 7295 }, { "epoch": 0.57, "learning_rate": 9.427899317807575e-05, "loss": 2.974, "step": 7296 }, { "epoch": 0.57, "learning_rate": 9.427820904885125e-05, "loss": 2.9684, "step": 7297 }, { "epoch": 0.57, "learning_rate": 9.427742491962675e-05, "loss": 3.3325, "step": 7298 }, { "epoch": 0.57, "learning_rate": 9.427664079040226e-05, "loss": 3.2352, "step": 7299 }, { "epoch": 0.57, "learning_rate": 9.427585666117776e-05, "loss": 2.933, "step": 7300 }, { "epoch": 0.57, "learning_rate": 9.427507253195326e-05, "loss": 2.9827, "step": 7301 }, { "epoch": 0.57, "learning_rate": 9.427428840272877e-05, "loss": 2.9036, "step": 7302 }, { "epoch": 0.57, "learning_rate": 9.427350427350427e-05, "loss": 2.9657, "step": 7303 }, { "epoch": 0.57, "learning_rate": 9.427272014427977e-05, "loss": 3.1208, "step": 7304 }, { "epoch": 0.57, "learning_rate": 9.427193601505528e-05, "loss": 3.3325, "step": 7305 }, { "epoch": 0.57, "learning_rate": 9.427115188583078e-05, "loss": 3.0288, "step": 7306 }, { "epoch": 0.57, "learning_rate": 9.42703677566063e-05, "loss": 2.9987, "step": 7307 }, { "epoch": 0.57, "learning_rate": 9.42695836273818e-05, "loss": 2.8962, "step": 7308 }, { "epoch": 0.57, "learning_rate": 9.42687994981573e-05, "loss": 2.8198, "step": 7309 }, { "epoch": 0.57, "learning_rate": 9.426801536893281e-05, "loss": 3.0112, "step": 7310 }, { "epoch": 0.57, "learning_rate": 9.426723123970831e-05, "loss": 3.0645, "step": 7311 }, { "epoch": 0.57, "learning_rate": 9.426644711048382e-05, "loss": 3.2081, "step": 7312 }, { "epoch": 0.57, "learning_rate": 9.426566298125932e-05, "loss": 2.9278, "step": 7313 }, { "epoch": 0.57, "learning_rate": 9.426487885203483e-05, "loss": 3.0232, "step": 7314 }, { "epoch": 0.57, "learning_rate": 9.426409472281033e-05, "loss": 2.9865, "step": 7315 }, { "epoch": 0.57, "learning_rate": 9.426331059358583e-05, "loss": 3.0856, "step": 7316 }, { "epoch": 0.57, "learning_rate": 9.426252646436134e-05, "loss": 2.883, "step": 7317 }, { "epoch": 0.57, "learning_rate": 9.426174233513684e-05, "loss": 3.0958, "step": 7318 }, { "epoch": 0.57, "learning_rate": 9.426095820591234e-05, "loss": 3.0396, "step": 7319 }, { "epoch": 0.57, "learning_rate": 9.426017407668785e-05, "loss": 3.1074, "step": 7320 }, { "epoch": 0.57, "learning_rate": 9.425938994746335e-05, "loss": 3.0517, "step": 7321 }, { "epoch": 0.57, "learning_rate": 9.425860581823885e-05, "loss": 2.9002, "step": 7322 }, { "epoch": 0.57, "learning_rate": 9.425782168901436e-05, "loss": 3.1311, "step": 7323 }, { "epoch": 0.57, "learning_rate": 9.425703755978986e-05, "loss": 3.1599, "step": 7324 }, { "epoch": 0.57, "learning_rate": 9.425625343056537e-05, "loss": 3.054, "step": 7325 }, { "epoch": 0.57, "learning_rate": 9.425546930134087e-05, "loss": 3.4144, "step": 7326 }, { "epoch": 0.57, "learning_rate": 9.425468517211637e-05, "loss": 2.927, "step": 7327 }, { "epoch": 0.57, "learning_rate": 9.425390104289186e-05, "loss": 2.998, "step": 7328 }, { "epoch": 0.57, "learning_rate": 9.425311691366737e-05, "loss": 2.664, "step": 7329 }, { "epoch": 0.57, "learning_rate": 9.425233278444287e-05, "loss": 2.9861, "step": 7330 }, { "epoch": 0.57, "learning_rate": 9.425154865521837e-05, "loss": 2.9908, "step": 7331 }, { "epoch": 0.57, "learning_rate": 9.425076452599388e-05, "loss": 3.0802, "step": 7332 }, { "epoch": 0.58, "learning_rate": 9.42499803967694e-05, "loss": 3.0529, "step": 7333 }, { "epoch": 0.58, "learning_rate": 9.42491962675449e-05, "loss": 3.0129, "step": 7334 }, { "epoch": 0.58, "learning_rate": 9.42484121383204e-05, "loss": 3.5335, "step": 7335 }, { "epoch": 0.58, "learning_rate": 9.42476280090959e-05, "loss": 2.9641, "step": 7336 }, { "epoch": 0.58, "learning_rate": 9.424684387987141e-05, "loss": 3.1077, "step": 7337 }, { "epoch": 0.58, "learning_rate": 9.424605975064691e-05, "loss": 3.1538, "step": 7338 }, { "epoch": 0.58, "learning_rate": 9.424527562142242e-05, "loss": 3.1939, "step": 7339 }, { "epoch": 0.58, "learning_rate": 9.424449149219792e-05, "loss": 2.9935, "step": 7340 }, { "epoch": 0.58, "learning_rate": 9.424370736297342e-05, "loss": 3.0378, "step": 7341 }, { "epoch": 0.58, "learning_rate": 9.424292323374893e-05, "loss": 3.2025, "step": 7342 }, { "epoch": 0.58, "learning_rate": 9.424213910452443e-05, "loss": 3.0396, "step": 7343 }, { "epoch": 0.58, "learning_rate": 9.424135497529994e-05, "loss": 3.2112, "step": 7344 }, { "epoch": 0.58, "learning_rate": 9.424057084607544e-05, "loss": 3.1288, "step": 7345 }, { "epoch": 0.58, "learning_rate": 9.423978671685094e-05, "loss": 2.8799, "step": 7346 }, { "epoch": 0.58, "learning_rate": 9.423900258762645e-05, "loss": 3.062, "step": 7347 }, { "epoch": 0.58, "learning_rate": 9.423821845840195e-05, "loss": 3.0402, "step": 7348 }, { "epoch": 0.58, "learning_rate": 9.423743432917745e-05, "loss": 3.1461, "step": 7349 }, { "epoch": 0.58, "learning_rate": 9.423665019995296e-05, "loss": 2.9279, "step": 7350 }, { "epoch": 0.58, "learning_rate": 9.423586607072846e-05, "loss": 2.9216, "step": 7351 }, { "epoch": 0.58, "learning_rate": 9.423508194150396e-05, "loss": 3.0706, "step": 7352 }, { "epoch": 0.58, "learning_rate": 9.423429781227947e-05, "loss": 2.9675, "step": 7353 }, { "epoch": 0.58, "learning_rate": 9.423351368305497e-05, "loss": 2.9043, "step": 7354 }, { "epoch": 0.58, "learning_rate": 9.423272955383048e-05, "loss": 2.8734, "step": 7355 }, { "epoch": 0.58, "learning_rate": 9.423194542460598e-05, "loss": 2.9916, "step": 7356 }, { "epoch": 0.58, "learning_rate": 9.423116129538148e-05, "loss": 2.871, "step": 7357 }, { "epoch": 0.58, "learning_rate": 9.423037716615699e-05, "loss": 3.1468, "step": 7358 }, { "epoch": 0.58, "learning_rate": 9.422959303693249e-05, "loss": 2.921, "step": 7359 }, { "epoch": 0.58, "learning_rate": 9.4228808907708e-05, "loss": 3.3367, "step": 7360 }, { "epoch": 0.58, "learning_rate": 9.42280247784835e-05, "loss": 2.9128, "step": 7361 }, { "epoch": 0.58, "learning_rate": 9.4227240649259e-05, "loss": 3.1028, "step": 7362 }, { "epoch": 0.58, "learning_rate": 9.42264565200345e-05, "loss": 2.9772, "step": 7363 }, { "epoch": 0.58, "learning_rate": 9.422567239081001e-05, "loss": 3.0732, "step": 7364 }, { "epoch": 0.58, "learning_rate": 9.422488826158551e-05, "loss": 3.2947, "step": 7365 }, { "epoch": 0.58, "learning_rate": 9.422410413236102e-05, "loss": 2.9993, "step": 7366 }, { "epoch": 0.58, "learning_rate": 9.422332000313652e-05, "loss": 3.0943, "step": 7367 }, { "epoch": 0.58, "learning_rate": 9.422253587391202e-05, "loss": 3.2304, "step": 7368 }, { "epoch": 0.58, "learning_rate": 9.422175174468753e-05, "loss": 3.3126, "step": 7369 }, { "epoch": 0.58, "learning_rate": 9.422096761546303e-05, "loss": 3.0557, "step": 7370 }, { "epoch": 0.58, "learning_rate": 9.422018348623853e-05, "loss": 3.0167, "step": 7371 }, { "epoch": 0.58, "learning_rate": 9.421939935701404e-05, "loss": 3.0249, "step": 7372 }, { "epoch": 0.58, "learning_rate": 9.421861522778954e-05, "loss": 3.0629, "step": 7373 }, { "epoch": 0.58, "learning_rate": 9.421783109856505e-05, "loss": 2.9731, "step": 7374 }, { "epoch": 0.58, "learning_rate": 9.421704696934055e-05, "loss": 3.1388, "step": 7375 }, { "epoch": 0.58, "learning_rate": 9.421626284011605e-05, "loss": 3.0815, "step": 7376 }, { "epoch": 0.58, "learning_rate": 9.421547871089156e-05, "loss": 2.7526, "step": 7377 }, { "epoch": 0.58, "learning_rate": 9.421469458166706e-05, "loss": 2.8734, "step": 7378 }, { "epoch": 0.58, "learning_rate": 9.421391045244256e-05, "loss": 3.2517, "step": 7379 }, { "epoch": 0.58, "learning_rate": 9.421312632321807e-05, "loss": 2.9542, "step": 7380 }, { "epoch": 0.58, "learning_rate": 9.421234219399357e-05, "loss": 2.8985, "step": 7381 }, { "epoch": 0.58, "learning_rate": 9.421155806476908e-05, "loss": 3.148, "step": 7382 }, { "epoch": 0.58, "learning_rate": 9.421077393554458e-05, "loss": 2.9048, "step": 7383 }, { "epoch": 0.58, "learning_rate": 9.420998980632008e-05, "loss": 2.9596, "step": 7384 }, { "epoch": 0.58, "learning_rate": 9.420920567709559e-05, "loss": 3.0405, "step": 7385 }, { "epoch": 0.58, "learning_rate": 9.420842154787109e-05, "loss": 3.1177, "step": 7386 }, { "epoch": 0.58, "learning_rate": 9.42076374186466e-05, "loss": 3.0347, "step": 7387 }, { "epoch": 0.58, "learning_rate": 9.42068532894221e-05, "loss": 3.1858, "step": 7388 }, { "epoch": 0.58, "learning_rate": 9.42060691601976e-05, "loss": 2.993, "step": 7389 }, { "epoch": 0.58, "learning_rate": 9.420528503097312e-05, "loss": 3.1236, "step": 7390 }, { "epoch": 0.58, "learning_rate": 9.420450090174862e-05, "loss": 3.0459, "step": 7391 }, { "epoch": 0.58, "learning_rate": 9.420371677252413e-05, "loss": 3.0842, "step": 7392 }, { "epoch": 0.58, "learning_rate": 9.420293264329963e-05, "loss": 2.9519, "step": 7393 }, { "epoch": 0.58, "learning_rate": 9.420214851407513e-05, "loss": 2.9748, "step": 7394 }, { "epoch": 0.58, "learning_rate": 9.420136438485064e-05, "loss": 3.1449, "step": 7395 }, { "epoch": 0.58, "learning_rate": 9.420058025562614e-05, "loss": 2.8156, "step": 7396 }, { "epoch": 0.58, "learning_rate": 9.419979612640164e-05, "loss": 3.3608, "step": 7397 }, { "epoch": 0.58, "learning_rate": 9.419901199717715e-05, "loss": 2.9384, "step": 7398 }, { "epoch": 0.58, "learning_rate": 9.419822786795265e-05, "loss": 3.3083, "step": 7399 }, { "epoch": 0.58, "learning_rate": 9.419744373872814e-05, "loss": 2.9094, "step": 7400 }, { "epoch": 0.58, "learning_rate": 9.419665960950365e-05, "loss": 2.8508, "step": 7401 }, { "epoch": 0.58, "learning_rate": 9.419587548027915e-05, "loss": 3.0132, "step": 7402 }, { "epoch": 0.58, "learning_rate": 9.419509135105465e-05, "loss": 3.051, "step": 7403 }, { "epoch": 0.58, "learning_rate": 9.419430722183016e-05, "loss": 2.9808, "step": 7404 }, { "epoch": 0.58, "learning_rate": 9.419352309260566e-05, "loss": 2.9303, "step": 7405 }, { "epoch": 0.58, "learning_rate": 9.419273896338116e-05, "loss": 3.2939, "step": 7406 }, { "epoch": 0.58, "learning_rate": 9.419195483415667e-05, "loss": 3.1333, "step": 7407 }, { "epoch": 0.58, "learning_rate": 9.419117070493217e-05, "loss": 3.0805, "step": 7408 }, { "epoch": 0.58, "learning_rate": 9.419038657570767e-05, "loss": 3.036, "step": 7409 }, { "epoch": 0.58, "learning_rate": 9.418960244648318e-05, "loss": 2.9173, "step": 7410 }, { "epoch": 0.58, "learning_rate": 9.418881831725868e-05, "loss": 3.1409, "step": 7411 }, { "epoch": 0.58, "learning_rate": 9.418803418803419e-05, "loss": 3.1967, "step": 7412 }, { "epoch": 0.58, "learning_rate": 9.418725005880969e-05, "loss": 3.1945, "step": 7413 }, { "epoch": 0.58, "learning_rate": 9.418646592958519e-05, "loss": 3.0392, "step": 7414 }, { "epoch": 0.58, "learning_rate": 9.418568180036071e-05, "loss": 3.1773, "step": 7415 }, { "epoch": 0.58, "learning_rate": 9.418489767113621e-05, "loss": 2.8328, "step": 7416 }, { "epoch": 0.58, "learning_rate": 9.418411354191172e-05, "loss": 2.8847, "step": 7417 }, { "epoch": 0.58, "learning_rate": 9.418332941268722e-05, "loss": 3.0493, "step": 7418 }, { "epoch": 0.58, "learning_rate": 9.418254528346272e-05, "loss": 2.7428, "step": 7419 }, { "epoch": 0.58, "learning_rate": 9.418176115423823e-05, "loss": 3.1278, "step": 7420 }, { "epoch": 0.58, "learning_rate": 9.418097702501373e-05, "loss": 3.1081, "step": 7421 }, { "epoch": 0.58, "learning_rate": 9.418019289578924e-05, "loss": 3.017, "step": 7422 }, { "epoch": 0.58, "learning_rate": 9.417940876656474e-05, "loss": 3.3234, "step": 7423 }, { "epoch": 0.58, "learning_rate": 9.417862463734024e-05, "loss": 3.0616, "step": 7424 }, { "epoch": 0.58, "learning_rate": 9.417784050811575e-05, "loss": 3.2167, "step": 7425 }, { "epoch": 0.58, "learning_rate": 9.417705637889125e-05, "loss": 3.0334, "step": 7426 }, { "epoch": 0.58, "learning_rate": 9.417627224966675e-05, "loss": 2.9719, "step": 7427 }, { "epoch": 0.58, "learning_rate": 9.417548812044226e-05, "loss": 2.9987, "step": 7428 }, { "epoch": 0.58, "learning_rate": 9.417470399121776e-05, "loss": 2.9564, "step": 7429 }, { "epoch": 0.58, "learning_rate": 9.417391986199327e-05, "loss": 3.0643, "step": 7430 }, { "epoch": 0.58, "learning_rate": 9.417313573276877e-05, "loss": 3.0876, "step": 7431 }, { "epoch": 0.58, "learning_rate": 9.417235160354427e-05, "loss": 3.0032, "step": 7432 }, { "epoch": 0.58, "learning_rate": 9.417156747431978e-05, "loss": 3.0455, "step": 7433 }, { "epoch": 0.58, "learning_rate": 9.417078334509528e-05, "loss": 2.994, "step": 7434 }, { "epoch": 0.58, "learning_rate": 9.416999921587078e-05, "loss": 3.2592, "step": 7435 }, { "epoch": 0.58, "learning_rate": 9.416921508664627e-05, "loss": 2.8972, "step": 7436 }, { "epoch": 0.58, "learning_rate": 9.416843095742178e-05, "loss": 3.2799, "step": 7437 }, { "epoch": 0.58, "learning_rate": 9.416764682819728e-05, "loss": 2.9842, "step": 7438 }, { "epoch": 0.58, "learning_rate": 9.416686269897278e-05, "loss": 3.1534, "step": 7439 }, { "epoch": 0.58, "learning_rate": 9.416607856974829e-05, "loss": 2.9227, "step": 7440 }, { "epoch": 0.58, "learning_rate": 9.41652944405238e-05, "loss": 3.0937, "step": 7441 }, { "epoch": 0.58, "learning_rate": 9.416451031129931e-05, "loss": 2.9289, "step": 7442 }, { "epoch": 0.58, "learning_rate": 9.416372618207481e-05, "loss": 3.0386, "step": 7443 }, { "epoch": 0.58, "learning_rate": 9.416294205285032e-05, "loss": 3.0212, "step": 7444 }, { "epoch": 0.58, "learning_rate": 9.416215792362582e-05, "loss": 3.0782, "step": 7445 }, { "epoch": 0.58, "learning_rate": 9.416137379440132e-05, "loss": 3.008, "step": 7446 }, { "epoch": 0.58, "learning_rate": 9.416058966517683e-05, "loss": 3.0894, "step": 7447 }, { "epoch": 0.58, "learning_rate": 9.415980553595233e-05, "loss": 2.9178, "step": 7448 }, { "epoch": 0.58, "learning_rate": 9.415902140672784e-05, "loss": 2.9011, "step": 7449 }, { "epoch": 0.58, "learning_rate": 9.415823727750334e-05, "loss": 3.3392, "step": 7450 }, { "epoch": 0.58, "learning_rate": 9.415745314827884e-05, "loss": 3.0772, "step": 7451 }, { "epoch": 0.58, "learning_rate": 9.415666901905435e-05, "loss": 2.9878, "step": 7452 }, { "epoch": 0.58, "learning_rate": 9.415588488982985e-05, "loss": 3.1088, "step": 7453 }, { "epoch": 0.58, "learning_rate": 9.415510076060535e-05, "loss": 3.0858, "step": 7454 }, { "epoch": 0.58, "learning_rate": 9.415431663138086e-05, "loss": 3.1119, "step": 7455 }, { "epoch": 0.58, "learning_rate": 9.415353250215636e-05, "loss": 3.0059, "step": 7456 }, { "epoch": 0.58, "learning_rate": 9.415274837293186e-05, "loss": 2.9256, "step": 7457 }, { "epoch": 0.58, "learning_rate": 9.415196424370737e-05, "loss": 3.0213, "step": 7458 }, { "epoch": 0.58, "learning_rate": 9.415118011448287e-05, "loss": 2.7026, "step": 7459 }, { "epoch": 0.58, "learning_rate": 9.415039598525838e-05, "loss": 3.1752, "step": 7460 }, { "epoch": 0.59, "learning_rate": 9.414961185603388e-05, "loss": 2.9141, "step": 7461 }, { "epoch": 0.59, "learning_rate": 9.414882772680938e-05, "loss": 2.8206, "step": 7462 }, { "epoch": 0.59, "learning_rate": 9.414804359758489e-05, "loss": 3.1587, "step": 7463 }, { "epoch": 0.59, "learning_rate": 9.414725946836039e-05, "loss": 3.019, "step": 7464 }, { "epoch": 0.59, "learning_rate": 9.41464753391359e-05, "loss": 2.7971, "step": 7465 }, { "epoch": 0.59, "learning_rate": 9.41456912099114e-05, "loss": 2.9922, "step": 7466 }, { "epoch": 0.59, "learning_rate": 9.41449070806869e-05, "loss": 3.153, "step": 7467 }, { "epoch": 0.59, "learning_rate": 9.41441229514624e-05, "loss": 2.9314, "step": 7468 }, { "epoch": 0.59, "learning_rate": 9.414333882223791e-05, "loss": 3.1103, "step": 7469 }, { "epoch": 0.59, "learning_rate": 9.414255469301341e-05, "loss": 3.0461, "step": 7470 }, { "epoch": 0.59, "learning_rate": 9.414177056378892e-05, "loss": 2.9967, "step": 7471 }, { "epoch": 0.59, "learning_rate": 9.414098643456442e-05, "loss": 2.8753, "step": 7472 }, { "epoch": 0.59, "learning_rate": 9.414020230533992e-05, "loss": 3.0577, "step": 7473 }, { "epoch": 0.59, "learning_rate": 9.413941817611543e-05, "loss": 2.9876, "step": 7474 }, { "epoch": 0.59, "learning_rate": 9.413863404689093e-05, "loss": 3.1425, "step": 7475 }, { "epoch": 0.59, "learning_rate": 9.413784991766643e-05, "loss": 3.0135, "step": 7476 }, { "epoch": 0.59, "learning_rate": 9.413706578844194e-05, "loss": 2.8696, "step": 7477 }, { "epoch": 0.59, "learning_rate": 9.413628165921744e-05, "loss": 2.958, "step": 7478 }, { "epoch": 0.59, "learning_rate": 9.413549752999295e-05, "loss": 3.3994, "step": 7479 }, { "epoch": 0.59, "learning_rate": 9.413471340076845e-05, "loss": 2.8777, "step": 7480 }, { "epoch": 0.59, "learning_rate": 9.413392927154395e-05, "loss": 3.1264, "step": 7481 }, { "epoch": 0.59, "learning_rate": 9.413314514231946e-05, "loss": 3.12, "step": 7482 }, { "epoch": 0.59, "learning_rate": 9.413236101309496e-05, "loss": 2.8833, "step": 7483 }, { "epoch": 0.59, "learning_rate": 9.413157688387046e-05, "loss": 2.7941, "step": 7484 }, { "epoch": 0.59, "learning_rate": 9.413079275464597e-05, "loss": 3.2235, "step": 7485 }, { "epoch": 0.59, "learning_rate": 9.413000862542147e-05, "loss": 3.1236, "step": 7486 }, { "epoch": 0.59, "learning_rate": 9.412922449619697e-05, "loss": 2.8516, "step": 7487 }, { "epoch": 0.59, "learning_rate": 9.412844036697248e-05, "loss": 3.3879, "step": 7488 }, { "epoch": 0.59, "learning_rate": 9.412765623774798e-05, "loss": 3.205, "step": 7489 }, { "epoch": 0.59, "learning_rate": 9.412687210852349e-05, "loss": 3.2303, "step": 7490 }, { "epoch": 0.59, "learning_rate": 9.412608797929899e-05, "loss": 3.0455, "step": 7491 }, { "epoch": 0.59, "learning_rate": 9.41253038500745e-05, "loss": 2.9384, "step": 7492 }, { "epoch": 0.59, "learning_rate": 9.412451972085e-05, "loss": 2.9685, "step": 7493 }, { "epoch": 0.59, "learning_rate": 9.41237355916255e-05, "loss": 3.0381, "step": 7494 }, { "epoch": 0.59, "learning_rate": 9.4122951462401e-05, "loss": 3.0358, "step": 7495 }, { "epoch": 0.59, "learning_rate": 9.412216733317651e-05, "loss": 2.9481, "step": 7496 }, { "epoch": 0.59, "learning_rate": 9.412138320395201e-05, "loss": 3.6005, "step": 7497 }, { "epoch": 0.59, "learning_rate": 9.412059907472753e-05, "loss": 3.3681, "step": 7498 }, { "epoch": 0.59, "learning_rate": 9.411981494550303e-05, "loss": 2.7531, "step": 7499 }, { "epoch": 0.59, "learning_rate": 9.411903081627854e-05, "loss": 3.245, "step": 7500 }, { "epoch": 0.59, "learning_rate": 9.411824668705404e-05, "loss": 3.0823, "step": 7501 }, { "epoch": 0.59, "learning_rate": 9.411746255782954e-05, "loss": 3.1449, "step": 7502 }, { "epoch": 0.59, "learning_rate": 9.411667842860505e-05, "loss": 2.8917, "step": 7503 }, { "epoch": 0.59, "learning_rate": 9.411589429938055e-05, "loss": 2.8797, "step": 7504 }, { "epoch": 0.59, "learning_rate": 9.411511017015605e-05, "loss": 3.1712, "step": 7505 }, { "epoch": 0.59, "learning_rate": 9.411432604093156e-05, "loss": 2.7968, "step": 7506 }, { "epoch": 0.59, "learning_rate": 9.411354191170706e-05, "loss": 3.3016, "step": 7507 }, { "epoch": 0.59, "learning_rate": 9.411275778248255e-05, "loss": 3.0424, "step": 7508 }, { "epoch": 0.59, "learning_rate": 9.411197365325806e-05, "loss": 3.0736, "step": 7509 }, { "epoch": 0.59, "learning_rate": 9.411118952403356e-05, "loss": 3.0505, "step": 7510 }, { "epoch": 0.59, "learning_rate": 9.411040539480906e-05, "loss": 3.1161, "step": 7511 }, { "epoch": 0.59, "learning_rate": 9.410962126558457e-05, "loss": 2.8877, "step": 7512 }, { "epoch": 0.59, "learning_rate": 9.410883713636007e-05, "loss": 2.9228, "step": 7513 }, { "epoch": 0.59, "learning_rate": 9.410805300713557e-05, "loss": 2.9638, "step": 7514 }, { "epoch": 0.59, "learning_rate": 9.410726887791108e-05, "loss": 3.0014, "step": 7515 }, { "epoch": 0.59, "learning_rate": 9.410648474868658e-05, "loss": 2.8812, "step": 7516 }, { "epoch": 0.59, "learning_rate": 9.410570061946209e-05, "loss": 2.9786, "step": 7517 }, { "epoch": 0.59, "learning_rate": 9.410491649023759e-05, "loss": 3.3497, "step": 7518 }, { "epoch": 0.59, "learning_rate": 9.410413236101309e-05, "loss": 2.8361, "step": 7519 }, { "epoch": 0.59, "learning_rate": 9.41033482317886e-05, "loss": 3.0637, "step": 7520 }, { "epoch": 0.59, "learning_rate": 9.41025641025641e-05, "loss": 2.9991, "step": 7521 }, { "epoch": 0.59, "learning_rate": 9.41017799733396e-05, "loss": 3.0923, "step": 7522 }, { "epoch": 0.59, "learning_rate": 9.410099584411512e-05, "loss": 3.1786, "step": 7523 }, { "epoch": 0.59, "learning_rate": 9.410021171489062e-05, "loss": 3.208, "step": 7524 }, { "epoch": 0.59, "learning_rate": 9.409942758566613e-05, "loss": 3.1571, "step": 7525 }, { "epoch": 0.59, "learning_rate": 9.409864345644163e-05, "loss": 2.928, "step": 7526 }, { "epoch": 0.59, "learning_rate": 9.409785932721714e-05, "loss": 3.1724, "step": 7527 }, { "epoch": 0.59, "learning_rate": 9.409707519799264e-05, "loss": 3.0468, "step": 7528 }, { "epoch": 0.59, "learning_rate": 9.409629106876814e-05, "loss": 3.1492, "step": 7529 }, { "epoch": 0.59, "learning_rate": 9.409550693954365e-05, "loss": 2.993, "step": 7530 }, { "epoch": 0.59, "learning_rate": 9.409472281031915e-05, "loss": 3.3136, "step": 7531 }, { "epoch": 0.59, "learning_rate": 9.409393868109465e-05, "loss": 3.166, "step": 7532 }, { "epoch": 0.59, "learning_rate": 9.409315455187016e-05, "loss": 3.0583, "step": 7533 }, { "epoch": 0.59, "learning_rate": 9.409237042264566e-05, "loss": 3.0834, "step": 7534 }, { "epoch": 0.59, "learning_rate": 9.409158629342116e-05, "loss": 3.098, "step": 7535 }, { "epoch": 0.59, "learning_rate": 9.409080216419667e-05, "loss": 2.9413, "step": 7536 }, { "epoch": 0.59, "learning_rate": 9.409001803497217e-05, "loss": 3.2514, "step": 7537 }, { "epoch": 0.59, "learning_rate": 9.408923390574768e-05, "loss": 2.9183, "step": 7538 }, { "epoch": 0.59, "learning_rate": 9.408844977652318e-05, "loss": 2.8857, "step": 7539 }, { "epoch": 0.59, "learning_rate": 9.408766564729868e-05, "loss": 2.8601, "step": 7540 }, { "epoch": 0.59, "learning_rate": 9.408688151807419e-05, "loss": 2.9841, "step": 7541 }, { "epoch": 0.59, "learning_rate": 9.408609738884969e-05, "loss": 2.9906, "step": 7542 }, { "epoch": 0.59, "learning_rate": 9.40853132596252e-05, "loss": 2.9635, "step": 7543 }, { "epoch": 0.59, "learning_rate": 9.408452913040068e-05, "loss": 2.9565, "step": 7544 }, { "epoch": 0.59, "learning_rate": 9.408374500117619e-05, "loss": 2.9438, "step": 7545 }, { "epoch": 0.59, "learning_rate": 9.408296087195169e-05, "loss": 2.944, "step": 7546 }, { "epoch": 0.59, "learning_rate": 9.40821767427272e-05, "loss": 2.9541, "step": 7547 }, { "epoch": 0.59, "learning_rate": 9.40813926135027e-05, "loss": 3.2104, "step": 7548 }, { "epoch": 0.59, "learning_rate": 9.408060848427822e-05, "loss": 3.0816, "step": 7549 }, { "epoch": 0.59, "learning_rate": 9.407982435505372e-05, "loss": 2.9932, "step": 7550 }, { "epoch": 0.59, "learning_rate": 9.407904022582922e-05, "loss": 2.9442, "step": 7551 }, { "epoch": 0.59, "learning_rate": 9.407825609660473e-05, "loss": 2.9475, "step": 7552 }, { "epoch": 0.59, "learning_rate": 9.407747196738023e-05, "loss": 3.1906, "step": 7553 }, { "epoch": 0.59, "learning_rate": 9.407668783815573e-05, "loss": 2.8525, "step": 7554 }, { "epoch": 0.59, "learning_rate": 9.407590370893124e-05, "loss": 3.1383, "step": 7555 }, { "epoch": 0.59, "learning_rate": 9.407511957970674e-05, "loss": 3.2634, "step": 7556 }, { "epoch": 0.59, "learning_rate": 9.407433545048225e-05, "loss": 3.0887, "step": 7557 }, { "epoch": 0.59, "learning_rate": 9.407355132125775e-05, "loss": 2.8401, "step": 7558 }, { "epoch": 0.59, "learning_rate": 9.407276719203325e-05, "loss": 2.9128, "step": 7559 }, { "epoch": 0.59, "learning_rate": 9.407198306280876e-05, "loss": 3.0457, "step": 7560 }, { "epoch": 0.59, "learning_rate": 9.407119893358426e-05, "loss": 2.9769, "step": 7561 }, { "epoch": 0.59, "learning_rate": 9.407041480435976e-05, "loss": 2.931, "step": 7562 }, { "epoch": 0.59, "learning_rate": 9.406963067513527e-05, "loss": 3.2016, "step": 7563 }, { "epoch": 0.59, "learning_rate": 9.406884654591077e-05, "loss": 3.1741, "step": 7564 }, { "epoch": 0.59, "learning_rate": 9.406806241668628e-05, "loss": 3.2536, "step": 7565 }, { "epoch": 0.59, "learning_rate": 9.406727828746178e-05, "loss": 3.0303, "step": 7566 }, { "epoch": 0.59, "learning_rate": 9.406649415823728e-05, "loss": 2.9424, "step": 7567 }, { "epoch": 0.59, "learning_rate": 9.406571002901279e-05, "loss": 3.2163, "step": 7568 }, { "epoch": 0.59, "learning_rate": 9.406492589978829e-05, "loss": 2.995, "step": 7569 }, { "epoch": 0.59, "learning_rate": 9.40641417705638e-05, "loss": 3.1511, "step": 7570 }, { "epoch": 0.59, "learning_rate": 9.40633576413393e-05, "loss": 3.1173, "step": 7571 }, { "epoch": 0.59, "learning_rate": 9.40625735121148e-05, "loss": 2.9911, "step": 7572 }, { "epoch": 0.59, "learning_rate": 9.40617893828903e-05, "loss": 3.058, "step": 7573 }, { "epoch": 0.59, "learning_rate": 9.406100525366581e-05, "loss": 3.1316, "step": 7574 }, { "epoch": 0.59, "learning_rate": 9.406022112444131e-05, "loss": 2.9736, "step": 7575 }, { "epoch": 0.59, "learning_rate": 9.405943699521682e-05, "loss": 2.9632, "step": 7576 }, { "epoch": 0.59, "learning_rate": 9.405865286599232e-05, "loss": 2.9882, "step": 7577 }, { "epoch": 0.59, "learning_rate": 9.405786873676782e-05, "loss": 2.9535, "step": 7578 }, { "epoch": 0.59, "learning_rate": 9.405708460754333e-05, "loss": 2.8593, "step": 7579 }, { "epoch": 0.59, "learning_rate": 9.405630047831883e-05, "loss": 2.9226, "step": 7580 }, { "epoch": 0.59, "learning_rate": 9.405551634909433e-05, "loss": 3.196, "step": 7581 }, { "epoch": 0.59, "learning_rate": 9.405473221986984e-05, "loss": 3.0637, "step": 7582 }, { "epoch": 0.59, "learning_rate": 9.405394809064534e-05, "loss": 2.8694, "step": 7583 }, { "epoch": 0.59, "learning_rate": 9.405316396142085e-05, "loss": 3.0234, "step": 7584 }, { "epoch": 0.59, "learning_rate": 9.405237983219635e-05, "loss": 3.3185, "step": 7585 }, { "epoch": 0.59, "learning_rate": 9.405159570297185e-05, "loss": 3.0303, "step": 7586 }, { "epoch": 0.59, "learning_rate": 9.405081157374736e-05, "loss": 3.0286, "step": 7587 }, { "epoch": 0.59, "learning_rate": 9.405002744452286e-05, "loss": 2.9175, "step": 7588 }, { "epoch": 0.6, "learning_rate": 9.404924331529836e-05, "loss": 3.0678, "step": 7589 }, { "epoch": 0.6, "learning_rate": 9.404845918607387e-05, "loss": 2.9356, "step": 7590 }, { "epoch": 0.6, "learning_rate": 9.404767505684937e-05, "loss": 3.2318, "step": 7591 }, { "epoch": 0.6, "learning_rate": 9.404689092762487e-05, "loss": 3.0515, "step": 7592 }, { "epoch": 0.6, "learning_rate": 9.404610679840038e-05, "loss": 2.9543, "step": 7593 }, { "epoch": 0.6, "learning_rate": 9.404532266917588e-05, "loss": 3.5054, "step": 7594 }, { "epoch": 0.6, "learning_rate": 9.404453853995139e-05, "loss": 3.1758, "step": 7595 }, { "epoch": 0.6, "learning_rate": 9.404375441072689e-05, "loss": 3.0521, "step": 7596 }, { "epoch": 0.6, "learning_rate": 9.404297028150239e-05, "loss": 3.1528, "step": 7597 }, { "epoch": 0.6, "learning_rate": 9.40421861522779e-05, "loss": 2.9462, "step": 7598 }, { "epoch": 0.6, "learning_rate": 9.40414020230534e-05, "loss": 3.2883, "step": 7599 }, { "epoch": 0.6, "learning_rate": 9.40406178938289e-05, "loss": 3.1076, "step": 7600 }, { "epoch": 0.6, "learning_rate": 9.403983376460441e-05, "loss": 3.3624, "step": 7601 }, { "epoch": 0.6, "learning_rate": 9.403904963537991e-05, "loss": 3.0444, "step": 7602 }, { "epoch": 0.6, "learning_rate": 9.403826550615542e-05, "loss": 2.9293, "step": 7603 }, { "epoch": 0.6, "learning_rate": 9.403748137693092e-05, "loss": 3.0268, "step": 7604 }, { "epoch": 0.6, "learning_rate": 9.403669724770642e-05, "loss": 2.8952, "step": 7605 }, { "epoch": 0.6, "learning_rate": 9.403591311848194e-05, "loss": 3.2754, "step": 7606 }, { "epoch": 0.6, "learning_rate": 9.403512898925744e-05, "loss": 3.1187, "step": 7607 }, { "epoch": 0.6, "learning_rate": 9.403434486003295e-05, "loss": 3.3278, "step": 7608 }, { "epoch": 0.6, "learning_rate": 9.403356073080845e-05, "loss": 3.1483, "step": 7609 }, { "epoch": 0.6, "learning_rate": 9.403277660158395e-05, "loss": 2.9409, "step": 7610 }, { "epoch": 0.6, "learning_rate": 9.403199247235946e-05, "loss": 2.9416, "step": 7611 }, { "epoch": 0.6, "learning_rate": 9.403120834313496e-05, "loss": 2.8876, "step": 7612 }, { "epoch": 0.6, "learning_rate": 9.403042421391047e-05, "loss": 3.0531, "step": 7613 }, { "epoch": 0.6, "learning_rate": 9.402964008468597e-05, "loss": 3.1742, "step": 7614 }, { "epoch": 0.6, "learning_rate": 9.402885595546147e-05, "loss": 3.0522, "step": 7615 }, { "epoch": 0.6, "learning_rate": 9.402807182623696e-05, "loss": 2.9319, "step": 7616 }, { "epoch": 0.6, "learning_rate": 9.402728769701247e-05, "loss": 2.9477, "step": 7617 }, { "epoch": 0.6, "learning_rate": 9.402650356778797e-05, "loss": 2.9847, "step": 7618 }, { "epoch": 0.6, "learning_rate": 9.402571943856347e-05, "loss": 3.1237, "step": 7619 }, { "epoch": 0.6, "learning_rate": 9.402493530933898e-05, "loss": 3.0492, "step": 7620 }, { "epoch": 0.6, "learning_rate": 9.402415118011448e-05, "loss": 3.0185, "step": 7621 }, { "epoch": 0.6, "learning_rate": 9.402336705088998e-05, "loss": 3.006, "step": 7622 }, { "epoch": 0.6, "learning_rate": 9.402258292166549e-05, "loss": 3.0067, "step": 7623 }, { "epoch": 0.6, "learning_rate": 9.402179879244099e-05, "loss": 2.8825, "step": 7624 }, { "epoch": 0.6, "learning_rate": 9.40210146632165e-05, "loss": 2.7613, "step": 7625 }, { "epoch": 0.6, "learning_rate": 9.4020230533992e-05, "loss": 3.0224, "step": 7626 }, { "epoch": 0.6, "learning_rate": 9.40194464047675e-05, "loss": 3.1687, "step": 7627 }, { "epoch": 0.6, "learning_rate": 9.401866227554301e-05, "loss": 3.0695, "step": 7628 }, { "epoch": 0.6, "learning_rate": 9.401787814631851e-05, "loss": 3.014, "step": 7629 }, { "epoch": 0.6, "learning_rate": 9.401709401709401e-05, "loss": 3.0851, "step": 7630 }, { "epoch": 0.6, "learning_rate": 9.401630988786953e-05, "loss": 2.7903, "step": 7631 }, { "epoch": 0.6, "learning_rate": 9.401552575864504e-05, "loss": 3.0111, "step": 7632 }, { "epoch": 0.6, "learning_rate": 9.401474162942054e-05, "loss": 3.0324, "step": 7633 }, { "epoch": 0.6, "learning_rate": 9.401395750019604e-05, "loss": 3.0978, "step": 7634 }, { "epoch": 0.6, "learning_rate": 9.401317337097155e-05, "loss": 3.5082, "step": 7635 }, { "epoch": 0.6, "learning_rate": 9.401238924174705e-05, "loss": 2.8485, "step": 7636 }, { "epoch": 0.6, "learning_rate": 9.401160511252255e-05, "loss": 2.9869, "step": 7637 }, { "epoch": 0.6, "learning_rate": 9.401082098329806e-05, "loss": 3.2105, "step": 7638 }, { "epoch": 0.6, "learning_rate": 9.401003685407356e-05, "loss": 2.9629, "step": 7639 }, { "epoch": 0.6, "learning_rate": 9.400925272484906e-05, "loss": 3.0539, "step": 7640 }, { "epoch": 0.6, "learning_rate": 9.400846859562457e-05, "loss": 3.0186, "step": 7641 }, { "epoch": 0.6, "learning_rate": 9.400768446640007e-05, "loss": 3.1273, "step": 7642 }, { "epoch": 0.6, "learning_rate": 9.400690033717558e-05, "loss": 2.9859, "step": 7643 }, { "epoch": 0.6, "learning_rate": 9.400611620795108e-05, "loss": 3.1342, "step": 7644 }, { "epoch": 0.6, "learning_rate": 9.400533207872658e-05, "loss": 3.0087, "step": 7645 }, { "epoch": 0.6, "learning_rate": 9.400454794950209e-05, "loss": 2.7906, "step": 7646 }, { "epoch": 0.6, "learning_rate": 9.400376382027759e-05, "loss": 3.4147, "step": 7647 }, { "epoch": 0.6, "learning_rate": 9.40029796910531e-05, "loss": 2.9668, "step": 7648 }, { "epoch": 0.6, "learning_rate": 9.40021955618286e-05, "loss": 2.8041, "step": 7649 }, { "epoch": 0.6, "learning_rate": 9.40014114326041e-05, "loss": 3.2658, "step": 7650 }, { "epoch": 0.6, "learning_rate": 9.40006273033796e-05, "loss": 3.0351, "step": 7651 }, { "epoch": 0.6, "learning_rate": 9.39998431741551e-05, "loss": 2.9326, "step": 7652 }, { "epoch": 0.6, "learning_rate": 9.39990590449306e-05, "loss": 2.9849, "step": 7653 }, { "epoch": 0.6, "learning_rate": 9.39982749157061e-05, "loss": 3.2141, "step": 7654 }, { "epoch": 0.6, "learning_rate": 9.39974907864816e-05, "loss": 2.8491, "step": 7655 }, { "epoch": 0.6, "learning_rate": 9.399670665725711e-05, "loss": 2.9945, "step": 7656 }, { "epoch": 0.6, "learning_rate": 9.399592252803263e-05, "loss": 2.7265, "step": 7657 }, { "epoch": 0.6, "learning_rate": 9.399513839880813e-05, "loss": 3.0044, "step": 7658 }, { "epoch": 0.6, "learning_rate": 9.399435426958363e-05, "loss": 3.1244, "step": 7659 }, { "epoch": 0.6, "learning_rate": 9.399357014035914e-05, "loss": 2.8158, "step": 7660 }, { "epoch": 0.6, "learning_rate": 9.399278601113464e-05, "loss": 2.9981, "step": 7661 }, { "epoch": 0.6, "learning_rate": 9.399200188191015e-05, "loss": 3.1509, "step": 7662 }, { "epoch": 0.6, "learning_rate": 9.399121775268565e-05, "loss": 3.3746, "step": 7663 }, { "epoch": 0.6, "learning_rate": 9.399043362346115e-05, "loss": 2.9476, "step": 7664 }, { "epoch": 0.6, "learning_rate": 9.398964949423666e-05, "loss": 3.008, "step": 7665 }, { "epoch": 0.6, "learning_rate": 9.398886536501216e-05, "loss": 3.0759, "step": 7666 }, { "epoch": 0.6, "learning_rate": 9.398808123578766e-05, "loss": 3.0974, "step": 7667 }, { "epoch": 0.6, "learning_rate": 9.398729710656317e-05, "loss": 3.0166, "step": 7668 }, { "epoch": 0.6, "learning_rate": 9.398651297733867e-05, "loss": 3.0243, "step": 7669 }, { "epoch": 0.6, "learning_rate": 9.398572884811417e-05, "loss": 3.022, "step": 7670 }, { "epoch": 0.6, "learning_rate": 9.398494471888968e-05, "loss": 3.2135, "step": 7671 }, { "epoch": 0.6, "learning_rate": 9.398416058966518e-05, "loss": 3.2479, "step": 7672 }, { "epoch": 0.6, "learning_rate": 9.398337646044069e-05, "loss": 3.0292, "step": 7673 }, { "epoch": 0.6, "learning_rate": 9.398259233121619e-05, "loss": 2.9797, "step": 7674 }, { "epoch": 0.6, "learning_rate": 9.39818082019917e-05, "loss": 2.9175, "step": 7675 }, { "epoch": 0.6, "learning_rate": 9.39810240727672e-05, "loss": 2.9512, "step": 7676 }, { "epoch": 0.6, "learning_rate": 9.39802399435427e-05, "loss": 3.038, "step": 7677 }, { "epoch": 0.6, "learning_rate": 9.39794558143182e-05, "loss": 3.2386, "step": 7678 }, { "epoch": 0.6, "learning_rate": 9.397867168509371e-05, "loss": 2.8384, "step": 7679 }, { "epoch": 0.6, "learning_rate": 9.397788755586921e-05, "loss": 3.011, "step": 7680 }, { "epoch": 0.6, "learning_rate": 9.397710342664472e-05, "loss": 2.9205, "step": 7681 }, { "epoch": 0.6, "learning_rate": 9.397631929742022e-05, "loss": 2.9481, "step": 7682 }, { "epoch": 0.6, "learning_rate": 9.397553516819572e-05, "loss": 3.1022, "step": 7683 }, { "epoch": 0.6, "learning_rate": 9.397475103897123e-05, "loss": 2.8943, "step": 7684 }, { "epoch": 0.6, "learning_rate": 9.397396690974673e-05, "loss": 2.9806, "step": 7685 }, { "epoch": 0.6, "learning_rate": 9.397318278052223e-05, "loss": 3.039, "step": 7686 }, { "epoch": 0.6, "learning_rate": 9.397239865129774e-05, "loss": 2.8936, "step": 7687 }, { "epoch": 0.6, "learning_rate": 9.397161452207324e-05, "loss": 2.954, "step": 7688 }, { "epoch": 0.6, "learning_rate": 9.397083039284874e-05, "loss": 3.388, "step": 7689 }, { "epoch": 0.6, "learning_rate": 9.397004626362425e-05, "loss": 3.4778, "step": 7690 }, { "epoch": 0.6, "learning_rate": 9.396926213439975e-05, "loss": 3.2258, "step": 7691 }, { "epoch": 0.6, "learning_rate": 9.396847800517526e-05, "loss": 3.1031, "step": 7692 }, { "epoch": 0.6, "learning_rate": 9.396769387595076e-05, "loss": 3.1163, "step": 7693 }, { "epoch": 0.6, "learning_rate": 9.396690974672626e-05, "loss": 3.2522, "step": 7694 }, { "epoch": 0.6, "learning_rate": 9.396612561750177e-05, "loss": 3.0238, "step": 7695 }, { "epoch": 0.6, "learning_rate": 9.396534148827727e-05, "loss": 3.0678, "step": 7696 }, { "epoch": 0.6, "learning_rate": 9.396455735905277e-05, "loss": 2.9924, "step": 7697 }, { "epoch": 0.6, "learning_rate": 9.396377322982828e-05, "loss": 3.1826, "step": 7698 }, { "epoch": 0.6, "learning_rate": 9.396298910060378e-05, "loss": 2.9641, "step": 7699 }, { "epoch": 0.6, "learning_rate": 9.396220497137929e-05, "loss": 3.0883, "step": 7700 }, { "epoch": 0.6, "learning_rate": 9.396142084215479e-05, "loss": 3.2514, "step": 7701 }, { "epoch": 0.6, "learning_rate": 9.396063671293029e-05, "loss": 2.8823, "step": 7702 }, { "epoch": 0.6, "learning_rate": 9.39598525837058e-05, "loss": 2.9957, "step": 7703 }, { "epoch": 0.6, "learning_rate": 9.39590684544813e-05, "loss": 2.8454, "step": 7704 }, { "epoch": 0.6, "learning_rate": 9.39582843252568e-05, "loss": 3.0359, "step": 7705 }, { "epoch": 0.6, "learning_rate": 9.395750019603231e-05, "loss": 2.8923, "step": 7706 }, { "epoch": 0.6, "learning_rate": 9.395671606680781e-05, "loss": 3.0366, "step": 7707 }, { "epoch": 0.6, "learning_rate": 9.395593193758331e-05, "loss": 3.2468, "step": 7708 }, { "epoch": 0.6, "learning_rate": 9.395514780835882e-05, "loss": 3.0932, "step": 7709 }, { "epoch": 0.6, "learning_rate": 9.395436367913432e-05, "loss": 3.1273, "step": 7710 }, { "epoch": 0.6, "learning_rate": 9.395357954990983e-05, "loss": 3.1337, "step": 7711 }, { "epoch": 0.6, "learning_rate": 9.395279542068533e-05, "loss": 2.9773, "step": 7712 }, { "epoch": 0.6, "learning_rate": 9.395201129146083e-05, "loss": 2.9459, "step": 7713 }, { "epoch": 0.6, "learning_rate": 9.395122716223635e-05, "loss": 3.041, "step": 7714 }, { "epoch": 0.6, "learning_rate": 9.395044303301185e-05, "loss": 2.9153, "step": 7715 }, { "epoch": 0.61, "learning_rate": 9.394965890378736e-05, "loss": 2.9, "step": 7716 }, { "epoch": 0.61, "learning_rate": 9.394887477456286e-05, "loss": 2.9333, "step": 7717 }, { "epoch": 0.61, "learning_rate": 9.394809064533837e-05, "loss": 2.9076, "step": 7718 }, { "epoch": 0.61, "learning_rate": 9.394730651611387e-05, "loss": 3.0708, "step": 7719 }, { "epoch": 0.61, "learning_rate": 9.394652238688937e-05, "loss": 3.3642, "step": 7720 }, { "epoch": 0.61, "learning_rate": 9.394573825766488e-05, "loss": 3.2202, "step": 7721 }, { "epoch": 0.61, "learning_rate": 9.394495412844038e-05, "loss": 3.0617, "step": 7722 }, { "epoch": 0.61, "learning_rate": 9.394416999921588e-05, "loss": 2.9104, "step": 7723 }, { "epoch": 0.61, "learning_rate": 9.394338586999137e-05, "loss": 2.9899, "step": 7724 }, { "epoch": 0.61, "learning_rate": 9.394260174076688e-05, "loss": 3.1032, "step": 7725 }, { "epoch": 0.61, "learning_rate": 9.394181761154238e-05, "loss": 2.9918, "step": 7726 }, { "epoch": 0.61, "learning_rate": 9.394103348231788e-05, "loss": 3.1068, "step": 7727 }, { "epoch": 0.61, "learning_rate": 9.394024935309339e-05, "loss": 3.1397, "step": 7728 }, { "epoch": 0.61, "learning_rate": 9.393946522386889e-05, "loss": 3.0382, "step": 7729 }, { "epoch": 0.61, "learning_rate": 9.39386810946444e-05, "loss": 2.6621, "step": 7730 }, { "epoch": 0.61, "learning_rate": 9.39378969654199e-05, "loss": 3.4745, "step": 7731 }, { "epoch": 0.61, "learning_rate": 9.39371128361954e-05, "loss": 2.9782, "step": 7732 }, { "epoch": 0.61, "learning_rate": 9.39363287069709e-05, "loss": 3.1246, "step": 7733 }, { "epoch": 0.61, "learning_rate": 9.393554457774641e-05, "loss": 3.0267, "step": 7734 }, { "epoch": 0.61, "learning_rate": 9.393476044852191e-05, "loss": 3.056, "step": 7735 }, { "epoch": 0.61, "learning_rate": 9.393397631929742e-05, "loss": 3.1767, "step": 7736 }, { "epoch": 0.61, "learning_rate": 9.393319219007292e-05, "loss": 3.2884, "step": 7737 }, { "epoch": 0.61, "learning_rate": 9.393240806084843e-05, "loss": 3.5044, "step": 7738 }, { "epoch": 0.61, "learning_rate": 9.393162393162393e-05, "loss": 3.1816, "step": 7739 }, { "epoch": 0.61, "learning_rate": 9.393083980239945e-05, "loss": 3.0897, "step": 7740 }, { "epoch": 0.61, "learning_rate": 9.393005567317495e-05, "loss": 2.9244, "step": 7741 }, { "epoch": 0.61, "learning_rate": 9.392927154395045e-05, "loss": 2.9614, "step": 7742 }, { "epoch": 0.61, "learning_rate": 9.392848741472596e-05, "loss": 3.0324, "step": 7743 }, { "epoch": 0.61, "learning_rate": 9.392770328550146e-05, "loss": 2.9159, "step": 7744 }, { "epoch": 0.61, "learning_rate": 9.392691915627696e-05, "loss": 2.9285, "step": 7745 }, { "epoch": 0.61, "learning_rate": 9.392613502705247e-05, "loss": 2.7715, "step": 7746 }, { "epoch": 0.61, "learning_rate": 9.392535089782797e-05, "loss": 3.1306, "step": 7747 }, { "epoch": 0.61, "learning_rate": 9.392456676860348e-05, "loss": 3.3616, "step": 7748 }, { "epoch": 0.61, "learning_rate": 9.392378263937898e-05, "loss": 3.4296, "step": 7749 }, { "epoch": 0.61, "learning_rate": 9.392299851015448e-05, "loss": 2.9713, "step": 7750 }, { "epoch": 0.61, "learning_rate": 9.392221438092999e-05, "loss": 3.1071, "step": 7751 }, { "epoch": 0.61, "learning_rate": 9.392143025170549e-05, "loss": 2.9725, "step": 7752 }, { "epoch": 0.61, "learning_rate": 9.3920646122481e-05, "loss": 3.0056, "step": 7753 }, { "epoch": 0.61, "learning_rate": 9.39198619932565e-05, "loss": 2.8283, "step": 7754 }, { "epoch": 0.61, "learning_rate": 9.3919077864032e-05, "loss": 2.9362, "step": 7755 }, { "epoch": 0.61, "learning_rate": 9.39182937348075e-05, "loss": 2.9829, "step": 7756 }, { "epoch": 0.61, "learning_rate": 9.391750960558301e-05, "loss": 2.9908, "step": 7757 }, { "epoch": 0.61, "learning_rate": 9.391672547635851e-05, "loss": 3.0287, "step": 7758 }, { "epoch": 0.61, "learning_rate": 9.391594134713402e-05, "loss": 3.0936, "step": 7759 }, { "epoch": 0.61, "learning_rate": 9.39151572179095e-05, "loss": 3.2296, "step": 7760 }, { "epoch": 0.61, "learning_rate": 9.391437308868501e-05, "loss": 3.1632, "step": 7761 }, { "epoch": 0.61, "learning_rate": 9.391358895946051e-05, "loss": 2.9972, "step": 7762 }, { "epoch": 0.61, "learning_rate": 9.391280483023602e-05, "loss": 3.3376, "step": 7763 }, { "epoch": 0.61, "learning_rate": 9.391202070101152e-05, "loss": 2.9813, "step": 7764 }, { "epoch": 0.61, "learning_rate": 9.391123657178704e-05, "loss": 2.8792, "step": 7765 }, { "epoch": 0.61, "learning_rate": 9.391045244256254e-05, "loss": 3.0299, "step": 7766 }, { "epoch": 0.61, "learning_rate": 9.390966831333805e-05, "loss": 3.0956, "step": 7767 }, { "epoch": 0.61, "learning_rate": 9.390888418411355e-05, "loss": 2.8311, "step": 7768 }, { "epoch": 0.61, "learning_rate": 9.390810005488905e-05, "loss": 2.9147, "step": 7769 }, { "epoch": 0.61, "learning_rate": 9.390731592566456e-05, "loss": 3.3531, "step": 7770 }, { "epoch": 0.61, "learning_rate": 9.390653179644006e-05, "loss": 3.1891, "step": 7771 }, { "epoch": 0.61, "learning_rate": 9.390574766721556e-05, "loss": 3.122, "step": 7772 }, { "epoch": 0.61, "learning_rate": 9.390496353799107e-05, "loss": 3.0728, "step": 7773 }, { "epoch": 0.61, "learning_rate": 9.390417940876657e-05, "loss": 2.9449, "step": 7774 }, { "epoch": 0.61, "learning_rate": 9.390339527954207e-05, "loss": 2.8682, "step": 7775 }, { "epoch": 0.61, "learning_rate": 9.390261115031758e-05, "loss": 2.9565, "step": 7776 }, { "epoch": 0.61, "learning_rate": 9.390182702109308e-05, "loss": 2.9397, "step": 7777 }, { "epoch": 0.61, "learning_rate": 9.390104289186859e-05, "loss": 3.0961, "step": 7778 }, { "epoch": 0.61, "learning_rate": 9.390025876264409e-05, "loss": 2.919, "step": 7779 }, { "epoch": 0.61, "learning_rate": 9.389947463341959e-05, "loss": 2.972, "step": 7780 }, { "epoch": 0.61, "learning_rate": 9.38986905041951e-05, "loss": 3.2677, "step": 7781 }, { "epoch": 0.61, "learning_rate": 9.38979063749706e-05, "loss": 3.1159, "step": 7782 }, { "epoch": 0.61, "learning_rate": 9.38971222457461e-05, "loss": 3.0789, "step": 7783 }, { "epoch": 0.61, "learning_rate": 9.389633811652161e-05, "loss": 3.2109, "step": 7784 }, { "epoch": 0.61, "learning_rate": 9.389555398729711e-05, "loss": 2.9421, "step": 7785 }, { "epoch": 0.61, "learning_rate": 9.389476985807262e-05, "loss": 2.9507, "step": 7786 }, { "epoch": 0.61, "learning_rate": 9.389398572884812e-05, "loss": 3.1226, "step": 7787 }, { "epoch": 0.61, "learning_rate": 9.389320159962362e-05, "loss": 3.1184, "step": 7788 }, { "epoch": 0.61, "learning_rate": 9.389241747039913e-05, "loss": 3.0958, "step": 7789 }, { "epoch": 0.61, "learning_rate": 9.389163334117463e-05, "loss": 2.9978, "step": 7790 }, { "epoch": 0.61, "learning_rate": 9.389084921195013e-05, "loss": 2.8329, "step": 7791 }, { "epoch": 0.61, "learning_rate": 9.389006508272564e-05, "loss": 2.912, "step": 7792 }, { "epoch": 0.61, "learning_rate": 9.388928095350114e-05, "loss": 3.0286, "step": 7793 }, { "epoch": 0.61, "learning_rate": 9.388849682427664e-05, "loss": 3.0676, "step": 7794 }, { "epoch": 0.61, "learning_rate": 9.388771269505215e-05, "loss": 3.1638, "step": 7795 }, { "epoch": 0.61, "learning_rate": 9.388692856582765e-05, "loss": 2.948, "step": 7796 }, { "epoch": 0.61, "learning_rate": 9.388614443660316e-05, "loss": 2.9032, "step": 7797 }, { "epoch": 0.61, "learning_rate": 9.388536030737866e-05, "loss": 2.9181, "step": 7798 }, { "epoch": 0.61, "learning_rate": 9.388457617815416e-05, "loss": 3.1215, "step": 7799 }, { "epoch": 0.61, "learning_rate": 9.388379204892967e-05, "loss": 3.1415, "step": 7800 }, { "epoch": 0.61, "learning_rate": 9.388300791970517e-05, "loss": 3.0582, "step": 7801 }, { "epoch": 0.61, "learning_rate": 9.388222379048067e-05, "loss": 2.9225, "step": 7802 }, { "epoch": 0.61, "learning_rate": 9.388143966125618e-05, "loss": 3.1745, "step": 7803 }, { "epoch": 0.61, "learning_rate": 9.388065553203168e-05, "loss": 2.8427, "step": 7804 }, { "epoch": 0.61, "learning_rate": 9.387987140280719e-05, "loss": 2.764, "step": 7805 }, { "epoch": 0.61, "learning_rate": 9.387908727358269e-05, "loss": 3.0558, "step": 7806 }, { "epoch": 0.61, "learning_rate": 9.387830314435819e-05, "loss": 3.0617, "step": 7807 }, { "epoch": 0.61, "learning_rate": 9.38775190151337e-05, "loss": 3.0629, "step": 7808 }, { "epoch": 0.61, "learning_rate": 9.38767348859092e-05, "loss": 3.0258, "step": 7809 }, { "epoch": 0.61, "learning_rate": 9.38759507566847e-05, "loss": 2.8385, "step": 7810 }, { "epoch": 0.61, "learning_rate": 9.387516662746021e-05, "loss": 2.9326, "step": 7811 }, { "epoch": 0.61, "learning_rate": 9.387438249823571e-05, "loss": 3.1162, "step": 7812 }, { "epoch": 0.61, "learning_rate": 9.387359836901121e-05, "loss": 2.8079, "step": 7813 }, { "epoch": 0.61, "learning_rate": 9.387281423978672e-05, "loss": 3.0635, "step": 7814 }, { "epoch": 0.61, "learning_rate": 9.387203011056222e-05, "loss": 3.0456, "step": 7815 }, { "epoch": 0.61, "learning_rate": 9.387124598133773e-05, "loss": 3.0405, "step": 7816 }, { "epoch": 0.61, "learning_rate": 9.387046185211323e-05, "loss": 3.1327, "step": 7817 }, { "epoch": 0.61, "learning_rate": 9.386967772288873e-05, "loss": 2.8096, "step": 7818 }, { "epoch": 0.61, "learning_rate": 9.386889359366424e-05, "loss": 3.3264, "step": 7819 }, { "epoch": 0.61, "learning_rate": 9.386810946443974e-05, "loss": 2.9817, "step": 7820 }, { "epoch": 0.61, "learning_rate": 9.386732533521524e-05, "loss": 3.1631, "step": 7821 }, { "epoch": 0.61, "learning_rate": 9.386654120599076e-05, "loss": 2.9435, "step": 7822 }, { "epoch": 0.61, "learning_rate": 9.386575707676626e-05, "loss": 2.9792, "step": 7823 }, { "epoch": 0.61, "learning_rate": 9.386497294754177e-05, "loss": 2.9783, "step": 7824 }, { "epoch": 0.61, "learning_rate": 9.386418881831727e-05, "loss": 2.9303, "step": 7825 }, { "epoch": 0.61, "learning_rate": 9.386340468909278e-05, "loss": 3.2628, "step": 7826 }, { "epoch": 0.61, "learning_rate": 9.386262055986828e-05, "loss": 3.0889, "step": 7827 }, { "epoch": 0.61, "learning_rate": 9.386183643064378e-05, "loss": 3.202, "step": 7828 }, { "epoch": 0.61, "learning_rate": 9.386105230141929e-05, "loss": 3.2233, "step": 7829 }, { "epoch": 0.61, "learning_rate": 9.386026817219479e-05, "loss": 2.9423, "step": 7830 }, { "epoch": 0.61, "learning_rate": 9.38594840429703e-05, "loss": 2.9813, "step": 7831 }, { "epoch": 0.61, "learning_rate": 9.385869991374578e-05, "loss": 3.1532, "step": 7832 }, { "epoch": 0.61, "learning_rate": 9.385791578452129e-05, "loss": 3.0907, "step": 7833 }, { "epoch": 0.61, "learning_rate": 9.385713165529679e-05, "loss": 3.0265, "step": 7834 }, { "epoch": 0.61, "learning_rate": 9.38563475260723e-05, "loss": 3.0661, "step": 7835 }, { "epoch": 0.61, "learning_rate": 9.38555633968478e-05, "loss": 2.709, "step": 7836 }, { "epoch": 0.61, "learning_rate": 9.38547792676233e-05, "loss": 3.0557, "step": 7837 }, { "epoch": 0.61, "learning_rate": 9.38539951383988e-05, "loss": 3.0391, "step": 7838 }, { "epoch": 0.61, "learning_rate": 9.385321100917431e-05, "loss": 2.9103, "step": 7839 }, { "epoch": 0.61, "learning_rate": 9.385242687994981e-05, "loss": 2.8823, "step": 7840 }, { "epoch": 0.61, "learning_rate": 9.385164275072532e-05, "loss": 3.026, "step": 7841 }, { "epoch": 0.61, "learning_rate": 9.385085862150082e-05, "loss": 3.2173, "step": 7842 }, { "epoch": 0.61, "learning_rate": 9.385007449227632e-05, "loss": 3.4287, "step": 7843 }, { "epoch": 0.62, "learning_rate": 9.384929036305183e-05, "loss": 3.0421, "step": 7844 }, { "epoch": 0.62, "learning_rate": 9.384850623382733e-05, "loss": 3.0508, "step": 7845 }, { "epoch": 0.62, "learning_rate": 9.384772210460284e-05, "loss": 3.1749, "step": 7846 }, { "epoch": 0.62, "learning_rate": 9.384693797537834e-05, "loss": 2.879, "step": 7847 }, { "epoch": 0.62, "learning_rate": 9.384615384615386e-05, "loss": 3.0182, "step": 7848 }, { "epoch": 0.62, "learning_rate": 9.384536971692936e-05, "loss": 2.9119, "step": 7849 }, { "epoch": 0.62, "learning_rate": 9.384458558770486e-05, "loss": 3.057, "step": 7850 }, { "epoch": 0.62, "learning_rate": 9.384380145848037e-05, "loss": 3.1376, "step": 7851 }, { "epoch": 0.62, "learning_rate": 9.384301732925587e-05, "loss": 3.1356, "step": 7852 }, { "epoch": 0.62, "learning_rate": 9.384223320003138e-05, "loss": 2.892, "step": 7853 }, { "epoch": 0.62, "learning_rate": 9.384144907080688e-05, "loss": 3.0487, "step": 7854 }, { "epoch": 0.62, "learning_rate": 9.384066494158238e-05, "loss": 2.8773, "step": 7855 }, { "epoch": 0.62, "learning_rate": 9.383988081235789e-05, "loss": 3.1493, "step": 7856 }, { "epoch": 0.62, "learning_rate": 9.383909668313339e-05, "loss": 2.918, "step": 7857 }, { "epoch": 0.62, "learning_rate": 9.38383125539089e-05, "loss": 2.9228, "step": 7858 }, { "epoch": 0.62, "learning_rate": 9.38375284246844e-05, "loss": 3.0506, "step": 7859 }, { "epoch": 0.62, "learning_rate": 9.38367442954599e-05, "loss": 3.0806, "step": 7860 }, { "epoch": 0.62, "learning_rate": 9.38359601662354e-05, "loss": 2.8513, "step": 7861 }, { "epoch": 0.62, "learning_rate": 9.383517603701091e-05, "loss": 3.1821, "step": 7862 }, { "epoch": 0.62, "learning_rate": 9.383439190778641e-05, "loss": 2.8717, "step": 7863 }, { "epoch": 0.62, "learning_rate": 9.383360777856192e-05, "loss": 3.012, "step": 7864 }, { "epoch": 0.62, "learning_rate": 9.383282364933742e-05, "loss": 3.0424, "step": 7865 }, { "epoch": 0.62, "learning_rate": 9.383203952011292e-05, "loss": 2.8955, "step": 7866 }, { "epoch": 0.62, "learning_rate": 9.383125539088843e-05, "loss": 2.9319, "step": 7867 }, { "epoch": 0.62, "learning_rate": 9.383047126166392e-05, "loss": 2.9229, "step": 7868 }, { "epoch": 0.62, "learning_rate": 9.382968713243942e-05, "loss": 3.0448, "step": 7869 }, { "epoch": 0.62, "learning_rate": 9.382890300321492e-05, "loss": 2.8857, "step": 7870 }, { "epoch": 0.62, "learning_rate": 9.382811887399043e-05, "loss": 2.93, "step": 7871 }, { "epoch": 0.62, "learning_rate": 9.382733474476593e-05, "loss": 2.832, "step": 7872 }, { "epoch": 0.62, "learning_rate": 9.382655061554145e-05, "loss": 3.029, "step": 7873 }, { "epoch": 0.62, "learning_rate": 9.382576648631695e-05, "loss": 2.9374, "step": 7874 }, { "epoch": 0.62, "learning_rate": 9.382498235709246e-05, "loss": 3.0727, "step": 7875 }, { "epoch": 0.62, "learning_rate": 9.382419822786796e-05, "loss": 2.981, "step": 7876 }, { "epoch": 0.62, "learning_rate": 9.382341409864346e-05, "loss": 3.172, "step": 7877 }, { "epoch": 0.62, "learning_rate": 9.382262996941897e-05, "loss": 3.0373, "step": 7878 }, { "epoch": 0.62, "learning_rate": 9.382184584019447e-05, "loss": 3.1091, "step": 7879 }, { "epoch": 0.62, "learning_rate": 9.382106171096997e-05, "loss": 3.1021, "step": 7880 }, { "epoch": 0.62, "learning_rate": 9.382027758174548e-05, "loss": 3.1863, "step": 7881 }, { "epoch": 0.62, "learning_rate": 9.381949345252098e-05, "loss": 3.0443, "step": 7882 }, { "epoch": 0.62, "learning_rate": 9.381870932329649e-05, "loss": 3.0874, "step": 7883 }, { "epoch": 0.62, "learning_rate": 9.381792519407199e-05, "loss": 2.8338, "step": 7884 }, { "epoch": 0.62, "learning_rate": 9.381714106484749e-05, "loss": 3.0471, "step": 7885 }, { "epoch": 0.62, "learning_rate": 9.3816356935623e-05, "loss": 3.1088, "step": 7886 }, { "epoch": 0.62, "learning_rate": 9.38155728063985e-05, "loss": 3.1674, "step": 7887 }, { "epoch": 0.62, "learning_rate": 9.3814788677174e-05, "loss": 3.1668, "step": 7888 }, { "epoch": 0.62, "learning_rate": 9.381400454794951e-05, "loss": 3.0605, "step": 7889 }, { "epoch": 0.62, "learning_rate": 9.381322041872501e-05, "loss": 2.9638, "step": 7890 }, { "epoch": 0.62, "learning_rate": 9.381243628950051e-05, "loss": 2.9351, "step": 7891 }, { "epoch": 0.62, "learning_rate": 9.381165216027602e-05, "loss": 3.0681, "step": 7892 }, { "epoch": 0.62, "learning_rate": 9.381086803105152e-05, "loss": 3.019, "step": 7893 }, { "epoch": 0.62, "learning_rate": 9.381008390182703e-05, "loss": 2.9018, "step": 7894 }, { "epoch": 0.62, "learning_rate": 9.380929977260253e-05, "loss": 3.0211, "step": 7895 }, { "epoch": 0.62, "learning_rate": 9.380851564337803e-05, "loss": 3.0085, "step": 7896 }, { "epoch": 0.62, "learning_rate": 9.380773151415354e-05, "loss": 2.9095, "step": 7897 }, { "epoch": 0.62, "learning_rate": 9.380694738492904e-05, "loss": 3.3378, "step": 7898 }, { "epoch": 0.62, "learning_rate": 9.380616325570454e-05, "loss": 3.0407, "step": 7899 }, { "epoch": 0.62, "learning_rate": 9.380537912648005e-05, "loss": 3.0934, "step": 7900 }, { "epoch": 0.62, "learning_rate": 9.380459499725555e-05, "loss": 2.8727, "step": 7901 }, { "epoch": 0.62, "learning_rate": 9.380381086803106e-05, "loss": 2.8347, "step": 7902 }, { "epoch": 0.62, "learning_rate": 9.380302673880656e-05, "loss": 2.9974, "step": 7903 }, { "epoch": 0.62, "learning_rate": 9.380224260958206e-05, "loss": 2.8978, "step": 7904 }, { "epoch": 0.62, "learning_rate": 9.380145848035757e-05, "loss": 2.8099, "step": 7905 }, { "epoch": 0.62, "learning_rate": 9.380067435113307e-05, "loss": 2.9421, "step": 7906 }, { "epoch": 0.62, "learning_rate": 9.379989022190857e-05, "loss": 3.0405, "step": 7907 }, { "epoch": 0.62, "learning_rate": 9.379910609268408e-05, "loss": 3.0245, "step": 7908 }, { "epoch": 0.62, "learning_rate": 9.379832196345958e-05, "loss": 3.0596, "step": 7909 }, { "epoch": 0.62, "learning_rate": 9.379753783423508e-05, "loss": 3.2482, "step": 7910 }, { "epoch": 0.62, "learning_rate": 9.379675370501059e-05, "loss": 3.2497, "step": 7911 }, { "epoch": 0.62, "learning_rate": 9.379596957578609e-05, "loss": 2.8265, "step": 7912 }, { "epoch": 0.62, "learning_rate": 9.37951854465616e-05, "loss": 2.9299, "step": 7913 }, { "epoch": 0.62, "learning_rate": 9.37944013173371e-05, "loss": 3.0266, "step": 7914 }, { "epoch": 0.62, "learning_rate": 9.37936171881126e-05, "loss": 3.0052, "step": 7915 }, { "epoch": 0.62, "learning_rate": 9.379283305888811e-05, "loss": 3.083, "step": 7916 }, { "epoch": 0.62, "learning_rate": 9.379204892966361e-05, "loss": 3.0423, "step": 7917 }, { "epoch": 0.62, "learning_rate": 9.379126480043911e-05, "loss": 3.2074, "step": 7918 }, { "epoch": 0.62, "learning_rate": 9.379048067121462e-05, "loss": 3.0429, "step": 7919 }, { "epoch": 0.62, "learning_rate": 9.378969654199012e-05, "loss": 2.8945, "step": 7920 }, { "epoch": 0.62, "learning_rate": 9.378891241276563e-05, "loss": 3.1186, "step": 7921 }, { "epoch": 0.62, "learning_rate": 9.378812828354113e-05, "loss": 2.8803, "step": 7922 }, { "epoch": 0.62, "learning_rate": 9.378734415431663e-05, "loss": 3.1876, "step": 7923 }, { "epoch": 0.62, "learning_rate": 9.378656002509214e-05, "loss": 3.0416, "step": 7924 }, { "epoch": 0.62, "learning_rate": 9.378577589586764e-05, "loss": 3.174, "step": 7925 }, { "epoch": 0.62, "learning_rate": 9.378499176664314e-05, "loss": 2.9104, "step": 7926 }, { "epoch": 0.62, "learning_rate": 9.378420763741865e-05, "loss": 2.9156, "step": 7927 }, { "epoch": 0.62, "learning_rate": 9.378342350819415e-05, "loss": 3.0728, "step": 7928 }, { "epoch": 0.62, "learning_rate": 9.378263937896965e-05, "loss": 3.1813, "step": 7929 }, { "epoch": 0.62, "learning_rate": 9.378185524974517e-05, "loss": 2.908, "step": 7930 }, { "epoch": 0.62, "learning_rate": 9.378107112052068e-05, "loss": 3.0396, "step": 7931 }, { "epoch": 0.62, "learning_rate": 9.378028699129618e-05, "loss": 3.1585, "step": 7932 }, { "epoch": 0.62, "learning_rate": 9.377950286207168e-05, "loss": 2.9537, "step": 7933 }, { "epoch": 0.62, "learning_rate": 9.377871873284719e-05, "loss": 3.3137, "step": 7934 }, { "epoch": 0.62, "learning_rate": 9.377793460362269e-05, "loss": 3.0569, "step": 7935 }, { "epoch": 0.62, "learning_rate": 9.37771504743982e-05, "loss": 3.2299, "step": 7936 }, { "epoch": 0.62, "learning_rate": 9.37763663451737e-05, "loss": 3.112, "step": 7937 }, { "epoch": 0.62, "learning_rate": 9.37755822159492e-05, "loss": 3.2154, "step": 7938 }, { "epoch": 0.62, "learning_rate": 9.37747980867247e-05, "loss": 3.1763, "step": 7939 }, { "epoch": 0.62, "learning_rate": 9.37740139575002e-05, "loss": 3.1481, "step": 7940 }, { "epoch": 0.62, "learning_rate": 9.37732298282757e-05, "loss": 2.9493, "step": 7941 }, { "epoch": 0.62, "learning_rate": 9.37724456990512e-05, "loss": 2.8959, "step": 7942 }, { "epoch": 0.62, "learning_rate": 9.37716615698267e-05, "loss": 3.3023, "step": 7943 }, { "epoch": 0.62, "learning_rate": 9.377087744060221e-05, "loss": 3.1036, "step": 7944 }, { "epoch": 0.62, "learning_rate": 9.377009331137771e-05, "loss": 2.9477, "step": 7945 }, { "epoch": 0.62, "learning_rate": 9.376930918215322e-05, "loss": 3.0826, "step": 7946 }, { "epoch": 0.62, "learning_rate": 9.376852505292872e-05, "loss": 2.9182, "step": 7947 }, { "epoch": 0.62, "learning_rate": 9.376774092370422e-05, "loss": 3.0392, "step": 7948 }, { "epoch": 0.62, "learning_rate": 9.376695679447973e-05, "loss": 3.0931, "step": 7949 }, { "epoch": 0.62, "learning_rate": 9.376617266525523e-05, "loss": 2.98, "step": 7950 }, { "epoch": 0.62, "learning_rate": 9.376538853603074e-05, "loss": 3.1929, "step": 7951 }, { "epoch": 0.62, "learning_rate": 9.376460440680624e-05, "loss": 3.1955, "step": 7952 }, { "epoch": 0.62, "learning_rate": 9.376382027758174e-05, "loss": 2.9473, "step": 7953 }, { "epoch": 0.62, "learning_rate": 9.376303614835725e-05, "loss": 3.0657, "step": 7954 }, { "epoch": 0.62, "learning_rate": 9.376225201913275e-05, "loss": 2.9496, "step": 7955 }, { "epoch": 0.62, "learning_rate": 9.376146788990827e-05, "loss": 3.1379, "step": 7956 }, { "epoch": 0.62, "learning_rate": 9.376068376068377e-05, "loss": 2.9369, "step": 7957 }, { "epoch": 0.62, "learning_rate": 9.375989963145927e-05, "loss": 3.0974, "step": 7958 }, { "epoch": 0.62, "learning_rate": 9.375911550223478e-05, "loss": 3.0768, "step": 7959 }, { "epoch": 0.62, "learning_rate": 9.375833137301028e-05, "loss": 3.2295, "step": 7960 }, { "epoch": 0.62, "learning_rate": 9.375754724378579e-05, "loss": 3.2656, "step": 7961 }, { "epoch": 0.62, "learning_rate": 9.375676311456129e-05, "loss": 3.1153, "step": 7962 }, { "epoch": 0.62, "learning_rate": 9.375597898533679e-05, "loss": 2.961, "step": 7963 }, { "epoch": 0.62, "learning_rate": 9.37551948561123e-05, "loss": 2.7931, "step": 7964 }, { "epoch": 0.62, "learning_rate": 9.37544107268878e-05, "loss": 2.9285, "step": 7965 }, { "epoch": 0.62, "learning_rate": 9.37536265976633e-05, "loss": 2.9118, "step": 7966 }, { "epoch": 0.62, "learning_rate": 9.375284246843881e-05, "loss": 3.1847, "step": 7967 }, { "epoch": 0.62, "learning_rate": 9.375205833921431e-05, "loss": 3.0135, "step": 7968 }, { "epoch": 0.62, "learning_rate": 9.375127420998982e-05, "loss": 2.9187, "step": 7969 }, { "epoch": 0.62, "learning_rate": 9.375049008076532e-05, "loss": 3.3518, "step": 7970 }, { "epoch": 0.63, "learning_rate": 9.374970595154082e-05, "loss": 3.2544, "step": 7971 }, { "epoch": 0.63, "learning_rate": 9.374892182231633e-05, "loss": 3.1769, "step": 7972 }, { "epoch": 0.63, "learning_rate": 9.374813769309183e-05, "loss": 2.9575, "step": 7973 }, { "epoch": 0.63, "learning_rate": 9.374735356386733e-05, "loss": 3.0124, "step": 7974 }, { "epoch": 0.63, "learning_rate": 9.374656943464284e-05, "loss": 3.0401, "step": 7975 }, { "epoch": 0.63, "learning_rate": 9.374578530541833e-05, "loss": 3.1424, "step": 7976 }, { "epoch": 0.63, "learning_rate": 9.374500117619383e-05, "loss": 2.8808, "step": 7977 }, { "epoch": 0.63, "learning_rate": 9.374421704696933e-05, "loss": 3.0772, "step": 7978 }, { "epoch": 0.63, "learning_rate": 9.374343291774484e-05, "loss": 2.9437, "step": 7979 }, { "epoch": 0.63, "learning_rate": 9.374264878852034e-05, "loss": 3.2913, "step": 7980 }, { "epoch": 0.63, "learning_rate": 9.374186465929586e-05, "loss": 3.103, "step": 7981 }, { "epoch": 0.63, "learning_rate": 9.374108053007136e-05, "loss": 3.0195, "step": 7982 }, { "epoch": 0.63, "learning_rate": 9.374029640084687e-05, "loss": 3.0209, "step": 7983 }, { "epoch": 0.63, "learning_rate": 9.373951227162237e-05, "loss": 3.031, "step": 7984 }, { "epoch": 0.63, "learning_rate": 9.373872814239787e-05, "loss": 3.4697, "step": 7985 }, { "epoch": 0.63, "learning_rate": 9.373794401317338e-05, "loss": 3.2917, "step": 7986 }, { "epoch": 0.63, "learning_rate": 9.373715988394888e-05, "loss": 2.9209, "step": 7987 }, { "epoch": 0.63, "learning_rate": 9.373637575472439e-05, "loss": 2.9791, "step": 7988 }, { "epoch": 0.63, "learning_rate": 9.373559162549989e-05, "loss": 3.1694, "step": 7989 }, { "epoch": 0.63, "learning_rate": 9.373480749627539e-05, "loss": 3.126, "step": 7990 }, { "epoch": 0.63, "learning_rate": 9.37340233670509e-05, "loss": 3.0377, "step": 7991 }, { "epoch": 0.63, "learning_rate": 9.37332392378264e-05, "loss": 2.8641, "step": 7992 }, { "epoch": 0.63, "learning_rate": 9.37324551086019e-05, "loss": 3.0152, "step": 7993 }, { "epoch": 0.63, "learning_rate": 9.373167097937741e-05, "loss": 3.2054, "step": 7994 }, { "epoch": 0.63, "learning_rate": 9.373088685015291e-05, "loss": 2.8521, "step": 7995 }, { "epoch": 0.63, "learning_rate": 9.373010272092841e-05, "loss": 3.1119, "step": 7996 }, { "epoch": 0.63, "learning_rate": 9.372931859170392e-05, "loss": 2.9397, "step": 7997 }, { "epoch": 0.63, "learning_rate": 9.372853446247942e-05, "loss": 2.7939, "step": 7998 }, { "epoch": 0.63, "learning_rate": 9.372775033325493e-05, "loss": 3.0259, "step": 7999 }, { "epoch": 0.63, "learning_rate": 9.372696620403043e-05, "loss": 2.9716, "step": 8000 }, { "epoch": 0.63, "eval_loss": 2.948347568511963, "eval_runtime": 218.2659, "eval_samples_per_second": 51.918, "eval_steps_per_second": 3.248, "step": 8000 }, { "epoch": 0.63, "learning_rate": 9.372618207480593e-05, "loss": 3.0458, "step": 8001 }, { "epoch": 0.63, "learning_rate": 9.372539794558144e-05, "loss": 3.0047, "step": 8002 }, { "epoch": 0.63, "learning_rate": 9.372461381635694e-05, "loss": 3.0964, "step": 8003 }, { "epoch": 0.63, "learning_rate": 9.372382968713244e-05, "loss": 3.0426, "step": 8004 }, { "epoch": 0.63, "learning_rate": 9.372304555790795e-05, "loss": 3.1118, "step": 8005 }, { "epoch": 0.63, "learning_rate": 9.372226142868345e-05, "loss": 2.9741, "step": 8006 }, { "epoch": 0.63, "learning_rate": 9.372147729945896e-05, "loss": 2.9081, "step": 8007 }, { "epoch": 0.63, "learning_rate": 9.372069317023446e-05, "loss": 3.2023, "step": 8008 }, { "epoch": 0.63, "learning_rate": 9.371990904100996e-05, "loss": 3.5184, "step": 8009 }, { "epoch": 0.63, "learning_rate": 9.371912491178547e-05, "loss": 2.8965, "step": 8010 }, { "epoch": 0.63, "learning_rate": 9.371834078256097e-05, "loss": 3.0189, "step": 8011 }, { "epoch": 0.63, "learning_rate": 9.371755665333647e-05, "loss": 2.9944, "step": 8012 }, { "epoch": 0.63, "learning_rate": 9.371677252411198e-05, "loss": 2.9608, "step": 8013 }, { "epoch": 0.63, "learning_rate": 9.371598839488748e-05, "loss": 2.9708, "step": 8014 }, { "epoch": 0.63, "learning_rate": 9.371520426566298e-05, "loss": 3.2366, "step": 8015 }, { "epoch": 0.63, "learning_rate": 9.371442013643849e-05, "loss": 2.7455, "step": 8016 }, { "epoch": 0.63, "learning_rate": 9.371363600721399e-05, "loss": 3.1463, "step": 8017 }, { "epoch": 0.63, "learning_rate": 9.37128518779895e-05, "loss": 2.9292, "step": 8018 }, { "epoch": 0.63, "learning_rate": 9.3712067748765e-05, "loss": 3.041, "step": 8019 }, { "epoch": 0.63, "learning_rate": 9.37112836195405e-05, "loss": 2.9152, "step": 8020 }, { "epoch": 0.63, "learning_rate": 9.3710499490316e-05, "loss": 3.0799, "step": 8021 }, { "epoch": 0.63, "learning_rate": 9.370971536109151e-05, "loss": 3.0369, "step": 8022 }, { "epoch": 0.63, "learning_rate": 9.370893123186701e-05, "loss": 3.2657, "step": 8023 }, { "epoch": 0.63, "learning_rate": 9.370814710264252e-05, "loss": 2.8514, "step": 8024 }, { "epoch": 0.63, "learning_rate": 9.370736297341802e-05, "loss": 2.7035, "step": 8025 }, { "epoch": 0.63, "learning_rate": 9.370657884419352e-05, "loss": 3.0072, "step": 8026 }, { "epoch": 0.63, "learning_rate": 9.370579471496903e-05, "loss": 3.2583, "step": 8027 }, { "epoch": 0.63, "learning_rate": 9.370501058574453e-05, "loss": 2.989, "step": 8028 }, { "epoch": 0.63, "learning_rate": 9.370422645652004e-05, "loss": 3.2128, "step": 8029 }, { "epoch": 0.63, "learning_rate": 9.370344232729554e-05, "loss": 3.0119, "step": 8030 }, { "epoch": 0.63, "learning_rate": 9.370265819807104e-05, "loss": 3.2586, "step": 8031 }, { "epoch": 0.63, "learning_rate": 9.370187406884655e-05, "loss": 3.0776, "step": 8032 }, { "epoch": 0.63, "learning_rate": 9.370108993962205e-05, "loss": 3.2089, "step": 8033 }, { "epoch": 0.63, "learning_rate": 9.370030581039755e-05, "loss": 2.9657, "step": 8034 }, { "epoch": 0.63, "learning_rate": 9.369952168117306e-05, "loss": 3.1393, "step": 8035 }, { "epoch": 0.63, "learning_rate": 9.369873755194856e-05, "loss": 2.8457, "step": 8036 }, { "epoch": 0.63, "learning_rate": 9.369795342272407e-05, "loss": 2.9231, "step": 8037 }, { "epoch": 0.63, "learning_rate": 9.369716929349958e-05, "loss": 3.3112, "step": 8038 }, { "epoch": 0.63, "learning_rate": 9.369638516427509e-05, "loss": 3.3004, "step": 8039 }, { "epoch": 0.63, "learning_rate": 9.369560103505059e-05, "loss": 3.1316, "step": 8040 }, { "epoch": 0.63, "learning_rate": 9.36948169058261e-05, "loss": 3.1437, "step": 8041 }, { "epoch": 0.63, "learning_rate": 9.36940327766016e-05, "loss": 2.8987, "step": 8042 }, { "epoch": 0.63, "learning_rate": 9.36932486473771e-05, "loss": 3.3341, "step": 8043 }, { "epoch": 0.63, "learning_rate": 9.36924645181526e-05, "loss": 3.2328, "step": 8044 }, { "epoch": 0.63, "learning_rate": 9.369168038892811e-05, "loss": 3.1823, "step": 8045 }, { "epoch": 0.63, "learning_rate": 9.369089625970361e-05, "loss": 3.0267, "step": 8046 }, { "epoch": 0.63, "learning_rate": 9.369011213047912e-05, "loss": 3.1916, "step": 8047 }, { "epoch": 0.63, "learning_rate": 9.36893280012546e-05, "loss": 3.0633, "step": 8048 }, { "epoch": 0.63, "learning_rate": 9.368854387203011e-05, "loss": 3.1054, "step": 8049 }, { "epoch": 0.63, "learning_rate": 9.368775974280561e-05, "loss": 2.9026, "step": 8050 }, { "epoch": 0.63, "learning_rate": 9.368697561358112e-05, "loss": 3.192, "step": 8051 }, { "epoch": 0.63, "learning_rate": 9.368619148435662e-05, "loss": 3.2936, "step": 8052 }, { "epoch": 0.63, "learning_rate": 9.368540735513212e-05, "loss": 2.7915, "step": 8053 }, { "epoch": 0.63, "learning_rate": 9.368462322590763e-05, "loss": 2.8901, "step": 8054 }, { "epoch": 0.63, "learning_rate": 9.368383909668313e-05, "loss": 3.1879, "step": 8055 }, { "epoch": 0.63, "learning_rate": 9.368305496745864e-05, "loss": 2.8287, "step": 8056 }, { "epoch": 0.63, "learning_rate": 9.368227083823414e-05, "loss": 2.8712, "step": 8057 }, { "epoch": 0.63, "learning_rate": 9.368148670900964e-05, "loss": 3.1177, "step": 8058 }, { "epoch": 0.63, "learning_rate": 9.368070257978515e-05, "loss": 2.9807, "step": 8059 }, { "epoch": 0.63, "learning_rate": 9.367991845056065e-05, "loss": 3.0883, "step": 8060 }, { "epoch": 0.63, "learning_rate": 9.367913432133615e-05, "loss": 2.9895, "step": 8061 }, { "epoch": 0.63, "learning_rate": 9.367835019211166e-05, "loss": 2.9193, "step": 8062 }, { "epoch": 0.63, "learning_rate": 9.367756606288716e-05, "loss": 2.9595, "step": 8063 }, { "epoch": 0.63, "learning_rate": 9.367678193366268e-05, "loss": 3.0477, "step": 8064 }, { "epoch": 0.63, "learning_rate": 9.367599780443818e-05, "loss": 2.7497, "step": 8065 }, { "epoch": 0.63, "learning_rate": 9.367521367521369e-05, "loss": 3.1227, "step": 8066 }, { "epoch": 0.63, "learning_rate": 9.367442954598919e-05, "loss": 3.1631, "step": 8067 }, { "epoch": 0.63, "learning_rate": 9.367364541676469e-05, "loss": 2.9457, "step": 8068 }, { "epoch": 0.63, "learning_rate": 9.36728612875402e-05, "loss": 2.9887, "step": 8069 }, { "epoch": 0.63, "learning_rate": 9.36720771583157e-05, "loss": 3.0456, "step": 8070 }, { "epoch": 0.63, "learning_rate": 9.36712930290912e-05, "loss": 2.8431, "step": 8071 }, { "epoch": 0.63, "learning_rate": 9.367050889986671e-05, "loss": 3.0449, "step": 8072 }, { "epoch": 0.63, "learning_rate": 9.366972477064221e-05, "loss": 3.199, "step": 8073 }, { "epoch": 0.63, "learning_rate": 9.366894064141771e-05, "loss": 2.9928, "step": 8074 }, { "epoch": 0.63, "learning_rate": 9.366815651219322e-05, "loss": 2.8172, "step": 8075 }, { "epoch": 0.63, "learning_rate": 9.366737238296872e-05, "loss": 3.1028, "step": 8076 }, { "epoch": 0.63, "learning_rate": 9.366658825374423e-05, "loss": 3.0163, "step": 8077 }, { "epoch": 0.63, "learning_rate": 9.366580412451973e-05, "loss": 2.883, "step": 8078 }, { "epoch": 0.63, "learning_rate": 9.366501999529523e-05, "loss": 3.0654, "step": 8079 }, { "epoch": 0.63, "learning_rate": 9.366423586607074e-05, "loss": 3.0476, "step": 8080 }, { "epoch": 0.63, "learning_rate": 9.366345173684624e-05, "loss": 3.0807, "step": 8081 }, { "epoch": 0.63, "learning_rate": 9.366266760762174e-05, "loss": 2.8591, "step": 8082 }, { "epoch": 0.63, "learning_rate": 9.366188347839725e-05, "loss": 2.9638, "step": 8083 }, { "epoch": 0.63, "learning_rate": 9.366109934917274e-05, "loss": 2.9838, "step": 8084 }, { "epoch": 0.63, "learning_rate": 9.366031521994824e-05, "loss": 2.7811, "step": 8085 }, { "epoch": 0.63, "learning_rate": 9.365953109072375e-05, "loss": 2.8744, "step": 8086 }, { "epoch": 0.63, "learning_rate": 9.365874696149925e-05, "loss": 3.0408, "step": 8087 }, { "epoch": 0.63, "learning_rate": 9.365796283227475e-05, "loss": 2.7265, "step": 8088 }, { "epoch": 0.63, "learning_rate": 9.365717870305027e-05, "loss": 3.2987, "step": 8089 }, { "epoch": 0.63, "learning_rate": 9.365639457382577e-05, "loss": 3.1603, "step": 8090 }, { "epoch": 0.63, "learning_rate": 9.365561044460128e-05, "loss": 2.9107, "step": 8091 }, { "epoch": 0.63, "learning_rate": 9.365482631537678e-05, "loss": 2.9936, "step": 8092 }, { "epoch": 0.63, "learning_rate": 9.365404218615228e-05, "loss": 2.8232, "step": 8093 }, { "epoch": 0.63, "learning_rate": 9.365325805692779e-05, "loss": 2.9148, "step": 8094 }, { "epoch": 0.63, "learning_rate": 9.365247392770329e-05, "loss": 2.924, "step": 8095 }, { "epoch": 0.63, "learning_rate": 9.36516897984788e-05, "loss": 3.2321, "step": 8096 }, { "epoch": 0.63, "learning_rate": 9.36509056692543e-05, "loss": 3.0168, "step": 8097 }, { "epoch": 0.63, "learning_rate": 9.36501215400298e-05, "loss": 3.0283, "step": 8098 }, { "epoch": 0.64, "learning_rate": 9.364933741080531e-05, "loss": 3.1549, "step": 8099 }, { "epoch": 0.64, "learning_rate": 9.364855328158081e-05, "loss": 3.1176, "step": 8100 }, { "epoch": 0.64, "learning_rate": 9.364776915235631e-05, "loss": 2.9435, "step": 8101 }, { "epoch": 0.64, "learning_rate": 9.364698502313182e-05, "loss": 3.1178, "step": 8102 }, { "epoch": 0.64, "learning_rate": 9.364620089390732e-05, "loss": 3.0433, "step": 8103 }, { "epoch": 0.64, "learning_rate": 9.364541676468283e-05, "loss": 3.087, "step": 8104 }, { "epoch": 0.64, "learning_rate": 9.364463263545833e-05, "loss": 3.1863, "step": 8105 }, { "epoch": 0.64, "learning_rate": 9.364384850623383e-05, "loss": 3.2196, "step": 8106 }, { "epoch": 0.64, "learning_rate": 9.364306437700934e-05, "loss": 3.0602, "step": 8107 }, { "epoch": 0.64, "learning_rate": 9.364228024778484e-05, "loss": 3.3747, "step": 8108 }, { "epoch": 0.64, "learning_rate": 9.364149611856034e-05, "loss": 3.076, "step": 8109 }, { "epoch": 0.64, "learning_rate": 9.364071198933585e-05, "loss": 2.9938, "step": 8110 }, { "epoch": 0.64, "learning_rate": 9.363992786011135e-05, "loss": 2.826, "step": 8111 }, { "epoch": 0.64, "learning_rate": 9.363914373088685e-05, "loss": 3.0627, "step": 8112 }, { "epoch": 0.64, "learning_rate": 9.363835960166236e-05, "loss": 3.0394, "step": 8113 }, { "epoch": 0.64, "learning_rate": 9.363757547243786e-05, "loss": 3.027, "step": 8114 }, { "epoch": 0.64, "learning_rate": 9.363679134321337e-05, "loss": 3.0493, "step": 8115 }, { "epoch": 0.64, "learning_rate": 9.363600721398887e-05, "loss": 3.2874, "step": 8116 }, { "epoch": 0.64, "learning_rate": 9.363522308476437e-05, "loss": 3.1276, "step": 8117 }, { "epoch": 0.64, "learning_rate": 9.363443895553988e-05, "loss": 3.016, "step": 8118 }, { "epoch": 0.64, "learning_rate": 9.363365482631538e-05, "loss": 3.0738, "step": 8119 }, { "epoch": 0.64, "learning_rate": 9.363287069709088e-05, "loss": 3.2039, "step": 8120 }, { "epoch": 0.64, "learning_rate": 9.363208656786639e-05, "loss": 2.9799, "step": 8121 }, { "epoch": 0.64, "learning_rate": 9.363130243864189e-05, "loss": 3.1204, "step": 8122 }, { "epoch": 0.64, "learning_rate": 9.36305183094174e-05, "loss": 3.089, "step": 8123 }, { "epoch": 0.64, "learning_rate": 9.36297341801929e-05, "loss": 3.1267, "step": 8124 }, { "epoch": 0.64, "learning_rate": 9.36289500509684e-05, "loss": 3.0555, "step": 8125 }, { "epoch": 0.64, "learning_rate": 9.36281659217439e-05, "loss": 2.8882, "step": 8126 }, { "epoch": 0.64, "learning_rate": 9.362738179251941e-05, "loss": 2.84, "step": 8127 }, { "epoch": 0.64, "learning_rate": 9.362659766329491e-05, "loss": 2.7995, "step": 8128 }, { "epoch": 0.64, "learning_rate": 9.362581353407042e-05, "loss": 3.0887, "step": 8129 }, { "epoch": 0.64, "learning_rate": 9.362502940484592e-05, "loss": 2.816, "step": 8130 }, { "epoch": 0.64, "learning_rate": 9.362424527562142e-05, "loss": 2.9734, "step": 8131 }, { "epoch": 0.64, "learning_rate": 9.362346114639693e-05, "loss": 2.9481, "step": 8132 }, { "epoch": 0.64, "learning_rate": 9.362267701717243e-05, "loss": 2.8342, "step": 8133 }, { "epoch": 0.64, "learning_rate": 9.362189288794794e-05, "loss": 2.9813, "step": 8134 }, { "epoch": 0.64, "learning_rate": 9.362110875872344e-05, "loss": 3.2383, "step": 8135 }, { "epoch": 0.64, "learning_rate": 9.362032462949894e-05, "loss": 2.9777, "step": 8136 }, { "epoch": 0.64, "learning_rate": 9.361954050027445e-05, "loss": 2.9311, "step": 8137 }, { "epoch": 0.64, "learning_rate": 9.361875637104995e-05, "loss": 3.1207, "step": 8138 }, { "epoch": 0.64, "learning_rate": 9.361797224182545e-05, "loss": 2.9456, "step": 8139 }, { "epoch": 0.64, "learning_rate": 9.361718811260096e-05, "loss": 3.001, "step": 8140 }, { "epoch": 0.64, "learning_rate": 9.361640398337646e-05, "loss": 3.0161, "step": 8141 }, { "epoch": 0.64, "learning_rate": 9.361561985415197e-05, "loss": 3.0215, "step": 8142 }, { "epoch": 0.64, "learning_rate": 9.361483572492747e-05, "loss": 3.3941, "step": 8143 }, { "epoch": 0.64, "learning_rate": 9.361405159570297e-05, "loss": 3.2096, "step": 8144 }, { "epoch": 0.64, "learning_rate": 9.361326746647848e-05, "loss": 3.3994, "step": 8145 }, { "epoch": 0.64, "learning_rate": 9.3612483337254e-05, "loss": 3.3052, "step": 8146 }, { "epoch": 0.64, "learning_rate": 9.36116992080295e-05, "loss": 2.936, "step": 8147 }, { "epoch": 0.64, "learning_rate": 9.3610915078805e-05, "loss": 2.9627, "step": 8148 }, { "epoch": 0.64, "learning_rate": 9.36101309495805e-05, "loss": 3.1276, "step": 8149 }, { "epoch": 0.64, "learning_rate": 9.360934682035601e-05, "loss": 3.0628, "step": 8150 }, { "epoch": 0.64, "learning_rate": 9.360856269113151e-05, "loss": 2.9891, "step": 8151 }, { "epoch": 0.64, "learning_rate": 9.360777856190702e-05, "loss": 3.0804, "step": 8152 }, { "epoch": 0.64, "learning_rate": 9.360699443268252e-05, "loss": 3.318, "step": 8153 }, { "epoch": 0.64, "learning_rate": 9.360621030345802e-05, "loss": 3.006, "step": 8154 }, { "epoch": 0.64, "learning_rate": 9.360542617423353e-05, "loss": 3.1143, "step": 8155 }, { "epoch": 0.64, "learning_rate": 9.360464204500902e-05, "loss": 3.2826, "step": 8156 }, { "epoch": 0.64, "learning_rate": 9.360385791578452e-05, "loss": 2.9427, "step": 8157 }, { "epoch": 0.64, "learning_rate": 9.360307378656002e-05, "loss": 2.9377, "step": 8158 }, { "epoch": 0.64, "learning_rate": 9.360228965733553e-05, "loss": 3.0208, "step": 8159 }, { "epoch": 0.64, "learning_rate": 9.360150552811103e-05, "loss": 3.0437, "step": 8160 }, { "epoch": 0.64, "learning_rate": 9.360072139888653e-05, "loss": 3.1858, "step": 8161 }, { "epoch": 0.64, "learning_rate": 9.359993726966204e-05, "loss": 3.15, "step": 8162 }, { "epoch": 0.64, "learning_rate": 9.359915314043754e-05, "loss": 3.0263, "step": 8163 }, { "epoch": 0.64, "learning_rate": 9.359836901121305e-05, "loss": 2.9675, "step": 8164 }, { "epoch": 0.64, "learning_rate": 9.359758488198855e-05, "loss": 3.0529, "step": 8165 }, { "epoch": 0.64, "learning_rate": 9.359680075276405e-05, "loss": 2.8015, "step": 8166 }, { "epoch": 0.64, "learning_rate": 9.359601662353956e-05, "loss": 3.3112, "step": 8167 }, { "epoch": 0.64, "learning_rate": 9.359523249431506e-05, "loss": 3.3894, "step": 8168 }, { "epoch": 0.64, "learning_rate": 9.359444836509056e-05, "loss": 2.6913, "step": 8169 }, { "epoch": 0.64, "learning_rate": 9.359366423586607e-05, "loss": 2.8752, "step": 8170 }, { "epoch": 0.64, "learning_rate": 9.359288010664157e-05, "loss": 2.8407, "step": 8171 }, { "epoch": 0.64, "learning_rate": 9.359209597741709e-05, "loss": 3.0262, "step": 8172 }, { "epoch": 0.64, "learning_rate": 9.359131184819259e-05, "loss": 3.2538, "step": 8173 }, { "epoch": 0.64, "learning_rate": 9.35905277189681e-05, "loss": 3.043, "step": 8174 }, { "epoch": 0.64, "learning_rate": 9.35897435897436e-05, "loss": 3.2383, "step": 8175 }, { "epoch": 0.64, "learning_rate": 9.35889594605191e-05, "loss": 2.8905, "step": 8176 }, { "epoch": 0.64, "learning_rate": 9.358817533129461e-05, "loss": 3.4325, "step": 8177 }, { "epoch": 0.64, "learning_rate": 9.358739120207011e-05, "loss": 3.2162, "step": 8178 }, { "epoch": 0.64, "learning_rate": 9.358660707284561e-05, "loss": 3.158, "step": 8179 }, { "epoch": 0.64, "learning_rate": 9.358582294362112e-05, "loss": 2.8776, "step": 8180 }, { "epoch": 0.64, "learning_rate": 9.358503881439662e-05, "loss": 2.9911, "step": 8181 }, { "epoch": 0.64, "learning_rate": 9.358425468517213e-05, "loss": 2.9393, "step": 8182 }, { "epoch": 0.64, "learning_rate": 9.358347055594763e-05, "loss": 2.9825, "step": 8183 }, { "epoch": 0.64, "learning_rate": 9.358268642672313e-05, "loss": 3.037, "step": 8184 }, { "epoch": 0.64, "learning_rate": 9.358190229749864e-05, "loss": 3.268, "step": 8185 }, { "epoch": 0.64, "learning_rate": 9.358111816827414e-05, "loss": 2.674, "step": 8186 }, { "epoch": 0.64, "learning_rate": 9.358033403904964e-05, "loss": 3.0648, "step": 8187 }, { "epoch": 0.64, "learning_rate": 9.357954990982515e-05, "loss": 3.4709, "step": 8188 }, { "epoch": 0.64, "learning_rate": 9.357876578060065e-05, "loss": 3.1863, "step": 8189 }, { "epoch": 0.64, "learning_rate": 9.357798165137616e-05, "loss": 3.3322, "step": 8190 }, { "epoch": 0.64, "learning_rate": 9.357719752215166e-05, "loss": 2.9728, "step": 8191 }, { "epoch": 0.64, "learning_rate": 9.357641339292715e-05, "loss": 3.1806, "step": 8192 }, { "epoch": 0.64, "learning_rate": 9.357562926370265e-05, "loss": 2.8976, "step": 8193 }, { "epoch": 0.64, "learning_rate": 9.357484513447816e-05, "loss": 3.106, "step": 8194 }, { "epoch": 0.64, "learning_rate": 9.357406100525366e-05, "loss": 3.0794, "step": 8195 }, { "epoch": 0.64, "learning_rate": 9.357327687602916e-05, "loss": 3.3202, "step": 8196 }, { "epoch": 0.64, "learning_rate": 9.357249274680468e-05, "loss": 3.0432, "step": 8197 }, { "epoch": 0.64, "learning_rate": 9.357170861758018e-05, "loss": 3.006, "step": 8198 }, { "epoch": 0.64, "learning_rate": 9.357092448835569e-05, "loss": 3.3898, "step": 8199 }, { "epoch": 0.64, "learning_rate": 9.357014035913119e-05, "loss": 2.9881, "step": 8200 }, { "epoch": 0.64, "learning_rate": 9.35693562299067e-05, "loss": 3.1563, "step": 8201 }, { "epoch": 0.64, "learning_rate": 9.35685721006822e-05, "loss": 2.9157, "step": 8202 }, { "epoch": 0.64, "learning_rate": 9.35677879714577e-05, "loss": 3.069, "step": 8203 }, { "epoch": 0.64, "learning_rate": 9.35670038422332e-05, "loss": 2.9635, "step": 8204 }, { "epoch": 0.64, "learning_rate": 9.356621971300871e-05, "loss": 3.0614, "step": 8205 }, { "epoch": 0.64, "learning_rate": 9.356543558378421e-05, "loss": 2.9638, "step": 8206 }, { "epoch": 0.64, "learning_rate": 9.356465145455972e-05, "loss": 2.7767, "step": 8207 }, { "epoch": 0.64, "learning_rate": 9.356386732533522e-05, "loss": 3.1342, "step": 8208 }, { "epoch": 0.64, "learning_rate": 9.356308319611073e-05, "loss": 2.9357, "step": 8209 }, { "epoch": 0.64, "learning_rate": 9.356229906688623e-05, "loss": 3.147, "step": 8210 }, { "epoch": 0.64, "learning_rate": 9.356151493766173e-05, "loss": 2.9861, "step": 8211 }, { "epoch": 0.64, "learning_rate": 9.356073080843724e-05, "loss": 3.0033, "step": 8212 }, { "epoch": 0.64, "learning_rate": 9.355994667921274e-05, "loss": 2.9767, "step": 8213 }, { "epoch": 0.64, "learning_rate": 9.355916254998824e-05, "loss": 3.0696, "step": 8214 }, { "epoch": 0.64, "learning_rate": 9.355837842076375e-05, "loss": 3.2109, "step": 8215 }, { "epoch": 0.64, "learning_rate": 9.355759429153925e-05, "loss": 3.23, "step": 8216 }, { "epoch": 0.64, "learning_rate": 9.355681016231475e-05, "loss": 3.0476, "step": 8217 }, { "epoch": 0.64, "learning_rate": 9.355602603309026e-05, "loss": 3.0375, "step": 8218 }, { "epoch": 0.64, "learning_rate": 9.355524190386576e-05, "loss": 3.1214, "step": 8219 }, { "epoch": 0.64, "learning_rate": 9.355445777464127e-05, "loss": 2.9846, "step": 8220 }, { "epoch": 0.64, "learning_rate": 9.355367364541677e-05, "loss": 3.0687, "step": 8221 }, { "epoch": 0.64, "learning_rate": 9.355288951619227e-05, "loss": 3.1081, "step": 8222 }, { "epoch": 0.64, "learning_rate": 9.355210538696778e-05, "loss": 3.1077, "step": 8223 }, { "epoch": 0.64, "learning_rate": 9.355132125774328e-05, "loss": 2.9689, "step": 8224 }, { "epoch": 0.64, "learning_rate": 9.355053712851878e-05, "loss": 2.8582, "step": 8225 }, { "epoch": 0.65, "learning_rate": 9.354975299929429e-05, "loss": 3.082, "step": 8226 }, { "epoch": 0.65, "learning_rate": 9.354896887006979e-05, "loss": 2.8949, "step": 8227 }, { "epoch": 0.65, "learning_rate": 9.35481847408453e-05, "loss": 2.8726, "step": 8228 }, { "epoch": 0.65, "learning_rate": 9.35474006116208e-05, "loss": 3.1224, "step": 8229 }, { "epoch": 0.65, "learning_rate": 9.35466164823963e-05, "loss": 2.7506, "step": 8230 }, { "epoch": 0.65, "learning_rate": 9.35458323531718e-05, "loss": 3.1549, "step": 8231 }, { "epoch": 0.65, "learning_rate": 9.354504822394731e-05, "loss": 3.4397, "step": 8232 }, { "epoch": 0.65, "learning_rate": 9.354426409472281e-05, "loss": 2.9223, "step": 8233 }, { "epoch": 0.65, "learning_rate": 9.354347996549832e-05, "loss": 3.12, "step": 8234 }, { "epoch": 0.65, "learning_rate": 9.354269583627382e-05, "loss": 2.9188, "step": 8235 }, { "epoch": 0.65, "learning_rate": 9.354191170704932e-05, "loss": 3.1021, "step": 8236 }, { "epoch": 0.65, "learning_rate": 9.354112757782483e-05, "loss": 2.9562, "step": 8237 }, { "epoch": 0.65, "learning_rate": 9.354034344860033e-05, "loss": 3.0717, "step": 8238 }, { "epoch": 0.65, "learning_rate": 9.353955931937584e-05, "loss": 3.0914, "step": 8239 }, { "epoch": 0.65, "learning_rate": 9.353877519015134e-05, "loss": 3.0379, "step": 8240 }, { "epoch": 0.65, "learning_rate": 9.353799106092684e-05, "loss": 2.8713, "step": 8241 }, { "epoch": 0.65, "learning_rate": 9.353720693170235e-05, "loss": 2.927, "step": 8242 }, { "epoch": 0.65, "learning_rate": 9.353642280247785e-05, "loss": 2.8043, "step": 8243 }, { "epoch": 0.65, "learning_rate": 9.353563867325335e-05, "loss": 2.9751, "step": 8244 }, { "epoch": 0.65, "learning_rate": 9.353485454402886e-05, "loss": 3.0468, "step": 8245 }, { "epoch": 0.65, "learning_rate": 9.353407041480436e-05, "loss": 3.0596, "step": 8246 }, { "epoch": 0.65, "learning_rate": 9.353328628557986e-05, "loss": 2.8752, "step": 8247 }, { "epoch": 0.65, "learning_rate": 9.353250215635537e-05, "loss": 3.0189, "step": 8248 }, { "epoch": 0.65, "learning_rate": 9.353171802713087e-05, "loss": 3.0049, "step": 8249 }, { "epoch": 0.65, "learning_rate": 9.353093389790638e-05, "loss": 3.0098, "step": 8250 }, { "epoch": 0.65, "learning_rate": 9.353014976868188e-05, "loss": 2.8714, "step": 8251 }, { "epoch": 0.65, "learning_rate": 9.352936563945738e-05, "loss": 3.1531, "step": 8252 }, { "epoch": 0.65, "learning_rate": 9.352858151023289e-05, "loss": 3.0647, "step": 8253 }, { "epoch": 0.65, "learning_rate": 9.35277973810084e-05, "loss": 3.1821, "step": 8254 }, { "epoch": 0.65, "learning_rate": 9.352701325178391e-05, "loss": 2.8074, "step": 8255 }, { "epoch": 0.65, "learning_rate": 9.352622912255941e-05, "loss": 3.1704, "step": 8256 }, { "epoch": 0.65, "learning_rate": 9.352544499333492e-05, "loss": 2.9281, "step": 8257 }, { "epoch": 0.65, "learning_rate": 9.352466086411042e-05, "loss": 2.8973, "step": 8258 }, { "epoch": 0.65, "learning_rate": 9.352387673488592e-05, "loss": 2.964, "step": 8259 }, { "epoch": 0.65, "learning_rate": 9.352309260566143e-05, "loss": 2.888, "step": 8260 }, { "epoch": 0.65, "learning_rate": 9.352230847643693e-05, "loss": 2.9012, "step": 8261 }, { "epoch": 0.65, "learning_rate": 9.352152434721243e-05, "loss": 2.9922, "step": 8262 }, { "epoch": 0.65, "learning_rate": 9.352074021798794e-05, "loss": 2.9619, "step": 8263 }, { "epoch": 0.65, "learning_rate": 9.351995608876343e-05, "loss": 2.9372, "step": 8264 }, { "epoch": 0.65, "learning_rate": 9.351917195953893e-05, "loss": 2.9073, "step": 8265 }, { "epoch": 0.65, "learning_rate": 9.351838783031443e-05, "loss": 2.7991, "step": 8266 }, { "epoch": 0.65, "learning_rate": 9.351760370108994e-05, "loss": 3.159, "step": 8267 }, { "epoch": 0.65, "learning_rate": 9.351681957186544e-05, "loss": 2.9884, "step": 8268 }, { "epoch": 0.65, "learning_rate": 9.351603544264095e-05, "loss": 3.0979, "step": 8269 }, { "epoch": 0.65, "learning_rate": 9.351525131341645e-05, "loss": 2.9714, "step": 8270 }, { "epoch": 0.65, "learning_rate": 9.351446718419195e-05, "loss": 3.3342, "step": 8271 }, { "epoch": 0.65, "learning_rate": 9.351368305496746e-05, "loss": 2.8037, "step": 8272 }, { "epoch": 0.65, "learning_rate": 9.351289892574296e-05, "loss": 2.9818, "step": 8273 }, { "epoch": 0.65, "learning_rate": 9.351211479651846e-05, "loss": 3.0475, "step": 8274 }, { "epoch": 0.65, "learning_rate": 9.351133066729397e-05, "loss": 2.9821, "step": 8275 }, { "epoch": 0.65, "learning_rate": 9.351054653806947e-05, "loss": 3.0986, "step": 8276 }, { "epoch": 0.65, "learning_rate": 9.350976240884498e-05, "loss": 3.2689, "step": 8277 }, { "epoch": 0.65, "learning_rate": 9.350897827962048e-05, "loss": 2.9202, "step": 8278 }, { "epoch": 0.65, "learning_rate": 9.350819415039598e-05, "loss": 3.1668, "step": 8279 }, { "epoch": 0.65, "learning_rate": 9.35074100211715e-05, "loss": 3.0814, "step": 8280 }, { "epoch": 0.65, "learning_rate": 9.3506625891947e-05, "loss": 2.9961, "step": 8281 }, { "epoch": 0.65, "learning_rate": 9.350584176272251e-05, "loss": 2.8815, "step": 8282 }, { "epoch": 0.65, "learning_rate": 9.350505763349801e-05, "loss": 3.0456, "step": 8283 }, { "epoch": 0.65, "learning_rate": 9.350427350427351e-05, "loss": 3.1969, "step": 8284 }, { "epoch": 0.65, "learning_rate": 9.350348937504902e-05, "loss": 2.8489, "step": 8285 }, { "epoch": 0.65, "learning_rate": 9.350270524582452e-05, "loss": 2.9147, "step": 8286 }, { "epoch": 0.65, "learning_rate": 9.350192111660003e-05, "loss": 2.8884, "step": 8287 }, { "epoch": 0.65, "learning_rate": 9.350113698737553e-05, "loss": 3.1954, "step": 8288 }, { "epoch": 0.65, "learning_rate": 9.350035285815103e-05, "loss": 3.0644, "step": 8289 }, { "epoch": 0.65, "learning_rate": 9.349956872892654e-05, "loss": 2.9282, "step": 8290 }, { "epoch": 0.65, "learning_rate": 9.349878459970204e-05, "loss": 2.8607, "step": 8291 }, { "epoch": 0.65, "learning_rate": 9.349800047047754e-05, "loss": 2.8887, "step": 8292 }, { "epoch": 0.65, "learning_rate": 9.349721634125305e-05, "loss": 3.0609, "step": 8293 }, { "epoch": 0.65, "learning_rate": 9.349643221202855e-05, "loss": 3.0115, "step": 8294 }, { "epoch": 0.65, "learning_rate": 9.349564808280405e-05, "loss": 3.1474, "step": 8295 }, { "epoch": 0.65, "learning_rate": 9.349486395357956e-05, "loss": 2.9939, "step": 8296 }, { "epoch": 0.65, "learning_rate": 9.349407982435506e-05, "loss": 2.8598, "step": 8297 }, { "epoch": 0.65, "learning_rate": 9.349329569513057e-05, "loss": 2.8924, "step": 8298 }, { "epoch": 0.65, "learning_rate": 9.349251156590607e-05, "loss": 3.2081, "step": 8299 }, { "epoch": 0.65, "learning_rate": 9.349172743668156e-05, "loss": 3.0129, "step": 8300 }, { "epoch": 0.65, "learning_rate": 9.349094330745706e-05, "loss": 3.1778, "step": 8301 }, { "epoch": 0.65, "learning_rate": 9.349015917823257e-05, "loss": 3.2909, "step": 8302 }, { "epoch": 0.65, "learning_rate": 9.348937504900807e-05, "loss": 2.8875, "step": 8303 }, { "epoch": 0.65, "learning_rate": 9.348859091978357e-05, "loss": 3.2286, "step": 8304 }, { "epoch": 0.65, "learning_rate": 9.348780679055909e-05, "loss": 3.0196, "step": 8305 }, { "epoch": 0.65, "learning_rate": 9.34870226613346e-05, "loss": 3.0146, "step": 8306 }, { "epoch": 0.65, "learning_rate": 9.34862385321101e-05, "loss": 3.0819, "step": 8307 }, { "epoch": 0.65, "learning_rate": 9.34854544028856e-05, "loss": 3.088, "step": 8308 }, { "epoch": 0.65, "learning_rate": 9.34846702736611e-05, "loss": 2.9944, "step": 8309 }, { "epoch": 0.65, "learning_rate": 9.348388614443661e-05, "loss": 2.6587, "step": 8310 }, { "epoch": 0.65, "learning_rate": 9.348310201521211e-05, "loss": 2.9971, "step": 8311 }, { "epoch": 0.65, "learning_rate": 9.348231788598762e-05, "loss": 3.0492, "step": 8312 }, { "epoch": 0.65, "learning_rate": 9.348153375676312e-05, "loss": 3.0998, "step": 8313 }, { "epoch": 0.65, "learning_rate": 9.348074962753862e-05, "loss": 2.6955, "step": 8314 }, { "epoch": 0.65, "learning_rate": 9.347996549831413e-05, "loss": 2.9282, "step": 8315 }, { "epoch": 0.65, "learning_rate": 9.347918136908963e-05, "loss": 2.9826, "step": 8316 }, { "epoch": 0.65, "learning_rate": 9.347839723986514e-05, "loss": 2.8816, "step": 8317 }, { "epoch": 0.65, "learning_rate": 9.347761311064064e-05, "loss": 2.9435, "step": 8318 }, { "epoch": 0.65, "learning_rate": 9.347682898141614e-05, "loss": 2.9606, "step": 8319 }, { "epoch": 0.65, "learning_rate": 9.347604485219165e-05, "loss": 3.1097, "step": 8320 }, { "epoch": 0.65, "learning_rate": 9.347526072296715e-05, "loss": 3.227, "step": 8321 }, { "epoch": 0.65, "learning_rate": 9.347447659374265e-05, "loss": 3.0154, "step": 8322 }, { "epoch": 0.65, "learning_rate": 9.347369246451816e-05, "loss": 2.9597, "step": 8323 }, { "epoch": 0.65, "learning_rate": 9.347290833529366e-05, "loss": 2.9954, "step": 8324 }, { "epoch": 0.65, "learning_rate": 9.347212420606917e-05, "loss": 3.0166, "step": 8325 }, { "epoch": 0.65, "learning_rate": 9.347134007684467e-05, "loss": 3.257, "step": 8326 }, { "epoch": 0.65, "learning_rate": 9.347055594762017e-05, "loss": 2.9987, "step": 8327 }, { "epoch": 0.65, "learning_rate": 9.346977181839568e-05, "loss": 3.2659, "step": 8328 }, { "epoch": 0.65, "learning_rate": 9.346898768917118e-05, "loss": 2.8205, "step": 8329 }, { "epoch": 0.65, "learning_rate": 9.346820355994668e-05, "loss": 2.9949, "step": 8330 }, { "epoch": 0.65, "learning_rate": 9.346741943072219e-05, "loss": 2.8248, "step": 8331 }, { "epoch": 0.65, "learning_rate": 9.346663530149769e-05, "loss": 3.1407, "step": 8332 }, { "epoch": 0.65, "learning_rate": 9.34658511722732e-05, "loss": 3.2124, "step": 8333 }, { "epoch": 0.65, "learning_rate": 9.34650670430487e-05, "loss": 3.0047, "step": 8334 }, { "epoch": 0.65, "learning_rate": 9.34642829138242e-05, "loss": 2.9891, "step": 8335 }, { "epoch": 0.65, "learning_rate": 9.34634987845997e-05, "loss": 2.8371, "step": 8336 }, { "epoch": 0.65, "learning_rate": 9.346271465537521e-05, "loss": 3.0853, "step": 8337 }, { "epoch": 0.65, "learning_rate": 9.346193052615071e-05, "loss": 3.477, "step": 8338 }, { "epoch": 0.65, "learning_rate": 9.346114639692622e-05, "loss": 2.9455, "step": 8339 }, { "epoch": 0.65, "learning_rate": 9.346036226770172e-05, "loss": 2.9211, "step": 8340 }, { "epoch": 0.65, "learning_rate": 9.345957813847722e-05, "loss": 3.1221, "step": 8341 }, { "epoch": 0.65, "learning_rate": 9.345879400925273e-05, "loss": 3.0852, "step": 8342 }, { "epoch": 0.65, "learning_rate": 9.345800988002823e-05, "loss": 2.8882, "step": 8343 }, { "epoch": 0.65, "learning_rate": 9.345722575080374e-05, "loss": 3.1745, "step": 8344 }, { "epoch": 0.65, "learning_rate": 9.345644162157924e-05, "loss": 3.1091, "step": 8345 }, { "epoch": 0.65, "learning_rate": 9.345565749235474e-05, "loss": 2.8531, "step": 8346 }, { "epoch": 0.65, "learning_rate": 9.345487336313025e-05, "loss": 2.8384, "step": 8347 }, { "epoch": 0.65, "learning_rate": 9.345408923390575e-05, "loss": 2.8824, "step": 8348 }, { "epoch": 0.65, "learning_rate": 9.345330510468125e-05, "loss": 2.9316, "step": 8349 }, { "epoch": 0.65, "learning_rate": 9.345252097545676e-05, "loss": 3.0569, "step": 8350 }, { "epoch": 0.65, "learning_rate": 9.345173684623226e-05, "loss": 2.9915, "step": 8351 }, { "epoch": 0.65, "learning_rate": 9.345095271700776e-05, "loss": 2.9651, "step": 8352 }, { "epoch": 0.65, "learning_rate": 9.345016858778327e-05, "loss": 2.9786, "step": 8353 }, { "epoch": 0.66, "learning_rate": 9.344938445855877e-05, "loss": 3.1488, "step": 8354 }, { "epoch": 0.66, "learning_rate": 9.344860032933428e-05, "loss": 3.2091, "step": 8355 }, { "epoch": 0.66, "learning_rate": 9.344781620010978e-05, "loss": 3.0086, "step": 8356 }, { "epoch": 0.66, "learning_rate": 9.344703207088528e-05, "loss": 3.2272, "step": 8357 }, { "epoch": 0.66, "learning_rate": 9.344624794166079e-05, "loss": 3.1816, "step": 8358 }, { "epoch": 0.66, "learning_rate": 9.344546381243629e-05, "loss": 3.001, "step": 8359 }, { "epoch": 0.66, "learning_rate": 9.34446796832118e-05, "loss": 2.844, "step": 8360 }, { "epoch": 0.66, "learning_rate": 9.34438955539873e-05, "loss": 2.913, "step": 8361 }, { "epoch": 0.66, "learning_rate": 9.344311142476281e-05, "loss": 2.9301, "step": 8362 }, { "epoch": 0.66, "learning_rate": 9.344232729553832e-05, "loss": 2.8284, "step": 8363 }, { "epoch": 0.66, "learning_rate": 9.344154316631382e-05, "loss": 3.1121, "step": 8364 }, { "epoch": 0.66, "learning_rate": 9.344075903708933e-05, "loss": 3.4696, "step": 8365 }, { "epoch": 0.66, "learning_rate": 9.343997490786483e-05, "loss": 2.8719, "step": 8366 }, { "epoch": 0.66, "learning_rate": 9.343919077864033e-05, "loss": 3.039, "step": 8367 }, { "epoch": 0.66, "learning_rate": 9.343840664941584e-05, "loss": 2.8198, "step": 8368 }, { "epoch": 0.66, "learning_rate": 9.343762252019134e-05, "loss": 3.1527, "step": 8369 }, { "epoch": 0.66, "learning_rate": 9.343683839096684e-05, "loss": 2.8126, "step": 8370 }, { "epoch": 0.66, "learning_rate": 9.343605426174235e-05, "loss": 3.3712, "step": 8371 }, { "epoch": 0.66, "learning_rate": 9.343527013251784e-05, "loss": 3.1151, "step": 8372 }, { "epoch": 0.66, "learning_rate": 9.343448600329334e-05, "loss": 3.1888, "step": 8373 }, { "epoch": 0.66, "learning_rate": 9.343370187406885e-05, "loss": 2.9619, "step": 8374 }, { "epoch": 0.66, "learning_rate": 9.343291774484435e-05, "loss": 3.0746, "step": 8375 }, { "epoch": 0.66, "learning_rate": 9.343213361561985e-05, "loss": 3.1687, "step": 8376 }, { "epoch": 0.66, "learning_rate": 9.343134948639536e-05, "loss": 3.179, "step": 8377 }, { "epoch": 0.66, "learning_rate": 9.343056535717086e-05, "loss": 3.311, "step": 8378 }, { "epoch": 0.66, "learning_rate": 9.342978122794636e-05, "loss": 3.086, "step": 8379 }, { "epoch": 0.66, "learning_rate": 9.342899709872187e-05, "loss": 2.7949, "step": 8380 }, { "epoch": 0.66, "learning_rate": 9.342821296949737e-05, "loss": 2.9361, "step": 8381 }, { "epoch": 0.66, "learning_rate": 9.342742884027287e-05, "loss": 3.1246, "step": 8382 }, { "epoch": 0.66, "learning_rate": 9.342664471104838e-05, "loss": 3.1337, "step": 8383 }, { "epoch": 0.66, "learning_rate": 9.342586058182388e-05, "loss": 2.9005, "step": 8384 }, { "epoch": 0.66, "learning_rate": 9.342507645259939e-05, "loss": 3.3043, "step": 8385 }, { "epoch": 0.66, "learning_rate": 9.342429232337489e-05, "loss": 2.8789, "step": 8386 }, { "epoch": 0.66, "learning_rate": 9.342350819415039e-05, "loss": 2.9675, "step": 8387 }, { "epoch": 0.66, "learning_rate": 9.342272406492591e-05, "loss": 3.0041, "step": 8388 }, { "epoch": 0.66, "learning_rate": 9.342193993570141e-05, "loss": 3.2329, "step": 8389 }, { "epoch": 0.66, "learning_rate": 9.342115580647692e-05, "loss": 2.9057, "step": 8390 }, { "epoch": 0.66, "learning_rate": 9.342037167725242e-05, "loss": 2.9894, "step": 8391 }, { "epoch": 0.66, "learning_rate": 9.341958754802793e-05, "loss": 3.1993, "step": 8392 }, { "epoch": 0.66, "learning_rate": 9.341880341880343e-05, "loss": 3.0653, "step": 8393 }, { "epoch": 0.66, "learning_rate": 9.341801928957893e-05, "loss": 3.1846, "step": 8394 }, { "epoch": 0.66, "learning_rate": 9.341723516035444e-05, "loss": 3.1175, "step": 8395 }, { "epoch": 0.66, "learning_rate": 9.341645103112994e-05, "loss": 2.7435, "step": 8396 }, { "epoch": 0.66, "learning_rate": 9.341566690190544e-05, "loss": 3.493, "step": 8397 }, { "epoch": 0.66, "learning_rate": 9.341488277268095e-05, "loss": 3.3924, "step": 8398 }, { "epoch": 0.66, "learning_rate": 9.341409864345645e-05, "loss": 2.8093, "step": 8399 }, { "epoch": 0.66, "learning_rate": 9.341331451423195e-05, "loss": 3.0125, "step": 8400 }, { "epoch": 0.66, "learning_rate": 9.341253038500746e-05, "loss": 3.0141, "step": 8401 }, { "epoch": 0.66, "learning_rate": 9.341174625578296e-05, "loss": 2.8792, "step": 8402 }, { "epoch": 0.66, "learning_rate": 9.341096212655847e-05, "loss": 2.8427, "step": 8403 }, { "epoch": 0.66, "learning_rate": 9.341017799733397e-05, "loss": 2.7589, "step": 8404 }, { "epoch": 0.66, "learning_rate": 9.340939386810947e-05, "loss": 3.1096, "step": 8405 }, { "epoch": 0.66, "learning_rate": 9.340860973888498e-05, "loss": 2.9106, "step": 8406 }, { "epoch": 0.66, "learning_rate": 9.340782560966048e-05, "loss": 2.9596, "step": 8407 }, { "epoch": 0.66, "learning_rate": 9.340704148043597e-05, "loss": 2.7638, "step": 8408 }, { "epoch": 0.66, "learning_rate": 9.340625735121147e-05, "loss": 3.0191, "step": 8409 }, { "epoch": 0.66, "learning_rate": 9.340547322198698e-05, "loss": 3.0057, "step": 8410 }, { "epoch": 0.66, "learning_rate": 9.340468909276248e-05, "loss": 3.0782, "step": 8411 }, { "epoch": 0.66, "learning_rate": 9.340390496353799e-05, "loss": 3.0046, "step": 8412 }, { "epoch": 0.66, "learning_rate": 9.340312083431349e-05, "loss": 3.1085, "step": 8413 }, { "epoch": 0.66, "learning_rate": 9.3402336705089e-05, "loss": 2.9798, "step": 8414 }, { "epoch": 0.66, "learning_rate": 9.340155257586451e-05, "loss": 2.889, "step": 8415 }, { "epoch": 0.66, "learning_rate": 9.340076844664001e-05, "loss": 3.1416, "step": 8416 }, { "epoch": 0.66, "learning_rate": 9.339998431741552e-05, "loss": 2.9643, "step": 8417 }, { "epoch": 0.66, "learning_rate": 9.339920018819102e-05, "loss": 3.2711, "step": 8418 }, { "epoch": 0.66, "learning_rate": 9.339841605896652e-05, "loss": 3.15, "step": 8419 }, { "epoch": 0.66, "learning_rate": 9.339763192974203e-05, "loss": 3.1091, "step": 8420 }, { "epoch": 0.66, "learning_rate": 9.339684780051753e-05, "loss": 3.2858, "step": 8421 }, { "epoch": 0.66, "learning_rate": 9.339606367129304e-05, "loss": 3.1573, "step": 8422 }, { "epoch": 0.66, "learning_rate": 9.339527954206854e-05, "loss": 2.8675, "step": 8423 }, { "epoch": 0.66, "learning_rate": 9.339449541284404e-05, "loss": 2.9293, "step": 8424 }, { "epoch": 0.66, "learning_rate": 9.339371128361955e-05, "loss": 2.7688, "step": 8425 }, { "epoch": 0.66, "learning_rate": 9.339292715439505e-05, "loss": 3.0156, "step": 8426 }, { "epoch": 0.66, "learning_rate": 9.339214302517055e-05, "loss": 3.0226, "step": 8427 }, { "epoch": 0.66, "learning_rate": 9.339135889594606e-05, "loss": 3.1638, "step": 8428 }, { "epoch": 0.66, "learning_rate": 9.339057476672156e-05, "loss": 3.3079, "step": 8429 }, { "epoch": 0.66, "learning_rate": 9.338979063749706e-05, "loss": 3.0701, "step": 8430 }, { "epoch": 0.66, "learning_rate": 9.338900650827257e-05, "loss": 2.8397, "step": 8431 }, { "epoch": 0.66, "learning_rate": 9.338822237904807e-05, "loss": 3.0937, "step": 8432 }, { "epoch": 0.66, "learning_rate": 9.338743824982358e-05, "loss": 3.0129, "step": 8433 }, { "epoch": 0.66, "learning_rate": 9.338665412059908e-05, "loss": 3.0951, "step": 8434 }, { "epoch": 0.66, "learning_rate": 9.338586999137458e-05, "loss": 3.0922, "step": 8435 }, { "epoch": 0.66, "learning_rate": 9.338508586215009e-05, "loss": 2.956, "step": 8436 }, { "epoch": 0.66, "learning_rate": 9.338430173292559e-05, "loss": 3.1616, "step": 8437 }, { "epoch": 0.66, "learning_rate": 9.33835176037011e-05, "loss": 3.0334, "step": 8438 }, { "epoch": 0.66, "learning_rate": 9.33827334744766e-05, "loss": 2.8696, "step": 8439 }, { "epoch": 0.66, "learning_rate": 9.33819493452521e-05, "loss": 3.2865, "step": 8440 }, { "epoch": 0.66, "learning_rate": 9.33811652160276e-05, "loss": 3.0711, "step": 8441 }, { "epoch": 0.66, "learning_rate": 9.338038108680311e-05, "loss": 3.1742, "step": 8442 }, { "epoch": 0.66, "learning_rate": 9.337959695757861e-05, "loss": 3.051, "step": 8443 }, { "epoch": 0.66, "learning_rate": 9.337881282835412e-05, "loss": 2.9927, "step": 8444 }, { "epoch": 0.66, "learning_rate": 9.337802869912962e-05, "loss": 3.2108, "step": 8445 }, { "epoch": 0.66, "learning_rate": 9.337724456990512e-05, "loss": 3.1128, "step": 8446 }, { "epoch": 0.66, "learning_rate": 9.337646044068063e-05, "loss": 2.9817, "step": 8447 }, { "epoch": 0.66, "learning_rate": 9.337567631145613e-05, "loss": 3.075, "step": 8448 }, { "epoch": 0.66, "learning_rate": 9.337489218223163e-05, "loss": 2.949, "step": 8449 }, { "epoch": 0.66, "learning_rate": 9.337410805300714e-05, "loss": 2.899, "step": 8450 }, { "epoch": 0.66, "learning_rate": 9.337332392378264e-05, "loss": 3.0794, "step": 8451 }, { "epoch": 0.66, "learning_rate": 9.337253979455815e-05, "loss": 3.0852, "step": 8452 }, { "epoch": 0.66, "learning_rate": 9.337175566533365e-05, "loss": 3.0601, "step": 8453 }, { "epoch": 0.66, "learning_rate": 9.337097153610915e-05, "loss": 2.9349, "step": 8454 }, { "epoch": 0.66, "learning_rate": 9.337018740688466e-05, "loss": 2.887, "step": 8455 }, { "epoch": 0.66, "learning_rate": 9.336940327766016e-05, "loss": 2.9006, "step": 8456 }, { "epoch": 0.66, "learning_rate": 9.336861914843566e-05, "loss": 2.9948, "step": 8457 }, { "epoch": 0.66, "learning_rate": 9.336783501921117e-05, "loss": 3.1413, "step": 8458 }, { "epoch": 0.66, "learning_rate": 9.336705088998667e-05, "loss": 3.287, "step": 8459 }, { "epoch": 0.66, "learning_rate": 9.336626676076218e-05, "loss": 3.041, "step": 8460 }, { "epoch": 0.66, "learning_rate": 9.336548263153768e-05, "loss": 2.8661, "step": 8461 }, { "epoch": 0.66, "learning_rate": 9.336469850231318e-05, "loss": 2.7958, "step": 8462 }, { "epoch": 0.66, "learning_rate": 9.336391437308869e-05, "loss": 2.9553, "step": 8463 }, { "epoch": 0.66, "learning_rate": 9.336313024386419e-05, "loss": 2.9874, "step": 8464 }, { "epoch": 0.66, "learning_rate": 9.33623461146397e-05, "loss": 3.2517, "step": 8465 }, { "epoch": 0.66, "learning_rate": 9.33615619854152e-05, "loss": 3.0541, "step": 8466 }, { "epoch": 0.66, "learning_rate": 9.33607778561907e-05, "loss": 2.9625, "step": 8467 }, { "epoch": 0.66, "learning_rate": 9.33599937269662e-05, "loss": 3.1663, "step": 8468 }, { "epoch": 0.66, "learning_rate": 9.335920959774171e-05, "loss": 2.9336, "step": 8469 }, { "epoch": 0.66, "learning_rate": 9.335842546851723e-05, "loss": 2.9295, "step": 8470 }, { "epoch": 0.66, "learning_rate": 9.335764133929273e-05, "loss": 3.0989, "step": 8471 }, { "epoch": 0.66, "learning_rate": 9.335685721006823e-05, "loss": 2.9738, "step": 8472 }, { "epoch": 0.66, "learning_rate": 9.335607308084374e-05, "loss": 3.1335, "step": 8473 }, { "epoch": 0.66, "learning_rate": 9.335528895161924e-05, "loss": 3.1052, "step": 8474 }, { "epoch": 0.66, "learning_rate": 9.335450482239474e-05, "loss": 2.9285, "step": 8475 }, { "epoch": 0.66, "learning_rate": 9.335372069317025e-05, "loss": 3.0461, "step": 8476 }, { "epoch": 0.66, "learning_rate": 9.335293656394575e-05, "loss": 3.2319, "step": 8477 }, { "epoch": 0.66, "learning_rate": 9.335215243472125e-05, "loss": 2.8658, "step": 8478 }, { "epoch": 0.66, "learning_rate": 9.335136830549676e-05, "loss": 2.9706, "step": 8479 }, { "epoch": 0.66, "learning_rate": 9.335058417627225e-05, "loss": 2.8324, "step": 8480 }, { "epoch": 0.67, "learning_rate": 9.334980004704775e-05, "loss": 3.0966, "step": 8481 }, { "epoch": 0.67, "learning_rate": 9.334901591782326e-05, "loss": 3.1635, "step": 8482 }, { "epoch": 0.67, "learning_rate": 9.334823178859876e-05, "loss": 3.1251, "step": 8483 }, { "epoch": 0.67, "learning_rate": 9.334744765937426e-05, "loss": 2.9985, "step": 8484 }, { "epoch": 0.67, "learning_rate": 9.334666353014977e-05, "loss": 2.9032, "step": 8485 }, { "epoch": 0.67, "learning_rate": 9.334587940092527e-05, "loss": 2.9605, "step": 8486 }, { "epoch": 0.67, "learning_rate": 9.334509527170077e-05, "loss": 3.0021, "step": 8487 }, { "epoch": 0.67, "learning_rate": 9.334431114247628e-05, "loss": 3.1727, "step": 8488 }, { "epoch": 0.67, "learning_rate": 9.334352701325178e-05, "loss": 3.0453, "step": 8489 }, { "epoch": 0.67, "learning_rate": 9.334274288402729e-05, "loss": 3.2383, "step": 8490 }, { "epoch": 0.67, "learning_rate": 9.334195875480279e-05, "loss": 2.9531, "step": 8491 }, { "epoch": 0.67, "learning_rate": 9.334117462557829e-05, "loss": 2.7564, "step": 8492 }, { "epoch": 0.67, "learning_rate": 9.33403904963538e-05, "loss": 3.2039, "step": 8493 }, { "epoch": 0.67, "learning_rate": 9.33396063671293e-05, "loss": 2.9606, "step": 8494 }, { "epoch": 0.67, "learning_rate": 9.33388222379048e-05, "loss": 3.0098, "step": 8495 }, { "epoch": 0.67, "learning_rate": 9.333803810868032e-05, "loss": 3.0442, "step": 8496 }, { "epoch": 0.67, "learning_rate": 9.333725397945582e-05, "loss": 3.1079, "step": 8497 }, { "epoch": 0.67, "learning_rate": 9.333646985023133e-05, "loss": 2.9673, "step": 8498 }, { "epoch": 0.67, "learning_rate": 9.333568572100683e-05, "loss": 2.9927, "step": 8499 }, { "epoch": 0.67, "learning_rate": 9.333490159178234e-05, "loss": 3.0813, "step": 8500 }, { "epoch": 0.67, "learning_rate": 9.333411746255784e-05, "loss": 3.0203, "step": 8501 }, { "epoch": 0.67, "learning_rate": 9.333333333333334e-05, "loss": 2.9692, "step": 8502 }, { "epoch": 0.67, "learning_rate": 9.333254920410885e-05, "loss": 3.0403, "step": 8503 }, { "epoch": 0.67, "learning_rate": 9.333176507488435e-05, "loss": 2.9369, "step": 8504 }, { "epoch": 0.67, "learning_rate": 9.333098094565985e-05, "loss": 3.1185, "step": 8505 }, { "epoch": 0.67, "learning_rate": 9.333019681643536e-05, "loss": 2.9251, "step": 8506 }, { "epoch": 0.67, "learning_rate": 9.332941268721086e-05, "loss": 3.001, "step": 8507 }, { "epoch": 0.67, "learning_rate": 9.332862855798637e-05, "loss": 3.1568, "step": 8508 }, { "epoch": 0.67, "learning_rate": 9.332784442876187e-05, "loss": 3.0516, "step": 8509 }, { "epoch": 0.67, "learning_rate": 9.332706029953737e-05, "loss": 3.0887, "step": 8510 }, { "epoch": 0.67, "learning_rate": 9.332627617031288e-05, "loss": 2.8939, "step": 8511 }, { "epoch": 0.67, "learning_rate": 9.332549204108838e-05, "loss": 3.0197, "step": 8512 }, { "epoch": 0.67, "learning_rate": 9.332470791186388e-05, "loss": 2.916, "step": 8513 }, { "epoch": 0.67, "learning_rate": 9.332392378263939e-05, "loss": 2.8838, "step": 8514 }, { "epoch": 0.67, "learning_rate": 9.332313965341489e-05, "loss": 3.0768, "step": 8515 }, { "epoch": 0.67, "learning_rate": 9.332235552419038e-05, "loss": 2.9561, "step": 8516 }, { "epoch": 0.67, "learning_rate": 9.332157139496588e-05, "loss": 2.9531, "step": 8517 }, { "epoch": 0.67, "learning_rate": 9.332078726574139e-05, "loss": 2.8423, "step": 8518 }, { "epoch": 0.67, "learning_rate": 9.332000313651689e-05, "loss": 2.9985, "step": 8519 }, { "epoch": 0.67, "learning_rate": 9.33192190072924e-05, "loss": 2.8776, "step": 8520 }, { "epoch": 0.67, "learning_rate": 9.33184348780679e-05, "loss": 2.8916, "step": 8521 }, { "epoch": 0.67, "learning_rate": 9.331765074884342e-05, "loss": 3.1146, "step": 8522 }, { "epoch": 0.67, "learning_rate": 9.331686661961892e-05, "loss": 2.9187, "step": 8523 }, { "epoch": 0.67, "learning_rate": 9.331608249039442e-05, "loss": 3.0188, "step": 8524 }, { "epoch": 0.67, "learning_rate": 9.331529836116993e-05, "loss": 2.9249, "step": 8525 }, { "epoch": 0.67, "learning_rate": 9.331451423194543e-05, "loss": 3.1028, "step": 8526 }, { "epoch": 0.67, "learning_rate": 9.331373010272094e-05, "loss": 3.022, "step": 8527 }, { "epoch": 0.67, "learning_rate": 9.331294597349644e-05, "loss": 3.2641, "step": 8528 }, { "epoch": 0.67, "learning_rate": 9.331216184427194e-05, "loss": 2.8918, "step": 8529 }, { "epoch": 0.67, "learning_rate": 9.331137771504745e-05, "loss": 3.0773, "step": 8530 }, { "epoch": 0.67, "learning_rate": 9.331059358582295e-05, "loss": 2.9968, "step": 8531 }, { "epoch": 0.67, "learning_rate": 9.330980945659845e-05, "loss": 2.9383, "step": 8532 }, { "epoch": 0.67, "learning_rate": 9.330902532737396e-05, "loss": 3.1203, "step": 8533 }, { "epoch": 0.67, "learning_rate": 9.330824119814946e-05, "loss": 2.9067, "step": 8534 }, { "epoch": 0.67, "learning_rate": 9.330745706892496e-05, "loss": 3.0113, "step": 8535 }, { "epoch": 0.67, "learning_rate": 9.330667293970047e-05, "loss": 3.0503, "step": 8536 }, { "epoch": 0.67, "learning_rate": 9.330588881047597e-05, "loss": 3.074, "step": 8537 }, { "epoch": 0.67, "learning_rate": 9.330510468125148e-05, "loss": 2.9037, "step": 8538 }, { "epoch": 0.67, "learning_rate": 9.330432055202698e-05, "loss": 2.9592, "step": 8539 }, { "epoch": 0.67, "learning_rate": 9.330353642280248e-05, "loss": 3.1013, "step": 8540 }, { "epoch": 0.67, "learning_rate": 9.330275229357799e-05, "loss": 2.9754, "step": 8541 }, { "epoch": 0.67, "learning_rate": 9.330196816435349e-05, "loss": 2.8379, "step": 8542 }, { "epoch": 0.67, "learning_rate": 9.3301184035129e-05, "loss": 3.3085, "step": 8543 }, { "epoch": 0.67, "learning_rate": 9.33003999059045e-05, "loss": 2.883, "step": 8544 }, { "epoch": 0.67, "learning_rate": 9.329961577668e-05, "loss": 3.145, "step": 8545 }, { "epoch": 0.67, "learning_rate": 9.32988316474555e-05, "loss": 3.1294, "step": 8546 }, { "epoch": 0.67, "learning_rate": 9.329804751823101e-05, "loss": 2.9753, "step": 8547 }, { "epoch": 0.67, "learning_rate": 9.329726338900651e-05, "loss": 3.0989, "step": 8548 }, { "epoch": 0.67, "learning_rate": 9.329647925978202e-05, "loss": 2.9174, "step": 8549 }, { "epoch": 0.67, "learning_rate": 9.329569513055752e-05, "loss": 3.1258, "step": 8550 }, { "epoch": 0.67, "learning_rate": 9.329491100133302e-05, "loss": 2.9852, "step": 8551 }, { "epoch": 0.67, "learning_rate": 9.329412687210853e-05, "loss": 3.0086, "step": 8552 }, { "epoch": 0.67, "learning_rate": 9.329334274288403e-05, "loss": 3.0037, "step": 8553 }, { "epoch": 0.67, "learning_rate": 9.329255861365953e-05, "loss": 3.0867, "step": 8554 }, { "epoch": 0.67, "learning_rate": 9.329177448443504e-05, "loss": 3.1646, "step": 8555 }, { "epoch": 0.67, "learning_rate": 9.329099035521054e-05, "loss": 2.8936, "step": 8556 }, { "epoch": 0.67, "learning_rate": 9.329020622598605e-05, "loss": 3.2477, "step": 8557 }, { "epoch": 0.67, "learning_rate": 9.328942209676155e-05, "loss": 3.0236, "step": 8558 }, { "epoch": 0.67, "learning_rate": 9.328863796753705e-05, "loss": 3.127, "step": 8559 }, { "epoch": 0.67, "learning_rate": 9.328785383831256e-05, "loss": 3.1092, "step": 8560 }, { "epoch": 0.67, "learning_rate": 9.328706970908806e-05, "loss": 2.8165, "step": 8561 }, { "epoch": 0.67, "learning_rate": 9.328628557986356e-05, "loss": 3.1545, "step": 8562 }, { "epoch": 0.67, "learning_rate": 9.328550145063907e-05, "loss": 3.0192, "step": 8563 }, { "epoch": 0.67, "learning_rate": 9.328471732141457e-05, "loss": 3.1385, "step": 8564 }, { "epoch": 0.67, "learning_rate": 9.328393319219007e-05, "loss": 2.9868, "step": 8565 }, { "epoch": 0.67, "learning_rate": 9.328314906296558e-05, "loss": 3.2523, "step": 8566 }, { "epoch": 0.67, "learning_rate": 9.328236493374108e-05, "loss": 3.0648, "step": 8567 }, { "epoch": 0.67, "learning_rate": 9.328158080451659e-05, "loss": 2.9706, "step": 8568 }, { "epoch": 0.67, "learning_rate": 9.328079667529209e-05, "loss": 2.994, "step": 8569 }, { "epoch": 0.67, "learning_rate": 9.32800125460676e-05, "loss": 3.1784, "step": 8570 }, { "epoch": 0.67, "learning_rate": 9.32792284168431e-05, "loss": 3.0493, "step": 8571 }, { "epoch": 0.67, "learning_rate": 9.32784442876186e-05, "loss": 3.1883, "step": 8572 }, { "epoch": 0.67, "learning_rate": 9.32776601583941e-05, "loss": 3.0069, "step": 8573 }, { "epoch": 0.67, "learning_rate": 9.327687602916961e-05, "loss": 2.8955, "step": 8574 }, { "epoch": 0.67, "learning_rate": 9.327609189994511e-05, "loss": 3.003, "step": 8575 }, { "epoch": 0.67, "learning_rate": 9.327530777072062e-05, "loss": 3.1368, "step": 8576 }, { "epoch": 0.67, "learning_rate": 9.327452364149612e-05, "loss": 3.0998, "step": 8577 }, { "epoch": 0.67, "learning_rate": 9.327373951227164e-05, "loss": 2.9678, "step": 8578 }, { "epoch": 0.67, "learning_rate": 9.327295538304714e-05, "loss": 3.0026, "step": 8579 }, { "epoch": 0.67, "learning_rate": 9.327217125382264e-05, "loss": 2.8638, "step": 8580 }, { "epoch": 0.67, "learning_rate": 9.327138712459815e-05, "loss": 3.4243, "step": 8581 }, { "epoch": 0.67, "learning_rate": 9.327060299537365e-05, "loss": 3.1735, "step": 8582 }, { "epoch": 0.67, "learning_rate": 9.326981886614915e-05, "loss": 2.93, "step": 8583 }, { "epoch": 0.67, "learning_rate": 9.326903473692466e-05, "loss": 2.9931, "step": 8584 }, { "epoch": 0.67, "learning_rate": 9.326825060770016e-05, "loss": 3.2354, "step": 8585 }, { "epoch": 0.67, "learning_rate": 9.326746647847567e-05, "loss": 2.844, "step": 8586 }, { "epoch": 0.67, "learning_rate": 9.326668234925117e-05, "loss": 2.973, "step": 8587 }, { "epoch": 0.67, "learning_rate": 9.326589822002666e-05, "loss": 2.9275, "step": 8588 }, { "epoch": 0.67, "learning_rate": 9.326511409080216e-05, "loss": 2.9874, "step": 8589 }, { "epoch": 0.67, "learning_rate": 9.326432996157767e-05, "loss": 2.8692, "step": 8590 }, { "epoch": 0.67, "learning_rate": 9.326354583235317e-05, "loss": 2.7948, "step": 8591 }, { "epoch": 0.67, "learning_rate": 9.326276170312867e-05, "loss": 3.1181, "step": 8592 }, { "epoch": 0.67, "learning_rate": 9.326197757390418e-05, "loss": 3.0576, "step": 8593 }, { "epoch": 0.67, "learning_rate": 9.326119344467968e-05, "loss": 3.0712, "step": 8594 }, { "epoch": 0.67, "learning_rate": 9.326040931545519e-05, "loss": 3.0105, "step": 8595 }, { "epoch": 0.67, "learning_rate": 9.325962518623069e-05, "loss": 3.3568, "step": 8596 }, { "epoch": 0.67, "learning_rate": 9.325884105700619e-05, "loss": 2.9313, "step": 8597 }, { "epoch": 0.67, "learning_rate": 9.32580569277817e-05, "loss": 3.287, "step": 8598 }, { "epoch": 0.67, "learning_rate": 9.32572727985572e-05, "loss": 3.0171, "step": 8599 }, { "epoch": 0.67, "learning_rate": 9.32564886693327e-05, "loss": 2.8626, "step": 8600 }, { "epoch": 0.67, "learning_rate": 9.325570454010821e-05, "loss": 2.9411, "step": 8601 }, { "epoch": 0.67, "learning_rate": 9.325492041088371e-05, "loss": 3.0223, "step": 8602 }, { "epoch": 0.67, "learning_rate": 9.325413628165921e-05, "loss": 3.0718, "step": 8603 }, { "epoch": 0.67, "learning_rate": 9.325335215243473e-05, "loss": 3.2001, "step": 8604 }, { "epoch": 0.67, "learning_rate": 9.325256802321024e-05, "loss": 3.1726, "step": 8605 }, { "epoch": 0.67, "learning_rate": 9.325178389398574e-05, "loss": 3.3441, "step": 8606 }, { "epoch": 0.67, "learning_rate": 9.325099976476124e-05, "loss": 2.9364, "step": 8607 }, { "epoch": 0.67, "learning_rate": 9.325021563553675e-05, "loss": 2.9337, "step": 8608 }, { "epoch": 0.68, "learning_rate": 9.324943150631225e-05, "loss": 3.0053, "step": 8609 }, { "epoch": 0.68, "learning_rate": 9.324864737708775e-05, "loss": 2.8651, "step": 8610 }, { "epoch": 0.68, "learning_rate": 9.324786324786326e-05, "loss": 3.0233, "step": 8611 }, { "epoch": 0.68, "learning_rate": 9.324707911863876e-05, "loss": 3.1281, "step": 8612 }, { "epoch": 0.68, "learning_rate": 9.324629498941427e-05, "loss": 3.0912, "step": 8613 }, { "epoch": 0.68, "learning_rate": 9.324551086018977e-05, "loss": 2.7812, "step": 8614 }, { "epoch": 0.68, "learning_rate": 9.324472673096527e-05, "loss": 2.896, "step": 8615 }, { "epoch": 0.68, "learning_rate": 9.324394260174078e-05, "loss": 2.8151, "step": 8616 }, { "epoch": 0.68, "learning_rate": 9.324315847251628e-05, "loss": 2.759, "step": 8617 }, { "epoch": 0.68, "learning_rate": 9.324237434329178e-05, "loss": 2.9344, "step": 8618 }, { "epoch": 0.68, "learning_rate": 9.324159021406729e-05, "loss": 3.0091, "step": 8619 }, { "epoch": 0.68, "learning_rate": 9.324080608484279e-05, "loss": 3.2435, "step": 8620 }, { "epoch": 0.68, "learning_rate": 9.32400219556183e-05, "loss": 3.1552, "step": 8621 }, { "epoch": 0.68, "learning_rate": 9.32392378263938e-05, "loss": 3.0696, "step": 8622 }, { "epoch": 0.68, "learning_rate": 9.32384536971693e-05, "loss": 3.235, "step": 8623 }, { "epoch": 0.68, "learning_rate": 9.323766956794479e-05, "loss": 3.2511, "step": 8624 }, { "epoch": 0.68, "learning_rate": 9.32368854387203e-05, "loss": 2.8544, "step": 8625 }, { "epoch": 0.68, "learning_rate": 9.32361013094958e-05, "loss": 3.0542, "step": 8626 }, { "epoch": 0.68, "learning_rate": 9.32353171802713e-05, "loss": 3.1267, "step": 8627 }, { "epoch": 0.68, "learning_rate": 9.32345330510468e-05, "loss": 3.209, "step": 8628 }, { "epoch": 0.68, "learning_rate": 9.323374892182231e-05, "loss": 3.2669, "step": 8629 }, { "epoch": 0.68, "learning_rate": 9.323296479259783e-05, "loss": 2.9745, "step": 8630 }, { "epoch": 0.68, "learning_rate": 9.323218066337333e-05, "loss": 3.0419, "step": 8631 }, { "epoch": 0.68, "learning_rate": 9.323139653414883e-05, "loss": 3.1685, "step": 8632 }, { "epoch": 0.68, "learning_rate": 9.323061240492434e-05, "loss": 3.0804, "step": 8633 }, { "epoch": 0.68, "learning_rate": 9.322982827569984e-05, "loss": 3.3202, "step": 8634 }, { "epoch": 0.68, "learning_rate": 9.322904414647535e-05, "loss": 3.0295, "step": 8635 }, { "epoch": 0.68, "learning_rate": 9.322826001725085e-05, "loss": 3.0243, "step": 8636 }, { "epoch": 0.68, "learning_rate": 9.322747588802635e-05, "loss": 2.9113, "step": 8637 }, { "epoch": 0.68, "learning_rate": 9.322669175880186e-05, "loss": 3.0458, "step": 8638 }, { "epoch": 0.68, "learning_rate": 9.322590762957736e-05, "loss": 3.1412, "step": 8639 }, { "epoch": 0.68, "learning_rate": 9.322512350035286e-05, "loss": 3.0232, "step": 8640 }, { "epoch": 0.68, "learning_rate": 9.322433937112837e-05, "loss": 2.9266, "step": 8641 }, { "epoch": 0.68, "learning_rate": 9.322355524190387e-05, "loss": 2.8454, "step": 8642 }, { "epoch": 0.68, "learning_rate": 9.322277111267938e-05, "loss": 3.1249, "step": 8643 }, { "epoch": 0.68, "learning_rate": 9.322198698345488e-05, "loss": 3.0944, "step": 8644 }, { "epoch": 0.68, "learning_rate": 9.322120285423038e-05, "loss": 2.8983, "step": 8645 }, { "epoch": 0.68, "learning_rate": 9.322041872500589e-05, "loss": 3.0253, "step": 8646 }, { "epoch": 0.68, "learning_rate": 9.321963459578139e-05, "loss": 3.3052, "step": 8647 }, { "epoch": 0.68, "learning_rate": 9.32188504665569e-05, "loss": 3.303, "step": 8648 }, { "epoch": 0.68, "learning_rate": 9.32180663373324e-05, "loss": 3.3188, "step": 8649 }, { "epoch": 0.68, "learning_rate": 9.32172822081079e-05, "loss": 3.0582, "step": 8650 }, { "epoch": 0.68, "learning_rate": 9.32164980788834e-05, "loss": 2.9145, "step": 8651 }, { "epoch": 0.68, "learning_rate": 9.321571394965891e-05, "loss": 3.0077, "step": 8652 }, { "epoch": 0.68, "learning_rate": 9.321492982043441e-05, "loss": 3.2177, "step": 8653 }, { "epoch": 0.68, "learning_rate": 9.321414569120992e-05, "loss": 2.8338, "step": 8654 }, { "epoch": 0.68, "learning_rate": 9.321336156198542e-05, "loss": 3.041, "step": 8655 }, { "epoch": 0.68, "learning_rate": 9.321257743276092e-05, "loss": 3.0062, "step": 8656 }, { "epoch": 0.68, "learning_rate": 9.321179330353643e-05, "loss": 2.8065, "step": 8657 }, { "epoch": 0.68, "learning_rate": 9.321100917431193e-05, "loss": 3.0776, "step": 8658 }, { "epoch": 0.68, "learning_rate": 9.321022504508743e-05, "loss": 3.1552, "step": 8659 }, { "epoch": 0.68, "learning_rate": 9.320944091586294e-05, "loss": 2.9373, "step": 8660 }, { "epoch": 0.68, "learning_rate": 9.320865678663844e-05, "loss": 3.1746, "step": 8661 }, { "epoch": 0.68, "learning_rate": 9.320787265741395e-05, "loss": 3.1754, "step": 8662 }, { "epoch": 0.68, "learning_rate": 9.320708852818945e-05, "loss": 3.3476, "step": 8663 }, { "epoch": 0.68, "learning_rate": 9.320630439896495e-05, "loss": 2.878, "step": 8664 }, { "epoch": 0.68, "learning_rate": 9.320552026974046e-05, "loss": 3.1926, "step": 8665 }, { "epoch": 0.68, "learning_rate": 9.320473614051596e-05, "loss": 3.0277, "step": 8666 }, { "epoch": 0.68, "learning_rate": 9.320395201129146e-05, "loss": 3.0624, "step": 8667 }, { "epoch": 0.68, "learning_rate": 9.320316788206697e-05, "loss": 2.9885, "step": 8668 }, { "epoch": 0.68, "learning_rate": 9.320238375284247e-05, "loss": 2.9029, "step": 8669 }, { "epoch": 0.68, "learning_rate": 9.320159962361797e-05, "loss": 3.0207, "step": 8670 }, { "epoch": 0.68, "learning_rate": 9.320081549439348e-05, "loss": 2.9163, "step": 8671 }, { "epoch": 0.68, "learning_rate": 9.320003136516898e-05, "loss": 2.6984, "step": 8672 }, { "epoch": 0.68, "learning_rate": 9.319924723594449e-05, "loss": 3.1508, "step": 8673 }, { "epoch": 0.68, "learning_rate": 9.319846310671999e-05, "loss": 3.157, "step": 8674 }, { "epoch": 0.68, "learning_rate": 9.319767897749549e-05, "loss": 3.058, "step": 8675 }, { "epoch": 0.68, "learning_rate": 9.3196894848271e-05, "loss": 3.0298, "step": 8676 }, { "epoch": 0.68, "learning_rate": 9.31961107190465e-05, "loss": 3.0345, "step": 8677 }, { "epoch": 0.68, "learning_rate": 9.3195326589822e-05, "loss": 2.9578, "step": 8678 }, { "epoch": 0.68, "learning_rate": 9.319454246059751e-05, "loss": 3.1261, "step": 8679 }, { "epoch": 0.68, "learning_rate": 9.319375833137301e-05, "loss": 3.1054, "step": 8680 }, { "epoch": 0.68, "learning_rate": 9.319297420214852e-05, "loss": 2.8811, "step": 8681 }, { "epoch": 0.68, "learning_rate": 9.319219007292402e-05, "loss": 3.0741, "step": 8682 }, { "epoch": 0.68, "learning_rate": 9.319140594369952e-05, "loss": 3.1909, "step": 8683 }, { "epoch": 0.68, "learning_rate": 9.319062181447503e-05, "loss": 2.8135, "step": 8684 }, { "epoch": 0.68, "learning_rate": 9.318983768525053e-05, "loss": 3.1698, "step": 8685 }, { "epoch": 0.68, "learning_rate": 9.318905355602605e-05, "loss": 2.8354, "step": 8686 }, { "epoch": 0.68, "learning_rate": 9.318826942680155e-05, "loss": 2.6776, "step": 8687 }, { "epoch": 0.68, "learning_rate": 9.318748529757705e-05, "loss": 3.5624, "step": 8688 }, { "epoch": 0.68, "learning_rate": 9.318670116835256e-05, "loss": 3.0257, "step": 8689 }, { "epoch": 0.68, "learning_rate": 9.318591703912806e-05, "loss": 3.2189, "step": 8690 }, { "epoch": 0.68, "learning_rate": 9.318513290990357e-05, "loss": 2.9244, "step": 8691 }, { "epoch": 0.68, "learning_rate": 9.318434878067907e-05, "loss": 3.095, "step": 8692 }, { "epoch": 0.68, "learning_rate": 9.318356465145457e-05, "loss": 3.1232, "step": 8693 }, { "epoch": 0.68, "learning_rate": 9.318278052223008e-05, "loss": 3.123, "step": 8694 }, { "epoch": 0.68, "learning_rate": 9.318199639300558e-05, "loss": 3.0348, "step": 8695 }, { "epoch": 0.68, "learning_rate": 9.318121226378107e-05, "loss": 3.0972, "step": 8696 }, { "epoch": 0.68, "learning_rate": 9.318042813455657e-05, "loss": 3.309, "step": 8697 }, { "epoch": 0.68, "learning_rate": 9.317964400533208e-05, "loss": 3.0738, "step": 8698 }, { "epoch": 0.68, "learning_rate": 9.317885987610758e-05, "loss": 2.9412, "step": 8699 }, { "epoch": 0.68, "learning_rate": 9.317807574688308e-05, "loss": 2.8381, "step": 8700 }, { "epoch": 0.68, "learning_rate": 9.317729161765859e-05, "loss": 3.1234, "step": 8701 }, { "epoch": 0.68, "learning_rate": 9.317650748843409e-05, "loss": 2.9884, "step": 8702 }, { "epoch": 0.68, "learning_rate": 9.31757233592096e-05, "loss": 2.9367, "step": 8703 }, { "epoch": 0.68, "learning_rate": 9.31749392299851e-05, "loss": 3.0042, "step": 8704 }, { "epoch": 0.68, "learning_rate": 9.31741551007606e-05, "loss": 2.9415, "step": 8705 }, { "epoch": 0.68, "learning_rate": 9.317337097153611e-05, "loss": 2.876, "step": 8706 }, { "epoch": 0.68, "learning_rate": 9.317258684231161e-05, "loss": 3.0691, "step": 8707 }, { "epoch": 0.68, "learning_rate": 9.317180271308711e-05, "loss": 3.1313, "step": 8708 }, { "epoch": 0.68, "learning_rate": 9.317101858386262e-05, "loss": 3.0605, "step": 8709 }, { "epoch": 0.68, "learning_rate": 9.317023445463812e-05, "loss": 2.9364, "step": 8710 }, { "epoch": 0.68, "learning_rate": 9.316945032541363e-05, "loss": 3.1641, "step": 8711 }, { "epoch": 0.68, "learning_rate": 9.316866619618914e-05, "loss": 3.1414, "step": 8712 }, { "epoch": 0.68, "learning_rate": 9.316788206696465e-05, "loss": 2.9016, "step": 8713 }, { "epoch": 0.68, "learning_rate": 9.316709793774015e-05, "loss": 3.1591, "step": 8714 }, { "epoch": 0.68, "learning_rate": 9.316631380851565e-05, "loss": 2.9963, "step": 8715 }, { "epoch": 0.68, "learning_rate": 9.316552967929116e-05, "loss": 3.107, "step": 8716 }, { "epoch": 0.68, "learning_rate": 9.316474555006666e-05, "loss": 2.9618, "step": 8717 }, { "epoch": 0.68, "learning_rate": 9.316396142084216e-05, "loss": 3.1124, "step": 8718 }, { "epoch": 0.68, "learning_rate": 9.316317729161767e-05, "loss": 3.1094, "step": 8719 }, { "epoch": 0.68, "learning_rate": 9.316239316239317e-05, "loss": 2.9423, "step": 8720 }, { "epoch": 0.68, "learning_rate": 9.316160903316868e-05, "loss": 3.0623, "step": 8721 }, { "epoch": 0.68, "learning_rate": 9.316082490394418e-05, "loss": 2.8279, "step": 8722 }, { "epoch": 0.68, "learning_rate": 9.316004077471968e-05, "loss": 3.0097, "step": 8723 }, { "epoch": 0.68, "learning_rate": 9.315925664549519e-05, "loss": 2.7702, "step": 8724 }, { "epoch": 0.68, "learning_rate": 9.315847251627069e-05, "loss": 2.9704, "step": 8725 }, { "epoch": 0.68, "learning_rate": 9.31576883870462e-05, "loss": 2.9572, "step": 8726 }, { "epoch": 0.68, "learning_rate": 9.31569042578217e-05, "loss": 3.0001, "step": 8727 }, { "epoch": 0.68, "learning_rate": 9.31561201285972e-05, "loss": 3.3601, "step": 8728 }, { "epoch": 0.68, "learning_rate": 9.31553359993727e-05, "loss": 2.8462, "step": 8729 }, { "epoch": 0.68, "learning_rate": 9.315455187014821e-05, "loss": 3.1168, "step": 8730 }, { "epoch": 0.68, "learning_rate": 9.315376774092371e-05, "loss": 2.7374, "step": 8731 }, { "epoch": 0.68, "learning_rate": 9.31529836116992e-05, "loss": 2.7883, "step": 8732 }, { "epoch": 0.68, "learning_rate": 9.31521994824747e-05, "loss": 2.966, "step": 8733 }, { "epoch": 0.68, "learning_rate": 9.315141535325021e-05, "loss": 2.9863, "step": 8734 }, { "epoch": 0.68, "learning_rate": 9.315063122402571e-05, "loss": 3.0656, "step": 8735 }, { "epoch": 0.69, "learning_rate": 9.314984709480122e-05, "loss": 2.912, "step": 8736 }, { "epoch": 0.69, "learning_rate": 9.314906296557672e-05, "loss": 3.0954, "step": 8737 }, { "epoch": 0.69, "learning_rate": 9.314827883635224e-05, "loss": 3.0479, "step": 8738 }, { "epoch": 0.69, "learning_rate": 9.314749470712774e-05, "loss": 2.9972, "step": 8739 }, { "epoch": 0.69, "learning_rate": 9.314671057790325e-05, "loss": 2.8577, "step": 8740 }, { "epoch": 0.69, "learning_rate": 9.314592644867875e-05, "loss": 2.9059, "step": 8741 }, { "epoch": 0.69, "learning_rate": 9.314514231945425e-05, "loss": 2.9699, "step": 8742 }, { "epoch": 0.69, "learning_rate": 9.314435819022976e-05, "loss": 3.1023, "step": 8743 }, { "epoch": 0.69, "learning_rate": 9.314357406100526e-05, "loss": 2.9228, "step": 8744 }, { "epoch": 0.69, "learning_rate": 9.314278993178076e-05, "loss": 2.9983, "step": 8745 }, { "epoch": 0.69, "learning_rate": 9.314200580255627e-05, "loss": 2.9554, "step": 8746 }, { "epoch": 0.69, "learning_rate": 9.314122167333177e-05, "loss": 2.8956, "step": 8747 }, { "epoch": 0.69, "learning_rate": 9.314043754410728e-05, "loss": 3.0495, "step": 8748 }, { "epoch": 0.69, "learning_rate": 9.313965341488278e-05, "loss": 3.2561, "step": 8749 }, { "epoch": 0.69, "learning_rate": 9.313886928565828e-05, "loss": 3.0625, "step": 8750 }, { "epoch": 0.69, "learning_rate": 9.313808515643379e-05, "loss": 2.8508, "step": 8751 }, { "epoch": 0.69, "learning_rate": 9.313730102720929e-05, "loss": 3.2233, "step": 8752 }, { "epoch": 0.69, "learning_rate": 9.31365168979848e-05, "loss": 2.983, "step": 8753 }, { "epoch": 0.69, "learning_rate": 9.31357327687603e-05, "loss": 3.1361, "step": 8754 }, { "epoch": 0.69, "learning_rate": 9.31349486395358e-05, "loss": 2.8871, "step": 8755 }, { "epoch": 0.69, "learning_rate": 9.31341645103113e-05, "loss": 3.1491, "step": 8756 }, { "epoch": 0.69, "learning_rate": 9.313338038108681e-05, "loss": 2.8992, "step": 8757 }, { "epoch": 0.69, "learning_rate": 9.313259625186231e-05, "loss": 3.1327, "step": 8758 }, { "epoch": 0.69, "learning_rate": 9.313181212263782e-05, "loss": 2.9322, "step": 8759 }, { "epoch": 0.69, "learning_rate": 9.313102799341332e-05, "loss": 3.0044, "step": 8760 }, { "epoch": 0.69, "learning_rate": 9.313024386418882e-05, "loss": 3.0142, "step": 8761 }, { "epoch": 0.69, "learning_rate": 9.312945973496433e-05, "loss": 3.1576, "step": 8762 }, { "epoch": 0.69, "learning_rate": 9.312867560573983e-05, "loss": 3.3542, "step": 8763 }, { "epoch": 0.69, "learning_rate": 9.312789147651533e-05, "loss": 3.0379, "step": 8764 }, { "epoch": 0.69, "learning_rate": 9.312710734729084e-05, "loss": 2.966, "step": 8765 }, { "epoch": 0.69, "learning_rate": 9.312632321806634e-05, "loss": 3.2443, "step": 8766 }, { "epoch": 0.69, "learning_rate": 9.312553908884184e-05, "loss": 2.8795, "step": 8767 }, { "epoch": 0.69, "learning_rate": 9.312475495961735e-05, "loss": 2.7945, "step": 8768 }, { "epoch": 0.69, "learning_rate": 9.312397083039285e-05, "loss": 3.1124, "step": 8769 }, { "epoch": 0.69, "learning_rate": 9.312318670116836e-05, "loss": 3.1828, "step": 8770 }, { "epoch": 0.69, "learning_rate": 9.312240257194386e-05, "loss": 3.0371, "step": 8771 }, { "epoch": 0.69, "learning_rate": 9.312161844271936e-05, "loss": 2.9745, "step": 8772 }, { "epoch": 0.69, "learning_rate": 9.312083431349487e-05, "loss": 2.7521, "step": 8773 }, { "epoch": 0.69, "learning_rate": 9.312005018427037e-05, "loss": 2.9743, "step": 8774 }, { "epoch": 0.69, "learning_rate": 9.311926605504587e-05, "loss": 3.3203, "step": 8775 }, { "epoch": 0.69, "learning_rate": 9.311848192582138e-05, "loss": 2.9753, "step": 8776 }, { "epoch": 0.69, "learning_rate": 9.311769779659688e-05, "loss": 2.88, "step": 8777 }, { "epoch": 0.69, "learning_rate": 9.311691366737239e-05, "loss": 3.1452, "step": 8778 }, { "epoch": 0.69, "learning_rate": 9.311612953814789e-05, "loss": 3.0001, "step": 8779 }, { "epoch": 0.69, "learning_rate": 9.311534540892339e-05, "loss": 3.2976, "step": 8780 }, { "epoch": 0.69, "learning_rate": 9.31145612796989e-05, "loss": 2.9603, "step": 8781 }, { "epoch": 0.69, "learning_rate": 9.31137771504744e-05, "loss": 3.2218, "step": 8782 }, { "epoch": 0.69, "learning_rate": 9.31129930212499e-05, "loss": 2.8241, "step": 8783 }, { "epoch": 0.69, "learning_rate": 9.311220889202541e-05, "loss": 2.7149, "step": 8784 }, { "epoch": 0.69, "learning_rate": 9.311142476280091e-05, "loss": 2.8651, "step": 8785 }, { "epoch": 0.69, "learning_rate": 9.311064063357641e-05, "loss": 3.1547, "step": 8786 }, { "epoch": 0.69, "learning_rate": 9.310985650435192e-05, "loss": 3.1331, "step": 8787 }, { "epoch": 0.69, "learning_rate": 9.310907237512742e-05, "loss": 2.7764, "step": 8788 }, { "epoch": 0.69, "learning_rate": 9.310828824590293e-05, "loss": 3.0453, "step": 8789 }, { "epoch": 0.69, "learning_rate": 9.310750411667843e-05, "loss": 3.3786, "step": 8790 }, { "epoch": 0.69, "learning_rate": 9.310671998745393e-05, "loss": 2.8061, "step": 8791 }, { "epoch": 0.69, "learning_rate": 9.310593585822944e-05, "loss": 2.8689, "step": 8792 }, { "epoch": 0.69, "learning_rate": 9.310515172900494e-05, "loss": 3.0273, "step": 8793 }, { "epoch": 0.69, "learning_rate": 9.310436759978044e-05, "loss": 3.0618, "step": 8794 }, { "epoch": 0.69, "learning_rate": 9.310358347055596e-05, "loss": 3.3964, "step": 8795 }, { "epoch": 0.69, "learning_rate": 9.310279934133147e-05, "loss": 3.2851, "step": 8796 }, { "epoch": 0.69, "learning_rate": 9.310201521210697e-05, "loss": 3.1116, "step": 8797 }, { "epoch": 0.69, "learning_rate": 9.310123108288247e-05, "loss": 2.9743, "step": 8798 }, { "epoch": 0.69, "learning_rate": 9.310044695365798e-05, "loss": 3.0627, "step": 8799 }, { "epoch": 0.69, "learning_rate": 9.309966282443348e-05, "loss": 3.2813, "step": 8800 }, { "epoch": 0.69, "learning_rate": 9.309887869520898e-05, "loss": 3.1593, "step": 8801 }, { "epoch": 0.69, "learning_rate": 9.309809456598449e-05, "loss": 2.7884, "step": 8802 }, { "epoch": 0.69, "learning_rate": 9.309731043675999e-05, "loss": 3.1345, "step": 8803 }, { "epoch": 0.69, "learning_rate": 9.309652630753548e-05, "loss": 2.9964, "step": 8804 }, { "epoch": 0.69, "learning_rate": 9.309574217831098e-05, "loss": 2.8138, "step": 8805 }, { "epoch": 0.69, "learning_rate": 9.309495804908649e-05, "loss": 3.1262, "step": 8806 }, { "epoch": 0.69, "learning_rate": 9.309417391986199e-05, "loss": 2.8629, "step": 8807 }, { "epoch": 0.69, "learning_rate": 9.30933897906375e-05, "loss": 3.2256, "step": 8808 }, { "epoch": 0.69, "learning_rate": 9.3092605661413e-05, "loss": 2.9973, "step": 8809 }, { "epoch": 0.69, "learning_rate": 9.30918215321885e-05, "loss": 3.1666, "step": 8810 }, { "epoch": 0.69, "learning_rate": 9.3091037402964e-05, "loss": 3.2124, "step": 8811 }, { "epoch": 0.69, "learning_rate": 9.309025327373951e-05, "loss": 2.8436, "step": 8812 }, { "epoch": 0.69, "learning_rate": 9.308946914451501e-05, "loss": 3.1248, "step": 8813 }, { "epoch": 0.69, "learning_rate": 9.308868501529052e-05, "loss": 2.9821, "step": 8814 }, { "epoch": 0.69, "learning_rate": 9.308790088606602e-05, "loss": 3.2684, "step": 8815 }, { "epoch": 0.69, "learning_rate": 9.308711675684153e-05, "loss": 3.2794, "step": 8816 }, { "epoch": 0.69, "learning_rate": 9.308633262761703e-05, "loss": 3.3008, "step": 8817 }, { "epoch": 0.69, "learning_rate": 9.308554849839253e-05, "loss": 2.9198, "step": 8818 }, { "epoch": 0.69, "learning_rate": 9.308476436916804e-05, "loss": 2.9078, "step": 8819 }, { "epoch": 0.69, "learning_rate": 9.308398023994355e-05, "loss": 2.8184, "step": 8820 }, { "epoch": 0.69, "learning_rate": 9.308319611071906e-05, "loss": 3.2306, "step": 8821 }, { "epoch": 0.69, "learning_rate": 9.308241198149456e-05, "loss": 2.987, "step": 8822 }, { "epoch": 0.69, "learning_rate": 9.308162785227006e-05, "loss": 3.2622, "step": 8823 }, { "epoch": 0.69, "learning_rate": 9.308084372304557e-05, "loss": 3.0488, "step": 8824 }, { "epoch": 0.69, "learning_rate": 9.308005959382107e-05, "loss": 2.9631, "step": 8825 }, { "epoch": 0.69, "learning_rate": 9.307927546459658e-05, "loss": 2.9948, "step": 8826 }, { "epoch": 0.69, "learning_rate": 9.307849133537208e-05, "loss": 3.0775, "step": 8827 }, { "epoch": 0.69, "learning_rate": 9.307770720614758e-05, "loss": 3.2417, "step": 8828 }, { "epoch": 0.69, "learning_rate": 9.307692307692309e-05, "loss": 2.9885, "step": 8829 }, { "epoch": 0.69, "learning_rate": 9.307613894769859e-05, "loss": 3.0636, "step": 8830 }, { "epoch": 0.69, "learning_rate": 9.30753548184741e-05, "loss": 2.9418, "step": 8831 }, { "epoch": 0.69, "learning_rate": 9.30745706892496e-05, "loss": 2.9399, "step": 8832 }, { "epoch": 0.69, "learning_rate": 9.30737865600251e-05, "loss": 2.9456, "step": 8833 }, { "epoch": 0.69, "learning_rate": 9.30730024308006e-05, "loss": 2.8835, "step": 8834 }, { "epoch": 0.69, "learning_rate": 9.307221830157611e-05, "loss": 2.9374, "step": 8835 }, { "epoch": 0.69, "learning_rate": 9.307143417235161e-05, "loss": 2.9381, "step": 8836 }, { "epoch": 0.69, "learning_rate": 9.307065004312712e-05, "loss": 3.003, "step": 8837 }, { "epoch": 0.69, "learning_rate": 9.306986591390262e-05, "loss": 3.2107, "step": 8838 }, { "epoch": 0.69, "learning_rate": 9.306908178467812e-05, "loss": 2.8687, "step": 8839 }, { "epoch": 0.69, "learning_rate": 9.306829765545361e-05, "loss": 2.811, "step": 8840 }, { "epoch": 0.69, "learning_rate": 9.306751352622912e-05, "loss": 3.0824, "step": 8841 }, { "epoch": 0.69, "learning_rate": 9.306672939700462e-05, "loss": 3.4412, "step": 8842 }, { "epoch": 0.69, "learning_rate": 9.306594526778012e-05, "loss": 2.93, "step": 8843 }, { "epoch": 0.69, "learning_rate": 9.306516113855563e-05, "loss": 3.0046, "step": 8844 }, { "epoch": 0.69, "learning_rate": 9.306437700933113e-05, "loss": 3.0521, "step": 8845 }, { "epoch": 0.69, "learning_rate": 9.306359288010665e-05, "loss": 3.0139, "step": 8846 }, { "epoch": 0.69, "learning_rate": 9.306280875088215e-05, "loss": 3.1643, "step": 8847 }, { "epoch": 0.69, "learning_rate": 9.306202462165766e-05, "loss": 2.8243, "step": 8848 }, { "epoch": 0.69, "learning_rate": 9.306124049243316e-05, "loss": 3.0489, "step": 8849 }, { "epoch": 0.69, "learning_rate": 9.306045636320866e-05, "loss": 3.3164, "step": 8850 }, { "epoch": 0.69, "learning_rate": 9.305967223398417e-05, "loss": 3.1603, "step": 8851 }, { "epoch": 0.69, "learning_rate": 9.305888810475967e-05, "loss": 3.0779, "step": 8852 }, { "epoch": 0.69, "learning_rate": 9.305810397553517e-05, "loss": 2.9748, "step": 8853 }, { "epoch": 0.69, "learning_rate": 9.305731984631068e-05, "loss": 3.1428, "step": 8854 }, { "epoch": 0.69, "learning_rate": 9.305653571708618e-05, "loss": 2.9112, "step": 8855 }, { "epoch": 0.69, "learning_rate": 9.305575158786169e-05, "loss": 3.2524, "step": 8856 }, { "epoch": 0.69, "learning_rate": 9.305496745863719e-05, "loss": 3.0619, "step": 8857 }, { "epoch": 0.69, "learning_rate": 9.305418332941269e-05, "loss": 3.2733, "step": 8858 }, { "epoch": 0.69, "learning_rate": 9.30533992001882e-05, "loss": 2.9865, "step": 8859 }, { "epoch": 0.69, "learning_rate": 9.30526150709637e-05, "loss": 3.3062, "step": 8860 }, { "epoch": 0.69, "learning_rate": 9.30518309417392e-05, "loss": 2.9103, "step": 8861 }, { "epoch": 0.69, "learning_rate": 9.305104681251471e-05, "loss": 3.3213, "step": 8862 }, { "epoch": 0.69, "learning_rate": 9.305026268329021e-05, "loss": 2.7901, "step": 8863 }, { "epoch": 0.7, "learning_rate": 9.304947855406572e-05, "loss": 3.4811, "step": 8864 }, { "epoch": 0.7, "learning_rate": 9.304869442484122e-05, "loss": 3.2127, "step": 8865 }, { "epoch": 0.7, "learning_rate": 9.304791029561672e-05, "loss": 2.9095, "step": 8866 }, { "epoch": 0.7, "learning_rate": 9.304712616639223e-05, "loss": 2.9593, "step": 8867 }, { "epoch": 0.7, "learning_rate": 9.304634203716773e-05, "loss": 2.9138, "step": 8868 }, { "epoch": 0.7, "learning_rate": 9.304555790794323e-05, "loss": 3.2947, "step": 8869 }, { "epoch": 0.7, "learning_rate": 9.304477377871874e-05, "loss": 2.8893, "step": 8870 }, { "epoch": 0.7, "learning_rate": 9.304398964949424e-05, "loss": 2.9448, "step": 8871 }, { "epoch": 0.7, "learning_rate": 9.304320552026974e-05, "loss": 2.986, "step": 8872 }, { "epoch": 0.7, "learning_rate": 9.304242139104525e-05, "loss": 3.1412, "step": 8873 }, { "epoch": 0.7, "learning_rate": 9.304163726182075e-05, "loss": 3.2757, "step": 8874 }, { "epoch": 0.7, "learning_rate": 9.304085313259626e-05, "loss": 2.93, "step": 8875 }, { "epoch": 0.7, "learning_rate": 9.304006900337176e-05, "loss": 3.1678, "step": 8876 }, { "epoch": 0.7, "learning_rate": 9.303928487414726e-05, "loss": 2.9039, "step": 8877 }, { "epoch": 0.7, "learning_rate": 9.303850074492277e-05, "loss": 2.9787, "step": 8878 }, { "epoch": 0.7, "learning_rate": 9.303771661569827e-05, "loss": 3.0147, "step": 8879 }, { "epoch": 0.7, "learning_rate": 9.303693248647377e-05, "loss": 3.0787, "step": 8880 }, { "epoch": 0.7, "learning_rate": 9.303614835724928e-05, "loss": 3.0081, "step": 8881 }, { "epoch": 0.7, "learning_rate": 9.303536422802478e-05, "loss": 2.913, "step": 8882 }, { "epoch": 0.7, "learning_rate": 9.303458009880029e-05, "loss": 3.4042, "step": 8883 }, { "epoch": 0.7, "learning_rate": 9.303379596957579e-05, "loss": 2.997, "step": 8884 }, { "epoch": 0.7, "learning_rate": 9.303301184035129e-05, "loss": 2.8123, "step": 8885 }, { "epoch": 0.7, "learning_rate": 9.30322277111268e-05, "loss": 3.0316, "step": 8886 }, { "epoch": 0.7, "learning_rate": 9.30314435819023e-05, "loss": 3.1294, "step": 8887 }, { "epoch": 0.7, "learning_rate": 9.30306594526778e-05, "loss": 3.0553, "step": 8888 }, { "epoch": 0.7, "learning_rate": 9.302987532345331e-05, "loss": 3.1234, "step": 8889 }, { "epoch": 0.7, "learning_rate": 9.302909119422881e-05, "loss": 3.1417, "step": 8890 }, { "epoch": 0.7, "learning_rate": 9.302830706500431e-05, "loss": 2.9666, "step": 8891 }, { "epoch": 0.7, "learning_rate": 9.302752293577982e-05, "loss": 3.1076, "step": 8892 }, { "epoch": 0.7, "learning_rate": 9.302673880655532e-05, "loss": 3.1238, "step": 8893 }, { "epoch": 0.7, "learning_rate": 9.302595467733083e-05, "loss": 2.7023, "step": 8894 }, { "epoch": 0.7, "learning_rate": 9.302517054810633e-05, "loss": 2.8937, "step": 8895 }, { "epoch": 0.7, "learning_rate": 9.302438641888183e-05, "loss": 3.1037, "step": 8896 }, { "epoch": 0.7, "learning_rate": 9.302360228965734e-05, "loss": 3.0885, "step": 8897 }, { "epoch": 0.7, "learning_rate": 9.302281816043284e-05, "loss": 3.0919, "step": 8898 }, { "epoch": 0.7, "learning_rate": 9.302203403120834e-05, "loss": 2.9495, "step": 8899 }, { "epoch": 0.7, "learning_rate": 9.302124990198385e-05, "loss": 3.0394, "step": 8900 }, { "epoch": 0.7, "learning_rate": 9.302046577275935e-05, "loss": 3.0411, "step": 8901 }, { "epoch": 0.7, "learning_rate": 9.301968164353485e-05, "loss": 2.7716, "step": 8902 }, { "epoch": 0.7, "learning_rate": 9.301889751431037e-05, "loss": 2.9972, "step": 8903 }, { "epoch": 0.7, "learning_rate": 9.301811338508588e-05, "loss": 3.1047, "step": 8904 }, { "epoch": 0.7, "learning_rate": 9.301732925586138e-05, "loss": 2.8021, "step": 8905 }, { "epoch": 0.7, "learning_rate": 9.301654512663688e-05, "loss": 3.3871, "step": 8906 }, { "epoch": 0.7, "learning_rate": 9.301576099741239e-05, "loss": 2.8931, "step": 8907 }, { "epoch": 0.7, "learning_rate": 9.301497686818789e-05, "loss": 2.8233, "step": 8908 }, { "epoch": 0.7, "learning_rate": 9.30141927389634e-05, "loss": 2.8935, "step": 8909 }, { "epoch": 0.7, "learning_rate": 9.30134086097389e-05, "loss": 3.0588, "step": 8910 }, { "epoch": 0.7, "learning_rate": 9.30126244805144e-05, "loss": 3.4602, "step": 8911 }, { "epoch": 0.7, "learning_rate": 9.301184035128989e-05, "loss": 3.0463, "step": 8912 }, { "epoch": 0.7, "learning_rate": 9.30110562220654e-05, "loss": 2.9942, "step": 8913 }, { "epoch": 0.7, "learning_rate": 9.30102720928409e-05, "loss": 2.8961, "step": 8914 }, { "epoch": 0.7, "learning_rate": 9.30094879636164e-05, "loss": 2.945, "step": 8915 }, { "epoch": 0.7, "learning_rate": 9.30087038343919e-05, "loss": 2.9444, "step": 8916 }, { "epoch": 0.7, "learning_rate": 9.300791970516741e-05, "loss": 3.0962, "step": 8917 }, { "epoch": 0.7, "learning_rate": 9.300713557594291e-05, "loss": 3.0472, "step": 8918 }, { "epoch": 0.7, "learning_rate": 9.300635144671842e-05, "loss": 3.0175, "step": 8919 }, { "epoch": 0.7, "learning_rate": 9.300556731749392e-05, "loss": 3.1661, "step": 8920 }, { "epoch": 0.7, "learning_rate": 9.300478318826942e-05, "loss": 2.9758, "step": 8921 }, { "epoch": 0.7, "learning_rate": 9.300399905904493e-05, "loss": 3.1919, "step": 8922 }, { "epoch": 0.7, "learning_rate": 9.300321492982043e-05, "loss": 3.1588, "step": 8923 }, { "epoch": 0.7, "learning_rate": 9.300243080059594e-05, "loss": 3.2943, "step": 8924 }, { "epoch": 0.7, "learning_rate": 9.300164667137144e-05, "loss": 2.9207, "step": 8925 }, { "epoch": 0.7, "learning_rate": 9.300086254214694e-05, "loss": 2.948, "step": 8926 }, { "epoch": 0.7, "learning_rate": 9.300007841292245e-05, "loss": 2.949, "step": 8927 }, { "epoch": 0.7, "learning_rate": 9.299929428369796e-05, "loss": 3.1733, "step": 8928 }, { "epoch": 0.7, "learning_rate": 9.299851015447347e-05, "loss": 2.9722, "step": 8929 }, { "epoch": 0.7, "learning_rate": 9.299772602524897e-05, "loss": 2.9449, "step": 8930 }, { "epoch": 0.7, "learning_rate": 9.299694189602448e-05, "loss": 2.8455, "step": 8931 }, { "epoch": 0.7, "learning_rate": 9.299615776679998e-05, "loss": 3.0274, "step": 8932 }, { "epoch": 0.7, "learning_rate": 9.299537363757548e-05, "loss": 2.9596, "step": 8933 }, { "epoch": 0.7, "learning_rate": 9.299458950835099e-05, "loss": 3.0018, "step": 8934 }, { "epoch": 0.7, "learning_rate": 9.299380537912649e-05, "loss": 3.0775, "step": 8935 }, { "epoch": 0.7, "learning_rate": 9.2993021249902e-05, "loss": 3.0718, "step": 8936 }, { "epoch": 0.7, "learning_rate": 9.29922371206775e-05, "loss": 3.0653, "step": 8937 }, { "epoch": 0.7, "learning_rate": 9.2991452991453e-05, "loss": 2.9284, "step": 8938 }, { "epoch": 0.7, "learning_rate": 9.29906688622285e-05, "loss": 3.1388, "step": 8939 }, { "epoch": 0.7, "learning_rate": 9.298988473300401e-05, "loss": 3.1141, "step": 8940 }, { "epoch": 0.7, "learning_rate": 9.298910060377951e-05, "loss": 3.067, "step": 8941 }, { "epoch": 0.7, "learning_rate": 9.298831647455502e-05, "loss": 3.0422, "step": 8942 }, { "epoch": 0.7, "learning_rate": 9.298753234533052e-05, "loss": 3.0212, "step": 8943 }, { "epoch": 0.7, "learning_rate": 9.298674821610602e-05, "loss": 2.9103, "step": 8944 }, { "epoch": 0.7, "learning_rate": 9.298596408688153e-05, "loss": 3.0651, "step": 8945 }, { "epoch": 0.7, "learning_rate": 9.298517995765703e-05, "loss": 2.7899, "step": 8946 }, { "epoch": 0.7, "learning_rate": 9.298439582843253e-05, "loss": 3.2712, "step": 8947 }, { "epoch": 0.7, "learning_rate": 9.298361169920802e-05, "loss": 3.0767, "step": 8948 }, { "epoch": 0.7, "learning_rate": 9.298282756998353e-05, "loss": 3.002, "step": 8949 }, { "epoch": 0.7, "learning_rate": 9.298204344075903e-05, "loss": 2.8047, "step": 8950 }, { "epoch": 0.7, "learning_rate": 9.298125931153454e-05, "loss": 3.062, "step": 8951 }, { "epoch": 0.7, "learning_rate": 9.298047518231004e-05, "loss": 3.0578, "step": 8952 }, { "epoch": 0.7, "learning_rate": 9.297969105308554e-05, "loss": 3.0466, "step": 8953 }, { "epoch": 0.7, "learning_rate": 9.297890692386106e-05, "loss": 2.9094, "step": 8954 }, { "epoch": 0.7, "learning_rate": 9.297812279463656e-05, "loss": 3.0774, "step": 8955 }, { "epoch": 0.7, "learning_rate": 9.297733866541207e-05, "loss": 3.0538, "step": 8956 }, { "epoch": 0.7, "learning_rate": 9.297655453618757e-05, "loss": 2.8968, "step": 8957 }, { "epoch": 0.7, "learning_rate": 9.297577040696307e-05, "loss": 3.0172, "step": 8958 }, { "epoch": 0.7, "learning_rate": 9.297498627773858e-05, "loss": 3.1605, "step": 8959 }, { "epoch": 0.7, "learning_rate": 9.297420214851408e-05, "loss": 3.1679, "step": 8960 }, { "epoch": 0.7, "learning_rate": 9.297341801928959e-05, "loss": 2.9759, "step": 8961 }, { "epoch": 0.7, "learning_rate": 9.297263389006509e-05, "loss": 3.155, "step": 8962 }, { "epoch": 0.7, "learning_rate": 9.297184976084059e-05, "loss": 2.8302, "step": 8963 }, { "epoch": 0.7, "learning_rate": 9.29710656316161e-05, "loss": 3.053, "step": 8964 }, { "epoch": 0.7, "learning_rate": 9.29702815023916e-05, "loss": 3.0533, "step": 8965 }, { "epoch": 0.7, "learning_rate": 9.29694973731671e-05, "loss": 3.0365, "step": 8966 }, { "epoch": 0.7, "learning_rate": 9.296871324394261e-05, "loss": 3.0945, "step": 8967 }, { "epoch": 0.7, "learning_rate": 9.296792911471811e-05, "loss": 2.8656, "step": 8968 }, { "epoch": 0.7, "learning_rate": 9.296714498549361e-05, "loss": 2.7803, "step": 8969 }, { "epoch": 0.7, "learning_rate": 9.296636085626912e-05, "loss": 3.0188, "step": 8970 }, { "epoch": 0.7, "learning_rate": 9.296557672704462e-05, "loss": 3.5454, "step": 8971 }, { "epoch": 0.7, "learning_rate": 9.296479259782013e-05, "loss": 2.9628, "step": 8972 }, { "epoch": 0.7, "learning_rate": 9.296400846859563e-05, "loss": 2.9069, "step": 8973 }, { "epoch": 0.7, "learning_rate": 9.296322433937113e-05, "loss": 2.8751, "step": 8974 }, { "epoch": 0.7, "learning_rate": 9.296244021014664e-05, "loss": 2.9808, "step": 8975 }, { "epoch": 0.7, "learning_rate": 9.296165608092214e-05, "loss": 3.0387, "step": 8976 }, { "epoch": 0.7, "learning_rate": 9.296087195169764e-05, "loss": 3.1183, "step": 8977 }, { "epoch": 0.7, "learning_rate": 9.296008782247315e-05, "loss": 3.0159, "step": 8978 }, { "epoch": 0.7, "learning_rate": 9.295930369324865e-05, "loss": 3.2014, "step": 8979 }, { "epoch": 0.7, "learning_rate": 9.295851956402416e-05, "loss": 2.8501, "step": 8980 }, { "epoch": 0.7, "learning_rate": 9.295773543479966e-05, "loss": 3.2406, "step": 8981 }, { "epoch": 0.7, "learning_rate": 9.295695130557516e-05, "loss": 3.385, "step": 8982 }, { "epoch": 0.7, "learning_rate": 9.295616717635067e-05, "loss": 2.8989, "step": 8983 }, { "epoch": 0.7, "learning_rate": 9.295538304712617e-05, "loss": 3.0764, "step": 8984 }, { "epoch": 0.7, "learning_rate": 9.295459891790167e-05, "loss": 3.2043, "step": 8985 }, { "epoch": 0.7, "learning_rate": 9.295381478867718e-05, "loss": 3.0014, "step": 8986 }, { "epoch": 0.7, "learning_rate": 9.295303065945268e-05, "loss": 3.2597, "step": 8987 }, { "epoch": 0.7, "learning_rate": 9.295224653022818e-05, "loss": 2.7739, "step": 8988 }, { "epoch": 0.7, "learning_rate": 9.295146240100369e-05, "loss": 3.1472, "step": 8989 }, { "epoch": 0.7, "learning_rate": 9.295067827177919e-05, "loss": 2.8583, "step": 8990 }, { "epoch": 0.71, "learning_rate": 9.29498941425547e-05, "loss": 3.0841, "step": 8991 }, { "epoch": 0.71, "learning_rate": 9.29491100133302e-05, "loss": 3.1767, "step": 8992 }, { "epoch": 0.71, "learning_rate": 9.29483258841057e-05, "loss": 2.9723, "step": 8993 }, { "epoch": 0.71, "learning_rate": 9.294754175488121e-05, "loss": 2.7682, "step": 8994 }, { "epoch": 0.71, "learning_rate": 9.294675762565671e-05, "loss": 3.0225, "step": 8995 }, { "epoch": 0.71, "learning_rate": 9.294597349643221e-05, "loss": 2.9268, "step": 8996 }, { "epoch": 0.71, "learning_rate": 9.294518936720772e-05, "loss": 2.9019, "step": 8997 }, { "epoch": 0.71, "learning_rate": 9.294440523798322e-05, "loss": 3.0762, "step": 8998 }, { "epoch": 0.71, "learning_rate": 9.294362110875873e-05, "loss": 2.8582, "step": 8999 }, { "epoch": 0.71, "learning_rate": 9.294283697953423e-05, "loss": 2.9851, "step": 9000 }, { "epoch": 0.71, "eval_loss": 2.9403252601623535, "eval_runtime": 218.1426, "eval_samples_per_second": 51.948, "eval_steps_per_second": 3.25, "step": 9000 }, { "epoch": 0.71, "learning_rate": 9.294205285030973e-05, "loss": 2.9123, "step": 9001 }, { "epoch": 0.71, "learning_rate": 9.294126872108524e-05, "loss": 3.1473, "step": 9002 }, { "epoch": 0.71, "learning_rate": 9.294048459186074e-05, "loss": 3.0825, "step": 9003 }, { "epoch": 0.71, "learning_rate": 9.293970046263624e-05, "loss": 2.9911, "step": 9004 }, { "epoch": 0.71, "learning_rate": 9.293891633341175e-05, "loss": 3.0518, "step": 9005 }, { "epoch": 0.71, "learning_rate": 9.293813220418725e-05, "loss": 2.9265, "step": 9006 }, { "epoch": 0.71, "learning_rate": 9.293734807496275e-05, "loss": 2.8925, "step": 9007 }, { "epoch": 0.71, "learning_rate": 9.293656394573826e-05, "loss": 3.2357, "step": 9008 }, { "epoch": 0.71, "learning_rate": 9.293577981651376e-05, "loss": 3.3555, "step": 9009 }, { "epoch": 0.71, "learning_rate": 9.293499568728927e-05, "loss": 3.0671, "step": 9010 }, { "epoch": 0.71, "learning_rate": 9.293421155806478e-05, "loss": 3.247, "step": 9011 }, { "epoch": 0.71, "learning_rate": 9.293342742884029e-05, "loss": 3.0349, "step": 9012 }, { "epoch": 0.71, "learning_rate": 9.293264329961579e-05, "loss": 2.7947, "step": 9013 }, { "epoch": 0.71, "learning_rate": 9.29318591703913e-05, "loss": 2.9634, "step": 9014 }, { "epoch": 0.71, "learning_rate": 9.29310750411668e-05, "loss": 3.098, "step": 9015 }, { "epoch": 0.71, "learning_rate": 9.29302909119423e-05, "loss": 3.0623, "step": 9016 }, { "epoch": 0.71, "learning_rate": 9.29295067827178e-05, "loss": 2.9337, "step": 9017 }, { "epoch": 0.71, "learning_rate": 9.292872265349331e-05, "loss": 3.0256, "step": 9018 }, { "epoch": 0.71, "learning_rate": 9.292793852426881e-05, "loss": 2.9399, "step": 9019 }, { "epoch": 0.71, "learning_rate": 9.29271543950443e-05, "loss": 3.3075, "step": 9020 }, { "epoch": 0.71, "learning_rate": 9.29263702658198e-05, "loss": 2.9537, "step": 9021 }, { "epoch": 0.71, "learning_rate": 9.292558613659531e-05, "loss": 3.0027, "step": 9022 }, { "epoch": 0.71, "learning_rate": 9.292480200737081e-05, "loss": 3.1394, "step": 9023 }, { "epoch": 0.71, "learning_rate": 9.292401787814632e-05, "loss": 3.0172, "step": 9024 }, { "epoch": 0.71, "learning_rate": 9.292323374892182e-05, "loss": 3.0417, "step": 9025 }, { "epoch": 0.71, "learning_rate": 9.292244961969732e-05, "loss": 3.1395, "step": 9026 }, { "epoch": 0.71, "learning_rate": 9.292166549047283e-05, "loss": 2.9084, "step": 9027 }, { "epoch": 0.71, "learning_rate": 9.292088136124833e-05, "loss": 3.1007, "step": 9028 }, { "epoch": 0.71, "learning_rate": 9.292009723202384e-05, "loss": 3.178, "step": 9029 }, { "epoch": 0.71, "learning_rate": 9.291931310279934e-05, "loss": 2.9186, "step": 9030 }, { "epoch": 0.71, "learning_rate": 9.291852897357484e-05, "loss": 3.1148, "step": 9031 }, { "epoch": 0.71, "learning_rate": 9.291774484435035e-05, "loss": 2.9859, "step": 9032 }, { "epoch": 0.71, "learning_rate": 9.291696071512585e-05, "loss": 3.1435, "step": 9033 }, { "epoch": 0.71, "learning_rate": 9.291617658590135e-05, "loss": 2.9451, "step": 9034 }, { "epoch": 0.71, "learning_rate": 9.291539245667686e-05, "loss": 3.0315, "step": 9035 }, { "epoch": 0.71, "learning_rate": 9.291460832745237e-05, "loss": 2.9521, "step": 9036 }, { "epoch": 0.71, "learning_rate": 9.291382419822788e-05, "loss": 3.2217, "step": 9037 }, { "epoch": 0.71, "learning_rate": 9.291304006900338e-05, "loss": 2.9739, "step": 9038 }, { "epoch": 0.71, "learning_rate": 9.291225593977889e-05, "loss": 3.2242, "step": 9039 }, { "epoch": 0.71, "learning_rate": 9.291147181055439e-05, "loss": 3.0093, "step": 9040 }, { "epoch": 0.71, "learning_rate": 9.29106876813299e-05, "loss": 3.217, "step": 9041 }, { "epoch": 0.71, "learning_rate": 9.29099035521054e-05, "loss": 3.0626, "step": 9042 }, { "epoch": 0.71, "learning_rate": 9.29091194228809e-05, "loss": 2.9332, "step": 9043 }, { "epoch": 0.71, "learning_rate": 9.29083352936564e-05, "loss": 3.1133, "step": 9044 }, { "epoch": 0.71, "learning_rate": 9.290755116443191e-05, "loss": 3.1204, "step": 9045 }, { "epoch": 0.71, "learning_rate": 9.290676703520741e-05, "loss": 3.1183, "step": 9046 }, { "epoch": 0.71, "learning_rate": 9.290598290598292e-05, "loss": 2.8895, "step": 9047 }, { "epoch": 0.71, "learning_rate": 9.290519877675842e-05, "loss": 3.046, "step": 9048 }, { "epoch": 0.71, "learning_rate": 9.290441464753392e-05, "loss": 3.0421, "step": 9049 }, { "epoch": 0.71, "learning_rate": 9.290363051830943e-05, "loss": 3.1483, "step": 9050 }, { "epoch": 0.71, "learning_rate": 9.290284638908493e-05, "loss": 3.054, "step": 9051 }, { "epoch": 0.71, "learning_rate": 9.290206225986043e-05, "loss": 2.9878, "step": 9052 }, { "epoch": 0.71, "learning_rate": 9.290127813063594e-05, "loss": 2.8764, "step": 9053 }, { "epoch": 0.71, "learning_rate": 9.290049400141144e-05, "loss": 2.9729, "step": 9054 }, { "epoch": 0.71, "learning_rate": 9.289970987218694e-05, "loss": 3.0152, "step": 9055 }, { "epoch": 0.71, "learning_rate": 9.289892574296243e-05, "loss": 2.9132, "step": 9056 }, { "epoch": 0.71, "learning_rate": 9.289814161373794e-05, "loss": 2.9984, "step": 9057 }, { "epoch": 0.71, "learning_rate": 9.289735748451344e-05, "loss": 2.842, "step": 9058 }, { "epoch": 0.71, "learning_rate": 9.289657335528895e-05, "loss": 2.8809, "step": 9059 }, { "epoch": 0.71, "learning_rate": 9.289578922606445e-05, "loss": 3.0039, "step": 9060 }, { "epoch": 0.71, "learning_rate": 9.289500509683995e-05, "loss": 3.2698, "step": 9061 }, { "epoch": 0.71, "learning_rate": 9.289422096761547e-05, "loss": 3.2526, "step": 9062 }, { "epoch": 0.71, "learning_rate": 9.289343683839097e-05, "loss": 3.0285, "step": 9063 }, { "epoch": 0.71, "learning_rate": 9.289265270916648e-05, "loss": 2.9392, "step": 9064 }, { "epoch": 0.71, "learning_rate": 9.289186857994198e-05, "loss": 3.2387, "step": 9065 }, { "epoch": 0.71, "learning_rate": 9.289108445071749e-05, "loss": 3.1242, "step": 9066 }, { "epoch": 0.71, "learning_rate": 9.289030032149299e-05, "loss": 3.0614, "step": 9067 }, { "epoch": 0.71, "learning_rate": 9.288951619226849e-05, "loss": 3.3201, "step": 9068 }, { "epoch": 0.71, "learning_rate": 9.2888732063044e-05, "loss": 2.7125, "step": 9069 }, { "epoch": 0.71, "learning_rate": 9.28879479338195e-05, "loss": 3.0799, "step": 9070 }, { "epoch": 0.71, "learning_rate": 9.2887163804595e-05, "loss": 2.9187, "step": 9071 }, { "epoch": 0.71, "learning_rate": 9.288637967537051e-05, "loss": 3.1641, "step": 9072 }, { "epoch": 0.71, "learning_rate": 9.288559554614601e-05, "loss": 2.972, "step": 9073 }, { "epoch": 0.71, "learning_rate": 9.288481141692151e-05, "loss": 3.0993, "step": 9074 }, { "epoch": 0.71, "learning_rate": 9.288402728769702e-05, "loss": 3.1934, "step": 9075 }, { "epoch": 0.71, "learning_rate": 9.288324315847252e-05, "loss": 3.2352, "step": 9076 }, { "epoch": 0.71, "learning_rate": 9.288245902924803e-05, "loss": 3.1592, "step": 9077 }, { "epoch": 0.71, "learning_rate": 9.288167490002353e-05, "loss": 3.0845, "step": 9078 }, { "epoch": 0.71, "learning_rate": 9.288089077079903e-05, "loss": 3.1518, "step": 9079 }, { "epoch": 0.71, "learning_rate": 9.288010664157454e-05, "loss": 2.8475, "step": 9080 }, { "epoch": 0.71, "learning_rate": 9.287932251235004e-05, "loss": 3.0143, "step": 9081 }, { "epoch": 0.71, "learning_rate": 9.287853838312554e-05, "loss": 3.0071, "step": 9082 }, { "epoch": 0.71, "learning_rate": 9.287775425390105e-05, "loss": 3.1325, "step": 9083 }, { "epoch": 0.71, "learning_rate": 9.287697012467655e-05, "loss": 2.9361, "step": 9084 }, { "epoch": 0.71, "learning_rate": 9.287618599545206e-05, "loss": 3.0297, "step": 9085 }, { "epoch": 0.71, "learning_rate": 9.287540186622756e-05, "loss": 3.0849, "step": 9086 }, { "epoch": 0.71, "learning_rate": 9.287461773700306e-05, "loss": 3.0971, "step": 9087 }, { "epoch": 0.71, "learning_rate": 9.287383360777857e-05, "loss": 3.527, "step": 9088 }, { "epoch": 0.71, "learning_rate": 9.287304947855407e-05, "loss": 3.1885, "step": 9089 }, { "epoch": 0.71, "learning_rate": 9.287226534932957e-05, "loss": 2.8486, "step": 9090 }, { "epoch": 0.71, "learning_rate": 9.287148122010508e-05, "loss": 2.8747, "step": 9091 }, { "epoch": 0.71, "learning_rate": 9.287069709088058e-05, "loss": 2.9316, "step": 9092 }, { "epoch": 0.71, "learning_rate": 9.286991296165608e-05, "loss": 3.0506, "step": 9093 }, { "epoch": 0.71, "learning_rate": 9.286912883243159e-05, "loss": 3.0438, "step": 9094 }, { "epoch": 0.71, "learning_rate": 9.286834470320709e-05, "loss": 3.164, "step": 9095 }, { "epoch": 0.71, "learning_rate": 9.28675605739826e-05, "loss": 2.9364, "step": 9096 }, { "epoch": 0.71, "learning_rate": 9.28667764447581e-05, "loss": 2.9837, "step": 9097 }, { "epoch": 0.71, "learning_rate": 9.28659923155336e-05, "loss": 2.9296, "step": 9098 }, { "epoch": 0.71, "learning_rate": 9.28652081863091e-05, "loss": 3.0477, "step": 9099 }, { "epoch": 0.71, "learning_rate": 9.286442405708461e-05, "loss": 3.1107, "step": 9100 }, { "epoch": 0.71, "learning_rate": 9.286363992786011e-05, "loss": 3.0899, "step": 9101 }, { "epoch": 0.71, "learning_rate": 9.286285579863562e-05, "loss": 2.9856, "step": 9102 }, { "epoch": 0.71, "learning_rate": 9.286207166941112e-05, "loss": 3.1973, "step": 9103 }, { "epoch": 0.71, "learning_rate": 9.286128754018662e-05, "loss": 2.8915, "step": 9104 }, { "epoch": 0.71, "learning_rate": 9.286050341096213e-05, "loss": 2.8896, "step": 9105 }, { "epoch": 0.71, "learning_rate": 9.285971928173763e-05, "loss": 2.9236, "step": 9106 }, { "epoch": 0.71, "learning_rate": 9.285893515251314e-05, "loss": 3.2312, "step": 9107 }, { "epoch": 0.71, "learning_rate": 9.285815102328864e-05, "loss": 3.2386, "step": 9108 }, { "epoch": 0.71, "learning_rate": 9.285736689406414e-05, "loss": 3.0611, "step": 9109 }, { "epoch": 0.71, "learning_rate": 9.285658276483965e-05, "loss": 3.1855, "step": 9110 }, { "epoch": 0.71, "learning_rate": 9.285579863561515e-05, "loss": 2.9954, "step": 9111 }, { "epoch": 0.71, "learning_rate": 9.285501450639065e-05, "loss": 2.9913, "step": 9112 }, { "epoch": 0.71, "learning_rate": 9.285423037716616e-05, "loss": 3.0272, "step": 9113 }, { "epoch": 0.71, "learning_rate": 9.285344624794166e-05, "loss": 3.0785, "step": 9114 }, { "epoch": 0.71, "learning_rate": 9.285266211871717e-05, "loss": 2.8983, "step": 9115 }, { "epoch": 0.71, "learning_rate": 9.285187798949267e-05, "loss": 2.8865, "step": 9116 }, { "epoch": 0.71, "learning_rate": 9.285109386026817e-05, "loss": 3.1302, "step": 9117 }, { "epoch": 0.71, "learning_rate": 9.285030973104368e-05, "loss": 3.0541, "step": 9118 }, { "epoch": 0.72, "learning_rate": 9.28495256018192e-05, "loss": 3.0215, "step": 9119 }, { "epoch": 0.72, "learning_rate": 9.28487414725947e-05, "loss": 3.2014, "step": 9120 }, { "epoch": 0.72, "learning_rate": 9.28479573433702e-05, "loss": 3.0968, "step": 9121 }, { "epoch": 0.72, "learning_rate": 9.28471732141457e-05, "loss": 2.9264, "step": 9122 }, { "epoch": 0.72, "learning_rate": 9.284638908492121e-05, "loss": 2.9766, "step": 9123 }, { "epoch": 0.72, "learning_rate": 9.284560495569671e-05, "loss": 2.9594, "step": 9124 }, { "epoch": 0.72, "learning_rate": 9.284482082647222e-05, "loss": 3.1684, "step": 9125 }, { "epoch": 0.72, "learning_rate": 9.284403669724772e-05, "loss": 3.125, "step": 9126 }, { "epoch": 0.72, "learning_rate": 9.284325256802322e-05, "loss": 3.0446, "step": 9127 }, { "epoch": 0.72, "learning_rate": 9.284246843879871e-05, "loss": 3.3141, "step": 9128 }, { "epoch": 0.72, "learning_rate": 9.284168430957422e-05, "loss": 3.1231, "step": 9129 }, { "epoch": 0.72, "learning_rate": 9.284090018034972e-05, "loss": 2.9251, "step": 9130 }, { "epoch": 0.72, "learning_rate": 9.284011605112522e-05, "loss": 2.9645, "step": 9131 }, { "epoch": 0.72, "learning_rate": 9.283933192190073e-05, "loss": 2.9101, "step": 9132 }, { "epoch": 0.72, "learning_rate": 9.283854779267623e-05, "loss": 3.0658, "step": 9133 }, { "epoch": 0.72, "learning_rate": 9.283776366345174e-05, "loss": 3.1328, "step": 9134 }, { "epoch": 0.72, "learning_rate": 9.283697953422724e-05, "loss": 3.0206, "step": 9135 }, { "epoch": 0.72, "learning_rate": 9.283619540500274e-05, "loss": 2.9442, "step": 9136 }, { "epoch": 0.72, "learning_rate": 9.283541127577825e-05, "loss": 2.8382, "step": 9137 }, { "epoch": 0.72, "learning_rate": 9.283462714655375e-05, "loss": 3.1055, "step": 9138 }, { "epoch": 0.72, "learning_rate": 9.283384301732925e-05, "loss": 3.066, "step": 9139 }, { "epoch": 0.72, "learning_rate": 9.283305888810476e-05, "loss": 3.0852, "step": 9140 }, { "epoch": 0.72, "learning_rate": 9.283227475888026e-05, "loss": 2.793, "step": 9141 }, { "epoch": 0.72, "learning_rate": 9.283149062965576e-05, "loss": 3.1616, "step": 9142 }, { "epoch": 0.72, "learning_rate": 9.283070650043127e-05, "loss": 2.8906, "step": 9143 }, { "epoch": 0.72, "learning_rate": 9.282992237120679e-05, "loss": 2.9696, "step": 9144 }, { "epoch": 0.72, "learning_rate": 9.282913824198229e-05, "loss": 3.0594, "step": 9145 }, { "epoch": 0.72, "learning_rate": 9.282835411275779e-05, "loss": 3.0307, "step": 9146 }, { "epoch": 0.72, "learning_rate": 9.28275699835333e-05, "loss": 3.0469, "step": 9147 }, { "epoch": 0.72, "learning_rate": 9.28267858543088e-05, "loss": 3.1303, "step": 9148 }, { "epoch": 0.72, "learning_rate": 9.28260017250843e-05, "loss": 2.9754, "step": 9149 }, { "epoch": 0.72, "learning_rate": 9.282521759585981e-05, "loss": 2.8617, "step": 9150 }, { "epoch": 0.72, "learning_rate": 9.282443346663531e-05, "loss": 3.018, "step": 9151 }, { "epoch": 0.72, "learning_rate": 9.282364933741082e-05, "loss": 3.1785, "step": 9152 }, { "epoch": 0.72, "learning_rate": 9.282286520818632e-05, "loss": 3.0498, "step": 9153 }, { "epoch": 0.72, "learning_rate": 9.282208107896182e-05, "loss": 3.0888, "step": 9154 }, { "epoch": 0.72, "learning_rate": 9.282129694973733e-05, "loss": 3.2259, "step": 9155 }, { "epoch": 0.72, "learning_rate": 9.282051282051283e-05, "loss": 3.1009, "step": 9156 }, { "epoch": 0.72, "learning_rate": 9.281972869128833e-05, "loss": 2.961, "step": 9157 }, { "epoch": 0.72, "learning_rate": 9.281894456206384e-05, "loss": 3.1343, "step": 9158 }, { "epoch": 0.72, "learning_rate": 9.281816043283934e-05, "loss": 2.8042, "step": 9159 }, { "epoch": 0.72, "learning_rate": 9.281737630361484e-05, "loss": 2.9084, "step": 9160 }, { "epoch": 0.72, "learning_rate": 9.281659217439035e-05, "loss": 3.0614, "step": 9161 }, { "epoch": 0.72, "learning_rate": 9.281580804516585e-05, "loss": 3.2033, "step": 9162 }, { "epoch": 0.72, "learning_rate": 9.281502391594136e-05, "loss": 3.1969, "step": 9163 }, { "epoch": 0.72, "learning_rate": 9.281423978671685e-05, "loss": 2.8784, "step": 9164 }, { "epoch": 0.72, "learning_rate": 9.281345565749235e-05, "loss": 2.965, "step": 9165 }, { "epoch": 0.72, "learning_rate": 9.281267152826785e-05, "loss": 3.2675, "step": 9166 }, { "epoch": 0.72, "learning_rate": 9.281188739904336e-05, "loss": 3.1178, "step": 9167 }, { "epoch": 0.72, "learning_rate": 9.281110326981886e-05, "loss": 2.9005, "step": 9168 }, { "epoch": 0.72, "learning_rate": 9.281031914059436e-05, "loss": 3.0081, "step": 9169 }, { "epoch": 0.72, "learning_rate": 9.280953501136988e-05, "loss": 3.0827, "step": 9170 }, { "epoch": 0.72, "learning_rate": 9.280875088214538e-05, "loss": 2.9221, "step": 9171 }, { "epoch": 0.72, "learning_rate": 9.280796675292089e-05, "loss": 2.9802, "step": 9172 }, { "epoch": 0.72, "learning_rate": 9.280718262369639e-05, "loss": 3.0895, "step": 9173 }, { "epoch": 0.72, "learning_rate": 9.28063984944719e-05, "loss": 3.3126, "step": 9174 }, { "epoch": 0.72, "learning_rate": 9.28056143652474e-05, "loss": 3.1052, "step": 9175 }, { "epoch": 0.72, "learning_rate": 9.28048302360229e-05, "loss": 3.0861, "step": 9176 }, { "epoch": 0.72, "learning_rate": 9.280404610679841e-05, "loss": 2.9548, "step": 9177 }, { "epoch": 0.72, "learning_rate": 9.280326197757391e-05, "loss": 3.2848, "step": 9178 }, { "epoch": 0.72, "learning_rate": 9.280247784834941e-05, "loss": 2.9451, "step": 9179 }, { "epoch": 0.72, "learning_rate": 9.280169371912492e-05, "loss": 3.1026, "step": 9180 }, { "epoch": 0.72, "learning_rate": 9.280090958990042e-05, "loss": 3.1068, "step": 9181 }, { "epoch": 0.72, "learning_rate": 9.280012546067593e-05, "loss": 3.0297, "step": 9182 }, { "epoch": 0.72, "learning_rate": 9.279934133145143e-05, "loss": 3.2723, "step": 9183 }, { "epoch": 0.72, "learning_rate": 9.279855720222693e-05, "loss": 3.1203, "step": 9184 }, { "epoch": 0.72, "learning_rate": 9.279777307300244e-05, "loss": 3.1637, "step": 9185 }, { "epoch": 0.72, "learning_rate": 9.279698894377794e-05, "loss": 3.0105, "step": 9186 }, { "epoch": 0.72, "learning_rate": 9.279620481455344e-05, "loss": 3.0734, "step": 9187 }, { "epoch": 0.72, "learning_rate": 9.279542068532895e-05, "loss": 3.0643, "step": 9188 }, { "epoch": 0.72, "learning_rate": 9.279463655610445e-05, "loss": 3.0702, "step": 9189 }, { "epoch": 0.72, "learning_rate": 9.279385242687995e-05, "loss": 3.194, "step": 9190 }, { "epoch": 0.72, "learning_rate": 9.279306829765546e-05, "loss": 2.9008, "step": 9191 }, { "epoch": 0.72, "learning_rate": 9.279228416843096e-05, "loss": 3.2066, "step": 9192 }, { "epoch": 0.72, "learning_rate": 9.279150003920647e-05, "loss": 2.916, "step": 9193 }, { "epoch": 0.72, "learning_rate": 9.279071590998197e-05, "loss": 3.1515, "step": 9194 }, { "epoch": 0.72, "learning_rate": 9.278993178075747e-05, "loss": 3.0276, "step": 9195 }, { "epoch": 0.72, "learning_rate": 9.278914765153298e-05, "loss": 3.2016, "step": 9196 }, { "epoch": 0.72, "learning_rate": 9.278836352230848e-05, "loss": 3.1638, "step": 9197 }, { "epoch": 0.72, "learning_rate": 9.278757939308398e-05, "loss": 3.1199, "step": 9198 }, { "epoch": 0.72, "learning_rate": 9.278679526385949e-05, "loss": 3.0412, "step": 9199 }, { "epoch": 0.72, "learning_rate": 9.278601113463499e-05, "loss": 2.9343, "step": 9200 }, { "epoch": 0.72, "learning_rate": 9.27852270054105e-05, "loss": 2.9021, "step": 9201 }, { "epoch": 0.72, "learning_rate": 9.2784442876186e-05, "loss": 3.1324, "step": 9202 }, { "epoch": 0.72, "learning_rate": 9.27836587469615e-05, "loss": 2.9384, "step": 9203 }, { "epoch": 0.72, "learning_rate": 9.2782874617737e-05, "loss": 3.1598, "step": 9204 }, { "epoch": 0.72, "learning_rate": 9.278209048851251e-05, "loss": 2.8438, "step": 9205 }, { "epoch": 0.72, "learning_rate": 9.278130635928801e-05, "loss": 3.0062, "step": 9206 }, { "epoch": 0.72, "learning_rate": 9.278052223006352e-05, "loss": 3.1106, "step": 9207 }, { "epoch": 0.72, "learning_rate": 9.277973810083902e-05, "loss": 3.0124, "step": 9208 }, { "epoch": 0.72, "learning_rate": 9.277895397161452e-05, "loss": 3.3772, "step": 9209 }, { "epoch": 0.72, "learning_rate": 9.277816984239003e-05, "loss": 2.9947, "step": 9210 }, { "epoch": 0.72, "learning_rate": 9.277738571316553e-05, "loss": 3.1916, "step": 9211 }, { "epoch": 0.72, "learning_rate": 9.277660158394104e-05, "loss": 3.0006, "step": 9212 }, { "epoch": 0.72, "learning_rate": 9.277581745471654e-05, "loss": 3.1734, "step": 9213 }, { "epoch": 0.72, "learning_rate": 9.277503332549204e-05, "loss": 3.0444, "step": 9214 }, { "epoch": 0.72, "learning_rate": 9.277424919626755e-05, "loss": 2.9396, "step": 9215 }, { "epoch": 0.72, "learning_rate": 9.277346506704305e-05, "loss": 2.9925, "step": 9216 }, { "epoch": 0.72, "learning_rate": 9.277268093781855e-05, "loss": 3.1226, "step": 9217 }, { "epoch": 0.72, "learning_rate": 9.277189680859406e-05, "loss": 2.887, "step": 9218 }, { "epoch": 0.72, "learning_rate": 9.277111267936956e-05, "loss": 3.4989, "step": 9219 }, { "epoch": 0.72, "learning_rate": 9.277032855014507e-05, "loss": 3.0101, "step": 9220 }, { "epoch": 0.72, "learning_rate": 9.276954442092057e-05, "loss": 3.2058, "step": 9221 }, { "epoch": 0.72, "learning_rate": 9.276876029169607e-05, "loss": 3.1495, "step": 9222 }, { "epoch": 0.72, "learning_rate": 9.276797616247158e-05, "loss": 3.079, "step": 9223 }, { "epoch": 0.72, "learning_rate": 9.276719203324708e-05, "loss": 3.2281, "step": 9224 }, { "epoch": 0.72, "learning_rate": 9.276640790402258e-05, "loss": 2.973, "step": 9225 }, { "epoch": 0.72, "learning_rate": 9.276562377479809e-05, "loss": 3.1356, "step": 9226 }, { "epoch": 0.72, "learning_rate": 9.27648396455736e-05, "loss": 2.9522, "step": 9227 }, { "epoch": 0.72, "learning_rate": 9.276405551634911e-05, "loss": 3.0605, "step": 9228 }, { "epoch": 0.72, "learning_rate": 9.276327138712461e-05, "loss": 2.9505, "step": 9229 }, { "epoch": 0.72, "learning_rate": 9.276248725790012e-05, "loss": 3.0576, "step": 9230 }, { "epoch": 0.72, "learning_rate": 9.276170312867562e-05, "loss": 3.1505, "step": 9231 }, { "epoch": 0.72, "learning_rate": 9.276091899945112e-05, "loss": 3.1163, "step": 9232 }, { "epoch": 0.72, "learning_rate": 9.276013487022663e-05, "loss": 3.1381, "step": 9233 }, { "epoch": 0.72, "learning_rate": 9.275935074100213e-05, "loss": 3.0146, "step": 9234 }, { "epoch": 0.72, "learning_rate": 9.275856661177763e-05, "loss": 3.1445, "step": 9235 }, { "epoch": 0.72, "learning_rate": 9.275778248255312e-05, "loss": 3.058, "step": 9236 }, { "epoch": 0.72, "learning_rate": 9.275699835332863e-05, "loss": 2.9346, "step": 9237 }, { "epoch": 0.72, "learning_rate": 9.275621422410413e-05, "loss": 2.9966, "step": 9238 }, { "epoch": 0.72, "learning_rate": 9.275543009487964e-05, "loss": 3.2229, "step": 9239 }, { "epoch": 0.72, "learning_rate": 9.275464596565514e-05, "loss": 3.157, "step": 9240 }, { "epoch": 0.72, "learning_rate": 9.275386183643064e-05, "loss": 3.0566, "step": 9241 }, { "epoch": 0.72, "learning_rate": 9.275307770720615e-05, "loss": 3.0738, "step": 9242 }, { "epoch": 0.72, "learning_rate": 9.275229357798165e-05, "loss": 2.9708, "step": 9243 }, { "epoch": 0.72, "learning_rate": 9.275150944875715e-05, "loss": 2.7206, "step": 9244 }, { "epoch": 0.72, "learning_rate": 9.275072531953266e-05, "loss": 3.253, "step": 9245 }, { "epoch": 0.73, "learning_rate": 9.274994119030816e-05, "loss": 2.9743, "step": 9246 }, { "epoch": 0.73, "learning_rate": 9.274915706108366e-05, "loss": 3.0863, "step": 9247 }, { "epoch": 0.73, "learning_rate": 9.274837293185917e-05, "loss": 3.171, "step": 9248 }, { "epoch": 0.73, "learning_rate": 9.274758880263467e-05, "loss": 3.4362, "step": 9249 }, { "epoch": 0.73, "learning_rate": 9.274680467341018e-05, "loss": 3.0044, "step": 9250 }, { "epoch": 0.73, "learning_rate": 9.274602054418568e-05, "loss": 3.0767, "step": 9251 }, { "epoch": 0.73, "learning_rate": 9.27452364149612e-05, "loss": 2.9213, "step": 9252 }, { "epoch": 0.73, "learning_rate": 9.27444522857367e-05, "loss": 3.12, "step": 9253 }, { "epoch": 0.73, "learning_rate": 9.27436681565122e-05, "loss": 3.1656, "step": 9254 }, { "epoch": 0.73, "learning_rate": 9.274288402728771e-05, "loss": 3.0632, "step": 9255 }, { "epoch": 0.73, "learning_rate": 9.274209989806321e-05, "loss": 3.1132, "step": 9256 }, { "epoch": 0.73, "learning_rate": 9.274131576883871e-05, "loss": 2.9131, "step": 9257 }, { "epoch": 0.73, "learning_rate": 9.274053163961422e-05, "loss": 3.0197, "step": 9258 }, { "epoch": 0.73, "learning_rate": 9.273974751038972e-05, "loss": 3.0786, "step": 9259 }, { "epoch": 0.73, "learning_rate": 9.273896338116523e-05, "loss": 3.0758, "step": 9260 }, { "epoch": 0.73, "learning_rate": 9.273817925194073e-05, "loss": 3.3359, "step": 9261 }, { "epoch": 0.73, "learning_rate": 9.273739512271623e-05, "loss": 3.3258, "step": 9262 }, { "epoch": 0.73, "learning_rate": 9.273661099349174e-05, "loss": 3.1681, "step": 9263 }, { "epoch": 0.73, "learning_rate": 9.273582686426724e-05, "loss": 3.0504, "step": 9264 }, { "epoch": 0.73, "learning_rate": 9.273504273504274e-05, "loss": 2.8179, "step": 9265 }, { "epoch": 0.73, "learning_rate": 9.273425860581825e-05, "loss": 2.887, "step": 9266 }, { "epoch": 0.73, "learning_rate": 9.273347447659375e-05, "loss": 3.2546, "step": 9267 }, { "epoch": 0.73, "learning_rate": 9.273269034736926e-05, "loss": 2.8615, "step": 9268 }, { "epoch": 0.73, "learning_rate": 9.273190621814476e-05, "loss": 3.2195, "step": 9269 }, { "epoch": 0.73, "learning_rate": 9.273112208892026e-05, "loss": 3.2257, "step": 9270 }, { "epoch": 0.73, "learning_rate": 9.273033795969577e-05, "loss": 3.0941, "step": 9271 }, { "epoch": 0.73, "learning_rate": 9.272955383047126e-05, "loss": 3.1208, "step": 9272 }, { "epoch": 0.73, "learning_rate": 9.272876970124676e-05, "loss": 2.995, "step": 9273 }, { "epoch": 0.73, "learning_rate": 9.272798557202226e-05, "loss": 2.9809, "step": 9274 }, { "epoch": 0.73, "learning_rate": 9.272720144279777e-05, "loss": 2.8931, "step": 9275 }, { "epoch": 0.73, "learning_rate": 9.272641731357327e-05, "loss": 3.0149, "step": 9276 }, { "epoch": 0.73, "learning_rate": 9.272563318434877e-05, "loss": 2.9022, "step": 9277 }, { "epoch": 0.73, "learning_rate": 9.272484905512429e-05, "loss": 2.8911, "step": 9278 }, { "epoch": 0.73, "learning_rate": 9.27240649258998e-05, "loss": 3.1688, "step": 9279 }, { "epoch": 0.73, "learning_rate": 9.27232807966753e-05, "loss": 2.8927, "step": 9280 }, { "epoch": 0.73, "learning_rate": 9.27224966674508e-05, "loss": 3.2255, "step": 9281 }, { "epoch": 0.73, "learning_rate": 9.27217125382263e-05, "loss": 3.1292, "step": 9282 }, { "epoch": 0.73, "learning_rate": 9.272092840900181e-05, "loss": 3.1258, "step": 9283 }, { "epoch": 0.73, "learning_rate": 9.272014427977731e-05, "loss": 2.8634, "step": 9284 }, { "epoch": 0.73, "learning_rate": 9.271936015055282e-05, "loss": 2.8827, "step": 9285 }, { "epoch": 0.73, "learning_rate": 9.271857602132832e-05, "loss": 3.0073, "step": 9286 }, { "epoch": 0.73, "learning_rate": 9.271779189210383e-05, "loss": 3.0082, "step": 9287 }, { "epoch": 0.73, "learning_rate": 9.271700776287933e-05, "loss": 2.9093, "step": 9288 }, { "epoch": 0.73, "learning_rate": 9.271622363365483e-05, "loss": 3.0121, "step": 9289 }, { "epoch": 0.73, "learning_rate": 9.271543950443034e-05, "loss": 2.8235, "step": 9290 }, { "epoch": 0.73, "learning_rate": 9.271465537520584e-05, "loss": 3.0605, "step": 9291 }, { "epoch": 0.73, "learning_rate": 9.271387124598134e-05, "loss": 2.9485, "step": 9292 }, { "epoch": 0.73, "learning_rate": 9.271308711675685e-05, "loss": 3.1348, "step": 9293 }, { "epoch": 0.73, "learning_rate": 9.271230298753235e-05, "loss": 3.4623, "step": 9294 }, { "epoch": 0.73, "learning_rate": 9.271151885830785e-05, "loss": 2.922, "step": 9295 }, { "epoch": 0.73, "learning_rate": 9.271073472908336e-05, "loss": 3.0847, "step": 9296 }, { "epoch": 0.73, "learning_rate": 9.270995059985886e-05, "loss": 3.0746, "step": 9297 }, { "epoch": 0.73, "learning_rate": 9.270916647063437e-05, "loss": 2.8734, "step": 9298 }, { "epoch": 0.73, "learning_rate": 9.270838234140987e-05, "loss": 2.9369, "step": 9299 }, { "epoch": 0.73, "learning_rate": 9.270759821218537e-05, "loss": 2.9195, "step": 9300 }, { "epoch": 0.73, "learning_rate": 9.270681408296088e-05, "loss": 2.8563, "step": 9301 }, { "epoch": 0.73, "learning_rate": 9.270602995373638e-05, "loss": 3.0896, "step": 9302 }, { "epoch": 0.73, "learning_rate": 9.270524582451188e-05, "loss": 3.0263, "step": 9303 }, { "epoch": 0.73, "learning_rate": 9.270446169528739e-05, "loss": 2.8394, "step": 9304 }, { "epoch": 0.73, "learning_rate": 9.270367756606289e-05, "loss": 3.0076, "step": 9305 }, { "epoch": 0.73, "learning_rate": 9.27028934368384e-05, "loss": 3.1165, "step": 9306 }, { "epoch": 0.73, "learning_rate": 9.27021093076139e-05, "loss": 3.3012, "step": 9307 }, { "epoch": 0.73, "learning_rate": 9.27013251783894e-05, "loss": 2.9711, "step": 9308 }, { "epoch": 0.73, "learning_rate": 9.27005410491649e-05, "loss": 3.022, "step": 9309 }, { "epoch": 0.73, "learning_rate": 9.269975691994041e-05, "loss": 2.9431, "step": 9310 }, { "epoch": 0.73, "learning_rate": 9.269897279071591e-05, "loss": 3.2011, "step": 9311 }, { "epoch": 0.73, "learning_rate": 9.269818866149142e-05, "loss": 3.0511, "step": 9312 }, { "epoch": 0.73, "learning_rate": 9.269740453226692e-05, "loss": 3.2321, "step": 9313 }, { "epoch": 0.73, "learning_rate": 9.269662040304242e-05, "loss": 2.8356, "step": 9314 }, { "epoch": 0.73, "learning_rate": 9.269583627381793e-05, "loss": 2.9677, "step": 9315 }, { "epoch": 0.73, "learning_rate": 9.269505214459343e-05, "loss": 2.8818, "step": 9316 }, { "epoch": 0.73, "learning_rate": 9.269426801536894e-05, "loss": 2.9545, "step": 9317 }, { "epoch": 0.73, "learning_rate": 9.269348388614444e-05, "loss": 2.9137, "step": 9318 }, { "epoch": 0.73, "learning_rate": 9.269269975691994e-05, "loss": 3.0751, "step": 9319 }, { "epoch": 0.73, "learning_rate": 9.269191562769545e-05, "loss": 2.8404, "step": 9320 }, { "epoch": 0.73, "learning_rate": 9.269113149847095e-05, "loss": 3.0292, "step": 9321 }, { "epoch": 0.73, "learning_rate": 9.269034736924645e-05, "loss": 3.1931, "step": 9322 }, { "epoch": 0.73, "learning_rate": 9.268956324002196e-05, "loss": 2.8829, "step": 9323 }, { "epoch": 0.73, "learning_rate": 9.268877911079746e-05, "loss": 2.9421, "step": 9324 }, { "epoch": 0.73, "learning_rate": 9.268799498157296e-05, "loss": 3.4903, "step": 9325 }, { "epoch": 0.73, "learning_rate": 9.268721085234847e-05, "loss": 3.2451, "step": 9326 }, { "epoch": 0.73, "learning_rate": 9.268642672312397e-05, "loss": 2.9903, "step": 9327 }, { "epoch": 0.73, "learning_rate": 9.268564259389948e-05, "loss": 2.9398, "step": 9328 }, { "epoch": 0.73, "learning_rate": 9.268485846467498e-05, "loss": 3.1252, "step": 9329 }, { "epoch": 0.73, "learning_rate": 9.268407433545048e-05, "loss": 2.8967, "step": 9330 }, { "epoch": 0.73, "learning_rate": 9.268329020622599e-05, "loss": 3.1134, "step": 9331 }, { "epoch": 0.73, "learning_rate": 9.268250607700149e-05, "loss": 2.8765, "step": 9332 }, { "epoch": 0.73, "learning_rate": 9.2681721947777e-05, "loss": 3.1283, "step": 9333 }, { "epoch": 0.73, "learning_rate": 9.26809378185525e-05, "loss": 2.9252, "step": 9334 }, { "epoch": 0.73, "learning_rate": 9.268015368932802e-05, "loss": 3.0537, "step": 9335 }, { "epoch": 0.73, "learning_rate": 9.267936956010352e-05, "loss": 2.9408, "step": 9336 }, { "epoch": 0.73, "learning_rate": 9.267858543087902e-05, "loss": 2.7188, "step": 9337 }, { "epoch": 0.73, "learning_rate": 9.267780130165453e-05, "loss": 3.1889, "step": 9338 }, { "epoch": 0.73, "learning_rate": 9.267701717243003e-05, "loss": 3.0235, "step": 9339 }, { "epoch": 0.73, "learning_rate": 9.267623304320553e-05, "loss": 2.9408, "step": 9340 }, { "epoch": 0.73, "learning_rate": 9.267544891398104e-05, "loss": 3.0952, "step": 9341 }, { "epoch": 0.73, "learning_rate": 9.267466478475654e-05, "loss": 3.0209, "step": 9342 }, { "epoch": 0.73, "learning_rate": 9.267388065553204e-05, "loss": 3.2314, "step": 9343 }, { "epoch": 0.73, "learning_rate": 9.267309652630753e-05, "loss": 3.1709, "step": 9344 }, { "epoch": 0.73, "learning_rate": 9.267231239708304e-05, "loss": 2.9753, "step": 9345 }, { "epoch": 0.73, "learning_rate": 9.267152826785854e-05, "loss": 3.0712, "step": 9346 }, { "epoch": 0.73, "learning_rate": 9.267074413863405e-05, "loss": 3.06, "step": 9347 }, { "epoch": 0.73, "learning_rate": 9.266996000940955e-05, "loss": 3.1947, "step": 9348 }, { "epoch": 0.73, "learning_rate": 9.266917588018505e-05, "loss": 3.0105, "step": 9349 }, { "epoch": 0.73, "learning_rate": 9.266839175096056e-05, "loss": 3.2575, "step": 9350 }, { "epoch": 0.73, "learning_rate": 9.266760762173606e-05, "loss": 3.0383, "step": 9351 }, { "epoch": 0.73, "learning_rate": 9.266682349251156e-05, "loss": 3.1637, "step": 9352 }, { "epoch": 0.73, "learning_rate": 9.266603936328707e-05, "loss": 3.0628, "step": 9353 }, { "epoch": 0.73, "learning_rate": 9.266525523406257e-05, "loss": 2.9867, "step": 9354 }, { "epoch": 0.73, "learning_rate": 9.266447110483808e-05, "loss": 3.006, "step": 9355 }, { "epoch": 0.73, "learning_rate": 9.266368697561358e-05, "loss": 2.7388, "step": 9356 }, { "epoch": 0.73, "learning_rate": 9.266290284638908e-05, "loss": 2.989, "step": 9357 }, { "epoch": 0.73, "learning_rate": 9.266211871716459e-05, "loss": 2.8539, "step": 9358 }, { "epoch": 0.73, "learning_rate": 9.266133458794009e-05, "loss": 3.0555, "step": 9359 }, { "epoch": 0.73, "learning_rate": 9.266055045871561e-05, "loss": 3.1266, "step": 9360 }, { "epoch": 0.73, "learning_rate": 9.265976632949111e-05, "loss": 3.3448, "step": 9361 }, { "epoch": 0.73, "learning_rate": 9.265898220026661e-05, "loss": 3.0338, "step": 9362 }, { "epoch": 0.73, "learning_rate": 9.265819807104212e-05, "loss": 3.0598, "step": 9363 }, { "epoch": 0.73, "learning_rate": 9.265741394181762e-05, "loss": 2.9371, "step": 9364 }, { "epoch": 0.73, "learning_rate": 9.265662981259313e-05, "loss": 3.2374, "step": 9365 }, { "epoch": 0.73, "learning_rate": 9.265584568336863e-05, "loss": 3.0866, "step": 9366 }, { "epoch": 0.73, "learning_rate": 9.265506155414413e-05, "loss": 2.9269, "step": 9367 }, { "epoch": 0.73, "learning_rate": 9.265427742491964e-05, "loss": 2.8797, "step": 9368 }, { "epoch": 0.73, "learning_rate": 9.265349329569514e-05, "loss": 3.0306, "step": 9369 }, { "epoch": 0.73, "learning_rate": 9.265270916647064e-05, "loss": 3.1805, "step": 9370 }, { "epoch": 0.73, "learning_rate": 9.265192503724615e-05, "loss": 2.818, "step": 9371 }, { "epoch": 0.73, "learning_rate": 9.265114090802165e-05, "loss": 2.7635, "step": 9372 }, { "epoch": 0.73, "learning_rate": 9.265035677879715e-05, "loss": 2.9898, "step": 9373 }, { "epoch": 0.74, "learning_rate": 9.264957264957266e-05, "loss": 3.0286, "step": 9374 }, { "epoch": 0.74, "learning_rate": 9.264878852034816e-05, "loss": 3.0096, "step": 9375 }, { "epoch": 0.74, "learning_rate": 9.264800439112367e-05, "loss": 3.008, "step": 9376 }, { "epoch": 0.74, "learning_rate": 9.264722026189917e-05, "loss": 2.9138, "step": 9377 }, { "epoch": 0.74, "learning_rate": 9.264643613267467e-05, "loss": 3.324, "step": 9378 }, { "epoch": 0.74, "learning_rate": 9.264565200345018e-05, "loss": 3.0438, "step": 9379 }, { "epoch": 0.74, "learning_rate": 9.264486787422567e-05, "loss": 3.0868, "step": 9380 }, { "epoch": 0.74, "learning_rate": 9.264408374500117e-05, "loss": 2.8805, "step": 9381 }, { "epoch": 0.74, "learning_rate": 9.264329961577667e-05, "loss": 2.9433, "step": 9382 }, { "epoch": 0.74, "learning_rate": 9.264251548655218e-05, "loss": 3.091, "step": 9383 }, { "epoch": 0.74, "learning_rate": 9.264173135732768e-05, "loss": 2.9739, "step": 9384 }, { "epoch": 0.74, "learning_rate": 9.264094722810319e-05, "loss": 3.1517, "step": 9385 }, { "epoch": 0.74, "learning_rate": 9.26401630988787e-05, "loss": 3.1402, "step": 9386 }, { "epoch": 0.74, "learning_rate": 9.26393789696542e-05, "loss": 2.8953, "step": 9387 }, { "epoch": 0.74, "learning_rate": 9.263859484042971e-05, "loss": 3.0744, "step": 9388 }, { "epoch": 0.74, "learning_rate": 9.263781071120521e-05, "loss": 2.9899, "step": 9389 }, { "epoch": 0.74, "learning_rate": 9.263702658198072e-05, "loss": 2.7997, "step": 9390 }, { "epoch": 0.74, "learning_rate": 9.263624245275622e-05, "loss": 3.4339, "step": 9391 }, { "epoch": 0.74, "learning_rate": 9.263545832353172e-05, "loss": 3.049, "step": 9392 }, { "epoch": 0.74, "learning_rate": 9.263467419430723e-05, "loss": 2.8715, "step": 9393 }, { "epoch": 0.74, "learning_rate": 9.263389006508273e-05, "loss": 3.0096, "step": 9394 }, { "epoch": 0.74, "learning_rate": 9.263310593585824e-05, "loss": 2.9001, "step": 9395 }, { "epoch": 0.74, "learning_rate": 9.263232180663374e-05, "loss": 2.9651, "step": 9396 }, { "epoch": 0.74, "learning_rate": 9.263153767740924e-05, "loss": 3.0395, "step": 9397 }, { "epoch": 0.74, "learning_rate": 9.263075354818475e-05, "loss": 3.0246, "step": 9398 }, { "epoch": 0.74, "learning_rate": 9.262996941896025e-05, "loss": 2.9244, "step": 9399 }, { "epoch": 0.74, "learning_rate": 9.262918528973575e-05, "loss": 2.7875, "step": 9400 }, { "epoch": 0.74, "learning_rate": 9.262840116051126e-05, "loss": 2.8275, "step": 9401 }, { "epoch": 0.74, "learning_rate": 9.262761703128676e-05, "loss": 2.9684, "step": 9402 }, { "epoch": 0.74, "learning_rate": 9.262683290206227e-05, "loss": 3.1609, "step": 9403 }, { "epoch": 0.74, "learning_rate": 9.262604877283777e-05, "loss": 3.221, "step": 9404 }, { "epoch": 0.74, "learning_rate": 9.262526464361327e-05, "loss": 2.7916, "step": 9405 }, { "epoch": 0.74, "learning_rate": 9.262448051438878e-05, "loss": 3.1362, "step": 9406 }, { "epoch": 0.74, "learning_rate": 9.262369638516428e-05, "loss": 3.2069, "step": 9407 }, { "epoch": 0.74, "learning_rate": 9.262291225593978e-05, "loss": 3.2136, "step": 9408 }, { "epoch": 0.74, "learning_rate": 9.262212812671529e-05, "loss": 3.061, "step": 9409 }, { "epoch": 0.74, "learning_rate": 9.262134399749079e-05, "loss": 2.9704, "step": 9410 }, { "epoch": 0.74, "learning_rate": 9.26205598682663e-05, "loss": 2.9872, "step": 9411 }, { "epoch": 0.74, "learning_rate": 9.26197757390418e-05, "loss": 2.8723, "step": 9412 }, { "epoch": 0.74, "learning_rate": 9.26189916098173e-05, "loss": 2.8544, "step": 9413 }, { "epoch": 0.74, "learning_rate": 9.26182074805928e-05, "loss": 2.9846, "step": 9414 }, { "epoch": 0.74, "learning_rate": 9.261742335136831e-05, "loss": 3.3154, "step": 9415 }, { "epoch": 0.74, "learning_rate": 9.261663922214381e-05, "loss": 3.2658, "step": 9416 }, { "epoch": 0.74, "learning_rate": 9.261585509291932e-05, "loss": 3.1112, "step": 9417 }, { "epoch": 0.74, "learning_rate": 9.261507096369482e-05, "loss": 2.9713, "step": 9418 }, { "epoch": 0.74, "learning_rate": 9.261428683447032e-05, "loss": 3.0209, "step": 9419 }, { "epoch": 0.74, "learning_rate": 9.261350270524583e-05, "loss": 3.0253, "step": 9420 }, { "epoch": 0.74, "learning_rate": 9.261271857602133e-05, "loss": 3.0321, "step": 9421 }, { "epoch": 0.74, "learning_rate": 9.261193444679684e-05, "loss": 3.4571, "step": 9422 }, { "epoch": 0.74, "learning_rate": 9.261115031757234e-05, "loss": 3.098, "step": 9423 }, { "epoch": 0.74, "learning_rate": 9.261036618834784e-05, "loss": 3.0147, "step": 9424 }, { "epoch": 0.74, "learning_rate": 9.260958205912335e-05, "loss": 2.987, "step": 9425 }, { "epoch": 0.74, "learning_rate": 9.260879792989885e-05, "loss": 3.0585, "step": 9426 }, { "epoch": 0.74, "learning_rate": 9.260801380067435e-05, "loss": 3.0879, "step": 9427 }, { "epoch": 0.74, "learning_rate": 9.260722967144986e-05, "loss": 3.1436, "step": 9428 }, { "epoch": 0.74, "learning_rate": 9.260644554222536e-05, "loss": 3.1451, "step": 9429 }, { "epoch": 0.74, "learning_rate": 9.260566141300086e-05, "loss": 2.9429, "step": 9430 }, { "epoch": 0.74, "learning_rate": 9.260487728377637e-05, "loss": 3.1376, "step": 9431 }, { "epoch": 0.74, "learning_rate": 9.260409315455187e-05, "loss": 3.079, "step": 9432 }, { "epoch": 0.74, "learning_rate": 9.260330902532738e-05, "loss": 3.5176, "step": 9433 }, { "epoch": 0.74, "learning_rate": 9.260252489610288e-05, "loss": 3.2033, "step": 9434 }, { "epoch": 0.74, "learning_rate": 9.260174076687838e-05, "loss": 2.8312, "step": 9435 }, { "epoch": 0.74, "learning_rate": 9.260095663765389e-05, "loss": 3.2307, "step": 9436 }, { "epoch": 0.74, "learning_rate": 9.260017250842939e-05, "loss": 3.0237, "step": 9437 }, { "epoch": 0.74, "learning_rate": 9.25993883792049e-05, "loss": 2.8609, "step": 9438 }, { "epoch": 0.74, "learning_rate": 9.25986042499804e-05, "loss": 3.2481, "step": 9439 }, { "epoch": 0.74, "learning_rate": 9.25978201207559e-05, "loss": 3.113, "step": 9440 }, { "epoch": 0.74, "learning_rate": 9.25970359915314e-05, "loss": 3.1508, "step": 9441 }, { "epoch": 0.74, "learning_rate": 9.259625186230691e-05, "loss": 3.1329, "step": 9442 }, { "epoch": 0.74, "learning_rate": 9.259546773308243e-05, "loss": 2.7524, "step": 9443 }, { "epoch": 0.74, "learning_rate": 9.259468360385793e-05, "loss": 2.861, "step": 9444 }, { "epoch": 0.74, "learning_rate": 9.259389947463343e-05, "loss": 3.0561, "step": 9445 }, { "epoch": 0.74, "learning_rate": 9.259311534540894e-05, "loss": 2.9935, "step": 9446 }, { "epoch": 0.74, "learning_rate": 9.259233121618444e-05, "loss": 2.8869, "step": 9447 }, { "epoch": 0.74, "learning_rate": 9.259154708695994e-05, "loss": 2.9808, "step": 9448 }, { "epoch": 0.74, "learning_rate": 9.259076295773545e-05, "loss": 2.9308, "step": 9449 }, { "epoch": 0.74, "learning_rate": 9.258997882851095e-05, "loss": 3.1048, "step": 9450 }, { "epoch": 0.74, "learning_rate": 9.258919469928646e-05, "loss": 3.1517, "step": 9451 }, { "epoch": 0.74, "learning_rate": 9.258841057006195e-05, "loss": 3.3215, "step": 9452 }, { "epoch": 0.74, "learning_rate": 9.258762644083745e-05, "loss": 3.0531, "step": 9453 }, { "epoch": 0.74, "learning_rate": 9.258684231161295e-05, "loss": 3.0335, "step": 9454 }, { "epoch": 0.74, "learning_rate": 9.258605818238846e-05, "loss": 2.8735, "step": 9455 }, { "epoch": 0.74, "learning_rate": 9.258527405316396e-05, "loss": 3.1238, "step": 9456 }, { "epoch": 0.74, "learning_rate": 9.258448992393946e-05, "loss": 2.9304, "step": 9457 }, { "epoch": 0.74, "learning_rate": 9.258370579471497e-05, "loss": 3.0334, "step": 9458 }, { "epoch": 0.74, "learning_rate": 9.258292166549047e-05, "loss": 2.9046, "step": 9459 }, { "epoch": 0.74, "learning_rate": 9.258213753626597e-05, "loss": 2.7911, "step": 9460 }, { "epoch": 0.74, "learning_rate": 9.258135340704148e-05, "loss": 2.9492, "step": 9461 }, { "epoch": 0.74, "learning_rate": 9.258056927781698e-05, "loss": 2.9561, "step": 9462 }, { "epoch": 0.74, "learning_rate": 9.257978514859249e-05, "loss": 3.0605, "step": 9463 }, { "epoch": 0.74, "learning_rate": 9.257900101936799e-05, "loss": 3.0709, "step": 9464 }, { "epoch": 0.74, "learning_rate": 9.25782168901435e-05, "loss": 2.8637, "step": 9465 }, { "epoch": 0.74, "learning_rate": 9.2577432760919e-05, "loss": 2.9825, "step": 9466 }, { "epoch": 0.74, "learning_rate": 9.25766486316945e-05, "loss": 3.1027, "step": 9467 }, { "epoch": 0.74, "learning_rate": 9.257586450247002e-05, "loss": 2.9469, "step": 9468 }, { "epoch": 0.74, "learning_rate": 9.257508037324552e-05, "loss": 3.0082, "step": 9469 }, { "epoch": 0.74, "learning_rate": 9.257429624402103e-05, "loss": 3.0201, "step": 9470 }, { "epoch": 0.74, "learning_rate": 9.257351211479653e-05, "loss": 2.8967, "step": 9471 }, { "epoch": 0.74, "learning_rate": 9.257272798557203e-05, "loss": 2.8878, "step": 9472 }, { "epoch": 0.74, "learning_rate": 9.257194385634754e-05, "loss": 3.1021, "step": 9473 }, { "epoch": 0.74, "learning_rate": 9.257115972712304e-05, "loss": 3.084, "step": 9474 }, { "epoch": 0.74, "learning_rate": 9.257037559789854e-05, "loss": 3.1212, "step": 9475 }, { "epoch": 0.74, "learning_rate": 9.256959146867405e-05, "loss": 3.318, "step": 9476 }, { "epoch": 0.74, "learning_rate": 9.256880733944955e-05, "loss": 3.1115, "step": 9477 }, { "epoch": 0.74, "learning_rate": 9.256802321022505e-05, "loss": 2.8465, "step": 9478 }, { "epoch": 0.74, "learning_rate": 9.256723908100056e-05, "loss": 3.0374, "step": 9479 }, { "epoch": 0.74, "learning_rate": 9.256645495177606e-05, "loss": 3.0155, "step": 9480 }, { "epoch": 0.74, "learning_rate": 9.256567082255157e-05, "loss": 2.9183, "step": 9481 }, { "epoch": 0.74, "learning_rate": 9.256488669332707e-05, "loss": 3.0637, "step": 9482 }, { "epoch": 0.74, "learning_rate": 9.256410256410257e-05, "loss": 2.8948, "step": 9483 }, { "epoch": 0.74, "learning_rate": 9.256331843487808e-05, "loss": 3.0312, "step": 9484 }, { "epoch": 0.74, "learning_rate": 9.256253430565358e-05, "loss": 3.6425, "step": 9485 }, { "epoch": 0.74, "learning_rate": 9.256175017642908e-05, "loss": 3.1413, "step": 9486 }, { "epoch": 0.74, "learning_rate": 9.256096604720459e-05, "loss": 2.9713, "step": 9487 }, { "epoch": 0.74, "learning_rate": 9.256018191798008e-05, "loss": 3.0566, "step": 9488 }, { "epoch": 0.74, "learning_rate": 9.255939778875558e-05, "loss": 2.9236, "step": 9489 }, { "epoch": 0.74, "learning_rate": 9.255861365953109e-05, "loss": 3.0838, "step": 9490 }, { "epoch": 0.74, "learning_rate": 9.255782953030659e-05, "loss": 2.8075, "step": 9491 }, { "epoch": 0.74, "learning_rate": 9.255704540108209e-05, "loss": 3.0457, "step": 9492 }, { "epoch": 0.74, "learning_rate": 9.25562612718576e-05, "loss": 2.9488, "step": 9493 }, { "epoch": 0.74, "learning_rate": 9.255547714263311e-05, "loss": 2.9706, "step": 9494 }, { "epoch": 0.74, "learning_rate": 9.255469301340862e-05, "loss": 2.9548, "step": 9495 }, { "epoch": 0.74, "learning_rate": 9.255390888418412e-05, "loss": 3.1575, "step": 9496 }, { "epoch": 0.74, "learning_rate": 9.255312475495962e-05, "loss": 3.2365, "step": 9497 }, { "epoch": 0.74, "learning_rate": 9.255234062573513e-05, "loss": 3.1062, "step": 9498 }, { "epoch": 0.74, "learning_rate": 9.255155649651063e-05, "loss": 3.0095, "step": 9499 }, { "epoch": 0.74, "learning_rate": 9.255077236728614e-05, "loss": 3.1002, "step": 9500 }, { "epoch": 0.75, "learning_rate": 9.254998823806164e-05, "loss": 3.0219, "step": 9501 }, { "epoch": 0.75, "learning_rate": 9.254920410883714e-05, "loss": 3.0699, "step": 9502 }, { "epoch": 0.75, "learning_rate": 9.254841997961265e-05, "loss": 3.0595, "step": 9503 }, { "epoch": 0.75, "learning_rate": 9.254763585038815e-05, "loss": 3.1189, "step": 9504 }, { "epoch": 0.75, "learning_rate": 9.254685172116365e-05, "loss": 3.14, "step": 9505 }, { "epoch": 0.75, "learning_rate": 9.254606759193916e-05, "loss": 3.2458, "step": 9506 }, { "epoch": 0.75, "learning_rate": 9.254528346271466e-05, "loss": 2.8046, "step": 9507 }, { "epoch": 0.75, "learning_rate": 9.254449933349016e-05, "loss": 3.0655, "step": 9508 }, { "epoch": 0.75, "learning_rate": 9.254371520426567e-05, "loss": 3.0333, "step": 9509 }, { "epoch": 0.75, "learning_rate": 9.254293107504117e-05, "loss": 2.9493, "step": 9510 }, { "epoch": 0.75, "learning_rate": 9.254214694581668e-05, "loss": 2.9337, "step": 9511 }, { "epoch": 0.75, "learning_rate": 9.254136281659218e-05, "loss": 2.9882, "step": 9512 }, { "epoch": 0.75, "learning_rate": 9.254057868736768e-05, "loss": 2.9667, "step": 9513 }, { "epoch": 0.75, "learning_rate": 9.253979455814319e-05, "loss": 2.7751, "step": 9514 }, { "epoch": 0.75, "learning_rate": 9.253901042891869e-05, "loss": 3.2193, "step": 9515 }, { "epoch": 0.75, "learning_rate": 9.25382262996942e-05, "loss": 3.0951, "step": 9516 }, { "epoch": 0.75, "learning_rate": 9.25374421704697e-05, "loss": 3.1535, "step": 9517 }, { "epoch": 0.75, "learning_rate": 9.25366580412452e-05, "loss": 2.8629, "step": 9518 }, { "epoch": 0.75, "learning_rate": 9.25358739120207e-05, "loss": 3.0098, "step": 9519 }, { "epoch": 0.75, "learning_rate": 9.253508978279621e-05, "loss": 3.285, "step": 9520 }, { "epoch": 0.75, "learning_rate": 9.253430565357171e-05, "loss": 2.9351, "step": 9521 }, { "epoch": 0.75, "learning_rate": 9.253352152434722e-05, "loss": 2.9985, "step": 9522 }, { "epoch": 0.75, "learning_rate": 9.253273739512272e-05, "loss": 3.0974, "step": 9523 }, { "epoch": 0.75, "learning_rate": 9.253195326589822e-05, "loss": 3.2245, "step": 9524 }, { "epoch": 0.75, "learning_rate": 9.253116913667373e-05, "loss": 3.2944, "step": 9525 }, { "epoch": 0.75, "learning_rate": 9.253038500744923e-05, "loss": 3.0721, "step": 9526 }, { "epoch": 0.75, "learning_rate": 9.252960087822473e-05, "loss": 3.1141, "step": 9527 }, { "epoch": 0.75, "learning_rate": 9.252881674900024e-05, "loss": 3.0871, "step": 9528 }, { "epoch": 0.75, "learning_rate": 9.252803261977574e-05, "loss": 2.943, "step": 9529 }, { "epoch": 0.75, "learning_rate": 9.252724849055125e-05, "loss": 3.043, "step": 9530 }, { "epoch": 0.75, "learning_rate": 9.252646436132675e-05, "loss": 3.2125, "step": 9531 }, { "epoch": 0.75, "learning_rate": 9.252568023210225e-05, "loss": 3.1775, "step": 9532 }, { "epoch": 0.75, "learning_rate": 9.252489610287776e-05, "loss": 2.9781, "step": 9533 }, { "epoch": 0.75, "learning_rate": 9.252411197365326e-05, "loss": 2.8108, "step": 9534 }, { "epoch": 0.75, "learning_rate": 9.252332784442876e-05, "loss": 2.9744, "step": 9535 }, { "epoch": 0.75, "learning_rate": 9.252254371520427e-05, "loss": 3.011, "step": 9536 }, { "epoch": 0.75, "learning_rate": 9.252175958597977e-05, "loss": 3.2338, "step": 9537 }, { "epoch": 0.75, "learning_rate": 9.252097545675528e-05, "loss": 3.0402, "step": 9538 }, { "epoch": 0.75, "learning_rate": 9.252019132753078e-05, "loss": 2.799, "step": 9539 }, { "epoch": 0.75, "learning_rate": 9.251940719830628e-05, "loss": 3.1184, "step": 9540 }, { "epoch": 0.75, "learning_rate": 9.251862306908179e-05, "loss": 3.1459, "step": 9541 }, { "epoch": 0.75, "learning_rate": 9.251783893985729e-05, "loss": 2.9018, "step": 9542 }, { "epoch": 0.75, "learning_rate": 9.25170548106328e-05, "loss": 3.0628, "step": 9543 }, { "epoch": 0.75, "learning_rate": 9.25162706814083e-05, "loss": 2.9786, "step": 9544 }, { "epoch": 0.75, "learning_rate": 9.25154865521838e-05, "loss": 2.9274, "step": 9545 }, { "epoch": 0.75, "learning_rate": 9.25147024229593e-05, "loss": 3.4547, "step": 9546 }, { "epoch": 0.75, "learning_rate": 9.251391829373481e-05, "loss": 2.9294, "step": 9547 }, { "epoch": 0.75, "learning_rate": 9.251313416451031e-05, "loss": 3.0231, "step": 9548 }, { "epoch": 0.75, "learning_rate": 9.251235003528582e-05, "loss": 3.0178, "step": 9549 }, { "epoch": 0.75, "learning_rate": 9.251156590606132e-05, "loss": 3.2293, "step": 9550 }, { "epoch": 0.75, "learning_rate": 9.251078177683684e-05, "loss": 3.2265, "step": 9551 }, { "epoch": 0.75, "learning_rate": 9.250999764761234e-05, "loss": 2.8807, "step": 9552 }, { "epoch": 0.75, "learning_rate": 9.250921351838784e-05, "loss": 2.9993, "step": 9553 }, { "epoch": 0.75, "learning_rate": 9.250842938916335e-05, "loss": 3.2014, "step": 9554 }, { "epoch": 0.75, "learning_rate": 9.250764525993885e-05, "loss": 3.0521, "step": 9555 }, { "epoch": 0.75, "learning_rate": 9.250686113071436e-05, "loss": 2.9501, "step": 9556 }, { "epoch": 0.75, "learning_rate": 9.250607700148986e-05, "loss": 3.0188, "step": 9557 }, { "epoch": 0.75, "learning_rate": 9.250529287226536e-05, "loss": 2.9631, "step": 9558 }, { "epoch": 0.75, "learning_rate": 9.250450874304087e-05, "loss": 3.0867, "step": 9559 }, { "epoch": 0.75, "learning_rate": 9.250372461381636e-05, "loss": 3.2481, "step": 9560 }, { "epoch": 0.75, "learning_rate": 9.250294048459186e-05, "loss": 3.192, "step": 9561 }, { "epoch": 0.75, "learning_rate": 9.250215635536736e-05, "loss": 3.0309, "step": 9562 }, { "epoch": 0.75, "learning_rate": 9.250137222614287e-05, "loss": 2.9334, "step": 9563 }, { "epoch": 0.75, "learning_rate": 9.250058809691837e-05, "loss": 3.0405, "step": 9564 }, { "epoch": 0.75, "learning_rate": 9.249980396769387e-05, "loss": 3.2149, "step": 9565 }, { "epoch": 0.75, "learning_rate": 9.249901983846938e-05, "loss": 3.192, "step": 9566 }, { "epoch": 0.75, "learning_rate": 9.249823570924488e-05, "loss": 2.9895, "step": 9567 }, { "epoch": 0.75, "learning_rate": 9.249745158002039e-05, "loss": 2.9406, "step": 9568 }, { "epoch": 0.75, "learning_rate": 9.249666745079589e-05, "loss": 2.9402, "step": 9569 }, { "epoch": 0.75, "learning_rate": 9.249588332157139e-05, "loss": 3.1619, "step": 9570 }, { "epoch": 0.75, "learning_rate": 9.24950991923469e-05, "loss": 2.901, "step": 9571 }, { "epoch": 0.75, "learning_rate": 9.24943150631224e-05, "loss": 3.085, "step": 9572 }, { "epoch": 0.75, "learning_rate": 9.24935309338979e-05, "loss": 2.8877, "step": 9573 }, { "epoch": 0.75, "learning_rate": 9.249274680467341e-05, "loss": 3.0669, "step": 9574 }, { "epoch": 0.75, "learning_rate": 9.249196267544891e-05, "loss": 2.9445, "step": 9575 }, { "epoch": 0.75, "learning_rate": 9.249117854622442e-05, "loss": 3.1426, "step": 9576 }, { "epoch": 0.75, "learning_rate": 9.249039441699993e-05, "loss": 2.8958, "step": 9577 }, { "epoch": 0.75, "learning_rate": 9.248961028777544e-05, "loss": 3.3969, "step": 9578 }, { "epoch": 0.75, "learning_rate": 9.248882615855094e-05, "loss": 2.8681, "step": 9579 }, { "epoch": 0.75, "learning_rate": 9.248804202932644e-05, "loss": 3.262, "step": 9580 }, { "epoch": 0.75, "learning_rate": 9.248725790010195e-05, "loss": 2.9246, "step": 9581 }, { "epoch": 0.75, "learning_rate": 9.248647377087745e-05, "loss": 3.4684, "step": 9582 }, { "epoch": 0.75, "learning_rate": 9.248568964165295e-05, "loss": 3.1573, "step": 9583 }, { "epoch": 0.75, "learning_rate": 9.248490551242846e-05, "loss": 3.2443, "step": 9584 }, { "epoch": 0.75, "learning_rate": 9.248412138320396e-05, "loss": 3.09, "step": 9585 }, { "epoch": 0.75, "learning_rate": 9.248333725397947e-05, "loss": 3.1245, "step": 9586 }, { "epoch": 0.75, "learning_rate": 9.248255312475497e-05, "loss": 3.0524, "step": 9587 }, { "epoch": 0.75, "learning_rate": 9.248176899553047e-05, "loss": 3.288, "step": 9588 }, { "epoch": 0.75, "learning_rate": 9.248098486630598e-05, "loss": 2.9395, "step": 9589 }, { "epoch": 0.75, "learning_rate": 9.248020073708148e-05, "loss": 2.9514, "step": 9590 }, { "epoch": 0.75, "learning_rate": 9.247941660785698e-05, "loss": 3.11, "step": 9591 }, { "epoch": 0.75, "learning_rate": 9.247863247863249e-05, "loss": 3.0778, "step": 9592 }, { "epoch": 0.75, "learning_rate": 9.247784834940799e-05, "loss": 3.161, "step": 9593 }, { "epoch": 0.75, "learning_rate": 9.24770642201835e-05, "loss": 3.2335, "step": 9594 }, { "epoch": 0.75, "learning_rate": 9.2476280090959e-05, "loss": 2.9909, "step": 9595 }, { "epoch": 0.75, "learning_rate": 9.247549596173449e-05, "loss": 3.0154, "step": 9596 }, { "epoch": 0.75, "learning_rate": 9.247471183250999e-05, "loss": 3.1266, "step": 9597 }, { "epoch": 0.75, "learning_rate": 9.24739277032855e-05, "loss": 2.8598, "step": 9598 }, { "epoch": 0.75, "learning_rate": 9.2473143574061e-05, "loss": 3.1658, "step": 9599 }, { "epoch": 0.75, "learning_rate": 9.24723594448365e-05, "loss": 3.0804, "step": 9600 }, { "epoch": 0.75, "learning_rate": 9.247157531561201e-05, "loss": 3.1455, "step": 9601 }, { "epoch": 0.75, "learning_rate": 9.247079118638752e-05, "loss": 3.1222, "step": 9602 }, { "epoch": 0.75, "learning_rate": 9.247000705716303e-05, "loss": 3.2318, "step": 9603 }, { "epoch": 0.75, "learning_rate": 9.246922292793853e-05, "loss": 2.7846, "step": 9604 }, { "epoch": 0.75, "learning_rate": 9.246843879871404e-05, "loss": 3.1041, "step": 9605 }, { "epoch": 0.75, "learning_rate": 9.246765466948954e-05, "loss": 3.1249, "step": 9606 }, { "epoch": 0.75, "learning_rate": 9.246687054026504e-05, "loss": 3.0852, "step": 9607 }, { "epoch": 0.75, "learning_rate": 9.246608641104055e-05, "loss": 3.0833, "step": 9608 }, { "epoch": 0.75, "learning_rate": 9.246530228181605e-05, "loss": 2.8709, "step": 9609 }, { "epoch": 0.75, "learning_rate": 9.246451815259155e-05, "loss": 3.0836, "step": 9610 }, { "epoch": 0.75, "learning_rate": 9.246373402336706e-05, "loss": 2.9637, "step": 9611 }, { "epoch": 0.75, "learning_rate": 9.246294989414256e-05, "loss": 2.9487, "step": 9612 }, { "epoch": 0.75, "learning_rate": 9.246216576491806e-05, "loss": 3.0445, "step": 9613 }, { "epoch": 0.75, "learning_rate": 9.246138163569357e-05, "loss": 3.0667, "step": 9614 }, { "epoch": 0.75, "learning_rate": 9.246059750646907e-05, "loss": 2.9585, "step": 9615 }, { "epoch": 0.75, "learning_rate": 9.245981337724458e-05, "loss": 2.9494, "step": 9616 }, { "epoch": 0.75, "learning_rate": 9.245902924802008e-05, "loss": 2.9216, "step": 9617 }, { "epoch": 0.75, "learning_rate": 9.245824511879558e-05, "loss": 2.7731, "step": 9618 }, { "epoch": 0.75, "learning_rate": 9.245746098957109e-05, "loss": 2.8677, "step": 9619 }, { "epoch": 0.75, "learning_rate": 9.245667686034659e-05, "loss": 3.1986, "step": 9620 }, { "epoch": 0.75, "learning_rate": 9.24558927311221e-05, "loss": 2.9593, "step": 9621 }, { "epoch": 0.75, "learning_rate": 9.24551086018976e-05, "loss": 3.1185, "step": 9622 }, { "epoch": 0.75, "learning_rate": 9.24543244726731e-05, "loss": 3.0144, "step": 9623 }, { "epoch": 0.75, "learning_rate": 9.24535403434486e-05, "loss": 3.1426, "step": 9624 }, { "epoch": 0.75, "learning_rate": 9.245275621422411e-05, "loss": 3.0748, "step": 9625 }, { "epoch": 0.75, "learning_rate": 9.245197208499961e-05, "loss": 3.2686, "step": 9626 }, { "epoch": 0.75, "learning_rate": 9.245118795577512e-05, "loss": 3.2857, "step": 9627 }, { "epoch": 0.75, "learning_rate": 9.245040382655062e-05, "loss": 3.4234, "step": 9628 }, { "epoch": 0.76, "learning_rate": 9.244961969732612e-05, "loss": 3.3981, "step": 9629 }, { "epoch": 0.76, "learning_rate": 9.244883556810163e-05, "loss": 3.0528, "step": 9630 }, { "epoch": 0.76, "learning_rate": 9.244805143887713e-05, "loss": 3.0122, "step": 9631 }, { "epoch": 0.76, "learning_rate": 9.244726730965263e-05, "loss": 3.0906, "step": 9632 }, { "epoch": 0.76, "learning_rate": 9.244648318042814e-05, "loss": 2.9464, "step": 9633 }, { "epoch": 0.76, "learning_rate": 9.244569905120364e-05, "loss": 3.2452, "step": 9634 }, { "epoch": 0.76, "learning_rate": 9.244491492197915e-05, "loss": 2.991, "step": 9635 }, { "epoch": 0.76, "learning_rate": 9.244413079275465e-05, "loss": 3.3865, "step": 9636 }, { "epoch": 0.76, "learning_rate": 9.244334666353015e-05, "loss": 2.7571, "step": 9637 }, { "epoch": 0.76, "learning_rate": 9.244256253430566e-05, "loss": 3.2114, "step": 9638 }, { "epoch": 0.76, "learning_rate": 9.244177840508116e-05, "loss": 2.9755, "step": 9639 }, { "epoch": 0.76, "learning_rate": 9.244099427585666e-05, "loss": 2.9393, "step": 9640 }, { "epoch": 0.76, "learning_rate": 9.244021014663217e-05, "loss": 3.3157, "step": 9641 }, { "epoch": 0.76, "learning_rate": 9.243942601740767e-05, "loss": 2.9131, "step": 9642 }, { "epoch": 0.76, "learning_rate": 9.243864188818318e-05, "loss": 2.9435, "step": 9643 }, { "epoch": 0.76, "learning_rate": 9.243785775895868e-05, "loss": 2.9569, "step": 9644 }, { "epoch": 0.76, "learning_rate": 9.243707362973418e-05, "loss": 3.2122, "step": 9645 }, { "epoch": 0.76, "learning_rate": 9.243628950050969e-05, "loss": 2.9956, "step": 9646 }, { "epoch": 0.76, "learning_rate": 9.243550537128519e-05, "loss": 3.1692, "step": 9647 }, { "epoch": 0.76, "learning_rate": 9.24347212420607e-05, "loss": 2.9069, "step": 9648 }, { "epoch": 0.76, "learning_rate": 9.24339371128362e-05, "loss": 2.9762, "step": 9649 }, { "epoch": 0.76, "learning_rate": 9.24331529836117e-05, "loss": 3.009, "step": 9650 }, { "epoch": 0.76, "learning_rate": 9.24323688543872e-05, "loss": 3.0692, "step": 9651 }, { "epoch": 0.76, "learning_rate": 9.243158472516271e-05, "loss": 2.9206, "step": 9652 }, { "epoch": 0.76, "learning_rate": 9.243080059593821e-05, "loss": 2.9728, "step": 9653 }, { "epoch": 0.76, "learning_rate": 9.243001646671372e-05, "loss": 2.8353, "step": 9654 }, { "epoch": 0.76, "learning_rate": 9.242923233748922e-05, "loss": 3.184, "step": 9655 }, { "epoch": 0.76, "learning_rate": 9.242844820826472e-05, "loss": 2.7451, "step": 9656 }, { "epoch": 0.76, "learning_rate": 9.242766407904023e-05, "loss": 3.1732, "step": 9657 }, { "epoch": 0.76, "learning_rate": 9.242687994981573e-05, "loss": 2.7026, "step": 9658 }, { "epoch": 0.76, "learning_rate": 9.242609582059125e-05, "loss": 3.1523, "step": 9659 }, { "epoch": 0.76, "learning_rate": 9.242531169136675e-05, "loss": 3.0315, "step": 9660 }, { "epoch": 0.76, "learning_rate": 9.242452756214225e-05, "loss": 3.0818, "step": 9661 }, { "epoch": 0.76, "learning_rate": 9.242374343291776e-05, "loss": 2.9113, "step": 9662 }, { "epoch": 0.76, "learning_rate": 9.242295930369326e-05, "loss": 3.118, "step": 9663 }, { "epoch": 0.76, "learning_rate": 9.242217517446877e-05, "loss": 2.9165, "step": 9664 }, { "epoch": 0.76, "learning_rate": 9.242139104524427e-05, "loss": 2.881, "step": 9665 }, { "epoch": 0.76, "learning_rate": 9.242060691601977e-05, "loss": 3.006, "step": 9666 }, { "epoch": 0.76, "learning_rate": 9.241982278679528e-05, "loss": 3.002, "step": 9667 }, { "epoch": 0.76, "learning_rate": 9.241903865757077e-05, "loss": 3.2034, "step": 9668 }, { "epoch": 0.76, "learning_rate": 9.241825452834627e-05, "loss": 3.2938, "step": 9669 }, { "epoch": 0.76, "learning_rate": 9.241747039912177e-05, "loss": 3.078, "step": 9670 }, { "epoch": 0.76, "learning_rate": 9.241668626989728e-05, "loss": 3.2116, "step": 9671 }, { "epoch": 0.76, "learning_rate": 9.241590214067278e-05, "loss": 3.0275, "step": 9672 }, { "epoch": 0.76, "learning_rate": 9.241511801144829e-05, "loss": 3.1212, "step": 9673 }, { "epoch": 0.76, "learning_rate": 9.241433388222379e-05, "loss": 2.7962, "step": 9674 }, { "epoch": 0.76, "learning_rate": 9.241354975299929e-05, "loss": 3.0333, "step": 9675 }, { "epoch": 0.76, "learning_rate": 9.24127656237748e-05, "loss": 3.0882, "step": 9676 }, { "epoch": 0.76, "learning_rate": 9.24119814945503e-05, "loss": 3.0102, "step": 9677 }, { "epoch": 0.76, "learning_rate": 9.24111973653258e-05, "loss": 2.9539, "step": 9678 }, { "epoch": 0.76, "learning_rate": 9.241041323610131e-05, "loss": 3.1471, "step": 9679 }, { "epoch": 0.76, "learning_rate": 9.240962910687681e-05, "loss": 2.903, "step": 9680 }, { "epoch": 0.76, "learning_rate": 9.240884497765231e-05, "loss": 2.9996, "step": 9681 }, { "epoch": 0.76, "learning_rate": 9.240806084842782e-05, "loss": 3.5048, "step": 9682 }, { "epoch": 0.76, "learning_rate": 9.240727671920332e-05, "loss": 2.9957, "step": 9683 }, { "epoch": 0.76, "learning_rate": 9.240649258997883e-05, "loss": 3.0029, "step": 9684 }, { "epoch": 0.76, "learning_rate": 9.240570846075434e-05, "loss": 3.0106, "step": 9685 }, { "epoch": 0.76, "learning_rate": 9.240492433152985e-05, "loss": 2.9194, "step": 9686 }, { "epoch": 0.76, "learning_rate": 9.240414020230535e-05, "loss": 3.0668, "step": 9687 }, { "epoch": 0.76, "learning_rate": 9.240335607308085e-05, "loss": 2.8673, "step": 9688 }, { "epoch": 0.76, "learning_rate": 9.240257194385636e-05, "loss": 2.914, "step": 9689 }, { "epoch": 0.76, "learning_rate": 9.240178781463186e-05, "loss": 3.0507, "step": 9690 }, { "epoch": 0.76, "learning_rate": 9.240100368540737e-05, "loss": 2.8426, "step": 9691 }, { "epoch": 0.76, "learning_rate": 9.240021955618287e-05, "loss": 2.9828, "step": 9692 }, { "epoch": 0.76, "learning_rate": 9.239943542695837e-05, "loss": 3.1391, "step": 9693 }, { "epoch": 0.76, "learning_rate": 9.239865129773388e-05, "loss": 3.1211, "step": 9694 }, { "epoch": 0.76, "learning_rate": 9.239786716850938e-05, "loss": 2.7243, "step": 9695 }, { "epoch": 0.76, "learning_rate": 9.239708303928488e-05, "loss": 3.1048, "step": 9696 }, { "epoch": 0.76, "learning_rate": 9.239629891006039e-05, "loss": 3.2744, "step": 9697 }, { "epoch": 0.76, "learning_rate": 9.239551478083589e-05, "loss": 3.0712, "step": 9698 }, { "epoch": 0.76, "learning_rate": 9.23947306516114e-05, "loss": 3.06, "step": 9699 }, { "epoch": 0.76, "learning_rate": 9.23939465223869e-05, "loss": 2.8935, "step": 9700 }, { "epoch": 0.76, "learning_rate": 9.23931623931624e-05, "loss": 3.0928, "step": 9701 }, { "epoch": 0.76, "learning_rate": 9.23923782639379e-05, "loss": 2.9971, "step": 9702 }, { "epoch": 0.76, "learning_rate": 9.239159413471341e-05, "loss": 3.3431, "step": 9703 }, { "epoch": 0.76, "learning_rate": 9.23908100054889e-05, "loss": 3.1487, "step": 9704 }, { "epoch": 0.76, "learning_rate": 9.23900258762644e-05, "loss": 3.1569, "step": 9705 }, { "epoch": 0.76, "learning_rate": 9.23892417470399e-05, "loss": 2.963, "step": 9706 }, { "epoch": 0.76, "learning_rate": 9.238845761781541e-05, "loss": 3.1931, "step": 9707 }, { "epoch": 0.76, "learning_rate": 9.238767348859091e-05, "loss": 3.1298, "step": 9708 }, { "epoch": 0.76, "learning_rate": 9.238688935936642e-05, "loss": 3.2383, "step": 9709 }, { "epoch": 0.76, "learning_rate": 9.238610523014193e-05, "loss": 2.9265, "step": 9710 }, { "epoch": 0.76, "learning_rate": 9.238532110091744e-05, "loss": 3.0115, "step": 9711 }, { "epoch": 0.76, "learning_rate": 9.238453697169294e-05, "loss": 2.9138, "step": 9712 }, { "epoch": 0.76, "learning_rate": 9.238375284246845e-05, "loss": 3.0066, "step": 9713 }, { "epoch": 0.76, "learning_rate": 9.238296871324395e-05, "loss": 3.157, "step": 9714 }, { "epoch": 0.76, "learning_rate": 9.238218458401945e-05, "loss": 2.9465, "step": 9715 }, { "epoch": 0.76, "learning_rate": 9.238140045479496e-05, "loss": 2.881, "step": 9716 }, { "epoch": 0.76, "learning_rate": 9.238061632557046e-05, "loss": 3.3934, "step": 9717 }, { "epoch": 0.76, "learning_rate": 9.237983219634596e-05, "loss": 2.8145, "step": 9718 }, { "epoch": 0.76, "learning_rate": 9.237904806712147e-05, "loss": 3.0079, "step": 9719 }, { "epoch": 0.76, "learning_rate": 9.237826393789697e-05, "loss": 3.1791, "step": 9720 }, { "epoch": 0.76, "learning_rate": 9.237747980867248e-05, "loss": 3.1999, "step": 9721 }, { "epoch": 0.76, "learning_rate": 9.237669567944798e-05, "loss": 2.997, "step": 9722 }, { "epoch": 0.76, "learning_rate": 9.237591155022348e-05, "loss": 3.0056, "step": 9723 }, { "epoch": 0.76, "learning_rate": 9.237512742099899e-05, "loss": 2.9927, "step": 9724 }, { "epoch": 0.76, "learning_rate": 9.237434329177449e-05, "loss": 3.1038, "step": 9725 }, { "epoch": 0.76, "learning_rate": 9.237355916255e-05, "loss": 3.2738, "step": 9726 }, { "epoch": 0.76, "learning_rate": 9.23727750333255e-05, "loss": 3.0129, "step": 9727 }, { "epoch": 0.76, "learning_rate": 9.2371990904101e-05, "loss": 3.0253, "step": 9728 }, { "epoch": 0.76, "learning_rate": 9.23712067748765e-05, "loss": 2.9341, "step": 9729 }, { "epoch": 0.76, "learning_rate": 9.237042264565201e-05, "loss": 2.7258, "step": 9730 }, { "epoch": 0.76, "learning_rate": 9.236963851642751e-05, "loss": 3.0429, "step": 9731 }, { "epoch": 0.76, "learning_rate": 9.236885438720302e-05, "loss": 2.9516, "step": 9732 }, { "epoch": 0.76, "learning_rate": 9.236807025797852e-05, "loss": 3.1106, "step": 9733 }, { "epoch": 0.76, "learning_rate": 9.236728612875402e-05, "loss": 3.0206, "step": 9734 }, { "epoch": 0.76, "learning_rate": 9.236650199952953e-05, "loss": 3.0472, "step": 9735 }, { "epoch": 0.76, "learning_rate": 9.236571787030503e-05, "loss": 2.7278, "step": 9736 }, { "epoch": 0.76, "learning_rate": 9.236493374108053e-05, "loss": 3.0984, "step": 9737 }, { "epoch": 0.76, "learning_rate": 9.236414961185604e-05, "loss": 3.0067, "step": 9738 }, { "epoch": 0.76, "learning_rate": 9.236336548263154e-05, "loss": 3.1004, "step": 9739 }, { "epoch": 0.76, "learning_rate": 9.236258135340705e-05, "loss": 3.1532, "step": 9740 }, { "epoch": 0.76, "learning_rate": 9.236179722418255e-05, "loss": 2.9653, "step": 9741 }, { "epoch": 0.76, "learning_rate": 9.236101309495805e-05, "loss": 3.3025, "step": 9742 }, { "epoch": 0.76, "learning_rate": 9.236022896573356e-05, "loss": 2.809, "step": 9743 }, { "epoch": 0.76, "learning_rate": 9.235944483650906e-05, "loss": 2.8518, "step": 9744 }, { "epoch": 0.76, "learning_rate": 9.235866070728456e-05, "loss": 3.0667, "step": 9745 }, { "epoch": 0.76, "learning_rate": 9.235787657806007e-05, "loss": 3.0177, "step": 9746 }, { "epoch": 0.76, "learning_rate": 9.235709244883557e-05, "loss": 3.0018, "step": 9747 }, { "epoch": 0.76, "learning_rate": 9.235630831961107e-05, "loss": 3.0973, "step": 9748 }, { "epoch": 0.76, "learning_rate": 9.235552419038658e-05, "loss": 2.8155, "step": 9749 }, { "epoch": 0.76, "learning_rate": 9.235474006116208e-05, "loss": 3.0883, "step": 9750 }, { "epoch": 0.76, "learning_rate": 9.235395593193759e-05, "loss": 2.9499, "step": 9751 }, { "epoch": 0.76, "learning_rate": 9.235317180271309e-05, "loss": 3.0033, "step": 9752 }, { "epoch": 0.76, "learning_rate": 9.235238767348859e-05, "loss": 2.6653, "step": 9753 }, { "epoch": 0.76, "learning_rate": 9.23516035442641e-05, "loss": 3.0449, "step": 9754 }, { "epoch": 0.76, "learning_rate": 9.23508194150396e-05, "loss": 3.1542, "step": 9755 }, { "epoch": 0.76, "learning_rate": 9.23500352858151e-05, "loss": 3.1818, "step": 9756 }, { "epoch": 0.77, "learning_rate": 9.234925115659061e-05, "loss": 2.9013, "step": 9757 }, { "epoch": 0.77, "learning_rate": 9.234846702736611e-05, "loss": 3.0071, "step": 9758 }, { "epoch": 0.77, "learning_rate": 9.234768289814162e-05, "loss": 2.9717, "step": 9759 }, { "epoch": 0.77, "learning_rate": 9.234689876891712e-05, "loss": 2.9801, "step": 9760 }, { "epoch": 0.77, "learning_rate": 9.234611463969262e-05, "loss": 3.3693, "step": 9761 }, { "epoch": 0.77, "learning_rate": 9.234533051046813e-05, "loss": 2.926, "step": 9762 }, { "epoch": 0.77, "learning_rate": 9.234454638124363e-05, "loss": 2.7791, "step": 9763 }, { "epoch": 0.77, "learning_rate": 9.234376225201913e-05, "loss": 2.7562, "step": 9764 }, { "epoch": 0.77, "learning_rate": 9.234297812279464e-05, "loss": 2.9064, "step": 9765 }, { "epoch": 0.77, "learning_rate": 9.234219399357014e-05, "loss": 2.9417, "step": 9766 }, { "epoch": 0.77, "learning_rate": 9.234140986434566e-05, "loss": 2.8907, "step": 9767 }, { "epoch": 0.77, "learning_rate": 9.234062573512116e-05, "loss": 3.2106, "step": 9768 }, { "epoch": 0.77, "learning_rate": 9.233984160589667e-05, "loss": 2.9017, "step": 9769 }, { "epoch": 0.77, "learning_rate": 9.233905747667217e-05, "loss": 3.2671, "step": 9770 }, { "epoch": 0.77, "learning_rate": 9.233827334744767e-05, "loss": 2.982, "step": 9771 }, { "epoch": 0.77, "learning_rate": 9.233748921822318e-05, "loss": 3.2332, "step": 9772 }, { "epoch": 0.77, "learning_rate": 9.233670508899868e-05, "loss": 2.9147, "step": 9773 }, { "epoch": 0.77, "learning_rate": 9.233592095977418e-05, "loss": 2.8883, "step": 9774 }, { "epoch": 0.77, "learning_rate": 9.233513683054969e-05, "loss": 3.2164, "step": 9775 }, { "epoch": 0.77, "learning_rate": 9.233435270132518e-05, "loss": 3.0637, "step": 9776 }, { "epoch": 0.77, "learning_rate": 9.233356857210068e-05, "loss": 3.0084, "step": 9777 }, { "epoch": 0.77, "learning_rate": 9.233278444287619e-05, "loss": 3.0425, "step": 9778 }, { "epoch": 0.77, "learning_rate": 9.233200031365169e-05, "loss": 3.0048, "step": 9779 }, { "epoch": 0.77, "learning_rate": 9.233121618442719e-05, "loss": 2.9357, "step": 9780 }, { "epoch": 0.77, "learning_rate": 9.23304320552027e-05, "loss": 2.8936, "step": 9781 }, { "epoch": 0.77, "learning_rate": 9.23296479259782e-05, "loss": 2.7425, "step": 9782 }, { "epoch": 0.77, "learning_rate": 9.23288637967537e-05, "loss": 2.8813, "step": 9783 }, { "epoch": 0.77, "learning_rate": 9.232807966752921e-05, "loss": 2.9853, "step": 9784 }, { "epoch": 0.77, "learning_rate": 9.232729553830471e-05, "loss": 3.0856, "step": 9785 }, { "epoch": 0.77, "learning_rate": 9.232651140908021e-05, "loss": 2.9719, "step": 9786 }, { "epoch": 0.77, "learning_rate": 9.232572727985572e-05, "loss": 2.8865, "step": 9787 }, { "epoch": 0.77, "learning_rate": 9.232494315063122e-05, "loss": 2.9198, "step": 9788 }, { "epoch": 0.77, "learning_rate": 9.232415902140673e-05, "loss": 3.029, "step": 9789 }, { "epoch": 0.77, "learning_rate": 9.232337489218223e-05, "loss": 2.9583, "step": 9790 }, { "epoch": 0.77, "learning_rate": 9.232259076295773e-05, "loss": 3.1618, "step": 9791 }, { "epoch": 0.77, "learning_rate": 9.232180663373324e-05, "loss": 2.8712, "step": 9792 }, { "epoch": 0.77, "learning_rate": 9.232102250450875e-05, "loss": 2.9746, "step": 9793 }, { "epoch": 0.77, "learning_rate": 9.232023837528426e-05, "loss": 3.0616, "step": 9794 }, { "epoch": 0.77, "learning_rate": 9.231945424605976e-05, "loss": 2.9973, "step": 9795 }, { "epoch": 0.77, "learning_rate": 9.231867011683526e-05, "loss": 2.9598, "step": 9796 }, { "epoch": 0.77, "learning_rate": 9.231788598761077e-05, "loss": 2.9404, "step": 9797 }, { "epoch": 0.77, "learning_rate": 9.231710185838627e-05, "loss": 3.0182, "step": 9798 }, { "epoch": 0.77, "learning_rate": 9.231631772916178e-05, "loss": 3.0086, "step": 9799 }, { "epoch": 0.77, "learning_rate": 9.231553359993728e-05, "loss": 2.891, "step": 9800 }, { "epoch": 0.77, "learning_rate": 9.231474947071278e-05, "loss": 2.8042, "step": 9801 }, { "epoch": 0.77, "learning_rate": 9.231396534148829e-05, "loss": 3.1545, "step": 9802 }, { "epoch": 0.77, "learning_rate": 9.231318121226379e-05, "loss": 3.2022, "step": 9803 }, { "epoch": 0.77, "learning_rate": 9.23123970830393e-05, "loss": 2.8115, "step": 9804 }, { "epoch": 0.77, "learning_rate": 9.23116129538148e-05, "loss": 3.0396, "step": 9805 }, { "epoch": 0.77, "learning_rate": 9.23108288245903e-05, "loss": 3.0063, "step": 9806 }, { "epoch": 0.77, "learning_rate": 9.23100446953658e-05, "loss": 3.2123, "step": 9807 }, { "epoch": 0.77, "learning_rate": 9.230926056614131e-05, "loss": 3.0132, "step": 9808 }, { "epoch": 0.77, "learning_rate": 9.230847643691681e-05, "loss": 2.9642, "step": 9809 }, { "epoch": 0.77, "learning_rate": 9.230769230769232e-05, "loss": 3.0202, "step": 9810 }, { "epoch": 0.77, "learning_rate": 9.230690817846782e-05, "loss": 3.0378, "step": 9811 }, { "epoch": 0.77, "learning_rate": 9.230612404924331e-05, "loss": 3.0541, "step": 9812 }, { "epoch": 0.77, "learning_rate": 9.230533992001881e-05, "loss": 2.8854, "step": 9813 }, { "epoch": 0.77, "learning_rate": 9.230455579079432e-05, "loss": 2.8999, "step": 9814 }, { "epoch": 0.77, "learning_rate": 9.230377166156982e-05, "loss": 3.049, "step": 9815 }, { "epoch": 0.77, "learning_rate": 9.230298753234532e-05, "loss": 3.1543, "step": 9816 }, { "epoch": 0.77, "learning_rate": 9.230220340312083e-05, "loss": 3.0662, "step": 9817 }, { "epoch": 0.77, "learning_rate": 9.230141927389635e-05, "loss": 3.022, "step": 9818 }, { "epoch": 0.77, "learning_rate": 9.230063514467185e-05, "loss": 2.928, "step": 9819 }, { "epoch": 0.77, "learning_rate": 9.229985101544735e-05, "loss": 3.1633, "step": 9820 }, { "epoch": 0.77, "learning_rate": 9.229906688622286e-05, "loss": 3.0139, "step": 9821 }, { "epoch": 0.77, "learning_rate": 9.229828275699836e-05, "loss": 3.1689, "step": 9822 }, { "epoch": 0.77, "learning_rate": 9.229749862777386e-05, "loss": 2.9416, "step": 9823 }, { "epoch": 0.77, "learning_rate": 9.229671449854937e-05, "loss": 2.7773, "step": 9824 }, { "epoch": 0.77, "learning_rate": 9.229593036932487e-05, "loss": 3.042, "step": 9825 }, { "epoch": 0.77, "learning_rate": 9.229514624010038e-05, "loss": 3.0722, "step": 9826 }, { "epoch": 0.77, "learning_rate": 9.229436211087588e-05, "loss": 2.927, "step": 9827 }, { "epoch": 0.77, "learning_rate": 9.229357798165138e-05, "loss": 3.0871, "step": 9828 }, { "epoch": 0.77, "learning_rate": 9.229279385242689e-05, "loss": 3.1627, "step": 9829 }, { "epoch": 0.77, "learning_rate": 9.229200972320239e-05, "loss": 2.881, "step": 9830 }, { "epoch": 0.77, "learning_rate": 9.22912255939779e-05, "loss": 3.2453, "step": 9831 }, { "epoch": 0.77, "learning_rate": 9.22904414647534e-05, "loss": 3.1002, "step": 9832 }, { "epoch": 0.77, "learning_rate": 9.22896573355289e-05, "loss": 3.0296, "step": 9833 }, { "epoch": 0.77, "learning_rate": 9.22888732063044e-05, "loss": 3.1126, "step": 9834 }, { "epoch": 0.77, "learning_rate": 9.228808907707991e-05, "loss": 2.8161, "step": 9835 }, { "epoch": 0.77, "learning_rate": 9.228730494785541e-05, "loss": 3.0749, "step": 9836 }, { "epoch": 0.77, "learning_rate": 9.228652081863092e-05, "loss": 3.0386, "step": 9837 }, { "epoch": 0.77, "learning_rate": 9.228573668940642e-05, "loss": 3.0458, "step": 9838 }, { "epoch": 0.77, "learning_rate": 9.228495256018192e-05, "loss": 2.9071, "step": 9839 }, { "epoch": 0.77, "learning_rate": 9.228416843095743e-05, "loss": 3.1817, "step": 9840 }, { "epoch": 0.77, "learning_rate": 9.228338430173293e-05, "loss": 2.9744, "step": 9841 }, { "epoch": 0.77, "learning_rate": 9.228260017250843e-05, "loss": 3.0126, "step": 9842 }, { "epoch": 0.77, "learning_rate": 9.228181604328394e-05, "loss": 2.8603, "step": 9843 }, { "epoch": 0.77, "learning_rate": 9.228103191405944e-05, "loss": 2.9132, "step": 9844 }, { "epoch": 0.77, "learning_rate": 9.228024778483495e-05, "loss": 2.7646, "step": 9845 }, { "epoch": 0.77, "learning_rate": 9.227946365561045e-05, "loss": 3.1031, "step": 9846 }, { "epoch": 0.77, "learning_rate": 9.227867952638595e-05, "loss": 2.9973, "step": 9847 }, { "epoch": 0.77, "learning_rate": 9.227789539716146e-05, "loss": 3.2946, "step": 9848 }, { "epoch": 0.77, "learning_rate": 9.227711126793696e-05, "loss": 3.1586, "step": 9849 }, { "epoch": 0.77, "learning_rate": 9.227632713871246e-05, "loss": 2.9008, "step": 9850 }, { "epoch": 0.77, "learning_rate": 9.227554300948797e-05, "loss": 3.0011, "step": 9851 }, { "epoch": 0.77, "learning_rate": 9.227475888026347e-05, "loss": 2.9815, "step": 9852 }, { "epoch": 0.77, "learning_rate": 9.227397475103897e-05, "loss": 2.9953, "step": 9853 }, { "epoch": 0.77, "learning_rate": 9.227319062181448e-05, "loss": 2.9815, "step": 9854 }, { "epoch": 0.77, "learning_rate": 9.227240649258998e-05, "loss": 3.1465, "step": 9855 }, { "epoch": 0.77, "learning_rate": 9.227162236336549e-05, "loss": 3.0603, "step": 9856 }, { "epoch": 0.77, "learning_rate": 9.227083823414099e-05, "loss": 3.1895, "step": 9857 }, { "epoch": 0.77, "learning_rate": 9.227005410491649e-05, "loss": 3.1117, "step": 9858 }, { "epoch": 0.77, "learning_rate": 9.2269269975692e-05, "loss": 3.0076, "step": 9859 }, { "epoch": 0.77, "learning_rate": 9.22684858464675e-05, "loss": 3.1087, "step": 9860 }, { "epoch": 0.77, "learning_rate": 9.2267701717243e-05, "loss": 3.2688, "step": 9861 }, { "epoch": 0.77, "learning_rate": 9.226691758801851e-05, "loss": 3.0206, "step": 9862 }, { "epoch": 0.77, "learning_rate": 9.226613345879401e-05, "loss": 2.7889, "step": 9863 }, { "epoch": 0.77, "learning_rate": 9.226534932956951e-05, "loss": 3.1666, "step": 9864 }, { "epoch": 0.77, "learning_rate": 9.226456520034502e-05, "loss": 2.7846, "step": 9865 }, { "epoch": 0.77, "learning_rate": 9.226378107112052e-05, "loss": 3.0202, "step": 9866 }, { "epoch": 0.77, "learning_rate": 9.226299694189603e-05, "loss": 2.8233, "step": 9867 }, { "epoch": 0.77, "learning_rate": 9.226221281267153e-05, "loss": 2.9885, "step": 9868 }, { "epoch": 0.77, "learning_rate": 9.226142868344703e-05, "loss": 2.9812, "step": 9869 }, { "epoch": 0.77, "learning_rate": 9.226064455422254e-05, "loss": 3.1511, "step": 9870 }, { "epoch": 0.77, "learning_rate": 9.225986042499804e-05, "loss": 3.1651, "step": 9871 }, { "epoch": 0.77, "learning_rate": 9.225907629577354e-05, "loss": 3.1243, "step": 9872 }, { "epoch": 0.77, "learning_rate": 9.225829216654905e-05, "loss": 3.2169, "step": 9873 }, { "epoch": 0.77, "learning_rate": 9.225750803732455e-05, "loss": 3.0858, "step": 9874 }, { "epoch": 0.77, "learning_rate": 9.225672390810007e-05, "loss": 2.8128, "step": 9875 }, { "epoch": 0.77, "learning_rate": 9.225593977887557e-05, "loss": 2.8484, "step": 9876 }, { "epoch": 0.77, "learning_rate": 9.225515564965108e-05, "loss": 2.9587, "step": 9877 }, { "epoch": 0.77, "learning_rate": 9.225437152042658e-05, "loss": 2.835, "step": 9878 }, { "epoch": 0.77, "learning_rate": 9.225358739120208e-05, "loss": 2.936, "step": 9879 }, { "epoch": 0.77, "learning_rate": 9.225280326197759e-05, "loss": 2.9735, "step": 9880 }, { "epoch": 0.77, "learning_rate": 9.225201913275309e-05, "loss": 3.0249, "step": 9881 }, { "epoch": 0.77, "learning_rate": 9.22512350035286e-05, "loss": 2.8481, "step": 9882 }, { "epoch": 0.77, "learning_rate": 9.22504508743041e-05, "loss": 3.1126, "step": 9883 }, { "epoch": 0.78, "learning_rate": 9.224966674507959e-05, "loss": 3.0232, "step": 9884 }, { "epoch": 0.78, "learning_rate": 9.224888261585509e-05, "loss": 3.2223, "step": 9885 }, { "epoch": 0.78, "learning_rate": 9.22480984866306e-05, "loss": 3.3069, "step": 9886 }, { "epoch": 0.78, "learning_rate": 9.22473143574061e-05, "loss": 2.9005, "step": 9887 }, { "epoch": 0.78, "learning_rate": 9.22465302281816e-05, "loss": 3.1861, "step": 9888 }, { "epoch": 0.78, "learning_rate": 9.224574609895711e-05, "loss": 2.8233, "step": 9889 }, { "epoch": 0.78, "learning_rate": 9.224496196973261e-05, "loss": 2.9715, "step": 9890 }, { "epoch": 0.78, "learning_rate": 9.224417784050811e-05, "loss": 3.2089, "step": 9891 }, { "epoch": 0.78, "learning_rate": 9.224339371128362e-05, "loss": 2.9185, "step": 9892 }, { "epoch": 0.78, "learning_rate": 9.224260958205912e-05, "loss": 2.8041, "step": 9893 }, { "epoch": 0.78, "learning_rate": 9.224182545283463e-05, "loss": 3.17, "step": 9894 }, { "epoch": 0.78, "learning_rate": 9.224104132361013e-05, "loss": 3.0039, "step": 9895 }, { "epoch": 0.78, "learning_rate": 9.224025719438563e-05, "loss": 3.0642, "step": 9896 }, { "epoch": 0.78, "learning_rate": 9.223947306516114e-05, "loss": 3.0407, "step": 9897 }, { "epoch": 0.78, "learning_rate": 9.223868893593664e-05, "loss": 2.8697, "step": 9898 }, { "epoch": 0.78, "learning_rate": 9.223790480671214e-05, "loss": 2.8093, "step": 9899 }, { "epoch": 0.78, "learning_rate": 9.223712067748765e-05, "loss": 3.1084, "step": 9900 }, { "epoch": 0.78, "learning_rate": 9.223633654826316e-05, "loss": 2.9478, "step": 9901 }, { "epoch": 0.78, "learning_rate": 9.223555241903867e-05, "loss": 2.8607, "step": 9902 }, { "epoch": 0.78, "learning_rate": 9.223476828981417e-05, "loss": 3.1042, "step": 9903 }, { "epoch": 0.78, "learning_rate": 9.223398416058968e-05, "loss": 3.1063, "step": 9904 }, { "epoch": 0.78, "learning_rate": 9.223320003136518e-05, "loss": 2.953, "step": 9905 }, { "epoch": 0.78, "learning_rate": 9.223241590214068e-05, "loss": 2.7819, "step": 9906 }, { "epoch": 0.78, "learning_rate": 9.223163177291619e-05, "loss": 3.1294, "step": 9907 }, { "epoch": 0.78, "learning_rate": 9.223084764369169e-05, "loss": 3.0364, "step": 9908 }, { "epoch": 0.78, "learning_rate": 9.22300635144672e-05, "loss": 2.9037, "step": 9909 }, { "epoch": 0.78, "learning_rate": 9.22292793852427e-05, "loss": 3.0512, "step": 9910 }, { "epoch": 0.78, "learning_rate": 9.22284952560182e-05, "loss": 2.991, "step": 9911 }, { "epoch": 0.78, "learning_rate": 9.22277111267937e-05, "loss": 3.0882, "step": 9912 }, { "epoch": 0.78, "learning_rate": 9.222692699756921e-05, "loss": 3.0137, "step": 9913 }, { "epoch": 0.78, "learning_rate": 9.222614286834471e-05, "loss": 3.2689, "step": 9914 }, { "epoch": 0.78, "learning_rate": 9.222535873912022e-05, "loss": 3.0101, "step": 9915 }, { "epoch": 0.78, "learning_rate": 9.222457460989572e-05, "loss": 2.9422, "step": 9916 }, { "epoch": 0.78, "learning_rate": 9.222379048067122e-05, "loss": 2.9247, "step": 9917 }, { "epoch": 0.78, "learning_rate": 9.222300635144673e-05, "loss": 3.0884, "step": 9918 }, { "epoch": 0.78, "learning_rate": 9.222222222222223e-05, "loss": 3.0882, "step": 9919 }, { "epoch": 0.78, "learning_rate": 9.222143809299772e-05, "loss": 3.1824, "step": 9920 }, { "epoch": 0.78, "learning_rate": 9.222065396377322e-05, "loss": 2.7828, "step": 9921 }, { "epoch": 0.78, "learning_rate": 9.221986983454873e-05, "loss": 2.9801, "step": 9922 }, { "epoch": 0.78, "learning_rate": 9.221908570532423e-05, "loss": 2.8527, "step": 9923 }, { "epoch": 0.78, "learning_rate": 9.221830157609974e-05, "loss": 2.9828, "step": 9924 }, { "epoch": 0.78, "learning_rate": 9.221751744687524e-05, "loss": 3.124, "step": 9925 }, { "epoch": 0.78, "learning_rate": 9.221673331765076e-05, "loss": 3.0905, "step": 9926 }, { "epoch": 0.78, "learning_rate": 9.221594918842626e-05, "loss": 3.1068, "step": 9927 }, { "epoch": 0.78, "learning_rate": 9.221516505920176e-05, "loss": 2.9454, "step": 9928 }, { "epoch": 0.78, "learning_rate": 9.221438092997727e-05, "loss": 3.0892, "step": 9929 }, { "epoch": 0.78, "learning_rate": 9.221359680075277e-05, "loss": 3.0556, "step": 9930 }, { "epoch": 0.78, "learning_rate": 9.221281267152827e-05, "loss": 3.0734, "step": 9931 }, { "epoch": 0.78, "learning_rate": 9.221202854230378e-05, "loss": 3.0254, "step": 9932 }, { "epoch": 0.78, "learning_rate": 9.221124441307928e-05, "loss": 3.3362, "step": 9933 }, { "epoch": 0.78, "learning_rate": 9.221046028385479e-05, "loss": 3.209, "step": 9934 }, { "epoch": 0.78, "learning_rate": 9.220967615463029e-05, "loss": 3.1488, "step": 9935 }, { "epoch": 0.78, "learning_rate": 9.22088920254058e-05, "loss": 2.8976, "step": 9936 }, { "epoch": 0.78, "learning_rate": 9.22081078961813e-05, "loss": 2.9247, "step": 9937 }, { "epoch": 0.78, "learning_rate": 9.22073237669568e-05, "loss": 2.9227, "step": 9938 }, { "epoch": 0.78, "learning_rate": 9.22065396377323e-05, "loss": 3.004, "step": 9939 }, { "epoch": 0.78, "learning_rate": 9.220575550850781e-05, "loss": 3.0265, "step": 9940 }, { "epoch": 0.78, "learning_rate": 9.220497137928331e-05, "loss": 3.2491, "step": 9941 }, { "epoch": 0.78, "learning_rate": 9.220418725005882e-05, "loss": 3.0688, "step": 9942 }, { "epoch": 0.78, "learning_rate": 9.220340312083432e-05, "loss": 2.8898, "step": 9943 }, { "epoch": 0.78, "learning_rate": 9.220261899160982e-05, "loss": 3.0429, "step": 9944 }, { "epoch": 0.78, "learning_rate": 9.220183486238533e-05, "loss": 2.9052, "step": 9945 }, { "epoch": 0.78, "learning_rate": 9.220105073316083e-05, "loss": 3.0103, "step": 9946 }, { "epoch": 0.78, "learning_rate": 9.220026660393633e-05, "loss": 2.9988, "step": 9947 }, { "epoch": 0.78, "learning_rate": 9.219948247471184e-05, "loss": 3.0346, "step": 9948 }, { "epoch": 0.78, "learning_rate": 9.219869834548734e-05, "loss": 3.1084, "step": 9949 }, { "epoch": 0.78, "learning_rate": 9.219791421626284e-05, "loss": 2.8295, "step": 9950 }, { "epoch": 0.78, "learning_rate": 9.219713008703835e-05, "loss": 3.047, "step": 9951 }, { "epoch": 0.78, "learning_rate": 9.219634595781385e-05, "loss": 2.8585, "step": 9952 }, { "epoch": 0.78, "learning_rate": 9.219556182858936e-05, "loss": 2.8736, "step": 9953 }, { "epoch": 0.78, "learning_rate": 9.219477769936486e-05, "loss": 2.7316, "step": 9954 }, { "epoch": 0.78, "learning_rate": 9.219399357014036e-05, "loss": 2.8713, "step": 9955 }, { "epoch": 0.78, "learning_rate": 9.219320944091587e-05, "loss": 2.9726, "step": 9956 }, { "epoch": 0.78, "learning_rate": 9.219242531169137e-05, "loss": 3.1377, "step": 9957 }, { "epoch": 0.78, "learning_rate": 9.219164118246687e-05, "loss": 3.024, "step": 9958 }, { "epoch": 0.78, "learning_rate": 9.219085705324238e-05, "loss": 3.0429, "step": 9959 }, { "epoch": 0.78, "learning_rate": 9.219007292401788e-05, "loss": 2.8865, "step": 9960 }, { "epoch": 0.78, "learning_rate": 9.218928879479339e-05, "loss": 2.8694, "step": 9961 }, { "epoch": 0.78, "learning_rate": 9.218850466556889e-05, "loss": 3.0961, "step": 9962 }, { "epoch": 0.78, "learning_rate": 9.218772053634439e-05, "loss": 3.3219, "step": 9963 }, { "epoch": 0.78, "learning_rate": 9.21869364071199e-05, "loss": 2.9968, "step": 9964 }, { "epoch": 0.78, "learning_rate": 9.21861522778954e-05, "loss": 2.8815, "step": 9965 }, { "epoch": 0.78, "learning_rate": 9.21853681486709e-05, "loss": 2.8808, "step": 9966 }, { "epoch": 0.78, "learning_rate": 9.218458401944641e-05, "loss": 2.9616, "step": 9967 }, { "epoch": 0.78, "learning_rate": 9.218379989022191e-05, "loss": 3.3149, "step": 9968 }, { "epoch": 0.78, "learning_rate": 9.218301576099741e-05, "loss": 3.0921, "step": 9969 }, { "epoch": 0.78, "learning_rate": 9.218223163177292e-05, "loss": 3.2989, "step": 9970 }, { "epoch": 0.78, "learning_rate": 9.218144750254842e-05, "loss": 3.0724, "step": 9971 }, { "epoch": 0.78, "learning_rate": 9.218066337332393e-05, "loss": 3.5623, "step": 9972 }, { "epoch": 0.78, "learning_rate": 9.217987924409943e-05, "loss": 2.9245, "step": 9973 }, { "epoch": 0.78, "learning_rate": 9.217909511487493e-05, "loss": 2.9034, "step": 9974 }, { "epoch": 0.78, "learning_rate": 9.217831098565044e-05, "loss": 2.8882, "step": 9975 }, { "epoch": 0.78, "learning_rate": 9.217752685642594e-05, "loss": 3.0078, "step": 9976 }, { "epoch": 0.78, "learning_rate": 9.217674272720144e-05, "loss": 2.8633, "step": 9977 }, { "epoch": 0.78, "learning_rate": 9.217595859797695e-05, "loss": 3.1243, "step": 9978 }, { "epoch": 0.78, "learning_rate": 9.217517446875245e-05, "loss": 3.135, "step": 9979 }, { "epoch": 0.78, "learning_rate": 9.217439033952796e-05, "loss": 3.0334, "step": 9980 }, { "epoch": 0.78, "learning_rate": 9.217360621030346e-05, "loss": 3.0091, "step": 9981 }, { "epoch": 0.78, "learning_rate": 9.217282208107896e-05, "loss": 2.8135, "step": 9982 }, { "epoch": 0.78, "learning_rate": 9.217203795185448e-05, "loss": 3.3347, "step": 9983 }, { "epoch": 0.78, "learning_rate": 9.217125382262998e-05, "loss": 2.9926, "step": 9984 }, { "epoch": 0.78, "learning_rate": 9.217046969340549e-05, "loss": 2.8798, "step": 9985 }, { "epoch": 0.78, "learning_rate": 9.216968556418099e-05, "loss": 3.0717, "step": 9986 }, { "epoch": 0.78, "learning_rate": 9.21689014349565e-05, "loss": 2.9951, "step": 9987 }, { "epoch": 0.78, "learning_rate": 9.2168117305732e-05, "loss": 3.0824, "step": 9988 }, { "epoch": 0.78, "learning_rate": 9.21673331765075e-05, "loss": 2.9172, "step": 9989 }, { "epoch": 0.78, "learning_rate": 9.2166549047283e-05, "loss": 3.1518, "step": 9990 }, { "epoch": 0.78, "learning_rate": 9.216576491805851e-05, "loss": 3.2703, "step": 9991 }, { "epoch": 0.78, "learning_rate": 9.2164980788834e-05, "loss": 2.9574, "step": 9992 }, { "epoch": 0.78, "learning_rate": 9.21641966596095e-05, "loss": 3.2275, "step": 9993 }, { "epoch": 0.78, "learning_rate": 9.2163412530385e-05, "loss": 3.2719, "step": 9994 }, { "epoch": 0.78, "learning_rate": 9.216262840116051e-05, "loss": 3.3324, "step": 9995 }, { "epoch": 0.78, "learning_rate": 9.216184427193601e-05, "loss": 2.9951, "step": 9996 }, { "epoch": 0.78, "learning_rate": 9.216106014271152e-05, "loss": 3.3072, "step": 9997 }, { "epoch": 0.78, "learning_rate": 9.216027601348702e-05, "loss": 3.0406, "step": 9998 }, { "epoch": 0.78, "learning_rate": 9.215949188426252e-05, "loss": 3.0753, "step": 9999 }, { "epoch": 0.78, "learning_rate": 9.215870775503803e-05, "loss": 3.1184, "step": 10000 }, { "epoch": 0.78, "eval_loss": 2.935558557510376, "eval_runtime": 217.9773, "eval_samples_per_second": 51.987, "eval_steps_per_second": 3.253, "step": 10000 }, { "epoch": 0.78, "learning_rate": 9.215792362581353e-05, "loss": 3.1525, "step": 10001 }, { "epoch": 0.78, "learning_rate": 9.215713949658904e-05, "loss": 3.0242, "step": 10002 }, { "epoch": 0.78, "learning_rate": 9.215635536736454e-05, "loss": 3.1237, "step": 10003 }, { "epoch": 0.78, "learning_rate": 9.215557123814004e-05, "loss": 3.0581, "step": 10004 }, { "epoch": 0.78, "learning_rate": 9.215478710891555e-05, "loss": 3.2987, "step": 10005 }, { "epoch": 0.78, "learning_rate": 9.215400297969105e-05, "loss": 3.2231, "step": 10006 }, { "epoch": 0.78, "learning_rate": 9.215321885046655e-05, "loss": 3.0622, "step": 10007 }, { "epoch": 0.78, "learning_rate": 9.215243472124206e-05, "loss": 3.0469, "step": 10008 }, { "epoch": 0.78, "learning_rate": 9.215165059201758e-05, "loss": 2.8633, "step": 10009 }, { "epoch": 0.78, "learning_rate": 9.215086646279308e-05, "loss": 3.0099, "step": 10010 }, { "epoch": 0.78, "learning_rate": 9.215008233356858e-05, "loss": 2.8166, "step": 10011 }, { "epoch": 0.79, "learning_rate": 9.214929820434409e-05, "loss": 3.14, "step": 10012 }, { "epoch": 0.79, "learning_rate": 9.214851407511959e-05, "loss": 3.1772, "step": 10013 }, { "epoch": 0.79, "learning_rate": 9.21477299458951e-05, "loss": 2.9101, "step": 10014 }, { "epoch": 0.79, "learning_rate": 9.21469458166706e-05, "loss": 2.9982, "step": 10015 }, { "epoch": 0.79, "learning_rate": 9.21461616874461e-05, "loss": 2.9391, "step": 10016 }, { "epoch": 0.79, "learning_rate": 9.21453775582216e-05, "loss": 2.9236, "step": 10017 }, { "epoch": 0.79, "learning_rate": 9.214459342899711e-05, "loss": 3.3197, "step": 10018 }, { "epoch": 0.79, "learning_rate": 9.214380929977261e-05, "loss": 3.0803, "step": 10019 }, { "epoch": 0.79, "learning_rate": 9.214302517054812e-05, "loss": 3.1769, "step": 10020 }, { "epoch": 0.79, "learning_rate": 9.214224104132362e-05, "loss": 3.1299, "step": 10021 }, { "epoch": 0.79, "learning_rate": 9.214145691209912e-05, "loss": 2.7868, "step": 10022 }, { "epoch": 0.79, "learning_rate": 9.214067278287463e-05, "loss": 2.9505, "step": 10023 }, { "epoch": 0.79, "learning_rate": 9.213988865365013e-05, "loss": 3.0917, "step": 10024 }, { "epoch": 0.79, "learning_rate": 9.213910452442563e-05, "loss": 3.0138, "step": 10025 }, { "epoch": 0.79, "learning_rate": 9.213832039520114e-05, "loss": 3.2499, "step": 10026 }, { "epoch": 0.79, "learning_rate": 9.213753626597664e-05, "loss": 3.0034, "step": 10027 }, { "epoch": 0.79, "learning_rate": 9.213675213675213e-05, "loss": 2.7946, "step": 10028 }, { "epoch": 0.79, "learning_rate": 9.213596800752764e-05, "loss": 3.1237, "step": 10029 }, { "epoch": 0.79, "learning_rate": 9.213518387830314e-05, "loss": 2.8275, "step": 10030 }, { "epoch": 0.79, "learning_rate": 9.213439974907864e-05, "loss": 2.9955, "step": 10031 }, { "epoch": 0.79, "learning_rate": 9.213361561985415e-05, "loss": 3.0843, "step": 10032 }, { "epoch": 0.79, "learning_rate": 9.213283149062965e-05, "loss": 3.1898, "step": 10033 }, { "epoch": 0.79, "learning_rate": 9.213204736140517e-05, "loss": 2.9874, "step": 10034 }, { "epoch": 0.79, "learning_rate": 9.213126323218067e-05, "loss": 2.9619, "step": 10035 }, { "epoch": 0.79, "learning_rate": 9.213047910295617e-05, "loss": 3.5106, "step": 10036 }, { "epoch": 0.79, "learning_rate": 9.212969497373168e-05, "loss": 2.8378, "step": 10037 }, { "epoch": 0.79, "learning_rate": 9.212891084450718e-05, "loss": 2.9365, "step": 10038 }, { "epoch": 0.79, "learning_rate": 9.212812671528269e-05, "loss": 3.1625, "step": 10039 }, { "epoch": 0.79, "learning_rate": 9.212734258605819e-05, "loss": 3.007, "step": 10040 }, { "epoch": 0.79, "learning_rate": 9.212655845683369e-05, "loss": 3.0909, "step": 10041 }, { "epoch": 0.79, "learning_rate": 9.21257743276092e-05, "loss": 2.848, "step": 10042 }, { "epoch": 0.79, "learning_rate": 9.21249901983847e-05, "loss": 3.1446, "step": 10043 }, { "epoch": 0.79, "learning_rate": 9.21242060691602e-05, "loss": 3.0, "step": 10044 }, { "epoch": 0.79, "learning_rate": 9.212342193993571e-05, "loss": 2.9396, "step": 10045 }, { "epoch": 0.79, "learning_rate": 9.212263781071121e-05, "loss": 2.7982, "step": 10046 }, { "epoch": 0.79, "learning_rate": 9.212185368148672e-05, "loss": 2.8307, "step": 10047 }, { "epoch": 0.79, "learning_rate": 9.212106955226222e-05, "loss": 2.9219, "step": 10048 }, { "epoch": 0.79, "learning_rate": 9.212028542303772e-05, "loss": 3.0043, "step": 10049 }, { "epoch": 0.79, "learning_rate": 9.211950129381323e-05, "loss": 3.0729, "step": 10050 }, { "epoch": 0.79, "learning_rate": 9.211871716458873e-05, "loss": 3.1745, "step": 10051 }, { "epoch": 0.79, "learning_rate": 9.211793303536423e-05, "loss": 2.7448, "step": 10052 }, { "epoch": 0.79, "learning_rate": 9.211714890613974e-05, "loss": 2.9386, "step": 10053 }, { "epoch": 0.79, "learning_rate": 9.211636477691524e-05, "loss": 3.0691, "step": 10054 }, { "epoch": 0.79, "learning_rate": 9.211558064769074e-05, "loss": 2.9984, "step": 10055 }, { "epoch": 0.79, "learning_rate": 9.211479651846625e-05, "loss": 2.936, "step": 10056 }, { "epoch": 0.79, "learning_rate": 9.211401238924175e-05, "loss": 2.9354, "step": 10057 }, { "epoch": 0.79, "learning_rate": 9.211322826001726e-05, "loss": 3.0273, "step": 10058 }, { "epoch": 0.79, "learning_rate": 9.211244413079276e-05, "loss": 3.2502, "step": 10059 }, { "epoch": 0.79, "learning_rate": 9.211166000156826e-05, "loss": 2.9814, "step": 10060 }, { "epoch": 0.79, "learning_rate": 9.211087587234377e-05, "loss": 3.1037, "step": 10061 }, { "epoch": 0.79, "learning_rate": 9.211009174311927e-05, "loss": 2.9694, "step": 10062 }, { "epoch": 0.79, "learning_rate": 9.210930761389477e-05, "loss": 2.9677, "step": 10063 }, { "epoch": 0.79, "learning_rate": 9.210852348467028e-05, "loss": 3.2364, "step": 10064 }, { "epoch": 0.79, "learning_rate": 9.210773935544578e-05, "loss": 2.8225, "step": 10065 }, { "epoch": 0.79, "learning_rate": 9.210695522622128e-05, "loss": 3.0002, "step": 10066 }, { "epoch": 0.79, "learning_rate": 9.210617109699679e-05, "loss": 3.1343, "step": 10067 }, { "epoch": 0.79, "learning_rate": 9.210538696777229e-05, "loss": 2.9238, "step": 10068 }, { "epoch": 0.79, "learning_rate": 9.21046028385478e-05, "loss": 3.1367, "step": 10069 }, { "epoch": 0.79, "learning_rate": 9.21038187093233e-05, "loss": 3.0074, "step": 10070 }, { "epoch": 0.79, "learning_rate": 9.21030345800988e-05, "loss": 2.8559, "step": 10071 }, { "epoch": 0.79, "learning_rate": 9.210225045087431e-05, "loss": 3.0722, "step": 10072 }, { "epoch": 0.79, "learning_rate": 9.210146632164981e-05, "loss": 3.0216, "step": 10073 }, { "epoch": 0.79, "learning_rate": 9.210068219242531e-05, "loss": 3.0164, "step": 10074 }, { "epoch": 0.79, "learning_rate": 9.209989806320082e-05, "loss": 2.9103, "step": 10075 }, { "epoch": 0.79, "learning_rate": 9.209911393397632e-05, "loss": 3.2132, "step": 10076 }, { "epoch": 0.79, "learning_rate": 9.209832980475183e-05, "loss": 2.9766, "step": 10077 }, { "epoch": 0.79, "learning_rate": 9.209754567552733e-05, "loss": 2.8673, "step": 10078 }, { "epoch": 0.79, "learning_rate": 9.209676154630283e-05, "loss": 3.0817, "step": 10079 }, { "epoch": 0.79, "learning_rate": 9.209597741707834e-05, "loss": 2.9663, "step": 10080 }, { "epoch": 0.79, "learning_rate": 9.209519328785384e-05, "loss": 2.869, "step": 10081 }, { "epoch": 0.79, "learning_rate": 9.209440915862934e-05, "loss": 3.0284, "step": 10082 }, { "epoch": 0.79, "learning_rate": 9.209362502940485e-05, "loss": 2.9606, "step": 10083 }, { "epoch": 0.79, "learning_rate": 9.209284090018035e-05, "loss": 3.1441, "step": 10084 }, { "epoch": 0.79, "learning_rate": 9.209205677095585e-05, "loss": 3.1204, "step": 10085 }, { "epoch": 0.79, "learning_rate": 9.209127264173136e-05, "loss": 3.3962, "step": 10086 }, { "epoch": 0.79, "learning_rate": 9.209048851250686e-05, "loss": 2.9469, "step": 10087 }, { "epoch": 0.79, "learning_rate": 9.208970438328237e-05, "loss": 2.8886, "step": 10088 }, { "epoch": 0.79, "learning_rate": 9.208892025405787e-05, "loss": 3.023, "step": 10089 }, { "epoch": 0.79, "learning_rate": 9.208813612483337e-05, "loss": 2.8978, "step": 10090 }, { "epoch": 0.79, "learning_rate": 9.208735199560889e-05, "loss": 2.9401, "step": 10091 }, { "epoch": 0.79, "learning_rate": 9.20865678663844e-05, "loss": 2.9312, "step": 10092 }, { "epoch": 0.79, "learning_rate": 9.20857837371599e-05, "loss": 2.9274, "step": 10093 }, { "epoch": 0.79, "learning_rate": 9.20849996079354e-05, "loss": 3.0734, "step": 10094 }, { "epoch": 0.79, "learning_rate": 9.20842154787109e-05, "loss": 2.9203, "step": 10095 }, { "epoch": 0.79, "learning_rate": 9.208343134948641e-05, "loss": 3.1971, "step": 10096 }, { "epoch": 0.79, "learning_rate": 9.208264722026191e-05, "loss": 3.1821, "step": 10097 }, { "epoch": 0.79, "learning_rate": 9.208186309103742e-05, "loss": 2.849, "step": 10098 }, { "epoch": 0.79, "learning_rate": 9.208107896181292e-05, "loss": 3.0419, "step": 10099 }, { "epoch": 0.79, "learning_rate": 9.208029483258841e-05, "loss": 3.0796, "step": 10100 }, { "epoch": 0.79, "learning_rate": 9.207951070336391e-05, "loss": 3.0701, "step": 10101 }, { "epoch": 0.79, "learning_rate": 9.207872657413942e-05, "loss": 3.2119, "step": 10102 }, { "epoch": 0.79, "learning_rate": 9.207794244491492e-05, "loss": 3.0544, "step": 10103 }, { "epoch": 0.79, "learning_rate": 9.207715831569042e-05, "loss": 2.7838, "step": 10104 }, { "epoch": 0.79, "learning_rate": 9.207637418646593e-05, "loss": 3.2673, "step": 10105 }, { "epoch": 0.79, "learning_rate": 9.207559005724143e-05, "loss": 2.8486, "step": 10106 }, { "epoch": 0.79, "learning_rate": 9.207480592801694e-05, "loss": 3.0062, "step": 10107 }, { "epoch": 0.79, "learning_rate": 9.207402179879244e-05, "loss": 2.784, "step": 10108 }, { "epoch": 0.79, "learning_rate": 9.207323766956794e-05, "loss": 2.8182, "step": 10109 }, { "epoch": 0.79, "learning_rate": 9.207245354034345e-05, "loss": 3.1056, "step": 10110 }, { "epoch": 0.79, "learning_rate": 9.207166941111895e-05, "loss": 3.0801, "step": 10111 }, { "epoch": 0.79, "learning_rate": 9.207088528189445e-05, "loss": 3.0783, "step": 10112 }, { "epoch": 0.79, "learning_rate": 9.207010115266996e-05, "loss": 3.3972, "step": 10113 }, { "epoch": 0.79, "learning_rate": 9.206931702344546e-05, "loss": 3.1016, "step": 10114 }, { "epoch": 0.79, "learning_rate": 9.206853289422097e-05, "loss": 2.9738, "step": 10115 }, { "epoch": 0.79, "learning_rate": 9.206774876499647e-05, "loss": 3.0, "step": 10116 }, { "epoch": 0.79, "learning_rate": 9.206696463577199e-05, "loss": 3.1202, "step": 10117 }, { "epoch": 0.79, "learning_rate": 9.206618050654749e-05, "loss": 2.8489, "step": 10118 }, { "epoch": 0.79, "learning_rate": 9.2065396377323e-05, "loss": 2.9876, "step": 10119 }, { "epoch": 0.79, "learning_rate": 9.20646122480985e-05, "loss": 3.0671, "step": 10120 }, { "epoch": 0.79, "learning_rate": 9.2063828118874e-05, "loss": 3.1306, "step": 10121 }, { "epoch": 0.79, "learning_rate": 9.20630439896495e-05, "loss": 2.8957, "step": 10122 }, { "epoch": 0.79, "learning_rate": 9.206225986042501e-05, "loss": 3.0288, "step": 10123 }, { "epoch": 0.79, "learning_rate": 9.206147573120051e-05, "loss": 2.967, "step": 10124 }, { "epoch": 0.79, "learning_rate": 9.206069160197602e-05, "loss": 3.182, "step": 10125 }, { "epoch": 0.79, "learning_rate": 9.205990747275152e-05, "loss": 2.9145, "step": 10126 }, { "epoch": 0.79, "learning_rate": 9.205912334352702e-05, "loss": 3.082, "step": 10127 }, { "epoch": 0.79, "learning_rate": 9.205833921430253e-05, "loss": 3.192, "step": 10128 }, { "epoch": 0.79, "learning_rate": 9.205755508507803e-05, "loss": 3.2724, "step": 10129 }, { "epoch": 0.79, "learning_rate": 9.205677095585353e-05, "loss": 2.9997, "step": 10130 }, { "epoch": 0.79, "learning_rate": 9.205598682662904e-05, "loss": 2.7643, "step": 10131 }, { "epoch": 0.79, "learning_rate": 9.205520269740454e-05, "loss": 2.9532, "step": 10132 }, { "epoch": 0.79, "learning_rate": 9.205441856818004e-05, "loss": 2.9387, "step": 10133 }, { "epoch": 0.79, "learning_rate": 9.205363443895555e-05, "loss": 2.9552, "step": 10134 }, { "epoch": 0.79, "learning_rate": 9.205285030973105e-05, "loss": 3.333, "step": 10135 }, { "epoch": 0.79, "learning_rate": 9.205206618050654e-05, "loss": 3.0753, "step": 10136 }, { "epoch": 0.79, "learning_rate": 9.205128205128205e-05, "loss": 2.9383, "step": 10137 }, { "epoch": 0.79, "learning_rate": 9.205049792205755e-05, "loss": 2.9278, "step": 10138 }, { "epoch": 0.8, "learning_rate": 9.204971379283305e-05, "loss": 2.8921, "step": 10139 }, { "epoch": 0.8, "learning_rate": 9.204892966360856e-05, "loss": 2.9937, "step": 10140 }, { "epoch": 0.8, "learning_rate": 9.204814553438406e-05, "loss": 3.0255, "step": 10141 }, { "epoch": 0.8, "learning_rate": 9.204736140515958e-05, "loss": 2.9463, "step": 10142 }, { "epoch": 0.8, "learning_rate": 9.204657727593508e-05, "loss": 2.9356, "step": 10143 }, { "epoch": 0.8, "learning_rate": 9.204579314671059e-05, "loss": 3.2608, "step": 10144 }, { "epoch": 0.8, "learning_rate": 9.204500901748609e-05, "loss": 3.1341, "step": 10145 }, { "epoch": 0.8, "learning_rate": 9.204422488826159e-05, "loss": 3.2801, "step": 10146 }, { "epoch": 0.8, "learning_rate": 9.20434407590371e-05, "loss": 3.1086, "step": 10147 }, { "epoch": 0.8, "learning_rate": 9.20426566298126e-05, "loss": 3.2239, "step": 10148 }, { "epoch": 0.8, "learning_rate": 9.20418725005881e-05, "loss": 2.927, "step": 10149 }, { "epoch": 0.8, "learning_rate": 9.204108837136361e-05, "loss": 3.008, "step": 10150 }, { "epoch": 0.8, "learning_rate": 9.204030424213911e-05, "loss": 3.0109, "step": 10151 }, { "epoch": 0.8, "learning_rate": 9.203952011291461e-05, "loss": 2.9724, "step": 10152 }, { "epoch": 0.8, "learning_rate": 9.203873598369012e-05, "loss": 3.0933, "step": 10153 }, { "epoch": 0.8, "learning_rate": 9.203795185446562e-05, "loss": 3.0484, "step": 10154 }, { "epoch": 0.8, "learning_rate": 9.203716772524113e-05, "loss": 2.8342, "step": 10155 }, { "epoch": 0.8, "learning_rate": 9.203638359601663e-05, "loss": 2.912, "step": 10156 }, { "epoch": 0.8, "learning_rate": 9.203559946679213e-05, "loss": 3.0286, "step": 10157 }, { "epoch": 0.8, "learning_rate": 9.203481533756764e-05, "loss": 3.0389, "step": 10158 }, { "epoch": 0.8, "learning_rate": 9.203403120834314e-05, "loss": 3.049, "step": 10159 }, { "epoch": 0.8, "learning_rate": 9.203324707911864e-05, "loss": 3.057, "step": 10160 }, { "epoch": 0.8, "learning_rate": 9.203246294989415e-05, "loss": 3.0856, "step": 10161 }, { "epoch": 0.8, "learning_rate": 9.203167882066965e-05, "loss": 3.1321, "step": 10162 }, { "epoch": 0.8, "learning_rate": 9.203089469144516e-05, "loss": 3.0226, "step": 10163 }, { "epoch": 0.8, "learning_rate": 9.203011056222066e-05, "loss": 3.1458, "step": 10164 }, { "epoch": 0.8, "learning_rate": 9.202932643299616e-05, "loss": 3.218, "step": 10165 }, { "epoch": 0.8, "learning_rate": 9.202854230377167e-05, "loss": 2.8911, "step": 10166 }, { "epoch": 0.8, "learning_rate": 9.202775817454717e-05, "loss": 2.8608, "step": 10167 }, { "epoch": 0.8, "learning_rate": 9.202697404532267e-05, "loss": 2.8629, "step": 10168 }, { "epoch": 0.8, "learning_rate": 9.202618991609818e-05, "loss": 3.0354, "step": 10169 }, { "epoch": 0.8, "learning_rate": 9.202540578687368e-05, "loss": 2.824, "step": 10170 }, { "epoch": 0.8, "learning_rate": 9.202462165764918e-05, "loss": 2.9726, "step": 10171 }, { "epoch": 0.8, "learning_rate": 9.202383752842469e-05, "loss": 3.028, "step": 10172 }, { "epoch": 0.8, "learning_rate": 9.202305339920019e-05, "loss": 2.9024, "step": 10173 }, { "epoch": 0.8, "learning_rate": 9.20222692699757e-05, "loss": 3.0672, "step": 10174 }, { "epoch": 0.8, "learning_rate": 9.20214851407512e-05, "loss": 2.7992, "step": 10175 }, { "epoch": 0.8, "learning_rate": 9.20207010115267e-05, "loss": 2.9538, "step": 10176 }, { "epoch": 0.8, "learning_rate": 9.20199168823022e-05, "loss": 2.8846, "step": 10177 }, { "epoch": 0.8, "learning_rate": 9.201913275307771e-05, "loss": 3.0176, "step": 10178 }, { "epoch": 0.8, "learning_rate": 9.201834862385321e-05, "loss": 2.9191, "step": 10179 }, { "epoch": 0.8, "learning_rate": 9.201756449462872e-05, "loss": 2.8378, "step": 10180 }, { "epoch": 0.8, "learning_rate": 9.201678036540422e-05, "loss": 3.2442, "step": 10181 }, { "epoch": 0.8, "learning_rate": 9.201599623617973e-05, "loss": 3.0941, "step": 10182 }, { "epoch": 0.8, "learning_rate": 9.201521210695523e-05, "loss": 3.3272, "step": 10183 }, { "epoch": 0.8, "learning_rate": 9.201442797773073e-05, "loss": 2.9154, "step": 10184 }, { "epoch": 0.8, "learning_rate": 9.201364384850624e-05, "loss": 3.0018, "step": 10185 }, { "epoch": 0.8, "learning_rate": 9.201285971928174e-05, "loss": 3.2219, "step": 10186 }, { "epoch": 0.8, "learning_rate": 9.201207559005724e-05, "loss": 3.1065, "step": 10187 }, { "epoch": 0.8, "learning_rate": 9.201129146083275e-05, "loss": 3.1988, "step": 10188 }, { "epoch": 0.8, "learning_rate": 9.201050733160825e-05, "loss": 3.0571, "step": 10189 }, { "epoch": 0.8, "learning_rate": 9.200972320238375e-05, "loss": 2.7411, "step": 10190 }, { "epoch": 0.8, "learning_rate": 9.200893907315926e-05, "loss": 3.1656, "step": 10191 }, { "epoch": 0.8, "learning_rate": 9.200815494393476e-05, "loss": 3.2254, "step": 10192 }, { "epoch": 0.8, "learning_rate": 9.200737081471027e-05, "loss": 3.1396, "step": 10193 }, { "epoch": 0.8, "learning_rate": 9.200658668548577e-05, "loss": 2.9828, "step": 10194 }, { "epoch": 0.8, "learning_rate": 9.200580255626127e-05, "loss": 3.0298, "step": 10195 }, { "epoch": 0.8, "learning_rate": 9.200501842703678e-05, "loss": 3.2227, "step": 10196 }, { "epoch": 0.8, "learning_rate": 9.200423429781228e-05, "loss": 3.0917, "step": 10197 }, { "epoch": 0.8, "learning_rate": 9.200345016858778e-05, "loss": 2.8332, "step": 10198 }, { "epoch": 0.8, "learning_rate": 9.20026660393633e-05, "loss": 2.7148, "step": 10199 }, { "epoch": 0.8, "learning_rate": 9.20018819101388e-05, "loss": 2.8885, "step": 10200 }, { "epoch": 0.8, "learning_rate": 9.200109778091431e-05, "loss": 3.0416, "step": 10201 }, { "epoch": 0.8, "learning_rate": 9.200031365168981e-05, "loss": 2.8099, "step": 10202 }, { "epoch": 0.8, "learning_rate": 9.199952952246532e-05, "loss": 3.2291, "step": 10203 }, { "epoch": 0.8, "learning_rate": 9.199874539324082e-05, "loss": 2.8269, "step": 10204 }, { "epoch": 0.8, "learning_rate": 9.199796126401632e-05, "loss": 2.9261, "step": 10205 }, { "epoch": 0.8, "learning_rate": 9.199717713479183e-05, "loss": 3.1569, "step": 10206 }, { "epoch": 0.8, "learning_rate": 9.199639300556733e-05, "loss": 3.096, "step": 10207 }, { "epoch": 0.8, "learning_rate": 9.199560887634282e-05, "loss": 2.9709, "step": 10208 }, { "epoch": 0.8, "learning_rate": 9.199482474711832e-05, "loss": 3.1941, "step": 10209 }, { "epoch": 0.8, "learning_rate": 9.199404061789383e-05, "loss": 3.2936, "step": 10210 }, { "epoch": 0.8, "learning_rate": 9.199325648866933e-05, "loss": 3.0176, "step": 10211 }, { "epoch": 0.8, "learning_rate": 9.199247235944484e-05, "loss": 2.7937, "step": 10212 }, { "epoch": 0.8, "learning_rate": 9.199168823022034e-05, "loss": 2.9617, "step": 10213 }, { "epoch": 0.8, "learning_rate": 9.199090410099584e-05, "loss": 2.9055, "step": 10214 }, { "epoch": 0.8, "learning_rate": 9.199011997177135e-05, "loss": 3.0218, "step": 10215 }, { "epoch": 0.8, "learning_rate": 9.198933584254685e-05, "loss": 3.1094, "step": 10216 }, { "epoch": 0.8, "learning_rate": 9.198855171332235e-05, "loss": 2.9372, "step": 10217 }, { "epoch": 0.8, "learning_rate": 9.198776758409786e-05, "loss": 2.9368, "step": 10218 }, { "epoch": 0.8, "learning_rate": 9.198698345487336e-05, "loss": 3.0238, "step": 10219 }, { "epoch": 0.8, "learning_rate": 9.198619932564886e-05, "loss": 3.1812, "step": 10220 }, { "epoch": 0.8, "learning_rate": 9.198541519642437e-05, "loss": 3.1821, "step": 10221 }, { "epoch": 0.8, "learning_rate": 9.198463106719987e-05, "loss": 2.9442, "step": 10222 }, { "epoch": 0.8, "learning_rate": 9.198384693797538e-05, "loss": 3.2668, "step": 10223 }, { "epoch": 0.8, "learning_rate": 9.198306280875088e-05, "loss": 2.8468, "step": 10224 }, { "epoch": 0.8, "learning_rate": 9.19822786795264e-05, "loss": 2.8504, "step": 10225 }, { "epoch": 0.8, "learning_rate": 9.19814945503019e-05, "loss": 2.9471, "step": 10226 }, { "epoch": 0.8, "learning_rate": 9.19807104210774e-05, "loss": 3.0562, "step": 10227 }, { "epoch": 0.8, "learning_rate": 9.197992629185291e-05, "loss": 3.1531, "step": 10228 }, { "epoch": 0.8, "learning_rate": 9.197914216262841e-05, "loss": 3.0568, "step": 10229 }, { "epoch": 0.8, "learning_rate": 9.197835803340392e-05, "loss": 3.2503, "step": 10230 }, { "epoch": 0.8, "learning_rate": 9.197757390417942e-05, "loss": 3.1101, "step": 10231 }, { "epoch": 0.8, "learning_rate": 9.197678977495492e-05, "loss": 3.0491, "step": 10232 }, { "epoch": 0.8, "learning_rate": 9.197600564573043e-05, "loss": 3.0865, "step": 10233 }, { "epoch": 0.8, "learning_rate": 9.197522151650593e-05, "loss": 2.9151, "step": 10234 }, { "epoch": 0.8, "learning_rate": 9.197443738728143e-05, "loss": 3.1885, "step": 10235 }, { "epoch": 0.8, "learning_rate": 9.197365325805694e-05, "loss": 3.0951, "step": 10236 }, { "epoch": 0.8, "learning_rate": 9.197286912883244e-05, "loss": 3.3244, "step": 10237 }, { "epoch": 0.8, "learning_rate": 9.197208499960794e-05, "loss": 3.0654, "step": 10238 }, { "epoch": 0.8, "learning_rate": 9.197130087038345e-05, "loss": 2.9243, "step": 10239 }, { "epoch": 0.8, "learning_rate": 9.197051674115895e-05, "loss": 3.154, "step": 10240 }, { "epoch": 0.8, "learning_rate": 9.196973261193446e-05, "loss": 3.0864, "step": 10241 }, { "epoch": 0.8, "learning_rate": 9.196894848270996e-05, "loss": 2.8119, "step": 10242 }, { "epoch": 0.8, "learning_rate": 9.196816435348546e-05, "loss": 3.0371, "step": 10243 }, { "epoch": 0.8, "learning_rate": 9.196738022426095e-05, "loss": 3.0294, "step": 10244 }, { "epoch": 0.8, "learning_rate": 9.196659609503646e-05, "loss": 3.1351, "step": 10245 }, { "epoch": 0.8, "learning_rate": 9.196581196581196e-05, "loss": 3.0432, "step": 10246 }, { "epoch": 0.8, "learning_rate": 9.196502783658746e-05, "loss": 3.1686, "step": 10247 }, { "epoch": 0.8, "learning_rate": 9.196424370736297e-05, "loss": 2.8489, "step": 10248 }, { "epoch": 0.8, "learning_rate": 9.196345957813847e-05, "loss": 2.9147, "step": 10249 }, { "epoch": 0.8, "learning_rate": 9.196267544891398e-05, "loss": 3.2623, "step": 10250 }, { "epoch": 0.8, "learning_rate": 9.196189131968949e-05, "loss": 2.9204, "step": 10251 }, { "epoch": 0.8, "learning_rate": 9.1961107190465e-05, "loss": 3.1097, "step": 10252 }, { "epoch": 0.8, "learning_rate": 9.19603230612405e-05, "loss": 2.9327, "step": 10253 }, { "epoch": 0.8, "learning_rate": 9.1959538932016e-05, "loss": 3.2707, "step": 10254 }, { "epoch": 0.8, "learning_rate": 9.195875480279151e-05, "loss": 3.0539, "step": 10255 }, { "epoch": 0.8, "learning_rate": 9.195797067356701e-05, "loss": 3.0055, "step": 10256 }, { "epoch": 0.8, "learning_rate": 9.195718654434251e-05, "loss": 2.9307, "step": 10257 }, { "epoch": 0.8, "learning_rate": 9.195640241511802e-05, "loss": 3.006, "step": 10258 }, { "epoch": 0.8, "learning_rate": 9.195561828589352e-05, "loss": 2.9412, "step": 10259 }, { "epoch": 0.8, "learning_rate": 9.195483415666903e-05, "loss": 2.9681, "step": 10260 }, { "epoch": 0.8, "learning_rate": 9.195405002744453e-05, "loss": 3.1347, "step": 10261 }, { "epoch": 0.8, "learning_rate": 9.195326589822003e-05, "loss": 2.764, "step": 10262 }, { "epoch": 0.8, "learning_rate": 9.195248176899554e-05, "loss": 2.8128, "step": 10263 }, { "epoch": 0.8, "learning_rate": 9.195169763977104e-05, "loss": 2.9645, "step": 10264 }, { "epoch": 0.8, "learning_rate": 9.195091351054654e-05, "loss": 2.8383, "step": 10265 }, { "epoch": 0.8, "learning_rate": 9.195012938132205e-05, "loss": 3.2885, "step": 10266 }, { "epoch": 0.81, "learning_rate": 9.194934525209755e-05, "loss": 3.1226, "step": 10267 }, { "epoch": 0.81, "learning_rate": 9.194856112287305e-05, "loss": 3.0717, "step": 10268 }, { "epoch": 0.81, "learning_rate": 9.194777699364856e-05, "loss": 3.2314, "step": 10269 }, { "epoch": 0.81, "learning_rate": 9.194699286442406e-05, "loss": 2.96, "step": 10270 }, { "epoch": 0.81, "learning_rate": 9.194620873519957e-05, "loss": 3.0367, "step": 10271 }, { "epoch": 0.81, "learning_rate": 9.194542460597507e-05, "loss": 3.0349, "step": 10272 }, { "epoch": 0.81, "learning_rate": 9.194464047675057e-05, "loss": 3.4487, "step": 10273 }, { "epoch": 0.81, "learning_rate": 9.194385634752608e-05, "loss": 2.8563, "step": 10274 }, { "epoch": 0.81, "learning_rate": 9.194307221830158e-05, "loss": 2.9174, "step": 10275 }, { "epoch": 0.81, "learning_rate": 9.194228808907708e-05, "loss": 3.2856, "step": 10276 }, { "epoch": 0.81, "learning_rate": 9.194150395985259e-05, "loss": 3.0443, "step": 10277 }, { "epoch": 0.81, "learning_rate": 9.194071983062809e-05, "loss": 2.9521, "step": 10278 }, { "epoch": 0.81, "learning_rate": 9.19399357014036e-05, "loss": 2.9963, "step": 10279 }, { "epoch": 0.81, "learning_rate": 9.19391515721791e-05, "loss": 3.1293, "step": 10280 }, { "epoch": 0.81, "learning_rate": 9.19383674429546e-05, "loss": 2.9442, "step": 10281 }, { "epoch": 0.81, "learning_rate": 9.19375833137301e-05, "loss": 3.1519, "step": 10282 }, { "epoch": 0.81, "learning_rate": 9.193679918450561e-05, "loss": 3.0083, "step": 10283 }, { "epoch": 0.81, "learning_rate": 9.193601505528111e-05, "loss": 3.0483, "step": 10284 }, { "epoch": 0.81, "learning_rate": 9.193523092605662e-05, "loss": 3.1509, "step": 10285 }, { "epoch": 0.81, "learning_rate": 9.193444679683212e-05, "loss": 3.278, "step": 10286 }, { "epoch": 0.81, "learning_rate": 9.193366266760762e-05, "loss": 2.8828, "step": 10287 }, { "epoch": 0.81, "learning_rate": 9.193287853838313e-05, "loss": 2.7109, "step": 10288 }, { "epoch": 0.81, "learning_rate": 9.193209440915863e-05, "loss": 3.1539, "step": 10289 }, { "epoch": 0.81, "learning_rate": 9.193131027993414e-05, "loss": 2.9753, "step": 10290 }, { "epoch": 0.81, "learning_rate": 9.193052615070964e-05, "loss": 3.0558, "step": 10291 }, { "epoch": 0.81, "learning_rate": 9.192974202148514e-05, "loss": 3.088, "step": 10292 }, { "epoch": 0.81, "learning_rate": 9.192895789226065e-05, "loss": 3.1059, "step": 10293 }, { "epoch": 0.81, "learning_rate": 9.192817376303615e-05, "loss": 2.9966, "step": 10294 }, { "epoch": 0.81, "learning_rate": 9.192738963381165e-05, "loss": 3.2177, "step": 10295 }, { "epoch": 0.81, "learning_rate": 9.192660550458716e-05, "loss": 2.9744, "step": 10296 }, { "epoch": 0.81, "learning_rate": 9.192582137536266e-05, "loss": 3.0396, "step": 10297 }, { "epoch": 0.81, "learning_rate": 9.192503724613817e-05, "loss": 3.0441, "step": 10298 }, { "epoch": 0.81, "learning_rate": 9.192425311691367e-05, "loss": 2.8002, "step": 10299 }, { "epoch": 0.81, "learning_rate": 9.192346898768917e-05, "loss": 2.9504, "step": 10300 }, { "epoch": 0.81, "learning_rate": 9.192268485846468e-05, "loss": 3.0175, "step": 10301 }, { "epoch": 0.81, "learning_rate": 9.192190072924018e-05, "loss": 2.821, "step": 10302 }, { "epoch": 0.81, "learning_rate": 9.192111660001568e-05, "loss": 2.8466, "step": 10303 }, { "epoch": 0.81, "learning_rate": 9.192033247079119e-05, "loss": 3.2355, "step": 10304 }, { "epoch": 0.81, "learning_rate": 9.191954834156669e-05, "loss": 2.907, "step": 10305 }, { "epoch": 0.81, "learning_rate": 9.19187642123422e-05, "loss": 3.0614, "step": 10306 }, { "epoch": 0.81, "learning_rate": 9.191798008311771e-05, "loss": 2.9675, "step": 10307 }, { "epoch": 0.81, "learning_rate": 9.191719595389322e-05, "loss": 3.069, "step": 10308 }, { "epoch": 0.81, "learning_rate": 9.191641182466872e-05, "loss": 3.0437, "step": 10309 }, { "epoch": 0.81, "learning_rate": 9.191562769544422e-05, "loss": 3.0343, "step": 10310 }, { "epoch": 0.81, "learning_rate": 9.191484356621973e-05, "loss": 3.0753, "step": 10311 }, { "epoch": 0.81, "learning_rate": 9.191405943699523e-05, "loss": 2.7879, "step": 10312 }, { "epoch": 0.81, "learning_rate": 9.191327530777073e-05, "loss": 2.9679, "step": 10313 }, { "epoch": 0.81, "learning_rate": 9.191249117854624e-05, "loss": 2.8401, "step": 10314 }, { "epoch": 0.81, "learning_rate": 9.191170704932174e-05, "loss": 2.7964, "step": 10315 }, { "epoch": 0.81, "learning_rate": 9.191092292009723e-05, "loss": 3.0681, "step": 10316 }, { "epoch": 0.81, "learning_rate": 9.191013879087274e-05, "loss": 3.0243, "step": 10317 }, { "epoch": 0.81, "learning_rate": 9.190935466164824e-05, "loss": 3.0084, "step": 10318 }, { "epoch": 0.81, "learning_rate": 9.190857053242374e-05, "loss": 2.918, "step": 10319 }, { "epoch": 0.81, "learning_rate": 9.190778640319925e-05, "loss": 2.9146, "step": 10320 }, { "epoch": 0.81, "learning_rate": 9.190700227397475e-05, "loss": 2.9792, "step": 10321 }, { "epoch": 0.81, "learning_rate": 9.190621814475025e-05, "loss": 2.8008, "step": 10322 }, { "epoch": 0.81, "learning_rate": 9.190543401552576e-05, "loss": 2.9866, "step": 10323 }, { "epoch": 0.81, "learning_rate": 9.190464988630126e-05, "loss": 3.1236, "step": 10324 }, { "epoch": 0.81, "learning_rate": 9.190386575707676e-05, "loss": 3.1634, "step": 10325 }, { "epoch": 0.81, "learning_rate": 9.190308162785227e-05, "loss": 2.9297, "step": 10326 }, { "epoch": 0.81, "learning_rate": 9.190229749862777e-05, "loss": 3.0078, "step": 10327 }, { "epoch": 0.81, "learning_rate": 9.190151336940328e-05, "loss": 3.0115, "step": 10328 }, { "epoch": 0.81, "learning_rate": 9.190072924017878e-05, "loss": 2.9408, "step": 10329 }, { "epoch": 0.81, "learning_rate": 9.189994511095428e-05, "loss": 2.9985, "step": 10330 }, { "epoch": 0.81, "learning_rate": 9.189916098172979e-05, "loss": 2.9915, "step": 10331 }, { "epoch": 0.81, "learning_rate": 9.189837685250529e-05, "loss": 2.813, "step": 10332 }, { "epoch": 0.81, "learning_rate": 9.189759272328081e-05, "loss": 3.0657, "step": 10333 }, { "epoch": 0.81, "learning_rate": 9.189680859405631e-05, "loss": 2.8169, "step": 10334 }, { "epoch": 0.81, "learning_rate": 9.189602446483181e-05, "loss": 3.0917, "step": 10335 }, { "epoch": 0.81, "learning_rate": 9.189524033560732e-05, "loss": 2.9249, "step": 10336 }, { "epoch": 0.81, "learning_rate": 9.189445620638282e-05, "loss": 3.1213, "step": 10337 }, { "epoch": 0.81, "learning_rate": 9.189367207715833e-05, "loss": 3.208, "step": 10338 }, { "epoch": 0.81, "learning_rate": 9.189288794793383e-05, "loss": 3.2015, "step": 10339 }, { "epoch": 0.81, "learning_rate": 9.189210381870933e-05, "loss": 3.1913, "step": 10340 }, { "epoch": 0.81, "learning_rate": 9.189131968948484e-05, "loss": 3.1174, "step": 10341 }, { "epoch": 0.81, "learning_rate": 9.189053556026034e-05, "loss": 3.2859, "step": 10342 }, { "epoch": 0.81, "learning_rate": 9.188975143103584e-05, "loss": 2.936, "step": 10343 }, { "epoch": 0.81, "learning_rate": 9.188896730181135e-05, "loss": 2.9885, "step": 10344 }, { "epoch": 0.81, "learning_rate": 9.188818317258685e-05, "loss": 3.0545, "step": 10345 }, { "epoch": 0.81, "learning_rate": 9.188739904336236e-05, "loss": 3.0519, "step": 10346 }, { "epoch": 0.81, "learning_rate": 9.188661491413786e-05, "loss": 2.9468, "step": 10347 }, { "epoch": 0.81, "learning_rate": 9.188583078491336e-05, "loss": 3.054, "step": 10348 }, { "epoch": 0.81, "learning_rate": 9.188504665568887e-05, "loss": 2.82, "step": 10349 }, { "epoch": 0.81, "learning_rate": 9.188426252646437e-05, "loss": 2.7926, "step": 10350 }, { "epoch": 0.81, "learning_rate": 9.188347839723987e-05, "loss": 3.1132, "step": 10351 }, { "epoch": 0.81, "learning_rate": 9.188269426801536e-05, "loss": 3.0613, "step": 10352 }, { "epoch": 0.81, "learning_rate": 9.188191013879087e-05, "loss": 2.8641, "step": 10353 }, { "epoch": 0.81, "learning_rate": 9.188112600956637e-05, "loss": 3.1719, "step": 10354 }, { "epoch": 0.81, "learning_rate": 9.188034188034187e-05, "loss": 3.1913, "step": 10355 }, { "epoch": 0.81, "learning_rate": 9.187955775111738e-05, "loss": 3.049, "step": 10356 }, { "epoch": 0.81, "learning_rate": 9.187877362189288e-05, "loss": 2.819, "step": 10357 }, { "epoch": 0.81, "learning_rate": 9.187798949266839e-05, "loss": 2.9801, "step": 10358 }, { "epoch": 0.81, "learning_rate": 9.18772053634439e-05, "loss": 2.9405, "step": 10359 }, { "epoch": 0.81, "learning_rate": 9.187642123421941e-05, "loss": 3.1054, "step": 10360 }, { "epoch": 0.81, "learning_rate": 9.187563710499491e-05, "loss": 3.0241, "step": 10361 }, { "epoch": 0.81, "learning_rate": 9.187485297577041e-05, "loss": 3.0381, "step": 10362 }, { "epoch": 0.81, "learning_rate": 9.187406884654592e-05, "loss": 3.1439, "step": 10363 }, { "epoch": 0.81, "learning_rate": 9.187328471732142e-05, "loss": 2.9159, "step": 10364 }, { "epoch": 0.81, "learning_rate": 9.187250058809693e-05, "loss": 3.2353, "step": 10365 }, { "epoch": 0.81, "learning_rate": 9.187171645887243e-05, "loss": 2.9154, "step": 10366 }, { "epoch": 0.81, "learning_rate": 9.187093232964793e-05, "loss": 2.9695, "step": 10367 }, { "epoch": 0.81, "learning_rate": 9.187014820042344e-05, "loss": 3.3833, "step": 10368 }, { "epoch": 0.81, "learning_rate": 9.186936407119894e-05, "loss": 3.0336, "step": 10369 }, { "epoch": 0.81, "learning_rate": 9.186857994197444e-05, "loss": 3.2107, "step": 10370 }, { "epoch": 0.81, "learning_rate": 9.186779581274995e-05, "loss": 3.0256, "step": 10371 }, { "epoch": 0.81, "learning_rate": 9.186701168352545e-05, "loss": 2.9725, "step": 10372 }, { "epoch": 0.81, "learning_rate": 9.186622755430095e-05, "loss": 2.7357, "step": 10373 }, { "epoch": 0.81, "learning_rate": 9.186544342507646e-05, "loss": 3.3059, "step": 10374 }, { "epoch": 0.81, "learning_rate": 9.186465929585196e-05, "loss": 2.9404, "step": 10375 }, { "epoch": 0.81, "learning_rate": 9.186387516662747e-05, "loss": 3.0049, "step": 10376 }, { "epoch": 0.81, "learning_rate": 9.186309103740297e-05, "loss": 3.1305, "step": 10377 }, { "epoch": 0.81, "learning_rate": 9.186230690817847e-05, "loss": 2.9057, "step": 10378 }, { "epoch": 0.81, "learning_rate": 9.186152277895398e-05, "loss": 3.1086, "step": 10379 }, { "epoch": 0.81, "learning_rate": 9.186073864972948e-05, "loss": 2.9087, "step": 10380 }, { "epoch": 0.81, "learning_rate": 9.185995452050498e-05, "loss": 3.1072, "step": 10381 }, { "epoch": 0.81, "learning_rate": 9.185917039128049e-05, "loss": 2.9631, "step": 10382 }, { "epoch": 0.81, "learning_rate": 9.185838626205599e-05, "loss": 3.2033, "step": 10383 }, { "epoch": 0.81, "learning_rate": 9.18576021328315e-05, "loss": 2.8714, "step": 10384 }, { "epoch": 0.81, "learning_rate": 9.1856818003607e-05, "loss": 3.1241, "step": 10385 }, { "epoch": 0.81, "learning_rate": 9.18560338743825e-05, "loss": 3.0243, "step": 10386 }, { "epoch": 0.81, "learning_rate": 9.1855249745158e-05, "loss": 3.0657, "step": 10387 }, { "epoch": 0.81, "learning_rate": 9.185446561593351e-05, "loss": 2.9085, "step": 10388 }, { "epoch": 0.81, "learning_rate": 9.185368148670901e-05, "loss": 3.0682, "step": 10389 }, { "epoch": 0.81, "learning_rate": 9.185289735748452e-05, "loss": 3.073, "step": 10390 }, { "epoch": 0.81, "learning_rate": 9.185211322826002e-05, "loss": 2.692, "step": 10391 }, { "epoch": 0.81, "learning_rate": 9.185132909903552e-05, "loss": 3.1157, "step": 10392 }, { "epoch": 0.81, "learning_rate": 9.185054496981103e-05, "loss": 2.9804, "step": 10393 }, { "epoch": 0.82, "learning_rate": 9.184976084058653e-05, "loss": 3.0779, "step": 10394 }, { "epoch": 0.82, "learning_rate": 9.184897671136204e-05, "loss": 3.2256, "step": 10395 }, { "epoch": 0.82, "learning_rate": 9.184819258213754e-05, "loss": 2.8239, "step": 10396 }, { "epoch": 0.82, "learning_rate": 9.184740845291304e-05, "loss": 3.0241, "step": 10397 }, { "epoch": 0.82, "learning_rate": 9.184662432368855e-05, "loss": 2.8197, "step": 10398 }, { "epoch": 0.82, "learning_rate": 9.184584019446405e-05, "loss": 3.0675, "step": 10399 }, { "epoch": 0.82, "learning_rate": 9.184505606523955e-05, "loss": 3.0908, "step": 10400 }, { "epoch": 0.82, "learning_rate": 9.184427193601506e-05, "loss": 2.9086, "step": 10401 }, { "epoch": 0.82, "learning_rate": 9.184348780679056e-05, "loss": 2.9232, "step": 10402 }, { "epoch": 0.82, "learning_rate": 9.184270367756606e-05, "loss": 3.0995, "step": 10403 }, { "epoch": 0.82, "learning_rate": 9.184191954834157e-05, "loss": 3.0938, "step": 10404 }, { "epoch": 0.82, "learning_rate": 9.184113541911707e-05, "loss": 3.0648, "step": 10405 }, { "epoch": 0.82, "learning_rate": 9.184035128989258e-05, "loss": 3.1012, "step": 10406 }, { "epoch": 0.82, "learning_rate": 9.183956716066808e-05, "loss": 3.2285, "step": 10407 }, { "epoch": 0.82, "learning_rate": 9.183878303144358e-05, "loss": 3.2222, "step": 10408 }, { "epoch": 0.82, "learning_rate": 9.183799890221909e-05, "loss": 3.3643, "step": 10409 }, { "epoch": 0.82, "learning_rate": 9.183721477299459e-05, "loss": 2.9775, "step": 10410 }, { "epoch": 0.82, "learning_rate": 9.18364306437701e-05, "loss": 3.1239, "step": 10411 }, { "epoch": 0.82, "learning_rate": 9.18356465145456e-05, "loss": 3.0743, "step": 10412 }, { "epoch": 0.82, "learning_rate": 9.18348623853211e-05, "loss": 3.0271, "step": 10413 }, { "epoch": 0.82, "learning_rate": 9.18340782560966e-05, "loss": 2.9009, "step": 10414 }, { "epoch": 0.82, "learning_rate": 9.183329412687212e-05, "loss": 3.0694, "step": 10415 }, { "epoch": 0.82, "learning_rate": 9.183250999764763e-05, "loss": 3.2399, "step": 10416 }, { "epoch": 0.82, "learning_rate": 9.183172586842313e-05, "loss": 3.3247, "step": 10417 }, { "epoch": 0.82, "learning_rate": 9.183094173919863e-05, "loss": 3.0284, "step": 10418 }, { "epoch": 0.82, "learning_rate": 9.183015760997414e-05, "loss": 3.1801, "step": 10419 }, { "epoch": 0.82, "learning_rate": 9.182937348074964e-05, "loss": 3.0503, "step": 10420 }, { "epoch": 0.82, "learning_rate": 9.182858935152514e-05, "loss": 3.2735, "step": 10421 }, { "epoch": 0.82, "learning_rate": 9.182780522230065e-05, "loss": 2.9036, "step": 10422 }, { "epoch": 0.82, "learning_rate": 9.182702109307615e-05, "loss": 2.6825, "step": 10423 }, { "epoch": 0.82, "learning_rate": 9.182623696385164e-05, "loss": 3.0068, "step": 10424 }, { "epoch": 0.82, "learning_rate": 9.182545283462715e-05, "loss": 2.8783, "step": 10425 }, { "epoch": 0.82, "learning_rate": 9.182466870540265e-05, "loss": 3.0461, "step": 10426 }, { "epoch": 0.82, "learning_rate": 9.182388457617815e-05, "loss": 3.4486, "step": 10427 }, { "epoch": 0.82, "learning_rate": 9.182310044695366e-05, "loss": 2.8217, "step": 10428 }, { "epoch": 0.82, "learning_rate": 9.182231631772916e-05, "loss": 2.8603, "step": 10429 }, { "epoch": 0.82, "learning_rate": 9.182153218850466e-05, "loss": 2.9599, "step": 10430 }, { "epoch": 0.82, "learning_rate": 9.182074805928017e-05, "loss": 3.0245, "step": 10431 }, { "epoch": 0.82, "learning_rate": 9.181996393005567e-05, "loss": 2.9048, "step": 10432 }, { "epoch": 0.82, "learning_rate": 9.181917980083118e-05, "loss": 3.2674, "step": 10433 }, { "epoch": 0.82, "learning_rate": 9.181839567160668e-05, "loss": 3.1004, "step": 10434 }, { "epoch": 0.82, "learning_rate": 9.181761154238218e-05, "loss": 2.8478, "step": 10435 }, { "epoch": 0.82, "learning_rate": 9.181682741315769e-05, "loss": 3.3109, "step": 10436 }, { "epoch": 0.82, "learning_rate": 9.181604328393319e-05, "loss": 2.9426, "step": 10437 }, { "epoch": 0.82, "learning_rate": 9.18152591547087e-05, "loss": 3.3029, "step": 10438 }, { "epoch": 0.82, "learning_rate": 9.18144750254842e-05, "loss": 2.8773, "step": 10439 }, { "epoch": 0.82, "learning_rate": 9.18136908962597e-05, "loss": 2.7911, "step": 10440 }, { "epoch": 0.82, "learning_rate": 9.181290676703522e-05, "loss": 2.8666, "step": 10441 }, { "epoch": 0.82, "learning_rate": 9.181212263781072e-05, "loss": 2.9847, "step": 10442 }, { "epoch": 0.82, "learning_rate": 9.181133850858623e-05, "loss": 2.9243, "step": 10443 }, { "epoch": 0.82, "learning_rate": 9.181055437936173e-05, "loss": 3.1279, "step": 10444 }, { "epoch": 0.82, "learning_rate": 9.180977025013723e-05, "loss": 3.0223, "step": 10445 }, { "epoch": 0.82, "learning_rate": 9.180898612091274e-05, "loss": 3.2175, "step": 10446 }, { "epoch": 0.82, "learning_rate": 9.180820199168824e-05, "loss": 2.9684, "step": 10447 }, { "epoch": 0.82, "learning_rate": 9.180741786246374e-05, "loss": 3.2025, "step": 10448 }, { "epoch": 0.82, "learning_rate": 9.180663373323925e-05, "loss": 2.9728, "step": 10449 }, { "epoch": 0.82, "learning_rate": 9.180584960401475e-05, "loss": 2.8947, "step": 10450 }, { "epoch": 0.82, "learning_rate": 9.180506547479026e-05, "loss": 3.0622, "step": 10451 }, { "epoch": 0.82, "learning_rate": 9.180428134556576e-05, "loss": 2.851, "step": 10452 }, { "epoch": 0.82, "learning_rate": 9.180349721634126e-05, "loss": 2.9577, "step": 10453 }, { "epoch": 0.82, "learning_rate": 9.180271308711677e-05, "loss": 3.0244, "step": 10454 }, { "epoch": 0.82, "learning_rate": 9.180192895789227e-05, "loss": 2.9587, "step": 10455 }, { "epoch": 0.82, "learning_rate": 9.180114482866777e-05, "loss": 2.9184, "step": 10456 }, { "epoch": 0.82, "learning_rate": 9.180036069944328e-05, "loss": 2.9974, "step": 10457 }, { "epoch": 0.82, "learning_rate": 9.179957657021878e-05, "loss": 2.9979, "step": 10458 }, { "epoch": 0.82, "learning_rate": 9.179879244099428e-05, "loss": 3.0451, "step": 10459 }, { "epoch": 0.82, "learning_rate": 9.179800831176977e-05, "loss": 2.8964, "step": 10460 }, { "epoch": 0.82, "learning_rate": 9.179722418254528e-05, "loss": 3.1629, "step": 10461 }, { "epoch": 0.82, "learning_rate": 9.179644005332078e-05, "loss": 2.8677, "step": 10462 }, { "epoch": 0.82, "learning_rate": 9.179565592409629e-05, "loss": 2.8676, "step": 10463 }, { "epoch": 0.82, "learning_rate": 9.179487179487179e-05, "loss": 2.8385, "step": 10464 }, { "epoch": 0.82, "learning_rate": 9.179408766564729e-05, "loss": 3.0474, "step": 10465 }, { "epoch": 0.82, "learning_rate": 9.17933035364228e-05, "loss": 2.9219, "step": 10466 }, { "epoch": 0.82, "learning_rate": 9.179251940719831e-05, "loss": 3.0449, "step": 10467 }, { "epoch": 0.82, "learning_rate": 9.179173527797382e-05, "loss": 2.9852, "step": 10468 }, { "epoch": 0.82, "learning_rate": 9.179095114874932e-05, "loss": 2.8514, "step": 10469 }, { "epoch": 0.82, "learning_rate": 9.179016701952482e-05, "loss": 3.1901, "step": 10470 }, { "epoch": 0.82, "learning_rate": 9.178938289030033e-05, "loss": 3.1743, "step": 10471 }, { "epoch": 0.82, "learning_rate": 9.178859876107583e-05, "loss": 3.0756, "step": 10472 }, { "epoch": 0.82, "learning_rate": 9.178781463185134e-05, "loss": 3.0507, "step": 10473 }, { "epoch": 0.82, "learning_rate": 9.178703050262684e-05, "loss": 2.8242, "step": 10474 }, { "epoch": 0.82, "learning_rate": 9.178624637340234e-05, "loss": 2.9178, "step": 10475 }, { "epoch": 0.82, "learning_rate": 9.178546224417785e-05, "loss": 2.8375, "step": 10476 }, { "epoch": 0.82, "learning_rate": 9.178467811495335e-05, "loss": 2.9073, "step": 10477 }, { "epoch": 0.82, "learning_rate": 9.178389398572885e-05, "loss": 3.1097, "step": 10478 }, { "epoch": 0.82, "learning_rate": 9.178310985650436e-05, "loss": 3.0792, "step": 10479 }, { "epoch": 0.82, "learning_rate": 9.178232572727986e-05, "loss": 3.0478, "step": 10480 }, { "epoch": 0.82, "learning_rate": 9.178154159805537e-05, "loss": 2.8075, "step": 10481 }, { "epoch": 0.82, "learning_rate": 9.178075746883087e-05, "loss": 3.1165, "step": 10482 }, { "epoch": 0.82, "learning_rate": 9.177997333960637e-05, "loss": 3.3168, "step": 10483 }, { "epoch": 0.82, "learning_rate": 9.177918921038188e-05, "loss": 3.1469, "step": 10484 }, { "epoch": 0.82, "learning_rate": 9.177840508115738e-05, "loss": 2.9469, "step": 10485 }, { "epoch": 0.82, "learning_rate": 9.177762095193288e-05, "loss": 2.8868, "step": 10486 }, { "epoch": 0.82, "learning_rate": 9.177683682270839e-05, "loss": 3.0595, "step": 10487 }, { "epoch": 0.82, "learning_rate": 9.177605269348389e-05, "loss": 2.9994, "step": 10488 }, { "epoch": 0.82, "learning_rate": 9.17752685642594e-05, "loss": 3.2647, "step": 10489 }, { "epoch": 0.82, "learning_rate": 9.17744844350349e-05, "loss": 2.7996, "step": 10490 }, { "epoch": 0.82, "learning_rate": 9.17737003058104e-05, "loss": 3.0168, "step": 10491 }, { "epoch": 0.82, "learning_rate": 9.17729161765859e-05, "loss": 2.8786, "step": 10492 }, { "epoch": 0.82, "learning_rate": 9.177213204736141e-05, "loss": 3.2201, "step": 10493 }, { "epoch": 0.82, "learning_rate": 9.177134791813691e-05, "loss": 3.152, "step": 10494 }, { "epoch": 0.82, "learning_rate": 9.177056378891242e-05, "loss": 3.1604, "step": 10495 }, { "epoch": 0.82, "learning_rate": 9.176977965968792e-05, "loss": 3.1087, "step": 10496 }, { "epoch": 0.82, "learning_rate": 9.176899553046342e-05, "loss": 3.1606, "step": 10497 }, { "epoch": 0.82, "learning_rate": 9.176821140123893e-05, "loss": 2.9978, "step": 10498 }, { "epoch": 0.82, "learning_rate": 9.176742727201443e-05, "loss": 3.0637, "step": 10499 }, { "epoch": 0.82, "learning_rate": 9.176664314278994e-05, "loss": 3.0073, "step": 10500 }, { "epoch": 0.82, "learning_rate": 9.176585901356544e-05, "loss": 2.7962, "step": 10501 }, { "epoch": 0.82, "learning_rate": 9.176507488434094e-05, "loss": 2.8753, "step": 10502 }, { "epoch": 0.82, "learning_rate": 9.176429075511645e-05, "loss": 2.8661, "step": 10503 }, { "epoch": 0.82, "learning_rate": 9.176350662589195e-05, "loss": 3.1024, "step": 10504 }, { "epoch": 0.82, "learning_rate": 9.176272249666745e-05, "loss": 3.0015, "step": 10505 }, { "epoch": 0.82, "learning_rate": 9.176193836744296e-05, "loss": 2.9922, "step": 10506 }, { "epoch": 0.82, "learning_rate": 9.176115423821846e-05, "loss": 2.9313, "step": 10507 }, { "epoch": 0.82, "learning_rate": 9.176037010899396e-05, "loss": 2.8741, "step": 10508 }, { "epoch": 0.82, "learning_rate": 9.175958597976947e-05, "loss": 2.983, "step": 10509 }, { "epoch": 0.82, "learning_rate": 9.175880185054497e-05, "loss": 3.01, "step": 10510 }, { "epoch": 0.82, "learning_rate": 9.175801772132048e-05, "loss": 2.8636, "step": 10511 }, { "epoch": 0.82, "learning_rate": 9.175723359209598e-05, "loss": 2.9265, "step": 10512 }, { "epoch": 0.82, "learning_rate": 9.175644946287148e-05, "loss": 3.1464, "step": 10513 }, { "epoch": 0.82, "learning_rate": 9.175566533364699e-05, "loss": 3.2566, "step": 10514 }, { "epoch": 0.82, "learning_rate": 9.175488120442249e-05, "loss": 3.0418, "step": 10515 }, { "epoch": 0.82, "learning_rate": 9.1754097075198e-05, "loss": 3.0384, "step": 10516 }, { "epoch": 0.82, "learning_rate": 9.17533129459735e-05, "loss": 2.8695, "step": 10517 }, { "epoch": 0.82, "learning_rate": 9.1752528816749e-05, "loss": 2.8927, "step": 10518 }, { "epoch": 0.82, "learning_rate": 9.17517446875245e-05, "loss": 3.1043, "step": 10519 }, { "epoch": 0.82, "learning_rate": 9.175096055830001e-05, "loss": 3.5063, "step": 10520 }, { "epoch": 0.82, "learning_rate": 9.175017642907551e-05, "loss": 2.7281, "step": 10521 }, { "epoch": 0.83, "learning_rate": 9.174939229985102e-05, "loss": 2.8478, "step": 10522 }, { "epoch": 0.83, "learning_rate": 9.174860817062653e-05, "loss": 2.9796, "step": 10523 }, { "epoch": 0.83, "learning_rate": 9.174782404140204e-05, "loss": 2.968, "step": 10524 }, { "epoch": 0.83, "learning_rate": 9.174703991217754e-05, "loss": 3.0483, "step": 10525 }, { "epoch": 0.83, "learning_rate": 9.174625578295304e-05, "loss": 2.8418, "step": 10526 }, { "epoch": 0.83, "learning_rate": 9.174547165372855e-05, "loss": 3.0192, "step": 10527 }, { "epoch": 0.83, "learning_rate": 9.174468752450405e-05, "loss": 3.2242, "step": 10528 }, { "epoch": 0.83, "learning_rate": 9.174390339527956e-05, "loss": 2.933, "step": 10529 }, { "epoch": 0.83, "learning_rate": 9.174311926605506e-05, "loss": 3.0298, "step": 10530 }, { "epoch": 0.83, "learning_rate": 9.174233513683056e-05, "loss": 2.8618, "step": 10531 }, { "epoch": 0.83, "learning_rate": 9.174155100760605e-05, "loss": 3.1613, "step": 10532 }, { "epoch": 0.83, "learning_rate": 9.174076687838156e-05, "loss": 3.2774, "step": 10533 }, { "epoch": 0.83, "learning_rate": 9.173998274915706e-05, "loss": 2.9279, "step": 10534 }, { "epoch": 0.83, "learning_rate": 9.173919861993256e-05, "loss": 3.0091, "step": 10535 }, { "epoch": 0.83, "learning_rate": 9.173841449070807e-05, "loss": 3.056, "step": 10536 }, { "epoch": 0.83, "learning_rate": 9.173763036148357e-05, "loss": 3.264, "step": 10537 }, { "epoch": 0.83, "learning_rate": 9.173684623225907e-05, "loss": 3.2428, "step": 10538 }, { "epoch": 0.83, "learning_rate": 9.173606210303458e-05, "loss": 2.8161, "step": 10539 }, { "epoch": 0.83, "learning_rate": 9.173527797381008e-05, "loss": 3.3284, "step": 10540 }, { "epoch": 0.83, "learning_rate": 9.173449384458559e-05, "loss": 3.2042, "step": 10541 }, { "epoch": 0.83, "learning_rate": 9.173370971536109e-05, "loss": 3.4359, "step": 10542 }, { "epoch": 0.83, "learning_rate": 9.17329255861366e-05, "loss": 3.2955, "step": 10543 }, { "epoch": 0.83, "learning_rate": 9.17321414569121e-05, "loss": 2.9714, "step": 10544 }, { "epoch": 0.83, "learning_rate": 9.17313573276876e-05, "loss": 2.954, "step": 10545 }, { "epoch": 0.83, "learning_rate": 9.17305731984631e-05, "loss": 3.0496, "step": 10546 }, { "epoch": 0.83, "learning_rate": 9.172978906923861e-05, "loss": 3.2196, "step": 10547 }, { "epoch": 0.83, "learning_rate": 9.172900494001411e-05, "loss": 3.2198, "step": 10548 }, { "epoch": 0.83, "learning_rate": 9.172822081078963e-05, "loss": 3.2668, "step": 10549 }, { "epoch": 0.83, "learning_rate": 9.172743668156513e-05, "loss": 2.8404, "step": 10550 }, { "epoch": 0.83, "learning_rate": 9.172665255234064e-05, "loss": 3.2125, "step": 10551 }, { "epoch": 0.83, "learning_rate": 9.172586842311614e-05, "loss": 2.8855, "step": 10552 }, { "epoch": 0.83, "learning_rate": 9.172508429389164e-05, "loss": 3.1509, "step": 10553 }, { "epoch": 0.83, "learning_rate": 9.172430016466715e-05, "loss": 3.0607, "step": 10554 }, { "epoch": 0.83, "learning_rate": 9.172351603544265e-05, "loss": 3.1644, "step": 10555 }, { "epoch": 0.83, "learning_rate": 9.172273190621815e-05, "loss": 3.4058, "step": 10556 }, { "epoch": 0.83, "learning_rate": 9.172194777699366e-05, "loss": 3.0905, "step": 10557 }, { "epoch": 0.83, "learning_rate": 9.172116364776916e-05, "loss": 3.274, "step": 10558 }, { "epoch": 0.83, "learning_rate": 9.172037951854467e-05, "loss": 2.9501, "step": 10559 }, { "epoch": 0.83, "learning_rate": 9.171959538932017e-05, "loss": 3.2555, "step": 10560 }, { "epoch": 0.83, "learning_rate": 9.171881126009567e-05, "loss": 3.0438, "step": 10561 }, { "epoch": 0.83, "learning_rate": 9.171802713087118e-05, "loss": 3.2213, "step": 10562 }, { "epoch": 0.83, "learning_rate": 9.171724300164668e-05, "loss": 3.1996, "step": 10563 }, { "epoch": 0.83, "learning_rate": 9.171645887242218e-05, "loss": 3.204, "step": 10564 }, { "epoch": 0.83, "learning_rate": 9.171567474319769e-05, "loss": 2.9911, "step": 10565 }, { "epoch": 0.83, "learning_rate": 9.171489061397319e-05, "loss": 2.9222, "step": 10566 }, { "epoch": 0.83, "learning_rate": 9.17141064847487e-05, "loss": 2.9976, "step": 10567 }, { "epoch": 0.83, "learning_rate": 9.171332235552419e-05, "loss": 3.3039, "step": 10568 }, { "epoch": 0.83, "learning_rate": 9.171253822629969e-05, "loss": 3.151, "step": 10569 }, { "epoch": 0.83, "learning_rate": 9.171175409707519e-05, "loss": 3.1638, "step": 10570 }, { "epoch": 0.83, "learning_rate": 9.17109699678507e-05, "loss": 3.085, "step": 10571 }, { "epoch": 0.83, "learning_rate": 9.17101858386262e-05, "loss": 3.0613, "step": 10572 }, { "epoch": 0.83, "learning_rate": 9.17094017094017e-05, "loss": 2.8494, "step": 10573 }, { "epoch": 0.83, "learning_rate": 9.170861758017721e-05, "loss": 2.9149, "step": 10574 }, { "epoch": 0.83, "learning_rate": 9.170783345095272e-05, "loss": 3.0238, "step": 10575 }, { "epoch": 0.83, "learning_rate": 9.170704932172823e-05, "loss": 2.9287, "step": 10576 }, { "epoch": 0.83, "learning_rate": 9.170626519250373e-05, "loss": 3.0118, "step": 10577 }, { "epoch": 0.83, "learning_rate": 9.170548106327924e-05, "loss": 2.8132, "step": 10578 }, { "epoch": 0.83, "learning_rate": 9.170469693405474e-05, "loss": 3.0859, "step": 10579 }, { "epoch": 0.83, "learning_rate": 9.170391280483024e-05, "loss": 3.0816, "step": 10580 }, { "epoch": 0.83, "learning_rate": 9.170312867560575e-05, "loss": 3.1386, "step": 10581 }, { "epoch": 0.83, "learning_rate": 9.170234454638125e-05, "loss": 3.0096, "step": 10582 }, { "epoch": 0.83, "learning_rate": 9.170156041715675e-05, "loss": 2.9079, "step": 10583 }, { "epoch": 0.83, "learning_rate": 9.170077628793226e-05, "loss": 2.8889, "step": 10584 }, { "epoch": 0.83, "learning_rate": 9.169999215870776e-05, "loss": 2.9745, "step": 10585 }, { "epoch": 0.83, "learning_rate": 9.169920802948327e-05, "loss": 2.9718, "step": 10586 }, { "epoch": 0.83, "learning_rate": 9.169842390025877e-05, "loss": 2.8193, "step": 10587 }, { "epoch": 0.83, "learning_rate": 9.169763977103427e-05, "loss": 3.071, "step": 10588 }, { "epoch": 0.83, "learning_rate": 9.169685564180978e-05, "loss": 3.061, "step": 10589 }, { "epoch": 0.83, "learning_rate": 9.169607151258528e-05, "loss": 2.9035, "step": 10590 }, { "epoch": 0.83, "learning_rate": 9.169528738336078e-05, "loss": 3.3156, "step": 10591 }, { "epoch": 0.83, "learning_rate": 9.169450325413629e-05, "loss": 3.1937, "step": 10592 }, { "epoch": 0.83, "learning_rate": 9.169371912491179e-05, "loss": 2.8793, "step": 10593 }, { "epoch": 0.83, "learning_rate": 9.16929349956873e-05, "loss": 3.0954, "step": 10594 }, { "epoch": 0.83, "learning_rate": 9.16921508664628e-05, "loss": 3.211, "step": 10595 }, { "epoch": 0.83, "learning_rate": 9.16913667372383e-05, "loss": 3.1638, "step": 10596 }, { "epoch": 0.83, "learning_rate": 9.16905826080138e-05, "loss": 2.9325, "step": 10597 }, { "epoch": 0.83, "learning_rate": 9.168979847878931e-05, "loss": 3.0335, "step": 10598 }, { "epoch": 0.83, "learning_rate": 9.168901434956481e-05, "loss": 2.8342, "step": 10599 }, { "epoch": 0.83, "learning_rate": 9.168823022034032e-05, "loss": 3.0324, "step": 10600 }, { "epoch": 0.83, "learning_rate": 9.168744609111582e-05, "loss": 3.0909, "step": 10601 }, { "epoch": 0.83, "learning_rate": 9.168666196189132e-05, "loss": 2.8569, "step": 10602 }, { "epoch": 0.83, "learning_rate": 9.168587783266683e-05, "loss": 2.595, "step": 10603 }, { "epoch": 0.83, "learning_rate": 9.168509370344233e-05, "loss": 3.0545, "step": 10604 }, { "epoch": 0.83, "learning_rate": 9.168430957421783e-05, "loss": 2.9456, "step": 10605 }, { "epoch": 0.83, "learning_rate": 9.168352544499334e-05, "loss": 2.9981, "step": 10606 }, { "epoch": 0.83, "learning_rate": 9.168274131576884e-05, "loss": 3.022, "step": 10607 }, { "epoch": 0.83, "learning_rate": 9.168195718654435e-05, "loss": 3.0697, "step": 10608 }, { "epoch": 0.83, "learning_rate": 9.168117305731985e-05, "loss": 3.0048, "step": 10609 }, { "epoch": 0.83, "learning_rate": 9.168038892809535e-05, "loss": 2.9058, "step": 10610 }, { "epoch": 0.83, "learning_rate": 9.167960479887086e-05, "loss": 2.9128, "step": 10611 }, { "epoch": 0.83, "learning_rate": 9.167882066964636e-05, "loss": 3.1701, "step": 10612 }, { "epoch": 0.83, "learning_rate": 9.167803654042186e-05, "loss": 3.2407, "step": 10613 }, { "epoch": 0.83, "learning_rate": 9.167725241119737e-05, "loss": 2.9043, "step": 10614 }, { "epoch": 0.83, "learning_rate": 9.167646828197287e-05, "loss": 3.099, "step": 10615 }, { "epoch": 0.83, "learning_rate": 9.167568415274838e-05, "loss": 2.8863, "step": 10616 }, { "epoch": 0.83, "learning_rate": 9.167490002352388e-05, "loss": 3.1192, "step": 10617 }, { "epoch": 0.83, "learning_rate": 9.167411589429938e-05, "loss": 2.9487, "step": 10618 }, { "epoch": 0.83, "learning_rate": 9.167333176507489e-05, "loss": 3.0014, "step": 10619 }, { "epoch": 0.83, "learning_rate": 9.167254763585039e-05, "loss": 3.0374, "step": 10620 }, { "epoch": 0.83, "learning_rate": 9.16717635066259e-05, "loss": 3.2116, "step": 10621 }, { "epoch": 0.83, "learning_rate": 9.16709793774014e-05, "loss": 2.9652, "step": 10622 }, { "epoch": 0.83, "learning_rate": 9.16701952481769e-05, "loss": 3.0085, "step": 10623 }, { "epoch": 0.83, "learning_rate": 9.16694111189524e-05, "loss": 3.0663, "step": 10624 }, { "epoch": 0.83, "learning_rate": 9.166862698972791e-05, "loss": 3.2502, "step": 10625 }, { "epoch": 0.83, "learning_rate": 9.166784286050341e-05, "loss": 2.9821, "step": 10626 }, { "epoch": 0.83, "learning_rate": 9.166705873127892e-05, "loss": 2.9276, "step": 10627 }, { "epoch": 0.83, "learning_rate": 9.166627460205442e-05, "loss": 2.7563, "step": 10628 }, { "epoch": 0.83, "learning_rate": 9.166549047282992e-05, "loss": 2.7681, "step": 10629 }, { "epoch": 0.83, "learning_rate": 9.166470634360543e-05, "loss": 2.9827, "step": 10630 }, { "epoch": 0.83, "learning_rate": 9.166392221438093e-05, "loss": 2.9933, "step": 10631 }, { "epoch": 0.83, "learning_rate": 9.166313808515645e-05, "loss": 3.0268, "step": 10632 }, { "epoch": 0.83, "learning_rate": 9.166235395593195e-05, "loss": 2.9202, "step": 10633 }, { "epoch": 0.83, "learning_rate": 9.166156982670746e-05, "loss": 2.994, "step": 10634 }, { "epoch": 0.83, "learning_rate": 9.166078569748296e-05, "loss": 3.2078, "step": 10635 }, { "epoch": 0.83, "learning_rate": 9.166000156825846e-05, "loss": 2.8541, "step": 10636 }, { "epoch": 0.83, "learning_rate": 9.165921743903397e-05, "loss": 2.8253, "step": 10637 }, { "epoch": 0.83, "learning_rate": 9.165843330980947e-05, "loss": 2.9374, "step": 10638 }, { "epoch": 0.83, "learning_rate": 9.165764918058497e-05, "loss": 3.069, "step": 10639 }, { "epoch": 0.83, "learning_rate": 9.165686505136046e-05, "loss": 2.9764, "step": 10640 }, { "epoch": 0.83, "learning_rate": 9.165608092213597e-05, "loss": 3.0048, "step": 10641 }, { "epoch": 0.83, "learning_rate": 9.165529679291147e-05, "loss": 2.961, "step": 10642 }, { "epoch": 0.83, "learning_rate": 9.165451266368697e-05, "loss": 3.0783, "step": 10643 }, { "epoch": 0.83, "learning_rate": 9.165372853446248e-05, "loss": 3.1212, "step": 10644 }, { "epoch": 0.83, "learning_rate": 9.165294440523798e-05, "loss": 3.0999, "step": 10645 }, { "epoch": 0.83, "learning_rate": 9.165216027601349e-05, "loss": 2.7827, "step": 10646 }, { "epoch": 0.83, "learning_rate": 9.165137614678899e-05, "loss": 2.8663, "step": 10647 }, { "epoch": 0.83, "learning_rate": 9.165059201756449e-05, "loss": 3.0432, "step": 10648 }, { "epoch": 0.84, "learning_rate": 9.164980788834e-05, "loss": 3.1175, "step": 10649 }, { "epoch": 0.84, "learning_rate": 9.16490237591155e-05, "loss": 2.9182, "step": 10650 }, { "epoch": 0.84, "learning_rate": 9.1648239629891e-05, "loss": 2.9701, "step": 10651 }, { "epoch": 0.84, "learning_rate": 9.164745550066651e-05, "loss": 3.2263, "step": 10652 }, { "epoch": 0.84, "learning_rate": 9.164667137144201e-05, "loss": 2.7775, "step": 10653 }, { "epoch": 0.84, "learning_rate": 9.164588724221752e-05, "loss": 2.8726, "step": 10654 }, { "epoch": 0.84, "learning_rate": 9.164510311299302e-05, "loss": 2.9542, "step": 10655 }, { "epoch": 0.84, "learning_rate": 9.164431898376852e-05, "loss": 2.7891, "step": 10656 }, { "epoch": 0.84, "learning_rate": 9.164353485454404e-05, "loss": 2.9004, "step": 10657 }, { "epoch": 0.84, "learning_rate": 9.164275072531954e-05, "loss": 3.1581, "step": 10658 }, { "epoch": 0.84, "learning_rate": 9.164196659609505e-05, "loss": 3.0554, "step": 10659 }, { "epoch": 0.84, "learning_rate": 9.164118246687055e-05, "loss": 2.8874, "step": 10660 }, { "epoch": 0.84, "learning_rate": 9.164039833764605e-05, "loss": 2.9266, "step": 10661 }, { "epoch": 0.84, "learning_rate": 9.163961420842156e-05, "loss": 2.9796, "step": 10662 }, { "epoch": 0.84, "learning_rate": 9.163883007919706e-05, "loss": 3.1023, "step": 10663 }, { "epoch": 0.84, "learning_rate": 9.163804594997257e-05, "loss": 3.1571, "step": 10664 }, { "epoch": 0.84, "learning_rate": 9.163726182074807e-05, "loss": 2.953, "step": 10665 }, { "epoch": 0.84, "learning_rate": 9.163647769152357e-05, "loss": 2.9203, "step": 10666 }, { "epoch": 0.84, "learning_rate": 9.163569356229908e-05, "loss": 2.988, "step": 10667 }, { "epoch": 0.84, "learning_rate": 9.163490943307458e-05, "loss": 2.8779, "step": 10668 }, { "epoch": 0.84, "learning_rate": 9.163412530385008e-05, "loss": 2.916, "step": 10669 }, { "epoch": 0.84, "learning_rate": 9.163334117462559e-05, "loss": 2.8685, "step": 10670 }, { "epoch": 0.84, "learning_rate": 9.163255704540109e-05, "loss": 2.8839, "step": 10671 }, { "epoch": 0.84, "learning_rate": 9.16317729161766e-05, "loss": 2.9925, "step": 10672 }, { "epoch": 0.84, "learning_rate": 9.16309887869521e-05, "loss": 2.9719, "step": 10673 }, { "epoch": 0.84, "learning_rate": 9.16302046577276e-05, "loss": 3.1667, "step": 10674 }, { "epoch": 0.84, "learning_rate": 9.16294205285031e-05, "loss": 3.0865, "step": 10675 }, { "epoch": 0.84, "learning_rate": 9.16286363992786e-05, "loss": 3.3594, "step": 10676 }, { "epoch": 0.84, "learning_rate": 9.16278522700541e-05, "loss": 2.8364, "step": 10677 }, { "epoch": 0.84, "learning_rate": 9.16270681408296e-05, "loss": 2.8515, "step": 10678 }, { "epoch": 0.84, "learning_rate": 9.162628401160511e-05, "loss": 2.8089, "step": 10679 }, { "epoch": 0.84, "learning_rate": 9.162549988238061e-05, "loss": 2.9728, "step": 10680 }, { "epoch": 0.84, "learning_rate": 9.162471575315611e-05, "loss": 2.9895, "step": 10681 }, { "epoch": 0.84, "learning_rate": 9.162393162393162e-05, "loss": 3.033, "step": 10682 }, { "epoch": 0.84, "learning_rate": 9.162314749470714e-05, "loss": 2.991, "step": 10683 }, { "epoch": 0.84, "learning_rate": 9.162236336548264e-05, "loss": 2.7841, "step": 10684 }, { "epoch": 0.84, "learning_rate": 9.162157923625814e-05, "loss": 2.8544, "step": 10685 }, { "epoch": 0.84, "learning_rate": 9.162079510703365e-05, "loss": 3.0416, "step": 10686 }, { "epoch": 0.84, "learning_rate": 9.162001097780915e-05, "loss": 2.8632, "step": 10687 }, { "epoch": 0.84, "learning_rate": 9.161922684858465e-05, "loss": 2.8664, "step": 10688 }, { "epoch": 0.84, "learning_rate": 9.161844271936016e-05, "loss": 2.9842, "step": 10689 }, { "epoch": 0.84, "learning_rate": 9.161765859013566e-05, "loss": 2.8956, "step": 10690 }, { "epoch": 0.84, "learning_rate": 9.161687446091116e-05, "loss": 3.1251, "step": 10691 }, { "epoch": 0.84, "learning_rate": 9.161609033168667e-05, "loss": 3.2377, "step": 10692 }, { "epoch": 0.84, "learning_rate": 9.161530620246217e-05, "loss": 2.9379, "step": 10693 }, { "epoch": 0.84, "learning_rate": 9.161452207323768e-05, "loss": 2.9171, "step": 10694 }, { "epoch": 0.84, "learning_rate": 9.161373794401318e-05, "loss": 3.1535, "step": 10695 }, { "epoch": 0.84, "learning_rate": 9.161295381478868e-05, "loss": 3.1185, "step": 10696 }, { "epoch": 0.84, "learning_rate": 9.161216968556419e-05, "loss": 2.917, "step": 10697 }, { "epoch": 0.84, "learning_rate": 9.161138555633969e-05, "loss": 2.8474, "step": 10698 }, { "epoch": 0.84, "learning_rate": 9.16106014271152e-05, "loss": 2.9189, "step": 10699 }, { "epoch": 0.84, "learning_rate": 9.16098172978907e-05, "loss": 2.8907, "step": 10700 }, { "epoch": 0.84, "learning_rate": 9.16090331686662e-05, "loss": 3.1062, "step": 10701 }, { "epoch": 0.84, "learning_rate": 9.16082490394417e-05, "loss": 3.0789, "step": 10702 }, { "epoch": 0.84, "learning_rate": 9.160746491021721e-05, "loss": 2.6146, "step": 10703 }, { "epoch": 0.84, "learning_rate": 9.160668078099271e-05, "loss": 3.1557, "step": 10704 }, { "epoch": 0.84, "learning_rate": 9.160589665176822e-05, "loss": 2.8964, "step": 10705 }, { "epoch": 0.84, "learning_rate": 9.160511252254372e-05, "loss": 3.1681, "step": 10706 }, { "epoch": 0.84, "learning_rate": 9.160432839331922e-05, "loss": 2.8852, "step": 10707 }, { "epoch": 0.84, "learning_rate": 9.160354426409473e-05, "loss": 3.2767, "step": 10708 }, { "epoch": 0.84, "learning_rate": 9.160276013487023e-05, "loss": 3.0795, "step": 10709 }, { "epoch": 0.84, "learning_rate": 9.160197600564573e-05, "loss": 3.1201, "step": 10710 }, { "epoch": 0.84, "learning_rate": 9.160119187642124e-05, "loss": 2.8259, "step": 10711 }, { "epoch": 0.84, "learning_rate": 9.160040774719674e-05, "loss": 2.8316, "step": 10712 }, { "epoch": 0.84, "learning_rate": 9.159962361797225e-05, "loss": 2.9471, "step": 10713 }, { "epoch": 0.84, "learning_rate": 9.159883948874775e-05, "loss": 3.0767, "step": 10714 }, { "epoch": 0.84, "learning_rate": 9.159805535952325e-05, "loss": 3.2382, "step": 10715 }, { "epoch": 0.84, "learning_rate": 9.159727123029876e-05, "loss": 3.0062, "step": 10716 }, { "epoch": 0.84, "learning_rate": 9.159648710107426e-05, "loss": 2.938, "step": 10717 }, { "epoch": 0.84, "learning_rate": 9.159570297184976e-05, "loss": 3.1248, "step": 10718 }, { "epoch": 0.84, "learning_rate": 9.159491884262527e-05, "loss": 3.0681, "step": 10719 }, { "epoch": 0.84, "learning_rate": 9.159413471340077e-05, "loss": 2.8569, "step": 10720 }, { "epoch": 0.84, "learning_rate": 9.159335058417628e-05, "loss": 2.96, "step": 10721 }, { "epoch": 0.84, "learning_rate": 9.159256645495178e-05, "loss": 2.9409, "step": 10722 }, { "epoch": 0.84, "learning_rate": 9.159178232572728e-05, "loss": 3.5683, "step": 10723 }, { "epoch": 0.84, "learning_rate": 9.159099819650279e-05, "loss": 3.3336, "step": 10724 }, { "epoch": 0.84, "learning_rate": 9.159021406727829e-05, "loss": 2.917, "step": 10725 }, { "epoch": 0.84, "learning_rate": 9.15894299380538e-05, "loss": 3.034, "step": 10726 }, { "epoch": 0.84, "learning_rate": 9.15886458088293e-05, "loss": 3.1088, "step": 10727 }, { "epoch": 0.84, "learning_rate": 9.15878616796048e-05, "loss": 2.7705, "step": 10728 }, { "epoch": 0.84, "learning_rate": 9.15870775503803e-05, "loss": 3.0895, "step": 10729 }, { "epoch": 0.84, "learning_rate": 9.158629342115581e-05, "loss": 3.0251, "step": 10730 }, { "epoch": 0.84, "learning_rate": 9.158550929193131e-05, "loss": 2.9729, "step": 10731 }, { "epoch": 0.84, "learning_rate": 9.158472516270682e-05, "loss": 3.1068, "step": 10732 }, { "epoch": 0.84, "learning_rate": 9.158394103348232e-05, "loss": 2.8505, "step": 10733 }, { "epoch": 0.84, "learning_rate": 9.158315690425782e-05, "loss": 3.2755, "step": 10734 }, { "epoch": 0.84, "learning_rate": 9.158237277503333e-05, "loss": 3.0024, "step": 10735 }, { "epoch": 0.84, "learning_rate": 9.158158864580883e-05, "loss": 2.9782, "step": 10736 }, { "epoch": 0.84, "learning_rate": 9.158080451658433e-05, "loss": 3.0072, "step": 10737 }, { "epoch": 0.84, "learning_rate": 9.158002038735984e-05, "loss": 3.0619, "step": 10738 }, { "epoch": 0.84, "learning_rate": 9.157923625813534e-05, "loss": 3.0076, "step": 10739 }, { "epoch": 0.84, "learning_rate": 9.157845212891086e-05, "loss": 2.9735, "step": 10740 }, { "epoch": 0.84, "learning_rate": 9.157766799968636e-05, "loss": 3.1239, "step": 10741 }, { "epoch": 0.84, "learning_rate": 9.157688387046187e-05, "loss": 3.2016, "step": 10742 }, { "epoch": 0.84, "learning_rate": 9.157609974123737e-05, "loss": 3.3666, "step": 10743 }, { "epoch": 0.84, "learning_rate": 9.157531561201287e-05, "loss": 3.076, "step": 10744 }, { "epoch": 0.84, "learning_rate": 9.157453148278838e-05, "loss": 3.0713, "step": 10745 }, { "epoch": 0.84, "learning_rate": 9.157374735356388e-05, "loss": 2.9733, "step": 10746 }, { "epoch": 0.84, "learning_rate": 9.157296322433938e-05, "loss": 3.0517, "step": 10747 }, { "epoch": 0.84, "learning_rate": 9.157217909511487e-05, "loss": 2.9331, "step": 10748 }, { "epoch": 0.84, "learning_rate": 9.157139496589038e-05, "loss": 2.9088, "step": 10749 }, { "epoch": 0.84, "learning_rate": 9.157061083666588e-05, "loss": 2.8261, "step": 10750 }, { "epoch": 0.84, "learning_rate": 9.156982670744139e-05, "loss": 2.984, "step": 10751 }, { "epoch": 0.84, "learning_rate": 9.156904257821689e-05, "loss": 3.1883, "step": 10752 }, { "epoch": 0.84, "learning_rate": 9.156825844899239e-05, "loss": 3.0712, "step": 10753 }, { "epoch": 0.84, "learning_rate": 9.15674743197679e-05, "loss": 2.7996, "step": 10754 }, { "epoch": 0.84, "learning_rate": 9.15666901905434e-05, "loss": 3.0727, "step": 10755 }, { "epoch": 0.84, "learning_rate": 9.15659060613189e-05, "loss": 2.9765, "step": 10756 }, { "epoch": 0.84, "learning_rate": 9.156512193209441e-05, "loss": 3.0257, "step": 10757 }, { "epoch": 0.84, "learning_rate": 9.156433780286991e-05, "loss": 2.9727, "step": 10758 }, { "epoch": 0.84, "learning_rate": 9.156355367364541e-05, "loss": 3.1688, "step": 10759 }, { "epoch": 0.84, "learning_rate": 9.156276954442092e-05, "loss": 3.3495, "step": 10760 }, { "epoch": 0.84, "learning_rate": 9.156198541519642e-05, "loss": 2.8525, "step": 10761 }, { "epoch": 0.84, "learning_rate": 9.156120128597193e-05, "loss": 3.0625, "step": 10762 }, { "epoch": 0.84, "learning_rate": 9.156041715674743e-05, "loss": 3.0327, "step": 10763 }, { "epoch": 0.84, "learning_rate": 9.155963302752293e-05, "loss": 3.0594, "step": 10764 }, { "epoch": 0.84, "learning_rate": 9.155884889829845e-05, "loss": 3.4887, "step": 10765 }, { "epoch": 0.84, "learning_rate": 9.155806476907395e-05, "loss": 2.7901, "step": 10766 }, { "epoch": 0.84, "learning_rate": 9.155728063984946e-05, "loss": 3.3246, "step": 10767 }, { "epoch": 0.84, "learning_rate": 9.155649651062496e-05, "loss": 3.2156, "step": 10768 }, { "epoch": 0.84, "learning_rate": 9.155571238140047e-05, "loss": 2.7751, "step": 10769 }, { "epoch": 0.84, "learning_rate": 9.155492825217597e-05, "loss": 2.9964, "step": 10770 }, { "epoch": 0.84, "learning_rate": 9.155414412295147e-05, "loss": 3.0305, "step": 10771 }, { "epoch": 0.84, "learning_rate": 9.155335999372698e-05, "loss": 3.0292, "step": 10772 }, { "epoch": 0.84, "learning_rate": 9.155257586450248e-05, "loss": 2.8758, "step": 10773 }, { "epoch": 0.84, "learning_rate": 9.155179173527798e-05, "loss": 2.898, "step": 10774 }, { "epoch": 0.84, "learning_rate": 9.155100760605349e-05, "loss": 3.1071, "step": 10775 }, { "epoch": 0.84, "learning_rate": 9.155022347682899e-05, "loss": 2.963, "step": 10776 }, { "epoch": 0.85, "learning_rate": 9.15494393476045e-05, "loss": 2.8732, "step": 10777 }, { "epoch": 0.85, "learning_rate": 9.154865521838e-05, "loss": 2.8547, "step": 10778 }, { "epoch": 0.85, "learning_rate": 9.15478710891555e-05, "loss": 3.0226, "step": 10779 }, { "epoch": 0.85, "learning_rate": 9.1547086959931e-05, "loss": 2.9904, "step": 10780 }, { "epoch": 0.85, "learning_rate": 9.154630283070651e-05, "loss": 3.2451, "step": 10781 }, { "epoch": 0.85, "learning_rate": 9.154551870148201e-05, "loss": 2.8813, "step": 10782 }, { "epoch": 0.85, "learning_rate": 9.154473457225752e-05, "loss": 2.9247, "step": 10783 }, { "epoch": 0.85, "learning_rate": 9.154395044303301e-05, "loss": 3.0813, "step": 10784 }, { "epoch": 0.85, "learning_rate": 9.154316631380851e-05, "loss": 3.044, "step": 10785 }, { "epoch": 0.85, "learning_rate": 9.154238218458401e-05, "loss": 2.7963, "step": 10786 }, { "epoch": 0.85, "learning_rate": 9.154159805535952e-05, "loss": 3.1354, "step": 10787 }, { "epoch": 0.85, "learning_rate": 9.154081392613502e-05, "loss": 3.0399, "step": 10788 }, { "epoch": 0.85, "learning_rate": 9.154002979691053e-05, "loss": 2.8917, "step": 10789 }, { "epoch": 0.85, "learning_rate": 9.153924566768603e-05, "loss": 3.0863, "step": 10790 }, { "epoch": 0.85, "learning_rate": 9.153846153846155e-05, "loss": 3.0776, "step": 10791 }, { "epoch": 0.85, "learning_rate": 9.153767740923705e-05, "loss": 3.3517, "step": 10792 }, { "epoch": 0.85, "learning_rate": 9.153689328001255e-05, "loss": 3.1342, "step": 10793 }, { "epoch": 0.85, "learning_rate": 9.153610915078806e-05, "loss": 3.0006, "step": 10794 }, { "epoch": 0.85, "learning_rate": 9.153532502156356e-05, "loss": 3.2088, "step": 10795 }, { "epoch": 0.85, "learning_rate": 9.153454089233906e-05, "loss": 3.0114, "step": 10796 }, { "epoch": 0.85, "learning_rate": 9.153375676311457e-05, "loss": 2.8175, "step": 10797 }, { "epoch": 0.85, "learning_rate": 9.153297263389007e-05, "loss": 2.8922, "step": 10798 }, { "epoch": 0.85, "learning_rate": 9.153218850466558e-05, "loss": 3.0461, "step": 10799 }, { "epoch": 0.85, "learning_rate": 9.153140437544108e-05, "loss": 3.2411, "step": 10800 }, { "epoch": 0.85, "learning_rate": 9.153062024621658e-05, "loss": 3.1525, "step": 10801 }, { "epoch": 0.85, "learning_rate": 9.152983611699209e-05, "loss": 3.1278, "step": 10802 }, { "epoch": 0.85, "learning_rate": 9.152905198776759e-05, "loss": 2.9998, "step": 10803 }, { "epoch": 0.85, "learning_rate": 9.15282678585431e-05, "loss": 2.8643, "step": 10804 }, { "epoch": 0.85, "learning_rate": 9.15274837293186e-05, "loss": 2.9429, "step": 10805 }, { "epoch": 0.85, "learning_rate": 9.15266996000941e-05, "loss": 2.8928, "step": 10806 }, { "epoch": 0.85, "learning_rate": 9.15259154708696e-05, "loss": 3.1796, "step": 10807 }, { "epoch": 0.85, "learning_rate": 9.152513134164511e-05, "loss": 3.4534, "step": 10808 }, { "epoch": 0.85, "learning_rate": 9.152434721242061e-05, "loss": 2.8734, "step": 10809 }, { "epoch": 0.85, "learning_rate": 9.152356308319612e-05, "loss": 3.0569, "step": 10810 }, { "epoch": 0.85, "learning_rate": 9.152277895397162e-05, "loss": 2.9053, "step": 10811 }, { "epoch": 0.85, "learning_rate": 9.152199482474712e-05, "loss": 2.9705, "step": 10812 }, { "epoch": 0.85, "learning_rate": 9.152121069552263e-05, "loss": 2.8774, "step": 10813 }, { "epoch": 0.85, "learning_rate": 9.152042656629813e-05, "loss": 3.0949, "step": 10814 }, { "epoch": 0.85, "learning_rate": 9.151964243707363e-05, "loss": 3.1166, "step": 10815 }, { "epoch": 0.85, "learning_rate": 9.151885830784914e-05, "loss": 2.8595, "step": 10816 }, { "epoch": 0.85, "learning_rate": 9.151807417862464e-05, "loss": 3.1988, "step": 10817 }, { "epoch": 0.85, "learning_rate": 9.151729004940015e-05, "loss": 2.9694, "step": 10818 }, { "epoch": 0.85, "learning_rate": 9.151650592017565e-05, "loss": 2.9063, "step": 10819 }, { "epoch": 0.85, "learning_rate": 9.151572179095115e-05, "loss": 2.8737, "step": 10820 }, { "epoch": 0.85, "learning_rate": 9.151493766172666e-05, "loss": 2.8721, "step": 10821 }, { "epoch": 0.85, "learning_rate": 9.151415353250216e-05, "loss": 2.8477, "step": 10822 }, { "epoch": 0.85, "learning_rate": 9.151336940327766e-05, "loss": 3.2625, "step": 10823 }, { "epoch": 0.85, "learning_rate": 9.151258527405317e-05, "loss": 2.9366, "step": 10824 }, { "epoch": 0.85, "learning_rate": 9.151180114482867e-05, "loss": 3.0265, "step": 10825 }, { "epoch": 0.85, "learning_rate": 9.151101701560417e-05, "loss": 3.0715, "step": 10826 }, { "epoch": 0.85, "learning_rate": 9.151023288637968e-05, "loss": 2.977, "step": 10827 }, { "epoch": 0.85, "learning_rate": 9.150944875715518e-05, "loss": 3.1922, "step": 10828 }, { "epoch": 0.85, "learning_rate": 9.150866462793069e-05, "loss": 2.8319, "step": 10829 }, { "epoch": 0.85, "learning_rate": 9.150788049870619e-05, "loss": 3.2287, "step": 10830 }, { "epoch": 0.85, "learning_rate": 9.150709636948169e-05, "loss": 3.1328, "step": 10831 }, { "epoch": 0.85, "learning_rate": 9.15063122402572e-05, "loss": 3.1286, "step": 10832 }, { "epoch": 0.85, "learning_rate": 9.15055281110327e-05, "loss": 3.1518, "step": 10833 }, { "epoch": 0.85, "learning_rate": 9.15047439818082e-05, "loss": 2.9143, "step": 10834 }, { "epoch": 0.85, "learning_rate": 9.150395985258371e-05, "loss": 3.1623, "step": 10835 }, { "epoch": 0.85, "learning_rate": 9.150317572335921e-05, "loss": 3.1646, "step": 10836 }, { "epoch": 0.85, "learning_rate": 9.150239159413472e-05, "loss": 2.989, "step": 10837 }, { "epoch": 0.85, "learning_rate": 9.150160746491022e-05, "loss": 2.9072, "step": 10838 }, { "epoch": 0.85, "learning_rate": 9.150082333568572e-05, "loss": 3.254, "step": 10839 }, { "epoch": 0.85, "learning_rate": 9.150003920646123e-05, "loss": 2.8202, "step": 10840 }, { "epoch": 0.85, "learning_rate": 9.149925507723673e-05, "loss": 2.788, "step": 10841 }, { "epoch": 0.85, "learning_rate": 9.149847094801223e-05, "loss": 3.1334, "step": 10842 }, { "epoch": 0.85, "learning_rate": 9.149768681878774e-05, "loss": 2.8821, "step": 10843 }, { "epoch": 0.85, "learning_rate": 9.149690268956324e-05, "loss": 2.8939, "step": 10844 }, { "epoch": 0.85, "learning_rate": 9.149611856033874e-05, "loss": 3.1548, "step": 10845 }, { "epoch": 0.85, "learning_rate": 9.149533443111425e-05, "loss": 3.1191, "step": 10846 }, { "epoch": 0.85, "learning_rate": 9.149455030188975e-05, "loss": 3.3986, "step": 10847 }, { "epoch": 0.85, "learning_rate": 9.149376617266527e-05, "loss": 3.0087, "step": 10848 }, { "epoch": 0.85, "learning_rate": 9.149298204344077e-05, "loss": 2.9366, "step": 10849 }, { "epoch": 0.85, "learning_rate": 9.149219791421628e-05, "loss": 3.4194, "step": 10850 }, { "epoch": 0.85, "learning_rate": 9.149141378499178e-05, "loss": 3.0258, "step": 10851 }, { "epoch": 0.85, "learning_rate": 9.149062965576728e-05, "loss": 3.3583, "step": 10852 }, { "epoch": 0.85, "learning_rate": 9.148984552654279e-05, "loss": 2.7764, "step": 10853 }, { "epoch": 0.85, "learning_rate": 9.148906139731829e-05, "loss": 3.0868, "step": 10854 }, { "epoch": 0.85, "learning_rate": 9.14882772680938e-05, "loss": 2.9323, "step": 10855 }, { "epoch": 0.85, "learning_rate": 9.148749313886929e-05, "loss": 2.7856, "step": 10856 }, { "epoch": 0.85, "learning_rate": 9.148670900964479e-05, "loss": 2.8222, "step": 10857 }, { "epoch": 0.85, "learning_rate": 9.148592488042029e-05, "loss": 2.9556, "step": 10858 }, { "epoch": 0.85, "learning_rate": 9.14851407511958e-05, "loss": 2.8751, "step": 10859 }, { "epoch": 0.85, "learning_rate": 9.14843566219713e-05, "loss": 2.9646, "step": 10860 }, { "epoch": 0.85, "learning_rate": 9.14835724927468e-05, "loss": 2.9717, "step": 10861 }, { "epoch": 0.85, "learning_rate": 9.148278836352231e-05, "loss": 2.8432, "step": 10862 }, { "epoch": 0.85, "learning_rate": 9.148200423429781e-05, "loss": 2.9174, "step": 10863 }, { "epoch": 0.85, "learning_rate": 9.148122010507331e-05, "loss": 3.1244, "step": 10864 }, { "epoch": 0.85, "learning_rate": 9.148043597584882e-05, "loss": 2.943, "step": 10865 }, { "epoch": 0.85, "learning_rate": 9.147965184662432e-05, "loss": 3.2919, "step": 10866 }, { "epoch": 0.85, "learning_rate": 9.147886771739983e-05, "loss": 3.016, "step": 10867 }, { "epoch": 0.85, "learning_rate": 9.147808358817533e-05, "loss": 3.0897, "step": 10868 }, { "epoch": 0.85, "learning_rate": 9.147729945895083e-05, "loss": 3.0402, "step": 10869 }, { "epoch": 0.85, "learning_rate": 9.147651532972634e-05, "loss": 2.8857, "step": 10870 }, { "epoch": 0.85, "learning_rate": 9.147573120050184e-05, "loss": 2.9588, "step": 10871 }, { "epoch": 0.85, "learning_rate": 9.147494707127734e-05, "loss": 2.9474, "step": 10872 }, { "epoch": 0.85, "learning_rate": 9.147416294205286e-05, "loss": 3.0114, "step": 10873 }, { "epoch": 0.85, "learning_rate": 9.147337881282836e-05, "loss": 2.964, "step": 10874 }, { "epoch": 0.85, "learning_rate": 9.147259468360387e-05, "loss": 3.0205, "step": 10875 }, { "epoch": 0.85, "learning_rate": 9.147181055437937e-05, "loss": 2.9622, "step": 10876 }, { "epoch": 0.85, "learning_rate": 9.147102642515488e-05, "loss": 2.9471, "step": 10877 }, { "epoch": 0.85, "learning_rate": 9.147024229593038e-05, "loss": 3.0628, "step": 10878 }, { "epoch": 0.85, "learning_rate": 9.146945816670588e-05, "loss": 2.9512, "step": 10879 }, { "epoch": 0.85, "learning_rate": 9.146867403748139e-05, "loss": 2.9022, "step": 10880 }, { "epoch": 0.85, "learning_rate": 9.146788990825689e-05, "loss": 3.1919, "step": 10881 }, { "epoch": 0.85, "learning_rate": 9.14671057790324e-05, "loss": 2.841, "step": 10882 }, { "epoch": 0.85, "learning_rate": 9.14663216498079e-05, "loss": 3.1463, "step": 10883 }, { "epoch": 0.85, "learning_rate": 9.14655375205834e-05, "loss": 3.0788, "step": 10884 }, { "epoch": 0.85, "learning_rate": 9.14647533913589e-05, "loss": 3.1624, "step": 10885 }, { "epoch": 0.85, "learning_rate": 9.146396926213441e-05, "loss": 2.9786, "step": 10886 }, { "epoch": 0.85, "learning_rate": 9.146318513290991e-05, "loss": 2.8754, "step": 10887 }, { "epoch": 0.85, "learning_rate": 9.146240100368542e-05, "loss": 3.1393, "step": 10888 }, { "epoch": 0.85, "learning_rate": 9.146161687446092e-05, "loss": 2.8651, "step": 10889 }, { "epoch": 0.85, "learning_rate": 9.146083274523642e-05, "loss": 2.8693, "step": 10890 }, { "epoch": 0.85, "learning_rate": 9.146004861601193e-05, "loss": 2.9273, "step": 10891 }, { "epoch": 0.85, "learning_rate": 9.145926448678742e-05, "loss": 2.9075, "step": 10892 }, { "epoch": 0.85, "learning_rate": 9.145848035756292e-05, "loss": 3.1938, "step": 10893 }, { "epoch": 0.85, "learning_rate": 9.145769622833842e-05, "loss": 2.8058, "step": 10894 }, { "epoch": 0.85, "learning_rate": 9.145691209911393e-05, "loss": 3.6851, "step": 10895 }, { "epoch": 0.85, "learning_rate": 9.145612796988943e-05, "loss": 3.1535, "step": 10896 }, { "epoch": 0.85, "learning_rate": 9.145534384066494e-05, "loss": 2.9522, "step": 10897 }, { "epoch": 0.85, "learning_rate": 9.145455971144044e-05, "loss": 3.1052, "step": 10898 }, { "epoch": 0.85, "learning_rate": 9.145377558221596e-05, "loss": 3.2097, "step": 10899 }, { "epoch": 0.85, "learning_rate": 9.145299145299146e-05, "loss": 2.8881, "step": 10900 }, { "epoch": 0.85, "learning_rate": 9.145220732376696e-05, "loss": 3.1175, "step": 10901 }, { "epoch": 0.85, "learning_rate": 9.145142319454247e-05, "loss": 3.3785, "step": 10902 }, { "epoch": 0.85, "learning_rate": 9.145063906531797e-05, "loss": 3.1252, "step": 10903 }, { "epoch": 0.86, "learning_rate": 9.144985493609348e-05, "loss": 3.2536, "step": 10904 }, { "epoch": 0.86, "learning_rate": 9.144907080686898e-05, "loss": 3.0202, "step": 10905 }, { "epoch": 0.86, "learning_rate": 9.144828667764448e-05, "loss": 3.0893, "step": 10906 }, { "epoch": 0.86, "learning_rate": 9.144750254841999e-05, "loss": 3.1232, "step": 10907 }, { "epoch": 0.86, "learning_rate": 9.144671841919549e-05, "loss": 3.2864, "step": 10908 }, { "epoch": 0.86, "learning_rate": 9.1445934289971e-05, "loss": 2.8831, "step": 10909 }, { "epoch": 0.86, "learning_rate": 9.14451501607465e-05, "loss": 2.8438, "step": 10910 }, { "epoch": 0.86, "learning_rate": 9.1444366031522e-05, "loss": 2.9384, "step": 10911 }, { "epoch": 0.86, "learning_rate": 9.14435819022975e-05, "loss": 2.8696, "step": 10912 }, { "epoch": 0.86, "learning_rate": 9.144279777307301e-05, "loss": 2.8752, "step": 10913 }, { "epoch": 0.86, "learning_rate": 9.144201364384851e-05, "loss": 3.1341, "step": 10914 }, { "epoch": 0.86, "learning_rate": 9.144122951462402e-05, "loss": 2.9996, "step": 10915 }, { "epoch": 0.86, "learning_rate": 9.144044538539952e-05, "loss": 2.8354, "step": 10916 }, { "epoch": 0.86, "learning_rate": 9.143966125617502e-05, "loss": 2.9818, "step": 10917 }, { "epoch": 0.86, "learning_rate": 9.143887712695053e-05, "loss": 3.0164, "step": 10918 }, { "epoch": 0.86, "learning_rate": 9.143809299772603e-05, "loss": 3.1299, "step": 10919 }, { "epoch": 0.86, "learning_rate": 9.143730886850153e-05, "loss": 3.0124, "step": 10920 }, { "epoch": 0.86, "learning_rate": 9.143652473927704e-05, "loss": 2.9859, "step": 10921 }, { "epoch": 0.86, "learning_rate": 9.143574061005254e-05, "loss": 2.8787, "step": 10922 }, { "epoch": 0.86, "learning_rate": 9.143495648082805e-05, "loss": 2.8831, "step": 10923 }, { "epoch": 0.86, "learning_rate": 9.143417235160355e-05, "loss": 3.1258, "step": 10924 }, { "epoch": 0.86, "learning_rate": 9.143338822237905e-05, "loss": 3.1742, "step": 10925 }, { "epoch": 0.86, "learning_rate": 9.143260409315456e-05, "loss": 2.8335, "step": 10926 }, { "epoch": 0.86, "learning_rate": 9.143181996393006e-05, "loss": 2.9506, "step": 10927 }, { "epoch": 0.86, "learning_rate": 9.143103583470556e-05, "loss": 2.9184, "step": 10928 }, { "epoch": 0.86, "learning_rate": 9.143025170548107e-05, "loss": 2.9507, "step": 10929 }, { "epoch": 0.86, "learning_rate": 9.142946757625657e-05, "loss": 2.9312, "step": 10930 }, { "epoch": 0.86, "learning_rate": 9.142868344703207e-05, "loss": 2.9573, "step": 10931 }, { "epoch": 0.86, "learning_rate": 9.142789931780758e-05, "loss": 2.8835, "step": 10932 }, { "epoch": 0.86, "learning_rate": 9.142711518858308e-05, "loss": 3.2452, "step": 10933 }, { "epoch": 0.86, "learning_rate": 9.142633105935859e-05, "loss": 3.187, "step": 10934 }, { "epoch": 0.86, "learning_rate": 9.142554693013409e-05, "loss": 3.0975, "step": 10935 }, { "epoch": 0.86, "learning_rate": 9.142476280090959e-05, "loss": 2.9119, "step": 10936 }, { "epoch": 0.86, "learning_rate": 9.14239786716851e-05, "loss": 3.3132, "step": 10937 }, { "epoch": 0.86, "learning_rate": 9.14231945424606e-05, "loss": 2.714, "step": 10938 }, { "epoch": 0.86, "learning_rate": 9.14224104132361e-05, "loss": 3.3547, "step": 10939 }, { "epoch": 0.86, "learning_rate": 9.142162628401161e-05, "loss": 3.2703, "step": 10940 }, { "epoch": 0.86, "learning_rate": 9.142084215478711e-05, "loss": 2.936, "step": 10941 }, { "epoch": 0.86, "learning_rate": 9.142005802556261e-05, "loss": 2.8657, "step": 10942 }, { "epoch": 0.86, "learning_rate": 9.141927389633812e-05, "loss": 3.188, "step": 10943 }, { "epoch": 0.86, "learning_rate": 9.141848976711362e-05, "loss": 3.3884, "step": 10944 }, { "epoch": 0.86, "learning_rate": 9.141770563788913e-05, "loss": 3.0936, "step": 10945 }, { "epoch": 0.86, "learning_rate": 9.141692150866463e-05, "loss": 3.0432, "step": 10946 }, { "epoch": 0.86, "learning_rate": 9.141613737944013e-05, "loss": 3.0352, "step": 10947 }, { "epoch": 0.86, "learning_rate": 9.141535325021564e-05, "loss": 2.8826, "step": 10948 }, { "epoch": 0.86, "learning_rate": 9.141456912099114e-05, "loss": 3.2546, "step": 10949 }, { "epoch": 0.86, "learning_rate": 9.141378499176664e-05, "loss": 3.2692, "step": 10950 }, { "epoch": 0.86, "learning_rate": 9.141300086254215e-05, "loss": 3.216, "step": 10951 }, { "epoch": 0.86, "learning_rate": 9.141221673331765e-05, "loss": 2.8152, "step": 10952 }, { "epoch": 0.86, "learning_rate": 9.141143260409316e-05, "loss": 3.0625, "step": 10953 }, { "epoch": 0.86, "learning_rate": 9.141064847486866e-05, "loss": 3.0528, "step": 10954 }, { "epoch": 0.86, "learning_rate": 9.140986434564416e-05, "loss": 2.9201, "step": 10955 }, { "epoch": 0.86, "learning_rate": 9.140908021641968e-05, "loss": 3.0065, "step": 10956 }, { "epoch": 0.86, "learning_rate": 9.140829608719518e-05, "loss": 2.8998, "step": 10957 }, { "epoch": 0.86, "learning_rate": 9.140751195797069e-05, "loss": 2.9588, "step": 10958 }, { "epoch": 0.86, "learning_rate": 9.140672782874619e-05, "loss": 2.8527, "step": 10959 }, { "epoch": 0.86, "learning_rate": 9.14059436995217e-05, "loss": 3.2586, "step": 10960 }, { "epoch": 0.86, "learning_rate": 9.14051595702972e-05, "loss": 2.8881, "step": 10961 }, { "epoch": 0.86, "learning_rate": 9.14043754410727e-05, "loss": 3.0264, "step": 10962 }, { "epoch": 0.86, "learning_rate": 9.14035913118482e-05, "loss": 2.9325, "step": 10963 }, { "epoch": 0.86, "learning_rate": 9.14028071826237e-05, "loss": 3.0325, "step": 10964 }, { "epoch": 0.86, "learning_rate": 9.14020230533992e-05, "loss": 3.3135, "step": 10965 }, { "epoch": 0.86, "learning_rate": 9.14012389241747e-05, "loss": 2.9838, "step": 10966 }, { "epoch": 0.86, "learning_rate": 9.140045479495021e-05, "loss": 2.8715, "step": 10967 }, { "epoch": 0.86, "learning_rate": 9.139967066572571e-05, "loss": 3.2402, "step": 10968 }, { "epoch": 0.86, "learning_rate": 9.139888653650121e-05, "loss": 2.7799, "step": 10969 }, { "epoch": 0.86, "learning_rate": 9.139810240727672e-05, "loss": 3.243, "step": 10970 }, { "epoch": 0.86, "learning_rate": 9.139731827805222e-05, "loss": 3.1516, "step": 10971 }, { "epoch": 0.86, "learning_rate": 9.139653414882773e-05, "loss": 3.2096, "step": 10972 }, { "epoch": 0.86, "learning_rate": 9.139575001960323e-05, "loss": 3.2005, "step": 10973 }, { "epoch": 0.86, "learning_rate": 9.139496589037873e-05, "loss": 3.1223, "step": 10974 }, { "epoch": 0.86, "learning_rate": 9.139418176115424e-05, "loss": 2.7783, "step": 10975 }, { "epoch": 0.86, "learning_rate": 9.139339763192974e-05, "loss": 3.0175, "step": 10976 }, { "epoch": 0.86, "learning_rate": 9.139261350270524e-05, "loss": 2.9199, "step": 10977 }, { "epoch": 0.86, "learning_rate": 9.139182937348075e-05, "loss": 2.9376, "step": 10978 }, { "epoch": 0.86, "learning_rate": 9.139104524425625e-05, "loss": 3.1217, "step": 10979 }, { "epoch": 0.86, "learning_rate": 9.139026111503175e-05, "loss": 2.9466, "step": 10980 }, { "epoch": 0.86, "learning_rate": 9.138947698580727e-05, "loss": 2.837, "step": 10981 }, { "epoch": 0.86, "learning_rate": 9.138869285658278e-05, "loss": 2.9237, "step": 10982 }, { "epoch": 0.86, "learning_rate": 9.138790872735828e-05, "loss": 2.9445, "step": 10983 }, { "epoch": 0.86, "learning_rate": 9.138712459813378e-05, "loss": 3.2071, "step": 10984 }, { "epoch": 0.86, "learning_rate": 9.138634046890929e-05, "loss": 2.9632, "step": 10985 }, { "epoch": 0.86, "learning_rate": 9.138555633968479e-05, "loss": 3.0686, "step": 10986 }, { "epoch": 0.86, "learning_rate": 9.13847722104603e-05, "loss": 2.916, "step": 10987 }, { "epoch": 0.86, "learning_rate": 9.13839880812358e-05, "loss": 3.1226, "step": 10988 }, { "epoch": 0.86, "learning_rate": 9.13832039520113e-05, "loss": 3.0342, "step": 10989 }, { "epoch": 0.86, "learning_rate": 9.13824198227868e-05, "loss": 2.9565, "step": 10990 }, { "epoch": 0.86, "learning_rate": 9.138163569356231e-05, "loss": 2.9515, "step": 10991 }, { "epoch": 0.86, "learning_rate": 9.138085156433781e-05, "loss": 3.0402, "step": 10992 }, { "epoch": 0.86, "learning_rate": 9.138006743511332e-05, "loss": 2.8551, "step": 10993 }, { "epoch": 0.86, "learning_rate": 9.137928330588882e-05, "loss": 3.0434, "step": 10994 }, { "epoch": 0.86, "learning_rate": 9.137849917666432e-05, "loss": 3.3747, "step": 10995 }, { "epoch": 0.86, "learning_rate": 9.137771504743983e-05, "loss": 2.8665, "step": 10996 }, { "epoch": 0.86, "learning_rate": 9.137693091821533e-05, "loss": 3.1343, "step": 10997 }, { "epoch": 0.86, "learning_rate": 9.137614678899083e-05, "loss": 2.7559, "step": 10998 }, { "epoch": 0.86, "learning_rate": 9.137536265976634e-05, "loss": 3.0023, "step": 10999 }, { "epoch": 0.86, "learning_rate": 9.137457853054183e-05, "loss": 2.914, "step": 11000 }, { "epoch": 0.86, "eval_loss": 2.9334206581115723, "eval_runtime": 217.8335, "eval_samples_per_second": 52.021, "eval_steps_per_second": 3.255, "step": 11000 }, { "epoch": 0.86, "learning_rate": 9.137379440131733e-05, "loss": 2.9765, "step": 11001 }, { "epoch": 0.86, "learning_rate": 9.137301027209284e-05, "loss": 2.9764, "step": 11002 }, { "epoch": 0.86, "learning_rate": 9.137222614286834e-05, "loss": 2.8169, "step": 11003 }, { "epoch": 0.86, "learning_rate": 9.137144201364384e-05, "loss": 3.005, "step": 11004 }, { "epoch": 0.86, "learning_rate": 9.137065788441935e-05, "loss": 3.2614, "step": 11005 }, { "epoch": 0.86, "learning_rate": 9.136987375519485e-05, "loss": 3.0316, "step": 11006 }, { "epoch": 0.86, "learning_rate": 9.136908962597037e-05, "loss": 3.0381, "step": 11007 }, { "epoch": 0.86, "learning_rate": 9.136830549674587e-05, "loss": 2.8927, "step": 11008 }, { "epoch": 0.86, "learning_rate": 9.136752136752137e-05, "loss": 3.0974, "step": 11009 }, { "epoch": 0.86, "learning_rate": 9.136673723829688e-05, "loss": 2.9799, "step": 11010 }, { "epoch": 0.86, "learning_rate": 9.136595310907238e-05, "loss": 2.8089, "step": 11011 }, { "epoch": 0.86, "learning_rate": 9.136516897984789e-05, "loss": 2.953, "step": 11012 }, { "epoch": 0.86, "learning_rate": 9.136438485062339e-05, "loss": 2.7728, "step": 11013 }, { "epoch": 0.86, "learning_rate": 9.13636007213989e-05, "loss": 2.9276, "step": 11014 }, { "epoch": 0.86, "learning_rate": 9.13628165921744e-05, "loss": 3.1148, "step": 11015 }, { "epoch": 0.86, "learning_rate": 9.13620324629499e-05, "loss": 2.992, "step": 11016 }, { "epoch": 0.86, "learning_rate": 9.13612483337254e-05, "loss": 2.9571, "step": 11017 }, { "epoch": 0.86, "learning_rate": 9.136046420450091e-05, "loss": 2.9784, "step": 11018 }, { "epoch": 0.86, "learning_rate": 9.135968007527641e-05, "loss": 3.2977, "step": 11019 }, { "epoch": 0.86, "learning_rate": 9.135889594605192e-05, "loss": 3.0024, "step": 11020 }, { "epoch": 0.86, "learning_rate": 9.135811181682742e-05, "loss": 3.0807, "step": 11021 }, { "epoch": 0.86, "learning_rate": 9.135732768760292e-05, "loss": 3.1584, "step": 11022 }, { "epoch": 0.86, "learning_rate": 9.135654355837843e-05, "loss": 3.0756, "step": 11023 }, { "epoch": 0.86, "learning_rate": 9.135575942915393e-05, "loss": 3.1786, "step": 11024 }, { "epoch": 0.86, "learning_rate": 9.135497529992943e-05, "loss": 2.9628, "step": 11025 }, { "epoch": 0.86, "learning_rate": 9.135419117070494e-05, "loss": 3.108, "step": 11026 }, { "epoch": 0.86, "learning_rate": 9.135340704148044e-05, "loss": 3.116, "step": 11027 }, { "epoch": 0.86, "learning_rate": 9.135262291225594e-05, "loss": 2.8597, "step": 11028 }, { "epoch": 0.86, "learning_rate": 9.135183878303145e-05, "loss": 2.8208, "step": 11029 }, { "epoch": 0.86, "learning_rate": 9.135105465380695e-05, "loss": 3.2549, "step": 11030 }, { "epoch": 0.86, "learning_rate": 9.135027052458246e-05, "loss": 3.0465, "step": 11031 }, { "epoch": 0.87, "learning_rate": 9.134948639535796e-05, "loss": 2.9526, "step": 11032 }, { "epoch": 0.87, "learning_rate": 9.134870226613346e-05, "loss": 2.9929, "step": 11033 }, { "epoch": 0.87, "learning_rate": 9.134791813690897e-05, "loss": 2.9919, "step": 11034 }, { "epoch": 0.87, "learning_rate": 9.134713400768447e-05, "loss": 2.9068, "step": 11035 }, { "epoch": 0.87, "learning_rate": 9.134634987845997e-05, "loss": 3.0222, "step": 11036 }, { "epoch": 0.87, "learning_rate": 9.134556574923548e-05, "loss": 3.2324, "step": 11037 }, { "epoch": 0.87, "learning_rate": 9.134478162001098e-05, "loss": 2.9085, "step": 11038 }, { "epoch": 0.87, "learning_rate": 9.134399749078649e-05, "loss": 3.0615, "step": 11039 }, { "epoch": 0.87, "learning_rate": 9.134321336156199e-05, "loss": 2.8964, "step": 11040 }, { "epoch": 0.87, "learning_rate": 9.134242923233749e-05, "loss": 2.7746, "step": 11041 }, { "epoch": 0.87, "learning_rate": 9.1341645103113e-05, "loss": 2.9194, "step": 11042 }, { "epoch": 0.87, "learning_rate": 9.13408609738885e-05, "loss": 2.7101, "step": 11043 }, { "epoch": 0.87, "learning_rate": 9.1340076844664e-05, "loss": 3.0867, "step": 11044 }, { "epoch": 0.87, "learning_rate": 9.133929271543951e-05, "loss": 3.1576, "step": 11045 }, { "epoch": 0.87, "learning_rate": 9.133850858621501e-05, "loss": 2.9752, "step": 11046 }, { "epoch": 0.87, "learning_rate": 9.133772445699051e-05, "loss": 3.1455, "step": 11047 }, { "epoch": 0.87, "learning_rate": 9.133694032776602e-05, "loss": 2.9987, "step": 11048 }, { "epoch": 0.87, "learning_rate": 9.133615619854152e-05, "loss": 3.1361, "step": 11049 }, { "epoch": 0.87, "learning_rate": 9.133537206931703e-05, "loss": 3.4176, "step": 11050 }, { "epoch": 0.87, "learning_rate": 9.133458794009253e-05, "loss": 3.1467, "step": 11051 }, { "epoch": 0.87, "learning_rate": 9.133380381086803e-05, "loss": 3.307, "step": 11052 }, { "epoch": 0.87, "learning_rate": 9.133301968164354e-05, "loss": 2.873, "step": 11053 }, { "epoch": 0.87, "learning_rate": 9.133223555241904e-05, "loss": 3.1649, "step": 11054 }, { "epoch": 0.87, "learning_rate": 9.133145142319454e-05, "loss": 2.8634, "step": 11055 }, { "epoch": 0.87, "learning_rate": 9.133066729397005e-05, "loss": 3.0445, "step": 11056 }, { "epoch": 0.87, "learning_rate": 9.132988316474555e-05, "loss": 2.8934, "step": 11057 }, { "epoch": 0.87, "learning_rate": 9.132909903552106e-05, "loss": 3.0715, "step": 11058 }, { "epoch": 0.87, "learning_rate": 9.132831490629656e-05, "loss": 3.0575, "step": 11059 }, { "epoch": 0.87, "learning_rate": 9.132753077707206e-05, "loss": 2.8199, "step": 11060 }, { "epoch": 0.87, "learning_rate": 9.132674664784757e-05, "loss": 3.2728, "step": 11061 }, { "epoch": 0.87, "learning_rate": 9.132596251862307e-05, "loss": 3.2452, "step": 11062 }, { "epoch": 0.87, "learning_rate": 9.132517838939857e-05, "loss": 2.8482, "step": 11063 }, { "epoch": 0.87, "learning_rate": 9.132439426017409e-05, "loss": 2.9422, "step": 11064 }, { "epoch": 0.87, "learning_rate": 9.13236101309496e-05, "loss": 2.9379, "step": 11065 }, { "epoch": 0.87, "learning_rate": 9.13228260017251e-05, "loss": 2.9938, "step": 11066 }, { "epoch": 0.87, "learning_rate": 9.13220418725006e-05, "loss": 2.9097, "step": 11067 }, { "epoch": 0.87, "learning_rate": 9.13212577432761e-05, "loss": 2.8387, "step": 11068 }, { "epoch": 0.87, "learning_rate": 9.132047361405161e-05, "loss": 2.9302, "step": 11069 }, { "epoch": 0.87, "learning_rate": 9.131968948482711e-05, "loss": 3.1826, "step": 11070 }, { "epoch": 0.87, "learning_rate": 9.131890535560262e-05, "loss": 2.9334, "step": 11071 }, { "epoch": 0.87, "learning_rate": 9.13181212263781e-05, "loss": 3.0962, "step": 11072 }, { "epoch": 0.87, "learning_rate": 9.131733709715361e-05, "loss": 3.0736, "step": 11073 }, { "epoch": 0.87, "learning_rate": 9.131655296792911e-05, "loss": 2.967, "step": 11074 }, { "epoch": 0.87, "learning_rate": 9.131576883870462e-05, "loss": 3.1454, "step": 11075 }, { "epoch": 0.87, "learning_rate": 9.131498470948012e-05, "loss": 2.9055, "step": 11076 }, { "epoch": 0.87, "learning_rate": 9.131420058025563e-05, "loss": 3.0647, "step": 11077 }, { "epoch": 0.87, "learning_rate": 9.131341645103113e-05, "loss": 2.9461, "step": 11078 }, { "epoch": 0.87, "learning_rate": 9.131263232180663e-05, "loss": 3.1997, "step": 11079 }, { "epoch": 0.87, "learning_rate": 9.131184819258214e-05, "loss": 3.1367, "step": 11080 }, { "epoch": 0.87, "learning_rate": 9.131106406335764e-05, "loss": 2.7676, "step": 11081 }, { "epoch": 0.87, "learning_rate": 9.131027993413314e-05, "loss": 3.0074, "step": 11082 }, { "epoch": 0.87, "learning_rate": 9.130949580490865e-05, "loss": 2.9891, "step": 11083 }, { "epoch": 0.87, "learning_rate": 9.130871167568415e-05, "loss": 3.019, "step": 11084 }, { "epoch": 0.87, "learning_rate": 9.130792754645965e-05, "loss": 2.7801, "step": 11085 }, { "epoch": 0.87, "learning_rate": 9.130714341723516e-05, "loss": 3.236, "step": 11086 }, { "epoch": 0.87, "learning_rate": 9.130635928801066e-05, "loss": 3.0269, "step": 11087 }, { "epoch": 0.87, "learning_rate": 9.130557515878617e-05, "loss": 3.096, "step": 11088 }, { "epoch": 0.87, "learning_rate": 9.130479102956168e-05, "loss": 3.1167, "step": 11089 }, { "epoch": 0.87, "learning_rate": 9.130400690033719e-05, "loss": 3.1417, "step": 11090 }, { "epoch": 0.87, "learning_rate": 9.130322277111269e-05, "loss": 3.057, "step": 11091 }, { "epoch": 0.87, "learning_rate": 9.13024386418882e-05, "loss": 3.1241, "step": 11092 }, { "epoch": 0.87, "learning_rate": 9.13016545126637e-05, "loss": 3.022, "step": 11093 }, { "epoch": 0.87, "learning_rate": 9.13008703834392e-05, "loss": 2.9132, "step": 11094 }, { "epoch": 0.87, "learning_rate": 9.13000862542147e-05, "loss": 3.1025, "step": 11095 }, { "epoch": 0.87, "learning_rate": 9.129930212499021e-05, "loss": 3.022, "step": 11096 }, { "epoch": 0.87, "learning_rate": 9.129851799576571e-05, "loss": 3.0745, "step": 11097 }, { "epoch": 0.87, "learning_rate": 9.129773386654122e-05, "loss": 3.3191, "step": 11098 }, { "epoch": 0.87, "learning_rate": 9.129694973731672e-05, "loss": 3.063, "step": 11099 }, { "epoch": 0.87, "learning_rate": 9.129616560809222e-05, "loss": 2.8544, "step": 11100 }, { "epoch": 0.87, "learning_rate": 9.129538147886773e-05, "loss": 2.8461, "step": 11101 }, { "epoch": 0.87, "learning_rate": 9.129459734964323e-05, "loss": 2.6955, "step": 11102 }, { "epoch": 0.87, "learning_rate": 9.129381322041873e-05, "loss": 2.8841, "step": 11103 }, { "epoch": 0.87, "learning_rate": 9.129302909119424e-05, "loss": 2.8768, "step": 11104 }, { "epoch": 0.87, "learning_rate": 9.129224496196974e-05, "loss": 3.3731, "step": 11105 }, { "epoch": 0.87, "learning_rate": 9.129146083274525e-05, "loss": 3.0159, "step": 11106 }, { "epoch": 0.87, "learning_rate": 9.129067670352075e-05, "loss": 3.0726, "step": 11107 }, { "epoch": 0.87, "learning_rate": 9.128989257429624e-05, "loss": 2.83, "step": 11108 }, { "epoch": 0.87, "learning_rate": 9.128910844507174e-05, "loss": 3.1057, "step": 11109 }, { "epoch": 0.87, "learning_rate": 9.128832431584725e-05, "loss": 3.0065, "step": 11110 }, { "epoch": 0.87, "learning_rate": 9.128754018662275e-05, "loss": 2.7185, "step": 11111 }, { "epoch": 0.87, "learning_rate": 9.128675605739825e-05, "loss": 3.0018, "step": 11112 }, { "epoch": 0.87, "learning_rate": 9.128597192817376e-05, "loss": 2.9708, "step": 11113 }, { "epoch": 0.87, "learning_rate": 9.128518779894926e-05, "loss": 2.8789, "step": 11114 }, { "epoch": 0.87, "learning_rate": 9.128440366972478e-05, "loss": 3.0831, "step": 11115 }, { "epoch": 0.87, "learning_rate": 9.128361954050028e-05, "loss": 3.1925, "step": 11116 }, { "epoch": 0.87, "learning_rate": 9.128283541127579e-05, "loss": 2.9224, "step": 11117 }, { "epoch": 0.87, "learning_rate": 9.128205128205129e-05, "loss": 3.2129, "step": 11118 }, { "epoch": 0.87, "learning_rate": 9.128126715282679e-05, "loss": 3.2503, "step": 11119 }, { "epoch": 0.87, "learning_rate": 9.12804830236023e-05, "loss": 3.1032, "step": 11120 }, { "epoch": 0.87, "learning_rate": 9.12796988943778e-05, "loss": 2.9108, "step": 11121 }, { "epoch": 0.87, "learning_rate": 9.12789147651533e-05, "loss": 3.1504, "step": 11122 }, { "epoch": 0.87, "learning_rate": 9.127813063592881e-05, "loss": 3.1875, "step": 11123 }, { "epoch": 0.87, "learning_rate": 9.127734650670431e-05, "loss": 3.1114, "step": 11124 }, { "epoch": 0.87, "learning_rate": 9.127656237747982e-05, "loss": 2.9257, "step": 11125 }, { "epoch": 0.87, "learning_rate": 9.127577824825532e-05, "loss": 3.1397, "step": 11126 }, { "epoch": 0.87, "learning_rate": 9.127499411903082e-05, "loss": 2.8024, "step": 11127 }, { "epoch": 0.87, "learning_rate": 9.127420998980633e-05, "loss": 2.7193, "step": 11128 }, { "epoch": 0.87, "learning_rate": 9.127342586058183e-05, "loss": 2.971, "step": 11129 }, { "epoch": 0.87, "learning_rate": 9.127264173135733e-05, "loss": 3.1441, "step": 11130 }, { "epoch": 0.87, "learning_rate": 9.127185760213284e-05, "loss": 2.9594, "step": 11131 }, { "epoch": 0.87, "learning_rate": 9.127107347290834e-05, "loss": 3.1146, "step": 11132 }, { "epoch": 0.87, "learning_rate": 9.127028934368384e-05, "loss": 3.0822, "step": 11133 }, { "epoch": 0.87, "learning_rate": 9.126950521445935e-05, "loss": 3.0358, "step": 11134 }, { "epoch": 0.87, "learning_rate": 9.126872108523485e-05, "loss": 2.8833, "step": 11135 }, { "epoch": 0.87, "learning_rate": 9.126793695601036e-05, "loss": 2.8551, "step": 11136 }, { "epoch": 0.87, "learning_rate": 9.126715282678586e-05, "loss": 3.0011, "step": 11137 }, { "epoch": 0.87, "learning_rate": 9.126636869756136e-05, "loss": 2.8879, "step": 11138 }, { "epoch": 0.87, "learning_rate": 9.126558456833687e-05, "loss": 2.8926, "step": 11139 }, { "epoch": 0.87, "learning_rate": 9.126480043911237e-05, "loss": 3.3487, "step": 11140 }, { "epoch": 0.87, "learning_rate": 9.126401630988787e-05, "loss": 2.9432, "step": 11141 }, { "epoch": 0.87, "learning_rate": 9.126323218066338e-05, "loss": 2.9846, "step": 11142 }, { "epoch": 0.87, "learning_rate": 9.126244805143888e-05, "loss": 2.882, "step": 11143 }, { "epoch": 0.87, "learning_rate": 9.126166392221438e-05, "loss": 3.1021, "step": 11144 }, { "epoch": 0.87, "learning_rate": 9.126087979298989e-05, "loss": 3.0237, "step": 11145 }, { "epoch": 0.87, "learning_rate": 9.126009566376539e-05, "loss": 3.2651, "step": 11146 }, { "epoch": 0.87, "learning_rate": 9.12593115345409e-05, "loss": 3.0693, "step": 11147 }, { "epoch": 0.87, "learning_rate": 9.12585274053164e-05, "loss": 3.2215, "step": 11148 }, { "epoch": 0.87, "learning_rate": 9.12577432760919e-05, "loss": 2.9862, "step": 11149 }, { "epoch": 0.87, "learning_rate": 9.125695914686741e-05, "loss": 2.7804, "step": 11150 }, { "epoch": 0.87, "learning_rate": 9.125617501764291e-05, "loss": 3.0026, "step": 11151 }, { "epoch": 0.87, "learning_rate": 9.125539088841841e-05, "loss": 3.0546, "step": 11152 }, { "epoch": 0.87, "learning_rate": 9.125460675919392e-05, "loss": 3.0208, "step": 11153 }, { "epoch": 0.87, "learning_rate": 9.125382262996942e-05, "loss": 3.0399, "step": 11154 }, { "epoch": 0.87, "learning_rate": 9.125303850074493e-05, "loss": 2.8447, "step": 11155 }, { "epoch": 0.87, "learning_rate": 9.125225437152043e-05, "loss": 3.0074, "step": 11156 }, { "epoch": 0.87, "learning_rate": 9.125147024229593e-05, "loss": 3.0223, "step": 11157 }, { "epoch": 0.87, "learning_rate": 9.125068611307144e-05, "loss": 3.205, "step": 11158 }, { "epoch": 0.88, "learning_rate": 9.124990198384694e-05, "loss": 3.1319, "step": 11159 }, { "epoch": 0.88, "learning_rate": 9.124911785462244e-05, "loss": 2.8557, "step": 11160 }, { "epoch": 0.88, "learning_rate": 9.124833372539795e-05, "loss": 2.9401, "step": 11161 }, { "epoch": 0.88, "learning_rate": 9.124754959617345e-05, "loss": 3.0031, "step": 11162 }, { "epoch": 0.88, "learning_rate": 9.124676546694895e-05, "loss": 3.0561, "step": 11163 }, { "epoch": 0.88, "learning_rate": 9.124598133772446e-05, "loss": 3.5271, "step": 11164 }, { "epoch": 0.88, "learning_rate": 9.124519720849996e-05, "loss": 2.815, "step": 11165 }, { "epoch": 0.88, "learning_rate": 9.124441307927547e-05, "loss": 2.9285, "step": 11166 }, { "epoch": 0.88, "learning_rate": 9.124362895005097e-05, "loss": 2.9152, "step": 11167 }, { "epoch": 0.88, "learning_rate": 9.124284482082647e-05, "loss": 3.4633, "step": 11168 }, { "epoch": 0.88, "learning_rate": 9.124206069160198e-05, "loss": 2.9035, "step": 11169 }, { "epoch": 0.88, "learning_rate": 9.124127656237748e-05, "loss": 2.973, "step": 11170 }, { "epoch": 0.88, "learning_rate": 9.124049243315298e-05, "loss": 3.0941, "step": 11171 }, { "epoch": 0.88, "learning_rate": 9.12397083039285e-05, "loss": 2.8442, "step": 11172 }, { "epoch": 0.88, "learning_rate": 9.1238924174704e-05, "loss": 2.9762, "step": 11173 }, { "epoch": 0.88, "learning_rate": 9.123814004547951e-05, "loss": 2.9753, "step": 11174 }, { "epoch": 0.88, "learning_rate": 9.123735591625501e-05, "loss": 3.2538, "step": 11175 }, { "epoch": 0.88, "learning_rate": 9.123657178703052e-05, "loss": 3.1112, "step": 11176 }, { "epoch": 0.88, "learning_rate": 9.123578765780602e-05, "loss": 3.0341, "step": 11177 }, { "epoch": 0.88, "learning_rate": 9.123500352858152e-05, "loss": 2.972, "step": 11178 }, { "epoch": 0.88, "learning_rate": 9.123421939935703e-05, "loss": 2.9795, "step": 11179 }, { "epoch": 0.88, "learning_rate": 9.123343527013252e-05, "loss": 2.736, "step": 11180 }, { "epoch": 0.88, "learning_rate": 9.123265114090802e-05, "loss": 2.8385, "step": 11181 }, { "epoch": 0.88, "learning_rate": 9.123186701168352e-05, "loss": 3.0195, "step": 11182 }, { "epoch": 0.88, "learning_rate": 9.123108288245903e-05, "loss": 3.0096, "step": 11183 }, { "epoch": 0.88, "learning_rate": 9.123029875323453e-05, "loss": 2.995, "step": 11184 }, { "epoch": 0.88, "learning_rate": 9.122951462401004e-05, "loss": 2.851, "step": 11185 }, { "epoch": 0.88, "learning_rate": 9.122873049478554e-05, "loss": 3.0763, "step": 11186 }, { "epoch": 0.88, "learning_rate": 9.122794636556104e-05, "loss": 3.0624, "step": 11187 }, { "epoch": 0.88, "learning_rate": 9.122716223633655e-05, "loss": 3.1408, "step": 11188 }, { "epoch": 0.88, "learning_rate": 9.122637810711205e-05, "loss": 3.5164, "step": 11189 }, { "epoch": 0.88, "learning_rate": 9.122559397788755e-05, "loss": 3.2962, "step": 11190 }, { "epoch": 0.88, "learning_rate": 9.122480984866306e-05, "loss": 2.8204, "step": 11191 }, { "epoch": 0.88, "learning_rate": 9.122402571943856e-05, "loss": 3.3994, "step": 11192 }, { "epoch": 0.88, "learning_rate": 9.122324159021407e-05, "loss": 3.1363, "step": 11193 }, { "epoch": 0.88, "learning_rate": 9.122245746098957e-05, "loss": 3.0053, "step": 11194 }, { "epoch": 0.88, "learning_rate": 9.122167333176507e-05, "loss": 3.2135, "step": 11195 }, { "epoch": 0.88, "learning_rate": 9.122088920254058e-05, "loss": 3.133, "step": 11196 }, { "epoch": 0.88, "learning_rate": 9.12201050733161e-05, "loss": 3.0154, "step": 11197 }, { "epoch": 0.88, "learning_rate": 9.12193209440916e-05, "loss": 2.9507, "step": 11198 }, { "epoch": 0.88, "learning_rate": 9.12185368148671e-05, "loss": 2.9112, "step": 11199 }, { "epoch": 0.88, "learning_rate": 9.12177526856426e-05, "loss": 3.1191, "step": 11200 }, { "epoch": 0.88, "learning_rate": 9.121696855641811e-05, "loss": 3.059, "step": 11201 }, { "epoch": 0.88, "learning_rate": 9.121618442719361e-05, "loss": 3.0678, "step": 11202 }, { "epoch": 0.88, "learning_rate": 9.121540029796912e-05, "loss": 3.0868, "step": 11203 }, { "epoch": 0.88, "learning_rate": 9.121461616874462e-05, "loss": 3.2776, "step": 11204 }, { "epoch": 0.88, "learning_rate": 9.121383203952012e-05, "loss": 3.2833, "step": 11205 }, { "epoch": 0.88, "learning_rate": 9.121304791029563e-05, "loss": 2.8909, "step": 11206 }, { "epoch": 0.88, "learning_rate": 9.121226378107113e-05, "loss": 3.1208, "step": 11207 }, { "epoch": 0.88, "learning_rate": 9.121147965184663e-05, "loss": 2.9968, "step": 11208 }, { "epoch": 0.88, "learning_rate": 9.121069552262214e-05, "loss": 2.9346, "step": 11209 }, { "epoch": 0.88, "learning_rate": 9.120991139339764e-05, "loss": 3.0013, "step": 11210 }, { "epoch": 0.88, "learning_rate": 9.120912726417314e-05, "loss": 3.0655, "step": 11211 }, { "epoch": 0.88, "learning_rate": 9.120834313494865e-05, "loss": 3.1387, "step": 11212 }, { "epoch": 0.88, "learning_rate": 9.120755900572415e-05, "loss": 3.272, "step": 11213 }, { "epoch": 0.88, "learning_rate": 9.120677487649966e-05, "loss": 2.9769, "step": 11214 }, { "epoch": 0.88, "learning_rate": 9.120599074727516e-05, "loss": 3.0494, "step": 11215 }, { "epoch": 0.88, "learning_rate": 9.120520661805065e-05, "loss": 3.243, "step": 11216 }, { "epoch": 0.88, "learning_rate": 9.120442248882615e-05, "loss": 2.9977, "step": 11217 }, { "epoch": 0.88, "learning_rate": 9.120363835960166e-05, "loss": 2.9116, "step": 11218 }, { "epoch": 0.88, "learning_rate": 9.120285423037716e-05, "loss": 3.1914, "step": 11219 }, { "epoch": 0.88, "learning_rate": 9.120207010115266e-05, "loss": 2.7406, "step": 11220 }, { "epoch": 0.88, "learning_rate": 9.120128597192817e-05, "loss": 2.9728, "step": 11221 }, { "epoch": 0.88, "learning_rate": 9.120050184270367e-05, "loss": 3.1296, "step": 11222 }, { "epoch": 0.88, "learning_rate": 9.119971771347919e-05, "loss": 3.0651, "step": 11223 }, { "epoch": 0.88, "learning_rate": 9.119893358425469e-05, "loss": 2.7101, "step": 11224 }, { "epoch": 0.88, "learning_rate": 9.11981494550302e-05, "loss": 3.0591, "step": 11225 }, { "epoch": 0.88, "learning_rate": 9.11973653258057e-05, "loss": 3.1111, "step": 11226 }, { "epoch": 0.88, "learning_rate": 9.11965811965812e-05, "loss": 2.9667, "step": 11227 }, { "epoch": 0.88, "learning_rate": 9.119579706735671e-05, "loss": 3.2127, "step": 11228 }, { "epoch": 0.88, "learning_rate": 9.119501293813221e-05, "loss": 3.1417, "step": 11229 }, { "epoch": 0.88, "learning_rate": 9.119422880890771e-05, "loss": 3.044, "step": 11230 }, { "epoch": 0.88, "learning_rate": 9.119344467968322e-05, "loss": 2.9126, "step": 11231 }, { "epoch": 0.88, "learning_rate": 9.119266055045872e-05, "loss": 3.0116, "step": 11232 }, { "epoch": 0.88, "learning_rate": 9.119187642123423e-05, "loss": 2.6922, "step": 11233 }, { "epoch": 0.88, "learning_rate": 9.119109229200973e-05, "loss": 2.9944, "step": 11234 }, { "epoch": 0.88, "learning_rate": 9.119030816278523e-05, "loss": 2.965, "step": 11235 }, { "epoch": 0.88, "learning_rate": 9.118952403356074e-05, "loss": 3.2909, "step": 11236 }, { "epoch": 0.88, "learning_rate": 9.118873990433624e-05, "loss": 3.005, "step": 11237 }, { "epoch": 0.88, "learning_rate": 9.118795577511174e-05, "loss": 3.0656, "step": 11238 }, { "epoch": 0.88, "learning_rate": 9.118717164588725e-05, "loss": 2.8388, "step": 11239 }, { "epoch": 0.88, "learning_rate": 9.118638751666275e-05, "loss": 2.8478, "step": 11240 }, { "epoch": 0.88, "learning_rate": 9.118560338743826e-05, "loss": 3.2784, "step": 11241 }, { "epoch": 0.88, "learning_rate": 9.118481925821376e-05, "loss": 3.2825, "step": 11242 }, { "epoch": 0.88, "learning_rate": 9.118403512898926e-05, "loss": 3.3066, "step": 11243 }, { "epoch": 0.88, "learning_rate": 9.118325099976477e-05, "loss": 2.741, "step": 11244 }, { "epoch": 0.88, "learning_rate": 9.118246687054027e-05, "loss": 3.157, "step": 11245 }, { "epoch": 0.88, "learning_rate": 9.118168274131577e-05, "loss": 2.9048, "step": 11246 }, { "epoch": 0.88, "learning_rate": 9.118089861209128e-05, "loss": 2.9506, "step": 11247 }, { "epoch": 0.88, "learning_rate": 9.118011448286678e-05, "loss": 2.9988, "step": 11248 }, { "epoch": 0.88, "learning_rate": 9.117933035364228e-05, "loss": 3.1536, "step": 11249 }, { "epoch": 0.88, "learning_rate": 9.117854622441779e-05, "loss": 2.8352, "step": 11250 }, { "epoch": 0.88, "learning_rate": 9.117776209519329e-05, "loss": 2.9131, "step": 11251 }, { "epoch": 0.88, "learning_rate": 9.11769779659688e-05, "loss": 2.8379, "step": 11252 }, { "epoch": 0.88, "learning_rate": 9.11761938367443e-05, "loss": 3.1266, "step": 11253 }, { "epoch": 0.88, "learning_rate": 9.11754097075198e-05, "loss": 2.8202, "step": 11254 }, { "epoch": 0.88, "learning_rate": 9.11746255782953e-05, "loss": 3.1162, "step": 11255 }, { "epoch": 0.88, "learning_rate": 9.117384144907081e-05, "loss": 2.8573, "step": 11256 }, { "epoch": 0.88, "learning_rate": 9.117305731984631e-05, "loss": 3.1589, "step": 11257 }, { "epoch": 0.88, "learning_rate": 9.117227319062182e-05, "loss": 2.9176, "step": 11258 }, { "epoch": 0.88, "learning_rate": 9.117148906139732e-05, "loss": 2.9246, "step": 11259 }, { "epoch": 0.88, "learning_rate": 9.117070493217283e-05, "loss": 2.8484, "step": 11260 }, { "epoch": 0.88, "learning_rate": 9.116992080294833e-05, "loss": 3.016, "step": 11261 }, { "epoch": 0.88, "learning_rate": 9.116913667372383e-05, "loss": 3.0757, "step": 11262 }, { "epoch": 0.88, "learning_rate": 9.116835254449934e-05, "loss": 3.2734, "step": 11263 }, { "epoch": 0.88, "learning_rate": 9.116756841527484e-05, "loss": 3.0159, "step": 11264 }, { "epoch": 0.88, "learning_rate": 9.116678428605034e-05, "loss": 3.0651, "step": 11265 }, { "epoch": 0.88, "learning_rate": 9.116600015682585e-05, "loss": 3.1475, "step": 11266 }, { "epoch": 0.88, "learning_rate": 9.116521602760135e-05, "loss": 2.8233, "step": 11267 }, { "epoch": 0.88, "learning_rate": 9.116443189837685e-05, "loss": 2.9257, "step": 11268 }, { "epoch": 0.88, "learning_rate": 9.116364776915236e-05, "loss": 3.1128, "step": 11269 }, { "epoch": 0.88, "learning_rate": 9.116286363992786e-05, "loss": 3.157, "step": 11270 }, { "epoch": 0.88, "learning_rate": 9.116207951070337e-05, "loss": 2.9765, "step": 11271 }, { "epoch": 0.88, "learning_rate": 9.116129538147887e-05, "loss": 2.8233, "step": 11272 }, { "epoch": 0.88, "learning_rate": 9.116051125225437e-05, "loss": 2.9551, "step": 11273 }, { "epoch": 0.88, "learning_rate": 9.115972712302988e-05, "loss": 3.181, "step": 11274 }, { "epoch": 0.88, "learning_rate": 9.115894299380538e-05, "loss": 3.11, "step": 11275 }, { "epoch": 0.88, "learning_rate": 9.115815886458088e-05, "loss": 2.879, "step": 11276 }, { "epoch": 0.88, "learning_rate": 9.115737473535639e-05, "loss": 2.9884, "step": 11277 }, { "epoch": 0.88, "learning_rate": 9.115659060613189e-05, "loss": 3.1675, "step": 11278 }, { "epoch": 0.88, "learning_rate": 9.11558064769074e-05, "loss": 2.832, "step": 11279 }, { "epoch": 0.88, "learning_rate": 9.115502234768291e-05, "loss": 2.847, "step": 11280 }, { "epoch": 0.88, "learning_rate": 9.115423821845842e-05, "loss": 2.987, "step": 11281 }, { "epoch": 0.88, "learning_rate": 9.115345408923392e-05, "loss": 2.9343, "step": 11282 }, { "epoch": 0.88, "learning_rate": 9.115266996000942e-05, "loss": 3.1721, "step": 11283 }, { "epoch": 0.88, "learning_rate": 9.115188583078493e-05, "loss": 3.1805, "step": 11284 }, { "epoch": 0.88, "learning_rate": 9.115110170156043e-05, "loss": 3.1268, "step": 11285 }, { "epoch": 0.88, "learning_rate": 9.115031757233593e-05, "loss": 2.8019, "step": 11286 }, { "epoch": 0.89, "learning_rate": 9.114953344311144e-05, "loss": 3.2777, "step": 11287 }, { "epoch": 0.89, "learning_rate": 9.114874931388693e-05, "loss": 2.8464, "step": 11288 }, { "epoch": 0.89, "learning_rate": 9.114796518466243e-05, "loss": 3.0863, "step": 11289 }, { "epoch": 0.89, "learning_rate": 9.114718105543794e-05, "loss": 3.3717, "step": 11290 }, { "epoch": 0.89, "learning_rate": 9.114639692621344e-05, "loss": 2.548, "step": 11291 }, { "epoch": 0.89, "learning_rate": 9.114561279698894e-05, "loss": 3.1671, "step": 11292 }, { "epoch": 0.89, "learning_rate": 9.114482866776445e-05, "loss": 2.8984, "step": 11293 }, { "epoch": 0.89, "learning_rate": 9.114404453853995e-05, "loss": 2.9353, "step": 11294 }, { "epoch": 0.89, "learning_rate": 9.114326040931545e-05, "loss": 3.0902, "step": 11295 }, { "epoch": 0.89, "learning_rate": 9.114247628009096e-05, "loss": 3.0829, "step": 11296 }, { "epoch": 0.89, "learning_rate": 9.114169215086646e-05, "loss": 2.8286, "step": 11297 }, { "epoch": 0.89, "learning_rate": 9.114090802164196e-05, "loss": 3.0056, "step": 11298 }, { "epoch": 0.89, "learning_rate": 9.114012389241747e-05, "loss": 2.9917, "step": 11299 }, { "epoch": 0.89, "learning_rate": 9.113933976319297e-05, "loss": 2.92, "step": 11300 }, { "epoch": 0.89, "learning_rate": 9.113855563396848e-05, "loss": 3.1332, "step": 11301 }, { "epoch": 0.89, "learning_rate": 9.113777150474398e-05, "loss": 3.0004, "step": 11302 }, { "epoch": 0.89, "learning_rate": 9.113698737551948e-05, "loss": 2.9876, "step": 11303 }, { "epoch": 0.89, "learning_rate": 9.113620324629499e-05, "loss": 2.8592, "step": 11304 }, { "epoch": 0.89, "learning_rate": 9.113541911707049e-05, "loss": 3.1706, "step": 11305 }, { "epoch": 0.89, "learning_rate": 9.113463498784601e-05, "loss": 2.9247, "step": 11306 }, { "epoch": 0.89, "learning_rate": 9.113385085862151e-05, "loss": 3.0155, "step": 11307 }, { "epoch": 0.89, "learning_rate": 9.113306672939702e-05, "loss": 3.0041, "step": 11308 }, { "epoch": 0.89, "learning_rate": 9.113228260017252e-05, "loss": 3.136, "step": 11309 }, { "epoch": 0.89, "learning_rate": 9.113149847094802e-05, "loss": 2.9332, "step": 11310 }, { "epoch": 0.89, "learning_rate": 9.113071434172353e-05, "loss": 3.1748, "step": 11311 }, { "epoch": 0.89, "learning_rate": 9.112993021249903e-05, "loss": 3.1197, "step": 11312 }, { "epoch": 0.89, "learning_rate": 9.112914608327453e-05, "loss": 2.9786, "step": 11313 }, { "epoch": 0.89, "learning_rate": 9.112836195405004e-05, "loss": 2.906, "step": 11314 }, { "epoch": 0.89, "learning_rate": 9.112757782482554e-05, "loss": 2.9972, "step": 11315 }, { "epoch": 0.89, "learning_rate": 9.112679369560104e-05, "loss": 2.9456, "step": 11316 }, { "epoch": 0.89, "learning_rate": 9.112600956637655e-05, "loss": 2.8931, "step": 11317 }, { "epoch": 0.89, "learning_rate": 9.112522543715205e-05, "loss": 2.9307, "step": 11318 }, { "epoch": 0.89, "learning_rate": 9.112444130792756e-05, "loss": 3.0371, "step": 11319 }, { "epoch": 0.89, "learning_rate": 9.112365717870306e-05, "loss": 2.9744, "step": 11320 }, { "epoch": 0.89, "learning_rate": 9.112287304947856e-05, "loss": 3.1336, "step": 11321 }, { "epoch": 0.89, "learning_rate": 9.112208892025407e-05, "loss": 3.0339, "step": 11322 }, { "epoch": 0.89, "learning_rate": 9.112130479102957e-05, "loss": 3.0495, "step": 11323 }, { "epoch": 0.89, "learning_rate": 9.112052066180506e-05, "loss": 2.9818, "step": 11324 }, { "epoch": 0.89, "learning_rate": 9.111973653258056e-05, "loss": 2.7748, "step": 11325 }, { "epoch": 0.89, "learning_rate": 9.111895240335607e-05, "loss": 2.8322, "step": 11326 }, { "epoch": 0.89, "learning_rate": 9.111816827413157e-05, "loss": 3.0828, "step": 11327 }, { "epoch": 0.89, "learning_rate": 9.111738414490708e-05, "loss": 3.0234, "step": 11328 }, { "epoch": 0.89, "learning_rate": 9.111660001568258e-05, "loss": 2.9701, "step": 11329 }, { "epoch": 0.89, "learning_rate": 9.111581588645808e-05, "loss": 2.7657, "step": 11330 }, { "epoch": 0.89, "learning_rate": 9.11150317572336e-05, "loss": 3.1759, "step": 11331 }, { "epoch": 0.89, "learning_rate": 9.11142476280091e-05, "loss": 3.1438, "step": 11332 }, { "epoch": 0.89, "learning_rate": 9.111346349878461e-05, "loss": 2.9931, "step": 11333 }, { "epoch": 0.89, "learning_rate": 9.111267936956011e-05, "loss": 2.9969, "step": 11334 }, { "epoch": 0.89, "learning_rate": 9.111189524033561e-05, "loss": 3.1704, "step": 11335 }, { "epoch": 0.89, "learning_rate": 9.111111111111112e-05, "loss": 3.0301, "step": 11336 }, { "epoch": 0.89, "learning_rate": 9.111032698188662e-05, "loss": 3.0278, "step": 11337 }, { "epoch": 0.89, "learning_rate": 9.110954285266213e-05, "loss": 3.0204, "step": 11338 }, { "epoch": 0.89, "learning_rate": 9.110875872343763e-05, "loss": 2.9175, "step": 11339 }, { "epoch": 0.89, "learning_rate": 9.110797459421313e-05, "loss": 3.0884, "step": 11340 }, { "epoch": 0.89, "learning_rate": 9.110719046498864e-05, "loss": 3.0013, "step": 11341 }, { "epoch": 0.89, "learning_rate": 9.110640633576414e-05, "loss": 2.8713, "step": 11342 }, { "epoch": 0.89, "learning_rate": 9.110562220653964e-05, "loss": 3.0211, "step": 11343 }, { "epoch": 0.89, "learning_rate": 9.110483807731515e-05, "loss": 3.0488, "step": 11344 }, { "epoch": 0.89, "learning_rate": 9.110405394809065e-05, "loss": 2.9047, "step": 11345 }, { "epoch": 0.89, "learning_rate": 9.110326981886615e-05, "loss": 3.159, "step": 11346 }, { "epoch": 0.89, "learning_rate": 9.110248568964166e-05, "loss": 3.1712, "step": 11347 }, { "epoch": 0.89, "learning_rate": 9.110170156041716e-05, "loss": 3.3071, "step": 11348 }, { "epoch": 0.89, "learning_rate": 9.110091743119267e-05, "loss": 2.9063, "step": 11349 }, { "epoch": 0.89, "learning_rate": 9.110013330196817e-05, "loss": 3.2787, "step": 11350 }, { "epoch": 0.89, "learning_rate": 9.109934917274367e-05, "loss": 3.0973, "step": 11351 }, { "epoch": 0.89, "learning_rate": 9.109856504351918e-05, "loss": 2.9418, "step": 11352 }, { "epoch": 0.89, "learning_rate": 9.109778091429468e-05, "loss": 2.7869, "step": 11353 }, { "epoch": 0.89, "learning_rate": 9.109699678507018e-05, "loss": 3.2999, "step": 11354 }, { "epoch": 0.89, "learning_rate": 9.109621265584569e-05, "loss": 2.736, "step": 11355 }, { "epoch": 0.89, "learning_rate": 9.109542852662119e-05, "loss": 2.7555, "step": 11356 }, { "epoch": 0.89, "learning_rate": 9.10946443973967e-05, "loss": 2.9957, "step": 11357 }, { "epoch": 0.89, "learning_rate": 9.10938602681722e-05, "loss": 3.3251, "step": 11358 }, { "epoch": 0.89, "learning_rate": 9.10930761389477e-05, "loss": 2.7922, "step": 11359 }, { "epoch": 0.89, "learning_rate": 9.10922920097232e-05, "loss": 3.0818, "step": 11360 }, { "epoch": 0.89, "learning_rate": 9.109150788049871e-05, "loss": 3.0924, "step": 11361 }, { "epoch": 0.89, "learning_rate": 9.109072375127421e-05, "loss": 2.9642, "step": 11362 }, { "epoch": 0.89, "learning_rate": 9.108993962204972e-05, "loss": 2.9447, "step": 11363 }, { "epoch": 0.89, "learning_rate": 9.108915549282522e-05, "loss": 3.0654, "step": 11364 }, { "epoch": 0.89, "learning_rate": 9.108837136360072e-05, "loss": 2.7214, "step": 11365 }, { "epoch": 0.89, "learning_rate": 9.108758723437623e-05, "loss": 3.0082, "step": 11366 }, { "epoch": 0.89, "learning_rate": 9.108680310515173e-05, "loss": 2.9676, "step": 11367 }, { "epoch": 0.89, "learning_rate": 9.108601897592724e-05, "loss": 3.2313, "step": 11368 }, { "epoch": 0.89, "learning_rate": 9.108523484670274e-05, "loss": 2.9799, "step": 11369 }, { "epoch": 0.89, "learning_rate": 9.108445071747824e-05, "loss": 2.9065, "step": 11370 }, { "epoch": 0.89, "learning_rate": 9.108366658825375e-05, "loss": 2.9935, "step": 11371 }, { "epoch": 0.89, "learning_rate": 9.108288245902925e-05, "loss": 3.0725, "step": 11372 }, { "epoch": 0.89, "learning_rate": 9.108209832980475e-05, "loss": 2.8932, "step": 11373 }, { "epoch": 0.89, "learning_rate": 9.108131420058026e-05, "loss": 2.9968, "step": 11374 }, { "epoch": 0.89, "learning_rate": 9.108053007135576e-05, "loss": 3.0874, "step": 11375 }, { "epoch": 0.89, "learning_rate": 9.107974594213127e-05, "loss": 3.0019, "step": 11376 }, { "epoch": 0.89, "learning_rate": 9.107896181290677e-05, "loss": 3.107, "step": 11377 }, { "epoch": 0.89, "learning_rate": 9.107817768368227e-05, "loss": 2.8891, "step": 11378 }, { "epoch": 0.89, "learning_rate": 9.107739355445778e-05, "loss": 2.8122, "step": 11379 }, { "epoch": 0.89, "learning_rate": 9.107660942523328e-05, "loss": 2.9919, "step": 11380 }, { "epoch": 0.89, "learning_rate": 9.107582529600878e-05, "loss": 2.8564, "step": 11381 }, { "epoch": 0.89, "learning_rate": 9.107504116678429e-05, "loss": 2.9529, "step": 11382 }, { "epoch": 0.89, "learning_rate": 9.107425703755979e-05, "loss": 3.0416, "step": 11383 }, { "epoch": 0.89, "learning_rate": 9.10734729083353e-05, "loss": 2.9876, "step": 11384 }, { "epoch": 0.89, "learning_rate": 9.10726887791108e-05, "loss": 3.1332, "step": 11385 }, { "epoch": 0.89, "learning_rate": 9.10719046498863e-05, "loss": 2.8101, "step": 11386 }, { "epoch": 0.89, "learning_rate": 9.10711205206618e-05, "loss": 3.1541, "step": 11387 }, { "epoch": 0.89, "learning_rate": 9.107033639143732e-05, "loss": 3.0544, "step": 11388 }, { "epoch": 0.89, "learning_rate": 9.106955226221283e-05, "loss": 2.9344, "step": 11389 }, { "epoch": 0.89, "learning_rate": 9.106876813298833e-05, "loss": 3.1696, "step": 11390 }, { "epoch": 0.89, "learning_rate": 9.106798400376383e-05, "loss": 2.9963, "step": 11391 }, { "epoch": 0.89, "learning_rate": 9.106719987453934e-05, "loss": 3.0443, "step": 11392 }, { "epoch": 0.89, "learning_rate": 9.106641574531484e-05, "loss": 3.0098, "step": 11393 }, { "epoch": 0.89, "learning_rate": 9.106563161609035e-05, "loss": 2.9975, "step": 11394 }, { "epoch": 0.89, "learning_rate": 9.106484748686585e-05, "loss": 3.19, "step": 11395 }, { "epoch": 0.89, "learning_rate": 9.106406335764134e-05, "loss": 2.9337, "step": 11396 }, { "epoch": 0.89, "learning_rate": 9.106327922841684e-05, "loss": 3.0986, "step": 11397 }, { "epoch": 0.89, "learning_rate": 9.106249509919235e-05, "loss": 3.0979, "step": 11398 }, { "epoch": 0.89, "learning_rate": 9.106171096996785e-05, "loss": 3.0331, "step": 11399 }, { "epoch": 0.89, "learning_rate": 9.106092684074335e-05, "loss": 3.0105, "step": 11400 }, { "epoch": 0.89, "learning_rate": 9.106014271151886e-05, "loss": 3.0611, "step": 11401 }, { "epoch": 0.89, "learning_rate": 9.105935858229436e-05, "loss": 3.011, "step": 11402 }, { "epoch": 0.89, "learning_rate": 9.105857445306986e-05, "loss": 3.0798, "step": 11403 }, { "epoch": 0.89, "learning_rate": 9.105779032384537e-05, "loss": 3.2759, "step": 11404 }, { "epoch": 0.89, "learning_rate": 9.105700619462087e-05, "loss": 2.7647, "step": 11405 }, { "epoch": 0.89, "learning_rate": 9.105622206539638e-05, "loss": 3.0826, "step": 11406 }, { "epoch": 0.89, "learning_rate": 9.105543793617188e-05, "loss": 3.1951, "step": 11407 }, { "epoch": 0.89, "learning_rate": 9.105465380694738e-05, "loss": 3.0237, "step": 11408 }, { "epoch": 0.89, "learning_rate": 9.105386967772289e-05, "loss": 2.904, "step": 11409 }, { "epoch": 0.89, "learning_rate": 9.105308554849839e-05, "loss": 2.9134, "step": 11410 }, { "epoch": 0.89, "learning_rate": 9.10523014192739e-05, "loss": 2.9041, "step": 11411 }, { "epoch": 0.89, "learning_rate": 9.10515172900494e-05, "loss": 2.855, "step": 11412 }, { "epoch": 0.89, "learning_rate": 9.10507331608249e-05, "loss": 3.2312, "step": 11413 }, { "epoch": 0.9, "learning_rate": 9.104994903160042e-05, "loss": 2.9858, "step": 11414 }, { "epoch": 0.9, "learning_rate": 9.104916490237592e-05, "loss": 3.0638, "step": 11415 }, { "epoch": 0.9, "learning_rate": 9.104838077315143e-05, "loss": 3.1367, "step": 11416 }, { "epoch": 0.9, "learning_rate": 9.104759664392693e-05, "loss": 3.0449, "step": 11417 }, { "epoch": 0.9, "learning_rate": 9.104681251470243e-05, "loss": 3.0544, "step": 11418 }, { "epoch": 0.9, "learning_rate": 9.104602838547794e-05, "loss": 3.163, "step": 11419 }, { "epoch": 0.9, "learning_rate": 9.104524425625344e-05, "loss": 3.1279, "step": 11420 }, { "epoch": 0.9, "learning_rate": 9.104446012702894e-05, "loss": 3.1411, "step": 11421 }, { "epoch": 0.9, "learning_rate": 9.104367599780445e-05, "loss": 2.9699, "step": 11422 }, { "epoch": 0.9, "learning_rate": 9.104289186857995e-05, "loss": 3.0235, "step": 11423 }, { "epoch": 0.9, "learning_rate": 9.104210773935546e-05, "loss": 2.8749, "step": 11424 }, { "epoch": 0.9, "learning_rate": 9.104132361013096e-05, "loss": 3.4554, "step": 11425 }, { "epoch": 0.9, "learning_rate": 9.104053948090646e-05, "loss": 2.9359, "step": 11426 }, { "epoch": 0.9, "learning_rate": 9.103975535168197e-05, "loss": 3.4578, "step": 11427 }, { "epoch": 0.9, "learning_rate": 9.103897122245747e-05, "loss": 2.9768, "step": 11428 }, { "epoch": 0.9, "learning_rate": 9.103818709323297e-05, "loss": 2.878, "step": 11429 }, { "epoch": 0.9, "learning_rate": 9.103740296400848e-05, "loss": 2.9478, "step": 11430 }, { "epoch": 0.9, "learning_rate": 9.103661883478398e-05, "loss": 3.301, "step": 11431 }, { "epoch": 0.9, "learning_rate": 9.103583470555947e-05, "loss": 3.0165, "step": 11432 }, { "epoch": 0.9, "learning_rate": 9.103505057633497e-05, "loss": 2.9179, "step": 11433 }, { "epoch": 0.9, "learning_rate": 9.103426644711048e-05, "loss": 3.0058, "step": 11434 }, { "epoch": 0.9, "learning_rate": 9.103348231788598e-05, "loss": 2.8951, "step": 11435 }, { "epoch": 0.9, "learning_rate": 9.103269818866149e-05, "loss": 2.8201, "step": 11436 }, { "epoch": 0.9, "learning_rate": 9.103191405943699e-05, "loss": 3.1346, "step": 11437 }, { "epoch": 0.9, "learning_rate": 9.10311299302125e-05, "loss": 3.1321, "step": 11438 }, { "epoch": 0.9, "learning_rate": 9.103034580098801e-05, "loss": 2.9079, "step": 11439 }, { "epoch": 0.9, "learning_rate": 9.102956167176351e-05, "loss": 3.0468, "step": 11440 }, { "epoch": 0.9, "learning_rate": 9.102877754253902e-05, "loss": 3.0001, "step": 11441 }, { "epoch": 0.9, "learning_rate": 9.102799341331452e-05, "loss": 2.7462, "step": 11442 }, { "epoch": 0.9, "learning_rate": 9.102720928409003e-05, "loss": 2.8883, "step": 11443 }, { "epoch": 0.9, "learning_rate": 9.102642515486553e-05, "loss": 2.8997, "step": 11444 }, { "epoch": 0.9, "learning_rate": 9.102564102564103e-05, "loss": 3.2281, "step": 11445 }, { "epoch": 0.9, "learning_rate": 9.102485689641654e-05, "loss": 3.0532, "step": 11446 }, { "epoch": 0.9, "learning_rate": 9.102407276719204e-05, "loss": 2.977, "step": 11447 }, { "epoch": 0.9, "learning_rate": 9.102328863796754e-05, "loss": 2.8814, "step": 11448 }, { "epoch": 0.9, "learning_rate": 9.102250450874305e-05, "loss": 3.0744, "step": 11449 }, { "epoch": 0.9, "learning_rate": 9.102172037951855e-05, "loss": 2.9586, "step": 11450 }, { "epoch": 0.9, "learning_rate": 9.102093625029405e-05, "loss": 2.9682, "step": 11451 }, { "epoch": 0.9, "learning_rate": 9.102015212106956e-05, "loss": 3.04, "step": 11452 }, { "epoch": 0.9, "learning_rate": 9.101936799184506e-05, "loss": 2.9547, "step": 11453 }, { "epoch": 0.9, "learning_rate": 9.101858386262057e-05, "loss": 2.9813, "step": 11454 }, { "epoch": 0.9, "learning_rate": 9.101779973339607e-05, "loss": 2.9665, "step": 11455 }, { "epoch": 0.9, "learning_rate": 9.101701560417157e-05, "loss": 3.3425, "step": 11456 }, { "epoch": 0.9, "learning_rate": 9.101623147494708e-05, "loss": 3.0772, "step": 11457 }, { "epoch": 0.9, "learning_rate": 9.101544734572258e-05, "loss": 3.0181, "step": 11458 }, { "epoch": 0.9, "learning_rate": 9.101466321649808e-05, "loss": 2.8129, "step": 11459 }, { "epoch": 0.9, "learning_rate": 9.101387908727359e-05, "loss": 3.1802, "step": 11460 }, { "epoch": 0.9, "learning_rate": 9.101309495804909e-05, "loss": 2.9273, "step": 11461 }, { "epoch": 0.9, "learning_rate": 9.10123108288246e-05, "loss": 3.0664, "step": 11462 }, { "epoch": 0.9, "learning_rate": 9.10115266996001e-05, "loss": 3.1691, "step": 11463 }, { "epoch": 0.9, "learning_rate": 9.10107425703756e-05, "loss": 2.9705, "step": 11464 }, { "epoch": 0.9, "learning_rate": 9.10099584411511e-05, "loss": 3.0595, "step": 11465 }, { "epoch": 0.9, "learning_rate": 9.100917431192661e-05, "loss": 2.9192, "step": 11466 }, { "epoch": 0.9, "learning_rate": 9.100839018270211e-05, "loss": 3.0341, "step": 11467 }, { "epoch": 0.9, "learning_rate": 9.100760605347762e-05, "loss": 3.1041, "step": 11468 }, { "epoch": 0.9, "learning_rate": 9.100682192425312e-05, "loss": 3.0887, "step": 11469 }, { "epoch": 0.9, "learning_rate": 9.100603779502862e-05, "loss": 2.9645, "step": 11470 }, { "epoch": 0.9, "learning_rate": 9.100525366580413e-05, "loss": 2.8383, "step": 11471 }, { "epoch": 0.9, "learning_rate": 9.100446953657963e-05, "loss": 2.959, "step": 11472 }, { "epoch": 0.9, "learning_rate": 9.100368540735514e-05, "loss": 3.1919, "step": 11473 }, { "epoch": 0.9, "learning_rate": 9.100290127813064e-05, "loss": 2.9535, "step": 11474 }, { "epoch": 0.9, "learning_rate": 9.100211714890614e-05, "loss": 2.9691, "step": 11475 }, { "epoch": 0.9, "learning_rate": 9.100133301968165e-05, "loss": 3.2833, "step": 11476 }, { "epoch": 0.9, "learning_rate": 9.100054889045715e-05, "loss": 3.1861, "step": 11477 }, { "epoch": 0.9, "learning_rate": 9.099976476123265e-05, "loss": 2.8862, "step": 11478 }, { "epoch": 0.9, "learning_rate": 9.099898063200816e-05, "loss": 3.1547, "step": 11479 }, { "epoch": 0.9, "learning_rate": 9.099819650278366e-05, "loss": 3.2771, "step": 11480 }, { "epoch": 0.9, "learning_rate": 9.099741237355917e-05, "loss": 3.0243, "step": 11481 }, { "epoch": 0.9, "learning_rate": 9.099662824433467e-05, "loss": 2.9078, "step": 11482 }, { "epoch": 0.9, "learning_rate": 9.099584411511017e-05, "loss": 2.9899, "step": 11483 }, { "epoch": 0.9, "learning_rate": 9.099505998588568e-05, "loss": 2.798, "step": 11484 }, { "epoch": 0.9, "learning_rate": 9.099427585666118e-05, "loss": 3.0454, "step": 11485 }, { "epoch": 0.9, "learning_rate": 9.099349172743668e-05, "loss": 3.1052, "step": 11486 }, { "epoch": 0.9, "learning_rate": 9.099270759821219e-05, "loss": 3.0135, "step": 11487 }, { "epoch": 0.9, "learning_rate": 9.099192346898769e-05, "loss": 2.9081, "step": 11488 }, { "epoch": 0.9, "learning_rate": 9.09911393397632e-05, "loss": 2.9976, "step": 11489 }, { "epoch": 0.9, "learning_rate": 9.09903552105387e-05, "loss": 3.0164, "step": 11490 }, { "epoch": 0.9, "learning_rate": 9.09895710813142e-05, "loss": 3.0723, "step": 11491 }, { "epoch": 0.9, "learning_rate": 9.09887869520897e-05, "loss": 3.0666, "step": 11492 }, { "epoch": 0.9, "learning_rate": 9.098800282286521e-05, "loss": 2.9967, "step": 11493 }, { "epoch": 0.9, "learning_rate": 9.098721869364071e-05, "loss": 3.0706, "step": 11494 }, { "epoch": 0.9, "learning_rate": 9.098643456441622e-05, "loss": 2.8864, "step": 11495 }, { "epoch": 0.9, "learning_rate": 9.098565043519173e-05, "loss": 3.0065, "step": 11496 }, { "epoch": 0.9, "learning_rate": 9.098486630596724e-05, "loss": 3.1399, "step": 11497 }, { "epoch": 0.9, "learning_rate": 9.098408217674274e-05, "loss": 3.005, "step": 11498 }, { "epoch": 0.9, "learning_rate": 9.098329804751824e-05, "loss": 3.2334, "step": 11499 }, { "epoch": 0.9, "learning_rate": 9.098251391829375e-05, "loss": 2.9739, "step": 11500 }, { "epoch": 0.9, "learning_rate": 9.098172978906925e-05, "loss": 2.869, "step": 11501 }, { "epoch": 0.9, "learning_rate": 9.098094565984476e-05, "loss": 3.0273, "step": 11502 }, { "epoch": 0.9, "learning_rate": 9.098016153062026e-05, "loss": 3.0547, "step": 11503 }, { "epoch": 0.9, "learning_rate": 9.097937740139575e-05, "loss": 2.9207, "step": 11504 }, { "epoch": 0.9, "learning_rate": 9.097859327217125e-05, "loss": 3.072, "step": 11505 }, { "epoch": 0.9, "learning_rate": 9.097780914294676e-05, "loss": 3.0588, "step": 11506 }, { "epoch": 0.9, "learning_rate": 9.097702501372226e-05, "loss": 3.0628, "step": 11507 }, { "epoch": 0.9, "learning_rate": 9.097624088449776e-05, "loss": 3.0394, "step": 11508 }, { "epoch": 0.9, "learning_rate": 9.097545675527327e-05, "loss": 3.0156, "step": 11509 }, { "epoch": 0.9, "learning_rate": 9.097467262604877e-05, "loss": 3.1781, "step": 11510 }, { "epoch": 0.9, "learning_rate": 9.097388849682428e-05, "loss": 3.3319, "step": 11511 }, { "epoch": 0.9, "learning_rate": 9.097310436759978e-05, "loss": 2.9856, "step": 11512 }, { "epoch": 0.9, "learning_rate": 9.097232023837528e-05, "loss": 2.9558, "step": 11513 }, { "epoch": 0.9, "learning_rate": 9.097153610915079e-05, "loss": 3.2124, "step": 11514 }, { "epoch": 0.9, "learning_rate": 9.097075197992629e-05, "loss": 3.0505, "step": 11515 }, { "epoch": 0.9, "learning_rate": 9.09699678507018e-05, "loss": 3.0438, "step": 11516 }, { "epoch": 0.9, "learning_rate": 9.09691837214773e-05, "loss": 2.9803, "step": 11517 }, { "epoch": 0.9, "learning_rate": 9.09683995922528e-05, "loss": 2.8424, "step": 11518 }, { "epoch": 0.9, "learning_rate": 9.09676154630283e-05, "loss": 2.7971, "step": 11519 }, { "epoch": 0.9, "learning_rate": 9.096683133380381e-05, "loss": 2.9565, "step": 11520 }, { "epoch": 0.9, "learning_rate": 9.096604720457931e-05, "loss": 3.0924, "step": 11521 }, { "epoch": 0.9, "learning_rate": 9.096526307535483e-05, "loss": 3.0892, "step": 11522 }, { "epoch": 0.9, "learning_rate": 9.096447894613033e-05, "loss": 2.9789, "step": 11523 }, { "epoch": 0.9, "learning_rate": 9.096369481690584e-05, "loss": 3.0877, "step": 11524 }, { "epoch": 0.9, "learning_rate": 9.096291068768134e-05, "loss": 2.8349, "step": 11525 }, { "epoch": 0.9, "learning_rate": 9.096212655845684e-05, "loss": 2.7764, "step": 11526 }, { "epoch": 0.9, "learning_rate": 9.096134242923235e-05, "loss": 3.1018, "step": 11527 }, { "epoch": 0.9, "learning_rate": 9.096055830000785e-05, "loss": 2.9886, "step": 11528 }, { "epoch": 0.9, "learning_rate": 9.095977417078336e-05, "loss": 3.1875, "step": 11529 }, { "epoch": 0.9, "learning_rate": 9.095899004155886e-05, "loss": 3.1159, "step": 11530 }, { "epoch": 0.9, "learning_rate": 9.095820591233436e-05, "loss": 2.9162, "step": 11531 }, { "epoch": 0.9, "learning_rate": 9.095742178310987e-05, "loss": 2.9362, "step": 11532 }, { "epoch": 0.9, "learning_rate": 9.095663765388537e-05, "loss": 2.8869, "step": 11533 }, { "epoch": 0.9, "learning_rate": 9.095585352466087e-05, "loss": 3.3182, "step": 11534 }, { "epoch": 0.9, "learning_rate": 9.095506939543638e-05, "loss": 3.189, "step": 11535 }, { "epoch": 0.9, "learning_rate": 9.095428526621188e-05, "loss": 3.4026, "step": 11536 }, { "epoch": 0.9, "learning_rate": 9.095350113698738e-05, "loss": 3.2325, "step": 11537 }, { "epoch": 0.9, "learning_rate": 9.095271700776289e-05, "loss": 3.0218, "step": 11538 }, { "epoch": 0.9, "learning_rate": 9.095193287853839e-05, "loss": 2.935, "step": 11539 }, { "epoch": 0.9, "learning_rate": 9.095114874931388e-05, "loss": 2.8279, "step": 11540 }, { "epoch": 0.9, "learning_rate": 9.095036462008939e-05, "loss": 2.8357, "step": 11541 }, { "epoch": 0.91, "learning_rate": 9.094958049086489e-05, "loss": 2.9369, "step": 11542 }, { "epoch": 0.91, "learning_rate": 9.094879636164039e-05, "loss": 3.219, "step": 11543 }, { "epoch": 0.91, "learning_rate": 9.09480122324159e-05, "loss": 2.9924, "step": 11544 }, { "epoch": 0.91, "learning_rate": 9.09472281031914e-05, "loss": 2.8439, "step": 11545 }, { "epoch": 0.91, "learning_rate": 9.09464439739669e-05, "loss": 2.9363, "step": 11546 }, { "epoch": 0.91, "learning_rate": 9.094565984474242e-05, "loss": 3.1555, "step": 11547 }, { "epoch": 0.91, "learning_rate": 9.094487571551792e-05, "loss": 2.843, "step": 11548 }, { "epoch": 0.91, "learning_rate": 9.094409158629343e-05, "loss": 2.9416, "step": 11549 }, { "epoch": 0.91, "learning_rate": 9.094330745706893e-05, "loss": 3.0053, "step": 11550 }, { "epoch": 0.91, "learning_rate": 9.094252332784444e-05, "loss": 3.1602, "step": 11551 }, { "epoch": 0.91, "learning_rate": 9.094173919861994e-05, "loss": 3.0081, "step": 11552 }, { "epoch": 0.91, "learning_rate": 9.094095506939544e-05, "loss": 2.9046, "step": 11553 }, { "epoch": 0.91, "learning_rate": 9.094017094017095e-05, "loss": 2.9886, "step": 11554 }, { "epoch": 0.91, "learning_rate": 9.093938681094645e-05, "loss": 3.0213, "step": 11555 }, { "epoch": 0.91, "learning_rate": 9.093860268172195e-05, "loss": 2.9573, "step": 11556 }, { "epoch": 0.91, "learning_rate": 9.093781855249746e-05, "loss": 2.76, "step": 11557 }, { "epoch": 0.91, "learning_rate": 9.093703442327296e-05, "loss": 3.1572, "step": 11558 }, { "epoch": 0.91, "learning_rate": 9.093625029404847e-05, "loss": 2.8257, "step": 11559 }, { "epoch": 0.91, "learning_rate": 9.093546616482397e-05, "loss": 2.8679, "step": 11560 }, { "epoch": 0.91, "learning_rate": 9.093468203559947e-05, "loss": 3.1965, "step": 11561 }, { "epoch": 0.91, "learning_rate": 9.093389790637498e-05, "loss": 3.2413, "step": 11562 }, { "epoch": 0.91, "learning_rate": 9.093311377715048e-05, "loss": 2.8781, "step": 11563 }, { "epoch": 0.91, "learning_rate": 9.093232964792598e-05, "loss": 3.0835, "step": 11564 }, { "epoch": 0.91, "learning_rate": 9.093154551870149e-05, "loss": 2.7065, "step": 11565 }, { "epoch": 0.91, "learning_rate": 9.093076138947699e-05, "loss": 2.897, "step": 11566 }, { "epoch": 0.91, "learning_rate": 9.09299772602525e-05, "loss": 2.8986, "step": 11567 }, { "epoch": 0.91, "learning_rate": 9.0929193131028e-05, "loss": 2.9281, "step": 11568 }, { "epoch": 0.91, "learning_rate": 9.09284090018035e-05, "loss": 2.9389, "step": 11569 }, { "epoch": 0.91, "learning_rate": 9.0927624872579e-05, "loss": 2.8761, "step": 11570 }, { "epoch": 0.91, "learning_rate": 9.092684074335451e-05, "loss": 3.0065, "step": 11571 }, { "epoch": 0.91, "learning_rate": 9.092605661413001e-05, "loss": 3.3254, "step": 11572 }, { "epoch": 0.91, "learning_rate": 9.092527248490552e-05, "loss": 3.2305, "step": 11573 }, { "epoch": 0.91, "learning_rate": 9.092448835568102e-05, "loss": 3.4225, "step": 11574 }, { "epoch": 0.91, "learning_rate": 9.092370422645652e-05, "loss": 3.0625, "step": 11575 }, { "epoch": 0.91, "learning_rate": 9.092292009723203e-05, "loss": 3.0229, "step": 11576 }, { "epoch": 0.91, "learning_rate": 9.092213596800753e-05, "loss": 3.0363, "step": 11577 }, { "epoch": 0.91, "learning_rate": 9.092135183878304e-05, "loss": 3.007, "step": 11578 }, { "epoch": 0.91, "learning_rate": 9.092056770955854e-05, "loss": 3.2671, "step": 11579 }, { "epoch": 0.91, "learning_rate": 9.091978358033404e-05, "loss": 2.9908, "step": 11580 }, { "epoch": 0.91, "learning_rate": 9.091899945110955e-05, "loss": 3.1101, "step": 11581 }, { "epoch": 0.91, "learning_rate": 9.091821532188505e-05, "loss": 3.0863, "step": 11582 }, { "epoch": 0.91, "learning_rate": 9.091743119266055e-05, "loss": 2.9833, "step": 11583 }, { "epoch": 0.91, "learning_rate": 9.091664706343606e-05, "loss": 3.4382, "step": 11584 }, { "epoch": 0.91, "learning_rate": 9.091586293421156e-05, "loss": 2.9966, "step": 11585 }, { "epoch": 0.91, "learning_rate": 9.091507880498706e-05, "loss": 3.1088, "step": 11586 }, { "epoch": 0.91, "learning_rate": 9.091429467576257e-05, "loss": 3.5798, "step": 11587 }, { "epoch": 0.91, "learning_rate": 9.091351054653807e-05, "loss": 2.8316, "step": 11588 }, { "epoch": 0.91, "learning_rate": 9.091272641731358e-05, "loss": 2.8286, "step": 11589 }, { "epoch": 0.91, "learning_rate": 9.091194228808908e-05, "loss": 2.8657, "step": 11590 }, { "epoch": 0.91, "learning_rate": 9.091115815886458e-05, "loss": 2.9501, "step": 11591 }, { "epoch": 0.91, "learning_rate": 9.091037402964009e-05, "loss": 2.9044, "step": 11592 }, { "epoch": 0.91, "learning_rate": 9.090958990041559e-05, "loss": 3.1972, "step": 11593 }, { "epoch": 0.91, "learning_rate": 9.09088057711911e-05, "loss": 3.2063, "step": 11594 }, { "epoch": 0.91, "learning_rate": 9.09080216419666e-05, "loss": 3.121, "step": 11595 }, { "epoch": 0.91, "learning_rate": 9.09072375127421e-05, "loss": 3.2122, "step": 11596 }, { "epoch": 0.91, "learning_rate": 9.09064533835176e-05, "loss": 3.0202, "step": 11597 }, { "epoch": 0.91, "learning_rate": 9.090566925429311e-05, "loss": 3.0684, "step": 11598 }, { "epoch": 0.91, "learning_rate": 9.090488512506861e-05, "loss": 2.7079, "step": 11599 }, { "epoch": 0.91, "learning_rate": 9.090410099584412e-05, "loss": 2.9397, "step": 11600 }, { "epoch": 0.91, "learning_rate": 9.090331686661962e-05, "loss": 3.1622, "step": 11601 }, { "epoch": 0.91, "learning_rate": 9.090253273739512e-05, "loss": 2.9183, "step": 11602 }, { "epoch": 0.91, "learning_rate": 9.090174860817063e-05, "loss": 3.0647, "step": 11603 }, { "epoch": 0.91, "learning_rate": 9.090096447894614e-05, "loss": 3.0124, "step": 11604 }, { "epoch": 0.91, "learning_rate": 9.090018034972165e-05, "loss": 2.7502, "step": 11605 }, { "epoch": 0.91, "learning_rate": 9.089939622049715e-05, "loss": 2.9622, "step": 11606 }, { "epoch": 0.91, "learning_rate": 9.089861209127266e-05, "loss": 3.1269, "step": 11607 }, { "epoch": 0.91, "learning_rate": 9.089782796204816e-05, "loss": 2.9525, "step": 11608 }, { "epoch": 0.91, "learning_rate": 9.089704383282366e-05, "loss": 3.0968, "step": 11609 }, { "epoch": 0.91, "learning_rate": 9.089625970359917e-05, "loss": 2.8146, "step": 11610 }, { "epoch": 0.91, "learning_rate": 9.089547557437467e-05, "loss": 2.9722, "step": 11611 }, { "epoch": 0.91, "learning_rate": 9.089469144515016e-05, "loss": 3.1192, "step": 11612 }, { "epoch": 0.91, "learning_rate": 9.089390731592566e-05, "loss": 3.0791, "step": 11613 }, { "epoch": 0.91, "learning_rate": 9.089312318670117e-05, "loss": 2.7919, "step": 11614 }, { "epoch": 0.91, "learning_rate": 9.089233905747667e-05, "loss": 2.9916, "step": 11615 }, { "epoch": 0.91, "learning_rate": 9.089155492825218e-05, "loss": 3.0747, "step": 11616 }, { "epoch": 0.91, "learning_rate": 9.089077079902768e-05, "loss": 3.1562, "step": 11617 }, { "epoch": 0.91, "learning_rate": 9.088998666980318e-05, "loss": 3.2389, "step": 11618 }, { "epoch": 0.91, "learning_rate": 9.088920254057869e-05, "loss": 3.0316, "step": 11619 }, { "epoch": 0.91, "learning_rate": 9.088841841135419e-05, "loss": 2.8481, "step": 11620 }, { "epoch": 0.91, "learning_rate": 9.08876342821297e-05, "loss": 3.087, "step": 11621 }, { "epoch": 0.91, "learning_rate": 9.08868501529052e-05, "loss": 3.0003, "step": 11622 }, { "epoch": 0.91, "learning_rate": 9.08860660236807e-05, "loss": 3.0542, "step": 11623 }, { "epoch": 0.91, "learning_rate": 9.08852818944562e-05, "loss": 3.142, "step": 11624 }, { "epoch": 0.91, "learning_rate": 9.088449776523171e-05, "loss": 3.1569, "step": 11625 }, { "epoch": 0.91, "learning_rate": 9.088371363600721e-05, "loss": 3.11, "step": 11626 }, { "epoch": 0.91, "learning_rate": 9.088292950678272e-05, "loss": 3.2237, "step": 11627 }, { "epoch": 0.91, "learning_rate": 9.088214537755822e-05, "loss": 2.9328, "step": 11628 }, { "epoch": 0.91, "learning_rate": 9.088136124833372e-05, "loss": 2.9166, "step": 11629 }, { "epoch": 0.91, "learning_rate": 9.088057711910924e-05, "loss": 3.0353, "step": 11630 }, { "epoch": 0.91, "learning_rate": 9.087979298988474e-05, "loss": 3.047, "step": 11631 }, { "epoch": 0.91, "learning_rate": 9.087900886066025e-05, "loss": 2.9254, "step": 11632 }, { "epoch": 0.91, "learning_rate": 9.087822473143575e-05, "loss": 2.9503, "step": 11633 }, { "epoch": 0.91, "learning_rate": 9.087744060221125e-05, "loss": 2.935, "step": 11634 }, { "epoch": 0.91, "learning_rate": 9.087665647298676e-05, "loss": 2.9314, "step": 11635 }, { "epoch": 0.91, "learning_rate": 9.087587234376226e-05, "loss": 3.2943, "step": 11636 }, { "epoch": 0.91, "learning_rate": 9.087508821453777e-05, "loss": 3.0224, "step": 11637 }, { "epoch": 0.91, "learning_rate": 9.087430408531327e-05, "loss": 3.0754, "step": 11638 }, { "epoch": 0.91, "learning_rate": 9.087351995608877e-05, "loss": 3.1303, "step": 11639 }, { "epoch": 0.91, "learning_rate": 9.087273582686428e-05, "loss": 3.1626, "step": 11640 }, { "epoch": 0.91, "learning_rate": 9.087195169763978e-05, "loss": 2.9836, "step": 11641 }, { "epoch": 0.91, "learning_rate": 9.087116756841528e-05, "loss": 3.0089, "step": 11642 }, { "epoch": 0.91, "learning_rate": 9.087038343919079e-05, "loss": 2.7919, "step": 11643 }, { "epoch": 0.91, "learning_rate": 9.086959930996629e-05, "loss": 2.9286, "step": 11644 }, { "epoch": 0.91, "learning_rate": 9.08688151807418e-05, "loss": 2.9511, "step": 11645 }, { "epoch": 0.91, "learning_rate": 9.08680310515173e-05, "loss": 2.8432, "step": 11646 }, { "epoch": 0.91, "learning_rate": 9.08672469222928e-05, "loss": 3.1468, "step": 11647 }, { "epoch": 0.91, "learning_rate": 9.086646279306829e-05, "loss": 3.0567, "step": 11648 }, { "epoch": 0.91, "learning_rate": 9.08656786638438e-05, "loss": 3.1158, "step": 11649 }, { "epoch": 0.91, "learning_rate": 9.08648945346193e-05, "loss": 2.9646, "step": 11650 }, { "epoch": 0.91, "learning_rate": 9.08641104053948e-05, "loss": 2.7464, "step": 11651 }, { "epoch": 0.91, "learning_rate": 9.086332627617031e-05, "loss": 3.0776, "step": 11652 }, { "epoch": 0.91, "learning_rate": 9.086254214694581e-05, "loss": 3.1161, "step": 11653 }, { "epoch": 0.91, "learning_rate": 9.086175801772131e-05, "loss": 3.106, "step": 11654 }, { "epoch": 0.91, "learning_rate": 9.086097388849683e-05, "loss": 3.0527, "step": 11655 }, { "epoch": 0.91, "learning_rate": 9.086018975927234e-05, "loss": 2.976, "step": 11656 }, { "epoch": 0.91, "learning_rate": 9.085940563004784e-05, "loss": 3.1309, "step": 11657 }, { "epoch": 0.91, "learning_rate": 9.085862150082334e-05, "loss": 3.1172, "step": 11658 }, { "epoch": 0.91, "learning_rate": 9.085783737159885e-05, "loss": 2.9968, "step": 11659 }, { "epoch": 0.91, "learning_rate": 9.085705324237435e-05, "loss": 3.0681, "step": 11660 }, { "epoch": 0.91, "learning_rate": 9.085626911314985e-05, "loss": 3.064, "step": 11661 }, { "epoch": 0.91, "learning_rate": 9.085548498392536e-05, "loss": 3.0801, "step": 11662 }, { "epoch": 0.91, "learning_rate": 9.085470085470086e-05, "loss": 3.1744, "step": 11663 }, { "epoch": 0.91, "learning_rate": 9.085391672547637e-05, "loss": 3.2623, "step": 11664 }, { "epoch": 0.91, "learning_rate": 9.085313259625187e-05, "loss": 3.2809, "step": 11665 }, { "epoch": 0.91, "learning_rate": 9.085234846702737e-05, "loss": 3.104, "step": 11666 }, { "epoch": 0.91, "learning_rate": 9.085156433780288e-05, "loss": 3.123, "step": 11667 }, { "epoch": 0.91, "learning_rate": 9.085078020857838e-05, "loss": 3.1807, "step": 11668 }, { "epoch": 0.92, "learning_rate": 9.084999607935388e-05, "loss": 2.8866, "step": 11669 }, { "epoch": 0.92, "learning_rate": 9.084921195012939e-05, "loss": 3.0879, "step": 11670 }, { "epoch": 0.92, "learning_rate": 9.084842782090489e-05, "loss": 3.1721, "step": 11671 }, { "epoch": 0.92, "learning_rate": 9.08476436916804e-05, "loss": 3.0887, "step": 11672 }, { "epoch": 0.92, "learning_rate": 9.08468595624559e-05, "loss": 2.9809, "step": 11673 }, { "epoch": 0.92, "learning_rate": 9.08460754332314e-05, "loss": 3.1285, "step": 11674 }, { "epoch": 0.92, "learning_rate": 9.08452913040069e-05, "loss": 3.2156, "step": 11675 }, { "epoch": 0.92, "learning_rate": 9.084450717478241e-05, "loss": 2.8451, "step": 11676 }, { "epoch": 0.92, "learning_rate": 9.084372304555791e-05, "loss": 3.0047, "step": 11677 }, { "epoch": 0.92, "learning_rate": 9.084293891633342e-05, "loss": 3.0515, "step": 11678 }, { "epoch": 0.92, "learning_rate": 9.084215478710892e-05, "loss": 3.4387, "step": 11679 }, { "epoch": 0.92, "learning_rate": 9.084137065788442e-05, "loss": 3.0617, "step": 11680 }, { "epoch": 0.92, "learning_rate": 9.084058652865993e-05, "loss": 3.0959, "step": 11681 }, { "epoch": 0.92, "learning_rate": 9.083980239943543e-05, "loss": 3.0818, "step": 11682 }, { "epoch": 0.92, "learning_rate": 9.083901827021094e-05, "loss": 2.9765, "step": 11683 }, { "epoch": 0.92, "learning_rate": 9.083823414098644e-05, "loss": 2.7396, "step": 11684 }, { "epoch": 0.92, "learning_rate": 9.083745001176194e-05, "loss": 3.9536, "step": 11685 }, { "epoch": 0.92, "learning_rate": 9.083666588253745e-05, "loss": 3.1541, "step": 11686 }, { "epoch": 0.92, "learning_rate": 9.083588175331295e-05, "loss": 2.7739, "step": 11687 }, { "epoch": 0.92, "learning_rate": 9.083509762408845e-05, "loss": 3.2206, "step": 11688 }, { "epoch": 0.92, "learning_rate": 9.083431349486396e-05, "loss": 3.0555, "step": 11689 }, { "epoch": 0.92, "learning_rate": 9.083352936563946e-05, "loss": 2.9185, "step": 11690 }, { "epoch": 0.92, "learning_rate": 9.083274523641496e-05, "loss": 2.9291, "step": 11691 }, { "epoch": 0.92, "learning_rate": 9.083196110719047e-05, "loss": 3.453, "step": 11692 }, { "epoch": 0.92, "learning_rate": 9.083117697796597e-05, "loss": 2.9532, "step": 11693 }, { "epoch": 0.92, "learning_rate": 9.083039284874148e-05, "loss": 3.1155, "step": 11694 }, { "epoch": 0.92, "learning_rate": 9.082960871951698e-05, "loss": 2.8829, "step": 11695 }, { "epoch": 0.92, "learning_rate": 9.082882459029248e-05, "loss": 2.9235, "step": 11696 }, { "epoch": 0.92, "learning_rate": 9.082804046106799e-05, "loss": 2.7892, "step": 11697 }, { "epoch": 0.92, "learning_rate": 9.082725633184349e-05, "loss": 2.9654, "step": 11698 }, { "epoch": 0.92, "learning_rate": 9.0826472202619e-05, "loss": 3.1291, "step": 11699 }, { "epoch": 0.92, "learning_rate": 9.08256880733945e-05, "loss": 2.8596, "step": 11700 }, { "epoch": 0.92, "learning_rate": 9.082490394417e-05, "loss": 3.0198, "step": 11701 }, { "epoch": 0.92, "learning_rate": 9.08241198149455e-05, "loss": 2.9149, "step": 11702 }, { "epoch": 0.92, "learning_rate": 9.082333568572101e-05, "loss": 2.7424, "step": 11703 }, { "epoch": 0.92, "learning_rate": 9.082255155649651e-05, "loss": 3.1464, "step": 11704 }, { "epoch": 0.92, "learning_rate": 9.082176742727202e-05, "loss": 3.1998, "step": 11705 }, { "epoch": 0.92, "learning_rate": 9.082098329804752e-05, "loss": 2.9868, "step": 11706 }, { "epoch": 0.92, "learning_rate": 9.082019916882302e-05, "loss": 3.0152, "step": 11707 }, { "epoch": 0.92, "learning_rate": 9.081941503959853e-05, "loss": 3.0207, "step": 11708 }, { "epoch": 0.92, "learning_rate": 9.081863091037403e-05, "loss": 2.9927, "step": 11709 }, { "epoch": 0.92, "learning_rate": 9.081784678114953e-05, "loss": 3.0767, "step": 11710 }, { "epoch": 0.92, "learning_rate": 9.081706265192504e-05, "loss": 3.0293, "step": 11711 }, { "epoch": 0.92, "learning_rate": 9.081627852270056e-05, "loss": 2.93, "step": 11712 }, { "epoch": 0.92, "learning_rate": 9.081549439347606e-05, "loss": 3.0356, "step": 11713 }, { "epoch": 0.92, "learning_rate": 9.081471026425156e-05, "loss": 3.0372, "step": 11714 }, { "epoch": 0.92, "learning_rate": 9.081392613502707e-05, "loss": 2.8969, "step": 11715 }, { "epoch": 0.92, "learning_rate": 9.081314200580257e-05, "loss": 3.0941, "step": 11716 }, { "epoch": 0.92, "learning_rate": 9.081235787657807e-05, "loss": 2.954, "step": 11717 }, { "epoch": 0.92, "learning_rate": 9.081157374735358e-05, "loss": 3.0585, "step": 11718 }, { "epoch": 0.92, "learning_rate": 9.081078961812908e-05, "loss": 2.7911, "step": 11719 }, { "epoch": 0.92, "learning_rate": 9.081000548890457e-05, "loss": 3.1851, "step": 11720 }, { "epoch": 0.92, "learning_rate": 9.080922135968007e-05, "loss": 2.9656, "step": 11721 }, { "epoch": 0.92, "learning_rate": 9.080843723045558e-05, "loss": 3.3064, "step": 11722 }, { "epoch": 0.92, "learning_rate": 9.080765310123108e-05, "loss": 3.5123, "step": 11723 }, { "epoch": 0.92, "learning_rate": 9.080686897200659e-05, "loss": 3.05, "step": 11724 }, { "epoch": 0.92, "learning_rate": 9.080608484278209e-05, "loss": 3.0143, "step": 11725 }, { "epoch": 0.92, "learning_rate": 9.080530071355759e-05, "loss": 2.9345, "step": 11726 }, { "epoch": 0.92, "learning_rate": 9.08045165843331e-05, "loss": 3.0901, "step": 11727 }, { "epoch": 0.92, "learning_rate": 9.08037324551086e-05, "loss": 3.23, "step": 11728 }, { "epoch": 0.92, "learning_rate": 9.08029483258841e-05, "loss": 2.9507, "step": 11729 }, { "epoch": 0.92, "learning_rate": 9.080216419665961e-05, "loss": 2.9561, "step": 11730 }, { "epoch": 0.92, "learning_rate": 9.080138006743511e-05, "loss": 2.9937, "step": 11731 }, { "epoch": 0.92, "learning_rate": 9.080059593821062e-05, "loss": 2.905, "step": 11732 }, { "epoch": 0.92, "learning_rate": 9.079981180898612e-05, "loss": 3.2736, "step": 11733 }, { "epoch": 0.92, "learning_rate": 9.079902767976162e-05, "loss": 3.1545, "step": 11734 }, { "epoch": 0.92, "learning_rate": 9.079824355053713e-05, "loss": 2.7685, "step": 11735 }, { "epoch": 0.92, "learning_rate": 9.079745942131263e-05, "loss": 2.994, "step": 11736 }, { "epoch": 0.92, "learning_rate": 9.079667529208813e-05, "loss": 2.9461, "step": 11737 }, { "epoch": 0.92, "learning_rate": 9.079589116286365e-05, "loss": 3.4236, "step": 11738 }, { "epoch": 0.92, "learning_rate": 9.079510703363915e-05, "loss": 2.976, "step": 11739 }, { "epoch": 0.92, "learning_rate": 9.079432290441466e-05, "loss": 2.9698, "step": 11740 }, { "epoch": 0.92, "learning_rate": 9.079353877519016e-05, "loss": 2.8264, "step": 11741 }, { "epoch": 0.92, "learning_rate": 9.079275464596567e-05, "loss": 3.0242, "step": 11742 }, { "epoch": 0.92, "learning_rate": 9.079197051674117e-05, "loss": 2.9974, "step": 11743 }, { "epoch": 0.92, "learning_rate": 9.079118638751667e-05, "loss": 2.8773, "step": 11744 }, { "epoch": 0.92, "learning_rate": 9.079040225829218e-05, "loss": 2.8389, "step": 11745 }, { "epoch": 0.92, "learning_rate": 9.078961812906768e-05, "loss": 3.4775, "step": 11746 }, { "epoch": 0.92, "learning_rate": 9.078883399984318e-05, "loss": 3.1476, "step": 11747 }, { "epoch": 0.92, "learning_rate": 9.078804987061869e-05, "loss": 2.9939, "step": 11748 }, { "epoch": 0.92, "learning_rate": 9.078726574139419e-05, "loss": 3.0149, "step": 11749 }, { "epoch": 0.92, "learning_rate": 9.07864816121697e-05, "loss": 2.9998, "step": 11750 }, { "epoch": 0.92, "learning_rate": 9.07856974829452e-05, "loss": 3.0213, "step": 11751 }, { "epoch": 0.92, "learning_rate": 9.07849133537207e-05, "loss": 3.1662, "step": 11752 }, { "epoch": 0.92, "learning_rate": 9.07841292244962e-05, "loss": 2.9604, "step": 11753 }, { "epoch": 0.92, "learning_rate": 9.078334509527171e-05, "loss": 3.1511, "step": 11754 }, { "epoch": 0.92, "learning_rate": 9.078256096604721e-05, "loss": 2.9961, "step": 11755 }, { "epoch": 0.92, "learning_rate": 9.07817768368227e-05, "loss": 3.1205, "step": 11756 }, { "epoch": 0.92, "learning_rate": 9.078099270759821e-05, "loss": 3.0858, "step": 11757 }, { "epoch": 0.92, "learning_rate": 9.078020857837371e-05, "loss": 2.767, "step": 11758 }, { "epoch": 0.92, "learning_rate": 9.077942444914921e-05, "loss": 2.8604, "step": 11759 }, { "epoch": 0.92, "learning_rate": 9.077864031992472e-05, "loss": 3.0101, "step": 11760 }, { "epoch": 0.92, "learning_rate": 9.077785619070022e-05, "loss": 2.8389, "step": 11761 }, { "epoch": 0.92, "learning_rate": 9.077707206147573e-05, "loss": 3.1496, "step": 11762 }, { "epoch": 0.92, "learning_rate": 9.077628793225124e-05, "loss": 2.977, "step": 11763 }, { "epoch": 0.92, "learning_rate": 9.077550380302675e-05, "loss": 3.3208, "step": 11764 }, { "epoch": 0.92, "learning_rate": 9.077471967380225e-05, "loss": 3.1629, "step": 11765 }, { "epoch": 0.92, "learning_rate": 9.077393554457775e-05, "loss": 2.7632, "step": 11766 }, { "epoch": 0.92, "learning_rate": 9.077315141535326e-05, "loss": 3.0094, "step": 11767 }, { "epoch": 0.92, "learning_rate": 9.077236728612876e-05, "loss": 2.9285, "step": 11768 }, { "epoch": 0.92, "learning_rate": 9.077158315690426e-05, "loss": 3.0849, "step": 11769 }, { "epoch": 0.92, "learning_rate": 9.077079902767977e-05, "loss": 3.2128, "step": 11770 }, { "epoch": 0.92, "learning_rate": 9.077001489845527e-05, "loss": 2.9506, "step": 11771 }, { "epoch": 0.92, "learning_rate": 9.076923076923078e-05, "loss": 3.0715, "step": 11772 }, { "epoch": 0.92, "learning_rate": 9.076844664000628e-05, "loss": 2.9465, "step": 11773 }, { "epoch": 0.92, "learning_rate": 9.076766251078178e-05, "loss": 3.0554, "step": 11774 }, { "epoch": 0.92, "learning_rate": 9.076687838155729e-05, "loss": 3.0858, "step": 11775 }, { "epoch": 0.92, "learning_rate": 9.076609425233279e-05, "loss": 2.8075, "step": 11776 }, { "epoch": 0.92, "learning_rate": 9.07653101231083e-05, "loss": 3.1119, "step": 11777 }, { "epoch": 0.92, "learning_rate": 9.07645259938838e-05, "loss": 2.7875, "step": 11778 }, { "epoch": 0.92, "learning_rate": 9.07637418646593e-05, "loss": 2.9769, "step": 11779 }, { "epoch": 0.92, "learning_rate": 9.07629577354348e-05, "loss": 3.0376, "step": 11780 }, { "epoch": 0.92, "learning_rate": 9.076217360621031e-05, "loss": 3.0322, "step": 11781 }, { "epoch": 0.92, "learning_rate": 9.076138947698581e-05, "loss": 3.206, "step": 11782 }, { "epoch": 0.92, "learning_rate": 9.076060534776132e-05, "loss": 2.9931, "step": 11783 }, { "epoch": 0.92, "learning_rate": 9.075982121853682e-05, "loss": 2.9564, "step": 11784 }, { "epoch": 0.92, "learning_rate": 9.075903708931232e-05, "loss": 2.8958, "step": 11785 }, { "epoch": 0.92, "learning_rate": 9.075825296008783e-05, "loss": 2.8745, "step": 11786 }, { "epoch": 0.92, "learning_rate": 9.075746883086333e-05, "loss": 2.8315, "step": 11787 }, { "epoch": 0.92, "learning_rate": 9.075668470163883e-05, "loss": 2.8315, "step": 11788 }, { "epoch": 0.92, "learning_rate": 9.075590057241434e-05, "loss": 3.0981, "step": 11789 }, { "epoch": 0.92, "learning_rate": 9.075511644318984e-05, "loss": 3.1616, "step": 11790 }, { "epoch": 0.92, "learning_rate": 9.075433231396535e-05, "loss": 3.1159, "step": 11791 }, { "epoch": 0.92, "learning_rate": 9.075354818474085e-05, "loss": 3.2079, "step": 11792 }, { "epoch": 0.92, "learning_rate": 9.075276405551635e-05, "loss": 2.9585, "step": 11793 }, { "epoch": 0.92, "learning_rate": 9.075197992629186e-05, "loss": 2.8568, "step": 11794 }, { "epoch": 0.92, "learning_rate": 9.075119579706736e-05, "loss": 3.0647, "step": 11795 }, { "epoch": 0.92, "learning_rate": 9.075041166784286e-05, "loss": 3.0188, "step": 11796 }, { "epoch": 0.93, "learning_rate": 9.074962753861837e-05, "loss": 3.2576, "step": 11797 }, { "epoch": 0.93, "learning_rate": 9.074884340939387e-05, "loss": 3.0257, "step": 11798 }, { "epoch": 0.93, "learning_rate": 9.074805928016938e-05, "loss": 2.8124, "step": 11799 }, { "epoch": 0.93, "learning_rate": 9.074727515094488e-05, "loss": 2.9098, "step": 11800 }, { "epoch": 0.93, "learning_rate": 9.074649102172038e-05, "loss": 2.8491, "step": 11801 }, { "epoch": 0.93, "learning_rate": 9.074570689249589e-05, "loss": 2.9204, "step": 11802 }, { "epoch": 0.93, "learning_rate": 9.074492276327139e-05, "loss": 3.2284, "step": 11803 }, { "epoch": 0.93, "learning_rate": 9.07441386340469e-05, "loss": 3.0431, "step": 11804 }, { "epoch": 0.93, "learning_rate": 9.07433545048224e-05, "loss": 3.5615, "step": 11805 }, { "epoch": 0.93, "learning_rate": 9.07425703755979e-05, "loss": 3.0881, "step": 11806 }, { "epoch": 0.93, "learning_rate": 9.07417862463734e-05, "loss": 3.2462, "step": 11807 }, { "epoch": 0.93, "learning_rate": 9.074100211714891e-05, "loss": 2.998, "step": 11808 }, { "epoch": 0.93, "learning_rate": 9.074021798792441e-05, "loss": 2.8325, "step": 11809 }, { "epoch": 0.93, "learning_rate": 9.073943385869992e-05, "loss": 2.95, "step": 11810 }, { "epoch": 0.93, "learning_rate": 9.073864972947542e-05, "loss": 3.2813, "step": 11811 }, { "epoch": 0.93, "learning_rate": 9.073786560025092e-05, "loss": 2.9462, "step": 11812 }, { "epoch": 0.93, "learning_rate": 9.073708147102643e-05, "loss": 3.1253, "step": 11813 }, { "epoch": 0.93, "learning_rate": 9.073629734180193e-05, "loss": 3.017, "step": 11814 }, { "epoch": 0.93, "learning_rate": 9.073551321257743e-05, "loss": 3.0626, "step": 11815 }, { "epoch": 0.93, "learning_rate": 9.073472908335294e-05, "loss": 2.8657, "step": 11816 }, { "epoch": 0.93, "learning_rate": 9.073394495412844e-05, "loss": 2.9295, "step": 11817 }, { "epoch": 0.93, "learning_rate": 9.073316082490395e-05, "loss": 3.5103, "step": 11818 }, { "epoch": 0.93, "learning_rate": 9.073237669567945e-05, "loss": 2.9783, "step": 11819 }, { "epoch": 0.93, "learning_rate": 9.073159256645497e-05, "loss": 2.9738, "step": 11820 }, { "epoch": 0.93, "learning_rate": 9.073080843723047e-05, "loss": 2.8964, "step": 11821 }, { "epoch": 0.93, "learning_rate": 9.073002430800597e-05, "loss": 3.0021, "step": 11822 }, { "epoch": 0.93, "learning_rate": 9.072924017878148e-05, "loss": 3.2539, "step": 11823 }, { "epoch": 0.93, "learning_rate": 9.072845604955698e-05, "loss": 3.0625, "step": 11824 }, { "epoch": 0.93, "learning_rate": 9.072767192033248e-05, "loss": 2.9056, "step": 11825 }, { "epoch": 0.93, "learning_rate": 9.072688779110799e-05, "loss": 3.156, "step": 11826 }, { "epoch": 0.93, "learning_rate": 9.072610366188349e-05, "loss": 3.0902, "step": 11827 }, { "epoch": 0.93, "learning_rate": 9.072531953265898e-05, "loss": 2.9757, "step": 11828 }, { "epoch": 0.93, "learning_rate": 9.072453540343449e-05, "loss": 2.8408, "step": 11829 }, { "epoch": 0.93, "learning_rate": 9.072375127420999e-05, "loss": 3.0076, "step": 11830 }, { "epoch": 0.93, "learning_rate": 9.072296714498549e-05, "loss": 2.8804, "step": 11831 }, { "epoch": 0.93, "learning_rate": 9.0722183015761e-05, "loss": 2.8812, "step": 11832 }, { "epoch": 0.93, "learning_rate": 9.07213988865365e-05, "loss": 3.1469, "step": 11833 }, { "epoch": 0.93, "learning_rate": 9.0720614757312e-05, "loss": 3.1523, "step": 11834 }, { "epoch": 0.93, "learning_rate": 9.071983062808751e-05, "loss": 3.1157, "step": 11835 }, { "epoch": 0.93, "learning_rate": 9.071904649886301e-05, "loss": 3.3202, "step": 11836 }, { "epoch": 0.93, "learning_rate": 9.071826236963851e-05, "loss": 3.2632, "step": 11837 }, { "epoch": 0.93, "learning_rate": 9.071747824041402e-05, "loss": 2.9381, "step": 11838 }, { "epoch": 0.93, "learning_rate": 9.071669411118952e-05, "loss": 3.1216, "step": 11839 }, { "epoch": 0.93, "learning_rate": 9.071590998196503e-05, "loss": 3.1921, "step": 11840 }, { "epoch": 0.93, "learning_rate": 9.071512585274053e-05, "loss": 3.0256, "step": 11841 }, { "epoch": 0.93, "learning_rate": 9.071434172351603e-05, "loss": 2.9145, "step": 11842 }, { "epoch": 0.93, "learning_rate": 9.071355759429154e-05, "loss": 3.1253, "step": 11843 }, { "epoch": 0.93, "learning_rate": 9.071277346506704e-05, "loss": 3.1855, "step": 11844 }, { "epoch": 0.93, "learning_rate": 9.071198933584254e-05, "loss": 3.2812, "step": 11845 }, { "epoch": 0.93, "learning_rate": 9.071120520661806e-05, "loss": 2.6894, "step": 11846 }, { "epoch": 0.93, "learning_rate": 9.071042107739357e-05, "loss": 2.9595, "step": 11847 }, { "epoch": 0.93, "learning_rate": 9.070963694816907e-05, "loss": 3.1514, "step": 11848 }, { "epoch": 0.93, "learning_rate": 9.070885281894457e-05, "loss": 3.0527, "step": 11849 }, { "epoch": 0.93, "learning_rate": 9.070806868972008e-05, "loss": 2.9863, "step": 11850 }, { "epoch": 0.93, "learning_rate": 9.070728456049558e-05, "loss": 2.9953, "step": 11851 }, { "epoch": 0.93, "learning_rate": 9.070650043127108e-05, "loss": 2.9019, "step": 11852 }, { "epoch": 0.93, "learning_rate": 9.070571630204659e-05, "loss": 2.8804, "step": 11853 }, { "epoch": 0.93, "learning_rate": 9.070493217282209e-05, "loss": 3.1407, "step": 11854 }, { "epoch": 0.93, "learning_rate": 9.07041480435976e-05, "loss": 3.1633, "step": 11855 }, { "epoch": 0.93, "learning_rate": 9.07033639143731e-05, "loss": 3.4271, "step": 11856 }, { "epoch": 0.93, "learning_rate": 9.07025797851486e-05, "loss": 3.1231, "step": 11857 }, { "epoch": 0.93, "learning_rate": 9.07017956559241e-05, "loss": 3.0962, "step": 11858 }, { "epoch": 0.93, "learning_rate": 9.070101152669961e-05, "loss": 2.7581, "step": 11859 }, { "epoch": 0.93, "learning_rate": 9.070022739747511e-05, "loss": 3.0823, "step": 11860 }, { "epoch": 0.93, "learning_rate": 9.069944326825062e-05, "loss": 3.0364, "step": 11861 }, { "epoch": 0.93, "learning_rate": 9.069865913902612e-05, "loss": 3.1996, "step": 11862 }, { "epoch": 0.93, "learning_rate": 9.069787500980162e-05, "loss": 3.127, "step": 11863 }, { "epoch": 0.93, "learning_rate": 9.069709088057711e-05, "loss": 2.7694, "step": 11864 }, { "epoch": 0.93, "learning_rate": 9.069630675135262e-05, "loss": 2.9184, "step": 11865 }, { "epoch": 0.93, "learning_rate": 9.069552262212812e-05, "loss": 3.0872, "step": 11866 }, { "epoch": 0.93, "learning_rate": 9.069473849290363e-05, "loss": 3.0896, "step": 11867 }, { "epoch": 0.93, "learning_rate": 9.069395436367913e-05, "loss": 3.09, "step": 11868 }, { "epoch": 0.93, "learning_rate": 9.069317023445463e-05, "loss": 2.83, "step": 11869 }, { "epoch": 0.93, "learning_rate": 9.069238610523014e-05, "loss": 2.762, "step": 11870 }, { "epoch": 0.93, "learning_rate": 9.069160197600565e-05, "loss": 2.8132, "step": 11871 }, { "epoch": 0.93, "learning_rate": 9.069081784678116e-05, "loss": 3.0322, "step": 11872 }, { "epoch": 0.93, "learning_rate": 9.069003371755666e-05, "loss": 3.0176, "step": 11873 }, { "epoch": 0.93, "learning_rate": 9.068924958833216e-05, "loss": 2.8021, "step": 11874 }, { "epoch": 0.93, "learning_rate": 9.068846545910767e-05, "loss": 3.0709, "step": 11875 }, { "epoch": 0.93, "learning_rate": 9.068768132988317e-05, "loss": 2.9703, "step": 11876 }, { "epoch": 0.93, "learning_rate": 9.068689720065868e-05, "loss": 3.0289, "step": 11877 }, { "epoch": 0.93, "learning_rate": 9.068611307143418e-05, "loss": 2.8728, "step": 11878 }, { "epoch": 0.93, "learning_rate": 9.068532894220968e-05, "loss": 3.0753, "step": 11879 }, { "epoch": 0.93, "learning_rate": 9.068454481298519e-05, "loss": 2.906, "step": 11880 }, { "epoch": 0.93, "learning_rate": 9.068376068376069e-05, "loss": 2.8346, "step": 11881 }, { "epoch": 0.93, "learning_rate": 9.06829765545362e-05, "loss": 3.0385, "step": 11882 }, { "epoch": 0.93, "learning_rate": 9.06821924253117e-05, "loss": 2.9317, "step": 11883 }, { "epoch": 0.93, "learning_rate": 9.06814082960872e-05, "loss": 3.1818, "step": 11884 }, { "epoch": 0.93, "learning_rate": 9.06806241668627e-05, "loss": 2.8633, "step": 11885 }, { "epoch": 0.93, "learning_rate": 9.067984003763821e-05, "loss": 3.1955, "step": 11886 }, { "epoch": 0.93, "learning_rate": 9.067905590841371e-05, "loss": 3.4617, "step": 11887 }, { "epoch": 0.93, "learning_rate": 9.067827177918922e-05, "loss": 2.9684, "step": 11888 }, { "epoch": 0.93, "learning_rate": 9.067748764996472e-05, "loss": 2.9533, "step": 11889 }, { "epoch": 0.93, "learning_rate": 9.067670352074022e-05, "loss": 2.9793, "step": 11890 }, { "epoch": 0.93, "learning_rate": 9.067591939151573e-05, "loss": 3.0737, "step": 11891 }, { "epoch": 0.93, "learning_rate": 9.067513526229123e-05, "loss": 3.2448, "step": 11892 }, { "epoch": 0.93, "learning_rate": 9.067435113306673e-05, "loss": 3.3441, "step": 11893 }, { "epoch": 0.93, "learning_rate": 9.067356700384224e-05, "loss": 2.9818, "step": 11894 }, { "epoch": 0.93, "learning_rate": 9.067278287461774e-05, "loss": 3.2198, "step": 11895 }, { "epoch": 0.93, "learning_rate": 9.067199874539325e-05, "loss": 2.9742, "step": 11896 }, { "epoch": 0.93, "learning_rate": 9.067121461616875e-05, "loss": 3.2863, "step": 11897 }, { "epoch": 0.93, "learning_rate": 9.067043048694425e-05, "loss": 2.9556, "step": 11898 }, { "epoch": 0.93, "learning_rate": 9.066964635771976e-05, "loss": 3.3371, "step": 11899 }, { "epoch": 0.93, "learning_rate": 9.066886222849526e-05, "loss": 2.9864, "step": 11900 }, { "epoch": 0.93, "learning_rate": 9.066807809927076e-05, "loss": 3.0706, "step": 11901 }, { "epoch": 0.93, "learning_rate": 9.066729397004627e-05, "loss": 3.0612, "step": 11902 }, { "epoch": 0.93, "learning_rate": 9.066650984082177e-05, "loss": 3.1182, "step": 11903 }, { "epoch": 0.93, "learning_rate": 9.066572571159727e-05, "loss": 3.0668, "step": 11904 }, { "epoch": 0.93, "learning_rate": 9.066494158237278e-05, "loss": 3.0184, "step": 11905 }, { "epoch": 0.93, "learning_rate": 9.066415745314828e-05, "loss": 3.0072, "step": 11906 }, { "epoch": 0.93, "learning_rate": 9.066337332392379e-05, "loss": 2.946, "step": 11907 }, { "epoch": 0.93, "learning_rate": 9.066258919469929e-05, "loss": 3.0858, "step": 11908 }, { "epoch": 0.93, "learning_rate": 9.06618050654748e-05, "loss": 3.0074, "step": 11909 }, { "epoch": 0.93, "learning_rate": 9.06610209362503e-05, "loss": 3.0893, "step": 11910 }, { "epoch": 0.93, "learning_rate": 9.06602368070258e-05, "loss": 2.8782, "step": 11911 }, { "epoch": 0.93, "learning_rate": 9.06594526778013e-05, "loss": 2.9705, "step": 11912 }, { "epoch": 0.93, "learning_rate": 9.065866854857681e-05, "loss": 3.0615, "step": 11913 }, { "epoch": 0.93, "learning_rate": 9.065788441935231e-05, "loss": 2.9221, "step": 11914 }, { "epoch": 0.93, "learning_rate": 9.065710029012782e-05, "loss": 2.9216, "step": 11915 }, { "epoch": 0.93, "learning_rate": 9.065631616090332e-05, "loss": 2.933, "step": 11916 }, { "epoch": 0.93, "learning_rate": 9.065553203167882e-05, "loss": 3.0493, "step": 11917 }, { "epoch": 0.93, "learning_rate": 9.065474790245433e-05, "loss": 3.1383, "step": 11918 }, { "epoch": 0.93, "learning_rate": 9.065396377322983e-05, "loss": 2.9855, "step": 11919 }, { "epoch": 0.93, "learning_rate": 9.065317964400533e-05, "loss": 2.9398, "step": 11920 }, { "epoch": 0.93, "learning_rate": 9.065239551478084e-05, "loss": 3.0757, "step": 11921 }, { "epoch": 0.93, "learning_rate": 9.065161138555634e-05, "loss": 2.8214, "step": 11922 }, { "epoch": 0.93, "learning_rate": 9.065082725633184e-05, "loss": 3.203, "step": 11923 }, { "epoch": 0.93, "learning_rate": 9.065004312710735e-05, "loss": 3.0486, "step": 11924 }, { "epoch": 0.94, "learning_rate": 9.064925899788285e-05, "loss": 3.0784, "step": 11925 }, { "epoch": 0.94, "learning_rate": 9.064847486865836e-05, "loss": 3.0706, "step": 11926 }, { "epoch": 0.94, "learning_rate": 9.064769073943386e-05, "loss": 3.3036, "step": 11927 }, { "epoch": 0.94, "learning_rate": 9.064690661020938e-05, "loss": 2.8886, "step": 11928 }, { "epoch": 0.94, "learning_rate": 9.064612248098488e-05, "loss": 2.9584, "step": 11929 }, { "epoch": 0.94, "learning_rate": 9.064533835176038e-05, "loss": 2.9209, "step": 11930 }, { "epoch": 0.94, "learning_rate": 9.064455422253589e-05, "loss": 3.0033, "step": 11931 }, { "epoch": 0.94, "learning_rate": 9.064377009331139e-05, "loss": 3.1943, "step": 11932 }, { "epoch": 0.94, "learning_rate": 9.06429859640869e-05, "loss": 2.9856, "step": 11933 }, { "epoch": 0.94, "learning_rate": 9.06422018348624e-05, "loss": 2.8589, "step": 11934 }, { "epoch": 0.94, "learning_rate": 9.06414177056379e-05, "loss": 2.8743, "step": 11935 }, { "epoch": 0.94, "learning_rate": 9.064063357641339e-05, "loss": 3.1102, "step": 11936 }, { "epoch": 0.94, "learning_rate": 9.06398494471889e-05, "loss": 3.2142, "step": 11937 }, { "epoch": 0.94, "learning_rate": 9.06390653179644e-05, "loss": 2.7906, "step": 11938 }, { "epoch": 0.94, "learning_rate": 9.06382811887399e-05, "loss": 2.818, "step": 11939 }, { "epoch": 0.94, "learning_rate": 9.063749705951541e-05, "loss": 2.99, "step": 11940 }, { "epoch": 0.94, "learning_rate": 9.063671293029091e-05, "loss": 2.8812, "step": 11941 }, { "epoch": 0.94, "learning_rate": 9.063592880106641e-05, "loss": 3.1152, "step": 11942 }, { "epoch": 0.94, "learning_rate": 9.063514467184192e-05, "loss": 2.9875, "step": 11943 }, { "epoch": 0.94, "learning_rate": 9.063436054261742e-05, "loss": 3.1182, "step": 11944 }, { "epoch": 0.94, "learning_rate": 9.063357641339293e-05, "loss": 3.2881, "step": 11945 }, { "epoch": 0.94, "learning_rate": 9.063279228416843e-05, "loss": 2.9707, "step": 11946 }, { "epoch": 0.94, "learning_rate": 9.063200815494393e-05, "loss": 2.894, "step": 11947 }, { "epoch": 0.94, "learning_rate": 9.063122402571944e-05, "loss": 2.7963, "step": 11948 }, { "epoch": 0.94, "learning_rate": 9.063043989649494e-05, "loss": 3.0962, "step": 11949 }, { "epoch": 0.94, "learning_rate": 9.062965576727044e-05, "loss": 3.1064, "step": 11950 }, { "epoch": 0.94, "learning_rate": 9.062887163804595e-05, "loss": 2.7916, "step": 11951 }, { "epoch": 0.94, "learning_rate": 9.062808750882145e-05, "loss": 3.1543, "step": 11952 }, { "epoch": 0.94, "learning_rate": 9.062730337959696e-05, "loss": 3.1142, "step": 11953 }, { "epoch": 0.94, "learning_rate": 9.062651925037247e-05, "loss": 3.0342, "step": 11954 }, { "epoch": 0.94, "learning_rate": 9.062573512114798e-05, "loss": 3.0914, "step": 11955 }, { "epoch": 0.94, "learning_rate": 9.062495099192348e-05, "loss": 2.928, "step": 11956 }, { "epoch": 0.94, "learning_rate": 9.062416686269898e-05, "loss": 2.8619, "step": 11957 }, { "epoch": 0.94, "learning_rate": 9.062338273347449e-05, "loss": 3.0523, "step": 11958 }, { "epoch": 0.94, "learning_rate": 9.062259860424999e-05, "loss": 3.2122, "step": 11959 }, { "epoch": 0.94, "learning_rate": 9.06218144750255e-05, "loss": 3.1819, "step": 11960 }, { "epoch": 0.94, "learning_rate": 9.0621030345801e-05, "loss": 2.8114, "step": 11961 }, { "epoch": 0.94, "learning_rate": 9.06202462165765e-05, "loss": 3.0819, "step": 11962 }, { "epoch": 0.94, "learning_rate": 9.0619462087352e-05, "loss": 3.0022, "step": 11963 }, { "epoch": 0.94, "learning_rate": 9.061867795812751e-05, "loss": 3.2084, "step": 11964 }, { "epoch": 0.94, "learning_rate": 9.061789382890301e-05, "loss": 2.923, "step": 11965 }, { "epoch": 0.94, "learning_rate": 9.061710969967852e-05, "loss": 3.0985, "step": 11966 }, { "epoch": 0.94, "learning_rate": 9.061632557045402e-05, "loss": 2.934, "step": 11967 }, { "epoch": 0.94, "learning_rate": 9.061554144122952e-05, "loss": 2.9875, "step": 11968 }, { "epoch": 0.94, "learning_rate": 9.061475731200503e-05, "loss": 2.8839, "step": 11969 }, { "epoch": 0.94, "learning_rate": 9.061397318278053e-05, "loss": 3.1439, "step": 11970 }, { "epoch": 0.94, "learning_rate": 9.061318905355603e-05, "loss": 2.9543, "step": 11971 }, { "epoch": 0.94, "learning_rate": 9.061240492433152e-05, "loss": 2.98, "step": 11972 }, { "epoch": 0.94, "learning_rate": 9.061162079510703e-05, "loss": 3.0744, "step": 11973 }, { "epoch": 0.94, "learning_rate": 9.061083666588253e-05, "loss": 2.8843, "step": 11974 }, { "epoch": 0.94, "learning_rate": 9.061005253665804e-05, "loss": 2.9099, "step": 11975 }, { "epoch": 0.94, "learning_rate": 9.060926840743354e-05, "loss": 3.2336, "step": 11976 }, { "epoch": 0.94, "learning_rate": 9.060848427820904e-05, "loss": 3.0496, "step": 11977 }, { "epoch": 0.94, "learning_rate": 9.060770014898455e-05, "loss": 3.2564, "step": 11978 }, { "epoch": 0.94, "learning_rate": 9.060691601976006e-05, "loss": 2.919, "step": 11979 }, { "epoch": 0.94, "learning_rate": 9.060613189053557e-05, "loss": 3.1648, "step": 11980 }, { "epoch": 0.94, "learning_rate": 9.060534776131107e-05, "loss": 3.0144, "step": 11981 }, { "epoch": 0.94, "learning_rate": 9.060456363208658e-05, "loss": 3.1216, "step": 11982 }, { "epoch": 0.94, "learning_rate": 9.060377950286208e-05, "loss": 2.8931, "step": 11983 }, { "epoch": 0.94, "learning_rate": 9.060299537363758e-05, "loss": 2.9277, "step": 11984 }, { "epoch": 0.94, "learning_rate": 9.060221124441309e-05, "loss": 3.2042, "step": 11985 }, { "epoch": 0.94, "learning_rate": 9.060142711518859e-05, "loss": 3.0661, "step": 11986 }, { "epoch": 0.94, "learning_rate": 9.06006429859641e-05, "loss": 2.873, "step": 11987 }, { "epoch": 0.94, "learning_rate": 9.05998588567396e-05, "loss": 2.9039, "step": 11988 }, { "epoch": 0.94, "learning_rate": 9.05990747275151e-05, "loss": 3.3374, "step": 11989 }, { "epoch": 0.94, "learning_rate": 9.05982905982906e-05, "loss": 3.0181, "step": 11990 }, { "epoch": 0.94, "learning_rate": 9.059750646906611e-05, "loss": 2.8233, "step": 11991 }, { "epoch": 0.94, "learning_rate": 9.059672233984161e-05, "loss": 2.9732, "step": 11992 }, { "epoch": 0.94, "learning_rate": 9.059593821061712e-05, "loss": 2.9754, "step": 11993 }, { "epoch": 0.94, "learning_rate": 9.059515408139262e-05, "loss": 2.968, "step": 11994 }, { "epoch": 0.94, "learning_rate": 9.059436995216812e-05, "loss": 2.9317, "step": 11995 }, { "epoch": 0.94, "learning_rate": 9.059358582294363e-05, "loss": 3.0459, "step": 11996 }, { "epoch": 0.94, "learning_rate": 9.059280169371913e-05, "loss": 2.9969, "step": 11997 }, { "epoch": 0.94, "learning_rate": 9.059201756449463e-05, "loss": 3.0742, "step": 11998 }, { "epoch": 0.94, "learning_rate": 9.059123343527014e-05, "loss": 3.1161, "step": 11999 }, { "epoch": 0.94, "learning_rate": 9.059044930604564e-05, "loss": 3.0143, "step": 12000 }, { "epoch": 0.94, "eval_loss": 2.9289560317993164, "eval_runtime": 217.9513, "eval_samples_per_second": 51.993, "eval_steps_per_second": 3.253, "step": 12000 }, { "epoch": 0.94, "learning_rate": 9.058966517682115e-05, "loss": 2.8666, "step": 12001 }, { "epoch": 0.94, "learning_rate": 9.058888104759665e-05, "loss": 2.9734, "step": 12002 }, { "epoch": 0.94, "learning_rate": 9.058809691837215e-05, "loss": 3.0127, "step": 12003 }, { "epoch": 0.94, "learning_rate": 9.058731278914766e-05, "loss": 2.7523, "step": 12004 }, { "epoch": 0.94, "learning_rate": 9.058652865992316e-05, "loss": 2.9562, "step": 12005 }, { "epoch": 0.94, "learning_rate": 9.058574453069866e-05, "loss": 3.0097, "step": 12006 }, { "epoch": 0.94, "learning_rate": 9.058496040147417e-05, "loss": 2.8573, "step": 12007 }, { "epoch": 0.94, "learning_rate": 9.058417627224967e-05, "loss": 2.9763, "step": 12008 }, { "epoch": 0.94, "learning_rate": 9.058339214302517e-05, "loss": 3.4809, "step": 12009 }, { "epoch": 0.94, "learning_rate": 9.058260801380068e-05, "loss": 2.7388, "step": 12010 }, { "epoch": 0.94, "learning_rate": 9.058182388457618e-05, "loss": 2.8875, "step": 12011 }, { "epoch": 0.94, "learning_rate": 9.058103975535169e-05, "loss": 2.8667, "step": 12012 }, { "epoch": 0.94, "learning_rate": 9.058025562612719e-05, "loss": 2.9555, "step": 12013 }, { "epoch": 0.94, "learning_rate": 9.057947149690269e-05, "loss": 3.1532, "step": 12014 }, { "epoch": 0.94, "learning_rate": 9.05786873676782e-05, "loss": 3.103, "step": 12015 }, { "epoch": 0.94, "learning_rate": 9.05779032384537e-05, "loss": 3.2617, "step": 12016 }, { "epoch": 0.94, "learning_rate": 9.05771191092292e-05, "loss": 3.2824, "step": 12017 }, { "epoch": 0.94, "learning_rate": 9.057633498000471e-05, "loss": 2.7734, "step": 12018 }, { "epoch": 0.94, "learning_rate": 9.057555085078021e-05, "loss": 2.9605, "step": 12019 }, { "epoch": 0.94, "learning_rate": 9.057476672155572e-05, "loss": 3.0525, "step": 12020 }, { "epoch": 0.94, "learning_rate": 9.057398259233122e-05, "loss": 2.9098, "step": 12021 }, { "epoch": 0.94, "learning_rate": 9.057319846310672e-05, "loss": 3.0098, "step": 12022 }, { "epoch": 0.94, "learning_rate": 9.057241433388223e-05, "loss": 3.0281, "step": 12023 }, { "epoch": 0.94, "learning_rate": 9.057163020465773e-05, "loss": 2.8303, "step": 12024 }, { "epoch": 0.94, "learning_rate": 9.057084607543323e-05, "loss": 2.9616, "step": 12025 }, { "epoch": 0.94, "learning_rate": 9.057006194620874e-05, "loss": 3.1067, "step": 12026 }, { "epoch": 0.94, "learning_rate": 9.056927781698424e-05, "loss": 2.9748, "step": 12027 }, { "epoch": 0.94, "learning_rate": 9.056849368775974e-05, "loss": 2.853, "step": 12028 }, { "epoch": 0.94, "learning_rate": 9.056770955853525e-05, "loss": 2.8981, "step": 12029 }, { "epoch": 0.94, "learning_rate": 9.056692542931075e-05, "loss": 3.2475, "step": 12030 }, { "epoch": 0.94, "learning_rate": 9.056614130008626e-05, "loss": 3.0891, "step": 12031 }, { "epoch": 0.94, "learning_rate": 9.056535717086176e-05, "loss": 3.2589, "step": 12032 }, { "epoch": 0.94, "learning_rate": 9.056457304163726e-05, "loss": 3.3899, "step": 12033 }, { "epoch": 0.94, "learning_rate": 9.056378891241277e-05, "loss": 3.0756, "step": 12034 }, { "epoch": 0.94, "learning_rate": 9.056300478318827e-05, "loss": 3.1056, "step": 12035 }, { "epoch": 0.94, "learning_rate": 9.056222065396379e-05, "loss": 3.1478, "step": 12036 }, { "epoch": 0.94, "learning_rate": 9.056143652473929e-05, "loss": 3.1516, "step": 12037 }, { "epoch": 0.94, "learning_rate": 9.05606523955148e-05, "loss": 2.8886, "step": 12038 }, { "epoch": 0.94, "learning_rate": 9.05598682662903e-05, "loss": 3.157, "step": 12039 }, { "epoch": 0.94, "learning_rate": 9.05590841370658e-05, "loss": 2.9925, "step": 12040 }, { "epoch": 0.94, "learning_rate": 9.05583000078413e-05, "loss": 3.0632, "step": 12041 }, { "epoch": 0.94, "learning_rate": 9.055751587861681e-05, "loss": 2.9138, "step": 12042 }, { "epoch": 0.94, "learning_rate": 9.055673174939231e-05, "loss": 3.0537, "step": 12043 }, { "epoch": 0.94, "learning_rate": 9.05559476201678e-05, "loss": 3.0055, "step": 12044 }, { "epoch": 0.94, "learning_rate": 9.055516349094331e-05, "loss": 2.9913, "step": 12045 }, { "epoch": 0.94, "learning_rate": 9.055437936171881e-05, "loss": 3.2426, "step": 12046 }, { "epoch": 0.94, "learning_rate": 9.055359523249431e-05, "loss": 3.3779, "step": 12047 }, { "epoch": 0.94, "learning_rate": 9.055281110326982e-05, "loss": 3.2371, "step": 12048 }, { "epoch": 0.94, "learning_rate": 9.055202697404532e-05, "loss": 3.1533, "step": 12049 }, { "epoch": 0.94, "learning_rate": 9.055124284482083e-05, "loss": 3.0042, "step": 12050 }, { "epoch": 0.94, "learning_rate": 9.055045871559633e-05, "loss": 3.3387, "step": 12051 }, { "epoch": 0.95, "learning_rate": 9.054967458637183e-05, "loss": 3.0373, "step": 12052 }, { "epoch": 0.95, "learning_rate": 9.054889045714734e-05, "loss": 2.9967, "step": 12053 }, { "epoch": 0.95, "learning_rate": 9.054810632792284e-05, "loss": 3.2325, "step": 12054 }, { "epoch": 0.95, "learning_rate": 9.054732219869834e-05, "loss": 3.0761, "step": 12055 }, { "epoch": 0.95, "learning_rate": 9.054653806947385e-05, "loss": 2.7944, "step": 12056 }, { "epoch": 0.95, "learning_rate": 9.054575394024935e-05, "loss": 3.0138, "step": 12057 }, { "epoch": 0.95, "learning_rate": 9.054496981102485e-05, "loss": 2.9569, "step": 12058 }, { "epoch": 0.95, "learning_rate": 9.054418568180036e-05, "loss": 3.0726, "step": 12059 }, { "epoch": 0.95, "learning_rate": 9.054340155257586e-05, "loss": 3.1644, "step": 12060 }, { "epoch": 0.95, "learning_rate": 9.054261742335137e-05, "loss": 2.8083, "step": 12061 }, { "epoch": 0.95, "learning_rate": 9.054183329412688e-05, "loss": 2.9011, "step": 12062 }, { "epoch": 0.95, "learning_rate": 9.054104916490239e-05, "loss": 3.0511, "step": 12063 }, { "epoch": 0.95, "learning_rate": 9.054026503567789e-05, "loss": 3.0759, "step": 12064 }, { "epoch": 0.95, "learning_rate": 9.05394809064534e-05, "loss": 3.1077, "step": 12065 }, { "epoch": 0.95, "learning_rate": 9.05386967772289e-05, "loss": 3.2543, "step": 12066 }, { "epoch": 0.95, "learning_rate": 9.05379126480044e-05, "loss": 3.1429, "step": 12067 }, { "epoch": 0.95, "learning_rate": 9.05371285187799e-05, "loss": 3.3037, "step": 12068 }, { "epoch": 0.95, "learning_rate": 9.053634438955541e-05, "loss": 2.8202, "step": 12069 }, { "epoch": 0.95, "learning_rate": 9.053556026033091e-05, "loss": 3.0264, "step": 12070 }, { "epoch": 0.95, "learning_rate": 9.053477613110642e-05, "loss": 3.0585, "step": 12071 }, { "epoch": 0.95, "learning_rate": 9.053399200188192e-05, "loss": 3.0753, "step": 12072 }, { "epoch": 0.95, "learning_rate": 9.053320787265742e-05, "loss": 3.1142, "step": 12073 }, { "epoch": 0.95, "learning_rate": 9.053242374343293e-05, "loss": 2.9687, "step": 12074 }, { "epoch": 0.95, "learning_rate": 9.053163961420843e-05, "loss": 3.009, "step": 12075 }, { "epoch": 0.95, "learning_rate": 9.053085548498393e-05, "loss": 2.9889, "step": 12076 }, { "epoch": 0.95, "learning_rate": 9.053007135575944e-05, "loss": 3.0983, "step": 12077 }, { "epoch": 0.95, "learning_rate": 9.052928722653494e-05, "loss": 2.9833, "step": 12078 }, { "epoch": 0.95, "learning_rate": 9.052850309731045e-05, "loss": 3.0057, "step": 12079 }, { "epoch": 0.95, "learning_rate": 9.052771896808594e-05, "loss": 3.3171, "step": 12080 }, { "epoch": 0.95, "learning_rate": 9.052693483886144e-05, "loss": 3.0154, "step": 12081 }, { "epoch": 0.95, "learning_rate": 9.052615070963694e-05, "loss": 3.0521, "step": 12082 }, { "epoch": 0.95, "learning_rate": 9.052536658041245e-05, "loss": 2.9968, "step": 12083 }, { "epoch": 0.95, "learning_rate": 9.052458245118795e-05, "loss": 3.2551, "step": 12084 }, { "epoch": 0.95, "learning_rate": 9.052379832196345e-05, "loss": 2.7944, "step": 12085 }, { "epoch": 0.95, "learning_rate": 9.052301419273896e-05, "loss": 2.8956, "step": 12086 }, { "epoch": 0.95, "learning_rate": 9.052223006351446e-05, "loss": 3.0886, "step": 12087 }, { "epoch": 0.95, "learning_rate": 9.052144593428998e-05, "loss": 2.8748, "step": 12088 }, { "epoch": 0.95, "learning_rate": 9.052066180506548e-05, "loss": 3.0129, "step": 12089 }, { "epoch": 0.95, "learning_rate": 9.051987767584099e-05, "loss": 3.0237, "step": 12090 }, { "epoch": 0.95, "learning_rate": 9.051909354661649e-05, "loss": 3.5676, "step": 12091 }, { "epoch": 0.95, "learning_rate": 9.0518309417392e-05, "loss": 3.1117, "step": 12092 }, { "epoch": 0.95, "learning_rate": 9.05175252881675e-05, "loss": 3.0679, "step": 12093 }, { "epoch": 0.95, "learning_rate": 9.0516741158943e-05, "loss": 2.9041, "step": 12094 }, { "epoch": 0.95, "learning_rate": 9.05159570297185e-05, "loss": 2.8247, "step": 12095 }, { "epoch": 0.95, "learning_rate": 9.051517290049401e-05, "loss": 3.1211, "step": 12096 }, { "epoch": 0.95, "learning_rate": 9.051438877126951e-05, "loss": 2.889, "step": 12097 }, { "epoch": 0.95, "learning_rate": 9.051360464204502e-05, "loss": 2.7415, "step": 12098 }, { "epoch": 0.95, "learning_rate": 9.051282051282052e-05, "loss": 3.0655, "step": 12099 }, { "epoch": 0.95, "learning_rate": 9.051203638359602e-05, "loss": 2.9575, "step": 12100 }, { "epoch": 0.95, "learning_rate": 9.051125225437153e-05, "loss": 3.4492, "step": 12101 }, { "epoch": 0.95, "learning_rate": 9.051046812514703e-05, "loss": 2.9244, "step": 12102 }, { "epoch": 0.95, "learning_rate": 9.050968399592253e-05, "loss": 3.0488, "step": 12103 }, { "epoch": 0.95, "learning_rate": 9.050889986669804e-05, "loss": 2.9783, "step": 12104 }, { "epoch": 0.95, "learning_rate": 9.050811573747354e-05, "loss": 2.8195, "step": 12105 }, { "epoch": 0.95, "learning_rate": 9.050733160824904e-05, "loss": 2.9939, "step": 12106 }, { "epoch": 0.95, "learning_rate": 9.050654747902455e-05, "loss": 2.9076, "step": 12107 }, { "epoch": 0.95, "learning_rate": 9.050576334980005e-05, "loss": 2.9779, "step": 12108 }, { "epoch": 0.95, "learning_rate": 9.050497922057556e-05, "loss": 3.1166, "step": 12109 }, { "epoch": 0.95, "learning_rate": 9.050419509135106e-05, "loss": 2.7853, "step": 12110 }, { "epoch": 0.95, "learning_rate": 9.050341096212656e-05, "loss": 2.9505, "step": 12111 }, { "epoch": 0.95, "learning_rate": 9.050262683290207e-05, "loss": 2.7801, "step": 12112 }, { "epoch": 0.95, "learning_rate": 9.050184270367757e-05, "loss": 2.9204, "step": 12113 }, { "epoch": 0.95, "learning_rate": 9.050105857445307e-05, "loss": 2.9465, "step": 12114 }, { "epoch": 0.95, "learning_rate": 9.050027444522858e-05, "loss": 2.9486, "step": 12115 }, { "epoch": 0.95, "learning_rate": 9.049949031600408e-05, "loss": 2.9388, "step": 12116 }, { "epoch": 0.95, "learning_rate": 9.049870618677959e-05, "loss": 2.8842, "step": 12117 }, { "epoch": 0.95, "learning_rate": 9.049792205755509e-05, "loss": 2.9015, "step": 12118 }, { "epoch": 0.95, "learning_rate": 9.049713792833059e-05, "loss": 3.1932, "step": 12119 }, { "epoch": 0.95, "learning_rate": 9.04963537991061e-05, "loss": 3.1206, "step": 12120 }, { "epoch": 0.95, "learning_rate": 9.04955696698816e-05, "loss": 3.3823, "step": 12121 }, { "epoch": 0.95, "learning_rate": 9.04947855406571e-05, "loss": 2.9533, "step": 12122 }, { "epoch": 0.95, "learning_rate": 9.049400141143261e-05, "loss": 3.0194, "step": 12123 }, { "epoch": 0.95, "learning_rate": 9.049321728220811e-05, "loss": 3.0513, "step": 12124 }, { "epoch": 0.95, "learning_rate": 9.049243315298361e-05, "loss": 2.9346, "step": 12125 }, { "epoch": 0.95, "learning_rate": 9.049164902375912e-05, "loss": 3.0668, "step": 12126 }, { "epoch": 0.95, "learning_rate": 9.049086489453462e-05, "loss": 3.1269, "step": 12127 }, { "epoch": 0.95, "learning_rate": 9.049008076531013e-05, "loss": 3.2037, "step": 12128 }, { "epoch": 0.95, "learning_rate": 9.048929663608563e-05, "loss": 3.3236, "step": 12129 }, { "epoch": 0.95, "learning_rate": 9.048851250686113e-05, "loss": 3.3936, "step": 12130 }, { "epoch": 0.95, "learning_rate": 9.048772837763664e-05, "loss": 3.1408, "step": 12131 }, { "epoch": 0.95, "learning_rate": 9.048694424841214e-05, "loss": 2.9409, "step": 12132 }, { "epoch": 0.95, "learning_rate": 9.048616011918764e-05, "loss": 3.1183, "step": 12133 }, { "epoch": 0.95, "learning_rate": 9.048537598996315e-05, "loss": 2.776, "step": 12134 }, { "epoch": 0.95, "learning_rate": 9.048459186073865e-05, "loss": 3.0518, "step": 12135 }, { "epoch": 0.95, "learning_rate": 9.048380773151416e-05, "loss": 3.1555, "step": 12136 }, { "epoch": 0.95, "learning_rate": 9.048302360228966e-05, "loss": 3.2635, "step": 12137 }, { "epoch": 0.95, "learning_rate": 9.048223947306516e-05, "loss": 2.9778, "step": 12138 }, { "epoch": 0.95, "learning_rate": 9.048145534384067e-05, "loss": 3.118, "step": 12139 }, { "epoch": 0.95, "learning_rate": 9.048067121461617e-05, "loss": 2.9559, "step": 12140 }, { "epoch": 0.95, "learning_rate": 9.047988708539167e-05, "loss": 3.1072, "step": 12141 }, { "epoch": 0.95, "learning_rate": 9.047910295616718e-05, "loss": 3.0104, "step": 12142 }, { "epoch": 0.95, "learning_rate": 9.047831882694268e-05, "loss": 2.7023, "step": 12143 }, { "epoch": 0.95, "learning_rate": 9.04775346977182e-05, "loss": 2.97, "step": 12144 }, { "epoch": 0.95, "learning_rate": 9.04767505684937e-05, "loss": 2.8998, "step": 12145 }, { "epoch": 0.95, "learning_rate": 9.04759664392692e-05, "loss": 3.0896, "step": 12146 }, { "epoch": 0.95, "learning_rate": 9.047518231004471e-05, "loss": 2.924, "step": 12147 }, { "epoch": 0.95, "learning_rate": 9.047439818082021e-05, "loss": 3.2618, "step": 12148 }, { "epoch": 0.95, "learning_rate": 9.047361405159572e-05, "loss": 3.0843, "step": 12149 }, { "epoch": 0.95, "learning_rate": 9.047282992237122e-05, "loss": 3.1072, "step": 12150 }, { "epoch": 0.95, "learning_rate": 9.047204579314672e-05, "loss": 2.8407, "step": 12151 }, { "epoch": 0.95, "learning_rate": 9.047126166392221e-05, "loss": 2.9293, "step": 12152 }, { "epoch": 0.95, "learning_rate": 9.047047753469772e-05, "loss": 2.9624, "step": 12153 }, { "epoch": 0.95, "learning_rate": 9.046969340547322e-05, "loss": 2.8682, "step": 12154 }, { "epoch": 0.95, "learning_rate": 9.046890927624873e-05, "loss": 3.0556, "step": 12155 }, { "epoch": 0.95, "learning_rate": 9.046812514702423e-05, "loss": 2.9853, "step": 12156 }, { "epoch": 0.95, "learning_rate": 9.046734101779973e-05, "loss": 3.1568, "step": 12157 }, { "epoch": 0.95, "learning_rate": 9.046655688857524e-05, "loss": 3.1446, "step": 12158 }, { "epoch": 0.95, "learning_rate": 9.046577275935074e-05, "loss": 3.0761, "step": 12159 }, { "epoch": 0.95, "learning_rate": 9.046498863012624e-05, "loss": 3.1473, "step": 12160 }, { "epoch": 0.95, "learning_rate": 9.046420450090175e-05, "loss": 2.8832, "step": 12161 }, { "epoch": 0.95, "learning_rate": 9.046342037167725e-05, "loss": 3.0714, "step": 12162 }, { "epoch": 0.95, "learning_rate": 9.046263624245275e-05, "loss": 2.9185, "step": 12163 }, { "epoch": 0.95, "learning_rate": 9.046185211322826e-05, "loss": 2.9633, "step": 12164 }, { "epoch": 0.95, "learning_rate": 9.046106798400376e-05, "loss": 3.1506, "step": 12165 }, { "epoch": 0.95, "learning_rate": 9.046028385477927e-05, "loss": 3.2216, "step": 12166 }, { "epoch": 0.95, "learning_rate": 9.045949972555477e-05, "loss": 2.941, "step": 12167 }, { "epoch": 0.95, "learning_rate": 9.045871559633027e-05, "loss": 3.2069, "step": 12168 }, { "epoch": 0.95, "learning_rate": 9.045793146710578e-05, "loss": 3.0648, "step": 12169 }, { "epoch": 0.95, "learning_rate": 9.04571473378813e-05, "loss": 2.9449, "step": 12170 }, { "epoch": 0.95, "learning_rate": 9.04563632086568e-05, "loss": 3.1383, "step": 12171 }, { "epoch": 0.95, "learning_rate": 9.04555790794323e-05, "loss": 3.1231, "step": 12172 }, { "epoch": 0.95, "learning_rate": 9.04547949502078e-05, "loss": 2.9373, "step": 12173 }, { "epoch": 0.95, "learning_rate": 9.045401082098331e-05, "loss": 3.0192, "step": 12174 }, { "epoch": 0.95, "learning_rate": 9.045322669175881e-05, "loss": 3.1427, "step": 12175 }, { "epoch": 0.95, "learning_rate": 9.045244256253432e-05, "loss": 3.1144, "step": 12176 }, { "epoch": 0.95, "learning_rate": 9.045165843330982e-05, "loss": 3.2077, "step": 12177 }, { "epoch": 0.95, "learning_rate": 9.045087430408532e-05, "loss": 3.0499, "step": 12178 }, { "epoch": 0.95, "learning_rate": 9.045009017486083e-05, "loss": 3.0635, "step": 12179 }, { "epoch": 0.96, "learning_rate": 9.044930604563633e-05, "loss": 3.3685, "step": 12180 }, { "epoch": 0.96, "learning_rate": 9.044852191641183e-05, "loss": 2.8758, "step": 12181 }, { "epoch": 0.96, "learning_rate": 9.044773778718734e-05, "loss": 3.042, "step": 12182 }, { "epoch": 0.96, "learning_rate": 9.044695365796284e-05, "loss": 3.0203, "step": 12183 }, { "epoch": 0.96, "learning_rate": 9.044616952873835e-05, "loss": 3.1314, "step": 12184 }, { "epoch": 0.96, "learning_rate": 9.044538539951385e-05, "loss": 2.9816, "step": 12185 }, { "epoch": 0.96, "learning_rate": 9.044460127028935e-05, "loss": 2.9142, "step": 12186 }, { "epoch": 0.96, "learning_rate": 9.044381714106486e-05, "loss": 2.9292, "step": 12187 }, { "epoch": 0.96, "learning_rate": 9.044303301184035e-05, "loss": 3.0031, "step": 12188 }, { "epoch": 0.96, "learning_rate": 9.044224888261585e-05, "loss": 3.0166, "step": 12189 }, { "epoch": 0.96, "learning_rate": 9.044146475339135e-05, "loss": 3.0233, "step": 12190 }, { "epoch": 0.96, "learning_rate": 9.044068062416686e-05, "loss": 2.9433, "step": 12191 }, { "epoch": 0.96, "learning_rate": 9.043989649494236e-05, "loss": 2.7943, "step": 12192 }, { "epoch": 0.96, "learning_rate": 9.043911236571786e-05, "loss": 2.7323, "step": 12193 }, { "epoch": 0.96, "learning_rate": 9.043832823649337e-05, "loss": 2.9708, "step": 12194 }, { "epoch": 0.96, "learning_rate": 9.043754410726887e-05, "loss": 3.1806, "step": 12195 }, { "epoch": 0.96, "learning_rate": 9.043675997804439e-05, "loss": 3.0047, "step": 12196 }, { "epoch": 0.96, "learning_rate": 9.043597584881989e-05, "loss": 3.035, "step": 12197 }, { "epoch": 0.96, "learning_rate": 9.04351917195954e-05, "loss": 3.2537, "step": 12198 }, { "epoch": 0.96, "learning_rate": 9.04344075903709e-05, "loss": 2.9533, "step": 12199 }, { "epoch": 0.96, "learning_rate": 9.04336234611464e-05, "loss": 3.0752, "step": 12200 }, { "epoch": 0.96, "learning_rate": 9.043283933192191e-05, "loss": 3.1348, "step": 12201 }, { "epoch": 0.96, "learning_rate": 9.043205520269741e-05, "loss": 3.0655, "step": 12202 }, { "epoch": 0.96, "learning_rate": 9.043127107347292e-05, "loss": 2.886, "step": 12203 }, { "epoch": 0.96, "learning_rate": 9.043048694424842e-05, "loss": 2.9531, "step": 12204 }, { "epoch": 0.96, "learning_rate": 9.042970281502392e-05, "loss": 3.0283, "step": 12205 }, { "epoch": 0.96, "learning_rate": 9.042891868579943e-05, "loss": 3.0741, "step": 12206 }, { "epoch": 0.96, "learning_rate": 9.042813455657493e-05, "loss": 2.9859, "step": 12207 }, { "epoch": 0.96, "learning_rate": 9.042735042735043e-05, "loss": 3.2836, "step": 12208 }, { "epoch": 0.96, "learning_rate": 9.042656629812594e-05, "loss": 3.0509, "step": 12209 }, { "epoch": 0.96, "learning_rate": 9.042578216890144e-05, "loss": 2.8713, "step": 12210 }, { "epoch": 0.96, "learning_rate": 9.042499803967694e-05, "loss": 3.2336, "step": 12211 }, { "epoch": 0.96, "learning_rate": 9.042421391045245e-05, "loss": 3.2107, "step": 12212 }, { "epoch": 0.96, "learning_rate": 9.042342978122795e-05, "loss": 2.868, "step": 12213 }, { "epoch": 0.96, "learning_rate": 9.042264565200346e-05, "loss": 3.0367, "step": 12214 }, { "epoch": 0.96, "learning_rate": 9.042186152277896e-05, "loss": 3.159, "step": 12215 }, { "epoch": 0.96, "learning_rate": 9.042107739355446e-05, "loss": 2.8227, "step": 12216 }, { "epoch": 0.96, "learning_rate": 9.042029326432997e-05, "loss": 2.9563, "step": 12217 }, { "epoch": 0.96, "learning_rate": 9.041950913510547e-05, "loss": 3.2382, "step": 12218 }, { "epoch": 0.96, "learning_rate": 9.041872500588097e-05, "loss": 2.9949, "step": 12219 }, { "epoch": 0.96, "learning_rate": 9.041794087665648e-05, "loss": 3.3383, "step": 12220 }, { "epoch": 0.96, "learning_rate": 9.041715674743198e-05, "loss": 3.0038, "step": 12221 }, { "epoch": 0.96, "learning_rate": 9.041637261820749e-05, "loss": 3.1529, "step": 12222 }, { "epoch": 0.96, "learning_rate": 9.041558848898299e-05, "loss": 2.9193, "step": 12223 }, { "epoch": 0.96, "learning_rate": 9.041480435975849e-05, "loss": 3.0255, "step": 12224 }, { "epoch": 0.96, "learning_rate": 9.0414020230534e-05, "loss": 2.9416, "step": 12225 }, { "epoch": 0.96, "learning_rate": 9.04132361013095e-05, "loss": 2.9311, "step": 12226 }, { "epoch": 0.96, "learning_rate": 9.0412451972085e-05, "loss": 2.8708, "step": 12227 }, { "epoch": 0.96, "learning_rate": 9.041166784286051e-05, "loss": 3.038, "step": 12228 }, { "epoch": 0.96, "learning_rate": 9.041088371363601e-05, "loss": 2.9174, "step": 12229 }, { "epoch": 0.96, "learning_rate": 9.041009958441151e-05, "loss": 2.9602, "step": 12230 }, { "epoch": 0.96, "learning_rate": 9.040931545518702e-05, "loss": 3.1651, "step": 12231 }, { "epoch": 0.96, "learning_rate": 9.040853132596252e-05, "loss": 2.8113, "step": 12232 }, { "epoch": 0.96, "learning_rate": 9.040774719673803e-05, "loss": 3.1455, "step": 12233 }, { "epoch": 0.96, "learning_rate": 9.040696306751353e-05, "loss": 3.0437, "step": 12234 }, { "epoch": 0.96, "learning_rate": 9.040617893828903e-05, "loss": 3.0191, "step": 12235 }, { "epoch": 0.96, "learning_rate": 9.040539480906454e-05, "loss": 2.8746, "step": 12236 }, { "epoch": 0.96, "learning_rate": 9.040461067984004e-05, "loss": 3.0724, "step": 12237 }, { "epoch": 0.96, "learning_rate": 9.040382655061554e-05, "loss": 3.0404, "step": 12238 }, { "epoch": 0.96, "learning_rate": 9.040304242139105e-05, "loss": 3.053, "step": 12239 }, { "epoch": 0.96, "learning_rate": 9.040225829216655e-05, "loss": 3.0963, "step": 12240 }, { "epoch": 0.96, "learning_rate": 9.040147416294205e-05, "loss": 2.8842, "step": 12241 }, { "epoch": 0.96, "learning_rate": 9.040069003371756e-05, "loss": 3.1357, "step": 12242 }, { "epoch": 0.96, "learning_rate": 9.039990590449306e-05, "loss": 2.8844, "step": 12243 }, { "epoch": 0.96, "learning_rate": 9.039912177526857e-05, "loss": 3.2473, "step": 12244 }, { "epoch": 0.96, "learning_rate": 9.039833764604407e-05, "loss": 3.4187, "step": 12245 }, { "epoch": 0.96, "learning_rate": 9.039755351681957e-05, "loss": 2.8213, "step": 12246 }, { "epoch": 0.96, "learning_rate": 9.039676938759508e-05, "loss": 2.7677, "step": 12247 }, { "epoch": 0.96, "learning_rate": 9.039598525837058e-05, "loss": 3.4699, "step": 12248 }, { "epoch": 0.96, "learning_rate": 9.039520112914608e-05, "loss": 2.9652, "step": 12249 }, { "epoch": 0.96, "learning_rate": 9.039441699992159e-05, "loss": 2.9139, "step": 12250 }, { "epoch": 0.96, "learning_rate": 9.039363287069709e-05, "loss": 2.929, "step": 12251 }, { "epoch": 0.96, "learning_rate": 9.039284874147261e-05, "loss": 3.0378, "step": 12252 }, { "epoch": 0.96, "learning_rate": 9.039206461224811e-05, "loss": 3.252, "step": 12253 }, { "epoch": 0.96, "learning_rate": 9.039128048302362e-05, "loss": 3.2157, "step": 12254 }, { "epoch": 0.96, "learning_rate": 9.039049635379912e-05, "loss": 3.0444, "step": 12255 }, { "epoch": 0.96, "learning_rate": 9.038971222457462e-05, "loss": 2.8929, "step": 12256 }, { "epoch": 0.96, "learning_rate": 9.038892809535013e-05, "loss": 3.0679, "step": 12257 }, { "epoch": 0.96, "learning_rate": 9.038814396612563e-05, "loss": 2.7396, "step": 12258 }, { "epoch": 0.96, "learning_rate": 9.038735983690113e-05, "loss": 2.8525, "step": 12259 }, { "epoch": 0.96, "learning_rate": 9.038657570767662e-05, "loss": 3.1061, "step": 12260 }, { "epoch": 0.96, "learning_rate": 9.038579157845213e-05, "loss": 3.1266, "step": 12261 }, { "epoch": 0.96, "learning_rate": 9.038500744922763e-05, "loss": 2.9718, "step": 12262 }, { "epoch": 0.96, "learning_rate": 9.038422332000314e-05, "loss": 2.9797, "step": 12263 }, { "epoch": 0.96, "learning_rate": 9.038343919077864e-05, "loss": 3.3331, "step": 12264 }, { "epoch": 0.96, "learning_rate": 9.038265506155414e-05, "loss": 3.0126, "step": 12265 }, { "epoch": 0.96, "learning_rate": 9.038187093232965e-05, "loss": 3.0998, "step": 12266 }, { "epoch": 0.96, "learning_rate": 9.038108680310515e-05, "loss": 2.9839, "step": 12267 }, { "epoch": 0.96, "learning_rate": 9.038030267388065e-05, "loss": 3.283, "step": 12268 }, { "epoch": 0.96, "learning_rate": 9.037951854465616e-05, "loss": 3.0893, "step": 12269 }, { "epoch": 0.96, "learning_rate": 9.037873441543166e-05, "loss": 3.1459, "step": 12270 }, { "epoch": 0.96, "learning_rate": 9.037795028620717e-05, "loss": 2.9816, "step": 12271 }, { "epoch": 0.96, "learning_rate": 9.037716615698267e-05, "loss": 2.9312, "step": 12272 }, { "epoch": 0.96, "learning_rate": 9.037638202775817e-05, "loss": 2.9431, "step": 12273 }, { "epoch": 0.96, "learning_rate": 9.037559789853368e-05, "loss": 3.0192, "step": 12274 }, { "epoch": 0.96, "learning_rate": 9.037481376930918e-05, "loss": 2.9742, "step": 12275 }, { "epoch": 0.96, "learning_rate": 9.037402964008468e-05, "loss": 3.214, "step": 12276 }, { "epoch": 0.96, "learning_rate": 9.037324551086019e-05, "loss": 2.7526, "step": 12277 }, { "epoch": 0.96, "learning_rate": 9.03724613816357e-05, "loss": 3.0522, "step": 12278 }, { "epoch": 0.96, "learning_rate": 9.037167725241121e-05, "loss": 3.0567, "step": 12279 }, { "epoch": 0.96, "learning_rate": 9.037089312318671e-05, "loss": 3.1999, "step": 12280 }, { "epoch": 0.96, "learning_rate": 9.037010899396222e-05, "loss": 2.8817, "step": 12281 }, { "epoch": 0.96, "learning_rate": 9.036932486473772e-05, "loss": 3.0685, "step": 12282 }, { "epoch": 0.96, "learning_rate": 9.036854073551322e-05, "loss": 2.7366, "step": 12283 }, { "epoch": 0.96, "learning_rate": 9.036775660628873e-05, "loss": 2.8053, "step": 12284 }, { "epoch": 0.96, "learning_rate": 9.036697247706423e-05, "loss": 2.9951, "step": 12285 }, { "epoch": 0.96, "learning_rate": 9.036618834783973e-05, "loss": 2.9438, "step": 12286 }, { "epoch": 0.96, "learning_rate": 9.036540421861524e-05, "loss": 3.2846, "step": 12287 }, { "epoch": 0.96, "learning_rate": 9.036462008939074e-05, "loss": 2.8769, "step": 12288 }, { "epoch": 0.96, "learning_rate": 9.036383596016625e-05, "loss": 3.0032, "step": 12289 }, { "epoch": 0.96, "learning_rate": 9.036305183094175e-05, "loss": 3.4594, "step": 12290 }, { "epoch": 0.96, "learning_rate": 9.036226770171725e-05, "loss": 2.8659, "step": 12291 }, { "epoch": 0.96, "learning_rate": 9.036148357249276e-05, "loss": 2.8476, "step": 12292 }, { "epoch": 0.96, "learning_rate": 9.036069944326826e-05, "loss": 2.8606, "step": 12293 }, { "epoch": 0.96, "learning_rate": 9.035991531404376e-05, "loss": 3.1792, "step": 12294 }, { "epoch": 0.96, "learning_rate": 9.035913118481927e-05, "loss": 2.9802, "step": 12295 }, { "epoch": 0.96, "learning_rate": 9.035834705559476e-05, "loss": 3.0975, "step": 12296 }, { "epoch": 0.96, "learning_rate": 9.035756292637026e-05, "loss": 3.4564, "step": 12297 }, { "epoch": 0.96, "learning_rate": 9.035677879714576e-05, "loss": 3.2618, "step": 12298 }, { "epoch": 0.96, "learning_rate": 9.035599466792127e-05, "loss": 3.2198, "step": 12299 }, { "epoch": 0.96, "learning_rate": 9.035521053869677e-05, "loss": 3.2064, "step": 12300 }, { "epoch": 0.96, "learning_rate": 9.035442640947228e-05, "loss": 3.1068, "step": 12301 }, { "epoch": 0.96, "learning_rate": 9.035364228024778e-05, "loss": 3.0039, "step": 12302 }, { "epoch": 0.96, "learning_rate": 9.035285815102328e-05, "loss": 2.8795, "step": 12303 }, { "epoch": 0.96, "learning_rate": 9.03520740217988e-05, "loss": 3.2107, "step": 12304 }, { "epoch": 0.96, "learning_rate": 9.03512898925743e-05, "loss": 2.8079, "step": 12305 }, { "epoch": 0.96, "learning_rate": 9.035050576334981e-05, "loss": 2.9046, "step": 12306 }, { "epoch": 0.97, "learning_rate": 9.034972163412531e-05, "loss": 3.2677, "step": 12307 }, { "epoch": 0.97, "learning_rate": 9.034893750490081e-05, "loss": 2.8696, "step": 12308 }, { "epoch": 0.97, "learning_rate": 9.034815337567632e-05, "loss": 2.8447, "step": 12309 }, { "epoch": 0.97, "learning_rate": 9.034736924645182e-05, "loss": 3.1, "step": 12310 }, { "epoch": 0.97, "learning_rate": 9.034658511722733e-05, "loss": 2.8803, "step": 12311 }, { "epoch": 0.97, "learning_rate": 9.034580098800283e-05, "loss": 2.8624, "step": 12312 }, { "epoch": 0.97, "learning_rate": 9.034501685877833e-05, "loss": 3.2608, "step": 12313 }, { "epoch": 0.97, "learning_rate": 9.034423272955384e-05, "loss": 2.9522, "step": 12314 }, { "epoch": 0.97, "learning_rate": 9.034344860032934e-05, "loss": 3.0019, "step": 12315 }, { "epoch": 0.97, "learning_rate": 9.034266447110484e-05, "loss": 3.197, "step": 12316 }, { "epoch": 0.97, "learning_rate": 9.034188034188035e-05, "loss": 2.6885, "step": 12317 }, { "epoch": 0.97, "learning_rate": 9.034109621265585e-05, "loss": 3.1159, "step": 12318 }, { "epoch": 0.97, "learning_rate": 9.034031208343136e-05, "loss": 3.3562, "step": 12319 }, { "epoch": 0.97, "learning_rate": 9.033952795420686e-05, "loss": 2.8327, "step": 12320 }, { "epoch": 0.97, "learning_rate": 9.033874382498236e-05, "loss": 2.754, "step": 12321 }, { "epoch": 0.97, "learning_rate": 9.033795969575787e-05, "loss": 3.0917, "step": 12322 }, { "epoch": 0.97, "learning_rate": 9.033717556653337e-05, "loss": 3.2063, "step": 12323 }, { "epoch": 0.97, "learning_rate": 9.033639143730887e-05, "loss": 3.0432, "step": 12324 }, { "epoch": 0.97, "learning_rate": 9.033560730808438e-05, "loss": 2.9144, "step": 12325 }, { "epoch": 0.97, "learning_rate": 9.033482317885988e-05, "loss": 3.1537, "step": 12326 }, { "epoch": 0.97, "learning_rate": 9.033403904963538e-05, "loss": 3.1167, "step": 12327 }, { "epoch": 0.97, "learning_rate": 9.033325492041089e-05, "loss": 3.1384, "step": 12328 }, { "epoch": 0.97, "learning_rate": 9.033247079118639e-05, "loss": 2.935, "step": 12329 }, { "epoch": 0.97, "learning_rate": 9.03316866619619e-05, "loss": 2.8733, "step": 12330 }, { "epoch": 0.97, "learning_rate": 9.03309025327374e-05, "loss": 2.7058, "step": 12331 }, { "epoch": 0.97, "learning_rate": 9.03301184035129e-05, "loss": 2.9902, "step": 12332 }, { "epoch": 0.97, "learning_rate": 9.032933427428841e-05, "loss": 3.4065, "step": 12333 }, { "epoch": 0.97, "learning_rate": 9.032855014506391e-05, "loss": 3.0498, "step": 12334 }, { "epoch": 0.97, "learning_rate": 9.032776601583941e-05, "loss": 2.8876, "step": 12335 }, { "epoch": 0.97, "learning_rate": 9.032698188661492e-05, "loss": 2.8499, "step": 12336 }, { "epoch": 0.97, "learning_rate": 9.032619775739042e-05, "loss": 3.3757, "step": 12337 }, { "epoch": 0.97, "learning_rate": 9.032541362816593e-05, "loss": 3.1195, "step": 12338 }, { "epoch": 0.97, "learning_rate": 9.032462949894143e-05, "loss": 2.7729, "step": 12339 }, { "epoch": 0.97, "learning_rate": 9.032384536971693e-05, "loss": 3.0373, "step": 12340 }, { "epoch": 0.97, "learning_rate": 9.032306124049244e-05, "loss": 2.9431, "step": 12341 }, { "epoch": 0.97, "learning_rate": 9.032227711126794e-05, "loss": 3.052, "step": 12342 }, { "epoch": 0.97, "learning_rate": 9.032149298204344e-05, "loss": 3.2149, "step": 12343 }, { "epoch": 0.97, "learning_rate": 9.032070885281895e-05, "loss": 2.857, "step": 12344 }, { "epoch": 0.97, "learning_rate": 9.031992472359445e-05, "loss": 2.9747, "step": 12345 }, { "epoch": 0.97, "learning_rate": 9.031914059436995e-05, "loss": 2.9074, "step": 12346 }, { "epoch": 0.97, "learning_rate": 9.031835646514546e-05, "loss": 2.9735, "step": 12347 }, { "epoch": 0.97, "learning_rate": 9.031757233592096e-05, "loss": 3.1003, "step": 12348 }, { "epoch": 0.97, "learning_rate": 9.031678820669647e-05, "loss": 3.1546, "step": 12349 }, { "epoch": 0.97, "learning_rate": 9.031600407747197e-05, "loss": 2.8433, "step": 12350 }, { "epoch": 0.97, "learning_rate": 9.031521994824747e-05, "loss": 2.972, "step": 12351 }, { "epoch": 0.97, "learning_rate": 9.031443581902298e-05, "loss": 3.0649, "step": 12352 }, { "epoch": 0.97, "learning_rate": 9.031365168979848e-05, "loss": 3.2457, "step": 12353 }, { "epoch": 0.97, "learning_rate": 9.031286756057398e-05, "loss": 3.2036, "step": 12354 }, { "epoch": 0.97, "learning_rate": 9.031208343134949e-05, "loss": 2.825, "step": 12355 }, { "epoch": 0.97, "learning_rate": 9.031129930212499e-05, "loss": 2.9279, "step": 12356 }, { "epoch": 0.97, "learning_rate": 9.03105151729005e-05, "loss": 3.2118, "step": 12357 }, { "epoch": 0.97, "learning_rate": 9.0309731043676e-05, "loss": 3.0108, "step": 12358 }, { "epoch": 0.97, "learning_rate": 9.03089469144515e-05, "loss": 3.1335, "step": 12359 }, { "epoch": 0.97, "learning_rate": 9.030816278522702e-05, "loss": 2.9987, "step": 12360 }, { "epoch": 0.97, "learning_rate": 9.030737865600252e-05, "loss": 2.9571, "step": 12361 }, { "epoch": 0.97, "learning_rate": 9.030659452677803e-05, "loss": 3.2749, "step": 12362 }, { "epoch": 0.97, "learning_rate": 9.030581039755353e-05, "loss": 3.0569, "step": 12363 }, { "epoch": 0.97, "learning_rate": 9.030502626832903e-05, "loss": 3.1441, "step": 12364 }, { "epoch": 0.97, "learning_rate": 9.030424213910454e-05, "loss": 2.7854, "step": 12365 }, { "epoch": 0.97, "learning_rate": 9.030345800988004e-05, "loss": 2.9668, "step": 12366 }, { "epoch": 0.97, "learning_rate": 9.030267388065555e-05, "loss": 3.049, "step": 12367 }, { "epoch": 0.97, "learning_rate": 9.030188975143104e-05, "loss": 2.9244, "step": 12368 }, { "epoch": 0.97, "learning_rate": 9.030110562220654e-05, "loss": 2.88, "step": 12369 }, { "epoch": 0.97, "learning_rate": 9.030032149298204e-05, "loss": 3.0951, "step": 12370 }, { "epoch": 0.97, "learning_rate": 9.029953736375755e-05, "loss": 2.863, "step": 12371 }, { "epoch": 0.97, "learning_rate": 9.029875323453305e-05, "loss": 3.0072, "step": 12372 }, { "epoch": 0.97, "learning_rate": 9.029796910530855e-05, "loss": 2.9773, "step": 12373 }, { "epoch": 0.97, "learning_rate": 9.029718497608406e-05, "loss": 2.8266, "step": 12374 }, { "epoch": 0.97, "learning_rate": 9.029640084685956e-05, "loss": 3.272, "step": 12375 }, { "epoch": 0.97, "learning_rate": 9.029561671763506e-05, "loss": 3.1077, "step": 12376 }, { "epoch": 0.97, "learning_rate": 9.029483258841057e-05, "loss": 3.0214, "step": 12377 }, { "epoch": 0.97, "learning_rate": 9.029404845918607e-05, "loss": 3.206, "step": 12378 }, { "epoch": 0.97, "learning_rate": 9.029326432996158e-05, "loss": 3.0062, "step": 12379 }, { "epoch": 0.97, "learning_rate": 9.029248020073708e-05, "loss": 2.7992, "step": 12380 }, { "epoch": 0.97, "learning_rate": 9.029169607151258e-05, "loss": 3.0107, "step": 12381 }, { "epoch": 0.97, "learning_rate": 9.029091194228809e-05, "loss": 3.1706, "step": 12382 }, { "epoch": 0.97, "learning_rate": 9.029012781306359e-05, "loss": 3.0051, "step": 12383 }, { "epoch": 0.97, "learning_rate": 9.02893436838391e-05, "loss": 3.0739, "step": 12384 }, { "epoch": 0.97, "learning_rate": 9.02885595546146e-05, "loss": 3.13, "step": 12385 }, { "epoch": 0.97, "learning_rate": 9.028777542539012e-05, "loss": 2.7099, "step": 12386 }, { "epoch": 0.97, "learning_rate": 9.028699129616562e-05, "loss": 3.0946, "step": 12387 }, { "epoch": 0.97, "learning_rate": 9.028620716694112e-05, "loss": 3.203, "step": 12388 }, { "epoch": 0.97, "learning_rate": 9.028542303771663e-05, "loss": 2.9339, "step": 12389 }, { "epoch": 0.97, "learning_rate": 9.028463890849213e-05, "loss": 2.8022, "step": 12390 }, { "epoch": 0.97, "learning_rate": 9.028385477926763e-05, "loss": 3.2095, "step": 12391 }, { "epoch": 0.97, "learning_rate": 9.028307065004314e-05, "loss": 2.955, "step": 12392 }, { "epoch": 0.97, "learning_rate": 9.028228652081864e-05, "loss": 3.1639, "step": 12393 }, { "epoch": 0.97, "learning_rate": 9.028150239159414e-05, "loss": 3.0359, "step": 12394 }, { "epoch": 0.97, "learning_rate": 9.028071826236965e-05, "loss": 3.2268, "step": 12395 }, { "epoch": 0.97, "learning_rate": 9.027993413314515e-05, "loss": 2.7666, "step": 12396 }, { "epoch": 0.97, "learning_rate": 9.027915000392066e-05, "loss": 3.174, "step": 12397 }, { "epoch": 0.97, "learning_rate": 9.027836587469616e-05, "loss": 3.0165, "step": 12398 }, { "epoch": 0.97, "learning_rate": 9.027758174547166e-05, "loss": 3.0217, "step": 12399 }, { "epoch": 0.97, "learning_rate": 9.027679761624717e-05, "loss": 2.9312, "step": 12400 }, { "epoch": 0.97, "learning_rate": 9.027601348702267e-05, "loss": 3.057, "step": 12401 }, { "epoch": 0.97, "learning_rate": 9.027522935779817e-05, "loss": 3.1433, "step": 12402 }, { "epoch": 0.97, "learning_rate": 9.027444522857368e-05, "loss": 3.0693, "step": 12403 }, { "epoch": 0.97, "learning_rate": 9.027366109934917e-05, "loss": 3.2245, "step": 12404 }, { "epoch": 0.97, "learning_rate": 9.027287697012467e-05, "loss": 3.1055, "step": 12405 }, { "epoch": 0.97, "learning_rate": 9.027209284090018e-05, "loss": 3.0582, "step": 12406 }, { "epoch": 0.97, "learning_rate": 9.027130871167568e-05, "loss": 3.3244, "step": 12407 }, { "epoch": 0.97, "learning_rate": 9.027052458245118e-05, "loss": 3.0161, "step": 12408 }, { "epoch": 0.97, "learning_rate": 9.026974045322669e-05, "loss": 2.938, "step": 12409 }, { "epoch": 0.97, "learning_rate": 9.026895632400219e-05, "loss": 3.0186, "step": 12410 }, { "epoch": 0.97, "learning_rate": 9.02681721947777e-05, "loss": 2.8867, "step": 12411 }, { "epoch": 0.97, "learning_rate": 9.026738806555321e-05, "loss": 2.8988, "step": 12412 }, { "epoch": 0.97, "learning_rate": 9.026660393632871e-05, "loss": 2.8266, "step": 12413 }, { "epoch": 0.97, "learning_rate": 9.026581980710422e-05, "loss": 2.9654, "step": 12414 }, { "epoch": 0.97, "learning_rate": 9.026503567787972e-05, "loss": 2.9055, "step": 12415 }, { "epoch": 0.97, "learning_rate": 9.026425154865523e-05, "loss": 2.9165, "step": 12416 }, { "epoch": 0.97, "learning_rate": 9.026346741943073e-05, "loss": 3.1635, "step": 12417 }, { "epoch": 0.97, "learning_rate": 9.026268329020623e-05, "loss": 2.993, "step": 12418 }, { "epoch": 0.97, "learning_rate": 9.026189916098174e-05, "loss": 2.9935, "step": 12419 }, { "epoch": 0.97, "learning_rate": 9.026111503175724e-05, "loss": 3.164, "step": 12420 }, { "epoch": 0.97, "learning_rate": 9.026033090253274e-05, "loss": 2.761, "step": 12421 }, { "epoch": 0.97, "learning_rate": 9.025954677330825e-05, "loss": 3.1837, "step": 12422 }, { "epoch": 0.97, "learning_rate": 9.025876264408375e-05, "loss": 3.1467, "step": 12423 }, { "epoch": 0.97, "learning_rate": 9.025797851485926e-05, "loss": 3.0805, "step": 12424 }, { "epoch": 0.97, "learning_rate": 9.025719438563476e-05, "loss": 2.8834, "step": 12425 }, { "epoch": 0.97, "learning_rate": 9.025641025641026e-05, "loss": 2.9878, "step": 12426 }, { "epoch": 0.97, "learning_rate": 9.025562612718577e-05, "loss": 3.2744, "step": 12427 }, { "epoch": 0.97, "learning_rate": 9.025484199796127e-05, "loss": 3.0213, "step": 12428 }, { "epoch": 0.97, "learning_rate": 9.025405786873677e-05, "loss": 3.0949, "step": 12429 }, { "epoch": 0.97, "learning_rate": 9.025327373951228e-05, "loss": 2.9219, "step": 12430 }, { "epoch": 0.97, "learning_rate": 9.025248961028778e-05, "loss": 2.8245, "step": 12431 }, { "epoch": 0.97, "learning_rate": 9.025170548106328e-05, "loss": 2.9036, "step": 12432 }, { "epoch": 0.97, "learning_rate": 9.025092135183879e-05, "loss": 3.0596, "step": 12433 }, { "epoch": 0.97, "learning_rate": 9.025013722261429e-05, "loss": 2.8506, "step": 12434 }, { "epoch": 0.98, "learning_rate": 9.02493530933898e-05, "loss": 2.868, "step": 12435 }, { "epoch": 0.98, "learning_rate": 9.02485689641653e-05, "loss": 2.8221, "step": 12436 }, { "epoch": 0.98, "learning_rate": 9.02477848349408e-05, "loss": 3.0649, "step": 12437 }, { "epoch": 0.98, "learning_rate": 9.02470007057163e-05, "loss": 2.8891, "step": 12438 }, { "epoch": 0.98, "learning_rate": 9.024621657649181e-05, "loss": 3.1294, "step": 12439 }, { "epoch": 0.98, "learning_rate": 9.024543244726731e-05, "loss": 3.1094, "step": 12440 }, { "epoch": 0.98, "learning_rate": 9.024464831804282e-05, "loss": 2.8039, "step": 12441 }, { "epoch": 0.98, "learning_rate": 9.024386418881832e-05, "loss": 2.7595, "step": 12442 }, { "epoch": 0.98, "learning_rate": 9.024308005959382e-05, "loss": 2.7791, "step": 12443 }, { "epoch": 0.98, "learning_rate": 9.024229593036933e-05, "loss": 3.1674, "step": 12444 }, { "epoch": 0.98, "learning_rate": 9.024151180114483e-05, "loss": 2.9648, "step": 12445 }, { "epoch": 0.98, "learning_rate": 9.024072767192034e-05, "loss": 2.9423, "step": 12446 }, { "epoch": 0.98, "learning_rate": 9.023994354269584e-05, "loss": 3.0901, "step": 12447 }, { "epoch": 0.98, "learning_rate": 9.023915941347134e-05, "loss": 3.01, "step": 12448 }, { "epoch": 0.98, "learning_rate": 9.023837528424685e-05, "loss": 3.0531, "step": 12449 }, { "epoch": 0.98, "learning_rate": 9.023759115502235e-05, "loss": 3.0437, "step": 12450 }, { "epoch": 0.98, "learning_rate": 9.023680702579785e-05, "loss": 2.9079, "step": 12451 }, { "epoch": 0.98, "learning_rate": 9.023602289657336e-05, "loss": 2.8738, "step": 12452 }, { "epoch": 0.98, "learning_rate": 9.023523876734886e-05, "loss": 2.9201, "step": 12453 }, { "epoch": 0.98, "learning_rate": 9.023445463812437e-05, "loss": 2.802, "step": 12454 }, { "epoch": 0.98, "learning_rate": 9.023367050889987e-05, "loss": 3.0323, "step": 12455 }, { "epoch": 0.98, "learning_rate": 9.023288637967537e-05, "loss": 3.0688, "step": 12456 }, { "epoch": 0.98, "learning_rate": 9.023210225045088e-05, "loss": 3.0691, "step": 12457 }, { "epoch": 0.98, "learning_rate": 9.023131812122638e-05, "loss": 3.2364, "step": 12458 }, { "epoch": 0.98, "learning_rate": 9.023053399200188e-05, "loss": 2.9495, "step": 12459 }, { "epoch": 0.98, "learning_rate": 9.022974986277739e-05, "loss": 2.8944, "step": 12460 }, { "epoch": 0.98, "learning_rate": 9.022896573355289e-05, "loss": 3.1424, "step": 12461 }, { "epoch": 0.98, "learning_rate": 9.02281816043284e-05, "loss": 3.197, "step": 12462 }, { "epoch": 0.98, "learning_rate": 9.02273974751039e-05, "loss": 2.9057, "step": 12463 }, { "epoch": 0.98, "learning_rate": 9.02266133458794e-05, "loss": 2.9845, "step": 12464 }, { "epoch": 0.98, "learning_rate": 9.02258292166549e-05, "loss": 3.1621, "step": 12465 }, { "epoch": 0.98, "learning_rate": 9.022504508743041e-05, "loss": 2.7913, "step": 12466 }, { "epoch": 0.98, "learning_rate": 9.022426095820591e-05, "loss": 3.2358, "step": 12467 }, { "epoch": 0.98, "learning_rate": 9.022347682898142e-05, "loss": 3.0974, "step": 12468 }, { "epoch": 0.98, "learning_rate": 9.022269269975693e-05, "loss": 3.1565, "step": 12469 }, { "epoch": 0.98, "learning_rate": 9.022190857053244e-05, "loss": 2.8269, "step": 12470 }, { "epoch": 0.98, "learning_rate": 9.022112444130794e-05, "loss": 3.3293, "step": 12471 }, { "epoch": 0.98, "learning_rate": 9.022034031208345e-05, "loss": 2.7971, "step": 12472 }, { "epoch": 0.98, "learning_rate": 9.021955618285895e-05, "loss": 2.834, "step": 12473 }, { "epoch": 0.98, "learning_rate": 9.021877205363445e-05, "loss": 3.3259, "step": 12474 }, { "epoch": 0.98, "learning_rate": 9.021798792440996e-05, "loss": 3.0615, "step": 12475 }, { "epoch": 0.98, "learning_rate": 9.021720379518545e-05, "loss": 3.0458, "step": 12476 }, { "epoch": 0.98, "learning_rate": 9.021641966596095e-05, "loss": 3.0254, "step": 12477 }, { "epoch": 0.98, "learning_rate": 9.021563553673645e-05, "loss": 3.0332, "step": 12478 }, { "epoch": 0.98, "learning_rate": 9.021485140751196e-05, "loss": 2.9497, "step": 12479 }, { "epoch": 0.98, "learning_rate": 9.021406727828746e-05, "loss": 2.9409, "step": 12480 }, { "epoch": 0.98, "learning_rate": 9.021328314906296e-05, "loss": 2.8756, "step": 12481 }, { "epoch": 0.98, "learning_rate": 9.021249901983847e-05, "loss": 2.7509, "step": 12482 }, { "epoch": 0.98, "learning_rate": 9.021171489061397e-05, "loss": 2.8738, "step": 12483 }, { "epoch": 0.98, "learning_rate": 9.021093076138948e-05, "loss": 3.0183, "step": 12484 }, { "epoch": 0.98, "learning_rate": 9.021014663216498e-05, "loss": 3.0661, "step": 12485 }, { "epoch": 0.98, "learning_rate": 9.020936250294048e-05, "loss": 3.0791, "step": 12486 }, { "epoch": 0.98, "learning_rate": 9.020857837371599e-05, "loss": 2.9837, "step": 12487 }, { "epoch": 0.98, "learning_rate": 9.020779424449149e-05, "loss": 3.1358, "step": 12488 }, { "epoch": 0.98, "learning_rate": 9.0207010115267e-05, "loss": 2.8862, "step": 12489 }, { "epoch": 0.98, "learning_rate": 9.02062259860425e-05, "loss": 3.1093, "step": 12490 }, { "epoch": 0.98, "learning_rate": 9.0205441856818e-05, "loss": 3.1002, "step": 12491 }, { "epoch": 0.98, "learning_rate": 9.02046577275935e-05, "loss": 3.0048, "step": 12492 }, { "epoch": 0.98, "learning_rate": 9.020387359836901e-05, "loss": 3.0033, "step": 12493 }, { "epoch": 0.98, "learning_rate": 9.020308946914453e-05, "loss": 2.8136, "step": 12494 }, { "epoch": 0.98, "learning_rate": 9.020230533992003e-05, "loss": 3.6089, "step": 12495 }, { "epoch": 0.98, "learning_rate": 9.020152121069553e-05, "loss": 2.9445, "step": 12496 }, { "epoch": 0.98, "learning_rate": 9.020073708147104e-05, "loss": 2.9975, "step": 12497 }, { "epoch": 0.98, "learning_rate": 9.019995295224654e-05, "loss": 2.8701, "step": 12498 }, { "epoch": 0.98, "learning_rate": 9.019916882302204e-05, "loss": 2.9394, "step": 12499 }, { "epoch": 0.98, "learning_rate": 9.019838469379755e-05, "loss": 2.9258, "step": 12500 }, { "epoch": 0.98, "learning_rate": 9.019760056457305e-05, "loss": 2.9598, "step": 12501 }, { "epoch": 0.98, "learning_rate": 9.019681643534856e-05, "loss": 2.9978, "step": 12502 }, { "epoch": 0.98, "learning_rate": 9.019603230612406e-05, "loss": 3.1439, "step": 12503 }, { "epoch": 0.98, "learning_rate": 9.019524817689956e-05, "loss": 3.1476, "step": 12504 }, { "epoch": 0.98, "learning_rate": 9.019446404767507e-05, "loss": 3.1241, "step": 12505 }, { "epoch": 0.98, "learning_rate": 9.019367991845057e-05, "loss": 2.9684, "step": 12506 }, { "epoch": 0.98, "learning_rate": 9.019289578922607e-05, "loss": 3.0617, "step": 12507 }, { "epoch": 0.98, "learning_rate": 9.019211166000158e-05, "loss": 3.0534, "step": 12508 }, { "epoch": 0.98, "learning_rate": 9.019132753077708e-05, "loss": 3.0799, "step": 12509 }, { "epoch": 0.98, "learning_rate": 9.019054340155258e-05, "loss": 2.8418, "step": 12510 }, { "epoch": 0.98, "learning_rate": 9.018975927232809e-05, "loss": 2.9807, "step": 12511 }, { "epoch": 0.98, "learning_rate": 9.018897514310358e-05, "loss": 2.7244, "step": 12512 }, { "epoch": 0.98, "learning_rate": 9.018819101387908e-05, "loss": 3.0694, "step": 12513 }, { "epoch": 0.98, "learning_rate": 9.018740688465459e-05, "loss": 3.0869, "step": 12514 }, { "epoch": 0.98, "learning_rate": 9.018662275543009e-05, "loss": 3.0672, "step": 12515 }, { "epoch": 0.98, "learning_rate": 9.01858386262056e-05, "loss": 2.9959, "step": 12516 }, { "epoch": 0.98, "learning_rate": 9.01850544969811e-05, "loss": 3.276, "step": 12517 }, { "epoch": 0.98, "learning_rate": 9.01842703677566e-05, "loss": 3.35, "step": 12518 }, { "epoch": 0.98, "learning_rate": 9.01834862385321e-05, "loss": 2.888, "step": 12519 }, { "epoch": 0.98, "learning_rate": 9.018270210930762e-05, "loss": 2.9665, "step": 12520 }, { "epoch": 0.98, "learning_rate": 9.018191798008313e-05, "loss": 3.1615, "step": 12521 }, { "epoch": 0.98, "learning_rate": 9.018113385085863e-05, "loss": 2.8898, "step": 12522 }, { "epoch": 0.98, "learning_rate": 9.018034972163413e-05, "loss": 3.0206, "step": 12523 }, { "epoch": 0.98, "learning_rate": 9.017956559240964e-05, "loss": 2.8947, "step": 12524 }, { "epoch": 0.98, "learning_rate": 9.017878146318514e-05, "loss": 2.9686, "step": 12525 }, { "epoch": 0.98, "learning_rate": 9.017799733396064e-05, "loss": 2.993, "step": 12526 }, { "epoch": 0.98, "learning_rate": 9.017721320473615e-05, "loss": 3.2635, "step": 12527 }, { "epoch": 0.98, "learning_rate": 9.017642907551165e-05, "loss": 2.673, "step": 12528 }, { "epoch": 0.98, "learning_rate": 9.017564494628715e-05, "loss": 3.0904, "step": 12529 }, { "epoch": 0.98, "learning_rate": 9.017486081706266e-05, "loss": 3.0626, "step": 12530 }, { "epoch": 0.98, "learning_rate": 9.017407668783816e-05, "loss": 2.9391, "step": 12531 }, { "epoch": 0.98, "learning_rate": 9.017329255861367e-05, "loss": 2.9773, "step": 12532 }, { "epoch": 0.98, "learning_rate": 9.017250842938917e-05, "loss": 2.903, "step": 12533 }, { "epoch": 0.98, "learning_rate": 9.017172430016467e-05, "loss": 2.819, "step": 12534 }, { "epoch": 0.98, "learning_rate": 9.017094017094018e-05, "loss": 2.7922, "step": 12535 }, { "epoch": 0.98, "learning_rate": 9.017015604171568e-05, "loss": 2.9199, "step": 12536 }, { "epoch": 0.98, "learning_rate": 9.016937191249118e-05, "loss": 3.0775, "step": 12537 }, { "epoch": 0.98, "learning_rate": 9.016858778326669e-05, "loss": 3.0109, "step": 12538 }, { "epoch": 0.98, "learning_rate": 9.016780365404219e-05, "loss": 2.8886, "step": 12539 }, { "epoch": 0.98, "learning_rate": 9.01670195248177e-05, "loss": 2.9544, "step": 12540 }, { "epoch": 0.98, "learning_rate": 9.01662353955932e-05, "loss": 3.1585, "step": 12541 }, { "epoch": 0.98, "learning_rate": 9.01654512663687e-05, "loss": 3.0223, "step": 12542 }, { "epoch": 0.98, "learning_rate": 9.01646671371442e-05, "loss": 2.9771, "step": 12543 }, { "epoch": 0.98, "learning_rate": 9.016388300791971e-05, "loss": 3.1016, "step": 12544 }, { "epoch": 0.98, "learning_rate": 9.016309887869521e-05, "loss": 3.0821, "step": 12545 }, { "epoch": 0.98, "learning_rate": 9.016231474947072e-05, "loss": 3.07, "step": 12546 }, { "epoch": 0.98, "learning_rate": 9.016153062024622e-05, "loss": 3.0936, "step": 12547 }, { "epoch": 0.98, "learning_rate": 9.016074649102172e-05, "loss": 2.9622, "step": 12548 }, { "epoch": 0.98, "learning_rate": 9.015996236179723e-05, "loss": 3.2515, "step": 12549 }, { "epoch": 0.98, "learning_rate": 9.015917823257273e-05, "loss": 2.9773, "step": 12550 }, { "epoch": 0.98, "learning_rate": 9.015839410334824e-05, "loss": 3.0562, "step": 12551 }, { "epoch": 0.98, "learning_rate": 9.015760997412374e-05, "loss": 2.8626, "step": 12552 }, { "epoch": 0.98, "learning_rate": 9.015682584489924e-05, "loss": 3.0929, "step": 12553 }, { "epoch": 0.98, "learning_rate": 9.015604171567475e-05, "loss": 2.9935, "step": 12554 }, { "epoch": 0.98, "learning_rate": 9.015525758645025e-05, "loss": 2.9305, "step": 12555 }, { "epoch": 0.98, "learning_rate": 9.015447345722575e-05, "loss": 3.0011, "step": 12556 }, { "epoch": 0.98, "learning_rate": 9.015368932800126e-05, "loss": 3.004, "step": 12557 }, { "epoch": 0.98, "learning_rate": 9.015290519877676e-05, "loss": 3.0742, "step": 12558 }, { "epoch": 0.98, "learning_rate": 9.015212106955227e-05, "loss": 3.132, "step": 12559 }, { "epoch": 0.98, "learning_rate": 9.015133694032777e-05, "loss": 3.0652, "step": 12560 }, { "epoch": 0.98, "learning_rate": 9.015055281110327e-05, "loss": 3.3007, "step": 12561 }, { "epoch": 0.99, "learning_rate": 9.014976868187878e-05, "loss": 3.0854, "step": 12562 }, { "epoch": 0.99, "learning_rate": 9.014898455265428e-05, "loss": 2.9347, "step": 12563 }, { "epoch": 0.99, "learning_rate": 9.014820042342978e-05, "loss": 3.1213, "step": 12564 }, { "epoch": 0.99, "learning_rate": 9.014741629420529e-05, "loss": 3.1391, "step": 12565 }, { "epoch": 0.99, "learning_rate": 9.014663216498079e-05, "loss": 3.1715, "step": 12566 }, { "epoch": 0.99, "learning_rate": 9.01458480357563e-05, "loss": 2.831, "step": 12567 }, { "epoch": 0.99, "learning_rate": 9.01450639065318e-05, "loss": 2.9506, "step": 12568 }, { "epoch": 0.99, "learning_rate": 9.01442797773073e-05, "loss": 3.0779, "step": 12569 }, { "epoch": 0.99, "learning_rate": 9.01434956480828e-05, "loss": 2.89, "step": 12570 }, { "epoch": 0.99, "learning_rate": 9.014271151885831e-05, "loss": 3.0737, "step": 12571 }, { "epoch": 0.99, "learning_rate": 9.014192738963381e-05, "loss": 2.9195, "step": 12572 }, { "epoch": 0.99, "learning_rate": 9.014114326040932e-05, "loss": 2.8993, "step": 12573 }, { "epoch": 0.99, "learning_rate": 9.014035913118482e-05, "loss": 3.0994, "step": 12574 }, { "epoch": 0.99, "learning_rate": 9.013957500196032e-05, "loss": 2.9243, "step": 12575 }, { "epoch": 0.99, "learning_rate": 9.013879087273583e-05, "loss": 2.782, "step": 12576 }, { "epoch": 0.99, "learning_rate": 9.013800674351134e-05, "loss": 3.1108, "step": 12577 }, { "epoch": 0.99, "learning_rate": 9.013722261428685e-05, "loss": 3.0199, "step": 12578 }, { "epoch": 0.99, "learning_rate": 9.013643848506235e-05, "loss": 3.1374, "step": 12579 }, { "epoch": 0.99, "learning_rate": 9.013565435583786e-05, "loss": 3.0146, "step": 12580 }, { "epoch": 0.99, "learning_rate": 9.013487022661336e-05, "loss": 2.9971, "step": 12581 }, { "epoch": 0.99, "learning_rate": 9.013408609738886e-05, "loss": 3.4812, "step": 12582 }, { "epoch": 0.99, "learning_rate": 9.013330196816437e-05, "loss": 2.7141, "step": 12583 }, { "epoch": 0.99, "learning_rate": 9.013251783893986e-05, "loss": 2.8841, "step": 12584 }, { "epoch": 0.99, "learning_rate": 9.013173370971536e-05, "loss": 2.9146, "step": 12585 }, { "epoch": 0.99, "learning_rate": 9.013094958049086e-05, "loss": 2.9994, "step": 12586 }, { "epoch": 0.99, "learning_rate": 9.013016545126637e-05, "loss": 3.0633, "step": 12587 }, { "epoch": 0.99, "learning_rate": 9.012938132204187e-05, "loss": 3.1493, "step": 12588 }, { "epoch": 0.99, "learning_rate": 9.012859719281738e-05, "loss": 3.0823, "step": 12589 }, { "epoch": 0.99, "learning_rate": 9.012781306359288e-05, "loss": 2.9445, "step": 12590 }, { "epoch": 0.99, "learning_rate": 9.012702893436838e-05, "loss": 3.0999, "step": 12591 }, { "epoch": 0.99, "learning_rate": 9.012624480514389e-05, "loss": 2.9142, "step": 12592 }, { "epoch": 0.99, "learning_rate": 9.012546067591939e-05, "loss": 3.0881, "step": 12593 }, { "epoch": 0.99, "learning_rate": 9.01246765466949e-05, "loss": 3.1497, "step": 12594 }, { "epoch": 0.99, "learning_rate": 9.01238924174704e-05, "loss": 2.8692, "step": 12595 }, { "epoch": 0.99, "learning_rate": 9.01231082882459e-05, "loss": 3.1447, "step": 12596 }, { "epoch": 0.99, "learning_rate": 9.01223241590214e-05, "loss": 2.924, "step": 12597 }, { "epoch": 0.99, "learning_rate": 9.012154002979691e-05, "loss": 2.9719, "step": 12598 }, { "epoch": 0.99, "learning_rate": 9.012075590057241e-05, "loss": 2.9941, "step": 12599 }, { "epoch": 0.99, "learning_rate": 9.011997177134792e-05, "loss": 3.0212, "step": 12600 }, { "epoch": 0.99, "learning_rate": 9.011918764212342e-05, "loss": 2.9348, "step": 12601 }, { "epoch": 0.99, "learning_rate": 9.011840351289894e-05, "loss": 2.8049, "step": 12602 }, { "epoch": 0.99, "learning_rate": 9.011761938367444e-05, "loss": 3.348, "step": 12603 }, { "epoch": 0.99, "learning_rate": 9.011683525444994e-05, "loss": 2.9238, "step": 12604 }, { "epoch": 0.99, "learning_rate": 9.011605112522545e-05, "loss": 3.1608, "step": 12605 }, { "epoch": 0.99, "learning_rate": 9.011526699600095e-05, "loss": 3.1899, "step": 12606 }, { "epoch": 0.99, "learning_rate": 9.011448286677646e-05, "loss": 2.8157, "step": 12607 }, { "epoch": 0.99, "learning_rate": 9.011369873755196e-05, "loss": 2.8379, "step": 12608 }, { "epoch": 0.99, "learning_rate": 9.011291460832746e-05, "loss": 2.9701, "step": 12609 }, { "epoch": 0.99, "learning_rate": 9.011213047910297e-05, "loss": 2.9449, "step": 12610 }, { "epoch": 0.99, "learning_rate": 9.011134634987847e-05, "loss": 3.1434, "step": 12611 }, { "epoch": 0.99, "learning_rate": 9.011056222065397e-05, "loss": 2.9246, "step": 12612 }, { "epoch": 0.99, "learning_rate": 9.010977809142948e-05, "loss": 2.9712, "step": 12613 }, { "epoch": 0.99, "learning_rate": 9.010899396220498e-05, "loss": 2.9017, "step": 12614 }, { "epoch": 0.99, "learning_rate": 9.010820983298048e-05, "loss": 3.0729, "step": 12615 }, { "epoch": 0.99, "learning_rate": 9.010742570375599e-05, "loss": 2.9986, "step": 12616 }, { "epoch": 0.99, "learning_rate": 9.010664157453149e-05, "loss": 2.9869, "step": 12617 }, { "epoch": 0.99, "learning_rate": 9.0105857445307e-05, "loss": 3.0557, "step": 12618 }, { "epoch": 0.99, "learning_rate": 9.01050733160825e-05, "loss": 3.2104, "step": 12619 }, { "epoch": 0.99, "learning_rate": 9.010428918685799e-05, "loss": 3.0272, "step": 12620 }, { "epoch": 0.99, "learning_rate": 9.010350505763349e-05, "loss": 2.7721, "step": 12621 }, { "epoch": 0.99, "learning_rate": 9.0102720928409e-05, "loss": 2.9552, "step": 12622 }, { "epoch": 0.99, "learning_rate": 9.01019367991845e-05, "loss": 3.0674, "step": 12623 }, { "epoch": 0.99, "learning_rate": 9.010115266996e-05, "loss": 3.1895, "step": 12624 }, { "epoch": 0.99, "learning_rate": 9.010036854073551e-05, "loss": 3.1064, "step": 12625 }, { "epoch": 0.99, "learning_rate": 9.009958441151101e-05, "loss": 3.0113, "step": 12626 }, { "epoch": 0.99, "learning_rate": 9.009880028228652e-05, "loss": 2.9503, "step": 12627 }, { "epoch": 0.99, "learning_rate": 9.009801615306203e-05, "loss": 2.8193, "step": 12628 }, { "epoch": 0.99, "learning_rate": 9.009723202383754e-05, "loss": 3.1456, "step": 12629 }, { "epoch": 0.99, "learning_rate": 9.009644789461304e-05, "loss": 2.9256, "step": 12630 }, { "epoch": 0.99, "learning_rate": 9.009566376538854e-05, "loss": 3.2488, "step": 12631 }, { "epoch": 0.99, "learning_rate": 9.009487963616405e-05, "loss": 2.9461, "step": 12632 }, { "epoch": 0.99, "learning_rate": 9.009409550693955e-05, "loss": 3.2369, "step": 12633 }, { "epoch": 0.99, "learning_rate": 9.009331137771505e-05, "loss": 3.2125, "step": 12634 }, { "epoch": 0.99, "learning_rate": 9.009252724849056e-05, "loss": 2.8117, "step": 12635 }, { "epoch": 0.99, "learning_rate": 9.009174311926606e-05, "loss": 2.9081, "step": 12636 }, { "epoch": 0.99, "learning_rate": 9.009095899004157e-05, "loss": 2.9546, "step": 12637 }, { "epoch": 0.99, "learning_rate": 9.009017486081707e-05, "loss": 3.1706, "step": 12638 }, { "epoch": 0.99, "learning_rate": 9.008939073159257e-05, "loss": 2.8555, "step": 12639 }, { "epoch": 0.99, "learning_rate": 9.008860660236808e-05, "loss": 2.9883, "step": 12640 }, { "epoch": 0.99, "learning_rate": 9.008782247314358e-05, "loss": 2.9242, "step": 12641 }, { "epoch": 0.99, "learning_rate": 9.008703834391908e-05, "loss": 3.2388, "step": 12642 }, { "epoch": 0.99, "learning_rate": 9.008625421469459e-05, "loss": 3.0814, "step": 12643 }, { "epoch": 0.99, "learning_rate": 9.008547008547009e-05, "loss": 2.8528, "step": 12644 }, { "epoch": 0.99, "learning_rate": 9.00846859562456e-05, "loss": 3.074, "step": 12645 }, { "epoch": 0.99, "learning_rate": 9.00839018270211e-05, "loss": 3.1318, "step": 12646 }, { "epoch": 0.99, "learning_rate": 9.00831176977966e-05, "loss": 2.9055, "step": 12647 }, { "epoch": 0.99, "learning_rate": 9.00823335685721e-05, "loss": 3.1296, "step": 12648 }, { "epoch": 0.99, "learning_rate": 9.008154943934761e-05, "loss": 3.0189, "step": 12649 }, { "epoch": 0.99, "learning_rate": 9.008076531012311e-05, "loss": 2.9723, "step": 12650 }, { "epoch": 0.99, "learning_rate": 9.007998118089862e-05, "loss": 3.0358, "step": 12651 }, { "epoch": 0.99, "learning_rate": 9.007919705167412e-05, "loss": 2.9087, "step": 12652 }, { "epoch": 0.99, "learning_rate": 9.007841292244962e-05, "loss": 2.9709, "step": 12653 }, { "epoch": 0.99, "learning_rate": 9.007762879322513e-05, "loss": 3.0651, "step": 12654 }, { "epoch": 0.99, "learning_rate": 9.007684466400063e-05, "loss": 3.2491, "step": 12655 }, { "epoch": 0.99, "learning_rate": 9.007606053477614e-05, "loss": 2.945, "step": 12656 }, { "epoch": 0.99, "learning_rate": 9.007527640555164e-05, "loss": 3.1299, "step": 12657 }, { "epoch": 0.99, "learning_rate": 9.007449227632714e-05, "loss": 3.0789, "step": 12658 }, { "epoch": 0.99, "learning_rate": 9.007370814710265e-05, "loss": 2.9751, "step": 12659 }, { "epoch": 0.99, "learning_rate": 9.007292401787815e-05, "loss": 3.0634, "step": 12660 }, { "epoch": 0.99, "learning_rate": 9.007213988865365e-05, "loss": 2.8765, "step": 12661 }, { "epoch": 0.99, "learning_rate": 9.007135575942916e-05, "loss": 2.9168, "step": 12662 }, { "epoch": 0.99, "learning_rate": 9.007057163020466e-05, "loss": 3.1639, "step": 12663 }, { "epoch": 0.99, "learning_rate": 9.006978750098016e-05, "loss": 2.8119, "step": 12664 }, { "epoch": 0.99, "learning_rate": 9.006900337175567e-05, "loss": 3.129, "step": 12665 }, { "epoch": 0.99, "learning_rate": 9.006821924253117e-05, "loss": 2.9023, "step": 12666 }, { "epoch": 0.99, "learning_rate": 9.006743511330668e-05, "loss": 3.0663, "step": 12667 }, { "epoch": 0.99, "learning_rate": 9.006665098408218e-05, "loss": 3.0041, "step": 12668 }, { "epoch": 0.99, "learning_rate": 9.006586685485768e-05, "loss": 3.0112, "step": 12669 }, { "epoch": 0.99, "learning_rate": 9.006508272563319e-05, "loss": 2.811, "step": 12670 }, { "epoch": 0.99, "learning_rate": 9.006429859640869e-05, "loss": 2.8129, "step": 12671 }, { "epoch": 0.99, "learning_rate": 9.00635144671842e-05, "loss": 3.1809, "step": 12672 }, { "epoch": 0.99, "learning_rate": 9.00627303379597e-05, "loss": 3.2455, "step": 12673 }, { "epoch": 0.99, "learning_rate": 9.00619462087352e-05, "loss": 2.9738, "step": 12674 }, { "epoch": 0.99, "learning_rate": 9.00611620795107e-05, "loss": 3.1344, "step": 12675 }, { "epoch": 0.99, "learning_rate": 9.006037795028621e-05, "loss": 2.967, "step": 12676 }, { "epoch": 0.99, "learning_rate": 9.005959382106171e-05, "loss": 3.0633, "step": 12677 }, { "epoch": 0.99, "learning_rate": 9.005880969183722e-05, "loss": 2.98, "step": 12678 }, { "epoch": 0.99, "learning_rate": 9.005802556261272e-05, "loss": 2.9875, "step": 12679 }, { "epoch": 0.99, "learning_rate": 9.005724143338822e-05, "loss": 2.9887, "step": 12680 }, { "epoch": 0.99, "learning_rate": 9.005645730416373e-05, "loss": 2.8842, "step": 12681 }, { "epoch": 0.99, "learning_rate": 9.005567317493923e-05, "loss": 3.1074, "step": 12682 }, { "epoch": 0.99, "learning_rate": 9.005488904571473e-05, "loss": 2.9004, "step": 12683 }, { "epoch": 0.99, "learning_rate": 9.005410491649024e-05, "loss": 2.7701, "step": 12684 }, { "epoch": 0.99, "learning_rate": 9.005332078726576e-05, "loss": 3.0217, "step": 12685 }, { "epoch": 0.99, "learning_rate": 9.005253665804126e-05, "loss": 2.7983, "step": 12686 }, { "epoch": 0.99, "learning_rate": 9.005175252881676e-05, "loss": 3.0314, "step": 12687 }, { "epoch": 0.99, "learning_rate": 9.005096839959227e-05, "loss": 3.1097, "step": 12688 }, { "epoch": 0.99, "learning_rate": 9.005018427036777e-05, "loss": 2.8359, "step": 12689 }, { "epoch": 1.0, "learning_rate": 9.004940014114327e-05, "loss": 3.0227, "step": 12690 }, { "epoch": 1.0, "learning_rate": 9.004861601191878e-05, "loss": 3.2281, "step": 12691 }, { "epoch": 1.0, "learning_rate": 9.004783188269427e-05, "loss": 3.1144, "step": 12692 }, { "epoch": 1.0, "learning_rate": 9.004704775346977e-05, "loss": 2.7872, "step": 12693 }, { "epoch": 1.0, "learning_rate": 9.004626362424528e-05, "loss": 2.8478, "step": 12694 }, { "epoch": 1.0, "learning_rate": 9.004547949502078e-05, "loss": 3.2607, "step": 12695 }, { "epoch": 1.0, "learning_rate": 9.004469536579628e-05, "loss": 2.8162, "step": 12696 }, { "epoch": 1.0, "learning_rate": 9.004391123657179e-05, "loss": 2.9595, "step": 12697 }, { "epoch": 1.0, "learning_rate": 9.004312710734729e-05, "loss": 3.1805, "step": 12698 }, { "epoch": 1.0, "learning_rate": 9.00423429781228e-05, "loss": 2.8524, "step": 12699 }, { "epoch": 1.0, "learning_rate": 9.00415588488983e-05, "loss": 2.9274, "step": 12700 }, { "epoch": 1.0, "learning_rate": 9.00407747196738e-05, "loss": 3.0503, "step": 12701 }, { "epoch": 1.0, "learning_rate": 9.00399905904493e-05, "loss": 3.0115, "step": 12702 }, { "epoch": 1.0, "learning_rate": 9.003920646122481e-05, "loss": 2.8638, "step": 12703 }, { "epoch": 1.0, "learning_rate": 9.003842233200031e-05, "loss": 3.2081, "step": 12704 }, { "epoch": 1.0, "learning_rate": 9.003763820277582e-05, "loss": 3.1363, "step": 12705 }, { "epoch": 1.0, "learning_rate": 9.003685407355132e-05, "loss": 2.8411, "step": 12706 }, { "epoch": 1.0, "learning_rate": 9.003606994432682e-05, "loss": 3.3842, "step": 12707 }, { "epoch": 1.0, "learning_rate": 9.003528581510233e-05, "loss": 2.8827, "step": 12708 }, { "epoch": 1.0, "learning_rate": 9.003450168587783e-05, "loss": 3.0155, "step": 12709 }, { "epoch": 1.0, "learning_rate": 9.003371755665335e-05, "loss": 3.0749, "step": 12710 }, { "epoch": 1.0, "learning_rate": 9.003293342742885e-05, "loss": 3.0104, "step": 12711 }, { "epoch": 1.0, "learning_rate": 9.003214929820435e-05, "loss": 2.9523, "step": 12712 }, { "epoch": 1.0, "learning_rate": 9.003136516897986e-05, "loss": 3.1147, "step": 12713 }, { "epoch": 1.0, "learning_rate": 9.003058103975536e-05, "loss": 2.9872, "step": 12714 }, { "epoch": 1.0, "learning_rate": 9.002979691053087e-05, "loss": 3.4499, "step": 12715 }, { "epoch": 1.0, "learning_rate": 9.002901278130637e-05, "loss": 3.1863, "step": 12716 }, { "epoch": 1.0, "learning_rate": 9.002822865208187e-05, "loss": 2.9777, "step": 12717 }, { "epoch": 1.0, "learning_rate": 9.002744452285738e-05, "loss": 2.8908, "step": 12718 }, { "epoch": 1.0, "learning_rate": 9.002666039363288e-05, "loss": 3.0689, "step": 12719 }, { "epoch": 1.0, "learning_rate": 9.002587626440838e-05, "loss": 2.8301, "step": 12720 }, { "epoch": 1.0, "learning_rate": 9.002509213518389e-05, "loss": 3.1832, "step": 12721 }, { "epoch": 1.0, "learning_rate": 9.002430800595939e-05, "loss": 3.0444, "step": 12722 }, { "epoch": 1.0, "learning_rate": 9.00235238767349e-05, "loss": 2.934, "step": 12723 }, { "epoch": 1.0, "learning_rate": 9.00227397475104e-05, "loss": 3.14, "step": 12724 }, { "epoch": 1.0, "learning_rate": 9.00219556182859e-05, "loss": 3.159, "step": 12725 }, { "epoch": 1.0, "learning_rate": 9.00211714890614e-05, "loss": 3.0513, "step": 12726 }, { "epoch": 1.0, "learning_rate": 9.002038735983691e-05, "loss": 3.134, "step": 12727 }, { "epoch": 1.0, "learning_rate": 9.00196032306124e-05, "loss": 3.4524, "step": 12728 }, { "epoch": 1.0, "learning_rate": 9.00188191013879e-05, "loss": 2.8263, "step": 12729 }, { "epoch": 1.0, "learning_rate": 9.001803497216341e-05, "loss": 3.1473, "step": 12730 }, { "epoch": 1.0, "learning_rate": 9.001725084293891e-05, "loss": 2.9752, "step": 12731 }, { "epoch": 1.0, "learning_rate": 9.001646671371441e-05, "loss": 2.9523, "step": 12732 }, { "epoch": 1.0, "learning_rate": 9.001568258448992e-05, "loss": 2.8991, "step": 12733 }, { "epoch": 1.0, "learning_rate": 9.001489845526542e-05, "loss": 2.9519, "step": 12734 }, { "epoch": 1.0, "learning_rate": 9.001411432604093e-05, "loss": 3.1092, "step": 12735 }, { "epoch": 1.0, "learning_rate": 9.001333019681644e-05, "loss": 3.2257, "step": 12736 }, { "epoch": 1.0, "learning_rate": 9.001254606759195e-05, "loss": 3.1224, "step": 12737 }, { "epoch": 1.0, "learning_rate": 9.001176193836745e-05, "loss": 2.7543, "step": 12738 }, { "epoch": 1.0, "learning_rate": 9.001097780914295e-05, "loss": 2.9376, "step": 12739 }, { "epoch": 1.0, "learning_rate": 9.001019367991846e-05, "loss": 3.0766, "step": 12740 }, { "epoch": 1.0, "learning_rate": 9.000940955069396e-05, "loss": 2.9562, "step": 12741 }, { "epoch": 1.0, "learning_rate": 9.000862542146947e-05, "loss": 3.0007, "step": 12742 }, { "epoch": 1.0, "learning_rate": 9.000784129224497e-05, "loss": 2.7549, "step": 12743 }, { "epoch": 1.0, "learning_rate": 9.000705716302047e-05, "loss": 3.1382, "step": 12744 }, { "epoch": 1.0, "learning_rate": 9.000627303379598e-05, "loss": 2.8889, "step": 12745 }, { "epoch": 1.0, "learning_rate": 9.000548890457148e-05, "loss": 3.1637, "step": 12746 }, { "epoch": 1.0, "learning_rate": 9.000470477534698e-05, "loss": 3.1415, "step": 12747 }, { "epoch": 1.0, "learning_rate": 9.000392064612249e-05, "loss": 2.8302, "step": 12748 }, { "epoch": 1.0, "learning_rate": 9.000313651689799e-05, "loss": 3.0335, "step": 12749 }, { "epoch": 1.0, "learning_rate": 9.00023523876735e-05, "loss": 2.9357, "step": 12750 }, { "epoch": 1.0, "learning_rate": 9.0001568258449e-05, "loss": 3.169, "step": 12751 }, { "epoch": 1.0, "learning_rate": 9.00007841292245e-05, "loss": 3.1678, "step": 12752 }, { "epoch": 1.0, "learning_rate": 9e-05, "loss": 2.7794, "step": 12753 }, { "epoch": 1.0, "learning_rate": 8.999921587077551e-05, "loss": 2.9567, "step": 12754 }, { "epoch": 1.0, "learning_rate": 8.999843174155101e-05, "loss": 3.0013, "step": 12755 }, { "epoch": 1.0, "learning_rate": 8.999764761232652e-05, "loss": 3.0699, "step": 12756 }, { "epoch": 1.0, "learning_rate": 8.999686348310202e-05, "loss": 3.0284, "step": 12757 }, { "epoch": 1.0, "learning_rate": 8.999607935387752e-05, "loss": 2.651, "step": 12758 }, { "epoch": 1.0, "learning_rate": 8.999529522465303e-05, "loss": 2.6559, "step": 12759 }, { "epoch": 1.0, "learning_rate": 8.999451109542853e-05, "loss": 2.8556, "step": 12760 }, { "epoch": 1.0, "learning_rate": 8.999372696620404e-05, "loss": 2.8558, "step": 12761 }, { "epoch": 1.0, "learning_rate": 8.999294283697954e-05, "loss": 2.6642, "step": 12762 }, { "epoch": 1.0, "learning_rate": 8.999215870775504e-05, "loss": 3.1007, "step": 12763 }, { "epoch": 1.0, "learning_rate": 8.999137457853055e-05, "loss": 2.9664, "step": 12764 }, { "epoch": 1.0, "learning_rate": 8.999059044930605e-05, "loss": 2.932, "step": 12765 }, { "epoch": 1.0, "learning_rate": 8.998980632008155e-05, "loss": 2.9597, "step": 12766 }, { "epoch": 1.0, "learning_rate": 8.998902219085706e-05, "loss": 2.7835, "step": 12767 }, { "epoch": 1.0, "learning_rate": 8.998823806163256e-05, "loss": 2.7091, "step": 12768 }, { "epoch": 1.0, "learning_rate": 8.998745393240806e-05, "loss": 3.0782, "step": 12769 }, { "epoch": 1.0, "learning_rate": 8.998666980318357e-05, "loss": 2.9513, "step": 12770 }, { "epoch": 1.0, "learning_rate": 8.998588567395907e-05, "loss": 2.9424, "step": 12771 }, { "epoch": 1.0, "learning_rate": 8.998510154473458e-05, "loss": 2.8268, "step": 12772 }, { "epoch": 1.0, "learning_rate": 8.998431741551008e-05, "loss": 3.0668, "step": 12773 }, { "epoch": 1.0, "learning_rate": 8.998353328628558e-05, "loss": 2.7453, "step": 12774 }, { "epoch": 1.0, "learning_rate": 8.998274915706109e-05, "loss": 2.8293, "step": 12775 }, { "epoch": 1.0, "learning_rate": 8.998196502783659e-05, "loss": 3.1747, "step": 12776 }, { "epoch": 1.0, "learning_rate": 8.99811808986121e-05, "loss": 2.9512, "step": 12777 }, { "epoch": 1.0, "learning_rate": 8.99803967693876e-05, "loss": 2.767, "step": 12778 }, { "epoch": 1.0, "learning_rate": 8.99796126401631e-05, "loss": 2.7915, "step": 12779 }, { "epoch": 1.0, "learning_rate": 8.99788285109386e-05, "loss": 2.9529, "step": 12780 }, { "epoch": 1.0, "learning_rate": 8.997804438171411e-05, "loss": 2.936, "step": 12781 }, { "epoch": 1.0, "learning_rate": 8.997726025248961e-05, "loss": 2.8468, "step": 12782 }, { "epoch": 1.0, "learning_rate": 8.997647612326512e-05, "loss": 2.757, "step": 12783 }, { "epoch": 1.0, "learning_rate": 8.997569199404062e-05, "loss": 2.8626, "step": 12784 }, { "epoch": 1.0, "learning_rate": 8.997490786481612e-05, "loss": 2.8067, "step": 12785 }, { "epoch": 1.0, "learning_rate": 8.997412373559163e-05, "loss": 2.9357, "step": 12786 }, { "epoch": 1.0, "learning_rate": 8.997333960636713e-05, "loss": 2.8196, "step": 12787 }, { "epoch": 1.0, "learning_rate": 8.997255547714263e-05, "loss": 3.0387, "step": 12788 }, { "epoch": 1.0, "learning_rate": 8.997177134791814e-05, "loss": 2.6808, "step": 12789 }, { "epoch": 1.0, "learning_rate": 8.997098721869364e-05, "loss": 2.8824, "step": 12790 }, { "epoch": 1.0, "learning_rate": 8.997020308946915e-05, "loss": 2.8818, "step": 12791 }, { "epoch": 1.0, "learning_rate": 8.996941896024465e-05, "loss": 3.0826, "step": 12792 }, { "epoch": 1.0, "learning_rate": 8.996863483102017e-05, "loss": 2.754, "step": 12793 }, { "epoch": 1.0, "learning_rate": 8.996785070179567e-05, "loss": 2.9581, "step": 12794 }, { "epoch": 1.0, "learning_rate": 8.996706657257117e-05, "loss": 2.7607, "step": 12795 }, { "epoch": 1.0, "learning_rate": 8.996628244334668e-05, "loss": 2.8373, "step": 12796 }, { "epoch": 1.0, "learning_rate": 8.996549831412218e-05, "loss": 2.9935, "step": 12797 }, { "epoch": 1.0, "learning_rate": 8.996471418489768e-05, "loss": 2.792, "step": 12798 }, { "epoch": 1.0, "learning_rate": 8.996393005567319e-05, "loss": 2.8898, "step": 12799 }, { "epoch": 1.0, "learning_rate": 8.996314592644868e-05, "loss": 2.6822, "step": 12800 }, { "epoch": 1.0, "learning_rate": 8.996236179722418e-05, "loss": 2.9535, "step": 12801 }, { "epoch": 1.0, "learning_rate": 8.996157766799969e-05, "loss": 2.9745, "step": 12802 }, { "epoch": 1.0, "learning_rate": 8.996079353877519e-05, "loss": 2.8671, "step": 12803 }, { "epoch": 1.0, "learning_rate": 8.99600094095507e-05, "loss": 2.9349, "step": 12804 }, { "epoch": 1.0, "learning_rate": 8.99592252803262e-05, "loss": 2.8784, "step": 12805 }, { "epoch": 1.0, "learning_rate": 8.99584411511017e-05, "loss": 2.8938, "step": 12806 }, { "epoch": 1.0, "learning_rate": 8.99576570218772e-05, "loss": 2.636, "step": 12807 }, { "epoch": 1.0, "learning_rate": 8.995687289265271e-05, "loss": 2.7484, "step": 12808 }, { "epoch": 1.0, "learning_rate": 8.995608876342821e-05, "loss": 2.9677, "step": 12809 }, { "epoch": 1.0, "learning_rate": 8.995530463420372e-05, "loss": 2.9082, "step": 12810 }, { "epoch": 1.0, "learning_rate": 8.995452050497922e-05, "loss": 2.6595, "step": 12811 }, { "epoch": 1.0, "learning_rate": 8.995373637575472e-05, "loss": 2.9773, "step": 12812 }, { "epoch": 1.0, "learning_rate": 8.995295224653023e-05, "loss": 2.7824, "step": 12813 }, { "epoch": 1.0, "learning_rate": 8.995216811730573e-05, "loss": 2.7683, "step": 12814 }, { "epoch": 1.0, "learning_rate": 8.995138398808123e-05, "loss": 2.7908, "step": 12815 }, { "epoch": 1.0, "learning_rate": 8.995059985885674e-05, "loss": 2.799, "step": 12816 }, { "epoch": 1.01, "learning_rate": 8.994981572963224e-05, "loss": 2.9698, "step": 12817 }, { "epoch": 1.01, "learning_rate": 8.994903160040776e-05, "loss": 2.8732, "step": 12818 }, { "epoch": 1.01, "learning_rate": 8.994824747118326e-05, "loss": 2.8272, "step": 12819 }, { "epoch": 1.01, "learning_rate": 8.994746334195877e-05, "loss": 2.8629, "step": 12820 }, { "epoch": 1.01, "learning_rate": 8.994667921273427e-05, "loss": 2.8002, "step": 12821 }, { "epoch": 1.01, "learning_rate": 8.994589508350977e-05, "loss": 2.9841, "step": 12822 }, { "epoch": 1.01, "learning_rate": 8.994511095428528e-05, "loss": 2.9237, "step": 12823 }, { "epoch": 1.01, "learning_rate": 8.994432682506078e-05, "loss": 2.9038, "step": 12824 }, { "epoch": 1.01, "learning_rate": 8.994354269583628e-05, "loss": 3.0217, "step": 12825 }, { "epoch": 1.01, "learning_rate": 8.994275856661179e-05, "loss": 2.8464, "step": 12826 }, { "epoch": 1.01, "learning_rate": 8.994197443738729e-05, "loss": 3.0808, "step": 12827 }, { "epoch": 1.01, "learning_rate": 8.99411903081628e-05, "loss": 2.7676, "step": 12828 }, { "epoch": 1.01, "learning_rate": 8.99404061789383e-05, "loss": 2.9627, "step": 12829 }, { "epoch": 1.01, "learning_rate": 8.99396220497138e-05, "loss": 2.751, "step": 12830 }, { "epoch": 1.01, "learning_rate": 8.99388379204893e-05, "loss": 2.8594, "step": 12831 }, { "epoch": 1.01, "learning_rate": 8.993805379126481e-05, "loss": 2.7846, "step": 12832 }, { "epoch": 1.01, "learning_rate": 8.993726966204031e-05, "loss": 2.985, "step": 12833 }, { "epoch": 1.01, "learning_rate": 8.993648553281582e-05, "loss": 3.0507, "step": 12834 }, { "epoch": 1.01, "learning_rate": 8.993570140359132e-05, "loss": 2.7314, "step": 12835 }, { "epoch": 1.01, "learning_rate": 8.993491727436681e-05, "loss": 2.9712, "step": 12836 }, { "epoch": 1.01, "learning_rate": 8.993413314514231e-05, "loss": 3.0394, "step": 12837 }, { "epoch": 1.01, "learning_rate": 8.993334901591782e-05, "loss": 2.7016, "step": 12838 }, { "epoch": 1.01, "learning_rate": 8.993256488669332e-05, "loss": 3.0636, "step": 12839 }, { "epoch": 1.01, "learning_rate": 8.993178075746883e-05, "loss": 3.0718, "step": 12840 }, { "epoch": 1.01, "learning_rate": 8.993099662824433e-05, "loss": 2.8828, "step": 12841 }, { "epoch": 1.01, "learning_rate": 8.993021249901983e-05, "loss": 2.9846, "step": 12842 }, { "epoch": 1.01, "learning_rate": 8.992942836979534e-05, "loss": 2.9164, "step": 12843 }, { "epoch": 1.01, "learning_rate": 8.992864424057085e-05, "loss": 2.9171, "step": 12844 }, { "epoch": 1.01, "learning_rate": 8.992786011134636e-05, "loss": 3.0522, "step": 12845 }, { "epoch": 1.01, "learning_rate": 8.992707598212186e-05, "loss": 2.9138, "step": 12846 }, { "epoch": 1.01, "learning_rate": 8.992629185289736e-05, "loss": 3.1098, "step": 12847 }, { "epoch": 1.01, "learning_rate": 8.992550772367287e-05, "loss": 2.7956, "step": 12848 }, { "epoch": 1.01, "learning_rate": 8.992472359444837e-05, "loss": 2.8425, "step": 12849 }, { "epoch": 1.01, "learning_rate": 8.992393946522388e-05, "loss": 2.8622, "step": 12850 }, { "epoch": 1.01, "learning_rate": 8.992315533599938e-05, "loss": 2.9697, "step": 12851 }, { "epoch": 1.01, "learning_rate": 8.992237120677488e-05, "loss": 3.0015, "step": 12852 }, { "epoch": 1.01, "learning_rate": 8.992158707755039e-05, "loss": 2.7009, "step": 12853 }, { "epoch": 1.01, "learning_rate": 8.992080294832589e-05, "loss": 2.8709, "step": 12854 }, { "epoch": 1.01, "learning_rate": 8.99200188191014e-05, "loss": 2.9056, "step": 12855 }, { "epoch": 1.01, "learning_rate": 8.99192346898769e-05, "loss": 2.8005, "step": 12856 }, { "epoch": 1.01, "learning_rate": 8.99184505606524e-05, "loss": 2.8482, "step": 12857 }, { "epoch": 1.01, "learning_rate": 8.99176664314279e-05, "loss": 2.7597, "step": 12858 }, { "epoch": 1.01, "learning_rate": 8.991688230220341e-05, "loss": 3.2585, "step": 12859 }, { "epoch": 1.01, "learning_rate": 8.991609817297891e-05, "loss": 2.8236, "step": 12860 }, { "epoch": 1.01, "learning_rate": 8.991531404375442e-05, "loss": 3.0792, "step": 12861 }, { "epoch": 1.01, "learning_rate": 8.991452991452992e-05, "loss": 2.8124, "step": 12862 }, { "epoch": 1.01, "learning_rate": 8.991374578530542e-05, "loss": 3.0147, "step": 12863 }, { "epoch": 1.01, "learning_rate": 8.991296165608093e-05, "loss": 2.971, "step": 12864 }, { "epoch": 1.01, "learning_rate": 8.991217752685643e-05, "loss": 3.201, "step": 12865 }, { "epoch": 1.01, "learning_rate": 8.991139339763193e-05, "loss": 2.7905, "step": 12866 }, { "epoch": 1.01, "learning_rate": 8.991060926840744e-05, "loss": 2.8551, "step": 12867 }, { "epoch": 1.01, "learning_rate": 8.990982513918294e-05, "loss": 2.986, "step": 12868 }, { "epoch": 1.01, "learning_rate": 8.990904100995845e-05, "loss": 2.8666, "step": 12869 }, { "epoch": 1.01, "learning_rate": 8.990825688073395e-05, "loss": 3.0012, "step": 12870 }, { "epoch": 1.01, "learning_rate": 8.990747275150945e-05, "loss": 2.9015, "step": 12871 }, { "epoch": 1.01, "learning_rate": 8.990668862228496e-05, "loss": 2.7911, "step": 12872 }, { "epoch": 1.01, "learning_rate": 8.990590449306046e-05, "loss": 2.8032, "step": 12873 }, { "epoch": 1.01, "learning_rate": 8.990512036383596e-05, "loss": 3.1474, "step": 12874 }, { "epoch": 1.01, "learning_rate": 8.990433623461147e-05, "loss": 2.8557, "step": 12875 }, { "epoch": 1.01, "learning_rate": 8.990355210538697e-05, "loss": 2.8067, "step": 12876 }, { "epoch": 1.01, "learning_rate": 8.990276797616248e-05, "loss": 2.8509, "step": 12877 }, { "epoch": 1.01, "learning_rate": 8.990198384693798e-05, "loss": 2.8517, "step": 12878 }, { "epoch": 1.01, "learning_rate": 8.990119971771348e-05, "loss": 2.8948, "step": 12879 }, { "epoch": 1.01, "learning_rate": 8.990041558848899e-05, "loss": 3.0132, "step": 12880 }, { "epoch": 1.01, "learning_rate": 8.989963145926449e-05, "loss": 2.9715, "step": 12881 }, { "epoch": 1.01, "learning_rate": 8.989884733004e-05, "loss": 2.722, "step": 12882 }, { "epoch": 1.01, "learning_rate": 8.98980632008155e-05, "loss": 2.8638, "step": 12883 }, { "epoch": 1.01, "learning_rate": 8.9897279071591e-05, "loss": 3.1305, "step": 12884 }, { "epoch": 1.01, "learning_rate": 8.98964949423665e-05, "loss": 2.822, "step": 12885 }, { "epoch": 1.01, "learning_rate": 8.989571081314201e-05, "loss": 3.0289, "step": 12886 }, { "epoch": 1.01, "learning_rate": 8.989492668391751e-05, "loss": 2.8024, "step": 12887 }, { "epoch": 1.01, "learning_rate": 8.989414255469302e-05, "loss": 2.957, "step": 12888 }, { "epoch": 1.01, "learning_rate": 8.989335842546852e-05, "loss": 3.1065, "step": 12889 }, { "epoch": 1.01, "learning_rate": 8.989257429624402e-05, "loss": 2.8588, "step": 12890 }, { "epoch": 1.01, "learning_rate": 8.989179016701953e-05, "loss": 2.9396, "step": 12891 }, { "epoch": 1.01, "learning_rate": 8.989100603779503e-05, "loss": 2.7547, "step": 12892 }, { "epoch": 1.01, "learning_rate": 8.989022190857053e-05, "loss": 3.1123, "step": 12893 }, { "epoch": 1.01, "learning_rate": 8.988943777934604e-05, "loss": 2.9735, "step": 12894 }, { "epoch": 1.01, "learning_rate": 8.988865365012154e-05, "loss": 2.9686, "step": 12895 }, { "epoch": 1.01, "learning_rate": 8.988786952089705e-05, "loss": 2.8675, "step": 12896 }, { "epoch": 1.01, "learning_rate": 8.988708539167255e-05, "loss": 2.9128, "step": 12897 }, { "epoch": 1.01, "learning_rate": 8.988630126244805e-05, "loss": 2.936, "step": 12898 }, { "epoch": 1.01, "learning_rate": 8.988551713322356e-05, "loss": 2.9994, "step": 12899 }, { "epoch": 1.01, "learning_rate": 8.988473300399906e-05, "loss": 3.2158, "step": 12900 }, { "epoch": 1.01, "learning_rate": 8.988394887477458e-05, "loss": 3.0734, "step": 12901 }, { "epoch": 1.01, "learning_rate": 8.988316474555008e-05, "loss": 2.9652, "step": 12902 }, { "epoch": 1.01, "learning_rate": 8.988238061632558e-05, "loss": 2.8401, "step": 12903 }, { "epoch": 1.01, "learning_rate": 8.988159648710109e-05, "loss": 2.7613, "step": 12904 }, { "epoch": 1.01, "learning_rate": 8.988081235787659e-05, "loss": 2.9068, "step": 12905 }, { "epoch": 1.01, "learning_rate": 8.98800282286521e-05, "loss": 2.8845, "step": 12906 }, { "epoch": 1.01, "learning_rate": 8.98792440994276e-05, "loss": 2.8424, "step": 12907 }, { "epoch": 1.01, "learning_rate": 8.987845997020309e-05, "loss": 2.7515, "step": 12908 }, { "epoch": 1.01, "learning_rate": 8.987767584097859e-05, "loss": 2.9403, "step": 12909 }, { "epoch": 1.01, "learning_rate": 8.98768917117541e-05, "loss": 2.8136, "step": 12910 }, { "epoch": 1.01, "learning_rate": 8.98761075825296e-05, "loss": 2.8125, "step": 12911 }, { "epoch": 1.01, "learning_rate": 8.98753234533051e-05, "loss": 2.8862, "step": 12912 }, { "epoch": 1.01, "learning_rate": 8.987453932408061e-05, "loss": 3.0706, "step": 12913 }, { "epoch": 1.01, "learning_rate": 8.987375519485611e-05, "loss": 2.8003, "step": 12914 }, { "epoch": 1.01, "learning_rate": 8.987297106563162e-05, "loss": 2.6607, "step": 12915 }, { "epoch": 1.01, "learning_rate": 8.987218693640712e-05, "loss": 3.0259, "step": 12916 }, { "epoch": 1.01, "learning_rate": 8.987140280718262e-05, "loss": 3.1053, "step": 12917 }, { "epoch": 1.01, "learning_rate": 8.987061867795813e-05, "loss": 2.6801, "step": 12918 }, { "epoch": 1.01, "learning_rate": 8.986983454873363e-05, "loss": 2.8465, "step": 12919 }, { "epoch": 1.01, "learning_rate": 8.986905041950913e-05, "loss": 2.7328, "step": 12920 }, { "epoch": 1.01, "learning_rate": 8.986826629028464e-05, "loss": 2.8299, "step": 12921 }, { "epoch": 1.01, "learning_rate": 8.986748216106014e-05, "loss": 3.0427, "step": 12922 }, { "epoch": 1.01, "learning_rate": 8.986669803183564e-05, "loss": 2.9171, "step": 12923 }, { "epoch": 1.01, "learning_rate": 8.986591390261115e-05, "loss": 2.889, "step": 12924 }, { "epoch": 1.01, "learning_rate": 8.986512977338665e-05, "loss": 2.9903, "step": 12925 }, { "epoch": 1.01, "learning_rate": 8.986434564416217e-05, "loss": 3.0391, "step": 12926 }, { "epoch": 1.01, "learning_rate": 8.986356151493767e-05, "loss": 2.9002, "step": 12927 }, { "epoch": 1.01, "learning_rate": 8.986277738571318e-05, "loss": 2.922, "step": 12928 }, { "epoch": 1.01, "learning_rate": 8.986199325648868e-05, "loss": 2.8443, "step": 12929 }, { "epoch": 1.01, "learning_rate": 8.986120912726418e-05, "loss": 2.8596, "step": 12930 }, { "epoch": 1.01, "learning_rate": 8.986042499803969e-05, "loss": 3.3285, "step": 12931 }, { "epoch": 1.01, "learning_rate": 8.985964086881519e-05, "loss": 2.8188, "step": 12932 }, { "epoch": 1.01, "learning_rate": 8.98588567395907e-05, "loss": 3.0002, "step": 12933 }, { "epoch": 1.01, "learning_rate": 8.98580726103662e-05, "loss": 2.7949, "step": 12934 }, { "epoch": 1.01, "learning_rate": 8.98572884811417e-05, "loss": 2.7012, "step": 12935 }, { "epoch": 1.01, "learning_rate": 8.98565043519172e-05, "loss": 2.8246, "step": 12936 }, { "epoch": 1.01, "learning_rate": 8.985572022269271e-05, "loss": 2.8815, "step": 12937 }, { "epoch": 1.01, "learning_rate": 8.985493609346821e-05, "loss": 3.1358, "step": 12938 }, { "epoch": 1.01, "learning_rate": 8.985415196424372e-05, "loss": 3.0926, "step": 12939 }, { "epoch": 1.01, "learning_rate": 8.985336783501922e-05, "loss": 2.8815, "step": 12940 }, { "epoch": 1.01, "learning_rate": 8.985258370579472e-05, "loss": 2.9743, "step": 12941 }, { "epoch": 1.01, "learning_rate": 8.985179957657023e-05, "loss": 2.7907, "step": 12942 }, { "epoch": 1.01, "learning_rate": 8.985101544734573e-05, "loss": 2.835, "step": 12943 }, { "epoch": 1.01, "learning_rate": 8.985023131812122e-05, "loss": 2.9985, "step": 12944 }, { "epoch": 1.02, "learning_rate": 8.984944718889673e-05, "loss": 2.8277, "step": 12945 }, { "epoch": 1.02, "learning_rate": 8.984866305967223e-05, "loss": 3.0746, "step": 12946 }, { "epoch": 1.02, "learning_rate": 8.984787893044773e-05, "loss": 2.872, "step": 12947 }, { "epoch": 1.02, "learning_rate": 8.984709480122324e-05, "loss": 2.6688, "step": 12948 }, { "epoch": 1.02, "learning_rate": 8.984631067199874e-05, "loss": 2.9108, "step": 12949 }, { "epoch": 1.02, "learning_rate": 8.984552654277424e-05, "loss": 2.7049, "step": 12950 }, { "epoch": 1.02, "learning_rate": 8.984474241354975e-05, "loss": 3.0261, "step": 12951 }, { "epoch": 1.02, "learning_rate": 8.984395828432526e-05, "loss": 3.0294, "step": 12952 }, { "epoch": 1.02, "learning_rate": 8.984317415510077e-05, "loss": 2.7903, "step": 12953 }, { "epoch": 1.02, "learning_rate": 8.984239002587627e-05, "loss": 3.0848, "step": 12954 }, { "epoch": 1.02, "learning_rate": 8.984160589665178e-05, "loss": 3.0368, "step": 12955 }, { "epoch": 1.02, "learning_rate": 8.984082176742728e-05, "loss": 2.7188, "step": 12956 }, { "epoch": 1.02, "learning_rate": 8.984003763820278e-05, "loss": 2.9445, "step": 12957 }, { "epoch": 1.02, "learning_rate": 8.983925350897829e-05, "loss": 2.9286, "step": 12958 }, { "epoch": 1.02, "learning_rate": 8.983846937975379e-05, "loss": 3.1895, "step": 12959 }, { "epoch": 1.02, "learning_rate": 8.98376852505293e-05, "loss": 2.7447, "step": 12960 }, { "epoch": 1.02, "learning_rate": 8.98369011213048e-05, "loss": 2.9293, "step": 12961 }, { "epoch": 1.02, "learning_rate": 8.98361169920803e-05, "loss": 2.799, "step": 12962 }, { "epoch": 1.02, "learning_rate": 8.98353328628558e-05, "loss": 2.8923, "step": 12963 }, { "epoch": 1.02, "learning_rate": 8.983454873363131e-05, "loss": 2.7684, "step": 12964 }, { "epoch": 1.02, "learning_rate": 8.983376460440681e-05, "loss": 2.9645, "step": 12965 }, { "epoch": 1.02, "learning_rate": 8.983298047518232e-05, "loss": 3.1257, "step": 12966 }, { "epoch": 1.02, "learning_rate": 8.983219634595782e-05, "loss": 3.0993, "step": 12967 }, { "epoch": 1.02, "learning_rate": 8.983141221673332e-05, "loss": 2.8211, "step": 12968 }, { "epoch": 1.02, "learning_rate": 8.983062808750883e-05, "loss": 2.8631, "step": 12969 }, { "epoch": 1.02, "learning_rate": 8.982984395828433e-05, "loss": 2.8765, "step": 12970 }, { "epoch": 1.02, "learning_rate": 8.982905982905983e-05, "loss": 2.9358, "step": 12971 }, { "epoch": 1.02, "learning_rate": 8.982827569983534e-05, "loss": 2.8434, "step": 12972 }, { "epoch": 1.02, "learning_rate": 8.982749157061084e-05, "loss": 2.7432, "step": 12973 }, { "epoch": 1.02, "learning_rate": 8.982670744138635e-05, "loss": 2.8245, "step": 12974 }, { "epoch": 1.02, "learning_rate": 8.982592331216185e-05, "loss": 2.9676, "step": 12975 }, { "epoch": 1.02, "learning_rate": 8.982513918293735e-05, "loss": 2.9581, "step": 12976 }, { "epoch": 1.02, "learning_rate": 8.982435505371286e-05, "loss": 2.9454, "step": 12977 }, { "epoch": 1.02, "learning_rate": 8.982357092448836e-05, "loss": 2.9731, "step": 12978 }, { "epoch": 1.02, "learning_rate": 8.982278679526386e-05, "loss": 3.0079, "step": 12979 }, { "epoch": 1.02, "learning_rate": 8.982200266603937e-05, "loss": 2.9669, "step": 12980 }, { "epoch": 1.02, "learning_rate": 8.982121853681487e-05, "loss": 2.9347, "step": 12981 }, { "epoch": 1.02, "learning_rate": 8.982043440759037e-05, "loss": 2.8557, "step": 12982 }, { "epoch": 1.02, "learning_rate": 8.981965027836588e-05, "loss": 2.8487, "step": 12983 }, { "epoch": 1.02, "learning_rate": 8.981886614914138e-05, "loss": 2.7391, "step": 12984 }, { "epoch": 1.02, "learning_rate": 8.981808201991689e-05, "loss": 2.8266, "step": 12985 }, { "epoch": 1.02, "learning_rate": 8.981729789069239e-05, "loss": 2.7887, "step": 12986 }, { "epoch": 1.02, "learning_rate": 8.98165137614679e-05, "loss": 2.634, "step": 12987 }, { "epoch": 1.02, "learning_rate": 8.98157296322434e-05, "loss": 3.1208, "step": 12988 }, { "epoch": 1.02, "learning_rate": 8.98149455030189e-05, "loss": 3.1733, "step": 12989 }, { "epoch": 1.02, "learning_rate": 8.98141613737944e-05, "loss": 2.7635, "step": 12990 }, { "epoch": 1.02, "learning_rate": 8.981337724456991e-05, "loss": 2.9202, "step": 12991 }, { "epoch": 1.02, "learning_rate": 8.981259311534541e-05, "loss": 2.9663, "step": 12992 }, { "epoch": 1.02, "learning_rate": 8.981180898612092e-05, "loss": 2.8566, "step": 12993 }, { "epoch": 1.02, "learning_rate": 8.981102485689642e-05, "loss": 2.8842, "step": 12994 }, { "epoch": 1.02, "learning_rate": 8.981024072767192e-05, "loss": 2.8745, "step": 12995 }, { "epoch": 1.02, "learning_rate": 8.980945659844743e-05, "loss": 3.0344, "step": 12996 }, { "epoch": 1.02, "learning_rate": 8.980867246922293e-05, "loss": 2.899, "step": 12997 }, { "epoch": 1.02, "learning_rate": 8.980788833999843e-05, "loss": 3.0347, "step": 12998 }, { "epoch": 1.02, "learning_rate": 8.980710421077394e-05, "loss": 3.0279, "step": 12999 }, { "epoch": 1.02, "learning_rate": 8.980632008154944e-05, "loss": 2.8451, "step": 13000 }, { "epoch": 1.02, "eval_loss": 2.926194190979004, "eval_runtime": 218.0348, "eval_samples_per_second": 51.973, "eval_steps_per_second": 3.252, "step": 13000 }, { "epoch": 1.02, "learning_rate": 8.980553595232494e-05, "loss": 2.9471, "step": 13001 }, { "epoch": 1.02, "learning_rate": 8.980475182310045e-05, "loss": 2.9117, "step": 13002 }, { "epoch": 1.02, "learning_rate": 8.980396769387595e-05, "loss": 2.8633, "step": 13003 }, { "epoch": 1.02, "learning_rate": 8.980318356465146e-05, "loss": 2.9123, "step": 13004 }, { "epoch": 1.02, "learning_rate": 8.980239943542696e-05, "loss": 2.8223, "step": 13005 }, { "epoch": 1.02, "learning_rate": 8.980161530620246e-05, "loss": 2.7856, "step": 13006 }, { "epoch": 1.02, "learning_rate": 8.980083117697797e-05, "loss": 2.7833, "step": 13007 }, { "epoch": 1.02, "learning_rate": 8.980004704775347e-05, "loss": 2.9774, "step": 13008 }, { "epoch": 1.02, "learning_rate": 8.979926291852899e-05, "loss": 2.9325, "step": 13009 }, { "epoch": 1.02, "learning_rate": 8.979847878930449e-05, "loss": 2.7844, "step": 13010 }, { "epoch": 1.02, "learning_rate": 8.979769466008e-05, "loss": 2.8322, "step": 13011 }, { "epoch": 1.02, "learning_rate": 8.97969105308555e-05, "loss": 2.9217, "step": 13012 }, { "epoch": 1.02, "learning_rate": 8.9796126401631e-05, "loss": 2.9169, "step": 13013 }, { "epoch": 1.02, "learning_rate": 8.97953422724065e-05, "loss": 2.8085, "step": 13014 }, { "epoch": 1.02, "learning_rate": 8.979455814318201e-05, "loss": 2.6603, "step": 13015 }, { "epoch": 1.02, "learning_rate": 8.97937740139575e-05, "loss": 3.0294, "step": 13016 }, { "epoch": 1.02, "learning_rate": 8.9792989884733e-05, "loss": 2.8749, "step": 13017 }, { "epoch": 1.02, "learning_rate": 8.979220575550851e-05, "loss": 3.0926, "step": 13018 }, { "epoch": 1.02, "learning_rate": 8.979142162628401e-05, "loss": 2.816, "step": 13019 }, { "epoch": 1.02, "learning_rate": 8.979063749705951e-05, "loss": 2.9767, "step": 13020 }, { "epoch": 1.02, "learning_rate": 8.978985336783502e-05, "loss": 2.7159, "step": 13021 }, { "epoch": 1.02, "learning_rate": 8.978906923861052e-05, "loss": 3.0092, "step": 13022 }, { "epoch": 1.02, "learning_rate": 8.978828510938603e-05, "loss": 2.6851, "step": 13023 }, { "epoch": 1.02, "learning_rate": 8.978750098016153e-05, "loss": 2.9826, "step": 13024 }, { "epoch": 1.02, "learning_rate": 8.978671685093703e-05, "loss": 2.8564, "step": 13025 }, { "epoch": 1.02, "learning_rate": 8.978593272171254e-05, "loss": 2.9837, "step": 13026 }, { "epoch": 1.02, "learning_rate": 8.978514859248804e-05, "loss": 2.781, "step": 13027 }, { "epoch": 1.02, "learning_rate": 8.978436446326354e-05, "loss": 2.9204, "step": 13028 }, { "epoch": 1.02, "learning_rate": 8.978358033403905e-05, "loss": 2.9048, "step": 13029 }, { "epoch": 1.02, "learning_rate": 8.978279620481455e-05, "loss": 2.826, "step": 13030 }, { "epoch": 1.02, "learning_rate": 8.978201207559006e-05, "loss": 2.9642, "step": 13031 }, { "epoch": 1.02, "learning_rate": 8.978122794636556e-05, "loss": 2.8113, "step": 13032 }, { "epoch": 1.02, "learning_rate": 8.978044381714106e-05, "loss": 2.8153, "step": 13033 }, { "epoch": 1.02, "learning_rate": 8.977965968791658e-05, "loss": 2.8049, "step": 13034 }, { "epoch": 1.02, "learning_rate": 8.977887555869208e-05, "loss": 2.9791, "step": 13035 }, { "epoch": 1.02, "learning_rate": 8.977809142946759e-05, "loss": 3.0501, "step": 13036 }, { "epoch": 1.02, "learning_rate": 8.977730730024309e-05, "loss": 2.6778, "step": 13037 }, { "epoch": 1.02, "learning_rate": 8.97765231710186e-05, "loss": 2.9891, "step": 13038 }, { "epoch": 1.02, "learning_rate": 8.97757390417941e-05, "loss": 2.8315, "step": 13039 }, { "epoch": 1.02, "learning_rate": 8.97749549125696e-05, "loss": 2.4473, "step": 13040 }, { "epoch": 1.02, "learning_rate": 8.97741707833451e-05, "loss": 2.932, "step": 13041 }, { "epoch": 1.02, "learning_rate": 8.977338665412061e-05, "loss": 2.8117, "step": 13042 }, { "epoch": 1.02, "learning_rate": 8.977260252489611e-05, "loss": 3.0899, "step": 13043 }, { "epoch": 1.02, "learning_rate": 8.977181839567162e-05, "loss": 2.7362, "step": 13044 }, { "epoch": 1.02, "learning_rate": 8.977103426644712e-05, "loss": 2.9788, "step": 13045 }, { "epoch": 1.02, "learning_rate": 8.977025013722262e-05, "loss": 2.8757, "step": 13046 }, { "epoch": 1.02, "learning_rate": 8.976946600799813e-05, "loss": 2.893, "step": 13047 }, { "epoch": 1.02, "learning_rate": 8.976868187877363e-05, "loss": 2.7777, "step": 13048 }, { "epoch": 1.02, "learning_rate": 8.976789774954913e-05, "loss": 2.9217, "step": 13049 }, { "epoch": 1.02, "learning_rate": 8.976711362032464e-05, "loss": 2.8845, "step": 13050 }, { "epoch": 1.02, "learning_rate": 8.976632949110014e-05, "loss": 2.8473, "step": 13051 }, { "epoch": 1.02, "learning_rate": 8.976554536187563e-05, "loss": 2.7977, "step": 13052 }, { "epoch": 1.02, "learning_rate": 8.976476123265114e-05, "loss": 3.0726, "step": 13053 }, { "epoch": 1.02, "learning_rate": 8.976397710342664e-05, "loss": 3.14, "step": 13054 }, { "epoch": 1.02, "learning_rate": 8.976319297420214e-05, "loss": 2.8634, "step": 13055 }, { "epoch": 1.02, "learning_rate": 8.976240884497765e-05, "loss": 2.7419, "step": 13056 }, { "epoch": 1.02, "learning_rate": 8.976162471575315e-05, "loss": 3.0136, "step": 13057 }, { "epoch": 1.02, "learning_rate": 8.976084058652865e-05, "loss": 2.7846, "step": 13058 }, { "epoch": 1.02, "learning_rate": 8.976005645730416e-05, "loss": 3.0275, "step": 13059 }, { "epoch": 1.02, "learning_rate": 8.975927232807968e-05, "loss": 2.8882, "step": 13060 }, { "epoch": 1.02, "learning_rate": 8.975848819885518e-05, "loss": 2.5701, "step": 13061 }, { "epoch": 1.02, "learning_rate": 8.975770406963068e-05, "loss": 3.0479, "step": 13062 }, { "epoch": 1.02, "learning_rate": 8.975691994040619e-05, "loss": 3.0018, "step": 13063 }, { "epoch": 1.02, "learning_rate": 8.975613581118169e-05, "loss": 2.8852, "step": 13064 }, { "epoch": 1.02, "learning_rate": 8.97553516819572e-05, "loss": 2.7764, "step": 13065 }, { "epoch": 1.02, "learning_rate": 8.97545675527327e-05, "loss": 3.0412, "step": 13066 }, { "epoch": 1.02, "learning_rate": 8.97537834235082e-05, "loss": 3.01, "step": 13067 }, { "epoch": 1.02, "learning_rate": 8.97529992942837e-05, "loss": 2.9843, "step": 13068 }, { "epoch": 1.02, "learning_rate": 8.975221516505921e-05, "loss": 3.1145, "step": 13069 }, { "epoch": 1.02, "learning_rate": 8.975143103583471e-05, "loss": 3.025, "step": 13070 }, { "epoch": 1.02, "learning_rate": 8.975064690661022e-05, "loss": 2.846, "step": 13071 }, { "epoch": 1.03, "learning_rate": 8.974986277738572e-05, "loss": 2.9742, "step": 13072 }, { "epoch": 1.03, "learning_rate": 8.974907864816122e-05, "loss": 2.9757, "step": 13073 }, { "epoch": 1.03, "learning_rate": 8.974829451893673e-05, "loss": 3.0642, "step": 13074 }, { "epoch": 1.03, "learning_rate": 8.974751038971223e-05, "loss": 2.7844, "step": 13075 }, { "epoch": 1.03, "learning_rate": 8.974672626048773e-05, "loss": 3.0597, "step": 13076 }, { "epoch": 1.03, "learning_rate": 8.974594213126324e-05, "loss": 3.0329, "step": 13077 }, { "epoch": 1.03, "learning_rate": 8.974515800203874e-05, "loss": 2.8415, "step": 13078 }, { "epoch": 1.03, "learning_rate": 8.974437387281425e-05, "loss": 2.9537, "step": 13079 }, { "epoch": 1.03, "learning_rate": 8.974358974358975e-05, "loss": 3.0817, "step": 13080 }, { "epoch": 1.03, "learning_rate": 8.974280561436525e-05, "loss": 2.8398, "step": 13081 }, { "epoch": 1.03, "learning_rate": 8.974202148514076e-05, "loss": 2.9506, "step": 13082 }, { "epoch": 1.03, "learning_rate": 8.974123735591626e-05, "loss": 2.9978, "step": 13083 }, { "epoch": 1.03, "learning_rate": 8.974045322669176e-05, "loss": 2.8003, "step": 13084 }, { "epoch": 1.03, "learning_rate": 8.973966909746727e-05, "loss": 2.8333, "step": 13085 }, { "epoch": 1.03, "learning_rate": 8.973888496824277e-05, "loss": 2.8368, "step": 13086 }, { "epoch": 1.03, "learning_rate": 8.973810083901827e-05, "loss": 2.8701, "step": 13087 }, { "epoch": 1.03, "learning_rate": 8.973731670979378e-05, "loss": 2.6684, "step": 13088 }, { "epoch": 1.03, "learning_rate": 8.973653258056928e-05, "loss": 3.0453, "step": 13089 }, { "epoch": 1.03, "learning_rate": 8.973574845134479e-05, "loss": 2.7374, "step": 13090 }, { "epoch": 1.03, "learning_rate": 8.973496432212029e-05, "loss": 2.9249, "step": 13091 }, { "epoch": 1.03, "learning_rate": 8.973418019289579e-05, "loss": 2.7197, "step": 13092 }, { "epoch": 1.03, "learning_rate": 8.97333960636713e-05, "loss": 2.628, "step": 13093 }, { "epoch": 1.03, "learning_rate": 8.97326119344468e-05, "loss": 2.8645, "step": 13094 }, { "epoch": 1.03, "learning_rate": 8.97318278052223e-05, "loss": 2.9209, "step": 13095 }, { "epoch": 1.03, "learning_rate": 8.973104367599781e-05, "loss": 2.8561, "step": 13096 }, { "epoch": 1.03, "learning_rate": 8.973025954677331e-05, "loss": 3.0518, "step": 13097 }, { "epoch": 1.03, "learning_rate": 8.972947541754882e-05, "loss": 2.7801, "step": 13098 }, { "epoch": 1.03, "learning_rate": 8.972869128832432e-05, "loss": 3.0827, "step": 13099 }, { "epoch": 1.03, "learning_rate": 8.972790715909982e-05, "loss": 2.8232, "step": 13100 }, { "epoch": 1.03, "learning_rate": 8.972712302987533e-05, "loss": 2.7919, "step": 13101 }, { "epoch": 1.03, "learning_rate": 8.972633890065083e-05, "loss": 2.9196, "step": 13102 }, { "epoch": 1.03, "learning_rate": 8.972555477142633e-05, "loss": 2.7598, "step": 13103 }, { "epoch": 1.03, "learning_rate": 8.972477064220184e-05, "loss": 2.9462, "step": 13104 }, { "epoch": 1.03, "learning_rate": 8.972398651297734e-05, "loss": 2.8972, "step": 13105 }, { "epoch": 1.03, "learning_rate": 8.972320238375284e-05, "loss": 2.9554, "step": 13106 }, { "epoch": 1.03, "learning_rate": 8.972241825452835e-05, "loss": 3.0262, "step": 13107 }, { "epoch": 1.03, "learning_rate": 8.972163412530385e-05, "loss": 2.8307, "step": 13108 }, { "epoch": 1.03, "learning_rate": 8.972084999607936e-05, "loss": 2.752, "step": 13109 }, { "epoch": 1.03, "learning_rate": 8.972006586685486e-05, "loss": 3.1023, "step": 13110 }, { "epoch": 1.03, "learning_rate": 8.971928173763036e-05, "loss": 2.7701, "step": 13111 }, { "epoch": 1.03, "learning_rate": 8.971849760840587e-05, "loss": 2.9335, "step": 13112 }, { "epoch": 1.03, "learning_rate": 8.971771347918137e-05, "loss": 3.1487, "step": 13113 }, { "epoch": 1.03, "learning_rate": 8.971692934995687e-05, "loss": 2.9674, "step": 13114 }, { "epoch": 1.03, "learning_rate": 8.971614522073238e-05, "loss": 2.8499, "step": 13115 }, { "epoch": 1.03, "learning_rate": 8.971536109150788e-05, "loss": 2.8311, "step": 13116 }, { "epoch": 1.03, "learning_rate": 8.97145769622834e-05, "loss": 2.9787, "step": 13117 }, { "epoch": 1.03, "learning_rate": 8.97137928330589e-05, "loss": 3.0851, "step": 13118 }, { "epoch": 1.03, "learning_rate": 8.97130087038344e-05, "loss": 2.9689, "step": 13119 }, { "epoch": 1.03, "learning_rate": 8.971222457460991e-05, "loss": 3.0132, "step": 13120 }, { "epoch": 1.03, "learning_rate": 8.971144044538541e-05, "loss": 2.9369, "step": 13121 }, { "epoch": 1.03, "learning_rate": 8.971065631616092e-05, "loss": 2.8981, "step": 13122 }, { "epoch": 1.03, "learning_rate": 8.970987218693642e-05, "loss": 2.8133, "step": 13123 }, { "epoch": 1.03, "learning_rate": 8.970908805771191e-05, "loss": 2.8028, "step": 13124 }, { "epoch": 1.03, "learning_rate": 8.970830392848741e-05, "loss": 2.8146, "step": 13125 }, { "epoch": 1.03, "learning_rate": 8.970751979926292e-05, "loss": 3.0197, "step": 13126 }, { "epoch": 1.03, "learning_rate": 8.970673567003842e-05, "loss": 2.8543, "step": 13127 }, { "epoch": 1.03, "learning_rate": 8.970595154081393e-05, "loss": 3.0087, "step": 13128 }, { "epoch": 1.03, "learning_rate": 8.970516741158943e-05, "loss": 2.7727, "step": 13129 }, { "epoch": 1.03, "learning_rate": 8.970438328236493e-05, "loss": 2.952, "step": 13130 }, { "epoch": 1.03, "learning_rate": 8.970359915314044e-05, "loss": 3.0437, "step": 13131 }, { "epoch": 1.03, "learning_rate": 8.970281502391594e-05, "loss": 2.8703, "step": 13132 }, { "epoch": 1.03, "learning_rate": 8.970203089469144e-05, "loss": 3.0226, "step": 13133 }, { "epoch": 1.03, "learning_rate": 8.970124676546695e-05, "loss": 2.7822, "step": 13134 }, { "epoch": 1.03, "learning_rate": 8.970046263624245e-05, "loss": 2.9467, "step": 13135 }, { "epoch": 1.03, "learning_rate": 8.969967850701795e-05, "loss": 2.7634, "step": 13136 }, { "epoch": 1.03, "learning_rate": 8.969889437779346e-05, "loss": 2.8968, "step": 13137 }, { "epoch": 1.03, "learning_rate": 8.969811024856896e-05, "loss": 3.1412, "step": 13138 }, { "epoch": 1.03, "learning_rate": 8.969732611934447e-05, "loss": 2.9324, "step": 13139 }, { "epoch": 1.03, "learning_rate": 8.969654199011997e-05, "loss": 2.8688, "step": 13140 }, { "epoch": 1.03, "learning_rate": 8.969575786089547e-05, "loss": 2.8566, "step": 13141 }, { "epoch": 1.03, "learning_rate": 8.969497373167098e-05, "loss": 2.7286, "step": 13142 }, { "epoch": 1.03, "learning_rate": 8.96941896024465e-05, "loss": 3.1941, "step": 13143 }, { "epoch": 1.03, "learning_rate": 8.9693405473222e-05, "loss": 2.8971, "step": 13144 }, { "epoch": 1.03, "learning_rate": 8.96926213439975e-05, "loss": 2.8585, "step": 13145 }, { "epoch": 1.03, "learning_rate": 8.9691837214773e-05, "loss": 2.8534, "step": 13146 }, { "epoch": 1.03, "learning_rate": 8.969105308554851e-05, "loss": 2.7143, "step": 13147 }, { "epoch": 1.03, "learning_rate": 8.969026895632401e-05, "loss": 2.9184, "step": 13148 }, { "epoch": 1.03, "learning_rate": 8.968948482709952e-05, "loss": 2.9838, "step": 13149 }, { "epoch": 1.03, "learning_rate": 8.968870069787502e-05, "loss": 2.9622, "step": 13150 }, { "epoch": 1.03, "learning_rate": 8.968791656865052e-05, "loss": 2.8, "step": 13151 }, { "epoch": 1.03, "learning_rate": 8.968713243942603e-05, "loss": 2.9851, "step": 13152 }, { "epoch": 1.03, "learning_rate": 8.968634831020153e-05, "loss": 2.8368, "step": 13153 }, { "epoch": 1.03, "learning_rate": 8.968556418097703e-05, "loss": 2.9114, "step": 13154 }, { "epoch": 1.03, "learning_rate": 8.968478005175254e-05, "loss": 2.9243, "step": 13155 }, { "epoch": 1.03, "learning_rate": 8.968399592252804e-05, "loss": 3.0083, "step": 13156 }, { "epoch": 1.03, "learning_rate": 8.968321179330355e-05, "loss": 2.8675, "step": 13157 }, { "epoch": 1.03, "learning_rate": 8.968242766407905e-05, "loss": 3.1491, "step": 13158 }, { "epoch": 1.03, "learning_rate": 8.968164353485455e-05, "loss": 3.0821, "step": 13159 }, { "epoch": 1.03, "learning_rate": 8.968085940563004e-05, "loss": 2.8245, "step": 13160 }, { "epoch": 1.03, "learning_rate": 8.968007527640555e-05, "loss": 2.9238, "step": 13161 }, { "epoch": 1.03, "learning_rate": 8.967929114718105e-05, "loss": 3.0797, "step": 13162 }, { "epoch": 1.03, "learning_rate": 8.967850701795655e-05, "loss": 3.1278, "step": 13163 }, { "epoch": 1.03, "learning_rate": 8.967772288873206e-05, "loss": 2.9928, "step": 13164 }, { "epoch": 1.03, "learning_rate": 8.967693875950756e-05, "loss": 2.7164, "step": 13165 }, { "epoch": 1.03, "learning_rate": 8.967615463028307e-05, "loss": 3.1307, "step": 13166 }, { "epoch": 1.03, "learning_rate": 8.967537050105857e-05, "loss": 2.8478, "step": 13167 }, { "epoch": 1.03, "learning_rate": 8.967458637183409e-05, "loss": 2.8653, "step": 13168 }, { "epoch": 1.03, "learning_rate": 8.967380224260959e-05, "loss": 2.7299, "step": 13169 }, { "epoch": 1.03, "learning_rate": 8.96730181133851e-05, "loss": 2.9704, "step": 13170 }, { "epoch": 1.03, "learning_rate": 8.96722339841606e-05, "loss": 2.9125, "step": 13171 }, { "epoch": 1.03, "learning_rate": 8.96714498549361e-05, "loss": 2.8434, "step": 13172 }, { "epoch": 1.03, "learning_rate": 8.96706657257116e-05, "loss": 3.0017, "step": 13173 }, { "epoch": 1.03, "learning_rate": 8.966988159648711e-05, "loss": 2.9637, "step": 13174 }, { "epoch": 1.03, "learning_rate": 8.966909746726261e-05, "loss": 2.8978, "step": 13175 }, { "epoch": 1.03, "learning_rate": 8.966831333803812e-05, "loss": 2.9859, "step": 13176 }, { "epoch": 1.03, "learning_rate": 8.966752920881362e-05, "loss": 2.8726, "step": 13177 }, { "epoch": 1.03, "learning_rate": 8.966674507958912e-05, "loss": 2.9078, "step": 13178 }, { "epoch": 1.03, "learning_rate": 8.966596095036463e-05, "loss": 2.9595, "step": 13179 }, { "epoch": 1.03, "learning_rate": 8.966517682114013e-05, "loss": 2.9037, "step": 13180 }, { "epoch": 1.03, "learning_rate": 8.966439269191563e-05, "loss": 2.9111, "step": 13181 }, { "epoch": 1.03, "learning_rate": 8.966360856269114e-05, "loss": 3.1102, "step": 13182 }, { "epoch": 1.03, "learning_rate": 8.966282443346664e-05, "loss": 3.0064, "step": 13183 }, { "epoch": 1.03, "learning_rate": 8.966204030424214e-05, "loss": 2.9597, "step": 13184 }, { "epoch": 1.03, "learning_rate": 8.966125617501765e-05, "loss": 3.1752, "step": 13185 }, { "epoch": 1.03, "learning_rate": 8.966047204579315e-05, "loss": 3.2005, "step": 13186 }, { "epoch": 1.03, "learning_rate": 8.965968791656866e-05, "loss": 2.7788, "step": 13187 }, { "epoch": 1.03, "learning_rate": 8.965890378734416e-05, "loss": 2.7083, "step": 13188 }, { "epoch": 1.03, "learning_rate": 8.965811965811966e-05, "loss": 2.778, "step": 13189 }, { "epoch": 1.03, "learning_rate": 8.965733552889517e-05, "loss": 3.0309, "step": 13190 }, { "epoch": 1.03, "learning_rate": 8.965655139967067e-05, "loss": 2.8668, "step": 13191 }, { "epoch": 1.03, "learning_rate": 8.965576727044617e-05, "loss": 2.976, "step": 13192 }, { "epoch": 1.03, "learning_rate": 8.965498314122168e-05, "loss": 2.9996, "step": 13193 }, { "epoch": 1.03, "learning_rate": 8.965419901199718e-05, "loss": 2.9514, "step": 13194 }, { "epoch": 1.03, "learning_rate": 8.965341488277269e-05, "loss": 2.7526, "step": 13195 }, { "epoch": 1.03, "learning_rate": 8.965263075354819e-05, "loss": 2.9171, "step": 13196 }, { "epoch": 1.03, "learning_rate": 8.965184662432369e-05, "loss": 2.7862, "step": 13197 }, { "epoch": 1.03, "learning_rate": 8.96510624950992e-05, "loss": 2.9366, "step": 13198 }, { "epoch": 1.03, "learning_rate": 8.96502783658747e-05, "loss": 3.1732, "step": 13199 }, { "epoch": 1.04, "learning_rate": 8.96494942366502e-05, "loss": 2.8947, "step": 13200 }, { "epoch": 1.04, "learning_rate": 8.964871010742571e-05, "loss": 2.8082, "step": 13201 }, { "epoch": 1.04, "learning_rate": 8.964792597820121e-05, "loss": 3.2009, "step": 13202 }, { "epoch": 1.04, "learning_rate": 8.964714184897671e-05, "loss": 2.7228, "step": 13203 }, { "epoch": 1.04, "learning_rate": 8.964635771975222e-05, "loss": 3.013, "step": 13204 }, { "epoch": 1.04, "learning_rate": 8.964557359052772e-05, "loss": 2.8342, "step": 13205 }, { "epoch": 1.04, "learning_rate": 8.964478946130323e-05, "loss": 2.8175, "step": 13206 }, { "epoch": 1.04, "learning_rate": 8.964400533207873e-05, "loss": 2.8408, "step": 13207 }, { "epoch": 1.04, "learning_rate": 8.964322120285423e-05, "loss": 2.9393, "step": 13208 }, { "epoch": 1.04, "learning_rate": 8.964243707362974e-05, "loss": 2.9747, "step": 13209 }, { "epoch": 1.04, "learning_rate": 8.964165294440524e-05, "loss": 2.8241, "step": 13210 }, { "epoch": 1.04, "learning_rate": 8.964086881518074e-05, "loss": 2.9381, "step": 13211 }, { "epoch": 1.04, "learning_rate": 8.964008468595625e-05, "loss": 2.887, "step": 13212 }, { "epoch": 1.04, "learning_rate": 8.963930055673175e-05, "loss": 3.1089, "step": 13213 }, { "epoch": 1.04, "learning_rate": 8.963851642750726e-05, "loss": 3.0199, "step": 13214 }, { "epoch": 1.04, "learning_rate": 8.963773229828276e-05, "loss": 2.7417, "step": 13215 }, { "epoch": 1.04, "learning_rate": 8.963694816905826e-05, "loss": 2.6572, "step": 13216 }, { "epoch": 1.04, "learning_rate": 8.963616403983377e-05, "loss": 2.7739, "step": 13217 }, { "epoch": 1.04, "learning_rate": 8.963537991060927e-05, "loss": 3.1616, "step": 13218 }, { "epoch": 1.04, "learning_rate": 8.963459578138477e-05, "loss": 2.8953, "step": 13219 }, { "epoch": 1.04, "learning_rate": 8.963381165216028e-05, "loss": 3.2579, "step": 13220 }, { "epoch": 1.04, "learning_rate": 8.963302752293578e-05, "loss": 2.8618, "step": 13221 }, { "epoch": 1.04, "learning_rate": 8.963224339371128e-05, "loss": 2.8787, "step": 13222 }, { "epoch": 1.04, "learning_rate": 8.963145926448679e-05, "loss": 2.8192, "step": 13223 }, { "epoch": 1.04, "learning_rate": 8.963067513526229e-05, "loss": 2.8927, "step": 13224 }, { "epoch": 1.04, "learning_rate": 8.962989100603781e-05, "loss": 2.9312, "step": 13225 }, { "epoch": 1.04, "learning_rate": 8.962910687681331e-05, "loss": 2.8768, "step": 13226 }, { "epoch": 1.04, "learning_rate": 8.962832274758882e-05, "loss": 2.8795, "step": 13227 }, { "epoch": 1.04, "learning_rate": 8.962753861836432e-05, "loss": 3.1317, "step": 13228 }, { "epoch": 1.04, "learning_rate": 8.962675448913982e-05, "loss": 3.005, "step": 13229 }, { "epoch": 1.04, "learning_rate": 8.962597035991533e-05, "loss": 2.9018, "step": 13230 }, { "epoch": 1.04, "learning_rate": 8.962518623069083e-05, "loss": 2.9845, "step": 13231 }, { "epoch": 1.04, "learning_rate": 8.962440210146632e-05, "loss": 2.6441, "step": 13232 }, { "epoch": 1.04, "learning_rate": 8.962361797224183e-05, "loss": 2.9219, "step": 13233 }, { "epoch": 1.04, "learning_rate": 8.962283384301733e-05, "loss": 2.9823, "step": 13234 }, { "epoch": 1.04, "learning_rate": 8.962204971379283e-05, "loss": 2.9719, "step": 13235 }, { "epoch": 1.04, "learning_rate": 8.962126558456834e-05, "loss": 2.6997, "step": 13236 }, { "epoch": 1.04, "learning_rate": 8.962048145534384e-05, "loss": 2.9783, "step": 13237 }, { "epoch": 1.04, "learning_rate": 8.961969732611934e-05, "loss": 3.0046, "step": 13238 }, { "epoch": 1.04, "learning_rate": 8.961891319689485e-05, "loss": 2.9031, "step": 13239 }, { "epoch": 1.04, "learning_rate": 8.961812906767035e-05, "loss": 3.0341, "step": 13240 }, { "epoch": 1.04, "learning_rate": 8.961734493844585e-05, "loss": 3.0218, "step": 13241 }, { "epoch": 1.04, "learning_rate": 8.961656080922136e-05, "loss": 2.9448, "step": 13242 }, { "epoch": 1.04, "learning_rate": 8.961577667999686e-05, "loss": 2.7453, "step": 13243 }, { "epoch": 1.04, "learning_rate": 8.961499255077237e-05, "loss": 2.7812, "step": 13244 }, { "epoch": 1.04, "learning_rate": 8.961420842154787e-05, "loss": 2.806, "step": 13245 }, { "epoch": 1.04, "learning_rate": 8.961342429232337e-05, "loss": 2.9331, "step": 13246 }, { "epoch": 1.04, "learning_rate": 8.961264016309888e-05, "loss": 3.1061, "step": 13247 }, { "epoch": 1.04, "learning_rate": 8.961185603387438e-05, "loss": 2.8655, "step": 13248 }, { "epoch": 1.04, "learning_rate": 8.961107190464988e-05, "loss": 2.9105, "step": 13249 }, { "epoch": 1.04, "learning_rate": 8.961028777542539e-05, "loss": 2.8017, "step": 13250 }, { "epoch": 1.04, "learning_rate": 8.96095036462009e-05, "loss": 2.7952, "step": 13251 }, { "epoch": 1.04, "learning_rate": 8.960871951697641e-05, "loss": 2.8827, "step": 13252 }, { "epoch": 1.04, "learning_rate": 8.960793538775191e-05, "loss": 2.9212, "step": 13253 }, { "epoch": 1.04, "learning_rate": 8.960715125852742e-05, "loss": 2.964, "step": 13254 }, { "epoch": 1.04, "learning_rate": 8.960636712930292e-05, "loss": 2.9043, "step": 13255 }, { "epoch": 1.04, "learning_rate": 8.960558300007842e-05, "loss": 2.8857, "step": 13256 }, { "epoch": 1.04, "learning_rate": 8.960479887085393e-05, "loss": 3.0857, "step": 13257 }, { "epoch": 1.04, "learning_rate": 8.960401474162943e-05, "loss": 2.8969, "step": 13258 }, { "epoch": 1.04, "learning_rate": 8.960323061240493e-05, "loss": 2.7691, "step": 13259 }, { "epoch": 1.04, "learning_rate": 8.960244648318044e-05, "loss": 2.8932, "step": 13260 }, { "epoch": 1.04, "learning_rate": 8.960166235395594e-05, "loss": 2.8513, "step": 13261 }, { "epoch": 1.04, "learning_rate": 8.960087822473145e-05, "loss": 2.9538, "step": 13262 }, { "epoch": 1.04, "learning_rate": 8.960009409550695e-05, "loss": 2.8382, "step": 13263 }, { "epoch": 1.04, "learning_rate": 8.959930996628245e-05, "loss": 3.1018, "step": 13264 }, { "epoch": 1.04, "learning_rate": 8.959852583705796e-05, "loss": 2.8243, "step": 13265 }, { "epoch": 1.04, "learning_rate": 8.959774170783346e-05, "loss": 3.0619, "step": 13266 }, { "epoch": 1.04, "learning_rate": 8.959695757860896e-05, "loss": 2.9006, "step": 13267 }, { "epoch": 1.04, "learning_rate": 8.959617344938445e-05, "loss": 2.8649, "step": 13268 }, { "epoch": 1.04, "learning_rate": 8.959538932015996e-05, "loss": 2.6545, "step": 13269 }, { "epoch": 1.04, "learning_rate": 8.959460519093546e-05, "loss": 2.8719, "step": 13270 }, { "epoch": 1.04, "learning_rate": 8.959382106171096e-05, "loss": 2.8798, "step": 13271 }, { "epoch": 1.04, "learning_rate": 8.959303693248647e-05, "loss": 2.8239, "step": 13272 }, { "epoch": 1.04, "learning_rate": 8.959225280326197e-05, "loss": 3.1026, "step": 13273 }, { "epoch": 1.04, "learning_rate": 8.959146867403748e-05, "loss": 2.9419, "step": 13274 }, { "epoch": 1.04, "learning_rate": 8.959068454481298e-05, "loss": 2.9693, "step": 13275 }, { "epoch": 1.04, "learning_rate": 8.95899004155885e-05, "loss": 2.9053, "step": 13276 }, { "epoch": 1.04, "learning_rate": 8.9589116286364e-05, "loss": 2.7017, "step": 13277 }, { "epoch": 1.04, "learning_rate": 8.95883321571395e-05, "loss": 3.098, "step": 13278 }, { "epoch": 1.04, "learning_rate": 8.958754802791501e-05, "loss": 2.8597, "step": 13279 }, { "epoch": 1.04, "learning_rate": 8.958676389869051e-05, "loss": 3.2733, "step": 13280 }, { "epoch": 1.04, "learning_rate": 8.958597976946602e-05, "loss": 2.8901, "step": 13281 }, { "epoch": 1.04, "learning_rate": 8.958519564024152e-05, "loss": 2.9726, "step": 13282 }, { "epoch": 1.04, "learning_rate": 8.958441151101702e-05, "loss": 2.9441, "step": 13283 }, { "epoch": 1.04, "learning_rate": 8.958362738179253e-05, "loss": 2.9985, "step": 13284 }, { "epoch": 1.04, "learning_rate": 8.958284325256803e-05, "loss": 2.6988, "step": 13285 }, { "epoch": 1.04, "learning_rate": 8.958205912334353e-05, "loss": 2.9137, "step": 13286 }, { "epoch": 1.04, "learning_rate": 8.958127499411904e-05, "loss": 2.9023, "step": 13287 }, { "epoch": 1.04, "learning_rate": 8.958049086489454e-05, "loss": 3.0713, "step": 13288 }, { "epoch": 1.04, "learning_rate": 8.957970673567004e-05, "loss": 2.897, "step": 13289 }, { "epoch": 1.04, "learning_rate": 8.957892260644555e-05, "loss": 2.8892, "step": 13290 }, { "epoch": 1.04, "learning_rate": 8.957813847722105e-05, "loss": 2.9761, "step": 13291 }, { "epoch": 1.04, "learning_rate": 8.957735434799656e-05, "loss": 3.0495, "step": 13292 }, { "epoch": 1.04, "learning_rate": 8.957657021877206e-05, "loss": 3.0531, "step": 13293 }, { "epoch": 1.04, "learning_rate": 8.957578608954756e-05, "loss": 2.8644, "step": 13294 }, { "epoch": 1.04, "learning_rate": 8.957500196032307e-05, "loss": 2.7407, "step": 13295 }, { "epoch": 1.04, "learning_rate": 8.957421783109857e-05, "loss": 2.6991, "step": 13296 }, { "epoch": 1.04, "learning_rate": 8.957343370187407e-05, "loss": 2.795, "step": 13297 }, { "epoch": 1.04, "learning_rate": 8.957264957264958e-05, "loss": 2.7655, "step": 13298 }, { "epoch": 1.04, "learning_rate": 8.957186544342508e-05, "loss": 2.5748, "step": 13299 }, { "epoch": 1.04, "learning_rate": 8.957108131420059e-05, "loss": 2.8468, "step": 13300 }, { "epoch": 1.04, "learning_rate": 8.957029718497609e-05, "loss": 2.9263, "step": 13301 }, { "epoch": 1.04, "learning_rate": 8.956951305575159e-05, "loss": 2.7097, "step": 13302 }, { "epoch": 1.04, "learning_rate": 8.95687289265271e-05, "loss": 2.6769, "step": 13303 }, { "epoch": 1.04, "learning_rate": 8.95679447973026e-05, "loss": 2.8754, "step": 13304 }, { "epoch": 1.04, "learning_rate": 8.95671606680781e-05, "loss": 2.8576, "step": 13305 }, { "epoch": 1.04, "learning_rate": 8.956637653885361e-05, "loss": 2.9735, "step": 13306 }, { "epoch": 1.04, "learning_rate": 8.956559240962911e-05, "loss": 3.1532, "step": 13307 }, { "epoch": 1.04, "learning_rate": 8.956480828040461e-05, "loss": 2.83, "step": 13308 }, { "epoch": 1.04, "learning_rate": 8.956402415118012e-05, "loss": 2.7699, "step": 13309 }, { "epoch": 1.04, "learning_rate": 8.956324002195562e-05, "loss": 2.8052, "step": 13310 }, { "epoch": 1.04, "learning_rate": 8.956245589273113e-05, "loss": 3.1713, "step": 13311 }, { "epoch": 1.04, "learning_rate": 8.956167176350663e-05, "loss": 3.1013, "step": 13312 }, { "epoch": 1.04, "learning_rate": 8.956088763428213e-05, "loss": 3.0323, "step": 13313 }, { "epoch": 1.04, "learning_rate": 8.956010350505764e-05, "loss": 3.0037, "step": 13314 }, { "epoch": 1.04, "learning_rate": 8.955931937583314e-05, "loss": 2.8671, "step": 13315 }, { "epoch": 1.04, "learning_rate": 8.955853524660864e-05, "loss": 3.0596, "step": 13316 }, { "epoch": 1.04, "learning_rate": 8.955775111738415e-05, "loss": 2.8561, "step": 13317 }, { "epoch": 1.04, "learning_rate": 8.955696698815965e-05, "loss": 2.7172, "step": 13318 }, { "epoch": 1.04, "learning_rate": 8.955618285893516e-05, "loss": 2.9832, "step": 13319 }, { "epoch": 1.04, "learning_rate": 8.955539872971066e-05, "loss": 2.9452, "step": 13320 }, { "epoch": 1.04, "learning_rate": 8.955461460048616e-05, "loss": 2.7009, "step": 13321 }, { "epoch": 1.04, "learning_rate": 8.955383047126167e-05, "loss": 2.9076, "step": 13322 }, { "epoch": 1.04, "learning_rate": 8.955304634203717e-05, "loss": 2.7329, "step": 13323 }, { "epoch": 1.04, "learning_rate": 8.955226221281267e-05, "loss": 2.9573, "step": 13324 }, { "epoch": 1.04, "learning_rate": 8.955147808358818e-05, "loss": 2.8106, "step": 13325 }, { "epoch": 1.04, "learning_rate": 8.955069395436368e-05, "loss": 2.8436, "step": 13326 }, { "epoch": 1.05, "learning_rate": 8.954990982513918e-05, "loss": 2.8575, "step": 13327 }, { "epoch": 1.05, "learning_rate": 8.954912569591469e-05, "loss": 3.1919, "step": 13328 }, { "epoch": 1.05, "learning_rate": 8.954834156669019e-05, "loss": 2.8373, "step": 13329 }, { "epoch": 1.05, "learning_rate": 8.95475574374657e-05, "loss": 2.8206, "step": 13330 }, { "epoch": 1.05, "learning_rate": 8.95467733082412e-05, "loss": 3.135, "step": 13331 }, { "epoch": 1.05, "learning_rate": 8.95459891790167e-05, "loss": 2.8333, "step": 13332 }, { "epoch": 1.05, "learning_rate": 8.954520504979222e-05, "loss": 2.8921, "step": 13333 }, { "epoch": 1.05, "learning_rate": 8.954442092056772e-05, "loss": 2.8671, "step": 13334 }, { "epoch": 1.05, "learning_rate": 8.954363679134323e-05, "loss": 2.7069, "step": 13335 }, { "epoch": 1.05, "learning_rate": 8.954285266211873e-05, "loss": 2.7902, "step": 13336 }, { "epoch": 1.05, "learning_rate": 8.954206853289423e-05, "loss": 3.1391, "step": 13337 }, { "epoch": 1.05, "learning_rate": 8.954128440366974e-05, "loss": 2.852, "step": 13338 }, { "epoch": 1.05, "learning_rate": 8.954050027444524e-05, "loss": 2.9067, "step": 13339 }, { "epoch": 1.05, "learning_rate": 8.953971614522073e-05, "loss": 3.1355, "step": 13340 }, { "epoch": 1.05, "learning_rate": 8.953893201599624e-05, "loss": 2.7963, "step": 13341 }, { "epoch": 1.05, "learning_rate": 8.953814788677174e-05, "loss": 2.8981, "step": 13342 }, { "epoch": 1.05, "learning_rate": 8.953736375754724e-05, "loss": 2.8765, "step": 13343 }, { "epoch": 1.05, "learning_rate": 8.953657962832275e-05, "loss": 2.9405, "step": 13344 }, { "epoch": 1.05, "learning_rate": 8.953579549909825e-05, "loss": 2.9101, "step": 13345 }, { "epoch": 1.05, "learning_rate": 8.953501136987375e-05, "loss": 3.0008, "step": 13346 }, { "epoch": 1.05, "learning_rate": 8.953422724064926e-05, "loss": 2.849, "step": 13347 }, { "epoch": 1.05, "learning_rate": 8.953344311142476e-05, "loss": 2.9132, "step": 13348 }, { "epoch": 1.05, "learning_rate": 8.953265898220027e-05, "loss": 3.0168, "step": 13349 }, { "epoch": 1.05, "learning_rate": 8.953187485297577e-05, "loss": 2.9689, "step": 13350 }, { "epoch": 1.05, "learning_rate": 8.953109072375127e-05, "loss": 2.9616, "step": 13351 }, { "epoch": 1.05, "learning_rate": 8.953030659452678e-05, "loss": 3.1537, "step": 13352 }, { "epoch": 1.05, "learning_rate": 8.952952246530228e-05, "loss": 2.9391, "step": 13353 }, { "epoch": 1.05, "learning_rate": 8.952873833607778e-05, "loss": 2.8872, "step": 13354 }, { "epoch": 1.05, "learning_rate": 8.952795420685329e-05, "loss": 2.7722, "step": 13355 }, { "epoch": 1.05, "learning_rate": 8.952717007762879e-05, "loss": 2.8667, "step": 13356 }, { "epoch": 1.05, "learning_rate": 8.95263859484043e-05, "loss": 2.86, "step": 13357 }, { "epoch": 1.05, "learning_rate": 8.95256018191798e-05, "loss": 2.8441, "step": 13358 }, { "epoch": 1.05, "learning_rate": 8.952481768995532e-05, "loss": 2.9436, "step": 13359 }, { "epoch": 1.05, "learning_rate": 8.952403356073082e-05, "loss": 2.7075, "step": 13360 }, { "epoch": 1.05, "learning_rate": 8.952324943150632e-05, "loss": 2.7071, "step": 13361 }, { "epoch": 1.05, "learning_rate": 8.952246530228183e-05, "loss": 2.905, "step": 13362 }, { "epoch": 1.05, "learning_rate": 8.952168117305733e-05, "loss": 2.7619, "step": 13363 }, { "epoch": 1.05, "learning_rate": 8.952089704383283e-05, "loss": 2.8502, "step": 13364 }, { "epoch": 1.05, "learning_rate": 8.952011291460834e-05, "loss": 2.7611, "step": 13365 }, { "epoch": 1.05, "learning_rate": 8.951932878538384e-05, "loss": 2.7715, "step": 13366 }, { "epoch": 1.05, "learning_rate": 8.951854465615935e-05, "loss": 2.726, "step": 13367 }, { "epoch": 1.05, "learning_rate": 8.951776052693485e-05, "loss": 2.8009, "step": 13368 }, { "epoch": 1.05, "learning_rate": 8.951697639771035e-05, "loss": 2.8874, "step": 13369 }, { "epoch": 1.05, "learning_rate": 8.951619226848586e-05, "loss": 3.0118, "step": 13370 }, { "epoch": 1.05, "learning_rate": 8.951540813926136e-05, "loss": 3.2904, "step": 13371 }, { "epoch": 1.05, "learning_rate": 8.951462401003686e-05, "loss": 2.8356, "step": 13372 }, { "epoch": 1.05, "learning_rate": 8.951383988081237e-05, "loss": 3.3284, "step": 13373 }, { "epoch": 1.05, "learning_rate": 8.951305575158787e-05, "loss": 3.0357, "step": 13374 }, { "epoch": 1.05, "learning_rate": 8.951227162236337e-05, "loss": 2.7992, "step": 13375 }, { "epoch": 1.05, "learning_rate": 8.951148749313886e-05, "loss": 2.9321, "step": 13376 }, { "epoch": 1.05, "learning_rate": 8.951070336391437e-05, "loss": 3.0933, "step": 13377 }, { "epoch": 1.05, "learning_rate": 8.950991923468987e-05, "loss": 3.0781, "step": 13378 }, { "epoch": 1.05, "learning_rate": 8.950913510546538e-05, "loss": 3.1172, "step": 13379 }, { "epoch": 1.05, "learning_rate": 8.950835097624088e-05, "loss": 2.9346, "step": 13380 }, { "epoch": 1.05, "learning_rate": 8.950756684701638e-05, "loss": 2.8786, "step": 13381 }, { "epoch": 1.05, "learning_rate": 8.950678271779189e-05, "loss": 3.1487, "step": 13382 }, { "epoch": 1.05, "learning_rate": 8.950599858856739e-05, "loss": 2.6352, "step": 13383 }, { "epoch": 1.05, "learning_rate": 8.950521445934291e-05, "loss": 2.7269, "step": 13384 }, { "epoch": 1.05, "learning_rate": 8.950443033011841e-05, "loss": 2.8944, "step": 13385 }, { "epoch": 1.05, "learning_rate": 8.950364620089391e-05, "loss": 3.1539, "step": 13386 }, { "epoch": 1.05, "learning_rate": 8.950286207166942e-05, "loss": 2.8565, "step": 13387 }, { "epoch": 1.05, "learning_rate": 8.950207794244492e-05, "loss": 2.7579, "step": 13388 }, { "epoch": 1.05, "learning_rate": 8.950129381322043e-05, "loss": 3.029, "step": 13389 }, { "epoch": 1.05, "learning_rate": 8.950050968399593e-05, "loss": 2.9742, "step": 13390 }, { "epoch": 1.05, "learning_rate": 8.949972555477143e-05, "loss": 3.101, "step": 13391 }, { "epoch": 1.05, "learning_rate": 8.949894142554694e-05, "loss": 2.8087, "step": 13392 }, { "epoch": 1.05, "learning_rate": 8.949815729632244e-05, "loss": 3.0029, "step": 13393 }, { "epoch": 1.05, "learning_rate": 8.949737316709794e-05, "loss": 3.1682, "step": 13394 }, { "epoch": 1.05, "learning_rate": 8.949658903787345e-05, "loss": 3.2792, "step": 13395 }, { "epoch": 1.05, "learning_rate": 8.949580490864895e-05, "loss": 2.7537, "step": 13396 }, { "epoch": 1.05, "learning_rate": 8.949502077942446e-05, "loss": 3.0368, "step": 13397 }, { "epoch": 1.05, "learning_rate": 8.949423665019996e-05, "loss": 3.0561, "step": 13398 }, { "epoch": 1.05, "learning_rate": 8.949345252097546e-05, "loss": 2.7118, "step": 13399 }, { "epoch": 1.05, "learning_rate": 8.949266839175097e-05, "loss": 2.7251, "step": 13400 }, { "epoch": 1.05, "learning_rate": 8.949188426252647e-05, "loss": 2.9296, "step": 13401 }, { "epoch": 1.05, "learning_rate": 8.949110013330197e-05, "loss": 2.7943, "step": 13402 }, { "epoch": 1.05, "learning_rate": 8.949031600407748e-05, "loss": 2.9152, "step": 13403 }, { "epoch": 1.05, "learning_rate": 8.948953187485298e-05, "loss": 2.8886, "step": 13404 }, { "epoch": 1.05, "learning_rate": 8.948874774562848e-05, "loss": 2.8841, "step": 13405 }, { "epoch": 1.05, "learning_rate": 8.948796361640399e-05, "loss": 2.8581, "step": 13406 }, { "epoch": 1.05, "learning_rate": 8.948717948717949e-05, "loss": 2.9357, "step": 13407 }, { "epoch": 1.05, "learning_rate": 8.9486395357955e-05, "loss": 3.1149, "step": 13408 }, { "epoch": 1.05, "learning_rate": 8.94856112287305e-05, "loss": 2.8003, "step": 13409 }, { "epoch": 1.05, "learning_rate": 8.9484827099506e-05, "loss": 2.7483, "step": 13410 }, { "epoch": 1.05, "learning_rate": 8.948404297028151e-05, "loss": 2.9001, "step": 13411 }, { "epoch": 1.05, "learning_rate": 8.948325884105701e-05, "loss": 2.873, "step": 13412 }, { "epoch": 1.05, "learning_rate": 8.948247471183251e-05, "loss": 2.9076, "step": 13413 }, { "epoch": 1.05, "learning_rate": 8.948169058260802e-05, "loss": 2.639, "step": 13414 }, { "epoch": 1.05, "learning_rate": 8.948090645338352e-05, "loss": 2.9929, "step": 13415 }, { "epoch": 1.05, "learning_rate": 8.948012232415903e-05, "loss": 2.8429, "step": 13416 }, { "epoch": 1.05, "learning_rate": 8.947933819493453e-05, "loss": 3.0603, "step": 13417 }, { "epoch": 1.05, "learning_rate": 8.947855406571003e-05, "loss": 3.093, "step": 13418 }, { "epoch": 1.05, "learning_rate": 8.947776993648554e-05, "loss": 2.9304, "step": 13419 }, { "epoch": 1.05, "learning_rate": 8.947698580726104e-05, "loss": 3.0255, "step": 13420 }, { "epoch": 1.05, "learning_rate": 8.947620167803654e-05, "loss": 2.712, "step": 13421 }, { "epoch": 1.05, "learning_rate": 8.947541754881205e-05, "loss": 2.9966, "step": 13422 }, { "epoch": 1.05, "learning_rate": 8.947463341958755e-05, "loss": 2.9843, "step": 13423 }, { "epoch": 1.05, "learning_rate": 8.947384929036305e-05, "loss": 2.9608, "step": 13424 }, { "epoch": 1.05, "learning_rate": 8.947306516113856e-05, "loss": 3.1472, "step": 13425 }, { "epoch": 1.05, "learning_rate": 8.947228103191406e-05, "loss": 2.9593, "step": 13426 }, { "epoch": 1.05, "learning_rate": 8.947149690268957e-05, "loss": 2.8338, "step": 13427 }, { "epoch": 1.05, "learning_rate": 8.947071277346507e-05, "loss": 2.8848, "step": 13428 }, { "epoch": 1.05, "learning_rate": 8.946992864424057e-05, "loss": 3.0426, "step": 13429 }, { "epoch": 1.05, "learning_rate": 8.946914451501608e-05, "loss": 2.6862, "step": 13430 }, { "epoch": 1.05, "learning_rate": 8.946836038579158e-05, "loss": 2.7598, "step": 13431 }, { "epoch": 1.05, "learning_rate": 8.946757625656708e-05, "loss": 2.688, "step": 13432 }, { "epoch": 1.05, "learning_rate": 8.946679212734259e-05, "loss": 2.6218, "step": 13433 }, { "epoch": 1.05, "learning_rate": 8.946600799811809e-05, "loss": 2.8105, "step": 13434 }, { "epoch": 1.05, "learning_rate": 8.94652238688936e-05, "loss": 2.9558, "step": 13435 }, { "epoch": 1.05, "learning_rate": 8.94644397396691e-05, "loss": 2.8223, "step": 13436 }, { "epoch": 1.05, "learning_rate": 8.94636556104446e-05, "loss": 3.0835, "step": 13437 }, { "epoch": 1.05, "learning_rate": 8.94628714812201e-05, "loss": 2.7764, "step": 13438 }, { "epoch": 1.05, "learning_rate": 8.946208735199561e-05, "loss": 2.5851, "step": 13439 }, { "epoch": 1.05, "learning_rate": 8.946130322277111e-05, "loss": 2.9569, "step": 13440 }, { "epoch": 1.05, "learning_rate": 8.946051909354663e-05, "loss": 2.979, "step": 13441 }, { "epoch": 1.05, "learning_rate": 8.945973496432213e-05, "loss": 3.0675, "step": 13442 }, { "epoch": 1.05, "learning_rate": 8.945895083509764e-05, "loss": 2.925, "step": 13443 }, { "epoch": 1.05, "learning_rate": 8.945816670587314e-05, "loss": 2.7986, "step": 13444 }, { "epoch": 1.05, "learning_rate": 8.945738257664865e-05, "loss": 2.8888, "step": 13445 }, { "epoch": 1.05, "learning_rate": 8.945659844742415e-05, "loss": 3.1384, "step": 13446 }, { "epoch": 1.05, "learning_rate": 8.945581431819965e-05, "loss": 2.9221, "step": 13447 }, { "epoch": 1.05, "learning_rate": 8.945503018897514e-05, "loss": 2.8813, "step": 13448 }, { "epoch": 1.05, "learning_rate": 8.945424605975065e-05, "loss": 3.0522, "step": 13449 }, { "epoch": 1.05, "learning_rate": 8.945346193052615e-05, "loss": 2.6629, "step": 13450 }, { "epoch": 1.05, "learning_rate": 8.945267780130165e-05, "loss": 2.8002, "step": 13451 }, { "epoch": 1.05, "learning_rate": 8.945189367207716e-05, "loss": 2.9563, "step": 13452 }, { "epoch": 1.05, "learning_rate": 8.945110954285266e-05, "loss": 2.9715, "step": 13453 }, { "epoch": 1.05, "learning_rate": 8.945032541362817e-05, "loss": 2.7817, "step": 13454 }, { "epoch": 1.06, "learning_rate": 8.944954128440367e-05, "loss": 3.0818, "step": 13455 }, { "epoch": 1.06, "learning_rate": 8.944875715517917e-05, "loss": 3.0619, "step": 13456 }, { "epoch": 1.06, "learning_rate": 8.944797302595468e-05, "loss": 2.8046, "step": 13457 }, { "epoch": 1.06, "learning_rate": 8.944718889673018e-05, "loss": 2.8903, "step": 13458 }, { "epoch": 1.06, "learning_rate": 8.944640476750568e-05, "loss": 2.8988, "step": 13459 }, { "epoch": 1.06, "learning_rate": 8.944562063828119e-05, "loss": 3.0713, "step": 13460 }, { "epoch": 1.06, "learning_rate": 8.944483650905669e-05, "loss": 2.7388, "step": 13461 }, { "epoch": 1.06, "learning_rate": 8.94440523798322e-05, "loss": 2.7925, "step": 13462 }, { "epoch": 1.06, "learning_rate": 8.94432682506077e-05, "loss": 2.989, "step": 13463 }, { "epoch": 1.06, "learning_rate": 8.94424841213832e-05, "loss": 2.7778, "step": 13464 }, { "epoch": 1.06, "learning_rate": 8.94416999921587e-05, "loss": 3.0457, "step": 13465 }, { "epoch": 1.06, "learning_rate": 8.944091586293421e-05, "loss": 2.8955, "step": 13466 }, { "epoch": 1.06, "learning_rate": 8.944013173370973e-05, "loss": 3.0716, "step": 13467 }, { "epoch": 1.06, "learning_rate": 8.943934760448523e-05, "loss": 2.9967, "step": 13468 }, { "epoch": 1.06, "learning_rate": 8.943856347526073e-05, "loss": 2.7514, "step": 13469 }, { "epoch": 1.06, "learning_rate": 8.943777934603624e-05, "loss": 3.0477, "step": 13470 }, { "epoch": 1.06, "learning_rate": 8.943699521681174e-05, "loss": 2.8683, "step": 13471 }, { "epoch": 1.06, "learning_rate": 8.943621108758724e-05, "loss": 2.9022, "step": 13472 }, { "epoch": 1.06, "learning_rate": 8.943542695836275e-05, "loss": 3.1517, "step": 13473 }, { "epoch": 1.06, "learning_rate": 8.943464282913825e-05, "loss": 2.8223, "step": 13474 }, { "epoch": 1.06, "learning_rate": 8.943385869991376e-05, "loss": 3.1868, "step": 13475 }, { "epoch": 1.06, "learning_rate": 8.943307457068926e-05, "loss": 3.1597, "step": 13476 }, { "epoch": 1.06, "learning_rate": 8.943229044146476e-05, "loss": 3.0709, "step": 13477 }, { "epoch": 1.06, "learning_rate": 8.943150631224027e-05, "loss": 2.8115, "step": 13478 }, { "epoch": 1.06, "learning_rate": 8.943072218301577e-05, "loss": 2.7129, "step": 13479 }, { "epoch": 1.06, "learning_rate": 8.942993805379127e-05, "loss": 3.1767, "step": 13480 }, { "epoch": 1.06, "learning_rate": 8.942915392456678e-05, "loss": 2.7862, "step": 13481 }, { "epoch": 1.06, "learning_rate": 8.942836979534228e-05, "loss": 3.1327, "step": 13482 }, { "epoch": 1.06, "learning_rate": 8.942758566611779e-05, "loss": 2.7479, "step": 13483 }, { "epoch": 1.06, "learning_rate": 8.942680153689328e-05, "loss": 2.8922, "step": 13484 }, { "epoch": 1.06, "learning_rate": 8.942601740766878e-05, "loss": 3.0193, "step": 13485 }, { "epoch": 1.06, "learning_rate": 8.942523327844428e-05, "loss": 3.2726, "step": 13486 }, { "epoch": 1.06, "learning_rate": 8.942444914921979e-05, "loss": 2.6496, "step": 13487 }, { "epoch": 1.06, "learning_rate": 8.942366501999529e-05, "loss": 2.8985, "step": 13488 }, { "epoch": 1.06, "learning_rate": 8.94228808907708e-05, "loss": 3.0044, "step": 13489 }, { "epoch": 1.06, "learning_rate": 8.94220967615463e-05, "loss": 3.1099, "step": 13490 }, { "epoch": 1.06, "learning_rate": 8.94213126323218e-05, "loss": 2.6913, "step": 13491 }, { "epoch": 1.06, "learning_rate": 8.942052850309732e-05, "loss": 2.9196, "step": 13492 }, { "epoch": 1.06, "learning_rate": 8.941974437387282e-05, "loss": 2.9207, "step": 13493 }, { "epoch": 1.06, "learning_rate": 8.941896024464833e-05, "loss": 2.9227, "step": 13494 }, { "epoch": 1.06, "learning_rate": 8.941817611542383e-05, "loss": 2.8936, "step": 13495 }, { "epoch": 1.06, "learning_rate": 8.941739198619933e-05, "loss": 2.9383, "step": 13496 }, { "epoch": 1.06, "learning_rate": 8.941660785697484e-05, "loss": 2.9409, "step": 13497 }, { "epoch": 1.06, "learning_rate": 8.941582372775034e-05, "loss": 2.8543, "step": 13498 }, { "epoch": 1.06, "learning_rate": 8.941503959852584e-05, "loss": 2.8227, "step": 13499 }, { "epoch": 1.06, "learning_rate": 8.941425546930135e-05, "loss": 3.0102, "step": 13500 }, { "epoch": 1.06, "learning_rate": 8.941347134007685e-05, "loss": 2.8471, "step": 13501 }, { "epoch": 1.06, "learning_rate": 8.941268721085236e-05, "loss": 3.0486, "step": 13502 }, { "epoch": 1.06, "learning_rate": 8.941190308162786e-05, "loss": 2.8613, "step": 13503 }, { "epoch": 1.06, "learning_rate": 8.941111895240336e-05, "loss": 2.8806, "step": 13504 }, { "epoch": 1.06, "learning_rate": 8.941033482317887e-05, "loss": 2.7798, "step": 13505 }, { "epoch": 1.06, "learning_rate": 8.940955069395437e-05, "loss": 2.9034, "step": 13506 }, { "epoch": 1.06, "learning_rate": 8.940876656472987e-05, "loss": 2.8471, "step": 13507 }, { "epoch": 1.06, "learning_rate": 8.940798243550538e-05, "loss": 2.9119, "step": 13508 }, { "epoch": 1.06, "learning_rate": 8.940719830628088e-05, "loss": 3.0183, "step": 13509 }, { "epoch": 1.06, "learning_rate": 8.940641417705638e-05, "loss": 3.001, "step": 13510 }, { "epoch": 1.06, "learning_rate": 8.940563004783189e-05, "loss": 3.0314, "step": 13511 }, { "epoch": 1.06, "learning_rate": 8.940484591860739e-05, "loss": 3.0469, "step": 13512 }, { "epoch": 1.06, "learning_rate": 8.94040617893829e-05, "loss": 2.8545, "step": 13513 }, { "epoch": 1.06, "learning_rate": 8.94032776601584e-05, "loss": 2.8614, "step": 13514 }, { "epoch": 1.06, "learning_rate": 8.94024935309339e-05, "loss": 2.9648, "step": 13515 }, { "epoch": 1.06, "learning_rate": 8.94017094017094e-05, "loss": 3.0425, "step": 13516 }, { "epoch": 1.06, "learning_rate": 8.940092527248491e-05, "loss": 2.9808, "step": 13517 }, { "epoch": 1.06, "learning_rate": 8.940014114326041e-05, "loss": 2.8018, "step": 13518 }, { "epoch": 1.06, "learning_rate": 8.939935701403592e-05, "loss": 2.7802, "step": 13519 }, { "epoch": 1.06, "learning_rate": 8.939857288481142e-05, "loss": 2.9396, "step": 13520 }, { "epoch": 1.06, "learning_rate": 8.939778875558693e-05, "loss": 2.9854, "step": 13521 }, { "epoch": 1.06, "learning_rate": 8.939700462636243e-05, "loss": 2.8095, "step": 13522 }, { "epoch": 1.06, "learning_rate": 8.939622049713793e-05, "loss": 3.1081, "step": 13523 }, { "epoch": 1.06, "learning_rate": 8.939543636791344e-05, "loss": 2.9742, "step": 13524 }, { "epoch": 1.06, "learning_rate": 8.939465223868894e-05, "loss": 3.1714, "step": 13525 }, { "epoch": 1.06, "learning_rate": 8.939386810946444e-05, "loss": 2.781, "step": 13526 }, { "epoch": 1.06, "learning_rate": 8.939308398023995e-05, "loss": 2.8901, "step": 13527 }, { "epoch": 1.06, "learning_rate": 8.939229985101545e-05, "loss": 2.881, "step": 13528 }, { "epoch": 1.06, "learning_rate": 8.939151572179095e-05, "loss": 2.9744, "step": 13529 }, { "epoch": 1.06, "learning_rate": 8.939073159256646e-05, "loss": 3.2209, "step": 13530 }, { "epoch": 1.06, "learning_rate": 8.938994746334196e-05, "loss": 2.9749, "step": 13531 }, { "epoch": 1.06, "learning_rate": 8.938916333411747e-05, "loss": 2.9327, "step": 13532 }, { "epoch": 1.06, "learning_rate": 8.938837920489297e-05, "loss": 2.8978, "step": 13533 }, { "epoch": 1.06, "learning_rate": 8.938759507566847e-05, "loss": 2.8373, "step": 13534 }, { "epoch": 1.06, "learning_rate": 8.938681094644398e-05, "loss": 2.974, "step": 13535 }, { "epoch": 1.06, "learning_rate": 8.938602681721948e-05, "loss": 3.3131, "step": 13536 }, { "epoch": 1.06, "learning_rate": 8.938524268799498e-05, "loss": 2.8664, "step": 13537 }, { "epoch": 1.06, "learning_rate": 8.938445855877049e-05, "loss": 3.0973, "step": 13538 }, { "epoch": 1.06, "learning_rate": 8.938367442954599e-05, "loss": 2.9971, "step": 13539 }, { "epoch": 1.06, "learning_rate": 8.93828903003215e-05, "loss": 2.5594, "step": 13540 }, { "epoch": 1.06, "learning_rate": 8.9382106171097e-05, "loss": 2.7218, "step": 13541 }, { "epoch": 1.06, "learning_rate": 8.93813220418725e-05, "loss": 2.7925, "step": 13542 }, { "epoch": 1.06, "learning_rate": 8.9380537912648e-05, "loss": 2.9876, "step": 13543 }, { "epoch": 1.06, "learning_rate": 8.937975378342351e-05, "loss": 2.8717, "step": 13544 }, { "epoch": 1.06, "learning_rate": 8.937896965419901e-05, "loss": 2.785, "step": 13545 }, { "epoch": 1.06, "learning_rate": 8.937818552497452e-05, "loss": 2.9, "step": 13546 }, { "epoch": 1.06, "learning_rate": 8.937740139575002e-05, "loss": 2.8556, "step": 13547 }, { "epoch": 1.06, "learning_rate": 8.937661726652552e-05, "loss": 3.1121, "step": 13548 }, { "epoch": 1.06, "learning_rate": 8.937583313730104e-05, "loss": 2.9373, "step": 13549 }, { "epoch": 1.06, "learning_rate": 8.937504900807655e-05, "loss": 2.9305, "step": 13550 }, { "epoch": 1.06, "learning_rate": 8.937426487885205e-05, "loss": 3.0357, "step": 13551 }, { "epoch": 1.06, "learning_rate": 8.937348074962755e-05, "loss": 3.0096, "step": 13552 }, { "epoch": 1.06, "learning_rate": 8.937269662040306e-05, "loss": 2.8586, "step": 13553 }, { "epoch": 1.06, "learning_rate": 8.937191249117856e-05, "loss": 2.9051, "step": 13554 }, { "epoch": 1.06, "learning_rate": 8.937112836195406e-05, "loss": 2.929, "step": 13555 }, { "epoch": 1.06, "learning_rate": 8.937034423272955e-05, "loss": 2.8336, "step": 13556 }, { "epoch": 1.06, "learning_rate": 8.936956010350506e-05, "loss": 3.0278, "step": 13557 }, { "epoch": 1.06, "learning_rate": 8.936877597428056e-05, "loss": 2.9119, "step": 13558 }, { "epoch": 1.06, "learning_rate": 8.936799184505606e-05, "loss": 2.9113, "step": 13559 }, { "epoch": 1.06, "learning_rate": 8.936720771583157e-05, "loss": 2.8855, "step": 13560 }, { "epoch": 1.06, "learning_rate": 8.936642358660707e-05, "loss": 2.7361, "step": 13561 }, { "epoch": 1.06, "learning_rate": 8.936563945738258e-05, "loss": 2.9373, "step": 13562 }, { "epoch": 1.06, "learning_rate": 8.936485532815808e-05, "loss": 2.8825, "step": 13563 }, { "epoch": 1.06, "learning_rate": 8.936407119893358e-05, "loss": 2.7292, "step": 13564 }, { "epoch": 1.06, "learning_rate": 8.936328706970909e-05, "loss": 2.7378, "step": 13565 }, { "epoch": 1.06, "learning_rate": 8.936250294048459e-05, "loss": 2.8751, "step": 13566 }, { "epoch": 1.06, "learning_rate": 8.93617188112601e-05, "loss": 2.7598, "step": 13567 }, { "epoch": 1.06, "learning_rate": 8.93609346820356e-05, "loss": 3.1018, "step": 13568 }, { "epoch": 1.06, "learning_rate": 8.93601505528111e-05, "loss": 2.8041, "step": 13569 }, { "epoch": 1.06, "learning_rate": 8.93593664235866e-05, "loss": 3.1595, "step": 13570 }, { "epoch": 1.06, "learning_rate": 8.935858229436211e-05, "loss": 2.7883, "step": 13571 }, { "epoch": 1.06, "learning_rate": 8.935779816513761e-05, "loss": 3.0567, "step": 13572 }, { "epoch": 1.06, "learning_rate": 8.935701403591312e-05, "loss": 3.0675, "step": 13573 }, { "epoch": 1.06, "learning_rate": 8.935622990668862e-05, "loss": 2.764, "step": 13574 }, { "epoch": 1.06, "learning_rate": 8.935544577746414e-05, "loss": 3.027, "step": 13575 }, { "epoch": 1.06, "learning_rate": 8.935466164823964e-05, "loss": 3.2874, "step": 13576 }, { "epoch": 1.06, "learning_rate": 8.935387751901514e-05, "loss": 2.8457, "step": 13577 }, { "epoch": 1.06, "learning_rate": 8.935309338979065e-05, "loss": 2.8294, "step": 13578 }, { "epoch": 1.06, "learning_rate": 8.935230926056615e-05, "loss": 2.8792, "step": 13579 }, { "epoch": 1.06, "learning_rate": 8.935152513134166e-05, "loss": 2.9882, "step": 13580 }, { "epoch": 1.06, "learning_rate": 8.935074100211716e-05, "loss": 2.9552, "step": 13581 }, { "epoch": 1.07, "learning_rate": 8.934995687289266e-05, "loss": 2.7379, "step": 13582 }, { "epoch": 1.07, "learning_rate": 8.934917274366817e-05, "loss": 2.6787, "step": 13583 }, { "epoch": 1.07, "learning_rate": 8.934838861444367e-05, "loss": 3.025, "step": 13584 }, { "epoch": 1.07, "learning_rate": 8.934760448521917e-05, "loss": 3.0273, "step": 13585 }, { "epoch": 1.07, "learning_rate": 8.934682035599468e-05, "loss": 2.7696, "step": 13586 }, { "epoch": 1.07, "learning_rate": 8.934603622677018e-05, "loss": 2.9974, "step": 13587 }, { "epoch": 1.07, "learning_rate": 8.934525209754568e-05, "loss": 3.041, "step": 13588 }, { "epoch": 1.07, "learning_rate": 8.934446796832119e-05, "loss": 2.7879, "step": 13589 }, { "epoch": 1.07, "learning_rate": 8.934368383909669e-05, "loss": 2.7693, "step": 13590 }, { "epoch": 1.07, "learning_rate": 8.93428997098722e-05, "loss": 2.6379, "step": 13591 }, { "epoch": 1.07, "learning_rate": 8.934211558064769e-05, "loss": 3.0134, "step": 13592 }, { "epoch": 1.07, "learning_rate": 8.934133145142319e-05, "loss": 2.9722, "step": 13593 }, { "epoch": 1.07, "learning_rate": 8.93405473221987e-05, "loss": 2.8315, "step": 13594 }, { "epoch": 1.07, "learning_rate": 8.93397631929742e-05, "loss": 2.656, "step": 13595 }, { "epoch": 1.07, "learning_rate": 8.93389790637497e-05, "loss": 3.101, "step": 13596 }, { "epoch": 1.07, "learning_rate": 8.93381949345252e-05, "loss": 2.9914, "step": 13597 }, { "epoch": 1.07, "learning_rate": 8.933741080530071e-05, "loss": 2.9166, "step": 13598 }, { "epoch": 1.07, "learning_rate": 8.933662667607621e-05, "loss": 2.8948, "step": 13599 }, { "epoch": 1.07, "learning_rate": 8.933584254685173e-05, "loss": 2.626, "step": 13600 }, { "epoch": 1.07, "learning_rate": 8.933505841762723e-05, "loss": 2.9477, "step": 13601 }, { "epoch": 1.07, "learning_rate": 8.933427428840274e-05, "loss": 2.932, "step": 13602 }, { "epoch": 1.07, "learning_rate": 8.933349015917824e-05, "loss": 2.8429, "step": 13603 }, { "epoch": 1.07, "learning_rate": 8.933270602995374e-05, "loss": 2.8126, "step": 13604 }, { "epoch": 1.07, "learning_rate": 8.933192190072925e-05, "loss": 2.877, "step": 13605 }, { "epoch": 1.07, "learning_rate": 8.933113777150475e-05, "loss": 3.1544, "step": 13606 }, { "epoch": 1.07, "learning_rate": 8.933035364228025e-05, "loss": 2.8168, "step": 13607 }, { "epoch": 1.07, "learning_rate": 8.932956951305576e-05, "loss": 2.8782, "step": 13608 }, { "epoch": 1.07, "learning_rate": 8.932878538383126e-05, "loss": 2.973, "step": 13609 }, { "epoch": 1.07, "learning_rate": 8.932800125460677e-05, "loss": 2.8607, "step": 13610 }, { "epoch": 1.07, "learning_rate": 8.932721712538227e-05, "loss": 3.0761, "step": 13611 }, { "epoch": 1.07, "learning_rate": 8.932643299615777e-05, "loss": 3.2846, "step": 13612 }, { "epoch": 1.07, "learning_rate": 8.932564886693328e-05, "loss": 2.8453, "step": 13613 }, { "epoch": 1.07, "learning_rate": 8.932486473770878e-05, "loss": 2.8237, "step": 13614 }, { "epoch": 1.07, "learning_rate": 8.932408060848428e-05, "loss": 3.2108, "step": 13615 }, { "epoch": 1.07, "learning_rate": 8.932329647925979e-05, "loss": 2.7467, "step": 13616 }, { "epoch": 1.07, "learning_rate": 8.932251235003529e-05, "loss": 2.617, "step": 13617 }, { "epoch": 1.07, "learning_rate": 8.93217282208108e-05, "loss": 2.964, "step": 13618 }, { "epoch": 1.07, "learning_rate": 8.93209440915863e-05, "loss": 2.7604, "step": 13619 }, { "epoch": 1.07, "learning_rate": 8.93201599623618e-05, "loss": 3.1178, "step": 13620 }, { "epoch": 1.07, "learning_rate": 8.93193758331373e-05, "loss": 2.7723, "step": 13621 }, { "epoch": 1.07, "learning_rate": 8.931859170391281e-05, "loss": 2.8105, "step": 13622 }, { "epoch": 1.07, "learning_rate": 8.931780757468831e-05, "loss": 3.0357, "step": 13623 }, { "epoch": 1.07, "learning_rate": 8.931702344546382e-05, "loss": 2.8926, "step": 13624 }, { "epoch": 1.07, "learning_rate": 8.931623931623932e-05, "loss": 2.9349, "step": 13625 }, { "epoch": 1.07, "learning_rate": 8.931545518701482e-05, "loss": 3.0467, "step": 13626 }, { "epoch": 1.07, "learning_rate": 8.931467105779033e-05, "loss": 2.9079, "step": 13627 }, { "epoch": 1.07, "learning_rate": 8.931388692856583e-05, "loss": 2.8702, "step": 13628 }, { "epoch": 1.07, "learning_rate": 8.931310279934134e-05, "loss": 3.0533, "step": 13629 }, { "epoch": 1.07, "learning_rate": 8.931231867011684e-05, "loss": 2.9743, "step": 13630 }, { "epoch": 1.07, "learning_rate": 8.931153454089234e-05, "loss": 2.7771, "step": 13631 }, { "epoch": 1.07, "learning_rate": 8.931075041166785e-05, "loss": 3.4068, "step": 13632 }, { "epoch": 1.07, "learning_rate": 8.930996628244335e-05, "loss": 2.822, "step": 13633 }, { "epoch": 1.07, "learning_rate": 8.930918215321885e-05, "loss": 2.9991, "step": 13634 }, { "epoch": 1.07, "learning_rate": 8.930839802399436e-05, "loss": 2.6331, "step": 13635 }, { "epoch": 1.07, "learning_rate": 8.930761389476986e-05, "loss": 2.8619, "step": 13636 }, { "epoch": 1.07, "learning_rate": 8.930682976554537e-05, "loss": 2.9338, "step": 13637 }, { "epoch": 1.07, "learning_rate": 8.930604563632087e-05, "loss": 2.6699, "step": 13638 }, { "epoch": 1.07, "learning_rate": 8.930526150709637e-05, "loss": 3.0091, "step": 13639 }, { "epoch": 1.07, "learning_rate": 8.930447737787188e-05, "loss": 2.6993, "step": 13640 }, { "epoch": 1.07, "learning_rate": 8.930369324864738e-05, "loss": 2.9205, "step": 13641 }, { "epoch": 1.07, "learning_rate": 8.930290911942288e-05, "loss": 2.9946, "step": 13642 }, { "epoch": 1.07, "learning_rate": 8.930212499019839e-05, "loss": 2.9154, "step": 13643 }, { "epoch": 1.07, "learning_rate": 8.930134086097389e-05, "loss": 3.2295, "step": 13644 }, { "epoch": 1.07, "learning_rate": 8.93005567317494e-05, "loss": 3.1518, "step": 13645 }, { "epoch": 1.07, "learning_rate": 8.92997726025249e-05, "loss": 3.0531, "step": 13646 }, { "epoch": 1.07, "learning_rate": 8.92989884733004e-05, "loss": 2.9048, "step": 13647 }, { "epoch": 1.07, "learning_rate": 8.92982043440759e-05, "loss": 2.7447, "step": 13648 }, { "epoch": 1.07, "learning_rate": 8.929742021485141e-05, "loss": 2.8275, "step": 13649 }, { "epoch": 1.07, "learning_rate": 8.929663608562691e-05, "loss": 3.0982, "step": 13650 }, { "epoch": 1.07, "learning_rate": 8.929585195640242e-05, "loss": 2.8323, "step": 13651 }, { "epoch": 1.07, "learning_rate": 8.929506782717792e-05, "loss": 2.8909, "step": 13652 }, { "epoch": 1.07, "learning_rate": 8.929428369795342e-05, "loss": 2.9083, "step": 13653 }, { "epoch": 1.07, "learning_rate": 8.929349956872893e-05, "loss": 2.6511, "step": 13654 }, { "epoch": 1.07, "learning_rate": 8.929271543950443e-05, "loss": 2.9461, "step": 13655 }, { "epoch": 1.07, "learning_rate": 8.929193131027994e-05, "loss": 2.6653, "step": 13656 }, { "epoch": 1.07, "learning_rate": 8.929114718105545e-05, "loss": 2.7687, "step": 13657 }, { "epoch": 1.07, "learning_rate": 8.929036305183096e-05, "loss": 3.0359, "step": 13658 }, { "epoch": 1.07, "learning_rate": 8.928957892260646e-05, "loss": 2.9695, "step": 13659 }, { "epoch": 1.07, "learning_rate": 8.928879479338196e-05, "loss": 2.5011, "step": 13660 }, { "epoch": 1.07, "learning_rate": 8.928801066415747e-05, "loss": 2.7758, "step": 13661 }, { "epoch": 1.07, "learning_rate": 8.928722653493297e-05, "loss": 2.7455, "step": 13662 }, { "epoch": 1.07, "learning_rate": 8.928644240570847e-05, "loss": 2.8701, "step": 13663 }, { "epoch": 1.07, "learning_rate": 8.928565827648396e-05, "loss": 2.7701, "step": 13664 }, { "epoch": 1.07, "learning_rate": 8.928487414725947e-05, "loss": 2.9646, "step": 13665 }, { "epoch": 1.07, "learning_rate": 8.928409001803497e-05, "loss": 2.8262, "step": 13666 }, { "epoch": 1.07, "learning_rate": 8.928330588881048e-05, "loss": 2.922, "step": 13667 }, { "epoch": 1.07, "learning_rate": 8.928252175958598e-05, "loss": 2.9521, "step": 13668 }, { "epoch": 1.07, "learning_rate": 8.928173763036148e-05, "loss": 2.9097, "step": 13669 }, { "epoch": 1.07, "learning_rate": 8.928095350113699e-05, "loss": 2.97, "step": 13670 }, { "epoch": 1.07, "learning_rate": 8.928016937191249e-05, "loss": 2.884, "step": 13671 }, { "epoch": 1.07, "learning_rate": 8.9279385242688e-05, "loss": 3.1041, "step": 13672 }, { "epoch": 1.07, "learning_rate": 8.92786011134635e-05, "loss": 3.0492, "step": 13673 }, { "epoch": 1.07, "learning_rate": 8.9277816984239e-05, "loss": 2.7753, "step": 13674 }, { "epoch": 1.07, "learning_rate": 8.92770328550145e-05, "loss": 3.1772, "step": 13675 }, { "epoch": 1.07, "learning_rate": 8.927624872579001e-05, "loss": 2.7546, "step": 13676 }, { "epoch": 1.07, "learning_rate": 8.927546459656551e-05, "loss": 3.112, "step": 13677 }, { "epoch": 1.07, "learning_rate": 8.927468046734102e-05, "loss": 3.1393, "step": 13678 }, { "epoch": 1.07, "learning_rate": 8.927389633811652e-05, "loss": 3.0473, "step": 13679 }, { "epoch": 1.07, "learning_rate": 8.927311220889202e-05, "loss": 2.9963, "step": 13680 }, { "epoch": 1.07, "learning_rate": 8.927232807966753e-05, "loss": 2.9466, "step": 13681 }, { "epoch": 1.07, "learning_rate": 8.927154395044303e-05, "loss": 2.8297, "step": 13682 }, { "epoch": 1.07, "learning_rate": 8.927075982121855e-05, "loss": 2.8848, "step": 13683 }, { "epoch": 1.07, "learning_rate": 8.926997569199405e-05, "loss": 2.7566, "step": 13684 }, { "epoch": 1.07, "learning_rate": 8.926919156276956e-05, "loss": 3.1346, "step": 13685 }, { "epoch": 1.07, "learning_rate": 8.926840743354506e-05, "loss": 2.927, "step": 13686 }, { "epoch": 1.07, "learning_rate": 8.926762330432056e-05, "loss": 2.9404, "step": 13687 }, { "epoch": 1.07, "learning_rate": 8.926683917509607e-05, "loss": 2.8211, "step": 13688 }, { "epoch": 1.07, "learning_rate": 8.926605504587157e-05, "loss": 3.0344, "step": 13689 }, { "epoch": 1.07, "learning_rate": 8.926527091664707e-05, "loss": 2.7967, "step": 13690 }, { "epoch": 1.07, "learning_rate": 8.926448678742258e-05, "loss": 2.954, "step": 13691 }, { "epoch": 1.07, "learning_rate": 8.926370265819808e-05, "loss": 2.8748, "step": 13692 }, { "epoch": 1.07, "learning_rate": 8.926291852897358e-05, "loss": 2.883, "step": 13693 }, { "epoch": 1.07, "learning_rate": 8.926213439974909e-05, "loss": 2.8805, "step": 13694 }, { "epoch": 1.07, "learning_rate": 8.926135027052459e-05, "loss": 2.9986, "step": 13695 }, { "epoch": 1.07, "learning_rate": 8.92605661413001e-05, "loss": 2.9187, "step": 13696 }, { "epoch": 1.07, "learning_rate": 8.92597820120756e-05, "loss": 3.1889, "step": 13697 }, { "epoch": 1.07, "learning_rate": 8.92589978828511e-05, "loss": 3.0612, "step": 13698 }, { "epoch": 1.07, "learning_rate": 8.92582137536266e-05, "loss": 2.7946, "step": 13699 }, { "epoch": 1.07, "learning_rate": 8.92574296244021e-05, "loss": 2.8067, "step": 13700 }, { "epoch": 1.07, "learning_rate": 8.92566454951776e-05, "loss": 2.8781, "step": 13701 }, { "epoch": 1.07, "learning_rate": 8.92558613659531e-05, "loss": 3.0299, "step": 13702 }, { "epoch": 1.07, "learning_rate": 8.925507723672861e-05, "loss": 2.9246, "step": 13703 }, { "epoch": 1.07, "learning_rate": 8.925429310750411e-05, "loss": 2.7418, "step": 13704 }, { "epoch": 1.07, "learning_rate": 8.925350897827962e-05, "loss": 2.7953, "step": 13705 }, { "epoch": 1.07, "learning_rate": 8.925272484905512e-05, "loss": 2.878, "step": 13706 }, { "epoch": 1.07, "learning_rate": 8.925194071983062e-05, "loss": 2.7672, "step": 13707 }, { "epoch": 1.07, "learning_rate": 8.925115659060614e-05, "loss": 2.7853, "step": 13708 }, { "epoch": 1.07, "learning_rate": 8.925037246138164e-05, "loss": 2.7533, "step": 13709 }, { "epoch": 1.08, "learning_rate": 8.924958833215715e-05, "loss": 3.0705, "step": 13710 }, { "epoch": 1.08, "learning_rate": 8.924880420293265e-05, "loss": 3.1375, "step": 13711 }, { "epoch": 1.08, "learning_rate": 8.924802007370815e-05, "loss": 2.7741, "step": 13712 }, { "epoch": 1.08, "learning_rate": 8.924723594448366e-05, "loss": 2.9889, "step": 13713 }, { "epoch": 1.08, "learning_rate": 8.924645181525916e-05, "loss": 3.1616, "step": 13714 }, { "epoch": 1.08, "learning_rate": 8.924566768603467e-05, "loss": 3.0252, "step": 13715 }, { "epoch": 1.08, "learning_rate": 8.924488355681017e-05, "loss": 2.8099, "step": 13716 }, { "epoch": 1.08, "learning_rate": 8.924409942758567e-05, "loss": 2.8651, "step": 13717 }, { "epoch": 1.08, "learning_rate": 8.924331529836118e-05, "loss": 3.0062, "step": 13718 }, { "epoch": 1.08, "learning_rate": 8.924253116913668e-05, "loss": 2.907, "step": 13719 }, { "epoch": 1.08, "learning_rate": 8.924174703991218e-05, "loss": 2.9558, "step": 13720 }, { "epoch": 1.08, "learning_rate": 8.924096291068769e-05, "loss": 2.7531, "step": 13721 }, { "epoch": 1.08, "learning_rate": 8.924017878146319e-05, "loss": 2.6849, "step": 13722 }, { "epoch": 1.08, "learning_rate": 8.92393946522387e-05, "loss": 2.8745, "step": 13723 }, { "epoch": 1.08, "learning_rate": 8.92386105230142e-05, "loss": 3.1192, "step": 13724 }, { "epoch": 1.08, "learning_rate": 8.92378263937897e-05, "loss": 3.0671, "step": 13725 }, { "epoch": 1.08, "learning_rate": 8.92370422645652e-05, "loss": 2.8994, "step": 13726 }, { "epoch": 1.08, "learning_rate": 8.923625813534071e-05, "loss": 2.7496, "step": 13727 }, { "epoch": 1.08, "learning_rate": 8.923547400611621e-05, "loss": 2.8255, "step": 13728 }, { "epoch": 1.08, "learning_rate": 8.923468987689172e-05, "loss": 2.9301, "step": 13729 }, { "epoch": 1.08, "learning_rate": 8.923390574766722e-05, "loss": 3.0662, "step": 13730 }, { "epoch": 1.08, "learning_rate": 8.923312161844272e-05, "loss": 2.9294, "step": 13731 }, { "epoch": 1.08, "learning_rate": 8.923233748921823e-05, "loss": 3.1027, "step": 13732 }, { "epoch": 1.08, "learning_rate": 8.923155335999373e-05, "loss": 2.8806, "step": 13733 }, { "epoch": 1.08, "learning_rate": 8.923076923076924e-05, "loss": 2.7951, "step": 13734 }, { "epoch": 1.08, "learning_rate": 8.922998510154474e-05, "loss": 2.8741, "step": 13735 }, { "epoch": 1.08, "learning_rate": 8.922920097232024e-05, "loss": 2.7379, "step": 13736 }, { "epoch": 1.08, "learning_rate": 8.922841684309575e-05, "loss": 2.8217, "step": 13737 }, { "epoch": 1.08, "learning_rate": 8.922763271387125e-05, "loss": 3.1671, "step": 13738 }, { "epoch": 1.08, "learning_rate": 8.922684858464675e-05, "loss": 3.0486, "step": 13739 }, { "epoch": 1.08, "learning_rate": 8.922606445542226e-05, "loss": 2.9474, "step": 13740 }, { "epoch": 1.08, "learning_rate": 8.922528032619776e-05, "loss": 2.9259, "step": 13741 }, { "epoch": 1.08, "learning_rate": 8.922449619697326e-05, "loss": 2.7699, "step": 13742 }, { "epoch": 1.08, "learning_rate": 8.922371206774877e-05, "loss": 3.1558, "step": 13743 }, { "epoch": 1.08, "learning_rate": 8.922292793852427e-05, "loss": 2.993, "step": 13744 }, { "epoch": 1.08, "learning_rate": 8.922214380929978e-05, "loss": 2.9467, "step": 13745 }, { "epoch": 1.08, "learning_rate": 8.922135968007528e-05, "loss": 2.6106, "step": 13746 }, { "epoch": 1.08, "learning_rate": 8.922057555085078e-05, "loss": 2.8064, "step": 13747 }, { "epoch": 1.08, "learning_rate": 8.921979142162629e-05, "loss": 3.06, "step": 13748 }, { "epoch": 1.08, "learning_rate": 8.921900729240179e-05, "loss": 2.9552, "step": 13749 }, { "epoch": 1.08, "learning_rate": 8.92182231631773e-05, "loss": 2.7289, "step": 13750 }, { "epoch": 1.08, "learning_rate": 8.92174390339528e-05, "loss": 2.885, "step": 13751 }, { "epoch": 1.08, "learning_rate": 8.92166549047283e-05, "loss": 2.8619, "step": 13752 }, { "epoch": 1.08, "learning_rate": 8.92158707755038e-05, "loss": 2.7361, "step": 13753 }, { "epoch": 1.08, "learning_rate": 8.921508664627931e-05, "loss": 3.0473, "step": 13754 }, { "epoch": 1.08, "learning_rate": 8.921430251705481e-05, "loss": 2.7661, "step": 13755 }, { "epoch": 1.08, "learning_rate": 8.921351838783032e-05, "loss": 2.9869, "step": 13756 }, { "epoch": 1.08, "learning_rate": 8.921273425860582e-05, "loss": 3.089, "step": 13757 }, { "epoch": 1.08, "learning_rate": 8.921195012938132e-05, "loss": 2.7857, "step": 13758 }, { "epoch": 1.08, "learning_rate": 8.921116600015683e-05, "loss": 2.8899, "step": 13759 }, { "epoch": 1.08, "learning_rate": 8.921038187093233e-05, "loss": 2.7767, "step": 13760 }, { "epoch": 1.08, "learning_rate": 8.920959774170783e-05, "loss": 2.877, "step": 13761 }, { "epoch": 1.08, "learning_rate": 8.920881361248334e-05, "loss": 3.0332, "step": 13762 }, { "epoch": 1.08, "learning_rate": 8.920802948325884e-05, "loss": 2.6809, "step": 13763 }, { "epoch": 1.08, "learning_rate": 8.920724535403435e-05, "loss": 3.0926, "step": 13764 }, { "epoch": 1.08, "learning_rate": 8.920646122480986e-05, "loss": 2.8839, "step": 13765 }, { "epoch": 1.08, "learning_rate": 8.920567709558537e-05, "loss": 3.042, "step": 13766 }, { "epoch": 1.08, "learning_rate": 8.920489296636087e-05, "loss": 2.7876, "step": 13767 }, { "epoch": 1.08, "learning_rate": 8.920410883713637e-05, "loss": 2.8932, "step": 13768 }, { "epoch": 1.08, "learning_rate": 8.920332470791188e-05, "loss": 2.9713, "step": 13769 }, { "epoch": 1.08, "learning_rate": 8.920254057868738e-05, "loss": 2.7443, "step": 13770 }, { "epoch": 1.08, "learning_rate": 8.920175644946289e-05, "loss": 2.8063, "step": 13771 }, { "epoch": 1.08, "learning_rate": 8.920097232023838e-05, "loss": 2.9653, "step": 13772 }, { "epoch": 1.08, "learning_rate": 8.920018819101388e-05, "loss": 3.2603, "step": 13773 }, { "epoch": 1.08, "learning_rate": 8.919940406178938e-05, "loss": 3.2032, "step": 13774 }, { "epoch": 1.08, "learning_rate": 8.919861993256489e-05, "loss": 2.9175, "step": 13775 }, { "epoch": 1.08, "learning_rate": 8.919783580334039e-05, "loss": 2.8704, "step": 13776 }, { "epoch": 1.08, "learning_rate": 8.91970516741159e-05, "loss": 2.9724, "step": 13777 }, { "epoch": 1.08, "learning_rate": 8.91962675448914e-05, "loss": 3.0481, "step": 13778 }, { "epoch": 1.08, "learning_rate": 8.91954834156669e-05, "loss": 3.0244, "step": 13779 }, { "epoch": 1.08, "learning_rate": 8.91946992864424e-05, "loss": 2.8451, "step": 13780 }, { "epoch": 1.08, "learning_rate": 8.919391515721791e-05, "loss": 3.0516, "step": 13781 }, { "epoch": 1.08, "learning_rate": 8.919313102799341e-05, "loss": 2.873, "step": 13782 }, { "epoch": 1.08, "learning_rate": 8.919234689876892e-05, "loss": 3.0027, "step": 13783 }, { "epoch": 1.08, "learning_rate": 8.919156276954442e-05, "loss": 2.9092, "step": 13784 }, { "epoch": 1.08, "learning_rate": 8.919077864031992e-05, "loss": 3.0021, "step": 13785 }, { "epoch": 1.08, "learning_rate": 8.918999451109543e-05, "loss": 3.0859, "step": 13786 }, { "epoch": 1.08, "learning_rate": 8.918921038187093e-05, "loss": 2.8595, "step": 13787 }, { "epoch": 1.08, "learning_rate": 8.918842625264643e-05, "loss": 2.807, "step": 13788 }, { "epoch": 1.08, "learning_rate": 8.918764212342194e-05, "loss": 2.8949, "step": 13789 }, { "epoch": 1.08, "learning_rate": 8.918685799419744e-05, "loss": 2.8606, "step": 13790 }, { "epoch": 1.08, "learning_rate": 8.918607386497296e-05, "loss": 2.9907, "step": 13791 }, { "epoch": 1.08, "learning_rate": 8.918528973574846e-05, "loss": 2.7482, "step": 13792 }, { "epoch": 1.08, "learning_rate": 8.918450560652397e-05, "loss": 2.8989, "step": 13793 }, { "epoch": 1.08, "learning_rate": 8.918372147729947e-05, "loss": 2.8469, "step": 13794 }, { "epoch": 1.08, "learning_rate": 8.918293734807497e-05, "loss": 2.847, "step": 13795 }, { "epoch": 1.08, "learning_rate": 8.918215321885048e-05, "loss": 3.0351, "step": 13796 }, { "epoch": 1.08, "learning_rate": 8.918136908962598e-05, "loss": 3.0531, "step": 13797 }, { "epoch": 1.08, "learning_rate": 8.918058496040148e-05, "loss": 2.9856, "step": 13798 }, { "epoch": 1.08, "learning_rate": 8.917980083117699e-05, "loss": 3.0829, "step": 13799 }, { "epoch": 1.08, "learning_rate": 8.917901670195249e-05, "loss": 3.069, "step": 13800 }, { "epoch": 1.08, "learning_rate": 8.9178232572728e-05, "loss": 2.9176, "step": 13801 }, { "epoch": 1.08, "learning_rate": 8.91774484435035e-05, "loss": 2.9848, "step": 13802 }, { "epoch": 1.08, "learning_rate": 8.9176664314279e-05, "loss": 3.1384, "step": 13803 }, { "epoch": 1.08, "learning_rate": 8.91758801850545e-05, "loss": 2.9325, "step": 13804 }, { "epoch": 1.08, "learning_rate": 8.917509605583001e-05, "loss": 2.9419, "step": 13805 }, { "epoch": 1.08, "learning_rate": 8.917431192660551e-05, "loss": 2.787, "step": 13806 }, { "epoch": 1.08, "learning_rate": 8.917352779738102e-05, "loss": 2.9025, "step": 13807 }, { "epoch": 1.08, "learning_rate": 8.917274366815651e-05, "loss": 2.9339, "step": 13808 }, { "epoch": 1.08, "learning_rate": 8.917195953893201e-05, "loss": 2.8577, "step": 13809 }, { "epoch": 1.08, "learning_rate": 8.917117540970751e-05, "loss": 2.6598, "step": 13810 }, { "epoch": 1.08, "learning_rate": 8.917039128048302e-05, "loss": 3.1059, "step": 13811 }, { "epoch": 1.08, "learning_rate": 8.916960715125852e-05, "loss": 2.9971, "step": 13812 }, { "epoch": 1.08, "learning_rate": 8.916882302203403e-05, "loss": 2.7008, "step": 13813 }, { "epoch": 1.08, "learning_rate": 8.916803889280953e-05, "loss": 2.8755, "step": 13814 }, { "epoch": 1.08, "learning_rate": 8.916725476358503e-05, "loss": 2.8544, "step": 13815 }, { "epoch": 1.08, "learning_rate": 8.916647063436055e-05, "loss": 2.815, "step": 13816 }, { "epoch": 1.08, "learning_rate": 8.916568650513605e-05, "loss": 2.7829, "step": 13817 }, { "epoch": 1.08, "learning_rate": 8.916490237591156e-05, "loss": 2.9063, "step": 13818 }, { "epoch": 1.08, "learning_rate": 8.916411824668706e-05, "loss": 2.961, "step": 13819 }, { "epoch": 1.08, "learning_rate": 8.916333411746257e-05, "loss": 2.8316, "step": 13820 }, { "epoch": 1.08, "learning_rate": 8.916254998823807e-05, "loss": 2.9063, "step": 13821 }, { "epoch": 1.08, "learning_rate": 8.916176585901357e-05, "loss": 2.82, "step": 13822 }, { "epoch": 1.08, "learning_rate": 8.916098172978908e-05, "loss": 2.7034, "step": 13823 }, { "epoch": 1.08, "learning_rate": 8.916019760056458e-05, "loss": 3.1533, "step": 13824 }, { "epoch": 1.08, "learning_rate": 8.915941347134008e-05, "loss": 2.9366, "step": 13825 }, { "epoch": 1.08, "learning_rate": 8.915862934211559e-05, "loss": 2.6978, "step": 13826 }, { "epoch": 1.08, "learning_rate": 8.915784521289109e-05, "loss": 3.0386, "step": 13827 }, { "epoch": 1.08, "learning_rate": 8.91570610836666e-05, "loss": 3.0087, "step": 13828 }, { "epoch": 1.08, "learning_rate": 8.91562769544421e-05, "loss": 2.8453, "step": 13829 }, { "epoch": 1.08, "learning_rate": 8.91554928252176e-05, "loss": 2.851, "step": 13830 }, { "epoch": 1.08, "learning_rate": 8.91547086959931e-05, "loss": 2.9152, "step": 13831 }, { "epoch": 1.08, "learning_rate": 8.915392456676861e-05, "loss": 2.9294, "step": 13832 }, { "epoch": 1.08, "learning_rate": 8.915314043754411e-05, "loss": 2.9292, "step": 13833 }, { "epoch": 1.08, "learning_rate": 8.915235630831962e-05, "loss": 2.8783, "step": 13834 }, { "epoch": 1.08, "learning_rate": 8.915157217909512e-05, "loss": 3.0584, "step": 13835 }, { "epoch": 1.08, "learning_rate": 8.915078804987062e-05, "loss": 2.9771, "step": 13836 }, { "epoch": 1.08, "learning_rate": 8.915000392064613e-05, "loss": 2.9022, "step": 13837 }, { "epoch": 1.09, "learning_rate": 8.914921979142163e-05, "loss": 2.9957, "step": 13838 }, { "epoch": 1.09, "learning_rate": 8.914843566219714e-05, "loss": 2.9398, "step": 13839 }, { "epoch": 1.09, "learning_rate": 8.914765153297264e-05, "loss": 2.8995, "step": 13840 }, { "epoch": 1.09, "learning_rate": 8.914686740374814e-05, "loss": 2.8883, "step": 13841 }, { "epoch": 1.09, "learning_rate": 8.914608327452365e-05, "loss": 3.1077, "step": 13842 }, { "epoch": 1.09, "learning_rate": 8.914529914529915e-05, "loss": 2.778, "step": 13843 }, { "epoch": 1.09, "learning_rate": 8.914451501607465e-05, "loss": 2.8053, "step": 13844 }, { "epoch": 1.09, "learning_rate": 8.914373088685016e-05, "loss": 2.8528, "step": 13845 }, { "epoch": 1.09, "learning_rate": 8.914294675762566e-05, "loss": 2.9225, "step": 13846 }, { "epoch": 1.09, "learning_rate": 8.914216262840116e-05, "loss": 2.8612, "step": 13847 }, { "epoch": 1.09, "learning_rate": 8.914137849917667e-05, "loss": 2.9866, "step": 13848 }, { "epoch": 1.09, "learning_rate": 8.914059436995217e-05, "loss": 3.0011, "step": 13849 }, { "epoch": 1.09, "learning_rate": 8.913981024072768e-05, "loss": 3.0663, "step": 13850 }, { "epoch": 1.09, "learning_rate": 8.913902611150318e-05, "loss": 3.0249, "step": 13851 }, { "epoch": 1.09, "learning_rate": 8.913824198227868e-05, "loss": 2.929, "step": 13852 }, { "epoch": 1.09, "learning_rate": 8.913745785305419e-05, "loss": 3.2912, "step": 13853 }, { "epoch": 1.09, "learning_rate": 8.913667372382969e-05, "loss": 2.6475, "step": 13854 }, { "epoch": 1.09, "learning_rate": 8.91358895946052e-05, "loss": 2.906, "step": 13855 }, { "epoch": 1.09, "learning_rate": 8.91351054653807e-05, "loss": 2.7472, "step": 13856 }, { "epoch": 1.09, "learning_rate": 8.91343213361562e-05, "loss": 2.7402, "step": 13857 }, { "epoch": 1.09, "learning_rate": 8.91335372069317e-05, "loss": 3.0111, "step": 13858 }, { "epoch": 1.09, "learning_rate": 8.913275307770721e-05, "loss": 2.8938, "step": 13859 }, { "epoch": 1.09, "learning_rate": 8.913196894848271e-05, "loss": 3.1412, "step": 13860 }, { "epoch": 1.09, "learning_rate": 8.913118481925822e-05, "loss": 2.8513, "step": 13861 }, { "epoch": 1.09, "learning_rate": 8.913040069003372e-05, "loss": 2.724, "step": 13862 }, { "epoch": 1.09, "learning_rate": 8.912961656080922e-05, "loss": 3.0144, "step": 13863 }, { "epoch": 1.09, "learning_rate": 8.912883243158473e-05, "loss": 2.7353, "step": 13864 }, { "epoch": 1.09, "learning_rate": 8.912804830236023e-05, "loss": 3.1204, "step": 13865 }, { "epoch": 1.09, "learning_rate": 8.912726417313573e-05, "loss": 2.9758, "step": 13866 }, { "epoch": 1.09, "learning_rate": 8.912648004391124e-05, "loss": 3.3264, "step": 13867 }, { "epoch": 1.09, "learning_rate": 8.912569591468674e-05, "loss": 2.8269, "step": 13868 }, { "epoch": 1.09, "learning_rate": 8.912491178546225e-05, "loss": 2.8767, "step": 13869 }, { "epoch": 1.09, "learning_rate": 8.912412765623775e-05, "loss": 3.2709, "step": 13870 }, { "epoch": 1.09, "learning_rate": 8.912334352701325e-05, "loss": 3.1786, "step": 13871 }, { "epoch": 1.09, "learning_rate": 8.912255939778876e-05, "loss": 2.8999, "step": 13872 }, { "epoch": 1.09, "learning_rate": 8.912177526856427e-05, "loss": 2.8121, "step": 13873 }, { "epoch": 1.09, "learning_rate": 8.912099113933978e-05, "loss": 3.0324, "step": 13874 }, { "epoch": 1.09, "learning_rate": 8.912020701011528e-05, "loss": 2.7, "step": 13875 }, { "epoch": 1.09, "learning_rate": 8.911942288089078e-05, "loss": 2.8301, "step": 13876 }, { "epoch": 1.09, "learning_rate": 8.911863875166629e-05, "loss": 2.806, "step": 13877 }, { "epoch": 1.09, "learning_rate": 8.911785462244179e-05, "loss": 3.1285, "step": 13878 }, { "epoch": 1.09, "learning_rate": 8.91170704932173e-05, "loss": 2.8344, "step": 13879 }, { "epoch": 1.09, "learning_rate": 8.911628636399279e-05, "loss": 2.896, "step": 13880 }, { "epoch": 1.09, "learning_rate": 8.911550223476829e-05, "loss": 2.7448, "step": 13881 }, { "epoch": 1.09, "learning_rate": 8.91147181055438e-05, "loss": 3.0235, "step": 13882 }, { "epoch": 1.09, "learning_rate": 8.91139339763193e-05, "loss": 2.8879, "step": 13883 }, { "epoch": 1.09, "learning_rate": 8.91131498470948e-05, "loss": 2.9375, "step": 13884 }, { "epoch": 1.09, "learning_rate": 8.91123657178703e-05, "loss": 2.7307, "step": 13885 }, { "epoch": 1.09, "learning_rate": 8.911158158864581e-05, "loss": 2.8849, "step": 13886 }, { "epoch": 1.09, "learning_rate": 8.911079745942131e-05, "loss": 2.6391, "step": 13887 }, { "epoch": 1.09, "learning_rate": 8.911001333019682e-05, "loss": 2.8623, "step": 13888 }, { "epoch": 1.09, "learning_rate": 8.910922920097232e-05, "loss": 2.7027, "step": 13889 }, { "epoch": 1.09, "learning_rate": 8.910844507174782e-05, "loss": 3.3027, "step": 13890 }, { "epoch": 1.09, "learning_rate": 8.910766094252333e-05, "loss": 3.0316, "step": 13891 }, { "epoch": 1.09, "learning_rate": 8.910687681329883e-05, "loss": 2.5904, "step": 13892 }, { "epoch": 1.09, "learning_rate": 8.910609268407433e-05, "loss": 3.1587, "step": 13893 }, { "epoch": 1.09, "learning_rate": 8.910530855484984e-05, "loss": 2.6271, "step": 13894 }, { "epoch": 1.09, "learning_rate": 8.910452442562534e-05, "loss": 3.1195, "step": 13895 }, { "epoch": 1.09, "learning_rate": 8.910374029640084e-05, "loss": 2.8299, "step": 13896 }, { "epoch": 1.09, "learning_rate": 8.910295616717635e-05, "loss": 2.8916, "step": 13897 }, { "epoch": 1.09, "learning_rate": 8.910217203795185e-05, "loss": 2.8089, "step": 13898 }, { "epoch": 1.09, "learning_rate": 8.910138790872737e-05, "loss": 2.9329, "step": 13899 }, { "epoch": 1.09, "learning_rate": 8.910060377950287e-05, "loss": 2.807, "step": 13900 }, { "epoch": 1.09, "learning_rate": 8.909981965027838e-05, "loss": 3.0241, "step": 13901 }, { "epoch": 1.09, "learning_rate": 8.909903552105388e-05, "loss": 3.0265, "step": 13902 }, { "epoch": 1.09, "learning_rate": 8.909825139182938e-05, "loss": 3.2847, "step": 13903 }, { "epoch": 1.09, "learning_rate": 8.909746726260489e-05, "loss": 2.8852, "step": 13904 }, { "epoch": 1.09, "learning_rate": 8.909668313338039e-05, "loss": 3.1811, "step": 13905 }, { "epoch": 1.09, "learning_rate": 8.90958990041559e-05, "loss": 2.8505, "step": 13906 }, { "epoch": 1.09, "learning_rate": 8.90951148749314e-05, "loss": 2.9388, "step": 13907 }, { "epoch": 1.09, "learning_rate": 8.90943307457069e-05, "loss": 2.8637, "step": 13908 }, { "epoch": 1.09, "learning_rate": 8.90935466164824e-05, "loss": 3.0215, "step": 13909 }, { "epoch": 1.09, "learning_rate": 8.909276248725791e-05, "loss": 3.1761, "step": 13910 }, { "epoch": 1.09, "learning_rate": 8.909197835803341e-05, "loss": 2.9361, "step": 13911 }, { "epoch": 1.09, "learning_rate": 8.909119422880892e-05, "loss": 2.947, "step": 13912 }, { "epoch": 1.09, "learning_rate": 8.909041009958442e-05, "loss": 2.6938, "step": 13913 }, { "epoch": 1.09, "learning_rate": 8.908962597035992e-05, "loss": 2.9755, "step": 13914 }, { "epoch": 1.09, "learning_rate": 8.908884184113543e-05, "loss": 2.8466, "step": 13915 }, { "epoch": 1.09, "learning_rate": 8.908805771191092e-05, "loss": 2.9275, "step": 13916 }, { "epoch": 1.09, "learning_rate": 8.908727358268642e-05, "loss": 2.7861, "step": 13917 }, { "epoch": 1.09, "learning_rate": 8.908648945346193e-05, "loss": 2.9405, "step": 13918 }, { "epoch": 1.09, "learning_rate": 8.908570532423743e-05, "loss": 2.6705, "step": 13919 }, { "epoch": 1.09, "learning_rate": 8.908492119501293e-05, "loss": 2.7724, "step": 13920 }, { "epoch": 1.09, "learning_rate": 8.908413706578844e-05, "loss": 2.8295, "step": 13921 }, { "epoch": 1.09, "learning_rate": 8.908335293656394e-05, "loss": 3.0293, "step": 13922 }, { "epoch": 1.09, "learning_rate": 8.908256880733944e-05, "loss": 2.8641, "step": 13923 }, { "epoch": 1.09, "learning_rate": 8.908178467811495e-05, "loss": 2.8691, "step": 13924 }, { "epoch": 1.09, "learning_rate": 8.908100054889047e-05, "loss": 3.0561, "step": 13925 }, { "epoch": 1.09, "learning_rate": 8.908021641966597e-05, "loss": 3.036, "step": 13926 }, { "epoch": 1.09, "learning_rate": 8.907943229044147e-05, "loss": 2.9987, "step": 13927 }, { "epoch": 1.09, "learning_rate": 8.907864816121698e-05, "loss": 2.8116, "step": 13928 }, { "epoch": 1.09, "learning_rate": 8.907786403199248e-05, "loss": 2.8613, "step": 13929 }, { "epoch": 1.09, "learning_rate": 8.907707990276798e-05, "loss": 2.8068, "step": 13930 }, { "epoch": 1.09, "learning_rate": 8.907629577354349e-05, "loss": 3.0191, "step": 13931 }, { "epoch": 1.09, "learning_rate": 8.907551164431899e-05, "loss": 3.3058, "step": 13932 }, { "epoch": 1.09, "learning_rate": 8.90747275150945e-05, "loss": 2.9601, "step": 13933 }, { "epoch": 1.09, "learning_rate": 8.907394338587e-05, "loss": 2.9623, "step": 13934 }, { "epoch": 1.09, "learning_rate": 8.90731592566455e-05, "loss": 2.8428, "step": 13935 }, { "epoch": 1.09, "learning_rate": 8.9072375127421e-05, "loss": 2.9005, "step": 13936 }, { "epoch": 1.09, "learning_rate": 8.907159099819651e-05, "loss": 2.8787, "step": 13937 }, { "epoch": 1.09, "learning_rate": 8.907080686897201e-05, "loss": 2.8565, "step": 13938 }, { "epoch": 1.09, "learning_rate": 8.907002273974752e-05, "loss": 2.7532, "step": 13939 }, { "epoch": 1.09, "learning_rate": 8.906923861052302e-05, "loss": 2.8567, "step": 13940 }, { "epoch": 1.09, "learning_rate": 8.906845448129852e-05, "loss": 2.8419, "step": 13941 }, { "epoch": 1.09, "learning_rate": 8.906767035207403e-05, "loss": 2.8515, "step": 13942 }, { "epoch": 1.09, "learning_rate": 8.906688622284953e-05, "loss": 2.7778, "step": 13943 }, { "epoch": 1.09, "learning_rate": 8.906610209362503e-05, "loss": 3.1641, "step": 13944 }, { "epoch": 1.09, "learning_rate": 8.906531796440054e-05, "loss": 2.8533, "step": 13945 }, { "epoch": 1.09, "learning_rate": 8.906453383517604e-05, "loss": 2.9563, "step": 13946 }, { "epoch": 1.09, "learning_rate": 8.906374970595155e-05, "loss": 2.6815, "step": 13947 }, { "epoch": 1.09, "learning_rate": 8.906296557672705e-05, "loss": 2.7932, "step": 13948 }, { "epoch": 1.09, "learning_rate": 8.906218144750255e-05, "loss": 3.0107, "step": 13949 }, { "epoch": 1.09, "learning_rate": 8.906139731827806e-05, "loss": 2.89, "step": 13950 }, { "epoch": 1.09, "learning_rate": 8.906061318905356e-05, "loss": 2.9587, "step": 13951 }, { "epoch": 1.09, "learning_rate": 8.905982905982906e-05, "loss": 3.155, "step": 13952 }, { "epoch": 1.09, "learning_rate": 8.905904493060457e-05, "loss": 2.8903, "step": 13953 }, { "epoch": 1.09, "learning_rate": 8.905826080138007e-05, "loss": 2.7009, "step": 13954 }, { "epoch": 1.09, "learning_rate": 8.905747667215558e-05, "loss": 2.9596, "step": 13955 }, { "epoch": 1.09, "learning_rate": 8.905669254293108e-05, "loss": 2.8759, "step": 13956 }, { "epoch": 1.09, "learning_rate": 8.905590841370658e-05, "loss": 2.7989, "step": 13957 }, { "epoch": 1.09, "learning_rate": 8.905512428448209e-05, "loss": 2.6735, "step": 13958 }, { "epoch": 1.09, "learning_rate": 8.905434015525759e-05, "loss": 2.8959, "step": 13959 }, { "epoch": 1.09, "learning_rate": 8.90535560260331e-05, "loss": 2.9091, "step": 13960 }, { "epoch": 1.09, "learning_rate": 8.90527718968086e-05, "loss": 2.7745, "step": 13961 }, { "epoch": 1.09, "learning_rate": 8.90519877675841e-05, "loss": 3.0834, "step": 13962 }, { "epoch": 1.09, "learning_rate": 8.90512036383596e-05, "loss": 2.8298, "step": 13963 }, { "epoch": 1.09, "learning_rate": 8.905041950913511e-05, "loss": 2.8956, "step": 13964 }, { "epoch": 1.1, "learning_rate": 8.904963537991061e-05, "loss": 2.8146, "step": 13965 }, { "epoch": 1.1, "learning_rate": 8.904885125068612e-05, "loss": 2.9327, "step": 13966 }, { "epoch": 1.1, "learning_rate": 8.904806712146162e-05, "loss": 2.9546, "step": 13967 }, { "epoch": 1.1, "learning_rate": 8.904728299223712e-05, "loss": 2.913, "step": 13968 }, { "epoch": 1.1, "learning_rate": 8.904649886301263e-05, "loss": 3.047, "step": 13969 }, { "epoch": 1.1, "learning_rate": 8.904571473378813e-05, "loss": 3.034, "step": 13970 }, { "epoch": 1.1, "learning_rate": 8.904493060456363e-05, "loss": 2.8411, "step": 13971 }, { "epoch": 1.1, "learning_rate": 8.904414647533914e-05, "loss": 2.7108, "step": 13972 }, { "epoch": 1.1, "learning_rate": 8.904336234611464e-05, "loss": 2.9402, "step": 13973 }, { "epoch": 1.1, "learning_rate": 8.904257821689015e-05, "loss": 3.0245, "step": 13974 }, { "epoch": 1.1, "learning_rate": 8.904179408766565e-05, "loss": 3.1407, "step": 13975 }, { "epoch": 1.1, "learning_rate": 8.904100995844115e-05, "loss": 3.1429, "step": 13976 }, { "epoch": 1.1, "learning_rate": 8.904022582921666e-05, "loss": 3.0678, "step": 13977 }, { "epoch": 1.1, "learning_rate": 8.903944169999216e-05, "loss": 2.7394, "step": 13978 }, { "epoch": 1.1, "learning_rate": 8.903865757076766e-05, "loss": 2.8424, "step": 13979 }, { "epoch": 1.1, "learning_rate": 8.903787344154317e-05, "loss": 2.8149, "step": 13980 }, { "epoch": 1.1, "learning_rate": 8.903708931231868e-05, "loss": 2.8738, "step": 13981 }, { "epoch": 1.1, "learning_rate": 8.903630518309419e-05, "loss": 3.0446, "step": 13982 }, { "epoch": 1.1, "learning_rate": 8.903552105386969e-05, "loss": 2.9308, "step": 13983 }, { "epoch": 1.1, "learning_rate": 8.90347369246452e-05, "loss": 2.6427, "step": 13984 }, { "epoch": 1.1, "learning_rate": 8.90339527954207e-05, "loss": 2.6552, "step": 13985 }, { "epoch": 1.1, "learning_rate": 8.90331686661962e-05, "loss": 2.8591, "step": 13986 }, { "epoch": 1.1, "learning_rate": 8.90323845369717e-05, "loss": 2.7677, "step": 13987 }, { "epoch": 1.1, "learning_rate": 8.90316004077472e-05, "loss": 2.8735, "step": 13988 }, { "epoch": 1.1, "learning_rate": 8.90308162785227e-05, "loss": 2.8793, "step": 13989 }, { "epoch": 1.1, "learning_rate": 8.90300321492982e-05, "loss": 2.84, "step": 13990 }, { "epoch": 1.1, "learning_rate": 8.902924802007371e-05, "loss": 2.8183, "step": 13991 }, { "epoch": 1.1, "learning_rate": 8.902846389084921e-05, "loss": 2.7832, "step": 13992 }, { "epoch": 1.1, "learning_rate": 8.902767976162472e-05, "loss": 2.9228, "step": 13993 }, { "epoch": 1.1, "learning_rate": 8.902689563240022e-05, "loss": 2.9442, "step": 13994 }, { "epoch": 1.1, "learning_rate": 8.902611150317572e-05, "loss": 2.9055, "step": 13995 }, { "epoch": 1.1, "learning_rate": 8.902532737395123e-05, "loss": 3.2099, "step": 13996 }, { "epoch": 1.1, "learning_rate": 8.902454324472673e-05, "loss": 2.7882, "step": 13997 }, { "epoch": 1.1, "learning_rate": 8.902375911550223e-05, "loss": 2.9835, "step": 13998 }, { "epoch": 1.1, "learning_rate": 8.902297498627774e-05, "loss": 2.7434, "step": 13999 }, { "epoch": 1.1, "learning_rate": 8.902219085705324e-05, "loss": 3.3254, "step": 14000 }, { "epoch": 1.1, "eval_loss": 2.9225692749023438, "eval_runtime": 218.0689, "eval_samples_per_second": 51.965, "eval_steps_per_second": 3.251, "step": 14000 }, { "epoch": 1.1, "learning_rate": 8.902140672782874e-05, "loss": 2.8176, "step": 14001 }, { "epoch": 1.1, "learning_rate": 8.902062259860425e-05, "loss": 3.2265, "step": 14002 }, { "epoch": 1.1, "learning_rate": 8.901983846937975e-05, "loss": 3.0125, "step": 14003 }, { "epoch": 1.1, "learning_rate": 8.901905434015526e-05, "loss": 2.9068, "step": 14004 }, { "epoch": 1.1, "learning_rate": 8.901827021093076e-05, "loss": 2.8653, "step": 14005 }, { "epoch": 1.1, "learning_rate": 8.901748608170626e-05, "loss": 3.0357, "step": 14006 }, { "epoch": 1.1, "learning_rate": 8.901670195248178e-05, "loss": 2.8335, "step": 14007 }, { "epoch": 1.1, "learning_rate": 8.901591782325728e-05, "loss": 3.0191, "step": 14008 }, { "epoch": 1.1, "learning_rate": 8.901513369403279e-05, "loss": 2.7592, "step": 14009 }, { "epoch": 1.1, "learning_rate": 8.901434956480829e-05, "loss": 2.9926, "step": 14010 }, { "epoch": 1.1, "learning_rate": 8.90135654355838e-05, "loss": 2.9278, "step": 14011 }, { "epoch": 1.1, "learning_rate": 8.90127813063593e-05, "loss": 2.8877, "step": 14012 }, { "epoch": 1.1, "learning_rate": 8.90119971771348e-05, "loss": 2.9688, "step": 14013 }, { "epoch": 1.1, "learning_rate": 8.90112130479103e-05, "loss": 2.9904, "step": 14014 }, { "epoch": 1.1, "learning_rate": 8.901042891868581e-05, "loss": 2.7599, "step": 14015 }, { "epoch": 1.1, "learning_rate": 8.900964478946131e-05, "loss": 2.6578, "step": 14016 }, { "epoch": 1.1, "learning_rate": 8.900886066023682e-05, "loss": 2.9984, "step": 14017 }, { "epoch": 1.1, "learning_rate": 8.900807653101232e-05, "loss": 3.113, "step": 14018 }, { "epoch": 1.1, "learning_rate": 8.900729240178782e-05, "loss": 2.7871, "step": 14019 }, { "epoch": 1.1, "learning_rate": 8.900650827256333e-05, "loss": 2.8506, "step": 14020 }, { "epoch": 1.1, "learning_rate": 8.900572414333883e-05, "loss": 3.0588, "step": 14021 }, { "epoch": 1.1, "learning_rate": 8.900494001411434e-05, "loss": 3.2976, "step": 14022 }, { "epoch": 1.1, "learning_rate": 8.900415588488984e-05, "loss": 2.7201, "step": 14023 }, { "epoch": 1.1, "learning_rate": 8.900337175566533e-05, "loss": 3.1587, "step": 14024 }, { "epoch": 1.1, "learning_rate": 8.900258762644083e-05, "loss": 3.0152, "step": 14025 }, { "epoch": 1.1, "learning_rate": 8.900180349721634e-05, "loss": 2.7908, "step": 14026 }, { "epoch": 1.1, "learning_rate": 8.900101936799184e-05, "loss": 2.7942, "step": 14027 }, { "epoch": 1.1, "learning_rate": 8.900023523876734e-05, "loss": 3.0319, "step": 14028 }, { "epoch": 1.1, "learning_rate": 8.899945110954285e-05, "loss": 2.7468, "step": 14029 }, { "epoch": 1.1, "learning_rate": 8.899866698031835e-05, "loss": 2.7296, "step": 14030 }, { "epoch": 1.1, "learning_rate": 8.899788285109385e-05, "loss": 2.8272, "step": 14031 }, { "epoch": 1.1, "learning_rate": 8.899709872186936e-05, "loss": 2.8279, "step": 14032 }, { "epoch": 1.1, "learning_rate": 8.899631459264488e-05, "loss": 3.0557, "step": 14033 }, { "epoch": 1.1, "learning_rate": 8.899553046342038e-05, "loss": 2.7288, "step": 14034 }, { "epoch": 1.1, "learning_rate": 8.899474633419588e-05, "loss": 2.9935, "step": 14035 }, { "epoch": 1.1, "learning_rate": 8.899396220497139e-05, "loss": 2.8311, "step": 14036 }, { "epoch": 1.1, "learning_rate": 8.899317807574689e-05, "loss": 2.8557, "step": 14037 }, { "epoch": 1.1, "learning_rate": 8.89923939465224e-05, "loss": 3.002, "step": 14038 }, { "epoch": 1.1, "learning_rate": 8.89916098172979e-05, "loss": 3.005, "step": 14039 }, { "epoch": 1.1, "learning_rate": 8.89908256880734e-05, "loss": 3.0266, "step": 14040 }, { "epoch": 1.1, "learning_rate": 8.89900415588489e-05, "loss": 2.9854, "step": 14041 }, { "epoch": 1.1, "learning_rate": 8.898925742962441e-05, "loss": 2.896, "step": 14042 }, { "epoch": 1.1, "learning_rate": 8.898847330039991e-05, "loss": 2.8848, "step": 14043 }, { "epoch": 1.1, "learning_rate": 8.898768917117542e-05, "loss": 2.8716, "step": 14044 }, { "epoch": 1.1, "learning_rate": 8.898690504195092e-05, "loss": 2.7548, "step": 14045 }, { "epoch": 1.1, "learning_rate": 8.898612091272642e-05, "loss": 3.0017, "step": 14046 }, { "epoch": 1.1, "learning_rate": 8.898533678350193e-05, "loss": 2.9725, "step": 14047 }, { "epoch": 1.1, "learning_rate": 8.898455265427743e-05, "loss": 2.791, "step": 14048 }, { "epoch": 1.1, "learning_rate": 8.898376852505293e-05, "loss": 3.0097, "step": 14049 }, { "epoch": 1.1, "learning_rate": 8.898298439582844e-05, "loss": 2.9801, "step": 14050 }, { "epoch": 1.1, "learning_rate": 8.898220026660394e-05, "loss": 2.8028, "step": 14051 }, { "epoch": 1.1, "learning_rate": 8.898141613737945e-05, "loss": 2.9151, "step": 14052 }, { "epoch": 1.1, "learning_rate": 8.898063200815495e-05, "loss": 2.87, "step": 14053 }, { "epoch": 1.1, "learning_rate": 8.897984787893045e-05, "loss": 3.0065, "step": 14054 }, { "epoch": 1.1, "learning_rate": 8.897906374970596e-05, "loss": 2.9863, "step": 14055 }, { "epoch": 1.1, "learning_rate": 8.897827962048146e-05, "loss": 2.8604, "step": 14056 }, { "epoch": 1.1, "learning_rate": 8.897749549125696e-05, "loss": 2.7772, "step": 14057 }, { "epoch": 1.1, "learning_rate": 8.897671136203247e-05, "loss": 2.8082, "step": 14058 }, { "epoch": 1.1, "learning_rate": 8.897592723280797e-05, "loss": 2.8234, "step": 14059 }, { "epoch": 1.1, "learning_rate": 8.897514310358348e-05, "loss": 2.9802, "step": 14060 }, { "epoch": 1.1, "learning_rate": 8.897435897435898e-05, "loss": 2.8832, "step": 14061 }, { "epoch": 1.1, "learning_rate": 8.897357484513448e-05, "loss": 2.8001, "step": 14062 }, { "epoch": 1.1, "learning_rate": 8.897279071590999e-05, "loss": 2.8239, "step": 14063 }, { "epoch": 1.1, "learning_rate": 8.897200658668549e-05, "loss": 2.8437, "step": 14064 }, { "epoch": 1.1, "learning_rate": 8.8971222457461e-05, "loss": 2.8165, "step": 14065 }, { "epoch": 1.1, "learning_rate": 8.89704383282365e-05, "loss": 2.7979, "step": 14066 }, { "epoch": 1.1, "learning_rate": 8.8969654199012e-05, "loss": 2.6431, "step": 14067 }, { "epoch": 1.1, "learning_rate": 8.89688700697875e-05, "loss": 2.8752, "step": 14068 }, { "epoch": 1.1, "learning_rate": 8.896808594056301e-05, "loss": 2.9622, "step": 14069 }, { "epoch": 1.1, "learning_rate": 8.896730181133851e-05, "loss": 3.1478, "step": 14070 }, { "epoch": 1.1, "learning_rate": 8.896651768211402e-05, "loss": 2.8438, "step": 14071 }, { "epoch": 1.1, "learning_rate": 8.896573355288952e-05, "loss": 2.9703, "step": 14072 }, { "epoch": 1.1, "learning_rate": 8.896494942366502e-05, "loss": 2.8249, "step": 14073 }, { "epoch": 1.1, "learning_rate": 8.896416529444053e-05, "loss": 3.0457, "step": 14074 }, { "epoch": 1.1, "learning_rate": 8.896338116521603e-05, "loss": 2.9113, "step": 14075 }, { "epoch": 1.1, "learning_rate": 8.896259703599153e-05, "loss": 3.0306, "step": 14076 }, { "epoch": 1.1, "learning_rate": 8.896181290676704e-05, "loss": 2.9944, "step": 14077 }, { "epoch": 1.1, "learning_rate": 8.896102877754254e-05, "loss": 2.8023, "step": 14078 }, { "epoch": 1.1, "learning_rate": 8.896024464831804e-05, "loss": 2.8643, "step": 14079 }, { "epoch": 1.1, "learning_rate": 8.895946051909355e-05, "loss": 2.808, "step": 14080 }, { "epoch": 1.1, "learning_rate": 8.895867638986905e-05, "loss": 2.7602, "step": 14081 }, { "epoch": 1.1, "learning_rate": 8.895789226064456e-05, "loss": 2.9912, "step": 14082 }, { "epoch": 1.1, "learning_rate": 8.895710813142006e-05, "loss": 2.9085, "step": 14083 }, { "epoch": 1.1, "learning_rate": 8.895632400219556e-05, "loss": 2.9798, "step": 14084 }, { "epoch": 1.1, "learning_rate": 8.895553987297107e-05, "loss": 2.8735, "step": 14085 }, { "epoch": 1.1, "learning_rate": 8.895475574374657e-05, "loss": 2.9879, "step": 14086 }, { "epoch": 1.1, "learning_rate": 8.895397161452207e-05, "loss": 2.8727, "step": 14087 }, { "epoch": 1.1, "learning_rate": 8.895318748529758e-05, "loss": 2.8511, "step": 14088 }, { "epoch": 1.1, "learning_rate": 8.89524033560731e-05, "loss": 3.0456, "step": 14089 }, { "epoch": 1.1, "learning_rate": 8.89516192268486e-05, "loss": 2.8032, "step": 14090 }, { "epoch": 1.1, "learning_rate": 8.89508350976241e-05, "loss": 2.9951, "step": 14091 }, { "epoch": 1.1, "learning_rate": 8.89500509683996e-05, "loss": 3.113, "step": 14092 }, { "epoch": 1.11, "learning_rate": 8.894926683917511e-05, "loss": 3.0417, "step": 14093 }, { "epoch": 1.11, "learning_rate": 8.894848270995061e-05, "loss": 2.8045, "step": 14094 }, { "epoch": 1.11, "learning_rate": 8.894769858072612e-05, "loss": 2.8707, "step": 14095 }, { "epoch": 1.11, "learning_rate": 8.894691445150161e-05, "loss": 2.8821, "step": 14096 }, { "epoch": 1.11, "learning_rate": 8.894613032227711e-05, "loss": 2.7168, "step": 14097 }, { "epoch": 1.11, "learning_rate": 8.894534619305261e-05, "loss": 2.8021, "step": 14098 }, { "epoch": 1.11, "learning_rate": 8.894456206382812e-05, "loss": 3.0165, "step": 14099 }, { "epoch": 1.11, "learning_rate": 8.894377793460362e-05, "loss": 2.9273, "step": 14100 }, { "epoch": 1.11, "learning_rate": 8.894299380537913e-05, "loss": 2.8444, "step": 14101 }, { "epoch": 1.11, "learning_rate": 8.894220967615463e-05, "loss": 2.8788, "step": 14102 }, { "epoch": 1.11, "learning_rate": 8.894142554693013e-05, "loss": 2.8065, "step": 14103 }, { "epoch": 1.11, "learning_rate": 8.894064141770564e-05, "loss": 2.7096, "step": 14104 }, { "epoch": 1.11, "learning_rate": 8.893985728848114e-05, "loss": 3.2475, "step": 14105 }, { "epoch": 1.11, "learning_rate": 8.893907315925664e-05, "loss": 2.8791, "step": 14106 }, { "epoch": 1.11, "learning_rate": 8.893828903003215e-05, "loss": 2.7371, "step": 14107 }, { "epoch": 1.11, "learning_rate": 8.893750490080765e-05, "loss": 3.024, "step": 14108 }, { "epoch": 1.11, "learning_rate": 8.893672077158316e-05, "loss": 2.9645, "step": 14109 }, { "epoch": 1.11, "learning_rate": 8.893593664235866e-05, "loss": 2.7393, "step": 14110 }, { "epoch": 1.11, "learning_rate": 8.893515251313416e-05, "loss": 2.9671, "step": 14111 }, { "epoch": 1.11, "learning_rate": 8.893436838390967e-05, "loss": 2.8036, "step": 14112 }, { "epoch": 1.11, "learning_rate": 8.893358425468517e-05, "loss": 2.8088, "step": 14113 }, { "epoch": 1.11, "learning_rate": 8.893280012546067e-05, "loss": 2.8026, "step": 14114 }, { "epoch": 1.11, "learning_rate": 8.893201599623619e-05, "loss": 2.86, "step": 14115 }, { "epoch": 1.11, "learning_rate": 8.89312318670117e-05, "loss": 2.9961, "step": 14116 }, { "epoch": 1.11, "learning_rate": 8.89304477377872e-05, "loss": 2.8523, "step": 14117 }, { "epoch": 1.11, "learning_rate": 8.89296636085627e-05, "loss": 3.0011, "step": 14118 }, { "epoch": 1.11, "learning_rate": 8.89288794793382e-05, "loss": 2.8318, "step": 14119 }, { "epoch": 1.11, "learning_rate": 8.892809535011371e-05, "loss": 3.0128, "step": 14120 }, { "epoch": 1.11, "learning_rate": 8.892731122088921e-05, "loss": 2.7157, "step": 14121 }, { "epoch": 1.11, "learning_rate": 8.892652709166472e-05, "loss": 2.8657, "step": 14122 }, { "epoch": 1.11, "learning_rate": 8.892574296244022e-05, "loss": 2.5908, "step": 14123 }, { "epoch": 1.11, "learning_rate": 8.892495883321572e-05, "loss": 2.86, "step": 14124 }, { "epoch": 1.11, "learning_rate": 8.892417470399123e-05, "loss": 2.6713, "step": 14125 }, { "epoch": 1.11, "learning_rate": 8.892339057476673e-05, "loss": 2.8578, "step": 14126 }, { "epoch": 1.11, "learning_rate": 8.892260644554224e-05, "loss": 2.8677, "step": 14127 }, { "epoch": 1.11, "learning_rate": 8.892182231631774e-05, "loss": 2.8472, "step": 14128 }, { "epoch": 1.11, "learning_rate": 8.892103818709324e-05, "loss": 3.0313, "step": 14129 }, { "epoch": 1.11, "learning_rate": 8.892025405786875e-05, "loss": 2.8456, "step": 14130 }, { "epoch": 1.11, "learning_rate": 8.891946992864425e-05, "loss": 2.6735, "step": 14131 }, { "epoch": 1.11, "learning_rate": 8.891868579941974e-05, "loss": 2.757, "step": 14132 }, { "epoch": 1.11, "learning_rate": 8.891790167019524e-05, "loss": 2.7831, "step": 14133 }, { "epoch": 1.11, "learning_rate": 8.891711754097075e-05, "loss": 2.9452, "step": 14134 }, { "epoch": 1.11, "learning_rate": 8.891633341174625e-05, "loss": 2.9574, "step": 14135 }, { "epoch": 1.11, "learning_rate": 8.891554928252175e-05, "loss": 3.0956, "step": 14136 }, { "epoch": 1.11, "learning_rate": 8.891476515329726e-05, "loss": 2.6828, "step": 14137 }, { "epoch": 1.11, "learning_rate": 8.891398102407276e-05, "loss": 3.1351, "step": 14138 }, { "epoch": 1.11, "learning_rate": 8.891319689484827e-05, "loss": 2.9266, "step": 14139 }, { "epoch": 1.11, "learning_rate": 8.891241276562377e-05, "loss": 2.9232, "step": 14140 }, { "epoch": 1.11, "learning_rate": 8.891162863639929e-05, "loss": 2.8299, "step": 14141 }, { "epoch": 1.11, "learning_rate": 8.891084450717479e-05, "loss": 3.0082, "step": 14142 }, { "epoch": 1.11, "learning_rate": 8.89100603779503e-05, "loss": 2.8668, "step": 14143 }, { "epoch": 1.11, "learning_rate": 8.89092762487258e-05, "loss": 2.838, "step": 14144 }, { "epoch": 1.11, "learning_rate": 8.89084921195013e-05, "loss": 2.8981, "step": 14145 }, { "epoch": 1.11, "learning_rate": 8.89077079902768e-05, "loss": 3.092, "step": 14146 }, { "epoch": 1.11, "learning_rate": 8.890692386105231e-05, "loss": 2.989, "step": 14147 }, { "epoch": 1.11, "learning_rate": 8.890613973182781e-05, "loss": 2.9743, "step": 14148 }, { "epoch": 1.11, "learning_rate": 8.890535560260332e-05, "loss": 3.0796, "step": 14149 }, { "epoch": 1.11, "learning_rate": 8.890457147337882e-05, "loss": 2.9127, "step": 14150 }, { "epoch": 1.11, "learning_rate": 8.890378734415432e-05, "loss": 3.311, "step": 14151 }, { "epoch": 1.11, "learning_rate": 8.890300321492983e-05, "loss": 2.9321, "step": 14152 }, { "epoch": 1.11, "learning_rate": 8.890221908570533e-05, "loss": 2.9817, "step": 14153 }, { "epoch": 1.11, "learning_rate": 8.890143495648083e-05, "loss": 2.9339, "step": 14154 }, { "epoch": 1.11, "learning_rate": 8.890065082725634e-05, "loss": 2.9742, "step": 14155 }, { "epoch": 1.11, "learning_rate": 8.889986669803184e-05, "loss": 2.7879, "step": 14156 }, { "epoch": 1.11, "learning_rate": 8.889908256880735e-05, "loss": 3.2924, "step": 14157 }, { "epoch": 1.11, "learning_rate": 8.889829843958285e-05, "loss": 2.853, "step": 14158 }, { "epoch": 1.11, "learning_rate": 8.889751431035835e-05, "loss": 2.7234, "step": 14159 }, { "epoch": 1.11, "learning_rate": 8.889673018113386e-05, "loss": 2.6262, "step": 14160 }, { "epoch": 1.11, "learning_rate": 8.889594605190936e-05, "loss": 3.0171, "step": 14161 }, { "epoch": 1.11, "learning_rate": 8.889516192268486e-05, "loss": 2.7089, "step": 14162 }, { "epoch": 1.11, "learning_rate": 8.889437779346037e-05, "loss": 2.7282, "step": 14163 }, { "epoch": 1.11, "learning_rate": 8.889359366423587e-05, "loss": 2.9899, "step": 14164 }, { "epoch": 1.11, "learning_rate": 8.889280953501137e-05, "loss": 2.9379, "step": 14165 }, { "epoch": 1.11, "learning_rate": 8.889202540578688e-05, "loss": 2.8593, "step": 14166 }, { "epoch": 1.11, "learning_rate": 8.889124127656238e-05, "loss": 2.6924, "step": 14167 }, { "epoch": 1.11, "learning_rate": 8.889045714733789e-05, "loss": 2.7647, "step": 14168 }, { "epoch": 1.11, "learning_rate": 8.888967301811339e-05, "loss": 3.1815, "step": 14169 }, { "epoch": 1.11, "learning_rate": 8.888888888888889e-05, "loss": 2.8557, "step": 14170 }, { "epoch": 1.11, "learning_rate": 8.88881047596644e-05, "loss": 2.9241, "step": 14171 }, { "epoch": 1.11, "learning_rate": 8.88873206304399e-05, "loss": 2.8853, "step": 14172 }, { "epoch": 1.11, "learning_rate": 8.88865365012154e-05, "loss": 2.8697, "step": 14173 }, { "epoch": 1.11, "learning_rate": 8.888575237199091e-05, "loss": 2.5892, "step": 14174 }, { "epoch": 1.11, "learning_rate": 8.888496824276641e-05, "loss": 3.0783, "step": 14175 }, { "epoch": 1.11, "learning_rate": 8.888418411354192e-05, "loss": 3.0939, "step": 14176 }, { "epoch": 1.11, "learning_rate": 8.888339998431742e-05, "loss": 2.8951, "step": 14177 }, { "epoch": 1.11, "learning_rate": 8.888261585509292e-05, "loss": 2.9344, "step": 14178 }, { "epoch": 1.11, "learning_rate": 8.888183172586843e-05, "loss": 2.8382, "step": 14179 }, { "epoch": 1.11, "learning_rate": 8.888104759664393e-05, "loss": 2.9629, "step": 14180 }, { "epoch": 1.11, "learning_rate": 8.888026346741943e-05, "loss": 2.9217, "step": 14181 }, { "epoch": 1.11, "learning_rate": 8.887947933819494e-05, "loss": 3.0343, "step": 14182 }, { "epoch": 1.11, "learning_rate": 8.887869520897044e-05, "loss": 2.7136, "step": 14183 }, { "epoch": 1.11, "learning_rate": 8.887791107974594e-05, "loss": 2.9053, "step": 14184 }, { "epoch": 1.11, "learning_rate": 8.887712695052145e-05, "loss": 3.1199, "step": 14185 }, { "epoch": 1.11, "learning_rate": 8.887634282129695e-05, "loss": 3.0001, "step": 14186 }, { "epoch": 1.11, "learning_rate": 8.887555869207246e-05, "loss": 3.0952, "step": 14187 }, { "epoch": 1.11, "learning_rate": 8.887477456284796e-05, "loss": 2.8673, "step": 14188 }, { "epoch": 1.11, "learning_rate": 8.887399043362346e-05, "loss": 2.8047, "step": 14189 }, { "epoch": 1.11, "learning_rate": 8.887320630439897e-05, "loss": 2.8654, "step": 14190 }, { "epoch": 1.11, "learning_rate": 8.887242217517447e-05, "loss": 2.8844, "step": 14191 }, { "epoch": 1.11, "learning_rate": 8.887163804594997e-05, "loss": 2.923, "step": 14192 }, { "epoch": 1.11, "learning_rate": 8.887085391672548e-05, "loss": 2.7054, "step": 14193 }, { "epoch": 1.11, "learning_rate": 8.887006978750098e-05, "loss": 2.9481, "step": 14194 }, { "epoch": 1.11, "learning_rate": 8.886928565827649e-05, "loss": 2.9804, "step": 14195 }, { "epoch": 1.11, "learning_rate": 8.886850152905199e-05, "loss": 2.7677, "step": 14196 }, { "epoch": 1.11, "learning_rate": 8.886771739982749e-05, "loss": 2.9311, "step": 14197 }, { "epoch": 1.11, "learning_rate": 8.886693327060301e-05, "loss": 3.0682, "step": 14198 }, { "epoch": 1.11, "learning_rate": 8.886614914137851e-05, "loss": 2.7002, "step": 14199 }, { "epoch": 1.11, "learning_rate": 8.886536501215402e-05, "loss": 2.8157, "step": 14200 }, { "epoch": 1.11, "learning_rate": 8.886458088292952e-05, "loss": 2.9076, "step": 14201 }, { "epoch": 1.11, "learning_rate": 8.886379675370502e-05, "loss": 2.9449, "step": 14202 }, { "epoch": 1.11, "learning_rate": 8.886301262448053e-05, "loss": 2.7814, "step": 14203 }, { "epoch": 1.11, "learning_rate": 8.886222849525603e-05, "loss": 2.8829, "step": 14204 }, { "epoch": 1.11, "learning_rate": 8.886144436603152e-05, "loss": 3.1152, "step": 14205 }, { "epoch": 1.11, "learning_rate": 8.886066023680703e-05, "loss": 2.7417, "step": 14206 }, { "epoch": 1.11, "learning_rate": 8.885987610758253e-05, "loss": 3.0117, "step": 14207 }, { "epoch": 1.11, "learning_rate": 8.885909197835803e-05, "loss": 2.872, "step": 14208 }, { "epoch": 1.11, "learning_rate": 8.885830784913354e-05, "loss": 3.0146, "step": 14209 }, { "epoch": 1.11, "learning_rate": 8.885752371990904e-05, "loss": 2.8769, "step": 14210 }, { "epoch": 1.11, "learning_rate": 8.885673959068454e-05, "loss": 3.0767, "step": 14211 }, { "epoch": 1.11, "learning_rate": 8.885595546146005e-05, "loss": 2.8256, "step": 14212 }, { "epoch": 1.11, "learning_rate": 8.885517133223555e-05, "loss": 3.0498, "step": 14213 }, { "epoch": 1.11, "learning_rate": 8.885438720301105e-05, "loss": 2.8123, "step": 14214 }, { "epoch": 1.11, "learning_rate": 8.885360307378656e-05, "loss": 2.7529, "step": 14215 }, { "epoch": 1.11, "learning_rate": 8.885281894456206e-05, "loss": 3.0772, "step": 14216 }, { "epoch": 1.11, "learning_rate": 8.885203481533757e-05, "loss": 2.9016, "step": 14217 }, { "epoch": 1.11, "learning_rate": 8.885125068611307e-05, "loss": 2.9003, "step": 14218 }, { "epoch": 1.11, "learning_rate": 8.885046655688857e-05, "loss": 2.9354, "step": 14219 }, { "epoch": 1.12, "learning_rate": 8.884968242766408e-05, "loss": 2.902, "step": 14220 }, { "epoch": 1.12, "learning_rate": 8.884889829843958e-05, "loss": 2.8976, "step": 14221 }, { "epoch": 1.12, "learning_rate": 8.884811416921508e-05, "loss": 2.8887, "step": 14222 }, { "epoch": 1.12, "learning_rate": 8.88473300399906e-05, "loss": 2.9021, "step": 14223 }, { "epoch": 1.12, "learning_rate": 8.88465459107661e-05, "loss": 2.9714, "step": 14224 }, { "epoch": 1.12, "learning_rate": 8.884576178154161e-05, "loss": 2.7947, "step": 14225 }, { "epoch": 1.12, "learning_rate": 8.884497765231711e-05, "loss": 2.9174, "step": 14226 }, { "epoch": 1.12, "learning_rate": 8.884419352309262e-05, "loss": 2.916, "step": 14227 }, { "epoch": 1.12, "learning_rate": 8.884340939386812e-05, "loss": 2.963, "step": 14228 }, { "epoch": 1.12, "learning_rate": 8.884262526464362e-05, "loss": 2.955, "step": 14229 }, { "epoch": 1.12, "learning_rate": 8.884184113541913e-05, "loss": 2.992, "step": 14230 }, { "epoch": 1.12, "learning_rate": 8.884105700619463e-05, "loss": 2.7748, "step": 14231 }, { "epoch": 1.12, "learning_rate": 8.884027287697013e-05, "loss": 2.8587, "step": 14232 }, { "epoch": 1.12, "learning_rate": 8.883948874774564e-05, "loss": 2.843, "step": 14233 }, { "epoch": 1.12, "learning_rate": 8.883870461852114e-05, "loss": 2.9207, "step": 14234 }, { "epoch": 1.12, "learning_rate": 8.883792048929665e-05, "loss": 2.9315, "step": 14235 }, { "epoch": 1.12, "learning_rate": 8.883713636007215e-05, "loss": 2.9385, "step": 14236 }, { "epoch": 1.12, "learning_rate": 8.883635223084765e-05, "loss": 2.9861, "step": 14237 }, { "epoch": 1.12, "learning_rate": 8.883556810162316e-05, "loss": 3.155, "step": 14238 }, { "epoch": 1.12, "learning_rate": 8.883478397239866e-05, "loss": 3.2582, "step": 14239 }, { "epoch": 1.12, "learning_rate": 8.883399984317416e-05, "loss": 3.0493, "step": 14240 }, { "epoch": 1.12, "learning_rate": 8.883321571394965e-05, "loss": 3.0577, "step": 14241 }, { "epoch": 1.12, "learning_rate": 8.883243158472516e-05, "loss": 2.8065, "step": 14242 }, { "epoch": 1.12, "learning_rate": 8.883164745550066e-05, "loss": 2.6807, "step": 14243 }, { "epoch": 1.12, "learning_rate": 8.883086332627617e-05, "loss": 2.9155, "step": 14244 }, { "epoch": 1.12, "learning_rate": 8.883007919705167e-05, "loss": 3.0782, "step": 14245 }, { "epoch": 1.12, "learning_rate": 8.882929506782717e-05, "loss": 2.846, "step": 14246 }, { "epoch": 1.12, "learning_rate": 8.882851093860268e-05, "loss": 3.0429, "step": 14247 }, { "epoch": 1.12, "learning_rate": 8.882772680937818e-05, "loss": 2.7796, "step": 14248 }, { "epoch": 1.12, "learning_rate": 8.88269426801537e-05, "loss": 2.8994, "step": 14249 }, { "epoch": 1.12, "learning_rate": 8.88261585509292e-05, "loss": 2.8949, "step": 14250 }, { "epoch": 1.12, "learning_rate": 8.88253744217047e-05, "loss": 2.9878, "step": 14251 }, { "epoch": 1.12, "learning_rate": 8.882459029248021e-05, "loss": 2.9176, "step": 14252 }, { "epoch": 1.12, "learning_rate": 8.882380616325571e-05, "loss": 2.9994, "step": 14253 }, { "epoch": 1.12, "learning_rate": 8.882302203403122e-05, "loss": 2.8887, "step": 14254 }, { "epoch": 1.12, "learning_rate": 8.882223790480672e-05, "loss": 2.8475, "step": 14255 }, { "epoch": 1.12, "learning_rate": 8.882145377558222e-05, "loss": 2.6149, "step": 14256 }, { "epoch": 1.12, "learning_rate": 8.882066964635773e-05, "loss": 2.8595, "step": 14257 }, { "epoch": 1.12, "learning_rate": 8.881988551713323e-05, "loss": 2.7192, "step": 14258 }, { "epoch": 1.12, "learning_rate": 8.881910138790873e-05, "loss": 2.8009, "step": 14259 }, { "epoch": 1.12, "learning_rate": 8.881831725868424e-05, "loss": 2.8376, "step": 14260 }, { "epoch": 1.12, "learning_rate": 8.881753312945974e-05, "loss": 2.8748, "step": 14261 }, { "epoch": 1.12, "learning_rate": 8.881674900023525e-05, "loss": 2.8095, "step": 14262 }, { "epoch": 1.12, "learning_rate": 8.881596487101075e-05, "loss": 2.9951, "step": 14263 }, { "epoch": 1.12, "learning_rate": 8.881518074178625e-05, "loss": 3.041, "step": 14264 }, { "epoch": 1.12, "learning_rate": 8.881439661256176e-05, "loss": 2.8034, "step": 14265 }, { "epoch": 1.12, "learning_rate": 8.881361248333726e-05, "loss": 2.7606, "step": 14266 }, { "epoch": 1.12, "learning_rate": 8.881282835411276e-05, "loss": 3.0038, "step": 14267 }, { "epoch": 1.12, "learning_rate": 8.881204422488827e-05, "loss": 3.0308, "step": 14268 }, { "epoch": 1.12, "learning_rate": 8.881126009566377e-05, "loss": 2.7133, "step": 14269 }, { "epoch": 1.12, "learning_rate": 8.881047596643927e-05, "loss": 2.8585, "step": 14270 }, { "epoch": 1.12, "learning_rate": 8.880969183721478e-05, "loss": 2.9029, "step": 14271 }, { "epoch": 1.12, "learning_rate": 8.880890770799028e-05, "loss": 3.1497, "step": 14272 }, { "epoch": 1.12, "learning_rate": 8.880812357876579e-05, "loss": 3.0573, "step": 14273 }, { "epoch": 1.12, "learning_rate": 8.880733944954129e-05, "loss": 2.7947, "step": 14274 }, { "epoch": 1.12, "learning_rate": 8.880655532031679e-05, "loss": 2.7216, "step": 14275 }, { "epoch": 1.12, "learning_rate": 8.88057711910923e-05, "loss": 2.9249, "step": 14276 }, { "epoch": 1.12, "learning_rate": 8.88049870618678e-05, "loss": 2.9204, "step": 14277 }, { "epoch": 1.12, "learning_rate": 8.88042029326433e-05, "loss": 3.3223, "step": 14278 }, { "epoch": 1.12, "learning_rate": 8.880341880341881e-05, "loss": 3.0669, "step": 14279 }, { "epoch": 1.12, "learning_rate": 8.880263467419431e-05, "loss": 3.0014, "step": 14280 }, { "epoch": 1.12, "learning_rate": 8.880185054496981e-05, "loss": 2.6607, "step": 14281 }, { "epoch": 1.12, "learning_rate": 8.880106641574532e-05, "loss": 2.9841, "step": 14282 }, { "epoch": 1.12, "learning_rate": 8.880028228652082e-05, "loss": 2.9938, "step": 14283 }, { "epoch": 1.12, "learning_rate": 8.879949815729633e-05, "loss": 2.9879, "step": 14284 }, { "epoch": 1.12, "learning_rate": 8.879871402807183e-05, "loss": 2.9953, "step": 14285 }, { "epoch": 1.12, "learning_rate": 8.879792989884733e-05, "loss": 2.893, "step": 14286 }, { "epoch": 1.12, "learning_rate": 8.879714576962284e-05, "loss": 2.9768, "step": 14287 }, { "epoch": 1.12, "learning_rate": 8.879636164039834e-05, "loss": 2.836, "step": 14288 }, { "epoch": 1.12, "learning_rate": 8.879557751117384e-05, "loss": 3.1694, "step": 14289 }, { "epoch": 1.12, "learning_rate": 8.879479338194935e-05, "loss": 2.8204, "step": 14290 }, { "epoch": 1.12, "learning_rate": 8.879400925272485e-05, "loss": 2.8271, "step": 14291 }, { "epoch": 1.12, "learning_rate": 8.879322512350036e-05, "loss": 2.7757, "step": 14292 }, { "epoch": 1.12, "learning_rate": 8.879244099427586e-05, "loss": 3.3164, "step": 14293 }, { "epoch": 1.12, "learning_rate": 8.879165686505136e-05, "loss": 2.87, "step": 14294 }, { "epoch": 1.12, "learning_rate": 8.879087273582687e-05, "loss": 3.2852, "step": 14295 }, { "epoch": 1.12, "learning_rate": 8.879008860660237e-05, "loss": 2.8304, "step": 14296 }, { "epoch": 1.12, "learning_rate": 8.878930447737787e-05, "loss": 2.9405, "step": 14297 }, { "epoch": 1.12, "learning_rate": 8.878852034815338e-05, "loss": 3.0403, "step": 14298 }, { "epoch": 1.12, "learning_rate": 8.878773621892888e-05, "loss": 3.2194, "step": 14299 }, { "epoch": 1.12, "learning_rate": 8.878695208970438e-05, "loss": 2.8087, "step": 14300 }, { "epoch": 1.12, "learning_rate": 8.878616796047989e-05, "loss": 3.1145, "step": 14301 }, { "epoch": 1.12, "learning_rate": 8.878538383125539e-05, "loss": 2.7816, "step": 14302 }, { "epoch": 1.12, "learning_rate": 8.87845997020309e-05, "loss": 2.9301, "step": 14303 }, { "epoch": 1.12, "learning_rate": 8.87838155728064e-05, "loss": 3.0389, "step": 14304 }, { "epoch": 1.12, "learning_rate": 8.87830314435819e-05, "loss": 2.8587, "step": 14305 }, { "epoch": 1.12, "learning_rate": 8.878224731435742e-05, "loss": 2.9284, "step": 14306 }, { "epoch": 1.12, "learning_rate": 8.878146318513292e-05, "loss": 2.7975, "step": 14307 }, { "epoch": 1.12, "learning_rate": 8.878067905590843e-05, "loss": 2.7556, "step": 14308 }, { "epoch": 1.12, "learning_rate": 8.877989492668393e-05, "loss": 2.9034, "step": 14309 }, { "epoch": 1.12, "learning_rate": 8.877911079745944e-05, "loss": 3.0055, "step": 14310 }, { "epoch": 1.12, "learning_rate": 8.877832666823494e-05, "loss": 3.0167, "step": 14311 }, { "epoch": 1.12, "learning_rate": 8.877754253901044e-05, "loss": 2.9184, "step": 14312 }, { "epoch": 1.12, "learning_rate": 8.877675840978593e-05, "loss": 2.6469, "step": 14313 }, { "epoch": 1.12, "learning_rate": 8.877597428056144e-05, "loss": 3.2677, "step": 14314 }, { "epoch": 1.12, "learning_rate": 8.877519015133694e-05, "loss": 2.7538, "step": 14315 }, { "epoch": 1.12, "learning_rate": 8.877440602211244e-05, "loss": 3.338, "step": 14316 }, { "epoch": 1.12, "learning_rate": 8.877362189288795e-05, "loss": 3.2723, "step": 14317 }, { "epoch": 1.12, "learning_rate": 8.877283776366345e-05, "loss": 3.0669, "step": 14318 }, { "epoch": 1.12, "learning_rate": 8.877205363443895e-05, "loss": 3.0412, "step": 14319 }, { "epoch": 1.12, "learning_rate": 8.877126950521446e-05, "loss": 3.0165, "step": 14320 }, { "epoch": 1.12, "learning_rate": 8.877048537598996e-05, "loss": 2.7668, "step": 14321 }, { "epoch": 1.12, "learning_rate": 8.876970124676547e-05, "loss": 3.0208, "step": 14322 }, { "epoch": 1.12, "learning_rate": 8.876891711754097e-05, "loss": 2.9453, "step": 14323 }, { "epoch": 1.12, "learning_rate": 8.876813298831647e-05, "loss": 3.0332, "step": 14324 }, { "epoch": 1.12, "learning_rate": 8.876734885909198e-05, "loss": 3.0055, "step": 14325 }, { "epoch": 1.12, "learning_rate": 8.876656472986748e-05, "loss": 2.8665, "step": 14326 }, { "epoch": 1.12, "learning_rate": 8.876578060064298e-05, "loss": 3.2519, "step": 14327 }, { "epoch": 1.12, "learning_rate": 8.876499647141849e-05, "loss": 2.8888, "step": 14328 }, { "epoch": 1.12, "learning_rate": 8.876421234219399e-05, "loss": 2.8118, "step": 14329 }, { "epoch": 1.12, "learning_rate": 8.87634282129695e-05, "loss": 2.8081, "step": 14330 }, { "epoch": 1.12, "learning_rate": 8.876264408374501e-05, "loss": 2.9288, "step": 14331 }, { "epoch": 1.12, "learning_rate": 8.876185995452052e-05, "loss": 2.758, "step": 14332 }, { "epoch": 1.12, "learning_rate": 8.876107582529602e-05, "loss": 2.7451, "step": 14333 }, { "epoch": 1.12, "learning_rate": 8.876029169607152e-05, "loss": 3.0639, "step": 14334 }, { "epoch": 1.12, "learning_rate": 8.875950756684703e-05, "loss": 2.8634, "step": 14335 }, { "epoch": 1.12, "learning_rate": 8.875872343762253e-05, "loss": 2.8052, "step": 14336 }, { "epoch": 1.12, "learning_rate": 8.875793930839803e-05, "loss": 3.0072, "step": 14337 }, { "epoch": 1.12, "learning_rate": 8.875715517917354e-05, "loss": 3.0075, "step": 14338 }, { "epoch": 1.12, "learning_rate": 8.875637104994904e-05, "loss": 2.533, "step": 14339 }, { "epoch": 1.12, "learning_rate": 8.875558692072455e-05, "loss": 3.0333, "step": 14340 }, { "epoch": 1.12, "learning_rate": 8.875480279150005e-05, "loss": 2.8725, "step": 14341 }, { "epoch": 1.12, "learning_rate": 8.875401866227555e-05, "loss": 2.8578, "step": 14342 }, { "epoch": 1.12, "learning_rate": 8.875323453305106e-05, "loss": 2.753, "step": 14343 }, { "epoch": 1.12, "learning_rate": 8.875245040382656e-05, "loss": 2.8098, "step": 14344 }, { "epoch": 1.12, "learning_rate": 8.875166627460206e-05, "loss": 2.8045, "step": 14345 }, { "epoch": 1.12, "learning_rate": 8.875088214537757e-05, "loss": 3.0183, "step": 14346 }, { "epoch": 1.12, "learning_rate": 8.875009801615307e-05, "loss": 3.0826, "step": 14347 }, { "epoch": 1.13, "learning_rate": 8.874931388692857e-05, "loss": 2.7828, "step": 14348 }, { "epoch": 1.13, "learning_rate": 8.874852975770407e-05, "loss": 2.8828, "step": 14349 }, { "epoch": 1.13, "learning_rate": 8.874774562847957e-05, "loss": 2.8188, "step": 14350 }, { "epoch": 1.13, "learning_rate": 8.874696149925507e-05, "loss": 2.8037, "step": 14351 }, { "epoch": 1.13, "learning_rate": 8.874617737003058e-05, "loss": 2.9687, "step": 14352 }, { "epoch": 1.13, "learning_rate": 8.874539324080608e-05, "loss": 2.8099, "step": 14353 }, { "epoch": 1.13, "learning_rate": 8.874460911158158e-05, "loss": 2.846, "step": 14354 }, { "epoch": 1.13, "learning_rate": 8.874382498235709e-05, "loss": 2.8835, "step": 14355 }, { "epoch": 1.13, "learning_rate": 8.874304085313259e-05, "loss": 2.8193, "step": 14356 }, { "epoch": 1.13, "learning_rate": 8.874225672390811e-05, "loss": 3.1005, "step": 14357 }, { "epoch": 1.13, "learning_rate": 8.874147259468361e-05, "loss": 2.901, "step": 14358 }, { "epoch": 1.13, "learning_rate": 8.874068846545912e-05, "loss": 2.9002, "step": 14359 }, { "epoch": 1.13, "learning_rate": 8.873990433623462e-05, "loss": 2.9034, "step": 14360 }, { "epoch": 1.13, "learning_rate": 8.873912020701012e-05, "loss": 2.7798, "step": 14361 }, { "epoch": 1.13, "learning_rate": 8.873833607778563e-05, "loss": 2.9682, "step": 14362 }, { "epoch": 1.13, "learning_rate": 8.873755194856113e-05, "loss": 3.0346, "step": 14363 }, { "epoch": 1.13, "learning_rate": 8.873676781933663e-05, "loss": 2.8413, "step": 14364 }, { "epoch": 1.13, "learning_rate": 8.873598369011214e-05, "loss": 2.8703, "step": 14365 }, { "epoch": 1.13, "learning_rate": 8.873519956088764e-05, "loss": 2.9787, "step": 14366 }, { "epoch": 1.13, "learning_rate": 8.873441543166314e-05, "loss": 2.948, "step": 14367 }, { "epoch": 1.13, "learning_rate": 8.873363130243865e-05, "loss": 2.8044, "step": 14368 }, { "epoch": 1.13, "learning_rate": 8.873284717321415e-05, "loss": 2.9895, "step": 14369 }, { "epoch": 1.13, "learning_rate": 8.873206304398966e-05, "loss": 2.892, "step": 14370 }, { "epoch": 1.13, "learning_rate": 8.873127891476516e-05, "loss": 3.0227, "step": 14371 }, { "epoch": 1.13, "learning_rate": 8.873049478554066e-05, "loss": 3.0437, "step": 14372 }, { "epoch": 1.13, "learning_rate": 8.872971065631617e-05, "loss": 2.8199, "step": 14373 }, { "epoch": 1.13, "learning_rate": 8.872892652709167e-05, "loss": 3.0339, "step": 14374 }, { "epoch": 1.13, "learning_rate": 8.872814239786717e-05, "loss": 2.9667, "step": 14375 }, { "epoch": 1.13, "learning_rate": 8.872735826864268e-05, "loss": 2.7904, "step": 14376 }, { "epoch": 1.13, "learning_rate": 8.872657413941818e-05, "loss": 2.8755, "step": 14377 }, { "epoch": 1.13, "learning_rate": 8.872579001019369e-05, "loss": 2.9309, "step": 14378 }, { "epoch": 1.13, "learning_rate": 8.872500588096919e-05, "loss": 2.6598, "step": 14379 }, { "epoch": 1.13, "learning_rate": 8.872422175174469e-05, "loss": 2.875, "step": 14380 }, { "epoch": 1.13, "learning_rate": 8.87234376225202e-05, "loss": 2.9117, "step": 14381 }, { "epoch": 1.13, "learning_rate": 8.87226534932957e-05, "loss": 2.7703, "step": 14382 }, { "epoch": 1.13, "learning_rate": 8.87218693640712e-05, "loss": 2.6542, "step": 14383 }, { "epoch": 1.13, "learning_rate": 8.872108523484671e-05, "loss": 2.9353, "step": 14384 }, { "epoch": 1.13, "learning_rate": 8.872030110562221e-05, "loss": 2.8471, "step": 14385 }, { "epoch": 1.13, "learning_rate": 8.871951697639771e-05, "loss": 2.8594, "step": 14386 }, { "epoch": 1.13, "learning_rate": 8.871873284717322e-05, "loss": 3.0186, "step": 14387 }, { "epoch": 1.13, "learning_rate": 8.871794871794872e-05, "loss": 2.9406, "step": 14388 }, { "epoch": 1.13, "learning_rate": 8.871716458872423e-05, "loss": 2.8335, "step": 14389 }, { "epoch": 1.13, "learning_rate": 8.871638045949973e-05, "loss": 2.8582, "step": 14390 }, { "epoch": 1.13, "learning_rate": 8.871559633027523e-05, "loss": 2.8372, "step": 14391 }, { "epoch": 1.13, "learning_rate": 8.871481220105074e-05, "loss": 2.9509, "step": 14392 }, { "epoch": 1.13, "learning_rate": 8.871402807182624e-05, "loss": 2.8974, "step": 14393 }, { "epoch": 1.13, "learning_rate": 8.871324394260174e-05, "loss": 2.7951, "step": 14394 }, { "epoch": 1.13, "learning_rate": 8.871245981337725e-05, "loss": 2.9139, "step": 14395 }, { "epoch": 1.13, "learning_rate": 8.871167568415275e-05, "loss": 3.0212, "step": 14396 }, { "epoch": 1.13, "learning_rate": 8.871089155492826e-05, "loss": 2.7692, "step": 14397 }, { "epoch": 1.13, "learning_rate": 8.871010742570376e-05, "loss": 2.9848, "step": 14398 }, { "epoch": 1.13, "learning_rate": 8.870932329647926e-05, "loss": 2.844, "step": 14399 }, { "epoch": 1.13, "learning_rate": 8.870853916725477e-05, "loss": 2.999, "step": 14400 }, { "epoch": 1.13, "learning_rate": 8.870775503803027e-05, "loss": 3.0593, "step": 14401 }, { "epoch": 1.13, "learning_rate": 8.870697090880577e-05, "loss": 2.9754, "step": 14402 }, { "epoch": 1.13, "learning_rate": 8.870618677958128e-05, "loss": 2.7455, "step": 14403 }, { "epoch": 1.13, "learning_rate": 8.870540265035678e-05, "loss": 2.9144, "step": 14404 }, { "epoch": 1.13, "learning_rate": 8.870461852113228e-05, "loss": 3.0426, "step": 14405 }, { "epoch": 1.13, "learning_rate": 8.870383439190779e-05, "loss": 3.045, "step": 14406 }, { "epoch": 1.13, "learning_rate": 8.870305026268329e-05, "loss": 2.9673, "step": 14407 }, { "epoch": 1.13, "learning_rate": 8.87022661334588e-05, "loss": 2.8961, "step": 14408 }, { "epoch": 1.13, "learning_rate": 8.87014820042343e-05, "loss": 2.8725, "step": 14409 }, { "epoch": 1.13, "learning_rate": 8.87006978750098e-05, "loss": 2.9079, "step": 14410 }, { "epoch": 1.13, "learning_rate": 8.86999137457853e-05, "loss": 2.8721, "step": 14411 }, { "epoch": 1.13, "learning_rate": 8.869912961656081e-05, "loss": 2.7794, "step": 14412 }, { "epoch": 1.13, "learning_rate": 8.869834548733631e-05, "loss": 3.0426, "step": 14413 }, { "epoch": 1.13, "learning_rate": 8.869756135811183e-05, "loss": 2.895, "step": 14414 }, { "epoch": 1.13, "learning_rate": 8.869677722888733e-05, "loss": 2.816, "step": 14415 }, { "epoch": 1.13, "learning_rate": 8.869599309966284e-05, "loss": 3.1279, "step": 14416 }, { "epoch": 1.13, "learning_rate": 8.869520897043834e-05, "loss": 2.8812, "step": 14417 }, { "epoch": 1.13, "learning_rate": 8.869442484121385e-05, "loss": 2.7851, "step": 14418 }, { "epoch": 1.13, "learning_rate": 8.869364071198935e-05, "loss": 3.0348, "step": 14419 }, { "epoch": 1.13, "learning_rate": 8.869285658276485e-05, "loss": 3.0766, "step": 14420 }, { "epoch": 1.13, "learning_rate": 8.869207245354034e-05, "loss": 3.0668, "step": 14421 }, { "epoch": 1.13, "learning_rate": 8.869128832431585e-05, "loss": 2.8638, "step": 14422 }, { "epoch": 1.13, "learning_rate": 8.869050419509135e-05, "loss": 3.1843, "step": 14423 }, { "epoch": 1.13, "learning_rate": 8.868972006586685e-05, "loss": 2.7198, "step": 14424 }, { "epoch": 1.13, "learning_rate": 8.868893593664236e-05, "loss": 2.9726, "step": 14425 }, { "epoch": 1.13, "learning_rate": 8.868815180741786e-05, "loss": 2.7765, "step": 14426 }, { "epoch": 1.13, "learning_rate": 8.868736767819337e-05, "loss": 3.2, "step": 14427 }, { "epoch": 1.13, "learning_rate": 8.868658354896887e-05, "loss": 2.7094, "step": 14428 }, { "epoch": 1.13, "learning_rate": 8.868579941974437e-05, "loss": 2.6904, "step": 14429 }, { "epoch": 1.13, "learning_rate": 8.868501529051988e-05, "loss": 2.8949, "step": 14430 }, { "epoch": 1.13, "learning_rate": 8.868423116129538e-05, "loss": 3.2133, "step": 14431 }, { "epoch": 1.13, "learning_rate": 8.868344703207088e-05, "loss": 3.071, "step": 14432 }, { "epoch": 1.13, "learning_rate": 8.868266290284639e-05, "loss": 2.836, "step": 14433 }, { "epoch": 1.13, "learning_rate": 8.868187877362189e-05, "loss": 3.1137, "step": 14434 }, { "epoch": 1.13, "learning_rate": 8.86810946443974e-05, "loss": 3.3068, "step": 14435 }, { "epoch": 1.13, "learning_rate": 8.86803105151729e-05, "loss": 2.8394, "step": 14436 }, { "epoch": 1.13, "learning_rate": 8.86795263859484e-05, "loss": 2.8252, "step": 14437 }, { "epoch": 1.13, "learning_rate": 8.86787422567239e-05, "loss": 3.1019, "step": 14438 }, { "epoch": 1.13, "learning_rate": 8.867795812749942e-05, "loss": 2.7743, "step": 14439 }, { "epoch": 1.13, "learning_rate": 8.867717399827493e-05, "loss": 3.0276, "step": 14440 }, { "epoch": 1.13, "learning_rate": 8.867638986905043e-05, "loss": 2.7457, "step": 14441 }, { "epoch": 1.13, "learning_rate": 8.867560573982593e-05, "loss": 3.1149, "step": 14442 }, { "epoch": 1.13, "learning_rate": 8.867482161060144e-05, "loss": 2.9928, "step": 14443 }, { "epoch": 1.13, "learning_rate": 8.867403748137694e-05, "loss": 2.7223, "step": 14444 }, { "epoch": 1.13, "learning_rate": 8.867325335215245e-05, "loss": 3.0578, "step": 14445 }, { "epoch": 1.13, "learning_rate": 8.867246922292795e-05, "loss": 2.8777, "step": 14446 }, { "epoch": 1.13, "learning_rate": 8.867168509370345e-05, "loss": 2.9246, "step": 14447 }, { "epoch": 1.13, "learning_rate": 8.867090096447896e-05, "loss": 2.908, "step": 14448 }, { "epoch": 1.13, "learning_rate": 8.867011683525446e-05, "loss": 2.9449, "step": 14449 }, { "epoch": 1.13, "learning_rate": 8.866933270602996e-05, "loss": 2.8528, "step": 14450 }, { "epoch": 1.13, "learning_rate": 8.866854857680547e-05, "loss": 2.6869, "step": 14451 }, { "epoch": 1.13, "learning_rate": 8.866776444758097e-05, "loss": 2.8427, "step": 14452 }, { "epoch": 1.13, "learning_rate": 8.866698031835647e-05, "loss": 2.9406, "step": 14453 }, { "epoch": 1.13, "learning_rate": 8.866619618913198e-05, "loss": 2.9301, "step": 14454 }, { "epoch": 1.13, "learning_rate": 8.866541205990748e-05, "loss": 3.2786, "step": 14455 }, { "epoch": 1.13, "learning_rate": 8.866462793068299e-05, "loss": 2.8624, "step": 14456 }, { "epoch": 1.13, "learning_rate": 8.866384380145848e-05, "loss": 3.0138, "step": 14457 }, { "epoch": 1.13, "learning_rate": 8.866305967223398e-05, "loss": 2.8337, "step": 14458 }, { "epoch": 1.13, "learning_rate": 8.866227554300948e-05, "loss": 2.7713, "step": 14459 }, { "epoch": 1.13, "learning_rate": 8.866149141378499e-05, "loss": 2.9522, "step": 14460 }, { "epoch": 1.13, "learning_rate": 8.866070728456049e-05, "loss": 2.8337, "step": 14461 }, { "epoch": 1.13, "learning_rate": 8.8659923155336e-05, "loss": 2.8116, "step": 14462 }, { "epoch": 1.13, "learning_rate": 8.86591390261115e-05, "loss": 2.8293, "step": 14463 }, { "epoch": 1.13, "learning_rate": 8.8658354896887e-05, "loss": 2.8278, "step": 14464 }, { "epoch": 1.13, "learning_rate": 8.865757076766252e-05, "loss": 2.9157, "step": 14465 }, { "epoch": 1.13, "learning_rate": 8.865678663843802e-05, "loss": 2.9667, "step": 14466 }, { "epoch": 1.13, "learning_rate": 8.865600250921353e-05, "loss": 2.9842, "step": 14467 }, { "epoch": 1.13, "learning_rate": 8.865521837998903e-05, "loss": 2.793, "step": 14468 }, { "epoch": 1.13, "learning_rate": 8.865443425076453e-05, "loss": 2.9186, "step": 14469 }, { "epoch": 1.13, "learning_rate": 8.865365012154004e-05, "loss": 2.9763, "step": 14470 }, { "epoch": 1.13, "learning_rate": 8.865286599231554e-05, "loss": 2.6981, "step": 14471 }, { "epoch": 1.13, "learning_rate": 8.865208186309104e-05, "loss": 2.6173, "step": 14472 }, { "epoch": 1.13, "learning_rate": 8.865129773386655e-05, "loss": 2.7389, "step": 14473 }, { "epoch": 1.13, "learning_rate": 8.865051360464205e-05, "loss": 3.0859, "step": 14474 }, { "epoch": 1.14, "learning_rate": 8.864972947541756e-05, "loss": 2.653, "step": 14475 }, { "epoch": 1.14, "learning_rate": 8.864894534619306e-05, "loss": 2.8496, "step": 14476 }, { "epoch": 1.14, "learning_rate": 8.864816121696856e-05, "loss": 2.8314, "step": 14477 }, { "epoch": 1.14, "learning_rate": 8.864737708774407e-05, "loss": 2.9055, "step": 14478 }, { "epoch": 1.14, "learning_rate": 8.864659295851957e-05, "loss": 2.5923, "step": 14479 }, { "epoch": 1.14, "learning_rate": 8.864580882929507e-05, "loss": 3.0683, "step": 14480 }, { "epoch": 1.14, "learning_rate": 8.864502470007058e-05, "loss": 3.0867, "step": 14481 }, { "epoch": 1.14, "learning_rate": 8.864424057084608e-05, "loss": 3.1141, "step": 14482 }, { "epoch": 1.14, "learning_rate": 8.864345644162158e-05, "loss": 2.9701, "step": 14483 }, { "epoch": 1.14, "learning_rate": 8.864267231239709e-05, "loss": 3.0382, "step": 14484 }, { "epoch": 1.14, "learning_rate": 8.864188818317259e-05, "loss": 3.0107, "step": 14485 }, { "epoch": 1.14, "learning_rate": 8.86411040539481e-05, "loss": 2.7998, "step": 14486 }, { "epoch": 1.14, "learning_rate": 8.86403199247236e-05, "loss": 2.8245, "step": 14487 }, { "epoch": 1.14, "learning_rate": 8.86395357954991e-05, "loss": 2.8634, "step": 14488 }, { "epoch": 1.14, "learning_rate": 8.863875166627461e-05, "loss": 2.9681, "step": 14489 }, { "epoch": 1.14, "learning_rate": 8.863796753705011e-05, "loss": 3.2345, "step": 14490 }, { "epoch": 1.14, "learning_rate": 8.863718340782561e-05, "loss": 3.0296, "step": 14491 }, { "epoch": 1.14, "learning_rate": 8.863639927860112e-05, "loss": 2.9385, "step": 14492 }, { "epoch": 1.14, "learning_rate": 8.863561514937662e-05, "loss": 2.6991, "step": 14493 }, { "epoch": 1.14, "learning_rate": 8.863483102015213e-05, "loss": 3.1281, "step": 14494 }, { "epoch": 1.14, "learning_rate": 8.863404689092763e-05, "loss": 3.0714, "step": 14495 }, { "epoch": 1.14, "learning_rate": 8.863326276170313e-05, "loss": 3.0174, "step": 14496 }, { "epoch": 1.14, "learning_rate": 8.863247863247864e-05, "loss": 2.8852, "step": 14497 }, { "epoch": 1.14, "learning_rate": 8.863169450325414e-05, "loss": 3.0125, "step": 14498 }, { "epoch": 1.14, "learning_rate": 8.863091037402964e-05, "loss": 2.7698, "step": 14499 }, { "epoch": 1.14, "learning_rate": 8.863012624480515e-05, "loss": 3.0707, "step": 14500 }, { "epoch": 1.14, "learning_rate": 8.862934211558065e-05, "loss": 2.8457, "step": 14501 }, { "epoch": 1.14, "learning_rate": 8.862855798635615e-05, "loss": 2.8303, "step": 14502 }, { "epoch": 1.14, "learning_rate": 8.862777385713166e-05, "loss": 3.0823, "step": 14503 }, { "epoch": 1.14, "learning_rate": 8.862698972790716e-05, "loss": 2.8528, "step": 14504 }, { "epoch": 1.14, "learning_rate": 8.862620559868267e-05, "loss": 2.6291, "step": 14505 }, { "epoch": 1.14, "learning_rate": 8.862542146945817e-05, "loss": 2.6391, "step": 14506 }, { "epoch": 1.14, "learning_rate": 8.862463734023367e-05, "loss": 3.0573, "step": 14507 }, { "epoch": 1.14, "learning_rate": 8.862385321100918e-05, "loss": 2.8741, "step": 14508 }, { "epoch": 1.14, "learning_rate": 8.862306908178468e-05, "loss": 2.9204, "step": 14509 }, { "epoch": 1.14, "learning_rate": 8.862228495256018e-05, "loss": 3.016, "step": 14510 }, { "epoch": 1.14, "learning_rate": 8.862150082333569e-05, "loss": 3.029, "step": 14511 }, { "epoch": 1.14, "learning_rate": 8.862071669411119e-05, "loss": 2.9645, "step": 14512 }, { "epoch": 1.14, "learning_rate": 8.86199325648867e-05, "loss": 2.9041, "step": 14513 }, { "epoch": 1.14, "learning_rate": 8.86191484356622e-05, "loss": 2.6665, "step": 14514 }, { "epoch": 1.14, "learning_rate": 8.86183643064377e-05, "loss": 2.7896, "step": 14515 }, { "epoch": 1.14, "learning_rate": 8.86175801772132e-05, "loss": 2.6675, "step": 14516 }, { "epoch": 1.14, "learning_rate": 8.861679604798871e-05, "loss": 2.9619, "step": 14517 }, { "epoch": 1.14, "learning_rate": 8.861601191876421e-05, "loss": 3.0645, "step": 14518 }, { "epoch": 1.14, "learning_rate": 8.861522778953972e-05, "loss": 2.9657, "step": 14519 }, { "epoch": 1.14, "learning_rate": 8.861444366031522e-05, "loss": 2.8593, "step": 14520 }, { "epoch": 1.14, "learning_rate": 8.861365953109072e-05, "loss": 2.9641, "step": 14521 }, { "epoch": 1.14, "learning_rate": 8.861287540186624e-05, "loss": 2.8464, "step": 14522 }, { "epoch": 1.14, "learning_rate": 8.861209127264175e-05, "loss": 2.9985, "step": 14523 }, { "epoch": 1.14, "learning_rate": 8.861130714341725e-05, "loss": 2.7819, "step": 14524 }, { "epoch": 1.14, "learning_rate": 8.861052301419275e-05, "loss": 2.8977, "step": 14525 }, { "epoch": 1.14, "learning_rate": 8.860973888496826e-05, "loss": 2.8259, "step": 14526 }, { "epoch": 1.14, "learning_rate": 8.860895475574376e-05, "loss": 3.0675, "step": 14527 }, { "epoch": 1.14, "learning_rate": 8.860817062651926e-05, "loss": 2.9298, "step": 14528 }, { "epoch": 1.14, "learning_rate": 8.860738649729475e-05, "loss": 2.8404, "step": 14529 }, { "epoch": 1.14, "learning_rate": 8.860660236807026e-05, "loss": 2.9582, "step": 14530 }, { "epoch": 1.14, "learning_rate": 8.860581823884576e-05, "loss": 2.7294, "step": 14531 }, { "epoch": 1.14, "learning_rate": 8.860503410962127e-05, "loss": 2.8182, "step": 14532 }, { "epoch": 1.14, "learning_rate": 8.860424998039677e-05, "loss": 2.8158, "step": 14533 }, { "epoch": 1.14, "learning_rate": 8.860346585117227e-05, "loss": 2.8592, "step": 14534 }, { "epoch": 1.14, "learning_rate": 8.860268172194778e-05, "loss": 2.7644, "step": 14535 }, { "epoch": 1.14, "learning_rate": 8.860189759272328e-05, "loss": 2.6749, "step": 14536 }, { "epoch": 1.14, "learning_rate": 8.860111346349878e-05, "loss": 3.0439, "step": 14537 }, { "epoch": 1.14, "learning_rate": 8.860032933427429e-05, "loss": 2.9281, "step": 14538 }, { "epoch": 1.14, "learning_rate": 8.859954520504979e-05, "loss": 3.0751, "step": 14539 }, { "epoch": 1.14, "learning_rate": 8.85987610758253e-05, "loss": 2.918, "step": 14540 }, { "epoch": 1.14, "learning_rate": 8.85979769466008e-05, "loss": 3.0168, "step": 14541 }, { "epoch": 1.14, "learning_rate": 8.85971928173763e-05, "loss": 2.8142, "step": 14542 }, { "epoch": 1.14, "learning_rate": 8.85964086881518e-05, "loss": 2.9352, "step": 14543 }, { "epoch": 1.14, "learning_rate": 8.859562455892731e-05, "loss": 2.8131, "step": 14544 }, { "epoch": 1.14, "learning_rate": 8.859484042970281e-05, "loss": 2.8573, "step": 14545 }, { "epoch": 1.14, "learning_rate": 8.859405630047832e-05, "loss": 2.7634, "step": 14546 }, { "epoch": 1.14, "learning_rate": 8.859327217125383e-05, "loss": 2.9157, "step": 14547 }, { "epoch": 1.14, "learning_rate": 8.859248804202934e-05, "loss": 2.8775, "step": 14548 }, { "epoch": 1.14, "learning_rate": 8.859170391280484e-05, "loss": 2.927, "step": 14549 }, { "epoch": 1.14, "learning_rate": 8.859091978358034e-05, "loss": 2.9318, "step": 14550 }, { "epoch": 1.14, "learning_rate": 8.859013565435585e-05, "loss": 2.986, "step": 14551 }, { "epoch": 1.14, "learning_rate": 8.858935152513135e-05, "loss": 2.9464, "step": 14552 }, { "epoch": 1.14, "learning_rate": 8.858856739590686e-05, "loss": 2.8578, "step": 14553 }, { "epoch": 1.14, "learning_rate": 8.858778326668236e-05, "loss": 3.1261, "step": 14554 }, { "epoch": 1.14, "learning_rate": 8.858699913745786e-05, "loss": 2.9631, "step": 14555 }, { "epoch": 1.14, "learning_rate": 8.858621500823337e-05, "loss": 2.9622, "step": 14556 }, { "epoch": 1.14, "learning_rate": 8.858543087900887e-05, "loss": 3.0179, "step": 14557 }, { "epoch": 1.14, "learning_rate": 8.858464674978437e-05, "loss": 3.027, "step": 14558 }, { "epoch": 1.14, "learning_rate": 8.858386262055988e-05, "loss": 2.8284, "step": 14559 }, { "epoch": 1.14, "learning_rate": 8.858307849133538e-05, "loss": 2.7905, "step": 14560 }, { "epoch": 1.14, "learning_rate": 8.858229436211089e-05, "loss": 2.8638, "step": 14561 }, { "epoch": 1.14, "learning_rate": 8.858151023288639e-05, "loss": 3.0735, "step": 14562 }, { "epoch": 1.14, "learning_rate": 8.858072610366189e-05, "loss": 2.7919, "step": 14563 }, { "epoch": 1.14, "learning_rate": 8.85799419744374e-05, "loss": 3.0986, "step": 14564 }, { "epoch": 1.14, "learning_rate": 8.857915784521289e-05, "loss": 2.8168, "step": 14565 }, { "epoch": 1.14, "learning_rate": 8.857837371598839e-05, "loss": 3.0177, "step": 14566 }, { "epoch": 1.14, "learning_rate": 8.85775895867639e-05, "loss": 3.2133, "step": 14567 }, { "epoch": 1.14, "learning_rate": 8.85768054575394e-05, "loss": 2.6726, "step": 14568 }, { "epoch": 1.14, "learning_rate": 8.85760213283149e-05, "loss": 2.7756, "step": 14569 }, { "epoch": 1.14, "learning_rate": 8.85752371990904e-05, "loss": 3.0112, "step": 14570 }, { "epoch": 1.14, "learning_rate": 8.857445306986591e-05, "loss": 2.7112, "step": 14571 }, { "epoch": 1.14, "learning_rate": 8.857366894064141e-05, "loss": 2.6522, "step": 14572 }, { "epoch": 1.14, "learning_rate": 8.857288481141693e-05, "loss": 2.9738, "step": 14573 }, { "epoch": 1.14, "learning_rate": 8.857210068219243e-05, "loss": 2.7593, "step": 14574 }, { "epoch": 1.14, "learning_rate": 8.857131655296794e-05, "loss": 2.9067, "step": 14575 }, { "epoch": 1.14, "learning_rate": 8.857053242374344e-05, "loss": 2.9523, "step": 14576 }, { "epoch": 1.14, "learning_rate": 8.856974829451894e-05, "loss": 3.0217, "step": 14577 }, { "epoch": 1.14, "learning_rate": 8.856896416529445e-05, "loss": 2.9243, "step": 14578 }, { "epoch": 1.14, "learning_rate": 8.856818003606995e-05, "loss": 2.9506, "step": 14579 }, { "epoch": 1.14, "learning_rate": 8.856739590684546e-05, "loss": 3.0057, "step": 14580 }, { "epoch": 1.14, "learning_rate": 8.856661177762096e-05, "loss": 2.9121, "step": 14581 }, { "epoch": 1.14, "learning_rate": 8.856582764839646e-05, "loss": 2.8627, "step": 14582 }, { "epoch": 1.14, "learning_rate": 8.856504351917197e-05, "loss": 2.7912, "step": 14583 }, { "epoch": 1.14, "learning_rate": 8.856425938994747e-05, "loss": 2.9826, "step": 14584 }, { "epoch": 1.14, "learning_rate": 8.856347526072297e-05, "loss": 2.9017, "step": 14585 }, { "epoch": 1.14, "learning_rate": 8.856269113149848e-05, "loss": 2.8592, "step": 14586 }, { "epoch": 1.14, "learning_rate": 8.856190700227398e-05, "loss": 3.116, "step": 14587 }, { "epoch": 1.14, "learning_rate": 8.856112287304948e-05, "loss": 3.1086, "step": 14588 }, { "epoch": 1.14, "learning_rate": 8.856033874382499e-05, "loss": 2.9801, "step": 14589 }, { "epoch": 1.14, "learning_rate": 8.855955461460049e-05, "loss": 3.1907, "step": 14590 }, { "epoch": 1.14, "learning_rate": 8.8558770485376e-05, "loss": 2.9865, "step": 14591 }, { "epoch": 1.14, "learning_rate": 8.85579863561515e-05, "loss": 2.8571, "step": 14592 }, { "epoch": 1.14, "learning_rate": 8.8557202226927e-05, "loss": 3.0399, "step": 14593 }, { "epoch": 1.14, "learning_rate": 8.85564180977025e-05, "loss": 3.1707, "step": 14594 }, { "epoch": 1.14, "learning_rate": 8.855563396847801e-05, "loss": 3.0951, "step": 14595 }, { "epoch": 1.14, "learning_rate": 8.855484983925351e-05, "loss": 2.7628, "step": 14596 }, { "epoch": 1.14, "learning_rate": 8.855406571002902e-05, "loss": 3.0739, "step": 14597 }, { "epoch": 1.14, "learning_rate": 8.855328158080452e-05, "loss": 3.1512, "step": 14598 }, { "epoch": 1.14, "learning_rate": 8.855249745158003e-05, "loss": 2.6845, "step": 14599 }, { "epoch": 1.14, "learning_rate": 8.855171332235553e-05, "loss": 3.1293, "step": 14600 }, { "epoch": 1.14, "learning_rate": 8.855092919313103e-05, "loss": 2.9955, "step": 14601 }, { "epoch": 1.14, "learning_rate": 8.855014506390654e-05, "loss": 2.8448, "step": 14602 }, { "epoch": 1.15, "learning_rate": 8.854936093468204e-05, "loss": 2.8583, "step": 14603 }, { "epoch": 1.15, "learning_rate": 8.854857680545754e-05, "loss": 2.8208, "step": 14604 }, { "epoch": 1.15, "learning_rate": 8.854779267623305e-05, "loss": 2.8975, "step": 14605 }, { "epoch": 1.15, "learning_rate": 8.854700854700855e-05, "loss": 2.7634, "step": 14606 }, { "epoch": 1.15, "learning_rate": 8.854622441778405e-05, "loss": 2.8817, "step": 14607 }, { "epoch": 1.15, "learning_rate": 8.854544028855956e-05, "loss": 2.9212, "step": 14608 }, { "epoch": 1.15, "learning_rate": 8.854465615933506e-05, "loss": 2.8641, "step": 14609 }, { "epoch": 1.15, "learning_rate": 8.854387203011057e-05, "loss": 2.7308, "step": 14610 }, { "epoch": 1.15, "learning_rate": 8.854308790088607e-05, "loss": 3.0549, "step": 14611 }, { "epoch": 1.15, "learning_rate": 8.854230377166157e-05, "loss": 2.7386, "step": 14612 }, { "epoch": 1.15, "learning_rate": 8.854151964243708e-05, "loss": 2.881, "step": 14613 }, { "epoch": 1.15, "learning_rate": 8.854073551321258e-05, "loss": 2.8196, "step": 14614 }, { "epoch": 1.15, "learning_rate": 8.853995138398808e-05, "loss": 2.8018, "step": 14615 }, { "epoch": 1.15, "learning_rate": 8.853916725476359e-05, "loss": 2.9706, "step": 14616 }, { "epoch": 1.15, "learning_rate": 8.853838312553909e-05, "loss": 2.8319, "step": 14617 }, { "epoch": 1.15, "learning_rate": 8.85375989963146e-05, "loss": 2.8337, "step": 14618 }, { "epoch": 1.15, "learning_rate": 8.85368148670901e-05, "loss": 3.1345, "step": 14619 }, { "epoch": 1.15, "learning_rate": 8.85360307378656e-05, "loss": 2.7806, "step": 14620 }, { "epoch": 1.15, "learning_rate": 8.85352466086411e-05, "loss": 2.8415, "step": 14621 }, { "epoch": 1.15, "learning_rate": 8.853446247941661e-05, "loss": 2.8165, "step": 14622 }, { "epoch": 1.15, "learning_rate": 8.853367835019211e-05, "loss": 3.0167, "step": 14623 }, { "epoch": 1.15, "learning_rate": 8.853289422096762e-05, "loss": 2.897, "step": 14624 }, { "epoch": 1.15, "learning_rate": 8.853211009174312e-05, "loss": 2.6999, "step": 14625 }, { "epoch": 1.15, "learning_rate": 8.853132596251862e-05, "loss": 3.2862, "step": 14626 }, { "epoch": 1.15, "learning_rate": 8.853054183329413e-05, "loss": 2.5716, "step": 14627 }, { "epoch": 1.15, "learning_rate": 8.852975770406963e-05, "loss": 2.8472, "step": 14628 }, { "epoch": 1.15, "learning_rate": 8.852897357484514e-05, "loss": 2.837, "step": 14629 }, { "epoch": 1.15, "learning_rate": 8.852818944562065e-05, "loss": 3.0229, "step": 14630 }, { "epoch": 1.15, "learning_rate": 8.852740531639616e-05, "loss": 2.9571, "step": 14631 }, { "epoch": 1.15, "learning_rate": 8.852662118717166e-05, "loss": 2.7665, "step": 14632 }, { "epoch": 1.15, "learning_rate": 8.852583705794716e-05, "loss": 2.8422, "step": 14633 }, { "epoch": 1.15, "learning_rate": 8.852505292872267e-05, "loss": 2.7492, "step": 14634 }, { "epoch": 1.15, "learning_rate": 8.852426879949817e-05, "loss": 2.697, "step": 14635 }, { "epoch": 1.15, "learning_rate": 8.852348467027367e-05, "loss": 2.962, "step": 14636 }, { "epoch": 1.15, "learning_rate": 8.852270054104916e-05, "loss": 2.8201, "step": 14637 }, { "epoch": 1.15, "learning_rate": 8.852191641182467e-05, "loss": 2.873, "step": 14638 }, { "epoch": 1.15, "learning_rate": 8.852113228260017e-05, "loss": 2.9502, "step": 14639 }, { "epoch": 1.15, "learning_rate": 8.852034815337568e-05, "loss": 2.8489, "step": 14640 }, { "epoch": 1.15, "learning_rate": 8.851956402415118e-05, "loss": 2.9575, "step": 14641 }, { "epoch": 1.15, "learning_rate": 8.851877989492668e-05, "loss": 2.8327, "step": 14642 }, { "epoch": 1.15, "learning_rate": 8.851799576570219e-05, "loss": 2.9707, "step": 14643 }, { "epoch": 1.15, "learning_rate": 8.851721163647769e-05, "loss": 2.8216, "step": 14644 }, { "epoch": 1.15, "learning_rate": 8.85164275072532e-05, "loss": 2.9186, "step": 14645 }, { "epoch": 1.15, "learning_rate": 8.85156433780287e-05, "loss": 3.1842, "step": 14646 }, { "epoch": 1.15, "learning_rate": 8.85148592488042e-05, "loss": 2.8829, "step": 14647 }, { "epoch": 1.15, "learning_rate": 8.85140751195797e-05, "loss": 2.9926, "step": 14648 }, { "epoch": 1.15, "learning_rate": 8.851329099035521e-05, "loss": 2.8327, "step": 14649 }, { "epoch": 1.15, "learning_rate": 8.851250686113071e-05, "loss": 3.0817, "step": 14650 }, { "epoch": 1.15, "learning_rate": 8.851172273190622e-05, "loss": 2.8549, "step": 14651 }, { "epoch": 1.15, "learning_rate": 8.851093860268172e-05, "loss": 2.8019, "step": 14652 }, { "epoch": 1.15, "learning_rate": 8.851015447345722e-05, "loss": 2.9254, "step": 14653 }, { "epoch": 1.15, "learning_rate": 8.850937034423273e-05, "loss": 2.6375, "step": 14654 }, { "epoch": 1.15, "learning_rate": 8.850858621500824e-05, "loss": 2.9777, "step": 14655 }, { "epoch": 1.15, "learning_rate": 8.850780208578375e-05, "loss": 2.7498, "step": 14656 }, { "epoch": 1.15, "learning_rate": 8.850701795655925e-05, "loss": 2.9757, "step": 14657 }, { "epoch": 1.15, "learning_rate": 8.850623382733476e-05, "loss": 3.038, "step": 14658 }, { "epoch": 1.15, "learning_rate": 8.850544969811026e-05, "loss": 2.7096, "step": 14659 }, { "epoch": 1.15, "learning_rate": 8.850466556888576e-05, "loss": 2.8809, "step": 14660 }, { "epoch": 1.15, "learning_rate": 8.850388143966127e-05, "loss": 3.0474, "step": 14661 }, { "epoch": 1.15, "learning_rate": 8.850309731043677e-05, "loss": 3.0217, "step": 14662 }, { "epoch": 1.15, "learning_rate": 8.850231318121227e-05, "loss": 3.112, "step": 14663 }, { "epoch": 1.15, "learning_rate": 8.850152905198778e-05, "loss": 2.8459, "step": 14664 }, { "epoch": 1.15, "learning_rate": 8.850074492276328e-05, "loss": 2.8664, "step": 14665 }, { "epoch": 1.15, "learning_rate": 8.849996079353879e-05, "loss": 2.9448, "step": 14666 }, { "epoch": 1.15, "learning_rate": 8.849917666431429e-05, "loss": 2.9715, "step": 14667 }, { "epoch": 1.15, "learning_rate": 8.849839253508979e-05, "loss": 2.7335, "step": 14668 }, { "epoch": 1.15, "learning_rate": 8.84976084058653e-05, "loss": 2.9333, "step": 14669 }, { "epoch": 1.15, "learning_rate": 8.84968242766408e-05, "loss": 2.7041, "step": 14670 }, { "epoch": 1.15, "learning_rate": 8.84960401474163e-05, "loss": 3.0009, "step": 14671 }, { "epoch": 1.15, "learning_rate": 8.849525601819181e-05, "loss": 2.8874, "step": 14672 }, { "epoch": 1.15, "learning_rate": 8.84944718889673e-05, "loss": 2.8637, "step": 14673 }, { "epoch": 1.15, "learning_rate": 8.84936877597428e-05, "loss": 2.9361, "step": 14674 }, { "epoch": 1.15, "learning_rate": 8.84929036305183e-05, "loss": 3.039, "step": 14675 }, { "epoch": 1.15, "learning_rate": 8.849211950129381e-05, "loss": 2.9139, "step": 14676 }, { "epoch": 1.15, "learning_rate": 8.849133537206931e-05, "loss": 2.9969, "step": 14677 }, { "epoch": 1.15, "learning_rate": 8.849055124284482e-05, "loss": 2.9389, "step": 14678 }, { "epoch": 1.15, "learning_rate": 8.848976711362032e-05, "loss": 3.1209, "step": 14679 }, { "epoch": 1.15, "learning_rate": 8.848898298439582e-05, "loss": 2.9238, "step": 14680 }, { "epoch": 1.15, "learning_rate": 8.848819885517134e-05, "loss": 2.7916, "step": 14681 }, { "epoch": 1.15, "learning_rate": 8.848741472594684e-05, "loss": 2.7745, "step": 14682 }, { "epoch": 1.15, "learning_rate": 8.848663059672235e-05, "loss": 3.3496, "step": 14683 }, { "epoch": 1.15, "learning_rate": 8.848584646749785e-05, "loss": 2.7244, "step": 14684 }, { "epoch": 1.15, "learning_rate": 8.848506233827335e-05, "loss": 2.9284, "step": 14685 }, { "epoch": 1.15, "learning_rate": 8.848427820904886e-05, "loss": 2.6637, "step": 14686 }, { "epoch": 1.15, "learning_rate": 8.848349407982436e-05, "loss": 3.1466, "step": 14687 }, { "epoch": 1.15, "learning_rate": 8.848270995059987e-05, "loss": 3.2192, "step": 14688 }, { "epoch": 1.15, "learning_rate": 8.848192582137537e-05, "loss": 2.6714, "step": 14689 }, { "epoch": 1.15, "learning_rate": 8.848114169215087e-05, "loss": 2.9718, "step": 14690 }, { "epoch": 1.15, "learning_rate": 8.848035756292638e-05, "loss": 2.8597, "step": 14691 }, { "epoch": 1.15, "learning_rate": 8.847957343370188e-05, "loss": 2.9444, "step": 14692 }, { "epoch": 1.15, "learning_rate": 8.847878930447738e-05, "loss": 2.8476, "step": 14693 }, { "epoch": 1.15, "learning_rate": 8.847800517525289e-05, "loss": 2.8389, "step": 14694 }, { "epoch": 1.15, "learning_rate": 8.847722104602839e-05, "loss": 2.8384, "step": 14695 }, { "epoch": 1.15, "learning_rate": 8.84764369168039e-05, "loss": 2.8013, "step": 14696 }, { "epoch": 1.15, "learning_rate": 8.84756527875794e-05, "loss": 2.677, "step": 14697 }, { "epoch": 1.15, "learning_rate": 8.84748686583549e-05, "loss": 2.8172, "step": 14698 }, { "epoch": 1.15, "learning_rate": 8.84740845291304e-05, "loss": 3.1136, "step": 14699 }, { "epoch": 1.15, "learning_rate": 8.847330039990591e-05, "loss": 3.0264, "step": 14700 }, { "epoch": 1.15, "learning_rate": 8.847251627068141e-05, "loss": 2.7597, "step": 14701 }, { "epoch": 1.15, "learning_rate": 8.847173214145692e-05, "loss": 2.7624, "step": 14702 }, { "epoch": 1.15, "learning_rate": 8.847094801223242e-05, "loss": 3.1904, "step": 14703 }, { "epoch": 1.15, "learning_rate": 8.847016388300792e-05, "loss": 3.01, "step": 14704 }, { "epoch": 1.15, "learning_rate": 8.846937975378343e-05, "loss": 2.9312, "step": 14705 }, { "epoch": 1.15, "learning_rate": 8.846859562455893e-05, "loss": 3.2439, "step": 14706 }, { "epoch": 1.15, "learning_rate": 8.846781149533444e-05, "loss": 2.8911, "step": 14707 }, { "epoch": 1.15, "learning_rate": 8.846702736610994e-05, "loss": 3.0967, "step": 14708 }, { "epoch": 1.15, "learning_rate": 8.846624323688544e-05, "loss": 3.0844, "step": 14709 }, { "epoch": 1.15, "learning_rate": 8.846545910766095e-05, "loss": 2.7736, "step": 14710 }, { "epoch": 1.15, "learning_rate": 8.846467497843645e-05, "loss": 2.9279, "step": 14711 }, { "epoch": 1.15, "learning_rate": 8.846389084921195e-05, "loss": 2.8619, "step": 14712 }, { "epoch": 1.15, "learning_rate": 8.846310671998746e-05, "loss": 3.0387, "step": 14713 }, { "epoch": 1.15, "learning_rate": 8.846232259076296e-05, "loss": 2.8355, "step": 14714 }, { "epoch": 1.15, "learning_rate": 8.846153846153847e-05, "loss": 2.9199, "step": 14715 }, { "epoch": 1.15, "learning_rate": 8.846075433231397e-05, "loss": 2.9929, "step": 14716 }, { "epoch": 1.15, "learning_rate": 8.845997020308947e-05, "loss": 2.6598, "step": 14717 }, { "epoch": 1.15, "learning_rate": 8.845918607386498e-05, "loss": 3.0943, "step": 14718 }, { "epoch": 1.15, "learning_rate": 8.845840194464048e-05, "loss": 3.0928, "step": 14719 }, { "epoch": 1.15, "learning_rate": 8.845761781541598e-05, "loss": 2.8923, "step": 14720 }, { "epoch": 1.15, "learning_rate": 8.845683368619149e-05, "loss": 2.822, "step": 14721 }, { "epoch": 1.15, "learning_rate": 8.845604955696699e-05, "loss": 2.9314, "step": 14722 }, { "epoch": 1.15, "learning_rate": 8.84552654277425e-05, "loss": 2.8677, "step": 14723 }, { "epoch": 1.15, "learning_rate": 8.8454481298518e-05, "loss": 2.7233, "step": 14724 }, { "epoch": 1.15, "learning_rate": 8.84536971692935e-05, "loss": 2.7283, "step": 14725 }, { "epoch": 1.15, "learning_rate": 8.8452913040069e-05, "loss": 2.7063, "step": 14726 }, { "epoch": 1.15, "learning_rate": 8.845212891084451e-05, "loss": 3.0211, "step": 14727 }, { "epoch": 1.15, "learning_rate": 8.845134478162001e-05, "loss": 2.8285, "step": 14728 }, { "epoch": 1.15, "learning_rate": 8.845056065239552e-05, "loss": 2.97, "step": 14729 }, { "epoch": 1.16, "learning_rate": 8.844977652317102e-05, "loss": 2.9288, "step": 14730 }, { "epoch": 1.16, "learning_rate": 8.844899239394652e-05, "loss": 2.9957, "step": 14731 }, { "epoch": 1.16, "learning_rate": 8.844820826472203e-05, "loss": 2.9745, "step": 14732 }, { "epoch": 1.16, "learning_rate": 8.844742413549753e-05, "loss": 2.9039, "step": 14733 }, { "epoch": 1.16, "learning_rate": 8.844664000627304e-05, "loss": 2.8936, "step": 14734 }, { "epoch": 1.16, "learning_rate": 8.844585587704854e-05, "loss": 3.0755, "step": 14735 }, { "epoch": 1.16, "learning_rate": 8.844507174782404e-05, "loss": 2.7509, "step": 14736 }, { "epoch": 1.16, "learning_rate": 8.844428761859955e-05, "loss": 2.9431, "step": 14737 }, { "epoch": 1.16, "learning_rate": 8.844350348937506e-05, "loss": 3.0992, "step": 14738 }, { "epoch": 1.16, "learning_rate": 8.844271936015057e-05, "loss": 2.8908, "step": 14739 }, { "epoch": 1.16, "learning_rate": 8.844193523092607e-05, "loss": 2.874, "step": 14740 }, { "epoch": 1.16, "learning_rate": 8.844115110170157e-05, "loss": 2.939, "step": 14741 }, { "epoch": 1.16, "learning_rate": 8.844036697247708e-05, "loss": 2.9383, "step": 14742 }, { "epoch": 1.16, "learning_rate": 8.843958284325258e-05, "loss": 3.025, "step": 14743 }, { "epoch": 1.16, "learning_rate": 8.843879871402809e-05, "loss": 2.8738, "step": 14744 }, { "epoch": 1.16, "learning_rate": 8.843801458480358e-05, "loss": 3.0256, "step": 14745 }, { "epoch": 1.16, "learning_rate": 8.843723045557908e-05, "loss": 3.1138, "step": 14746 }, { "epoch": 1.16, "learning_rate": 8.843644632635458e-05, "loss": 2.7543, "step": 14747 }, { "epoch": 1.16, "learning_rate": 8.843566219713009e-05, "loss": 2.9089, "step": 14748 }, { "epoch": 1.16, "learning_rate": 8.843487806790559e-05, "loss": 3.0178, "step": 14749 }, { "epoch": 1.16, "learning_rate": 8.84340939386811e-05, "loss": 2.8847, "step": 14750 }, { "epoch": 1.16, "learning_rate": 8.84333098094566e-05, "loss": 2.9812, "step": 14751 }, { "epoch": 1.16, "learning_rate": 8.84325256802321e-05, "loss": 3.0402, "step": 14752 }, { "epoch": 1.16, "learning_rate": 8.84317415510076e-05, "loss": 2.9012, "step": 14753 }, { "epoch": 1.16, "learning_rate": 8.843095742178311e-05, "loss": 2.7416, "step": 14754 }, { "epoch": 1.16, "learning_rate": 8.843017329255861e-05, "loss": 2.941, "step": 14755 }, { "epoch": 1.16, "learning_rate": 8.842938916333412e-05, "loss": 2.7809, "step": 14756 }, { "epoch": 1.16, "learning_rate": 8.842860503410962e-05, "loss": 2.9203, "step": 14757 }, { "epoch": 1.16, "learning_rate": 8.842782090488512e-05, "loss": 2.8694, "step": 14758 }, { "epoch": 1.16, "learning_rate": 8.842703677566063e-05, "loss": 2.7364, "step": 14759 }, { "epoch": 1.16, "learning_rate": 8.842625264643613e-05, "loss": 2.8665, "step": 14760 }, { "epoch": 1.16, "learning_rate": 8.842546851721163e-05, "loss": 3.1068, "step": 14761 }, { "epoch": 1.16, "learning_rate": 8.842468438798714e-05, "loss": 3.1846, "step": 14762 }, { "epoch": 1.16, "learning_rate": 8.842390025876266e-05, "loss": 3.0692, "step": 14763 }, { "epoch": 1.16, "learning_rate": 8.842311612953816e-05, "loss": 3.1584, "step": 14764 }, { "epoch": 1.16, "learning_rate": 8.842233200031366e-05, "loss": 2.9167, "step": 14765 }, { "epoch": 1.16, "learning_rate": 8.842154787108917e-05, "loss": 3.2205, "step": 14766 }, { "epoch": 1.16, "learning_rate": 8.842076374186467e-05, "loss": 2.8561, "step": 14767 }, { "epoch": 1.16, "learning_rate": 8.841997961264017e-05, "loss": 3.0813, "step": 14768 }, { "epoch": 1.16, "learning_rate": 8.841919548341568e-05, "loss": 2.8762, "step": 14769 }, { "epoch": 1.16, "learning_rate": 8.841841135419118e-05, "loss": 2.7246, "step": 14770 }, { "epoch": 1.16, "learning_rate": 8.841762722496668e-05, "loss": 2.8066, "step": 14771 }, { "epoch": 1.16, "learning_rate": 8.841684309574219e-05, "loss": 2.8764, "step": 14772 }, { "epoch": 1.16, "learning_rate": 8.841605896651769e-05, "loss": 2.8222, "step": 14773 }, { "epoch": 1.16, "learning_rate": 8.84152748372932e-05, "loss": 3.0389, "step": 14774 }, { "epoch": 1.16, "learning_rate": 8.84144907080687e-05, "loss": 3.0047, "step": 14775 }, { "epoch": 1.16, "learning_rate": 8.84137065788442e-05, "loss": 2.9191, "step": 14776 }, { "epoch": 1.16, "learning_rate": 8.841292244961971e-05, "loss": 2.9613, "step": 14777 }, { "epoch": 1.16, "learning_rate": 8.841213832039521e-05, "loss": 2.9132, "step": 14778 }, { "epoch": 1.16, "learning_rate": 8.841135419117071e-05, "loss": 2.8775, "step": 14779 }, { "epoch": 1.16, "learning_rate": 8.841057006194622e-05, "loss": 3.0915, "step": 14780 }, { "epoch": 1.16, "learning_rate": 8.840978593272171e-05, "loss": 2.5921, "step": 14781 }, { "epoch": 1.16, "learning_rate": 8.840900180349721e-05, "loss": 2.9604, "step": 14782 }, { "epoch": 1.16, "learning_rate": 8.840821767427272e-05, "loss": 3.0991, "step": 14783 }, { "epoch": 1.16, "learning_rate": 8.840743354504822e-05, "loss": 2.8869, "step": 14784 }, { "epoch": 1.16, "learning_rate": 8.840664941582372e-05, "loss": 2.947, "step": 14785 }, { "epoch": 1.16, "learning_rate": 8.840586528659923e-05, "loss": 2.859, "step": 14786 }, { "epoch": 1.16, "learning_rate": 8.840508115737473e-05, "loss": 2.921, "step": 14787 }, { "epoch": 1.16, "learning_rate": 8.840429702815023e-05, "loss": 2.9282, "step": 14788 }, { "epoch": 1.16, "learning_rate": 8.840351289892575e-05, "loss": 2.9833, "step": 14789 }, { "epoch": 1.16, "learning_rate": 8.840272876970125e-05, "loss": 3.0041, "step": 14790 }, { "epoch": 1.16, "learning_rate": 8.840194464047676e-05, "loss": 3.0674, "step": 14791 }, { "epoch": 1.16, "learning_rate": 8.840116051125226e-05, "loss": 2.7146, "step": 14792 }, { "epoch": 1.16, "learning_rate": 8.840037638202777e-05, "loss": 2.8522, "step": 14793 }, { "epoch": 1.16, "learning_rate": 8.839959225280327e-05, "loss": 2.7698, "step": 14794 }, { "epoch": 1.16, "learning_rate": 8.839880812357877e-05, "loss": 2.7581, "step": 14795 }, { "epoch": 1.16, "learning_rate": 8.839802399435428e-05, "loss": 3.0336, "step": 14796 }, { "epoch": 1.16, "learning_rate": 8.839723986512978e-05, "loss": 2.8121, "step": 14797 }, { "epoch": 1.16, "learning_rate": 8.839645573590528e-05, "loss": 2.866, "step": 14798 }, { "epoch": 1.16, "learning_rate": 8.839567160668079e-05, "loss": 2.651, "step": 14799 }, { "epoch": 1.16, "learning_rate": 8.839488747745629e-05, "loss": 2.8593, "step": 14800 }, { "epoch": 1.16, "learning_rate": 8.83941033482318e-05, "loss": 2.8182, "step": 14801 }, { "epoch": 1.16, "learning_rate": 8.83933192190073e-05, "loss": 2.8611, "step": 14802 }, { "epoch": 1.16, "learning_rate": 8.83925350897828e-05, "loss": 2.8115, "step": 14803 }, { "epoch": 1.16, "learning_rate": 8.83917509605583e-05, "loss": 2.907, "step": 14804 }, { "epoch": 1.16, "learning_rate": 8.839096683133381e-05, "loss": 3.0109, "step": 14805 }, { "epoch": 1.16, "learning_rate": 8.839018270210931e-05, "loss": 2.7343, "step": 14806 }, { "epoch": 1.16, "learning_rate": 8.838939857288482e-05, "loss": 2.6067, "step": 14807 }, { "epoch": 1.16, "learning_rate": 8.838861444366032e-05, "loss": 2.7373, "step": 14808 }, { "epoch": 1.16, "learning_rate": 8.838783031443582e-05, "loss": 2.8136, "step": 14809 }, { "epoch": 1.16, "learning_rate": 8.838704618521133e-05, "loss": 2.9172, "step": 14810 }, { "epoch": 1.16, "learning_rate": 8.838626205598683e-05, "loss": 2.9762, "step": 14811 }, { "epoch": 1.16, "learning_rate": 8.838547792676234e-05, "loss": 3.0373, "step": 14812 }, { "epoch": 1.16, "learning_rate": 8.838469379753784e-05, "loss": 2.9622, "step": 14813 }, { "epoch": 1.16, "learning_rate": 8.838390966831334e-05, "loss": 2.9907, "step": 14814 }, { "epoch": 1.16, "learning_rate": 8.838312553908885e-05, "loss": 2.8461, "step": 14815 }, { "epoch": 1.16, "learning_rate": 8.838234140986435e-05, "loss": 2.9154, "step": 14816 }, { "epoch": 1.16, "learning_rate": 8.838155728063985e-05, "loss": 3.0178, "step": 14817 }, { "epoch": 1.16, "learning_rate": 8.838077315141536e-05, "loss": 3.3009, "step": 14818 }, { "epoch": 1.16, "learning_rate": 8.837998902219086e-05, "loss": 3.2079, "step": 14819 }, { "epoch": 1.16, "learning_rate": 8.837920489296636e-05, "loss": 3.2315, "step": 14820 }, { "epoch": 1.16, "learning_rate": 8.837842076374187e-05, "loss": 3.0405, "step": 14821 }, { "epoch": 1.16, "learning_rate": 8.837763663451737e-05, "loss": 3.0856, "step": 14822 }, { "epoch": 1.16, "learning_rate": 8.837685250529288e-05, "loss": 2.7335, "step": 14823 }, { "epoch": 1.16, "learning_rate": 8.837606837606838e-05, "loss": 3.1135, "step": 14824 }, { "epoch": 1.16, "learning_rate": 8.837528424684388e-05, "loss": 2.9392, "step": 14825 }, { "epoch": 1.16, "learning_rate": 8.837450011761939e-05, "loss": 2.9699, "step": 14826 }, { "epoch": 1.16, "learning_rate": 8.837371598839489e-05, "loss": 3.0772, "step": 14827 }, { "epoch": 1.16, "learning_rate": 8.83729318591704e-05, "loss": 2.818, "step": 14828 }, { "epoch": 1.16, "learning_rate": 8.83721477299459e-05, "loss": 2.8463, "step": 14829 }, { "epoch": 1.16, "learning_rate": 8.83713636007214e-05, "loss": 3.1828, "step": 14830 }, { "epoch": 1.16, "learning_rate": 8.83705794714969e-05, "loss": 2.9743, "step": 14831 }, { "epoch": 1.16, "learning_rate": 8.836979534227241e-05, "loss": 2.7249, "step": 14832 }, { "epoch": 1.16, "learning_rate": 8.836901121304791e-05, "loss": 2.9005, "step": 14833 }, { "epoch": 1.16, "learning_rate": 8.836822708382342e-05, "loss": 2.8644, "step": 14834 }, { "epoch": 1.16, "learning_rate": 8.836744295459892e-05, "loss": 2.9419, "step": 14835 }, { "epoch": 1.16, "learning_rate": 8.836665882537442e-05, "loss": 2.6847, "step": 14836 }, { "epoch": 1.16, "learning_rate": 8.836587469614993e-05, "loss": 3.1038, "step": 14837 }, { "epoch": 1.16, "learning_rate": 8.836509056692543e-05, "loss": 3.012, "step": 14838 }, { "epoch": 1.16, "learning_rate": 8.836430643770093e-05, "loss": 2.9036, "step": 14839 }, { "epoch": 1.16, "learning_rate": 8.836352230847644e-05, "loss": 2.9949, "step": 14840 }, { "epoch": 1.16, "learning_rate": 8.836273817925194e-05, "loss": 2.9707, "step": 14841 }, { "epoch": 1.16, "learning_rate": 8.836195405002745e-05, "loss": 2.9522, "step": 14842 }, { "epoch": 1.16, "learning_rate": 8.836116992080295e-05, "loss": 2.8431, "step": 14843 }, { "epoch": 1.16, "learning_rate": 8.836038579157845e-05, "loss": 2.9868, "step": 14844 }, { "epoch": 1.16, "learning_rate": 8.835960166235396e-05, "loss": 2.8128, "step": 14845 }, { "epoch": 1.16, "learning_rate": 8.835881753312947e-05, "loss": 3.1406, "step": 14846 }, { "epoch": 1.16, "learning_rate": 8.835803340390498e-05, "loss": 3.0004, "step": 14847 }, { "epoch": 1.16, "learning_rate": 8.835724927468048e-05, "loss": 2.9496, "step": 14848 }, { "epoch": 1.16, "learning_rate": 8.835646514545599e-05, "loss": 3.0918, "step": 14849 }, { "epoch": 1.16, "learning_rate": 8.835568101623149e-05, "loss": 2.855, "step": 14850 }, { "epoch": 1.16, "learning_rate": 8.835489688700699e-05, "loss": 2.7685, "step": 14851 }, { "epoch": 1.16, "learning_rate": 8.83541127577825e-05, "loss": 2.6281, "step": 14852 }, { "epoch": 1.16, "learning_rate": 8.835332862855799e-05, "loss": 2.9099, "step": 14853 }, { "epoch": 1.16, "learning_rate": 8.835254449933349e-05, "loss": 2.8859, "step": 14854 }, { "epoch": 1.16, "learning_rate": 8.8351760370109e-05, "loss": 2.8263, "step": 14855 }, { "epoch": 1.16, "learning_rate": 8.83509762408845e-05, "loss": 2.9547, "step": 14856 }, { "epoch": 1.16, "learning_rate": 8.835019211166e-05, "loss": 3.0327, "step": 14857 }, { "epoch": 1.17, "learning_rate": 8.83494079824355e-05, "loss": 2.9365, "step": 14858 }, { "epoch": 1.17, "learning_rate": 8.834862385321101e-05, "loss": 2.7553, "step": 14859 }, { "epoch": 1.17, "learning_rate": 8.834783972398651e-05, "loss": 2.739, "step": 14860 }, { "epoch": 1.17, "learning_rate": 8.834705559476202e-05, "loss": 3.0397, "step": 14861 }, { "epoch": 1.17, "learning_rate": 8.834627146553752e-05, "loss": 3.0837, "step": 14862 }, { "epoch": 1.17, "learning_rate": 8.834548733631302e-05, "loss": 2.7878, "step": 14863 }, { "epoch": 1.17, "learning_rate": 8.834470320708853e-05, "loss": 2.9153, "step": 14864 }, { "epoch": 1.17, "learning_rate": 8.834391907786403e-05, "loss": 2.8054, "step": 14865 }, { "epoch": 1.17, "learning_rate": 8.834313494863953e-05, "loss": 2.8126, "step": 14866 }, { "epoch": 1.17, "learning_rate": 8.834235081941504e-05, "loss": 2.9825, "step": 14867 }, { "epoch": 1.17, "learning_rate": 8.834156669019054e-05, "loss": 2.9791, "step": 14868 }, { "epoch": 1.17, "learning_rate": 8.834078256096605e-05, "loss": 3.2583, "step": 14869 }, { "epoch": 1.17, "learning_rate": 8.833999843174155e-05, "loss": 2.8567, "step": 14870 }, { "epoch": 1.17, "learning_rate": 8.833921430251707e-05, "loss": 2.7691, "step": 14871 }, { "epoch": 1.17, "learning_rate": 8.833843017329257e-05, "loss": 2.8584, "step": 14872 }, { "epoch": 1.17, "learning_rate": 8.833764604406807e-05, "loss": 2.979, "step": 14873 }, { "epoch": 1.17, "learning_rate": 8.833686191484358e-05, "loss": 2.7478, "step": 14874 }, { "epoch": 1.17, "learning_rate": 8.833607778561908e-05, "loss": 2.7163, "step": 14875 }, { "epoch": 1.17, "learning_rate": 8.833529365639458e-05, "loss": 3.0474, "step": 14876 }, { "epoch": 1.17, "learning_rate": 8.833450952717009e-05, "loss": 2.8774, "step": 14877 }, { "epoch": 1.17, "learning_rate": 8.833372539794559e-05, "loss": 3.1465, "step": 14878 }, { "epoch": 1.17, "learning_rate": 8.83329412687211e-05, "loss": 2.9733, "step": 14879 }, { "epoch": 1.17, "learning_rate": 8.83321571394966e-05, "loss": 3.1275, "step": 14880 }, { "epoch": 1.17, "learning_rate": 8.83313730102721e-05, "loss": 2.9233, "step": 14881 }, { "epoch": 1.17, "learning_rate": 8.83305888810476e-05, "loss": 2.9652, "step": 14882 }, { "epoch": 1.17, "learning_rate": 8.832980475182311e-05, "loss": 2.8444, "step": 14883 }, { "epoch": 1.17, "learning_rate": 8.832902062259861e-05, "loss": 2.9537, "step": 14884 }, { "epoch": 1.17, "learning_rate": 8.832823649337412e-05, "loss": 2.7688, "step": 14885 }, { "epoch": 1.17, "learning_rate": 8.832745236414962e-05, "loss": 2.9943, "step": 14886 }, { "epoch": 1.17, "learning_rate": 8.832666823492512e-05, "loss": 2.9405, "step": 14887 }, { "epoch": 1.17, "learning_rate": 8.832588410570063e-05, "loss": 2.9595, "step": 14888 }, { "epoch": 1.17, "learning_rate": 8.832509997647612e-05, "loss": 3.0683, "step": 14889 }, { "epoch": 1.17, "learning_rate": 8.832431584725162e-05, "loss": 2.6988, "step": 14890 }, { "epoch": 1.17, "learning_rate": 8.832353171802713e-05, "loss": 2.7314, "step": 14891 }, { "epoch": 1.17, "learning_rate": 8.832274758880263e-05, "loss": 2.7001, "step": 14892 }, { "epoch": 1.17, "learning_rate": 8.832196345957813e-05, "loss": 2.9826, "step": 14893 }, { "epoch": 1.17, "learning_rate": 8.832117933035364e-05, "loss": 2.9099, "step": 14894 }, { "epoch": 1.17, "learning_rate": 8.832039520112914e-05, "loss": 3.1673, "step": 14895 }, { "epoch": 1.17, "learning_rate": 8.831961107190464e-05, "loss": 2.8472, "step": 14896 }, { "epoch": 1.17, "learning_rate": 8.831882694268016e-05, "loss": 2.8895, "step": 14897 }, { "epoch": 1.17, "learning_rate": 8.831804281345567e-05, "loss": 3.0382, "step": 14898 }, { "epoch": 1.17, "learning_rate": 8.831725868423117e-05, "loss": 2.9878, "step": 14899 }, { "epoch": 1.17, "learning_rate": 8.831647455500667e-05, "loss": 3.0935, "step": 14900 }, { "epoch": 1.17, "learning_rate": 8.831569042578218e-05, "loss": 2.9303, "step": 14901 }, { "epoch": 1.17, "learning_rate": 8.831490629655768e-05, "loss": 2.9891, "step": 14902 }, { "epoch": 1.17, "learning_rate": 8.831412216733318e-05, "loss": 2.9161, "step": 14903 }, { "epoch": 1.17, "learning_rate": 8.831333803810869e-05, "loss": 2.7866, "step": 14904 }, { "epoch": 1.17, "learning_rate": 8.831255390888419e-05, "loss": 3.2091, "step": 14905 }, { "epoch": 1.17, "learning_rate": 8.83117697796597e-05, "loss": 2.795, "step": 14906 }, { "epoch": 1.17, "learning_rate": 8.83109856504352e-05, "loss": 2.9763, "step": 14907 }, { "epoch": 1.17, "learning_rate": 8.83102015212107e-05, "loss": 2.8096, "step": 14908 }, { "epoch": 1.17, "learning_rate": 8.83094173919862e-05, "loss": 2.9343, "step": 14909 }, { "epoch": 1.17, "learning_rate": 8.830863326276171e-05, "loss": 2.8084, "step": 14910 }, { "epoch": 1.17, "learning_rate": 8.830784913353721e-05, "loss": 3.171, "step": 14911 }, { "epoch": 1.17, "learning_rate": 8.830706500431272e-05, "loss": 2.7849, "step": 14912 }, { "epoch": 1.17, "learning_rate": 8.830628087508822e-05, "loss": 3.0609, "step": 14913 }, { "epoch": 1.17, "learning_rate": 8.830549674586372e-05, "loss": 3.0763, "step": 14914 }, { "epoch": 1.17, "learning_rate": 8.830471261663923e-05, "loss": 3.1938, "step": 14915 }, { "epoch": 1.17, "learning_rate": 8.830392848741473e-05, "loss": 2.976, "step": 14916 }, { "epoch": 1.17, "learning_rate": 8.830314435819024e-05, "loss": 2.9797, "step": 14917 }, { "epoch": 1.17, "learning_rate": 8.830236022896574e-05, "loss": 2.7901, "step": 14918 }, { "epoch": 1.17, "learning_rate": 8.830157609974124e-05, "loss": 2.9536, "step": 14919 }, { "epoch": 1.17, "learning_rate": 8.830079197051675e-05, "loss": 2.9471, "step": 14920 }, { "epoch": 1.17, "learning_rate": 8.830000784129225e-05, "loss": 2.912, "step": 14921 }, { "epoch": 1.17, "learning_rate": 8.829922371206775e-05, "loss": 3.0927, "step": 14922 }, { "epoch": 1.17, "learning_rate": 8.829843958284326e-05, "loss": 2.8049, "step": 14923 }, { "epoch": 1.17, "learning_rate": 8.829765545361876e-05, "loss": 2.8677, "step": 14924 }, { "epoch": 1.17, "learning_rate": 8.829687132439426e-05, "loss": 3.0004, "step": 14925 }, { "epoch": 1.17, "learning_rate": 8.829608719516977e-05, "loss": 2.7915, "step": 14926 }, { "epoch": 1.17, "learning_rate": 8.829530306594527e-05, "loss": 2.9282, "step": 14927 }, { "epoch": 1.17, "learning_rate": 8.829451893672078e-05, "loss": 2.7478, "step": 14928 }, { "epoch": 1.17, "learning_rate": 8.829373480749628e-05, "loss": 2.926, "step": 14929 }, { "epoch": 1.17, "learning_rate": 8.829295067827178e-05, "loss": 3.0, "step": 14930 }, { "epoch": 1.17, "learning_rate": 8.829216654904729e-05, "loss": 2.8973, "step": 14931 }, { "epoch": 1.17, "learning_rate": 8.829138241982279e-05, "loss": 2.849, "step": 14932 }, { "epoch": 1.17, "learning_rate": 8.82905982905983e-05, "loss": 2.8736, "step": 14933 }, { "epoch": 1.17, "learning_rate": 8.82898141613738e-05, "loss": 2.7686, "step": 14934 }, { "epoch": 1.17, "learning_rate": 8.82890300321493e-05, "loss": 2.8861, "step": 14935 }, { "epoch": 1.17, "learning_rate": 8.82882459029248e-05, "loss": 2.8583, "step": 14936 }, { "epoch": 1.17, "learning_rate": 8.828746177370031e-05, "loss": 2.891, "step": 14937 }, { "epoch": 1.17, "learning_rate": 8.828667764447581e-05, "loss": 3.0528, "step": 14938 }, { "epoch": 1.17, "learning_rate": 8.828589351525132e-05, "loss": 2.8873, "step": 14939 }, { "epoch": 1.17, "learning_rate": 8.828510938602682e-05, "loss": 2.8843, "step": 14940 }, { "epoch": 1.17, "learning_rate": 8.828432525680232e-05, "loss": 2.8266, "step": 14941 }, { "epoch": 1.17, "learning_rate": 8.828354112757783e-05, "loss": 2.7813, "step": 14942 }, { "epoch": 1.17, "learning_rate": 8.828275699835333e-05, "loss": 2.8367, "step": 14943 }, { "epoch": 1.17, "learning_rate": 8.828197286912883e-05, "loss": 2.8575, "step": 14944 }, { "epoch": 1.17, "learning_rate": 8.828118873990434e-05, "loss": 2.8755, "step": 14945 }, { "epoch": 1.17, "learning_rate": 8.828040461067984e-05, "loss": 2.7792, "step": 14946 }, { "epoch": 1.17, "learning_rate": 8.827962048145535e-05, "loss": 2.8015, "step": 14947 }, { "epoch": 1.17, "learning_rate": 8.827883635223085e-05, "loss": 3.1621, "step": 14948 }, { "epoch": 1.17, "learning_rate": 8.827805222300635e-05, "loss": 2.8529, "step": 14949 }, { "epoch": 1.17, "learning_rate": 8.827726809378186e-05, "loss": 2.7143, "step": 14950 }, { "epoch": 1.17, "learning_rate": 8.827648396455736e-05, "loss": 2.976, "step": 14951 }, { "epoch": 1.17, "learning_rate": 8.827569983533286e-05, "loss": 2.7849, "step": 14952 }, { "epoch": 1.17, "learning_rate": 8.827491570610837e-05, "loss": 3.0501, "step": 14953 }, { "epoch": 1.17, "learning_rate": 8.827413157688388e-05, "loss": 2.995, "step": 14954 }, { "epoch": 1.17, "learning_rate": 8.827334744765939e-05, "loss": 3.0988, "step": 14955 }, { "epoch": 1.17, "learning_rate": 8.827256331843489e-05, "loss": 2.8777, "step": 14956 }, { "epoch": 1.17, "learning_rate": 8.82717791892104e-05, "loss": 2.8289, "step": 14957 }, { "epoch": 1.17, "learning_rate": 8.82709950599859e-05, "loss": 2.9961, "step": 14958 }, { "epoch": 1.17, "learning_rate": 8.82702109307614e-05, "loss": 2.7763, "step": 14959 }, { "epoch": 1.17, "learning_rate": 8.826942680153691e-05, "loss": 2.7834, "step": 14960 }, { "epoch": 1.17, "learning_rate": 8.82686426723124e-05, "loss": 2.8807, "step": 14961 }, { "epoch": 1.17, "learning_rate": 8.82678585430879e-05, "loss": 2.7871, "step": 14962 }, { "epoch": 1.17, "learning_rate": 8.82670744138634e-05, "loss": 3.0655, "step": 14963 }, { "epoch": 1.17, "learning_rate": 8.826629028463891e-05, "loss": 2.8519, "step": 14964 }, { "epoch": 1.17, "learning_rate": 8.826550615541441e-05, "loss": 3.2385, "step": 14965 }, { "epoch": 1.17, "learning_rate": 8.826472202618992e-05, "loss": 2.8175, "step": 14966 }, { "epoch": 1.17, "learning_rate": 8.826393789696542e-05, "loss": 2.8094, "step": 14967 }, { "epoch": 1.17, "learning_rate": 8.826315376774092e-05, "loss": 2.8735, "step": 14968 }, { "epoch": 1.17, "learning_rate": 8.826236963851643e-05, "loss": 2.9657, "step": 14969 }, { "epoch": 1.17, "learning_rate": 8.826158550929193e-05, "loss": 2.8504, "step": 14970 }, { "epoch": 1.17, "learning_rate": 8.826080138006743e-05, "loss": 2.9857, "step": 14971 }, { "epoch": 1.17, "learning_rate": 8.826001725084294e-05, "loss": 2.8414, "step": 14972 }, { "epoch": 1.17, "learning_rate": 8.825923312161844e-05, "loss": 2.8534, "step": 14973 }, { "epoch": 1.17, "learning_rate": 8.825844899239394e-05, "loss": 2.8296, "step": 14974 }, { "epoch": 1.17, "learning_rate": 8.825766486316945e-05, "loss": 2.92, "step": 14975 }, { "epoch": 1.17, "learning_rate": 8.825688073394495e-05, "loss": 3.0033, "step": 14976 }, { "epoch": 1.17, "learning_rate": 8.825609660472046e-05, "loss": 2.9543, "step": 14977 }, { "epoch": 1.17, "learning_rate": 8.825531247549596e-05, "loss": 2.9854, "step": 14978 }, { "epoch": 1.17, "learning_rate": 8.825452834627146e-05, "loss": 2.8239, "step": 14979 }, { "epoch": 1.17, "learning_rate": 8.825374421704698e-05, "loss": 2.894, "step": 14980 }, { "epoch": 1.17, "learning_rate": 8.825296008782248e-05, "loss": 3.0819, "step": 14981 }, { "epoch": 1.17, "learning_rate": 8.825217595859799e-05, "loss": 3.1606, "step": 14982 }, { "epoch": 1.17, "learning_rate": 8.825139182937349e-05, "loss": 2.8755, "step": 14983 }, { "epoch": 1.17, "learning_rate": 8.8250607700149e-05, "loss": 3.0018, "step": 14984 }, { "epoch": 1.18, "learning_rate": 8.82498235709245e-05, "loss": 2.9359, "step": 14985 }, { "epoch": 1.18, "learning_rate": 8.82490394417e-05, "loss": 2.8741, "step": 14986 }, { "epoch": 1.18, "learning_rate": 8.82482553124755e-05, "loss": 2.8019, "step": 14987 }, { "epoch": 1.18, "learning_rate": 8.824747118325101e-05, "loss": 2.9098, "step": 14988 }, { "epoch": 1.18, "learning_rate": 8.824668705402651e-05, "loss": 2.8685, "step": 14989 }, { "epoch": 1.18, "learning_rate": 8.824590292480202e-05, "loss": 3.1497, "step": 14990 }, { "epoch": 1.18, "learning_rate": 8.824511879557752e-05, "loss": 2.7956, "step": 14991 }, { "epoch": 1.18, "learning_rate": 8.824433466635302e-05, "loss": 2.8735, "step": 14992 }, { "epoch": 1.18, "learning_rate": 8.824355053712853e-05, "loss": 2.9771, "step": 14993 }, { "epoch": 1.18, "learning_rate": 8.824276640790403e-05, "loss": 2.6882, "step": 14994 }, { "epoch": 1.18, "learning_rate": 8.824198227867954e-05, "loss": 2.9978, "step": 14995 }, { "epoch": 1.18, "learning_rate": 8.824119814945504e-05, "loss": 2.9365, "step": 14996 }, { "epoch": 1.18, "learning_rate": 8.824041402023053e-05, "loss": 2.8766, "step": 14997 }, { "epoch": 1.18, "learning_rate": 8.823962989100603e-05, "loss": 2.7321, "step": 14998 }, { "epoch": 1.18, "learning_rate": 8.823884576178154e-05, "loss": 2.8462, "step": 14999 }, { "epoch": 1.18, "learning_rate": 8.823806163255704e-05, "loss": 2.7004, "step": 15000 }, { "epoch": 1.18, "eval_loss": 2.9245760440826416, "eval_runtime": 218.0329, "eval_samples_per_second": 51.974, "eval_steps_per_second": 3.252, "step": 15000 }, { "epoch": 1.18, "learning_rate": 8.823727750333254e-05, "loss": 2.7854, "step": 15001 }, { "epoch": 1.18, "learning_rate": 8.823649337410805e-05, "loss": 2.9255, "step": 15002 }, { "epoch": 1.18, "learning_rate": 8.823570924488355e-05, "loss": 2.9471, "step": 15003 }, { "epoch": 1.18, "learning_rate": 8.823492511565906e-05, "loss": 2.9349, "step": 15004 }, { "epoch": 1.18, "learning_rate": 8.823414098643457e-05, "loss": 2.7417, "step": 15005 }, { "epoch": 1.18, "learning_rate": 8.823335685721008e-05, "loss": 2.9484, "step": 15006 }, { "epoch": 1.18, "learning_rate": 8.823257272798558e-05, "loss": 2.8152, "step": 15007 }, { "epoch": 1.18, "learning_rate": 8.823178859876108e-05, "loss": 2.8121, "step": 15008 }, { "epoch": 1.18, "learning_rate": 8.823100446953659e-05, "loss": 2.7914, "step": 15009 }, { "epoch": 1.18, "learning_rate": 8.823022034031209e-05, "loss": 2.933, "step": 15010 }, { "epoch": 1.18, "learning_rate": 8.82294362110876e-05, "loss": 2.7231, "step": 15011 }, { "epoch": 1.18, "learning_rate": 8.82286520818631e-05, "loss": 2.9507, "step": 15012 }, { "epoch": 1.18, "learning_rate": 8.82278679526386e-05, "loss": 2.8637, "step": 15013 }, { "epoch": 1.18, "learning_rate": 8.82270838234141e-05, "loss": 2.9307, "step": 15014 }, { "epoch": 1.18, "learning_rate": 8.822629969418961e-05, "loss": 2.8628, "step": 15015 }, { "epoch": 1.18, "learning_rate": 8.822551556496511e-05, "loss": 3.0598, "step": 15016 }, { "epoch": 1.18, "learning_rate": 8.822473143574062e-05, "loss": 2.9314, "step": 15017 }, { "epoch": 1.18, "learning_rate": 8.822394730651612e-05, "loss": 2.6648, "step": 15018 }, { "epoch": 1.18, "learning_rate": 8.822316317729162e-05, "loss": 2.7624, "step": 15019 }, { "epoch": 1.18, "learning_rate": 8.822237904806713e-05, "loss": 2.9654, "step": 15020 }, { "epoch": 1.18, "learning_rate": 8.822159491884263e-05, "loss": 2.902, "step": 15021 }, { "epoch": 1.18, "learning_rate": 8.822081078961813e-05, "loss": 2.9742, "step": 15022 }, { "epoch": 1.18, "learning_rate": 8.822002666039364e-05, "loss": 2.9196, "step": 15023 }, { "epoch": 1.18, "learning_rate": 8.821924253116914e-05, "loss": 2.9189, "step": 15024 }, { "epoch": 1.18, "learning_rate": 8.821845840194465e-05, "loss": 2.8363, "step": 15025 }, { "epoch": 1.18, "learning_rate": 8.821767427272015e-05, "loss": 2.8333, "step": 15026 }, { "epoch": 1.18, "learning_rate": 8.821689014349565e-05, "loss": 2.9424, "step": 15027 }, { "epoch": 1.18, "learning_rate": 8.821610601427116e-05, "loss": 2.8171, "step": 15028 }, { "epoch": 1.18, "learning_rate": 8.821532188504666e-05, "loss": 3.1285, "step": 15029 }, { "epoch": 1.18, "learning_rate": 8.821453775582216e-05, "loss": 2.7205, "step": 15030 }, { "epoch": 1.18, "learning_rate": 8.821375362659767e-05, "loss": 2.8937, "step": 15031 }, { "epoch": 1.18, "learning_rate": 8.821296949737317e-05, "loss": 2.6284, "step": 15032 }, { "epoch": 1.18, "learning_rate": 8.821218536814868e-05, "loss": 2.8921, "step": 15033 }, { "epoch": 1.18, "learning_rate": 8.821140123892418e-05, "loss": 2.9869, "step": 15034 }, { "epoch": 1.18, "learning_rate": 8.821061710969968e-05, "loss": 3.0162, "step": 15035 }, { "epoch": 1.18, "learning_rate": 8.820983298047519e-05, "loss": 2.9037, "step": 15036 }, { "epoch": 1.18, "learning_rate": 8.820904885125069e-05, "loss": 2.8976, "step": 15037 }, { "epoch": 1.18, "learning_rate": 8.82082647220262e-05, "loss": 2.8615, "step": 15038 }, { "epoch": 1.18, "learning_rate": 8.82074805928017e-05, "loss": 2.9554, "step": 15039 }, { "epoch": 1.18, "learning_rate": 8.82066964635772e-05, "loss": 2.9461, "step": 15040 }, { "epoch": 1.18, "learning_rate": 8.82059123343527e-05, "loss": 2.9535, "step": 15041 }, { "epoch": 1.18, "learning_rate": 8.820512820512821e-05, "loss": 2.9178, "step": 15042 }, { "epoch": 1.18, "learning_rate": 8.820434407590371e-05, "loss": 2.7828, "step": 15043 }, { "epoch": 1.18, "learning_rate": 8.820355994667922e-05, "loss": 2.7171, "step": 15044 }, { "epoch": 1.18, "learning_rate": 8.820277581745472e-05, "loss": 2.8866, "step": 15045 }, { "epoch": 1.18, "learning_rate": 8.820199168823022e-05, "loss": 2.8775, "step": 15046 }, { "epoch": 1.18, "learning_rate": 8.820120755900573e-05, "loss": 3.3882, "step": 15047 }, { "epoch": 1.18, "learning_rate": 8.820042342978123e-05, "loss": 3.076, "step": 15048 }, { "epoch": 1.18, "learning_rate": 8.819963930055673e-05, "loss": 3.0366, "step": 15049 }, { "epoch": 1.18, "learning_rate": 8.819885517133224e-05, "loss": 2.8734, "step": 15050 }, { "epoch": 1.18, "learning_rate": 8.819807104210774e-05, "loss": 2.7718, "step": 15051 }, { "epoch": 1.18, "learning_rate": 8.819728691288325e-05, "loss": 2.7014, "step": 15052 }, { "epoch": 1.18, "learning_rate": 8.819650278365875e-05, "loss": 3.2426, "step": 15053 }, { "epoch": 1.18, "learning_rate": 8.819571865443425e-05, "loss": 2.8896, "step": 15054 }, { "epoch": 1.18, "learning_rate": 8.819493452520976e-05, "loss": 2.7352, "step": 15055 }, { "epoch": 1.18, "learning_rate": 8.819415039598526e-05, "loss": 2.8211, "step": 15056 }, { "epoch": 1.18, "learning_rate": 8.819336626676076e-05, "loss": 2.9166, "step": 15057 }, { "epoch": 1.18, "learning_rate": 8.819258213753627e-05, "loss": 2.9583, "step": 15058 }, { "epoch": 1.18, "learning_rate": 8.819179800831177e-05, "loss": 2.68, "step": 15059 }, { "epoch": 1.18, "learning_rate": 8.819101387908727e-05, "loss": 2.7772, "step": 15060 }, { "epoch": 1.18, "learning_rate": 8.819022974986278e-05, "loss": 2.985, "step": 15061 }, { "epoch": 1.18, "learning_rate": 8.81894456206383e-05, "loss": 3.083, "step": 15062 }, { "epoch": 1.18, "learning_rate": 8.81886614914138e-05, "loss": 3.0141, "step": 15063 }, { "epoch": 1.18, "learning_rate": 8.81878773621893e-05, "loss": 2.8159, "step": 15064 }, { "epoch": 1.18, "learning_rate": 8.81870932329648e-05, "loss": 2.8445, "step": 15065 }, { "epoch": 1.18, "learning_rate": 8.818630910374031e-05, "loss": 2.9079, "step": 15066 }, { "epoch": 1.18, "learning_rate": 8.818552497451581e-05, "loss": 2.8909, "step": 15067 }, { "epoch": 1.18, "learning_rate": 8.818474084529132e-05, "loss": 2.9402, "step": 15068 }, { "epoch": 1.18, "learning_rate": 8.818395671606681e-05, "loss": 2.8397, "step": 15069 }, { "epoch": 1.18, "learning_rate": 8.818317258684231e-05, "loss": 2.7514, "step": 15070 }, { "epoch": 1.18, "learning_rate": 8.818238845761782e-05, "loss": 3.1347, "step": 15071 }, { "epoch": 1.18, "learning_rate": 8.818160432839332e-05, "loss": 3.0509, "step": 15072 }, { "epoch": 1.18, "learning_rate": 8.818082019916882e-05, "loss": 2.7401, "step": 15073 }, { "epoch": 1.18, "learning_rate": 8.818003606994433e-05, "loss": 3.0475, "step": 15074 }, { "epoch": 1.18, "learning_rate": 8.817925194071983e-05, "loss": 2.7317, "step": 15075 }, { "epoch": 1.18, "learning_rate": 8.817846781149533e-05, "loss": 2.7341, "step": 15076 }, { "epoch": 1.18, "learning_rate": 8.817768368227084e-05, "loss": 2.8759, "step": 15077 }, { "epoch": 1.18, "learning_rate": 8.817689955304634e-05, "loss": 3.2477, "step": 15078 }, { "epoch": 1.18, "learning_rate": 8.817611542382184e-05, "loss": 2.9111, "step": 15079 }, { "epoch": 1.18, "learning_rate": 8.817533129459735e-05, "loss": 2.8425, "step": 15080 }, { "epoch": 1.18, "learning_rate": 8.817454716537285e-05, "loss": 2.7742, "step": 15081 }, { "epoch": 1.18, "learning_rate": 8.817376303614836e-05, "loss": 2.9204, "step": 15082 }, { "epoch": 1.18, "learning_rate": 8.817297890692386e-05, "loss": 2.8182, "step": 15083 }, { "epoch": 1.18, "learning_rate": 8.817219477769936e-05, "loss": 2.872, "step": 15084 }, { "epoch": 1.18, "learning_rate": 8.817141064847487e-05, "loss": 2.8995, "step": 15085 }, { "epoch": 1.18, "learning_rate": 8.817062651925037e-05, "loss": 2.9228, "step": 15086 }, { "epoch": 1.18, "learning_rate": 8.816984239002587e-05, "loss": 2.9999, "step": 15087 }, { "epoch": 1.18, "learning_rate": 8.816905826080139e-05, "loss": 2.8191, "step": 15088 }, { "epoch": 1.18, "learning_rate": 8.81682741315769e-05, "loss": 3.1103, "step": 15089 }, { "epoch": 1.18, "learning_rate": 8.81674900023524e-05, "loss": 2.98, "step": 15090 }, { "epoch": 1.18, "learning_rate": 8.81667058731279e-05, "loss": 2.8478, "step": 15091 }, { "epoch": 1.18, "learning_rate": 8.81659217439034e-05, "loss": 2.9141, "step": 15092 }, { "epoch": 1.18, "learning_rate": 8.816513761467891e-05, "loss": 2.9784, "step": 15093 }, { "epoch": 1.18, "learning_rate": 8.816435348545441e-05, "loss": 2.9101, "step": 15094 }, { "epoch": 1.18, "learning_rate": 8.816356935622992e-05, "loss": 3.0667, "step": 15095 }, { "epoch": 1.18, "learning_rate": 8.816278522700542e-05, "loss": 2.9993, "step": 15096 }, { "epoch": 1.18, "learning_rate": 8.816200109778092e-05, "loss": 2.8484, "step": 15097 }, { "epoch": 1.18, "learning_rate": 8.816121696855643e-05, "loss": 2.7665, "step": 15098 }, { "epoch": 1.18, "learning_rate": 8.816043283933193e-05, "loss": 2.7664, "step": 15099 }, { "epoch": 1.18, "learning_rate": 8.815964871010744e-05, "loss": 2.9041, "step": 15100 }, { "epoch": 1.18, "learning_rate": 8.815886458088294e-05, "loss": 2.9967, "step": 15101 }, { "epoch": 1.18, "learning_rate": 8.815808045165844e-05, "loss": 2.8477, "step": 15102 }, { "epoch": 1.18, "learning_rate": 8.815729632243395e-05, "loss": 3.0525, "step": 15103 }, { "epoch": 1.18, "learning_rate": 8.815651219320945e-05, "loss": 3.0712, "step": 15104 }, { "epoch": 1.18, "learning_rate": 8.815572806398494e-05, "loss": 3.2679, "step": 15105 }, { "epoch": 1.18, "learning_rate": 8.815494393476044e-05, "loss": 2.8691, "step": 15106 }, { "epoch": 1.18, "learning_rate": 8.815415980553595e-05, "loss": 2.929, "step": 15107 }, { "epoch": 1.18, "learning_rate": 8.815337567631145e-05, "loss": 3.1311, "step": 15108 }, { "epoch": 1.18, "learning_rate": 8.815259154708695e-05, "loss": 2.9062, "step": 15109 }, { "epoch": 1.18, "learning_rate": 8.815180741786246e-05, "loss": 2.861, "step": 15110 }, { "epoch": 1.18, "learning_rate": 8.815102328863796e-05, "loss": 3.0121, "step": 15111 }, { "epoch": 1.18, "learning_rate": 8.815023915941347e-05, "loss": 2.8891, "step": 15112 }, { "epoch": 1.19, "learning_rate": 8.814945503018898e-05, "loss": 2.9172, "step": 15113 }, { "epoch": 1.19, "learning_rate": 8.814867090096449e-05, "loss": 2.9361, "step": 15114 }, { "epoch": 1.19, "learning_rate": 8.814788677173999e-05, "loss": 2.8332, "step": 15115 }, { "epoch": 1.19, "learning_rate": 8.81471026425155e-05, "loss": 2.7698, "step": 15116 }, { "epoch": 1.19, "learning_rate": 8.8146318513291e-05, "loss": 3.0637, "step": 15117 }, { "epoch": 1.19, "learning_rate": 8.81455343840665e-05, "loss": 2.8053, "step": 15118 }, { "epoch": 1.19, "learning_rate": 8.8144750254842e-05, "loss": 2.8564, "step": 15119 }, { "epoch": 1.19, "learning_rate": 8.814396612561751e-05, "loss": 3.1661, "step": 15120 }, { "epoch": 1.19, "learning_rate": 8.814318199639301e-05, "loss": 2.9426, "step": 15121 }, { "epoch": 1.19, "learning_rate": 8.814239786716852e-05, "loss": 3.3756, "step": 15122 }, { "epoch": 1.19, "learning_rate": 8.814161373794402e-05, "loss": 3.0607, "step": 15123 }, { "epoch": 1.19, "learning_rate": 8.814082960871952e-05, "loss": 2.931, "step": 15124 }, { "epoch": 1.19, "learning_rate": 8.814004547949503e-05, "loss": 3.146, "step": 15125 }, { "epoch": 1.19, "learning_rate": 8.813926135027053e-05, "loss": 3.0345, "step": 15126 }, { "epoch": 1.19, "learning_rate": 8.813847722104603e-05, "loss": 2.7619, "step": 15127 }, { "epoch": 1.19, "learning_rate": 8.813769309182154e-05, "loss": 2.7496, "step": 15128 }, { "epoch": 1.19, "learning_rate": 8.813690896259704e-05, "loss": 2.9265, "step": 15129 }, { "epoch": 1.19, "learning_rate": 8.813612483337255e-05, "loss": 2.7207, "step": 15130 }, { "epoch": 1.19, "learning_rate": 8.813534070414805e-05, "loss": 2.8854, "step": 15131 }, { "epoch": 1.19, "learning_rate": 8.813455657492355e-05, "loss": 3.1428, "step": 15132 }, { "epoch": 1.19, "learning_rate": 8.813377244569906e-05, "loss": 2.724, "step": 15133 }, { "epoch": 1.19, "learning_rate": 8.813298831647456e-05, "loss": 2.8176, "step": 15134 }, { "epoch": 1.19, "learning_rate": 8.813220418725006e-05, "loss": 2.9364, "step": 15135 }, { "epoch": 1.19, "learning_rate": 8.813142005802557e-05, "loss": 2.8072, "step": 15136 }, { "epoch": 1.19, "learning_rate": 8.813063592880107e-05, "loss": 2.9411, "step": 15137 }, { "epoch": 1.19, "learning_rate": 8.812985179957658e-05, "loss": 2.7904, "step": 15138 }, { "epoch": 1.19, "learning_rate": 8.812906767035208e-05, "loss": 2.8127, "step": 15139 }, { "epoch": 1.19, "learning_rate": 8.812828354112758e-05, "loss": 2.865, "step": 15140 }, { "epoch": 1.19, "learning_rate": 8.812749941190309e-05, "loss": 2.9838, "step": 15141 }, { "epoch": 1.19, "learning_rate": 8.812671528267859e-05, "loss": 2.8214, "step": 15142 }, { "epoch": 1.19, "learning_rate": 8.81259311534541e-05, "loss": 3.1796, "step": 15143 }, { "epoch": 1.19, "learning_rate": 8.81251470242296e-05, "loss": 2.9967, "step": 15144 }, { "epoch": 1.19, "learning_rate": 8.81243628950051e-05, "loss": 2.972, "step": 15145 }, { "epoch": 1.19, "learning_rate": 8.81235787657806e-05, "loss": 2.7988, "step": 15146 }, { "epoch": 1.19, "learning_rate": 8.812279463655611e-05, "loss": 2.8194, "step": 15147 }, { "epoch": 1.19, "learning_rate": 8.812201050733161e-05, "loss": 2.7951, "step": 15148 }, { "epoch": 1.19, "learning_rate": 8.812122637810712e-05, "loss": 2.9403, "step": 15149 }, { "epoch": 1.19, "learning_rate": 8.812044224888262e-05, "loss": 2.981, "step": 15150 }, { "epoch": 1.19, "learning_rate": 8.811965811965812e-05, "loss": 3.0797, "step": 15151 }, { "epoch": 1.19, "learning_rate": 8.811887399043363e-05, "loss": 3.0391, "step": 15152 }, { "epoch": 1.19, "learning_rate": 8.811808986120913e-05, "loss": 2.9699, "step": 15153 }, { "epoch": 1.19, "learning_rate": 8.811730573198463e-05, "loss": 2.8536, "step": 15154 }, { "epoch": 1.19, "learning_rate": 8.811652160276014e-05, "loss": 3.2947, "step": 15155 }, { "epoch": 1.19, "learning_rate": 8.811573747353564e-05, "loss": 3.05, "step": 15156 }, { "epoch": 1.19, "learning_rate": 8.811495334431115e-05, "loss": 2.9905, "step": 15157 }, { "epoch": 1.19, "learning_rate": 8.811416921508665e-05, "loss": 3.1547, "step": 15158 }, { "epoch": 1.19, "learning_rate": 8.811338508586215e-05, "loss": 2.872, "step": 15159 }, { "epoch": 1.19, "learning_rate": 8.811260095663766e-05, "loss": 2.9185, "step": 15160 }, { "epoch": 1.19, "learning_rate": 8.811181682741316e-05, "loss": 2.7268, "step": 15161 }, { "epoch": 1.19, "learning_rate": 8.811103269818866e-05, "loss": 3.0193, "step": 15162 }, { "epoch": 1.19, "learning_rate": 8.811024856896417e-05, "loss": 2.9987, "step": 15163 }, { "epoch": 1.19, "learning_rate": 8.810946443973967e-05, "loss": 2.8362, "step": 15164 }, { "epoch": 1.19, "learning_rate": 8.810868031051517e-05, "loss": 2.78, "step": 15165 }, { "epoch": 1.19, "learning_rate": 8.810789618129068e-05, "loss": 2.9136, "step": 15166 }, { "epoch": 1.19, "learning_rate": 8.810711205206618e-05, "loss": 2.8418, "step": 15167 }, { "epoch": 1.19, "learning_rate": 8.810632792284169e-05, "loss": 2.9096, "step": 15168 }, { "epoch": 1.19, "learning_rate": 8.810554379361719e-05, "loss": 3.0504, "step": 15169 }, { "epoch": 1.19, "learning_rate": 8.81047596643927e-05, "loss": 2.8804, "step": 15170 }, { "epoch": 1.19, "learning_rate": 8.810397553516821e-05, "loss": 2.9442, "step": 15171 }, { "epoch": 1.19, "learning_rate": 8.810319140594371e-05, "loss": 2.9736, "step": 15172 }, { "epoch": 1.19, "learning_rate": 8.810240727671922e-05, "loss": 3.0596, "step": 15173 }, { "epoch": 1.19, "learning_rate": 8.810162314749472e-05, "loss": 3.0731, "step": 15174 }, { "epoch": 1.19, "learning_rate": 8.810083901827022e-05, "loss": 2.7976, "step": 15175 }, { "epoch": 1.19, "learning_rate": 8.810005488904573e-05, "loss": 2.9308, "step": 15176 }, { "epoch": 1.19, "learning_rate": 8.809927075982122e-05, "loss": 2.6872, "step": 15177 }, { "epoch": 1.19, "learning_rate": 8.809848663059672e-05, "loss": 2.7391, "step": 15178 }, { "epoch": 1.19, "learning_rate": 8.809770250137223e-05, "loss": 2.6526, "step": 15179 }, { "epoch": 1.19, "learning_rate": 8.809691837214773e-05, "loss": 3.0014, "step": 15180 }, { "epoch": 1.19, "learning_rate": 8.809613424292323e-05, "loss": 2.9655, "step": 15181 }, { "epoch": 1.19, "learning_rate": 8.809535011369874e-05, "loss": 2.7726, "step": 15182 }, { "epoch": 1.19, "learning_rate": 8.809456598447424e-05, "loss": 2.8795, "step": 15183 }, { "epoch": 1.19, "learning_rate": 8.809378185524974e-05, "loss": 3.1014, "step": 15184 }, { "epoch": 1.19, "learning_rate": 8.809299772602525e-05, "loss": 2.837, "step": 15185 }, { "epoch": 1.19, "learning_rate": 8.809221359680075e-05, "loss": 2.8308, "step": 15186 }, { "epoch": 1.19, "learning_rate": 8.809142946757626e-05, "loss": 2.813, "step": 15187 }, { "epoch": 1.19, "learning_rate": 8.809064533835176e-05, "loss": 2.9778, "step": 15188 }, { "epoch": 1.19, "learning_rate": 8.808986120912726e-05, "loss": 2.8175, "step": 15189 }, { "epoch": 1.19, "learning_rate": 8.808907707990277e-05, "loss": 3.0821, "step": 15190 }, { "epoch": 1.19, "learning_rate": 8.808829295067827e-05, "loss": 2.5858, "step": 15191 }, { "epoch": 1.19, "learning_rate": 8.808750882145377e-05, "loss": 2.9625, "step": 15192 }, { "epoch": 1.19, "learning_rate": 8.808672469222928e-05, "loss": 2.7113, "step": 15193 }, { "epoch": 1.19, "learning_rate": 8.808594056300478e-05, "loss": 2.9899, "step": 15194 }, { "epoch": 1.19, "learning_rate": 8.808515643378028e-05, "loss": 2.9398, "step": 15195 }, { "epoch": 1.19, "learning_rate": 8.80843723045558e-05, "loss": 2.6509, "step": 15196 }, { "epoch": 1.19, "learning_rate": 8.80835881753313e-05, "loss": 2.7993, "step": 15197 }, { "epoch": 1.19, "learning_rate": 8.808280404610681e-05, "loss": 2.8686, "step": 15198 }, { "epoch": 1.19, "learning_rate": 8.808201991688231e-05, "loss": 2.95, "step": 15199 }, { "epoch": 1.19, "learning_rate": 8.808123578765782e-05, "loss": 2.8032, "step": 15200 }, { "epoch": 1.19, "learning_rate": 8.808045165843332e-05, "loss": 2.9411, "step": 15201 }, { "epoch": 1.19, "learning_rate": 8.807966752920882e-05, "loss": 2.8923, "step": 15202 }, { "epoch": 1.19, "learning_rate": 8.807888339998433e-05, "loss": 3.0095, "step": 15203 }, { "epoch": 1.19, "learning_rate": 8.807809927075983e-05, "loss": 2.9811, "step": 15204 }, { "epoch": 1.19, "learning_rate": 8.807731514153534e-05, "loss": 3.1358, "step": 15205 }, { "epoch": 1.19, "learning_rate": 8.807653101231084e-05, "loss": 3.0558, "step": 15206 }, { "epoch": 1.19, "learning_rate": 8.807574688308634e-05, "loss": 2.893, "step": 15207 }, { "epoch": 1.19, "learning_rate": 8.807496275386185e-05, "loss": 2.8781, "step": 15208 }, { "epoch": 1.19, "learning_rate": 8.807417862463735e-05, "loss": 3.399, "step": 15209 }, { "epoch": 1.19, "learning_rate": 8.807339449541285e-05, "loss": 2.9943, "step": 15210 }, { "epoch": 1.19, "learning_rate": 8.807261036618836e-05, "loss": 2.959, "step": 15211 }, { "epoch": 1.19, "learning_rate": 8.807182623696386e-05, "loss": 2.9995, "step": 15212 }, { "epoch": 1.19, "learning_rate": 8.807104210773935e-05, "loss": 2.8996, "step": 15213 }, { "epoch": 1.19, "learning_rate": 8.807025797851485e-05, "loss": 3.0292, "step": 15214 }, { "epoch": 1.19, "learning_rate": 8.806947384929036e-05, "loss": 2.9399, "step": 15215 }, { "epoch": 1.19, "learning_rate": 8.806868972006586e-05, "loss": 2.7893, "step": 15216 }, { "epoch": 1.19, "learning_rate": 8.806790559084137e-05, "loss": 2.8678, "step": 15217 }, { "epoch": 1.19, "learning_rate": 8.806712146161687e-05, "loss": 2.7406, "step": 15218 }, { "epoch": 1.19, "learning_rate": 8.806633733239237e-05, "loss": 2.7703, "step": 15219 }, { "epoch": 1.19, "learning_rate": 8.806555320316788e-05, "loss": 3.0808, "step": 15220 }, { "epoch": 1.19, "learning_rate": 8.80647690739434e-05, "loss": 2.8802, "step": 15221 }, { "epoch": 1.19, "learning_rate": 8.80639849447189e-05, "loss": 2.9488, "step": 15222 }, { "epoch": 1.19, "learning_rate": 8.80632008154944e-05, "loss": 2.8615, "step": 15223 }, { "epoch": 1.19, "learning_rate": 8.80624166862699e-05, "loss": 3.082, "step": 15224 }, { "epoch": 1.19, "learning_rate": 8.806163255704541e-05, "loss": 2.8105, "step": 15225 }, { "epoch": 1.19, "learning_rate": 8.806084842782091e-05, "loss": 3.0346, "step": 15226 }, { "epoch": 1.19, "learning_rate": 8.806006429859642e-05, "loss": 2.8496, "step": 15227 }, { "epoch": 1.19, "learning_rate": 8.805928016937192e-05, "loss": 3.0124, "step": 15228 }, { "epoch": 1.19, "learning_rate": 8.805849604014742e-05, "loss": 2.9477, "step": 15229 }, { "epoch": 1.19, "learning_rate": 8.805771191092293e-05, "loss": 3.3834, "step": 15230 }, { "epoch": 1.19, "learning_rate": 8.805692778169843e-05, "loss": 3.0349, "step": 15231 }, { "epoch": 1.19, "learning_rate": 8.805614365247393e-05, "loss": 2.8482, "step": 15232 }, { "epoch": 1.19, "learning_rate": 8.805535952324944e-05, "loss": 2.8206, "step": 15233 }, { "epoch": 1.19, "learning_rate": 8.805457539402494e-05, "loss": 2.8376, "step": 15234 }, { "epoch": 1.19, "learning_rate": 8.805379126480045e-05, "loss": 2.7348, "step": 15235 }, { "epoch": 1.19, "learning_rate": 8.805300713557595e-05, "loss": 2.7152, "step": 15236 }, { "epoch": 1.19, "learning_rate": 8.805222300635145e-05, "loss": 2.8139, "step": 15237 }, { "epoch": 1.19, "learning_rate": 8.805143887712696e-05, "loss": 2.794, "step": 15238 }, { "epoch": 1.19, "learning_rate": 8.805065474790246e-05, "loss": 2.7308, "step": 15239 }, { "epoch": 1.2, "learning_rate": 8.804987061867796e-05, "loss": 2.7718, "step": 15240 }, { "epoch": 1.2, "learning_rate": 8.804908648945347e-05, "loss": 3.0582, "step": 15241 }, { "epoch": 1.2, "learning_rate": 8.804830236022897e-05, "loss": 2.722, "step": 15242 }, { "epoch": 1.2, "learning_rate": 8.804751823100447e-05, "loss": 2.8439, "step": 15243 }, { "epoch": 1.2, "learning_rate": 8.804673410177998e-05, "loss": 2.9139, "step": 15244 }, { "epoch": 1.2, "learning_rate": 8.804594997255548e-05, "loss": 2.9646, "step": 15245 }, { "epoch": 1.2, "learning_rate": 8.804516584333099e-05, "loss": 2.8409, "step": 15246 }, { "epoch": 1.2, "learning_rate": 8.804438171410649e-05, "loss": 2.9795, "step": 15247 }, { "epoch": 1.2, "learning_rate": 8.8043597584882e-05, "loss": 2.9726, "step": 15248 }, { "epoch": 1.2, "learning_rate": 8.80428134556575e-05, "loss": 2.8499, "step": 15249 }, { "epoch": 1.2, "learning_rate": 8.8042029326433e-05, "loss": 3.0098, "step": 15250 }, { "epoch": 1.2, "learning_rate": 8.80412451972085e-05, "loss": 2.9724, "step": 15251 }, { "epoch": 1.2, "learning_rate": 8.804046106798401e-05, "loss": 2.9228, "step": 15252 }, { "epoch": 1.2, "learning_rate": 8.803967693875951e-05, "loss": 2.721, "step": 15253 }, { "epoch": 1.2, "learning_rate": 8.803889280953502e-05, "loss": 2.6442, "step": 15254 }, { "epoch": 1.2, "learning_rate": 8.803810868031052e-05, "loss": 2.8537, "step": 15255 }, { "epoch": 1.2, "learning_rate": 8.803732455108602e-05, "loss": 2.7687, "step": 15256 }, { "epoch": 1.2, "learning_rate": 8.803654042186153e-05, "loss": 2.8832, "step": 15257 }, { "epoch": 1.2, "learning_rate": 8.803575629263703e-05, "loss": 2.7971, "step": 15258 }, { "epoch": 1.2, "learning_rate": 8.803497216341253e-05, "loss": 2.849, "step": 15259 }, { "epoch": 1.2, "learning_rate": 8.803418803418804e-05, "loss": 2.8273, "step": 15260 }, { "epoch": 1.2, "learning_rate": 8.803340390496354e-05, "loss": 2.8925, "step": 15261 }, { "epoch": 1.2, "learning_rate": 8.803261977573904e-05, "loss": 2.9888, "step": 15262 }, { "epoch": 1.2, "learning_rate": 8.803183564651455e-05, "loss": 3.0199, "step": 15263 }, { "epoch": 1.2, "learning_rate": 8.803105151729005e-05, "loss": 2.7354, "step": 15264 }, { "epoch": 1.2, "learning_rate": 8.803026738806556e-05, "loss": 3.0106, "step": 15265 }, { "epoch": 1.2, "learning_rate": 8.802948325884106e-05, "loss": 2.9533, "step": 15266 }, { "epoch": 1.2, "learning_rate": 8.802869912961656e-05, "loss": 2.9037, "step": 15267 }, { "epoch": 1.2, "learning_rate": 8.802791500039207e-05, "loss": 2.9564, "step": 15268 }, { "epoch": 1.2, "learning_rate": 8.802713087116757e-05, "loss": 2.97, "step": 15269 }, { "epoch": 1.2, "learning_rate": 8.802634674194307e-05, "loss": 3.0212, "step": 15270 }, { "epoch": 1.2, "learning_rate": 8.802556261271858e-05, "loss": 3.0796, "step": 15271 }, { "epoch": 1.2, "learning_rate": 8.802477848349408e-05, "loss": 2.9115, "step": 15272 }, { "epoch": 1.2, "learning_rate": 8.802399435426959e-05, "loss": 2.7974, "step": 15273 }, { "epoch": 1.2, "learning_rate": 8.802321022504509e-05, "loss": 2.9304, "step": 15274 }, { "epoch": 1.2, "learning_rate": 8.802242609582059e-05, "loss": 2.9451, "step": 15275 }, { "epoch": 1.2, "learning_rate": 8.80216419665961e-05, "loss": 2.9897, "step": 15276 }, { "epoch": 1.2, "learning_rate": 8.80208578373716e-05, "loss": 2.9403, "step": 15277 }, { "epoch": 1.2, "learning_rate": 8.802007370814712e-05, "loss": 2.9969, "step": 15278 }, { "epoch": 1.2, "learning_rate": 8.801928957892262e-05, "loss": 3.0332, "step": 15279 }, { "epoch": 1.2, "learning_rate": 8.801850544969812e-05, "loss": 2.9816, "step": 15280 }, { "epoch": 1.2, "learning_rate": 8.801772132047363e-05, "loss": 3.1129, "step": 15281 }, { "epoch": 1.2, "learning_rate": 8.801693719124913e-05, "loss": 2.8982, "step": 15282 }, { "epoch": 1.2, "learning_rate": 8.801615306202464e-05, "loss": 2.8146, "step": 15283 }, { "epoch": 1.2, "learning_rate": 8.801536893280014e-05, "loss": 2.6706, "step": 15284 }, { "epoch": 1.2, "learning_rate": 8.801458480357563e-05, "loss": 2.7932, "step": 15285 }, { "epoch": 1.2, "learning_rate": 8.801380067435113e-05, "loss": 2.9615, "step": 15286 }, { "epoch": 1.2, "learning_rate": 8.801301654512664e-05, "loss": 2.8158, "step": 15287 }, { "epoch": 1.2, "learning_rate": 8.801223241590214e-05, "loss": 2.9128, "step": 15288 }, { "epoch": 1.2, "learning_rate": 8.801144828667764e-05, "loss": 2.7521, "step": 15289 }, { "epoch": 1.2, "learning_rate": 8.801066415745315e-05, "loss": 2.7714, "step": 15290 }, { "epoch": 1.2, "learning_rate": 8.800988002822865e-05, "loss": 2.9256, "step": 15291 }, { "epoch": 1.2, "learning_rate": 8.800909589900416e-05, "loss": 3.0382, "step": 15292 }, { "epoch": 1.2, "learning_rate": 8.800831176977966e-05, "loss": 2.99, "step": 15293 }, { "epoch": 1.2, "learning_rate": 8.800752764055516e-05, "loss": 2.8067, "step": 15294 }, { "epoch": 1.2, "learning_rate": 8.800674351133067e-05, "loss": 2.9876, "step": 15295 }, { "epoch": 1.2, "learning_rate": 8.800595938210617e-05, "loss": 2.8886, "step": 15296 }, { "epoch": 1.2, "learning_rate": 8.800517525288167e-05, "loss": 2.981, "step": 15297 }, { "epoch": 1.2, "learning_rate": 8.800439112365718e-05, "loss": 2.8448, "step": 15298 }, { "epoch": 1.2, "learning_rate": 8.800360699443268e-05, "loss": 2.8587, "step": 15299 }, { "epoch": 1.2, "learning_rate": 8.800282286520818e-05, "loss": 2.6323, "step": 15300 }, { "epoch": 1.2, "learning_rate": 8.800203873598369e-05, "loss": 2.7824, "step": 15301 }, { "epoch": 1.2, "learning_rate": 8.800125460675919e-05, "loss": 3.0347, "step": 15302 }, { "epoch": 1.2, "learning_rate": 8.80004704775347e-05, "loss": 3.1694, "step": 15303 }, { "epoch": 1.2, "learning_rate": 8.799968634831021e-05, "loss": 2.696, "step": 15304 }, { "epoch": 1.2, "learning_rate": 8.799890221908572e-05, "loss": 2.8883, "step": 15305 }, { "epoch": 1.2, "learning_rate": 8.799811808986122e-05, "loss": 3.05, "step": 15306 }, { "epoch": 1.2, "learning_rate": 8.799733396063672e-05, "loss": 2.9562, "step": 15307 }, { "epoch": 1.2, "learning_rate": 8.799654983141223e-05, "loss": 2.9421, "step": 15308 }, { "epoch": 1.2, "learning_rate": 8.799576570218773e-05, "loss": 3.1106, "step": 15309 }, { "epoch": 1.2, "learning_rate": 8.799498157296323e-05, "loss": 2.8535, "step": 15310 }, { "epoch": 1.2, "learning_rate": 8.799419744373874e-05, "loss": 2.8536, "step": 15311 }, { "epoch": 1.2, "learning_rate": 8.799341331451424e-05, "loss": 2.9355, "step": 15312 }, { "epoch": 1.2, "learning_rate": 8.799262918528975e-05, "loss": 2.811, "step": 15313 }, { "epoch": 1.2, "learning_rate": 8.799184505606525e-05, "loss": 2.8743, "step": 15314 }, { "epoch": 1.2, "learning_rate": 8.799106092684075e-05, "loss": 2.9319, "step": 15315 }, { "epoch": 1.2, "learning_rate": 8.799027679761626e-05, "loss": 3.1664, "step": 15316 }, { "epoch": 1.2, "learning_rate": 8.798949266839176e-05, "loss": 3.1359, "step": 15317 }, { "epoch": 1.2, "learning_rate": 8.798870853916726e-05, "loss": 2.7466, "step": 15318 }, { "epoch": 1.2, "learning_rate": 8.798792440994277e-05, "loss": 2.9924, "step": 15319 }, { "epoch": 1.2, "learning_rate": 8.798714028071827e-05, "loss": 2.9045, "step": 15320 }, { "epoch": 1.2, "learning_rate": 8.798635615149376e-05, "loss": 2.8917, "step": 15321 }, { "epoch": 1.2, "learning_rate": 8.798557202226927e-05, "loss": 2.9117, "step": 15322 }, { "epoch": 1.2, "learning_rate": 8.798478789304477e-05, "loss": 2.8738, "step": 15323 }, { "epoch": 1.2, "learning_rate": 8.798400376382027e-05, "loss": 3.105, "step": 15324 }, { "epoch": 1.2, "learning_rate": 8.798321963459578e-05, "loss": 2.6885, "step": 15325 }, { "epoch": 1.2, "learning_rate": 8.798243550537128e-05, "loss": 3.1289, "step": 15326 }, { "epoch": 1.2, "learning_rate": 8.798165137614678e-05, "loss": 2.8615, "step": 15327 }, { "epoch": 1.2, "learning_rate": 8.798086724692229e-05, "loss": 2.8388, "step": 15328 }, { "epoch": 1.2, "learning_rate": 8.79800831176978e-05, "loss": 2.9415, "step": 15329 }, { "epoch": 1.2, "learning_rate": 8.797929898847331e-05, "loss": 2.8407, "step": 15330 }, { "epoch": 1.2, "learning_rate": 8.797851485924881e-05, "loss": 2.7709, "step": 15331 }, { "epoch": 1.2, "learning_rate": 8.797773073002432e-05, "loss": 3.0604, "step": 15332 }, { "epoch": 1.2, "learning_rate": 8.797694660079982e-05, "loss": 3.0137, "step": 15333 }, { "epoch": 1.2, "learning_rate": 8.797616247157532e-05, "loss": 3.1189, "step": 15334 }, { "epoch": 1.2, "learning_rate": 8.797537834235083e-05, "loss": 2.7631, "step": 15335 }, { "epoch": 1.2, "learning_rate": 8.797459421312633e-05, "loss": 2.9096, "step": 15336 }, { "epoch": 1.2, "learning_rate": 8.797381008390183e-05, "loss": 2.8679, "step": 15337 }, { "epoch": 1.2, "learning_rate": 8.797302595467734e-05, "loss": 2.8089, "step": 15338 }, { "epoch": 1.2, "learning_rate": 8.797224182545284e-05, "loss": 2.8642, "step": 15339 }, { "epoch": 1.2, "learning_rate": 8.797145769622835e-05, "loss": 2.8436, "step": 15340 }, { "epoch": 1.2, "learning_rate": 8.797067356700385e-05, "loss": 2.8859, "step": 15341 }, { "epoch": 1.2, "learning_rate": 8.796988943777935e-05, "loss": 2.9595, "step": 15342 }, { "epoch": 1.2, "learning_rate": 8.796910530855486e-05, "loss": 2.7018, "step": 15343 }, { "epoch": 1.2, "learning_rate": 8.796832117933036e-05, "loss": 2.9574, "step": 15344 }, { "epoch": 1.2, "learning_rate": 8.796753705010586e-05, "loss": 2.9227, "step": 15345 }, { "epoch": 1.2, "learning_rate": 8.796675292088137e-05, "loss": 2.6806, "step": 15346 }, { "epoch": 1.2, "learning_rate": 8.796596879165687e-05, "loss": 3.0465, "step": 15347 }, { "epoch": 1.2, "learning_rate": 8.796518466243237e-05, "loss": 2.6819, "step": 15348 }, { "epoch": 1.2, "learning_rate": 8.796440053320788e-05, "loss": 2.8933, "step": 15349 }, { "epoch": 1.2, "learning_rate": 8.796361640398338e-05, "loss": 2.9476, "step": 15350 }, { "epoch": 1.2, "learning_rate": 8.796283227475889e-05, "loss": 3.0296, "step": 15351 }, { "epoch": 1.2, "learning_rate": 8.796204814553439e-05, "loss": 2.8669, "step": 15352 }, { "epoch": 1.2, "learning_rate": 8.796126401630989e-05, "loss": 2.7938, "step": 15353 }, { "epoch": 1.2, "learning_rate": 8.79604798870854e-05, "loss": 3.2697, "step": 15354 }, { "epoch": 1.2, "learning_rate": 8.79596957578609e-05, "loss": 3.0745, "step": 15355 }, { "epoch": 1.2, "learning_rate": 8.79589116286364e-05, "loss": 2.9068, "step": 15356 }, { "epoch": 1.2, "learning_rate": 8.795812749941191e-05, "loss": 2.75, "step": 15357 }, { "epoch": 1.2, "learning_rate": 8.795734337018741e-05, "loss": 2.939, "step": 15358 }, { "epoch": 1.2, "learning_rate": 8.795655924096292e-05, "loss": 2.9672, "step": 15359 }, { "epoch": 1.2, "learning_rate": 8.795577511173842e-05, "loss": 2.6806, "step": 15360 }, { "epoch": 1.2, "learning_rate": 8.795499098251392e-05, "loss": 2.7071, "step": 15361 }, { "epoch": 1.2, "learning_rate": 8.795420685328943e-05, "loss": 2.9666, "step": 15362 }, { "epoch": 1.2, "learning_rate": 8.795342272406493e-05, "loss": 3.2951, "step": 15363 }, { "epoch": 1.2, "learning_rate": 8.795263859484043e-05, "loss": 2.8595, "step": 15364 }, { "epoch": 1.2, "learning_rate": 8.795185446561594e-05, "loss": 2.9596, "step": 15365 }, { "epoch": 1.2, "learning_rate": 8.795107033639144e-05, "loss": 2.7303, "step": 15366 }, { "epoch": 1.2, "learning_rate": 8.795028620716694e-05, "loss": 3.0543, "step": 15367 }, { "epoch": 1.21, "learning_rate": 8.794950207794245e-05, "loss": 2.8139, "step": 15368 }, { "epoch": 1.21, "learning_rate": 8.794871794871795e-05, "loss": 2.8578, "step": 15369 }, { "epoch": 1.21, "learning_rate": 8.794793381949346e-05, "loss": 2.8866, "step": 15370 }, { "epoch": 1.21, "learning_rate": 8.794714969026896e-05, "loss": 3.0066, "step": 15371 }, { "epoch": 1.21, "learning_rate": 8.794636556104446e-05, "loss": 2.9332, "step": 15372 }, { "epoch": 1.21, "learning_rate": 8.794558143181997e-05, "loss": 2.9656, "step": 15373 }, { "epoch": 1.21, "learning_rate": 8.794479730259547e-05, "loss": 2.748, "step": 15374 }, { "epoch": 1.21, "learning_rate": 8.794401317337097e-05, "loss": 2.9285, "step": 15375 }, { "epoch": 1.21, "learning_rate": 8.794322904414648e-05, "loss": 2.8569, "step": 15376 }, { "epoch": 1.21, "learning_rate": 8.794244491492198e-05, "loss": 2.7922, "step": 15377 }, { "epoch": 1.21, "learning_rate": 8.794166078569748e-05, "loss": 2.6053, "step": 15378 }, { "epoch": 1.21, "learning_rate": 8.794087665647299e-05, "loss": 2.9208, "step": 15379 }, { "epoch": 1.21, "learning_rate": 8.794009252724849e-05, "loss": 3.248, "step": 15380 }, { "epoch": 1.21, "learning_rate": 8.7939308398024e-05, "loss": 2.734, "step": 15381 }, { "epoch": 1.21, "learning_rate": 8.79385242687995e-05, "loss": 3.0786, "step": 15382 }, { "epoch": 1.21, "learning_rate": 8.7937740139575e-05, "loss": 3.0957, "step": 15383 }, { "epoch": 1.21, "learning_rate": 8.793695601035051e-05, "loss": 2.8868, "step": 15384 }, { "epoch": 1.21, "learning_rate": 8.793617188112601e-05, "loss": 3.0255, "step": 15385 }, { "epoch": 1.21, "learning_rate": 8.793538775190153e-05, "loss": 2.9836, "step": 15386 }, { "epoch": 1.21, "learning_rate": 8.793460362267703e-05, "loss": 2.9177, "step": 15387 }, { "epoch": 1.21, "learning_rate": 8.793381949345254e-05, "loss": 2.998, "step": 15388 }, { "epoch": 1.21, "learning_rate": 8.793303536422804e-05, "loss": 2.8219, "step": 15389 }, { "epoch": 1.21, "learning_rate": 8.793225123500354e-05, "loss": 2.8786, "step": 15390 }, { "epoch": 1.21, "learning_rate": 8.793146710577905e-05, "loss": 2.8022, "step": 15391 }, { "epoch": 1.21, "learning_rate": 8.793068297655455e-05, "loss": 2.8866, "step": 15392 }, { "epoch": 1.21, "learning_rate": 8.792989884733004e-05, "loss": 2.9262, "step": 15393 }, { "epoch": 1.21, "learning_rate": 8.792911471810554e-05, "loss": 2.8644, "step": 15394 }, { "epoch": 1.21, "learning_rate": 8.792833058888105e-05, "loss": 2.9058, "step": 15395 }, { "epoch": 1.21, "learning_rate": 8.792754645965655e-05, "loss": 2.9614, "step": 15396 }, { "epoch": 1.21, "learning_rate": 8.792676233043205e-05, "loss": 2.9396, "step": 15397 }, { "epoch": 1.21, "learning_rate": 8.792597820120756e-05, "loss": 2.8929, "step": 15398 }, { "epoch": 1.21, "learning_rate": 8.792519407198306e-05, "loss": 2.9757, "step": 15399 }, { "epoch": 1.21, "learning_rate": 8.792440994275857e-05, "loss": 2.9516, "step": 15400 }, { "epoch": 1.21, "learning_rate": 8.792362581353407e-05, "loss": 2.8301, "step": 15401 }, { "epoch": 1.21, "learning_rate": 8.792284168430957e-05, "loss": 2.5744, "step": 15402 }, { "epoch": 1.21, "learning_rate": 8.792205755508508e-05, "loss": 2.8775, "step": 15403 }, { "epoch": 1.21, "learning_rate": 8.792127342586058e-05, "loss": 2.9438, "step": 15404 }, { "epoch": 1.21, "learning_rate": 8.792048929663608e-05, "loss": 2.9078, "step": 15405 }, { "epoch": 1.21, "learning_rate": 8.791970516741159e-05, "loss": 2.7415, "step": 15406 }, { "epoch": 1.21, "learning_rate": 8.791892103818709e-05, "loss": 2.8887, "step": 15407 }, { "epoch": 1.21, "learning_rate": 8.79181369089626e-05, "loss": 2.8431, "step": 15408 }, { "epoch": 1.21, "learning_rate": 8.79173527797381e-05, "loss": 3.1124, "step": 15409 }, { "epoch": 1.21, "learning_rate": 8.79165686505136e-05, "loss": 2.8211, "step": 15410 }, { "epoch": 1.21, "learning_rate": 8.79157845212891e-05, "loss": 2.9679, "step": 15411 }, { "epoch": 1.21, "learning_rate": 8.791500039206462e-05, "loss": 3.232, "step": 15412 }, { "epoch": 1.21, "learning_rate": 8.791421626284013e-05, "loss": 2.996, "step": 15413 }, { "epoch": 1.21, "learning_rate": 8.791343213361563e-05, "loss": 2.8374, "step": 15414 }, { "epoch": 1.21, "learning_rate": 8.791264800439113e-05, "loss": 2.9852, "step": 15415 }, { "epoch": 1.21, "learning_rate": 8.791186387516664e-05, "loss": 3.0536, "step": 15416 }, { "epoch": 1.21, "learning_rate": 8.791107974594214e-05, "loss": 2.8898, "step": 15417 }, { "epoch": 1.21, "learning_rate": 8.791029561671765e-05, "loss": 2.8364, "step": 15418 }, { "epoch": 1.21, "learning_rate": 8.790951148749315e-05, "loss": 2.8421, "step": 15419 }, { "epoch": 1.21, "learning_rate": 8.790872735826865e-05, "loss": 2.9248, "step": 15420 }, { "epoch": 1.21, "learning_rate": 8.790794322904416e-05, "loss": 3.0232, "step": 15421 }, { "epoch": 1.21, "learning_rate": 8.790715909981966e-05, "loss": 3.0383, "step": 15422 }, { "epoch": 1.21, "learning_rate": 8.790637497059516e-05, "loss": 2.9182, "step": 15423 }, { "epoch": 1.21, "learning_rate": 8.790559084137067e-05, "loss": 2.9091, "step": 15424 }, { "epoch": 1.21, "learning_rate": 8.790480671214617e-05, "loss": 2.7333, "step": 15425 }, { "epoch": 1.21, "learning_rate": 8.790402258292167e-05, "loss": 2.9012, "step": 15426 }, { "epoch": 1.21, "learning_rate": 8.790323845369718e-05, "loss": 2.8479, "step": 15427 }, { "epoch": 1.21, "learning_rate": 8.790245432447268e-05, "loss": 2.9944, "step": 15428 }, { "epoch": 1.21, "learning_rate": 8.790167019524817e-05, "loss": 2.7231, "step": 15429 }, { "epoch": 1.21, "learning_rate": 8.790088606602368e-05, "loss": 2.7726, "step": 15430 }, { "epoch": 1.21, "learning_rate": 8.790010193679918e-05, "loss": 3.2486, "step": 15431 }, { "epoch": 1.21, "learning_rate": 8.789931780757468e-05, "loss": 3.0851, "step": 15432 }, { "epoch": 1.21, "learning_rate": 8.789853367835019e-05, "loss": 2.8462, "step": 15433 }, { "epoch": 1.21, "learning_rate": 8.789774954912569e-05, "loss": 2.7738, "step": 15434 }, { "epoch": 1.21, "learning_rate": 8.78969654199012e-05, "loss": 2.8403, "step": 15435 }, { "epoch": 1.21, "learning_rate": 8.78961812906767e-05, "loss": 2.8457, "step": 15436 }, { "epoch": 1.21, "learning_rate": 8.789539716145222e-05, "loss": 2.7423, "step": 15437 }, { "epoch": 1.21, "learning_rate": 8.789461303222772e-05, "loss": 2.8319, "step": 15438 }, { "epoch": 1.21, "learning_rate": 8.789382890300322e-05, "loss": 3.1966, "step": 15439 }, { "epoch": 1.21, "learning_rate": 8.789304477377873e-05, "loss": 2.7587, "step": 15440 }, { "epoch": 1.21, "learning_rate": 8.789226064455423e-05, "loss": 2.721, "step": 15441 }, { "epoch": 1.21, "learning_rate": 8.789147651532973e-05, "loss": 2.8068, "step": 15442 }, { "epoch": 1.21, "learning_rate": 8.789069238610524e-05, "loss": 2.9384, "step": 15443 }, { "epoch": 1.21, "learning_rate": 8.788990825688074e-05, "loss": 3.187, "step": 15444 }, { "epoch": 1.21, "learning_rate": 8.788912412765624e-05, "loss": 2.9522, "step": 15445 }, { "epoch": 1.21, "learning_rate": 8.788833999843175e-05, "loss": 2.9046, "step": 15446 }, { "epoch": 1.21, "learning_rate": 8.788755586920725e-05, "loss": 2.7836, "step": 15447 }, { "epoch": 1.21, "learning_rate": 8.788677173998276e-05, "loss": 2.7341, "step": 15448 }, { "epoch": 1.21, "learning_rate": 8.788598761075826e-05, "loss": 2.7915, "step": 15449 }, { "epoch": 1.21, "learning_rate": 8.788520348153376e-05, "loss": 3.1056, "step": 15450 }, { "epoch": 1.21, "learning_rate": 8.788441935230927e-05, "loss": 2.9099, "step": 15451 }, { "epoch": 1.21, "learning_rate": 8.788363522308477e-05, "loss": 3.105, "step": 15452 }, { "epoch": 1.21, "learning_rate": 8.788285109386027e-05, "loss": 2.9111, "step": 15453 }, { "epoch": 1.21, "learning_rate": 8.788206696463578e-05, "loss": 2.6615, "step": 15454 }, { "epoch": 1.21, "learning_rate": 8.788128283541128e-05, "loss": 2.9088, "step": 15455 }, { "epoch": 1.21, "learning_rate": 8.788049870618679e-05, "loss": 3.1122, "step": 15456 }, { "epoch": 1.21, "learning_rate": 8.787971457696229e-05, "loss": 2.7419, "step": 15457 }, { "epoch": 1.21, "learning_rate": 8.787893044773779e-05, "loss": 2.7521, "step": 15458 }, { "epoch": 1.21, "learning_rate": 8.78781463185133e-05, "loss": 2.8817, "step": 15459 }, { "epoch": 1.21, "learning_rate": 8.78773621892888e-05, "loss": 2.8025, "step": 15460 }, { "epoch": 1.21, "learning_rate": 8.78765780600643e-05, "loss": 2.8025, "step": 15461 }, { "epoch": 1.21, "learning_rate": 8.787579393083981e-05, "loss": 2.8196, "step": 15462 }, { "epoch": 1.21, "learning_rate": 8.787500980161531e-05, "loss": 2.9682, "step": 15463 }, { "epoch": 1.21, "learning_rate": 8.787422567239081e-05, "loss": 3.0515, "step": 15464 }, { "epoch": 1.21, "learning_rate": 8.787344154316632e-05, "loss": 2.8162, "step": 15465 }, { "epoch": 1.21, "learning_rate": 8.787265741394182e-05, "loss": 3.0667, "step": 15466 }, { "epoch": 1.21, "learning_rate": 8.787187328471733e-05, "loss": 2.9362, "step": 15467 }, { "epoch": 1.21, "learning_rate": 8.787108915549283e-05, "loss": 2.9474, "step": 15468 }, { "epoch": 1.21, "learning_rate": 8.787030502626833e-05, "loss": 3.0424, "step": 15469 }, { "epoch": 1.21, "learning_rate": 8.786952089704384e-05, "loss": 2.8063, "step": 15470 }, { "epoch": 1.21, "learning_rate": 8.786873676781934e-05, "loss": 3.0806, "step": 15471 }, { "epoch": 1.21, "learning_rate": 8.786795263859484e-05, "loss": 3.1559, "step": 15472 }, { "epoch": 1.21, "learning_rate": 8.786716850937035e-05, "loss": 2.8492, "step": 15473 }, { "epoch": 1.21, "learning_rate": 8.786638438014585e-05, "loss": 2.9869, "step": 15474 }, { "epoch": 1.21, "learning_rate": 8.786560025092136e-05, "loss": 2.9274, "step": 15475 }, { "epoch": 1.21, "learning_rate": 8.786481612169686e-05, "loss": 2.8335, "step": 15476 }, { "epoch": 1.21, "learning_rate": 8.786403199247236e-05, "loss": 2.9458, "step": 15477 }, { "epoch": 1.21, "learning_rate": 8.786324786324787e-05, "loss": 2.9919, "step": 15478 }, { "epoch": 1.21, "learning_rate": 8.786246373402337e-05, "loss": 2.7664, "step": 15479 }, { "epoch": 1.21, "learning_rate": 8.786167960479887e-05, "loss": 2.7339, "step": 15480 }, { "epoch": 1.21, "learning_rate": 8.786089547557438e-05, "loss": 3.1021, "step": 15481 }, { "epoch": 1.21, "learning_rate": 8.786011134634988e-05, "loss": 3.0387, "step": 15482 }, { "epoch": 1.21, "learning_rate": 8.785932721712538e-05, "loss": 2.9784, "step": 15483 }, { "epoch": 1.21, "learning_rate": 8.785854308790089e-05, "loss": 2.9774, "step": 15484 }, { "epoch": 1.21, "learning_rate": 8.785775895867639e-05, "loss": 2.7957, "step": 15485 }, { "epoch": 1.21, "learning_rate": 8.78569748294519e-05, "loss": 2.9496, "step": 15486 }, { "epoch": 1.21, "learning_rate": 8.78561907002274e-05, "loss": 3.0415, "step": 15487 }, { "epoch": 1.21, "learning_rate": 8.78554065710029e-05, "loss": 2.8566, "step": 15488 }, { "epoch": 1.21, "learning_rate": 8.78546224417784e-05, "loss": 2.8449, "step": 15489 }, { "epoch": 1.21, "learning_rate": 8.785383831255391e-05, "loss": 3.0751, "step": 15490 }, { "epoch": 1.21, "learning_rate": 8.785305418332941e-05, "loss": 2.8162, "step": 15491 }, { "epoch": 1.21, "learning_rate": 8.785227005410492e-05, "loss": 2.878, "step": 15492 }, { "epoch": 1.21, "learning_rate": 8.785148592488042e-05, "loss": 3.0316, "step": 15493 }, { "epoch": 1.21, "learning_rate": 8.785070179565594e-05, "loss": 2.9717, "step": 15494 }, { "epoch": 1.22, "learning_rate": 8.784991766643144e-05, "loss": 2.8046, "step": 15495 }, { "epoch": 1.22, "learning_rate": 8.784913353720695e-05, "loss": 2.9211, "step": 15496 }, { "epoch": 1.22, "learning_rate": 8.784834940798245e-05, "loss": 3.0866, "step": 15497 }, { "epoch": 1.22, "learning_rate": 8.784756527875795e-05, "loss": 3.0339, "step": 15498 }, { "epoch": 1.22, "learning_rate": 8.784678114953346e-05, "loss": 2.8501, "step": 15499 }, { "epoch": 1.22, "learning_rate": 8.784599702030896e-05, "loss": 2.8653, "step": 15500 }, { "epoch": 1.22, "learning_rate": 8.784521289108445e-05, "loss": 2.9925, "step": 15501 }, { "epoch": 1.22, "learning_rate": 8.784442876185995e-05, "loss": 2.8837, "step": 15502 }, { "epoch": 1.22, "learning_rate": 8.784364463263546e-05, "loss": 2.9433, "step": 15503 }, { "epoch": 1.22, "learning_rate": 8.784286050341096e-05, "loss": 2.9263, "step": 15504 }, { "epoch": 1.22, "learning_rate": 8.784207637418647e-05, "loss": 2.6771, "step": 15505 }, { "epoch": 1.22, "learning_rate": 8.784129224496197e-05, "loss": 3.1044, "step": 15506 }, { "epoch": 1.22, "learning_rate": 8.784050811573747e-05, "loss": 2.8049, "step": 15507 }, { "epoch": 1.22, "learning_rate": 8.783972398651298e-05, "loss": 2.9452, "step": 15508 }, { "epoch": 1.22, "learning_rate": 8.783893985728848e-05, "loss": 2.8314, "step": 15509 }, { "epoch": 1.22, "learning_rate": 8.783815572806398e-05, "loss": 2.7438, "step": 15510 }, { "epoch": 1.22, "learning_rate": 8.783737159883949e-05, "loss": 2.8743, "step": 15511 }, { "epoch": 1.22, "learning_rate": 8.783658746961499e-05, "loss": 2.9371, "step": 15512 }, { "epoch": 1.22, "learning_rate": 8.78358033403905e-05, "loss": 2.943, "step": 15513 }, { "epoch": 1.22, "learning_rate": 8.7835019211166e-05, "loss": 2.9242, "step": 15514 }, { "epoch": 1.22, "learning_rate": 8.78342350819415e-05, "loss": 2.8894, "step": 15515 }, { "epoch": 1.22, "learning_rate": 8.7833450952717e-05, "loss": 3.1867, "step": 15516 }, { "epoch": 1.22, "learning_rate": 8.783266682349251e-05, "loss": 2.7748, "step": 15517 }, { "epoch": 1.22, "learning_rate": 8.783188269426801e-05, "loss": 2.852, "step": 15518 }, { "epoch": 1.22, "learning_rate": 8.783109856504352e-05, "loss": 2.9716, "step": 15519 }, { "epoch": 1.22, "learning_rate": 8.783031443581903e-05, "loss": 3.0595, "step": 15520 }, { "epoch": 1.22, "learning_rate": 8.782953030659454e-05, "loss": 2.8845, "step": 15521 }, { "epoch": 1.22, "learning_rate": 8.782874617737004e-05, "loss": 2.9576, "step": 15522 }, { "epoch": 1.22, "learning_rate": 8.782796204814555e-05, "loss": 2.739, "step": 15523 }, { "epoch": 1.22, "learning_rate": 8.782717791892105e-05, "loss": 3.0165, "step": 15524 }, { "epoch": 1.22, "learning_rate": 8.782639378969655e-05, "loss": 2.9523, "step": 15525 }, { "epoch": 1.22, "learning_rate": 8.782560966047206e-05, "loss": 2.7733, "step": 15526 }, { "epoch": 1.22, "learning_rate": 8.782482553124756e-05, "loss": 2.8973, "step": 15527 }, { "epoch": 1.22, "learning_rate": 8.782404140202306e-05, "loss": 2.9588, "step": 15528 }, { "epoch": 1.22, "learning_rate": 8.782325727279857e-05, "loss": 3.0177, "step": 15529 }, { "epoch": 1.22, "learning_rate": 8.782247314357407e-05, "loss": 2.7861, "step": 15530 }, { "epoch": 1.22, "learning_rate": 8.782168901434957e-05, "loss": 2.8754, "step": 15531 }, { "epoch": 1.22, "learning_rate": 8.782090488512508e-05, "loss": 2.8815, "step": 15532 }, { "epoch": 1.22, "learning_rate": 8.782012075590058e-05, "loss": 2.9076, "step": 15533 }, { "epoch": 1.22, "learning_rate": 8.781933662667609e-05, "loss": 3.0117, "step": 15534 }, { "epoch": 1.22, "learning_rate": 8.781855249745159e-05, "loss": 3.1484, "step": 15535 }, { "epoch": 1.22, "learning_rate": 8.781776836822709e-05, "loss": 2.7799, "step": 15536 }, { "epoch": 1.22, "learning_rate": 8.781698423900258e-05, "loss": 2.65, "step": 15537 }, { "epoch": 1.22, "learning_rate": 8.781620010977809e-05, "loss": 2.7949, "step": 15538 }, { "epoch": 1.22, "learning_rate": 8.781541598055359e-05, "loss": 3.1208, "step": 15539 }, { "epoch": 1.22, "learning_rate": 8.78146318513291e-05, "loss": 2.9444, "step": 15540 }, { "epoch": 1.22, "learning_rate": 8.78138477221046e-05, "loss": 2.8531, "step": 15541 }, { "epoch": 1.22, "learning_rate": 8.78130635928801e-05, "loss": 2.8571, "step": 15542 }, { "epoch": 1.22, "learning_rate": 8.78122794636556e-05, "loss": 2.8642, "step": 15543 }, { "epoch": 1.22, "learning_rate": 8.781149533443111e-05, "loss": 2.9853, "step": 15544 }, { "epoch": 1.22, "learning_rate": 8.781071120520663e-05, "loss": 2.8443, "step": 15545 }, { "epoch": 1.22, "learning_rate": 8.780992707598213e-05, "loss": 2.6736, "step": 15546 }, { "epoch": 1.22, "learning_rate": 8.780914294675763e-05, "loss": 3.0121, "step": 15547 }, { "epoch": 1.22, "learning_rate": 8.780835881753314e-05, "loss": 2.6904, "step": 15548 }, { "epoch": 1.22, "learning_rate": 8.780757468830864e-05, "loss": 3.0293, "step": 15549 }, { "epoch": 1.22, "learning_rate": 8.780679055908414e-05, "loss": 3.0233, "step": 15550 }, { "epoch": 1.22, "learning_rate": 8.780600642985965e-05, "loss": 2.8712, "step": 15551 }, { "epoch": 1.22, "learning_rate": 8.780522230063515e-05, "loss": 3.1239, "step": 15552 }, { "epoch": 1.22, "learning_rate": 8.780443817141066e-05, "loss": 2.7357, "step": 15553 }, { "epoch": 1.22, "learning_rate": 8.780365404218616e-05, "loss": 2.673, "step": 15554 }, { "epoch": 1.22, "learning_rate": 8.780286991296166e-05, "loss": 2.8225, "step": 15555 }, { "epoch": 1.22, "learning_rate": 8.780208578373717e-05, "loss": 2.8036, "step": 15556 }, { "epoch": 1.22, "learning_rate": 8.780130165451267e-05, "loss": 2.8751, "step": 15557 }, { "epoch": 1.22, "learning_rate": 8.780051752528817e-05, "loss": 2.7475, "step": 15558 }, { "epoch": 1.22, "learning_rate": 8.779973339606368e-05, "loss": 2.7995, "step": 15559 }, { "epoch": 1.22, "learning_rate": 8.779894926683918e-05, "loss": 3.0126, "step": 15560 }, { "epoch": 1.22, "learning_rate": 8.779816513761469e-05, "loss": 2.8417, "step": 15561 }, { "epoch": 1.22, "learning_rate": 8.779738100839019e-05, "loss": 2.8849, "step": 15562 }, { "epoch": 1.22, "learning_rate": 8.779659687916569e-05, "loss": 3.1226, "step": 15563 }, { "epoch": 1.22, "learning_rate": 8.77958127499412e-05, "loss": 3.117, "step": 15564 }, { "epoch": 1.22, "learning_rate": 8.77950286207167e-05, "loss": 2.8883, "step": 15565 }, { "epoch": 1.22, "learning_rate": 8.77942444914922e-05, "loss": 2.5853, "step": 15566 }, { "epoch": 1.22, "learning_rate": 8.779346036226771e-05, "loss": 2.7942, "step": 15567 }, { "epoch": 1.22, "learning_rate": 8.779267623304321e-05, "loss": 2.7425, "step": 15568 }, { "epoch": 1.22, "learning_rate": 8.779189210381871e-05, "loss": 2.826, "step": 15569 }, { "epoch": 1.22, "learning_rate": 8.779110797459422e-05, "loss": 2.781, "step": 15570 }, { "epoch": 1.22, "learning_rate": 8.779032384536972e-05, "loss": 2.9462, "step": 15571 }, { "epoch": 1.22, "learning_rate": 8.778953971614523e-05, "loss": 2.7643, "step": 15572 }, { "epoch": 1.22, "learning_rate": 8.778875558692073e-05, "loss": 2.8978, "step": 15573 }, { "epoch": 1.22, "learning_rate": 8.778797145769623e-05, "loss": 3.0737, "step": 15574 }, { "epoch": 1.22, "learning_rate": 8.778718732847174e-05, "loss": 2.8181, "step": 15575 }, { "epoch": 1.22, "learning_rate": 8.778640319924724e-05, "loss": 3.1289, "step": 15576 }, { "epoch": 1.22, "learning_rate": 8.778561907002274e-05, "loss": 2.9406, "step": 15577 }, { "epoch": 1.22, "learning_rate": 8.778483494079825e-05, "loss": 3.1013, "step": 15578 }, { "epoch": 1.22, "learning_rate": 8.778405081157375e-05, "loss": 2.6851, "step": 15579 }, { "epoch": 1.22, "learning_rate": 8.778326668234925e-05, "loss": 2.947, "step": 15580 }, { "epoch": 1.22, "learning_rate": 8.778248255312476e-05, "loss": 2.9142, "step": 15581 }, { "epoch": 1.22, "learning_rate": 8.778169842390026e-05, "loss": 3.0318, "step": 15582 }, { "epoch": 1.22, "learning_rate": 8.778091429467577e-05, "loss": 3.0392, "step": 15583 }, { "epoch": 1.22, "learning_rate": 8.778013016545127e-05, "loss": 2.9171, "step": 15584 }, { "epoch": 1.22, "learning_rate": 8.777934603622677e-05, "loss": 2.8463, "step": 15585 }, { "epoch": 1.22, "learning_rate": 8.777856190700228e-05, "loss": 2.9848, "step": 15586 }, { "epoch": 1.22, "learning_rate": 8.777777777777778e-05, "loss": 2.9872, "step": 15587 }, { "epoch": 1.22, "learning_rate": 8.777699364855328e-05, "loss": 2.9664, "step": 15588 }, { "epoch": 1.22, "learning_rate": 8.777620951932879e-05, "loss": 2.8279, "step": 15589 }, { "epoch": 1.22, "learning_rate": 8.777542539010429e-05, "loss": 2.9885, "step": 15590 }, { "epoch": 1.22, "learning_rate": 8.77746412608798e-05, "loss": 2.9872, "step": 15591 }, { "epoch": 1.22, "learning_rate": 8.77738571316553e-05, "loss": 2.8284, "step": 15592 }, { "epoch": 1.22, "learning_rate": 8.77730730024308e-05, "loss": 2.7183, "step": 15593 }, { "epoch": 1.22, "learning_rate": 8.77722888732063e-05, "loss": 3.0841, "step": 15594 }, { "epoch": 1.22, "learning_rate": 8.777150474398181e-05, "loss": 3.1742, "step": 15595 }, { "epoch": 1.22, "learning_rate": 8.777072061475731e-05, "loss": 3.2058, "step": 15596 }, { "epoch": 1.22, "learning_rate": 8.776993648553282e-05, "loss": 2.9789, "step": 15597 }, { "epoch": 1.22, "learning_rate": 8.776915235630832e-05, "loss": 2.8013, "step": 15598 }, { "epoch": 1.22, "learning_rate": 8.776836822708382e-05, "loss": 2.7402, "step": 15599 }, { "epoch": 1.22, "learning_rate": 8.776758409785933e-05, "loss": 3.1051, "step": 15600 }, { "epoch": 1.22, "learning_rate": 8.776679996863483e-05, "loss": 2.7941, "step": 15601 }, { "epoch": 1.22, "learning_rate": 8.776601583941035e-05, "loss": 2.8518, "step": 15602 }, { "epoch": 1.22, "learning_rate": 8.776523171018585e-05, "loss": 3.0433, "step": 15603 }, { "epoch": 1.22, "learning_rate": 8.776444758096136e-05, "loss": 2.8579, "step": 15604 }, { "epoch": 1.22, "learning_rate": 8.776366345173686e-05, "loss": 3.0054, "step": 15605 }, { "epoch": 1.22, "learning_rate": 8.776287932251236e-05, "loss": 2.9938, "step": 15606 }, { "epoch": 1.22, "learning_rate": 8.776209519328787e-05, "loss": 2.8385, "step": 15607 }, { "epoch": 1.22, "learning_rate": 8.776131106406337e-05, "loss": 3.2559, "step": 15608 }, { "epoch": 1.22, "learning_rate": 8.776052693483886e-05, "loss": 2.6611, "step": 15609 }, { "epoch": 1.22, "learning_rate": 8.775974280561437e-05, "loss": 2.7674, "step": 15610 }, { "epoch": 1.22, "learning_rate": 8.775895867638987e-05, "loss": 2.9055, "step": 15611 }, { "epoch": 1.22, "learning_rate": 8.775817454716537e-05, "loss": 2.8422, "step": 15612 }, { "epoch": 1.22, "learning_rate": 8.775739041794088e-05, "loss": 2.9028, "step": 15613 }, { "epoch": 1.22, "learning_rate": 8.775660628871638e-05, "loss": 2.8782, "step": 15614 }, { "epoch": 1.22, "learning_rate": 8.775582215949188e-05, "loss": 3.0379, "step": 15615 }, { "epoch": 1.22, "learning_rate": 8.775503803026739e-05, "loss": 2.9517, "step": 15616 }, { "epoch": 1.22, "learning_rate": 8.775425390104289e-05, "loss": 2.8391, "step": 15617 }, { "epoch": 1.22, "learning_rate": 8.77534697718184e-05, "loss": 2.9091, "step": 15618 }, { "epoch": 1.22, "learning_rate": 8.77526856425939e-05, "loss": 2.88, "step": 15619 }, { "epoch": 1.22, "learning_rate": 8.77519015133694e-05, "loss": 2.7804, "step": 15620 }, { "epoch": 1.22, "learning_rate": 8.77511173841449e-05, "loss": 2.9833, "step": 15621 }, { "epoch": 1.22, "learning_rate": 8.775033325492041e-05, "loss": 2.8117, "step": 15622 }, { "epoch": 1.23, "learning_rate": 8.774954912569591e-05, "loss": 2.7687, "step": 15623 }, { "epoch": 1.23, "learning_rate": 8.774876499647142e-05, "loss": 2.8258, "step": 15624 }, { "epoch": 1.23, "learning_rate": 8.774798086724692e-05, "loss": 2.9942, "step": 15625 }, { "epoch": 1.23, "learning_rate": 8.774719673802242e-05, "loss": 2.7778, "step": 15626 }, { "epoch": 1.23, "learning_rate": 8.774641260879793e-05, "loss": 3.0087, "step": 15627 }, { "epoch": 1.23, "learning_rate": 8.774562847957344e-05, "loss": 3.1495, "step": 15628 }, { "epoch": 1.23, "learning_rate": 8.774484435034895e-05, "loss": 3.0062, "step": 15629 }, { "epoch": 1.23, "learning_rate": 8.774406022112445e-05, "loss": 2.7995, "step": 15630 }, { "epoch": 1.23, "learning_rate": 8.774327609189996e-05, "loss": 2.9659, "step": 15631 }, { "epoch": 1.23, "learning_rate": 8.774249196267546e-05, "loss": 2.9103, "step": 15632 }, { "epoch": 1.23, "learning_rate": 8.774170783345096e-05, "loss": 3.2019, "step": 15633 }, { "epoch": 1.23, "learning_rate": 8.774092370422647e-05, "loss": 2.8896, "step": 15634 }, { "epoch": 1.23, "learning_rate": 8.774013957500197e-05, "loss": 3.0243, "step": 15635 }, { "epoch": 1.23, "learning_rate": 8.773935544577747e-05, "loss": 3.0156, "step": 15636 }, { "epoch": 1.23, "learning_rate": 8.773857131655298e-05, "loss": 2.6763, "step": 15637 }, { "epoch": 1.23, "learning_rate": 8.773778718732848e-05, "loss": 2.8104, "step": 15638 }, { "epoch": 1.23, "learning_rate": 8.773700305810399e-05, "loss": 2.7996, "step": 15639 }, { "epoch": 1.23, "learning_rate": 8.773621892887949e-05, "loss": 2.7795, "step": 15640 }, { "epoch": 1.23, "learning_rate": 8.773543479965499e-05, "loss": 2.8293, "step": 15641 }, { "epoch": 1.23, "learning_rate": 8.77346506704305e-05, "loss": 2.7692, "step": 15642 }, { "epoch": 1.23, "learning_rate": 8.7733866541206e-05, "loss": 2.81, "step": 15643 }, { "epoch": 1.23, "learning_rate": 8.77330824119815e-05, "loss": 2.8485, "step": 15644 }, { "epoch": 1.23, "learning_rate": 8.7732298282757e-05, "loss": 2.9491, "step": 15645 }, { "epoch": 1.23, "learning_rate": 8.77315141535325e-05, "loss": 2.911, "step": 15646 }, { "epoch": 1.23, "learning_rate": 8.7730730024308e-05, "loss": 3.0051, "step": 15647 }, { "epoch": 1.23, "learning_rate": 8.77299458950835e-05, "loss": 2.9951, "step": 15648 }, { "epoch": 1.23, "learning_rate": 8.772916176585901e-05, "loss": 2.9211, "step": 15649 }, { "epoch": 1.23, "learning_rate": 8.772837763663451e-05, "loss": 2.8751, "step": 15650 }, { "epoch": 1.23, "learning_rate": 8.772759350741002e-05, "loss": 3.133, "step": 15651 }, { "epoch": 1.23, "learning_rate": 8.772680937818552e-05, "loss": 2.8675, "step": 15652 }, { "epoch": 1.23, "learning_rate": 8.772602524896104e-05, "loss": 2.9692, "step": 15653 }, { "epoch": 1.23, "learning_rate": 8.772524111973654e-05, "loss": 2.995, "step": 15654 }, { "epoch": 1.23, "learning_rate": 8.772445699051204e-05, "loss": 2.8523, "step": 15655 }, { "epoch": 1.23, "learning_rate": 8.772367286128755e-05, "loss": 2.7901, "step": 15656 }, { "epoch": 1.23, "learning_rate": 8.772288873206305e-05, "loss": 2.7973, "step": 15657 }, { "epoch": 1.23, "learning_rate": 8.772210460283856e-05, "loss": 2.9916, "step": 15658 }, { "epoch": 1.23, "learning_rate": 8.772132047361406e-05, "loss": 2.8585, "step": 15659 }, { "epoch": 1.23, "learning_rate": 8.772053634438956e-05, "loss": 3.0323, "step": 15660 }, { "epoch": 1.23, "learning_rate": 8.771975221516507e-05, "loss": 2.9998, "step": 15661 }, { "epoch": 1.23, "learning_rate": 8.771896808594057e-05, "loss": 2.7423, "step": 15662 }, { "epoch": 1.23, "learning_rate": 8.771818395671607e-05, "loss": 2.8273, "step": 15663 }, { "epoch": 1.23, "learning_rate": 8.771739982749158e-05, "loss": 3.1131, "step": 15664 }, { "epoch": 1.23, "learning_rate": 8.771661569826708e-05, "loss": 2.9229, "step": 15665 }, { "epoch": 1.23, "learning_rate": 8.771583156904258e-05, "loss": 3.0511, "step": 15666 }, { "epoch": 1.23, "learning_rate": 8.771504743981809e-05, "loss": 2.968, "step": 15667 }, { "epoch": 1.23, "learning_rate": 8.771426331059359e-05, "loss": 2.9113, "step": 15668 }, { "epoch": 1.23, "learning_rate": 8.77134791813691e-05, "loss": 2.8159, "step": 15669 }, { "epoch": 1.23, "learning_rate": 8.77126950521446e-05, "loss": 2.8945, "step": 15670 }, { "epoch": 1.23, "learning_rate": 8.77119109229201e-05, "loss": 2.8333, "step": 15671 }, { "epoch": 1.23, "learning_rate": 8.771112679369561e-05, "loss": 3.1078, "step": 15672 }, { "epoch": 1.23, "learning_rate": 8.771034266447111e-05, "loss": 2.9747, "step": 15673 }, { "epoch": 1.23, "learning_rate": 8.770955853524661e-05, "loss": 2.9037, "step": 15674 }, { "epoch": 1.23, "learning_rate": 8.770877440602212e-05, "loss": 2.9418, "step": 15675 }, { "epoch": 1.23, "learning_rate": 8.770799027679762e-05, "loss": 2.9852, "step": 15676 }, { "epoch": 1.23, "learning_rate": 8.770720614757313e-05, "loss": 2.8413, "step": 15677 }, { "epoch": 1.23, "learning_rate": 8.770642201834863e-05, "loss": 2.8688, "step": 15678 }, { "epoch": 1.23, "learning_rate": 8.770563788912413e-05, "loss": 3.0051, "step": 15679 }, { "epoch": 1.23, "learning_rate": 8.770485375989964e-05, "loss": 3.0521, "step": 15680 }, { "epoch": 1.23, "learning_rate": 8.770406963067514e-05, "loss": 2.8738, "step": 15681 }, { "epoch": 1.23, "learning_rate": 8.770328550145064e-05, "loss": 2.8985, "step": 15682 }, { "epoch": 1.23, "learning_rate": 8.770250137222615e-05, "loss": 2.742, "step": 15683 }, { "epoch": 1.23, "learning_rate": 8.770171724300165e-05, "loss": 2.8579, "step": 15684 }, { "epoch": 1.23, "learning_rate": 8.770093311377715e-05, "loss": 2.8986, "step": 15685 }, { "epoch": 1.23, "learning_rate": 8.770014898455266e-05, "loss": 2.8366, "step": 15686 }, { "epoch": 1.23, "learning_rate": 8.769936485532816e-05, "loss": 2.9713, "step": 15687 }, { "epoch": 1.23, "learning_rate": 8.769858072610367e-05, "loss": 2.9883, "step": 15688 }, { "epoch": 1.23, "learning_rate": 8.769779659687917e-05, "loss": 2.8622, "step": 15689 }, { "epoch": 1.23, "learning_rate": 8.769701246765467e-05, "loss": 2.9746, "step": 15690 }, { "epoch": 1.23, "learning_rate": 8.769622833843018e-05, "loss": 2.8372, "step": 15691 }, { "epoch": 1.23, "learning_rate": 8.769544420920568e-05, "loss": 2.7651, "step": 15692 }, { "epoch": 1.23, "learning_rate": 8.769466007998118e-05, "loss": 2.9239, "step": 15693 }, { "epoch": 1.23, "learning_rate": 8.769387595075669e-05, "loss": 2.8528, "step": 15694 }, { "epoch": 1.23, "learning_rate": 8.769309182153219e-05, "loss": 2.9287, "step": 15695 }, { "epoch": 1.23, "learning_rate": 8.76923076923077e-05, "loss": 2.9141, "step": 15696 }, { "epoch": 1.23, "learning_rate": 8.76915235630832e-05, "loss": 3.1035, "step": 15697 }, { "epoch": 1.23, "learning_rate": 8.76907394338587e-05, "loss": 2.8777, "step": 15698 }, { "epoch": 1.23, "learning_rate": 8.76899553046342e-05, "loss": 2.8761, "step": 15699 }, { "epoch": 1.23, "learning_rate": 8.768917117540971e-05, "loss": 2.9391, "step": 15700 }, { "epoch": 1.23, "learning_rate": 8.768838704618521e-05, "loss": 3.0564, "step": 15701 }, { "epoch": 1.23, "learning_rate": 8.768760291696072e-05, "loss": 2.7152, "step": 15702 }, { "epoch": 1.23, "learning_rate": 8.768681878773622e-05, "loss": 2.9063, "step": 15703 }, { "epoch": 1.23, "learning_rate": 8.768603465851172e-05, "loss": 2.9964, "step": 15704 }, { "epoch": 1.23, "learning_rate": 8.768525052928723e-05, "loss": 2.8025, "step": 15705 }, { "epoch": 1.23, "learning_rate": 8.768446640006273e-05, "loss": 2.9655, "step": 15706 }, { "epoch": 1.23, "learning_rate": 8.768368227083824e-05, "loss": 2.6997, "step": 15707 }, { "epoch": 1.23, "learning_rate": 8.768289814161374e-05, "loss": 3.4127, "step": 15708 }, { "epoch": 1.23, "learning_rate": 8.768211401238924e-05, "loss": 3.016, "step": 15709 }, { "epoch": 1.23, "learning_rate": 8.768132988316476e-05, "loss": 2.8365, "step": 15710 }, { "epoch": 1.23, "learning_rate": 8.768054575394026e-05, "loss": 3.0815, "step": 15711 }, { "epoch": 1.23, "learning_rate": 8.767976162471577e-05, "loss": 2.8866, "step": 15712 }, { "epoch": 1.23, "learning_rate": 8.767897749549127e-05, "loss": 2.8568, "step": 15713 }, { "epoch": 1.23, "learning_rate": 8.767819336626677e-05, "loss": 2.9597, "step": 15714 }, { "epoch": 1.23, "learning_rate": 8.767740923704228e-05, "loss": 3.0094, "step": 15715 }, { "epoch": 1.23, "learning_rate": 8.767662510781778e-05, "loss": 2.7746, "step": 15716 }, { "epoch": 1.23, "learning_rate": 8.767584097859327e-05, "loss": 2.7615, "step": 15717 }, { "epoch": 1.23, "learning_rate": 8.767505684936878e-05, "loss": 2.888, "step": 15718 }, { "epoch": 1.23, "learning_rate": 8.767427272014428e-05, "loss": 2.9049, "step": 15719 }, { "epoch": 1.23, "learning_rate": 8.767348859091978e-05, "loss": 2.7126, "step": 15720 }, { "epoch": 1.23, "learning_rate": 8.767270446169529e-05, "loss": 2.8287, "step": 15721 }, { "epoch": 1.23, "learning_rate": 8.767192033247079e-05, "loss": 2.7159, "step": 15722 }, { "epoch": 1.23, "learning_rate": 8.76711362032463e-05, "loss": 2.8913, "step": 15723 }, { "epoch": 1.23, "learning_rate": 8.76703520740218e-05, "loss": 2.9556, "step": 15724 }, { "epoch": 1.23, "learning_rate": 8.76695679447973e-05, "loss": 2.7525, "step": 15725 }, { "epoch": 1.23, "learning_rate": 8.76687838155728e-05, "loss": 3.0125, "step": 15726 }, { "epoch": 1.23, "learning_rate": 8.766799968634831e-05, "loss": 2.9865, "step": 15727 }, { "epoch": 1.23, "learning_rate": 8.766721555712381e-05, "loss": 3.0579, "step": 15728 }, { "epoch": 1.23, "learning_rate": 8.766643142789932e-05, "loss": 2.8116, "step": 15729 }, { "epoch": 1.23, "learning_rate": 8.766564729867482e-05, "loss": 2.832, "step": 15730 }, { "epoch": 1.23, "learning_rate": 8.766486316945032e-05, "loss": 2.9106, "step": 15731 }, { "epoch": 1.23, "learning_rate": 8.766407904022583e-05, "loss": 2.9353, "step": 15732 }, { "epoch": 1.23, "learning_rate": 8.766329491100133e-05, "loss": 2.7667, "step": 15733 }, { "epoch": 1.23, "learning_rate": 8.766251078177683e-05, "loss": 2.9445, "step": 15734 }, { "epoch": 1.23, "learning_rate": 8.766172665255234e-05, "loss": 2.9297, "step": 15735 }, { "epoch": 1.23, "learning_rate": 8.766094252332786e-05, "loss": 2.9344, "step": 15736 }, { "epoch": 1.23, "learning_rate": 8.766015839410336e-05, "loss": 2.8635, "step": 15737 }, { "epoch": 1.23, "learning_rate": 8.765937426487886e-05, "loss": 2.9835, "step": 15738 }, { "epoch": 1.23, "learning_rate": 8.765859013565437e-05, "loss": 3.014, "step": 15739 }, { "epoch": 1.23, "learning_rate": 8.765780600642987e-05, "loss": 2.7949, "step": 15740 }, { "epoch": 1.23, "learning_rate": 8.765702187720537e-05, "loss": 2.8174, "step": 15741 }, { "epoch": 1.23, "learning_rate": 8.765623774798088e-05, "loss": 2.9255, "step": 15742 }, { "epoch": 1.23, "learning_rate": 8.765545361875638e-05, "loss": 2.7192, "step": 15743 }, { "epoch": 1.23, "learning_rate": 8.765466948953189e-05, "loss": 2.8653, "step": 15744 }, { "epoch": 1.23, "learning_rate": 8.765388536030739e-05, "loss": 2.7243, "step": 15745 }, { "epoch": 1.23, "learning_rate": 8.765310123108289e-05, "loss": 2.9652, "step": 15746 }, { "epoch": 1.23, "learning_rate": 8.76523171018584e-05, "loss": 2.9838, "step": 15747 }, { "epoch": 1.23, "learning_rate": 8.76515329726339e-05, "loss": 2.8641, "step": 15748 }, { "epoch": 1.23, "learning_rate": 8.76507488434094e-05, "loss": 2.7985, "step": 15749 }, { "epoch": 1.24, "learning_rate": 8.764996471418491e-05, "loss": 2.9816, "step": 15750 }, { "epoch": 1.24, "learning_rate": 8.764918058496041e-05, "loss": 2.8025, "step": 15751 }, { "epoch": 1.24, "learning_rate": 8.764839645573591e-05, "loss": 2.7894, "step": 15752 }, { "epoch": 1.24, "learning_rate": 8.76476123265114e-05, "loss": 2.9447, "step": 15753 }, { "epoch": 1.24, "learning_rate": 8.764682819728691e-05, "loss": 2.8775, "step": 15754 }, { "epoch": 1.24, "learning_rate": 8.764604406806241e-05, "loss": 2.7969, "step": 15755 }, { "epoch": 1.24, "learning_rate": 8.764525993883792e-05, "loss": 2.8408, "step": 15756 }, { "epoch": 1.24, "learning_rate": 8.764447580961342e-05, "loss": 2.7723, "step": 15757 }, { "epoch": 1.24, "learning_rate": 8.764369168038892e-05, "loss": 2.7855, "step": 15758 }, { "epoch": 1.24, "learning_rate": 8.764290755116443e-05, "loss": 2.9939, "step": 15759 }, { "epoch": 1.24, "learning_rate": 8.764212342193993e-05, "loss": 3.028, "step": 15760 }, { "epoch": 1.24, "learning_rate": 8.764133929271543e-05, "loss": 2.9455, "step": 15761 }, { "epoch": 1.24, "learning_rate": 8.764055516349095e-05, "loss": 3.0866, "step": 15762 }, { "epoch": 1.24, "learning_rate": 8.763977103426646e-05, "loss": 3.0045, "step": 15763 }, { "epoch": 1.24, "learning_rate": 8.763898690504196e-05, "loss": 2.8978, "step": 15764 }, { "epoch": 1.24, "learning_rate": 8.763820277581746e-05, "loss": 2.8306, "step": 15765 }, { "epoch": 1.24, "learning_rate": 8.763741864659297e-05, "loss": 3.0163, "step": 15766 }, { "epoch": 1.24, "learning_rate": 8.763663451736847e-05, "loss": 2.8898, "step": 15767 }, { "epoch": 1.24, "learning_rate": 8.763585038814397e-05, "loss": 2.9368, "step": 15768 }, { "epoch": 1.24, "learning_rate": 8.763506625891948e-05, "loss": 2.6882, "step": 15769 }, { "epoch": 1.24, "learning_rate": 8.763428212969498e-05, "loss": 2.9642, "step": 15770 }, { "epoch": 1.24, "learning_rate": 8.763349800047048e-05, "loss": 2.9088, "step": 15771 }, { "epoch": 1.24, "learning_rate": 8.763271387124599e-05, "loss": 2.7989, "step": 15772 }, { "epoch": 1.24, "learning_rate": 8.763192974202149e-05, "loss": 2.9216, "step": 15773 }, { "epoch": 1.24, "learning_rate": 8.7631145612797e-05, "loss": 2.9643, "step": 15774 }, { "epoch": 1.24, "learning_rate": 8.76303614835725e-05, "loss": 2.7436, "step": 15775 }, { "epoch": 1.24, "learning_rate": 8.7629577354348e-05, "loss": 2.8207, "step": 15776 }, { "epoch": 1.24, "learning_rate": 8.76287932251235e-05, "loss": 2.8136, "step": 15777 }, { "epoch": 1.24, "learning_rate": 8.762800909589901e-05, "loss": 2.872, "step": 15778 }, { "epoch": 1.24, "learning_rate": 8.762722496667451e-05, "loss": 2.9622, "step": 15779 }, { "epoch": 1.24, "learning_rate": 8.762644083745002e-05, "loss": 2.8647, "step": 15780 }, { "epoch": 1.24, "learning_rate": 8.762565670822552e-05, "loss": 2.7704, "step": 15781 }, { "epoch": 1.24, "learning_rate": 8.762487257900102e-05, "loss": 2.9183, "step": 15782 }, { "epoch": 1.24, "learning_rate": 8.762408844977653e-05, "loss": 2.9219, "step": 15783 }, { "epoch": 1.24, "learning_rate": 8.762330432055203e-05, "loss": 2.9356, "step": 15784 }, { "epoch": 1.24, "learning_rate": 8.762252019132754e-05, "loss": 2.8806, "step": 15785 }, { "epoch": 1.24, "learning_rate": 8.762173606210304e-05, "loss": 2.9687, "step": 15786 }, { "epoch": 1.24, "learning_rate": 8.762095193287854e-05, "loss": 2.9684, "step": 15787 }, { "epoch": 1.24, "learning_rate": 8.762016780365405e-05, "loss": 2.9802, "step": 15788 }, { "epoch": 1.24, "learning_rate": 8.761938367442955e-05, "loss": 2.9773, "step": 15789 }, { "epoch": 1.24, "learning_rate": 8.761859954520505e-05, "loss": 2.614, "step": 15790 }, { "epoch": 1.24, "learning_rate": 8.761781541598056e-05, "loss": 3.0835, "step": 15791 }, { "epoch": 1.24, "learning_rate": 8.761703128675606e-05, "loss": 2.8894, "step": 15792 }, { "epoch": 1.24, "learning_rate": 8.761624715753157e-05, "loss": 2.7387, "step": 15793 }, { "epoch": 1.24, "learning_rate": 8.761546302830707e-05, "loss": 2.7987, "step": 15794 }, { "epoch": 1.24, "learning_rate": 8.761467889908257e-05, "loss": 2.8609, "step": 15795 }, { "epoch": 1.24, "learning_rate": 8.761389476985808e-05, "loss": 2.9807, "step": 15796 }, { "epoch": 1.24, "learning_rate": 8.761311064063358e-05, "loss": 2.9891, "step": 15797 }, { "epoch": 1.24, "learning_rate": 8.761232651140908e-05, "loss": 3.1126, "step": 15798 }, { "epoch": 1.24, "learning_rate": 8.761154238218459e-05, "loss": 2.9483, "step": 15799 }, { "epoch": 1.24, "learning_rate": 8.761075825296009e-05, "loss": 3.054, "step": 15800 }, { "epoch": 1.24, "learning_rate": 8.76099741237356e-05, "loss": 2.7494, "step": 15801 }, { "epoch": 1.24, "learning_rate": 8.76091899945111e-05, "loss": 2.9164, "step": 15802 }, { "epoch": 1.24, "learning_rate": 8.76084058652866e-05, "loss": 2.7212, "step": 15803 }, { "epoch": 1.24, "learning_rate": 8.76076217360621e-05, "loss": 3.1103, "step": 15804 }, { "epoch": 1.24, "learning_rate": 8.760683760683761e-05, "loss": 2.8802, "step": 15805 }, { "epoch": 1.24, "learning_rate": 8.760605347761311e-05, "loss": 3.1484, "step": 15806 }, { "epoch": 1.24, "learning_rate": 8.760526934838862e-05, "loss": 3.054, "step": 15807 }, { "epoch": 1.24, "learning_rate": 8.760448521916412e-05, "loss": 3.0334, "step": 15808 }, { "epoch": 1.24, "learning_rate": 8.760370108993962e-05, "loss": 3.0596, "step": 15809 }, { "epoch": 1.24, "learning_rate": 8.760291696071513e-05, "loss": 2.8605, "step": 15810 }, { "epoch": 1.24, "learning_rate": 8.760213283149063e-05, "loss": 2.9455, "step": 15811 }, { "epoch": 1.24, "learning_rate": 8.760134870226614e-05, "loss": 3.167, "step": 15812 }, { "epoch": 1.24, "learning_rate": 8.760056457304164e-05, "loss": 2.901, "step": 15813 }, { "epoch": 1.24, "learning_rate": 8.759978044381714e-05, "loss": 3.0245, "step": 15814 }, { "epoch": 1.24, "learning_rate": 8.759899631459265e-05, "loss": 3.0074, "step": 15815 }, { "epoch": 1.24, "learning_rate": 8.759821218536815e-05, "loss": 2.927, "step": 15816 }, { "epoch": 1.24, "learning_rate": 8.759742805614365e-05, "loss": 2.9159, "step": 15817 }, { "epoch": 1.24, "learning_rate": 8.759664392691917e-05, "loss": 3.006, "step": 15818 }, { "epoch": 1.24, "learning_rate": 8.759585979769467e-05, "loss": 2.7807, "step": 15819 }, { "epoch": 1.24, "learning_rate": 8.759507566847018e-05, "loss": 2.5833, "step": 15820 }, { "epoch": 1.24, "learning_rate": 8.759429153924568e-05, "loss": 2.8453, "step": 15821 }, { "epoch": 1.24, "learning_rate": 8.759350741002119e-05, "loss": 2.7225, "step": 15822 }, { "epoch": 1.24, "learning_rate": 8.759272328079669e-05, "loss": 2.7999, "step": 15823 }, { "epoch": 1.24, "learning_rate": 8.759193915157219e-05, "loss": 2.7003, "step": 15824 }, { "epoch": 1.24, "learning_rate": 8.759115502234768e-05, "loss": 3.0494, "step": 15825 }, { "epoch": 1.24, "learning_rate": 8.759037089312319e-05, "loss": 2.6252, "step": 15826 }, { "epoch": 1.24, "learning_rate": 8.758958676389869e-05, "loss": 2.9477, "step": 15827 }, { "epoch": 1.24, "learning_rate": 8.75888026346742e-05, "loss": 2.9051, "step": 15828 }, { "epoch": 1.24, "learning_rate": 8.75880185054497e-05, "loss": 3.0032, "step": 15829 }, { "epoch": 1.24, "learning_rate": 8.75872343762252e-05, "loss": 2.9214, "step": 15830 }, { "epoch": 1.24, "learning_rate": 8.75864502470007e-05, "loss": 2.915, "step": 15831 }, { "epoch": 1.24, "learning_rate": 8.758566611777621e-05, "loss": 3.0069, "step": 15832 }, { "epoch": 1.24, "learning_rate": 8.758488198855171e-05, "loss": 2.9517, "step": 15833 }, { "epoch": 1.24, "learning_rate": 8.758409785932722e-05, "loss": 2.7826, "step": 15834 }, { "epoch": 1.24, "learning_rate": 8.758331373010272e-05, "loss": 3.0159, "step": 15835 }, { "epoch": 1.24, "learning_rate": 8.758252960087822e-05, "loss": 2.9674, "step": 15836 }, { "epoch": 1.24, "learning_rate": 8.758174547165373e-05, "loss": 2.6665, "step": 15837 }, { "epoch": 1.24, "learning_rate": 8.758096134242923e-05, "loss": 2.7283, "step": 15838 }, { "epoch": 1.24, "learning_rate": 8.758017721320473e-05, "loss": 3.2157, "step": 15839 }, { "epoch": 1.24, "learning_rate": 8.757939308398024e-05, "loss": 2.8639, "step": 15840 }, { "epoch": 1.24, "learning_rate": 8.757860895475574e-05, "loss": 2.8961, "step": 15841 }, { "epoch": 1.24, "learning_rate": 8.757782482553125e-05, "loss": 3.0408, "step": 15842 }, { "epoch": 1.24, "learning_rate": 8.757704069630675e-05, "loss": 2.9033, "step": 15843 }, { "epoch": 1.24, "learning_rate": 8.757625656708227e-05, "loss": 2.9313, "step": 15844 }, { "epoch": 1.24, "learning_rate": 8.757547243785777e-05, "loss": 2.8404, "step": 15845 }, { "epoch": 1.24, "learning_rate": 8.757468830863327e-05, "loss": 2.9702, "step": 15846 }, { "epoch": 1.24, "learning_rate": 8.757390417940878e-05, "loss": 2.9824, "step": 15847 }, { "epoch": 1.24, "learning_rate": 8.757312005018428e-05, "loss": 3.0807, "step": 15848 }, { "epoch": 1.24, "learning_rate": 8.757233592095978e-05, "loss": 2.813, "step": 15849 }, { "epoch": 1.24, "learning_rate": 8.757155179173529e-05, "loss": 3.002, "step": 15850 }, { "epoch": 1.24, "learning_rate": 8.757076766251079e-05, "loss": 2.8321, "step": 15851 }, { "epoch": 1.24, "learning_rate": 8.75699835332863e-05, "loss": 3.0732, "step": 15852 }, { "epoch": 1.24, "learning_rate": 8.75691994040618e-05, "loss": 2.9881, "step": 15853 }, { "epoch": 1.24, "learning_rate": 8.75684152748373e-05, "loss": 2.8797, "step": 15854 }, { "epoch": 1.24, "learning_rate": 8.756763114561281e-05, "loss": 2.8599, "step": 15855 }, { "epoch": 1.24, "learning_rate": 8.756684701638831e-05, "loss": 2.9928, "step": 15856 }, { "epoch": 1.24, "learning_rate": 8.756606288716381e-05, "loss": 2.933, "step": 15857 }, { "epoch": 1.24, "learning_rate": 8.756527875793932e-05, "loss": 2.8091, "step": 15858 }, { "epoch": 1.24, "learning_rate": 8.756449462871482e-05, "loss": 3.2034, "step": 15859 }, { "epoch": 1.24, "learning_rate": 8.756371049949033e-05, "loss": 2.8552, "step": 15860 }, { "epoch": 1.24, "learning_rate": 8.756292637026582e-05, "loss": 2.7807, "step": 15861 }, { "epoch": 1.24, "learning_rate": 8.756214224104132e-05, "loss": 2.799, "step": 15862 }, { "epoch": 1.24, "learning_rate": 8.756135811181682e-05, "loss": 2.6962, "step": 15863 }, { "epoch": 1.24, "learning_rate": 8.756057398259233e-05, "loss": 2.8386, "step": 15864 }, { "epoch": 1.24, "learning_rate": 8.755978985336783e-05, "loss": 2.9466, "step": 15865 }, { "epoch": 1.24, "learning_rate": 8.755900572414333e-05, "loss": 2.9022, "step": 15866 }, { "epoch": 1.24, "learning_rate": 8.755822159491884e-05, "loss": 3.092, "step": 15867 }, { "epoch": 1.24, "learning_rate": 8.755743746569434e-05, "loss": 3.0505, "step": 15868 }, { "epoch": 1.24, "learning_rate": 8.755665333646984e-05, "loss": 2.8305, "step": 15869 }, { "epoch": 1.24, "learning_rate": 8.755586920724536e-05, "loss": 3.1642, "step": 15870 }, { "epoch": 1.24, "learning_rate": 8.755508507802087e-05, "loss": 2.7839, "step": 15871 }, { "epoch": 1.24, "learning_rate": 8.755430094879637e-05, "loss": 2.8809, "step": 15872 }, { "epoch": 1.24, "learning_rate": 8.755351681957187e-05, "loss": 2.6916, "step": 15873 }, { "epoch": 1.24, "learning_rate": 8.755273269034738e-05, "loss": 3.3333, "step": 15874 }, { "epoch": 1.24, "learning_rate": 8.755194856112288e-05, "loss": 2.9307, "step": 15875 }, { "epoch": 1.24, "learning_rate": 8.755116443189838e-05, "loss": 2.8691, "step": 15876 }, { "epoch": 1.24, "learning_rate": 8.755038030267389e-05, "loss": 2.9384, "step": 15877 }, { "epoch": 1.25, "learning_rate": 8.754959617344939e-05, "loss": 2.8059, "step": 15878 }, { "epoch": 1.25, "learning_rate": 8.75488120442249e-05, "loss": 2.9821, "step": 15879 }, { "epoch": 1.25, "learning_rate": 8.75480279150004e-05, "loss": 3.047, "step": 15880 }, { "epoch": 1.25, "learning_rate": 8.75472437857759e-05, "loss": 2.7871, "step": 15881 }, { "epoch": 1.25, "learning_rate": 8.75464596565514e-05, "loss": 2.7431, "step": 15882 }, { "epoch": 1.25, "learning_rate": 8.754567552732691e-05, "loss": 2.9676, "step": 15883 }, { "epoch": 1.25, "learning_rate": 8.754489139810241e-05, "loss": 3.1829, "step": 15884 }, { "epoch": 1.25, "learning_rate": 8.754410726887792e-05, "loss": 2.7136, "step": 15885 }, { "epoch": 1.25, "learning_rate": 8.754332313965342e-05, "loss": 2.8074, "step": 15886 }, { "epoch": 1.25, "learning_rate": 8.754253901042892e-05, "loss": 3.0472, "step": 15887 }, { "epoch": 1.25, "learning_rate": 8.754175488120443e-05, "loss": 2.9799, "step": 15888 }, { "epoch": 1.25, "learning_rate": 8.754097075197993e-05, "loss": 2.9319, "step": 15889 }, { "epoch": 1.25, "learning_rate": 8.754018662275544e-05, "loss": 2.8702, "step": 15890 }, { "epoch": 1.25, "learning_rate": 8.753940249353094e-05, "loss": 2.9713, "step": 15891 }, { "epoch": 1.25, "learning_rate": 8.753861836430644e-05, "loss": 2.7728, "step": 15892 }, { "epoch": 1.25, "learning_rate": 8.753783423508195e-05, "loss": 3.1965, "step": 15893 }, { "epoch": 1.25, "learning_rate": 8.753705010585745e-05, "loss": 2.6032, "step": 15894 }, { "epoch": 1.25, "learning_rate": 8.753626597663295e-05, "loss": 2.8424, "step": 15895 }, { "epoch": 1.25, "learning_rate": 8.753548184740846e-05, "loss": 2.8775, "step": 15896 }, { "epoch": 1.25, "learning_rate": 8.753469771818396e-05, "loss": 2.9315, "step": 15897 }, { "epoch": 1.25, "learning_rate": 8.753391358895947e-05, "loss": 3.1034, "step": 15898 }, { "epoch": 1.25, "learning_rate": 8.753312945973497e-05, "loss": 2.9711, "step": 15899 }, { "epoch": 1.25, "learning_rate": 8.753234533051047e-05, "loss": 2.7979, "step": 15900 }, { "epoch": 1.25, "learning_rate": 8.753156120128598e-05, "loss": 2.8069, "step": 15901 }, { "epoch": 1.25, "learning_rate": 8.753077707206148e-05, "loss": 3.0871, "step": 15902 }, { "epoch": 1.25, "learning_rate": 8.752999294283698e-05, "loss": 2.837, "step": 15903 }, { "epoch": 1.25, "learning_rate": 8.752920881361249e-05, "loss": 3.1577, "step": 15904 }, { "epoch": 1.25, "learning_rate": 8.752842468438799e-05, "loss": 2.9829, "step": 15905 }, { "epoch": 1.25, "learning_rate": 8.75276405551635e-05, "loss": 2.9392, "step": 15906 }, { "epoch": 1.25, "learning_rate": 8.7526856425939e-05, "loss": 2.8038, "step": 15907 }, { "epoch": 1.25, "learning_rate": 8.75260722967145e-05, "loss": 2.9729, "step": 15908 }, { "epoch": 1.25, "learning_rate": 8.752528816749e-05, "loss": 2.7348, "step": 15909 }, { "epoch": 1.25, "learning_rate": 8.752450403826551e-05, "loss": 2.895, "step": 15910 }, { "epoch": 1.25, "learning_rate": 8.752371990904101e-05, "loss": 3.0676, "step": 15911 }, { "epoch": 1.25, "learning_rate": 8.752293577981652e-05, "loss": 2.9113, "step": 15912 }, { "epoch": 1.25, "learning_rate": 8.752215165059202e-05, "loss": 2.9839, "step": 15913 }, { "epoch": 1.25, "learning_rate": 8.752136752136752e-05, "loss": 2.7048, "step": 15914 }, { "epoch": 1.25, "learning_rate": 8.752058339214303e-05, "loss": 2.8696, "step": 15915 }, { "epoch": 1.25, "learning_rate": 8.751979926291853e-05, "loss": 2.917, "step": 15916 }, { "epoch": 1.25, "learning_rate": 8.751901513369403e-05, "loss": 2.7877, "step": 15917 }, { "epoch": 1.25, "learning_rate": 8.751823100446954e-05, "loss": 2.7926, "step": 15918 }, { "epoch": 1.25, "learning_rate": 8.751744687524504e-05, "loss": 2.8751, "step": 15919 }, { "epoch": 1.25, "learning_rate": 8.751666274602055e-05, "loss": 2.9016, "step": 15920 }, { "epoch": 1.25, "learning_rate": 8.751587861679605e-05, "loss": 2.7661, "step": 15921 }, { "epoch": 1.25, "learning_rate": 8.751509448757155e-05, "loss": 2.7045, "step": 15922 }, { "epoch": 1.25, "learning_rate": 8.751431035834706e-05, "loss": 3.2229, "step": 15923 }, { "epoch": 1.25, "learning_rate": 8.751352622912256e-05, "loss": 2.8069, "step": 15924 }, { "epoch": 1.25, "learning_rate": 8.751274209989806e-05, "loss": 2.7958, "step": 15925 }, { "epoch": 1.25, "learning_rate": 8.751195797067358e-05, "loss": 2.9528, "step": 15926 }, { "epoch": 1.25, "learning_rate": 8.751117384144909e-05, "loss": 2.7011, "step": 15927 }, { "epoch": 1.25, "learning_rate": 8.751038971222459e-05, "loss": 2.7553, "step": 15928 }, { "epoch": 1.25, "learning_rate": 8.750960558300009e-05, "loss": 2.754, "step": 15929 }, { "epoch": 1.25, "learning_rate": 8.75088214537756e-05, "loss": 2.9895, "step": 15930 }, { "epoch": 1.25, "learning_rate": 8.75080373245511e-05, "loss": 2.8482, "step": 15931 }, { "epoch": 1.25, "learning_rate": 8.75072531953266e-05, "loss": 3.0868, "step": 15932 }, { "epoch": 1.25, "learning_rate": 8.75064690661021e-05, "loss": 3.1011, "step": 15933 }, { "epoch": 1.25, "learning_rate": 8.75056849368776e-05, "loss": 3.0935, "step": 15934 }, { "epoch": 1.25, "learning_rate": 8.75049008076531e-05, "loss": 2.817, "step": 15935 }, { "epoch": 1.25, "learning_rate": 8.75041166784286e-05, "loss": 2.9594, "step": 15936 }, { "epoch": 1.25, "learning_rate": 8.750333254920411e-05, "loss": 3.0263, "step": 15937 }, { "epoch": 1.25, "learning_rate": 8.750254841997961e-05, "loss": 2.8541, "step": 15938 }, { "epoch": 1.25, "learning_rate": 8.750176429075512e-05, "loss": 2.7746, "step": 15939 }, { "epoch": 1.25, "learning_rate": 8.750098016153062e-05, "loss": 2.6419, "step": 15940 }, { "epoch": 1.25, "learning_rate": 8.750019603230612e-05, "loss": 2.934, "step": 15941 }, { "epoch": 1.25, "learning_rate": 8.749941190308163e-05, "loss": 3.1039, "step": 15942 }, { "epoch": 1.25, "learning_rate": 8.749862777385713e-05, "loss": 2.7254, "step": 15943 }, { "epoch": 1.25, "learning_rate": 8.749784364463263e-05, "loss": 3.1732, "step": 15944 }, { "epoch": 1.25, "learning_rate": 8.749705951540814e-05, "loss": 2.6434, "step": 15945 }, { "epoch": 1.25, "learning_rate": 8.749627538618364e-05, "loss": 2.989, "step": 15946 }, { "epoch": 1.25, "learning_rate": 8.749549125695915e-05, "loss": 2.8151, "step": 15947 }, { "epoch": 1.25, "learning_rate": 8.749470712773465e-05, "loss": 2.792, "step": 15948 }, { "epoch": 1.25, "learning_rate": 8.749392299851015e-05, "loss": 2.8003, "step": 15949 }, { "epoch": 1.25, "learning_rate": 8.749313886928566e-05, "loss": 2.8309, "step": 15950 }, { "epoch": 1.25, "learning_rate": 8.749235474006116e-05, "loss": 2.8857, "step": 15951 }, { "epoch": 1.25, "learning_rate": 8.749157061083668e-05, "loss": 2.7661, "step": 15952 }, { "epoch": 1.25, "learning_rate": 8.749078648161218e-05, "loss": 2.8365, "step": 15953 }, { "epoch": 1.25, "learning_rate": 8.749000235238768e-05, "loss": 2.7377, "step": 15954 }, { "epoch": 1.25, "learning_rate": 8.748921822316319e-05, "loss": 3.0588, "step": 15955 }, { "epoch": 1.25, "learning_rate": 8.748843409393869e-05, "loss": 3.1215, "step": 15956 }, { "epoch": 1.25, "learning_rate": 8.74876499647142e-05, "loss": 3.0273, "step": 15957 }, { "epoch": 1.25, "learning_rate": 8.74868658354897e-05, "loss": 3.0983, "step": 15958 }, { "epoch": 1.25, "learning_rate": 8.74860817062652e-05, "loss": 2.7857, "step": 15959 }, { "epoch": 1.25, "learning_rate": 8.74852975770407e-05, "loss": 2.8586, "step": 15960 }, { "epoch": 1.25, "learning_rate": 8.748451344781621e-05, "loss": 2.8909, "step": 15961 }, { "epoch": 1.25, "learning_rate": 8.748372931859171e-05, "loss": 2.8731, "step": 15962 }, { "epoch": 1.25, "learning_rate": 8.748294518936722e-05, "loss": 3.0362, "step": 15963 }, { "epoch": 1.25, "learning_rate": 8.748216106014272e-05, "loss": 2.8082, "step": 15964 }, { "epoch": 1.25, "learning_rate": 8.748137693091823e-05, "loss": 2.9734, "step": 15965 }, { "epoch": 1.25, "learning_rate": 8.748059280169373e-05, "loss": 2.9791, "step": 15966 }, { "epoch": 1.25, "learning_rate": 8.747980867246923e-05, "loss": 3.1943, "step": 15967 }, { "epoch": 1.25, "learning_rate": 8.747902454324474e-05, "loss": 2.974, "step": 15968 }, { "epoch": 1.25, "learning_rate": 8.747824041402023e-05, "loss": 3.0332, "step": 15969 }, { "epoch": 1.25, "learning_rate": 8.747745628479573e-05, "loss": 2.9802, "step": 15970 }, { "epoch": 1.25, "learning_rate": 8.747667215557123e-05, "loss": 2.6038, "step": 15971 }, { "epoch": 1.25, "learning_rate": 8.747588802634674e-05, "loss": 3.2742, "step": 15972 }, { "epoch": 1.25, "learning_rate": 8.747510389712224e-05, "loss": 2.9212, "step": 15973 }, { "epoch": 1.25, "learning_rate": 8.747431976789774e-05, "loss": 2.921, "step": 15974 }, { "epoch": 1.25, "learning_rate": 8.747353563867325e-05, "loss": 3.0353, "step": 15975 }, { "epoch": 1.25, "learning_rate": 8.747275150944875e-05, "loss": 3.0661, "step": 15976 }, { "epoch": 1.25, "learning_rate": 8.747196738022426e-05, "loss": 2.9321, "step": 15977 }, { "epoch": 1.25, "learning_rate": 8.747118325099977e-05, "loss": 2.9578, "step": 15978 }, { "epoch": 1.25, "learning_rate": 8.747039912177528e-05, "loss": 2.8875, "step": 15979 }, { "epoch": 1.25, "learning_rate": 8.746961499255078e-05, "loss": 2.7707, "step": 15980 }, { "epoch": 1.25, "learning_rate": 8.746883086332628e-05, "loss": 2.9102, "step": 15981 }, { "epoch": 1.25, "learning_rate": 8.746804673410179e-05, "loss": 2.7827, "step": 15982 }, { "epoch": 1.25, "learning_rate": 8.746726260487729e-05, "loss": 2.8375, "step": 15983 }, { "epoch": 1.25, "learning_rate": 8.74664784756528e-05, "loss": 2.9137, "step": 15984 }, { "epoch": 1.25, "learning_rate": 8.74656943464283e-05, "loss": 2.8149, "step": 15985 }, { "epoch": 1.25, "learning_rate": 8.74649102172038e-05, "loss": 2.887, "step": 15986 }, { "epoch": 1.25, "learning_rate": 8.74641260879793e-05, "loss": 3.0015, "step": 15987 }, { "epoch": 1.25, "learning_rate": 8.746334195875481e-05, "loss": 2.8014, "step": 15988 }, { "epoch": 1.25, "learning_rate": 8.746255782953031e-05, "loss": 2.762, "step": 15989 }, { "epoch": 1.25, "learning_rate": 8.746177370030582e-05, "loss": 2.7455, "step": 15990 }, { "epoch": 1.25, "learning_rate": 8.746098957108132e-05, "loss": 2.968, "step": 15991 }, { "epoch": 1.25, "learning_rate": 8.746020544185682e-05, "loss": 2.9275, "step": 15992 }, { "epoch": 1.25, "learning_rate": 8.745942131263233e-05, "loss": 2.7263, "step": 15993 }, { "epoch": 1.25, "learning_rate": 8.745863718340783e-05, "loss": 3.02, "step": 15994 }, { "epoch": 1.25, "learning_rate": 8.745785305418334e-05, "loss": 2.9485, "step": 15995 }, { "epoch": 1.25, "learning_rate": 8.745706892495884e-05, "loss": 3.3359, "step": 15996 }, { "epoch": 1.25, "learning_rate": 8.745628479573434e-05, "loss": 2.7864, "step": 15997 }, { "epoch": 1.25, "learning_rate": 8.745550066650985e-05, "loss": 2.8713, "step": 15998 }, { "epoch": 1.25, "learning_rate": 8.745471653728535e-05, "loss": 2.8521, "step": 15999 }, { "epoch": 1.25, "learning_rate": 8.745393240806085e-05, "loss": 2.7407, "step": 16000 }, { "epoch": 1.25, "eval_loss": 2.9206058979034424, "eval_runtime": 218.0399, "eval_samples_per_second": 51.972, "eval_steps_per_second": 3.252, "step": 16000 }, { "epoch": 1.25, "learning_rate": 8.745314827883636e-05, "loss": 2.7978, "step": 16001 }, { "epoch": 1.25, "learning_rate": 8.745236414961186e-05, "loss": 2.7753, "step": 16002 }, { "epoch": 1.25, "learning_rate": 8.745158002038736e-05, "loss": 2.9006, "step": 16003 }, { "epoch": 1.25, "learning_rate": 8.745079589116287e-05, "loss": 2.8939, "step": 16004 }, { "epoch": 1.25, "learning_rate": 8.745001176193837e-05, "loss": 2.6125, "step": 16005 }, { "epoch": 1.26, "learning_rate": 8.744922763271388e-05, "loss": 2.9475, "step": 16006 }, { "epoch": 1.26, "learning_rate": 8.744844350348938e-05, "loss": 3.0065, "step": 16007 }, { "epoch": 1.26, "learning_rate": 8.744765937426488e-05, "loss": 2.7499, "step": 16008 }, { "epoch": 1.26, "learning_rate": 8.744687524504039e-05, "loss": 2.7753, "step": 16009 }, { "epoch": 1.26, "learning_rate": 8.744609111581589e-05, "loss": 3.0283, "step": 16010 }, { "epoch": 1.26, "learning_rate": 8.74453069865914e-05, "loss": 3.0033, "step": 16011 }, { "epoch": 1.26, "learning_rate": 8.74445228573669e-05, "loss": 2.8164, "step": 16012 }, { "epoch": 1.26, "learning_rate": 8.74437387281424e-05, "loss": 2.9024, "step": 16013 }, { "epoch": 1.26, "learning_rate": 8.74429545989179e-05, "loss": 2.7474, "step": 16014 }, { "epoch": 1.26, "learning_rate": 8.744217046969341e-05, "loss": 3.0043, "step": 16015 }, { "epoch": 1.26, "learning_rate": 8.744138634046891e-05, "loss": 2.9188, "step": 16016 }, { "epoch": 1.26, "learning_rate": 8.744060221124442e-05, "loss": 2.7744, "step": 16017 }, { "epoch": 1.26, "learning_rate": 8.743981808201992e-05, "loss": 2.8301, "step": 16018 }, { "epoch": 1.26, "learning_rate": 8.743903395279542e-05, "loss": 2.7965, "step": 16019 }, { "epoch": 1.26, "learning_rate": 8.743824982357093e-05, "loss": 2.9277, "step": 16020 }, { "epoch": 1.26, "learning_rate": 8.743746569434643e-05, "loss": 2.8814, "step": 16021 }, { "epoch": 1.26, "learning_rate": 8.743668156512193e-05, "loss": 3.0842, "step": 16022 }, { "epoch": 1.26, "learning_rate": 8.743589743589744e-05, "loss": 3.0658, "step": 16023 }, { "epoch": 1.26, "learning_rate": 8.743511330667294e-05, "loss": 2.7298, "step": 16024 }, { "epoch": 1.26, "learning_rate": 8.743432917744845e-05, "loss": 3.1808, "step": 16025 }, { "epoch": 1.26, "learning_rate": 8.743354504822395e-05, "loss": 3.0254, "step": 16026 }, { "epoch": 1.26, "learning_rate": 8.743276091899945e-05, "loss": 2.8667, "step": 16027 }, { "epoch": 1.26, "learning_rate": 8.743197678977496e-05, "loss": 2.6631, "step": 16028 }, { "epoch": 1.26, "learning_rate": 8.743119266055046e-05, "loss": 2.8967, "step": 16029 }, { "epoch": 1.26, "learning_rate": 8.743040853132596e-05, "loss": 2.7843, "step": 16030 }, { "epoch": 1.26, "learning_rate": 8.742962440210147e-05, "loss": 2.8941, "step": 16031 }, { "epoch": 1.26, "learning_rate": 8.742884027287697e-05, "loss": 3.0287, "step": 16032 }, { "epoch": 1.26, "learning_rate": 8.742805614365248e-05, "loss": 2.7389, "step": 16033 }, { "epoch": 1.26, "learning_rate": 8.742727201442798e-05, "loss": 2.9182, "step": 16034 }, { "epoch": 1.26, "learning_rate": 8.74264878852035e-05, "loss": 3.203, "step": 16035 }, { "epoch": 1.26, "learning_rate": 8.7425703755979e-05, "loss": 3.114, "step": 16036 }, { "epoch": 1.26, "learning_rate": 8.74249196267545e-05, "loss": 2.9523, "step": 16037 }, { "epoch": 1.26, "learning_rate": 8.742413549753001e-05, "loss": 2.983, "step": 16038 }, { "epoch": 1.26, "learning_rate": 8.742335136830551e-05, "loss": 2.9489, "step": 16039 }, { "epoch": 1.26, "learning_rate": 8.742256723908101e-05, "loss": 2.8907, "step": 16040 }, { "epoch": 1.26, "learning_rate": 8.74217831098565e-05, "loss": 2.6753, "step": 16041 }, { "epoch": 1.26, "learning_rate": 8.742099898063201e-05, "loss": 2.859, "step": 16042 }, { "epoch": 1.26, "learning_rate": 8.742021485140751e-05, "loss": 3.1506, "step": 16043 }, { "epoch": 1.26, "learning_rate": 8.741943072218302e-05, "loss": 2.861, "step": 16044 }, { "epoch": 1.26, "learning_rate": 8.741864659295852e-05, "loss": 2.7321, "step": 16045 }, { "epoch": 1.26, "learning_rate": 8.741786246373402e-05, "loss": 2.9693, "step": 16046 }, { "epoch": 1.26, "learning_rate": 8.741707833450953e-05, "loss": 2.9027, "step": 16047 }, { "epoch": 1.26, "learning_rate": 8.741629420528503e-05, "loss": 2.9049, "step": 16048 }, { "epoch": 1.26, "learning_rate": 8.741551007606053e-05, "loss": 3.1764, "step": 16049 }, { "epoch": 1.26, "learning_rate": 8.741472594683604e-05, "loss": 3.0041, "step": 16050 }, { "epoch": 1.26, "learning_rate": 8.741394181761154e-05, "loss": 2.9126, "step": 16051 }, { "epoch": 1.26, "learning_rate": 8.741315768838704e-05, "loss": 3.2203, "step": 16052 }, { "epoch": 1.26, "learning_rate": 8.741237355916255e-05, "loss": 2.7943, "step": 16053 }, { "epoch": 1.26, "learning_rate": 8.741158942993805e-05, "loss": 2.8417, "step": 16054 }, { "epoch": 1.26, "learning_rate": 8.741080530071356e-05, "loss": 2.9875, "step": 16055 }, { "epoch": 1.26, "learning_rate": 8.741002117148906e-05, "loss": 2.894, "step": 16056 }, { "epoch": 1.26, "learning_rate": 8.740923704226456e-05, "loss": 3.0508, "step": 16057 }, { "epoch": 1.26, "learning_rate": 8.740845291304007e-05, "loss": 2.8098, "step": 16058 }, { "epoch": 1.26, "learning_rate": 8.740766878381557e-05, "loss": 2.8427, "step": 16059 }, { "epoch": 1.26, "learning_rate": 8.740688465459109e-05, "loss": 2.9146, "step": 16060 }, { "epoch": 1.26, "learning_rate": 8.740610052536659e-05, "loss": 2.7261, "step": 16061 }, { "epoch": 1.26, "learning_rate": 8.74053163961421e-05, "loss": 2.8874, "step": 16062 }, { "epoch": 1.26, "learning_rate": 8.74045322669176e-05, "loss": 3.0538, "step": 16063 }, { "epoch": 1.26, "learning_rate": 8.74037481376931e-05, "loss": 2.7133, "step": 16064 }, { "epoch": 1.26, "learning_rate": 8.74029640084686e-05, "loss": 2.839, "step": 16065 }, { "epoch": 1.26, "learning_rate": 8.740217987924411e-05, "loss": 2.7773, "step": 16066 }, { "epoch": 1.26, "learning_rate": 8.740139575001961e-05, "loss": 3.0965, "step": 16067 }, { "epoch": 1.26, "learning_rate": 8.740061162079512e-05, "loss": 2.976, "step": 16068 }, { "epoch": 1.26, "learning_rate": 8.739982749157062e-05, "loss": 2.8387, "step": 16069 }, { "epoch": 1.26, "learning_rate": 8.739904336234612e-05, "loss": 2.9984, "step": 16070 }, { "epoch": 1.26, "learning_rate": 8.739825923312163e-05, "loss": 3.1446, "step": 16071 }, { "epoch": 1.26, "learning_rate": 8.739747510389713e-05, "loss": 2.9592, "step": 16072 }, { "epoch": 1.26, "learning_rate": 8.739669097467264e-05, "loss": 2.8368, "step": 16073 }, { "epoch": 1.26, "learning_rate": 8.739590684544814e-05, "loss": 2.8084, "step": 16074 }, { "epoch": 1.26, "learning_rate": 8.739512271622364e-05, "loss": 2.9011, "step": 16075 }, { "epoch": 1.26, "learning_rate": 8.739433858699915e-05, "loss": 2.6721, "step": 16076 }, { "epoch": 1.26, "learning_rate": 8.739355445777464e-05, "loss": 3.0268, "step": 16077 }, { "epoch": 1.26, "learning_rate": 8.739277032855014e-05, "loss": 2.9799, "step": 16078 }, { "epoch": 1.26, "learning_rate": 8.739198619932564e-05, "loss": 3.1364, "step": 16079 }, { "epoch": 1.26, "learning_rate": 8.739120207010115e-05, "loss": 2.9302, "step": 16080 }, { "epoch": 1.26, "learning_rate": 8.739041794087665e-05, "loss": 3.2936, "step": 16081 }, { "epoch": 1.26, "learning_rate": 8.738963381165216e-05, "loss": 2.8514, "step": 16082 }, { "epoch": 1.26, "learning_rate": 8.738884968242766e-05, "loss": 3.1849, "step": 16083 }, { "epoch": 1.26, "learning_rate": 8.738806555320316e-05, "loss": 2.7585, "step": 16084 }, { "epoch": 1.26, "learning_rate": 8.738728142397867e-05, "loss": 2.9684, "step": 16085 }, { "epoch": 1.26, "learning_rate": 8.738649729475418e-05, "loss": 2.9761, "step": 16086 }, { "epoch": 1.26, "learning_rate": 8.738571316552969e-05, "loss": 2.8455, "step": 16087 }, { "epoch": 1.26, "learning_rate": 8.738492903630519e-05, "loss": 2.9083, "step": 16088 }, { "epoch": 1.26, "learning_rate": 8.73841449070807e-05, "loss": 2.8545, "step": 16089 }, { "epoch": 1.26, "learning_rate": 8.73833607778562e-05, "loss": 2.7798, "step": 16090 }, { "epoch": 1.26, "learning_rate": 8.73825766486317e-05, "loss": 2.9695, "step": 16091 }, { "epoch": 1.26, "learning_rate": 8.73817925194072e-05, "loss": 3.2084, "step": 16092 }, { "epoch": 1.26, "learning_rate": 8.738100839018271e-05, "loss": 3.0542, "step": 16093 }, { "epoch": 1.26, "learning_rate": 8.738022426095821e-05, "loss": 2.895, "step": 16094 }, { "epoch": 1.26, "learning_rate": 8.737944013173372e-05, "loss": 2.9911, "step": 16095 }, { "epoch": 1.26, "learning_rate": 8.737865600250922e-05, "loss": 3.0208, "step": 16096 }, { "epoch": 1.26, "learning_rate": 8.737787187328472e-05, "loss": 2.8478, "step": 16097 }, { "epoch": 1.26, "learning_rate": 8.737708774406023e-05, "loss": 2.9992, "step": 16098 }, { "epoch": 1.26, "learning_rate": 8.737630361483573e-05, "loss": 2.9219, "step": 16099 }, { "epoch": 1.26, "learning_rate": 8.737551948561124e-05, "loss": 2.9205, "step": 16100 }, { "epoch": 1.26, "learning_rate": 8.737473535638674e-05, "loss": 2.7905, "step": 16101 }, { "epoch": 1.26, "learning_rate": 8.737395122716224e-05, "loss": 2.9144, "step": 16102 }, { "epoch": 1.26, "learning_rate": 8.737316709793775e-05, "loss": 2.964, "step": 16103 }, { "epoch": 1.26, "learning_rate": 8.737238296871325e-05, "loss": 2.8786, "step": 16104 }, { "epoch": 1.26, "learning_rate": 8.737159883948875e-05, "loss": 2.8494, "step": 16105 }, { "epoch": 1.26, "learning_rate": 8.737081471026426e-05, "loss": 2.9632, "step": 16106 }, { "epoch": 1.26, "learning_rate": 8.737003058103976e-05, "loss": 2.9176, "step": 16107 }, { "epoch": 1.26, "learning_rate": 8.736924645181526e-05, "loss": 2.9171, "step": 16108 }, { "epoch": 1.26, "learning_rate": 8.736846232259077e-05, "loss": 3.2036, "step": 16109 }, { "epoch": 1.26, "learning_rate": 8.736767819336627e-05, "loss": 3.292, "step": 16110 }, { "epoch": 1.26, "learning_rate": 8.736689406414178e-05, "loss": 2.8238, "step": 16111 }, { "epoch": 1.26, "learning_rate": 8.736610993491728e-05, "loss": 2.8376, "step": 16112 }, { "epoch": 1.26, "learning_rate": 8.736532580569278e-05, "loss": 2.8146, "step": 16113 }, { "epoch": 1.26, "learning_rate": 8.736454167646829e-05, "loss": 2.8476, "step": 16114 }, { "epoch": 1.26, "learning_rate": 8.736375754724379e-05, "loss": 3.0286, "step": 16115 }, { "epoch": 1.26, "learning_rate": 8.73629734180193e-05, "loss": 2.9806, "step": 16116 }, { "epoch": 1.26, "learning_rate": 8.73621892887948e-05, "loss": 2.7675, "step": 16117 }, { "epoch": 1.26, "learning_rate": 8.73614051595703e-05, "loss": 2.9462, "step": 16118 }, { "epoch": 1.26, "learning_rate": 8.73606210303458e-05, "loss": 2.7571, "step": 16119 }, { "epoch": 1.26, "learning_rate": 8.735983690112131e-05, "loss": 2.752, "step": 16120 }, { "epoch": 1.26, "learning_rate": 8.735905277189681e-05, "loss": 2.728, "step": 16121 }, { "epoch": 1.26, "learning_rate": 8.735826864267232e-05, "loss": 2.7747, "step": 16122 }, { "epoch": 1.26, "learning_rate": 8.735748451344782e-05, "loss": 2.9339, "step": 16123 }, { "epoch": 1.26, "learning_rate": 8.735670038422332e-05, "loss": 2.8008, "step": 16124 }, { "epoch": 1.26, "learning_rate": 8.735591625499883e-05, "loss": 2.9046, "step": 16125 }, { "epoch": 1.26, "learning_rate": 8.735513212577433e-05, "loss": 2.9235, "step": 16126 }, { "epoch": 1.26, "learning_rate": 8.735434799654983e-05, "loss": 2.8802, "step": 16127 }, { "epoch": 1.26, "learning_rate": 8.735356386732534e-05, "loss": 2.8067, "step": 16128 }, { "epoch": 1.26, "learning_rate": 8.735277973810084e-05, "loss": 2.9891, "step": 16129 }, { "epoch": 1.26, "learning_rate": 8.735199560887635e-05, "loss": 2.852, "step": 16130 }, { "epoch": 1.26, "learning_rate": 8.735121147965185e-05, "loss": 3.0307, "step": 16131 }, { "epoch": 1.26, "learning_rate": 8.735042735042735e-05, "loss": 2.8974, "step": 16132 }, { "epoch": 1.27, "learning_rate": 8.734964322120286e-05, "loss": 2.9485, "step": 16133 }, { "epoch": 1.27, "learning_rate": 8.734885909197836e-05, "loss": 3.1245, "step": 16134 }, { "epoch": 1.27, "learning_rate": 8.734807496275386e-05, "loss": 3.1351, "step": 16135 }, { "epoch": 1.27, "learning_rate": 8.734729083352937e-05, "loss": 2.9112, "step": 16136 }, { "epoch": 1.27, "learning_rate": 8.734650670430487e-05, "loss": 3.041, "step": 16137 }, { "epoch": 1.27, "learning_rate": 8.734572257508037e-05, "loss": 3.0103, "step": 16138 }, { "epoch": 1.27, "learning_rate": 8.734493844585588e-05, "loss": 3.0292, "step": 16139 }, { "epoch": 1.27, "learning_rate": 8.734415431663138e-05, "loss": 3.0396, "step": 16140 }, { "epoch": 1.27, "learning_rate": 8.734337018740689e-05, "loss": 2.7971, "step": 16141 }, { "epoch": 1.27, "learning_rate": 8.734258605818239e-05, "loss": 2.8319, "step": 16142 }, { "epoch": 1.27, "learning_rate": 8.73418019289579e-05, "loss": 2.8194, "step": 16143 }, { "epoch": 1.27, "learning_rate": 8.734101779973341e-05, "loss": 2.9807, "step": 16144 }, { "epoch": 1.27, "learning_rate": 8.734023367050891e-05, "loss": 2.8831, "step": 16145 }, { "epoch": 1.27, "learning_rate": 8.733944954128442e-05, "loss": 2.9018, "step": 16146 }, { "epoch": 1.27, "learning_rate": 8.733866541205992e-05, "loss": 2.8943, "step": 16147 }, { "epoch": 1.27, "learning_rate": 8.733788128283543e-05, "loss": 2.8419, "step": 16148 }, { "epoch": 1.27, "learning_rate": 8.733709715361092e-05, "loss": 3.0632, "step": 16149 }, { "epoch": 1.27, "learning_rate": 8.733631302438642e-05, "loss": 2.8464, "step": 16150 }, { "epoch": 1.27, "learning_rate": 8.733552889516192e-05, "loss": 2.8131, "step": 16151 }, { "epoch": 1.27, "learning_rate": 8.733474476593743e-05, "loss": 2.7753, "step": 16152 }, { "epoch": 1.27, "learning_rate": 8.733396063671293e-05, "loss": 2.9422, "step": 16153 }, { "epoch": 1.27, "learning_rate": 8.733317650748843e-05, "loss": 2.945, "step": 16154 }, { "epoch": 1.27, "learning_rate": 8.733239237826394e-05, "loss": 2.8607, "step": 16155 }, { "epoch": 1.27, "learning_rate": 8.733160824903944e-05, "loss": 2.767, "step": 16156 }, { "epoch": 1.27, "learning_rate": 8.733082411981494e-05, "loss": 2.9299, "step": 16157 }, { "epoch": 1.27, "learning_rate": 8.733003999059045e-05, "loss": 2.8789, "step": 16158 }, { "epoch": 1.27, "learning_rate": 8.732925586136595e-05, "loss": 2.9787, "step": 16159 }, { "epoch": 1.27, "learning_rate": 8.732847173214146e-05, "loss": 2.7888, "step": 16160 }, { "epoch": 1.27, "learning_rate": 8.732768760291696e-05, "loss": 2.9656, "step": 16161 }, { "epoch": 1.27, "learning_rate": 8.732690347369246e-05, "loss": 2.7296, "step": 16162 }, { "epoch": 1.27, "learning_rate": 8.732611934446797e-05, "loss": 2.6887, "step": 16163 }, { "epoch": 1.27, "learning_rate": 8.732533521524347e-05, "loss": 3.0976, "step": 16164 }, { "epoch": 1.27, "learning_rate": 8.732455108601897e-05, "loss": 3.1131, "step": 16165 }, { "epoch": 1.27, "learning_rate": 8.732376695679448e-05, "loss": 3.2119, "step": 16166 }, { "epoch": 1.27, "learning_rate": 8.732298282756998e-05, "loss": 3.0581, "step": 16167 }, { "epoch": 1.27, "learning_rate": 8.73221986983455e-05, "loss": 3.0844, "step": 16168 }, { "epoch": 1.27, "learning_rate": 8.7321414569121e-05, "loss": 3.305, "step": 16169 }, { "epoch": 1.27, "learning_rate": 8.73206304398965e-05, "loss": 2.6934, "step": 16170 }, { "epoch": 1.27, "learning_rate": 8.731984631067201e-05, "loss": 2.9015, "step": 16171 }, { "epoch": 1.27, "learning_rate": 8.731906218144751e-05, "loss": 3.008, "step": 16172 }, { "epoch": 1.27, "learning_rate": 8.731827805222302e-05, "loss": 2.9104, "step": 16173 }, { "epoch": 1.27, "learning_rate": 8.731749392299852e-05, "loss": 2.8006, "step": 16174 }, { "epoch": 1.27, "learning_rate": 8.731670979377402e-05, "loss": 2.9545, "step": 16175 }, { "epoch": 1.27, "learning_rate": 8.731592566454953e-05, "loss": 2.8161, "step": 16176 }, { "epoch": 1.27, "learning_rate": 8.731514153532503e-05, "loss": 2.9725, "step": 16177 }, { "epoch": 1.27, "learning_rate": 8.731435740610054e-05, "loss": 2.7257, "step": 16178 }, { "epoch": 1.27, "learning_rate": 8.731357327687604e-05, "loss": 2.9196, "step": 16179 }, { "epoch": 1.27, "learning_rate": 8.731278914765154e-05, "loss": 2.9254, "step": 16180 }, { "epoch": 1.27, "learning_rate": 8.731200501842705e-05, "loss": 2.9743, "step": 16181 }, { "epoch": 1.27, "learning_rate": 8.731122088920255e-05, "loss": 2.7478, "step": 16182 }, { "epoch": 1.27, "learning_rate": 8.731043675997805e-05, "loss": 2.9277, "step": 16183 }, { "epoch": 1.27, "learning_rate": 8.730965263075356e-05, "loss": 3.1701, "step": 16184 }, { "epoch": 1.27, "learning_rate": 8.730886850152905e-05, "loss": 2.8353, "step": 16185 }, { "epoch": 1.27, "learning_rate": 8.730808437230455e-05, "loss": 2.7095, "step": 16186 }, { "epoch": 1.27, "learning_rate": 8.730730024308006e-05, "loss": 2.7329, "step": 16187 }, { "epoch": 1.27, "learning_rate": 8.730651611385556e-05, "loss": 3.0337, "step": 16188 }, { "epoch": 1.27, "learning_rate": 8.730573198463106e-05, "loss": 2.8919, "step": 16189 }, { "epoch": 1.27, "learning_rate": 8.730494785540657e-05, "loss": 2.8426, "step": 16190 }, { "epoch": 1.27, "learning_rate": 8.730416372618207e-05, "loss": 2.7465, "step": 16191 }, { "epoch": 1.27, "learning_rate": 8.730337959695757e-05, "loss": 2.7779, "step": 16192 }, { "epoch": 1.27, "learning_rate": 8.730259546773308e-05, "loss": 3.1116, "step": 16193 }, { "epoch": 1.27, "learning_rate": 8.73018113385086e-05, "loss": 2.8506, "step": 16194 }, { "epoch": 1.27, "learning_rate": 8.73010272092841e-05, "loss": 2.9848, "step": 16195 }, { "epoch": 1.27, "learning_rate": 8.73002430800596e-05, "loss": 2.8559, "step": 16196 }, { "epoch": 1.27, "learning_rate": 8.72994589508351e-05, "loss": 2.8367, "step": 16197 }, { "epoch": 1.27, "learning_rate": 8.729867482161061e-05, "loss": 3.092, "step": 16198 }, { "epoch": 1.27, "learning_rate": 8.729789069238611e-05, "loss": 3.0612, "step": 16199 }, { "epoch": 1.27, "learning_rate": 8.729710656316162e-05, "loss": 3.2299, "step": 16200 }, { "epoch": 1.27, "learning_rate": 8.729632243393712e-05, "loss": 2.8612, "step": 16201 }, { "epoch": 1.27, "learning_rate": 8.729553830471262e-05, "loss": 3.0752, "step": 16202 }, { "epoch": 1.27, "learning_rate": 8.729475417548813e-05, "loss": 3.1719, "step": 16203 }, { "epoch": 1.27, "learning_rate": 8.729397004626363e-05, "loss": 2.9036, "step": 16204 }, { "epoch": 1.27, "learning_rate": 8.729318591703913e-05, "loss": 2.9766, "step": 16205 }, { "epoch": 1.27, "learning_rate": 8.729240178781464e-05, "loss": 3.0043, "step": 16206 }, { "epoch": 1.27, "learning_rate": 8.729161765859014e-05, "loss": 3.1198, "step": 16207 }, { "epoch": 1.27, "learning_rate": 8.729083352936565e-05, "loss": 2.9359, "step": 16208 }, { "epoch": 1.27, "learning_rate": 8.729004940014115e-05, "loss": 2.8692, "step": 16209 }, { "epoch": 1.27, "learning_rate": 8.728926527091665e-05, "loss": 3.0218, "step": 16210 }, { "epoch": 1.27, "learning_rate": 8.728848114169216e-05, "loss": 2.6569, "step": 16211 }, { "epoch": 1.27, "learning_rate": 8.728769701246766e-05, "loss": 2.8213, "step": 16212 }, { "epoch": 1.27, "learning_rate": 8.728691288324316e-05, "loss": 2.7537, "step": 16213 }, { "epoch": 1.27, "learning_rate": 8.728612875401867e-05, "loss": 2.9217, "step": 16214 }, { "epoch": 1.27, "learning_rate": 8.728534462479417e-05, "loss": 3.0127, "step": 16215 }, { "epoch": 1.27, "learning_rate": 8.728456049556968e-05, "loss": 2.9305, "step": 16216 }, { "epoch": 1.27, "learning_rate": 8.728377636634518e-05, "loss": 2.955, "step": 16217 }, { "epoch": 1.27, "learning_rate": 8.728299223712068e-05, "loss": 2.8378, "step": 16218 }, { "epoch": 1.27, "learning_rate": 8.728220810789619e-05, "loss": 2.813, "step": 16219 }, { "epoch": 1.27, "learning_rate": 8.728142397867169e-05, "loss": 2.81, "step": 16220 }, { "epoch": 1.27, "learning_rate": 8.72806398494472e-05, "loss": 2.7723, "step": 16221 }, { "epoch": 1.27, "learning_rate": 8.72798557202227e-05, "loss": 2.9319, "step": 16222 }, { "epoch": 1.27, "learning_rate": 8.72790715909982e-05, "loss": 3.1744, "step": 16223 }, { "epoch": 1.27, "learning_rate": 8.72782874617737e-05, "loss": 3.1177, "step": 16224 }, { "epoch": 1.27, "learning_rate": 8.727750333254921e-05, "loss": 3.0357, "step": 16225 }, { "epoch": 1.27, "learning_rate": 8.727671920332471e-05, "loss": 2.8051, "step": 16226 }, { "epoch": 1.27, "learning_rate": 8.727593507410022e-05, "loss": 2.8575, "step": 16227 }, { "epoch": 1.27, "learning_rate": 8.727515094487572e-05, "loss": 3.101, "step": 16228 }, { "epoch": 1.27, "learning_rate": 8.727436681565122e-05, "loss": 2.8679, "step": 16229 }, { "epoch": 1.27, "learning_rate": 8.727358268642673e-05, "loss": 2.8711, "step": 16230 }, { "epoch": 1.27, "learning_rate": 8.727279855720223e-05, "loss": 3.1504, "step": 16231 }, { "epoch": 1.27, "learning_rate": 8.727201442797773e-05, "loss": 2.8444, "step": 16232 }, { "epoch": 1.27, "learning_rate": 8.727123029875324e-05, "loss": 3.0142, "step": 16233 }, { "epoch": 1.27, "learning_rate": 8.727044616952874e-05, "loss": 2.8085, "step": 16234 }, { "epoch": 1.27, "learning_rate": 8.726966204030425e-05, "loss": 2.8858, "step": 16235 }, { "epoch": 1.27, "learning_rate": 8.726887791107975e-05, "loss": 2.9453, "step": 16236 }, { "epoch": 1.27, "learning_rate": 8.726809378185525e-05, "loss": 2.8498, "step": 16237 }, { "epoch": 1.27, "learning_rate": 8.726730965263076e-05, "loss": 2.8801, "step": 16238 }, { "epoch": 1.27, "learning_rate": 8.726652552340626e-05, "loss": 2.6901, "step": 16239 }, { "epoch": 1.27, "learning_rate": 8.726574139418176e-05, "loss": 2.9507, "step": 16240 }, { "epoch": 1.27, "learning_rate": 8.726495726495727e-05, "loss": 2.856, "step": 16241 }, { "epoch": 1.27, "learning_rate": 8.726417313573277e-05, "loss": 2.8554, "step": 16242 }, { "epoch": 1.27, "learning_rate": 8.726338900650827e-05, "loss": 2.7981, "step": 16243 }, { "epoch": 1.27, "learning_rate": 8.726260487728378e-05, "loss": 3.0294, "step": 16244 }, { "epoch": 1.27, "learning_rate": 8.726182074805928e-05, "loss": 2.8981, "step": 16245 }, { "epoch": 1.27, "learning_rate": 8.726103661883479e-05, "loss": 2.8415, "step": 16246 }, { "epoch": 1.27, "learning_rate": 8.726025248961029e-05, "loss": 2.8256, "step": 16247 }, { "epoch": 1.27, "learning_rate": 8.725946836038579e-05, "loss": 3.0354, "step": 16248 }, { "epoch": 1.27, "learning_rate": 8.72586842311613e-05, "loss": 2.8286, "step": 16249 }, { "epoch": 1.27, "learning_rate": 8.72579001019368e-05, "loss": 2.7517, "step": 16250 }, { "epoch": 1.27, "learning_rate": 8.725711597271232e-05, "loss": 2.848, "step": 16251 }, { "epoch": 1.27, "learning_rate": 8.725633184348782e-05, "loss": 2.884, "step": 16252 }, { "epoch": 1.27, "learning_rate": 8.725554771426332e-05, "loss": 2.9178, "step": 16253 }, { "epoch": 1.27, "learning_rate": 8.725476358503883e-05, "loss": 2.9213, "step": 16254 }, { "epoch": 1.27, "learning_rate": 8.725397945581433e-05, "loss": 2.8831, "step": 16255 }, { "epoch": 1.27, "learning_rate": 8.725319532658984e-05, "loss": 3.0202, "step": 16256 }, { "epoch": 1.27, "learning_rate": 8.725241119736533e-05, "loss": 2.7091, "step": 16257 }, { "epoch": 1.27, "learning_rate": 8.725162706814083e-05, "loss": 2.7196, "step": 16258 }, { "epoch": 1.27, "learning_rate": 8.725084293891633e-05, "loss": 2.8901, "step": 16259 }, { "epoch": 1.27, "learning_rate": 8.725005880969184e-05, "loss": 2.8342, "step": 16260 }, { "epoch": 1.28, "learning_rate": 8.724927468046734e-05, "loss": 3.2326, "step": 16261 }, { "epoch": 1.28, "learning_rate": 8.724849055124284e-05, "loss": 2.8727, "step": 16262 }, { "epoch": 1.28, "learning_rate": 8.724770642201835e-05, "loss": 2.8984, "step": 16263 }, { "epoch": 1.28, "learning_rate": 8.724692229279385e-05, "loss": 3.1748, "step": 16264 }, { "epoch": 1.28, "learning_rate": 8.724613816356936e-05, "loss": 2.6767, "step": 16265 }, { "epoch": 1.28, "learning_rate": 8.724535403434486e-05, "loss": 2.9513, "step": 16266 }, { "epoch": 1.28, "learning_rate": 8.724456990512036e-05, "loss": 2.8226, "step": 16267 }, { "epoch": 1.28, "learning_rate": 8.724378577589587e-05, "loss": 2.9191, "step": 16268 }, { "epoch": 1.28, "learning_rate": 8.724300164667137e-05, "loss": 2.8861, "step": 16269 }, { "epoch": 1.28, "learning_rate": 8.724221751744687e-05, "loss": 2.7139, "step": 16270 }, { "epoch": 1.28, "learning_rate": 8.724143338822238e-05, "loss": 2.9922, "step": 16271 }, { "epoch": 1.28, "learning_rate": 8.724064925899788e-05, "loss": 2.8663, "step": 16272 }, { "epoch": 1.28, "learning_rate": 8.723986512977338e-05, "loss": 2.7813, "step": 16273 }, { "epoch": 1.28, "learning_rate": 8.723908100054889e-05, "loss": 2.9193, "step": 16274 }, { "epoch": 1.28, "learning_rate": 8.723829687132439e-05, "loss": 2.7508, "step": 16275 }, { "epoch": 1.28, "learning_rate": 8.723751274209991e-05, "loss": 2.9331, "step": 16276 }, { "epoch": 1.28, "learning_rate": 8.723672861287541e-05, "loss": 2.8148, "step": 16277 }, { "epoch": 1.28, "learning_rate": 8.723594448365092e-05, "loss": 2.8771, "step": 16278 }, { "epoch": 1.28, "learning_rate": 8.723516035442642e-05, "loss": 3.1093, "step": 16279 }, { "epoch": 1.28, "learning_rate": 8.723437622520192e-05, "loss": 2.7141, "step": 16280 }, { "epoch": 1.28, "learning_rate": 8.723359209597743e-05, "loss": 2.6881, "step": 16281 }, { "epoch": 1.28, "learning_rate": 8.723280796675293e-05, "loss": 2.6507, "step": 16282 }, { "epoch": 1.28, "learning_rate": 8.723202383752844e-05, "loss": 2.9958, "step": 16283 }, { "epoch": 1.28, "learning_rate": 8.723123970830394e-05, "loss": 2.8977, "step": 16284 }, { "epoch": 1.28, "learning_rate": 8.723045557907944e-05, "loss": 2.9926, "step": 16285 }, { "epoch": 1.28, "learning_rate": 8.722967144985495e-05, "loss": 3.0825, "step": 16286 }, { "epoch": 1.28, "learning_rate": 8.722888732063045e-05, "loss": 2.9599, "step": 16287 }, { "epoch": 1.28, "learning_rate": 8.722810319140595e-05, "loss": 2.7278, "step": 16288 }, { "epoch": 1.28, "learning_rate": 8.722731906218146e-05, "loss": 2.8233, "step": 16289 }, { "epoch": 1.28, "learning_rate": 8.722653493295696e-05, "loss": 2.8458, "step": 16290 }, { "epoch": 1.28, "learning_rate": 8.722575080373246e-05, "loss": 2.8033, "step": 16291 }, { "epoch": 1.28, "learning_rate": 8.722496667450797e-05, "loss": 2.9842, "step": 16292 }, { "epoch": 1.28, "learning_rate": 8.722418254528346e-05, "loss": 2.8802, "step": 16293 }, { "epoch": 1.28, "learning_rate": 8.722339841605896e-05, "loss": 3.0252, "step": 16294 }, { "epoch": 1.28, "learning_rate": 8.722261428683447e-05, "loss": 2.9041, "step": 16295 }, { "epoch": 1.28, "learning_rate": 8.722183015760997e-05, "loss": 2.762, "step": 16296 }, { "epoch": 1.28, "learning_rate": 8.722104602838547e-05, "loss": 3.1416, "step": 16297 }, { "epoch": 1.28, "learning_rate": 8.722026189916098e-05, "loss": 2.8284, "step": 16298 }, { "epoch": 1.28, "learning_rate": 8.721947776993648e-05, "loss": 2.9342, "step": 16299 }, { "epoch": 1.28, "learning_rate": 8.721869364071198e-05, "loss": 3.1316, "step": 16300 }, { "epoch": 1.28, "learning_rate": 8.721790951148749e-05, "loss": 3.053, "step": 16301 }, { "epoch": 1.28, "learning_rate": 8.7217125382263e-05, "loss": 2.8743, "step": 16302 }, { "epoch": 1.28, "learning_rate": 8.721634125303851e-05, "loss": 3.0678, "step": 16303 }, { "epoch": 1.28, "learning_rate": 8.721555712381401e-05, "loss": 2.6433, "step": 16304 }, { "epoch": 1.28, "learning_rate": 8.721477299458952e-05, "loss": 2.8964, "step": 16305 }, { "epoch": 1.28, "learning_rate": 8.721398886536502e-05, "loss": 3.0991, "step": 16306 }, { "epoch": 1.28, "learning_rate": 8.721320473614052e-05, "loss": 2.8874, "step": 16307 }, { "epoch": 1.28, "learning_rate": 8.721242060691603e-05, "loss": 2.8573, "step": 16308 }, { "epoch": 1.28, "learning_rate": 8.721163647769153e-05, "loss": 2.8334, "step": 16309 }, { "epoch": 1.28, "learning_rate": 8.721085234846703e-05, "loss": 3.0139, "step": 16310 }, { "epoch": 1.28, "learning_rate": 8.721006821924254e-05, "loss": 3.0209, "step": 16311 }, { "epoch": 1.28, "learning_rate": 8.720928409001804e-05, "loss": 2.681, "step": 16312 }, { "epoch": 1.28, "learning_rate": 8.720849996079355e-05, "loss": 2.9347, "step": 16313 }, { "epoch": 1.28, "learning_rate": 8.720771583156905e-05, "loss": 2.9463, "step": 16314 }, { "epoch": 1.28, "learning_rate": 8.720693170234455e-05, "loss": 3.1697, "step": 16315 }, { "epoch": 1.28, "learning_rate": 8.720614757312006e-05, "loss": 2.8335, "step": 16316 }, { "epoch": 1.28, "learning_rate": 8.720536344389556e-05, "loss": 2.8948, "step": 16317 }, { "epoch": 1.28, "learning_rate": 8.720457931467106e-05, "loss": 2.8241, "step": 16318 }, { "epoch": 1.28, "learning_rate": 8.720379518544657e-05, "loss": 3.1249, "step": 16319 }, { "epoch": 1.28, "learning_rate": 8.720301105622207e-05, "loss": 3.1424, "step": 16320 }, { "epoch": 1.28, "learning_rate": 8.720222692699757e-05, "loss": 2.8534, "step": 16321 }, { "epoch": 1.28, "learning_rate": 8.720144279777308e-05, "loss": 3.1001, "step": 16322 }, { "epoch": 1.28, "learning_rate": 8.720065866854858e-05, "loss": 2.9366, "step": 16323 }, { "epoch": 1.28, "learning_rate": 8.719987453932409e-05, "loss": 3.0189, "step": 16324 }, { "epoch": 1.28, "learning_rate": 8.719909041009959e-05, "loss": 3.038, "step": 16325 }, { "epoch": 1.28, "learning_rate": 8.71983062808751e-05, "loss": 2.6251, "step": 16326 }, { "epoch": 1.28, "learning_rate": 8.71975221516506e-05, "loss": 2.966, "step": 16327 }, { "epoch": 1.28, "learning_rate": 8.71967380224261e-05, "loss": 3.1435, "step": 16328 }, { "epoch": 1.28, "learning_rate": 8.71959538932016e-05, "loss": 2.7188, "step": 16329 }, { "epoch": 1.28, "learning_rate": 8.719516976397711e-05, "loss": 2.9107, "step": 16330 }, { "epoch": 1.28, "learning_rate": 8.719438563475261e-05, "loss": 2.9871, "step": 16331 }, { "epoch": 1.28, "learning_rate": 8.719360150552812e-05, "loss": 2.8872, "step": 16332 }, { "epoch": 1.28, "learning_rate": 8.719281737630362e-05, "loss": 2.8356, "step": 16333 }, { "epoch": 1.28, "learning_rate": 8.719203324707912e-05, "loss": 3.0195, "step": 16334 }, { "epoch": 1.28, "learning_rate": 8.719124911785463e-05, "loss": 3.1213, "step": 16335 }, { "epoch": 1.28, "learning_rate": 8.719046498863013e-05, "loss": 2.8645, "step": 16336 }, { "epoch": 1.28, "learning_rate": 8.718968085940563e-05, "loss": 2.8653, "step": 16337 }, { "epoch": 1.28, "learning_rate": 8.718889673018114e-05, "loss": 2.876, "step": 16338 }, { "epoch": 1.28, "learning_rate": 8.718811260095664e-05, "loss": 2.9436, "step": 16339 }, { "epoch": 1.28, "learning_rate": 8.718732847173214e-05, "loss": 2.7919, "step": 16340 }, { "epoch": 1.28, "learning_rate": 8.718654434250765e-05, "loss": 2.8204, "step": 16341 }, { "epoch": 1.28, "learning_rate": 8.718576021328315e-05, "loss": 2.8747, "step": 16342 }, { "epoch": 1.28, "learning_rate": 8.718497608405866e-05, "loss": 2.8371, "step": 16343 }, { "epoch": 1.28, "learning_rate": 8.718419195483416e-05, "loss": 2.8115, "step": 16344 }, { "epoch": 1.28, "learning_rate": 8.718340782560966e-05, "loss": 3.1034, "step": 16345 }, { "epoch": 1.28, "learning_rate": 8.718262369638517e-05, "loss": 2.7818, "step": 16346 }, { "epoch": 1.28, "learning_rate": 8.718183956716067e-05, "loss": 2.7009, "step": 16347 }, { "epoch": 1.28, "learning_rate": 8.718105543793617e-05, "loss": 3.0099, "step": 16348 }, { "epoch": 1.28, "learning_rate": 8.718027130871168e-05, "loss": 2.7246, "step": 16349 }, { "epoch": 1.28, "learning_rate": 8.717948717948718e-05, "loss": 2.8388, "step": 16350 }, { "epoch": 1.28, "learning_rate": 8.717870305026269e-05, "loss": 3.0037, "step": 16351 }, { "epoch": 1.28, "learning_rate": 8.717791892103819e-05, "loss": 2.7607, "step": 16352 }, { "epoch": 1.28, "learning_rate": 8.717713479181369e-05, "loss": 2.9272, "step": 16353 }, { "epoch": 1.28, "learning_rate": 8.71763506625892e-05, "loss": 2.896, "step": 16354 }, { "epoch": 1.28, "learning_rate": 8.71755665333647e-05, "loss": 2.7802, "step": 16355 }, { "epoch": 1.28, "learning_rate": 8.71747824041402e-05, "loss": 2.9125, "step": 16356 }, { "epoch": 1.28, "learning_rate": 8.717399827491571e-05, "loss": 3.0388, "step": 16357 }, { "epoch": 1.28, "learning_rate": 8.717321414569121e-05, "loss": 2.8287, "step": 16358 }, { "epoch": 1.28, "learning_rate": 8.717243001646673e-05, "loss": 3.2097, "step": 16359 }, { "epoch": 1.28, "learning_rate": 8.717164588724223e-05, "loss": 2.9582, "step": 16360 }, { "epoch": 1.28, "learning_rate": 8.717086175801774e-05, "loss": 2.7655, "step": 16361 }, { "epoch": 1.28, "learning_rate": 8.717007762879324e-05, "loss": 2.9542, "step": 16362 }, { "epoch": 1.28, "learning_rate": 8.716929349956874e-05, "loss": 3.1078, "step": 16363 }, { "epoch": 1.28, "learning_rate": 8.716850937034425e-05, "loss": 2.8226, "step": 16364 }, { "epoch": 1.28, "learning_rate": 8.716772524111974e-05, "loss": 3.0319, "step": 16365 }, { "epoch": 1.28, "learning_rate": 8.716694111189524e-05, "loss": 2.9944, "step": 16366 }, { "epoch": 1.28, "learning_rate": 8.716615698267074e-05, "loss": 2.8969, "step": 16367 }, { "epoch": 1.28, "learning_rate": 8.716537285344625e-05, "loss": 3.0955, "step": 16368 }, { "epoch": 1.28, "learning_rate": 8.716458872422175e-05, "loss": 2.8802, "step": 16369 }, { "epoch": 1.28, "learning_rate": 8.716380459499726e-05, "loss": 2.7765, "step": 16370 }, { "epoch": 1.28, "learning_rate": 8.716302046577276e-05, "loss": 3.0194, "step": 16371 }, { "epoch": 1.28, "learning_rate": 8.716223633654826e-05, "loss": 2.9673, "step": 16372 }, { "epoch": 1.28, "learning_rate": 8.716145220732377e-05, "loss": 2.7329, "step": 16373 }, { "epoch": 1.28, "learning_rate": 8.716066807809927e-05, "loss": 2.9409, "step": 16374 }, { "epoch": 1.28, "learning_rate": 8.715988394887477e-05, "loss": 2.6603, "step": 16375 }, { "epoch": 1.28, "learning_rate": 8.715909981965028e-05, "loss": 3.0365, "step": 16376 }, { "epoch": 1.28, "learning_rate": 8.715831569042578e-05, "loss": 3.1239, "step": 16377 }, { "epoch": 1.28, "learning_rate": 8.715753156120128e-05, "loss": 2.7691, "step": 16378 }, { "epoch": 1.28, "learning_rate": 8.715674743197679e-05, "loss": 2.7424, "step": 16379 }, { "epoch": 1.28, "learning_rate": 8.715596330275229e-05, "loss": 2.8298, "step": 16380 }, { "epoch": 1.28, "learning_rate": 8.71551791735278e-05, "loss": 2.7822, "step": 16381 }, { "epoch": 1.28, "learning_rate": 8.71543950443033e-05, "loss": 2.736, "step": 16382 }, { "epoch": 1.28, "learning_rate": 8.71536109150788e-05, "loss": 3.0987, "step": 16383 }, { "epoch": 1.28, "learning_rate": 8.715282678585432e-05, "loss": 2.7489, "step": 16384 }, { "epoch": 1.28, "learning_rate": 8.715204265662982e-05, "loss": 2.9236, "step": 16385 }, { "epoch": 1.28, "learning_rate": 8.715125852740533e-05, "loss": 3.0873, "step": 16386 }, { "epoch": 1.28, "learning_rate": 8.715047439818083e-05, "loss": 2.9019, "step": 16387 }, { "epoch": 1.29, "learning_rate": 8.714969026895633e-05, "loss": 2.8391, "step": 16388 }, { "epoch": 1.29, "learning_rate": 8.714890613973184e-05, "loss": 2.806, "step": 16389 }, { "epoch": 1.29, "learning_rate": 8.714812201050734e-05, "loss": 2.7192, "step": 16390 }, { "epoch": 1.29, "learning_rate": 8.714733788128285e-05, "loss": 2.9263, "step": 16391 }, { "epoch": 1.29, "learning_rate": 8.714655375205835e-05, "loss": 2.853, "step": 16392 }, { "epoch": 1.29, "learning_rate": 8.714576962283385e-05, "loss": 3.0436, "step": 16393 }, { "epoch": 1.29, "learning_rate": 8.714498549360936e-05, "loss": 2.9488, "step": 16394 }, { "epoch": 1.29, "learning_rate": 8.714420136438486e-05, "loss": 3.1903, "step": 16395 }, { "epoch": 1.29, "learning_rate": 8.714341723516036e-05, "loss": 3.0187, "step": 16396 }, { "epoch": 1.29, "learning_rate": 8.714263310593587e-05, "loss": 2.9494, "step": 16397 }, { "epoch": 1.29, "learning_rate": 8.714184897671137e-05, "loss": 2.8115, "step": 16398 }, { "epoch": 1.29, "learning_rate": 8.714106484748688e-05, "loss": 2.8189, "step": 16399 }, { "epoch": 1.29, "learning_rate": 8.714028071826238e-05, "loss": 2.9105, "step": 16400 }, { "epoch": 1.29, "learning_rate": 8.713949658903787e-05, "loss": 2.8771, "step": 16401 }, { "epoch": 1.29, "learning_rate": 8.713871245981337e-05, "loss": 3.0436, "step": 16402 }, { "epoch": 1.29, "learning_rate": 8.713792833058888e-05, "loss": 2.7705, "step": 16403 }, { "epoch": 1.29, "learning_rate": 8.713714420136438e-05, "loss": 2.7235, "step": 16404 }, { "epoch": 1.29, "learning_rate": 8.713636007213988e-05, "loss": 2.9948, "step": 16405 }, { "epoch": 1.29, "learning_rate": 8.713557594291539e-05, "loss": 2.8975, "step": 16406 }, { "epoch": 1.29, "learning_rate": 8.713479181369089e-05, "loss": 2.8963, "step": 16407 }, { "epoch": 1.29, "learning_rate": 8.71340076844664e-05, "loss": 2.8757, "step": 16408 }, { "epoch": 1.29, "learning_rate": 8.71332235552419e-05, "loss": 2.8474, "step": 16409 }, { "epoch": 1.29, "learning_rate": 8.713243942601742e-05, "loss": 2.674, "step": 16410 }, { "epoch": 1.29, "learning_rate": 8.713165529679292e-05, "loss": 3.0012, "step": 16411 }, { "epoch": 1.29, "learning_rate": 8.713087116756842e-05, "loss": 2.8103, "step": 16412 }, { "epoch": 1.29, "learning_rate": 8.713008703834393e-05, "loss": 2.892, "step": 16413 }, { "epoch": 1.29, "learning_rate": 8.712930290911943e-05, "loss": 3.0841, "step": 16414 }, { "epoch": 1.29, "learning_rate": 8.712851877989493e-05, "loss": 2.8359, "step": 16415 }, { "epoch": 1.29, "learning_rate": 8.712773465067044e-05, "loss": 3.0352, "step": 16416 }, { "epoch": 1.29, "learning_rate": 8.712695052144594e-05, "loss": 2.88, "step": 16417 }, { "epoch": 1.29, "learning_rate": 8.712616639222145e-05, "loss": 2.9432, "step": 16418 }, { "epoch": 1.29, "learning_rate": 8.712538226299695e-05, "loss": 2.9127, "step": 16419 }, { "epoch": 1.29, "learning_rate": 8.712459813377245e-05, "loss": 3.0751, "step": 16420 }, { "epoch": 1.29, "learning_rate": 8.712381400454796e-05, "loss": 3.0058, "step": 16421 }, { "epoch": 1.29, "learning_rate": 8.712302987532346e-05, "loss": 2.9334, "step": 16422 }, { "epoch": 1.29, "learning_rate": 8.712224574609896e-05, "loss": 2.9678, "step": 16423 }, { "epoch": 1.29, "learning_rate": 8.712146161687447e-05, "loss": 3.0005, "step": 16424 }, { "epoch": 1.29, "learning_rate": 8.712067748764997e-05, "loss": 2.6409, "step": 16425 }, { "epoch": 1.29, "learning_rate": 8.711989335842547e-05, "loss": 3.0533, "step": 16426 }, { "epoch": 1.29, "learning_rate": 8.711910922920098e-05, "loss": 3.0608, "step": 16427 }, { "epoch": 1.29, "learning_rate": 8.711832509997648e-05, "loss": 2.897, "step": 16428 }, { "epoch": 1.29, "learning_rate": 8.711754097075199e-05, "loss": 2.9903, "step": 16429 }, { "epoch": 1.29, "learning_rate": 8.711675684152749e-05, "loss": 2.845, "step": 16430 }, { "epoch": 1.29, "learning_rate": 8.711597271230299e-05, "loss": 2.6648, "step": 16431 }, { "epoch": 1.29, "learning_rate": 8.71151885830785e-05, "loss": 2.8686, "step": 16432 }, { "epoch": 1.29, "learning_rate": 8.7114404453854e-05, "loss": 2.6906, "step": 16433 }, { "epoch": 1.29, "learning_rate": 8.71136203246295e-05, "loss": 3.0285, "step": 16434 }, { "epoch": 1.29, "learning_rate": 8.711283619540501e-05, "loss": 2.8837, "step": 16435 }, { "epoch": 1.29, "learning_rate": 8.711205206618051e-05, "loss": 2.6426, "step": 16436 }, { "epoch": 1.29, "learning_rate": 8.711126793695602e-05, "loss": 3.0901, "step": 16437 }, { "epoch": 1.29, "learning_rate": 8.711048380773152e-05, "loss": 2.6285, "step": 16438 }, { "epoch": 1.29, "learning_rate": 8.710969967850702e-05, "loss": 2.8512, "step": 16439 }, { "epoch": 1.29, "learning_rate": 8.710891554928253e-05, "loss": 2.9433, "step": 16440 }, { "epoch": 1.29, "learning_rate": 8.710813142005803e-05, "loss": 2.9419, "step": 16441 }, { "epoch": 1.29, "learning_rate": 8.710734729083353e-05, "loss": 2.8793, "step": 16442 }, { "epoch": 1.29, "learning_rate": 8.710656316160904e-05, "loss": 3.2047, "step": 16443 }, { "epoch": 1.29, "learning_rate": 8.710577903238454e-05, "loss": 2.963, "step": 16444 }, { "epoch": 1.29, "learning_rate": 8.710499490316004e-05, "loss": 2.7648, "step": 16445 }, { "epoch": 1.29, "learning_rate": 8.710421077393555e-05, "loss": 2.917, "step": 16446 }, { "epoch": 1.29, "learning_rate": 8.710342664471105e-05, "loss": 2.8067, "step": 16447 }, { "epoch": 1.29, "learning_rate": 8.710264251548656e-05, "loss": 2.9036, "step": 16448 }, { "epoch": 1.29, "learning_rate": 8.710185838626206e-05, "loss": 2.9017, "step": 16449 }, { "epoch": 1.29, "learning_rate": 8.710107425703756e-05, "loss": 2.9736, "step": 16450 }, { "epoch": 1.29, "learning_rate": 8.710029012781307e-05, "loss": 3.0114, "step": 16451 }, { "epoch": 1.29, "learning_rate": 8.709950599858857e-05, "loss": 2.8315, "step": 16452 }, { "epoch": 1.29, "learning_rate": 8.709872186936407e-05, "loss": 3.2063, "step": 16453 }, { "epoch": 1.29, "learning_rate": 8.709793774013958e-05, "loss": 2.9544, "step": 16454 }, { "epoch": 1.29, "learning_rate": 8.709715361091508e-05, "loss": 2.9062, "step": 16455 }, { "epoch": 1.29, "learning_rate": 8.709636948169058e-05, "loss": 3.0984, "step": 16456 }, { "epoch": 1.29, "learning_rate": 8.709558535246609e-05, "loss": 3.1679, "step": 16457 }, { "epoch": 1.29, "learning_rate": 8.709480122324159e-05, "loss": 2.8411, "step": 16458 }, { "epoch": 1.29, "learning_rate": 8.70940170940171e-05, "loss": 2.8644, "step": 16459 }, { "epoch": 1.29, "learning_rate": 8.70932329647926e-05, "loss": 2.8499, "step": 16460 }, { "epoch": 1.29, "learning_rate": 8.70924488355681e-05, "loss": 2.8364, "step": 16461 }, { "epoch": 1.29, "learning_rate": 8.709166470634361e-05, "loss": 2.7216, "step": 16462 }, { "epoch": 1.29, "learning_rate": 8.709088057711911e-05, "loss": 2.9393, "step": 16463 }, { "epoch": 1.29, "learning_rate": 8.709009644789461e-05, "loss": 2.8598, "step": 16464 }, { "epoch": 1.29, "learning_rate": 8.708931231867012e-05, "loss": 2.8423, "step": 16465 }, { "epoch": 1.29, "learning_rate": 8.708852818944562e-05, "loss": 2.7746, "step": 16466 }, { "epoch": 1.29, "learning_rate": 8.708774406022114e-05, "loss": 2.8009, "step": 16467 }, { "epoch": 1.29, "learning_rate": 8.708695993099664e-05, "loss": 2.9141, "step": 16468 }, { "epoch": 1.29, "learning_rate": 8.708617580177215e-05, "loss": 2.8639, "step": 16469 }, { "epoch": 1.29, "learning_rate": 8.708539167254765e-05, "loss": 2.946, "step": 16470 }, { "epoch": 1.29, "learning_rate": 8.708460754332315e-05, "loss": 2.9531, "step": 16471 }, { "epoch": 1.29, "learning_rate": 8.708382341409866e-05, "loss": 3.0695, "step": 16472 }, { "epoch": 1.29, "learning_rate": 8.708303928487415e-05, "loss": 2.6964, "step": 16473 }, { "epoch": 1.29, "learning_rate": 8.708225515564965e-05, "loss": 2.9673, "step": 16474 }, { "epoch": 1.29, "learning_rate": 8.708147102642515e-05, "loss": 2.8022, "step": 16475 }, { "epoch": 1.29, "learning_rate": 8.708068689720066e-05, "loss": 2.9719, "step": 16476 }, { "epoch": 1.29, "learning_rate": 8.707990276797616e-05, "loss": 2.7833, "step": 16477 }, { "epoch": 1.29, "learning_rate": 8.707911863875167e-05, "loss": 2.8064, "step": 16478 }, { "epoch": 1.29, "learning_rate": 8.707833450952717e-05, "loss": 2.736, "step": 16479 }, { "epoch": 1.29, "learning_rate": 8.707755038030267e-05, "loss": 2.8828, "step": 16480 }, { "epoch": 1.29, "learning_rate": 8.707676625107818e-05, "loss": 2.8773, "step": 16481 }, { "epoch": 1.29, "learning_rate": 8.707598212185368e-05, "loss": 3.348, "step": 16482 }, { "epoch": 1.29, "learning_rate": 8.707519799262918e-05, "loss": 2.9608, "step": 16483 }, { "epoch": 1.29, "learning_rate": 8.707441386340469e-05, "loss": 2.8711, "step": 16484 }, { "epoch": 1.29, "learning_rate": 8.707362973418019e-05, "loss": 3.0388, "step": 16485 }, { "epoch": 1.29, "learning_rate": 8.70728456049557e-05, "loss": 3.03, "step": 16486 }, { "epoch": 1.29, "learning_rate": 8.70720614757312e-05, "loss": 3.1209, "step": 16487 }, { "epoch": 1.29, "learning_rate": 8.70712773465067e-05, "loss": 2.8903, "step": 16488 }, { "epoch": 1.29, "learning_rate": 8.70704932172822e-05, "loss": 3.0702, "step": 16489 }, { "epoch": 1.29, "learning_rate": 8.706970908805771e-05, "loss": 2.7456, "step": 16490 }, { "epoch": 1.29, "learning_rate": 8.706892495883321e-05, "loss": 2.8459, "step": 16491 }, { "epoch": 1.29, "learning_rate": 8.706814082960873e-05, "loss": 3.0754, "step": 16492 }, { "epoch": 1.29, "learning_rate": 8.706735670038423e-05, "loss": 2.772, "step": 16493 }, { "epoch": 1.29, "learning_rate": 8.706657257115974e-05, "loss": 2.7561, "step": 16494 }, { "epoch": 1.29, "learning_rate": 8.706578844193524e-05, "loss": 3.2333, "step": 16495 }, { "epoch": 1.29, "learning_rate": 8.706500431271075e-05, "loss": 2.6288, "step": 16496 }, { "epoch": 1.29, "learning_rate": 8.706422018348625e-05, "loss": 2.9246, "step": 16497 }, { "epoch": 1.29, "learning_rate": 8.706343605426175e-05, "loss": 2.8401, "step": 16498 }, { "epoch": 1.29, "learning_rate": 8.706265192503726e-05, "loss": 2.9004, "step": 16499 }, { "epoch": 1.29, "learning_rate": 8.706186779581276e-05, "loss": 2.8202, "step": 16500 }, { "epoch": 1.29, "learning_rate": 8.706108366658826e-05, "loss": 3.0144, "step": 16501 }, { "epoch": 1.29, "learning_rate": 8.706029953736377e-05, "loss": 2.8145, "step": 16502 }, { "epoch": 1.29, "learning_rate": 8.705951540813927e-05, "loss": 2.8753, "step": 16503 }, { "epoch": 1.29, "learning_rate": 8.705873127891478e-05, "loss": 2.8375, "step": 16504 }, { "epoch": 1.29, "learning_rate": 8.705794714969028e-05, "loss": 2.9307, "step": 16505 }, { "epoch": 1.29, "learning_rate": 8.705716302046578e-05, "loss": 2.8987, "step": 16506 }, { "epoch": 1.29, "learning_rate": 8.705637889124129e-05, "loss": 2.9161, "step": 16507 }, { "epoch": 1.29, "learning_rate": 8.705559476201679e-05, "loss": 2.7956, "step": 16508 }, { "epoch": 1.29, "learning_rate": 8.705481063279228e-05, "loss": 2.8568, "step": 16509 }, { "epoch": 1.29, "learning_rate": 8.705402650356778e-05, "loss": 3.0148, "step": 16510 }, { "epoch": 1.29, "learning_rate": 8.705324237434329e-05, "loss": 2.875, "step": 16511 }, { "epoch": 1.29, "learning_rate": 8.705245824511879e-05, "loss": 2.8755, "step": 16512 }, { "epoch": 1.29, "learning_rate": 8.70516741158943e-05, "loss": 2.6492, "step": 16513 }, { "epoch": 1.29, "learning_rate": 8.70508899866698e-05, "loss": 2.9582, "step": 16514 }, { "epoch": 1.29, "learning_rate": 8.70501058574453e-05, "loss": 3.1396, "step": 16515 }, { "epoch": 1.3, "learning_rate": 8.70493217282208e-05, "loss": 2.8677, "step": 16516 }, { "epoch": 1.3, "learning_rate": 8.704853759899631e-05, "loss": 2.7571, "step": 16517 }, { "epoch": 1.3, "learning_rate": 8.704775346977183e-05, "loss": 2.8362, "step": 16518 }, { "epoch": 1.3, "learning_rate": 8.704696934054733e-05, "loss": 2.9889, "step": 16519 }, { "epoch": 1.3, "learning_rate": 8.704618521132283e-05, "loss": 2.9998, "step": 16520 }, { "epoch": 1.3, "learning_rate": 8.704540108209834e-05, "loss": 2.9858, "step": 16521 }, { "epoch": 1.3, "learning_rate": 8.704461695287384e-05, "loss": 2.7407, "step": 16522 }, { "epoch": 1.3, "learning_rate": 8.704383282364934e-05, "loss": 2.8833, "step": 16523 }, { "epoch": 1.3, "learning_rate": 8.704304869442485e-05, "loss": 2.8362, "step": 16524 }, { "epoch": 1.3, "learning_rate": 8.704226456520035e-05, "loss": 2.6081, "step": 16525 }, { "epoch": 1.3, "learning_rate": 8.704148043597586e-05, "loss": 3.0682, "step": 16526 }, { "epoch": 1.3, "learning_rate": 8.704069630675136e-05, "loss": 2.8136, "step": 16527 }, { "epoch": 1.3, "learning_rate": 8.703991217752686e-05, "loss": 3.1107, "step": 16528 }, { "epoch": 1.3, "learning_rate": 8.703912804830237e-05, "loss": 3.2166, "step": 16529 }, { "epoch": 1.3, "learning_rate": 8.703834391907787e-05, "loss": 2.9022, "step": 16530 }, { "epoch": 1.3, "learning_rate": 8.703755978985337e-05, "loss": 2.8208, "step": 16531 }, { "epoch": 1.3, "learning_rate": 8.703677566062888e-05, "loss": 2.598, "step": 16532 }, { "epoch": 1.3, "learning_rate": 8.703599153140438e-05, "loss": 3.1131, "step": 16533 }, { "epoch": 1.3, "learning_rate": 8.703520740217989e-05, "loss": 2.9447, "step": 16534 }, { "epoch": 1.3, "learning_rate": 8.703442327295539e-05, "loss": 2.8409, "step": 16535 }, { "epoch": 1.3, "learning_rate": 8.703363914373089e-05, "loss": 3.0678, "step": 16536 }, { "epoch": 1.3, "learning_rate": 8.70328550145064e-05, "loss": 2.9075, "step": 16537 }, { "epoch": 1.3, "learning_rate": 8.70320708852819e-05, "loss": 3.1074, "step": 16538 }, { "epoch": 1.3, "learning_rate": 8.70312867560574e-05, "loss": 2.8779, "step": 16539 }, { "epoch": 1.3, "learning_rate": 8.703050262683291e-05, "loss": 2.8935, "step": 16540 }, { "epoch": 1.3, "learning_rate": 8.702971849760841e-05, "loss": 2.6438, "step": 16541 }, { "epoch": 1.3, "learning_rate": 8.702893436838391e-05, "loss": 2.9974, "step": 16542 }, { "epoch": 1.3, "learning_rate": 8.702815023915942e-05, "loss": 2.8565, "step": 16543 }, { "epoch": 1.3, "learning_rate": 8.702736610993492e-05, "loss": 2.7591, "step": 16544 }, { "epoch": 1.3, "learning_rate": 8.702658198071043e-05, "loss": 3.0678, "step": 16545 }, { "epoch": 1.3, "learning_rate": 8.702579785148593e-05, "loss": 2.7996, "step": 16546 }, { "epoch": 1.3, "learning_rate": 8.702501372226143e-05, "loss": 2.56, "step": 16547 }, { "epoch": 1.3, "learning_rate": 8.702422959303694e-05, "loss": 3.0218, "step": 16548 }, { "epoch": 1.3, "learning_rate": 8.702344546381244e-05, "loss": 2.971, "step": 16549 }, { "epoch": 1.3, "learning_rate": 8.702266133458794e-05, "loss": 3.0905, "step": 16550 }, { "epoch": 1.3, "learning_rate": 8.702187720536345e-05, "loss": 2.7725, "step": 16551 }, { "epoch": 1.3, "learning_rate": 8.702109307613895e-05, "loss": 2.8944, "step": 16552 }, { "epoch": 1.3, "learning_rate": 8.702030894691446e-05, "loss": 2.9981, "step": 16553 }, { "epoch": 1.3, "learning_rate": 8.701952481768996e-05, "loss": 2.9505, "step": 16554 }, { "epoch": 1.3, "learning_rate": 8.701874068846546e-05, "loss": 2.8849, "step": 16555 }, { "epoch": 1.3, "learning_rate": 8.701795655924097e-05, "loss": 2.8739, "step": 16556 }, { "epoch": 1.3, "learning_rate": 8.701717243001647e-05, "loss": 3.0991, "step": 16557 }, { "epoch": 1.3, "learning_rate": 8.701638830079197e-05, "loss": 2.9072, "step": 16558 }, { "epoch": 1.3, "learning_rate": 8.701560417156748e-05, "loss": 3.0211, "step": 16559 }, { "epoch": 1.3, "learning_rate": 8.701482004234298e-05, "loss": 2.8887, "step": 16560 }, { "epoch": 1.3, "learning_rate": 8.701403591311848e-05, "loss": 2.8747, "step": 16561 }, { "epoch": 1.3, "learning_rate": 8.701325178389399e-05, "loss": 2.8092, "step": 16562 }, { "epoch": 1.3, "learning_rate": 8.701246765466949e-05, "loss": 3.0697, "step": 16563 }, { "epoch": 1.3, "learning_rate": 8.7011683525445e-05, "loss": 2.6978, "step": 16564 }, { "epoch": 1.3, "learning_rate": 8.70108993962205e-05, "loss": 2.9085, "step": 16565 }, { "epoch": 1.3, "learning_rate": 8.7010115266996e-05, "loss": 2.9299, "step": 16566 }, { "epoch": 1.3, "learning_rate": 8.70093311377715e-05, "loss": 3.2321, "step": 16567 }, { "epoch": 1.3, "learning_rate": 8.700854700854701e-05, "loss": 2.9008, "step": 16568 }, { "epoch": 1.3, "learning_rate": 8.700776287932251e-05, "loss": 2.8929, "step": 16569 }, { "epoch": 1.3, "learning_rate": 8.700697875009802e-05, "loss": 2.7422, "step": 16570 }, { "epoch": 1.3, "learning_rate": 8.700619462087352e-05, "loss": 2.8932, "step": 16571 }, { "epoch": 1.3, "learning_rate": 8.700541049164903e-05, "loss": 2.7896, "step": 16572 }, { "epoch": 1.3, "learning_rate": 8.700462636242453e-05, "loss": 3.1443, "step": 16573 }, { "epoch": 1.3, "learning_rate": 8.700384223320003e-05, "loss": 2.7415, "step": 16574 }, { "epoch": 1.3, "learning_rate": 8.700305810397555e-05, "loss": 2.7438, "step": 16575 }, { "epoch": 1.3, "learning_rate": 8.700227397475105e-05, "loss": 3.0212, "step": 16576 }, { "epoch": 1.3, "learning_rate": 8.700148984552656e-05, "loss": 3.0473, "step": 16577 }, { "epoch": 1.3, "learning_rate": 8.700070571630206e-05, "loss": 3.0648, "step": 16578 }, { "epoch": 1.3, "learning_rate": 8.699992158707756e-05, "loss": 3.1562, "step": 16579 }, { "epoch": 1.3, "learning_rate": 8.699913745785307e-05, "loss": 2.997, "step": 16580 }, { "epoch": 1.3, "learning_rate": 8.699835332862856e-05, "loss": 2.9656, "step": 16581 }, { "epoch": 1.3, "learning_rate": 8.699756919940406e-05, "loss": 3.1941, "step": 16582 }, { "epoch": 1.3, "learning_rate": 8.699678507017957e-05, "loss": 2.893, "step": 16583 }, { "epoch": 1.3, "learning_rate": 8.699600094095507e-05, "loss": 2.875, "step": 16584 }, { "epoch": 1.3, "learning_rate": 8.699521681173057e-05, "loss": 3.1703, "step": 16585 }, { "epoch": 1.3, "learning_rate": 8.699443268250608e-05, "loss": 3.3014, "step": 16586 }, { "epoch": 1.3, "learning_rate": 8.699364855328158e-05, "loss": 2.8969, "step": 16587 }, { "epoch": 1.3, "learning_rate": 8.699286442405708e-05, "loss": 3.2075, "step": 16588 }, { "epoch": 1.3, "learning_rate": 8.699208029483259e-05, "loss": 3.085, "step": 16589 }, { "epoch": 1.3, "learning_rate": 8.699129616560809e-05, "loss": 2.8843, "step": 16590 }, { "epoch": 1.3, "learning_rate": 8.69905120363836e-05, "loss": 2.891, "step": 16591 }, { "epoch": 1.3, "learning_rate": 8.69897279071591e-05, "loss": 2.9454, "step": 16592 }, { "epoch": 1.3, "learning_rate": 8.69889437779346e-05, "loss": 2.7707, "step": 16593 }, { "epoch": 1.3, "learning_rate": 8.69881596487101e-05, "loss": 3.0616, "step": 16594 }, { "epoch": 1.3, "learning_rate": 8.698737551948561e-05, "loss": 3.2494, "step": 16595 }, { "epoch": 1.3, "learning_rate": 8.698659139026111e-05, "loss": 2.9811, "step": 16596 }, { "epoch": 1.3, "learning_rate": 8.698580726103662e-05, "loss": 2.7518, "step": 16597 }, { "epoch": 1.3, "learning_rate": 8.698502313181212e-05, "loss": 3.1078, "step": 16598 }, { "epoch": 1.3, "learning_rate": 8.698423900258762e-05, "loss": 2.7709, "step": 16599 }, { "epoch": 1.3, "learning_rate": 8.698345487336314e-05, "loss": 2.7265, "step": 16600 }, { "epoch": 1.3, "learning_rate": 8.698267074413865e-05, "loss": 3.0156, "step": 16601 }, { "epoch": 1.3, "learning_rate": 8.698188661491415e-05, "loss": 2.8021, "step": 16602 }, { "epoch": 1.3, "learning_rate": 8.698110248568965e-05, "loss": 2.9327, "step": 16603 }, { "epoch": 1.3, "learning_rate": 8.698031835646516e-05, "loss": 3.0344, "step": 16604 }, { "epoch": 1.3, "learning_rate": 8.697953422724066e-05, "loss": 2.7814, "step": 16605 }, { "epoch": 1.3, "learning_rate": 8.697875009801616e-05, "loss": 2.9083, "step": 16606 }, { "epoch": 1.3, "learning_rate": 8.697796596879167e-05, "loss": 2.9864, "step": 16607 }, { "epoch": 1.3, "learning_rate": 8.697718183956717e-05, "loss": 2.8188, "step": 16608 }, { "epoch": 1.3, "learning_rate": 8.697639771034267e-05, "loss": 3.0164, "step": 16609 }, { "epoch": 1.3, "learning_rate": 8.697561358111818e-05, "loss": 3.2445, "step": 16610 }, { "epoch": 1.3, "learning_rate": 8.697482945189368e-05, "loss": 2.8387, "step": 16611 }, { "epoch": 1.3, "learning_rate": 8.697404532266919e-05, "loss": 2.8783, "step": 16612 }, { "epoch": 1.3, "learning_rate": 8.697326119344469e-05, "loss": 2.9337, "step": 16613 }, { "epoch": 1.3, "learning_rate": 8.697247706422019e-05, "loss": 3.0497, "step": 16614 }, { "epoch": 1.3, "learning_rate": 8.69716929349957e-05, "loss": 2.7607, "step": 16615 }, { "epoch": 1.3, "learning_rate": 8.69709088057712e-05, "loss": 2.9793, "step": 16616 }, { "epoch": 1.3, "learning_rate": 8.697012467654669e-05, "loss": 3.0153, "step": 16617 }, { "epoch": 1.3, "learning_rate": 8.69693405473222e-05, "loss": 3.4198, "step": 16618 }, { "epoch": 1.3, "learning_rate": 8.69685564180977e-05, "loss": 3.0549, "step": 16619 }, { "epoch": 1.3, "learning_rate": 8.69677722888732e-05, "loss": 2.7199, "step": 16620 }, { "epoch": 1.3, "learning_rate": 8.69669881596487e-05, "loss": 2.8811, "step": 16621 }, { "epoch": 1.3, "learning_rate": 8.696620403042421e-05, "loss": 3.0, "step": 16622 }, { "epoch": 1.3, "learning_rate": 8.696541990119971e-05, "loss": 3.0324, "step": 16623 }, { "epoch": 1.3, "learning_rate": 8.696463577197522e-05, "loss": 2.8854, "step": 16624 }, { "epoch": 1.3, "learning_rate": 8.696385164275072e-05, "loss": 2.7714, "step": 16625 }, { "epoch": 1.3, "learning_rate": 8.696306751352624e-05, "loss": 2.9277, "step": 16626 }, { "epoch": 1.3, "learning_rate": 8.696228338430174e-05, "loss": 2.868, "step": 16627 }, { "epoch": 1.3, "learning_rate": 8.696149925507724e-05, "loss": 2.8537, "step": 16628 }, { "epoch": 1.3, "learning_rate": 8.696071512585275e-05, "loss": 2.9559, "step": 16629 }, { "epoch": 1.3, "learning_rate": 8.695993099662825e-05, "loss": 2.978, "step": 16630 }, { "epoch": 1.3, "learning_rate": 8.695914686740376e-05, "loss": 2.8379, "step": 16631 }, { "epoch": 1.3, "learning_rate": 8.695836273817926e-05, "loss": 2.9539, "step": 16632 }, { "epoch": 1.3, "learning_rate": 8.695757860895476e-05, "loss": 2.8472, "step": 16633 }, { "epoch": 1.3, "learning_rate": 8.695679447973027e-05, "loss": 2.8029, "step": 16634 }, { "epoch": 1.3, "learning_rate": 8.695601035050577e-05, "loss": 2.7926, "step": 16635 }, { "epoch": 1.3, "learning_rate": 8.695522622128127e-05, "loss": 2.7652, "step": 16636 }, { "epoch": 1.3, "learning_rate": 8.695444209205678e-05, "loss": 2.7544, "step": 16637 }, { "epoch": 1.3, "learning_rate": 8.695365796283228e-05, "loss": 2.8881, "step": 16638 }, { "epoch": 1.3, "learning_rate": 8.695287383360779e-05, "loss": 2.9462, "step": 16639 }, { "epoch": 1.3, "learning_rate": 8.695208970438329e-05, "loss": 3.1142, "step": 16640 }, { "epoch": 1.3, "learning_rate": 8.695130557515879e-05, "loss": 2.9895, "step": 16641 }, { "epoch": 1.3, "learning_rate": 8.69505214459343e-05, "loss": 2.9914, "step": 16642 }, { "epoch": 1.31, "learning_rate": 8.69497373167098e-05, "loss": 2.9432, "step": 16643 }, { "epoch": 1.31, "learning_rate": 8.69489531874853e-05, "loss": 2.9311, "step": 16644 }, { "epoch": 1.31, "learning_rate": 8.694816905826081e-05, "loss": 2.8245, "step": 16645 }, { "epoch": 1.31, "learning_rate": 8.694738492903631e-05, "loss": 2.9954, "step": 16646 }, { "epoch": 1.31, "learning_rate": 8.694660079981181e-05, "loss": 3.0283, "step": 16647 }, { "epoch": 1.31, "learning_rate": 8.694581667058732e-05, "loss": 2.9685, "step": 16648 }, { "epoch": 1.31, "learning_rate": 8.694503254136282e-05, "loss": 3.089, "step": 16649 }, { "epoch": 1.31, "learning_rate": 8.694424841213833e-05, "loss": 2.7459, "step": 16650 }, { "epoch": 1.31, "learning_rate": 8.694346428291383e-05, "loss": 2.9114, "step": 16651 }, { "epoch": 1.31, "learning_rate": 8.694268015368933e-05, "loss": 2.8814, "step": 16652 }, { "epoch": 1.31, "learning_rate": 8.694189602446484e-05, "loss": 2.9212, "step": 16653 }, { "epoch": 1.31, "learning_rate": 8.694111189524034e-05, "loss": 2.9494, "step": 16654 }, { "epoch": 1.31, "learning_rate": 8.694032776601584e-05, "loss": 2.8097, "step": 16655 }, { "epoch": 1.31, "learning_rate": 8.693954363679135e-05, "loss": 2.8896, "step": 16656 }, { "epoch": 1.31, "learning_rate": 8.693875950756685e-05, "loss": 2.7266, "step": 16657 }, { "epoch": 1.31, "learning_rate": 8.693797537834235e-05, "loss": 2.8223, "step": 16658 }, { "epoch": 1.31, "learning_rate": 8.693719124911786e-05, "loss": 2.8575, "step": 16659 }, { "epoch": 1.31, "learning_rate": 8.693640711989336e-05, "loss": 2.7623, "step": 16660 }, { "epoch": 1.31, "learning_rate": 8.693562299066887e-05, "loss": 2.9242, "step": 16661 }, { "epoch": 1.31, "learning_rate": 8.693483886144437e-05, "loss": 2.968, "step": 16662 }, { "epoch": 1.31, "learning_rate": 8.693405473221987e-05, "loss": 2.9552, "step": 16663 }, { "epoch": 1.31, "learning_rate": 8.693327060299538e-05, "loss": 2.7389, "step": 16664 }, { "epoch": 1.31, "learning_rate": 8.693248647377088e-05, "loss": 2.9649, "step": 16665 }, { "epoch": 1.31, "learning_rate": 8.693170234454638e-05, "loss": 2.9201, "step": 16666 }, { "epoch": 1.31, "learning_rate": 8.693091821532189e-05, "loss": 2.7572, "step": 16667 }, { "epoch": 1.31, "learning_rate": 8.693013408609739e-05, "loss": 3.1278, "step": 16668 }, { "epoch": 1.31, "learning_rate": 8.69293499568729e-05, "loss": 2.8882, "step": 16669 }, { "epoch": 1.31, "learning_rate": 8.69285658276484e-05, "loss": 2.8168, "step": 16670 }, { "epoch": 1.31, "learning_rate": 8.69277816984239e-05, "loss": 3.0692, "step": 16671 }, { "epoch": 1.31, "learning_rate": 8.69269975691994e-05, "loss": 2.8584, "step": 16672 }, { "epoch": 1.31, "learning_rate": 8.692621343997491e-05, "loss": 3.0502, "step": 16673 }, { "epoch": 1.31, "learning_rate": 8.692542931075041e-05, "loss": 2.978, "step": 16674 }, { "epoch": 1.31, "learning_rate": 8.692464518152592e-05, "loss": 2.7022, "step": 16675 }, { "epoch": 1.31, "learning_rate": 8.692386105230142e-05, "loss": 3.136, "step": 16676 }, { "epoch": 1.31, "learning_rate": 8.692307692307692e-05, "loss": 2.9445, "step": 16677 }, { "epoch": 1.31, "learning_rate": 8.692229279385243e-05, "loss": 3.0051, "step": 16678 }, { "epoch": 1.31, "learning_rate": 8.692150866462793e-05, "loss": 2.8147, "step": 16679 }, { "epoch": 1.31, "learning_rate": 8.692072453540344e-05, "loss": 2.7314, "step": 16680 }, { "epoch": 1.31, "learning_rate": 8.691994040617894e-05, "loss": 2.9397, "step": 16681 }, { "epoch": 1.31, "learning_rate": 8.691915627695444e-05, "loss": 2.9391, "step": 16682 }, { "epoch": 1.31, "learning_rate": 8.691837214772996e-05, "loss": 2.8025, "step": 16683 }, { "epoch": 1.31, "learning_rate": 8.691758801850546e-05, "loss": 2.7577, "step": 16684 }, { "epoch": 1.31, "learning_rate": 8.691680388928097e-05, "loss": 3.1901, "step": 16685 }, { "epoch": 1.31, "learning_rate": 8.691601976005647e-05, "loss": 2.8929, "step": 16686 }, { "epoch": 1.31, "learning_rate": 8.691523563083198e-05, "loss": 2.7656, "step": 16687 }, { "epoch": 1.31, "learning_rate": 8.691445150160748e-05, "loss": 2.9836, "step": 16688 }, { "epoch": 1.31, "learning_rate": 8.691366737238297e-05, "loss": 2.9459, "step": 16689 }, { "epoch": 1.31, "learning_rate": 8.691288324315847e-05, "loss": 2.8373, "step": 16690 }, { "epoch": 1.31, "learning_rate": 8.691209911393398e-05, "loss": 3.0235, "step": 16691 }, { "epoch": 1.31, "learning_rate": 8.691131498470948e-05, "loss": 3.2046, "step": 16692 }, { "epoch": 1.31, "learning_rate": 8.691053085548498e-05, "loss": 2.6616, "step": 16693 }, { "epoch": 1.31, "learning_rate": 8.690974672626049e-05, "loss": 2.8754, "step": 16694 }, { "epoch": 1.31, "learning_rate": 8.690896259703599e-05, "loss": 2.7761, "step": 16695 }, { "epoch": 1.31, "learning_rate": 8.69081784678115e-05, "loss": 2.9733, "step": 16696 }, { "epoch": 1.31, "learning_rate": 8.6907394338587e-05, "loss": 3.1193, "step": 16697 }, { "epoch": 1.31, "learning_rate": 8.69066102093625e-05, "loss": 2.827, "step": 16698 }, { "epoch": 1.31, "learning_rate": 8.6905826080138e-05, "loss": 2.907, "step": 16699 }, { "epoch": 1.31, "learning_rate": 8.690504195091351e-05, "loss": 2.8804, "step": 16700 }, { "epoch": 1.31, "learning_rate": 8.690425782168901e-05, "loss": 3.0688, "step": 16701 }, { "epoch": 1.31, "learning_rate": 8.690347369246452e-05, "loss": 2.8714, "step": 16702 }, { "epoch": 1.31, "learning_rate": 8.690268956324002e-05, "loss": 2.8905, "step": 16703 }, { "epoch": 1.31, "learning_rate": 8.690190543401552e-05, "loss": 2.8659, "step": 16704 }, { "epoch": 1.31, "learning_rate": 8.690112130479103e-05, "loss": 2.8819, "step": 16705 }, { "epoch": 1.31, "learning_rate": 8.690033717556653e-05, "loss": 3.1563, "step": 16706 }, { "epoch": 1.31, "learning_rate": 8.689955304634204e-05, "loss": 2.6569, "step": 16707 }, { "epoch": 1.31, "learning_rate": 8.689876891711755e-05, "loss": 2.8964, "step": 16708 }, { "epoch": 1.31, "learning_rate": 8.689798478789306e-05, "loss": 2.893, "step": 16709 }, { "epoch": 1.31, "learning_rate": 8.689720065866856e-05, "loss": 3.0854, "step": 16710 }, { "epoch": 1.31, "learning_rate": 8.689641652944406e-05, "loss": 2.8519, "step": 16711 }, { "epoch": 1.31, "learning_rate": 8.689563240021957e-05, "loss": 2.7544, "step": 16712 }, { "epoch": 1.31, "learning_rate": 8.689484827099507e-05, "loss": 2.8072, "step": 16713 }, { "epoch": 1.31, "learning_rate": 8.689406414177057e-05, "loss": 2.7192, "step": 16714 }, { "epoch": 1.31, "learning_rate": 8.689328001254608e-05, "loss": 2.8518, "step": 16715 }, { "epoch": 1.31, "learning_rate": 8.689249588332158e-05, "loss": 2.7214, "step": 16716 }, { "epoch": 1.31, "learning_rate": 8.689171175409709e-05, "loss": 2.9399, "step": 16717 }, { "epoch": 1.31, "learning_rate": 8.689092762487259e-05, "loss": 2.9919, "step": 16718 }, { "epoch": 1.31, "learning_rate": 8.689014349564809e-05, "loss": 2.8662, "step": 16719 }, { "epoch": 1.31, "learning_rate": 8.68893593664236e-05, "loss": 3.0579, "step": 16720 }, { "epoch": 1.31, "learning_rate": 8.68885752371991e-05, "loss": 2.6419, "step": 16721 }, { "epoch": 1.31, "learning_rate": 8.68877911079746e-05, "loss": 2.7823, "step": 16722 }, { "epoch": 1.31, "learning_rate": 8.688700697875011e-05, "loss": 3.089, "step": 16723 }, { "epoch": 1.31, "learning_rate": 8.688622284952561e-05, "loss": 2.9106, "step": 16724 }, { "epoch": 1.31, "learning_rate": 8.68854387203011e-05, "loss": 3.4931, "step": 16725 }, { "epoch": 1.31, "learning_rate": 8.68846545910766e-05, "loss": 3.0714, "step": 16726 }, { "epoch": 1.31, "learning_rate": 8.688387046185211e-05, "loss": 2.7912, "step": 16727 }, { "epoch": 1.31, "learning_rate": 8.688308633262761e-05, "loss": 3.0936, "step": 16728 }, { "epoch": 1.31, "learning_rate": 8.688230220340312e-05, "loss": 2.6442, "step": 16729 }, { "epoch": 1.31, "learning_rate": 8.688151807417862e-05, "loss": 2.8592, "step": 16730 }, { "epoch": 1.31, "learning_rate": 8.688073394495412e-05, "loss": 2.9652, "step": 16731 }, { "epoch": 1.31, "learning_rate": 8.687994981572963e-05, "loss": 3.1003, "step": 16732 }, { "epoch": 1.31, "learning_rate": 8.687916568650513e-05, "loss": 3.1719, "step": 16733 }, { "epoch": 1.31, "learning_rate": 8.687838155728065e-05, "loss": 2.9752, "step": 16734 }, { "epoch": 1.31, "learning_rate": 8.687759742805615e-05, "loss": 2.6634, "step": 16735 }, { "epoch": 1.31, "learning_rate": 8.687681329883166e-05, "loss": 2.8658, "step": 16736 }, { "epoch": 1.31, "learning_rate": 8.687602916960716e-05, "loss": 2.7862, "step": 16737 }, { "epoch": 1.31, "learning_rate": 8.687524504038266e-05, "loss": 2.8231, "step": 16738 }, { "epoch": 1.31, "learning_rate": 8.687446091115817e-05, "loss": 2.7188, "step": 16739 }, { "epoch": 1.31, "learning_rate": 8.687367678193367e-05, "loss": 2.9403, "step": 16740 }, { "epoch": 1.31, "learning_rate": 8.687289265270917e-05, "loss": 3.1483, "step": 16741 }, { "epoch": 1.31, "learning_rate": 8.687210852348468e-05, "loss": 2.8099, "step": 16742 }, { "epoch": 1.31, "learning_rate": 8.687132439426018e-05, "loss": 2.9251, "step": 16743 }, { "epoch": 1.31, "learning_rate": 8.687054026503568e-05, "loss": 3.0759, "step": 16744 }, { "epoch": 1.31, "learning_rate": 8.686975613581119e-05, "loss": 2.9552, "step": 16745 }, { "epoch": 1.31, "learning_rate": 8.686897200658669e-05, "loss": 2.8011, "step": 16746 }, { "epoch": 1.31, "learning_rate": 8.68681878773622e-05, "loss": 2.8819, "step": 16747 }, { "epoch": 1.31, "learning_rate": 8.68674037481377e-05, "loss": 3.1237, "step": 16748 }, { "epoch": 1.31, "learning_rate": 8.68666196189132e-05, "loss": 2.8486, "step": 16749 }, { "epoch": 1.31, "learning_rate": 8.686583548968871e-05, "loss": 2.9694, "step": 16750 }, { "epoch": 1.31, "learning_rate": 8.686505136046421e-05, "loss": 2.8405, "step": 16751 }, { "epoch": 1.31, "learning_rate": 8.686426723123971e-05, "loss": 2.8492, "step": 16752 }, { "epoch": 1.31, "learning_rate": 8.686348310201522e-05, "loss": 2.8153, "step": 16753 }, { "epoch": 1.31, "learning_rate": 8.686269897279072e-05, "loss": 2.9629, "step": 16754 }, { "epoch": 1.31, "learning_rate": 8.686191484356623e-05, "loss": 2.874, "step": 16755 }, { "epoch": 1.31, "learning_rate": 8.686113071434173e-05, "loss": 2.7204, "step": 16756 }, { "epoch": 1.31, "learning_rate": 8.686034658511723e-05, "loss": 2.7059, "step": 16757 }, { "epoch": 1.31, "learning_rate": 8.685956245589274e-05, "loss": 2.92, "step": 16758 }, { "epoch": 1.31, "learning_rate": 8.685877832666824e-05, "loss": 2.7886, "step": 16759 }, { "epoch": 1.31, "learning_rate": 8.685799419744374e-05, "loss": 2.9311, "step": 16760 }, { "epoch": 1.31, "learning_rate": 8.685721006821925e-05, "loss": 2.848, "step": 16761 }, { "epoch": 1.31, "learning_rate": 8.685642593899475e-05, "loss": 2.8048, "step": 16762 }, { "epoch": 1.31, "learning_rate": 8.685564180977025e-05, "loss": 2.7586, "step": 16763 }, { "epoch": 1.31, "learning_rate": 8.685485768054576e-05, "loss": 3.0851, "step": 16764 }, { "epoch": 1.31, "learning_rate": 8.685407355132126e-05, "loss": 3.0252, "step": 16765 }, { "epoch": 1.31, "learning_rate": 8.685328942209677e-05, "loss": 2.7083, "step": 16766 }, { "epoch": 1.31, "learning_rate": 8.685250529287227e-05, "loss": 2.8183, "step": 16767 }, { "epoch": 1.31, "learning_rate": 8.685172116364777e-05, "loss": 3.0319, "step": 16768 }, { "epoch": 1.31, "learning_rate": 8.685093703442328e-05, "loss": 2.8919, "step": 16769 }, { "epoch": 1.31, "learning_rate": 8.685015290519878e-05, "loss": 2.7495, "step": 16770 }, { "epoch": 1.32, "learning_rate": 8.684936877597428e-05, "loss": 2.7115, "step": 16771 }, { "epoch": 1.32, "learning_rate": 8.684858464674979e-05, "loss": 2.8291, "step": 16772 }, { "epoch": 1.32, "learning_rate": 8.684780051752529e-05, "loss": 2.774, "step": 16773 }, { "epoch": 1.32, "learning_rate": 8.68470163883008e-05, "loss": 2.827, "step": 16774 }, { "epoch": 1.32, "learning_rate": 8.68462322590763e-05, "loss": 3.2537, "step": 16775 }, { "epoch": 1.32, "learning_rate": 8.68454481298518e-05, "loss": 2.714, "step": 16776 }, { "epoch": 1.32, "learning_rate": 8.68446640006273e-05, "loss": 2.8529, "step": 16777 }, { "epoch": 1.32, "learning_rate": 8.684387987140281e-05, "loss": 2.8846, "step": 16778 }, { "epoch": 1.32, "learning_rate": 8.684309574217831e-05, "loss": 2.8715, "step": 16779 }, { "epoch": 1.32, "learning_rate": 8.684231161295382e-05, "loss": 2.9361, "step": 16780 }, { "epoch": 1.32, "learning_rate": 8.684152748372932e-05, "loss": 3.0577, "step": 16781 }, { "epoch": 1.32, "learning_rate": 8.684074335450482e-05, "loss": 2.8556, "step": 16782 }, { "epoch": 1.32, "learning_rate": 8.683995922528033e-05, "loss": 3.0522, "step": 16783 }, { "epoch": 1.32, "learning_rate": 8.683917509605583e-05, "loss": 3.0477, "step": 16784 }, { "epoch": 1.32, "learning_rate": 8.683839096683134e-05, "loss": 2.8918, "step": 16785 }, { "epoch": 1.32, "learning_rate": 8.683760683760684e-05, "loss": 2.7316, "step": 16786 }, { "epoch": 1.32, "learning_rate": 8.683682270838234e-05, "loss": 2.8983, "step": 16787 }, { "epoch": 1.32, "learning_rate": 8.683603857915785e-05, "loss": 2.8751, "step": 16788 }, { "epoch": 1.32, "learning_rate": 8.683525444993335e-05, "loss": 2.8016, "step": 16789 }, { "epoch": 1.32, "learning_rate": 8.683447032070885e-05, "loss": 3.1146, "step": 16790 }, { "epoch": 1.32, "learning_rate": 8.683368619148437e-05, "loss": 2.8619, "step": 16791 }, { "epoch": 1.32, "learning_rate": 8.683290206225987e-05, "loss": 2.6643, "step": 16792 }, { "epoch": 1.32, "learning_rate": 8.683211793303538e-05, "loss": 3.1694, "step": 16793 }, { "epoch": 1.32, "learning_rate": 8.683133380381088e-05, "loss": 2.8844, "step": 16794 }, { "epoch": 1.32, "learning_rate": 8.683054967458639e-05, "loss": 3.115, "step": 16795 }, { "epoch": 1.32, "learning_rate": 8.682976554536189e-05, "loss": 2.8572, "step": 16796 }, { "epoch": 1.32, "learning_rate": 8.682898141613738e-05, "loss": 2.7817, "step": 16797 }, { "epoch": 1.32, "learning_rate": 8.682819728691288e-05, "loss": 3.1841, "step": 16798 }, { "epoch": 1.32, "learning_rate": 8.682741315768839e-05, "loss": 3.0012, "step": 16799 }, { "epoch": 1.32, "learning_rate": 8.682662902846389e-05, "loss": 2.7899, "step": 16800 }, { "epoch": 1.32, "learning_rate": 8.68258448992394e-05, "loss": 2.895, "step": 16801 }, { "epoch": 1.32, "learning_rate": 8.68250607700149e-05, "loss": 2.632, "step": 16802 }, { "epoch": 1.32, "learning_rate": 8.68242766407904e-05, "loss": 2.866, "step": 16803 }, { "epoch": 1.32, "learning_rate": 8.68234925115659e-05, "loss": 2.8827, "step": 16804 }, { "epoch": 1.32, "learning_rate": 8.682270838234141e-05, "loss": 3.0132, "step": 16805 }, { "epoch": 1.32, "learning_rate": 8.682192425311691e-05, "loss": 2.9428, "step": 16806 }, { "epoch": 1.32, "learning_rate": 8.682114012389242e-05, "loss": 2.9928, "step": 16807 }, { "epoch": 1.32, "learning_rate": 8.682035599466792e-05, "loss": 2.6568, "step": 16808 }, { "epoch": 1.32, "learning_rate": 8.681957186544342e-05, "loss": 3.0179, "step": 16809 }, { "epoch": 1.32, "learning_rate": 8.681878773621893e-05, "loss": 3.0764, "step": 16810 }, { "epoch": 1.32, "learning_rate": 8.681800360699443e-05, "loss": 2.877, "step": 16811 }, { "epoch": 1.32, "learning_rate": 8.681721947776993e-05, "loss": 2.7008, "step": 16812 }, { "epoch": 1.32, "learning_rate": 8.681643534854544e-05, "loss": 3.051, "step": 16813 }, { "epoch": 1.32, "learning_rate": 8.681565121932094e-05, "loss": 2.6707, "step": 16814 }, { "epoch": 1.32, "learning_rate": 8.681486709009645e-05, "loss": 3.0513, "step": 16815 }, { "epoch": 1.32, "learning_rate": 8.681408296087195e-05, "loss": 2.855, "step": 16816 }, { "epoch": 1.32, "learning_rate": 8.681329883164747e-05, "loss": 2.8804, "step": 16817 }, { "epoch": 1.32, "learning_rate": 8.681251470242297e-05, "loss": 3.0515, "step": 16818 }, { "epoch": 1.32, "learning_rate": 8.681173057319847e-05, "loss": 2.9167, "step": 16819 }, { "epoch": 1.32, "learning_rate": 8.681094644397398e-05, "loss": 2.538, "step": 16820 }, { "epoch": 1.32, "learning_rate": 8.681016231474948e-05, "loss": 2.861, "step": 16821 }, { "epoch": 1.32, "learning_rate": 8.680937818552499e-05, "loss": 2.9052, "step": 16822 }, { "epoch": 1.32, "learning_rate": 8.680859405630049e-05, "loss": 2.8533, "step": 16823 }, { "epoch": 1.32, "learning_rate": 8.680780992707599e-05, "loss": 2.915, "step": 16824 }, { "epoch": 1.32, "learning_rate": 8.68070257978515e-05, "loss": 2.7965, "step": 16825 }, { "epoch": 1.32, "learning_rate": 8.6806241668627e-05, "loss": 2.8636, "step": 16826 }, { "epoch": 1.32, "learning_rate": 8.68054575394025e-05, "loss": 2.9048, "step": 16827 }, { "epoch": 1.32, "learning_rate": 8.680467341017801e-05, "loss": 2.9631, "step": 16828 }, { "epoch": 1.32, "learning_rate": 8.680388928095351e-05, "loss": 2.893, "step": 16829 }, { "epoch": 1.32, "learning_rate": 8.680310515172901e-05, "loss": 2.7329, "step": 16830 }, { "epoch": 1.32, "learning_rate": 8.680232102250452e-05, "loss": 2.9973, "step": 16831 }, { "epoch": 1.32, "learning_rate": 8.680153689328002e-05, "loss": 3.0207, "step": 16832 }, { "epoch": 1.32, "learning_rate": 8.680075276405551e-05, "loss": 2.7943, "step": 16833 }, { "epoch": 1.32, "learning_rate": 8.679996863483102e-05, "loss": 2.7808, "step": 16834 }, { "epoch": 1.32, "learning_rate": 8.679918450560652e-05, "loss": 2.9232, "step": 16835 }, { "epoch": 1.32, "learning_rate": 8.679840037638202e-05, "loss": 2.9628, "step": 16836 }, { "epoch": 1.32, "learning_rate": 8.679761624715753e-05, "loss": 3.1537, "step": 16837 }, { "epoch": 1.32, "learning_rate": 8.679683211793303e-05, "loss": 2.935, "step": 16838 }, { "epoch": 1.32, "learning_rate": 8.679604798870853e-05, "loss": 2.9593, "step": 16839 }, { "epoch": 1.32, "learning_rate": 8.679526385948404e-05, "loss": 3.0965, "step": 16840 }, { "epoch": 1.32, "learning_rate": 8.679447973025954e-05, "loss": 2.7646, "step": 16841 }, { "epoch": 1.32, "learning_rate": 8.679369560103506e-05, "loss": 2.9768, "step": 16842 }, { "epoch": 1.32, "learning_rate": 8.679291147181056e-05, "loss": 2.9621, "step": 16843 }, { "epoch": 1.32, "learning_rate": 8.679212734258607e-05, "loss": 3.0918, "step": 16844 }, { "epoch": 1.32, "learning_rate": 8.679134321336157e-05, "loss": 2.9438, "step": 16845 }, { "epoch": 1.32, "learning_rate": 8.679055908413707e-05, "loss": 2.7871, "step": 16846 }, { "epoch": 1.32, "learning_rate": 8.678977495491258e-05, "loss": 2.8585, "step": 16847 }, { "epoch": 1.32, "learning_rate": 8.678899082568808e-05, "loss": 2.9042, "step": 16848 }, { "epoch": 1.32, "learning_rate": 8.678820669646358e-05, "loss": 2.7807, "step": 16849 }, { "epoch": 1.32, "learning_rate": 8.678742256723909e-05, "loss": 3.0746, "step": 16850 }, { "epoch": 1.32, "learning_rate": 8.678663843801459e-05, "loss": 3.0698, "step": 16851 }, { "epoch": 1.32, "learning_rate": 8.67858543087901e-05, "loss": 2.6967, "step": 16852 }, { "epoch": 1.32, "learning_rate": 8.67850701795656e-05, "loss": 3.0725, "step": 16853 }, { "epoch": 1.32, "learning_rate": 8.67842860503411e-05, "loss": 2.8671, "step": 16854 }, { "epoch": 1.32, "learning_rate": 8.67835019211166e-05, "loss": 2.8701, "step": 16855 }, { "epoch": 1.32, "learning_rate": 8.678271779189211e-05, "loss": 2.7078, "step": 16856 }, { "epoch": 1.32, "learning_rate": 8.678193366266761e-05, "loss": 2.9838, "step": 16857 }, { "epoch": 1.32, "learning_rate": 8.678114953344312e-05, "loss": 2.9753, "step": 16858 }, { "epoch": 1.32, "learning_rate": 8.678036540421862e-05, "loss": 2.9007, "step": 16859 }, { "epoch": 1.32, "learning_rate": 8.677958127499412e-05, "loss": 3.002, "step": 16860 }, { "epoch": 1.32, "learning_rate": 8.677879714576963e-05, "loss": 3.0653, "step": 16861 }, { "epoch": 1.32, "learning_rate": 8.677801301654513e-05, "loss": 2.7879, "step": 16862 }, { "epoch": 1.32, "learning_rate": 8.677722888732064e-05, "loss": 3.2793, "step": 16863 }, { "epoch": 1.32, "learning_rate": 8.677644475809614e-05, "loss": 2.8852, "step": 16864 }, { "epoch": 1.32, "learning_rate": 8.677566062887164e-05, "loss": 2.9303, "step": 16865 }, { "epoch": 1.32, "learning_rate": 8.677487649964715e-05, "loss": 2.9111, "step": 16866 }, { "epoch": 1.32, "learning_rate": 8.677409237042265e-05, "loss": 2.8934, "step": 16867 }, { "epoch": 1.32, "learning_rate": 8.677330824119815e-05, "loss": 3.026, "step": 16868 }, { "epoch": 1.32, "learning_rate": 8.677252411197366e-05, "loss": 2.9833, "step": 16869 }, { "epoch": 1.32, "learning_rate": 8.677173998274916e-05, "loss": 2.5933, "step": 16870 }, { "epoch": 1.32, "learning_rate": 8.677095585352467e-05, "loss": 2.7556, "step": 16871 }, { "epoch": 1.32, "learning_rate": 8.677017172430017e-05, "loss": 2.8343, "step": 16872 }, { "epoch": 1.32, "learning_rate": 8.676938759507567e-05, "loss": 2.9173, "step": 16873 }, { "epoch": 1.32, "learning_rate": 8.676860346585118e-05, "loss": 2.7303, "step": 16874 }, { "epoch": 1.32, "learning_rate": 8.676781933662668e-05, "loss": 2.8997, "step": 16875 }, { "epoch": 1.32, "learning_rate": 8.676703520740218e-05, "loss": 2.9311, "step": 16876 }, { "epoch": 1.32, "learning_rate": 8.676625107817769e-05, "loss": 2.9514, "step": 16877 }, { "epoch": 1.32, "learning_rate": 8.676546694895319e-05, "loss": 2.8679, "step": 16878 }, { "epoch": 1.32, "learning_rate": 8.67646828197287e-05, "loss": 2.9389, "step": 16879 }, { "epoch": 1.32, "learning_rate": 8.67638986905042e-05, "loss": 2.6275, "step": 16880 }, { "epoch": 1.32, "learning_rate": 8.67631145612797e-05, "loss": 3.2206, "step": 16881 }, { "epoch": 1.32, "learning_rate": 8.67623304320552e-05, "loss": 2.954, "step": 16882 }, { "epoch": 1.32, "learning_rate": 8.676154630283071e-05, "loss": 2.7896, "step": 16883 }, { "epoch": 1.32, "learning_rate": 8.676076217360621e-05, "loss": 2.7896, "step": 16884 }, { "epoch": 1.32, "learning_rate": 8.675997804438172e-05, "loss": 3.0487, "step": 16885 }, { "epoch": 1.32, "learning_rate": 8.675919391515722e-05, "loss": 2.8806, "step": 16886 }, { "epoch": 1.32, "learning_rate": 8.675840978593272e-05, "loss": 2.963, "step": 16887 }, { "epoch": 1.32, "learning_rate": 8.675762565670823e-05, "loss": 2.8194, "step": 16888 }, { "epoch": 1.32, "learning_rate": 8.675684152748373e-05, "loss": 2.8782, "step": 16889 }, { "epoch": 1.32, "learning_rate": 8.675605739825924e-05, "loss": 3.1348, "step": 16890 }, { "epoch": 1.32, "learning_rate": 8.675527326903474e-05, "loss": 2.8771, "step": 16891 }, { "epoch": 1.32, "learning_rate": 8.675448913981024e-05, "loss": 2.9484, "step": 16892 }, { "epoch": 1.32, "learning_rate": 8.675370501058575e-05, "loss": 3.1441, "step": 16893 }, { "epoch": 1.32, "learning_rate": 8.675292088136125e-05, "loss": 2.918, "step": 16894 }, { "epoch": 1.32, "learning_rate": 8.675213675213675e-05, "loss": 2.8253, "step": 16895 }, { "epoch": 1.32, "learning_rate": 8.675135262291226e-05, "loss": 2.8841, "step": 16896 }, { "epoch": 1.32, "learning_rate": 8.675056849368776e-05, "loss": 3.2161, "step": 16897 }, { "epoch": 1.33, "learning_rate": 8.674978436446326e-05, "loss": 2.8817, "step": 16898 }, { "epoch": 1.33, "learning_rate": 8.674900023523878e-05, "loss": 2.9892, "step": 16899 }, { "epoch": 1.33, "learning_rate": 8.674821610601429e-05, "loss": 2.9816, "step": 16900 }, { "epoch": 1.33, "learning_rate": 8.674743197678979e-05, "loss": 2.7519, "step": 16901 }, { "epoch": 1.33, "learning_rate": 8.674664784756529e-05, "loss": 2.8599, "step": 16902 }, { "epoch": 1.33, "learning_rate": 8.67458637183408e-05, "loss": 3.0538, "step": 16903 }, { "epoch": 1.33, "learning_rate": 8.67450795891163e-05, "loss": 2.6887, "step": 16904 }, { "epoch": 1.33, "learning_rate": 8.674429545989179e-05, "loss": 2.7438, "step": 16905 }, { "epoch": 1.33, "learning_rate": 8.67435113306673e-05, "loss": 2.909, "step": 16906 }, { "epoch": 1.33, "learning_rate": 8.67427272014428e-05, "loss": 3.0512, "step": 16907 }, { "epoch": 1.33, "learning_rate": 8.67419430722183e-05, "loss": 2.7187, "step": 16908 }, { "epoch": 1.33, "learning_rate": 8.67411589429938e-05, "loss": 2.9086, "step": 16909 }, { "epoch": 1.33, "learning_rate": 8.674037481376931e-05, "loss": 2.9181, "step": 16910 }, { "epoch": 1.33, "learning_rate": 8.673959068454481e-05, "loss": 3.0168, "step": 16911 }, { "epoch": 1.33, "learning_rate": 8.673880655532032e-05, "loss": 2.8298, "step": 16912 }, { "epoch": 1.33, "learning_rate": 8.673802242609582e-05, "loss": 2.9508, "step": 16913 }, { "epoch": 1.33, "learning_rate": 8.673723829687132e-05, "loss": 2.6925, "step": 16914 }, { "epoch": 1.33, "learning_rate": 8.673645416764683e-05, "loss": 2.8423, "step": 16915 }, { "epoch": 1.33, "learning_rate": 8.673567003842233e-05, "loss": 3.301, "step": 16916 }, { "epoch": 1.33, "learning_rate": 8.673488590919783e-05, "loss": 2.8541, "step": 16917 }, { "epoch": 1.33, "learning_rate": 8.673410177997334e-05, "loss": 2.8505, "step": 16918 }, { "epoch": 1.33, "learning_rate": 8.673331765074884e-05, "loss": 2.7543, "step": 16919 }, { "epoch": 1.33, "learning_rate": 8.673253352152435e-05, "loss": 3.1453, "step": 16920 }, { "epoch": 1.33, "learning_rate": 8.673174939229985e-05, "loss": 2.6832, "step": 16921 }, { "epoch": 1.33, "learning_rate": 8.673096526307535e-05, "loss": 3.1111, "step": 16922 }, { "epoch": 1.33, "learning_rate": 8.673018113385086e-05, "loss": 3.0264, "step": 16923 }, { "epoch": 1.33, "learning_rate": 8.672939700462636e-05, "loss": 2.9481, "step": 16924 }, { "epoch": 1.33, "learning_rate": 8.672861287540188e-05, "loss": 2.9503, "step": 16925 }, { "epoch": 1.33, "learning_rate": 8.672782874617738e-05, "loss": 3.0048, "step": 16926 }, { "epoch": 1.33, "learning_rate": 8.672704461695288e-05, "loss": 2.8851, "step": 16927 }, { "epoch": 1.33, "learning_rate": 8.672626048772839e-05, "loss": 2.9101, "step": 16928 }, { "epoch": 1.33, "learning_rate": 8.672547635850389e-05, "loss": 2.9109, "step": 16929 }, { "epoch": 1.33, "learning_rate": 8.67246922292794e-05, "loss": 2.7977, "step": 16930 }, { "epoch": 1.33, "learning_rate": 8.67239081000549e-05, "loss": 2.887, "step": 16931 }, { "epoch": 1.33, "learning_rate": 8.67231239708304e-05, "loss": 2.8729, "step": 16932 }, { "epoch": 1.33, "learning_rate": 8.672233984160591e-05, "loss": 2.904, "step": 16933 }, { "epoch": 1.33, "learning_rate": 8.672155571238141e-05, "loss": 2.8561, "step": 16934 }, { "epoch": 1.33, "learning_rate": 8.672077158315691e-05, "loss": 2.6869, "step": 16935 }, { "epoch": 1.33, "learning_rate": 8.671998745393242e-05, "loss": 2.7013, "step": 16936 }, { "epoch": 1.33, "learning_rate": 8.671920332470792e-05, "loss": 3.0494, "step": 16937 }, { "epoch": 1.33, "learning_rate": 8.671841919548343e-05, "loss": 2.799, "step": 16938 }, { "epoch": 1.33, "learning_rate": 8.671763506625893e-05, "loss": 2.8765, "step": 16939 }, { "epoch": 1.33, "learning_rate": 8.671685093703443e-05, "loss": 2.9423, "step": 16940 }, { "epoch": 1.33, "learning_rate": 8.671606680780992e-05, "loss": 2.8676, "step": 16941 }, { "epoch": 1.33, "learning_rate": 8.671528267858543e-05, "loss": 2.9317, "step": 16942 }, { "epoch": 1.33, "learning_rate": 8.671449854936093e-05, "loss": 2.8442, "step": 16943 }, { "epoch": 1.33, "learning_rate": 8.671371442013643e-05, "loss": 2.7641, "step": 16944 }, { "epoch": 1.33, "learning_rate": 8.671293029091194e-05, "loss": 2.8095, "step": 16945 }, { "epoch": 1.33, "learning_rate": 8.671214616168744e-05, "loss": 2.9608, "step": 16946 }, { "epoch": 1.33, "learning_rate": 8.671136203246294e-05, "loss": 3.2325, "step": 16947 }, { "epoch": 1.33, "learning_rate": 8.671057790323845e-05, "loss": 2.8433, "step": 16948 }, { "epoch": 1.33, "learning_rate": 8.670979377401395e-05, "loss": 3.1275, "step": 16949 }, { "epoch": 1.33, "learning_rate": 8.670900964478947e-05, "loss": 2.8501, "step": 16950 }, { "epoch": 1.33, "learning_rate": 8.670822551556497e-05, "loss": 3.1909, "step": 16951 }, { "epoch": 1.33, "learning_rate": 8.670744138634048e-05, "loss": 3.0482, "step": 16952 }, { "epoch": 1.33, "learning_rate": 8.670665725711598e-05, "loss": 2.8205, "step": 16953 }, { "epoch": 1.33, "learning_rate": 8.670587312789148e-05, "loss": 2.8029, "step": 16954 }, { "epoch": 1.33, "learning_rate": 8.670508899866699e-05, "loss": 2.9826, "step": 16955 }, { "epoch": 1.33, "learning_rate": 8.670430486944249e-05, "loss": 2.7719, "step": 16956 }, { "epoch": 1.33, "learning_rate": 8.6703520740218e-05, "loss": 3.0885, "step": 16957 }, { "epoch": 1.33, "learning_rate": 8.67027366109935e-05, "loss": 2.7898, "step": 16958 }, { "epoch": 1.33, "learning_rate": 8.6701952481769e-05, "loss": 2.8783, "step": 16959 }, { "epoch": 1.33, "learning_rate": 8.67011683525445e-05, "loss": 2.8748, "step": 16960 }, { "epoch": 1.33, "learning_rate": 8.670038422332001e-05, "loss": 2.7828, "step": 16961 }, { "epoch": 1.33, "learning_rate": 8.669960009409551e-05, "loss": 2.6802, "step": 16962 }, { "epoch": 1.33, "learning_rate": 8.669881596487102e-05, "loss": 2.839, "step": 16963 }, { "epoch": 1.33, "learning_rate": 8.669803183564652e-05, "loss": 2.8659, "step": 16964 }, { "epoch": 1.33, "learning_rate": 8.669724770642202e-05, "loss": 3.069, "step": 16965 }, { "epoch": 1.33, "learning_rate": 8.669646357719753e-05, "loss": 2.7837, "step": 16966 }, { "epoch": 1.33, "learning_rate": 8.669567944797303e-05, "loss": 2.8679, "step": 16967 }, { "epoch": 1.33, "learning_rate": 8.669489531874854e-05, "loss": 2.8758, "step": 16968 }, { "epoch": 1.33, "learning_rate": 8.669411118952404e-05, "loss": 2.894, "step": 16969 }, { "epoch": 1.33, "learning_rate": 8.669332706029954e-05, "loss": 2.9721, "step": 16970 }, { "epoch": 1.33, "learning_rate": 8.669254293107505e-05, "loss": 3.0872, "step": 16971 }, { "epoch": 1.33, "learning_rate": 8.669175880185055e-05, "loss": 3.0008, "step": 16972 }, { "epoch": 1.33, "learning_rate": 8.669097467262605e-05, "loss": 2.8444, "step": 16973 }, { "epoch": 1.33, "learning_rate": 8.669019054340156e-05, "loss": 3.3523, "step": 16974 }, { "epoch": 1.33, "learning_rate": 8.668940641417706e-05, "loss": 2.8059, "step": 16975 }, { "epoch": 1.33, "learning_rate": 8.668862228495257e-05, "loss": 2.9839, "step": 16976 }, { "epoch": 1.33, "learning_rate": 8.668783815572807e-05, "loss": 3.0857, "step": 16977 }, { "epoch": 1.33, "learning_rate": 8.668705402650357e-05, "loss": 3.1531, "step": 16978 }, { "epoch": 1.33, "learning_rate": 8.668626989727908e-05, "loss": 2.8376, "step": 16979 }, { "epoch": 1.33, "learning_rate": 8.668548576805458e-05, "loss": 3.0112, "step": 16980 }, { "epoch": 1.33, "learning_rate": 8.668470163883008e-05, "loss": 2.8354, "step": 16981 }, { "epoch": 1.33, "learning_rate": 8.668391750960559e-05, "loss": 3.043, "step": 16982 }, { "epoch": 1.33, "learning_rate": 8.668313338038109e-05, "loss": 3.1192, "step": 16983 }, { "epoch": 1.33, "learning_rate": 8.66823492511566e-05, "loss": 2.9264, "step": 16984 }, { "epoch": 1.33, "learning_rate": 8.66815651219321e-05, "loss": 2.8452, "step": 16985 }, { "epoch": 1.33, "learning_rate": 8.66807809927076e-05, "loss": 2.7615, "step": 16986 }, { "epoch": 1.33, "learning_rate": 8.66799968634831e-05, "loss": 3.2784, "step": 16987 }, { "epoch": 1.33, "learning_rate": 8.667921273425861e-05, "loss": 2.7735, "step": 16988 }, { "epoch": 1.33, "learning_rate": 8.667842860503411e-05, "loss": 2.9935, "step": 16989 }, { "epoch": 1.33, "learning_rate": 8.667764447580962e-05, "loss": 3.0808, "step": 16990 }, { "epoch": 1.33, "learning_rate": 8.667686034658512e-05, "loss": 3.1377, "step": 16991 }, { "epoch": 1.33, "learning_rate": 8.667607621736062e-05, "loss": 2.9102, "step": 16992 }, { "epoch": 1.33, "learning_rate": 8.667529208813613e-05, "loss": 2.7491, "step": 16993 }, { "epoch": 1.33, "learning_rate": 8.667450795891163e-05, "loss": 2.7699, "step": 16994 }, { "epoch": 1.33, "learning_rate": 8.667372382968714e-05, "loss": 2.9011, "step": 16995 }, { "epoch": 1.33, "learning_rate": 8.667293970046264e-05, "loss": 2.7144, "step": 16996 }, { "epoch": 1.33, "learning_rate": 8.667215557123814e-05, "loss": 2.8185, "step": 16997 }, { "epoch": 1.33, "learning_rate": 8.667137144201365e-05, "loss": 2.7362, "step": 16998 }, { "epoch": 1.33, "learning_rate": 8.667058731278915e-05, "loss": 2.7222, "step": 16999 }, { "epoch": 1.33, "learning_rate": 8.666980318356465e-05, "loss": 2.9213, "step": 17000 }, { "epoch": 1.33, "eval_loss": 2.91564679145813, "eval_runtime": 217.9837, "eval_samples_per_second": 51.986, "eval_steps_per_second": 3.253, "step": 17000 }, { "epoch": 1.33, "learning_rate": 8.666901905434016e-05, "loss": 2.7037, "step": 17001 }, { "epoch": 1.33, "learning_rate": 8.666823492511566e-05, "loss": 3.1592, "step": 17002 }, { "epoch": 1.33, "learning_rate": 8.666745079589116e-05, "loss": 2.7348, "step": 17003 }, { "epoch": 1.33, "learning_rate": 8.666666666666667e-05, "loss": 2.677, "step": 17004 }, { "epoch": 1.33, "learning_rate": 8.666588253744217e-05, "loss": 2.9375, "step": 17005 }, { "epoch": 1.33, "learning_rate": 8.666509840821768e-05, "loss": 2.8269, "step": 17006 }, { "epoch": 1.33, "learning_rate": 8.666431427899319e-05, "loss": 2.6369, "step": 17007 }, { "epoch": 1.33, "learning_rate": 8.66635301497687e-05, "loss": 2.914, "step": 17008 }, { "epoch": 1.33, "learning_rate": 8.66627460205442e-05, "loss": 2.9114, "step": 17009 }, { "epoch": 1.33, "learning_rate": 8.66619618913197e-05, "loss": 2.9714, "step": 17010 }, { "epoch": 1.33, "learning_rate": 8.666117776209521e-05, "loss": 2.9303, "step": 17011 }, { "epoch": 1.33, "learning_rate": 8.666039363287071e-05, "loss": 2.7853, "step": 17012 }, { "epoch": 1.33, "learning_rate": 8.66596095036462e-05, "loss": 2.7608, "step": 17013 }, { "epoch": 1.33, "learning_rate": 8.66588253744217e-05, "loss": 2.8798, "step": 17014 }, { "epoch": 1.33, "learning_rate": 8.665804124519721e-05, "loss": 3.2273, "step": 17015 }, { "epoch": 1.33, "learning_rate": 8.665725711597271e-05, "loss": 2.8629, "step": 17016 }, { "epoch": 1.33, "learning_rate": 8.665647298674822e-05, "loss": 2.8689, "step": 17017 }, { "epoch": 1.33, "learning_rate": 8.665568885752372e-05, "loss": 2.9186, "step": 17018 }, { "epoch": 1.33, "learning_rate": 8.665490472829922e-05, "loss": 3.1356, "step": 17019 }, { "epoch": 1.33, "learning_rate": 8.665412059907473e-05, "loss": 2.9256, "step": 17020 }, { "epoch": 1.33, "learning_rate": 8.665333646985023e-05, "loss": 2.7794, "step": 17021 }, { "epoch": 1.33, "learning_rate": 8.665255234062573e-05, "loss": 3.1692, "step": 17022 }, { "epoch": 1.33, "learning_rate": 8.665176821140124e-05, "loss": 2.941, "step": 17023 }, { "epoch": 1.33, "learning_rate": 8.665098408217674e-05, "loss": 2.8975, "step": 17024 }, { "epoch": 1.33, "learning_rate": 8.665019995295225e-05, "loss": 3.1116, "step": 17025 }, { "epoch": 1.34, "learning_rate": 8.664941582372775e-05, "loss": 2.9947, "step": 17026 }, { "epoch": 1.34, "learning_rate": 8.664863169450325e-05, "loss": 2.9218, "step": 17027 }, { "epoch": 1.34, "learning_rate": 8.664784756527876e-05, "loss": 2.8989, "step": 17028 }, { "epoch": 1.34, "learning_rate": 8.664706343605426e-05, "loss": 2.8097, "step": 17029 }, { "epoch": 1.34, "learning_rate": 8.664627930682976e-05, "loss": 2.8719, "step": 17030 }, { "epoch": 1.34, "learning_rate": 8.664549517760527e-05, "loss": 3.0148, "step": 17031 }, { "epoch": 1.34, "learning_rate": 8.664471104838077e-05, "loss": 2.6644, "step": 17032 }, { "epoch": 1.34, "learning_rate": 8.664392691915629e-05, "loss": 2.871, "step": 17033 }, { "epoch": 1.34, "learning_rate": 8.664314278993179e-05, "loss": 2.9309, "step": 17034 }, { "epoch": 1.34, "learning_rate": 8.66423586607073e-05, "loss": 3.1266, "step": 17035 }, { "epoch": 1.34, "learning_rate": 8.66415745314828e-05, "loss": 3.3012, "step": 17036 }, { "epoch": 1.34, "learning_rate": 8.66407904022583e-05, "loss": 2.8522, "step": 17037 }, { "epoch": 1.34, "learning_rate": 8.66400062730338e-05, "loss": 2.7398, "step": 17038 }, { "epoch": 1.34, "learning_rate": 8.663922214380931e-05, "loss": 3.2018, "step": 17039 }, { "epoch": 1.34, "learning_rate": 8.663843801458481e-05, "loss": 2.7336, "step": 17040 }, { "epoch": 1.34, "learning_rate": 8.663765388536032e-05, "loss": 3.4783, "step": 17041 }, { "epoch": 1.34, "learning_rate": 8.663686975613582e-05, "loss": 2.6798, "step": 17042 }, { "epoch": 1.34, "learning_rate": 8.663608562691133e-05, "loss": 2.7229, "step": 17043 }, { "epoch": 1.34, "learning_rate": 8.663530149768683e-05, "loss": 3.0238, "step": 17044 }, { "epoch": 1.34, "learning_rate": 8.663451736846233e-05, "loss": 2.9313, "step": 17045 }, { "epoch": 1.34, "learning_rate": 8.663373323923784e-05, "loss": 2.9574, "step": 17046 }, { "epoch": 1.34, "learning_rate": 8.663294911001334e-05, "loss": 3.0258, "step": 17047 }, { "epoch": 1.34, "learning_rate": 8.663216498078884e-05, "loss": 2.9822, "step": 17048 }, { "epoch": 1.34, "learning_rate": 8.663138085156433e-05, "loss": 2.9092, "step": 17049 }, { "epoch": 1.34, "learning_rate": 8.663059672233984e-05, "loss": 2.8362, "step": 17050 }, { "epoch": 1.34, "learning_rate": 8.662981259311534e-05, "loss": 2.9577, "step": 17051 }, { "epoch": 1.34, "learning_rate": 8.662902846389084e-05, "loss": 2.9444, "step": 17052 }, { "epoch": 1.34, "learning_rate": 8.662824433466635e-05, "loss": 2.7579, "step": 17053 }, { "epoch": 1.34, "learning_rate": 8.662746020544185e-05, "loss": 2.91, "step": 17054 }, { "epoch": 1.34, "learning_rate": 8.662667607621736e-05, "loss": 3.126, "step": 17055 }, { "epoch": 1.34, "learning_rate": 8.662589194699286e-05, "loss": 2.8087, "step": 17056 }, { "epoch": 1.34, "learning_rate": 8.662510781776836e-05, "loss": 2.8146, "step": 17057 }, { "epoch": 1.34, "learning_rate": 8.662432368854388e-05, "loss": 3.1862, "step": 17058 }, { "epoch": 1.34, "learning_rate": 8.662353955931938e-05, "loss": 2.6853, "step": 17059 }, { "epoch": 1.34, "learning_rate": 8.662275543009489e-05, "loss": 2.8703, "step": 17060 }, { "epoch": 1.34, "learning_rate": 8.662197130087039e-05, "loss": 3.0009, "step": 17061 }, { "epoch": 1.34, "learning_rate": 8.66211871716459e-05, "loss": 2.7383, "step": 17062 }, { "epoch": 1.34, "learning_rate": 8.66204030424214e-05, "loss": 2.7847, "step": 17063 }, { "epoch": 1.34, "learning_rate": 8.66196189131969e-05, "loss": 3.0069, "step": 17064 }, { "epoch": 1.34, "learning_rate": 8.66188347839724e-05, "loss": 2.7903, "step": 17065 }, { "epoch": 1.34, "learning_rate": 8.661805065474791e-05, "loss": 2.8113, "step": 17066 }, { "epoch": 1.34, "learning_rate": 8.661726652552341e-05, "loss": 3.2901, "step": 17067 }, { "epoch": 1.34, "learning_rate": 8.661648239629892e-05, "loss": 2.834, "step": 17068 }, { "epoch": 1.34, "learning_rate": 8.661569826707442e-05, "loss": 2.7255, "step": 17069 }, { "epoch": 1.34, "learning_rate": 8.661491413784992e-05, "loss": 2.9338, "step": 17070 }, { "epoch": 1.34, "learning_rate": 8.661413000862543e-05, "loss": 3.0084, "step": 17071 }, { "epoch": 1.34, "learning_rate": 8.661334587940093e-05, "loss": 2.7135, "step": 17072 }, { "epoch": 1.34, "learning_rate": 8.661256175017644e-05, "loss": 2.7557, "step": 17073 }, { "epoch": 1.34, "learning_rate": 8.661177762095194e-05, "loss": 2.9076, "step": 17074 }, { "epoch": 1.34, "learning_rate": 8.661099349172744e-05, "loss": 3.1523, "step": 17075 }, { "epoch": 1.34, "learning_rate": 8.661020936250295e-05, "loss": 3.0323, "step": 17076 }, { "epoch": 1.34, "learning_rate": 8.660942523327845e-05, "loss": 2.8738, "step": 17077 }, { "epoch": 1.34, "learning_rate": 8.660864110405395e-05, "loss": 3.2378, "step": 17078 }, { "epoch": 1.34, "learning_rate": 8.660785697482946e-05, "loss": 3.414, "step": 17079 }, { "epoch": 1.34, "learning_rate": 8.660707284560496e-05, "loss": 3.0283, "step": 17080 }, { "epoch": 1.34, "learning_rate": 8.660628871638046e-05, "loss": 3.0674, "step": 17081 }, { "epoch": 1.34, "learning_rate": 8.660550458715597e-05, "loss": 2.9167, "step": 17082 }, { "epoch": 1.34, "learning_rate": 8.660472045793147e-05, "loss": 2.827, "step": 17083 }, { "epoch": 1.34, "learning_rate": 8.660393632870698e-05, "loss": 2.83, "step": 17084 }, { "epoch": 1.34, "learning_rate": 8.660315219948248e-05, "loss": 2.936, "step": 17085 }, { "epoch": 1.34, "learning_rate": 8.660236807025798e-05, "loss": 2.9329, "step": 17086 }, { "epoch": 1.34, "learning_rate": 8.660158394103349e-05, "loss": 2.9041, "step": 17087 }, { "epoch": 1.34, "learning_rate": 8.660079981180899e-05, "loss": 2.8726, "step": 17088 }, { "epoch": 1.34, "learning_rate": 8.66000156825845e-05, "loss": 2.7265, "step": 17089 }, { "epoch": 1.34, "learning_rate": 8.659923155336e-05, "loss": 2.932, "step": 17090 }, { "epoch": 1.34, "learning_rate": 8.65984474241355e-05, "loss": 3.1154, "step": 17091 }, { "epoch": 1.34, "learning_rate": 8.6597663294911e-05, "loss": 2.6751, "step": 17092 }, { "epoch": 1.34, "learning_rate": 8.659687916568651e-05, "loss": 2.8732, "step": 17093 }, { "epoch": 1.34, "learning_rate": 8.659609503646201e-05, "loss": 2.9564, "step": 17094 }, { "epoch": 1.34, "learning_rate": 8.659531090723752e-05, "loss": 2.9234, "step": 17095 }, { "epoch": 1.34, "learning_rate": 8.659452677801302e-05, "loss": 3.0676, "step": 17096 }, { "epoch": 1.34, "learning_rate": 8.659374264878852e-05, "loss": 2.8886, "step": 17097 }, { "epoch": 1.34, "learning_rate": 8.659295851956403e-05, "loss": 2.8198, "step": 17098 }, { "epoch": 1.34, "learning_rate": 8.659217439033953e-05, "loss": 2.8668, "step": 17099 }, { "epoch": 1.34, "learning_rate": 8.659139026111503e-05, "loss": 2.9829, "step": 17100 }, { "epoch": 1.34, "learning_rate": 8.659060613189054e-05, "loss": 3.1031, "step": 17101 }, { "epoch": 1.34, "learning_rate": 8.658982200266604e-05, "loss": 2.9817, "step": 17102 }, { "epoch": 1.34, "learning_rate": 8.658903787344155e-05, "loss": 2.7854, "step": 17103 }, { "epoch": 1.34, "learning_rate": 8.658825374421705e-05, "loss": 2.9866, "step": 17104 }, { "epoch": 1.34, "learning_rate": 8.658746961499255e-05, "loss": 3.0744, "step": 17105 }, { "epoch": 1.34, "learning_rate": 8.658668548576806e-05, "loss": 2.9563, "step": 17106 }, { "epoch": 1.34, "learning_rate": 8.658590135654356e-05, "loss": 2.8709, "step": 17107 }, { "epoch": 1.34, "learning_rate": 8.658511722731906e-05, "loss": 2.7491, "step": 17108 }, { "epoch": 1.34, "learning_rate": 8.658433309809457e-05, "loss": 2.9073, "step": 17109 }, { "epoch": 1.34, "learning_rate": 8.658354896887007e-05, "loss": 2.7757, "step": 17110 }, { "epoch": 1.34, "learning_rate": 8.658276483964558e-05, "loss": 2.9539, "step": 17111 }, { "epoch": 1.34, "learning_rate": 8.658198071042108e-05, "loss": 2.9988, "step": 17112 }, { "epoch": 1.34, "learning_rate": 8.658119658119658e-05, "loss": 2.8335, "step": 17113 }, { "epoch": 1.34, "learning_rate": 8.658041245197209e-05, "loss": 2.7817, "step": 17114 }, { "epoch": 1.34, "learning_rate": 8.65796283227476e-05, "loss": 3.0603, "step": 17115 }, { "epoch": 1.34, "learning_rate": 8.657884419352311e-05, "loss": 2.8613, "step": 17116 }, { "epoch": 1.34, "learning_rate": 8.657806006429861e-05, "loss": 3.017, "step": 17117 }, { "epoch": 1.34, "learning_rate": 8.657727593507411e-05, "loss": 2.7905, "step": 17118 }, { "epoch": 1.34, "learning_rate": 8.657649180584962e-05, "loss": 2.962, "step": 17119 }, { "epoch": 1.34, "learning_rate": 8.657570767662512e-05, "loss": 2.9993, "step": 17120 }, { "epoch": 1.34, "learning_rate": 8.657492354740061e-05, "loss": 3.0484, "step": 17121 }, { "epoch": 1.34, "learning_rate": 8.657413941817612e-05, "loss": 2.9867, "step": 17122 }, { "epoch": 1.34, "learning_rate": 8.657335528895162e-05, "loss": 3.0052, "step": 17123 }, { "epoch": 1.34, "learning_rate": 8.657257115972712e-05, "loss": 3.1528, "step": 17124 }, { "epoch": 1.34, "learning_rate": 8.657178703050263e-05, "loss": 2.9051, "step": 17125 }, { "epoch": 1.34, "learning_rate": 8.657100290127813e-05, "loss": 3.0439, "step": 17126 }, { "epoch": 1.34, "learning_rate": 8.657021877205363e-05, "loss": 2.8214, "step": 17127 }, { "epoch": 1.34, "learning_rate": 8.656943464282914e-05, "loss": 2.959, "step": 17128 }, { "epoch": 1.34, "learning_rate": 8.656865051360464e-05, "loss": 2.8245, "step": 17129 }, { "epoch": 1.34, "learning_rate": 8.656786638438015e-05, "loss": 2.7698, "step": 17130 }, { "epoch": 1.34, "learning_rate": 8.656708225515565e-05, "loss": 2.8841, "step": 17131 }, { "epoch": 1.34, "learning_rate": 8.656629812593115e-05, "loss": 2.9904, "step": 17132 }, { "epoch": 1.34, "learning_rate": 8.656551399670666e-05, "loss": 2.8255, "step": 17133 }, { "epoch": 1.34, "learning_rate": 8.656472986748216e-05, "loss": 2.8614, "step": 17134 }, { "epoch": 1.34, "learning_rate": 8.656394573825766e-05, "loss": 3.0247, "step": 17135 }, { "epoch": 1.34, "learning_rate": 8.656316160903317e-05, "loss": 2.9136, "step": 17136 }, { "epoch": 1.34, "learning_rate": 8.656237747980867e-05, "loss": 3.0428, "step": 17137 }, { "epoch": 1.34, "learning_rate": 8.656159335058417e-05, "loss": 3.0323, "step": 17138 }, { "epoch": 1.34, "learning_rate": 8.656080922135968e-05, "loss": 3.1358, "step": 17139 }, { "epoch": 1.34, "learning_rate": 8.656002509213518e-05, "loss": 3.1086, "step": 17140 }, { "epoch": 1.34, "learning_rate": 8.65592409629107e-05, "loss": 2.7109, "step": 17141 }, { "epoch": 1.34, "learning_rate": 8.65584568336862e-05, "loss": 2.8101, "step": 17142 }, { "epoch": 1.34, "learning_rate": 8.65576727044617e-05, "loss": 2.8643, "step": 17143 }, { "epoch": 1.34, "learning_rate": 8.655688857523721e-05, "loss": 2.9787, "step": 17144 }, { "epoch": 1.34, "learning_rate": 8.655610444601271e-05, "loss": 2.9196, "step": 17145 }, { "epoch": 1.34, "learning_rate": 8.655532031678822e-05, "loss": 3.0081, "step": 17146 }, { "epoch": 1.34, "learning_rate": 8.655453618756372e-05, "loss": 3.0593, "step": 17147 }, { "epoch": 1.34, "learning_rate": 8.655375205833922e-05, "loss": 3.0189, "step": 17148 }, { "epoch": 1.34, "learning_rate": 8.655296792911473e-05, "loss": 2.5987, "step": 17149 }, { "epoch": 1.34, "learning_rate": 8.655218379989023e-05, "loss": 2.8531, "step": 17150 }, { "epoch": 1.34, "learning_rate": 8.655139967066574e-05, "loss": 3.1827, "step": 17151 }, { "epoch": 1.34, "learning_rate": 8.655061554144124e-05, "loss": 2.8224, "step": 17152 }, { "epoch": 1.35, "learning_rate": 8.654983141221674e-05, "loss": 2.9159, "step": 17153 }, { "epoch": 1.35, "learning_rate": 8.654904728299225e-05, "loss": 2.8585, "step": 17154 }, { "epoch": 1.35, "learning_rate": 8.654826315376775e-05, "loss": 2.9699, "step": 17155 }, { "epoch": 1.35, "learning_rate": 8.654747902454325e-05, "loss": 2.8755, "step": 17156 }, { "epoch": 1.35, "learning_rate": 8.654669489531874e-05, "loss": 2.8095, "step": 17157 }, { "epoch": 1.35, "learning_rate": 8.654591076609425e-05, "loss": 2.8093, "step": 17158 }, { "epoch": 1.35, "learning_rate": 8.654512663686975e-05, "loss": 2.9002, "step": 17159 }, { "epoch": 1.35, "learning_rate": 8.654434250764526e-05, "loss": 2.775, "step": 17160 }, { "epoch": 1.35, "learning_rate": 8.654355837842076e-05, "loss": 2.9738, "step": 17161 }, { "epoch": 1.35, "learning_rate": 8.654277424919626e-05, "loss": 2.9377, "step": 17162 }, { "epoch": 1.35, "learning_rate": 8.654199011997177e-05, "loss": 2.8121, "step": 17163 }, { "epoch": 1.35, "learning_rate": 8.654120599074727e-05, "loss": 3.1318, "step": 17164 }, { "epoch": 1.35, "learning_rate": 8.654042186152277e-05, "loss": 2.9656, "step": 17165 }, { "epoch": 1.35, "learning_rate": 8.653963773229829e-05, "loss": 2.7794, "step": 17166 }, { "epoch": 1.35, "learning_rate": 8.65388536030738e-05, "loss": 2.7197, "step": 17167 }, { "epoch": 1.35, "learning_rate": 8.65380694738493e-05, "loss": 3.1711, "step": 17168 }, { "epoch": 1.35, "learning_rate": 8.65372853446248e-05, "loss": 3.0535, "step": 17169 }, { "epoch": 1.35, "learning_rate": 8.65365012154003e-05, "loss": 2.7697, "step": 17170 }, { "epoch": 1.35, "learning_rate": 8.653571708617581e-05, "loss": 3.0086, "step": 17171 }, { "epoch": 1.35, "learning_rate": 8.653493295695131e-05, "loss": 2.9326, "step": 17172 }, { "epoch": 1.35, "learning_rate": 8.653414882772682e-05, "loss": 3.0278, "step": 17173 }, { "epoch": 1.35, "learning_rate": 8.653336469850232e-05, "loss": 2.8328, "step": 17174 }, { "epoch": 1.35, "learning_rate": 8.653258056927782e-05, "loss": 2.9642, "step": 17175 }, { "epoch": 1.35, "learning_rate": 8.653179644005333e-05, "loss": 3.0401, "step": 17176 }, { "epoch": 1.35, "learning_rate": 8.653101231082883e-05, "loss": 2.7318, "step": 17177 }, { "epoch": 1.35, "learning_rate": 8.653022818160434e-05, "loss": 2.8563, "step": 17178 }, { "epoch": 1.35, "learning_rate": 8.652944405237984e-05, "loss": 3.1491, "step": 17179 }, { "epoch": 1.35, "learning_rate": 8.652865992315534e-05, "loss": 2.9126, "step": 17180 }, { "epoch": 1.35, "learning_rate": 8.652787579393085e-05, "loss": 2.7657, "step": 17181 }, { "epoch": 1.35, "learning_rate": 8.652709166470635e-05, "loss": 2.956, "step": 17182 }, { "epoch": 1.35, "learning_rate": 8.652630753548185e-05, "loss": 2.8831, "step": 17183 }, { "epoch": 1.35, "learning_rate": 8.652552340625736e-05, "loss": 2.9551, "step": 17184 }, { "epoch": 1.35, "learning_rate": 8.652473927703286e-05, "loss": 2.8845, "step": 17185 }, { "epoch": 1.35, "learning_rate": 8.652395514780836e-05, "loss": 2.8415, "step": 17186 }, { "epoch": 1.35, "learning_rate": 8.652317101858387e-05, "loss": 2.9621, "step": 17187 }, { "epoch": 1.35, "learning_rate": 8.652238688935937e-05, "loss": 2.6795, "step": 17188 }, { "epoch": 1.35, "learning_rate": 8.652160276013488e-05, "loss": 2.8386, "step": 17189 }, { "epoch": 1.35, "learning_rate": 8.652081863091038e-05, "loss": 2.9138, "step": 17190 }, { "epoch": 1.35, "learning_rate": 8.652003450168588e-05, "loss": 2.899, "step": 17191 }, { "epoch": 1.35, "learning_rate": 8.651925037246139e-05, "loss": 3.0982, "step": 17192 }, { "epoch": 1.35, "learning_rate": 8.651846624323689e-05, "loss": 2.7679, "step": 17193 }, { "epoch": 1.35, "learning_rate": 8.65176821140124e-05, "loss": 2.5994, "step": 17194 }, { "epoch": 1.35, "learning_rate": 8.65168979847879e-05, "loss": 2.8427, "step": 17195 }, { "epoch": 1.35, "learning_rate": 8.65161138555634e-05, "loss": 2.7645, "step": 17196 }, { "epoch": 1.35, "learning_rate": 8.65153297263389e-05, "loss": 2.8947, "step": 17197 }, { "epoch": 1.35, "learning_rate": 8.651454559711441e-05, "loss": 2.834, "step": 17198 }, { "epoch": 1.35, "learning_rate": 8.651376146788991e-05, "loss": 2.8885, "step": 17199 }, { "epoch": 1.35, "learning_rate": 8.651297733866542e-05, "loss": 3.0658, "step": 17200 }, { "epoch": 1.35, "learning_rate": 8.651219320944092e-05, "loss": 2.8861, "step": 17201 }, { "epoch": 1.35, "learning_rate": 8.651140908021642e-05, "loss": 2.8374, "step": 17202 }, { "epoch": 1.35, "learning_rate": 8.651062495099193e-05, "loss": 2.7001, "step": 17203 }, { "epoch": 1.35, "learning_rate": 8.650984082176743e-05, "loss": 2.9034, "step": 17204 }, { "epoch": 1.35, "learning_rate": 8.650905669254293e-05, "loss": 3.0333, "step": 17205 }, { "epoch": 1.35, "learning_rate": 8.650827256331844e-05, "loss": 3.0041, "step": 17206 }, { "epoch": 1.35, "learning_rate": 8.650748843409394e-05, "loss": 2.9769, "step": 17207 }, { "epoch": 1.35, "learning_rate": 8.650670430486945e-05, "loss": 2.8486, "step": 17208 }, { "epoch": 1.35, "learning_rate": 8.650592017564495e-05, "loss": 2.8984, "step": 17209 }, { "epoch": 1.35, "learning_rate": 8.650513604642045e-05, "loss": 2.856, "step": 17210 }, { "epoch": 1.35, "learning_rate": 8.650435191719596e-05, "loss": 2.6323, "step": 17211 }, { "epoch": 1.35, "learning_rate": 8.650356778797146e-05, "loss": 2.9609, "step": 17212 }, { "epoch": 1.35, "learning_rate": 8.650278365874696e-05, "loss": 2.8306, "step": 17213 }, { "epoch": 1.35, "learning_rate": 8.650199952952247e-05, "loss": 2.8453, "step": 17214 }, { "epoch": 1.35, "learning_rate": 8.650121540029797e-05, "loss": 2.9146, "step": 17215 }, { "epoch": 1.35, "learning_rate": 8.650043127107347e-05, "loss": 2.934, "step": 17216 }, { "epoch": 1.35, "learning_rate": 8.649964714184898e-05, "loss": 2.6891, "step": 17217 }, { "epoch": 1.35, "learning_rate": 8.649886301262448e-05, "loss": 2.9146, "step": 17218 }, { "epoch": 1.35, "learning_rate": 8.649807888339999e-05, "loss": 2.9701, "step": 17219 }, { "epoch": 1.35, "learning_rate": 8.649729475417549e-05, "loss": 2.7623, "step": 17220 }, { "epoch": 1.35, "learning_rate": 8.6496510624951e-05, "loss": 2.8413, "step": 17221 }, { "epoch": 1.35, "learning_rate": 8.64957264957265e-05, "loss": 2.7954, "step": 17222 }, { "epoch": 1.35, "learning_rate": 8.649494236650201e-05, "loss": 2.9776, "step": 17223 }, { "epoch": 1.35, "learning_rate": 8.649415823727752e-05, "loss": 2.8942, "step": 17224 }, { "epoch": 1.35, "learning_rate": 8.649337410805302e-05, "loss": 3.2518, "step": 17225 }, { "epoch": 1.35, "learning_rate": 8.649258997882853e-05, "loss": 3.0056, "step": 17226 }, { "epoch": 1.35, "learning_rate": 8.649180584960403e-05, "loss": 2.8839, "step": 17227 }, { "epoch": 1.35, "learning_rate": 8.649102172037953e-05, "loss": 2.8147, "step": 17228 }, { "epoch": 1.35, "learning_rate": 8.649023759115502e-05, "loss": 2.7462, "step": 17229 }, { "epoch": 1.35, "learning_rate": 8.648945346193053e-05, "loss": 3.0433, "step": 17230 }, { "epoch": 1.35, "learning_rate": 8.648866933270603e-05, "loss": 2.9283, "step": 17231 }, { "epoch": 1.35, "learning_rate": 8.648788520348153e-05, "loss": 2.7902, "step": 17232 }, { "epoch": 1.35, "learning_rate": 8.648710107425704e-05, "loss": 2.8081, "step": 17233 }, { "epoch": 1.35, "learning_rate": 8.648631694503254e-05, "loss": 3.1081, "step": 17234 }, { "epoch": 1.35, "learning_rate": 8.648553281580804e-05, "loss": 3.0798, "step": 17235 }, { "epoch": 1.35, "learning_rate": 8.648474868658355e-05, "loss": 2.8487, "step": 17236 }, { "epoch": 1.35, "learning_rate": 8.648396455735905e-05, "loss": 2.6667, "step": 17237 }, { "epoch": 1.35, "learning_rate": 8.648318042813456e-05, "loss": 2.7051, "step": 17238 }, { "epoch": 1.35, "learning_rate": 8.648239629891006e-05, "loss": 2.7551, "step": 17239 }, { "epoch": 1.35, "learning_rate": 8.648161216968556e-05, "loss": 2.9747, "step": 17240 }, { "epoch": 1.35, "learning_rate": 8.648082804046107e-05, "loss": 2.904, "step": 17241 }, { "epoch": 1.35, "learning_rate": 8.648004391123657e-05, "loss": 2.8932, "step": 17242 }, { "epoch": 1.35, "learning_rate": 8.647925978201207e-05, "loss": 2.8068, "step": 17243 }, { "epoch": 1.35, "learning_rate": 8.647847565278758e-05, "loss": 2.7567, "step": 17244 }, { "epoch": 1.35, "learning_rate": 8.647769152356308e-05, "loss": 3.0042, "step": 17245 }, { "epoch": 1.35, "learning_rate": 8.647690739433859e-05, "loss": 2.7489, "step": 17246 }, { "epoch": 1.35, "learning_rate": 8.647612326511409e-05, "loss": 2.7236, "step": 17247 }, { "epoch": 1.35, "learning_rate": 8.647533913588959e-05, "loss": 3.0692, "step": 17248 }, { "epoch": 1.35, "learning_rate": 8.647455500666511e-05, "loss": 2.8542, "step": 17249 }, { "epoch": 1.35, "learning_rate": 8.647377087744061e-05, "loss": 2.7901, "step": 17250 }, { "epoch": 1.35, "learning_rate": 8.647298674821612e-05, "loss": 2.7881, "step": 17251 }, { "epoch": 1.35, "learning_rate": 8.647220261899162e-05, "loss": 2.8226, "step": 17252 }, { "epoch": 1.35, "learning_rate": 8.647141848976712e-05, "loss": 2.939, "step": 17253 }, { "epoch": 1.35, "learning_rate": 8.647063436054263e-05, "loss": 2.7677, "step": 17254 }, { "epoch": 1.35, "learning_rate": 8.646985023131813e-05, "loss": 2.8059, "step": 17255 }, { "epoch": 1.35, "learning_rate": 8.646906610209364e-05, "loss": 2.814, "step": 17256 }, { "epoch": 1.35, "learning_rate": 8.646828197286914e-05, "loss": 3.0157, "step": 17257 }, { "epoch": 1.35, "learning_rate": 8.646749784364464e-05, "loss": 2.8081, "step": 17258 }, { "epoch": 1.35, "learning_rate": 8.646671371442015e-05, "loss": 2.7751, "step": 17259 }, { "epoch": 1.35, "learning_rate": 8.646592958519565e-05, "loss": 2.8998, "step": 17260 }, { "epoch": 1.35, "learning_rate": 8.646514545597115e-05, "loss": 2.8152, "step": 17261 }, { "epoch": 1.35, "learning_rate": 8.646436132674666e-05, "loss": 2.8796, "step": 17262 }, { "epoch": 1.35, "learning_rate": 8.646357719752216e-05, "loss": 3.0846, "step": 17263 }, { "epoch": 1.35, "learning_rate": 8.646279306829766e-05, "loss": 3.0068, "step": 17264 }, { "epoch": 1.35, "learning_rate": 8.646200893907316e-05, "loss": 2.8505, "step": 17265 }, { "epoch": 1.35, "learning_rate": 8.646122480984866e-05, "loss": 2.9552, "step": 17266 }, { "epoch": 1.35, "learning_rate": 8.646044068062416e-05, "loss": 3.0437, "step": 17267 }, { "epoch": 1.35, "learning_rate": 8.645965655139967e-05, "loss": 3.0522, "step": 17268 }, { "epoch": 1.35, "learning_rate": 8.645887242217517e-05, "loss": 2.8701, "step": 17269 }, { "epoch": 1.35, "learning_rate": 8.645808829295067e-05, "loss": 2.6944, "step": 17270 }, { "epoch": 1.35, "learning_rate": 8.645730416372618e-05, "loss": 2.911, "step": 17271 }, { "epoch": 1.35, "learning_rate": 8.645652003450168e-05, "loss": 2.9504, "step": 17272 }, { "epoch": 1.35, "learning_rate": 8.645573590527718e-05, "loss": 2.8657, "step": 17273 }, { "epoch": 1.35, "learning_rate": 8.64549517760527e-05, "loss": 3.0198, "step": 17274 }, { "epoch": 1.35, "learning_rate": 8.64541676468282e-05, "loss": 2.7298, "step": 17275 }, { "epoch": 1.35, "learning_rate": 8.645338351760371e-05, "loss": 2.8313, "step": 17276 }, { "epoch": 1.35, "learning_rate": 8.645259938837921e-05, "loss": 3.2657, "step": 17277 }, { "epoch": 1.35, "learning_rate": 8.645181525915472e-05, "loss": 2.8463, "step": 17278 }, { "epoch": 1.35, "learning_rate": 8.645103112993022e-05, "loss": 2.762, "step": 17279 }, { "epoch": 1.35, "learning_rate": 8.645024700070572e-05, "loss": 3.0816, "step": 17280 }, { "epoch": 1.36, "learning_rate": 8.644946287148123e-05, "loss": 3.1487, "step": 17281 }, { "epoch": 1.36, "learning_rate": 8.644867874225673e-05, "loss": 2.8911, "step": 17282 }, { "epoch": 1.36, "learning_rate": 8.644789461303223e-05, "loss": 2.898, "step": 17283 }, { "epoch": 1.36, "learning_rate": 8.644711048380774e-05, "loss": 2.8161, "step": 17284 }, { "epoch": 1.36, "learning_rate": 8.644632635458324e-05, "loss": 3.0016, "step": 17285 }, { "epoch": 1.36, "learning_rate": 8.644554222535875e-05, "loss": 3.1386, "step": 17286 }, { "epoch": 1.36, "learning_rate": 8.644475809613425e-05, "loss": 2.8083, "step": 17287 }, { "epoch": 1.36, "learning_rate": 8.644397396690975e-05, "loss": 3.002, "step": 17288 }, { "epoch": 1.36, "learning_rate": 8.644318983768526e-05, "loss": 2.7573, "step": 17289 }, { "epoch": 1.36, "learning_rate": 8.644240570846076e-05, "loss": 2.9012, "step": 17290 }, { "epoch": 1.36, "learning_rate": 8.644162157923626e-05, "loss": 2.8871, "step": 17291 }, { "epoch": 1.36, "learning_rate": 8.644083745001177e-05, "loss": 2.9446, "step": 17292 }, { "epoch": 1.36, "learning_rate": 8.644005332078727e-05, "loss": 2.9893, "step": 17293 }, { "epoch": 1.36, "learning_rate": 8.643926919156278e-05, "loss": 2.7074, "step": 17294 }, { "epoch": 1.36, "learning_rate": 8.643848506233828e-05, "loss": 2.6997, "step": 17295 }, { "epoch": 1.36, "learning_rate": 8.643770093311378e-05, "loss": 2.7455, "step": 17296 }, { "epoch": 1.36, "learning_rate": 8.643691680388929e-05, "loss": 2.8717, "step": 17297 }, { "epoch": 1.36, "learning_rate": 8.643613267466479e-05, "loss": 2.9021, "step": 17298 }, { "epoch": 1.36, "learning_rate": 8.64353485454403e-05, "loss": 3.2271, "step": 17299 }, { "epoch": 1.36, "learning_rate": 8.64345644162158e-05, "loss": 3.0209, "step": 17300 }, { "epoch": 1.36, "learning_rate": 8.64337802869913e-05, "loss": 2.8616, "step": 17301 }, { "epoch": 1.36, "learning_rate": 8.64329961577668e-05, "loss": 3.199, "step": 17302 }, { "epoch": 1.36, "learning_rate": 8.643221202854231e-05, "loss": 2.8544, "step": 17303 }, { "epoch": 1.36, "learning_rate": 8.643142789931781e-05, "loss": 2.7055, "step": 17304 }, { "epoch": 1.36, "learning_rate": 8.643064377009332e-05, "loss": 3.1581, "step": 17305 }, { "epoch": 1.36, "learning_rate": 8.642985964086882e-05, "loss": 2.8446, "step": 17306 }, { "epoch": 1.36, "learning_rate": 8.642907551164432e-05, "loss": 2.9796, "step": 17307 }, { "epoch": 1.36, "learning_rate": 8.642829138241983e-05, "loss": 2.7561, "step": 17308 }, { "epoch": 1.36, "learning_rate": 8.642750725319533e-05, "loss": 3.0201, "step": 17309 }, { "epoch": 1.36, "learning_rate": 8.642672312397083e-05, "loss": 2.8248, "step": 17310 }, { "epoch": 1.36, "learning_rate": 8.642593899474634e-05, "loss": 3.1101, "step": 17311 }, { "epoch": 1.36, "learning_rate": 8.642515486552184e-05, "loss": 2.9274, "step": 17312 }, { "epoch": 1.36, "learning_rate": 8.642437073629735e-05, "loss": 3.0533, "step": 17313 }, { "epoch": 1.36, "learning_rate": 8.642358660707285e-05, "loss": 3.0085, "step": 17314 }, { "epoch": 1.36, "learning_rate": 8.642280247784835e-05, "loss": 3.0585, "step": 17315 }, { "epoch": 1.36, "learning_rate": 8.642201834862386e-05, "loss": 2.9274, "step": 17316 }, { "epoch": 1.36, "learning_rate": 8.642123421939936e-05, "loss": 2.9355, "step": 17317 }, { "epoch": 1.36, "learning_rate": 8.642045009017486e-05, "loss": 2.9471, "step": 17318 }, { "epoch": 1.36, "learning_rate": 8.641966596095037e-05, "loss": 2.7343, "step": 17319 }, { "epoch": 1.36, "learning_rate": 8.641888183172587e-05, "loss": 2.8896, "step": 17320 }, { "epoch": 1.36, "learning_rate": 8.641809770250137e-05, "loss": 2.7981, "step": 17321 }, { "epoch": 1.36, "learning_rate": 8.641731357327688e-05, "loss": 2.9093, "step": 17322 }, { "epoch": 1.36, "learning_rate": 8.641652944405238e-05, "loss": 2.7589, "step": 17323 }, { "epoch": 1.36, "learning_rate": 8.641574531482789e-05, "loss": 3.0979, "step": 17324 }, { "epoch": 1.36, "learning_rate": 8.641496118560339e-05, "loss": 3.032, "step": 17325 }, { "epoch": 1.36, "learning_rate": 8.641417705637889e-05, "loss": 2.8326, "step": 17326 }, { "epoch": 1.36, "learning_rate": 8.64133929271544e-05, "loss": 2.8866, "step": 17327 }, { "epoch": 1.36, "learning_rate": 8.64126087979299e-05, "loss": 3.0438, "step": 17328 }, { "epoch": 1.36, "learning_rate": 8.64118246687054e-05, "loss": 2.9176, "step": 17329 }, { "epoch": 1.36, "learning_rate": 8.641104053948091e-05, "loss": 2.9874, "step": 17330 }, { "epoch": 1.36, "learning_rate": 8.641025641025642e-05, "loss": 2.9167, "step": 17331 }, { "epoch": 1.36, "learning_rate": 8.640947228103193e-05, "loss": 2.8938, "step": 17332 }, { "epoch": 1.36, "learning_rate": 8.640868815180743e-05, "loss": 2.8133, "step": 17333 }, { "epoch": 1.36, "learning_rate": 8.640790402258294e-05, "loss": 2.7315, "step": 17334 }, { "epoch": 1.36, "learning_rate": 8.640711989335844e-05, "loss": 2.9187, "step": 17335 }, { "epoch": 1.36, "learning_rate": 8.640633576413394e-05, "loss": 3.0585, "step": 17336 }, { "epoch": 1.36, "learning_rate": 8.640555163490943e-05, "loss": 3.1909, "step": 17337 }, { "epoch": 1.36, "learning_rate": 8.640476750568494e-05, "loss": 2.7396, "step": 17338 }, { "epoch": 1.36, "learning_rate": 8.640398337646044e-05, "loss": 2.7221, "step": 17339 }, { "epoch": 1.36, "learning_rate": 8.640319924723594e-05, "loss": 3.0159, "step": 17340 }, { "epoch": 1.36, "learning_rate": 8.640241511801145e-05, "loss": 2.9268, "step": 17341 }, { "epoch": 1.36, "learning_rate": 8.640163098878695e-05, "loss": 2.7994, "step": 17342 }, { "epoch": 1.36, "learning_rate": 8.640084685956246e-05, "loss": 2.9946, "step": 17343 }, { "epoch": 1.36, "learning_rate": 8.640006273033796e-05, "loss": 2.9548, "step": 17344 }, { "epoch": 1.36, "learning_rate": 8.639927860111346e-05, "loss": 2.7461, "step": 17345 }, { "epoch": 1.36, "learning_rate": 8.639849447188897e-05, "loss": 2.8871, "step": 17346 }, { "epoch": 1.36, "learning_rate": 8.639771034266447e-05, "loss": 2.7342, "step": 17347 }, { "epoch": 1.36, "learning_rate": 8.639692621343997e-05, "loss": 2.8631, "step": 17348 }, { "epoch": 1.36, "learning_rate": 8.639614208421548e-05, "loss": 2.8271, "step": 17349 }, { "epoch": 1.36, "learning_rate": 8.639535795499098e-05, "loss": 2.8328, "step": 17350 }, { "epoch": 1.36, "learning_rate": 8.639457382576648e-05, "loss": 3.0944, "step": 17351 }, { "epoch": 1.36, "learning_rate": 8.639378969654199e-05, "loss": 2.7156, "step": 17352 }, { "epoch": 1.36, "learning_rate": 8.639300556731749e-05, "loss": 2.9268, "step": 17353 }, { "epoch": 1.36, "learning_rate": 8.6392221438093e-05, "loss": 2.9463, "step": 17354 }, { "epoch": 1.36, "learning_rate": 8.63914373088685e-05, "loss": 2.7256, "step": 17355 }, { "epoch": 1.36, "learning_rate": 8.6390653179644e-05, "loss": 2.73, "step": 17356 }, { "epoch": 1.36, "learning_rate": 8.638986905041952e-05, "loss": 3.2333, "step": 17357 }, { "epoch": 1.36, "learning_rate": 8.638908492119502e-05, "loss": 2.9233, "step": 17358 }, { "epoch": 1.36, "learning_rate": 8.638830079197053e-05, "loss": 3.1362, "step": 17359 }, { "epoch": 1.36, "learning_rate": 8.638751666274603e-05, "loss": 2.941, "step": 17360 }, { "epoch": 1.36, "learning_rate": 8.638673253352154e-05, "loss": 2.7954, "step": 17361 }, { "epoch": 1.36, "learning_rate": 8.638594840429704e-05, "loss": 3.2109, "step": 17362 }, { "epoch": 1.36, "learning_rate": 8.638516427507254e-05, "loss": 3.0607, "step": 17363 }, { "epoch": 1.36, "learning_rate": 8.638438014584805e-05, "loss": 2.8782, "step": 17364 }, { "epoch": 1.36, "learning_rate": 8.638359601662355e-05, "loss": 2.7386, "step": 17365 }, { "epoch": 1.36, "learning_rate": 8.638281188739905e-05, "loss": 2.9204, "step": 17366 }, { "epoch": 1.36, "learning_rate": 8.638202775817456e-05, "loss": 2.8857, "step": 17367 }, { "epoch": 1.36, "learning_rate": 8.638124362895006e-05, "loss": 2.9421, "step": 17368 }, { "epoch": 1.36, "learning_rate": 8.638045949972556e-05, "loss": 2.7539, "step": 17369 }, { "epoch": 1.36, "learning_rate": 8.637967537050107e-05, "loss": 3.1265, "step": 17370 }, { "epoch": 1.36, "learning_rate": 8.637889124127657e-05, "loss": 3.0012, "step": 17371 }, { "epoch": 1.36, "learning_rate": 8.637810711205208e-05, "loss": 2.9897, "step": 17372 }, { "epoch": 1.36, "learning_rate": 8.637732298282757e-05, "loss": 3.0337, "step": 17373 }, { "epoch": 1.36, "learning_rate": 8.637653885360307e-05, "loss": 2.8371, "step": 17374 }, { "epoch": 1.36, "learning_rate": 8.637575472437857e-05, "loss": 3.2223, "step": 17375 }, { "epoch": 1.36, "learning_rate": 8.637497059515408e-05, "loss": 3.0036, "step": 17376 }, { "epoch": 1.36, "learning_rate": 8.637418646592958e-05, "loss": 3.0743, "step": 17377 }, { "epoch": 1.36, "learning_rate": 8.637340233670508e-05, "loss": 2.8872, "step": 17378 }, { "epoch": 1.36, "learning_rate": 8.637261820748059e-05, "loss": 2.9004, "step": 17379 }, { "epoch": 1.36, "learning_rate": 8.637183407825609e-05, "loss": 3.1127, "step": 17380 }, { "epoch": 1.36, "learning_rate": 8.63710499490316e-05, "loss": 3.3307, "step": 17381 }, { "epoch": 1.36, "learning_rate": 8.637026581980711e-05, "loss": 2.7268, "step": 17382 }, { "epoch": 1.36, "learning_rate": 8.636948169058262e-05, "loss": 2.9886, "step": 17383 }, { "epoch": 1.36, "learning_rate": 8.636869756135812e-05, "loss": 3.0126, "step": 17384 }, { "epoch": 1.36, "learning_rate": 8.636791343213362e-05, "loss": 2.8958, "step": 17385 }, { "epoch": 1.36, "learning_rate": 8.636712930290913e-05, "loss": 2.7817, "step": 17386 }, { "epoch": 1.36, "learning_rate": 8.636634517368463e-05, "loss": 2.9565, "step": 17387 }, { "epoch": 1.36, "learning_rate": 8.636556104446013e-05, "loss": 2.8539, "step": 17388 }, { "epoch": 1.36, "learning_rate": 8.636477691523564e-05, "loss": 3.0631, "step": 17389 }, { "epoch": 1.36, "learning_rate": 8.636399278601114e-05, "loss": 2.9544, "step": 17390 }, { "epoch": 1.36, "learning_rate": 8.636320865678665e-05, "loss": 3.0922, "step": 17391 }, { "epoch": 1.36, "learning_rate": 8.636242452756215e-05, "loss": 2.8485, "step": 17392 }, { "epoch": 1.36, "learning_rate": 8.636164039833765e-05, "loss": 2.7708, "step": 17393 }, { "epoch": 1.36, "learning_rate": 8.636085626911316e-05, "loss": 2.9151, "step": 17394 }, { "epoch": 1.36, "learning_rate": 8.636007213988866e-05, "loss": 3.1387, "step": 17395 }, { "epoch": 1.36, "learning_rate": 8.635928801066416e-05, "loss": 3.2751, "step": 17396 }, { "epoch": 1.36, "learning_rate": 8.635850388143967e-05, "loss": 2.8739, "step": 17397 }, { "epoch": 1.36, "learning_rate": 8.635771975221517e-05, "loss": 2.7888, "step": 17398 }, { "epoch": 1.36, "learning_rate": 8.635693562299068e-05, "loss": 3.1387, "step": 17399 }, { "epoch": 1.36, "learning_rate": 8.635615149376618e-05, "loss": 3.0401, "step": 17400 }, { "epoch": 1.36, "learning_rate": 8.635536736454168e-05, "loss": 3.1207, "step": 17401 }, { "epoch": 1.36, "learning_rate": 8.635458323531719e-05, "loss": 3.0034, "step": 17402 }, { "epoch": 1.36, "learning_rate": 8.635379910609269e-05, "loss": 3.1387, "step": 17403 }, { "epoch": 1.36, "learning_rate": 8.63530149768682e-05, "loss": 2.8977, "step": 17404 }, { "epoch": 1.36, "learning_rate": 8.63522308476437e-05, "loss": 2.9895, "step": 17405 }, { "epoch": 1.36, "learning_rate": 8.63514467184192e-05, "loss": 3.0649, "step": 17406 }, { "epoch": 1.36, "learning_rate": 8.63506625891947e-05, "loss": 2.9652, "step": 17407 }, { "epoch": 1.37, "learning_rate": 8.634987845997021e-05, "loss": 2.8409, "step": 17408 }, { "epoch": 1.37, "learning_rate": 8.634909433074571e-05, "loss": 2.8411, "step": 17409 }, { "epoch": 1.37, "learning_rate": 8.634831020152122e-05, "loss": 2.8997, "step": 17410 }, { "epoch": 1.37, "learning_rate": 8.634752607229672e-05, "loss": 2.6742, "step": 17411 }, { "epoch": 1.37, "learning_rate": 8.634674194307222e-05, "loss": 2.8539, "step": 17412 }, { "epoch": 1.37, "learning_rate": 8.634595781384773e-05, "loss": 3.0641, "step": 17413 }, { "epoch": 1.37, "learning_rate": 8.634517368462323e-05, "loss": 3.0788, "step": 17414 }, { "epoch": 1.37, "learning_rate": 8.634438955539873e-05, "loss": 3.0539, "step": 17415 }, { "epoch": 1.37, "learning_rate": 8.634360542617424e-05, "loss": 2.9501, "step": 17416 }, { "epoch": 1.37, "learning_rate": 8.634282129694974e-05, "loss": 3.0597, "step": 17417 }, { "epoch": 1.37, "learning_rate": 8.634203716772524e-05, "loss": 2.92, "step": 17418 }, { "epoch": 1.37, "learning_rate": 8.634125303850075e-05, "loss": 2.9095, "step": 17419 }, { "epoch": 1.37, "learning_rate": 8.634046890927625e-05, "loss": 2.9423, "step": 17420 }, { "epoch": 1.37, "learning_rate": 8.633968478005176e-05, "loss": 2.8025, "step": 17421 }, { "epoch": 1.37, "learning_rate": 8.633890065082726e-05, "loss": 2.8302, "step": 17422 }, { "epoch": 1.37, "learning_rate": 8.633811652160276e-05, "loss": 2.8099, "step": 17423 }, { "epoch": 1.37, "learning_rate": 8.633733239237827e-05, "loss": 2.8755, "step": 17424 }, { "epoch": 1.37, "learning_rate": 8.633654826315377e-05, "loss": 2.8771, "step": 17425 }, { "epoch": 1.37, "learning_rate": 8.633576413392927e-05, "loss": 2.6983, "step": 17426 }, { "epoch": 1.37, "learning_rate": 8.633498000470478e-05, "loss": 2.718, "step": 17427 }, { "epoch": 1.37, "learning_rate": 8.633419587548028e-05, "loss": 2.9608, "step": 17428 }, { "epoch": 1.37, "learning_rate": 8.633341174625579e-05, "loss": 2.9751, "step": 17429 }, { "epoch": 1.37, "learning_rate": 8.633262761703129e-05, "loss": 2.7074, "step": 17430 }, { "epoch": 1.37, "learning_rate": 8.633184348780679e-05, "loss": 2.724, "step": 17431 }, { "epoch": 1.37, "learning_rate": 8.63310593585823e-05, "loss": 2.9228, "step": 17432 }, { "epoch": 1.37, "learning_rate": 8.63302752293578e-05, "loss": 2.8798, "step": 17433 }, { "epoch": 1.37, "learning_rate": 8.63294911001333e-05, "loss": 2.693, "step": 17434 }, { "epoch": 1.37, "learning_rate": 8.632870697090881e-05, "loss": 3.0025, "step": 17435 }, { "epoch": 1.37, "learning_rate": 8.632792284168431e-05, "loss": 2.9575, "step": 17436 }, { "epoch": 1.37, "learning_rate": 8.632713871245981e-05, "loss": 2.8141, "step": 17437 }, { "epoch": 1.37, "learning_rate": 8.632635458323532e-05, "loss": 2.8627, "step": 17438 }, { "epoch": 1.37, "learning_rate": 8.632557045401084e-05, "loss": 2.9399, "step": 17439 }, { "epoch": 1.37, "learning_rate": 8.632478632478634e-05, "loss": 3.0935, "step": 17440 }, { "epoch": 1.37, "learning_rate": 8.632400219556184e-05, "loss": 2.9137, "step": 17441 }, { "epoch": 1.37, "learning_rate": 8.632321806633735e-05, "loss": 2.7946, "step": 17442 }, { "epoch": 1.37, "learning_rate": 8.632243393711285e-05, "loss": 2.8399, "step": 17443 }, { "epoch": 1.37, "learning_rate": 8.632164980788835e-05, "loss": 2.871, "step": 17444 }, { "epoch": 1.37, "learning_rate": 8.632086567866384e-05, "loss": 2.8951, "step": 17445 }, { "epoch": 1.37, "learning_rate": 8.632008154943935e-05, "loss": 3.0129, "step": 17446 }, { "epoch": 1.37, "learning_rate": 8.631929742021485e-05, "loss": 2.7047, "step": 17447 }, { "epoch": 1.37, "learning_rate": 8.631851329099036e-05, "loss": 2.8695, "step": 17448 }, { "epoch": 1.37, "learning_rate": 8.631772916176586e-05, "loss": 2.7603, "step": 17449 }, { "epoch": 1.37, "learning_rate": 8.631694503254136e-05, "loss": 3.1514, "step": 17450 }, { "epoch": 1.37, "learning_rate": 8.631616090331687e-05, "loss": 3.0393, "step": 17451 }, { "epoch": 1.37, "learning_rate": 8.631537677409237e-05, "loss": 2.7798, "step": 17452 }, { "epoch": 1.37, "learning_rate": 8.631459264486787e-05, "loss": 2.8677, "step": 17453 }, { "epoch": 1.37, "learning_rate": 8.631380851564338e-05, "loss": 2.8349, "step": 17454 }, { "epoch": 1.37, "learning_rate": 8.631302438641888e-05, "loss": 2.833, "step": 17455 }, { "epoch": 1.37, "learning_rate": 8.631224025719438e-05, "loss": 2.8876, "step": 17456 }, { "epoch": 1.37, "learning_rate": 8.631145612796989e-05, "loss": 2.9255, "step": 17457 }, { "epoch": 1.37, "learning_rate": 8.631067199874539e-05, "loss": 2.9446, "step": 17458 }, { "epoch": 1.37, "learning_rate": 8.63098878695209e-05, "loss": 3.0404, "step": 17459 }, { "epoch": 1.37, "learning_rate": 8.63091037402964e-05, "loss": 2.9119, "step": 17460 }, { "epoch": 1.37, "learning_rate": 8.63083196110719e-05, "loss": 2.6216, "step": 17461 }, { "epoch": 1.37, "learning_rate": 8.63075354818474e-05, "loss": 3.0927, "step": 17462 }, { "epoch": 1.37, "learning_rate": 8.630675135262291e-05, "loss": 3.1389, "step": 17463 }, { "epoch": 1.37, "learning_rate": 8.630596722339841e-05, "loss": 2.8392, "step": 17464 }, { "epoch": 1.37, "learning_rate": 8.630518309417393e-05, "loss": 2.8506, "step": 17465 }, { "epoch": 1.37, "learning_rate": 8.630439896494943e-05, "loss": 3.0811, "step": 17466 }, { "epoch": 1.37, "learning_rate": 8.630361483572494e-05, "loss": 3.0412, "step": 17467 }, { "epoch": 1.37, "learning_rate": 8.630283070650044e-05, "loss": 2.8392, "step": 17468 }, { "epoch": 1.37, "learning_rate": 8.630204657727595e-05, "loss": 2.8447, "step": 17469 }, { "epoch": 1.37, "learning_rate": 8.630126244805145e-05, "loss": 2.7955, "step": 17470 }, { "epoch": 1.37, "learning_rate": 8.630047831882695e-05, "loss": 2.9889, "step": 17471 }, { "epoch": 1.37, "learning_rate": 8.629969418960246e-05, "loss": 3.3964, "step": 17472 }, { "epoch": 1.37, "learning_rate": 8.629891006037796e-05, "loss": 2.9019, "step": 17473 }, { "epoch": 1.37, "learning_rate": 8.629812593115346e-05, "loss": 3.0502, "step": 17474 }, { "epoch": 1.37, "learning_rate": 8.629734180192897e-05, "loss": 3.0873, "step": 17475 }, { "epoch": 1.37, "learning_rate": 8.629655767270447e-05, "loss": 2.5993, "step": 17476 }, { "epoch": 1.37, "learning_rate": 8.629577354347998e-05, "loss": 2.9817, "step": 17477 }, { "epoch": 1.37, "learning_rate": 8.629498941425548e-05, "loss": 2.8983, "step": 17478 }, { "epoch": 1.37, "learning_rate": 8.629420528503098e-05, "loss": 2.7392, "step": 17479 }, { "epoch": 1.37, "learning_rate": 8.629342115580649e-05, "loss": 2.9505, "step": 17480 }, { "epoch": 1.37, "learning_rate": 8.629263702658198e-05, "loss": 2.8103, "step": 17481 }, { "epoch": 1.37, "learning_rate": 8.629185289735748e-05, "loss": 2.9232, "step": 17482 }, { "epoch": 1.37, "learning_rate": 8.629106876813298e-05, "loss": 2.9222, "step": 17483 }, { "epoch": 1.37, "learning_rate": 8.629028463890849e-05, "loss": 2.8366, "step": 17484 }, { "epoch": 1.37, "learning_rate": 8.628950050968399e-05, "loss": 2.8713, "step": 17485 }, { "epoch": 1.37, "learning_rate": 8.62887163804595e-05, "loss": 2.8465, "step": 17486 }, { "epoch": 1.37, "learning_rate": 8.6287932251235e-05, "loss": 2.8709, "step": 17487 }, { "epoch": 1.37, "learning_rate": 8.62871481220105e-05, "loss": 2.9554, "step": 17488 }, { "epoch": 1.37, "learning_rate": 8.6286363992786e-05, "loss": 3.0694, "step": 17489 }, { "epoch": 1.37, "learning_rate": 8.628557986356152e-05, "loss": 2.7846, "step": 17490 }, { "epoch": 1.37, "learning_rate": 8.628479573433703e-05, "loss": 2.9066, "step": 17491 }, { "epoch": 1.37, "learning_rate": 8.628401160511253e-05, "loss": 2.8076, "step": 17492 }, { "epoch": 1.37, "learning_rate": 8.628322747588803e-05, "loss": 3.0815, "step": 17493 }, { "epoch": 1.37, "learning_rate": 8.628244334666354e-05, "loss": 2.8413, "step": 17494 }, { "epoch": 1.37, "learning_rate": 8.628165921743904e-05, "loss": 2.8322, "step": 17495 }, { "epoch": 1.37, "learning_rate": 8.628087508821455e-05, "loss": 2.8823, "step": 17496 }, { "epoch": 1.37, "learning_rate": 8.628009095899005e-05, "loss": 2.9551, "step": 17497 }, { "epoch": 1.37, "learning_rate": 8.627930682976555e-05, "loss": 3.0929, "step": 17498 }, { "epoch": 1.37, "learning_rate": 8.627852270054106e-05, "loss": 3.0144, "step": 17499 }, { "epoch": 1.37, "learning_rate": 8.627773857131656e-05, "loss": 3.1092, "step": 17500 }, { "epoch": 1.37, "learning_rate": 8.627695444209206e-05, "loss": 2.8179, "step": 17501 }, { "epoch": 1.37, "learning_rate": 8.627617031286757e-05, "loss": 3.0076, "step": 17502 }, { "epoch": 1.37, "learning_rate": 8.627538618364307e-05, "loss": 3.0167, "step": 17503 }, { "epoch": 1.37, "learning_rate": 8.627460205441857e-05, "loss": 2.7377, "step": 17504 }, { "epoch": 1.37, "learning_rate": 8.627381792519408e-05, "loss": 2.6651, "step": 17505 }, { "epoch": 1.37, "learning_rate": 8.627303379596958e-05, "loss": 2.656, "step": 17506 }, { "epoch": 1.37, "learning_rate": 8.627224966674509e-05, "loss": 2.9393, "step": 17507 }, { "epoch": 1.37, "learning_rate": 8.627146553752059e-05, "loss": 2.9806, "step": 17508 }, { "epoch": 1.37, "learning_rate": 8.627068140829609e-05, "loss": 2.8965, "step": 17509 }, { "epoch": 1.37, "learning_rate": 8.62698972790716e-05, "loss": 2.725, "step": 17510 }, { "epoch": 1.37, "learning_rate": 8.62691131498471e-05, "loss": 2.9886, "step": 17511 }, { "epoch": 1.37, "learning_rate": 8.62683290206226e-05, "loss": 2.9392, "step": 17512 }, { "epoch": 1.37, "learning_rate": 8.626754489139811e-05, "loss": 2.9437, "step": 17513 }, { "epoch": 1.37, "learning_rate": 8.626676076217361e-05, "loss": 2.9454, "step": 17514 }, { "epoch": 1.37, "learning_rate": 8.626597663294912e-05, "loss": 2.8966, "step": 17515 }, { "epoch": 1.37, "learning_rate": 8.626519250372462e-05, "loss": 2.8303, "step": 17516 }, { "epoch": 1.37, "learning_rate": 8.626440837450012e-05, "loss": 2.8883, "step": 17517 }, { "epoch": 1.37, "learning_rate": 8.626362424527563e-05, "loss": 2.9163, "step": 17518 }, { "epoch": 1.37, "learning_rate": 8.626284011605113e-05, "loss": 2.911, "step": 17519 }, { "epoch": 1.37, "learning_rate": 8.626205598682663e-05, "loss": 3.041, "step": 17520 }, { "epoch": 1.37, "learning_rate": 8.626127185760214e-05, "loss": 2.7872, "step": 17521 }, { "epoch": 1.37, "learning_rate": 8.626048772837764e-05, "loss": 2.8781, "step": 17522 }, { "epoch": 1.37, "learning_rate": 8.625970359915314e-05, "loss": 3.2543, "step": 17523 }, { "epoch": 1.37, "learning_rate": 8.625891946992865e-05, "loss": 2.7481, "step": 17524 }, { "epoch": 1.37, "learning_rate": 8.625813534070415e-05, "loss": 2.7114, "step": 17525 }, { "epoch": 1.37, "learning_rate": 8.625735121147966e-05, "loss": 3.0282, "step": 17526 }, { "epoch": 1.37, "learning_rate": 8.625656708225516e-05, "loss": 2.7848, "step": 17527 }, { "epoch": 1.37, "learning_rate": 8.625578295303066e-05, "loss": 2.8302, "step": 17528 }, { "epoch": 1.37, "learning_rate": 8.625499882380617e-05, "loss": 2.9095, "step": 17529 }, { "epoch": 1.37, "learning_rate": 8.625421469458167e-05, "loss": 2.8385, "step": 17530 }, { "epoch": 1.37, "learning_rate": 8.625343056535717e-05, "loss": 2.9341, "step": 17531 }, { "epoch": 1.37, "learning_rate": 8.625264643613268e-05, "loss": 2.9919, "step": 17532 }, { "epoch": 1.37, "learning_rate": 8.625186230690818e-05, "loss": 2.7974, "step": 17533 }, { "epoch": 1.37, "learning_rate": 8.625107817768369e-05, "loss": 3.0098, "step": 17534 }, { "epoch": 1.37, "learning_rate": 8.625029404845919e-05, "loss": 2.8769, "step": 17535 }, { "epoch": 1.38, "learning_rate": 8.624950991923469e-05, "loss": 2.8041, "step": 17536 }, { "epoch": 1.38, "learning_rate": 8.62487257900102e-05, "loss": 2.904, "step": 17537 }, { "epoch": 1.38, "learning_rate": 8.62479416607857e-05, "loss": 2.9799, "step": 17538 }, { "epoch": 1.38, "learning_rate": 8.62471575315612e-05, "loss": 3.1044, "step": 17539 }, { "epoch": 1.38, "learning_rate": 8.624637340233671e-05, "loss": 2.7333, "step": 17540 }, { "epoch": 1.38, "learning_rate": 8.624558927311221e-05, "loss": 3.0023, "step": 17541 }, { "epoch": 1.38, "learning_rate": 8.624480514388771e-05, "loss": 2.854, "step": 17542 }, { "epoch": 1.38, "learning_rate": 8.624402101466322e-05, "loss": 2.9401, "step": 17543 }, { "epoch": 1.38, "learning_rate": 8.624323688543872e-05, "loss": 2.9741, "step": 17544 }, { "epoch": 1.38, "learning_rate": 8.624245275621423e-05, "loss": 2.9043, "step": 17545 }, { "epoch": 1.38, "learning_rate": 8.624166862698973e-05, "loss": 2.7065, "step": 17546 }, { "epoch": 1.38, "learning_rate": 8.624088449776525e-05, "loss": 2.823, "step": 17547 }, { "epoch": 1.38, "learning_rate": 8.624010036854075e-05, "loss": 3.2405, "step": 17548 }, { "epoch": 1.38, "learning_rate": 8.623931623931625e-05, "loss": 2.7581, "step": 17549 }, { "epoch": 1.38, "learning_rate": 8.623853211009176e-05, "loss": 2.9387, "step": 17550 }, { "epoch": 1.38, "learning_rate": 8.623774798086726e-05, "loss": 2.7654, "step": 17551 }, { "epoch": 1.38, "learning_rate": 8.623696385164276e-05, "loss": 2.9426, "step": 17552 }, { "epoch": 1.38, "learning_rate": 8.623617972241825e-05, "loss": 3.034, "step": 17553 }, { "epoch": 1.38, "learning_rate": 8.623539559319376e-05, "loss": 2.8354, "step": 17554 }, { "epoch": 1.38, "learning_rate": 8.623461146396926e-05, "loss": 2.8319, "step": 17555 }, { "epoch": 1.38, "learning_rate": 8.623382733474477e-05, "loss": 2.7649, "step": 17556 }, { "epoch": 1.38, "learning_rate": 8.623304320552027e-05, "loss": 2.8382, "step": 17557 }, { "epoch": 1.38, "learning_rate": 8.623225907629577e-05, "loss": 3.0344, "step": 17558 }, { "epoch": 1.38, "learning_rate": 8.623147494707128e-05, "loss": 3.0407, "step": 17559 }, { "epoch": 1.38, "learning_rate": 8.623069081784678e-05, "loss": 2.7501, "step": 17560 }, { "epoch": 1.38, "learning_rate": 8.622990668862228e-05, "loss": 2.9005, "step": 17561 }, { "epoch": 1.38, "learning_rate": 8.622912255939779e-05, "loss": 3.1491, "step": 17562 }, { "epoch": 1.38, "learning_rate": 8.622833843017329e-05, "loss": 2.9397, "step": 17563 }, { "epoch": 1.38, "learning_rate": 8.62275543009488e-05, "loss": 2.7321, "step": 17564 }, { "epoch": 1.38, "learning_rate": 8.62267701717243e-05, "loss": 3.0774, "step": 17565 }, { "epoch": 1.38, "learning_rate": 8.62259860424998e-05, "loss": 2.7806, "step": 17566 }, { "epoch": 1.38, "learning_rate": 8.62252019132753e-05, "loss": 3.1602, "step": 17567 }, { "epoch": 1.38, "learning_rate": 8.622441778405081e-05, "loss": 3.0175, "step": 17568 }, { "epoch": 1.38, "learning_rate": 8.622363365482631e-05, "loss": 3.241, "step": 17569 }, { "epoch": 1.38, "learning_rate": 8.622284952560182e-05, "loss": 2.9809, "step": 17570 }, { "epoch": 1.38, "learning_rate": 8.622206539637732e-05, "loss": 3.0445, "step": 17571 }, { "epoch": 1.38, "learning_rate": 8.622128126715282e-05, "loss": 2.8206, "step": 17572 }, { "epoch": 1.38, "learning_rate": 8.622049713792834e-05, "loss": 2.8158, "step": 17573 }, { "epoch": 1.38, "learning_rate": 8.621971300870385e-05, "loss": 2.7973, "step": 17574 }, { "epoch": 1.38, "learning_rate": 8.621892887947935e-05, "loss": 3.0485, "step": 17575 }, { "epoch": 1.38, "learning_rate": 8.621814475025485e-05, "loss": 2.8016, "step": 17576 }, { "epoch": 1.38, "learning_rate": 8.621736062103036e-05, "loss": 2.9848, "step": 17577 }, { "epoch": 1.38, "learning_rate": 8.621657649180586e-05, "loss": 2.7944, "step": 17578 }, { "epoch": 1.38, "learning_rate": 8.621579236258136e-05, "loss": 3.1152, "step": 17579 }, { "epoch": 1.38, "learning_rate": 8.621500823335687e-05, "loss": 2.9727, "step": 17580 }, { "epoch": 1.38, "learning_rate": 8.621422410413237e-05, "loss": 3.1049, "step": 17581 }, { "epoch": 1.38, "learning_rate": 8.621343997490788e-05, "loss": 3.2199, "step": 17582 }, { "epoch": 1.38, "learning_rate": 8.621265584568338e-05, "loss": 3.1181, "step": 17583 }, { "epoch": 1.38, "learning_rate": 8.621187171645888e-05, "loss": 2.7185, "step": 17584 }, { "epoch": 1.38, "learning_rate": 8.621108758723439e-05, "loss": 2.8988, "step": 17585 }, { "epoch": 1.38, "learning_rate": 8.621030345800989e-05, "loss": 2.7057, "step": 17586 }, { "epoch": 1.38, "learning_rate": 8.62095193287854e-05, "loss": 3.0807, "step": 17587 }, { "epoch": 1.38, "learning_rate": 8.62087351995609e-05, "loss": 2.9727, "step": 17588 }, { "epoch": 1.38, "learning_rate": 8.620795107033639e-05, "loss": 2.9143, "step": 17589 }, { "epoch": 1.38, "learning_rate": 8.620716694111189e-05, "loss": 2.7621, "step": 17590 }, { "epoch": 1.38, "learning_rate": 8.62063828118874e-05, "loss": 2.8426, "step": 17591 }, { "epoch": 1.38, "learning_rate": 8.62055986826629e-05, "loss": 2.887, "step": 17592 }, { "epoch": 1.38, "learning_rate": 8.62048145534384e-05, "loss": 2.873, "step": 17593 }, { "epoch": 1.38, "learning_rate": 8.62040304242139e-05, "loss": 2.9769, "step": 17594 }, { "epoch": 1.38, "learning_rate": 8.620324629498941e-05, "loss": 2.746, "step": 17595 }, { "epoch": 1.38, "learning_rate": 8.620246216576491e-05, "loss": 2.677, "step": 17596 }, { "epoch": 1.38, "learning_rate": 8.620167803654042e-05, "loss": 2.8165, "step": 17597 }, { "epoch": 1.38, "learning_rate": 8.620089390731592e-05, "loss": 3.2304, "step": 17598 }, { "epoch": 1.38, "learning_rate": 8.620010977809144e-05, "loss": 3.0641, "step": 17599 }, { "epoch": 1.38, "learning_rate": 8.619932564886694e-05, "loss": 2.7938, "step": 17600 }, { "epoch": 1.38, "learning_rate": 8.619854151964245e-05, "loss": 2.7325, "step": 17601 }, { "epoch": 1.38, "learning_rate": 8.619775739041795e-05, "loss": 2.9171, "step": 17602 }, { "epoch": 1.38, "learning_rate": 8.619697326119345e-05, "loss": 2.8484, "step": 17603 }, { "epoch": 1.38, "learning_rate": 8.619618913196896e-05, "loss": 2.8553, "step": 17604 }, { "epoch": 1.38, "learning_rate": 8.619540500274446e-05, "loss": 2.9551, "step": 17605 }, { "epoch": 1.38, "learning_rate": 8.619462087351996e-05, "loss": 2.8078, "step": 17606 }, { "epoch": 1.38, "learning_rate": 8.619383674429547e-05, "loss": 2.9015, "step": 17607 }, { "epoch": 1.38, "learning_rate": 8.619305261507097e-05, "loss": 2.9496, "step": 17608 }, { "epoch": 1.38, "learning_rate": 8.619226848584647e-05, "loss": 3.0371, "step": 17609 }, { "epoch": 1.38, "learning_rate": 8.619148435662198e-05, "loss": 2.9912, "step": 17610 }, { "epoch": 1.38, "learning_rate": 8.619070022739748e-05, "loss": 2.7302, "step": 17611 }, { "epoch": 1.38, "learning_rate": 8.618991609817299e-05, "loss": 2.6201, "step": 17612 }, { "epoch": 1.38, "learning_rate": 8.618913196894849e-05, "loss": 2.9785, "step": 17613 }, { "epoch": 1.38, "learning_rate": 8.618834783972399e-05, "loss": 2.8257, "step": 17614 }, { "epoch": 1.38, "learning_rate": 8.61875637104995e-05, "loss": 2.9928, "step": 17615 }, { "epoch": 1.38, "learning_rate": 8.6186779581275e-05, "loss": 2.857, "step": 17616 }, { "epoch": 1.38, "learning_rate": 8.61859954520505e-05, "loss": 2.8304, "step": 17617 }, { "epoch": 1.38, "learning_rate": 8.618521132282601e-05, "loss": 2.8437, "step": 17618 }, { "epoch": 1.38, "learning_rate": 8.618442719360151e-05, "loss": 2.7719, "step": 17619 }, { "epoch": 1.38, "learning_rate": 8.618364306437701e-05, "loss": 2.9949, "step": 17620 }, { "epoch": 1.38, "learning_rate": 8.618285893515252e-05, "loss": 2.8246, "step": 17621 }, { "epoch": 1.38, "learning_rate": 8.618207480592802e-05, "loss": 2.8737, "step": 17622 }, { "epoch": 1.38, "learning_rate": 8.618129067670353e-05, "loss": 2.7686, "step": 17623 }, { "epoch": 1.38, "learning_rate": 8.618050654747903e-05, "loss": 2.8073, "step": 17624 }, { "epoch": 1.38, "learning_rate": 8.617972241825453e-05, "loss": 2.872, "step": 17625 }, { "epoch": 1.38, "learning_rate": 8.617893828903004e-05, "loss": 2.8608, "step": 17626 }, { "epoch": 1.38, "learning_rate": 8.617815415980554e-05, "loss": 2.7106, "step": 17627 }, { "epoch": 1.38, "learning_rate": 8.617737003058104e-05, "loss": 2.9555, "step": 17628 }, { "epoch": 1.38, "learning_rate": 8.617658590135655e-05, "loss": 3.0028, "step": 17629 }, { "epoch": 1.38, "learning_rate": 8.617580177213205e-05, "loss": 3.1476, "step": 17630 }, { "epoch": 1.38, "learning_rate": 8.617501764290756e-05, "loss": 2.8645, "step": 17631 }, { "epoch": 1.38, "learning_rate": 8.617423351368306e-05, "loss": 3.0272, "step": 17632 }, { "epoch": 1.38, "learning_rate": 8.617344938445856e-05, "loss": 2.7735, "step": 17633 }, { "epoch": 1.38, "learning_rate": 8.617266525523407e-05, "loss": 3.0086, "step": 17634 }, { "epoch": 1.38, "learning_rate": 8.617188112600957e-05, "loss": 2.8317, "step": 17635 }, { "epoch": 1.38, "learning_rate": 8.617109699678507e-05, "loss": 2.8243, "step": 17636 }, { "epoch": 1.38, "learning_rate": 8.617031286756058e-05, "loss": 3.3265, "step": 17637 }, { "epoch": 1.38, "learning_rate": 8.616952873833608e-05, "loss": 2.9138, "step": 17638 }, { "epoch": 1.38, "learning_rate": 8.616874460911158e-05, "loss": 2.8713, "step": 17639 }, { "epoch": 1.38, "learning_rate": 8.616796047988709e-05, "loss": 2.7982, "step": 17640 }, { "epoch": 1.38, "learning_rate": 8.616717635066259e-05, "loss": 2.9057, "step": 17641 }, { "epoch": 1.38, "learning_rate": 8.61663922214381e-05, "loss": 2.8475, "step": 17642 }, { "epoch": 1.38, "learning_rate": 8.61656080922136e-05, "loss": 2.9521, "step": 17643 }, { "epoch": 1.38, "learning_rate": 8.61648239629891e-05, "loss": 3.0733, "step": 17644 }, { "epoch": 1.38, "learning_rate": 8.616403983376461e-05, "loss": 2.7853, "step": 17645 }, { "epoch": 1.38, "learning_rate": 8.616325570454011e-05, "loss": 3.0213, "step": 17646 }, { "epoch": 1.38, "learning_rate": 8.616247157531561e-05, "loss": 2.8287, "step": 17647 }, { "epoch": 1.38, "learning_rate": 8.616168744609112e-05, "loss": 2.7124, "step": 17648 }, { "epoch": 1.38, "learning_rate": 8.616090331686662e-05, "loss": 3.0, "step": 17649 }, { "epoch": 1.38, "learning_rate": 8.616011918764213e-05, "loss": 3.0286, "step": 17650 }, { "epoch": 1.38, "learning_rate": 8.615933505841763e-05, "loss": 2.8723, "step": 17651 }, { "epoch": 1.38, "learning_rate": 8.615855092919313e-05, "loss": 2.7558, "step": 17652 }, { "epoch": 1.38, "learning_rate": 8.615776679996864e-05, "loss": 2.8844, "step": 17653 }, { "epoch": 1.38, "learning_rate": 8.615698267074414e-05, "loss": 2.9239, "step": 17654 }, { "epoch": 1.38, "learning_rate": 8.615619854151966e-05, "loss": 3.1032, "step": 17655 }, { "epoch": 1.38, "learning_rate": 8.615541441229516e-05, "loss": 2.8138, "step": 17656 }, { "epoch": 1.38, "learning_rate": 8.615463028307066e-05, "loss": 2.7146, "step": 17657 }, { "epoch": 1.38, "learning_rate": 8.615384615384617e-05, "loss": 3.0171, "step": 17658 }, { "epoch": 1.38, "learning_rate": 8.615306202462167e-05, "loss": 2.8059, "step": 17659 }, { "epoch": 1.38, "learning_rate": 8.615227789539718e-05, "loss": 2.8463, "step": 17660 }, { "epoch": 1.38, "learning_rate": 8.615149376617267e-05, "loss": 3.1546, "step": 17661 }, { "epoch": 1.38, "learning_rate": 8.615070963694817e-05, "loss": 2.9325, "step": 17662 }, { "epoch": 1.39, "learning_rate": 8.614992550772367e-05, "loss": 2.7954, "step": 17663 }, { "epoch": 1.39, "learning_rate": 8.614914137849918e-05, "loss": 2.7699, "step": 17664 }, { "epoch": 1.39, "learning_rate": 8.614835724927468e-05, "loss": 2.898, "step": 17665 }, { "epoch": 1.39, "learning_rate": 8.614757312005018e-05, "loss": 3.1586, "step": 17666 }, { "epoch": 1.39, "learning_rate": 8.614678899082569e-05, "loss": 2.9424, "step": 17667 }, { "epoch": 1.39, "learning_rate": 8.614600486160119e-05, "loss": 2.7666, "step": 17668 }, { "epoch": 1.39, "learning_rate": 8.61452207323767e-05, "loss": 2.8218, "step": 17669 }, { "epoch": 1.39, "learning_rate": 8.61444366031522e-05, "loss": 2.7305, "step": 17670 }, { "epoch": 1.39, "learning_rate": 8.61436524739277e-05, "loss": 2.9302, "step": 17671 }, { "epoch": 1.39, "learning_rate": 8.61428683447032e-05, "loss": 2.7674, "step": 17672 }, { "epoch": 1.39, "learning_rate": 8.614208421547871e-05, "loss": 2.9624, "step": 17673 }, { "epoch": 1.39, "learning_rate": 8.614130008625421e-05, "loss": 2.8808, "step": 17674 }, { "epoch": 1.39, "learning_rate": 8.614051595702972e-05, "loss": 2.8373, "step": 17675 }, { "epoch": 1.39, "learning_rate": 8.613973182780522e-05, "loss": 2.801, "step": 17676 }, { "epoch": 1.39, "learning_rate": 8.613894769858072e-05, "loss": 2.8022, "step": 17677 }, { "epoch": 1.39, "learning_rate": 8.613816356935623e-05, "loss": 3.231, "step": 17678 }, { "epoch": 1.39, "learning_rate": 8.613737944013173e-05, "loss": 2.7664, "step": 17679 }, { "epoch": 1.39, "learning_rate": 8.613659531090724e-05, "loss": 2.9179, "step": 17680 }, { "epoch": 1.39, "learning_rate": 8.613581118168275e-05, "loss": 3.0199, "step": 17681 }, { "epoch": 1.39, "learning_rate": 8.613502705245826e-05, "loss": 2.8394, "step": 17682 }, { "epoch": 1.39, "learning_rate": 8.613424292323376e-05, "loss": 3.1444, "step": 17683 }, { "epoch": 1.39, "learning_rate": 8.613345879400926e-05, "loss": 2.897, "step": 17684 }, { "epoch": 1.39, "learning_rate": 8.613267466478477e-05, "loss": 3.1988, "step": 17685 }, { "epoch": 1.39, "learning_rate": 8.613189053556027e-05, "loss": 2.858, "step": 17686 }, { "epoch": 1.39, "learning_rate": 8.613110640633577e-05, "loss": 2.9477, "step": 17687 }, { "epoch": 1.39, "learning_rate": 8.613032227711128e-05, "loss": 2.9705, "step": 17688 }, { "epoch": 1.39, "learning_rate": 8.612953814788678e-05, "loss": 3.0061, "step": 17689 }, { "epoch": 1.39, "learning_rate": 8.612875401866229e-05, "loss": 3.3786, "step": 17690 }, { "epoch": 1.39, "learning_rate": 8.612796988943779e-05, "loss": 2.7554, "step": 17691 }, { "epoch": 1.39, "learning_rate": 8.61271857602133e-05, "loss": 2.9235, "step": 17692 }, { "epoch": 1.39, "learning_rate": 8.61264016309888e-05, "loss": 2.778, "step": 17693 }, { "epoch": 1.39, "learning_rate": 8.61256175017643e-05, "loss": 2.9127, "step": 17694 }, { "epoch": 1.39, "learning_rate": 8.61248333725398e-05, "loss": 3.0986, "step": 17695 }, { "epoch": 1.39, "learning_rate": 8.612404924331531e-05, "loss": 2.9068, "step": 17696 }, { "epoch": 1.39, "learning_rate": 8.61232651140908e-05, "loss": 3.1556, "step": 17697 }, { "epoch": 1.39, "learning_rate": 8.61224809848663e-05, "loss": 2.8227, "step": 17698 }, { "epoch": 1.39, "learning_rate": 8.61216968556418e-05, "loss": 3.0935, "step": 17699 }, { "epoch": 1.39, "learning_rate": 8.612091272641731e-05, "loss": 2.9729, "step": 17700 }, { "epoch": 1.39, "learning_rate": 8.612012859719281e-05, "loss": 2.8352, "step": 17701 }, { "epoch": 1.39, "learning_rate": 8.611934446796832e-05, "loss": 2.8511, "step": 17702 }, { "epoch": 1.39, "learning_rate": 8.611856033874382e-05, "loss": 2.9584, "step": 17703 }, { "epoch": 1.39, "learning_rate": 8.611777620951932e-05, "loss": 2.6635, "step": 17704 }, { "epoch": 1.39, "learning_rate": 8.611699208029483e-05, "loss": 2.8983, "step": 17705 }, { "epoch": 1.39, "learning_rate": 8.611620795107033e-05, "loss": 3.0577, "step": 17706 }, { "epoch": 1.39, "learning_rate": 8.611542382184585e-05, "loss": 2.9438, "step": 17707 }, { "epoch": 1.39, "learning_rate": 8.611463969262135e-05, "loss": 3.0527, "step": 17708 }, { "epoch": 1.39, "learning_rate": 8.611385556339686e-05, "loss": 2.7679, "step": 17709 }, { "epoch": 1.39, "learning_rate": 8.611307143417236e-05, "loss": 2.8829, "step": 17710 }, { "epoch": 1.39, "learning_rate": 8.611228730494786e-05, "loss": 3.132, "step": 17711 }, { "epoch": 1.39, "learning_rate": 8.611150317572337e-05, "loss": 2.9302, "step": 17712 }, { "epoch": 1.39, "learning_rate": 8.611071904649887e-05, "loss": 2.7706, "step": 17713 }, { "epoch": 1.39, "learning_rate": 8.610993491727437e-05, "loss": 2.873, "step": 17714 }, { "epoch": 1.39, "learning_rate": 8.610915078804988e-05, "loss": 2.8068, "step": 17715 }, { "epoch": 1.39, "learning_rate": 8.610836665882538e-05, "loss": 3.02, "step": 17716 }, { "epoch": 1.39, "learning_rate": 8.610758252960089e-05, "loss": 2.9633, "step": 17717 }, { "epoch": 1.39, "learning_rate": 8.610679840037639e-05, "loss": 3.0801, "step": 17718 }, { "epoch": 1.39, "learning_rate": 8.610601427115189e-05, "loss": 2.7828, "step": 17719 }, { "epoch": 1.39, "learning_rate": 8.61052301419274e-05, "loss": 2.9752, "step": 17720 }, { "epoch": 1.39, "learning_rate": 8.61044460127029e-05, "loss": 2.8301, "step": 17721 }, { "epoch": 1.39, "learning_rate": 8.61036618834784e-05, "loss": 2.7823, "step": 17722 }, { "epoch": 1.39, "learning_rate": 8.610287775425391e-05, "loss": 2.8369, "step": 17723 }, { "epoch": 1.39, "learning_rate": 8.610209362502941e-05, "loss": 2.7766, "step": 17724 }, { "epoch": 1.39, "learning_rate": 8.610130949580491e-05, "loss": 2.8881, "step": 17725 }, { "epoch": 1.39, "learning_rate": 8.610052536658042e-05, "loss": 2.816, "step": 17726 }, { "epoch": 1.39, "learning_rate": 8.609974123735592e-05, "loss": 3.076, "step": 17727 }, { "epoch": 1.39, "learning_rate": 8.609895710813143e-05, "loss": 3.1755, "step": 17728 }, { "epoch": 1.39, "learning_rate": 8.609817297890693e-05, "loss": 2.6665, "step": 17729 }, { "epoch": 1.39, "learning_rate": 8.609738884968243e-05, "loss": 3.1503, "step": 17730 }, { "epoch": 1.39, "learning_rate": 8.609660472045794e-05, "loss": 2.9274, "step": 17731 }, { "epoch": 1.39, "learning_rate": 8.609582059123344e-05, "loss": 2.8634, "step": 17732 }, { "epoch": 1.39, "learning_rate": 8.609503646200894e-05, "loss": 2.9784, "step": 17733 }, { "epoch": 1.39, "learning_rate": 8.609425233278445e-05, "loss": 2.8658, "step": 17734 }, { "epoch": 1.39, "learning_rate": 8.609346820355995e-05, "loss": 2.8155, "step": 17735 }, { "epoch": 1.39, "learning_rate": 8.609268407433546e-05, "loss": 3.0344, "step": 17736 }, { "epoch": 1.39, "learning_rate": 8.609189994511096e-05, "loss": 2.7497, "step": 17737 }, { "epoch": 1.39, "learning_rate": 8.609111581588646e-05, "loss": 2.9742, "step": 17738 }, { "epoch": 1.39, "learning_rate": 8.609033168666197e-05, "loss": 3.0744, "step": 17739 }, { "epoch": 1.39, "learning_rate": 8.608954755743747e-05, "loss": 2.831, "step": 17740 }, { "epoch": 1.39, "learning_rate": 8.608876342821297e-05, "loss": 2.812, "step": 17741 }, { "epoch": 1.39, "learning_rate": 8.608797929898848e-05, "loss": 2.9273, "step": 17742 }, { "epoch": 1.39, "learning_rate": 8.608719516976398e-05, "loss": 2.9669, "step": 17743 }, { "epoch": 1.39, "learning_rate": 8.608641104053948e-05, "loss": 2.6773, "step": 17744 }, { "epoch": 1.39, "learning_rate": 8.608562691131499e-05, "loss": 2.9818, "step": 17745 }, { "epoch": 1.39, "learning_rate": 8.608484278209049e-05, "loss": 2.8053, "step": 17746 }, { "epoch": 1.39, "learning_rate": 8.6084058652866e-05, "loss": 2.9977, "step": 17747 }, { "epoch": 1.39, "learning_rate": 8.60832745236415e-05, "loss": 2.7615, "step": 17748 }, { "epoch": 1.39, "learning_rate": 8.6082490394417e-05, "loss": 2.8085, "step": 17749 }, { "epoch": 1.39, "learning_rate": 8.60817062651925e-05, "loss": 2.9538, "step": 17750 }, { "epoch": 1.39, "learning_rate": 8.608092213596801e-05, "loss": 2.9488, "step": 17751 }, { "epoch": 1.39, "learning_rate": 8.608013800674351e-05, "loss": 3.0633, "step": 17752 }, { "epoch": 1.39, "learning_rate": 8.607935387751902e-05, "loss": 3.1404, "step": 17753 }, { "epoch": 1.39, "learning_rate": 8.607856974829452e-05, "loss": 2.5525, "step": 17754 }, { "epoch": 1.39, "learning_rate": 8.607778561907002e-05, "loss": 3.0026, "step": 17755 }, { "epoch": 1.39, "learning_rate": 8.607700148984553e-05, "loss": 2.8758, "step": 17756 }, { "epoch": 1.39, "learning_rate": 8.607621736062103e-05, "loss": 2.7409, "step": 17757 }, { "epoch": 1.39, "learning_rate": 8.607543323139654e-05, "loss": 2.8939, "step": 17758 }, { "epoch": 1.39, "learning_rate": 8.607464910217204e-05, "loss": 3.1576, "step": 17759 }, { "epoch": 1.39, "learning_rate": 8.607386497294754e-05, "loss": 2.8549, "step": 17760 }, { "epoch": 1.39, "learning_rate": 8.607308084372305e-05, "loss": 2.8713, "step": 17761 }, { "epoch": 1.39, "learning_rate": 8.607229671449855e-05, "loss": 2.9166, "step": 17762 }, { "epoch": 1.39, "learning_rate": 8.607151258527407e-05, "loss": 3.1898, "step": 17763 }, { "epoch": 1.39, "learning_rate": 8.607072845604957e-05, "loss": 2.7909, "step": 17764 }, { "epoch": 1.39, "learning_rate": 8.606994432682508e-05, "loss": 2.9117, "step": 17765 }, { "epoch": 1.39, "learning_rate": 8.606916019760058e-05, "loss": 3.2133, "step": 17766 }, { "epoch": 1.39, "learning_rate": 8.606837606837608e-05, "loss": 3.0511, "step": 17767 }, { "epoch": 1.39, "learning_rate": 8.606759193915159e-05, "loss": 2.7405, "step": 17768 }, { "epoch": 1.39, "learning_rate": 8.606680780992708e-05, "loss": 2.8696, "step": 17769 }, { "epoch": 1.39, "learning_rate": 8.606602368070258e-05, "loss": 2.8603, "step": 17770 }, { "epoch": 1.39, "learning_rate": 8.606523955147808e-05, "loss": 3.1204, "step": 17771 }, { "epoch": 1.39, "learning_rate": 8.606445542225359e-05, "loss": 2.7224, "step": 17772 }, { "epoch": 1.39, "learning_rate": 8.606367129302909e-05, "loss": 2.7428, "step": 17773 }, { "epoch": 1.39, "learning_rate": 8.60628871638046e-05, "loss": 3.1005, "step": 17774 }, { "epoch": 1.39, "learning_rate": 8.60621030345801e-05, "loss": 3.1273, "step": 17775 }, { "epoch": 1.39, "learning_rate": 8.60613189053556e-05, "loss": 2.9604, "step": 17776 }, { "epoch": 1.39, "learning_rate": 8.60605347761311e-05, "loss": 3.0575, "step": 17777 }, { "epoch": 1.39, "learning_rate": 8.605975064690661e-05, "loss": 2.776, "step": 17778 }, { "epoch": 1.39, "learning_rate": 8.605896651768211e-05, "loss": 2.925, "step": 17779 }, { "epoch": 1.39, "learning_rate": 8.605818238845762e-05, "loss": 2.9761, "step": 17780 }, { "epoch": 1.39, "learning_rate": 8.605739825923312e-05, "loss": 2.9719, "step": 17781 }, { "epoch": 1.39, "learning_rate": 8.605661413000862e-05, "loss": 2.9695, "step": 17782 }, { "epoch": 1.39, "learning_rate": 8.605583000078413e-05, "loss": 3.1428, "step": 17783 }, { "epoch": 1.39, "learning_rate": 8.605504587155963e-05, "loss": 2.8535, "step": 17784 }, { "epoch": 1.39, "learning_rate": 8.605426174233514e-05, "loss": 2.7522, "step": 17785 }, { "epoch": 1.39, "learning_rate": 8.605347761311064e-05, "loss": 2.8521, "step": 17786 }, { "epoch": 1.39, "learning_rate": 8.605269348388614e-05, "loss": 2.8655, "step": 17787 }, { "epoch": 1.39, "learning_rate": 8.605190935466165e-05, "loss": 3.2123, "step": 17788 }, { "epoch": 1.39, "learning_rate": 8.605112522543716e-05, "loss": 3.3361, "step": 17789 }, { "epoch": 1.39, "learning_rate": 8.605034109621267e-05, "loss": 2.8997, "step": 17790 }, { "epoch": 1.4, "learning_rate": 8.604955696698817e-05, "loss": 3.0499, "step": 17791 }, { "epoch": 1.4, "learning_rate": 8.604877283776367e-05, "loss": 2.8006, "step": 17792 }, { "epoch": 1.4, "learning_rate": 8.604798870853918e-05, "loss": 2.8302, "step": 17793 }, { "epoch": 1.4, "learning_rate": 8.604720457931468e-05, "loss": 2.7869, "step": 17794 }, { "epoch": 1.4, "learning_rate": 8.604642045009019e-05, "loss": 2.9188, "step": 17795 }, { "epoch": 1.4, "learning_rate": 8.604563632086569e-05, "loss": 2.8597, "step": 17796 }, { "epoch": 1.4, "learning_rate": 8.604485219164119e-05, "loss": 3.0646, "step": 17797 }, { "epoch": 1.4, "learning_rate": 8.60440680624167e-05, "loss": 2.8003, "step": 17798 }, { "epoch": 1.4, "learning_rate": 8.60432839331922e-05, "loss": 3.0406, "step": 17799 }, { "epoch": 1.4, "learning_rate": 8.60424998039677e-05, "loss": 2.8356, "step": 17800 }, { "epoch": 1.4, "learning_rate": 8.604171567474321e-05, "loss": 2.7636, "step": 17801 }, { "epoch": 1.4, "learning_rate": 8.604093154551871e-05, "loss": 2.7788, "step": 17802 }, { "epoch": 1.4, "learning_rate": 8.604014741629422e-05, "loss": 2.9445, "step": 17803 }, { "epoch": 1.4, "learning_rate": 8.603936328706972e-05, "loss": 2.9832, "step": 17804 }, { "epoch": 1.4, "learning_rate": 8.603857915784521e-05, "loss": 3.0836, "step": 17805 }, { "epoch": 1.4, "learning_rate": 8.603779502862071e-05, "loss": 3.0836, "step": 17806 }, { "epoch": 1.4, "learning_rate": 8.603701089939622e-05, "loss": 3.0432, "step": 17807 }, { "epoch": 1.4, "learning_rate": 8.603622677017172e-05, "loss": 2.7229, "step": 17808 }, { "epoch": 1.4, "learning_rate": 8.603544264094722e-05, "loss": 3.0494, "step": 17809 }, { "epoch": 1.4, "learning_rate": 8.603465851172273e-05, "loss": 3.0387, "step": 17810 }, { "epoch": 1.4, "learning_rate": 8.603387438249823e-05, "loss": 2.6239, "step": 17811 }, { "epoch": 1.4, "learning_rate": 8.603309025327373e-05, "loss": 2.7268, "step": 17812 }, { "epoch": 1.4, "learning_rate": 8.603230612404924e-05, "loss": 2.6442, "step": 17813 }, { "epoch": 1.4, "learning_rate": 8.603152199482474e-05, "loss": 3.0092, "step": 17814 }, { "epoch": 1.4, "learning_rate": 8.603073786560026e-05, "loss": 2.9846, "step": 17815 }, { "epoch": 1.4, "learning_rate": 8.602995373637576e-05, "loss": 2.9007, "step": 17816 }, { "epoch": 1.4, "learning_rate": 8.602916960715127e-05, "loss": 3.0112, "step": 17817 }, { "epoch": 1.4, "learning_rate": 8.602838547792677e-05, "loss": 2.997, "step": 17818 }, { "epoch": 1.4, "learning_rate": 8.602760134870227e-05, "loss": 2.8849, "step": 17819 }, { "epoch": 1.4, "learning_rate": 8.602681721947778e-05, "loss": 2.9642, "step": 17820 }, { "epoch": 1.4, "learning_rate": 8.602603309025328e-05, "loss": 3.2219, "step": 17821 }, { "epoch": 1.4, "learning_rate": 8.602524896102878e-05, "loss": 2.8704, "step": 17822 }, { "epoch": 1.4, "learning_rate": 8.602446483180429e-05, "loss": 2.8803, "step": 17823 }, { "epoch": 1.4, "learning_rate": 8.602368070257979e-05, "loss": 2.8336, "step": 17824 }, { "epoch": 1.4, "learning_rate": 8.60228965733553e-05, "loss": 2.6943, "step": 17825 }, { "epoch": 1.4, "learning_rate": 8.60221124441308e-05, "loss": 2.9462, "step": 17826 }, { "epoch": 1.4, "learning_rate": 8.60213283149063e-05, "loss": 2.864, "step": 17827 }, { "epoch": 1.4, "learning_rate": 8.602054418568181e-05, "loss": 3.1413, "step": 17828 }, { "epoch": 1.4, "learning_rate": 8.601976005645731e-05, "loss": 2.9178, "step": 17829 }, { "epoch": 1.4, "learning_rate": 8.601897592723281e-05, "loss": 2.7675, "step": 17830 }, { "epoch": 1.4, "learning_rate": 8.601819179800832e-05, "loss": 3.0117, "step": 17831 }, { "epoch": 1.4, "learning_rate": 8.601740766878382e-05, "loss": 3.0572, "step": 17832 }, { "epoch": 1.4, "learning_rate": 8.601662353955933e-05, "loss": 3.0679, "step": 17833 }, { "epoch": 1.4, "learning_rate": 8.601583941033483e-05, "loss": 2.7359, "step": 17834 }, { "epoch": 1.4, "learning_rate": 8.601505528111033e-05, "loss": 2.9893, "step": 17835 }, { "epoch": 1.4, "learning_rate": 8.601427115188584e-05, "loss": 2.6317, "step": 17836 }, { "epoch": 1.4, "learning_rate": 8.601348702266134e-05, "loss": 2.9718, "step": 17837 }, { "epoch": 1.4, "learning_rate": 8.601270289343684e-05, "loss": 2.7852, "step": 17838 }, { "epoch": 1.4, "learning_rate": 8.601191876421235e-05, "loss": 2.9878, "step": 17839 }, { "epoch": 1.4, "learning_rate": 8.601113463498785e-05, "loss": 2.8602, "step": 17840 }, { "epoch": 1.4, "learning_rate": 8.601035050576335e-05, "loss": 3.0074, "step": 17841 }, { "epoch": 1.4, "learning_rate": 8.600956637653886e-05, "loss": 2.931, "step": 17842 }, { "epoch": 1.4, "learning_rate": 8.600878224731436e-05, "loss": 2.7775, "step": 17843 }, { "epoch": 1.4, "learning_rate": 8.600799811808987e-05, "loss": 2.6721, "step": 17844 }, { "epoch": 1.4, "learning_rate": 8.600721398886537e-05, "loss": 2.8058, "step": 17845 }, { "epoch": 1.4, "learning_rate": 8.600642985964087e-05, "loss": 2.7917, "step": 17846 }, { "epoch": 1.4, "learning_rate": 8.600564573041638e-05, "loss": 2.7076, "step": 17847 }, { "epoch": 1.4, "learning_rate": 8.600486160119188e-05, "loss": 3.0498, "step": 17848 }, { "epoch": 1.4, "learning_rate": 8.600407747196738e-05, "loss": 2.9213, "step": 17849 }, { "epoch": 1.4, "learning_rate": 8.600329334274289e-05, "loss": 2.8389, "step": 17850 }, { "epoch": 1.4, "learning_rate": 8.600250921351839e-05, "loss": 3.0372, "step": 17851 }, { "epoch": 1.4, "learning_rate": 8.60017250842939e-05, "loss": 2.774, "step": 17852 }, { "epoch": 1.4, "learning_rate": 8.60009409550694e-05, "loss": 2.8975, "step": 17853 }, { "epoch": 1.4, "learning_rate": 8.60001568258449e-05, "loss": 3.2151, "step": 17854 }, { "epoch": 1.4, "learning_rate": 8.59993726966204e-05, "loss": 2.8649, "step": 17855 }, { "epoch": 1.4, "learning_rate": 8.599858856739591e-05, "loss": 2.8787, "step": 17856 }, { "epoch": 1.4, "learning_rate": 8.599780443817141e-05, "loss": 2.841, "step": 17857 }, { "epoch": 1.4, "learning_rate": 8.599702030894692e-05, "loss": 2.8972, "step": 17858 }, { "epoch": 1.4, "learning_rate": 8.599623617972242e-05, "loss": 2.7772, "step": 17859 }, { "epoch": 1.4, "learning_rate": 8.599545205049792e-05, "loss": 2.773, "step": 17860 }, { "epoch": 1.4, "learning_rate": 8.599466792127343e-05, "loss": 2.91, "step": 17861 }, { "epoch": 1.4, "learning_rate": 8.599388379204893e-05, "loss": 2.7047, "step": 17862 }, { "epoch": 1.4, "learning_rate": 8.599309966282444e-05, "loss": 2.8668, "step": 17863 }, { "epoch": 1.4, "learning_rate": 8.599231553359994e-05, "loss": 2.8957, "step": 17864 }, { "epoch": 1.4, "learning_rate": 8.599153140437544e-05, "loss": 2.7824, "step": 17865 }, { "epoch": 1.4, "learning_rate": 8.599074727515095e-05, "loss": 2.7751, "step": 17866 }, { "epoch": 1.4, "learning_rate": 8.598996314592645e-05, "loss": 2.7544, "step": 17867 }, { "epoch": 1.4, "learning_rate": 8.598917901670195e-05, "loss": 3.0313, "step": 17868 }, { "epoch": 1.4, "learning_rate": 8.598839488747746e-05, "loss": 2.8671, "step": 17869 }, { "epoch": 1.4, "learning_rate": 8.598761075825296e-05, "loss": 2.6425, "step": 17870 }, { "epoch": 1.4, "learning_rate": 8.598682662902847e-05, "loss": 3.113, "step": 17871 }, { "epoch": 1.4, "learning_rate": 8.598604249980398e-05, "loss": 2.9452, "step": 17872 }, { "epoch": 1.4, "learning_rate": 8.598525837057949e-05, "loss": 2.7859, "step": 17873 }, { "epoch": 1.4, "learning_rate": 8.598447424135499e-05, "loss": 2.8344, "step": 17874 }, { "epoch": 1.4, "learning_rate": 8.59836901121305e-05, "loss": 2.8624, "step": 17875 }, { "epoch": 1.4, "learning_rate": 8.5982905982906e-05, "loss": 2.8949, "step": 17876 }, { "epoch": 1.4, "learning_rate": 8.598212185368149e-05, "loss": 2.9822, "step": 17877 }, { "epoch": 1.4, "learning_rate": 8.598133772445699e-05, "loss": 2.6838, "step": 17878 }, { "epoch": 1.4, "learning_rate": 8.59805535952325e-05, "loss": 2.6813, "step": 17879 }, { "epoch": 1.4, "learning_rate": 8.5979769466008e-05, "loss": 2.8605, "step": 17880 }, { "epoch": 1.4, "learning_rate": 8.59789853367835e-05, "loss": 3.0824, "step": 17881 }, { "epoch": 1.4, "learning_rate": 8.5978201207559e-05, "loss": 2.8358, "step": 17882 }, { "epoch": 1.4, "learning_rate": 8.597741707833451e-05, "loss": 2.994, "step": 17883 }, { "epoch": 1.4, "learning_rate": 8.597663294911001e-05, "loss": 2.9842, "step": 17884 }, { "epoch": 1.4, "learning_rate": 8.597584881988552e-05, "loss": 3.1592, "step": 17885 }, { "epoch": 1.4, "learning_rate": 8.597506469066102e-05, "loss": 2.8975, "step": 17886 }, { "epoch": 1.4, "learning_rate": 8.597428056143652e-05, "loss": 2.9616, "step": 17887 }, { "epoch": 1.4, "learning_rate": 8.597349643221203e-05, "loss": 2.8347, "step": 17888 }, { "epoch": 1.4, "learning_rate": 8.597271230298753e-05, "loss": 2.7778, "step": 17889 }, { "epoch": 1.4, "learning_rate": 8.597192817376303e-05, "loss": 2.8423, "step": 17890 }, { "epoch": 1.4, "learning_rate": 8.597114404453854e-05, "loss": 2.8683, "step": 17891 }, { "epoch": 1.4, "learning_rate": 8.597035991531404e-05, "loss": 2.6699, "step": 17892 }, { "epoch": 1.4, "learning_rate": 8.596957578608955e-05, "loss": 2.872, "step": 17893 }, { "epoch": 1.4, "learning_rate": 8.596879165686505e-05, "loss": 3.1405, "step": 17894 }, { "epoch": 1.4, "learning_rate": 8.596800752764055e-05, "loss": 3.0081, "step": 17895 }, { "epoch": 1.4, "learning_rate": 8.596722339841606e-05, "loss": 3.1039, "step": 17896 }, { "epoch": 1.4, "learning_rate": 8.596643926919157e-05, "loss": 3.1319, "step": 17897 }, { "epoch": 1.4, "learning_rate": 8.596565513996708e-05, "loss": 3.1476, "step": 17898 }, { "epoch": 1.4, "learning_rate": 8.596487101074258e-05, "loss": 2.7802, "step": 17899 }, { "epoch": 1.4, "learning_rate": 8.596408688151809e-05, "loss": 2.9468, "step": 17900 }, { "epoch": 1.4, "learning_rate": 8.596330275229359e-05, "loss": 3.0294, "step": 17901 }, { "epoch": 1.4, "learning_rate": 8.596251862306909e-05, "loss": 2.8028, "step": 17902 }, { "epoch": 1.4, "learning_rate": 8.59617344938446e-05, "loss": 2.7345, "step": 17903 }, { "epoch": 1.4, "learning_rate": 8.59609503646201e-05, "loss": 2.8682, "step": 17904 }, { "epoch": 1.4, "learning_rate": 8.59601662353956e-05, "loss": 2.9502, "step": 17905 }, { "epoch": 1.4, "learning_rate": 8.595938210617111e-05, "loss": 2.8265, "step": 17906 }, { "epoch": 1.4, "learning_rate": 8.595859797694661e-05, "loss": 2.9187, "step": 17907 }, { "epoch": 1.4, "learning_rate": 8.595781384772211e-05, "loss": 3.0175, "step": 17908 }, { "epoch": 1.4, "learning_rate": 8.595702971849762e-05, "loss": 2.8732, "step": 17909 }, { "epoch": 1.4, "learning_rate": 8.595624558927312e-05, "loss": 2.8843, "step": 17910 }, { "epoch": 1.4, "learning_rate": 8.595546146004863e-05, "loss": 2.8877, "step": 17911 }, { "epoch": 1.4, "learning_rate": 8.595467733082413e-05, "loss": 2.9109, "step": 17912 }, { "epoch": 1.4, "learning_rate": 8.595389320159962e-05, "loss": 2.6705, "step": 17913 }, { "epoch": 1.4, "learning_rate": 8.595310907237512e-05, "loss": 2.8748, "step": 17914 }, { "epoch": 1.4, "learning_rate": 8.595232494315063e-05, "loss": 2.7039, "step": 17915 }, { "epoch": 1.4, "learning_rate": 8.595154081392613e-05, "loss": 2.8259, "step": 17916 }, { "epoch": 1.4, "learning_rate": 8.595075668470163e-05, "loss": 2.9412, "step": 17917 }, { "epoch": 1.41, "learning_rate": 8.594997255547714e-05, "loss": 2.7999, "step": 17918 }, { "epoch": 1.41, "learning_rate": 8.594918842625264e-05, "loss": 3.0772, "step": 17919 }, { "epoch": 1.41, "learning_rate": 8.594840429702815e-05, "loss": 3.2096, "step": 17920 }, { "epoch": 1.41, "learning_rate": 8.594762016780365e-05, "loss": 2.8326, "step": 17921 }, { "epoch": 1.41, "learning_rate": 8.594683603857915e-05, "loss": 2.9184, "step": 17922 }, { "epoch": 1.41, "learning_rate": 8.594605190935467e-05, "loss": 2.9985, "step": 17923 }, { "epoch": 1.41, "learning_rate": 8.594526778013017e-05, "loss": 3.0133, "step": 17924 }, { "epoch": 1.41, "learning_rate": 8.594448365090568e-05, "loss": 2.7296, "step": 17925 }, { "epoch": 1.41, "learning_rate": 8.594369952168118e-05, "loss": 3.0045, "step": 17926 }, { "epoch": 1.41, "learning_rate": 8.594291539245668e-05, "loss": 3.1509, "step": 17927 }, { "epoch": 1.41, "learning_rate": 8.594213126323219e-05, "loss": 2.8222, "step": 17928 }, { "epoch": 1.41, "learning_rate": 8.594134713400769e-05, "loss": 2.8442, "step": 17929 }, { "epoch": 1.41, "learning_rate": 8.59405630047832e-05, "loss": 2.754, "step": 17930 }, { "epoch": 1.41, "learning_rate": 8.59397788755587e-05, "loss": 3.0173, "step": 17931 }, { "epoch": 1.41, "learning_rate": 8.59389947463342e-05, "loss": 2.911, "step": 17932 }, { "epoch": 1.41, "learning_rate": 8.59382106171097e-05, "loss": 3.1585, "step": 17933 }, { "epoch": 1.41, "learning_rate": 8.593742648788521e-05, "loss": 2.8645, "step": 17934 }, { "epoch": 1.41, "learning_rate": 8.593664235866071e-05, "loss": 2.9653, "step": 17935 }, { "epoch": 1.41, "learning_rate": 8.593585822943622e-05, "loss": 3.0018, "step": 17936 }, { "epoch": 1.41, "learning_rate": 8.593507410021172e-05, "loss": 2.6942, "step": 17937 }, { "epoch": 1.41, "learning_rate": 8.593428997098723e-05, "loss": 3.2966, "step": 17938 }, { "epoch": 1.41, "learning_rate": 8.593350584176273e-05, "loss": 2.7257, "step": 17939 }, { "epoch": 1.41, "learning_rate": 8.593272171253823e-05, "loss": 2.9937, "step": 17940 }, { "epoch": 1.41, "learning_rate": 8.593193758331374e-05, "loss": 2.8163, "step": 17941 }, { "epoch": 1.41, "learning_rate": 8.593115345408924e-05, "loss": 2.6441, "step": 17942 }, { "epoch": 1.41, "learning_rate": 8.593036932486474e-05, "loss": 3.072, "step": 17943 }, { "epoch": 1.41, "learning_rate": 8.592958519564025e-05, "loss": 2.904, "step": 17944 }, { "epoch": 1.41, "learning_rate": 8.592880106641575e-05, "loss": 2.7075, "step": 17945 }, { "epoch": 1.41, "learning_rate": 8.592801693719125e-05, "loss": 2.75, "step": 17946 }, { "epoch": 1.41, "learning_rate": 8.592723280796676e-05, "loss": 2.9535, "step": 17947 }, { "epoch": 1.41, "learning_rate": 8.592644867874226e-05, "loss": 2.9626, "step": 17948 }, { "epoch": 1.41, "learning_rate": 8.592566454951777e-05, "loss": 2.8501, "step": 17949 }, { "epoch": 1.41, "learning_rate": 8.592488042029327e-05, "loss": 2.899, "step": 17950 }, { "epoch": 1.41, "learning_rate": 8.592409629106877e-05, "loss": 3.0531, "step": 17951 }, { "epoch": 1.41, "learning_rate": 8.592331216184428e-05, "loss": 3.2024, "step": 17952 }, { "epoch": 1.41, "learning_rate": 8.592252803261978e-05, "loss": 2.8548, "step": 17953 }, { "epoch": 1.41, "learning_rate": 8.592174390339528e-05, "loss": 2.818, "step": 17954 }, { "epoch": 1.41, "learning_rate": 8.592095977417079e-05, "loss": 2.8977, "step": 17955 }, { "epoch": 1.41, "learning_rate": 8.592017564494629e-05, "loss": 2.711, "step": 17956 }, { "epoch": 1.41, "learning_rate": 8.59193915157218e-05, "loss": 2.8356, "step": 17957 }, { "epoch": 1.41, "learning_rate": 8.59186073864973e-05, "loss": 2.6899, "step": 17958 }, { "epoch": 1.41, "learning_rate": 8.59178232572728e-05, "loss": 2.8458, "step": 17959 }, { "epoch": 1.41, "learning_rate": 8.59170391280483e-05, "loss": 2.6891, "step": 17960 }, { "epoch": 1.41, "learning_rate": 8.591625499882381e-05, "loss": 2.9252, "step": 17961 }, { "epoch": 1.41, "learning_rate": 8.591547086959931e-05, "loss": 2.7552, "step": 17962 }, { "epoch": 1.41, "learning_rate": 8.591468674037482e-05, "loss": 3.2996, "step": 17963 }, { "epoch": 1.41, "learning_rate": 8.591390261115032e-05, "loss": 2.9947, "step": 17964 }, { "epoch": 1.41, "learning_rate": 8.591311848192582e-05, "loss": 2.907, "step": 17965 }, { "epoch": 1.41, "learning_rate": 8.591233435270133e-05, "loss": 2.7895, "step": 17966 }, { "epoch": 1.41, "learning_rate": 8.591155022347683e-05, "loss": 3.1014, "step": 17967 }, { "epoch": 1.41, "learning_rate": 8.591076609425234e-05, "loss": 2.9416, "step": 17968 }, { "epoch": 1.41, "learning_rate": 8.590998196502784e-05, "loss": 3.0766, "step": 17969 }, { "epoch": 1.41, "learning_rate": 8.590919783580334e-05, "loss": 2.8526, "step": 17970 }, { "epoch": 1.41, "learning_rate": 8.590841370657885e-05, "loss": 3.0233, "step": 17971 }, { "epoch": 1.41, "learning_rate": 8.590762957735435e-05, "loss": 2.8179, "step": 17972 }, { "epoch": 1.41, "learning_rate": 8.590684544812985e-05, "loss": 2.8008, "step": 17973 }, { "epoch": 1.41, "learning_rate": 8.590606131890536e-05, "loss": 2.8802, "step": 17974 }, { "epoch": 1.41, "learning_rate": 8.590527718968086e-05, "loss": 2.8159, "step": 17975 }, { "epoch": 1.41, "learning_rate": 8.590449306045636e-05, "loss": 2.9944, "step": 17976 }, { "epoch": 1.41, "learning_rate": 8.590370893123187e-05, "loss": 2.877, "step": 17977 }, { "epoch": 1.41, "learning_rate": 8.590292480200737e-05, "loss": 2.9534, "step": 17978 }, { "epoch": 1.41, "learning_rate": 8.590214067278288e-05, "loss": 3.0212, "step": 17979 }, { "epoch": 1.41, "learning_rate": 8.590135654355839e-05, "loss": 2.7701, "step": 17980 }, { "epoch": 1.41, "learning_rate": 8.59005724143339e-05, "loss": 2.9349, "step": 17981 }, { "epoch": 1.41, "learning_rate": 8.58997882851094e-05, "loss": 2.8934, "step": 17982 }, { "epoch": 1.41, "learning_rate": 8.58990041558849e-05, "loss": 2.8586, "step": 17983 }, { "epoch": 1.41, "learning_rate": 8.589822002666041e-05, "loss": 2.905, "step": 17984 }, { "epoch": 1.41, "learning_rate": 8.58974358974359e-05, "loss": 2.8539, "step": 17985 }, { "epoch": 1.41, "learning_rate": 8.58966517682114e-05, "loss": 2.9093, "step": 17986 }, { "epoch": 1.41, "learning_rate": 8.58958676389869e-05, "loss": 3.1037, "step": 17987 }, { "epoch": 1.41, "learning_rate": 8.589508350976241e-05, "loss": 2.9568, "step": 17988 }, { "epoch": 1.41, "learning_rate": 8.589429938053791e-05, "loss": 3.1124, "step": 17989 }, { "epoch": 1.41, "learning_rate": 8.589351525131342e-05, "loss": 2.7707, "step": 17990 }, { "epoch": 1.41, "learning_rate": 8.589273112208892e-05, "loss": 2.9566, "step": 17991 }, { "epoch": 1.41, "learning_rate": 8.589194699286442e-05, "loss": 2.8594, "step": 17992 }, { "epoch": 1.41, "learning_rate": 8.589116286363993e-05, "loss": 3.0536, "step": 17993 }, { "epoch": 1.41, "learning_rate": 8.589037873441543e-05, "loss": 2.8295, "step": 17994 }, { "epoch": 1.41, "learning_rate": 8.588959460519093e-05, "loss": 2.8664, "step": 17995 }, { "epoch": 1.41, "learning_rate": 8.588881047596644e-05, "loss": 3.2844, "step": 17996 }, { "epoch": 1.41, "learning_rate": 8.588802634674194e-05, "loss": 2.9105, "step": 17997 }, { "epoch": 1.41, "learning_rate": 8.588724221751745e-05, "loss": 2.6484, "step": 17998 }, { "epoch": 1.41, "learning_rate": 8.588645808829295e-05, "loss": 2.8854, "step": 17999 }, { "epoch": 1.41, "learning_rate": 8.588567395906845e-05, "loss": 2.9913, "step": 18000 }, { "epoch": 1.41, "eval_loss": 2.9159839153289795, "eval_runtime": 218.0193, "eval_samples_per_second": 51.977, "eval_steps_per_second": 3.252, "step": 18000 }, { "epoch": 1.41, "learning_rate": 8.588488982984396e-05, "loss": 2.9089, "step": 18001 }, { "epoch": 1.41, "learning_rate": 8.588410570061946e-05, "loss": 3.1122, "step": 18002 }, { "epoch": 1.41, "learning_rate": 8.588332157139496e-05, "loss": 2.7752, "step": 18003 }, { "epoch": 1.41, "learning_rate": 8.588253744217047e-05, "loss": 2.7813, "step": 18004 }, { "epoch": 1.41, "learning_rate": 8.588175331294599e-05, "loss": 2.8541, "step": 18005 }, { "epoch": 1.41, "learning_rate": 8.588096918372149e-05, "loss": 2.765, "step": 18006 }, { "epoch": 1.41, "learning_rate": 8.588018505449699e-05, "loss": 2.9346, "step": 18007 }, { "epoch": 1.41, "learning_rate": 8.58794009252725e-05, "loss": 2.7557, "step": 18008 }, { "epoch": 1.41, "learning_rate": 8.5878616796048e-05, "loss": 2.8897, "step": 18009 }, { "epoch": 1.41, "learning_rate": 8.58778326668235e-05, "loss": 2.7749, "step": 18010 }, { "epoch": 1.41, "learning_rate": 8.587704853759901e-05, "loss": 2.8178, "step": 18011 }, { "epoch": 1.41, "learning_rate": 8.587626440837451e-05, "loss": 3.0978, "step": 18012 }, { "epoch": 1.41, "learning_rate": 8.587548027915001e-05, "loss": 3.0098, "step": 18013 }, { "epoch": 1.41, "learning_rate": 8.587469614992552e-05, "loss": 3.0143, "step": 18014 }, { "epoch": 1.41, "learning_rate": 8.587391202070102e-05, "loss": 2.5949, "step": 18015 }, { "epoch": 1.41, "learning_rate": 8.587312789147653e-05, "loss": 3.3996, "step": 18016 }, { "epoch": 1.41, "learning_rate": 8.587234376225203e-05, "loss": 3.0069, "step": 18017 }, { "epoch": 1.41, "learning_rate": 8.587155963302753e-05, "loss": 2.8874, "step": 18018 }, { "epoch": 1.41, "learning_rate": 8.587077550380304e-05, "loss": 2.7941, "step": 18019 }, { "epoch": 1.41, "learning_rate": 8.586999137457854e-05, "loss": 2.6201, "step": 18020 }, { "epoch": 1.41, "learning_rate": 8.586920724535403e-05, "loss": 2.9275, "step": 18021 }, { "epoch": 1.41, "learning_rate": 8.586842311612953e-05, "loss": 2.8689, "step": 18022 }, { "epoch": 1.41, "learning_rate": 8.586763898690504e-05, "loss": 2.7597, "step": 18023 }, { "epoch": 1.41, "learning_rate": 8.586685485768054e-05, "loss": 2.7891, "step": 18024 }, { "epoch": 1.41, "learning_rate": 8.586607072845605e-05, "loss": 2.942, "step": 18025 }, { "epoch": 1.41, "learning_rate": 8.586528659923155e-05, "loss": 2.8785, "step": 18026 }, { "epoch": 1.41, "learning_rate": 8.586450247000705e-05, "loss": 3.1555, "step": 18027 }, { "epoch": 1.41, "learning_rate": 8.586371834078256e-05, "loss": 2.9143, "step": 18028 }, { "epoch": 1.41, "learning_rate": 8.586293421155806e-05, "loss": 2.7844, "step": 18029 }, { "epoch": 1.41, "learning_rate": 8.586215008233356e-05, "loss": 2.7232, "step": 18030 }, { "epoch": 1.41, "learning_rate": 8.586136595310908e-05, "loss": 2.8881, "step": 18031 }, { "epoch": 1.41, "learning_rate": 8.586058182388458e-05, "loss": 2.8885, "step": 18032 }, { "epoch": 1.41, "learning_rate": 8.585979769466009e-05, "loss": 2.9589, "step": 18033 }, { "epoch": 1.41, "learning_rate": 8.585901356543559e-05, "loss": 2.8683, "step": 18034 }, { "epoch": 1.41, "learning_rate": 8.58582294362111e-05, "loss": 3.1869, "step": 18035 }, { "epoch": 1.41, "learning_rate": 8.58574453069866e-05, "loss": 2.7794, "step": 18036 }, { "epoch": 1.41, "learning_rate": 8.58566611777621e-05, "loss": 2.7536, "step": 18037 }, { "epoch": 1.41, "learning_rate": 8.58558770485376e-05, "loss": 2.764, "step": 18038 }, { "epoch": 1.41, "learning_rate": 8.585509291931311e-05, "loss": 2.8261, "step": 18039 }, { "epoch": 1.41, "learning_rate": 8.585430879008861e-05, "loss": 2.7676, "step": 18040 }, { "epoch": 1.41, "learning_rate": 8.585352466086412e-05, "loss": 2.8972, "step": 18041 }, { "epoch": 1.41, "learning_rate": 8.585274053163962e-05, "loss": 2.7365, "step": 18042 }, { "epoch": 1.41, "learning_rate": 8.585195640241512e-05, "loss": 2.7439, "step": 18043 }, { "epoch": 1.41, "learning_rate": 8.585117227319063e-05, "loss": 2.9943, "step": 18044 }, { "epoch": 1.41, "learning_rate": 8.585038814396613e-05, "loss": 2.8697, "step": 18045 }, { "epoch": 1.42, "learning_rate": 8.584960401474164e-05, "loss": 2.9162, "step": 18046 }, { "epoch": 1.42, "learning_rate": 8.584881988551714e-05, "loss": 2.9418, "step": 18047 }, { "epoch": 1.42, "learning_rate": 8.584803575629264e-05, "loss": 2.6922, "step": 18048 }, { "epoch": 1.42, "learning_rate": 8.584725162706815e-05, "loss": 2.7454, "step": 18049 }, { "epoch": 1.42, "learning_rate": 8.584646749784365e-05, "loss": 2.8274, "step": 18050 }, { "epoch": 1.42, "learning_rate": 8.584568336861915e-05, "loss": 2.8679, "step": 18051 }, { "epoch": 1.42, "learning_rate": 8.584489923939466e-05, "loss": 2.8286, "step": 18052 }, { "epoch": 1.42, "learning_rate": 8.584411511017016e-05, "loss": 2.8104, "step": 18053 }, { "epoch": 1.42, "learning_rate": 8.584333098094567e-05, "loss": 2.9367, "step": 18054 }, { "epoch": 1.42, "learning_rate": 8.584254685172117e-05, "loss": 2.9425, "step": 18055 }, { "epoch": 1.42, "learning_rate": 8.584176272249667e-05, "loss": 3.3995, "step": 18056 }, { "epoch": 1.42, "learning_rate": 8.584097859327218e-05, "loss": 2.8054, "step": 18057 }, { "epoch": 1.42, "learning_rate": 8.584019446404768e-05, "loss": 2.7946, "step": 18058 }, { "epoch": 1.42, "learning_rate": 8.583941033482318e-05, "loss": 2.9022, "step": 18059 }, { "epoch": 1.42, "learning_rate": 8.583862620559869e-05, "loss": 3.045, "step": 18060 }, { "epoch": 1.42, "learning_rate": 8.583784207637419e-05, "loss": 3.3059, "step": 18061 }, { "epoch": 1.42, "learning_rate": 8.58370579471497e-05, "loss": 2.6619, "step": 18062 }, { "epoch": 1.42, "learning_rate": 8.58362738179252e-05, "loss": 2.8706, "step": 18063 }, { "epoch": 1.42, "learning_rate": 8.58354896887007e-05, "loss": 2.7931, "step": 18064 }, { "epoch": 1.42, "learning_rate": 8.58347055594762e-05, "loss": 2.8387, "step": 18065 }, { "epoch": 1.42, "learning_rate": 8.583392143025171e-05, "loss": 2.9625, "step": 18066 }, { "epoch": 1.42, "learning_rate": 8.583313730102721e-05, "loss": 2.7695, "step": 18067 }, { "epoch": 1.42, "learning_rate": 8.583235317180272e-05, "loss": 3.1629, "step": 18068 }, { "epoch": 1.42, "learning_rate": 8.583156904257822e-05, "loss": 3.1416, "step": 18069 }, { "epoch": 1.42, "learning_rate": 8.583078491335372e-05, "loss": 3.1786, "step": 18070 }, { "epoch": 1.42, "learning_rate": 8.583000078412923e-05, "loss": 2.9465, "step": 18071 }, { "epoch": 1.42, "learning_rate": 8.582921665490473e-05, "loss": 2.9281, "step": 18072 }, { "epoch": 1.42, "learning_rate": 8.582843252568024e-05, "loss": 2.8636, "step": 18073 }, { "epoch": 1.42, "learning_rate": 8.582764839645574e-05, "loss": 2.7393, "step": 18074 }, { "epoch": 1.42, "learning_rate": 8.582686426723124e-05, "loss": 2.5753, "step": 18075 }, { "epoch": 1.42, "learning_rate": 8.582608013800675e-05, "loss": 2.9056, "step": 18076 }, { "epoch": 1.42, "learning_rate": 8.582529600878225e-05, "loss": 2.9334, "step": 18077 }, { "epoch": 1.42, "learning_rate": 8.582451187955775e-05, "loss": 3.0723, "step": 18078 }, { "epoch": 1.42, "learning_rate": 8.582372775033326e-05, "loss": 2.7288, "step": 18079 }, { "epoch": 1.42, "learning_rate": 8.582294362110876e-05, "loss": 2.8821, "step": 18080 }, { "epoch": 1.42, "learning_rate": 8.582215949188426e-05, "loss": 3.0018, "step": 18081 }, { "epoch": 1.42, "learning_rate": 8.582137536265977e-05, "loss": 2.7484, "step": 18082 }, { "epoch": 1.42, "learning_rate": 8.582059123343527e-05, "loss": 2.8822, "step": 18083 }, { "epoch": 1.42, "learning_rate": 8.581980710421078e-05, "loss": 3.0184, "step": 18084 }, { "epoch": 1.42, "learning_rate": 8.581902297498628e-05, "loss": 2.8421, "step": 18085 }, { "epoch": 1.42, "learning_rate": 8.581823884576178e-05, "loss": 2.9256, "step": 18086 }, { "epoch": 1.42, "learning_rate": 8.581745471653729e-05, "loss": 3.0206, "step": 18087 }, { "epoch": 1.42, "learning_rate": 8.58166705873128e-05, "loss": 2.9857, "step": 18088 }, { "epoch": 1.42, "learning_rate": 8.581588645808831e-05, "loss": 2.8296, "step": 18089 }, { "epoch": 1.42, "learning_rate": 8.581510232886381e-05, "loss": 2.8709, "step": 18090 }, { "epoch": 1.42, "learning_rate": 8.581431819963931e-05, "loss": 2.8689, "step": 18091 }, { "epoch": 1.42, "learning_rate": 8.581353407041482e-05, "loss": 2.9182, "step": 18092 }, { "epoch": 1.42, "learning_rate": 8.581274994119031e-05, "loss": 2.9945, "step": 18093 }, { "epoch": 1.42, "learning_rate": 8.581196581196581e-05, "loss": 2.6814, "step": 18094 }, { "epoch": 1.42, "learning_rate": 8.581118168274132e-05, "loss": 3.0463, "step": 18095 }, { "epoch": 1.42, "learning_rate": 8.581039755351682e-05, "loss": 2.8792, "step": 18096 }, { "epoch": 1.42, "learning_rate": 8.580961342429232e-05, "loss": 2.9806, "step": 18097 }, { "epoch": 1.42, "learning_rate": 8.580882929506783e-05, "loss": 3.0728, "step": 18098 }, { "epoch": 1.42, "learning_rate": 8.580804516584333e-05, "loss": 2.7868, "step": 18099 }, { "epoch": 1.42, "learning_rate": 8.580726103661883e-05, "loss": 2.8021, "step": 18100 }, { "epoch": 1.42, "learning_rate": 8.580647690739434e-05, "loss": 3.041, "step": 18101 }, { "epoch": 1.42, "learning_rate": 8.580569277816984e-05, "loss": 2.9484, "step": 18102 }, { "epoch": 1.42, "learning_rate": 8.580490864894535e-05, "loss": 2.9451, "step": 18103 }, { "epoch": 1.42, "learning_rate": 8.580412451972085e-05, "loss": 2.996, "step": 18104 }, { "epoch": 1.42, "learning_rate": 8.580334039049635e-05, "loss": 2.7516, "step": 18105 }, { "epoch": 1.42, "learning_rate": 8.580255626127186e-05, "loss": 2.8666, "step": 18106 }, { "epoch": 1.42, "learning_rate": 8.580177213204736e-05, "loss": 3.0368, "step": 18107 }, { "epoch": 1.42, "learning_rate": 8.580098800282286e-05, "loss": 2.9106, "step": 18108 }, { "epoch": 1.42, "learning_rate": 8.580020387359837e-05, "loss": 2.8949, "step": 18109 }, { "epoch": 1.42, "learning_rate": 8.579941974437387e-05, "loss": 2.9757, "step": 18110 }, { "epoch": 1.42, "learning_rate": 8.579863561514937e-05, "loss": 3.0439, "step": 18111 }, { "epoch": 1.42, "learning_rate": 8.579785148592488e-05, "loss": 2.7981, "step": 18112 }, { "epoch": 1.42, "learning_rate": 8.57970673567004e-05, "loss": 2.9938, "step": 18113 }, { "epoch": 1.42, "learning_rate": 8.57962832274759e-05, "loss": 3.0595, "step": 18114 }, { "epoch": 1.42, "learning_rate": 8.57954990982514e-05, "loss": 2.7997, "step": 18115 }, { "epoch": 1.42, "learning_rate": 8.579471496902691e-05, "loss": 2.6572, "step": 18116 }, { "epoch": 1.42, "learning_rate": 8.579393083980241e-05, "loss": 3.139, "step": 18117 }, { "epoch": 1.42, "learning_rate": 8.579314671057791e-05, "loss": 2.8691, "step": 18118 }, { "epoch": 1.42, "learning_rate": 8.579236258135342e-05, "loss": 2.7892, "step": 18119 }, { "epoch": 1.42, "learning_rate": 8.579157845212892e-05, "loss": 2.7628, "step": 18120 }, { "epoch": 1.42, "learning_rate": 8.579079432290443e-05, "loss": 2.7951, "step": 18121 }, { "epoch": 1.42, "learning_rate": 8.579001019367993e-05, "loss": 2.9807, "step": 18122 }, { "epoch": 1.42, "learning_rate": 8.578922606445543e-05, "loss": 2.7492, "step": 18123 }, { "epoch": 1.42, "learning_rate": 8.578844193523094e-05, "loss": 2.8946, "step": 18124 }, { "epoch": 1.42, "learning_rate": 8.578765780600644e-05, "loss": 2.9148, "step": 18125 }, { "epoch": 1.42, "learning_rate": 8.578687367678194e-05, "loss": 3.0253, "step": 18126 }, { "epoch": 1.42, "learning_rate": 8.578608954755745e-05, "loss": 3.1151, "step": 18127 }, { "epoch": 1.42, "learning_rate": 8.578530541833295e-05, "loss": 2.7549, "step": 18128 }, { "epoch": 1.42, "learning_rate": 8.578452128910844e-05, "loss": 2.9047, "step": 18129 }, { "epoch": 1.42, "learning_rate": 8.578373715988394e-05, "loss": 2.7611, "step": 18130 }, { "epoch": 1.42, "learning_rate": 8.578295303065945e-05, "loss": 2.9237, "step": 18131 }, { "epoch": 1.42, "learning_rate": 8.578216890143495e-05, "loss": 2.6832, "step": 18132 }, { "epoch": 1.42, "learning_rate": 8.578138477221046e-05, "loss": 2.5604, "step": 18133 }, { "epoch": 1.42, "learning_rate": 8.578060064298596e-05, "loss": 2.9275, "step": 18134 }, { "epoch": 1.42, "learning_rate": 8.577981651376146e-05, "loss": 2.7682, "step": 18135 }, { "epoch": 1.42, "learning_rate": 8.577903238453697e-05, "loss": 2.7101, "step": 18136 }, { "epoch": 1.42, "learning_rate": 8.577824825531247e-05, "loss": 2.9111, "step": 18137 }, { "epoch": 1.42, "learning_rate": 8.577746412608797e-05, "loss": 2.794, "step": 18138 }, { "epoch": 1.42, "learning_rate": 8.577667999686349e-05, "loss": 2.8827, "step": 18139 }, { "epoch": 1.42, "learning_rate": 8.5775895867639e-05, "loss": 3.0948, "step": 18140 }, { "epoch": 1.42, "learning_rate": 8.57751117384145e-05, "loss": 2.9949, "step": 18141 }, { "epoch": 1.42, "learning_rate": 8.577432760919e-05, "loss": 2.756, "step": 18142 }, { "epoch": 1.42, "learning_rate": 8.57735434799655e-05, "loss": 3.0051, "step": 18143 }, { "epoch": 1.42, "learning_rate": 8.577275935074101e-05, "loss": 2.7626, "step": 18144 }, { "epoch": 1.42, "learning_rate": 8.577197522151651e-05, "loss": 2.9697, "step": 18145 }, { "epoch": 1.42, "learning_rate": 8.577119109229202e-05, "loss": 3.1242, "step": 18146 }, { "epoch": 1.42, "learning_rate": 8.577040696306752e-05, "loss": 3.0319, "step": 18147 }, { "epoch": 1.42, "learning_rate": 8.576962283384302e-05, "loss": 2.6816, "step": 18148 }, { "epoch": 1.42, "learning_rate": 8.576883870461853e-05, "loss": 2.7882, "step": 18149 }, { "epoch": 1.42, "learning_rate": 8.576805457539403e-05, "loss": 2.8564, "step": 18150 }, { "epoch": 1.42, "learning_rate": 8.576727044616954e-05, "loss": 2.8181, "step": 18151 }, { "epoch": 1.42, "learning_rate": 8.576648631694504e-05, "loss": 2.7198, "step": 18152 }, { "epoch": 1.42, "learning_rate": 8.576570218772054e-05, "loss": 3.118, "step": 18153 }, { "epoch": 1.42, "learning_rate": 8.576491805849605e-05, "loss": 2.9439, "step": 18154 }, { "epoch": 1.42, "learning_rate": 8.576413392927155e-05, "loss": 2.8557, "step": 18155 }, { "epoch": 1.42, "learning_rate": 8.576334980004705e-05, "loss": 2.906, "step": 18156 }, { "epoch": 1.42, "learning_rate": 8.576256567082256e-05, "loss": 2.7764, "step": 18157 }, { "epoch": 1.42, "learning_rate": 8.576178154159806e-05, "loss": 2.8753, "step": 18158 }, { "epoch": 1.42, "learning_rate": 8.576099741237356e-05, "loss": 3.023, "step": 18159 }, { "epoch": 1.42, "learning_rate": 8.576021328314907e-05, "loss": 2.7746, "step": 18160 }, { "epoch": 1.42, "learning_rate": 8.575942915392457e-05, "loss": 2.8538, "step": 18161 }, { "epoch": 1.42, "learning_rate": 8.575864502470008e-05, "loss": 2.89, "step": 18162 }, { "epoch": 1.42, "learning_rate": 8.575786089547558e-05, "loss": 2.6768, "step": 18163 }, { "epoch": 1.42, "learning_rate": 8.575707676625108e-05, "loss": 2.8134, "step": 18164 }, { "epoch": 1.42, "learning_rate": 8.575629263702659e-05, "loss": 2.8982, "step": 18165 }, { "epoch": 1.42, "learning_rate": 8.575550850780209e-05, "loss": 3.0125, "step": 18166 }, { "epoch": 1.42, "learning_rate": 8.57547243785776e-05, "loss": 3.1051, "step": 18167 }, { "epoch": 1.42, "learning_rate": 8.57539402493531e-05, "loss": 3.068, "step": 18168 }, { "epoch": 1.42, "learning_rate": 8.57531561201286e-05, "loss": 3.1172, "step": 18169 }, { "epoch": 1.42, "learning_rate": 8.57523719909041e-05, "loss": 3.0403, "step": 18170 }, { "epoch": 1.42, "learning_rate": 8.575158786167961e-05, "loss": 2.8147, "step": 18171 }, { "epoch": 1.42, "learning_rate": 8.575080373245511e-05, "loss": 3.0007, "step": 18172 }, { "epoch": 1.42, "learning_rate": 8.575001960323062e-05, "loss": 2.9157, "step": 18173 }, { "epoch": 1.43, "learning_rate": 8.574923547400612e-05, "loss": 2.8552, "step": 18174 }, { "epoch": 1.43, "learning_rate": 8.574845134478162e-05, "loss": 2.8411, "step": 18175 }, { "epoch": 1.43, "learning_rate": 8.574766721555713e-05, "loss": 2.8053, "step": 18176 }, { "epoch": 1.43, "learning_rate": 8.574688308633263e-05, "loss": 2.7293, "step": 18177 }, { "epoch": 1.43, "learning_rate": 8.574609895710813e-05, "loss": 2.8155, "step": 18178 }, { "epoch": 1.43, "learning_rate": 8.574531482788364e-05, "loss": 3.1059, "step": 18179 }, { "epoch": 1.43, "learning_rate": 8.574453069865914e-05, "loss": 2.898, "step": 18180 }, { "epoch": 1.43, "learning_rate": 8.574374656943465e-05, "loss": 3.1333, "step": 18181 }, { "epoch": 1.43, "learning_rate": 8.574296244021015e-05, "loss": 2.8793, "step": 18182 }, { "epoch": 1.43, "learning_rate": 8.574217831098565e-05, "loss": 2.7221, "step": 18183 }, { "epoch": 1.43, "learning_rate": 8.574139418176116e-05, "loss": 2.7688, "step": 18184 }, { "epoch": 1.43, "learning_rate": 8.574061005253666e-05, "loss": 2.7624, "step": 18185 }, { "epoch": 1.43, "learning_rate": 8.573982592331216e-05, "loss": 3.1911, "step": 18186 }, { "epoch": 1.43, "learning_rate": 8.573904179408767e-05, "loss": 2.7522, "step": 18187 }, { "epoch": 1.43, "learning_rate": 8.573825766486317e-05, "loss": 2.7557, "step": 18188 }, { "epoch": 1.43, "learning_rate": 8.573747353563868e-05, "loss": 2.8517, "step": 18189 }, { "epoch": 1.43, "learning_rate": 8.573668940641418e-05, "loss": 2.7616, "step": 18190 }, { "epoch": 1.43, "learning_rate": 8.573590527718968e-05, "loss": 3.0577, "step": 18191 }, { "epoch": 1.43, "learning_rate": 8.573512114796519e-05, "loss": 2.9872, "step": 18192 }, { "epoch": 1.43, "learning_rate": 8.573433701874069e-05, "loss": 2.7963, "step": 18193 }, { "epoch": 1.43, "learning_rate": 8.57335528895162e-05, "loss": 3.0682, "step": 18194 }, { "epoch": 1.43, "learning_rate": 8.57327687602917e-05, "loss": 2.7654, "step": 18195 }, { "epoch": 1.43, "learning_rate": 8.573198463106721e-05, "loss": 2.6946, "step": 18196 }, { "epoch": 1.43, "learning_rate": 8.573120050184272e-05, "loss": 3.0046, "step": 18197 }, { "epoch": 1.43, "learning_rate": 8.573041637261822e-05, "loss": 2.7488, "step": 18198 }, { "epoch": 1.43, "learning_rate": 8.572963224339373e-05, "loss": 2.8575, "step": 18199 }, { "epoch": 1.43, "learning_rate": 8.572884811416923e-05, "loss": 2.7488, "step": 18200 }, { "epoch": 1.43, "learning_rate": 8.572806398494472e-05, "loss": 2.9757, "step": 18201 }, { "epoch": 1.43, "learning_rate": 8.572727985572022e-05, "loss": 3.0174, "step": 18202 }, { "epoch": 1.43, "learning_rate": 8.572649572649573e-05, "loss": 2.7897, "step": 18203 }, { "epoch": 1.43, "learning_rate": 8.572571159727123e-05, "loss": 2.8968, "step": 18204 }, { "epoch": 1.43, "learning_rate": 8.572492746804673e-05, "loss": 2.8941, "step": 18205 }, { "epoch": 1.43, "learning_rate": 8.572414333882224e-05, "loss": 2.8556, "step": 18206 }, { "epoch": 1.43, "learning_rate": 8.572335920959774e-05, "loss": 2.8132, "step": 18207 }, { "epoch": 1.43, "learning_rate": 8.572257508037325e-05, "loss": 3.0546, "step": 18208 }, { "epoch": 1.43, "learning_rate": 8.572179095114875e-05, "loss": 2.9371, "step": 18209 }, { "epoch": 1.43, "learning_rate": 8.572100682192425e-05, "loss": 3.0933, "step": 18210 }, { "epoch": 1.43, "learning_rate": 8.572022269269976e-05, "loss": 3.088, "step": 18211 }, { "epoch": 1.43, "learning_rate": 8.571943856347526e-05, "loss": 2.9875, "step": 18212 }, { "epoch": 1.43, "learning_rate": 8.571865443425076e-05, "loss": 3.0819, "step": 18213 }, { "epoch": 1.43, "learning_rate": 8.571787030502627e-05, "loss": 2.8601, "step": 18214 }, { "epoch": 1.43, "learning_rate": 8.571708617580177e-05, "loss": 2.9813, "step": 18215 }, { "epoch": 1.43, "learning_rate": 8.571630204657727e-05, "loss": 2.8846, "step": 18216 }, { "epoch": 1.43, "learning_rate": 8.571551791735278e-05, "loss": 2.8963, "step": 18217 }, { "epoch": 1.43, "learning_rate": 8.571473378812828e-05, "loss": 2.834, "step": 18218 }, { "epoch": 1.43, "learning_rate": 8.571394965890379e-05, "loss": 2.6911, "step": 18219 }, { "epoch": 1.43, "learning_rate": 8.571316552967929e-05, "loss": 2.9068, "step": 18220 }, { "epoch": 1.43, "learning_rate": 8.57123814004548e-05, "loss": 3.0247, "step": 18221 }, { "epoch": 1.43, "learning_rate": 8.571159727123031e-05, "loss": 2.6815, "step": 18222 }, { "epoch": 1.43, "learning_rate": 8.571081314200581e-05, "loss": 3.0021, "step": 18223 }, { "epoch": 1.43, "learning_rate": 8.571002901278132e-05, "loss": 2.8285, "step": 18224 }, { "epoch": 1.43, "learning_rate": 8.570924488355682e-05, "loss": 2.9511, "step": 18225 }, { "epoch": 1.43, "learning_rate": 8.570846075433232e-05, "loss": 2.804, "step": 18226 }, { "epoch": 1.43, "learning_rate": 8.570767662510783e-05, "loss": 2.9035, "step": 18227 }, { "epoch": 1.43, "learning_rate": 8.570689249588333e-05, "loss": 2.9877, "step": 18228 }, { "epoch": 1.43, "learning_rate": 8.570610836665884e-05, "loss": 3.0216, "step": 18229 }, { "epoch": 1.43, "learning_rate": 8.570532423743434e-05, "loss": 2.8982, "step": 18230 }, { "epoch": 1.43, "learning_rate": 8.570454010820984e-05, "loss": 2.8878, "step": 18231 }, { "epoch": 1.43, "learning_rate": 8.570375597898535e-05, "loss": 2.743, "step": 18232 }, { "epoch": 1.43, "learning_rate": 8.570297184976085e-05, "loss": 2.775, "step": 18233 }, { "epoch": 1.43, "learning_rate": 8.570218772053635e-05, "loss": 3.1882, "step": 18234 }, { "epoch": 1.43, "learning_rate": 8.570140359131186e-05, "loss": 2.729, "step": 18235 }, { "epoch": 1.43, "learning_rate": 8.570061946208736e-05, "loss": 2.9001, "step": 18236 }, { "epoch": 1.43, "learning_rate": 8.569983533286285e-05, "loss": 2.7307, "step": 18237 }, { "epoch": 1.43, "learning_rate": 8.569905120363836e-05, "loss": 2.8541, "step": 18238 }, { "epoch": 1.43, "learning_rate": 8.569826707441386e-05, "loss": 2.9762, "step": 18239 }, { "epoch": 1.43, "learning_rate": 8.569748294518936e-05, "loss": 2.7542, "step": 18240 }, { "epoch": 1.43, "learning_rate": 8.569669881596487e-05, "loss": 2.9022, "step": 18241 }, { "epoch": 1.43, "learning_rate": 8.569591468674037e-05, "loss": 2.7556, "step": 18242 }, { "epoch": 1.43, "learning_rate": 8.569513055751587e-05, "loss": 2.873, "step": 18243 }, { "epoch": 1.43, "learning_rate": 8.569434642829138e-05, "loss": 2.8093, "step": 18244 }, { "epoch": 1.43, "learning_rate": 8.569356229906688e-05, "loss": 3.0293, "step": 18245 }, { "epoch": 1.43, "learning_rate": 8.569277816984238e-05, "loss": 2.9133, "step": 18246 }, { "epoch": 1.43, "learning_rate": 8.56919940406179e-05, "loss": 2.9998, "step": 18247 }, { "epoch": 1.43, "learning_rate": 8.56912099113934e-05, "loss": 3.0194, "step": 18248 }, { "epoch": 1.43, "learning_rate": 8.569042578216891e-05, "loss": 2.9728, "step": 18249 }, { "epoch": 1.43, "learning_rate": 8.568964165294441e-05, "loss": 2.7775, "step": 18250 }, { "epoch": 1.43, "learning_rate": 8.568885752371992e-05, "loss": 2.738, "step": 18251 }, { "epoch": 1.43, "learning_rate": 8.568807339449542e-05, "loss": 2.8796, "step": 18252 }, { "epoch": 1.43, "learning_rate": 8.568728926527092e-05, "loss": 2.9098, "step": 18253 }, { "epoch": 1.43, "learning_rate": 8.568650513604643e-05, "loss": 3.0625, "step": 18254 }, { "epoch": 1.43, "learning_rate": 8.568572100682193e-05, "loss": 2.7775, "step": 18255 }, { "epoch": 1.43, "learning_rate": 8.568493687759744e-05, "loss": 2.718, "step": 18256 }, { "epoch": 1.43, "learning_rate": 8.568415274837294e-05, "loss": 3.0532, "step": 18257 }, { "epoch": 1.43, "learning_rate": 8.568336861914844e-05, "loss": 2.7032, "step": 18258 }, { "epoch": 1.43, "learning_rate": 8.568258448992395e-05, "loss": 3.0243, "step": 18259 }, { "epoch": 1.43, "learning_rate": 8.568180036069945e-05, "loss": 2.8555, "step": 18260 }, { "epoch": 1.43, "learning_rate": 8.568101623147495e-05, "loss": 2.9343, "step": 18261 }, { "epoch": 1.43, "learning_rate": 8.568023210225046e-05, "loss": 3.0137, "step": 18262 }, { "epoch": 1.43, "learning_rate": 8.567944797302596e-05, "loss": 3.1103, "step": 18263 }, { "epoch": 1.43, "learning_rate": 8.567866384380146e-05, "loss": 3.0211, "step": 18264 }, { "epoch": 1.43, "learning_rate": 8.567787971457697e-05, "loss": 2.8117, "step": 18265 }, { "epoch": 1.43, "learning_rate": 8.567709558535247e-05, "loss": 3.0678, "step": 18266 }, { "epoch": 1.43, "learning_rate": 8.567631145612798e-05, "loss": 2.9405, "step": 18267 }, { "epoch": 1.43, "learning_rate": 8.567552732690348e-05, "loss": 2.6167, "step": 18268 }, { "epoch": 1.43, "learning_rate": 8.567474319767898e-05, "loss": 2.6698, "step": 18269 }, { "epoch": 1.43, "learning_rate": 8.567395906845449e-05, "loss": 2.8682, "step": 18270 }, { "epoch": 1.43, "learning_rate": 8.567317493922999e-05, "loss": 3.0485, "step": 18271 }, { "epoch": 1.43, "learning_rate": 8.56723908100055e-05, "loss": 2.8148, "step": 18272 }, { "epoch": 1.43, "learning_rate": 8.5671606680781e-05, "loss": 2.9945, "step": 18273 }, { "epoch": 1.43, "learning_rate": 8.56708225515565e-05, "loss": 2.7609, "step": 18274 }, { "epoch": 1.43, "learning_rate": 8.5670038422332e-05, "loss": 2.881, "step": 18275 }, { "epoch": 1.43, "learning_rate": 8.566925429310751e-05, "loss": 2.8531, "step": 18276 }, { "epoch": 1.43, "learning_rate": 8.566847016388301e-05, "loss": 3.1102, "step": 18277 }, { "epoch": 1.43, "learning_rate": 8.566768603465852e-05, "loss": 2.624, "step": 18278 }, { "epoch": 1.43, "learning_rate": 8.566690190543402e-05, "loss": 3.1587, "step": 18279 }, { "epoch": 1.43, "learning_rate": 8.566611777620952e-05, "loss": 2.9344, "step": 18280 }, { "epoch": 1.43, "learning_rate": 8.566533364698503e-05, "loss": 2.8312, "step": 18281 }, { "epoch": 1.43, "learning_rate": 8.566454951776053e-05, "loss": 2.9282, "step": 18282 }, { "epoch": 1.43, "learning_rate": 8.566376538853603e-05, "loss": 2.7736, "step": 18283 }, { "epoch": 1.43, "learning_rate": 8.566298125931154e-05, "loss": 2.6924, "step": 18284 }, { "epoch": 1.43, "learning_rate": 8.566219713008704e-05, "loss": 2.8937, "step": 18285 }, { "epoch": 1.43, "learning_rate": 8.566141300086255e-05, "loss": 2.8856, "step": 18286 }, { "epoch": 1.43, "learning_rate": 8.566062887163805e-05, "loss": 2.9456, "step": 18287 }, { "epoch": 1.43, "learning_rate": 8.565984474241355e-05, "loss": 2.8551, "step": 18288 }, { "epoch": 1.43, "learning_rate": 8.565906061318906e-05, "loss": 2.9546, "step": 18289 }, { "epoch": 1.43, "learning_rate": 8.565827648396456e-05, "loss": 3.0988, "step": 18290 }, { "epoch": 1.43, "learning_rate": 8.565749235474006e-05, "loss": 3.0027, "step": 18291 }, { "epoch": 1.43, "learning_rate": 8.565670822551557e-05, "loss": 2.8281, "step": 18292 }, { "epoch": 1.43, "learning_rate": 8.565592409629107e-05, "loss": 2.7789, "step": 18293 }, { "epoch": 1.43, "learning_rate": 8.565513996706657e-05, "loss": 2.6487, "step": 18294 }, { "epoch": 1.43, "learning_rate": 8.565435583784208e-05, "loss": 2.86, "step": 18295 }, { "epoch": 1.43, "learning_rate": 8.565357170861758e-05, "loss": 2.876, "step": 18296 }, { "epoch": 1.43, "learning_rate": 8.565278757939309e-05, "loss": 2.8313, "step": 18297 }, { "epoch": 1.43, "learning_rate": 8.565200345016859e-05, "loss": 2.9531, "step": 18298 }, { "epoch": 1.43, "learning_rate": 8.56512193209441e-05, "loss": 2.8238, "step": 18299 }, { "epoch": 1.43, "learning_rate": 8.56504351917196e-05, "loss": 2.7997, "step": 18300 }, { "epoch": 1.44, "learning_rate": 8.56496510624951e-05, "loss": 2.9648, "step": 18301 }, { "epoch": 1.44, "learning_rate": 8.56488669332706e-05, "loss": 2.6684, "step": 18302 }, { "epoch": 1.44, "learning_rate": 8.564808280404611e-05, "loss": 2.9834, "step": 18303 }, { "epoch": 1.44, "learning_rate": 8.564729867482163e-05, "loss": 2.9143, "step": 18304 }, { "epoch": 1.44, "learning_rate": 8.564651454559713e-05, "loss": 3.0042, "step": 18305 }, { "epoch": 1.44, "learning_rate": 8.564573041637263e-05, "loss": 2.742, "step": 18306 }, { "epoch": 1.44, "learning_rate": 8.564494628714814e-05, "loss": 2.6605, "step": 18307 }, { "epoch": 1.44, "learning_rate": 8.564416215792364e-05, "loss": 3.0778, "step": 18308 }, { "epoch": 1.44, "learning_rate": 8.564337802869913e-05, "loss": 2.9542, "step": 18309 }, { "epoch": 1.44, "learning_rate": 8.564259389947463e-05, "loss": 2.9123, "step": 18310 }, { "epoch": 1.44, "learning_rate": 8.564180977025014e-05, "loss": 2.7645, "step": 18311 }, { "epoch": 1.44, "learning_rate": 8.564102564102564e-05, "loss": 2.7194, "step": 18312 }, { "epoch": 1.44, "learning_rate": 8.564024151180114e-05, "loss": 2.7819, "step": 18313 }, { "epoch": 1.44, "learning_rate": 8.563945738257665e-05, "loss": 2.8692, "step": 18314 }, { "epoch": 1.44, "learning_rate": 8.563867325335215e-05, "loss": 3.0087, "step": 18315 }, { "epoch": 1.44, "learning_rate": 8.563788912412766e-05, "loss": 2.9214, "step": 18316 }, { "epoch": 1.44, "learning_rate": 8.563710499490316e-05, "loss": 3.2214, "step": 18317 }, { "epoch": 1.44, "learning_rate": 8.563632086567866e-05, "loss": 2.8293, "step": 18318 }, { "epoch": 1.44, "learning_rate": 8.563553673645417e-05, "loss": 2.8616, "step": 18319 }, { "epoch": 1.44, "learning_rate": 8.563475260722967e-05, "loss": 2.9949, "step": 18320 }, { "epoch": 1.44, "learning_rate": 8.563396847800517e-05, "loss": 3.0373, "step": 18321 }, { "epoch": 1.44, "learning_rate": 8.563318434878068e-05, "loss": 3.151, "step": 18322 }, { "epoch": 1.44, "learning_rate": 8.563240021955618e-05, "loss": 2.8936, "step": 18323 }, { "epoch": 1.44, "learning_rate": 8.563161609033169e-05, "loss": 2.8962, "step": 18324 }, { "epoch": 1.44, "learning_rate": 8.563083196110719e-05, "loss": 2.8872, "step": 18325 }, { "epoch": 1.44, "learning_rate": 8.563004783188269e-05, "loss": 2.9036, "step": 18326 }, { "epoch": 1.44, "learning_rate": 8.56292637026582e-05, "loss": 2.8324, "step": 18327 }, { "epoch": 1.44, "learning_rate": 8.56284795734337e-05, "loss": 2.8395, "step": 18328 }, { "epoch": 1.44, "learning_rate": 8.562769544420922e-05, "loss": 2.9965, "step": 18329 }, { "epoch": 1.44, "learning_rate": 8.562691131498472e-05, "loss": 2.8071, "step": 18330 }, { "epoch": 1.44, "learning_rate": 8.562612718576022e-05, "loss": 2.8823, "step": 18331 }, { "epoch": 1.44, "learning_rate": 8.562534305653573e-05, "loss": 2.9553, "step": 18332 }, { "epoch": 1.44, "learning_rate": 8.562455892731123e-05, "loss": 2.8606, "step": 18333 }, { "epoch": 1.44, "learning_rate": 8.562377479808674e-05, "loss": 2.9167, "step": 18334 }, { "epoch": 1.44, "learning_rate": 8.562299066886224e-05, "loss": 2.9897, "step": 18335 }, { "epoch": 1.44, "learning_rate": 8.562220653963774e-05, "loss": 2.8328, "step": 18336 }, { "epoch": 1.44, "learning_rate": 8.562142241041325e-05, "loss": 2.8775, "step": 18337 }, { "epoch": 1.44, "learning_rate": 8.562063828118875e-05, "loss": 2.9984, "step": 18338 }, { "epoch": 1.44, "learning_rate": 8.561985415196425e-05, "loss": 2.7859, "step": 18339 }, { "epoch": 1.44, "learning_rate": 8.561907002273976e-05, "loss": 2.8346, "step": 18340 }, { "epoch": 1.44, "learning_rate": 8.561828589351526e-05, "loss": 3.122, "step": 18341 }, { "epoch": 1.44, "learning_rate": 8.561750176429077e-05, "loss": 2.9287, "step": 18342 }, { "epoch": 1.44, "learning_rate": 8.561671763506627e-05, "loss": 2.9603, "step": 18343 }, { "epoch": 1.44, "learning_rate": 8.561593350584177e-05, "loss": 2.9292, "step": 18344 }, { "epoch": 1.44, "learning_rate": 8.561514937661726e-05, "loss": 2.9036, "step": 18345 }, { "epoch": 1.44, "learning_rate": 8.561436524739277e-05, "loss": 2.9181, "step": 18346 }, { "epoch": 1.44, "learning_rate": 8.561358111816827e-05, "loss": 2.9887, "step": 18347 }, { "epoch": 1.44, "learning_rate": 8.561279698894377e-05, "loss": 2.8668, "step": 18348 }, { "epoch": 1.44, "learning_rate": 8.561201285971928e-05, "loss": 2.8803, "step": 18349 }, { "epoch": 1.44, "learning_rate": 8.561122873049478e-05, "loss": 2.8863, "step": 18350 }, { "epoch": 1.44, "learning_rate": 8.561044460127028e-05, "loss": 2.7787, "step": 18351 }, { "epoch": 1.44, "learning_rate": 8.560966047204579e-05, "loss": 2.7787, "step": 18352 }, { "epoch": 1.44, "learning_rate": 8.560887634282129e-05, "loss": 3.0428, "step": 18353 }, { "epoch": 1.44, "learning_rate": 8.56080922135968e-05, "loss": 3.0631, "step": 18354 }, { "epoch": 1.44, "learning_rate": 8.560730808437231e-05, "loss": 2.7865, "step": 18355 }, { "epoch": 1.44, "learning_rate": 8.560652395514782e-05, "loss": 3.159, "step": 18356 }, { "epoch": 1.44, "learning_rate": 8.560573982592332e-05, "loss": 2.8837, "step": 18357 }, { "epoch": 1.44, "learning_rate": 8.560495569669882e-05, "loss": 2.9208, "step": 18358 }, { "epoch": 1.44, "learning_rate": 8.560417156747433e-05, "loss": 3.0212, "step": 18359 }, { "epoch": 1.44, "learning_rate": 8.560338743824983e-05, "loss": 2.7437, "step": 18360 }, { "epoch": 1.44, "learning_rate": 8.560260330902533e-05, "loss": 2.983, "step": 18361 }, { "epoch": 1.44, "learning_rate": 8.560181917980084e-05, "loss": 3.0459, "step": 18362 }, { "epoch": 1.44, "learning_rate": 8.560103505057634e-05, "loss": 2.7976, "step": 18363 }, { "epoch": 1.44, "learning_rate": 8.560025092135185e-05, "loss": 2.9399, "step": 18364 }, { "epoch": 1.44, "learning_rate": 8.559946679212735e-05, "loss": 2.7535, "step": 18365 }, { "epoch": 1.44, "learning_rate": 8.559868266290285e-05, "loss": 3.0361, "step": 18366 }, { "epoch": 1.44, "learning_rate": 8.559789853367836e-05, "loss": 2.7416, "step": 18367 }, { "epoch": 1.44, "learning_rate": 8.559711440445386e-05, "loss": 3.0189, "step": 18368 }, { "epoch": 1.44, "learning_rate": 8.559633027522936e-05, "loss": 2.8278, "step": 18369 }, { "epoch": 1.44, "learning_rate": 8.559554614600487e-05, "loss": 2.7268, "step": 18370 }, { "epoch": 1.44, "learning_rate": 8.559476201678037e-05, "loss": 2.778, "step": 18371 }, { "epoch": 1.44, "learning_rate": 8.559397788755588e-05, "loss": 2.8595, "step": 18372 }, { "epoch": 1.44, "learning_rate": 8.559319375833138e-05, "loss": 2.954, "step": 18373 }, { "epoch": 1.44, "learning_rate": 8.559240962910688e-05, "loss": 2.9997, "step": 18374 }, { "epoch": 1.44, "learning_rate": 8.559162549988239e-05, "loss": 2.8307, "step": 18375 }, { "epoch": 1.44, "learning_rate": 8.559084137065789e-05, "loss": 2.7886, "step": 18376 }, { "epoch": 1.44, "learning_rate": 8.55900572414334e-05, "loss": 2.8662, "step": 18377 }, { "epoch": 1.44, "learning_rate": 8.55892731122089e-05, "loss": 2.8484, "step": 18378 }, { "epoch": 1.44, "learning_rate": 8.55884889829844e-05, "loss": 2.8324, "step": 18379 }, { "epoch": 1.44, "learning_rate": 8.55877048537599e-05, "loss": 3.0639, "step": 18380 }, { "epoch": 1.44, "learning_rate": 8.558692072453541e-05, "loss": 2.9401, "step": 18381 }, { "epoch": 1.44, "learning_rate": 8.558613659531091e-05, "loss": 3.1369, "step": 18382 }, { "epoch": 1.44, "learning_rate": 8.558535246608642e-05, "loss": 2.6357, "step": 18383 }, { "epoch": 1.44, "learning_rate": 8.558456833686192e-05, "loss": 3.0184, "step": 18384 }, { "epoch": 1.44, "learning_rate": 8.558378420763742e-05, "loss": 2.8682, "step": 18385 }, { "epoch": 1.44, "learning_rate": 8.558300007841293e-05, "loss": 2.9177, "step": 18386 }, { "epoch": 1.44, "learning_rate": 8.558221594918843e-05, "loss": 2.8658, "step": 18387 }, { "epoch": 1.44, "learning_rate": 8.558143181996393e-05, "loss": 2.7859, "step": 18388 }, { "epoch": 1.44, "learning_rate": 8.558064769073944e-05, "loss": 2.9714, "step": 18389 }, { "epoch": 1.44, "learning_rate": 8.557986356151494e-05, "loss": 2.8243, "step": 18390 }, { "epoch": 1.44, "learning_rate": 8.557907943229045e-05, "loss": 2.7734, "step": 18391 }, { "epoch": 1.44, "learning_rate": 8.557829530306595e-05, "loss": 2.9238, "step": 18392 }, { "epoch": 1.44, "learning_rate": 8.557751117384145e-05, "loss": 2.881, "step": 18393 }, { "epoch": 1.44, "learning_rate": 8.557672704461696e-05, "loss": 2.7689, "step": 18394 }, { "epoch": 1.44, "learning_rate": 8.557594291539246e-05, "loss": 2.8426, "step": 18395 }, { "epoch": 1.44, "learning_rate": 8.557515878616796e-05, "loss": 3.1788, "step": 18396 }, { "epoch": 1.44, "learning_rate": 8.557437465694347e-05, "loss": 2.8722, "step": 18397 }, { "epoch": 1.44, "learning_rate": 8.557359052771897e-05, "loss": 2.7047, "step": 18398 }, { "epoch": 1.44, "learning_rate": 8.557280639849447e-05, "loss": 3.0097, "step": 18399 }, { "epoch": 1.44, "learning_rate": 8.557202226926998e-05, "loss": 2.9909, "step": 18400 }, { "epoch": 1.44, "learning_rate": 8.557123814004548e-05, "loss": 2.8045, "step": 18401 }, { "epoch": 1.44, "learning_rate": 8.557045401082099e-05, "loss": 3.0101, "step": 18402 }, { "epoch": 1.44, "learning_rate": 8.556966988159649e-05, "loss": 2.8255, "step": 18403 }, { "epoch": 1.44, "learning_rate": 8.556888575237199e-05, "loss": 2.9751, "step": 18404 }, { "epoch": 1.44, "learning_rate": 8.55681016231475e-05, "loss": 2.6988, "step": 18405 }, { "epoch": 1.44, "learning_rate": 8.5567317493923e-05, "loss": 2.9186, "step": 18406 }, { "epoch": 1.44, "learning_rate": 8.55665333646985e-05, "loss": 2.9961, "step": 18407 }, { "epoch": 1.44, "learning_rate": 8.556574923547401e-05, "loss": 2.8782, "step": 18408 }, { "epoch": 1.44, "learning_rate": 8.556496510624951e-05, "loss": 2.7014, "step": 18409 }, { "epoch": 1.44, "learning_rate": 8.556418097702502e-05, "loss": 2.7465, "step": 18410 }, { "epoch": 1.44, "learning_rate": 8.556339684780052e-05, "loss": 2.7881, "step": 18411 }, { "epoch": 1.44, "learning_rate": 8.556261271857604e-05, "loss": 2.7688, "step": 18412 }, { "epoch": 1.44, "learning_rate": 8.556182858935154e-05, "loss": 2.8484, "step": 18413 }, { "epoch": 1.44, "learning_rate": 8.556104446012704e-05, "loss": 2.8075, "step": 18414 }, { "epoch": 1.44, "learning_rate": 8.556026033090255e-05, "loss": 2.6627, "step": 18415 }, { "epoch": 1.44, "learning_rate": 8.555947620167805e-05, "loss": 2.8671, "step": 18416 }, { "epoch": 1.44, "learning_rate": 8.555869207245354e-05, "loss": 2.946, "step": 18417 }, { "epoch": 1.44, "learning_rate": 8.555790794322904e-05, "loss": 2.9678, "step": 18418 }, { "epoch": 1.44, "learning_rate": 8.555712381400455e-05, "loss": 2.9398, "step": 18419 }, { "epoch": 1.44, "learning_rate": 8.555633968478005e-05, "loss": 2.9147, "step": 18420 }, { "epoch": 1.44, "learning_rate": 8.555555555555556e-05, "loss": 2.9857, "step": 18421 }, { "epoch": 1.44, "learning_rate": 8.555477142633106e-05, "loss": 2.9784, "step": 18422 }, { "epoch": 1.44, "learning_rate": 8.555398729710656e-05, "loss": 3.0822, "step": 18423 }, { "epoch": 1.44, "learning_rate": 8.555320316788207e-05, "loss": 3.1343, "step": 18424 }, { "epoch": 1.44, "learning_rate": 8.555241903865757e-05, "loss": 3.0468, "step": 18425 }, { "epoch": 1.44, "learning_rate": 8.555163490943307e-05, "loss": 2.8503, "step": 18426 }, { "epoch": 1.44, "learning_rate": 8.555085078020858e-05, "loss": 3.0273, "step": 18427 }, { "epoch": 1.44, "learning_rate": 8.555006665098408e-05, "loss": 3.0087, "step": 18428 }, { "epoch": 1.45, "learning_rate": 8.554928252175959e-05, "loss": 2.728, "step": 18429 }, { "epoch": 1.45, "learning_rate": 8.554849839253509e-05, "loss": 2.8073, "step": 18430 }, { "epoch": 1.45, "learning_rate": 8.554771426331059e-05, "loss": 2.7621, "step": 18431 }, { "epoch": 1.45, "learning_rate": 8.55469301340861e-05, "loss": 2.8126, "step": 18432 }, { "epoch": 1.45, "learning_rate": 8.55461460048616e-05, "loss": 2.6883, "step": 18433 }, { "epoch": 1.45, "learning_rate": 8.55453618756371e-05, "loss": 2.9298, "step": 18434 }, { "epoch": 1.45, "learning_rate": 8.554457774641261e-05, "loss": 3.1796, "step": 18435 }, { "epoch": 1.45, "learning_rate": 8.554379361718811e-05, "loss": 2.9234, "step": 18436 }, { "epoch": 1.45, "learning_rate": 8.554300948796363e-05, "loss": 3.0633, "step": 18437 }, { "epoch": 1.45, "learning_rate": 8.554222535873913e-05, "loss": 2.9915, "step": 18438 }, { "epoch": 1.45, "learning_rate": 8.554144122951464e-05, "loss": 3.1158, "step": 18439 }, { "epoch": 1.45, "learning_rate": 8.554065710029014e-05, "loss": 2.9328, "step": 18440 }, { "epoch": 1.45, "learning_rate": 8.553987297106564e-05, "loss": 3.0391, "step": 18441 }, { "epoch": 1.45, "learning_rate": 8.553908884184115e-05, "loss": 2.8503, "step": 18442 }, { "epoch": 1.45, "learning_rate": 8.553830471261665e-05, "loss": 3.0564, "step": 18443 }, { "epoch": 1.45, "learning_rate": 8.553752058339215e-05, "loss": 3.0129, "step": 18444 }, { "epoch": 1.45, "learning_rate": 8.553673645416766e-05, "loss": 2.6652, "step": 18445 }, { "epoch": 1.45, "learning_rate": 8.553595232494316e-05, "loss": 3.0826, "step": 18446 }, { "epoch": 1.45, "learning_rate": 8.553516819571866e-05, "loss": 2.994, "step": 18447 }, { "epoch": 1.45, "learning_rate": 8.553438406649417e-05, "loss": 2.8626, "step": 18448 }, { "epoch": 1.45, "learning_rate": 8.553359993726967e-05, "loss": 2.7504, "step": 18449 }, { "epoch": 1.45, "learning_rate": 8.553281580804518e-05, "loss": 2.8758, "step": 18450 }, { "epoch": 1.45, "learning_rate": 8.553203167882068e-05, "loss": 2.8559, "step": 18451 }, { "epoch": 1.45, "learning_rate": 8.553124754959618e-05, "loss": 2.8826, "step": 18452 }, { "epoch": 1.45, "learning_rate": 8.553046342037167e-05, "loss": 2.8843, "step": 18453 }, { "epoch": 1.45, "learning_rate": 8.552967929114718e-05, "loss": 3.0161, "step": 18454 }, { "epoch": 1.45, "learning_rate": 8.552889516192268e-05, "loss": 2.8871, "step": 18455 }, { "epoch": 1.45, "learning_rate": 8.552811103269818e-05, "loss": 2.8569, "step": 18456 }, { "epoch": 1.45, "learning_rate": 8.552732690347369e-05, "loss": 3.1754, "step": 18457 }, { "epoch": 1.45, "learning_rate": 8.552654277424919e-05, "loss": 2.8413, "step": 18458 }, { "epoch": 1.45, "learning_rate": 8.55257586450247e-05, "loss": 2.7244, "step": 18459 }, { "epoch": 1.45, "learning_rate": 8.55249745158002e-05, "loss": 3.1028, "step": 18460 }, { "epoch": 1.45, "learning_rate": 8.55241903865757e-05, "loss": 2.879, "step": 18461 }, { "epoch": 1.45, "learning_rate": 8.55234062573512e-05, "loss": 2.7711, "step": 18462 }, { "epoch": 1.45, "learning_rate": 8.552262212812672e-05, "loss": 3.0508, "step": 18463 }, { "epoch": 1.45, "learning_rate": 8.552183799890223e-05, "loss": 2.8201, "step": 18464 }, { "epoch": 1.45, "learning_rate": 8.552105386967773e-05, "loss": 2.6263, "step": 18465 }, { "epoch": 1.45, "learning_rate": 8.552026974045323e-05, "loss": 2.8806, "step": 18466 }, { "epoch": 1.45, "learning_rate": 8.551948561122874e-05, "loss": 3.1645, "step": 18467 }, { "epoch": 1.45, "learning_rate": 8.551870148200424e-05, "loss": 3.0218, "step": 18468 }, { "epoch": 1.45, "learning_rate": 8.551791735277975e-05, "loss": 2.8448, "step": 18469 }, { "epoch": 1.45, "learning_rate": 8.551713322355525e-05, "loss": 2.7856, "step": 18470 }, { "epoch": 1.45, "learning_rate": 8.551634909433075e-05, "loss": 2.9896, "step": 18471 }, { "epoch": 1.45, "learning_rate": 8.551556496510626e-05, "loss": 2.9849, "step": 18472 }, { "epoch": 1.45, "learning_rate": 8.551478083588176e-05, "loss": 3.2366, "step": 18473 }, { "epoch": 1.45, "learning_rate": 8.551399670665726e-05, "loss": 2.9853, "step": 18474 }, { "epoch": 1.45, "learning_rate": 8.551321257743277e-05, "loss": 3.0055, "step": 18475 }, { "epoch": 1.45, "learning_rate": 8.551242844820827e-05, "loss": 2.9772, "step": 18476 }, { "epoch": 1.45, "learning_rate": 8.551164431898378e-05, "loss": 3.0152, "step": 18477 }, { "epoch": 1.45, "learning_rate": 8.551086018975928e-05, "loss": 2.7713, "step": 18478 }, { "epoch": 1.45, "learning_rate": 8.551007606053478e-05, "loss": 2.9576, "step": 18479 }, { "epoch": 1.45, "learning_rate": 8.550929193131029e-05, "loss": 2.9215, "step": 18480 }, { "epoch": 1.45, "learning_rate": 8.550850780208579e-05, "loss": 2.8803, "step": 18481 }, { "epoch": 1.45, "learning_rate": 8.55077236728613e-05, "loss": 3.2542, "step": 18482 }, { "epoch": 1.45, "learning_rate": 8.55069395436368e-05, "loss": 3.1162, "step": 18483 }, { "epoch": 1.45, "learning_rate": 8.55061554144123e-05, "loss": 2.7, "step": 18484 }, { "epoch": 1.45, "learning_rate": 8.55053712851878e-05, "loss": 2.8892, "step": 18485 }, { "epoch": 1.45, "learning_rate": 8.550458715596331e-05, "loss": 2.8239, "step": 18486 }, { "epoch": 1.45, "learning_rate": 8.550380302673881e-05, "loss": 2.6942, "step": 18487 }, { "epoch": 1.45, "learning_rate": 8.550301889751432e-05, "loss": 2.8624, "step": 18488 }, { "epoch": 1.45, "learning_rate": 8.550223476828982e-05, "loss": 3.0836, "step": 18489 }, { "epoch": 1.45, "learning_rate": 8.550145063906532e-05, "loss": 2.9034, "step": 18490 }, { "epoch": 1.45, "learning_rate": 8.550066650984083e-05, "loss": 2.7366, "step": 18491 }, { "epoch": 1.45, "learning_rate": 8.549988238061633e-05, "loss": 2.9534, "step": 18492 }, { "epoch": 1.45, "learning_rate": 8.549909825139183e-05, "loss": 2.864, "step": 18493 }, { "epoch": 1.45, "learning_rate": 8.549831412216734e-05, "loss": 2.7245, "step": 18494 }, { "epoch": 1.45, "learning_rate": 8.549752999294284e-05, "loss": 2.8592, "step": 18495 }, { "epoch": 1.45, "learning_rate": 8.549674586371834e-05, "loss": 2.953, "step": 18496 }, { "epoch": 1.45, "learning_rate": 8.549596173449385e-05, "loss": 2.8276, "step": 18497 }, { "epoch": 1.45, "learning_rate": 8.549517760526935e-05, "loss": 3.063, "step": 18498 }, { "epoch": 1.45, "learning_rate": 8.549439347604486e-05, "loss": 2.9907, "step": 18499 }, { "epoch": 1.45, "learning_rate": 8.549360934682036e-05, "loss": 2.7712, "step": 18500 }, { "epoch": 1.45, "learning_rate": 8.549282521759586e-05, "loss": 2.9434, "step": 18501 }, { "epoch": 1.45, "learning_rate": 8.549204108837137e-05, "loss": 2.8574, "step": 18502 }, { "epoch": 1.45, "learning_rate": 8.549125695914687e-05, "loss": 2.921, "step": 18503 }, { "epoch": 1.45, "learning_rate": 8.549047282992237e-05, "loss": 3.0912, "step": 18504 }, { "epoch": 1.45, "learning_rate": 8.548968870069788e-05, "loss": 2.7939, "step": 18505 }, { "epoch": 1.45, "learning_rate": 8.548890457147338e-05, "loss": 2.8404, "step": 18506 }, { "epoch": 1.45, "learning_rate": 8.548812044224889e-05, "loss": 2.8873, "step": 18507 }, { "epoch": 1.45, "learning_rate": 8.548733631302439e-05, "loss": 3.0248, "step": 18508 }, { "epoch": 1.45, "learning_rate": 8.548655218379989e-05, "loss": 2.9286, "step": 18509 }, { "epoch": 1.45, "learning_rate": 8.54857680545754e-05, "loss": 2.9532, "step": 18510 }, { "epoch": 1.45, "learning_rate": 8.54849839253509e-05, "loss": 2.7545, "step": 18511 }, { "epoch": 1.45, "learning_rate": 8.54841997961264e-05, "loss": 3.1357, "step": 18512 }, { "epoch": 1.45, "learning_rate": 8.548341566690191e-05, "loss": 3.0217, "step": 18513 }, { "epoch": 1.45, "learning_rate": 8.548263153767741e-05, "loss": 2.9612, "step": 18514 }, { "epoch": 1.45, "learning_rate": 8.548184740845291e-05, "loss": 3.1335, "step": 18515 }, { "epoch": 1.45, "learning_rate": 8.548106327922842e-05, "loss": 3.0109, "step": 18516 }, { "epoch": 1.45, "learning_rate": 8.548027915000392e-05, "loss": 3.0123, "step": 18517 }, { "epoch": 1.45, "learning_rate": 8.547949502077943e-05, "loss": 2.8383, "step": 18518 }, { "epoch": 1.45, "learning_rate": 8.547871089155493e-05, "loss": 2.8982, "step": 18519 }, { "epoch": 1.45, "learning_rate": 8.547792676233045e-05, "loss": 2.7784, "step": 18520 }, { "epoch": 1.45, "learning_rate": 8.547714263310595e-05, "loss": 2.9828, "step": 18521 }, { "epoch": 1.45, "learning_rate": 8.547635850388145e-05, "loss": 3.0158, "step": 18522 }, { "epoch": 1.45, "learning_rate": 8.547557437465696e-05, "loss": 2.8884, "step": 18523 }, { "epoch": 1.45, "learning_rate": 8.547479024543246e-05, "loss": 2.7241, "step": 18524 }, { "epoch": 1.45, "learning_rate": 8.547400611620795e-05, "loss": 2.5739, "step": 18525 }, { "epoch": 1.45, "learning_rate": 8.547322198698346e-05, "loss": 2.8893, "step": 18526 }, { "epoch": 1.45, "learning_rate": 8.547243785775896e-05, "loss": 2.9189, "step": 18527 }, { "epoch": 1.45, "learning_rate": 8.547165372853446e-05, "loss": 2.9667, "step": 18528 }, { "epoch": 1.45, "learning_rate": 8.547086959930997e-05, "loss": 2.7101, "step": 18529 }, { "epoch": 1.45, "learning_rate": 8.547008547008547e-05, "loss": 2.8521, "step": 18530 }, { "epoch": 1.45, "learning_rate": 8.546930134086097e-05, "loss": 2.9147, "step": 18531 }, { "epoch": 1.45, "learning_rate": 8.546851721163648e-05, "loss": 2.6105, "step": 18532 }, { "epoch": 1.45, "learning_rate": 8.546773308241198e-05, "loss": 3.0713, "step": 18533 }, { "epoch": 1.45, "learning_rate": 8.546694895318748e-05, "loss": 3.131, "step": 18534 }, { "epoch": 1.45, "learning_rate": 8.546616482396299e-05, "loss": 2.9552, "step": 18535 }, { "epoch": 1.45, "learning_rate": 8.546538069473849e-05, "loss": 2.8531, "step": 18536 }, { "epoch": 1.45, "learning_rate": 8.5464596565514e-05, "loss": 2.8852, "step": 18537 }, { "epoch": 1.45, "learning_rate": 8.54638124362895e-05, "loss": 3.0043, "step": 18538 }, { "epoch": 1.45, "learning_rate": 8.5463028307065e-05, "loss": 2.9487, "step": 18539 }, { "epoch": 1.45, "learning_rate": 8.546224417784051e-05, "loss": 2.8998, "step": 18540 }, { "epoch": 1.45, "learning_rate": 8.546146004861601e-05, "loss": 2.8543, "step": 18541 }, { "epoch": 1.45, "learning_rate": 8.546067591939151e-05, "loss": 3.1437, "step": 18542 }, { "epoch": 1.45, "learning_rate": 8.545989179016702e-05, "loss": 3.0382, "step": 18543 }, { "epoch": 1.45, "learning_rate": 8.545910766094252e-05, "loss": 3.0953, "step": 18544 }, { "epoch": 1.45, "learning_rate": 8.545832353171804e-05, "loss": 2.7576, "step": 18545 }, { "epoch": 1.45, "learning_rate": 8.545753940249354e-05, "loss": 3.0473, "step": 18546 }, { "epoch": 1.45, "learning_rate": 8.545675527326905e-05, "loss": 2.871, "step": 18547 }, { "epoch": 1.45, "learning_rate": 8.545597114404455e-05, "loss": 2.8958, "step": 18548 }, { "epoch": 1.45, "learning_rate": 8.545518701482005e-05, "loss": 3.1019, "step": 18549 }, { "epoch": 1.45, "learning_rate": 8.545440288559556e-05, "loss": 2.8133, "step": 18550 }, { "epoch": 1.45, "learning_rate": 8.545361875637106e-05, "loss": 2.6912, "step": 18551 }, { "epoch": 1.45, "learning_rate": 8.545283462714656e-05, "loss": 3.1797, "step": 18552 }, { "epoch": 1.45, "learning_rate": 8.545205049792207e-05, "loss": 3.0178, "step": 18553 }, { "epoch": 1.45, "learning_rate": 8.545126636869757e-05, "loss": 2.9354, "step": 18554 }, { "epoch": 1.45, "learning_rate": 8.545048223947308e-05, "loss": 2.9218, "step": 18555 }, { "epoch": 1.46, "learning_rate": 8.544969811024858e-05, "loss": 3.2274, "step": 18556 }, { "epoch": 1.46, "learning_rate": 8.544891398102408e-05, "loss": 2.8698, "step": 18557 }, { "epoch": 1.46, "learning_rate": 8.544812985179959e-05, "loss": 2.7879, "step": 18558 }, { "epoch": 1.46, "learning_rate": 8.544734572257509e-05, "loss": 2.9157, "step": 18559 }, { "epoch": 1.46, "learning_rate": 8.54465615933506e-05, "loss": 2.8681, "step": 18560 }, { "epoch": 1.46, "learning_rate": 8.544577746412608e-05, "loss": 2.9746, "step": 18561 }, { "epoch": 1.46, "learning_rate": 8.544499333490159e-05, "loss": 2.9821, "step": 18562 }, { "epoch": 1.46, "learning_rate": 8.544420920567709e-05, "loss": 2.8635, "step": 18563 }, { "epoch": 1.46, "learning_rate": 8.54434250764526e-05, "loss": 2.6714, "step": 18564 }, { "epoch": 1.46, "learning_rate": 8.54426409472281e-05, "loss": 3.0542, "step": 18565 }, { "epoch": 1.46, "learning_rate": 8.54418568180036e-05, "loss": 2.758, "step": 18566 }, { "epoch": 1.46, "learning_rate": 8.54410726887791e-05, "loss": 2.7746, "step": 18567 }, { "epoch": 1.46, "learning_rate": 8.544028855955461e-05, "loss": 2.9284, "step": 18568 }, { "epoch": 1.46, "learning_rate": 8.543950443033011e-05, "loss": 2.806, "step": 18569 }, { "epoch": 1.46, "learning_rate": 8.543872030110562e-05, "loss": 2.8193, "step": 18570 }, { "epoch": 1.46, "learning_rate": 8.543793617188113e-05, "loss": 3.1769, "step": 18571 }, { "epoch": 1.46, "learning_rate": 8.543715204265664e-05, "loss": 3.0635, "step": 18572 }, { "epoch": 1.46, "learning_rate": 8.543636791343214e-05, "loss": 2.8593, "step": 18573 }, { "epoch": 1.46, "learning_rate": 8.543558378420765e-05, "loss": 3.1434, "step": 18574 }, { "epoch": 1.46, "learning_rate": 8.543479965498315e-05, "loss": 2.9246, "step": 18575 }, { "epoch": 1.46, "learning_rate": 8.543401552575865e-05, "loss": 3.1447, "step": 18576 }, { "epoch": 1.46, "learning_rate": 8.543323139653416e-05, "loss": 2.7159, "step": 18577 }, { "epoch": 1.46, "learning_rate": 8.543244726730966e-05, "loss": 2.7933, "step": 18578 }, { "epoch": 1.46, "learning_rate": 8.543166313808516e-05, "loss": 2.8055, "step": 18579 }, { "epoch": 1.46, "learning_rate": 8.543087900886067e-05, "loss": 2.7873, "step": 18580 }, { "epoch": 1.46, "learning_rate": 8.543009487963617e-05, "loss": 2.9032, "step": 18581 }, { "epoch": 1.46, "learning_rate": 8.542931075041167e-05, "loss": 3.1637, "step": 18582 }, { "epoch": 1.46, "learning_rate": 8.542852662118718e-05, "loss": 2.8179, "step": 18583 }, { "epoch": 1.46, "learning_rate": 8.542774249196268e-05, "loss": 3.0072, "step": 18584 }, { "epoch": 1.46, "learning_rate": 8.542695836273819e-05, "loss": 3.1241, "step": 18585 }, { "epoch": 1.46, "learning_rate": 8.542617423351369e-05, "loss": 3.2219, "step": 18586 }, { "epoch": 1.46, "learning_rate": 8.542539010428919e-05, "loss": 2.7992, "step": 18587 }, { "epoch": 1.46, "learning_rate": 8.54246059750647e-05, "loss": 2.962, "step": 18588 }, { "epoch": 1.46, "learning_rate": 8.54238218458402e-05, "loss": 2.8494, "step": 18589 }, { "epoch": 1.46, "learning_rate": 8.54230377166157e-05, "loss": 2.793, "step": 18590 }, { "epoch": 1.46, "learning_rate": 8.542225358739121e-05, "loss": 2.8525, "step": 18591 }, { "epoch": 1.46, "learning_rate": 8.542146945816671e-05, "loss": 2.6763, "step": 18592 }, { "epoch": 1.46, "learning_rate": 8.542068532894222e-05, "loss": 3.0254, "step": 18593 }, { "epoch": 1.46, "learning_rate": 8.541990119971772e-05, "loss": 2.8846, "step": 18594 }, { "epoch": 1.46, "learning_rate": 8.541911707049322e-05, "loss": 2.6868, "step": 18595 }, { "epoch": 1.46, "learning_rate": 8.541833294126873e-05, "loss": 3.007, "step": 18596 }, { "epoch": 1.46, "learning_rate": 8.541754881204423e-05, "loss": 2.8989, "step": 18597 }, { "epoch": 1.46, "learning_rate": 8.541676468281973e-05, "loss": 2.9252, "step": 18598 }, { "epoch": 1.46, "learning_rate": 8.541598055359524e-05, "loss": 2.709, "step": 18599 }, { "epoch": 1.46, "learning_rate": 8.541519642437074e-05, "loss": 3.0352, "step": 18600 }, { "epoch": 1.46, "learning_rate": 8.541441229514624e-05, "loss": 2.8352, "step": 18601 }, { "epoch": 1.46, "learning_rate": 8.541362816592175e-05, "loss": 3.0947, "step": 18602 }, { "epoch": 1.46, "learning_rate": 8.541284403669725e-05, "loss": 3.027, "step": 18603 }, { "epoch": 1.46, "learning_rate": 8.541205990747276e-05, "loss": 3.0878, "step": 18604 }, { "epoch": 1.46, "learning_rate": 8.541127577824826e-05, "loss": 2.934, "step": 18605 }, { "epoch": 1.46, "learning_rate": 8.541049164902376e-05, "loss": 2.694, "step": 18606 }, { "epoch": 1.46, "learning_rate": 8.540970751979927e-05, "loss": 2.7284, "step": 18607 }, { "epoch": 1.46, "learning_rate": 8.540892339057477e-05, "loss": 2.9826, "step": 18608 }, { "epoch": 1.46, "learning_rate": 8.540813926135027e-05, "loss": 2.9819, "step": 18609 }, { "epoch": 1.46, "learning_rate": 8.540735513212578e-05, "loss": 3.1292, "step": 18610 }, { "epoch": 1.46, "learning_rate": 8.540657100290128e-05, "loss": 2.97, "step": 18611 }, { "epoch": 1.46, "learning_rate": 8.540578687367679e-05, "loss": 2.9465, "step": 18612 }, { "epoch": 1.46, "learning_rate": 8.540500274445229e-05, "loss": 2.9446, "step": 18613 }, { "epoch": 1.46, "learning_rate": 8.540421861522779e-05, "loss": 2.9792, "step": 18614 }, { "epoch": 1.46, "learning_rate": 8.54034344860033e-05, "loss": 3.0468, "step": 18615 }, { "epoch": 1.46, "learning_rate": 8.54026503567788e-05, "loss": 2.943, "step": 18616 }, { "epoch": 1.46, "learning_rate": 8.54018662275543e-05, "loss": 2.9443, "step": 18617 }, { "epoch": 1.46, "learning_rate": 8.540108209832981e-05, "loss": 2.7781, "step": 18618 }, { "epoch": 1.46, "learning_rate": 8.540029796910531e-05, "loss": 2.9894, "step": 18619 }, { "epoch": 1.46, "learning_rate": 8.539951383988081e-05, "loss": 3.0032, "step": 18620 }, { "epoch": 1.46, "learning_rate": 8.539872971065632e-05, "loss": 2.8844, "step": 18621 }, { "epoch": 1.46, "learning_rate": 8.539794558143182e-05, "loss": 2.9002, "step": 18622 }, { "epoch": 1.46, "learning_rate": 8.539716145220733e-05, "loss": 2.9804, "step": 18623 }, { "epoch": 1.46, "learning_rate": 8.539637732298283e-05, "loss": 2.9772, "step": 18624 }, { "epoch": 1.46, "learning_rate": 8.539559319375833e-05, "loss": 2.8849, "step": 18625 }, { "epoch": 1.46, "learning_rate": 8.539480906453384e-05, "loss": 2.7185, "step": 18626 }, { "epoch": 1.46, "learning_rate": 8.539402493530934e-05, "loss": 3.0329, "step": 18627 }, { "epoch": 1.46, "learning_rate": 8.539324080608486e-05, "loss": 3.0368, "step": 18628 }, { "epoch": 1.46, "learning_rate": 8.539245667686036e-05, "loss": 2.9502, "step": 18629 }, { "epoch": 1.46, "learning_rate": 8.539167254763586e-05, "loss": 3.0657, "step": 18630 }, { "epoch": 1.46, "learning_rate": 8.539088841841137e-05, "loss": 2.784, "step": 18631 }, { "epoch": 1.46, "learning_rate": 8.539010428918687e-05, "loss": 3.1442, "step": 18632 }, { "epoch": 1.46, "learning_rate": 8.538932015996236e-05, "loss": 2.7969, "step": 18633 }, { "epoch": 1.46, "learning_rate": 8.538853603073787e-05, "loss": 2.8974, "step": 18634 }, { "epoch": 1.46, "learning_rate": 8.538775190151337e-05, "loss": 3.0012, "step": 18635 }, { "epoch": 1.46, "learning_rate": 8.538696777228887e-05, "loss": 2.9813, "step": 18636 }, { "epoch": 1.46, "learning_rate": 8.538618364306438e-05, "loss": 2.6745, "step": 18637 }, { "epoch": 1.46, "learning_rate": 8.538539951383988e-05, "loss": 2.9104, "step": 18638 }, { "epoch": 1.46, "learning_rate": 8.538461538461538e-05, "loss": 2.9623, "step": 18639 }, { "epoch": 1.46, "learning_rate": 8.538383125539089e-05, "loss": 3.0108, "step": 18640 }, { "epoch": 1.46, "learning_rate": 8.538304712616639e-05, "loss": 2.9164, "step": 18641 }, { "epoch": 1.46, "learning_rate": 8.53822629969419e-05, "loss": 3.1967, "step": 18642 }, { "epoch": 1.46, "learning_rate": 8.53814788677174e-05, "loss": 2.84, "step": 18643 }, { "epoch": 1.46, "learning_rate": 8.53806947384929e-05, "loss": 2.5406, "step": 18644 }, { "epoch": 1.46, "learning_rate": 8.53799106092684e-05, "loss": 2.7067, "step": 18645 }, { "epoch": 1.46, "learning_rate": 8.537912648004391e-05, "loss": 2.9863, "step": 18646 }, { "epoch": 1.46, "learning_rate": 8.537834235081941e-05, "loss": 2.8943, "step": 18647 }, { "epoch": 1.46, "learning_rate": 8.537755822159492e-05, "loss": 2.9116, "step": 18648 }, { "epoch": 1.46, "learning_rate": 8.537677409237042e-05, "loss": 2.9861, "step": 18649 }, { "epoch": 1.46, "learning_rate": 8.537598996314592e-05, "loss": 2.8621, "step": 18650 }, { "epoch": 1.46, "learning_rate": 8.537520583392143e-05, "loss": 2.9671, "step": 18651 }, { "epoch": 1.46, "learning_rate": 8.537442170469693e-05, "loss": 2.7927, "step": 18652 }, { "epoch": 1.46, "learning_rate": 8.537363757547244e-05, "loss": 2.8387, "step": 18653 }, { "epoch": 1.46, "learning_rate": 8.537285344624795e-05, "loss": 2.7795, "step": 18654 }, { "epoch": 1.46, "learning_rate": 8.537206931702346e-05, "loss": 3.0405, "step": 18655 }, { "epoch": 1.46, "learning_rate": 8.537128518779896e-05, "loss": 2.7724, "step": 18656 }, { "epoch": 1.46, "learning_rate": 8.537050105857446e-05, "loss": 2.9046, "step": 18657 }, { "epoch": 1.46, "learning_rate": 8.536971692934997e-05, "loss": 3.1591, "step": 18658 }, { "epoch": 1.46, "learning_rate": 8.536893280012547e-05, "loss": 2.8058, "step": 18659 }, { "epoch": 1.46, "learning_rate": 8.536814867090098e-05, "loss": 2.8291, "step": 18660 }, { "epoch": 1.46, "learning_rate": 8.536736454167648e-05, "loss": 2.8579, "step": 18661 }, { "epoch": 1.46, "learning_rate": 8.536658041245198e-05, "loss": 2.8686, "step": 18662 }, { "epoch": 1.46, "learning_rate": 8.536579628322749e-05, "loss": 2.7318, "step": 18663 }, { "epoch": 1.46, "learning_rate": 8.536501215400299e-05, "loss": 2.9195, "step": 18664 }, { "epoch": 1.46, "learning_rate": 8.53642280247785e-05, "loss": 2.7249, "step": 18665 }, { "epoch": 1.46, "learning_rate": 8.5363443895554e-05, "loss": 2.9341, "step": 18666 }, { "epoch": 1.46, "learning_rate": 8.53626597663295e-05, "loss": 2.8829, "step": 18667 }, { "epoch": 1.46, "learning_rate": 8.5361875637105e-05, "loss": 3.2012, "step": 18668 }, { "epoch": 1.46, "learning_rate": 8.53610915078805e-05, "loss": 2.9551, "step": 18669 }, { "epoch": 1.46, "learning_rate": 8.5360307378656e-05, "loss": 3.0231, "step": 18670 }, { "epoch": 1.46, "learning_rate": 8.53595232494315e-05, "loss": 2.8929, "step": 18671 }, { "epoch": 1.46, "learning_rate": 8.5358739120207e-05, "loss": 2.8963, "step": 18672 }, { "epoch": 1.46, "learning_rate": 8.535795499098251e-05, "loss": 2.6914, "step": 18673 }, { "epoch": 1.46, "learning_rate": 8.535717086175801e-05, "loss": 3.2152, "step": 18674 }, { "epoch": 1.46, "learning_rate": 8.535638673253352e-05, "loss": 3.0437, "step": 18675 }, { "epoch": 1.46, "learning_rate": 8.535560260330902e-05, "loss": 2.7404, "step": 18676 }, { "epoch": 1.46, "learning_rate": 8.535481847408452e-05, "loss": 3.0006, "step": 18677 }, { "epoch": 1.46, "learning_rate": 8.535403434486003e-05, "loss": 2.858, "step": 18678 }, { "epoch": 1.46, "learning_rate": 8.535325021563555e-05, "loss": 2.8182, "step": 18679 }, { "epoch": 1.46, "learning_rate": 8.535246608641105e-05, "loss": 2.6613, "step": 18680 }, { "epoch": 1.46, "learning_rate": 8.535168195718655e-05, "loss": 2.6698, "step": 18681 }, { "epoch": 1.46, "learning_rate": 8.535089782796206e-05, "loss": 3.0042, "step": 18682 }, { "epoch": 1.46, "learning_rate": 8.535011369873756e-05, "loss": 2.8373, "step": 18683 }, { "epoch": 1.47, "learning_rate": 8.534932956951306e-05, "loss": 2.9979, "step": 18684 }, { "epoch": 1.47, "learning_rate": 8.534854544028857e-05, "loss": 2.9826, "step": 18685 }, { "epoch": 1.47, "learning_rate": 8.534776131106407e-05, "loss": 2.773, "step": 18686 }, { "epoch": 1.47, "learning_rate": 8.534697718183957e-05, "loss": 2.8771, "step": 18687 }, { "epoch": 1.47, "learning_rate": 8.534619305261508e-05, "loss": 2.7709, "step": 18688 }, { "epoch": 1.47, "learning_rate": 8.534540892339058e-05, "loss": 2.8516, "step": 18689 }, { "epoch": 1.47, "learning_rate": 8.534462479416609e-05, "loss": 2.9569, "step": 18690 }, { "epoch": 1.47, "learning_rate": 8.534384066494159e-05, "loss": 2.99, "step": 18691 }, { "epoch": 1.47, "learning_rate": 8.534305653571709e-05, "loss": 2.8243, "step": 18692 }, { "epoch": 1.47, "learning_rate": 8.53422724064926e-05, "loss": 2.8582, "step": 18693 }, { "epoch": 1.47, "learning_rate": 8.53414882772681e-05, "loss": 2.9867, "step": 18694 }, { "epoch": 1.47, "learning_rate": 8.53407041480436e-05, "loss": 2.8112, "step": 18695 }, { "epoch": 1.47, "learning_rate": 8.533992001881911e-05, "loss": 2.9, "step": 18696 }, { "epoch": 1.47, "learning_rate": 8.533913588959461e-05, "loss": 3.0154, "step": 18697 }, { "epoch": 1.47, "learning_rate": 8.533835176037011e-05, "loss": 2.8461, "step": 18698 }, { "epoch": 1.47, "learning_rate": 8.533756763114562e-05, "loss": 2.9669, "step": 18699 }, { "epoch": 1.47, "learning_rate": 8.533678350192112e-05, "loss": 2.8579, "step": 18700 }, { "epoch": 1.47, "learning_rate": 8.533599937269663e-05, "loss": 2.8607, "step": 18701 }, { "epoch": 1.47, "learning_rate": 8.533521524347213e-05, "loss": 3.2656, "step": 18702 }, { "epoch": 1.47, "learning_rate": 8.533443111424763e-05, "loss": 3.0097, "step": 18703 }, { "epoch": 1.47, "learning_rate": 8.533364698502314e-05, "loss": 3.0279, "step": 18704 }, { "epoch": 1.47, "learning_rate": 8.533286285579864e-05, "loss": 3.0735, "step": 18705 }, { "epoch": 1.47, "learning_rate": 8.533207872657414e-05, "loss": 2.8067, "step": 18706 }, { "epoch": 1.47, "learning_rate": 8.533129459734965e-05, "loss": 3.3348, "step": 18707 }, { "epoch": 1.47, "learning_rate": 8.533051046812515e-05, "loss": 2.8887, "step": 18708 }, { "epoch": 1.47, "learning_rate": 8.532972633890066e-05, "loss": 2.9187, "step": 18709 }, { "epoch": 1.47, "learning_rate": 8.532894220967616e-05, "loss": 2.9123, "step": 18710 }, { "epoch": 1.47, "learning_rate": 8.532815808045166e-05, "loss": 3.0142, "step": 18711 }, { "epoch": 1.47, "learning_rate": 8.532737395122717e-05, "loss": 2.9768, "step": 18712 }, { "epoch": 1.47, "learning_rate": 8.532658982200267e-05, "loss": 2.6972, "step": 18713 }, { "epoch": 1.47, "learning_rate": 8.532580569277817e-05, "loss": 2.9278, "step": 18714 }, { "epoch": 1.47, "learning_rate": 8.532502156355368e-05, "loss": 2.8237, "step": 18715 }, { "epoch": 1.47, "learning_rate": 8.532423743432918e-05, "loss": 2.7176, "step": 18716 }, { "epoch": 1.47, "learning_rate": 8.532345330510468e-05, "loss": 2.9961, "step": 18717 }, { "epoch": 1.47, "learning_rate": 8.532266917588019e-05, "loss": 2.814, "step": 18718 }, { "epoch": 1.47, "learning_rate": 8.532188504665569e-05, "loss": 2.6543, "step": 18719 }, { "epoch": 1.47, "learning_rate": 8.53211009174312e-05, "loss": 2.6963, "step": 18720 }, { "epoch": 1.47, "learning_rate": 8.53203167882067e-05, "loss": 2.9665, "step": 18721 }, { "epoch": 1.47, "learning_rate": 8.53195326589822e-05, "loss": 2.8721, "step": 18722 }, { "epoch": 1.47, "learning_rate": 8.531874852975771e-05, "loss": 2.9974, "step": 18723 }, { "epoch": 1.47, "learning_rate": 8.531796440053321e-05, "loss": 3.0263, "step": 18724 }, { "epoch": 1.47, "learning_rate": 8.531718027130871e-05, "loss": 3.0658, "step": 18725 }, { "epoch": 1.47, "learning_rate": 8.531639614208422e-05, "loss": 3.0458, "step": 18726 }, { "epoch": 1.47, "learning_rate": 8.531561201285972e-05, "loss": 2.8576, "step": 18727 }, { "epoch": 1.47, "learning_rate": 8.531482788363523e-05, "loss": 2.9641, "step": 18728 }, { "epoch": 1.47, "learning_rate": 8.531404375441073e-05, "loss": 2.7426, "step": 18729 }, { "epoch": 1.47, "learning_rate": 8.531325962518623e-05, "loss": 2.7398, "step": 18730 }, { "epoch": 1.47, "learning_rate": 8.531247549596174e-05, "loss": 2.82, "step": 18731 }, { "epoch": 1.47, "learning_rate": 8.531169136673724e-05, "loss": 2.8894, "step": 18732 }, { "epoch": 1.47, "learning_rate": 8.531090723751274e-05, "loss": 2.9582, "step": 18733 }, { "epoch": 1.47, "learning_rate": 8.531012310828825e-05, "loss": 2.8559, "step": 18734 }, { "epoch": 1.47, "learning_rate": 8.530933897906375e-05, "loss": 2.9242, "step": 18735 }, { "epoch": 1.47, "learning_rate": 8.530855484983927e-05, "loss": 3.0547, "step": 18736 }, { "epoch": 1.47, "learning_rate": 8.530777072061477e-05, "loss": 2.8314, "step": 18737 }, { "epoch": 1.47, "learning_rate": 8.530698659139028e-05, "loss": 2.718, "step": 18738 }, { "epoch": 1.47, "learning_rate": 8.530620246216578e-05, "loss": 2.827, "step": 18739 }, { "epoch": 1.47, "learning_rate": 8.530541833294128e-05, "loss": 2.8435, "step": 18740 }, { "epoch": 1.47, "learning_rate": 8.530463420371677e-05, "loss": 3.1451, "step": 18741 }, { "epoch": 1.47, "learning_rate": 8.530385007449228e-05, "loss": 2.8935, "step": 18742 }, { "epoch": 1.47, "learning_rate": 8.530306594526778e-05, "loss": 2.7378, "step": 18743 }, { "epoch": 1.47, "learning_rate": 8.530228181604328e-05, "loss": 3.0969, "step": 18744 }, { "epoch": 1.47, "learning_rate": 8.530149768681879e-05, "loss": 2.8284, "step": 18745 }, { "epoch": 1.47, "learning_rate": 8.530071355759429e-05, "loss": 2.8667, "step": 18746 }, { "epoch": 1.47, "learning_rate": 8.52999294283698e-05, "loss": 2.8347, "step": 18747 }, { "epoch": 1.47, "learning_rate": 8.52991452991453e-05, "loss": 3.0358, "step": 18748 }, { "epoch": 1.47, "learning_rate": 8.52983611699208e-05, "loss": 2.9892, "step": 18749 }, { "epoch": 1.47, "learning_rate": 8.52975770406963e-05, "loss": 2.7692, "step": 18750 }, { "epoch": 1.47, "learning_rate": 8.529679291147181e-05, "loss": 2.9408, "step": 18751 }, { "epoch": 1.47, "learning_rate": 8.529600878224731e-05, "loss": 2.6376, "step": 18752 }, { "epoch": 1.47, "learning_rate": 8.529522465302282e-05, "loss": 2.7812, "step": 18753 }, { "epoch": 1.47, "learning_rate": 8.529444052379832e-05, "loss": 2.832, "step": 18754 }, { "epoch": 1.47, "learning_rate": 8.529365639457382e-05, "loss": 2.7929, "step": 18755 }, { "epoch": 1.47, "learning_rate": 8.529287226534933e-05, "loss": 2.965, "step": 18756 }, { "epoch": 1.47, "learning_rate": 8.529208813612483e-05, "loss": 3.1663, "step": 18757 }, { "epoch": 1.47, "learning_rate": 8.529130400690034e-05, "loss": 2.9672, "step": 18758 }, { "epoch": 1.47, "learning_rate": 8.529051987767584e-05, "loss": 3.0584, "step": 18759 }, { "epoch": 1.47, "learning_rate": 8.528973574845134e-05, "loss": 2.8507, "step": 18760 }, { "epoch": 1.47, "learning_rate": 8.528895161922685e-05, "loss": 2.9231, "step": 18761 }, { "epoch": 1.47, "learning_rate": 8.528816749000236e-05, "loss": 2.8775, "step": 18762 }, { "epoch": 1.47, "learning_rate": 8.528738336077787e-05, "loss": 2.9952, "step": 18763 }, { "epoch": 1.47, "learning_rate": 8.528659923155337e-05, "loss": 3.1691, "step": 18764 }, { "epoch": 1.47, "learning_rate": 8.528581510232887e-05, "loss": 2.6635, "step": 18765 }, { "epoch": 1.47, "learning_rate": 8.528503097310438e-05, "loss": 2.914, "step": 18766 }, { "epoch": 1.47, "learning_rate": 8.528424684387988e-05, "loss": 2.6508, "step": 18767 }, { "epoch": 1.47, "learning_rate": 8.528346271465539e-05, "loss": 2.534, "step": 18768 }, { "epoch": 1.47, "learning_rate": 8.528267858543089e-05, "loss": 2.7118, "step": 18769 }, { "epoch": 1.47, "learning_rate": 8.52818944562064e-05, "loss": 3.1854, "step": 18770 }, { "epoch": 1.47, "learning_rate": 8.52811103269819e-05, "loss": 2.6729, "step": 18771 }, { "epoch": 1.47, "learning_rate": 8.52803261977574e-05, "loss": 2.9401, "step": 18772 }, { "epoch": 1.47, "learning_rate": 8.52795420685329e-05, "loss": 3.0214, "step": 18773 }, { "epoch": 1.47, "learning_rate": 8.527875793930841e-05, "loss": 3.1272, "step": 18774 }, { "epoch": 1.47, "learning_rate": 8.527797381008391e-05, "loss": 2.9644, "step": 18775 }, { "epoch": 1.47, "learning_rate": 8.527718968085942e-05, "loss": 2.8852, "step": 18776 }, { "epoch": 1.47, "learning_rate": 8.52764055516349e-05, "loss": 2.8308, "step": 18777 }, { "epoch": 1.47, "learning_rate": 8.527562142241041e-05, "loss": 2.8543, "step": 18778 }, { "epoch": 1.47, "learning_rate": 8.527483729318591e-05, "loss": 2.788, "step": 18779 }, { "epoch": 1.47, "learning_rate": 8.527405316396142e-05, "loss": 3.0349, "step": 18780 }, { "epoch": 1.47, "learning_rate": 8.527326903473692e-05, "loss": 2.943, "step": 18781 }, { "epoch": 1.47, "learning_rate": 8.527248490551242e-05, "loss": 2.7471, "step": 18782 }, { "epoch": 1.47, "learning_rate": 8.527170077628793e-05, "loss": 2.8601, "step": 18783 }, { "epoch": 1.47, "learning_rate": 8.527091664706343e-05, "loss": 2.9406, "step": 18784 }, { "epoch": 1.47, "learning_rate": 8.527013251783893e-05, "loss": 3.06, "step": 18785 }, { "epoch": 1.47, "learning_rate": 8.526934838861444e-05, "loss": 2.8455, "step": 18786 }, { "epoch": 1.47, "learning_rate": 8.526856425938996e-05, "loss": 2.8744, "step": 18787 }, { "epoch": 1.47, "learning_rate": 8.526778013016546e-05, "loss": 2.9742, "step": 18788 }, { "epoch": 1.47, "learning_rate": 8.526699600094096e-05, "loss": 3.12, "step": 18789 }, { "epoch": 1.47, "learning_rate": 8.526621187171647e-05, "loss": 2.7198, "step": 18790 }, { "epoch": 1.47, "learning_rate": 8.526542774249197e-05, "loss": 3.043, "step": 18791 }, { "epoch": 1.47, "learning_rate": 8.526464361326747e-05, "loss": 2.9818, "step": 18792 }, { "epoch": 1.47, "learning_rate": 8.526385948404298e-05, "loss": 2.9256, "step": 18793 }, { "epoch": 1.47, "learning_rate": 8.526307535481848e-05, "loss": 2.7001, "step": 18794 }, { "epoch": 1.47, "learning_rate": 8.526229122559399e-05, "loss": 2.8475, "step": 18795 }, { "epoch": 1.47, "learning_rate": 8.526150709636949e-05, "loss": 2.7456, "step": 18796 }, { "epoch": 1.47, "learning_rate": 8.526072296714499e-05, "loss": 2.8773, "step": 18797 }, { "epoch": 1.47, "learning_rate": 8.52599388379205e-05, "loss": 3.4352, "step": 18798 }, { "epoch": 1.47, "learning_rate": 8.5259154708696e-05, "loss": 2.7788, "step": 18799 }, { "epoch": 1.47, "learning_rate": 8.52583705794715e-05, "loss": 2.9508, "step": 18800 }, { "epoch": 1.47, "learning_rate": 8.525758645024701e-05, "loss": 3.1576, "step": 18801 }, { "epoch": 1.47, "learning_rate": 8.525680232102251e-05, "loss": 2.8441, "step": 18802 }, { "epoch": 1.47, "learning_rate": 8.525601819179801e-05, "loss": 2.8224, "step": 18803 }, { "epoch": 1.47, "learning_rate": 8.525523406257352e-05, "loss": 2.7321, "step": 18804 }, { "epoch": 1.47, "learning_rate": 8.525444993334902e-05, "loss": 2.6704, "step": 18805 }, { "epoch": 1.47, "learning_rate": 8.525366580412453e-05, "loss": 2.8282, "step": 18806 }, { "epoch": 1.47, "learning_rate": 8.525288167490003e-05, "loss": 3.0092, "step": 18807 }, { "epoch": 1.47, "learning_rate": 8.525209754567553e-05, "loss": 2.689, "step": 18808 }, { "epoch": 1.47, "learning_rate": 8.525131341645104e-05, "loss": 2.974, "step": 18809 }, { "epoch": 1.47, "learning_rate": 8.525052928722654e-05, "loss": 2.9692, "step": 18810 }, { "epoch": 1.48, "learning_rate": 8.524974515800204e-05, "loss": 3.0446, "step": 18811 }, { "epoch": 1.48, "learning_rate": 8.524896102877755e-05, "loss": 2.9372, "step": 18812 }, { "epoch": 1.48, "learning_rate": 8.524817689955305e-05, "loss": 3.0074, "step": 18813 }, { "epoch": 1.48, "learning_rate": 8.524739277032856e-05, "loss": 2.8269, "step": 18814 }, { "epoch": 1.48, "learning_rate": 8.524660864110406e-05, "loss": 2.9854, "step": 18815 }, { "epoch": 1.48, "learning_rate": 8.524582451187956e-05, "loss": 2.9491, "step": 18816 }, { "epoch": 1.48, "learning_rate": 8.524504038265507e-05, "loss": 3.0754, "step": 18817 }, { "epoch": 1.48, "learning_rate": 8.524425625343057e-05, "loss": 2.6096, "step": 18818 }, { "epoch": 1.48, "learning_rate": 8.524347212420607e-05, "loss": 3.0047, "step": 18819 }, { "epoch": 1.48, "learning_rate": 8.524268799498158e-05, "loss": 3.0502, "step": 18820 }, { "epoch": 1.48, "learning_rate": 8.524190386575708e-05, "loss": 3.0914, "step": 18821 }, { "epoch": 1.48, "learning_rate": 8.524111973653258e-05, "loss": 2.784, "step": 18822 }, { "epoch": 1.48, "learning_rate": 8.524033560730809e-05, "loss": 2.9022, "step": 18823 }, { "epoch": 1.48, "learning_rate": 8.523955147808359e-05, "loss": 3.0772, "step": 18824 }, { "epoch": 1.48, "learning_rate": 8.52387673488591e-05, "loss": 2.9882, "step": 18825 }, { "epoch": 1.48, "learning_rate": 8.52379832196346e-05, "loss": 2.9396, "step": 18826 }, { "epoch": 1.48, "learning_rate": 8.52371990904101e-05, "loss": 2.9565, "step": 18827 }, { "epoch": 1.48, "learning_rate": 8.52364149611856e-05, "loss": 3.0684, "step": 18828 }, { "epoch": 1.48, "learning_rate": 8.523563083196111e-05, "loss": 2.8252, "step": 18829 }, { "epoch": 1.48, "learning_rate": 8.523484670273661e-05, "loss": 2.8338, "step": 18830 }, { "epoch": 1.48, "learning_rate": 8.523406257351212e-05, "loss": 2.8953, "step": 18831 }, { "epoch": 1.48, "learning_rate": 8.523327844428762e-05, "loss": 3.0314, "step": 18832 }, { "epoch": 1.48, "learning_rate": 8.523249431506313e-05, "loss": 2.8104, "step": 18833 }, { "epoch": 1.48, "learning_rate": 8.523171018583863e-05, "loss": 3.0245, "step": 18834 }, { "epoch": 1.48, "learning_rate": 8.523092605661413e-05, "loss": 3.0522, "step": 18835 }, { "epoch": 1.48, "learning_rate": 8.523014192738964e-05, "loss": 2.9441, "step": 18836 }, { "epoch": 1.48, "learning_rate": 8.522935779816514e-05, "loss": 2.6997, "step": 18837 }, { "epoch": 1.48, "learning_rate": 8.522857366894064e-05, "loss": 2.7713, "step": 18838 }, { "epoch": 1.48, "learning_rate": 8.522778953971615e-05, "loss": 2.7085, "step": 18839 }, { "epoch": 1.48, "learning_rate": 8.522700541049165e-05, "loss": 3.2027, "step": 18840 }, { "epoch": 1.48, "learning_rate": 8.522622128126715e-05, "loss": 2.8563, "step": 18841 }, { "epoch": 1.48, "learning_rate": 8.522543715204266e-05, "loss": 2.9976, "step": 18842 }, { "epoch": 1.48, "learning_rate": 8.522465302281816e-05, "loss": 2.8376, "step": 18843 }, { "epoch": 1.48, "learning_rate": 8.522386889359368e-05, "loss": 2.6649, "step": 18844 }, { "epoch": 1.48, "learning_rate": 8.522308476436918e-05, "loss": 2.5613, "step": 18845 }, { "epoch": 1.48, "learning_rate": 8.522230063514469e-05, "loss": 3.2469, "step": 18846 }, { "epoch": 1.48, "learning_rate": 8.522151650592019e-05, "loss": 2.9993, "step": 18847 }, { "epoch": 1.48, "learning_rate": 8.52207323766957e-05, "loss": 3.0139, "step": 18848 }, { "epoch": 1.48, "learning_rate": 8.521994824747118e-05, "loss": 3.0064, "step": 18849 }, { "epoch": 1.48, "learning_rate": 8.521916411824669e-05, "loss": 2.9477, "step": 18850 }, { "epoch": 1.48, "learning_rate": 8.521837998902219e-05, "loss": 2.8196, "step": 18851 }, { "epoch": 1.48, "learning_rate": 8.52175958597977e-05, "loss": 3.1338, "step": 18852 }, { "epoch": 1.48, "learning_rate": 8.52168117305732e-05, "loss": 2.7361, "step": 18853 }, { "epoch": 1.48, "learning_rate": 8.52160276013487e-05, "loss": 2.9829, "step": 18854 }, { "epoch": 1.48, "learning_rate": 8.52152434721242e-05, "loss": 3.2829, "step": 18855 }, { "epoch": 1.48, "learning_rate": 8.521445934289971e-05, "loss": 2.962, "step": 18856 }, { "epoch": 1.48, "learning_rate": 8.521367521367521e-05, "loss": 2.9234, "step": 18857 }, { "epoch": 1.48, "learning_rate": 8.521289108445072e-05, "loss": 2.7865, "step": 18858 }, { "epoch": 1.48, "learning_rate": 8.521210695522622e-05, "loss": 2.9119, "step": 18859 }, { "epoch": 1.48, "learning_rate": 8.521132282600172e-05, "loss": 2.8793, "step": 18860 }, { "epoch": 1.48, "learning_rate": 8.521053869677723e-05, "loss": 2.7661, "step": 18861 }, { "epoch": 1.48, "learning_rate": 8.520975456755273e-05, "loss": 2.9356, "step": 18862 }, { "epoch": 1.48, "learning_rate": 8.520897043832824e-05, "loss": 2.9182, "step": 18863 }, { "epoch": 1.48, "learning_rate": 8.520818630910374e-05, "loss": 2.883, "step": 18864 }, { "epoch": 1.48, "learning_rate": 8.520740217987924e-05, "loss": 2.7514, "step": 18865 }, { "epoch": 1.48, "learning_rate": 8.520661805065475e-05, "loss": 3.0545, "step": 18866 }, { "epoch": 1.48, "learning_rate": 8.520583392143025e-05, "loss": 2.7773, "step": 18867 }, { "epoch": 1.48, "learning_rate": 8.520504979220575e-05, "loss": 2.8264, "step": 18868 }, { "epoch": 1.48, "learning_rate": 8.520426566298126e-05, "loss": 2.814, "step": 18869 }, { "epoch": 1.48, "learning_rate": 8.520348153375677e-05, "loss": 2.9145, "step": 18870 }, { "epoch": 1.48, "learning_rate": 8.520269740453228e-05, "loss": 2.7569, "step": 18871 }, { "epoch": 1.48, "learning_rate": 8.520191327530778e-05, "loss": 3.0055, "step": 18872 }, { "epoch": 1.48, "learning_rate": 8.520112914608329e-05, "loss": 3.0226, "step": 18873 }, { "epoch": 1.48, "learning_rate": 8.520034501685879e-05, "loss": 2.9465, "step": 18874 }, { "epoch": 1.48, "learning_rate": 8.519956088763429e-05, "loss": 2.9427, "step": 18875 }, { "epoch": 1.48, "learning_rate": 8.51987767584098e-05, "loss": 2.6955, "step": 18876 }, { "epoch": 1.48, "learning_rate": 8.51979926291853e-05, "loss": 2.8008, "step": 18877 }, { "epoch": 1.48, "learning_rate": 8.51972084999608e-05, "loss": 2.8912, "step": 18878 }, { "epoch": 1.48, "learning_rate": 8.519642437073631e-05, "loss": 2.9799, "step": 18879 }, { "epoch": 1.48, "learning_rate": 8.519564024151181e-05, "loss": 2.8275, "step": 18880 }, { "epoch": 1.48, "learning_rate": 8.519485611228732e-05, "loss": 2.7605, "step": 18881 }, { "epoch": 1.48, "learning_rate": 8.519407198306282e-05, "loss": 2.8713, "step": 18882 }, { "epoch": 1.48, "learning_rate": 8.519328785383832e-05, "loss": 3.3325, "step": 18883 }, { "epoch": 1.48, "learning_rate": 8.519250372461383e-05, "loss": 2.8531, "step": 18884 }, { "epoch": 1.48, "learning_rate": 8.519171959538932e-05, "loss": 2.8646, "step": 18885 }, { "epoch": 1.48, "learning_rate": 8.519093546616482e-05, "loss": 2.8203, "step": 18886 }, { "epoch": 1.48, "learning_rate": 8.519015133694032e-05, "loss": 2.693, "step": 18887 }, { "epoch": 1.48, "learning_rate": 8.518936720771583e-05, "loss": 2.8278, "step": 18888 }, { "epoch": 1.48, "learning_rate": 8.518858307849133e-05, "loss": 3.0056, "step": 18889 }, { "epoch": 1.48, "learning_rate": 8.518779894926683e-05, "loss": 2.8896, "step": 18890 }, { "epoch": 1.48, "learning_rate": 8.518701482004234e-05, "loss": 2.987, "step": 18891 }, { "epoch": 1.48, "learning_rate": 8.518623069081784e-05, "loss": 3.1481, "step": 18892 }, { "epoch": 1.48, "learning_rate": 8.518544656159335e-05, "loss": 3.1098, "step": 18893 }, { "epoch": 1.48, "learning_rate": 8.518466243236885e-05, "loss": 3.0456, "step": 18894 }, { "epoch": 1.48, "learning_rate": 8.518387830314437e-05, "loss": 3.0266, "step": 18895 }, { "epoch": 1.48, "learning_rate": 8.518309417391987e-05, "loss": 3.041, "step": 18896 }, { "epoch": 1.48, "learning_rate": 8.518231004469537e-05, "loss": 2.9763, "step": 18897 }, { "epoch": 1.48, "learning_rate": 8.518152591547088e-05, "loss": 2.9003, "step": 18898 }, { "epoch": 1.48, "learning_rate": 8.518074178624638e-05, "loss": 2.9022, "step": 18899 }, { "epoch": 1.48, "learning_rate": 8.517995765702188e-05, "loss": 2.8706, "step": 18900 }, { "epoch": 1.48, "learning_rate": 8.517917352779739e-05, "loss": 2.8598, "step": 18901 }, { "epoch": 1.48, "learning_rate": 8.517838939857289e-05, "loss": 2.8493, "step": 18902 }, { "epoch": 1.48, "learning_rate": 8.51776052693484e-05, "loss": 2.7375, "step": 18903 }, { "epoch": 1.48, "learning_rate": 8.51768211401239e-05, "loss": 2.9244, "step": 18904 }, { "epoch": 1.48, "learning_rate": 8.51760370108994e-05, "loss": 2.8448, "step": 18905 }, { "epoch": 1.48, "learning_rate": 8.517525288167491e-05, "loss": 2.8363, "step": 18906 }, { "epoch": 1.48, "learning_rate": 8.517446875245041e-05, "loss": 2.7965, "step": 18907 }, { "epoch": 1.48, "learning_rate": 8.517368462322591e-05, "loss": 2.9253, "step": 18908 }, { "epoch": 1.48, "learning_rate": 8.517290049400142e-05, "loss": 2.9663, "step": 18909 }, { "epoch": 1.48, "learning_rate": 8.517211636477692e-05, "loss": 2.9769, "step": 18910 }, { "epoch": 1.48, "learning_rate": 8.517133223555243e-05, "loss": 2.7973, "step": 18911 }, { "epoch": 1.48, "learning_rate": 8.517054810632793e-05, "loss": 2.8549, "step": 18912 }, { "epoch": 1.48, "learning_rate": 8.516976397710343e-05, "loss": 3.0707, "step": 18913 }, { "epoch": 1.48, "learning_rate": 8.516897984787894e-05, "loss": 2.7894, "step": 18914 }, { "epoch": 1.48, "learning_rate": 8.516819571865444e-05, "loss": 2.9605, "step": 18915 }, { "epoch": 1.48, "learning_rate": 8.516741158942994e-05, "loss": 3.0086, "step": 18916 }, { "epoch": 1.48, "learning_rate": 8.516662746020545e-05, "loss": 2.8582, "step": 18917 }, { "epoch": 1.48, "learning_rate": 8.516584333098095e-05, "loss": 2.8261, "step": 18918 }, { "epoch": 1.48, "learning_rate": 8.516505920175645e-05, "loss": 2.7912, "step": 18919 }, { "epoch": 1.48, "learning_rate": 8.516427507253196e-05, "loss": 2.7561, "step": 18920 }, { "epoch": 1.48, "learning_rate": 8.516349094330746e-05, "loss": 2.8407, "step": 18921 }, { "epoch": 1.48, "learning_rate": 8.516270681408297e-05, "loss": 2.943, "step": 18922 }, { "epoch": 1.48, "learning_rate": 8.516192268485847e-05, "loss": 2.881, "step": 18923 }, { "epoch": 1.48, "learning_rate": 8.516113855563397e-05, "loss": 2.8804, "step": 18924 }, { "epoch": 1.48, "learning_rate": 8.516035442640948e-05, "loss": 2.9307, "step": 18925 }, { "epoch": 1.48, "learning_rate": 8.515957029718498e-05, "loss": 2.7063, "step": 18926 }, { "epoch": 1.48, "learning_rate": 8.515878616796048e-05, "loss": 2.9951, "step": 18927 }, { "epoch": 1.48, "learning_rate": 8.515800203873599e-05, "loss": 2.8902, "step": 18928 }, { "epoch": 1.48, "learning_rate": 8.515721790951149e-05, "loss": 2.9184, "step": 18929 }, { "epoch": 1.48, "learning_rate": 8.5156433780287e-05, "loss": 3.0254, "step": 18930 }, { "epoch": 1.48, "learning_rate": 8.51556496510625e-05, "loss": 3.492, "step": 18931 }, { "epoch": 1.48, "learning_rate": 8.5154865521838e-05, "loss": 3.0332, "step": 18932 }, { "epoch": 1.48, "learning_rate": 8.51540813926135e-05, "loss": 2.7745, "step": 18933 }, { "epoch": 1.48, "learning_rate": 8.515329726338901e-05, "loss": 3.045, "step": 18934 }, { "epoch": 1.48, "learning_rate": 8.515251313416451e-05, "loss": 2.721, "step": 18935 }, { "epoch": 1.48, "learning_rate": 8.515172900494002e-05, "loss": 2.9234, "step": 18936 }, { "epoch": 1.48, "learning_rate": 8.515094487571552e-05, "loss": 3.0672, "step": 18937 }, { "epoch": 1.48, "learning_rate": 8.515016074649102e-05, "loss": 2.9629, "step": 18938 }, { "epoch": 1.49, "learning_rate": 8.514937661726653e-05, "loss": 2.9499, "step": 18939 }, { "epoch": 1.49, "learning_rate": 8.514859248804203e-05, "loss": 2.8824, "step": 18940 }, { "epoch": 1.49, "learning_rate": 8.514780835881754e-05, "loss": 2.9634, "step": 18941 }, { "epoch": 1.49, "learning_rate": 8.514702422959304e-05, "loss": 3.0476, "step": 18942 }, { "epoch": 1.49, "learning_rate": 8.514624010036854e-05, "loss": 2.7496, "step": 18943 }, { "epoch": 1.49, "learning_rate": 8.514545597114405e-05, "loss": 2.8887, "step": 18944 }, { "epoch": 1.49, "learning_rate": 8.514467184191955e-05, "loss": 3.0608, "step": 18945 }, { "epoch": 1.49, "learning_rate": 8.514388771269505e-05, "loss": 2.8765, "step": 18946 }, { "epoch": 1.49, "learning_rate": 8.514310358347056e-05, "loss": 2.817, "step": 18947 }, { "epoch": 1.49, "learning_rate": 8.514231945424606e-05, "loss": 2.776, "step": 18948 }, { "epoch": 1.49, "learning_rate": 8.514153532502157e-05, "loss": 3.055, "step": 18949 }, { "epoch": 1.49, "learning_rate": 8.514075119579707e-05, "loss": 3.0057, "step": 18950 }, { "epoch": 1.49, "learning_rate": 8.513996706657257e-05, "loss": 3.1716, "step": 18951 }, { "epoch": 1.49, "learning_rate": 8.513918293734809e-05, "loss": 2.8208, "step": 18952 }, { "epoch": 1.49, "learning_rate": 8.51383988081236e-05, "loss": 2.8664, "step": 18953 }, { "epoch": 1.49, "learning_rate": 8.51376146788991e-05, "loss": 3.1273, "step": 18954 }, { "epoch": 1.49, "learning_rate": 8.51368305496746e-05, "loss": 2.8602, "step": 18955 }, { "epoch": 1.49, "learning_rate": 8.51360464204501e-05, "loss": 2.8204, "step": 18956 }, { "epoch": 1.49, "learning_rate": 8.51352622912256e-05, "loss": 2.9424, "step": 18957 }, { "epoch": 1.49, "learning_rate": 8.51344781620011e-05, "loss": 3.2761, "step": 18958 }, { "epoch": 1.49, "learning_rate": 8.51336940327766e-05, "loss": 2.7964, "step": 18959 }, { "epoch": 1.49, "learning_rate": 8.51329099035521e-05, "loss": 2.8322, "step": 18960 }, { "epoch": 1.49, "learning_rate": 8.513212577432761e-05, "loss": 3.028, "step": 18961 }, { "epoch": 1.49, "learning_rate": 8.513134164510311e-05, "loss": 2.9193, "step": 18962 }, { "epoch": 1.49, "learning_rate": 8.513055751587862e-05, "loss": 2.7597, "step": 18963 }, { "epoch": 1.49, "learning_rate": 8.512977338665412e-05, "loss": 2.9385, "step": 18964 }, { "epoch": 1.49, "learning_rate": 8.512898925742962e-05, "loss": 2.9944, "step": 18965 }, { "epoch": 1.49, "learning_rate": 8.512820512820513e-05, "loss": 2.8958, "step": 18966 }, { "epoch": 1.49, "learning_rate": 8.512742099898063e-05, "loss": 2.7164, "step": 18967 }, { "epoch": 1.49, "learning_rate": 8.512663686975614e-05, "loss": 2.7359, "step": 18968 }, { "epoch": 1.49, "learning_rate": 8.512585274053164e-05, "loss": 2.7933, "step": 18969 }, { "epoch": 1.49, "learning_rate": 8.512506861130714e-05, "loss": 2.7375, "step": 18970 }, { "epoch": 1.49, "learning_rate": 8.512428448208265e-05, "loss": 3.083, "step": 18971 }, { "epoch": 1.49, "learning_rate": 8.512350035285815e-05, "loss": 3.1483, "step": 18972 }, { "epoch": 1.49, "learning_rate": 8.512271622363365e-05, "loss": 3.0126, "step": 18973 }, { "epoch": 1.49, "learning_rate": 8.512193209440916e-05, "loss": 2.8076, "step": 18974 }, { "epoch": 1.49, "learning_rate": 8.512114796518466e-05, "loss": 3.0964, "step": 18975 }, { "epoch": 1.49, "learning_rate": 8.512036383596016e-05, "loss": 3.0255, "step": 18976 }, { "epoch": 1.49, "learning_rate": 8.511957970673567e-05, "loss": 2.8805, "step": 18977 }, { "epoch": 1.49, "learning_rate": 8.511879557751119e-05, "loss": 2.8329, "step": 18978 }, { "epoch": 1.49, "learning_rate": 8.511801144828669e-05, "loss": 2.7509, "step": 18979 }, { "epoch": 1.49, "learning_rate": 8.511722731906219e-05, "loss": 2.7701, "step": 18980 }, { "epoch": 1.49, "learning_rate": 8.51164431898377e-05, "loss": 2.6642, "step": 18981 }, { "epoch": 1.49, "learning_rate": 8.51156590606132e-05, "loss": 2.822, "step": 18982 }, { "epoch": 1.49, "learning_rate": 8.51148749313887e-05, "loss": 3.128, "step": 18983 }, { "epoch": 1.49, "learning_rate": 8.511409080216421e-05, "loss": 2.8192, "step": 18984 }, { "epoch": 1.49, "learning_rate": 8.511330667293971e-05, "loss": 2.8583, "step": 18985 }, { "epoch": 1.49, "learning_rate": 8.511252254371521e-05, "loss": 3.054, "step": 18986 }, { "epoch": 1.49, "learning_rate": 8.511173841449072e-05, "loss": 2.8003, "step": 18987 }, { "epoch": 1.49, "learning_rate": 8.511095428526622e-05, "loss": 2.9134, "step": 18988 }, { "epoch": 1.49, "learning_rate": 8.511017015604173e-05, "loss": 2.9086, "step": 18989 }, { "epoch": 1.49, "learning_rate": 8.510938602681723e-05, "loss": 3.0726, "step": 18990 }, { "epoch": 1.49, "learning_rate": 8.510860189759273e-05, "loss": 2.9537, "step": 18991 }, { "epoch": 1.49, "learning_rate": 8.510781776836824e-05, "loss": 3.0217, "step": 18992 }, { "epoch": 1.49, "learning_rate": 8.510703363914373e-05, "loss": 2.9093, "step": 18993 }, { "epoch": 1.49, "learning_rate": 8.510624950991923e-05, "loss": 2.9355, "step": 18994 }, { "epoch": 1.49, "learning_rate": 8.510546538069473e-05, "loss": 2.8745, "step": 18995 }, { "epoch": 1.49, "learning_rate": 8.510468125147024e-05, "loss": 2.8949, "step": 18996 }, { "epoch": 1.49, "learning_rate": 8.510389712224574e-05, "loss": 2.9744, "step": 18997 }, { "epoch": 1.49, "learning_rate": 8.510311299302125e-05, "loss": 2.8051, "step": 18998 }, { "epoch": 1.49, "learning_rate": 8.510232886379675e-05, "loss": 2.9711, "step": 18999 }, { "epoch": 1.49, "learning_rate": 8.510154473457225e-05, "loss": 3.0291, "step": 19000 }, { "epoch": 1.49, "eval_loss": 2.9106497764587402, "eval_runtime": 217.8896, "eval_samples_per_second": 52.008, "eval_steps_per_second": 3.254, "step": 19000 }, { "epoch": 1.49, "learning_rate": 8.510076060534776e-05, "loss": 2.8699, "step": 19001 }, { "epoch": 1.49, "learning_rate": 8.509997647612326e-05, "loss": 2.9906, "step": 19002 }, { "epoch": 1.49, "learning_rate": 8.509919234689878e-05, "loss": 3.2911, "step": 19003 }, { "epoch": 1.49, "learning_rate": 8.509840821767428e-05, "loss": 2.9124, "step": 19004 }, { "epoch": 1.49, "learning_rate": 8.509762408844978e-05, "loss": 2.9876, "step": 19005 }, { "epoch": 1.49, "learning_rate": 8.509683995922529e-05, "loss": 2.8731, "step": 19006 }, { "epoch": 1.49, "learning_rate": 8.509605583000079e-05, "loss": 3.1761, "step": 19007 }, { "epoch": 1.49, "learning_rate": 8.50952717007763e-05, "loss": 3.0422, "step": 19008 }, { "epoch": 1.49, "learning_rate": 8.50944875715518e-05, "loss": 3.1703, "step": 19009 }, { "epoch": 1.49, "learning_rate": 8.50937034423273e-05, "loss": 2.7501, "step": 19010 }, { "epoch": 1.49, "learning_rate": 8.50929193131028e-05, "loss": 3.2891, "step": 19011 }, { "epoch": 1.49, "learning_rate": 8.509213518387831e-05, "loss": 2.923, "step": 19012 }, { "epoch": 1.49, "learning_rate": 8.509135105465381e-05, "loss": 2.7612, "step": 19013 }, { "epoch": 1.49, "learning_rate": 8.509056692542932e-05, "loss": 2.9202, "step": 19014 }, { "epoch": 1.49, "learning_rate": 8.508978279620482e-05, "loss": 2.8112, "step": 19015 }, { "epoch": 1.49, "learning_rate": 8.508899866698033e-05, "loss": 2.9964, "step": 19016 }, { "epoch": 1.49, "learning_rate": 8.508821453775583e-05, "loss": 3.0118, "step": 19017 }, { "epoch": 1.49, "learning_rate": 8.508743040853133e-05, "loss": 2.8102, "step": 19018 }, { "epoch": 1.49, "learning_rate": 8.508664627930684e-05, "loss": 2.8833, "step": 19019 }, { "epoch": 1.49, "learning_rate": 8.508586215008234e-05, "loss": 2.7427, "step": 19020 }, { "epoch": 1.49, "learning_rate": 8.508507802085784e-05, "loss": 3.1197, "step": 19021 }, { "epoch": 1.49, "learning_rate": 8.508429389163335e-05, "loss": 3.1147, "step": 19022 }, { "epoch": 1.49, "learning_rate": 8.508350976240885e-05, "loss": 2.8261, "step": 19023 }, { "epoch": 1.49, "learning_rate": 8.508272563318435e-05, "loss": 2.8756, "step": 19024 }, { "epoch": 1.49, "learning_rate": 8.508194150395986e-05, "loss": 2.6545, "step": 19025 }, { "epoch": 1.49, "learning_rate": 8.508115737473536e-05, "loss": 2.7438, "step": 19026 }, { "epoch": 1.49, "learning_rate": 8.508037324551087e-05, "loss": 2.8054, "step": 19027 }, { "epoch": 1.49, "learning_rate": 8.507958911628637e-05, "loss": 2.8171, "step": 19028 }, { "epoch": 1.49, "learning_rate": 8.507880498706187e-05, "loss": 3.0328, "step": 19029 }, { "epoch": 1.49, "learning_rate": 8.507802085783738e-05, "loss": 2.9443, "step": 19030 }, { "epoch": 1.49, "learning_rate": 8.507723672861288e-05, "loss": 2.8346, "step": 19031 }, { "epoch": 1.49, "learning_rate": 8.507645259938838e-05, "loss": 2.7966, "step": 19032 }, { "epoch": 1.49, "learning_rate": 8.507566847016389e-05, "loss": 2.8566, "step": 19033 }, { "epoch": 1.49, "learning_rate": 8.507488434093939e-05, "loss": 2.8506, "step": 19034 }, { "epoch": 1.49, "learning_rate": 8.50741002117149e-05, "loss": 3.0122, "step": 19035 }, { "epoch": 1.49, "learning_rate": 8.50733160824904e-05, "loss": 3.0323, "step": 19036 }, { "epoch": 1.49, "learning_rate": 8.50725319532659e-05, "loss": 2.8885, "step": 19037 }, { "epoch": 1.49, "learning_rate": 8.50717478240414e-05, "loss": 2.8361, "step": 19038 }, { "epoch": 1.49, "learning_rate": 8.507096369481691e-05, "loss": 2.7375, "step": 19039 }, { "epoch": 1.49, "learning_rate": 8.507017956559241e-05, "loss": 2.7742, "step": 19040 }, { "epoch": 1.49, "learning_rate": 8.506939543636792e-05, "loss": 3.0008, "step": 19041 }, { "epoch": 1.49, "learning_rate": 8.506861130714342e-05, "loss": 2.9075, "step": 19042 }, { "epoch": 1.49, "learning_rate": 8.506782717791892e-05, "loss": 2.852, "step": 19043 }, { "epoch": 1.49, "learning_rate": 8.506704304869443e-05, "loss": 3.097, "step": 19044 }, { "epoch": 1.49, "learning_rate": 8.506625891946993e-05, "loss": 2.7357, "step": 19045 }, { "epoch": 1.49, "learning_rate": 8.506547479024544e-05, "loss": 2.9659, "step": 19046 }, { "epoch": 1.49, "learning_rate": 8.506469066102094e-05, "loss": 2.9447, "step": 19047 }, { "epoch": 1.49, "learning_rate": 8.506390653179644e-05, "loss": 2.7614, "step": 19048 }, { "epoch": 1.49, "learning_rate": 8.506312240257195e-05, "loss": 2.6765, "step": 19049 }, { "epoch": 1.49, "learning_rate": 8.506233827334745e-05, "loss": 2.7202, "step": 19050 }, { "epoch": 1.49, "learning_rate": 8.506155414412295e-05, "loss": 2.8793, "step": 19051 }, { "epoch": 1.49, "learning_rate": 8.506077001489846e-05, "loss": 3.0145, "step": 19052 }, { "epoch": 1.49, "learning_rate": 8.505998588567396e-05, "loss": 2.7513, "step": 19053 }, { "epoch": 1.49, "learning_rate": 8.505920175644946e-05, "loss": 3.0627, "step": 19054 }, { "epoch": 1.49, "learning_rate": 8.505841762722497e-05, "loss": 2.8858, "step": 19055 }, { "epoch": 1.49, "learning_rate": 8.505763349800047e-05, "loss": 2.8756, "step": 19056 }, { "epoch": 1.49, "learning_rate": 8.505684936877598e-05, "loss": 2.9088, "step": 19057 }, { "epoch": 1.49, "learning_rate": 8.505606523955148e-05, "loss": 2.8958, "step": 19058 }, { "epoch": 1.49, "learning_rate": 8.505528111032698e-05, "loss": 3.0335, "step": 19059 }, { "epoch": 1.49, "learning_rate": 8.50544969811025e-05, "loss": 3.0977, "step": 19060 }, { "epoch": 1.49, "learning_rate": 8.5053712851878e-05, "loss": 2.7997, "step": 19061 }, { "epoch": 1.49, "learning_rate": 8.505292872265351e-05, "loss": 2.9826, "step": 19062 }, { "epoch": 1.49, "learning_rate": 8.505214459342901e-05, "loss": 2.9867, "step": 19063 }, { "epoch": 1.49, "learning_rate": 8.505136046420452e-05, "loss": 2.947, "step": 19064 }, { "epoch": 1.49, "learning_rate": 8.505057633498e-05, "loss": 3.2099, "step": 19065 }, { "epoch": 1.5, "learning_rate": 8.504979220575551e-05, "loss": 2.9318, "step": 19066 }, { "epoch": 1.5, "learning_rate": 8.504900807653101e-05, "loss": 2.7741, "step": 19067 }, { "epoch": 1.5, "learning_rate": 8.504822394730652e-05, "loss": 2.6034, "step": 19068 }, { "epoch": 1.5, "learning_rate": 8.504743981808202e-05, "loss": 2.8342, "step": 19069 }, { "epoch": 1.5, "learning_rate": 8.504665568885752e-05, "loss": 2.832, "step": 19070 }, { "epoch": 1.5, "learning_rate": 8.504587155963303e-05, "loss": 2.8321, "step": 19071 }, { "epoch": 1.5, "learning_rate": 8.504508743040853e-05, "loss": 3.1556, "step": 19072 }, { "epoch": 1.5, "learning_rate": 8.504430330118403e-05, "loss": 2.6951, "step": 19073 }, { "epoch": 1.5, "learning_rate": 8.504351917195954e-05, "loss": 3.0592, "step": 19074 }, { "epoch": 1.5, "learning_rate": 8.504273504273504e-05, "loss": 2.664, "step": 19075 }, { "epoch": 1.5, "learning_rate": 8.504195091351055e-05, "loss": 2.7077, "step": 19076 }, { "epoch": 1.5, "learning_rate": 8.504116678428605e-05, "loss": 2.6498, "step": 19077 }, { "epoch": 1.5, "learning_rate": 8.504038265506155e-05, "loss": 2.8705, "step": 19078 }, { "epoch": 1.5, "learning_rate": 8.503959852583706e-05, "loss": 2.7887, "step": 19079 }, { "epoch": 1.5, "learning_rate": 8.503881439661256e-05, "loss": 2.8294, "step": 19080 }, { "epoch": 1.5, "learning_rate": 8.503803026738806e-05, "loss": 2.6926, "step": 19081 }, { "epoch": 1.5, "learning_rate": 8.503724613816357e-05, "loss": 2.712, "step": 19082 }, { "epoch": 1.5, "learning_rate": 8.503646200893907e-05, "loss": 2.8803, "step": 19083 }, { "epoch": 1.5, "learning_rate": 8.503567787971458e-05, "loss": 2.8479, "step": 19084 }, { "epoch": 1.5, "learning_rate": 8.503489375049008e-05, "loss": 2.8735, "step": 19085 }, { "epoch": 1.5, "learning_rate": 8.50341096212656e-05, "loss": 2.9065, "step": 19086 }, { "epoch": 1.5, "learning_rate": 8.50333254920411e-05, "loss": 2.8562, "step": 19087 }, { "epoch": 1.5, "learning_rate": 8.50325413628166e-05, "loss": 2.8831, "step": 19088 }, { "epoch": 1.5, "learning_rate": 8.503175723359211e-05, "loss": 2.7665, "step": 19089 }, { "epoch": 1.5, "learning_rate": 8.503097310436761e-05, "loss": 3.0507, "step": 19090 }, { "epoch": 1.5, "learning_rate": 8.503018897514311e-05, "loss": 2.8969, "step": 19091 }, { "epoch": 1.5, "learning_rate": 8.502940484591862e-05, "loss": 2.7616, "step": 19092 }, { "epoch": 1.5, "learning_rate": 8.502862071669412e-05, "loss": 2.824, "step": 19093 }, { "epoch": 1.5, "learning_rate": 8.502783658746963e-05, "loss": 2.9307, "step": 19094 }, { "epoch": 1.5, "learning_rate": 8.502705245824513e-05, "loss": 3.0854, "step": 19095 }, { "epoch": 1.5, "learning_rate": 8.502626832902063e-05, "loss": 2.9761, "step": 19096 }, { "epoch": 1.5, "learning_rate": 8.502548419979614e-05, "loss": 2.8536, "step": 19097 }, { "epoch": 1.5, "learning_rate": 8.502470007057164e-05, "loss": 2.7249, "step": 19098 }, { "epoch": 1.5, "learning_rate": 8.502391594134714e-05, "loss": 2.8688, "step": 19099 }, { "epoch": 1.5, "learning_rate": 8.502313181212265e-05, "loss": 2.8269, "step": 19100 }, { "epoch": 1.5, "learning_rate": 8.502234768289814e-05, "loss": 3.0894, "step": 19101 }, { "epoch": 1.5, "learning_rate": 8.502156355367364e-05, "loss": 2.9507, "step": 19102 }, { "epoch": 1.5, "learning_rate": 8.502077942444915e-05, "loss": 2.9992, "step": 19103 }, { "epoch": 1.5, "learning_rate": 8.501999529522465e-05, "loss": 2.6793, "step": 19104 }, { "epoch": 1.5, "learning_rate": 8.501921116600015e-05, "loss": 3.1394, "step": 19105 }, { "epoch": 1.5, "learning_rate": 8.501842703677566e-05, "loss": 2.8637, "step": 19106 }, { "epoch": 1.5, "learning_rate": 8.501764290755116e-05, "loss": 2.809, "step": 19107 }, { "epoch": 1.5, "learning_rate": 8.501685877832666e-05, "loss": 2.7348, "step": 19108 }, { "epoch": 1.5, "learning_rate": 8.501607464910217e-05, "loss": 3.105, "step": 19109 }, { "epoch": 1.5, "learning_rate": 8.501529051987767e-05, "loss": 2.8021, "step": 19110 }, { "epoch": 1.5, "learning_rate": 8.501450639065319e-05, "loss": 3.0844, "step": 19111 }, { "epoch": 1.5, "learning_rate": 8.501372226142869e-05, "loss": 3.1952, "step": 19112 }, { "epoch": 1.5, "learning_rate": 8.50129381322042e-05, "loss": 2.7405, "step": 19113 }, { "epoch": 1.5, "learning_rate": 8.50121540029797e-05, "loss": 3.0222, "step": 19114 }, { "epoch": 1.5, "learning_rate": 8.50113698737552e-05, "loss": 2.876, "step": 19115 }, { "epoch": 1.5, "learning_rate": 8.50105857445307e-05, "loss": 3.0103, "step": 19116 }, { "epoch": 1.5, "learning_rate": 8.500980161530621e-05, "loss": 2.9203, "step": 19117 }, { "epoch": 1.5, "learning_rate": 8.500901748608171e-05, "loss": 2.7567, "step": 19118 }, { "epoch": 1.5, "learning_rate": 8.500823335685722e-05, "loss": 2.7047, "step": 19119 }, { "epoch": 1.5, "learning_rate": 8.500744922763272e-05, "loss": 2.9491, "step": 19120 }, { "epoch": 1.5, "learning_rate": 8.500666509840822e-05, "loss": 2.893, "step": 19121 }, { "epoch": 1.5, "learning_rate": 8.500588096918373e-05, "loss": 2.7093, "step": 19122 }, { "epoch": 1.5, "learning_rate": 8.500509683995923e-05, "loss": 2.8234, "step": 19123 }, { "epoch": 1.5, "learning_rate": 8.500431271073474e-05, "loss": 2.9435, "step": 19124 }, { "epoch": 1.5, "learning_rate": 8.500352858151024e-05, "loss": 2.8245, "step": 19125 }, { "epoch": 1.5, "learning_rate": 8.500274445228574e-05, "loss": 2.8848, "step": 19126 }, { "epoch": 1.5, "learning_rate": 8.500196032306125e-05, "loss": 3.2841, "step": 19127 }, { "epoch": 1.5, "learning_rate": 8.500117619383675e-05, "loss": 3.0459, "step": 19128 }, { "epoch": 1.5, "learning_rate": 8.500039206461225e-05, "loss": 2.8951, "step": 19129 }, { "epoch": 1.5, "learning_rate": 8.499960793538776e-05, "loss": 2.9809, "step": 19130 }, { "epoch": 1.5, "learning_rate": 8.499882380616326e-05, "loss": 2.7187, "step": 19131 }, { "epoch": 1.5, "learning_rate": 8.499803967693877e-05, "loss": 2.7367, "step": 19132 }, { "epoch": 1.5, "learning_rate": 8.499725554771427e-05, "loss": 2.6612, "step": 19133 }, { "epoch": 1.5, "learning_rate": 8.499647141848977e-05, "loss": 2.6441, "step": 19134 }, { "epoch": 1.5, "learning_rate": 8.499568728926528e-05, "loss": 2.9508, "step": 19135 }, { "epoch": 1.5, "learning_rate": 8.499490316004078e-05, "loss": 3.1327, "step": 19136 }, { "epoch": 1.5, "learning_rate": 8.499411903081628e-05, "loss": 2.7837, "step": 19137 }, { "epoch": 1.5, "learning_rate": 8.499333490159179e-05, "loss": 3.0702, "step": 19138 }, { "epoch": 1.5, "learning_rate": 8.499255077236729e-05, "loss": 2.856, "step": 19139 }, { "epoch": 1.5, "learning_rate": 8.49917666431428e-05, "loss": 2.8297, "step": 19140 }, { "epoch": 1.5, "learning_rate": 8.49909825139183e-05, "loss": 2.9105, "step": 19141 }, { "epoch": 1.5, "learning_rate": 8.49901983846938e-05, "loss": 2.6942, "step": 19142 }, { "epoch": 1.5, "learning_rate": 8.49894142554693e-05, "loss": 2.7237, "step": 19143 }, { "epoch": 1.5, "learning_rate": 8.498863012624481e-05, "loss": 3.0909, "step": 19144 }, { "epoch": 1.5, "learning_rate": 8.498784599702031e-05, "loss": 2.8476, "step": 19145 }, { "epoch": 1.5, "learning_rate": 8.498706186779582e-05, "loss": 2.8243, "step": 19146 }, { "epoch": 1.5, "learning_rate": 8.498627773857132e-05, "loss": 2.8793, "step": 19147 }, { "epoch": 1.5, "learning_rate": 8.498549360934682e-05, "loss": 2.6916, "step": 19148 }, { "epoch": 1.5, "learning_rate": 8.498470948012233e-05, "loss": 2.8069, "step": 19149 }, { "epoch": 1.5, "learning_rate": 8.498392535089783e-05, "loss": 3.0803, "step": 19150 }, { "epoch": 1.5, "learning_rate": 8.498314122167334e-05, "loss": 2.9611, "step": 19151 }, { "epoch": 1.5, "learning_rate": 8.498235709244884e-05, "loss": 2.7569, "step": 19152 }, { "epoch": 1.5, "learning_rate": 8.498157296322434e-05, "loss": 2.8419, "step": 19153 }, { "epoch": 1.5, "learning_rate": 8.498078883399985e-05, "loss": 2.6626, "step": 19154 }, { "epoch": 1.5, "learning_rate": 8.498000470477535e-05, "loss": 2.9358, "step": 19155 }, { "epoch": 1.5, "learning_rate": 8.497922057555085e-05, "loss": 2.9096, "step": 19156 }, { "epoch": 1.5, "learning_rate": 8.497843644632636e-05, "loss": 2.9065, "step": 19157 }, { "epoch": 1.5, "learning_rate": 8.497765231710186e-05, "loss": 3.2613, "step": 19158 }, { "epoch": 1.5, "learning_rate": 8.497686818787736e-05, "loss": 2.9014, "step": 19159 }, { "epoch": 1.5, "learning_rate": 8.497608405865287e-05, "loss": 2.8525, "step": 19160 }, { "epoch": 1.5, "learning_rate": 8.497529992942837e-05, "loss": 2.8381, "step": 19161 }, { "epoch": 1.5, "learning_rate": 8.497451580020388e-05, "loss": 3.1874, "step": 19162 }, { "epoch": 1.5, "learning_rate": 8.497373167097938e-05, "loss": 3.057, "step": 19163 }, { "epoch": 1.5, "learning_rate": 8.497294754175488e-05, "loss": 2.7985, "step": 19164 }, { "epoch": 1.5, "learning_rate": 8.497216341253039e-05, "loss": 2.9777, "step": 19165 }, { "epoch": 1.5, "learning_rate": 8.497137928330589e-05, "loss": 2.9514, "step": 19166 }, { "epoch": 1.5, "learning_rate": 8.49705951540814e-05, "loss": 3.1206, "step": 19167 }, { "epoch": 1.5, "learning_rate": 8.496981102485691e-05, "loss": 2.6759, "step": 19168 }, { "epoch": 1.5, "learning_rate": 8.496902689563241e-05, "loss": 2.9773, "step": 19169 }, { "epoch": 1.5, "learning_rate": 8.496824276640792e-05, "loss": 2.8876, "step": 19170 }, { "epoch": 1.5, "learning_rate": 8.496745863718342e-05, "loss": 2.779, "step": 19171 }, { "epoch": 1.5, "learning_rate": 8.496667450795893e-05, "loss": 3.055, "step": 19172 }, { "epoch": 1.5, "learning_rate": 8.496589037873442e-05, "loss": 2.8455, "step": 19173 }, { "epoch": 1.5, "learning_rate": 8.496510624950992e-05, "loss": 3.2685, "step": 19174 }, { "epoch": 1.5, "learning_rate": 8.496432212028542e-05, "loss": 2.8233, "step": 19175 }, { "epoch": 1.5, "learning_rate": 8.496353799106093e-05, "loss": 2.9398, "step": 19176 }, { "epoch": 1.5, "learning_rate": 8.496275386183643e-05, "loss": 3.0506, "step": 19177 }, { "epoch": 1.5, "learning_rate": 8.496196973261193e-05, "loss": 2.9547, "step": 19178 }, { "epoch": 1.5, "learning_rate": 8.496118560338744e-05, "loss": 2.7806, "step": 19179 }, { "epoch": 1.5, "learning_rate": 8.496040147416294e-05, "loss": 3.3835, "step": 19180 }, { "epoch": 1.5, "learning_rate": 8.495961734493845e-05, "loss": 2.6802, "step": 19181 }, { "epoch": 1.5, "learning_rate": 8.495883321571395e-05, "loss": 2.7778, "step": 19182 }, { "epoch": 1.5, "learning_rate": 8.495804908648945e-05, "loss": 2.785, "step": 19183 }, { "epoch": 1.5, "learning_rate": 8.495726495726496e-05, "loss": 2.7366, "step": 19184 }, { "epoch": 1.5, "learning_rate": 8.495648082804046e-05, "loss": 2.8733, "step": 19185 }, { "epoch": 1.5, "learning_rate": 8.495569669881596e-05, "loss": 2.7091, "step": 19186 }, { "epoch": 1.5, "learning_rate": 8.495491256959147e-05, "loss": 2.9206, "step": 19187 }, { "epoch": 1.5, "learning_rate": 8.495412844036697e-05, "loss": 2.7315, "step": 19188 }, { "epoch": 1.5, "learning_rate": 8.495334431114247e-05, "loss": 3.0831, "step": 19189 }, { "epoch": 1.5, "learning_rate": 8.495256018191798e-05, "loss": 3.1008, "step": 19190 }, { "epoch": 1.5, "learning_rate": 8.495177605269348e-05, "loss": 2.93, "step": 19191 }, { "epoch": 1.5, "learning_rate": 8.495099192346899e-05, "loss": 2.8252, "step": 19192 }, { "epoch": 1.5, "learning_rate": 8.495020779424449e-05, "loss": 2.8867, "step": 19193 }, { "epoch": 1.51, "learning_rate": 8.494942366502001e-05, "loss": 2.9872, "step": 19194 }, { "epoch": 1.51, "learning_rate": 8.494863953579551e-05, "loss": 2.8602, "step": 19195 }, { "epoch": 1.51, "learning_rate": 8.494785540657101e-05, "loss": 3.1613, "step": 19196 }, { "epoch": 1.51, "learning_rate": 8.494707127734652e-05, "loss": 2.9935, "step": 19197 }, { "epoch": 1.51, "learning_rate": 8.494628714812202e-05, "loss": 2.7982, "step": 19198 }, { "epoch": 1.51, "learning_rate": 8.494550301889753e-05, "loss": 2.6504, "step": 19199 }, { "epoch": 1.51, "learning_rate": 8.494471888967303e-05, "loss": 2.9438, "step": 19200 }, { "epoch": 1.51, "learning_rate": 8.494393476044853e-05, "loss": 2.9335, "step": 19201 }, { "epoch": 1.51, "learning_rate": 8.494315063122404e-05, "loss": 3.046, "step": 19202 }, { "epoch": 1.51, "learning_rate": 8.494236650199954e-05, "loss": 2.9675, "step": 19203 }, { "epoch": 1.51, "learning_rate": 8.494158237277504e-05, "loss": 3.1831, "step": 19204 }, { "epoch": 1.51, "learning_rate": 8.494079824355055e-05, "loss": 2.7408, "step": 19205 }, { "epoch": 1.51, "learning_rate": 8.494001411432605e-05, "loss": 2.9931, "step": 19206 }, { "epoch": 1.51, "learning_rate": 8.493922998510155e-05, "loss": 2.9268, "step": 19207 }, { "epoch": 1.51, "learning_rate": 8.493844585587706e-05, "loss": 3.182, "step": 19208 }, { "epoch": 1.51, "learning_rate": 8.493766172665255e-05, "loss": 3.0171, "step": 19209 }, { "epoch": 1.51, "learning_rate": 8.493687759742805e-05, "loss": 2.7328, "step": 19210 }, { "epoch": 1.51, "learning_rate": 8.493609346820356e-05, "loss": 2.9876, "step": 19211 }, { "epoch": 1.51, "learning_rate": 8.493530933897906e-05, "loss": 2.909, "step": 19212 }, { "epoch": 1.51, "learning_rate": 8.493452520975456e-05, "loss": 3.1632, "step": 19213 }, { "epoch": 1.51, "learning_rate": 8.493374108053007e-05, "loss": 2.8312, "step": 19214 }, { "epoch": 1.51, "learning_rate": 8.493295695130557e-05, "loss": 2.8734, "step": 19215 }, { "epoch": 1.51, "learning_rate": 8.493217282208107e-05, "loss": 2.8167, "step": 19216 }, { "epoch": 1.51, "learning_rate": 8.493138869285658e-05, "loss": 2.7349, "step": 19217 }, { "epoch": 1.51, "learning_rate": 8.493060456363208e-05, "loss": 2.9383, "step": 19218 }, { "epoch": 1.51, "learning_rate": 8.49298204344076e-05, "loss": 2.8326, "step": 19219 }, { "epoch": 1.51, "learning_rate": 8.49290363051831e-05, "loss": 2.8263, "step": 19220 }, { "epoch": 1.51, "learning_rate": 8.49282521759586e-05, "loss": 3.3007, "step": 19221 }, { "epoch": 1.51, "learning_rate": 8.492746804673411e-05, "loss": 2.9442, "step": 19222 }, { "epoch": 1.51, "learning_rate": 8.492668391750961e-05, "loss": 2.7972, "step": 19223 }, { "epoch": 1.51, "learning_rate": 8.492589978828512e-05, "loss": 2.8885, "step": 19224 }, { "epoch": 1.51, "learning_rate": 8.492511565906062e-05, "loss": 2.8454, "step": 19225 }, { "epoch": 1.51, "learning_rate": 8.492433152983612e-05, "loss": 2.6454, "step": 19226 }, { "epoch": 1.51, "learning_rate": 8.492354740061163e-05, "loss": 2.9213, "step": 19227 }, { "epoch": 1.51, "learning_rate": 8.492276327138713e-05, "loss": 3.0584, "step": 19228 }, { "epoch": 1.51, "learning_rate": 8.492197914216264e-05, "loss": 2.9553, "step": 19229 }, { "epoch": 1.51, "learning_rate": 8.492119501293814e-05, "loss": 2.7709, "step": 19230 }, { "epoch": 1.51, "learning_rate": 8.492041088371364e-05, "loss": 2.856, "step": 19231 }, { "epoch": 1.51, "learning_rate": 8.491962675448915e-05, "loss": 2.8759, "step": 19232 }, { "epoch": 1.51, "learning_rate": 8.491884262526465e-05, "loss": 2.9081, "step": 19233 }, { "epoch": 1.51, "learning_rate": 8.491805849604015e-05, "loss": 3.3048, "step": 19234 }, { "epoch": 1.51, "learning_rate": 8.491727436681566e-05, "loss": 2.7801, "step": 19235 }, { "epoch": 1.51, "learning_rate": 8.491649023759116e-05, "loss": 3.1689, "step": 19236 }, { "epoch": 1.51, "learning_rate": 8.491570610836667e-05, "loss": 2.7764, "step": 19237 }, { "epoch": 1.51, "learning_rate": 8.491492197914217e-05, "loss": 2.8976, "step": 19238 }, { "epoch": 1.51, "learning_rate": 8.491413784991767e-05, "loss": 2.9182, "step": 19239 }, { "epoch": 1.51, "learning_rate": 8.491335372069318e-05, "loss": 2.9423, "step": 19240 }, { "epoch": 1.51, "learning_rate": 8.491256959146868e-05, "loss": 2.9176, "step": 19241 }, { "epoch": 1.51, "learning_rate": 8.491178546224418e-05, "loss": 3.1758, "step": 19242 }, { "epoch": 1.51, "learning_rate": 8.491100133301969e-05, "loss": 2.7643, "step": 19243 }, { "epoch": 1.51, "learning_rate": 8.491021720379519e-05, "loss": 2.9659, "step": 19244 }, { "epoch": 1.51, "learning_rate": 8.49094330745707e-05, "loss": 2.8652, "step": 19245 }, { "epoch": 1.51, "learning_rate": 8.49086489453462e-05, "loss": 3.2848, "step": 19246 }, { "epoch": 1.51, "learning_rate": 8.49078648161217e-05, "loss": 2.8111, "step": 19247 }, { "epoch": 1.51, "learning_rate": 8.49070806868972e-05, "loss": 3.025, "step": 19248 }, { "epoch": 1.51, "learning_rate": 8.490629655767271e-05, "loss": 3.0786, "step": 19249 }, { "epoch": 1.51, "learning_rate": 8.490551242844821e-05, "loss": 2.7003, "step": 19250 }, { "epoch": 1.51, "learning_rate": 8.490472829922372e-05, "loss": 2.8172, "step": 19251 }, { "epoch": 1.51, "learning_rate": 8.490394416999922e-05, "loss": 2.9273, "step": 19252 }, { "epoch": 1.51, "learning_rate": 8.490316004077472e-05, "loss": 2.9958, "step": 19253 }, { "epoch": 1.51, "learning_rate": 8.490237591155023e-05, "loss": 2.9539, "step": 19254 }, { "epoch": 1.51, "learning_rate": 8.490159178232573e-05, "loss": 3.1306, "step": 19255 }, { "epoch": 1.51, "learning_rate": 8.490080765310123e-05, "loss": 2.9663, "step": 19256 }, { "epoch": 1.51, "learning_rate": 8.490002352387674e-05, "loss": 2.7032, "step": 19257 }, { "epoch": 1.51, "learning_rate": 8.489923939465224e-05, "loss": 2.9725, "step": 19258 }, { "epoch": 1.51, "learning_rate": 8.489845526542775e-05, "loss": 2.857, "step": 19259 }, { "epoch": 1.51, "learning_rate": 8.489767113620325e-05, "loss": 2.76, "step": 19260 }, { "epoch": 1.51, "learning_rate": 8.489688700697875e-05, "loss": 2.889, "step": 19261 }, { "epoch": 1.51, "learning_rate": 8.489610287775426e-05, "loss": 2.7889, "step": 19262 }, { "epoch": 1.51, "learning_rate": 8.489531874852976e-05, "loss": 3.1449, "step": 19263 }, { "epoch": 1.51, "learning_rate": 8.489453461930526e-05, "loss": 2.8425, "step": 19264 }, { "epoch": 1.51, "learning_rate": 8.489375049008077e-05, "loss": 2.8433, "step": 19265 }, { "epoch": 1.51, "learning_rate": 8.489296636085627e-05, "loss": 2.752, "step": 19266 }, { "epoch": 1.51, "learning_rate": 8.489218223163178e-05, "loss": 3.0837, "step": 19267 }, { "epoch": 1.51, "learning_rate": 8.489139810240728e-05, "loss": 2.7517, "step": 19268 }, { "epoch": 1.51, "learning_rate": 8.489061397318278e-05, "loss": 2.8056, "step": 19269 }, { "epoch": 1.51, "learning_rate": 8.488982984395829e-05, "loss": 2.8958, "step": 19270 }, { "epoch": 1.51, "learning_rate": 8.488904571473379e-05, "loss": 2.773, "step": 19271 }, { "epoch": 1.51, "learning_rate": 8.48882615855093e-05, "loss": 2.8133, "step": 19272 }, { "epoch": 1.51, "learning_rate": 8.48874774562848e-05, "loss": 3.2376, "step": 19273 }, { "epoch": 1.51, "learning_rate": 8.48866933270603e-05, "loss": 2.9311, "step": 19274 }, { "epoch": 1.51, "learning_rate": 8.48859091978358e-05, "loss": 2.935, "step": 19275 }, { "epoch": 1.51, "learning_rate": 8.488512506861132e-05, "loss": 2.7486, "step": 19276 }, { "epoch": 1.51, "learning_rate": 8.488434093938683e-05, "loss": 2.7416, "step": 19277 }, { "epoch": 1.51, "learning_rate": 8.488355681016233e-05, "loss": 2.6868, "step": 19278 }, { "epoch": 1.51, "learning_rate": 8.488277268093783e-05, "loss": 2.8444, "step": 19279 }, { "epoch": 1.51, "learning_rate": 8.488198855171334e-05, "loss": 3.0907, "step": 19280 }, { "epoch": 1.51, "learning_rate": 8.488120442248883e-05, "loss": 3.2418, "step": 19281 }, { "epoch": 1.51, "learning_rate": 8.488042029326433e-05, "loss": 2.6892, "step": 19282 }, { "epoch": 1.51, "learning_rate": 8.487963616403983e-05, "loss": 2.9798, "step": 19283 }, { "epoch": 1.51, "learning_rate": 8.487885203481534e-05, "loss": 2.999, "step": 19284 }, { "epoch": 1.51, "learning_rate": 8.487806790559084e-05, "loss": 3.0992, "step": 19285 }, { "epoch": 1.51, "learning_rate": 8.487728377636635e-05, "loss": 2.7102, "step": 19286 }, { "epoch": 1.51, "learning_rate": 8.487649964714185e-05, "loss": 2.8328, "step": 19287 }, { "epoch": 1.51, "learning_rate": 8.487571551791735e-05, "loss": 2.9085, "step": 19288 }, { "epoch": 1.51, "learning_rate": 8.487493138869286e-05, "loss": 2.9119, "step": 19289 }, { "epoch": 1.51, "learning_rate": 8.487414725946836e-05, "loss": 3.0556, "step": 19290 }, { "epoch": 1.51, "learning_rate": 8.487336313024386e-05, "loss": 2.7145, "step": 19291 }, { "epoch": 1.51, "learning_rate": 8.487257900101937e-05, "loss": 2.9571, "step": 19292 }, { "epoch": 1.51, "learning_rate": 8.487179487179487e-05, "loss": 3.0617, "step": 19293 }, { "epoch": 1.51, "learning_rate": 8.487101074257037e-05, "loss": 2.9414, "step": 19294 }, { "epoch": 1.51, "learning_rate": 8.487022661334588e-05, "loss": 2.9213, "step": 19295 }, { "epoch": 1.51, "learning_rate": 8.486944248412138e-05, "loss": 2.9959, "step": 19296 }, { "epoch": 1.51, "learning_rate": 8.486865835489689e-05, "loss": 3.3165, "step": 19297 }, { "epoch": 1.51, "learning_rate": 8.486787422567239e-05, "loss": 2.7804, "step": 19298 }, { "epoch": 1.51, "learning_rate": 8.486709009644789e-05, "loss": 2.8212, "step": 19299 }, { "epoch": 1.51, "learning_rate": 8.48663059672234e-05, "loss": 2.8351, "step": 19300 }, { "epoch": 1.51, "learning_rate": 8.48655218379989e-05, "loss": 2.9582, "step": 19301 }, { "epoch": 1.51, "learning_rate": 8.486473770877442e-05, "loss": 3.0266, "step": 19302 }, { "epoch": 1.51, "learning_rate": 8.486395357954992e-05, "loss": 2.8363, "step": 19303 }, { "epoch": 1.51, "learning_rate": 8.486316945032542e-05, "loss": 2.9373, "step": 19304 }, { "epoch": 1.51, "learning_rate": 8.486238532110093e-05, "loss": 2.9831, "step": 19305 }, { "epoch": 1.51, "learning_rate": 8.486160119187643e-05, "loss": 2.8048, "step": 19306 }, { "epoch": 1.51, "learning_rate": 8.486081706265194e-05, "loss": 3.2205, "step": 19307 }, { "epoch": 1.51, "learning_rate": 8.486003293342744e-05, "loss": 3.1977, "step": 19308 }, { "epoch": 1.51, "learning_rate": 8.485924880420294e-05, "loss": 3.0241, "step": 19309 }, { "epoch": 1.51, "learning_rate": 8.485846467497845e-05, "loss": 2.7624, "step": 19310 }, { "epoch": 1.51, "learning_rate": 8.485768054575395e-05, "loss": 2.8335, "step": 19311 }, { "epoch": 1.51, "learning_rate": 8.485689641652945e-05, "loss": 2.9645, "step": 19312 }, { "epoch": 1.51, "learning_rate": 8.485611228730496e-05, "loss": 3.0374, "step": 19313 }, { "epoch": 1.51, "learning_rate": 8.485532815808046e-05, "loss": 2.9913, "step": 19314 }, { "epoch": 1.51, "learning_rate": 8.485454402885597e-05, "loss": 3.0288, "step": 19315 }, { "epoch": 1.51, "learning_rate": 8.485375989963147e-05, "loss": 2.8211, "step": 19316 }, { "epoch": 1.51, "learning_rate": 8.485297577040696e-05, "loss": 2.8842, "step": 19317 }, { "epoch": 1.51, "learning_rate": 8.485219164118246e-05, "loss": 2.8467, "step": 19318 }, { "epoch": 1.51, "learning_rate": 8.485140751195797e-05, "loss": 2.9849, "step": 19319 }, { "epoch": 1.51, "learning_rate": 8.485062338273347e-05, "loss": 2.9932, "step": 19320 }, { "epoch": 1.52, "learning_rate": 8.484983925350897e-05, "loss": 2.647, "step": 19321 }, { "epoch": 1.52, "learning_rate": 8.484905512428448e-05, "loss": 2.903, "step": 19322 }, { "epoch": 1.52, "learning_rate": 8.484827099505998e-05, "loss": 3.0008, "step": 19323 }, { "epoch": 1.52, "learning_rate": 8.484748686583548e-05, "loss": 2.7364, "step": 19324 }, { "epoch": 1.52, "learning_rate": 8.484670273661099e-05, "loss": 2.9656, "step": 19325 }, { "epoch": 1.52, "learning_rate": 8.484591860738649e-05, "loss": 2.7841, "step": 19326 }, { "epoch": 1.52, "learning_rate": 8.484513447816201e-05, "loss": 2.9609, "step": 19327 }, { "epoch": 1.52, "learning_rate": 8.484435034893751e-05, "loss": 2.9568, "step": 19328 }, { "epoch": 1.52, "learning_rate": 8.484356621971302e-05, "loss": 2.7331, "step": 19329 }, { "epoch": 1.52, "learning_rate": 8.484278209048852e-05, "loss": 2.9448, "step": 19330 }, { "epoch": 1.52, "learning_rate": 8.484199796126402e-05, "loss": 2.9265, "step": 19331 }, { "epoch": 1.52, "learning_rate": 8.484121383203953e-05, "loss": 2.9826, "step": 19332 }, { "epoch": 1.52, "learning_rate": 8.484042970281503e-05, "loss": 2.9145, "step": 19333 }, { "epoch": 1.52, "learning_rate": 8.483964557359054e-05, "loss": 2.8768, "step": 19334 }, { "epoch": 1.52, "learning_rate": 8.483886144436604e-05, "loss": 2.6329, "step": 19335 }, { "epoch": 1.52, "learning_rate": 8.483807731514154e-05, "loss": 2.8696, "step": 19336 }, { "epoch": 1.52, "learning_rate": 8.483729318591705e-05, "loss": 2.8539, "step": 19337 }, { "epoch": 1.52, "learning_rate": 8.483650905669255e-05, "loss": 2.9184, "step": 19338 }, { "epoch": 1.52, "learning_rate": 8.483572492746805e-05, "loss": 3.1448, "step": 19339 }, { "epoch": 1.52, "learning_rate": 8.483494079824356e-05, "loss": 2.7962, "step": 19340 }, { "epoch": 1.52, "learning_rate": 8.483415666901906e-05, "loss": 3.0353, "step": 19341 }, { "epoch": 1.52, "learning_rate": 8.483337253979456e-05, "loss": 2.9509, "step": 19342 }, { "epoch": 1.52, "learning_rate": 8.483258841057007e-05, "loss": 3.1101, "step": 19343 }, { "epoch": 1.52, "learning_rate": 8.483180428134557e-05, "loss": 2.7861, "step": 19344 }, { "epoch": 1.52, "learning_rate": 8.483102015212108e-05, "loss": 2.9676, "step": 19345 }, { "epoch": 1.52, "learning_rate": 8.483023602289658e-05, "loss": 2.997, "step": 19346 }, { "epoch": 1.52, "learning_rate": 8.482945189367208e-05, "loss": 3.0138, "step": 19347 }, { "epoch": 1.52, "learning_rate": 8.482866776444759e-05, "loss": 2.8634, "step": 19348 }, { "epoch": 1.52, "learning_rate": 8.482788363522309e-05, "loss": 2.8725, "step": 19349 }, { "epoch": 1.52, "learning_rate": 8.48270995059986e-05, "loss": 2.7365, "step": 19350 }, { "epoch": 1.52, "learning_rate": 8.48263153767741e-05, "loss": 2.8436, "step": 19351 }, { "epoch": 1.52, "learning_rate": 8.48255312475496e-05, "loss": 2.7954, "step": 19352 }, { "epoch": 1.52, "learning_rate": 8.48247471183251e-05, "loss": 3.1191, "step": 19353 }, { "epoch": 1.52, "learning_rate": 8.482396298910061e-05, "loss": 2.7039, "step": 19354 }, { "epoch": 1.52, "learning_rate": 8.482317885987611e-05, "loss": 2.9627, "step": 19355 }, { "epoch": 1.52, "learning_rate": 8.482239473065162e-05, "loss": 2.7707, "step": 19356 }, { "epoch": 1.52, "learning_rate": 8.482161060142712e-05, "loss": 2.9138, "step": 19357 }, { "epoch": 1.52, "learning_rate": 8.482082647220262e-05, "loss": 2.9697, "step": 19358 }, { "epoch": 1.52, "learning_rate": 8.482004234297813e-05, "loss": 2.8785, "step": 19359 }, { "epoch": 1.52, "learning_rate": 8.481925821375363e-05, "loss": 3.2361, "step": 19360 }, { "epoch": 1.52, "learning_rate": 8.481847408452913e-05, "loss": 2.9256, "step": 19361 }, { "epoch": 1.52, "learning_rate": 8.481768995530464e-05, "loss": 3.1576, "step": 19362 }, { "epoch": 1.52, "learning_rate": 8.481690582608014e-05, "loss": 2.7531, "step": 19363 }, { "epoch": 1.52, "learning_rate": 8.481612169685565e-05, "loss": 2.7484, "step": 19364 }, { "epoch": 1.52, "learning_rate": 8.481533756763115e-05, "loss": 2.7573, "step": 19365 }, { "epoch": 1.52, "learning_rate": 8.481455343840665e-05, "loss": 2.9632, "step": 19366 }, { "epoch": 1.52, "learning_rate": 8.481376930918216e-05, "loss": 2.8713, "step": 19367 }, { "epoch": 1.52, "learning_rate": 8.481298517995766e-05, "loss": 2.8943, "step": 19368 }, { "epoch": 1.52, "learning_rate": 8.481220105073316e-05, "loss": 2.6579, "step": 19369 }, { "epoch": 1.52, "learning_rate": 8.481141692150867e-05, "loss": 2.8898, "step": 19370 }, { "epoch": 1.52, "learning_rate": 8.481063279228417e-05, "loss": 2.9539, "step": 19371 }, { "epoch": 1.52, "learning_rate": 8.480984866305968e-05, "loss": 2.8744, "step": 19372 }, { "epoch": 1.52, "learning_rate": 8.480906453383518e-05, "loss": 3.1788, "step": 19373 }, { "epoch": 1.52, "learning_rate": 8.480828040461068e-05, "loss": 3.0166, "step": 19374 }, { "epoch": 1.52, "learning_rate": 8.480749627538619e-05, "loss": 2.9526, "step": 19375 }, { "epoch": 1.52, "learning_rate": 8.480671214616169e-05, "loss": 3.0088, "step": 19376 }, { "epoch": 1.52, "learning_rate": 8.48059280169372e-05, "loss": 2.9248, "step": 19377 }, { "epoch": 1.52, "learning_rate": 8.48051438877127e-05, "loss": 2.6031, "step": 19378 }, { "epoch": 1.52, "learning_rate": 8.48043597584882e-05, "loss": 2.8892, "step": 19379 }, { "epoch": 1.52, "learning_rate": 8.48035756292637e-05, "loss": 2.8595, "step": 19380 }, { "epoch": 1.52, "learning_rate": 8.480279150003921e-05, "loss": 2.9779, "step": 19381 }, { "epoch": 1.52, "learning_rate": 8.480200737081471e-05, "loss": 2.8039, "step": 19382 }, { "epoch": 1.52, "learning_rate": 8.480122324159022e-05, "loss": 2.9449, "step": 19383 }, { "epoch": 1.52, "learning_rate": 8.480043911236573e-05, "loss": 3.2152, "step": 19384 }, { "epoch": 1.52, "learning_rate": 8.479965498314124e-05, "loss": 2.8469, "step": 19385 }, { "epoch": 1.52, "learning_rate": 8.479887085391674e-05, "loss": 2.933, "step": 19386 }, { "epoch": 1.52, "learning_rate": 8.479808672469224e-05, "loss": 2.929, "step": 19387 }, { "epoch": 1.52, "learning_rate": 8.479730259546775e-05, "loss": 3.0376, "step": 19388 }, { "epoch": 1.52, "learning_rate": 8.479651846624324e-05, "loss": 2.829, "step": 19389 }, { "epoch": 1.52, "learning_rate": 8.479573433701874e-05, "loss": 2.8763, "step": 19390 }, { "epoch": 1.52, "learning_rate": 8.479495020779424e-05, "loss": 3.1418, "step": 19391 }, { "epoch": 1.52, "learning_rate": 8.479416607856975e-05, "loss": 2.8832, "step": 19392 }, { "epoch": 1.52, "learning_rate": 8.479338194934525e-05, "loss": 2.8684, "step": 19393 }, { "epoch": 1.52, "learning_rate": 8.479259782012076e-05, "loss": 3.0777, "step": 19394 }, { "epoch": 1.52, "learning_rate": 8.479181369089626e-05, "loss": 2.7906, "step": 19395 }, { "epoch": 1.52, "learning_rate": 8.479102956167176e-05, "loss": 2.8701, "step": 19396 }, { "epoch": 1.52, "learning_rate": 8.479024543244727e-05, "loss": 3.0434, "step": 19397 }, { "epoch": 1.52, "learning_rate": 8.478946130322277e-05, "loss": 2.9599, "step": 19398 }, { "epoch": 1.52, "learning_rate": 8.478867717399827e-05, "loss": 3.0505, "step": 19399 }, { "epoch": 1.52, "learning_rate": 8.478789304477378e-05, "loss": 2.762, "step": 19400 }, { "epoch": 1.52, "learning_rate": 8.478710891554928e-05, "loss": 2.7662, "step": 19401 }, { "epoch": 1.52, "learning_rate": 8.478632478632479e-05, "loss": 2.6835, "step": 19402 }, { "epoch": 1.52, "learning_rate": 8.478554065710029e-05, "loss": 2.9991, "step": 19403 }, { "epoch": 1.52, "learning_rate": 8.478475652787579e-05, "loss": 2.9702, "step": 19404 }, { "epoch": 1.52, "learning_rate": 8.47839723986513e-05, "loss": 2.9011, "step": 19405 }, { "epoch": 1.52, "learning_rate": 8.47831882694268e-05, "loss": 3.0933, "step": 19406 }, { "epoch": 1.52, "learning_rate": 8.47824041402023e-05, "loss": 3.1553, "step": 19407 }, { "epoch": 1.52, "learning_rate": 8.478162001097781e-05, "loss": 2.7386, "step": 19408 }, { "epoch": 1.52, "learning_rate": 8.478083588175331e-05, "loss": 2.76, "step": 19409 }, { "epoch": 1.52, "learning_rate": 8.478005175252883e-05, "loss": 3.0474, "step": 19410 }, { "epoch": 1.52, "learning_rate": 8.477926762330433e-05, "loss": 2.7988, "step": 19411 }, { "epoch": 1.52, "learning_rate": 8.477848349407984e-05, "loss": 2.8525, "step": 19412 }, { "epoch": 1.52, "learning_rate": 8.477769936485534e-05, "loss": 3.1183, "step": 19413 }, { "epoch": 1.52, "learning_rate": 8.477691523563084e-05, "loss": 2.8438, "step": 19414 }, { "epoch": 1.52, "learning_rate": 8.477613110640635e-05, "loss": 3.0856, "step": 19415 }, { "epoch": 1.52, "learning_rate": 8.477534697718185e-05, "loss": 2.8212, "step": 19416 }, { "epoch": 1.52, "learning_rate": 8.477456284795735e-05, "loss": 2.7601, "step": 19417 }, { "epoch": 1.52, "learning_rate": 8.477377871873286e-05, "loss": 2.7414, "step": 19418 }, { "epoch": 1.52, "learning_rate": 8.477299458950836e-05, "loss": 2.9194, "step": 19419 }, { "epoch": 1.52, "learning_rate": 8.477221046028387e-05, "loss": 2.9537, "step": 19420 }, { "epoch": 1.52, "learning_rate": 8.477142633105937e-05, "loss": 2.9776, "step": 19421 }, { "epoch": 1.52, "learning_rate": 8.477064220183487e-05, "loss": 3.1967, "step": 19422 }, { "epoch": 1.52, "learning_rate": 8.476985807261038e-05, "loss": 2.8176, "step": 19423 }, { "epoch": 1.52, "learning_rate": 8.476907394338588e-05, "loss": 2.6923, "step": 19424 }, { "epoch": 1.52, "learning_rate": 8.476828981416137e-05, "loss": 2.8752, "step": 19425 }, { "epoch": 1.52, "learning_rate": 8.476750568493687e-05, "loss": 2.8582, "step": 19426 }, { "epoch": 1.52, "learning_rate": 8.476672155571238e-05, "loss": 3.4847, "step": 19427 }, { "epoch": 1.52, "learning_rate": 8.476593742648788e-05, "loss": 2.9333, "step": 19428 }, { "epoch": 1.52, "learning_rate": 8.476515329726338e-05, "loss": 3.0405, "step": 19429 }, { "epoch": 1.52, "learning_rate": 8.476436916803889e-05, "loss": 2.7798, "step": 19430 }, { "epoch": 1.52, "learning_rate": 8.476358503881439e-05, "loss": 2.9606, "step": 19431 }, { "epoch": 1.52, "learning_rate": 8.47628009095899e-05, "loss": 2.9457, "step": 19432 }, { "epoch": 1.52, "learning_rate": 8.47620167803654e-05, "loss": 2.9906, "step": 19433 }, { "epoch": 1.52, "learning_rate": 8.47612326511409e-05, "loss": 2.8534, "step": 19434 }, { "epoch": 1.52, "learning_rate": 8.47604485219164e-05, "loss": 3.0014, "step": 19435 }, { "epoch": 1.52, "learning_rate": 8.475966439269192e-05, "loss": 2.7999, "step": 19436 }, { "epoch": 1.52, "learning_rate": 8.475888026346743e-05, "loss": 2.8914, "step": 19437 }, { "epoch": 1.52, "learning_rate": 8.475809613424293e-05, "loss": 2.9948, "step": 19438 }, { "epoch": 1.52, "learning_rate": 8.475731200501844e-05, "loss": 2.7254, "step": 19439 }, { "epoch": 1.52, "learning_rate": 8.475652787579394e-05, "loss": 2.7534, "step": 19440 }, { "epoch": 1.52, "learning_rate": 8.475574374656944e-05, "loss": 2.8561, "step": 19441 }, { "epoch": 1.52, "learning_rate": 8.475495961734495e-05, "loss": 2.8721, "step": 19442 }, { "epoch": 1.52, "learning_rate": 8.475417548812045e-05, "loss": 3.2443, "step": 19443 }, { "epoch": 1.52, "learning_rate": 8.475339135889595e-05, "loss": 2.8579, "step": 19444 }, { "epoch": 1.52, "learning_rate": 8.475260722967146e-05, "loss": 2.946, "step": 19445 }, { "epoch": 1.52, "learning_rate": 8.475182310044696e-05, "loss": 2.7269, "step": 19446 }, { "epoch": 1.52, "learning_rate": 8.475103897122246e-05, "loss": 2.8904, "step": 19447 }, { "epoch": 1.52, "learning_rate": 8.475025484199797e-05, "loss": 2.9423, "step": 19448 }, { "epoch": 1.53, "learning_rate": 8.474947071277347e-05, "loss": 2.8309, "step": 19449 }, { "epoch": 1.53, "learning_rate": 8.474868658354898e-05, "loss": 2.5515, "step": 19450 }, { "epoch": 1.53, "learning_rate": 8.474790245432448e-05, "loss": 2.9224, "step": 19451 }, { "epoch": 1.53, "learning_rate": 8.474711832509998e-05, "loss": 3.043, "step": 19452 }, { "epoch": 1.53, "learning_rate": 8.474633419587549e-05, "loss": 2.9667, "step": 19453 }, { "epoch": 1.53, "learning_rate": 8.474555006665099e-05, "loss": 2.744, "step": 19454 }, { "epoch": 1.53, "learning_rate": 8.47447659374265e-05, "loss": 2.9046, "step": 19455 }, { "epoch": 1.53, "learning_rate": 8.4743981808202e-05, "loss": 3.1026, "step": 19456 }, { "epoch": 1.53, "learning_rate": 8.47431976789775e-05, "loss": 2.7256, "step": 19457 }, { "epoch": 1.53, "learning_rate": 8.4742413549753e-05, "loss": 2.7292, "step": 19458 }, { "epoch": 1.53, "learning_rate": 8.474162942052851e-05, "loss": 2.7814, "step": 19459 }, { "epoch": 1.53, "learning_rate": 8.474084529130401e-05, "loss": 2.7752, "step": 19460 }, { "epoch": 1.53, "learning_rate": 8.474006116207952e-05, "loss": 2.9737, "step": 19461 }, { "epoch": 1.53, "learning_rate": 8.473927703285502e-05, "loss": 2.9614, "step": 19462 }, { "epoch": 1.53, "learning_rate": 8.473849290363052e-05, "loss": 2.7925, "step": 19463 }, { "epoch": 1.53, "learning_rate": 8.473770877440603e-05, "loss": 2.8095, "step": 19464 }, { "epoch": 1.53, "learning_rate": 8.473692464518153e-05, "loss": 2.8562, "step": 19465 }, { "epoch": 1.53, "learning_rate": 8.473614051595703e-05, "loss": 3.1862, "step": 19466 }, { "epoch": 1.53, "learning_rate": 8.473535638673254e-05, "loss": 3.031, "step": 19467 }, { "epoch": 1.53, "learning_rate": 8.473457225750804e-05, "loss": 2.7807, "step": 19468 }, { "epoch": 1.53, "learning_rate": 8.473378812828355e-05, "loss": 2.9665, "step": 19469 }, { "epoch": 1.53, "learning_rate": 8.473300399905905e-05, "loss": 2.8025, "step": 19470 }, { "epoch": 1.53, "learning_rate": 8.473221986983455e-05, "loss": 2.7517, "step": 19471 }, { "epoch": 1.53, "learning_rate": 8.473143574061006e-05, "loss": 2.7549, "step": 19472 }, { "epoch": 1.53, "learning_rate": 8.473065161138556e-05, "loss": 3.0999, "step": 19473 }, { "epoch": 1.53, "learning_rate": 8.472986748216106e-05, "loss": 2.9398, "step": 19474 }, { "epoch": 1.53, "learning_rate": 8.472908335293657e-05, "loss": 2.9303, "step": 19475 }, { "epoch": 1.53, "learning_rate": 8.472829922371207e-05, "loss": 2.9459, "step": 19476 }, { "epoch": 1.53, "learning_rate": 8.472751509448757e-05, "loss": 2.8726, "step": 19477 }, { "epoch": 1.53, "learning_rate": 8.472673096526308e-05, "loss": 2.771, "step": 19478 }, { "epoch": 1.53, "learning_rate": 8.472594683603858e-05, "loss": 2.7642, "step": 19479 }, { "epoch": 1.53, "learning_rate": 8.472516270681409e-05, "loss": 2.7983, "step": 19480 }, { "epoch": 1.53, "learning_rate": 8.472437857758959e-05, "loss": 3.197, "step": 19481 }, { "epoch": 1.53, "learning_rate": 8.472359444836509e-05, "loss": 2.8776, "step": 19482 }, { "epoch": 1.53, "learning_rate": 8.47228103191406e-05, "loss": 2.7594, "step": 19483 }, { "epoch": 1.53, "learning_rate": 8.47220261899161e-05, "loss": 3.0816, "step": 19484 }, { "epoch": 1.53, "learning_rate": 8.47212420606916e-05, "loss": 2.8234, "step": 19485 }, { "epoch": 1.53, "learning_rate": 8.472045793146711e-05, "loss": 2.9249, "step": 19486 }, { "epoch": 1.53, "learning_rate": 8.471967380224261e-05, "loss": 2.9549, "step": 19487 }, { "epoch": 1.53, "learning_rate": 8.471888967301812e-05, "loss": 3.1186, "step": 19488 }, { "epoch": 1.53, "learning_rate": 8.471810554379362e-05, "loss": 2.927, "step": 19489 }, { "epoch": 1.53, "learning_rate": 8.471732141456912e-05, "loss": 2.9916, "step": 19490 }, { "epoch": 1.53, "learning_rate": 8.471653728534463e-05, "loss": 2.9021, "step": 19491 }, { "epoch": 1.53, "learning_rate": 8.471575315612014e-05, "loss": 2.883, "step": 19492 }, { "epoch": 1.53, "learning_rate": 8.471496902689565e-05, "loss": 2.8619, "step": 19493 }, { "epoch": 1.53, "learning_rate": 8.471418489767115e-05, "loss": 3.0174, "step": 19494 }, { "epoch": 1.53, "learning_rate": 8.471340076844665e-05, "loss": 2.8925, "step": 19495 }, { "epoch": 1.53, "learning_rate": 8.471261663922216e-05, "loss": 2.9689, "step": 19496 }, { "epoch": 1.53, "learning_rate": 8.471183250999765e-05, "loss": 2.7565, "step": 19497 }, { "epoch": 1.53, "learning_rate": 8.471104838077315e-05, "loss": 2.9367, "step": 19498 }, { "epoch": 1.53, "learning_rate": 8.471026425154866e-05, "loss": 2.758, "step": 19499 }, { "epoch": 1.53, "learning_rate": 8.470948012232416e-05, "loss": 2.8874, "step": 19500 }, { "epoch": 1.53, "learning_rate": 8.470869599309966e-05, "loss": 2.8699, "step": 19501 }, { "epoch": 1.53, "learning_rate": 8.470791186387517e-05, "loss": 3.0555, "step": 19502 }, { "epoch": 1.53, "learning_rate": 8.470712773465067e-05, "loss": 3.0427, "step": 19503 }, { "epoch": 1.53, "learning_rate": 8.470634360542617e-05, "loss": 3.0677, "step": 19504 }, { "epoch": 1.53, "learning_rate": 8.470555947620168e-05, "loss": 3.0882, "step": 19505 }, { "epoch": 1.53, "learning_rate": 8.470477534697718e-05, "loss": 2.6407, "step": 19506 }, { "epoch": 1.53, "learning_rate": 8.470399121775269e-05, "loss": 2.7808, "step": 19507 }, { "epoch": 1.53, "learning_rate": 8.470320708852819e-05, "loss": 2.9512, "step": 19508 }, { "epoch": 1.53, "learning_rate": 8.470242295930369e-05, "loss": 3.1323, "step": 19509 }, { "epoch": 1.53, "learning_rate": 8.47016388300792e-05, "loss": 3.0571, "step": 19510 }, { "epoch": 1.53, "learning_rate": 8.47008547008547e-05, "loss": 2.992, "step": 19511 }, { "epoch": 1.53, "learning_rate": 8.47000705716302e-05, "loss": 2.9077, "step": 19512 }, { "epoch": 1.53, "learning_rate": 8.469928644240571e-05, "loss": 2.738, "step": 19513 }, { "epoch": 1.53, "learning_rate": 8.469850231318121e-05, "loss": 2.7785, "step": 19514 }, { "epoch": 1.53, "learning_rate": 8.469771818395671e-05, "loss": 2.7219, "step": 19515 }, { "epoch": 1.53, "learning_rate": 8.469693405473222e-05, "loss": 2.7645, "step": 19516 }, { "epoch": 1.53, "learning_rate": 8.469614992550772e-05, "loss": 2.7122, "step": 19517 }, { "epoch": 1.53, "learning_rate": 8.469536579628324e-05, "loss": 2.7494, "step": 19518 }, { "epoch": 1.53, "learning_rate": 8.469458166705874e-05, "loss": 2.8418, "step": 19519 }, { "epoch": 1.53, "learning_rate": 8.469379753783425e-05, "loss": 3.0052, "step": 19520 }, { "epoch": 1.53, "learning_rate": 8.469301340860975e-05, "loss": 2.9747, "step": 19521 }, { "epoch": 1.53, "learning_rate": 8.469222927938525e-05, "loss": 2.8988, "step": 19522 }, { "epoch": 1.53, "learning_rate": 8.469144515016076e-05, "loss": 2.8952, "step": 19523 }, { "epoch": 1.53, "learning_rate": 8.469066102093626e-05, "loss": 2.9656, "step": 19524 }, { "epoch": 1.53, "learning_rate": 8.468987689171176e-05, "loss": 2.8568, "step": 19525 }, { "epoch": 1.53, "learning_rate": 8.468909276248727e-05, "loss": 2.7645, "step": 19526 }, { "epoch": 1.53, "learning_rate": 8.468830863326277e-05, "loss": 2.8051, "step": 19527 }, { "epoch": 1.53, "learning_rate": 8.468752450403828e-05, "loss": 2.9144, "step": 19528 }, { "epoch": 1.53, "learning_rate": 8.468674037481378e-05, "loss": 2.7912, "step": 19529 }, { "epoch": 1.53, "learning_rate": 8.468595624558928e-05, "loss": 2.9941, "step": 19530 }, { "epoch": 1.53, "learning_rate": 8.468517211636479e-05, "loss": 2.7379, "step": 19531 }, { "epoch": 1.53, "learning_rate": 8.468438798714029e-05, "loss": 3.0778, "step": 19532 }, { "epoch": 1.53, "learning_rate": 8.468360385791578e-05, "loss": 3.0152, "step": 19533 }, { "epoch": 1.53, "learning_rate": 8.468281972869128e-05, "loss": 3.0359, "step": 19534 }, { "epoch": 1.53, "learning_rate": 8.468203559946679e-05, "loss": 3.1325, "step": 19535 }, { "epoch": 1.53, "learning_rate": 8.468125147024229e-05, "loss": 2.6766, "step": 19536 }, { "epoch": 1.53, "learning_rate": 8.46804673410178e-05, "loss": 2.9621, "step": 19537 }, { "epoch": 1.53, "learning_rate": 8.46796832117933e-05, "loss": 2.7515, "step": 19538 }, { "epoch": 1.53, "learning_rate": 8.46788990825688e-05, "loss": 2.6447, "step": 19539 }, { "epoch": 1.53, "learning_rate": 8.46781149533443e-05, "loss": 2.9509, "step": 19540 }, { "epoch": 1.53, "learning_rate": 8.467733082411981e-05, "loss": 2.8403, "step": 19541 }, { "epoch": 1.53, "learning_rate": 8.467654669489531e-05, "loss": 2.9319, "step": 19542 }, { "epoch": 1.53, "learning_rate": 8.467576256567082e-05, "loss": 3.1017, "step": 19543 }, { "epoch": 1.53, "learning_rate": 8.467497843644633e-05, "loss": 3.001, "step": 19544 }, { "epoch": 1.53, "learning_rate": 8.467419430722184e-05, "loss": 3.0688, "step": 19545 }, { "epoch": 1.53, "learning_rate": 8.467341017799734e-05, "loss": 3.1121, "step": 19546 }, { "epoch": 1.53, "learning_rate": 8.467262604877285e-05, "loss": 3.0501, "step": 19547 }, { "epoch": 1.53, "learning_rate": 8.467184191954835e-05, "loss": 2.8067, "step": 19548 }, { "epoch": 1.53, "learning_rate": 8.467105779032385e-05, "loss": 3.0388, "step": 19549 }, { "epoch": 1.53, "learning_rate": 8.467027366109936e-05, "loss": 2.8878, "step": 19550 }, { "epoch": 1.53, "learning_rate": 8.466948953187486e-05, "loss": 2.8396, "step": 19551 }, { "epoch": 1.53, "learning_rate": 8.466870540265036e-05, "loss": 2.7301, "step": 19552 }, { "epoch": 1.53, "learning_rate": 8.466792127342587e-05, "loss": 2.8008, "step": 19553 }, { "epoch": 1.53, "learning_rate": 8.466713714420137e-05, "loss": 2.9338, "step": 19554 }, { "epoch": 1.53, "learning_rate": 8.466635301497688e-05, "loss": 2.8386, "step": 19555 }, { "epoch": 1.53, "learning_rate": 8.466556888575238e-05, "loss": 2.7453, "step": 19556 }, { "epoch": 1.53, "learning_rate": 8.466478475652788e-05, "loss": 2.6764, "step": 19557 }, { "epoch": 1.53, "learning_rate": 8.466400062730339e-05, "loss": 2.8887, "step": 19558 }, { "epoch": 1.53, "learning_rate": 8.466321649807889e-05, "loss": 2.8677, "step": 19559 }, { "epoch": 1.53, "learning_rate": 8.46624323688544e-05, "loss": 2.8863, "step": 19560 }, { "epoch": 1.53, "learning_rate": 8.46616482396299e-05, "loss": 2.7505, "step": 19561 }, { "epoch": 1.53, "learning_rate": 8.46608641104054e-05, "loss": 2.9405, "step": 19562 }, { "epoch": 1.53, "learning_rate": 8.46600799811809e-05, "loss": 2.923, "step": 19563 }, { "epoch": 1.53, "learning_rate": 8.465929585195641e-05, "loss": 2.8495, "step": 19564 }, { "epoch": 1.53, "learning_rate": 8.465851172273191e-05, "loss": 2.8468, "step": 19565 }, { "epoch": 1.53, "learning_rate": 8.465772759350742e-05, "loss": 3.0255, "step": 19566 }, { "epoch": 1.53, "learning_rate": 8.465694346428292e-05, "loss": 2.8791, "step": 19567 }, { "epoch": 1.53, "learning_rate": 8.465615933505842e-05, "loss": 2.7877, "step": 19568 }, { "epoch": 1.53, "learning_rate": 8.465537520583393e-05, "loss": 2.8604, "step": 19569 }, { "epoch": 1.53, "learning_rate": 8.465459107660943e-05, "loss": 2.9288, "step": 19570 }, { "epoch": 1.53, "learning_rate": 8.465380694738493e-05, "loss": 2.9804, "step": 19571 }, { "epoch": 1.53, "learning_rate": 8.465302281816044e-05, "loss": 3.3108, "step": 19572 }, { "epoch": 1.53, "learning_rate": 8.465223868893594e-05, "loss": 2.8495, "step": 19573 }, { "epoch": 1.53, "learning_rate": 8.465145455971145e-05, "loss": 2.6449, "step": 19574 }, { "epoch": 1.53, "learning_rate": 8.465067043048695e-05, "loss": 2.9361, "step": 19575 }, { "epoch": 1.54, "learning_rate": 8.464988630126245e-05, "loss": 3.1321, "step": 19576 }, { "epoch": 1.54, "learning_rate": 8.464910217203796e-05, "loss": 2.7325, "step": 19577 }, { "epoch": 1.54, "learning_rate": 8.464831804281346e-05, "loss": 2.9269, "step": 19578 }, { "epoch": 1.54, "learning_rate": 8.464753391358896e-05, "loss": 3.1617, "step": 19579 }, { "epoch": 1.54, "learning_rate": 8.464674978436447e-05, "loss": 2.8667, "step": 19580 }, { "epoch": 1.54, "learning_rate": 8.464596565513997e-05, "loss": 2.6813, "step": 19581 }, { "epoch": 1.54, "learning_rate": 8.464518152591547e-05, "loss": 2.99, "step": 19582 }, { "epoch": 1.54, "learning_rate": 8.464439739669098e-05, "loss": 3.2026, "step": 19583 }, { "epoch": 1.54, "learning_rate": 8.464361326746648e-05, "loss": 3.2598, "step": 19584 }, { "epoch": 1.54, "learning_rate": 8.464282913824199e-05, "loss": 2.836, "step": 19585 }, { "epoch": 1.54, "learning_rate": 8.464204500901749e-05, "loss": 3.0915, "step": 19586 }, { "epoch": 1.54, "learning_rate": 8.464126087979299e-05, "loss": 2.7824, "step": 19587 }, { "epoch": 1.54, "learning_rate": 8.46404767505685e-05, "loss": 2.8797, "step": 19588 }, { "epoch": 1.54, "learning_rate": 8.4639692621344e-05, "loss": 2.8344, "step": 19589 }, { "epoch": 1.54, "learning_rate": 8.46389084921195e-05, "loss": 2.8771, "step": 19590 }, { "epoch": 1.54, "learning_rate": 8.463812436289501e-05, "loss": 3.1237, "step": 19591 }, { "epoch": 1.54, "learning_rate": 8.463734023367051e-05, "loss": 3.2297, "step": 19592 }, { "epoch": 1.54, "learning_rate": 8.463655610444601e-05, "loss": 2.9386, "step": 19593 }, { "epoch": 1.54, "learning_rate": 8.463577197522152e-05, "loss": 2.7927, "step": 19594 }, { "epoch": 1.54, "learning_rate": 8.463498784599702e-05, "loss": 3.0256, "step": 19595 }, { "epoch": 1.54, "learning_rate": 8.463420371677253e-05, "loss": 2.7927, "step": 19596 }, { "epoch": 1.54, "learning_rate": 8.463341958754803e-05, "loss": 2.9144, "step": 19597 }, { "epoch": 1.54, "learning_rate": 8.463263545832353e-05, "loss": 2.9809, "step": 19598 }, { "epoch": 1.54, "learning_rate": 8.463185132909904e-05, "loss": 2.8414, "step": 19599 }, { "epoch": 1.54, "learning_rate": 8.463106719987455e-05, "loss": 2.8597, "step": 19600 }, { "epoch": 1.54, "learning_rate": 8.463028307065006e-05, "loss": 2.8576, "step": 19601 }, { "epoch": 1.54, "learning_rate": 8.462949894142556e-05, "loss": 2.7976, "step": 19602 }, { "epoch": 1.54, "learning_rate": 8.462871481220107e-05, "loss": 3.0662, "step": 19603 }, { "epoch": 1.54, "learning_rate": 8.462793068297657e-05, "loss": 3.1117, "step": 19604 }, { "epoch": 1.54, "learning_rate": 8.462714655375206e-05, "loss": 2.7399, "step": 19605 }, { "epoch": 1.54, "learning_rate": 8.462636242452756e-05, "loss": 2.7936, "step": 19606 }, { "epoch": 1.54, "learning_rate": 8.462557829530307e-05, "loss": 2.8084, "step": 19607 }, { "epoch": 1.54, "learning_rate": 8.462479416607857e-05, "loss": 2.7012, "step": 19608 }, { "epoch": 1.54, "learning_rate": 8.462401003685407e-05, "loss": 2.9311, "step": 19609 }, { "epoch": 1.54, "learning_rate": 8.462322590762958e-05, "loss": 2.9866, "step": 19610 }, { "epoch": 1.54, "learning_rate": 8.462244177840508e-05, "loss": 2.8981, "step": 19611 }, { "epoch": 1.54, "learning_rate": 8.462165764918058e-05, "loss": 3.0387, "step": 19612 }, { "epoch": 1.54, "learning_rate": 8.462087351995609e-05, "loss": 2.7214, "step": 19613 }, { "epoch": 1.54, "learning_rate": 8.462008939073159e-05, "loss": 2.8593, "step": 19614 }, { "epoch": 1.54, "learning_rate": 8.46193052615071e-05, "loss": 2.9583, "step": 19615 }, { "epoch": 1.54, "learning_rate": 8.46185211322826e-05, "loss": 2.8485, "step": 19616 }, { "epoch": 1.54, "learning_rate": 8.46177370030581e-05, "loss": 3.147, "step": 19617 }, { "epoch": 1.54, "learning_rate": 8.461695287383361e-05, "loss": 2.6819, "step": 19618 }, { "epoch": 1.54, "learning_rate": 8.461616874460911e-05, "loss": 2.9254, "step": 19619 }, { "epoch": 1.54, "learning_rate": 8.461538461538461e-05, "loss": 3.097, "step": 19620 }, { "epoch": 1.54, "learning_rate": 8.461460048616012e-05, "loss": 2.8359, "step": 19621 }, { "epoch": 1.54, "learning_rate": 8.461381635693562e-05, "loss": 2.8748, "step": 19622 }, { "epoch": 1.54, "learning_rate": 8.461303222771113e-05, "loss": 2.7055, "step": 19623 }, { "epoch": 1.54, "learning_rate": 8.461224809848663e-05, "loss": 3.0815, "step": 19624 }, { "epoch": 1.54, "learning_rate": 8.461146396926213e-05, "loss": 2.8619, "step": 19625 }, { "epoch": 1.54, "learning_rate": 8.461067984003765e-05, "loss": 2.9391, "step": 19626 }, { "epoch": 1.54, "learning_rate": 8.460989571081315e-05, "loss": 2.9023, "step": 19627 }, { "epoch": 1.54, "learning_rate": 8.460911158158866e-05, "loss": 2.8314, "step": 19628 }, { "epoch": 1.54, "learning_rate": 8.460832745236416e-05, "loss": 2.5906, "step": 19629 }, { "epoch": 1.54, "learning_rate": 8.460754332313966e-05, "loss": 2.9654, "step": 19630 }, { "epoch": 1.54, "learning_rate": 8.460675919391517e-05, "loss": 2.731, "step": 19631 }, { "epoch": 1.54, "learning_rate": 8.460597506469067e-05, "loss": 3.0751, "step": 19632 }, { "epoch": 1.54, "learning_rate": 8.460519093546618e-05, "loss": 3.0302, "step": 19633 }, { "epoch": 1.54, "learning_rate": 8.460440680624168e-05, "loss": 2.921, "step": 19634 }, { "epoch": 1.54, "learning_rate": 8.460362267701718e-05, "loss": 2.8691, "step": 19635 }, { "epoch": 1.54, "learning_rate": 8.460283854779269e-05, "loss": 2.9857, "step": 19636 }, { "epoch": 1.54, "learning_rate": 8.460205441856819e-05, "loss": 2.7509, "step": 19637 }, { "epoch": 1.54, "learning_rate": 8.46012702893437e-05, "loss": 2.9722, "step": 19638 }, { "epoch": 1.54, "learning_rate": 8.46004861601192e-05, "loss": 2.8604, "step": 19639 }, { "epoch": 1.54, "learning_rate": 8.45997020308947e-05, "loss": 2.9647, "step": 19640 }, { "epoch": 1.54, "learning_rate": 8.459891790167019e-05, "loss": 2.7426, "step": 19641 }, { "epoch": 1.54, "learning_rate": 8.45981337724457e-05, "loss": 2.8526, "step": 19642 }, { "epoch": 1.54, "learning_rate": 8.45973496432212e-05, "loss": 2.7003, "step": 19643 }, { "epoch": 1.54, "learning_rate": 8.45965655139967e-05, "loss": 2.5686, "step": 19644 }, { "epoch": 1.54, "learning_rate": 8.45957813847722e-05, "loss": 3.1119, "step": 19645 }, { "epoch": 1.54, "learning_rate": 8.459499725554771e-05, "loss": 2.9185, "step": 19646 }, { "epoch": 1.54, "learning_rate": 8.459421312632321e-05, "loss": 2.8806, "step": 19647 }, { "epoch": 1.54, "learning_rate": 8.459342899709872e-05, "loss": 2.8426, "step": 19648 }, { "epoch": 1.54, "learning_rate": 8.459264486787422e-05, "loss": 2.9183, "step": 19649 }, { "epoch": 1.54, "learning_rate": 8.459186073864972e-05, "loss": 2.9793, "step": 19650 }, { "epoch": 1.54, "learning_rate": 8.459107660942523e-05, "loss": 2.8831, "step": 19651 }, { "epoch": 1.54, "learning_rate": 8.459029248020075e-05, "loss": 2.7939, "step": 19652 }, { "epoch": 1.54, "learning_rate": 8.458950835097625e-05, "loss": 2.93, "step": 19653 }, { "epoch": 1.54, "learning_rate": 8.458872422175175e-05, "loss": 2.9958, "step": 19654 }, { "epoch": 1.54, "learning_rate": 8.458794009252726e-05, "loss": 3.1335, "step": 19655 }, { "epoch": 1.54, "learning_rate": 8.458715596330276e-05, "loss": 2.9464, "step": 19656 }, { "epoch": 1.54, "learning_rate": 8.458637183407826e-05, "loss": 2.9036, "step": 19657 }, { "epoch": 1.54, "learning_rate": 8.458558770485377e-05, "loss": 3.0566, "step": 19658 }, { "epoch": 1.54, "learning_rate": 8.458480357562927e-05, "loss": 2.716, "step": 19659 }, { "epoch": 1.54, "learning_rate": 8.458401944640477e-05, "loss": 2.6657, "step": 19660 }, { "epoch": 1.54, "learning_rate": 8.458323531718028e-05, "loss": 2.8904, "step": 19661 }, { "epoch": 1.54, "learning_rate": 8.458245118795578e-05, "loss": 3.0232, "step": 19662 }, { "epoch": 1.54, "learning_rate": 8.458166705873129e-05, "loss": 2.9571, "step": 19663 }, { "epoch": 1.54, "learning_rate": 8.458088292950679e-05, "loss": 3.0196, "step": 19664 }, { "epoch": 1.54, "learning_rate": 8.45800988002823e-05, "loss": 2.9713, "step": 19665 }, { "epoch": 1.54, "learning_rate": 8.45793146710578e-05, "loss": 2.8667, "step": 19666 }, { "epoch": 1.54, "learning_rate": 8.45785305418333e-05, "loss": 2.7595, "step": 19667 }, { "epoch": 1.54, "learning_rate": 8.45777464126088e-05, "loss": 2.928, "step": 19668 }, { "epoch": 1.54, "learning_rate": 8.457696228338431e-05, "loss": 3.0365, "step": 19669 }, { "epoch": 1.54, "learning_rate": 8.457617815415981e-05, "loss": 2.7817, "step": 19670 }, { "epoch": 1.54, "learning_rate": 8.457539402493532e-05, "loss": 2.7872, "step": 19671 }, { "epoch": 1.54, "learning_rate": 8.457460989571082e-05, "loss": 2.8968, "step": 19672 }, { "epoch": 1.54, "learning_rate": 8.457382576648632e-05, "loss": 2.8548, "step": 19673 }, { "epoch": 1.54, "learning_rate": 8.457304163726183e-05, "loss": 2.9026, "step": 19674 }, { "epoch": 1.54, "learning_rate": 8.457225750803733e-05, "loss": 2.8738, "step": 19675 }, { "epoch": 1.54, "learning_rate": 8.457147337881283e-05, "loss": 2.9489, "step": 19676 }, { "epoch": 1.54, "learning_rate": 8.457068924958834e-05, "loss": 2.8234, "step": 19677 }, { "epoch": 1.54, "learning_rate": 8.456990512036384e-05, "loss": 2.8591, "step": 19678 }, { "epoch": 1.54, "learning_rate": 8.456912099113934e-05, "loss": 2.8047, "step": 19679 }, { "epoch": 1.54, "learning_rate": 8.456833686191485e-05, "loss": 2.8841, "step": 19680 }, { "epoch": 1.54, "learning_rate": 8.456755273269035e-05, "loss": 2.7209, "step": 19681 }, { "epoch": 1.54, "learning_rate": 8.456676860346586e-05, "loss": 3.0466, "step": 19682 }, { "epoch": 1.54, "learning_rate": 8.456598447424136e-05, "loss": 2.7638, "step": 19683 }, { "epoch": 1.54, "learning_rate": 8.456520034501686e-05, "loss": 2.7911, "step": 19684 }, { "epoch": 1.54, "learning_rate": 8.456441621579237e-05, "loss": 2.8962, "step": 19685 }, { "epoch": 1.54, "learning_rate": 8.456363208656787e-05, "loss": 2.9291, "step": 19686 }, { "epoch": 1.54, "learning_rate": 8.456284795734337e-05, "loss": 2.7699, "step": 19687 }, { "epoch": 1.54, "learning_rate": 8.456206382811888e-05, "loss": 2.8421, "step": 19688 }, { "epoch": 1.54, "learning_rate": 8.456127969889438e-05, "loss": 2.9644, "step": 19689 }, { "epoch": 1.54, "learning_rate": 8.456049556966989e-05, "loss": 3.0961, "step": 19690 }, { "epoch": 1.54, "learning_rate": 8.455971144044539e-05, "loss": 2.6311, "step": 19691 }, { "epoch": 1.54, "learning_rate": 8.455892731122089e-05, "loss": 2.8493, "step": 19692 }, { "epoch": 1.54, "learning_rate": 8.45581431819964e-05, "loss": 2.8083, "step": 19693 }, { "epoch": 1.54, "learning_rate": 8.45573590527719e-05, "loss": 2.8586, "step": 19694 }, { "epoch": 1.54, "learning_rate": 8.45565749235474e-05, "loss": 2.7821, "step": 19695 }, { "epoch": 1.54, "learning_rate": 8.455579079432291e-05, "loss": 2.7024, "step": 19696 }, { "epoch": 1.54, "learning_rate": 8.455500666509841e-05, "loss": 3.0778, "step": 19697 }, { "epoch": 1.54, "learning_rate": 8.455422253587391e-05, "loss": 3.0156, "step": 19698 }, { "epoch": 1.54, "learning_rate": 8.455343840664942e-05, "loss": 2.8937, "step": 19699 }, { "epoch": 1.54, "learning_rate": 8.455265427742492e-05, "loss": 3.5261, "step": 19700 }, { "epoch": 1.54, "learning_rate": 8.455187014820043e-05, "loss": 2.7016, "step": 19701 }, { "epoch": 1.54, "learning_rate": 8.455108601897593e-05, "loss": 2.8616, "step": 19702 }, { "epoch": 1.54, "learning_rate": 8.455030188975143e-05, "loss": 2.784, "step": 19703 }, { "epoch": 1.55, "learning_rate": 8.454951776052694e-05, "loss": 2.9128, "step": 19704 }, { "epoch": 1.55, "learning_rate": 8.454873363130244e-05, "loss": 2.9082, "step": 19705 }, { "epoch": 1.55, "learning_rate": 8.454794950207794e-05, "loss": 2.7094, "step": 19706 }, { "epoch": 1.55, "learning_rate": 8.454716537285345e-05, "loss": 3.0159, "step": 19707 }, { "epoch": 1.55, "learning_rate": 8.454638124362895e-05, "loss": 2.7664, "step": 19708 }, { "epoch": 1.55, "learning_rate": 8.454559711440447e-05, "loss": 2.8761, "step": 19709 }, { "epoch": 1.55, "learning_rate": 8.454481298517997e-05, "loss": 2.7896, "step": 19710 }, { "epoch": 1.55, "learning_rate": 8.454402885595548e-05, "loss": 2.9081, "step": 19711 }, { "epoch": 1.55, "learning_rate": 8.454324472673098e-05, "loss": 2.7609, "step": 19712 }, { "epoch": 1.55, "learning_rate": 8.454246059750647e-05, "loss": 2.8052, "step": 19713 }, { "epoch": 1.55, "learning_rate": 8.454167646828197e-05, "loss": 2.9037, "step": 19714 }, { "epoch": 1.55, "learning_rate": 8.454089233905748e-05, "loss": 2.9804, "step": 19715 }, { "epoch": 1.55, "learning_rate": 8.454010820983298e-05, "loss": 2.8807, "step": 19716 }, { "epoch": 1.55, "learning_rate": 8.453932408060848e-05, "loss": 2.9458, "step": 19717 }, { "epoch": 1.55, "learning_rate": 8.453853995138399e-05, "loss": 2.8223, "step": 19718 }, { "epoch": 1.55, "learning_rate": 8.453775582215949e-05, "loss": 2.9192, "step": 19719 }, { "epoch": 1.55, "learning_rate": 8.4536971692935e-05, "loss": 2.8136, "step": 19720 }, { "epoch": 1.55, "learning_rate": 8.45361875637105e-05, "loss": 2.7558, "step": 19721 }, { "epoch": 1.55, "learning_rate": 8.4535403434486e-05, "loss": 2.8479, "step": 19722 }, { "epoch": 1.55, "learning_rate": 8.45346193052615e-05, "loss": 2.8961, "step": 19723 }, { "epoch": 1.55, "learning_rate": 8.453383517603701e-05, "loss": 2.9581, "step": 19724 }, { "epoch": 1.55, "learning_rate": 8.453305104681251e-05, "loss": 3.066, "step": 19725 }, { "epoch": 1.55, "learning_rate": 8.453226691758802e-05, "loss": 2.763, "step": 19726 }, { "epoch": 1.55, "learning_rate": 8.453148278836352e-05, "loss": 2.9871, "step": 19727 }, { "epoch": 1.55, "learning_rate": 8.453069865913902e-05, "loss": 2.8663, "step": 19728 }, { "epoch": 1.55, "learning_rate": 8.452991452991453e-05, "loss": 3.1852, "step": 19729 }, { "epoch": 1.55, "learning_rate": 8.452913040069003e-05, "loss": 2.9298, "step": 19730 }, { "epoch": 1.55, "learning_rate": 8.452834627146554e-05, "loss": 2.6964, "step": 19731 }, { "epoch": 1.55, "learning_rate": 8.452756214224104e-05, "loss": 3.1433, "step": 19732 }, { "epoch": 1.55, "learning_rate": 8.452677801301654e-05, "loss": 2.7497, "step": 19733 }, { "epoch": 1.55, "learning_rate": 8.452599388379206e-05, "loss": 3.0739, "step": 19734 }, { "epoch": 1.55, "learning_rate": 8.452520975456756e-05, "loss": 2.6838, "step": 19735 }, { "epoch": 1.55, "learning_rate": 8.452442562534307e-05, "loss": 2.5815, "step": 19736 }, { "epoch": 1.55, "learning_rate": 8.452364149611857e-05, "loss": 2.9094, "step": 19737 }, { "epoch": 1.55, "learning_rate": 8.452285736689408e-05, "loss": 2.8165, "step": 19738 }, { "epoch": 1.55, "learning_rate": 8.452207323766958e-05, "loss": 2.7783, "step": 19739 }, { "epoch": 1.55, "learning_rate": 8.452128910844508e-05, "loss": 2.8838, "step": 19740 }, { "epoch": 1.55, "learning_rate": 8.452050497922059e-05, "loss": 2.8072, "step": 19741 }, { "epoch": 1.55, "learning_rate": 8.451972084999609e-05, "loss": 2.807, "step": 19742 }, { "epoch": 1.55, "learning_rate": 8.45189367207716e-05, "loss": 2.9272, "step": 19743 }, { "epoch": 1.55, "learning_rate": 8.45181525915471e-05, "loss": 3.2927, "step": 19744 }, { "epoch": 1.55, "learning_rate": 8.45173684623226e-05, "loss": 2.8864, "step": 19745 }, { "epoch": 1.55, "learning_rate": 8.45165843330981e-05, "loss": 3.052, "step": 19746 }, { "epoch": 1.55, "learning_rate": 8.451580020387361e-05, "loss": 2.9585, "step": 19747 }, { "epoch": 1.55, "learning_rate": 8.451501607464911e-05, "loss": 3.0646, "step": 19748 }, { "epoch": 1.55, "learning_rate": 8.45142319454246e-05, "loss": 2.6254, "step": 19749 }, { "epoch": 1.55, "learning_rate": 8.45134478162001e-05, "loss": 2.6267, "step": 19750 }, { "epoch": 1.55, "learning_rate": 8.451266368697561e-05, "loss": 3.0332, "step": 19751 }, { "epoch": 1.55, "learning_rate": 8.451187955775111e-05, "loss": 2.9349, "step": 19752 }, { "epoch": 1.55, "learning_rate": 8.451109542852662e-05, "loss": 2.7832, "step": 19753 }, { "epoch": 1.55, "learning_rate": 8.451031129930212e-05, "loss": 2.8327, "step": 19754 }, { "epoch": 1.55, "learning_rate": 8.450952717007762e-05, "loss": 3.0186, "step": 19755 }, { "epoch": 1.55, "learning_rate": 8.450874304085313e-05, "loss": 2.9255, "step": 19756 }, { "epoch": 1.55, "learning_rate": 8.450795891162863e-05, "loss": 2.8364, "step": 19757 }, { "epoch": 1.55, "learning_rate": 8.450717478240414e-05, "loss": 2.9308, "step": 19758 }, { "epoch": 1.55, "learning_rate": 8.450639065317964e-05, "loss": 2.9264, "step": 19759 }, { "epoch": 1.55, "learning_rate": 8.450560652395516e-05, "loss": 2.8852, "step": 19760 }, { "epoch": 1.55, "learning_rate": 8.450482239473066e-05, "loss": 3.0423, "step": 19761 }, { "epoch": 1.55, "learning_rate": 8.450403826550616e-05, "loss": 2.8981, "step": 19762 }, { "epoch": 1.55, "learning_rate": 8.450325413628167e-05, "loss": 2.8194, "step": 19763 }, { "epoch": 1.55, "learning_rate": 8.450247000705717e-05, "loss": 2.8007, "step": 19764 }, { "epoch": 1.55, "learning_rate": 8.450168587783267e-05, "loss": 2.9482, "step": 19765 }, { "epoch": 1.55, "learning_rate": 8.450090174860818e-05, "loss": 2.9598, "step": 19766 }, { "epoch": 1.55, "learning_rate": 8.450011761938368e-05, "loss": 3.0348, "step": 19767 }, { "epoch": 1.55, "learning_rate": 8.449933349015919e-05, "loss": 2.8311, "step": 19768 }, { "epoch": 1.55, "learning_rate": 8.449854936093469e-05, "loss": 2.8403, "step": 19769 }, { "epoch": 1.55, "learning_rate": 8.449776523171019e-05, "loss": 2.8623, "step": 19770 }, { "epoch": 1.55, "learning_rate": 8.44969811024857e-05, "loss": 2.7124, "step": 19771 }, { "epoch": 1.55, "learning_rate": 8.44961969732612e-05, "loss": 2.7529, "step": 19772 }, { "epoch": 1.55, "learning_rate": 8.44954128440367e-05, "loss": 2.6664, "step": 19773 }, { "epoch": 1.55, "learning_rate": 8.449462871481221e-05, "loss": 3.4429, "step": 19774 }, { "epoch": 1.55, "learning_rate": 8.449384458558771e-05, "loss": 2.6833, "step": 19775 }, { "epoch": 1.55, "learning_rate": 8.449306045636322e-05, "loss": 2.8821, "step": 19776 }, { "epoch": 1.55, "learning_rate": 8.449227632713872e-05, "loss": 2.8034, "step": 19777 }, { "epoch": 1.55, "learning_rate": 8.449149219791422e-05, "loss": 2.9671, "step": 19778 }, { "epoch": 1.55, "learning_rate": 8.449070806868973e-05, "loss": 2.8702, "step": 19779 }, { "epoch": 1.55, "learning_rate": 8.448992393946523e-05, "loss": 3.0942, "step": 19780 }, { "epoch": 1.55, "learning_rate": 8.448913981024073e-05, "loss": 2.7184, "step": 19781 }, { "epoch": 1.55, "learning_rate": 8.448835568101624e-05, "loss": 2.9335, "step": 19782 }, { "epoch": 1.55, "learning_rate": 8.448757155179174e-05, "loss": 2.969, "step": 19783 }, { "epoch": 1.55, "learning_rate": 8.448678742256724e-05, "loss": 2.8564, "step": 19784 }, { "epoch": 1.55, "learning_rate": 8.448600329334275e-05, "loss": 2.8513, "step": 19785 }, { "epoch": 1.55, "learning_rate": 8.448521916411825e-05, "loss": 2.6584, "step": 19786 }, { "epoch": 1.55, "learning_rate": 8.448443503489376e-05, "loss": 2.8848, "step": 19787 }, { "epoch": 1.55, "learning_rate": 8.448365090566926e-05, "loss": 2.8817, "step": 19788 }, { "epoch": 1.55, "learning_rate": 8.448286677644476e-05, "loss": 2.7994, "step": 19789 }, { "epoch": 1.55, "learning_rate": 8.448208264722027e-05, "loss": 2.9615, "step": 19790 }, { "epoch": 1.55, "learning_rate": 8.448129851799577e-05, "loss": 2.9133, "step": 19791 }, { "epoch": 1.55, "learning_rate": 8.448051438877127e-05, "loss": 3.0584, "step": 19792 }, { "epoch": 1.55, "learning_rate": 8.447973025954678e-05, "loss": 2.895, "step": 19793 }, { "epoch": 1.55, "learning_rate": 8.447894613032228e-05, "loss": 2.9323, "step": 19794 }, { "epoch": 1.55, "learning_rate": 8.447816200109778e-05, "loss": 2.9842, "step": 19795 }, { "epoch": 1.55, "learning_rate": 8.447737787187329e-05, "loss": 2.983, "step": 19796 }, { "epoch": 1.55, "learning_rate": 8.447659374264879e-05, "loss": 2.9284, "step": 19797 }, { "epoch": 1.55, "learning_rate": 8.44758096134243e-05, "loss": 2.8083, "step": 19798 }, { "epoch": 1.55, "learning_rate": 8.44750254841998e-05, "loss": 2.763, "step": 19799 }, { "epoch": 1.55, "learning_rate": 8.44742413549753e-05, "loss": 2.7902, "step": 19800 }, { "epoch": 1.55, "learning_rate": 8.447345722575081e-05, "loss": 2.7125, "step": 19801 }, { "epoch": 1.55, "learning_rate": 8.447267309652631e-05, "loss": 2.6748, "step": 19802 }, { "epoch": 1.55, "learning_rate": 8.447188896730181e-05, "loss": 3.0101, "step": 19803 }, { "epoch": 1.55, "learning_rate": 8.447110483807732e-05, "loss": 3.0118, "step": 19804 }, { "epoch": 1.55, "learning_rate": 8.447032070885282e-05, "loss": 3.0205, "step": 19805 }, { "epoch": 1.55, "learning_rate": 8.446953657962833e-05, "loss": 2.5813, "step": 19806 }, { "epoch": 1.55, "learning_rate": 8.446875245040383e-05, "loss": 2.9725, "step": 19807 }, { "epoch": 1.55, "learning_rate": 8.446796832117933e-05, "loss": 2.8456, "step": 19808 }, { "epoch": 1.55, "learning_rate": 8.446718419195484e-05, "loss": 3.024, "step": 19809 }, { "epoch": 1.55, "learning_rate": 8.446640006273034e-05, "loss": 2.8143, "step": 19810 }, { "epoch": 1.55, "learning_rate": 8.446561593350584e-05, "loss": 2.9589, "step": 19811 }, { "epoch": 1.55, "learning_rate": 8.446483180428135e-05, "loss": 2.9292, "step": 19812 }, { "epoch": 1.55, "learning_rate": 8.446404767505685e-05, "loss": 2.9969, "step": 19813 }, { "epoch": 1.55, "learning_rate": 8.446326354583235e-05, "loss": 2.971, "step": 19814 }, { "epoch": 1.55, "learning_rate": 8.446247941660786e-05, "loss": 2.873, "step": 19815 }, { "epoch": 1.55, "learning_rate": 8.446169528738336e-05, "loss": 2.8458, "step": 19816 }, { "epoch": 1.55, "learning_rate": 8.446091115815888e-05, "loss": 2.7876, "step": 19817 }, { "epoch": 1.55, "learning_rate": 8.446012702893438e-05, "loss": 2.8246, "step": 19818 }, { "epoch": 1.55, "learning_rate": 8.445934289970989e-05, "loss": 2.6385, "step": 19819 }, { "epoch": 1.55, "learning_rate": 8.445855877048539e-05, "loss": 2.9926, "step": 19820 }, { "epoch": 1.55, "learning_rate": 8.445777464126088e-05, "loss": 3.0614, "step": 19821 }, { "epoch": 1.55, "learning_rate": 8.445699051203638e-05, "loss": 2.7846, "step": 19822 }, { "epoch": 1.55, "learning_rate": 8.445620638281189e-05, "loss": 2.9993, "step": 19823 }, { "epoch": 1.55, "learning_rate": 8.445542225358739e-05, "loss": 3.2909, "step": 19824 }, { "epoch": 1.55, "learning_rate": 8.44546381243629e-05, "loss": 2.7803, "step": 19825 }, { "epoch": 1.55, "learning_rate": 8.44538539951384e-05, "loss": 3.0203, "step": 19826 }, { "epoch": 1.55, "learning_rate": 8.44530698659139e-05, "loss": 2.8616, "step": 19827 }, { "epoch": 1.55, "learning_rate": 8.44522857366894e-05, "loss": 2.8883, "step": 19828 }, { "epoch": 1.55, "learning_rate": 8.445150160746491e-05, "loss": 2.8088, "step": 19829 }, { "epoch": 1.55, "learning_rate": 8.445071747824041e-05, "loss": 2.7448, "step": 19830 }, { "epoch": 1.56, "learning_rate": 8.444993334901592e-05, "loss": 2.7234, "step": 19831 }, { "epoch": 1.56, "learning_rate": 8.444914921979142e-05, "loss": 2.9972, "step": 19832 }, { "epoch": 1.56, "learning_rate": 8.444836509056692e-05, "loss": 3.0484, "step": 19833 }, { "epoch": 1.56, "learning_rate": 8.444758096134243e-05, "loss": 2.7563, "step": 19834 }, { "epoch": 1.56, "learning_rate": 8.444679683211793e-05, "loss": 2.8617, "step": 19835 }, { "epoch": 1.56, "learning_rate": 8.444601270289344e-05, "loss": 3.0859, "step": 19836 }, { "epoch": 1.56, "learning_rate": 8.444522857366894e-05, "loss": 2.9101, "step": 19837 }, { "epoch": 1.56, "learning_rate": 8.444444444444444e-05, "loss": 3.1205, "step": 19838 }, { "epoch": 1.56, "learning_rate": 8.444366031521995e-05, "loss": 2.7904, "step": 19839 }, { "epoch": 1.56, "learning_rate": 8.444287618599545e-05, "loss": 2.7623, "step": 19840 }, { "epoch": 1.56, "learning_rate": 8.444209205677095e-05, "loss": 3.1938, "step": 19841 }, { "epoch": 1.56, "learning_rate": 8.444130792754647e-05, "loss": 2.6757, "step": 19842 }, { "epoch": 1.56, "learning_rate": 8.444052379832198e-05, "loss": 2.7229, "step": 19843 }, { "epoch": 1.56, "learning_rate": 8.443973966909748e-05, "loss": 2.8011, "step": 19844 }, { "epoch": 1.56, "learning_rate": 8.443895553987298e-05, "loss": 3.255, "step": 19845 }, { "epoch": 1.56, "learning_rate": 8.443817141064849e-05, "loss": 2.9877, "step": 19846 }, { "epoch": 1.56, "learning_rate": 8.443738728142399e-05, "loss": 3.1981, "step": 19847 }, { "epoch": 1.56, "learning_rate": 8.44366031521995e-05, "loss": 2.8248, "step": 19848 }, { "epoch": 1.56, "learning_rate": 8.4435819022975e-05, "loss": 2.9822, "step": 19849 }, { "epoch": 1.56, "learning_rate": 8.44350348937505e-05, "loss": 2.9963, "step": 19850 }, { "epoch": 1.56, "learning_rate": 8.4434250764526e-05, "loss": 3.0039, "step": 19851 }, { "epoch": 1.56, "learning_rate": 8.443346663530151e-05, "loss": 3.0418, "step": 19852 }, { "epoch": 1.56, "learning_rate": 8.443268250607701e-05, "loss": 2.9298, "step": 19853 }, { "epoch": 1.56, "learning_rate": 8.443189837685252e-05, "loss": 2.9649, "step": 19854 }, { "epoch": 1.56, "learning_rate": 8.443111424762802e-05, "loss": 3.011, "step": 19855 }, { "epoch": 1.56, "learning_rate": 8.443033011840352e-05, "loss": 2.9703, "step": 19856 }, { "epoch": 1.56, "learning_rate": 8.442954598917901e-05, "loss": 2.8004, "step": 19857 }, { "epoch": 1.56, "learning_rate": 8.442876185995452e-05, "loss": 2.9168, "step": 19858 }, { "epoch": 1.56, "learning_rate": 8.442797773073002e-05, "loss": 2.7527, "step": 19859 }, { "epoch": 1.56, "learning_rate": 8.442719360150552e-05, "loss": 2.9647, "step": 19860 }, { "epoch": 1.56, "learning_rate": 8.442640947228103e-05, "loss": 2.8605, "step": 19861 }, { "epoch": 1.56, "learning_rate": 8.442562534305653e-05, "loss": 3.0204, "step": 19862 }, { "epoch": 1.56, "learning_rate": 8.442484121383204e-05, "loss": 2.863, "step": 19863 }, { "epoch": 1.56, "learning_rate": 8.442405708460754e-05, "loss": 2.9433, "step": 19864 }, { "epoch": 1.56, "learning_rate": 8.442327295538304e-05, "loss": 2.7899, "step": 19865 }, { "epoch": 1.56, "learning_rate": 8.442248882615855e-05, "loss": 2.9049, "step": 19866 }, { "epoch": 1.56, "learning_rate": 8.442170469693405e-05, "loss": 2.7967, "step": 19867 }, { "epoch": 1.56, "learning_rate": 8.442092056770957e-05, "loss": 2.5566, "step": 19868 }, { "epoch": 1.56, "learning_rate": 8.442013643848507e-05, "loss": 2.8022, "step": 19869 }, { "epoch": 1.56, "learning_rate": 8.441935230926057e-05, "loss": 2.948, "step": 19870 }, { "epoch": 1.56, "learning_rate": 8.441856818003608e-05, "loss": 2.9482, "step": 19871 }, { "epoch": 1.56, "learning_rate": 8.441778405081158e-05, "loss": 2.8274, "step": 19872 }, { "epoch": 1.56, "learning_rate": 8.441699992158709e-05, "loss": 2.7992, "step": 19873 }, { "epoch": 1.56, "learning_rate": 8.441621579236259e-05, "loss": 2.7138, "step": 19874 }, { "epoch": 1.56, "learning_rate": 8.441543166313809e-05, "loss": 3.0653, "step": 19875 }, { "epoch": 1.56, "learning_rate": 8.44146475339136e-05, "loss": 3.1163, "step": 19876 }, { "epoch": 1.56, "learning_rate": 8.44138634046891e-05, "loss": 2.9525, "step": 19877 }, { "epoch": 1.56, "learning_rate": 8.44130792754646e-05, "loss": 2.8575, "step": 19878 }, { "epoch": 1.56, "learning_rate": 8.441229514624011e-05, "loss": 3.0684, "step": 19879 }, { "epoch": 1.56, "learning_rate": 8.441151101701561e-05, "loss": 2.8368, "step": 19880 }, { "epoch": 1.56, "learning_rate": 8.441072688779111e-05, "loss": 2.9527, "step": 19881 }, { "epoch": 1.56, "learning_rate": 8.440994275856662e-05, "loss": 2.9623, "step": 19882 }, { "epoch": 1.56, "learning_rate": 8.440915862934212e-05, "loss": 2.864, "step": 19883 }, { "epoch": 1.56, "learning_rate": 8.440837450011763e-05, "loss": 3.2392, "step": 19884 }, { "epoch": 1.56, "learning_rate": 8.440759037089313e-05, "loss": 2.7714, "step": 19885 }, { "epoch": 1.56, "learning_rate": 8.440680624166863e-05, "loss": 3.0689, "step": 19886 }, { "epoch": 1.56, "learning_rate": 8.440602211244414e-05, "loss": 2.9075, "step": 19887 }, { "epoch": 1.56, "learning_rate": 8.440523798321964e-05, "loss": 2.7587, "step": 19888 }, { "epoch": 1.56, "learning_rate": 8.440445385399514e-05, "loss": 3.0491, "step": 19889 }, { "epoch": 1.56, "learning_rate": 8.440366972477065e-05, "loss": 2.997, "step": 19890 }, { "epoch": 1.56, "learning_rate": 8.440288559554615e-05, "loss": 3.0465, "step": 19891 }, { "epoch": 1.56, "learning_rate": 8.440210146632166e-05, "loss": 2.9695, "step": 19892 }, { "epoch": 1.56, "learning_rate": 8.440131733709716e-05, "loss": 2.6491, "step": 19893 }, { "epoch": 1.56, "learning_rate": 8.440053320787266e-05, "loss": 3.0993, "step": 19894 }, { "epoch": 1.56, "learning_rate": 8.439974907864817e-05, "loss": 3.1346, "step": 19895 }, { "epoch": 1.56, "learning_rate": 8.439896494942367e-05, "loss": 2.9401, "step": 19896 }, { "epoch": 1.56, "learning_rate": 8.439818082019917e-05, "loss": 2.7863, "step": 19897 }, { "epoch": 1.56, "learning_rate": 8.439739669097468e-05, "loss": 3.0219, "step": 19898 }, { "epoch": 1.56, "learning_rate": 8.439661256175018e-05, "loss": 2.603, "step": 19899 }, { "epoch": 1.56, "learning_rate": 8.439582843252568e-05, "loss": 2.9461, "step": 19900 }, { "epoch": 1.56, "learning_rate": 8.439504430330119e-05, "loss": 2.9266, "step": 19901 }, { "epoch": 1.56, "learning_rate": 8.439426017407669e-05, "loss": 2.8795, "step": 19902 }, { "epoch": 1.56, "learning_rate": 8.43934760448522e-05, "loss": 2.8436, "step": 19903 }, { "epoch": 1.56, "learning_rate": 8.43926919156277e-05, "loss": 3.0195, "step": 19904 }, { "epoch": 1.56, "learning_rate": 8.43919077864032e-05, "loss": 2.7905, "step": 19905 }, { "epoch": 1.56, "learning_rate": 8.43911236571787e-05, "loss": 3.0813, "step": 19906 }, { "epoch": 1.56, "learning_rate": 8.439033952795421e-05, "loss": 2.8169, "step": 19907 }, { "epoch": 1.56, "learning_rate": 8.438955539872971e-05, "loss": 3.1165, "step": 19908 }, { "epoch": 1.56, "learning_rate": 8.438877126950522e-05, "loss": 2.869, "step": 19909 }, { "epoch": 1.56, "learning_rate": 8.438798714028072e-05, "loss": 2.9473, "step": 19910 }, { "epoch": 1.56, "learning_rate": 8.438720301105623e-05, "loss": 2.963, "step": 19911 }, { "epoch": 1.56, "learning_rate": 8.438641888183173e-05, "loss": 2.7614, "step": 19912 }, { "epoch": 1.56, "learning_rate": 8.438563475260723e-05, "loss": 2.7862, "step": 19913 }, { "epoch": 1.56, "learning_rate": 8.438485062338274e-05, "loss": 2.9799, "step": 19914 }, { "epoch": 1.56, "learning_rate": 8.438406649415824e-05, "loss": 2.855, "step": 19915 }, { "epoch": 1.56, "learning_rate": 8.438328236493374e-05, "loss": 2.9949, "step": 19916 }, { "epoch": 1.56, "learning_rate": 8.438249823570925e-05, "loss": 3.0569, "step": 19917 }, { "epoch": 1.56, "learning_rate": 8.438171410648475e-05, "loss": 2.8677, "step": 19918 }, { "epoch": 1.56, "learning_rate": 8.438092997726025e-05, "loss": 2.7462, "step": 19919 }, { "epoch": 1.56, "learning_rate": 8.438014584803576e-05, "loss": 2.953, "step": 19920 }, { "epoch": 1.56, "learning_rate": 8.437936171881126e-05, "loss": 3.039, "step": 19921 }, { "epoch": 1.56, "learning_rate": 8.437857758958677e-05, "loss": 2.882, "step": 19922 }, { "epoch": 1.56, "learning_rate": 8.437779346036227e-05, "loss": 2.9898, "step": 19923 }, { "epoch": 1.56, "learning_rate": 8.437700933113777e-05, "loss": 3.1065, "step": 19924 }, { "epoch": 1.56, "learning_rate": 8.437622520191329e-05, "loss": 3.0252, "step": 19925 }, { "epoch": 1.56, "learning_rate": 8.43754410726888e-05, "loss": 3.0504, "step": 19926 }, { "epoch": 1.56, "learning_rate": 8.43746569434643e-05, "loss": 2.8671, "step": 19927 }, { "epoch": 1.56, "learning_rate": 8.43738728142398e-05, "loss": 2.9973, "step": 19928 }, { "epoch": 1.56, "learning_rate": 8.437308868501529e-05, "loss": 2.9097, "step": 19929 }, { "epoch": 1.56, "learning_rate": 8.43723045557908e-05, "loss": 2.7322, "step": 19930 }, { "epoch": 1.56, "learning_rate": 8.43715204265663e-05, "loss": 2.8469, "step": 19931 }, { "epoch": 1.56, "learning_rate": 8.43707362973418e-05, "loss": 3.0403, "step": 19932 }, { "epoch": 1.56, "learning_rate": 8.43699521681173e-05, "loss": 3.1229, "step": 19933 }, { "epoch": 1.56, "learning_rate": 8.436916803889281e-05, "loss": 2.7629, "step": 19934 }, { "epoch": 1.56, "learning_rate": 8.436838390966831e-05, "loss": 2.9316, "step": 19935 }, { "epoch": 1.56, "learning_rate": 8.436759978044382e-05, "loss": 2.9756, "step": 19936 }, { "epoch": 1.56, "learning_rate": 8.436681565121932e-05, "loss": 2.9551, "step": 19937 }, { "epoch": 1.56, "learning_rate": 8.436603152199482e-05, "loss": 2.8818, "step": 19938 }, { "epoch": 1.56, "learning_rate": 8.436524739277033e-05, "loss": 2.6559, "step": 19939 }, { "epoch": 1.56, "learning_rate": 8.436446326354583e-05, "loss": 2.7008, "step": 19940 }, { "epoch": 1.56, "learning_rate": 8.436367913432134e-05, "loss": 2.8403, "step": 19941 }, { "epoch": 1.56, "learning_rate": 8.436289500509684e-05, "loss": 3.0422, "step": 19942 }, { "epoch": 1.56, "learning_rate": 8.436211087587234e-05, "loss": 2.7645, "step": 19943 }, { "epoch": 1.56, "learning_rate": 8.436132674664785e-05, "loss": 2.9996, "step": 19944 }, { "epoch": 1.56, "learning_rate": 8.436054261742335e-05, "loss": 2.8429, "step": 19945 }, { "epoch": 1.56, "learning_rate": 8.435975848819885e-05, "loss": 2.8614, "step": 19946 }, { "epoch": 1.56, "learning_rate": 8.435897435897436e-05, "loss": 2.7358, "step": 19947 }, { "epoch": 1.56, "learning_rate": 8.435819022974986e-05, "loss": 2.815, "step": 19948 }, { "epoch": 1.56, "learning_rate": 8.435740610052536e-05, "loss": 2.8497, "step": 19949 }, { "epoch": 1.56, "learning_rate": 8.435662197130088e-05, "loss": 2.7752, "step": 19950 }, { "epoch": 1.56, "learning_rate": 8.435583784207639e-05, "loss": 2.8504, "step": 19951 }, { "epoch": 1.56, "learning_rate": 8.435505371285189e-05, "loss": 2.7909, "step": 19952 }, { "epoch": 1.56, "learning_rate": 8.435426958362739e-05, "loss": 3.0084, "step": 19953 }, { "epoch": 1.56, "learning_rate": 8.43534854544029e-05, "loss": 3.2089, "step": 19954 }, { "epoch": 1.56, "learning_rate": 8.43527013251784e-05, "loss": 2.7025, "step": 19955 }, { "epoch": 1.56, "learning_rate": 8.43519171959539e-05, "loss": 2.8584, "step": 19956 }, { "epoch": 1.56, "learning_rate": 8.435113306672941e-05, "loss": 2.7947, "step": 19957 }, { "epoch": 1.56, "learning_rate": 8.435034893750491e-05, "loss": 2.7706, "step": 19958 }, { "epoch": 1.57, "learning_rate": 8.434956480828042e-05, "loss": 3.1967, "step": 19959 }, { "epoch": 1.57, "learning_rate": 8.434878067905592e-05, "loss": 3.162, "step": 19960 }, { "epoch": 1.57, "learning_rate": 8.434799654983142e-05, "loss": 3.013, "step": 19961 }, { "epoch": 1.57, "learning_rate": 8.434721242060693e-05, "loss": 3.0493, "step": 19962 }, { "epoch": 1.57, "learning_rate": 8.434642829138243e-05, "loss": 2.8179, "step": 19963 }, { "epoch": 1.57, "learning_rate": 8.434564416215793e-05, "loss": 3.0207, "step": 19964 }, { "epoch": 1.57, "learning_rate": 8.434486003293342e-05, "loss": 3.2057, "step": 19965 }, { "epoch": 1.57, "learning_rate": 8.434407590370893e-05, "loss": 2.8244, "step": 19966 }, { "epoch": 1.57, "learning_rate": 8.434329177448443e-05, "loss": 2.8135, "step": 19967 }, { "epoch": 1.57, "learning_rate": 8.434250764525993e-05, "loss": 2.8943, "step": 19968 }, { "epoch": 1.57, "learning_rate": 8.434172351603544e-05, "loss": 2.867, "step": 19969 }, { "epoch": 1.57, "learning_rate": 8.434093938681094e-05, "loss": 2.6989, "step": 19970 }, { "epoch": 1.57, "learning_rate": 8.434015525758645e-05, "loss": 3.0032, "step": 19971 }, { "epoch": 1.57, "learning_rate": 8.433937112836195e-05, "loss": 2.8473, "step": 19972 }, { "epoch": 1.57, "learning_rate": 8.433858699913745e-05, "loss": 2.9769, "step": 19973 }, { "epoch": 1.57, "learning_rate": 8.433780286991296e-05, "loss": 2.7984, "step": 19974 }, { "epoch": 1.57, "learning_rate": 8.433701874068846e-05, "loss": 3.1543, "step": 19975 }, { "epoch": 1.57, "learning_rate": 8.433623461146398e-05, "loss": 2.8702, "step": 19976 }, { "epoch": 1.57, "learning_rate": 8.433545048223948e-05, "loss": 2.8495, "step": 19977 }, { "epoch": 1.57, "learning_rate": 8.433466635301499e-05, "loss": 2.6093, "step": 19978 }, { "epoch": 1.57, "learning_rate": 8.433388222379049e-05, "loss": 2.8914, "step": 19979 }, { "epoch": 1.57, "learning_rate": 8.433309809456599e-05, "loss": 3.0363, "step": 19980 }, { "epoch": 1.57, "learning_rate": 8.43323139653415e-05, "loss": 2.8015, "step": 19981 }, { "epoch": 1.57, "learning_rate": 8.4331529836117e-05, "loss": 3.2966, "step": 19982 }, { "epoch": 1.57, "learning_rate": 8.43307457068925e-05, "loss": 2.875, "step": 19983 }, { "epoch": 1.57, "learning_rate": 8.432996157766801e-05, "loss": 2.7625, "step": 19984 }, { "epoch": 1.57, "learning_rate": 8.432917744844351e-05, "loss": 2.8902, "step": 19985 }, { "epoch": 1.57, "learning_rate": 8.432839331921901e-05, "loss": 3.0016, "step": 19986 }, { "epoch": 1.57, "learning_rate": 8.432760918999452e-05, "loss": 2.75, "step": 19987 }, { "epoch": 1.57, "learning_rate": 8.432682506077002e-05, "loss": 2.9965, "step": 19988 }, { "epoch": 1.57, "learning_rate": 8.432604093154553e-05, "loss": 2.8168, "step": 19989 }, { "epoch": 1.57, "learning_rate": 8.432525680232103e-05, "loss": 3.0488, "step": 19990 }, { "epoch": 1.57, "learning_rate": 8.432447267309653e-05, "loss": 3.0049, "step": 19991 }, { "epoch": 1.57, "learning_rate": 8.432368854387204e-05, "loss": 3.0078, "step": 19992 }, { "epoch": 1.57, "learning_rate": 8.432290441464754e-05, "loss": 2.8901, "step": 19993 }, { "epoch": 1.57, "learning_rate": 8.432212028542304e-05, "loss": 2.729, "step": 19994 }, { "epoch": 1.57, "learning_rate": 8.432133615619855e-05, "loss": 3.0765, "step": 19995 }, { "epoch": 1.57, "learning_rate": 8.432055202697405e-05, "loss": 3.0575, "step": 19996 }, { "epoch": 1.57, "learning_rate": 8.431976789774955e-05, "loss": 2.8302, "step": 19997 }, { "epoch": 1.57, "learning_rate": 8.431898376852506e-05, "loss": 2.6606, "step": 19998 }, { "epoch": 1.57, "learning_rate": 8.431819963930056e-05, "loss": 2.733, "step": 19999 }, { "epoch": 1.57, "learning_rate": 8.431741551007607e-05, "loss": 2.7298, "step": 20000 }, { "epoch": 1.57, "eval_loss": 2.912069320678711, "eval_runtime": 217.9108, "eval_samples_per_second": 52.003, "eval_steps_per_second": 3.254, "step": 20000 }, { "epoch": 1.57, "learning_rate": 8.431663138085157e-05, "loss": 2.8373, "step": 20001 }, { "epoch": 1.57, "learning_rate": 8.431584725162707e-05, "loss": 2.9821, "step": 20002 }, { "epoch": 1.57, "learning_rate": 8.431506312240258e-05, "loss": 2.8625, "step": 20003 }, { "epoch": 1.57, "learning_rate": 8.431427899317808e-05, "loss": 3.1, "step": 20004 }, { "epoch": 1.57, "learning_rate": 8.431349486395358e-05, "loss": 3.1664, "step": 20005 }, { "epoch": 1.57, "learning_rate": 8.431271073472909e-05, "loss": 2.6456, "step": 20006 }, { "epoch": 1.57, "learning_rate": 8.431192660550459e-05, "loss": 3.1711, "step": 20007 }, { "epoch": 1.57, "learning_rate": 8.43111424762801e-05, "loss": 2.7737, "step": 20008 }, { "epoch": 1.57, "learning_rate": 8.43103583470556e-05, "loss": 2.6253, "step": 20009 }, { "epoch": 1.57, "learning_rate": 8.43095742178311e-05, "loss": 3.0058, "step": 20010 }, { "epoch": 1.57, "learning_rate": 8.43087900886066e-05, "loss": 3.0709, "step": 20011 }, { "epoch": 1.57, "learning_rate": 8.430800595938211e-05, "loss": 2.8687, "step": 20012 }, { "epoch": 1.57, "learning_rate": 8.430722183015761e-05, "loss": 2.5237, "step": 20013 }, { "epoch": 1.57, "learning_rate": 8.430643770093312e-05, "loss": 2.8548, "step": 20014 }, { "epoch": 1.57, "learning_rate": 8.430565357170862e-05, "loss": 2.9059, "step": 20015 }, { "epoch": 1.57, "learning_rate": 8.430486944248412e-05, "loss": 2.9309, "step": 20016 }, { "epoch": 1.57, "learning_rate": 8.430408531325963e-05, "loss": 3.4167, "step": 20017 }, { "epoch": 1.57, "learning_rate": 8.430330118403513e-05, "loss": 2.7494, "step": 20018 }, { "epoch": 1.57, "learning_rate": 8.430251705481064e-05, "loss": 3.05, "step": 20019 }, { "epoch": 1.57, "learning_rate": 8.430173292558614e-05, "loss": 2.7027, "step": 20020 }, { "epoch": 1.57, "learning_rate": 8.430094879636164e-05, "loss": 2.7333, "step": 20021 }, { "epoch": 1.57, "learning_rate": 8.430016466713715e-05, "loss": 3.0206, "step": 20022 }, { "epoch": 1.57, "learning_rate": 8.429938053791265e-05, "loss": 2.7448, "step": 20023 }, { "epoch": 1.57, "learning_rate": 8.429859640868815e-05, "loss": 3.1031, "step": 20024 }, { "epoch": 1.57, "learning_rate": 8.429781227946366e-05, "loss": 2.7259, "step": 20025 }, { "epoch": 1.57, "learning_rate": 8.429702815023916e-05, "loss": 2.8551, "step": 20026 }, { "epoch": 1.57, "learning_rate": 8.429624402101467e-05, "loss": 3.0599, "step": 20027 }, { "epoch": 1.57, "learning_rate": 8.429545989179017e-05, "loss": 2.944, "step": 20028 }, { "epoch": 1.57, "learning_rate": 8.429467576256567e-05, "loss": 2.8817, "step": 20029 }, { "epoch": 1.57, "learning_rate": 8.429389163334118e-05, "loss": 2.7177, "step": 20030 }, { "epoch": 1.57, "learning_rate": 8.429310750411668e-05, "loss": 2.8767, "step": 20031 }, { "epoch": 1.57, "learning_rate": 8.429232337489218e-05, "loss": 2.8054, "step": 20032 }, { "epoch": 1.57, "learning_rate": 8.42915392456677e-05, "loss": 2.6239, "step": 20033 }, { "epoch": 1.57, "learning_rate": 8.42907551164432e-05, "loss": 3.0198, "step": 20034 }, { "epoch": 1.57, "learning_rate": 8.428997098721871e-05, "loss": 3.0292, "step": 20035 }, { "epoch": 1.57, "learning_rate": 8.428918685799421e-05, "loss": 2.8558, "step": 20036 }, { "epoch": 1.57, "learning_rate": 8.42884027287697e-05, "loss": 3.0542, "step": 20037 }, { "epoch": 1.57, "learning_rate": 8.42876185995452e-05, "loss": 2.7212, "step": 20038 }, { "epoch": 1.57, "learning_rate": 8.428683447032071e-05, "loss": 3.1221, "step": 20039 }, { "epoch": 1.57, "learning_rate": 8.428605034109621e-05, "loss": 2.8273, "step": 20040 }, { "epoch": 1.57, "learning_rate": 8.428526621187172e-05, "loss": 3.0046, "step": 20041 }, { "epoch": 1.57, "learning_rate": 8.428448208264722e-05, "loss": 2.6222, "step": 20042 }, { "epoch": 1.57, "learning_rate": 8.428369795342272e-05, "loss": 2.8563, "step": 20043 }, { "epoch": 1.57, "learning_rate": 8.428291382419823e-05, "loss": 2.7992, "step": 20044 }, { "epoch": 1.57, "learning_rate": 8.428212969497373e-05, "loss": 2.8456, "step": 20045 }, { "epoch": 1.57, "learning_rate": 8.428134556574924e-05, "loss": 2.8295, "step": 20046 }, { "epoch": 1.57, "learning_rate": 8.428056143652474e-05, "loss": 2.6722, "step": 20047 }, { "epoch": 1.57, "learning_rate": 8.427977730730024e-05, "loss": 2.9073, "step": 20048 }, { "epoch": 1.57, "learning_rate": 8.427899317807575e-05, "loss": 3.1003, "step": 20049 }, { "epoch": 1.57, "learning_rate": 8.427820904885125e-05, "loss": 2.7709, "step": 20050 }, { "epoch": 1.57, "learning_rate": 8.427742491962675e-05, "loss": 2.9348, "step": 20051 }, { "epoch": 1.57, "learning_rate": 8.427664079040226e-05, "loss": 2.9564, "step": 20052 }, { "epoch": 1.57, "learning_rate": 8.427585666117776e-05, "loss": 2.8218, "step": 20053 }, { "epoch": 1.57, "learning_rate": 8.427507253195326e-05, "loss": 2.9502, "step": 20054 }, { "epoch": 1.57, "learning_rate": 8.427428840272877e-05, "loss": 2.9539, "step": 20055 }, { "epoch": 1.57, "learning_rate": 8.427350427350427e-05, "loss": 2.9523, "step": 20056 }, { "epoch": 1.57, "learning_rate": 8.427272014427978e-05, "loss": 2.7329, "step": 20057 }, { "epoch": 1.57, "learning_rate": 8.427193601505529e-05, "loss": 3.0247, "step": 20058 }, { "epoch": 1.57, "learning_rate": 8.42711518858308e-05, "loss": 2.8943, "step": 20059 }, { "epoch": 1.57, "learning_rate": 8.42703677566063e-05, "loss": 3.0212, "step": 20060 }, { "epoch": 1.57, "learning_rate": 8.42695836273818e-05, "loss": 2.8737, "step": 20061 }, { "epoch": 1.57, "learning_rate": 8.426879949815731e-05, "loss": 2.9175, "step": 20062 }, { "epoch": 1.57, "learning_rate": 8.426801536893281e-05, "loss": 2.764, "step": 20063 }, { "epoch": 1.57, "learning_rate": 8.426723123970831e-05, "loss": 2.9975, "step": 20064 }, { "epoch": 1.57, "learning_rate": 8.426644711048382e-05, "loss": 2.7531, "step": 20065 }, { "epoch": 1.57, "learning_rate": 8.426566298125932e-05, "loss": 2.7619, "step": 20066 }, { "epoch": 1.57, "learning_rate": 8.426487885203483e-05, "loss": 2.8336, "step": 20067 }, { "epoch": 1.57, "learning_rate": 8.426409472281033e-05, "loss": 2.8905, "step": 20068 }, { "epoch": 1.57, "learning_rate": 8.426331059358583e-05, "loss": 2.817, "step": 20069 }, { "epoch": 1.57, "learning_rate": 8.426252646436134e-05, "loss": 3.2146, "step": 20070 }, { "epoch": 1.57, "learning_rate": 8.426174233513684e-05, "loss": 2.8479, "step": 20071 }, { "epoch": 1.57, "learning_rate": 8.426095820591234e-05, "loss": 3.0481, "step": 20072 }, { "epoch": 1.57, "learning_rate": 8.426017407668783e-05, "loss": 2.8529, "step": 20073 }, { "epoch": 1.57, "learning_rate": 8.425938994746334e-05, "loss": 2.8205, "step": 20074 }, { "epoch": 1.57, "learning_rate": 8.425860581823884e-05, "loss": 2.9748, "step": 20075 }, { "epoch": 1.57, "learning_rate": 8.425782168901435e-05, "loss": 2.7504, "step": 20076 }, { "epoch": 1.57, "learning_rate": 8.425703755978985e-05, "loss": 2.9783, "step": 20077 }, { "epoch": 1.57, "learning_rate": 8.425625343056535e-05, "loss": 3.0991, "step": 20078 }, { "epoch": 1.57, "learning_rate": 8.425546930134086e-05, "loss": 3.0116, "step": 20079 }, { "epoch": 1.57, "learning_rate": 8.425468517211636e-05, "loss": 2.9393, "step": 20080 }, { "epoch": 1.57, "learning_rate": 8.425390104289186e-05, "loss": 3.0884, "step": 20081 }, { "epoch": 1.57, "learning_rate": 8.425311691366737e-05, "loss": 2.6832, "step": 20082 }, { "epoch": 1.57, "learning_rate": 8.425233278444287e-05, "loss": 3.1111, "step": 20083 }, { "epoch": 1.57, "learning_rate": 8.425154865521839e-05, "loss": 2.7836, "step": 20084 }, { "epoch": 1.57, "learning_rate": 8.425076452599389e-05, "loss": 2.6129, "step": 20085 }, { "epoch": 1.58, "learning_rate": 8.42499803967694e-05, "loss": 2.8497, "step": 20086 }, { "epoch": 1.58, "learning_rate": 8.42491962675449e-05, "loss": 3.1461, "step": 20087 }, { "epoch": 1.58, "learning_rate": 8.42484121383204e-05, "loss": 2.6767, "step": 20088 }, { "epoch": 1.58, "learning_rate": 8.424762800909591e-05, "loss": 2.9917, "step": 20089 }, { "epoch": 1.58, "learning_rate": 8.424684387987141e-05, "loss": 3.2792, "step": 20090 }, { "epoch": 1.58, "learning_rate": 8.424605975064691e-05, "loss": 2.8133, "step": 20091 }, { "epoch": 1.58, "learning_rate": 8.424527562142242e-05, "loss": 2.8218, "step": 20092 }, { "epoch": 1.58, "learning_rate": 8.424449149219792e-05, "loss": 2.8194, "step": 20093 }, { "epoch": 1.58, "learning_rate": 8.424370736297343e-05, "loss": 2.9022, "step": 20094 }, { "epoch": 1.58, "learning_rate": 8.424292323374893e-05, "loss": 2.7147, "step": 20095 }, { "epoch": 1.58, "learning_rate": 8.424213910452443e-05, "loss": 2.7712, "step": 20096 }, { "epoch": 1.58, "learning_rate": 8.424135497529994e-05, "loss": 2.8401, "step": 20097 }, { "epoch": 1.58, "learning_rate": 8.424057084607544e-05, "loss": 2.9616, "step": 20098 }, { "epoch": 1.58, "learning_rate": 8.423978671685094e-05, "loss": 2.8278, "step": 20099 }, { "epoch": 1.58, "learning_rate": 8.423900258762645e-05, "loss": 2.9226, "step": 20100 }, { "epoch": 1.58, "learning_rate": 8.423821845840195e-05, "loss": 3.0368, "step": 20101 }, { "epoch": 1.58, "learning_rate": 8.423743432917745e-05, "loss": 2.7761, "step": 20102 }, { "epoch": 1.58, "learning_rate": 8.423665019995296e-05, "loss": 2.7141, "step": 20103 }, { "epoch": 1.58, "learning_rate": 8.423586607072846e-05, "loss": 2.7511, "step": 20104 }, { "epoch": 1.58, "learning_rate": 8.423508194150397e-05, "loss": 2.8217, "step": 20105 }, { "epoch": 1.58, "learning_rate": 8.423429781227947e-05, "loss": 2.8904, "step": 20106 }, { "epoch": 1.58, "learning_rate": 8.423351368305497e-05, "loss": 2.7133, "step": 20107 }, { "epoch": 1.58, "learning_rate": 8.423272955383048e-05, "loss": 2.9633, "step": 20108 }, { "epoch": 1.58, "learning_rate": 8.423194542460597e-05, "loss": 3.3018, "step": 20109 }, { "epoch": 1.58, "learning_rate": 8.423116129538148e-05, "loss": 2.7959, "step": 20110 }, { "epoch": 1.58, "learning_rate": 8.423037716615699e-05, "loss": 3.1291, "step": 20111 }, { "epoch": 1.58, "learning_rate": 8.422959303693249e-05, "loss": 2.767, "step": 20112 }, { "epoch": 1.58, "learning_rate": 8.4228808907708e-05, "loss": 3.0584, "step": 20113 }, { "epoch": 1.58, "learning_rate": 8.42280247784835e-05, "loss": 2.8327, "step": 20114 }, { "epoch": 1.58, "learning_rate": 8.4227240649259e-05, "loss": 2.8697, "step": 20115 }, { "epoch": 1.58, "learning_rate": 8.42264565200345e-05, "loss": 2.8972, "step": 20116 }, { "epoch": 1.58, "learning_rate": 8.422567239081001e-05, "loss": 3.3079, "step": 20117 }, { "epoch": 1.58, "learning_rate": 8.422488826158551e-05, "loss": 2.8432, "step": 20118 }, { "epoch": 1.58, "learning_rate": 8.422410413236102e-05, "loss": 2.9403, "step": 20119 }, { "epoch": 1.58, "learning_rate": 8.422332000313652e-05, "loss": 3.1219, "step": 20120 }, { "epoch": 1.58, "learning_rate": 8.422253587391202e-05, "loss": 2.6576, "step": 20121 }, { "epoch": 1.58, "learning_rate": 8.422175174468753e-05, "loss": 3.0888, "step": 20122 }, { "epoch": 1.58, "learning_rate": 8.422096761546303e-05, "loss": 2.8803, "step": 20123 }, { "epoch": 1.58, "learning_rate": 8.422018348623854e-05, "loss": 2.9388, "step": 20124 }, { "epoch": 1.58, "learning_rate": 8.421939935701404e-05, "loss": 3.1195, "step": 20125 }, { "epoch": 1.58, "learning_rate": 8.421861522778954e-05, "loss": 3.012, "step": 20126 }, { "epoch": 1.58, "learning_rate": 8.421783109856505e-05, "loss": 2.7608, "step": 20127 }, { "epoch": 1.58, "learning_rate": 8.421704696934055e-05, "loss": 2.6649, "step": 20128 }, { "epoch": 1.58, "learning_rate": 8.421626284011605e-05, "loss": 2.6496, "step": 20129 }, { "epoch": 1.58, "learning_rate": 8.421547871089156e-05, "loss": 2.7614, "step": 20130 }, { "epoch": 1.58, "learning_rate": 8.421469458166706e-05, "loss": 3.0892, "step": 20131 }, { "epoch": 1.58, "learning_rate": 8.421391045244256e-05, "loss": 2.7599, "step": 20132 }, { "epoch": 1.58, "learning_rate": 8.421312632321807e-05, "loss": 2.8534, "step": 20133 }, { "epoch": 1.58, "learning_rate": 8.421234219399357e-05, "loss": 2.9833, "step": 20134 }, { "epoch": 1.58, "learning_rate": 8.421155806476908e-05, "loss": 2.9027, "step": 20135 }, { "epoch": 1.58, "learning_rate": 8.421077393554458e-05, "loss": 2.8261, "step": 20136 }, { "epoch": 1.58, "learning_rate": 8.420998980632008e-05, "loss": 3.0371, "step": 20137 }, { "epoch": 1.58, "learning_rate": 8.420920567709559e-05, "loss": 2.8429, "step": 20138 }, { "epoch": 1.58, "learning_rate": 8.420842154787109e-05, "loss": 3.2105, "step": 20139 }, { "epoch": 1.58, "learning_rate": 8.42076374186466e-05, "loss": 2.9836, "step": 20140 }, { "epoch": 1.58, "learning_rate": 8.420685328942211e-05, "loss": 2.8275, "step": 20141 }, { "epoch": 1.58, "learning_rate": 8.420606916019762e-05, "loss": 2.8508, "step": 20142 }, { "epoch": 1.58, "learning_rate": 8.420528503097312e-05, "loss": 2.9947, "step": 20143 }, { "epoch": 1.58, "learning_rate": 8.420450090174862e-05, "loss": 3.0745, "step": 20144 }, { "epoch": 1.58, "learning_rate": 8.420371677252411e-05, "loss": 2.964, "step": 20145 }, { "epoch": 1.58, "learning_rate": 8.420293264329962e-05, "loss": 2.895, "step": 20146 }, { "epoch": 1.58, "learning_rate": 8.420214851407512e-05, "loss": 2.8084, "step": 20147 }, { "epoch": 1.58, "learning_rate": 8.420136438485062e-05, "loss": 3.0017, "step": 20148 }, { "epoch": 1.58, "learning_rate": 8.420058025562613e-05, "loss": 3.0744, "step": 20149 }, { "epoch": 1.58, "learning_rate": 8.419979612640163e-05, "loss": 2.8892, "step": 20150 }, { "epoch": 1.58, "learning_rate": 8.419901199717713e-05, "loss": 3.1628, "step": 20151 }, { "epoch": 1.58, "learning_rate": 8.419822786795264e-05, "loss": 2.9813, "step": 20152 }, { "epoch": 1.58, "learning_rate": 8.419744373872814e-05, "loss": 2.6753, "step": 20153 }, { "epoch": 1.58, "learning_rate": 8.419665960950365e-05, "loss": 2.8699, "step": 20154 }, { "epoch": 1.58, "learning_rate": 8.419587548027915e-05, "loss": 2.7236, "step": 20155 }, { "epoch": 1.58, "learning_rate": 8.419509135105465e-05, "loss": 2.864, "step": 20156 }, { "epoch": 1.58, "learning_rate": 8.419430722183016e-05, "loss": 2.9669, "step": 20157 }, { "epoch": 1.58, "learning_rate": 8.419352309260566e-05, "loss": 2.6854, "step": 20158 }, { "epoch": 1.58, "learning_rate": 8.419273896338116e-05, "loss": 2.7301, "step": 20159 }, { "epoch": 1.58, "learning_rate": 8.419195483415667e-05, "loss": 2.8716, "step": 20160 }, { "epoch": 1.58, "learning_rate": 8.419117070493217e-05, "loss": 2.8729, "step": 20161 }, { "epoch": 1.58, "learning_rate": 8.419038657570768e-05, "loss": 2.9084, "step": 20162 }, { "epoch": 1.58, "learning_rate": 8.418960244648318e-05, "loss": 2.9816, "step": 20163 }, { "epoch": 1.58, "learning_rate": 8.418881831725868e-05, "loss": 2.9599, "step": 20164 }, { "epoch": 1.58, "learning_rate": 8.418803418803419e-05, "loss": 2.9631, "step": 20165 }, { "epoch": 1.58, "learning_rate": 8.41872500588097e-05, "loss": 2.7638, "step": 20166 }, { "epoch": 1.58, "learning_rate": 8.418646592958521e-05, "loss": 2.8558, "step": 20167 }, { "epoch": 1.58, "learning_rate": 8.418568180036071e-05, "loss": 2.7018, "step": 20168 }, { "epoch": 1.58, "learning_rate": 8.418489767113621e-05, "loss": 2.8793, "step": 20169 }, { "epoch": 1.58, "learning_rate": 8.418411354191172e-05, "loss": 2.9106, "step": 20170 }, { "epoch": 1.58, "learning_rate": 8.418332941268722e-05, "loss": 2.82, "step": 20171 }, { "epoch": 1.58, "learning_rate": 8.418254528346273e-05, "loss": 3.1718, "step": 20172 }, { "epoch": 1.58, "learning_rate": 8.418176115423823e-05, "loss": 3.1141, "step": 20173 }, { "epoch": 1.58, "learning_rate": 8.418097702501373e-05, "loss": 2.743, "step": 20174 }, { "epoch": 1.58, "learning_rate": 8.418019289578924e-05, "loss": 2.9162, "step": 20175 }, { "epoch": 1.58, "learning_rate": 8.417940876656474e-05, "loss": 2.8954, "step": 20176 }, { "epoch": 1.58, "learning_rate": 8.417862463734024e-05, "loss": 2.9261, "step": 20177 }, { "epoch": 1.58, "learning_rate": 8.417784050811575e-05, "loss": 3.0781, "step": 20178 }, { "epoch": 1.58, "learning_rate": 8.417705637889125e-05, "loss": 3.0306, "step": 20179 }, { "epoch": 1.58, "learning_rate": 8.417627224966676e-05, "loss": 2.8397, "step": 20180 }, { "epoch": 1.58, "learning_rate": 8.417548812044225e-05, "loss": 2.7011, "step": 20181 }, { "epoch": 1.58, "learning_rate": 8.417470399121775e-05, "loss": 2.777, "step": 20182 }, { "epoch": 1.58, "learning_rate": 8.417391986199325e-05, "loss": 2.8862, "step": 20183 }, { "epoch": 1.58, "learning_rate": 8.417313573276876e-05, "loss": 2.8437, "step": 20184 }, { "epoch": 1.58, "learning_rate": 8.417235160354426e-05, "loss": 2.8296, "step": 20185 }, { "epoch": 1.58, "learning_rate": 8.417156747431976e-05, "loss": 3.0079, "step": 20186 }, { "epoch": 1.58, "learning_rate": 8.417078334509527e-05, "loss": 2.8277, "step": 20187 }, { "epoch": 1.58, "learning_rate": 8.416999921587077e-05, "loss": 2.8092, "step": 20188 }, { "epoch": 1.58, "learning_rate": 8.416921508664627e-05, "loss": 3.0137, "step": 20189 }, { "epoch": 1.58, "learning_rate": 8.416843095742178e-05, "loss": 2.9812, "step": 20190 }, { "epoch": 1.58, "learning_rate": 8.416764682819728e-05, "loss": 2.8783, "step": 20191 }, { "epoch": 1.58, "learning_rate": 8.41668626989728e-05, "loss": 2.6938, "step": 20192 }, { "epoch": 1.58, "learning_rate": 8.41660785697483e-05, "loss": 2.6757, "step": 20193 }, { "epoch": 1.58, "learning_rate": 8.41652944405238e-05, "loss": 3.0007, "step": 20194 }, { "epoch": 1.58, "learning_rate": 8.416451031129931e-05, "loss": 2.8421, "step": 20195 }, { "epoch": 1.58, "learning_rate": 8.416372618207481e-05, "loss": 2.605, "step": 20196 }, { "epoch": 1.58, "learning_rate": 8.416294205285032e-05, "loss": 3.085, "step": 20197 }, { "epoch": 1.58, "learning_rate": 8.416215792362582e-05, "loss": 2.8173, "step": 20198 }, { "epoch": 1.58, "learning_rate": 8.416137379440132e-05, "loss": 2.8323, "step": 20199 }, { "epoch": 1.58, "learning_rate": 8.416058966517683e-05, "loss": 2.8579, "step": 20200 }, { "epoch": 1.58, "learning_rate": 8.415980553595233e-05, "loss": 2.8113, "step": 20201 }, { "epoch": 1.58, "learning_rate": 8.415902140672784e-05, "loss": 2.8159, "step": 20202 }, { "epoch": 1.58, "learning_rate": 8.415823727750334e-05, "loss": 2.8091, "step": 20203 }, { "epoch": 1.58, "learning_rate": 8.415745314827884e-05, "loss": 2.9242, "step": 20204 }, { "epoch": 1.58, "learning_rate": 8.415666901905435e-05, "loss": 2.7534, "step": 20205 }, { "epoch": 1.58, "learning_rate": 8.415588488982985e-05, "loss": 3.0359, "step": 20206 }, { "epoch": 1.58, "learning_rate": 8.415510076060535e-05, "loss": 3.1228, "step": 20207 }, { "epoch": 1.58, "learning_rate": 8.415431663138086e-05, "loss": 2.9374, "step": 20208 }, { "epoch": 1.58, "learning_rate": 8.415353250215636e-05, "loss": 2.9469, "step": 20209 }, { "epoch": 1.58, "learning_rate": 8.415274837293187e-05, "loss": 3.188, "step": 20210 }, { "epoch": 1.58, "learning_rate": 8.415196424370737e-05, "loss": 3.0538, "step": 20211 }, { "epoch": 1.58, "learning_rate": 8.415118011448287e-05, "loss": 2.85, "step": 20212 }, { "epoch": 1.58, "learning_rate": 8.415039598525838e-05, "loss": 3.1158, "step": 20213 }, { "epoch": 1.59, "learning_rate": 8.414961185603388e-05, "loss": 2.8181, "step": 20214 }, { "epoch": 1.59, "learning_rate": 8.414882772680938e-05, "loss": 2.9892, "step": 20215 }, { "epoch": 1.59, "learning_rate": 8.414804359758489e-05, "loss": 2.9854, "step": 20216 }, { "epoch": 1.59, "learning_rate": 8.414725946836038e-05, "loss": 2.7649, "step": 20217 }, { "epoch": 1.59, "learning_rate": 8.41464753391359e-05, "loss": 2.9186, "step": 20218 }, { "epoch": 1.59, "learning_rate": 8.41456912099114e-05, "loss": 3.16, "step": 20219 }, { "epoch": 1.59, "learning_rate": 8.41449070806869e-05, "loss": 2.6948, "step": 20220 }, { "epoch": 1.59, "learning_rate": 8.41441229514624e-05, "loss": 2.751, "step": 20221 }, { "epoch": 1.59, "learning_rate": 8.414333882223791e-05, "loss": 3.0891, "step": 20222 }, { "epoch": 1.59, "learning_rate": 8.414255469301341e-05, "loss": 3.0252, "step": 20223 }, { "epoch": 1.59, "learning_rate": 8.414177056378892e-05, "loss": 3.2003, "step": 20224 }, { "epoch": 1.59, "learning_rate": 8.414098643456442e-05, "loss": 3.0202, "step": 20225 }, { "epoch": 1.59, "learning_rate": 8.414020230533992e-05, "loss": 2.9201, "step": 20226 }, { "epoch": 1.59, "learning_rate": 8.413941817611543e-05, "loss": 2.8708, "step": 20227 }, { "epoch": 1.59, "learning_rate": 8.413863404689093e-05, "loss": 2.798, "step": 20228 }, { "epoch": 1.59, "learning_rate": 8.413784991766644e-05, "loss": 2.9765, "step": 20229 }, { "epoch": 1.59, "learning_rate": 8.413706578844194e-05, "loss": 2.93, "step": 20230 }, { "epoch": 1.59, "learning_rate": 8.413628165921744e-05, "loss": 2.6607, "step": 20231 }, { "epoch": 1.59, "learning_rate": 8.413549752999295e-05, "loss": 3.1998, "step": 20232 }, { "epoch": 1.59, "learning_rate": 8.413471340076845e-05, "loss": 2.8207, "step": 20233 }, { "epoch": 1.59, "learning_rate": 8.413392927154395e-05, "loss": 2.6159, "step": 20234 }, { "epoch": 1.59, "learning_rate": 8.413314514231946e-05, "loss": 2.7726, "step": 20235 }, { "epoch": 1.59, "learning_rate": 8.413236101309496e-05, "loss": 2.8561, "step": 20236 }, { "epoch": 1.59, "learning_rate": 8.413157688387046e-05, "loss": 3.1643, "step": 20237 }, { "epoch": 1.59, "learning_rate": 8.413079275464597e-05, "loss": 2.9672, "step": 20238 }, { "epoch": 1.59, "learning_rate": 8.413000862542147e-05, "loss": 2.7225, "step": 20239 }, { "epoch": 1.59, "learning_rate": 8.412922449619698e-05, "loss": 3.011, "step": 20240 }, { "epoch": 1.59, "learning_rate": 8.412844036697248e-05, "loss": 2.9224, "step": 20241 }, { "epoch": 1.59, "learning_rate": 8.412765623774798e-05, "loss": 2.9015, "step": 20242 }, { "epoch": 1.59, "learning_rate": 8.412687210852349e-05, "loss": 2.8978, "step": 20243 }, { "epoch": 1.59, "learning_rate": 8.412608797929899e-05, "loss": 2.9872, "step": 20244 }, { "epoch": 1.59, "learning_rate": 8.41253038500745e-05, "loss": 3.0644, "step": 20245 }, { "epoch": 1.59, "learning_rate": 8.412451972085e-05, "loss": 2.7743, "step": 20246 }, { "epoch": 1.59, "learning_rate": 8.41237355916255e-05, "loss": 2.8025, "step": 20247 }, { "epoch": 1.59, "learning_rate": 8.4122951462401e-05, "loss": 3.2172, "step": 20248 }, { "epoch": 1.59, "learning_rate": 8.412216733317652e-05, "loss": 3.1383, "step": 20249 }, { "epoch": 1.59, "learning_rate": 8.412138320395203e-05, "loss": 2.7433, "step": 20250 }, { "epoch": 1.59, "learning_rate": 8.412059907472753e-05, "loss": 2.9045, "step": 20251 }, { "epoch": 1.59, "learning_rate": 8.411981494550303e-05, "loss": 2.7727, "step": 20252 }, { "epoch": 1.59, "learning_rate": 8.411903081627852e-05, "loss": 2.795, "step": 20253 }, { "epoch": 1.59, "learning_rate": 8.411824668705403e-05, "loss": 2.9399, "step": 20254 }, { "epoch": 1.59, "learning_rate": 8.411746255782953e-05, "loss": 2.7759, "step": 20255 }, { "epoch": 1.59, "learning_rate": 8.411667842860503e-05, "loss": 2.9031, "step": 20256 }, { "epoch": 1.59, "learning_rate": 8.411589429938054e-05, "loss": 2.6513, "step": 20257 }, { "epoch": 1.59, "learning_rate": 8.411511017015604e-05, "loss": 2.9065, "step": 20258 }, { "epoch": 1.59, "learning_rate": 8.411432604093155e-05, "loss": 2.9412, "step": 20259 }, { "epoch": 1.59, "learning_rate": 8.411354191170705e-05, "loss": 2.728, "step": 20260 }, { "epoch": 1.59, "learning_rate": 8.411275778248255e-05, "loss": 2.8729, "step": 20261 }, { "epoch": 1.59, "learning_rate": 8.411197365325806e-05, "loss": 2.9791, "step": 20262 }, { "epoch": 1.59, "learning_rate": 8.411118952403356e-05, "loss": 2.9812, "step": 20263 }, { "epoch": 1.59, "learning_rate": 8.411040539480906e-05, "loss": 2.9251, "step": 20264 }, { "epoch": 1.59, "learning_rate": 8.410962126558457e-05, "loss": 2.9899, "step": 20265 }, { "epoch": 1.59, "learning_rate": 8.410883713636007e-05, "loss": 3.0269, "step": 20266 }, { "epoch": 1.59, "learning_rate": 8.410805300713558e-05, "loss": 2.8413, "step": 20267 }, { "epoch": 1.59, "learning_rate": 8.410726887791108e-05, "loss": 2.8598, "step": 20268 }, { "epoch": 1.59, "learning_rate": 8.410648474868658e-05, "loss": 2.9817, "step": 20269 }, { "epoch": 1.59, "learning_rate": 8.410570061946209e-05, "loss": 2.9954, "step": 20270 }, { "epoch": 1.59, "learning_rate": 8.410491649023759e-05, "loss": 2.7342, "step": 20271 }, { "epoch": 1.59, "learning_rate": 8.41041323610131e-05, "loss": 2.842, "step": 20272 }, { "epoch": 1.59, "learning_rate": 8.41033482317886e-05, "loss": 2.8136, "step": 20273 }, { "epoch": 1.59, "learning_rate": 8.410256410256411e-05, "loss": 2.9862, "step": 20274 }, { "epoch": 1.59, "learning_rate": 8.410177997333962e-05, "loss": 3.1235, "step": 20275 }, { "epoch": 1.59, "learning_rate": 8.410099584411512e-05, "loss": 2.9177, "step": 20276 }, { "epoch": 1.59, "learning_rate": 8.410021171489063e-05, "loss": 2.8625, "step": 20277 }, { "epoch": 1.59, "learning_rate": 8.409942758566613e-05, "loss": 2.9106, "step": 20278 }, { "epoch": 1.59, "learning_rate": 8.409864345644163e-05, "loss": 2.9072, "step": 20279 }, { "epoch": 1.59, "learning_rate": 8.409785932721714e-05, "loss": 3.0707, "step": 20280 }, { "epoch": 1.59, "learning_rate": 8.409707519799264e-05, "loss": 2.7179, "step": 20281 }, { "epoch": 1.59, "learning_rate": 8.409629106876814e-05, "loss": 2.7749, "step": 20282 }, { "epoch": 1.59, "learning_rate": 8.409550693954365e-05, "loss": 3.0377, "step": 20283 }, { "epoch": 1.59, "learning_rate": 8.409472281031915e-05, "loss": 2.7561, "step": 20284 }, { "epoch": 1.59, "learning_rate": 8.409393868109465e-05, "loss": 2.9168, "step": 20285 }, { "epoch": 1.59, "learning_rate": 8.409315455187016e-05, "loss": 2.7553, "step": 20286 }, { "epoch": 1.59, "learning_rate": 8.409237042264566e-05, "loss": 2.8506, "step": 20287 }, { "epoch": 1.59, "learning_rate": 8.409158629342117e-05, "loss": 2.6796, "step": 20288 }, { "epoch": 1.59, "learning_rate": 8.409080216419666e-05, "loss": 2.7578, "step": 20289 }, { "epoch": 1.59, "learning_rate": 8.409001803497216e-05, "loss": 2.8788, "step": 20290 }, { "epoch": 1.59, "learning_rate": 8.408923390574766e-05, "loss": 3.0082, "step": 20291 }, { "epoch": 1.59, "learning_rate": 8.408844977652317e-05, "loss": 2.8768, "step": 20292 }, { "epoch": 1.59, "learning_rate": 8.408766564729867e-05, "loss": 2.7633, "step": 20293 }, { "epoch": 1.59, "learning_rate": 8.408688151807417e-05, "loss": 2.8549, "step": 20294 }, { "epoch": 1.59, "learning_rate": 8.408609738884968e-05, "loss": 3.0942, "step": 20295 }, { "epoch": 1.59, "learning_rate": 8.408531325962518e-05, "loss": 2.7901, "step": 20296 }, { "epoch": 1.59, "learning_rate": 8.408452913040069e-05, "loss": 2.9171, "step": 20297 }, { "epoch": 1.59, "learning_rate": 8.408374500117619e-05, "loss": 2.7435, "step": 20298 }, { "epoch": 1.59, "learning_rate": 8.408296087195169e-05, "loss": 3.0183, "step": 20299 }, { "epoch": 1.59, "learning_rate": 8.408217674272721e-05, "loss": 2.9779, "step": 20300 }, { "epoch": 1.59, "learning_rate": 8.408139261350271e-05, "loss": 2.9476, "step": 20301 }, { "epoch": 1.59, "learning_rate": 8.408060848427822e-05, "loss": 2.827, "step": 20302 }, { "epoch": 1.59, "learning_rate": 8.407982435505372e-05, "loss": 2.7405, "step": 20303 }, { "epoch": 1.59, "learning_rate": 8.407904022582922e-05, "loss": 2.7624, "step": 20304 }, { "epoch": 1.59, "learning_rate": 8.407825609660473e-05, "loss": 2.8042, "step": 20305 }, { "epoch": 1.59, "learning_rate": 8.407747196738023e-05, "loss": 2.8035, "step": 20306 }, { "epoch": 1.59, "learning_rate": 8.407668783815574e-05, "loss": 3.1404, "step": 20307 }, { "epoch": 1.59, "learning_rate": 8.407590370893124e-05, "loss": 3.1921, "step": 20308 }, { "epoch": 1.59, "learning_rate": 8.407511957970674e-05, "loss": 2.8995, "step": 20309 }, { "epoch": 1.59, "learning_rate": 8.407433545048225e-05, "loss": 2.9588, "step": 20310 }, { "epoch": 1.59, "learning_rate": 8.407355132125775e-05, "loss": 3.2757, "step": 20311 }, { "epoch": 1.59, "learning_rate": 8.407276719203325e-05, "loss": 2.9134, "step": 20312 }, { "epoch": 1.59, "learning_rate": 8.407198306280876e-05, "loss": 2.8074, "step": 20313 }, { "epoch": 1.59, "learning_rate": 8.407119893358426e-05, "loss": 3.0217, "step": 20314 }, { "epoch": 1.59, "learning_rate": 8.407041480435977e-05, "loss": 3.1277, "step": 20315 }, { "epoch": 1.59, "learning_rate": 8.406963067513527e-05, "loss": 2.8319, "step": 20316 }, { "epoch": 1.59, "learning_rate": 8.406884654591077e-05, "loss": 2.745, "step": 20317 }, { "epoch": 1.59, "learning_rate": 8.406806241668628e-05, "loss": 2.9626, "step": 20318 }, { "epoch": 1.59, "learning_rate": 8.406727828746178e-05, "loss": 2.9507, "step": 20319 }, { "epoch": 1.59, "learning_rate": 8.406649415823728e-05, "loss": 2.8541, "step": 20320 }, { "epoch": 1.59, "learning_rate": 8.406571002901279e-05, "loss": 2.6916, "step": 20321 }, { "epoch": 1.59, "learning_rate": 8.406492589978829e-05, "loss": 2.9432, "step": 20322 }, { "epoch": 1.59, "learning_rate": 8.40641417705638e-05, "loss": 2.8307, "step": 20323 }, { "epoch": 1.59, "learning_rate": 8.40633576413393e-05, "loss": 2.8054, "step": 20324 }, { "epoch": 1.59, "learning_rate": 8.406257351211479e-05, "loss": 2.9565, "step": 20325 }, { "epoch": 1.59, "learning_rate": 8.40617893828903e-05, "loss": 2.9739, "step": 20326 }, { "epoch": 1.59, "learning_rate": 8.406100525366581e-05, "loss": 3.2578, "step": 20327 }, { "epoch": 1.59, "learning_rate": 8.406022112444131e-05, "loss": 3.0186, "step": 20328 }, { "epoch": 1.59, "learning_rate": 8.405943699521682e-05, "loss": 3.2041, "step": 20329 }, { "epoch": 1.59, "learning_rate": 8.405865286599232e-05, "loss": 3.0497, "step": 20330 }, { "epoch": 1.59, "learning_rate": 8.405786873676782e-05, "loss": 2.867, "step": 20331 }, { "epoch": 1.59, "learning_rate": 8.405708460754333e-05, "loss": 2.9355, "step": 20332 }, { "epoch": 1.59, "learning_rate": 8.405630047831883e-05, "loss": 2.9907, "step": 20333 }, { "epoch": 1.59, "learning_rate": 8.405551634909433e-05, "loss": 2.7999, "step": 20334 }, { "epoch": 1.59, "learning_rate": 8.405473221986984e-05, "loss": 2.9909, "step": 20335 }, { "epoch": 1.59, "learning_rate": 8.405394809064534e-05, "loss": 2.8947, "step": 20336 }, { "epoch": 1.59, "learning_rate": 8.405316396142085e-05, "loss": 3.0846, "step": 20337 }, { "epoch": 1.59, "learning_rate": 8.405237983219635e-05, "loss": 3.0541, "step": 20338 }, { "epoch": 1.59, "learning_rate": 8.405159570297185e-05, "loss": 2.8282, "step": 20339 }, { "epoch": 1.59, "learning_rate": 8.405081157374736e-05, "loss": 2.6765, "step": 20340 }, { "epoch": 1.59, "learning_rate": 8.405002744452286e-05, "loss": 2.9507, "step": 20341 }, { "epoch": 1.6, "learning_rate": 8.404924331529836e-05, "loss": 2.6162, "step": 20342 }, { "epoch": 1.6, "learning_rate": 8.404845918607387e-05, "loss": 3.0542, "step": 20343 }, { "epoch": 1.6, "learning_rate": 8.404767505684937e-05, "loss": 2.7754, "step": 20344 }, { "epoch": 1.6, "learning_rate": 8.404689092762488e-05, "loss": 2.8728, "step": 20345 }, { "epoch": 1.6, "learning_rate": 8.404610679840038e-05, "loss": 2.7282, "step": 20346 }, { "epoch": 1.6, "learning_rate": 8.404532266917588e-05, "loss": 3.1552, "step": 20347 }, { "epoch": 1.6, "learning_rate": 8.404453853995139e-05, "loss": 2.9165, "step": 20348 }, { "epoch": 1.6, "learning_rate": 8.404375441072689e-05, "loss": 3.0382, "step": 20349 }, { "epoch": 1.6, "learning_rate": 8.40429702815024e-05, "loss": 2.9498, "step": 20350 }, { "epoch": 1.6, "learning_rate": 8.40421861522779e-05, "loss": 3.1172, "step": 20351 }, { "epoch": 1.6, "learning_rate": 8.40414020230534e-05, "loss": 2.8855, "step": 20352 }, { "epoch": 1.6, "learning_rate": 8.40406178938289e-05, "loss": 3.2285, "step": 20353 }, { "epoch": 1.6, "learning_rate": 8.403983376460441e-05, "loss": 3.0127, "step": 20354 }, { "epoch": 1.6, "learning_rate": 8.403904963537991e-05, "loss": 2.7848, "step": 20355 }, { "epoch": 1.6, "learning_rate": 8.403826550615542e-05, "loss": 2.8193, "step": 20356 }, { "epoch": 1.6, "learning_rate": 8.403748137693093e-05, "loss": 2.6539, "step": 20357 }, { "epoch": 1.6, "learning_rate": 8.403669724770644e-05, "loss": 2.8383, "step": 20358 }, { "epoch": 1.6, "learning_rate": 8.403591311848194e-05, "loss": 2.8301, "step": 20359 }, { "epoch": 1.6, "learning_rate": 8.403512898925744e-05, "loss": 2.7492, "step": 20360 }, { "epoch": 1.6, "learning_rate": 8.403434486003293e-05, "loss": 3.1032, "step": 20361 }, { "epoch": 1.6, "learning_rate": 8.403356073080844e-05, "loss": 2.8289, "step": 20362 }, { "epoch": 1.6, "learning_rate": 8.403277660158394e-05, "loss": 2.9913, "step": 20363 }, { "epoch": 1.6, "learning_rate": 8.403199247235945e-05, "loss": 2.6985, "step": 20364 }, { "epoch": 1.6, "learning_rate": 8.403120834313495e-05, "loss": 2.9948, "step": 20365 }, { "epoch": 1.6, "learning_rate": 8.403042421391045e-05, "loss": 2.9029, "step": 20366 }, { "epoch": 1.6, "learning_rate": 8.402964008468596e-05, "loss": 2.742, "step": 20367 }, { "epoch": 1.6, "learning_rate": 8.402885595546146e-05, "loss": 2.891, "step": 20368 }, { "epoch": 1.6, "learning_rate": 8.402807182623696e-05, "loss": 2.7241, "step": 20369 }, { "epoch": 1.6, "learning_rate": 8.402728769701247e-05, "loss": 2.7116, "step": 20370 }, { "epoch": 1.6, "learning_rate": 8.402650356778797e-05, "loss": 3.0051, "step": 20371 }, { "epoch": 1.6, "learning_rate": 8.402571943856347e-05, "loss": 2.7635, "step": 20372 }, { "epoch": 1.6, "learning_rate": 8.402493530933898e-05, "loss": 3.1523, "step": 20373 }, { "epoch": 1.6, "learning_rate": 8.402415118011448e-05, "loss": 2.9356, "step": 20374 }, { "epoch": 1.6, "learning_rate": 8.402336705088999e-05, "loss": 2.9604, "step": 20375 }, { "epoch": 1.6, "learning_rate": 8.402258292166549e-05, "loss": 2.9388, "step": 20376 }, { "epoch": 1.6, "learning_rate": 8.402179879244099e-05, "loss": 3.078, "step": 20377 }, { "epoch": 1.6, "learning_rate": 8.40210146632165e-05, "loss": 2.8015, "step": 20378 }, { "epoch": 1.6, "learning_rate": 8.4020230533992e-05, "loss": 2.9182, "step": 20379 }, { "epoch": 1.6, "learning_rate": 8.40194464047675e-05, "loss": 2.8287, "step": 20380 }, { "epoch": 1.6, "learning_rate": 8.401866227554301e-05, "loss": 2.7801, "step": 20381 }, { "epoch": 1.6, "learning_rate": 8.401787814631853e-05, "loss": 3.2397, "step": 20382 }, { "epoch": 1.6, "learning_rate": 8.401709401709403e-05, "loss": 2.6589, "step": 20383 }, { "epoch": 1.6, "learning_rate": 8.401630988786953e-05, "loss": 2.8286, "step": 20384 }, { "epoch": 1.6, "learning_rate": 8.401552575864504e-05, "loss": 2.8294, "step": 20385 }, { "epoch": 1.6, "learning_rate": 8.401474162942054e-05, "loss": 2.842, "step": 20386 }, { "epoch": 1.6, "learning_rate": 8.401395750019604e-05, "loss": 2.8369, "step": 20387 }, { "epoch": 1.6, "learning_rate": 8.401317337097155e-05, "loss": 2.7463, "step": 20388 }, { "epoch": 1.6, "learning_rate": 8.401238924174705e-05, "loss": 3.1785, "step": 20389 }, { "epoch": 1.6, "learning_rate": 8.401160511252255e-05, "loss": 3.0238, "step": 20390 }, { "epoch": 1.6, "learning_rate": 8.401082098329806e-05, "loss": 2.8315, "step": 20391 }, { "epoch": 1.6, "learning_rate": 8.401003685407356e-05, "loss": 2.8341, "step": 20392 }, { "epoch": 1.6, "learning_rate": 8.400925272484907e-05, "loss": 2.9106, "step": 20393 }, { "epoch": 1.6, "learning_rate": 8.400846859562457e-05, "loss": 3.0788, "step": 20394 }, { "epoch": 1.6, "learning_rate": 8.400768446640007e-05, "loss": 2.8556, "step": 20395 }, { "epoch": 1.6, "learning_rate": 8.400690033717558e-05, "loss": 2.8643, "step": 20396 }, { "epoch": 1.6, "learning_rate": 8.400611620795107e-05, "loss": 2.883, "step": 20397 }, { "epoch": 1.6, "learning_rate": 8.400533207872657e-05, "loss": 3.0254, "step": 20398 }, { "epoch": 1.6, "learning_rate": 8.400454794950207e-05, "loss": 3.0254, "step": 20399 }, { "epoch": 1.6, "learning_rate": 8.400376382027758e-05, "loss": 2.9673, "step": 20400 }, { "epoch": 1.6, "learning_rate": 8.400297969105308e-05, "loss": 2.8674, "step": 20401 }, { "epoch": 1.6, "learning_rate": 8.400219556182859e-05, "loss": 2.7678, "step": 20402 }, { "epoch": 1.6, "learning_rate": 8.400141143260409e-05, "loss": 2.8336, "step": 20403 }, { "epoch": 1.6, "learning_rate": 8.400062730337959e-05, "loss": 3.1516, "step": 20404 }, { "epoch": 1.6, "learning_rate": 8.39998431741551e-05, "loss": 2.6267, "step": 20405 }, { "epoch": 1.6, "learning_rate": 8.39990590449306e-05, "loss": 2.7369, "step": 20406 }, { "epoch": 1.6, "learning_rate": 8.39982749157061e-05, "loss": 3.072, "step": 20407 }, { "epoch": 1.6, "learning_rate": 8.399749078648162e-05, "loss": 2.8428, "step": 20408 }, { "epoch": 1.6, "learning_rate": 8.399670665725712e-05, "loss": 2.8915, "step": 20409 }, { "epoch": 1.6, "learning_rate": 8.399592252803263e-05, "loss": 3.0223, "step": 20410 }, { "epoch": 1.6, "learning_rate": 8.399513839880813e-05, "loss": 2.7959, "step": 20411 }, { "epoch": 1.6, "learning_rate": 8.399435426958364e-05, "loss": 2.7945, "step": 20412 }, { "epoch": 1.6, "learning_rate": 8.399357014035914e-05, "loss": 2.7781, "step": 20413 }, { "epoch": 1.6, "learning_rate": 8.399278601113464e-05, "loss": 3.0293, "step": 20414 }, { "epoch": 1.6, "learning_rate": 8.399200188191015e-05, "loss": 2.8802, "step": 20415 }, { "epoch": 1.6, "learning_rate": 8.399121775268565e-05, "loss": 2.8728, "step": 20416 }, { "epoch": 1.6, "learning_rate": 8.399043362346115e-05, "loss": 3.0207, "step": 20417 }, { "epoch": 1.6, "learning_rate": 8.398964949423666e-05, "loss": 2.9647, "step": 20418 }, { "epoch": 1.6, "learning_rate": 8.398886536501216e-05, "loss": 3.0079, "step": 20419 }, { "epoch": 1.6, "learning_rate": 8.398808123578766e-05, "loss": 2.7035, "step": 20420 }, { "epoch": 1.6, "learning_rate": 8.398729710656317e-05, "loss": 3.0488, "step": 20421 }, { "epoch": 1.6, "learning_rate": 8.398651297733867e-05, "loss": 2.7172, "step": 20422 }, { "epoch": 1.6, "learning_rate": 8.398572884811418e-05, "loss": 2.8062, "step": 20423 }, { "epoch": 1.6, "learning_rate": 8.398494471888968e-05, "loss": 2.8548, "step": 20424 }, { "epoch": 1.6, "learning_rate": 8.398416058966518e-05, "loss": 2.7345, "step": 20425 }, { "epoch": 1.6, "learning_rate": 8.398337646044069e-05, "loss": 2.821, "step": 20426 }, { "epoch": 1.6, "learning_rate": 8.398259233121619e-05, "loss": 2.9188, "step": 20427 }, { "epoch": 1.6, "learning_rate": 8.39818082019917e-05, "loss": 2.8977, "step": 20428 }, { "epoch": 1.6, "learning_rate": 8.39810240727672e-05, "loss": 3.0115, "step": 20429 }, { "epoch": 1.6, "learning_rate": 8.39802399435427e-05, "loss": 2.7587, "step": 20430 }, { "epoch": 1.6, "learning_rate": 8.39794558143182e-05, "loss": 2.7946, "step": 20431 }, { "epoch": 1.6, "learning_rate": 8.397867168509371e-05, "loss": 3.0057, "step": 20432 }, { "epoch": 1.6, "learning_rate": 8.39778875558692e-05, "loss": 2.9383, "step": 20433 }, { "epoch": 1.6, "learning_rate": 8.397710342664472e-05, "loss": 2.954, "step": 20434 }, { "epoch": 1.6, "learning_rate": 8.397631929742022e-05, "loss": 3.0086, "step": 20435 }, { "epoch": 1.6, "learning_rate": 8.397553516819572e-05, "loss": 2.7821, "step": 20436 }, { "epoch": 1.6, "learning_rate": 8.397475103897123e-05, "loss": 2.8571, "step": 20437 }, { "epoch": 1.6, "learning_rate": 8.397396690974673e-05, "loss": 3.0059, "step": 20438 }, { "epoch": 1.6, "learning_rate": 8.397318278052223e-05, "loss": 3.2244, "step": 20439 }, { "epoch": 1.6, "learning_rate": 8.397239865129774e-05, "loss": 2.7076, "step": 20440 }, { "epoch": 1.6, "learning_rate": 8.397161452207324e-05, "loss": 3.0298, "step": 20441 }, { "epoch": 1.6, "learning_rate": 8.397083039284875e-05, "loss": 3.1429, "step": 20442 }, { "epoch": 1.6, "learning_rate": 8.397004626362425e-05, "loss": 2.8908, "step": 20443 }, { "epoch": 1.6, "learning_rate": 8.396926213439975e-05, "loss": 2.6884, "step": 20444 }, { "epoch": 1.6, "learning_rate": 8.396847800517526e-05, "loss": 2.8685, "step": 20445 }, { "epoch": 1.6, "learning_rate": 8.396769387595076e-05, "loss": 3.0752, "step": 20446 }, { "epoch": 1.6, "learning_rate": 8.396690974672626e-05, "loss": 2.8673, "step": 20447 }, { "epoch": 1.6, "learning_rate": 8.396612561750177e-05, "loss": 2.9594, "step": 20448 }, { "epoch": 1.6, "learning_rate": 8.396534148827727e-05, "loss": 2.9162, "step": 20449 }, { "epoch": 1.6, "learning_rate": 8.396455735905278e-05, "loss": 2.8317, "step": 20450 }, { "epoch": 1.6, "learning_rate": 8.396377322982828e-05, "loss": 2.9134, "step": 20451 }, { "epoch": 1.6, "learning_rate": 8.396298910060378e-05, "loss": 2.8175, "step": 20452 }, { "epoch": 1.6, "learning_rate": 8.396220497137929e-05, "loss": 3.1439, "step": 20453 }, { "epoch": 1.6, "learning_rate": 8.396142084215479e-05, "loss": 3.013, "step": 20454 }, { "epoch": 1.6, "learning_rate": 8.39606367129303e-05, "loss": 2.9679, "step": 20455 }, { "epoch": 1.6, "learning_rate": 8.39598525837058e-05, "loss": 2.8591, "step": 20456 }, { "epoch": 1.6, "learning_rate": 8.39590684544813e-05, "loss": 2.7722, "step": 20457 }, { "epoch": 1.6, "learning_rate": 8.39582843252568e-05, "loss": 2.6912, "step": 20458 }, { "epoch": 1.6, "learning_rate": 8.395750019603231e-05, "loss": 2.7113, "step": 20459 }, { "epoch": 1.6, "learning_rate": 8.395671606680781e-05, "loss": 2.8604, "step": 20460 }, { "epoch": 1.6, "learning_rate": 8.395593193758332e-05, "loss": 3.1117, "step": 20461 }, { "epoch": 1.6, "learning_rate": 8.395514780835882e-05, "loss": 3.267, "step": 20462 }, { "epoch": 1.6, "learning_rate": 8.395436367913432e-05, "loss": 2.7225, "step": 20463 }, { "epoch": 1.6, "learning_rate": 8.395357954990983e-05, "loss": 2.9043, "step": 20464 }, { "epoch": 1.6, "learning_rate": 8.395279542068534e-05, "loss": 3.029, "step": 20465 }, { "epoch": 1.6, "learning_rate": 8.395201129146085e-05, "loss": 2.8346, "step": 20466 }, { "epoch": 1.6, "learning_rate": 8.395122716223635e-05, "loss": 2.8626, "step": 20467 }, { "epoch": 1.6, "learning_rate": 8.395044303301185e-05, "loss": 2.7788, "step": 20468 }, { "epoch": 1.61, "learning_rate": 8.394965890378735e-05, "loss": 3.1007, "step": 20469 }, { "epoch": 1.61, "learning_rate": 8.394887477456285e-05, "loss": 3.0167, "step": 20470 }, { "epoch": 1.61, "learning_rate": 8.394809064533835e-05, "loss": 2.8789, "step": 20471 }, { "epoch": 1.61, "learning_rate": 8.394730651611386e-05, "loss": 2.9693, "step": 20472 }, { "epoch": 1.61, "learning_rate": 8.394652238688936e-05, "loss": 2.8495, "step": 20473 }, { "epoch": 1.61, "learning_rate": 8.394573825766486e-05, "loss": 2.8419, "step": 20474 }, { "epoch": 1.61, "learning_rate": 8.394495412844037e-05, "loss": 2.8288, "step": 20475 }, { "epoch": 1.61, "learning_rate": 8.394416999921587e-05, "loss": 2.9611, "step": 20476 }, { "epoch": 1.61, "learning_rate": 8.394338586999137e-05, "loss": 2.7784, "step": 20477 }, { "epoch": 1.61, "learning_rate": 8.394260174076688e-05, "loss": 2.9214, "step": 20478 }, { "epoch": 1.61, "learning_rate": 8.394181761154238e-05, "loss": 2.7552, "step": 20479 }, { "epoch": 1.61, "learning_rate": 8.394103348231789e-05, "loss": 2.8688, "step": 20480 }, { "epoch": 1.61, "learning_rate": 8.394024935309339e-05, "loss": 3.0687, "step": 20481 }, { "epoch": 1.61, "learning_rate": 8.393946522386889e-05, "loss": 2.8897, "step": 20482 }, { "epoch": 1.61, "learning_rate": 8.39386810946444e-05, "loss": 2.7857, "step": 20483 }, { "epoch": 1.61, "learning_rate": 8.39378969654199e-05, "loss": 3.0488, "step": 20484 }, { "epoch": 1.61, "learning_rate": 8.39371128361954e-05, "loss": 2.7963, "step": 20485 }, { "epoch": 1.61, "learning_rate": 8.393632870697091e-05, "loss": 2.7173, "step": 20486 }, { "epoch": 1.61, "learning_rate": 8.393554457774641e-05, "loss": 2.9004, "step": 20487 }, { "epoch": 1.61, "learning_rate": 8.393476044852191e-05, "loss": 3.1144, "step": 20488 }, { "epoch": 1.61, "learning_rate": 8.393397631929742e-05, "loss": 2.8478, "step": 20489 }, { "epoch": 1.61, "learning_rate": 8.393319219007292e-05, "loss": 2.843, "step": 20490 }, { "epoch": 1.61, "learning_rate": 8.393240806084844e-05, "loss": 3.0722, "step": 20491 }, { "epoch": 1.61, "learning_rate": 8.393162393162394e-05, "loss": 2.8436, "step": 20492 }, { "epoch": 1.61, "learning_rate": 8.393083980239945e-05, "loss": 2.7486, "step": 20493 }, { "epoch": 1.61, "learning_rate": 8.393005567317495e-05, "loss": 2.8249, "step": 20494 }, { "epoch": 1.61, "learning_rate": 8.392927154395045e-05, "loss": 2.8912, "step": 20495 }, { "epoch": 1.61, "learning_rate": 8.392848741472596e-05, "loss": 2.8469, "step": 20496 }, { "epoch": 1.61, "learning_rate": 8.392770328550146e-05, "loss": 2.9349, "step": 20497 }, { "epoch": 1.61, "learning_rate": 8.392691915627697e-05, "loss": 2.8603, "step": 20498 }, { "epoch": 1.61, "learning_rate": 8.392613502705247e-05, "loss": 3.1069, "step": 20499 }, { "epoch": 1.61, "learning_rate": 8.392535089782797e-05, "loss": 2.8925, "step": 20500 }, { "epoch": 1.61, "learning_rate": 8.392456676860348e-05, "loss": 2.9292, "step": 20501 }, { "epoch": 1.61, "learning_rate": 8.392378263937898e-05, "loss": 2.8117, "step": 20502 }, { "epoch": 1.61, "learning_rate": 8.392299851015448e-05, "loss": 2.8053, "step": 20503 }, { "epoch": 1.61, "learning_rate": 8.392221438092999e-05, "loss": 3.0364, "step": 20504 }, { "epoch": 1.61, "learning_rate": 8.392143025170548e-05, "loss": 2.7561, "step": 20505 }, { "epoch": 1.61, "learning_rate": 8.392064612248098e-05, "loss": 3.0078, "step": 20506 }, { "epoch": 1.61, "learning_rate": 8.391986199325648e-05, "loss": 2.6912, "step": 20507 }, { "epoch": 1.61, "learning_rate": 8.391907786403199e-05, "loss": 2.7524, "step": 20508 }, { "epoch": 1.61, "learning_rate": 8.391829373480749e-05, "loss": 2.778, "step": 20509 }, { "epoch": 1.61, "learning_rate": 8.3917509605583e-05, "loss": 2.7151, "step": 20510 }, { "epoch": 1.61, "learning_rate": 8.39167254763585e-05, "loss": 2.8506, "step": 20511 }, { "epoch": 1.61, "learning_rate": 8.3915941347134e-05, "loss": 3.1116, "step": 20512 }, { "epoch": 1.61, "learning_rate": 8.391515721790951e-05, "loss": 2.7343, "step": 20513 }, { "epoch": 1.61, "learning_rate": 8.391437308868501e-05, "loss": 3.0025, "step": 20514 }, { "epoch": 1.61, "learning_rate": 8.391358895946051e-05, "loss": 2.7763, "step": 20515 }, { "epoch": 1.61, "learning_rate": 8.391280483023603e-05, "loss": 2.9855, "step": 20516 }, { "epoch": 1.61, "learning_rate": 8.391202070101154e-05, "loss": 2.7951, "step": 20517 }, { "epoch": 1.61, "learning_rate": 8.391123657178704e-05, "loss": 2.7961, "step": 20518 }, { "epoch": 1.61, "learning_rate": 8.391045244256254e-05, "loss": 2.7745, "step": 20519 }, { "epoch": 1.61, "learning_rate": 8.390966831333805e-05, "loss": 2.8422, "step": 20520 }, { "epoch": 1.61, "learning_rate": 8.390888418411355e-05, "loss": 2.8851, "step": 20521 }, { "epoch": 1.61, "learning_rate": 8.390810005488905e-05, "loss": 3.2332, "step": 20522 }, { "epoch": 1.61, "learning_rate": 8.390731592566456e-05, "loss": 2.9033, "step": 20523 }, { "epoch": 1.61, "learning_rate": 8.390653179644006e-05, "loss": 2.8536, "step": 20524 }, { "epoch": 1.61, "learning_rate": 8.390574766721556e-05, "loss": 2.967, "step": 20525 }, { "epoch": 1.61, "learning_rate": 8.390496353799107e-05, "loss": 3.1604, "step": 20526 }, { "epoch": 1.61, "learning_rate": 8.390417940876657e-05, "loss": 3.2687, "step": 20527 }, { "epoch": 1.61, "learning_rate": 8.390339527954208e-05, "loss": 3.1356, "step": 20528 }, { "epoch": 1.61, "learning_rate": 8.390261115031758e-05, "loss": 3.0201, "step": 20529 }, { "epoch": 1.61, "learning_rate": 8.390182702109308e-05, "loss": 2.9942, "step": 20530 }, { "epoch": 1.61, "learning_rate": 8.390104289186859e-05, "loss": 2.652, "step": 20531 }, { "epoch": 1.61, "learning_rate": 8.390025876264409e-05, "loss": 2.9704, "step": 20532 }, { "epoch": 1.61, "learning_rate": 8.38994746334196e-05, "loss": 2.8467, "step": 20533 }, { "epoch": 1.61, "learning_rate": 8.38986905041951e-05, "loss": 3.1674, "step": 20534 }, { "epoch": 1.61, "learning_rate": 8.38979063749706e-05, "loss": 3.0185, "step": 20535 }, { "epoch": 1.61, "learning_rate": 8.38971222457461e-05, "loss": 2.7956, "step": 20536 }, { "epoch": 1.61, "learning_rate": 8.389633811652161e-05, "loss": 2.8648, "step": 20537 }, { "epoch": 1.61, "learning_rate": 8.389555398729711e-05, "loss": 2.9571, "step": 20538 }, { "epoch": 1.61, "learning_rate": 8.389476985807262e-05, "loss": 2.8346, "step": 20539 }, { "epoch": 1.61, "learning_rate": 8.389398572884812e-05, "loss": 2.7895, "step": 20540 }, { "epoch": 1.61, "learning_rate": 8.389320159962361e-05, "loss": 2.8062, "step": 20541 }, { "epoch": 1.61, "learning_rate": 8.389241747039913e-05, "loss": 2.9853, "step": 20542 }, { "epoch": 1.61, "learning_rate": 8.389163334117463e-05, "loss": 3.1384, "step": 20543 }, { "epoch": 1.61, "learning_rate": 8.389084921195013e-05, "loss": 2.8664, "step": 20544 }, { "epoch": 1.61, "learning_rate": 8.389006508272564e-05, "loss": 3.0004, "step": 20545 }, { "epoch": 1.61, "learning_rate": 8.388928095350114e-05, "loss": 2.7022, "step": 20546 }, { "epoch": 1.61, "learning_rate": 8.388849682427665e-05, "loss": 2.824, "step": 20547 }, { "epoch": 1.61, "learning_rate": 8.388771269505215e-05, "loss": 2.931, "step": 20548 }, { "epoch": 1.61, "learning_rate": 8.388692856582765e-05, "loss": 2.7661, "step": 20549 }, { "epoch": 1.61, "learning_rate": 8.388614443660316e-05, "loss": 2.8569, "step": 20550 }, { "epoch": 1.61, "learning_rate": 8.388536030737866e-05, "loss": 2.8066, "step": 20551 }, { "epoch": 1.61, "learning_rate": 8.388457617815416e-05, "loss": 2.898, "step": 20552 }, { "epoch": 1.61, "learning_rate": 8.388379204892967e-05, "loss": 2.8388, "step": 20553 }, { "epoch": 1.61, "learning_rate": 8.388300791970517e-05, "loss": 2.8444, "step": 20554 }, { "epoch": 1.61, "learning_rate": 8.388222379048067e-05, "loss": 3.0787, "step": 20555 }, { "epoch": 1.61, "learning_rate": 8.388143966125618e-05, "loss": 2.7661, "step": 20556 }, { "epoch": 1.61, "learning_rate": 8.388065553203168e-05, "loss": 2.8042, "step": 20557 }, { "epoch": 1.61, "learning_rate": 8.387987140280719e-05, "loss": 2.6367, "step": 20558 }, { "epoch": 1.61, "learning_rate": 8.387908727358269e-05, "loss": 2.7071, "step": 20559 }, { "epoch": 1.61, "learning_rate": 8.38783031443582e-05, "loss": 3.0955, "step": 20560 }, { "epoch": 1.61, "learning_rate": 8.38775190151337e-05, "loss": 3.0115, "step": 20561 }, { "epoch": 1.61, "learning_rate": 8.38767348859092e-05, "loss": 2.8589, "step": 20562 }, { "epoch": 1.61, "learning_rate": 8.38759507566847e-05, "loss": 2.8675, "step": 20563 }, { "epoch": 1.61, "learning_rate": 8.387516662746021e-05, "loss": 2.6761, "step": 20564 }, { "epoch": 1.61, "learning_rate": 8.387438249823571e-05, "loss": 2.734, "step": 20565 }, { "epoch": 1.61, "learning_rate": 8.387359836901122e-05, "loss": 2.9492, "step": 20566 }, { "epoch": 1.61, "learning_rate": 8.387281423978672e-05, "loss": 2.8765, "step": 20567 }, { "epoch": 1.61, "learning_rate": 8.387203011056222e-05, "loss": 3.0195, "step": 20568 }, { "epoch": 1.61, "learning_rate": 8.387124598133773e-05, "loss": 2.8964, "step": 20569 }, { "epoch": 1.61, "learning_rate": 8.387046185211323e-05, "loss": 2.7801, "step": 20570 }, { "epoch": 1.61, "learning_rate": 8.386967772288873e-05, "loss": 2.8979, "step": 20571 }, { "epoch": 1.61, "learning_rate": 8.386889359366424e-05, "loss": 2.8998, "step": 20572 }, { "epoch": 1.61, "learning_rate": 8.386810946443975e-05, "loss": 2.7275, "step": 20573 }, { "epoch": 1.61, "learning_rate": 8.386732533521526e-05, "loss": 2.8121, "step": 20574 }, { "epoch": 1.61, "learning_rate": 8.386654120599076e-05, "loss": 2.9527, "step": 20575 }, { "epoch": 1.61, "learning_rate": 8.386575707676627e-05, "loss": 2.952, "step": 20576 }, { "epoch": 1.61, "learning_rate": 8.386497294754176e-05, "loss": 2.867, "step": 20577 }, { "epoch": 1.61, "learning_rate": 8.386418881831726e-05, "loss": 3.0157, "step": 20578 }, { "epoch": 1.61, "learning_rate": 8.386340468909276e-05, "loss": 2.7857, "step": 20579 }, { "epoch": 1.61, "learning_rate": 8.386262055986827e-05, "loss": 2.8489, "step": 20580 }, { "epoch": 1.61, "learning_rate": 8.386183643064377e-05, "loss": 2.9815, "step": 20581 }, { "epoch": 1.61, "learning_rate": 8.386105230141927e-05, "loss": 2.8777, "step": 20582 }, { "epoch": 1.61, "learning_rate": 8.386026817219478e-05, "loss": 2.9954, "step": 20583 }, { "epoch": 1.61, "learning_rate": 8.385948404297028e-05, "loss": 2.9006, "step": 20584 }, { "epoch": 1.61, "learning_rate": 8.385869991374579e-05, "loss": 2.6617, "step": 20585 }, { "epoch": 1.61, "learning_rate": 8.385791578452129e-05, "loss": 2.8733, "step": 20586 }, { "epoch": 1.61, "learning_rate": 8.385713165529679e-05, "loss": 2.9005, "step": 20587 }, { "epoch": 1.61, "learning_rate": 8.38563475260723e-05, "loss": 2.7, "step": 20588 }, { "epoch": 1.61, "learning_rate": 8.38555633968478e-05, "loss": 2.8088, "step": 20589 }, { "epoch": 1.61, "learning_rate": 8.38547792676233e-05, "loss": 2.6652, "step": 20590 }, { "epoch": 1.61, "learning_rate": 8.385399513839881e-05, "loss": 2.9668, "step": 20591 }, { "epoch": 1.61, "learning_rate": 8.385321100917431e-05, "loss": 2.8838, "step": 20592 }, { "epoch": 1.61, "learning_rate": 8.385242687994981e-05, "loss": 2.9333, "step": 20593 }, { "epoch": 1.61, "learning_rate": 8.385164275072532e-05, "loss": 2.7275, "step": 20594 }, { "epoch": 1.61, "learning_rate": 8.385085862150082e-05, "loss": 2.8074, "step": 20595 }, { "epoch": 1.61, "learning_rate": 8.385007449227633e-05, "loss": 2.8722, "step": 20596 }, { "epoch": 1.62, "learning_rate": 8.384929036305183e-05, "loss": 3.1421, "step": 20597 }, { "epoch": 1.62, "learning_rate": 8.384850623382733e-05, "loss": 2.9751, "step": 20598 }, { "epoch": 1.62, "learning_rate": 8.384772210460285e-05, "loss": 2.9044, "step": 20599 }, { "epoch": 1.62, "learning_rate": 8.384693797537835e-05, "loss": 2.8489, "step": 20600 }, { "epoch": 1.62, "learning_rate": 8.384615384615386e-05, "loss": 3.0511, "step": 20601 }, { "epoch": 1.62, "learning_rate": 8.384536971692936e-05, "loss": 2.8196, "step": 20602 }, { "epoch": 1.62, "learning_rate": 8.384458558770486e-05, "loss": 2.8752, "step": 20603 }, { "epoch": 1.62, "learning_rate": 8.384380145848037e-05, "loss": 2.8641, "step": 20604 }, { "epoch": 1.62, "learning_rate": 8.384301732925587e-05, "loss": 3.0337, "step": 20605 }, { "epoch": 1.62, "learning_rate": 8.384223320003138e-05, "loss": 3.0549, "step": 20606 }, { "epoch": 1.62, "learning_rate": 8.384144907080688e-05, "loss": 2.911, "step": 20607 }, { "epoch": 1.62, "learning_rate": 8.384066494158238e-05, "loss": 3.1215, "step": 20608 }, { "epoch": 1.62, "learning_rate": 8.383988081235789e-05, "loss": 3.018, "step": 20609 }, { "epoch": 1.62, "learning_rate": 8.383909668313339e-05, "loss": 2.8052, "step": 20610 }, { "epoch": 1.62, "learning_rate": 8.38383125539089e-05, "loss": 2.7466, "step": 20611 }, { "epoch": 1.62, "learning_rate": 8.38375284246844e-05, "loss": 3.0048, "step": 20612 }, { "epoch": 1.62, "learning_rate": 8.383674429545989e-05, "loss": 2.9447, "step": 20613 }, { "epoch": 1.62, "learning_rate": 8.383596016623539e-05, "loss": 2.7848, "step": 20614 }, { "epoch": 1.62, "learning_rate": 8.38351760370109e-05, "loss": 2.7734, "step": 20615 }, { "epoch": 1.62, "learning_rate": 8.38343919077864e-05, "loss": 2.8315, "step": 20616 }, { "epoch": 1.62, "learning_rate": 8.38336077785619e-05, "loss": 3.0978, "step": 20617 }, { "epoch": 1.62, "learning_rate": 8.38328236493374e-05, "loss": 2.7912, "step": 20618 }, { "epoch": 1.62, "learning_rate": 8.383203952011291e-05, "loss": 2.6519, "step": 20619 }, { "epoch": 1.62, "learning_rate": 8.383125539088841e-05, "loss": 2.7038, "step": 20620 }, { "epoch": 1.62, "learning_rate": 8.383047126166392e-05, "loss": 2.8568, "step": 20621 }, { "epoch": 1.62, "learning_rate": 8.382968713243942e-05, "loss": 2.9262, "step": 20622 }, { "epoch": 1.62, "learning_rate": 8.382890300321492e-05, "loss": 2.8715, "step": 20623 }, { "epoch": 1.62, "learning_rate": 8.382811887399044e-05, "loss": 2.8781, "step": 20624 }, { "epoch": 1.62, "learning_rate": 8.382733474476595e-05, "loss": 2.7389, "step": 20625 }, { "epoch": 1.62, "learning_rate": 8.382655061554145e-05, "loss": 2.9056, "step": 20626 }, { "epoch": 1.62, "learning_rate": 8.382576648631695e-05, "loss": 2.615, "step": 20627 }, { "epoch": 1.62, "learning_rate": 8.382498235709246e-05, "loss": 2.8919, "step": 20628 }, { "epoch": 1.62, "learning_rate": 8.382419822786796e-05, "loss": 2.9727, "step": 20629 }, { "epoch": 1.62, "learning_rate": 8.382341409864346e-05, "loss": 2.812, "step": 20630 }, { "epoch": 1.62, "learning_rate": 8.382262996941897e-05, "loss": 2.8565, "step": 20631 }, { "epoch": 1.62, "learning_rate": 8.382184584019447e-05, "loss": 3.0715, "step": 20632 }, { "epoch": 1.62, "learning_rate": 8.382106171096998e-05, "loss": 2.9157, "step": 20633 }, { "epoch": 1.62, "learning_rate": 8.382027758174548e-05, "loss": 2.8135, "step": 20634 }, { "epoch": 1.62, "learning_rate": 8.381949345252098e-05, "loss": 2.9207, "step": 20635 }, { "epoch": 1.62, "learning_rate": 8.381870932329649e-05, "loss": 2.8362, "step": 20636 }, { "epoch": 1.62, "learning_rate": 8.381792519407199e-05, "loss": 2.876, "step": 20637 }, { "epoch": 1.62, "learning_rate": 8.38171410648475e-05, "loss": 2.9348, "step": 20638 }, { "epoch": 1.62, "learning_rate": 8.3816356935623e-05, "loss": 2.8111, "step": 20639 }, { "epoch": 1.62, "learning_rate": 8.38155728063985e-05, "loss": 2.995, "step": 20640 }, { "epoch": 1.62, "learning_rate": 8.3814788677174e-05, "loss": 2.6992, "step": 20641 }, { "epoch": 1.62, "learning_rate": 8.381400454794951e-05, "loss": 2.8273, "step": 20642 }, { "epoch": 1.62, "learning_rate": 8.381322041872501e-05, "loss": 3.0526, "step": 20643 }, { "epoch": 1.62, "learning_rate": 8.381243628950052e-05, "loss": 2.9413, "step": 20644 }, { "epoch": 1.62, "learning_rate": 8.381165216027602e-05, "loss": 2.9476, "step": 20645 }, { "epoch": 1.62, "learning_rate": 8.381086803105152e-05, "loss": 2.9201, "step": 20646 }, { "epoch": 1.62, "learning_rate": 8.381008390182703e-05, "loss": 2.743, "step": 20647 }, { "epoch": 1.62, "learning_rate": 8.380929977260253e-05, "loss": 2.9406, "step": 20648 }, { "epoch": 1.62, "learning_rate": 8.380851564337802e-05, "loss": 2.7808, "step": 20649 }, { "epoch": 1.62, "learning_rate": 8.380773151415354e-05, "loss": 2.7333, "step": 20650 }, { "epoch": 1.62, "learning_rate": 8.380694738492904e-05, "loss": 2.9953, "step": 20651 }, { "epoch": 1.62, "learning_rate": 8.380616325570455e-05, "loss": 3.0442, "step": 20652 }, { "epoch": 1.62, "learning_rate": 8.380537912648005e-05, "loss": 2.8076, "step": 20653 }, { "epoch": 1.62, "learning_rate": 8.380459499725555e-05, "loss": 2.8578, "step": 20654 }, { "epoch": 1.62, "learning_rate": 8.380381086803106e-05, "loss": 2.8557, "step": 20655 }, { "epoch": 1.62, "learning_rate": 8.380302673880656e-05, "loss": 2.9677, "step": 20656 }, { "epoch": 1.62, "learning_rate": 8.380224260958206e-05, "loss": 3.2299, "step": 20657 }, { "epoch": 1.62, "learning_rate": 8.380145848035757e-05, "loss": 3.0768, "step": 20658 }, { "epoch": 1.62, "learning_rate": 8.380067435113307e-05, "loss": 2.9533, "step": 20659 }, { "epoch": 1.62, "learning_rate": 8.379989022190857e-05, "loss": 2.8494, "step": 20660 }, { "epoch": 1.62, "learning_rate": 8.379910609268408e-05, "loss": 2.9931, "step": 20661 }, { "epoch": 1.62, "learning_rate": 8.379832196345958e-05, "loss": 2.7375, "step": 20662 }, { "epoch": 1.62, "learning_rate": 8.379753783423509e-05, "loss": 3.0837, "step": 20663 }, { "epoch": 1.62, "learning_rate": 8.379675370501059e-05, "loss": 2.8112, "step": 20664 }, { "epoch": 1.62, "learning_rate": 8.379596957578609e-05, "loss": 3.0505, "step": 20665 }, { "epoch": 1.62, "learning_rate": 8.37951854465616e-05, "loss": 2.8046, "step": 20666 }, { "epoch": 1.62, "learning_rate": 8.37944013173371e-05, "loss": 2.8486, "step": 20667 }, { "epoch": 1.62, "learning_rate": 8.37936171881126e-05, "loss": 2.8832, "step": 20668 }, { "epoch": 1.62, "learning_rate": 8.379283305888811e-05, "loss": 2.8263, "step": 20669 }, { "epoch": 1.62, "learning_rate": 8.379204892966361e-05, "loss": 3.0323, "step": 20670 }, { "epoch": 1.62, "learning_rate": 8.379126480043912e-05, "loss": 3.127, "step": 20671 }, { "epoch": 1.62, "learning_rate": 8.379048067121462e-05, "loss": 2.7256, "step": 20672 }, { "epoch": 1.62, "learning_rate": 8.378969654199012e-05, "loss": 2.8178, "step": 20673 }, { "epoch": 1.62, "learning_rate": 8.378891241276563e-05, "loss": 2.754, "step": 20674 }, { "epoch": 1.62, "learning_rate": 8.378812828354113e-05, "loss": 2.9522, "step": 20675 }, { "epoch": 1.62, "learning_rate": 8.378734415431663e-05, "loss": 2.7187, "step": 20676 }, { "epoch": 1.62, "learning_rate": 8.378656002509214e-05, "loss": 2.9212, "step": 20677 }, { "epoch": 1.62, "learning_rate": 8.378577589586764e-05, "loss": 2.8116, "step": 20678 }, { "epoch": 1.62, "learning_rate": 8.378499176664314e-05, "loss": 2.9525, "step": 20679 }, { "epoch": 1.62, "learning_rate": 8.378420763741865e-05, "loss": 2.845, "step": 20680 }, { "epoch": 1.62, "learning_rate": 8.378342350819417e-05, "loss": 2.7853, "step": 20681 }, { "epoch": 1.62, "learning_rate": 8.378263937896967e-05, "loss": 2.7828, "step": 20682 }, { "epoch": 1.62, "learning_rate": 8.378185524974517e-05, "loss": 2.964, "step": 20683 }, { "epoch": 1.62, "learning_rate": 8.378107112052068e-05, "loss": 3.1414, "step": 20684 }, { "epoch": 1.62, "learning_rate": 8.378028699129617e-05, "loss": 2.8935, "step": 20685 }, { "epoch": 1.62, "learning_rate": 8.377950286207167e-05, "loss": 2.88, "step": 20686 }, { "epoch": 1.62, "learning_rate": 8.377871873284717e-05, "loss": 2.8224, "step": 20687 }, { "epoch": 1.62, "learning_rate": 8.377793460362268e-05, "loss": 2.9552, "step": 20688 }, { "epoch": 1.62, "learning_rate": 8.377715047439818e-05, "loss": 2.8699, "step": 20689 }, { "epoch": 1.62, "learning_rate": 8.377636634517368e-05, "loss": 2.884, "step": 20690 }, { "epoch": 1.62, "learning_rate": 8.377558221594919e-05, "loss": 2.8777, "step": 20691 }, { "epoch": 1.62, "learning_rate": 8.377479808672469e-05, "loss": 2.8203, "step": 20692 }, { "epoch": 1.62, "learning_rate": 8.37740139575002e-05, "loss": 3.0538, "step": 20693 }, { "epoch": 1.62, "learning_rate": 8.37732298282757e-05, "loss": 3.0008, "step": 20694 }, { "epoch": 1.62, "learning_rate": 8.37724456990512e-05, "loss": 2.8828, "step": 20695 }, { "epoch": 1.62, "learning_rate": 8.377166156982671e-05, "loss": 3.0237, "step": 20696 }, { "epoch": 1.62, "learning_rate": 8.377087744060221e-05, "loss": 2.9653, "step": 20697 }, { "epoch": 1.62, "learning_rate": 8.377009331137771e-05, "loss": 2.7506, "step": 20698 }, { "epoch": 1.62, "learning_rate": 8.376930918215322e-05, "loss": 2.901, "step": 20699 }, { "epoch": 1.62, "learning_rate": 8.376852505292872e-05, "loss": 2.8594, "step": 20700 }, { "epoch": 1.62, "learning_rate": 8.376774092370423e-05, "loss": 2.9339, "step": 20701 }, { "epoch": 1.62, "learning_rate": 8.376695679447973e-05, "loss": 2.9571, "step": 20702 }, { "epoch": 1.62, "learning_rate": 8.376617266525523e-05, "loss": 2.9735, "step": 20703 }, { "epoch": 1.62, "learning_rate": 8.376538853603074e-05, "loss": 2.8709, "step": 20704 }, { "epoch": 1.62, "learning_rate": 8.376460440680624e-05, "loss": 2.8757, "step": 20705 }, { "epoch": 1.62, "learning_rate": 8.376382027758174e-05, "loss": 2.9249, "step": 20706 }, { "epoch": 1.62, "learning_rate": 8.376303614835726e-05, "loss": 3.2274, "step": 20707 }, { "epoch": 1.62, "learning_rate": 8.376225201913276e-05, "loss": 2.6663, "step": 20708 }, { "epoch": 1.62, "learning_rate": 8.376146788990827e-05, "loss": 2.9447, "step": 20709 }, { "epoch": 1.62, "learning_rate": 8.376068376068377e-05, "loss": 3.1197, "step": 20710 }, { "epoch": 1.62, "learning_rate": 8.375989963145928e-05, "loss": 2.8302, "step": 20711 }, { "epoch": 1.62, "learning_rate": 8.375911550223478e-05, "loss": 2.7621, "step": 20712 }, { "epoch": 1.62, "learning_rate": 8.375833137301028e-05, "loss": 2.8649, "step": 20713 }, { "epoch": 1.62, "learning_rate": 8.375754724378579e-05, "loss": 3.123, "step": 20714 }, { "epoch": 1.62, "learning_rate": 8.375676311456129e-05, "loss": 3.0031, "step": 20715 }, { "epoch": 1.62, "learning_rate": 8.37559789853368e-05, "loss": 2.788, "step": 20716 }, { "epoch": 1.62, "learning_rate": 8.37551948561123e-05, "loss": 2.5988, "step": 20717 }, { "epoch": 1.62, "learning_rate": 8.37544107268878e-05, "loss": 2.9833, "step": 20718 }, { "epoch": 1.62, "learning_rate": 8.37536265976633e-05, "loss": 3.0686, "step": 20719 }, { "epoch": 1.62, "learning_rate": 8.375284246843881e-05, "loss": 2.8832, "step": 20720 }, { "epoch": 1.62, "learning_rate": 8.37520583392143e-05, "loss": 2.9416, "step": 20721 }, { "epoch": 1.62, "learning_rate": 8.37512742099898e-05, "loss": 3.0768, "step": 20722 }, { "epoch": 1.62, "learning_rate": 8.37504900807653e-05, "loss": 2.9365, "step": 20723 }, { "epoch": 1.63, "learning_rate": 8.374970595154081e-05, "loss": 3.0121, "step": 20724 }, { "epoch": 1.63, "learning_rate": 8.374892182231631e-05, "loss": 2.89, "step": 20725 }, { "epoch": 1.63, "learning_rate": 8.374813769309182e-05, "loss": 2.7409, "step": 20726 }, { "epoch": 1.63, "learning_rate": 8.374735356386732e-05, "loss": 2.8872, "step": 20727 }, { "epoch": 1.63, "learning_rate": 8.374656943464282e-05, "loss": 2.9943, "step": 20728 }, { "epoch": 1.63, "learning_rate": 8.374578530541833e-05, "loss": 2.9246, "step": 20729 }, { "epoch": 1.63, "learning_rate": 8.374500117619383e-05, "loss": 3.0789, "step": 20730 }, { "epoch": 1.63, "learning_rate": 8.374421704696934e-05, "loss": 3.0171, "step": 20731 }, { "epoch": 1.63, "learning_rate": 8.374343291774485e-05, "loss": 2.7672, "step": 20732 }, { "epoch": 1.63, "learning_rate": 8.374264878852036e-05, "loss": 3.0166, "step": 20733 }, { "epoch": 1.63, "learning_rate": 8.374186465929586e-05, "loss": 2.874, "step": 20734 }, { "epoch": 1.63, "learning_rate": 8.374108053007136e-05, "loss": 3.1922, "step": 20735 }, { "epoch": 1.63, "learning_rate": 8.374029640084687e-05, "loss": 2.9106, "step": 20736 }, { "epoch": 1.63, "learning_rate": 8.373951227162237e-05, "loss": 2.8242, "step": 20737 }, { "epoch": 1.63, "learning_rate": 8.373872814239787e-05, "loss": 2.7468, "step": 20738 }, { "epoch": 1.63, "learning_rate": 8.373794401317338e-05, "loss": 2.9033, "step": 20739 }, { "epoch": 1.63, "learning_rate": 8.373715988394888e-05, "loss": 2.862, "step": 20740 }, { "epoch": 1.63, "learning_rate": 8.373637575472439e-05, "loss": 2.8595, "step": 20741 }, { "epoch": 1.63, "learning_rate": 8.373559162549989e-05, "loss": 2.7977, "step": 20742 }, { "epoch": 1.63, "learning_rate": 8.37348074962754e-05, "loss": 2.8843, "step": 20743 }, { "epoch": 1.63, "learning_rate": 8.37340233670509e-05, "loss": 2.9174, "step": 20744 }, { "epoch": 1.63, "learning_rate": 8.37332392378264e-05, "loss": 2.9691, "step": 20745 }, { "epoch": 1.63, "learning_rate": 8.37324551086019e-05, "loss": 3.0239, "step": 20746 }, { "epoch": 1.63, "learning_rate": 8.373167097937741e-05, "loss": 3.01, "step": 20747 }, { "epoch": 1.63, "learning_rate": 8.373088685015291e-05, "loss": 2.7031, "step": 20748 }, { "epoch": 1.63, "learning_rate": 8.373010272092842e-05, "loss": 3.0632, "step": 20749 }, { "epoch": 1.63, "learning_rate": 8.372931859170392e-05, "loss": 3.033, "step": 20750 }, { "epoch": 1.63, "learning_rate": 8.372853446247942e-05, "loss": 2.9837, "step": 20751 }, { "epoch": 1.63, "learning_rate": 8.372775033325493e-05, "loss": 2.8863, "step": 20752 }, { "epoch": 1.63, "learning_rate": 8.372696620403043e-05, "loss": 3.0078, "step": 20753 }, { "epoch": 1.63, "learning_rate": 8.372618207480593e-05, "loss": 2.6886, "step": 20754 }, { "epoch": 1.63, "learning_rate": 8.372539794558144e-05, "loss": 2.7389, "step": 20755 }, { "epoch": 1.63, "learning_rate": 8.372461381635694e-05, "loss": 2.9908, "step": 20756 }, { "epoch": 1.63, "learning_rate": 8.372382968713243e-05, "loss": 2.8051, "step": 20757 }, { "epoch": 1.63, "learning_rate": 8.372304555790795e-05, "loss": 2.9382, "step": 20758 }, { "epoch": 1.63, "learning_rate": 8.372226142868345e-05, "loss": 2.6153, "step": 20759 }, { "epoch": 1.63, "learning_rate": 8.372147729945896e-05, "loss": 2.8238, "step": 20760 }, { "epoch": 1.63, "learning_rate": 8.372069317023446e-05, "loss": 2.59, "step": 20761 }, { "epoch": 1.63, "learning_rate": 8.371990904100996e-05, "loss": 2.8418, "step": 20762 }, { "epoch": 1.63, "learning_rate": 8.371912491178547e-05, "loss": 2.8504, "step": 20763 }, { "epoch": 1.63, "learning_rate": 8.371834078256097e-05, "loss": 2.7099, "step": 20764 }, { "epoch": 1.63, "learning_rate": 8.371755665333647e-05, "loss": 2.6768, "step": 20765 }, { "epoch": 1.63, "learning_rate": 8.371677252411198e-05, "loss": 2.8521, "step": 20766 }, { "epoch": 1.63, "learning_rate": 8.371598839488748e-05, "loss": 3.0069, "step": 20767 }, { "epoch": 1.63, "learning_rate": 8.371520426566299e-05, "loss": 2.7727, "step": 20768 }, { "epoch": 1.63, "learning_rate": 8.371442013643849e-05, "loss": 2.9089, "step": 20769 }, { "epoch": 1.63, "learning_rate": 8.371363600721399e-05, "loss": 2.8462, "step": 20770 }, { "epoch": 1.63, "learning_rate": 8.37128518779895e-05, "loss": 3.1034, "step": 20771 }, { "epoch": 1.63, "learning_rate": 8.3712067748765e-05, "loss": 2.9585, "step": 20772 }, { "epoch": 1.63, "learning_rate": 8.37112836195405e-05, "loss": 2.9422, "step": 20773 }, { "epoch": 1.63, "learning_rate": 8.371049949031601e-05, "loss": 2.7822, "step": 20774 }, { "epoch": 1.63, "learning_rate": 8.370971536109151e-05, "loss": 2.7831, "step": 20775 }, { "epoch": 1.63, "learning_rate": 8.370893123186701e-05, "loss": 2.8017, "step": 20776 }, { "epoch": 1.63, "learning_rate": 8.370814710264252e-05, "loss": 2.7754, "step": 20777 }, { "epoch": 1.63, "learning_rate": 8.370736297341802e-05, "loss": 2.8701, "step": 20778 }, { "epoch": 1.63, "learning_rate": 8.370657884419353e-05, "loss": 3.1743, "step": 20779 }, { "epoch": 1.63, "learning_rate": 8.370579471496903e-05, "loss": 3.0723, "step": 20780 }, { "epoch": 1.63, "learning_rate": 8.370501058574453e-05, "loss": 2.9384, "step": 20781 }, { "epoch": 1.63, "learning_rate": 8.370422645652004e-05, "loss": 2.8405, "step": 20782 }, { "epoch": 1.63, "learning_rate": 8.370344232729554e-05, "loss": 2.8237, "step": 20783 }, { "epoch": 1.63, "learning_rate": 8.370265819807104e-05, "loss": 2.7806, "step": 20784 }, { "epoch": 1.63, "learning_rate": 8.370187406884655e-05, "loss": 3.0233, "step": 20785 }, { "epoch": 1.63, "learning_rate": 8.370108993962205e-05, "loss": 3.0636, "step": 20786 }, { "epoch": 1.63, "learning_rate": 8.370030581039756e-05, "loss": 2.8698, "step": 20787 }, { "epoch": 1.63, "learning_rate": 8.369952168117306e-05, "loss": 3.0103, "step": 20788 }, { "epoch": 1.63, "learning_rate": 8.369873755194858e-05, "loss": 2.9649, "step": 20789 }, { "epoch": 1.63, "learning_rate": 8.369795342272408e-05, "loss": 2.9143, "step": 20790 }, { "epoch": 1.63, "learning_rate": 8.369716929349958e-05, "loss": 2.8084, "step": 20791 }, { "epoch": 1.63, "learning_rate": 8.369638516427509e-05, "loss": 2.7567, "step": 20792 }, { "epoch": 1.63, "learning_rate": 8.369560103505058e-05, "loss": 2.8285, "step": 20793 }, { "epoch": 1.63, "learning_rate": 8.369481690582608e-05, "loss": 3.1343, "step": 20794 }, { "epoch": 1.63, "learning_rate": 8.369403277660158e-05, "loss": 2.9033, "step": 20795 }, { "epoch": 1.63, "learning_rate": 8.369324864737709e-05, "loss": 2.77, "step": 20796 }, { "epoch": 1.63, "learning_rate": 8.369246451815259e-05, "loss": 3.0321, "step": 20797 }, { "epoch": 1.63, "learning_rate": 8.36916803889281e-05, "loss": 2.8764, "step": 20798 }, { "epoch": 1.63, "learning_rate": 8.36908962597036e-05, "loss": 2.7542, "step": 20799 }, { "epoch": 1.63, "learning_rate": 8.36901121304791e-05, "loss": 3.0049, "step": 20800 }, { "epoch": 1.63, "learning_rate": 8.36893280012546e-05, "loss": 2.8075, "step": 20801 }, { "epoch": 1.63, "learning_rate": 8.368854387203011e-05, "loss": 2.8849, "step": 20802 }, { "epoch": 1.63, "learning_rate": 8.368775974280561e-05, "loss": 2.7634, "step": 20803 }, { "epoch": 1.63, "learning_rate": 8.368697561358112e-05, "loss": 2.7375, "step": 20804 }, { "epoch": 1.63, "learning_rate": 8.368619148435662e-05, "loss": 2.9016, "step": 20805 }, { "epoch": 1.63, "learning_rate": 8.368540735513213e-05, "loss": 2.9548, "step": 20806 }, { "epoch": 1.63, "learning_rate": 8.368462322590763e-05, "loss": 2.8168, "step": 20807 }, { "epoch": 1.63, "learning_rate": 8.368383909668313e-05, "loss": 3.1261, "step": 20808 }, { "epoch": 1.63, "learning_rate": 8.368305496745864e-05, "loss": 2.8019, "step": 20809 }, { "epoch": 1.63, "learning_rate": 8.368227083823414e-05, "loss": 3.0495, "step": 20810 }, { "epoch": 1.63, "learning_rate": 8.368148670900964e-05, "loss": 2.9242, "step": 20811 }, { "epoch": 1.63, "learning_rate": 8.368070257978515e-05, "loss": 2.8713, "step": 20812 }, { "epoch": 1.63, "learning_rate": 8.367991845056065e-05, "loss": 3.189, "step": 20813 }, { "epoch": 1.63, "learning_rate": 8.367913432133615e-05, "loss": 2.8846, "step": 20814 }, { "epoch": 1.63, "learning_rate": 8.367835019211167e-05, "loss": 2.9441, "step": 20815 }, { "epoch": 1.63, "learning_rate": 8.367756606288718e-05, "loss": 2.9096, "step": 20816 }, { "epoch": 1.63, "learning_rate": 8.367678193366268e-05, "loss": 2.9552, "step": 20817 }, { "epoch": 1.63, "learning_rate": 8.367599780443818e-05, "loss": 2.9159, "step": 20818 }, { "epoch": 1.63, "learning_rate": 8.367521367521369e-05, "loss": 2.8674, "step": 20819 }, { "epoch": 1.63, "learning_rate": 8.367442954598919e-05, "loss": 2.9087, "step": 20820 }, { "epoch": 1.63, "learning_rate": 8.36736454167647e-05, "loss": 3.018, "step": 20821 }, { "epoch": 1.63, "learning_rate": 8.36728612875402e-05, "loss": 2.8112, "step": 20822 }, { "epoch": 1.63, "learning_rate": 8.36720771583157e-05, "loss": 3.1081, "step": 20823 }, { "epoch": 1.63, "learning_rate": 8.36712930290912e-05, "loss": 2.6953, "step": 20824 }, { "epoch": 1.63, "learning_rate": 8.367050889986671e-05, "loss": 3.2087, "step": 20825 }, { "epoch": 1.63, "learning_rate": 8.366972477064221e-05, "loss": 2.9527, "step": 20826 }, { "epoch": 1.63, "learning_rate": 8.366894064141772e-05, "loss": 3.0125, "step": 20827 }, { "epoch": 1.63, "learning_rate": 8.366815651219322e-05, "loss": 2.8127, "step": 20828 }, { "epoch": 1.63, "learning_rate": 8.366737238296871e-05, "loss": 2.7128, "step": 20829 }, { "epoch": 1.63, "learning_rate": 8.366658825374421e-05, "loss": 2.8481, "step": 20830 }, { "epoch": 1.63, "learning_rate": 8.366580412451972e-05, "loss": 2.9569, "step": 20831 }, { "epoch": 1.63, "learning_rate": 8.366501999529522e-05, "loss": 2.7184, "step": 20832 }, { "epoch": 1.63, "learning_rate": 8.366423586607072e-05, "loss": 2.811, "step": 20833 }, { "epoch": 1.63, "learning_rate": 8.366345173684623e-05, "loss": 3.1244, "step": 20834 }, { "epoch": 1.63, "learning_rate": 8.366266760762173e-05, "loss": 2.9848, "step": 20835 }, { "epoch": 1.63, "learning_rate": 8.366188347839724e-05, "loss": 2.9307, "step": 20836 }, { "epoch": 1.63, "learning_rate": 8.366109934917274e-05, "loss": 2.6872, "step": 20837 }, { "epoch": 1.63, "learning_rate": 8.366031521994824e-05, "loss": 2.8438, "step": 20838 }, { "epoch": 1.63, "learning_rate": 8.365953109072375e-05, "loss": 2.8847, "step": 20839 }, { "epoch": 1.63, "learning_rate": 8.365874696149926e-05, "loss": 2.8667, "step": 20840 }, { "epoch": 1.63, "learning_rate": 8.365796283227477e-05, "loss": 2.9224, "step": 20841 }, { "epoch": 1.63, "learning_rate": 8.365717870305027e-05, "loss": 3.0568, "step": 20842 }, { "epoch": 1.63, "learning_rate": 8.365639457382577e-05, "loss": 2.8966, "step": 20843 }, { "epoch": 1.63, "learning_rate": 8.365561044460128e-05, "loss": 2.9077, "step": 20844 }, { "epoch": 1.63, "learning_rate": 8.365482631537678e-05, "loss": 2.9469, "step": 20845 }, { "epoch": 1.63, "learning_rate": 8.365404218615229e-05, "loss": 2.8654, "step": 20846 }, { "epoch": 1.63, "learning_rate": 8.365325805692779e-05, "loss": 2.8599, "step": 20847 }, { "epoch": 1.63, "learning_rate": 8.365247392770329e-05, "loss": 3.2785, "step": 20848 }, { "epoch": 1.63, "learning_rate": 8.36516897984788e-05, "loss": 3.2486, "step": 20849 }, { "epoch": 1.63, "learning_rate": 8.36509056692543e-05, "loss": 2.8284, "step": 20850 }, { "epoch": 1.63, "learning_rate": 8.36501215400298e-05, "loss": 2.8281, "step": 20851 }, { "epoch": 1.64, "learning_rate": 8.364933741080531e-05, "loss": 2.83, "step": 20852 }, { "epoch": 1.64, "learning_rate": 8.364855328158081e-05, "loss": 2.8451, "step": 20853 }, { "epoch": 1.64, "learning_rate": 8.364776915235632e-05, "loss": 2.7911, "step": 20854 }, { "epoch": 1.64, "learning_rate": 8.364698502313182e-05, "loss": 3.1032, "step": 20855 }, { "epoch": 1.64, "learning_rate": 8.364620089390732e-05, "loss": 2.7667, "step": 20856 }, { "epoch": 1.64, "learning_rate": 8.364541676468283e-05, "loss": 2.9384, "step": 20857 }, { "epoch": 1.64, "learning_rate": 8.364463263545833e-05, "loss": 2.8613, "step": 20858 }, { "epoch": 1.64, "learning_rate": 8.364384850623383e-05, "loss": 2.9629, "step": 20859 }, { "epoch": 1.64, "learning_rate": 8.364306437700934e-05, "loss": 2.832, "step": 20860 }, { "epoch": 1.64, "learning_rate": 8.364228024778484e-05, "loss": 2.7764, "step": 20861 }, { "epoch": 1.64, "learning_rate": 8.364149611856034e-05, "loss": 2.8508, "step": 20862 }, { "epoch": 1.64, "learning_rate": 8.364071198933585e-05, "loss": 3.0352, "step": 20863 }, { "epoch": 1.64, "learning_rate": 8.363992786011135e-05, "loss": 2.8428, "step": 20864 }, { "epoch": 1.64, "learning_rate": 8.363914373088684e-05, "loss": 2.7689, "step": 20865 }, { "epoch": 1.64, "learning_rate": 8.363835960166236e-05, "loss": 3.0608, "step": 20866 }, { "epoch": 1.64, "learning_rate": 8.363757547243786e-05, "loss": 3.0278, "step": 20867 }, { "epoch": 1.64, "learning_rate": 8.363679134321337e-05, "loss": 2.7292, "step": 20868 }, { "epoch": 1.64, "learning_rate": 8.363600721398887e-05, "loss": 2.9224, "step": 20869 }, { "epoch": 1.64, "learning_rate": 8.363522308476437e-05, "loss": 2.8878, "step": 20870 }, { "epoch": 1.64, "learning_rate": 8.363443895553988e-05, "loss": 2.7532, "step": 20871 }, { "epoch": 1.64, "learning_rate": 8.363365482631538e-05, "loss": 2.8172, "step": 20872 }, { "epoch": 1.64, "learning_rate": 8.363287069709089e-05, "loss": 2.6635, "step": 20873 }, { "epoch": 1.64, "learning_rate": 8.363208656786639e-05, "loss": 2.9284, "step": 20874 }, { "epoch": 1.64, "learning_rate": 8.363130243864189e-05, "loss": 2.9807, "step": 20875 }, { "epoch": 1.64, "learning_rate": 8.36305183094174e-05, "loss": 3.0337, "step": 20876 }, { "epoch": 1.64, "learning_rate": 8.36297341801929e-05, "loss": 2.86, "step": 20877 }, { "epoch": 1.64, "learning_rate": 8.36289500509684e-05, "loss": 2.9684, "step": 20878 }, { "epoch": 1.64, "learning_rate": 8.362816592174391e-05, "loss": 3.3389, "step": 20879 }, { "epoch": 1.64, "learning_rate": 8.362738179251941e-05, "loss": 2.7634, "step": 20880 }, { "epoch": 1.64, "learning_rate": 8.362659766329491e-05, "loss": 2.7829, "step": 20881 }, { "epoch": 1.64, "learning_rate": 8.362581353407042e-05, "loss": 2.8962, "step": 20882 }, { "epoch": 1.64, "learning_rate": 8.362502940484592e-05, "loss": 3.0092, "step": 20883 }, { "epoch": 1.64, "learning_rate": 8.362424527562143e-05, "loss": 2.9524, "step": 20884 }, { "epoch": 1.64, "learning_rate": 8.362346114639693e-05, "loss": 3.0352, "step": 20885 }, { "epoch": 1.64, "learning_rate": 8.362267701717243e-05, "loss": 2.9523, "step": 20886 }, { "epoch": 1.64, "learning_rate": 8.362189288794794e-05, "loss": 2.9583, "step": 20887 }, { "epoch": 1.64, "learning_rate": 8.362110875872344e-05, "loss": 2.7567, "step": 20888 }, { "epoch": 1.64, "learning_rate": 8.362032462949894e-05, "loss": 3.2227, "step": 20889 }, { "epoch": 1.64, "learning_rate": 8.361954050027445e-05, "loss": 2.7985, "step": 20890 }, { "epoch": 1.64, "learning_rate": 8.361875637104995e-05, "loss": 3.0079, "step": 20891 }, { "epoch": 1.64, "learning_rate": 8.361797224182545e-05, "loss": 2.6593, "step": 20892 }, { "epoch": 1.64, "learning_rate": 8.361718811260096e-05, "loss": 2.9003, "step": 20893 }, { "epoch": 1.64, "learning_rate": 8.361640398337646e-05, "loss": 2.6201, "step": 20894 }, { "epoch": 1.64, "learning_rate": 8.361561985415197e-05, "loss": 2.8628, "step": 20895 }, { "epoch": 1.64, "learning_rate": 8.361483572492747e-05, "loss": 2.978, "step": 20896 }, { "epoch": 1.64, "learning_rate": 8.361405159570299e-05, "loss": 2.8528, "step": 20897 }, { "epoch": 1.64, "learning_rate": 8.361326746647849e-05, "loss": 2.9295, "step": 20898 }, { "epoch": 1.64, "learning_rate": 8.3612483337254e-05, "loss": 2.7709, "step": 20899 }, { "epoch": 1.64, "learning_rate": 8.36116992080295e-05, "loss": 2.9857, "step": 20900 }, { "epoch": 1.64, "learning_rate": 8.361091507880499e-05, "loss": 2.9159, "step": 20901 }, { "epoch": 1.64, "learning_rate": 8.361013094958049e-05, "loss": 2.8252, "step": 20902 }, { "epoch": 1.64, "learning_rate": 8.3609346820356e-05, "loss": 3.0276, "step": 20903 }, { "epoch": 1.64, "learning_rate": 8.36085626911315e-05, "loss": 3.0004, "step": 20904 }, { "epoch": 1.64, "learning_rate": 8.3607778561907e-05, "loss": 2.9759, "step": 20905 }, { "epoch": 1.64, "learning_rate": 8.36069944326825e-05, "loss": 2.7274, "step": 20906 }, { "epoch": 1.64, "learning_rate": 8.360621030345801e-05, "loss": 2.9274, "step": 20907 }, { "epoch": 1.64, "learning_rate": 8.360542617423351e-05, "loss": 3.1255, "step": 20908 }, { "epoch": 1.64, "learning_rate": 8.360464204500902e-05, "loss": 2.8356, "step": 20909 }, { "epoch": 1.64, "learning_rate": 8.360385791578452e-05, "loss": 2.9507, "step": 20910 }, { "epoch": 1.64, "learning_rate": 8.360307378656002e-05, "loss": 2.9114, "step": 20911 }, { "epoch": 1.64, "learning_rate": 8.360228965733553e-05, "loss": 3.2031, "step": 20912 }, { "epoch": 1.64, "learning_rate": 8.360150552811103e-05, "loss": 2.7436, "step": 20913 }, { "epoch": 1.64, "learning_rate": 8.360072139888654e-05, "loss": 2.9725, "step": 20914 }, { "epoch": 1.64, "learning_rate": 8.359993726966204e-05, "loss": 2.8078, "step": 20915 }, { "epoch": 1.64, "learning_rate": 8.359915314043754e-05, "loss": 2.8626, "step": 20916 }, { "epoch": 1.64, "learning_rate": 8.359836901121305e-05, "loss": 2.7629, "step": 20917 }, { "epoch": 1.64, "learning_rate": 8.359758488198855e-05, "loss": 2.9354, "step": 20918 }, { "epoch": 1.64, "learning_rate": 8.359680075276405e-05, "loss": 3.0245, "step": 20919 }, { "epoch": 1.64, "learning_rate": 8.359601662353956e-05, "loss": 2.8347, "step": 20920 }, { "epoch": 1.64, "learning_rate": 8.359523249431506e-05, "loss": 2.8676, "step": 20921 }, { "epoch": 1.64, "learning_rate": 8.359444836509057e-05, "loss": 2.8017, "step": 20922 }, { "epoch": 1.64, "learning_rate": 8.359366423586608e-05, "loss": 2.9227, "step": 20923 }, { "epoch": 1.64, "learning_rate": 8.359288010664159e-05, "loss": 2.9184, "step": 20924 }, { "epoch": 1.64, "learning_rate": 8.359209597741709e-05, "loss": 2.8076, "step": 20925 }, { "epoch": 1.64, "learning_rate": 8.35913118481926e-05, "loss": 2.8727, "step": 20926 }, { "epoch": 1.64, "learning_rate": 8.35905277189681e-05, "loss": 2.8296, "step": 20927 }, { "epoch": 1.64, "learning_rate": 8.35897435897436e-05, "loss": 2.9346, "step": 20928 }, { "epoch": 1.64, "learning_rate": 8.35889594605191e-05, "loss": 2.9803, "step": 20929 }, { "epoch": 1.64, "learning_rate": 8.358817533129461e-05, "loss": 2.8397, "step": 20930 }, { "epoch": 1.64, "learning_rate": 8.358739120207011e-05, "loss": 2.8618, "step": 20931 }, { "epoch": 1.64, "learning_rate": 8.358660707284562e-05, "loss": 2.8773, "step": 20932 }, { "epoch": 1.64, "learning_rate": 8.358582294362112e-05, "loss": 2.9338, "step": 20933 }, { "epoch": 1.64, "learning_rate": 8.358503881439662e-05, "loss": 3.1153, "step": 20934 }, { "epoch": 1.64, "learning_rate": 8.358425468517213e-05, "loss": 2.7987, "step": 20935 }, { "epoch": 1.64, "learning_rate": 8.358347055594763e-05, "loss": 2.9299, "step": 20936 }, { "epoch": 1.64, "learning_rate": 8.358268642672312e-05, "loss": 2.874, "step": 20937 }, { "epoch": 1.64, "learning_rate": 8.358190229749862e-05, "loss": 3.1063, "step": 20938 }, { "epoch": 1.64, "learning_rate": 8.358111816827413e-05, "loss": 3.1887, "step": 20939 }, { "epoch": 1.64, "learning_rate": 8.358033403904963e-05, "loss": 3.0823, "step": 20940 }, { "epoch": 1.64, "learning_rate": 8.357954990982514e-05, "loss": 2.7981, "step": 20941 }, { "epoch": 1.64, "learning_rate": 8.357876578060064e-05, "loss": 2.7354, "step": 20942 }, { "epoch": 1.64, "learning_rate": 8.357798165137614e-05, "loss": 3.0688, "step": 20943 }, { "epoch": 1.64, "learning_rate": 8.357719752215165e-05, "loss": 2.7755, "step": 20944 }, { "epoch": 1.64, "learning_rate": 8.357641339292715e-05, "loss": 2.7517, "step": 20945 }, { "epoch": 1.64, "learning_rate": 8.357562926370265e-05, "loss": 2.9123, "step": 20946 }, { "epoch": 1.64, "learning_rate": 8.357484513447816e-05, "loss": 2.9792, "step": 20947 }, { "epoch": 1.64, "learning_rate": 8.357406100525367e-05, "loss": 2.912, "step": 20948 }, { "epoch": 1.64, "learning_rate": 8.357327687602918e-05, "loss": 2.93, "step": 20949 }, { "epoch": 1.64, "learning_rate": 8.357249274680468e-05, "loss": 2.8132, "step": 20950 }, { "epoch": 1.64, "learning_rate": 8.357170861758019e-05, "loss": 2.9911, "step": 20951 }, { "epoch": 1.64, "learning_rate": 8.357092448835569e-05, "loss": 2.9039, "step": 20952 }, { "epoch": 1.64, "learning_rate": 8.357014035913119e-05, "loss": 2.9618, "step": 20953 }, { "epoch": 1.64, "learning_rate": 8.35693562299067e-05, "loss": 3.0231, "step": 20954 }, { "epoch": 1.64, "learning_rate": 8.35685721006822e-05, "loss": 3.2137, "step": 20955 }, { "epoch": 1.64, "learning_rate": 8.35677879714577e-05, "loss": 2.8785, "step": 20956 }, { "epoch": 1.64, "learning_rate": 8.356700384223321e-05, "loss": 2.8815, "step": 20957 }, { "epoch": 1.64, "learning_rate": 8.356621971300871e-05, "loss": 2.8738, "step": 20958 }, { "epoch": 1.64, "learning_rate": 8.356543558378421e-05, "loss": 2.7166, "step": 20959 }, { "epoch": 1.64, "learning_rate": 8.356465145455972e-05, "loss": 3.1303, "step": 20960 }, { "epoch": 1.64, "learning_rate": 8.356386732533522e-05, "loss": 3.1061, "step": 20961 }, { "epoch": 1.64, "learning_rate": 8.356308319611073e-05, "loss": 2.9855, "step": 20962 }, { "epoch": 1.64, "learning_rate": 8.356229906688623e-05, "loss": 3.0375, "step": 20963 }, { "epoch": 1.64, "learning_rate": 8.356151493766173e-05, "loss": 2.9872, "step": 20964 }, { "epoch": 1.64, "learning_rate": 8.356073080843724e-05, "loss": 2.8396, "step": 20965 }, { "epoch": 1.64, "learning_rate": 8.355994667921274e-05, "loss": 3.0309, "step": 20966 }, { "epoch": 1.64, "learning_rate": 8.355916254998824e-05, "loss": 3.0054, "step": 20967 }, { "epoch": 1.64, "learning_rate": 8.355837842076375e-05, "loss": 3.0505, "step": 20968 }, { "epoch": 1.64, "learning_rate": 8.355759429153925e-05, "loss": 2.77, "step": 20969 }, { "epoch": 1.64, "learning_rate": 8.355681016231476e-05, "loss": 2.9634, "step": 20970 }, { "epoch": 1.64, "learning_rate": 8.355602603309026e-05, "loss": 2.9686, "step": 20971 }, { "epoch": 1.64, "learning_rate": 8.355524190386576e-05, "loss": 2.8025, "step": 20972 }, { "epoch": 1.64, "learning_rate": 8.355445777464125e-05, "loss": 2.9401, "step": 20973 }, { "epoch": 1.64, "learning_rate": 8.355367364541677e-05, "loss": 3.1881, "step": 20974 }, { "epoch": 1.64, "learning_rate": 8.355288951619227e-05, "loss": 2.8527, "step": 20975 }, { "epoch": 1.64, "learning_rate": 8.355210538696778e-05, "loss": 2.8934, "step": 20976 }, { "epoch": 1.64, "learning_rate": 8.355132125774328e-05, "loss": 2.9745, "step": 20977 }, { "epoch": 1.64, "learning_rate": 8.355053712851878e-05, "loss": 2.8587, "step": 20978 }, { "epoch": 1.65, "learning_rate": 8.354975299929429e-05, "loss": 3.0858, "step": 20979 }, { "epoch": 1.65, "learning_rate": 8.354896887006979e-05, "loss": 2.7428, "step": 20980 }, { "epoch": 1.65, "learning_rate": 8.35481847408453e-05, "loss": 2.9634, "step": 20981 }, { "epoch": 1.65, "learning_rate": 8.35474006116208e-05, "loss": 2.8286, "step": 20982 }, { "epoch": 1.65, "learning_rate": 8.35466164823963e-05, "loss": 3.0222, "step": 20983 }, { "epoch": 1.65, "learning_rate": 8.354583235317181e-05, "loss": 2.9847, "step": 20984 }, { "epoch": 1.65, "learning_rate": 8.354504822394731e-05, "loss": 2.7511, "step": 20985 }, { "epoch": 1.65, "learning_rate": 8.354426409472281e-05, "loss": 2.8134, "step": 20986 }, { "epoch": 1.65, "learning_rate": 8.354347996549832e-05, "loss": 2.6426, "step": 20987 }, { "epoch": 1.65, "learning_rate": 8.354269583627382e-05, "loss": 2.9271, "step": 20988 }, { "epoch": 1.65, "learning_rate": 8.354191170704933e-05, "loss": 2.8308, "step": 20989 }, { "epoch": 1.65, "learning_rate": 8.354112757782483e-05, "loss": 2.895, "step": 20990 }, { "epoch": 1.65, "learning_rate": 8.354034344860033e-05, "loss": 2.8943, "step": 20991 }, { "epoch": 1.65, "learning_rate": 8.353955931937584e-05, "loss": 3.0734, "step": 20992 }, { "epoch": 1.65, "learning_rate": 8.353877519015134e-05, "loss": 2.7202, "step": 20993 }, { "epoch": 1.65, "learning_rate": 8.353799106092684e-05, "loss": 2.8286, "step": 20994 }, { "epoch": 1.65, "learning_rate": 8.353720693170235e-05, "loss": 2.9787, "step": 20995 }, { "epoch": 1.65, "learning_rate": 8.353642280247785e-05, "loss": 2.9751, "step": 20996 }, { "epoch": 1.65, "learning_rate": 8.353563867325335e-05, "loss": 2.9067, "step": 20997 }, { "epoch": 1.65, "learning_rate": 8.353485454402886e-05, "loss": 2.9289, "step": 20998 }, { "epoch": 1.65, "learning_rate": 8.353407041480436e-05, "loss": 2.8623, "step": 20999 }, { "epoch": 1.65, "learning_rate": 8.353328628557987e-05, "loss": 2.8228, "step": 21000 }, { "epoch": 1.65, "eval_loss": 2.907252311706543, "eval_runtime": 217.9805, "eval_samples_per_second": 51.986, "eval_steps_per_second": 3.253, "step": 21000 }, { "epoch": 1.65, "learning_rate": 8.353250215635537e-05, "loss": 2.7822, "step": 21001 }, { "epoch": 1.65, "learning_rate": 8.353171802713087e-05, "loss": 3.0542, "step": 21002 }, { "epoch": 1.65, "learning_rate": 8.353093389790638e-05, "loss": 2.9337, "step": 21003 }, { "epoch": 1.65, "learning_rate": 8.353014976868188e-05, "loss": 2.9294, "step": 21004 }, { "epoch": 1.65, "learning_rate": 8.35293656394574e-05, "loss": 2.8613, "step": 21005 }, { "epoch": 1.65, "learning_rate": 8.35285815102329e-05, "loss": 2.6883, "step": 21006 }, { "epoch": 1.65, "learning_rate": 8.35277973810084e-05, "loss": 3.0377, "step": 21007 }, { "epoch": 1.65, "learning_rate": 8.352701325178391e-05, "loss": 2.8601, "step": 21008 }, { "epoch": 1.65, "learning_rate": 8.35262291225594e-05, "loss": 2.8183, "step": 21009 }, { "epoch": 1.65, "learning_rate": 8.35254449933349e-05, "loss": 3.1538, "step": 21010 }, { "epoch": 1.65, "learning_rate": 8.35246608641104e-05, "loss": 3.2144, "step": 21011 }, { "epoch": 1.65, "learning_rate": 8.352387673488591e-05, "loss": 2.8812, "step": 21012 }, { "epoch": 1.65, "learning_rate": 8.352309260566141e-05, "loss": 3.1721, "step": 21013 }, { "epoch": 1.65, "learning_rate": 8.352230847643692e-05, "loss": 2.7524, "step": 21014 }, { "epoch": 1.65, "learning_rate": 8.352152434721242e-05, "loss": 2.9808, "step": 21015 }, { "epoch": 1.65, "learning_rate": 8.352074021798792e-05, "loss": 2.8092, "step": 21016 }, { "epoch": 1.65, "learning_rate": 8.351995608876343e-05, "loss": 2.7709, "step": 21017 }, { "epoch": 1.65, "learning_rate": 8.351917195953893e-05, "loss": 3.05, "step": 21018 }, { "epoch": 1.65, "learning_rate": 8.351838783031444e-05, "loss": 2.8248, "step": 21019 }, { "epoch": 1.65, "learning_rate": 8.351760370108994e-05, "loss": 2.7468, "step": 21020 }, { "epoch": 1.65, "learning_rate": 8.351681957186544e-05, "loss": 2.6612, "step": 21021 }, { "epoch": 1.65, "learning_rate": 8.351603544264095e-05, "loss": 2.7264, "step": 21022 }, { "epoch": 1.65, "learning_rate": 8.351525131341645e-05, "loss": 2.8818, "step": 21023 }, { "epoch": 1.65, "learning_rate": 8.351446718419195e-05, "loss": 2.919, "step": 21024 }, { "epoch": 1.65, "learning_rate": 8.351368305496746e-05, "loss": 2.7659, "step": 21025 }, { "epoch": 1.65, "learning_rate": 8.351289892574296e-05, "loss": 2.9078, "step": 21026 }, { "epoch": 1.65, "learning_rate": 8.351211479651846e-05, "loss": 3.0268, "step": 21027 }, { "epoch": 1.65, "learning_rate": 8.351133066729397e-05, "loss": 2.9547, "step": 21028 }, { "epoch": 1.65, "learning_rate": 8.351054653806947e-05, "loss": 2.8244, "step": 21029 }, { "epoch": 1.65, "learning_rate": 8.350976240884498e-05, "loss": 2.7974, "step": 21030 }, { "epoch": 1.65, "learning_rate": 8.350897827962049e-05, "loss": 2.9316, "step": 21031 }, { "epoch": 1.65, "learning_rate": 8.3508194150396e-05, "loss": 2.9065, "step": 21032 }, { "epoch": 1.65, "learning_rate": 8.35074100211715e-05, "loss": 2.7978, "step": 21033 }, { "epoch": 1.65, "learning_rate": 8.3506625891947e-05, "loss": 2.6332, "step": 21034 }, { "epoch": 1.65, "learning_rate": 8.350584176272251e-05, "loss": 3.2894, "step": 21035 }, { "epoch": 1.65, "learning_rate": 8.350505763349801e-05, "loss": 2.6199, "step": 21036 }, { "epoch": 1.65, "learning_rate": 8.350427350427352e-05, "loss": 2.8979, "step": 21037 }, { "epoch": 1.65, "learning_rate": 8.350348937504902e-05, "loss": 3.0409, "step": 21038 }, { "epoch": 1.65, "learning_rate": 8.350270524582452e-05, "loss": 2.9798, "step": 21039 }, { "epoch": 1.65, "learning_rate": 8.350192111660003e-05, "loss": 2.8633, "step": 21040 }, { "epoch": 1.65, "learning_rate": 8.350113698737553e-05, "loss": 2.9041, "step": 21041 }, { "epoch": 1.65, "learning_rate": 8.350035285815103e-05, "loss": 2.8259, "step": 21042 }, { "epoch": 1.65, "learning_rate": 8.349956872892654e-05, "loss": 2.8568, "step": 21043 }, { "epoch": 1.65, "learning_rate": 8.349878459970204e-05, "loss": 2.9674, "step": 21044 }, { "epoch": 1.65, "learning_rate": 8.349800047047753e-05, "loss": 2.9806, "step": 21045 }, { "epoch": 1.65, "learning_rate": 8.349721634125303e-05, "loss": 3.0846, "step": 21046 }, { "epoch": 1.65, "learning_rate": 8.349643221202854e-05, "loss": 2.7871, "step": 21047 }, { "epoch": 1.65, "learning_rate": 8.349564808280404e-05, "loss": 2.9322, "step": 21048 }, { "epoch": 1.65, "learning_rate": 8.349486395357955e-05, "loss": 2.8469, "step": 21049 }, { "epoch": 1.65, "learning_rate": 8.349407982435505e-05, "loss": 3.0637, "step": 21050 }, { "epoch": 1.65, "learning_rate": 8.349329569513055e-05, "loss": 2.6445, "step": 21051 }, { "epoch": 1.65, "learning_rate": 8.349251156590606e-05, "loss": 2.8481, "step": 21052 }, { "epoch": 1.65, "learning_rate": 8.349172743668156e-05, "loss": 2.7722, "step": 21053 }, { "epoch": 1.65, "learning_rate": 8.349094330745706e-05, "loss": 2.9737, "step": 21054 }, { "epoch": 1.65, "learning_rate": 8.349015917823257e-05, "loss": 2.8948, "step": 21055 }, { "epoch": 1.65, "learning_rate": 8.348937504900809e-05, "loss": 2.8344, "step": 21056 }, { "epoch": 1.65, "learning_rate": 8.348859091978359e-05, "loss": 2.7973, "step": 21057 }, { "epoch": 1.65, "learning_rate": 8.348780679055909e-05, "loss": 2.8081, "step": 21058 }, { "epoch": 1.65, "learning_rate": 8.34870226613346e-05, "loss": 3.1013, "step": 21059 }, { "epoch": 1.65, "learning_rate": 8.34862385321101e-05, "loss": 3.0809, "step": 21060 }, { "epoch": 1.65, "learning_rate": 8.34854544028856e-05, "loss": 2.9179, "step": 21061 }, { "epoch": 1.65, "learning_rate": 8.348467027366111e-05, "loss": 2.7336, "step": 21062 }, { "epoch": 1.65, "learning_rate": 8.348388614443661e-05, "loss": 2.8598, "step": 21063 }, { "epoch": 1.65, "learning_rate": 8.348310201521211e-05, "loss": 2.9948, "step": 21064 }, { "epoch": 1.65, "learning_rate": 8.348231788598762e-05, "loss": 2.9475, "step": 21065 }, { "epoch": 1.65, "learning_rate": 8.348153375676312e-05, "loss": 2.9915, "step": 21066 }, { "epoch": 1.65, "learning_rate": 8.348074962753863e-05, "loss": 2.8137, "step": 21067 }, { "epoch": 1.65, "learning_rate": 8.347996549831413e-05, "loss": 3.1209, "step": 21068 }, { "epoch": 1.65, "learning_rate": 8.347918136908963e-05, "loss": 2.7978, "step": 21069 }, { "epoch": 1.65, "learning_rate": 8.347839723986514e-05, "loss": 2.9336, "step": 21070 }, { "epoch": 1.65, "learning_rate": 8.347761311064064e-05, "loss": 2.7875, "step": 21071 }, { "epoch": 1.65, "learning_rate": 8.347682898141614e-05, "loss": 2.8645, "step": 21072 }, { "epoch": 1.65, "learning_rate": 8.347604485219165e-05, "loss": 2.7483, "step": 21073 }, { "epoch": 1.65, "learning_rate": 8.347526072296715e-05, "loss": 2.948, "step": 21074 }, { "epoch": 1.65, "learning_rate": 8.347447659374266e-05, "loss": 2.706, "step": 21075 }, { "epoch": 1.65, "learning_rate": 8.347369246451816e-05, "loss": 2.877, "step": 21076 }, { "epoch": 1.65, "learning_rate": 8.347290833529366e-05, "loss": 2.7348, "step": 21077 }, { "epoch": 1.65, "learning_rate": 8.347212420606917e-05, "loss": 2.8116, "step": 21078 }, { "epoch": 1.65, "learning_rate": 8.347134007684467e-05, "loss": 2.9028, "step": 21079 }, { "epoch": 1.65, "learning_rate": 8.347055594762017e-05, "loss": 3.054, "step": 21080 }, { "epoch": 1.65, "learning_rate": 8.346977181839566e-05, "loss": 2.9402, "step": 21081 }, { "epoch": 1.65, "learning_rate": 8.346898768917118e-05, "loss": 2.8986, "step": 21082 }, { "epoch": 1.65, "learning_rate": 8.346820355994668e-05, "loss": 2.9209, "step": 21083 }, { "epoch": 1.65, "learning_rate": 8.346741943072219e-05, "loss": 2.854, "step": 21084 }, { "epoch": 1.65, "learning_rate": 8.346663530149769e-05, "loss": 2.9948, "step": 21085 }, { "epoch": 1.65, "learning_rate": 8.34658511722732e-05, "loss": 3.0838, "step": 21086 }, { "epoch": 1.65, "learning_rate": 8.34650670430487e-05, "loss": 2.841, "step": 21087 }, { "epoch": 1.65, "learning_rate": 8.34642829138242e-05, "loss": 2.7309, "step": 21088 }, { "epoch": 1.65, "learning_rate": 8.34634987845997e-05, "loss": 2.8205, "step": 21089 }, { "epoch": 1.65, "learning_rate": 8.346271465537521e-05, "loss": 2.7569, "step": 21090 }, { "epoch": 1.65, "learning_rate": 8.346193052615071e-05, "loss": 2.864, "step": 21091 }, { "epoch": 1.65, "learning_rate": 8.346114639692622e-05, "loss": 2.8594, "step": 21092 }, { "epoch": 1.65, "learning_rate": 8.346036226770172e-05, "loss": 2.8322, "step": 21093 }, { "epoch": 1.65, "learning_rate": 8.345957813847722e-05, "loss": 2.9892, "step": 21094 }, { "epoch": 1.65, "learning_rate": 8.345879400925273e-05, "loss": 2.6073, "step": 21095 }, { "epoch": 1.65, "learning_rate": 8.345800988002823e-05, "loss": 3.1629, "step": 21096 }, { "epoch": 1.65, "learning_rate": 8.345722575080374e-05, "loss": 2.9071, "step": 21097 }, { "epoch": 1.65, "learning_rate": 8.345644162157924e-05, "loss": 2.7406, "step": 21098 }, { "epoch": 1.65, "learning_rate": 8.345565749235474e-05, "loss": 2.7834, "step": 21099 }, { "epoch": 1.65, "learning_rate": 8.345487336313025e-05, "loss": 2.757, "step": 21100 }, { "epoch": 1.65, "learning_rate": 8.345408923390575e-05, "loss": 2.9917, "step": 21101 }, { "epoch": 1.65, "learning_rate": 8.345330510468125e-05, "loss": 3.0401, "step": 21102 }, { "epoch": 1.65, "learning_rate": 8.345252097545676e-05, "loss": 2.8332, "step": 21103 }, { "epoch": 1.65, "learning_rate": 8.345173684623226e-05, "loss": 3.0248, "step": 21104 }, { "epoch": 1.65, "learning_rate": 8.345095271700777e-05, "loss": 2.7564, "step": 21105 }, { "epoch": 1.65, "learning_rate": 8.345016858778327e-05, "loss": 2.9273, "step": 21106 }, { "epoch": 1.66, "learning_rate": 8.344938445855877e-05, "loss": 2.7267, "step": 21107 }, { "epoch": 1.66, "learning_rate": 8.344860032933428e-05, "loss": 2.8244, "step": 21108 }, { "epoch": 1.66, "learning_rate": 8.344781620010978e-05, "loss": 2.5865, "step": 21109 }, { "epoch": 1.66, "learning_rate": 8.344703207088528e-05, "loss": 3.0227, "step": 21110 }, { "epoch": 1.66, "learning_rate": 8.344624794166079e-05, "loss": 2.9256, "step": 21111 }, { "epoch": 1.66, "learning_rate": 8.344546381243629e-05, "loss": 2.8327, "step": 21112 }, { "epoch": 1.66, "learning_rate": 8.344467968321181e-05, "loss": 2.7013, "step": 21113 }, { "epoch": 1.66, "learning_rate": 8.344389555398731e-05, "loss": 2.7326, "step": 21114 }, { "epoch": 1.66, "learning_rate": 8.344311142476282e-05, "loss": 2.8504, "step": 21115 }, { "epoch": 1.66, "learning_rate": 8.344232729553832e-05, "loss": 2.9156, "step": 21116 }, { "epoch": 1.66, "learning_rate": 8.344154316631381e-05, "loss": 3.0554, "step": 21117 }, { "epoch": 1.66, "learning_rate": 8.344075903708931e-05, "loss": 3.0275, "step": 21118 }, { "epoch": 1.66, "learning_rate": 8.343997490786482e-05, "loss": 3.0262, "step": 21119 }, { "epoch": 1.66, "learning_rate": 8.343919077864032e-05, "loss": 3.0673, "step": 21120 }, { "epoch": 1.66, "learning_rate": 8.343840664941582e-05, "loss": 3.0255, "step": 21121 }, { "epoch": 1.66, "learning_rate": 8.343762252019133e-05, "loss": 2.9406, "step": 21122 }, { "epoch": 1.66, "learning_rate": 8.343683839096683e-05, "loss": 2.9479, "step": 21123 }, { "epoch": 1.66, "learning_rate": 8.343605426174234e-05, "loss": 2.7926, "step": 21124 }, { "epoch": 1.66, "learning_rate": 8.343527013251784e-05, "loss": 3.0454, "step": 21125 }, { "epoch": 1.66, "learning_rate": 8.343448600329334e-05, "loss": 2.8222, "step": 21126 }, { "epoch": 1.66, "learning_rate": 8.343370187406885e-05, "loss": 3.0604, "step": 21127 }, { "epoch": 1.66, "learning_rate": 8.343291774484435e-05, "loss": 2.9709, "step": 21128 }, { "epoch": 1.66, "learning_rate": 8.343213361561985e-05, "loss": 3.0845, "step": 21129 }, { "epoch": 1.66, "learning_rate": 8.343134948639536e-05, "loss": 3.1735, "step": 21130 }, { "epoch": 1.66, "learning_rate": 8.343056535717086e-05, "loss": 2.916, "step": 21131 }, { "epoch": 1.66, "learning_rate": 8.342978122794636e-05, "loss": 2.8464, "step": 21132 }, { "epoch": 1.66, "learning_rate": 8.342899709872187e-05, "loss": 2.7412, "step": 21133 }, { "epoch": 1.66, "learning_rate": 8.342821296949737e-05, "loss": 2.7127, "step": 21134 }, { "epoch": 1.66, "learning_rate": 8.342742884027288e-05, "loss": 2.9928, "step": 21135 }, { "epoch": 1.66, "learning_rate": 8.342664471104838e-05, "loss": 2.7837, "step": 21136 }, { "epoch": 1.66, "learning_rate": 8.342586058182388e-05, "loss": 2.8229, "step": 21137 }, { "epoch": 1.66, "learning_rate": 8.342507645259939e-05, "loss": 2.9899, "step": 21138 }, { "epoch": 1.66, "learning_rate": 8.34242923233749e-05, "loss": 2.9174, "step": 21139 }, { "epoch": 1.66, "learning_rate": 8.342350819415041e-05, "loss": 2.9653, "step": 21140 }, { "epoch": 1.66, "learning_rate": 8.342272406492591e-05, "loss": 2.966, "step": 21141 }, { "epoch": 1.66, "learning_rate": 8.342193993570141e-05, "loss": 2.7977, "step": 21142 }, { "epoch": 1.66, "learning_rate": 8.342115580647692e-05, "loss": 3.0939, "step": 21143 }, { "epoch": 1.66, "learning_rate": 8.342037167725242e-05, "loss": 2.9566, "step": 21144 }, { "epoch": 1.66, "learning_rate": 8.341958754802793e-05, "loss": 2.836, "step": 21145 }, { "epoch": 1.66, "learning_rate": 8.341880341880343e-05, "loss": 2.8449, "step": 21146 }, { "epoch": 1.66, "learning_rate": 8.341801928957893e-05, "loss": 3.0946, "step": 21147 }, { "epoch": 1.66, "learning_rate": 8.341723516035444e-05, "loss": 2.8153, "step": 21148 }, { "epoch": 1.66, "learning_rate": 8.341645103112994e-05, "loss": 3.0538, "step": 21149 }, { "epoch": 1.66, "learning_rate": 8.341566690190544e-05, "loss": 2.7851, "step": 21150 }, { "epoch": 1.66, "learning_rate": 8.341488277268095e-05, "loss": 3.2327, "step": 21151 }, { "epoch": 1.66, "learning_rate": 8.341409864345645e-05, "loss": 2.9976, "step": 21152 }, { "epoch": 1.66, "learning_rate": 8.341331451423194e-05, "loss": 2.8347, "step": 21153 }, { "epoch": 1.66, "learning_rate": 8.341253038500745e-05, "loss": 2.714, "step": 21154 }, { "epoch": 1.66, "learning_rate": 8.341174625578295e-05, "loss": 3.0485, "step": 21155 }, { "epoch": 1.66, "learning_rate": 8.341096212655845e-05, "loss": 2.9023, "step": 21156 }, { "epoch": 1.66, "learning_rate": 8.341017799733396e-05, "loss": 2.8577, "step": 21157 }, { "epoch": 1.66, "learning_rate": 8.340939386810946e-05, "loss": 2.8536, "step": 21158 }, { "epoch": 1.66, "learning_rate": 8.340860973888496e-05, "loss": 2.6911, "step": 21159 }, { "epoch": 1.66, "learning_rate": 8.340782560966047e-05, "loss": 2.7074, "step": 21160 }, { "epoch": 1.66, "learning_rate": 8.340704148043597e-05, "loss": 2.9985, "step": 21161 }, { "epoch": 1.66, "learning_rate": 8.340625735121147e-05, "loss": 3.0703, "step": 21162 }, { "epoch": 1.66, "learning_rate": 8.340547322198698e-05, "loss": 2.7525, "step": 21163 }, { "epoch": 1.66, "learning_rate": 8.340468909276248e-05, "loss": 3.0357, "step": 21164 }, { "epoch": 1.66, "learning_rate": 8.3403904963538e-05, "loss": 2.9015, "step": 21165 }, { "epoch": 1.66, "learning_rate": 8.34031208343135e-05, "loss": 2.8651, "step": 21166 }, { "epoch": 1.66, "learning_rate": 8.340233670508901e-05, "loss": 2.9773, "step": 21167 }, { "epoch": 1.66, "learning_rate": 8.340155257586451e-05, "loss": 3.1256, "step": 21168 }, { "epoch": 1.66, "learning_rate": 8.340076844664001e-05, "loss": 2.752, "step": 21169 }, { "epoch": 1.66, "learning_rate": 8.339998431741552e-05, "loss": 2.6873, "step": 21170 }, { "epoch": 1.66, "learning_rate": 8.339920018819102e-05, "loss": 2.8923, "step": 21171 }, { "epoch": 1.66, "learning_rate": 8.339841605896653e-05, "loss": 2.8805, "step": 21172 }, { "epoch": 1.66, "learning_rate": 8.339763192974203e-05, "loss": 2.6705, "step": 21173 }, { "epoch": 1.66, "learning_rate": 8.339684780051753e-05, "loss": 2.9881, "step": 21174 }, { "epoch": 1.66, "learning_rate": 8.339606367129304e-05, "loss": 2.8113, "step": 21175 }, { "epoch": 1.66, "learning_rate": 8.339527954206854e-05, "loss": 2.7652, "step": 21176 }, { "epoch": 1.66, "learning_rate": 8.339449541284404e-05, "loss": 3.0174, "step": 21177 }, { "epoch": 1.66, "learning_rate": 8.339371128361955e-05, "loss": 3.1283, "step": 21178 }, { "epoch": 1.66, "learning_rate": 8.339292715439505e-05, "loss": 2.7646, "step": 21179 }, { "epoch": 1.66, "learning_rate": 8.339214302517055e-05, "loss": 2.8978, "step": 21180 }, { "epoch": 1.66, "learning_rate": 8.339135889594606e-05, "loss": 2.7115, "step": 21181 }, { "epoch": 1.66, "learning_rate": 8.339057476672156e-05, "loss": 2.8926, "step": 21182 }, { "epoch": 1.66, "learning_rate": 8.338979063749707e-05, "loss": 2.9336, "step": 21183 }, { "epoch": 1.66, "learning_rate": 8.338900650827257e-05, "loss": 3.2645, "step": 21184 }, { "epoch": 1.66, "learning_rate": 8.338822237904807e-05, "loss": 2.8608, "step": 21185 }, { "epoch": 1.66, "learning_rate": 8.338743824982358e-05, "loss": 2.9346, "step": 21186 }, { "epoch": 1.66, "learning_rate": 8.338665412059908e-05, "loss": 2.8886, "step": 21187 }, { "epoch": 1.66, "learning_rate": 8.338586999137458e-05, "loss": 2.8891, "step": 21188 }, { "epoch": 1.66, "learning_rate": 8.338508586215007e-05, "loss": 3.074, "step": 21189 }, { "epoch": 1.66, "learning_rate": 8.338430173292559e-05, "loss": 3.0081, "step": 21190 }, { "epoch": 1.66, "learning_rate": 8.33835176037011e-05, "loss": 2.9861, "step": 21191 }, { "epoch": 1.66, "learning_rate": 8.33827334744766e-05, "loss": 2.8343, "step": 21192 }, { "epoch": 1.66, "learning_rate": 8.33819493452521e-05, "loss": 2.7252, "step": 21193 }, { "epoch": 1.66, "learning_rate": 8.33811652160276e-05, "loss": 3.137, "step": 21194 }, { "epoch": 1.66, "learning_rate": 8.338038108680311e-05, "loss": 2.8389, "step": 21195 }, { "epoch": 1.66, "learning_rate": 8.337959695757861e-05, "loss": 3.0028, "step": 21196 }, { "epoch": 1.66, "learning_rate": 8.337881282835412e-05, "loss": 2.7649, "step": 21197 }, { "epoch": 1.66, "learning_rate": 8.337802869912962e-05, "loss": 2.8221, "step": 21198 }, { "epoch": 1.66, "learning_rate": 8.337724456990512e-05, "loss": 2.7722, "step": 21199 }, { "epoch": 1.66, "learning_rate": 8.337646044068063e-05, "loss": 3.0618, "step": 21200 }, { "epoch": 1.66, "learning_rate": 8.337567631145613e-05, "loss": 2.887, "step": 21201 }, { "epoch": 1.66, "learning_rate": 8.337489218223164e-05, "loss": 2.8583, "step": 21202 }, { "epoch": 1.66, "learning_rate": 8.337410805300714e-05, "loss": 2.7546, "step": 21203 }, { "epoch": 1.66, "learning_rate": 8.337332392378264e-05, "loss": 2.6876, "step": 21204 }, { "epoch": 1.66, "learning_rate": 8.337253979455815e-05, "loss": 2.9219, "step": 21205 }, { "epoch": 1.66, "learning_rate": 8.337175566533365e-05, "loss": 2.9148, "step": 21206 }, { "epoch": 1.66, "learning_rate": 8.337097153610915e-05, "loss": 2.9584, "step": 21207 }, { "epoch": 1.66, "learning_rate": 8.337018740688466e-05, "loss": 2.8698, "step": 21208 }, { "epoch": 1.66, "learning_rate": 8.336940327766016e-05, "loss": 2.9929, "step": 21209 }, { "epoch": 1.66, "learning_rate": 8.336861914843567e-05, "loss": 2.8351, "step": 21210 }, { "epoch": 1.66, "learning_rate": 8.336783501921117e-05, "loss": 3.0621, "step": 21211 }, { "epoch": 1.66, "learning_rate": 8.336705088998667e-05, "loss": 2.8902, "step": 21212 }, { "epoch": 1.66, "learning_rate": 8.336626676076218e-05, "loss": 2.8887, "step": 21213 }, { "epoch": 1.66, "learning_rate": 8.336548263153768e-05, "loss": 3.0325, "step": 21214 }, { "epoch": 1.66, "learning_rate": 8.336469850231318e-05, "loss": 3.0153, "step": 21215 }, { "epoch": 1.66, "learning_rate": 8.336391437308869e-05, "loss": 2.805, "step": 21216 }, { "epoch": 1.66, "learning_rate": 8.336313024386419e-05, "loss": 2.9345, "step": 21217 }, { "epoch": 1.66, "learning_rate": 8.33623461146397e-05, "loss": 3.0312, "step": 21218 }, { "epoch": 1.66, "learning_rate": 8.33615619854152e-05, "loss": 2.7103, "step": 21219 }, { "epoch": 1.66, "learning_rate": 8.33607778561907e-05, "loss": 2.8562, "step": 21220 }, { "epoch": 1.66, "learning_rate": 8.335999372696622e-05, "loss": 2.7896, "step": 21221 }, { "epoch": 1.66, "learning_rate": 8.335920959774172e-05, "loss": 2.6735, "step": 21222 }, { "epoch": 1.66, "learning_rate": 8.335842546851723e-05, "loss": 2.8638, "step": 21223 }, { "epoch": 1.66, "learning_rate": 8.335764133929273e-05, "loss": 2.9852, "step": 21224 }, { "epoch": 1.66, "learning_rate": 8.335685721006822e-05, "loss": 2.7164, "step": 21225 }, { "epoch": 1.66, "learning_rate": 8.335607308084372e-05, "loss": 2.8105, "step": 21226 }, { "epoch": 1.66, "learning_rate": 8.335528895161923e-05, "loss": 2.7603, "step": 21227 }, { "epoch": 1.66, "learning_rate": 8.335450482239473e-05, "loss": 3.0281, "step": 21228 }, { "epoch": 1.66, "learning_rate": 8.335372069317023e-05, "loss": 3.0557, "step": 21229 }, { "epoch": 1.66, "learning_rate": 8.335293656394574e-05, "loss": 2.8222, "step": 21230 }, { "epoch": 1.66, "learning_rate": 8.335215243472124e-05, "loss": 3.0235, "step": 21231 }, { "epoch": 1.66, "learning_rate": 8.335136830549675e-05, "loss": 2.9084, "step": 21232 }, { "epoch": 1.66, "learning_rate": 8.335058417627225e-05, "loss": 2.8869, "step": 21233 }, { "epoch": 1.67, "learning_rate": 8.334980004704775e-05, "loss": 3.0347, "step": 21234 }, { "epoch": 1.67, "learning_rate": 8.334901591782326e-05, "loss": 2.7443, "step": 21235 }, { "epoch": 1.67, "learning_rate": 8.334823178859876e-05, "loss": 2.9998, "step": 21236 }, { "epoch": 1.67, "learning_rate": 8.334744765937426e-05, "loss": 2.8348, "step": 21237 }, { "epoch": 1.67, "learning_rate": 8.334666353014977e-05, "loss": 2.8288, "step": 21238 }, { "epoch": 1.67, "learning_rate": 8.334587940092527e-05, "loss": 2.6448, "step": 21239 }, { "epoch": 1.67, "learning_rate": 8.334509527170078e-05, "loss": 2.9707, "step": 21240 }, { "epoch": 1.67, "learning_rate": 8.334431114247628e-05, "loss": 3.0323, "step": 21241 }, { "epoch": 1.67, "learning_rate": 8.334352701325178e-05, "loss": 3.191, "step": 21242 }, { "epoch": 1.67, "learning_rate": 8.334274288402729e-05, "loss": 2.7947, "step": 21243 }, { "epoch": 1.67, "learning_rate": 8.334195875480279e-05, "loss": 2.8165, "step": 21244 }, { "epoch": 1.67, "learning_rate": 8.33411746255783e-05, "loss": 2.9336, "step": 21245 }, { "epoch": 1.67, "learning_rate": 8.33403904963538e-05, "loss": 2.9779, "step": 21246 }, { "epoch": 1.67, "learning_rate": 8.333960636712931e-05, "loss": 3.0672, "step": 21247 }, { "epoch": 1.67, "learning_rate": 8.333882223790482e-05, "loss": 2.8301, "step": 21248 }, { "epoch": 1.67, "learning_rate": 8.333803810868032e-05, "loss": 2.9933, "step": 21249 }, { "epoch": 1.67, "learning_rate": 8.333725397945583e-05, "loss": 2.9403, "step": 21250 }, { "epoch": 1.67, "learning_rate": 8.333646985023133e-05, "loss": 2.669, "step": 21251 }, { "epoch": 1.67, "learning_rate": 8.333568572100683e-05, "loss": 2.8122, "step": 21252 }, { "epoch": 1.67, "learning_rate": 8.333490159178234e-05, "loss": 2.8401, "step": 21253 }, { "epoch": 1.67, "learning_rate": 8.333411746255784e-05, "loss": 2.8022, "step": 21254 }, { "epoch": 1.67, "learning_rate": 8.333333333333334e-05, "loss": 2.9042, "step": 21255 }, { "epoch": 1.67, "learning_rate": 8.333254920410885e-05, "loss": 2.7561, "step": 21256 }, { "epoch": 1.67, "learning_rate": 8.333176507488435e-05, "loss": 2.8422, "step": 21257 }, { "epoch": 1.67, "learning_rate": 8.333098094565986e-05, "loss": 3.0217, "step": 21258 }, { "epoch": 1.67, "learning_rate": 8.333019681643536e-05, "loss": 3.1767, "step": 21259 }, { "epoch": 1.67, "learning_rate": 8.332941268721086e-05, "loss": 3.3521, "step": 21260 }, { "epoch": 1.67, "learning_rate": 8.332862855798637e-05, "loss": 2.6846, "step": 21261 }, { "epoch": 1.67, "learning_rate": 8.332784442876186e-05, "loss": 3.0312, "step": 21262 }, { "epoch": 1.67, "learning_rate": 8.332706029953736e-05, "loss": 2.9641, "step": 21263 }, { "epoch": 1.67, "learning_rate": 8.332627617031286e-05, "loss": 2.8834, "step": 21264 }, { "epoch": 1.67, "learning_rate": 8.332549204108837e-05, "loss": 3.0797, "step": 21265 }, { "epoch": 1.67, "learning_rate": 8.332470791186387e-05, "loss": 2.8303, "step": 21266 }, { "epoch": 1.67, "learning_rate": 8.332392378263937e-05, "loss": 2.7801, "step": 21267 }, { "epoch": 1.67, "learning_rate": 8.332313965341488e-05, "loss": 2.8991, "step": 21268 }, { "epoch": 1.67, "learning_rate": 8.332235552419038e-05, "loss": 3.0436, "step": 21269 }, { "epoch": 1.67, "learning_rate": 8.332157139496589e-05, "loss": 3.1222, "step": 21270 }, { "epoch": 1.67, "learning_rate": 8.332078726574139e-05, "loss": 3.1215, "step": 21271 }, { "epoch": 1.67, "learning_rate": 8.332000313651689e-05, "loss": 3.0255, "step": 21272 }, { "epoch": 1.67, "learning_rate": 8.331921900729241e-05, "loss": 3.1941, "step": 21273 }, { "epoch": 1.67, "learning_rate": 8.331843487806791e-05, "loss": 2.9943, "step": 21274 }, { "epoch": 1.67, "learning_rate": 8.331765074884342e-05, "loss": 3.0396, "step": 21275 }, { "epoch": 1.67, "learning_rate": 8.331686661961892e-05, "loss": 2.9779, "step": 21276 }, { "epoch": 1.67, "learning_rate": 8.331608249039443e-05, "loss": 2.7189, "step": 21277 }, { "epoch": 1.67, "learning_rate": 8.331529836116993e-05, "loss": 2.9401, "step": 21278 }, { "epoch": 1.67, "learning_rate": 8.331451423194543e-05, "loss": 2.9247, "step": 21279 }, { "epoch": 1.67, "learning_rate": 8.331373010272094e-05, "loss": 2.8799, "step": 21280 }, { "epoch": 1.67, "learning_rate": 8.331294597349644e-05, "loss": 2.9004, "step": 21281 }, { "epoch": 1.67, "learning_rate": 8.331216184427194e-05, "loss": 2.7541, "step": 21282 }, { "epoch": 1.67, "learning_rate": 8.331137771504745e-05, "loss": 2.9211, "step": 21283 }, { "epoch": 1.67, "learning_rate": 8.331059358582295e-05, "loss": 3.0164, "step": 21284 }, { "epoch": 1.67, "learning_rate": 8.330980945659845e-05, "loss": 2.897, "step": 21285 }, { "epoch": 1.67, "learning_rate": 8.330902532737396e-05, "loss": 2.8341, "step": 21286 }, { "epoch": 1.67, "learning_rate": 8.330824119814946e-05, "loss": 2.8929, "step": 21287 }, { "epoch": 1.67, "learning_rate": 8.330745706892497e-05, "loss": 2.609, "step": 21288 }, { "epoch": 1.67, "learning_rate": 8.330667293970047e-05, "loss": 2.8906, "step": 21289 }, { "epoch": 1.67, "learning_rate": 8.330588881047597e-05, "loss": 2.7787, "step": 21290 }, { "epoch": 1.67, "learning_rate": 8.330510468125148e-05, "loss": 2.9121, "step": 21291 }, { "epoch": 1.67, "learning_rate": 8.330432055202698e-05, "loss": 2.7321, "step": 21292 }, { "epoch": 1.67, "learning_rate": 8.330353642280248e-05, "loss": 2.9853, "step": 21293 }, { "epoch": 1.67, "learning_rate": 8.330275229357799e-05, "loss": 2.8153, "step": 21294 }, { "epoch": 1.67, "learning_rate": 8.330196816435349e-05, "loss": 3.061, "step": 21295 }, { "epoch": 1.67, "learning_rate": 8.3301184035129e-05, "loss": 3.1125, "step": 21296 }, { "epoch": 1.67, "learning_rate": 8.33003999059045e-05, "loss": 2.7953, "step": 21297 }, { "epoch": 1.67, "learning_rate": 8.329961577668e-05, "loss": 3.0247, "step": 21298 }, { "epoch": 1.67, "learning_rate": 8.32988316474555e-05, "loss": 3.0976, "step": 21299 }, { "epoch": 1.67, "learning_rate": 8.329804751823101e-05, "loss": 2.9126, "step": 21300 }, { "epoch": 1.67, "learning_rate": 8.329726338900651e-05, "loss": 2.8053, "step": 21301 }, { "epoch": 1.67, "learning_rate": 8.329647925978202e-05, "loss": 3.1016, "step": 21302 }, { "epoch": 1.67, "learning_rate": 8.329569513055752e-05, "loss": 3.0985, "step": 21303 }, { "epoch": 1.67, "learning_rate": 8.329491100133302e-05, "loss": 2.902, "step": 21304 }, { "epoch": 1.67, "learning_rate": 8.329412687210853e-05, "loss": 2.696, "step": 21305 }, { "epoch": 1.67, "learning_rate": 8.329334274288403e-05, "loss": 2.9449, "step": 21306 }, { "epoch": 1.67, "learning_rate": 8.329255861365954e-05, "loss": 2.9781, "step": 21307 }, { "epoch": 1.67, "learning_rate": 8.329177448443504e-05, "loss": 2.8544, "step": 21308 }, { "epoch": 1.67, "learning_rate": 8.329099035521054e-05, "loss": 3.2863, "step": 21309 }, { "epoch": 1.67, "learning_rate": 8.329020622598605e-05, "loss": 3.0721, "step": 21310 }, { "epoch": 1.67, "learning_rate": 8.328942209676155e-05, "loss": 3.1948, "step": 21311 }, { "epoch": 1.67, "learning_rate": 8.328863796753705e-05, "loss": 3.1192, "step": 21312 }, { "epoch": 1.67, "learning_rate": 8.328785383831256e-05, "loss": 3.1045, "step": 21313 }, { "epoch": 1.67, "learning_rate": 8.328706970908806e-05, "loss": 2.8136, "step": 21314 }, { "epoch": 1.67, "learning_rate": 8.328628557986356e-05, "loss": 2.8655, "step": 21315 }, { "epoch": 1.67, "learning_rate": 8.328550145063907e-05, "loss": 2.8339, "step": 21316 }, { "epoch": 1.67, "learning_rate": 8.328471732141457e-05, "loss": 2.8301, "step": 21317 }, { "epoch": 1.67, "learning_rate": 8.328393319219008e-05, "loss": 2.8774, "step": 21318 }, { "epoch": 1.67, "learning_rate": 8.328314906296558e-05, "loss": 3.0445, "step": 21319 }, { "epoch": 1.67, "learning_rate": 8.328236493374108e-05, "loss": 3.0629, "step": 21320 }, { "epoch": 1.67, "learning_rate": 8.328158080451659e-05, "loss": 3.2874, "step": 21321 }, { "epoch": 1.67, "learning_rate": 8.328079667529209e-05, "loss": 2.964, "step": 21322 }, { "epoch": 1.67, "learning_rate": 8.32800125460676e-05, "loss": 2.8068, "step": 21323 }, { "epoch": 1.67, "learning_rate": 8.32792284168431e-05, "loss": 3.0844, "step": 21324 }, { "epoch": 1.67, "learning_rate": 8.32784442876186e-05, "loss": 2.8613, "step": 21325 }, { "epoch": 1.67, "learning_rate": 8.32776601583941e-05, "loss": 3.1139, "step": 21326 }, { "epoch": 1.67, "learning_rate": 8.327687602916961e-05, "loss": 2.8958, "step": 21327 }, { "epoch": 1.67, "learning_rate": 8.327609189994511e-05, "loss": 2.7687, "step": 21328 }, { "epoch": 1.67, "learning_rate": 8.327530777072063e-05, "loss": 2.8577, "step": 21329 }, { "epoch": 1.67, "learning_rate": 8.327452364149613e-05, "loss": 3.1851, "step": 21330 }, { "epoch": 1.67, "learning_rate": 8.327373951227164e-05, "loss": 3.1886, "step": 21331 }, { "epoch": 1.67, "learning_rate": 8.327295538304714e-05, "loss": 2.9712, "step": 21332 }, { "epoch": 1.67, "learning_rate": 8.327217125382264e-05, "loss": 2.9448, "step": 21333 }, { "epoch": 1.67, "learning_rate": 8.327138712459813e-05, "loss": 2.7989, "step": 21334 }, { "epoch": 1.67, "learning_rate": 8.327060299537364e-05, "loss": 2.9217, "step": 21335 }, { "epoch": 1.67, "learning_rate": 8.326981886614914e-05, "loss": 2.7689, "step": 21336 }, { "epoch": 1.67, "learning_rate": 8.326903473692465e-05, "loss": 2.9191, "step": 21337 }, { "epoch": 1.67, "learning_rate": 8.326825060770015e-05, "loss": 2.791, "step": 21338 }, { "epoch": 1.67, "learning_rate": 8.326746647847565e-05, "loss": 3.0295, "step": 21339 }, { "epoch": 1.67, "learning_rate": 8.326668234925116e-05, "loss": 3.0508, "step": 21340 }, { "epoch": 1.67, "learning_rate": 8.326589822002666e-05, "loss": 2.9972, "step": 21341 }, { "epoch": 1.67, "learning_rate": 8.326511409080216e-05, "loss": 2.7435, "step": 21342 }, { "epoch": 1.67, "learning_rate": 8.326432996157767e-05, "loss": 3.1584, "step": 21343 }, { "epoch": 1.67, "learning_rate": 8.326354583235317e-05, "loss": 3.063, "step": 21344 }, { "epoch": 1.67, "learning_rate": 8.326276170312868e-05, "loss": 2.7456, "step": 21345 }, { "epoch": 1.67, "learning_rate": 8.326197757390418e-05, "loss": 2.8076, "step": 21346 }, { "epoch": 1.67, "learning_rate": 8.326119344467968e-05, "loss": 3.007, "step": 21347 }, { "epoch": 1.67, "learning_rate": 8.326040931545519e-05, "loss": 2.7672, "step": 21348 }, { "epoch": 1.67, "learning_rate": 8.325962518623069e-05, "loss": 2.8401, "step": 21349 }, { "epoch": 1.67, "learning_rate": 8.32588410570062e-05, "loss": 3.0982, "step": 21350 }, { "epoch": 1.67, "learning_rate": 8.32580569277817e-05, "loss": 2.9164, "step": 21351 }, { "epoch": 1.67, "learning_rate": 8.32572727985572e-05, "loss": 2.8975, "step": 21352 }, { "epoch": 1.67, "learning_rate": 8.32564886693327e-05, "loss": 2.9717, "step": 21353 }, { "epoch": 1.67, "learning_rate": 8.325570454010821e-05, "loss": 3.0839, "step": 21354 }, { "epoch": 1.67, "learning_rate": 8.325492041088373e-05, "loss": 2.937, "step": 21355 }, { "epoch": 1.67, "learning_rate": 8.325413628165923e-05, "loss": 3.093, "step": 21356 }, { "epoch": 1.67, "learning_rate": 8.325335215243473e-05, "loss": 2.7873, "step": 21357 }, { "epoch": 1.67, "learning_rate": 8.325256802321024e-05, "loss": 2.8881, "step": 21358 }, { "epoch": 1.67, "learning_rate": 8.325178389398574e-05, "loss": 3.0223, "step": 21359 }, { "epoch": 1.67, "learning_rate": 8.325099976476124e-05, "loss": 2.6518, "step": 21360 }, { "epoch": 1.67, "learning_rate": 8.325021563553675e-05, "loss": 2.8432, "step": 21361 }, { "epoch": 1.68, "learning_rate": 8.324943150631225e-05, "loss": 2.6655, "step": 21362 }, { "epoch": 1.68, "learning_rate": 8.324864737708775e-05, "loss": 2.8116, "step": 21363 }, { "epoch": 1.68, "learning_rate": 8.324786324786326e-05, "loss": 2.848, "step": 21364 }, { "epoch": 1.68, "learning_rate": 8.324707911863876e-05, "loss": 3.1068, "step": 21365 }, { "epoch": 1.68, "learning_rate": 8.324629498941427e-05, "loss": 3.0127, "step": 21366 }, { "epoch": 1.68, "learning_rate": 8.324551086018977e-05, "loss": 2.9464, "step": 21367 }, { "epoch": 1.68, "learning_rate": 8.324472673096527e-05, "loss": 2.8369, "step": 21368 }, { "epoch": 1.68, "learning_rate": 8.324394260174078e-05, "loss": 2.761, "step": 21369 }, { "epoch": 1.68, "learning_rate": 8.324315847251627e-05, "loss": 2.7461, "step": 21370 }, { "epoch": 1.68, "learning_rate": 8.324237434329177e-05, "loss": 3.0427, "step": 21371 }, { "epoch": 1.68, "learning_rate": 8.324159021406727e-05, "loss": 2.9562, "step": 21372 }, { "epoch": 1.68, "learning_rate": 8.324080608484278e-05, "loss": 3.0144, "step": 21373 }, { "epoch": 1.68, "learning_rate": 8.324002195561828e-05, "loss": 2.9104, "step": 21374 }, { "epoch": 1.68, "learning_rate": 8.323923782639379e-05, "loss": 2.8148, "step": 21375 }, { "epoch": 1.68, "learning_rate": 8.323845369716929e-05, "loss": 2.7198, "step": 21376 }, { "epoch": 1.68, "learning_rate": 8.323766956794479e-05, "loss": 2.9056, "step": 21377 }, { "epoch": 1.68, "learning_rate": 8.32368854387203e-05, "loss": 3.1343, "step": 21378 }, { "epoch": 1.68, "learning_rate": 8.32361013094958e-05, "loss": 2.7687, "step": 21379 }, { "epoch": 1.68, "learning_rate": 8.32353171802713e-05, "loss": 2.8857, "step": 21380 }, { "epoch": 1.68, "learning_rate": 8.323453305104682e-05, "loss": 3.2356, "step": 21381 }, { "epoch": 1.68, "learning_rate": 8.323374892182232e-05, "loss": 2.7618, "step": 21382 }, { "epoch": 1.68, "learning_rate": 8.323296479259783e-05, "loss": 2.7649, "step": 21383 }, { "epoch": 1.68, "learning_rate": 8.323218066337333e-05, "loss": 2.771, "step": 21384 }, { "epoch": 1.68, "learning_rate": 8.323139653414884e-05, "loss": 2.7803, "step": 21385 }, { "epoch": 1.68, "learning_rate": 8.323061240492434e-05, "loss": 2.7879, "step": 21386 }, { "epoch": 1.68, "learning_rate": 8.322982827569984e-05, "loss": 2.8579, "step": 21387 }, { "epoch": 1.68, "learning_rate": 8.322904414647535e-05, "loss": 3.019, "step": 21388 }, { "epoch": 1.68, "learning_rate": 8.322826001725085e-05, "loss": 2.7818, "step": 21389 }, { "epoch": 1.68, "learning_rate": 8.322747588802635e-05, "loss": 3.107, "step": 21390 }, { "epoch": 1.68, "learning_rate": 8.322669175880186e-05, "loss": 2.95, "step": 21391 }, { "epoch": 1.68, "learning_rate": 8.322590762957736e-05, "loss": 2.9003, "step": 21392 }, { "epoch": 1.68, "learning_rate": 8.322512350035287e-05, "loss": 3.0882, "step": 21393 }, { "epoch": 1.68, "learning_rate": 8.322433937112837e-05, "loss": 2.8301, "step": 21394 }, { "epoch": 1.68, "learning_rate": 8.322355524190387e-05, "loss": 2.799, "step": 21395 }, { "epoch": 1.68, "learning_rate": 8.322277111267938e-05, "loss": 2.833, "step": 21396 }, { "epoch": 1.68, "learning_rate": 8.322198698345488e-05, "loss": 2.9844, "step": 21397 }, { "epoch": 1.68, "learning_rate": 8.322120285423038e-05, "loss": 2.8002, "step": 21398 }, { "epoch": 1.68, "learning_rate": 8.322041872500589e-05, "loss": 2.7482, "step": 21399 }, { "epoch": 1.68, "learning_rate": 8.321963459578139e-05, "loss": 2.7022, "step": 21400 }, { "epoch": 1.68, "learning_rate": 8.32188504665569e-05, "loss": 2.9741, "step": 21401 }, { "epoch": 1.68, "learning_rate": 8.32180663373324e-05, "loss": 3.0351, "step": 21402 }, { "epoch": 1.68, "learning_rate": 8.32172822081079e-05, "loss": 3.0054, "step": 21403 }, { "epoch": 1.68, "learning_rate": 8.32164980788834e-05, "loss": 2.76, "step": 21404 }, { "epoch": 1.68, "learning_rate": 8.321571394965891e-05, "loss": 2.8789, "step": 21405 }, { "epoch": 1.68, "learning_rate": 8.321492982043441e-05, "loss": 2.9681, "step": 21406 }, { "epoch": 1.68, "learning_rate": 8.321414569120992e-05, "loss": 3.0454, "step": 21407 }, { "epoch": 1.68, "learning_rate": 8.321336156198542e-05, "loss": 2.7939, "step": 21408 }, { "epoch": 1.68, "learning_rate": 8.321257743276092e-05, "loss": 2.8306, "step": 21409 }, { "epoch": 1.68, "learning_rate": 8.321179330353643e-05, "loss": 3.2054, "step": 21410 }, { "epoch": 1.68, "learning_rate": 8.321100917431193e-05, "loss": 2.7272, "step": 21411 }, { "epoch": 1.68, "learning_rate": 8.321022504508744e-05, "loss": 3.0605, "step": 21412 }, { "epoch": 1.68, "learning_rate": 8.320944091586294e-05, "loss": 2.8822, "step": 21413 }, { "epoch": 1.68, "learning_rate": 8.320865678663844e-05, "loss": 2.8305, "step": 21414 }, { "epoch": 1.68, "learning_rate": 8.320787265741395e-05, "loss": 2.6565, "step": 21415 }, { "epoch": 1.68, "learning_rate": 8.320708852818945e-05, "loss": 2.9396, "step": 21416 }, { "epoch": 1.68, "learning_rate": 8.320630439896495e-05, "loss": 2.8162, "step": 21417 }, { "epoch": 1.68, "learning_rate": 8.320552026974046e-05, "loss": 3.0285, "step": 21418 }, { "epoch": 1.68, "learning_rate": 8.320473614051596e-05, "loss": 3.0128, "step": 21419 }, { "epoch": 1.68, "learning_rate": 8.320395201129146e-05, "loss": 2.8843, "step": 21420 }, { "epoch": 1.68, "learning_rate": 8.320316788206697e-05, "loss": 2.7551, "step": 21421 }, { "epoch": 1.68, "learning_rate": 8.320238375284247e-05, "loss": 3.0167, "step": 21422 }, { "epoch": 1.68, "learning_rate": 8.320159962361798e-05, "loss": 2.7434, "step": 21423 }, { "epoch": 1.68, "learning_rate": 8.320081549439348e-05, "loss": 3.0397, "step": 21424 }, { "epoch": 1.68, "learning_rate": 8.320003136516898e-05, "loss": 2.9415, "step": 21425 }, { "epoch": 1.68, "learning_rate": 8.319924723594449e-05, "loss": 2.8, "step": 21426 }, { "epoch": 1.68, "learning_rate": 8.319846310671999e-05, "loss": 2.9108, "step": 21427 }, { "epoch": 1.68, "learning_rate": 8.31976789774955e-05, "loss": 2.8078, "step": 21428 }, { "epoch": 1.68, "learning_rate": 8.3196894848271e-05, "loss": 2.7821, "step": 21429 }, { "epoch": 1.68, "learning_rate": 8.31961107190465e-05, "loss": 2.9197, "step": 21430 }, { "epoch": 1.68, "learning_rate": 8.3195326589822e-05, "loss": 2.9385, "step": 21431 }, { "epoch": 1.68, "learning_rate": 8.319454246059751e-05, "loss": 2.8515, "step": 21432 }, { "epoch": 1.68, "learning_rate": 8.319375833137301e-05, "loss": 2.8772, "step": 21433 }, { "epoch": 1.68, "learning_rate": 8.319297420214852e-05, "loss": 2.8425, "step": 21434 }, { "epoch": 1.68, "learning_rate": 8.319219007292402e-05, "loss": 2.9137, "step": 21435 }, { "epoch": 1.68, "learning_rate": 8.319140594369952e-05, "loss": 2.8043, "step": 21436 }, { "epoch": 1.68, "learning_rate": 8.319062181447504e-05, "loss": 2.9573, "step": 21437 }, { "epoch": 1.68, "learning_rate": 8.318983768525054e-05, "loss": 3.0815, "step": 21438 }, { "epoch": 1.68, "learning_rate": 8.318905355602605e-05, "loss": 2.8957, "step": 21439 }, { "epoch": 1.68, "learning_rate": 8.318826942680155e-05, "loss": 3.0797, "step": 21440 }, { "epoch": 1.68, "learning_rate": 8.318748529757706e-05, "loss": 2.8921, "step": 21441 }, { "epoch": 1.68, "learning_rate": 8.318670116835255e-05, "loss": 2.6505, "step": 21442 }, { "epoch": 1.68, "learning_rate": 8.318591703912805e-05, "loss": 3.0229, "step": 21443 }, { "epoch": 1.68, "learning_rate": 8.318513290990355e-05, "loss": 3.0182, "step": 21444 }, { "epoch": 1.68, "learning_rate": 8.318434878067906e-05, "loss": 2.8814, "step": 21445 }, { "epoch": 1.68, "learning_rate": 8.318356465145456e-05, "loss": 2.6373, "step": 21446 }, { "epoch": 1.68, "learning_rate": 8.318278052223006e-05, "loss": 2.7902, "step": 21447 }, { "epoch": 1.68, "learning_rate": 8.318199639300557e-05, "loss": 2.8781, "step": 21448 }, { "epoch": 1.68, "learning_rate": 8.318121226378107e-05, "loss": 2.9017, "step": 21449 }, { "epoch": 1.68, "learning_rate": 8.318042813455657e-05, "loss": 2.6008, "step": 21450 }, { "epoch": 1.68, "learning_rate": 8.317964400533208e-05, "loss": 2.7498, "step": 21451 }, { "epoch": 1.68, "learning_rate": 8.317885987610758e-05, "loss": 3.0088, "step": 21452 }, { "epoch": 1.68, "learning_rate": 8.317807574688309e-05, "loss": 2.802, "step": 21453 }, { "epoch": 1.68, "learning_rate": 8.317729161765859e-05, "loss": 3.0116, "step": 21454 }, { "epoch": 1.68, "learning_rate": 8.317650748843409e-05, "loss": 2.9273, "step": 21455 }, { "epoch": 1.68, "learning_rate": 8.31757233592096e-05, "loss": 2.9946, "step": 21456 }, { "epoch": 1.68, "learning_rate": 8.31749392299851e-05, "loss": 2.6185, "step": 21457 }, { "epoch": 1.68, "learning_rate": 8.31741551007606e-05, "loss": 3.2521, "step": 21458 }, { "epoch": 1.68, "learning_rate": 8.317337097153611e-05, "loss": 2.9286, "step": 21459 }, { "epoch": 1.68, "learning_rate": 8.317258684231161e-05, "loss": 2.9113, "step": 21460 }, { "epoch": 1.68, "learning_rate": 8.317180271308712e-05, "loss": 2.6788, "step": 21461 }, { "epoch": 1.68, "learning_rate": 8.317101858386262e-05, "loss": 2.86, "step": 21462 }, { "epoch": 1.68, "learning_rate": 8.317023445463814e-05, "loss": 2.9611, "step": 21463 }, { "epoch": 1.68, "learning_rate": 8.316945032541364e-05, "loss": 3.0423, "step": 21464 }, { "epoch": 1.68, "learning_rate": 8.316866619618914e-05, "loss": 2.868, "step": 21465 }, { "epoch": 1.68, "learning_rate": 8.316788206696465e-05, "loss": 2.8527, "step": 21466 }, { "epoch": 1.68, "learning_rate": 8.316709793774015e-05, "loss": 2.7428, "step": 21467 }, { "epoch": 1.68, "learning_rate": 8.316631380851565e-05, "loss": 3.0611, "step": 21468 }, { "epoch": 1.68, "learning_rate": 8.316552967929116e-05, "loss": 2.7556, "step": 21469 }, { "epoch": 1.68, "learning_rate": 8.316474555006666e-05, "loss": 2.9392, "step": 21470 }, { "epoch": 1.68, "learning_rate": 8.316396142084217e-05, "loss": 3.0727, "step": 21471 }, { "epoch": 1.68, "learning_rate": 8.316317729161767e-05, "loss": 2.9308, "step": 21472 }, { "epoch": 1.68, "learning_rate": 8.316239316239317e-05, "loss": 2.8733, "step": 21473 }, { "epoch": 1.68, "learning_rate": 8.316160903316868e-05, "loss": 2.9323, "step": 21474 }, { "epoch": 1.68, "learning_rate": 8.316082490394418e-05, "loss": 2.7594, "step": 21475 }, { "epoch": 1.68, "learning_rate": 8.316004077471968e-05, "loss": 2.9055, "step": 21476 }, { "epoch": 1.68, "learning_rate": 8.315925664549519e-05, "loss": 2.9548, "step": 21477 }, { "epoch": 1.68, "learning_rate": 8.315847251627068e-05, "loss": 3.0744, "step": 21478 }, { "epoch": 1.68, "learning_rate": 8.315768838704618e-05, "loss": 3.0927, "step": 21479 }, { "epoch": 1.68, "learning_rate": 8.315690425782169e-05, "loss": 3.0263, "step": 21480 }, { "epoch": 1.68, "learning_rate": 8.315612012859719e-05, "loss": 2.8058, "step": 21481 }, { "epoch": 1.68, "learning_rate": 8.315533599937269e-05, "loss": 2.9279, "step": 21482 }, { "epoch": 1.68, "learning_rate": 8.31545518701482e-05, "loss": 2.8965, "step": 21483 }, { "epoch": 1.68, "learning_rate": 8.31537677409237e-05, "loss": 3.154, "step": 21484 }, { "epoch": 1.68, "learning_rate": 8.31529836116992e-05, "loss": 2.8479, "step": 21485 }, { "epoch": 1.68, "learning_rate": 8.315219948247471e-05, "loss": 2.9293, "step": 21486 }, { "epoch": 1.68, "learning_rate": 8.315141535325021e-05, "loss": 2.8499, "step": 21487 }, { "epoch": 1.68, "learning_rate": 8.315063122402571e-05, "loss": 2.7738, "step": 21488 }, { "epoch": 1.69, "learning_rate": 8.314984709480123e-05, "loss": 2.9012, "step": 21489 }, { "epoch": 1.69, "learning_rate": 8.314906296557674e-05, "loss": 2.9989, "step": 21490 }, { "epoch": 1.69, "learning_rate": 8.314827883635224e-05, "loss": 2.9108, "step": 21491 }, { "epoch": 1.69, "learning_rate": 8.314749470712774e-05, "loss": 2.6884, "step": 21492 }, { "epoch": 1.69, "learning_rate": 8.314671057790325e-05, "loss": 2.8697, "step": 21493 }, { "epoch": 1.69, "learning_rate": 8.314592644867875e-05, "loss": 2.7791, "step": 21494 }, { "epoch": 1.69, "learning_rate": 8.314514231945425e-05, "loss": 3.096, "step": 21495 }, { "epoch": 1.69, "learning_rate": 8.314435819022976e-05, "loss": 2.9938, "step": 21496 }, { "epoch": 1.69, "learning_rate": 8.314357406100526e-05, "loss": 2.8494, "step": 21497 }, { "epoch": 1.69, "learning_rate": 8.314278993178076e-05, "loss": 2.9949, "step": 21498 }, { "epoch": 1.69, "learning_rate": 8.314200580255627e-05, "loss": 3.124, "step": 21499 }, { "epoch": 1.69, "learning_rate": 8.314122167333177e-05, "loss": 2.8939, "step": 21500 }, { "epoch": 1.69, "learning_rate": 8.314043754410728e-05, "loss": 2.8719, "step": 21501 }, { "epoch": 1.69, "learning_rate": 8.313965341488278e-05, "loss": 2.8088, "step": 21502 }, { "epoch": 1.69, "learning_rate": 8.313886928565828e-05, "loss": 2.9236, "step": 21503 }, { "epoch": 1.69, "learning_rate": 8.313808515643379e-05, "loss": 3.1307, "step": 21504 }, { "epoch": 1.69, "learning_rate": 8.313730102720929e-05, "loss": 2.9477, "step": 21505 }, { "epoch": 1.69, "learning_rate": 8.31365168979848e-05, "loss": 2.9556, "step": 21506 }, { "epoch": 1.69, "learning_rate": 8.31357327687603e-05, "loss": 2.7675, "step": 21507 }, { "epoch": 1.69, "learning_rate": 8.31349486395358e-05, "loss": 3.2209, "step": 21508 }, { "epoch": 1.69, "learning_rate": 8.31341645103113e-05, "loss": 3.1285, "step": 21509 }, { "epoch": 1.69, "learning_rate": 8.313338038108681e-05, "loss": 2.8239, "step": 21510 }, { "epoch": 1.69, "learning_rate": 8.313259625186231e-05, "loss": 2.9603, "step": 21511 }, { "epoch": 1.69, "learning_rate": 8.313181212263782e-05, "loss": 3.025, "step": 21512 }, { "epoch": 1.69, "learning_rate": 8.313102799341332e-05, "loss": 3.2656, "step": 21513 }, { "epoch": 1.69, "learning_rate": 8.313024386418882e-05, "loss": 2.86, "step": 21514 }, { "epoch": 1.69, "learning_rate": 8.312945973496433e-05, "loss": 2.6849, "step": 21515 }, { "epoch": 1.69, "learning_rate": 8.312867560573983e-05, "loss": 2.9209, "step": 21516 }, { "epoch": 1.69, "learning_rate": 8.312789147651533e-05, "loss": 2.6813, "step": 21517 }, { "epoch": 1.69, "learning_rate": 8.312710734729084e-05, "loss": 3.1496, "step": 21518 }, { "epoch": 1.69, "learning_rate": 8.312632321806634e-05, "loss": 2.868, "step": 21519 }, { "epoch": 1.69, "learning_rate": 8.312553908884185e-05, "loss": 2.8727, "step": 21520 }, { "epoch": 1.69, "learning_rate": 8.312475495961735e-05, "loss": 2.7665, "step": 21521 }, { "epoch": 1.69, "learning_rate": 8.312397083039285e-05, "loss": 2.6206, "step": 21522 }, { "epoch": 1.69, "learning_rate": 8.312318670116836e-05, "loss": 2.8426, "step": 21523 }, { "epoch": 1.69, "learning_rate": 8.312240257194386e-05, "loss": 2.7857, "step": 21524 }, { "epoch": 1.69, "learning_rate": 8.312161844271936e-05, "loss": 2.9059, "step": 21525 }, { "epoch": 1.69, "learning_rate": 8.312083431349487e-05, "loss": 2.9708, "step": 21526 }, { "epoch": 1.69, "learning_rate": 8.312005018427037e-05, "loss": 2.8087, "step": 21527 }, { "epoch": 1.69, "learning_rate": 8.311926605504588e-05, "loss": 2.7224, "step": 21528 }, { "epoch": 1.69, "learning_rate": 8.311848192582138e-05, "loss": 3.0204, "step": 21529 }, { "epoch": 1.69, "learning_rate": 8.311769779659688e-05, "loss": 2.7803, "step": 21530 }, { "epoch": 1.69, "learning_rate": 8.311691366737239e-05, "loss": 2.9615, "step": 21531 }, { "epoch": 1.69, "learning_rate": 8.311612953814789e-05, "loss": 2.9577, "step": 21532 }, { "epoch": 1.69, "learning_rate": 8.31153454089234e-05, "loss": 2.9259, "step": 21533 }, { "epoch": 1.69, "learning_rate": 8.31145612796989e-05, "loss": 2.9915, "step": 21534 }, { "epoch": 1.69, "learning_rate": 8.31137771504744e-05, "loss": 2.9291, "step": 21535 }, { "epoch": 1.69, "learning_rate": 8.31129930212499e-05, "loss": 2.8709, "step": 21536 }, { "epoch": 1.69, "learning_rate": 8.311220889202541e-05, "loss": 2.924, "step": 21537 }, { "epoch": 1.69, "learning_rate": 8.311142476280091e-05, "loss": 2.9487, "step": 21538 }, { "epoch": 1.69, "learning_rate": 8.311064063357642e-05, "loss": 2.8461, "step": 21539 }, { "epoch": 1.69, "learning_rate": 8.310985650435192e-05, "loss": 2.8445, "step": 21540 }, { "epoch": 1.69, "learning_rate": 8.310907237512742e-05, "loss": 2.9215, "step": 21541 }, { "epoch": 1.69, "learning_rate": 8.310828824590293e-05, "loss": 2.8323, "step": 21542 }, { "epoch": 1.69, "learning_rate": 8.310750411667843e-05, "loss": 2.654, "step": 21543 }, { "epoch": 1.69, "learning_rate": 8.310671998745393e-05, "loss": 2.8398, "step": 21544 }, { "epoch": 1.69, "learning_rate": 8.310593585822944e-05, "loss": 2.7003, "step": 21545 }, { "epoch": 1.69, "learning_rate": 8.310515172900495e-05, "loss": 2.8012, "step": 21546 }, { "epoch": 1.69, "learning_rate": 8.310436759978046e-05, "loss": 2.9336, "step": 21547 }, { "epoch": 1.69, "learning_rate": 8.310358347055596e-05, "loss": 2.9191, "step": 21548 }, { "epoch": 1.69, "learning_rate": 8.310279934133147e-05, "loss": 2.9127, "step": 21549 }, { "epoch": 1.69, "learning_rate": 8.310201521210696e-05, "loss": 3.1164, "step": 21550 }, { "epoch": 1.69, "learning_rate": 8.310123108288246e-05, "loss": 2.8916, "step": 21551 }, { "epoch": 1.69, "learning_rate": 8.310044695365796e-05, "loss": 3.1803, "step": 21552 }, { "epoch": 1.69, "learning_rate": 8.309966282443347e-05, "loss": 3.0074, "step": 21553 }, { "epoch": 1.69, "learning_rate": 8.309887869520897e-05, "loss": 2.8797, "step": 21554 }, { "epoch": 1.69, "learning_rate": 8.309809456598447e-05, "loss": 3.1073, "step": 21555 }, { "epoch": 1.69, "learning_rate": 8.309731043675998e-05, "loss": 2.9049, "step": 21556 }, { "epoch": 1.69, "learning_rate": 8.309652630753548e-05, "loss": 2.7908, "step": 21557 }, { "epoch": 1.69, "learning_rate": 8.309574217831099e-05, "loss": 2.7843, "step": 21558 }, { "epoch": 1.69, "learning_rate": 8.309495804908649e-05, "loss": 2.8099, "step": 21559 }, { "epoch": 1.69, "learning_rate": 8.309417391986199e-05, "loss": 2.714, "step": 21560 }, { "epoch": 1.69, "learning_rate": 8.30933897906375e-05, "loss": 2.7801, "step": 21561 }, { "epoch": 1.69, "learning_rate": 8.3092605661413e-05, "loss": 2.8804, "step": 21562 }, { "epoch": 1.69, "learning_rate": 8.30918215321885e-05, "loss": 2.8332, "step": 21563 }, { "epoch": 1.69, "learning_rate": 8.309103740296401e-05, "loss": 2.8597, "step": 21564 }, { "epoch": 1.69, "learning_rate": 8.309025327373951e-05, "loss": 2.7941, "step": 21565 }, { "epoch": 1.69, "learning_rate": 8.308946914451501e-05, "loss": 2.6762, "step": 21566 }, { "epoch": 1.69, "learning_rate": 8.308868501529052e-05, "loss": 2.8797, "step": 21567 }, { "epoch": 1.69, "learning_rate": 8.308790088606602e-05, "loss": 2.9568, "step": 21568 }, { "epoch": 1.69, "learning_rate": 8.308711675684153e-05, "loss": 2.7298, "step": 21569 }, { "epoch": 1.69, "learning_rate": 8.308633262761703e-05, "loss": 3.0908, "step": 21570 }, { "epoch": 1.69, "learning_rate": 8.308554849839255e-05, "loss": 2.8403, "step": 21571 }, { "epoch": 1.69, "learning_rate": 8.308476436916805e-05, "loss": 2.9443, "step": 21572 }, { "epoch": 1.69, "learning_rate": 8.308398023994355e-05, "loss": 2.9936, "step": 21573 }, { "epoch": 1.69, "learning_rate": 8.308319611071906e-05, "loss": 2.9555, "step": 21574 }, { "epoch": 1.69, "learning_rate": 8.308241198149456e-05, "loss": 2.9895, "step": 21575 }, { "epoch": 1.69, "learning_rate": 8.308162785227007e-05, "loss": 2.7535, "step": 21576 }, { "epoch": 1.69, "learning_rate": 8.308084372304557e-05, "loss": 2.9593, "step": 21577 }, { "epoch": 1.69, "learning_rate": 8.308005959382107e-05, "loss": 3.0519, "step": 21578 }, { "epoch": 1.69, "learning_rate": 8.307927546459658e-05, "loss": 2.869, "step": 21579 }, { "epoch": 1.69, "learning_rate": 8.307849133537208e-05, "loss": 2.9541, "step": 21580 }, { "epoch": 1.69, "learning_rate": 8.307770720614758e-05, "loss": 2.7446, "step": 21581 }, { "epoch": 1.69, "learning_rate": 8.307692307692309e-05, "loss": 2.8725, "step": 21582 }, { "epoch": 1.69, "learning_rate": 8.307613894769859e-05, "loss": 2.8562, "step": 21583 }, { "epoch": 1.69, "learning_rate": 8.30753548184741e-05, "loss": 2.9503, "step": 21584 }, { "epoch": 1.69, "learning_rate": 8.30745706892496e-05, "loss": 2.8049, "step": 21585 }, { "epoch": 1.69, "learning_rate": 8.307378656002509e-05, "loss": 2.7505, "step": 21586 }, { "epoch": 1.69, "learning_rate": 8.307300243080059e-05, "loss": 3.0081, "step": 21587 }, { "epoch": 1.69, "learning_rate": 8.30722183015761e-05, "loss": 2.9221, "step": 21588 }, { "epoch": 1.69, "learning_rate": 8.30714341723516e-05, "loss": 2.7908, "step": 21589 }, { "epoch": 1.69, "learning_rate": 8.30706500431271e-05, "loss": 2.997, "step": 21590 }, { "epoch": 1.69, "learning_rate": 8.306986591390261e-05, "loss": 2.8636, "step": 21591 }, { "epoch": 1.69, "learning_rate": 8.306908178467811e-05, "loss": 2.9345, "step": 21592 }, { "epoch": 1.69, "learning_rate": 8.306829765545361e-05, "loss": 2.9244, "step": 21593 }, { "epoch": 1.69, "learning_rate": 8.306751352622912e-05, "loss": 3.1234, "step": 21594 }, { "epoch": 1.69, "learning_rate": 8.306672939700462e-05, "loss": 3.1674, "step": 21595 }, { "epoch": 1.69, "learning_rate": 8.306594526778013e-05, "loss": 2.6686, "step": 21596 }, { "epoch": 1.69, "learning_rate": 8.306516113855564e-05, "loss": 3.0729, "step": 21597 }, { "epoch": 1.69, "learning_rate": 8.306437700933115e-05, "loss": 2.9519, "step": 21598 }, { "epoch": 1.69, "learning_rate": 8.306359288010665e-05, "loss": 2.8591, "step": 21599 }, { "epoch": 1.69, "learning_rate": 8.306280875088215e-05, "loss": 3.1071, "step": 21600 }, { "epoch": 1.69, "learning_rate": 8.306202462165766e-05, "loss": 2.9918, "step": 21601 }, { "epoch": 1.69, "learning_rate": 8.306124049243316e-05, "loss": 2.7717, "step": 21602 }, { "epoch": 1.69, "learning_rate": 8.306045636320866e-05, "loss": 2.8088, "step": 21603 }, { "epoch": 1.69, "learning_rate": 8.305967223398417e-05, "loss": 2.9988, "step": 21604 }, { "epoch": 1.69, "learning_rate": 8.305888810475967e-05, "loss": 2.8154, "step": 21605 }, { "epoch": 1.69, "learning_rate": 8.305810397553518e-05, "loss": 2.8912, "step": 21606 }, { "epoch": 1.69, "learning_rate": 8.305731984631068e-05, "loss": 2.9064, "step": 21607 }, { "epoch": 1.69, "learning_rate": 8.305653571708618e-05, "loss": 2.9529, "step": 21608 }, { "epoch": 1.69, "learning_rate": 8.305575158786169e-05, "loss": 2.9016, "step": 21609 }, { "epoch": 1.69, "learning_rate": 8.305496745863719e-05, "loss": 3.0543, "step": 21610 }, { "epoch": 1.69, "learning_rate": 8.30541833294127e-05, "loss": 3.193, "step": 21611 }, { "epoch": 1.69, "learning_rate": 8.30533992001882e-05, "loss": 2.9119, "step": 21612 }, { "epoch": 1.69, "learning_rate": 8.30526150709637e-05, "loss": 2.6417, "step": 21613 }, { "epoch": 1.69, "learning_rate": 8.30518309417392e-05, "loss": 3.0745, "step": 21614 }, { "epoch": 1.69, "learning_rate": 8.305104681251471e-05, "loss": 2.8749, "step": 21615 }, { "epoch": 1.69, "learning_rate": 8.305026268329021e-05, "loss": 2.9374, "step": 21616 }, { "epoch": 1.7, "learning_rate": 8.304947855406572e-05, "loss": 3.0561, "step": 21617 }, { "epoch": 1.7, "learning_rate": 8.304869442484122e-05, "loss": 2.8832, "step": 21618 }, { "epoch": 1.7, "learning_rate": 8.304791029561672e-05, "loss": 2.8598, "step": 21619 }, { "epoch": 1.7, "learning_rate": 8.304712616639223e-05, "loss": 2.8751, "step": 21620 }, { "epoch": 1.7, "learning_rate": 8.304634203716773e-05, "loss": 2.9973, "step": 21621 }, { "epoch": 1.7, "learning_rate": 8.304555790794323e-05, "loss": 2.8804, "step": 21622 }, { "epoch": 1.7, "learning_rate": 8.304477377871874e-05, "loss": 3.007, "step": 21623 }, { "epoch": 1.7, "learning_rate": 8.304398964949424e-05, "loss": 2.6097, "step": 21624 }, { "epoch": 1.7, "learning_rate": 8.304320552026975e-05, "loss": 2.8789, "step": 21625 }, { "epoch": 1.7, "learning_rate": 8.304242139104525e-05, "loss": 2.9549, "step": 21626 }, { "epoch": 1.7, "learning_rate": 8.304163726182075e-05, "loss": 2.8639, "step": 21627 }, { "epoch": 1.7, "learning_rate": 8.304085313259626e-05, "loss": 2.8578, "step": 21628 }, { "epoch": 1.7, "learning_rate": 8.304006900337176e-05, "loss": 2.9359, "step": 21629 }, { "epoch": 1.7, "learning_rate": 8.303928487414726e-05, "loss": 2.974, "step": 21630 }, { "epoch": 1.7, "learning_rate": 8.303850074492277e-05, "loss": 2.9048, "step": 21631 }, { "epoch": 1.7, "learning_rate": 8.303771661569827e-05, "loss": 2.8422, "step": 21632 }, { "epoch": 1.7, "learning_rate": 8.303693248647377e-05, "loss": 2.8344, "step": 21633 }, { "epoch": 1.7, "learning_rate": 8.303614835724928e-05, "loss": 2.8794, "step": 21634 }, { "epoch": 1.7, "learning_rate": 8.303536422802478e-05, "loss": 2.8247, "step": 21635 }, { "epoch": 1.7, "learning_rate": 8.303458009880029e-05, "loss": 2.9072, "step": 21636 }, { "epoch": 1.7, "learning_rate": 8.303379596957579e-05, "loss": 3.1422, "step": 21637 }, { "epoch": 1.7, "learning_rate": 8.30330118403513e-05, "loss": 2.8359, "step": 21638 }, { "epoch": 1.7, "learning_rate": 8.30322277111268e-05, "loss": 2.7673, "step": 21639 }, { "epoch": 1.7, "learning_rate": 8.30314435819023e-05, "loss": 2.7398, "step": 21640 }, { "epoch": 1.7, "learning_rate": 8.30306594526778e-05, "loss": 2.9965, "step": 21641 }, { "epoch": 1.7, "learning_rate": 8.302987532345331e-05, "loss": 3.0485, "step": 21642 }, { "epoch": 1.7, "learning_rate": 8.302909119422881e-05, "loss": 3.088, "step": 21643 }, { "epoch": 1.7, "learning_rate": 8.302830706500432e-05, "loss": 2.8009, "step": 21644 }, { "epoch": 1.7, "learning_rate": 8.302752293577982e-05, "loss": 2.7489, "step": 21645 }, { "epoch": 1.7, "learning_rate": 8.302673880655532e-05, "loss": 2.9549, "step": 21646 }, { "epoch": 1.7, "learning_rate": 8.302595467733083e-05, "loss": 2.9081, "step": 21647 }, { "epoch": 1.7, "learning_rate": 8.302517054810633e-05, "loss": 2.8855, "step": 21648 }, { "epoch": 1.7, "learning_rate": 8.302438641888183e-05, "loss": 2.8487, "step": 21649 }, { "epoch": 1.7, "learning_rate": 8.302360228965734e-05, "loss": 3.1201, "step": 21650 }, { "epoch": 1.7, "learning_rate": 8.302281816043284e-05, "loss": 2.9537, "step": 21651 }, { "epoch": 1.7, "learning_rate": 8.302203403120834e-05, "loss": 3.0889, "step": 21652 }, { "epoch": 1.7, "learning_rate": 8.302124990198385e-05, "loss": 2.8899, "step": 21653 }, { "epoch": 1.7, "learning_rate": 8.302046577275937e-05, "loss": 2.7801, "step": 21654 }, { "epoch": 1.7, "learning_rate": 8.301968164353487e-05, "loss": 3.1773, "step": 21655 }, { "epoch": 1.7, "learning_rate": 8.301889751431037e-05, "loss": 3.0274, "step": 21656 }, { "epoch": 1.7, "learning_rate": 8.301811338508588e-05, "loss": 2.9387, "step": 21657 }, { "epoch": 1.7, "learning_rate": 8.301732925586137e-05, "loss": 2.7198, "step": 21658 }, { "epoch": 1.7, "learning_rate": 8.301654512663687e-05, "loss": 2.7434, "step": 21659 }, { "epoch": 1.7, "learning_rate": 8.301576099741237e-05, "loss": 2.545, "step": 21660 }, { "epoch": 1.7, "learning_rate": 8.301497686818788e-05, "loss": 2.8695, "step": 21661 }, { "epoch": 1.7, "learning_rate": 8.301419273896338e-05, "loss": 2.83, "step": 21662 }, { "epoch": 1.7, "learning_rate": 8.301340860973889e-05, "loss": 2.9592, "step": 21663 }, { "epoch": 1.7, "learning_rate": 8.301262448051439e-05, "loss": 2.9502, "step": 21664 }, { "epoch": 1.7, "learning_rate": 8.301184035128989e-05, "loss": 2.7304, "step": 21665 }, { "epoch": 1.7, "learning_rate": 8.30110562220654e-05, "loss": 2.9533, "step": 21666 }, { "epoch": 1.7, "learning_rate": 8.30102720928409e-05, "loss": 2.9469, "step": 21667 }, { "epoch": 1.7, "learning_rate": 8.30094879636164e-05, "loss": 2.7662, "step": 21668 }, { "epoch": 1.7, "learning_rate": 8.300870383439191e-05, "loss": 2.8895, "step": 21669 }, { "epoch": 1.7, "learning_rate": 8.300791970516741e-05, "loss": 2.7004, "step": 21670 }, { "epoch": 1.7, "learning_rate": 8.300713557594291e-05, "loss": 3.0523, "step": 21671 }, { "epoch": 1.7, "learning_rate": 8.300635144671842e-05, "loss": 3.1262, "step": 21672 }, { "epoch": 1.7, "learning_rate": 8.300556731749392e-05, "loss": 2.7567, "step": 21673 }, { "epoch": 1.7, "learning_rate": 8.300478318826943e-05, "loss": 3.1003, "step": 21674 }, { "epoch": 1.7, "learning_rate": 8.300399905904493e-05, "loss": 3.037, "step": 21675 }, { "epoch": 1.7, "learning_rate": 8.300321492982043e-05, "loss": 2.881, "step": 21676 }, { "epoch": 1.7, "learning_rate": 8.300243080059594e-05, "loss": 2.9275, "step": 21677 }, { "epoch": 1.7, "learning_rate": 8.300164667137144e-05, "loss": 2.8206, "step": 21678 }, { "epoch": 1.7, "learning_rate": 8.300086254214696e-05, "loss": 2.7506, "step": 21679 }, { "epoch": 1.7, "learning_rate": 8.300007841292246e-05, "loss": 3.0347, "step": 21680 }, { "epoch": 1.7, "learning_rate": 8.299929428369797e-05, "loss": 2.809, "step": 21681 }, { "epoch": 1.7, "learning_rate": 8.299851015447347e-05, "loss": 2.9506, "step": 21682 }, { "epoch": 1.7, "learning_rate": 8.299772602524897e-05, "loss": 3.0468, "step": 21683 }, { "epoch": 1.7, "learning_rate": 8.299694189602448e-05, "loss": 2.9284, "step": 21684 }, { "epoch": 1.7, "learning_rate": 8.299615776679998e-05, "loss": 2.8725, "step": 21685 }, { "epoch": 1.7, "learning_rate": 8.299537363757548e-05, "loss": 3.0769, "step": 21686 }, { "epoch": 1.7, "learning_rate": 8.299458950835099e-05, "loss": 3.0579, "step": 21687 }, { "epoch": 1.7, "learning_rate": 8.299380537912649e-05, "loss": 2.9666, "step": 21688 }, { "epoch": 1.7, "learning_rate": 8.2993021249902e-05, "loss": 3.3593, "step": 21689 }, { "epoch": 1.7, "learning_rate": 8.29922371206775e-05, "loss": 3.0784, "step": 21690 }, { "epoch": 1.7, "learning_rate": 8.2991452991453e-05, "loss": 3.029, "step": 21691 }, { "epoch": 1.7, "learning_rate": 8.29906688622285e-05, "loss": 3.0977, "step": 21692 }, { "epoch": 1.7, "learning_rate": 8.298988473300401e-05, "loss": 2.8267, "step": 21693 }, { "epoch": 1.7, "learning_rate": 8.29891006037795e-05, "loss": 2.7824, "step": 21694 }, { "epoch": 1.7, "learning_rate": 8.2988316474555e-05, "loss": 2.8783, "step": 21695 }, { "epoch": 1.7, "learning_rate": 8.29875323453305e-05, "loss": 2.9706, "step": 21696 }, { "epoch": 1.7, "learning_rate": 8.298674821610601e-05, "loss": 2.928, "step": 21697 }, { "epoch": 1.7, "learning_rate": 8.298596408688151e-05, "loss": 2.7938, "step": 21698 }, { "epoch": 1.7, "learning_rate": 8.298517995765702e-05, "loss": 2.8559, "step": 21699 }, { "epoch": 1.7, "learning_rate": 8.298439582843252e-05, "loss": 2.8534, "step": 21700 }, { "epoch": 1.7, "learning_rate": 8.298361169920803e-05, "loss": 3.0478, "step": 21701 }, { "epoch": 1.7, "learning_rate": 8.298282756998353e-05, "loss": 2.8831, "step": 21702 }, { "epoch": 1.7, "learning_rate": 8.298204344075903e-05, "loss": 3.1017, "step": 21703 }, { "epoch": 1.7, "learning_rate": 8.298125931153454e-05, "loss": 2.7776, "step": 21704 }, { "epoch": 1.7, "learning_rate": 8.298047518231005e-05, "loss": 2.8514, "step": 21705 }, { "epoch": 1.7, "learning_rate": 8.297969105308556e-05, "loss": 2.8434, "step": 21706 }, { "epoch": 1.7, "learning_rate": 8.297890692386106e-05, "loss": 3.0174, "step": 21707 }, { "epoch": 1.7, "learning_rate": 8.297812279463656e-05, "loss": 2.7189, "step": 21708 }, { "epoch": 1.7, "learning_rate": 8.297733866541207e-05, "loss": 2.7048, "step": 21709 }, { "epoch": 1.7, "learning_rate": 8.297655453618757e-05, "loss": 2.7906, "step": 21710 }, { "epoch": 1.7, "learning_rate": 8.297577040696308e-05, "loss": 2.9771, "step": 21711 }, { "epoch": 1.7, "learning_rate": 8.297498627773858e-05, "loss": 3.1603, "step": 21712 }, { "epoch": 1.7, "learning_rate": 8.297420214851408e-05, "loss": 2.898, "step": 21713 }, { "epoch": 1.7, "learning_rate": 8.297341801928959e-05, "loss": 2.7196, "step": 21714 }, { "epoch": 1.7, "learning_rate": 8.297263389006509e-05, "loss": 2.6486, "step": 21715 }, { "epoch": 1.7, "learning_rate": 8.29718497608406e-05, "loss": 2.9502, "step": 21716 }, { "epoch": 1.7, "learning_rate": 8.29710656316161e-05, "loss": 2.823, "step": 21717 }, { "epoch": 1.7, "learning_rate": 8.29702815023916e-05, "loss": 2.5804, "step": 21718 }, { "epoch": 1.7, "learning_rate": 8.29694973731671e-05, "loss": 3.0238, "step": 21719 }, { "epoch": 1.7, "learning_rate": 8.296871324394261e-05, "loss": 3.0269, "step": 21720 }, { "epoch": 1.7, "learning_rate": 8.296792911471811e-05, "loss": 3.2086, "step": 21721 }, { "epoch": 1.7, "learning_rate": 8.296714498549362e-05, "loss": 3.0667, "step": 21722 }, { "epoch": 1.7, "learning_rate": 8.296636085626912e-05, "loss": 3.0282, "step": 21723 }, { "epoch": 1.7, "learning_rate": 8.296557672704462e-05, "loss": 2.8008, "step": 21724 }, { "epoch": 1.7, "learning_rate": 8.296479259782013e-05, "loss": 2.8676, "step": 21725 }, { "epoch": 1.7, "learning_rate": 8.296400846859563e-05, "loss": 2.6893, "step": 21726 }, { "epoch": 1.7, "learning_rate": 8.296322433937113e-05, "loss": 3.122, "step": 21727 }, { "epoch": 1.7, "learning_rate": 8.296244021014664e-05, "loss": 2.9903, "step": 21728 }, { "epoch": 1.7, "learning_rate": 8.296165608092214e-05, "loss": 2.7807, "step": 21729 }, { "epoch": 1.7, "learning_rate": 8.296087195169765e-05, "loss": 2.877, "step": 21730 }, { "epoch": 1.7, "learning_rate": 8.296008782247315e-05, "loss": 2.9359, "step": 21731 }, { "epoch": 1.7, "learning_rate": 8.295930369324865e-05, "loss": 2.8427, "step": 21732 }, { "epoch": 1.7, "learning_rate": 8.295851956402416e-05, "loss": 3.0886, "step": 21733 }, { "epoch": 1.7, "learning_rate": 8.295773543479966e-05, "loss": 2.9712, "step": 21734 }, { "epoch": 1.7, "learning_rate": 8.295695130557516e-05, "loss": 3.0778, "step": 21735 }, { "epoch": 1.7, "learning_rate": 8.295616717635067e-05, "loss": 2.8034, "step": 21736 }, { "epoch": 1.7, "learning_rate": 8.295538304712617e-05, "loss": 3.08, "step": 21737 }, { "epoch": 1.7, "learning_rate": 8.295459891790167e-05, "loss": 3.024, "step": 21738 }, { "epoch": 1.7, "learning_rate": 8.295381478867718e-05, "loss": 3.0372, "step": 21739 }, { "epoch": 1.7, "learning_rate": 8.295303065945268e-05, "loss": 3.0053, "step": 21740 }, { "epoch": 1.7, "learning_rate": 8.295224653022819e-05, "loss": 2.8974, "step": 21741 }, { "epoch": 1.7, "learning_rate": 8.295146240100369e-05, "loss": 3.0472, "step": 21742 }, { "epoch": 1.7, "learning_rate": 8.295067827177919e-05, "loss": 2.8901, "step": 21743 }, { "epoch": 1.71, "learning_rate": 8.29498941425547e-05, "loss": 3.1743, "step": 21744 }, { "epoch": 1.71, "learning_rate": 8.29491100133302e-05, "loss": 3.1089, "step": 21745 }, { "epoch": 1.71, "learning_rate": 8.29483258841057e-05, "loss": 2.7172, "step": 21746 }, { "epoch": 1.71, "learning_rate": 8.294754175488121e-05, "loss": 2.7106, "step": 21747 }, { "epoch": 1.71, "learning_rate": 8.294675762565671e-05, "loss": 2.8775, "step": 21748 }, { "epoch": 1.71, "learning_rate": 8.294597349643222e-05, "loss": 3.0121, "step": 21749 }, { "epoch": 1.71, "learning_rate": 8.294518936720772e-05, "loss": 2.7251, "step": 21750 }, { "epoch": 1.71, "learning_rate": 8.294440523798322e-05, "loss": 3.0248, "step": 21751 }, { "epoch": 1.71, "learning_rate": 8.294362110875873e-05, "loss": 2.7936, "step": 21752 }, { "epoch": 1.71, "learning_rate": 8.294283697953423e-05, "loss": 2.8902, "step": 21753 }, { "epoch": 1.71, "learning_rate": 8.294205285030973e-05, "loss": 2.8533, "step": 21754 }, { "epoch": 1.71, "learning_rate": 8.294126872108524e-05, "loss": 2.8374, "step": 21755 }, { "epoch": 1.71, "learning_rate": 8.294048459186074e-05, "loss": 2.8073, "step": 21756 }, { "epoch": 1.71, "learning_rate": 8.293970046263624e-05, "loss": 2.8545, "step": 21757 }, { "epoch": 1.71, "learning_rate": 8.293891633341175e-05, "loss": 2.7331, "step": 21758 }, { "epoch": 1.71, "learning_rate": 8.293813220418725e-05, "loss": 3.139, "step": 21759 }, { "epoch": 1.71, "learning_rate": 8.293734807496276e-05, "loss": 2.9781, "step": 21760 }, { "epoch": 1.71, "learning_rate": 8.293656394573826e-05, "loss": 2.9561, "step": 21761 }, { "epoch": 1.71, "learning_rate": 8.293577981651378e-05, "loss": 3.0255, "step": 21762 }, { "epoch": 1.71, "learning_rate": 8.293499568728928e-05, "loss": 2.8001, "step": 21763 }, { "epoch": 1.71, "learning_rate": 8.293421155806478e-05, "loss": 3.0382, "step": 21764 }, { "epoch": 1.71, "learning_rate": 8.293342742884029e-05, "loss": 2.9719, "step": 21765 }, { "epoch": 1.71, "learning_rate": 8.293264329961578e-05, "loss": 2.9548, "step": 21766 }, { "epoch": 1.71, "learning_rate": 8.293185917039128e-05, "loss": 2.8438, "step": 21767 }, { "epoch": 1.71, "learning_rate": 8.293107504116678e-05, "loss": 3.0629, "step": 21768 }, { "epoch": 1.71, "learning_rate": 8.293029091194229e-05, "loss": 2.9829, "step": 21769 }, { "epoch": 1.71, "learning_rate": 8.292950678271779e-05, "loss": 3.0924, "step": 21770 }, { "epoch": 1.71, "learning_rate": 8.29287226534933e-05, "loss": 2.7372, "step": 21771 }, { "epoch": 1.71, "learning_rate": 8.29279385242688e-05, "loss": 2.8978, "step": 21772 }, { "epoch": 1.71, "learning_rate": 8.29271543950443e-05, "loss": 2.9666, "step": 21773 }, { "epoch": 1.71, "learning_rate": 8.292637026581981e-05, "loss": 2.8641, "step": 21774 }, { "epoch": 1.71, "learning_rate": 8.292558613659531e-05, "loss": 2.8007, "step": 21775 }, { "epoch": 1.71, "learning_rate": 8.292480200737081e-05, "loss": 3.0731, "step": 21776 }, { "epoch": 1.71, "learning_rate": 8.292401787814632e-05, "loss": 2.6715, "step": 21777 }, { "epoch": 1.71, "learning_rate": 8.292323374892182e-05, "loss": 2.9475, "step": 21778 }, { "epoch": 1.71, "learning_rate": 8.292244961969733e-05, "loss": 2.9338, "step": 21779 }, { "epoch": 1.71, "learning_rate": 8.292166549047283e-05, "loss": 3.0504, "step": 21780 }, { "epoch": 1.71, "learning_rate": 8.292088136124833e-05, "loss": 3.1052, "step": 21781 }, { "epoch": 1.71, "learning_rate": 8.292009723202384e-05, "loss": 3.2004, "step": 21782 }, { "epoch": 1.71, "learning_rate": 8.291931310279934e-05, "loss": 2.9527, "step": 21783 }, { "epoch": 1.71, "learning_rate": 8.291852897357484e-05, "loss": 2.9484, "step": 21784 }, { "epoch": 1.71, "learning_rate": 8.291774484435035e-05, "loss": 2.9841, "step": 21785 }, { "epoch": 1.71, "learning_rate": 8.291696071512585e-05, "loss": 2.8751, "step": 21786 }, { "epoch": 1.71, "learning_rate": 8.291617658590137e-05, "loss": 3.062, "step": 21787 }, { "epoch": 1.71, "learning_rate": 8.291539245667687e-05, "loss": 2.9627, "step": 21788 }, { "epoch": 1.71, "learning_rate": 8.291460832745238e-05, "loss": 3.1156, "step": 21789 }, { "epoch": 1.71, "learning_rate": 8.291382419822788e-05, "loss": 2.9042, "step": 21790 }, { "epoch": 1.71, "learning_rate": 8.291304006900338e-05, "loss": 3.0615, "step": 21791 }, { "epoch": 1.71, "learning_rate": 8.291225593977889e-05, "loss": 2.9781, "step": 21792 }, { "epoch": 1.71, "learning_rate": 8.291147181055439e-05, "loss": 2.9098, "step": 21793 }, { "epoch": 1.71, "learning_rate": 8.29106876813299e-05, "loss": 2.902, "step": 21794 }, { "epoch": 1.71, "learning_rate": 8.29099035521054e-05, "loss": 2.8057, "step": 21795 }, { "epoch": 1.71, "learning_rate": 8.29091194228809e-05, "loss": 2.9835, "step": 21796 }, { "epoch": 1.71, "learning_rate": 8.29083352936564e-05, "loss": 2.9692, "step": 21797 }, { "epoch": 1.71, "learning_rate": 8.290755116443191e-05, "loss": 2.9878, "step": 21798 }, { "epoch": 1.71, "learning_rate": 8.290676703520741e-05, "loss": 3.0189, "step": 21799 }, { "epoch": 1.71, "learning_rate": 8.290598290598292e-05, "loss": 2.7105, "step": 21800 }, { "epoch": 1.71, "learning_rate": 8.290519877675842e-05, "loss": 2.8756, "step": 21801 }, { "epoch": 1.71, "learning_rate": 8.290441464753391e-05, "loss": 2.9394, "step": 21802 }, { "epoch": 1.71, "learning_rate": 8.290363051830941e-05, "loss": 2.8213, "step": 21803 }, { "epoch": 1.71, "learning_rate": 8.290284638908492e-05, "loss": 2.7546, "step": 21804 }, { "epoch": 1.71, "learning_rate": 8.290206225986042e-05, "loss": 2.6544, "step": 21805 }, { "epoch": 1.71, "learning_rate": 8.290127813063592e-05, "loss": 3.1272, "step": 21806 }, { "epoch": 1.71, "learning_rate": 8.290049400141143e-05, "loss": 2.7932, "step": 21807 }, { "epoch": 1.71, "learning_rate": 8.289970987218693e-05, "loss": 2.8712, "step": 21808 }, { "epoch": 1.71, "learning_rate": 8.289892574296244e-05, "loss": 3.0076, "step": 21809 }, { "epoch": 1.71, "learning_rate": 8.289814161373794e-05, "loss": 2.9006, "step": 21810 }, { "epoch": 1.71, "learning_rate": 8.289735748451344e-05, "loss": 3.1488, "step": 21811 }, { "epoch": 1.71, "learning_rate": 8.289657335528895e-05, "loss": 2.6847, "step": 21812 }, { "epoch": 1.71, "learning_rate": 8.289578922606446e-05, "loss": 2.9591, "step": 21813 }, { "epoch": 1.71, "learning_rate": 8.289500509683997e-05, "loss": 2.6872, "step": 21814 }, { "epoch": 1.71, "learning_rate": 8.289422096761547e-05, "loss": 3.0336, "step": 21815 }, { "epoch": 1.71, "learning_rate": 8.289343683839098e-05, "loss": 2.8578, "step": 21816 }, { "epoch": 1.71, "learning_rate": 8.289265270916648e-05, "loss": 2.9213, "step": 21817 }, { "epoch": 1.71, "learning_rate": 8.289186857994198e-05, "loss": 2.811, "step": 21818 }, { "epoch": 1.71, "learning_rate": 8.289108445071749e-05, "loss": 2.8641, "step": 21819 }, { "epoch": 1.71, "learning_rate": 8.289030032149299e-05, "loss": 2.9778, "step": 21820 }, { "epoch": 1.71, "learning_rate": 8.28895161922685e-05, "loss": 2.6992, "step": 21821 }, { "epoch": 1.71, "learning_rate": 8.2888732063044e-05, "loss": 2.9206, "step": 21822 }, { "epoch": 1.71, "learning_rate": 8.28879479338195e-05, "loss": 3.1001, "step": 21823 }, { "epoch": 1.71, "learning_rate": 8.2887163804595e-05, "loss": 3.0475, "step": 21824 }, { "epoch": 1.71, "learning_rate": 8.288637967537051e-05, "loss": 2.9307, "step": 21825 }, { "epoch": 1.71, "learning_rate": 8.288559554614601e-05, "loss": 2.7791, "step": 21826 }, { "epoch": 1.71, "learning_rate": 8.288481141692152e-05, "loss": 2.8436, "step": 21827 }, { "epoch": 1.71, "learning_rate": 8.288402728769702e-05, "loss": 3.0176, "step": 21828 }, { "epoch": 1.71, "learning_rate": 8.288324315847252e-05, "loss": 3.0787, "step": 21829 }, { "epoch": 1.71, "learning_rate": 8.288245902924803e-05, "loss": 2.7768, "step": 21830 }, { "epoch": 1.71, "learning_rate": 8.288167490002353e-05, "loss": 2.7662, "step": 21831 }, { "epoch": 1.71, "learning_rate": 8.288089077079903e-05, "loss": 2.9781, "step": 21832 }, { "epoch": 1.71, "learning_rate": 8.288010664157454e-05, "loss": 3.0914, "step": 21833 }, { "epoch": 1.71, "learning_rate": 8.287932251235004e-05, "loss": 2.7861, "step": 21834 }, { "epoch": 1.71, "learning_rate": 8.287853838312554e-05, "loss": 2.9607, "step": 21835 }, { "epoch": 1.71, "learning_rate": 8.287775425390105e-05, "loss": 2.8089, "step": 21836 }, { "epoch": 1.71, "learning_rate": 8.287697012467655e-05, "loss": 2.8091, "step": 21837 }, { "epoch": 1.71, "learning_rate": 8.287618599545206e-05, "loss": 3.0051, "step": 21838 }, { "epoch": 1.71, "learning_rate": 8.287540186622756e-05, "loss": 2.6653, "step": 21839 }, { "epoch": 1.71, "learning_rate": 8.287461773700306e-05, "loss": 2.7456, "step": 21840 }, { "epoch": 1.71, "learning_rate": 8.287383360777857e-05, "loss": 3.0256, "step": 21841 }, { "epoch": 1.71, "learning_rate": 8.287304947855407e-05, "loss": 2.867, "step": 21842 }, { "epoch": 1.71, "learning_rate": 8.287226534932957e-05, "loss": 2.965, "step": 21843 }, { "epoch": 1.71, "learning_rate": 8.287148122010508e-05, "loss": 2.9186, "step": 21844 }, { "epoch": 1.71, "learning_rate": 8.287069709088058e-05, "loss": 3.0097, "step": 21845 }, { "epoch": 1.71, "learning_rate": 8.286991296165609e-05, "loss": 2.8636, "step": 21846 }, { "epoch": 1.71, "learning_rate": 8.286912883243159e-05, "loss": 3.1337, "step": 21847 }, { "epoch": 1.71, "learning_rate": 8.286834470320709e-05, "loss": 2.8903, "step": 21848 }, { "epoch": 1.71, "learning_rate": 8.28675605739826e-05, "loss": 2.9224, "step": 21849 }, { "epoch": 1.71, "learning_rate": 8.28667764447581e-05, "loss": 2.935, "step": 21850 }, { "epoch": 1.71, "learning_rate": 8.28659923155336e-05, "loss": 2.7803, "step": 21851 }, { "epoch": 1.71, "learning_rate": 8.286520818630911e-05, "loss": 3.0294, "step": 21852 }, { "epoch": 1.71, "learning_rate": 8.286442405708461e-05, "loss": 2.8844, "step": 21853 }, { "epoch": 1.71, "learning_rate": 8.286363992786011e-05, "loss": 2.6651, "step": 21854 }, { "epoch": 1.71, "learning_rate": 8.286285579863562e-05, "loss": 2.8559, "step": 21855 }, { "epoch": 1.71, "learning_rate": 8.286207166941112e-05, "loss": 3.0639, "step": 21856 }, { "epoch": 1.71, "learning_rate": 8.286128754018663e-05, "loss": 2.9661, "step": 21857 }, { "epoch": 1.71, "learning_rate": 8.286050341096213e-05, "loss": 3.3525, "step": 21858 }, { "epoch": 1.71, "learning_rate": 8.285971928173763e-05, "loss": 2.8693, "step": 21859 }, { "epoch": 1.71, "learning_rate": 8.285893515251314e-05, "loss": 2.6354, "step": 21860 }, { "epoch": 1.71, "learning_rate": 8.285815102328864e-05, "loss": 2.7586, "step": 21861 }, { "epoch": 1.71, "learning_rate": 8.285736689406414e-05, "loss": 3.0122, "step": 21862 }, { "epoch": 1.71, "learning_rate": 8.285658276483965e-05, "loss": 2.8181, "step": 21863 }, { "epoch": 1.71, "learning_rate": 8.285579863561515e-05, "loss": 2.7995, "step": 21864 }, { "epoch": 1.71, "learning_rate": 8.285501450639066e-05, "loss": 2.8293, "step": 21865 }, { "epoch": 1.71, "learning_rate": 8.285423037716616e-05, "loss": 2.9242, "step": 21866 }, { "epoch": 1.71, "learning_rate": 8.285344624794166e-05, "loss": 2.7431, "step": 21867 }, { "epoch": 1.71, "learning_rate": 8.285266211871717e-05, "loss": 2.8685, "step": 21868 }, { "epoch": 1.71, "learning_rate": 8.285187798949267e-05, "loss": 2.6372, "step": 21869 }, { "epoch": 1.71, "learning_rate": 8.285109386026819e-05, "loss": 2.7738, "step": 21870 }, { "epoch": 1.71, "learning_rate": 8.285030973104369e-05, "loss": 3.1285, "step": 21871 }, { "epoch": 1.72, "learning_rate": 8.28495256018192e-05, "loss": 2.9771, "step": 21872 }, { "epoch": 1.72, "learning_rate": 8.28487414725947e-05, "loss": 2.7973, "step": 21873 }, { "epoch": 1.72, "learning_rate": 8.284795734337019e-05, "loss": 2.9291, "step": 21874 }, { "epoch": 1.72, "learning_rate": 8.284717321414569e-05, "loss": 2.9845, "step": 21875 }, { "epoch": 1.72, "learning_rate": 8.28463890849212e-05, "loss": 2.8928, "step": 21876 }, { "epoch": 1.72, "learning_rate": 8.28456049556967e-05, "loss": 2.8164, "step": 21877 }, { "epoch": 1.72, "learning_rate": 8.28448208264722e-05, "loss": 2.8033, "step": 21878 }, { "epoch": 1.72, "learning_rate": 8.28440366972477e-05, "loss": 2.8313, "step": 21879 }, { "epoch": 1.72, "learning_rate": 8.284325256802321e-05, "loss": 2.7907, "step": 21880 }, { "epoch": 1.72, "learning_rate": 8.284246843879871e-05, "loss": 2.8309, "step": 21881 }, { "epoch": 1.72, "learning_rate": 8.284168430957422e-05, "loss": 3.1295, "step": 21882 }, { "epoch": 1.72, "learning_rate": 8.284090018034972e-05, "loss": 2.9941, "step": 21883 }, { "epoch": 1.72, "learning_rate": 8.284011605112523e-05, "loss": 2.8141, "step": 21884 }, { "epoch": 1.72, "learning_rate": 8.283933192190073e-05, "loss": 2.8158, "step": 21885 }, { "epoch": 1.72, "learning_rate": 8.283854779267623e-05, "loss": 2.6819, "step": 21886 }, { "epoch": 1.72, "learning_rate": 8.283776366345174e-05, "loss": 3.016, "step": 21887 }, { "epoch": 1.72, "learning_rate": 8.283697953422724e-05, "loss": 2.6911, "step": 21888 }, { "epoch": 1.72, "learning_rate": 8.283619540500274e-05, "loss": 2.7829, "step": 21889 }, { "epoch": 1.72, "learning_rate": 8.283541127577825e-05, "loss": 2.8788, "step": 21890 }, { "epoch": 1.72, "learning_rate": 8.283462714655375e-05, "loss": 2.9631, "step": 21891 }, { "epoch": 1.72, "learning_rate": 8.283384301732925e-05, "loss": 2.8552, "step": 21892 }, { "epoch": 1.72, "learning_rate": 8.283305888810476e-05, "loss": 2.7582, "step": 21893 }, { "epoch": 1.72, "learning_rate": 8.283227475888026e-05, "loss": 2.8466, "step": 21894 }, { "epoch": 1.72, "learning_rate": 8.283149062965578e-05, "loss": 3.118, "step": 21895 }, { "epoch": 1.72, "learning_rate": 8.283070650043128e-05, "loss": 3.0561, "step": 21896 }, { "epoch": 1.72, "learning_rate": 8.282992237120679e-05, "loss": 2.7586, "step": 21897 }, { "epoch": 1.72, "learning_rate": 8.282913824198229e-05, "loss": 2.8249, "step": 21898 }, { "epoch": 1.72, "learning_rate": 8.28283541127578e-05, "loss": 3.1555, "step": 21899 }, { "epoch": 1.72, "learning_rate": 8.28275699835333e-05, "loss": 2.8069, "step": 21900 }, { "epoch": 1.72, "learning_rate": 8.28267858543088e-05, "loss": 2.8396, "step": 21901 }, { "epoch": 1.72, "learning_rate": 8.28260017250843e-05, "loss": 2.8984, "step": 21902 }, { "epoch": 1.72, "learning_rate": 8.282521759585981e-05, "loss": 2.9642, "step": 21903 }, { "epoch": 1.72, "learning_rate": 8.282443346663531e-05, "loss": 3.0064, "step": 21904 }, { "epoch": 1.72, "learning_rate": 8.282364933741082e-05, "loss": 2.6255, "step": 21905 }, { "epoch": 1.72, "learning_rate": 8.282286520818632e-05, "loss": 2.8343, "step": 21906 }, { "epoch": 1.72, "learning_rate": 8.282208107896182e-05, "loss": 3.1453, "step": 21907 }, { "epoch": 1.72, "learning_rate": 8.282129694973733e-05, "loss": 3.1026, "step": 21908 }, { "epoch": 1.72, "learning_rate": 8.282051282051283e-05, "loss": 3.075, "step": 21909 }, { "epoch": 1.72, "learning_rate": 8.281972869128832e-05, "loss": 2.8568, "step": 21910 }, { "epoch": 1.72, "learning_rate": 8.281894456206382e-05, "loss": 2.9448, "step": 21911 }, { "epoch": 1.72, "learning_rate": 8.281816043283933e-05, "loss": 2.9956, "step": 21912 }, { "epoch": 1.72, "learning_rate": 8.281737630361483e-05, "loss": 3.0212, "step": 21913 }, { "epoch": 1.72, "learning_rate": 8.281659217439034e-05, "loss": 3.0093, "step": 21914 }, { "epoch": 1.72, "learning_rate": 8.281580804516584e-05, "loss": 2.6766, "step": 21915 }, { "epoch": 1.72, "learning_rate": 8.281502391594134e-05, "loss": 2.852, "step": 21916 }, { "epoch": 1.72, "learning_rate": 8.281423978671685e-05, "loss": 2.9307, "step": 21917 }, { "epoch": 1.72, "learning_rate": 8.281345565749235e-05, "loss": 3.0325, "step": 21918 }, { "epoch": 1.72, "learning_rate": 8.281267152826785e-05, "loss": 2.7705, "step": 21919 }, { "epoch": 1.72, "learning_rate": 8.281188739904336e-05, "loss": 2.9851, "step": 21920 }, { "epoch": 1.72, "learning_rate": 8.281110326981887e-05, "loss": 2.8903, "step": 21921 }, { "epoch": 1.72, "learning_rate": 8.281031914059438e-05, "loss": 2.8121, "step": 21922 }, { "epoch": 1.72, "learning_rate": 8.280953501136988e-05, "loss": 3.0348, "step": 21923 }, { "epoch": 1.72, "learning_rate": 8.280875088214539e-05, "loss": 2.9003, "step": 21924 }, { "epoch": 1.72, "learning_rate": 8.280796675292089e-05, "loss": 3.0557, "step": 21925 }, { "epoch": 1.72, "learning_rate": 8.280718262369639e-05, "loss": 2.7468, "step": 21926 }, { "epoch": 1.72, "learning_rate": 8.28063984944719e-05, "loss": 3.1233, "step": 21927 }, { "epoch": 1.72, "learning_rate": 8.28056143652474e-05, "loss": 2.7694, "step": 21928 }, { "epoch": 1.72, "learning_rate": 8.28048302360229e-05, "loss": 2.8795, "step": 21929 }, { "epoch": 1.72, "learning_rate": 8.280404610679841e-05, "loss": 2.8619, "step": 21930 }, { "epoch": 1.72, "learning_rate": 8.280326197757391e-05, "loss": 2.8532, "step": 21931 }, { "epoch": 1.72, "learning_rate": 8.280247784834942e-05, "loss": 2.9423, "step": 21932 }, { "epoch": 1.72, "learning_rate": 8.280169371912492e-05, "loss": 2.9901, "step": 21933 }, { "epoch": 1.72, "learning_rate": 8.280090958990042e-05, "loss": 2.9379, "step": 21934 }, { "epoch": 1.72, "learning_rate": 8.280012546067593e-05, "loss": 3.0538, "step": 21935 }, { "epoch": 1.72, "learning_rate": 8.279934133145143e-05, "loss": 3.0276, "step": 21936 }, { "epoch": 1.72, "learning_rate": 8.279855720222693e-05, "loss": 2.8423, "step": 21937 }, { "epoch": 1.72, "learning_rate": 8.279777307300244e-05, "loss": 2.7646, "step": 21938 }, { "epoch": 1.72, "learning_rate": 8.279698894377794e-05, "loss": 2.9501, "step": 21939 }, { "epoch": 1.72, "learning_rate": 8.279620481455344e-05, "loss": 2.8117, "step": 21940 }, { "epoch": 1.72, "learning_rate": 8.279542068532895e-05, "loss": 2.8919, "step": 21941 }, { "epoch": 1.72, "learning_rate": 8.279463655610445e-05, "loss": 2.648, "step": 21942 }, { "epoch": 1.72, "learning_rate": 8.279385242687996e-05, "loss": 2.9002, "step": 21943 }, { "epoch": 1.72, "learning_rate": 8.279306829765546e-05, "loss": 3.0007, "step": 21944 }, { "epoch": 1.72, "learning_rate": 8.279228416843096e-05, "loss": 2.6849, "step": 21945 }, { "epoch": 1.72, "learning_rate": 8.279150003920645e-05, "loss": 3.1117, "step": 21946 }, { "epoch": 1.72, "learning_rate": 8.279071590998197e-05, "loss": 2.8497, "step": 21947 }, { "epoch": 1.72, "learning_rate": 8.278993178075747e-05, "loss": 2.9179, "step": 21948 }, { "epoch": 1.72, "learning_rate": 8.278914765153298e-05, "loss": 2.9355, "step": 21949 }, { "epoch": 1.72, "learning_rate": 8.278836352230848e-05, "loss": 3.1548, "step": 21950 }, { "epoch": 1.72, "learning_rate": 8.278757939308399e-05, "loss": 2.9501, "step": 21951 }, { "epoch": 1.72, "learning_rate": 8.278679526385949e-05, "loss": 2.744, "step": 21952 }, { "epoch": 1.72, "learning_rate": 8.278601113463499e-05, "loss": 3.0678, "step": 21953 }, { "epoch": 1.72, "learning_rate": 8.27852270054105e-05, "loss": 2.7227, "step": 21954 }, { "epoch": 1.72, "learning_rate": 8.2784442876186e-05, "loss": 3.0169, "step": 21955 }, { "epoch": 1.72, "learning_rate": 8.27836587469615e-05, "loss": 2.8412, "step": 21956 }, { "epoch": 1.72, "learning_rate": 8.278287461773701e-05, "loss": 2.7957, "step": 21957 }, { "epoch": 1.72, "learning_rate": 8.278209048851251e-05, "loss": 3.0204, "step": 21958 }, { "epoch": 1.72, "learning_rate": 8.278130635928801e-05, "loss": 3.0894, "step": 21959 }, { "epoch": 1.72, "learning_rate": 8.278052223006352e-05, "loss": 2.861, "step": 21960 }, { "epoch": 1.72, "learning_rate": 8.277973810083902e-05, "loss": 2.894, "step": 21961 }, { "epoch": 1.72, "learning_rate": 8.277895397161453e-05, "loss": 2.7729, "step": 21962 }, { "epoch": 1.72, "learning_rate": 8.277816984239003e-05, "loss": 3.0249, "step": 21963 }, { "epoch": 1.72, "learning_rate": 8.277738571316553e-05, "loss": 2.8205, "step": 21964 }, { "epoch": 1.72, "learning_rate": 8.277660158394104e-05, "loss": 2.9679, "step": 21965 }, { "epoch": 1.72, "learning_rate": 8.277581745471654e-05, "loss": 2.9954, "step": 21966 }, { "epoch": 1.72, "learning_rate": 8.277503332549204e-05, "loss": 3.0663, "step": 21967 }, { "epoch": 1.72, "learning_rate": 8.277424919626755e-05, "loss": 2.6975, "step": 21968 }, { "epoch": 1.72, "learning_rate": 8.277346506704305e-05, "loss": 2.904, "step": 21969 }, { "epoch": 1.72, "learning_rate": 8.277268093781855e-05, "loss": 3.1573, "step": 21970 }, { "epoch": 1.72, "learning_rate": 8.277189680859406e-05, "loss": 2.7906, "step": 21971 }, { "epoch": 1.72, "learning_rate": 8.277111267936956e-05, "loss": 2.8005, "step": 21972 }, { "epoch": 1.72, "learning_rate": 8.277032855014507e-05, "loss": 2.7443, "step": 21973 }, { "epoch": 1.72, "learning_rate": 8.276954442092057e-05, "loss": 2.8621, "step": 21974 }, { "epoch": 1.72, "learning_rate": 8.276876029169607e-05, "loss": 2.8003, "step": 21975 }, { "epoch": 1.72, "learning_rate": 8.276797616247158e-05, "loss": 2.8425, "step": 21976 }, { "epoch": 1.72, "learning_rate": 8.276719203324708e-05, "loss": 2.9026, "step": 21977 }, { "epoch": 1.72, "learning_rate": 8.27664079040226e-05, "loss": 2.8233, "step": 21978 }, { "epoch": 1.72, "learning_rate": 8.27656237747981e-05, "loss": 2.9348, "step": 21979 }, { "epoch": 1.72, "learning_rate": 8.27648396455736e-05, "loss": 2.9857, "step": 21980 }, { "epoch": 1.72, "learning_rate": 8.276405551634911e-05, "loss": 2.977, "step": 21981 }, { "epoch": 1.72, "learning_rate": 8.27632713871246e-05, "loss": 3.1233, "step": 21982 }, { "epoch": 1.72, "learning_rate": 8.27624872579001e-05, "loss": 2.868, "step": 21983 }, { "epoch": 1.72, "learning_rate": 8.27617031286756e-05, "loss": 3.0065, "step": 21984 }, { "epoch": 1.72, "learning_rate": 8.276091899945111e-05, "loss": 2.8282, "step": 21985 }, { "epoch": 1.72, "learning_rate": 8.276013487022661e-05, "loss": 2.6686, "step": 21986 }, { "epoch": 1.72, "learning_rate": 8.275935074100212e-05, "loss": 2.663, "step": 21987 }, { "epoch": 1.72, "learning_rate": 8.275856661177762e-05, "loss": 3.0054, "step": 21988 }, { "epoch": 1.72, "learning_rate": 8.275778248255312e-05, "loss": 2.9026, "step": 21989 }, { "epoch": 1.72, "learning_rate": 8.275699835332863e-05, "loss": 2.73, "step": 21990 }, { "epoch": 1.72, "learning_rate": 8.275621422410413e-05, "loss": 2.9211, "step": 21991 }, { "epoch": 1.72, "learning_rate": 8.275543009487964e-05, "loss": 2.9562, "step": 21992 }, { "epoch": 1.72, "learning_rate": 8.275464596565514e-05, "loss": 2.9479, "step": 21993 }, { "epoch": 1.72, "learning_rate": 8.275386183643064e-05, "loss": 2.9745, "step": 21994 }, { "epoch": 1.72, "learning_rate": 8.275307770720615e-05, "loss": 2.9143, "step": 21995 }, { "epoch": 1.72, "learning_rate": 8.275229357798165e-05, "loss": 3.0937, "step": 21996 }, { "epoch": 1.72, "learning_rate": 8.275150944875715e-05, "loss": 2.8848, "step": 21997 }, { "epoch": 1.72, "learning_rate": 8.275072531953266e-05, "loss": 3.0367, "step": 21998 }, { "epoch": 1.73, "learning_rate": 8.274994119030816e-05, "loss": 2.8642, "step": 21999 }, { "epoch": 1.73, "learning_rate": 8.274915706108367e-05, "loss": 2.7488, "step": 22000 }, { "epoch": 1.73, "eval_loss": 2.9070091247558594, "eval_runtime": 218.0202, "eval_samples_per_second": 51.977, "eval_steps_per_second": 3.252, "step": 22000 }, { "epoch": 1.73, "learning_rate": 8.274837293185917e-05, "loss": 2.8268, "step": 22001 }, { "epoch": 1.73, "learning_rate": 8.274758880263467e-05, "loss": 2.9195, "step": 22002 }, { "epoch": 1.73, "learning_rate": 8.274680467341019e-05, "loss": 2.7161, "step": 22003 }, { "epoch": 1.73, "learning_rate": 8.27460205441857e-05, "loss": 2.946, "step": 22004 }, { "epoch": 1.73, "learning_rate": 8.27452364149612e-05, "loss": 2.7956, "step": 22005 }, { "epoch": 1.73, "learning_rate": 8.27444522857367e-05, "loss": 2.9648, "step": 22006 }, { "epoch": 1.73, "learning_rate": 8.27436681565122e-05, "loss": 2.9476, "step": 22007 }, { "epoch": 1.73, "learning_rate": 8.274288402728771e-05, "loss": 2.9786, "step": 22008 }, { "epoch": 1.73, "learning_rate": 8.274209989806321e-05, "loss": 2.887, "step": 22009 }, { "epoch": 1.73, "learning_rate": 8.274131576883872e-05, "loss": 3.1642, "step": 22010 }, { "epoch": 1.73, "learning_rate": 8.274053163961422e-05, "loss": 2.6396, "step": 22011 }, { "epoch": 1.73, "learning_rate": 8.273974751038972e-05, "loss": 2.7796, "step": 22012 }, { "epoch": 1.73, "learning_rate": 8.273896338116523e-05, "loss": 2.9741, "step": 22013 }, { "epoch": 1.73, "learning_rate": 8.273817925194073e-05, "loss": 2.9861, "step": 22014 }, { "epoch": 1.73, "learning_rate": 8.273739512271623e-05, "loss": 2.8725, "step": 22015 }, { "epoch": 1.73, "learning_rate": 8.273661099349174e-05, "loss": 2.8335, "step": 22016 }, { "epoch": 1.73, "learning_rate": 8.273582686426724e-05, "loss": 2.7633, "step": 22017 }, { "epoch": 1.73, "learning_rate": 8.273504273504273e-05, "loss": 2.8572, "step": 22018 }, { "epoch": 1.73, "learning_rate": 8.273425860581824e-05, "loss": 2.9056, "step": 22019 }, { "epoch": 1.73, "learning_rate": 8.273347447659374e-05, "loss": 2.9258, "step": 22020 }, { "epoch": 1.73, "learning_rate": 8.273269034736924e-05, "loss": 2.9384, "step": 22021 }, { "epoch": 1.73, "learning_rate": 8.273190621814475e-05, "loss": 2.8993, "step": 22022 }, { "epoch": 1.73, "learning_rate": 8.273112208892025e-05, "loss": 3.0839, "step": 22023 }, { "epoch": 1.73, "learning_rate": 8.273033795969575e-05, "loss": 2.896, "step": 22024 }, { "epoch": 1.73, "learning_rate": 8.272955383047126e-05, "loss": 2.8829, "step": 22025 }, { "epoch": 1.73, "learning_rate": 8.272876970124676e-05, "loss": 2.9617, "step": 22026 }, { "epoch": 1.73, "learning_rate": 8.272798557202226e-05, "loss": 2.8545, "step": 22027 }, { "epoch": 1.73, "learning_rate": 8.272720144279777e-05, "loss": 2.7771, "step": 22028 }, { "epoch": 1.73, "learning_rate": 8.272641731357329e-05, "loss": 2.8602, "step": 22029 }, { "epoch": 1.73, "learning_rate": 8.272563318434879e-05, "loss": 2.9213, "step": 22030 }, { "epoch": 1.73, "learning_rate": 8.272484905512429e-05, "loss": 2.9925, "step": 22031 }, { "epoch": 1.73, "learning_rate": 8.27240649258998e-05, "loss": 2.8769, "step": 22032 }, { "epoch": 1.73, "learning_rate": 8.27232807966753e-05, "loss": 2.9007, "step": 22033 }, { "epoch": 1.73, "learning_rate": 8.27224966674508e-05, "loss": 2.8297, "step": 22034 }, { "epoch": 1.73, "learning_rate": 8.272171253822631e-05, "loss": 2.9216, "step": 22035 }, { "epoch": 1.73, "learning_rate": 8.272092840900181e-05, "loss": 2.9193, "step": 22036 }, { "epoch": 1.73, "learning_rate": 8.272014427977731e-05, "loss": 2.7632, "step": 22037 }, { "epoch": 1.73, "learning_rate": 8.271936015055282e-05, "loss": 2.8298, "step": 22038 }, { "epoch": 1.73, "learning_rate": 8.271857602132832e-05, "loss": 3.0029, "step": 22039 }, { "epoch": 1.73, "learning_rate": 8.271779189210383e-05, "loss": 3.1047, "step": 22040 }, { "epoch": 1.73, "learning_rate": 8.271700776287933e-05, "loss": 2.8669, "step": 22041 }, { "epoch": 1.73, "learning_rate": 8.271622363365483e-05, "loss": 2.7615, "step": 22042 }, { "epoch": 1.73, "learning_rate": 8.271543950443034e-05, "loss": 3.3151, "step": 22043 }, { "epoch": 1.73, "learning_rate": 8.271465537520584e-05, "loss": 2.7291, "step": 22044 }, { "epoch": 1.73, "learning_rate": 8.271387124598134e-05, "loss": 2.7996, "step": 22045 }, { "epoch": 1.73, "learning_rate": 8.271308711675685e-05, "loss": 2.7916, "step": 22046 }, { "epoch": 1.73, "learning_rate": 8.271230298753235e-05, "loss": 3.1085, "step": 22047 }, { "epoch": 1.73, "learning_rate": 8.271151885830786e-05, "loss": 2.8829, "step": 22048 }, { "epoch": 1.73, "learning_rate": 8.271073472908336e-05, "loss": 2.8851, "step": 22049 }, { "epoch": 1.73, "learning_rate": 8.270995059985886e-05, "loss": 2.6288, "step": 22050 }, { "epoch": 1.73, "learning_rate": 8.270916647063437e-05, "loss": 2.9541, "step": 22051 }, { "epoch": 1.73, "learning_rate": 8.270838234140987e-05, "loss": 2.9439, "step": 22052 }, { "epoch": 1.73, "learning_rate": 8.270759821218537e-05, "loss": 3.049, "step": 22053 }, { "epoch": 1.73, "learning_rate": 8.270681408296086e-05, "loss": 2.9987, "step": 22054 }, { "epoch": 1.73, "learning_rate": 8.270602995373638e-05, "loss": 3.0448, "step": 22055 }, { "epoch": 1.73, "learning_rate": 8.270524582451188e-05, "loss": 2.778, "step": 22056 }, { "epoch": 1.73, "learning_rate": 8.270446169528739e-05, "loss": 2.9982, "step": 22057 }, { "epoch": 1.73, "learning_rate": 8.270367756606289e-05, "loss": 2.9363, "step": 22058 }, { "epoch": 1.73, "learning_rate": 8.27028934368384e-05, "loss": 2.9178, "step": 22059 }, { "epoch": 1.73, "learning_rate": 8.27021093076139e-05, "loss": 2.816, "step": 22060 }, { "epoch": 1.73, "learning_rate": 8.27013251783894e-05, "loss": 2.9502, "step": 22061 }, { "epoch": 1.73, "learning_rate": 8.270054104916491e-05, "loss": 2.7332, "step": 22062 }, { "epoch": 1.73, "learning_rate": 8.269975691994041e-05, "loss": 2.9262, "step": 22063 }, { "epoch": 1.73, "learning_rate": 8.269897279071591e-05, "loss": 2.6209, "step": 22064 }, { "epoch": 1.73, "learning_rate": 8.269818866149142e-05, "loss": 2.7938, "step": 22065 }, { "epoch": 1.73, "learning_rate": 8.269740453226692e-05, "loss": 2.7663, "step": 22066 }, { "epoch": 1.73, "learning_rate": 8.269662040304243e-05, "loss": 3.1207, "step": 22067 }, { "epoch": 1.73, "learning_rate": 8.269583627381793e-05, "loss": 3.0358, "step": 22068 }, { "epoch": 1.73, "learning_rate": 8.269505214459343e-05, "loss": 2.79, "step": 22069 }, { "epoch": 1.73, "learning_rate": 8.269426801536894e-05, "loss": 3.0506, "step": 22070 }, { "epoch": 1.73, "learning_rate": 8.269348388614444e-05, "loss": 2.9357, "step": 22071 }, { "epoch": 1.73, "learning_rate": 8.269269975691994e-05, "loss": 2.7171, "step": 22072 }, { "epoch": 1.73, "learning_rate": 8.269191562769545e-05, "loss": 2.7163, "step": 22073 }, { "epoch": 1.73, "learning_rate": 8.269113149847095e-05, "loss": 2.7013, "step": 22074 }, { "epoch": 1.73, "learning_rate": 8.269034736924645e-05, "loss": 3.1055, "step": 22075 }, { "epoch": 1.73, "learning_rate": 8.268956324002196e-05, "loss": 3.0685, "step": 22076 }, { "epoch": 1.73, "learning_rate": 8.268877911079746e-05, "loss": 2.8494, "step": 22077 }, { "epoch": 1.73, "learning_rate": 8.268799498157297e-05, "loss": 3.1027, "step": 22078 }, { "epoch": 1.73, "learning_rate": 8.268721085234847e-05, "loss": 2.891, "step": 22079 }, { "epoch": 1.73, "learning_rate": 8.268642672312397e-05, "loss": 2.8501, "step": 22080 }, { "epoch": 1.73, "learning_rate": 8.268564259389948e-05, "loss": 2.762, "step": 22081 }, { "epoch": 1.73, "learning_rate": 8.268485846467498e-05, "loss": 2.9583, "step": 22082 }, { "epoch": 1.73, "learning_rate": 8.268407433545048e-05, "loss": 2.9025, "step": 22083 }, { "epoch": 1.73, "learning_rate": 8.268329020622599e-05, "loss": 3.0393, "step": 22084 }, { "epoch": 1.73, "learning_rate": 8.268250607700149e-05, "loss": 2.8449, "step": 22085 }, { "epoch": 1.73, "learning_rate": 8.268172194777701e-05, "loss": 2.8059, "step": 22086 }, { "epoch": 1.73, "learning_rate": 8.268093781855251e-05, "loss": 2.6622, "step": 22087 }, { "epoch": 1.73, "learning_rate": 8.268015368932802e-05, "loss": 3.1105, "step": 22088 }, { "epoch": 1.73, "learning_rate": 8.267936956010352e-05, "loss": 2.8236, "step": 22089 }, { "epoch": 1.73, "learning_rate": 8.267858543087901e-05, "loss": 3.1902, "step": 22090 }, { "epoch": 1.73, "learning_rate": 8.267780130165451e-05, "loss": 2.8808, "step": 22091 }, { "epoch": 1.73, "learning_rate": 8.267701717243002e-05, "loss": 2.7252, "step": 22092 }, { "epoch": 1.73, "learning_rate": 8.267623304320552e-05, "loss": 2.8817, "step": 22093 }, { "epoch": 1.73, "learning_rate": 8.267544891398102e-05, "loss": 3.08, "step": 22094 }, { "epoch": 1.73, "learning_rate": 8.267466478475653e-05, "loss": 2.9375, "step": 22095 }, { "epoch": 1.73, "learning_rate": 8.267388065553203e-05, "loss": 3.0079, "step": 22096 }, { "epoch": 1.73, "learning_rate": 8.267309652630754e-05, "loss": 2.8984, "step": 22097 }, { "epoch": 1.73, "learning_rate": 8.267231239708304e-05, "loss": 2.8731, "step": 22098 }, { "epoch": 1.73, "learning_rate": 8.267152826785854e-05, "loss": 2.9295, "step": 22099 }, { "epoch": 1.73, "learning_rate": 8.267074413863405e-05, "loss": 2.9521, "step": 22100 }, { "epoch": 1.73, "learning_rate": 8.266996000940955e-05, "loss": 2.8054, "step": 22101 }, { "epoch": 1.73, "learning_rate": 8.266917588018505e-05, "loss": 2.9613, "step": 22102 }, { "epoch": 1.73, "learning_rate": 8.266839175096056e-05, "loss": 2.8875, "step": 22103 }, { "epoch": 1.73, "learning_rate": 8.266760762173606e-05, "loss": 3.0136, "step": 22104 }, { "epoch": 1.73, "learning_rate": 8.266682349251157e-05, "loss": 2.8436, "step": 22105 }, { "epoch": 1.73, "learning_rate": 8.266603936328707e-05, "loss": 2.7427, "step": 22106 }, { "epoch": 1.73, "learning_rate": 8.266525523406257e-05, "loss": 2.866, "step": 22107 }, { "epoch": 1.73, "learning_rate": 8.266447110483808e-05, "loss": 2.7562, "step": 22108 }, { "epoch": 1.73, "learning_rate": 8.266368697561358e-05, "loss": 3.0219, "step": 22109 }, { "epoch": 1.73, "learning_rate": 8.266290284638908e-05, "loss": 2.9297, "step": 22110 }, { "epoch": 1.73, "learning_rate": 8.26621187171646e-05, "loss": 2.6814, "step": 22111 }, { "epoch": 1.73, "learning_rate": 8.26613345879401e-05, "loss": 2.8682, "step": 22112 }, { "epoch": 1.73, "learning_rate": 8.266055045871561e-05, "loss": 2.9317, "step": 22113 }, { "epoch": 1.73, "learning_rate": 8.265976632949111e-05, "loss": 2.7344, "step": 22114 }, { "epoch": 1.73, "learning_rate": 8.265898220026662e-05, "loss": 3.0057, "step": 22115 }, { "epoch": 1.73, "learning_rate": 8.265819807104212e-05, "loss": 2.7484, "step": 22116 }, { "epoch": 1.73, "learning_rate": 8.265741394181762e-05, "loss": 2.9569, "step": 22117 }, { "epoch": 1.73, "learning_rate": 8.265662981259313e-05, "loss": 2.8183, "step": 22118 }, { "epoch": 1.73, "learning_rate": 8.265584568336863e-05, "loss": 2.9692, "step": 22119 }, { "epoch": 1.73, "learning_rate": 8.265506155414413e-05, "loss": 2.99, "step": 22120 }, { "epoch": 1.73, "learning_rate": 8.265427742491964e-05, "loss": 2.9225, "step": 22121 }, { "epoch": 1.73, "learning_rate": 8.265349329569514e-05, "loss": 2.8236, "step": 22122 }, { "epoch": 1.73, "learning_rate": 8.265270916647064e-05, "loss": 2.8843, "step": 22123 }, { "epoch": 1.73, "learning_rate": 8.265192503724615e-05, "loss": 2.933, "step": 22124 }, { "epoch": 1.73, "learning_rate": 8.265114090802165e-05, "loss": 3.01, "step": 22125 }, { "epoch": 1.73, "learning_rate": 8.265035677879714e-05, "loss": 2.774, "step": 22126 }, { "epoch": 1.74, "learning_rate": 8.264957264957265e-05, "loss": 3.1386, "step": 22127 }, { "epoch": 1.74, "learning_rate": 8.264878852034815e-05, "loss": 2.8944, "step": 22128 }, { "epoch": 1.74, "learning_rate": 8.264800439112365e-05, "loss": 3.0642, "step": 22129 }, { "epoch": 1.74, "learning_rate": 8.264722026189916e-05, "loss": 2.782, "step": 22130 }, { "epoch": 1.74, "learning_rate": 8.264643613267466e-05, "loss": 2.8157, "step": 22131 }, { "epoch": 1.74, "learning_rate": 8.264565200345016e-05, "loss": 2.8348, "step": 22132 }, { "epoch": 1.74, "learning_rate": 8.264486787422567e-05, "loss": 3.0105, "step": 22133 }, { "epoch": 1.74, "learning_rate": 8.264408374500117e-05, "loss": 2.7345, "step": 22134 }, { "epoch": 1.74, "learning_rate": 8.264329961577668e-05, "loss": 2.7659, "step": 22135 }, { "epoch": 1.74, "learning_rate": 8.264251548655218e-05, "loss": 2.9134, "step": 22136 }, { "epoch": 1.74, "learning_rate": 8.26417313573277e-05, "loss": 3.0202, "step": 22137 }, { "epoch": 1.74, "learning_rate": 8.26409472281032e-05, "loss": 2.8411, "step": 22138 }, { "epoch": 1.74, "learning_rate": 8.26401630988787e-05, "loss": 2.9041, "step": 22139 }, { "epoch": 1.74, "learning_rate": 8.263937896965421e-05, "loss": 2.7509, "step": 22140 }, { "epoch": 1.74, "learning_rate": 8.263859484042971e-05, "loss": 2.8113, "step": 22141 }, { "epoch": 1.74, "learning_rate": 8.263781071120521e-05, "loss": 2.7668, "step": 22142 }, { "epoch": 1.74, "learning_rate": 8.263702658198072e-05, "loss": 2.6447, "step": 22143 }, { "epoch": 1.74, "learning_rate": 8.263624245275622e-05, "loss": 2.7298, "step": 22144 }, { "epoch": 1.74, "learning_rate": 8.263545832353173e-05, "loss": 2.8502, "step": 22145 }, { "epoch": 1.74, "learning_rate": 8.263467419430723e-05, "loss": 2.6798, "step": 22146 }, { "epoch": 1.74, "learning_rate": 8.263389006508273e-05, "loss": 3.0791, "step": 22147 }, { "epoch": 1.74, "learning_rate": 8.263310593585824e-05, "loss": 2.887, "step": 22148 }, { "epoch": 1.74, "learning_rate": 8.263232180663374e-05, "loss": 2.7633, "step": 22149 }, { "epoch": 1.74, "learning_rate": 8.263153767740924e-05, "loss": 2.8668, "step": 22150 }, { "epoch": 1.74, "learning_rate": 8.263075354818475e-05, "loss": 3.0021, "step": 22151 }, { "epoch": 1.74, "learning_rate": 8.262996941896025e-05, "loss": 2.8496, "step": 22152 }, { "epoch": 1.74, "learning_rate": 8.262918528973576e-05, "loss": 2.8054, "step": 22153 }, { "epoch": 1.74, "learning_rate": 8.262840116051126e-05, "loss": 2.7618, "step": 22154 }, { "epoch": 1.74, "learning_rate": 8.262761703128676e-05, "loss": 3.0218, "step": 22155 }, { "epoch": 1.74, "learning_rate": 8.262683290206227e-05, "loss": 3.0307, "step": 22156 }, { "epoch": 1.74, "learning_rate": 8.262604877283777e-05, "loss": 2.9315, "step": 22157 }, { "epoch": 1.74, "learning_rate": 8.262526464361327e-05, "loss": 2.8452, "step": 22158 }, { "epoch": 1.74, "learning_rate": 8.262448051438878e-05, "loss": 3.0216, "step": 22159 }, { "epoch": 1.74, "learning_rate": 8.262369638516428e-05, "loss": 2.8721, "step": 22160 }, { "epoch": 1.74, "learning_rate": 8.262291225593978e-05, "loss": 2.8522, "step": 22161 }, { "epoch": 1.74, "learning_rate": 8.262212812671527e-05, "loss": 2.8838, "step": 22162 }, { "epoch": 1.74, "learning_rate": 8.262134399749079e-05, "loss": 2.849, "step": 22163 }, { "epoch": 1.74, "learning_rate": 8.26205598682663e-05, "loss": 3.053, "step": 22164 }, { "epoch": 1.74, "learning_rate": 8.26197757390418e-05, "loss": 2.8654, "step": 22165 }, { "epoch": 1.74, "learning_rate": 8.26189916098173e-05, "loss": 2.8866, "step": 22166 }, { "epoch": 1.74, "learning_rate": 8.26182074805928e-05, "loss": 3.0416, "step": 22167 }, { "epoch": 1.74, "learning_rate": 8.261742335136831e-05, "loss": 2.9742, "step": 22168 }, { "epoch": 1.74, "learning_rate": 8.261663922214381e-05, "loss": 2.819, "step": 22169 }, { "epoch": 1.74, "learning_rate": 8.261585509291932e-05, "loss": 2.9182, "step": 22170 }, { "epoch": 1.74, "learning_rate": 8.261507096369482e-05, "loss": 2.5795, "step": 22171 }, { "epoch": 1.74, "learning_rate": 8.261428683447032e-05, "loss": 2.803, "step": 22172 }, { "epoch": 1.74, "learning_rate": 8.261350270524583e-05, "loss": 2.9025, "step": 22173 }, { "epoch": 1.74, "learning_rate": 8.261271857602133e-05, "loss": 3.0434, "step": 22174 }, { "epoch": 1.74, "learning_rate": 8.261193444679684e-05, "loss": 2.7602, "step": 22175 }, { "epoch": 1.74, "learning_rate": 8.261115031757234e-05, "loss": 2.9368, "step": 22176 }, { "epoch": 1.74, "learning_rate": 8.261036618834784e-05, "loss": 2.8868, "step": 22177 }, { "epoch": 1.74, "learning_rate": 8.260958205912335e-05, "loss": 2.9513, "step": 22178 }, { "epoch": 1.74, "learning_rate": 8.260879792989885e-05, "loss": 3.0082, "step": 22179 }, { "epoch": 1.74, "learning_rate": 8.260801380067435e-05, "loss": 2.6629, "step": 22180 }, { "epoch": 1.74, "learning_rate": 8.260722967144986e-05, "loss": 2.9387, "step": 22181 }, { "epoch": 1.74, "learning_rate": 8.260644554222536e-05, "loss": 3.0566, "step": 22182 }, { "epoch": 1.74, "learning_rate": 8.260566141300087e-05, "loss": 2.984, "step": 22183 }, { "epoch": 1.74, "learning_rate": 8.260487728377637e-05, "loss": 2.8271, "step": 22184 }, { "epoch": 1.74, "learning_rate": 8.260409315455187e-05, "loss": 2.8682, "step": 22185 }, { "epoch": 1.74, "learning_rate": 8.260330902532738e-05, "loss": 2.8202, "step": 22186 }, { "epoch": 1.74, "learning_rate": 8.260252489610288e-05, "loss": 3.1921, "step": 22187 }, { "epoch": 1.74, "learning_rate": 8.260174076687838e-05, "loss": 3.0384, "step": 22188 }, { "epoch": 1.74, "learning_rate": 8.260095663765389e-05, "loss": 2.7622, "step": 22189 }, { "epoch": 1.74, "learning_rate": 8.260017250842939e-05, "loss": 2.8955, "step": 22190 }, { "epoch": 1.74, "learning_rate": 8.25993883792049e-05, "loss": 2.8025, "step": 22191 }, { "epoch": 1.74, "learning_rate": 8.25986042499804e-05, "loss": 2.9813, "step": 22192 }, { "epoch": 1.74, "learning_rate": 8.25978201207559e-05, "loss": 2.7561, "step": 22193 }, { "epoch": 1.74, "learning_rate": 8.259703599153142e-05, "loss": 2.7954, "step": 22194 }, { "epoch": 1.74, "learning_rate": 8.259625186230692e-05, "loss": 2.8292, "step": 22195 }, { "epoch": 1.74, "learning_rate": 8.259546773308243e-05, "loss": 2.8878, "step": 22196 }, { "epoch": 1.74, "learning_rate": 8.259468360385793e-05, "loss": 2.7314, "step": 22197 }, { "epoch": 1.74, "learning_rate": 8.259389947463342e-05, "loss": 3.1591, "step": 22198 }, { "epoch": 1.74, "learning_rate": 8.259311534540892e-05, "loss": 2.8747, "step": 22199 }, { "epoch": 1.74, "learning_rate": 8.259233121618443e-05, "loss": 2.8297, "step": 22200 }, { "epoch": 1.74, "learning_rate": 8.259154708695993e-05, "loss": 2.7203, "step": 22201 }, { "epoch": 1.74, "learning_rate": 8.259076295773544e-05, "loss": 3.0582, "step": 22202 }, { "epoch": 1.74, "learning_rate": 8.258997882851094e-05, "loss": 2.8256, "step": 22203 }, { "epoch": 1.74, "learning_rate": 8.258919469928644e-05, "loss": 2.9516, "step": 22204 }, { "epoch": 1.74, "learning_rate": 8.258841057006195e-05, "loss": 2.7969, "step": 22205 }, { "epoch": 1.74, "learning_rate": 8.258762644083745e-05, "loss": 3.0758, "step": 22206 }, { "epoch": 1.74, "learning_rate": 8.258684231161295e-05, "loss": 3.0493, "step": 22207 }, { "epoch": 1.74, "learning_rate": 8.258605818238846e-05, "loss": 2.9639, "step": 22208 }, { "epoch": 1.74, "learning_rate": 8.258527405316396e-05, "loss": 3.0836, "step": 22209 }, { "epoch": 1.74, "learning_rate": 8.258448992393946e-05, "loss": 2.9529, "step": 22210 }, { "epoch": 1.74, "learning_rate": 8.258370579471497e-05, "loss": 2.9969, "step": 22211 }, { "epoch": 1.74, "learning_rate": 8.258292166549047e-05, "loss": 2.7561, "step": 22212 }, { "epoch": 1.74, "learning_rate": 8.258213753626598e-05, "loss": 2.817, "step": 22213 }, { "epoch": 1.74, "learning_rate": 8.258135340704148e-05, "loss": 2.7669, "step": 22214 }, { "epoch": 1.74, "learning_rate": 8.258056927781698e-05, "loss": 3.0575, "step": 22215 }, { "epoch": 1.74, "learning_rate": 8.257978514859249e-05, "loss": 2.9299, "step": 22216 }, { "epoch": 1.74, "learning_rate": 8.257900101936799e-05, "loss": 2.8973, "step": 22217 }, { "epoch": 1.74, "learning_rate": 8.25782168901435e-05, "loss": 2.8983, "step": 22218 }, { "epoch": 1.74, "learning_rate": 8.2577432760919e-05, "loss": 2.8834, "step": 22219 }, { "epoch": 1.74, "learning_rate": 8.257664863169452e-05, "loss": 2.7108, "step": 22220 }, { "epoch": 1.74, "learning_rate": 8.257586450247002e-05, "loss": 3.0048, "step": 22221 }, { "epoch": 1.74, "learning_rate": 8.257508037324552e-05, "loss": 3.1666, "step": 22222 }, { "epoch": 1.74, "learning_rate": 8.257429624402103e-05, "loss": 2.8257, "step": 22223 }, { "epoch": 1.74, "learning_rate": 8.257351211479653e-05, "loss": 2.6694, "step": 22224 }, { "epoch": 1.74, "learning_rate": 8.257272798557203e-05, "loss": 2.859, "step": 22225 }, { "epoch": 1.74, "learning_rate": 8.257194385634754e-05, "loss": 2.8258, "step": 22226 }, { "epoch": 1.74, "learning_rate": 8.257115972712304e-05, "loss": 2.8809, "step": 22227 }, { "epoch": 1.74, "learning_rate": 8.257037559789854e-05, "loss": 2.7951, "step": 22228 }, { "epoch": 1.74, "learning_rate": 8.256959146867405e-05, "loss": 2.8646, "step": 22229 }, { "epoch": 1.74, "learning_rate": 8.256880733944955e-05, "loss": 3.0265, "step": 22230 }, { "epoch": 1.74, "learning_rate": 8.256802321022506e-05, "loss": 2.8253, "step": 22231 }, { "epoch": 1.74, "learning_rate": 8.256723908100056e-05, "loss": 2.7721, "step": 22232 }, { "epoch": 1.74, "learning_rate": 8.256645495177606e-05, "loss": 2.6109, "step": 22233 }, { "epoch": 1.74, "learning_rate": 8.256567082255155e-05, "loss": 2.8059, "step": 22234 }, { "epoch": 1.74, "learning_rate": 8.256488669332706e-05, "loss": 2.9455, "step": 22235 }, { "epoch": 1.74, "learning_rate": 8.256410256410256e-05, "loss": 2.8434, "step": 22236 }, { "epoch": 1.74, "learning_rate": 8.256331843487806e-05, "loss": 2.8601, "step": 22237 }, { "epoch": 1.74, "learning_rate": 8.256253430565357e-05, "loss": 2.8156, "step": 22238 }, { "epoch": 1.74, "learning_rate": 8.256175017642907e-05, "loss": 3.1084, "step": 22239 }, { "epoch": 1.74, "learning_rate": 8.256096604720458e-05, "loss": 2.7945, "step": 22240 }, { "epoch": 1.74, "learning_rate": 8.256018191798008e-05, "loss": 2.9774, "step": 22241 }, { "epoch": 1.74, "learning_rate": 8.255939778875558e-05, "loss": 2.9784, "step": 22242 }, { "epoch": 1.74, "learning_rate": 8.255861365953109e-05, "loss": 2.9861, "step": 22243 }, { "epoch": 1.74, "learning_rate": 8.255782953030659e-05, "loss": 3.0903, "step": 22244 }, { "epoch": 1.74, "learning_rate": 8.255704540108211e-05, "loss": 2.7836, "step": 22245 }, { "epoch": 1.74, "learning_rate": 8.255626127185761e-05, "loss": 2.9129, "step": 22246 }, { "epoch": 1.74, "learning_rate": 8.255547714263311e-05, "loss": 2.9657, "step": 22247 }, { "epoch": 1.74, "learning_rate": 8.255469301340862e-05, "loss": 2.906, "step": 22248 }, { "epoch": 1.74, "learning_rate": 8.255390888418412e-05, "loss": 2.99, "step": 22249 }, { "epoch": 1.74, "learning_rate": 8.255312475495963e-05, "loss": 2.9962, "step": 22250 }, { "epoch": 1.74, "learning_rate": 8.255234062573513e-05, "loss": 2.9283, "step": 22251 }, { "epoch": 1.74, "learning_rate": 8.255155649651063e-05, "loss": 3.2289, "step": 22252 }, { "epoch": 1.74, "learning_rate": 8.255077236728614e-05, "loss": 2.8849, "step": 22253 }, { "epoch": 1.75, "learning_rate": 8.254998823806164e-05, "loss": 2.8568, "step": 22254 }, { "epoch": 1.75, "learning_rate": 8.254920410883714e-05, "loss": 2.6695, "step": 22255 }, { "epoch": 1.75, "learning_rate": 8.254841997961265e-05, "loss": 2.9974, "step": 22256 }, { "epoch": 1.75, "learning_rate": 8.254763585038815e-05, "loss": 2.7589, "step": 22257 }, { "epoch": 1.75, "learning_rate": 8.254685172116365e-05, "loss": 2.9598, "step": 22258 }, { "epoch": 1.75, "learning_rate": 8.254606759193916e-05, "loss": 2.8829, "step": 22259 }, { "epoch": 1.75, "learning_rate": 8.254528346271466e-05, "loss": 3.2633, "step": 22260 }, { "epoch": 1.75, "learning_rate": 8.254449933349017e-05, "loss": 2.9196, "step": 22261 }, { "epoch": 1.75, "learning_rate": 8.254371520426567e-05, "loss": 3.1058, "step": 22262 }, { "epoch": 1.75, "learning_rate": 8.254293107504117e-05, "loss": 2.9058, "step": 22263 }, { "epoch": 1.75, "learning_rate": 8.254214694581668e-05, "loss": 2.752, "step": 22264 }, { "epoch": 1.75, "learning_rate": 8.254136281659218e-05, "loss": 3.1222, "step": 22265 }, { "epoch": 1.75, "learning_rate": 8.254057868736768e-05, "loss": 2.6971, "step": 22266 }, { "epoch": 1.75, "learning_rate": 8.253979455814319e-05, "loss": 2.9655, "step": 22267 }, { "epoch": 1.75, "learning_rate": 8.253901042891869e-05, "loss": 2.8242, "step": 22268 }, { "epoch": 1.75, "learning_rate": 8.25382262996942e-05, "loss": 3.0524, "step": 22269 }, { "epoch": 1.75, "learning_rate": 8.253744217046969e-05, "loss": 2.65, "step": 22270 }, { "epoch": 1.75, "learning_rate": 8.25366580412452e-05, "loss": 2.8844, "step": 22271 }, { "epoch": 1.75, "learning_rate": 8.25358739120207e-05, "loss": 2.9974, "step": 22272 }, { "epoch": 1.75, "learning_rate": 8.253508978279621e-05, "loss": 2.8559, "step": 22273 }, { "epoch": 1.75, "learning_rate": 8.253430565357171e-05, "loss": 2.6733, "step": 22274 }, { "epoch": 1.75, "learning_rate": 8.253352152434722e-05, "loss": 2.7855, "step": 22275 }, { "epoch": 1.75, "learning_rate": 8.253273739512272e-05, "loss": 2.7265, "step": 22276 }, { "epoch": 1.75, "learning_rate": 8.253195326589822e-05, "loss": 2.8986, "step": 22277 }, { "epoch": 1.75, "learning_rate": 8.253116913667373e-05, "loss": 3.0054, "step": 22278 }, { "epoch": 1.75, "learning_rate": 8.253038500744923e-05, "loss": 2.9958, "step": 22279 }, { "epoch": 1.75, "learning_rate": 8.252960087822474e-05, "loss": 2.7296, "step": 22280 }, { "epoch": 1.75, "learning_rate": 8.252881674900024e-05, "loss": 2.7921, "step": 22281 }, { "epoch": 1.75, "learning_rate": 8.252803261977574e-05, "loss": 3.0252, "step": 22282 }, { "epoch": 1.75, "learning_rate": 8.252724849055125e-05, "loss": 2.8332, "step": 22283 }, { "epoch": 1.75, "learning_rate": 8.252646436132675e-05, "loss": 2.8291, "step": 22284 }, { "epoch": 1.75, "learning_rate": 8.252568023210225e-05, "loss": 2.693, "step": 22285 }, { "epoch": 1.75, "learning_rate": 8.252489610287776e-05, "loss": 2.9424, "step": 22286 }, { "epoch": 1.75, "learning_rate": 8.252411197365326e-05, "loss": 2.8709, "step": 22287 }, { "epoch": 1.75, "learning_rate": 8.252332784442877e-05, "loss": 2.9844, "step": 22288 }, { "epoch": 1.75, "learning_rate": 8.252254371520427e-05, "loss": 2.8583, "step": 22289 }, { "epoch": 1.75, "learning_rate": 8.252175958597977e-05, "loss": 2.996, "step": 22290 }, { "epoch": 1.75, "learning_rate": 8.252097545675528e-05, "loss": 2.8106, "step": 22291 }, { "epoch": 1.75, "learning_rate": 8.252019132753078e-05, "loss": 2.8889, "step": 22292 }, { "epoch": 1.75, "learning_rate": 8.251940719830628e-05, "loss": 2.9314, "step": 22293 }, { "epoch": 1.75, "learning_rate": 8.251862306908179e-05, "loss": 3.0907, "step": 22294 }, { "epoch": 1.75, "learning_rate": 8.251783893985729e-05, "loss": 2.7777, "step": 22295 }, { "epoch": 1.75, "learning_rate": 8.25170548106328e-05, "loss": 3.1071, "step": 22296 }, { "epoch": 1.75, "learning_rate": 8.25162706814083e-05, "loss": 2.8895, "step": 22297 }, { "epoch": 1.75, "learning_rate": 8.25154865521838e-05, "loss": 2.7628, "step": 22298 }, { "epoch": 1.75, "learning_rate": 8.25147024229593e-05, "loss": 2.8723, "step": 22299 }, { "epoch": 1.75, "learning_rate": 8.251391829373481e-05, "loss": 2.95, "step": 22300 }, { "epoch": 1.75, "learning_rate": 8.251313416451031e-05, "loss": 2.9776, "step": 22301 }, { "epoch": 1.75, "learning_rate": 8.251235003528583e-05, "loss": 2.9232, "step": 22302 }, { "epoch": 1.75, "learning_rate": 8.251156590606133e-05, "loss": 3.0107, "step": 22303 }, { "epoch": 1.75, "learning_rate": 8.251078177683684e-05, "loss": 2.8454, "step": 22304 }, { "epoch": 1.75, "learning_rate": 8.250999764761234e-05, "loss": 2.9749, "step": 22305 }, { "epoch": 1.75, "learning_rate": 8.250921351838783e-05, "loss": 2.8515, "step": 22306 }, { "epoch": 1.75, "learning_rate": 8.250842938916334e-05, "loss": 2.746, "step": 22307 }, { "epoch": 1.75, "learning_rate": 8.250764525993884e-05, "loss": 2.9712, "step": 22308 }, { "epoch": 1.75, "learning_rate": 8.250686113071434e-05, "loss": 2.7485, "step": 22309 }, { "epoch": 1.75, "learning_rate": 8.250607700148985e-05, "loss": 2.9306, "step": 22310 }, { "epoch": 1.75, "learning_rate": 8.250529287226535e-05, "loss": 3.0642, "step": 22311 }, { "epoch": 1.75, "learning_rate": 8.250450874304085e-05, "loss": 2.8194, "step": 22312 }, { "epoch": 1.75, "learning_rate": 8.250372461381636e-05, "loss": 3.0207, "step": 22313 }, { "epoch": 1.75, "learning_rate": 8.250294048459186e-05, "loss": 2.8668, "step": 22314 }, { "epoch": 1.75, "learning_rate": 8.250215635536736e-05, "loss": 3.2441, "step": 22315 }, { "epoch": 1.75, "learning_rate": 8.250137222614287e-05, "loss": 3.0571, "step": 22316 }, { "epoch": 1.75, "learning_rate": 8.250058809691837e-05, "loss": 3.1325, "step": 22317 }, { "epoch": 1.75, "learning_rate": 8.249980396769388e-05, "loss": 3.2116, "step": 22318 }, { "epoch": 1.75, "learning_rate": 8.249901983846938e-05, "loss": 2.9411, "step": 22319 }, { "epoch": 1.75, "learning_rate": 8.249823570924488e-05, "loss": 2.8715, "step": 22320 }, { "epoch": 1.75, "learning_rate": 8.249745158002039e-05, "loss": 3.0652, "step": 22321 }, { "epoch": 1.75, "learning_rate": 8.249666745079589e-05, "loss": 2.8812, "step": 22322 }, { "epoch": 1.75, "learning_rate": 8.24958833215714e-05, "loss": 2.7301, "step": 22323 }, { "epoch": 1.75, "learning_rate": 8.24950991923469e-05, "loss": 2.9448, "step": 22324 }, { "epoch": 1.75, "learning_rate": 8.24943150631224e-05, "loss": 2.6195, "step": 22325 }, { "epoch": 1.75, "learning_rate": 8.24935309338979e-05, "loss": 2.96, "step": 22326 }, { "epoch": 1.75, "learning_rate": 8.249274680467341e-05, "loss": 2.9681, "step": 22327 }, { "epoch": 1.75, "learning_rate": 8.249196267544893e-05, "loss": 3.187, "step": 22328 }, { "epoch": 1.75, "learning_rate": 8.249117854622443e-05, "loss": 2.9339, "step": 22329 }, { "epoch": 1.75, "learning_rate": 8.249039441699993e-05, "loss": 2.7573, "step": 22330 }, { "epoch": 1.75, "learning_rate": 8.248961028777544e-05, "loss": 2.896, "step": 22331 }, { "epoch": 1.75, "learning_rate": 8.248882615855094e-05, "loss": 2.7677, "step": 22332 }, { "epoch": 1.75, "learning_rate": 8.248804202932644e-05, "loss": 2.7326, "step": 22333 }, { "epoch": 1.75, "learning_rate": 8.248725790010195e-05, "loss": 2.8883, "step": 22334 }, { "epoch": 1.75, "learning_rate": 8.248647377087745e-05, "loss": 2.8396, "step": 22335 }, { "epoch": 1.75, "learning_rate": 8.248568964165296e-05, "loss": 2.7202, "step": 22336 }, { "epoch": 1.75, "learning_rate": 8.248490551242846e-05, "loss": 2.8267, "step": 22337 }, { "epoch": 1.75, "learning_rate": 8.248412138320396e-05, "loss": 2.6544, "step": 22338 }, { "epoch": 1.75, "learning_rate": 8.248333725397947e-05, "loss": 2.762, "step": 22339 }, { "epoch": 1.75, "learning_rate": 8.248255312475497e-05, "loss": 3.1317, "step": 22340 }, { "epoch": 1.75, "learning_rate": 8.248176899553047e-05, "loss": 2.8844, "step": 22341 }, { "epoch": 1.75, "learning_rate": 8.248098486630596e-05, "loss": 2.911, "step": 22342 }, { "epoch": 1.75, "learning_rate": 8.248020073708147e-05, "loss": 3.1695, "step": 22343 }, { "epoch": 1.75, "learning_rate": 8.247941660785697e-05, "loss": 3.1621, "step": 22344 }, { "epoch": 1.75, "learning_rate": 8.247863247863247e-05, "loss": 3.0957, "step": 22345 }, { "epoch": 1.75, "learning_rate": 8.247784834940798e-05, "loss": 2.6635, "step": 22346 }, { "epoch": 1.75, "learning_rate": 8.247706422018348e-05, "loss": 2.6931, "step": 22347 }, { "epoch": 1.75, "learning_rate": 8.247628009095899e-05, "loss": 2.6772, "step": 22348 }, { "epoch": 1.75, "learning_rate": 8.247549596173449e-05, "loss": 2.9884, "step": 22349 }, { "epoch": 1.75, "learning_rate": 8.247471183250999e-05, "loss": 3.0046, "step": 22350 }, { "epoch": 1.75, "learning_rate": 8.24739277032855e-05, "loss": 3.2041, "step": 22351 }, { "epoch": 1.75, "learning_rate": 8.2473143574061e-05, "loss": 2.839, "step": 22352 }, { "epoch": 1.75, "learning_rate": 8.247235944483652e-05, "loss": 2.7348, "step": 22353 }, { "epoch": 1.75, "learning_rate": 8.247157531561202e-05, "loss": 2.7708, "step": 22354 }, { "epoch": 1.75, "learning_rate": 8.247079118638753e-05, "loss": 2.933, "step": 22355 }, { "epoch": 1.75, "learning_rate": 8.247000705716303e-05, "loss": 2.8235, "step": 22356 }, { "epoch": 1.75, "learning_rate": 8.246922292793853e-05, "loss": 3.2346, "step": 22357 }, { "epoch": 1.75, "learning_rate": 8.246843879871404e-05, "loss": 2.6635, "step": 22358 }, { "epoch": 1.75, "learning_rate": 8.246765466948954e-05, "loss": 2.8202, "step": 22359 }, { "epoch": 1.75, "learning_rate": 8.246687054026504e-05, "loss": 2.8402, "step": 22360 }, { "epoch": 1.75, "learning_rate": 8.246608641104055e-05, "loss": 2.9861, "step": 22361 }, { "epoch": 1.75, "learning_rate": 8.246530228181605e-05, "loss": 2.6577, "step": 22362 }, { "epoch": 1.75, "learning_rate": 8.246451815259155e-05, "loss": 3.0929, "step": 22363 }, { "epoch": 1.75, "learning_rate": 8.246373402336706e-05, "loss": 3.1737, "step": 22364 }, { "epoch": 1.75, "learning_rate": 8.246294989414256e-05, "loss": 2.8918, "step": 22365 }, { "epoch": 1.75, "learning_rate": 8.246216576491807e-05, "loss": 2.9101, "step": 22366 }, { "epoch": 1.75, "learning_rate": 8.246138163569357e-05, "loss": 2.6595, "step": 22367 }, { "epoch": 1.75, "learning_rate": 8.246059750646907e-05, "loss": 2.8855, "step": 22368 }, { "epoch": 1.75, "learning_rate": 8.245981337724458e-05, "loss": 2.8766, "step": 22369 }, { "epoch": 1.75, "learning_rate": 8.245902924802008e-05, "loss": 2.8183, "step": 22370 }, { "epoch": 1.75, "learning_rate": 8.245824511879558e-05, "loss": 2.8887, "step": 22371 }, { "epoch": 1.75, "learning_rate": 8.245746098957109e-05, "loss": 2.8297, "step": 22372 }, { "epoch": 1.75, "learning_rate": 8.245667686034659e-05, "loss": 2.9797, "step": 22373 }, { "epoch": 1.75, "learning_rate": 8.24558927311221e-05, "loss": 2.7592, "step": 22374 }, { "epoch": 1.75, "learning_rate": 8.24551086018976e-05, "loss": 3.0567, "step": 22375 }, { "epoch": 1.75, "learning_rate": 8.24543244726731e-05, "loss": 2.9371, "step": 22376 }, { "epoch": 1.75, "learning_rate": 8.24535403434486e-05, "loss": 2.7102, "step": 22377 }, { "epoch": 1.75, "learning_rate": 8.24527562142241e-05, "loss": 2.7123, "step": 22378 }, { "epoch": 1.75, "learning_rate": 8.245197208499961e-05, "loss": 3.0736, "step": 22379 }, { "epoch": 1.75, "learning_rate": 8.245118795577512e-05, "loss": 2.9152, "step": 22380 }, { "epoch": 1.75, "learning_rate": 8.245040382655062e-05, "loss": 3.0698, "step": 22381 }, { "epoch": 1.76, "learning_rate": 8.244961969732612e-05, "loss": 2.8435, "step": 22382 }, { "epoch": 1.76, "learning_rate": 8.244883556810163e-05, "loss": 2.8168, "step": 22383 }, { "epoch": 1.76, "learning_rate": 8.244805143887713e-05, "loss": 2.8233, "step": 22384 }, { "epoch": 1.76, "learning_rate": 8.244726730965264e-05, "loss": 3.0086, "step": 22385 }, { "epoch": 1.76, "learning_rate": 8.244648318042814e-05, "loss": 2.8215, "step": 22386 }, { "epoch": 1.76, "learning_rate": 8.244569905120364e-05, "loss": 2.842, "step": 22387 }, { "epoch": 1.76, "learning_rate": 8.244491492197915e-05, "loss": 2.9952, "step": 22388 }, { "epoch": 1.76, "learning_rate": 8.244413079275465e-05, "loss": 2.9533, "step": 22389 }, { "epoch": 1.76, "learning_rate": 8.244334666353015e-05, "loss": 2.9471, "step": 22390 }, { "epoch": 1.76, "learning_rate": 8.244256253430566e-05, "loss": 3.0184, "step": 22391 }, { "epoch": 1.76, "learning_rate": 8.244177840508116e-05, "loss": 2.7935, "step": 22392 }, { "epoch": 1.76, "learning_rate": 8.244099427585666e-05, "loss": 2.8817, "step": 22393 }, { "epoch": 1.76, "learning_rate": 8.244021014663217e-05, "loss": 2.6095, "step": 22394 }, { "epoch": 1.76, "learning_rate": 8.243942601740767e-05, "loss": 2.9732, "step": 22395 }, { "epoch": 1.76, "learning_rate": 8.243864188818318e-05, "loss": 2.9081, "step": 22396 }, { "epoch": 1.76, "learning_rate": 8.243785775895868e-05, "loss": 2.6807, "step": 22397 }, { "epoch": 1.76, "learning_rate": 8.243707362973418e-05, "loss": 2.7484, "step": 22398 }, { "epoch": 1.76, "learning_rate": 8.243628950050969e-05, "loss": 3.0087, "step": 22399 }, { "epoch": 1.76, "learning_rate": 8.243550537128519e-05, "loss": 2.7034, "step": 22400 }, { "epoch": 1.76, "learning_rate": 8.24347212420607e-05, "loss": 2.7413, "step": 22401 }, { "epoch": 1.76, "learning_rate": 8.24339371128362e-05, "loss": 2.9105, "step": 22402 }, { "epoch": 1.76, "learning_rate": 8.24331529836117e-05, "loss": 2.7315, "step": 22403 }, { "epoch": 1.76, "learning_rate": 8.24323688543872e-05, "loss": 2.8969, "step": 22404 }, { "epoch": 1.76, "learning_rate": 8.243158472516271e-05, "loss": 2.8775, "step": 22405 }, { "epoch": 1.76, "learning_rate": 8.243080059593821e-05, "loss": 2.878, "step": 22406 }, { "epoch": 1.76, "learning_rate": 8.243001646671372e-05, "loss": 3.133, "step": 22407 }, { "epoch": 1.76, "learning_rate": 8.242923233748922e-05, "loss": 3.1962, "step": 22408 }, { "epoch": 1.76, "learning_rate": 8.242844820826472e-05, "loss": 3.135, "step": 22409 }, { "epoch": 1.76, "learning_rate": 8.242766407904024e-05, "loss": 3.1834, "step": 22410 }, { "epoch": 1.76, "learning_rate": 8.242687994981574e-05, "loss": 2.8273, "step": 22411 }, { "epoch": 1.76, "learning_rate": 8.242609582059125e-05, "loss": 2.9333, "step": 22412 }, { "epoch": 1.76, "learning_rate": 8.242531169136675e-05, "loss": 2.7362, "step": 22413 }, { "epoch": 1.76, "learning_rate": 8.242452756214224e-05, "loss": 2.8352, "step": 22414 }, { "epoch": 1.76, "learning_rate": 8.242374343291775e-05, "loss": 3.155, "step": 22415 }, { "epoch": 1.76, "learning_rate": 8.242295930369325e-05, "loss": 2.8636, "step": 22416 }, { "epoch": 1.76, "learning_rate": 8.242217517446875e-05, "loss": 2.9057, "step": 22417 }, { "epoch": 1.76, "learning_rate": 8.242139104524426e-05, "loss": 2.9085, "step": 22418 }, { "epoch": 1.76, "learning_rate": 8.242060691601976e-05, "loss": 2.8322, "step": 22419 }, { "epoch": 1.76, "learning_rate": 8.241982278679526e-05, "loss": 2.9758, "step": 22420 }, { "epoch": 1.76, "learning_rate": 8.241903865757077e-05, "loss": 2.9295, "step": 22421 }, { "epoch": 1.76, "learning_rate": 8.241825452834627e-05, "loss": 2.7291, "step": 22422 }, { "epoch": 1.76, "learning_rate": 8.241747039912178e-05, "loss": 3.0425, "step": 22423 }, { "epoch": 1.76, "learning_rate": 8.241668626989728e-05, "loss": 2.8327, "step": 22424 }, { "epoch": 1.76, "learning_rate": 8.241590214067278e-05, "loss": 2.6706, "step": 22425 }, { "epoch": 1.76, "learning_rate": 8.241511801144829e-05, "loss": 2.7618, "step": 22426 }, { "epoch": 1.76, "learning_rate": 8.241433388222379e-05, "loss": 2.9653, "step": 22427 }, { "epoch": 1.76, "learning_rate": 8.24135497529993e-05, "loss": 3.0996, "step": 22428 }, { "epoch": 1.76, "learning_rate": 8.24127656237748e-05, "loss": 2.7447, "step": 22429 }, { "epoch": 1.76, "learning_rate": 8.24119814945503e-05, "loss": 2.6734, "step": 22430 }, { "epoch": 1.76, "learning_rate": 8.24111973653258e-05, "loss": 3.1389, "step": 22431 }, { "epoch": 1.76, "learning_rate": 8.241041323610131e-05, "loss": 2.602, "step": 22432 }, { "epoch": 1.76, "learning_rate": 8.240962910687681e-05, "loss": 3.0418, "step": 22433 }, { "epoch": 1.76, "learning_rate": 8.240884497765232e-05, "loss": 2.8404, "step": 22434 }, { "epoch": 1.76, "learning_rate": 8.240806084842782e-05, "loss": 2.6603, "step": 22435 }, { "epoch": 1.76, "learning_rate": 8.240727671920334e-05, "loss": 3.0458, "step": 22436 }, { "epoch": 1.76, "learning_rate": 8.240649258997884e-05, "loss": 2.8279, "step": 22437 }, { "epoch": 1.76, "learning_rate": 8.240570846075434e-05, "loss": 2.905, "step": 22438 }, { "epoch": 1.76, "learning_rate": 8.240492433152985e-05, "loss": 3.2653, "step": 22439 }, { "epoch": 1.76, "learning_rate": 8.240414020230535e-05, "loss": 3.0286, "step": 22440 }, { "epoch": 1.76, "learning_rate": 8.240335607308085e-05, "loss": 3.103, "step": 22441 }, { "epoch": 1.76, "learning_rate": 8.240257194385636e-05, "loss": 2.9796, "step": 22442 }, { "epoch": 1.76, "learning_rate": 8.240178781463186e-05, "loss": 2.8555, "step": 22443 }, { "epoch": 1.76, "learning_rate": 8.240100368540737e-05, "loss": 2.9292, "step": 22444 }, { "epoch": 1.76, "learning_rate": 8.240021955618287e-05, "loss": 2.9507, "step": 22445 }, { "epoch": 1.76, "learning_rate": 8.239943542695837e-05, "loss": 2.8229, "step": 22446 }, { "epoch": 1.76, "learning_rate": 8.239865129773388e-05, "loss": 2.9391, "step": 22447 }, { "epoch": 1.76, "learning_rate": 8.239786716850938e-05, "loss": 2.8633, "step": 22448 }, { "epoch": 1.76, "learning_rate": 8.239708303928488e-05, "loss": 2.7324, "step": 22449 }, { "epoch": 1.76, "learning_rate": 8.239629891006037e-05, "loss": 2.927, "step": 22450 }, { "epoch": 1.76, "learning_rate": 8.239551478083588e-05, "loss": 2.9012, "step": 22451 }, { "epoch": 1.76, "learning_rate": 8.239473065161138e-05, "loss": 3.0051, "step": 22452 }, { "epoch": 1.76, "learning_rate": 8.239394652238689e-05, "loss": 2.6893, "step": 22453 }, { "epoch": 1.76, "learning_rate": 8.239316239316239e-05, "loss": 3.0127, "step": 22454 }, { "epoch": 1.76, "learning_rate": 8.239237826393789e-05, "loss": 2.6959, "step": 22455 }, { "epoch": 1.76, "learning_rate": 8.23915941347134e-05, "loss": 2.8474, "step": 22456 }, { "epoch": 1.76, "learning_rate": 8.23908100054889e-05, "loss": 3.168, "step": 22457 }, { "epoch": 1.76, "learning_rate": 8.23900258762644e-05, "loss": 3.0067, "step": 22458 }, { "epoch": 1.76, "learning_rate": 8.238924174703991e-05, "loss": 2.9753, "step": 22459 }, { "epoch": 1.76, "learning_rate": 8.238845761781541e-05, "loss": 2.7948, "step": 22460 }, { "epoch": 1.76, "learning_rate": 8.238767348859093e-05, "loss": 3.1064, "step": 22461 }, { "epoch": 1.76, "learning_rate": 8.238688935936643e-05, "loss": 2.8717, "step": 22462 }, { "epoch": 1.76, "learning_rate": 8.238610523014194e-05, "loss": 2.8401, "step": 22463 }, { "epoch": 1.76, "learning_rate": 8.238532110091744e-05, "loss": 2.9414, "step": 22464 }, { "epoch": 1.76, "learning_rate": 8.238453697169294e-05, "loss": 2.8971, "step": 22465 }, { "epoch": 1.76, "learning_rate": 8.238375284246845e-05, "loss": 2.9873, "step": 22466 }, { "epoch": 1.76, "learning_rate": 8.238296871324395e-05, "loss": 2.8093, "step": 22467 }, { "epoch": 1.76, "learning_rate": 8.238218458401945e-05, "loss": 3.1686, "step": 22468 }, { "epoch": 1.76, "learning_rate": 8.238140045479496e-05, "loss": 2.87, "step": 22469 }, { "epoch": 1.76, "learning_rate": 8.238061632557046e-05, "loss": 2.8567, "step": 22470 }, { "epoch": 1.76, "learning_rate": 8.237983219634597e-05, "loss": 2.9881, "step": 22471 }, { "epoch": 1.76, "learning_rate": 8.237904806712147e-05, "loss": 3.0104, "step": 22472 }, { "epoch": 1.76, "learning_rate": 8.237826393789697e-05, "loss": 2.8421, "step": 22473 }, { "epoch": 1.76, "learning_rate": 8.237747980867248e-05, "loss": 2.887, "step": 22474 }, { "epoch": 1.76, "learning_rate": 8.237669567944798e-05, "loss": 2.8024, "step": 22475 }, { "epoch": 1.76, "learning_rate": 8.237591155022348e-05, "loss": 2.9963, "step": 22476 }, { "epoch": 1.76, "learning_rate": 8.237512742099899e-05, "loss": 2.7707, "step": 22477 }, { "epoch": 1.76, "learning_rate": 8.237434329177449e-05, "loss": 2.8779, "step": 22478 }, { "epoch": 1.76, "learning_rate": 8.237355916255e-05, "loss": 2.875, "step": 22479 }, { "epoch": 1.76, "learning_rate": 8.23727750333255e-05, "loss": 2.8846, "step": 22480 }, { "epoch": 1.76, "learning_rate": 8.2371990904101e-05, "loss": 2.8935, "step": 22481 }, { "epoch": 1.76, "learning_rate": 8.23712067748765e-05, "loss": 2.9896, "step": 22482 }, { "epoch": 1.76, "learning_rate": 8.237042264565201e-05, "loss": 3.098, "step": 22483 }, { "epoch": 1.76, "learning_rate": 8.236963851642751e-05, "loss": 2.8516, "step": 22484 }, { "epoch": 1.76, "learning_rate": 8.236885438720302e-05, "loss": 3.0847, "step": 22485 }, { "epoch": 1.76, "learning_rate": 8.236807025797851e-05, "loss": 2.8342, "step": 22486 }, { "epoch": 1.76, "learning_rate": 8.236728612875402e-05, "loss": 2.9867, "step": 22487 }, { "epoch": 1.76, "learning_rate": 8.236650199952953e-05, "loss": 2.7697, "step": 22488 }, { "epoch": 1.76, "learning_rate": 8.236571787030503e-05, "loss": 2.7703, "step": 22489 }, { "epoch": 1.76, "learning_rate": 8.236493374108054e-05, "loss": 2.8182, "step": 22490 }, { "epoch": 1.76, "learning_rate": 8.236414961185604e-05, "loss": 2.8631, "step": 22491 }, { "epoch": 1.76, "learning_rate": 8.236336548263154e-05, "loss": 2.9181, "step": 22492 }, { "epoch": 1.76, "learning_rate": 8.236258135340705e-05, "loss": 3.0196, "step": 22493 }, { "epoch": 1.76, "learning_rate": 8.236179722418255e-05, "loss": 2.7611, "step": 22494 }, { "epoch": 1.76, "learning_rate": 8.236101309495805e-05, "loss": 3.0471, "step": 22495 }, { "epoch": 1.76, "learning_rate": 8.236022896573356e-05, "loss": 3.0122, "step": 22496 }, { "epoch": 1.76, "learning_rate": 8.235944483650906e-05, "loss": 2.8512, "step": 22497 }, { "epoch": 1.76, "learning_rate": 8.235866070728456e-05, "loss": 2.8445, "step": 22498 }, { "epoch": 1.76, "learning_rate": 8.235787657806007e-05, "loss": 2.7141, "step": 22499 }, { "epoch": 1.76, "learning_rate": 8.235709244883557e-05, "loss": 2.9051, "step": 22500 }, { "epoch": 1.76, "learning_rate": 8.235630831961108e-05, "loss": 3.2165, "step": 22501 }, { "epoch": 1.76, "learning_rate": 8.235552419038658e-05, "loss": 2.7694, "step": 22502 }, { "epoch": 1.76, "learning_rate": 8.235474006116208e-05, "loss": 2.994, "step": 22503 }, { "epoch": 1.76, "learning_rate": 8.235395593193759e-05, "loss": 2.842, "step": 22504 }, { "epoch": 1.76, "learning_rate": 8.235317180271309e-05, "loss": 3.067, "step": 22505 }, { "epoch": 1.76, "learning_rate": 8.23523876734886e-05, "loss": 3.0262, "step": 22506 }, { "epoch": 1.76, "learning_rate": 8.23516035442641e-05, "loss": 2.8164, "step": 22507 }, { "epoch": 1.76, "learning_rate": 8.23508194150396e-05, "loss": 2.847, "step": 22508 }, { "epoch": 1.76, "learning_rate": 8.23500352858151e-05, "loss": 2.9428, "step": 22509 }, { "epoch": 1.77, "learning_rate": 8.234925115659061e-05, "loss": 2.9346, "step": 22510 }, { "epoch": 1.77, "learning_rate": 8.234846702736611e-05, "loss": 3.0531, "step": 22511 }, { "epoch": 1.77, "learning_rate": 8.234768289814162e-05, "loss": 2.9525, "step": 22512 }, { "epoch": 1.77, "learning_rate": 8.234689876891712e-05, "loss": 2.8204, "step": 22513 }, { "epoch": 1.77, "learning_rate": 8.234611463969262e-05, "loss": 2.9021, "step": 22514 }, { "epoch": 1.77, "learning_rate": 8.234533051046813e-05, "loss": 2.7871, "step": 22515 }, { "epoch": 1.77, "learning_rate": 8.234454638124363e-05, "loss": 3.1531, "step": 22516 }, { "epoch": 1.77, "learning_rate": 8.234376225201913e-05, "loss": 3.0114, "step": 22517 }, { "epoch": 1.77, "learning_rate": 8.234297812279465e-05, "loss": 3.2203, "step": 22518 }, { "epoch": 1.77, "learning_rate": 8.234219399357016e-05, "loss": 3.0889, "step": 22519 }, { "epoch": 1.77, "learning_rate": 8.234140986434566e-05, "loss": 2.968, "step": 22520 }, { "epoch": 1.77, "learning_rate": 8.234062573512116e-05, "loss": 2.8012, "step": 22521 }, { "epoch": 1.77, "learning_rate": 8.233984160589665e-05, "loss": 2.9292, "step": 22522 }, { "epoch": 1.77, "learning_rate": 8.233905747667216e-05, "loss": 2.6462, "step": 22523 }, { "epoch": 1.77, "learning_rate": 8.233827334744766e-05, "loss": 2.8656, "step": 22524 }, { "epoch": 1.77, "learning_rate": 8.233748921822316e-05, "loss": 2.7823, "step": 22525 }, { "epoch": 1.77, "learning_rate": 8.233670508899867e-05, "loss": 2.9494, "step": 22526 }, { "epoch": 1.77, "learning_rate": 8.233592095977417e-05, "loss": 3.0905, "step": 22527 }, { "epoch": 1.77, "learning_rate": 8.233513683054967e-05, "loss": 2.7425, "step": 22528 }, { "epoch": 1.77, "learning_rate": 8.233435270132518e-05, "loss": 3.0446, "step": 22529 }, { "epoch": 1.77, "learning_rate": 8.233356857210068e-05, "loss": 2.8024, "step": 22530 }, { "epoch": 1.77, "learning_rate": 8.233278444287619e-05, "loss": 2.9411, "step": 22531 }, { "epoch": 1.77, "learning_rate": 8.233200031365169e-05, "loss": 2.8618, "step": 22532 }, { "epoch": 1.77, "learning_rate": 8.23312161844272e-05, "loss": 2.8523, "step": 22533 }, { "epoch": 1.77, "learning_rate": 8.23304320552027e-05, "loss": 2.8682, "step": 22534 }, { "epoch": 1.77, "learning_rate": 8.23296479259782e-05, "loss": 3.1301, "step": 22535 }, { "epoch": 1.77, "learning_rate": 8.23288637967537e-05, "loss": 2.7713, "step": 22536 }, { "epoch": 1.77, "learning_rate": 8.232807966752921e-05, "loss": 2.9375, "step": 22537 }, { "epoch": 1.77, "learning_rate": 8.232729553830471e-05, "loss": 2.9746, "step": 22538 }, { "epoch": 1.77, "learning_rate": 8.232651140908022e-05, "loss": 2.8692, "step": 22539 }, { "epoch": 1.77, "learning_rate": 8.232572727985572e-05, "loss": 3.075, "step": 22540 }, { "epoch": 1.77, "learning_rate": 8.232494315063122e-05, "loss": 2.5978, "step": 22541 }, { "epoch": 1.77, "learning_rate": 8.232415902140673e-05, "loss": 3.5584, "step": 22542 }, { "epoch": 1.77, "learning_rate": 8.232337489218223e-05, "loss": 3.1067, "step": 22543 }, { "epoch": 1.77, "learning_rate": 8.232259076295775e-05, "loss": 2.7874, "step": 22544 }, { "epoch": 1.77, "learning_rate": 8.232180663373325e-05, "loss": 2.7784, "step": 22545 }, { "epoch": 1.77, "learning_rate": 8.232102250450875e-05, "loss": 3.1248, "step": 22546 }, { "epoch": 1.77, "learning_rate": 8.232023837528426e-05, "loss": 2.9264, "step": 22547 }, { "epoch": 1.77, "learning_rate": 8.231945424605976e-05, "loss": 2.8175, "step": 22548 }, { "epoch": 1.77, "learning_rate": 8.231867011683527e-05, "loss": 2.897, "step": 22549 }, { "epoch": 1.77, "learning_rate": 8.231788598761077e-05, "loss": 2.9941, "step": 22550 }, { "epoch": 1.77, "learning_rate": 8.231710185838627e-05, "loss": 3.0829, "step": 22551 }, { "epoch": 1.77, "learning_rate": 8.231631772916178e-05, "loss": 2.9792, "step": 22552 }, { "epoch": 1.77, "learning_rate": 8.231553359993728e-05, "loss": 2.7506, "step": 22553 }, { "epoch": 1.77, "learning_rate": 8.231474947071278e-05, "loss": 3.0185, "step": 22554 }, { "epoch": 1.77, "learning_rate": 8.231396534148829e-05, "loss": 2.7097, "step": 22555 }, { "epoch": 1.77, "learning_rate": 8.231318121226379e-05, "loss": 3.0069, "step": 22556 }, { "epoch": 1.77, "learning_rate": 8.23123970830393e-05, "loss": 3.0474, "step": 22557 }, { "epoch": 1.77, "learning_rate": 8.231161295381479e-05, "loss": 3.0059, "step": 22558 }, { "epoch": 1.77, "learning_rate": 8.231082882459029e-05, "loss": 2.9601, "step": 22559 }, { "epoch": 1.77, "learning_rate": 8.231004469536579e-05, "loss": 3.1826, "step": 22560 }, { "epoch": 1.77, "learning_rate": 8.23092605661413e-05, "loss": 2.9675, "step": 22561 }, { "epoch": 1.77, "learning_rate": 8.23084764369168e-05, "loss": 2.7411, "step": 22562 }, { "epoch": 1.77, "learning_rate": 8.23076923076923e-05, "loss": 2.7983, "step": 22563 }, { "epoch": 1.77, "learning_rate": 8.230690817846781e-05, "loss": 3.0574, "step": 22564 }, { "epoch": 1.77, "learning_rate": 8.230612404924331e-05, "loss": 2.875, "step": 22565 }, { "epoch": 1.77, "learning_rate": 8.230533992001881e-05, "loss": 2.9307, "step": 22566 }, { "epoch": 1.77, "learning_rate": 8.230455579079432e-05, "loss": 2.8686, "step": 22567 }, { "epoch": 1.77, "learning_rate": 8.230377166156982e-05, "loss": 2.7287, "step": 22568 }, { "epoch": 1.77, "learning_rate": 8.230298753234534e-05, "loss": 2.7875, "step": 22569 }, { "epoch": 1.77, "learning_rate": 8.230220340312084e-05, "loss": 2.7427, "step": 22570 }, { "epoch": 1.77, "learning_rate": 8.230141927389635e-05, "loss": 2.932, "step": 22571 }, { "epoch": 1.77, "learning_rate": 8.230063514467185e-05, "loss": 2.9065, "step": 22572 }, { "epoch": 1.77, "learning_rate": 8.229985101544735e-05, "loss": 3.0389, "step": 22573 }, { "epoch": 1.77, "learning_rate": 8.229906688622286e-05, "loss": 3.0276, "step": 22574 }, { "epoch": 1.77, "learning_rate": 8.229828275699836e-05, "loss": 2.958, "step": 22575 }, { "epoch": 1.77, "learning_rate": 8.229749862777386e-05, "loss": 3.1059, "step": 22576 }, { "epoch": 1.77, "learning_rate": 8.229671449854937e-05, "loss": 2.9055, "step": 22577 }, { "epoch": 1.77, "learning_rate": 8.229593036932487e-05, "loss": 2.9342, "step": 22578 }, { "epoch": 1.77, "learning_rate": 8.229514624010038e-05, "loss": 2.8471, "step": 22579 }, { "epoch": 1.77, "learning_rate": 8.229436211087588e-05, "loss": 2.7557, "step": 22580 }, { "epoch": 1.77, "learning_rate": 8.229357798165138e-05, "loss": 3.0738, "step": 22581 }, { "epoch": 1.77, "learning_rate": 8.229279385242689e-05, "loss": 2.9288, "step": 22582 }, { "epoch": 1.77, "learning_rate": 8.229200972320239e-05, "loss": 2.7798, "step": 22583 }, { "epoch": 1.77, "learning_rate": 8.22912255939779e-05, "loss": 3.1132, "step": 22584 }, { "epoch": 1.77, "learning_rate": 8.22904414647534e-05, "loss": 2.8066, "step": 22585 }, { "epoch": 1.77, "learning_rate": 8.22896573355289e-05, "loss": 3.1974, "step": 22586 }, { "epoch": 1.77, "learning_rate": 8.22888732063044e-05, "loss": 2.8688, "step": 22587 }, { "epoch": 1.77, "learning_rate": 8.228808907707991e-05, "loss": 3.1896, "step": 22588 }, { "epoch": 1.77, "learning_rate": 8.228730494785541e-05, "loss": 3.0443, "step": 22589 }, { "epoch": 1.77, "learning_rate": 8.228652081863092e-05, "loss": 3.0238, "step": 22590 }, { "epoch": 1.77, "learning_rate": 8.228573668940642e-05, "loss": 2.6861, "step": 22591 }, { "epoch": 1.77, "learning_rate": 8.228495256018192e-05, "loss": 2.945, "step": 22592 }, { "epoch": 1.77, "learning_rate": 8.228416843095743e-05, "loss": 3.0181, "step": 22593 }, { "epoch": 1.77, "learning_rate": 8.228338430173292e-05, "loss": 2.8193, "step": 22594 }, { "epoch": 1.77, "learning_rate": 8.228260017250843e-05, "loss": 2.7636, "step": 22595 }, { "epoch": 1.77, "learning_rate": 8.228181604328394e-05, "loss": 2.8973, "step": 22596 }, { "epoch": 1.77, "learning_rate": 8.228103191405944e-05, "loss": 3.0801, "step": 22597 }, { "epoch": 1.77, "learning_rate": 8.228024778483495e-05, "loss": 3.0104, "step": 22598 }, { "epoch": 1.77, "learning_rate": 8.227946365561045e-05, "loss": 2.8352, "step": 22599 }, { "epoch": 1.77, "learning_rate": 8.227867952638595e-05, "loss": 2.8482, "step": 22600 }, { "epoch": 1.77, "learning_rate": 8.227789539716146e-05, "loss": 2.8847, "step": 22601 }, { "epoch": 1.77, "learning_rate": 8.227711126793696e-05, "loss": 2.8497, "step": 22602 }, { "epoch": 1.77, "learning_rate": 8.227632713871246e-05, "loss": 2.9539, "step": 22603 }, { "epoch": 1.77, "learning_rate": 8.227554300948797e-05, "loss": 2.8058, "step": 22604 }, { "epoch": 1.77, "learning_rate": 8.227475888026347e-05, "loss": 3.0191, "step": 22605 }, { "epoch": 1.77, "learning_rate": 8.227397475103898e-05, "loss": 2.916, "step": 22606 }, { "epoch": 1.77, "learning_rate": 8.227319062181448e-05, "loss": 2.9139, "step": 22607 }, { "epoch": 1.77, "learning_rate": 8.227240649258998e-05, "loss": 3.2842, "step": 22608 }, { "epoch": 1.77, "learning_rate": 8.227162236336549e-05, "loss": 2.8004, "step": 22609 }, { "epoch": 1.77, "learning_rate": 8.227083823414099e-05, "loss": 3.5645, "step": 22610 }, { "epoch": 1.77, "learning_rate": 8.22700541049165e-05, "loss": 2.9128, "step": 22611 }, { "epoch": 1.77, "learning_rate": 8.2269269975692e-05, "loss": 2.8537, "step": 22612 }, { "epoch": 1.77, "learning_rate": 8.22684858464675e-05, "loss": 3.0827, "step": 22613 }, { "epoch": 1.77, "learning_rate": 8.2267701717243e-05, "loss": 2.8946, "step": 22614 }, { "epoch": 1.77, "learning_rate": 8.226691758801851e-05, "loss": 2.9888, "step": 22615 }, { "epoch": 1.77, "learning_rate": 8.226613345879401e-05, "loss": 3.0903, "step": 22616 }, { "epoch": 1.77, "learning_rate": 8.226534932956952e-05, "loss": 2.9529, "step": 22617 }, { "epoch": 1.77, "learning_rate": 8.226456520034502e-05, "loss": 2.866, "step": 22618 }, { "epoch": 1.77, "learning_rate": 8.226378107112052e-05, "loss": 2.9506, "step": 22619 }, { "epoch": 1.77, "learning_rate": 8.226299694189603e-05, "loss": 2.8, "step": 22620 }, { "epoch": 1.77, "learning_rate": 8.226221281267153e-05, "loss": 2.9822, "step": 22621 }, { "epoch": 1.77, "learning_rate": 8.226142868344703e-05, "loss": 3.1527, "step": 22622 }, { "epoch": 1.77, "learning_rate": 8.226064455422254e-05, "loss": 2.7557, "step": 22623 }, { "epoch": 1.77, "learning_rate": 8.225986042499804e-05, "loss": 3.0013, "step": 22624 }, { "epoch": 1.77, "learning_rate": 8.225907629577355e-05, "loss": 2.8869, "step": 22625 }, { "epoch": 1.77, "learning_rate": 8.225829216654906e-05, "loss": 3.1165, "step": 22626 }, { "epoch": 1.77, "learning_rate": 8.225750803732457e-05, "loss": 2.8024, "step": 22627 }, { "epoch": 1.77, "learning_rate": 8.225672390810007e-05, "loss": 2.8558, "step": 22628 }, { "epoch": 1.77, "learning_rate": 8.225593977887557e-05, "loss": 2.8181, "step": 22629 }, { "epoch": 1.77, "learning_rate": 8.225515564965106e-05, "loss": 2.8634, "step": 22630 }, { "epoch": 1.77, "learning_rate": 8.225437152042657e-05, "loss": 3.0497, "step": 22631 }, { "epoch": 1.77, "learning_rate": 8.225358739120207e-05, "loss": 2.7812, "step": 22632 }, { "epoch": 1.77, "learning_rate": 8.225280326197757e-05, "loss": 2.8722, "step": 22633 }, { "epoch": 1.77, "learning_rate": 8.225201913275308e-05, "loss": 2.9054, "step": 22634 }, { "epoch": 1.77, "learning_rate": 8.225123500352858e-05, "loss": 3.0579, "step": 22635 }, { "epoch": 1.77, "learning_rate": 8.225045087430409e-05, "loss": 2.9578, "step": 22636 }, { "epoch": 1.78, "learning_rate": 8.224966674507959e-05, "loss": 2.883, "step": 22637 }, { "epoch": 1.78, "learning_rate": 8.224888261585509e-05, "loss": 2.9356, "step": 22638 }, { "epoch": 1.78, "learning_rate": 8.22480984866306e-05, "loss": 3.0576, "step": 22639 }, { "epoch": 1.78, "learning_rate": 8.22473143574061e-05, "loss": 2.6996, "step": 22640 }, { "epoch": 1.78, "learning_rate": 8.22465302281816e-05, "loss": 3.0786, "step": 22641 }, { "epoch": 1.78, "learning_rate": 8.224574609895711e-05, "loss": 3.2414, "step": 22642 }, { "epoch": 1.78, "learning_rate": 8.224496196973261e-05, "loss": 3.2862, "step": 22643 }, { "epoch": 1.78, "learning_rate": 8.224417784050812e-05, "loss": 2.6501, "step": 22644 }, { "epoch": 1.78, "learning_rate": 8.224339371128362e-05, "loss": 2.6941, "step": 22645 }, { "epoch": 1.78, "learning_rate": 8.224260958205912e-05, "loss": 2.8554, "step": 22646 }, { "epoch": 1.78, "learning_rate": 8.224182545283463e-05, "loss": 2.9566, "step": 22647 }, { "epoch": 1.78, "learning_rate": 8.224104132361013e-05, "loss": 2.8558, "step": 22648 }, { "epoch": 1.78, "learning_rate": 8.224025719438563e-05, "loss": 2.9998, "step": 22649 }, { "epoch": 1.78, "learning_rate": 8.223947306516114e-05, "loss": 2.7339, "step": 22650 }, { "epoch": 1.78, "learning_rate": 8.223868893593664e-05, "loss": 2.9043, "step": 22651 }, { "epoch": 1.78, "learning_rate": 8.223790480671216e-05, "loss": 2.9125, "step": 22652 }, { "epoch": 1.78, "learning_rate": 8.223712067748766e-05, "loss": 2.8661, "step": 22653 }, { "epoch": 1.78, "learning_rate": 8.223633654826317e-05, "loss": 2.6676, "step": 22654 }, { "epoch": 1.78, "learning_rate": 8.223555241903867e-05, "loss": 2.8993, "step": 22655 }, { "epoch": 1.78, "learning_rate": 8.223476828981417e-05, "loss": 2.8536, "step": 22656 }, { "epoch": 1.78, "learning_rate": 8.223398416058968e-05, "loss": 2.9022, "step": 22657 }, { "epoch": 1.78, "learning_rate": 8.223320003136518e-05, "loss": 3.1831, "step": 22658 }, { "epoch": 1.78, "learning_rate": 8.223241590214068e-05, "loss": 3.2281, "step": 22659 }, { "epoch": 1.78, "learning_rate": 8.223163177291619e-05, "loss": 3.0038, "step": 22660 }, { "epoch": 1.78, "learning_rate": 8.223084764369169e-05, "loss": 2.8852, "step": 22661 }, { "epoch": 1.78, "learning_rate": 8.22300635144672e-05, "loss": 3.0476, "step": 22662 }, { "epoch": 1.78, "learning_rate": 8.22292793852427e-05, "loss": 2.6977, "step": 22663 }, { "epoch": 1.78, "learning_rate": 8.22284952560182e-05, "loss": 3.0218, "step": 22664 }, { "epoch": 1.78, "learning_rate": 8.22277111267937e-05, "loss": 2.8665, "step": 22665 }, { "epoch": 1.78, "learning_rate": 8.22269269975692e-05, "loss": 2.7594, "step": 22666 }, { "epoch": 1.78, "learning_rate": 8.22261428683447e-05, "loss": 2.93, "step": 22667 }, { "epoch": 1.78, "learning_rate": 8.22253587391202e-05, "loss": 2.9139, "step": 22668 }, { "epoch": 1.78, "learning_rate": 8.222457460989571e-05, "loss": 2.7945, "step": 22669 }, { "epoch": 1.78, "learning_rate": 8.222379048067121e-05, "loss": 3.0038, "step": 22670 }, { "epoch": 1.78, "learning_rate": 8.222300635144671e-05, "loss": 2.9769, "step": 22671 }, { "epoch": 1.78, "learning_rate": 8.222222222222222e-05, "loss": 2.8342, "step": 22672 }, { "epoch": 1.78, "learning_rate": 8.222143809299772e-05, "loss": 2.7442, "step": 22673 }, { "epoch": 1.78, "learning_rate": 8.222065396377323e-05, "loss": 3.1342, "step": 22674 }, { "epoch": 1.78, "learning_rate": 8.221986983454873e-05, "loss": 3.0092, "step": 22675 }, { "epoch": 1.78, "learning_rate": 8.221908570532423e-05, "loss": 2.9272, "step": 22676 }, { "epoch": 1.78, "learning_rate": 8.221830157609975e-05, "loss": 2.7479, "step": 22677 }, { "epoch": 1.78, "learning_rate": 8.221751744687525e-05, "loss": 2.8513, "step": 22678 }, { "epoch": 1.78, "learning_rate": 8.221673331765076e-05, "loss": 2.8744, "step": 22679 }, { "epoch": 1.78, "learning_rate": 8.221594918842626e-05, "loss": 2.9519, "step": 22680 }, { "epoch": 1.78, "learning_rate": 8.221516505920176e-05, "loss": 2.9594, "step": 22681 }, { "epoch": 1.78, "learning_rate": 8.221438092997727e-05, "loss": 2.8075, "step": 22682 }, { "epoch": 1.78, "learning_rate": 8.221359680075277e-05, "loss": 2.8735, "step": 22683 }, { "epoch": 1.78, "learning_rate": 8.221281267152828e-05, "loss": 2.8662, "step": 22684 }, { "epoch": 1.78, "learning_rate": 8.221202854230378e-05, "loss": 2.7348, "step": 22685 }, { "epoch": 1.78, "learning_rate": 8.221124441307928e-05, "loss": 3.2206, "step": 22686 }, { "epoch": 1.78, "learning_rate": 8.221046028385479e-05, "loss": 2.8868, "step": 22687 }, { "epoch": 1.78, "learning_rate": 8.220967615463029e-05, "loss": 3.027, "step": 22688 }, { "epoch": 1.78, "learning_rate": 8.22088920254058e-05, "loss": 2.9849, "step": 22689 }, { "epoch": 1.78, "learning_rate": 8.22081078961813e-05, "loss": 2.8064, "step": 22690 }, { "epoch": 1.78, "learning_rate": 8.22073237669568e-05, "loss": 2.8799, "step": 22691 }, { "epoch": 1.78, "learning_rate": 8.22065396377323e-05, "loss": 2.6614, "step": 22692 }, { "epoch": 1.78, "learning_rate": 8.220575550850781e-05, "loss": 2.9004, "step": 22693 }, { "epoch": 1.78, "learning_rate": 8.220497137928331e-05, "loss": 2.9875, "step": 22694 }, { "epoch": 1.78, "learning_rate": 8.220418725005882e-05, "loss": 2.979, "step": 22695 }, { "epoch": 1.78, "learning_rate": 8.220340312083432e-05, "loss": 3.0536, "step": 22696 }, { "epoch": 1.78, "learning_rate": 8.220261899160982e-05, "loss": 2.991, "step": 22697 }, { "epoch": 1.78, "learning_rate": 8.220183486238533e-05, "loss": 2.8074, "step": 22698 }, { "epoch": 1.78, "learning_rate": 8.220105073316083e-05, "loss": 2.7873, "step": 22699 }, { "epoch": 1.78, "learning_rate": 8.220026660393633e-05, "loss": 2.9152, "step": 22700 }, { "epoch": 1.78, "learning_rate": 8.219948247471184e-05, "loss": 2.7137, "step": 22701 }, { "epoch": 1.78, "learning_rate": 8.219869834548733e-05, "loss": 2.8142, "step": 22702 }, { "epoch": 1.78, "learning_rate": 8.219791421626285e-05, "loss": 2.9725, "step": 22703 }, { "epoch": 1.78, "learning_rate": 8.219713008703835e-05, "loss": 2.9234, "step": 22704 }, { "epoch": 1.78, "learning_rate": 8.219634595781385e-05, "loss": 2.8318, "step": 22705 }, { "epoch": 1.78, "learning_rate": 8.219556182858936e-05, "loss": 3.0779, "step": 22706 }, { "epoch": 1.78, "learning_rate": 8.219477769936486e-05, "loss": 2.8392, "step": 22707 }, { "epoch": 1.78, "learning_rate": 8.219399357014036e-05, "loss": 2.7896, "step": 22708 }, { "epoch": 1.78, "learning_rate": 8.219320944091587e-05, "loss": 2.9592, "step": 22709 }, { "epoch": 1.78, "learning_rate": 8.219242531169137e-05, "loss": 3.1046, "step": 22710 }, { "epoch": 1.78, "learning_rate": 8.219164118246688e-05, "loss": 2.999, "step": 22711 }, { "epoch": 1.78, "learning_rate": 8.219085705324238e-05, "loss": 2.7566, "step": 22712 }, { "epoch": 1.78, "learning_rate": 8.219007292401788e-05, "loss": 2.7111, "step": 22713 }, { "epoch": 1.78, "learning_rate": 8.218928879479339e-05, "loss": 2.8583, "step": 22714 }, { "epoch": 1.78, "learning_rate": 8.218850466556889e-05, "loss": 3.0831, "step": 22715 }, { "epoch": 1.78, "learning_rate": 8.21877205363444e-05, "loss": 2.8168, "step": 22716 }, { "epoch": 1.78, "learning_rate": 8.21869364071199e-05, "loss": 2.8647, "step": 22717 }, { "epoch": 1.78, "learning_rate": 8.21861522778954e-05, "loss": 2.8992, "step": 22718 }, { "epoch": 1.78, "learning_rate": 8.21853681486709e-05, "loss": 3.0664, "step": 22719 }, { "epoch": 1.78, "learning_rate": 8.218458401944641e-05, "loss": 2.9658, "step": 22720 }, { "epoch": 1.78, "learning_rate": 8.218379989022191e-05, "loss": 2.9815, "step": 22721 }, { "epoch": 1.78, "learning_rate": 8.218301576099742e-05, "loss": 2.8723, "step": 22722 }, { "epoch": 1.78, "learning_rate": 8.218223163177292e-05, "loss": 3.1311, "step": 22723 }, { "epoch": 1.78, "learning_rate": 8.218144750254842e-05, "loss": 3.0864, "step": 22724 }, { "epoch": 1.78, "learning_rate": 8.218066337332393e-05, "loss": 3.0786, "step": 22725 }, { "epoch": 1.78, "learning_rate": 8.217987924409943e-05, "loss": 2.8732, "step": 22726 }, { "epoch": 1.78, "learning_rate": 8.217909511487493e-05, "loss": 2.8348, "step": 22727 }, { "epoch": 1.78, "learning_rate": 8.217831098565044e-05, "loss": 3.0792, "step": 22728 }, { "epoch": 1.78, "learning_rate": 8.217752685642594e-05, "loss": 3.1184, "step": 22729 }, { "epoch": 1.78, "learning_rate": 8.217674272720144e-05, "loss": 3.179, "step": 22730 }, { "epoch": 1.78, "learning_rate": 8.217595859797695e-05, "loss": 2.7245, "step": 22731 }, { "epoch": 1.78, "learning_rate": 8.217517446875245e-05, "loss": 3.1077, "step": 22732 }, { "epoch": 1.78, "learning_rate": 8.217439033952796e-05, "loss": 3.2493, "step": 22733 }, { "epoch": 1.78, "learning_rate": 8.217360621030347e-05, "loss": 2.9665, "step": 22734 }, { "epoch": 1.78, "learning_rate": 8.217282208107898e-05, "loss": 2.9288, "step": 22735 }, { "epoch": 1.78, "learning_rate": 8.217203795185448e-05, "loss": 2.9603, "step": 22736 }, { "epoch": 1.78, "learning_rate": 8.217125382262998e-05, "loss": 2.8339, "step": 22737 }, { "epoch": 1.78, "learning_rate": 8.217046969340547e-05, "loss": 2.8618, "step": 22738 }, { "epoch": 1.78, "learning_rate": 8.216968556418098e-05, "loss": 2.9989, "step": 22739 }, { "epoch": 1.78, "learning_rate": 8.216890143495648e-05, "loss": 2.9755, "step": 22740 }, { "epoch": 1.78, "learning_rate": 8.216811730573199e-05, "loss": 3.0346, "step": 22741 }, { "epoch": 1.78, "learning_rate": 8.216733317650749e-05, "loss": 2.7573, "step": 22742 }, { "epoch": 1.78, "learning_rate": 8.216654904728299e-05, "loss": 2.7444, "step": 22743 }, { "epoch": 1.78, "learning_rate": 8.21657649180585e-05, "loss": 2.9402, "step": 22744 }, { "epoch": 1.78, "learning_rate": 8.2164980788834e-05, "loss": 2.8883, "step": 22745 }, { "epoch": 1.78, "learning_rate": 8.21641966596095e-05, "loss": 2.8816, "step": 22746 }, { "epoch": 1.78, "learning_rate": 8.216341253038501e-05, "loss": 2.836, "step": 22747 }, { "epoch": 1.78, "learning_rate": 8.216262840116051e-05, "loss": 3.0395, "step": 22748 }, { "epoch": 1.78, "learning_rate": 8.216184427193601e-05, "loss": 2.9381, "step": 22749 }, { "epoch": 1.78, "learning_rate": 8.216106014271152e-05, "loss": 2.9456, "step": 22750 }, { "epoch": 1.78, "learning_rate": 8.216027601348702e-05, "loss": 2.9423, "step": 22751 }, { "epoch": 1.78, "learning_rate": 8.215949188426253e-05, "loss": 2.9348, "step": 22752 }, { "epoch": 1.78, "learning_rate": 8.215870775503803e-05, "loss": 2.8768, "step": 22753 }, { "epoch": 1.78, "learning_rate": 8.215792362581353e-05, "loss": 2.8491, "step": 22754 }, { "epoch": 1.78, "learning_rate": 8.215713949658904e-05, "loss": 2.9073, "step": 22755 }, { "epoch": 1.78, "learning_rate": 8.215635536736454e-05, "loss": 2.8503, "step": 22756 }, { "epoch": 1.78, "learning_rate": 8.215557123814004e-05, "loss": 2.7709, "step": 22757 }, { "epoch": 1.78, "learning_rate": 8.215478710891555e-05, "loss": 3.1214, "step": 22758 }, { "epoch": 1.78, "learning_rate": 8.215400297969105e-05, "loss": 3.0495, "step": 22759 }, { "epoch": 1.78, "learning_rate": 8.215321885046657e-05, "loss": 2.9187, "step": 22760 }, { "epoch": 1.78, "learning_rate": 8.215243472124207e-05, "loss": 2.9855, "step": 22761 }, { "epoch": 1.78, "learning_rate": 8.215165059201758e-05, "loss": 2.6939, "step": 22762 }, { "epoch": 1.78, "learning_rate": 8.215086646279308e-05, "loss": 3.1641, "step": 22763 }, { "epoch": 1.78, "learning_rate": 8.215008233356858e-05, "loss": 2.9091, "step": 22764 }, { "epoch": 1.79, "learning_rate": 8.214929820434409e-05, "loss": 2.9183, "step": 22765 }, { "epoch": 1.79, "learning_rate": 8.214851407511959e-05, "loss": 3.13, "step": 22766 }, { "epoch": 1.79, "learning_rate": 8.21477299458951e-05, "loss": 3.1411, "step": 22767 }, { "epoch": 1.79, "learning_rate": 8.21469458166706e-05, "loss": 2.7145, "step": 22768 }, { "epoch": 1.79, "learning_rate": 8.21461616874461e-05, "loss": 2.7141, "step": 22769 }, { "epoch": 1.79, "learning_rate": 8.21453775582216e-05, "loss": 2.8461, "step": 22770 }, { "epoch": 1.79, "learning_rate": 8.214459342899711e-05, "loss": 2.8936, "step": 22771 }, { "epoch": 1.79, "learning_rate": 8.214380929977261e-05, "loss": 2.9261, "step": 22772 }, { "epoch": 1.79, "learning_rate": 8.214302517054812e-05, "loss": 2.9964, "step": 22773 }, { "epoch": 1.79, "learning_rate": 8.21422410413236e-05, "loss": 2.6713, "step": 22774 }, { "epoch": 1.79, "learning_rate": 8.214145691209911e-05, "loss": 2.871, "step": 22775 }, { "epoch": 1.79, "learning_rate": 8.214067278287461e-05, "loss": 2.7206, "step": 22776 }, { "epoch": 1.79, "learning_rate": 8.213988865365012e-05, "loss": 2.806, "step": 22777 }, { "epoch": 1.79, "learning_rate": 8.213910452442562e-05, "loss": 2.9831, "step": 22778 }, { "epoch": 1.79, "learning_rate": 8.213832039520113e-05, "loss": 2.7217, "step": 22779 }, { "epoch": 1.79, "learning_rate": 8.213753626597663e-05, "loss": 2.9261, "step": 22780 }, { "epoch": 1.79, "learning_rate": 8.213675213675213e-05, "loss": 2.8412, "step": 22781 }, { "epoch": 1.79, "learning_rate": 8.213596800752764e-05, "loss": 2.9207, "step": 22782 }, { "epoch": 1.79, "learning_rate": 8.213518387830314e-05, "loss": 2.8442, "step": 22783 }, { "epoch": 1.79, "learning_rate": 8.213439974907864e-05, "loss": 2.99, "step": 22784 }, { "epoch": 1.79, "learning_rate": 8.213361561985416e-05, "loss": 3.0311, "step": 22785 }, { "epoch": 1.79, "learning_rate": 8.213283149062966e-05, "loss": 2.9276, "step": 22786 }, { "epoch": 1.79, "learning_rate": 8.213204736140517e-05, "loss": 2.9054, "step": 22787 }, { "epoch": 1.79, "learning_rate": 8.213126323218067e-05, "loss": 3.1757, "step": 22788 }, { "epoch": 1.79, "learning_rate": 8.213047910295618e-05, "loss": 2.8013, "step": 22789 }, { "epoch": 1.79, "learning_rate": 8.212969497373168e-05, "loss": 2.8219, "step": 22790 }, { "epoch": 1.79, "learning_rate": 8.212891084450718e-05, "loss": 2.7373, "step": 22791 }, { "epoch": 1.79, "learning_rate": 8.212812671528269e-05, "loss": 3.261, "step": 22792 }, { "epoch": 1.79, "learning_rate": 8.212734258605819e-05, "loss": 2.8955, "step": 22793 }, { "epoch": 1.79, "learning_rate": 8.21265584568337e-05, "loss": 2.9112, "step": 22794 }, { "epoch": 1.79, "learning_rate": 8.21257743276092e-05, "loss": 3.0844, "step": 22795 }, { "epoch": 1.79, "learning_rate": 8.21249901983847e-05, "loss": 2.8834, "step": 22796 }, { "epoch": 1.79, "learning_rate": 8.21242060691602e-05, "loss": 2.887, "step": 22797 }, { "epoch": 1.79, "learning_rate": 8.212342193993571e-05, "loss": 2.907, "step": 22798 }, { "epoch": 1.79, "learning_rate": 8.212263781071121e-05, "loss": 2.8648, "step": 22799 }, { "epoch": 1.79, "learning_rate": 8.212185368148672e-05, "loss": 2.7167, "step": 22800 }, { "epoch": 1.79, "learning_rate": 8.212106955226222e-05, "loss": 2.9042, "step": 22801 }, { "epoch": 1.79, "learning_rate": 8.212028542303772e-05, "loss": 2.875, "step": 22802 }, { "epoch": 1.79, "learning_rate": 8.211950129381323e-05, "loss": 2.9099, "step": 22803 }, { "epoch": 1.79, "learning_rate": 8.211871716458873e-05, "loss": 2.9158, "step": 22804 }, { "epoch": 1.79, "learning_rate": 8.211793303536423e-05, "loss": 2.8171, "step": 22805 }, { "epoch": 1.79, "learning_rate": 8.211714890613974e-05, "loss": 2.9016, "step": 22806 }, { "epoch": 1.79, "learning_rate": 8.211636477691524e-05, "loss": 2.911, "step": 22807 }, { "epoch": 1.79, "learning_rate": 8.211558064769075e-05, "loss": 2.7729, "step": 22808 }, { "epoch": 1.79, "learning_rate": 8.211479651846625e-05, "loss": 3.1095, "step": 22809 }, { "epoch": 1.79, "learning_rate": 8.211401238924174e-05, "loss": 2.8329, "step": 22810 }, { "epoch": 1.79, "learning_rate": 8.211322826001726e-05, "loss": 2.7752, "step": 22811 }, { "epoch": 1.79, "learning_rate": 8.211244413079276e-05, "loss": 2.937, "step": 22812 }, { "epoch": 1.79, "learning_rate": 8.211166000156826e-05, "loss": 3.0663, "step": 22813 }, { "epoch": 1.79, "learning_rate": 8.211087587234377e-05, "loss": 2.8763, "step": 22814 }, { "epoch": 1.79, "learning_rate": 8.211009174311927e-05, "loss": 2.7864, "step": 22815 }, { "epoch": 1.79, "learning_rate": 8.210930761389477e-05, "loss": 2.852, "step": 22816 }, { "epoch": 1.79, "learning_rate": 8.210852348467028e-05, "loss": 2.6514, "step": 22817 }, { "epoch": 1.79, "learning_rate": 8.210773935544578e-05, "loss": 2.9207, "step": 22818 }, { "epoch": 1.79, "learning_rate": 8.210695522622129e-05, "loss": 2.8876, "step": 22819 }, { "epoch": 1.79, "learning_rate": 8.210617109699679e-05, "loss": 2.8004, "step": 22820 }, { "epoch": 1.79, "learning_rate": 8.210538696777229e-05, "loss": 3.0565, "step": 22821 }, { "epoch": 1.79, "learning_rate": 8.21046028385478e-05, "loss": 2.8588, "step": 22822 }, { "epoch": 1.79, "learning_rate": 8.21038187093233e-05, "loss": 3.0139, "step": 22823 }, { "epoch": 1.79, "learning_rate": 8.21030345800988e-05, "loss": 2.9642, "step": 22824 }, { "epoch": 1.79, "learning_rate": 8.210225045087431e-05, "loss": 2.9208, "step": 22825 }, { "epoch": 1.79, "learning_rate": 8.210146632164981e-05, "loss": 2.921, "step": 22826 }, { "epoch": 1.79, "learning_rate": 8.210068219242532e-05, "loss": 2.8756, "step": 22827 }, { "epoch": 1.79, "learning_rate": 8.209989806320082e-05, "loss": 2.8021, "step": 22828 }, { "epoch": 1.79, "learning_rate": 8.209911393397632e-05, "loss": 2.7244, "step": 22829 }, { "epoch": 1.79, "learning_rate": 8.209832980475183e-05, "loss": 2.7762, "step": 22830 }, { "epoch": 1.79, "learning_rate": 8.209754567552733e-05, "loss": 2.9601, "step": 22831 }, { "epoch": 1.79, "learning_rate": 8.209676154630283e-05, "loss": 2.8759, "step": 22832 }, { "epoch": 1.79, "learning_rate": 8.209597741707834e-05, "loss": 3.0511, "step": 22833 }, { "epoch": 1.79, "learning_rate": 8.209519328785384e-05, "loss": 3.0108, "step": 22834 }, { "epoch": 1.79, "learning_rate": 8.209440915862934e-05, "loss": 2.9696, "step": 22835 }, { "epoch": 1.79, "learning_rate": 8.209362502940485e-05, "loss": 2.728, "step": 22836 }, { "epoch": 1.79, "learning_rate": 8.209284090018035e-05, "loss": 2.8754, "step": 22837 }, { "epoch": 1.79, "learning_rate": 8.209205677095586e-05, "loss": 3.1675, "step": 22838 }, { "epoch": 1.79, "learning_rate": 8.209127264173136e-05, "loss": 2.9807, "step": 22839 }, { "epoch": 1.79, "learning_rate": 8.209048851250686e-05, "loss": 2.9064, "step": 22840 }, { "epoch": 1.79, "learning_rate": 8.208970438328237e-05, "loss": 2.9909, "step": 22841 }, { "epoch": 1.79, "learning_rate": 8.208892025405788e-05, "loss": 2.9634, "step": 22842 }, { "epoch": 1.79, "learning_rate": 8.208813612483339e-05, "loss": 2.9861, "step": 22843 }, { "epoch": 1.79, "learning_rate": 8.208735199560889e-05, "loss": 2.9452, "step": 22844 }, { "epoch": 1.79, "learning_rate": 8.20865678663844e-05, "loss": 2.8694, "step": 22845 }, { "epoch": 1.79, "learning_rate": 8.208578373715989e-05, "loss": 2.6274, "step": 22846 }, { "epoch": 1.79, "learning_rate": 8.208499960793539e-05, "loss": 2.9766, "step": 22847 }, { "epoch": 1.79, "learning_rate": 8.208421547871089e-05, "loss": 2.9477, "step": 22848 }, { "epoch": 1.79, "learning_rate": 8.20834313494864e-05, "loss": 2.9209, "step": 22849 }, { "epoch": 1.79, "learning_rate": 8.20826472202619e-05, "loss": 2.8742, "step": 22850 }, { "epoch": 1.79, "learning_rate": 8.20818630910374e-05, "loss": 2.6845, "step": 22851 }, { "epoch": 1.79, "learning_rate": 8.208107896181291e-05, "loss": 2.8563, "step": 22852 }, { "epoch": 1.79, "learning_rate": 8.208029483258841e-05, "loss": 2.9212, "step": 22853 }, { "epoch": 1.79, "learning_rate": 8.207951070336391e-05, "loss": 3.0199, "step": 22854 }, { "epoch": 1.79, "learning_rate": 8.207872657413942e-05, "loss": 2.6855, "step": 22855 }, { "epoch": 1.79, "learning_rate": 8.207794244491492e-05, "loss": 3.0932, "step": 22856 }, { "epoch": 1.79, "learning_rate": 8.207715831569043e-05, "loss": 2.87, "step": 22857 }, { "epoch": 1.79, "learning_rate": 8.207637418646593e-05, "loss": 2.8981, "step": 22858 }, { "epoch": 1.79, "learning_rate": 8.207559005724143e-05, "loss": 2.8381, "step": 22859 }, { "epoch": 1.79, "learning_rate": 8.207480592801694e-05, "loss": 2.9214, "step": 22860 }, { "epoch": 1.79, "learning_rate": 8.207402179879244e-05, "loss": 2.892, "step": 22861 }, { "epoch": 1.79, "learning_rate": 8.207323766956794e-05, "loss": 2.9336, "step": 22862 }, { "epoch": 1.79, "learning_rate": 8.207245354034345e-05, "loss": 2.7567, "step": 22863 }, { "epoch": 1.79, "learning_rate": 8.207166941111895e-05, "loss": 2.8774, "step": 22864 }, { "epoch": 1.79, "learning_rate": 8.207088528189445e-05, "loss": 2.847, "step": 22865 }, { "epoch": 1.79, "learning_rate": 8.207010115266996e-05, "loss": 3.3144, "step": 22866 }, { "epoch": 1.79, "learning_rate": 8.206931702344546e-05, "loss": 3.0402, "step": 22867 }, { "epoch": 1.79, "learning_rate": 8.206853289422098e-05, "loss": 3.0794, "step": 22868 }, { "epoch": 1.79, "learning_rate": 8.206774876499648e-05, "loss": 3.0304, "step": 22869 }, { "epoch": 1.79, "learning_rate": 8.206696463577199e-05, "loss": 2.7439, "step": 22870 }, { "epoch": 1.79, "learning_rate": 8.206618050654749e-05, "loss": 2.6747, "step": 22871 }, { "epoch": 1.79, "learning_rate": 8.2065396377323e-05, "loss": 2.8859, "step": 22872 }, { "epoch": 1.79, "learning_rate": 8.20646122480985e-05, "loss": 2.8714, "step": 22873 }, { "epoch": 1.79, "learning_rate": 8.2063828118874e-05, "loss": 2.9773, "step": 22874 }, { "epoch": 1.79, "learning_rate": 8.20630439896495e-05, "loss": 2.8278, "step": 22875 }, { "epoch": 1.79, "learning_rate": 8.206225986042501e-05, "loss": 2.8461, "step": 22876 }, { "epoch": 1.79, "learning_rate": 8.206147573120051e-05, "loss": 2.9674, "step": 22877 }, { "epoch": 1.79, "learning_rate": 8.206069160197602e-05, "loss": 3.0481, "step": 22878 }, { "epoch": 1.79, "learning_rate": 8.205990747275152e-05, "loss": 3.0274, "step": 22879 }, { "epoch": 1.79, "learning_rate": 8.205912334352702e-05, "loss": 2.6515, "step": 22880 }, { "epoch": 1.79, "learning_rate": 8.205833921430253e-05, "loss": 2.9117, "step": 22881 }, { "epoch": 1.79, "learning_rate": 8.205755508507802e-05, "loss": 2.8706, "step": 22882 }, { "epoch": 1.79, "learning_rate": 8.205677095585352e-05, "loss": 2.7982, "step": 22883 }, { "epoch": 1.79, "learning_rate": 8.205598682662902e-05, "loss": 3.1948, "step": 22884 }, { "epoch": 1.79, "learning_rate": 8.205520269740453e-05, "loss": 2.9612, "step": 22885 }, { "epoch": 1.79, "learning_rate": 8.205441856818003e-05, "loss": 2.9509, "step": 22886 }, { "epoch": 1.79, "learning_rate": 8.205363443895554e-05, "loss": 3.0187, "step": 22887 }, { "epoch": 1.79, "learning_rate": 8.205285030973104e-05, "loss": 2.7367, "step": 22888 }, { "epoch": 1.79, "learning_rate": 8.205206618050654e-05, "loss": 2.8218, "step": 22889 }, { "epoch": 1.79, "learning_rate": 8.205128205128205e-05, "loss": 3.1032, "step": 22890 }, { "epoch": 1.79, "learning_rate": 8.205049792205755e-05, "loss": 3.0204, "step": 22891 }, { "epoch": 1.8, "learning_rate": 8.204971379283305e-05, "loss": 2.9341, "step": 22892 }, { "epoch": 1.8, "learning_rate": 8.204892966360857e-05, "loss": 2.985, "step": 22893 }, { "epoch": 1.8, "learning_rate": 8.204814553438408e-05, "loss": 2.9442, "step": 22894 }, { "epoch": 1.8, "learning_rate": 8.204736140515958e-05, "loss": 2.8483, "step": 22895 }, { "epoch": 1.8, "learning_rate": 8.204657727593508e-05, "loss": 2.9792, "step": 22896 }, { "epoch": 1.8, "learning_rate": 8.204579314671059e-05, "loss": 2.8137, "step": 22897 }, { "epoch": 1.8, "learning_rate": 8.204500901748609e-05, "loss": 2.9841, "step": 22898 }, { "epoch": 1.8, "learning_rate": 8.20442248882616e-05, "loss": 2.9551, "step": 22899 }, { "epoch": 1.8, "learning_rate": 8.20434407590371e-05, "loss": 2.9537, "step": 22900 }, { "epoch": 1.8, "learning_rate": 8.20426566298126e-05, "loss": 2.7809, "step": 22901 }, { "epoch": 1.8, "learning_rate": 8.20418725005881e-05, "loss": 2.8605, "step": 22902 }, { "epoch": 1.8, "learning_rate": 8.204108837136361e-05, "loss": 2.7774, "step": 22903 }, { "epoch": 1.8, "learning_rate": 8.204030424213911e-05, "loss": 2.9898, "step": 22904 }, { "epoch": 1.8, "learning_rate": 8.203952011291462e-05, "loss": 3.0913, "step": 22905 }, { "epoch": 1.8, "learning_rate": 8.203873598369012e-05, "loss": 2.9713, "step": 22906 }, { "epoch": 1.8, "learning_rate": 8.203795185446562e-05, "loss": 2.9323, "step": 22907 }, { "epoch": 1.8, "learning_rate": 8.203716772524113e-05, "loss": 2.8278, "step": 22908 }, { "epoch": 1.8, "learning_rate": 8.203638359601663e-05, "loss": 2.8725, "step": 22909 }, { "epoch": 1.8, "learning_rate": 8.203559946679213e-05, "loss": 3.1181, "step": 22910 }, { "epoch": 1.8, "learning_rate": 8.203481533756764e-05, "loss": 3.0262, "step": 22911 }, { "epoch": 1.8, "learning_rate": 8.203403120834314e-05, "loss": 2.8859, "step": 22912 }, { "epoch": 1.8, "learning_rate": 8.203324707911865e-05, "loss": 2.8246, "step": 22913 }, { "epoch": 1.8, "learning_rate": 8.203246294989415e-05, "loss": 3.0092, "step": 22914 }, { "epoch": 1.8, "learning_rate": 8.203167882066965e-05, "loss": 2.8715, "step": 22915 }, { "epoch": 1.8, "learning_rate": 8.203089469144516e-05, "loss": 3.0531, "step": 22916 }, { "epoch": 1.8, "learning_rate": 8.203011056222066e-05, "loss": 2.8052, "step": 22917 }, { "epoch": 1.8, "learning_rate": 8.202932643299615e-05, "loss": 2.8389, "step": 22918 }, { "epoch": 1.8, "learning_rate": 8.202854230377167e-05, "loss": 2.9582, "step": 22919 }, { "epoch": 1.8, "learning_rate": 8.202775817454717e-05, "loss": 2.7651, "step": 22920 }, { "epoch": 1.8, "learning_rate": 8.202697404532267e-05, "loss": 3.0634, "step": 22921 }, { "epoch": 1.8, "learning_rate": 8.202618991609818e-05, "loss": 2.8527, "step": 22922 }, { "epoch": 1.8, "learning_rate": 8.202540578687368e-05, "loss": 3.1129, "step": 22923 }, { "epoch": 1.8, "learning_rate": 8.202462165764919e-05, "loss": 2.7578, "step": 22924 }, { "epoch": 1.8, "learning_rate": 8.202383752842469e-05, "loss": 2.7377, "step": 22925 }, { "epoch": 1.8, "learning_rate": 8.202305339920019e-05, "loss": 3.2487, "step": 22926 }, { "epoch": 1.8, "learning_rate": 8.20222692699757e-05, "loss": 3.0572, "step": 22927 }, { "epoch": 1.8, "learning_rate": 8.20214851407512e-05, "loss": 3.2445, "step": 22928 }, { "epoch": 1.8, "learning_rate": 8.20207010115267e-05, "loss": 2.7955, "step": 22929 }, { "epoch": 1.8, "learning_rate": 8.201991688230221e-05, "loss": 2.8931, "step": 22930 }, { "epoch": 1.8, "learning_rate": 8.201913275307771e-05, "loss": 2.7396, "step": 22931 }, { "epoch": 1.8, "learning_rate": 8.201834862385321e-05, "loss": 2.9651, "step": 22932 }, { "epoch": 1.8, "learning_rate": 8.201756449462872e-05, "loss": 2.7271, "step": 22933 }, { "epoch": 1.8, "learning_rate": 8.201678036540422e-05, "loss": 2.8385, "step": 22934 }, { "epoch": 1.8, "learning_rate": 8.201599623617973e-05, "loss": 2.8281, "step": 22935 }, { "epoch": 1.8, "learning_rate": 8.201521210695523e-05, "loss": 2.7668, "step": 22936 }, { "epoch": 1.8, "learning_rate": 8.201442797773073e-05, "loss": 2.9615, "step": 22937 }, { "epoch": 1.8, "learning_rate": 8.201364384850624e-05, "loss": 2.8882, "step": 22938 }, { "epoch": 1.8, "learning_rate": 8.201285971928174e-05, "loss": 2.8412, "step": 22939 }, { "epoch": 1.8, "learning_rate": 8.201207559005724e-05, "loss": 3.1264, "step": 22940 }, { "epoch": 1.8, "learning_rate": 8.201129146083275e-05, "loss": 2.7748, "step": 22941 }, { "epoch": 1.8, "learning_rate": 8.201050733160825e-05, "loss": 2.9293, "step": 22942 }, { "epoch": 1.8, "learning_rate": 8.200972320238376e-05, "loss": 3.1838, "step": 22943 }, { "epoch": 1.8, "learning_rate": 8.200893907315926e-05, "loss": 2.8064, "step": 22944 }, { "epoch": 1.8, "learning_rate": 8.200815494393476e-05, "loss": 2.9377, "step": 22945 }, { "epoch": 1.8, "learning_rate": 8.200737081471027e-05, "loss": 2.8987, "step": 22946 }, { "epoch": 1.8, "learning_rate": 8.200658668548577e-05, "loss": 2.7546, "step": 22947 }, { "epoch": 1.8, "learning_rate": 8.200580255626127e-05, "loss": 2.5135, "step": 22948 }, { "epoch": 1.8, "learning_rate": 8.200501842703678e-05, "loss": 3.0259, "step": 22949 }, { "epoch": 1.8, "learning_rate": 8.20042342978123e-05, "loss": 2.8429, "step": 22950 }, { "epoch": 1.8, "learning_rate": 8.20034501685878e-05, "loss": 2.9502, "step": 22951 }, { "epoch": 1.8, "learning_rate": 8.20026660393633e-05, "loss": 2.8867, "step": 22952 }, { "epoch": 1.8, "learning_rate": 8.20018819101388e-05, "loss": 2.8011, "step": 22953 }, { "epoch": 1.8, "learning_rate": 8.20010977809143e-05, "loss": 3.016, "step": 22954 }, { "epoch": 1.8, "learning_rate": 8.20003136516898e-05, "loss": 3.1005, "step": 22955 }, { "epoch": 1.8, "learning_rate": 8.19995295224653e-05, "loss": 2.8604, "step": 22956 }, { "epoch": 1.8, "learning_rate": 8.199874539324081e-05, "loss": 3.1952, "step": 22957 }, { "epoch": 1.8, "learning_rate": 8.199796126401631e-05, "loss": 2.8446, "step": 22958 }, { "epoch": 1.8, "learning_rate": 8.199717713479181e-05, "loss": 2.9453, "step": 22959 }, { "epoch": 1.8, "learning_rate": 8.199639300556732e-05, "loss": 2.8453, "step": 22960 }, { "epoch": 1.8, "learning_rate": 8.199560887634282e-05, "loss": 2.8682, "step": 22961 }, { "epoch": 1.8, "learning_rate": 8.199482474711833e-05, "loss": 2.9564, "step": 22962 }, { "epoch": 1.8, "learning_rate": 8.199404061789383e-05, "loss": 2.8935, "step": 22963 }, { "epoch": 1.8, "learning_rate": 8.199325648866933e-05, "loss": 2.8849, "step": 22964 }, { "epoch": 1.8, "learning_rate": 8.199247235944484e-05, "loss": 2.948, "step": 22965 }, { "epoch": 1.8, "learning_rate": 8.199168823022034e-05, "loss": 2.8055, "step": 22966 }, { "epoch": 1.8, "learning_rate": 8.199090410099584e-05, "loss": 2.945, "step": 22967 }, { "epoch": 1.8, "learning_rate": 8.199011997177135e-05, "loss": 2.7505, "step": 22968 }, { "epoch": 1.8, "learning_rate": 8.198933584254685e-05, "loss": 2.8115, "step": 22969 }, { "epoch": 1.8, "learning_rate": 8.198855171332235e-05, "loss": 2.7735, "step": 22970 }, { "epoch": 1.8, "learning_rate": 8.198776758409786e-05, "loss": 2.9902, "step": 22971 }, { "epoch": 1.8, "learning_rate": 8.198698345487336e-05, "loss": 2.7096, "step": 22972 }, { "epoch": 1.8, "learning_rate": 8.198619932564887e-05, "loss": 3.0049, "step": 22973 }, { "epoch": 1.8, "learning_rate": 8.198541519642437e-05, "loss": 2.9899, "step": 22974 }, { "epoch": 1.8, "learning_rate": 8.198463106719987e-05, "loss": 2.8952, "step": 22975 }, { "epoch": 1.8, "learning_rate": 8.198384693797539e-05, "loss": 2.8736, "step": 22976 }, { "epoch": 1.8, "learning_rate": 8.19830628087509e-05, "loss": 3.2659, "step": 22977 }, { "epoch": 1.8, "learning_rate": 8.19822786795264e-05, "loss": 2.8148, "step": 22978 }, { "epoch": 1.8, "learning_rate": 8.19814945503019e-05, "loss": 3.0151, "step": 22979 }, { "epoch": 1.8, "learning_rate": 8.19807104210774e-05, "loss": 2.7736, "step": 22980 }, { "epoch": 1.8, "learning_rate": 8.197992629185291e-05, "loss": 2.9611, "step": 22981 }, { "epoch": 1.8, "learning_rate": 8.197914216262841e-05, "loss": 2.9635, "step": 22982 }, { "epoch": 1.8, "learning_rate": 8.197835803340392e-05, "loss": 3.2097, "step": 22983 }, { "epoch": 1.8, "learning_rate": 8.197757390417942e-05, "loss": 3.0722, "step": 22984 }, { "epoch": 1.8, "learning_rate": 8.197678977495492e-05, "loss": 2.7802, "step": 22985 }, { "epoch": 1.8, "learning_rate": 8.197600564573043e-05, "loss": 2.9684, "step": 22986 }, { "epoch": 1.8, "learning_rate": 8.197522151650593e-05, "loss": 2.7825, "step": 22987 }, { "epoch": 1.8, "learning_rate": 8.197443738728143e-05, "loss": 2.9647, "step": 22988 }, { "epoch": 1.8, "learning_rate": 8.197365325805694e-05, "loss": 2.7543, "step": 22989 }, { "epoch": 1.8, "learning_rate": 8.197286912883243e-05, "loss": 3.1532, "step": 22990 }, { "epoch": 1.8, "learning_rate": 8.197208499960793e-05, "loss": 2.9066, "step": 22991 }, { "epoch": 1.8, "learning_rate": 8.197130087038344e-05, "loss": 2.8424, "step": 22992 }, { "epoch": 1.8, "learning_rate": 8.197051674115894e-05, "loss": 2.8681, "step": 22993 }, { "epoch": 1.8, "learning_rate": 8.196973261193444e-05, "loss": 2.7493, "step": 22994 }, { "epoch": 1.8, "learning_rate": 8.196894848270995e-05, "loss": 2.9244, "step": 22995 }, { "epoch": 1.8, "learning_rate": 8.196816435348545e-05, "loss": 2.9127, "step": 22996 }, { "epoch": 1.8, "learning_rate": 8.196738022426095e-05, "loss": 2.7637, "step": 22997 }, { "epoch": 1.8, "learning_rate": 8.196659609503646e-05, "loss": 2.8728, "step": 22998 }, { "epoch": 1.8, "learning_rate": 8.196581196581196e-05, "loss": 2.9458, "step": 22999 }, { "epoch": 1.8, "learning_rate": 8.196502783658746e-05, "loss": 2.7359, "step": 23000 }, { "epoch": 1.8, "eval_loss": 2.9029297828674316, "eval_runtime": 218.0054, "eval_samples_per_second": 51.98, "eval_steps_per_second": 3.252, "step": 23000 }, { "epoch": 1.8, "learning_rate": 8.196424370736297e-05, "loss": 3.0999, "step": 23001 }, { "epoch": 1.8, "learning_rate": 8.196345957813849e-05, "loss": 3.0297, "step": 23002 }, { "epoch": 1.8, "learning_rate": 8.196267544891399e-05, "loss": 2.8997, "step": 23003 }, { "epoch": 1.8, "learning_rate": 8.19618913196895e-05, "loss": 2.8029, "step": 23004 }, { "epoch": 1.8, "learning_rate": 8.1961107190465e-05, "loss": 2.8908, "step": 23005 }, { "epoch": 1.8, "learning_rate": 8.19603230612405e-05, "loss": 2.76, "step": 23006 }, { "epoch": 1.8, "learning_rate": 8.1959538932016e-05, "loss": 2.8505, "step": 23007 }, { "epoch": 1.8, "learning_rate": 8.195875480279151e-05, "loss": 2.7866, "step": 23008 }, { "epoch": 1.8, "learning_rate": 8.195797067356701e-05, "loss": 2.7673, "step": 23009 }, { "epoch": 1.8, "learning_rate": 8.195718654434252e-05, "loss": 3.0678, "step": 23010 }, { "epoch": 1.8, "learning_rate": 8.195640241511802e-05, "loss": 3.0478, "step": 23011 }, { "epoch": 1.8, "learning_rate": 8.195561828589352e-05, "loss": 2.7972, "step": 23012 }, { "epoch": 1.8, "learning_rate": 8.195483415666903e-05, "loss": 2.6958, "step": 23013 }, { "epoch": 1.8, "learning_rate": 8.195405002744453e-05, "loss": 2.9042, "step": 23014 }, { "epoch": 1.8, "learning_rate": 8.195326589822003e-05, "loss": 2.9659, "step": 23015 }, { "epoch": 1.8, "learning_rate": 8.195248176899554e-05, "loss": 2.8369, "step": 23016 }, { "epoch": 1.8, "learning_rate": 8.195169763977104e-05, "loss": 2.8912, "step": 23017 }, { "epoch": 1.8, "learning_rate": 8.195091351054654e-05, "loss": 2.9855, "step": 23018 }, { "epoch": 1.8, "learning_rate": 8.195012938132205e-05, "loss": 2.9831, "step": 23019 }, { "epoch": 1.81, "learning_rate": 8.194934525209755e-05, "loss": 2.834, "step": 23020 }, { "epoch": 1.81, "learning_rate": 8.194856112287306e-05, "loss": 2.9772, "step": 23021 }, { "epoch": 1.81, "learning_rate": 8.194777699364856e-05, "loss": 2.6897, "step": 23022 }, { "epoch": 1.81, "learning_rate": 8.194699286442406e-05, "loss": 2.8215, "step": 23023 }, { "epoch": 1.81, "learning_rate": 8.194620873519957e-05, "loss": 2.8658, "step": 23024 }, { "epoch": 1.81, "learning_rate": 8.194542460597507e-05, "loss": 3.0874, "step": 23025 }, { "epoch": 1.81, "learning_rate": 8.194464047675056e-05, "loss": 2.7092, "step": 23026 }, { "epoch": 1.81, "learning_rate": 8.194385634752608e-05, "loss": 3.0283, "step": 23027 }, { "epoch": 1.81, "learning_rate": 8.194307221830158e-05, "loss": 2.8911, "step": 23028 }, { "epoch": 1.81, "learning_rate": 8.194228808907709e-05, "loss": 2.8228, "step": 23029 }, { "epoch": 1.81, "learning_rate": 8.194150395985259e-05, "loss": 2.7521, "step": 23030 }, { "epoch": 1.81, "learning_rate": 8.194071983062809e-05, "loss": 2.7808, "step": 23031 }, { "epoch": 1.81, "learning_rate": 8.19399357014036e-05, "loss": 2.9214, "step": 23032 }, { "epoch": 1.81, "learning_rate": 8.19391515721791e-05, "loss": 2.8971, "step": 23033 }, { "epoch": 1.81, "learning_rate": 8.19383674429546e-05, "loss": 2.6727, "step": 23034 }, { "epoch": 1.81, "learning_rate": 8.193758331373011e-05, "loss": 2.8678, "step": 23035 }, { "epoch": 1.81, "learning_rate": 8.193679918450561e-05, "loss": 3.1408, "step": 23036 }, { "epoch": 1.81, "learning_rate": 8.193601505528111e-05, "loss": 2.7326, "step": 23037 }, { "epoch": 1.81, "learning_rate": 8.193523092605662e-05, "loss": 2.9582, "step": 23038 }, { "epoch": 1.81, "learning_rate": 8.193444679683212e-05, "loss": 2.7817, "step": 23039 }, { "epoch": 1.81, "learning_rate": 8.193366266760763e-05, "loss": 2.7607, "step": 23040 }, { "epoch": 1.81, "learning_rate": 8.193287853838313e-05, "loss": 2.9456, "step": 23041 }, { "epoch": 1.81, "learning_rate": 8.193209440915863e-05, "loss": 2.9399, "step": 23042 }, { "epoch": 1.81, "learning_rate": 8.193131027993414e-05, "loss": 2.7657, "step": 23043 }, { "epoch": 1.81, "learning_rate": 8.193052615070964e-05, "loss": 2.8473, "step": 23044 }, { "epoch": 1.81, "learning_rate": 8.192974202148514e-05, "loss": 2.9217, "step": 23045 }, { "epoch": 1.81, "learning_rate": 8.192895789226065e-05, "loss": 2.9079, "step": 23046 }, { "epoch": 1.81, "learning_rate": 8.192817376303615e-05, "loss": 2.9455, "step": 23047 }, { "epoch": 1.81, "learning_rate": 8.192738963381166e-05, "loss": 2.9176, "step": 23048 }, { "epoch": 1.81, "learning_rate": 8.192660550458716e-05, "loss": 2.8824, "step": 23049 }, { "epoch": 1.81, "learning_rate": 8.192582137536266e-05, "loss": 3.075, "step": 23050 }, { "epoch": 1.81, "learning_rate": 8.192503724613817e-05, "loss": 2.8651, "step": 23051 }, { "epoch": 1.81, "learning_rate": 8.192425311691367e-05, "loss": 2.9359, "step": 23052 }, { "epoch": 1.81, "learning_rate": 8.192346898768917e-05, "loss": 2.7909, "step": 23053 }, { "epoch": 1.81, "learning_rate": 8.192268485846468e-05, "loss": 2.9628, "step": 23054 }, { "epoch": 1.81, "learning_rate": 8.192190072924018e-05, "loss": 2.9796, "step": 23055 }, { "epoch": 1.81, "learning_rate": 8.192111660001568e-05, "loss": 2.8966, "step": 23056 }, { "epoch": 1.81, "learning_rate": 8.192033247079119e-05, "loss": 2.8283, "step": 23057 }, { "epoch": 1.81, "learning_rate": 8.19195483415667e-05, "loss": 3.0406, "step": 23058 }, { "epoch": 1.81, "learning_rate": 8.191876421234221e-05, "loss": 2.9577, "step": 23059 }, { "epoch": 1.81, "learning_rate": 8.191798008311771e-05, "loss": 3.1145, "step": 23060 }, { "epoch": 1.81, "learning_rate": 8.191719595389322e-05, "loss": 2.8498, "step": 23061 }, { "epoch": 1.81, "learning_rate": 8.19164118246687e-05, "loss": 3.1001, "step": 23062 }, { "epoch": 1.81, "learning_rate": 8.191562769544421e-05, "loss": 2.8378, "step": 23063 }, { "epoch": 1.81, "learning_rate": 8.191484356621971e-05, "loss": 2.9821, "step": 23064 }, { "epoch": 1.81, "learning_rate": 8.191405943699522e-05, "loss": 2.8594, "step": 23065 }, { "epoch": 1.81, "learning_rate": 8.191327530777072e-05, "loss": 2.9032, "step": 23066 }, { "epoch": 1.81, "learning_rate": 8.191249117854622e-05, "loss": 3.0514, "step": 23067 }, { "epoch": 1.81, "learning_rate": 8.191170704932173e-05, "loss": 2.9197, "step": 23068 }, { "epoch": 1.81, "learning_rate": 8.191092292009723e-05, "loss": 3.0585, "step": 23069 }, { "epoch": 1.81, "learning_rate": 8.191013879087274e-05, "loss": 2.972, "step": 23070 }, { "epoch": 1.81, "learning_rate": 8.190935466164824e-05, "loss": 2.946, "step": 23071 }, { "epoch": 1.81, "learning_rate": 8.190857053242374e-05, "loss": 2.9438, "step": 23072 }, { "epoch": 1.81, "learning_rate": 8.190778640319925e-05, "loss": 3.0018, "step": 23073 }, { "epoch": 1.81, "learning_rate": 8.190700227397475e-05, "loss": 3.0786, "step": 23074 }, { "epoch": 1.81, "learning_rate": 8.190621814475025e-05, "loss": 3.1454, "step": 23075 }, { "epoch": 1.81, "learning_rate": 8.190543401552576e-05, "loss": 2.8412, "step": 23076 }, { "epoch": 1.81, "learning_rate": 8.190464988630126e-05, "loss": 2.7882, "step": 23077 }, { "epoch": 1.81, "learning_rate": 8.190386575707677e-05, "loss": 2.8078, "step": 23078 }, { "epoch": 1.81, "learning_rate": 8.190308162785227e-05, "loss": 3.0219, "step": 23079 }, { "epoch": 1.81, "learning_rate": 8.190229749862777e-05, "loss": 2.9465, "step": 23080 }, { "epoch": 1.81, "learning_rate": 8.190151336940328e-05, "loss": 2.9664, "step": 23081 }, { "epoch": 1.81, "learning_rate": 8.190072924017878e-05, "loss": 2.8416, "step": 23082 }, { "epoch": 1.81, "learning_rate": 8.189994511095428e-05, "loss": 2.9754, "step": 23083 }, { "epoch": 1.81, "learning_rate": 8.18991609817298e-05, "loss": 2.8869, "step": 23084 }, { "epoch": 1.81, "learning_rate": 8.18983768525053e-05, "loss": 2.8721, "step": 23085 }, { "epoch": 1.81, "learning_rate": 8.189759272328081e-05, "loss": 2.8242, "step": 23086 }, { "epoch": 1.81, "learning_rate": 8.189680859405631e-05, "loss": 3.0544, "step": 23087 }, { "epoch": 1.81, "learning_rate": 8.189602446483182e-05, "loss": 2.9602, "step": 23088 }, { "epoch": 1.81, "learning_rate": 8.189524033560732e-05, "loss": 2.8175, "step": 23089 }, { "epoch": 1.81, "learning_rate": 8.189445620638282e-05, "loss": 2.9181, "step": 23090 }, { "epoch": 1.81, "learning_rate": 8.189367207715833e-05, "loss": 2.8554, "step": 23091 }, { "epoch": 1.81, "learning_rate": 8.189288794793383e-05, "loss": 2.6595, "step": 23092 }, { "epoch": 1.81, "learning_rate": 8.189210381870933e-05, "loss": 2.94, "step": 23093 }, { "epoch": 1.81, "learning_rate": 8.189131968948484e-05, "loss": 2.8909, "step": 23094 }, { "epoch": 1.81, "learning_rate": 8.189053556026034e-05, "loss": 2.8958, "step": 23095 }, { "epoch": 1.81, "learning_rate": 8.188975143103585e-05, "loss": 3.0634, "step": 23096 }, { "epoch": 1.81, "learning_rate": 8.188896730181135e-05, "loss": 3.2424, "step": 23097 }, { "epoch": 1.81, "learning_rate": 8.188818317258684e-05, "loss": 2.9191, "step": 23098 }, { "epoch": 1.81, "learning_rate": 8.188739904336234e-05, "loss": 2.914, "step": 23099 }, { "epoch": 1.81, "learning_rate": 8.188661491413785e-05, "loss": 2.9815, "step": 23100 }, { "epoch": 1.81, "learning_rate": 8.188583078491335e-05, "loss": 2.9727, "step": 23101 }, { "epoch": 1.81, "learning_rate": 8.188504665568885e-05, "loss": 3.1718, "step": 23102 }, { "epoch": 1.81, "learning_rate": 8.188426252646436e-05, "loss": 3.0714, "step": 23103 }, { "epoch": 1.81, "learning_rate": 8.188347839723986e-05, "loss": 2.9136, "step": 23104 }, { "epoch": 1.81, "learning_rate": 8.188269426801536e-05, "loss": 2.8447, "step": 23105 }, { "epoch": 1.81, "learning_rate": 8.188191013879087e-05, "loss": 2.9662, "step": 23106 }, { "epoch": 1.81, "learning_rate": 8.188112600956637e-05, "loss": 2.8621, "step": 23107 }, { "epoch": 1.81, "learning_rate": 8.188034188034188e-05, "loss": 2.991, "step": 23108 }, { "epoch": 1.81, "learning_rate": 8.187955775111738e-05, "loss": 3.1426, "step": 23109 }, { "epoch": 1.81, "learning_rate": 8.18787736218929e-05, "loss": 2.8287, "step": 23110 }, { "epoch": 1.81, "learning_rate": 8.18779894926684e-05, "loss": 2.7383, "step": 23111 }, { "epoch": 1.81, "learning_rate": 8.18772053634439e-05, "loss": 2.8466, "step": 23112 }, { "epoch": 1.81, "learning_rate": 8.187642123421941e-05, "loss": 3.139, "step": 23113 }, { "epoch": 1.81, "learning_rate": 8.187563710499491e-05, "loss": 2.7736, "step": 23114 }, { "epoch": 1.81, "learning_rate": 8.187485297577042e-05, "loss": 2.8344, "step": 23115 }, { "epoch": 1.81, "learning_rate": 8.187406884654592e-05, "loss": 3.424, "step": 23116 }, { "epoch": 1.81, "learning_rate": 8.187328471732142e-05, "loss": 2.6308, "step": 23117 }, { "epoch": 1.81, "learning_rate": 8.187250058809693e-05, "loss": 2.9929, "step": 23118 }, { "epoch": 1.81, "learning_rate": 8.187171645887243e-05, "loss": 2.8023, "step": 23119 }, { "epoch": 1.81, "learning_rate": 8.187093232964793e-05, "loss": 3.054, "step": 23120 }, { "epoch": 1.81, "learning_rate": 8.187014820042344e-05, "loss": 2.7572, "step": 23121 }, { "epoch": 1.81, "learning_rate": 8.186936407119894e-05, "loss": 3.0755, "step": 23122 }, { "epoch": 1.81, "learning_rate": 8.186857994197444e-05, "loss": 2.9395, "step": 23123 }, { "epoch": 1.81, "learning_rate": 8.186779581274995e-05, "loss": 2.7474, "step": 23124 }, { "epoch": 1.81, "learning_rate": 8.186701168352545e-05, "loss": 2.8458, "step": 23125 }, { "epoch": 1.81, "learning_rate": 8.186622755430096e-05, "loss": 2.9175, "step": 23126 }, { "epoch": 1.81, "learning_rate": 8.186544342507646e-05, "loss": 2.7881, "step": 23127 }, { "epoch": 1.81, "learning_rate": 8.186465929585196e-05, "loss": 3.0678, "step": 23128 }, { "epoch": 1.81, "learning_rate": 8.186387516662747e-05, "loss": 2.7354, "step": 23129 }, { "epoch": 1.81, "learning_rate": 8.186309103740297e-05, "loss": 2.9727, "step": 23130 }, { "epoch": 1.81, "learning_rate": 8.186230690817847e-05, "loss": 2.8888, "step": 23131 }, { "epoch": 1.81, "learning_rate": 8.186152277895398e-05, "loss": 2.8713, "step": 23132 }, { "epoch": 1.81, "learning_rate": 8.186073864972948e-05, "loss": 2.9471, "step": 23133 }, { "epoch": 1.81, "learning_rate": 8.185995452050497e-05, "loss": 3.1052, "step": 23134 }, { "epoch": 1.81, "learning_rate": 8.185917039128049e-05, "loss": 3.0996, "step": 23135 }, { "epoch": 1.81, "learning_rate": 8.185838626205599e-05, "loss": 2.9408, "step": 23136 }, { "epoch": 1.81, "learning_rate": 8.18576021328315e-05, "loss": 2.8089, "step": 23137 }, { "epoch": 1.81, "learning_rate": 8.1856818003607e-05, "loss": 2.9261, "step": 23138 }, { "epoch": 1.81, "learning_rate": 8.18560338743825e-05, "loss": 2.6338, "step": 23139 }, { "epoch": 1.81, "learning_rate": 8.185524974515801e-05, "loss": 2.9584, "step": 23140 }, { "epoch": 1.81, "learning_rate": 8.185446561593351e-05, "loss": 2.7336, "step": 23141 }, { "epoch": 1.81, "learning_rate": 8.185368148670901e-05, "loss": 2.8196, "step": 23142 }, { "epoch": 1.81, "learning_rate": 8.185289735748452e-05, "loss": 2.9746, "step": 23143 }, { "epoch": 1.81, "learning_rate": 8.185211322826002e-05, "loss": 2.9003, "step": 23144 }, { "epoch": 1.81, "learning_rate": 8.185132909903553e-05, "loss": 2.8439, "step": 23145 }, { "epoch": 1.81, "learning_rate": 8.185054496981103e-05, "loss": 3.2313, "step": 23146 }, { "epoch": 1.82, "learning_rate": 8.184976084058653e-05, "loss": 3.071, "step": 23147 }, { "epoch": 1.82, "learning_rate": 8.184897671136204e-05, "loss": 3.1326, "step": 23148 }, { "epoch": 1.82, "learning_rate": 8.184819258213754e-05, "loss": 3.1235, "step": 23149 }, { "epoch": 1.82, "learning_rate": 8.184740845291304e-05, "loss": 3.1037, "step": 23150 }, { "epoch": 1.82, "learning_rate": 8.184662432368855e-05, "loss": 2.8318, "step": 23151 }, { "epoch": 1.82, "learning_rate": 8.184584019446405e-05, "loss": 2.8731, "step": 23152 }, { "epoch": 1.82, "learning_rate": 8.184505606523955e-05, "loss": 2.7657, "step": 23153 }, { "epoch": 1.82, "learning_rate": 8.184427193601506e-05, "loss": 2.9076, "step": 23154 }, { "epoch": 1.82, "learning_rate": 8.184348780679056e-05, "loss": 2.722, "step": 23155 }, { "epoch": 1.82, "learning_rate": 8.184270367756607e-05, "loss": 2.9452, "step": 23156 }, { "epoch": 1.82, "learning_rate": 8.184191954834157e-05, "loss": 3.1022, "step": 23157 }, { "epoch": 1.82, "learning_rate": 8.184113541911707e-05, "loss": 2.7641, "step": 23158 }, { "epoch": 1.82, "learning_rate": 8.184035128989258e-05, "loss": 2.9251, "step": 23159 }, { "epoch": 1.82, "learning_rate": 8.183956716066808e-05, "loss": 2.6401, "step": 23160 }, { "epoch": 1.82, "learning_rate": 8.183878303144358e-05, "loss": 2.8532, "step": 23161 }, { "epoch": 1.82, "learning_rate": 8.183799890221909e-05, "loss": 2.8483, "step": 23162 }, { "epoch": 1.82, "learning_rate": 8.183721477299459e-05, "loss": 2.7653, "step": 23163 }, { "epoch": 1.82, "learning_rate": 8.18364306437701e-05, "loss": 2.635, "step": 23164 }, { "epoch": 1.82, "learning_rate": 8.18356465145456e-05, "loss": 2.9924, "step": 23165 }, { "epoch": 1.82, "learning_rate": 8.183486238532112e-05, "loss": 2.8156, "step": 23166 }, { "epoch": 1.82, "learning_rate": 8.183407825609662e-05, "loss": 2.8387, "step": 23167 }, { "epoch": 1.82, "learning_rate": 8.183329412687212e-05, "loss": 2.7132, "step": 23168 }, { "epoch": 1.82, "learning_rate": 8.183250999764763e-05, "loss": 2.7137, "step": 23169 }, { "epoch": 1.82, "learning_rate": 8.183172586842312e-05, "loss": 3.0091, "step": 23170 }, { "epoch": 1.82, "learning_rate": 8.183094173919862e-05, "loss": 2.7499, "step": 23171 }, { "epoch": 1.82, "learning_rate": 8.183015760997412e-05, "loss": 2.8186, "step": 23172 }, { "epoch": 1.82, "learning_rate": 8.182937348074963e-05, "loss": 2.8784, "step": 23173 }, { "epoch": 1.82, "learning_rate": 8.182858935152513e-05, "loss": 2.9792, "step": 23174 }, { "epoch": 1.82, "learning_rate": 8.182780522230064e-05, "loss": 2.6248, "step": 23175 }, { "epoch": 1.82, "learning_rate": 8.182702109307614e-05, "loss": 2.831, "step": 23176 }, { "epoch": 1.82, "learning_rate": 8.182623696385164e-05, "loss": 2.8743, "step": 23177 }, { "epoch": 1.82, "learning_rate": 8.182545283462715e-05, "loss": 2.7804, "step": 23178 }, { "epoch": 1.82, "learning_rate": 8.182466870540265e-05, "loss": 3.2243, "step": 23179 }, { "epoch": 1.82, "learning_rate": 8.182388457617815e-05, "loss": 3.0066, "step": 23180 }, { "epoch": 1.82, "learning_rate": 8.182310044695366e-05, "loss": 2.8669, "step": 23181 }, { "epoch": 1.82, "learning_rate": 8.182231631772916e-05, "loss": 2.9181, "step": 23182 }, { "epoch": 1.82, "learning_rate": 8.182153218850467e-05, "loss": 3.0172, "step": 23183 }, { "epoch": 1.82, "learning_rate": 8.182074805928017e-05, "loss": 2.8801, "step": 23184 }, { "epoch": 1.82, "learning_rate": 8.181996393005567e-05, "loss": 2.8534, "step": 23185 }, { "epoch": 1.82, "learning_rate": 8.181917980083118e-05, "loss": 2.6411, "step": 23186 }, { "epoch": 1.82, "learning_rate": 8.181839567160668e-05, "loss": 2.8544, "step": 23187 }, { "epoch": 1.82, "learning_rate": 8.181761154238218e-05, "loss": 3.0551, "step": 23188 }, { "epoch": 1.82, "learning_rate": 8.181682741315769e-05, "loss": 3.0908, "step": 23189 }, { "epoch": 1.82, "learning_rate": 8.181604328393319e-05, "loss": 2.8278, "step": 23190 }, { "epoch": 1.82, "learning_rate": 8.18152591547087e-05, "loss": 3.1434, "step": 23191 }, { "epoch": 1.82, "learning_rate": 8.181447502548421e-05, "loss": 2.7665, "step": 23192 }, { "epoch": 1.82, "learning_rate": 8.181369089625972e-05, "loss": 3.061, "step": 23193 }, { "epoch": 1.82, "learning_rate": 8.181290676703522e-05, "loss": 2.8423, "step": 23194 }, { "epoch": 1.82, "learning_rate": 8.181212263781072e-05, "loss": 2.7847, "step": 23195 }, { "epoch": 1.82, "learning_rate": 8.181133850858623e-05, "loss": 3.0001, "step": 23196 }, { "epoch": 1.82, "learning_rate": 8.181055437936173e-05, "loss": 2.953, "step": 23197 }, { "epoch": 1.82, "learning_rate": 8.180977025013723e-05, "loss": 2.9838, "step": 23198 }, { "epoch": 1.82, "learning_rate": 8.180898612091274e-05, "loss": 2.944, "step": 23199 }, { "epoch": 1.82, "learning_rate": 8.180820199168824e-05, "loss": 2.9101, "step": 23200 }, { "epoch": 1.82, "learning_rate": 8.180741786246374e-05, "loss": 2.8562, "step": 23201 }, { "epoch": 1.82, "learning_rate": 8.180663373323925e-05, "loss": 2.9857, "step": 23202 }, { "epoch": 1.82, "learning_rate": 8.180584960401475e-05, "loss": 2.945, "step": 23203 }, { "epoch": 1.82, "learning_rate": 8.180506547479026e-05, "loss": 3.0914, "step": 23204 }, { "epoch": 1.82, "learning_rate": 8.180428134556576e-05, "loss": 2.7889, "step": 23205 }, { "epoch": 1.82, "learning_rate": 8.180349721634125e-05, "loss": 2.9032, "step": 23206 }, { "epoch": 1.82, "learning_rate": 8.180271308711675e-05, "loss": 3.0463, "step": 23207 }, { "epoch": 1.82, "learning_rate": 8.180192895789226e-05, "loss": 3.0079, "step": 23208 }, { "epoch": 1.82, "learning_rate": 8.180114482866776e-05, "loss": 2.9183, "step": 23209 }, { "epoch": 1.82, "learning_rate": 8.180036069944326e-05, "loss": 3.1043, "step": 23210 }, { "epoch": 1.82, "learning_rate": 8.179957657021877e-05, "loss": 2.8953, "step": 23211 }, { "epoch": 1.82, "learning_rate": 8.179879244099427e-05, "loss": 2.7884, "step": 23212 }, { "epoch": 1.82, "learning_rate": 8.179800831176978e-05, "loss": 2.9924, "step": 23213 }, { "epoch": 1.82, "learning_rate": 8.179722418254528e-05, "loss": 2.9911, "step": 23214 }, { "epoch": 1.82, "learning_rate": 8.179644005332078e-05, "loss": 2.8627, "step": 23215 }, { "epoch": 1.82, "learning_rate": 8.179565592409629e-05, "loss": 2.9593, "step": 23216 }, { "epoch": 1.82, "learning_rate": 8.179487179487179e-05, "loss": 2.9722, "step": 23217 }, { "epoch": 1.82, "learning_rate": 8.179408766564731e-05, "loss": 2.9084, "step": 23218 }, { "epoch": 1.82, "learning_rate": 8.179330353642281e-05, "loss": 2.7549, "step": 23219 }, { "epoch": 1.82, "learning_rate": 8.179251940719831e-05, "loss": 3.0723, "step": 23220 }, { "epoch": 1.82, "learning_rate": 8.179173527797382e-05, "loss": 2.98, "step": 23221 }, { "epoch": 1.82, "learning_rate": 8.179095114874932e-05, "loss": 2.8905, "step": 23222 }, { "epoch": 1.82, "learning_rate": 8.179016701952483e-05, "loss": 3.0899, "step": 23223 }, { "epoch": 1.82, "learning_rate": 8.178938289030033e-05, "loss": 2.8331, "step": 23224 }, { "epoch": 1.82, "learning_rate": 8.178859876107583e-05, "loss": 3.3153, "step": 23225 }, { "epoch": 1.82, "learning_rate": 8.178781463185134e-05, "loss": 2.8901, "step": 23226 }, { "epoch": 1.82, "learning_rate": 8.178703050262684e-05, "loss": 3.0213, "step": 23227 }, { "epoch": 1.82, "learning_rate": 8.178624637340234e-05, "loss": 2.9175, "step": 23228 }, { "epoch": 1.82, "learning_rate": 8.178546224417785e-05, "loss": 2.9563, "step": 23229 }, { "epoch": 1.82, "learning_rate": 8.178467811495335e-05, "loss": 2.8485, "step": 23230 }, { "epoch": 1.82, "learning_rate": 8.178389398572886e-05, "loss": 2.9277, "step": 23231 }, { "epoch": 1.82, "learning_rate": 8.178310985650436e-05, "loss": 2.8949, "step": 23232 }, { "epoch": 1.82, "learning_rate": 8.178232572727986e-05, "loss": 3.0735, "step": 23233 }, { "epoch": 1.82, "learning_rate": 8.178154159805537e-05, "loss": 2.6442, "step": 23234 }, { "epoch": 1.82, "learning_rate": 8.178075746883087e-05, "loss": 3.0267, "step": 23235 }, { "epoch": 1.82, "learning_rate": 8.177997333960637e-05, "loss": 2.9598, "step": 23236 }, { "epoch": 1.82, "learning_rate": 8.177918921038188e-05, "loss": 3.0991, "step": 23237 }, { "epoch": 1.82, "learning_rate": 8.177840508115738e-05, "loss": 3.0271, "step": 23238 }, { "epoch": 1.82, "learning_rate": 8.177762095193288e-05, "loss": 2.8494, "step": 23239 }, { "epoch": 1.82, "learning_rate": 8.177683682270839e-05, "loss": 3.1828, "step": 23240 }, { "epoch": 1.82, "learning_rate": 8.177605269348389e-05, "loss": 3.2203, "step": 23241 }, { "epoch": 1.82, "learning_rate": 8.177526856425938e-05, "loss": 2.6924, "step": 23242 }, { "epoch": 1.82, "learning_rate": 8.17744844350349e-05, "loss": 2.8776, "step": 23243 }, { "epoch": 1.82, "learning_rate": 8.17737003058104e-05, "loss": 3.0021, "step": 23244 }, { "epoch": 1.82, "learning_rate": 8.17729161765859e-05, "loss": 2.801, "step": 23245 }, { "epoch": 1.82, "learning_rate": 8.177213204736141e-05, "loss": 2.9611, "step": 23246 }, { "epoch": 1.82, "learning_rate": 8.177134791813691e-05, "loss": 2.8352, "step": 23247 }, { "epoch": 1.82, "learning_rate": 8.177056378891242e-05, "loss": 2.8464, "step": 23248 }, { "epoch": 1.82, "learning_rate": 8.176977965968792e-05, "loss": 2.9819, "step": 23249 }, { "epoch": 1.82, "learning_rate": 8.176899553046343e-05, "loss": 2.7531, "step": 23250 }, { "epoch": 1.82, "learning_rate": 8.176821140123893e-05, "loss": 3.2178, "step": 23251 }, { "epoch": 1.82, "learning_rate": 8.176742727201443e-05, "loss": 2.9349, "step": 23252 }, { "epoch": 1.82, "learning_rate": 8.176664314278994e-05, "loss": 2.9012, "step": 23253 }, { "epoch": 1.82, "learning_rate": 8.176585901356544e-05, "loss": 2.9857, "step": 23254 }, { "epoch": 1.82, "learning_rate": 8.176507488434094e-05, "loss": 2.8906, "step": 23255 }, { "epoch": 1.82, "learning_rate": 8.176429075511645e-05, "loss": 2.7725, "step": 23256 }, { "epoch": 1.82, "learning_rate": 8.176350662589195e-05, "loss": 3.1165, "step": 23257 }, { "epoch": 1.82, "learning_rate": 8.176272249666745e-05, "loss": 3.0742, "step": 23258 }, { "epoch": 1.82, "learning_rate": 8.176193836744296e-05, "loss": 2.6831, "step": 23259 }, { "epoch": 1.82, "learning_rate": 8.176115423821846e-05, "loss": 2.8859, "step": 23260 }, { "epoch": 1.82, "learning_rate": 8.176037010899397e-05, "loss": 3.0055, "step": 23261 }, { "epoch": 1.82, "learning_rate": 8.175958597976947e-05, "loss": 2.7297, "step": 23262 }, { "epoch": 1.82, "learning_rate": 8.175880185054497e-05, "loss": 2.716, "step": 23263 }, { "epoch": 1.82, "learning_rate": 8.175801772132048e-05, "loss": 3.2176, "step": 23264 }, { "epoch": 1.82, "learning_rate": 8.175723359209598e-05, "loss": 2.9351, "step": 23265 }, { "epoch": 1.82, "learning_rate": 8.175644946287148e-05, "loss": 2.8045, "step": 23266 }, { "epoch": 1.82, "learning_rate": 8.175566533364699e-05, "loss": 3.0764, "step": 23267 }, { "epoch": 1.82, "learning_rate": 8.175488120442249e-05, "loss": 2.9583, "step": 23268 }, { "epoch": 1.82, "learning_rate": 8.1754097075198e-05, "loss": 2.8786, "step": 23269 }, { "epoch": 1.82, "learning_rate": 8.17533129459735e-05, "loss": 3.0071, "step": 23270 }, { "epoch": 1.82, "learning_rate": 8.1752528816749e-05, "loss": 2.785, "step": 23271 }, { "epoch": 1.82, "learning_rate": 8.17517446875245e-05, "loss": 2.93, "step": 23272 }, { "epoch": 1.82, "learning_rate": 8.175096055830001e-05, "loss": 3.034, "step": 23273 }, { "epoch": 1.82, "learning_rate": 8.175017642907553e-05, "loss": 2.7982, "step": 23274 }, { "epoch": 1.83, "learning_rate": 8.174939229985103e-05, "loss": 2.8357, "step": 23275 }, { "epoch": 1.83, "learning_rate": 8.174860817062653e-05, "loss": 2.9669, "step": 23276 }, { "epoch": 1.83, "learning_rate": 8.174782404140204e-05, "loss": 2.8741, "step": 23277 }, { "epoch": 1.83, "learning_rate": 8.174703991217753e-05, "loss": 2.7024, "step": 23278 }, { "epoch": 1.83, "learning_rate": 8.174625578295303e-05, "loss": 2.8804, "step": 23279 }, { "epoch": 1.83, "learning_rate": 8.174547165372854e-05, "loss": 3.0506, "step": 23280 }, { "epoch": 1.83, "learning_rate": 8.174468752450404e-05, "loss": 3.0802, "step": 23281 }, { "epoch": 1.83, "learning_rate": 8.174390339527954e-05, "loss": 3.0542, "step": 23282 }, { "epoch": 1.83, "learning_rate": 8.174311926605505e-05, "loss": 2.7447, "step": 23283 }, { "epoch": 1.83, "learning_rate": 8.174233513683055e-05, "loss": 2.8726, "step": 23284 }, { "epoch": 1.83, "learning_rate": 8.174155100760605e-05, "loss": 2.8785, "step": 23285 }, { "epoch": 1.83, "learning_rate": 8.174076687838156e-05, "loss": 2.6437, "step": 23286 }, { "epoch": 1.83, "learning_rate": 8.173998274915706e-05, "loss": 3.0121, "step": 23287 }, { "epoch": 1.83, "learning_rate": 8.173919861993256e-05, "loss": 2.7673, "step": 23288 }, { "epoch": 1.83, "learning_rate": 8.173841449070807e-05, "loss": 3.0092, "step": 23289 }, { "epoch": 1.83, "learning_rate": 8.173763036148357e-05, "loss": 2.9221, "step": 23290 }, { "epoch": 1.83, "learning_rate": 8.173684623225908e-05, "loss": 2.8692, "step": 23291 }, { "epoch": 1.83, "learning_rate": 8.173606210303458e-05, "loss": 2.7898, "step": 23292 }, { "epoch": 1.83, "learning_rate": 8.173527797381008e-05, "loss": 2.8572, "step": 23293 }, { "epoch": 1.83, "learning_rate": 8.173449384458559e-05, "loss": 3.0838, "step": 23294 }, { "epoch": 1.83, "learning_rate": 8.173370971536109e-05, "loss": 2.9208, "step": 23295 }, { "epoch": 1.83, "learning_rate": 8.17329255861366e-05, "loss": 2.8914, "step": 23296 }, { "epoch": 1.83, "learning_rate": 8.17321414569121e-05, "loss": 2.7856, "step": 23297 }, { "epoch": 1.83, "learning_rate": 8.17313573276876e-05, "loss": 2.7788, "step": 23298 }, { "epoch": 1.83, "learning_rate": 8.17305731984631e-05, "loss": 2.8043, "step": 23299 }, { "epoch": 1.83, "learning_rate": 8.172978906923862e-05, "loss": 2.9809, "step": 23300 }, { "epoch": 1.83, "learning_rate": 8.172900494001413e-05, "loss": 2.9817, "step": 23301 }, { "epoch": 1.83, "learning_rate": 8.172822081078963e-05, "loss": 3.0287, "step": 23302 }, { "epoch": 1.83, "learning_rate": 8.172743668156513e-05, "loss": 3.0953, "step": 23303 }, { "epoch": 1.83, "learning_rate": 8.172665255234064e-05, "loss": 2.8443, "step": 23304 }, { "epoch": 1.83, "learning_rate": 8.172586842311614e-05, "loss": 3.0655, "step": 23305 }, { "epoch": 1.83, "learning_rate": 8.172508429389164e-05, "loss": 2.7326, "step": 23306 }, { "epoch": 1.83, "learning_rate": 8.172430016466715e-05, "loss": 2.9034, "step": 23307 }, { "epoch": 1.83, "learning_rate": 8.172351603544265e-05, "loss": 2.9186, "step": 23308 }, { "epoch": 1.83, "learning_rate": 8.172273190621816e-05, "loss": 2.9013, "step": 23309 }, { "epoch": 1.83, "learning_rate": 8.172194777699366e-05, "loss": 2.719, "step": 23310 }, { "epoch": 1.83, "learning_rate": 8.172116364776916e-05, "loss": 2.9715, "step": 23311 }, { "epoch": 1.83, "learning_rate": 8.172037951854467e-05, "loss": 2.8504, "step": 23312 }, { "epoch": 1.83, "learning_rate": 8.171959538932017e-05, "loss": 2.9479, "step": 23313 }, { "epoch": 1.83, "learning_rate": 8.171881126009566e-05, "loss": 2.9962, "step": 23314 }, { "epoch": 1.83, "learning_rate": 8.171802713087116e-05, "loss": 2.9608, "step": 23315 }, { "epoch": 1.83, "learning_rate": 8.171724300164667e-05, "loss": 2.8584, "step": 23316 }, { "epoch": 1.83, "learning_rate": 8.171645887242217e-05, "loss": 2.9004, "step": 23317 }, { "epoch": 1.83, "learning_rate": 8.171567474319768e-05, "loss": 2.792, "step": 23318 }, { "epoch": 1.83, "learning_rate": 8.171489061397318e-05, "loss": 2.8391, "step": 23319 }, { "epoch": 1.83, "learning_rate": 8.171410648474868e-05, "loss": 3.113, "step": 23320 }, { "epoch": 1.83, "learning_rate": 8.171332235552419e-05, "loss": 2.8599, "step": 23321 }, { "epoch": 1.83, "learning_rate": 8.171253822629969e-05, "loss": 2.7075, "step": 23322 }, { "epoch": 1.83, "learning_rate": 8.17117540970752e-05, "loss": 2.7938, "step": 23323 }, { "epoch": 1.83, "learning_rate": 8.17109699678507e-05, "loss": 2.8472, "step": 23324 }, { "epoch": 1.83, "learning_rate": 8.17101858386262e-05, "loss": 2.7494, "step": 23325 }, { "epoch": 1.83, "learning_rate": 8.170940170940172e-05, "loss": 3.0678, "step": 23326 }, { "epoch": 1.83, "learning_rate": 8.170861758017722e-05, "loss": 3.0141, "step": 23327 }, { "epoch": 1.83, "learning_rate": 8.170783345095273e-05, "loss": 2.8204, "step": 23328 }, { "epoch": 1.83, "learning_rate": 8.170704932172823e-05, "loss": 2.7091, "step": 23329 }, { "epoch": 1.83, "learning_rate": 8.170626519250373e-05, "loss": 2.7368, "step": 23330 }, { "epoch": 1.83, "learning_rate": 8.170548106327924e-05, "loss": 2.9205, "step": 23331 }, { "epoch": 1.83, "learning_rate": 8.170469693405474e-05, "loss": 2.7058, "step": 23332 }, { "epoch": 1.83, "learning_rate": 8.170391280483024e-05, "loss": 2.7006, "step": 23333 }, { "epoch": 1.83, "learning_rate": 8.170312867560575e-05, "loss": 2.9303, "step": 23334 }, { "epoch": 1.83, "learning_rate": 8.170234454638125e-05, "loss": 2.828, "step": 23335 }, { "epoch": 1.83, "learning_rate": 8.170156041715675e-05, "loss": 2.7358, "step": 23336 }, { "epoch": 1.83, "learning_rate": 8.170077628793226e-05, "loss": 2.9638, "step": 23337 }, { "epoch": 1.83, "learning_rate": 8.169999215870776e-05, "loss": 2.8028, "step": 23338 }, { "epoch": 1.83, "learning_rate": 8.169920802948327e-05, "loss": 3.0234, "step": 23339 }, { "epoch": 1.83, "learning_rate": 8.169842390025877e-05, "loss": 2.8555, "step": 23340 }, { "epoch": 1.83, "learning_rate": 8.169763977103427e-05, "loss": 2.8527, "step": 23341 }, { "epoch": 1.83, "learning_rate": 8.169685564180978e-05, "loss": 2.6477, "step": 23342 }, { "epoch": 1.83, "learning_rate": 8.169607151258528e-05, "loss": 3.0392, "step": 23343 }, { "epoch": 1.83, "learning_rate": 8.169528738336078e-05, "loss": 2.6969, "step": 23344 }, { "epoch": 1.83, "learning_rate": 8.169450325413629e-05, "loss": 2.9924, "step": 23345 }, { "epoch": 1.83, "learning_rate": 8.169371912491179e-05, "loss": 2.7624, "step": 23346 }, { "epoch": 1.83, "learning_rate": 8.16929349956873e-05, "loss": 2.8285, "step": 23347 }, { "epoch": 1.83, "learning_rate": 8.16921508664628e-05, "loss": 2.9316, "step": 23348 }, { "epoch": 1.83, "learning_rate": 8.16913667372383e-05, "loss": 2.7334, "step": 23349 }, { "epoch": 1.83, "learning_rate": 8.169058260801379e-05, "loss": 2.7931, "step": 23350 }, { "epoch": 1.83, "learning_rate": 8.168979847878931e-05, "loss": 2.9308, "step": 23351 }, { "epoch": 1.83, "learning_rate": 8.168901434956481e-05, "loss": 3.0885, "step": 23352 }, { "epoch": 1.83, "learning_rate": 8.168823022034032e-05, "loss": 2.8927, "step": 23353 }, { "epoch": 1.83, "learning_rate": 8.168744609111582e-05, "loss": 2.8417, "step": 23354 }, { "epoch": 1.83, "learning_rate": 8.168666196189132e-05, "loss": 2.7775, "step": 23355 }, { "epoch": 1.83, "learning_rate": 8.168587783266683e-05, "loss": 2.9836, "step": 23356 }, { "epoch": 1.83, "learning_rate": 8.168509370344233e-05, "loss": 3.0183, "step": 23357 }, { "epoch": 1.83, "learning_rate": 8.168430957421784e-05, "loss": 2.9537, "step": 23358 }, { "epoch": 1.83, "learning_rate": 8.168352544499334e-05, "loss": 2.7662, "step": 23359 }, { "epoch": 1.83, "learning_rate": 8.168274131576884e-05, "loss": 2.865, "step": 23360 }, { "epoch": 1.83, "learning_rate": 8.168195718654435e-05, "loss": 2.9308, "step": 23361 }, { "epoch": 1.83, "learning_rate": 8.168117305731985e-05, "loss": 2.8282, "step": 23362 }, { "epoch": 1.83, "learning_rate": 8.168038892809535e-05, "loss": 2.7929, "step": 23363 }, { "epoch": 1.83, "learning_rate": 8.167960479887086e-05, "loss": 2.9413, "step": 23364 }, { "epoch": 1.83, "learning_rate": 8.167882066964636e-05, "loss": 2.8041, "step": 23365 }, { "epoch": 1.83, "learning_rate": 8.167803654042187e-05, "loss": 2.8021, "step": 23366 }, { "epoch": 1.83, "learning_rate": 8.167725241119737e-05, "loss": 3.1797, "step": 23367 }, { "epoch": 1.83, "learning_rate": 8.167646828197287e-05, "loss": 2.8933, "step": 23368 }, { "epoch": 1.83, "learning_rate": 8.167568415274838e-05, "loss": 2.6651, "step": 23369 }, { "epoch": 1.83, "learning_rate": 8.167490002352388e-05, "loss": 2.7468, "step": 23370 }, { "epoch": 1.83, "learning_rate": 8.167411589429938e-05, "loss": 3.0344, "step": 23371 }, { "epoch": 1.83, "learning_rate": 8.167333176507489e-05, "loss": 3.1243, "step": 23372 }, { "epoch": 1.83, "learning_rate": 8.167254763585039e-05, "loss": 2.8542, "step": 23373 }, { "epoch": 1.83, "learning_rate": 8.16717635066259e-05, "loss": 2.9607, "step": 23374 }, { "epoch": 1.83, "learning_rate": 8.16709793774014e-05, "loss": 2.8002, "step": 23375 }, { "epoch": 1.83, "learning_rate": 8.16701952481769e-05, "loss": 2.8475, "step": 23376 }, { "epoch": 1.83, "learning_rate": 8.16694111189524e-05, "loss": 2.9323, "step": 23377 }, { "epoch": 1.83, "learning_rate": 8.166862698972791e-05, "loss": 2.9362, "step": 23378 }, { "epoch": 1.83, "learning_rate": 8.166784286050341e-05, "loss": 2.857, "step": 23379 }, { "epoch": 1.83, "learning_rate": 8.166705873127892e-05, "loss": 3.0708, "step": 23380 }, { "epoch": 1.83, "learning_rate": 8.166627460205442e-05, "loss": 2.9327, "step": 23381 }, { "epoch": 1.83, "learning_rate": 8.166549047282992e-05, "loss": 3.0207, "step": 23382 }, { "epoch": 1.83, "learning_rate": 8.166470634360544e-05, "loss": 2.9844, "step": 23383 }, { "epoch": 1.83, "learning_rate": 8.166392221438094e-05, "loss": 2.9532, "step": 23384 }, { "epoch": 1.83, "learning_rate": 8.166313808515645e-05, "loss": 2.9374, "step": 23385 }, { "epoch": 1.83, "learning_rate": 8.166235395593194e-05, "loss": 2.8681, "step": 23386 }, { "epoch": 1.83, "learning_rate": 8.166156982670744e-05, "loss": 2.7319, "step": 23387 }, { "epoch": 1.83, "learning_rate": 8.166078569748295e-05, "loss": 3.0668, "step": 23388 }, { "epoch": 1.83, "learning_rate": 8.166000156825845e-05, "loss": 2.8805, "step": 23389 }, { "epoch": 1.83, "learning_rate": 8.165921743903395e-05, "loss": 3.0718, "step": 23390 }, { "epoch": 1.83, "learning_rate": 8.165843330980946e-05, "loss": 2.9568, "step": 23391 }, { "epoch": 1.83, "learning_rate": 8.165764918058496e-05, "loss": 2.775, "step": 23392 }, { "epoch": 1.83, "learning_rate": 8.165686505136046e-05, "loss": 2.6617, "step": 23393 }, { "epoch": 1.83, "learning_rate": 8.165608092213597e-05, "loss": 2.8744, "step": 23394 }, { "epoch": 1.83, "learning_rate": 8.165529679291147e-05, "loss": 2.85, "step": 23395 }, { "epoch": 1.83, "learning_rate": 8.165451266368698e-05, "loss": 2.9665, "step": 23396 }, { "epoch": 1.83, "learning_rate": 8.165372853446248e-05, "loss": 3.0074, "step": 23397 }, { "epoch": 1.83, "learning_rate": 8.165294440523798e-05, "loss": 2.9721, "step": 23398 }, { "epoch": 1.83, "learning_rate": 8.165216027601349e-05, "loss": 2.8985, "step": 23399 }, { "epoch": 1.83, "learning_rate": 8.165137614678899e-05, "loss": 2.7917, "step": 23400 }, { "epoch": 1.83, "learning_rate": 8.16505920175645e-05, "loss": 2.9382, "step": 23401 }, { "epoch": 1.84, "learning_rate": 8.164980788834e-05, "loss": 3.0175, "step": 23402 }, { "epoch": 1.84, "learning_rate": 8.16490237591155e-05, "loss": 3.1269, "step": 23403 }, { "epoch": 1.84, "learning_rate": 8.1648239629891e-05, "loss": 2.9183, "step": 23404 }, { "epoch": 1.84, "learning_rate": 8.164745550066651e-05, "loss": 2.7956, "step": 23405 }, { "epoch": 1.84, "learning_rate": 8.164667137144201e-05, "loss": 3.0124, "step": 23406 }, { "epoch": 1.84, "learning_rate": 8.164588724221752e-05, "loss": 3.0654, "step": 23407 }, { "epoch": 1.84, "learning_rate": 8.164510311299303e-05, "loss": 2.8638, "step": 23408 }, { "epoch": 1.84, "learning_rate": 8.164431898376854e-05, "loss": 2.924, "step": 23409 }, { "epoch": 1.84, "learning_rate": 8.164353485454404e-05, "loss": 3.1336, "step": 23410 }, { "epoch": 1.84, "learning_rate": 8.164275072531954e-05, "loss": 2.8942, "step": 23411 }, { "epoch": 1.84, "learning_rate": 8.164196659609505e-05, "loss": 2.6535, "step": 23412 }, { "epoch": 1.84, "learning_rate": 8.164118246687055e-05, "loss": 2.7745, "step": 23413 }, { "epoch": 1.84, "learning_rate": 8.164039833764606e-05, "loss": 2.8655, "step": 23414 }, { "epoch": 1.84, "learning_rate": 8.163961420842156e-05, "loss": 2.9583, "step": 23415 }, { "epoch": 1.84, "learning_rate": 8.163883007919706e-05, "loss": 2.9363, "step": 23416 }, { "epoch": 1.84, "learning_rate": 8.163804594997257e-05, "loss": 2.912, "step": 23417 }, { "epoch": 1.84, "learning_rate": 8.163726182074807e-05, "loss": 3.0477, "step": 23418 }, { "epoch": 1.84, "learning_rate": 8.163647769152357e-05, "loss": 3.0001, "step": 23419 }, { "epoch": 1.84, "learning_rate": 8.163569356229908e-05, "loss": 2.9209, "step": 23420 }, { "epoch": 1.84, "learning_rate": 8.163490943307458e-05, "loss": 2.9552, "step": 23421 }, { "epoch": 1.84, "learning_rate": 8.163412530385007e-05, "loss": 2.7915, "step": 23422 }, { "epoch": 1.84, "learning_rate": 8.163334117462557e-05, "loss": 2.9644, "step": 23423 }, { "epoch": 1.84, "learning_rate": 8.163255704540108e-05, "loss": 2.6494, "step": 23424 }, { "epoch": 1.84, "learning_rate": 8.163177291617658e-05, "loss": 3.1718, "step": 23425 }, { "epoch": 1.84, "learning_rate": 8.163098878695209e-05, "loss": 3.0497, "step": 23426 }, { "epoch": 1.84, "learning_rate": 8.163020465772759e-05, "loss": 2.8992, "step": 23427 }, { "epoch": 1.84, "learning_rate": 8.16294205285031e-05, "loss": 2.8578, "step": 23428 }, { "epoch": 1.84, "learning_rate": 8.16286363992786e-05, "loss": 3.0281, "step": 23429 }, { "epoch": 1.84, "learning_rate": 8.16278522700541e-05, "loss": 2.8164, "step": 23430 }, { "epoch": 1.84, "learning_rate": 8.16270681408296e-05, "loss": 2.7819, "step": 23431 }, { "epoch": 1.84, "learning_rate": 8.162628401160511e-05, "loss": 2.961, "step": 23432 }, { "epoch": 1.84, "learning_rate": 8.162549988238061e-05, "loss": 3.1023, "step": 23433 }, { "epoch": 1.84, "learning_rate": 8.162471575315613e-05, "loss": 2.9052, "step": 23434 }, { "epoch": 1.84, "learning_rate": 8.162393162393163e-05, "loss": 2.8021, "step": 23435 }, { "epoch": 1.84, "learning_rate": 8.162314749470714e-05, "loss": 2.8513, "step": 23436 }, { "epoch": 1.84, "learning_rate": 8.162236336548264e-05, "loss": 2.7705, "step": 23437 }, { "epoch": 1.84, "learning_rate": 8.162157923625814e-05, "loss": 2.937, "step": 23438 }, { "epoch": 1.84, "learning_rate": 8.162079510703365e-05, "loss": 2.9188, "step": 23439 }, { "epoch": 1.84, "learning_rate": 8.162001097780915e-05, "loss": 2.8535, "step": 23440 }, { "epoch": 1.84, "learning_rate": 8.161922684858465e-05, "loss": 2.9962, "step": 23441 }, { "epoch": 1.84, "learning_rate": 8.161844271936016e-05, "loss": 2.6364, "step": 23442 }, { "epoch": 1.84, "learning_rate": 8.161765859013566e-05, "loss": 3.0226, "step": 23443 }, { "epoch": 1.84, "learning_rate": 8.161687446091117e-05, "loss": 2.9198, "step": 23444 }, { "epoch": 1.84, "learning_rate": 8.161609033168667e-05, "loss": 3.3458, "step": 23445 }, { "epoch": 1.84, "learning_rate": 8.161530620246217e-05, "loss": 2.9484, "step": 23446 }, { "epoch": 1.84, "learning_rate": 8.161452207323768e-05, "loss": 2.6417, "step": 23447 }, { "epoch": 1.84, "learning_rate": 8.161373794401318e-05, "loss": 2.8234, "step": 23448 }, { "epoch": 1.84, "learning_rate": 8.161295381478868e-05, "loss": 2.7331, "step": 23449 }, { "epoch": 1.84, "learning_rate": 8.161216968556419e-05, "loss": 3.0565, "step": 23450 }, { "epoch": 1.84, "learning_rate": 8.161138555633969e-05, "loss": 2.6099, "step": 23451 }, { "epoch": 1.84, "learning_rate": 8.16106014271152e-05, "loss": 2.9204, "step": 23452 }, { "epoch": 1.84, "learning_rate": 8.16098172978907e-05, "loss": 2.8694, "step": 23453 }, { "epoch": 1.84, "learning_rate": 8.16090331686662e-05, "loss": 3.076, "step": 23454 }, { "epoch": 1.84, "learning_rate": 8.16082490394417e-05, "loss": 2.7596, "step": 23455 }, { "epoch": 1.84, "learning_rate": 8.160746491021721e-05, "loss": 3.0645, "step": 23456 }, { "epoch": 1.84, "learning_rate": 8.160668078099271e-05, "loss": 2.9148, "step": 23457 }, { "epoch": 1.84, "learning_rate": 8.16058966517682e-05, "loss": 2.8856, "step": 23458 }, { "epoch": 1.84, "learning_rate": 8.160511252254372e-05, "loss": 2.8412, "step": 23459 }, { "epoch": 1.84, "learning_rate": 8.160432839331922e-05, "loss": 2.8401, "step": 23460 }, { "epoch": 1.84, "learning_rate": 8.160354426409473e-05, "loss": 2.8878, "step": 23461 }, { "epoch": 1.84, "learning_rate": 8.160276013487023e-05, "loss": 3.2146, "step": 23462 }, { "epoch": 1.84, "learning_rate": 8.160197600564574e-05, "loss": 2.8957, "step": 23463 }, { "epoch": 1.84, "learning_rate": 8.160119187642124e-05, "loss": 2.8542, "step": 23464 }, { "epoch": 1.84, "learning_rate": 8.160040774719674e-05, "loss": 2.6463, "step": 23465 }, { "epoch": 1.84, "learning_rate": 8.159962361797225e-05, "loss": 2.8385, "step": 23466 }, { "epoch": 1.84, "learning_rate": 8.159883948874775e-05, "loss": 2.6623, "step": 23467 }, { "epoch": 1.84, "learning_rate": 8.159805535952325e-05, "loss": 3.2222, "step": 23468 }, { "epoch": 1.84, "learning_rate": 8.159727123029876e-05, "loss": 2.8801, "step": 23469 }, { "epoch": 1.84, "learning_rate": 8.159648710107426e-05, "loss": 2.8067, "step": 23470 }, { "epoch": 1.84, "learning_rate": 8.159570297184976e-05, "loss": 2.8958, "step": 23471 }, { "epoch": 1.84, "learning_rate": 8.159491884262527e-05, "loss": 2.8605, "step": 23472 }, { "epoch": 1.84, "learning_rate": 8.159413471340077e-05, "loss": 2.7557, "step": 23473 }, { "epoch": 1.84, "learning_rate": 8.159335058417628e-05, "loss": 2.8708, "step": 23474 }, { "epoch": 1.84, "learning_rate": 8.159256645495178e-05, "loss": 2.8974, "step": 23475 }, { "epoch": 1.84, "learning_rate": 8.159178232572728e-05, "loss": 2.9616, "step": 23476 }, { "epoch": 1.84, "learning_rate": 8.159099819650279e-05, "loss": 2.8983, "step": 23477 }, { "epoch": 1.84, "learning_rate": 8.159021406727829e-05, "loss": 2.7905, "step": 23478 }, { "epoch": 1.84, "learning_rate": 8.15894299380538e-05, "loss": 2.974, "step": 23479 }, { "epoch": 1.84, "learning_rate": 8.15886458088293e-05, "loss": 2.905, "step": 23480 }, { "epoch": 1.84, "learning_rate": 8.15878616796048e-05, "loss": 3.0714, "step": 23481 }, { "epoch": 1.84, "learning_rate": 8.15870775503803e-05, "loss": 2.6315, "step": 23482 }, { "epoch": 1.84, "learning_rate": 8.158629342115581e-05, "loss": 2.7124, "step": 23483 }, { "epoch": 1.84, "learning_rate": 8.158550929193131e-05, "loss": 2.9196, "step": 23484 }, { "epoch": 1.84, "learning_rate": 8.158472516270682e-05, "loss": 2.8653, "step": 23485 }, { "epoch": 1.84, "learning_rate": 8.158394103348232e-05, "loss": 2.8714, "step": 23486 }, { "epoch": 1.84, "learning_rate": 8.158315690425782e-05, "loss": 3.13, "step": 23487 }, { "epoch": 1.84, "learning_rate": 8.158237277503333e-05, "loss": 2.954, "step": 23488 }, { "epoch": 1.84, "learning_rate": 8.158158864580883e-05, "loss": 2.831, "step": 23489 }, { "epoch": 1.84, "learning_rate": 8.158080451658433e-05, "loss": 3.3208, "step": 23490 }, { "epoch": 1.84, "learning_rate": 8.158002038735985e-05, "loss": 2.973, "step": 23491 }, { "epoch": 1.84, "learning_rate": 8.157923625813536e-05, "loss": 2.936, "step": 23492 }, { "epoch": 1.84, "learning_rate": 8.157845212891086e-05, "loss": 2.8514, "step": 23493 }, { "epoch": 1.84, "learning_rate": 8.157766799968635e-05, "loss": 2.7291, "step": 23494 }, { "epoch": 1.84, "learning_rate": 8.157688387046185e-05, "loss": 2.8389, "step": 23495 }, { "epoch": 1.84, "learning_rate": 8.157609974123736e-05, "loss": 3.2497, "step": 23496 }, { "epoch": 1.84, "learning_rate": 8.157531561201286e-05, "loss": 2.71, "step": 23497 }, { "epoch": 1.84, "learning_rate": 8.157453148278836e-05, "loss": 2.7589, "step": 23498 }, { "epoch": 1.84, "learning_rate": 8.157374735356387e-05, "loss": 3.2226, "step": 23499 }, { "epoch": 1.84, "learning_rate": 8.157296322433937e-05, "loss": 2.8441, "step": 23500 }, { "epoch": 1.84, "learning_rate": 8.157217909511488e-05, "loss": 2.7844, "step": 23501 }, { "epoch": 1.84, "learning_rate": 8.157139496589038e-05, "loss": 3.033, "step": 23502 }, { "epoch": 1.84, "learning_rate": 8.157061083666588e-05, "loss": 2.9815, "step": 23503 }, { "epoch": 1.84, "learning_rate": 8.156982670744139e-05, "loss": 2.8362, "step": 23504 }, { "epoch": 1.84, "learning_rate": 8.156904257821689e-05, "loss": 2.9436, "step": 23505 }, { "epoch": 1.84, "learning_rate": 8.15682584489924e-05, "loss": 2.8589, "step": 23506 }, { "epoch": 1.84, "learning_rate": 8.15674743197679e-05, "loss": 2.8483, "step": 23507 }, { "epoch": 1.84, "learning_rate": 8.15666901905434e-05, "loss": 2.9017, "step": 23508 }, { "epoch": 1.84, "learning_rate": 8.15659060613189e-05, "loss": 3.1347, "step": 23509 }, { "epoch": 1.84, "learning_rate": 8.156512193209441e-05, "loss": 2.9183, "step": 23510 }, { "epoch": 1.84, "learning_rate": 8.156433780286991e-05, "loss": 2.7869, "step": 23511 }, { "epoch": 1.84, "learning_rate": 8.156355367364542e-05, "loss": 2.6914, "step": 23512 }, { "epoch": 1.84, "learning_rate": 8.156276954442092e-05, "loss": 3.1128, "step": 23513 }, { "epoch": 1.84, "learning_rate": 8.156198541519642e-05, "loss": 2.8809, "step": 23514 }, { "epoch": 1.84, "learning_rate": 8.156120128597193e-05, "loss": 2.6835, "step": 23515 }, { "epoch": 1.84, "learning_rate": 8.156041715674744e-05, "loss": 2.7377, "step": 23516 }, { "epoch": 1.84, "learning_rate": 8.155963302752295e-05, "loss": 2.8768, "step": 23517 }, { "epoch": 1.84, "learning_rate": 8.155884889829845e-05, "loss": 2.8946, "step": 23518 }, { "epoch": 1.84, "learning_rate": 8.155806476907396e-05, "loss": 2.8259, "step": 23519 }, { "epoch": 1.84, "learning_rate": 8.155728063984946e-05, "loss": 3.1007, "step": 23520 }, { "epoch": 1.84, "learning_rate": 8.155649651062496e-05, "loss": 2.9492, "step": 23521 }, { "epoch": 1.84, "learning_rate": 8.155571238140047e-05, "loss": 2.928, "step": 23522 }, { "epoch": 1.84, "learning_rate": 8.155492825217597e-05, "loss": 2.9285, "step": 23523 }, { "epoch": 1.84, "learning_rate": 8.155414412295147e-05, "loss": 2.883, "step": 23524 }, { "epoch": 1.84, "learning_rate": 8.155335999372698e-05, "loss": 3.1247, "step": 23525 }, { "epoch": 1.84, "learning_rate": 8.155257586450248e-05, "loss": 2.861, "step": 23526 }, { "epoch": 1.84, "learning_rate": 8.155179173527798e-05, "loss": 3.0279, "step": 23527 }, { "epoch": 1.84, "learning_rate": 8.155100760605349e-05, "loss": 2.9066, "step": 23528 }, { "epoch": 1.84, "learning_rate": 8.155022347682899e-05, "loss": 2.9168, "step": 23529 }, { "epoch": 1.85, "learning_rate": 8.154943934760448e-05, "loss": 2.9305, "step": 23530 }, { "epoch": 1.85, "learning_rate": 8.154865521837999e-05, "loss": 2.9966, "step": 23531 }, { "epoch": 1.85, "learning_rate": 8.154787108915549e-05, "loss": 2.9301, "step": 23532 }, { "epoch": 1.85, "learning_rate": 8.154708695993099e-05, "loss": 2.8212, "step": 23533 }, { "epoch": 1.85, "learning_rate": 8.15463028307065e-05, "loss": 2.7381, "step": 23534 }, { "epoch": 1.85, "learning_rate": 8.1545518701482e-05, "loss": 2.9621, "step": 23535 }, { "epoch": 1.85, "learning_rate": 8.15447345722575e-05, "loss": 2.8498, "step": 23536 }, { "epoch": 1.85, "learning_rate": 8.154395044303301e-05, "loss": 2.7828, "step": 23537 }, { "epoch": 1.85, "learning_rate": 8.154316631380851e-05, "loss": 2.8531, "step": 23538 }, { "epoch": 1.85, "learning_rate": 8.154238218458402e-05, "loss": 2.7896, "step": 23539 }, { "epoch": 1.85, "learning_rate": 8.154159805535952e-05, "loss": 2.8068, "step": 23540 }, { "epoch": 1.85, "learning_rate": 8.154081392613502e-05, "loss": 2.8899, "step": 23541 }, { "epoch": 1.85, "learning_rate": 8.154002979691054e-05, "loss": 2.8196, "step": 23542 }, { "epoch": 1.85, "learning_rate": 8.153924566768604e-05, "loss": 2.8269, "step": 23543 }, { "epoch": 1.85, "learning_rate": 8.153846153846155e-05, "loss": 2.8304, "step": 23544 }, { "epoch": 1.85, "learning_rate": 8.153767740923705e-05, "loss": 3.0489, "step": 23545 }, { "epoch": 1.85, "learning_rate": 8.153689328001255e-05, "loss": 2.7498, "step": 23546 }, { "epoch": 1.85, "learning_rate": 8.153610915078806e-05, "loss": 2.9463, "step": 23547 }, { "epoch": 1.85, "learning_rate": 8.153532502156356e-05, "loss": 2.958, "step": 23548 }, { "epoch": 1.85, "learning_rate": 8.153454089233907e-05, "loss": 2.881, "step": 23549 }, { "epoch": 1.85, "learning_rate": 8.153375676311457e-05, "loss": 2.912, "step": 23550 }, { "epoch": 1.85, "learning_rate": 8.153297263389007e-05, "loss": 2.9207, "step": 23551 }, { "epoch": 1.85, "learning_rate": 8.153218850466558e-05, "loss": 2.8786, "step": 23552 }, { "epoch": 1.85, "learning_rate": 8.153140437544108e-05, "loss": 2.9647, "step": 23553 }, { "epoch": 1.85, "learning_rate": 8.153062024621658e-05, "loss": 2.8169, "step": 23554 }, { "epoch": 1.85, "learning_rate": 8.152983611699209e-05, "loss": 2.7131, "step": 23555 }, { "epoch": 1.85, "learning_rate": 8.152905198776759e-05, "loss": 2.9331, "step": 23556 }, { "epoch": 1.85, "learning_rate": 8.15282678585431e-05, "loss": 3.2491, "step": 23557 }, { "epoch": 1.85, "learning_rate": 8.15274837293186e-05, "loss": 2.8276, "step": 23558 }, { "epoch": 1.85, "learning_rate": 8.15266996000941e-05, "loss": 2.8007, "step": 23559 }, { "epoch": 1.85, "learning_rate": 8.15259154708696e-05, "loss": 2.8818, "step": 23560 }, { "epoch": 1.85, "learning_rate": 8.152513134164511e-05, "loss": 2.8064, "step": 23561 }, { "epoch": 1.85, "learning_rate": 8.152434721242061e-05, "loss": 2.7613, "step": 23562 }, { "epoch": 1.85, "learning_rate": 8.152356308319612e-05, "loss": 2.9485, "step": 23563 }, { "epoch": 1.85, "learning_rate": 8.152277895397162e-05, "loss": 2.7205, "step": 23564 }, { "epoch": 1.85, "learning_rate": 8.152199482474712e-05, "loss": 2.8935, "step": 23565 }, { "epoch": 1.85, "learning_rate": 8.152121069552261e-05, "loss": 2.9237, "step": 23566 }, { "epoch": 1.85, "learning_rate": 8.152042656629813e-05, "loss": 3.0465, "step": 23567 }, { "epoch": 1.85, "learning_rate": 8.151964243707364e-05, "loss": 2.977, "step": 23568 }, { "epoch": 1.85, "learning_rate": 8.151885830784914e-05, "loss": 2.9896, "step": 23569 }, { "epoch": 1.85, "learning_rate": 8.151807417862464e-05, "loss": 2.8831, "step": 23570 }, { "epoch": 1.85, "learning_rate": 8.151729004940015e-05, "loss": 3.1155, "step": 23571 }, { "epoch": 1.85, "learning_rate": 8.151650592017565e-05, "loss": 2.9441, "step": 23572 }, { "epoch": 1.85, "learning_rate": 8.151572179095115e-05, "loss": 2.8991, "step": 23573 }, { "epoch": 1.85, "learning_rate": 8.151493766172666e-05, "loss": 2.7629, "step": 23574 }, { "epoch": 1.85, "learning_rate": 8.151415353250216e-05, "loss": 3.1701, "step": 23575 }, { "epoch": 1.85, "learning_rate": 8.151336940327766e-05, "loss": 2.7374, "step": 23576 }, { "epoch": 1.85, "learning_rate": 8.151258527405317e-05, "loss": 2.9576, "step": 23577 }, { "epoch": 1.85, "learning_rate": 8.151180114482867e-05, "loss": 2.809, "step": 23578 }, { "epoch": 1.85, "learning_rate": 8.151101701560418e-05, "loss": 2.9728, "step": 23579 }, { "epoch": 1.85, "learning_rate": 8.151023288637968e-05, "loss": 2.9756, "step": 23580 }, { "epoch": 1.85, "learning_rate": 8.150944875715518e-05, "loss": 2.9093, "step": 23581 }, { "epoch": 1.85, "learning_rate": 8.150866462793069e-05, "loss": 2.9054, "step": 23582 }, { "epoch": 1.85, "learning_rate": 8.150788049870619e-05, "loss": 3.0394, "step": 23583 }, { "epoch": 1.85, "learning_rate": 8.15070963694817e-05, "loss": 3.0716, "step": 23584 }, { "epoch": 1.85, "learning_rate": 8.15063122402572e-05, "loss": 2.7938, "step": 23585 }, { "epoch": 1.85, "learning_rate": 8.15055281110327e-05, "loss": 2.981, "step": 23586 }, { "epoch": 1.85, "learning_rate": 8.15047439818082e-05, "loss": 3.1304, "step": 23587 }, { "epoch": 1.85, "learning_rate": 8.150395985258371e-05, "loss": 3.0731, "step": 23588 }, { "epoch": 1.85, "learning_rate": 8.150317572335921e-05, "loss": 2.9663, "step": 23589 }, { "epoch": 1.85, "learning_rate": 8.150239159413472e-05, "loss": 2.9574, "step": 23590 }, { "epoch": 1.85, "learning_rate": 8.150160746491022e-05, "loss": 2.8113, "step": 23591 }, { "epoch": 1.85, "learning_rate": 8.150082333568572e-05, "loss": 2.8809, "step": 23592 }, { "epoch": 1.85, "learning_rate": 8.150003920646123e-05, "loss": 3.0472, "step": 23593 }, { "epoch": 1.85, "learning_rate": 8.149925507723673e-05, "loss": 3.0735, "step": 23594 }, { "epoch": 1.85, "learning_rate": 8.149847094801223e-05, "loss": 3.0024, "step": 23595 }, { "epoch": 1.85, "learning_rate": 8.149768681878774e-05, "loss": 2.8598, "step": 23596 }, { "epoch": 1.85, "learning_rate": 8.149690268956324e-05, "loss": 2.9398, "step": 23597 }, { "epoch": 1.85, "learning_rate": 8.149611856033875e-05, "loss": 2.809, "step": 23598 }, { "epoch": 1.85, "learning_rate": 8.149533443111426e-05, "loss": 2.8622, "step": 23599 }, { "epoch": 1.85, "learning_rate": 8.149455030188977e-05, "loss": 2.9578, "step": 23600 }, { "epoch": 1.85, "learning_rate": 8.149376617266527e-05, "loss": 2.8199, "step": 23601 }, { "epoch": 1.85, "learning_rate": 8.149298204344076e-05, "loss": 2.7763, "step": 23602 }, { "epoch": 1.85, "learning_rate": 8.149219791421626e-05, "loss": 2.8342, "step": 23603 }, { "epoch": 1.85, "learning_rate": 8.149141378499177e-05, "loss": 2.8646, "step": 23604 }, { "epoch": 1.85, "learning_rate": 8.149062965576727e-05, "loss": 2.7063, "step": 23605 }, { "epoch": 1.85, "learning_rate": 8.148984552654277e-05, "loss": 2.8497, "step": 23606 }, { "epoch": 1.85, "learning_rate": 8.148906139731828e-05, "loss": 2.9881, "step": 23607 }, { "epoch": 1.85, "learning_rate": 8.148827726809378e-05, "loss": 3.0369, "step": 23608 }, { "epoch": 1.85, "learning_rate": 8.148749313886929e-05, "loss": 2.8327, "step": 23609 }, { "epoch": 1.85, "learning_rate": 8.148670900964479e-05, "loss": 2.8208, "step": 23610 }, { "epoch": 1.85, "learning_rate": 8.14859248804203e-05, "loss": 2.692, "step": 23611 }, { "epoch": 1.85, "learning_rate": 8.14851407511958e-05, "loss": 2.9188, "step": 23612 }, { "epoch": 1.85, "learning_rate": 8.14843566219713e-05, "loss": 2.6527, "step": 23613 }, { "epoch": 1.85, "learning_rate": 8.14835724927468e-05, "loss": 2.8747, "step": 23614 }, { "epoch": 1.85, "learning_rate": 8.148278836352231e-05, "loss": 3.064, "step": 23615 }, { "epoch": 1.85, "learning_rate": 8.148200423429781e-05, "loss": 2.9896, "step": 23616 }, { "epoch": 1.85, "learning_rate": 8.148122010507332e-05, "loss": 3.0849, "step": 23617 }, { "epoch": 1.85, "learning_rate": 8.148043597584882e-05, "loss": 2.7609, "step": 23618 }, { "epoch": 1.85, "learning_rate": 8.147965184662432e-05, "loss": 3.1152, "step": 23619 }, { "epoch": 1.85, "learning_rate": 8.147886771739983e-05, "loss": 2.9096, "step": 23620 }, { "epoch": 1.85, "learning_rate": 8.147808358817533e-05, "loss": 2.9721, "step": 23621 }, { "epoch": 1.85, "learning_rate": 8.147729945895083e-05, "loss": 2.9788, "step": 23622 }, { "epoch": 1.85, "learning_rate": 8.147651532972634e-05, "loss": 2.9718, "step": 23623 }, { "epoch": 1.85, "learning_rate": 8.147573120050185e-05, "loss": 2.9981, "step": 23624 }, { "epoch": 1.85, "learning_rate": 8.147494707127736e-05, "loss": 2.6632, "step": 23625 }, { "epoch": 1.85, "learning_rate": 8.147416294205286e-05, "loss": 2.7962, "step": 23626 }, { "epoch": 1.85, "learning_rate": 8.147337881282837e-05, "loss": 2.9382, "step": 23627 }, { "epoch": 1.85, "learning_rate": 8.147259468360387e-05, "loss": 3.3045, "step": 23628 }, { "epoch": 1.85, "learning_rate": 8.147181055437937e-05, "loss": 2.8423, "step": 23629 }, { "epoch": 1.85, "learning_rate": 8.147102642515488e-05, "loss": 2.8579, "step": 23630 }, { "epoch": 1.85, "learning_rate": 8.147024229593038e-05, "loss": 2.9269, "step": 23631 }, { "epoch": 1.85, "learning_rate": 8.146945816670588e-05, "loss": 3.0136, "step": 23632 }, { "epoch": 1.85, "learning_rate": 8.146867403748139e-05, "loss": 2.8809, "step": 23633 }, { "epoch": 1.85, "learning_rate": 8.146788990825689e-05, "loss": 2.7157, "step": 23634 }, { "epoch": 1.85, "learning_rate": 8.14671057790324e-05, "loss": 2.8392, "step": 23635 }, { "epoch": 1.85, "learning_rate": 8.14663216498079e-05, "loss": 3.0252, "step": 23636 }, { "epoch": 1.85, "learning_rate": 8.14655375205834e-05, "loss": 3.0356, "step": 23637 }, { "epoch": 1.85, "learning_rate": 8.146475339135889e-05, "loss": 3.0213, "step": 23638 }, { "epoch": 1.85, "learning_rate": 8.14639692621344e-05, "loss": 2.6571, "step": 23639 }, { "epoch": 1.85, "learning_rate": 8.14631851329099e-05, "loss": 2.7493, "step": 23640 }, { "epoch": 1.85, "learning_rate": 8.14624010036854e-05, "loss": 2.7449, "step": 23641 }, { "epoch": 1.85, "learning_rate": 8.146161687446091e-05, "loss": 2.9834, "step": 23642 }, { "epoch": 1.85, "learning_rate": 8.146083274523641e-05, "loss": 2.573, "step": 23643 }, { "epoch": 1.85, "learning_rate": 8.146004861601191e-05, "loss": 2.9454, "step": 23644 }, { "epoch": 1.85, "learning_rate": 8.145926448678742e-05, "loss": 3.013, "step": 23645 }, { "epoch": 1.85, "learning_rate": 8.145848035756292e-05, "loss": 2.8918, "step": 23646 }, { "epoch": 1.85, "learning_rate": 8.145769622833843e-05, "loss": 2.8484, "step": 23647 }, { "epoch": 1.85, "learning_rate": 8.145691209911393e-05, "loss": 3.0252, "step": 23648 }, { "epoch": 1.85, "learning_rate": 8.145612796988943e-05, "loss": 2.846, "step": 23649 }, { "epoch": 1.85, "learning_rate": 8.145534384066495e-05, "loss": 2.7641, "step": 23650 }, { "epoch": 1.85, "learning_rate": 8.145455971144045e-05, "loss": 2.9898, "step": 23651 }, { "epoch": 1.85, "learning_rate": 8.145377558221596e-05, "loss": 3.0303, "step": 23652 }, { "epoch": 1.85, "learning_rate": 8.145299145299146e-05, "loss": 3.0793, "step": 23653 }, { "epoch": 1.85, "learning_rate": 8.145220732376697e-05, "loss": 2.9621, "step": 23654 }, { "epoch": 1.85, "learning_rate": 8.145142319454247e-05, "loss": 2.8693, "step": 23655 }, { "epoch": 1.85, "learning_rate": 8.145063906531797e-05, "loss": 2.7987, "step": 23656 }, { "epoch": 1.86, "learning_rate": 8.144985493609348e-05, "loss": 2.6398, "step": 23657 }, { "epoch": 1.86, "learning_rate": 8.144907080686898e-05, "loss": 2.9058, "step": 23658 }, { "epoch": 1.86, "learning_rate": 8.144828667764448e-05, "loss": 2.9821, "step": 23659 }, { "epoch": 1.86, "learning_rate": 8.144750254841999e-05, "loss": 2.9582, "step": 23660 }, { "epoch": 1.86, "learning_rate": 8.144671841919549e-05, "loss": 2.7441, "step": 23661 }, { "epoch": 1.86, "learning_rate": 8.1445934289971e-05, "loss": 2.7726, "step": 23662 }, { "epoch": 1.86, "learning_rate": 8.14451501607465e-05, "loss": 2.742, "step": 23663 }, { "epoch": 1.86, "learning_rate": 8.1444366031522e-05, "loss": 2.8632, "step": 23664 }, { "epoch": 1.86, "learning_rate": 8.14435819022975e-05, "loss": 2.9843, "step": 23665 }, { "epoch": 1.86, "learning_rate": 8.144279777307301e-05, "loss": 2.6967, "step": 23666 }, { "epoch": 1.86, "learning_rate": 8.144201364384851e-05, "loss": 2.795, "step": 23667 }, { "epoch": 1.86, "learning_rate": 8.144122951462402e-05, "loss": 2.804, "step": 23668 }, { "epoch": 1.86, "learning_rate": 8.144044538539952e-05, "loss": 2.9944, "step": 23669 }, { "epoch": 1.86, "learning_rate": 8.143966125617502e-05, "loss": 3.097, "step": 23670 }, { "epoch": 1.86, "learning_rate": 8.143887712695053e-05, "loss": 2.8258, "step": 23671 }, { "epoch": 1.86, "learning_rate": 8.143809299772603e-05, "loss": 2.8254, "step": 23672 }, { "epoch": 1.86, "learning_rate": 8.143730886850153e-05, "loss": 3.2407, "step": 23673 }, { "epoch": 1.86, "learning_rate": 8.143652473927703e-05, "loss": 2.9334, "step": 23674 }, { "epoch": 1.86, "learning_rate": 8.143574061005254e-05, "loss": 2.8474, "step": 23675 }, { "epoch": 1.86, "learning_rate": 8.143495648082805e-05, "loss": 2.9241, "step": 23676 }, { "epoch": 1.86, "learning_rate": 8.143417235160355e-05, "loss": 2.9743, "step": 23677 }, { "epoch": 1.86, "learning_rate": 8.143338822237905e-05, "loss": 2.7591, "step": 23678 }, { "epoch": 1.86, "learning_rate": 8.143260409315456e-05, "loss": 3.1142, "step": 23679 }, { "epoch": 1.86, "learning_rate": 8.143181996393006e-05, "loss": 2.8356, "step": 23680 }, { "epoch": 1.86, "learning_rate": 8.143103583470556e-05, "loss": 2.8533, "step": 23681 }, { "epoch": 1.86, "learning_rate": 8.143025170548107e-05, "loss": 2.8188, "step": 23682 }, { "epoch": 1.86, "learning_rate": 8.142946757625657e-05, "loss": 2.7775, "step": 23683 }, { "epoch": 1.86, "learning_rate": 8.142868344703208e-05, "loss": 3.092, "step": 23684 }, { "epoch": 1.86, "learning_rate": 8.142789931780758e-05, "loss": 2.8523, "step": 23685 }, { "epoch": 1.86, "learning_rate": 8.142711518858308e-05, "loss": 2.7942, "step": 23686 }, { "epoch": 1.86, "learning_rate": 8.142633105935859e-05, "loss": 3.0448, "step": 23687 }, { "epoch": 1.86, "learning_rate": 8.142554693013409e-05, "loss": 2.6564, "step": 23688 }, { "epoch": 1.86, "learning_rate": 8.14247628009096e-05, "loss": 2.7275, "step": 23689 }, { "epoch": 1.86, "learning_rate": 8.14239786716851e-05, "loss": 3.022, "step": 23690 }, { "epoch": 1.86, "learning_rate": 8.14231945424606e-05, "loss": 2.7709, "step": 23691 }, { "epoch": 1.86, "learning_rate": 8.14224104132361e-05, "loss": 2.7736, "step": 23692 }, { "epoch": 1.86, "learning_rate": 8.142162628401161e-05, "loss": 2.8569, "step": 23693 }, { "epoch": 1.86, "learning_rate": 8.142084215478711e-05, "loss": 2.96, "step": 23694 }, { "epoch": 1.86, "learning_rate": 8.142005802556262e-05, "loss": 2.7082, "step": 23695 }, { "epoch": 1.86, "learning_rate": 8.141927389633812e-05, "loss": 2.8744, "step": 23696 }, { "epoch": 1.86, "learning_rate": 8.141848976711362e-05, "loss": 2.8415, "step": 23697 }, { "epoch": 1.86, "learning_rate": 8.141770563788913e-05, "loss": 2.8062, "step": 23698 }, { "epoch": 1.86, "learning_rate": 8.141692150866463e-05, "loss": 2.9584, "step": 23699 }, { "epoch": 1.86, "learning_rate": 8.141613737944013e-05, "loss": 2.8756, "step": 23700 }, { "epoch": 1.86, "learning_rate": 8.141535325021564e-05, "loss": 2.764, "step": 23701 }, { "epoch": 1.86, "learning_rate": 8.141456912099114e-05, "loss": 2.7534, "step": 23702 }, { "epoch": 1.86, "learning_rate": 8.141378499176665e-05, "loss": 2.9785, "step": 23703 }, { "epoch": 1.86, "learning_rate": 8.141300086254215e-05, "loss": 3.0611, "step": 23704 }, { "epoch": 1.86, "learning_rate": 8.141221673331765e-05, "loss": 2.8222, "step": 23705 }, { "epoch": 1.86, "learning_rate": 8.141143260409316e-05, "loss": 2.736, "step": 23706 }, { "epoch": 1.86, "learning_rate": 8.141064847486867e-05, "loss": 3.0363, "step": 23707 }, { "epoch": 1.86, "learning_rate": 8.140986434564418e-05, "loss": 3.2223, "step": 23708 }, { "epoch": 1.86, "learning_rate": 8.140908021641968e-05, "loss": 2.7143, "step": 23709 }, { "epoch": 1.86, "learning_rate": 8.140829608719517e-05, "loss": 3.025, "step": 23710 }, { "epoch": 1.86, "learning_rate": 8.140751195797067e-05, "loss": 2.8154, "step": 23711 }, { "epoch": 1.86, "learning_rate": 8.140672782874618e-05, "loss": 2.8589, "step": 23712 }, { "epoch": 1.86, "learning_rate": 8.140594369952168e-05, "loss": 2.8883, "step": 23713 }, { "epoch": 1.86, "learning_rate": 8.140515957029719e-05, "loss": 2.7684, "step": 23714 }, { "epoch": 1.86, "learning_rate": 8.140437544107269e-05, "loss": 2.5445, "step": 23715 }, { "epoch": 1.86, "learning_rate": 8.140359131184819e-05, "loss": 2.8239, "step": 23716 }, { "epoch": 1.86, "learning_rate": 8.14028071826237e-05, "loss": 2.9738, "step": 23717 }, { "epoch": 1.86, "learning_rate": 8.14020230533992e-05, "loss": 2.8968, "step": 23718 }, { "epoch": 1.86, "learning_rate": 8.14012389241747e-05, "loss": 3.14, "step": 23719 }, { "epoch": 1.86, "learning_rate": 8.140045479495021e-05, "loss": 2.6933, "step": 23720 }, { "epoch": 1.86, "learning_rate": 8.139967066572571e-05, "loss": 2.8172, "step": 23721 }, { "epoch": 1.86, "learning_rate": 8.139888653650122e-05, "loss": 2.8894, "step": 23722 }, { "epoch": 1.86, "learning_rate": 8.139810240727672e-05, "loss": 2.871, "step": 23723 }, { "epoch": 1.86, "learning_rate": 8.139731827805222e-05, "loss": 2.911, "step": 23724 }, { "epoch": 1.86, "learning_rate": 8.139653414882773e-05, "loss": 3.0069, "step": 23725 }, { "epoch": 1.86, "learning_rate": 8.139575001960323e-05, "loss": 2.7983, "step": 23726 }, { "epoch": 1.86, "learning_rate": 8.139496589037873e-05, "loss": 2.9944, "step": 23727 }, { "epoch": 1.86, "learning_rate": 8.139418176115424e-05, "loss": 2.8, "step": 23728 }, { "epoch": 1.86, "learning_rate": 8.139339763192974e-05, "loss": 2.8056, "step": 23729 }, { "epoch": 1.86, "learning_rate": 8.139261350270524e-05, "loss": 3.0457, "step": 23730 }, { "epoch": 1.86, "learning_rate": 8.139182937348075e-05, "loss": 2.7899, "step": 23731 }, { "epoch": 1.86, "learning_rate": 8.139104524425627e-05, "loss": 2.9145, "step": 23732 }, { "epoch": 1.86, "learning_rate": 8.139026111503177e-05, "loss": 2.848, "step": 23733 }, { "epoch": 1.86, "learning_rate": 8.138947698580727e-05, "loss": 2.8485, "step": 23734 }, { "epoch": 1.86, "learning_rate": 8.138869285658278e-05, "loss": 3.0121, "step": 23735 }, { "epoch": 1.86, "learning_rate": 8.138790872735828e-05, "loss": 2.8357, "step": 23736 }, { "epoch": 1.86, "learning_rate": 8.138712459813378e-05, "loss": 2.7823, "step": 23737 }, { "epoch": 1.86, "learning_rate": 8.138634046890929e-05, "loss": 2.8716, "step": 23738 }, { "epoch": 1.86, "learning_rate": 8.138555633968479e-05, "loss": 2.9028, "step": 23739 }, { "epoch": 1.86, "learning_rate": 8.13847722104603e-05, "loss": 2.8955, "step": 23740 }, { "epoch": 1.86, "learning_rate": 8.13839880812358e-05, "loss": 2.6919, "step": 23741 }, { "epoch": 1.86, "learning_rate": 8.13832039520113e-05, "loss": 2.8723, "step": 23742 }, { "epoch": 1.86, "learning_rate": 8.13824198227868e-05, "loss": 2.9354, "step": 23743 }, { "epoch": 1.86, "learning_rate": 8.138163569356231e-05, "loss": 2.7525, "step": 23744 }, { "epoch": 1.86, "learning_rate": 8.138085156433781e-05, "loss": 2.6389, "step": 23745 }, { "epoch": 1.86, "learning_rate": 8.13800674351133e-05, "loss": 3.1174, "step": 23746 }, { "epoch": 1.86, "learning_rate": 8.137928330588881e-05, "loss": 3.0076, "step": 23747 }, { "epoch": 1.86, "learning_rate": 8.137849917666431e-05, "loss": 3.0577, "step": 23748 }, { "epoch": 1.86, "learning_rate": 8.137771504743981e-05, "loss": 2.8541, "step": 23749 }, { "epoch": 1.86, "learning_rate": 8.137693091821532e-05, "loss": 3.01, "step": 23750 }, { "epoch": 1.86, "learning_rate": 8.137614678899082e-05, "loss": 2.9934, "step": 23751 }, { "epoch": 1.86, "learning_rate": 8.137536265976633e-05, "loss": 2.8416, "step": 23752 }, { "epoch": 1.86, "learning_rate": 8.137457853054183e-05, "loss": 2.9137, "step": 23753 }, { "epoch": 1.86, "learning_rate": 8.137379440131733e-05, "loss": 2.7847, "step": 23754 }, { "epoch": 1.86, "learning_rate": 8.137301027209284e-05, "loss": 2.9414, "step": 23755 }, { "epoch": 1.86, "learning_rate": 8.137222614286834e-05, "loss": 2.9733, "step": 23756 }, { "epoch": 1.86, "learning_rate": 8.137144201364384e-05, "loss": 2.7753, "step": 23757 }, { "epoch": 1.86, "learning_rate": 8.137065788441936e-05, "loss": 2.8413, "step": 23758 }, { "epoch": 1.86, "learning_rate": 8.136987375519486e-05, "loss": 2.8767, "step": 23759 }, { "epoch": 1.86, "learning_rate": 8.136908962597037e-05, "loss": 2.8195, "step": 23760 }, { "epoch": 1.86, "learning_rate": 8.136830549674587e-05, "loss": 2.9999, "step": 23761 }, { "epoch": 1.86, "learning_rate": 8.136752136752138e-05, "loss": 2.9662, "step": 23762 }, { "epoch": 1.86, "learning_rate": 8.136673723829688e-05, "loss": 2.6186, "step": 23763 }, { "epoch": 1.86, "learning_rate": 8.136595310907238e-05, "loss": 2.754, "step": 23764 }, { "epoch": 1.86, "learning_rate": 8.136516897984789e-05, "loss": 2.8555, "step": 23765 }, { "epoch": 1.86, "learning_rate": 8.136438485062339e-05, "loss": 2.7716, "step": 23766 }, { "epoch": 1.86, "learning_rate": 8.13636007213989e-05, "loss": 3.179, "step": 23767 }, { "epoch": 1.86, "learning_rate": 8.13628165921744e-05, "loss": 3.1621, "step": 23768 }, { "epoch": 1.86, "learning_rate": 8.13620324629499e-05, "loss": 3.0742, "step": 23769 }, { "epoch": 1.86, "learning_rate": 8.13612483337254e-05, "loss": 2.8528, "step": 23770 }, { "epoch": 1.86, "learning_rate": 8.136046420450091e-05, "loss": 2.9128, "step": 23771 }, { "epoch": 1.86, "learning_rate": 8.135968007527641e-05, "loss": 2.8539, "step": 23772 }, { "epoch": 1.86, "learning_rate": 8.135889594605192e-05, "loss": 3.0492, "step": 23773 }, { "epoch": 1.86, "learning_rate": 8.135811181682742e-05, "loss": 2.7799, "step": 23774 }, { "epoch": 1.86, "learning_rate": 8.135732768760292e-05, "loss": 2.9608, "step": 23775 }, { "epoch": 1.86, "learning_rate": 8.135654355837843e-05, "loss": 3.1102, "step": 23776 }, { "epoch": 1.86, "learning_rate": 8.135575942915393e-05, "loss": 2.9404, "step": 23777 }, { "epoch": 1.86, "learning_rate": 8.135497529992943e-05, "loss": 2.8426, "step": 23778 }, { "epoch": 1.86, "learning_rate": 8.135419117070494e-05, "loss": 2.8521, "step": 23779 }, { "epoch": 1.86, "learning_rate": 8.135340704148044e-05, "loss": 2.7798, "step": 23780 }, { "epoch": 1.86, "learning_rate": 8.135262291225595e-05, "loss": 2.8739, "step": 23781 }, { "epoch": 1.86, "learning_rate": 8.135183878303144e-05, "loss": 2.6305, "step": 23782 }, { "epoch": 1.86, "learning_rate": 8.135105465380694e-05, "loss": 2.9402, "step": 23783 }, { "epoch": 1.86, "learning_rate": 8.135027052458246e-05, "loss": 3.1336, "step": 23784 }, { "epoch": 1.87, "learning_rate": 8.134948639535796e-05, "loss": 3.023, "step": 23785 }, { "epoch": 1.87, "learning_rate": 8.134870226613346e-05, "loss": 2.9346, "step": 23786 }, { "epoch": 1.87, "learning_rate": 8.134791813690897e-05, "loss": 2.8838, "step": 23787 }, { "epoch": 1.87, "learning_rate": 8.134713400768447e-05, "loss": 2.9223, "step": 23788 }, { "epoch": 1.87, "learning_rate": 8.134634987845998e-05, "loss": 3.0278, "step": 23789 }, { "epoch": 1.87, "learning_rate": 8.134556574923548e-05, "loss": 2.9842, "step": 23790 }, { "epoch": 1.87, "learning_rate": 8.134478162001098e-05, "loss": 2.793, "step": 23791 }, { "epoch": 1.87, "learning_rate": 8.134399749078649e-05, "loss": 2.7172, "step": 23792 }, { "epoch": 1.87, "learning_rate": 8.134321336156199e-05, "loss": 2.7896, "step": 23793 }, { "epoch": 1.87, "learning_rate": 8.13424292323375e-05, "loss": 2.8557, "step": 23794 }, { "epoch": 1.87, "learning_rate": 8.1341645103113e-05, "loss": 2.8862, "step": 23795 }, { "epoch": 1.87, "learning_rate": 8.13408609738885e-05, "loss": 2.82, "step": 23796 }, { "epoch": 1.87, "learning_rate": 8.1340076844664e-05, "loss": 2.8649, "step": 23797 }, { "epoch": 1.87, "learning_rate": 8.133929271543951e-05, "loss": 2.796, "step": 23798 }, { "epoch": 1.87, "learning_rate": 8.133850858621501e-05, "loss": 2.9537, "step": 23799 }, { "epoch": 1.87, "learning_rate": 8.133772445699052e-05, "loss": 2.8622, "step": 23800 }, { "epoch": 1.87, "learning_rate": 8.133694032776602e-05, "loss": 2.8227, "step": 23801 }, { "epoch": 1.87, "learning_rate": 8.133615619854152e-05, "loss": 2.8185, "step": 23802 }, { "epoch": 1.87, "learning_rate": 8.133537206931703e-05, "loss": 3.1337, "step": 23803 }, { "epoch": 1.87, "learning_rate": 8.133458794009253e-05, "loss": 2.9454, "step": 23804 }, { "epoch": 1.87, "learning_rate": 8.133380381086803e-05, "loss": 2.9914, "step": 23805 }, { "epoch": 1.87, "learning_rate": 8.133301968164354e-05, "loss": 2.9672, "step": 23806 }, { "epoch": 1.87, "learning_rate": 8.133223555241904e-05, "loss": 2.952, "step": 23807 }, { "epoch": 1.87, "learning_rate": 8.133145142319454e-05, "loss": 3.1076, "step": 23808 }, { "epoch": 1.87, "learning_rate": 8.133066729397005e-05, "loss": 3.1046, "step": 23809 }, { "epoch": 1.87, "learning_rate": 8.132988316474555e-05, "loss": 3.0264, "step": 23810 }, { "epoch": 1.87, "learning_rate": 8.132909903552106e-05, "loss": 2.7847, "step": 23811 }, { "epoch": 1.87, "learning_rate": 8.132831490629656e-05, "loss": 2.7739, "step": 23812 }, { "epoch": 1.87, "learning_rate": 8.132753077707206e-05, "loss": 2.8774, "step": 23813 }, { "epoch": 1.87, "learning_rate": 8.132674664784757e-05, "loss": 2.8417, "step": 23814 }, { "epoch": 1.87, "learning_rate": 8.132596251862308e-05, "loss": 2.8823, "step": 23815 }, { "epoch": 1.87, "learning_rate": 8.132517838939859e-05, "loss": 3.1373, "step": 23816 }, { "epoch": 1.87, "learning_rate": 8.132439426017409e-05, "loss": 2.9538, "step": 23817 }, { "epoch": 1.87, "learning_rate": 8.132361013094958e-05, "loss": 2.84, "step": 23818 }, { "epoch": 1.87, "learning_rate": 8.132282600172509e-05, "loss": 2.9268, "step": 23819 }, { "epoch": 1.87, "learning_rate": 8.132204187250059e-05, "loss": 2.9397, "step": 23820 }, { "epoch": 1.87, "learning_rate": 8.132125774327609e-05, "loss": 2.8379, "step": 23821 }, { "epoch": 1.87, "learning_rate": 8.13204736140516e-05, "loss": 2.8338, "step": 23822 }, { "epoch": 1.87, "learning_rate": 8.13196894848271e-05, "loss": 2.8363, "step": 23823 }, { "epoch": 1.87, "learning_rate": 8.13189053556026e-05, "loss": 2.8878, "step": 23824 }, { "epoch": 1.87, "learning_rate": 8.131812122637811e-05, "loss": 2.8623, "step": 23825 }, { "epoch": 1.87, "learning_rate": 8.131733709715361e-05, "loss": 2.7072, "step": 23826 }, { "epoch": 1.87, "learning_rate": 8.131655296792911e-05, "loss": 3.0047, "step": 23827 }, { "epoch": 1.87, "learning_rate": 8.131576883870462e-05, "loss": 2.8553, "step": 23828 }, { "epoch": 1.87, "learning_rate": 8.131498470948012e-05, "loss": 3.1471, "step": 23829 }, { "epoch": 1.87, "learning_rate": 8.131420058025563e-05, "loss": 2.8065, "step": 23830 }, { "epoch": 1.87, "learning_rate": 8.131341645103113e-05, "loss": 2.5849, "step": 23831 }, { "epoch": 1.87, "learning_rate": 8.131263232180663e-05, "loss": 3.0304, "step": 23832 }, { "epoch": 1.87, "learning_rate": 8.131184819258214e-05, "loss": 2.8707, "step": 23833 }, { "epoch": 1.87, "learning_rate": 8.131106406335764e-05, "loss": 2.8874, "step": 23834 }, { "epoch": 1.87, "learning_rate": 8.131027993413314e-05, "loss": 3.0041, "step": 23835 }, { "epoch": 1.87, "learning_rate": 8.130949580490865e-05, "loss": 2.8713, "step": 23836 }, { "epoch": 1.87, "learning_rate": 8.130871167568415e-05, "loss": 2.9659, "step": 23837 }, { "epoch": 1.87, "learning_rate": 8.130792754645966e-05, "loss": 2.774, "step": 23838 }, { "epoch": 1.87, "learning_rate": 8.130714341723516e-05, "loss": 2.7831, "step": 23839 }, { "epoch": 1.87, "learning_rate": 8.130635928801068e-05, "loss": 2.8547, "step": 23840 }, { "epoch": 1.87, "learning_rate": 8.130557515878618e-05, "loss": 2.8274, "step": 23841 }, { "epoch": 1.87, "learning_rate": 8.130479102956168e-05, "loss": 3.2009, "step": 23842 }, { "epoch": 1.87, "learning_rate": 8.130400690033719e-05, "loss": 2.8319, "step": 23843 }, { "epoch": 1.87, "learning_rate": 8.130322277111269e-05, "loss": 3.0401, "step": 23844 }, { "epoch": 1.87, "learning_rate": 8.13024386418882e-05, "loss": 2.8672, "step": 23845 }, { "epoch": 1.87, "learning_rate": 8.13016545126637e-05, "loss": 2.8453, "step": 23846 }, { "epoch": 1.87, "learning_rate": 8.13008703834392e-05, "loss": 2.7998, "step": 23847 }, { "epoch": 1.87, "learning_rate": 8.13000862542147e-05, "loss": 3.1316, "step": 23848 }, { "epoch": 1.87, "learning_rate": 8.129930212499021e-05, "loss": 2.83, "step": 23849 }, { "epoch": 1.87, "learning_rate": 8.129851799576571e-05, "loss": 2.9363, "step": 23850 }, { "epoch": 1.87, "learning_rate": 8.129773386654122e-05, "loss": 2.9154, "step": 23851 }, { "epoch": 1.87, "learning_rate": 8.129694973731672e-05, "loss": 2.6796, "step": 23852 }, { "epoch": 1.87, "learning_rate": 8.129616560809222e-05, "loss": 2.9499, "step": 23853 }, { "epoch": 1.87, "learning_rate": 8.129538147886771e-05, "loss": 2.8724, "step": 23854 }, { "epoch": 1.87, "learning_rate": 8.129459734964322e-05, "loss": 2.8649, "step": 23855 }, { "epoch": 1.87, "learning_rate": 8.129381322041872e-05, "loss": 2.9244, "step": 23856 }, { "epoch": 1.87, "learning_rate": 8.129302909119423e-05, "loss": 2.9103, "step": 23857 }, { "epoch": 1.87, "learning_rate": 8.129224496196973e-05, "loss": 2.9322, "step": 23858 }, { "epoch": 1.87, "learning_rate": 8.129146083274523e-05, "loss": 2.9956, "step": 23859 }, { "epoch": 1.87, "learning_rate": 8.129067670352074e-05, "loss": 3.016, "step": 23860 }, { "epoch": 1.87, "learning_rate": 8.128989257429624e-05, "loss": 2.7278, "step": 23861 }, { "epoch": 1.87, "learning_rate": 8.128910844507174e-05, "loss": 2.957, "step": 23862 }, { "epoch": 1.87, "learning_rate": 8.128832431584725e-05, "loss": 2.6723, "step": 23863 }, { "epoch": 1.87, "learning_rate": 8.128754018662275e-05, "loss": 2.949, "step": 23864 }, { "epoch": 1.87, "learning_rate": 8.128675605739825e-05, "loss": 2.992, "step": 23865 }, { "epoch": 1.87, "learning_rate": 8.128597192817377e-05, "loss": 2.8245, "step": 23866 }, { "epoch": 1.87, "learning_rate": 8.128518779894928e-05, "loss": 3.0904, "step": 23867 }, { "epoch": 1.87, "learning_rate": 8.128440366972478e-05, "loss": 3.0182, "step": 23868 }, { "epoch": 1.87, "learning_rate": 8.128361954050028e-05, "loss": 2.988, "step": 23869 }, { "epoch": 1.87, "learning_rate": 8.128283541127579e-05, "loss": 2.9624, "step": 23870 }, { "epoch": 1.87, "learning_rate": 8.128205128205129e-05, "loss": 2.925, "step": 23871 }, { "epoch": 1.87, "learning_rate": 8.12812671528268e-05, "loss": 2.9139, "step": 23872 }, { "epoch": 1.87, "learning_rate": 8.12804830236023e-05, "loss": 2.9964, "step": 23873 }, { "epoch": 1.87, "learning_rate": 8.12796988943778e-05, "loss": 2.7553, "step": 23874 }, { "epoch": 1.87, "learning_rate": 8.12789147651533e-05, "loss": 2.8908, "step": 23875 }, { "epoch": 1.87, "learning_rate": 8.127813063592881e-05, "loss": 2.8652, "step": 23876 }, { "epoch": 1.87, "learning_rate": 8.127734650670431e-05, "loss": 2.987, "step": 23877 }, { "epoch": 1.87, "learning_rate": 8.127656237747982e-05, "loss": 2.9447, "step": 23878 }, { "epoch": 1.87, "learning_rate": 8.127577824825532e-05, "loss": 2.8294, "step": 23879 }, { "epoch": 1.87, "learning_rate": 8.127499411903082e-05, "loss": 2.9376, "step": 23880 }, { "epoch": 1.87, "learning_rate": 8.127420998980633e-05, "loss": 3.0036, "step": 23881 }, { "epoch": 1.87, "learning_rate": 8.127342586058183e-05, "loss": 2.8976, "step": 23882 }, { "epoch": 1.87, "learning_rate": 8.127264173135733e-05, "loss": 2.9077, "step": 23883 }, { "epoch": 1.87, "learning_rate": 8.127185760213284e-05, "loss": 3.084, "step": 23884 }, { "epoch": 1.87, "learning_rate": 8.127107347290834e-05, "loss": 2.9026, "step": 23885 }, { "epoch": 1.87, "learning_rate": 8.127028934368385e-05, "loss": 2.8849, "step": 23886 }, { "epoch": 1.87, "learning_rate": 8.126950521445935e-05, "loss": 2.7746, "step": 23887 }, { "epoch": 1.87, "learning_rate": 8.126872108523485e-05, "loss": 2.8608, "step": 23888 }, { "epoch": 1.87, "learning_rate": 8.126793695601036e-05, "loss": 3.003, "step": 23889 }, { "epoch": 1.87, "learning_rate": 8.126715282678585e-05, "loss": 2.8512, "step": 23890 }, { "epoch": 1.87, "learning_rate": 8.126636869756135e-05, "loss": 2.8135, "step": 23891 }, { "epoch": 1.87, "learning_rate": 8.126558456833687e-05, "loss": 2.8668, "step": 23892 }, { "epoch": 1.87, "learning_rate": 8.126480043911237e-05, "loss": 2.7889, "step": 23893 }, { "epoch": 1.87, "learning_rate": 8.126401630988787e-05, "loss": 2.9992, "step": 23894 }, { "epoch": 1.87, "learning_rate": 8.126323218066338e-05, "loss": 3.0531, "step": 23895 }, { "epoch": 1.87, "learning_rate": 8.126244805143888e-05, "loss": 2.9332, "step": 23896 }, { "epoch": 1.87, "learning_rate": 8.126166392221439e-05, "loss": 2.7528, "step": 23897 }, { "epoch": 1.87, "learning_rate": 8.126087979298989e-05, "loss": 2.8156, "step": 23898 }, { "epoch": 1.87, "learning_rate": 8.126009566376539e-05, "loss": 3.0239, "step": 23899 }, { "epoch": 1.87, "learning_rate": 8.12593115345409e-05, "loss": 3.2032, "step": 23900 }, { "epoch": 1.87, "learning_rate": 8.12585274053164e-05, "loss": 2.8648, "step": 23901 }, { "epoch": 1.87, "learning_rate": 8.12577432760919e-05, "loss": 2.8455, "step": 23902 }, { "epoch": 1.87, "learning_rate": 8.125695914686741e-05, "loss": 2.8448, "step": 23903 }, { "epoch": 1.87, "learning_rate": 8.125617501764291e-05, "loss": 2.8741, "step": 23904 }, { "epoch": 1.87, "learning_rate": 8.125539088841842e-05, "loss": 2.7895, "step": 23905 }, { "epoch": 1.87, "learning_rate": 8.125460675919392e-05, "loss": 2.7776, "step": 23906 }, { "epoch": 1.87, "learning_rate": 8.125382262996942e-05, "loss": 2.9495, "step": 23907 }, { "epoch": 1.87, "learning_rate": 8.125303850074493e-05, "loss": 2.9142, "step": 23908 }, { "epoch": 1.87, "learning_rate": 8.125225437152043e-05, "loss": 2.8599, "step": 23909 }, { "epoch": 1.87, "learning_rate": 8.125147024229593e-05, "loss": 3.1803, "step": 23910 }, { "epoch": 1.87, "learning_rate": 8.125068611307144e-05, "loss": 2.959, "step": 23911 }, { "epoch": 1.88, "learning_rate": 8.124990198384694e-05, "loss": 3.1544, "step": 23912 }, { "epoch": 1.88, "learning_rate": 8.124911785462244e-05, "loss": 2.9162, "step": 23913 }, { "epoch": 1.88, "learning_rate": 8.124833372539795e-05, "loss": 3.0533, "step": 23914 }, { "epoch": 1.88, "learning_rate": 8.124754959617345e-05, "loss": 2.9578, "step": 23915 }, { "epoch": 1.88, "learning_rate": 8.124676546694896e-05, "loss": 2.8405, "step": 23916 }, { "epoch": 1.88, "learning_rate": 8.124598133772446e-05, "loss": 2.8334, "step": 23917 }, { "epoch": 1.88, "learning_rate": 8.124519720849996e-05, "loss": 2.8568, "step": 23918 }, { "epoch": 1.88, "learning_rate": 8.124441307927547e-05, "loss": 2.8134, "step": 23919 }, { "epoch": 1.88, "learning_rate": 8.124362895005097e-05, "loss": 2.985, "step": 23920 }, { "epoch": 1.88, "learning_rate": 8.124284482082647e-05, "loss": 2.6671, "step": 23921 }, { "epoch": 1.88, "learning_rate": 8.124206069160198e-05, "loss": 2.7938, "step": 23922 }, { "epoch": 1.88, "learning_rate": 8.12412765623775e-05, "loss": 3.0532, "step": 23923 }, { "epoch": 1.88, "learning_rate": 8.1240492433153e-05, "loss": 2.8081, "step": 23924 }, { "epoch": 1.88, "learning_rate": 8.12397083039285e-05, "loss": 2.9454, "step": 23925 }, { "epoch": 1.88, "learning_rate": 8.123892417470399e-05, "loss": 2.9842, "step": 23926 }, { "epoch": 1.88, "learning_rate": 8.12381400454795e-05, "loss": 2.6848, "step": 23927 }, { "epoch": 1.88, "learning_rate": 8.1237355916255e-05, "loss": 3.0698, "step": 23928 }, { "epoch": 1.88, "learning_rate": 8.12365717870305e-05, "loss": 2.8211, "step": 23929 }, { "epoch": 1.88, "learning_rate": 8.123578765780601e-05, "loss": 2.9255, "step": 23930 }, { "epoch": 1.88, "learning_rate": 8.123500352858151e-05, "loss": 2.9762, "step": 23931 }, { "epoch": 1.88, "learning_rate": 8.123421939935701e-05, "loss": 2.8513, "step": 23932 }, { "epoch": 1.88, "learning_rate": 8.123343527013252e-05, "loss": 3.0233, "step": 23933 }, { "epoch": 1.88, "learning_rate": 8.123265114090802e-05, "loss": 2.7684, "step": 23934 }, { "epoch": 1.88, "learning_rate": 8.123186701168353e-05, "loss": 2.9445, "step": 23935 }, { "epoch": 1.88, "learning_rate": 8.123108288245903e-05, "loss": 3.1941, "step": 23936 }, { "epoch": 1.88, "learning_rate": 8.123029875323453e-05, "loss": 2.9909, "step": 23937 }, { "epoch": 1.88, "learning_rate": 8.122951462401004e-05, "loss": 2.8794, "step": 23938 }, { "epoch": 1.88, "learning_rate": 8.122873049478554e-05, "loss": 3.1558, "step": 23939 }, { "epoch": 1.88, "learning_rate": 8.122794636556104e-05, "loss": 2.9434, "step": 23940 }, { "epoch": 1.88, "learning_rate": 8.122716223633655e-05, "loss": 2.7217, "step": 23941 }, { "epoch": 1.88, "learning_rate": 8.122637810711205e-05, "loss": 2.9625, "step": 23942 }, { "epoch": 1.88, "learning_rate": 8.122559397788756e-05, "loss": 2.6729, "step": 23943 }, { "epoch": 1.88, "learning_rate": 8.122480984866306e-05, "loss": 3.0249, "step": 23944 }, { "epoch": 1.88, "learning_rate": 8.122402571943856e-05, "loss": 2.9327, "step": 23945 }, { "epoch": 1.88, "learning_rate": 8.122324159021407e-05, "loss": 3.1133, "step": 23946 }, { "epoch": 1.88, "learning_rate": 8.122245746098957e-05, "loss": 2.8364, "step": 23947 }, { "epoch": 1.88, "learning_rate": 8.122167333176509e-05, "loss": 3.1925, "step": 23948 }, { "epoch": 1.88, "learning_rate": 8.122088920254059e-05, "loss": 2.74, "step": 23949 }, { "epoch": 1.88, "learning_rate": 8.12201050733161e-05, "loss": 2.9988, "step": 23950 }, { "epoch": 1.88, "learning_rate": 8.12193209440916e-05, "loss": 2.8457, "step": 23951 }, { "epoch": 1.88, "learning_rate": 8.12185368148671e-05, "loss": 3.1201, "step": 23952 }, { "epoch": 1.88, "learning_rate": 8.12177526856426e-05, "loss": 2.7928, "step": 23953 }, { "epoch": 1.88, "learning_rate": 8.121696855641811e-05, "loss": 2.9515, "step": 23954 }, { "epoch": 1.88, "learning_rate": 8.121618442719361e-05, "loss": 3.0117, "step": 23955 }, { "epoch": 1.88, "learning_rate": 8.121540029796912e-05, "loss": 2.8723, "step": 23956 }, { "epoch": 1.88, "learning_rate": 8.121461616874462e-05, "loss": 2.9385, "step": 23957 }, { "epoch": 1.88, "learning_rate": 8.121383203952012e-05, "loss": 2.6727, "step": 23958 }, { "epoch": 1.88, "learning_rate": 8.121304791029563e-05, "loss": 2.9791, "step": 23959 }, { "epoch": 1.88, "learning_rate": 8.121226378107113e-05, "loss": 2.77, "step": 23960 }, { "epoch": 1.88, "learning_rate": 8.121147965184663e-05, "loss": 3.0683, "step": 23961 }, { "epoch": 1.88, "learning_rate": 8.121069552262212e-05, "loss": 3.1559, "step": 23962 }, { "epoch": 1.88, "learning_rate": 8.120991139339763e-05, "loss": 2.7298, "step": 23963 }, { "epoch": 1.88, "learning_rate": 8.120912726417313e-05, "loss": 3.232, "step": 23964 }, { "epoch": 1.88, "learning_rate": 8.120834313494864e-05, "loss": 3.134, "step": 23965 }, { "epoch": 1.88, "learning_rate": 8.120755900572414e-05, "loss": 2.7326, "step": 23966 }, { "epoch": 1.88, "learning_rate": 8.120677487649964e-05, "loss": 2.8605, "step": 23967 }, { "epoch": 1.88, "learning_rate": 8.120599074727515e-05, "loss": 2.8015, "step": 23968 }, { "epoch": 1.88, "learning_rate": 8.120520661805065e-05, "loss": 2.7128, "step": 23969 }, { "epoch": 1.88, "learning_rate": 8.120442248882615e-05, "loss": 2.8856, "step": 23970 }, { "epoch": 1.88, "learning_rate": 8.120363835960166e-05, "loss": 2.8752, "step": 23971 }, { "epoch": 1.88, "learning_rate": 8.120285423037716e-05, "loss": 3.0176, "step": 23972 }, { "epoch": 1.88, "learning_rate": 8.120207010115267e-05, "loss": 2.9875, "step": 23973 }, { "epoch": 1.88, "learning_rate": 8.120128597192818e-05, "loss": 2.814, "step": 23974 }, { "epoch": 1.88, "learning_rate": 8.120050184270369e-05, "loss": 3.0093, "step": 23975 }, { "epoch": 1.88, "learning_rate": 8.119971771347919e-05, "loss": 2.7776, "step": 23976 }, { "epoch": 1.88, "learning_rate": 8.11989335842547e-05, "loss": 2.8867, "step": 23977 }, { "epoch": 1.88, "learning_rate": 8.11981494550302e-05, "loss": 2.9171, "step": 23978 }, { "epoch": 1.88, "learning_rate": 8.11973653258057e-05, "loss": 2.9965, "step": 23979 }, { "epoch": 1.88, "learning_rate": 8.11965811965812e-05, "loss": 2.6929, "step": 23980 }, { "epoch": 1.88, "learning_rate": 8.119579706735671e-05, "loss": 3.1997, "step": 23981 }, { "epoch": 1.88, "learning_rate": 8.119501293813221e-05, "loss": 2.8665, "step": 23982 }, { "epoch": 1.88, "learning_rate": 8.119422880890772e-05, "loss": 2.7525, "step": 23983 }, { "epoch": 1.88, "learning_rate": 8.119344467968322e-05, "loss": 2.8335, "step": 23984 }, { "epoch": 1.88, "learning_rate": 8.119266055045872e-05, "loss": 2.8663, "step": 23985 }, { "epoch": 1.88, "learning_rate": 8.119187642123423e-05, "loss": 2.9749, "step": 23986 }, { "epoch": 1.88, "learning_rate": 8.119109229200973e-05, "loss": 3.1026, "step": 23987 }, { "epoch": 1.88, "learning_rate": 8.119030816278523e-05, "loss": 3.2136, "step": 23988 }, { "epoch": 1.88, "learning_rate": 8.118952403356074e-05, "loss": 2.8314, "step": 23989 }, { "epoch": 1.88, "learning_rate": 8.118873990433624e-05, "loss": 2.8258, "step": 23990 }, { "epoch": 1.88, "learning_rate": 8.118795577511175e-05, "loss": 2.7874, "step": 23991 }, { "epoch": 1.88, "learning_rate": 8.118717164588725e-05, "loss": 3.1317, "step": 23992 }, { "epoch": 1.88, "learning_rate": 8.118638751666275e-05, "loss": 3.1704, "step": 23993 }, { "epoch": 1.88, "learning_rate": 8.118560338743826e-05, "loss": 2.9376, "step": 23994 }, { "epoch": 1.88, "learning_rate": 8.118481925821376e-05, "loss": 2.8631, "step": 23995 }, { "epoch": 1.88, "learning_rate": 8.118403512898926e-05, "loss": 2.9407, "step": 23996 }, { "epoch": 1.88, "learning_rate": 8.118325099976477e-05, "loss": 2.8807, "step": 23997 }, { "epoch": 1.88, "learning_rate": 8.118246687054026e-05, "loss": 2.9254, "step": 23998 }, { "epoch": 1.88, "learning_rate": 8.118168274131576e-05, "loss": 3.0275, "step": 23999 }, { "epoch": 1.88, "learning_rate": 8.118089861209128e-05, "loss": 2.9174, "step": 24000 }, { "epoch": 1.88, "eval_loss": 2.9015953540802, "eval_runtime": 218.0662, "eval_samples_per_second": 51.966, "eval_steps_per_second": 3.251, "step": 24000 }, { "epoch": 1.88, "learning_rate": 8.118011448286678e-05, "loss": 2.8097, "step": 24001 }, { "epoch": 1.88, "learning_rate": 8.117933035364229e-05, "loss": 2.9815, "step": 24002 }, { "epoch": 1.88, "learning_rate": 8.117854622441779e-05, "loss": 2.9594, "step": 24003 }, { "epoch": 1.88, "learning_rate": 8.117776209519329e-05, "loss": 2.9685, "step": 24004 }, { "epoch": 1.88, "learning_rate": 8.11769779659688e-05, "loss": 3.0967, "step": 24005 }, { "epoch": 1.88, "learning_rate": 8.11761938367443e-05, "loss": 2.8251, "step": 24006 }, { "epoch": 1.88, "learning_rate": 8.11754097075198e-05, "loss": 2.7072, "step": 24007 }, { "epoch": 1.88, "learning_rate": 8.117462557829531e-05, "loss": 2.9889, "step": 24008 }, { "epoch": 1.88, "learning_rate": 8.117384144907081e-05, "loss": 3.3617, "step": 24009 }, { "epoch": 1.88, "learning_rate": 8.117305731984631e-05, "loss": 2.7793, "step": 24010 }, { "epoch": 1.88, "learning_rate": 8.117227319062182e-05, "loss": 3.0113, "step": 24011 }, { "epoch": 1.88, "learning_rate": 8.117148906139732e-05, "loss": 2.8138, "step": 24012 }, { "epoch": 1.88, "learning_rate": 8.117070493217283e-05, "loss": 2.8861, "step": 24013 }, { "epoch": 1.88, "learning_rate": 8.116992080294833e-05, "loss": 2.8924, "step": 24014 }, { "epoch": 1.88, "learning_rate": 8.116913667372383e-05, "loss": 2.9394, "step": 24015 }, { "epoch": 1.88, "learning_rate": 8.116835254449934e-05, "loss": 2.6909, "step": 24016 }, { "epoch": 1.88, "learning_rate": 8.116756841527484e-05, "loss": 2.715, "step": 24017 }, { "epoch": 1.88, "learning_rate": 8.116678428605034e-05, "loss": 2.9886, "step": 24018 }, { "epoch": 1.88, "learning_rate": 8.116600015682585e-05, "loss": 2.704, "step": 24019 }, { "epoch": 1.88, "learning_rate": 8.116521602760135e-05, "loss": 2.917, "step": 24020 }, { "epoch": 1.88, "learning_rate": 8.116443189837686e-05, "loss": 2.9528, "step": 24021 }, { "epoch": 1.88, "learning_rate": 8.116364776915236e-05, "loss": 2.7791, "step": 24022 }, { "epoch": 1.88, "learning_rate": 8.116286363992786e-05, "loss": 2.7838, "step": 24023 }, { "epoch": 1.88, "learning_rate": 8.116207951070337e-05, "loss": 2.9492, "step": 24024 }, { "epoch": 1.88, "learning_rate": 8.116129538147887e-05, "loss": 2.6206, "step": 24025 }, { "epoch": 1.88, "learning_rate": 8.116051125225437e-05, "loss": 2.9398, "step": 24026 }, { "epoch": 1.88, "learning_rate": 8.115972712302988e-05, "loss": 2.8414, "step": 24027 }, { "epoch": 1.88, "learning_rate": 8.115894299380538e-05, "loss": 2.7048, "step": 24028 }, { "epoch": 1.88, "learning_rate": 8.115815886458088e-05, "loss": 3.101, "step": 24029 }, { "epoch": 1.88, "learning_rate": 8.115737473535639e-05, "loss": 2.7081, "step": 24030 }, { "epoch": 1.88, "learning_rate": 8.11565906061319e-05, "loss": 3.09, "step": 24031 }, { "epoch": 1.88, "learning_rate": 8.115580647690741e-05, "loss": 3.024, "step": 24032 }, { "epoch": 1.88, "learning_rate": 8.115502234768291e-05, "loss": 2.9573, "step": 24033 }, { "epoch": 1.88, "learning_rate": 8.11542382184584e-05, "loss": 2.6158, "step": 24034 }, { "epoch": 1.88, "learning_rate": 8.115345408923391e-05, "loss": 2.622, "step": 24035 }, { "epoch": 1.88, "learning_rate": 8.115266996000941e-05, "loss": 3.0364, "step": 24036 }, { "epoch": 1.88, "learning_rate": 8.115188583078491e-05, "loss": 2.8849, "step": 24037 }, { "epoch": 1.88, "learning_rate": 8.115110170156042e-05, "loss": 2.9044, "step": 24038 }, { "epoch": 1.88, "learning_rate": 8.115031757233592e-05, "loss": 2.8515, "step": 24039 }, { "epoch": 1.89, "learning_rate": 8.114953344311143e-05, "loss": 2.882, "step": 24040 }, { "epoch": 1.89, "learning_rate": 8.114874931388693e-05, "loss": 2.8537, "step": 24041 }, { "epoch": 1.89, "learning_rate": 8.114796518466243e-05, "loss": 2.8824, "step": 24042 }, { "epoch": 1.89, "learning_rate": 8.114718105543794e-05, "loss": 2.9389, "step": 24043 }, { "epoch": 1.89, "learning_rate": 8.114639692621344e-05, "loss": 2.9227, "step": 24044 }, { "epoch": 1.89, "learning_rate": 8.114561279698894e-05, "loss": 3.0467, "step": 24045 }, { "epoch": 1.89, "learning_rate": 8.114482866776445e-05, "loss": 3.0354, "step": 24046 }, { "epoch": 1.89, "learning_rate": 8.114404453853995e-05, "loss": 2.9759, "step": 24047 }, { "epoch": 1.89, "learning_rate": 8.114326040931545e-05, "loss": 2.6964, "step": 24048 }, { "epoch": 1.89, "learning_rate": 8.114247628009096e-05, "loss": 2.8343, "step": 24049 }, { "epoch": 1.89, "learning_rate": 8.114169215086646e-05, "loss": 3.0371, "step": 24050 }, { "epoch": 1.89, "learning_rate": 8.114090802164197e-05, "loss": 2.8993, "step": 24051 }, { "epoch": 1.89, "learning_rate": 8.114012389241747e-05, "loss": 2.7321, "step": 24052 }, { "epoch": 1.89, "learning_rate": 8.113933976319297e-05, "loss": 2.7549, "step": 24053 }, { "epoch": 1.89, "learning_rate": 8.113855563396848e-05, "loss": 2.7304, "step": 24054 }, { "epoch": 1.89, "learning_rate": 8.113777150474398e-05, "loss": 3.2299, "step": 24055 }, { "epoch": 1.89, "learning_rate": 8.113698737551948e-05, "loss": 2.9481, "step": 24056 }, { "epoch": 1.89, "learning_rate": 8.1136203246295e-05, "loss": 2.9132, "step": 24057 }, { "epoch": 1.89, "learning_rate": 8.11354191170705e-05, "loss": 2.8856, "step": 24058 }, { "epoch": 1.89, "learning_rate": 8.113463498784601e-05, "loss": 2.9693, "step": 24059 }, { "epoch": 1.89, "learning_rate": 8.113385085862151e-05, "loss": 2.8503, "step": 24060 }, { "epoch": 1.89, "learning_rate": 8.113306672939702e-05, "loss": 2.9233, "step": 24061 }, { "epoch": 1.89, "learning_rate": 8.113228260017252e-05, "loss": 2.7587, "step": 24062 }, { "epoch": 1.89, "learning_rate": 8.113149847094802e-05, "loss": 2.7928, "step": 24063 }, { "epoch": 1.89, "learning_rate": 8.113071434172353e-05, "loss": 2.989, "step": 24064 }, { "epoch": 1.89, "learning_rate": 8.112993021249903e-05, "loss": 2.8795, "step": 24065 }, { "epoch": 1.89, "learning_rate": 8.112914608327453e-05, "loss": 2.865, "step": 24066 }, { "epoch": 1.89, "learning_rate": 8.112836195405004e-05, "loss": 2.7524, "step": 24067 }, { "epoch": 1.89, "learning_rate": 8.112757782482554e-05, "loss": 3.1284, "step": 24068 }, { "epoch": 1.89, "learning_rate": 8.112679369560105e-05, "loss": 2.727, "step": 24069 }, { "epoch": 1.89, "learning_rate": 8.112600956637654e-05, "loss": 2.841, "step": 24070 }, { "epoch": 1.89, "learning_rate": 8.112522543715204e-05, "loss": 2.9002, "step": 24071 }, { "epoch": 1.89, "learning_rate": 8.112444130792754e-05, "loss": 2.8447, "step": 24072 }, { "epoch": 1.89, "learning_rate": 8.112365717870305e-05, "loss": 2.8437, "step": 24073 }, { "epoch": 1.89, "learning_rate": 8.112287304947855e-05, "loss": 2.8101, "step": 24074 }, { "epoch": 1.89, "learning_rate": 8.112208892025405e-05, "loss": 2.8437, "step": 24075 }, { "epoch": 1.89, "learning_rate": 8.112130479102956e-05, "loss": 2.8873, "step": 24076 }, { "epoch": 1.89, "learning_rate": 8.112052066180506e-05, "loss": 2.8047, "step": 24077 }, { "epoch": 1.89, "learning_rate": 8.111973653258057e-05, "loss": 2.9735, "step": 24078 }, { "epoch": 1.89, "learning_rate": 8.111895240335607e-05, "loss": 2.9771, "step": 24079 }, { "epoch": 1.89, "learning_rate": 8.111816827413157e-05, "loss": 2.992, "step": 24080 }, { "epoch": 1.89, "learning_rate": 8.111738414490708e-05, "loss": 2.8143, "step": 24081 }, { "epoch": 1.89, "learning_rate": 8.11166000156826e-05, "loss": 2.7885, "step": 24082 }, { "epoch": 1.89, "learning_rate": 8.11158158864581e-05, "loss": 2.9388, "step": 24083 }, { "epoch": 1.89, "learning_rate": 8.11150317572336e-05, "loss": 2.8482, "step": 24084 }, { "epoch": 1.89, "learning_rate": 8.11142476280091e-05, "loss": 2.8685, "step": 24085 }, { "epoch": 1.89, "learning_rate": 8.111346349878461e-05, "loss": 2.8038, "step": 24086 }, { "epoch": 1.89, "learning_rate": 8.111267936956011e-05, "loss": 3.1127, "step": 24087 }, { "epoch": 1.89, "learning_rate": 8.111189524033562e-05, "loss": 2.8475, "step": 24088 }, { "epoch": 1.89, "learning_rate": 8.111111111111112e-05, "loss": 3.0498, "step": 24089 }, { "epoch": 1.89, "learning_rate": 8.111032698188662e-05, "loss": 2.9778, "step": 24090 }, { "epoch": 1.89, "learning_rate": 8.110954285266213e-05, "loss": 3.0013, "step": 24091 }, { "epoch": 1.89, "learning_rate": 8.110875872343763e-05, "loss": 3.1172, "step": 24092 }, { "epoch": 1.89, "learning_rate": 8.110797459421313e-05, "loss": 3.0096, "step": 24093 }, { "epoch": 1.89, "learning_rate": 8.110719046498864e-05, "loss": 3.0443, "step": 24094 }, { "epoch": 1.89, "learning_rate": 8.110640633576414e-05, "loss": 2.9715, "step": 24095 }, { "epoch": 1.89, "learning_rate": 8.110562220653964e-05, "loss": 2.8047, "step": 24096 }, { "epoch": 1.89, "learning_rate": 8.110483807731515e-05, "loss": 2.7987, "step": 24097 }, { "epoch": 1.89, "learning_rate": 8.110405394809065e-05, "loss": 2.9072, "step": 24098 }, { "epoch": 1.89, "learning_rate": 8.110326981886616e-05, "loss": 2.7189, "step": 24099 }, { "epoch": 1.89, "learning_rate": 8.110248568964166e-05, "loss": 3.1521, "step": 24100 }, { "epoch": 1.89, "learning_rate": 8.110170156041716e-05, "loss": 2.7605, "step": 24101 }, { "epoch": 1.89, "learning_rate": 8.110091743119267e-05, "loss": 3.1477, "step": 24102 }, { "epoch": 1.89, "learning_rate": 8.110013330196817e-05, "loss": 2.6806, "step": 24103 }, { "epoch": 1.89, "learning_rate": 8.109934917274367e-05, "loss": 3.0466, "step": 24104 }, { "epoch": 1.89, "learning_rate": 8.109856504351918e-05, "loss": 2.8313, "step": 24105 }, { "epoch": 1.89, "learning_rate": 8.109778091429467e-05, "loss": 2.85, "step": 24106 }, { "epoch": 1.89, "learning_rate": 8.109699678507017e-05, "loss": 2.8249, "step": 24107 }, { "epoch": 1.89, "learning_rate": 8.109621265584569e-05, "loss": 2.9092, "step": 24108 }, { "epoch": 1.89, "learning_rate": 8.109542852662119e-05, "loss": 2.9443, "step": 24109 }, { "epoch": 1.89, "learning_rate": 8.10946443973967e-05, "loss": 2.7951, "step": 24110 }, { "epoch": 1.89, "learning_rate": 8.10938602681722e-05, "loss": 2.6912, "step": 24111 }, { "epoch": 1.89, "learning_rate": 8.10930761389477e-05, "loss": 2.8709, "step": 24112 }, { "epoch": 1.89, "learning_rate": 8.109229200972321e-05, "loss": 2.9735, "step": 24113 }, { "epoch": 1.89, "learning_rate": 8.109150788049871e-05, "loss": 2.7879, "step": 24114 }, { "epoch": 1.89, "learning_rate": 8.109072375127421e-05, "loss": 3.0553, "step": 24115 }, { "epoch": 1.89, "learning_rate": 8.108993962204972e-05, "loss": 2.932, "step": 24116 }, { "epoch": 1.89, "learning_rate": 8.108915549282522e-05, "loss": 2.9597, "step": 24117 }, { "epoch": 1.89, "learning_rate": 8.108837136360073e-05, "loss": 2.9626, "step": 24118 }, { "epoch": 1.89, "learning_rate": 8.108758723437623e-05, "loss": 3.0134, "step": 24119 }, { "epoch": 1.89, "learning_rate": 8.108680310515173e-05, "loss": 2.9538, "step": 24120 }, { "epoch": 1.89, "learning_rate": 8.108601897592724e-05, "loss": 3.0368, "step": 24121 }, { "epoch": 1.89, "learning_rate": 8.108523484670274e-05, "loss": 2.986, "step": 24122 }, { "epoch": 1.89, "learning_rate": 8.108445071747824e-05, "loss": 3.1674, "step": 24123 }, { "epoch": 1.89, "learning_rate": 8.108366658825375e-05, "loss": 3.0635, "step": 24124 }, { "epoch": 1.89, "learning_rate": 8.108288245902925e-05, "loss": 2.7148, "step": 24125 }, { "epoch": 1.89, "learning_rate": 8.108209832980476e-05, "loss": 2.9601, "step": 24126 }, { "epoch": 1.89, "learning_rate": 8.108131420058026e-05, "loss": 2.838, "step": 24127 }, { "epoch": 1.89, "learning_rate": 8.108053007135576e-05, "loss": 2.9433, "step": 24128 }, { "epoch": 1.89, "learning_rate": 8.107974594213127e-05, "loss": 2.7224, "step": 24129 }, { "epoch": 1.89, "learning_rate": 8.107896181290677e-05, "loss": 3.1487, "step": 24130 }, { "epoch": 1.89, "learning_rate": 8.107817768368227e-05, "loss": 2.8261, "step": 24131 }, { "epoch": 1.89, "learning_rate": 8.107739355445778e-05, "loss": 2.8743, "step": 24132 }, { "epoch": 1.89, "learning_rate": 8.107660942523328e-05, "loss": 2.7539, "step": 24133 }, { "epoch": 1.89, "learning_rate": 8.107582529600878e-05, "loss": 2.7981, "step": 24134 }, { "epoch": 1.89, "learning_rate": 8.107504116678429e-05, "loss": 2.9274, "step": 24135 }, { "epoch": 1.89, "learning_rate": 8.107425703755979e-05, "loss": 2.9136, "step": 24136 }, { "epoch": 1.89, "learning_rate": 8.10734729083353e-05, "loss": 2.7118, "step": 24137 }, { "epoch": 1.89, "learning_rate": 8.10726887791108e-05, "loss": 2.6843, "step": 24138 }, { "epoch": 1.89, "learning_rate": 8.107190464988632e-05, "loss": 2.9708, "step": 24139 }, { "epoch": 1.89, "learning_rate": 8.107112052066182e-05, "loss": 3.0069, "step": 24140 }, { "epoch": 1.89, "learning_rate": 8.107033639143732e-05, "loss": 2.7257, "step": 24141 }, { "epoch": 1.89, "learning_rate": 8.106955226221281e-05, "loss": 3.1623, "step": 24142 }, { "epoch": 1.89, "learning_rate": 8.106876813298832e-05, "loss": 2.8657, "step": 24143 }, { "epoch": 1.89, "learning_rate": 8.106798400376382e-05, "loss": 2.8145, "step": 24144 }, { "epoch": 1.89, "learning_rate": 8.106719987453933e-05, "loss": 3.1101, "step": 24145 }, { "epoch": 1.89, "learning_rate": 8.106641574531483e-05, "loss": 2.8872, "step": 24146 }, { "epoch": 1.89, "learning_rate": 8.106563161609033e-05, "loss": 2.6827, "step": 24147 }, { "epoch": 1.89, "learning_rate": 8.106484748686584e-05, "loss": 3.1394, "step": 24148 }, { "epoch": 1.89, "learning_rate": 8.106406335764134e-05, "loss": 2.9639, "step": 24149 }, { "epoch": 1.89, "learning_rate": 8.106327922841684e-05, "loss": 3.0748, "step": 24150 }, { "epoch": 1.89, "learning_rate": 8.106249509919235e-05, "loss": 2.9964, "step": 24151 }, { "epoch": 1.89, "learning_rate": 8.106171096996785e-05, "loss": 2.9402, "step": 24152 }, { "epoch": 1.89, "learning_rate": 8.106092684074335e-05, "loss": 2.8485, "step": 24153 }, { "epoch": 1.89, "learning_rate": 8.106014271151886e-05, "loss": 2.7653, "step": 24154 }, { "epoch": 1.89, "learning_rate": 8.105935858229436e-05, "loss": 2.8614, "step": 24155 }, { "epoch": 1.89, "learning_rate": 8.105857445306987e-05, "loss": 2.9238, "step": 24156 }, { "epoch": 1.89, "learning_rate": 8.105779032384537e-05, "loss": 2.826, "step": 24157 }, { "epoch": 1.89, "learning_rate": 8.105700619462087e-05, "loss": 2.8, "step": 24158 }, { "epoch": 1.89, "learning_rate": 8.105622206539638e-05, "loss": 2.9783, "step": 24159 }, { "epoch": 1.89, "learning_rate": 8.105543793617188e-05, "loss": 2.688, "step": 24160 }, { "epoch": 1.89, "learning_rate": 8.105465380694738e-05, "loss": 2.8857, "step": 24161 }, { "epoch": 1.89, "learning_rate": 8.105386967772289e-05, "loss": 2.7508, "step": 24162 }, { "epoch": 1.89, "learning_rate": 8.105308554849839e-05, "loss": 2.8472, "step": 24163 }, { "epoch": 1.89, "learning_rate": 8.10523014192739e-05, "loss": 2.6916, "step": 24164 }, { "epoch": 1.89, "learning_rate": 8.105151729004941e-05, "loss": 2.7481, "step": 24165 }, { "epoch": 1.89, "learning_rate": 8.105073316082492e-05, "loss": 2.8126, "step": 24166 }, { "epoch": 1.9, "learning_rate": 8.104994903160042e-05, "loss": 2.8843, "step": 24167 }, { "epoch": 1.9, "learning_rate": 8.104916490237592e-05, "loss": 3.2167, "step": 24168 }, { "epoch": 1.9, "learning_rate": 8.104838077315143e-05, "loss": 2.8925, "step": 24169 }, { "epoch": 1.9, "learning_rate": 8.104759664392693e-05, "loss": 3.0152, "step": 24170 }, { "epoch": 1.9, "learning_rate": 8.104681251470243e-05, "loss": 2.9266, "step": 24171 }, { "epoch": 1.9, "learning_rate": 8.104602838547794e-05, "loss": 2.8366, "step": 24172 }, { "epoch": 1.9, "learning_rate": 8.104524425625344e-05, "loss": 3.0021, "step": 24173 }, { "epoch": 1.9, "learning_rate": 8.104446012702895e-05, "loss": 3.1572, "step": 24174 }, { "epoch": 1.9, "learning_rate": 8.104367599780445e-05, "loss": 2.9175, "step": 24175 }, { "epoch": 1.9, "learning_rate": 8.104289186857995e-05, "loss": 2.9114, "step": 24176 }, { "epoch": 1.9, "learning_rate": 8.104210773935546e-05, "loss": 2.6992, "step": 24177 }, { "epoch": 1.9, "learning_rate": 8.104132361013095e-05, "loss": 2.793, "step": 24178 }, { "epoch": 1.9, "learning_rate": 8.104053948090645e-05, "loss": 2.828, "step": 24179 }, { "epoch": 1.9, "learning_rate": 8.103975535168195e-05, "loss": 2.7586, "step": 24180 }, { "epoch": 1.9, "learning_rate": 8.103897122245746e-05, "loss": 2.9184, "step": 24181 }, { "epoch": 1.9, "learning_rate": 8.103818709323296e-05, "loss": 3.142, "step": 24182 }, { "epoch": 1.9, "learning_rate": 8.103740296400846e-05, "loss": 2.9761, "step": 24183 }, { "epoch": 1.9, "learning_rate": 8.103661883478397e-05, "loss": 2.9505, "step": 24184 }, { "epoch": 1.9, "learning_rate": 8.103583470555947e-05, "loss": 2.8892, "step": 24185 }, { "epoch": 1.9, "learning_rate": 8.103505057633498e-05, "loss": 3.2606, "step": 24186 }, { "epoch": 1.9, "learning_rate": 8.103426644711048e-05, "loss": 3.0039, "step": 24187 }, { "epoch": 1.9, "learning_rate": 8.103348231788598e-05, "loss": 2.6573, "step": 24188 }, { "epoch": 1.9, "learning_rate": 8.103269818866149e-05, "loss": 2.934, "step": 24189 }, { "epoch": 1.9, "learning_rate": 8.1031914059437e-05, "loss": 2.8959, "step": 24190 }, { "epoch": 1.9, "learning_rate": 8.103112993021251e-05, "loss": 2.6917, "step": 24191 }, { "epoch": 1.9, "learning_rate": 8.103034580098801e-05, "loss": 2.9929, "step": 24192 }, { "epoch": 1.9, "learning_rate": 8.102956167176352e-05, "loss": 2.7888, "step": 24193 }, { "epoch": 1.9, "learning_rate": 8.102877754253902e-05, "loss": 3.0934, "step": 24194 }, { "epoch": 1.9, "learning_rate": 8.102799341331452e-05, "loss": 2.8122, "step": 24195 }, { "epoch": 1.9, "learning_rate": 8.102720928409003e-05, "loss": 2.9327, "step": 24196 }, { "epoch": 1.9, "learning_rate": 8.102642515486553e-05, "loss": 2.7183, "step": 24197 }, { "epoch": 1.9, "learning_rate": 8.102564102564103e-05, "loss": 2.9654, "step": 24198 }, { "epoch": 1.9, "learning_rate": 8.102485689641654e-05, "loss": 2.8488, "step": 24199 }, { "epoch": 1.9, "learning_rate": 8.102407276719204e-05, "loss": 2.8092, "step": 24200 }, { "epoch": 1.9, "learning_rate": 8.102328863796754e-05, "loss": 2.9238, "step": 24201 }, { "epoch": 1.9, "learning_rate": 8.102250450874305e-05, "loss": 2.7989, "step": 24202 }, { "epoch": 1.9, "learning_rate": 8.102172037951855e-05, "loss": 2.7681, "step": 24203 }, { "epoch": 1.9, "learning_rate": 8.102093625029406e-05, "loss": 2.6896, "step": 24204 }, { "epoch": 1.9, "learning_rate": 8.102015212106956e-05, "loss": 2.934, "step": 24205 }, { "epoch": 1.9, "learning_rate": 8.101936799184506e-05, "loss": 3.167, "step": 24206 }, { "epoch": 1.9, "learning_rate": 8.101858386262057e-05, "loss": 2.674, "step": 24207 }, { "epoch": 1.9, "learning_rate": 8.101779973339607e-05, "loss": 3.0903, "step": 24208 }, { "epoch": 1.9, "learning_rate": 8.101701560417157e-05, "loss": 2.7492, "step": 24209 }, { "epoch": 1.9, "learning_rate": 8.101623147494708e-05, "loss": 3.0858, "step": 24210 }, { "epoch": 1.9, "learning_rate": 8.101544734572258e-05, "loss": 3.0213, "step": 24211 }, { "epoch": 1.9, "learning_rate": 8.101466321649808e-05, "loss": 2.766, "step": 24212 }, { "epoch": 1.9, "learning_rate": 8.101387908727359e-05, "loss": 2.8919, "step": 24213 }, { "epoch": 1.9, "learning_rate": 8.101309495804908e-05, "loss": 3.1655, "step": 24214 }, { "epoch": 1.9, "learning_rate": 8.101231082882458e-05, "loss": 2.5202, "step": 24215 }, { "epoch": 1.9, "learning_rate": 8.10115266996001e-05, "loss": 2.8362, "step": 24216 }, { "epoch": 1.9, "learning_rate": 8.10107425703756e-05, "loss": 2.7995, "step": 24217 }, { "epoch": 1.9, "learning_rate": 8.100995844115111e-05, "loss": 2.8356, "step": 24218 }, { "epoch": 1.9, "learning_rate": 8.100917431192661e-05, "loss": 2.9876, "step": 24219 }, { "epoch": 1.9, "learning_rate": 8.100839018270211e-05, "loss": 2.9419, "step": 24220 }, { "epoch": 1.9, "learning_rate": 8.100760605347762e-05, "loss": 2.9057, "step": 24221 }, { "epoch": 1.9, "learning_rate": 8.100682192425312e-05, "loss": 2.784, "step": 24222 }, { "epoch": 1.9, "learning_rate": 8.100603779502863e-05, "loss": 2.8001, "step": 24223 }, { "epoch": 1.9, "learning_rate": 8.100525366580413e-05, "loss": 2.8953, "step": 24224 }, { "epoch": 1.9, "learning_rate": 8.100446953657963e-05, "loss": 2.9556, "step": 24225 }, { "epoch": 1.9, "learning_rate": 8.100368540735514e-05, "loss": 3.0243, "step": 24226 }, { "epoch": 1.9, "learning_rate": 8.100290127813064e-05, "loss": 3.0978, "step": 24227 }, { "epoch": 1.9, "learning_rate": 8.100211714890614e-05, "loss": 2.8324, "step": 24228 }, { "epoch": 1.9, "learning_rate": 8.100133301968165e-05, "loss": 2.8436, "step": 24229 }, { "epoch": 1.9, "learning_rate": 8.100054889045715e-05, "loss": 2.8254, "step": 24230 }, { "epoch": 1.9, "learning_rate": 8.099976476123265e-05, "loss": 2.8893, "step": 24231 }, { "epoch": 1.9, "learning_rate": 8.099898063200816e-05, "loss": 3.0378, "step": 24232 }, { "epoch": 1.9, "learning_rate": 8.099819650278366e-05, "loss": 2.9308, "step": 24233 }, { "epoch": 1.9, "learning_rate": 8.099741237355917e-05, "loss": 2.8383, "step": 24234 }, { "epoch": 1.9, "learning_rate": 8.099662824433467e-05, "loss": 2.8528, "step": 24235 }, { "epoch": 1.9, "learning_rate": 8.099584411511017e-05, "loss": 2.7421, "step": 24236 }, { "epoch": 1.9, "learning_rate": 8.099505998588568e-05, "loss": 2.6707, "step": 24237 }, { "epoch": 1.9, "learning_rate": 8.099427585666118e-05, "loss": 2.7317, "step": 24238 }, { "epoch": 1.9, "learning_rate": 8.099349172743668e-05, "loss": 2.923, "step": 24239 }, { "epoch": 1.9, "learning_rate": 8.099270759821219e-05, "loss": 2.7173, "step": 24240 }, { "epoch": 1.9, "learning_rate": 8.099192346898769e-05, "loss": 2.7759, "step": 24241 }, { "epoch": 1.9, "learning_rate": 8.09911393397632e-05, "loss": 2.8459, "step": 24242 }, { "epoch": 1.9, "learning_rate": 8.09903552105387e-05, "loss": 3.1099, "step": 24243 }, { "epoch": 1.9, "learning_rate": 8.09895710813142e-05, "loss": 2.7457, "step": 24244 }, { "epoch": 1.9, "learning_rate": 8.09887869520897e-05, "loss": 2.922, "step": 24245 }, { "epoch": 1.9, "learning_rate": 8.098800282286521e-05, "loss": 2.845, "step": 24246 }, { "epoch": 1.9, "learning_rate": 8.098721869364073e-05, "loss": 2.8318, "step": 24247 }, { "epoch": 1.9, "learning_rate": 8.098643456441623e-05, "loss": 3.0424, "step": 24248 }, { "epoch": 1.9, "learning_rate": 8.098565043519173e-05, "loss": 3.0697, "step": 24249 }, { "epoch": 1.9, "learning_rate": 8.098486630596722e-05, "loss": 2.9117, "step": 24250 }, { "epoch": 1.9, "learning_rate": 8.098408217674273e-05, "loss": 2.8578, "step": 24251 }, { "epoch": 1.9, "learning_rate": 8.098329804751823e-05, "loss": 2.8197, "step": 24252 }, { "epoch": 1.9, "learning_rate": 8.098251391829374e-05, "loss": 2.8679, "step": 24253 }, { "epoch": 1.9, "learning_rate": 8.098172978906924e-05, "loss": 2.836, "step": 24254 }, { "epoch": 1.9, "learning_rate": 8.098094565984474e-05, "loss": 2.8513, "step": 24255 }, { "epoch": 1.9, "learning_rate": 8.098016153062025e-05, "loss": 3.0107, "step": 24256 }, { "epoch": 1.9, "learning_rate": 8.097937740139575e-05, "loss": 2.9414, "step": 24257 }, { "epoch": 1.9, "learning_rate": 8.097859327217125e-05, "loss": 2.9, "step": 24258 }, { "epoch": 1.9, "learning_rate": 8.097780914294676e-05, "loss": 2.8625, "step": 24259 }, { "epoch": 1.9, "learning_rate": 8.097702501372226e-05, "loss": 2.8363, "step": 24260 }, { "epoch": 1.9, "learning_rate": 8.097624088449777e-05, "loss": 2.7366, "step": 24261 }, { "epoch": 1.9, "learning_rate": 8.097545675527327e-05, "loss": 2.8881, "step": 24262 }, { "epoch": 1.9, "learning_rate": 8.097467262604877e-05, "loss": 2.8615, "step": 24263 }, { "epoch": 1.9, "learning_rate": 8.097388849682428e-05, "loss": 3.0909, "step": 24264 }, { "epoch": 1.9, "learning_rate": 8.097310436759978e-05, "loss": 2.8444, "step": 24265 }, { "epoch": 1.9, "learning_rate": 8.097232023837528e-05, "loss": 2.7215, "step": 24266 }, { "epoch": 1.9, "learning_rate": 8.097153610915079e-05, "loss": 2.8354, "step": 24267 }, { "epoch": 1.9, "learning_rate": 8.097075197992629e-05, "loss": 2.7841, "step": 24268 }, { "epoch": 1.9, "learning_rate": 8.09699678507018e-05, "loss": 2.7475, "step": 24269 }, { "epoch": 1.9, "learning_rate": 8.09691837214773e-05, "loss": 2.9123, "step": 24270 }, { "epoch": 1.9, "learning_rate": 8.09683995922528e-05, "loss": 3.1894, "step": 24271 }, { "epoch": 1.9, "learning_rate": 8.09676154630283e-05, "loss": 3.0687, "step": 24272 }, { "epoch": 1.9, "learning_rate": 8.096683133380382e-05, "loss": 2.9131, "step": 24273 }, { "epoch": 1.9, "learning_rate": 8.096604720457933e-05, "loss": 2.9377, "step": 24274 }, { "epoch": 1.9, "learning_rate": 8.096526307535483e-05, "loss": 2.8906, "step": 24275 }, { "epoch": 1.9, "learning_rate": 8.096447894613033e-05, "loss": 2.8797, "step": 24276 }, { "epoch": 1.9, "learning_rate": 8.096369481690584e-05, "loss": 2.7762, "step": 24277 }, { "epoch": 1.9, "learning_rate": 8.096291068768134e-05, "loss": 3.0108, "step": 24278 }, { "epoch": 1.9, "learning_rate": 8.096212655845684e-05, "loss": 2.9724, "step": 24279 }, { "epoch": 1.9, "learning_rate": 8.096134242923235e-05, "loss": 2.9669, "step": 24280 }, { "epoch": 1.9, "learning_rate": 8.096055830000785e-05, "loss": 3.1621, "step": 24281 }, { "epoch": 1.9, "learning_rate": 8.095977417078336e-05, "loss": 2.7519, "step": 24282 }, { "epoch": 1.9, "learning_rate": 8.095899004155886e-05, "loss": 2.8801, "step": 24283 }, { "epoch": 1.9, "learning_rate": 8.095820591233436e-05, "loss": 2.843, "step": 24284 }, { "epoch": 1.9, "learning_rate": 8.095742178310987e-05, "loss": 3.005, "step": 24285 }, { "epoch": 1.9, "learning_rate": 8.095663765388536e-05, "loss": 2.681, "step": 24286 }, { "epoch": 1.9, "learning_rate": 8.095585352466086e-05, "loss": 2.9213, "step": 24287 }, { "epoch": 1.9, "learning_rate": 8.095506939543636e-05, "loss": 2.8426, "step": 24288 }, { "epoch": 1.9, "learning_rate": 8.095428526621187e-05, "loss": 2.7536, "step": 24289 }, { "epoch": 1.9, "learning_rate": 8.095350113698737e-05, "loss": 2.9756, "step": 24290 }, { "epoch": 1.9, "learning_rate": 8.095271700776288e-05, "loss": 2.929, "step": 24291 }, { "epoch": 1.9, "learning_rate": 8.095193287853838e-05, "loss": 2.841, "step": 24292 }, { "epoch": 1.9, "learning_rate": 8.095114874931388e-05, "loss": 2.8563, "step": 24293 }, { "epoch": 1.9, "learning_rate": 8.095036462008939e-05, "loss": 2.9735, "step": 24294 }, { "epoch": 1.91, "learning_rate": 8.094958049086489e-05, "loss": 2.8671, "step": 24295 }, { "epoch": 1.91, "learning_rate": 8.09487963616404e-05, "loss": 3.0404, "step": 24296 }, { "epoch": 1.91, "learning_rate": 8.09480122324159e-05, "loss": 2.9948, "step": 24297 }, { "epoch": 1.91, "learning_rate": 8.094722810319141e-05, "loss": 2.8888, "step": 24298 }, { "epoch": 1.91, "learning_rate": 8.094644397396692e-05, "loss": 3.0142, "step": 24299 }, { "epoch": 1.91, "learning_rate": 8.094565984474242e-05, "loss": 2.941, "step": 24300 }, { "epoch": 1.91, "learning_rate": 8.094487571551793e-05, "loss": 2.7692, "step": 24301 }, { "epoch": 1.91, "learning_rate": 8.094409158629343e-05, "loss": 2.8817, "step": 24302 }, { "epoch": 1.91, "learning_rate": 8.094330745706893e-05, "loss": 2.7758, "step": 24303 }, { "epoch": 1.91, "learning_rate": 8.094252332784444e-05, "loss": 2.8667, "step": 24304 }, { "epoch": 1.91, "learning_rate": 8.094173919861994e-05, "loss": 2.972, "step": 24305 }, { "epoch": 1.91, "learning_rate": 8.094095506939544e-05, "loss": 2.8514, "step": 24306 }, { "epoch": 1.91, "learning_rate": 8.094017094017095e-05, "loss": 2.8539, "step": 24307 }, { "epoch": 1.91, "learning_rate": 8.093938681094645e-05, "loss": 2.9851, "step": 24308 }, { "epoch": 1.91, "learning_rate": 8.093860268172196e-05, "loss": 2.8454, "step": 24309 }, { "epoch": 1.91, "learning_rate": 8.093781855249746e-05, "loss": 3.0183, "step": 24310 }, { "epoch": 1.91, "learning_rate": 8.093703442327296e-05, "loss": 3.0082, "step": 24311 }, { "epoch": 1.91, "learning_rate": 8.093625029404847e-05, "loss": 3.0016, "step": 24312 }, { "epoch": 1.91, "learning_rate": 8.093546616482397e-05, "loss": 2.9405, "step": 24313 }, { "epoch": 1.91, "learning_rate": 8.093468203559947e-05, "loss": 2.6455, "step": 24314 }, { "epoch": 1.91, "learning_rate": 8.093389790637498e-05, "loss": 2.7186, "step": 24315 }, { "epoch": 1.91, "learning_rate": 8.093311377715048e-05, "loss": 3.082, "step": 24316 }, { "epoch": 1.91, "learning_rate": 8.093232964792598e-05, "loss": 2.9646, "step": 24317 }, { "epoch": 1.91, "learning_rate": 8.093154551870149e-05, "loss": 2.7913, "step": 24318 }, { "epoch": 1.91, "learning_rate": 8.093076138947699e-05, "loss": 2.979, "step": 24319 }, { "epoch": 1.91, "learning_rate": 8.09299772602525e-05, "loss": 2.6838, "step": 24320 }, { "epoch": 1.91, "learning_rate": 8.0929193131028e-05, "loss": 2.8205, "step": 24321 }, { "epoch": 1.91, "learning_rate": 8.092840900180349e-05, "loss": 2.9578, "step": 24322 }, { "epoch": 1.91, "learning_rate": 8.092762487257899e-05, "loss": 2.8995, "step": 24323 }, { "epoch": 1.91, "learning_rate": 8.092684074335451e-05, "loss": 2.8456, "step": 24324 }, { "epoch": 1.91, "learning_rate": 8.092605661413001e-05, "loss": 2.8348, "step": 24325 }, { "epoch": 1.91, "learning_rate": 8.092527248490552e-05, "loss": 3.0467, "step": 24326 }, { "epoch": 1.91, "learning_rate": 8.092448835568102e-05, "loss": 2.7555, "step": 24327 }, { "epoch": 1.91, "learning_rate": 8.092370422645653e-05, "loss": 2.9346, "step": 24328 }, { "epoch": 1.91, "learning_rate": 8.092292009723203e-05, "loss": 2.9155, "step": 24329 }, { "epoch": 1.91, "learning_rate": 8.092213596800753e-05, "loss": 2.7379, "step": 24330 }, { "epoch": 1.91, "learning_rate": 8.092135183878304e-05, "loss": 2.9892, "step": 24331 }, { "epoch": 1.91, "learning_rate": 8.092056770955854e-05, "loss": 2.8277, "step": 24332 }, { "epoch": 1.91, "learning_rate": 8.091978358033404e-05, "loss": 2.7176, "step": 24333 }, { "epoch": 1.91, "learning_rate": 8.091899945110955e-05, "loss": 3.0472, "step": 24334 }, { "epoch": 1.91, "learning_rate": 8.091821532188505e-05, "loss": 3.2814, "step": 24335 }, { "epoch": 1.91, "learning_rate": 8.091743119266055e-05, "loss": 2.7549, "step": 24336 }, { "epoch": 1.91, "learning_rate": 8.091664706343606e-05, "loss": 2.9949, "step": 24337 }, { "epoch": 1.91, "learning_rate": 8.091586293421156e-05, "loss": 2.757, "step": 24338 }, { "epoch": 1.91, "learning_rate": 8.091507880498707e-05, "loss": 2.803, "step": 24339 }, { "epoch": 1.91, "learning_rate": 8.091429467576257e-05, "loss": 3.0241, "step": 24340 }, { "epoch": 1.91, "learning_rate": 8.091351054653807e-05, "loss": 2.8997, "step": 24341 }, { "epoch": 1.91, "learning_rate": 8.091272641731358e-05, "loss": 3.1274, "step": 24342 }, { "epoch": 1.91, "learning_rate": 8.091194228808908e-05, "loss": 3.0477, "step": 24343 }, { "epoch": 1.91, "learning_rate": 8.091115815886458e-05, "loss": 2.9079, "step": 24344 }, { "epoch": 1.91, "learning_rate": 8.091037402964009e-05, "loss": 2.746, "step": 24345 }, { "epoch": 1.91, "learning_rate": 8.090958990041559e-05, "loss": 2.9885, "step": 24346 }, { "epoch": 1.91, "learning_rate": 8.09088057711911e-05, "loss": 2.8831, "step": 24347 }, { "epoch": 1.91, "learning_rate": 8.09080216419666e-05, "loss": 3.0775, "step": 24348 }, { "epoch": 1.91, "learning_rate": 8.09072375127421e-05, "loss": 2.8991, "step": 24349 }, { "epoch": 1.91, "learning_rate": 8.09064533835176e-05, "loss": 2.983, "step": 24350 }, { "epoch": 1.91, "learning_rate": 8.090566925429311e-05, "loss": 2.7412, "step": 24351 }, { "epoch": 1.91, "learning_rate": 8.090488512506861e-05, "loss": 2.8167, "step": 24352 }, { "epoch": 1.91, "learning_rate": 8.090410099584412e-05, "loss": 2.7472, "step": 24353 }, { "epoch": 1.91, "learning_rate": 8.090331686661962e-05, "loss": 3.0286, "step": 24354 }, { "epoch": 1.91, "learning_rate": 8.090253273739514e-05, "loss": 2.9397, "step": 24355 }, { "epoch": 1.91, "learning_rate": 8.090174860817064e-05, "loss": 2.9898, "step": 24356 }, { "epoch": 1.91, "learning_rate": 8.090096447894615e-05, "loss": 2.731, "step": 24357 }, { "epoch": 1.91, "learning_rate": 8.090018034972164e-05, "loss": 3.0325, "step": 24358 }, { "epoch": 1.91, "learning_rate": 8.089939622049714e-05, "loss": 2.994, "step": 24359 }, { "epoch": 1.91, "learning_rate": 8.089861209127264e-05, "loss": 2.8168, "step": 24360 }, { "epoch": 1.91, "learning_rate": 8.089782796204815e-05, "loss": 2.6654, "step": 24361 }, { "epoch": 1.91, "learning_rate": 8.089704383282365e-05, "loss": 2.9832, "step": 24362 }, { "epoch": 1.91, "learning_rate": 8.089625970359915e-05, "loss": 2.6054, "step": 24363 }, { "epoch": 1.91, "learning_rate": 8.089547557437466e-05, "loss": 2.8941, "step": 24364 }, { "epoch": 1.91, "learning_rate": 8.089469144515016e-05, "loss": 2.6926, "step": 24365 }, { "epoch": 1.91, "learning_rate": 8.089390731592566e-05, "loss": 2.8911, "step": 24366 }, { "epoch": 1.91, "learning_rate": 8.089312318670117e-05, "loss": 3.0077, "step": 24367 }, { "epoch": 1.91, "learning_rate": 8.089233905747667e-05, "loss": 2.9872, "step": 24368 }, { "epoch": 1.91, "learning_rate": 8.089155492825218e-05, "loss": 2.7046, "step": 24369 }, { "epoch": 1.91, "learning_rate": 8.089077079902768e-05, "loss": 2.7386, "step": 24370 }, { "epoch": 1.91, "learning_rate": 8.088998666980318e-05, "loss": 3.063, "step": 24371 }, { "epoch": 1.91, "learning_rate": 8.088920254057869e-05, "loss": 2.8909, "step": 24372 }, { "epoch": 1.91, "learning_rate": 8.088841841135419e-05, "loss": 3.0287, "step": 24373 }, { "epoch": 1.91, "learning_rate": 8.08876342821297e-05, "loss": 2.6361, "step": 24374 }, { "epoch": 1.91, "learning_rate": 8.08868501529052e-05, "loss": 3.0005, "step": 24375 }, { "epoch": 1.91, "learning_rate": 8.08860660236807e-05, "loss": 3.1638, "step": 24376 }, { "epoch": 1.91, "learning_rate": 8.08852818944562e-05, "loss": 2.8903, "step": 24377 }, { "epoch": 1.91, "learning_rate": 8.088449776523171e-05, "loss": 2.9049, "step": 24378 }, { "epoch": 1.91, "learning_rate": 8.088371363600721e-05, "loss": 3.0017, "step": 24379 }, { "epoch": 1.91, "learning_rate": 8.088292950678272e-05, "loss": 2.7846, "step": 24380 }, { "epoch": 1.91, "learning_rate": 8.088214537755823e-05, "loss": 2.8653, "step": 24381 }, { "epoch": 1.91, "learning_rate": 8.088136124833374e-05, "loss": 2.7571, "step": 24382 }, { "epoch": 1.91, "learning_rate": 8.088057711910924e-05, "loss": 2.8121, "step": 24383 }, { "epoch": 1.91, "learning_rate": 8.087979298988474e-05, "loss": 2.6531, "step": 24384 }, { "epoch": 1.91, "learning_rate": 8.087900886066025e-05, "loss": 2.9566, "step": 24385 }, { "epoch": 1.91, "learning_rate": 8.087822473143575e-05, "loss": 2.8245, "step": 24386 }, { "epoch": 1.91, "learning_rate": 8.087744060221126e-05, "loss": 2.9681, "step": 24387 }, { "epoch": 1.91, "learning_rate": 8.087665647298676e-05, "loss": 2.8817, "step": 24388 }, { "epoch": 1.91, "learning_rate": 8.087587234376226e-05, "loss": 2.7367, "step": 24389 }, { "epoch": 1.91, "learning_rate": 8.087508821453777e-05, "loss": 2.8353, "step": 24390 }, { "epoch": 1.91, "learning_rate": 8.087430408531327e-05, "loss": 3.1186, "step": 24391 }, { "epoch": 1.91, "learning_rate": 8.087351995608877e-05, "loss": 2.8484, "step": 24392 }, { "epoch": 1.91, "learning_rate": 8.087273582686428e-05, "loss": 2.7897, "step": 24393 }, { "epoch": 1.91, "learning_rate": 8.087195169763977e-05, "loss": 2.8874, "step": 24394 }, { "epoch": 1.91, "learning_rate": 8.087116756841527e-05, "loss": 2.915, "step": 24395 }, { "epoch": 1.91, "learning_rate": 8.087038343919078e-05, "loss": 3.0702, "step": 24396 }, { "epoch": 1.91, "learning_rate": 8.086959930996628e-05, "loss": 2.8755, "step": 24397 }, { "epoch": 1.91, "learning_rate": 8.086881518074178e-05, "loss": 2.9816, "step": 24398 }, { "epoch": 1.91, "learning_rate": 8.086803105151729e-05, "loss": 2.8745, "step": 24399 }, { "epoch": 1.91, "learning_rate": 8.086724692229279e-05, "loss": 3.1674, "step": 24400 }, { "epoch": 1.91, "learning_rate": 8.08664627930683e-05, "loss": 2.7728, "step": 24401 }, { "epoch": 1.91, "learning_rate": 8.08656786638438e-05, "loss": 2.9083, "step": 24402 }, { "epoch": 1.91, "learning_rate": 8.08648945346193e-05, "loss": 3.0961, "step": 24403 }, { "epoch": 1.91, "learning_rate": 8.08641104053948e-05, "loss": 2.9422, "step": 24404 }, { "epoch": 1.91, "learning_rate": 8.086332627617031e-05, "loss": 3.0594, "step": 24405 }, { "epoch": 1.91, "learning_rate": 8.086254214694583e-05, "loss": 2.9261, "step": 24406 }, { "epoch": 1.91, "learning_rate": 8.086175801772133e-05, "loss": 2.7663, "step": 24407 }, { "epoch": 1.91, "learning_rate": 8.086097388849683e-05, "loss": 2.7738, "step": 24408 }, { "epoch": 1.91, "learning_rate": 8.086018975927234e-05, "loss": 2.9068, "step": 24409 }, { "epoch": 1.91, "learning_rate": 8.085940563004784e-05, "loss": 2.8132, "step": 24410 }, { "epoch": 1.91, "learning_rate": 8.085862150082334e-05, "loss": 2.8091, "step": 24411 }, { "epoch": 1.91, "learning_rate": 8.085783737159885e-05, "loss": 2.8073, "step": 24412 }, { "epoch": 1.91, "learning_rate": 8.085705324237435e-05, "loss": 2.7254, "step": 24413 }, { "epoch": 1.91, "learning_rate": 8.085626911314985e-05, "loss": 2.6617, "step": 24414 }, { "epoch": 1.91, "learning_rate": 8.085548498392536e-05, "loss": 2.9695, "step": 24415 }, { "epoch": 1.91, "learning_rate": 8.085470085470086e-05, "loss": 2.9013, "step": 24416 }, { "epoch": 1.91, "learning_rate": 8.085391672547637e-05, "loss": 3.1533, "step": 24417 }, { "epoch": 1.91, "learning_rate": 8.085313259625187e-05, "loss": 2.8984, "step": 24418 }, { "epoch": 1.91, "learning_rate": 8.085234846702737e-05, "loss": 2.6973, "step": 24419 }, { "epoch": 1.91, "learning_rate": 8.085156433780288e-05, "loss": 2.8208, "step": 24420 }, { "epoch": 1.91, "learning_rate": 8.085078020857838e-05, "loss": 3.1098, "step": 24421 }, { "epoch": 1.92, "learning_rate": 8.084999607935388e-05, "loss": 2.9539, "step": 24422 }, { "epoch": 1.92, "learning_rate": 8.084921195012939e-05, "loss": 2.7165, "step": 24423 }, { "epoch": 1.92, "learning_rate": 8.084842782090489e-05, "loss": 2.9312, "step": 24424 }, { "epoch": 1.92, "learning_rate": 8.08476436916804e-05, "loss": 2.7435, "step": 24425 }, { "epoch": 1.92, "learning_rate": 8.08468595624559e-05, "loss": 2.936, "step": 24426 }, { "epoch": 1.92, "learning_rate": 8.08460754332314e-05, "loss": 2.6515, "step": 24427 }, { "epoch": 1.92, "learning_rate": 8.08452913040069e-05, "loss": 3.1071, "step": 24428 }, { "epoch": 1.92, "learning_rate": 8.084450717478241e-05, "loss": 2.7576, "step": 24429 }, { "epoch": 1.92, "learning_rate": 8.08437230455579e-05, "loss": 2.8021, "step": 24430 }, { "epoch": 1.92, "learning_rate": 8.08429389163334e-05, "loss": 2.9748, "step": 24431 }, { "epoch": 1.92, "learning_rate": 8.084215478710892e-05, "loss": 2.7783, "step": 24432 }, { "epoch": 1.92, "learning_rate": 8.084137065788442e-05, "loss": 2.674, "step": 24433 }, { "epoch": 1.92, "learning_rate": 8.084058652865993e-05, "loss": 2.8543, "step": 24434 }, { "epoch": 1.92, "learning_rate": 8.083980239943543e-05, "loss": 2.9318, "step": 24435 }, { "epoch": 1.92, "learning_rate": 8.083901827021094e-05, "loss": 3.074, "step": 24436 }, { "epoch": 1.92, "learning_rate": 8.083823414098644e-05, "loss": 2.9095, "step": 24437 }, { "epoch": 1.92, "learning_rate": 8.083745001176194e-05, "loss": 3.0658, "step": 24438 }, { "epoch": 1.92, "learning_rate": 8.083666588253745e-05, "loss": 3.0715, "step": 24439 }, { "epoch": 1.92, "learning_rate": 8.083588175331295e-05, "loss": 3.0005, "step": 24440 }, { "epoch": 1.92, "learning_rate": 8.083509762408845e-05, "loss": 2.6636, "step": 24441 }, { "epoch": 1.92, "learning_rate": 8.083431349486396e-05, "loss": 2.8768, "step": 24442 }, { "epoch": 1.92, "learning_rate": 8.083352936563946e-05, "loss": 2.9317, "step": 24443 }, { "epoch": 1.92, "learning_rate": 8.083274523641497e-05, "loss": 2.9771, "step": 24444 }, { "epoch": 1.92, "learning_rate": 8.083196110719047e-05, "loss": 2.6725, "step": 24445 }, { "epoch": 1.92, "learning_rate": 8.083117697796597e-05, "loss": 2.8574, "step": 24446 }, { "epoch": 1.92, "learning_rate": 8.083039284874148e-05, "loss": 2.9401, "step": 24447 }, { "epoch": 1.92, "learning_rate": 8.082960871951698e-05, "loss": 2.9028, "step": 24448 }, { "epoch": 1.92, "learning_rate": 8.082882459029248e-05, "loss": 2.7155, "step": 24449 }, { "epoch": 1.92, "learning_rate": 8.082804046106799e-05, "loss": 3.15, "step": 24450 }, { "epoch": 1.92, "learning_rate": 8.082725633184349e-05, "loss": 3.0989, "step": 24451 }, { "epoch": 1.92, "learning_rate": 8.0826472202619e-05, "loss": 2.9034, "step": 24452 }, { "epoch": 1.92, "learning_rate": 8.08256880733945e-05, "loss": 2.868, "step": 24453 }, { "epoch": 1.92, "learning_rate": 8.082490394417e-05, "loss": 2.8869, "step": 24454 }, { "epoch": 1.92, "learning_rate": 8.08241198149455e-05, "loss": 2.8669, "step": 24455 }, { "epoch": 1.92, "learning_rate": 8.082333568572101e-05, "loss": 2.7764, "step": 24456 }, { "epoch": 1.92, "learning_rate": 8.082255155649651e-05, "loss": 2.8707, "step": 24457 }, { "epoch": 1.92, "learning_rate": 8.082176742727202e-05, "loss": 2.9431, "step": 24458 }, { "epoch": 1.92, "learning_rate": 8.082098329804752e-05, "loss": 2.7597, "step": 24459 }, { "epoch": 1.92, "learning_rate": 8.082019916882302e-05, "loss": 3.1079, "step": 24460 }, { "epoch": 1.92, "learning_rate": 8.081941503959853e-05, "loss": 2.83, "step": 24461 }, { "epoch": 1.92, "learning_rate": 8.081863091037403e-05, "loss": 2.7779, "step": 24462 }, { "epoch": 1.92, "learning_rate": 8.081784678114955e-05, "loss": 3.0234, "step": 24463 }, { "epoch": 1.92, "learning_rate": 8.081706265192505e-05, "loss": 3.0675, "step": 24464 }, { "epoch": 1.92, "learning_rate": 8.081627852270056e-05, "loss": 2.9678, "step": 24465 }, { "epoch": 1.92, "learning_rate": 8.081549439347605e-05, "loss": 3.2726, "step": 24466 }, { "epoch": 1.92, "learning_rate": 8.081471026425155e-05, "loss": 2.8916, "step": 24467 }, { "epoch": 1.92, "learning_rate": 8.081392613502705e-05, "loss": 2.9991, "step": 24468 }, { "epoch": 1.92, "learning_rate": 8.081314200580256e-05, "loss": 2.8665, "step": 24469 }, { "epoch": 1.92, "learning_rate": 8.081235787657806e-05, "loss": 2.6362, "step": 24470 }, { "epoch": 1.92, "learning_rate": 8.081157374735356e-05, "loss": 2.5256, "step": 24471 }, { "epoch": 1.92, "learning_rate": 8.081078961812907e-05, "loss": 2.9408, "step": 24472 }, { "epoch": 1.92, "learning_rate": 8.081000548890457e-05, "loss": 2.8685, "step": 24473 }, { "epoch": 1.92, "learning_rate": 8.080922135968008e-05, "loss": 2.7766, "step": 24474 }, { "epoch": 1.92, "learning_rate": 8.080843723045558e-05, "loss": 2.7447, "step": 24475 }, { "epoch": 1.92, "learning_rate": 8.080765310123108e-05, "loss": 2.8444, "step": 24476 }, { "epoch": 1.92, "learning_rate": 8.080686897200659e-05, "loss": 2.8235, "step": 24477 }, { "epoch": 1.92, "learning_rate": 8.080608484278209e-05, "loss": 2.9607, "step": 24478 }, { "epoch": 1.92, "learning_rate": 8.08053007135576e-05, "loss": 2.8075, "step": 24479 }, { "epoch": 1.92, "learning_rate": 8.08045165843331e-05, "loss": 3.1028, "step": 24480 }, { "epoch": 1.92, "learning_rate": 8.08037324551086e-05, "loss": 2.8914, "step": 24481 }, { "epoch": 1.92, "learning_rate": 8.08029483258841e-05, "loss": 3.2471, "step": 24482 }, { "epoch": 1.92, "learning_rate": 8.080216419665961e-05, "loss": 2.7717, "step": 24483 }, { "epoch": 1.92, "learning_rate": 8.080138006743511e-05, "loss": 2.8698, "step": 24484 }, { "epoch": 1.92, "learning_rate": 8.080059593821062e-05, "loss": 3.0287, "step": 24485 }, { "epoch": 1.92, "learning_rate": 8.079981180898612e-05, "loss": 2.9828, "step": 24486 }, { "epoch": 1.92, "learning_rate": 8.079902767976162e-05, "loss": 3.0622, "step": 24487 }, { "epoch": 1.92, "learning_rate": 8.079824355053713e-05, "loss": 2.9444, "step": 24488 }, { "epoch": 1.92, "learning_rate": 8.079745942131264e-05, "loss": 2.873, "step": 24489 }, { "epoch": 1.92, "learning_rate": 8.079667529208815e-05, "loss": 2.884, "step": 24490 }, { "epoch": 1.92, "learning_rate": 8.079589116286365e-05, "loss": 2.8331, "step": 24491 }, { "epoch": 1.92, "learning_rate": 8.079510703363916e-05, "loss": 2.9726, "step": 24492 }, { "epoch": 1.92, "learning_rate": 8.079432290441466e-05, "loss": 2.9062, "step": 24493 }, { "epoch": 1.92, "learning_rate": 8.079353877519016e-05, "loss": 2.906, "step": 24494 }, { "epoch": 1.92, "learning_rate": 8.079275464596567e-05, "loss": 2.985, "step": 24495 }, { "epoch": 1.92, "learning_rate": 8.079197051674117e-05, "loss": 3.0813, "step": 24496 }, { "epoch": 1.92, "learning_rate": 8.079118638751667e-05, "loss": 2.9498, "step": 24497 }, { "epoch": 1.92, "learning_rate": 8.079040225829218e-05, "loss": 2.7348, "step": 24498 }, { "epoch": 1.92, "learning_rate": 8.078961812906768e-05, "loss": 2.8661, "step": 24499 }, { "epoch": 1.92, "learning_rate": 8.078883399984318e-05, "loss": 2.9645, "step": 24500 }, { "epoch": 1.92, "learning_rate": 8.078804987061869e-05, "loss": 3.1049, "step": 24501 }, { "epoch": 1.92, "learning_rate": 8.078726574139418e-05, "loss": 2.8921, "step": 24502 }, { "epoch": 1.92, "learning_rate": 8.078648161216968e-05, "loss": 2.9022, "step": 24503 }, { "epoch": 1.92, "learning_rate": 8.078569748294519e-05, "loss": 2.8778, "step": 24504 }, { "epoch": 1.92, "learning_rate": 8.078491335372069e-05, "loss": 2.9236, "step": 24505 }, { "epoch": 1.92, "learning_rate": 8.07841292244962e-05, "loss": 2.8469, "step": 24506 }, { "epoch": 1.92, "learning_rate": 8.07833450952717e-05, "loss": 3.0344, "step": 24507 }, { "epoch": 1.92, "learning_rate": 8.07825609660472e-05, "loss": 3.2319, "step": 24508 }, { "epoch": 1.92, "learning_rate": 8.07817768368227e-05, "loss": 2.8565, "step": 24509 }, { "epoch": 1.92, "learning_rate": 8.078099270759821e-05, "loss": 2.9448, "step": 24510 }, { "epoch": 1.92, "learning_rate": 8.078020857837371e-05, "loss": 3.1117, "step": 24511 }, { "epoch": 1.92, "learning_rate": 8.077942444914922e-05, "loss": 2.8505, "step": 24512 }, { "epoch": 1.92, "learning_rate": 8.077864031992472e-05, "loss": 2.9373, "step": 24513 }, { "epoch": 1.92, "learning_rate": 8.077785619070024e-05, "loss": 3.0394, "step": 24514 }, { "epoch": 1.92, "learning_rate": 8.077707206147574e-05, "loss": 2.9367, "step": 24515 }, { "epoch": 1.92, "learning_rate": 8.077628793225124e-05, "loss": 2.7249, "step": 24516 }, { "epoch": 1.92, "learning_rate": 8.077550380302675e-05, "loss": 3.1585, "step": 24517 }, { "epoch": 1.92, "learning_rate": 8.077471967380225e-05, "loss": 2.803, "step": 24518 }, { "epoch": 1.92, "learning_rate": 8.077393554457775e-05, "loss": 3.0622, "step": 24519 }, { "epoch": 1.92, "learning_rate": 8.077315141535326e-05, "loss": 2.7231, "step": 24520 }, { "epoch": 1.92, "learning_rate": 8.077236728612876e-05, "loss": 3.2004, "step": 24521 }, { "epoch": 1.92, "learning_rate": 8.077158315690427e-05, "loss": 2.8455, "step": 24522 }, { "epoch": 1.92, "learning_rate": 8.077079902767977e-05, "loss": 2.9764, "step": 24523 }, { "epoch": 1.92, "learning_rate": 8.077001489845527e-05, "loss": 2.8427, "step": 24524 }, { "epoch": 1.92, "learning_rate": 8.076923076923078e-05, "loss": 2.8835, "step": 24525 }, { "epoch": 1.92, "learning_rate": 8.076844664000628e-05, "loss": 3.0702, "step": 24526 }, { "epoch": 1.92, "learning_rate": 8.076766251078178e-05, "loss": 2.9361, "step": 24527 }, { "epoch": 1.92, "learning_rate": 8.076687838155729e-05, "loss": 2.8097, "step": 24528 }, { "epoch": 1.92, "learning_rate": 8.076609425233279e-05, "loss": 2.9955, "step": 24529 }, { "epoch": 1.92, "learning_rate": 8.07653101231083e-05, "loss": 2.9128, "step": 24530 }, { "epoch": 1.92, "learning_rate": 8.07645259938838e-05, "loss": 2.9468, "step": 24531 }, { "epoch": 1.92, "learning_rate": 8.07637418646593e-05, "loss": 2.8934, "step": 24532 }, { "epoch": 1.92, "learning_rate": 8.07629577354348e-05, "loss": 2.9373, "step": 24533 }, { "epoch": 1.92, "learning_rate": 8.076217360621031e-05, "loss": 2.9389, "step": 24534 }, { "epoch": 1.92, "learning_rate": 8.076138947698581e-05, "loss": 2.6982, "step": 24535 }, { "epoch": 1.92, "learning_rate": 8.076060534776132e-05, "loss": 3.106, "step": 24536 }, { "epoch": 1.92, "learning_rate": 8.075982121853682e-05, "loss": 2.6913, "step": 24537 }, { "epoch": 1.92, "learning_rate": 8.075903708931231e-05, "loss": 2.9845, "step": 24538 }, { "epoch": 1.92, "learning_rate": 8.075825296008781e-05, "loss": 2.8539, "step": 24539 }, { "epoch": 1.92, "learning_rate": 8.075746883086333e-05, "loss": 2.8857, "step": 24540 }, { "epoch": 1.92, "learning_rate": 8.075668470163884e-05, "loss": 2.8388, "step": 24541 }, { "epoch": 1.92, "learning_rate": 8.075590057241434e-05, "loss": 2.8831, "step": 24542 }, { "epoch": 1.92, "learning_rate": 8.075511644318984e-05, "loss": 3.0348, "step": 24543 }, { "epoch": 1.92, "learning_rate": 8.075433231396535e-05, "loss": 3.1461, "step": 24544 }, { "epoch": 1.92, "learning_rate": 8.075354818474085e-05, "loss": 2.9477, "step": 24545 }, { "epoch": 1.92, "learning_rate": 8.075276405551635e-05, "loss": 2.8208, "step": 24546 }, { "epoch": 1.92, "learning_rate": 8.075197992629186e-05, "loss": 2.8346, "step": 24547 }, { "epoch": 1.92, "learning_rate": 8.075119579706736e-05, "loss": 2.9397, "step": 24548 }, { "epoch": 1.92, "learning_rate": 8.075041166784287e-05, "loss": 2.7051, "step": 24549 }, { "epoch": 1.93, "learning_rate": 8.074962753861837e-05, "loss": 2.8788, "step": 24550 }, { "epoch": 1.93, "learning_rate": 8.074884340939387e-05, "loss": 3.0368, "step": 24551 }, { "epoch": 1.93, "learning_rate": 8.074805928016938e-05, "loss": 3.1001, "step": 24552 }, { "epoch": 1.93, "learning_rate": 8.074727515094488e-05, "loss": 2.8753, "step": 24553 }, { "epoch": 1.93, "learning_rate": 8.074649102172038e-05, "loss": 2.7674, "step": 24554 }, { "epoch": 1.93, "learning_rate": 8.074570689249589e-05, "loss": 2.7814, "step": 24555 }, { "epoch": 1.93, "learning_rate": 8.074492276327139e-05, "loss": 2.8288, "step": 24556 }, { "epoch": 1.93, "learning_rate": 8.07441386340469e-05, "loss": 2.8516, "step": 24557 }, { "epoch": 1.93, "learning_rate": 8.07433545048224e-05, "loss": 2.8274, "step": 24558 }, { "epoch": 1.93, "learning_rate": 8.07425703755979e-05, "loss": 3.1885, "step": 24559 }, { "epoch": 1.93, "learning_rate": 8.07417862463734e-05, "loss": 2.7105, "step": 24560 }, { "epoch": 1.93, "learning_rate": 8.074100211714891e-05, "loss": 3.1928, "step": 24561 }, { "epoch": 1.93, "learning_rate": 8.074021798792441e-05, "loss": 2.8666, "step": 24562 }, { "epoch": 1.93, "learning_rate": 8.073943385869992e-05, "loss": 2.9395, "step": 24563 }, { "epoch": 1.93, "learning_rate": 8.073864972947542e-05, "loss": 2.9128, "step": 24564 }, { "epoch": 1.93, "learning_rate": 8.073786560025092e-05, "loss": 3.1499, "step": 24565 }, { "epoch": 1.93, "learning_rate": 8.073708147102643e-05, "loss": 2.8775, "step": 24566 }, { "epoch": 1.93, "learning_rate": 8.073629734180193e-05, "loss": 2.7698, "step": 24567 }, { "epoch": 1.93, "learning_rate": 8.073551321257743e-05, "loss": 2.794, "step": 24568 }, { "epoch": 1.93, "learning_rate": 8.073472908335294e-05, "loss": 3.1086, "step": 24569 }, { "epoch": 1.93, "learning_rate": 8.073394495412844e-05, "loss": 2.8206, "step": 24570 }, { "epoch": 1.93, "learning_rate": 8.073316082490396e-05, "loss": 3.0429, "step": 24571 }, { "epoch": 1.93, "learning_rate": 8.073237669567946e-05, "loss": 2.796, "step": 24572 }, { "epoch": 1.93, "learning_rate": 8.073159256645497e-05, "loss": 2.7178, "step": 24573 }, { "epoch": 1.93, "learning_rate": 8.073080843723046e-05, "loss": 2.8441, "step": 24574 }, { "epoch": 1.93, "learning_rate": 8.073002430800596e-05, "loss": 2.6893, "step": 24575 }, { "epoch": 1.93, "learning_rate": 8.072924017878146e-05, "loss": 3.0928, "step": 24576 }, { "epoch": 1.93, "learning_rate": 8.072845604955697e-05, "loss": 3.1484, "step": 24577 }, { "epoch": 1.93, "learning_rate": 8.072767192033247e-05, "loss": 2.8387, "step": 24578 }, { "epoch": 1.93, "learning_rate": 8.072688779110798e-05, "loss": 2.8876, "step": 24579 }, { "epoch": 1.93, "learning_rate": 8.072610366188348e-05, "loss": 2.8391, "step": 24580 }, { "epoch": 1.93, "learning_rate": 8.072531953265898e-05, "loss": 2.9867, "step": 24581 }, { "epoch": 1.93, "learning_rate": 8.072453540343449e-05, "loss": 3.0825, "step": 24582 }, { "epoch": 1.93, "learning_rate": 8.072375127420999e-05, "loss": 2.8603, "step": 24583 }, { "epoch": 1.93, "learning_rate": 8.07229671449855e-05, "loss": 2.9898, "step": 24584 }, { "epoch": 1.93, "learning_rate": 8.0722183015761e-05, "loss": 2.7451, "step": 24585 }, { "epoch": 1.93, "learning_rate": 8.07213988865365e-05, "loss": 2.947, "step": 24586 }, { "epoch": 1.93, "learning_rate": 8.0720614757312e-05, "loss": 2.955, "step": 24587 }, { "epoch": 1.93, "learning_rate": 8.071983062808751e-05, "loss": 2.956, "step": 24588 }, { "epoch": 1.93, "learning_rate": 8.071904649886301e-05, "loss": 2.7363, "step": 24589 }, { "epoch": 1.93, "learning_rate": 8.071826236963852e-05, "loss": 2.9113, "step": 24590 }, { "epoch": 1.93, "learning_rate": 8.071747824041402e-05, "loss": 2.9468, "step": 24591 }, { "epoch": 1.93, "learning_rate": 8.071669411118952e-05, "loss": 2.7113, "step": 24592 }, { "epoch": 1.93, "learning_rate": 8.071590998196503e-05, "loss": 2.7477, "step": 24593 }, { "epoch": 1.93, "learning_rate": 8.071512585274053e-05, "loss": 3.1, "step": 24594 }, { "epoch": 1.93, "learning_rate": 8.071434172351603e-05, "loss": 2.7435, "step": 24595 }, { "epoch": 1.93, "learning_rate": 8.071355759429154e-05, "loss": 3.1044, "step": 24596 }, { "epoch": 1.93, "learning_rate": 8.071277346506706e-05, "loss": 2.8036, "step": 24597 }, { "epoch": 1.93, "learning_rate": 8.071198933584256e-05, "loss": 2.7038, "step": 24598 }, { "epoch": 1.93, "learning_rate": 8.071120520661806e-05, "loss": 2.8991, "step": 24599 }, { "epoch": 1.93, "learning_rate": 8.071042107739357e-05, "loss": 3.0052, "step": 24600 }, { "epoch": 1.93, "learning_rate": 8.070963694816907e-05, "loss": 3.0221, "step": 24601 }, { "epoch": 1.93, "learning_rate": 8.070885281894457e-05, "loss": 2.6678, "step": 24602 }, { "epoch": 1.93, "learning_rate": 8.070806868972008e-05, "loss": 2.8073, "step": 24603 }, { "epoch": 1.93, "learning_rate": 8.070728456049558e-05, "loss": 2.6994, "step": 24604 }, { "epoch": 1.93, "learning_rate": 8.070650043127108e-05, "loss": 2.7102, "step": 24605 }, { "epoch": 1.93, "learning_rate": 8.070571630204659e-05, "loss": 2.7166, "step": 24606 }, { "epoch": 1.93, "learning_rate": 8.070493217282209e-05, "loss": 2.8112, "step": 24607 }, { "epoch": 1.93, "learning_rate": 8.07041480435976e-05, "loss": 3.0027, "step": 24608 }, { "epoch": 1.93, "learning_rate": 8.07033639143731e-05, "loss": 2.6986, "step": 24609 }, { "epoch": 1.93, "learning_rate": 8.070257978514859e-05, "loss": 2.8651, "step": 24610 }, { "epoch": 1.93, "learning_rate": 8.070179565592409e-05, "loss": 2.7109, "step": 24611 }, { "epoch": 1.93, "learning_rate": 8.07010115266996e-05, "loss": 2.8512, "step": 24612 }, { "epoch": 1.93, "learning_rate": 8.07002273974751e-05, "loss": 2.7362, "step": 24613 }, { "epoch": 1.93, "learning_rate": 8.06994432682506e-05, "loss": 2.9536, "step": 24614 }, { "epoch": 1.93, "learning_rate": 8.069865913902611e-05, "loss": 2.9185, "step": 24615 }, { "epoch": 1.93, "learning_rate": 8.069787500980161e-05, "loss": 2.79, "step": 24616 }, { "epoch": 1.93, "learning_rate": 8.069709088057712e-05, "loss": 2.6906, "step": 24617 }, { "epoch": 1.93, "learning_rate": 8.069630675135262e-05, "loss": 2.6987, "step": 24618 }, { "epoch": 1.93, "learning_rate": 8.069552262212812e-05, "loss": 2.9956, "step": 24619 }, { "epoch": 1.93, "learning_rate": 8.069473849290363e-05, "loss": 2.7691, "step": 24620 }, { "epoch": 1.93, "learning_rate": 8.069395436367913e-05, "loss": 2.7979, "step": 24621 }, { "epoch": 1.93, "learning_rate": 8.069317023445465e-05, "loss": 2.8217, "step": 24622 }, { "epoch": 1.93, "learning_rate": 8.069238610523015e-05, "loss": 2.9931, "step": 24623 }, { "epoch": 1.93, "learning_rate": 8.069160197600565e-05, "loss": 3.0278, "step": 24624 }, { "epoch": 1.93, "learning_rate": 8.069081784678116e-05, "loss": 3.0413, "step": 24625 }, { "epoch": 1.93, "learning_rate": 8.069003371755666e-05, "loss": 2.8059, "step": 24626 }, { "epoch": 1.93, "learning_rate": 8.068924958833217e-05, "loss": 3.1125, "step": 24627 }, { "epoch": 1.93, "learning_rate": 8.068846545910767e-05, "loss": 2.7527, "step": 24628 }, { "epoch": 1.93, "learning_rate": 8.068768132988317e-05, "loss": 2.8947, "step": 24629 }, { "epoch": 1.93, "learning_rate": 8.068689720065868e-05, "loss": 2.7948, "step": 24630 }, { "epoch": 1.93, "learning_rate": 8.068611307143418e-05, "loss": 3.1476, "step": 24631 }, { "epoch": 1.93, "learning_rate": 8.068532894220968e-05, "loss": 2.9549, "step": 24632 }, { "epoch": 1.93, "learning_rate": 8.068454481298519e-05, "loss": 2.7441, "step": 24633 }, { "epoch": 1.93, "learning_rate": 8.068376068376069e-05, "loss": 2.7378, "step": 24634 }, { "epoch": 1.93, "learning_rate": 8.06829765545362e-05, "loss": 2.8696, "step": 24635 }, { "epoch": 1.93, "learning_rate": 8.06821924253117e-05, "loss": 3.0239, "step": 24636 }, { "epoch": 1.93, "learning_rate": 8.06814082960872e-05, "loss": 3.0716, "step": 24637 }, { "epoch": 1.93, "learning_rate": 8.06806241668627e-05, "loss": 2.5979, "step": 24638 }, { "epoch": 1.93, "learning_rate": 8.067984003763821e-05, "loss": 2.9087, "step": 24639 }, { "epoch": 1.93, "learning_rate": 8.067905590841371e-05, "loss": 2.8152, "step": 24640 }, { "epoch": 1.93, "learning_rate": 8.067827177918922e-05, "loss": 2.9981, "step": 24641 }, { "epoch": 1.93, "learning_rate": 8.067748764996472e-05, "loss": 2.708, "step": 24642 }, { "epoch": 1.93, "learning_rate": 8.067670352074022e-05, "loss": 2.7429, "step": 24643 }, { "epoch": 1.93, "learning_rate": 8.067591939151573e-05, "loss": 3.0621, "step": 24644 }, { "epoch": 1.93, "learning_rate": 8.067513526229123e-05, "loss": 3.0303, "step": 24645 }, { "epoch": 1.93, "learning_rate": 8.067435113306672e-05, "loss": 2.8534, "step": 24646 }, { "epoch": 1.93, "learning_rate": 8.067356700384223e-05, "loss": 3.0921, "step": 24647 }, { "epoch": 1.93, "learning_rate": 8.067278287461774e-05, "loss": 2.8073, "step": 24648 }, { "epoch": 1.93, "learning_rate": 8.067199874539325e-05, "loss": 2.7826, "step": 24649 }, { "epoch": 1.93, "learning_rate": 8.067121461616875e-05, "loss": 3.1146, "step": 24650 }, { "epoch": 1.93, "learning_rate": 8.067043048694425e-05, "loss": 2.9588, "step": 24651 }, { "epoch": 1.93, "learning_rate": 8.066964635771976e-05, "loss": 2.9374, "step": 24652 }, { "epoch": 1.93, "learning_rate": 8.066886222849526e-05, "loss": 3.0783, "step": 24653 }, { "epoch": 1.93, "learning_rate": 8.066807809927076e-05, "loss": 2.8106, "step": 24654 }, { "epoch": 1.93, "learning_rate": 8.066729397004627e-05, "loss": 3.0854, "step": 24655 }, { "epoch": 1.93, "learning_rate": 8.066650984082177e-05, "loss": 2.9831, "step": 24656 }, { "epoch": 1.93, "learning_rate": 8.066572571159728e-05, "loss": 2.7142, "step": 24657 }, { "epoch": 1.93, "learning_rate": 8.066494158237278e-05, "loss": 3.2102, "step": 24658 }, { "epoch": 1.93, "learning_rate": 8.066415745314828e-05, "loss": 3.1408, "step": 24659 }, { "epoch": 1.93, "learning_rate": 8.066337332392379e-05, "loss": 3.1542, "step": 24660 }, { "epoch": 1.93, "learning_rate": 8.066258919469929e-05, "loss": 2.8985, "step": 24661 }, { "epoch": 1.93, "learning_rate": 8.06618050654748e-05, "loss": 2.887, "step": 24662 }, { "epoch": 1.93, "learning_rate": 8.06610209362503e-05, "loss": 2.5925, "step": 24663 }, { "epoch": 1.93, "learning_rate": 8.06602368070258e-05, "loss": 3.0421, "step": 24664 }, { "epoch": 1.93, "learning_rate": 8.06594526778013e-05, "loss": 2.6831, "step": 24665 }, { "epoch": 1.93, "learning_rate": 8.065866854857681e-05, "loss": 2.8381, "step": 24666 }, { "epoch": 1.93, "learning_rate": 8.065788441935231e-05, "loss": 2.7382, "step": 24667 }, { "epoch": 1.93, "learning_rate": 8.065710029012782e-05, "loss": 2.9488, "step": 24668 }, { "epoch": 1.93, "learning_rate": 8.065631616090332e-05, "loss": 2.9303, "step": 24669 }, { "epoch": 1.93, "learning_rate": 8.065553203167882e-05, "loss": 3.3704, "step": 24670 }, { "epoch": 1.93, "learning_rate": 8.065474790245433e-05, "loss": 2.9563, "step": 24671 }, { "epoch": 1.93, "learning_rate": 8.065396377322983e-05, "loss": 2.7411, "step": 24672 }, { "epoch": 1.93, "learning_rate": 8.065317964400533e-05, "loss": 2.8943, "step": 24673 }, { "epoch": 1.93, "learning_rate": 8.065239551478084e-05, "loss": 2.9509, "step": 24674 }, { "epoch": 1.93, "learning_rate": 8.065161138555634e-05, "loss": 2.7233, "step": 24675 }, { "epoch": 1.93, "learning_rate": 8.065082725633185e-05, "loss": 2.6649, "step": 24676 }, { "epoch": 1.93, "learning_rate": 8.065004312710735e-05, "loss": 2.8787, "step": 24677 }, { "epoch": 1.94, "learning_rate": 8.064925899788285e-05, "loss": 2.9248, "step": 24678 }, { "epoch": 1.94, "learning_rate": 8.064847486865837e-05, "loss": 2.9405, "step": 24679 }, { "epoch": 1.94, "learning_rate": 8.064769073943387e-05, "loss": 3.0916, "step": 24680 }, { "epoch": 1.94, "learning_rate": 8.064690661020938e-05, "loss": 2.9176, "step": 24681 }, { "epoch": 1.94, "learning_rate": 8.064612248098487e-05, "loss": 2.9904, "step": 24682 }, { "epoch": 1.94, "learning_rate": 8.064533835176037e-05, "loss": 2.9124, "step": 24683 }, { "epoch": 1.94, "learning_rate": 8.064455422253588e-05, "loss": 2.8437, "step": 24684 }, { "epoch": 1.94, "learning_rate": 8.064377009331138e-05, "loss": 3.0502, "step": 24685 }, { "epoch": 1.94, "learning_rate": 8.064298596408688e-05, "loss": 2.9369, "step": 24686 }, { "epoch": 1.94, "learning_rate": 8.064220183486239e-05, "loss": 2.8102, "step": 24687 }, { "epoch": 1.94, "learning_rate": 8.064141770563789e-05, "loss": 2.9854, "step": 24688 }, { "epoch": 1.94, "learning_rate": 8.06406335764134e-05, "loss": 2.9155, "step": 24689 }, { "epoch": 1.94, "learning_rate": 8.06398494471889e-05, "loss": 3.0672, "step": 24690 }, { "epoch": 1.94, "learning_rate": 8.06390653179644e-05, "loss": 2.778, "step": 24691 }, { "epoch": 1.94, "learning_rate": 8.06382811887399e-05, "loss": 2.7725, "step": 24692 }, { "epoch": 1.94, "learning_rate": 8.063749705951541e-05, "loss": 3.0435, "step": 24693 }, { "epoch": 1.94, "learning_rate": 8.063671293029091e-05, "loss": 2.977, "step": 24694 }, { "epoch": 1.94, "learning_rate": 8.063592880106642e-05, "loss": 2.9882, "step": 24695 }, { "epoch": 1.94, "learning_rate": 8.063514467184192e-05, "loss": 2.6923, "step": 24696 }, { "epoch": 1.94, "learning_rate": 8.063436054261742e-05, "loss": 2.9067, "step": 24697 }, { "epoch": 1.94, "learning_rate": 8.063357641339293e-05, "loss": 2.9783, "step": 24698 }, { "epoch": 1.94, "learning_rate": 8.063279228416843e-05, "loss": 2.709, "step": 24699 }, { "epoch": 1.94, "learning_rate": 8.063200815494393e-05, "loss": 3.078, "step": 24700 }, { "epoch": 1.94, "learning_rate": 8.063122402571944e-05, "loss": 2.7252, "step": 24701 }, { "epoch": 1.94, "learning_rate": 8.063043989649494e-05, "loss": 2.7769, "step": 24702 }, { "epoch": 1.94, "learning_rate": 8.062965576727044e-05, "loss": 2.8745, "step": 24703 }, { "epoch": 1.94, "learning_rate": 8.062887163804595e-05, "loss": 3.0356, "step": 24704 }, { "epoch": 1.94, "learning_rate": 8.062808750882147e-05, "loss": 2.9122, "step": 24705 }, { "epoch": 1.94, "learning_rate": 8.062730337959697e-05, "loss": 2.8362, "step": 24706 }, { "epoch": 1.94, "learning_rate": 8.062651925037247e-05, "loss": 2.9922, "step": 24707 }, { "epoch": 1.94, "learning_rate": 8.062573512114798e-05, "loss": 2.8837, "step": 24708 }, { "epoch": 1.94, "learning_rate": 8.062495099192348e-05, "loss": 2.897, "step": 24709 }, { "epoch": 1.94, "learning_rate": 8.062416686269898e-05, "loss": 2.8914, "step": 24710 }, { "epoch": 1.94, "learning_rate": 8.062338273347449e-05, "loss": 2.7499, "step": 24711 }, { "epoch": 1.94, "learning_rate": 8.062259860424999e-05, "loss": 2.7945, "step": 24712 }, { "epoch": 1.94, "learning_rate": 8.06218144750255e-05, "loss": 2.9364, "step": 24713 }, { "epoch": 1.94, "learning_rate": 8.0621030345801e-05, "loss": 2.736, "step": 24714 }, { "epoch": 1.94, "learning_rate": 8.06202462165765e-05, "loss": 2.8067, "step": 24715 }, { "epoch": 1.94, "learning_rate": 8.0619462087352e-05, "loss": 2.8156, "step": 24716 }, { "epoch": 1.94, "learning_rate": 8.061867795812751e-05, "loss": 2.8443, "step": 24717 }, { "epoch": 1.94, "learning_rate": 8.0617893828903e-05, "loss": 3.049, "step": 24718 }, { "epoch": 1.94, "learning_rate": 8.06171096996785e-05, "loss": 3.2489, "step": 24719 }, { "epoch": 1.94, "learning_rate": 8.061632557045401e-05, "loss": 2.8595, "step": 24720 }, { "epoch": 1.94, "learning_rate": 8.061554144122951e-05, "loss": 2.9743, "step": 24721 }, { "epoch": 1.94, "learning_rate": 8.061475731200501e-05, "loss": 2.9181, "step": 24722 }, { "epoch": 1.94, "learning_rate": 8.061397318278052e-05, "loss": 3.0945, "step": 24723 }, { "epoch": 1.94, "learning_rate": 8.061318905355602e-05, "loss": 2.9197, "step": 24724 }, { "epoch": 1.94, "learning_rate": 8.061240492433153e-05, "loss": 2.7964, "step": 24725 }, { "epoch": 1.94, "learning_rate": 8.061162079510703e-05, "loss": 2.864, "step": 24726 }, { "epoch": 1.94, "learning_rate": 8.061083666588253e-05, "loss": 3.0255, "step": 24727 }, { "epoch": 1.94, "learning_rate": 8.061005253665804e-05, "loss": 2.7703, "step": 24728 }, { "epoch": 1.94, "learning_rate": 8.060926840743354e-05, "loss": 2.9339, "step": 24729 }, { "epoch": 1.94, "learning_rate": 8.060848427820906e-05, "loss": 2.8535, "step": 24730 }, { "epoch": 1.94, "learning_rate": 8.060770014898456e-05, "loss": 2.87, "step": 24731 }, { "epoch": 1.94, "learning_rate": 8.060691601976007e-05, "loss": 2.9504, "step": 24732 }, { "epoch": 1.94, "learning_rate": 8.060613189053557e-05, "loss": 2.9776, "step": 24733 }, { "epoch": 1.94, "learning_rate": 8.060534776131107e-05, "loss": 2.674, "step": 24734 }, { "epoch": 1.94, "learning_rate": 8.060456363208658e-05, "loss": 2.8485, "step": 24735 }, { "epoch": 1.94, "learning_rate": 8.060377950286208e-05, "loss": 2.9001, "step": 24736 }, { "epoch": 1.94, "learning_rate": 8.060299537363758e-05, "loss": 3.0316, "step": 24737 }, { "epoch": 1.94, "learning_rate": 8.060221124441309e-05, "loss": 2.6443, "step": 24738 }, { "epoch": 1.94, "learning_rate": 8.060142711518859e-05, "loss": 2.7948, "step": 24739 }, { "epoch": 1.94, "learning_rate": 8.06006429859641e-05, "loss": 3.0073, "step": 24740 }, { "epoch": 1.94, "learning_rate": 8.05998588567396e-05, "loss": 2.828, "step": 24741 }, { "epoch": 1.94, "learning_rate": 8.05990747275151e-05, "loss": 3.0165, "step": 24742 }, { "epoch": 1.94, "learning_rate": 8.05982905982906e-05, "loss": 3.0763, "step": 24743 }, { "epoch": 1.94, "learning_rate": 8.059750646906611e-05, "loss": 2.9561, "step": 24744 }, { "epoch": 1.94, "learning_rate": 8.059672233984161e-05, "loss": 2.8873, "step": 24745 }, { "epoch": 1.94, "learning_rate": 8.059593821061712e-05, "loss": 3.0713, "step": 24746 }, { "epoch": 1.94, "learning_rate": 8.059515408139262e-05, "loss": 3.1091, "step": 24747 }, { "epoch": 1.94, "learning_rate": 8.059436995216812e-05, "loss": 2.761, "step": 24748 }, { "epoch": 1.94, "learning_rate": 8.059358582294363e-05, "loss": 2.7823, "step": 24749 }, { "epoch": 1.94, "learning_rate": 8.059280169371913e-05, "loss": 2.9267, "step": 24750 }, { "epoch": 1.94, "learning_rate": 8.059201756449464e-05, "loss": 2.8204, "step": 24751 }, { "epoch": 1.94, "learning_rate": 8.059123343527014e-05, "loss": 2.9919, "step": 24752 }, { "epoch": 1.94, "learning_rate": 8.059044930604564e-05, "loss": 2.8494, "step": 24753 }, { "epoch": 1.94, "learning_rate": 8.058966517682113e-05, "loss": 2.8966, "step": 24754 }, { "epoch": 1.94, "learning_rate": 8.058888104759664e-05, "loss": 2.8529, "step": 24755 }, { "epoch": 1.94, "learning_rate": 8.058809691837215e-05, "loss": 2.9478, "step": 24756 }, { "epoch": 1.94, "learning_rate": 8.058731278914766e-05, "loss": 3.0798, "step": 24757 }, { "epoch": 1.94, "learning_rate": 8.058652865992316e-05, "loss": 2.8616, "step": 24758 }, { "epoch": 1.94, "learning_rate": 8.058574453069866e-05, "loss": 3.0811, "step": 24759 }, { "epoch": 1.94, "learning_rate": 8.058496040147417e-05, "loss": 2.7646, "step": 24760 }, { "epoch": 1.94, "learning_rate": 8.058417627224967e-05, "loss": 3.0705, "step": 24761 }, { "epoch": 1.94, "learning_rate": 8.058339214302518e-05, "loss": 2.8763, "step": 24762 }, { "epoch": 1.94, "learning_rate": 8.058260801380068e-05, "loss": 2.8538, "step": 24763 }, { "epoch": 1.94, "learning_rate": 8.058182388457618e-05, "loss": 2.9898, "step": 24764 }, { "epoch": 1.94, "learning_rate": 8.058103975535169e-05, "loss": 3.1701, "step": 24765 }, { "epoch": 1.94, "learning_rate": 8.058025562612719e-05, "loss": 2.9786, "step": 24766 }, { "epoch": 1.94, "learning_rate": 8.05794714969027e-05, "loss": 2.9278, "step": 24767 }, { "epoch": 1.94, "learning_rate": 8.05786873676782e-05, "loss": 2.8084, "step": 24768 }, { "epoch": 1.94, "learning_rate": 8.05779032384537e-05, "loss": 3.0988, "step": 24769 }, { "epoch": 1.94, "learning_rate": 8.05771191092292e-05, "loss": 2.9652, "step": 24770 }, { "epoch": 1.94, "learning_rate": 8.057633498000471e-05, "loss": 2.9613, "step": 24771 }, { "epoch": 1.94, "learning_rate": 8.057555085078021e-05, "loss": 2.8628, "step": 24772 }, { "epoch": 1.94, "learning_rate": 8.057476672155572e-05, "loss": 3.0865, "step": 24773 }, { "epoch": 1.94, "learning_rate": 8.057398259233122e-05, "loss": 2.856, "step": 24774 }, { "epoch": 1.94, "learning_rate": 8.057319846310672e-05, "loss": 2.9273, "step": 24775 }, { "epoch": 1.94, "learning_rate": 8.057241433388223e-05, "loss": 2.9875, "step": 24776 }, { "epoch": 1.94, "learning_rate": 8.057163020465773e-05, "loss": 3.0534, "step": 24777 }, { "epoch": 1.94, "learning_rate": 8.057084607543323e-05, "loss": 2.7888, "step": 24778 }, { "epoch": 1.94, "learning_rate": 8.057006194620874e-05, "loss": 2.8627, "step": 24779 }, { "epoch": 1.94, "learning_rate": 8.056927781698424e-05, "loss": 3.1301, "step": 24780 }, { "epoch": 1.94, "learning_rate": 8.056849368775975e-05, "loss": 2.8252, "step": 24781 }, { "epoch": 1.94, "learning_rate": 8.056770955853525e-05, "loss": 2.6959, "step": 24782 }, { "epoch": 1.94, "learning_rate": 8.056692542931075e-05, "loss": 3.1066, "step": 24783 }, { "epoch": 1.94, "learning_rate": 8.056614130008626e-05, "loss": 2.9157, "step": 24784 }, { "epoch": 1.94, "learning_rate": 8.056535717086176e-05, "loss": 2.677, "step": 24785 }, { "epoch": 1.94, "learning_rate": 8.056457304163726e-05, "loss": 2.8363, "step": 24786 }, { "epoch": 1.94, "learning_rate": 8.056378891241278e-05, "loss": 2.7539, "step": 24787 }, { "epoch": 1.94, "learning_rate": 8.056300478318828e-05, "loss": 2.8666, "step": 24788 }, { "epoch": 1.94, "learning_rate": 8.056222065396379e-05, "loss": 2.762, "step": 24789 }, { "epoch": 1.94, "learning_rate": 8.056143652473928e-05, "loss": 2.8982, "step": 24790 }, { "epoch": 1.94, "learning_rate": 8.056065239551478e-05, "loss": 3.0417, "step": 24791 }, { "epoch": 1.94, "learning_rate": 8.055986826629029e-05, "loss": 2.5553, "step": 24792 }, { "epoch": 1.94, "learning_rate": 8.055908413706579e-05, "loss": 3.07, "step": 24793 }, { "epoch": 1.94, "learning_rate": 8.055830000784129e-05, "loss": 2.8227, "step": 24794 }, { "epoch": 1.94, "learning_rate": 8.05575158786168e-05, "loss": 3.0362, "step": 24795 }, { "epoch": 1.94, "learning_rate": 8.05567317493923e-05, "loss": 2.9272, "step": 24796 }, { "epoch": 1.94, "learning_rate": 8.05559476201678e-05, "loss": 2.8065, "step": 24797 }, { "epoch": 1.94, "learning_rate": 8.055516349094331e-05, "loss": 2.7248, "step": 24798 }, { "epoch": 1.94, "learning_rate": 8.055437936171881e-05, "loss": 2.9865, "step": 24799 }, { "epoch": 1.94, "learning_rate": 8.055359523249432e-05, "loss": 2.7529, "step": 24800 }, { "epoch": 1.94, "learning_rate": 8.055281110326982e-05, "loss": 2.9786, "step": 24801 }, { "epoch": 1.94, "learning_rate": 8.055202697404532e-05, "loss": 2.9116, "step": 24802 }, { "epoch": 1.94, "learning_rate": 8.055124284482083e-05, "loss": 2.7827, "step": 24803 }, { "epoch": 1.94, "learning_rate": 8.055045871559633e-05, "loss": 2.5596, "step": 24804 }, { "epoch": 1.95, "learning_rate": 8.054967458637183e-05, "loss": 2.8908, "step": 24805 }, { "epoch": 1.95, "learning_rate": 8.054889045714734e-05, "loss": 3.0551, "step": 24806 }, { "epoch": 1.95, "learning_rate": 8.054810632792284e-05, "loss": 2.7305, "step": 24807 }, { "epoch": 1.95, "learning_rate": 8.054732219869834e-05, "loss": 3.1325, "step": 24808 }, { "epoch": 1.95, "learning_rate": 8.054653806947385e-05, "loss": 2.8037, "step": 24809 }, { "epoch": 1.95, "learning_rate": 8.054575394024935e-05, "loss": 2.76, "step": 24810 }, { "epoch": 1.95, "learning_rate": 8.054496981102486e-05, "loss": 2.8439, "step": 24811 }, { "epoch": 1.95, "learning_rate": 8.054418568180036e-05, "loss": 2.8409, "step": 24812 }, { "epoch": 1.95, "learning_rate": 8.054340155257588e-05, "loss": 2.6978, "step": 24813 }, { "epoch": 1.95, "learning_rate": 8.054261742335138e-05, "loss": 2.9338, "step": 24814 }, { "epoch": 1.95, "learning_rate": 8.054183329412688e-05, "loss": 2.7611, "step": 24815 }, { "epoch": 1.95, "learning_rate": 8.054104916490239e-05, "loss": 3.1973, "step": 24816 }, { "epoch": 1.95, "learning_rate": 8.054026503567789e-05, "loss": 2.8947, "step": 24817 }, { "epoch": 1.95, "learning_rate": 8.05394809064534e-05, "loss": 2.6977, "step": 24818 }, { "epoch": 1.95, "learning_rate": 8.05386967772289e-05, "loss": 3.0429, "step": 24819 }, { "epoch": 1.95, "learning_rate": 8.05379126480044e-05, "loss": 2.8957, "step": 24820 }, { "epoch": 1.95, "learning_rate": 8.05371285187799e-05, "loss": 2.7855, "step": 24821 }, { "epoch": 1.95, "learning_rate": 8.053634438955541e-05, "loss": 2.9299, "step": 24822 }, { "epoch": 1.95, "learning_rate": 8.053556026033091e-05, "loss": 2.8588, "step": 24823 }, { "epoch": 1.95, "learning_rate": 8.053477613110642e-05, "loss": 2.9037, "step": 24824 }, { "epoch": 1.95, "learning_rate": 8.053399200188192e-05, "loss": 2.7226, "step": 24825 }, { "epoch": 1.95, "learning_rate": 8.053320787265741e-05, "loss": 2.9166, "step": 24826 }, { "epoch": 1.95, "learning_rate": 8.053242374343291e-05, "loss": 3.0904, "step": 24827 }, { "epoch": 1.95, "learning_rate": 8.053163961420842e-05, "loss": 2.7532, "step": 24828 }, { "epoch": 1.95, "learning_rate": 8.053085548498392e-05, "loss": 2.7845, "step": 24829 }, { "epoch": 1.95, "learning_rate": 8.053007135575943e-05, "loss": 2.9757, "step": 24830 }, { "epoch": 1.95, "learning_rate": 8.052928722653493e-05, "loss": 2.942, "step": 24831 }, { "epoch": 1.95, "learning_rate": 8.052850309731043e-05, "loss": 2.7984, "step": 24832 }, { "epoch": 1.95, "learning_rate": 8.052771896808594e-05, "loss": 2.8299, "step": 24833 }, { "epoch": 1.95, "learning_rate": 8.052693483886144e-05, "loss": 2.8679, "step": 24834 }, { "epoch": 1.95, "learning_rate": 8.052615070963694e-05, "loss": 2.674, "step": 24835 }, { "epoch": 1.95, "learning_rate": 8.052536658041245e-05, "loss": 2.968, "step": 24836 }, { "epoch": 1.95, "learning_rate": 8.052458245118795e-05, "loss": 2.7766, "step": 24837 }, { "epoch": 1.95, "learning_rate": 8.052379832196345e-05, "loss": 2.9549, "step": 24838 }, { "epoch": 1.95, "learning_rate": 8.052301419273897e-05, "loss": 2.8502, "step": 24839 }, { "epoch": 1.95, "learning_rate": 8.052223006351448e-05, "loss": 2.938, "step": 24840 }, { "epoch": 1.95, "learning_rate": 8.052144593428998e-05, "loss": 2.6474, "step": 24841 }, { "epoch": 1.95, "learning_rate": 8.052066180506548e-05, "loss": 2.8861, "step": 24842 }, { "epoch": 1.95, "learning_rate": 8.051987767584099e-05, "loss": 2.8612, "step": 24843 }, { "epoch": 1.95, "learning_rate": 8.051909354661649e-05, "loss": 2.8973, "step": 24844 }, { "epoch": 1.95, "learning_rate": 8.0518309417392e-05, "loss": 2.8655, "step": 24845 }, { "epoch": 1.95, "learning_rate": 8.05175252881675e-05, "loss": 3.1576, "step": 24846 }, { "epoch": 1.95, "learning_rate": 8.0516741158943e-05, "loss": 3.1145, "step": 24847 }, { "epoch": 1.95, "learning_rate": 8.05159570297185e-05, "loss": 2.8359, "step": 24848 }, { "epoch": 1.95, "learning_rate": 8.051517290049401e-05, "loss": 3.013, "step": 24849 }, { "epoch": 1.95, "learning_rate": 8.051438877126951e-05, "loss": 2.7354, "step": 24850 }, { "epoch": 1.95, "learning_rate": 8.051360464204502e-05, "loss": 2.9722, "step": 24851 }, { "epoch": 1.95, "learning_rate": 8.051282051282052e-05, "loss": 2.9822, "step": 24852 }, { "epoch": 1.95, "learning_rate": 8.051203638359602e-05, "loss": 2.7242, "step": 24853 }, { "epoch": 1.95, "learning_rate": 8.051125225437153e-05, "loss": 2.9688, "step": 24854 }, { "epoch": 1.95, "learning_rate": 8.051046812514703e-05, "loss": 2.9355, "step": 24855 }, { "epoch": 1.95, "learning_rate": 8.050968399592253e-05, "loss": 3.0886, "step": 24856 }, { "epoch": 1.95, "learning_rate": 8.050889986669804e-05, "loss": 2.8934, "step": 24857 }, { "epoch": 1.95, "learning_rate": 8.050811573747354e-05, "loss": 2.767, "step": 24858 }, { "epoch": 1.95, "learning_rate": 8.050733160824905e-05, "loss": 3.0013, "step": 24859 }, { "epoch": 1.95, "learning_rate": 8.050654747902455e-05, "loss": 2.6851, "step": 24860 }, { "epoch": 1.95, "learning_rate": 8.050576334980005e-05, "loss": 2.8609, "step": 24861 }, { "epoch": 1.95, "learning_rate": 8.050497922057554e-05, "loss": 2.893, "step": 24862 }, { "epoch": 1.95, "learning_rate": 8.050419509135105e-05, "loss": 2.8279, "step": 24863 }, { "epoch": 1.95, "learning_rate": 8.050341096212656e-05, "loss": 2.8756, "step": 24864 }, { "epoch": 1.95, "learning_rate": 8.050262683290207e-05, "loss": 2.8671, "step": 24865 }, { "epoch": 1.95, "learning_rate": 8.050184270367757e-05, "loss": 2.729, "step": 24866 }, { "epoch": 1.95, "learning_rate": 8.050105857445308e-05, "loss": 2.9785, "step": 24867 }, { "epoch": 1.95, "learning_rate": 8.050027444522858e-05, "loss": 3.2062, "step": 24868 }, { "epoch": 1.95, "learning_rate": 8.049949031600408e-05, "loss": 2.7826, "step": 24869 }, { "epoch": 1.95, "learning_rate": 8.049870618677959e-05, "loss": 3.0829, "step": 24870 }, { "epoch": 1.95, "learning_rate": 8.049792205755509e-05, "loss": 2.7688, "step": 24871 }, { "epoch": 1.95, "learning_rate": 8.04971379283306e-05, "loss": 3.0057, "step": 24872 }, { "epoch": 1.95, "learning_rate": 8.04963537991061e-05, "loss": 2.8763, "step": 24873 }, { "epoch": 1.95, "learning_rate": 8.04955696698816e-05, "loss": 2.7922, "step": 24874 }, { "epoch": 1.95, "learning_rate": 8.04947855406571e-05, "loss": 3.0332, "step": 24875 }, { "epoch": 1.95, "learning_rate": 8.049400141143261e-05, "loss": 3.1664, "step": 24876 }, { "epoch": 1.95, "learning_rate": 8.049321728220811e-05, "loss": 2.8444, "step": 24877 }, { "epoch": 1.95, "learning_rate": 8.049243315298362e-05, "loss": 2.7806, "step": 24878 }, { "epoch": 1.95, "learning_rate": 8.049164902375912e-05, "loss": 2.939, "step": 24879 }, { "epoch": 1.95, "learning_rate": 8.049086489453462e-05, "loss": 2.5792, "step": 24880 }, { "epoch": 1.95, "learning_rate": 8.049008076531013e-05, "loss": 2.7013, "step": 24881 }, { "epoch": 1.95, "learning_rate": 8.048929663608563e-05, "loss": 2.7974, "step": 24882 }, { "epoch": 1.95, "learning_rate": 8.048851250686113e-05, "loss": 2.7161, "step": 24883 }, { "epoch": 1.95, "learning_rate": 8.048772837763664e-05, "loss": 3.0383, "step": 24884 }, { "epoch": 1.95, "learning_rate": 8.048694424841214e-05, "loss": 2.8049, "step": 24885 }, { "epoch": 1.95, "learning_rate": 8.048616011918765e-05, "loss": 2.904, "step": 24886 }, { "epoch": 1.95, "learning_rate": 8.048537598996315e-05, "loss": 3.0534, "step": 24887 }, { "epoch": 1.95, "learning_rate": 8.048459186073865e-05, "loss": 2.8976, "step": 24888 }, { "epoch": 1.95, "learning_rate": 8.048380773151416e-05, "loss": 3.0558, "step": 24889 }, { "epoch": 1.95, "learning_rate": 8.048302360228966e-05, "loss": 2.9717, "step": 24890 }, { "epoch": 1.95, "learning_rate": 8.048223947306516e-05, "loss": 2.7161, "step": 24891 }, { "epoch": 1.95, "learning_rate": 8.048145534384067e-05, "loss": 3.0371, "step": 24892 }, { "epoch": 1.95, "learning_rate": 8.048067121461617e-05, "loss": 2.8641, "step": 24893 }, { "epoch": 1.95, "learning_rate": 8.047988708539167e-05, "loss": 2.7484, "step": 24894 }, { "epoch": 1.95, "learning_rate": 8.047910295616719e-05, "loss": 2.8242, "step": 24895 }, { "epoch": 1.95, "learning_rate": 8.04783188269427e-05, "loss": 2.833, "step": 24896 }, { "epoch": 1.95, "learning_rate": 8.04775346977182e-05, "loss": 2.8727, "step": 24897 }, { "epoch": 1.95, "learning_rate": 8.047675056849369e-05, "loss": 2.8626, "step": 24898 }, { "epoch": 1.95, "learning_rate": 8.047596643926919e-05, "loss": 2.795, "step": 24899 }, { "epoch": 1.95, "learning_rate": 8.04751823100447e-05, "loss": 2.9092, "step": 24900 }, { "epoch": 1.95, "learning_rate": 8.04743981808202e-05, "loss": 2.7391, "step": 24901 }, { "epoch": 1.95, "learning_rate": 8.04736140515957e-05, "loss": 2.8259, "step": 24902 }, { "epoch": 1.95, "learning_rate": 8.047282992237121e-05, "loss": 2.7312, "step": 24903 }, { "epoch": 1.95, "learning_rate": 8.047204579314671e-05, "loss": 3.0685, "step": 24904 }, { "epoch": 1.95, "learning_rate": 8.047126166392221e-05, "loss": 2.6862, "step": 24905 }, { "epoch": 1.95, "learning_rate": 8.047047753469772e-05, "loss": 2.7728, "step": 24906 }, { "epoch": 1.95, "learning_rate": 8.046969340547322e-05, "loss": 2.6105, "step": 24907 }, { "epoch": 1.95, "learning_rate": 8.046890927624873e-05, "loss": 3.2985, "step": 24908 }, { "epoch": 1.95, "learning_rate": 8.046812514702423e-05, "loss": 2.8646, "step": 24909 }, { "epoch": 1.95, "learning_rate": 8.046734101779973e-05, "loss": 2.9903, "step": 24910 }, { "epoch": 1.95, "learning_rate": 8.046655688857524e-05, "loss": 2.9093, "step": 24911 }, { "epoch": 1.95, "learning_rate": 8.046577275935074e-05, "loss": 3.0261, "step": 24912 }, { "epoch": 1.95, "learning_rate": 8.046498863012624e-05, "loss": 3.3475, "step": 24913 }, { "epoch": 1.95, "learning_rate": 8.046420450090175e-05, "loss": 2.8107, "step": 24914 }, { "epoch": 1.95, "learning_rate": 8.046342037167725e-05, "loss": 2.7871, "step": 24915 }, { "epoch": 1.95, "learning_rate": 8.046263624245276e-05, "loss": 3.0041, "step": 24916 }, { "epoch": 1.95, "learning_rate": 8.046185211322826e-05, "loss": 2.6819, "step": 24917 }, { "epoch": 1.95, "learning_rate": 8.046106798400376e-05, "loss": 3.1705, "step": 24918 }, { "epoch": 1.95, "learning_rate": 8.046028385477927e-05, "loss": 3.0577, "step": 24919 }, { "epoch": 1.95, "learning_rate": 8.045949972555477e-05, "loss": 2.8663, "step": 24920 }, { "epoch": 1.95, "learning_rate": 8.045871559633029e-05, "loss": 3.1101, "step": 24921 }, { "epoch": 1.95, "learning_rate": 8.045793146710579e-05, "loss": 2.8985, "step": 24922 }, { "epoch": 1.95, "learning_rate": 8.04571473378813e-05, "loss": 2.859, "step": 24923 }, { "epoch": 1.95, "learning_rate": 8.04563632086568e-05, "loss": 2.8557, "step": 24924 }, { "epoch": 1.95, "learning_rate": 8.04555790794323e-05, "loss": 2.9855, "step": 24925 }, { "epoch": 1.95, "learning_rate": 8.04547949502078e-05, "loss": 2.8917, "step": 24926 }, { "epoch": 1.95, "learning_rate": 8.045401082098331e-05, "loss": 2.8852, "step": 24927 }, { "epoch": 1.95, "learning_rate": 8.045322669175881e-05, "loss": 2.8639, "step": 24928 }, { "epoch": 1.95, "learning_rate": 8.045244256253432e-05, "loss": 2.8655, "step": 24929 }, { "epoch": 1.95, "learning_rate": 8.045165843330982e-05, "loss": 2.931, "step": 24930 }, { "epoch": 1.95, "learning_rate": 8.045087430408532e-05, "loss": 2.7979, "step": 24931 }, { "epoch": 1.95, "learning_rate": 8.045009017486083e-05, "loss": 2.9069, "step": 24932 }, { "epoch": 1.96, "learning_rate": 8.044930604563633e-05, "loss": 2.9172, "step": 24933 }, { "epoch": 1.96, "learning_rate": 8.044852191641182e-05, "loss": 3.0701, "step": 24934 }, { "epoch": 1.96, "learning_rate": 8.044773778718733e-05, "loss": 2.7383, "step": 24935 }, { "epoch": 1.96, "learning_rate": 8.044695365796283e-05, "loss": 2.8444, "step": 24936 }, { "epoch": 1.96, "learning_rate": 8.044616952873833e-05, "loss": 2.9875, "step": 24937 }, { "epoch": 1.96, "learning_rate": 8.044538539951384e-05, "loss": 2.7497, "step": 24938 }, { "epoch": 1.96, "learning_rate": 8.044460127028934e-05, "loss": 2.8733, "step": 24939 }, { "epoch": 1.96, "learning_rate": 8.044381714106484e-05, "loss": 2.9959, "step": 24940 }, { "epoch": 1.96, "learning_rate": 8.044303301184035e-05, "loss": 2.9053, "step": 24941 }, { "epoch": 1.96, "learning_rate": 8.044224888261585e-05, "loss": 2.7338, "step": 24942 }, { "epoch": 1.96, "learning_rate": 8.044146475339135e-05, "loss": 2.9864, "step": 24943 }, { "epoch": 1.96, "learning_rate": 8.044068062416686e-05, "loss": 3.1011, "step": 24944 }, { "epoch": 1.96, "learning_rate": 8.043989649494236e-05, "loss": 3.0951, "step": 24945 }, { "epoch": 1.96, "learning_rate": 8.043911236571787e-05, "loss": 2.8608, "step": 24946 }, { "epoch": 1.96, "learning_rate": 8.043832823649338e-05, "loss": 2.7668, "step": 24947 }, { "epoch": 1.96, "learning_rate": 8.043754410726889e-05, "loss": 2.943, "step": 24948 }, { "epoch": 1.96, "learning_rate": 8.043675997804439e-05, "loss": 3.0055, "step": 24949 }, { "epoch": 1.96, "learning_rate": 8.04359758488199e-05, "loss": 2.7525, "step": 24950 }, { "epoch": 1.96, "learning_rate": 8.04351917195954e-05, "loss": 2.9586, "step": 24951 }, { "epoch": 1.96, "learning_rate": 8.04344075903709e-05, "loss": 2.9969, "step": 24952 }, { "epoch": 1.96, "learning_rate": 8.04336234611464e-05, "loss": 3.0043, "step": 24953 }, { "epoch": 1.96, "learning_rate": 8.043283933192191e-05, "loss": 2.788, "step": 24954 }, { "epoch": 1.96, "learning_rate": 8.043205520269741e-05, "loss": 2.7222, "step": 24955 }, { "epoch": 1.96, "learning_rate": 8.043127107347292e-05, "loss": 2.8523, "step": 24956 }, { "epoch": 1.96, "learning_rate": 8.043048694424842e-05, "loss": 2.6933, "step": 24957 }, { "epoch": 1.96, "learning_rate": 8.042970281502392e-05, "loss": 2.9665, "step": 24958 }, { "epoch": 1.96, "learning_rate": 8.042891868579943e-05, "loss": 2.9671, "step": 24959 }, { "epoch": 1.96, "learning_rate": 8.042813455657493e-05, "loss": 2.7199, "step": 24960 }, { "epoch": 1.96, "learning_rate": 8.042735042735043e-05, "loss": 3.2264, "step": 24961 }, { "epoch": 1.96, "learning_rate": 8.042656629812594e-05, "loss": 2.7602, "step": 24962 }, { "epoch": 1.96, "learning_rate": 8.042578216890144e-05, "loss": 2.8443, "step": 24963 }, { "epoch": 1.96, "learning_rate": 8.042499803967695e-05, "loss": 2.923, "step": 24964 }, { "epoch": 1.96, "learning_rate": 8.042421391045245e-05, "loss": 2.9307, "step": 24965 }, { "epoch": 1.96, "learning_rate": 8.042342978122795e-05, "loss": 2.9453, "step": 24966 }, { "epoch": 1.96, "learning_rate": 8.042264565200346e-05, "loss": 2.7391, "step": 24967 }, { "epoch": 1.96, "learning_rate": 8.042186152277896e-05, "loss": 2.9883, "step": 24968 }, { "epoch": 1.96, "learning_rate": 8.042107739355446e-05, "loss": 2.6867, "step": 24969 }, { "epoch": 1.96, "learning_rate": 8.042029326432995e-05, "loss": 3.1127, "step": 24970 }, { "epoch": 1.96, "learning_rate": 8.041950913510546e-05, "loss": 2.8399, "step": 24971 }, { "epoch": 1.96, "learning_rate": 8.041872500588097e-05, "loss": 2.9063, "step": 24972 }, { "epoch": 1.96, "learning_rate": 8.041794087665648e-05, "loss": 3.1749, "step": 24973 }, { "epoch": 1.96, "learning_rate": 8.041715674743198e-05, "loss": 2.7716, "step": 24974 }, { "epoch": 1.96, "learning_rate": 8.041637261820749e-05, "loss": 3.2098, "step": 24975 }, { "epoch": 1.96, "learning_rate": 8.041558848898299e-05, "loss": 2.8782, "step": 24976 }, { "epoch": 1.96, "learning_rate": 8.04148043597585e-05, "loss": 2.923, "step": 24977 }, { "epoch": 1.96, "learning_rate": 8.0414020230534e-05, "loss": 3.0218, "step": 24978 }, { "epoch": 1.96, "learning_rate": 8.04132361013095e-05, "loss": 2.809, "step": 24979 }, { "epoch": 1.96, "learning_rate": 8.0412451972085e-05, "loss": 3.0055, "step": 24980 }, { "epoch": 1.96, "learning_rate": 8.041166784286051e-05, "loss": 3.057, "step": 24981 }, { "epoch": 1.96, "learning_rate": 8.041088371363601e-05, "loss": 2.8678, "step": 24982 }, { "epoch": 1.96, "learning_rate": 8.041009958441152e-05, "loss": 2.8402, "step": 24983 }, { "epoch": 1.96, "learning_rate": 8.040931545518702e-05, "loss": 2.965, "step": 24984 }, { "epoch": 1.96, "learning_rate": 8.040853132596252e-05, "loss": 2.8482, "step": 24985 }, { "epoch": 1.96, "learning_rate": 8.040774719673803e-05, "loss": 3.2353, "step": 24986 }, { "epoch": 1.96, "learning_rate": 8.040696306751353e-05, "loss": 2.8116, "step": 24987 }, { "epoch": 1.96, "learning_rate": 8.040617893828903e-05, "loss": 2.6863, "step": 24988 }, { "epoch": 1.96, "learning_rate": 8.040539480906454e-05, "loss": 2.7685, "step": 24989 }, { "epoch": 1.96, "learning_rate": 8.040461067984004e-05, "loss": 2.9138, "step": 24990 }, { "epoch": 1.96, "learning_rate": 8.040382655061554e-05, "loss": 3.0198, "step": 24991 }, { "epoch": 1.96, "learning_rate": 8.040304242139105e-05, "loss": 2.9712, "step": 24992 }, { "epoch": 1.96, "learning_rate": 8.040225829216655e-05, "loss": 3.2671, "step": 24993 }, { "epoch": 1.96, "learning_rate": 8.040147416294206e-05, "loss": 2.8458, "step": 24994 }, { "epoch": 1.96, "learning_rate": 8.040069003371756e-05, "loss": 3.0473, "step": 24995 }, { "epoch": 1.96, "learning_rate": 8.039990590449306e-05, "loss": 2.914, "step": 24996 }, { "epoch": 1.96, "learning_rate": 8.039912177526857e-05, "loss": 3.0394, "step": 24997 }, { "epoch": 1.96, "learning_rate": 8.039833764604407e-05, "loss": 3.0422, "step": 24998 }, { "epoch": 1.96, "learning_rate": 8.039755351681957e-05, "loss": 2.9324, "step": 24999 }, { "epoch": 1.96, "learning_rate": 8.039676938759508e-05, "loss": 2.6621, "step": 25000 }, { "epoch": 1.96, "eval_loss": 2.9018239974975586, "eval_runtime": 218.0047, "eval_samples_per_second": 51.981, "eval_steps_per_second": 3.252, "step": 25000 }, { "epoch": 1.96, "learning_rate": 8.039598525837058e-05, "loss": 2.9242, "step": 25001 }, { "epoch": 1.96, "learning_rate": 8.039520112914609e-05, "loss": 3.0672, "step": 25002 }, { "epoch": 1.96, "learning_rate": 8.03944169999216e-05, "loss": 2.9826, "step": 25003 }, { "epoch": 1.96, "learning_rate": 8.03936328706971e-05, "loss": 2.6694, "step": 25004 }, { "epoch": 1.96, "learning_rate": 8.039284874147261e-05, "loss": 2.6674, "step": 25005 }, { "epoch": 1.96, "learning_rate": 8.03920646122481e-05, "loss": 3.0708, "step": 25006 }, { "epoch": 1.96, "learning_rate": 8.03912804830236e-05, "loss": 2.9541, "step": 25007 }, { "epoch": 1.96, "learning_rate": 8.039049635379911e-05, "loss": 3.1018, "step": 25008 }, { "epoch": 1.96, "learning_rate": 8.038971222457461e-05, "loss": 2.7936, "step": 25009 }, { "epoch": 1.96, "learning_rate": 8.038892809535011e-05, "loss": 2.9146, "step": 25010 }, { "epoch": 1.96, "learning_rate": 8.038814396612562e-05, "loss": 2.8459, "step": 25011 }, { "epoch": 1.96, "learning_rate": 8.038735983690112e-05, "loss": 2.9527, "step": 25012 }, { "epoch": 1.96, "learning_rate": 8.038657570767663e-05, "loss": 3.0301, "step": 25013 }, { "epoch": 1.96, "learning_rate": 8.038579157845213e-05, "loss": 3.1896, "step": 25014 }, { "epoch": 1.96, "learning_rate": 8.038500744922763e-05, "loss": 2.8679, "step": 25015 }, { "epoch": 1.96, "learning_rate": 8.038422332000314e-05, "loss": 2.8244, "step": 25016 }, { "epoch": 1.96, "learning_rate": 8.038343919077864e-05, "loss": 2.972, "step": 25017 }, { "epoch": 1.96, "learning_rate": 8.038265506155414e-05, "loss": 2.9516, "step": 25018 }, { "epoch": 1.96, "learning_rate": 8.038187093232965e-05, "loss": 2.9543, "step": 25019 }, { "epoch": 1.96, "learning_rate": 8.038108680310515e-05, "loss": 2.8298, "step": 25020 }, { "epoch": 1.96, "learning_rate": 8.038030267388066e-05, "loss": 2.7757, "step": 25021 }, { "epoch": 1.96, "learning_rate": 8.037951854465616e-05, "loss": 2.9794, "step": 25022 }, { "epoch": 1.96, "learning_rate": 8.037873441543166e-05, "loss": 2.9674, "step": 25023 }, { "epoch": 1.96, "learning_rate": 8.037795028620717e-05, "loss": 3.1285, "step": 25024 }, { "epoch": 1.96, "learning_rate": 8.037716615698267e-05, "loss": 2.701, "step": 25025 }, { "epoch": 1.96, "learning_rate": 8.037638202775817e-05, "loss": 2.9646, "step": 25026 }, { "epoch": 1.96, "learning_rate": 8.037559789853368e-05, "loss": 2.8664, "step": 25027 }, { "epoch": 1.96, "learning_rate": 8.037481376930918e-05, "loss": 2.9064, "step": 25028 }, { "epoch": 1.96, "learning_rate": 8.03740296400847e-05, "loss": 2.823, "step": 25029 }, { "epoch": 1.96, "learning_rate": 8.03732455108602e-05, "loss": 2.8233, "step": 25030 }, { "epoch": 1.96, "learning_rate": 8.03724613816357e-05, "loss": 2.8246, "step": 25031 }, { "epoch": 1.96, "learning_rate": 8.037167725241121e-05, "loss": 2.7135, "step": 25032 }, { "epoch": 1.96, "learning_rate": 8.037089312318671e-05, "loss": 2.9322, "step": 25033 }, { "epoch": 1.96, "learning_rate": 8.037010899396222e-05, "loss": 2.6931, "step": 25034 }, { "epoch": 1.96, "learning_rate": 8.036932486473772e-05, "loss": 2.7401, "step": 25035 }, { "epoch": 1.96, "learning_rate": 8.036854073551322e-05, "loss": 3.1418, "step": 25036 }, { "epoch": 1.96, "learning_rate": 8.036775660628873e-05, "loss": 2.974, "step": 25037 }, { "epoch": 1.96, "learning_rate": 8.036697247706423e-05, "loss": 2.7768, "step": 25038 }, { "epoch": 1.96, "learning_rate": 8.036618834783973e-05, "loss": 2.834, "step": 25039 }, { "epoch": 1.96, "learning_rate": 8.036540421861524e-05, "loss": 2.8179, "step": 25040 }, { "epoch": 1.96, "learning_rate": 8.036462008939074e-05, "loss": 3.1945, "step": 25041 }, { "epoch": 1.96, "learning_rate": 8.036383596016623e-05, "loss": 2.6553, "step": 25042 }, { "epoch": 1.96, "learning_rate": 8.036305183094174e-05, "loss": 2.7994, "step": 25043 }, { "epoch": 1.96, "learning_rate": 8.036226770171724e-05, "loss": 2.8988, "step": 25044 }, { "epoch": 1.96, "learning_rate": 8.036148357249274e-05, "loss": 2.9395, "step": 25045 }, { "epoch": 1.96, "learning_rate": 8.036069944326825e-05, "loss": 2.8246, "step": 25046 }, { "epoch": 1.96, "learning_rate": 8.035991531404375e-05, "loss": 2.95, "step": 25047 }, { "epoch": 1.96, "learning_rate": 8.035913118481925e-05, "loss": 2.7144, "step": 25048 }, { "epoch": 1.96, "learning_rate": 8.035834705559476e-05, "loss": 2.9491, "step": 25049 }, { "epoch": 1.96, "learning_rate": 8.035756292637026e-05, "loss": 2.6928, "step": 25050 }, { "epoch": 1.96, "learning_rate": 8.035677879714577e-05, "loss": 2.8538, "step": 25051 }, { "epoch": 1.96, "learning_rate": 8.035599466792127e-05, "loss": 2.9665, "step": 25052 }, { "epoch": 1.96, "learning_rate": 8.035521053869677e-05, "loss": 2.8388, "step": 25053 }, { "epoch": 1.96, "learning_rate": 8.035442640947228e-05, "loss": 2.874, "step": 25054 }, { "epoch": 1.96, "learning_rate": 8.03536422802478e-05, "loss": 2.9044, "step": 25055 }, { "epoch": 1.96, "learning_rate": 8.03528581510233e-05, "loss": 3.0176, "step": 25056 }, { "epoch": 1.96, "learning_rate": 8.03520740217988e-05, "loss": 2.9095, "step": 25057 }, { "epoch": 1.96, "learning_rate": 8.03512898925743e-05, "loss": 2.9669, "step": 25058 }, { "epoch": 1.96, "learning_rate": 8.035050576334981e-05, "loss": 2.9588, "step": 25059 }, { "epoch": 1.97, "learning_rate": 8.034972163412531e-05, "loss": 2.9256, "step": 25060 }, { "epoch": 1.97, "learning_rate": 8.034893750490082e-05, "loss": 2.9433, "step": 25061 }, { "epoch": 1.97, "learning_rate": 8.034815337567632e-05, "loss": 2.9647, "step": 25062 }, { "epoch": 1.97, "learning_rate": 8.034736924645182e-05, "loss": 2.6724, "step": 25063 }, { "epoch": 1.97, "learning_rate": 8.034658511722733e-05, "loss": 2.806, "step": 25064 }, { "epoch": 1.97, "learning_rate": 8.034580098800283e-05, "loss": 2.9022, "step": 25065 }, { "epoch": 1.97, "learning_rate": 8.034501685877833e-05, "loss": 2.8959, "step": 25066 }, { "epoch": 1.97, "learning_rate": 8.034423272955384e-05, "loss": 3.1515, "step": 25067 }, { "epoch": 1.97, "learning_rate": 8.034344860032934e-05, "loss": 2.8095, "step": 25068 }, { "epoch": 1.97, "learning_rate": 8.034266447110485e-05, "loss": 3.1494, "step": 25069 }, { "epoch": 1.97, "learning_rate": 8.034188034188035e-05, "loss": 2.8903, "step": 25070 }, { "epoch": 1.97, "learning_rate": 8.034109621265585e-05, "loss": 2.8253, "step": 25071 }, { "epoch": 1.97, "learning_rate": 8.034031208343136e-05, "loss": 2.9469, "step": 25072 }, { "epoch": 1.97, "learning_rate": 8.033952795420686e-05, "loss": 3.0848, "step": 25073 }, { "epoch": 1.97, "learning_rate": 8.033874382498236e-05, "loss": 2.8793, "step": 25074 }, { "epoch": 1.97, "learning_rate": 8.033795969575787e-05, "loss": 2.7171, "step": 25075 }, { "epoch": 1.97, "learning_rate": 8.033717556653337e-05, "loss": 2.8195, "step": 25076 }, { "epoch": 1.97, "learning_rate": 8.033639143730887e-05, "loss": 3.0109, "step": 25077 }, { "epoch": 1.97, "learning_rate": 8.033560730808436e-05, "loss": 2.7651, "step": 25078 }, { "epoch": 1.97, "learning_rate": 8.033482317885987e-05, "loss": 2.9639, "step": 25079 }, { "epoch": 1.97, "learning_rate": 8.033403904963539e-05, "loss": 2.8917, "step": 25080 }, { "epoch": 1.97, "learning_rate": 8.033325492041089e-05, "loss": 2.8775, "step": 25081 }, { "epoch": 1.97, "learning_rate": 8.033247079118639e-05, "loss": 2.8826, "step": 25082 }, { "epoch": 1.97, "learning_rate": 8.03316866619619e-05, "loss": 2.7241, "step": 25083 }, { "epoch": 1.97, "learning_rate": 8.03309025327374e-05, "loss": 2.8637, "step": 25084 }, { "epoch": 1.97, "learning_rate": 8.03301184035129e-05, "loss": 2.8785, "step": 25085 }, { "epoch": 1.97, "learning_rate": 8.032933427428841e-05, "loss": 2.6646, "step": 25086 }, { "epoch": 1.97, "learning_rate": 8.032855014506391e-05, "loss": 2.7822, "step": 25087 }, { "epoch": 1.97, "learning_rate": 8.032776601583942e-05, "loss": 3.0002, "step": 25088 }, { "epoch": 1.97, "learning_rate": 8.032698188661492e-05, "loss": 2.9251, "step": 25089 }, { "epoch": 1.97, "learning_rate": 8.032619775739042e-05, "loss": 2.9805, "step": 25090 }, { "epoch": 1.97, "learning_rate": 8.032541362816593e-05, "loss": 2.8804, "step": 25091 }, { "epoch": 1.97, "learning_rate": 8.032462949894143e-05, "loss": 2.8892, "step": 25092 }, { "epoch": 1.97, "learning_rate": 8.032384536971693e-05, "loss": 2.6894, "step": 25093 }, { "epoch": 1.97, "learning_rate": 8.032306124049244e-05, "loss": 3.0597, "step": 25094 }, { "epoch": 1.97, "learning_rate": 8.032227711126794e-05, "loss": 2.838, "step": 25095 }, { "epoch": 1.97, "learning_rate": 8.032149298204344e-05, "loss": 2.9349, "step": 25096 }, { "epoch": 1.97, "learning_rate": 8.032070885281895e-05, "loss": 2.8575, "step": 25097 }, { "epoch": 1.97, "learning_rate": 8.031992472359445e-05, "loss": 3.0246, "step": 25098 }, { "epoch": 1.97, "learning_rate": 8.031914059436996e-05, "loss": 2.5718, "step": 25099 }, { "epoch": 1.97, "learning_rate": 8.031835646514546e-05, "loss": 2.8344, "step": 25100 }, { "epoch": 1.97, "learning_rate": 8.031757233592096e-05, "loss": 2.9511, "step": 25101 }, { "epoch": 1.97, "learning_rate": 8.031678820669647e-05, "loss": 2.8909, "step": 25102 }, { "epoch": 1.97, "learning_rate": 8.031600407747197e-05, "loss": 2.9529, "step": 25103 }, { "epoch": 1.97, "learning_rate": 8.031521994824747e-05, "loss": 2.691, "step": 25104 }, { "epoch": 1.97, "learning_rate": 8.031443581902298e-05, "loss": 2.8387, "step": 25105 }, { "epoch": 1.97, "learning_rate": 8.031365168979848e-05, "loss": 2.6358, "step": 25106 }, { "epoch": 1.97, "learning_rate": 8.031286756057398e-05, "loss": 2.9987, "step": 25107 }, { "epoch": 1.97, "learning_rate": 8.031208343134949e-05, "loss": 2.8291, "step": 25108 }, { "epoch": 1.97, "learning_rate": 8.031129930212499e-05, "loss": 2.9763, "step": 25109 }, { "epoch": 1.97, "learning_rate": 8.03105151729005e-05, "loss": 3.0349, "step": 25110 }, { "epoch": 1.97, "learning_rate": 8.0309731043676e-05, "loss": 2.9359, "step": 25111 }, { "epoch": 1.97, "learning_rate": 8.030894691445152e-05, "loss": 2.826, "step": 25112 }, { "epoch": 1.97, "learning_rate": 8.030816278522702e-05, "loss": 3.0403, "step": 25113 }, { "epoch": 1.97, "learning_rate": 8.030737865600251e-05, "loss": 2.9741, "step": 25114 }, { "epoch": 1.97, "learning_rate": 8.030659452677801e-05, "loss": 2.9315, "step": 25115 }, { "epoch": 1.97, "learning_rate": 8.030581039755352e-05, "loss": 2.9792, "step": 25116 }, { "epoch": 1.97, "learning_rate": 8.030502626832902e-05, "loss": 2.8154, "step": 25117 }, { "epoch": 1.97, "learning_rate": 8.030424213910453e-05, "loss": 2.87, "step": 25118 }, { "epoch": 1.97, "learning_rate": 8.030345800988003e-05, "loss": 2.9146, "step": 25119 }, { "epoch": 1.97, "learning_rate": 8.030267388065553e-05, "loss": 2.9288, "step": 25120 }, { "epoch": 1.97, "learning_rate": 8.030188975143104e-05, "loss": 3.0209, "step": 25121 }, { "epoch": 1.97, "learning_rate": 8.030110562220654e-05, "loss": 2.7734, "step": 25122 }, { "epoch": 1.97, "learning_rate": 8.030032149298204e-05, "loss": 2.677, "step": 25123 }, { "epoch": 1.97, "learning_rate": 8.029953736375755e-05, "loss": 2.7839, "step": 25124 }, { "epoch": 1.97, "learning_rate": 8.029875323453305e-05, "loss": 2.7824, "step": 25125 }, { "epoch": 1.97, "learning_rate": 8.029796910530855e-05, "loss": 2.9875, "step": 25126 }, { "epoch": 1.97, "learning_rate": 8.029718497608406e-05, "loss": 2.747, "step": 25127 }, { "epoch": 1.97, "learning_rate": 8.029640084685956e-05, "loss": 3.0164, "step": 25128 }, { "epoch": 1.97, "learning_rate": 8.029561671763507e-05, "loss": 2.7469, "step": 25129 }, { "epoch": 1.97, "learning_rate": 8.029483258841057e-05, "loss": 2.9531, "step": 25130 }, { "epoch": 1.97, "learning_rate": 8.029404845918607e-05, "loss": 2.754, "step": 25131 }, { "epoch": 1.97, "learning_rate": 8.029326432996158e-05, "loss": 2.9676, "step": 25132 }, { "epoch": 1.97, "learning_rate": 8.029248020073708e-05, "loss": 2.8624, "step": 25133 }, { "epoch": 1.97, "learning_rate": 8.029169607151258e-05, "loss": 2.5924, "step": 25134 }, { "epoch": 1.97, "learning_rate": 8.029091194228809e-05, "loss": 2.8306, "step": 25135 }, { "epoch": 1.97, "learning_rate": 8.029012781306359e-05, "loss": 2.8756, "step": 25136 }, { "epoch": 1.97, "learning_rate": 8.028934368383911e-05, "loss": 3.0821, "step": 25137 }, { "epoch": 1.97, "learning_rate": 8.028855955461461e-05, "loss": 2.7218, "step": 25138 }, { "epoch": 1.97, "learning_rate": 8.028777542539012e-05, "loss": 2.8763, "step": 25139 }, { "epoch": 1.97, "learning_rate": 8.028699129616562e-05, "loss": 2.8648, "step": 25140 }, { "epoch": 1.97, "learning_rate": 8.028620716694112e-05, "loss": 2.7578, "step": 25141 }, { "epoch": 1.97, "learning_rate": 8.028542303771663e-05, "loss": 2.8032, "step": 25142 }, { "epoch": 1.97, "learning_rate": 8.028463890849213e-05, "loss": 2.7845, "step": 25143 }, { "epoch": 1.97, "learning_rate": 8.028385477926763e-05, "loss": 3.0234, "step": 25144 }, { "epoch": 1.97, "learning_rate": 8.028307065004314e-05, "loss": 2.7, "step": 25145 }, { "epoch": 1.97, "learning_rate": 8.028228652081864e-05, "loss": 2.9201, "step": 25146 }, { "epoch": 1.97, "learning_rate": 8.028150239159415e-05, "loss": 3.1979, "step": 25147 }, { "epoch": 1.97, "learning_rate": 8.028071826236965e-05, "loss": 3.1249, "step": 25148 }, { "epoch": 1.97, "learning_rate": 8.027993413314515e-05, "loss": 2.9738, "step": 25149 }, { "epoch": 1.97, "learning_rate": 8.027915000392064e-05, "loss": 3.0298, "step": 25150 }, { "epoch": 1.97, "learning_rate": 8.027836587469615e-05, "loss": 2.847, "step": 25151 }, { "epoch": 1.97, "learning_rate": 8.027758174547165e-05, "loss": 2.8324, "step": 25152 }, { "epoch": 1.97, "learning_rate": 8.027679761624715e-05, "loss": 2.9107, "step": 25153 }, { "epoch": 1.97, "learning_rate": 8.027601348702266e-05, "loss": 2.7666, "step": 25154 }, { "epoch": 1.97, "learning_rate": 8.027522935779816e-05, "loss": 2.874, "step": 25155 }, { "epoch": 1.97, "learning_rate": 8.027444522857367e-05, "loss": 2.8529, "step": 25156 }, { "epoch": 1.97, "learning_rate": 8.027366109934917e-05, "loss": 2.7851, "step": 25157 }, { "epoch": 1.97, "learning_rate": 8.027287697012467e-05, "loss": 2.868, "step": 25158 }, { "epoch": 1.97, "learning_rate": 8.027209284090018e-05, "loss": 3.018, "step": 25159 }, { "epoch": 1.97, "learning_rate": 8.027130871167568e-05, "loss": 2.8903, "step": 25160 }, { "epoch": 1.97, "learning_rate": 8.027052458245118e-05, "loss": 2.7205, "step": 25161 }, { "epoch": 1.97, "learning_rate": 8.026974045322669e-05, "loss": 2.8405, "step": 25162 }, { "epoch": 1.97, "learning_rate": 8.02689563240022e-05, "loss": 3.0751, "step": 25163 }, { "epoch": 1.97, "learning_rate": 8.026817219477771e-05, "loss": 2.7234, "step": 25164 }, { "epoch": 1.97, "learning_rate": 8.026738806555321e-05, "loss": 3.0581, "step": 25165 }, { "epoch": 1.97, "learning_rate": 8.026660393632872e-05, "loss": 2.8663, "step": 25166 }, { "epoch": 1.97, "learning_rate": 8.026581980710422e-05, "loss": 2.8689, "step": 25167 }, { "epoch": 1.97, "learning_rate": 8.026503567787972e-05, "loss": 3.118, "step": 25168 }, { "epoch": 1.97, "learning_rate": 8.026425154865523e-05, "loss": 3.0298, "step": 25169 }, { "epoch": 1.97, "learning_rate": 8.026346741943073e-05, "loss": 2.7355, "step": 25170 }, { "epoch": 1.97, "learning_rate": 8.026268329020623e-05, "loss": 2.9987, "step": 25171 }, { "epoch": 1.97, "learning_rate": 8.026189916098174e-05, "loss": 2.9921, "step": 25172 }, { "epoch": 1.97, "learning_rate": 8.026111503175724e-05, "loss": 2.8036, "step": 25173 }, { "epoch": 1.97, "learning_rate": 8.026033090253274e-05, "loss": 2.6096, "step": 25174 }, { "epoch": 1.97, "learning_rate": 8.025954677330825e-05, "loss": 2.7714, "step": 25175 }, { "epoch": 1.97, "learning_rate": 8.025876264408375e-05, "loss": 2.9235, "step": 25176 }, { "epoch": 1.97, "learning_rate": 8.025797851485926e-05, "loss": 3.0369, "step": 25177 }, { "epoch": 1.97, "learning_rate": 8.025719438563476e-05, "loss": 2.9583, "step": 25178 }, { "epoch": 1.97, "learning_rate": 8.025641025641026e-05, "loss": 2.9283, "step": 25179 }, { "epoch": 1.97, "learning_rate": 8.025562612718577e-05, "loss": 3.013, "step": 25180 }, { "epoch": 1.97, "learning_rate": 8.025484199796127e-05, "loss": 2.8107, "step": 25181 }, { "epoch": 1.97, "learning_rate": 8.025405786873677e-05, "loss": 2.8709, "step": 25182 }, { "epoch": 1.97, "learning_rate": 8.025327373951228e-05, "loss": 2.9576, "step": 25183 }, { "epoch": 1.97, "learning_rate": 8.025248961028778e-05, "loss": 2.7537, "step": 25184 }, { "epoch": 1.97, "learning_rate": 8.025170548106329e-05, "loss": 3.1812, "step": 25185 }, { "epoch": 1.97, "learning_rate": 8.025092135183878e-05, "loss": 2.9962, "step": 25186 }, { "epoch": 1.97, "learning_rate": 8.025013722261428e-05, "loss": 2.8085, "step": 25187 }, { "epoch": 1.98, "learning_rate": 8.02493530933898e-05, "loss": 2.9731, "step": 25188 }, { "epoch": 1.98, "learning_rate": 8.02485689641653e-05, "loss": 2.8835, "step": 25189 }, { "epoch": 1.98, "learning_rate": 8.02477848349408e-05, "loss": 2.9644, "step": 25190 }, { "epoch": 1.98, "learning_rate": 8.024700070571631e-05, "loss": 2.9126, "step": 25191 }, { "epoch": 1.98, "learning_rate": 8.024621657649181e-05, "loss": 2.7825, "step": 25192 }, { "epoch": 1.98, "learning_rate": 8.024543244726731e-05, "loss": 3.0023, "step": 25193 }, { "epoch": 1.98, "learning_rate": 8.024464831804282e-05, "loss": 3.0418, "step": 25194 }, { "epoch": 1.98, "learning_rate": 8.024386418881832e-05, "loss": 2.7439, "step": 25195 }, { "epoch": 1.98, "learning_rate": 8.024308005959383e-05, "loss": 2.7828, "step": 25196 }, { "epoch": 1.98, "learning_rate": 8.024229593036933e-05, "loss": 2.7634, "step": 25197 }, { "epoch": 1.98, "learning_rate": 8.024151180114483e-05, "loss": 2.9883, "step": 25198 }, { "epoch": 1.98, "learning_rate": 8.024072767192034e-05, "loss": 2.8263, "step": 25199 }, { "epoch": 1.98, "learning_rate": 8.023994354269584e-05, "loss": 3.1274, "step": 25200 }, { "epoch": 1.98, "learning_rate": 8.023915941347134e-05, "loss": 3.0348, "step": 25201 }, { "epoch": 1.98, "learning_rate": 8.023837528424685e-05, "loss": 2.6953, "step": 25202 }, { "epoch": 1.98, "learning_rate": 8.023759115502235e-05, "loss": 2.7216, "step": 25203 }, { "epoch": 1.98, "learning_rate": 8.023680702579786e-05, "loss": 2.7897, "step": 25204 }, { "epoch": 1.98, "learning_rate": 8.023602289657336e-05, "loss": 2.9617, "step": 25205 }, { "epoch": 1.98, "learning_rate": 8.023523876734886e-05, "loss": 3.578, "step": 25206 }, { "epoch": 1.98, "learning_rate": 8.023445463812437e-05, "loss": 2.8419, "step": 25207 }, { "epoch": 1.98, "learning_rate": 8.023367050889987e-05, "loss": 2.8021, "step": 25208 }, { "epoch": 1.98, "learning_rate": 8.023288637967537e-05, "loss": 2.8423, "step": 25209 }, { "epoch": 1.98, "learning_rate": 8.023210225045088e-05, "loss": 2.9873, "step": 25210 }, { "epoch": 1.98, "learning_rate": 8.023131812122638e-05, "loss": 2.8824, "step": 25211 }, { "epoch": 1.98, "learning_rate": 8.023053399200188e-05, "loss": 3.1316, "step": 25212 }, { "epoch": 1.98, "learning_rate": 8.022974986277739e-05, "loss": 2.8786, "step": 25213 }, { "epoch": 1.98, "learning_rate": 8.022896573355289e-05, "loss": 2.8982, "step": 25214 }, { "epoch": 1.98, "learning_rate": 8.02281816043284e-05, "loss": 2.8053, "step": 25215 }, { "epoch": 1.98, "learning_rate": 8.02273974751039e-05, "loss": 2.8983, "step": 25216 }, { "epoch": 1.98, "learning_rate": 8.02266133458794e-05, "loss": 2.9707, "step": 25217 }, { "epoch": 1.98, "learning_rate": 8.02258292166549e-05, "loss": 2.8393, "step": 25218 }, { "epoch": 1.98, "learning_rate": 8.022504508743041e-05, "loss": 3.2166, "step": 25219 }, { "epoch": 1.98, "learning_rate": 8.022426095820593e-05, "loss": 2.861, "step": 25220 }, { "epoch": 1.98, "learning_rate": 8.022347682898143e-05, "loss": 2.9732, "step": 25221 }, { "epoch": 1.98, "learning_rate": 8.022269269975692e-05, "loss": 2.786, "step": 25222 }, { "epoch": 1.98, "learning_rate": 8.022190857053243e-05, "loss": 3.0844, "step": 25223 }, { "epoch": 1.98, "learning_rate": 8.022112444130793e-05, "loss": 2.8606, "step": 25224 }, { "epoch": 1.98, "learning_rate": 8.022034031208343e-05, "loss": 2.9152, "step": 25225 }, { "epoch": 1.98, "learning_rate": 8.021955618285894e-05, "loss": 2.8448, "step": 25226 }, { "epoch": 1.98, "learning_rate": 8.021877205363444e-05, "loss": 3.0812, "step": 25227 }, { "epoch": 1.98, "learning_rate": 8.021798792440994e-05, "loss": 2.8462, "step": 25228 }, { "epoch": 1.98, "learning_rate": 8.021720379518545e-05, "loss": 2.9221, "step": 25229 }, { "epoch": 1.98, "learning_rate": 8.021641966596095e-05, "loss": 2.6532, "step": 25230 }, { "epoch": 1.98, "learning_rate": 8.021563553673645e-05, "loss": 2.9534, "step": 25231 }, { "epoch": 1.98, "learning_rate": 8.021485140751196e-05, "loss": 3.1619, "step": 25232 }, { "epoch": 1.98, "learning_rate": 8.021406727828746e-05, "loss": 3.0069, "step": 25233 }, { "epoch": 1.98, "learning_rate": 8.021328314906297e-05, "loss": 2.987, "step": 25234 }, { "epoch": 1.98, "learning_rate": 8.021249901983847e-05, "loss": 3.2883, "step": 25235 }, { "epoch": 1.98, "learning_rate": 8.021171489061397e-05, "loss": 2.9174, "step": 25236 }, { "epoch": 1.98, "learning_rate": 8.021093076138948e-05, "loss": 2.9245, "step": 25237 }, { "epoch": 1.98, "learning_rate": 8.021014663216498e-05, "loss": 3.1354, "step": 25238 }, { "epoch": 1.98, "learning_rate": 8.020936250294048e-05, "loss": 2.8301, "step": 25239 }, { "epoch": 1.98, "learning_rate": 8.020857837371599e-05, "loss": 2.8007, "step": 25240 }, { "epoch": 1.98, "learning_rate": 8.020779424449149e-05, "loss": 2.8739, "step": 25241 }, { "epoch": 1.98, "learning_rate": 8.0207010115267e-05, "loss": 2.9917, "step": 25242 }, { "epoch": 1.98, "learning_rate": 8.02062259860425e-05, "loss": 2.767, "step": 25243 }, { "epoch": 1.98, "learning_rate": 8.0205441856818e-05, "loss": 3.064, "step": 25244 }, { "epoch": 1.98, "learning_rate": 8.020465772759352e-05, "loss": 2.9582, "step": 25245 }, { "epoch": 1.98, "learning_rate": 8.020387359836902e-05, "loss": 2.8518, "step": 25246 }, { "epoch": 1.98, "learning_rate": 8.020308946914453e-05, "loss": 2.7732, "step": 25247 }, { "epoch": 1.98, "learning_rate": 8.020230533992003e-05, "loss": 2.9004, "step": 25248 }, { "epoch": 1.98, "learning_rate": 8.020152121069553e-05, "loss": 3.0388, "step": 25249 }, { "epoch": 1.98, "learning_rate": 8.020073708147104e-05, "loss": 3.1008, "step": 25250 }, { "epoch": 1.98, "learning_rate": 8.019995295224654e-05, "loss": 2.8338, "step": 25251 }, { "epoch": 1.98, "learning_rate": 8.019916882302205e-05, "loss": 2.7638, "step": 25252 }, { "epoch": 1.98, "learning_rate": 8.019838469379755e-05, "loss": 2.9975, "step": 25253 }, { "epoch": 1.98, "learning_rate": 8.019760056457305e-05, "loss": 2.8253, "step": 25254 }, { "epoch": 1.98, "learning_rate": 8.019681643534856e-05, "loss": 2.7324, "step": 25255 }, { "epoch": 1.98, "learning_rate": 8.019603230612406e-05, "loss": 2.6622, "step": 25256 }, { "epoch": 1.98, "learning_rate": 8.019524817689956e-05, "loss": 2.9357, "step": 25257 }, { "epoch": 1.98, "learning_rate": 8.019446404767505e-05, "loss": 2.7586, "step": 25258 }, { "epoch": 1.98, "learning_rate": 8.019367991845056e-05, "loss": 2.961, "step": 25259 }, { "epoch": 1.98, "learning_rate": 8.019289578922606e-05, "loss": 2.842, "step": 25260 }, { "epoch": 1.98, "learning_rate": 8.019211166000156e-05, "loss": 2.9363, "step": 25261 }, { "epoch": 1.98, "learning_rate": 8.019132753077707e-05, "loss": 3.0059, "step": 25262 }, { "epoch": 1.98, "learning_rate": 8.019054340155257e-05, "loss": 2.8434, "step": 25263 }, { "epoch": 1.98, "learning_rate": 8.018975927232808e-05, "loss": 3.0581, "step": 25264 }, { "epoch": 1.98, "learning_rate": 8.018897514310358e-05, "loss": 3.0853, "step": 25265 }, { "epoch": 1.98, "learning_rate": 8.018819101387908e-05, "loss": 3.0737, "step": 25266 }, { "epoch": 1.98, "learning_rate": 8.018740688465459e-05, "loss": 2.8595, "step": 25267 }, { "epoch": 1.98, "learning_rate": 8.018662275543009e-05, "loss": 2.9076, "step": 25268 }, { "epoch": 1.98, "learning_rate": 8.01858386262056e-05, "loss": 2.8686, "step": 25269 }, { "epoch": 1.98, "learning_rate": 8.01850544969811e-05, "loss": 2.915, "step": 25270 }, { "epoch": 1.98, "learning_rate": 8.018427036775662e-05, "loss": 2.876, "step": 25271 }, { "epoch": 1.98, "learning_rate": 8.018348623853212e-05, "loss": 2.8799, "step": 25272 }, { "epoch": 1.98, "learning_rate": 8.018270210930762e-05, "loss": 2.9889, "step": 25273 }, { "epoch": 1.98, "learning_rate": 8.018191798008313e-05, "loss": 2.9661, "step": 25274 }, { "epoch": 1.98, "learning_rate": 8.018113385085863e-05, "loss": 2.9879, "step": 25275 }, { "epoch": 1.98, "learning_rate": 8.018034972163413e-05, "loss": 2.8481, "step": 25276 }, { "epoch": 1.98, "learning_rate": 8.017956559240964e-05, "loss": 3.0792, "step": 25277 }, { "epoch": 1.98, "learning_rate": 8.017878146318514e-05, "loss": 2.9354, "step": 25278 }, { "epoch": 1.98, "learning_rate": 8.017799733396064e-05, "loss": 2.9726, "step": 25279 }, { "epoch": 1.98, "learning_rate": 8.017721320473615e-05, "loss": 2.9347, "step": 25280 }, { "epoch": 1.98, "learning_rate": 8.017642907551165e-05, "loss": 2.9061, "step": 25281 }, { "epoch": 1.98, "learning_rate": 8.017564494628716e-05, "loss": 2.9495, "step": 25282 }, { "epoch": 1.98, "learning_rate": 8.017486081706266e-05, "loss": 3.0203, "step": 25283 }, { "epoch": 1.98, "learning_rate": 8.017407668783816e-05, "loss": 3.0691, "step": 25284 }, { "epoch": 1.98, "learning_rate": 8.017329255861367e-05, "loss": 2.8867, "step": 25285 }, { "epoch": 1.98, "learning_rate": 8.017250842938917e-05, "loss": 3.0119, "step": 25286 }, { "epoch": 1.98, "learning_rate": 8.017172430016467e-05, "loss": 2.8391, "step": 25287 }, { "epoch": 1.98, "learning_rate": 8.017094017094018e-05, "loss": 2.8973, "step": 25288 }, { "epoch": 1.98, "learning_rate": 8.017015604171568e-05, "loss": 2.9938, "step": 25289 }, { "epoch": 1.98, "learning_rate": 8.016937191249119e-05, "loss": 2.8695, "step": 25290 }, { "epoch": 1.98, "learning_rate": 8.016858778326669e-05, "loss": 3.0486, "step": 25291 }, { "epoch": 1.98, "learning_rate": 8.016780365404219e-05, "loss": 2.7812, "step": 25292 }, { "epoch": 1.98, "learning_rate": 8.01670195248177e-05, "loss": 2.7034, "step": 25293 }, { "epoch": 1.98, "learning_rate": 8.016623539559319e-05, "loss": 3.1668, "step": 25294 }, { "epoch": 1.98, "learning_rate": 8.016545126636869e-05, "loss": 2.8975, "step": 25295 }, { "epoch": 1.98, "learning_rate": 8.016466713714421e-05, "loss": 2.8461, "step": 25296 }, { "epoch": 1.98, "learning_rate": 8.016388300791971e-05, "loss": 2.8316, "step": 25297 }, { "epoch": 1.98, "learning_rate": 8.016309887869521e-05, "loss": 2.8427, "step": 25298 }, { "epoch": 1.98, "learning_rate": 8.016231474947072e-05, "loss": 2.8632, "step": 25299 }, { "epoch": 1.98, "learning_rate": 8.016153062024622e-05, "loss": 2.9865, "step": 25300 }, { "epoch": 1.98, "learning_rate": 8.016074649102173e-05, "loss": 2.9182, "step": 25301 }, { "epoch": 1.98, "learning_rate": 8.015996236179723e-05, "loss": 3.0269, "step": 25302 }, { "epoch": 1.98, "learning_rate": 8.015917823257273e-05, "loss": 2.6886, "step": 25303 }, { "epoch": 1.98, "learning_rate": 8.015839410334824e-05, "loss": 2.7661, "step": 25304 }, { "epoch": 1.98, "learning_rate": 8.015760997412374e-05, "loss": 2.7653, "step": 25305 }, { "epoch": 1.98, "learning_rate": 8.015682584489924e-05, "loss": 3.0464, "step": 25306 }, { "epoch": 1.98, "learning_rate": 8.015604171567475e-05, "loss": 3.183, "step": 25307 }, { "epoch": 1.98, "learning_rate": 8.015525758645025e-05, "loss": 2.8913, "step": 25308 }, { "epoch": 1.98, "learning_rate": 8.015447345722575e-05, "loss": 2.9788, "step": 25309 }, { "epoch": 1.98, "learning_rate": 8.015368932800126e-05, "loss": 2.7703, "step": 25310 }, { "epoch": 1.98, "learning_rate": 8.015290519877676e-05, "loss": 2.7547, "step": 25311 }, { "epoch": 1.98, "learning_rate": 8.015212106955227e-05, "loss": 3.1171, "step": 25312 }, { "epoch": 1.98, "learning_rate": 8.015133694032777e-05, "loss": 2.9654, "step": 25313 }, { "epoch": 1.98, "learning_rate": 8.015055281110327e-05, "loss": 2.945, "step": 25314 }, { "epoch": 1.99, "learning_rate": 8.014976868187878e-05, "loss": 2.7125, "step": 25315 }, { "epoch": 1.99, "learning_rate": 8.014898455265428e-05, "loss": 2.8696, "step": 25316 }, { "epoch": 1.99, "learning_rate": 8.014820042342978e-05, "loss": 2.8202, "step": 25317 }, { "epoch": 1.99, "learning_rate": 8.014741629420529e-05, "loss": 2.8275, "step": 25318 }, { "epoch": 1.99, "learning_rate": 8.014663216498079e-05, "loss": 2.7941, "step": 25319 }, { "epoch": 1.99, "learning_rate": 8.01458480357563e-05, "loss": 2.8038, "step": 25320 }, { "epoch": 1.99, "learning_rate": 8.01450639065318e-05, "loss": 2.955, "step": 25321 }, { "epoch": 1.99, "learning_rate": 8.01442797773073e-05, "loss": 2.9616, "step": 25322 }, { "epoch": 1.99, "learning_rate": 8.01434956480828e-05, "loss": 2.9656, "step": 25323 }, { "epoch": 1.99, "learning_rate": 8.014271151885831e-05, "loss": 2.7156, "step": 25324 }, { "epoch": 1.99, "learning_rate": 8.014192738963381e-05, "loss": 2.907, "step": 25325 }, { "epoch": 1.99, "learning_rate": 8.014114326040932e-05, "loss": 3.0056, "step": 25326 }, { "epoch": 1.99, "learning_rate": 8.014035913118482e-05, "loss": 3.0543, "step": 25327 }, { "epoch": 1.99, "learning_rate": 8.013957500196034e-05, "loss": 2.7736, "step": 25328 }, { "epoch": 1.99, "learning_rate": 8.013879087273584e-05, "loss": 2.7837, "step": 25329 }, { "epoch": 1.99, "learning_rate": 8.013800674351133e-05, "loss": 2.8783, "step": 25330 }, { "epoch": 1.99, "learning_rate": 8.013722261428684e-05, "loss": 2.8315, "step": 25331 }, { "epoch": 1.99, "learning_rate": 8.013643848506234e-05, "loss": 2.6802, "step": 25332 }, { "epoch": 1.99, "learning_rate": 8.013565435583784e-05, "loss": 2.8528, "step": 25333 }, { "epoch": 1.99, "learning_rate": 8.013487022661335e-05, "loss": 2.8683, "step": 25334 }, { "epoch": 1.99, "learning_rate": 8.013408609738885e-05, "loss": 2.6741, "step": 25335 }, { "epoch": 1.99, "learning_rate": 8.013330196816435e-05, "loss": 2.9758, "step": 25336 }, { "epoch": 1.99, "learning_rate": 8.013251783893986e-05, "loss": 2.6406, "step": 25337 }, { "epoch": 1.99, "learning_rate": 8.013173370971536e-05, "loss": 3.0153, "step": 25338 }, { "epoch": 1.99, "learning_rate": 8.013094958049087e-05, "loss": 2.7611, "step": 25339 }, { "epoch": 1.99, "learning_rate": 8.013016545126637e-05, "loss": 2.7828, "step": 25340 }, { "epoch": 1.99, "learning_rate": 8.012938132204187e-05, "loss": 2.8379, "step": 25341 }, { "epoch": 1.99, "learning_rate": 8.012859719281738e-05, "loss": 2.7968, "step": 25342 }, { "epoch": 1.99, "learning_rate": 8.012781306359288e-05, "loss": 3.03, "step": 25343 }, { "epoch": 1.99, "learning_rate": 8.012702893436838e-05, "loss": 2.8852, "step": 25344 }, { "epoch": 1.99, "learning_rate": 8.012624480514389e-05, "loss": 2.7298, "step": 25345 }, { "epoch": 1.99, "learning_rate": 8.012546067591939e-05, "loss": 3.1477, "step": 25346 }, { "epoch": 1.99, "learning_rate": 8.01246765466949e-05, "loss": 2.7563, "step": 25347 }, { "epoch": 1.99, "learning_rate": 8.01238924174704e-05, "loss": 2.7224, "step": 25348 }, { "epoch": 1.99, "learning_rate": 8.01231082882459e-05, "loss": 2.7483, "step": 25349 }, { "epoch": 1.99, "learning_rate": 8.01223241590214e-05, "loss": 3.1409, "step": 25350 }, { "epoch": 1.99, "learning_rate": 8.012154002979691e-05, "loss": 3.1076, "step": 25351 }, { "epoch": 1.99, "learning_rate": 8.012075590057241e-05, "loss": 2.9498, "step": 25352 }, { "epoch": 1.99, "learning_rate": 8.011997177134793e-05, "loss": 2.9514, "step": 25353 }, { "epoch": 1.99, "learning_rate": 8.011918764212343e-05, "loss": 3.0157, "step": 25354 }, { "epoch": 1.99, "learning_rate": 8.011840351289894e-05, "loss": 2.8051, "step": 25355 }, { "epoch": 1.99, "learning_rate": 8.011761938367444e-05, "loss": 2.7671, "step": 25356 }, { "epoch": 1.99, "learning_rate": 8.011683525444995e-05, "loss": 2.9003, "step": 25357 }, { "epoch": 1.99, "learning_rate": 8.011605112522545e-05, "loss": 2.5159, "step": 25358 }, { "epoch": 1.99, "learning_rate": 8.011526699600095e-05, "loss": 2.9098, "step": 25359 }, { "epoch": 1.99, "learning_rate": 8.011448286677646e-05, "loss": 2.8443, "step": 25360 }, { "epoch": 1.99, "learning_rate": 8.011369873755196e-05, "loss": 3.0072, "step": 25361 }, { "epoch": 1.99, "learning_rate": 8.011291460832746e-05, "loss": 2.8922, "step": 25362 }, { "epoch": 1.99, "learning_rate": 8.011213047910297e-05, "loss": 2.7731, "step": 25363 }, { "epoch": 1.99, "learning_rate": 8.011134634987847e-05, "loss": 2.8668, "step": 25364 }, { "epoch": 1.99, "learning_rate": 8.011056222065397e-05, "loss": 2.8634, "step": 25365 }, { "epoch": 1.99, "learning_rate": 8.010977809142946e-05, "loss": 2.8245, "step": 25366 }, { "epoch": 1.99, "learning_rate": 8.010899396220497e-05, "loss": 2.8232, "step": 25367 }, { "epoch": 1.99, "learning_rate": 8.010820983298047e-05, "loss": 2.8334, "step": 25368 }, { "epoch": 1.99, "learning_rate": 8.010742570375598e-05, "loss": 2.9796, "step": 25369 }, { "epoch": 1.99, "learning_rate": 8.010664157453148e-05, "loss": 3.0186, "step": 25370 }, { "epoch": 1.99, "learning_rate": 8.010585744530698e-05, "loss": 3.0104, "step": 25371 }, { "epoch": 1.99, "learning_rate": 8.010507331608249e-05, "loss": 3.1338, "step": 25372 }, { "epoch": 1.99, "learning_rate": 8.010428918685799e-05, "loss": 2.8516, "step": 25373 }, { "epoch": 1.99, "learning_rate": 8.01035050576335e-05, "loss": 3.2456, "step": 25374 }, { "epoch": 1.99, "learning_rate": 8.0102720928409e-05, "loss": 2.6834, "step": 25375 }, { "epoch": 1.99, "learning_rate": 8.01019367991845e-05, "loss": 2.7168, "step": 25376 }, { "epoch": 1.99, "learning_rate": 8.010115266996e-05, "loss": 2.8946, "step": 25377 }, { "epoch": 1.99, "learning_rate": 8.010036854073551e-05, "loss": 3.284, "step": 25378 }, { "epoch": 1.99, "learning_rate": 8.009958441151103e-05, "loss": 2.7004, "step": 25379 }, { "epoch": 1.99, "learning_rate": 8.009880028228653e-05, "loss": 3.0824, "step": 25380 }, { "epoch": 1.99, "learning_rate": 8.009801615306203e-05, "loss": 2.8707, "step": 25381 }, { "epoch": 1.99, "learning_rate": 8.009723202383754e-05, "loss": 2.926, "step": 25382 }, { "epoch": 1.99, "learning_rate": 8.009644789461304e-05, "loss": 3.1945, "step": 25383 }, { "epoch": 1.99, "learning_rate": 8.009566376538854e-05, "loss": 2.7536, "step": 25384 }, { "epoch": 1.99, "learning_rate": 8.009487963616405e-05, "loss": 3.0533, "step": 25385 }, { "epoch": 1.99, "learning_rate": 8.009409550693955e-05, "loss": 2.9291, "step": 25386 }, { "epoch": 1.99, "learning_rate": 8.009331137771506e-05, "loss": 3.0424, "step": 25387 }, { "epoch": 1.99, "learning_rate": 8.009252724849056e-05, "loss": 2.6656, "step": 25388 }, { "epoch": 1.99, "learning_rate": 8.009174311926606e-05, "loss": 3.117, "step": 25389 }, { "epoch": 1.99, "learning_rate": 8.009095899004157e-05, "loss": 2.8029, "step": 25390 }, { "epoch": 1.99, "learning_rate": 8.009017486081707e-05, "loss": 3.0445, "step": 25391 }, { "epoch": 1.99, "learning_rate": 8.008939073159257e-05, "loss": 2.5326, "step": 25392 }, { "epoch": 1.99, "learning_rate": 8.008860660236808e-05, "loss": 2.7256, "step": 25393 }, { "epoch": 1.99, "learning_rate": 8.008782247314358e-05, "loss": 3.0182, "step": 25394 }, { "epoch": 1.99, "learning_rate": 8.008703834391908e-05, "loss": 2.8516, "step": 25395 }, { "epoch": 1.99, "learning_rate": 8.008625421469459e-05, "loss": 2.9863, "step": 25396 }, { "epoch": 1.99, "learning_rate": 8.008547008547009e-05, "loss": 2.8887, "step": 25397 }, { "epoch": 1.99, "learning_rate": 8.00846859562456e-05, "loss": 2.8386, "step": 25398 }, { "epoch": 1.99, "learning_rate": 8.00839018270211e-05, "loss": 2.8628, "step": 25399 }, { "epoch": 1.99, "learning_rate": 8.00831176977966e-05, "loss": 2.7496, "step": 25400 }, { "epoch": 1.99, "learning_rate": 8.008233356857211e-05, "loss": 2.8297, "step": 25401 }, { "epoch": 1.99, "learning_rate": 8.00815494393476e-05, "loss": 2.8187, "step": 25402 }, { "epoch": 1.99, "learning_rate": 8.00807653101231e-05, "loss": 2.6464, "step": 25403 }, { "epoch": 1.99, "learning_rate": 8.007998118089862e-05, "loss": 3.2948, "step": 25404 }, { "epoch": 1.99, "learning_rate": 8.007919705167412e-05, "loss": 3.0244, "step": 25405 }, { "epoch": 1.99, "learning_rate": 8.007841292244963e-05, "loss": 2.9433, "step": 25406 }, { "epoch": 1.99, "learning_rate": 8.007762879322513e-05, "loss": 3.0697, "step": 25407 }, { "epoch": 1.99, "learning_rate": 8.007684466400063e-05, "loss": 2.9909, "step": 25408 }, { "epoch": 1.99, "learning_rate": 8.007606053477614e-05, "loss": 2.7986, "step": 25409 }, { "epoch": 1.99, "learning_rate": 8.007527640555164e-05, "loss": 2.84, "step": 25410 }, { "epoch": 1.99, "learning_rate": 8.007449227632714e-05, "loss": 3.078, "step": 25411 }, { "epoch": 1.99, "learning_rate": 8.007370814710265e-05, "loss": 2.836, "step": 25412 }, { "epoch": 1.99, "learning_rate": 8.007292401787815e-05, "loss": 2.8499, "step": 25413 }, { "epoch": 1.99, "learning_rate": 8.007213988865365e-05, "loss": 2.9511, "step": 25414 }, { "epoch": 1.99, "learning_rate": 8.007135575942916e-05, "loss": 2.8162, "step": 25415 }, { "epoch": 1.99, "learning_rate": 8.007057163020466e-05, "loss": 2.9343, "step": 25416 }, { "epoch": 1.99, "learning_rate": 8.006978750098017e-05, "loss": 2.7754, "step": 25417 }, { "epoch": 1.99, "learning_rate": 8.006900337175567e-05, "loss": 2.9104, "step": 25418 }, { "epoch": 1.99, "learning_rate": 8.006821924253117e-05, "loss": 3.1194, "step": 25419 }, { "epoch": 1.99, "learning_rate": 8.006743511330668e-05, "loss": 2.9198, "step": 25420 }, { "epoch": 1.99, "learning_rate": 8.006665098408218e-05, "loss": 2.7222, "step": 25421 }, { "epoch": 1.99, "learning_rate": 8.006586685485768e-05, "loss": 3.1176, "step": 25422 }, { "epoch": 1.99, "learning_rate": 8.006508272563319e-05, "loss": 2.7608, "step": 25423 }, { "epoch": 1.99, "learning_rate": 8.006429859640869e-05, "loss": 2.8419, "step": 25424 }, { "epoch": 1.99, "learning_rate": 8.00635144671842e-05, "loss": 3.0074, "step": 25425 }, { "epoch": 1.99, "learning_rate": 8.00627303379597e-05, "loss": 3.0405, "step": 25426 }, { "epoch": 1.99, "learning_rate": 8.00619462087352e-05, "loss": 2.9338, "step": 25427 }, { "epoch": 1.99, "learning_rate": 8.00611620795107e-05, "loss": 2.8409, "step": 25428 }, { "epoch": 1.99, "learning_rate": 8.006037795028621e-05, "loss": 2.8999, "step": 25429 }, { "epoch": 1.99, "learning_rate": 8.005959382106171e-05, "loss": 2.9456, "step": 25430 }, { "epoch": 1.99, "learning_rate": 8.005880969183722e-05, "loss": 2.5947, "step": 25431 }, { "epoch": 1.99, "learning_rate": 8.005802556261272e-05, "loss": 3.1474, "step": 25432 }, { "epoch": 1.99, "learning_rate": 8.005724143338822e-05, "loss": 2.8708, "step": 25433 }, { "epoch": 1.99, "learning_rate": 8.005645730416373e-05, "loss": 2.9189, "step": 25434 }, { "epoch": 1.99, "learning_rate": 8.005567317493923e-05, "loss": 2.9655, "step": 25435 }, { "epoch": 1.99, "learning_rate": 8.005488904571475e-05, "loss": 3.4451, "step": 25436 }, { "epoch": 1.99, "learning_rate": 8.005410491649025e-05, "loss": 2.9426, "step": 25437 }, { "epoch": 1.99, "learning_rate": 8.005332078726574e-05, "loss": 2.9845, "step": 25438 }, { "epoch": 1.99, "learning_rate": 8.005253665804125e-05, "loss": 3.0367, "step": 25439 }, { "epoch": 1.99, "learning_rate": 8.005175252881675e-05, "loss": 2.7208, "step": 25440 }, { "epoch": 1.99, "learning_rate": 8.005096839959225e-05, "loss": 2.7686, "step": 25441 }, { "epoch": 1.99, "learning_rate": 8.005018427036776e-05, "loss": 2.9372, "step": 25442 }, { "epoch": 2.0, "learning_rate": 8.004940014114326e-05, "loss": 2.5697, "step": 25443 }, { "epoch": 2.0, "learning_rate": 8.004861601191876e-05, "loss": 2.7429, "step": 25444 }, { "epoch": 2.0, "learning_rate": 8.004783188269427e-05, "loss": 2.8438, "step": 25445 }, { "epoch": 2.0, "learning_rate": 8.004704775346977e-05, "loss": 2.9761, "step": 25446 }, { "epoch": 2.0, "learning_rate": 8.004626362424528e-05, "loss": 2.9059, "step": 25447 }, { "epoch": 2.0, "learning_rate": 8.004547949502078e-05, "loss": 2.9453, "step": 25448 }, { "epoch": 2.0, "learning_rate": 8.004469536579628e-05, "loss": 2.8868, "step": 25449 }, { "epoch": 2.0, "learning_rate": 8.004391123657179e-05, "loss": 2.9654, "step": 25450 }, { "epoch": 2.0, "learning_rate": 8.004312710734729e-05, "loss": 2.944, "step": 25451 }, { "epoch": 2.0, "learning_rate": 8.00423429781228e-05, "loss": 3.0131, "step": 25452 }, { "epoch": 2.0, "learning_rate": 8.00415588488983e-05, "loss": 2.7068, "step": 25453 }, { "epoch": 2.0, "learning_rate": 8.00407747196738e-05, "loss": 2.6775, "step": 25454 }, { "epoch": 2.0, "learning_rate": 8.00399905904493e-05, "loss": 2.8377, "step": 25455 }, { "epoch": 2.0, "learning_rate": 8.003920646122481e-05, "loss": 2.7942, "step": 25456 }, { "epoch": 2.0, "learning_rate": 8.003842233200031e-05, "loss": 2.9689, "step": 25457 }, { "epoch": 2.0, "learning_rate": 8.003763820277582e-05, "loss": 2.5718, "step": 25458 }, { "epoch": 2.0, "learning_rate": 8.003685407355132e-05, "loss": 2.8635, "step": 25459 }, { "epoch": 2.0, "learning_rate": 8.003606994432682e-05, "loss": 2.6215, "step": 25460 }, { "epoch": 2.0, "learning_rate": 8.003528581510234e-05, "loss": 3.0539, "step": 25461 }, { "epoch": 2.0, "learning_rate": 8.003450168587784e-05, "loss": 2.9386, "step": 25462 }, { "epoch": 2.0, "learning_rate": 8.003371755665335e-05, "loss": 2.805, "step": 25463 }, { "epoch": 2.0, "learning_rate": 8.003293342742885e-05, "loss": 2.9521, "step": 25464 }, { "epoch": 2.0, "learning_rate": 8.003214929820436e-05, "loss": 2.8625, "step": 25465 }, { "epoch": 2.0, "learning_rate": 8.003136516897986e-05, "loss": 3.0653, "step": 25466 }, { "epoch": 2.0, "learning_rate": 8.003058103975536e-05, "loss": 3.0772, "step": 25467 }, { "epoch": 2.0, "learning_rate": 8.002979691053087e-05, "loss": 2.7998, "step": 25468 }, { "epoch": 2.0, "learning_rate": 8.002901278130637e-05, "loss": 3.1262, "step": 25469 }, { "epoch": 2.0, "learning_rate": 8.002822865208187e-05, "loss": 2.809, "step": 25470 }, { "epoch": 2.0, "learning_rate": 8.002744452285738e-05, "loss": 3.0227, "step": 25471 }, { "epoch": 2.0, "learning_rate": 8.002666039363288e-05, "loss": 2.9149, "step": 25472 }, { "epoch": 2.0, "learning_rate": 8.002587626440839e-05, "loss": 3.1147, "step": 25473 }, { "epoch": 2.0, "learning_rate": 8.002509213518388e-05, "loss": 3.0431, "step": 25474 }, { "epoch": 2.0, "learning_rate": 8.002430800595938e-05, "loss": 2.9897, "step": 25475 }, { "epoch": 2.0, "learning_rate": 8.002352387673488e-05, "loss": 3.0586, "step": 25476 }, { "epoch": 2.0, "learning_rate": 8.002273974751039e-05, "loss": 2.6487, "step": 25477 }, { "epoch": 2.0, "learning_rate": 8.002195561828589e-05, "loss": 3.1261, "step": 25478 }, { "epoch": 2.0, "learning_rate": 8.00211714890614e-05, "loss": 2.8096, "step": 25479 }, { "epoch": 2.0, "learning_rate": 8.00203873598369e-05, "loss": 2.9651, "step": 25480 }, { "epoch": 2.0, "learning_rate": 8.00196032306124e-05, "loss": 2.9439, "step": 25481 }, { "epoch": 2.0, "learning_rate": 8.00188191013879e-05, "loss": 2.9721, "step": 25482 }, { "epoch": 2.0, "learning_rate": 8.001803497216341e-05, "loss": 2.7091, "step": 25483 }, { "epoch": 2.0, "learning_rate": 8.001725084293891e-05, "loss": 2.7273, "step": 25484 }, { "epoch": 2.0, "learning_rate": 8.001646671371442e-05, "loss": 2.7729, "step": 25485 }, { "epoch": 2.0, "learning_rate": 8.001568258448992e-05, "loss": 2.7474, "step": 25486 }, { "epoch": 2.0, "learning_rate": 8.001489845526544e-05, "loss": 2.7119, "step": 25487 }, { "epoch": 2.0, "learning_rate": 8.001411432604094e-05, "loss": 2.982, "step": 25488 }, { "epoch": 2.0, "learning_rate": 8.001333019681644e-05, "loss": 2.9098, "step": 25489 }, { "epoch": 2.0, "learning_rate": 8.001254606759195e-05, "loss": 2.6813, "step": 25490 }, { "epoch": 2.0, "learning_rate": 8.001176193836745e-05, "loss": 2.9231, "step": 25491 }, { "epoch": 2.0, "learning_rate": 8.001097780914296e-05, "loss": 2.8204, "step": 25492 }, { "epoch": 2.0, "learning_rate": 8.001019367991846e-05, "loss": 2.9524, "step": 25493 }, { "epoch": 2.0, "learning_rate": 8.000940955069396e-05, "loss": 3.0809, "step": 25494 }, { "epoch": 2.0, "learning_rate": 8.000862542146947e-05, "loss": 2.7407, "step": 25495 }, { "epoch": 2.0, "learning_rate": 8.000784129224497e-05, "loss": 2.9782, "step": 25496 }, { "epoch": 2.0, "learning_rate": 8.000705716302047e-05, "loss": 2.9742, "step": 25497 }, { "epoch": 2.0, "learning_rate": 8.000627303379598e-05, "loss": 2.9806, "step": 25498 }, { "epoch": 2.0, "learning_rate": 8.000548890457148e-05, "loss": 3.0517, "step": 25499 }, { "epoch": 2.0, "learning_rate": 8.000470477534698e-05, "loss": 2.9957, "step": 25500 }, { "epoch": 2.0, "learning_rate": 8.000392064612249e-05, "loss": 2.8181, "step": 25501 }, { "epoch": 2.0, "learning_rate": 8.000313651689799e-05, "loss": 3.0227, "step": 25502 }, { "epoch": 2.0, "learning_rate": 8.00023523876735e-05, "loss": 2.964, "step": 25503 }, { "epoch": 2.0, "learning_rate": 8.0001568258449e-05, "loss": 2.9234, "step": 25504 }, { "epoch": 2.0, "learning_rate": 8.00007841292245e-05, "loss": 3.0344, "step": 25505 }, { "epoch": 2.0, "learning_rate": 8e-05, "loss": 2.9236, "step": 25506 }, { "epoch": 2.0, "learning_rate": 7.999921587077551e-05, "loss": 3.0585, "step": 25507 }, { "epoch": 2.0, "learning_rate": 7.999843174155101e-05, "loss": 3.0091, "step": 25508 }, { "epoch": 2.0, "learning_rate": 7.999764761232652e-05, "loss": 2.7176, "step": 25509 }, { "epoch": 2.0, "learning_rate": 7.999686348310201e-05, "loss": 2.8252, "step": 25510 }, { "epoch": 2.0, "learning_rate": 7.999607935387751e-05, "loss": 2.7456, "step": 25511 }, { "epoch": 2.0, "learning_rate": 7.999529522465303e-05, "loss": 2.7065, "step": 25512 }, { "epoch": 2.0, "learning_rate": 7.999451109542853e-05, "loss": 2.8925, "step": 25513 }, { "epoch": 2.0, "learning_rate": 7.999372696620404e-05, "loss": 2.8782, "step": 25514 }, { "epoch": 2.0, "learning_rate": 7.999294283697954e-05, "loss": 2.7381, "step": 25515 }, { "epoch": 2.0, "learning_rate": 7.999215870775504e-05, "loss": 2.9839, "step": 25516 }, { "epoch": 2.0, "learning_rate": 7.999137457853055e-05, "loss": 2.7915, "step": 25517 }, { "epoch": 2.0, "learning_rate": 7.999059044930605e-05, "loss": 2.6514, "step": 25518 }, { "epoch": 2.0, "learning_rate": 7.998980632008155e-05, "loss": 3.1297, "step": 25519 }, { "epoch": 2.0, "learning_rate": 7.998902219085706e-05, "loss": 2.7905, "step": 25520 }, { "epoch": 2.0, "learning_rate": 7.998823806163256e-05, "loss": 2.9473, "step": 25521 }, { "epoch": 2.0, "learning_rate": 7.998745393240807e-05, "loss": 2.8152, "step": 25522 }, { "epoch": 2.0, "learning_rate": 7.998666980318357e-05, "loss": 2.5353, "step": 25523 }, { "epoch": 2.0, "learning_rate": 7.998588567395907e-05, "loss": 2.7963, "step": 25524 }, { "epoch": 2.0, "learning_rate": 7.998510154473458e-05, "loss": 2.8182, "step": 25525 }, { "epoch": 2.0, "learning_rate": 7.998431741551008e-05, "loss": 2.8897, "step": 25526 }, { "epoch": 2.0, "learning_rate": 7.998353328628558e-05, "loss": 2.9776, "step": 25527 }, { "epoch": 2.0, "learning_rate": 7.998274915706109e-05, "loss": 2.8132, "step": 25528 }, { "epoch": 2.0, "learning_rate": 7.998196502783659e-05, "loss": 2.888, "step": 25529 }, { "epoch": 2.0, "learning_rate": 7.99811808986121e-05, "loss": 3.2829, "step": 25530 }, { "epoch": 2.0, "learning_rate": 7.99803967693876e-05, "loss": 2.6748, "step": 25531 }, { "epoch": 2.0, "learning_rate": 7.99796126401631e-05, "loss": 2.7102, "step": 25532 }, { "epoch": 2.0, "learning_rate": 7.99788285109386e-05, "loss": 2.8316, "step": 25533 }, { "epoch": 2.0, "learning_rate": 7.997804438171411e-05, "loss": 2.7286, "step": 25534 }, { "epoch": 2.0, "learning_rate": 7.997726025248961e-05, "loss": 2.6875, "step": 25535 }, { "epoch": 2.0, "learning_rate": 7.997647612326512e-05, "loss": 2.7441, "step": 25536 }, { "epoch": 2.0, "learning_rate": 7.997569199404062e-05, "loss": 2.6109, "step": 25537 }, { "epoch": 2.0, "learning_rate": 7.997490786481612e-05, "loss": 2.747, "step": 25538 }, { "epoch": 2.0, "learning_rate": 7.997412373559163e-05, "loss": 2.917, "step": 25539 }, { "epoch": 2.0, "learning_rate": 7.997333960636713e-05, "loss": 2.6308, "step": 25540 }, { "epoch": 2.0, "learning_rate": 7.997255547714264e-05, "loss": 3.0809, "step": 25541 }, { "epoch": 2.0, "learning_rate": 7.997177134791814e-05, "loss": 2.6189, "step": 25542 }, { "epoch": 2.0, "learning_rate": 7.997098721869364e-05, "loss": 2.7683, "step": 25543 }, { "epoch": 2.0, "learning_rate": 7.997020308946916e-05, "loss": 2.7408, "step": 25544 }, { "epoch": 2.0, "learning_rate": 7.996941896024466e-05, "loss": 2.7934, "step": 25545 }, { "epoch": 2.0, "learning_rate": 7.996863483102015e-05, "loss": 2.7285, "step": 25546 }, { "epoch": 2.0, "learning_rate": 7.996785070179566e-05, "loss": 2.7607, "step": 25547 }, { "epoch": 2.0, "learning_rate": 7.996706657257116e-05, "loss": 2.6613, "step": 25548 }, { "epoch": 2.0, "learning_rate": 7.996628244334666e-05, "loss": 2.7206, "step": 25549 }, { "epoch": 2.0, "learning_rate": 7.996549831412217e-05, "loss": 2.6705, "step": 25550 }, { "epoch": 2.0, "learning_rate": 7.996471418489767e-05, "loss": 3.0829, "step": 25551 }, { "epoch": 2.0, "learning_rate": 7.996393005567318e-05, "loss": 2.6929, "step": 25552 }, { "epoch": 2.0, "learning_rate": 7.996314592644868e-05, "loss": 2.7678, "step": 25553 }, { "epoch": 2.0, "learning_rate": 7.996236179722418e-05, "loss": 2.8672, "step": 25554 }, { "epoch": 2.0, "learning_rate": 7.996157766799969e-05, "loss": 2.6213, "step": 25555 }, { "epoch": 2.0, "learning_rate": 7.996079353877519e-05, "loss": 2.5632, "step": 25556 }, { "epoch": 2.0, "learning_rate": 7.99600094095507e-05, "loss": 2.9354, "step": 25557 }, { "epoch": 2.0, "learning_rate": 7.99592252803262e-05, "loss": 2.7594, "step": 25558 }, { "epoch": 2.0, "learning_rate": 7.99584411511017e-05, "loss": 2.9844, "step": 25559 }, { "epoch": 2.0, "learning_rate": 7.99576570218772e-05, "loss": 2.9039, "step": 25560 }, { "epoch": 2.0, "learning_rate": 7.995687289265271e-05, "loss": 2.7811, "step": 25561 }, { "epoch": 2.0, "learning_rate": 7.995608876342821e-05, "loss": 2.7445, "step": 25562 }, { "epoch": 2.0, "learning_rate": 7.995530463420372e-05, "loss": 2.7874, "step": 25563 }, { "epoch": 2.0, "learning_rate": 7.995452050497922e-05, "loss": 2.6953, "step": 25564 }, { "epoch": 2.0, "learning_rate": 7.995373637575472e-05, "loss": 2.8551, "step": 25565 }, { "epoch": 2.0, "learning_rate": 7.995295224653023e-05, "loss": 2.8436, "step": 25566 }, { "epoch": 2.0, "learning_rate": 7.995216811730573e-05, "loss": 2.6672, "step": 25567 }, { "epoch": 2.0, "learning_rate": 7.995138398808123e-05, "loss": 2.6792, "step": 25568 }, { "epoch": 2.0, "learning_rate": 7.995059985885675e-05, "loss": 2.7473, "step": 25569 }, { "epoch": 2.01, "learning_rate": 7.994981572963226e-05, "loss": 2.753, "step": 25570 }, { "epoch": 2.01, "learning_rate": 7.994903160040776e-05, "loss": 2.8797, "step": 25571 }, { "epoch": 2.01, "learning_rate": 7.994824747118326e-05, "loss": 2.6569, "step": 25572 }, { "epoch": 2.01, "learning_rate": 7.994746334195877e-05, "loss": 2.9399, "step": 25573 }, { "epoch": 2.01, "learning_rate": 7.994667921273427e-05, "loss": 2.6755, "step": 25574 }, { "epoch": 2.01, "learning_rate": 7.994589508350977e-05, "loss": 2.7575, "step": 25575 }, { "epoch": 2.01, "learning_rate": 7.994511095428528e-05, "loss": 2.7842, "step": 25576 }, { "epoch": 2.01, "learning_rate": 7.994432682506078e-05, "loss": 2.7708, "step": 25577 }, { "epoch": 2.01, "learning_rate": 7.994354269583628e-05, "loss": 2.7872, "step": 25578 }, { "epoch": 2.01, "learning_rate": 7.994275856661179e-05, "loss": 2.7898, "step": 25579 }, { "epoch": 2.01, "learning_rate": 7.994197443738729e-05, "loss": 2.6931, "step": 25580 }, { "epoch": 2.01, "learning_rate": 7.99411903081628e-05, "loss": 3.0043, "step": 25581 }, { "epoch": 2.01, "learning_rate": 7.994040617893829e-05, "loss": 2.6003, "step": 25582 }, { "epoch": 2.01, "learning_rate": 7.993962204971379e-05, "loss": 2.934, "step": 25583 }, { "epoch": 2.01, "learning_rate": 7.99388379204893e-05, "loss": 2.8045, "step": 25584 }, { "epoch": 2.01, "learning_rate": 7.99380537912648e-05, "loss": 2.9317, "step": 25585 }, { "epoch": 2.01, "learning_rate": 7.99372696620403e-05, "loss": 2.9097, "step": 25586 }, { "epoch": 2.01, "learning_rate": 7.99364855328158e-05, "loss": 2.7612, "step": 25587 }, { "epoch": 2.01, "learning_rate": 7.993570140359131e-05, "loss": 2.8231, "step": 25588 }, { "epoch": 2.01, "learning_rate": 7.993491727436681e-05, "loss": 3.0322, "step": 25589 }, { "epoch": 2.01, "learning_rate": 7.993413314514232e-05, "loss": 2.8039, "step": 25590 }, { "epoch": 2.01, "learning_rate": 7.993334901591782e-05, "loss": 3.0124, "step": 25591 }, { "epoch": 2.01, "learning_rate": 7.993256488669332e-05, "loss": 2.6043, "step": 25592 }, { "epoch": 2.01, "learning_rate": 7.993178075746883e-05, "loss": 2.8785, "step": 25593 }, { "epoch": 2.01, "learning_rate": 7.993099662824433e-05, "loss": 2.5705, "step": 25594 }, { "epoch": 2.01, "learning_rate": 7.993021249901985e-05, "loss": 2.9227, "step": 25595 }, { "epoch": 2.01, "learning_rate": 7.992942836979535e-05, "loss": 2.8311, "step": 25596 }, { "epoch": 2.01, "learning_rate": 7.992864424057085e-05, "loss": 2.9273, "step": 25597 }, { "epoch": 2.01, "learning_rate": 7.992786011134636e-05, "loss": 2.8962, "step": 25598 }, { "epoch": 2.01, "learning_rate": 7.992707598212186e-05, "loss": 2.7312, "step": 25599 }, { "epoch": 2.01, "learning_rate": 7.992629185289737e-05, "loss": 2.6343, "step": 25600 }, { "epoch": 2.01, "learning_rate": 7.992550772367287e-05, "loss": 2.9749, "step": 25601 }, { "epoch": 2.01, "learning_rate": 7.992472359444837e-05, "loss": 2.7533, "step": 25602 }, { "epoch": 2.01, "learning_rate": 7.992393946522388e-05, "loss": 2.9729, "step": 25603 }, { "epoch": 2.01, "learning_rate": 7.992315533599938e-05, "loss": 2.6694, "step": 25604 }, { "epoch": 2.01, "learning_rate": 7.992237120677488e-05, "loss": 2.7741, "step": 25605 }, { "epoch": 2.01, "learning_rate": 7.992158707755039e-05, "loss": 2.8295, "step": 25606 }, { "epoch": 2.01, "learning_rate": 7.992080294832589e-05, "loss": 3.017, "step": 25607 }, { "epoch": 2.01, "learning_rate": 7.99200188191014e-05, "loss": 2.7576, "step": 25608 }, { "epoch": 2.01, "learning_rate": 7.99192346898769e-05, "loss": 2.8312, "step": 25609 }, { "epoch": 2.01, "learning_rate": 7.99184505606524e-05, "loss": 2.8487, "step": 25610 }, { "epoch": 2.01, "learning_rate": 7.99176664314279e-05, "loss": 2.7097, "step": 25611 }, { "epoch": 2.01, "learning_rate": 7.991688230220341e-05, "loss": 2.8023, "step": 25612 }, { "epoch": 2.01, "learning_rate": 7.991609817297891e-05, "loss": 2.5493, "step": 25613 }, { "epoch": 2.01, "learning_rate": 7.991531404375442e-05, "loss": 2.7164, "step": 25614 }, { "epoch": 2.01, "learning_rate": 7.991452991452992e-05, "loss": 2.6242, "step": 25615 }, { "epoch": 2.01, "learning_rate": 7.991374578530542e-05, "loss": 2.5861, "step": 25616 }, { "epoch": 2.01, "learning_rate": 7.991296165608093e-05, "loss": 2.9132, "step": 25617 }, { "epoch": 2.01, "learning_rate": 7.991217752685642e-05, "loss": 2.7582, "step": 25618 }, { "epoch": 2.01, "learning_rate": 7.991139339763192e-05, "loss": 2.8436, "step": 25619 }, { "epoch": 2.01, "learning_rate": 7.991060926840743e-05, "loss": 2.7133, "step": 25620 }, { "epoch": 2.01, "learning_rate": 7.990982513918294e-05, "loss": 2.7836, "step": 25621 }, { "epoch": 2.01, "learning_rate": 7.990904100995845e-05, "loss": 2.7045, "step": 25622 }, { "epoch": 2.01, "learning_rate": 7.990825688073395e-05, "loss": 2.7944, "step": 25623 }, { "epoch": 2.01, "learning_rate": 7.990747275150945e-05, "loss": 2.7963, "step": 25624 }, { "epoch": 2.01, "learning_rate": 7.990668862228496e-05, "loss": 3.0123, "step": 25625 }, { "epoch": 2.01, "learning_rate": 7.990590449306046e-05, "loss": 2.8067, "step": 25626 }, { "epoch": 2.01, "learning_rate": 7.990512036383597e-05, "loss": 2.7829, "step": 25627 }, { "epoch": 2.01, "learning_rate": 7.990433623461147e-05, "loss": 2.7594, "step": 25628 }, { "epoch": 2.01, "learning_rate": 7.990355210538697e-05, "loss": 2.9343, "step": 25629 }, { "epoch": 2.01, "learning_rate": 7.990276797616248e-05, "loss": 3.0223, "step": 25630 }, { "epoch": 2.01, "learning_rate": 7.990198384693798e-05, "loss": 2.7634, "step": 25631 }, { "epoch": 2.01, "learning_rate": 7.990119971771348e-05, "loss": 2.8712, "step": 25632 }, { "epoch": 2.01, "learning_rate": 7.990041558848899e-05, "loss": 3.0594, "step": 25633 }, { "epoch": 2.01, "learning_rate": 7.989963145926449e-05, "loss": 2.9031, "step": 25634 }, { "epoch": 2.01, "learning_rate": 7.989884733004e-05, "loss": 2.7293, "step": 25635 }, { "epoch": 2.01, "learning_rate": 7.98980632008155e-05, "loss": 2.5373, "step": 25636 }, { "epoch": 2.01, "learning_rate": 7.9897279071591e-05, "loss": 2.9444, "step": 25637 }, { "epoch": 2.01, "learning_rate": 7.98964949423665e-05, "loss": 2.838, "step": 25638 }, { "epoch": 2.01, "learning_rate": 7.989571081314201e-05, "loss": 2.9047, "step": 25639 }, { "epoch": 2.01, "learning_rate": 7.989492668391751e-05, "loss": 3.1077, "step": 25640 }, { "epoch": 2.01, "learning_rate": 7.989414255469302e-05, "loss": 2.8899, "step": 25641 }, { "epoch": 2.01, "learning_rate": 7.989335842546852e-05, "loss": 2.7926, "step": 25642 }, { "epoch": 2.01, "learning_rate": 7.989257429624402e-05, "loss": 2.5816, "step": 25643 }, { "epoch": 2.01, "learning_rate": 7.989179016701953e-05, "loss": 2.5751, "step": 25644 }, { "epoch": 2.01, "learning_rate": 7.989100603779503e-05, "loss": 2.8104, "step": 25645 }, { "epoch": 2.01, "learning_rate": 7.989022190857053e-05, "loss": 2.9601, "step": 25646 }, { "epoch": 2.01, "learning_rate": 7.988943777934604e-05, "loss": 2.7113, "step": 25647 }, { "epoch": 2.01, "learning_rate": 7.988865365012154e-05, "loss": 3.0318, "step": 25648 }, { "epoch": 2.01, "learning_rate": 7.988786952089705e-05, "loss": 2.6667, "step": 25649 }, { "epoch": 2.01, "learning_rate": 7.988708539167255e-05, "loss": 2.8314, "step": 25650 }, { "epoch": 2.01, "learning_rate": 7.988630126244805e-05, "loss": 2.8381, "step": 25651 }, { "epoch": 2.01, "learning_rate": 7.988551713322357e-05, "loss": 2.718, "step": 25652 }, { "epoch": 2.01, "learning_rate": 7.988473300399907e-05, "loss": 2.8675, "step": 25653 }, { "epoch": 2.01, "learning_rate": 7.988394887477456e-05, "loss": 2.8145, "step": 25654 }, { "epoch": 2.01, "learning_rate": 7.988316474555007e-05, "loss": 2.8536, "step": 25655 }, { "epoch": 2.01, "learning_rate": 7.988238061632557e-05, "loss": 2.7397, "step": 25656 }, { "epoch": 2.01, "learning_rate": 7.988159648710108e-05, "loss": 2.8747, "step": 25657 }, { "epoch": 2.01, "learning_rate": 7.988081235787658e-05, "loss": 2.6545, "step": 25658 }, { "epoch": 2.01, "learning_rate": 7.988002822865208e-05, "loss": 2.7132, "step": 25659 }, { "epoch": 2.01, "learning_rate": 7.987924409942759e-05, "loss": 2.778, "step": 25660 }, { "epoch": 2.01, "learning_rate": 7.987845997020309e-05, "loss": 2.8469, "step": 25661 }, { "epoch": 2.01, "learning_rate": 7.98776758409786e-05, "loss": 2.5997, "step": 25662 }, { "epoch": 2.01, "learning_rate": 7.98768917117541e-05, "loss": 2.9589, "step": 25663 }, { "epoch": 2.01, "learning_rate": 7.98761075825296e-05, "loss": 2.8091, "step": 25664 }, { "epoch": 2.01, "learning_rate": 7.98753234533051e-05, "loss": 2.6905, "step": 25665 }, { "epoch": 2.01, "learning_rate": 7.987453932408061e-05, "loss": 2.7004, "step": 25666 }, { "epoch": 2.01, "learning_rate": 7.987375519485611e-05, "loss": 2.6651, "step": 25667 }, { "epoch": 2.01, "learning_rate": 7.987297106563162e-05, "loss": 2.6062, "step": 25668 }, { "epoch": 2.01, "learning_rate": 7.987218693640712e-05, "loss": 2.5693, "step": 25669 }, { "epoch": 2.01, "learning_rate": 7.987140280718262e-05, "loss": 2.806, "step": 25670 }, { "epoch": 2.01, "learning_rate": 7.987061867795813e-05, "loss": 2.9165, "step": 25671 }, { "epoch": 2.01, "learning_rate": 7.986983454873363e-05, "loss": 2.6563, "step": 25672 }, { "epoch": 2.01, "learning_rate": 7.986905041950913e-05, "loss": 2.9871, "step": 25673 }, { "epoch": 2.01, "learning_rate": 7.986826629028464e-05, "loss": 2.8075, "step": 25674 }, { "epoch": 2.01, "learning_rate": 7.986748216106014e-05, "loss": 2.7627, "step": 25675 }, { "epoch": 2.01, "learning_rate": 7.986669803183565e-05, "loss": 2.916, "step": 25676 }, { "epoch": 2.01, "learning_rate": 7.986591390261116e-05, "loss": 2.9076, "step": 25677 }, { "epoch": 2.01, "learning_rate": 7.986512977338667e-05, "loss": 2.7942, "step": 25678 }, { "epoch": 2.01, "learning_rate": 7.986434564416217e-05, "loss": 2.7065, "step": 25679 }, { "epoch": 2.01, "learning_rate": 7.986356151493767e-05, "loss": 2.5691, "step": 25680 }, { "epoch": 2.01, "learning_rate": 7.986277738571318e-05, "loss": 2.8755, "step": 25681 }, { "epoch": 2.01, "learning_rate": 7.986199325648868e-05, "loss": 2.8451, "step": 25682 }, { "epoch": 2.01, "learning_rate": 7.986120912726418e-05, "loss": 2.6941, "step": 25683 }, { "epoch": 2.01, "learning_rate": 7.986042499803969e-05, "loss": 2.7112, "step": 25684 }, { "epoch": 2.01, "learning_rate": 7.985964086881519e-05, "loss": 2.7362, "step": 25685 }, { "epoch": 2.01, "learning_rate": 7.98588567395907e-05, "loss": 2.716, "step": 25686 }, { "epoch": 2.01, "learning_rate": 7.98580726103662e-05, "loss": 2.7498, "step": 25687 }, { "epoch": 2.01, "learning_rate": 7.98572884811417e-05, "loss": 2.842, "step": 25688 }, { "epoch": 2.01, "learning_rate": 7.98565043519172e-05, "loss": 2.5565, "step": 25689 }, { "epoch": 2.01, "learning_rate": 7.98557202226927e-05, "loss": 2.8431, "step": 25690 }, { "epoch": 2.01, "learning_rate": 7.98549360934682e-05, "loss": 2.8208, "step": 25691 }, { "epoch": 2.01, "learning_rate": 7.98541519642437e-05, "loss": 2.8181, "step": 25692 }, { "epoch": 2.01, "learning_rate": 7.985336783501921e-05, "loss": 2.7968, "step": 25693 }, { "epoch": 2.01, "learning_rate": 7.985258370579471e-05, "loss": 2.8434, "step": 25694 }, { "epoch": 2.01, "learning_rate": 7.985179957657022e-05, "loss": 2.8316, "step": 25695 }, { "epoch": 2.01, "learning_rate": 7.985101544734572e-05, "loss": 2.776, "step": 25696 }, { "epoch": 2.01, "learning_rate": 7.985023131812122e-05, "loss": 2.9285, "step": 25697 }, { "epoch": 2.02, "learning_rate": 7.984944718889673e-05, "loss": 2.7736, "step": 25698 }, { "epoch": 2.02, "learning_rate": 7.984866305967223e-05, "loss": 2.8913, "step": 25699 }, { "epoch": 2.02, "learning_rate": 7.984787893044773e-05, "loss": 3.0218, "step": 25700 }, { "epoch": 2.02, "learning_rate": 7.984709480122324e-05, "loss": 2.788, "step": 25701 }, { "epoch": 2.02, "learning_rate": 7.984631067199874e-05, "loss": 2.8609, "step": 25702 }, { "epoch": 2.02, "learning_rate": 7.984552654277426e-05, "loss": 2.8112, "step": 25703 }, { "epoch": 2.02, "learning_rate": 7.984474241354976e-05, "loss": 2.824, "step": 25704 }, { "epoch": 2.02, "learning_rate": 7.984395828432527e-05, "loss": 2.832, "step": 25705 }, { "epoch": 2.02, "learning_rate": 7.984317415510077e-05, "loss": 2.6991, "step": 25706 }, { "epoch": 2.02, "learning_rate": 7.984239002587627e-05, "loss": 2.6351, "step": 25707 }, { "epoch": 2.02, "learning_rate": 7.984160589665178e-05, "loss": 2.7522, "step": 25708 }, { "epoch": 2.02, "learning_rate": 7.984082176742728e-05, "loss": 2.6883, "step": 25709 }, { "epoch": 2.02, "learning_rate": 7.984003763820278e-05, "loss": 2.9111, "step": 25710 }, { "epoch": 2.02, "learning_rate": 7.983925350897829e-05, "loss": 2.8476, "step": 25711 }, { "epoch": 2.02, "learning_rate": 7.983846937975379e-05, "loss": 2.7062, "step": 25712 }, { "epoch": 2.02, "learning_rate": 7.98376852505293e-05, "loss": 2.8503, "step": 25713 }, { "epoch": 2.02, "learning_rate": 7.98369011213048e-05, "loss": 3.254, "step": 25714 }, { "epoch": 2.02, "learning_rate": 7.98361169920803e-05, "loss": 2.6947, "step": 25715 }, { "epoch": 2.02, "learning_rate": 7.98353328628558e-05, "loss": 2.9821, "step": 25716 }, { "epoch": 2.02, "learning_rate": 7.983454873363131e-05, "loss": 3.0127, "step": 25717 }, { "epoch": 2.02, "learning_rate": 7.983376460440681e-05, "loss": 2.984, "step": 25718 }, { "epoch": 2.02, "learning_rate": 7.983298047518232e-05, "loss": 2.9775, "step": 25719 }, { "epoch": 2.02, "learning_rate": 7.983219634595782e-05, "loss": 3.2425, "step": 25720 }, { "epoch": 2.02, "learning_rate": 7.983141221673332e-05, "loss": 2.9385, "step": 25721 }, { "epoch": 2.02, "learning_rate": 7.983062808750883e-05, "loss": 2.7528, "step": 25722 }, { "epoch": 2.02, "learning_rate": 7.982984395828433e-05, "loss": 2.8653, "step": 25723 }, { "epoch": 2.02, "learning_rate": 7.982905982905984e-05, "loss": 3.1344, "step": 25724 }, { "epoch": 2.02, "learning_rate": 7.982827569983534e-05, "loss": 2.8903, "step": 25725 }, { "epoch": 2.02, "learning_rate": 7.982749157061083e-05, "loss": 2.9355, "step": 25726 }, { "epoch": 2.02, "learning_rate": 7.982670744138633e-05, "loss": 2.7588, "step": 25727 }, { "epoch": 2.02, "learning_rate": 7.982592331216184e-05, "loss": 2.9399, "step": 25728 }, { "epoch": 2.02, "learning_rate": 7.982513918293735e-05, "loss": 2.8684, "step": 25729 }, { "epoch": 2.02, "learning_rate": 7.982435505371286e-05, "loss": 2.9855, "step": 25730 }, { "epoch": 2.02, "learning_rate": 7.982357092448836e-05, "loss": 2.9913, "step": 25731 }, { "epoch": 2.02, "learning_rate": 7.982278679526386e-05, "loss": 2.7499, "step": 25732 }, { "epoch": 2.02, "learning_rate": 7.982200266603937e-05, "loss": 2.7462, "step": 25733 }, { "epoch": 2.02, "learning_rate": 7.982121853681487e-05, "loss": 2.7827, "step": 25734 }, { "epoch": 2.02, "learning_rate": 7.982043440759038e-05, "loss": 2.7925, "step": 25735 }, { "epoch": 2.02, "learning_rate": 7.981965027836588e-05, "loss": 2.7259, "step": 25736 }, { "epoch": 2.02, "learning_rate": 7.981886614914138e-05, "loss": 3.0224, "step": 25737 }, { "epoch": 2.02, "learning_rate": 7.981808201991689e-05, "loss": 2.7457, "step": 25738 }, { "epoch": 2.02, "learning_rate": 7.981729789069239e-05, "loss": 2.6807, "step": 25739 }, { "epoch": 2.02, "learning_rate": 7.98165137614679e-05, "loss": 2.8468, "step": 25740 }, { "epoch": 2.02, "learning_rate": 7.98157296322434e-05, "loss": 2.4521, "step": 25741 }, { "epoch": 2.02, "learning_rate": 7.98149455030189e-05, "loss": 2.9311, "step": 25742 }, { "epoch": 2.02, "learning_rate": 7.98141613737944e-05, "loss": 2.7418, "step": 25743 }, { "epoch": 2.02, "learning_rate": 7.981337724456991e-05, "loss": 2.9611, "step": 25744 }, { "epoch": 2.02, "learning_rate": 7.981259311534541e-05, "loss": 2.677, "step": 25745 }, { "epoch": 2.02, "learning_rate": 7.981180898612092e-05, "loss": 2.8026, "step": 25746 }, { "epoch": 2.02, "learning_rate": 7.981102485689642e-05, "loss": 2.9366, "step": 25747 }, { "epoch": 2.02, "learning_rate": 7.981024072767192e-05, "loss": 2.8188, "step": 25748 }, { "epoch": 2.02, "learning_rate": 7.980945659844743e-05, "loss": 2.7304, "step": 25749 }, { "epoch": 2.02, "learning_rate": 7.980867246922293e-05, "loss": 2.9505, "step": 25750 }, { "epoch": 2.02, "learning_rate": 7.980788833999843e-05, "loss": 2.852, "step": 25751 }, { "epoch": 2.02, "learning_rate": 7.980710421077394e-05, "loss": 2.638, "step": 25752 }, { "epoch": 2.02, "learning_rate": 7.980632008154944e-05, "loss": 2.8312, "step": 25753 }, { "epoch": 2.02, "learning_rate": 7.980553595232495e-05, "loss": 2.8547, "step": 25754 }, { "epoch": 2.02, "learning_rate": 7.980475182310045e-05, "loss": 2.66, "step": 25755 }, { "epoch": 2.02, "learning_rate": 7.980396769387595e-05, "loss": 2.9396, "step": 25756 }, { "epoch": 2.02, "learning_rate": 7.980318356465146e-05, "loss": 2.9044, "step": 25757 }, { "epoch": 2.02, "learning_rate": 7.980239943542696e-05, "loss": 2.8257, "step": 25758 }, { "epoch": 2.02, "learning_rate": 7.980161530620246e-05, "loss": 2.9899, "step": 25759 }, { "epoch": 2.02, "learning_rate": 7.980083117697798e-05, "loss": 2.8312, "step": 25760 }, { "epoch": 2.02, "learning_rate": 7.980004704775349e-05, "loss": 2.6214, "step": 25761 }, { "epoch": 2.02, "learning_rate": 7.979926291852898e-05, "loss": 2.9132, "step": 25762 }, { "epoch": 2.02, "learning_rate": 7.979847878930448e-05, "loss": 2.9173, "step": 25763 }, { "epoch": 2.02, "learning_rate": 7.979769466007998e-05, "loss": 2.8925, "step": 25764 }, { "epoch": 2.02, "learning_rate": 7.979691053085549e-05, "loss": 2.8443, "step": 25765 }, { "epoch": 2.02, "learning_rate": 7.979612640163099e-05, "loss": 2.7735, "step": 25766 }, { "epoch": 2.02, "learning_rate": 7.97953422724065e-05, "loss": 2.7737, "step": 25767 }, { "epoch": 2.02, "learning_rate": 7.9794558143182e-05, "loss": 2.5247, "step": 25768 }, { "epoch": 2.02, "learning_rate": 7.97937740139575e-05, "loss": 2.8548, "step": 25769 }, { "epoch": 2.02, "learning_rate": 7.9792989884733e-05, "loss": 2.8636, "step": 25770 }, { "epoch": 2.02, "learning_rate": 7.979220575550851e-05, "loss": 2.9584, "step": 25771 }, { "epoch": 2.02, "learning_rate": 7.979142162628401e-05, "loss": 2.7862, "step": 25772 }, { "epoch": 2.02, "learning_rate": 7.979063749705952e-05, "loss": 2.7156, "step": 25773 }, { "epoch": 2.02, "learning_rate": 7.978985336783502e-05, "loss": 2.9238, "step": 25774 }, { "epoch": 2.02, "learning_rate": 7.978906923861052e-05, "loss": 3.1229, "step": 25775 }, { "epoch": 2.02, "learning_rate": 7.978828510938603e-05, "loss": 2.7213, "step": 25776 }, { "epoch": 2.02, "learning_rate": 7.978750098016153e-05, "loss": 2.6878, "step": 25777 }, { "epoch": 2.02, "learning_rate": 7.978671685093703e-05, "loss": 2.6045, "step": 25778 }, { "epoch": 2.02, "learning_rate": 7.978593272171254e-05, "loss": 2.9519, "step": 25779 }, { "epoch": 2.02, "learning_rate": 7.978514859248804e-05, "loss": 2.8708, "step": 25780 }, { "epoch": 2.02, "learning_rate": 7.978436446326355e-05, "loss": 2.7116, "step": 25781 }, { "epoch": 2.02, "learning_rate": 7.978358033403905e-05, "loss": 2.6793, "step": 25782 }, { "epoch": 2.02, "learning_rate": 7.978279620481455e-05, "loss": 2.6983, "step": 25783 }, { "epoch": 2.02, "learning_rate": 7.978201207559006e-05, "loss": 2.7299, "step": 25784 }, { "epoch": 2.02, "learning_rate": 7.978122794636557e-05, "loss": 2.6004, "step": 25785 }, { "epoch": 2.02, "learning_rate": 7.978044381714108e-05, "loss": 2.744, "step": 25786 }, { "epoch": 2.02, "learning_rate": 7.977965968791658e-05, "loss": 2.6783, "step": 25787 }, { "epoch": 2.02, "learning_rate": 7.977887555869208e-05, "loss": 2.6566, "step": 25788 }, { "epoch": 2.02, "learning_rate": 7.977809142946759e-05, "loss": 2.6585, "step": 25789 }, { "epoch": 2.02, "learning_rate": 7.977730730024309e-05, "loss": 3.1284, "step": 25790 }, { "epoch": 2.02, "learning_rate": 7.97765231710186e-05, "loss": 2.8792, "step": 25791 }, { "epoch": 2.02, "learning_rate": 7.97757390417941e-05, "loss": 2.8597, "step": 25792 }, { "epoch": 2.02, "learning_rate": 7.97749549125696e-05, "loss": 2.8303, "step": 25793 }, { "epoch": 2.02, "learning_rate": 7.97741707833451e-05, "loss": 2.7011, "step": 25794 }, { "epoch": 2.02, "learning_rate": 7.977338665412061e-05, "loss": 2.755, "step": 25795 }, { "epoch": 2.02, "learning_rate": 7.977260252489611e-05, "loss": 2.7621, "step": 25796 }, { "epoch": 2.02, "learning_rate": 7.977181839567162e-05, "loss": 2.7308, "step": 25797 }, { "epoch": 2.02, "learning_rate": 7.977103426644711e-05, "loss": 2.8795, "step": 25798 }, { "epoch": 2.02, "learning_rate": 7.977025013722261e-05, "loss": 2.7836, "step": 25799 }, { "epoch": 2.02, "learning_rate": 7.976946600799811e-05, "loss": 2.6786, "step": 25800 }, { "epoch": 2.02, "learning_rate": 7.976868187877362e-05, "loss": 2.6949, "step": 25801 }, { "epoch": 2.02, "learning_rate": 7.976789774954912e-05, "loss": 2.807, "step": 25802 }, { "epoch": 2.02, "learning_rate": 7.976711362032463e-05, "loss": 2.7684, "step": 25803 }, { "epoch": 2.02, "learning_rate": 7.976632949110013e-05, "loss": 2.8071, "step": 25804 }, { "epoch": 2.02, "learning_rate": 7.976554536187563e-05, "loss": 2.8112, "step": 25805 }, { "epoch": 2.02, "learning_rate": 7.976476123265114e-05, "loss": 2.964, "step": 25806 }, { "epoch": 2.02, "learning_rate": 7.976397710342664e-05, "loss": 2.8607, "step": 25807 }, { "epoch": 2.02, "learning_rate": 7.976319297420214e-05, "loss": 2.819, "step": 25808 }, { "epoch": 2.02, "learning_rate": 7.976240884497765e-05, "loss": 2.7358, "step": 25809 }, { "epoch": 2.02, "learning_rate": 7.976162471575315e-05, "loss": 2.8234, "step": 25810 }, { "epoch": 2.02, "learning_rate": 7.976084058652867e-05, "loss": 2.8161, "step": 25811 }, { "epoch": 2.02, "learning_rate": 7.976005645730417e-05, "loss": 3.0742, "step": 25812 }, { "epoch": 2.02, "learning_rate": 7.975927232807968e-05, "loss": 3.1172, "step": 25813 }, { "epoch": 2.02, "learning_rate": 7.975848819885518e-05, "loss": 2.7729, "step": 25814 }, { "epoch": 2.02, "learning_rate": 7.975770406963068e-05, "loss": 2.894, "step": 25815 }, { "epoch": 2.02, "learning_rate": 7.975691994040619e-05, "loss": 2.9605, "step": 25816 }, { "epoch": 2.02, "learning_rate": 7.975613581118169e-05, "loss": 2.9928, "step": 25817 }, { "epoch": 2.02, "learning_rate": 7.97553516819572e-05, "loss": 2.8396, "step": 25818 }, { "epoch": 2.02, "learning_rate": 7.97545675527327e-05, "loss": 2.8556, "step": 25819 }, { "epoch": 2.02, "learning_rate": 7.97537834235082e-05, "loss": 2.7549, "step": 25820 }, { "epoch": 2.02, "learning_rate": 7.97529992942837e-05, "loss": 2.7181, "step": 25821 }, { "epoch": 2.02, "learning_rate": 7.975221516505921e-05, "loss": 2.5875, "step": 25822 }, { "epoch": 2.02, "learning_rate": 7.975143103583471e-05, "loss": 2.6624, "step": 25823 }, { "epoch": 2.02, "learning_rate": 7.975064690661022e-05, "loss": 3.113, "step": 25824 }, { "epoch": 2.03, "learning_rate": 7.974986277738572e-05, "loss": 2.7597, "step": 25825 }, { "epoch": 2.03, "learning_rate": 7.974907864816122e-05, "loss": 2.8764, "step": 25826 }, { "epoch": 2.03, "learning_rate": 7.974829451893673e-05, "loss": 2.9226, "step": 25827 }, { "epoch": 2.03, "learning_rate": 7.974751038971223e-05, "loss": 2.8476, "step": 25828 }, { "epoch": 2.03, "learning_rate": 7.974672626048774e-05, "loss": 3.0186, "step": 25829 }, { "epoch": 2.03, "learning_rate": 7.974594213126324e-05, "loss": 2.6833, "step": 25830 }, { "epoch": 2.03, "learning_rate": 7.974515800203874e-05, "loss": 3.0271, "step": 25831 }, { "epoch": 2.03, "learning_rate": 7.974437387281425e-05, "loss": 3.0553, "step": 25832 }, { "epoch": 2.03, "learning_rate": 7.974358974358975e-05, "loss": 2.7134, "step": 25833 }, { "epoch": 2.03, "learning_rate": 7.974280561436524e-05, "loss": 2.7437, "step": 25834 }, { "epoch": 2.03, "learning_rate": 7.974202148514074e-05, "loss": 2.7734, "step": 25835 }, { "epoch": 2.03, "learning_rate": 7.974123735591625e-05, "loss": 2.8406, "step": 25836 }, { "epoch": 2.03, "learning_rate": 7.974045322669176e-05, "loss": 2.8379, "step": 25837 }, { "epoch": 2.03, "learning_rate": 7.973966909746727e-05, "loss": 3.1772, "step": 25838 }, { "epoch": 2.03, "learning_rate": 7.973888496824277e-05, "loss": 2.7622, "step": 25839 }, { "epoch": 2.03, "learning_rate": 7.973810083901828e-05, "loss": 2.903, "step": 25840 }, { "epoch": 2.03, "learning_rate": 7.973731670979378e-05, "loss": 2.7668, "step": 25841 }, { "epoch": 2.03, "learning_rate": 7.973653258056928e-05, "loss": 2.674, "step": 25842 }, { "epoch": 2.03, "learning_rate": 7.973574845134479e-05, "loss": 3.0329, "step": 25843 }, { "epoch": 2.03, "learning_rate": 7.973496432212029e-05, "loss": 2.8475, "step": 25844 }, { "epoch": 2.03, "learning_rate": 7.97341801928958e-05, "loss": 2.8412, "step": 25845 }, { "epoch": 2.03, "learning_rate": 7.97333960636713e-05, "loss": 2.7342, "step": 25846 }, { "epoch": 2.03, "learning_rate": 7.97326119344468e-05, "loss": 2.8803, "step": 25847 }, { "epoch": 2.03, "learning_rate": 7.97318278052223e-05, "loss": 2.761, "step": 25848 }, { "epoch": 2.03, "learning_rate": 7.973104367599781e-05, "loss": 2.7306, "step": 25849 }, { "epoch": 2.03, "learning_rate": 7.973025954677331e-05, "loss": 2.9317, "step": 25850 }, { "epoch": 2.03, "learning_rate": 7.972947541754882e-05, "loss": 2.797, "step": 25851 }, { "epoch": 2.03, "learning_rate": 7.972869128832432e-05, "loss": 2.8929, "step": 25852 }, { "epoch": 2.03, "learning_rate": 7.972790715909982e-05, "loss": 2.9555, "step": 25853 }, { "epoch": 2.03, "learning_rate": 7.972712302987533e-05, "loss": 2.8158, "step": 25854 }, { "epoch": 2.03, "learning_rate": 7.972633890065083e-05, "loss": 2.825, "step": 25855 }, { "epoch": 2.03, "learning_rate": 7.972555477142633e-05, "loss": 2.9778, "step": 25856 }, { "epoch": 2.03, "learning_rate": 7.972477064220184e-05, "loss": 2.7072, "step": 25857 }, { "epoch": 2.03, "learning_rate": 7.972398651297734e-05, "loss": 2.8517, "step": 25858 }, { "epoch": 2.03, "learning_rate": 7.972320238375285e-05, "loss": 2.8734, "step": 25859 }, { "epoch": 2.03, "learning_rate": 7.972241825452835e-05, "loss": 2.7836, "step": 25860 }, { "epoch": 2.03, "learning_rate": 7.972163412530385e-05, "loss": 2.6799, "step": 25861 }, { "epoch": 2.03, "learning_rate": 7.972084999607936e-05, "loss": 2.9607, "step": 25862 }, { "epoch": 2.03, "learning_rate": 7.972006586685486e-05, "loss": 2.7251, "step": 25863 }, { "epoch": 2.03, "learning_rate": 7.971928173763036e-05, "loss": 2.8043, "step": 25864 }, { "epoch": 2.03, "learning_rate": 7.971849760840587e-05, "loss": 2.7534, "step": 25865 }, { "epoch": 2.03, "learning_rate": 7.971771347918137e-05, "loss": 2.781, "step": 25866 }, { "epoch": 2.03, "learning_rate": 7.971692934995687e-05, "loss": 2.9962, "step": 25867 }, { "epoch": 2.03, "learning_rate": 7.971614522073239e-05, "loss": 2.6664, "step": 25868 }, { "epoch": 2.03, "learning_rate": 7.97153610915079e-05, "loss": 2.8199, "step": 25869 }, { "epoch": 2.03, "learning_rate": 7.971457696228339e-05, "loss": 2.5962, "step": 25870 }, { "epoch": 2.03, "learning_rate": 7.971379283305889e-05, "loss": 2.7621, "step": 25871 }, { "epoch": 2.03, "learning_rate": 7.97130087038344e-05, "loss": 2.637, "step": 25872 }, { "epoch": 2.03, "learning_rate": 7.97122245746099e-05, "loss": 2.588, "step": 25873 }, { "epoch": 2.03, "learning_rate": 7.97114404453854e-05, "loss": 2.7023, "step": 25874 }, { "epoch": 2.03, "learning_rate": 7.97106563161609e-05, "loss": 2.7027, "step": 25875 }, { "epoch": 2.03, "learning_rate": 7.970987218693641e-05, "loss": 2.8664, "step": 25876 }, { "epoch": 2.03, "learning_rate": 7.970908805771191e-05, "loss": 2.8448, "step": 25877 }, { "epoch": 2.03, "learning_rate": 7.970830392848742e-05, "loss": 3.0522, "step": 25878 }, { "epoch": 2.03, "learning_rate": 7.970751979926292e-05, "loss": 2.6181, "step": 25879 }, { "epoch": 2.03, "learning_rate": 7.970673567003842e-05, "loss": 2.6312, "step": 25880 }, { "epoch": 2.03, "learning_rate": 7.970595154081393e-05, "loss": 2.6309, "step": 25881 }, { "epoch": 2.03, "learning_rate": 7.970516741158943e-05, "loss": 3.0561, "step": 25882 }, { "epoch": 2.03, "learning_rate": 7.970438328236493e-05, "loss": 2.8543, "step": 25883 }, { "epoch": 2.03, "learning_rate": 7.970359915314044e-05, "loss": 2.9896, "step": 25884 }, { "epoch": 2.03, "learning_rate": 7.970281502391594e-05, "loss": 2.7854, "step": 25885 }, { "epoch": 2.03, "learning_rate": 7.970203089469144e-05, "loss": 2.696, "step": 25886 }, { "epoch": 2.03, "learning_rate": 7.970124676546695e-05, "loss": 2.6175, "step": 25887 }, { "epoch": 2.03, "learning_rate": 7.970046263624245e-05, "loss": 2.6122, "step": 25888 }, { "epoch": 2.03, "learning_rate": 7.969967850701796e-05, "loss": 2.8777, "step": 25889 }, { "epoch": 2.03, "learning_rate": 7.969889437779346e-05, "loss": 2.9754, "step": 25890 }, { "epoch": 2.03, "learning_rate": 7.969811024856896e-05, "loss": 3.0137, "step": 25891 }, { "epoch": 2.03, "learning_rate": 7.969732611934447e-05, "loss": 2.915, "step": 25892 }, { "epoch": 2.03, "learning_rate": 7.969654199011997e-05, "loss": 2.6578, "step": 25893 }, { "epoch": 2.03, "learning_rate": 7.969575786089549e-05, "loss": 2.7854, "step": 25894 }, { "epoch": 2.03, "learning_rate": 7.969497373167099e-05, "loss": 3.1582, "step": 25895 }, { "epoch": 2.03, "learning_rate": 7.96941896024465e-05, "loss": 2.8867, "step": 25896 }, { "epoch": 2.03, "learning_rate": 7.9693405473222e-05, "loss": 2.6531, "step": 25897 }, { "epoch": 2.03, "learning_rate": 7.96926213439975e-05, "loss": 2.8446, "step": 25898 }, { "epoch": 2.03, "learning_rate": 7.9691837214773e-05, "loss": 2.6802, "step": 25899 }, { "epoch": 2.03, "learning_rate": 7.969105308554851e-05, "loss": 2.8614, "step": 25900 }, { "epoch": 2.03, "learning_rate": 7.969026895632401e-05, "loss": 2.9377, "step": 25901 }, { "epoch": 2.03, "learning_rate": 7.968948482709952e-05, "loss": 2.7005, "step": 25902 }, { "epoch": 2.03, "learning_rate": 7.968870069787502e-05, "loss": 2.6373, "step": 25903 }, { "epoch": 2.03, "learning_rate": 7.968791656865052e-05, "loss": 2.8718, "step": 25904 }, { "epoch": 2.03, "learning_rate": 7.968713243942603e-05, "loss": 2.8883, "step": 25905 }, { "epoch": 2.03, "learning_rate": 7.968634831020152e-05, "loss": 3.1056, "step": 25906 }, { "epoch": 2.03, "learning_rate": 7.968556418097702e-05, "loss": 2.7177, "step": 25907 }, { "epoch": 2.03, "learning_rate": 7.968478005175253e-05, "loss": 2.7248, "step": 25908 }, { "epoch": 2.03, "learning_rate": 7.968399592252803e-05, "loss": 2.8035, "step": 25909 }, { "epoch": 2.03, "learning_rate": 7.968321179330353e-05, "loss": 2.9952, "step": 25910 }, { "epoch": 2.03, "learning_rate": 7.968242766407904e-05, "loss": 2.8563, "step": 25911 }, { "epoch": 2.03, "learning_rate": 7.968164353485454e-05, "loss": 2.8694, "step": 25912 }, { "epoch": 2.03, "learning_rate": 7.968085940563004e-05, "loss": 2.8337, "step": 25913 }, { "epoch": 2.03, "learning_rate": 7.968007527640555e-05, "loss": 3.0345, "step": 25914 }, { "epoch": 2.03, "learning_rate": 7.967929114718105e-05, "loss": 2.7513, "step": 25915 }, { "epoch": 2.03, "learning_rate": 7.967850701795656e-05, "loss": 2.5262, "step": 25916 }, { "epoch": 2.03, "learning_rate": 7.967772288873206e-05, "loss": 2.905, "step": 25917 }, { "epoch": 2.03, "learning_rate": 7.967693875950756e-05, "loss": 2.6959, "step": 25918 }, { "epoch": 2.03, "learning_rate": 7.967615463028308e-05, "loss": 2.8546, "step": 25919 }, { "epoch": 2.03, "learning_rate": 7.967537050105858e-05, "loss": 2.82, "step": 25920 }, { "epoch": 2.03, "learning_rate": 7.967458637183409e-05, "loss": 2.6073, "step": 25921 }, { "epoch": 2.03, "learning_rate": 7.967380224260959e-05, "loss": 2.6797, "step": 25922 }, { "epoch": 2.03, "learning_rate": 7.96730181133851e-05, "loss": 2.7518, "step": 25923 }, { "epoch": 2.03, "learning_rate": 7.96722339841606e-05, "loss": 3.272, "step": 25924 }, { "epoch": 2.03, "learning_rate": 7.96714498549361e-05, "loss": 2.7938, "step": 25925 }, { "epoch": 2.03, "learning_rate": 7.96706657257116e-05, "loss": 2.727, "step": 25926 }, { "epoch": 2.03, "learning_rate": 7.966988159648711e-05, "loss": 2.7168, "step": 25927 }, { "epoch": 2.03, "learning_rate": 7.966909746726261e-05, "loss": 3.0709, "step": 25928 }, { "epoch": 2.03, "learning_rate": 7.966831333803812e-05, "loss": 2.5836, "step": 25929 }, { "epoch": 2.03, "learning_rate": 7.966752920881362e-05, "loss": 2.8291, "step": 25930 }, { "epoch": 2.03, "learning_rate": 7.966674507958912e-05, "loss": 3.0587, "step": 25931 }, { "epoch": 2.03, "learning_rate": 7.966596095036463e-05, "loss": 2.769, "step": 25932 }, { "epoch": 2.03, "learning_rate": 7.966517682114013e-05, "loss": 2.6991, "step": 25933 }, { "epoch": 2.03, "learning_rate": 7.966439269191563e-05, "loss": 2.8605, "step": 25934 }, { "epoch": 2.03, "learning_rate": 7.966360856269114e-05, "loss": 2.881, "step": 25935 }, { "epoch": 2.03, "learning_rate": 7.966282443346664e-05, "loss": 2.5697, "step": 25936 }, { "epoch": 2.03, "learning_rate": 7.966204030424215e-05, "loss": 2.7687, "step": 25937 }, { "epoch": 2.03, "learning_rate": 7.966125617501765e-05, "loss": 2.8302, "step": 25938 }, { "epoch": 2.03, "learning_rate": 7.966047204579315e-05, "loss": 3.0478, "step": 25939 }, { "epoch": 2.03, "learning_rate": 7.965968791656866e-05, "loss": 2.8485, "step": 25940 }, { "epoch": 2.03, "learning_rate": 7.965890378734416e-05, "loss": 2.9518, "step": 25941 }, { "epoch": 2.03, "learning_rate": 7.965811965811965e-05, "loss": 2.6432, "step": 25942 }, { "epoch": 2.03, "learning_rate": 7.965733552889515e-05, "loss": 2.7523, "step": 25943 }, { "epoch": 2.03, "learning_rate": 7.965655139967066e-05, "loss": 2.6643, "step": 25944 }, { "epoch": 2.03, "learning_rate": 7.965576727044618e-05, "loss": 2.8558, "step": 25945 }, { "epoch": 2.03, "learning_rate": 7.965498314122168e-05, "loss": 2.8824, "step": 25946 }, { "epoch": 2.03, "learning_rate": 7.965419901199718e-05, "loss": 2.8941, "step": 25947 }, { "epoch": 2.03, "learning_rate": 7.965341488277269e-05, "loss": 2.8788, "step": 25948 }, { "epoch": 2.03, "learning_rate": 7.965263075354819e-05, "loss": 2.8013, "step": 25949 }, { "epoch": 2.03, "learning_rate": 7.96518466243237e-05, "loss": 2.7918, "step": 25950 }, { "epoch": 2.03, "learning_rate": 7.96510624950992e-05, "loss": 2.9048, "step": 25951 }, { "epoch": 2.03, "learning_rate": 7.96502783658747e-05, "loss": 2.6325, "step": 25952 }, { "epoch": 2.04, "learning_rate": 7.96494942366502e-05, "loss": 2.7497, "step": 25953 }, { "epoch": 2.04, "learning_rate": 7.964871010742571e-05, "loss": 2.939, "step": 25954 }, { "epoch": 2.04, "learning_rate": 7.964792597820121e-05, "loss": 2.9348, "step": 25955 }, { "epoch": 2.04, "learning_rate": 7.964714184897672e-05, "loss": 2.7096, "step": 25956 }, { "epoch": 2.04, "learning_rate": 7.964635771975222e-05, "loss": 3.0342, "step": 25957 }, { "epoch": 2.04, "learning_rate": 7.964557359052772e-05, "loss": 2.8257, "step": 25958 }, { "epoch": 2.04, "learning_rate": 7.964478946130323e-05, "loss": 3.006, "step": 25959 }, { "epoch": 2.04, "learning_rate": 7.964400533207873e-05, "loss": 2.8497, "step": 25960 }, { "epoch": 2.04, "learning_rate": 7.964322120285423e-05, "loss": 3.0584, "step": 25961 }, { "epoch": 2.04, "learning_rate": 7.964243707362974e-05, "loss": 2.6562, "step": 25962 }, { "epoch": 2.04, "learning_rate": 7.964165294440524e-05, "loss": 2.6753, "step": 25963 }, { "epoch": 2.04, "learning_rate": 7.964086881518075e-05, "loss": 2.8145, "step": 25964 }, { "epoch": 2.04, "learning_rate": 7.964008468595625e-05, "loss": 2.7781, "step": 25965 }, { "epoch": 2.04, "learning_rate": 7.963930055673175e-05, "loss": 3.0717, "step": 25966 }, { "epoch": 2.04, "learning_rate": 7.963851642750726e-05, "loss": 2.7413, "step": 25967 }, { "epoch": 2.04, "learning_rate": 7.963773229828276e-05, "loss": 3.0724, "step": 25968 }, { "epoch": 2.04, "learning_rate": 7.963694816905826e-05, "loss": 2.842, "step": 25969 }, { "epoch": 2.04, "learning_rate": 7.963616403983377e-05, "loss": 2.7797, "step": 25970 }, { "epoch": 2.04, "learning_rate": 7.963537991060927e-05, "loss": 2.6738, "step": 25971 }, { "epoch": 2.04, "learning_rate": 7.963459578138477e-05, "loss": 2.7815, "step": 25972 }, { "epoch": 2.04, "learning_rate": 7.963381165216028e-05, "loss": 2.6986, "step": 25973 }, { "epoch": 2.04, "learning_rate": 7.963302752293578e-05, "loss": 2.9245, "step": 25974 }, { "epoch": 2.04, "learning_rate": 7.963224339371129e-05, "loss": 2.6717, "step": 25975 }, { "epoch": 2.04, "learning_rate": 7.96314592644868e-05, "loss": 2.642, "step": 25976 }, { "epoch": 2.04, "learning_rate": 7.96306751352623e-05, "loss": 3.132, "step": 25977 }, { "epoch": 2.04, "learning_rate": 7.96298910060378e-05, "loss": 2.6708, "step": 25978 }, { "epoch": 2.04, "learning_rate": 7.96291068768133e-05, "loss": 3.1367, "step": 25979 }, { "epoch": 2.04, "learning_rate": 7.96283227475888e-05, "loss": 2.8787, "step": 25980 }, { "epoch": 2.04, "learning_rate": 7.962753861836431e-05, "loss": 2.8242, "step": 25981 }, { "epoch": 2.04, "learning_rate": 7.962675448913981e-05, "loss": 2.788, "step": 25982 }, { "epoch": 2.04, "learning_rate": 7.962597035991532e-05, "loss": 2.7426, "step": 25983 }, { "epoch": 2.04, "learning_rate": 7.962518623069082e-05, "loss": 2.8212, "step": 25984 }, { "epoch": 2.04, "learning_rate": 7.962440210146632e-05, "loss": 2.7298, "step": 25985 }, { "epoch": 2.04, "learning_rate": 7.962361797224183e-05, "loss": 2.8383, "step": 25986 }, { "epoch": 2.04, "learning_rate": 7.962283384301733e-05, "loss": 2.7381, "step": 25987 }, { "epoch": 2.04, "learning_rate": 7.962204971379283e-05, "loss": 2.9615, "step": 25988 }, { "epoch": 2.04, "learning_rate": 7.962126558456834e-05, "loss": 2.8296, "step": 25989 }, { "epoch": 2.04, "learning_rate": 7.962048145534384e-05, "loss": 2.9143, "step": 25990 }, { "epoch": 2.04, "learning_rate": 7.961969732611934e-05, "loss": 2.808, "step": 25991 }, { "epoch": 2.04, "learning_rate": 7.961891319689485e-05, "loss": 2.6799, "step": 25992 }, { "epoch": 2.04, "learning_rate": 7.961812906767035e-05, "loss": 2.8585, "step": 25993 }, { "epoch": 2.04, "learning_rate": 7.961734493844586e-05, "loss": 2.7413, "step": 25994 }, { "epoch": 2.04, "learning_rate": 7.961656080922136e-05, "loss": 2.7762, "step": 25995 }, { "epoch": 2.04, "learning_rate": 7.961577667999686e-05, "loss": 2.9067, "step": 25996 }, { "epoch": 2.04, "learning_rate": 7.961499255077237e-05, "loss": 2.7556, "step": 25997 }, { "epoch": 2.04, "learning_rate": 7.961420842154787e-05, "loss": 2.9541, "step": 25998 }, { "epoch": 2.04, "learning_rate": 7.961342429232337e-05, "loss": 2.931, "step": 25999 }, { "epoch": 2.04, "learning_rate": 7.961264016309888e-05, "loss": 2.6503, "step": 26000 }, { "epoch": 2.04, "eval_loss": 2.9073386192321777, "eval_runtime": 218.0389, "eval_samples_per_second": 51.972, "eval_steps_per_second": 3.252, "step": 26000 }, { "epoch": 2.04, "learning_rate": 7.961185603387438e-05, "loss": 2.6306, "step": 26001 }, { "epoch": 2.04, "learning_rate": 7.96110719046499e-05, "loss": 2.7643, "step": 26002 }, { "epoch": 2.04, "learning_rate": 7.96102877754254e-05, "loss": 2.7533, "step": 26003 }, { "epoch": 2.04, "learning_rate": 7.96095036462009e-05, "loss": 2.788, "step": 26004 }, { "epoch": 2.04, "learning_rate": 7.960871951697641e-05, "loss": 3.1504, "step": 26005 }, { "epoch": 2.04, "learning_rate": 7.960793538775191e-05, "loss": 2.7441, "step": 26006 }, { "epoch": 2.04, "learning_rate": 7.960715125852742e-05, "loss": 2.8158, "step": 26007 }, { "epoch": 2.04, "learning_rate": 7.960636712930292e-05, "loss": 2.6602, "step": 26008 }, { "epoch": 2.04, "learning_rate": 7.960558300007842e-05, "loss": 2.8301, "step": 26009 }, { "epoch": 2.04, "learning_rate": 7.960479887085393e-05, "loss": 2.7598, "step": 26010 }, { "epoch": 2.04, "learning_rate": 7.960401474162943e-05, "loss": 2.9984, "step": 26011 }, { "epoch": 2.04, "learning_rate": 7.960323061240494e-05, "loss": 2.8962, "step": 26012 }, { "epoch": 2.04, "learning_rate": 7.960244648318044e-05, "loss": 2.7624, "step": 26013 }, { "epoch": 2.04, "learning_rate": 7.960166235395593e-05, "loss": 2.7037, "step": 26014 }, { "epoch": 2.04, "learning_rate": 7.960087822473143e-05, "loss": 2.6996, "step": 26015 }, { "epoch": 2.04, "learning_rate": 7.960009409550694e-05, "loss": 2.67, "step": 26016 }, { "epoch": 2.04, "learning_rate": 7.959930996628244e-05, "loss": 2.7231, "step": 26017 }, { "epoch": 2.04, "learning_rate": 7.959852583705794e-05, "loss": 2.7323, "step": 26018 }, { "epoch": 2.04, "learning_rate": 7.959774170783345e-05, "loss": 2.784, "step": 26019 }, { "epoch": 2.04, "learning_rate": 7.959695757860895e-05, "loss": 2.7473, "step": 26020 }, { "epoch": 2.04, "learning_rate": 7.959617344938445e-05, "loss": 2.8781, "step": 26021 }, { "epoch": 2.04, "learning_rate": 7.959538932015996e-05, "loss": 2.8765, "step": 26022 }, { "epoch": 2.04, "learning_rate": 7.959460519093546e-05, "loss": 2.9156, "step": 26023 }, { "epoch": 2.04, "learning_rate": 7.959382106171097e-05, "loss": 2.708, "step": 26024 }, { "epoch": 2.04, "learning_rate": 7.959303693248647e-05, "loss": 2.78, "step": 26025 }, { "epoch": 2.04, "learning_rate": 7.959225280326197e-05, "loss": 2.7004, "step": 26026 }, { "epoch": 2.04, "learning_rate": 7.959146867403749e-05, "loss": 2.8494, "step": 26027 }, { "epoch": 2.04, "learning_rate": 7.9590684544813e-05, "loss": 2.6192, "step": 26028 }, { "epoch": 2.04, "learning_rate": 7.95899004155885e-05, "loss": 2.7049, "step": 26029 }, { "epoch": 2.04, "learning_rate": 7.9589116286364e-05, "loss": 2.7918, "step": 26030 }, { "epoch": 2.04, "learning_rate": 7.95883321571395e-05, "loss": 2.8689, "step": 26031 }, { "epoch": 2.04, "learning_rate": 7.958754802791501e-05, "loss": 3.0611, "step": 26032 }, { "epoch": 2.04, "learning_rate": 7.958676389869051e-05, "loss": 2.8774, "step": 26033 }, { "epoch": 2.04, "learning_rate": 7.958597976946602e-05, "loss": 2.5543, "step": 26034 }, { "epoch": 2.04, "learning_rate": 7.958519564024152e-05, "loss": 2.8563, "step": 26035 }, { "epoch": 2.04, "learning_rate": 7.958441151101702e-05, "loss": 2.6776, "step": 26036 }, { "epoch": 2.04, "learning_rate": 7.958362738179253e-05, "loss": 2.7928, "step": 26037 }, { "epoch": 2.04, "learning_rate": 7.958284325256803e-05, "loss": 2.7291, "step": 26038 }, { "epoch": 2.04, "learning_rate": 7.958205912334353e-05, "loss": 2.621, "step": 26039 }, { "epoch": 2.04, "learning_rate": 7.958127499411904e-05, "loss": 2.9028, "step": 26040 }, { "epoch": 2.04, "learning_rate": 7.958049086489454e-05, "loss": 2.8108, "step": 26041 }, { "epoch": 2.04, "learning_rate": 7.957970673567005e-05, "loss": 3.1328, "step": 26042 }, { "epoch": 2.04, "learning_rate": 7.957892260644555e-05, "loss": 3.0205, "step": 26043 }, { "epoch": 2.04, "learning_rate": 7.957813847722105e-05, "loss": 2.7087, "step": 26044 }, { "epoch": 2.04, "learning_rate": 7.957735434799656e-05, "loss": 2.8423, "step": 26045 }, { "epoch": 2.04, "learning_rate": 7.957657021877206e-05, "loss": 2.8711, "step": 26046 }, { "epoch": 2.04, "learning_rate": 7.957578608954756e-05, "loss": 2.6861, "step": 26047 }, { "epoch": 2.04, "learning_rate": 7.957500196032307e-05, "loss": 2.5938, "step": 26048 }, { "epoch": 2.04, "learning_rate": 7.957421783109857e-05, "loss": 2.7864, "step": 26049 }, { "epoch": 2.04, "learning_rate": 7.957343370187406e-05, "loss": 2.7325, "step": 26050 }, { "epoch": 2.04, "learning_rate": 7.957264957264957e-05, "loss": 2.7995, "step": 26051 }, { "epoch": 2.04, "learning_rate": 7.957186544342507e-05, "loss": 3.072, "step": 26052 }, { "epoch": 2.04, "learning_rate": 7.957108131420059e-05, "loss": 3.0231, "step": 26053 }, { "epoch": 2.04, "learning_rate": 7.957029718497609e-05, "loss": 2.7491, "step": 26054 }, { "epoch": 2.04, "learning_rate": 7.95695130557516e-05, "loss": 2.8142, "step": 26055 }, { "epoch": 2.04, "learning_rate": 7.95687289265271e-05, "loss": 2.7823, "step": 26056 }, { "epoch": 2.04, "learning_rate": 7.95679447973026e-05, "loss": 2.7228, "step": 26057 }, { "epoch": 2.04, "learning_rate": 7.95671606680781e-05, "loss": 2.752, "step": 26058 }, { "epoch": 2.04, "learning_rate": 7.956637653885361e-05, "loss": 2.667, "step": 26059 }, { "epoch": 2.04, "learning_rate": 7.956559240962911e-05, "loss": 2.7076, "step": 26060 }, { "epoch": 2.04, "learning_rate": 7.956480828040462e-05, "loss": 2.6188, "step": 26061 }, { "epoch": 2.04, "learning_rate": 7.956402415118012e-05, "loss": 2.8376, "step": 26062 }, { "epoch": 2.04, "learning_rate": 7.956324002195562e-05, "loss": 2.7872, "step": 26063 }, { "epoch": 2.04, "learning_rate": 7.956245589273113e-05, "loss": 2.8426, "step": 26064 }, { "epoch": 2.04, "learning_rate": 7.956167176350663e-05, "loss": 2.8963, "step": 26065 }, { "epoch": 2.04, "learning_rate": 7.956088763428213e-05, "loss": 2.7351, "step": 26066 }, { "epoch": 2.04, "learning_rate": 7.956010350505764e-05, "loss": 2.7159, "step": 26067 }, { "epoch": 2.04, "learning_rate": 7.955931937583314e-05, "loss": 2.6939, "step": 26068 }, { "epoch": 2.04, "learning_rate": 7.955853524660864e-05, "loss": 2.856, "step": 26069 }, { "epoch": 2.04, "learning_rate": 7.955775111738415e-05, "loss": 2.6589, "step": 26070 }, { "epoch": 2.04, "learning_rate": 7.955696698815965e-05, "loss": 2.9064, "step": 26071 }, { "epoch": 2.04, "learning_rate": 7.955618285893516e-05, "loss": 2.845, "step": 26072 }, { "epoch": 2.04, "learning_rate": 7.955539872971066e-05, "loss": 2.6769, "step": 26073 }, { "epoch": 2.04, "learning_rate": 7.955461460048616e-05, "loss": 2.6848, "step": 26074 }, { "epoch": 2.04, "learning_rate": 7.955383047126167e-05, "loss": 2.8189, "step": 26075 }, { "epoch": 2.04, "learning_rate": 7.955304634203717e-05, "loss": 2.7756, "step": 26076 }, { "epoch": 2.04, "learning_rate": 7.955226221281267e-05, "loss": 2.8029, "step": 26077 }, { "epoch": 2.04, "learning_rate": 7.955147808358818e-05, "loss": 2.8949, "step": 26078 }, { "epoch": 2.04, "learning_rate": 7.955069395436368e-05, "loss": 2.8599, "step": 26079 }, { "epoch": 2.05, "learning_rate": 7.954990982513919e-05, "loss": 2.9139, "step": 26080 }, { "epoch": 2.05, "learning_rate": 7.954912569591469e-05, "loss": 2.6478, "step": 26081 }, { "epoch": 2.05, "learning_rate": 7.954834156669019e-05, "loss": 2.7748, "step": 26082 }, { "epoch": 2.05, "learning_rate": 7.95475574374657e-05, "loss": 2.5019, "step": 26083 }, { "epoch": 2.05, "learning_rate": 7.954677330824121e-05, "loss": 2.915, "step": 26084 }, { "epoch": 2.05, "learning_rate": 7.954598917901672e-05, "loss": 2.7597, "step": 26085 }, { "epoch": 2.05, "learning_rate": 7.954520504979221e-05, "loss": 2.8213, "step": 26086 }, { "epoch": 2.05, "learning_rate": 7.954442092056771e-05, "loss": 2.8131, "step": 26087 }, { "epoch": 2.05, "learning_rate": 7.954363679134321e-05, "loss": 2.6855, "step": 26088 }, { "epoch": 2.05, "learning_rate": 7.954285266211872e-05, "loss": 2.7099, "step": 26089 }, { "epoch": 2.05, "learning_rate": 7.954206853289422e-05, "loss": 2.9258, "step": 26090 }, { "epoch": 2.05, "learning_rate": 7.954128440366973e-05, "loss": 2.6473, "step": 26091 }, { "epoch": 2.05, "learning_rate": 7.954050027444523e-05, "loss": 2.9154, "step": 26092 }, { "epoch": 2.05, "learning_rate": 7.953971614522073e-05, "loss": 2.9529, "step": 26093 }, { "epoch": 2.05, "learning_rate": 7.953893201599624e-05, "loss": 2.862, "step": 26094 }, { "epoch": 2.05, "learning_rate": 7.953814788677174e-05, "loss": 2.7579, "step": 26095 }, { "epoch": 2.05, "learning_rate": 7.953736375754724e-05, "loss": 2.9037, "step": 26096 }, { "epoch": 2.05, "learning_rate": 7.953657962832275e-05, "loss": 2.753, "step": 26097 }, { "epoch": 2.05, "learning_rate": 7.953579549909825e-05, "loss": 2.9608, "step": 26098 }, { "epoch": 2.05, "learning_rate": 7.953501136987376e-05, "loss": 2.8489, "step": 26099 }, { "epoch": 2.05, "learning_rate": 7.953422724064926e-05, "loss": 2.6342, "step": 26100 }, { "epoch": 2.05, "learning_rate": 7.953344311142476e-05, "loss": 2.8016, "step": 26101 }, { "epoch": 2.05, "learning_rate": 7.953265898220027e-05, "loss": 2.7241, "step": 26102 }, { "epoch": 2.05, "learning_rate": 7.953187485297577e-05, "loss": 2.6468, "step": 26103 }, { "epoch": 2.05, "learning_rate": 7.953109072375127e-05, "loss": 2.9042, "step": 26104 }, { "epoch": 2.05, "learning_rate": 7.953030659452678e-05, "loss": 3.078, "step": 26105 }, { "epoch": 2.05, "learning_rate": 7.952952246530228e-05, "loss": 2.8874, "step": 26106 }, { "epoch": 2.05, "learning_rate": 7.952873833607778e-05, "loss": 2.6505, "step": 26107 }, { "epoch": 2.05, "learning_rate": 7.952795420685329e-05, "loss": 2.5586, "step": 26108 }, { "epoch": 2.05, "learning_rate": 7.952717007762879e-05, "loss": 2.7726, "step": 26109 }, { "epoch": 2.05, "learning_rate": 7.952638594840431e-05, "loss": 2.7405, "step": 26110 }, { "epoch": 2.05, "learning_rate": 7.952560181917981e-05, "loss": 2.8388, "step": 26111 }, { "epoch": 2.05, "learning_rate": 7.952481768995532e-05, "loss": 2.9109, "step": 26112 }, { "epoch": 2.05, "learning_rate": 7.952403356073082e-05, "loss": 2.6182, "step": 26113 }, { "epoch": 2.05, "learning_rate": 7.952324943150632e-05, "loss": 2.781, "step": 26114 }, { "epoch": 2.05, "learning_rate": 7.952246530228183e-05, "loss": 3.1183, "step": 26115 }, { "epoch": 2.05, "learning_rate": 7.952168117305733e-05, "loss": 2.9544, "step": 26116 }, { "epoch": 2.05, "learning_rate": 7.952089704383283e-05, "loss": 2.7393, "step": 26117 }, { "epoch": 2.05, "learning_rate": 7.952011291460834e-05, "loss": 2.7251, "step": 26118 }, { "epoch": 2.05, "learning_rate": 7.951932878538384e-05, "loss": 2.8759, "step": 26119 }, { "epoch": 2.05, "learning_rate": 7.951854465615935e-05, "loss": 2.8111, "step": 26120 }, { "epoch": 2.05, "learning_rate": 7.951776052693485e-05, "loss": 2.8725, "step": 26121 }, { "epoch": 2.05, "learning_rate": 7.951697639771034e-05, "loss": 2.7044, "step": 26122 }, { "epoch": 2.05, "learning_rate": 7.951619226848584e-05, "loss": 2.6762, "step": 26123 }, { "epoch": 2.05, "learning_rate": 7.951540813926135e-05, "loss": 2.6881, "step": 26124 }, { "epoch": 2.05, "learning_rate": 7.951462401003685e-05, "loss": 3.0137, "step": 26125 }, { "epoch": 2.05, "learning_rate": 7.951383988081235e-05, "loss": 2.7622, "step": 26126 }, { "epoch": 2.05, "learning_rate": 7.951305575158786e-05, "loss": 2.6974, "step": 26127 }, { "epoch": 2.05, "learning_rate": 7.951227162236336e-05, "loss": 2.8856, "step": 26128 }, { "epoch": 2.05, "learning_rate": 7.951148749313887e-05, "loss": 2.7309, "step": 26129 }, { "epoch": 2.05, "learning_rate": 7.951070336391437e-05, "loss": 2.7877, "step": 26130 }, { "epoch": 2.05, "learning_rate": 7.950991923468987e-05, "loss": 2.5974, "step": 26131 }, { "epoch": 2.05, "learning_rate": 7.950913510546538e-05, "loss": 2.718, "step": 26132 }, { "epoch": 2.05, "learning_rate": 7.950835097624088e-05, "loss": 2.7634, "step": 26133 }, { "epoch": 2.05, "learning_rate": 7.950756684701638e-05, "loss": 2.7651, "step": 26134 }, { "epoch": 2.05, "learning_rate": 7.95067827177919e-05, "loss": 2.632, "step": 26135 }, { "epoch": 2.05, "learning_rate": 7.95059985885674e-05, "loss": 2.6404, "step": 26136 }, { "epoch": 2.05, "learning_rate": 7.950521445934291e-05, "loss": 2.6463, "step": 26137 }, { "epoch": 2.05, "learning_rate": 7.950443033011841e-05, "loss": 2.9326, "step": 26138 }, { "epoch": 2.05, "learning_rate": 7.950364620089392e-05, "loss": 2.7095, "step": 26139 }, { "epoch": 2.05, "learning_rate": 7.950286207166942e-05, "loss": 2.7482, "step": 26140 }, { "epoch": 2.05, "learning_rate": 7.950207794244492e-05, "loss": 2.8429, "step": 26141 }, { "epoch": 2.05, "learning_rate": 7.950129381322043e-05, "loss": 2.862, "step": 26142 }, { "epoch": 2.05, "learning_rate": 7.950050968399593e-05, "loss": 2.7911, "step": 26143 }, { "epoch": 2.05, "learning_rate": 7.949972555477143e-05, "loss": 2.7511, "step": 26144 }, { "epoch": 2.05, "learning_rate": 7.949894142554694e-05, "loss": 2.6651, "step": 26145 }, { "epoch": 2.05, "learning_rate": 7.949815729632244e-05, "loss": 2.8362, "step": 26146 }, { "epoch": 2.05, "learning_rate": 7.949737316709795e-05, "loss": 2.6514, "step": 26147 }, { "epoch": 2.05, "learning_rate": 7.949658903787345e-05, "loss": 2.7195, "step": 26148 }, { "epoch": 2.05, "learning_rate": 7.949580490864895e-05, "loss": 2.6162, "step": 26149 }, { "epoch": 2.05, "learning_rate": 7.949502077942446e-05, "loss": 2.8709, "step": 26150 }, { "epoch": 2.05, "learning_rate": 7.949423665019996e-05, "loss": 3.0157, "step": 26151 }, { "epoch": 2.05, "learning_rate": 7.949345252097546e-05, "loss": 2.8254, "step": 26152 }, { "epoch": 2.05, "learning_rate": 7.949266839175097e-05, "loss": 2.708, "step": 26153 }, { "epoch": 2.05, "learning_rate": 7.949188426252647e-05, "loss": 2.6665, "step": 26154 }, { "epoch": 2.05, "learning_rate": 7.949110013330197e-05, "loss": 2.8078, "step": 26155 }, { "epoch": 2.05, "learning_rate": 7.949031600407748e-05, "loss": 3.0404, "step": 26156 }, { "epoch": 2.05, "learning_rate": 7.948953187485298e-05, "loss": 2.7992, "step": 26157 }, { "epoch": 2.05, "learning_rate": 7.948874774562847e-05, "loss": 2.8995, "step": 26158 }, { "epoch": 2.05, "learning_rate": 7.948796361640398e-05, "loss": 2.751, "step": 26159 }, { "epoch": 2.05, "learning_rate": 7.948717948717948e-05, "loss": 2.8122, "step": 26160 }, { "epoch": 2.05, "learning_rate": 7.9486395357955e-05, "loss": 2.8502, "step": 26161 }, { "epoch": 2.05, "learning_rate": 7.94856112287305e-05, "loss": 2.7544, "step": 26162 }, { "epoch": 2.05, "learning_rate": 7.9484827099506e-05, "loss": 2.8058, "step": 26163 }, { "epoch": 2.05, "learning_rate": 7.948404297028151e-05, "loss": 2.7892, "step": 26164 }, { "epoch": 2.05, "learning_rate": 7.948325884105701e-05, "loss": 2.8405, "step": 26165 }, { "epoch": 2.05, "learning_rate": 7.948247471183252e-05, "loss": 2.8322, "step": 26166 }, { "epoch": 2.05, "learning_rate": 7.948169058260802e-05, "loss": 2.5538, "step": 26167 }, { "epoch": 2.05, "learning_rate": 7.948090645338352e-05, "loss": 2.8667, "step": 26168 }, { "epoch": 2.05, "learning_rate": 7.948012232415903e-05, "loss": 2.5996, "step": 26169 }, { "epoch": 2.05, "learning_rate": 7.947933819493453e-05, "loss": 2.9127, "step": 26170 }, { "epoch": 2.05, "learning_rate": 7.947855406571003e-05, "loss": 2.8871, "step": 26171 }, { "epoch": 2.05, "learning_rate": 7.947776993648554e-05, "loss": 2.6285, "step": 26172 }, { "epoch": 2.05, "learning_rate": 7.947698580726104e-05, "loss": 3.0303, "step": 26173 }, { "epoch": 2.05, "learning_rate": 7.947620167803654e-05, "loss": 2.9387, "step": 26174 }, { "epoch": 2.05, "learning_rate": 7.947541754881205e-05, "loss": 2.7664, "step": 26175 }, { "epoch": 2.05, "learning_rate": 7.947463341958755e-05, "loss": 2.646, "step": 26176 }, { "epoch": 2.05, "learning_rate": 7.947384929036306e-05, "loss": 2.6962, "step": 26177 }, { "epoch": 2.05, "learning_rate": 7.947306516113856e-05, "loss": 3.0182, "step": 26178 }, { "epoch": 2.05, "learning_rate": 7.947228103191406e-05, "loss": 2.7831, "step": 26179 }, { "epoch": 2.05, "learning_rate": 7.947149690268957e-05, "loss": 2.7948, "step": 26180 }, { "epoch": 2.05, "learning_rate": 7.947071277346507e-05, "loss": 2.4943, "step": 26181 }, { "epoch": 2.05, "learning_rate": 7.946992864424057e-05, "loss": 2.6809, "step": 26182 }, { "epoch": 2.05, "learning_rate": 7.946914451501608e-05, "loss": 2.9151, "step": 26183 }, { "epoch": 2.05, "learning_rate": 7.946836038579158e-05, "loss": 2.9731, "step": 26184 }, { "epoch": 2.05, "learning_rate": 7.946757625656709e-05, "loss": 2.7538, "step": 26185 }, { "epoch": 2.05, "learning_rate": 7.946679212734259e-05, "loss": 2.7632, "step": 26186 }, { "epoch": 2.05, "learning_rate": 7.946600799811809e-05, "loss": 2.8844, "step": 26187 }, { "epoch": 2.05, "learning_rate": 7.94652238688936e-05, "loss": 2.7472, "step": 26188 }, { "epoch": 2.05, "learning_rate": 7.94644397396691e-05, "loss": 2.7926, "step": 26189 }, { "epoch": 2.05, "learning_rate": 7.94636556104446e-05, "loss": 2.8132, "step": 26190 }, { "epoch": 2.05, "learning_rate": 7.946287148122011e-05, "loss": 2.8106, "step": 26191 }, { "epoch": 2.05, "learning_rate": 7.946208735199562e-05, "loss": 2.6826, "step": 26192 }, { "epoch": 2.05, "learning_rate": 7.946130322277113e-05, "loss": 2.9994, "step": 26193 }, { "epoch": 2.05, "learning_rate": 7.946051909354662e-05, "loss": 2.8494, "step": 26194 }, { "epoch": 2.05, "learning_rate": 7.945973496432212e-05, "loss": 2.6895, "step": 26195 }, { "epoch": 2.05, "learning_rate": 7.945895083509763e-05, "loss": 2.7862, "step": 26196 }, { "epoch": 2.05, "learning_rate": 7.945816670587313e-05, "loss": 2.6799, "step": 26197 }, { "epoch": 2.05, "learning_rate": 7.945738257664863e-05, "loss": 2.7422, "step": 26198 }, { "epoch": 2.05, "learning_rate": 7.945659844742414e-05, "loss": 2.6197, "step": 26199 }, { "epoch": 2.05, "learning_rate": 7.945581431819964e-05, "loss": 3.0272, "step": 26200 }, { "epoch": 2.05, "learning_rate": 7.945503018897514e-05, "loss": 2.8158, "step": 26201 }, { "epoch": 2.05, "learning_rate": 7.945424605975065e-05, "loss": 3.095, "step": 26202 }, { "epoch": 2.05, "learning_rate": 7.945346193052615e-05, "loss": 2.8414, "step": 26203 }, { "epoch": 2.05, "learning_rate": 7.945267780130165e-05, "loss": 2.8397, "step": 26204 }, { "epoch": 2.05, "learning_rate": 7.945189367207716e-05, "loss": 2.7903, "step": 26205 }, { "epoch": 2.05, "learning_rate": 7.945110954285266e-05, "loss": 3.0197, "step": 26206 }, { "epoch": 2.05, "learning_rate": 7.945032541362817e-05, "loss": 3.0583, "step": 26207 }, { "epoch": 2.06, "learning_rate": 7.944954128440367e-05, "loss": 2.857, "step": 26208 }, { "epoch": 2.06, "learning_rate": 7.944875715517917e-05, "loss": 2.7872, "step": 26209 }, { "epoch": 2.06, "learning_rate": 7.944797302595468e-05, "loss": 2.9028, "step": 26210 }, { "epoch": 2.06, "learning_rate": 7.944718889673018e-05, "loss": 2.8665, "step": 26211 }, { "epoch": 2.06, "learning_rate": 7.944640476750568e-05, "loss": 2.8206, "step": 26212 }, { "epoch": 2.06, "learning_rate": 7.944562063828119e-05, "loss": 3.0997, "step": 26213 }, { "epoch": 2.06, "learning_rate": 7.944483650905669e-05, "loss": 2.67, "step": 26214 }, { "epoch": 2.06, "learning_rate": 7.94440523798322e-05, "loss": 2.774, "step": 26215 }, { "epoch": 2.06, "learning_rate": 7.94432682506077e-05, "loss": 2.7495, "step": 26216 }, { "epoch": 2.06, "learning_rate": 7.94424841213832e-05, "loss": 2.7984, "step": 26217 }, { "epoch": 2.06, "learning_rate": 7.944169999215872e-05, "loss": 2.8665, "step": 26218 }, { "epoch": 2.06, "learning_rate": 7.944091586293422e-05, "loss": 2.7709, "step": 26219 }, { "epoch": 2.06, "learning_rate": 7.944013173370973e-05, "loss": 3.1146, "step": 26220 }, { "epoch": 2.06, "learning_rate": 7.943934760448523e-05, "loss": 2.7586, "step": 26221 }, { "epoch": 2.06, "learning_rate": 7.943856347526073e-05, "loss": 2.6091, "step": 26222 }, { "epoch": 2.06, "learning_rate": 7.943777934603624e-05, "loss": 2.7719, "step": 26223 }, { "epoch": 2.06, "learning_rate": 7.943699521681174e-05, "loss": 2.8393, "step": 26224 }, { "epoch": 2.06, "learning_rate": 7.943621108758725e-05, "loss": 2.7769, "step": 26225 }, { "epoch": 2.06, "learning_rate": 7.943542695836275e-05, "loss": 2.7631, "step": 26226 }, { "epoch": 2.06, "learning_rate": 7.943464282913825e-05, "loss": 2.7677, "step": 26227 }, { "epoch": 2.06, "learning_rate": 7.943385869991376e-05, "loss": 2.781, "step": 26228 }, { "epoch": 2.06, "learning_rate": 7.943307457068926e-05, "loss": 2.8776, "step": 26229 }, { "epoch": 2.06, "learning_rate": 7.943229044146475e-05, "loss": 2.7621, "step": 26230 }, { "epoch": 2.06, "learning_rate": 7.943150631224025e-05, "loss": 2.776, "step": 26231 }, { "epoch": 2.06, "learning_rate": 7.943072218301576e-05, "loss": 2.7581, "step": 26232 }, { "epoch": 2.06, "learning_rate": 7.942993805379126e-05, "loss": 2.8236, "step": 26233 }, { "epoch": 2.06, "learning_rate": 7.942915392456677e-05, "loss": 2.7243, "step": 26234 }, { "epoch": 2.06, "learning_rate": 7.942836979534227e-05, "loss": 2.7062, "step": 26235 }, { "epoch": 2.06, "learning_rate": 7.942758566611777e-05, "loss": 2.7536, "step": 26236 }, { "epoch": 2.06, "learning_rate": 7.942680153689328e-05, "loss": 2.7584, "step": 26237 }, { "epoch": 2.06, "learning_rate": 7.942601740766878e-05, "loss": 2.6019, "step": 26238 }, { "epoch": 2.06, "learning_rate": 7.942523327844428e-05, "loss": 2.6662, "step": 26239 }, { "epoch": 2.06, "learning_rate": 7.942444914921979e-05, "loss": 2.8202, "step": 26240 }, { "epoch": 2.06, "learning_rate": 7.942366501999529e-05, "loss": 2.7868, "step": 26241 }, { "epoch": 2.06, "learning_rate": 7.94228808907708e-05, "loss": 2.9142, "step": 26242 }, { "epoch": 2.06, "learning_rate": 7.942209676154631e-05, "loss": 2.8167, "step": 26243 }, { "epoch": 2.06, "learning_rate": 7.942131263232182e-05, "loss": 2.8888, "step": 26244 }, { "epoch": 2.06, "learning_rate": 7.942052850309732e-05, "loss": 2.9991, "step": 26245 }, { "epoch": 2.06, "learning_rate": 7.941974437387282e-05, "loss": 2.5678, "step": 26246 }, { "epoch": 2.06, "learning_rate": 7.941896024464833e-05, "loss": 2.7161, "step": 26247 }, { "epoch": 2.06, "learning_rate": 7.941817611542383e-05, "loss": 2.5994, "step": 26248 }, { "epoch": 2.06, "learning_rate": 7.941739198619933e-05, "loss": 2.7539, "step": 26249 }, { "epoch": 2.06, "learning_rate": 7.941660785697484e-05, "loss": 2.8, "step": 26250 }, { "epoch": 2.06, "learning_rate": 7.941582372775034e-05, "loss": 2.6808, "step": 26251 }, { "epoch": 2.06, "learning_rate": 7.941503959852584e-05, "loss": 2.7023, "step": 26252 }, { "epoch": 2.06, "learning_rate": 7.941425546930135e-05, "loss": 2.7704, "step": 26253 }, { "epoch": 2.06, "learning_rate": 7.941347134007685e-05, "loss": 2.7594, "step": 26254 }, { "epoch": 2.06, "learning_rate": 7.941268721085236e-05, "loss": 2.7864, "step": 26255 }, { "epoch": 2.06, "learning_rate": 7.941190308162786e-05, "loss": 2.764, "step": 26256 }, { "epoch": 2.06, "learning_rate": 7.941111895240336e-05, "loss": 2.8074, "step": 26257 }, { "epoch": 2.06, "learning_rate": 7.941033482317887e-05, "loss": 2.8298, "step": 26258 }, { "epoch": 2.06, "learning_rate": 7.940955069395437e-05, "loss": 2.7859, "step": 26259 }, { "epoch": 2.06, "learning_rate": 7.940876656472987e-05, "loss": 2.6765, "step": 26260 }, { "epoch": 2.06, "learning_rate": 7.940798243550538e-05, "loss": 3.055, "step": 26261 }, { "epoch": 2.06, "learning_rate": 7.940719830628088e-05, "loss": 2.6766, "step": 26262 }, { "epoch": 2.06, "learning_rate": 7.940641417705639e-05, "loss": 2.8689, "step": 26263 }, { "epoch": 2.06, "learning_rate": 7.940563004783189e-05, "loss": 2.7587, "step": 26264 }, { "epoch": 2.06, "learning_rate": 7.940484591860739e-05, "loss": 2.8376, "step": 26265 }, { "epoch": 2.06, "learning_rate": 7.940406178938288e-05, "loss": 2.8979, "step": 26266 }, { "epoch": 2.06, "learning_rate": 7.940327766015839e-05, "loss": 2.539, "step": 26267 }, { "epoch": 2.06, "learning_rate": 7.940249353093389e-05, "loss": 2.7502, "step": 26268 }, { "epoch": 2.06, "learning_rate": 7.940170940170941e-05, "loss": 2.7858, "step": 26269 }, { "epoch": 2.06, "learning_rate": 7.940092527248491e-05, "loss": 2.9145, "step": 26270 }, { "epoch": 2.06, "learning_rate": 7.940014114326041e-05, "loss": 2.6676, "step": 26271 }, { "epoch": 2.06, "learning_rate": 7.939935701403592e-05, "loss": 2.818, "step": 26272 }, { "epoch": 2.06, "learning_rate": 7.939857288481142e-05, "loss": 2.8354, "step": 26273 }, { "epoch": 2.06, "learning_rate": 7.939778875558693e-05, "loss": 2.8655, "step": 26274 }, { "epoch": 2.06, "learning_rate": 7.939700462636243e-05, "loss": 2.7706, "step": 26275 }, { "epoch": 2.06, "learning_rate": 7.939622049713793e-05, "loss": 2.9467, "step": 26276 }, { "epoch": 2.06, "learning_rate": 7.939543636791344e-05, "loss": 2.5151, "step": 26277 }, { "epoch": 2.06, "learning_rate": 7.939465223868894e-05, "loss": 2.8535, "step": 26278 }, { "epoch": 2.06, "learning_rate": 7.939386810946444e-05, "loss": 2.8719, "step": 26279 }, { "epoch": 2.06, "learning_rate": 7.939308398023995e-05, "loss": 2.7823, "step": 26280 }, { "epoch": 2.06, "learning_rate": 7.939229985101545e-05, "loss": 2.8072, "step": 26281 }, { "epoch": 2.06, "learning_rate": 7.939151572179096e-05, "loss": 2.8848, "step": 26282 }, { "epoch": 2.06, "learning_rate": 7.939073159256646e-05, "loss": 2.9293, "step": 26283 }, { "epoch": 2.06, "learning_rate": 7.938994746334196e-05, "loss": 2.8381, "step": 26284 }, { "epoch": 2.06, "learning_rate": 7.938916333411747e-05, "loss": 2.9712, "step": 26285 }, { "epoch": 2.06, "learning_rate": 7.938837920489297e-05, "loss": 2.8279, "step": 26286 }, { "epoch": 2.06, "learning_rate": 7.938759507566847e-05, "loss": 2.9087, "step": 26287 }, { "epoch": 2.06, "learning_rate": 7.938681094644398e-05, "loss": 2.7702, "step": 26288 }, { "epoch": 2.06, "learning_rate": 7.938602681721948e-05, "loss": 2.6379, "step": 26289 }, { "epoch": 2.06, "learning_rate": 7.938524268799498e-05, "loss": 2.8707, "step": 26290 }, { "epoch": 2.06, "learning_rate": 7.938445855877049e-05, "loss": 2.7103, "step": 26291 }, { "epoch": 2.06, "learning_rate": 7.938367442954599e-05, "loss": 2.6327, "step": 26292 }, { "epoch": 2.06, "learning_rate": 7.93828903003215e-05, "loss": 2.7845, "step": 26293 }, { "epoch": 2.06, "learning_rate": 7.9382106171097e-05, "loss": 2.7394, "step": 26294 }, { "epoch": 2.06, "learning_rate": 7.93813220418725e-05, "loss": 2.7958, "step": 26295 }, { "epoch": 2.06, "learning_rate": 7.938053791264801e-05, "loss": 3.0211, "step": 26296 }, { "epoch": 2.06, "learning_rate": 7.937975378342351e-05, "loss": 2.6858, "step": 26297 }, { "epoch": 2.06, "learning_rate": 7.937896965419901e-05, "loss": 2.7387, "step": 26298 }, { "epoch": 2.06, "learning_rate": 7.937818552497452e-05, "loss": 3.0153, "step": 26299 }, { "epoch": 2.06, "learning_rate": 7.937740139575004e-05, "loss": 2.5713, "step": 26300 }, { "epoch": 2.06, "learning_rate": 7.937661726652554e-05, "loss": 2.81, "step": 26301 }, { "epoch": 2.06, "learning_rate": 7.937583313730103e-05, "loss": 3.0434, "step": 26302 }, { "epoch": 2.06, "learning_rate": 7.937504900807653e-05, "loss": 2.7483, "step": 26303 }, { "epoch": 2.06, "learning_rate": 7.937426487885204e-05, "loss": 2.7502, "step": 26304 }, { "epoch": 2.06, "learning_rate": 7.937348074962754e-05, "loss": 2.8138, "step": 26305 }, { "epoch": 2.06, "learning_rate": 7.937269662040304e-05, "loss": 2.7841, "step": 26306 }, { "epoch": 2.06, "learning_rate": 7.937191249117855e-05, "loss": 2.9039, "step": 26307 }, { "epoch": 2.06, "learning_rate": 7.937112836195405e-05, "loss": 2.9398, "step": 26308 }, { "epoch": 2.06, "learning_rate": 7.937034423272955e-05, "loss": 2.8833, "step": 26309 }, { "epoch": 2.06, "learning_rate": 7.936956010350506e-05, "loss": 2.9929, "step": 26310 }, { "epoch": 2.06, "learning_rate": 7.936877597428056e-05, "loss": 2.8614, "step": 26311 }, { "epoch": 2.06, "learning_rate": 7.936799184505607e-05, "loss": 2.6813, "step": 26312 }, { "epoch": 2.06, "learning_rate": 7.936720771583157e-05, "loss": 2.7633, "step": 26313 }, { "epoch": 2.06, "learning_rate": 7.936642358660707e-05, "loss": 2.8709, "step": 26314 }, { "epoch": 2.06, "learning_rate": 7.936563945738258e-05, "loss": 2.9689, "step": 26315 }, { "epoch": 2.06, "learning_rate": 7.936485532815808e-05, "loss": 2.6404, "step": 26316 }, { "epoch": 2.06, "learning_rate": 7.936407119893358e-05, "loss": 2.7315, "step": 26317 }, { "epoch": 2.06, "learning_rate": 7.936328706970909e-05, "loss": 2.6943, "step": 26318 }, { "epoch": 2.06, "learning_rate": 7.936250294048459e-05, "loss": 2.7104, "step": 26319 }, { "epoch": 2.06, "learning_rate": 7.93617188112601e-05, "loss": 2.7578, "step": 26320 }, { "epoch": 2.06, "learning_rate": 7.93609346820356e-05, "loss": 2.7534, "step": 26321 }, { "epoch": 2.06, "learning_rate": 7.93601505528111e-05, "loss": 2.7332, "step": 26322 }, { "epoch": 2.06, "learning_rate": 7.93593664235866e-05, "loss": 2.8037, "step": 26323 }, { "epoch": 2.06, "learning_rate": 7.935858229436211e-05, "loss": 2.738, "step": 26324 }, { "epoch": 2.06, "learning_rate": 7.935779816513761e-05, "loss": 2.8214, "step": 26325 }, { "epoch": 2.06, "learning_rate": 7.935701403591313e-05, "loss": 2.7248, "step": 26326 }, { "epoch": 2.06, "learning_rate": 7.935622990668863e-05, "loss": 2.6691, "step": 26327 }, { "epoch": 2.06, "learning_rate": 7.935544577746414e-05, "loss": 2.8347, "step": 26328 }, { "epoch": 2.06, "learning_rate": 7.935466164823964e-05, "loss": 3.1163, "step": 26329 }, { "epoch": 2.06, "learning_rate": 7.935387751901515e-05, "loss": 2.7397, "step": 26330 }, { "epoch": 2.06, "learning_rate": 7.935309338979065e-05, "loss": 2.8944, "step": 26331 }, { "epoch": 2.06, "learning_rate": 7.935230926056615e-05, "loss": 2.9351, "step": 26332 }, { "epoch": 2.06, "learning_rate": 7.935152513134166e-05, "loss": 2.7332, "step": 26333 }, { "epoch": 2.06, "learning_rate": 7.935074100211716e-05, "loss": 2.8731, "step": 26334 }, { "epoch": 2.07, "learning_rate": 7.934995687289266e-05, "loss": 2.8916, "step": 26335 }, { "epoch": 2.07, "learning_rate": 7.934917274366817e-05, "loss": 2.7333, "step": 26336 }, { "epoch": 2.07, "learning_rate": 7.934838861444367e-05, "loss": 2.7096, "step": 26337 }, { "epoch": 2.07, "learning_rate": 7.934760448521916e-05, "loss": 2.9042, "step": 26338 }, { "epoch": 2.07, "learning_rate": 7.934682035599466e-05, "loss": 2.688, "step": 26339 }, { "epoch": 2.07, "learning_rate": 7.934603622677017e-05, "loss": 2.6799, "step": 26340 }, { "epoch": 2.07, "learning_rate": 7.934525209754567e-05, "loss": 2.7727, "step": 26341 }, { "epoch": 2.07, "learning_rate": 7.934446796832118e-05, "loss": 3.2704, "step": 26342 }, { "epoch": 2.07, "learning_rate": 7.934368383909668e-05, "loss": 2.8606, "step": 26343 }, { "epoch": 2.07, "learning_rate": 7.934289970987218e-05, "loss": 3.0177, "step": 26344 }, { "epoch": 2.07, "learning_rate": 7.934211558064769e-05, "loss": 2.6634, "step": 26345 }, { "epoch": 2.07, "learning_rate": 7.934133145142319e-05, "loss": 2.8423, "step": 26346 }, { "epoch": 2.07, "learning_rate": 7.93405473221987e-05, "loss": 2.8787, "step": 26347 }, { "epoch": 2.07, "learning_rate": 7.93397631929742e-05, "loss": 2.9108, "step": 26348 }, { "epoch": 2.07, "learning_rate": 7.93389790637497e-05, "loss": 2.8312, "step": 26349 }, { "epoch": 2.07, "learning_rate": 7.93381949345252e-05, "loss": 2.7983, "step": 26350 }, { "epoch": 2.07, "learning_rate": 7.933741080530072e-05, "loss": 2.7564, "step": 26351 }, { "epoch": 2.07, "learning_rate": 7.933662667607623e-05, "loss": 2.9897, "step": 26352 }, { "epoch": 2.07, "learning_rate": 7.933584254685173e-05, "loss": 2.9475, "step": 26353 }, { "epoch": 2.07, "learning_rate": 7.933505841762723e-05, "loss": 2.8223, "step": 26354 }, { "epoch": 2.07, "learning_rate": 7.933427428840274e-05, "loss": 2.5953, "step": 26355 }, { "epoch": 2.07, "learning_rate": 7.933349015917824e-05, "loss": 2.7499, "step": 26356 }, { "epoch": 2.07, "learning_rate": 7.933270602995374e-05, "loss": 2.6094, "step": 26357 }, { "epoch": 2.07, "learning_rate": 7.933192190072925e-05, "loss": 3.0037, "step": 26358 }, { "epoch": 2.07, "learning_rate": 7.933113777150475e-05, "loss": 3.0488, "step": 26359 }, { "epoch": 2.07, "learning_rate": 7.933035364228026e-05, "loss": 3.0149, "step": 26360 }, { "epoch": 2.07, "learning_rate": 7.932956951305576e-05, "loss": 2.7543, "step": 26361 }, { "epoch": 2.07, "learning_rate": 7.932878538383126e-05, "loss": 2.9212, "step": 26362 }, { "epoch": 2.07, "learning_rate": 7.932800125460677e-05, "loss": 2.9062, "step": 26363 }, { "epoch": 2.07, "learning_rate": 7.932721712538227e-05, "loss": 2.6845, "step": 26364 }, { "epoch": 2.07, "learning_rate": 7.932643299615777e-05, "loss": 2.7372, "step": 26365 }, { "epoch": 2.07, "learning_rate": 7.932564886693328e-05, "loss": 2.6103, "step": 26366 }, { "epoch": 2.07, "learning_rate": 7.932486473770878e-05, "loss": 2.6823, "step": 26367 }, { "epoch": 2.07, "learning_rate": 7.932408060848429e-05, "loss": 2.8151, "step": 26368 }, { "epoch": 2.07, "learning_rate": 7.932329647925979e-05, "loss": 2.769, "step": 26369 }, { "epoch": 2.07, "learning_rate": 7.932251235003529e-05, "loss": 2.7382, "step": 26370 }, { "epoch": 2.07, "learning_rate": 7.93217282208108e-05, "loss": 2.9513, "step": 26371 }, { "epoch": 2.07, "learning_rate": 7.93209440915863e-05, "loss": 2.7986, "step": 26372 }, { "epoch": 2.07, "learning_rate": 7.93201599623618e-05, "loss": 2.8601, "step": 26373 }, { "epoch": 2.07, "learning_rate": 7.93193758331373e-05, "loss": 2.6013, "step": 26374 }, { "epoch": 2.07, "learning_rate": 7.93185917039128e-05, "loss": 2.6084, "step": 26375 }, { "epoch": 2.07, "learning_rate": 7.93178075746883e-05, "loss": 3.0745, "step": 26376 }, { "epoch": 2.07, "learning_rate": 7.931702344546382e-05, "loss": 2.7221, "step": 26377 }, { "epoch": 2.07, "learning_rate": 7.931623931623932e-05, "loss": 2.8237, "step": 26378 }, { "epoch": 2.07, "learning_rate": 7.931545518701483e-05, "loss": 3.0265, "step": 26379 }, { "epoch": 2.07, "learning_rate": 7.931467105779033e-05, "loss": 2.8172, "step": 26380 }, { "epoch": 2.07, "learning_rate": 7.931388692856583e-05, "loss": 2.8647, "step": 26381 }, { "epoch": 2.07, "learning_rate": 7.931310279934134e-05, "loss": 2.7161, "step": 26382 }, { "epoch": 2.07, "learning_rate": 7.931231867011684e-05, "loss": 2.7545, "step": 26383 }, { "epoch": 2.07, "learning_rate": 7.931153454089234e-05, "loss": 2.8044, "step": 26384 }, { "epoch": 2.07, "learning_rate": 7.931075041166785e-05, "loss": 2.924, "step": 26385 }, { "epoch": 2.07, "learning_rate": 7.930996628244335e-05, "loss": 2.5578, "step": 26386 }, { "epoch": 2.07, "learning_rate": 7.930918215321886e-05, "loss": 2.6171, "step": 26387 }, { "epoch": 2.07, "learning_rate": 7.930839802399436e-05, "loss": 2.7496, "step": 26388 }, { "epoch": 2.07, "learning_rate": 7.930761389476986e-05, "loss": 2.7509, "step": 26389 }, { "epoch": 2.07, "learning_rate": 7.930682976554537e-05, "loss": 2.8758, "step": 26390 }, { "epoch": 2.07, "learning_rate": 7.930604563632087e-05, "loss": 2.6935, "step": 26391 }, { "epoch": 2.07, "learning_rate": 7.930526150709637e-05, "loss": 2.7111, "step": 26392 }, { "epoch": 2.07, "learning_rate": 7.930447737787188e-05, "loss": 2.9327, "step": 26393 }, { "epoch": 2.07, "learning_rate": 7.930369324864738e-05, "loss": 2.5849, "step": 26394 }, { "epoch": 2.07, "learning_rate": 7.930290911942288e-05, "loss": 2.6645, "step": 26395 }, { "epoch": 2.07, "learning_rate": 7.930212499019839e-05, "loss": 2.7669, "step": 26396 }, { "epoch": 2.07, "learning_rate": 7.930134086097389e-05, "loss": 2.5577, "step": 26397 }, { "epoch": 2.07, "learning_rate": 7.93005567317494e-05, "loss": 2.7389, "step": 26398 }, { "epoch": 2.07, "learning_rate": 7.92997726025249e-05, "loss": 2.7596, "step": 26399 }, { "epoch": 2.07, "learning_rate": 7.92989884733004e-05, "loss": 2.748, "step": 26400 }, { "epoch": 2.07, "learning_rate": 7.92982043440759e-05, "loss": 2.6303, "step": 26401 }, { "epoch": 2.07, "learning_rate": 7.929742021485141e-05, "loss": 2.812, "step": 26402 }, { "epoch": 2.07, "learning_rate": 7.929663608562691e-05, "loss": 2.8923, "step": 26403 }, { "epoch": 2.07, "learning_rate": 7.929585195640242e-05, "loss": 2.7229, "step": 26404 }, { "epoch": 2.07, "learning_rate": 7.929506782717792e-05, "loss": 2.7461, "step": 26405 }, { "epoch": 2.07, "learning_rate": 7.929428369795342e-05, "loss": 2.8764, "step": 26406 }, { "epoch": 2.07, "learning_rate": 7.929349956872893e-05, "loss": 3.0349, "step": 26407 }, { "epoch": 2.07, "learning_rate": 7.929271543950445e-05, "loss": 2.818, "step": 26408 }, { "epoch": 2.07, "learning_rate": 7.929193131027995e-05, "loss": 3.0162, "step": 26409 }, { "epoch": 2.07, "learning_rate": 7.929114718105544e-05, "loss": 2.9342, "step": 26410 }, { "epoch": 2.07, "learning_rate": 7.929036305183094e-05, "loss": 2.71, "step": 26411 }, { "epoch": 2.07, "learning_rate": 7.928957892260645e-05, "loss": 2.784, "step": 26412 }, { "epoch": 2.07, "learning_rate": 7.928879479338195e-05, "loss": 2.7439, "step": 26413 }, { "epoch": 2.07, "learning_rate": 7.928801066415745e-05, "loss": 3.0889, "step": 26414 }, { "epoch": 2.07, "learning_rate": 7.928722653493296e-05, "loss": 3.078, "step": 26415 }, { "epoch": 2.07, "learning_rate": 7.928644240570846e-05, "loss": 2.8187, "step": 26416 }, { "epoch": 2.07, "learning_rate": 7.928565827648397e-05, "loss": 2.7654, "step": 26417 }, { "epoch": 2.07, "learning_rate": 7.928487414725947e-05, "loss": 2.7436, "step": 26418 }, { "epoch": 2.07, "learning_rate": 7.928409001803497e-05, "loss": 2.9135, "step": 26419 }, { "epoch": 2.07, "learning_rate": 7.928330588881048e-05, "loss": 2.8378, "step": 26420 }, { "epoch": 2.07, "learning_rate": 7.928252175958598e-05, "loss": 2.5912, "step": 26421 }, { "epoch": 2.07, "learning_rate": 7.928173763036148e-05, "loss": 3.0847, "step": 26422 }, { "epoch": 2.07, "learning_rate": 7.928095350113699e-05, "loss": 2.9801, "step": 26423 }, { "epoch": 2.07, "learning_rate": 7.928016937191249e-05, "loss": 2.8255, "step": 26424 }, { "epoch": 2.07, "learning_rate": 7.9279385242688e-05, "loss": 2.8633, "step": 26425 }, { "epoch": 2.07, "learning_rate": 7.92786011134635e-05, "loss": 2.6694, "step": 26426 }, { "epoch": 2.07, "learning_rate": 7.9277816984239e-05, "loss": 2.696, "step": 26427 }, { "epoch": 2.07, "learning_rate": 7.92770328550145e-05, "loss": 2.7145, "step": 26428 }, { "epoch": 2.07, "learning_rate": 7.927624872579001e-05, "loss": 2.7075, "step": 26429 }, { "epoch": 2.07, "learning_rate": 7.927546459656551e-05, "loss": 2.5901, "step": 26430 }, { "epoch": 2.07, "learning_rate": 7.927468046734102e-05, "loss": 2.4698, "step": 26431 }, { "epoch": 2.07, "learning_rate": 7.927389633811652e-05, "loss": 2.8941, "step": 26432 }, { "epoch": 2.07, "learning_rate": 7.927311220889202e-05, "loss": 2.8115, "step": 26433 }, { "epoch": 2.07, "learning_rate": 7.927232807966754e-05, "loss": 2.7659, "step": 26434 }, { "epoch": 2.07, "learning_rate": 7.927154395044305e-05, "loss": 2.6305, "step": 26435 }, { "epoch": 2.07, "learning_rate": 7.927075982121855e-05, "loss": 2.6411, "step": 26436 }, { "epoch": 2.07, "learning_rate": 7.926997569199405e-05, "loss": 2.6311, "step": 26437 }, { "epoch": 2.07, "learning_rate": 7.926919156276956e-05, "loss": 2.6779, "step": 26438 }, { "epoch": 2.07, "learning_rate": 7.926840743354506e-05, "loss": 2.7736, "step": 26439 }, { "epoch": 2.07, "learning_rate": 7.926762330432056e-05, "loss": 2.8061, "step": 26440 }, { "epoch": 2.07, "learning_rate": 7.926683917509607e-05, "loss": 2.6918, "step": 26441 }, { "epoch": 2.07, "learning_rate": 7.926605504587157e-05, "loss": 2.7816, "step": 26442 }, { "epoch": 2.07, "learning_rate": 7.926527091664707e-05, "loss": 2.8197, "step": 26443 }, { "epoch": 2.07, "learning_rate": 7.926448678742258e-05, "loss": 2.7006, "step": 26444 }, { "epoch": 2.07, "learning_rate": 7.926370265819808e-05, "loss": 2.7699, "step": 26445 }, { "epoch": 2.07, "learning_rate": 7.926291852897357e-05, "loss": 2.7321, "step": 26446 }, { "epoch": 2.07, "learning_rate": 7.926213439974908e-05, "loss": 2.8621, "step": 26447 }, { "epoch": 2.07, "learning_rate": 7.926135027052458e-05, "loss": 2.6367, "step": 26448 }, { "epoch": 2.07, "learning_rate": 7.926056614130008e-05, "loss": 2.8458, "step": 26449 }, { "epoch": 2.07, "learning_rate": 7.925978201207559e-05, "loss": 2.6851, "step": 26450 }, { "epoch": 2.07, "learning_rate": 7.925899788285109e-05, "loss": 2.8121, "step": 26451 }, { "epoch": 2.07, "learning_rate": 7.92582137536266e-05, "loss": 2.8522, "step": 26452 }, { "epoch": 2.07, "learning_rate": 7.92574296244021e-05, "loss": 2.9285, "step": 26453 }, { "epoch": 2.07, "learning_rate": 7.92566454951776e-05, "loss": 2.9489, "step": 26454 }, { "epoch": 2.07, "learning_rate": 7.92558613659531e-05, "loss": 2.7081, "step": 26455 }, { "epoch": 2.07, "learning_rate": 7.925507723672861e-05, "loss": 2.8707, "step": 26456 }, { "epoch": 2.07, "learning_rate": 7.925429310750411e-05, "loss": 2.7713, "step": 26457 }, { "epoch": 2.07, "learning_rate": 7.925350897827962e-05, "loss": 2.7942, "step": 26458 }, { "epoch": 2.07, "learning_rate": 7.925272484905513e-05, "loss": 2.8664, "step": 26459 }, { "epoch": 2.07, "learning_rate": 7.925194071983064e-05, "loss": 2.7054, "step": 26460 }, { "epoch": 2.07, "learning_rate": 7.925115659060614e-05, "loss": 2.6141, "step": 26461 }, { "epoch": 2.07, "learning_rate": 7.925037246138164e-05, "loss": 2.9029, "step": 26462 }, { "epoch": 2.08, "learning_rate": 7.924958833215715e-05, "loss": 2.8393, "step": 26463 }, { "epoch": 2.08, "learning_rate": 7.924880420293265e-05, "loss": 2.5245, "step": 26464 }, { "epoch": 2.08, "learning_rate": 7.924802007370816e-05, "loss": 2.9229, "step": 26465 }, { "epoch": 2.08, "learning_rate": 7.924723594448366e-05, "loss": 2.9874, "step": 26466 }, { "epoch": 2.08, "learning_rate": 7.924645181525916e-05, "loss": 2.7566, "step": 26467 }, { "epoch": 2.08, "learning_rate": 7.924566768603467e-05, "loss": 2.6437, "step": 26468 }, { "epoch": 2.08, "learning_rate": 7.924488355681017e-05, "loss": 2.8301, "step": 26469 }, { "epoch": 2.08, "learning_rate": 7.924409942758567e-05, "loss": 2.8986, "step": 26470 }, { "epoch": 2.08, "learning_rate": 7.924331529836118e-05, "loss": 2.777, "step": 26471 }, { "epoch": 2.08, "learning_rate": 7.924253116913668e-05, "loss": 2.9872, "step": 26472 }, { "epoch": 2.08, "learning_rate": 7.924174703991218e-05, "loss": 2.7336, "step": 26473 }, { "epoch": 2.08, "learning_rate": 7.924096291068769e-05, "loss": 2.7376, "step": 26474 }, { "epoch": 2.08, "learning_rate": 7.924017878146319e-05, "loss": 2.6492, "step": 26475 }, { "epoch": 2.08, "learning_rate": 7.92393946522387e-05, "loss": 2.971, "step": 26476 }, { "epoch": 2.08, "learning_rate": 7.92386105230142e-05, "loss": 2.9444, "step": 26477 }, { "epoch": 2.08, "learning_rate": 7.92378263937897e-05, "loss": 2.9196, "step": 26478 }, { "epoch": 2.08, "learning_rate": 7.923704226456521e-05, "loss": 2.7754, "step": 26479 }, { "epoch": 2.08, "learning_rate": 7.923625813534071e-05, "loss": 2.652, "step": 26480 }, { "epoch": 2.08, "learning_rate": 7.923547400611621e-05, "loss": 2.749, "step": 26481 }, { "epoch": 2.08, "learning_rate": 7.92346898768917e-05, "loss": 2.6078, "step": 26482 }, { "epoch": 2.08, "learning_rate": 7.923390574766721e-05, "loss": 2.7609, "step": 26483 }, { "epoch": 2.08, "learning_rate": 7.923312161844271e-05, "loss": 2.9381, "step": 26484 }, { "epoch": 2.08, "learning_rate": 7.923233748921823e-05, "loss": 2.6428, "step": 26485 }, { "epoch": 2.08, "learning_rate": 7.923155335999373e-05, "loss": 3.0074, "step": 26486 }, { "epoch": 2.08, "learning_rate": 7.923076923076924e-05, "loss": 2.665, "step": 26487 }, { "epoch": 2.08, "learning_rate": 7.922998510154474e-05, "loss": 2.6509, "step": 26488 }, { "epoch": 2.08, "learning_rate": 7.922920097232024e-05, "loss": 2.8832, "step": 26489 }, { "epoch": 2.08, "learning_rate": 7.922841684309575e-05, "loss": 2.7561, "step": 26490 }, { "epoch": 2.08, "learning_rate": 7.922763271387125e-05, "loss": 2.6787, "step": 26491 }, { "epoch": 2.08, "learning_rate": 7.922684858464675e-05, "loss": 2.8199, "step": 26492 }, { "epoch": 2.08, "learning_rate": 7.922606445542226e-05, "loss": 2.8425, "step": 26493 }, { "epoch": 2.08, "learning_rate": 7.922528032619776e-05, "loss": 2.7249, "step": 26494 }, { "epoch": 2.08, "learning_rate": 7.922449619697327e-05, "loss": 2.907, "step": 26495 }, { "epoch": 2.08, "learning_rate": 7.922371206774877e-05, "loss": 2.8167, "step": 26496 }, { "epoch": 2.08, "learning_rate": 7.922292793852427e-05, "loss": 2.6776, "step": 26497 }, { "epoch": 2.08, "learning_rate": 7.922214380929978e-05, "loss": 2.8307, "step": 26498 }, { "epoch": 2.08, "learning_rate": 7.922135968007528e-05, "loss": 2.8996, "step": 26499 }, { "epoch": 2.08, "learning_rate": 7.922057555085078e-05, "loss": 2.7445, "step": 26500 }, { "epoch": 2.08, "learning_rate": 7.921979142162629e-05, "loss": 3.0352, "step": 26501 }, { "epoch": 2.08, "learning_rate": 7.921900729240179e-05, "loss": 2.8292, "step": 26502 }, { "epoch": 2.08, "learning_rate": 7.92182231631773e-05, "loss": 2.6731, "step": 26503 }, { "epoch": 2.08, "learning_rate": 7.92174390339528e-05, "loss": 2.7123, "step": 26504 }, { "epoch": 2.08, "learning_rate": 7.92166549047283e-05, "loss": 2.7339, "step": 26505 }, { "epoch": 2.08, "learning_rate": 7.92158707755038e-05, "loss": 2.748, "step": 26506 }, { "epoch": 2.08, "learning_rate": 7.921508664627931e-05, "loss": 2.8816, "step": 26507 }, { "epoch": 2.08, "learning_rate": 7.921430251705481e-05, "loss": 2.7375, "step": 26508 }, { "epoch": 2.08, "learning_rate": 7.921351838783032e-05, "loss": 2.7589, "step": 26509 }, { "epoch": 2.08, "learning_rate": 7.921273425860582e-05, "loss": 2.9424, "step": 26510 }, { "epoch": 2.08, "learning_rate": 7.921195012938132e-05, "loss": 3.0121, "step": 26511 }, { "epoch": 2.08, "learning_rate": 7.921116600015683e-05, "loss": 2.8271, "step": 26512 }, { "epoch": 2.08, "learning_rate": 7.921038187093233e-05, "loss": 2.7498, "step": 26513 }, { "epoch": 2.08, "learning_rate": 7.920959774170784e-05, "loss": 2.7368, "step": 26514 }, { "epoch": 2.08, "learning_rate": 7.920881361248334e-05, "loss": 2.646, "step": 26515 }, { "epoch": 2.08, "learning_rate": 7.920802948325886e-05, "loss": 2.8449, "step": 26516 }, { "epoch": 2.08, "learning_rate": 7.920724535403436e-05, "loss": 2.8021, "step": 26517 }, { "epoch": 2.08, "learning_rate": 7.920646122480985e-05, "loss": 2.819, "step": 26518 }, { "epoch": 2.08, "learning_rate": 7.920567709558535e-05, "loss": 3.0008, "step": 26519 }, { "epoch": 2.08, "learning_rate": 7.920489296636086e-05, "loss": 2.8122, "step": 26520 }, { "epoch": 2.08, "learning_rate": 7.920410883713636e-05, "loss": 2.9947, "step": 26521 }, { "epoch": 2.08, "learning_rate": 7.920332470791187e-05, "loss": 2.6091, "step": 26522 }, { "epoch": 2.08, "learning_rate": 7.920254057868737e-05, "loss": 2.8558, "step": 26523 }, { "epoch": 2.08, "learning_rate": 7.920175644946287e-05, "loss": 2.894, "step": 26524 }, { "epoch": 2.08, "learning_rate": 7.920097232023838e-05, "loss": 2.8102, "step": 26525 }, { "epoch": 2.08, "learning_rate": 7.920018819101388e-05, "loss": 2.8948, "step": 26526 }, { "epoch": 2.08, "learning_rate": 7.919940406178938e-05, "loss": 2.6787, "step": 26527 }, { "epoch": 2.08, "learning_rate": 7.919861993256489e-05, "loss": 2.9134, "step": 26528 }, { "epoch": 2.08, "learning_rate": 7.919783580334039e-05, "loss": 2.7855, "step": 26529 }, { "epoch": 2.08, "learning_rate": 7.91970516741159e-05, "loss": 2.6874, "step": 26530 }, { "epoch": 2.08, "learning_rate": 7.91962675448914e-05, "loss": 2.9044, "step": 26531 }, { "epoch": 2.08, "learning_rate": 7.91954834156669e-05, "loss": 2.9397, "step": 26532 }, { "epoch": 2.08, "learning_rate": 7.91946992864424e-05, "loss": 2.7919, "step": 26533 }, { "epoch": 2.08, "learning_rate": 7.919391515721791e-05, "loss": 2.6632, "step": 26534 }, { "epoch": 2.08, "learning_rate": 7.919313102799341e-05, "loss": 2.6799, "step": 26535 }, { "epoch": 2.08, "learning_rate": 7.919234689876892e-05, "loss": 2.8567, "step": 26536 }, { "epoch": 2.08, "learning_rate": 7.919156276954442e-05, "loss": 2.7669, "step": 26537 }, { "epoch": 2.08, "learning_rate": 7.919077864031992e-05, "loss": 2.7651, "step": 26538 }, { "epoch": 2.08, "learning_rate": 7.918999451109543e-05, "loss": 2.9135, "step": 26539 }, { "epoch": 2.08, "learning_rate": 7.918921038187093e-05, "loss": 2.6905, "step": 26540 }, { "epoch": 2.08, "learning_rate": 7.918842625264643e-05, "loss": 2.9381, "step": 26541 }, { "epoch": 2.08, "learning_rate": 7.918764212342195e-05, "loss": 2.624, "step": 26542 }, { "epoch": 2.08, "learning_rate": 7.918685799419746e-05, "loss": 2.6879, "step": 26543 }, { "epoch": 2.08, "learning_rate": 7.918607386497296e-05, "loss": 2.9036, "step": 26544 }, { "epoch": 2.08, "learning_rate": 7.918528973574846e-05, "loss": 2.763, "step": 26545 }, { "epoch": 2.08, "learning_rate": 7.918450560652397e-05, "loss": 2.9497, "step": 26546 }, { "epoch": 2.08, "learning_rate": 7.918372147729947e-05, "loss": 2.9102, "step": 26547 }, { "epoch": 2.08, "learning_rate": 7.918293734807497e-05, "loss": 2.7796, "step": 26548 }, { "epoch": 2.08, "learning_rate": 7.918215321885048e-05, "loss": 2.9616, "step": 26549 }, { "epoch": 2.08, "learning_rate": 7.918136908962598e-05, "loss": 2.9753, "step": 26550 }, { "epoch": 2.08, "learning_rate": 7.918058496040149e-05, "loss": 2.8841, "step": 26551 }, { "epoch": 2.08, "learning_rate": 7.917980083117699e-05, "loss": 2.8307, "step": 26552 }, { "epoch": 2.08, "learning_rate": 7.917901670195249e-05, "loss": 2.7926, "step": 26553 }, { "epoch": 2.08, "learning_rate": 7.917823257272798e-05, "loss": 2.8828, "step": 26554 }, { "epoch": 2.08, "learning_rate": 7.917744844350349e-05, "loss": 2.8947, "step": 26555 }, { "epoch": 2.08, "learning_rate": 7.917666431427899e-05, "loss": 2.8104, "step": 26556 }, { "epoch": 2.08, "learning_rate": 7.91758801850545e-05, "loss": 2.7972, "step": 26557 }, { "epoch": 2.08, "learning_rate": 7.917509605583e-05, "loss": 2.9152, "step": 26558 }, { "epoch": 2.08, "learning_rate": 7.91743119266055e-05, "loss": 2.8801, "step": 26559 }, { "epoch": 2.08, "learning_rate": 7.9173527797381e-05, "loss": 2.6965, "step": 26560 }, { "epoch": 2.08, "learning_rate": 7.917274366815651e-05, "loss": 2.7011, "step": 26561 }, { "epoch": 2.08, "learning_rate": 7.917195953893201e-05, "loss": 2.8998, "step": 26562 }, { "epoch": 2.08, "learning_rate": 7.917117540970752e-05, "loss": 2.8017, "step": 26563 }, { "epoch": 2.08, "learning_rate": 7.917039128048302e-05, "loss": 2.6711, "step": 26564 }, { "epoch": 2.08, "learning_rate": 7.916960715125852e-05, "loss": 2.6786, "step": 26565 }, { "epoch": 2.08, "learning_rate": 7.916882302203403e-05, "loss": 2.7951, "step": 26566 }, { "epoch": 2.08, "learning_rate": 7.916803889280954e-05, "loss": 2.6483, "step": 26567 }, { "epoch": 2.08, "learning_rate": 7.916725476358505e-05, "loss": 2.6903, "step": 26568 }, { "epoch": 2.08, "learning_rate": 7.916647063436055e-05, "loss": 2.7103, "step": 26569 }, { "epoch": 2.08, "learning_rate": 7.916568650513606e-05, "loss": 2.9079, "step": 26570 }, { "epoch": 2.08, "learning_rate": 7.916490237591156e-05, "loss": 2.6325, "step": 26571 }, { "epoch": 2.08, "learning_rate": 7.916411824668706e-05, "loss": 2.6879, "step": 26572 }, { "epoch": 2.08, "learning_rate": 7.916333411746257e-05, "loss": 2.905, "step": 26573 }, { "epoch": 2.08, "learning_rate": 7.916254998823807e-05, "loss": 2.7504, "step": 26574 }, { "epoch": 2.08, "learning_rate": 7.916176585901357e-05, "loss": 2.6673, "step": 26575 }, { "epoch": 2.08, "learning_rate": 7.916098172978908e-05, "loss": 2.8789, "step": 26576 }, { "epoch": 2.08, "learning_rate": 7.916019760056458e-05, "loss": 2.6313, "step": 26577 }, { "epoch": 2.08, "learning_rate": 7.915941347134008e-05, "loss": 2.9246, "step": 26578 }, { "epoch": 2.08, "learning_rate": 7.915862934211559e-05, "loss": 2.7022, "step": 26579 }, { "epoch": 2.08, "learning_rate": 7.915784521289109e-05, "loss": 2.5911, "step": 26580 }, { "epoch": 2.08, "learning_rate": 7.91570610836666e-05, "loss": 3.1297, "step": 26581 }, { "epoch": 2.08, "learning_rate": 7.91562769544421e-05, "loss": 2.8519, "step": 26582 }, { "epoch": 2.08, "learning_rate": 7.91554928252176e-05, "loss": 2.7586, "step": 26583 }, { "epoch": 2.08, "learning_rate": 7.91547086959931e-05, "loss": 2.9437, "step": 26584 }, { "epoch": 2.08, "learning_rate": 7.915392456676861e-05, "loss": 2.8229, "step": 26585 }, { "epoch": 2.08, "learning_rate": 7.915314043754411e-05, "loss": 2.8723, "step": 26586 }, { "epoch": 2.08, "learning_rate": 7.915235630831962e-05, "loss": 2.812, "step": 26587 }, { "epoch": 2.08, "learning_rate": 7.915157217909512e-05, "loss": 2.6666, "step": 26588 }, { "epoch": 2.08, "learning_rate": 7.915078804987063e-05, "loss": 2.9041, "step": 26589 }, { "epoch": 2.08, "learning_rate": 7.915000392064612e-05, "loss": 2.6856, "step": 26590 }, { "epoch": 2.09, "learning_rate": 7.914921979142162e-05, "loss": 2.8245, "step": 26591 }, { "epoch": 2.09, "learning_rate": 7.914843566219712e-05, "loss": 2.7998, "step": 26592 }, { "epoch": 2.09, "learning_rate": 7.914765153297264e-05, "loss": 2.7669, "step": 26593 }, { "epoch": 2.09, "learning_rate": 7.914686740374814e-05, "loss": 2.7776, "step": 26594 }, { "epoch": 2.09, "learning_rate": 7.914608327452365e-05, "loss": 2.7035, "step": 26595 }, { "epoch": 2.09, "learning_rate": 7.914529914529915e-05, "loss": 2.8065, "step": 26596 }, { "epoch": 2.09, "learning_rate": 7.914451501607465e-05, "loss": 2.9113, "step": 26597 }, { "epoch": 2.09, "learning_rate": 7.914373088685016e-05, "loss": 2.7108, "step": 26598 }, { "epoch": 2.09, "learning_rate": 7.914294675762566e-05, "loss": 2.7627, "step": 26599 }, { "epoch": 2.09, "learning_rate": 7.914216262840117e-05, "loss": 2.7188, "step": 26600 }, { "epoch": 2.09, "learning_rate": 7.914137849917667e-05, "loss": 2.9252, "step": 26601 }, { "epoch": 2.09, "learning_rate": 7.914059436995217e-05, "loss": 2.6524, "step": 26602 }, { "epoch": 2.09, "learning_rate": 7.913981024072768e-05, "loss": 2.7752, "step": 26603 }, { "epoch": 2.09, "learning_rate": 7.913902611150318e-05, "loss": 2.861, "step": 26604 }, { "epoch": 2.09, "learning_rate": 7.913824198227868e-05, "loss": 2.8573, "step": 26605 }, { "epoch": 2.09, "learning_rate": 7.913745785305419e-05, "loss": 2.7751, "step": 26606 }, { "epoch": 2.09, "learning_rate": 7.913667372382969e-05, "loss": 2.9454, "step": 26607 }, { "epoch": 2.09, "learning_rate": 7.91358895946052e-05, "loss": 2.728, "step": 26608 }, { "epoch": 2.09, "learning_rate": 7.91351054653807e-05, "loss": 2.771, "step": 26609 }, { "epoch": 2.09, "learning_rate": 7.91343213361562e-05, "loss": 2.8915, "step": 26610 }, { "epoch": 2.09, "learning_rate": 7.91335372069317e-05, "loss": 2.8449, "step": 26611 }, { "epoch": 2.09, "learning_rate": 7.913275307770721e-05, "loss": 3.0109, "step": 26612 }, { "epoch": 2.09, "learning_rate": 7.913196894848271e-05, "loss": 2.7293, "step": 26613 }, { "epoch": 2.09, "learning_rate": 7.913118481925822e-05, "loss": 2.8067, "step": 26614 }, { "epoch": 2.09, "learning_rate": 7.913040069003372e-05, "loss": 2.7788, "step": 26615 }, { "epoch": 2.09, "learning_rate": 7.912961656080922e-05, "loss": 2.6147, "step": 26616 }, { "epoch": 2.09, "learning_rate": 7.912883243158473e-05, "loss": 2.7811, "step": 26617 }, { "epoch": 2.09, "learning_rate": 7.912804830236023e-05, "loss": 2.9817, "step": 26618 }, { "epoch": 2.09, "learning_rate": 7.912726417313574e-05, "loss": 2.9007, "step": 26619 }, { "epoch": 2.09, "learning_rate": 7.912648004391124e-05, "loss": 2.6837, "step": 26620 }, { "epoch": 2.09, "learning_rate": 7.912569591468674e-05, "loss": 2.9021, "step": 26621 }, { "epoch": 2.09, "learning_rate": 7.912491178546225e-05, "loss": 2.7249, "step": 26622 }, { "epoch": 2.09, "learning_rate": 7.912412765623775e-05, "loss": 2.7737, "step": 26623 }, { "epoch": 2.09, "learning_rate": 7.912334352701327e-05, "loss": 2.7751, "step": 26624 }, { "epoch": 2.09, "learning_rate": 7.912255939778877e-05, "loss": 2.7794, "step": 26625 }, { "epoch": 2.09, "learning_rate": 7.912177526856426e-05, "loss": 2.8192, "step": 26626 }, { "epoch": 2.09, "learning_rate": 7.912099113933976e-05, "loss": 2.645, "step": 26627 }, { "epoch": 2.09, "learning_rate": 7.912020701011527e-05, "loss": 2.7927, "step": 26628 }, { "epoch": 2.09, "learning_rate": 7.911942288089077e-05, "loss": 2.8242, "step": 26629 }, { "epoch": 2.09, "learning_rate": 7.911863875166628e-05, "loss": 3.0006, "step": 26630 }, { "epoch": 2.09, "learning_rate": 7.911785462244178e-05, "loss": 2.6497, "step": 26631 }, { "epoch": 2.09, "learning_rate": 7.911707049321728e-05, "loss": 2.936, "step": 26632 }, { "epoch": 2.09, "learning_rate": 7.911628636399279e-05, "loss": 2.8652, "step": 26633 }, { "epoch": 2.09, "learning_rate": 7.911550223476829e-05, "loss": 3.0817, "step": 26634 }, { "epoch": 2.09, "learning_rate": 7.91147181055438e-05, "loss": 2.8054, "step": 26635 }, { "epoch": 2.09, "learning_rate": 7.91139339763193e-05, "loss": 2.8097, "step": 26636 }, { "epoch": 2.09, "learning_rate": 7.91131498470948e-05, "loss": 2.7424, "step": 26637 }, { "epoch": 2.09, "learning_rate": 7.91123657178703e-05, "loss": 2.7744, "step": 26638 }, { "epoch": 2.09, "learning_rate": 7.911158158864581e-05, "loss": 2.6746, "step": 26639 }, { "epoch": 2.09, "learning_rate": 7.911079745942131e-05, "loss": 2.8603, "step": 26640 }, { "epoch": 2.09, "learning_rate": 7.911001333019682e-05, "loss": 2.7616, "step": 26641 }, { "epoch": 2.09, "learning_rate": 7.910922920097232e-05, "loss": 2.7288, "step": 26642 }, { "epoch": 2.09, "learning_rate": 7.910844507174782e-05, "loss": 2.576, "step": 26643 }, { "epoch": 2.09, "learning_rate": 7.910766094252333e-05, "loss": 2.7875, "step": 26644 }, { "epoch": 2.09, "learning_rate": 7.910687681329883e-05, "loss": 2.995, "step": 26645 }, { "epoch": 2.09, "learning_rate": 7.910609268407433e-05, "loss": 3.1122, "step": 26646 }, { "epoch": 2.09, "learning_rate": 7.910530855484984e-05, "loss": 2.6724, "step": 26647 }, { "epoch": 2.09, "learning_rate": 7.910452442562534e-05, "loss": 2.8912, "step": 26648 }, { "epoch": 2.09, "learning_rate": 7.910374029640085e-05, "loss": 2.6802, "step": 26649 }, { "epoch": 2.09, "learning_rate": 7.910295616717636e-05, "loss": 3.1431, "step": 26650 }, { "epoch": 2.09, "learning_rate": 7.910217203795187e-05, "loss": 2.8019, "step": 26651 }, { "epoch": 2.09, "learning_rate": 7.910138790872737e-05, "loss": 2.6773, "step": 26652 }, { "epoch": 2.09, "learning_rate": 7.910060377950287e-05, "loss": 2.7781, "step": 26653 }, { "epoch": 2.09, "learning_rate": 7.909981965027838e-05, "loss": 2.7293, "step": 26654 }, { "epoch": 2.09, "learning_rate": 7.909903552105388e-05, "loss": 2.7626, "step": 26655 }, { "epoch": 2.09, "learning_rate": 7.909825139182938e-05, "loss": 2.9452, "step": 26656 }, { "epoch": 2.09, "learning_rate": 7.909746726260489e-05, "loss": 2.7693, "step": 26657 }, { "epoch": 2.09, "learning_rate": 7.909668313338039e-05, "loss": 2.8882, "step": 26658 }, { "epoch": 2.09, "learning_rate": 7.90958990041559e-05, "loss": 2.7627, "step": 26659 }, { "epoch": 2.09, "learning_rate": 7.90951148749314e-05, "loss": 2.7989, "step": 26660 }, { "epoch": 2.09, "learning_rate": 7.90943307457069e-05, "loss": 2.8929, "step": 26661 }, { "epoch": 2.09, "learning_rate": 7.90935466164824e-05, "loss": 2.8265, "step": 26662 }, { "epoch": 2.09, "learning_rate": 7.90927624872579e-05, "loss": 2.821, "step": 26663 }, { "epoch": 2.09, "learning_rate": 7.90919783580334e-05, "loss": 2.7902, "step": 26664 }, { "epoch": 2.09, "learning_rate": 7.90911942288089e-05, "loss": 2.7029, "step": 26665 }, { "epoch": 2.09, "learning_rate": 7.909041009958441e-05, "loss": 2.8773, "step": 26666 }, { "epoch": 2.09, "learning_rate": 7.908962597035991e-05, "loss": 3.0405, "step": 26667 }, { "epoch": 2.09, "learning_rate": 7.908884184113542e-05, "loss": 2.8412, "step": 26668 }, { "epoch": 2.09, "learning_rate": 7.908805771191092e-05, "loss": 2.7728, "step": 26669 }, { "epoch": 2.09, "learning_rate": 7.908727358268642e-05, "loss": 2.8755, "step": 26670 }, { "epoch": 2.09, "learning_rate": 7.908648945346193e-05, "loss": 2.9006, "step": 26671 }, { "epoch": 2.09, "learning_rate": 7.908570532423743e-05, "loss": 2.9338, "step": 26672 }, { "epoch": 2.09, "learning_rate": 7.908492119501293e-05, "loss": 2.8078, "step": 26673 }, { "epoch": 2.09, "learning_rate": 7.908413706578844e-05, "loss": 2.8458, "step": 26674 }, { "epoch": 2.09, "learning_rate": 7.908335293656394e-05, "loss": 2.7854, "step": 26675 }, { "epoch": 2.09, "learning_rate": 7.908256880733946e-05, "loss": 2.7232, "step": 26676 }, { "epoch": 2.09, "learning_rate": 7.908178467811496e-05, "loss": 2.7105, "step": 26677 }, { "epoch": 2.09, "learning_rate": 7.908100054889047e-05, "loss": 2.8364, "step": 26678 }, { "epoch": 2.09, "learning_rate": 7.908021641966597e-05, "loss": 2.8928, "step": 26679 }, { "epoch": 2.09, "learning_rate": 7.907943229044147e-05, "loss": 2.8366, "step": 26680 }, { "epoch": 2.09, "learning_rate": 7.907864816121698e-05, "loss": 2.6826, "step": 26681 }, { "epoch": 2.09, "learning_rate": 7.907786403199248e-05, "loss": 2.7387, "step": 26682 }, { "epoch": 2.09, "learning_rate": 7.907707990276798e-05, "loss": 2.7712, "step": 26683 }, { "epoch": 2.09, "learning_rate": 7.907629577354349e-05, "loss": 2.7755, "step": 26684 }, { "epoch": 2.09, "learning_rate": 7.907551164431899e-05, "loss": 2.7421, "step": 26685 }, { "epoch": 2.09, "learning_rate": 7.90747275150945e-05, "loss": 3.051, "step": 26686 }, { "epoch": 2.09, "learning_rate": 7.907394338587e-05, "loss": 2.7385, "step": 26687 }, { "epoch": 2.09, "learning_rate": 7.90731592566455e-05, "loss": 2.7399, "step": 26688 }, { "epoch": 2.09, "learning_rate": 7.9072375127421e-05, "loss": 2.867, "step": 26689 }, { "epoch": 2.09, "learning_rate": 7.907159099819651e-05, "loss": 2.74, "step": 26690 }, { "epoch": 2.09, "learning_rate": 7.907080686897201e-05, "loss": 2.7284, "step": 26691 }, { "epoch": 2.09, "learning_rate": 7.907002273974752e-05, "loss": 2.7605, "step": 26692 }, { "epoch": 2.09, "learning_rate": 7.906923861052302e-05, "loss": 2.7843, "step": 26693 }, { "epoch": 2.09, "learning_rate": 7.906845448129852e-05, "loss": 2.7001, "step": 26694 }, { "epoch": 2.09, "learning_rate": 7.906767035207403e-05, "loss": 2.9302, "step": 26695 }, { "epoch": 2.09, "learning_rate": 7.906688622284953e-05, "loss": 2.7598, "step": 26696 }, { "epoch": 2.09, "learning_rate": 7.906610209362504e-05, "loss": 2.6797, "step": 26697 }, { "epoch": 2.09, "learning_rate": 7.906531796440053e-05, "loss": 2.7697, "step": 26698 }, { "epoch": 2.09, "learning_rate": 7.906453383517603e-05, "loss": 2.8034, "step": 26699 }, { "epoch": 2.09, "learning_rate": 7.906374970595153e-05, "loss": 2.7665, "step": 26700 }, { "epoch": 2.09, "learning_rate": 7.906296557672705e-05, "loss": 2.8208, "step": 26701 }, { "epoch": 2.09, "learning_rate": 7.906218144750255e-05, "loss": 2.7055, "step": 26702 }, { "epoch": 2.09, "learning_rate": 7.906139731827806e-05, "loss": 2.6134, "step": 26703 }, { "epoch": 2.09, "learning_rate": 7.906061318905356e-05, "loss": 2.7844, "step": 26704 }, { "epoch": 2.09, "learning_rate": 7.905982905982907e-05, "loss": 3.0252, "step": 26705 }, { "epoch": 2.09, "learning_rate": 7.905904493060457e-05, "loss": 3.0965, "step": 26706 }, { "epoch": 2.09, "learning_rate": 7.905826080138007e-05, "loss": 2.8343, "step": 26707 }, { "epoch": 2.09, "learning_rate": 7.905747667215558e-05, "loss": 2.7313, "step": 26708 }, { "epoch": 2.09, "learning_rate": 7.905669254293108e-05, "loss": 2.7707, "step": 26709 }, { "epoch": 2.09, "learning_rate": 7.905590841370658e-05, "loss": 2.8561, "step": 26710 }, { "epoch": 2.09, "learning_rate": 7.905512428448209e-05, "loss": 3.0021, "step": 26711 }, { "epoch": 2.09, "learning_rate": 7.905434015525759e-05, "loss": 2.6658, "step": 26712 }, { "epoch": 2.09, "learning_rate": 7.90535560260331e-05, "loss": 2.712, "step": 26713 }, { "epoch": 2.09, "learning_rate": 7.90527718968086e-05, "loss": 2.8611, "step": 26714 }, { "epoch": 2.09, "learning_rate": 7.90519877675841e-05, "loss": 2.6392, "step": 26715 }, { "epoch": 2.09, "learning_rate": 7.90512036383596e-05, "loss": 2.8003, "step": 26716 }, { "epoch": 2.09, "learning_rate": 7.905041950913511e-05, "loss": 2.6068, "step": 26717 }, { "epoch": 2.1, "learning_rate": 7.904963537991061e-05, "loss": 2.7076, "step": 26718 }, { "epoch": 2.1, "learning_rate": 7.904885125068612e-05, "loss": 2.8335, "step": 26719 }, { "epoch": 2.1, "learning_rate": 7.904806712146162e-05, "loss": 3.0137, "step": 26720 }, { "epoch": 2.1, "learning_rate": 7.904728299223712e-05, "loss": 2.6909, "step": 26721 }, { "epoch": 2.1, "learning_rate": 7.904649886301263e-05, "loss": 2.7465, "step": 26722 }, { "epoch": 2.1, "learning_rate": 7.904571473378813e-05, "loss": 2.7347, "step": 26723 }, { "epoch": 2.1, "learning_rate": 7.904493060456364e-05, "loss": 2.8826, "step": 26724 }, { "epoch": 2.1, "learning_rate": 7.904414647533914e-05, "loss": 2.772, "step": 26725 }, { "epoch": 2.1, "learning_rate": 7.904336234611464e-05, "loss": 2.7418, "step": 26726 }, { "epoch": 2.1, "learning_rate": 7.904257821689015e-05, "loss": 2.9412, "step": 26727 }, { "epoch": 2.1, "learning_rate": 7.904179408766565e-05, "loss": 2.8559, "step": 26728 }, { "epoch": 2.1, "learning_rate": 7.904100995844115e-05, "loss": 2.7034, "step": 26729 }, { "epoch": 2.1, "learning_rate": 7.904022582921666e-05, "loss": 2.9485, "step": 26730 }, { "epoch": 2.1, "learning_rate": 7.903944169999216e-05, "loss": 2.6275, "step": 26731 }, { "epoch": 2.1, "learning_rate": 7.903865757076768e-05, "loss": 2.7614, "step": 26732 }, { "epoch": 2.1, "learning_rate": 7.903787344154318e-05, "loss": 2.7864, "step": 26733 }, { "epoch": 2.1, "learning_rate": 7.903708931231867e-05, "loss": 2.7681, "step": 26734 }, { "epoch": 2.1, "learning_rate": 7.903630518309418e-05, "loss": 2.7022, "step": 26735 }, { "epoch": 2.1, "learning_rate": 7.903552105386968e-05, "loss": 2.8494, "step": 26736 }, { "epoch": 2.1, "learning_rate": 7.903473692464518e-05, "loss": 2.7325, "step": 26737 }, { "epoch": 2.1, "learning_rate": 7.903395279542069e-05, "loss": 2.9429, "step": 26738 }, { "epoch": 2.1, "learning_rate": 7.903316866619619e-05, "loss": 2.8296, "step": 26739 }, { "epoch": 2.1, "learning_rate": 7.90323845369717e-05, "loss": 2.8398, "step": 26740 }, { "epoch": 2.1, "learning_rate": 7.90316004077472e-05, "loss": 2.696, "step": 26741 }, { "epoch": 2.1, "learning_rate": 7.90308162785227e-05, "loss": 2.8376, "step": 26742 }, { "epoch": 2.1, "learning_rate": 7.90300321492982e-05, "loss": 2.762, "step": 26743 }, { "epoch": 2.1, "learning_rate": 7.902924802007371e-05, "loss": 2.762, "step": 26744 }, { "epoch": 2.1, "learning_rate": 7.902846389084921e-05, "loss": 2.9228, "step": 26745 }, { "epoch": 2.1, "learning_rate": 7.902767976162472e-05, "loss": 2.7835, "step": 26746 }, { "epoch": 2.1, "learning_rate": 7.902689563240022e-05, "loss": 2.9123, "step": 26747 }, { "epoch": 2.1, "learning_rate": 7.902611150317572e-05, "loss": 2.8742, "step": 26748 }, { "epoch": 2.1, "learning_rate": 7.902532737395123e-05, "loss": 2.7628, "step": 26749 }, { "epoch": 2.1, "learning_rate": 7.902454324472673e-05, "loss": 2.7304, "step": 26750 }, { "epoch": 2.1, "learning_rate": 7.902375911550223e-05, "loss": 2.7456, "step": 26751 }, { "epoch": 2.1, "learning_rate": 7.902297498627774e-05, "loss": 2.7995, "step": 26752 }, { "epoch": 2.1, "learning_rate": 7.902219085705324e-05, "loss": 2.6851, "step": 26753 }, { "epoch": 2.1, "learning_rate": 7.902140672782875e-05, "loss": 2.6462, "step": 26754 }, { "epoch": 2.1, "learning_rate": 7.902062259860425e-05, "loss": 2.9286, "step": 26755 }, { "epoch": 2.1, "learning_rate": 7.901983846937975e-05, "loss": 3.0013, "step": 26756 }, { "epoch": 2.1, "learning_rate": 7.901905434015526e-05, "loss": 2.9512, "step": 26757 }, { "epoch": 2.1, "learning_rate": 7.901827021093077e-05, "loss": 2.7881, "step": 26758 }, { "epoch": 2.1, "learning_rate": 7.901748608170628e-05, "loss": 2.5972, "step": 26759 }, { "epoch": 2.1, "learning_rate": 7.901670195248178e-05, "loss": 2.9541, "step": 26760 }, { "epoch": 2.1, "learning_rate": 7.901591782325728e-05, "loss": 2.7842, "step": 26761 }, { "epoch": 2.1, "learning_rate": 7.901513369403279e-05, "loss": 2.9132, "step": 26762 }, { "epoch": 2.1, "learning_rate": 7.901434956480829e-05, "loss": 2.7552, "step": 26763 }, { "epoch": 2.1, "learning_rate": 7.90135654355838e-05, "loss": 2.7385, "step": 26764 }, { "epoch": 2.1, "learning_rate": 7.90127813063593e-05, "loss": 3.0013, "step": 26765 }, { "epoch": 2.1, "learning_rate": 7.90119971771348e-05, "loss": 2.8429, "step": 26766 }, { "epoch": 2.1, "learning_rate": 7.90112130479103e-05, "loss": 2.7173, "step": 26767 }, { "epoch": 2.1, "learning_rate": 7.901042891868581e-05, "loss": 2.8797, "step": 26768 }, { "epoch": 2.1, "learning_rate": 7.900964478946131e-05, "loss": 2.7581, "step": 26769 }, { "epoch": 2.1, "learning_rate": 7.90088606602368e-05, "loss": 2.6547, "step": 26770 }, { "epoch": 2.1, "learning_rate": 7.900807653101231e-05, "loss": 2.5657, "step": 26771 }, { "epoch": 2.1, "learning_rate": 7.900729240178781e-05, "loss": 2.9497, "step": 26772 }, { "epoch": 2.1, "learning_rate": 7.900650827256332e-05, "loss": 2.8003, "step": 26773 }, { "epoch": 2.1, "learning_rate": 7.900572414333882e-05, "loss": 2.6894, "step": 26774 }, { "epoch": 2.1, "learning_rate": 7.900494001411432e-05, "loss": 2.7549, "step": 26775 }, { "epoch": 2.1, "learning_rate": 7.900415588488983e-05, "loss": 3.0352, "step": 26776 }, { "epoch": 2.1, "learning_rate": 7.900337175566533e-05, "loss": 2.7588, "step": 26777 }, { "epoch": 2.1, "learning_rate": 7.900258762644083e-05, "loss": 2.6165, "step": 26778 }, { "epoch": 2.1, "learning_rate": 7.900180349721634e-05, "loss": 2.8838, "step": 26779 }, { "epoch": 2.1, "learning_rate": 7.900101936799184e-05, "loss": 2.7623, "step": 26780 }, { "epoch": 2.1, "learning_rate": 7.900023523876734e-05, "loss": 2.4798, "step": 26781 }, { "epoch": 2.1, "learning_rate": 7.899945110954285e-05, "loss": 2.7792, "step": 26782 }, { "epoch": 2.1, "learning_rate": 7.899866698031835e-05, "loss": 2.9489, "step": 26783 }, { "epoch": 2.1, "learning_rate": 7.899788285109387e-05, "loss": 2.7888, "step": 26784 }, { "epoch": 2.1, "learning_rate": 7.899709872186937e-05, "loss": 2.7788, "step": 26785 }, { "epoch": 2.1, "learning_rate": 7.899631459264488e-05, "loss": 2.7676, "step": 26786 }, { "epoch": 2.1, "learning_rate": 7.899553046342038e-05, "loss": 2.7143, "step": 26787 }, { "epoch": 2.1, "learning_rate": 7.899474633419588e-05, "loss": 2.7254, "step": 26788 }, { "epoch": 2.1, "learning_rate": 7.899396220497139e-05, "loss": 2.7371, "step": 26789 }, { "epoch": 2.1, "learning_rate": 7.899317807574689e-05, "loss": 2.9684, "step": 26790 }, { "epoch": 2.1, "learning_rate": 7.89923939465224e-05, "loss": 2.6102, "step": 26791 }, { "epoch": 2.1, "learning_rate": 7.89916098172979e-05, "loss": 2.6862, "step": 26792 }, { "epoch": 2.1, "learning_rate": 7.89908256880734e-05, "loss": 2.5586, "step": 26793 }, { "epoch": 2.1, "learning_rate": 7.89900415588489e-05, "loss": 2.8288, "step": 26794 }, { "epoch": 2.1, "learning_rate": 7.898925742962441e-05, "loss": 2.844, "step": 26795 }, { "epoch": 2.1, "learning_rate": 7.898847330039991e-05, "loss": 2.9595, "step": 26796 }, { "epoch": 2.1, "learning_rate": 7.898768917117542e-05, "loss": 2.8809, "step": 26797 }, { "epoch": 2.1, "learning_rate": 7.898690504195092e-05, "loss": 2.7547, "step": 26798 }, { "epoch": 2.1, "learning_rate": 7.898612091272642e-05, "loss": 2.8164, "step": 26799 }, { "epoch": 2.1, "learning_rate": 7.898533678350193e-05, "loss": 2.8998, "step": 26800 }, { "epoch": 2.1, "learning_rate": 7.898455265427743e-05, "loss": 2.9009, "step": 26801 }, { "epoch": 2.1, "learning_rate": 7.898376852505294e-05, "loss": 2.8979, "step": 26802 }, { "epoch": 2.1, "learning_rate": 7.898298439582844e-05, "loss": 2.8643, "step": 26803 }, { "epoch": 2.1, "learning_rate": 7.898220026660394e-05, "loss": 2.8304, "step": 26804 }, { "epoch": 2.1, "learning_rate": 7.898141613737945e-05, "loss": 2.7074, "step": 26805 }, { "epoch": 2.1, "learning_rate": 7.898063200815494e-05, "loss": 2.6609, "step": 26806 }, { "epoch": 2.1, "learning_rate": 7.897984787893044e-05, "loss": 3.0857, "step": 26807 }, { "epoch": 2.1, "learning_rate": 7.897906374970594e-05, "loss": 2.7941, "step": 26808 }, { "epoch": 2.1, "learning_rate": 7.897827962048146e-05, "loss": 2.6547, "step": 26809 }, { "epoch": 2.1, "learning_rate": 7.897749549125696e-05, "loss": 2.6533, "step": 26810 }, { "epoch": 2.1, "learning_rate": 7.897671136203247e-05, "loss": 2.7418, "step": 26811 }, { "epoch": 2.1, "learning_rate": 7.897592723280797e-05, "loss": 2.7115, "step": 26812 }, { "epoch": 2.1, "learning_rate": 7.897514310358348e-05, "loss": 2.9296, "step": 26813 }, { "epoch": 2.1, "learning_rate": 7.897435897435898e-05, "loss": 2.5963, "step": 26814 }, { "epoch": 2.1, "learning_rate": 7.897357484513448e-05, "loss": 2.6529, "step": 26815 }, { "epoch": 2.1, "learning_rate": 7.897279071590999e-05, "loss": 2.7503, "step": 26816 }, { "epoch": 2.1, "learning_rate": 7.897200658668549e-05, "loss": 2.5519, "step": 26817 }, { "epoch": 2.1, "learning_rate": 7.8971222457461e-05, "loss": 2.7443, "step": 26818 }, { "epoch": 2.1, "learning_rate": 7.89704383282365e-05, "loss": 2.6703, "step": 26819 }, { "epoch": 2.1, "learning_rate": 7.8969654199012e-05, "loss": 2.9021, "step": 26820 }, { "epoch": 2.1, "learning_rate": 7.89688700697875e-05, "loss": 2.8272, "step": 26821 }, { "epoch": 2.1, "learning_rate": 7.896808594056301e-05, "loss": 2.978, "step": 26822 }, { "epoch": 2.1, "learning_rate": 7.896730181133851e-05, "loss": 3.0384, "step": 26823 }, { "epoch": 2.1, "learning_rate": 7.896651768211402e-05, "loss": 2.9214, "step": 26824 }, { "epoch": 2.1, "learning_rate": 7.896573355288952e-05, "loss": 2.7574, "step": 26825 }, { "epoch": 2.1, "learning_rate": 7.896494942366502e-05, "loss": 2.8435, "step": 26826 }, { "epoch": 2.1, "learning_rate": 7.896416529444053e-05, "loss": 2.9129, "step": 26827 }, { "epoch": 2.1, "learning_rate": 7.896338116521603e-05, "loss": 2.9155, "step": 26828 }, { "epoch": 2.1, "learning_rate": 7.896259703599153e-05, "loss": 2.9819, "step": 26829 }, { "epoch": 2.1, "learning_rate": 7.896181290676704e-05, "loss": 2.8913, "step": 26830 }, { "epoch": 2.1, "learning_rate": 7.896102877754254e-05, "loss": 2.4143, "step": 26831 }, { "epoch": 2.1, "learning_rate": 7.896024464831805e-05, "loss": 2.7088, "step": 26832 }, { "epoch": 2.1, "learning_rate": 7.895946051909355e-05, "loss": 3.1394, "step": 26833 }, { "epoch": 2.1, "learning_rate": 7.895867638986905e-05, "loss": 2.8655, "step": 26834 }, { "epoch": 2.1, "learning_rate": 7.895789226064456e-05, "loss": 2.723, "step": 26835 }, { "epoch": 2.1, "learning_rate": 7.895710813142006e-05, "loss": 2.6627, "step": 26836 }, { "epoch": 2.1, "learning_rate": 7.895632400219556e-05, "loss": 2.9517, "step": 26837 }, { "epoch": 2.1, "learning_rate": 7.895553987297107e-05, "loss": 2.8365, "step": 26838 }, { "epoch": 2.1, "learning_rate": 7.895475574374657e-05, "loss": 2.8064, "step": 26839 }, { "epoch": 2.1, "learning_rate": 7.895397161452209e-05, "loss": 2.9281, "step": 26840 }, { "epoch": 2.1, "learning_rate": 7.895318748529759e-05, "loss": 2.8173, "step": 26841 }, { "epoch": 2.1, "learning_rate": 7.895240335607308e-05, "loss": 2.8632, "step": 26842 }, { "epoch": 2.1, "learning_rate": 7.895161922684859e-05, "loss": 2.8567, "step": 26843 }, { "epoch": 2.1, "learning_rate": 7.895083509762409e-05, "loss": 2.5675, "step": 26844 }, { "epoch": 2.1, "learning_rate": 7.89500509683996e-05, "loss": 2.7807, "step": 26845 }, { "epoch": 2.11, "learning_rate": 7.89492668391751e-05, "loss": 2.6888, "step": 26846 }, { "epoch": 2.11, "learning_rate": 7.89484827099506e-05, "loss": 2.933, "step": 26847 }, { "epoch": 2.11, "learning_rate": 7.89476985807261e-05, "loss": 2.6303, "step": 26848 }, { "epoch": 2.11, "learning_rate": 7.894691445150161e-05, "loss": 2.6307, "step": 26849 }, { "epoch": 2.11, "learning_rate": 7.894613032227711e-05, "loss": 2.6981, "step": 26850 }, { "epoch": 2.11, "learning_rate": 7.894534619305262e-05, "loss": 2.6493, "step": 26851 }, { "epoch": 2.11, "learning_rate": 7.894456206382812e-05, "loss": 2.9949, "step": 26852 }, { "epoch": 2.11, "learning_rate": 7.894377793460362e-05, "loss": 3.0234, "step": 26853 }, { "epoch": 2.11, "learning_rate": 7.894299380537913e-05, "loss": 2.7626, "step": 26854 }, { "epoch": 2.11, "learning_rate": 7.894220967615463e-05, "loss": 3.0012, "step": 26855 }, { "epoch": 2.11, "learning_rate": 7.894142554693013e-05, "loss": 2.844, "step": 26856 }, { "epoch": 2.11, "learning_rate": 7.894064141770564e-05, "loss": 2.8826, "step": 26857 }, { "epoch": 2.11, "learning_rate": 7.893985728848114e-05, "loss": 2.7097, "step": 26858 }, { "epoch": 2.11, "learning_rate": 7.893907315925665e-05, "loss": 2.9595, "step": 26859 }, { "epoch": 2.11, "learning_rate": 7.893828903003215e-05, "loss": 2.6804, "step": 26860 }, { "epoch": 2.11, "learning_rate": 7.893750490080765e-05, "loss": 2.8617, "step": 26861 }, { "epoch": 2.11, "learning_rate": 7.893672077158316e-05, "loss": 2.915, "step": 26862 }, { "epoch": 2.11, "learning_rate": 7.893593664235866e-05, "loss": 2.7559, "step": 26863 }, { "epoch": 2.11, "learning_rate": 7.893515251313416e-05, "loss": 2.7837, "step": 26864 }, { "epoch": 2.11, "learning_rate": 7.893436838390967e-05, "loss": 2.6515, "step": 26865 }, { "epoch": 2.11, "learning_rate": 7.893358425468518e-05, "loss": 2.897, "step": 26866 }, { "epoch": 2.11, "learning_rate": 7.893280012546069e-05, "loss": 2.9075, "step": 26867 }, { "epoch": 2.11, "learning_rate": 7.893201599623619e-05, "loss": 2.7971, "step": 26868 }, { "epoch": 2.11, "learning_rate": 7.89312318670117e-05, "loss": 2.8898, "step": 26869 }, { "epoch": 2.11, "learning_rate": 7.89304477377872e-05, "loss": 2.7939, "step": 26870 }, { "epoch": 2.11, "learning_rate": 7.89296636085627e-05, "loss": 2.8333, "step": 26871 }, { "epoch": 2.11, "learning_rate": 7.89288794793382e-05, "loss": 2.9011, "step": 26872 }, { "epoch": 2.11, "learning_rate": 7.892809535011371e-05, "loss": 2.7496, "step": 26873 }, { "epoch": 2.11, "learning_rate": 7.892731122088921e-05, "loss": 2.8734, "step": 26874 }, { "epoch": 2.11, "learning_rate": 7.892652709166472e-05, "loss": 2.9199, "step": 26875 }, { "epoch": 2.11, "learning_rate": 7.892574296244022e-05, "loss": 2.6252, "step": 26876 }, { "epoch": 2.11, "learning_rate": 7.892495883321572e-05, "loss": 2.739, "step": 26877 }, { "epoch": 2.11, "learning_rate": 7.892417470399121e-05, "loss": 3.1691, "step": 26878 }, { "epoch": 2.11, "learning_rate": 7.892339057476672e-05, "loss": 2.8157, "step": 26879 }, { "epoch": 2.11, "learning_rate": 7.892260644554222e-05, "loss": 2.775, "step": 26880 }, { "epoch": 2.11, "learning_rate": 7.892182231631773e-05, "loss": 2.821, "step": 26881 }, { "epoch": 2.11, "learning_rate": 7.892103818709323e-05, "loss": 3.0797, "step": 26882 }, { "epoch": 2.11, "learning_rate": 7.892025405786873e-05, "loss": 2.7653, "step": 26883 }, { "epoch": 2.11, "learning_rate": 7.891946992864424e-05, "loss": 3.1261, "step": 26884 }, { "epoch": 2.11, "learning_rate": 7.891868579941974e-05, "loss": 2.585, "step": 26885 }, { "epoch": 2.11, "learning_rate": 7.891790167019524e-05, "loss": 2.5793, "step": 26886 }, { "epoch": 2.11, "learning_rate": 7.891711754097075e-05, "loss": 2.7897, "step": 26887 }, { "epoch": 2.11, "learning_rate": 7.891633341174625e-05, "loss": 2.6333, "step": 26888 }, { "epoch": 2.11, "learning_rate": 7.891554928252176e-05, "loss": 2.8051, "step": 26889 }, { "epoch": 2.11, "learning_rate": 7.891476515329726e-05, "loss": 2.8739, "step": 26890 }, { "epoch": 2.11, "learning_rate": 7.891398102407276e-05, "loss": 2.9275, "step": 26891 }, { "epoch": 2.11, "learning_rate": 7.891319689484828e-05, "loss": 2.8329, "step": 26892 }, { "epoch": 2.11, "learning_rate": 7.891241276562378e-05, "loss": 2.9343, "step": 26893 }, { "epoch": 2.11, "learning_rate": 7.891162863639929e-05, "loss": 2.8763, "step": 26894 }, { "epoch": 2.11, "learning_rate": 7.891084450717479e-05, "loss": 2.8353, "step": 26895 }, { "epoch": 2.11, "learning_rate": 7.89100603779503e-05, "loss": 2.8709, "step": 26896 }, { "epoch": 2.11, "learning_rate": 7.89092762487258e-05, "loss": 2.7169, "step": 26897 }, { "epoch": 2.11, "learning_rate": 7.89084921195013e-05, "loss": 2.8119, "step": 26898 }, { "epoch": 2.11, "learning_rate": 7.89077079902768e-05, "loss": 2.7637, "step": 26899 }, { "epoch": 2.11, "learning_rate": 7.890692386105231e-05, "loss": 2.9406, "step": 26900 }, { "epoch": 2.11, "learning_rate": 7.890613973182781e-05, "loss": 2.7462, "step": 26901 }, { "epoch": 2.11, "learning_rate": 7.890535560260332e-05, "loss": 2.791, "step": 26902 }, { "epoch": 2.11, "learning_rate": 7.890457147337882e-05, "loss": 2.7443, "step": 26903 }, { "epoch": 2.11, "learning_rate": 7.890378734415432e-05, "loss": 2.9441, "step": 26904 }, { "epoch": 2.11, "learning_rate": 7.890300321492983e-05, "loss": 2.7093, "step": 26905 }, { "epoch": 2.11, "learning_rate": 7.890221908570533e-05, "loss": 2.7708, "step": 26906 }, { "epoch": 2.11, "learning_rate": 7.890143495648084e-05, "loss": 2.7345, "step": 26907 }, { "epoch": 2.11, "learning_rate": 7.890065082725634e-05, "loss": 2.9933, "step": 26908 }, { "epoch": 2.11, "learning_rate": 7.889986669803184e-05, "loss": 2.9334, "step": 26909 }, { "epoch": 2.11, "learning_rate": 7.889908256880735e-05, "loss": 2.6747, "step": 26910 }, { "epoch": 2.11, "learning_rate": 7.889829843958285e-05, "loss": 2.7554, "step": 26911 }, { "epoch": 2.11, "learning_rate": 7.889751431035835e-05, "loss": 2.7809, "step": 26912 }, { "epoch": 2.11, "learning_rate": 7.889673018113386e-05, "loss": 2.928, "step": 26913 }, { "epoch": 2.11, "learning_rate": 7.889594605190935e-05, "loss": 2.6475, "step": 26914 }, { "epoch": 2.11, "learning_rate": 7.889516192268485e-05, "loss": 2.8416, "step": 26915 }, { "epoch": 2.11, "learning_rate": 7.889437779346035e-05, "loss": 2.7841, "step": 26916 }, { "epoch": 2.11, "learning_rate": 7.889359366423587e-05, "loss": 2.725, "step": 26917 }, { "epoch": 2.11, "learning_rate": 7.889280953501138e-05, "loss": 3.0214, "step": 26918 }, { "epoch": 2.11, "learning_rate": 7.889202540578688e-05, "loss": 2.957, "step": 26919 }, { "epoch": 2.11, "learning_rate": 7.889124127656238e-05, "loss": 2.8596, "step": 26920 }, { "epoch": 2.11, "learning_rate": 7.889045714733789e-05, "loss": 2.8706, "step": 26921 }, { "epoch": 2.11, "learning_rate": 7.888967301811339e-05, "loss": 2.7984, "step": 26922 }, { "epoch": 2.11, "learning_rate": 7.88888888888889e-05, "loss": 2.9881, "step": 26923 }, { "epoch": 2.11, "learning_rate": 7.88881047596644e-05, "loss": 2.9421, "step": 26924 }, { "epoch": 2.11, "learning_rate": 7.88873206304399e-05, "loss": 2.8129, "step": 26925 }, { "epoch": 2.11, "learning_rate": 7.88865365012154e-05, "loss": 2.93, "step": 26926 }, { "epoch": 2.11, "learning_rate": 7.888575237199091e-05, "loss": 2.7295, "step": 26927 }, { "epoch": 2.11, "learning_rate": 7.888496824276641e-05, "loss": 2.8667, "step": 26928 }, { "epoch": 2.11, "learning_rate": 7.888418411354192e-05, "loss": 2.7999, "step": 26929 }, { "epoch": 2.11, "learning_rate": 7.888339998431742e-05, "loss": 2.7777, "step": 26930 }, { "epoch": 2.11, "learning_rate": 7.888261585509292e-05, "loss": 2.7901, "step": 26931 }, { "epoch": 2.11, "learning_rate": 7.888183172586843e-05, "loss": 2.8961, "step": 26932 }, { "epoch": 2.11, "learning_rate": 7.888104759664393e-05, "loss": 2.8627, "step": 26933 }, { "epoch": 2.11, "learning_rate": 7.888026346741943e-05, "loss": 3.0669, "step": 26934 }, { "epoch": 2.11, "learning_rate": 7.887947933819494e-05, "loss": 2.8152, "step": 26935 }, { "epoch": 2.11, "learning_rate": 7.887869520897044e-05, "loss": 2.6243, "step": 26936 }, { "epoch": 2.11, "learning_rate": 7.887791107974595e-05, "loss": 2.7586, "step": 26937 }, { "epoch": 2.11, "learning_rate": 7.887712695052145e-05, "loss": 2.7588, "step": 26938 }, { "epoch": 2.11, "learning_rate": 7.887634282129695e-05, "loss": 2.6143, "step": 26939 }, { "epoch": 2.11, "learning_rate": 7.887555869207246e-05, "loss": 2.8485, "step": 26940 }, { "epoch": 2.11, "learning_rate": 7.887477456284796e-05, "loss": 2.7882, "step": 26941 }, { "epoch": 2.11, "learning_rate": 7.887399043362346e-05, "loss": 2.8007, "step": 26942 }, { "epoch": 2.11, "learning_rate": 7.887320630439897e-05, "loss": 2.8938, "step": 26943 }, { "epoch": 2.11, "learning_rate": 7.887242217517447e-05, "loss": 2.8102, "step": 26944 }, { "epoch": 2.11, "learning_rate": 7.887163804594997e-05, "loss": 2.6341, "step": 26945 }, { "epoch": 2.11, "learning_rate": 7.887085391672548e-05, "loss": 2.8992, "step": 26946 }, { "epoch": 2.11, "learning_rate": 7.887006978750098e-05, "loss": 2.7342, "step": 26947 }, { "epoch": 2.11, "learning_rate": 7.886928565827649e-05, "loss": 2.7203, "step": 26948 }, { "epoch": 2.11, "learning_rate": 7.8868501529052e-05, "loss": 2.803, "step": 26949 }, { "epoch": 2.11, "learning_rate": 7.88677173998275e-05, "loss": 2.9061, "step": 26950 }, { "epoch": 2.11, "learning_rate": 7.8866933270603e-05, "loss": 2.7639, "step": 26951 }, { "epoch": 2.11, "learning_rate": 7.88661491413785e-05, "loss": 3.0648, "step": 26952 }, { "epoch": 2.11, "learning_rate": 7.8865365012154e-05, "loss": 2.7378, "step": 26953 }, { "epoch": 2.11, "learning_rate": 7.886458088292951e-05, "loss": 2.8437, "step": 26954 }, { "epoch": 2.11, "learning_rate": 7.886379675370501e-05, "loss": 2.6799, "step": 26955 }, { "epoch": 2.11, "learning_rate": 7.886301262448052e-05, "loss": 2.612, "step": 26956 }, { "epoch": 2.11, "learning_rate": 7.886222849525602e-05, "loss": 2.5696, "step": 26957 }, { "epoch": 2.11, "learning_rate": 7.886144436603152e-05, "loss": 2.8961, "step": 26958 }, { "epoch": 2.11, "learning_rate": 7.886066023680703e-05, "loss": 2.6883, "step": 26959 }, { "epoch": 2.11, "learning_rate": 7.885987610758253e-05, "loss": 2.715, "step": 26960 }, { "epoch": 2.11, "learning_rate": 7.885909197835803e-05, "loss": 2.8991, "step": 26961 }, { "epoch": 2.11, "learning_rate": 7.885830784913354e-05, "loss": 2.7774, "step": 26962 }, { "epoch": 2.11, "learning_rate": 7.885752371990904e-05, "loss": 2.6924, "step": 26963 }, { "epoch": 2.11, "learning_rate": 7.885673959068454e-05, "loss": 2.8328, "step": 26964 }, { "epoch": 2.11, "learning_rate": 7.885595546146005e-05, "loss": 2.6877, "step": 26965 }, { "epoch": 2.11, "learning_rate": 7.885517133223555e-05, "loss": 2.4564, "step": 26966 }, { "epoch": 2.11, "learning_rate": 7.885438720301106e-05, "loss": 2.9318, "step": 26967 }, { "epoch": 2.11, "learning_rate": 7.885360307378656e-05, "loss": 2.7928, "step": 26968 }, { "epoch": 2.11, "learning_rate": 7.885281894456206e-05, "loss": 2.75, "step": 26969 }, { "epoch": 2.11, "learning_rate": 7.885203481533757e-05, "loss": 3.0627, "step": 26970 }, { "epoch": 2.11, "learning_rate": 7.885125068611307e-05, "loss": 2.7538, "step": 26971 }, { "epoch": 2.11, "learning_rate": 7.885046655688857e-05, "loss": 2.6866, "step": 26972 }, { "epoch": 2.12, "learning_rate": 7.884968242766408e-05, "loss": 3.1337, "step": 26973 }, { "epoch": 2.12, "learning_rate": 7.88488982984396e-05, "loss": 2.948, "step": 26974 }, { "epoch": 2.12, "learning_rate": 7.88481141692151e-05, "loss": 3.0026, "step": 26975 }, { "epoch": 2.12, "learning_rate": 7.88473300399906e-05, "loss": 2.6694, "step": 26976 }, { "epoch": 2.12, "learning_rate": 7.88465459107661e-05, "loss": 2.8031, "step": 26977 }, { "epoch": 2.12, "learning_rate": 7.884576178154161e-05, "loss": 2.9756, "step": 26978 }, { "epoch": 2.12, "learning_rate": 7.884497765231711e-05, "loss": 3.0195, "step": 26979 }, { "epoch": 2.12, "learning_rate": 7.884419352309262e-05, "loss": 2.7807, "step": 26980 }, { "epoch": 2.12, "learning_rate": 7.884340939386812e-05, "loss": 2.9399, "step": 26981 }, { "epoch": 2.12, "learning_rate": 7.884262526464362e-05, "loss": 2.7161, "step": 26982 }, { "epoch": 2.12, "learning_rate": 7.884184113541913e-05, "loss": 2.8711, "step": 26983 }, { "epoch": 2.12, "learning_rate": 7.884105700619463e-05, "loss": 2.7425, "step": 26984 }, { "epoch": 2.12, "learning_rate": 7.884027287697014e-05, "loss": 2.7239, "step": 26985 }, { "epoch": 2.12, "learning_rate": 7.883948874774563e-05, "loss": 2.5406, "step": 26986 }, { "epoch": 2.12, "learning_rate": 7.883870461852113e-05, "loss": 2.8726, "step": 26987 }, { "epoch": 2.12, "learning_rate": 7.883792048929663e-05, "loss": 2.7965, "step": 26988 }, { "epoch": 2.12, "learning_rate": 7.883713636007214e-05, "loss": 2.7478, "step": 26989 }, { "epoch": 2.12, "learning_rate": 7.883635223084764e-05, "loss": 2.6574, "step": 26990 }, { "epoch": 2.12, "learning_rate": 7.883556810162314e-05, "loss": 2.816, "step": 26991 }, { "epoch": 2.12, "learning_rate": 7.883478397239865e-05, "loss": 2.9478, "step": 26992 }, { "epoch": 2.12, "learning_rate": 7.883399984317415e-05, "loss": 2.8925, "step": 26993 }, { "epoch": 2.12, "learning_rate": 7.883321571394966e-05, "loss": 2.8957, "step": 26994 }, { "epoch": 2.12, "learning_rate": 7.883243158472516e-05, "loss": 2.9656, "step": 26995 }, { "epoch": 2.12, "learning_rate": 7.883164745550066e-05, "loss": 2.7754, "step": 26996 }, { "epoch": 2.12, "learning_rate": 7.883086332627617e-05, "loss": 2.8094, "step": 26997 }, { "epoch": 2.12, "learning_rate": 7.883007919705167e-05, "loss": 2.6723, "step": 26998 }, { "epoch": 2.12, "learning_rate": 7.882929506782717e-05, "loss": 2.8346, "step": 26999 }, { "epoch": 2.12, "learning_rate": 7.882851093860269e-05, "loss": 2.7652, "step": 27000 }, { "epoch": 2.12, "eval_loss": 2.9066474437713623, "eval_runtime": 218.0377, "eval_samples_per_second": 51.973, "eval_steps_per_second": 3.252, "step": 27000 }, { "epoch": 2.12, "learning_rate": 7.88277268093782e-05, "loss": 2.7332, "step": 27001 }, { "epoch": 2.12, "learning_rate": 7.88269426801537e-05, "loss": 2.665, "step": 27002 }, { "epoch": 2.12, "learning_rate": 7.88261585509292e-05, "loss": 2.8497, "step": 27003 }, { "epoch": 2.12, "learning_rate": 7.88253744217047e-05, "loss": 2.6289, "step": 27004 }, { "epoch": 2.12, "learning_rate": 7.882459029248021e-05, "loss": 2.8812, "step": 27005 }, { "epoch": 2.12, "learning_rate": 7.882380616325571e-05, "loss": 2.6726, "step": 27006 }, { "epoch": 2.12, "learning_rate": 7.882302203403122e-05, "loss": 2.7567, "step": 27007 }, { "epoch": 2.12, "learning_rate": 7.882223790480672e-05, "loss": 2.571, "step": 27008 }, { "epoch": 2.12, "learning_rate": 7.882145377558222e-05, "loss": 3.1717, "step": 27009 }, { "epoch": 2.12, "learning_rate": 7.882066964635773e-05, "loss": 2.7151, "step": 27010 }, { "epoch": 2.12, "learning_rate": 7.881988551713323e-05, "loss": 2.7565, "step": 27011 }, { "epoch": 2.12, "learning_rate": 7.881910138790873e-05, "loss": 2.8867, "step": 27012 }, { "epoch": 2.12, "learning_rate": 7.881831725868424e-05, "loss": 2.643, "step": 27013 }, { "epoch": 2.12, "learning_rate": 7.881753312945974e-05, "loss": 2.815, "step": 27014 }, { "epoch": 2.12, "learning_rate": 7.881674900023525e-05, "loss": 2.7299, "step": 27015 }, { "epoch": 2.12, "learning_rate": 7.881596487101075e-05, "loss": 2.7351, "step": 27016 }, { "epoch": 2.12, "learning_rate": 7.881518074178625e-05, "loss": 2.9251, "step": 27017 }, { "epoch": 2.12, "learning_rate": 7.881439661256176e-05, "loss": 2.8806, "step": 27018 }, { "epoch": 2.12, "learning_rate": 7.881361248333726e-05, "loss": 2.8126, "step": 27019 }, { "epoch": 2.12, "learning_rate": 7.881282835411276e-05, "loss": 2.7603, "step": 27020 }, { "epoch": 2.12, "learning_rate": 7.881204422488827e-05, "loss": 2.9523, "step": 27021 }, { "epoch": 2.12, "learning_rate": 7.881126009566376e-05, "loss": 2.7383, "step": 27022 }, { "epoch": 2.12, "learning_rate": 7.881047596643926e-05, "loss": 2.7466, "step": 27023 }, { "epoch": 2.12, "learning_rate": 7.880969183721477e-05, "loss": 2.8041, "step": 27024 }, { "epoch": 2.12, "learning_rate": 7.880890770799028e-05, "loss": 2.8337, "step": 27025 }, { "epoch": 2.12, "learning_rate": 7.880812357876579e-05, "loss": 2.7512, "step": 27026 }, { "epoch": 2.12, "learning_rate": 7.880733944954129e-05, "loss": 2.7943, "step": 27027 }, { "epoch": 2.12, "learning_rate": 7.88065553203168e-05, "loss": 2.7762, "step": 27028 }, { "epoch": 2.12, "learning_rate": 7.88057711910923e-05, "loss": 3.308, "step": 27029 }, { "epoch": 2.12, "learning_rate": 7.88049870618678e-05, "loss": 2.7473, "step": 27030 }, { "epoch": 2.12, "learning_rate": 7.88042029326433e-05, "loss": 2.6426, "step": 27031 }, { "epoch": 2.12, "learning_rate": 7.880341880341881e-05, "loss": 3.0218, "step": 27032 }, { "epoch": 2.12, "learning_rate": 7.880263467419431e-05, "loss": 2.9107, "step": 27033 }, { "epoch": 2.12, "learning_rate": 7.880185054496982e-05, "loss": 2.6512, "step": 27034 }, { "epoch": 2.12, "learning_rate": 7.880106641574532e-05, "loss": 3.1499, "step": 27035 }, { "epoch": 2.12, "learning_rate": 7.880028228652082e-05, "loss": 2.8131, "step": 27036 }, { "epoch": 2.12, "learning_rate": 7.879949815729633e-05, "loss": 2.8607, "step": 27037 }, { "epoch": 2.12, "learning_rate": 7.879871402807183e-05, "loss": 3.0144, "step": 27038 }, { "epoch": 2.12, "learning_rate": 7.879792989884733e-05, "loss": 2.9002, "step": 27039 }, { "epoch": 2.12, "learning_rate": 7.879714576962284e-05, "loss": 2.9874, "step": 27040 }, { "epoch": 2.12, "learning_rate": 7.879636164039834e-05, "loss": 2.6543, "step": 27041 }, { "epoch": 2.12, "learning_rate": 7.879557751117385e-05, "loss": 2.8088, "step": 27042 }, { "epoch": 2.12, "learning_rate": 7.879479338194935e-05, "loss": 3.0435, "step": 27043 }, { "epoch": 2.12, "learning_rate": 7.879400925272485e-05, "loss": 2.8384, "step": 27044 }, { "epoch": 2.12, "learning_rate": 7.879322512350036e-05, "loss": 2.6321, "step": 27045 }, { "epoch": 2.12, "learning_rate": 7.879244099427586e-05, "loss": 2.71, "step": 27046 }, { "epoch": 2.12, "learning_rate": 7.879165686505136e-05, "loss": 2.7423, "step": 27047 }, { "epoch": 2.12, "learning_rate": 7.879087273582687e-05, "loss": 2.7939, "step": 27048 }, { "epoch": 2.12, "learning_rate": 7.879008860660237e-05, "loss": 2.6914, "step": 27049 }, { "epoch": 2.12, "learning_rate": 7.878930447737787e-05, "loss": 2.7154, "step": 27050 }, { "epoch": 2.12, "learning_rate": 7.878852034815338e-05, "loss": 2.9528, "step": 27051 }, { "epoch": 2.12, "learning_rate": 7.878773621892888e-05, "loss": 2.8239, "step": 27052 }, { "epoch": 2.12, "learning_rate": 7.878695208970439e-05, "loss": 2.748, "step": 27053 }, { "epoch": 2.12, "learning_rate": 7.878616796047989e-05, "loss": 3.0821, "step": 27054 }, { "epoch": 2.12, "learning_rate": 7.878538383125539e-05, "loss": 2.919, "step": 27055 }, { "epoch": 2.12, "learning_rate": 7.87845997020309e-05, "loss": 2.9107, "step": 27056 }, { "epoch": 2.12, "learning_rate": 7.878381557280641e-05, "loss": 2.8396, "step": 27057 }, { "epoch": 2.12, "learning_rate": 7.87830314435819e-05, "loss": 2.6958, "step": 27058 }, { "epoch": 2.12, "learning_rate": 7.878224731435741e-05, "loss": 2.7996, "step": 27059 }, { "epoch": 2.12, "learning_rate": 7.878146318513291e-05, "loss": 2.8662, "step": 27060 }, { "epoch": 2.12, "learning_rate": 7.878067905590842e-05, "loss": 3.0135, "step": 27061 }, { "epoch": 2.12, "learning_rate": 7.877989492668392e-05, "loss": 2.8571, "step": 27062 }, { "epoch": 2.12, "learning_rate": 7.877911079745942e-05, "loss": 2.8431, "step": 27063 }, { "epoch": 2.12, "learning_rate": 7.877832666823493e-05, "loss": 2.9637, "step": 27064 }, { "epoch": 2.12, "learning_rate": 7.877754253901043e-05, "loss": 2.7797, "step": 27065 }, { "epoch": 2.12, "learning_rate": 7.877675840978593e-05, "loss": 2.865, "step": 27066 }, { "epoch": 2.12, "learning_rate": 7.877597428056144e-05, "loss": 2.7716, "step": 27067 }, { "epoch": 2.12, "learning_rate": 7.877519015133694e-05, "loss": 2.8081, "step": 27068 }, { "epoch": 2.12, "learning_rate": 7.877440602211244e-05, "loss": 2.8204, "step": 27069 }, { "epoch": 2.12, "learning_rate": 7.877362189288795e-05, "loss": 2.8714, "step": 27070 }, { "epoch": 2.12, "learning_rate": 7.877283776366345e-05, "loss": 2.8598, "step": 27071 }, { "epoch": 2.12, "learning_rate": 7.877205363443896e-05, "loss": 2.5259, "step": 27072 }, { "epoch": 2.12, "learning_rate": 7.877126950521446e-05, "loss": 2.7284, "step": 27073 }, { "epoch": 2.12, "learning_rate": 7.877048537598996e-05, "loss": 2.8366, "step": 27074 }, { "epoch": 2.12, "learning_rate": 7.876970124676547e-05, "loss": 2.894, "step": 27075 }, { "epoch": 2.12, "learning_rate": 7.876891711754097e-05, "loss": 2.7143, "step": 27076 }, { "epoch": 2.12, "learning_rate": 7.876813298831647e-05, "loss": 3.0425, "step": 27077 }, { "epoch": 2.12, "learning_rate": 7.876734885909198e-05, "loss": 2.7036, "step": 27078 }, { "epoch": 2.12, "learning_rate": 7.876656472986748e-05, "loss": 2.7601, "step": 27079 }, { "epoch": 2.12, "learning_rate": 7.876578060064298e-05, "loss": 2.8514, "step": 27080 }, { "epoch": 2.12, "learning_rate": 7.876499647141849e-05, "loss": 2.7535, "step": 27081 }, { "epoch": 2.12, "learning_rate": 7.8764212342194e-05, "loss": 2.8178, "step": 27082 }, { "epoch": 2.12, "learning_rate": 7.876342821296951e-05, "loss": 2.7735, "step": 27083 }, { "epoch": 2.12, "learning_rate": 7.876264408374501e-05, "loss": 2.7394, "step": 27084 }, { "epoch": 2.12, "learning_rate": 7.876185995452052e-05, "loss": 2.8385, "step": 27085 }, { "epoch": 2.12, "learning_rate": 7.876107582529602e-05, "loss": 2.6966, "step": 27086 }, { "epoch": 2.12, "learning_rate": 7.876029169607152e-05, "loss": 2.9837, "step": 27087 }, { "epoch": 2.12, "learning_rate": 7.875950756684703e-05, "loss": 2.8421, "step": 27088 }, { "epoch": 2.12, "learning_rate": 7.875872343762253e-05, "loss": 2.6408, "step": 27089 }, { "epoch": 2.12, "learning_rate": 7.875793930839804e-05, "loss": 2.7451, "step": 27090 }, { "epoch": 2.12, "learning_rate": 7.875715517917354e-05, "loss": 2.8467, "step": 27091 }, { "epoch": 2.12, "learning_rate": 7.875637104994904e-05, "loss": 2.9102, "step": 27092 }, { "epoch": 2.12, "learning_rate": 7.875558692072455e-05, "loss": 2.7654, "step": 27093 }, { "epoch": 2.12, "learning_rate": 7.875480279150004e-05, "loss": 2.6506, "step": 27094 }, { "epoch": 2.12, "learning_rate": 7.875401866227554e-05, "loss": 2.5563, "step": 27095 }, { "epoch": 2.12, "learning_rate": 7.875323453305104e-05, "loss": 2.7288, "step": 27096 }, { "epoch": 2.12, "learning_rate": 7.875245040382655e-05, "loss": 2.7112, "step": 27097 }, { "epoch": 2.12, "learning_rate": 7.875166627460205e-05, "loss": 2.7745, "step": 27098 }, { "epoch": 2.12, "learning_rate": 7.875088214537755e-05, "loss": 2.6748, "step": 27099 }, { "epoch": 2.12, "learning_rate": 7.875009801615306e-05, "loss": 2.8328, "step": 27100 }, { "epoch": 2.13, "learning_rate": 7.874931388692856e-05, "loss": 2.8691, "step": 27101 }, { "epoch": 2.13, "learning_rate": 7.874852975770407e-05, "loss": 2.6839, "step": 27102 }, { "epoch": 2.13, "learning_rate": 7.874774562847957e-05, "loss": 2.788, "step": 27103 }, { "epoch": 2.13, "learning_rate": 7.874696149925507e-05, "loss": 2.9761, "step": 27104 }, { "epoch": 2.13, "learning_rate": 7.874617737003058e-05, "loss": 2.7239, "step": 27105 }, { "epoch": 2.13, "learning_rate": 7.874539324080608e-05, "loss": 2.7615, "step": 27106 }, { "epoch": 2.13, "learning_rate": 7.874460911158158e-05, "loss": 2.7994, "step": 27107 }, { "epoch": 2.13, "learning_rate": 7.87438249823571e-05, "loss": 2.6679, "step": 27108 }, { "epoch": 2.13, "learning_rate": 7.87430408531326e-05, "loss": 2.7713, "step": 27109 }, { "epoch": 2.13, "learning_rate": 7.874225672390811e-05, "loss": 2.7033, "step": 27110 }, { "epoch": 2.13, "learning_rate": 7.874147259468361e-05, "loss": 2.9055, "step": 27111 }, { "epoch": 2.13, "learning_rate": 7.874068846545912e-05, "loss": 2.9104, "step": 27112 }, { "epoch": 2.13, "learning_rate": 7.873990433623462e-05, "loss": 2.7355, "step": 27113 }, { "epoch": 2.13, "learning_rate": 7.873912020701012e-05, "loss": 2.6842, "step": 27114 }, { "epoch": 2.13, "learning_rate": 7.873833607778563e-05, "loss": 2.901, "step": 27115 }, { "epoch": 2.13, "learning_rate": 7.873755194856113e-05, "loss": 2.8811, "step": 27116 }, { "epoch": 2.13, "learning_rate": 7.873676781933663e-05, "loss": 2.7323, "step": 27117 }, { "epoch": 2.13, "learning_rate": 7.873598369011214e-05, "loss": 2.6622, "step": 27118 }, { "epoch": 2.13, "learning_rate": 7.873519956088764e-05, "loss": 2.9819, "step": 27119 }, { "epoch": 2.13, "learning_rate": 7.873441543166315e-05, "loss": 2.7058, "step": 27120 }, { "epoch": 2.13, "learning_rate": 7.873363130243865e-05, "loss": 2.6641, "step": 27121 }, { "epoch": 2.13, "learning_rate": 7.873284717321415e-05, "loss": 2.658, "step": 27122 }, { "epoch": 2.13, "learning_rate": 7.873206304398966e-05, "loss": 2.7829, "step": 27123 }, { "epoch": 2.13, "learning_rate": 7.873127891476516e-05, "loss": 2.6972, "step": 27124 }, { "epoch": 2.13, "learning_rate": 7.873049478554066e-05, "loss": 2.567, "step": 27125 }, { "epoch": 2.13, "learning_rate": 7.872971065631617e-05, "loss": 2.888, "step": 27126 }, { "epoch": 2.13, "learning_rate": 7.872892652709167e-05, "loss": 2.8634, "step": 27127 }, { "epoch": 2.13, "learning_rate": 7.872814239786718e-05, "loss": 2.62, "step": 27128 }, { "epoch": 2.13, "learning_rate": 7.872735826864268e-05, "loss": 2.8747, "step": 27129 }, { "epoch": 2.13, "learning_rate": 7.872657413941817e-05, "loss": 2.7338, "step": 27130 }, { "epoch": 2.13, "learning_rate": 7.872579001019367e-05, "loss": 2.7352, "step": 27131 }, { "epoch": 2.13, "learning_rate": 7.872500588096918e-05, "loss": 3.1717, "step": 27132 }, { "epoch": 2.13, "learning_rate": 7.87242217517447e-05, "loss": 3.0916, "step": 27133 }, { "epoch": 2.13, "learning_rate": 7.87234376225202e-05, "loss": 2.9978, "step": 27134 }, { "epoch": 2.13, "learning_rate": 7.87226534932957e-05, "loss": 3.0824, "step": 27135 }, { "epoch": 2.13, "learning_rate": 7.87218693640712e-05, "loss": 2.7517, "step": 27136 }, { "epoch": 2.13, "learning_rate": 7.872108523484671e-05, "loss": 2.7633, "step": 27137 }, { "epoch": 2.13, "learning_rate": 7.872030110562221e-05, "loss": 2.804, "step": 27138 }, { "epoch": 2.13, "learning_rate": 7.871951697639772e-05, "loss": 2.8646, "step": 27139 }, { "epoch": 2.13, "learning_rate": 7.871873284717322e-05, "loss": 3.0268, "step": 27140 }, { "epoch": 2.13, "learning_rate": 7.871794871794872e-05, "loss": 2.8041, "step": 27141 }, { "epoch": 2.13, "learning_rate": 7.871716458872423e-05, "loss": 2.7657, "step": 27142 }, { "epoch": 2.13, "learning_rate": 7.871638045949973e-05, "loss": 2.7775, "step": 27143 }, { "epoch": 2.13, "learning_rate": 7.871559633027523e-05, "loss": 2.8564, "step": 27144 }, { "epoch": 2.13, "learning_rate": 7.871481220105074e-05, "loss": 2.7944, "step": 27145 }, { "epoch": 2.13, "learning_rate": 7.871402807182624e-05, "loss": 2.7014, "step": 27146 }, { "epoch": 2.13, "learning_rate": 7.871324394260174e-05, "loss": 2.7457, "step": 27147 }, { "epoch": 2.13, "learning_rate": 7.871245981337725e-05, "loss": 2.7961, "step": 27148 }, { "epoch": 2.13, "learning_rate": 7.871167568415275e-05, "loss": 2.7909, "step": 27149 }, { "epoch": 2.13, "learning_rate": 7.871089155492826e-05, "loss": 2.9974, "step": 27150 }, { "epoch": 2.13, "learning_rate": 7.871010742570376e-05, "loss": 2.6625, "step": 27151 }, { "epoch": 2.13, "learning_rate": 7.870932329647926e-05, "loss": 3.279, "step": 27152 }, { "epoch": 2.13, "learning_rate": 7.870853916725477e-05, "loss": 2.6321, "step": 27153 }, { "epoch": 2.13, "learning_rate": 7.870775503803027e-05, "loss": 2.8269, "step": 27154 }, { "epoch": 2.13, "learning_rate": 7.870697090880577e-05, "loss": 2.6386, "step": 27155 }, { "epoch": 2.13, "learning_rate": 7.870618677958128e-05, "loss": 2.8214, "step": 27156 }, { "epoch": 2.13, "learning_rate": 7.870540265035678e-05, "loss": 3.0474, "step": 27157 }, { "epoch": 2.13, "learning_rate": 7.870461852113229e-05, "loss": 2.7361, "step": 27158 }, { "epoch": 2.13, "learning_rate": 7.870383439190779e-05, "loss": 2.8552, "step": 27159 }, { "epoch": 2.13, "learning_rate": 7.870305026268329e-05, "loss": 2.8775, "step": 27160 }, { "epoch": 2.13, "learning_rate": 7.87022661334588e-05, "loss": 2.9492, "step": 27161 }, { "epoch": 2.13, "learning_rate": 7.87014820042343e-05, "loss": 2.8678, "step": 27162 }, { "epoch": 2.13, "learning_rate": 7.87006978750098e-05, "loss": 2.8462, "step": 27163 }, { "epoch": 2.13, "learning_rate": 7.869991374578531e-05, "loss": 2.7221, "step": 27164 }, { "epoch": 2.13, "learning_rate": 7.869912961656082e-05, "loss": 2.7954, "step": 27165 }, { "epoch": 2.13, "learning_rate": 7.869834548733631e-05, "loss": 2.9086, "step": 27166 }, { "epoch": 2.13, "learning_rate": 7.869756135811182e-05, "loss": 2.8759, "step": 27167 }, { "epoch": 2.13, "learning_rate": 7.869677722888732e-05, "loss": 2.721, "step": 27168 }, { "epoch": 2.13, "learning_rate": 7.869599309966283e-05, "loss": 2.675, "step": 27169 }, { "epoch": 2.13, "learning_rate": 7.869520897043833e-05, "loss": 2.665, "step": 27170 }, { "epoch": 2.13, "learning_rate": 7.869442484121383e-05, "loss": 2.5934, "step": 27171 }, { "epoch": 2.13, "learning_rate": 7.869364071198934e-05, "loss": 2.5893, "step": 27172 }, { "epoch": 2.13, "learning_rate": 7.869285658276484e-05, "loss": 2.8281, "step": 27173 }, { "epoch": 2.13, "learning_rate": 7.869207245354034e-05, "loss": 2.8981, "step": 27174 }, { "epoch": 2.13, "learning_rate": 7.869128832431585e-05, "loss": 2.8512, "step": 27175 }, { "epoch": 2.13, "learning_rate": 7.869050419509135e-05, "loss": 2.6517, "step": 27176 }, { "epoch": 2.13, "learning_rate": 7.868972006586686e-05, "loss": 2.7099, "step": 27177 }, { "epoch": 2.13, "learning_rate": 7.868893593664236e-05, "loss": 2.869, "step": 27178 }, { "epoch": 2.13, "learning_rate": 7.868815180741786e-05, "loss": 2.7633, "step": 27179 }, { "epoch": 2.13, "learning_rate": 7.868736767819337e-05, "loss": 3.0788, "step": 27180 }, { "epoch": 2.13, "learning_rate": 7.868658354896887e-05, "loss": 2.723, "step": 27181 }, { "epoch": 2.13, "learning_rate": 7.868579941974437e-05, "loss": 2.8466, "step": 27182 }, { "epoch": 2.13, "learning_rate": 7.868501529051988e-05, "loss": 2.6393, "step": 27183 }, { "epoch": 2.13, "learning_rate": 7.868423116129538e-05, "loss": 2.9487, "step": 27184 }, { "epoch": 2.13, "learning_rate": 7.868344703207088e-05, "loss": 2.7924, "step": 27185 }, { "epoch": 2.13, "learning_rate": 7.868266290284639e-05, "loss": 2.863, "step": 27186 }, { "epoch": 2.13, "learning_rate": 7.868187877362189e-05, "loss": 2.7227, "step": 27187 }, { "epoch": 2.13, "learning_rate": 7.86810946443974e-05, "loss": 2.5666, "step": 27188 }, { "epoch": 2.13, "learning_rate": 7.86803105151729e-05, "loss": 2.7949, "step": 27189 }, { "epoch": 2.13, "learning_rate": 7.867952638594842e-05, "loss": 2.6823, "step": 27190 }, { "epoch": 2.13, "learning_rate": 7.867874225672392e-05, "loss": 2.8036, "step": 27191 }, { "epoch": 2.13, "learning_rate": 7.867795812749942e-05, "loss": 2.6326, "step": 27192 }, { "epoch": 2.13, "learning_rate": 7.867717399827493e-05, "loss": 2.9131, "step": 27193 }, { "epoch": 2.13, "learning_rate": 7.867638986905043e-05, "loss": 2.7506, "step": 27194 }, { "epoch": 2.13, "learning_rate": 7.867560573982594e-05, "loss": 2.6025, "step": 27195 }, { "epoch": 2.13, "learning_rate": 7.867482161060144e-05, "loss": 2.7667, "step": 27196 }, { "epoch": 2.13, "learning_rate": 7.867403748137694e-05, "loss": 2.9428, "step": 27197 }, { "epoch": 2.13, "learning_rate": 7.867325335215245e-05, "loss": 3.0668, "step": 27198 }, { "epoch": 2.13, "learning_rate": 7.867246922292795e-05, "loss": 2.9982, "step": 27199 }, { "epoch": 2.13, "learning_rate": 7.867168509370345e-05, "loss": 2.8829, "step": 27200 }, { "epoch": 2.13, "learning_rate": 7.867090096447896e-05, "loss": 2.6512, "step": 27201 }, { "epoch": 2.13, "learning_rate": 7.867011683525445e-05, "loss": 2.774, "step": 27202 }, { "epoch": 2.13, "learning_rate": 7.866933270602995e-05, "loss": 2.7369, "step": 27203 }, { "epoch": 2.13, "learning_rate": 7.866854857680545e-05, "loss": 2.7446, "step": 27204 }, { "epoch": 2.13, "learning_rate": 7.866776444758096e-05, "loss": 2.7457, "step": 27205 }, { "epoch": 2.13, "learning_rate": 7.866698031835646e-05, "loss": 2.8218, "step": 27206 }, { "epoch": 2.13, "learning_rate": 7.866619618913197e-05, "loss": 2.7282, "step": 27207 }, { "epoch": 2.13, "learning_rate": 7.866541205990747e-05, "loss": 2.8099, "step": 27208 }, { "epoch": 2.13, "learning_rate": 7.866462793068297e-05, "loss": 2.831, "step": 27209 }, { "epoch": 2.13, "learning_rate": 7.866384380145848e-05, "loss": 2.7355, "step": 27210 }, { "epoch": 2.13, "learning_rate": 7.866305967223398e-05, "loss": 2.8573, "step": 27211 }, { "epoch": 2.13, "learning_rate": 7.866227554300948e-05, "loss": 2.9785, "step": 27212 }, { "epoch": 2.13, "learning_rate": 7.866149141378499e-05, "loss": 2.8569, "step": 27213 }, { "epoch": 2.13, "learning_rate": 7.866070728456049e-05, "loss": 2.7018, "step": 27214 }, { "epoch": 2.13, "learning_rate": 7.8659923155336e-05, "loss": 3.0511, "step": 27215 }, { "epoch": 2.13, "learning_rate": 7.865913902611151e-05, "loss": 2.7202, "step": 27216 }, { "epoch": 2.13, "learning_rate": 7.865835489688702e-05, "loss": 2.6368, "step": 27217 }, { "epoch": 2.13, "learning_rate": 7.865757076766252e-05, "loss": 2.5511, "step": 27218 }, { "epoch": 2.13, "learning_rate": 7.865678663843802e-05, "loss": 2.9337, "step": 27219 }, { "epoch": 2.13, "learning_rate": 7.865600250921353e-05, "loss": 2.9981, "step": 27220 }, { "epoch": 2.13, "learning_rate": 7.865521837998903e-05, "loss": 2.8269, "step": 27221 }, { "epoch": 2.13, "learning_rate": 7.865443425076453e-05, "loss": 3.1011, "step": 27222 }, { "epoch": 2.13, "learning_rate": 7.865365012154004e-05, "loss": 2.8817, "step": 27223 }, { "epoch": 2.13, "learning_rate": 7.865286599231554e-05, "loss": 3.0433, "step": 27224 }, { "epoch": 2.13, "learning_rate": 7.865208186309105e-05, "loss": 2.9252, "step": 27225 }, { "epoch": 2.13, "learning_rate": 7.865129773386655e-05, "loss": 2.8588, "step": 27226 }, { "epoch": 2.13, "learning_rate": 7.865051360464205e-05, "loss": 2.9106, "step": 27227 }, { "epoch": 2.14, "learning_rate": 7.864972947541756e-05, "loss": 2.7541, "step": 27228 }, { "epoch": 2.14, "learning_rate": 7.864894534619306e-05, "loss": 3.1491, "step": 27229 }, { "epoch": 2.14, "learning_rate": 7.864816121696856e-05, "loss": 2.7988, "step": 27230 }, { "epoch": 2.14, "learning_rate": 7.864737708774407e-05, "loss": 2.7257, "step": 27231 }, { "epoch": 2.14, "learning_rate": 7.864659295851957e-05, "loss": 2.7225, "step": 27232 }, { "epoch": 2.14, "learning_rate": 7.864580882929507e-05, "loss": 2.8171, "step": 27233 }, { "epoch": 2.14, "learning_rate": 7.864502470007058e-05, "loss": 2.7901, "step": 27234 }, { "epoch": 2.14, "learning_rate": 7.864424057084608e-05, "loss": 2.6135, "step": 27235 }, { "epoch": 2.14, "learning_rate": 7.864345644162159e-05, "loss": 3.0802, "step": 27236 }, { "epoch": 2.14, "learning_rate": 7.864267231239709e-05, "loss": 2.5803, "step": 27237 }, { "epoch": 2.14, "learning_rate": 7.864188818317258e-05, "loss": 2.7251, "step": 27238 }, { "epoch": 2.14, "learning_rate": 7.864110405394808e-05, "loss": 2.8007, "step": 27239 }, { "epoch": 2.14, "learning_rate": 7.864031992472359e-05, "loss": 2.7697, "step": 27240 }, { "epoch": 2.14, "learning_rate": 7.86395357954991e-05, "loss": 2.8192, "step": 27241 }, { "epoch": 2.14, "learning_rate": 7.863875166627461e-05, "loss": 2.6833, "step": 27242 }, { "epoch": 2.14, "learning_rate": 7.863796753705011e-05, "loss": 2.9344, "step": 27243 }, { "epoch": 2.14, "learning_rate": 7.863718340782562e-05, "loss": 2.7141, "step": 27244 }, { "epoch": 2.14, "learning_rate": 7.863639927860112e-05, "loss": 2.7938, "step": 27245 }, { "epoch": 2.14, "learning_rate": 7.863561514937662e-05, "loss": 2.7557, "step": 27246 }, { "epoch": 2.14, "learning_rate": 7.863483102015213e-05, "loss": 2.8986, "step": 27247 }, { "epoch": 2.14, "learning_rate": 7.863404689092763e-05, "loss": 2.6725, "step": 27248 }, { "epoch": 2.14, "learning_rate": 7.863326276170313e-05, "loss": 2.6196, "step": 27249 }, { "epoch": 2.14, "learning_rate": 7.863247863247864e-05, "loss": 2.7744, "step": 27250 }, { "epoch": 2.14, "learning_rate": 7.863169450325414e-05, "loss": 2.6201, "step": 27251 }, { "epoch": 2.14, "learning_rate": 7.863091037402964e-05, "loss": 2.5884, "step": 27252 }, { "epoch": 2.14, "learning_rate": 7.863012624480515e-05, "loss": 2.8394, "step": 27253 }, { "epoch": 2.14, "learning_rate": 7.862934211558065e-05, "loss": 2.7989, "step": 27254 }, { "epoch": 2.14, "learning_rate": 7.862855798635616e-05, "loss": 2.8993, "step": 27255 }, { "epoch": 2.14, "learning_rate": 7.862777385713166e-05, "loss": 2.9611, "step": 27256 }, { "epoch": 2.14, "learning_rate": 7.862698972790716e-05, "loss": 3.2108, "step": 27257 }, { "epoch": 2.14, "learning_rate": 7.862620559868267e-05, "loss": 2.7848, "step": 27258 }, { "epoch": 2.14, "learning_rate": 7.862542146945817e-05, "loss": 2.7729, "step": 27259 }, { "epoch": 2.14, "learning_rate": 7.862463734023367e-05, "loss": 2.79, "step": 27260 }, { "epoch": 2.14, "learning_rate": 7.862385321100918e-05, "loss": 2.4748, "step": 27261 }, { "epoch": 2.14, "learning_rate": 7.862306908178468e-05, "loss": 2.575, "step": 27262 }, { "epoch": 2.14, "learning_rate": 7.862228495256019e-05, "loss": 2.6514, "step": 27263 }, { "epoch": 2.14, "learning_rate": 7.862150082333569e-05, "loss": 2.9923, "step": 27264 }, { "epoch": 2.14, "learning_rate": 7.862071669411119e-05, "loss": 2.7506, "step": 27265 }, { "epoch": 2.14, "learning_rate": 7.86199325648867e-05, "loss": 2.8371, "step": 27266 }, { "epoch": 2.14, "learning_rate": 7.86191484356622e-05, "loss": 2.7236, "step": 27267 }, { "epoch": 2.14, "learning_rate": 7.86183643064377e-05, "loss": 2.7444, "step": 27268 }, { "epoch": 2.14, "learning_rate": 7.861758017721321e-05, "loss": 2.9901, "step": 27269 }, { "epoch": 2.14, "learning_rate": 7.861679604798871e-05, "loss": 2.6653, "step": 27270 }, { "epoch": 2.14, "learning_rate": 7.861601191876421e-05, "loss": 2.7927, "step": 27271 }, { "epoch": 2.14, "learning_rate": 7.861522778953972e-05, "loss": 2.8384, "step": 27272 }, { "epoch": 2.14, "learning_rate": 7.861444366031524e-05, "loss": 2.6665, "step": 27273 }, { "epoch": 2.14, "learning_rate": 7.861365953109073e-05, "loss": 2.7259, "step": 27274 }, { "epoch": 2.14, "learning_rate": 7.861287540186623e-05, "loss": 2.6459, "step": 27275 }, { "epoch": 2.14, "learning_rate": 7.861209127264173e-05, "loss": 2.9006, "step": 27276 }, { "epoch": 2.14, "learning_rate": 7.861130714341724e-05, "loss": 2.5971, "step": 27277 }, { "epoch": 2.14, "learning_rate": 7.861052301419274e-05, "loss": 2.9189, "step": 27278 }, { "epoch": 2.14, "learning_rate": 7.860973888496824e-05, "loss": 2.821, "step": 27279 }, { "epoch": 2.14, "learning_rate": 7.860895475574375e-05, "loss": 2.7467, "step": 27280 }, { "epoch": 2.14, "learning_rate": 7.860817062651925e-05, "loss": 2.8129, "step": 27281 }, { "epoch": 2.14, "learning_rate": 7.860738649729475e-05, "loss": 2.7278, "step": 27282 }, { "epoch": 2.14, "learning_rate": 7.860660236807026e-05, "loss": 2.6995, "step": 27283 }, { "epoch": 2.14, "learning_rate": 7.860581823884576e-05, "loss": 2.706, "step": 27284 }, { "epoch": 2.14, "learning_rate": 7.860503410962127e-05, "loss": 2.9124, "step": 27285 }, { "epoch": 2.14, "learning_rate": 7.860424998039677e-05, "loss": 2.784, "step": 27286 }, { "epoch": 2.14, "learning_rate": 7.860346585117227e-05, "loss": 2.8134, "step": 27287 }, { "epoch": 2.14, "learning_rate": 7.860268172194778e-05, "loss": 2.6372, "step": 27288 }, { "epoch": 2.14, "learning_rate": 7.860189759272328e-05, "loss": 2.7946, "step": 27289 }, { "epoch": 2.14, "learning_rate": 7.860111346349878e-05, "loss": 2.7184, "step": 27290 }, { "epoch": 2.14, "learning_rate": 7.860032933427429e-05, "loss": 2.9013, "step": 27291 }, { "epoch": 2.14, "learning_rate": 7.859954520504979e-05, "loss": 2.7546, "step": 27292 }, { "epoch": 2.14, "learning_rate": 7.85987610758253e-05, "loss": 2.7637, "step": 27293 }, { "epoch": 2.14, "learning_rate": 7.85979769466008e-05, "loss": 2.6538, "step": 27294 }, { "epoch": 2.14, "learning_rate": 7.85971928173763e-05, "loss": 2.7283, "step": 27295 }, { "epoch": 2.14, "learning_rate": 7.85964086881518e-05, "loss": 2.6816, "step": 27296 }, { "epoch": 2.14, "learning_rate": 7.859562455892731e-05, "loss": 3.0378, "step": 27297 }, { "epoch": 2.14, "learning_rate": 7.859484042970283e-05, "loss": 2.7397, "step": 27298 }, { "epoch": 2.14, "learning_rate": 7.859405630047833e-05, "loss": 2.6514, "step": 27299 }, { "epoch": 2.14, "learning_rate": 7.859327217125383e-05, "loss": 3.1265, "step": 27300 }, { "epoch": 2.14, "learning_rate": 7.859248804202934e-05, "loss": 2.7857, "step": 27301 }, { "epoch": 2.14, "learning_rate": 7.859170391280484e-05, "loss": 2.9485, "step": 27302 }, { "epoch": 2.14, "learning_rate": 7.859091978358035e-05, "loss": 2.7294, "step": 27303 }, { "epoch": 2.14, "learning_rate": 7.859013565435585e-05, "loss": 2.8215, "step": 27304 }, { "epoch": 2.14, "learning_rate": 7.858935152513135e-05, "loss": 2.8885, "step": 27305 }, { "epoch": 2.14, "learning_rate": 7.858856739590686e-05, "loss": 2.7582, "step": 27306 }, { "epoch": 2.14, "learning_rate": 7.858778326668236e-05, "loss": 2.7594, "step": 27307 }, { "epoch": 2.14, "learning_rate": 7.858699913745786e-05, "loss": 2.6144, "step": 27308 }, { "epoch": 2.14, "learning_rate": 7.858621500823337e-05, "loss": 2.7074, "step": 27309 }, { "epoch": 2.14, "learning_rate": 7.858543087900886e-05, "loss": 2.8417, "step": 27310 }, { "epoch": 2.14, "learning_rate": 7.858464674978436e-05, "loss": 2.7482, "step": 27311 }, { "epoch": 2.14, "learning_rate": 7.858386262055987e-05, "loss": 2.7877, "step": 27312 }, { "epoch": 2.14, "learning_rate": 7.858307849133537e-05, "loss": 2.7142, "step": 27313 }, { "epoch": 2.14, "learning_rate": 7.858229436211087e-05, "loss": 2.7942, "step": 27314 }, { "epoch": 2.14, "learning_rate": 7.858151023288638e-05, "loss": 2.5704, "step": 27315 }, { "epoch": 2.14, "learning_rate": 7.858072610366188e-05, "loss": 2.8415, "step": 27316 }, { "epoch": 2.14, "learning_rate": 7.857994197443738e-05, "loss": 2.7573, "step": 27317 }, { "epoch": 2.14, "learning_rate": 7.857915784521289e-05, "loss": 2.8655, "step": 27318 }, { "epoch": 2.14, "learning_rate": 7.857837371598839e-05, "loss": 2.7788, "step": 27319 }, { "epoch": 2.14, "learning_rate": 7.85775895867639e-05, "loss": 2.7154, "step": 27320 }, { "epoch": 2.14, "learning_rate": 7.85768054575394e-05, "loss": 3.1018, "step": 27321 }, { "epoch": 2.14, "learning_rate": 7.85760213283149e-05, "loss": 2.9849, "step": 27322 }, { "epoch": 2.14, "learning_rate": 7.85752371990904e-05, "loss": 2.823, "step": 27323 }, { "epoch": 2.14, "learning_rate": 7.857445306986592e-05, "loss": 2.701, "step": 27324 }, { "epoch": 2.14, "learning_rate": 7.857366894064143e-05, "loss": 2.7533, "step": 27325 }, { "epoch": 2.14, "learning_rate": 7.857288481141693e-05, "loss": 2.8726, "step": 27326 }, { "epoch": 2.14, "learning_rate": 7.857210068219243e-05, "loss": 2.6871, "step": 27327 }, { "epoch": 2.14, "learning_rate": 7.857131655296794e-05, "loss": 3.0773, "step": 27328 }, { "epoch": 2.14, "learning_rate": 7.857053242374344e-05, "loss": 2.989, "step": 27329 }, { "epoch": 2.14, "learning_rate": 7.856974829451895e-05, "loss": 2.8658, "step": 27330 }, { "epoch": 2.14, "learning_rate": 7.856896416529445e-05, "loss": 2.7945, "step": 27331 }, { "epoch": 2.14, "learning_rate": 7.856818003606995e-05, "loss": 2.7088, "step": 27332 }, { "epoch": 2.14, "learning_rate": 7.856739590684546e-05, "loss": 2.8024, "step": 27333 }, { "epoch": 2.14, "learning_rate": 7.856661177762096e-05, "loss": 2.7247, "step": 27334 }, { "epoch": 2.14, "learning_rate": 7.856582764839646e-05, "loss": 2.7462, "step": 27335 }, { "epoch": 2.14, "learning_rate": 7.856504351917197e-05, "loss": 2.8866, "step": 27336 }, { "epoch": 2.14, "learning_rate": 7.856425938994747e-05, "loss": 2.9044, "step": 27337 }, { "epoch": 2.14, "learning_rate": 7.856347526072297e-05, "loss": 2.672, "step": 27338 }, { "epoch": 2.14, "learning_rate": 7.856269113149848e-05, "loss": 2.8429, "step": 27339 }, { "epoch": 2.14, "learning_rate": 7.856190700227398e-05, "loss": 2.7488, "step": 27340 }, { "epoch": 2.14, "learning_rate": 7.856112287304949e-05, "loss": 2.7041, "step": 27341 }, { "epoch": 2.14, "learning_rate": 7.856033874382499e-05, "loss": 2.7766, "step": 27342 }, { "epoch": 2.14, "learning_rate": 7.855955461460049e-05, "loss": 2.7678, "step": 27343 }, { "epoch": 2.14, "learning_rate": 7.8558770485376e-05, "loss": 2.8345, "step": 27344 }, { "epoch": 2.14, "learning_rate": 7.85579863561515e-05, "loss": 2.7637, "step": 27345 }, { "epoch": 2.14, "learning_rate": 7.855720222692699e-05, "loss": 2.6084, "step": 27346 }, { "epoch": 2.14, "learning_rate": 7.85564180977025e-05, "loss": 2.7922, "step": 27347 }, { "epoch": 2.14, "learning_rate": 7.8555633968478e-05, "loss": 2.7789, "step": 27348 }, { "epoch": 2.14, "learning_rate": 7.855484983925351e-05, "loss": 2.6937, "step": 27349 }, { "epoch": 2.14, "learning_rate": 7.855406571002902e-05, "loss": 2.7824, "step": 27350 }, { "epoch": 2.14, "learning_rate": 7.855328158080452e-05, "loss": 2.8036, "step": 27351 }, { "epoch": 2.14, "learning_rate": 7.855249745158003e-05, "loss": 2.5882, "step": 27352 }, { "epoch": 2.14, "learning_rate": 7.855171332235553e-05, "loss": 2.5686, "step": 27353 }, { "epoch": 2.14, "learning_rate": 7.855092919313103e-05, "loss": 2.9727, "step": 27354 }, { "epoch": 2.14, "learning_rate": 7.855014506390654e-05, "loss": 2.7432, "step": 27355 }, { "epoch": 2.15, "learning_rate": 7.854936093468204e-05, "loss": 2.8721, "step": 27356 }, { "epoch": 2.15, "learning_rate": 7.854857680545754e-05, "loss": 2.7973, "step": 27357 }, { "epoch": 2.15, "learning_rate": 7.854779267623305e-05, "loss": 2.6711, "step": 27358 }, { "epoch": 2.15, "learning_rate": 7.854700854700855e-05, "loss": 2.6248, "step": 27359 }, { "epoch": 2.15, "learning_rate": 7.854622441778406e-05, "loss": 2.7148, "step": 27360 }, { "epoch": 2.15, "learning_rate": 7.854544028855956e-05, "loss": 2.7928, "step": 27361 }, { "epoch": 2.15, "learning_rate": 7.854465615933506e-05, "loss": 2.6877, "step": 27362 }, { "epoch": 2.15, "learning_rate": 7.854387203011057e-05, "loss": 2.7114, "step": 27363 }, { "epoch": 2.15, "learning_rate": 7.854308790088607e-05, "loss": 2.7348, "step": 27364 }, { "epoch": 2.15, "learning_rate": 7.854230377166157e-05, "loss": 2.5978, "step": 27365 }, { "epoch": 2.15, "learning_rate": 7.854151964243708e-05, "loss": 2.7399, "step": 27366 }, { "epoch": 2.15, "learning_rate": 7.854073551321258e-05, "loss": 2.7221, "step": 27367 }, { "epoch": 2.15, "learning_rate": 7.853995138398808e-05, "loss": 2.8624, "step": 27368 }, { "epoch": 2.15, "learning_rate": 7.853916725476359e-05, "loss": 2.7895, "step": 27369 }, { "epoch": 2.15, "learning_rate": 7.853838312553909e-05, "loss": 2.9871, "step": 27370 }, { "epoch": 2.15, "learning_rate": 7.85375989963146e-05, "loss": 2.7945, "step": 27371 }, { "epoch": 2.15, "learning_rate": 7.85368148670901e-05, "loss": 2.8485, "step": 27372 }, { "epoch": 2.15, "learning_rate": 7.85360307378656e-05, "loss": 2.784, "step": 27373 }, { "epoch": 2.15, "learning_rate": 7.853524660864111e-05, "loss": 2.7027, "step": 27374 }, { "epoch": 2.15, "learning_rate": 7.853446247941661e-05, "loss": 2.7332, "step": 27375 }, { "epoch": 2.15, "learning_rate": 7.853367835019211e-05, "loss": 2.5839, "step": 27376 }, { "epoch": 2.15, "learning_rate": 7.853289422096762e-05, "loss": 2.9232, "step": 27377 }, { "epoch": 2.15, "learning_rate": 7.853211009174312e-05, "loss": 2.736, "step": 27378 }, { "epoch": 2.15, "learning_rate": 7.853132596251863e-05, "loss": 2.9056, "step": 27379 }, { "epoch": 2.15, "learning_rate": 7.853054183329413e-05, "loss": 2.8423, "step": 27380 }, { "epoch": 2.15, "learning_rate": 7.852975770406965e-05, "loss": 2.842, "step": 27381 }, { "epoch": 2.15, "learning_rate": 7.852897357484514e-05, "loss": 2.8001, "step": 27382 }, { "epoch": 2.15, "learning_rate": 7.852818944562064e-05, "loss": 2.6967, "step": 27383 }, { "epoch": 2.15, "learning_rate": 7.852740531639614e-05, "loss": 2.8203, "step": 27384 }, { "epoch": 2.15, "learning_rate": 7.852662118717165e-05, "loss": 2.7752, "step": 27385 }, { "epoch": 2.15, "learning_rate": 7.852583705794715e-05, "loss": 3.0778, "step": 27386 }, { "epoch": 2.15, "learning_rate": 7.852505292872265e-05, "loss": 2.6929, "step": 27387 }, { "epoch": 2.15, "learning_rate": 7.852426879949816e-05, "loss": 2.644, "step": 27388 }, { "epoch": 2.15, "learning_rate": 7.852348467027366e-05, "loss": 3.0057, "step": 27389 }, { "epoch": 2.15, "learning_rate": 7.852270054104917e-05, "loss": 2.7481, "step": 27390 }, { "epoch": 2.15, "learning_rate": 7.852191641182467e-05, "loss": 2.7811, "step": 27391 }, { "epoch": 2.15, "learning_rate": 7.852113228260017e-05, "loss": 2.626, "step": 27392 }, { "epoch": 2.15, "learning_rate": 7.852034815337568e-05, "loss": 2.632, "step": 27393 }, { "epoch": 2.15, "learning_rate": 7.851956402415118e-05, "loss": 2.8582, "step": 27394 }, { "epoch": 2.15, "learning_rate": 7.851877989492668e-05, "loss": 2.893, "step": 27395 }, { "epoch": 2.15, "learning_rate": 7.851799576570219e-05, "loss": 2.7796, "step": 27396 }, { "epoch": 2.15, "learning_rate": 7.851721163647769e-05, "loss": 2.5646, "step": 27397 }, { "epoch": 2.15, "learning_rate": 7.85164275072532e-05, "loss": 3.005, "step": 27398 }, { "epoch": 2.15, "learning_rate": 7.85156433780287e-05, "loss": 2.7158, "step": 27399 }, { "epoch": 2.15, "learning_rate": 7.85148592488042e-05, "loss": 2.8419, "step": 27400 }, { "epoch": 2.15, "learning_rate": 7.85140751195797e-05, "loss": 2.8486, "step": 27401 }, { "epoch": 2.15, "learning_rate": 7.851329099035521e-05, "loss": 2.8151, "step": 27402 }, { "epoch": 2.15, "learning_rate": 7.851250686113071e-05, "loss": 2.8958, "step": 27403 }, { "epoch": 2.15, "learning_rate": 7.851172273190622e-05, "loss": 2.7423, "step": 27404 }, { "epoch": 2.15, "learning_rate": 7.851093860268172e-05, "loss": 2.7173, "step": 27405 }, { "epoch": 2.15, "learning_rate": 7.851015447345724e-05, "loss": 2.9185, "step": 27406 }, { "epoch": 2.15, "learning_rate": 7.850937034423274e-05, "loss": 2.9956, "step": 27407 }, { "epoch": 2.15, "learning_rate": 7.850858621500825e-05, "loss": 3.0721, "step": 27408 }, { "epoch": 2.15, "learning_rate": 7.850780208578375e-05, "loss": 2.708, "step": 27409 }, { "epoch": 2.15, "learning_rate": 7.850701795655925e-05, "loss": 2.7887, "step": 27410 }, { "epoch": 2.15, "learning_rate": 7.850623382733476e-05, "loss": 2.6895, "step": 27411 }, { "epoch": 2.15, "learning_rate": 7.850544969811026e-05, "loss": 2.7931, "step": 27412 }, { "epoch": 2.15, "learning_rate": 7.850466556888576e-05, "loss": 2.9288, "step": 27413 }, { "epoch": 2.15, "learning_rate": 7.850388143966127e-05, "loss": 2.9433, "step": 27414 }, { "epoch": 2.15, "learning_rate": 7.850309731043677e-05, "loss": 3.071, "step": 27415 }, { "epoch": 2.15, "learning_rate": 7.850231318121227e-05, "loss": 2.7721, "step": 27416 }, { "epoch": 2.15, "learning_rate": 7.850152905198778e-05, "loss": 2.8192, "step": 27417 }, { "epoch": 2.15, "learning_rate": 7.850074492276327e-05, "loss": 2.8359, "step": 27418 }, { "epoch": 2.15, "learning_rate": 7.849996079353877e-05, "loss": 3.0719, "step": 27419 }, { "epoch": 2.15, "learning_rate": 7.849917666431428e-05, "loss": 2.793, "step": 27420 }, { "epoch": 2.15, "learning_rate": 7.849839253508978e-05, "loss": 2.7903, "step": 27421 }, { "epoch": 2.15, "learning_rate": 7.849760840586528e-05, "loss": 2.7253, "step": 27422 }, { "epoch": 2.15, "learning_rate": 7.849682427664079e-05, "loss": 2.8509, "step": 27423 }, { "epoch": 2.15, "learning_rate": 7.849604014741629e-05, "loss": 2.6576, "step": 27424 }, { "epoch": 2.15, "learning_rate": 7.84952560181918e-05, "loss": 2.7181, "step": 27425 }, { "epoch": 2.15, "learning_rate": 7.84944718889673e-05, "loss": 2.5892, "step": 27426 }, { "epoch": 2.15, "learning_rate": 7.84936877597428e-05, "loss": 2.7124, "step": 27427 }, { "epoch": 2.15, "learning_rate": 7.84929036305183e-05, "loss": 2.8351, "step": 27428 }, { "epoch": 2.15, "learning_rate": 7.849211950129381e-05, "loss": 3.0501, "step": 27429 }, { "epoch": 2.15, "learning_rate": 7.849133537206931e-05, "loss": 2.8575, "step": 27430 }, { "epoch": 2.15, "learning_rate": 7.849055124284482e-05, "loss": 2.8831, "step": 27431 }, { "epoch": 2.15, "learning_rate": 7.848976711362033e-05, "loss": 2.5605, "step": 27432 }, { "epoch": 2.15, "learning_rate": 7.848898298439584e-05, "loss": 2.817, "step": 27433 }, { "epoch": 2.15, "learning_rate": 7.848819885517134e-05, "loss": 2.7241, "step": 27434 }, { "epoch": 2.15, "learning_rate": 7.848741472594684e-05, "loss": 2.9965, "step": 27435 }, { "epoch": 2.15, "learning_rate": 7.848663059672235e-05, "loss": 2.9225, "step": 27436 }, { "epoch": 2.15, "learning_rate": 7.848584646749785e-05, "loss": 2.8903, "step": 27437 }, { "epoch": 2.15, "learning_rate": 7.848506233827336e-05, "loss": 3.0065, "step": 27438 }, { "epoch": 2.15, "learning_rate": 7.848427820904886e-05, "loss": 2.663, "step": 27439 }, { "epoch": 2.15, "learning_rate": 7.848349407982436e-05, "loss": 2.7128, "step": 27440 }, { "epoch": 2.15, "learning_rate": 7.848270995059987e-05, "loss": 2.9609, "step": 27441 }, { "epoch": 2.15, "learning_rate": 7.848192582137537e-05, "loss": 2.8589, "step": 27442 }, { "epoch": 2.15, "learning_rate": 7.848114169215087e-05, "loss": 2.5067, "step": 27443 }, { "epoch": 2.15, "learning_rate": 7.848035756292638e-05, "loss": 2.842, "step": 27444 }, { "epoch": 2.15, "learning_rate": 7.847957343370188e-05, "loss": 2.8201, "step": 27445 }, { "epoch": 2.15, "learning_rate": 7.847878930447739e-05, "loss": 2.9919, "step": 27446 }, { "epoch": 2.15, "learning_rate": 7.847800517525289e-05, "loss": 2.848, "step": 27447 }, { "epoch": 2.15, "learning_rate": 7.847722104602839e-05, "loss": 2.6746, "step": 27448 }, { "epoch": 2.15, "learning_rate": 7.84764369168039e-05, "loss": 2.8839, "step": 27449 }, { "epoch": 2.15, "learning_rate": 7.84756527875794e-05, "loss": 2.8147, "step": 27450 }, { "epoch": 2.15, "learning_rate": 7.84748686583549e-05, "loss": 2.6857, "step": 27451 }, { "epoch": 2.15, "learning_rate": 7.847408452913041e-05, "loss": 2.7623, "step": 27452 }, { "epoch": 2.15, "learning_rate": 7.847330039990591e-05, "loss": 2.7869, "step": 27453 }, { "epoch": 2.15, "learning_rate": 7.84725162706814e-05, "loss": 2.8323, "step": 27454 }, { "epoch": 2.15, "learning_rate": 7.84717321414569e-05, "loss": 2.7735, "step": 27455 }, { "epoch": 2.15, "learning_rate": 7.847094801223241e-05, "loss": 2.6754, "step": 27456 }, { "epoch": 2.15, "learning_rate": 7.847016388300791e-05, "loss": 2.8631, "step": 27457 }, { "epoch": 2.15, "learning_rate": 7.846937975378343e-05, "loss": 3.0411, "step": 27458 }, { "epoch": 2.15, "learning_rate": 7.846859562455893e-05, "loss": 2.7737, "step": 27459 }, { "epoch": 2.15, "learning_rate": 7.846781149533444e-05, "loss": 2.7724, "step": 27460 }, { "epoch": 2.15, "learning_rate": 7.846702736610994e-05, "loss": 2.9308, "step": 27461 }, { "epoch": 2.15, "learning_rate": 7.846624323688544e-05, "loss": 2.7608, "step": 27462 }, { "epoch": 2.15, "learning_rate": 7.846545910766095e-05, "loss": 3.0207, "step": 27463 }, { "epoch": 2.15, "learning_rate": 7.846467497843645e-05, "loss": 2.9351, "step": 27464 }, { "epoch": 2.15, "learning_rate": 7.846389084921196e-05, "loss": 2.8136, "step": 27465 }, { "epoch": 2.15, "learning_rate": 7.846310671998746e-05, "loss": 2.7296, "step": 27466 }, { "epoch": 2.15, "learning_rate": 7.846232259076296e-05, "loss": 2.8125, "step": 27467 }, { "epoch": 2.15, "learning_rate": 7.846153846153847e-05, "loss": 2.9704, "step": 27468 }, { "epoch": 2.15, "learning_rate": 7.846075433231397e-05, "loss": 2.7375, "step": 27469 }, { "epoch": 2.15, "learning_rate": 7.845997020308947e-05, "loss": 2.8307, "step": 27470 }, { "epoch": 2.15, "learning_rate": 7.845918607386498e-05, "loss": 2.9301, "step": 27471 }, { "epoch": 2.15, "learning_rate": 7.845840194464048e-05, "loss": 2.7017, "step": 27472 }, { "epoch": 2.15, "learning_rate": 7.845761781541598e-05, "loss": 2.7628, "step": 27473 }, { "epoch": 2.15, "learning_rate": 7.845683368619149e-05, "loss": 2.8256, "step": 27474 }, { "epoch": 2.15, "learning_rate": 7.845604955696699e-05, "loss": 2.9214, "step": 27475 }, { "epoch": 2.15, "learning_rate": 7.84552654277425e-05, "loss": 2.7041, "step": 27476 }, { "epoch": 2.15, "learning_rate": 7.8454481298518e-05, "loss": 2.8215, "step": 27477 }, { "epoch": 2.15, "learning_rate": 7.84536971692935e-05, "loss": 2.6802, "step": 27478 }, { "epoch": 2.15, "learning_rate": 7.8452913040069e-05, "loss": 2.7773, "step": 27479 }, { "epoch": 2.15, "learning_rate": 7.845212891084451e-05, "loss": 2.916, "step": 27480 }, { "epoch": 2.15, "learning_rate": 7.845134478162001e-05, "loss": 2.6828, "step": 27481 }, { "epoch": 2.15, "learning_rate": 7.845056065239552e-05, "loss": 2.8466, "step": 27482 }, { "epoch": 2.16, "learning_rate": 7.844977652317102e-05, "loss": 2.8073, "step": 27483 }, { "epoch": 2.16, "learning_rate": 7.844899239394652e-05, "loss": 2.7523, "step": 27484 }, { "epoch": 2.16, "learning_rate": 7.844820826472203e-05, "loss": 2.851, "step": 27485 }, { "epoch": 2.16, "learning_rate": 7.844742413549753e-05, "loss": 2.9297, "step": 27486 }, { "epoch": 2.16, "learning_rate": 7.844664000627304e-05, "loss": 2.7031, "step": 27487 }, { "epoch": 2.16, "learning_rate": 7.844585587704854e-05, "loss": 2.8824, "step": 27488 }, { "epoch": 2.16, "learning_rate": 7.844507174782406e-05, "loss": 2.6976, "step": 27489 }, { "epoch": 2.16, "learning_rate": 7.844428761859955e-05, "loss": 3.014, "step": 27490 }, { "epoch": 2.16, "learning_rate": 7.844350348937505e-05, "loss": 2.7214, "step": 27491 }, { "epoch": 2.16, "learning_rate": 7.844271936015055e-05, "loss": 2.7808, "step": 27492 }, { "epoch": 2.16, "learning_rate": 7.844193523092606e-05, "loss": 2.7587, "step": 27493 }, { "epoch": 2.16, "learning_rate": 7.844115110170156e-05, "loss": 2.8364, "step": 27494 }, { "epoch": 2.16, "learning_rate": 7.844036697247707e-05, "loss": 2.6569, "step": 27495 }, { "epoch": 2.16, "learning_rate": 7.843958284325257e-05, "loss": 3.005, "step": 27496 }, { "epoch": 2.16, "learning_rate": 7.843879871402807e-05, "loss": 2.7445, "step": 27497 }, { "epoch": 2.16, "learning_rate": 7.843801458480358e-05, "loss": 2.6533, "step": 27498 }, { "epoch": 2.16, "learning_rate": 7.843723045557908e-05, "loss": 2.7924, "step": 27499 }, { "epoch": 2.16, "learning_rate": 7.843644632635458e-05, "loss": 2.8641, "step": 27500 }, { "epoch": 2.16, "learning_rate": 7.843566219713009e-05, "loss": 2.7643, "step": 27501 }, { "epoch": 2.16, "learning_rate": 7.843487806790559e-05, "loss": 2.8983, "step": 27502 }, { "epoch": 2.16, "learning_rate": 7.84340939386811e-05, "loss": 2.8383, "step": 27503 }, { "epoch": 2.16, "learning_rate": 7.84333098094566e-05, "loss": 2.8867, "step": 27504 }, { "epoch": 2.16, "learning_rate": 7.84325256802321e-05, "loss": 2.934, "step": 27505 }, { "epoch": 2.16, "learning_rate": 7.84317415510076e-05, "loss": 2.8163, "step": 27506 }, { "epoch": 2.16, "learning_rate": 7.843095742178311e-05, "loss": 2.8246, "step": 27507 }, { "epoch": 2.16, "learning_rate": 7.843017329255861e-05, "loss": 2.8306, "step": 27508 }, { "epoch": 2.16, "learning_rate": 7.842938916333412e-05, "loss": 2.5512, "step": 27509 }, { "epoch": 2.16, "learning_rate": 7.842860503410962e-05, "loss": 3.1047, "step": 27510 }, { "epoch": 2.16, "learning_rate": 7.842782090488512e-05, "loss": 2.8117, "step": 27511 }, { "epoch": 2.16, "learning_rate": 7.842703677566063e-05, "loss": 2.8829, "step": 27512 }, { "epoch": 2.16, "learning_rate": 7.842625264643613e-05, "loss": 2.884, "step": 27513 }, { "epoch": 2.16, "learning_rate": 7.842546851721165e-05, "loss": 2.7779, "step": 27514 }, { "epoch": 2.16, "learning_rate": 7.842468438798715e-05, "loss": 2.6466, "step": 27515 }, { "epoch": 2.16, "learning_rate": 7.842390025876266e-05, "loss": 2.9327, "step": 27516 }, { "epoch": 2.16, "learning_rate": 7.842311612953816e-05, "loss": 2.7439, "step": 27517 }, { "epoch": 2.16, "learning_rate": 7.842233200031366e-05, "loss": 2.7836, "step": 27518 }, { "epoch": 2.16, "learning_rate": 7.842154787108917e-05, "loss": 2.9123, "step": 27519 }, { "epoch": 2.16, "learning_rate": 7.842076374186467e-05, "loss": 2.6233, "step": 27520 }, { "epoch": 2.16, "learning_rate": 7.841997961264017e-05, "loss": 2.8411, "step": 27521 }, { "epoch": 2.16, "learning_rate": 7.841919548341568e-05, "loss": 2.7134, "step": 27522 }, { "epoch": 2.16, "learning_rate": 7.841841135419118e-05, "loss": 2.8626, "step": 27523 }, { "epoch": 2.16, "learning_rate": 7.841762722496669e-05, "loss": 3.0519, "step": 27524 }, { "epoch": 2.16, "learning_rate": 7.841684309574219e-05, "loss": 2.8681, "step": 27525 }, { "epoch": 2.16, "learning_rate": 7.841605896651768e-05, "loss": 2.6586, "step": 27526 }, { "epoch": 2.16, "learning_rate": 7.841527483729318e-05, "loss": 2.8506, "step": 27527 }, { "epoch": 2.16, "learning_rate": 7.841449070806869e-05, "loss": 2.8751, "step": 27528 }, { "epoch": 2.16, "learning_rate": 7.841370657884419e-05, "loss": 2.8454, "step": 27529 }, { "epoch": 2.16, "learning_rate": 7.84129224496197e-05, "loss": 2.8422, "step": 27530 }, { "epoch": 2.16, "learning_rate": 7.84121383203952e-05, "loss": 2.5681, "step": 27531 }, { "epoch": 2.16, "learning_rate": 7.84113541911707e-05, "loss": 2.8158, "step": 27532 }, { "epoch": 2.16, "learning_rate": 7.84105700619462e-05, "loss": 2.6966, "step": 27533 }, { "epoch": 2.16, "learning_rate": 7.840978593272171e-05, "loss": 2.7851, "step": 27534 }, { "epoch": 2.16, "learning_rate": 7.840900180349721e-05, "loss": 2.6712, "step": 27535 }, { "epoch": 2.16, "learning_rate": 7.840821767427272e-05, "loss": 2.7195, "step": 27536 }, { "epoch": 2.16, "learning_rate": 7.840743354504822e-05, "loss": 2.6971, "step": 27537 }, { "epoch": 2.16, "learning_rate": 7.840664941582372e-05, "loss": 2.8547, "step": 27538 }, { "epoch": 2.16, "learning_rate": 7.840586528659923e-05, "loss": 2.8444, "step": 27539 }, { "epoch": 2.16, "learning_rate": 7.840508115737474e-05, "loss": 2.8584, "step": 27540 }, { "epoch": 2.16, "learning_rate": 7.840429702815025e-05, "loss": 2.7692, "step": 27541 }, { "epoch": 2.16, "learning_rate": 7.840351289892575e-05, "loss": 2.6943, "step": 27542 }, { "epoch": 2.16, "learning_rate": 7.840272876970126e-05, "loss": 2.9975, "step": 27543 }, { "epoch": 2.16, "learning_rate": 7.840194464047676e-05, "loss": 2.8976, "step": 27544 }, { "epoch": 2.16, "learning_rate": 7.840116051125226e-05, "loss": 2.6016, "step": 27545 }, { "epoch": 2.16, "learning_rate": 7.840037638202777e-05, "loss": 2.7174, "step": 27546 }, { "epoch": 2.16, "learning_rate": 7.839959225280327e-05, "loss": 2.9565, "step": 27547 }, { "epoch": 2.16, "learning_rate": 7.839880812357877e-05, "loss": 2.9153, "step": 27548 }, { "epoch": 2.16, "learning_rate": 7.839802399435428e-05, "loss": 2.8093, "step": 27549 }, { "epoch": 2.16, "learning_rate": 7.839723986512978e-05, "loss": 2.7627, "step": 27550 }, { "epoch": 2.16, "learning_rate": 7.839645573590528e-05, "loss": 2.8467, "step": 27551 }, { "epoch": 2.16, "learning_rate": 7.839567160668079e-05, "loss": 2.7296, "step": 27552 }, { "epoch": 2.16, "learning_rate": 7.839488747745629e-05, "loss": 2.7333, "step": 27553 }, { "epoch": 2.16, "learning_rate": 7.83941033482318e-05, "loss": 2.7949, "step": 27554 }, { "epoch": 2.16, "learning_rate": 7.83933192190073e-05, "loss": 2.7161, "step": 27555 }, { "epoch": 2.16, "learning_rate": 7.83925350897828e-05, "loss": 2.7095, "step": 27556 }, { "epoch": 2.16, "learning_rate": 7.839175096055831e-05, "loss": 2.7832, "step": 27557 }, { "epoch": 2.16, "learning_rate": 7.839096683133381e-05, "loss": 2.6678, "step": 27558 }, { "epoch": 2.16, "learning_rate": 7.839018270210931e-05, "loss": 2.6879, "step": 27559 }, { "epoch": 2.16, "learning_rate": 7.838939857288482e-05, "loss": 2.7372, "step": 27560 }, { "epoch": 2.16, "learning_rate": 7.838861444366032e-05, "loss": 2.6059, "step": 27561 }, { "epoch": 2.16, "learning_rate": 7.838783031443581e-05, "loss": 2.758, "step": 27562 }, { "epoch": 2.16, "learning_rate": 7.838704618521132e-05, "loss": 2.5695, "step": 27563 }, { "epoch": 2.16, "learning_rate": 7.838626205598682e-05, "loss": 2.8268, "step": 27564 }, { "epoch": 2.16, "learning_rate": 7.838547792676232e-05, "loss": 2.6829, "step": 27565 }, { "epoch": 2.16, "learning_rate": 7.838469379753784e-05, "loss": 2.8151, "step": 27566 }, { "epoch": 2.16, "learning_rate": 7.838390966831334e-05, "loss": 2.9627, "step": 27567 }, { "epoch": 2.16, "learning_rate": 7.838312553908885e-05, "loss": 2.9232, "step": 27568 }, { "epoch": 2.16, "learning_rate": 7.838234140986435e-05, "loss": 2.6016, "step": 27569 }, { "epoch": 2.16, "learning_rate": 7.838155728063985e-05, "loss": 2.7946, "step": 27570 }, { "epoch": 2.16, "learning_rate": 7.838077315141536e-05, "loss": 2.7744, "step": 27571 }, { "epoch": 2.16, "learning_rate": 7.837998902219086e-05, "loss": 2.7657, "step": 27572 }, { "epoch": 2.16, "learning_rate": 7.837920489296637e-05, "loss": 2.7165, "step": 27573 }, { "epoch": 2.16, "learning_rate": 7.837842076374187e-05, "loss": 2.9034, "step": 27574 }, { "epoch": 2.16, "learning_rate": 7.837763663451737e-05, "loss": 2.7872, "step": 27575 }, { "epoch": 2.16, "learning_rate": 7.837685250529288e-05, "loss": 2.7902, "step": 27576 }, { "epoch": 2.16, "learning_rate": 7.837606837606838e-05, "loss": 2.9728, "step": 27577 }, { "epoch": 2.16, "learning_rate": 7.837528424684388e-05, "loss": 2.6856, "step": 27578 }, { "epoch": 2.16, "learning_rate": 7.837450011761939e-05, "loss": 2.7431, "step": 27579 }, { "epoch": 2.16, "learning_rate": 7.837371598839489e-05, "loss": 2.9501, "step": 27580 }, { "epoch": 2.16, "learning_rate": 7.83729318591704e-05, "loss": 2.6483, "step": 27581 }, { "epoch": 2.16, "learning_rate": 7.83721477299459e-05, "loss": 2.8366, "step": 27582 }, { "epoch": 2.16, "learning_rate": 7.83713636007214e-05, "loss": 2.7523, "step": 27583 }, { "epoch": 2.16, "learning_rate": 7.83705794714969e-05, "loss": 2.9362, "step": 27584 }, { "epoch": 2.16, "learning_rate": 7.836979534227241e-05, "loss": 2.6596, "step": 27585 }, { "epoch": 2.16, "learning_rate": 7.836901121304791e-05, "loss": 2.851, "step": 27586 }, { "epoch": 2.16, "learning_rate": 7.836822708382342e-05, "loss": 2.9597, "step": 27587 }, { "epoch": 2.16, "learning_rate": 7.836744295459892e-05, "loss": 2.599, "step": 27588 }, { "epoch": 2.16, "learning_rate": 7.836665882537442e-05, "loss": 2.6245, "step": 27589 }, { "epoch": 2.16, "learning_rate": 7.836587469614993e-05, "loss": 3.1056, "step": 27590 }, { "epoch": 2.16, "learning_rate": 7.836509056692543e-05, "loss": 2.7926, "step": 27591 }, { "epoch": 2.16, "learning_rate": 7.836430643770094e-05, "loss": 2.7799, "step": 27592 }, { "epoch": 2.16, "learning_rate": 7.836352230847644e-05, "loss": 2.6813, "step": 27593 }, { "epoch": 2.16, "learning_rate": 7.836273817925194e-05, "loss": 2.714, "step": 27594 }, { "epoch": 2.16, "learning_rate": 7.836195405002745e-05, "loss": 2.6339, "step": 27595 }, { "epoch": 2.16, "learning_rate": 7.836116992080295e-05, "loss": 2.9172, "step": 27596 }, { "epoch": 2.16, "learning_rate": 7.836038579157847e-05, "loss": 2.7392, "step": 27597 }, { "epoch": 2.16, "learning_rate": 7.835960166235396e-05, "loss": 2.644, "step": 27598 }, { "epoch": 2.16, "learning_rate": 7.835881753312946e-05, "loss": 2.8041, "step": 27599 }, { "epoch": 2.16, "learning_rate": 7.835803340390497e-05, "loss": 2.7935, "step": 27600 }, { "epoch": 2.16, "learning_rate": 7.835724927468047e-05, "loss": 2.9604, "step": 27601 }, { "epoch": 2.16, "learning_rate": 7.835646514545597e-05, "loss": 2.9959, "step": 27602 }, { "epoch": 2.16, "learning_rate": 7.835568101623148e-05, "loss": 2.8297, "step": 27603 }, { "epoch": 2.16, "learning_rate": 7.835489688700698e-05, "loss": 2.8051, "step": 27604 }, { "epoch": 2.16, "learning_rate": 7.835411275778248e-05, "loss": 2.7675, "step": 27605 }, { "epoch": 2.16, "learning_rate": 7.835332862855799e-05, "loss": 2.9663, "step": 27606 }, { "epoch": 2.16, "learning_rate": 7.835254449933349e-05, "loss": 2.7492, "step": 27607 }, { "epoch": 2.16, "learning_rate": 7.8351760370109e-05, "loss": 2.8406, "step": 27608 }, { "epoch": 2.16, "learning_rate": 7.83509762408845e-05, "loss": 3.1, "step": 27609 }, { "epoch": 2.16, "learning_rate": 7.835019211166e-05, "loss": 2.8048, "step": 27610 }, { "epoch": 2.17, "learning_rate": 7.83494079824355e-05, "loss": 2.8809, "step": 27611 }, { "epoch": 2.17, "learning_rate": 7.834862385321101e-05, "loss": 2.9685, "step": 27612 }, { "epoch": 2.17, "learning_rate": 7.834783972398651e-05, "loss": 2.5806, "step": 27613 }, { "epoch": 2.17, "learning_rate": 7.834705559476202e-05, "loss": 3.2466, "step": 27614 }, { "epoch": 2.17, "learning_rate": 7.834627146553752e-05, "loss": 2.8588, "step": 27615 }, { "epoch": 2.17, "learning_rate": 7.834548733631302e-05, "loss": 2.8906, "step": 27616 }, { "epoch": 2.17, "learning_rate": 7.834470320708853e-05, "loss": 2.7062, "step": 27617 }, { "epoch": 2.17, "learning_rate": 7.834391907786403e-05, "loss": 2.8045, "step": 27618 }, { "epoch": 2.17, "learning_rate": 7.834313494863954e-05, "loss": 2.4964, "step": 27619 }, { "epoch": 2.17, "learning_rate": 7.834235081941504e-05, "loss": 2.9114, "step": 27620 }, { "epoch": 2.17, "learning_rate": 7.834156669019054e-05, "loss": 2.9195, "step": 27621 }, { "epoch": 2.17, "learning_rate": 7.834078256096606e-05, "loss": 2.743, "step": 27622 }, { "epoch": 2.17, "learning_rate": 7.833999843174156e-05, "loss": 2.8083, "step": 27623 }, { "epoch": 2.17, "learning_rate": 7.833921430251707e-05, "loss": 2.9189, "step": 27624 }, { "epoch": 2.17, "learning_rate": 7.833843017329257e-05, "loss": 2.6462, "step": 27625 }, { "epoch": 2.17, "learning_rate": 7.833764604406807e-05, "loss": 2.7035, "step": 27626 }, { "epoch": 2.17, "learning_rate": 7.833686191484358e-05, "loss": 2.469, "step": 27627 }, { "epoch": 2.17, "learning_rate": 7.833607778561908e-05, "loss": 2.6298, "step": 27628 }, { "epoch": 2.17, "learning_rate": 7.833529365639459e-05, "loss": 2.8169, "step": 27629 }, { "epoch": 2.17, "learning_rate": 7.833450952717009e-05, "loss": 2.8786, "step": 27630 }, { "epoch": 2.17, "learning_rate": 7.833372539794559e-05, "loss": 2.8342, "step": 27631 }, { "epoch": 2.17, "learning_rate": 7.83329412687211e-05, "loss": 2.912, "step": 27632 }, { "epoch": 2.17, "learning_rate": 7.83321571394966e-05, "loss": 2.9116, "step": 27633 }, { "epoch": 2.17, "learning_rate": 7.833137301027209e-05, "loss": 2.7181, "step": 27634 }, { "epoch": 2.17, "learning_rate": 7.83305888810476e-05, "loss": 2.5018, "step": 27635 }, { "epoch": 2.17, "learning_rate": 7.83298047518231e-05, "loss": 2.8719, "step": 27636 }, { "epoch": 2.17, "learning_rate": 7.83290206225986e-05, "loss": 2.8841, "step": 27637 }, { "epoch": 2.17, "learning_rate": 7.83282364933741e-05, "loss": 2.7453, "step": 27638 }, { "epoch": 2.17, "learning_rate": 7.832745236414961e-05, "loss": 2.8528, "step": 27639 }, { "epoch": 2.17, "learning_rate": 7.832666823492511e-05, "loss": 2.8578, "step": 27640 }, { "epoch": 2.17, "learning_rate": 7.832588410570062e-05, "loss": 2.7261, "step": 27641 }, { "epoch": 2.17, "learning_rate": 7.832509997647612e-05, "loss": 2.7454, "step": 27642 }, { "epoch": 2.17, "learning_rate": 7.832431584725162e-05, "loss": 2.7321, "step": 27643 }, { "epoch": 2.17, "learning_rate": 7.832353171802713e-05, "loss": 2.5253, "step": 27644 }, { "epoch": 2.17, "learning_rate": 7.832274758880263e-05, "loss": 2.7587, "step": 27645 }, { "epoch": 2.17, "learning_rate": 7.832196345957813e-05, "loss": 2.7557, "step": 27646 }, { "epoch": 2.17, "learning_rate": 7.832117933035364e-05, "loss": 2.7113, "step": 27647 }, { "epoch": 2.17, "learning_rate": 7.832039520112916e-05, "loss": 2.8235, "step": 27648 }, { "epoch": 2.17, "learning_rate": 7.831961107190466e-05, "loss": 2.7579, "step": 27649 }, { "epoch": 2.17, "learning_rate": 7.831882694268016e-05, "loss": 2.6081, "step": 27650 }, { "epoch": 2.17, "learning_rate": 7.831804281345567e-05, "loss": 2.7049, "step": 27651 }, { "epoch": 2.17, "learning_rate": 7.831725868423117e-05, "loss": 2.7352, "step": 27652 }, { "epoch": 2.17, "learning_rate": 7.831647455500667e-05, "loss": 2.6637, "step": 27653 }, { "epoch": 2.17, "learning_rate": 7.831569042578218e-05, "loss": 3.0059, "step": 27654 }, { "epoch": 2.17, "learning_rate": 7.831490629655768e-05, "loss": 2.83, "step": 27655 }, { "epoch": 2.17, "learning_rate": 7.831412216733318e-05, "loss": 2.6211, "step": 27656 }, { "epoch": 2.17, "learning_rate": 7.831333803810869e-05, "loss": 2.7115, "step": 27657 }, { "epoch": 2.17, "learning_rate": 7.831255390888419e-05, "loss": 2.6285, "step": 27658 }, { "epoch": 2.17, "learning_rate": 7.83117697796597e-05, "loss": 2.7417, "step": 27659 }, { "epoch": 2.17, "learning_rate": 7.83109856504352e-05, "loss": 2.7652, "step": 27660 }, { "epoch": 2.17, "learning_rate": 7.83102015212107e-05, "loss": 2.9361, "step": 27661 }, { "epoch": 2.17, "learning_rate": 7.83094173919862e-05, "loss": 2.7899, "step": 27662 }, { "epoch": 2.17, "learning_rate": 7.830863326276171e-05, "loss": 2.8732, "step": 27663 }, { "epoch": 2.17, "learning_rate": 7.830784913353721e-05, "loss": 2.9484, "step": 27664 }, { "epoch": 2.17, "learning_rate": 7.830706500431272e-05, "loss": 2.7873, "step": 27665 }, { "epoch": 2.17, "learning_rate": 7.830628087508822e-05, "loss": 2.7307, "step": 27666 }, { "epoch": 2.17, "learning_rate": 7.830549674586373e-05, "loss": 2.9844, "step": 27667 }, { "epoch": 2.17, "learning_rate": 7.830471261663923e-05, "loss": 2.7269, "step": 27668 }, { "epoch": 2.17, "learning_rate": 7.830392848741473e-05, "loss": 2.5843, "step": 27669 }, { "epoch": 2.17, "learning_rate": 7.830314435819022e-05, "loss": 2.8094, "step": 27670 }, { "epoch": 2.17, "learning_rate": 7.830236022896573e-05, "loss": 2.7741, "step": 27671 }, { "epoch": 2.17, "learning_rate": 7.830157609974123e-05, "loss": 2.7377, "step": 27672 }, { "epoch": 2.17, "learning_rate": 7.830079197051673e-05, "loss": 2.9151, "step": 27673 }, { "epoch": 2.17, "learning_rate": 7.830000784129225e-05, "loss": 3.0184, "step": 27674 }, { "epoch": 2.17, "learning_rate": 7.829922371206775e-05, "loss": 2.6336, "step": 27675 }, { "epoch": 2.17, "learning_rate": 7.829843958284326e-05, "loss": 2.7683, "step": 27676 }, { "epoch": 2.17, "learning_rate": 7.829765545361876e-05, "loss": 2.6333, "step": 27677 }, { "epoch": 2.17, "learning_rate": 7.829687132439427e-05, "loss": 2.8133, "step": 27678 }, { "epoch": 2.17, "learning_rate": 7.829608719516977e-05, "loss": 2.9028, "step": 27679 }, { "epoch": 2.17, "learning_rate": 7.829530306594527e-05, "loss": 2.9663, "step": 27680 }, { "epoch": 2.17, "learning_rate": 7.829451893672078e-05, "loss": 2.815, "step": 27681 }, { "epoch": 2.17, "learning_rate": 7.829373480749628e-05, "loss": 2.9236, "step": 27682 }, { "epoch": 2.17, "learning_rate": 7.829295067827178e-05, "loss": 2.5999, "step": 27683 }, { "epoch": 2.17, "learning_rate": 7.829216654904729e-05, "loss": 2.9019, "step": 27684 }, { "epoch": 2.17, "learning_rate": 7.829138241982279e-05, "loss": 2.9248, "step": 27685 }, { "epoch": 2.17, "learning_rate": 7.82905982905983e-05, "loss": 2.659, "step": 27686 }, { "epoch": 2.17, "learning_rate": 7.82898141613738e-05, "loss": 2.7448, "step": 27687 }, { "epoch": 2.17, "learning_rate": 7.82890300321493e-05, "loss": 2.7815, "step": 27688 }, { "epoch": 2.17, "learning_rate": 7.82882459029248e-05, "loss": 2.7073, "step": 27689 }, { "epoch": 2.17, "learning_rate": 7.828746177370031e-05, "loss": 3.0765, "step": 27690 }, { "epoch": 2.17, "learning_rate": 7.828667764447581e-05, "loss": 2.8202, "step": 27691 }, { "epoch": 2.17, "learning_rate": 7.828589351525132e-05, "loss": 3.1047, "step": 27692 }, { "epoch": 2.17, "learning_rate": 7.828510938602682e-05, "loss": 2.6828, "step": 27693 }, { "epoch": 2.17, "learning_rate": 7.828432525680232e-05, "loss": 2.8479, "step": 27694 }, { "epoch": 2.17, "learning_rate": 7.828354112757783e-05, "loss": 2.6696, "step": 27695 }, { "epoch": 2.17, "learning_rate": 7.828275699835333e-05, "loss": 2.9074, "step": 27696 }, { "epoch": 2.17, "learning_rate": 7.828197286912884e-05, "loss": 2.8022, "step": 27697 }, { "epoch": 2.17, "learning_rate": 7.828118873990434e-05, "loss": 3.1064, "step": 27698 }, { "epoch": 2.17, "learning_rate": 7.828040461067984e-05, "loss": 2.5844, "step": 27699 }, { "epoch": 2.17, "learning_rate": 7.827962048145535e-05, "loss": 2.6642, "step": 27700 }, { "epoch": 2.17, "learning_rate": 7.827883635223085e-05, "loss": 2.7073, "step": 27701 }, { "epoch": 2.17, "learning_rate": 7.827805222300635e-05, "loss": 2.9341, "step": 27702 }, { "epoch": 2.17, "learning_rate": 7.827726809378186e-05, "loss": 2.8066, "step": 27703 }, { "epoch": 2.17, "learning_rate": 7.827648396455736e-05, "loss": 2.6316, "step": 27704 }, { "epoch": 2.17, "learning_rate": 7.827569983533288e-05, "loss": 2.8494, "step": 27705 }, { "epoch": 2.17, "learning_rate": 7.827491570610837e-05, "loss": 2.7809, "step": 27706 }, { "epoch": 2.17, "learning_rate": 7.827413157688387e-05, "loss": 2.905, "step": 27707 }, { "epoch": 2.17, "learning_rate": 7.827334744765938e-05, "loss": 2.747, "step": 27708 }, { "epoch": 2.17, "learning_rate": 7.827256331843488e-05, "loss": 2.734, "step": 27709 }, { "epoch": 2.17, "learning_rate": 7.827177918921038e-05, "loss": 2.8562, "step": 27710 }, { "epoch": 2.17, "learning_rate": 7.827099505998589e-05, "loss": 2.999, "step": 27711 }, { "epoch": 2.17, "learning_rate": 7.827021093076139e-05, "loss": 2.8689, "step": 27712 }, { "epoch": 2.17, "learning_rate": 7.82694268015369e-05, "loss": 2.865, "step": 27713 }, { "epoch": 2.17, "learning_rate": 7.82686426723124e-05, "loss": 2.7634, "step": 27714 }, { "epoch": 2.17, "learning_rate": 7.82678585430879e-05, "loss": 2.7555, "step": 27715 }, { "epoch": 2.17, "learning_rate": 7.82670744138634e-05, "loss": 2.9853, "step": 27716 }, { "epoch": 2.17, "learning_rate": 7.826629028463891e-05, "loss": 2.7965, "step": 27717 }, { "epoch": 2.17, "learning_rate": 7.826550615541441e-05, "loss": 2.7756, "step": 27718 }, { "epoch": 2.17, "learning_rate": 7.826472202618992e-05, "loss": 2.7517, "step": 27719 }, { "epoch": 2.17, "learning_rate": 7.826393789696542e-05, "loss": 2.7561, "step": 27720 }, { "epoch": 2.17, "learning_rate": 7.826315376774092e-05, "loss": 2.6936, "step": 27721 }, { "epoch": 2.17, "learning_rate": 7.826236963851643e-05, "loss": 2.7387, "step": 27722 }, { "epoch": 2.17, "learning_rate": 7.826158550929193e-05, "loss": 2.8958, "step": 27723 }, { "epoch": 2.17, "learning_rate": 7.826080138006743e-05, "loss": 3.0862, "step": 27724 }, { "epoch": 2.17, "learning_rate": 7.826001725084294e-05, "loss": 2.8422, "step": 27725 }, { "epoch": 2.17, "learning_rate": 7.825923312161844e-05, "loss": 2.8839, "step": 27726 }, { "epoch": 2.17, "learning_rate": 7.825844899239395e-05, "loss": 2.7011, "step": 27727 }, { "epoch": 2.17, "learning_rate": 7.825766486316945e-05, "loss": 3.0162, "step": 27728 }, { "epoch": 2.17, "learning_rate": 7.825688073394495e-05, "loss": 2.9622, "step": 27729 }, { "epoch": 2.17, "learning_rate": 7.825609660472046e-05, "loss": 2.7773, "step": 27730 }, { "epoch": 2.17, "learning_rate": 7.825531247549597e-05, "loss": 2.8535, "step": 27731 }, { "epoch": 2.17, "learning_rate": 7.825452834627148e-05, "loss": 2.9237, "step": 27732 }, { "epoch": 2.17, "learning_rate": 7.825374421704698e-05, "loss": 2.7548, "step": 27733 }, { "epoch": 2.17, "learning_rate": 7.825296008782249e-05, "loss": 2.7371, "step": 27734 }, { "epoch": 2.17, "learning_rate": 7.825217595859799e-05, "loss": 2.8587, "step": 27735 }, { "epoch": 2.17, "learning_rate": 7.825139182937349e-05, "loss": 2.9758, "step": 27736 }, { "epoch": 2.17, "learning_rate": 7.8250607700149e-05, "loss": 2.9534, "step": 27737 }, { "epoch": 2.18, "learning_rate": 7.82498235709245e-05, "loss": 2.7729, "step": 27738 }, { "epoch": 2.18, "learning_rate": 7.82490394417e-05, "loss": 2.6726, "step": 27739 }, { "epoch": 2.18, "learning_rate": 7.824825531247551e-05, "loss": 2.7946, "step": 27740 }, { "epoch": 2.18, "learning_rate": 7.824747118325101e-05, "loss": 2.6943, "step": 27741 }, { "epoch": 2.18, "learning_rate": 7.82466870540265e-05, "loss": 2.6688, "step": 27742 }, { "epoch": 2.18, "learning_rate": 7.8245902924802e-05, "loss": 2.6884, "step": 27743 }, { "epoch": 2.18, "learning_rate": 7.824511879557751e-05, "loss": 2.6338, "step": 27744 }, { "epoch": 2.18, "learning_rate": 7.824433466635301e-05, "loss": 2.8695, "step": 27745 }, { "epoch": 2.18, "learning_rate": 7.824355053712852e-05, "loss": 2.7865, "step": 27746 }, { "epoch": 2.18, "learning_rate": 7.824276640790402e-05, "loss": 2.8162, "step": 27747 }, { "epoch": 2.18, "learning_rate": 7.824198227867952e-05, "loss": 2.874, "step": 27748 }, { "epoch": 2.18, "learning_rate": 7.824119814945503e-05, "loss": 3.1057, "step": 27749 }, { "epoch": 2.18, "learning_rate": 7.824041402023053e-05, "loss": 2.6617, "step": 27750 }, { "epoch": 2.18, "learning_rate": 7.823962989100603e-05, "loss": 2.801, "step": 27751 }, { "epoch": 2.18, "learning_rate": 7.823884576178154e-05, "loss": 3.0162, "step": 27752 }, { "epoch": 2.18, "learning_rate": 7.823806163255704e-05, "loss": 2.8557, "step": 27753 }, { "epoch": 2.18, "learning_rate": 7.823727750333255e-05, "loss": 2.6154, "step": 27754 }, { "epoch": 2.18, "learning_rate": 7.823649337410805e-05, "loss": 3.0749, "step": 27755 }, { "epoch": 2.18, "learning_rate": 7.823570924488357e-05, "loss": 2.592, "step": 27756 }, { "epoch": 2.18, "learning_rate": 7.823492511565907e-05, "loss": 2.6795, "step": 27757 }, { "epoch": 2.18, "learning_rate": 7.823414098643457e-05, "loss": 3.0103, "step": 27758 }, { "epoch": 2.18, "learning_rate": 7.823335685721008e-05, "loss": 2.7025, "step": 27759 }, { "epoch": 2.18, "learning_rate": 7.823257272798558e-05, "loss": 2.7252, "step": 27760 }, { "epoch": 2.18, "learning_rate": 7.823178859876108e-05, "loss": 2.7887, "step": 27761 }, { "epoch": 2.18, "learning_rate": 7.823100446953659e-05, "loss": 2.9686, "step": 27762 }, { "epoch": 2.18, "learning_rate": 7.823022034031209e-05, "loss": 3.0756, "step": 27763 }, { "epoch": 2.18, "learning_rate": 7.82294362110876e-05, "loss": 2.4738, "step": 27764 }, { "epoch": 2.18, "learning_rate": 7.82286520818631e-05, "loss": 2.6925, "step": 27765 }, { "epoch": 2.18, "learning_rate": 7.82278679526386e-05, "loss": 2.7249, "step": 27766 }, { "epoch": 2.18, "learning_rate": 7.82270838234141e-05, "loss": 3.0421, "step": 27767 }, { "epoch": 2.18, "learning_rate": 7.822629969418961e-05, "loss": 3.0879, "step": 27768 }, { "epoch": 2.18, "learning_rate": 7.822551556496511e-05, "loss": 3.0161, "step": 27769 }, { "epoch": 2.18, "learning_rate": 7.822473143574062e-05, "loss": 2.7435, "step": 27770 }, { "epoch": 2.18, "learning_rate": 7.822394730651612e-05, "loss": 2.6947, "step": 27771 }, { "epoch": 2.18, "learning_rate": 7.822316317729162e-05, "loss": 2.7043, "step": 27772 }, { "epoch": 2.18, "learning_rate": 7.822237904806713e-05, "loss": 2.9042, "step": 27773 }, { "epoch": 2.18, "learning_rate": 7.822159491884263e-05, "loss": 2.7432, "step": 27774 }, { "epoch": 2.18, "learning_rate": 7.822081078961814e-05, "loss": 2.8812, "step": 27775 }, { "epoch": 2.18, "learning_rate": 7.822002666039364e-05, "loss": 2.7994, "step": 27776 }, { "epoch": 2.18, "learning_rate": 7.821924253116914e-05, "loss": 2.8561, "step": 27777 }, { "epoch": 2.18, "learning_rate": 7.821845840194463e-05, "loss": 2.8637, "step": 27778 }, { "epoch": 2.18, "learning_rate": 7.821767427272014e-05, "loss": 2.8065, "step": 27779 }, { "epoch": 2.18, "learning_rate": 7.821689014349564e-05, "loss": 2.9082, "step": 27780 }, { "epoch": 2.18, "learning_rate": 7.821610601427114e-05, "loss": 2.7777, "step": 27781 }, { "epoch": 2.18, "learning_rate": 7.821532188504666e-05, "loss": 2.8892, "step": 27782 }, { "epoch": 2.18, "learning_rate": 7.821453775582217e-05, "loss": 2.8777, "step": 27783 }, { "epoch": 2.18, "learning_rate": 7.821375362659767e-05, "loss": 2.9799, "step": 27784 }, { "epoch": 2.18, "learning_rate": 7.821296949737317e-05, "loss": 2.8521, "step": 27785 }, { "epoch": 2.18, "learning_rate": 7.821218536814868e-05, "loss": 2.857, "step": 27786 }, { "epoch": 2.18, "learning_rate": 7.821140123892418e-05, "loss": 2.6997, "step": 27787 }, { "epoch": 2.18, "learning_rate": 7.821061710969968e-05, "loss": 2.7743, "step": 27788 }, { "epoch": 2.18, "learning_rate": 7.820983298047519e-05, "loss": 3.073, "step": 27789 }, { "epoch": 2.18, "learning_rate": 7.820904885125069e-05, "loss": 2.7494, "step": 27790 }, { "epoch": 2.18, "learning_rate": 7.82082647220262e-05, "loss": 2.8506, "step": 27791 }, { "epoch": 2.18, "learning_rate": 7.82074805928017e-05, "loss": 2.9752, "step": 27792 }, { "epoch": 2.18, "learning_rate": 7.82066964635772e-05, "loss": 2.8054, "step": 27793 }, { "epoch": 2.18, "learning_rate": 7.82059123343527e-05, "loss": 2.7141, "step": 27794 }, { "epoch": 2.18, "learning_rate": 7.820512820512821e-05, "loss": 2.9944, "step": 27795 }, { "epoch": 2.18, "learning_rate": 7.820434407590371e-05, "loss": 2.8818, "step": 27796 }, { "epoch": 2.18, "learning_rate": 7.820355994667922e-05, "loss": 2.7034, "step": 27797 }, { "epoch": 2.18, "learning_rate": 7.820277581745472e-05, "loss": 2.9079, "step": 27798 }, { "epoch": 2.18, "learning_rate": 7.820199168823022e-05, "loss": 2.6593, "step": 27799 }, { "epoch": 2.18, "learning_rate": 7.820120755900573e-05, "loss": 3.04, "step": 27800 }, { "epoch": 2.18, "learning_rate": 7.820042342978123e-05, "loss": 2.6872, "step": 27801 }, { "epoch": 2.18, "learning_rate": 7.819963930055674e-05, "loss": 3.0289, "step": 27802 }, { "epoch": 2.18, "learning_rate": 7.819885517133224e-05, "loss": 2.7149, "step": 27803 }, { "epoch": 2.18, "learning_rate": 7.819807104210774e-05, "loss": 3.0734, "step": 27804 }, { "epoch": 2.18, "learning_rate": 7.819728691288325e-05, "loss": 2.8578, "step": 27805 }, { "epoch": 2.18, "learning_rate": 7.819650278365875e-05, "loss": 2.7482, "step": 27806 }, { "epoch": 2.18, "learning_rate": 7.819571865443425e-05, "loss": 2.863, "step": 27807 }, { "epoch": 2.18, "learning_rate": 7.819493452520976e-05, "loss": 2.4926, "step": 27808 }, { "epoch": 2.18, "learning_rate": 7.819415039598526e-05, "loss": 3.0307, "step": 27809 }, { "epoch": 2.18, "learning_rate": 7.819336626676076e-05, "loss": 2.901, "step": 27810 }, { "epoch": 2.18, "learning_rate": 7.819258213753627e-05, "loss": 2.8062, "step": 27811 }, { "epoch": 2.18, "learning_rate": 7.819179800831177e-05, "loss": 2.8649, "step": 27812 }, { "epoch": 2.18, "learning_rate": 7.819101387908729e-05, "loss": 2.7353, "step": 27813 }, { "epoch": 2.18, "learning_rate": 7.819022974986278e-05, "loss": 2.8599, "step": 27814 }, { "epoch": 2.18, "learning_rate": 7.818944562063828e-05, "loss": 2.8995, "step": 27815 }, { "epoch": 2.18, "learning_rate": 7.818866149141379e-05, "loss": 2.8179, "step": 27816 }, { "epoch": 2.18, "learning_rate": 7.818787736218929e-05, "loss": 2.9461, "step": 27817 }, { "epoch": 2.18, "learning_rate": 7.81870932329648e-05, "loss": 2.6865, "step": 27818 }, { "epoch": 2.18, "learning_rate": 7.81863091037403e-05, "loss": 2.9397, "step": 27819 }, { "epoch": 2.18, "learning_rate": 7.81855249745158e-05, "loss": 2.8339, "step": 27820 }, { "epoch": 2.18, "learning_rate": 7.81847408452913e-05, "loss": 2.6216, "step": 27821 }, { "epoch": 2.18, "learning_rate": 7.818395671606681e-05, "loss": 2.7474, "step": 27822 }, { "epoch": 2.18, "learning_rate": 7.818317258684231e-05, "loss": 2.613, "step": 27823 }, { "epoch": 2.18, "learning_rate": 7.818238845761782e-05, "loss": 2.7995, "step": 27824 }, { "epoch": 2.18, "learning_rate": 7.818160432839332e-05, "loss": 2.7809, "step": 27825 }, { "epoch": 2.18, "learning_rate": 7.818082019916882e-05, "loss": 2.8647, "step": 27826 }, { "epoch": 2.18, "learning_rate": 7.818003606994433e-05, "loss": 2.7551, "step": 27827 }, { "epoch": 2.18, "learning_rate": 7.817925194071983e-05, "loss": 2.9681, "step": 27828 }, { "epoch": 2.18, "learning_rate": 7.817846781149533e-05, "loss": 2.7441, "step": 27829 }, { "epoch": 2.18, "learning_rate": 7.817768368227084e-05, "loss": 3.0883, "step": 27830 }, { "epoch": 2.18, "learning_rate": 7.817689955304634e-05, "loss": 2.7376, "step": 27831 }, { "epoch": 2.18, "learning_rate": 7.817611542382185e-05, "loss": 2.7021, "step": 27832 }, { "epoch": 2.18, "learning_rate": 7.817533129459735e-05, "loss": 2.7046, "step": 27833 }, { "epoch": 2.18, "learning_rate": 7.817454716537285e-05, "loss": 2.6019, "step": 27834 }, { "epoch": 2.18, "learning_rate": 7.817376303614836e-05, "loss": 2.9667, "step": 27835 }, { "epoch": 2.18, "learning_rate": 7.817297890692386e-05, "loss": 2.9191, "step": 27836 }, { "epoch": 2.18, "learning_rate": 7.817219477769936e-05, "loss": 2.7143, "step": 27837 }, { "epoch": 2.18, "learning_rate": 7.817141064847487e-05, "loss": 2.955, "step": 27838 }, { "epoch": 2.18, "learning_rate": 7.817062651925038e-05, "loss": 3.0081, "step": 27839 }, { "epoch": 2.18, "learning_rate": 7.816984239002589e-05, "loss": 2.8008, "step": 27840 }, { "epoch": 2.18, "learning_rate": 7.816905826080139e-05, "loss": 2.8646, "step": 27841 }, { "epoch": 2.18, "learning_rate": 7.81682741315769e-05, "loss": 2.7056, "step": 27842 }, { "epoch": 2.18, "learning_rate": 7.81674900023524e-05, "loss": 2.756, "step": 27843 }, { "epoch": 2.18, "learning_rate": 7.81667058731279e-05, "loss": 2.6771, "step": 27844 }, { "epoch": 2.18, "learning_rate": 7.816592174390341e-05, "loss": 2.6869, "step": 27845 }, { "epoch": 2.18, "learning_rate": 7.816513761467891e-05, "loss": 2.9006, "step": 27846 }, { "epoch": 2.18, "learning_rate": 7.816435348545441e-05, "loss": 2.8724, "step": 27847 }, { "epoch": 2.18, "learning_rate": 7.816356935622992e-05, "loss": 2.8333, "step": 27848 }, { "epoch": 2.18, "learning_rate": 7.816278522700542e-05, "loss": 2.8898, "step": 27849 }, { "epoch": 2.18, "learning_rate": 7.816200109778091e-05, "loss": 3.0178, "step": 27850 }, { "epoch": 2.18, "learning_rate": 7.816121696855642e-05, "loss": 2.8692, "step": 27851 }, { "epoch": 2.18, "learning_rate": 7.816043283933192e-05, "loss": 2.8715, "step": 27852 }, { "epoch": 2.18, "learning_rate": 7.815964871010742e-05, "loss": 2.6198, "step": 27853 }, { "epoch": 2.18, "learning_rate": 7.815886458088293e-05, "loss": 3.0392, "step": 27854 }, { "epoch": 2.18, "learning_rate": 7.815808045165843e-05, "loss": 2.7042, "step": 27855 }, { "epoch": 2.18, "learning_rate": 7.815729632243393e-05, "loss": 2.8365, "step": 27856 }, { "epoch": 2.18, "learning_rate": 7.815651219320944e-05, "loss": 2.7643, "step": 27857 }, { "epoch": 2.18, "learning_rate": 7.815572806398494e-05, "loss": 2.8858, "step": 27858 }, { "epoch": 2.18, "learning_rate": 7.815494393476044e-05, "loss": 2.6067, "step": 27859 }, { "epoch": 2.18, "learning_rate": 7.815415980553595e-05, "loss": 2.8416, "step": 27860 }, { "epoch": 2.18, "learning_rate": 7.815337567631145e-05, "loss": 2.8668, "step": 27861 }, { "epoch": 2.18, "learning_rate": 7.815259154708696e-05, "loss": 2.7493, "step": 27862 }, { "epoch": 2.18, "learning_rate": 7.815180741786246e-05, "loss": 2.9057, "step": 27863 }, { "epoch": 2.18, "learning_rate": 7.815102328863798e-05, "loss": 2.7579, "step": 27864 }, { "epoch": 2.18, "learning_rate": 7.815023915941348e-05, "loss": 2.5025, "step": 27865 }, { "epoch": 2.19, "learning_rate": 7.814945503018898e-05, "loss": 3.0631, "step": 27866 }, { "epoch": 2.19, "learning_rate": 7.814867090096449e-05, "loss": 2.7015, "step": 27867 }, { "epoch": 2.19, "learning_rate": 7.814788677173999e-05, "loss": 2.6284, "step": 27868 }, { "epoch": 2.19, "learning_rate": 7.81471026425155e-05, "loss": 2.6, "step": 27869 }, { "epoch": 2.19, "learning_rate": 7.8146318513291e-05, "loss": 2.6523, "step": 27870 }, { "epoch": 2.19, "learning_rate": 7.81455343840665e-05, "loss": 2.7027, "step": 27871 }, { "epoch": 2.19, "learning_rate": 7.8144750254842e-05, "loss": 2.7867, "step": 27872 }, { "epoch": 2.19, "learning_rate": 7.814396612561751e-05, "loss": 2.8117, "step": 27873 }, { "epoch": 2.19, "learning_rate": 7.814318199639301e-05, "loss": 2.5888, "step": 27874 }, { "epoch": 2.19, "learning_rate": 7.814239786716852e-05, "loss": 2.6361, "step": 27875 }, { "epoch": 2.19, "learning_rate": 7.814161373794402e-05, "loss": 2.801, "step": 27876 }, { "epoch": 2.19, "learning_rate": 7.814082960871952e-05, "loss": 2.6446, "step": 27877 }, { "epoch": 2.19, "learning_rate": 7.814004547949503e-05, "loss": 3.0726, "step": 27878 }, { "epoch": 2.19, "learning_rate": 7.813926135027053e-05, "loss": 2.6596, "step": 27879 }, { "epoch": 2.19, "learning_rate": 7.813847722104604e-05, "loss": 2.7192, "step": 27880 }, { "epoch": 2.19, "learning_rate": 7.813769309182154e-05, "loss": 2.7976, "step": 27881 }, { "epoch": 2.19, "learning_rate": 7.813690896259704e-05, "loss": 2.8535, "step": 27882 }, { "epoch": 2.19, "learning_rate": 7.813612483337255e-05, "loss": 2.8565, "step": 27883 }, { "epoch": 2.19, "learning_rate": 7.813534070414805e-05, "loss": 2.9373, "step": 27884 }, { "epoch": 2.19, "learning_rate": 7.813455657492355e-05, "loss": 2.8194, "step": 27885 }, { "epoch": 2.19, "learning_rate": 7.813377244569904e-05, "loss": 2.9499, "step": 27886 }, { "epoch": 2.19, "learning_rate": 7.813298831647455e-05, "loss": 2.7682, "step": 27887 }, { "epoch": 2.19, "learning_rate": 7.813220418725005e-05, "loss": 2.7597, "step": 27888 }, { "epoch": 2.19, "learning_rate": 7.813142005802556e-05, "loss": 2.7282, "step": 27889 }, { "epoch": 2.19, "learning_rate": 7.813063592880107e-05, "loss": 3.0998, "step": 27890 }, { "epoch": 2.19, "learning_rate": 7.812985179957658e-05, "loss": 2.6234, "step": 27891 }, { "epoch": 2.19, "learning_rate": 7.812906767035208e-05, "loss": 2.6839, "step": 27892 }, { "epoch": 2.19, "learning_rate": 7.812828354112758e-05, "loss": 2.6803, "step": 27893 }, { "epoch": 2.19, "learning_rate": 7.812749941190309e-05, "loss": 2.7281, "step": 27894 }, { "epoch": 2.19, "learning_rate": 7.812671528267859e-05, "loss": 2.6272, "step": 27895 }, { "epoch": 2.19, "learning_rate": 7.81259311534541e-05, "loss": 2.7408, "step": 27896 }, { "epoch": 2.19, "learning_rate": 7.81251470242296e-05, "loss": 2.8081, "step": 27897 }, { "epoch": 2.19, "learning_rate": 7.81243628950051e-05, "loss": 2.698, "step": 27898 }, { "epoch": 2.19, "learning_rate": 7.81235787657806e-05, "loss": 2.655, "step": 27899 }, { "epoch": 2.19, "learning_rate": 7.812279463655611e-05, "loss": 2.695, "step": 27900 }, { "epoch": 2.19, "learning_rate": 7.812201050733161e-05, "loss": 2.7076, "step": 27901 }, { "epoch": 2.19, "learning_rate": 7.812122637810712e-05, "loss": 2.9316, "step": 27902 }, { "epoch": 2.19, "learning_rate": 7.812044224888262e-05, "loss": 2.6682, "step": 27903 }, { "epoch": 2.19, "learning_rate": 7.811965811965812e-05, "loss": 3.0322, "step": 27904 }, { "epoch": 2.19, "learning_rate": 7.811887399043363e-05, "loss": 2.9349, "step": 27905 }, { "epoch": 2.19, "learning_rate": 7.811808986120913e-05, "loss": 2.8139, "step": 27906 }, { "epoch": 2.19, "learning_rate": 7.811730573198463e-05, "loss": 2.795, "step": 27907 }, { "epoch": 2.19, "learning_rate": 7.811652160276014e-05, "loss": 2.8432, "step": 27908 }, { "epoch": 2.19, "learning_rate": 7.811573747353564e-05, "loss": 2.6386, "step": 27909 }, { "epoch": 2.19, "learning_rate": 7.811495334431115e-05, "loss": 2.8776, "step": 27910 }, { "epoch": 2.19, "learning_rate": 7.811416921508665e-05, "loss": 2.6957, "step": 27911 }, { "epoch": 2.19, "learning_rate": 7.811338508586215e-05, "loss": 2.622, "step": 27912 }, { "epoch": 2.19, "learning_rate": 7.811260095663766e-05, "loss": 2.8305, "step": 27913 }, { "epoch": 2.19, "learning_rate": 7.811181682741316e-05, "loss": 2.763, "step": 27914 }, { "epoch": 2.19, "learning_rate": 7.811103269818866e-05, "loss": 2.5873, "step": 27915 }, { "epoch": 2.19, "learning_rate": 7.811024856896417e-05, "loss": 2.6166, "step": 27916 }, { "epoch": 2.19, "learning_rate": 7.810946443973967e-05, "loss": 2.7384, "step": 27917 }, { "epoch": 2.19, "learning_rate": 7.810868031051518e-05, "loss": 2.9728, "step": 27918 }, { "epoch": 2.19, "learning_rate": 7.810789618129068e-05, "loss": 2.7528, "step": 27919 }, { "epoch": 2.19, "learning_rate": 7.810711205206618e-05, "loss": 3.1314, "step": 27920 }, { "epoch": 2.19, "learning_rate": 7.81063279228417e-05, "loss": 2.9341, "step": 27921 }, { "epoch": 2.19, "learning_rate": 7.810554379361719e-05, "loss": 2.6467, "step": 27922 }, { "epoch": 2.19, "learning_rate": 7.81047596643927e-05, "loss": 2.791, "step": 27923 }, { "epoch": 2.19, "learning_rate": 7.81039755351682e-05, "loss": 3.127, "step": 27924 }, { "epoch": 2.19, "learning_rate": 7.81031914059437e-05, "loss": 3.1744, "step": 27925 }, { "epoch": 2.19, "learning_rate": 7.81024072767192e-05, "loss": 2.9682, "step": 27926 }, { "epoch": 2.19, "learning_rate": 7.810162314749471e-05, "loss": 2.7327, "step": 27927 }, { "epoch": 2.19, "learning_rate": 7.810083901827021e-05, "loss": 2.6319, "step": 27928 }, { "epoch": 2.19, "learning_rate": 7.810005488904572e-05, "loss": 2.8638, "step": 27929 }, { "epoch": 2.19, "learning_rate": 7.809927075982122e-05, "loss": 2.7214, "step": 27930 }, { "epoch": 2.19, "learning_rate": 7.809848663059672e-05, "loss": 2.9741, "step": 27931 }, { "epoch": 2.19, "learning_rate": 7.809770250137223e-05, "loss": 2.7295, "step": 27932 }, { "epoch": 2.19, "learning_rate": 7.809691837214773e-05, "loss": 2.9524, "step": 27933 }, { "epoch": 2.19, "learning_rate": 7.809613424292323e-05, "loss": 2.7302, "step": 27934 }, { "epoch": 2.19, "learning_rate": 7.809535011369874e-05, "loss": 2.9483, "step": 27935 }, { "epoch": 2.19, "learning_rate": 7.809456598447424e-05, "loss": 2.6507, "step": 27936 }, { "epoch": 2.19, "learning_rate": 7.809378185524975e-05, "loss": 2.8548, "step": 27937 }, { "epoch": 2.19, "learning_rate": 7.809299772602525e-05, "loss": 2.8954, "step": 27938 }, { "epoch": 2.19, "learning_rate": 7.809221359680075e-05, "loss": 2.7488, "step": 27939 }, { "epoch": 2.19, "learning_rate": 7.809142946757626e-05, "loss": 2.7215, "step": 27940 }, { "epoch": 2.19, "learning_rate": 7.809064533835176e-05, "loss": 2.752, "step": 27941 }, { "epoch": 2.19, "learning_rate": 7.808986120912726e-05, "loss": 2.8237, "step": 27942 }, { "epoch": 2.19, "learning_rate": 7.808907707990277e-05, "loss": 2.6489, "step": 27943 }, { "epoch": 2.19, "learning_rate": 7.808829295067827e-05, "loss": 2.7793, "step": 27944 }, { "epoch": 2.19, "learning_rate": 7.808750882145377e-05, "loss": 2.7353, "step": 27945 }, { "epoch": 2.19, "learning_rate": 7.808672469222928e-05, "loss": 3.1067, "step": 27946 }, { "epoch": 2.19, "learning_rate": 7.80859405630048e-05, "loss": 2.9127, "step": 27947 }, { "epoch": 2.19, "learning_rate": 7.80851564337803e-05, "loss": 2.9751, "step": 27948 }, { "epoch": 2.19, "learning_rate": 7.80843723045558e-05, "loss": 2.8683, "step": 27949 }, { "epoch": 2.19, "learning_rate": 7.80835881753313e-05, "loss": 2.8052, "step": 27950 }, { "epoch": 2.19, "learning_rate": 7.808280404610681e-05, "loss": 2.8002, "step": 27951 }, { "epoch": 2.19, "learning_rate": 7.808201991688231e-05, "loss": 2.6963, "step": 27952 }, { "epoch": 2.19, "learning_rate": 7.808123578765782e-05, "loss": 2.8631, "step": 27953 }, { "epoch": 2.19, "learning_rate": 7.808045165843332e-05, "loss": 2.8945, "step": 27954 }, { "epoch": 2.19, "learning_rate": 7.807966752920882e-05, "loss": 2.7349, "step": 27955 }, { "epoch": 2.19, "learning_rate": 7.807888339998433e-05, "loss": 2.8319, "step": 27956 }, { "epoch": 2.19, "learning_rate": 7.807809927075983e-05, "loss": 2.9108, "step": 27957 }, { "epoch": 2.19, "learning_rate": 7.807731514153532e-05, "loss": 2.7303, "step": 27958 }, { "epoch": 2.19, "learning_rate": 7.807653101231083e-05, "loss": 2.8549, "step": 27959 }, { "epoch": 2.19, "learning_rate": 7.807574688308633e-05, "loss": 2.77, "step": 27960 }, { "epoch": 2.19, "learning_rate": 7.807496275386183e-05, "loss": 2.7252, "step": 27961 }, { "epoch": 2.19, "learning_rate": 7.807417862463734e-05, "loss": 2.6498, "step": 27962 }, { "epoch": 2.19, "learning_rate": 7.807339449541284e-05, "loss": 2.7939, "step": 27963 }, { "epoch": 2.19, "learning_rate": 7.807261036618834e-05, "loss": 2.8493, "step": 27964 }, { "epoch": 2.19, "learning_rate": 7.807182623696385e-05, "loss": 2.7663, "step": 27965 }, { "epoch": 2.19, "learning_rate": 7.807104210773935e-05, "loss": 2.7918, "step": 27966 }, { "epoch": 2.19, "learning_rate": 7.807025797851486e-05, "loss": 3.0202, "step": 27967 }, { "epoch": 2.19, "learning_rate": 7.806947384929036e-05, "loss": 2.8029, "step": 27968 }, { "epoch": 2.19, "learning_rate": 7.806868972006586e-05, "loss": 2.8099, "step": 27969 }, { "epoch": 2.19, "learning_rate": 7.806790559084137e-05, "loss": 2.8234, "step": 27970 }, { "epoch": 2.19, "learning_rate": 7.806712146161687e-05, "loss": 2.8357, "step": 27971 }, { "epoch": 2.19, "learning_rate": 7.806633733239239e-05, "loss": 2.903, "step": 27972 }, { "epoch": 2.19, "learning_rate": 7.806555320316789e-05, "loss": 2.6546, "step": 27973 }, { "epoch": 2.19, "learning_rate": 7.80647690739434e-05, "loss": 2.8555, "step": 27974 }, { "epoch": 2.19, "learning_rate": 7.80639849447189e-05, "loss": 2.7096, "step": 27975 }, { "epoch": 2.19, "learning_rate": 7.80632008154944e-05, "loss": 2.7793, "step": 27976 }, { "epoch": 2.19, "learning_rate": 7.80624166862699e-05, "loss": 2.806, "step": 27977 }, { "epoch": 2.19, "learning_rate": 7.806163255704541e-05, "loss": 2.9554, "step": 27978 }, { "epoch": 2.19, "learning_rate": 7.806084842782091e-05, "loss": 2.7791, "step": 27979 }, { "epoch": 2.19, "learning_rate": 7.806006429859642e-05, "loss": 2.9862, "step": 27980 }, { "epoch": 2.19, "learning_rate": 7.805928016937192e-05, "loss": 2.6931, "step": 27981 }, { "epoch": 2.19, "learning_rate": 7.805849604014742e-05, "loss": 2.8864, "step": 27982 }, { "epoch": 2.19, "learning_rate": 7.805771191092293e-05, "loss": 2.8942, "step": 27983 }, { "epoch": 2.19, "learning_rate": 7.805692778169843e-05, "loss": 2.8412, "step": 27984 }, { "epoch": 2.19, "learning_rate": 7.805614365247394e-05, "loss": 2.7516, "step": 27985 }, { "epoch": 2.19, "learning_rate": 7.805535952324944e-05, "loss": 3.0905, "step": 27986 }, { "epoch": 2.19, "learning_rate": 7.805457539402494e-05, "loss": 2.7817, "step": 27987 }, { "epoch": 2.19, "learning_rate": 7.805379126480045e-05, "loss": 2.7154, "step": 27988 }, { "epoch": 2.19, "learning_rate": 7.805300713557595e-05, "loss": 2.9778, "step": 27989 }, { "epoch": 2.19, "learning_rate": 7.805222300635145e-05, "loss": 3.2361, "step": 27990 }, { "epoch": 2.19, "learning_rate": 7.805143887712696e-05, "loss": 2.8367, "step": 27991 }, { "epoch": 2.19, "learning_rate": 7.805065474790246e-05, "loss": 2.947, "step": 27992 }, { "epoch": 2.2, "learning_rate": 7.804987061867796e-05, "loss": 2.8463, "step": 27993 }, { "epoch": 2.2, "learning_rate": 7.804908648945345e-05, "loss": 2.7389, "step": 27994 }, { "epoch": 2.2, "learning_rate": 7.804830236022896e-05, "loss": 2.9658, "step": 27995 }, { "epoch": 2.2, "learning_rate": 7.804751823100446e-05, "loss": 2.8811, "step": 27996 }, { "epoch": 2.2, "learning_rate": 7.804673410177997e-05, "loss": 2.7489, "step": 27997 }, { "epoch": 2.2, "learning_rate": 7.804594997255548e-05, "loss": 2.6039, "step": 27998 }, { "epoch": 2.2, "learning_rate": 7.804516584333099e-05, "loss": 2.7777, "step": 27999 }, { "epoch": 2.2, "learning_rate": 7.804438171410649e-05, "loss": 2.739, "step": 28000 }, { "epoch": 2.2, "eval_loss": 2.9025726318359375, "eval_runtime": 217.9788, "eval_samples_per_second": 51.987, "eval_steps_per_second": 3.253, "step": 28000 }, { "epoch": 2.2, "learning_rate": 7.8043597584882e-05, "loss": 2.673, "step": 28001 }, { "epoch": 2.2, "learning_rate": 7.80428134556575e-05, "loss": 3.004, "step": 28002 }, { "epoch": 2.2, "learning_rate": 7.8042029326433e-05, "loss": 2.6266, "step": 28003 }, { "epoch": 2.2, "learning_rate": 7.80412451972085e-05, "loss": 2.6936, "step": 28004 }, { "epoch": 2.2, "learning_rate": 7.804046106798401e-05, "loss": 2.7555, "step": 28005 }, { "epoch": 2.2, "learning_rate": 7.803967693875951e-05, "loss": 3.0682, "step": 28006 }, { "epoch": 2.2, "learning_rate": 7.803889280953502e-05, "loss": 3.0092, "step": 28007 }, { "epoch": 2.2, "learning_rate": 7.803810868031052e-05, "loss": 2.7591, "step": 28008 }, { "epoch": 2.2, "learning_rate": 7.803732455108602e-05, "loss": 2.7977, "step": 28009 }, { "epoch": 2.2, "learning_rate": 7.803654042186153e-05, "loss": 2.7055, "step": 28010 }, { "epoch": 2.2, "learning_rate": 7.803575629263703e-05, "loss": 2.8071, "step": 28011 }, { "epoch": 2.2, "learning_rate": 7.803497216341253e-05, "loss": 2.6911, "step": 28012 }, { "epoch": 2.2, "learning_rate": 7.803418803418804e-05, "loss": 2.8712, "step": 28013 }, { "epoch": 2.2, "learning_rate": 7.803340390496354e-05, "loss": 2.6028, "step": 28014 }, { "epoch": 2.2, "learning_rate": 7.803261977573905e-05, "loss": 2.6534, "step": 28015 }, { "epoch": 2.2, "learning_rate": 7.803183564651455e-05, "loss": 2.8725, "step": 28016 }, { "epoch": 2.2, "learning_rate": 7.803105151729005e-05, "loss": 2.8097, "step": 28017 }, { "epoch": 2.2, "learning_rate": 7.803026738806556e-05, "loss": 2.4534, "step": 28018 }, { "epoch": 2.2, "learning_rate": 7.802948325884106e-05, "loss": 2.8777, "step": 28019 }, { "epoch": 2.2, "learning_rate": 7.802869912961656e-05, "loss": 2.9948, "step": 28020 }, { "epoch": 2.2, "learning_rate": 7.802791500039207e-05, "loss": 2.8124, "step": 28021 }, { "epoch": 2.2, "learning_rate": 7.802713087116757e-05, "loss": 2.9711, "step": 28022 }, { "epoch": 2.2, "learning_rate": 7.802634674194308e-05, "loss": 3.0132, "step": 28023 }, { "epoch": 2.2, "learning_rate": 7.802556261271858e-05, "loss": 2.7818, "step": 28024 }, { "epoch": 2.2, "learning_rate": 7.802477848349408e-05, "loss": 2.7134, "step": 28025 }, { "epoch": 2.2, "learning_rate": 7.802399435426959e-05, "loss": 2.6912, "step": 28026 }, { "epoch": 2.2, "learning_rate": 7.802321022504509e-05, "loss": 2.8348, "step": 28027 }, { "epoch": 2.2, "learning_rate": 7.80224260958206e-05, "loss": 2.8013, "step": 28028 }, { "epoch": 2.2, "learning_rate": 7.802164196659611e-05, "loss": 2.7059, "step": 28029 }, { "epoch": 2.2, "learning_rate": 7.80208578373716e-05, "loss": 2.864, "step": 28030 }, { "epoch": 2.2, "learning_rate": 7.80200737081471e-05, "loss": 2.6075, "step": 28031 }, { "epoch": 2.2, "learning_rate": 7.801928957892261e-05, "loss": 2.6733, "step": 28032 }, { "epoch": 2.2, "learning_rate": 7.801850544969811e-05, "loss": 2.8592, "step": 28033 }, { "epoch": 2.2, "learning_rate": 7.801772132047362e-05, "loss": 2.7447, "step": 28034 }, { "epoch": 2.2, "learning_rate": 7.801693719124912e-05, "loss": 2.8312, "step": 28035 }, { "epoch": 2.2, "learning_rate": 7.801615306202462e-05, "loss": 2.858, "step": 28036 }, { "epoch": 2.2, "learning_rate": 7.801536893280013e-05, "loss": 2.6021, "step": 28037 }, { "epoch": 2.2, "learning_rate": 7.801458480357563e-05, "loss": 2.8, "step": 28038 }, { "epoch": 2.2, "learning_rate": 7.801380067435113e-05, "loss": 2.852, "step": 28039 }, { "epoch": 2.2, "learning_rate": 7.801301654512664e-05, "loss": 2.7918, "step": 28040 }, { "epoch": 2.2, "learning_rate": 7.801223241590214e-05, "loss": 3.0512, "step": 28041 }, { "epoch": 2.2, "learning_rate": 7.801144828667764e-05, "loss": 2.7569, "step": 28042 }, { "epoch": 2.2, "learning_rate": 7.801066415745315e-05, "loss": 2.886, "step": 28043 }, { "epoch": 2.2, "learning_rate": 7.800988002822865e-05, "loss": 2.9052, "step": 28044 }, { "epoch": 2.2, "learning_rate": 7.800909589900416e-05, "loss": 2.7819, "step": 28045 }, { "epoch": 2.2, "learning_rate": 7.800831176977966e-05, "loss": 3.0075, "step": 28046 }, { "epoch": 2.2, "learning_rate": 7.800752764055516e-05, "loss": 3.0178, "step": 28047 }, { "epoch": 2.2, "learning_rate": 7.800674351133067e-05, "loss": 2.8286, "step": 28048 }, { "epoch": 2.2, "learning_rate": 7.800595938210617e-05, "loss": 2.718, "step": 28049 }, { "epoch": 2.2, "learning_rate": 7.800517525288167e-05, "loss": 2.6602, "step": 28050 }, { "epoch": 2.2, "learning_rate": 7.800439112365718e-05, "loss": 2.8183, "step": 28051 }, { "epoch": 2.2, "learning_rate": 7.800360699443268e-05, "loss": 2.8696, "step": 28052 }, { "epoch": 2.2, "learning_rate": 7.800282286520819e-05, "loss": 2.8296, "step": 28053 }, { "epoch": 2.2, "learning_rate": 7.800203873598369e-05, "loss": 2.8209, "step": 28054 }, { "epoch": 2.2, "learning_rate": 7.80012546067592e-05, "loss": 2.7752, "step": 28055 }, { "epoch": 2.2, "learning_rate": 7.800047047753471e-05, "loss": 2.9457, "step": 28056 }, { "epoch": 2.2, "learning_rate": 7.799968634831021e-05, "loss": 2.6512, "step": 28057 }, { "epoch": 2.2, "learning_rate": 7.799890221908572e-05, "loss": 2.9477, "step": 28058 }, { "epoch": 2.2, "learning_rate": 7.799811808986122e-05, "loss": 2.6052, "step": 28059 }, { "epoch": 2.2, "learning_rate": 7.799733396063672e-05, "loss": 2.8874, "step": 28060 }, { "epoch": 2.2, "learning_rate": 7.799654983141223e-05, "loss": 2.8935, "step": 28061 }, { "epoch": 2.2, "learning_rate": 7.799576570218773e-05, "loss": 2.8758, "step": 28062 }, { "epoch": 2.2, "learning_rate": 7.799498157296324e-05, "loss": 2.9315, "step": 28063 }, { "epoch": 2.2, "learning_rate": 7.799419744373874e-05, "loss": 2.9046, "step": 28064 }, { "epoch": 2.2, "learning_rate": 7.799341331451424e-05, "loss": 2.7635, "step": 28065 }, { "epoch": 2.2, "learning_rate": 7.799262918528973e-05, "loss": 2.7466, "step": 28066 }, { "epoch": 2.2, "learning_rate": 7.799184505606524e-05, "loss": 2.8381, "step": 28067 }, { "epoch": 2.2, "learning_rate": 7.799106092684074e-05, "loss": 2.7283, "step": 28068 }, { "epoch": 2.2, "learning_rate": 7.799027679761624e-05, "loss": 2.6719, "step": 28069 }, { "epoch": 2.2, "learning_rate": 7.798949266839175e-05, "loss": 2.6264, "step": 28070 }, { "epoch": 2.2, "learning_rate": 7.798870853916725e-05, "loss": 2.8004, "step": 28071 }, { "epoch": 2.2, "learning_rate": 7.798792440994276e-05, "loss": 2.839, "step": 28072 }, { "epoch": 2.2, "learning_rate": 7.798714028071826e-05, "loss": 2.7469, "step": 28073 }, { "epoch": 2.2, "learning_rate": 7.798635615149376e-05, "loss": 2.8704, "step": 28074 }, { "epoch": 2.2, "learning_rate": 7.798557202226927e-05, "loss": 2.8111, "step": 28075 }, { "epoch": 2.2, "learning_rate": 7.798478789304477e-05, "loss": 3.0103, "step": 28076 }, { "epoch": 2.2, "learning_rate": 7.798400376382027e-05, "loss": 2.842, "step": 28077 }, { "epoch": 2.2, "learning_rate": 7.798321963459578e-05, "loss": 2.9876, "step": 28078 }, { "epoch": 2.2, "learning_rate": 7.798243550537128e-05, "loss": 2.8468, "step": 28079 }, { "epoch": 2.2, "learning_rate": 7.79816513761468e-05, "loss": 2.7659, "step": 28080 }, { "epoch": 2.2, "learning_rate": 7.79808672469223e-05, "loss": 2.9432, "step": 28081 }, { "epoch": 2.2, "learning_rate": 7.79800831176978e-05, "loss": 2.5791, "step": 28082 }, { "epoch": 2.2, "learning_rate": 7.797929898847331e-05, "loss": 2.8536, "step": 28083 }, { "epoch": 2.2, "learning_rate": 7.797851485924881e-05, "loss": 2.8313, "step": 28084 }, { "epoch": 2.2, "learning_rate": 7.797773073002432e-05, "loss": 2.971, "step": 28085 }, { "epoch": 2.2, "learning_rate": 7.797694660079982e-05, "loss": 2.7666, "step": 28086 }, { "epoch": 2.2, "learning_rate": 7.797616247157532e-05, "loss": 2.9767, "step": 28087 }, { "epoch": 2.2, "learning_rate": 7.797537834235083e-05, "loss": 2.7635, "step": 28088 }, { "epoch": 2.2, "learning_rate": 7.797459421312633e-05, "loss": 2.6245, "step": 28089 }, { "epoch": 2.2, "learning_rate": 7.797381008390183e-05, "loss": 2.8058, "step": 28090 }, { "epoch": 2.2, "learning_rate": 7.797302595467734e-05, "loss": 2.9903, "step": 28091 }, { "epoch": 2.2, "learning_rate": 7.797224182545284e-05, "loss": 2.9194, "step": 28092 }, { "epoch": 2.2, "learning_rate": 7.797145769622835e-05, "loss": 2.835, "step": 28093 }, { "epoch": 2.2, "learning_rate": 7.797067356700385e-05, "loss": 2.7367, "step": 28094 }, { "epoch": 2.2, "learning_rate": 7.796988943777935e-05, "loss": 2.7358, "step": 28095 }, { "epoch": 2.2, "learning_rate": 7.796910530855486e-05, "loss": 2.9765, "step": 28096 }, { "epoch": 2.2, "learning_rate": 7.796832117933036e-05, "loss": 2.782, "step": 28097 }, { "epoch": 2.2, "learning_rate": 7.796753705010586e-05, "loss": 2.6616, "step": 28098 }, { "epoch": 2.2, "learning_rate": 7.796675292088137e-05, "loss": 2.7572, "step": 28099 }, { "epoch": 2.2, "learning_rate": 7.796596879165687e-05, "loss": 2.6195, "step": 28100 }, { "epoch": 2.2, "learning_rate": 7.796518466243238e-05, "loss": 2.6134, "step": 28101 }, { "epoch": 2.2, "learning_rate": 7.796440053320787e-05, "loss": 2.8266, "step": 28102 }, { "epoch": 2.2, "learning_rate": 7.796361640398337e-05, "loss": 2.9808, "step": 28103 }, { "epoch": 2.2, "learning_rate": 7.796283227475887e-05, "loss": 3.0654, "step": 28104 }, { "epoch": 2.2, "learning_rate": 7.796204814553438e-05, "loss": 2.7327, "step": 28105 }, { "epoch": 2.2, "learning_rate": 7.79612640163099e-05, "loss": 2.6627, "step": 28106 }, { "epoch": 2.2, "learning_rate": 7.79604798870854e-05, "loss": 2.8902, "step": 28107 }, { "epoch": 2.2, "learning_rate": 7.79596957578609e-05, "loss": 2.7758, "step": 28108 }, { "epoch": 2.2, "learning_rate": 7.79589116286364e-05, "loss": 2.6498, "step": 28109 }, { "epoch": 2.2, "learning_rate": 7.795812749941191e-05, "loss": 2.9235, "step": 28110 }, { "epoch": 2.2, "learning_rate": 7.795734337018741e-05, "loss": 2.7476, "step": 28111 }, { "epoch": 2.2, "learning_rate": 7.795655924096292e-05, "loss": 2.8039, "step": 28112 }, { "epoch": 2.2, "learning_rate": 7.795577511173842e-05, "loss": 2.7349, "step": 28113 }, { "epoch": 2.2, "learning_rate": 7.795499098251392e-05, "loss": 3.0232, "step": 28114 }, { "epoch": 2.2, "learning_rate": 7.795420685328943e-05, "loss": 2.6546, "step": 28115 }, { "epoch": 2.2, "learning_rate": 7.795342272406493e-05, "loss": 2.6042, "step": 28116 }, { "epoch": 2.2, "learning_rate": 7.795263859484043e-05, "loss": 2.6825, "step": 28117 }, { "epoch": 2.2, "learning_rate": 7.795185446561594e-05, "loss": 3.0242, "step": 28118 }, { "epoch": 2.2, "learning_rate": 7.795107033639144e-05, "loss": 2.928, "step": 28119 }, { "epoch": 2.2, "learning_rate": 7.795028620716695e-05, "loss": 2.7132, "step": 28120 }, { "epoch": 2.21, "learning_rate": 7.794950207794245e-05, "loss": 2.7353, "step": 28121 }, { "epoch": 2.21, "learning_rate": 7.794871794871795e-05, "loss": 3.1272, "step": 28122 }, { "epoch": 2.21, "learning_rate": 7.794793381949346e-05, "loss": 2.8277, "step": 28123 }, { "epoch": 2.21, "learning_rate": 7.794714969026896e-05, "loss": 3.1961, "step": 28124 }, { "epoch": 2.21, "learning_rate": 7.794636556104446e-05, "loss": 2.5811, "step": 28125 }, { "epoch": 2.21, "learning_rate": 7.794558143181997e-05, "loss": 2.5638, "step": 28126 }, { "epoch": 2.21, "learning_rate": 7.794479730259547e-05, "loss": 2.7262, "step": 28127 }, { "epoch": 2.21, "learning_rate": 7.794401317337097e-05, "loss": 2.8532, "step": 28128 }, { "epoch": 2.21, "learning_rate": 7.794322904414648e-05, "loss": 2.7816, "step": 28129 }, { "epoch": 2.21, "learning_rate": 7.794244491492198e-05, "loss": 2.7628, "step": 28130 }, { "epoch": 2.21, "learning_rate": 7.794166078569749e-05, "loss": 2.6929, "step": 28131 }, { "epoch": 2.21, "learning_rate": 7.794087665647299e-05, "loss": 2.779, "step": 28132 }, { "epoch": 2.21, "learning_rate": 7.794009252724849e-05, "loss": 2.7602, "step": 28133 }, { "epoch": 2.21, "learning_rate": 7.7939308398024e-05, "loss": 2.6468, "step": 28134 }, { "epoch": 2.21, "learning_rate": 7.79385242687995e-05, "loss": 2.7355, "step": 28135 }, { "epoch": 2.21, "learning_rate": 7.7937740139575e-05, "loss": 2.7956, "step": 28136 }, { "epoch": 2.21, "learning_rate": 7.793695601035052e-05, "loss": 2.734, "step": 28137 }, { "epoch": 2.21, "learning_rate": 7.793617188112601e-05, "loss": 3.1064, "step": 28138 }, { "epoch": 2.21, "learning_rate": 7.793538775190152e-05, "loss": 2.9659, "step": 28139 }, { "epoch": 2.21, "learning_rate": 7.793460362267702e-05, "loss": 3.0907, "step": 28140 }, { "epoch": 2.21, "learning_rate": 7.793381949345252e-05, "loss": 2.832, "step": 28141 }, { "epoch": 2.21, "learning_rate": 7.793303536422803e-05, "loss": 2.8291, "step": 28142 }, { "epoch": 2.21, "learning_rate": 7.793225123500353e-05, "loss": 2.8126, "step": 28143 }, { "epoch": 2.21, "learning_rate": 7.793146710577903e-05, "loss": 2.801, "step": 28144 }, { "epoch": 2.21, "learning_rate": 7.793068297655454e-05, "loss": 2.9723, "step": 28145 }, { "epoch": 2.21, "learning_rate": 7.792989884733004e-05, "loss": 2.8768, "step": 28146 }, { "epoch": 2.21, "learning_rate": 7.792911471810554e-05, "loss": 2.7547, "step": 28147 }, { "epoch": 2.21, "learning_rate": 7.792833058888105e-05, "loss": 2.6744, "step": 28148 }, { "epoch": 2.21, "learning_rate": 7.792754645965655e-05, "loss": 2.8029, "step": 28149 }, { "epoch": 2.21, "learning_rate": 7.792676233043206e-05, "loss": 2.6854, "step": 28150 }, { "epoch": 2.21, "learning_rate": 7.792597820120756e-05, "loss": 2.5725, "step": 28151 }, { "epoch": 2.21, "learning_rate": 7.792519407198306e-05, "loss": 2.7895, "step": 28152 }, { "epoch": 2.21, "learning_rate": 7.792440994275857e-05, "loss": 2.7668, "step": 28153 }, { "epoch": 2.21, "learning_rate": 7.792362581353407e-05, "loss": 2.9381, "step": 28154 }, { "epoch": 2.21, "learning_rate": 7.792284168430957e-05, "loss": 2.777, "step": 28155 }, { "epoch": 2.21, "learning_rate": 7.792205755508508e-05, "loss": 2.8187, "step": 28156 }, { "epoch": 2.21, "learning_rate": 7.792127342586058e-05, "loss": 2.9807, "step": 28157 }, { "epoch": 2.21, "learning_rate": 7.792048929663609e-05, "loss": 2.6077, "step": 28158 }, { "epoch": 2.21, "learning_rate": 7.791970516741159e-05, "loss": 3.0225, "step": 28159 }, { "epoch": 2.21, "learning_rate": 7.791892103818709e-05, "loss": 2.6702, "step": 28160 }, { "epoch": 2.21, "learning_rate": 7.79181369089626e-05, "loss": 2.8484, "step": 28161 }, { "epoch": 2.21, "learning_rate": 7.79173527797381e-05, "loss": 2.7979, "step": 28162 }, { "epoch": 2.21, "learning_rate": 7.791656865051362e-05, "loss": 2.8737, "step": 28163 }, { "epoch": 2.21, "learning_rate": 7.791578452128912e-05, "loss": 2.549, "step": 28164 }, { "epoch": 2.21, "learning_rate": 7.791500039206462e-05, "loss": 2.6946, "step": 28165 }, { "epoch": 2.21, "learning_rate": 7.791421626284013e-05, "loss": 2.7088, "step": 28166 }, { "epoch": 2.21, "learning_rate": 7.791343213361563e-05, "loss": 2.8732, "step": 28167 }, { "epoch": 2.21, "learning_rate": 7.791264800439114e-05, "loss": 2.9059, "step": 28168 }, { "epoch": 2.21, "learning_rate": 7.791186387516664e-05, "loss": 3.1283, "step": 28169 }, { "epoch": 2.21, "learning_rate": 7.791107974594214e-05, "loss": 2.824, "step": 28170 }, { "epoch": 2.21, "learning_rate": 7.791029561671765e-05, "loss": 2.9469, "step": 28171 }, { "epoch": 2.21, "learning_rate": 7.790951148749315e-05, "loss": 2.8256, "step": 28172 }, { "epoch": 2.21, "learning_rate": 7.790872735826865e-05, "loss": 2.6256, "step": 28173 }, { "epoch": 2.21, "learning_rate": 7.790794322904414e-05, "loss": 2.7879, "step": 28174 }, { "epoch": 2.21, "learning_rate": 7.790715909981965e-05, "loss": 2.8076, "step": 28175 }, { "epoch": 2.21, "learning_rate": 7.790637497059515e-05, "loss": 2.8522, "step": 28176 }, { "epoch": 2.21, "learning_rate": 7.790559084137065e-05, "loss": 2.647, "step": 28177 }, { "epoch": 2.21, "learning_rate": 7.790480671214616e-05, "loss": 2.7667, "step": 28178 }, { "epoch": 2.21, "learning_rate": 7.790402258292166e-05, "loss": 2.6856, "step": 28179 }, { "epoch": 2.21, "learning_rate": 7.790323845369717e-05, "loss": 2.8408, "step": 28180 }, { "epoch": 2.21, "learning_rate": 7.790245432447267e-05, "loss": 2.9044, "step": 28181 }, { "epoch": 2.21, "learning_rate": 7.790167019524817e-05, "loss": 2.8626, "step": 28182 }, { "epoch": 2.21, "learning_rate": 7.790088606602368e-05, "loss": 2.648, "step": 28183 }, { "epoch": 2.21, "learning_rate": 7.790010193679918e-05, "loss": 2.5881, "step": 28184 }, { "epoch": 2.21, "learning_rate": 7.789931780757468e-05, "loss": 2.8326, "step": 28185 }, { "epoch": 2.21, "learning_rate": 7.789853367835019e-05, "loss": 2.7798, "step": 28186 }, { "epoch": 2.21, "learning_rate": 7.789774954912569e-05, "loss": 2.8918, "step": 28187 }, { "epoch": 2.21, "learning_rate": 7.789696541990121e-05, "loss": 2.8904, "step": 28188 }, { "epoch": 2.21, "learning_rate": 7.789618129067671e-05, "loss": 3.2014, "step": 28189 }, { "epoch": 2.21, "learning_rate": 7.789539716145222e-05, "loss": 3.1561, "step": 28190 }, { "epoch": 2.21, "learning_rate": 7.789461303222772e-05, "loss": 2.9854, "step": 28191 }, { "epoch": 2.21, "learning_rate": 7.789382890300322e-05, "loss": 2.7818, "step": 28192 }, { "epoch": 2.21, "learning_rate": 7.789304477377873e-05, "loss": 2.9333, "step": 28193 }, { "epoch": 2.21, "learning_rate": 7.789226064455423e-05, "loss": 2.7248, "step": 28194 }, { "epoch": 2.21, "learning_rate": 7.789147651532973e-05, "loss": 2.5718, "step": 28195 }, { "epoch": 2.21, "learning_rate": 7.789069238610524e-05, "loss": 2.8631, "step": 28196 }, { "epoch": 2.21, "learning_rate": 7.788990825688074e-05, "loss": 2.9709, "step": 28197 }, { "epoch": 2.21, "learning_rate": 7.788912412765625e-05, "loss": 2.683, "step": 28198 }, { "epoch": 2.21, "learning_rate": 7.788833999843175e-05, "loss": 2.5979, "step": 28199 }, { "epoch": 2.21, "learning_rate": 7.788755586920725e-05, "loss": 2.8744, "step": 28200 }, { "epoch": 2.21, "learning_rate": 7.788677173998276e-05, "loss": 2.7351, "step": 28201 }, { "epoch": 2.21, "learning_rate": 7.788598761075826e-05, "loss": 2.807, "step": 28202 }, { "epoch": 2.21, "learning_rate": 7.788520348153376e-05, "loss": 2.6701, "step": 28203 }, { "epoch": 2.21, "learning_rate": 7.788441935230927e-05, "loss": 2.9078, "step": 28204 }, { "epoch": 2.21, "learning_rate": 7.788363522308477e-05, "loss": 2.6202, "step": 28205 }, { "epoch": 2.21, "learning_rate": 7.788285109386028e-05, "loss": 2.6563, "step": 28206 }, { "epoch": 2.21, "learning_rate": 7.788206696463578e-05, "loss": 2.9326, "step": 28207 }, { "epoch": 2.21, "learning_rate": 7.788128283541128e-05, "loss": 2.7457, "step": 28208 }, { "epoch": 2.21, "learning_rate": 7.788049870618679e-05, "loss": 2.8147, "step": 28209 }, { "epoch": 2.21, "learning_rate": 7.787971457696228e-05, "loss": 2.7155, "step": 28210 }, { "epoch": 2.21, "learning_rate": 7.787893044773778e-05, "loss": 2.7655, "step": 28211 }, { "epoch": 2.21, "learning_rate": 7.787814631851328e-05, "loss": 2.9063, "step": 28212 }, { "epoch": 2.21, "learning_rate": 7.787736218928879e-05, "loss": 2.8266, "step": 28213 }, { "epoch": 2.21, "learning_rate": 7.78765780600643e-05, "loss": 2.5562, "step": 28214 }, { "epoch": 2.21, "learning_rate": 7.787579393083981e-05, "loss": 2.9476, "step": 28215 }, { "epoch": 2.21, "learning_rate": 7.787500980161531e-05, "loss": 2.5936, "step": 28216 }, { "epoch": 2.21, "learning_rate": 7.787422567239082e-05, "loss": 2.8511, "step": 28217 }, { "epoch": 2.21, "learning_rate": 7.787344154316632e-05, "loss": 3.1303, "step": 28218 }, { "epoch": 2.21, "learning_rate": 7.787265741394182e-05, "loss": 2.9017, "step": 28219 }, { "epoch": 2.21, "learning_rate": 7.787187328471733e-05, "loss": 2.9476, "step": 28220 }, { "epoch": 2.21, "learning_rate": 7.787108915549283e-05, "loss": 2.9015, "step": 28221 }, { "epoch": 2.21, "learning_rate": 7.787030502626833e-05, "loss": 2.7921, "step": 28222 }, { "epoch": 2.21, "learning_rate": 7.786952089704384e-05, "loss": 2.7512, "step": 28223 }, { "epoch": 2.21, "learning_rate": 7.786873676781934e-05, "loss": 2.8522, "step": 28224 }, { "epoch": 2.21, "learning_rate": 7.786795263859485e-05, "loss": 2.9486, "step": 28225 }, { "epoch": 2.21, "learning_rate": 7.786716850937035e-05, "loss": 2.6322, "step": 28226 }, { "epoch": 2.21, "learning_rate": 7.786638438014585e-05, "loss": 2.8629, "step": 28227 }, { "epoch": 2.21, "learning_rate": 7.786560025092136e-05, "loss": 2.8875, "step": 28228 }, { "epoch": 2.21, "learning_rate": 7.786481612169686e-05, "loss": 2.8414, "step": 28229 }, { "epoch": 2.21, "learning_rate": 7.786403199247236e-05, "loss": 2.7984, "step": 28230 }, { "epoch": 2.21, "learning_rate": 7.786324786324787e-05, "loss": 3.0942, "step": 28231 }, { "epoch": 2.21, "learning_rate": 7.786246373402337e-05, "loss": 2.6164, "step": 28232 }, { "epoch": 2.21, "learning_rate": 7.786167960479887e-05, "loss": 2.8536, "step": 28233 }, { "epoch": 2.21, "learning_rate": 7.786089547557438e-05, "loss": 3.0726, "step": 28234 }, { "epoch": 2.21, "learning_rate": 7.786011134634988e-05, "loss": 2.7413, "step": 28235 }, { "epoch": 2.21, "learning_rate": 7.785932721712539e-05, "loss": 2.6103, "step": 28236 }, { "epoch": 2.21, "learning_rate": 7.785854308790089e-05, "loss": 2.7992, "step": 28237 }, { "epoch": 2.21, "learning_rate": 7.785775895867639e-05, "loss": 2.9716, "step": 28238 }, { "epoch": 2.21, "learning_rate": 7.78569748294519e-05, "loss": 2.6304, "step": 28239 }, { "epoch": 2.21, "learning_rate": 7.78561907002274e-05, "loss": 2.6814, "step": 28240 }, { "epoch": 2.21, "learning_rate": 7.78554065710029e-05, "loss": 2.7105, "step": 28241 }, { "epoch": 2.21, "learning_rate": 7.785462244177841e-05, "loss": 2.5325, "step": 28242 }, { "epoch": 2.21, "learning_rate": 7.785383831255391e-05, "loss": 2.9298, "step": 28243 }, { "epoch": 2.21, "learning_rate": 7.785305418332941e-05, "loss": 2.6446, "step": 28244 }, { "epoch": 2.21, "learning_rate": 7.785227005410493e-05, "loss": 2.8036, "step": 28245 }, { "epoch": 2.21, "learning_rate": 7.785148592488042e-05, "loss": 2.9562, "step": 28246 }, { "epoch": 2.21, "learning_rate": 7.785070179565593e-05, "loss": 2.8165, "step": 28247 }, { "epoch": 2.22, "learning_rate": 7.784991766643143e-05, "loss": 2.8945, "step": 28248 }, { "epoch": 2.22, "learning_rate": 7.784913353720693e-05, "loss": 2.6724, "step": 28249 }, { "epoch": 2.22, "learning_rate": 7.784834940798244e-05, "loss": 2.812, "step": 28250 }, { "epoch": 2.22, "learning_rate": 7.784756527875794e-05, "loss": 2.7064, "step": 28251 }, { "epoch": 2.22, "learning_rate": 7.784678114953344e-05, "loss": 2.8189, "step": 28252 }, { "epoch": 2.22, "learning_rate": 7.784599702030895e-05, "loss": 2.8698, "step": 28253 }, { "epoch": 2.22, "learning_rate": 7.784521289108445e-05, "loss": 2.6235, "step": 28254 }, { "epoch": 2.22, "learning_rate": 7.784442876185996e-05, "loss": 2.6998, "step": 28255 }, { "epoch": 2.22, "learning_rate": 7.784364463263546e-05, "loss": 2.7144, "step": 28256 }, { "epoch": 2.22, "learning_rate": 7.784286050341096e-05, "loss": 2.7466, "step": 28257 }, { "epoch": 2.22, "learning_rate": 7.784207637418647e-05, "loss": 3.0338, "step": 28258 }, { "epoch": 2.22, "learning_rate": 7.784129224496197e-05, "loss": 2.7919, "step": 28259 }, { "epoch": 2.22, "learning_rate": 7.784050811573747e-05, "loss": 2.8994, "step": 28260 }, { "epoch": 2.22, "learning_rate": 7.783972398651298e-05, "loss": 2.845, "step": 28261 }, { "epoch": 2.22, "learning_rate": 7.783893985728848e-05, "loss": 2.8375, "step": 28262 }, { "epoch": 2.22, "learning_rate": 7.783815572806398e-05, "loss": 2.6384, "step": 28263 }, { "epoch": 2.22, "learning_rate": 7.783737159883949e-05, "loss": 2.7758, "step": 28264 }, { "epoch": 2.22, "learning_rate": 7.783658746961499e-05, "loss": 2.5815, "step": 28265 }, { "epoch": 2.22, "learning_rate": 7.78358033403905e-05, "loss": 2.851, "step": 28266 }, { "epoch": 2.22, "learning_rate": 7.7835019211166e-05, "loss": 2.9737, "step": 28267 }, { "epoch": 2.22, "learning_rate": 7.78342350819415e-05, "loss": 2.7557, "step": 28268 }, { "epoch": 2.22, "learning_rate": 7.783345095271701e-05, "loss": 2.9467, "step": 28269 }, { "epoch": 2.22, "learning_rate": 7.783266682349251e-05, "loss": 2.7614, "step": 28270 }, { "epoch": 2.22, "learning_rate": 7.783188269426803e-05, "loss": 2.6877, "step": 28271 }, { "epoch": 2.22, "learning_rate": 7.783109856504353e-05, "loss": 2.8286, "step": 28272 }, { "epoch": 2.22, "learning_rate": 7.783031443581904e-05, "loss": 2.6569, "step": 28273 }, { "epoch": 2.22, "learning_rate": 7.782953030659454e-05, "loss": 2.7957, "step": 28274 }, { "epoch": 2.22, "learning_rate": 7.782874617737004e-05, "loss": 2.943, "step": 28275 }, { "epoch": 2.22, "learning_rate": 7.782796204814555e-05, "loss": 2.7888, "step": 28276 }, { "epoch": 2.22, "learning_rate": 7.782717791892105e-05, "loss": 2.6603, "step": 28277 }, { "epoch": 2.22, "learning_rate": 7.782639378969655e-05, "loss": 2.9531, "step": 28278 }, { "epoch": 2.22, "learning_rate": 7.782560966047206e-05, "loss": 2.8022, "step": 28279 }, { "epoch": 2.22, "learning_rate": 7.782482553124756e-05, "loss": 3.0103, "step": 28280 }, { "epoch": 2.22, "learning_rate": 7.782404140202306e-05, "loss": 2.8084, "step": 28281 }, { "epoch": 2.22, "learning_rate": 7.782325727279855e-05, "loss": 2.708, "step": 28282 }, { "epoch": 2.22, "learning_rate": 7.782247314357406e-05, "loss": 2.6943, "step": 28283 }, { "epoch": 2.22, "learning_rate": 7.782168901434956e-05, "loss": 2.7283, "step": 28284 }, { "epoch": 2.22, "learning_rate": 7.782090488512507e-05, "loss": 2.8042, "step": 28285 }, { "epoch": 2.22, "learning_rate": 7.782012075590057e-05, "loss": 2.8156, "step": 28286 }, { "epoch": 2.22, "learning_rate": 7.781933662667607e-05, "loss": 2.7689, "step": 28287 }, { "epoch": 2.22, "learning_rate": 7.781855249745158e-05, "loss": 2.8027, "step": 28288 }, { "epoch": 2.22, "learning_rate": 7.781776836822708e-05, "loss": 2.7554, "step": 28289 }, { "epoch": 2.22, "learning_rate": 7.781698423900258e-05, "loss": 2.9149, "step": 28290 }, { "epoch": 2.22, "learning_rate": 7.781620010977809e-05, "loss": 3.0009, "step": 28291 }, { "epoch": 2.22, "learning_rate": 7.781541598055359e-05, "loss": 2.887, "step": 28292 }, { "epoch": 2.22, "learning_rate": 7.78146318513291e-05, "loss": 2.7004, "step": 28293 }, { "epoch": 2.22, "learning_rate": 7.78138477221046e-05, "loss": 2.751, "step": 28294 }, { "epoch": 2.22, "learning_rate": 7.78130635928801e-05, "loss": 2.607, "step": 28295 }, { "epoch": 2.22, "learning_rate": 7.781227946365562e-05, "loss": 2.6818, "step": 28296 }, { "epoch": 2.22, "learning_rate": 7.781149533443112e-05, "loss": 2.883, "step": 28297 }, { "epoch": 2.22, "learning_rate": 7.781071120520663e-05, "loss": 2.658, "step": 28298 }, { "epoch": 2.22, "learning_rate": 7.780992707598213e-05, "loss": 2.7269, "step": 28299 }, { "epoch": 2.22, "learning_rate": 7.780914294675763e-05, "loss": 2.7186, "step": 28300 }, { "epoch": 2.22, "learning_rate": 7.780835881753314e-05, "loss": 2.7829, "step": 28301 }, { "epoch": 2.22, "learning_rate": 7.780757468830864e-05, "loss": 2.6807, "step": 28302 }, { "epoch": 2.22, "learning_rate": 7.780679055908415e-05, "loss": 2.877, "step": 28303 }, { "epoch": 2.22, "learning_rate": 7.780600642985965e-05, "loss": 2.8385, "step": 28304 }, { "epoch": 2.22, "learning_rate": 7.780522230063515e-05, "loss": 2.6698, "step": 28305 }, { "epoch": 2.22, "learning_rate": 7.780443817141066e-05, "loss": 2.7423, "step": 28306 }, { "epoch": 2.22, "learning_rate": 7.780365404218616e-05, "loss": 2.683, "step": 28307 }, { "epoch": 2.22, "learning_rate": 7.780286991296166e-05, "loss": 2.7183, "step": 28308 }, { "epoch": 2.22, "learning_rate": 7.780208578373717e-05, "loss": 2.7052, "step": 28309 }, { "epoch": 2.22, "learning_rate": 7.780130165451267e-05, "loss": 2.7703, "step": 28310 }, { "epoch": 2.22, "learning_rate": 7.780051752528817e-05, "loss": 2.8155, "step": 28311 }, { "epoch": 2.22, "learning_rate": 7.779973339606368e-05, "loss": 2.8524, "step": 28312 }, { "epoch": 2.22, "learning_rate": 7.779894926683918e-05, "loss": 3.1129, "step": 28313 }, { "epoch": 2.22, "learning_rate": 7.779816513761469e-05, "loss": 2.8592, "step": 28314 }, { "epoch": 2.22, "learning_rate": 7.779738100839019e-05, "loss": 2.6167, "step": 28315 }, { "epoch": 2.22, "learning_rate": 7.77965968791657e-05, "loss": 2.6472, "step": 28316 }, { "epoch": 2.22, "learning_rate": 7.77958127499412e-05, "loss": 2.9447, "step": 28317 }, { "epoch": 2.22, "learning_rate": 7.779502862071669e-05, "loss": 2.9142, "step": 28318 }, { "epoch": 2.22, "learning_rate": 7.779424449149219e-05, "loss": 2.8927, "step": 28319 }, { "epoch": 2.22, "learning_rate": 7.77934603622677e-05, "loss": 2.7355, "step": 28320 }, { "epoch": 2.22, "learning_rate": 7.77926762330432e-05, "loss": 2.7414, "step": 28321 }, { "epoch": 2.22, "learning_rate": 7.779189210381872e-05, "loss": 2.7656, "step": 28322 }, { "epoch": 2.22, "learning_rate": 7.779110797459422e-05, "loss": 2.8181, "step": 28323 }, { "epoch": 2.22, "learning_rate": 7.779032384536972e-05, "loss": 2.708, "step": 28324 }, { "epoch": 2.22, "learning_rate": 7.778953971614523e-05, "loss": 2.9244, "step": 28325 }, { "epoch": 2.22, "learning_rate": 7.778875558692073e-05, "loss": 2.6552, "step": 28326 }, { "epoch": 2.22, "learning_rate": 7.778797145769623e-05, "loss": 2.8507, "step": 28327 }, { "epoch": 2.22, "learning_rate": 7.778718732847174e-05, "loss": 2.4913, "step": 28328 }, { "epoch": 2.22, "learning_rate": 7.778640319924724e-05, "loss": 3.286, "step": 28329 }, { "epoch": 2.22, "learning_rate": 7.778561907002274e-05, "loss": 2.8017, "step": 28330 }, { "epoch": 2.22, "learning_rate": 7.778483494079825e-05, "loss": 2.7409, "step": 28331 }, { "epoch": 2.22, "learning_rate": 7.778405081157375e-05, "loss": 2.678, "step": 28332 }, { "epoch": 2.22, "learning_rate": 7.778326668234926e-05, "loss": 2.9584, "step": 28333 }, { "epoch": 2.22, "learning_rate": 7.778248255312476e-05, "loss": 2.6626, "step": 28334 }, { "epoch": 2.22, "learning_rate": 7.778169842390026e-05, "loss": 2.8213, "step": 28335 }, { "epoch": 2.22, "learning_rate": 7.778091429467577e-05, "loss": 2.6635, "step": 28336 }, { "epoch": 2.22, "learning_rate": 7.778013016545127e-05, "loss": 2.6364, "step": 28337 }, { "epoch": 2.22, "learning_rate": 7.777934603622677e-05, "loss": 2.8117, "step": 28338 }, { "epoch": 2.22, "learning_rate": 7.777856190700228e-05, "loss": 2.8574, "step": 28339 }, { "epoch": 2.22, "learning_rate": 7.777777777777778e-05, "loss": 2.9668, "step": 28340 }, { "epoch": 2.22, "learning_rate": 7.777699364855329e-05, "loss": 2.9987, "step": 28341 }, { "epoch": 2.22, "learning_rate": 7.777620951932879e-05, "loss": 2.9843, "step": 28342 }, { "epoch": 2.22, "learning_rate": 7.777542539010429e-05, "loss": 2.6711, "step": 28343 }, { "epoch": 2.22, "learning_rate": 7.77746412608798e-05, "loss": 2.88, "step": 28344 }, { "epoch": 2.22, "learning_rate": 7.77738571316553e-05, "loss": 2.8018, "step": 28345 }, { "epoch": 2.22, "learning_rate": 7.77730730024308e-05, "loss": 2.7277, "step": 28346 }, { "epoch": 2.22, "learning_rate": 7.777228887320631e-05, "loss": 2.7109, "step": 28347 }, { "epoch": 2.22, "learning_rate": 7.777150474398181e-05, "loss": 2.7878, "step": 28348 }, { "epoch": 2.22, "learning_rate": 7.777072061475731e-05, "loss": 2.8447, "step": 28349 }, { "epoch": 2.22, "learning_rate": 7.776993648553282e-05, "loss": 2.8173, "step": 28350 }, { "epoch": 2.22, "learning_rate": 7.776915235630832e-05, "loss": 3.1054, "step": 28351 }, { "epoch": 2.22, "learning_rate": 7.776836822708383e-05, "loss": 2.8641, "step": 28352 }, { "epoch": 2.22, "learning_rate": 7.776758409785934e-05, "loss": 2.8459, "step": 28353 }, { "epoch": 2.22, "learning_rate": 7.776679996863485e-05, "loss": 2.7482, "step": 28354 }, { "epoch": 2.22, "learning_rate": 7.776601583941034e-05, "loss": 2.7027, "step": 28355 }, { "epoch": 2.22, "learning_rate": 7.776523171018584e-05, "loss": 2.8572, "step": 28356 }, { "epoch": 2.22, "learning_rate": 7.776444758096134e-05, "loss": 2.8925, "step": 28357 }, { "epoch": 2.22, "learning_rate": 7.776366345173685e-05, "loss": 2.6218, "step": 28358 }, { "epoch": 2.22, "learning_rate": 7.776287932251235e-05, "loss": 2.7209, "step": 28359 }, { "epoch": 2.22, "learning_rate": 7.776209519328786e-05, "loss": 2.8834, "step": 28360 }, { "epoch": 2.22, "learning_rate": 7.776131106406336e-05, "loss": 3.0334, "step": 28361 }, { "epoch": 2.22, "learning_rate": 7.776052693483886e-05, "loss": 2.7226, "step": 28362 }, { "epoch": 2.22, "learning_rate": 7.775974280561437e-05, "loss": 2.839, "step": 28363 }, { "epoch": 2.22, "learning_rate": 7.775895867638987e-05, "loss": 2.7845, "step": 28364 }, { "epoch": 2.22, "learning_rate": 7.775817454716537e-05, "loss": 2.7666, "step": 28365 }, { "epoch": 2.22, "learning_rate": 7.775739041794088e-05, "loss": 2.7848, "step": 28366 }, { "epoch": 2.22, "learning_rate": 7.775660628871638e-05, "loss": 3.0308, "step": 28367 }, { "epoch": 2.22, "learning_rate": 7.775582215949188e-05, "loss": 2.9047, "step": 28368 }, { "epoch": 2.22, "learning_rate": 7.775503803026739e-05, "loss": 2.9949, "step": 28369 }, { "epoch": 2.22, "learning_rate": 7.775425390104289e-05, "loss": 2.64, "step": 28370 }, { "epoch": 2.22, "learning_rate": 7.77534697718184e-05, "loss": 3.1407, "step": 28371 }, { "epoch": 2.22, "learning_rate": 7.77526856425939e-05, "loss": 2.6226, "step": 28372 }, { "epoch": 2.22, "learning_rate": 7.77519015133694e-05, "loss": 2.7685, "step": 28373 }, { "epoch": 2.22, "learning_rate": 7.77511173841449e-05, "loss": 2.9348, "step": 28374 }, { "epoch": 2.22, "learning_rate": 7.775033325492041e-05, "loss": 2.8715, "step": 28375 }, { "epoch": 2.23, "learning_rate": 7.774954912569591e-05, "loss": 2.7926, "step": 28376 }, { "epoch": 2.23, "learning_rate": 7.774876499647142e-05, "loss": 2.8546, "step": 28377 }, { "epoch": 2.23, "learning_rate": 7.774798086724692e-05, "loss": 2.7273, "step": 28378 }, { "epoch": 2.23, "learning_rate": 7.774719673802244e-05, "loss": 2.6275, "step": 28379 }, { "epoch": 2.23, "learning_rate": 7.774641260879794e-05, "loss": 2.6956, "step": 28380 }, { "epoch": 2.23, "learning_rate": 7.774562847957345e-05, "loss": 2.9209, "step": 28381 }, { "epoch": 2.23, "learning_rate": 7.774484435034895e-05, "loss": 2.5686, "step": 28382 }, { "epoch": 2.23, "learning_rate": 7.774406022112445e-05, "loss": 2.8374, "step": 28383 }, { "epoch": 2.23, "learning_rate": 7.774327609189996e-05, "loss": 2.7216, "step": 28384 }, { "epoch": 2.23, "learning_rate": 7.774249196267546e-05, "loss": 2.8021, "step": 28385 }, { "epoch": 2.23, "learning_rate": 7.774170783345096e-05, "loss": 2.9498, "step": 28386 }, { "epoch": 2.23, "learning_rate": 7.774092370422647e-05, "loss": 2.8857, "step": 28387 }, { "epoch": 2.23, "learning_rate": 7.774013957500197e-05, "loss": 2.8601, "step": 28388 }, { "epoch": 2.23, "learning_rate": 7.773935544577748e-05, "loss": 2.8586, "step": 28389 }, { "epoch": 2.23, "learning_rate": 7.773857131655298e-05, "loss": 2.7885, "step": 28390 }, { "epoch": 2.23, "learning_rate": 7.773778718732847e-05, "loss": 2.7451, "step": 28391 }, { "epoch": 2.23, "learning_rate": 7.773700305810397e-05, "loss": 2.9365, "step": 28392 }, { "epoch": 2.23, "learning_rate": 7.773621892887948e-05, "loss": 2.599, "step": 28393 }, { "epoch": 2.23, "learning_rate": 7.773543479965498e-05, "loss": 2.7503, "step": 28394 }, { "epoch": 2.23, "learning_rate": 7.773465067043048e-05, "loss": 2.874, "step": 28395 }, { "epoch": 2.23, "learning_rate": 7.773386654120599e-05, "loss": 2.5663, "step": 28396 }, { "epoch": 2.23, "learning_rate": 7.773308241198149e-05, "loss": 2.7318, "step": 28397 }, { "epoch": 2.23, "learning_rate": 7.7732298282757e-05, "loss": 2.9457, "step": 28398 }, { "epoch": 2.23, "learning_rate": 7.77315141535325e-05, "loss": 2.7182, "step": 28399 }, { "epoch": 2.23, "learning_rate": 7.7730730024308e-05, "loss": 2.713, "step": 28400 }, { "epoch": 2.23, "learning_rate": 7.77299458950835e-05, "loss": 2.7947, "step": 28401 }, { "epoch": 2.23, "learning_rate": 7.772916176585901e-05, "loss": 2.8422, "step": 28402 }, { "epoch": 2.23, "learning_rate": 7.772837763663451e-05, "loss": 2.7967, "step": 28403 }, { "epoch": 2.23, "learning_rate": 7.772759350741003e-05, "loss": 2.867, "step": 28404 }, { "epoch": 2.23, "learning_rate": 7.772680937818553e-05, "loss": 2.927, "step": 28405 }, { "epoch": 2.23, "learning_rate": 7.772602524896104e-05, "loss": 2.5454, "step": 28406 }, { "epoch": 2.23, "learning_rate": 7.772524111973654e-05, "loss": 2.7706, "step": 28407 }, { "epoch": 2.23, "learning_rate": 7.772445699051205e-05, "loss": 2.825, "step": 28408 }, { "epoch": 2.23, "learning_rate": 7.772367286128755e-05, "loss": 2.5962, "step": 28409 }, { "epoch": 2.23, "learning_rate": 7.772288873206305e-05, "loss": 2.7301, "step": 28410 }, { "epoch": 2.23, "learning_rate": 7.772210460283856e-05, "loss": 2.9498, "step": 28411 }, { "epoch": 2.23, "learning_rate": 7.772132047361406e-05, "loss": 2.7124, "step": 28412 }, { "epoch": 2.23, "learning_rate": 7.772053634438956e-05, "loss": 2.9488, "step": 28413 }, { "epoch": 2.23, "learning_rate": 7.771975221516507e-05, "loss": 2.7348, "step": 28414 }, { "epoch": 2.23, "learning_rate": 7.771896808594057e-05, "loss": 2.7087, "step": 28415 }, { "epoch": 2.23, "learning_rate": 7.771818395671607e-05, "loss": 2.8089, "step": 28416 }, { "epoch": 2.23, "learning_rate": 7.771739982749158e-05, "loss": 2.5918, "step": 28417 }, { "epoch": 2.23, "learning_rate": 7.771661569826708e-05, "loss": 2.9862, "step": 28418 }, { "epoch": 2.23, "learning_rate": 7.771583156904259e-05, "loss": 2.8628, "step": 28419 }, { "epoch": 2.23, "learning_rate": 7.771504743981809e-05, "loss": 2.9684, "step": 28420 }, { "epoch": 2.23, "learning_rate": 7.771426331059359e-05, "loss": 2.7112, "step": 28421 }, { "epoch": 2.23, "learning_rate": 7.77134791813691e-05, "loss": 2.7871, "step": 28422 }, { "epoch": 2.23, "learning_rate": 7.77126950521446e-05, "loss": 3.0747, "step": 28423 }, { "epoch": 2.23, "learning_rate": 7.77119109229201e-05, "loss": 2.6986, "step": 28424 }, { "epoch": 2.23, "learning_rate": 7.771112679369561e-05, "loss": 2.8082, "step": 28425 }, { "epoch": 2.23, "learning_rate": 7.771034266447111e-05, "loss": 3.0291, "step": 28426 }, { "epoch": 2.23, "learning_rate": 7.77095585352466e-05, "loss": 2.6124, "step": 28427 }, { "epoch": 2.23, "learning_rate": 7.77087744060221e-05, "loss": 2.6578, "step": 28428 }, { "epoch": 2.23, "learning_rate": 7.770799027679761e-05, "loss": 2.6641, "step": 28429 }, { "epoch": 2.23, "learning_rate": 7.770720614757313e-05, "loss": 2.6426, "step": 28430 }, { "epoch": 2.23, "learning_rate": 7.770642201834863e-05, "loss": 2.8416, "step": 28431 }, { "epoch": 2.23, "learning_rate": 7.770563788912413e-05, "loss": 2.7396, "step": 28432 }, { "epoch": 2.23, "learning_rate": 7.770485375989964e-05, "loss": 3.1, "step": 28433 }, { "epoch": 2.23, "learning_rate": 7.770406963067514e-05, "loss": 2.7178, "step": 28434 }, { "epoch": 2.23, "learning_rate": 7.770328550145064e-05, "loss": 2.856, "step": 28435 }, { "epoch": 2.23, "learning_rate": 7.770250137222615e-05, "loss": 2.7167, "step": 28436 }, { "epoch": 2.23, "learning_rate": 7.770171724300165e-05, "loss": 2.8237, "step": 28437 }, { "epoch": 2.23, "learning_rate": 7.770093311377716e-05, "loss": 2.6801, "step": 28438 }, { "epoch": 2.23, "learning_rate": 7.770014898455266e-05, "loss": 2.617, "step": 28439 }, { "epoch": 2.23, "learning_rate": 7.769936485532816e-05, "loss": 2.7374, "step": 28440 }, { "epoch": 2.23, "learning_rate": 7.769858072610367e-05, "loss": 2.6217, "step": 28441 }, { "epoch": 2.23, "learning_rate": 7.769779659687917e-05, "loss": 2.6966, "step": 28442 }, { "epoch": 2.23, "learning_rate": 7.769701246765467e-05, "loss": 2.8123, "step": 28443 }, { "epoch": 2.23, "learning_rate": 7.769622833843018e-05, "loss": 2.8223, "step": 28444 }, { "epoch": 2.23, "learning_rate": 7.769544420920568e-05, "loss": 2.6375, "step": 28445 }, { "epoch": 2.23, "learning_rate": 7.769466007998118e-05, "loss": 2.9687, "step": 28446 }, { "epoch": 2.23, "learning_rate": 7.769387595075669e-05, "loss": 2.5623, "step": 28447 }, { "epoch": 2.23, "learning_rate": 7.769309182153219e-05, "loss": 2.793, "step": 28448 }, { "epoch": 2.23, "learning_rate": 7.76923076923077e-05, "loss": 2.7631, "step": 28449 }, { "epoch": 2.23, "learning_rate": 7.76915235630832e-05, "loss": 3.0858, "step": 28450 }, { "epoch": 2.23, "learning_rate": 7.76907394338587e-05, "loss": 2.7458, "step": 28451 }, { "epoch": 2.23, "learning_rate": 7.768995530463421e-05, "loss": 2.7304, "step": 28452 }, { "epoch": 2.23, "learning_rate": 7.768917117540971e-05, "loss": 2.8098, "step": 28453 }, { "epoch": 2.23, "learning_rate": 7.768838704618521e-05, "loss": 2.6407, "step": 28454 }, { "epoch": 2.23, "learning_rate": 7.768760291696072e-05, "loss": 2.7796, "step": 28455 }, { "epoch": 2.23, "learning_rate": 7.768681878773622e-05, "loss": 2.8533, "step": 28456 }, { "epoch": 2.23, "learning_rate": 7.768603465851173e-05, "loss": 2.7714, "step": 28457 }, { "epoch": 2.23, "learning_rate": 7.768525052928723e-05, "loss": 2.6938, "step": 28458 }, { "epoch": 2.23, "learning_rate": 7.768446640006273e-05, "loss": 2.7418, "step": 28459 }, { "epoch": 2.23, "learning_rate": 7.768368227083824e-05, "loss": 2.9261, "step": 28460 }, { "epoch": 2.23, "learning_rate": 7.768289814161375e-05, "loss": 2.7811, "step": 28461 }, { "epoch": 2.23, "learning_rate": 7.768211401238926e-05, "loss": 2.8358, "step": 28462 }, { "epoch": 2.23, "learning_rate": 7.768132988316475e-05, "loss": 2.8601, "step": 28463 }, { "epoch": 2.23, "learning_rate": 7.768054575394025e-05, "loss": 2.826, "step": 28464 }, { "epoch": 2.23, "learning_rate": 7.767976162471575e-05, "loss": 2.7027, "step": 28465 }, { "epoch": 2.23, "learning_rate": 7.767897749549126e-05, "loss": 2.7371, "step": 28466 }, { "epoch": 2.23, "learning_rate": 7.767819336626676e-05, "loss": 2.9334, "step": 28467 }, { "epoch": 2.23, "learning_rate": 7.767740923704227e-05, "loss": 2.8511, "step": 28468 }, { "epoch": 2.23, "learning_rate": 7.767662510781777e-05, "loss": 2.8227, "step": 28469 }, { "epoch": 2.23, "learning_rate": 7.767584097859327e-05, "loss": 2.8482, "step": 28470 }, { "epoch": 2.23, "learning_rate": 7.767505684936878e-05, "loss": 2.6025, "step": 28471 }, { "epoch": 2.23, "learning_rate": 7.767427272014428e-05, "loss": 3.1277, "step": 28472 }, { "epoch": 2.23, "learning_rate": 7.767348859091978e-05, "loss": 2.9981, "step": 28473 }, { "epoch": 2.23, "learning_rate": 7.767270446169529e-05, "loss": 2.8629, "step": 28474 }, { "epoch": 2.23, "learning_rate": 7.767192033247079e-05, "loss": 3.0907, "step": 28475 }, { "epoch": 2.23, "learning_rate": 7.76711362032463e-05, "loss": 2.7791, "step": 28476 }, { "epoch": 2.23, "learning_rate": 7.76703520740218e-05, "loss": 2.9819, "step": 28477 }, { "epoch": 2.23, "learning_rate": 7.76695679447973e-05, "loss": 2.8762, "step": 28478 }, { "epoch": 2.23, "learning_rate": 7.76687838155728e-05, "loss": 2.8114, "step": 28479 }, { "epoch": 2.23, "learning_rate": 7.766799968634831e-05, "loss": 2.8046, "step": 28480 }, { "epoch": 2.23, "learning_rate": 7.766721555712381e-05, "loss": 2.8341, "step": 28481 }, { "epoch": 2.23, "learning_rate": 7.766643142789932e-05, "loss": 2.8318, "step": 28482 }, { "epoch": 2.23, "learning_rate": 7.766564729867482e-05, "loss": 2.7979, "step": 28483 }, { "epoch": 2.23, "learning_rate": 7.766486316945032e-05, "loss": 2.688, "step": 28484 }, { "epoch": 2.23, "learning_rate": 7.766407904022583e-05, "loss": 2.9025, "step": 28485 }, { "epoch": 2.23, "learning_rate": 7.766329491100133e-05, "loss": 2.6916, "step": 28486 }, { "epoch": 2.23, "learning_rate": 7.766251078177685e-05, "loss": 2.8803, "step": 28487 }, { "epoch": 2.23, "learning_rate": 7.766172665255235e-05, "loss": 2.8204, "step": 28488 }, { "epoch": 2.23, "learning_rate": 7.766094252332786e-05, "loss": 2.7408, "step": 28489 }, { "epoch": 2.23, "learning_rate": 7.766015839410336e-05, "loss": 2.7811, "step": 28490 }, { "epoch": 2.23, "learning_rate": 7.765937426487886e-05, "loss": 2.7655, "step": 28491 }, { "epoch": 2.23, "learning_rate": 7.765859013565437e-05, "loss": 2.8252, "step": 28492 }, { "epoch": 2.23, "learning_rate": 7.765780600642987e-05, "loss": 3.0283, "step": 28493 }, { "epoch": 2.23, "learning_rate": 7.765702187720537e-05, "loss": 3.03, "step": 28494 }, { "epoch": 2.23, "learning_rate": 7.765623774798088e-05, "loss": 2.7863, "step": 28495 }, { "epoch": 2.23, "learning_rate": 7.765545361875638e-05, "loss": 3.0331, "step": 28496 }, { "epoch": 2.23, "learning_rate": 7.765466948953189e-05, "loss": 2.9666, "step": 28497 }, { "epoch": 2.23, "learning_rate": 7.765388536030739e-05, "loss": 2.7068, "step": 28498 }, { "epoch": 2.23, "learning_rate": 7.765310123108288e-05, "loss": 2.6495, "step": 28499 }, { "epoch": 2.23, "learning_rate": 7.765231710185838e-05, "loss": 2.9521, "step": 28500 }, { "epoch": 2.23, "learning_rate": 7.765153297263389e-05, "loss": 2.7055, "step": 28501 }, { "epoch": 2.23, "learning_rate": 7.765074884340939e-05, "loss": 2.7019, "step": 28502 }, { "epoch": 2.24, "learning_rate": 7.76499647141849e-05, "loss": 2.9314, "step": 28503 }, { "epoch": 2.24, "learning_rate": 7.76491805849604e-05, "loss": 3.0968, "step": 28504 }, { "epoch": 2.24, "learning_rate": 7.76483964557359e-05, "loss": 2.7255, "step": 28505 }, { "epoch": 2.24, "learning_rate": 7.76476123265114e-05, "loss": 2.9478, "step": 28506 }, { "epoch": 2.24, "learning_rate": 7.764682819728691e-05, "loss": 2.8347, "step": 28507 }, { "epoch": 2.24, "learning_rate": 7.764604406806241e-05, "loss": 2.561, "step": 28508 }, { "epoch": 2.24, "learning_rate": 7.764525993883792e-05, "loss": 2.9682, "step": 28509 }, { "epoch": 2.24, "learning_rate": 7.764447580961342e-05, "loss": 2.6599, "step": 28510 }, { "epoch": 2.24, "learning_rate": 7.764369168038892e-05, "loss": 2.9179, "step": 28511 }, { "epoch": 2.24, "learning_rate": 7.764290755116443e-05, "loss": 2.8493, "step": 28512 }, { "epoch": 2.24, "learning_rate": 7.764212342193994e-05, "loss": 2.6713, "step": 28513 }, { "epoch": 2.24, "learning_rate": 7.764133929271545e-05, "loss": 2.6928, "step": 28514 }, { "epoch": 2.24, "learning_rate": 7.764055516349095e-05, "loss": 2.8276, "step": 28515 }, { "epoch": 2.24, "learning_rate": 7.763977103426646e-05, "loss": 2.8156, "step": 28516 }, { "epoch": 2.24, "learning_rate": 7.763898690504196e-05, "loss": 2.8107, "step": 28517 }, { "epoch": 2.24, "learning_rate": 7.763820277581746e-05, "loss": 2.7053, "step": 28518 }, { "epoch": 2.24, "learning_rate": 7.763741864659297e-05, "loss": 2.7146, "step": 28519 }, { "epoch": 2.24, "learning_rate": 7.763663451736847e-05, "loss": 2.7515, "step": 28520 }, { "epoch": 2.24, "learning_rate": 7.763585038814397e-05, "loss": 2.7406, "step": 28521 }, { "epoch": 2.24, "learning_rate": 7.763506625891948e-05, "loss": 2.776, "step": 28522 }, { "epoch": 2.24, "learning_rate": 7.763428212969498e-05, "loss": 2.7012, "step": 28523 }, { "epoch": 2.24, "learning_rate": 7.763349800047049e-05, "loss": 2.7497, "step": 28524 }, { "epoch": 2.24, "learning_rate": 7.763271387124599e-05, "loss": 2.8702, "step": 28525 }, { "epoch": 2.24, "learning_rate": 7.763192974202149e-05, "loss": 2.8175, "step": 28526 }, { "epoch": 2.24, "learning_rate": 7.7631145612797e-05, "loss": 2.8773, "step": 28527 }, { "epoch": 2.24, "learning_rate": 7.76303614835725e-05, "loss": 3.2307, "step": 28528 }, { "epoch": 2.24, "learning_rate": 7.7629577354348e-05, "loss": 2.8131, "step": 28529 }, { "epoch": 2.24, "learning_rate": 7.762879322512351e-05, "loss": 2.7677, "step": 28530 }, { "epoch": 2.24, "learning_rate": 7.762800909589901e-05, "loss": 2.8098, "step": 28531 }, { "epoch": 2.24, "learning_rate": 7.762722496667451e-05, "loss": 2.7844, "step": 28532 }, { "epoch": 2.24, "learning_rate": 7.762644083745002e-05, "loss": 2.6106, "step": 28533 }, { "epoch": 2.24, "learning_rate": 7.762565670822552e-05, "loss": 2.7613, "step": 28534 }, { "epoch": 2.24, "learning_rate": 7.762487257900101e-05, "loss": 2.7508, "step": 28535 }, { "epoch": 2.24, "learning_rate": 7.762408844977652e-05, "loss": 2.8497, "step": 28536 }, { "epoch": 2.24, "learning_rate": 7.762330432055202e-05, "loss": 2.6128, "step": 28537 }, { "epoch": 2.24, "learning_rate": 7.762252019132754e-05, "loss": 2.8973, "step": 28538 }, { "epoch": 2.24, "learning_rate": 7.762173606210304e-05, "loss": 2.6825, "step": 28539 }, { "epoch": 2.24, "learning_rate": 7.762095193287854e-05, "loss": 2.759, "step": 28540 }, { "epoch": 2.24, "learning_rate": 7.762016780365405e-05, "loss": 3.1076, "step": 28541 }, { "epoch": 2.24, "learning_rate": 7.761938367442955e-05, "loss": 2.776, "step": 28542 }, { "epoch": 2.24, "learning_rate": 7.761859954520506e-05, "loss": 2.6648, "step": 28543 }, { "epoch": 2.24, "learning_rate": 7.761781541598056e-05, "loss": 2.7894, "step": 28544 }, { "epoch": 2.24, "learning_rate": 7.761703128675606e-05, "loss": 2.7166, "step": 28545 }, { "epoch": 2.24, "learning_rate": 7.761624715753157e-05, "loss": 2.8504, "step": 28546 }, { "epoch": 2.24, "learning_rate": 7.761546302830707e-05, "loss": 2.8326, "step": 28547 }, { "epoch": 2.24, "learning_rate": 7.761467889908257e-05, "loss": 2.6168, "step": 28548 }, { "epoch": 2.24, "learning_rate": 7.761389476985808e-05, "loss": 2.7903, "step": 28549 }, { "epoch": 2.24, "learning_rate": 7.761311064063358e-05, "loss": 2.862, "step": 28550 }, { "epoch": 2.24, "learning_rate": 7.761232651140908e-05, "loss": 2.6816, "step": 28551 }, { "epoch": 2.24, "learning_rate": 7.761154238218459e-05, "loss": 2.7121, "step": 28552 }, { "epoch": 2.24, "learning_rate": 7.761075825296009e-05, "loss": 2.9932, "step": 28553 }, { "epoch": 2.24, "learning_rate": 7.76099741237356e-05, "loss": 2.9462, "step": 28554 }, { "epoch": 2.24, "learning_rate": 7.76091899945111e-05, "loss": 2.8567, "step": 28555 }, { "epoch": 2.24, "learning_rate": 7.76084058652866e-05, "loss": 2.6927, "step": 28556 }, { "epoch": 2.24, "learning_rate": 7.76076217360621e-05, "loss": 2.7305, "step": 28557 }, { "epoch": 2.24, "learning_rate": 7.760683760683761e-05, "loss": 2.6251, "step": 28558 }, { "epoch": 2.24, "learning_rate": 7.760605347761311e-05, "loss": 3.0444, "step": 28559 }, { "epoch": 2.24, "learning_rate": 7.760526934838862e-05, "loss": 2.6351, "step": 28560 }, { "epoch": 2.24, "learning_rate": 7.760448521916412e-05, "loss": 2.891, "step": 28561 }, { "epoch": 2.24, "learning_rate": 7.760370108993963e-05, "loss": 2.613, "step": 28562 }, { "epoch": 2.24, "learning_rate": 7.760291696071513e-05, "loss": 2.7611, "step": 28563 }, { "epoch": 2.24, "learning_rate": 7.760213283149063e-05, "loss": 3.086, "step": 28564 }, { "epoch": 2.24, "learning_rate": 7.760134870226614e-05, "loss": 2.7288, "step": 28565 }, { "epoch": 2.24, "learning_rate": 7.760056457304164e-05, "loss": 2.8367, "step": 28566 }, { "epoch": 2.24, "learning_rate": 7.759978044381714e-05, "loss": 2.8836, "step": 28567 }, { "epoch": 2.24, "learning_rate": 7.759899631459265e-05, "loss": 2.6756, "step": 28568 }, { "epoch": 2.24, "learning_rate": 7.759821218536816e-05, "loss": 2.6927, "step": 28569 }, { "epoch": 2.24, "learning_rate": 7.759742805614367e-05, "loss": 2.8694, "step": 28570 }, { "epoch": 2.24, "learning_rate": 7.759664392691916e-05, "loss": 2.9019, "step": 28571 }, { "epoch": 2.24, "learning_rate": 7.759585979769466e-05, "loss": 2.8155, "step": 28572 }, { "epoch": 2.24, "learning_rate": 7.759507566847017e-05, "loss": 2.9119, "step": 28573 }, { "epoch": 2.24, "learning_rate": 7.759429153924567e-05, "loss": 2.757, "step": 28574 }, { "epoch": 2.24, "learning_rate": 7.759350741002117e-05, "loss": 2.7849, "step": 28575 }, { "epoch": 2.24, "learning_rate": 7.759272328079668e-05, "loss": 2.6254, "step": 28576 }, { "epoch": 2.24, "learning_rate": 7.759193915157218e-05, "loss": 3.063, "step": 28577 }, { "epoch": 2.24, "learning_rate": 7.759115502234768e-05, "loss": 2.7973, "step": 28578 }, { "epoch": 2.24, "learning_rate": 7.759037089312319e-05, "loss": 2.7056, "step": 28579 }, { "epoch": 2.24, "learning_rate": 7.758958676389869e-05, "loss": 2.5459, "step": 28580 }, { "epoch": 2.24, "learning_rate": 7.75888026346742e-05, "loss": 2.8266, "step": 28581 }, { "epoch": 2.24, "learning_rate": 7.75880185054497e-05, "loss": 2.9198, "step": 28582 }, { "epoch": 2.24, "learning_rate": 7.75872343762252e-05, "loss": 3.149, "step": 28583 }, { "epoch": 2.24, "learning_rate": 7.75864502470007e-05, "loss": 2.7385, "step": 28584 }, { "epoch": 2.24, "learning_rate": 7.758566611777621e-05, "loss": 2.8295, "step": 28585 }, { "epoch": 2.24, "learning_rate": 7.758488198855171e-05, "loss": 2.7364, "step": 28586 }, { "epoch": 2.24, "learning_rate": 7.758409785932722e-05, "loss": 2.7414, "step": 28587 }, { "epoch": 2.24, "learning_rate": 7.758331373010272e-05, "loss": 3.0038, "step": 28588 }, { "epoch": 2.24, "learning_rate": 7.758252960087822e-05, "loss": 2.8165, "step": 28589 }, { "epoch": 2.24, "learning_rate": 7.758174547165373e-05, "loss": 2.4857, "step": 28590 }, { "epoch": 2.24, "learning_rate": 7.758096134242923e-05, "loss": 3.1687, "step": 28591 }, { "epoch": 2.24, "learning_rate": 7.758017721320474e-05, "loss": 3.0496, "step": 28592 }, { "epoch": 2.24, "learning_rate": 7.757939308398024e-05, "loss": 2.6196, "step": 28593 }, { "epoch": 2.24, "learning_rate": 7.757860895475574e-05, "loss": 2.8543, "step": 28594 }, { "epoch": 2.24, "learning_rate": 7.757782482553126e-05, "loss": 2.7501, "step": 28595 }, { "epoch": 2.24, "learning_rate": 7.757704069630676e-05, "loss": 2.6592, "step": 28596 }, { "epoch": 2.24, "learning_rate": 7.757625656708227e-05, "loss": 2.817, "step": 28597 }, { "epoch": 2.24, "learning_rate": 7.757547243785777e-05, "loss": 2.9066, "step": 28598 }, { "epoch": 2.24, "learning_rate": 7.757468830863327e-05, "loss": 2.8648, "step": 28599 }, { "epoch": 2.24, "learning_rate": 7.757390417940878e-05, "loss": 2.8378, "step": 28600 }, { "epoch": 2.24, "learning_rate": 7.757312005018428e-05, "loss": 2.6182, "step": 28601 }, { "epoch": 2.24, "learning_rate": 7.757233592095979e-05, "loss": 2.7751, "step": 28602 }, { "epoch": 2.24, "learning_rate": 7.757155179173529e-05, "loss": 2.6814, "step": 28603 }, { "epoch": 2.24, "learning_rate": 7.757076766251079e-05, "loss": 2.9363, "step": 28604 }, { "epoch": 2.24, "learning_rate": 7.75699835332863e-05, "loss": 2.9121, "step": 28605 }, { "epoch": 2.24, "learning_rate": 7.75691994040618e-05, "loss": 2.8392, "step": 28606 }, { "epoch": 2.24, "learning_rate": 7.756841527483729e-05, "loss": 2.7389, "step": 28607 }, { "epoch": 2.24, "learning_rate": 7.75676311456128e-05, "loss": 2.7234, "step": 28608 }, { "epoch": 2.24, "learning_rate": 7.75668470163883e-05, "loss": 2.7658, "step": 28609 }, { "epoch": 2.24, "learning_rate": 7.75660628871638e-05, "loss": 3.0362, "step": 28610 }, { "epoch": 2.24, "learning_rate": 7.75652787579393e-05, "loss": 2.732, "step": 28611 }, { "epoch": 2.24, "learning_rate": 7.756449462871481e-05, "loss": 2.7544, "step": 28612 }, { "epoch": 2.24, "learning_rate": 7.756371049949031e-05, "loss": 2.8423, "step": 28613 }, { "epoch": 2.24, "learning_rate": 7.756292637026582e-05, "loss": 2.7867, "step": 28614 }, { "epoch": 2.24, "learning_rate": 7.756214224104132e-05, "loss": 2.6883, "step": 28615 }, { "epoch": 2.24, "learning_rate": 7.756135811181682e-05, "loss": 2.6209, "step": 28616 }, { "epoch": 2.24, "learning_rate": 7.756057398259233e-05, "loss": 2.8014, "step": 28617 }, { "epoch": 2.24, "learning_rate": 7.755978985336783e-05, "loss": 2.7313, "step": 28618 }, { "epoch": 2.24, "learning_rate": 7.755900572414333e-05, "loss": 2.7789, "step": 28619 }, { "epoch": 2.24, "learning_rate": 7.755822159491884e-05, "loss": 2.6785, "step": 28620 }, { "epoch": 2.24, "learning_rate": 7.755743746569436e-05, "loss": 3.1744, "step": 28621 }, { "epoch": 2.24, "learning_rate": 7.755665333646986e-05, "loss": 3.0476, "step": 28622 }, { "epoch": 2.24, "learning_rate": 7.755586920724536e-05, "loss": 2.6888, "step": 28623 }, { "epoch": 2.24, "learning_rate": 7.755508507802087e-05, "loss": 2.8587, "step": 28624 }, { "epoch": 2.24, "learning_rate": 7.755430094879637e-05, "loss": 2.9856, "step": 28625 }, { "epoch": 2.24, "learning_rate": 7.755351681957187e-05, "loss": 2.8433, "step": 28626 }, { "epoch": 2.24, "learning_rate": 7.755273269034738e-05, "loss": 2.9781, "step": 28627 }, { "epoch": 2.24, "learning_rate": 7.755194856112288e-05, "loss": 2.7985, "step": 28628 }, { "epoch": 2.24, "learning_rate": 7.755116443189839e-05, "loss": 2.6425, "step": 28629 }, { "epoch": 2.24, "learning_rate": 7.755038030267389e-05, "loss": 2.763, "step": 28630 }, { "epoch": 2.25, "learning_rate": 7.754959617344939e-05, "loss": 3.0445, "step": 28631 }, { "epoch": 2.25, "learning_rate": 7.75488120442249e-05, "loss": 2.7582, "step": 28632 }, { "epoch": 2.25, "learning_rate": 7.75480279150004e-05, "loss": 2.6697, "step": 28633 }, { "epoch": 2.25, "learning_rate": 7.75472437857759e-05, "loss": 2.7153, "step": 28634 }, { "epoch": 2.25, "learning_rate": 7.754645965655141e-05, "loss": 2.8284, "step": 28635 }, { "epoch": 2.25, "learning_rate": 7.754567552732691e-05, "loss": 2.6842, "step": 28636 }, { "epoch": 2.25, "learning_rate": 7.754489139810241e-05, "loss": 2.6952, "step": 28637 }, { "epoch": 2.25, "learning_rate": 7.754410726887792e-05, "loss": 2.9905, "step": 28638 }, { "epoch": 2.25, "learning_rate": 7.754332313965342e-05, "loss": 2.9536, "step": 28639 }, { "epoch": 2.25, "learning_rate": 7.754253901042893e-05, "loss": 3.0446, "step": 28640 }, { "epoch": 2.25, "learning_rate": 7.754175488120443e-05, "loss": 2.9088, "step": 28641 }, { "epoch": 2.25, "learning_rate": 7.754097075197993e-05, "loss": 2.7673, "step": 28642 }, { "epoch": 2.25, "learning_rate": 7.754018662275542e-05, "loss": 2.7027, "step": 28643 }, { "epoch": 2.25, "learning_rate": 7.753940249353093e-05, "loss": 2.9163, "step": 28644 }, { "epoch": 2.25, "learning_rate": 7.753861836430643e-05, "loss": 2.73, "step": 28645 }, { "epoch": 2.25, "learning_rate": 7.753783423508195e-05, "loss": 3.0352, "step": 28646 }, { "epoch": 2.25, "learning_rate": 7.753705010585745e-05, "loss": 3.0278, "step": 28647 }, { "epoch": 2.25, "learning_rate": 7.753626597663295e-05, "loss": 2.7538, "step": 28648 }, { "epoch": 2.25, "learning_rate": 7.753548184740846e-05, "loss": 2.8113, "step": 28649 }, { "epoch": 2.25, "learning_rate": 7.753469771818396e-05, "loss": 2.786, "step": 28650 }, { "epoch": 2.25, "learning_rate": 7.753391358895947e-05, "loss": 2.8412, "step": 28651 }, { "epoch": 2.25, "learning_rate": 7.753312945973497e-05, "loss": 2.6613, "step": 28652 }, { "epoch": 2.25, "learning_rate": 7.753234533051047e-05, "loss": 2.9151, "step": 28653 }, { "epoch": 2.25, "learning_rate": 7.753156120128598e-05, "loss": 2.7308, "step": 28654 }, { "epoch": 2.25, "learning_rate": 7.753077707206148e-05, "loss": 2.6399, "step": 28655 }, { "epoch": 2.25, "learning_rate": 7.752999294283698e-05, "loss": 2.8483, "step": 28656 }, { "epoch": 2.25, "learning_rate": 7.752920881361249e-05, "loss": 2.6831, "step": 28657 }, { "epoch": 2.25, "learning_rate": 7.752842468438799e-05, "loss": 3.0163, "step": 28658 }, { "epoch": 2.25, "learning_rate": 7.75276405551635e-05, "loss": 3.0066, "step": 28659 }, { "epoch": 2.25, "learning_rate": 7.7526856425939e-05, "loss": 2.7764, "step": 28660 }, { "epoch": 2.25, "learning_rate": 7.75260722967145e-05, "loss": 3.024, "step": 28661 }, { "epoch": 2.25, "learning_rate": 7.752528816749e-05, "loss": 2.7912, "step": 28662 }, { "epoch": 2.25, "learning_rate": 7.752450403826551e-05, "loss": 2.7145, "step": 28663 }, { "epoch": 2.25, "learning_rate": 7.752371990904101e-05, "loss": 2.886, "step": 28664 }, { "epoch": 2.25, "learning_rate": 7.752293577981652e-05, "loss": 3.0152, "step": 28665 }, { "epoch": 2.25, "learning_rate": 7.752215165059202e-05, "loss": 2.8734, "step": 28666 }, { "epoch": 2.25, "learning_rate": 7.752136752136752e-05, "loss": 2.7927, "step": 28667 }, { "epoch": 2.25, "learning_rate": 7.752058339214303e-05, "loss": 2.8121, "step": 28668 }, { "epoch": 2.25, "learning_rate": 7.751979926291853e-05, "loss": 2.7741, "step": 28669 }, { "epoch": 2.25, "learning_rate": 7.751901513369404e-05, "loss": 2.7556, "step": 28670 }, { "epoch": 2.25, "learning_rate": 7.751823100446954e-05, "loss": 2.6626, "step": 28671 }, { "epoch": 2.25, "learning_rate": 7.751744687524504e-05, "loss": 2.869, "step": 28672 }, { "epoch": 2.25, "learning_rate": 7.751666274602055e-05, "loss": 2.7439, "step": 28673 }, { "epoch": 2.25, "learning_rate": 7.751587861679605e-05, "loss": 2.7942, "step": 28674 }, { "epoch": 2.25, "learning_rate": 7.751509448757155e-05, "loss": 2.9122, "step": 28675 }, { "epoch": 2.25, "learning_rate": 7.751431035834706e-05, "loss": 2.7093, "step": 28676 }, { "epoch": 2.25, "learning_rate": 7.751352622912258e-05, "loss": 2.6858, "step": 28677 }, { "epoch": 2.25, "learning_rate": 7.751274209989808e-05, "loss": 2.8732, "step": 28678 }, { "epoch": 2.25, "learning_rate": 7.751195797067357e-05, "loss": 2.6319, "step": 28679 }, { "epoch": 2.25, "learning_rate": 7.751117384144907e-05, "loss": 2.9107, "step": 28680 }, { "epoch": 2.25, "learning_rate": 7.751038971222458e-05, "loss": 2.8426, "step": 28681 }, { "epoch": 2.25, "learning_rate": 7.750960558300008e-05, "loss": 2.8211, "step": 28682 }, { "epoch": 2.25, "learning_rate": 7.750882145377558e-05, "loss": 2.8155, "step": 28683 }, { "epoch": 2.25, "learning_rate": 7.750803732455109e-05, "loss": 2.6118, "step": 28684 }, { "epoch": 2.25, "learning_rate": 7.750725319532659e-05, "loss": 2.8991, "step": 28685 }, { "epoch": 2.25, "learning_rate": 7.75064690661021e-05, "loss": 2.7503, "step": 28686 }, { "epoch": 2.25, "learning_rate": 7.75056849368776e-05, "loss": 2.8785, "step": 28687 }, { "epoch": 2.25, "learning_rate": 7.75049008076531e-05, "loss": 2.7629, "step": 28688 }, { "epoch": 2.25, "learning_rate": 7.75041166784286e-05, "loss": 2.9318, "step": 28689 }, { "epoch": 2.25, "learning_rate": 7.750333254920411e-05, "loss": 2.9087, "step": 28690 }, { "epoch": 2.25, "learning_rate": 7.750254841997961e-05, "loss": 2.867, "step": 28691 }, { "epoch": 2.25, "learning_rate": 7.750176429075512e-05, "loss": 2.9899, "step": 28692 }, { "epoch": 2.25, "learning_rate": 7.750098016153062e-05, "loss": 2.826, "step": 28693 }, { "epoch": 2.25, "learning_rate": 7.750019603230612e-05, "loss": 2.8399, "step": 28694 }, { "epoch": 2.25, "learning_rate": 7.749941190308163e-05, "loss": 2.9144, "step": 28695 }, { "epoch": 2.25, "learning_rate": 7.749862777385713e-05, "loss": 2.7746, "step": 28696 }, { "epoch": 2.25, "learning_rate": 7.749784364463264e-05, "loss": 2.981, "step": 28697 }, { "epoch": 2.25, "learning_rate": 7.749705951540814e-05, "loss": 2.8676, "step": 28698 }, { "epoch": 2.25, "learning_rate": 7.749627538618364e-05, "loss": 2.8847, "step": 28699 }, { "epoch": 2.25, "learning_rate": 7.749549125695915e-05, "loss": 2.8172, "step": 28700 }, { "epoch": 2.25, "learning_rate": 7.749470712773465e-05, "loss": 2.7577, "step": 28701 }, { "epoch": 2.25, "learning_rate": 7.749392299851015e-05, "loss": 2.5907, "step": 28702 }, { "epoch": 2.25, "learning_rate": 7.749313886928567e-05, "loss": 2.9677, "step": 28703 }, { "epoch": 2.25, "learning_rate": 7.749235474006117e-05, "loss": 2.6482, "step": 28704 }, { "epoch": 2.25, "learning_rate": 7.749157061083668e-05, "loss": 2.8953, "step": 28705 }, { "epoch": 2.25, "learning_rate": 7.749078648161218e-05, "loss": 2.7551, "step": 28706 }, { "epoch": 2.25, "learning_rate": 7.749000235238769e-05, "loss": 2.7994, "step": 28707 }, { "epoch": 2.25, "learning_rate": 7.748921822316319e-05, "loss": 2.6197, "step": 28708 }, { "epoch": 2.25, "learning_rate": 7.748843409393869e-05, "loss": 2.713, "step": 28709 }, { "epoch": 2.25, "learning_rate": 7.74876499647142e-05, "loss": 2.68, "step": 28710 }, { "epoch": 2.25, "learning_rate": 7.74868658354897e-05, "loss": 2.7146, "step": 28711 }, { "epoch": 2.25, "learning_rate": 7.74860817062652e-05, "loss": 3.0259, "step": 28712 }, { "epoch": 2.25, "learning_rate": 7.748529757704071e-05, "loss": 2.8505, "step": 28713 }, { "epoch": 2.25, "learning_rate": 7.748451344781621e-05, "loss": 2.9177, "step": 28714 }, { "epoch": 2.25, "learning_rate": 7.74837293185917e-05, "loss": 2.7583, "step": 28715 }, { "epoch": 2.25, "learning_rate": 7.74829451893672e-05, "loss": 2.9143, "step": 28716 }, { "epoch": 2.25, "learning_rate": 7.748216106014271e-05, "loss": 2.961, "step": 28717 }, { "epoch": 2.25, "learning_rate": 7.748137693091821e-05, "loss": 2.7972, "step": 28718 }, { "epoch": 2.25, "learning_rate": 7.748059280169372e-05, "loss": 2.6326, "step": 28719 }, { "epoch": 2.25, "learning_rate": 7.747980867246922e-05, "loss": 2.7393, "step": 28720 }, { "epoch": 2.25, "learning_rate": 7.747902454324472e-05, "loss": 2.9162, "step": 28721 }, { "epoch": 2.25, "learning_rate": 7.747824041402023e-05, "loss": 2.6936, "step": 28722 }, { "epoch": 2.25, "learning_rate": 7.747745628479573e-05, "loss": 2.6062, "step": 28723 }, { "epoch": 2.25, "learning_rate": 7.747667215557123e-05, "loss": 2.7528, "step": 28724 }, { "epoch": 2.25, "learning_rate": 7.747588802634674e-05, "loss": 2.6737, "step": 28725 }, { "epoch": 2.25, "learning_rate": 7.747510389712224e-05, "loss": 2.8392, "step": 28726 }, { "epoch": 2.25, "learning_rate": 7.747431976789775e-05, "loss": 2.6441, "step": 28727 }, { "epoch": 2.25, "learning_rate": 7.747353563867325e-05, "loss": 2.7943, "step": 28728 }, { "epoch": 2.25, "learning_rate": 7.747275150944877e-05, "loss": 3.032, "step": 28729 }, { "epoch": 2.25, "learning_rate": 7.747196738022427e-05, "loss": 2.6137, "step": 28730 }, { "epoch": 2.25, "learning_rate": 7.747118325099977e-05, "loss": 2.7789, "step": 28731 }, { "epoch": 2.25, "learning_rate": 7.747039912177528e-05, "loss": 2.8164, "step": 28732 }, { "epoch": 2.25, "learning_rate": 7.746961499255078e-05, "loss": 2.6617, "step": 28733 }, { "epoch": 2.25, "learning_rate": 7.746883086332628e-05, "loss": 2.9981, "step": 28734 }, { "epoch": 2.25, "learning_rate": 7.746804673410179e-05, "loss": 2.9787, "step": 28735 }, { "epoch": 2.25, "learning_rate": 7.746726260487729e-05, "loss": 2.7683, "step": 28736 }, { "epoch": 2.25, "learning_rate": 7.74664784756528e-05, "loss": 2.6915, "step": 28737 }, { "epoch": 2.25, "learning_rate": 7.74656943464283e-05, "loss": 2.7407, "step": 28738 }, { "epoch": 2.25, "learning_rate": 7.74649102172038e-05, "loss": 2.9219, "step": 28739 }, { "epoch": 2.25, "learning_rate": 7.746412608797931e-05, "loss": 2.911, "step": 28740 }, { "epoch": 2.25, "learning_rate": 7.746334195875481e-05, "loss": 2.8446, "step": 28741 }, { "epoch": 2.25, "learning_rate": 7.746255782953031e-05, "loss": 2.7568, "step": 28742 }, { "epoch": 2.25, "learning_rate": 7.746177370030582e-05, "loss": 2.7968, "step": 28743 }, { "epoch": 2.25, "learning_rate": 7.746098957108132e-05, "loss": 2.9692, "step": 28744 }, { "epoch": 2.25, "learning_rate": 7.746020544185683e-05, "loss": 2.7943, "step": 28745 }, { "epoch": 2.25, "learning_rate": 7.745942131263233e-05, "loss": 2.6544, "step": 28746 }, { "epoch": 2.25, "learning_rate": 7.745863718340783e-05, "loss": 2.8237, "step": 28747 }, { "epoch": 2.25, "learning_rate": 7.745785305418334e-05, "loss": 2.7839, "step": 28748 }, { "epoch": 2.25, "learning_rate": 7.745706892495884e-05, "loss": 2.7605, "step": 28749 }, { "epoch": 2.25, "learning_rate": 7.745628479573434e-05, "loss": 3.0084, "step": 28750 }, { "epoch": 2.25, "learning_rate": 7.745550066650983e-05, "loss": 2.845, "step": 28751 }, { "epoch": 2.25, "learning_rate": 7.745471653728534e-05, "loss": 2.8931, "step": 28752 }, { "epoch": 2.25, "learning_rate": 7.745393240806084e-05, "loss": 3.0161, "step": 28753 }, { "epoch": 2.25, "learning_rate": 7.745314827883636e-05, "loss": 2.4981, "step": 28754 }, { "epoch": 2.25, "learning_rate": 7.745236414961186e-05, "loss": 2.5936, "step": 28755 }, { "epoch": 2.25, "learning_rate": 7.745158002038737e-05, "loss": 2.9894, "step": 28756 }, { "epoch": 2.25, "learning_rate": 7.745079589116287e-05, "loss": 2.7209, "step": 28757 }, { "epoch": 2.25, "learning_rate": 7.745001176193837e-05, "loss": 2.8957, "step": 28758 }, { "epoch": 2.26, "learning_rate": 7.744922763271388e-05, "loss": 2.8551, "step": 28759 }, { "epoch": 2.26, "learning_rate": 7.744844350348938e-05, "loss": 2.8506, "step": 28760 }, { "epoch": 2.26, "learning_rate": 7.744765937426488e-05, "loss": 2.8155, "step": 28761 }, { "epoch": 2.26, "learning_rate": 7.744687524504039e-05, "loss": 2.8141, "step": 28762 }, { "epoch": 2.26, "learning_rate": 7.744609111581589e-05, "loss": 2.9015, "step": 28763 }, { "epoch": 2.26, "learning_rate": 7.74453069865914e-05, "loss": 2.8107, "step": 28764 }, { "epoch": 2.26, "learning_rate": 7.74445228573669e-05, "loss": 2.6867, "step": 28765 }, { "epoch": 2.26, "learning_rate": 7.74437387281424e-05, "loss": 2.8342, "step": 28766 }, { "epoch": 2.26, "learning_rate": 7.74429545989179e-05, "loss": 2.8833, "step": 28767 }, { "epoch": 2.26, "learning_rate": 7.744217046969341e-05, "loss": 2.671, "step": 28768 }, { "epoch": 2.26, "learning_rate": 7.744138634046891e-05, "loss": 2.6664, "step": 28769 }, { "epoch": 2.26, "learning_rate": 7.744060221124442e-05, "loss": 2.8328, "step": 28770 }, { "epoch": 2.26, "learning_rate": 7.743981808201992e-05, "loss": 2.9632, "step": 28771 }, { "epoch": 2.26, "learning_rate": 7.743903395279542e-05, "loss": 2.6097, "step": 28772 }, { "epoch": 2.26, "learning_rate": 7.743824982357093e-05, "loss": 2.8236, "step": 28773 }, { "epoch": 2.26, "learning_rate": 7.743746569434643e-05, "loss": 2.7751, "step": 28774 }, { "epoch": 2.26, "learning_rate": 7.743668156512194e-05, "loss": 2.7765, "step": 28775 }, { "epoch": 2.26, "learning_rate": 7.743589743589744e-05, "loss": 2.7849, "step": 28776 }, { "epoch": 2.26, "learning_rate": 7.743511330667294e-05, "loss": 2.6679, "step": 28777 }, { "epoch": 2.26, "learning_rate": 7.743432917744845e-05, "loss": 3.0041, "step": 28778 }, { "epoch": 2.26, "learning_rate": 7.743354504822395e-05, "loss": 2.8308, "step": 28779 }, { "epoch": 2.26, "learning_rate": 7.743276091899945e-05, "loss": 2.8244, "step": 28780 }, { "epoch": 2.26, "learning_rate": 7.743197678977496e-05, "loss": 2.7361, "step": 28781 }, { "epoch": 2.26, "learning_rate": 7.743119266055046e-05, "loss": 3.0859, "step": 28782 }, { "epoch": 2.26, "learning_rate": 7.743040853132596e-05, "loss": 2.8957, "step": 28783 }, { "epoch": 2.26, "learning_rate": 7.742962440210147e-05, "loss": 2.8058, "step": 28784 }, { "epoch": 2.26, "learning_rate": 7.742884027287697e-05, "loss": 2.739, "step": 28785 }, { "epoch": 2.26, "learning_rate": 7.742805614365249e-05, "loss": 2.9623, "step": 28786 }, { "epoch": 2.26, "learning_rate": 7.742727201442798e-05, "loss": 2.8876, "step": 28787 }, { "epoch": 2.26, "learning_rate": 7.742648788520348e-05, "loss": 2.739, "step": 28788 }, { "epoch": 2.26, "learning_rate": 7.742570375597899e-05, "loss": 2.7387, "step": 28789 }, { "epoch": 2.26, "learning_rate": 7.742491962675449e-05, "loss": 2.8871, "step": 28790 }, { "epoch": 2.26, "learning_rate": 7.742413549753e-05, "loss": 2.829, "step": 28791 }, { "epoch": 2.26, "learning_rate": 7.74233513683055e-05, "loss": 2.9613, "step": 28792 }, { "epoch": 2.26, "learning_rate": 7.7422567239081e-05, "loss": 2.7949, "step": 28793 }, { "epoch": 2.26, "learning_rate": 7.74217831098565e-05, "loss": 2.8548, "step": 28794 }, { "epoch": 2.26, "learning_rate": 7.742099898063201e-05, "loss": 2.5882, "step": 28795 }, { "epoch": 2.26, "learning_rate": 7.742021485140751e-05, "loss": 3.051, "step": 28796 }, { "epoch": 2.26, "learning_rate": 7.741943072218302e-05, "loss": 2.716, "step": 28797 }, { "epoch": 2.26, "learning_rate": 7.741864659295852e-05, "loss": 2.7201, "step": 28798 }, { "epoch": 2.26, "learning_rate": 7.741786246373402e-05, "loss": 2.6322, "step": 28799 }, { "epoch": 2.26, "learning_rate": 7.741707833450953e-05, "loss": 2.7549, "step": 28800 }, { "epoch": 2.26, "learning_rate": 7.741629420528503e-05, "loss": 2.8617, "step": 28801 }, { "epoch": 2.26, "learning_rate": 7.741551007606053e-05, "loss": 2.9143, "step": 28802 }, { "epoch": 2.26, "learning_rate": 7.741472594683604e-05, "loss": 2.8203, "step": 28803 }, { "epoch": 2.26, "learning_rate": 7.741394181761154e-05, "loss": 2.7118, "step": 28804 }, { "epoch": 2.26, "learning_rate": 7.741315768838705e-05, "loss": 2.6837, "step": 28805 }, { "epoch": 2.26, "learning_rate": 7.741237355916255e-05, "loss": 2.7735, "step": 28806 }, { "epoch": 2.26, "learning_rate": 7.741158942993805e-05, "loss": 2.8546, "step": 28807 }, { "epoch": 2.26, "learning_rate": 7.741080530071356e-05, "loss": 2.5979, "step": 28808 }, { "epoch": 2.26, "learning_rate": 7.741002117148906e-05, "loss": 2.7398, "step": 28809 }, { "epoch": 2.26, "learning_rate": 7.740923704226456e-05, "loss": 2.9441, "step": 28810 }, { "epoch": 2.26, "learning_rate": 7.740845291304008e-05, "loss": 2.7866, "step": 28811 }, { "epoch": 2.26, "learning_rate": 7.740766878381559e-05, "loss": 2.631, "step": 28812 }, { "epoch": 2.26, "learning_rate": 7.740688465459109e-05, "loss": 2.6007, "step": 28813 }, { "epoch": 2.26, "learning_rate": 7.740610052536659e-05, "loss": 2.7285, "step": 28814 }, { "epoch": 2.26, "learning_rate": 7.74053163961421e-05, "loss": 2.7001, "step": 28815 }, { "epoch": 2.26, "learning_rate": 7.74045322669176e-05, "loss": 2.9518, "step": 28816 }, { "epoch": 2.26, "learning_rate": 7.74037481376931e-05, "loss": 2.7516, "step": 28817 }, { "epoch": 2.26, "learning_rate": 7.740296400846861e-05, "loss": 2.7616, "step": 28818 }, { "epoch": 2.26, "learning_rate": 7.740217987924411e-05, "loss": 2.7609, "step": 28819 }, { "epoch": 2.26, "learning_rate": 7.740139575001961e-05, "loss": 2.5955, "step": 28820 }, { "epoch": 2.26, "learning_rate": 7.740061162079512e-05, "loss": 2.7794, "step": 28821 }, { "epoch": 2.26, "learning_rate": 7.739982749157062e-05, "loss": 2.6518, "step": 28822 }, { "epoch": 2.26, "learning_rate": 7.739904336234611e-05, "loss": 2.8016, "step": 28823 }, { "epoch": 2.26, "learning_rate": 7.739825923312162e-05, "loss": 2.8543, "step": 28824 }, { "epoch": 2.26, "learning_rate": 7.739747510389712e-05, "loss": 2.6919, "step": 28825 }, { "epoch": 2.26, "learning_rate": 7.739669097467262e-05, "loss": 2.7558, "step": 28826 }, { "epoch": 2.26, "learning_rate": 7.739590684544813e-05, "loss": 2.6527, "step": 28827 }, { "epoch": 2.26, "learning_rate": 7.739512271622363e-05, "loss": 2.7575, "step": 28828 }, { "epoch": 2.26, "learning_rate": 7.739433858699913e-05, "loss": 2.4983, "step": 28829 }, { "epoch": 2.26, "learning_rate": 7.739355445777464e-05, "loss": 2.7354, "step": 28830 }, { "epoch": 2.26, "learning_rate": 7.739277032855014e-05, "loss": 2.8354, "step": 28831 }, { "epoch": 2.26, "learning_rate": 7.739198619932565e-05, "loss": 2.8011, "step": 28832 }, { "epoch": 2.26, "learning_rate": 7.739120207010115e-05, "loss": 2.7225, "step": 28833 }, { "epoch": 2.26, "learning_rate": 7.739041794087665e-05, "loss": 2.7214, "step": 28834 }, { "epoch": 2.26, "learning_rate": 7.738963381165216e-05, "loss": 2.8197, "step": 28835 }, { "epoch": 2.26, "learning_rate": 7.738884968242766e-05, "loss": 2.6202, "step": 28836 }, { "epoch": 2.26, "learning_rate": 7.738806555320318e-05, "loss": 2.985, "step": 28837 }, { "epoch": 2.26, "learning_rate": 7.738728142397868e-05, "loss": 2.9287, "step": 28838 }, { "epoch": 2.26, "learning_rate": 7.738649729475418e-05, "loss": 2.6095, "step": 28839 }, { "epoch": 2.26, "learning_rate": 7.738571316552969e-05, "loss": 2.8649, "step": 28840 }, { "epoch": 2.26, "learning_rate": 7.738492903630519e-05, "loss": 2.7556, "step": 28841 }, { "epoch": 2.26, "learning_rate": 7.73841449070807e-05, "loss": 2.8705, "step": 28842 }, { "epoch": 2.26, "learning_rate": 7.73833607778562e-05, "loss": 2.7242, "step": 28843 }, { "epoch": 2.26, "learning_rate": 7.73825766486317e-05, "loss": 2.7238, "step": 28844 }, { "epoch": 2.26, "learning_rate": 7.73817925194072e-05, "loss": 2.8283, "step": 28845 }, { "epoch": 2.26, "learning_rate": 7.738100839018271e-05, "loss": 2.8931, "step": 28846 }, { "epoch": 2.26, "learning_rate": 7.738022426095821e-05, "loss": 2.9582, "step": 28847 }, { "epoch": 2.26, "learning_rate": 7.737944013173372e-05, "loss": 2.6628, "step": 28848 }, { "epoch": 2.26, "learning_rate": 7.737865600250922e-05, "loss": 2.79, "step": 28849 }, { "epoch": 2.26, "learning_rate": 7.737787187328472e-05, "loss": 2.9087, "step": 28850 }, { "epoch": 2.26, "learning_rate": 7.737708774406023e-05, "loss": 2.5933, "step": 28851 }, { "epoch": 2.26, "learning_rate": 7.737630361483573e-05, "loss": 2.6595, "step": 28852 }, { "epoch": 2.26, "learning_rate": 7.737551948561124e-05, "loss": 2.9418, "step": 28853 }, { "epoch": 2.26, "learning_rate": 7.737473535638674e-05, "loss": 2.6278, "step": 28854 }, { "epoch": 2.26, "learning_rate": 7.737395122716224e-05, "loss": 2.6326, "step": 28855 }, { "epoch": 2.26, "learning_rate": 7.737316709793775e-05, "loss": 2.8588, "step": 28856 }, { "epoch": 2.26, "learning_rate": 7.737238296871325e-05, "loss": 2.8197, "step": 28857 }, { "epoch": 2.26, "learning_rate": 7.737159883948875e-05, "loss": 2.5047, "step": 28858 }, { "epoch": 2.26, "learning_rate": 7.737081471026424e-05, "loss": 2.6872, "step": 28859 }, { "epoch": 2.26, "learning_rate": 7.737003058103975e-05, "loss": 2.8462, "step": 28860 }, { "epoch": 2.26, "learning_rate": 7.736924645181525e-05, "loss": 2.7156, "step": 28861 }, { "epoch": 2.26, "learning_rate": 7.736846232259077e-05, "loss": 2.6769, "step": 28862 }, { "epoch": 2.26, "learning_rate": 7.736767819336627e-05, "loss": 2.7199, "step": 28863 }, { "epoch": 2.26, "learning_rate": 7.736689406414178e-05, "loss": 2.566, "step": 28864 }, { "epoch": 2.26, "learning_rate": 7.736610993491728e-05, "loss": 2.8611, "step": 28865 }, { "epoch": 2.26, "learning_rate": 7.736532580569278e-05, "loss": 2.7283, "step": 28866 }, { "epoch": 2.26, "learning_rate": 7.736454167646829e-05, "loss": 2.6631, "step": 28867 }, { "epoch": 2.26, "learning_rate": 7.736375754724379e-05, "loss": 2.6933, "step": 28868 }, { "epoch": 2.26, "learning_rate": 7.73629734180193e-05, "loss": 2.8437, "step": 28869 }, { "epoch": 2.26, "learning_rate": 7.73621892887948e-05, "loss": 3.0301, "step": 28870 }, { "epoch": 2.26, "learning_rate": 7.73614051595703e-05, "loss": 2.6942, "step": 28871 }, { "epoch": 2.26, "learning_rate": 7.73606210303458e-05, "loss": 2.7598, "step": 28872 }, { "epoch": 2.26, "learning_rate": 7.735983690112131e-05, "loss": 2.6803, "step": 28873 }, { "epoch": 2.26, "learning_rate": 7.735905277189681e-05, "loss": 2.7719, "step": 28874 }, { "epoch": 2.26, "learning_rate": 7.735826864267232e-05, "loss": 2.892, "step": 28875 }, { "epoch": 2.26, "learning_rate": 7.735748451344782e-05, "loss": 2.8591, "step": 28876 }, { "epoch": 2.26, "learning_rate": 7.735670038422332e-05, "loss": 2.9416, "step": 28877 }, { "epoch": 2.26, "learning_rate": 7.735591625499883e-05, "loss": 2.7519, "step": 28878 }, { "epoch": 2.26, "learning_rate": 7.735513212577433e-05, "loss": 2.9461, "step": 28879 }, { "epoch": 2.26, "learning_rate": 7.735434799654984e-05, "loss": 2.6993, "step": 28880 }, { "epoch": 2.26, "learning_rate": 7.735356386732534e-05, "loss": 2.8424, "step": 28881 }, { "epoch": 2.26, "learning_rate": 7.735277973810084e-05, "loss": 2.8489, "step": 28882 }, { "epoch": 2.26, "learning_rate": 7.735199560887635e-05, "loss": 2.6801, "step": 28883 }, { "epoch": 2.26, "learning_rate": 7.735121147965185e-05, "loss": 2.9266, "step": 28884 }, { "epoch": 2.26, "learning_rate": 7.735042735042735e-05, "loss": 2.7378, "step": 28885 }, { "epoch": 2.27, "learning_rate": 7.734964322120286e-05, "loss": 2.5483, "step": 28886 }, { "epoch": 2.27, "learning_rate": 7.734885909197836e-05, "loss": 2.8895, "step": 28887 }, { "epoch": 2.27, "learning_rate": 7.734807496275386e-05, "loss": 2.9564, "step": 28888 }, { "epoch": 2.27, "learning_rate": 7.734729083352937e-05, "loss": 2.8464, "step": 28889 }, { "epoch": 2.27, "learning_rate": 7.734650670430487e-05, "loss": 2.8525, "step": 28890 }, { "epoch": 2.27, "learning_rate": 7.734572257508038e-05, "loss": 2.7923, "step": 28891 }, { "epoch": 2.27, "learning_rate": 7.734493844585588e-05, "loss": 2.6629, "step": 28892 }, { "epoch": 2.27, "learning_rate": 7.734415431663138e-05, "loss": 2.6705, "step": 28893 }, { "epoch": 2.27, "learning_rate": 7.73433701874069e-05, "loss": 2.6896, "step": 28894 }, { "epoch": 2.27, "learning_rate": 7.734258605818239e-05, "loss": 2.6881, "step": 28895 }, { "epoch": 2.27, "learning_rate": 7.73418019289579e-05, "loss": 2.8461, "step": 28896 }, { "epoch": 2.27, "learning_rate": 7.73410177997334e-05, "loss": 2.7556, "step": 28897 }, { "epoch": 2.27, "learning_rate": 7.73402336705089e-05, "loss": 2.6925, "step": 28898 }, { "epoch": 2.27, "learning_rate": 7.73394495412844e-05, "loss": 3.134, "step": 28899 }, { "epoch": 2.27, "learning_rate": 7.733866541205991e-05, "loss": 2.9228, "step": 28900 }, { "epoch": 2.27, "learning_rate": 7.733788128283541e-05, "loss": 2.7537, "step": 28901 }, { "epoch": 2.27, "learning_rate": 7.733709715361092e-05, "loss": 3.0013, "step": 28902 }, { "epoch": 2.27, "learning_rate": 7.733631302438642e-05, "loss": 2.6917, "step": 28903 }, { "epoch": 2.27, "learning_rate": 7.733552889516192e-05, "loss": 2.7689, "step": 28904 }, { "epoch": 2.27, "learning_rate": 7.733474476593743e-05, "loss": 2.7158, "step": 28905 }, { "epoch": 2.27, "learning_rate": 7.733396063671293e-05, "loss": 2.8156, "step": 28906 }, { "epoch": 2.27, "learning_rate": 7.733317650748843e-05, "loss": 2.8763, "step": 28907 }, { "epoch": 2.27, "learning_rate": 7.733239237826394e-05, "loss": 2.7799, "step": 28908 }, { "epoch": 2.27, "learning_rate": 7.733160824903944e-05, "loss": 2.7287, "step": 28909 }, { "epoch": 2.27, "learning_rate": 7.733082411981495e-05, "loss": 2.869, "step": 28910 }, { "epoch": 2.27, "learning_rate": 7.733003999059045e-05, "loss": 3.0453, "step": 28911 }, { "epoch": 2.27, "learning_rate": 7.732925586136595e-05, "loss": 3.0343, "step": 28912 }, { "epoch": 2.27, "learning_rate": 7.732847173214146e-05, "loss": 2.5415, "step": 28913 }, { "epoch": 2.27, "learning_rate": 7.732768760291696e-05, "loss": 2.777, "step": 28914 }, { "epoch": 2.27, "learning_rate": 7.732690347369246e-05, "loss": 2.7006, "step": 28915 }, { "epoch": 2.27, "learning_rate": 7.732611934446797e-05, "loss": 2.8653, "step": 28916 }, { "epoch": 2.27, "learning_rate": 7.732533521524347e-05, "loss": 2.7134, "step": 28917 }, { "epoch": 2.27, "learning_rate": 7.732455108601897e-05, "loss": 2.6677, "step": 28918 }, { "epoch": 2.27, "learning_rate": 7.732376695679449e-05, "loss": 3.0146, "step": 28919 }, { "epoch": 2.27, "learning_rate": 7.732298282757e-05, "loss": 2.9661, "step": 28920 }, { "epoch": 2.27, "learning_rate": 7.73221986983455e-05, "loss": 2.69, "step": 28921 }, { "epoch": 2.27, "learning_rate": 7.7321414569121e-05, "loss": 2.8529, "step": 28922 }, { "epoch": 2.27, "learning_rate": 7.732063043989651e-05, "loss": 2.7809, "step": 28923 }, { "epoch": 2.27, "learning_rate": 7.731984631067201e-05, "loss": 2.7948, "step": 28924 }, { "epoch": 2.27, "learning_rate": 7.731906218144751e-05, "loss": 2.6388, "step": 28925 }, { "epoch": 2.27, "learning_rate": 7.731827805222302e-05, "loss": 2.6788, "step": 28926 }, { "epoch": 2.27, "learning_rate": 7.731749392299852e-05, "loss": 2.8627, "step": 28927 }, { "epoch": 2.27, "learning_rate": 7.731670979377403e-05, "loss": 2.6947, "step": 28928 }, { "epoch": 2.27, "learning_rate": 7.731592566454953e-05, "loss": 2.9604, "step": 28929 }, { "epoch": 2.27, "learning_rate": 7.731514153532503e-05, "loss": 2.7956, "step": 28930 }, { "epoch": 2.27, "learning_rate": 7.731435740610052e-05, "loss": 2.9292, "step": 28931 }, { "epoch": 2.27, "learning_rate": 7.731357327687603e-05, "loss": 2.7768, "step": 28932 }, { "epoch": 2.27, "learning_rate": 7.731278914765153e-05, "loss": 2.8872, "step": 28933 }, { "epoch": 2.27, "learning_rate": 7.731200501842703e-05, "loss": 3.1141, "step": 28934 }, { "epoch": 2.27, "learning_rate": 7.731122088920254e-05, "loss": 2.959, "step": 28935 }, { "epoch": 2.27, "learning_rate": 7.731043675997804e-05, "loss": 2.7322, "step": 28936 }, { "epoch": 2.27, "learning_rate": 7.730965263075354e-05, "loss": 2.578, "step": 28937 }, { "epoch": 2.27, "learning_rate": 7.730886850152905e-05, "loss": 2.697, "step": 28938 }, { "epoch": 2.27, "learning_rate": 7.730808437230455e-05, "loss": 2.8226, "step": 28939 }, { "epoch": 2.27, "learning_rate": 7.730730024308006e-05, "loss": 3.0552, "step": 28940 }, { "epoch": 2.27, "learning_rate": 7.730651611385556e-05, "loss": 2.7056, "step": 28941 }, { "epoch": 2.27, "learning_rate": 7.730573198463106e-05, "loss": 2.8593, "step": 28942 }, { "epoch": 2.27, "learning_rate": 7.730494785540657e-05, "loss": 3.0899, "step": 28943 }, { "epoch": 2.27, "learning_rate": 7.730416372618207e-05, "loss": 2.6449, "step": 28944 }, { "epoch": 2.27, "learning_rate": 7.730337959695759e-05, "loss": 2.8245, "step": 28945 }, { "epoch": 2.27, "learning_rate": 7.730259546773309e-05, "loss": 2.6812, "step": 28946 }, { "epoch": 2.27, "learning_rate": 7.73018113385086e-05, "loss": 2.8193, "step": 28947 }, { "epoch": 2.27, "learning_rate": 7.73010272092841e-05, "loss": 2.9425, "step": 28948 }, { "epoch": 2.27, "learning_rate": 7.73002430800596e-05, "loss": 2.7466, "step": 28949 }, { "epoch": 2.27, "learning_rate": 7.72994589508351e-05, "loss": 2.821, "step": 28950 }, { "epoch": 2.27, "learning_rate": 7.729867482161061e-05, "loss": 2.731, "step": 28951 }, { "epoch": 2.27, "learning_rate": 7.729789069238611e-05, "loss": 2.9934, "step": 28952 }, { "epoch": 2.27, "learning_rate": 7.729710656316162e-05, "loss": 2.7934, "step": 28953 }, { "epoch": 2.27, "learning_rate": 7.729632243393712e-05, "loss": 2.6969, "step": 28954 }, { "epoch": 2.27, "learning_rate": 7.729553830471262e-05, "loss": 2.4389, "step": 28955 }, { "epoch": 2.27, "learning_rate": 7.729475417548813e-05, "loss": 2.9654, "step": 28956 }, { "epoch": 2.27, "learning_rate": 7.729397004626363e-05, "loss": 2.7575, "step": 28957 }, { "epoch": 2.27, "learning_rate": 7.729318591703914e-05, "loss": 2.9469, "step": 28958 }, { "epoch": 2.27, "learning_rate": 7.729240178781464e-05, "loss": 2.7838, "step": 28959 }, { "epoch": 2.27, "learning_rate": 7.729161765859014e-05, "loss": 2.5852, "step": 28960 }, { "epoch": 2.27, "learning_rate": 7.729083352936565e-05, "loss": 2.7269, "step": 28961 }, { "epoch": 2.27, "learning_rate": 7.729004940014115e-05, "loss": 2.6052, "step": 28962 }, { "epoch": 2.27, "learning_rate": 7.728926527091665e-05, "loss": 2.8684, "step": 28963 }, { "epoch": 2.27, "learning_rate": 7.728848114169216e-05, "loss": 2.6892, "step": 28964 }, { "epoch": 2.27, "learning_rate": 7.728769701246766e-05, "loss": 2.9102, "step": 28965 }, { "epoch": 2.27, "learning_rate": 7.728691288324317e-05, "loss": 3.1366, "step": 28966 }, { "epoch": 2.27, "learning_rate": 7.728612875401866e-05, "loss": 3.0171, "step": 28967 }, { "epoch": 2.27, "learning_rate": 7.728534462479416e-05, "loss": 2.8885, "step": 28968 }, { "epoch": 2.27, "learning_rate": 7.728456049556966e-05, "loss": 2.7731, "step": 28969 }, { "epoch": 2.27, "learning_rate": 7.728377636634518e-05, "loss": 2.9527, "step": 28970 }, { "epoch": 2.27, "learning_rate": 7.728299223712068e-05, "loss": 2.7541, "step": 28971 }, { "epoch": 2.27, "learning_rate": 7.728220810789619e-05, "loss": 2.8365, "step": 28972 }, { "epoch": 2.27, "learning_rate": 7.728142397867169e-05, "loss": 3.1315, "step": 28973 }, { "epoch": 2.27, "learning_rate": 7.72806398494472e-05, "loss": 2.8971, "step": 28974 }, { "epoch": 2.27, "learning_rate": 7.72798557202227e-05, "loss": 2.8683, "step": 28975 }, { "epoch": 2.27, "learning_rate": 7.72790715909982e-05, "loss": 2.8088, "step": 28976 }, { "epoch": 2.27, "learning_rate": 7.72782874617737e-05, "loss": 2.8903, "step": 28977 }, { "epoch": 2.27, "learning_rate": 7.727750333254921e-05, "loss": 2.8284, "step": 28978 }, { "epoch": 2.27, "learning_rate": 7.727671920332471e-05, "loss": 2.8126, "step": 28979 }, { "epoch": 2.27, "learning_rate": 7.727593507410022e-05, "loss": 2.9975, "step": 28980 }, { "epoch": 2.27, "learning_rate": 7.727515094487572e-05, "loss": 2.6104, "step": 28981 }, { "epoch": 2.27, "learning_rate": 7.727436681565122e-05, "loss": 2.8687, "step": 28982 }, { "epoch": 2.27, "learning_rate": 7.727358268642673e-05, "loss": 2.7924, "step": 28983 }, { "epoch": 2.27, "learning_rate": 7.727279855720223e-05, "loss": 2.807, "step": 28984 }, { "epoch": 2.27, "learning_rate": 7.727201442797773e-05, "loss": 2.7116, "step": 28985 }, { "epoch": 2.27, "learning_rate": 7.727123029875324e-05, "loss": 2.6544, "step": 28986 }, { "epoch": 2.27, "learning_rate": 7.727044616952874e-05, "loss": 2.8217, "step": 28987 }, { "epoch": 2.27, "learning_rate": 7.726966204030425e-05, "loss": 2.8141, "step": 28988 }, { "epoch": 2.27, "learning_rate": 7.726887791107975e-05, "loss": 2.7469, "step": 28989 }, { "epoch": 2.27, "learning_rate": 7.726809378185525e-05, "loss": 2.8049, "step": 28990 }, { "epoch": 2.27, "learning_rate": 7.726730965263076e-05, "loss": 2.6022, "step": 28991 }, { "epoch": 2.27, "learning_rate": 7.726652552340626e-05, "loss": 2.8936, "step": 28992 }, { "epoch": 2.27, "learning_rate": 7.726574139418176e-05, "loss": 2.8032, "step": 28993 }, { "epoch": 2.27, "learning_rate": 7.726495726495727e-05, "loss": 2.902, "step": 28994 }, { "epoch": 2.27, "learning_rate": 7.726417313573277e-05, "loss": 2.8754, "step": 28995 }, { "epoch": 2.27, "learning_rate": 7.726338900650828e-05, "loss": 2.8245, "step": 28996 }, { "epoch": 2.27, "learning_rate": 7.726260487728378e-05, "loss": 2.8878, "step": 28997 }, { "epoch": 2.27, "learning_rate": 7.726182074805928e-05, "loss": 2.8747, "step": 28998 }, { "epoch": 2.27, "learning_rate": 7.726103661883479e-05, "loss": 2.8006, "step": 28999 }, { "epoch": 2.27, "learning_rate": 7.726025248961029e-05, "loss": 2.6762, "step": 29000 }, { "epoch": 2.27, "eval_loss": 2.905728816986084, "eval_runtime": 218.0229, "eval_samples_per_second": 51.976, "eval_steps_per_second": 3.252, "step": 29000 }, { "epoch": 2.27, "learning_rate": 7.72594683603858e-05, "loss": 3.0134, "step": 29001 }, { "epoch": 2.27, "learning_rate": 7.725868423116131e-05, "loss": 3.0216, "step": 29002 }, { "epoch": 2.27, "learning_rate": 7.72579001019368e-05, "loss": 2.8054, "step": 29003 }, { "epoch": 2.27, "learning_rate": 7.72571159727123e-05, "loss": 2.9351, "step": 29004 }, { "epoch": 2.27, "learning_rate": 7.725633184348781e-05, "loss": 2.8044, "step": 29005 }, { "epoch": 2.27, "learning_rate": 7.725554771426331e-05, "loss": 2.9039, "step": 29006 }, { "epoch": 2.27, "learning_rate": 7.725476358503882e-05, "loss": 2.7489, "step": 29007 }, { "epoch": 2.27, "learning_rate": 7.725397945581432e-05, "loss": 2.9218, "step": 29008 }, { "epoch": 2.27, "learning_rate": 7.725319532658982e-05, "loss": 3.0446, "step": 29009 }, { "epoch": 2.27, "learning_rate": 7.725241119736533e-05, "loss": 2.7484, "step": 29010 }, { "epoch": 2.27, "learning_rate": 7.725162706814083e-05, "loss": 3.0791, "step": 29011 }, { "epoch": 2.27, "learning_rate": 7.725084293891633e-05, "loss": 2.6822, "step": 29012 }, { "epoch": 2.27, "learning_rate": 7.725005880969184e-05, "loss": 2.8413, "step": 29013 }, { "epoch": 2.28, "learning_rate": 7.724927468046734e-05, "loss": 2.8287, "step": 29014 }, { "epoch": 2.28, "learning_rate": 7.724849055124285e-05, "loss": 2.6266, "step": 29015 }, { "epoch": 2.28, "learning_rate": 7.724770642201835e-05, "loss": 2.7945, "step": 29016 }, { "epoch": 2.28, "learning_rate": 7.724692229279385e-05, "loss": 2.7999, "step": 29017 }, { "epoch": 2.28, "learning_rate": 7.724613816356936e-05, "loss": 2.6073, "step": 29018 }, { "epoch": 2.28, "learning_rate": 7.724535403434486e-05, "loss": 2.7899, "step": 29019 }, { "epoch": 2.28, "learning_rate": 7.724456990512036e-05, "loss": 2.7506, "step": 29020 }, { "epoch": 2.28, "learning_rate": 7.724378577589587e-05, "loss": 2.8902, "step": 29021 }, { "epoch": 2.28, "learning_rate": 7.724300164667137e-05, "loss": 2.7346, "step": 29022 }, { "epoch": 2.28, "learning_rate": 7.724221751744687e-05, "loss": 2.92, "step": 29023 }, { "epoch": 2.28, "learning_rate": 7.724143338822238e-05, "loss": 2.9643, "step": 29024 }, { "epoch": 2.28, "learning_rate": 7.724064925899788e-05, "loss": 2.902, "step": 29025 }, { "epoch": 2.28, "learning_rate": 7.723986512977339e-05, "loss": 2.98, "step": 29026 }, { "epoch": 2.28, "learning_rate": 7.72390810005489e-05, "loss": 3.1033, "step": 29027 }, { "epoch": 2.28, "learning_rate": 7.72382968713244e-05, "loss": 2.7791, "step": 29028 }, { "epoch": 2.28, "learning_rate": 7.723751274209991e-05, "loss": 2.7362, "step": 29029 }, { "epoch": 2.28, "learning_rate": 7.723672861287541e-05, "loss": 2.6609, "step": 29030 }, { "epoch": 2.28, "learning_rate": 7.723594448365092e-05, "loss": 2.7233, "step": 29031 }, { "epoch": 2.28, "learning_rate": 7.723516035442642e-05, "loss": 2.7157, "step": 29032 }, { "epoch": 2.28, "learning_rate": 7.723437622520193e-05, "loss": 2.7983, "step": 29033 }, { "epoch": 2.28, "learning_rate": 7.723359209597743e-05, "loss": 2.6988, "step": 29034 }, { "epoch": 2.28, "learning_rate": 7.723280796675293e-05, "loss": 3.0103, "step": 29035 }, { "epoch": 2.28, "learning_rate": 7.723202383752844e-05, "loss": 2.8532, "step": 29036 }, { "epoch": 2.28, "learning_rate": 7.723123970830394e-05, "loss": 2.6799, "step": 29037 }, { "epoch": 2.28, "learning_rate": 7.723045557907944e-05, "loss": 2.8506, "step": 29038 }, { "epoch": 2.28, "learning_rate": 7.722967144985493e-05, "loss": 2.8222, "step": 29039 }, { "epoch": 2.28, "learning_rate": 7.722888732063044e-05, "loss": 2.754, "step": 29040 }, { "epoch": 2.28, "learning_rate": 7.722810319140594e-05, "loss": 2.7466, "step": 29041 }, { "epoch": 2.28, "learning_rate": 7.722731906218144e-05, "loss": 2.831, "step": 29042 }, { "epoch": 2.28, "learning_rate": 7.722653493295695e-05, "loss": 2.8424, "step": 29043 }, { "epoch": 2.28, "learning_rate": 7.722575080373245e-05, "loss": 3.081, "step": 29044 }, { "epoch": 2.28, "learning_rate": 7.722496667450796e-05, "loss": 2.7642, "step": 29045 }, { "epoch": 2.28, "learning_rate": 7.722418254528346e-05, "loss": 2.6991, "step": 29046 }, { "epoch": 2.28, "learning_rate": 7.722339841605896e-05, "loss": 2.6794, "step": 29047 }, { "epoch": 2.28, "learning_rate": 7.722261428683447e-05, "loss": 2.9779, "step": 29048 }, { "epoch": 2.28, "learning_rate": 7.722183015760997e-05, "loss": 2.6775, "step": 29049 }, { "epoch": 2.28, "learning_rate": 7.722104602838547e-05, "loss": 2.9001, "step": 29050 }, { "epoch": 2.28, "learning_rate": 7.722026189916098e-05, "loss": 2.9518, "step": 29051 }, { "epoch": 2.28, "learning_rate": 7.721947776993648e-05, "loss": 2.6123, "step": 29052 }, { "epoch": 2.28, "learning_rate": 7.7218693640712e-05, "loss": 2.9251, "step": 29053 }, { "epoch": 2.28, "learning_rate": 7.72179095114875e-05, "loss": 2.825, "step": 29054 }, { "epoch": 2.28, "learning_rate": 7.7217125382263e-05, "loss": 2.7762, "step": 29055 }, { "epoch": 2.28, "learning_rate": 7.721634125303851e-05, "loss": 2.8293, "step": 29056 }, { "epoch": 2.28, "learning_rate": 7.721555712381401e-05, "loss": 2.6322, "step": 29057 }, { "epoch": 2.28, "learning_rate": 7.721477299458952e-05, "loss": 2.735, "step": 29058 }, { "epoch": 2.28, "learning_rate": 7.721398886536502e-05, "loss": 2.7489, "step": 29059 }, { "epoch": 2.28, "learning_rate": 7.721320473614052e-05, "loss": 2.7188, "step": 29060 }, { "epoch": 2.28, "learning_rate": 7.721242060691603e-05, "loss": 2.7186, "step": 29061 }, { "epoch": 2.28, "learning_rate": 7.721163647769153e-05, "loss": 2.767, "step": 29062 }, { "epoch": 2.28, "learning_rate": 7.721085234846704e-05, "loss": 2.8851, "step": 29063 }, { "epoch": 2.28, "learning_rate": 7.721006821924254e-05, "loss": 2.7644, "step": 29064 }, { "epoch": 2.28, "learning_rate": 7.720928409001804e-05, "loss": 2.7612, "step": 29065 }, { "epoch": 2.28, "learning_rate": 7.720849996079355e-05, "loss": 2.8674, "step": 29066 }, { "epoch": 2.28, "learning_rate": 7.720771583156905e-05, "loss": 2.7456, "step": 29067 }, { "epoch": 2.28, "learning_rate": 7.720693170234455e-05, "loss": 2.8927, "step": 29068 }, { "epoch": 2.28, "learning_rate": 7.720614757312006e-05, "loss": 2.7932, "step": 29069 }, { "epoch": 2.28, "learning_rate": 7.720536344389556e-05, "loss": 2.6601, "step": 29070 }, { "epoch": 2.28, "learning_rate": 7.720457931467106e-05, "loss": 2.8225, "step": 29071 }, { "epoch": 2.28, "learning_rate": 7.720379518544657e-05, "loss": 2.7475, "step": 29072 }, { "epoch": 2.28, "learning_rate": 7.720301105622207e-05, "loss": 2.6924, "step": 29073 }, { "epoch": 2.28, "learning_rate": 7.720222692699758e-05, "loss": 2.7288, "step": 29074 }, { "epoch": 2.28, "learning_rate": 7.720144279777307e-05, "loss": 2.8513, "step": 29075 }, { "epoch": 2.28, "learning_rate": 7.720065866854857e-05, "loss": 2.9412, "step": 29076 }, { "epoch": 2.28, "learning_rate": 7.719987453932407e-05, "loss": 2.8593, "step": 29077 }, { "epoch": 2.28, "learning_rate": 7.719909041009959e-05, "loss": 2.7809, "step": 29078 }, { "epoch": 2.28, "learning_rate": 7.71983062808751e-05, "loss": 2.6484, "step": 29079 }, { "epoch": 2.28, "learning_rate": 7.71975221516506e-05, "loss": 2.7985, "step": 29080 }, { "epoch": 2.28, "learning_rate": 7.71967380224261e-05, "loss": 2.9705, "step": 29081 }, { "epoch": 2.28, "learning_rate": 7.71959538932016e-05, "loss": 2.635, "step": 29082 }, { "epoch": 2.28, "learning_rate": 7.719516976397711e-05, "loss": 2.8023, "step": 29083 }, { "epoch": 2.28, "learning_rate": 7.719438563475261e-05, "loss": 2.7534, "step": 29084 }, { "epoch": 2.28, "learning_rate": 7.719360150552812e-05, "loss": 2.7793, "step": 29085 }, { "epoch": 2.28, "learning_rate": 7.719281737630362e-05, "loss": 2.7942, "step": 29086 }, { "epoch": 2.28, "learning_rate": 7.719203324707912e-05, "loss": 2.7606, "step": 29087 }, { "epoch": 2.28, "learning_rate": 7.719124911785463e-05, "loss": 2.9378, "step": 29088 }, { "epoch": 2.28, "learning_rate": 7.719046498863013e-05, "loss": 2.8934, "step": 29089 }, { "epoch": 2.28, "learning_rate": 7.718968085940563e-05, "loss": 2.9007, "step": 29090 }, { "epoch": 2.28, "learning_rate": 7.718889673018114e-05, "loss": 2.5632, "step": 29091 }, { "epoch": 2.28, "learning_rate": 7.718811260095664e-05, "loss": 2.4789, "step": 29092 }, { "epoch": 2.28, "learning_rate": 7.718732847173215e-05, "loss": 2.7286, "step": 29093 }, { "epoch": 2.28, "learning_rate": 7.718654434250765e-05, "loss": 2.8592, "step": 29094 }, { "epoch": 2.28, "learning_rate": 7.718576021328315e-05, "loss": 2.7978, "step": 29095 }, { "epoch": 2.28, "learning_rate": 7.718497608405866e-05, "loss": 2.5145, "step": 29096 }, { "epoch": 2.28, "learning_rate": 7.718419195483416e-05, "loss": 2.7066, "step": 29097 }, { "epoch": 2.28, "learning_rate": 7.718340782560966e-05, "loss": 2.8287, "step": 29098 }, { "epoch": 2.28, "learning_rate": 7.718262369638517e-05, "loss": 2.705, "step": 29099 }, { "epoch": 2.28, "learning_rate": 7.718183956716067e-05, "loss": 2.8538, "step": 29100 }, { "epoch": 2.28, "learning_rate": 7.718105543793618e-05, "loss": 2.8486, "step": 29101 }, { "epoch": 2.28, "learning_rate": 7.718027130871168e-05, "loss": 2.7979, "step": 29102 }, { "epoch": 2.28, "learning_rate": 7.717948717948718e-05, "loss": 2.6678, "step": 29103 }, { "epoch": 2.28, "learning_rate": 7.717870305026269e-05, "loss": 2.7482, "step": 29104 }, { "epoch": 2.28, "learning_rate": 7.717791892103819e-05, "loss": 2.7072, "step": 29105 }, { "epoch": 2.28, "learning_rate": 7.71771347918137e-05, "loss": 2.6143, "step": 29106 }, { "epoch": 2.28, "learning_rate": 7.71763506625892e-05, "loss": 2.5657, "step": 29107 }, { "epoch": 2.28, "learning_rate": 7.71755665333647e-05, "loss": 2.9191, "step": 29108 }, { "epoch": 2.28, "learning_rate": 7.71747824041402e-05, "loss": 2.6775, "step": 29109 }, { "epoch": 2.28, "learning_rate": 7.717399827491572e-05, "loss": 2.6284, "step": 29110 }, { "epoch": 2.28, "learning_rate": 7.717321414569121e-05, "loss": 2.893, "step": 29111 }, { "epoch": 2.28, "learning_rate": 7.717243001646672e-05, "loss": 2.7716, "step": 29112 }, { "epoch": 2.28, "learning_rate": 7.717164588724222e-05, "loss": 2.8416, "step": 29113 }, { "epoch": 2.28, "learning_rate": 7.717086175801772e-05, "loss": 2.7375, "step": 29114 }, { "epoch": 2.28, "learning_rate": 7.717007762879323e-05, "loss": 2.9176, "step": 29115 }, { "epoch": 2.28, "learning_rate": 7.716929349956873e-05, "loss": 3.1101, "step": 29116 }, { "epoch": 2.28, "learning_rate": 7.716850937034423e-05, "loss": 2.6808, "step": 29117 }, { "epoch": 2.28, "learning_rate": 7.716772524111974e-05, "loss": 2.9083, "step": 29118 }, { "epoch": 2.28, "learning_rate": 7.716694111189524e-05, "loss": 2.6269, "step": 29119 }, { "epoch": 2.28, "learning_rate": 7.716615698267074e-05, "loss": 2.8556, "step": 29120 }, { "epoch": 2.28, "learning_rate": 7.716537285344625e-05, "loss": 2.7686, "step": 29121 }, { "epoch": 2.28, "learning_rate": 7.716458872422175e-05, "loss": 2.9616, "step": 29122 }, { "epoch": 2.28, "learning_rate": 7.716380459499726e-05, "loss": 2.8927, "step": 29123 }, { "epoch": 2.28, "learning_rate": 7.716302046577276e-05, "loss": 2.8169, "step": 29124 }, { "epoch": 2.28, "learning_rate": 7.716223633654826e-05, "loss": 3.1158, "step": 29125 }, { "epoch": 2.28, "learning_rate": 7.716145220732377e-05, "loss": 2.7167, "step": 29126 }, { "epoch": 2.28, "learning_rate": 7.716066807809927e-05, "loss": 2.7346, "step": 29127 }, { "epoch": 2.28, "learning_rate": 7.715988394887477e-05, "loss": 2.69, "step": 29128 }, { "epoch": 2.28, "learning_rate": 7.715909981965028e-05, "loss": 2.7979, "step": 29129 }, { "epoch": 2.28, "learning_rate": 7.715831569042578e-05, "loss": 3.0854, "step": 29130 }, { "epoch": 2.28, "learning_rate": 7.715753156120129e-05, "loss": 2.7137, "step": 29131 }, { "epoch": 2.28, "learning_rate": 7.715674743197679e-05, "loss": 2.9004, "step": 29132 }, { "epoch": 2.28, "learning_rate": 7.715596330275229e-05, "loss": 2.8471, "step": 29133 }, { "epoch": 2.28, "learning_rate": 7.71551791735278e-05, "loss": 2.6962, "step": 29134 }, { "epoch": 2.28, "learning_rate": 7.715439504430331e-05, "loss": 2.8151, "step": 29135 }, { "epoch": 2.28, "learning_rate": 7.715361091507882e-05, "loss": 2.7258, "step": 29136 }, { "epoch": 2.28, "learning_rate": 7.715282678585432e-05, "loss": 2.8427, "step": 29137 }, { "epoch": 2.28, "learning_rate": 7.715204265662982e-05, "loss": 2.8181, "step": 29138 }, { "epoch": 2.28, "learning_rate": 7.715125852740533e-05, "loss": 2.9138, "step": 29139 }, { "epoch": 2.28, "learning_rate": 7.715047439818083e-05, "loss": 2.9118, "step": 29140 }, { "epoch": 2.29, "learning_rate": 7.714969026895634e-05, "loss": 2.7162, "step": 29141 }, { "epoch": 2.29, "learning_rate": 7.714890613973184e-05, "loss": 2.5364, "step": 29142 }, { "epoch": 2.29, "learning_rate": 7.714812201050734e-05, "loss": 2.7173, "step": 29143 }, { "epoch": 2.29, "learning_rate": 7.714733788128285e-05, "loss": 2.7028, "step": 29144 }, { "epoch": 2.29, "learning_rate": 7.714655375205835e-05, "loss": 2.9378, "step": 29145 }, { "epoch": 2.29, "learning_rate": 7.714576962283385e-05, "loss": 2.7996, "step": 29146 }, { "epoch": 2.29, "learning_rate": 7.714498549360934e-05, "loss": 2.8751, "step": 29147 }, { "epoch": 2.29, "learning_rate": 7.714420136438485e-05, "loss": 2.8808, "step": 29148 }, { "epoch": 2.29, "learning_rate": 7.714341723516035e-05, "loss": 2.8625, "step": 29149 }, { "epoch": 2.29, "learning_rate": 7.714263310593586e-05, "loss": 2.8388, "step": 29150 }, { "epoch": 2.29, "learning_rate": 7.714184897671136e-05, "loss": 2.8862, "step": 29151 }, { "epoch": 2.29, "learning_rate": 7.714106484748686e-05, "loss": 2.7502, "step": 29152 }, { "epoch": 2.29, "learning_rate": 7.714028071826237e-05, "loss": 2.9394, "step": 29153 }, { "epoch": 2.29, "learning_rate": 7.713949658903787e-05, "loss": 2.6185, "step": 29154 }, { "epoch": 2.29, "learning_rate": 7.713871245981337e-05, "loss": 2.9376, "step": 29155 }, { "epoch": 2.29, "learning_rate": 7.713792833058888e-05, "loss": 2.7305, "step": 29156 }, { "epoch": 2.29, "learning_rate": 7.713714420136438e-05, "loss": 2.7624, "step": 29157 }, { "epoch": 2.29, "learning_rate": 7.713636007213988e-05, "loss": 3.1198, "step": 29158 }, { "epoch": 2.29, "learning_rate": 7.713557594291539e-05, "loss": 2.606, "step": 29159 }, { "epoch": 2.29, "learning_rate": 7.713479181369089e-05, "loss": 2.9093, "step": 29160 }, { "epoch": 2.29, "learning_rate": 7.713400768446641e-05, "loss": 2.8934, "step": 29161 }, { "epoch": 2.29, "learning_rate": 7.713322355524191e-05, "loss": 2.7316, "step": 29162 }, { "epoch": 2.29, "learning_rate": 7.713243942601742e-05, "loss": 2.8052, "step": 29163 }, { "epoch": 2.29, "learning_rate": 7.713165529679292e-05, "loss": 2.9446, "step": 29164 }, { "epoch": 2.29, "learning_rate": 7.713087116756842e-05, "loss": 2.7569, "step": 29165 }, { "epoch": 2.29, "learning_rate": 7.713008703834393e-05, "loss": 2.6496, "step": 29166 }, { "epoch": 2.29, "learning_rate": 7.712930290911943e-05, "loss": 3.0272, "step": 29167 }, { "epoch": 2.29, "learning_rate": 7.712851877989494e-05, "loss": 3.2511, "step": 29168 }, { "epoch": 2.29, "learning_rate": 7.712773465067044e-05, "loss": 2.8304, "step": 29169 }, { "epoch": 2.29, "learning_rate": 7.712695052144594e-05, "loss": 2.8043, "step": 29170 }, { "epoch": 2.29, "learning_rate": 7.712616639222145e-05, "loss": 2.7292, "step": 29171 }, { "epoch": 2.29, "learning_rate": 7.712538226299695e-05, "loss": 2.8816, "step": 29172 }, { "epoch": 2.29, "learning_rate": 7.712459813377245e-05, "loss": 2.7488, "step": 29173 }, { "epoch": 2.29, "learning_rate": 7.712381400454796e-05, "loss": 3.1364, "step": 29174 }, { "epoch": 2.29, "learning_rate": 7.712302987532346e-05, "loss": 2.7358, "step": 29175 }, { "epoch": 2.29, "learning_rate": 7.712224574609896e-05, "loss": 3.0234, "step": 29176 }, { "epoch": 2.29, "learning_rate": 7.712146161687447e-05, "loss": 2.6457, "step": 29177 }, { "epoch": 2.29, "learning_rate": 7.712067748764997e-05, "loss": 2.8473, "step": 29178 }, { "epoch": 2.29, "learning_rate": 7.711989335842548e-05, "loss": 2.757, "step": 29179 }, { "epoch": 2.29, "learning_rate": 7.711910922920098e-05, "loss": 2.898, "step": 29180 }, { "epoch": 2.29, "learning_rate": 7.711832509997648e-05, "loss": 2.6829, "step": 29181 }, { "epoch": 2.29, "learning_rate": 7.711754097075199e-05, "loss": 2.8799, "step": 29182 }, { "epoch": 2.29, "learning_rate": 7.711675684152748e-05, "loss": 2.982, "step": 29183 }, { "epoch": 2.29, "learning_rate": 7.711597271230298e-05, "loss": 2.846, "step": 29184 }, { "epoch": 2.29, "learning_rate": 7.711518858307848e-05, "loss": 3.0985, "step": 29185 }, { "epoch": 2.29, "learning_rate": 7.711440445385399e-05, "loss": 2.8643, "step": 29186 }, { "epoch": 2.29, "learning_rate": 7.71136203246295e-05, "loss": 2.7971, "step": 29187 }, { "epoch": 2.29, "learning_rate": 7.711283619540501e-05, "loss": 2.7644, "step": 29188 }, { "epoch": 2.29, "learning_rate": 7.711205206618051e-05, "loss": 2.8698, "step": 29189 }, { "epoch": 2.29, "learning_rate": 7.711126793695602e-05, "loss": 2.7359, "step": 29190 }, { "epoch": 2.29, "learning_rate": 7.711048380773152e-05, "loss": 2.8508, "step": 29191 }, { "epoch": 2.29, "learning_rate": 7.710969967850702e-05, "loss": 2.6466, "step": 29192 }, { "epoch": 2.29, "learning_rate": 7.710891554928253e-05, "loss": 2.8412, "step": 29193 }, { "epoch": 2.29, "learning_rate": 7.710813142005803e-05, "loss": 2.9419, "step": 29194 }, { "epoch": 2.29, "learning_rate": 7.710734729083353e-05, "loss": 2.7256, "step": 29195 }, { "epoch": 2.29, "learning_rate": 7.710656316160904e-05, "loss": 2.6599, "step": 29196 }, { "epoch": 2.29, "learning_rate": 7.710577903238454e-05, "loss": 3.0074, "step": 29197 }, { "epoch": 2.29, "learning_rate": 7.710499490316005e-05, "loss": 2.878, "step": 29198 }, { "epoch": 2.29, "learning_rate": 7.710421077393555e-05, "loss": 2.9258, "step": 29199 }, { "epoch": 2.29, "learning_rate": 7.710342664471105e-05, "loss": 2.8938, "step": 29200 }, { "epoch": 2.29, "learning_rate": 7.710264251548656e-05, "loss": 3.2291, "step": 29201 }, { "epoch": 2.29, "learning_rate": 7.710185838626206e-05, "loss": 2.7503, "step": 29202 }, { "epoch": 2.29, "learning_rate": 7.710107425703756e-05, "loss": 2.805, "step": 29203 }, { "epoch": 2.29, "learning_rate": 7.710029012781307e-05, "loss": 2.7951, "step": 29204 }, { "epoch": 2.29, "learning_rate": 7.709950599858857e-05, "loss": 2.788, "step": 29205 }, { "epoch": 2.29, "learning_rate": 7.709872186936407e-05, "loss": 2.5871, "step": 29206 }, { "epoch": 2.29, "learning_rate": 7.709793774013958e-05, "loss": 2.717, "step": 29207 }, { "epoch": 2.29, "learning_rate": 7.709715361091508e-05, "loss": 2.6254, "step": 29208 }, { "epoch": 2.29, "learning_rate": 7.709636948169059e-05, "loss": 2.956, "step": 29209 }, { "epoch": 2.29, "learning_rate": 7.709558535246609e-05, "loss": 2.8625, "step": 29210 }, { "epoch": 2.29, "learning_rate": 7.709480122324159e-05, "loss": 2.7407, "step": 29211 }, { "epoch": 2.29, "learning_rate": 7.70940170940171e-05, "loss": 2.7706, "step": 29212 }, { "epoch": 2.29, "learning_rate": 7.70932329647926e-05, "loss": 2.8456, "step": 29213 }, { "epoch": 2.29, "learning_rate": 7.70924488355681e-05, "loss": 2.8072, "step": 29214 }, { "epoch": 2.29, "learning_rate": 7.709166470634361e-05, "loss": 2.8424, "step": 29215 }, { "epoch": 2.29, "learning_rate": 7.709088057711911e-05, "loss": 3.0112, "step": 29216 }, { "epoch": 2.29, "learning_rate": 7.709009644789462e-05, "loss": 2.8091, "step": 29217 }, { "epoch": 2.29, "learning_rate": 7.708931231867013e-05, "loss": 2.9111, "step": 29218 }, { "epoch": 2.29, "learning_rate": 7.708852818944562e-05, "loss": 2.6541, "step": 29219 }, { "epoch": 2.29, "learning_rate": 7.708774406022113e-05, "loss": 2.8957, "step": 29220 }, { "epoch": 2.29, "learning_rate": 7.708695993099663e-05, "loss": 2.9606, "step": 29221 }, { "epoch": 2.29, "learning_rate": 7.708617580177213e-05, "loss": 2.9294, "step": 29222 }, { "epoch": 2.29, "learning_rate": 7.708539167254764e-05, "loss": 2.7084, "step": 29223 }, { "epoch": 2.29, "learning_rate": 7.708460754332314e-05, "loss": 2.5428, "step": 29224 }, { "epoch": 2.29, "learning_rate": 7.708382341409864e-05, "loss": 2.7226, "step": 29225 }, { "epoch": 2.29, "learning_rate": 7.708303928487415e-05, "loss": 2.9716, "step": 29226 }, { "epoch": 2.29, "learning_rate": 7.708225515564965e-05, "loss": 2.8599, "step": 29227 }, { "epoch": 2.29, "learning_rate": 7.708147102642516e-05, "loss": 2.6117, "step": 29228 }, { "epoch": 2.29, "learning_rate": 7.708068689720066e-05, "loss": 2.6737, "step": 29229 }, { "epoch": 2.29, "learning_rate": 7.707990276797616e-05, "loss": 2.6866, "step": 29230 }, { "epoch": 2.29, "learning_rate": 7.707911863875167e-05, "loss": 2.8711, "step": 29231 }, { "epoch": 2.29, "learning_rate": 7.707833450952717e-05, "loss": 2.48, "step": 29232 }, { "epoch": 2.29, "learning_rate": 7.707755038030267e-05, "loss": 2.6886, "step": 29233 }, { "epoch": 2.29, "learning_rate": 7.707676625107818e-05, "loss": 2.8123, "step": 29234 }, { "epoch": 2.29, "learning_rate": 7.707598212185368e-05, "loss": 2.6961, "step": 29235 }, { "epoch": 2.29, "learning_rate": 7.707519799262919e-05, "loss": 2.9384, "step": 29236 }, { "epoch": 2.29, "learning_rate": 7.707441386340469e-05, "loss": 2.7433, "step": 29237 }, { "epoch": 2.29, "learning_rate": 7.707362973418019e-05, "loss": 2.6684, "step": 29238 }, { "epoch": 2.29, "learning_rate": 7.70728456049557e-05, "loss": 2.9526, "step": 29239 }, { "epoch": 2.29, "learning_rate": 7.70720614757312e-05, "loss": 2.7608, "step": 29240 }, { "epoch": 2.29, "learning_rate": 7.70712773465067e-05, "loss": 2.8774, "step": 29241 }, { "epoch": 2.29, "learning_rate": 7.707049321728221e-05, "loss": 2.7975, "step": 29242 }, { "epoch": 2.29, "learning_rate": 7.706970908805772e-05, "loss": 2.8854, "step": 29243 }, { "epoch": 2.29, "learning_rate": 7.706892495883323e-05, "loss": 2.7441, "step": 29244 }, { "epoch": 2.29, "learning_rate": 7.706814082960873e-05, "loss": 2.6894, "step": 29245 }, { "epoch": 2.29, "learning_rate": 7.706735670038424e-05, "loss": 3.0794, "step": 29246 }, { "epoch": 2.29, "learning_rate": 7.706657257115974e-05, "loss": 2.5917, "step": 29247 }, { "epoch": 2.29, "learning_rate": 7.706578844193524e-05, "loss": 2.6231, "step": 29248 }, { "epoch": 2.29, "learning_rate": 7.706500431271075e-05, "loss": 2.6831, "step": 29249 }, { "epoch": 2.29, "learning_rate": 7.706422018348625e-05, "loss": 2.7557, "step": 29250 }, { "epoch": 2.29, "learning_rate": 7.706343605426175e-05, "loss": 2.6936, "step": 29251 }, { "epoch": 2.29, "learning_rate": 7.706265192503726e-05, "loss": 2.8993, "step": 29252 }, { "epoch": 2.29, "learning_rate": 7.706186779581276e-05, "loss": 2.9057, "step": 29253 }, { "epoch": 2.29, "learning_rate": 7.706108366658826e-05, "loss": 2.8455, "step": 29254 }, { "epoch": 2.29, "learning_rate": 7.706029953736376e-05, "loss": 2.8658, "step": 29255 }, { "epoch": 2.29, "learning_rate": 7.705951540813926e-05, "loss": 2.7253, "step": 29256 }, { "epoch": 2.29, "learning_rate": 7.705873127891476e-05, "loss": 2.6916, "step": 29257 }, { "epoch": 2.29, "learning_rate": 7.705794714969027e-05, "loss": 2.877, "step": 29258 }, { "epoch": 2.29, "learning_rate": 7.705716302046577e-05, "loss": 2.584, "step": 29259 }, { "epoch": 2.29, "learning_rate": 7.705637889124127e-05, "loss": 3.0115, "step": 29260 }, { "epoch": 2.29, "learning_rate": 7.705559476201678e-05, "loss": 2.8408, "step": 29261 }, { "epoch": 2.29, "learning_rate": 7.705481063279228e-05, "loss": 2.8142, "step": 29262 }, { "epoch": 2.29, "learning_rate": 7.705402650356778e-05, "loss": 2.8497, "step": 29263 }, { "epoch": 2.29, "learning_rate": 7.705324237434329e-05, "loss": 2.6876, "step": 29264 }, { "epoch": 2.29, "learning_rate": 7.705245824511879e-05, "loss": 2.7375, "step": 29265 }, { "epoch": 2.29, "learning_rate": 7.70516741158943e-05, "loss": 2.5942, "step": 29266 }, { "epoch": 2.29, "learning_rate": 7.70508899866698e-05, "loss": 2.7385, "step": 29267 }, { "epoch": 2.29, "learning_rate": 7.70501058574453e-05, "loss": 2.7108, "step": 29268 }, { "epoch": 2.3, "learning_rate": 7.704932172822082e-05, "loss": 2.7126, "step": 29269 }, { "epoch": 2.3, "learning_rate": 7.704853759899632e-05, "loss": 3.0566, "step": 29270 }, { "epoch": 2.3, "learning_rate": 7.704775346977183e-05, "loss": 2.8645, "step": 29271 }, { "epoch": 2.3, "learning_rate": 7.704696934054733e-05, "loss": 2.9425, "step": 29272 }, { "epoch": 2.3, "learning_rate": 7.704618521132283e-05, "loss": 2.7317, "step": 29273 }, { "epoch": 2.3, "learning_rate": 7.704540108209834e-05, "loss": 2.7884, "step": 29274 }, { "epoch": 2.3, "learning_rate": 7.704461695287384e-05, "loss": 2.7374, "step": 29275 }, { "epoch": 2.3, "learning_rate": 7.704383282364935e-05, "loss": 2.8296, "step": 29276 }, { "epoch": 2.3, "learning_rate": 7.704304869442485e-05, "loss": 2.8681, "step": 29277 }, { "epoch": 2.3, "learning_rate": 7.704226456520035e-05, "loss": 2.8399, "step": 29278 }, { "epoch": 2.3, "learning_rate": 7.704148043597586e-05, "loss": 2.7676, "step": 29279 }, { "epoch": 2.3, "learning_rate": 7.704069630675136e-05, "loss": 2.9266, "step": 29280 }, { "epoch": 2.3, "learning_rate": 7.703991217752686e-05, "loss": 2.7039, "step": 29281 }, { "epoch": 2.3, "learning_rate": 7.703912804830237e-05, "loss": 2.7334, "step": 29282 }, { "epoch": 2.3, "learning_rate": 7.703834391907787e-05, "loss": 2.6817, "step": 29283 }, { "epoch": 2.3, "learning_rate": 7.703755978985338e-05, "loss": 2.8927, "step": 29284 }, { "epoch": 2.3, "learning_rate": 7.703677566062888e-05, "loss": 2.6543, "step": 29285 }, { "epoch": 2.3, "learning_rate": 7.703599153140438e-05, "loss": 2.935, "step": 29286 }, { "epoch": 2.3, "learning_rate": 7.703520740217989e-05, "loss": 2.8514, "step": 29287 }, { "epoch": 2.3, "learning_rate": 7.703442327295539e-05, "loss": 2.9932, "step": 29288 }, { "epoch": 2.3, "learning_rate": 7.70336391437309e-05, "loss": 2.7107, "step": 29289 }, { "epoch": 2.3, "learning_rate": 7.70328550145064e-05, "loss": 3.0129, "step": 29290 }, { "epoch": 2.3, "learning_rate": 7.703207088528189e-05, "loss": 3.1922, "step": 29291 }, { "epoch": 2.3, "learning_rate": 7.703128675605739e-05, "loss": 2.9483, "step": 29292 }, { "epoch": 2.3, "learning_rate": 7.70305026268329e-05, "loss": 2.8553, "step": 29293 }, { "epoch": 2.3, "learning_rate": 7.70297184976084e-05, "loss": 2.694, "step": 29294 }, { "epoch": 2.3, "learning_rate": 7.702893436838392e-05, "loss": 2.8079, "step": 29295 }, { "epoch": 2.3, "learning_rate": 7.702815023915942e-05, "loss": 2.7379, "step": 29296 }, { "epoch": 2.3, "learning_rate": 7.702736610993492e-05, "loss": 3.0131, "step": 29297 }, { "epoch": 2.3, "learning_rate": 7.702658198071043e-05, "loss": 3.1554, "step": 29298 }, { "epoch": 2.3, "learning_rate": 7.702579785148593e-05, "loss": 2.8306, "step": 29299 }, { "epoch": 2.3, "learning_rate": 7.702501372226143e-05, "loss": 2.9038, "step": 29300 }, { "epoch": 2.3, "learning_rate": 7.702422959303694e-05, "loss": 2.8136, "step": 29301 }, { "epoch": 2.3, "learning_rate": 7.702344546381244e-05, "loss": 2.9016, "step": 29302 }, { "epoch": 2.3, "learning_rate": 7.702266133458795e-05, "loss": 2.8904, "step": 29303 }, { "epoch": 2.3, "learning_rate": 7.702187720536345e-05, "loss": 2.8135, "step": 29304 }, { "epoch": 2.3, "learning_rate": 7.702109307613895e-05, "loss": 2.8068, "step": 29305 }, { "epoch": 2.3, "learning_rate": 7.702030894691446e-05, "loss": 2.9374, "step": 29306 }, { "epoch": 2.3, "learning_rate": 7.701952481768996e-05, "loss": 2.8275, "step": 29307 }, { "epoch": 2.3, "learning_rate": 7.701874068846546e-05, "loss": 2.9449, "step": 29308 }, { "epoch": 2.3, "learning_rate": 7.701795655924097e-05, "loss": 2.8105, "step": 29309 }, { "epoch": 2.3, "learning_rate": 7.701717243001647e-05, "loss": 2.7042, "step": 29310 }, { "epoch": 2.3, "learning_rate": 7.701638830079197e-05, "loss": 2.7651, "step": 29311 }, { "epoch": 2.3, "learning_rate": 7.701560417156748e-05, "loss": 2.8695, "step": 29312 }, { "epoch": 2.3, "learning_rate": 7.701482004234298e-05, "loss": 2.8677, "step": 29313 }, { "epoch": 2.3, "learning_rate": 7.701403591311849e-05, "loss": 2.8012, "step": 29314 }, { "epoch": 2.3, "learning_rate": 7.701325178389399e-05, "loss": 2.8762, "step": 29315 }, { "epoch": 2.3, "learning_rate": 7.701246765466949e-05, "loss": 2.8589, "step": 29316 }, { "epoch": 2.3, "learning_rate": 7.7011683525445e-05, "loss": 2.799, "step": 29317 }, { "epoch": 2.3, "learning_rate": 7.70108993962205e-05, "loss": 3.0579, "step": 29318 }, { "epoch": 2.3, "learning_rate": 7.7010115266996e-05, "loss": 2.7889, "step": 29319 }, { "epoch": 2.3, "learning_rate": 7.700933113777151e-05, "loss": 2.5824, "step": 29320 }, { "epoch": 2.3, "learning_rate": 7.700854700854701e-05, "loss": 2.7231, "step": 29321 }, { "epoch": 2.3, "learning_rate": 7.700776287932251e-05, "loss": 2.6718, "step": 29322 }, { "epoch": 2.3, "learning_rate": 7.700697875009802e-05, "loss": 2.5469, "step": 29323 }, { "epoch": 2.3, "learning_rate": 7.700619462087352e-05, "loss": 2.6241, "step": 29324 }, { "epoch": 2.3, "learning_rate": 7.700541049164903e-05, "loss": 2.801, "step": 29325 }, { "epoch": 2.3, "learning_rate": 7.700462636242454e-05, "loss": 2.9631, "step": 29326 }, { "epoch": 2.3, "learning_rate": 7.700384223320003e-05, "loss": 3.0524, "step": 29327 }, { "epoch": 2.3, "learning_rate": 7.700305810397554e-05, "loss": 2.6665, "step": 29328 }, { "epoch": 2.3, "learning_rate": 7.700227397475104e-05, "loss": 2.8445, "step": 29329 }, { "epoch": 2.3, "learning_rate": 7.700148984552654e-05, "loss": 2.9065, "step": 29330 }, { "epoch": 2.3, "learning_rate": 7.700070571630205e-05, "loss": 2.7022, "step": 29331 }, { "epoch": 2.3, "learning_rate": 7.699992158707755e-05, "loss": 2.9961, "step": 29332 }, { "epoch": 2.3, "learning_rate": 7.699913745785306e-05, "loss": 2.9149, "step": 29333 }, { "epoch": 2.3, "learning_rate": 7.699835332862856e-05, "loss": 2.9355, "step": 29334 }, { "epoch": 2.3, "learning_rate": 7.699756919940406e-05, "loss": 2.7187, "step": 29335 }, { "epoch": 2.3, "learning_rate": 7.699678507017957e-05, "loss": 2.7451, "step": 29336 }, { "epoch": 2.3, "learning_rate": 7.699600094095507e-05, "loss": 2.7714, "step": 29337 }, { "epoch": 2.3, "learning_rate": 7.699521681173057e-05, "loss": 2.9708, "step": 29338 }, { "epoch": 2.3, "learning_rate": 7.699443268250608e-05, "loss": 2.8363, "step": 29339 }, { "epoch": 2.3, "learning_rate": 7.699364855328158e-05, "loss": 2.8751, "step": 29340 }, { "epoch": 2.3, "learning_rate": 7.699286442405708e-05, "loss": 2.798, "step": 29341 }, { "epoch": 2.3, "learning_rate": 7.699208029483259e-05, "loss": 2.7553, "step": 29342 }, { "epoch": 2.3, "learning_rate": 7.699129616560809e-05, "loss": 2.9917, "step": 29343 }, { "epoch": 2.3, "learning_rate": 7.69905120363836e-05, "loss": 2.8442, "step": 29344 }, { "epoch": 2.3, "learning_rate": 7.69897279071591e-05, "loss": 2.9162, "step": 29345 }, { "epoch": 2.3, "learning_rate": 7.69889437779346e-05, "loss": 2.8789, "step": 29346 }, { "epoch": 2.3, "learning_rate": 7.698815964871011e-05, "loss": 2.9105, "step": 29347 }, { "epoch": 2.3, "learning_rate": 7.698737551948561e-05, "loss": 2.9594, "step": 29348 }, { "epoch": 2.3, "learning_rate": 7.698659139026111e-05, "loss": 2.756, "step": 29349 }, { "epoch": 2.3, "learning_rate": 7.698580726103662e-05, "loss": 2.6014, "step": 29350 }, { "epoch": 2.3, "learning_rate": 7.698502313181214e-05, "loss": 2.5237, "step": 29351 }, { "epoch": 2.3, "learning_rate": 7.698423900258764e-05, "loss": 2.6879, "step": 29352 }, { "epoch": 2.3, "learning_rate": 7.698345487336314e-05, "loss": 3.1224, "step": 29353 }, { "epoch": 2.3, "learning_rate": 7.698267074413865e-05, "loss": 2.7451, "step": 29354 }, { "epoch": 2.3, "learning_rate": 7.698188661491415e-05, "loss": 3.0113, "step": 29355 }, { "epoch": 2.3, "learning_rate": 7.698110248568965e-05, "loss": 2.8752, "step": 29356 }, { "epoch": 2.3, "learning_rate": 7.698031835646516e-05, "loss": 2.8432, "step": 29357 }, { "epoch": 2.3, "learning_rate": 7.697953422724066e-05, "loss": 2.7373, "step": 29358 }, { "epoch": 2.3, "learning_rate": 7.697875009801616e-05, "loss": 2.695, "step": 29359 }, { "epoch": 2.3, "learning_rate": 7.697796596879167e-05, "loss": 2.6632, "step": 29360 }, { "epoch": 2.3, "learning_rate": 7.697718183956717e-05, "loss": 2.7568, "step": 29361 }, { "epoch": 2.3, "learning_rate": 7.697639771034268e-05, "loss": 2.9226, "step": 29362 }, { "epoch": 2.3, "learning_rate": 7.697561358111817e-05, "loss": 2.7311, "step": 29363 }, { "epoch": 2.3, "learning_rate": 7.697482945189367e-05, "loss": 2.9318, "step": 29364 }, { "epoch": 2.3, "learning_rate": 7.697404532266917e-05, "loss": 2.6924, "step": 29365 }, { "epoch": 2.3, "learning_rate": 7.697326119344468e-05, "loss": 2.6809, "step": 29366 }, { "epoch": 2.3, "learning_rate": 7.697247706422018e-05, "loss": 2.8829, "step": 29367 }, { "epoch": 2.3, "learning_rate": 7.697169293499568e-05, "loss": 2.7962, "step": 29368 }, { "epoch": 2.3, "learning_rate": 7.697090880577119e-05, "loss": 2.7267, "step": 29369 }, { "epoch": 2.3, "learning_rate": 7.697012467654669e-05, "loss": 2.7676, "step": 29370 }, { "epoch": 2.3, "learning_rate": 7.69693405473222e-05, "loss": 2.791, "step": 29371 }, { "epoch": 2.3, "learning_rate": 7.69685564180977e-05, "loss": 2.6583, "step": 29372 }, { "epoch": 2.3, "learning_rate": 7.69677722888732e-05, "loss": 2.5082, "step": 29373 }, { "epoch": 2.3, "learning_rate": 7.69669881596487e-05, "loss": 2.6059, "step": 29374 }, { "epoch": 2.3, "learning_rate": 7.696620403042421e-05, "loss": 2.7089, "step": 29375 }, { "epoch": 2.3, "learning_rate": 7.696541990119971e-05, "loss": 2.91, "step": 29376 }, { "epoch": 2.3, "learning_rate": 7.696463577197523e-05, "loss": 2.8951, "step": 29377 }, { "epoch": 2.3, "learning_rate": 7.696385164275073e-05, "loss": 2.8104, "step": 29378 }, { "epoch": 2.3, "learning_rate": 7.696306751352624e-05, "loss": 3.264, "step": 29379 }, { "epoch": 2.3, "learning_rate": 7.696228338430174e-05, "loss": 2.8229, "step": 29380 }, { "epoch": 2.3, "learning_rate": 7.696149925507725e-05, "loss": 3.0146, "step": 29381 }, { "epoch": 2.3, "learning_rate": 7.696071512585275e-05, "loss": 2.6854, "step": 29382 }, { "epoch": 2.3, "learning_rate": 7.695993099662825e-05, "loss": 2.9445, "step": 29383 }, { "epoch": 2.3, "learning_rate": 7.695914686740376e-05, "loss": 2.6213, "step": 29384 }, { "epoch": 2.3, "learning_rate": 7.695836273817926e-05, "loss": 2.7942, "step": 29385 }, { "epoch": 2.3, "learning_rate": 7.695757860895476e-05, "loss": 2.7943, "step": 29386 }, { "epoch": 2.3, "learning_rate": 7.695679447973027e-05, "loss": 2.8067, "step": 29387 }, { "epoch": 2.3, "learning_rate": 7.695601035050577e-05, "loss": 2.9605, "step": 29388 }, { "epoch": 2.3, "learning_rate": 7.695522622128127e-05, "loss": 2.7259, "step": 29389 }, { "epoch": 2.3, "learning_rate": 7.695444209205678e-05, "loss": 2.8831, "step": 29390 }, { "epoch": 2.3, "learning_rate": 7.695365796283228e-05, "loss": 2.8028, "step": 29391 }, { "epoch": 2.3, "learning_rate": 7.695287383360779e-05, "loss": 2.8215, "step": 29392 }, { "epoch": 2.3, "learning_rate": 7.695208970438329e-05, "loss": 2.7023, "step": 29393 }, { "epoch": 2.3, "learning_rate": 7.69513055751588e-05, "loss": 2.7299, "step": 29394 }, { "epoch": 2.3, "learning_rate": 7.69505214459343e-05, "loss": 2.9508, "step": 29395 }, { "epoch": 2.31, "learning_rate": 7.69497373167098e-05, "loss": 2.7947, "step": 29396 }, { "epoch": 2.31, "learning_rate": 7.69489531874853e-05, "loss": 2.8451, "step": 29397 }, { "epoch": 2.31, "learning_rate": 7.694816905826081e-05, "loss": 2.8282, "step": 29398 }, { "epoch": 2.31, "learning_rate": 7.69473849290363e-05, "loss": 2.7694, "step": 29399 }, { "epoch": 2.31, "learning_rate": 7.69466007998118e-05, "loss": 2.7849, "step": 29400 }, { "epoch": 2.31, "learning_rate": 7.69458166705873e-05, "loss": 2.6703, "step": 29401 }, { "epoch": 2.31, "learning_rate": 7.694503254136281e-05, "loss": 2.8816, "step": 29402 }, { "epoch": 2.31, "learning_rate": 7.694424841213833e-05, "loss": 2.7185, "step": 29403 }, { "epoch": 2.31, "learning_rate": 7.694346428291383e-05, "loss": 2.7862, "step": 29404 }, { "epoch": 2.31, "learning_rate": 7.694268015368933e-05, "loss": 2.6987, "step": 29405 }, { "epoch": 2.31, "learning_rate": 7.694189602446484e-05, "loss": 2.9696, "step": 29406 }, { "epoch": 2.31, "learning_rate": 7.694111189524034e-05, "loss": 2.8808, "step": 29407 }, { "epoch": 2.31, "learning_rate": 7.694032776601584e-05, "loss": 3.0259, "step": 29408 }, { "epoch": 2.31, "learning_rate": 7.693954363679135e-05, "loss": 2.8954, "step": 29409 }, { "epoch": 2.31, "learning_rate": 7.693875950756685e-05, "loss": 2.9466, "step": 29410 }, { "epoch": 2.31, "learning_rate": 7.693797537834236e-05, "loss": 2.5642, "step": 29411 }, { "epoch": 2.31, "learning_rate": 7.693719124911786e-05, "loss": 2.8775, "step": 29412 }, { "epoch": 2.31, "learning_rate": 7.693640711989336e-05, "loss": 2.8856, "step": 29413 }, { "epoch": 2.31, "learning_rate": 7.693562299066887e-05, "loss": 2.6412, "step": 29414 }, { "epoch": 2.31, "learning_rate": 7.693483886144437e-05, "loss": 2.8951, "step": 29415 }, { "epoch": 2.31, "learning_rate": 7.693405473221987e-05, "loss": 2.7664, "step": 29416 }, { "epoch": 2.31, "learning_rate": 7.693327060299538e-05, "loss": 2.7513, "step": 29417 }, { "epoch": 2.31, "learning_rate": 7.693248647377088e-05, "loss": 2.8711, "step": 29418 }, { "epoch": 2.31, "learning_rate": 7.693170234454639e-05, "loss": 2.9544, "step": 29419 }, { "epoch": 2.31, "learning_rate": 7.693091821532189e-05, "loss": 2.5589, "step": 29420 }, { "epoch": 2.31, "learning_rate": 7.693013408609739e-05, "loss": 2.9264, "step": 29421 }, { "epoch": 2.31, "learning_rate": 7.69293499568729e-05, "loss": 2.6088, "step": 29422 }, { "epoch": 2.31, "learning_rate": 7.69285658276484e-05, "loss": 2.8241, "step": 29423 }, { "epoch": 2.31, "learning_rate": 7.69277816984239e-05, "loss": 2.8507, "step": 29424 }, { "epoch": 2.31, "learning_rate": 7.692699756919941e-05, "loss": 2.945, "step": 29425 }, { "epoch": 2.31, "learning_rate": 7.692621343997491e-05, "loss": 2.5759, "step": 29426 }, { "epoch": 2.31, "learning_rate": 7.692542931075041e-05, "loss": 2.7466, "step": 29427 }, { "epoch": 2.31, "learning_rate": 7.692464518152592e-05, "loss": 2.8625, "step": 29428 }, { "epoch": 2.31, "learning_rate": 7.692386105230142e-05, "loss": 2.7462, "step": 29429 }, { "epoch": 2.31, "learning_rate": 7.692307692307693e-05, "loss": 2.8322, "step": 29430 }, { "epoch": 2.31, "learning_rate": 7.692229279385243e-05, "loss": 2.7453, "step": 29431 }, { "epoch": 2.31, "learning_rate": 7.692150866462793e-05, "loss": 2.8599, "step": 29432 }, { "epoch": 2.31, "learning_rate": 7.692072453540344e-05, "loss": 2.8878, "step": 29433 }, { "epoch": 2.31, "learning_rate": 7.691994040617895e-05, "loss": 2.5434, "step": 29434 }, { "epoch": 2.31, "learning_rate": 7.691915627695444e-05, "loss": 2.6462, "step": 29435 }, { "epoch": 2.31, "learning_rate": 7.691837214772995e-05, "loss": 2.7898, "step": 29436 }, { "epoch": 2.31, "learning_rate": 7.691758801850545e-05, "loss": 2.6421, "step": 29437 }, { "epoch": 2.31, "learning_rate": 7.691680388928096e-05, "loss": 2.722, "step": 29438 }, { "epoch": 2.31, "learning_rate": 7.691601976005646e-05, "loss": 2.7635, "step": 29439 }, { "epoch": 2.31, "learning_rate": 7.691523563083196e-05, "loss": 2.9063, "step": 29440 }, { "epoch": 2.31, "learning_rate": 7.691445150160747e-05, "loss": 2.8297, "step": 29441 }, { "epoch": 2.31, "learning_rate": 7.691366737238297e-05, "loss": 2.9238, "step": 29442 }, { "epoch": 2.31, "learning_rate": 7.691288324315847e-05, "loss": 2.9192, "step": 29443 }, { "epoch": 2.31, "learning_rate": 7.691209911393398e-05, "loss": 2.6834, "step": 29444 }, { "epoch": 2.31, "learning_rate": 7.691131498470948e-05, "loss": 2.7363, "step": 29445 }, { "epoch": 2.31, "learning_rate": 7.691053085548498e-05, "loss": 2.9078, "step": 29446 }, { "epoch": 2.31, "learning_rate": 7.690974672626049e-05, "loss": 3.2242, "step": 29447 }, { "epoch": 2.31, "learning_rate": 7.690896259703599e-05, "loss": 2.575, "step": 29448 }, { "epoch": 2.31, "learning_rate": 7.69081784678115e-05, "loss": 2.7949, "step": 29449 }, { "epoch": 2.31, "learning_rate": 7.6907394338587e-05, "loss": 2.7249, "step": 29450 }, { "epoch": 2.31, "learning_rate": 7.69066102093625e-05, "loss": 2.8904, "step": 29451 }, { "epoch": 2.31, "learning_rate": 7.6905826080138e-05, "loss": 2.6612, "step": 29452 }, { "epoch": 2.31, "learning_rate": 7.690504195091351e-05, "loss": 2.7283, "step": 29453 }, { "epoch": 2.31, "learning_rate": 7.690425782168901e-05, "loss": 2.918, "step": 29454 }, { "epoch": 2.31, "learning_rate": 7.690347369246452e-05, "loss": 2.9501, "step": 29455 }, { "epoch": 2.31, "learning_rate": 7.690268956324002e-05, "loss": 2.8785, "step": 29456 }, { "epoch": 2.31, "learning_rate": 7.690190543401553e-05, "loss": 2.9032, "step": 29457 }, { "epoch": 2.31, "learning_rate": 7.690112130479103e-05, "loss": 2.7399, "step": 29458 }, { "epoch": 2.31, "learning_rate": 7.690033717556655e-05, "loss": 2.9472, "step": 29459 }, { "epoch": 2.31, "learning_rate": 7.689955304634205e-05, "loss": 2.864, "step": 29460 }, { "epoch": 2.31, "learning_rate": 7.689876891711755e-05, "loss": 2.7456, "step": 29461 }, { "epoch": 2.31, "learning_rate": 7.689798478789306e-05, "loss": 2.6326, "step": 29462 }, { "epoch": 2.31, "learning_rate": 7.689720065866856e-05, "loss": 2.7714, "step": 29463 }, { "epoch": 2.31, "learning_rate": 7.689641652944406e-05, "loss": 2.716, "step": 29464 }, { "epoch": 2.31, "learning_rate": 7.689563240021957e-05, "loss": 2.9921, "step": 29465 }, { "epoch": 2.31, "learning_rate": 7.689484827099507e-05, "loss": 2.7757, "step": 29466 }, { "epoch": 2.31, "learning_rate": 7.689406414177058e-05, "loss": 2.8152, "step": 29467 }, { "epoch": 2.31, "learning_rate": 7.689328001254608e-05, "loss": 2.5868, "step": 29468 }, { "epoch": 2.31, "learning_rate": 7.689249588332158e-05, "loss": 3.1009, "step": 29469 }, { "epoch": 2.31, "learning_rate": 7.689171175409709e-05, "loss": 2.5507, "step": 29470 }, { "epoch": 2.31, "learning_rate": 7.689092762487258e-05, "loss": 2.8207, "step": 29471 }, { "epoch": 2.31, "learning_rate": 7.689014349564808e-05, "loss": 3.0749, "step": 29472 }, { "epoch": 2.31, "learning_rate": 7.688935936642358e-05, "loss": 2.8845, "step": 29473 }, { "epoch": 2.31, "learning_rate": 7.688857523719909e-05, "loss": 2.5954, "step": 29474 }, { "epoch": 2.31, "learning_rate": 7.688779110797459e-05, "loss": 2.7355, "step": 29475 }, { "epoch": 2.31, "learning_rate": 7.68870069787501e-05, "loss": 2.7046, "step": 29476 }, { "epoch": 2.31, "learning_rate": 7.68862228495256e-05, "loss": 2.7573, "step": 29477 }, { "epoch": 2.31, "learning_rate": 7.68854387203011e-05, "loss": 2.6548, "step": 29478 }, { "epoch": 2.31, "learning_rate": 7.68846545910766e-05, "loss": 2.7931, "step": 29479 }, { "epoch": 2.31, "learning_rate": 7.688387046185211e-05, "loss": 2.6545, "step": 29480 }, { "epoch": 2.31, "learning_rate": 7.688308633262761e-05, "loss": 2.9596, "step": 29481 }, { "epoch": 2.31, "learning_rate": 7.688230220340312e-05, "loss": 2.6603, "step": 29482 }, { "epoch": 2.31, "learning_rate": 7.688151807417862e-05, "loss": 2.818, "step": 29483 }, { "epoch": 2.31, "learning_rate": 7.688073394495412e-05, "loss": 2.7664, "step": 29484 }, { "epoch": 2.31, "learning_rate": 7.687994981572964e-05, "loss": 2.8451, "step": 29485 }, { "epoch": 2.31, "learning_rate": 7.687916568650515e-05, "loss": 2.9679, "step": 29486 }, { "epoch": 2.31, "learning_rate": 7.687838155728065e-05, "loss": 2.6842, "step": 29487 }, { "epoch": 2.31, "learning_rate": 7.687759742805615e-05, "loss": 2.7312, "step": 29488 }, { "epoch": 2.31, "learning_rate": 7.687681329883166e-05, "loss": 2.959, "step": 29489 }, { "epoch": 2.31, "learning_rate": 7.687602916960716e-05, "loss": 2.9047, "step": 29490 }, { "epoch": 2.31, "learning_rate": 7.687524504038266e-05, "loss": 2.9158, "step": 29491 }, { "epoch": 2.31, "learning_rate": 7.687446091115817e-05, "loss": 3.044, "step": 29492 }, { "epoch": 2.31, "learning_rate": 7.687367678193367e-05, "loss": 2.9615, "step": 29493 }, { "epoch": 2.31, "learning_rate": 7.687289265270917e-05, "loss": 2.7435, "step": 29494 }, { "epoch": 2.31, "learning_rate": 7.687210852348468e-05, "loss": 2.697, "step": 29495 }, { "epoch": 2.31, "learning_rate": 7.687132439426018e-05, "loss": 2.7788, "step": 29496 }, { "epoch": 2.31, "learning_rate": 7.687054026503569e-05, "loss": 2.7534, "step": 29497 }, { "epoch": 2.31, "learning_rate": 7.686975613581119e-05, "loss": 3.2072, "step": 29498 }, { "epoch": 2.31, "learning_rate": 7.686897200658669e-05, "loss": 2.8739, "step": 29499 }, { "epoch": 2.31, "learning_rate": 7.68681878773622e-05, "loss": 2.6142, "step": 29500 }, { "epoch": 2.31, "learning_rate": 7.68674037481377e-05, "loss": 2.975, "step": 29501 }, { "epoch": 2.31, "learning_rate": 7.68666196189132e-05, "loss": 2.9963, "step": 29502 }, { "epoch": 2.31, "learning_rate": 7.686583548968871e-05, "loss": 2.9752, "step": 29503 }, { "epoch": 2.31, "learning_rate": 7.686505136046421e-05, "loss": 2.6967, "step": 29504 }, { "epoch": 2.31, "learning_rate": 7.686426723123972e-05, "loss": 2.7326, "step": 29505 }, { "epoch": 2.31, "learning_rate": 7.686348310201522e-05, "loss": 2.6388, "step": 29506 }, { "epoch": 2.31, "learning_rate": 7.686269897279071e-05, "loss": 2.6975, "step": 29507 }, { "epoch": 2.31, "learning_rate": 7.686191484356621e-05, "loss": 2.7686, "step": 29508 }, { "epoch": 2.31, "learning_rate": 7.686113071434172e-05, "loss": 2.7292, "step": 29509 }, { "epoch": 2.31, "learning_rate": 7.686034658511722e-05, "loss": 2.6814, "step": 29510 }, { "epoch": 2.31, "learning_rate": 7.685956245589274e-05, "loss": 2.7404, "step": 29511 }, { "epoch": 2.31, "learning_rate": 7.685877832666824e-05, "loss": 3.0566, "step": 29512 }, { "epoch": 2.31, "learning_rate": 7.685799419744374e-05, "loss": 2.8715, "step": 29513 }, { "epoch": 2.31, "learning_rate": 7.685721006821925e-05, "loss": 2.7794, "step": 29514 }, { "epoch": 2.31, "learning_rate": 7.685642593899475e-05, "loss": 2.7077, "step": 29515 }, { "epoch": 2.31, "learning_rate": 7.685564180977026e-05, "loss": 2.8473, "step": 29516 }, { "epoch": 2.31, "learning_rate": 7.685485768054576e-05, "loss": 2.9026, "step": 29517 }, { "epoch": 2.31, "learning_rate": 7.685407355132126e-05, "loss": 2.793, "step": 29518 }, { "epoch": 2.31, "learning_rate": 7.685328942209677e-05, "loss": 2.9293, "step": 29519 }, { "epoch": 2.31, "learning_rate": 7.685250529287227e-05, "loss": 2.8114, "step": 29520 }, { "epoch": 2.31, "learning_rate": 7.685172116364777e-05, "loss": 2.9586, "step": 29521 }, { "epoch": 2.31, "learning_rate": 7.685093703442328e-05, "loss": 2.7491, "step": 29522 }, { "epoch": 2.31, "learning_rate": 7.685015290519878e-05, "loss": 2.8333, "step": 29523 }, { "epoch": 2.32, "learning_rate": 7.684936877597428e-05, "loss": 2.6497, "step": 29524 }, { "epoch": 2.32, "learning_rate": 7.684858464674979e-05, "loss": 2.8725, "step": 29525 }, { "epoch": 2.32, "learning_rate": 7.684780051752529e-05, "loss": 2.4672, "step": 29526 }, { "epoch": 2.32, "learning_rate": 7.68470163883008e-05, "loss": 2.8035, "step": 29527 }, { "epoch": 2.32, "learning_rate": 7.68462322590763e-05, "loss": 2.8931, "step": 29528 }, { "epoch": 2.32, "learning_rate": 7.68454481298518e-05, "loss": 2.8562, "step": 29529 }, { "epoch": 2.32, "learning_rate": 7.684466400062731e-05, "loss": 2.7851, "step": 29530 }, { "epoch": 2.32, "learning_rate": 7.684387987140281e-05, "loss": 2.7664, "step": 29531 }, { "epoch": 2.32, "learning_rate": 7.684309574217831e-05, "loss": 2.7931, "step": 29532 }, { "epoch": 2.32, "learning_rate": 7.684231161295382e-05, "loss": 2.9103, "step": 29533 }, { "epoch": 2.32, "learning_rate": 7.684152748372932e-05, "loss": 2.9112, "step": 29534 }, { "epoch": 2.32, "learning_rate": 7.684074335450483e-05, "loss": 2.6514, "step": 29535 }, { "epoch": 2.32, "learning_rate": 7.683995922528033e-05, "loss": 2.6894, "step": 29536 }, { "epoch": 2.32, "learning_rate": 7.683917509605583e-05, "loss": 2.8542, "step": 29537 }, { "epoch": 2.32, "learning_rate": 7.683839096683134e-05, "loss": 2.8592, "step": 29538 }, { "epoch": 2.32, "learning_rate": 7.683760683760684e-05, "loss": 2.9456, "step": 29539 }, { "epoch": 2.32, "learning_rate": 7.683682270838234e-05, "loss": 2.8869, "step": 29540 }, { "epoch": 2.32, "learning_rate": 7.683603857915785e-05, "loss": 2.7167, "step": 29541 }, { "epoch": 2.32, "learning_rate": 7.683525444993336e-05, "loss": 2.7724, "step": 29542 }, { "epoch": 2.32, "learning_rate": 7.683447032070885e-05, "loss": 2.6136, "step": 29543 }, { "epoch": 2.32, "learning_rate": 7.683368619148436e-05, "loss": 2.7547, "step": 29544 }, { "epoch": 2.32, "learning_rate": 7.683290206225986e-05, "loss": 2.8909, "step": 29545 }, { "epoch": 2.32, "learning_rate": 7.683211793303537e-05, "loss": 2.6523, "step": 29546 }, { "epoch": 2.32, "learning_rate": 7.683133380381087e-05, "loss": 2.8104, "step": 29547 }, { "epoch": 2.32, "learning_rate": 7.683054967458637e-05, "loss": 2.7158, "step": 29548 }, { "epoch": 2.32, "learning_rate": 7.682976554536188e-05, "loss": 2.6733, "step": 29549 }, { "epoch": 2.32, "learning_rate": 7.682898141613738e-05, "loss": 2.6989, "step": 29550 }, { "epoch": 2.32, "learning_rate": 7.682819728691288e-05, "loss": 2.717, "step": 29551 }, { "epoch": 2.32, "learning_rate": 7.682741315768839e-05, "loss": 2.8543, "step": 29552 }, { "epoch": 2.32, "learning_rate": 7.682662902846389e-05, "loss": 2.9811, "step": 29553 }, { "epoch": 2.32, "learning_rate": 7.68258448992394e-05, "loss": 2.9497, "step": 29554 }, { "epoch": 2.32, "learning_rate": 7.68250607700149e-05, "loss": 2.697, "step": 29555 }, { "epoch": 2.32, "learning_rate": 7.68242766407904e-05, "loss": 2.7765, "step": 29556 }, { "epoch": 2.32, "learning_rate": 7.68234925115659e-05, "loss": 2.7429, "step": 29557 }, { "epoch": 2.32, "learning_rate": 7.682270838234141e-05, "loss": 2.7901, "step": 29558 }, { "epoch": 2.32, "learning_rate": 7.682192425311691e-05, "loss": 2.9396, "step": 29559 }, { "epoch": 2.32, "learning_rate": 7.682114012389242e-05, "loss": 2.7465, "step": 29560 }, { "epoch": 2.32, "learning_rate": 7.682035599466792e-05, "loss": 3.0502, "step": 29561 }, { "epoch": 2.32, "learning_rate": 7.681957186544342e-05, "loss": 2.8709, "step": 29562 }, { "epoch": 2.32, "learning_rate": 7.681878773621893e-05, "loss": 2.9884, "step": 29563 }, { "epoch": 2.32, "learning_rate": 7.681800360699443e-05, "loss": 2.6962, "step": 29564 }, { "epoch": 2.32, "learning_rate": 7.681721947776994e-05, "loss": 2.9206, "step": 29565 }, { "epoch": 2.32, "learning_rate": 7.681643534854544e-05, "loss": 2.6717, "step": 29566 }, { "epoch": 2.32, "learning_rate": 7.681565121932094e-05, "loss": 2.8192, "step": 29567 }, { "epoch": 2.32, "learning_rate": 7.681486709009646e-05, "loss": 2.8584, "step": 29568 }, { "epoch": 2.32, "learning_rate": 7.681408296087196e-05, "loss": 2.8182, "step": 29569 }, { "epoch": 2.32, "learning_rate": 7.681329883164747e-05, "loss": 2.7608, "step": 29570 }, { "epoch": 2.32, "learning_rate": 7.681251470242297e-05, "loss": 2.6405, "step": 29571 }, { "epoch": 2.32, "learning_rate": 7.681173057319848e-05, "loss": 2.85, "step": 29572 }, { "epoch": 2.32, "learning_rate": 7.681094644397398e-05, "loss": 2.9167, "step": 29573 }, { "epoch": 2.32, "learning_rate": 7.681016231474948e-05, "loss": 2.8104, "step": 29574 }, { "epoch": 2.32, "learning_rate": 7.680937818552499e-05, "loss": 2.731, "step": 29575 }, { "epoch": 2.32, "learning_rate": 7.680859405630049e-05, "loss": 2.9533, "step": 29576 }, { "epoch": 2.32, "learning_rate": 7.6807809927076e-05, "loss": 3.0001, "step": 29577 }, { "epoch": 2.32, "learning_rate": 7.68070257978515e-05, "loss": 3.0188, "step": 29578 }, { "epoch": 2.32, "learning_rate": 7.680624166862699e-05, "loss": 2.7174, "step": 29579 }, { "epoch": 2.32, "learning_rate": 7.680545753940249e-05, "loss": 2.6943, "step": 29580 }, { "epoch": 2.32, "learning_rate": 7.6804673410178e-05, "loss": 2.8133, "step": 29581 }, { "epoch": 2.32, "learning_rate": 7.68038892809535e-05, "loss": 2.4638, "step": 29582 }, { "epoch": 2.32, "learning_rate": 7.6803105151729e-05, "loss": 2.7992, "step": 29583 }, { "epoch": 2.32, "learning_rate": 7.68023210225045e-05, "loss": 2.7975, "step": 29584 }, { "epoch": 2.32, "learning_rate": 7.680153689328001e-05, "loss": 2.6938, "step": 29585 }, { "epoch": 2.32, "learning_rate": 7.680075276405551e-05, "loss": 2.7871, "step": 29586 }, { "epoch": 2.32, "learning_rate": 7.679996863483102e-05, "loss": 2.9078, "step": 29587 }, { "epoch": 2.32, "learning_rate": 7.679918450560652e-05, "loss": 2.9564, "step": 29588 }, { "epoch": 2.32, "learning_rate": 7.679840037638202e-05, "loss": 2.7923, "step": 29589 }, { "epoch": 2.32, "learning_rate": 7.679761624715753e-05, "loss": 2.7677, "step": 29590 }, { "epoch": 2.32, "learning_rate": 7.679683211793303e-05, "loss": 2.6407, "step": 29591 }, { "epoch": 2.32, "learning_rate": 7.679604798870854e-05, "loss": 2.5881, "step": 29592 }, { "epoch": 2.32, "learning_rate": 7.679526385948405e-05, "loss": 2.6814, "step": 29593 }, { "epoch": 2.32, "learning_rate": 7.679447973025956e-05, "loss": 2.7634, "step": 29594 }, { "epoch": 2.32, "learning_rate": 7.679369560103506e-05, "loss": 2.7051, "step": 29595 }, { "epoch": 2.32, "learning_rate": 7.679291147181056e-05, "loss": 2.6811, "step": 29596 }, { "epoch": 2.32, "learning_rate": 7.679212734258607e-05, "loss": 2.6885, "step": 29597 }, { "epoch": 2.32, "learning_rate": 7.679134321336157e-05, "loss": 2.6262, "step": 29598 }, { "epoch": 2.32, "learning_rate": 7.679055908413707e-05, "loss": 2.6411, "step": 29599 }, { "epoch": 2.32, "learning_rate": 7.678977495491258e-05, "loss": 2.6185, "step": 29600 }, { "epoch": 2.32, "learning_rate": 7.678899082568808e-05, "loss": 2.7515, "step": 29601 }, { "epoch": 2.32, "learning_rate": 7.678820669646359e-05, "loss": 2.7538, "step": 29602 }, { "epoch": 2.32, "learning_rate": 7.678742256723909e-05, "loss": 2.5256, "step": 29603 }, { "epoch": 2.32, "learning_rate": 7.678663843801459e-05, "loss": 2.829, "step": 29604 }, { "epoch": 2.32, "learning_rate": 7.67858543087901e-05, "loss": 2.6225, "step": 29605 }, { "epoch": 2.32, "learning_rate": 7.67850701795656e-05, "loss": 2.8742, "step": 29606 }, { "epoch": 2.32, "learning_rate": 7.67842860503411e-05, "loss": 2.7489, "step": 29607 }, { "epoch": 2.32, "learning_rate": 7.678350192111661e-05, "loss": 2.8141, "step": 29608 }, { "epoch": 2.32, "learning_rate": 7.678271779189211e-05, "loss": 2.6516, "step": 29609 }, { "epoch": 2.32, "learning_rate": 7.678193366266761e-05, "loss": 2.7399, "step": 29610 }, { "epoch": 2.32, "learning_rate": 7.678114953344312e-05, "loss": 2.7733, "step": 29611 }, { "epoch": 2.32, "learning_rate": 7.678036540421862e-05, "loss": 2.8359, "step": 29612 }, { "epoch": 2.32, "learning_rate": 7.677958127499413e-05, "loss": 2.8741, "step": 29613 }, { "epoch": 2.32, "learning_rate": 7.677879714576963e-05, "loss": 2.9903, "step": 29614 }, { "epoch": 2.32, "learning_rate": 7.677801301654512e-05, "loss": 2.9004, "step": 29615 }, { "epoch": 2.32, "learning_rate": 7.677722888732062e-05, "loss": 2.8223, "step": 29616 }, { "epoch": 2.32, "learning_rate": 7.677644475809613e-05, "loss": 2.9968, "step": 29617 }, { "epoch": 2.32, "learning_rate": 7.677566062887163e-05, "loss": 3.0624, "step": 29618 }, { "epoch": 2.32, "learning_rate": 7.677487649964715e-05, "loss": 2.9205, "step": 29619 }, { "epoch": 2.32, "learning_rate": 7.677409237042265e-05, "loss": 2.8413, "step": 29620 }, { "epoch": 2.32, "learning_rate": 7.677330824119816e-05, "loss": 2.7988, "step": 29621 }, { "epoch": 2.32, "learning_rate": 7.677252411197366e-05, "loss": 3.1388, "step": 29622 }, { "epoch": 2.32, "learning_rate": 7.677173998274916e-05, "loss": 2.8371, "step": 29623 }, { "epoch": 2.32, "learning_rate": 7.677095585352467e-05, "loss": 2.7651, "step": 29624 }, { "epoch": 2.32, "learning_rate": 7.677017172430017e-05, "loss": 2.6693, "step": 29625 }, { "epoch": 2.32, "learning_rate": 7.676938759507567e-05, "loss": 2.8922, "step": 29626 }, { "epoch": 2.32, "learning_rate": 7.676860346585118e-05, "loss": 3.0076, "step": 29627 }, { "epoch": 2.32, "learning_rate": 7.676781933662668e-05, "loss": 3.0986, "step": 29628 }, { "epoch": 2.32, "learning_rate": 7.676703520740218e-05, "loss": 2.7401, "step": 29629 }, { "epoch": 2.32, "learning_rate": 7.676625107817769e-05, "loss": 2.8313, "step": 29630 }, { "epoch": 2.32, "learning_rate": 7.676546694895319e-05, "loss": 2.9325, "step": 29631 }, { "epoch": 2.32, "learning_rate": 7.67646828197287e-05, "loss": 2.6857, "step": 29632 }, { "epoch": 2.32, "learning_rate": 7.67638986905042e-05, "loss": 2.8263, "step": 29633 }, { "epoch": 2.32, "learning_rate": 7.67631145612797e-05, "loss": 2.9565, "step": 29634 }, { "epoch": 2.32, "learning_rate": 7.67623304320552e-05, "loss": 2.6404, "step": 29635 }, { "epoch": 2.32, "learning_rate": 7.676154630283071e-05, "loss": 2.6669, "step": 29636 }, { "epoch": 2.32, "learning_rate": 7.676076217360621e-05, "loss": 2.7666, "step": 29637 }, { "epoch": 2.32, "learning_rate": 7.675997804438172e-05, "loss": 2.803, "step": 29638 }, { "epoch": 2.32, "learning_rate": 7.675919391515722e-05, "loss": 2.7127, "step": 29639 }, { "epoch": 2.32, "learning_rate": 7.675840978593273e-05, "loss": 2.9653, "step": 29640 }, { "epoch": 2.32, "learning_rate": 7.675762565670823e-05, "loss": 2.7666, "step": 29641 }, { "epoch": 2.32, "learning_rate": 7.675684152748373e-05, "loss": 2.8551, "step": 29642 }, { "epoch": 2.32, "learning_rate": 7.675605739825924e-05, "loss": 2.7424, "step": 29643 }, { "epoch": 2.32, "learning_rate": 7.675527326903474e-05, "loss": 2.7474, "step": 29644 }, { "epoch": 2.32, "learning_rate": 7.675448913981024e-05, "loss": 2.7738, "step": 29645 }, { "epoch": 2.32, "learning_rate": 7.675370501058575e-05, "loss": 2.6675, "step": 29646 }, { "epoch": 2.32, "learning_rate": 7.675292088136125e-05, "loss": 2.8133, "step": 29647 }, { "epoch": 2.32, "learning_rate": 7.675213675213675e-05, "loss": 2.7515, "step": 29648 }, { "epoch": 2.32, "learning_rate": 7.675135262291226e-05, "loss": 2.8956, "step": 29649 }, { "epoch": 2.32, "learning_rate": 7.675056849368778e-05, "loss": 2.8527, "step": 29650 }, { "epoch": 2.33, "learning_rate": 7.674978436446327e-05, "loss": 2.9477, "step": 29651 }, { "epoch": 2.33, "learning_rate": 7.674900023523877e-05, "loss": 2.7731, "step": 29652 }, { "epoch": 2.33, "learning_rate": 7.674821610601427e-05, "loss": 2.8488, "step": 29653 }, { "epoch": 2.33, "learning_rate": 7.674743197678978e-05, "loss": 2.8291, "step": 29654 }, { "epoch": 2.33, "learning_rate": 7.674664784756528e-05, "loss": 2.7106, "step": 29655 }, { "epoch": 2.33, "learning_rate": 7.674586371834078e-05, "loss": 2.8952, "step": 29656 }, { "epoch": 2.33, "learning_rate": 7.674507958911629e-05, "loss": 2.7577, "step": 29657 }, { "epoch": 2.33, "learning_rate": 7.674429545989179e-05, "loss": 2.6459, "step": 29658 }, { "epoch": 2.33, "learning_rate": 7.67435113306673e-05, "loss": 2.7232, "step": 29659 }, { "epoch": 2.33, "learning_rate": 7.67427272014428e-05, "loss": 2.6808, "step": 29660 }, { "epoch": 2.33, "learning_rate": 7.67419430722183e-05, "loss": 2.8374, "step": 29661 }, { "epoch": 2.33, "learning_rate": 7.67411589429938e-05, "loss": 2.7029, "step": 29662 }, { "epoch": 2.33, "learning_rate": 7.674037481376931e-05, "loss": 2.8397, "step": 29663 }, { "epoch": 2.33, "learning_rate": 7.673959068454481e-05, "loss": 2.9489, "step": 29664 }, { "epoch": 2.33, "learning_rate": 7.673880655532032e-05, "loss": 2.7471, "step": 29665 }, { "epoch": 2.33, "learning_rate": 7.673802242609582e-05, "loss": 2.7587, "step": 29666 }, { "epoch": 2.33, "learning_rate": 7.673723829687132e-05, "loss": 2.7838, "step": 29667 }, { "epoch": 2.33, "learning_rate": 7.673645416764683e-05, "loss": 2.8998, "step": 29668 }, { "epoch": 2.33, "learning_rate": 7.673567003842233e-05, "loss": 2.6074, "step": 29669 }, { "epoch": 2.33, "learning_rate": 7.673488590919784e-05, "loss": 2.706, "step": 29670 }, { "epoch": 2.33, "learning_rate": 7.673410177997334e-05, "loss": 3.0053, "step": 29671 }, { "epoch": 2.33, "learning_rate": 7.673331765074884e-05, "loss": 2.8709, "step": 29672 }, { "epoch": 2.33, "learning_rate": 7.673253352152435e-05, "loss": 2.8293, "step": 29673 }, { "epoch": 2.33, "learning_rate": 7.673174939229985e-05, "loss": 2.7449, "step": 29674 }, { "epoch": 2.33, "learning_rate": 7.673096526307535e-05, "loss": 2.9862, "step": 29675 }, { "epoch": 2.33, "learning_rate": 7.673018113385087e-05, "loss": 2.7522, "step": 29676 }, { "epoch": 2.33, "learning_rate": 7.672939700462637e-05, "loss": 2.7949, "step": 29677 }, { "epoch": 2.33, "learning_rate": 7.672861287540188e-05, "loss": 2.8203, "step": 29678 }, { "epoch": 2.33, "learning_rate": 7.672782874617738e-05, "loss": 2.9259, "step": 29679 }, { "epoch": 2.33, "learning_rate": 7.672704461695289e-05, "loss": 2.8451, "step": 29680 }, { "epoch": 2.33, "learning_rate": 7.672626048772839e-05, "loss": 3.0235, "step": 29681 }, { "epoch": 2.33, "learning_rate": 7.672547635850389e-05, "loss": 2.7591, "step": 29682 }, { "epoch": 2.33, "learning_rate": 7.67246922292794e-05, "loss": 2.756, "step": 29683 }, { "epoch": 2.33, "learning_rate": 7.67239081000549e-05, "loss": 2.9611, "step": 29684 }, { "epoch": 2.33, "learning_rate": 7.67231239708304e-05, "loss": 2.793, "step": 29685 }, { "epoch": 2.33, "learning_rate": 7.672233984160591e-05, "loss": 2.7698, "step": 29686 }, { "epoch": 2.33, "learning_rate": 7.67215557123814e-05, "loss": 2.8701, "step": 29687 }, { "epoch": 2.33, "learning_rate": 7.67207715831569e-05, "loss": 2.8456, "step": 29688 }, { "epoch": 2.33, "learning_rate": 7.67199874539324e-05, "loss": 2.8929, "step": 29689 }, { "epoch": 2.33, "learning_rate": 7.671920332470791e-05, "loss": 2.7241, "step": 29690 }, { "epoch": 2.33, "learning_rate": 7.671841919548341e-05, "loss": 2.8729, "step": 29691 }, { "epoch": 2.33, "learning_rate": 7.671763506625892e-05, "loss": 2.6947, "step": 29692 }, { "epoch": 2.33, "learning_rate": 7.671685093703442e-05, "loss": 2.5842, "step": 29693 }, { "epoch": 2.33, "learning_rate": 7.671606680780992e-05, "loss": 2.9082, "step": 29694 }, { "epoch": 2.33, "learning_rate": 7.671528267858543e-05, "loss": 3.1179, "step": 29695 }, { "epoch": 2.33, "learning_rate": 7.671449854936093e-05, "loss": 2.9162, "step": 29696 }, { "epoch": 2.33, "learning_rate": 7.671371442013643e-05, "loss": 2.8268, "step": 29697 }, { "epoch": 2.33, "learning_rate": 7.671293029091194e-05, "loss": 2.8067, "step": 29698 }, { "epoch": 2.33, "learning_rate": 7.671214616168744e-05, "loss": 2.7046, "step": 29699 }, { "epoch": 2.33, "learning_rate": 7.671136203246295e-05, "loss": 2.9686, "step": 29700 }, { "epoch": 2.33, "learning_rate": 7.671057790323846e-05, "loss": 2.9342, "step": 29701 }, { "epoch": 2.33, "learning_rate": 7.670979377401397e-05, "loss": 2.6384, "step": 29702 }, { "epoch": 2.33, "learning_rate": 7.670900964478947e-05, "loss": 2.8716, "step": 29703 }, { "epoch": 2.33, "learning_rate": 7.670822551556497e-05, "loss": 3.0234, "step": 29704 }, { "epoch": 2.33, "learning_rate": 7.670744138634048e-05, "loss": 2.6375, "step": 29705 }, { "epoch": 2.33, "learning_rate": 7.670665725711598e-05, "loss": 2.8253, "step": 29706 }, { "epoch": 2.33, "learning_rate": 7.670587312789149e-05, "loss": 2.8801, "step": 29707 }, { "epoch": 2.33, "learning_rate": 7.670508899866699e-05, "loss": 2.8272, "step": 29708 }, { "epoch": 2.33, "learning_rate": 7.670430486944249e-05, "loss": 2.673, "step": 29709 }, { "epoch": 2.33, "learning_rate": 7.6703520740218e-05, "loss": 3.0978, "step": 29710 }, { "epoch": 2.33, "learning_rate": 7.67027366109935e-05, "loss": 2.6993, "step": 29711 }, { "epoch": 2.33, "learning_rate": 7.6701952481769e-05, "loss": 2.8113, "step": 29712 }, { "epoch": 2.33, "learning_rate": 7.670116835254451e-05, "loss": 2.9367, "step": 29713 }, { "epoch": 2.33, "learning_rate": 7.670038422332001e-05, "loss": 2.74, "step": 29714 }, { "epoch": 2.33, "learning_rate": 7.669960009409551e-05, "loss": 2.8534, "step": 29715 }, { "epoch": 2.33, "learning_rate": 7.669881596487102e-05, "loss": 2.7766, "step": 29716 }, { "epoch": 2.33, "learning_rate": 7.669803183564652e-05, "loss": 2.8904, "step": 29717 }, { "epoch": 2.33, "learning_rate": 7.669724770642203e-05, "loss": 2.9582, "step": 29718 }, { "epoch": 2.33, "learning_rate": 7.669646357719753e-05, "loss": 2.7573, "step": 29719 }, { "epoch": 2.33, "learning_rate": 7.669567944797303e-05, "loss": 2.8299, "step": 29720 }, { "epoch": 2.33, "learning_rate": 7.669489531874854e-05, "loss": 2.831, "step": 29721 }, { "epoch": 2.33, "learning_rate": 7.669411118952404e-05, "loss": 2.7465, "step": 29722 }, { "epoch": 2.33, "learning_rate": 7.669332706029953e-05, "loss": 2.6939, "step": 29723 }, { "epoch": 2.33, "learning_rate": 7.669254293107503e-05, "loss": 2.7971, "step": 29724 }, { "epoch": 2.33, "learning_rate": 7.669175880185054e-05, "loss": 2.7218, "step": 29725 }, { "epoch": 2.33, "learning_rate": 7.669097467262604e-05, "loss": 2.6157, "step": 29726 }, { "epoch": 2.33, "learning_rate": 7.669019054340156e-05, "loss": 2.6816, "step": 29727 }, { "epoch": 2.33, "learning_rate": 7.668940641417706e-05, "loss": 2.6479, "step": 29728 }, { "epoch": 2.33, "learning_rate": 7.668862228495257e-05, "loss": 2.6711, "step": 29729 }, { "epoch": 2.33, "learning_rate": 7.668783815572807e-05, "loss": 2.6884, "step": 29730 }, { "epoch": 2.33, "learning_rate": 7.668705402650357e-05, "loss": 2.9497, "step": 29731 }, { "epoch": 2.33, "learning_rate": 7.668626989727908e-05, "loss": 2.6657, "step": 29732 }, { "epoch": 2.33, "learning_rate": 7.668548576805458e-05, "loss": 2.7082, "step": 29733 }, { "epoch": 2.33, "learning_rate": 7.668470163883008e-05, "loss": 2.7877, "step": 29734 }, { "epoch": 2.33, "learning_rate": 7.668391750960559e-05, "loss": 2.7953, "step": 29735 }, { "epoch": 2.33, "learning_rate": 7.668313338038109e-05, "loss": 2.6323, "step": 29736 }, { "epoch": 2.33, "learning_rate": 7.66823492511566e-05, "loss": 2.727, "step": 29737 }, { "epoch": 2.33, "learning_rate": 7.66815651219321e-05, "loss": 3.1798, "step": 29738 }, { "epoch": 2.33, "learning_rate": 7.66807809927076e-05, "loss": 2.8031, "step": 29739 }, { "epoch": 2.33, "learning_rate": 7.66799968634831e-05, "loss": 2.8857, "step": 29740 }, { "epoch": 2.33, "learning_rate": 7.667921273425861e-05, "loss": 2.859, "step": 29741 }, { "epoch": 2.33, "learning_rate": 7.667842860503411e-05, "loss": 2.6349, "step": 29742 }, { "epoch": 2.33, "learning_rate": 7.667764447580962e-05, "loss": 2.7221, "step": 29743 }, { "epoch": 2.33, "learning_rate": 7.667686034658512e-05, "loss": 2.6497, "step": 29744 }, { "epoch": 2.33, "learning_rate": 7.667607621736062e-05, "loss": 2.8245, "step": 29745 }, { "epoch": 2.33, "learning_rate": 7.667529208813613e-05, "loss": 2.7348, "step": 29746 }, { "epoch": 2.33, "learning_rate": 7.667450795891163e-05, "loss": 2.7247, "step": 29747 }, { "epoch": 2.33, "learning_rate": 7.667372382968714e-05, "loss": 2.8994, "step": 29748 }, { "epoch": 2.33, "learning_rate": 7.667293970046264e-05, "loss": 2.7729, "step": 29749 }, { "epoch": 2.33, "learning_rate": 7.667215557123814e-05, "loss": 2.7735, "step": 29750 }, { "epoch": 2.33, "learning_rate": 7.667137144201365e-05, "loss": 2.8959, "step": 29751 }, { "epoch": 2.33, "learning_rate": 7.667058731278915e-05, "loss": 2.9215, "step": 29752 }, { "epoch": 2.33, "learning_rate": 7.666980318356465e-05, "loss": 2.54, "step": 29753 }, { "epoch": 2.33, "learning_rate": 7.666901905434016e-05, "loss": 2.7874, "step": 29754 }, { "epoch": 2.33, "learning_rate": 7.666823492511566e-05, "loss": 2.7262, "step": 29755 }, { "epoch": 2.33, "learning_rate": 7.666745079589117e-05, "loss": 2.8881, "step": 29756 }, { "epoch": 2.33, "learning_rate": 7.666666666666667e-05, "loss": 2.8621, "step": 29757 }, { "epoch": 2.33, "learning_rate": 7.666588253744219e-05, "loss": 2.7168, "step": 29758 }, { "epoch": 2.33, "learning_rate": 7.666509840821768e-05, "loss": 2.7844, "step": 29759 }, { "epoch": 2.33, "learning_rate": 7.666431427899318e-05, "loss": 2.8307, "step": 29760 }, { "epoch": 2.33, "learning_rate": 7.666353014976868e-05, "loss": 2.9217, "step": 29761 }, { "epoch": 2.33, "learning_rate": 7.666274602054419e-05, "loss": 2.6484, "step": 29762 }, { "epoch": 2.33, "learning_rate": 7.666196189131969e-05, "loss": 2.8471, "step": 29763 }, { "epoch": 2.33, "learning_rate": 7.66611777620952e-05, "loss": 2.7767, "step": 29764 }, { "epoch": 2.33, "learning_rate": 7.66603936328707e-05, "loss": 2.7415, "step": 29765 }, { "epoch": 2.33, "learning_rate": 7.66596095036462e-05, "loss": 2.9083, "step": 29766 }, { "epoch": 2.33, "learning_rate": 7.66588253744217e-05, "loss": 2.8444, "step": 29767 }, { "epoch": 2.33, "learning_rate": 7.665804124519721e-05, "loss": 3.0157, "step": 29768 }, { "epoch": 2.33, "learning_rate": 7.665725711597271e-05, "loss": 3.0259, "step": 29769 }, { "epoch": 2.33, "learning_rate": 7.665647298674822e-05, "loss": 2.9728, "step": 29770 }, { "epoch": 2.33, "learning_rate": 7.665568885752372e-05, "loss": 3.081, "step": 29771 }, { "epoch": 2.33, "learning_rate": 7.665490472829922e-05, "loss": 2.8761, "step": 29772 }, { "epoch": 2.33, "learning_rate": 7.665412059907473e-05, "loss": 2.8473, "step": 29773 }, { "epoch": 2.33, "learning_rate": 7.665333646985023e-05, "loss": 2.7344, "step": 29774 }, { "epoch": 2.33, "learning_rate": 7.665255234062574e-05, "loss": 2.7704, "step": 29775 }, { "epoch": 2.33, "learning_rate": 7.665176821140124e-05, "loss": 2.9363, "step": 29776 }, { "epoch": 2.33, "learning_rate": 7.665098408217674e-05, "loss": 3.1774, "step": 29777 }, { "epoch": 2.33, "learning_rate": 7.665019995295225e-05, "loss": 2.706, "step": 29778 }, { "epoch": 2.34, "learning_rate": 7.664941582372775e-05, "loss": 3.0288, "step": 29779 }, { "epoch": 2.34, "learning_rate": 7.664863169450325e-05, "loss": 2.7394, "step": 29780 }, { "epoch": 2.34, "learning_rate": 7.664784756527876e-05, "loss": 2.8967, "step": 29781 }, { "epoch": 2.34, "learning_rate": 7.664706343605426e-05, "loss": 2.9204, "step": 29782 }, { "epoch": 2.34, "learning_rate": 7.664627930682976e-05, "loss": 2.8014, "step": 29783 }, { "epoch": 2.34, "learning_rate": 7.664549517760528e-05, "loss": 2.8467, "step": 29784 }, { "epoch": 2.34, "learning_rate": 7.664471104838079e-05, "loss": 2.585, "step": 29785 }, { "epoch": 2.34, "learning_rate": 7.664392691915629e-05, "loss": 2.6655, "step": 29786 }, { "epoch": 2.34, "learning_rate": 7.664314278993179e-05, "loss": 2.5493, "step": 29787 }, { "epoch": 2.34, "learning_rate": 7.66423586607073e-05, "loss": 2.7695, "step": 29788 }, { "epoch": 2.34, "learning_rate": 7.66415745314828e-05, "loss": 2.7733, "step": 29789 }, { "epoch": 2.34, "learning_rate": 7.66407904022583e-05, "loss": 2.7922, "step": 29790 }, { "epoch": 2.34, "learning_rate": 7.664000627303381e-05, "loss": 2.7809, "step": 29791 }, { "epoch": 2.34, "learning_rate": 7.663922214380931e-05, "loss": 2.5407, "step": 29792 }, { "epoch": 2.34, "learning_rate": 7.663843801458481e-05, "loss": 2.7384, "step": 29793 }, { "epoch": 2.34, "learning_rate": 7.663765388536032e-05, "loss": 2.7568, "step": 29794 }, { "epoch": 2.34, "learning_rate": 7.663686975613581e-05, "loss": 2.705, "step": 29795 }, { "epoch": 2.34, "learning_rate": 7.663608562691131e-05, "loss": 2.7158, "step": 29796 }, { "epoch": 2.34, "learning_rate": 7.663530149768682e-05, "loss": 2.7521, "step": 29797 }, { "epoch": 2.34, "learning_rate": 7.663451736846232e-05, "loss": 2.7745, "step": 29798 }, { "epoch": 2.34, "learning_rate": 7.663373323923782e-05, "loss": 2.655, "step": 29799 }, { "epoch": 2.34, "learning_rate": 7.663294911001333e-05, "loss": 2.8033, "step": 29800 }, { "epoch": 2.34, "learning_rate": 7.663216498078883e-05, "loss": 2.9544, "step": 29801 }, { "epoch": 2.34, "learning_rate": 7.663138085156433e-05, "loss": 2.6905, "step": 29802 }, { "epoch": 2.34, "learning_rate": 7.663059672233984e-05, "loss": 2.9505, "step": 29803 }, { "epoch": 2.34, "learning_rate": 7.662981259311534e-05, "loss": 2.6899, "step": 29804 }, { "epoch": 2.34, "learning_rate": 7.662902846389085e-05, "loss": 2.8878, "step": 29805 }, { "epoch": 2.34, "learning_rate": 7.662824433466635e-05, "loss": 2.7917, "step": 29806 }, { "epoch": 2.34, "learning_rate": 7.662746020544185e-05, "loss": 2.9573, "step": 29807 }, { "epoch": 2.34, "learning_rate": 7.662667607621736e-05, "loss": 3.1077, "step": 29808 }, { "epoch": 2.34, "learning_rate": 7.662589194699287e-05, "loss": 2.7231, "step": 29809 }, { "epoch": 2.34, "learning_rate": 7.662510781776838e-05, "loss": 2.8452, "step": 29810 }, { "epoch": 2.34, "learning_rate": 7.662432368854388e-05, "loss": 2.7357, "step": 29811 }, { "epoch": 2.34, "learning_rate": 7.662353955931938e-05, "loss": 2.9245, "step": 29812 }, { "epoch": 2.34, "learning_rate": 7.662275543009489e-05, "loss": 2.8791, "step": 29813 }, { "epoch": 2.34, "learning_rate": 7.662197130087039e-05, "loss": 2.779, "step": 29814 }, { "epoch": 2.34, "learning_rate": 7.66211871716459e-05, "loss": 2.6859, "step": 29815 }, { "epoch": 2.34, "learning_rate": 7.66204030424214e-05, "loss": 2.7365, "step": 29816 }, { "epoch": 2.34, "learning_rate": 7.66196189131969e-05, "loss": 2.7835, "step": 29817 }, { "epoch": 2.34, "learning_rate": 7.661883478397241e-05, "loss": 2.9043, "step": 29818 }, { "epoch": 2.34, "learning_rate": 7.661805065474791e-05, "loss": 2.8547, "step": 29819 }, { "epoch": 2.34, "learning_rate": 7.661726652552341e-05, "loss": 2.5484, "step": 29820 }, { "epoch": 2.34, "learning_rate": 7.661648239629892e-05, "loss": 2.942, "step": 29821 }, { "epoch": 2.34, "learning_rate": 7.661569826707442e-05, "loss": 2.7566, "step": 29822 }, { "epoch": 2.34, "learning_rate": 7.661491413784993e-05, "loss": 2.9394, "step": 29823 }, { "epoch": 2.34, "learning_rate": 7.661413000862543e-05, "loss": 2.7155, "step": 29824 }, { "epoch": 2.34, "learning_rate": 7.661334587940093e-05, "loss": 2.7591, "step": 29825 }, { "epoch": 2.34, "learning_rate": 7.661256175017644e-05, "loss": 2.7836, "step": 29826 }, { "epoch": 2.34, "learning_rate": 7.661177762095194e-05, "loss": 3.058, "step": 29827 }, { "epoch": 2.34, "learning_rate": 7.661099349172744e-05, "loss": 2.7327, "step": 29828 }, { "epoch": 2.34, "learning_rate": 7.661020936250295e-05, "loss": 2.8807, "step": 29829 }, { "epoch": 2.34, "learning_rate": 7.660942523327845e-05, "loss": 2.8827, "step": 29830 }, { "epoch": 2.34, "learning_rate": 7.660864110405394e-05, "loss": 2.8452, "step": 29831 }, { "epoch": 2.34, "learning_rate": 7.660785697482944e-05, "loss": 2.9289, "step": 29832 }, { "epoch": 2.34, "learning_rate": 7.660707284560495e-05, "loss": 2.8475, "step": 29833 }, { "epoch": 2.34, "learning_rate": 7.660628871638045e-05, "loss": 2.7091, "step": 29834 }, { "epoch": 2.34, "learning_rate": 7.660550458715597e-05, "loss": 2.6113, "step": 29835 }, { "epoch": 2.34, "learning_rate": 7.660472045793147e-05, "loss": 2.7173, "step": 29836 }, { "epoch": 2.34, "learning_rate": 7.660393632870698e-05, "loss": 2.8166, "step": 29837 }, { "epoch": 2.34, "learning_rate": 7.660315219948248e-05, "loss": 2.7485, "step": 29838 }, { "epoch": 2.34, "learning_rate": 7.660236807025798e-05, "loss": 2.6623, "step": 29839 }, { "epoch": 2.34, "learning_rate": 7.660158394103349e-05, "loss": 3.0998, "step": 29840 }, { "epoch": 2.34, "learning_rate": 7.660079981180899e-05, "loss": 2.6021, "step": 29841 }, { "epoch": 2.34, "learning_rate": 7.66000156825845e-05, "loss": 2.5827, "step": 29842 }, { "epoch": 2.34, "learning_rate": 7.659923155336e-05, "loss": 2.9056, "step": 29843 }, { "epoch": 2.34, "learning_rate": 7.65984474241355e-05, "loss": 2.8073, "step": 29844 }, { "epoch": 2.34, "learning_rate": 7.6597663294911e-05, "loss": 3.0853, "step": 29845 }, { "epoch": 2.34, "learning_rate": 7.659687916568651e-05, "loss": 2.948, "step": 29846 }, { "epoch": 2.34, "learning_rate": 7.659609503646201e-05, "loss": 2.7845, "step": 29847 }, { "epoch": 2.34, "learning_rate": 7.659531090723752e-05, "loss": 2.8501, "step": 29848 }, { "epoch": 2.34, "learning_rate": 7.659452677801302e-05, "loss": 2.9949, "step": 29849 }, { "epoch": 2.34, "learning_rate": 7.659374264878852e-05, "loss": 2.8198, "step": 29850 }, { "epoch": 2.34, "learning_rate": 7.659295851956403e-05, "loss": 2.8583, "step": 29851 }, { "epoch": 2.34, "learning_rate": 7.659217439033953e-05, "loss": 2.6124, "step": 29852 }, { "epoch": 2.34, "learning_rate": 7.659139026111504e-05, "loss": 2.4734, "step": 29853 }, { "epoch": 2.34, "learning_rate": 7.659060613189054e-05, "loss": 2.6309, "step": 29854 }, { "epoch": 2.34, "learning_rate": 7.658982200266604e-05, "loss": 2.8176, "step": 29855 }, { "epoch": 2.34, "learning_rate": 7.658903787344155e-05, "loss": 2.8631, "step": 29856 }, { "epoch": 2.34, "learning_rate": 7.658825374421705e-05, "loss": 2.906, "step": 29857 }, { "epoch": 2.34, "learning_rate": 7.658746961499255e-05, "loss": 2.6407, "step": 29858 }, { "epoch": 2.34, "learning_rate": 7.658668548576806e-05, "loss": 2.8432, "step": 29859 }, { "epoch": 2.34, "learning_rate": 7.658590135654356e-05, "loss": 2.8101, "step": 29860 }, { "epoch": 2.34, "learning_rate": 7.658511722731907e-05, "loss": 2.7155, "step": 29861 }, { "epoch": 2.34, "learning_rate": 7.658433309809457e-05, "loss": 2.8484, "step": 29862 }, { "epoch": 2.34, "learning_rate": 7.658354896887007e-05, "loss": 3.1393, "step": 29863 }, { "epoch": 2.34, "learning_rate": 7.658276483964558e-05, "loss": 2.9814, "step": 29864 }, { "epoch": 2.34, "learning_rate": 7.658198071042108e-05, "loss": 2.7329, "step": 29865 }, { "epoch": 2.34, "learning_rate": 7.65811965811966e-05, "loss": 2.7571, "step": 29866 }, { "epoch": 2.34, "learning_rate": 7.658041245197209e-05, "loss": 2.9578, "step": 29867 }, { "epoch": 2.34, "learning_rate": 7.657962832274759e-05, "loss": 2.7871, "step": 29868 }, { "epoch": 2.34, "learning_rate": 7.65788441935231e-05, "loss": 2.6541, "step": 29869 }, { "epoch": 2.34, "learning_rate": 7.65780600642986e-05, "loss": 2.5822, "step": 29870 }, { "epoch": 2.34, "learning_rate": 7.65772759350741e-05, "loss": 2.8852, "step": 29871 }, { "epoch": 2.34, "learning_rate": 7.65764918058496e-05, "loss": 2.7096, "step": 29872 }, { "epoch": 2.34, "learning_rate": 7.657570767662511e-05, "loss": 2.7189, "step": 29873 }, { "epoch": 2.34, "learning_rate": 7.657492354740061e-05, "loss": 2.6856, "step": 29874 }, { "epoch": 2.34, "learning_rate": 7.657413941817612e-05, "loss": 2.8301, "step": 29875 }, { "epoch": 2.34, "learning_rate": 7.657335528895162e-05, "loss": 2.7861, "step": 29876 }, { "epoch": 2.34, "learning_rate": 7.657257115972712e-05, "loss": 2.7858, "step": 29877 }, { "epoch": 2.34, "learning_rate": 7.657178703050263e-05, "loss": 2.7938, "step": 29878 }, { "epoch": 2.34, "learning_rate": 7.657100290127813e-05, "loss": 2.7115, "step": 29879 }, { "epoch": 2.34, "learning_rate": 7.657021877205363e-05, "loss": 2.8781, "step": 29880 }, { "epoch": 2.34, "learning_rate": 7.656943464282914e-05, "loss": 2.8177, "step": 29881 }, { "epoch": 2.34, "learning_rate": 7.656865051360464e-05, "loss": 2.5752, "step": 29882 }, { "epoch": 2.34, "learning_rate": 7.656786638438015e-05, "loss": 2.7144, "step": 29883 }, { "epoch": 2.34, "learning_rate": 7.656708225515565e-05, "loss": 2.6839, "step": 29884 }, { "epoch": 2.34, "learning_rate": 7.656629812593115e-05, "loss": 2.596, "step": 29885 }, { "epoch": 2.34, "learning_rate": 7.656551399670666e-05, "loss": 2.878, "step": 29886 }, { "epoch": 2.34, "learning_rate": 7.656472986748216e-05, "loss": 2.6937, "step": 29887 }, { "epoch": 2.34, "learning_rate": 7.656394573825766e-05, "loss": 2.8723, "step": 29888 }, { "epoch": 2.34, "learning_rate": 7.656316160903317e-05, "loss": 2.7822, "step": 29889 }, { "epoch": 2.34, "learning_rate": 7.656237747980867e-05, "loss": 2.6501, "step": 29890 }, { "epoch": 2.34, "learning_rate": 7.656159335058418e-05, "loss": 2.8643, "step": 29891 }, { "epoch": 2.34, "learning_rate": 7.656080922135969e-05, "loss": 2.7457, "step": 29892 }, { "epoch": 2.34, "learning_rate": 7.65600250921352e-05, "loss": 2.9223, "step": 29893 }, { "epoch": 2.34, "learning_rate": 7.65592409629107e-05, "loss": 2.6961, "step": 29894 }, { "epoch": 2.34, "learning_rate": 7.65584568336862e-05, "loss": 2.8125, "step": 29895 }, { "epoch": 2.34, "learning_rate": 7.655767270446171e-05, "loss": 2.81, "step": 29896 }, { "epoch": 2.34, "learning_rate": 7.655688857523721e-05, "loss": 3.1713, "step": 29897 }, { "epoch": 2.34, "learning_rate": 7.655610444601271e-05, "loss": 2.7499, "step": 29898 }, { "epoch": 2.34, "learning_rate": 7.655532031678822e-05, "loss": 2.9691, "step": 29899 }, { "epoch": 2.34, "learning_rate": 7.655453618756372e-05, "loss": 2.6374, "step": 29900 }, { "epoch": 2.34, "learning_rate": 7.655375205833923e-05, "loss": 2.9831, "step": 29901 }, { "epoch": 2.34, "learning_rate": 7.655296792911473e-05, "loss": 2.6469, "step": 29902 }, { "epoch": 2.34, "learning_rate": 7.655218379989022e-05, "loss": 2.5584, "step": 29903 }, { "epoch": 2.34, "learning_rate": 7.655139967066572e-05, "loss": 2.942, "step": 29904 }, { "epoch": 2.34, "learning_rate": 7.655061554144123e-05, "loss": 2.6188, "step": 29905 }, { "epoch": 2.35, "learning_rate": 7.654983141221673e-05, "loss": 2.713, "step": 29906 }, { "epoch": 2.35, "learning_rate": 7.654904728299223e-05, "loss": 2.8883, "step": 29907 }, { "epoch": 2.35, "learning_rate": 7.654826315376774e-05, "loss": 2.7157, "step": 29908 }, { "epoch": 2.35, "learning_rate": 7.654747902454324e-05, "loss": 2.6289, "step": 29909 }, { "epoch": 2.35, "learning_rate": 7.654669489531875e-05, "loss": 2.8255, "step": 29910 }, { "epoch": 2.35, "learning_rate": 7.654591076609425e-05, "loss": 2.9147, "step": 29911 }, { "epoch": 2.35, "learning_rate": 7.654512663686975e-05, "loss": 3.014, "step": 29912 }, { "epoch": 2.35, "learning_rate": 7.654434250764526e-05, "loss": 2.566, "step": 29913 }, { "epoch": 2.35, "learning_rate": 7.654355837842076e-05, "loss": 2.837, "step": 29914 }, { "epoch": 2.35, "learning_rate": 7.654277424919626e-05, "loss": 2.8423, "step": 29915 }, { "epoch": 2.35, "learning_rate": 7.654199011997177e-05, "loss": 2.634, "step": 29916 }, { "epoch": 2.35, "learning_rate": 7.654120599074728e-05, "loss": 2.7689, "step": 29917 }, { "epoch": 2.35, "learning_rate": 7.654042186152279e-05, "loss": 2.7176, "step": 29918 }, { "epoch": 2.35, "learning_rate": 7.653963773229829e-05, "loss": 2.677, "step": 29919 }, { "epoch": 2.35, "learning_rate": 7.65388536030738e-05, "loss": 2.7749, "step": 29920 }, { "epoch": 2.35, "learning_rate": 7.65380694738493e-05, "loss": 2.8045, "step": 29921 }, { "epoch": 2.35, "learning_rate": 7.65372853446248e-05, "loss": 2.9051, "step": 29922 }, { "epoch": 2.35, "learning_rate": 7.65365012154003e-05, "loss": 2.6359, "step": 29923 }, { "epoch": 2.35, "learning_rate": 7.653571708617581e-05, "loss": 2.9137, "step": 29924 }, { "epoch": 2.35, "learning_rate": 7.653493295695131e-05, "loss": 2.77, "step": 29925 }, { "epoch": 2.35, "learning_rate": 7.653414882772682e-05, "loss": 2.6802, "step": 29926 }, { "epoch": 2.35, "learning_rate": 7.653336469850232e-05, "loss": 2.7208, "step": 29927 }, { "epoch": 2.35, "learning_rate": 7.653258056927782e-05, "loss": 2.7478, "step": 29928 }, { "epoch": 2.35, "learning_rate": 7.653179644005333e-05, "loss": 2.829, "step": 29929 }, { "epoch": 2.35, "learning_rate": 7.653101231082883e-05, "loss": 2.7408, "step": 29930 }, { "epoch": 2.35, "learning_rate": 7.653022818160434e-05, "loss": 2.935, "step": 29931 }, { "epoch": 2.35, "learning_rate": 7.652944405237984e-05, "loss": 2.7747, "step": 29932 }, { "epoch": 2.35, "learning_rate": 7.652865992315534e-05, "loss": 2.8985, "step": 29933 }, { "epoch": 2.35, "learning_rate": 7.652787579393085e-05, "loss": 2.6695, "step": 29934 }, { "epoch": 2.35, "learning_rate": 7.652709166470635e-05, "loss": 2.8832, "step": 29935 }, { "epoch": 2.35, "learning_rate": 7.652630753548185e-05, "loss": 2.7908, "step": 29936 }, { "epoch": 2.35, "learning_rate": 7.652552340625736e-05, "loss": 2.8232, "step": 29937 }, { "epoch": 2.35, "learning_rate": 7.652473927703286e-05, "loss": 2.8161, "step": 29938 }, { "epoch": 2.35, "learning_rate": 7.652395514780835e-05, "loss": 2.882, "step": 29939 }, { "epoch": 2.35, "learning_rate": 7.652317101858386e-05, "loss": 2.8769, "step": 29940 }, { "epoch": 2.35, "learning_rate": 7.652238688935936e-05, "loss": 2.7353, "step": 29941 }, { "epoch": 2.35, "learning_rate": 7.652160276013486e-05, "loss": 2.7306, "step": 29942 }, { "epoch": 2.35, "learning_rate": 7.652081863091038e-05, "loss": 2.7111, "step": 29943 }, { "epoch": 2.35, "learning_rate": 7.652003450168588e-05, "loss": 2.5696, "step": 29944 }, { "epoch": 2.35, "learning_rate": 7.651925037246139e-05, "loss": 2.7405, "step": 29945 }, { "epoch": 2.35, "learning_rate": 7.651846624323689e-05, "loss": 2.8862, "step": 29946 }, { "epoch": 2.35, "learning_rate": 7.65176821140124e-05, "loss": 2.8192, "step": 29947 }, { "epoch": 2.35, "learning_rate": 7.65168979847879e-05, "loss": 2.9437, "step": 29948 }, { "epoch": 2.35, "learning_rate": 7.65161138555634e-05, "loss": 2.7973, "step": 29949 }, { "epoch": 2.35, "learning_rate": 7.65153297263389e-05, "loss": 2.8376, "step": 29950 }, { "epoch": 2.35, "learning_rate": 7.651454559711441e-05, "loss": 3.0281, "step": 29951 }, { "epoch": 2.35, "learning_rate": 7.651376146788991e-05, "loss": 2.8372, "step": 29952 }, { "epoch": 2.35, "learning_rate": 7.651297733866542e-05, "loss": 2.7829, "step": 29953 }, { "epoch": 2.35, "learning_rate": 7.651219320944092e-05, "loss": 2.8457, "step": 29954 }, { "epoch": 2.35, "learning_rate": 7.651140908021642e-05, "loss": 2.6954, "step": 29955 }, { "epoch": 2.35, "learning_rate": 7.651062495099193e-05, "loss": 2.9304, "step": 29956 }, { "epoch": 2.35, "learning_rate": 7.650984082176743e-05, "loss": 2.6246, "step": 29957 }, { "epoch": 2.35, "learning_rate": 7.650905669254294e-05, "loss": 2.8307, "step": 29958 }, { "epoch": 2.35, "learning_rate": 7.650827256331844e-05, "loss": 2.94, "step": 29959 }, { "epoch": 2.35, "learning_rate": 7.650748843409394e-05, "loss": 2.6538, "step": 29960 }, { "epoch": 2.35, "learning_rate": 7.650670430486945e-05, "loss": 2.5549, "step": 29961 }, { "epoch": 2.35, "learning_rate": 7.650592017564495e-05, "loss": 2.8399, "step": 29962 }, { "epoch": 2.35, "learning_rate": 7.650513604642045e-05, "loss": 2.7352, "step": 29963 }, { "epoch": 2.35, "learning_rate": 7.650435191719596e-05, "loss": 2.7994, "step": 29964 }, { "epoch": 2.35, "learning_rate": 7.650356778797146e-05, "loss": 2.7832, "step": 29965 }, { "epoch": 2.35, "learning_rate": 7.650278365874696e-05, "loss": 2.8438, "step": 29966 }, { "epoch": 2.35, "learning_rate": 7.650199952952247e-05, "loss": 2.628, "step": 29967 }, { "epoch": 2.35, "learning_rate": 7.650121540029797e-05, "loss": 2.8037, "step": 29968 }, { "epoch": 2.35, "learning_rate": 7.650043127107348e-05, "loss": 2.7718, "step": 29969 }, { "epoch": 2.35, "learning_rate": 7.649964714184898e-05, "loss": 2.9584, "step": 29970 }, { "epoch": 2.35, "learning_rate": 7.649886301262448e-05, "loss": 2.6, "step": 29971 }, { "epoch": 2.35, "learning_rate": 7.649807888339999e-05, "loss": 2.7874, "step": 29972 }, { "epoch": 2.35, "learning_rate": 7.649729475417549e-05, "loss": 3.0388, "step": 29973 }, { "epoch": 2.35, "learning_rate": 7.649651062495101e-05, "loss": 2.9539, "step": 29974 }, { "epoch": 2.35, "learning_rate": 7.64957264957265e-05, "loss": 2.7671, "step": 29975 }, { "epoch": 2.35, "learning_rate": 7.6494942366502e-05, "loss": 2.6843, "step": 29976 }, { "epoch": 2.35, "learning_rate": 7.64941582372775e-05, "loss": 2.7032, "step": 29977 }, { "epoch": 2.35, "learning_rate": 7.649337410805301e-05, "loss": 2.762, "step": 29978 }, { "epoch": 2.35, "learning_rate": 7.649258997882851e-05, "loss": 2.8524, "step": 29979 }, { "epoch": 2.35, "learning_rate": 7.649180584960402e-05, "loss": 2.9137, "step": 29980 }, { "epoch": 2.35, "learning_rate": 7.649102172037952e-05, "loss": 2.7691, "step": 29981 }, { "epoch": 2.35, "learning_rate": 7.649023759115502e-05, "loss": 2.8293, "step": 29982 }, { "epoch": 2.35, "learning_rate": 7.648945346193053e-05, "loss": 2.9167, "step": 29983 }, { "epoch": 2.35, "learning_rate": 7.648866933270603e-05, "loss": 2.6689, "step": 29984 }, { "epoch": 2.35, "learning_rate": 7.648788520348153e-05, "loss": 2.8222, "step": 29985 }, { "epoch": 2.35, "learning_rate": 7.648710107425704e-05, "loss": 2.6871, "step": 29986 }, { "epoch": 2.35, "learning_rate": 7.648631694503254e-05, "loss": 2.7786, "step": 29987 }, { "epoch": 2.35, "learning_rate": 7.648553281580805e-05, "loss": 2.7776, "step": 29988 }, { "epoch": 2.35, "learning_rate": 7.648474868658355e-05, "loss": 3.0439, "step": 29989 }, { "epoch": 2.35, "learning_rate": 7.648396455735905e-05, "loss": 2.9395, "step": 29990 }, { "epoch": 2.35, "learning_rate": 7.648318042813456e-05, "loss": 2.8777, "step": 29991 }, { "epoch": 2.35, "learning_rate": 7.648239629891006e-05, "loss": 3.1836, "step": 29992 }, { "epoch": 2.35, "learning_rate": 7.648161216968556e-05, "loss": 2.8893, "step": 29993 }, { "epoch": 2.35, "learning_rate": 7.648082804046107e-05, "loss": 2.7019, "step": 29994 }, { "epoch": 2.35, "learning_rate": 7.648004391123657e-05, "loss": 2.7169, "step": 29995 }, { "epoch": 2.35, "learning_rate": 7.647925978201208e-05, "loss": 2.9518, "step": 29996 }, { "epoch": 2.35, "learning_rate": 7.647847565278758e-05, "loss": 2.6519, "step": 29997 }, { "epoch": 2.35, "learning_rate": 7.647769152356308e-05, "loss": 2.8488, "step": 29998 }, { "epoch": 2.35, "learning_rate": 7.647690739433859e-05, "loss": 2.7276, "step": 29999 }, { "epoch": 2.35, "learning_rate": 7.64761232651141e-05, "loss": 2.8603, "step": 30000 }, { "epoch": 2.35, "eval_loss": 2.9008233547210693, "eval_runtime": 218.0802, "eval_samples_per_second": 51.963, "eval_steps_per_second": 3.251, "step": 30000 }, { "epoch": 2.35, "learning_rate": 7.647533913588961e-05, "loss": 2.9169, "step": 30001 }, { "epoch": 2.35, "learning_rate": 7.647455500666511e-05, "loss": 2.7678, "step": 30002 }, { "epoch": 2.35, "learning_rate": 7.647377087744061e-05, "loss": 3.0657, "step": 30003 }, { "epoch": 2.35, "learning_rate": 7.647298674821612e-05, "loss": 2.6745, "step": 30004 }, { "epoch": 2.35, "learning_rate": 7.647220261899162e-05, "loss": 2.6178, "step": 30005 }, { "epoch": 2.35, "learning_rate": 7.647141848976713e-05, "loss": 2.8689, "step": 30006 }, { "epoch": 2.35, "learning_rate": 7.647063436054263e-05, "loss": 2.8531, "step": 30007 }, { "epoch": 2.35, "learning_rate": 7.646985023131813e-05, "loss": 2.6973, "step": 30008 }, { "epoch": 2.35, "learning_rate": 7.646906610209364e-05, "loss": 2.8038, "step": 30009 }, { "epoch": 2.35, "learning_rate": 7.646828197286914e-05, "loss": 2.8641, "step": 30010 }, { "epoch": 2.35, "learning_rate": 7.646749784364463e-05, "loss": 3.0476, "step": 30011 }, { "epoch": 2.35, "learning_rate": 7.646671371442013e-05, "loss": 2.7385, "step": 30012 }, { "epoch": 2.35, "learning_rate": 7.646592958519564e-05, "loss": 2.7932, "step": 30013 }, { "epoch": 2.35, "learning_rate": 7.646514545597114e-05, "loss": 2.6476, "step": 30014 }, { "epoch": 2.35, "learning_rate": 7.646436132674664e-05, "loss": 2.9858, "step": 30015 }, { "epoch": 2.35, "learning_rate": 7.646357719752215e-05, "loss": 2.7506, "step": 30016 }, { "epoch": 2.35, "learning_rate": 7.646279306829765e-05, "loss": 3.0142, "step": 30017 }, { "epoch": 2.35, "learning_rate": 7.646200893907316e-05, "loss": 3.1932, "step": 30018 }, { "epoch": 2.35, "learning_rate": 7.646122480984866e-05, "loss": 2.856, "step": 30019 }, { "epoch": 2.35, "learning_rate": 7.646044068062416e-05, "loss": 2.7888, "step": 30020 }, { "epoch": 2.35, "learning_rate": 7.645965655139967e-05, "loss": 2.923, "step": 30021 }, { "epoch": 2.35, "learning_rate": 7.645887242217517e-05, "loss": 2.8134, "step": 30022 }, { "epoch": 2.35, "learning_rate": 7.645808829295067e-05, "loss": 2.9351, "step": 30023 }, { "epoch": 2.35, "learning_rate": 7.645730416372618e-05, "loss": 2.809, "step": 30024 }, { "epoch": 2.35, "learning_rate": 7.64565200345017e-05, "loss": 2.7742, "step": 30025 }, { "epoch": 2.35, "learning_rate": 7.64557359052772e-05, "loss": 2.7801, "step": 30026 }, { "epoch": 2.35, "learning_rate": 7.64549517760527e-05, "loss": 2.693, "step": 30027 }, { "epoch": 2.35, "learning_rate": 7.64541676468282e-05, "loss": 2.7941, "step": 30028 }, { "epoch": 2.35, "learning_rate": 7.645338351760371e-05, "loss": 2.8192, "step": 30029 }, { "epoch": 2.35, "learning_rate": 7.645259938837921e-05, "loss": 2.6613, "step": 30030 }, { "epoch": 2.35, "learning_rate": 7.645181525915472e-05, "loss": 2.656, "step": 30031 }, { "epoch": 2.35, "learning_rate": 7.645103112993022e-05, "loss": 2.7003, "step": 30032 }, { "epoch": 2.35, "learning_rate": 7.645024700070572e-05, "loss": 2.7629, "step": 30033 }, { "epoch": 2.36, "learning_rate": 7.644946287148123e-05, "loss": 2.9458, "step": 30034 }, { "epoch": 2.36, "learning_rate": 7.644867874225673e-05, "loss": 2.7211, "step": 30035 }, { "epoch": 2.36, "learning_rate": 7.644789461303224e-05, "loss": 2.8837, "step": 30036 }, { "epoch": 2.36, "learning_rate": 7.644711048380774e-05, "loss": 2.7049, "step": 30037 }, { "epoch": 2.36, "learning_rate": 7.644632635458324e-05, "loss": 2.8379, "step": 30038 }, { "epoch": 2.36, "learning_rate": 7.644554222535875e-05, "loss": 2.8503, "step": 30039 }, { "epoch": 2.36, "learning_rate": 7.644475809613425e-05, "loss": 2.8272, "step": 30040 }, { "epoch": 2.36, "learning_rate": 7.644397396690975e-05, "loss": 2.6818, "step": 30041 }, { "epoch": 2.36, "learning_rate": 7.644318983768526e-05, "loss": 2.7624, "step": 30042 }, { "epoch": 2.36, "learning_rate": 7.644240570846076e-05, "loss": 2.8479, "step": 30043 }, { "epoch": 2.36, "learning_rate": 7.644162157923627e-05, "loss": 2.5856, "step": 30044 }, { "epoch": 2.36, "learning_rate": 7.644083745001177e-05, "loss": 3.0091, "step": 30045 }, { "epoch": 2.36, "learning_rate": 7.644005332078727e-05, "loss": 2.8084, "step": 30046 }, { "epoch": 2.36, "learning_rate": 7.643926919156276e-05, "loss": 2.9337, "step": 30047 }, { "epoch": 2.36, "learning_rate": 7.643848506233827e-05, "loss": 2.8789, "step": 30048 }, { "epoch": 2.36, "learning_rate": 7.643770093311377e-05, "loss": 2.7708, "step": 30049 }, { "epoch": 2.36, "learning_rate": 7.643691680388927e-05, "loss": 2.7724, "step": 30050 }, { "epoch": 2.36, "learning_rate": 7.643613267466479e-05, "loss": 2.8555, "step": 30051 }, { "epoch": 2.36, "learning_rate": 7.64353485454403e-05, "loss": 3.0972, "step": 30052 }, { "epoch": 2.36, "learning_rate": 7.64345644162158e-05, "loss": 2.9005, "step": 30053 }, { "epoch": 2.36, "learning_rate": 7.64337802869913e-05, "loss": 2.6842, "step": 30054 }, { "epoch": 2.36, "learning_rate": 7.64329961577668e-05, "loss": 2.9214, "step": 30055 }, { "epoch": 2.36, "learning_rate": 7.643221202854231e-05, "loss": 2.6205, "step": 30056 }, { "epoch": 2.36, "learning_rate": 7.643142789931781e-05, "loss": 3.2124, "step": 30057 }, { "epoch": 2.36, "learning_rate": 7.643064377009332e-05, "loss": 2.7722, "step": 30058 }, { "epoch": 2.36, "learning_rate": 7.642985964086882e-05, "loss": 2.7342, "step": 30059 }, { "epoch": 2.36, "learning_rate": 7.642907551164432e-05, "loss": 2.6042, "step": 30060 }, { "epoch": 2.36, "learning_rate": 7.642829138241983e-05, "loss": 2.7399, "step": 30061 }, { "epoch": 2.36, "learning_rate": 7.642750725319533e-05, "loss": 2.9782, "step": 30062 }, { "epoch": 2.36, "learning_rate": 7.642672312397084e-05, "loss": 2.6232, "step": 30063 }, { "epoch": 2.36, "learning_rate": 7.642593899474634e-05, "loss": 2.7483, "step": 30064 }, { "epoch": 2.36, "learning_rate": 7.642515486552184e-05, "loss": 2.845, "step": 30065 }, { "epoch": 2.36, "learning_rate": 7.642437073629735e-05, "loss": 2.6817, "step": 30066 }, { "epoch": 2.36, "learning_rate": 7.642358660707285e-05, "loss": 3.0762, "step": 30067 }, { "epoch": 2.36, "learning_rate": 7.642280247784835e-05, "loss": 2.8477, "step": 30068 }, { "epoch": 2.36, "learning_rate": 7.642201834862386e-05, "loss": 2.7166, "step": 30069 }, { "epoch": 2.36, "learning_rate": 7.642123421939936e-05, "loss": 2.7411, "step": 30070 }, { "epoch": 2.36, "learning_rate": 7.642045009017486e-05, "loss": 2.9171, "step": 30071 }, { "epoch": 2.36, "learning_rate": 7.641966596095037e-05, "loss": 2.8483, "step": 30072 }, { "epoch": 2.36, "learning_rate": 7.641888183172587e-05, "loss": 2.9258, "step": 30073 }, { "epoch": 2.36, "learning_rate": 7.641809770250138e-05, "loss": 2.7332, "step": 30074 }, { "epoch": 2.36, "learning_rate": 7.641731357327688e-05, "loss": 2.7308, "step": 30075 }, { "epoch": 2.36, "learning_rate": 7.641652944405238e-05, "loss": 2.9883, "step": 30076 }, { "epoch": 2.36, "learning_rate": 7.641574531482789e-05, "loss": 2.6985, "step": 30077 }, { "epoch": 2.36, "learning_rate": 7.641496118560339e-05, "loss": 2.8368, "step": 30078 }, { "epoch": 2.36, "learning_rate": 7.64141770563789e-05, "loss": 2.8306, "step": 30079 }, { "epoch": 2.36, "learning_rate": 7.64133929271544e-05, "loss": 2.7511, "step": 30080 }, { "epoch": 2.36, "learning_rate": 7.64126087979299e-05, "loss": 2.7973, "step": 30081 }, { "epoch": 2.36, "learning_rate": 7.641182466870542e-05, "loss": 2.9364, "step": 30082 }, { "epoch": 2.36, "learning_rate": 7.641104053948091e-05, "loss": 3.112, "step": 30083 }, { "epoch": 2.36, "learning_rate": 7.641025641025641e-05, "loss": 2.9978, "step": 30084 }, { "epoch": 2.36, "learning_rate": 7.640947228103192e-05, "loss": 3.0029, "step": 30085 }, { "epoch": 2.36, "learning_rate": 7.640868815180742e-05, "loss": 2.8841, "step": 30086 }, { "epoch": 2.36, "learning_rate": 7.640790402258292e-05, "loss": 2.6325, "step": 30087 }, { "epoch": 2.36, "learning_rate": 7.640711989335843e-05, "loss": 2.6525, "step": 30088 }, { "epoch": 2.36, "learning_rate": 7.640633576413393e-05, "loss": 2.7458, "step": 30089 }, { "epoch": 2.36, "learning_rate": 7.640555163490943e-05, "loss": 2.7597, "step": 30090 }, { "epoch": 2.36, "learning_rate": 7.640476750568494e-05, "loss": 3.0304, "step": 30091 }, { "epoch": 2.36, "learning_rate": 7.640398337646044e-05, "loss": 2.6874, "step": 30092 }, { "epoch": 2.36, "learning_rate": 7.640319924723595e-05, "loss": 2.9212, "step": 30093 }, { "epoch": 2.36, "learning_rate": 7.640241511801145e-05, "loss": 2.8084, "step": 30094 }, { "epoch": 2.36, "learning_rate": 7.640163098878695e-05, "loss": 2.5947, "step": 30095 }, { "epoch": 2.36, "learning_rate": 7.640084685956246e-05, "loss": 2.8202, "step": 30096 }, { "epoch": 2.36, "learning_rate": 7.640006273033796e-05, "loss": 2.6968, "step": 30097 }, { "epoch": 2.36, "learning_rate": 7.639927860111346e-05, "loss": 2.8117, "step": 30098 }, { "epoch": 2.36, "learning_rate": 7.639849447188897e-05, "loss": 2.6525, "step": 30099 }, { "epoch": 2.36, "learning_rate": 7.639771034266447e-05, "loss": 2.7549, "step": 30100 }, { "epoch": 2.36, "learning_rate": 7.639692621343997e-05, "loss": 2.5922, "step": 30101 }, { "epoch": 2.36, "learning_rate": 7.639614208421548e-05, "loss": 2.8533, "step": 30102 }, { "epoch": 2.36, "learning_rate": 7.639535795499098e-05, "loss": 2.873, "step": 30103 }, { "epoch": 2.36, "learning_rate": 7.639457382576649e-05, "loss": 2.6815, "step": 30104 }, { "epoch": 2.36, "learning_rate": 7.639378969654199e-05, "loss": 2.8045, "step": 30105 }, { "epoch": 2.36, "learning_rate": 7.639300556731749e-05, "loss": 2.8209, "step": 30106 }, { "epoch": 2.36, "learning_rate": 7.6392221438093e-05, "loss": 2.7836, "step": 30107 }, { "epoch": 2.36, "learning_rate": 7.639143730886851e-05, "loss": 2.8617, "step": 30108 }, { "epoch": 2.36, "learning_rate": 7.639065317964402e-05, "loss": 2.6303, "step": 30109 }, { "epoch": 2.36, "learning_rate": 7.638986905041952e-05, "loss": 2.8023, "step": 30110 }, { "epoch": 2.36, "learning_rate": 7.638908492119503e-05, "loss": 3.073, "step": 30111 }, { "epoch": 2.36, "learning_rate": 7.638830079197053e-05, "loss": 2.7236, "step": 30112 }, { "epoch": 2.36, "learning_rate": 7.638751666274603e-05, "loss": 2.7921, "step": 30113 }, { "epoch": 2.36, "learning_rate": 7.638673253352154e-05, "loss": 2.7225, "step": 30114 }, { "epoch": 2.36, "learning_rate": 7.638594840429704e-05, "loss": 2.7446, "step": 30115 }, { "epoch": 2.36, "learning_rate": 7.638516427507254e-05, "loss": 2.9481, "step": 30116 }, { "epoch": 2.36, "learning_rate": 7.638438014584805e-05, "loss": 2.7967, "step": 30117 }, { "epoch": 2.36, "learning_rate": 7.638359601662355e-05, "loss": 2.6979, "step": 30118 }, { "epoch": 2.36, "learning_rate": 7.638281188739904e-05, "loss": 2.6722, "step": 30119 }, { "epoch": 2.36, "learning_rate": 7.638202775817454e-05, "loss": 2.9701, "step": 30120 }, { "epoch": 2.36, "learning_rate": 7.638124362895005e-05, "loss": 2.9781, "step": 30121 }, { "epoch": 2.36, "learning_rate": 7.638045949972555e-05, "loss": 2.8831, "step": 30122 }, { "epoch": 2.36, "learning_rate": 7.637967537050106e-05, "loss": 2.8137, "step": 30123 }, { "epoch": 2.36, "learning_rate": 7.637889124127656e-05, "loss": 2.7381, "step": 30124 }, { "epoch": 2.36, "learning_rate": 7.637810711205206e-05, "loss": 2.8366, "step": 30125 }, { "epoch": 2.36, "learning_rate": 7.637732298282757e-05, "loss": 3.0242, "step": 30126 }, { "epoch": 2.36, "learning_rate": 7.637653885360307e-05, "loss": 2.9347, "step": 30127 }, { "epoch": 2.36, "learning_rate": 7.637575472437857e-05, "loss": 2.6258, "step": 30128 }, { "epoch": 2.36, "learning_rate": 7.637497059515408e-05, "loss": 3.0353, "step": 30129 }, { "epoch": 2.36, "learning_rate": 7.637418646592958e-05, "loss": 2.6177, "step": 30130 }, { "epoch": 2.36, "learning_rate": 7.637340233670509e-05, "loss": 2.967, "step": 30131 }, { "epoch": 2.36, "learning_rate": 7.637261820748059e-05, "loss": 2.5618, "step": 30132 }, { "epoch": 2.36, "learning_rate": 7.63718340782561e-05, "loss": 2.781, "step": 30133 }, { "epoch": 2.36, "learning_rate": 7.637104994903161e-05, "loss": 2.7448, "step": 30134 }, { "epoch": 2.36, "learning_rate": 7.637026581980711e-05, "loss": 3.1897, "step": 30135 }, { "epoch": 2.36, "learning_rate": 7.636948169058262e-05, "loss": 2.779, "step": 30136 }, { "epoch": 2.36, "learning_rate": 7.636869756135812e-05, "loss": 2.769, "step": 30137 }, { "epoch": 2.36, "learning_rate": 7.636791343213362e-05, "loss": 2.7343, "step": 30138 }, { "epoch": 2.36, "learning_rate": 7.636712930290913e-05, "loss": 2.8747, "step": 30139 }, { "epoch": 2.36, "learning_rate": 7.636634517368463e-05, "loss": 2.88, "step": 30140 }, { "epoch": 2.36, "learning_rate": 7.636556104446014e-05, "loss": 2.8585, "step": 30141 }, { "epoch": 2.36, "learning_rate": 7.636477691523564e-05, "loss": 2.9567, "step": 30142 }, { "epoch": 2.36, "learning_rate": 7.636399278601114e-05, "loss": 3.0709, "step": 30143 }, { "epoch": 2.36, "learning_rate": 7.636320865678665e-05, "loss": 2.76, "step": 30144 }, { "epoch": 2.36, "learning_rate": 7.636242452756215e-05, "loss": 2.7107, "step": 30145 }, { "epoch": 2.36, "learning_rate": 7.636164039833765e-05, "loss": 2.7352, "step": 30146 }, { "epoch": 2.36, "learning_rate": 7.636085626911316e-05, "loss": 2.8241, "step": 30147 }, { "epoch": 2.36, "learning_rate": 7.636007213988866e-05, "loss": 2.5421, "step": 30148 }, { "epoch": 2.36, "learning_rate": 7.635928801066416e-05, "loss": 2.8077, "step": 30149 }, { "epoch": 2.36, "learning_rate": 7.635850388143967e-05, "loss": 2.5859, "step": 30150 }, { "epoch": 2.36, "learning_rate": 7.635771975221517e-05, "loss": 2.769, "step": 30151 }, { "epoch": 2.36, "learning_rate": 7.635693562299068e-05, "loss": 2.6209, "step": 30152 }, { "epoch": 2.36, "learning_rate": 7.635615149376618e-05, "loss": 2.8034, "step": 30153 }, { "epoch": 2.36, "learning_rate": 7.635536736454168e-05, "loss": 2.7362, "step": 30154 }, { "epoch": 2.36, "learning_rate": 7.635458323531717e-05, "loss": 2.7018, "step": 30155 }, { "epoch": 2.36, "learning_rate": 7.635379910609268e-05, "loss": 2.7154, "step": 30156 }, { "epoch": 2.36, "learning_rate": 7.635301497686818e-05, "loss": 2.8574, "step": 30157 }, { "epoch": 2.36, "learning_rate": 7.635223084764368e-05, "loss": 2.9188, "step": 30158 }, { "epoch": 2.36, "learning_rate": 7.63514467184192e-05, "loss": 2.881, "step": 30159 }, { "epoch": 2.36, "learning_rate": 7.63506625891947e-05, "loss": 2.8078, "step": 30160 }, { "epoch": 2.37, "learning_rate": 7.634987845997021e-05, "loss": 2.8699, "step": 30161 }, { "epoch": 2.37, "learning_rate": 7.634909433074571e-05, "loss": 2.9349, "step": 30162 }, { "epoch": 2.37, "learning_rate": 7.634831020152122e-05, "loss": 2.7923, "step": 30163 }, { "epoch": 2.37, "learning_rate": 7.634752607229672e-05, "loss": 2.9434, "step": 30164 }, { "epoch": 2.37, "learning_rate": 7.634674194307222e-05, "loss": 2.8075, "step": 30165 }, { "epoch": 2.37, "learning_rate": 7.634595781384773e-05, "loss": 2.858, "step": 30166 }, { "epoch": 2.37, "learning_rate": 7.634517368462323e-05, "loss": 2.8458, "step": 30167 }, { "epoch": 2.37, "learning_rate": 7.634438955539873e-05, "loss": 2.7073, "step": 30168 }, { "epoch": 2.37, "learning_rate": 7.634360542617424e-05, "loss": 2.9822, "step": 30169 }, { "epoch": 2.37, "learning_rate": 7.634282129694974e-05, "loss": 2.8543, "step": 30170 }, { "epoch": 2.37, "learning_rate": 7.634203716772525e-05, "loss": 2.8282, "step": 30171 }, { "epoch": 2.37, "learning_rate": 7.634125303850075e-05, "loss": 2.9739, "step": 30172 }, { "epoch": 2.37, "learning_rate": 7.634046890927625e-05, "loss": 2.725, "step": 30173 }, { "epoch": 2.37, "learning_rate": 7.633968478005176e-05, "loss": 3.1056, "step": 30174 }, { "epoch": 2.37, "learning_rate": 7.633890065082726e-05, "loss": 2.8643, "step": 30175 }, { "epoch": 2.37, "learning_rate": 7.633811652160276e-05, "loss": 2.9437, "step": 30176 }, { "epoch": 2.37, "learning_rate": 7.633733239237827e-05, "loss": 2.5984, "step": 30177 }, { "epoch": 2.37, "learning_rate": 7.633654826315377e-05, "loss": 2.7061, "step": 30178 }, { "epoch": 2.37, "learning_rate": 7.633576413392928e-05, "loss": 2.7395, "step": 30179 }, { "epoch": 2.37, "learning_rate": 7.633498000470478e-05, "loss": 2.9778, "step": 30180 }, { "epoch": 2.37, "learning_rate": 7.633419587548028e-05, "loss": 2.7822, "step": 30181 }, { "epoch": 2.37, "learning_rate": 7.633341174625579e-05, "loss": 2.7521, "step": 30182 }, { "epoch": 2.37, "learning_rate": 7.633262761703129e-05, "loss": 2.8283, "step": 30183 }, { "epoch": 2.37, "learning_rate": 7.63318434878068e-05, "loss": 2.7698, "step": 30184 }, { "epoch": 2.37, "learning_rate": 7.63310593585823e-05, "loss": 2.8686, "step": 30185 }, { "epoch": 2.37, "learning_rate": 7.63302752293578e-05, "loss": 2.8022, "step": 30186 }, { "epoch": 2.37, "learning_rate": 7.63294911001333e-05, "loss": 3.0022, "step": 30187 }, { "epoch": 2.37, "learning_rate": 7.632870697090881e-05, "loss": 2.7791, "step": 30188 }, { "epoch": 2.37, "learning_rate": 7.632792284168431e-05, "loss": 2.9117, "step": 30189 }, { "epoch": 2.37, "learning_rate": 7.632713871245983e-05, "loss": 2.8084, "step": 30190 }, { "epoch": 2.37, "learning_rate": 7.632635458323532e-05, "loss": 2.7549, "step": 30191 }, { "epoch": 2.37, "learning_rate": 7.632557045401082e-05, "loss": 2.6256, "step": 30192 }, { "epoch": 2.37, "learning_rate": 7.632478632478633e-05, "loss": 2.7292, "step": 30193 }, { "epoch": 2.37, "learning_rate": 7.632400219556183e-05, "loss": 2.6969, "step": 30194 }, { "epoch": 2.37, "learning_rate": 7.632321806633733e-05, "loss": 2.7703, "step": 30195 }, { "epoch": 2.37, "learning_rate": 7.632243393711284e-05, "loss": 2.893, "step": 30196 }, { "epoch": 2.37, "learning_rate": 7.632164980788834e-05, "loss": 2.7568, "step": 30197 }, { "epoch": 2.37, "learning_rate": 7.632086567866385e-05, "loss": 2.9503, "step": 30198 }, { "epoch": 2.37, "learning_rate": 7.632008154943935e-05, "loss": 2.8974, "step": 30199 }, { "epoch": 2.37, "learning_rate": 7.631929742021485e-05, "loss": 2.6242, "step": 30200 }, { "epoch": 2.37, "learning_rate": 7.631851329099036e-05, "loss": 3.108, "step": 30201 }, { "epoch": 2.37, "learning_rate": 7.631772916176586e-05, "loss": 2.8489, "step": 30202 }, { "epoch": 2.37, "learning_rate": 7.631694503254136e-05, "loss": 2.985, "step": 30203 }, { "epoch": 2.37, "learning_rate": 7.631616090331687e-05, "loss": 2.6899, "step": 30204 }, { "epoch": 2.37, "learning_rate": 7.631537677409237e-05, "loss": 2.9479, "step": 30205 }, { "epoch": 2.37, "learning_rate": 7.631459264486787e-05, "loss": 2.9066, "step": 30206 }, { "epoch": 2.37, "learning_rate": 7.631380851564338e-05, "loss": 2.8007, "step": 30207 }, { "epoch": 2.37, "learning_rate": 7.631302438641888e-05, "loss": 2.6837, "step": 30208 }, { "epoch": 2.37, "learning_rate": 7.631224025719439e-05, "loss": 2.8633, "step": 30209 }, { "epoch": 2.37, "learning_rate": 7.631145612796989e-05, "loss": 2.7398, "step": 30210 }, { "epoch": 2.37, "learning_rate": 7.631067199874539e-05, "loss": 2.821, "step": 30211 }, { "epoch": 2.37, "learning_rate": 7.63098878695209e-05, "loss": 2.8756, "step": 30212 }, { "epoch": 2.37, "learning_rate": 7.63091037402964e-05, "loss": 2.8844, "step": 30213 }, { "epoch": 2.37, "learning_rate": 7.63083196110719e-05, "loss": 2.6629, "step": 30214 }, { "epoch": 2.37, "learning_rate": 7.630753548184741e-05, "loss": 2.6317, "step": 30215 }, { "epoch": 2.37, "learning_rate": 7.630675135262292e-05, "loss": 3.1276, "step": 30216 }, { "epoch": 2.37, "learning_rate": 7.630596722339843e-05, "loss": 3.1563, "step": 30217 }, { "epoch": 2.37, "learning_rate": 7.630518309417393e-05, "loss": 2.7722, "step": 30218 }, { "epoch": 2.37, "learning_rate": 7.630439896494944e-05, "loss": 2.8255, "step": 30219 }, { "epoch": 2.37, "learning_rate": 7.630361483572494e-05, "loss": 3.1797, "step": 30220 }, { "epoch": 2.37, "learning_rate": 7.630283070650044e-05, "loss": 2.7874, "step": 30221 }, { "epoch": 2.37, "learning_rate": 7.630204657727595e-05, "loss": 3.0203, "step": 30222 }, { "epoch": 2.37, "learning_rate": 7.630126244805145e-05, "loss": 2.5081, "step": 30223 }, { "epoch": 2.37, "learning_rate": 7.630047831882695e-05, "loss": 2.6597, "step": 30224 }, { "epoch": 2.37, "learning_rate": 7.629969418960246e-05, "loss": 2.9399, "step": 30225 }, { "epoch": 2.37, "learning_rate": 7.629891006037796e-05, "loss": 2.7292, "step": 30226 }, { "epoch": 2.37, "learning_rate": 7.629812593115345e-05, "loss": 2.8591, "step": 30227 }, { "epoch": 2.37, "learning_rate": 7.629734180192896e-05, "loss": 2.7802, "step": 30228 }, { "epoch": 2.37, "learning_rate": 7.629655767270446e-05, "loss": 2.8238, "step": 30229 }, { "epoch": 2.37, "learning_rate": 7.629577354347996e-05, "loss": 2.7401, "step": 30230 }, { "epoch": 2.37, "learning_rate": 7.629498941425547e-05, "loss": 2.9424, "step": 30231 }, { "epoch": 2.37, "learning_rate": 7.629420528503097e-05, "loss": 2.7996, "step": 30232 }, { "epoch": 2.37, "learning_rate": 7.629342115580647e-05, "loss": 2.5676, "step": 30233 }, { "epoch": 2.37, "learning_rate": 7.629263702658198e-05, "loss": 2.8342, "step": 30234 }, { "epoch": 2.37, "learning_rate": 7.629185289735748e-05, "loss": 2.9055, "step": 30235 }, { "epoch": 2.37, "learning_rate": 7.629106876813298e-05, "loss": 2.8798, "step": 30236 }, { "epoch": 2.37, "learning_rate": 7.629028463890849e-05, "loss": 2.7449, "step": 30237 }, { "epoch": 2.37, "learning_rate": 7.628950050968399e-05, "loss": 2.8887, "step": 30238 }, { "epoch": 2.37, "learning_rate": 7.62887163804595e-05, "loss": 2.8534, "step": 30239 }, { "epoch": 2.37, "learning_rate": 7.6287932251235e-05, "loss": 3.0153, "step": 30240 }, { "epoch": 2.37, "learning_rate": 7.628714812201052e-05, "loss": 3.0894, "step": 30241 }, { "epoch": 2.37, "learning_rate": 7.628636399278602e-05, "loss": 2.8632, "step": 30242 }, { "epoch": 2.37, "learning_rate": 7.628557986356152e-05, "loss": 2.8545, "step": 30243 }, { "epoch": 2.37, "learning_rate": 7.628479573433703e-05, "loss": 3.0381, "step": 30244 }, { "epoch": 2.37, "learning_rate": 7.628401160511253e-05, "loss": 3.0723, "step": 30245 }, { "epoch": 2.37, "learning_rate": 7.628322747588804e-05, "loss": 2.7479, "step": 30246 }, { "epoch": 2.37, "learning_rate": 7.628244334666354e-05, "loss": 2.6116, "step": 30247 }, { "epoch": 2.37, "learning_rate": 7.628165921743904e-05, "loss": 2.948, "step": 30248 }, { "epoch": 2.37, "learning_rate": 7.628087508821455e-05, "loss": 3.0096, "step": 30249 }, { "epoch": 2.37, "learning_rate": 7.628009095899005e-05, "loss": 2.7897, "step": 30250 }, { "epoch": 2.37, "learning_rate": 7.627930682976555e-05, "loss": 2.9742, "step": 30251 }, { "epoch": 2.37, "learning_rate": 7.627852270054106e-05, "loss": 2.7821, "step": 30252 }, { "epoch": 2.37, "learning_rate": 7.627773857131656e-05, "loss": 2.7829, "step": 30253 }, { "epoch": 2.37, "learning_rate": 7.627695444209206e-05, "loss": 2.7128, "step": 30254 }, { "epoch": 2.37, "learning_rate": 7.627617031286757e-05, "loss": 2.7549, "step": 30255 }, { "epoch": 2.37, "learning_rate": 7.627538618364307e-05, "loss": 2.7971, "step": 30256 }, { "epoch": 2.37, "learning_rate": 7.627460205441858e-05, "loss": 2.936, "step": 30257 }, { "epoch": 2.37, "learning_rate": 7.627381792519408e-05, "loss": 2.8469, "step": 30258 }, { "epoch": 2.37, "learning_rate": 7.627303379596958e-05, "loss": 2.8175, "step": 30259 }, { "epoch": 2.37, "learning_rate": 7.627224966674509e-05, "loss": 2.8269, "step": 30260 }, { "epoch": 2.37, "learning_rate": 7.627146553752059e-05, "loss": 2.8067, "step": 30261 }, { "epoch": 2.37, "learning_rate": 7.62706814082961e-05, "loss": 2.8146, "step": 30262 }, { "epoch": 2.37, "learning_rate": 7.626989727907158e-05, "loss": 2.7737, "step": 30263 }, { "epoch": 2.37, "learning_rate": 7.626911314984709e-05, "loss": 2.6418, "step": 30264 }, { "epoch": 2.37, "learning_rate": 7.626832902062259e-05, "loss": 2.8009, "step": 30265 }, { "epoch": 2.37, "learning_rate": 7.62675448913981e-05, "loss": 2.8344, "step": 30266 }, { "epoch": 2.37, "learning_rate": 7.626676076217361e-05, "loss": 2.6797, "step": 30267 }, { "epoch": 2.37, "learning_rate": 7.626597663294912e-05, "loss": 2.8441, "step": 30268 }, { "epoch": 2.37, "learning_rate": 7.626519250372462e-05, "loss": 2.695, "step": 30269 }, { "epoch": 2.37, "learning_rate": 7.626440837450012e-05, "loss": 2.8577, "step": 30270 }, { "epoch": 2.37, "learning_rate": 7.626362424527563e-05, "loss": 2.8373, "step": 30271 }, { "epoch": 2.37, "learning_rate": 7.626284011605113e-05, "loss": 2.8032, "step": 30272 }, { "epoch": 2.37, "learning_rate": 7.626205598682663e-05, "loss": 2.6819, "step": 30273 }, { "epoch": 2.37, "learning_rate": 7.626127185760214e-05, "loss": 2.879, "step": 30274 }, { "epoch": 2.37, "learning_rate": 7.626048772837764e-05, "loss": 2.7583, "step": 30275 }, { "epoch": 2.37, "learning_rate": 7.625970359915315e-05, "loss": 2.8182, "step": 30276 }, { "epoch": 2.37, "learning_rate": 7.625891946992865e-05, "loss": 2.6966, "step": 30277 }, { "epoch": 2.37, "learning_rate": 7.625813534070415e-05, "loss": 2.8282, "step": 30278 }, { "epoch": 2.37, "learning_rate": 7.625735121147966e-05, "loss": 2.845, "step": 30279 }, { "epoch": 2.37, "learning_rate": 7.625656708225516e-05, "loss": 2.6824, "step": 30280 }, { "epoch": 2.37, "learning_rate": 7.625578295303066e-05, "loss": 2.8406, "step": 30281 }, { "epoch": 2.37, "learning_rate": 7.625499882380617e-05, "loss": 2.8867, "step": 30282 }, { "epoch": 2.37, "learning_rate": 7.625421469458167e-05, "loss": 2.897, "step": 30283 }, { "epoch": 2.37, "learning_rate": 7.625343056535717e-05, "loss": 2.8125, "step": 30284 }, { "epoch": 2.37, "learning_rate": 7.625264643613268e-05, "loss": 2.8742, "step": 30285 }, { "epoch": 2.37, "learning_rate": 7.625186230690818e-05, "loss": 2.7909, "step": 30286 }, { "epoch": 2.37, "learning_rate": 7.625107817768369e-05, "loss": 2.6668, "step": 30287 }, { "epoch": 2.37, "learning_rate": 7.625029404845919e-05, "loss": 2.875, "step": 30288 }, { "epoch": 2.38, "learning_rate": 7.62495099192347e-05, "loss": 2.8484, "step": 30289 }, { "epoch": 2.38, "learning_rate": 7.62487257900102e-05, "loss": 2.8123, "step": 30290 }, { "epoch": 2.38, "learning_rate": 7.62479416607857e-05, "loss": 3.0052, "step": 30291 }, { "epoch": 2.38, "learning_rate": 7.62471575315612e-05, "loss": 2.6475, "step": 30292 }, { "epoch": 2.38, "learning_rate": 7.624637340233671e-05, "loss": 3.0307, "step": 30293 }, { "epoch": 2.38, "learning_rate": 7.624558927311221e-05, "loss": 2.9005, "step": 30294 }, { "epoch": 2.38, "learning_rate": 7.624480514388772e-05, "loss": 2.9325, "step": 30295 }, { "epoch": 2.38, "learning_rate": 7.624402101466322e-05, "loss": 2.9469, "step": 30296 }, { "epoch": 2.38, "learning_rate": 7.624323688543872e-05, "loss": 2.6625, "step": 30297 }, { "epoch": 2.38, "learning_rate": 7.624245275621424e-05, "loss": 2.7565, "step": 30298 }, { "epoch": 2.38, "learning_rate": 7.624166862698973e-05, "loss": 2.8497, "step": 30299 }, { "epoch": 2.38, "learning_rate": 7.624088449776523e-05, "loss": 2.6813, "step": 30300 }, { "epoch": 2.38, "learning_rate": 7.624010036854074e-05, "loss": 2.5928, "step": 30301 }, { "epoch": 2.38, "learning_rate": 7.623931623931624e-05, "loss": 2.9468, "step": 30302 }, { "epoch": 2.38, "learning_rate": 7.623853211009174e-05, "loss": 2.6499, "step": 30303 }, { "epoch": 2.38, "learning_rate": 7.623774798086725e-05, "loss": 2.7236, "step": 30304 }, { "epoch": 2.38, "learning_rate": 7.623696385164275e-05, "loss": 2.7203, "step": 30305 }, { "epoch": 2.38, "learning_rate": 7.623617972241826e-05, "loss": 2.9124, "step": 30306 }, { "epoch": 2.38, "learning_rate": 7.623539559319376e-05, "loss": 2.8564, "step": 30307 }, { "epoch": 2.38, "learning_rate": 7.623461146396926e-05, "loss": 2.938, "step": 30308 }, { "epoch": 2.38, "learning_rate": 7.623382733474477e-05, "loss": 2.8088, "step": 30309 }, { "epoch": 2.38, "learning_rate": 7.623304320552027e-05, "loss": 2.6971, "step": 30310 }, { "epoch": 2.38, "learning_rate": 7.623225907629577e-05, "loss": 2.8433, "step": 30311 }, { "epoch": 2.38, "learning_rate": 7.623147494707128e-05, "loss": 2.9125, "step": 30312 }, { "epoch": 2.38, "learning_rate": 7.623069081784678e-05, "loss": 3.1257, "step": 30313 }, { "epoch": 2.38, "learning_rate": 7.622990668862229e-05, "loss": 3.1871, "step": 30314 }, { "epoch": 2.38, "learning_rate": 7.622912255939779e-05, "loss": 2.7649, "step": 30315 }, { "epoch": 2.38, "learning_rate": 7.622833843017329e-05, "loss": 2.6219, "step": 30316 }, { "epoch": 2.38, "learning_rate": 7.62275543009488e-05, "loss": 2.5951, "step": 30317 }, { "epoch": 2.38, "learning_rate": 7.62267701717243e-05, "loss": 2.5141, "step": 30318 }, { "epoch": 2.38, "learning_rate": 7.62259860424998e-05, "loss": 3.0216, "step": 30319 }, { "epoch": 2.38, "learning_rate": 7.622520191327531e-05, "loss": 3.0805, "step": 30320 }, { "epoch": 2.38, "learning_rate": 7.622441778405081e-05, "loss": 2.9007, "step": 30321 }, { "epoch": 2.38, "learning_rate": 7.622363365482631e-05, "loss": 2.8841, "step": 30322 }, { "epoch": 2.38, "learning_rate": 7.622284952560182e-05, "loss": 2.6265, "step": 30323 }, { "epoch": 2.38, "learning_rate": 7.622206539637734e-05, "loss": 2.7387, "step": 30324 }, { "epoch": 2.38, "learning_rate": 7.622128126715284e-05, "loss": 2.7745, "step": 30325 }, { "epoch": 2.38, "learning_rate": 7.622049713792834e-05, "loss": 2.6328, "step": 30326 }, { "epoch": 2.38, "learning_rate": 7.621971300870385e-05, "loss": 2.7379, "step": 30327 }, { "epoch": 2.38, "learning_rate": 7.621892887947935e-05, "loss": 2.693, "step": 30328 }, { "epoch": 2.38, "learning_rate": 7.621814475025485e-05, "loss": 2.7699, "step": 30329 }, { "epoch": 2.38, "learning_rate": 7.621736062103036e-05, "loss": 2.9631, "step": 30330 }, { "epoch": 2.38, "learning_rate": 7.621657649180586e-05, "loss": 2.99, "step": 30331 }, { "epoch": 2.38, "learning_rate": 7.621579236258136e-05, "loss": 2.6765, "step": 30332 }, { "epoch": 2.38, "learning_rate": 7.621500823335687e-05, "loss": 2.8268, "step": 30333 }, { "epoch": 2.38, "learning_rate": 7.621422410413237e-05, "loss": 2.8142, "step": 30334 }, { "epoch": 2.38, "learning_rate": 7.621343997490786e-05, "loss": 2.9623, "step": 30335 }, { "epoch": 2.38, "learning_rate": 7.621265584568337e-05, "loss": 2.7607, "step": 30336 }, { "epoch": 2.38, "learning_rate": 7.621187171645887e-05, "loss": 2.8424, "step": 30337 }, { "epoch": 2.38, "learning_rate": 7.621108758723437e-05, "loss": 2.7411, "step": 30338 }, { "epoch": 2.38, "learning_rate": 7.621030345800988e-05, "loss": 2.7595, "step": 30339 }, { "epoch": 2.38, "learning_rate": 7.620951932878538e-05, "loss": 2.8372, "step": 30340 }, { "epoch": 2.38, "learning_rate": 7.620873519956088e-05, "loss": 2.8431, "step": 30341 }, { "epoch": 2.38, "learning_rate": 7.620795107033639e-05, "loss": 2.9463, "step": 30342 }, { "epoch": 2.38, "learning_rate": 7.620716694111189e-05, "loss": 2.7724, "step": 30343 }, { "epoch": 2.38, "learning_rate": 7.62063828118874e-05, "loss": 2.7222, "step": 30344 }, { "epoch": 2.38, "learning_rate": 7.62055986826629e-05, "loss": 3.044, "step": 30345 }, { "epoch": 2.38, "learning_rate": 7.62048145534384e-05, "loss": 2.897, "step": 30346 }, { "epoch": 2.38, "learning_rate": 7.62040304242139e-05, "loss": 2.6639, "step": 30347 }, { "epoch": 2.38, "learning_rate": 7.620324629498941e-05, "loss": 2.6683, "step": 30348 }, { "epoch": 2.38, "learning_rate": 7.620246216576491e-05, "loss": 2.697, "step": 30349 }, { "epoch": 2.38, "learning_rate": 7.620167803654043e-05, "loss": 2.945, "step": 30350 }, { "epoch": 2.38, "learning_rate": 7.620089390731593e-05, "loss": 2.7558, "step": 30351 }, { "epoch": 2.38, "learning_rate": 7.620010977809144e-05, "loss": 2.6939, "step": 30352 }, { "epoch": 2.38, "learning_rate": 7.619932564886694e-05, "loss": 2.6873, "step": 30353 }, { "epoch": 2.38, "learning_rate": 7.619854151964245e-05, "loss": 2.7467, "step": 30354 }, { "epoch": 2.38, "learning_rate": 7.619775739041795e-05, "loss": 2.8301, "step": 30355 }, { "epoch": 2.38, "learning_rate": 7.619697326119345e-05, "loss": 2.8476, "step": 30356 }, { "epoch": 2.38, "learning_rate": 7.619618913196896e-05, "loss": 2.6563, "step": 30357 }, { "epoch": 2.38, "learning_rate": 7.619540500274446e-05, "loss": 2.7475, "step": 30358 }, { "epoch": 2.38, "learning_rate": 7.619462087351996e-05, "loss": 2.8185, "step": 30359 }, { "epoch": 2.38, "learning_rate": 7.619383674429547e-05, "loss": 2.6258, "step": 30360 }, { "epoch": 2.38, "learning_rate": 7.619305261507097e-05, "loss": 2.8774, "step": 30361 }, { "epoch": 2.38, "learning_rate": 7.619226848584648e-05, "loss": 2.8364, "step": 30362 }, { "epoch": 2.38, "learning_rate": 7.619148435662198e-05, "loss": 2.7295, "step": 30363 }, { "epoch": 2.38, "learning_rate": 7.619070022739748e-05, "loss": 2.9226, "step": 30364 }, { "epoch": 2.38, "learning_rate": 7.618991609817299e-05, "loss": 3.0723, "step": 30365 }, { "epoch": 2.38, "learning_rate": 7.618913196894849e-05, "loss": 2.8378, "step": 30366 }, { "epoch": 2.38, "learning_rate": 7.6188347839724e-05, "loss": 2.7615, "step": 30367 }, { "epoch": 2.38, "learning_rate": 7.61875637104995e-05, "loss": 3.0442, "step": 30368 }, { "epoch": 2.38, "learning_rate": 7.6186779581275e-05, "loss": 3.055, "step": 30369 }, { "epoch": 2.38, "learning_rate": 7.61859954520505e-05, "loss": 2.7752, "step": 30370 }, { "epoch": 2.38, "learning_rate": 7.6185211322826e-05, "loss": 2.8026, "step": 30371 }, { "epoch": 2.38, "learning_rate": 7.61844271936015e-05, "loss": 2.7747, "step": 30372 }, { "epoch": 2.38, "learning_rate": 7.6183643064377e-05, "loss": 2.6902, "step": 30373 }, { "epoch": 2.38, "learning_rate": 7.61828589351525e-05, "loss": 2.9355, "step": 30374 }, { "epoch": 2.38, "learning_rate": 7.618207480592802e-05, "loss": 2.6739, "step": 30375 }, { "epoch": 2.38, "learning_rate": 7.618129067670353e-05, "loss": 2.8701, "step": 30376 }, { "epoch": 2.38, "learning_rate": 7.618050654747903e-05, "loss": 2.8227, "step": 30377 }, { "epoch": 2.38, "learning_rate": 7.617972241825453e-05, "loss": 2.5661, "step": 30378 }, { "epoch": 2.38, "learning_rate": 7.617893828903004e-05, "loss": 2.6164, "step": 30379 }, { "epoch": 2.38, "learning_rate": 7.617815415980554e-05, "loss": 2.7885, "step": 30380 }, { "epoch": 2.38, "learning_rate": 7.617737003058105e-05, "loss": 2.8109, "step": 30381 }, { "epoch": 2.38, "learning_rate": 7.617658590135655e-05, "loss": 2.6488, "step": 30382 }, { "epoch": 2.38, "learning_rate": 7.617580177213205e-05, "loss": 2.7468, "step": 30383 }, { "epoch": 2.38, "learning_rate": 7.617501764290756e-05, "loss": 2.889, "step": 30384 }, { "epoch": 2.38, "learning_rate": 7.617423351368306e-05, "loss": 2.7718, "step": 30385 }, { "epoch": 2.38, "learning_rate": 7.617344938445856e-05, "loss": 2.819, "step": 30386 }, { "epoch": 2.38, "learning_rate": 7.617266525523407e-05, "loss": 2.9213, "step": 30387 }, { "epoch": 2.38, "learning_rate": 7.617188112600957e-05, "loss": 2.6935, "step": 30388 }, { "epoch": 2.38, "learning_rate": 7.617109699678507e-05, "loss": 2.6654, "step": 30389 }, { "epoch": 2.38, "learning_rate": 7.617031286756058e-05, "loss": 2.8968, "step": 30390 }, { "epoch": 2.38, "learning_rate": 7.616952873833608e-05, "loss": 2.7946, "step": 30391 }, { "epoch": 2.38, "learning_rate": 7.616874460911159e-05, "loss": 2.8005, "step": 30392 }, { "epoch": 2.38, "learning_rate": 7.616796047988709e-05, "loss": 2.8759, "step": 30393 }, { "epoch": 2.38, "learning_rate": 7.616717635066259e-05, "loss": 2.9589, "step": 30394 }, { "epoch": 2.38, "learning_rate": 7.61663922214381e-05, "loss": 3.0152, "step": 30395 }, { "epoch": 2.38, "learning_rate": 7.61656080922136e-05, "loss": 2.6555, "step": 30396 }, { "epoch": 2.38, "learning_rate": 7.61648239629891e-05, "loss": 2.9404, "step": 30397 }, { "epoch": 2.38, "learning_rate": 7.616403983376461e-05, "loss": 2.7879, "step": 30398 }, { "epoch": 2.38, "learning_rate": 7.616325570454011e-05, "loss": 2.7813, "step": 30399 }, { "epoch": 2.38, "learning_rate": 7.616247157531562e-05, "loss": 2.8593, "step": 30400 }, { "epoch": 2.38, "learning_rate": 7.616168744609112e-05, "loss": 2.7528, "step": 30401 }, { "epoch": 2.38, "learning_rate": 7.616090331686662e-05, "loss": 2.7796, "step": 30402 }, { "epoch": 2.38, "learning_rate": 7.616011918764213e-05, "loss": 2.6025, "step": 30403 }, { "epoch": 2.38, "learning_rate": 7.615933505841763e-05, "loss": 2.9366, "step": 30404 }, { "epoch": 2.38, "learning_rate": 7.615855092919313e-05, "loss": 2.7786, "step": 30405 }, { "epoch": 2.38, "learning_rate": 7.615776679996865e-05, "loss": 2.7045, "step": 30406 }, { "epoch": 2.38, "learning_rate": 7.615698267074414e-05, "loss": 2.5831, "step": 30407 }, { "epoch": 2.38, "learning_rate": 7.615619854151964e-05, "loss": 2.628, "step": 30408 }, { "epoch": 2.38, "learning_rate": 7.615541441229515e-05, "loss": 2.7467, "step": 30409 }, { "epoch": 2.38, "learning_rate": 7.615463028307065e-05, "loss": 2.7845, "step": 30410 }, { "epoch": 2.38, "learning_rate": 7.615384615384616e-05, "loss": 2.924, "step": 30411 }, { "epoch": 2.38, "learning_rate": 7.615306202462166e-05, "loss": 2.6973, "step": 30412 }, { "epoch": 2.38, "learning_rate": 7.615227789539716e-05, "loss": 2.8689, "step": 30413 }, { "epoch": 2.38, "learning_rate": 7.615149376617267e-05, "loss": 2.8552, "step": 30414 }, { "epoch": 2.38, "learning_rate": 7.615070963694817e-05, "loss": 2.7844, "step": 30415 }, { "epoch": 2.39, "learning_rate": 7.614992550772367e-05, "loss": 2.8891, "step": 30416 }, { "epoch": 2.39, "learning_rate": 7.614914137849918e-05, "loss": 2.7916, "step": 30417 }, { "epoch": 2.39, "learning_rate": 7.614835724927468e-05, "loss": 2.8984, "step": 30418 }, { "epoch": 2.39, "learning_rate": 7.614757312005018e-05, "loss": 2.693, "step": 30419 }, { "epoch": 2.39, "learning_rate": 7.614678899082569e-05, "loss": 2.6545, "step": 30420 }, { "epoch": 2.39, "learning_rate": 7.614600486160119e-05, "loss": 2.9686, "step": 30421 }, { "epoch": 2.39, "learning_rate": 7.61452207323767e-05, "loss": 2.7184, "step": 30422 }, { "epoch": 2.39, "learning_rate": 7.61444366031522e-05, "loss": 2.8698, "step": 30423 }, { "epoch": 2.39, "learning_rate": 7.61436524739277e-05, "loss": 2.7919, "step": 30424 }, { "epoch": 2.39, "learning_rate": 7.614286834470321e-05, "loss": 2.7319, "step": 30425 }, { "epoch": 2.39, "learning_rate": 7.614208421547871e-05, "loss": 2.6804, "step": 30426 }, { "epoch": 2.39, "learning_rate": 7.614130008625421e-05, "loss": 2.8724, "step": 30427 }, { "epoch": 2.39, "learning_rate": 7.614051595702972e-05, "loss": 2.8536, "step": 30428 }, { "epoch": 2.39, "learning_rate": 7.613973182780522e-05, "loss": 2.8932, "step": 30429 }, { "epoch": 2.39, "learning_rate": 7.613894769858073e-05, "loss": 2.6467, "step": 30430 }, { "epoch": 2.39, "learning_rate": 7.613816356935623e-05, "loss": 2.776, "step": 30431 }, { "epoch": 2.39, "learning_rate": 7.613737944013175e-05, "loss": 2.8563, "step": 30432 }, { "epoch": 2.39, "learning_rate": 7.613659531090725e-05, "loss": 2.8552, "step": 30433 }, { "epoch": 2.39, "learning_rate": 7.613581118168275e-05, "loss": 2.9074, "step": 30434 }, { "epoch": 2.39, "learning_rate": 7.613502705245826e-05, "loss": 2.9138, "step": 30435 }, { "epoch": 2.39, "learning_rate": 7.613424292323376e-05, "loss": 2.8993, "step": 30436 }, { "epoch": 2.39, "learning_rate": 7.613345879400926e-05, "loss": 2.6793, "step": 30437 }, { "epoch": 2.39, "learning_rate": 7.613267466478477e-05, "loss": 2.868, "step": 30438 }, { "epoch": 2.39, "learning_rate": 7.613189053556027e-05, "loss": 2.6163, "step": 30439 }, { "epoch": 2.39, "learning_rate": 7.613110640633578e-05, "loss": 2.8169, "step": 30440 }, { "epoch": 2.39, "learning_rate": 7.613032227711128e-05, "loss": 2.7105, "step": 30441 }, { "epoch": 2.39, "learning_rate": 7.612953814788678e-05, "loss": 2.8046, "step": 30442 }, { "epoch": 2.39, "learning_rate": 7.612875401866227e-05, "loss": 2.8165, "step": 30443 }, { "epoch": 2.39, "learning_rate": 7.612796988943778e-05, "loss": 2.7839, "step": 30444 }, { "epoch": 2.39, "learning_rate": 7.612718576021328e-05, "loss": 2.9786, "step": 30445 }, { "epoch": 2.39, "learning_rate": 7.612640163098878e-05, "loss": 2.7566, "step": 30446 }, { "epoch": 2.39, "learning_rate": 7.612561750176429e-05, "loss": 2.8075, "step": 30447 }, { "epoch": 2.39, "learning_rate": 7.612483337253979e-05, "loss": 2.7536, "step": 30448 }, { "epoch": 2.39, "learning_rate": 7.61240492433153e-05, "loss": 2.7657, "step": 30449 }, { "epoch": 2.39, "learning_rate": 7.61232651140908e-05, "loss": 2.8131, "step": 30450 }, { "epoch": 2.39, "learning_rate": 7.61224809848663e-05, "loss": 2.6915, "step": 30451 }, { "epoch": 2.39, "learning_rate": 7.61216968556418e-05, "loss": 2.7829, "step": 30452 }, { "epoch": 2.39, "learning_rate": 7.612091272641731e-05, "loss": 2.902, "step": 30453 }, { "epoch": 2.39, "learning_rate": 7.612012859719281e-05, "loss": 2.7153, "step": 30454 }, { "epoch": 2.39, "learning_rate": 7.611934446796832e-05, "loss": 2.7409, "step": 30455 }, { "epoch": 2.39, "learning_rate": 7.611856033874382e-05, "loss": 2.8691, "step": 30456 }, { "epoch": 2.39, "learning_rate": 7.611777620951932e-05, "loss": 2.848, "step": 30457 }, { "epoch": 2.39, "learning_rate": 7.611699208029484e-05, "loss": 2.7977, "step": 30458 }, { "epoch": 2.39, "learning_rate": 7.611620795107035e-05, "loss": 2.9006, "step": 30459 }, { "epoch": 2.39, "learning_rate": 7.611542382184585e-05, "loss": 2.7731, "step": 30460 }, { "epoch": 2.39, "learning_rate": 7.611463969262135e-05, "loss": 2.8178, "step": 30461 }, { "epoch": 2.39, "learning_rate": 7.611385556339686e-05, "loss": 2.818, "step": 30462 }, { "epoch": 2.39, "learning_rate": 7.611307143417236e-05, "loss": 2.6636, "step": 30463 }, { "epoch": 2.39, "learning_rate": 7.611228730494786e-05, "loss": 2.7902, "step": 30464 }, { "epoch": 2.39, "learning_rate": 7.611150317572337e-05, "loss": 2.7927, "step": 30465 }, { "epoch": 2.39, "learning_rate": 7.611071904649887e-05, "loss": 2.8368, "step": 30466 }, { "epoch": 2.39, "learning_rate": 7.610993491727438e-05, "loss": 2.6063, "step": 30467 }, { "epoch": 2.39, "learning_rate": 7.610915078804988e-05, "loss": 2.6701, "step": 30468 }, { "epoch": 2.39, "learning_rate": 7.610836665882538e-05, "loss": 2.7324, "step": 30469 }, { "epoch": 2.39, "learning_rate": 7.610758252960089e-05, "loss": 2.7232, "step": 30470 }, { "epoch": 2.39, "learning_rate": 7.610679840037639e-05, "loss": 2.9495, "step": 30471 }, { "epoch": 2.39, "learning_rate": 7.61060142711519e-05, "loss": 2.6572, "step": 30472 }, { "epoch": 2.39, "learning_rate": 7.61052301419274e-05, "loss": 2.8882, "step": 30473 }, { "epoch": 2.39, "learning_rate": 7.61044460127029e-05, "loss": 2.8723, "step": 30474 }, { "epoch": 2.39, "learning_rate": 7.61036618834784e-05, "loss": 2.621, "step": 30475 }, { "epoch": 2.39, "learning_rate": 7.610287775425391e-05, "loss": 2.6868, "step": 30476 }, { "epoch": 2.39, "learning_rate": 7.610209362502941e-05, "loss": 2.9236, "step": 30477 }, { "epoch": 2.39, "learning_rate": 7.610130949580492e-05, "loss": 3.1369, "step": 30478 }, { "epoch": 2.39, "learning_rate": 7.61005253665804e-05, "loss": 2.7274, "step": 30479 }, { "epoch": 2.39, "learning_rate": 7.609974123735591e-05, "loss": 2.9315, "step": 30480 }, { "epoch": 2.39, "learning_rate": 7.609895710813141e-05, "loss": 2.839, "step": 30481 }, { "epoch": 2.39, "learning_rate": 7.609817297890692e-05, "loss": 2.7338, "step": 30482 }, { "epoch": 2.39, "learning_rate": 7.609738884968243e-05, "loss": 3.0409, "step": 30483 }, { "epoch": 2.39, "learning_rate": 7.609660472045794e-05, "loss": 2.7253, "step": 30484 }, { "epoch": 2.39, "learning_rate": 7.609582059123344e-05, "loss": 2.7921, "step": 30485 }, { "epoch": 2.39, "learning_rate": 7.609503646200894e-05, "loss": 2.7913, "step": 30486 }, { "epoch": 2.39, "learning_rate": 7.609425233278445e-05, "loss": 2.9275, "step": 30487 }, { "epoch": 2.39, "learning_rate": 7.609346820355995e-05, "loss": 2.8849, "step": 30488 }, { "epoch": 2.39, "learning_rate": 7.609268407433546e-05, "loss": 2.965, "step": 30489 }, { "epoch": 2.39, "learning_rate": 7.609189994511096e-05, "loss": 3.0467, "step": 30490 }, { "epoch": 2.39, "learning_rate": 7.609111581588646e-05, "loss": 2.6599, "step": 30491 }, { "epoch": 2.39, "learning_rate": 7.609033168666197e-05, "loss": 2.7348, "step": 30492 }, { "epoch": 2.39, "learning_rate": 7.608954755743747e-05, "loss": 2.6067, "step": 30493 }, { "epoch": 2.39, "learning_rate": 7.608876342821297e-05, "loss": 2.7508, "step": 30494 }, { "epoch": 2.39, "learning_rate": 7.608797929898848e-05, "loss": 2.6416, "step": 30495 }, { "epoch": 2.39, "learning_rate": 7.608719516976398e-05, "loss": 2.818, "step": 30496 }, { "epoch": 2.39, "learning_rate": 7.608641104053949e-05, "loss": 2.7435, "step": 30497 }, { "epoch": 2.39, "learning_rate": 7.608562691131499e-05, "loss": 2.6819, "step": 30498 }, { "epoch": 2.39, "learning_rate": 7.608484278209049e-05, "loss": 2.7679, "step": 30499 }, { "epoch": 2.39, "learning_rate": 7.6084058652866e-05, "loss": 2.8751, "step": 30500 }, { "epoch": 2.39, "learning_rate": 7.60832745236415e-05, "loss": 2.7979, "step": 30501 }, { "epoch": 2.39, "learning_rate": 7.6082490394417e-05, "loss": 3.0481, "step": 30502 }, { "epoch": 2.39, "learning_rate": 7.608170626519251e-05, "loss": 2.9816, "step": 30503 }, { "epoch": 2.39, "learning_rate": 7.608092213596801e-05, "loss": 2.8577, "step": 30504 }, { "epoch": 2.39, "learning_rate": 7.608013800674351e-05, "loss": 2.7431, "step": 30505 }, { "epoch": 2.39, "learning_rate": 7.607935387751902e-05, "loss": 2.8751, "step": 30506 }, { "epoch": 2.39, "learning_rate": 7.607856974829452e-05, "loss": 2.7967, "step": 30507 }, { "epoch": 2.39, "learning_rate": 7.607778561907003e-05, "loss": 2.8027, "step": 30508 }, { "epoch": 2.39, "learning_rate": 7.607700148984553e-05, "loss": 2.6934, "step": 30509 }, { "epoch": 2.39, "learning_rate": 7.607621736062103e-05, "loss": 2.5933, "step": 30510 }, { "epoch": 2.39, "learning_rate": 7.607543323139654e-05, "loss": 2.7547, "step": 30511 }, { "epoch": 2.39, "learning_rate": 7.607464910217204e-05, "loss": 2.823, "step": 30512 }, { "epoch": 2.39, "learning_rate": 7.607386497294754e-05, "loss": 2.9251, "step": 30513 }, { "epoch": 2.39, "learning_rate": 7.607308084372306e-05, "loss": 2.6801, "step": 30514 }, { "epoch": 2.39, "learning_rate": 7.607229671449855e-05, "loss": 2.8337, "step": 30515 }, { "epoch": 2.39, "learning_rate": 7.607151258527406e-05, "loss": 2.7978, "step": 30516 }, { "epoch": 2.39, "learning_rate": 7.607072845604956e-05, "loss": 2.8644, "step": 30517 }, { "epoch": 2.39, "learning_rate": 7.606994432682506e-05, "loss": 2.6038, "step": 30518 }, { "epoch": 2.39, "learning_rate": 7.606916019760057e-05, "loss": 2.8346, "step": 30519 }, { "epoch": 2.39, "learning_rate": 7.606837606837607e-05, "loss": 2.7692, "step": 30520 }, { "epoch": 2.39, "learning_rate": 7.606759193915157e-05, "loss": 2.6955, "step": 30521 }, { "epoch": 2.39, "learning_rate": 7.606680780992708e-05, "loss": 2.7405, "step": 30522 }, { "epoch": 2.39, "learning_rate": 7.606602368070258e-05, "loss": 2.8811, "step": 30523 }, { "epoch": 2.39, "learning_rate": 7.606523955147808e-05, "loss": 2.7772, "step": 30524 }, { "epoch": 2.39, "learning_rate": 7.606445542225359e-05, "loss": 2.7653, "step": 30525 }, { "epoch": 2.39, "learning_rate": 7.606367129302909e-05, "loss": 2.9554, "step": 30526 }, { "epoch": 2.39, "learning_rate": 7.60628871638046e-05, "loss": 2.7217, "step": 30527 }, { "epoch": 2.39, "learning_rate": 7.60621030345801e-05, "loss": 2.6723, "step": 30528 }, { "epoch": 2.39, "learning_rate": 7.60613189053556e-05, "loss": 2.7767, "step": 30529 }, { "epoch": 2.39, "learning_rate": 7.60605347761311e-05, "loss": 2.8392, "step": 30530 }, { "epoch": 2.39, "learning_rate": 7.605975064690661e-05, "loss": 2.9387, "step": 30531 }, { "epoch": 2.39, "learning_rate": 7.605896651768211e-05, "loss": 2.8191, "step": 30532 }, { "epoch": 2.39, "learning_rate": 7.605818238845762e-05, "loss": 2.9253, "step": 30533 }, { "epoch": 2.39, "learning_rate": 7.605739825923312e-05, "loss": 2.6538, "step": 30534 }, { "epoch": 2.39, "learning_rate": 7.605661413000863e-05, "loss": 2.9134, "step": 30535 }, { "epoch": 2.39, "learning_rate": 7.605583000078413e-05, "loss": 2.7708, "step": 30536 }, { "epoch": 2.39, "learning_rate": 7.605504587155963e-05, "loss": 2.7169, "step": 30537 }, { "epoch": 2.39, "learning_rate": 7.605426174233514e-05, "loss": 2.9823, "step": 30538 }, { "epoch": 2.39, "learning_rate": 7.605347761311064e-05, "loss": 2.7473, "step": 30539 }, { "epoch": 2.39, "learning_rate": 7.605269348388616e-05, "loss": 2.8001, "step": 30540 }, { "epoch": 2.39, "learning_rate": 7.605190935466166e-05, "loss": 2.8284, "step": 30541 }, { "epoch": 2.39, "learning_rate": 7.605112522543716e-05, "loss": 2.675, "step": 30542 }, { "epoch": 2.39, "learning_rate": 7.605034109621267e-05, "loss": 2.663, "step": 30543 }, { "epoch": 2.4, "learning_rate": 7.604955696698817e-05, "loss": 2.7542, "step": 30544 }, { "epoch": 2.4, "learning_rate": 7.604877283776368e-05, "loss": 2.7134, "step": 30545 }, { "epoch": 2.4, "learning_rate": 7.604798870853918e-05, "loss": 2.7691, "step": 30546 }, { "epoch": 2.4, "learning_rate": 7.604720457931468e-05, "loss": 2.96, "step": 30547 }, { "epoch": 2.4, "learning_rate": 7.604642045009019e-05, "loss": 2.84, "step": 30548 }, { "epoch": 2.4, "learning_rate": 7.604563632086569e-05, "loss": 2.6774, "step": 30549 }, { "epoch": 2.4, "learning_rate": 7.60448521916412e-05, "loss": 2.8437, "step": 30550 }, { "epoch": 2.4, "learning_rate": 7.604406806241668e-05, "loss": 2.7052, "step": 30551 }, { "epoch": 2.4, "learning_rate": 7.604328393319219e-05, "loss": 2.6641, "step": 30552 }, { "epoch": 2.4, "learning_rate": 7.604249980396769e-05, "loss": 2.8784, "step": 30553 }, { "epoch": 2.4, "learning_rate": 7.60417156747432e-05, "loss": 2.682, "step": 30554 }, { "epoch": 2.4, "learning_rate": 7.60409315455187e-05, "loss": 2.6452, "step": 30555 }, { "epoch": 2.4, "learning_rate": 7.60401474162942e-05, "loss": 2.7039, "step": 30556 }, { "epoch": 2.4, "learning_rate": 7.60393632870697e-05, "loss": 2.9338, "step": 30557 }, { "epoch": 2.4, "learning_rate": 7.603857915784521e-05, "loss": 2.8588, "step": 30558 }, { "epoch": 2.4, "learning_rate": 7.603779502862071e-05, "loss": 2.9255, "step": 30559 }, { "epoch": 2.4, "learning_rate": 7.603701089939622e-05, "loss": 2.9106, "step": 30560 }, { "epoch": 2.4, "learning_rate": 7.603622677017172e-05, "loss": 3.1398, "step": 30561 }, { "epoch": 2.4, "learning_rate": 7.603544264094722e-05, "loss": 2.8598, "step": 30562 }, { "epoch": 2.4, "learning_rate": 7.603465851172273e-05, "loss": 2.9438, "step": 30563 }, { "epoch": 2.4, "learning_rate": 7.603387438249823e-05, "loss": 2.8268, "step": 30564 }, { "epoch": 2.4, "learning_rate": 7.603309025327374e-05, "loss": 2.8438, "step": 30565 }, { "epoch": 2.4, "learning_rate": 7.603230612404925e-05, "loss": 2.7756, "step": 30566 }, { "epoch": 2.4, "learning_rate": 7.603152199482476e-05, "loss": 2.8613, "step": 30567 }, { "epoch": 2.4, "learning_rate": 7.603073786560026e-05, "loss": 2.8425, "step": 30568 }, { "epoch": 2.4, "learning_rate": 7.602995373637576e-05, "loss": 2.8223, "step": 30569 }, { "epoch": 2.4, "learning_rate": 7.602916960715127e-05, "loss": 3.1181, "step": 30570 }, { "epoch": 2.4, "learning_rate": 7.602838547792677e-05, "loss": 2.9519, "step": 30571 }, { "epoch": 2.4, "learning_rate": 7.602760134870227e-05, "loss": 2.9838, "step": 30572 }, { "epoch": 2.4, "learning_rate": 7.602681721947778e-05, "loss": 2.782, "step": 30573 }, { "epoch": 2.4, "learning_rate": 7.602603309025328e-05, "loss": 2.7167, "step": 30574 }, { "epoch": 2.4, "learning_rate": 7.602524896102879e-05, "loss": 2.7873, "step": 30575 }, { "epoch": 2.4, "learning_rate": 7.602446483180429e-05, "loss": 2.986, "step": 30576 }, { "epoch": 2.4, "learning_rate": 7.602368070257979e-05, "loss": 2.9206, "step": 30577 }, { "epoch": 2.4, "learning_rate": 7.60228965733553e-05, "loss": 2.8877, "step": 30578 }, { "epoch": 2.4, "learning_rate": 7.60221124441308e-05, "loss": 2.7728, "step": 30579 }, { "epoch": 2.4, "learning_rate": 7.60213283149063e-05, "loss": 2.8263, "step": 30580 }, { "epoch": 2.4, "learning_rate": 7.602054418568181e-05, "loss": 2.8582, "step": 30581 }, { "epoch": 2.4, "learning_rate": 7.601976005645731e-05, "loss": 2.7666, "step": 30582 }, { "epoch": 2.4, "learning_rate": 7.601897592723282e-05, "loss": 2.6531, "step": 30583 }, { "epoch": 2.4, "learning_rate": 7.601819179800832e-05, "loss": 2.7993, "step": 30584 }, { "epoch": 2.4, "learning_rate": 7.601740766878382e-05, "loss": 2.9136, "step": 30585 }, { "epoch": 2.4, "learning_rate": 7.601662353955933e-05, "loss": 2.691, "step": 30586 }, { "epoch": 2.4, "learning_rate": 7.601583941033482e-05, "loss": 3.052, "step": 30587 }, { "epoch": 2.4, "learning_rate": 7.601505528111032e-05, "loss": 2.7253, "step": 30588 }, { "epoch": 2.4, "learning_rate": 7.601427115188582e-05, "loss": 2.6647, "step": 30589 }, { "epoch": 2.4, "learning_rate": 7.601348702266133e-05, "loss": 2.8009, "step": 30590 }, { "epoch": 2.4, "learning_rate": 7.601270289343684e-05, "loss": 2.9434, "step": 30591 }, { "epoch": 2.4, "learning_rate": 7.601191876421235e-05, "loss": 2.9216, "step": 30592 }, { "epoch": 2.4, "learning_rate": 7.601113463498785e-05, "loss": 2.9736, "step": 30593 }, { "epoch": 2.4, "learning_rate": 7.601035050576336e-05, "loss": 3.0594, "step": 30594 }, { "epoch": 2.4, "learning_rate": 7.600956637653886e-05, "loss": 2.8041, "step": 30595 }, { "epoch": 2.4, "learning_rate": 7.600878224731436e-05, "loss": 2.8097, "step": 30596 }, { "epoch": 2.4, "learning_rate": 7.600799811808987e-05, "loss": 2.6907, "step": 30597 }, { "epoch": 2.4, "learning_rate": 7.600721398886537e-05, "loss": 2.6259, "step": 30598 }, { "epoch": 2.4, "learning_rate": 7.600642985964087e-05, "loss": 2.9784, "step": 30599 }, { "epoch": 2.4, "learning_rate": 7.600564573041638e-05, "loss": 2.6711, "step": 30600 }, { "epoch": 2.4, "learning_rate": 7.600486160119188e-05, "loss": 2.7376, "step": 30601 }, { "epoch": 2.4, "learning_rate": 7.600407747196739e-05, "loss": 2.8487, "step": 30602 }, { "epoch": 2.4, "learning_rate": 7.600329334274289e-05, "loss": 2.727, "step": 30603 }, { "epoch": 2.4, "learning_rate": 7.600250921351839e-05, "loss": 2.6491, "step": 30604 }, { "epoch": 2.4, "learning_rate": 7.60017250842939e-05, "loss": 2.818, "step": 30605 }, { "epoch": 2.4, "learning_rate": 7.60009409550694e-05, "loss": 2.7637, "step": 30606 }, { "epoch": 2.4, "learning_rate": 7.60001568258449e-05, "loss": 2.5953, "step": 30607 }, { "epoch": 2.4, "learning_rate": 7.599937269662041e-05, "loss": 3.0482, "step": 30608 }, { "epoch": 2.4, "learning_rate": 7.599858856739591e-05, "loss": 2.7198, "step": 30609 }, { "epoch": 2.4, "learning_rate": 7.599780443817141e-05, "loss": 2.8965, "step": 30610 }, { "epoch": 2.4, "learning_rate": 7.599702030894692e-05, "loss": 2.9321, "step": 30611 }, { "epoch": 2.4, "learning_rate": 7.599623617972242e-05, "loss": 2.6655, "step": 30612 }, { "epoch": 2.4, "learning_rate": 7.599545205049793e-05, "loss": 2.5987, "step": 30613 }, { "epoch": 2.4, "learning_rate": 7.599466792127343e-05, "loss": 2.8521, "step": 30614 }, { "epoch": 2.4, "learning_rate": 7.599388379204893e-05, "loss": 2.8237, "step": 30615 }, { "epoch": 2.4, "learning_rate": 7.599309966282444e-05, "loss": 2.8438, "step": 30616 }, { "epoch": 2.4, "learning_rate": 7.599231553359994e-05, "loss": 2.8228, "step": 30617 }, { "epoch": 2.4, "learning_rate": 7.599153140437544e-05, "loss": 2.7434, "step": 30618 }, { "epoch": 2.4, "learning_rate": 7.599074727515095e-05, "loss": 2.9919, "step": 30619 }, { "epoch": 2.4, "learning_rate": 7.598996314592645e-05, "loss": 2.5821, "step": 30620 }, { "epoch": 2.4, "learning_rate": 7.598917901670195e-05, "loss": 2.6724, "step": 30621 }, { "epoch": 2.4, "learning_rate": 7.598839488747746e-05, "loss": 3.0987, "step": 30622 }, { "epoch": 2.4, "learning_rate": 7.598761075825296e-05, "loss": 2.7741, "step": 30623 }, { "epoch": 2.4, "learning_rate": 7.598682662902847e-05, "loss": 2.7566, "step": 30624 }, { "epoch": 2.4, "learning_rate": 7.598604249980397e-05, "loss": 2.6106, "step": 30625 }, { "epoch": 2.4, "learning_rate": 7.598525837057947e-05, "loss": 2.8137, "step": 30626 }, { "epoch": 2.4, "learning_rate": 7.598447424135498e-05, "loss": 2.5992, "step": 30627 }, { "epoch": 2.4, "learning_rate": 7.598369011213048e-05, "loss": 2.6359, "step": 30628 }, { "epoch": 2.4, "learning_rate": 7.598290598290598e-05, "loss": 2.8002, "step": 30629 }, { "epoch": 2.4, "learning_rate": 7.598212185368149e-05, "loss": 2.7119, "step": 30630 }, { "epoch": 2.4, "learning_rate": 7.598133772445699e-05, "loss": 2.7532, "step": 30631 }, { "epoch": 2.4, "learning_rate": 7.59805535952325e-05, "loss": 2.7238, "step": 30632 }, { "epoch": 2.4, "learning_rate": 7.5979769466008e-05, "loss": 2.8536, "step": 30633 }, { "epoch": 2.4, "learning_rate": 7.59789853367835e-05, "loss": 2.8175, "step": 30634 }, { "epoch": 2.4, "learning_rate": 7.5978201207559e-05, "loss": 2.8569, "step": 30635 }, { "epoch": 2.4, "learning_rate": 7.597741707833451e-05, "loss": 2.7955, "step": 30636 }, { "epoch": 2.4, "learning_rate": 7.597663294911001e-05, "loss": 2.7126, "step": 30637 }, { "epoch": 2.4, "learning_rate": 7.597584881988552e-05, "loss": 2.9597, "step": 30638 }, { "epoch": 2.4, "learning_rate": 7.597506469066102e-05, "loss": 3.054, "step": 30639 }, { "epoch": 2.4, "learning_rate": 7.597428056143652e-05, "loss": 2.7832, "step": 30640 }, { "epoch": 2.4, "learning_rate": 7.597349643221203e-05, "loss": 2.8625, "step": 30641 }, { "epoch": 2.4, "learning_rate": 7.597271230298753e-05, "loss": 2.7723, "step": 30642 }, { "epoch": 2.4, "learning_rate": 7.597192817376304e-05, "loss": 2.8721, "step": 30643 }, { "epoch": 2.4, "learning_rate": 7.597114404453854e-05, "loss": 2.905, "step": 30644 }, { "epoch": 2.4, "learning_rate": 7.597035991531404e-05, "loss": 2.6401, "step": 30645 }, { "epoch": 2.4, "learning_rate": 7.596957578608955e-05, "loss": 2.6706, "step": 30646 }, { "epoch": 2.4, "learning_rate": 7.596879165686505e-05, "loss": 2.9968, "step": 30647 }, { "epoch": 2.4, "learning_rate": 7.596800752764057e-05, "loss": 2.8811, "step": 30648 }, { "epoch": 2.4, "learning_rate": 7.596722339841607e-05, "loss": 2.8092, "step": 30649 }, { "epoch": 2.4, "learning_rate": 7.596643926919158e-05, "loss": 2.8173, "step": 30650 }, { "epoch": 2.4, "learning_rate": 7.596565513996708e-05, "loss": 2.9179, "step": 30651 }, { "epoch": 2.4, "learning_rate": 7.596487101074258e-05, "loss": 2.6734, "step": 30652 }, { "epoch": 2.4, "learning_rate": 7.596408688151809e-05, "loss": 2.7377, "step": 30653 }, { "epoch": 2.4, "learning_rate": 7.596330275229359e-05, "loss": 2.8759, "step": 30654 }, { "epoch": 2.4, "learning_rate": 7.59625186230691e-05, "loss": 2.7329, "step": 30655 }, { "epoch": 2.4, "learning_rate": 7.59617344938446e-05, "loss": 2.6547, "step": 30656 }, { "epoch": 2.4, "learning_rate": 7.59609503646201e-05, "loss": 2.7617, "step": 30657 }, { "epoch": 2.4, "learning_rate": 7.59601662353956e-05, "loss": 2.8205, "step": 30658 }, { "epoch": 2.4, "learning_rate": 7.59593821061711e-05, "loss": 2.8543, "step": 30659 }, { "epoch": 2.4, "learning_rate": 7.59585979769466e-05, "loss": 2.7657, "step": 30660 }, { "epoch": 2.4, "learning_rate": 7.59578138477221e-05, "loss": 2.8865, "step": 30661 }, { "epoch": 2.4, "learning_rate": 7.59570297184976e-05, "loss": 2.9456, "step": 30662 }, { "epoch": 2.4, "learning_rate": 7.595624558927311e-05, "loss": 2.7214, "step": 30663 }, { "epoch": 2.4, "learning_rate": 7.595546146004861e-05, "loss": 2.8763, "step": 30664 }, { "epoch": 2.4, "learning_rate": 7.595467733082412e-05, "loss": 2.6637, "step": 30665 }, { "epoch": 2.4, "learning_rate": 7.595389320159962e-05, "loss": 2.8133, "step": 30666 }, { "epoch": 2.4, "learning_rate": 7.595310907237512e-05, "loss": 2.7333, "step": 30667 }, { "epoch": 2.4, "learning_rate": 7.595232494315063e-05, "loss": 2.6935, "step": 30668 }, { "epoch": 2.4, "learning_rate": 7.595154081392613e-05, "loss": 2.9328, "step": 30669 }, { "epoch": 2.4, "learning_rate": 7.595075668470164e-05, "loss": 2.9363, "step": 30670 }, { "epoch": 2.41, "learning_rate": 7.594997255547714e-05, "loss": 2.5339, "step": 30671 }, { "epoch": 2.41, "learning_rate": 7.594918842625264e-05, "loss": 2.9336, "step": 30672 }, { "epoch": 2.41, "learning_rate": 7.594840429702815e-05, "loss": 2.6486, "step": 30673 }, { "epoch": 2.41, "learning_rate": 7.594762016780366e-05, "loss": 2.7913, "step": 30674 }, { "epoch": 2.41, "learning_rate": 7.594683603857917e-05, "loss": 2.6676, "step": 30675 }, { "epoch": 2.41, "learning_rate": 7.594605190935467e-05, "loss": 2.6643, "step": 30676 }, { "epoch": 2.41, "learning_rate": 7.594526778013017e-05, "loss": 3.0219, "step": 30677 }, { "epoch": 2.41, "learning_rate": 7.594448365090568e-05, "loss": 2.752, "step": 30678 }, { "epoch": 2.41, "learning_rate": 7.594369952168118e-05, "loss": 2.7011, "step": 30679 }, { "epoch": 2.41, "learning_rate": 7.594291539245669e-05, "loss": 2.9605, "step": 30680 }, { "epoch": 2.41, "learning_rate": 7.594213126323219e-05, "loss": 3.0185, "step": 30681 }, { "epoch": 2.41, "learning_rate": 7.594134713400769e-05, "loss": 2.8286, "step": 30682 }, { "epoch": 2.41, "learning_rate": 7.59405630047832e-05, "loss": 2.923, "step": 30683 }, { "epoch": 2.41, "learning_rate": 7.59397788755587e-05, "loss": 2.6759, "step": 30684 }, { "epoch": 2.41, "learning_rate": 7.59389947463342e-05, "loss": 2.8427, "step": 30685 }, { "epoch": 2.41, "learning_rate": 7.593821061710971e-05, "loss": 2.933, "step": 30686 }, { "epoch": 2.41, "learning_rate": 7.593742648788521e-05, "loss": 2.8192, "step": 30687 }, { "epoch": 2.41, "learning_rate": 7.593664235866071e-05, "loss": 2.8583, "step": 30688 }, { "epoch": 2.41, "learning_rate": 7.593585822943622e-05, "loss": 3.0008, "step": 30689 }, { "epoch": 2.41, "learning_rate": 7.593507410021172e-05, "loss": 2.667, "step": 30690 }, { "epoch": 2.41, "learning_rate": 7.593428997098723e-05, "loss": 2.9021, "step": 30691 }, { "epoch": 2.41, "learning_rate": 7.593350584176273e-05, "loss": 2.6591, "step": 30692 }, { "epoch": 2.41, "learning_rate": 7.593272171253823e-05, "loss": 2.7589, "step": 30693 }, { "epoch": 2.41, "learning_rate": 7.593193758331374e-05, "loss": 2.8446, "step": 30694 }, { "epoch": 2.41, "learning_rate": 7.593115345408923e-05, "loss": 2.7933, "step": 30695 }, { "epoch": 2.41, "learning_rate": 7.593036932486473e-05, "loss": 2.7388, "step": 30696 }, { "epoch": 2.41, "learning_rate": 7.592958519564023e-05, "loss": 2.8149, "step": 30697 }, { "epoch": 2.41, "learning_rate": 7.592880106641574e-05, "loss": 2.6937, "step": 30698 }, { "epoch": 2.41, "learning_rate": 7.592801693719126e-05, "loss": 2.8301, "step": 30699 }, { "epoch": 2.41, "learning_rate": 7.592723280796676e-05, "loss": 2.7914, "step": 30700 }, { "epoch": 2.41, "learning_rate": 7.592644867874226e-05, "loss": 2.615, "step": 30701 }, { "epoch": 2.41, "learning_rate": 7.592566454951777e-05, "loss": 2.7739, "step": 30702 }, { "epoch": 2.41, "learning_rate": 7.592488042029327e-05, "loss": 2.7384, "step": 30703 }, { "epoch": 2.41, "learning_rate": 7.592409629106877e-05, "loss": 2.6995, "step": 30704 }, { "epoch": 2.41, "learning_rate": 7.592331216184428e-05, "loss": 2.655, "step": 30705 }, { "epoch": 2.41, "learning_rate": 7.592252803261978e-05, "loss": 2.9441, "step": 30706 }, { "epoch": 2.41, "learning_rate": 7.592174390339528e-05, "loss": 2.7252, "step": 30707 }, { "epoch": 2.41, "learning_rate": 7.592095977417079e-05, "loss": 3.067, "step": 30708 }, { "epoch": 2.41, "learning_rate": 7.592017564494629e-05, "loss": 2.8685, "step": 30709 }, { "epoch": 2.41, "learning_rate": 7.59193915157218e-05, "loss": 2.9258, "step": 30710 }, { "epoch": 2.41, "learning_rate": 7.59186073864973e-05, "loss": 2.8201, "step": 30711 }, { "epoch": 2.41, "learning_rate": 7.59178232572728e-05, "loss": 2.779, "step": 30712 }, { "epoch": 2.41, "learning_rate": 7.591703912804831e-05, "loss": 2.7953, "step": 30713 }, { "epoch": 2.41, "learning_rate": 7.591625499882381e-05, "loss": 2.8472, "step": 30714 }, { "epoch": 2.41, "learning_rate": 7.591547086959931e-05, "loss": 2.6953, "step": 30715 }, { "epoch": 2.41, "learning_rate": 7.591468674037482e-05, "loss": 2.5992, "step": 30716 }, { "epoch": 2.41, "learning_rate": 7.591390261115032e-05, "loss": 2.8455, "step": 30717 }, { "epoch": 2.41, "learning_rate": 7.591311848192583e-05, "loss": 2.7877, "step": 30718 }, { "epoch": 2.41, "learning_rate": 7.591233435270133e-05, "loss": 2.8857, "step": 30719 }, { "epoch": 2.41, "learning_rate": 7.591155022347683e-05, "loss": 2.8754, "step": 30720 }, { "epoch": 2.41, "learning_rate": 7.591076609425234e-05, "loss": 2.7315, "step": 30721 }, { "epoch": 2.41, "learning_rate": 7.590998196502784e-05, "loss": 3.1137, "step": 30722 }, { "epoch": 2.41, "learning_rate": 7.590919783580334e-05, "loss": 2.8482, "step": 30723 }, { "epoch": 2.41, "learning_rate": 7.590841370657885e-05, "loss": 2.791, "step": 30724 }, { "epoch": 2.41, "learning_rate": 7.590762957735435e-05, "loss": 2.7721, "step": 30725 }, { "epoch": 2.41, "learning_rate": 7.590684544812985e-05, "loss": 2.5783, "step": 30726 }, { "epoch": 2.41, "learning_rate": 7.590606131890536e-05, "loss": 2.6244, "step": 30727 }, { "epoch": 2.41, "learning_rate": 7.590527718968086e-05, "loss": 2.9861, "step": 30728 }, { "epoch": 2.41, "learning_rate": 7.590449306045637e-05, "loss": 2.5485, "step": 30729 }, { "epoch": 2.41, "learning_rate": 7.590370893123187e-05, "loss": 2.8041, "step": 30730 }, { "epoch": 2.41, "learning_rate": 7.590292480200737e-05, "loss": 2.9772, "step": 30731 }, { "epoch": 2.41, "learning_rate": 7.590214067278288e-05, "loss": 2.7146, "step": 30732 }, { "epoch": 2.41, "learning_rate": 7.590135654355838e-05, "loss": 2.9345, "step": 30733 }, { "epoch": 2.41, "learning_rate": 7.590057241433388e-05, "loss": 2.6341, "step": 30734 }, { "epoch": 2.41, "learning_rate": 7.589978828510939e-05, "loss": 2.8454, "step": 30735 }, { "epoch": 2.41, "learning_rate": 7.589900415588489e-05, "loss": 2.8653, "step": 30736 }, { "epoch": 2.41, "learning_rate": 7.58982200266604e-05, "loss": 2.5398, "step": 30737 }, { "epoch": 2.41, "learning_rate": 7.58974358974359e-05, "loss": 2.7316, "step": 30738 }, { "epoch": 2.41, "learning_rate": 7.58966517682114e-05, "loss": 3.0301, "step": 30739 }, { "epoch": 2.41, "learning_rate": 7.58958676389869e-05, "loss": 2.6611, "step": 30740 }, { "epoch": 2.41, "learning_rate": 7.589508350976241e-05, "loss": 2.9116, "step": 30741 }, { "epoch": 2.41, "learning_rate": 7.589429938053791e-05, "loss": 2.706, "step": 30742 }, { "epoch": 2.41, "learning_rate": 7.589351525131342e-05, "loss": 2.8504, "step": 30743 }, { "epoch": 2.41, "learning_rate": 7.589273112208892e-05, "loss": 2.9487, "step": 30744 }, { "epoch": 2.41, "learning_rate": 7.589194699286442e-05, "loss": 3.1351, "step": 30745 }, { "epoch": 2.41, "learning_rate": 7.589116286363993e-05, "loss": 2.9108, "step": 30746 }, { "epoch": 2.41, "learning_rate": 7.589037873441543e-05, "loss": 2.6675, "step": 30747 }, { "epoch": 2.41, "learning_rate": 7.588959460519094e-05, "loss": 2.7651, "step": 30748 }, { "epoch": 2.41, "learning_rate": 7.588881047596644e-05, "loss": 2.7374, "step": 30749 }, { "epoch": 2.41, "learning_rate": 7.588802634674194e-05, "loss": 2.925, "step": 30750 }, { "epoch": 2.41, "learning_rate": 7.588724221751745e-05, "loss": 2.868, "step": 30751 }, { "epoch": 2.41, "learning_rate": 7.588645808829295e-05, "loss": 2.8485, "step": 30752 }, { "epoch": 2.41, "learning_rate": 7.588567395906845e-05, "loss": 2.9986, "step": 30753 }, { "epoch": 2.41, "learning_rate": 7.588488982984396e-05, "loss": 2.8532, "step": 30754 }, { "epoch": 2.41, "learning_rate": 7.588410570061946e-05, "loss": 2.7801, "step": 30755 }, { "epoch": 2.41, "learning_rate": 7.588332157139498e-05, "loss": 2.8136, "step": 30756 }, { "epoch": 2.41, "learning_rate": 7.588253744217048e-05, "loss": 3.0964, "step": 30757 }, { "epoch": 2.41, "learning_rate": 7.588175331294599e-05, "loss": 2.6622, "step": 30758 }, { "epoch": 2.41, "learning_rate": 7.588096918372149e-05, "loss": 2.9023, "step": 30759 }, { "epoch": 2.41, "learning_rate": 7.5880185054497e-05, "loss": 3.077, "step": 30760 }, { "epoch": 2.41, "learning_rate": 7.58794009252725e-05, "loss": 2.7646, "step": 30761 }, { "epoch": 2.41, "learning_rate": 7.5878616796048e-05, "loss": 2.8343, "step": 30762 }, { "epoch": 2.41, "learning_rate": 7.58778326668235e-05, "loss": 2.9615, "step": 30763 }, { "epoch": 2.41, "learning_rate": 7.587704853759901e-05, "loss": 2.6241, "step": 30764 }, { "epoch": 2.41, "learning_rate": 7.587626440837451e-05, "loss": 2.6596, "step": 30765 }, { "epoch": 2.41, "learning_rate": 7.587548027915002e-05, "loss": 2.7208, "step": 30766 }, { "epoch": 2.41, "learning_rate": 7.58746961499255e-05, "loss": 2.6515, "step": 30767 }, { "epoch": 2.41, "learning_rate": 7.587391202070101e-05, "loss": 2.8436, "step": 30768 }, { "epoch": 2.41, "learning_rate": 7.587312789147651e-05, "loss": 2.8852, "step": 30769 }, { "epoch": 2.41, "learning_rate": 7.587234376225202e-05, "loss": 2.7851, "step": 30770 }, { "epoch": 2.41, "learning_rate": 7.587155963302752e-05, "loss": 2.8115, "step": 30771 }, { "epoch": 2.41, "learning_rate": 7.587077550380302e-05, "loss": 2.9288, "step": 30772 }, { "epoch": 2.41, "learning_rate": 7.586999137457853e-05, "loss": 2.9039, "step": 30773 }, { "epoch": 2.41, "learning_rate": 7.586920724535403e-05, "loss": 2.8073, "step": 30774 }, { "epoch": 2.41, "learning_rate": 7.586842311612953e-05, "loss": 2.7092, "step": 30775 }, { "epoch": 2.41, "learning_rate": 7.586763898690504e-05, "loss": 2.8295, "step": 30776 }, { "epoch": 2.41, "learning_rate": 7.586685485768054e-05, "loss": 2.8571, "step": 30777 }, { "epoch": 2.41, "learning_rate": 7.586607072845605e-05, "loss": 2.7349, "step": 30778 }, { "epoch": 2.41, "learning_rate": 7.586528659923155e-05, "loss": 2.9488, "step": 30779 }, { "epoch": 2.41, "learning_rate": 7.586450247000705e-05, "loss": 2.9031, "step": 30780 }, { "epoch": 2.41, "learning_rate": 7.586371834078256e-05, "loss": 2.758, "step": 30781 }, { "epoch": 2.41, "learning_rate": 7.586293421155807e-05, "loss": 2.9164, "step": 30782 }, { "epoch": 2.41, "learning_rate": 7.586215008233358e-05, "loss": 2.6878, "step": 30783 }, { "epoch": 2.41, "learning_rate": 7.586136595310908e-05, "loss": 2.5123, "step": 30784 }, { "epoch": 2.41, "learning_rate": 7.586058182388459e-05, "loss": 2.7981, "step": 30785 }, { "epoch": 2.41, "learning_rate": 7.585979769466009e-05, "loss": 3.0619, "step": 30786 }, { "epoch": 2.41, "learning_rate": 7.585901356543559e-05, "loss": 2.794, "step": 30787 }, { "epoch": 2.41, "learning_rate": 7.58582294362111e-05, "loss": 3.0899, "step": 30788 }, { "epoch": 2.41, "learning_rate": 7.58574453069866e-05, "loss": 2.8698, "step": 30789 }, { "epoch": 2.41, "learning_rate": 7.58566611777621e-05, "loss": 2.6235, "step": 30790 }, { "epoch": 2.41, "learning_rate": 7.585587704853761e-05, "loss": 2.9968, "step": 30791 }, { "epoch": 2.41, "learning_rate": 7.585509291931311e-05, "loss": 3.043, "step": 30792 }, { "epoch": 2.41, "learning_rate": 7.585430879008861e-05, "loss": 2.8482, "step": 30793 }, { "epoch": 2.41, "learning_rate": 7.585352466086412e-05, "loss": 2.8204, "step": 30794 }, { "epoch": 2.41, "learning_rate": 7.585274053163962e-05, "loss": 2.6821, "step": 30795 }, { "epoch": 2.41, "learning_rate": 7.585195640241513e-05, "loss": 2.8079, "step": 30796 }, { "epoch": 2.41, "learning_rate": 7.585117227319063e-05, "loss": 2.7451, "step": 30797 }, { "epoch": 2.41, "learning_rate": 7.585038814396613e-05, "loss": 2.7581, "step": 30798 }, { "epoch": 2.42, "learning_rate": 7.584960401474164e-05, "loss": 2.7973, "step": 30799 }, { "epoch": 2.42, "learning_rate": 7.584881988551714e-05, "loss": 2.7198, "step": 30800 }, { "epoch": 2.42, "learning_rate": 7.584803575629264e-05, "loss": 2.8893, "step": 30801 }, { "epoch": 2.42, "learning_rate": 7.584725162706815e-05, "loss": 2.9507, "step": 30802 }, { "epoch": 2.42, "learning_rate": 7.584646749784364e-05, "loss": 2.9315, "step": 30803 }, { "epoch": 2.42, "learning_rate": 7.584568336861914e-05, "loss": 2.7827, "step": 30804 }, { "epoch": 2.42, "learning_rate": 7.584489923939465e-05, "loss": 2.9569, "step": 30805 }, { "epoch": 2.42, "learning_rate": 7.584411511017015e-05, "loss": 2.8763, "step": 30806 }, { "epoch": 2.42, "learning_rate": 7.584333098094567e-05, "loss": 2.8283, "step": 30807 }, { "epoch": 2.42, "learning_rate": 7.584254685172117e-05, "loss": 2.9362, "step": 30808 }, { "epoch": 2.42, "learning_rate": 7.584176272249667e-05, "loss": 2.7387, "step": 30809 }, { "epoch": 2.42, "learning_rate": 7.584097859327218e-05, "loss": 2.6761, "step": 30810 }, { "epoch": 2.42, "learning_rate": 7.584019446404768e-05, "loss": 2.7554, "step": 30811 }, { "epoch": 2.42, "learning_rate": 7.583941033482318e-05, "loss": 2.7241, "step": 30812 }, { "epoch": 2.42, "learning_rate": 7.583862620559869e-05, "loss": 2.8145, "step": 30813 }, { "epoch": 2.42, "learning_rate": 7.583784207637419e-05, "loss": 2.751, "step": 30814 }, { "epoch": 2.42, "learning_rate": 7.58370579471497e-05, "loss": 2.9869, "step": 30815 }, { "epoch": 2.42, "learning_rate": 7.58362738179252e-05, "loss": 2.8991, "step": 30816 }, { "epoch": 2.42, "learning_rate": 7.58354896887007e-05, "loss": 2.6938, "step": 30817 }, { "epoch": 2.42, "learning_rate": 7.58347055594762e-05, "loss": 2.9288, "step": 30818 }, { "epoch": 2.42, "learning_rate": 7.583392143025171e-05, "loss": 2.7687, "step": 30819 }, { "epoch": 2.42, "learning_rate": 7.583313730102721e-05, "loss": 2.9367, "step": 30820 }, { "epoch": 2.42, "learning_rate": 7.583235317180272e-05, "loss": 2.5799, "step": 30821 }, { "epoch": 2.42, "learning_rate": 7.583156904257822e-05, "loss": 2.7167, "step": 30822 }, { "epoch": 2.42, "learning_rate": 7.583078491335372e-05, "loss": 2.7835, "step": 30823 }, { "epoch": 2.42, "learning_rate": 7.583000078412923e-05, "loss": 2.8192, "step": 30824 }, { "epoch": 2.42, "learning_rate": 7.582921665490473e-05, "loss": 2.7324, "step": 30825 }, { "epoch": 2.42, "learning_rate": 7.582843252568024e-05, "loss": 2.6709, "step": 30826 }, { "epoch": 2.42, "learning_rate": 7.582764839645574e-05, "loss": 2.9241, "step": 30827 }, { "epoch": 2.42, "learning_rate": 7.582686426723124e-05, "loss": 2.8022, "step": 30828 }, { "epoch": 2.42, "learning_rate": 7.582608013800675e-05, "loss": 2.7341, "step": 30829 }, { "epoch": 2.42, "learning_rate": 7.582529600878225e-05, "loss": 2.7214, "step": 30830 }, { "epoch": 2.42, "learning_rate": 7.582451187955775e-05, "loss": 2.6955, "step": 30831 }, { "epoch": 2.42, "learning_rate": 7.582372775033326e-05, "loss": 2.7774, "step": 30832 }, { "epoch": 2.42, "learning_rate": 7.582294362110876e-05, "loss": 2.8947, "step": 30833 }, { "epoch": 2.42, "learning_rate": 7.582215949188427e-05, "loss": 2.6258, "step": 30834 }, { "epoch": 2.42, "learning_rate": 7.582137536265977e-05, "loss": 2.6856, "step": 30835 }, { "epoch": 2.42, "learning_rate": 7.582059123343527e-05, "loss": 2.8691, "step": 30836 }, { "epoch": 2.42, "learning_rate": 7.581980710421078e-05, "loss": 2.7944, "step": 30837 }, { "epoch": 2.42, "learning_rate": 7.581902297498628e-05, "loss": 2.77, "step": 30838 }, { "epoch": 2.42, "learning_rate": 7.581823884576178e-05, "loss": 2.8869, "step": 30839 }, { "epoch": 2.42, "learning_rate": 7.581745471653729e-05, "loss": 2.7007, "step": 30840 }, { "epoch": 2.42, "learning_rate": 7.581667058731279e-05, "loss": 2.751, "step": 30841 }, { "epoch": 2.42, "learning_rate": 7.58158864580883e-05, "loss": 3.0181, "step": 30842 }, { "epoch": 2.42, "learning_rate": 7.58151023288638e-05, "loss": 2.8264, "step": 30843 }, { "epoch": 2.42, "learning_rate": 7.58143181996393e-05, "loss": 2.8079, "step": 30844 }, { "epoch": 2.42, "learning_rate": 7.58135340704148e-05, "loss": 2.6124, "step": 30845 }, { "epoch": 2.42, "learning_rate": 7.581274994119031e-05, "loss": 3.1844, "step": 30846 }, { "epoch": 2.42, "learning_rate": 7.581196581196581e-05, "loss": 2.7553, "step": 30847 }, { "epoch": 2.42, "learning_rate": 7.581118168274132e-05, "loss": 2.9367, "step": 30848 }, { "epoch": 2.42, "learning_rate": 7.581039755351682e-05, "loss": 2.7773, "step": 30849 }, { "epoch": 2.42, "learning_rate": 7.580961342429232e-05, "loss": 2.6803, "step": 30850 }, { "epoch": 2.42, "learning_rate": 7.580882929506783e-05, "loss": 2.8167, "step": 30851 }, { "epoch": 2.42, "learning_rate": 7.580804516584333e-05, "loss": 2.8093, "step": 30852 }, { "epoch": 2.42, "learning_rate": 7.580726103661884e-05, "loss": 2.8235, "step": 30853 }, { "epoch": 2.42, "learning_rate": 7.580647690739434e-05, "loss": 2.7748, "step": 30854 }, { "epoch": 2.42, "learning_rate": 7.580569277816984e-05, "loss": 2.6726, "step": 30855 }, { "epoch": 2.42, "learning_rate": 7.580490864894535e-05, "loss": 2.8773, "step": 30856 }, { "epoch": 2.42, "learning_rate": 7.580412451972085e-05, "loss": 2.8439, "step": 30857 }, { "epoch": 2.42, "learning_rate": 7.580334039049635e-05, "loss": 2.553, "step": 30858 }, { "epoch": 2.42, "learning_rate": 7.580255626127186e-05, "loss": 2.6654, "step": 30859 }, { "epoch": 2.42, "learning_rate": 7.580177213204736e-05, "loss": 2.8254, "step": 30860 }, { "epoch": 2.42, "learning_rate": 7.580098800282286e-05, "loss": 2.8418, "step": 30861 }, { "epoch": 2.42, "learning_rate": 7.580020387359837e-05, "loss": 2.7919, "step": 30862 }, { "epoch": 2.42, "learning_rate": 7.579941974437387e-05, "loss": 2.7344, "step": 30863 }, { "epoch": 2.42, "learning_rate": 7.579863561514939e-05, "loss": 2.5653, "step": 30864 }, { "epoch": 2.42, "learning_rate": 7.579785148592489e-05, "loss": 2.8646, "step": 30865 }, { "epoch": 2.42, "learning_rate": 7.57970673567004e-05, "loss": 2.9199, "step": 30866 }, { "epoch": 2.42, "learning_rate": 7.57962832274759e-05, "loss": 2.8238, "step": 30867 }, { "epoch": 2.42, "learning_rate": 7.57954990982514e-05, "loss": 2.7988, "step": 30868 }, { "epoch": 2.42, "learning_rate": 7.579471496902691e-05, "loss": 2.5645, "step": 30869 }, { "epoch": 2.42, "learning_rate": 7.579393083980241e-05, "loss": 3.0101, "step": 30870 }, { "epoch": 2.42, "learning_rate": 7.579314671057792e-05, "loss": 2.6882, "step": 30871 }, { "epoch": 2.42, "learning_rate": 7.579236258135342e-05, "loss": 2.7117, "step": 30872 }, { "epoch": 2.42, "learning_rate": 7.579157845212892e-05, "loss": 2.6841, "step": 30873 }, { "epoch": 2.42, "learning_rate": 7.579079432290443e-05, "loss": 2.64, "step": 30874 }, { "epoch": 2.42, "learning_rate": 7.579001019367992e-05, "loss": 2.9872, "step": 30875 }, { "epoch": 2.42, "learning_rate": 7.578922606445542e-05, "loss": 2.831, "step": 30876 }, { "epoch": 2.42, "learning_rate": 7.578844193523092e-05, "loss": 2.7502, "step": 30877 }, { "epoch": 2.42, "learning_rate": 7.578765780600643e-05, "loss": 2.8043, "step": 30878 }, { "epoch": 2.42, "learning_rate": 7.578687367678193e-05, "loss": 2.8448, "step": 30879 }, { "epoch": 2.42, "learning_rate": 7.578608954755743e-05, "loss": 2.8544, "step": 30880 }, { "epoch": 2.42, "learning_rate": 7.578530541833294e-05, "loss": 2.6736, "step": 30881 }, { "epoch": 2.42, "learning_rate": 7.578452128910844e-05, "loss": 2.6589, "step": 30882 }, { "epoch": 2.42, "learning_rate": 7.578373715988395e-05, "loss": 2.5995, "step": 30883 }, { "epoch": 2.42, "learning_rate": 7.578295303065945e-05, "loss": 2.9306, "step": 30884 }, { "epoch": 2.42, "learning_rate": 7.578216890143495e-05, "loss": 2.7284, "step": 30885 }, { "epoch": 2.42, "learning_rate": 7.578138477221046e-05, "loss": 2.7485, "step": 30886 }, { "epoch": 2.42, "learning_rate": 7.578060064298596e-05, "loss": 2.9172, "step": 30887 }, { "epoch": 2.42, "learning_rate": 7.577981651376146e-05, "loss": 2.7306, "step": 30888 }, { "epoch": 2.42, "learning_rate": 7.577903238453697e-05, "loss": 2.9967, "step": 30889 }, { "epoch": 2.42, "learning_rate": 7.577824825531248e-05, "loss": 2.7661, "step": 30890 }, { "epoch": 2.42, "learning_rate": 7.577746412608799e-05, "loss": 2.537, "step": 30891 }, { "epoch": 2.42, "learning_rate": 7.577667999686349e-05, "loss": 2.7796, "step": 30892 }, { "epoch": 2.42, "learning_rate": 7.5775895867639e-05, "loss": 2.7686, "step": 30893 }, { "epoch": 2.42, "learning_rate": 7.57751117384145e-05, "loss": 2.8066, "step": 30894 }, { "epoch": 2.42, "learning_rate": 7.577432760919e-05, "loss": 2.7836, "step": 30895 }, { "epoch": 2.42, "learning_rate": 7.577354347996551e-05, "loss": 2.7576, "step": 30896 }, { "epoch": 2.42, "learning_rate": 7.577275935074101e-05, "loss": 2.8056, "step": 30897 }, { "epoch": 2.42, "learning_rate": 7.577197522151651e-05, "loss": 2.751, "step": 30898 }, { "epoch": 2.42, "learning_rate": 7.577119109229202e-05, "loss": 3.0881, "step": 30899 }, { "epoch": 2.42, "learning_rate": 7.577040696306752e-05, "loss": 3.0836, "step": 30900 }, { "epoch": 2.42, "learning_rate": 7.576962283384303e-05, "loss": 2.8995, "step": 30901 }, { "epoch": 2.42, "learning_rate": 7.576883870461853e-05, "loss": 2.6784, "step": 30902 }, { "epoch": 2.42, "learning_rate": 7.576805457539403e-05, "loss": 2.8123, "step": 30903 }, { "epoch": 2.42, "learning_rate": 7.576727044616954e-05, "loss": 2.9801, "step": 30904 }, { "epoch": 2.42, "learning_rate": 7.576648631694504e-05, "loss": 2.9646, "step": 30905 }, { "epoch": 2.42, "learning_rate": 7.576570218772054e-05, "loss": 2.7627, "step": 30906 }, { "epoch": 2.42, "learning_rate": 7.576491805849605e-05, "loss": 2.8696, "step": 30907 }, { "epoch": 2.42, "learning_rate": 7.576413392927155e-05, "loss": 2.8471, "step": 30908 }, { "epoch": 2.42, "learning_rate": 7.576334980004705e-05, "loss": 2.8805, "step": 30909 }, { "epoch": 2.42, "learning_rate": 7.576256567082256e-05, "loss": 2.8983, "step": 30910 }, { "epoch": 2.42, "learning_rate": 7.576178154159805e-05, "loss": 2.8225, "step": 30911 }, { "epoch": 2.42, "learning_rate": 7.576099741237355e-05, "loss": 2.7382, "step": 30912 }, { "epoch": 2.42, "learning_rate": 7.576021328314906e-05, "loss": 2.8169, "step": 30913 }, { "epoch": 2.42, "learning_rate": 7.575942915392456e-05, "loss": 2.9124, "step": 30914 }, { "epoch": 2.42, "learning_rate": 7.575864502470008e-05, "loss": 2.869, "step": 30915 }, { "epoch": 2.42, "learning_rate": 7.575786089547558e-05, "loss": 2.8168, "step": 30916 }, { "epoch": 2.42, "learning_rate": 7.575707676625108e-05, "loss": 2.595, "step": 30917 }, { "epoch": 2.42, "learning_rate": 7.575629263702659e-05, "loss": 2.8924, "step": 30918 }, { "epoch": 2.42, "learning_rate": 7.575550850780209e-05, "loss": 2.9067, "step": 30919 }, { "epoch": 2.42, "learning_rate": 7.57547243785776e-05, "loss": 2.809, "step": 30920 }, { "epoch": 2.42, "learning_rate": 7.57539402493531e-05, "loss": 2.8923, "step": 30921 }, { "epoch": 2.42, "learning_rate": 7.57531561201286e-05, "loss": 2.7741, "step": 30922 }, { "epoch": 2.42, "learning_rate": 7.57523719909041e-05, "loss": 2.7028, "step": 30923 }, { "epoch": 2.42, "learning_rate": 7.575158786167961e-05, "loss": 2.875, "step": 30924 }, { "epoch": 2.42, "learning_rate": 7.575080373245511e-05, "loss": 2.7251, "step": 30925 }, { "epoch": 2.42, "learning_rate": 7.575001960323062e-05, "loss": 2.5733, "step": 30926 }, { "epoch": 2.43, "learning_rate": 7.574923547400612e-05, "loss": 2.763, "step": 30927 }, { "epoch": 2.43, "learning_rate": 7.574845134478162e-05, "loss": 2.8529, "step": 30928 }, { "epoch": 2.43, "learning_rate": 7.574766721555713e-05, "loss": 2.9007, "step": 30929 }, { "epoch": 2.43, "learning_rate": 7.574688308633263e-05, "loss": 2.9708, "step": 30930 }, { "epoch": 2.43, "learning_rate": 7.574609895710814e-05, "loss": 2.8708, "step": 30931 }, { "epoch": 2.43, "learning_rate": 7.574531482788364e-05, "loss": 2.8835, "step": 30932 }, { "epoch": 2.43, "learning_rate": 7.574453069865914e-05, "loss": 2.931, "step": 30933 }, { "epoch": 2.43, "learning_rate": 7.574374656943465e-05, "loss": 2.8156, "step": 30934 }, { "epoch": 2.43, "learning_rate": 7.574296244021015e-05, "loss": 2.8919, "step": 30935 }, { "epoch": 2.43, "learning_rate": 7.574217831098565e-05, "loss": 2.7411, "step": 30936 }, { "epoch": 2.43, "learning_rate": 7.574139418176116e-05, "loss": 2.8484, "step": 30937 }, { "epoch": 2.43, "learning_rate": 7.574061005253666e-05, "loss": 2.6428, "step": 30938 }, { "epoch": 2.43, "learning_rate": 7.573982592331217e-05, "loss": 2.8691, "step": 30939 }, { "epoch": 2.43, "learning_rate": 7.573904179408767e-05, "loss": 2.9117, "step": 30940 }, { "epoch": 2.43, "learning_rate": 7.573825766486317e-05, "loss": 2.7521, "step": 30941 }, { "epoch": 2.43, "learning_rate": 7.573747353563868e-05, "loss": 2.7624, "step": 30942 }, { "epoch": 2.43, "learning_rate": 7.573668940641418e-05, "loss": 2.7341, "step": 30943 }, { "epoch": 2.43, "learning_rate": 7.573590527718968e-05, "loss": 2.6735, "step": 30944 }, { "epoch": 2.43, "learning_rate": 7.573512114796519e-05, "loss": 2.7805, "step": 30945 }, { "epoch": 2.43, "learning_rate": 7.573433701874069e-05, "loss": 2.7287, "step": 30946 }, { "epoch": 2.43, "learning_rate": 7.57335528895162e-05, "loss": 2.6349, "step": 30947 }, { "epoch": 2.43, "learning_rate": 7.57327687602917e-05, "loss": 2.6584, "step": 30948 }, { "epoch": 2.43, "learning_rate": 7.57319846310672e-05, "loss": 2.7488, "step": 30949 }, { "epoch": 2.43, "learning_rate": 7.57312005018427e-05, "loss": 2.9715, "step": 30950 }, { "epoch": 2.43, "learning_rate": 7.573041637261821e-05, "loss": 2.8052, "step": 30951 }, { "epoch": 2.43, "learning_rate": 7.572963224339371e-05, "loss": 2.8674, "step": 30952 }, { "epoch": 2.43, "learning_rate": 7.572884811416922e-05, "loss": 2.9242, "step": 30953 }, { "epoch": 2.43, "learning_rate": 7.572806398494472e-05, "loss": 2.869, "step": 30954 }, { "epoch": 2.43, "learning_rate": 7.572727985572022e-05, "loss": 2.8009, "step": 30955 }, { "epoch": 2.43, "learning_rate": 7.572649572649573e-05, "loss": 2.7129, "step": 30956 }, { "epoch": 2.43, "learning_rate": 7.572571159727123e-05, "loss": 2.7451, "step": 30957 }, { "epoch": 2.43, "learning_rate": 7.572492746804673e-05, "loss": 3.0381, "step": 30958 }, { "epoch": 2.43, "learning_rate": 7.572414333882224e-05, "loss": 2.7301, "step": 30959 }, { "epoch": 2.43, "learning_rate": 7.572335920959774e-05, "loss": 2.8122, "step": 30960 }, { "epoch": 2.43, "learning_rate": 7.572257508037325e-05, "loss": 2.7641, "step": 30961 }, { "epoch": 2.43, "learning_rate": 7.572179095114875e-05, "loss": 2.8289, "step": 30962 }, { "epoch": 2.43, "learning_rate": 7.572100682192425e-05, "loss": 2.73, "step": 30963 }, { "epoch": 2.43, "learning_rate": 7.572022269269976e-05, "loss": 2.7603, "step": 30964 }, { "epoch": 2.43, "learning_rate": 7.571943856347526e-05, "loss": 2.8825, "step": 30965 }, { "epoch": 2.43, "learning_rate": 7.571865443425076e-05, "loss": 2.8992, "step": 30966 }, { "epoch": 2.43, "learning_rate": 7.571787030502627e-05, "loss": 2.7924, "step": 30967 }, { "epoch": 2.43, "learning_rate": 7.571708617580177e-05, "loss": 2.6794, "step": 30968 }, { "epoch": 2.43, "learning_rate": 7.571630204657728e-05, "loss": 2.8671, "step": 30969 }, { "epoch": 2.43, "learning_rate": 7.571551791735278e-05, "loss": 3.1492, "step": 30970 }, { "epoch": 2.43, "learning_rate": 7.571473378812828e-05, "loss": 3.1648, "step": 30971 }, { "epoch": 2.43, "learning_rate": 7.57139496589038e-05, "loss": 2.7229, "step": 30972 }, { "epoch": 2.43, "learning_rate": 7.57131655296793e-05, "loss": 2.8284, "step": 30973 }, { "epoch": 2.43, "learning_rate": 7.571238140045481e-05, "loss": 2.8473, "step": 30974 }, { "epoch": 2.43, "learning_rate": 7.571159727123031e-05, "loss": 2.7037, "step": 30975 }, { "epoch": 2.43, "learning_rate": 7.571081314200581e-05, "loss": 2.8524, "step": 30976 }, { "epoch": 2.43, "learning_rate": 7.571002901278132e-05, "loss": 2.854, "step": 30977 }, { "epoch": 2.43, "learning_rate": 7.570924488355682e-05, "loss": 2.5591, "step": 30978 }, { "epoch": 2.43, "learning_rate": 7.570846075433233e-05, "loss": 2.7935, "step": 30979 }, { "epoch": 2.43, "learning_rate": 7.570767662510783e-05, "loss": 2.7533, "step": 30980 }, { "epoch": 2.43, "learning_rate": 7.570689249588333e-05, "loss": 2.833, "step": 30981 }, { "epoch": 2.43, "learning_rate": 7.570610836665884e-05, "loss": 3.0698, "step": 30982 }, { "epoch": 2.43, "learning_rate": 7.570532423743433e-05, "loss": 2.8496, "step": 30983 }, { "epoch": 2.43, "learning_rate": 7.570454010820983e-05, "loss": 2.7949, "step": 30984 }, { "epoch": 2.43, "learning_rate": 7.570375597898533e-05, "loss": 2.9008, "step": 30985 }, { "epoch": 2.43, "learning_rate": 7.570297184976084e-05, "loss": 2.7512, "step": 30986 }, { "epoch": 2.43, "learning_rate": 7.570218772053634e-05, "loss": 2.7153, "step": 30987 }, { "epoch": 2.43, "learning_rate": 7.570140359131185e-05, "loss": 2.9182, "step": 30988 }, { "epoch": 2.43, "learning_rate": 7.570061946208735e-05, "loss": 2.9479, "step": 30989 }, { "epoch": 2.43, "learning_rate": 7.569983533286285e-05, "loss": 2.898, "step": 30990 }, { "epoch": 2.43, "learning_rate": 7.569905120363836e-05, "loss": 2.8288, "step": 30991 }, { "epoch": 2.43, "learning_rate": 7.569826707441386e-05, "loss": 2.7815, "step": 30992 }, { "epoch": 2.43, "learning_rate": 7.569748294518936e-05, "loss": 2.7653, "step": 30993 }, { "epoch": 2.43, "learning_rate": 7.569669881596487e-05, "loss": 2.7918, "step": 30994 }, { "epoch": 2.43, "learning_rate": 7.569591468674037e-05, "loss": 2.6929, "step": 30995 }, { "epoch": 2.43, "learning_rate": 7.569513055751587e-05, "loss": 2.7257, "step": 30996 }, { "epoch": 2.43, "learning_rate": 7.569434642829138e-05, "loss": 2.7252, "step": 30997 }, { "epoch": 2.43, "learning_rate": 7.56935622990669e-05, "loss": 2.7251, "step": 30998 }, { "epoch": 2.43, "learning_rate": 7.56927781698424e-05, "loss": 2.795, "step": 30999 }, { "epoch": 2.43, "learning_rate": 7.56919940406179e-05, "loss": 2.8167, "step": 31000 }, { "epoch": 2.43, "eval_loss": 2.900953769683838, "eval_runtime": 218.4466, "eval_samples_per_second": 51.875, "eval_steps_per_second": 3.246, "step": 31000 }, { "epoch": 2.43, "learning_rate": 7.56912099113934e-05, "loss": 2.8363, "step": 31001 }, { "epoch": 2.43, "learning_rate": 7.569042578216891e-05, "loss": 2.7294, "step": 31002 }, { "epoch": 2.43, "learning_rate": 7.568964165294441e-05, "loss": 2.861, "step": 31003 }, { "epoch": 2.43, "learning_rate": 7.568885752371992e-05, "loss": 2.8825, "step": 31004 }, { "epoch": 2.43, "learning_rate": 7.568807339449542e-05, "loss": 2.9705, "step": 31005 }, { "epoch": 2.43, "learning_rate": 7.568728926527093e-05, "loss": 2.7758, "step": 31006 }, { "epoch": 2.43, "learning_rate": 7.568650513604643e-05, "loss": 2.7877, "step": 31007 }, { "epoch": 2.43, "learning_rate": 7.568572100682193e-05, "loss": 2.8346, "step": 31008 }, { "epoch": 2.43, "learning_rate": 7.568493687759744e-05, "loss": 2.5734, "step": 31009 }, { "epoch": 2.43, "learning_rate": 7.568415274837294e-05, "loss": 2.7825, "step": 31010 }, { "epoch": 2.43, "learning_rate": 7.568336861914844e-05, "loss": 2.9726, "step": 31011 }, { "epoch": 2.43, "learning_rate": 7.568258448992395e-05, "loss": 2.6657, "step": 31012 }, { "epoch": 2.43, "learning_rate": 7.568180036069945e-05, "loss": 2.756, "step": 31013 }, { "epoch": 2.43, "learning_rate": 7.568101623147495e-05, "loss": 2.8537, "step": 31014 }, { "epoch": 2.43, "learning_rate": 7.568023210225046e-05, "loss": 2.7041, "step": 31015 }, { "epoch": 2.43, "learning_rate": 7.567944797302596e-05, "loss": 2.8022, "step": 31016 }, { "epoch": 2.43, "learning_rate": 7.567866384380147e-05, "loss": 2.9639, "step": 31017 }, { "epoch": 2.43, "learning_rate": 7.567787971457697e-05, "loss": 3.0169, "step": 31018 }, { "epoch": 2.43, "learning_rate": 7.567709558535246e-05, "loss": 2.6582, "step": 31019 }, { "epoch": 2.43, "learning_rate": 7.567631145612796e-05, "loss": 2.6982, "step": 31020 }, { "epoch": 2.43, "learning_rate": 7.567552732690347e-05, "loss": 2.8082, "step": 31021 }, { "epoch": 2.43, "learning_rate": 7.567474319767897e-05, "loss": 2.9491, "step": 31022 }, { "epoch": 2.43, "learning_rate": 7.567395906845447e-05, "loss": 2.6117, "step": 31023 }, { "epoch": 2.43, "learning_rate": 7.567317493922999e-05, "loss": 2.7783, "step": 31024 }, { "epoch": 2.43, "learning_rate": 7.56723908100055e-05, "loss": 2.8359, "step": 31025 }, { "epoch": 2.43, "learning_rate": 7.5671606680781e-05, "loss": 3.021, "step": 31026 }, { "epoch": 2.43, "learning_rate": 7.56708225515565e-05, "loss": 2.8561, "step": 31027 }, { "epoch": 2.43, "learning_rate": 7.5670038422332e-05, "loss": 2.734, "step": 31028 }, { "epoch": 2.43, "learning_rate": 7.566925429310751e-05, "loss": 3.001, "step": 31029 }, { "epoch": 2.43, "learning_rate": 7.566847016388301e-05, "loss": 2.9076, "step": 31030 }, { "epoch": 2.43, "learning_rate": 7.566768603465852e-05, "loss": 2.8805, "step": 31031 }, { "epoch": 2.43, "learning_rate": 7.566690190543402e-05, "loss": 2.8845, "step": 31032 }, { "epoch": 2.43, "learning_rate": 7.566611777620952e-05, "loss": 2.7597, "step": 31033 }, { "epoch": 2.43, "learning_rate": 7.566533364698503e-05, "loss": 2.8089, "step": 31034 }, { "epoch": 2.43, "learning_rate": 7.566454951776053e-05, "loss": 2.6729, "step": 31035 }, { "epoch": 2.43, "learning_rate": 7.566376538853604e-05, "loss": 2.9931, "step": 31036 }, { "epoch": 2.43, "learning_rate": 7.566298125931154e-05, "loss": 2.8612, "step": 31037 }, { "epoch": 2.43, "learning_rate": 7.566219713008704e-05, "loss": 2.6215, "step": 31038 }, { "epoch": 2.43, "learning_rate": 7.566141300086255e-05, "loss": 2.6505, "step": 31039 }, { "epoch": 2.43, "learning_rate": 7.566062887163805e-05, "loss": 2.7504, "step": 31040 }, { "epoch": 2.43, "learning_rate": 7.565984474241355e-05, "loss": 2.5295, "step": 31041 }, { "epoch": 2.43, "learning_rate": 7.565906061318906e-05, "loss": 2.6556, "step": 31042 }, { "epoch": 2.43, "learning_rate": 7.565827648396456e-05, "loss": 2.8555, "step": 31043 }, { "epoch": 2.43, "learning_rate": 7.565749235474006e-05, "loss": 2.6821, "step": 31044 }, { "epoch": 2.43, "learning_rate": 7.565670822551557e-05, "loss": 3.0536, "step": 31045 }, { "epoch": 2.43, "learning_rate": 7.565592409629107e-05, "loss": 2.5961, "step": 31046 }, { "epoch": 2.43, "learning_rate": 7.565513996706658e-05, "loss": 2.923, "step": 31047 }, { "epoch": 2.43, "learning_rate": 7.565435583784208e-05, "loss": 2.6877, "step": 31048 }, { "epoch": 2.43, "learning_rate": 7.565357170861758e-05, "loss": 3.0243, "step": 31049 }, { "epoch": 2.43, "learning_rate": 7.565278757939309e-05, "loss": 3.1241, "step": 31050 }, { "epoch": 2.43, "learning_rate": 7.565200345016859e-05, "loss": 2.6815, "step": 31051 }, { "epoch": 2.43, "learning_rate": 7.56512193209441e-05, "loss": 2.8353, "step": 31052 }, { "epoch": 2.43, "learning_rate": 7.56504351917196e-05, "loss": 2.6138, "step": 31053 }, { "epoch": 2.44, "learning_rate": 7.56496510624951e-05, "loss": 3.0585, "step": 31054 }, { "epoch": 2.44, "learning_rate": 7.56488669332706e-05, "loss": 2.7426, "step": 31055 }, { "epoch": 2.44, "learning_rate": 7.564808280404611e-05, "loss": 2.8303, "step": 31056 }, { "epoch": 2.44, "learning_rate": 7.564729867482161e-05, "loss": 2.8708, "step": 31057 }, { "epoch": 2.44, "learning_rate": 7.564651454559712e-05, "loss": 2.8371, "step": 31058 }, { "epoch": 2.44, "learning_rate": 7.564573041637262e-05, "loss": 2.9144, "step": 31059 }, { "epoch": 2.44, "learning_rate": 7.564494628714812e-05, "loss": 2.72, "step": 31060 }, { "epoch": 2.44, "learning_rate": 7.564416215792363e-05, "loss": 2.8445, "step": 31061 }, { "epoch": 2.44, "learning_rate": 7.564337802869913e-05, "loss": 2.5682, "step": 31062 }, { "epoch": 2.44, "learning_rate": 7.564259389947463e-05, "loss": 2.8931, "step": 31063 }, { "epoch": 2.44, "learning_rate": 7.564180977025014e-05, "loss": 2.6043, "step": 31064 }, { "epoch": 2.44, "learning_rate": 7.564102564102564e-05, "loss": 2.8219, "step": 31065 }, { "epoch": 2.44, "learning_rate": 7.564024151180115e-05, "loss": 2.8092, "step": 31066 }, { "epoch": 2.44, "learning_rate": 7.563945738257665e-05, "loss": 2.9281, "step": 31067 }, { "epoch": 2.44, "learning_rate": 7.563867325335215e-05, "loss": 2.7287, "step": 31068 }, { "epoch": 2.44, "learning_rate": 7.563788912412766e-05, "loss": 2.7162, "step": 31069 }, { "epoch": 2.44, "learning_rate": 7.563710499490316e-05, "loss": 2.6471, "step": 31070 }, { "epoch": 2.44, "learning_rate": 7.563632086567866e-05, "loss": 2.7622, "step": 31071 }, { "epoch": 2.44, "learning_rate": 7.563553673645417e-05, "loss": 2.8792, "step": 31072 }, { "epoch": 2.44, "learning_rate": 7.563475260722967e-05, "loss": 2.7491, "step": 31073 }, { "epoch": 2.44, "learning_rate": 7.563396847800518e-05, "loss": 2.8008, "step": 31074 }, { "epoch": 2.44, "learning_rate": 7.563318434878068e-05, "loss": 2.7249, "step": 31075 }, { "epoch": 2.44, "learning_rate": 7.563240021955618e-05, "loss": 2.8534, "step": 31076 }, { "epoch": 2.44, "learning_rate": 7.563161609033169e-05, "loss": 2.6748, "step": 31077 }, { "epoch": 2.44, "learning_rate": 7.563083196110719e-05, "loss": 2.7805, "step": 31078 }, { "epoch": 2.44, "learning_rate": 7.56300478318827e-05, "loss": 3.0654, "step": 31079 }, { "epoch": 2.44, "learning_rate": 7.562926370265821e-05, "loss": 2.916, "step": 31080 }, { "epoch": 2.44, "learning_rate": 7.562847957343371e-05, "loss": 2.8462, "step": 31081 }, { "epoch": 2.44, "learning_rate": 7.562769544420922e-05, "loss": 2.7295, "step": 31082 }, { "epoch": 2.44, "learning_rate": 7.562691131498472e-05, "loss": 2.6128, "step": 31083 }, { "epoch": 2.44, "learning_rate": 7.562612718576023e-05, "loss": 2.6228, "step": 31084 }, { "epoch": 2.44, "learning_rate": 7.562534305653573e-05, "loss": 3.0376, "step": 31085 }, { "epoch": 2.44, "learning_rate": 7.562455892731123e-05, "loss": 2.8504, "step": 31086 }, { "epoch": 2.44, "learning_rate": 7.562377479808674e-05, "loss": 2.6762, "step": 31087 }, { "epoch": 2.44, "learning_rate": 7.562299066886224e-05, "loss": 2.7002, "step": 31088 }, { "epoch": 2.44, "learning_rate": 7.562220653963774e-05, "loss": 2.7595, "step": 31089 }, { "epoch": 2.44, "learning_rate": 7.562142241041325e-05, "loss": 2.8535, "step": 31090 }, { "epoch": 2.44, "learning_rate": 7.562063828118874e-05, "loss": 2.9115, "step": 31091 }, { "epoch": 2.44, "learning_rate": 7.561985415196424e-05, "loss": 2.918, "step": 31092 }, { "epoch": 2.44, "learning_rate": 7.561907002273975e-05, "loss": 2.8495, "step": 31093 }, { "epoch": 2.44, "learning_rate": 7.561828589351525e-05, "loss": 3.1195, "step": 31094 }, { "epoch": 2.44, "learning_rate": 7.561750176429075e-05, "loss": 2.6872, "step": 31095 }, { "epoch": 2.44, "learning_rate": 7.561671763506626e-05, "loss": 2.7504, "step": 31096 }, { "epoch": 2.44, "learning_rate": 7.561593350584176e-05, "loss": 2.9944, "step": 31097 }, { "epoch": 2.44, "learning_rate": 7.561514937661726e-05, "loss": 2.8584, "step": 31098 }, { "epoch": 2.44, "learning_rate": 7.561436524739277e-05, "loss": 2.7214, "step": 31099 }, { "epoch": 2.44, "learning_rate": 7.561358111816827e-05, "loss": 2.7373, "step": 31100 }, { "epoch": 2.44, "learning_rate": 7.561279698894377e-05, "loss": 2.8527, "step": 31101 }, { "epoch": 2.44, "learning_rate": 7.561201285971928e-05, "loss": 2.8307, "step": 31102 }, { "epoch": 2.44, "learning_rate": 7.561122873049478e-05, "loss": 2.9021, "step": 31103 }, { "epoch": 2.44, "learning_rate": 7.561044460127029e-05, "loss": 2.8357, "step": 31104 }, { "epoch": 2.44, "learning_rate": 7.560966047204579e-05, "loss": 2.8129, "step": 31105 }, { "epoch": 2.44, "learning_rate": 7.56088763428213e-05, "loss": 2.8805, "step": 31106 }, { "epoch": 2.44, "learning_rate": 7.560809221359681e-05, "loss": 2.9029, "step": 31107 }, { "epoch": 2.44, "learning_rate": 7.560730808437231e-05, "loss": 2.7552, "step": 31108 }, { "epoch": 2.44, "learning_rate": 7.560652395514782e-05, "loss": 2.7369, "step": 31109 }, { "epoch": 2.44, "learning_rate": 7.560573982592332e-05, "loss": 2.6515, "step": 31110 }, { "epoch": 2.44, "learning_rate": 7.560495569669882e-05, "loss": 2.8394, "step": 31111 }, { "epoch": 2.44, "learning_rate": 7.560417156747433e-05, "loss": 3.1368, "step": 31112 }, { "epoch": 2.44, "learning_rate": 7.560338743824983e-05, "loss": 2.6956, "step": 31113 }, { "epoch": 2.44, "learning_rate": 7.560260330902534e-05, "loss": 2.8352, "step": 31114 }, { "epoch": 2.44, "learning_rate": 7.560181917980084e-05, "loss": 2.7033, "step": 31115 }, { "epoch": 2.44, "learning_rate": 7.560103505057634e-05, "loss": 2.7174, "step": 31116 }, { "epoch": 2.44, "learning_rate": 7.560025092135185e-05, "loss": 2.7675, "step": 31117 }, { "epoch": 2.44, "learning_rate": 7.559946679212735e-05, "loss": 2.5833, "step": 31118 }, { "epoch": 2.44, "learning_rate": 7.559868266290285e-05, "loss": 2.7843, "step": 31119 }, { "epoch": 2.44, "learning_rate": 7.559789853367836e-05, "loss": 2.7216, "step": 31120 }, { "epoch": 2.44, "learning_rate": 7.559711440445386e-05, "loss": 3.0261, "step": 31121 }, { "epoch": 2.44, "learning_rate": 7.559633027522937e-05, "loss": 2.7563, "step": 31122 }, { "epoch": 2.44, "learning_rate": 7.559554614600487e-05, "loss": 2.7466, "step": 31123 }, { "epoch": 2.44, "learning_rate": 7.559476201678037e-05, "loss": 2.694, "step": 31124 }, { "epoch": 2.44, "learning_rate": 7.559397788755588e-05, "loss": 2.8597, "step": 31125 }, { "epoch": 2.44, "learning_rate": 7.559319375833138e-05, "loss": 3.0775, "step": 31126 }, { "epoch": 2.44, "learning_rate": 7.559240962910687e-05, "loss": 2.8261, "step": 31127 }, { "epoch": 2.44, "learning_rate": 7.559162549988237e-05, "loss": 2.9709, "step": 31128 }, { "epoch": 2.44, "learning_rate": 7.559084137065788e-05, "loss": 2.7329, "step": 31129 }, { "epoch": 2.44, "learning_rate": 7.559005724143338e-05, "loss": 2.6728, "step": 31130 }, { "epoch": 2.44, "learning_rate": 7.558927311220888e-05, "loss": 2.8484, "step": 31131 }, { "epoch": 2.44, "learning_rate": 7.55884889829844e-05, "loss": 2.9245, "step": 31132 }, { "epoch": 2.44, "learning_rate": 7.55877048537599e-05, "loss": 2.9072, "step": 31133 }, { "epoch": 2.44, "learning_rate": 7.558692072453541e-05, "loss": 2.7981, "step": 31134 }, { "epoch": 2.44, "learning_rate": 7.558613659531091e-05, "loss": 2.9194, "step": 31135 }, { "epoch": 2.44, "learning_rate": 7.558535246608642e-05, "loss": 3.1052, "step": 31136 }, { "epoch": 2.44, "learning_rate": 7.558456833686192e-05, "loss": 2.7906, "step": 31137 }, { "epoch": 2.44, "learning_rate": 7.558378420763742e-05, "loss": 2.982, "step": 31138 }, { "epoch": 2.44, "learning_rate": 7.558300007841293e-05, "loss": 2.9264, "step": 31139 }, { "epoch": 2.44, "learning_rate": 7.558221594918843e-05, "loss": 2.8456, "step": 31140 }, { "epoch": 2.44, "learning_rate": 7.558143181996394e-05, "loss": 2.9088, "step": 31141 }, { "epoch": 2.44, "learning_rate": 7.558064769073944e-05, "loss": 2.7827, "step": 31142 }, { "epoch": 2.44, "learning_rate": 7.557986356151494e-05, "loss": 2.8742, "step": 31143 }, { "epoch": 2.44, "learning_rate": 7.557907943229045e-05, "loss": 2.7195, "step": 31144 }, { "epoch": 2.44, "learning_rate": 7.557829530306595e-05, "loss": 2.8801, "step": 31145 }, { "epoch": 2.44, "learning_rate": 7.557751117384145e-05, "loss": 2.7668, "step": 31146 }, { "epoch": 2.44, "learning_rate": 7.557672704461696e-05, "loss": 2.7011, "step": 31147 }, { "epoch": 2.44, "learning_rate": 7.557594291539246e-05, "loss": 2.9253, "step": 31148 }, { "epoch": 2.44, "learning_rate": 7.557515878616796e-05, "loss": 2.8116, "step": 31149 }, { "epoch": 2.44, "learning_rate": 7.557437465694347e-05, "loss": 2.6151, "step": 31150 }, { "epoch": 2.44, "learning_rate": 7.557359052771897e-05, "loss": 2.7107, "step": 31151 }, { "epoch": 2.44, "learning_rate": 7.557280639849448e-05, "loss": 2.867, "step": 31152 }, { "epoch": 2.44, "learning_rate": 7.557202226926998e-05, "loss": 2.7004, "step": 31153 }, { "epoch": 2.44, "learning_rate": 7.557123814004548e-05, "loss": 2.8978, "step": 31154 }, { "epoch": 2.44, "learning_rate": 7.557045401082099e-05, "loss": 2.8376, "step": 31155 }, { "epoch": 2.44, "learning_rate": 7.556966988159649e-05, "loss": 2.6362, "step": 31156 }, { "epoch": 2.44, "learning_rate": 7.5568885752372e-05, "loss": 3.1371, "step": 31157 }, { "epoch": 2.44, "learning_rate": 7.55681016231475e-05, "loss": 2.934, "step": 31158 }, { "epoch": 2.44, "learning_rate": 7.5567317493923e-05, "loss": 2.4676, "step": 31159 }, { "epoch": 2.44, "learning_rate": 7.55665333646985e-05, "loss": 2.821, "step": 31160 }, { "epoch": 2.44, "learning_rate": 7.556574923547401e-05, "loss": 2.8149, "step": 31161 }, { "epoch": 2.44, "learning_rate": 7.556496510624951e-05, "loss": 2.6073, "step": 31162 }, { "epoch": 2.44, "learning_rate": 7.556418097702502e-05, "loss": 2.6126, "step": 31163 }, { "epoch": 2.44, "learning_rate": 7.556339684780052e-05, "loss": 2.7474, "step": 31164 }, { "epoch": 2.44, "learning_rate": 7.556261271857602e-05, "loss": 2.7216, "step": 31165 }, { "epoch": 2.44, "learning_rate": 7.556182858935153e-05, "loss": 2.6733, "step": 31166 }, { "epoch": 2.44, "learning_rate": 7.556104446012703e-05, "loss": 3.0152, "step": 31167 }, { "epoch": 2.44, "learning_rate": 7.556026033090253e-05, "loss": 2.6487, "step": 31168 }, { "epoch": 2.44, "learning_rate": 7.555947620167804e-05, "loss": 2.9349, "step": 31169 }, { "epoch": 2.44, "learning_rate": 7.555869207245354e-05, "loss": 2.6745, "step": 31170 }, { "epoch": 2.44, "learning_rate": 7.555790794322905e-05, "loss": 2.9754, "step": 31171 }, { "epoch": 2.44, "learning_rate": 7.555712381400455e-05, "loss": 2.6763, "step": 31172 }, { "epoch": 2.44, "learning_rate": 7.555633968478005e-05, "loss": 2.7612, "step": 31173 }, { "epoch": 2.44, "learning_rate": 7.555555555555556e-05, "loss": 2.7527, "step": 31174 }, { "epoch": 2.44, "learning_rate": 7.555477142633106e-05, "loss": 2.5933, "step": 31175 }, { "epoch": 2.44, "learning_rate": 7.555398729710656e-05, "loss": 2.8536, "step": 31176 }, { "epoch": 2.44, "learning_rate": 7.555320316788207e-05, "loss": 3.0627, "step": 31177 }, { "epoch": 2.44, "learning_rate": 7.555241903865757e-05, "loss": 2.9126, "step": 31178 }, { "epoch": 2.44, "learning_rate": 7.555163490943307e-05, "loss": 3.0075, "step": 31179 }, { "epoch": 2.44, "learning_rate": 7.555085078020858e-05, "loss": 2.7984, "step": 31180 }, { "epoch": 2.44, "learning_rate": 7.555006665098408e-05, "loss": 3.0927, "step": 31181 }, { "epoch": 2.45, "learning_rate": 7.554928252175959e-05, "loss": 2.6766, "step": 31182 }, { "epoch": 2.45, "learning_rate": 7.554849839253509e-05, "loss": 2.9057, "step": 31183 }, { "epoch": 2.45, "learning_rate": 7.55477142633106e-05, "loss": 2.7726, "step": 31184 }, { "epoch": 2.45, "learning_rate": 7.55469301340861e-05, "loss": 2.8885, "step": 31185 }, { "epoch": 2.45, "learning_rate": 7.55461460048616e-05, "loss": 2.5395, "step": 31186 }, { "epoch": 2.45, "learning_rate": 7.55453618756371e-05, "loss": 2.8411, "step": 31187 }, { "epoch": 2.45, "learning_rate": 7.554457774641262e-05, "loss": 2.8452, "step": 31188 }, { "epoch": 2.45, "learning_rate": 7.554379361718813e-05, "loss": 2.9077, "step": 31189 }, { "epoch": 2.45, "learning_rate": 7.554300948796363e-05, "loss": 2.9319, "step": 31190 }, { "epoch": 2.45, "learning_rate": 7.554222535873913e-05, "loss": 2.912, "step": 31191 }, { "epoch": 2.45, "learning_rate": 7.554144122951464e-05, "loss": 2.8893, "step": 31192 }, { "epoch": 2.45, "learning_rate": 7.554065710029014e-05, "loss": 2.6568, "step": 31193 }, { "epoch": 2.45, "learning_rate": 7.553987297106564e-05, "loss": 3.1095, "step": 31194 }, { "epoch": 2.45, "learning_rate": 7.553908884184115e-05, "loss": 2.7406, "step": 31195 }, { "epoch": 2.45, "learning_rate": 7.553830471261665e-05, "loss": 2.7799, "step": 31196 }, { "epoch": 2.45, "learning_rate": 7.553752058339215e-05, "loss": 2.6861, "step": 31197 }, { "epoch": 2.45, "learning_rate": 7.553673645416766e-05, "loss": 2.6272, "step": 31198 }, { "epoch": 2.45, "learning_rate": 7.553595232494315e-05, "loss": 2.9151, "step": 31199 }, { "epoch": 2.45, "learning_rate": 7.553516819571865e-05, "loss": 3.0244, "step": 31200 }, { "epoch": 2.45, "learning_rate": 7.553438406649416e-05, "loss": 2.7439, "step": 31201 }, { "epoch": 2.45, "learning_rate": 7.553359993726966e-05, "loss": 2.7854, "step": 31202 }, { "epoch": 2.45, "learning_rate": 7.553281580804516e-05, "loss": 2.7395, "step": 31203 }, { "epoch": 2.45, "learning_rate": 7.553203167882067e-05, "loss": 2.8783, "step": 31204 }, { "epoch": 2.45, "learning_rate": 7.553124754959617e-05, "loss": 2.8118, "step": 31205 }, { "epoch": 2.45, "learning_rate": 7.553046342037167e-05, "loss": 2.6515, "step": 31206 }, { "epoch": 2.45, "learning_rate": 7.552967929114718e-05, "loss": 2.8542, "step": 31207 }, { "epoch": 2.45, "learning_rate": 7.552889516192268e-05, "loss": 2.902, "step": 31208 }, { "epoch": 2.45, "learning_rate": 7.552811103269819e-05, "loss": 2.9612, "step": 31209 }, { "epoch": 2.45, "learning_rate": 7.552732690347369e-05, "loss": 2.6908, "step": 31210 }, { "epoch": 2.45, "learning_rate": 7.552654277424919e-05, "loss": 2.9414, "step": 31211 }, { "epoch": 2.45, "learning_rate": 7.55257586450247e-05, "loss": 2.9574, "step": 31212 }, { "epoch": 2.45, "learning_rate": 7.55249745158002e-05, "loss": 2.8415, "step": 31213 }, { "epoch": 2.45, "learning_rate": 7.552419038657572e-05, "loss": 2.851, "step": 31214 }, { "epoch": 2.45, "learning_rate": 7.552340625735122e-05, "loss": 2.8226, "step": 31215 }, { "epoch": 2.45, "learning_rate": 7.552262212812672e-05, "loss": 2.6249, "step": 31216 }, { "epoch": 2.45, "learning_rate": 7.552183799890223e-05, "loss": 2.9303, "step": 31217 }, { "epoch": 2.45, "learning_rate": 7.552105386967773e-05, "loss": 2.8434, "step": 31218 }, { "epoch": 2.45, "learning_rate": 7.552026974045324e-05, "loss": 2.8083, "step": 31219 }, { "epoch": 2.45, "learning_rate": 7.551948561122874e-05, "loss": 3.0548, "step": 31220 }, { "epoch": 2.45, "learning_rate": 7.551870148200424e-05, "loss": 2.5592, "step": 31221 }, { "epoch": 2.45, "learning_rate": 7.551791735277975e-05, "loss": 2.6256, "step": 31222 }, { "epoch": 2.45, "learning_rate": 7.551713322355525e-05, "loss": 2.5828, "step": 31223 }, { "epoch": 2.45, "learning_rate": 7.551634909433075e-05, "loss": 2.9232, "step": 31224 }, { "epoch": 2.45, "learning_rate": 7.551556496510626e-05, "loss": 3.0514, "step": 31225 }, { "epoch": 2.45, "learning_rate": 7.551478083588176e-05, "loss": 3.153, "step": 31226 }, { "epoch": 2.45, "learning_rate": 7.551399670665726e-05, "loss": 2.6061, "step": 31227 }, { "epoch": 2.45, "learning_rate": 7.551321257743277e-05, "loss": 2.7359, "step": 31228 }, { "epoch": 2.45, "learning_rate": 7.551242844820827e-05, "loss": 2.9958, "step": 31229 }, { "epoch": 2.45, "learning_rate": 7.551164431898378e-05, "loss": 2.8042, "step": 31230 }, { "epoch": 2.45, "learning_rate": 7.551086018975928e-05, "loss": 2.8734, "step": 31231 }, { "epoch": 2.45, "learning_rate": 7.551007606053478e-05, "loss": 3.0354, "step": 31232 }, { "epoch": 2.45, "learning_rate": 7.550929193131029e-05, "loss": 2.6266, "step": 31233 }, { "epoch": 2.45, "learning_rate": 7.550850780208579e-05, "loss": 2.8541, "step": 31234 }, { "epoch": 2.45, "learning_rate": 7.550772367286128e-05, "loss": 2.6341, "step": 31235 }, { "epoch": 2.45, "learning_rate": 7.550693954363678e-05, "loss": 2.8679, "step": 31236 }, { "epoch": 2.45, "learning_rate": 7.550615541441229e-05, "loss": 2.8989, "step": 31237 }, { "epoch": 2.45, "learning_rate": 7.550537128518779e-05, "loss": 2.5949, "step": 31238 }, { "epoch": 2.45, "learning_rate": 7.55045871559633e-05, "loss": 2.7471, "step": 31239 }, { "epoch": 2.45, "learning_rate": 7.550380302673881e-05, "loss": 2.8103, "step": 31240 }, { "epoch": 2.45, "learning_rate": 7.550301889751432e-05, "loss": 2.7805, "step": 31241 }, { "epoch": 2.45, "learning_rate": 7.550223476828982e-05, "loss": 2.5736, "step": 31242 }, { "epoch": 2.45, "learning_rate": 7.550145063906532e-05, "loss": 2.4759, "step": 31243 }, { "epoch": 2.45, "learning_rate": 7.550066650984083e-05, "loss": 2.6838, "step": 31244 }, { "epoch": 2.45, "learning_rate": 7.549988238061633e-05, "loss": 2.79, "step": 31245 }, { "epoch": 2.45, "learning_rate": 7.549909825139183e-05, "loss": 2.7985, "step": 31246 }, { "epoch": 2.45, "learning_rate": 7.549831412216734e-05, "loss": 2.7258, "step": 31247 }, { "epoch": 2.45, "learning_rate": 7.549752999294284e-05, "loss": 2.8563, "step": 31248 }, { "epoch": 2.45, "learning_rate": 7.549674586371835e-05, "loss": 2.8288, "step": 31249 }, { "epoch": 2.45, "learning_rate": 7.549596173449385e-05, "loss": 2.6584, "step": 31250 }, { "epoch": 2.45, "learning_rate": 7.549517760526935e-05, "loss": 2.8759, "step": 31251 }, { "epoch": 2.45, "learning_rate": 7.549439347604486e-05, "loss": 2.9537, "step": 31252 }, { "epoch": 2.45, "learning_rate": 7.549360934682036e-05, "loss": 2.7824, "step": 31253 }, { "epoch": 2.45, "learning_rate": 7.549282521759586e-05, "loss": 2.9342, "step": 31254 }, { "epoch": 2.45, "learning_rate": 7.549204108837137e-05, "loss": 2.8832, "step": 31255 }, { "epoch": 2.45, "learning_rate": 7.549125695914687e-05, "loss": 2.7116, "step": 31256 }, { "epoch": 2.45, "learning_rate": 7.549047282992238e-05, "loss": 2.7379, "step": 31257 }, { "epoch": 2.45, "learning_rate": 7.548968870069788e-05, "loss": 2.6927, "step": 31258 }, { "epoch": 2.45, "learning_rate": 7.548890457147338e-05, "loss": 2.9653, "step": 31259 }, { "epoch": 2.45, "learning_rate": 7.548812044224889e-05, "loss": 2.7432, "step": 31260 }, { "epoch": 2.45, "learning_rate": 7.548733631302439e-05, "loss": 2.7949, "step": 31261 }, { "epoch": 2.45, "learning_rate": 7.54865521837999e-05, "loss": 2.9609, "step": 31262 }, { "epoch": 2.45, "learning_rate": 7.54857680545754e-05, "loss": 2.6065, "step": 31263 }, { "epoch": 2.45, "learning_rate": 7.54849839253509e-05, "loss": 2.688, "step": 31264 }, { "epoch": 2.45, "learning_rate": 7.54841997961264e-05, "loss": 2.5793, "step": 31265 }, { "epoch": 2.45, "learning_rate": 7.548341566690191e-05, "loss": 2.6351, "step": 31266 }, { "epoch": 2.45, "learning_rate": 7.548263153767741e-05, "loss": 2.9314, "step": 31267 }, { "epoch": 2.45, "learning_rate": 7.548184740845292e-05, "loss": 2.9186, "step": 31268 }, { "epoch": 2.45, "learning_rate": 7.548106327922842e-05, "loss": 2.6505, "step": 31269 }, { "epoch": 2.45, "learning_rate": 7.548027915000392e-05, "loss": 3.0259, "step": 31270 }, { "epoch": 2.45, "learning_rate": 7.547949502077943e-05, "loss": 2.9456, "step": 31271 }, { "epoch": 2.45, "learning_rate": 7.547871089155493e-05, "loss": 2.8635, "step": 31272 }, { "epoch": 2.45, "learning_rate": 7.547792676233043e-05, "loss": 2.7907, "step": 31273 }, { "epoch": 2.45, "learning_rate": 7.547714263310594e-05, "loss": 3.0031, "step": 31274 }, { "epoch": 2.45, "learning_rate": 7.547635850388144e-05, "loss": 2.6556, "step": 31275 }, { "epoch": 2.45, "learning_rate": 7.547557437465695e-05, "loss": 2.738, "step": 31276 }, { "epoch": 2.45, "learning_rate": 7.547479024543245e-05, "loss": 2.5959, "step": 31277 }, { "epoch": 2.45, "learning_rate": 7.547400611620795e-05, "loss": 2.7875, "step": 31278 }, { "epoch": 2.45, "learning_rate": 7.547322198698346e-05, "loss": 2.7262, "step": 31279 }, { "epoch": 2.45, "learning_rate": 7.547243785775896e-05, "loss": 2.954, "step": 31280 }, { "epoch": 2.45, "learning_rate": 7.547165372853446e-05, "loss": 2.7012, "step": 31281 }, { "epoch": 2.45, "learning_rate": 7.547086959930997e-05, "loss": 2.8658, "step": 31282 }, { "epoch": 2.45, "learning_rate": 7.547008547008547e-05, "loss": 2.8264, "step": 31283 }, { "epoch": 2.45, "learning_rate": 7.546930134086097e-05, "loss": 2.6872, "step": 31284 }, { "epoch": 2.45, "learning_rate": 7.546851721163648e-05, "loss": 2.937, "step": 31285 }, { "epoch": 2.45, "learning_rate": 7.546773308241198e-05, "loss": 2.8423, "step": 31286 }, { "epoch": 2.45, "learning_rate": 7.546694895318749e-05, "loss": 2.7768, "step": 31287 }, { "epoch": 2.45, "learning_rate": 7.546616482396299e-05, "loss": 2.7693, "step": 31288 }, { "epoch": 2.45, "learning_rate": 7.546538069473849e-05, "loss": 2.5098, "step": 31289 }, { "epoch": 2.45, "learning_rate": 7.5464596565514e-05, "loss": 2.8561, "step": 31290 }, { "epoch": 2.45, "learning_rate": 7.54638124362895e-05, "loss": 2.9469, "step": 31291 }, { "epoch": 2.45, "learning_rate": 7.5463028307065e-05, "loss": 2.7908, "step": 31292 }, { "epoch": 2.45, "learning_rate": 7.546224417784051e-05, "loss": 2.9361, "step": 31293 }, { "epoch": 2.45, "learning_rate": 7.546146004861601e-05, "loss": 2.7677, "step": 31294 }, { "epoch": 2.45, "learning_rate": 7.546067591939152e-05, "loss": 2.8678, "step": 31295 }, { "epoch": 2.45, "learning_rate": 7.545989179016703e-05, "loss": 3.0031, "step": 31296 }, { "epoch": 2.45, "learning_rate": 7.545910766094254e-05, "loss": 2.8944, "step": 31297 }, { "epoch": 2.45, "learning_rate": 7.545832353171804e-05, "loss": 2.7375, "step": 31298 }, { "epoch": 2.45, "learning_rate": 7.545753940249354e-05, "loss": 2.9349, "step": 31299 }, { "epoch": 2.45, "learning_rate": 7.545675527326905e-05, "loss": 3.006, "step": 31300 }, { "epoch": 2.45, "learning_rate": 7.545597114404455e-05, "loss": 2.7385, "step": 31301 }, { "epoch": 2.45, "learning_rate": 7.545518701482005e-05, "loss": 2.5891, "step": 31302 }, { "epoch": 2.45, "learning_rate": 7.545440288559556e-05, "loss": 3.1377, "step": 31303 }, { "epoch": 2.45, "learning_rate": 7.545361875637106e-05, "loss": 2.8181, "step": 31304 }, { "epoch": 2.45, "learning_rate": 7.545283462714657e-05, "loss": 2.7095, "step": 31305 }, { "epoch": 2.45, "learning_rate": 7.545205049792207e-05, "loss": 2.9187, "step": 31306 }, { "epoch": 2.45, "learning_rate": 7.545126636869756e-05, "loss": 2.7949, "step": 31307 }, { "epoch": 2.45, "learning_rate": 7.545048223947306e-05, "loss": 2.8809, "step": 31308 }, { "epoch": 2.46, "learning_rate": 7.544969811024857e-05, "loss": 2.8271, "step": 31309 }, { "epoch": 2.46, "learning_rate": 7.544891398102407e-05, "loss": 3.015, "step": 31310 }, { "epoch": 2.46, "learning_rate": 7.544812985179957e-05, "loss": 2.7645, "step": 31311 }, { "epoch": 2.46, "learning_rate": 7.544734572257508e-05, "loss": 2.9829, "step": 31312 }, { "epoch": 2.46, "learning_rate": 7.544656159335058e-05, "loss": 2.7581, "step": 31313 }, { "epoch": 2.46, "learning_rate": 7.544577746412608e-05, "loss": 2.7618, "step": 31314 }, { "epoch": 2.46, "learning_rate": 7.544499333490159e-05, "loss": 2.747, "step": 31315 }, { "epoch": 2.46, "learning_rate": 7.544420920567709e-05, "loss": 2.8314, "step": 31316 }, { "epoch": 2.46, "learning_rate": 7.54434250764526e-05, "loss": 3.0207, "step": 31317 }, { "epoch": 2.46, "learning_rate": 7.54426409472281e-05, "loss": 2.8616, "step": 31318 }, { "epoch": 2.46, "learning_rate": 7.54418568180036e-05, "loss": 2.9141, "step": 31319 }, { "epoch": 2.46, "learning_rate": 7.544107268877911e-05, "loss": 2.5566, "step": 31320 }, { "epoch": 2.46, "learning_rate": 7.544028855955461e-05, "loss": 2.7548, "step": 31321 }, { "epoch": 2.46, "learning_rate": 7.543950443033013e-05, "loss": 2.7065, "step": 31322 }, { "epoch": 2.46, "learning_rate": 7.543872030110563e-05, "loss": 2.7274, "step": 31323 }, { "epoch": 2.46, "learning_rate": 7.543793617188114e-05, "loss": 2.686, "step": 31324 }, { "epoch": 2.46, "learning_rate": 7.543715204265664e-05, "loss": 2.9868, "step": 31325 }, { "epoch": 2.46, "learning_rate": 7.543636791343214e-05, "loss": 2.814, "step": 31326 }, { "epoch": 2.46, "learning_rate": 7.543558378420765e-05, "loss": 2.8034, "step": 31327 }, { "epoch": 2.46, "learning_rate": 7.543479965498315e-05, "loss": 2.6562, "step": 31328 }, { "epoch": 2.46, "learning_rate": 7.543401552575865e-05, "loss": 2.7775, "step": 31329 }, { "epoch": 2.46, "learning_rate": 7.543323139653416e-05, "loss": 2.6773, "step": 31330 }, { "epoch": 2.46, "learning_rate": 7.543244726730966e-05, "loss": 2.7528, "step": 31331 }, { "epoch": 2.46, "learning_rate": 7.543166313808516e-05, "loss": 2.587, "step": 31332 }, { "epoch": 2.46, "learning_rate": 7.543087900886067e-05, "loss": 2.9353, "step": 31333 }, { "epoch": 2.46, "learning_rate": 7.543009487963617e-05, "loss": 2.7973, "step": 31334 }, { "epoch": 2.46, "learning_rate": 7.542931075041168e-05, "loss": 2.9497, "step": 31335 }, { "epoch": 2.46, "learning_rate": 7.542852662118718e-05, "loss": 3.0979, "step": 31336 }, { "epoch": 2.46, "learning_rate": 7.542774249196268e-05, "loss": 2.9182, "step": 31337 }, { "epoch": 2.46, "learning_rate": 7.542695836273819e-05, "loss": 2.7633, "step": 31338 }, { "epoch": 2.46, "learning_rate": 7.542617423351369e-05, "loss": 2.6514, "step": 31339 }, { "epoch": 2.46, "learning_rate": 7.54253901042892e-05, "loss": 3.1775, "step": 31340 }, { "epoch": 2.46, "learning_rate": 7.54246059750647e-05, "loss": 2.6843, "step": 31341 }, { "epoch": 2.46, "learning_rate": 7.54238218458402e-05, "loss": 2.7208, "step": 31342 }, { "epoch": 2.46, "learning_rate": 7.542303771661569e-05, "loss": 2.6661, "step": 31343 }, { "epoch": 2.46, "learning_rate": 7.54222535873912e-05, "loss": 2.6695, "step": 31344 }, { "epoch": 2.46, "learning_rate": 7.54214694581667e-05, "loss": 2.9019, "step": 31345 }, { "epoch": 2.46, "learning_rate": 7.54206853289422e-05, "loss": 2.8025, "step": 31346 }, { "epoch": 2.46, "learning_rate": 7.54199011997177e-05, "loss": 2.7271, "step": 31347 }, { "epoch": 2.46, "learning_rate": 7.541911707049322e-05, "loss": 2.8022, "step": 31348 }, { "epoch": 2.46, "learning_rate": 7.541833294126873e-05, "loss": 2.6366, "step": 31349 }, { "epoch": 2.46, "learning_rate": 7.541754881204423e-05, "loss": 2.8211, "step": 31350 }, { "epoch": 2.46, "learning_rate": 7.541676468281973e-05, "loss": 2.7201, "step": 31351 }, { "epoch": 2.46, "learning_rate": 7.541598055359524e-05, "loss": 2.978, "step": 31352 }, { "epoch": 2.46, "learning_rate": 7.541519642437074e-05, "loss": 2.8191, "step": 31353 }, { "epoch": 2.46, "learning_rate": 7.541441229514625e-05, "loss": 2.7636, "step": 31354 }, { "epoch": 2.46, "learning_rate": 7.541362816592175e-05, "loss": 3.3484, "step": 31355 }, { "epoch": 2.46, "learning_rate": 7.541284403669725e-05, "loss": 2.661, "step": 31356 }, { "epoch": 2.46, "learning_rate": 7.541205990747276e-05, "loss": 2.9001, "step": 31357 }, { "epoch": 2.46, "learning_rate": 7.541127577824826e-05, "loss": 2.6494, "step": 31358 }, { "epoch": 2.46, "learning_rate": 7.541049164902376e-05, "loss": 2.6611, "step": 31359 }, { "epoch": 2.46, "learning_rate": 7.540970751979927e-05, "loss": 2.6578, "step": 31360 }, { "epoch": 2.46, "learning_rate": 7.540892339057477e-05, "loss": 2.7715, "step": 31361 }, { "epoch": 2.46, "learning_rate": 7.540813926135027e-05, "loss": 2.7124, "step": 31362 }, { "epoch": 2.46, "learning_rate": 7.540735513212578e-05, "loss": 2.6406, "step": 31363 }, { "epoch": 2.46, "learning_rate": 7.540657100290128e-05, "loss": 2.7962, "step": 31364 }, { "epoch": 2.46, "learning_rate": 7.540578687367679e-05, "loss": 3.0729, "step": 31365 }, { "epoch": 2.46, "learning_rate": 7.540500274445229e-05, "loss": 2.6716, "step": 31366 }, { "epoch": 2.46, "learning_rate": 7.54042186152278e-05, "loss": 2.8432, "step": 31367 }, { "epoch": 2.46, "learning_rate": 7.54034344860033e-05, "loss": 2.8751, "step": 31368 }, { "epoch": 2.46, "learning_rate": 7.54026503567788e-05, "loss": 2.6417, "step": 31369 }, { "epoch": 2.46, "learning_rate": 7.54018662275543e-05, "loss": 2.7732, "step": 31370 }, { "epoch": 2.46, "learning_rate": 7.540108209832981e-05, "loss": 2.7449, "step": 31371 }, { "epoch": 2.46, "learning_rate": 7.540029796910531e-05, "loss": 2.7958, "step": 31372 }, { "epoch": 2.46, "learning_rate": 7.539951383988082e-05, "loss": 3.0537, "step": 31373 }, { "epoch": 2.46, "learning_rate": 7.539872971065632e-05, "loss": 2.8359, "step": 31374 }, { "epoch": 2.46, "learning_rate": 7.539794558143182e-05, "loss": 2.6327, "step": 31375 }, { "epoch": 2.46, "learning_rate": 7.539716145220733e-05, "loss": 2.8558, "step": 31376 }, { "epoch": 2.46, "learning_rate": 7.539637732298283e-05, "loss": 2.9103, "step": 31377 }, { "epoch": 2.46, "learning_rate": 7.539559319375833e-05, "loss": 2.7468, "step": 31378 }, { "epoch": 2.46, "learning_rate": 7.539480906453384e-05, "loss": 3.0772, "step": 31379 }, { "epoch": 2.46, "learning_rate": 7.539402493530934e-05, "loss": 2.8761, "step": 31380 }, { "epoch": 2.46, "learning_rate": 7.539324080608484e-05, "loss": 2.841, "step": 31381 }, { "epoch": 2.46, "learning_rate": 7.539245667686035e-05, "loss": 2.5115, "step": 31382 }, { "epoch": 2.46, "learning_rate": 7.539167254763585e-05, "loss": 2.9136, "step": 31383 }, { "epoch": 2.46, "learning_rate": 7.539088841841136e-05, "loss": 2.7385, "step": 31384 }, { "epoch": 2.46, "learning_rate": 7.539010428918686e-05, "loss": 2.7144, "step": 31385 }, { "epoch": 2.46, "learning_rate": 7.538932015996236e-05, "loss": 2.7909, "step": 31386 }, { "epoch": 2.46, "learning_rate": 7.538853603073787e-05, "loss": 2.7406, "step": 31387 }, { "epoch": 2.46, "learning_rate": 7.538775190151337e-05, "loss": 2.8925, "step": 31388 }, { "epoch": 2.46, "learning_rate": 7.538696777228887e-05, "loss": 2.8293, "step": 31389 }, { "epoch": 2.46, "learning_rate": 7.538618364306438e-05, "loss": 2.8975, "step": 31390 }, { "epoch": 2.46, "learning_rate": 7.538539951383988e-05, "loss": 2.4662, "step": 31391 }, { "epoch": 2.46, "learning_rate": 7.538461538461539e-05, "loss": 2.8092, "step": 31392 }, { "epoch": 2.46, "learning_rate": 7.538383125539089e-05, "loss": 2.6842, "step": 31393 }, { "epoch": 2.46, "learning_rate": 7.538304712616639e-05, "loss": 2.9339, "step": 31394 }, { "epoch": 2.46, "learning_rate": 7.53822629969419e-05, "loss": 2.668, "step": 31395 }, { "epoch": 2.46, "learning_rate": 7.53814788677174e-05, "loss": 2.9439, "step": 31396 }, { "epoch": 2.46, "learning_rate": 7.53806947384929e-05, "loss": 2.8376, "step": 31397 }, { "epoch": 2.46, "learning_rate": 7.537991060926841e-05, "loss": 2.6357, "step": 31398 }, { "epoch": 2.46, "learning_rate": 7.537912648004391e-05, "loss": 2.8189, "step": 31399 }, { "epoch": 2.46, "learning_rate": 7.537834235081941e-05, "loss": 2.85, "step": 31400 }, { "epoch": 2.46, "learning_rate": 7.537755822159492e-05, "loss": 2.7129, "step": 31401 }, { "epoch": 2.46, "learning_rate": 7.537677409237042e-05, "loss": 2.7707, "step": 31402 }, { "epoch": 2.46, "learning_rate": 7.537598996314593e-05, "loss": 2.8154, "step": 31403 }, { "epoch": 2.46, "learning_rate": 7.537520583392143e-05, "loss": 3.0039, "step": 31404 }, { "epoch": 2.46, "learning_rate": 7.537442170469695e-05, "loss": 2.5229, "step": 31405 }, { "epoch": 2.46, "learning_rate": 7.537363757547245e-05, "loss": 2.8292, "step": 31406 }, { "epoch": 2.46, "learning_rate": 7.537285344624795e-05, "loss": 2.6027, "step": 31407 }, { "epoch": 2.46, "learning_rate": 7.537206931702346e-05, "loss": 3.1573, "step": 31408 }, { "epoch": 2.46, "learning_rate": 7.537128518779896e-05, "loss": 3.0673, "step": 31409 }, { "epoch": 2.46, "learning_rate": 7.537050105857447e-05, "loss": 2.7305, "step": 31410 }, { "epoch": 2.46, "learning_rate": 7.536971692934997e-05, "loss": 2.8212, "step": 31411 }, { "epoch": 2.46, "learning_rate": 7.536893280012547e-05, "loss": 2.7688, "step": 31412 }, { "epoch": 2.46, "learning_rate": 7.536814867090098e-05, "loss": 3.066, "step": 31413 }, { "epoch": 2.46, "learning_rate": 7.536736454167648e-05, "loss": 2.7451, "step": 31414 }, { "epoch": 2.46, "learning_rate": 7.536658041245197e-05, "loss": 2.8185, "step": 31415 }, { "epoch": 2.46, "learning_rate": 7.536579628322747e-05, "loss": 2.9127, "step": 31416 }, { "epoch": 2.46, "learning_rate": 7.536501215400298e-05, "loss": 2.81, "step": 31417 }, { "epoch": 2.46, "learning_rate": 7.536422802477848e-05, "loss": 2.9387, "step": 31418 }, { "epoch": 2.46, "learning_rate": 7.536344389555398e-05, "loss": 2.6856, "step": 31419 }, { "epoch": 2.46, "learning_rate": 7.536265976632949e-05, "loss": 2.8866, "step": 31420 }, { "epoch": 2.46, "learning_rate": 7.536187563710499e-05, "loss": 2.6791, "step": 31421 }, { "epoch": 2.46, "learning_rate": 7.53610915078805e-05, "loss": 3.1471, "step": 31422 }, { "epoch": 2.46, "learning_rate": 7.5360307378656e-05, "loss": 2.6359, "step": 31423 }, { "epoch": 2.46, "learning_rate": 7.53595232494315e-05, "loss": 2.784, "step": 31424 }, { "epoch": 2.46, "learning_rate": 7.5358739120207e-05, "loss": 2.5544, "step": 31425 }, { "epoch": 2.46, "learning_rate": 7.535795499098251e-05, "loss": 2.8434, "step": 31426 }, { "epoch": 2.46, "learning_rate": 7.535717086175801e-05, "loss": 2.6537, "step": 31427 }, { "epoch": 2.46, "learning_rate": 7.535638673253352e-05, "loss": 2.9214, "step": 31428 }, { "epoch": 2.46, "learning_rate": 7.535560260330902e-05, "loss": 2.723, "step": 31429 }, { "epoch": 2.46, "learning_rate": 7.535481847408454e-05, "loss": 2.9164, "step": 31430 }, { "epoch": 2.46, "learning_rate": 7.535403434486004e-05, "loss": 2.817, "step": 31431 }, { "epoch": 2.46, "learning_rate": 7.535325021563555e-05, "loss": 2.9817, "step": 31432 }, { "epoch": 2.46, "learning_rate": 7.535246608641105e-05, "loss": 2.7692, "step": 31433 }, { "epoch": 2.46, "learning_rate": 7.535168195718655e-05, "loss": 2.7817, "step": 31434 }, { "epoch": 2.46, "learning_rate": 7.535089782796206e-05, "loss": 2.8881, "step": 31435 }, { "epoch": 2.46, "learning_rate": 7.535011369873756e-05, "loss": 2.7156, "step": 31436 }, { "epoch": 2.47, "learning_rate": 7.534932956951306e-05, "loss": 2.8881, "step": 31437 }, { "epoch": 2.47, "learning_rate": 7.534854544028857e-05, "loss": 2.9402, "step": 31438 }, { "epoch": 2.47, "learning_rate": 7.534776131106407e-05, "loss": 2.7595, "step": 31439 }, { "epoch": 2.47, "learning_rate": 7.534697718183958e-05, "loss": 2.7411, "step": 31440 }, { "epoch": 2.47, "learning_rate": 7.534619305261508e-05, "loss": 2.7152, "step": 31441 }, { "epoch": 2.47, "learning_rate": 7.534540892339058e-05, "loss": 2.9885, "step": 31442 }, { "epoch": 2.47, "learning_rate": 7.534462479416609e-05, "loss": 2.7655, "step": 31443 }, { "epoch": 2.47, "learning_rate": 7.534384066494159e-05, "loss": 2.8772, "step": 31444 }, { "epoch": 2.47, "learning_rate": 7.53430565357171e-05, "loss": 2.6988, "step": 31445 }, { "epoch": 2.47, "learning_rate": 7.53422724064926e-05, "loss": 2.9123, "step": 31446 }, { "epoch": 2.47, "learning_rate": 7.53414882772681e-05, "loss": 2.8049, "step": 31447 }, { "epoch": 2.47, "learning_rate": 7.53407041480436e-05, "loss": 2.8311, "step": 31448 }, { "epoch": 2.47, "learning_rate": 7.533992001881911e-05, "loss": 2.946, "step": 31449 }, { "epoch": 2.47, "learning_rate": 7.533913588959461e-05, "loss": 2.9163, "step": 31450 }, { "epoch": 2.47, "learning_rate": 7.53383517603701e-05, "loss": 2.9725, "step": 31451 }, { "epoch": 2.47, "learning_rate": 7.53375676311456e-05, "loss": 2.5998, "step": 31452 }, { "epoch": 2.47, "learning_rate": 7.533678350192111e-05, "loss": 2.7629, "step": 31453 }, { "epoch": 2.47, "learning_rate": 7.533599937269661e-05, "loss": 2.8002, "step": 31454 }, { "epoch": 2.47, "learning_rate": 7.533521524347212e-05, "loss": 2.9093, "step": 31455 }, { "epoch": 2.47, "learning_rate": 7.533443111424763e-05, "loss": 2.9087, "step": 31456 }, { "epoch": 2.47, "learning_rate": 7.533364698502314e-05, "loss": 2.6728, "step": 31457 }, { "epoch": 2.47, "learning_rate": 7.533286285579864e-05, "loss": 2.6627, "step": 31458 }, { "epoch": 2.47, "learning_rate": 7.533207872657415e-05, "loss": 2.8189, "step": 31459 }, { "epoch": 2.47, "learning_rate": 7.533129459734965e-05, "loss": 2.828, "step": 31460 }, { "epoch": 2.47, "learning_rate": 7.533051046812515e-05, "loss": 3.0493, "step": 31461 }, { "epoch": 2.47, "learning_rate": 7.532972633890066e-05, "loss": 2.6783, "step": 31462 }, { "epoch": 2.47, "learning_rate": 7.532894220967616e-05, "loss": 2.5866, "step": 31463 }, { "epoch": 2.47, "learning_rate": 7.532815808045166e-05, "loss": 2.95, "step": 31464 }, { "epoch": 2.47, "learning_rate": 7.532737395122717e-05, "loss": 2.8916, "step": 31465 }, { "epoch": 2.47, "learning_rate": 7.532658982200267e-05, "loss": 2.8585, "step": 31466 }, { "epoch": 2.47, "learning_rate": 7.532580569277817e-05, "loss": 2.822, "step": 31467 }, { "epoch": 2.47, "learning_rate": 7.532502156355368e-05, "loss": 2.7985, "step": 31468 }, { "epoch": 2.47, "learning_rate": 7.532423743432918e-05, "loss": 2.7947, "step": 31469 }, { "epoch": 2.47, "learning_rate": 7.532345330510469e-05, "loss": 2.9933, "step": 31470 }, { "epoch": 2.47, "learning_rate": 7.532266917588019e-05, "loss": 3.2642, "step": 31471 }, { "epoch": 2.47, "learning_rate": 7.532188504665569e-05, "loss": 3.0197, "step": 31472 }, { "epoch": 2.47, "learning_rate": 7.53211009174312e-05, "loss": 2.8856, "step": 31473 }, { "epoch": 2.47, "learning_rate": 7.53203167882067e-05, "loss": 2.7473, "step": 31474 }, { "epoch": 2.47, "learning_rate": 7.53195326589822e-05, "loss": 2.9717, "step": 31475 }, { "epoch": 2.47, "learning_rate": 7.531874852975771e-05, "loss": 2.7708, "step": 31476 }, { "epoch": 2.47, "learning_rate": 7.531796440053321e-05, "loss": 2.7273, "step": 31477 }, { "epoch": 2.47, "learning_rate": 7.531718027130872e-05, "loss": 2.756, "step": 31478 }, { "epoch": 2.47, "learning_rate": 7.531639614208422e-05, "loss": 2.7148, "step": 31479 }, { "epoch": 2.47, "learning_rate": 7.531561201285972e-05, "loss": 2.6624, "step": 31480 }, { "epoch": 2.47, "learning_rate": 7.531482788363523e-05, "loss": 2.7779, "step": 31481 }, { "epoch": 2.47, "learning_rate": 7.531404375441073e-05, "loss": 2.9125, "step": 31482 }, { "epoch": 2.47, "learning_rate": 7.531325962518623e-05, "loss": 3.0592, "step": 31483 }, { "epoch": 2.47, "learning_rate": 7.531247549596174e-05, "loss": 2.7658, "step": 31484 }, { "epoch": 2.47, "learning_rate": 7.531169136673724e-05, "loss": 2.8889, "step": 31485 }, { "epoch": 2.47, "learning_rate": 7.531090723751274e-05, "loss": 2.8446, "step": 31486 }, { "epoch": 2.47, "learning_rate": 7.531012310828825e-05, "loss": 2.7692, "step": 31487 }, { "epoch": 2.47, "learning_rate": 7.530933897906375e-05, "loss": 2.7032, "step": 31488 }, { "epoch": 2.47, "learning_rate": 7.530855484983926e-05, "loss": 2.7859, "step": 31489 }, { "epoch": 2.47, "learning_rate": 7.530777072061476e-05, "loss": 2.8268, "step": 31490 }, { "epoch": 2.47, "learning_rate": 7.530698659139026e-05, "loss": 2.8425, "step": 31491 }, { "epoch": 2.47, "learning_rate": 7.530620246216577e-05, "loss": 2.9059, "step": 31492 }, { "epoch": 2.47, "learning_rate": 7.530541833294127e-05, "loss": 2.9537, "step": 31493 }, { "epoch": 2.47, "learning_rate": 7.530463420371677e-05, "loss": 2.8709, "step": 31494 }, { "epoch": 2.47, "learning_rate": 7.530385007449228e-05, "loss": 2.6231, "step": 31495 }, { "epoch": 2.47, "learning_rate": 7.530306594526778e-05, "loss": 2.7331, "step": 31496 }, { "epoch": 2.47, "learning_rate": 7.530228181604329e-05, "loss": 2.9899, "step": 31497 }, { "epoch": 2.47, "learning_rate": 7.530149768681879e-05, "loss": 2.9867, "step": 31498 }, { "epoch": 2.47, "learning_rate": 7.530071355759429e-05, "loss": 2.8613, "step": 31499 }, { "epoch": 2.47, "learning_rate": 7.52999294283698e-05, "loss": 2.7882, "step": 31500 }, { "epoch": 2.47, "learning_rate": 7.52991452991453e-05, "loss": 2.9206, "step": 31501 }, { "epoch": 2.47, "learning_rate": 7.52983611699208e-05, "loss": 2.8932, "step": 31502 }, { "epoch": 2.47, "learning_rate": 7.529757704069631e-05, "loss": 2.7608, "step": 31503 }, { "epoch": 2.47, "learning_rate": 7.529679291147181e-05, "loss": 2.7388, "step": 31504 }, { "epoch": 2.47, "learning_rate": 7.529600878224731e-05, "loss": 2.7735, "step": 31505 }, { "epoch": 2.47, "learning_rate": 7.529522465302282e-05, "loss": 2.9183, "step": 31506 }, { "epoch": 2.47, "learning_rate": 7.529444052379832e-05, "loss": 2.657, "step": 31507 }, { "epoch": 2.47, "learning_rate": 7.529365639457383e-05, "loss": 2.9076, "step": 31508 }, { "epoch": 2.47, "learning_rate": 7.529287226534933e-05, "loss": 2.9974, "step": 31509 }, { "epoch": 2.47, "learning_rate": 7.529208813612483e-05, "loss": 2.7459, "step": 31510 }, { "epoch": 2.47, "learning_rate": 7.529130400690034e-05, "loss": 2.8173, "step": 31511 }, { "epoch": 2.47, "learning_rate": 7.529051987767584e-05, "loss": 3.0046, "step": 31512 }, { "epoch": 2.47, "learning_rate": 7.528973574845136e-05, "loss": 2.5792, "step": 31513 }, { "epoch": 2.47, "learning_rate": 7.528895161922686e-05, "loss": 2.6842, "step": 31514 }, { "epoch": 2.47, "learning_rate": 7.528816749000236e-05, "loss": 2.7918, "step": 31515 }, { "epoch": 2.47, "learning_rate": 7.528738336077787e-05, "loss": 2.8875, "step": 31516 }, { "epoch": 2.47, "learning_rate": 7.528659923155337e-05, "loss": 2.946, "step": 31517 }, { "epoch": 2.47, "learning_rate": 7.528581510232888e-05, "loss": 2.673, "step": 31518 }, { "epoch": 2.47, "learning_rate": 7.528503097310438e-05, "loss": 2.8551, "step": 31519 }, { "epoch": 2.47, "learning_rate": 7.528424684387988e-05, "loss": 2.9344, "step": 31520 }, { "epoch": 2.47, "learning_rate": 7.528346271465539e-05, "loss": 2.6572, "step": 31521 }, { "epoch": 2.47, "learning_rate": 7.528267858543089e-05, "loss": 3.0967, "step": 31522 }, { "epoch": 2.47, "learning_rate": 7.528189445620638e-05, "loss": 2.8504, "step": 31523 }, { "epoch": 2.47, "learning_rate": 7.528111032698188e-05, "loss": 2.8434, "step": 31524 }, { "epoch": 2.47, "learning_rate": 7.528032619775739e-05, "loss": 2.942, "step": 31525 }, { "epoch": 2.47, "learning_rate": 7.527954206853289e-05, "loss": 2.8082, "step": 31526 }, { "epoch": 2.47, "learning_rate": 7.52787579393084e-05, "loss": 2.7786, "step": 31527 }, { "epoch": 2.47, "learning_rate": 7.52779738100839e-05, "loss": 2.616, "step": 31528 }, { "epoch": 2.47, "learning_rate": 7.52771896808594e-05, "loss": 2.6735, "step": 31529 }, { "epoch": 2.47, "learning_rate": 7.52764055516349e-05, "loss": 3.0276, "step": 31530 }, { "epoch": 2.47, "learning_rate": 7.527562142241041e-05, "loss": 2.6081, "step": 31531 }, { "epoch": 2.47, "learning_rate": 7.527483729318591e-05, "loss": 2.7806, "step": 31532 }, { "epoch": 2.47, "learning_rate": 7.527405316396142e-05, "loss": 2.7818, "step": 31533 }, { "epoch": 2.47, "learning_rate": 7.527326903473692e-05, "loss": 2.8385, "step": 31534 }, { "epoch": 2.47, "learning_rate": 7.527248490551242e-05, "loss": 2.7932, "step": 31535 }, { "epoch": 2.47, "learning_rate": 7.527170077628793e-05, "loss": 2.9667, "step": 31536 }, { "epoch": 2.47, "learning_rate": 7.527091664706343e-05, "loss": 2.804, "step": 31537 }, { "epoch": 2.47, "learning_rate": 7.527013251783895e-05, "loss": 2.8469, "step": 31538 }, { "epoch": 2.47, "learning_rate": 7.526934838861445e-05, "loss": 2.7468, "step": 31539 }, { "epoch": 2.47, "learning_rate": 7.526856425938996e-05, "loss": 2.7547, "step": 31540 }, { "epoch": 2.47, "learning_rate": 7.526778013016546e-05, "loss": 2.6855, "step": 31541 }, { "epoch": 2.47, "learning_rate": 7.526699600094096e-05, "loss": 2.8465, "step": 31542 }, { "epoch": 2.47, "learning_rate": 7.526621187171647e-05, "loss": 2.8412, "step": 31543 }, { "epoch": 2.47, "learning_rate": 7.526542774249197e-05, "loss": 2.7383, "step": 31544 }, { "epoch": 2.47, "learning_rate": 7.526464361326748e-05, "loss": 2.5953, "step": 31545 }, { "epoch": 2.47, "learning_rate": 7.526385948404298e-05, "loss": 2.7771, "step": 31546 }, { "epoch": 2.47, "learning_rate": 7.526307535481848e-05, "loss": 2.8678, "step": 31547 }, { "epoch": 2.47, "learning_rate": 7.526229122559399e-05, "loss": 3.0364, "step": 31548 }, { "epoch": 2.47, "learning_rate": 7.526150709636949e-05, "loss": 2.7158, "step": 31549 }, { "epoch": 2.47, "learning_rate": 7.5260722967145e-05, "loss": 2.8114, "step": 31550 }, { "epoch": 2.47, "learning_rate": 7.52599388379205e-05, "loss": 2.6321, "step": 31551 }, { "epoch": 2.47, "learning_rate": 7.5259154708696e-05, "loss": 2.9461, "step": 31552 }, { "epoch": 2.47, "learning_rate": 7.52583705794715e-05, "loss": 2.8131, "step": 31553 }, { "epoch": 2.47, "learning_rate": 7.525758645024701e-05, "loss": 2.7174, "step": 31554 }, { "epoch": 2.47, "learning_rate": 7.525680232102251e-05, "loss": 2.6924, "step": 31555 }, { "epoch": 2.47, "learning_rate": 7.525601819179802e-05, "loss": 2.7967, "step": 31556 }, { "epoch": 2.47, "learning_rate": 7.525523406257352e-05, "loss": 2.6822, "step": 31557 }, { "epoch": 2.47, "learning_rate": 7.525444993334902e-05, "loss": 2.6691, "step": 31558 }, { "epoch": 2.47, "learning_rate": 7.525366580412451e-05, "loss": 3.0399, "step": 31559 }, { "epoch": 2.47, "learning_rate": 7.525288167490002e-05, "loss": 2.8658, "step": 31560 }, { "epoch": 2.47, "learning_rate": 7.525209754567552e-05, "loss": 2.7211, "step": 31561 }, { "epoch": 2.47, "learning_rate": 7.525131341645102e-05, "loss": 2.6994, "step": 31562 }, { "epoch": 2.47, "learning_rate": 7.525052928722653e-05, "loss": 2.8824, "step": 31563 }, { "epoch": 2.48, "learning_rate": 7.524974515800204e-05, "loss": 2.9331, "step": 31564 }, { "epoch": 2.48, "learning_rate": 7.524896102877755e-05, "loss": 2.6344, "step": 31565 }, { "epoch": 2.48, "learning_rate": 7.524817689955305e-05, "loss": 2.6195, "step": 31566 }, { "epoch": 2.48, "learning_rate": 7.524739277032856e-05, "loss": 2.6536, "step": 31567 }, { "epoch": 2.48, "learning_rate": 7.524660864110406e-05, "loss": 2.8386, "step": 31568 }, { "epoch": 2.48, "learning_rate": 7.524582451187956e-05, "loss": 2.8146, "step": 31569 }, { "epoch": 2.48, "learning_rate": 7.524504038265507e-05, "loss": 2.8156, "step": 31570 }, { "epoch": 2.48, "learning_rate": 7.524425625343057e-05, "loss": 2.8375, "step": 31571 }, { "epoch": 2.48, "learning_rate": 7.524347212420607e-05, "loss": 2.9401, "step": 31572 }, { "epoch": 2.48, "learning_rate": 7.524268799498158e-05, "loss": 2.8454, "step": 31573 }, { "epoch": 2.48, "learning_rate": 7.524190386575708e-05, "loss": 2.9103, "step": 31574 }, { "epoch": 2.48, "learning_rate": 7.524111973653259e-05, "loss": 2.8088, "step": 31575 }, { "epoch": 2.48, "learning_rate": 7.524033560730809e-05, "loss": 2.8383, "step": 31576 }, { "epoch": 2.48, "learning_rate": 7.523955147808359e-05, "loss": 2.7021, "step": 31577 }, { "epoch": 2.48, "learning_rate": 7.52387673488591e-05, "loss": 2.8144, "step": 31578 }, { "epoch": 2.48, "learning_rate": 7.52379832196346e-05, "loss": 2.547, "step": 31579 }, { "epoch": 2.48, "learning_rate": 7.52371990904101e-05, "loss": 2.6193, "step": 31580 }, { "epoch": 2.48, "learning_rate": 7.523641496118561e-05, "loss": 3.0407, "step": 31581 }, { "epoch": 2.48, "learning_rate": 7.523563083196111e-05, "loss": 2.6425, "step": 31582 }, { "epoch": 2.48, "learning_rate": 7.523484670273661e-05, "loss": 2.7872, "step": 31583 }, { "epoch": 2.48, "learning_rate": 7.523406257351212e-05, "loss": 2.7318, "step": 31584 }, { "epoch": 2.48, "learning_rate": 7.523327844428762e-05, "loss": 2.6998, "step": 31585 }, { "epoch": 2.48, "learning_rate": 7.523249431506313e-05, "loss": 2.7033, "step": 31586 }, { "epoch": 2.48, "learning_rate": 7.523171018583863e-05, "loss": 2.7304, "step": 31587 }, { "epoch": 2.48, "learning_rate": 7.523092605661413e-05, "loss": 3.2237, "step": 31588 }, { "epoch": 2.48, "learning_rate": 7.523014192738964e-05, "loss": 2.8815, "step": 31589 }, { "epoch": 2.48, "learning_rate": 7.522935779816514e-05, "loss": 2.628, "step": 31590 }, { "epoch": 2.48, "learning_rate": 7.522857366894064e-05, "loss": 2.8499, "step": 31591 }, { "epoch": 2.48, "learning_rate": 7.522778953971615e-05, "loss": 2.8864, "step": 31592 }, { "epoch": 2.48, "learning_rate": 7.522700541049165e-05, "loss": 2.5677, "step": 31593 }, { "epoch": 2.48, "learning_rate": 7.522622128126716e-05, "loss": 2.7177, "step": 31594 }, { "epoch": 2.48, "learning_rate": 7.522543715204266e-05, "loss": 2.7456, "step": 31595 }, { "epoch": 2.48, "learning_rate": 7.522465302281816e-05, "loss": 2.8047, "step": 31596 }, { "epoch": 2.48, "learning_rate": 7.522386889359367e-05, "loss": 2.5711, "step": 31597 }, { "epoch": 2.48, "learning_rate": 7.522308476436917e-05, "loss": 2.9379, "step": 31598 }, { "epoch": 2.48, "learning_rate": 7.522230063514467e-05, "loss": 2.7143, "step": 31599 }, { "epoch": 2.48, "learning_rate": 7.522151650592018e-05, "loss": 2.9117, "step": 31600 }, { "epoch": 2.48, "learning_rate": 7.522073237669568e-05, "loss": 2.8653, "step": 31601 }, { "epoch": 2.48, "learning_rate": 7.521994824747118e-05, "loss": 2.9498, "step": 31602 }, { "epoch": 2.48, "learning_rate": 7.521916411824669e-05, "loss": 2.7073, "step": 31603 }, { "epoch": 2.48, "learning_rate": 7.521837998902219e-05, "loss": 2.7904, "step": 31604 }, { "epoch": 2.48, "learning_rate": 7.52175958597977e-05, "loss": 2.6755, "step": 31605 }, { "epoch": 2.48, "learning_rate": 7.52168117305732e-05, "loss": 2.7987, "step": 31606 }, { "epoch": 2.48, "learning_rate": 7.52160276013487e-05, "loss": 2.8114, "step": 31607 }, { "epoch": 2.48, "learning_rate": 7.521524347212421e-05, "loss": 2.6208, "step": 31608 }, { "epoch": 2.48, "learning_rate": 7.521445934289971e-05, "loss": 3.1478, "step": 31609 }, { "epoch": 2.48, "learning_rate": 7.521367521367521e-05, "loss": 2.8689, "step": 31610 }, { "epoch": 2.48, "learning_rate": 7.521289108445072e-05, "loss": 2.8307, "step": 31611 }, { "epoch": 2.48, "learning_rate": 7.521210695522622e-05, "loss": 2.8014, "step": 31612 }, { "epoch": 2.48, "learning_rate": 7.521132282600173e-05, "loss": 2.6535, "step": 31613 }, { "epoch": 2.48, "learning_rate": 7.521053869677723e-05, "loss": 2.6961, "step": 31614 }, { "epoch": 2.48, "learning_rate": 7.520975456755273e-05, "loss": 2.9733, "step": 31615 }, { "epoch": 2.48, "learning_rate": 7.520897043832824e-05, "loss": 3.0312, "step": 31616 }, { "epoch": 2.48, "learning_rate": 7.520818630910374e-05, "loss": 2.9405, "step": 31617 }, { "epoch": 2.48, "learning_rate": 7.520740217987924e-05, "loss": 2.9149, "step": 31618 }, { "epoch": 2.48, "learning_rate": 7.520661805065475e-05, "loss": 2.7408, "step": 31619 }, { "epoch": 2.48, "learning_rate": 7.520583392143025e-05, "loss": 2.6062, "step": 31620 }, { "epoch": 2.48, "learning_rate": 7.520504979220577e-05, "loss": 2.8736, "step": 31621 }, { "epoch": 2.48, "learning_rate": 7.520426566298127e-05, "loss": 2.8115, "step": 31622 }, { "epoch": 2.48, "learning_rate": 7.520348153375678e-05, "loss": 2.7301, "step": 31623 }, { "epoch": 2.48, "learning_rate": 7.520269740453228e-05, "loss": 2.9214, "step": 31624 }, { "epoch": 2.48, "learning_rate": 7.520191327530778e-05, "loss": 2.7108, "step": 31625 }, { "epoch": 2.48, "learning_rate": 7.520112914608329e-05, "loss": 2.8027, "step": 31626 }, { "epoch": 2.48, "learning_rate": 7.520034501685879e-05, "loss": 2.9973, "step": 31627 }, { "epoch": 2.48, "learning_rate": 7.51995608876343e-05, "loss": 2.658, "step": 31628 }, { "epoch": 2.48, "learning_rate": 7.51987767584098e-05, "loss": 2.9061, "step": 31629 }, { "epoch": 2.48, "learning_rate": 7.51979926291853e-05, "loss": 2.8901, "step": 31630 }, { "epoch": 2.48, "learning_rate": 7.519720849996079e-05, "loss": 2.6741, "step": 31631 }, { "epoch": 2.48, "learning_rate": 7.51964243707363e-05, "loss": 3.062, "step": 31632 }, { "epoch": 2.48, "learning_rate": 7.51956402415118e-05, "loss": 2.8359, "step": 31633 }, { "epoch": 2.48, "learning_rate": 7.51948561122873e-05, "loss": 2.8136, "step": 31634 }, { "epoch": 2.48, "learning_rate": 7.51940719830628e-05, "loss": 2.8602, "step": 31635 }, { "epoch": 2.48, "learning_rate": 7.519328785383831e-05, "loss": 2.6404, "step": 31636 }, { "epoch": 2.48, "learning_rate": 7.519250372461381e-05, "loss": 3.2087, "step": 31637 }, { "epoch": 2.48, "learning_rate": 7.519171959538932e-05, "loss": 2.7817, "step": 31638 }, { "epoch": 2.48, "learning_rate": 7.519093546616482e-05, "loss": 2.6411, "step": 31639 }, { "epoch": 2.48, "learning_rate": 7.519015133694032e-05, "loss": 2.7947, "step": 31640 }, { "epoch": 2.48, "learning_rate": 7.518936720771583e-05, "loss": 2.7568, "step": 31641 }, { "epoch": 2.48, "learning_rate": 7.518858307849133e-05, "loss": 2.788, "step": 31642 }, { "epoch": 2.48, "learning_rate": 7.518779894926684e-05, "loss": 2.8337, "step": 31643 }, { "epoch": 2.48, "learning_rate": 7.518701482004234e-05, "loss": 2.9507, "step": 31644 }, { "epoch": 2.48, "learning_rate": 7.518623069081784e-05, "loss": 2.6016, "step": 31645 }, { "epoch": 2.48, "learning_rate": 7.518544656159336e-05, "loss": 2.7412, "step": 31646 }, { "epoch": 2.48, "learning_rate": 7.518466243236886e-05, "loss": 2.7112, "step": 31647 }, { "epoch": 2.48, "learning_rate": 7.518387830314437e-05, "loss": 2.782, "step": 31648 }, { "epoch": 2.48, "learning_rate": 7.518309417391987e-05, "loss": 2.7638, "step": 31649 }, { "epoch": 2.48, "learning_rate": 7.518231004469537e-05, "loss": 2.8479, "step": 31650 }, { "epoch": 2.48, "learning_rate": 7.518152591547088e-05, "loss": 2.9813, "step": 31651 }, { "epoch": 2.48, "learning_rate": 7.518074178624638e-05, "loss": 2.6855, "step": 31652 }, { "epoch": 2.48, "learning_rate": 7.517995765702189e-05, "loss": 2.6977, "step": 31653 }, { "epoch": 2.48, "learning_rate": 7.517917352779739e-05, "loss": 3.1739, "step": 31654 }, { "epoch": 2.48, "learning_rate": 7.517838939857289e-05, "loss": 2.8043, "step": 31655 }, { "epoch": 2.48, "learning_rate": 7.51776052693484e-05, "loss": 2.932, "step": 31656 }, { "epoch": 2.48, "learning_rate": 7.51768211401239e-05, "loss": 2.826, "step": 31657 }, { "epoch": 2.48, "learning_rate": 7.51760370108994e-05, "loss": 2.7142, "step": 31658 }, { "epoch": 2.48, "learning_rate": 7.517525288167491e-05, "loss": 2.8009, "step": 31659 }, { "epoch": 2.48, "learning_rate": 7.517446875245041e-05, "loss": 2.7236, "step": 31660 }, { "epoch": 2.48, "learning_rate": 7.517368462322592e-05, "loss": 2.8549, "step": 31661 }, { "epoch": 2.48, "learning_rate": 7.517290049400142e-05, "loss": 2.7305, "step": 31662 }, { "epoch": 2.48, "learning_rate": 7.517211636477692e-05, "loss": 2.7265, "step": 31663 }, { "epoch": 2.48, "learning_rate": 7.517133223555243e-05, "loss": 2.8753, "step": 31664 }, { "epoch": 2.48, "learning_rate": 7.517054810632793e-05, "loss": 2.7585, "step": 31665 }, { "epoch": 2.48, "learning_rate": 7.516976397710343e-05, "loss": 2.6926, "step": 31666 }, { "epoch": 2.48, "learning_rate": 7.516897984787892e-05, "loss": 2.8103, "step": 31667 }, { "epoch": 2.48, "learning_rate": 7.516819571865443e-05, "loss": 2.8336, "step": 31668 }, { "epoch": 2.48, "learning_rate": 7.516741158942993e-05, "loss": 2.8596, "step": 31669 }, { "epoch": 2.48, "learning_rate": 7.516662746020543e-05, "loss": 2.7365, "step": 31670 }, { "epoch": 2.48, "learning_rate": 7.516584333098094e-05, "loss": 2.7863, "step": 31671 }, { "epoch": 2.48, "learning_rate": 7.516505920175646e-05, "loss": 2.9096, "step": 31672 }, { "epoch": 2.48, "learning_rate": 7.516427507253196e-05, "loss": 2.6364, "step": 31673 }, { "epoch": 2.48, "learning_rate": 7.516349094330746e-05, "loss": 2.8, "step": 31674 }, { "epoch": 2.48, "learning_rate": 7.516270681408297e-05, "loss": 2.6774, "step": 31675 }, { "epoch": 2.48, "learning_rate": 7.516192268485847e-05, "loss": 2.8685, "step": 31676 }, { "epoch": 2.48, "learning_rate": 7.516113855563397e-05, "loss": 2.835, "step": 31677 }, { "epoch": 2.48, "learning_rate": 7.516035442640948e-05, "loss": 2.6251, "step": 31678 }, { "epoch": 2.48, "learning_rate": 7.515957029718498e-05, "loss": 2.7474, "step": 31679 }, { "epoch": 2.48, "learning_rate": 7.515878616796049e-05, "loss": 3.0792, "step": 31680 }, { "epoch": 2.48, "learning_rate": 7.515800203873599e-05, "loss": 2.6539, "step": 31681 }, { "epoch": 2.48, "learning_rate": 7.515721790951149e-05, "loss": 3.0771, "step": 31682 }, { "epoch": 2.48, "learning_rate": 7.5156433780287e-05, "loss": 3.0321, "step": 31683 }, { "epoch": 2.48, "learning_rate": 7.51556496510625e-05, "loss": 2.6803, "step": 31684 }, { "epoch": 2.48, "learning_rate": 7.5154865521838e-05, "loss": 2.7758, "step": 31685 }, { "epoch": 2.48, "learning_rate": 7.515408139261351e-05, "loss": 2.8942, "step": 31686 }, { "epoch": 2.48, "learning_rate": 7.515329726338901e-05, "loss": 2.9441, "step": 31687 }, { "epoch": 2.48, "learning_rate": 7.515251313416451e-05, "loss": 2.8504, "step": 31688 }, { "epoch": 2.48, "learning_rate": 7.515172900494002e-05, "loss": 2.698, "step": 31689 }, { "epoch": 2.48, "learning_rate": 7.515094487571552e-05, "loss": 2.8931, "step": 31690 }, { "epoch": 2.48, "learning_rate": 7.515016074649103e-05, "loss": 2.9214, "step": 31691 }, { "epoch": 2.49, "learning_rate": 7.514937661726653e-05, "loss": 2.9455, "step": 31692 }, { "epoch": 2.49, "learning_rate": 7.514859248804203e-05, "loss": 2.5065, "step": 31693 }, { "epoch": 2.49, "learning_rate": 7.514780835881754e-05, "loss": 2.7569, "step": 31694 }, { "epoch": 2.49, "learning_rate": 7.514702422959304e-05, "loss": 2.752, "step": 31695 }, { "epoch": 2.49, "learning_rate": 7.514624010036854e-05, "loss": 2.9186, "step": 31696 }, { "epoch": 2.49, "learning_rate": 7.514545597114405e-05, "loss": 2.782, "step": 31697 }, { "epoch": 2.49, "learning_rate": 7.514467184191955e-05, "loss": 2.9168, "step": 31698 }, { "epoch": 2.49, "learning_rate": 7.514388771269506e-05, "loss": 2.8078, "step": 31699 }, { "epoch": 2.49, "learning_rate": 7.514310358347056e-05, "loss": 2.7233, "step": 31700 }, { "epoch": 2.49, "learning_rate": 7.514231945424606e-05, "loss": 2.619, "step": 31701 }, { "epoch": 2.49, "learning_rate": 7.514153532502157e-05, "loss": 2.6625, "step": 31702 }, { "epoch": 2.49, "learning_rate": 7.514075119579707e-05, "loss": 2.7471, "step": 31703 }, { "epoch": 2.49, "learning_rate": 7.513996706657257e-05, "loss": 2.721, "step": 31704 }, { "epoch": 2.49, "learning_rate": 7.513918293734808e-05, "loss": 2.979, "step": 31705 }, { "epoch": 2.49, "learning_rate": 7.513839880812358e-05, "loss": 2.9701, "step": 31706 }, { "epoch": 2.49, "learning_rate": 7.513761467889908e-05, "loss": 2.7979, "step": 31707 }, { "epoch": 2.49, "learning_rate": 7.513683054967459e-05, "loss": 3.0058, "step": 31708 }, { "epoch": 2.49, "learning_rate": 7.513604642045009e-05, "loss": 2.7285, "step": 31709 }, { "epoch": 2.49, "learning_rate": 7.51352622912256e-05, "loss": 2.6492, "step": 31710 }, { "epoch": 2.49, "learning_rate": 7.51344781620011e-05, "loss": 2.8107, "step": 31711 }, { "epoch": 2.49, "learning_rate": 7.51336940327766e-05, "loss": 2.5419, "step": 31712 }, { "epoch": 2.49, "learning_rate": 7.51329099035521e-05, "loss": 2.7176, "step": 31713 }, { "epoch": 2.49, "learning_rate": 7.513212577432761e-05, "loss": 2.7132, "step": 31714 }, { "epoch": 2.49, "learning_rate": 7.513134164510311e-05, "loss": 2.8577, "step": 31715 }, { "epoch": 2.49, "learning_rate": 7.513055751587862e-05, "loss": 2.7014, "step": 31716 }, { "epoch": 2.49, "learning_rate": 7.512977338665412e-05, "loss": 2.873, "step": 31717 }, { "epoch": 2.49, "learning_rate": 7.512898925742962e-05, "loss": 2.8081, "step": 31718 }, { "epoch": 2.49, "learning_rate": 7.512820512820513e-05, "loss": 2.9807, "step": 31719 }, { "epoch": 2.49, "learning_rate": 7.512742099898063e-05, "loss": 2.7358, "step": 31720 }, { "epoch": 2.49, "learning_rate": 7.512663686975614e-05, "loss": 2.9008, "step": 31721 }, { "epoch": 2.49, "learning_rate": 7.512585274053164e-05, "loss": 2.975, "step": 31722 }, { "epoch": 2.49, "learning_rate": 7.512506861130714e-05, "loss": 2.894, "step": 31723 }, { "epoch": 2.49, "learning_rate": 7.512428448208265e-05, "loss": 2.8558, "step": 31724 }, { "epoch": 2.49, "learning_rate": 7.512350035285815e-05, "loss": 2.8766, "step": 31725 }, { "epoch": 2.49, "learning_rate": 7.512271622363365e-05, "loss": 2.6819, "step": 31726 }, { "epoch": 2.49, "learning_rate": 7.512193209440916e-05, "loss": 2.7315, "step": 31727 }, { "epoch": 2.49, "learning_rate": 7.512114796518466e-05, "loss": 3.1995, "step": 31728 }, { "epoch": 2.49, "learning_rate": 7.512036383596018e-05, "loss": 2.6745, "step": 31729 }, { "epoch": 2.49, "learning_rate": 7.511957970673568e-05, "loss": 2.992, "step": 31730 }, { "epoch": 2.49, "learning_rate": 7.511879557751119e-05, "loss": 2.7276, "step": 31731 }, { "epoch": 2.49, "learning_rate": 7.511801144828669e-05, "loss": 2.777, "step": 31732 }, { "epoch": 2.49, "learning_rate": 7.51172273190622e-05, "loss": 2.5976, "step": 31733 }, { "epoch": 2.49, "learning_rate": 7.51164431898377e-05, "loss": 3.0845, "step": 31734 }, { "epoch": 2.49, "learning_rate": 7.51156590606132e-05, "loss": 2.7579, "step": 31735 }, { "epoch": 2.49, "learning_rate": 7.51148749313887e-05, "loss": 2.747, "step": 31736 }, { "epoch": 2.49, "learning_rate": 7.511409080216421e-05, "loss": 2.6746, "step": 31737 }, { "epoch": 2.49, "learning_rate": 7.511330667293971e-05, "loss": 2.6194, "step": 31738 }, { "epoch": 2.49, "learning_rate": 7.51125225437152e-05, "loss": 2.7354, "step": 31739 }, { "epoch": 2.49, "learning_rate": 7.51117384144907e-05, "loss": 2.9328, "step": 31740 }, { "epoch": 2.49, "learning_rate": 7.511095428526621e-05, "loss": 3.0109, "step": 31741 }, { "epoch": 2.49, "learning_rate": 7.511017015604171e-05, "loss": 2.9437, "step": 31742 }, { "epoch": 2.49, "learning_rate": 7.510938602681722e-05, "loss": 2.7241, "step": 31743 }, { "epoch": 2.49, "learning_rate": 7.510860189759272e-05, "loss": 2.7747, "step": 31744 }, { "epoch": 2.49, "learning_rate": 7.510781776836822e-05, "loss": 2.7873, "step": 31745 }, { "epoch": 2.49, "learning_rate": 7.510703363914373e-05, "loss": 2.6582, "step": 31746 }, { "epoch": 2.49, "learning_rate": 7.510624950991923e-05, "loss": 2.6867, "step": 31747 }, { "epoch": 2.49, "learning_rate": 7.510546538069474e-05, "loss": 2.6348, "step": 31748 }, { "epoch": 2.49, "learning_rate": 7.510468125147024e-05, "loss": 2.7457, "step": 31749 }, { "epoch": 2.49, "learning_rate": 7.510389712224574e-05, "loss": 2.9229, "step": 31750 }, { "epoch": 2.49, "learning_rate": 7.510311299302125e-05, "loss": 2.894, "step": 31751 }, { "epoch": 2.49, "learning_rate": 7.510232886379675e-05, "loss": 2.9094, "step": 31752 }, { "epoch": 2.49, "learning_rate": 7.510154473457225e-05, "loss": 2.7224, "step": 31753 }, { "epoch": 2.49, "learning_rate": 7.510076060534777e-05, "loss": 2.7091, "step": 31754 }, { "epoch": 2.49, "learning_rate": 7.509997647612327e-05, "loss": 2.5676, "step": 31755 }, { "epoch": 2.49, "learning_rate": 7.509919234689878e-05, "loss": 2.7544, "step": 31756 }, { "epoch": 2.49, "learning_rate": 7.509840821767428e-05, "loss": 3.0035, "step": 31757 }, { "epoch": 2.49, "learning_rate": 7.509762408844979e-05, "loss": 2.7961, "step": 31758 }, { "epoch": 2.49, "learning_rate": 7.509683995922529e-05, "loss": 2.8388, "step": 31759 }, { "epoch": 2.49, "learning_rate": 7.509605583000079e-05, "loss": 2.6799, "step": 31760 }, { "epoch": 2.49, "learning_rate": 7.50952717007763e-05, "loss": 2.7186, "step": 31761 }, { "epoch": 2.49, "learning_rate": 7.50944875715518e-05, "loss": 2.7589, "step": 31762 }, { "epoch": 2.49, "learning_rate": 7.50937034423273e-05, "loss": 2.7503, "step": 31763 }, { "epoch": 2.49, "learning_rate": 7.509291931310281e-05, "loss": 2.7985, "step": 31764 }, { "epoch": 2.49, "learning_rate": 7.509213518387831e-05, "loss": 2.776, "step": 31765 }, { "epoch": 2.49, "learning_rate": 7.509135105465381e-05, "loss": 2.8588, "step": 31766 }, { "epoch": 2.49, "learning_rate": 7.509056692542932e-05, "loss": 2.6595, "step": 31767 }, { "epoch": 2.49, "learning_rate": 7.508978279620482e-05, "loss": 2.9755, "step": 31768 }, { "epoch": 2.49, "learning_rate": 7.508899866698033e-05, "loss": 3.2317, "step": 31769 }, { "epoch": 2.49, "learning_rate": 7.508821453775583e-05, "loss": 2.5112, "step": 31770 }, { "epoch": 2.49, "learning_rate": 7.508743040853133e-05, "loss": 2.8982, "step": 31771 }, { "epoch": 2.49, "learning_rate": 7.508664627930684e-05, "loss": 2.7076, "step": 31772 }, { "epoch": 2.49, "learning_rate": 7.508586215008234e-05, "loss": 2.9385, "step": 31773 }, { "epoch": 2.49, "learning_rate": 7.508507802085784e-05, "loss": 2.6903, "step": 31774 }, { "epoch": 2.49, "learning_rate": 7.508429389163333e-05, "loss": 2.8207, "step": 31775 }, { "epoch": 2.49, "learning_rate": 7.508350976240884e-05, "loss": 2.8559, "step": 31776 }, { "epoch": 2.49, "learning_rate": 7.508272563318434e-05, "loss": 2.719, "step": 31777 }, { "epoch": 2.49, "learning_rate": 7.508194150395985e-05, "loss": 2.5861, "step": 31778 }, { "epoch": 2.49, "learning_rate": 7.508115737473535e-05, "loss": 2.9171, "step": 31779 }, { "epoch": 2.49, "learning_rate": 7.508037324551087e-05, "loss": 2.9771, "step": 31780 }, { "epoch": 2.49, "learning_rate": 7.507958911628637e-05, "loss": 2.925, "step": 31781 }, { "epoch": 2.49, "learning_rate": 7.507880498706187e-05, "loss": 2.8332, "step": 31782 }, { "epoch": 2.49, "learning_rate": 7.507802085783738e-05, "loss": 3.0849, "step": 31783 }, { "epoch": 2.49, "learning_rate": 7.507723672861288e-05, "loss": 2.979, "step": 31784 }, { "epoch": 2.49, "learning_rate": 7.507645259938838e-05, "loss": 2.8026, "step": 31785 }, { "epoch": 2.49, "learning_rate": 7.507566847016389e-05, "loss": 2.6386, "step": 31786 }, { "epoch": 2.49, "learning_rate": 7.507488434093939e-05, "loss": 2.9716, "step": 31787 }, { "epoch": 2.49, "learning_rate": 7.50741002117149e-05, "loss": 2.8448, "step": 31788 }, { "epoch": 2.49, "learning_rate": 7.50733160824904e-05, "loss": 2.761, "step": 31789 }, { "epoch": 2.49, "learning_rate": 7.50725319532659e-05, "loss": 3.0208, "step": 31790 }, { "epoch": 2.49, "learning_rate": 7.507174782404141e-05, "loss": 2.7664, "step": 31791 }, { "epoch": 2.49, "learning_rate": 7.507096369481691e-05, "loss": 3.0045, "step": 31792 }, { "epoch": 2.49, "learning_rate": 7.507017956559241e-05, "loss": 2.8289, "step": 31793 }, { "epoch": 2.49, "learning_rate": 7.506939543636792e-05, "loss": 2.7429, "step": 31794 }, { "epoch": 2.49, "learning_rate": 7.506861130714342e-05, "loss": 2.7889, "step": 31795 }, { "epoch": 2.49, "learning_rate": 7.506782717791893e-05, "loss": 2.5474, "step": 31796 }, { "epoch": 2.49, "learning_rate": 7.506704304869443e-05, "loss": 2.7285, "step": 31797 }, { "epoch": 2.49, "learning_rate": 7.506625891946993e-05, "loss": 2.6332, "step": 31798 }, { "epoch": 2.49, "learning_rate": 7.506547479024544e-05, "loss": 3.1466, "step": 31799 }, { "epoch": 2.49, "learning_rate": 7.506469066102094e-05, "loss": 2.7354, "step": 31800 }, { "epoch": 2.49, "learning_rate": 7.506390653179644e-05, "loss": 2.8465, "step": 31801 }, { "epoch": 2.49, "learning_rate": 7.506312240257195e-05, "loss": 2.8314, "step": 31802 }, { "epoch": 2.49, "learning_rate": 7.506233827334745e-05, "loss": 2.8589, "step": 31803 }, { "epoch": 2.49, "learning_rate": 7.506155414412295e-05, "loss": 2.813, "step": 31804 }, { "epoch": 2.49, "learning_rate": 7.506077001489846e-05, "loss": 2.8207, "step": 31805 }, { "epoch": 2.49, "learning_rate": 7.505998588567396e-05, "loss": 2.7827, "step": 31806 }, { "epoch": 2.49, "learning_rate": 7.505920175644947e-05, "loss": 2.9472, "step": 31807 }, { "epoch": 2.49, "learning_rate": 7.505841762722497e-05, "loss": 2.5981, "step": 31808 }, { "epoch": 2.49, "learning_rate": 7.505763349800047e-05, "loss": 2.7855, "step": 31809 }, { "epoch": 2.49, "learning_rate": 7.505684936877598e-05, "loss": 2.9375, "step": 31810 }, { "epoch": 2.49, "learning_rate": 7.505606523955148e-05, "loss": 2.8571, "step": 31811 }, { "epoch": 2.49, "learning_rate": 7.505528111032698e-05, "loss": 2.7115, "step": 31812 }, { "epoch": 2.49, "learning_rate": 7.505449698110249e-05, "loss": 2.8064, "step": 31813 }, { "epoch": 2.49, "learning_rate": 7.505371285187799e-05, "loss": 2.8161, "step": 31814 }, { "epoch": 2.49, "learning_rate": 7.50529287226535e-05, "loss": 2.7232, "step": 31815 }, { "epoch": 2.49, "learning_rate": 7.5052144593429e-05, "loss": 2.8488, "step": 31816 }, { "epoch": 2.49, "learning_rate": 7.50513604642045e-05, "loss": 2.8154, "step": 31817 }, { "epoch": 2.49, "learning_rate": 7.505057633498e-05, "loss": 2.6996, "step": 31818 }, { "epoch": 2.5, "learning_rate": 7.504979220575551e-05, "loss": 2.9564, "step": 31819 }, { "epoch": 2.5, "learning_rate": 7.504900807653101e-05, "loss": 2.8089, "step": 31820 }, { "epoch": 2.5, "learning_rate": 7.504822394730652e-05, "loss": 2.9875, "step": 31821 }, { "epoch": 2.5, "learning_rate": 7.504743981808202e-05, "loss": 2.7123, "step": 31822 }, { "epoch": 2.5, "learning_rate": 7.504665568885752e-05, "loss": 2.719, "step": 31823 }, { "epoch": 2.5, "learning_rate": 7.504587155963303e-05, "loss": 2.9622, "step": 31824 }, { "epoch": 2.5, "learning_rate": 7.504508743040853e-05, "loss": 2.7855, "step": 31825 }, { "epoch": 2.5, "learning_rate": 7.504430330118404e-05, "loss": 2.9003, "step": 31826 }, { "epoch": 2.5, "learning_rate": 7.504351917195954e-05, "loss": 2.8993, "step": 31827 }, { "epoch": 2.5, "learning_rate": 7.504273504273504e-05, "loss": 3.1277, "step": 31828 }, { "epoch": 2.5, "learning_rate": 7.504195091351055e-05, "loss": 2.8488, "step": 31829 }, { "epoch": 2.5, "learning_rate": 7.504116678428605e-05, "loss": 2.8566, "step": 31830 }, { "epoch": 2.5, "learning_rate": 7.504038265506155e-05, "loss": 2.6398, "step": 31831 }, { "epoch": 2.5, "learning_rate": 7.503959852583706e-05, "loss": 2.8784, "step": 31832 }, { "epoch": 2.5, "learning_rate": 7.503881439661256e-05, "loss": 2.7788, "step": 31833 }, { "epoch": 2.5, "learning_rate": 7.503803026738807e-05, "loss": 2.7724, "step": 31834 }, { "epoch": 2.5, "learning_rate": 7.503724613816357e-05, "loss": 2.8537, "step": 31835 }, { "epoch": 2.5, "learning_rate": 7.503646200893907e-05, "loss": 2.7596, "step": 31836 }, { "epoch": 2.5, "learning_rate": 7.503567787971459e-05, "loss": 2.8165, "step": 31837 }, { "epoch": 2.5, "learning_rate": 7.50348937504901e-05, "loss": 2.8085, "step": 31838 }, { "epoch": 2.5, "learning_rate": 7.50341096212656e-05, "loss": 2.8356, "step": 31839 }, { "epoch": 2.5, "learning_rate": 7.50333254920411e-05, "loss": 2.5613, "step": 31840 }, { "epoch": 2.5, "learning_rate": 7.50325413628166e-05, "loss": 2.7412, "step": 31841 }, { "epoch": 2.5, "learning_rate": 7.503175723359211e-05, "loss": 2.923, "step": 31842 }, { "epoch": 2.5, "learning_rate": 7.503097310436761e-05, "loss": 2.7376, "step": 31843 }, { "epoch": 2.5, "learning_rate": 7.503018897514312e-05, "loss": 2.8888, "step": 31844 }, { "epoch": 2.5, "learning_rate": 7.502940484591862e-05, "loss": 2.8482, "step": 31845 }, { "epoch": 2.5, "learning_rate": 7.502862071669412e-05, "loss": 2.6042, "step": 31846 }, { "epoch": 2.5, "learning_rate": 7.502783658746961e-05, "loss": 2.7272, "step": 31847 }, { "epoch": 2.5, "learning_rate": 7.502705245824512e-05, "loss": 2.9554, "step": 31848 }, { "epoch": 2.5, "learning_rate": 7.502626832902062e-05, "loss": 2.7753, "step": 31849 }, { "epoch": 2.5, "learning_rate": 7.502548419979612e-05, "loss": 2.7244, "step": 31850 }, { "epoch": 2.5, "learning_rate": 7.502470007057163e-05, "loss": 2.6671, "step": 31851 }, { "epoch": 2.5, "learning_rate": 7.502391594134713e-05, "loss": 3.1303, "step": 31852 }, { "epoch": 2.5, "learning_rate": 7.502313181212263e-05, "loss": 2.7752, "step": 31853 }, { "epoch": 2.5, "learning_rate": 7.502234768289814e-05, "loss": 2.7271, "step": 31854 }, { "epoch": 2.5, "learning_rate": 7.502156355367364e-05, "loss": 2.8247, "step": 31855 }, { "epoch": 2.5, "learning_rate": 7.502077942444915e-05, "loss": 3.1147, "step": 31856 }, { "epoch": 2.5, "learning_rate": 7.501999529522465e-05, "loss": 2.7581, "step": 31857 }, { "epoch": 2.5, "learning_rate": 7.501921116600015e-05, "loss": 2.67, "step": 31858 }, { "epoch": 2.5, "learning_rate": 7.501842703677566e-05, "loss": 2.8263, "step": 31859 }, { "epoch": 2.5, "learning_rate": 7.501764290755116e-05, "loss": 2.8132, "step": 31860 }, { "epoch": 2.5, "learning_rate": 7.501685877832666e-05, "loss": 2.9323, "step": 31861 }, { "epoch": 2.5, "learning_rate": 7.501607464910218e-05, "loss": 2.9838, "step": 31862 }, { "epoch": 2.5, "learning_rate": 7.501529051987769e-05, "loss": 2.989, "step": 31863 }, { "epoch": 2.5, "learning_rate": 7.501450639065319e-05, "loss": 2.8293, "step": 31864 }, { "epoch": 2.5, "learning_rate": 7.501372226142869e-05, "loss": 2.8068, "step": 31865 }, { "epoch": 2.5, "learning_rate": 7.50129381322042e-05, "loss": 2.9094, "step": 31866 }, { "epoch": 2.5, "learning_rate": 7.50121540029797e-05, "loss": 2.8681, "step": 31867 }, { "epoch": 2.5, "learning_rate": 7.50113698737552e-05, "loss": 2.6297, "step": 31868 }, { "epoch": 2.5, "learning_rate": 7.501058574453071e-05, "loss": 2.7981, "step": 31869 }, { "epoch": 2.5, "learning_rate": 7.500980161530621e-05, "loss": 2.7757, "step": 31870 }, { "epoch": 2.5, "learning_rate": 7.500901748608171e-05, "loss": 2.7589, "step": 31871 }, { "epoch": 2.5, "learning_rate": 7.500823335685722e-05, "loss": 2.9023, "step": 31872 }, { "epoch": 2.5, "learning_rate": 7.500744922763272e-05, "loss": 3.0126, "step": 31873 }, { "epoch": 2.5, "learning_rate": 7.500666509840823e-05, "loss": 3.0442, "step": 31874 }, { "epoch": 2.5, "learning_rate": 7.500588096918373e-05, "loss": 2.8444, "step": 31875 }, { "epoch": 2.5, "learning_rate": 7.500509683995923e-05, "loss": 2.9098, "step": 31876 }, { "epoch": 2.5, "learning_rate": 7.500431271073474e-05, "loss": 2.8615, "step": 31877 }, { "epoch": 2.5, "learning_rate": 7.500352858151024e-05, "loss": 2.6829, "step": 31878 }, { "epoch": 2.5, "learning_rate": 7.500274445228574e-05, "loss": 2.7186, "step": 31879 }, { "epoch": 2.5, "learning_rate": 7.500196032306125e-05, "loss": 2.7889, "step": 31880 }, { "epoch": 2.5, "learning_rate": 7.500117619383675e-05, "loss": 2.6361, "step": 31881 }, { "epoch": 2.5, "learning_rate": 7.500039206461226e-05, "loss": 2.8385, "step": 31882 }, { "epoch": 2.5, "learning_rate": 7.499960793538775e-05, "loss": 3.0251, "step": 31883 }, { "epoch": 2.5, "learning_rate": 7.499882380616325e-05, "loss": 2.9741, "step": 31884 }, { "epoch": 2.5, "learning_rate": 7.499803967693875e-05, "loss": 2.9864, "step": 31885 }, { "epoch": 2.5, "learning_rate": 7.499725554771426e-05, "loss": 2.904, "step": 31886 }, { "epoch": 2.5, "learning_rate": 7.499647141848976e-05, "loss": 2.6825, "step": 31887 }, { "epoch": 2.5, "learning_rate": 7.499568728926528e-05, "loss": 2.9741, "step": 31888 }, { "epoch": 2.5, "learning_rate": 7.499490316004078e-05, "loss": 2.7591, "step": 31889 }, { "epoch": 2.5, "learning_rate": 7.499411903081628e-05, "loss": 2.646, "step": 31890 }, { "epoch": 2.5, "learning_rate": 7.499333490159179e-05, "loss": 2.8165, "step": 31891 }, { "epoch": 2.5, "learning_rate": 7.499255077236729e-05, "loss": 2.6565, "step": 31892 }, { "epoch": 2.5, "learning_rate": 7.49917666431428e-05, "loss": 2.8586, "step": 31893 }, { "epoch": 2.5, "learning_rate": 7.49909825139183e-05, "loss": 2.8348, "step": 31894 }, { "epoch": 2.5, "learning_rate": 7.49901983846938e-05, "loss": 2.7165, "step": 31895 }, { "epoch": 2.5, "learning_rate": 7.49894142554693e-05, "loss": 2.6661, "step": 31896 }, { "epoch": 2.5, "learning_rate": 7.498863012624481e-05, "loss": 2.7668, "step": 31897 }, { "epoch": 2.5, "learning_rate": 7.498784599702031e-05, "loss": 2.8522, "step": 31898 }, { "epoch": 2.5, "learning_rate": 7.498706186779582e-05, "loss": 2.8914, "step": 31899 }, { "epoch": 2.5, "learning_rate": 7.498627773857132e-05, "loss": 2.5475, "step": 31900 }, { "epoch": 2.5, "learning_rate": 7.498549360934683e-05, "loss": 2.6349, "step": 31901 }, { "epoch": 2.5, "learning_rate": 7.498470948012233e-05, "loss": 2.8347, "step": 31902 }, { "epoch": 2.5, "learning_rate": 7.498392535089783e-05, "loss": 2.7365, "step": 31903 }, { "epoch": 2.5, "learning_rate": 7.498314122167334e-05, "loss": 2.9548, "step": 31904 }, { "epoch": 2.5, "learning_rate": 7.498235709244884e-05, "loss": 2.8216, "step": 31905 }, { "epoch": 2.5, "learning_rate": 7.498157296322434e-05, "loss": 2.6881, "step": 31906 }, { "epoch": 2.5, "learning_rate": 7.498078883399985e-05, "loss": 3.0888, "step": 31907 }, { "epoch": 2.5, "learning_rate": 7.498000470477535e-05, "loss": 2.8234, "step": 31908 }, { "epoch": 2.5, "learning_rate": 7.497922057555085e-05, "loss": 2.7376, "step": 31909 }, { "epoch": 2.5, "learning_rate": 7.497843644632636e-05, "loss": 2.822, "step": 31910 }, { "epoch": 2.5, "learning_rate": 7.497765231710186e-05, "loss": 2.7379, "step": 31911 }, { "epoch": 2.5, "learning_rate": 7.497686818787737e-05, "loss": 2.8923, "step": 31912 }, { "epoch": 2.5, "learning_rate": 7.497608405865287e-05, "loss": 3.0043, "step": 31913 }, { "epoch": 2.5, "learning_rate": 7.497529992942837e-05, "loss": 2.5971, "step": 31914 }, { "epoch": 2.5, "learning_rate": 7.497451580020388e-05, "loss": 2.7678, "step": 31915 }, { "epoch": 2.5, "learning_rate": 7.497373167097938e-05, "loss": 3.0019, "step": 31916 }, { "epoch": 2.5, "learning_rate": 7.497294754175488e-05, "loss": 2.9352, "step": 31917 }, { "epoch": 2.5, "learning_rate": 7.497216341253039e-05, "loss": 2.8222, "step": 31918 }, { "epoch": 2.5, "learning_rate": 7.497137928330589e-05, "loss": 2.8955, "step": 31919 }, { "epoch": 2.5, "learning_rate": 7.49705951540814e-05, "loss": 2.9372, "step": 31920 }, { "epoch": 2.5, "learning_rate": 7.49698110248569e-05, "loss": 2.8387, "step": 31921 }, { "epoch": 2.5, "learning_rate": 7.49690268956324e-05, "loss": 2.8169, "step": 31922 }, { "epoch": 2.5, "learning_rate": 7.49682427664079e-05, "loss": 2.806, "step": 31923 }, { "epoch": 2.5, "learning_rate": 7.496745863718341e-05, "loss": 3.0026, "step": 31924 }, { "epoch": 2.5, "learning_rate": 7.496667450795891e-05, "loss": 2.8096, "step": 31925 }, { "epoch": 2.5, "learning_rate": 7.496589037873442e-05, "loss": 3.0746, "step": 31926 }, { "epoch": 2.5, "learning_rate": 7.496510624950992e-05, "loss": 2.4609, "step": 31927 }, { "epoch": 2.5, "learning_rate": 7.496432212028542e-05, "loss": 2.9924, "step": 31928 }, { "epoch": 2.5, "learning_rate": 7.496353799106093e-05, "loss": 2.8178, "step": 31929 }, { "epoch": 2.5, "learning_rate": 7.496275386183643e-05, "loss": 2.668, "step": 31930 }, { "epoch": 2.5, "learning_rate": 7.496196973261194e-05, "loss": 2.7445, "step": 31931 }, { "epoch": 2.5, "learning_rate": 7.496118560338744e-05, "loss": 2.8268, "step": 31932 }, { "epoch": 2.5, "learning_rate": 7.496040147416294e-05, "loss": 2.9122, "step": 31933 }, { "epoch": 2.5, "learning_rate": 7.495961734493845e-05, "loss": 2.841, "step": 31934 }, { "epoch": 2.5, "learning_rate": 7.495883321571395e-05, "loss": 2.6435, "step": 31935 }, { "epoch": 2.5, "learning_rate": 7.495804908648945e-05, "loss": 2.7177, "step": 31936 }, { "epoch": 2.5, "learning_rate": 7.495726495726496e-05, "loss": 2.6621, "step": 31937 }, { "epoch": 2.5, "learning_rate": 7.495648082804046e-05, "loss": 2.8058, "step": 31938 }, { "epoch": 2.5, "learning_rate": 7.495569669881596e-05, "loss": 2.5358, "step": 31939 }, { "epoch": 2.5, "learning_rate": 7.495491256959147e-05, "loss": 2.7413, "step": 31940 }, { "epoch": 2.5, "learning_rate": 7.495412844036697e-05, "loss": 2.6814, "step": 31941 }, { "epoch": 2.5, "learning_rate": 7.495334431114248e-05, "loss": 2.9122, "step": 31942 }, { "epoch": 2.5, "learning_rate": 7.495256018191798e-05, "loss": 2.8114, "step": 31943 }, { "epoch": 2.5, "learning_rate": 7.495177605269348e-05, "loss": 2.6723, "step": 31944 }, { "epoch": 2.5, "learning_rate": 7.4950991923469e-05, "loss": 2.9864, "step": 31945 }, { "epoch": 2.5, "learning_rate": 7.49502077942445e-05, "loss": 2.8836, "step": 31946 }, { "epoch": 2.51, "learning_rate": 7.494942366502001e-05, "loss": 2.7152, "step": 31947 }, { "epoch": 2.51, "learning_rate": 7.494863953579551e-05, "loss": 2.9008, "step": 31948 }, { "epoch": 2.51, "learning_rate": 7.494785540657102e-05, "loss": 2.6841, "step": 31949 }, { "epoch": 2.51, "learning_rate": 7.494707127734652e-05, "loss": 2.5795, "step": 31950 }, { "epoch": 2.51, "learning_rate": 7.494628714812202e-05, "loss": 3.0441, "step": 31951 }, { "epoch": 2.51, "learning_rate": 7.494550301889753e-05, "loss": 2.7638, "step": 31952 }, { "epoch": 2.51, "learning_rate": 7.494471888967303e-05, "loss": 2.868, "step": 31953 }, { "epoch": 2.51, "learning_rate": 7.494393476044853e-05, "loss": 2.7442, "step": 31954 }, { "epoch": 2.51, "learning_rate": 7.494315063122402e-05, "loss": 2.7066, "step": 31955 }, { "epoch": 2.51, "learning_rate": 7.494236650199953e-05, "loss": 2.6297, "step": 31956 }, { "epoch": 2.51, "learning_rate": 7.494158237277503e-05, "loss": 2.7614, "step": 31957 }, { "epoch": 2.51, "learning_rate": 7.494079824355053e-05, "loss": 2.7979, "step": 31958 }, { "epoch": 2.51, "learning_rate": 7.494001411432604e-05, "loss": 2.5913, "step": 31959 }, { "epoch": 2.51, "learning_rate": 7.493922998510154e-05, "loss": 2.8221, "step": 31960 }, { "epoch": 2.51, "learning_rate": 7.493844585587705e-05, "loss": 2.8812, "step": 31961 }, { "epoch": 2.51, "learning_rate": 7.493766172665255e-05, "loss": 2.8658, "step": 31962 }, { "epoch": 2.51, "learning_rate": 7.493687759742805e-05, "loss": 2.7061, "step": 31963 }, { "epoch": 2.51, "learning_rate": 7.493609346820356e-05, "loss": 2.6849, "step": 31964 }, { "epoch": 2.51, "learning_rate": 7.493530933897906e-05, "loss": 2.9511, "step": 31965 }, { "epoch": 2.51, "learning_rate": 7.493452520975456e-05, "loss": 2.82, "step": 31966 }, { "epoch": 2.51, "learning_rate": 7.493374108053007e-05, "loss": 3.0486, "step": 31967 }, { "epoch": 2.51, "learning_rate": 7.493295695130557e-05, "loss": 2.7156, "step": 31968 }, { "epoch": 2.51, "learning_rate": 7.493217282208108e-05, "loss": 2.6262, "step": 31969 }, { "epoch": 2.51, "learning_rate": 7.493138869285659e-05, "loss": 2.8533, "step": 31970 }, { "epoch": 2.51, "learning_rate": 7.49306045636321e-05, "loss": 2.8484, "step": 31971 }, { "epoch": 2.51, "learning_rate": 7.49298204344076e-05, "loss": 2.9201, "step": 31972 }, { "epoch": 2.51, "learning_rate": 7.49290363051831e-05, "loss": 2.7745, "step": 31973 }, { "epoch": 2.51, "learning_rate": 7.492825217595861e-05, "loss": 2.6435, "step": 31974 }, { "epoch": 2.51, "learning_rate": 7.492746804673411e-05, "loss": 2.6093, "step": 31975 }, { "epoch": 2.51, "learning_rate": 7.492668391750961e-05, "loss": 2.823, "step": 31976 }, { "epoch": 2.51, "learning_rate": 7.492589978828512e-05, "loss": 2.9156, "step": 31977 }, { "epoch": 2.51, "learning_rate": 7.492511565906062e-05, "loss": 2.5355, "step": 31978 }, { "epoch": 2.51, "learning_rate": 7.492433152983613e-05, "loss": 2.9221, "step": 31979 }, { "epoch": 2.51, "learning_rate": 7.492354740061163e-05, "loss": 2.6435, "step": 31980 }, { "epoch": 2.51, "learning_rate": 7.492276327138713e-05, "loss": 3.0156, "step": 31981 }, { "epoch": 2.51, "learning_rate": 7.492197914216264e-05, "loss": 2.726, "step": 31982 }, { "epoch": 2.51, "learning_rate": 7.492119501293814e-05, "loss": 2.8137, "step": 31983 }, { "epoch": 2.51, "learning_rate": 7.492041088371364e-05, "loss": 2.7689, "step": 31984 }, { "epoch": 2.51, "learning_rate": 7.491962675448915e-05, "loss": 2.886, "step": 31985 }, { "epoch": 2.51, "learning_rate": 7.491884262526465e-05, "loss": 2.8575, "step": 31986 }, { "epoch": 2.51, "learning_rate": 7.491805849604015e-05, "loss": 2.8707, "step": 31987 }, { "epoch": 2.51, "learning_rate": 7.491727436681566e-05, "loss": 2.8164, "step": 31988 }, { "epoch": 2.51, "learning_rate": 7.491649023759116e-05, "loss": 2.824, "step": 31989 }, { "epoch": 2.51, "learning_rate": 7.491570610836667e-05, "loss": 3.0421, "step": 31990 }, { "epoch": 2.51, "learning_rate": 7.491492197914216e-05, "loss": 2.7136, "step": 31991 }, { "epoch": 2.51, "learning_rate": 7.491413784991766e-05, "loss": 2.7734, "step": 31992 }, { "epoch": 2.51, "learning_rate": 7.491335372069316e-05, "loss": 2.7477, "step": 31993 }, { "epoch": 2.51, "learning_rate": 7.491256959146867e-05, "loss": 2.585, "step": 31994 }, { "epoch": 2.51, "learning_rate": 7.491178546224417e-05, "loss": 2.654, "step": 31995 }, { "epoch": 2.51, "learning_rate": 7.491100133301969e-05, "loss": 2.6424, "step": 31996 }, { "epoch": 2.51, "learning_rate": 7.491021720379519e-05, "loss": 2.7778, "step": 31997 }, { "epoch": 2.51, "learning_rate": 7.49094330745707e-05, "loss": 2.9207, "step": 31998 }, { "epoch": 2.51, "learning_rate": 7.49086489453462e-05, "loss": 3.0949, "step": 31999 }, { "epoch": 2.51, "learning_rate": 7.49078648161217e-05, "loss": 2.914, "step": 32000 }, { "epoch": 2.51, "eval_loss": 2.899564743041992, "eval_runtime": 218.1034, "eval_samples_per_second": 51.957, "eval_steps_per_second": 3.251, "step": 32000 }, { "epoch": 2.51, "learning_rate": 7.49070806868972e-05, "loss": 2.9093, "step": 32001 }, { "epoch": 2.51, "learning_rate": 7.490629655767271e-05, "loss": 2.6845, "step": 32002 }, { "epoch": 2.51, "learning_rate": 7.490551242844821e-05, "loss": 2.9133, "step": 32003 }, { "epoch": 2.51, "learning_rate": 7.490472829922372e-05, "loss": 2.6743, "step": 32004 }, { "epoch": 2.51, "learning_rate": 7.490394416999922e-05, "loss": 2.8602, "step": 32005 }, { "epoch": 2.51, "learning_rate": 7.490316004077472e-05, "loss": 2.7023, "step": 32006 }, { "epoch": 2.51, "learning_rate": 7.490237591155023e-05, "loss": 2.7396, "step": 32007 }, { "epoch": 2.51, "learning_rate": 7.490159178232573e-05, "loss": 2.747, "step": 32008 }, { "epoch": 2.51, "learning_rate": 7.490080765310124e-05, "loss": 2.9241, "step": 32009 }, { "epoch": 2.51, "learning_rate": 7.490002352387674e-05, "loss": 2.7945, "step": 32010 }, { "epoch": 2.51, "learning_rate": 7.489923939465224e-05, "loss": 2.9699, "step": 32011 }, { "epoch": 2.51, "learning_rate": 7.489845526542775e-05, "loss": 2.9375, "step": 32012 }, { "epoch": 2.51, "learning_rate": 7.489767113620325e-05, "loss": 2.6822, "step": 32013 }, { "epoch": 2.51, "learning_rate": 7.489688700697875e-05, "loss": 2.9516, "step": 32014 }, { "epoch": 2.51, "learning_rate": 7.489610287775426e-05, "loss": 2.5755, "step": 32015 }, { "epoch": 2.51, "learning_rate": 7.489531874852976e-05, "loss": 3.0868, "step": 32016 }, { "epoch": 2.51, "learning_rate": 7.489453461930527e-05, "loss": 2.7544, "step": 32017 }, { "epoch": 2.51, "learning_rate": 7.489375049008077e-05, "loss": 3.0093, "step": 32018 }, { "epoch": 2.51, "learning_rate": 7.489296636085627e-05, "loss": 2.9718, "step": 32019 }, { "epoch": 2.51, "learning_rate": 7.489218223163178e-05, "loss": 2.6485, "step": 32020 }, { "epoch": 2.51, "learning_rate": 7.489139810240728e-05, "loss": 3.1646, "step": 32021 }, { "epoch": 2.51, "learning_rate": 7.489061397318278e-05, "loss": 2.8901, "step": 32022 }, { "epoch": 2.51, "learning_rate": 7.488982984395829e-05, "loss": 2.6385, "step": 32023 }, { "epoch": 2.51, "learning_rate": 7.488904571473379e-05, "loss": 2.7234, "step": 32024 }, { "epoch": 2.51, "learning_rate": 7.48882615855093e-05, "loss": 2.9009, "step": 32025 }, { "epoch": 2.51, "learning_rate": 7.48874774562848e-05, "loss": 2.6817, "step": 32026 }, { "epoch": 2.51, "learning_rate": 7.48866933270603e-05, "loss": 2.784, "step": 32027 }, { "epoch": 2.51, "learning_rate": 7.48859091978358e-05, "loss": 2.6682, "step": 32028 }, { "epoch": 2.51, "learning_rate": 7.488512506861131e-05, "loss": 2.8863, "step": 32029 }, { "epoch": 2.51, "learning_rate": 7.488434093938681e-05, "loss": 2.726, "step": 32030 }, { "epoch": 2.51, "learning_rate": 7.488355681016232e-05, "loss": 2.6315, "step": 32031 }, { "epoch": 2.51, "learning_rate": 7.488277268093782e-05, "loss": 2.9202, "step": 32032 }, { "epoch": 2.51, "learning_rate": 7.488198855171332e-05, "loss": 2.8486, "step": 32033 }, { "epoch": 2.51, "learning_rate": 7.488120442248883e-05, "loss": 2.7097, "step": 32034 }, { "epoch": 2.51, "learning_rate": 7.488042029326433e-05, "loss": 2.8379, "step": 32035 }, { "epoch": 2.51, "learning_rate": 7.487963616403984e-05, "loss": 2.7271, "step": 32036 }, { "epoch": 2.51, "learning_rate": 7.487885203481534e-05, "loss": 2.8234, "step": 32037 }, { "epoch": 2.51, "learning_rate": 7.487806790559084e-05, "loss": 2.8274, "step": 32038 }, { "epoch": 2.51, "learning_rate": 7.487728377636635e-05, "loss": 2.8489, "step": 32039 }, { "epoch": 2.51, "learning_rate": 7.487649964714185e-05, "loss": 2.941, "step": 32040 }, { "epoch": 2.51, "learning_rate": 7.487571551791735e-05, "loss": 2.7016, "step": 32041 }, { "epoch": 2.51, "learning_rate": 7.487493138869286e-05, "loss": 2.7144, "step": 32042 }, { "epoch": 2.51, "learning_rate": 7.487414725946836e-05, "loss": 3.0841, "step": 32043 }, { "epoch": 2.51, "learning_rate": 7.487336313024386e-05, "loss": 2.7881, "step": 32044 }, { "epoch": 2.51, "learning_rate": 7.487257900101937e-05, "loss": 2.8094, "step": 32045 }, { "epoch": 2.51, "learning_rate": 7.487179487179487e-05, "loss": 2.6676, "step": 32046 }, { "epoch": 2.51, "learning_rate": 7.487101074257038e-05, "loss": 2.7693, "step": 32047 }, { "epoch": 2.51, "learning_rate": 7.487022661334588e-05, "loss": 2.6777, "step": 32048 }, { "epoch": 2.51, "learning_rate": 7.486944248412138e-05, "loss": 2.8938, "step": 32049 }, { "epoch": 2.51, "learning_rate": 7.486865835489689e-05, "loss": 2.8357, "step": 32050 }, { "epoch": 2.51, "learning_rate": 7.486787422567239e-05, "loss": 2.6956, "step": 32051 }, { "epoch": 2.51, "learning_rate": 7.48670900964479e-05, "loss": 2.7035, "step": 32052 }, { "epoch": 2.51, "learning_rate": 7.486630596722341e-05, "loss": 2.8406, "step": 32053 }, { "epoch": 2.51, "learning_rate": 7.486552183799891e-05, "loss": 3.0299, "step": 32054 }, { "epoch": 2.51, "learning_rate": 7.486473770877442e-05, "loss": 2.9091, "step": 32055 }, { "epoch": 2.51, "learning_rate": 7.486395357954992e-05, "loss": 2.6855, "step": 32056 }, { "epoch": 2.51, "learning_rate": 7.486316945032543e-05, "loss": 2.625, "step": 32057 }, { "epoch": 2.51, "learning_rate": 7.486238532110093e-05, "loss": 2.7308, "step": 32058 }, { "epoch": 2.51, "learning_rate": 7.486160119187643e-05, "loss": 2.6454, "step": 32059 }, { "epoch": 2.51, "learning_rate": 7.486081706265194e-05, "loss": 2.7342, "step": 32060 }, { "epoch": 2.51, "learning_rate": 7.486003293342744e-05, "loss": 2.72, "step": 32061 }, { "epoch": 2.51, "learning_rate": 7.485924880420294e-05, "loss": 2.7536, "step": 32062 }, { "epoch": 2.51, "learning_rate": 7.485846467497843e-05, "loss": 2.8256, "step": 32063 }, { "epoch": 2.51, "learning_rate": 7.485768054575394e-05, "loss": 2.7658, "step": 32064 }, { "epoch": 2.51, "learning_rate": 7.485689641652944e-05, "loss": 2.7291, "step": 32065 }, { "epoch": 2.51, "learning_rate": 7.485611228730495e-05, "loss": 2.7171, "step": 32066 }, { "epoch": 2.51, "learning_rate": 7.485532815808045e-05, "loss": 2.8028, "step": 32067 }, { "epoch": 2.51, "learning_rate": 7.485454402885595e-05, "loss": 2.8216, "step": 32068 }, { "epoch": 2.51, "learning_rate": 7.485375989963146e-05, "loss": 2.8205, "step": 32069 }, { "epoch": 2.51, "learning_rate": 7.485297577040696e-05, "loss": 2.7219, "step": 32070 }, { "epoch": 2.51, "learning_rate": 7.485219164118246e-05, "loss": 3.0106, "step": 32071 }, { "epoch": 2.51, "learning_rate": 7.485140751195797e-05, "loss": 2.9428, "step": 32072 }, { "epoch": 2.51, "learning_rate": 7.485062338273347e-05, "loss": 2.8283, "step": 32073 }, { "epoch": 2.52, "learning_rate": 7.484983925350897e-05, "loss": 2.8276, "step": 32074 }, { "epoch": 2.52, "learning_rate": 7.484905512428448e-05, "loss": 2.8904, "step": 32075 }, { "epoch": 2.52, "learning_rate": 7.484827099505998e-05, "loss": 2.6546, "step": 32076 }, { "epoch": 2.52, "learning_rate": 7.484748686583549e-05, "loss": 2.7427, "step": 32077 }, { "epoch": 2.52, "learning_rate": 7.484670273661099e-05, "loss": 2.7803, "step": 32078 }, { "epoch": 2.52, "learning_rate": 7.48459186073865e-05, "loss": 2.8984, "step": 32079 }, { "epoch": 2.52, "learning_rate": 7.484513447816201e-05, "loss": 2.9299, "step": 32080 }, { "epoch": 2.52, "learning_rate": 7.484435034893751e-05, "loss": 2.8671, "step": 32081 }, { "epoch": 2.52, "learning_rate": 7.484356621971302e-05, "loss": 2.9328, "step": 32082 }, { "epoch": 2.52, "learning_rate": 7.484278209048852e-05, "loss": 3.0632, "step": 32083 }, { "epoch": 2.52, "learning_rate": 7.484199796126403e-05, "loss": 2.778, "step": 32084 }, { "epoch": 2.52, "learning_rate": 7.484121383203953e-05, "loss": 3.0632, "step": 32085 }, { "epoch": 2.52, "learning_rate": 7.484042970281503e-05, "loss": 2.6085, "step": 32086 }, { "epoch": 2.52, "learning_rate": 7.483964557359054e-05, "loss": 2.8409, "step": 32087 }, { "epoch": 2.52, "learning_rate": 7.483886144436604e-05, "loss": 2.7867, "step": 32088 }, { "epoch": 2.52, "learning_rate": 7.483807731514154e-05, "loss": 3.0788, "step": 32089 }, { "epoch": 2.52, "learning_rate": 7.483729318591705e-05, "loss": 2.7562, "step": 32090 }, { "epoch": 2.52, "learning_rate": 7.483650905669255e-05, "loss": 3.0121, "step": 32091 }, { "epoch": 2.52, "learning_rate": 7.483572492746805e-05, "loss": 2.9004, "step": 32092 }, { "epoch": 2.52, "learning_rate": 7.483494079824356e-05, "loss": 2.7222, "step": 32093 }, { "epoch": 2.52, "learning_rate": 7.483415666901906e-05, "loss": 2.9363, "step": 32094 }, { "epoch": 2.52, "learning_rate": 7.483337253979457e-05, "loss": 2.7503, "step": 32095 }, { "epoch": 2.52, "learning_rate": 7.483258841057007e-05, "loss": 2.8913, "step": 32096 }, { "epoch": 2.52, "learning_rate": 7.483180428134557e-05, "loss": 2.7596, "step": 32097 }, { "epoch": 2.52, "learning_rate": 7.483102015212108e-05, "loss": 2.5585, "step": 32098 }, { "epoch": 2.52, "learning_rate": 7.483023602289657e-05, "loss": 2.7317, "step": 32099 }, { "epoch": 2.52, "learning_rate": 7.482945189367207e-05, "loss": 2.7081, "step": 32100 }, { "epoch": 2.52, "learning_rate": 7.482866776444757e-05, "loss": 2.7668, "step": 32101 }, { "epoch": 2.52, "learning_rate": 7.482788363522308e-05, "loss": 2.7698, "step": 32102 }, { "epoch": 2.52, "learning_rate": 7.482709950599858e-05, "loss": 2.8342, "step": 32103 }, { "epoch": 2.52, "learning_rate": 7.48263153767741e-05, "loss": 2.6122, "step": 32104 }, { "epoch": 2.52, "learning_rate": 7.48255312475496e-05, "loss": 2.7536, "step": 32105 }, { "epoch": 2.52, "learning_rate": 7.48247471183251e-05, "loss": 2.8773, "step": 32106 }, { "epoch": 2.52, "learning_rate": 7.482396298910061e-05, "loss": 2.9296, "step": 32107 }, { "epoch": 2.52, "learning_rate": 7.482317885987611e-05, "loss": 3.1578, "step": 32108 }, { "epoch": 2.52, "learning_rate": 7.482239473065162e-05, "loss": 2.9453, "step": 32109 }, { "epoch": 2.52, "learning_rate": 7.482161060142712e-05, "loss": 2.7791, "step": 32110 }, { "epoch": 2.52, "learning_rate": 7.482082647220262e-05, "loss": 2.5167, "step": 32111 }, { "epoch": 2.52, "learning_rate": 7.482004234297813e-05, "loss": 2.8509, "step": 32112 }, { "epoch": 2.52, "learning_rate": 7.481925821375363e-05, "loss": 2.7262, "step": 32113 }, { "epoch": 2.52, "learning_rate": 7.481847408452914e-05, "loss": 2.6911, "step": 32114 }, { "epoch": 2.52, "learning_rate": 7.481768995530464e-05, "loss": 2.75, "step": 32115 }, { "epoch": 2.52, "learning_rate": 7.481690582608014e-05, "loss": 3.0634, "step": 32116 }, { "epoch": 2.52, "learning_rate": 7.481612169685565e-05, "loss": 2.8624, "step": 32117 }, { "epoch": 2.52, "learning_rate": 7.481533756763115e-05, "loss": 2.7673, "step": 32118 }, { "epoch": 2.52, "learning_rate": 7.481455343840665e-05, "loss": 2.7735, "step": 32119 }, { "epoch": 2.52, "learning_rate": 7.481376930918216e-05, "loss": 2.8081, "step": 32120 }, { "epoch": 2.52, "learning_rate": 7.481298517995766e-05, "loss": 2.8772, "step": 32121 }, { "epoch": 2.52, "learning_rate": 7.481220105073316e-05, "loss": 2.8631, "step": 32122 }, { "epoch": 2.52, "learning_rate": 7.481141692150867e-05, "loss": 2.7481, "step": 32123 }, { "epoch": 2.52, "learning_rate": 7.481063279228417e-05, "loss": 2.6028, "step": 32124 }, { "epoch": 2.52, "learning_rate": 7.480984866305968e-05, "loss": 3.0514, "step": 32125 }, { "epoch": 2.52, "learning_rate": 7.480906453383518e-05, "loss": 2.8278, "step": 32126 }, { "epoch": 2.52, "learning_rate": 7.480828040461068e-05, "loss": 2.7001, "step": 32127 }, { "epoch": 2.52, "learning_rate": 7.480749627538619e-05, "loss": 2.7688, "step": 32128 }, { "epoch": 2.52, "learning_rate": 7.480671214616169e-05, "loss": 2.889, "step": 32129 }, { "epoch": 2.52, "learning_rate": 7.48059280169372e-05, "loss": 2.855, "step": 32130 }, { "epoch": 2.52, "learning_rate": 7.48051438877127e-05, "loss": 2.9672, "step": 32131 }, { "epoch": 2.52, "learning_rate": 7.48043597584882e-05, "loss": 2.7283, "step": 32132 }, { "epoch": 2.52, "learning_rate": 7.48035756292637e-05, "loss": 2.8267, "step": 32133 }, { "epoch": 2.52, "learning_rate": 7.480279150003921e-05, "loss": 2.8088, "step": 32134 }, { "epoch": 2.52, "learning_rate": 7.480200737081471e-05, "loss": 2.7895, "step": 32135 }, { "epoch": 2.52, "learning_rate": 7.480122324159022e-05, "loss": 2.6212, "step": 32136 }, { "epoch": 2.52, "learning_rate": 7.480043911236572e-05, "loss": 2.7993, "step": 32137 }, { "epoch": 2.52, "learning_rate": 7.479965498314122e-05, "loss": 2.7074, "step": 32138 }, { "epoch": 2.52, "learning_rate": 7.479887085391673e-05, "loss": 2.853, "step": 32139 }, { "epoch": 2.52, "learning_rate": 7.479808672469223e-05, "loss": 2.7924, "step": 32140 }, { "epoch": 2.52, "learning_rate": 7.479730259546773e-05, "loss": 2.6344, "step": 32141 }, { "epoch": 2.52, "learning_rate": 7.479651846624324e-05, "loss": 3.1644, "step": 32142 }, { "epoch": 2.52, "learning_rate": 7.479573433701874e-05, "loss": 2.9135, "step": 32143 }, { "epoch": 2.52, "learning_rate": 7.479495020779425e-05, "loss": 2.6724, "step": 32144 }, { "epoch": 2.52, "learning_rate": 7.479416607856975e-05, "loss": 2.9633, "step": 32145 }, { "epoch": 2.52, "learning_rate": 7.479338194934525e-05, "loss": 2.7507, "step": 32146 }, { "epoch": 2.52, "learning_rate": 7.479259782012076e-05, "loss": 2.7641, "step": 32147 }, { "epoch": 2.52, "learning_rate": 7.479181369089626e-05, "loss": 2.9351, "step": 32148 }, { "epoch": 2.52, "learning_rate": 7.479102956167176e-05, "loss": 2.7535, "step": 32149 }, { "epoch": 2.52, "learning_rate": 7.479024543244727e-05, "loss": 2.8604, "step": 32150 }, { "epoch": 2.52, "learning_rate": 7.478946130322277e-05, "loss": 2.7851, "step": 32151 }, { "epoch": 2.52, "learning_rate": 7.478867717399828e-05, "loss": 2.7523, "step": 32152 }, { "epoch": 2.52, "learning_rate": 7.478789304477378e-05, "loss": 2.825, "step": 32153 }, { "epoch": 2.52, "learning_rate": 7.478710891554928e-05, "loss": 2.848, "step": 32154 }, { "epoch": 2.52, "learning_rate": 7.478632478632479e-05, "loss": 2.9653, "step": 32155 }, { "epoch": 2.52, "learning_rate": 7.478554065710029e-05, "loss": 2.8983, "step": 32156 }, { "epoch": 2.52, "learning_rate": 7.47847565278758e-05, "loss": 2.9176, "step": 32157 }, { "epoch": 2.52, "learning_rate": 7.47839723986513e-05, "loss": 2.8433, "step": 32158 }, { "epoch": 2.52, "learning_rate": 7.47831882694268e-05, "loss": 2.7667, "step": 32159 }, { "epoch": 2.52, "learning_rate": 7.47824041402023e-05, "loss": 2.8835, "step": 32160 }, { "epoch": 2.52, "learning_rate": 7.478162001097782e-05, "loss": 2.8155, "step": 32161 }, { "epoch": 2.52, "learning_rate": 7.478083588175333e-05, "loss": 2.8521, "step": 32162 }, { "epoch": 2.52, "learning_rate": 7.478005175252883e-05, "loss": 2.7889, "step": 32163 }, { "epoch": 2.52, "learning_rate": 7.477926762330433e-05, "loss": 2.858, "step": 32164 }, { "epoch": 2.52, "learning_rate": 7.477848349407984e-05, "loss": 3.0225, "step": 32165 }, { "epoch": 2.52, "learning_rate": 7.477769936485534e-05, "loss": 2.8212, "step": 32166 }, { "epoch": 2.52, "learning_rate": 7.477691523563084e-05, "loss": 2.7554, "step": 32167 }, { "epoch": 2.52, "learning_rate": 7.477613110640635e-05, "loss": 2.6719, "step": 32168 }, { "epoch": 2.52, "learning_rate": 7.477534697718185e-05, "loss": 2.6789, "step": 32169 }, { "epoch": 2.52, "learning_rate": 7.477456284795735e-05, "loss": 2.8815, "step": 32170 }, { "epoch": 2.52, "learning_rate": 7.477377871873285e-05, "loss": 2.7304, "step": 32171 }, { "epoch": 2.52, "learning_rate": 7.477299458950835e-05, "loss": 2.908, "step": 32172 }, { "epoch": 2.52, "learning_rate": 7.477221046028385e-05, "loss": 2.7612, "step": 32173 }, { "epoch": 2.52, "learning_rate": 7.477142633105936e-05, "loss": 2.8561, "step": 32174 }, { "epoch": 2.52, "learning_rate": 7.477064220183486e-05, "loss": 2.7675, "step": 32175 }, { "epoch": 2.52, "learning_rate": 7.476985807261036e-05, "loss": 2.8724, "step": 32176 }, { "epoch": 2.52, "learning_rate": 7.476907394338587e-05, "loss": 2.6813, "step": 32177 }, { "epoch": 2.52, "learning_rate": 7.476828981416137e-05, "loss": 2.8341, "step": 32178 }, { "epoch": 2.52, "learning_rate": 7.476750568493687e-05, "loss": 2.7264, "step": 32179 }, { "epoch": 2.52, "learning_rate": 7.476672155571238e-05, "loss": 2.6098, "step": 32180 }, { "epoch": 2.52, "learning_rate": 7.476593742648788e-05, "loss": 2.8715, "step": 32181 }, { "epoch": 2.52, "learning_rate": 7.476515329726339e-05, "loss": 2.8436, "step": 32182 }, { "epoch": 2.52, "learning_rate": 7.476436916803889e-05, "loss": 2.8936, "step": 32183 }, { "epoch": 2.52, "learning_rate": 7.476358503881439e-05, "loss": 2.7637, "step": 32184 }, { "epoch": 2.52, "learning_rate": 7.47628009095899e-05, "loss": 2.8312, "step": 32185 }, { "epoch": 2.52, "learning_rate": 7.47620167803654e-05, "loss": 2.8226, "step": 32186 }, { "epoch": 2.52, "learning_rate": 7.476123265114092e-05, "loss": 2.8463, "step": 32187 }, { "epoch": 2.52, "learning_rate": 7.476044852191642e-05, "loss": 2.8536, "step": 32188 }, { "epoch": 2.52, "learning_rate": 7.475966439269192e-05, "loss": 2.9309, "step": 32189 }, { "epoch": 2.52, "learning_rate": 7.475888026346743e-05, "loss": 2.7575, "step": 32190 }, { "epoch": 2.52, "learning_rate": 7.475809613424293e-05, "loss": 2.7187, "step": 32191 }, { "epoch": 2.52, "learning_rate": 7.475731200501844e-05, "loss": 2.6307, "step": 32192 }, { "epoch": 2.52, "learning_rate": 7.475652787579394e-05, "loss": 2.613, "step": 32193 }, { "epoch": 2.52, "learning_rate": 7.475574374656944e-05, "loss": 2.7443, "step": 32194 }, { "epoch": 2.52, "learning_rate": 7.475495961734495e-05, "loss": 2.9819, "step": 32195 }, { "epoch": 2.52, "learning_rate": 7.475417548812045e-05, "loss": 2.5706, "step": 32196 }, { "epoch": 2.52, "learning_rate": 7.475339135889595e-05, "loss": 2.9848, "step": 32197 }, { "epoch": 2.52, "learning_rate": 7.475260722967146e-05, "loss": 2.879, "step": 32198 }, { "epoch": 2.52, "learning_rate": 7.475182310044696e-05, "loss": 2.9036, "step": 32199 }, { "epoch": 2.52, "learning_rate": 7.475103897122247e-05, "loss": 2.7251, "step": 32200 }, { "epoch": 2.52, "learning_rate": 7.475025484199797e-05, "loss": 2.8101, "step": 32201 }, { "epoch": 2.53, "learning_rate": 7.474947071277347e-05, "loss": 2.7077, "step": 32202 }, { "epoch": 2.53, "learning_rate": 7.474868658354898e-05, "loss": 2.6155, "step": 32203 }, { "epoch": 2.53, "learning_rate": 7.474790245432448e-05, "loss": 3.0585, "step": 32204 }, { "epoch": 2.53, "learning_rate": 7.474711832509998e-05, "loss": 2.8395, "step": 32205 }, { "epoch": 2.53, "learning_rate": 7.474633419587549e-05, "loss": 2.9053, "step": 32206 }, { "epoch": 2.53, "learning_rate": 7.474555006665098e-05, "loss": 2.8113, "step": 32207 }, { "epoch": 2.53, "learning_rate": 7.474476593742648e-05, "loss": 2.8904, "step": 32208 }, { "epoch": 2.53, "learning_rate": 7.474398180820198e-05, "loss": 2.9517, "step": 32209 }, { "epoch": 2.53, "learning_rate": 7.474319767897749e-05, "loss": 2.9994, "step": 32210 }, { "epoch": 2.53, "learning_rate": 7.474241354975299e-05, "loss": 2.876, "step": 32211 }, { "epoch": 2.53, "learning_rate": 7.474162942052851e-05, "loss": 2.9869, "step": 32212 }, { "epoch": 2.53, "learning_rate": 7.474084529130401e-05, "loss": 2.7357, "step": 32213 }, { "epoch": 2.53, "learning_rate": 7.474006116207952e-05, "loss": 2.7656, "step": 32214 }, { "epoch": 2.53, "learning_rate": 7.473927703285502e-05, "loss": 2.8645, "step": 32215 }, { "epoch": 2.53, "learning_rate": 7.473849290363052e-05, "loss": 3.0367, "step": 32216 }, { "epoch": 2.53, "learning_rate": 7.473770877440603e-05, "loss": 2.7674, "step": 32217 }, { "epoch": 2.53, "learning_rate": 7.473692464518153e-05, "loss": 2.956, "step": 32218 }, { "epoch": 2.53, "learning_rate": 7.473614051595704e-05, "loss": 2.78, "step": 32219 }, { "epoch": 2.53, "learning_rate": 7.473535638673254e-05, "loss": 2.9173, "step": 32220 }, { "epoch": 2.53, "learning_rate": 7.473457225750804e-05, "loss": 2.6712, "step": 32221 }, { "epoch": 2.53, "learning_rate": 7.473378812828355e-05, "loss": 2.6619, "step": 32222 }, { "epoch": 2.53, "learning_rate": 7.473300399905905e-05, "loss": 3.0409, "step": 32223 }, { "epoch": 2.53, "learning_rate": 7.473221986983455e-05, "loss": 2.6956, "step": 32224 }, { "epoch": 2.53, "learning_rate": 7.473143574061006e-05, "loss": 2.7329, "step": 32225 }, { "epoch": 2.53, "learning_rate": 7.473065161138556e-05, "loss": 2.7328, "step": 32226 }, { "epoch": 2.53, "learning_rate": 7.472986748216106e-05, "loss": 2.6797, "step": 32227 }, { "epoch": 2.53, "learning_rate": 7.472908335293657e-05, "loss": 2.8259, "step": 32228 }, { "epoch": 2.53, "learning_rate": 7.472829922371207e-05, "loss": 2.6857, "step": 32229 }, { "epoch": 2.53, "learning_rate": 7.472751509448758e-05, "loss": 2.9645, "step": 32230 }, { "epoch": 2.53, "learning_rate": 7.472673096526308e-05, "loss": 2.713, "step": 32231 }, { "epoch": 2.53, "learning_rate": 7.472594683603858e-05, "loss": 2.7335, "step": 32232 }, { "epoch": 2.53, "learning_rate": 7.472516270681409e-05, "loss": 2.9581, "step": 32233 }, { "epoch": 2.53, "learning_rate": 7.472437857758959e-05, "loss": 2.8433, "step": 32234 }, { "epoch": 2.53, "learning_rate": 7.47235944483651e-05, "loss": 2.8467, "step": 32235 }, { "epoch": 2.53, "learning_rate": 7.47228103191406e-05, "loss": 2.8297, "step": 32236 }, { "epoch": 2.53, "learning_rate": 7.47220261899161e-05, "loss": 3.0295, "step": 32237 }, { "epoch": 2.53, "learning_rate": 7.47212420606916e-05, "loss": 2.7372, "step": 32238 }, { "epoch": 2.53, "learning_rate": 7.472045793146711e-05, "loss": 2.7579, "step": 32239 }, { "epoch": 2.53, "learning_rate": 7.471967380224261e-05, "loss": 2.7524, "step": 32240 }, { "epoch": 2.53, "learning_rate": 7.471888967301812e-05, "loss": 2.7518, "step": 32241 }, { "epoch": 2.53, "learning_rate": 7.471810554379362e-05, "loss": 2.7859, "step": 32242 }, { "epoch": 2.53, "learning_rate": 7.471732141456912e-05, "loss": 2.7882, "step": 32243 }, { "epoch": 2.53, "learning_rate": 7.471653728534463e-05, "loss": 3.0742, "step": 32244 }, { "epoch": 2.53, "learning_rate": 7.471575315612013e-05, "loss": 2.8964, "step": 32245 }, { "epoch": 2.53, "learning_rate": 7.471496902689563e-05, "loss": 2.8428, "step": 32246 }, { "epoch": 2.53, "learning_rate": 7.471418489767114e-05, "loss": 2.8044, "step": 32247 }, { "epoch": 2.53, "learning_rate": 7.471340076844664e-05, "loss": 2.6782, "step": 32248 }, { "epoch": 2.53, "learning_rate": 7.471261663922215e-05, "loss": 2.8256, "step": 32249 }, { "epoch": 2.53, "learning_rate": 7.471183250999765e-05, "loss": 2.8814, "step": 32250 }, { "epoch": 2.53, "learning_rate": 7.471104838077315e-05, "loss": 2.8717, "step": 32251 }, { "epoch": 2.53, "learning_rate": 7.471026425154866e-05, "loss": 2.6117, "step": 32252 }, { "epoch": 2.53, "learning_rate": 7.470948012232416e-05, "loss": 2.6843, "step": 32253 }, { "epoch": 2.53, "learning_rate": 7.470869599309966e-05, "loss": 2.9195, "step": 32254 }, { "epoch": 2.53, "learning_rate": 7.470791186387517e-05, "loss": 2.8016, "step": 32255 }, { "epoch": 2.53, "learning_rate": 7.470712773465067e-05, "loss": 3.0605, "step": 32256 }, { "epoch": 2.53, "learning_rate": 7.470634360542617e-05, "loss": 2.7738, "step": 32257 }, { "epoch": 2.53, "learning_rate": 7.470555947620168e-05, "loss": 2.8646, "step": 32258 }, { "epoch": 2.53, "learning_rate": 7.470477534697718e-05, "loss": 2.8137, "step": 32259 }, { "epoch": 2.53, "learning_rate": 7.470399121775269e-05, "loss": 2.8431, "step": 32260 }, { "epoch": 2.53, "learning_rate": 7.470320708852819e-05, "loss": 2.6671, "step": 32261 }, { "epoch": 2.53, "learning_rate": 7.47024229593037e-05, "loss": 2.9064, "step": 32262 }, { "epoch": 2.53, "learning_rate": 7.47016388300792e-05, "loss": 2.7417, "step": 32263 }, { "epoch": 2.53, "learning_rate": 7.47008547008547e-05, "loss": 2.7539, "step": 32264 }, { "epoch": 2.53, "learning_rate": 7.47000705716302e-05, "loss": 2.8534, "step": 32265 }, { "epoch": 2.53, "learning_rate": 7.469928644240571e-05, "loss": 2.9479, "step": 32266 }, { "epoch": 2.53, "learning_rate": 7.469850231318121e-05, "loss": 2.7604, "step": 32267 }, { "epoch": 2.53, "learning_rate": 7.469771818395672e-05, "loss": 2.7774, "step": 32268 }, { "epoch": 2.53, "learning_rate": 7.469693405473223e-05, "loss": 2.6961, "step": 32269 }, { "epoch": 2.53, "learning_rate": 7.469614992550774e-05, "loss": 2.8872, "step": 32270 }, { "epoch": 2.53, "learning_rate": 7.469536579628324e-05, "loss": 2.9397, "step": 32271 }, { "epoch": 2.53, "learning_rate": 7.469458166705874e-05, "loss": 2.8807, "step": 32272 }, { "epoch": 2.53, "learning_rate": 7.469379753783425e-05, "loss": 2.7812, "step": 32273 }, { "epoch": 2.53, "learning_rate": 7.469301340860975e-05, "loss": 2.7473, "step": 32274 }, { "epoch": 2.53, "learning_rate": 7.469222927938525e-05, "loss": 2.8739, "step": 32275 }, { "epoch": 2.53, "learning_rate": 7.469144515016076e-05, "loss": 2.8192, "step": 32276 }, { "epoch": 2.53, "learning_rate": 7.469066102093626e-05, "loss": 2.6089, "step": 32277 }, { "epoch": 2.53, "learning_rate": 7.468987689171177e-05, "loss": 2.6487, "step": 32278 }, { "epoch": 2.53, "learning_rate": 7.468909276248726e-05, "loss": 2.6674, "step": 32279 }, { "epoch": 2.53, "learning_rate": 7.468830863326276e-05, "loss": 2.8052, "step": 32280 }, { "epoch": 2.53, "learning_rate": 7.468752450403826e-05, "loss": 2.8766, "step": 32281 }, { "epoch": 2.53, "learning_rate": 7.468674037481377e-05, "loss": 2.9265, "step": 32282 }, { "epoch": 2.53, "learning_rate": 7.468595624558927e-05, "loss": 2.7404, "step": 32283 }, { "epoch": 2.53, "learning_rate": 7.468517211636477e-05, "loss": 2.844, "step": 32284 }, { "epoch": 2.53, "learning_rate": 7.468438798714028e-05, "loss": 2.946, "step": 32285 }, { "epoch": 2.53, "learning_rate": 7.468360385791578e-05, "loss": 2.6643, "step": 32286 }, { "epoch": 2.53, "learning_rate": 7.468281972869129e-05, "loss": 2.8043, "step": 32287 }, { "epoch": 2.53, "learning_rate": 7.468203559946679e-05, "loss": 2.6959, "step": 32288 }, { "epoch": 2.53, "learning_rate": 7.468125147024229e-05, "loss": 2.7927, "step": 32289 }, { "epoch": 2.53, "learning_rate": 7.46804673410178e-05, "loss": 2.7639, "step": 32290 }, { "epoch": 2.53, "learning_rate": 7.46796832117933e-05, "loss": 2.6701, "step": 32291 }, { "epoch": 2.53, "learning_rate": 7.46788990825688e-05, "loss": 2.6843, "step": 32292 }, { "epoch": 2.53, "learning_rate": 7.467811495334431e-05, "loss": 2.8885, "step": 32293 }, { "epoch": 2.53, "learning_rate": 7.467733082411981e-05, "loss": 2.7616, "step": 32294 }, { "epoch": 2.53, "learning_rate": 7.467654669489533e-05, "loss": 3.1258, "step": 32295 }, { "epoch": 2.53, "learning_rate": 7.467576256567083e-05, "loss": 2.7035, "step": 32296 }, { "epoch": 2.53, "learning_rate": 7.467497843644634e-05, "loss": 2.6122, "step": 32297 }, { "epoch": 2.53, "learning_rate": 7.467419430722184e-05, "loss": 2.6441, "step": 32298 }, { "epoch": 2.53, "learning_rate": 7.467341017799734e-05, "loss": 2.8781, "step": 32299 }, { "epoch": 2.53, "learning_rate": 7.467262604877285e-05, "loss": 3.1749, "step": 32300 }, { "epoch": 2.53, "learning_rate": 7.467184191954835e-05, "loss": 2.7601, "step": 32301 }, { "epoch": 2.53, "learning_rate": 7.467105779032385e-05, "loss": 2.7362, "step": 32302 }, { "epoch": 2.53, "learning_rate": 7.467027366109936e-05, "loss": 2.6699, "step": 32303 }, { "epoch": 2.53, "learning_rate": 7.466948953187486e-05, "loss": 2.9174, "step": 32304 }, { "epoch": 2.53, "learning_rate": 7.466870540265037e-05, "loss": 2.7638, "step": 32305 }, { "epoch": 2.53, "learning_rate": 7.466792127342587e-05, "loss": 2.938, "step": 32306 }, { "epoch": 2.53, "learning_rate": 7.466713714420137e-05, "loss": 2.6572, "step": 32307 }, { "epoch": 2.53, "learning_rate": 7.466635301497688e-05, "loss": 2.5714, "step": 32308 }, { "epoch": 2.53, "learning_rate": 7.466556888575238e-05, "loss": 2.6186, "step": 32309 }, { "epoch": 2.53, "learning_rate": 7.466478475652788e-05, "loss": 2.7305, "step": 32310 }, { "epoch": 2.53, "learning_rate": 7.466400062730339e-05, "loss": 2.7004, "step": 32311 }, { "epoch": 2.53, "learning_rate": 7.466321649807889e-05, "loss": 2.9997, "step": 32312 }, { "epoch": 2.53, "learning_rate": 7.46624323688544e-05, "loss": 2.9529, "step": 32313 }, { "epoch": 2.53, "learning_rate": 7.46616482396299e-05, "loss": 2.6723, "step": 32314 }, { "epoch": 2.53, "learning_rate": 7.466086411040539e-05, "loss": 2.732, "step": 32315 }, { "epoch": 2.53, "learning_rate": 7.466007998118089e-05, "loss": 2.9437, "step": 32316 }, { "epoch": 2.53, "learning_rate": 7.46592958519564e-05, "loss": 3.0135, "step": 32317 }, { "epoch": 2.53, "learning_rate": 7.46585117227319e-05, "loss": 2.7272, "step": 32318 }, { "epoch": 2.53, "learning_rate": 7.46577275935074e-05, "loss": 2.978, "step": 32319 }, { "epoch": 2.53, "learning_rate": 7.465694346428292e-05, "loss": 2.7577, "step": 32320 }, { "epoch": 2.53, "learning_rate": 7.465615933505842e-05, "loss": 2.7538, "step": 32321 }, { "epoch": 2.53, "learning_rate": 7.465537520583393e-05, "loss": 3.0084, "step": 32322 }, { "epoch": 2.53, "learning_rate": 7.465459107660943e-05, "loss": 2.7988, "step": 32323 }, { "epoch": 2.53, "learning_rate": 7.465380694738493e-05, "loss": 3.1354, "step": 32324 }, { "epoch": 2.53, "learning_rate": 7.465302281816044e-05, "loss": 2.8627, "step": 32325 }, { "epoch": 2.53, "learning_rate": 7.465223868893594e-05, "loss": 2.6666, "step": 32326 }, { "epoch": 2.53, "learning_rate": 7.465145455971145e-05, "loss": 2.6614, "step": 32327 }, { "epoch": 2.53, "learning_rate": 7.465067043048695e-05, "loss": 2.7705, "step": 32328 }, { "epoch": 2.54, "learning_rate": 7.464988630126245e-05, "loss": 2.8499, "step": 32329 }, { "epoch": 2.54, "learning_rate": 7.464910217203796e-05, "loss": 2.7981, "step": 32330 }, { "epoch": 2.54, "learning_rate": 7.464831804281346e-05, "loss": 2.7927, "step": 32331 }, { "epoch": 2.54, "learning_rate": 7.464753391358896e-05, "loss": 3.0025, "step": 32332 }, { "epoch": 2.54, "learning_rate": 7.464674978436447e-05, "loss": 3.0862, "step": 32333 }, { "epoch": 2.54, "learning_rate": 7.464596565513997e-05, "loss": 2.7082, "step": 32334 }, { "epoch": 2.54, "learning_rate": 7.464518152591548e-05, "loss": 2.897, "step": 32335 }, { "epoch": 2.54, "learning_rate": 7.464439739669098e-05, "loss": 2.7882, "step": 32336 }, { "epoch": 2.54, "learning_rate": 7.464361326746648e-05, "loss": 2.6025, "step": 32337 }, { "epoch": 2.54, "learning_rate": 7.464282913824199e-05, "loss": 2.9813, "step": 32338 }, { "epoch": 2.54, "learning_rate": 7.464204500901749e-05, "loss": 2.7424, "step": 32339 }, { "epoch": 2.54, "learning_rate": 7.4641260879793e-05, "loss": 2.9336, "step": 32340 }, { "epoch": 2.54, "learning_rate": 7.46404767505685e-05, "loss": 2.7414, "step": 32341 }, { "epoch": 2.54, "learning_rate": 7.4639692621344e-05, "loss": 2.6698, "step": 32342 }, { "epoch": 2.54, "learning_rate": 7.46389084921195e-05, "loss": 2.778, "step": 32343 }, { "epoch": 2.54, "learning_rate": 7.463812436289501e-05, "loss": 2.894, "step": 32344 }, { "epoch": 2.54, "learning_rate": 7.463734023367051e-05, "loss": 2.6322, "step": 32345 }, { "epoch": 2.54, "learning_rate": 7.463655610444602e-05, "loss": 2.686, "step": 32346 }, { "epoch": 2.54, "learning_rate": 7.463577197522152e-05, "loss": 2.8743, "step": 32347 }, { "epoch": 2.54, "learning_rate": 7.463498784599702e-05, "loss": 2.9302, "step": 32348 }, { "epoch": 2.54, "learning_rate": 7.463420371677253e-05, "loss": 2.877, "step": 32349 }, { "epoch": 2.54, "learning_rate": 7.463341958754803e-05, "loss": 2.8578, "step": 32350 }, { "epoch": 2.54, "learning_rate": 7.463263545832353e-05, "loss": 2.9299, "step": 32351 }, { "epoch": 2.54, "learning_rate": 7.463185132909904e-05, "loss": 2.7863, "step": 32352 }, { "epoch": 2.54, "learning_rate": 7.463106719987454e-05, "loss": 2.8423, "step": 32353 }, { "epoch": 2.54, "learning_rate": 7.463028307065005e-05, "loss": 2.8012, "step": 32354 }, { "epoch": 2.54, "learning_rate": 7.462949894142555e-05, "loss": 2.6901, "step": 32355 }, { "epoch": 2.54, "learning_rate": 7.462871481220105e-05, "loss": 2.9659, "step": 32356 }, { "epoch": 2.54, "learning_rate": 7.462793068297656e-05, "loss": 2.9823, "step": 32357 }, { "epoch": 2.54, "learning_rate": 7.462714655375206e-05, "loss": 2.8195, "step": 32358 }, { "epoch": 2.54, "learning_rate": 7.462636242452756e-05, "loss": 2.7329, "step": 32359 }, { "epoch": 2.54, "learning_rate": 7.462557829530307e-05, "loss": 2.9466, "step": 32360 }, { "epoch": 2.54, "learning_rate": 7.462479416607857e-05, "loss": 2.8857, "step": 32361 }, { "epoch": 2.54, "learning_rate": 7.462401003685407e-05, "loss": 2.7564, "step": 32362 }, { "epoch": 2.54, "learning_rate": 7.462322590762958e-05, "loss": 2.8795, "step": 32363 }, { "epoch": 2.54, "learning_rate": 7.462244177840508e-05, "loss": 2.9022, "step": 32364 }, { "epoch": 2.54, "learning_rate": 7.462165764918059e-05, "loss": 2.7511, "step": 32365 }, { "epoch": 2.54, "learning_rate": 7.462087351995609e-05, "loss": 2.8053, "step": 32366 }, { "epoch": 2.54, "learning_rate": 7.462008939073159e-05, "loss": 2.862, "step": 32367 }, { "epoch": 2.54, "learning_rate": 7.46193052615071e-05, "loss": 2.8468, "step": 32368 }, { "epoch": 2.54, "learning_rate": 7.46185211322826e-05, "loss": 2.8341, "step": 32369 }, { "epoch": 2.54, "learning_rate": 7.46177370030581e-05, "loss": 2.8176, "step": 32370 }, { "epoch": 2.54, "learning_rate": 7.461695287383361e-05, "loss": 2.7726, "step": 32371 }, { "epoch": 2.54, "learning_rate": 7.461616874460911e-05, "loss": 2.6607, "step": 32372 }, { "epoch": 2.54, "learning_rate": 7.461538461538462e-05, "loss": 2.8583, "step": 32373 }, { "epoch": 2.54, "learning_rate": 7.461460048616012e-05, "loss": 3.0529, "step": 32374 }, { "epoch": 2.54, "learning_rate": 7.461381635693562e-05, "loss": 2.9543, "step": 32375 }, { "epoch": 2.54, "learning_rate": 7.461303222771113e-05, "loss": 2.8114, "step": 32376 }, { "epoch": 2.54, "learning_rate": 7.461224809848664e-05, "loss": 2.6809, "step": 32377 }, { "epoch": 2.54, "learning_rate": 7.461146396926215e-05, "loss": 2.5761, "step": 32378 }, { "epoch": 2.54, "learning_rate": 7.461067984003765e-05, "loss": 2.7595, "step": 32379 }, { "epoch": 2.54, "learning_rate": 7.460989571081315e-05, "loss": 3.2629, "step": 32380 }, { "epoch": 2.54, "learning_rate": 7.460911158158866e-05, "loss": 2.8095, "step": 32381 }, { "epoch": 2.54, "learning_rate": 7.460832745236416e-05, "loss": 2.802, "step": 32382 }, { "epoch": 2.54, "learning_rate": 7.460754332313967e-05, "loss": 2.7682, "step": 32383 }, { "epoch": 2.54, "learning_rate": 7.460675919391517e-05, "loss": 2.757, "step": 32384 }, { "epoch": 2.54, "learning_rate": 7.460597506469067e-05, "loss": 3.0848, "step": 32385 }, { "epoch": 2.54, "learning_rate": 7.460519093546618e-05, "loss": 2.8106, "step": 32386 }, { "epoch": 2.54, "learning_rate": 7.460440680624167e-05, "loss": 2.7949, "step": 32387 }, { "epoch": 2.54, "learning_rate": 7.460362267701717e-05, "loss": 2.8788, "step": 32388 }, { "epoch": 2.54, "learning_rate": 7.460283854779267e-05, "loss": 2.7896, "step": 32389 }, { "epoch": 2.54, "learning_rate": 7.460205441856818e-05, "loss": 2.9114, "step": 32390 }, { "epoch": 2.54, "learning_rate": 7.460127028934368e-05, "loss": 3.0913, "step": 32391 }, { "epoch": 2.54, "learning_rate": 7.460048616011918e-05, "loss": 2.7987, "step": 32392 }, { "epoch": 2.54, "learning_rate": 7.459970203089469e-05, "loss": 2.8465, "step": 32393 }, { "epoch": 2.54, "learning_rate": 7.459891790167019e-05, "loss": 2.7825, "step": 32394 }, { "epoch": 2.54, "learning_rate": 7.45981337724457e-05, "loss": 2.7266, "step": 32395 }, { "epoch": 2.54, "learning_rate": 7.45973496432212e-05, "loss": 2.9126, "step": 32396 }, { "epoch": 2.54, "learning_rate": 7.45965655139967e-05, "loss": 2.9486, "step": 32397 }, { "epoch": 2.54, "learning_rate": 7.459578138477221e-05, "loss": 2.9227, "step": 32398 }, { "epoch": 2.54, "learning_rate": 7.459499725554771e-05, "loss": 2.7442, "step": 32399 }, { "epoch": 2.54, "learning_rate": 7.459421312632321e-05, "loss": 2.8968, "step": 32400 }, { "epoch": 2.54, "learning_rate": 7.459342899709872e-05, "loss": 2.5969, "step": 32401 }, { "epoch": 2.54, "learning_rate": 7.459264486787422e-05, "loss": 2.8844, "step": 32402 }, { "epoch": 2.54, "learning_rate": 7.459186073864974e-05, "loss": 2.7105, "step": 32403 }, { "epoch": 2.54, "learning_rate": 7.459107660942524e-05, "loss": 2.7799, "step": 32404 }, { "epoch": 2.54, "learning_rate": 7.459029248020075e-05, "loss": 2.6896, "step": 32405 }, { "epoch": 2.54, "learning_rate": 7.458950835097625e-05, "loss": 3.1709, "step": 32406 }, { "epoch": 2.54, "learning_rate": 7.458872422175175e-05, "loss": 2.8047, "step": 32407 }, { "epoch": 2.54, "learning_rate": 7.458794009252726e-05, "loss": 3.067, "step": 32408 }, { "epoch": 2.54, "learning_rate": 7.458715596330276e-05, "loss": 2.85, "step": 32409 }, { "epoch": 2.54, "learning_rate": 7.458637183407826e-05, "loss": 2.9503, "step": 32410 }, { "epoch": 2.54, "learning_rate": 7.458558770485377e-05, "loss": 2.7705, "step": 32411 }, { "epoch": 2.54, "learning_rate": 7.458480357562927e-05, "loss": 2.6614, "step": 32412 }, { "epoch": 2.54, "learning_rate": 7.458401944640478e-05, "loss": 2.8914, "step": 32413 }, { "epoch": 2.54, "learning_rate": 7.458323531718028e-05, "loss": 2.7041, "step": 32414 }, { "epoch": 2.54, "learning_rate": 7.458245118795578e-05, "loss": 2.8327, "step": 32415 }, { "epoch": 2.54, "learning_rate": 7.458166705873129e-05, "loss": 2.6174, "step": 32416 }, { "epoch": 2.54, "learning_rate": 7.458088292950679e-05, "loss": 2.7582, "step": 32417 }, { "epoch": 2.54, "learning_rate": 7.45800988002823e-05, "loss": 2.6622, "step": 32418 }, { "epoch": 2.54, "learning_rate": 7.45793146710578e-05, "loss": 2.6575, "step": 32419 }, { "epoch": 2.54, "learning_rate": 7.45785305418333e-05, "loss": 2.8868, "step": 32420 }, { "epoch": 2.54, "learning_rate": 7.45777464126088e-05, "loss": 2.5673, "step": 32421 }, { "epoch": 2.54, "learning_rate": 7.457696228338431e-05, "loss": 2.8585, "step": 32422 }, { "epoch": 2.54, "learning_rate": 7.45761781541598e-05, "loss": 2.6717, "step": 32423 }, { "epoch": 2.54, "learning_rate": 7.45753940249353e-05, "loss": 2.9102, "step": 32424 }, { "epoch": 2.54, "learning_rate": 7.45746098957108e-05, "loss": 3.0248, "step": 32425 }, { "epoch": 2.54, "learning_rate": 7.457382576648631e-05, "loss": 3.0302, "step": 32426 }, { "epoch": 2.54, "learning_rate": 7.457304163726181e-05, "loss": 2.7977, "step": 32427 }, { "epoch": 2.54, "learning_rate": 7.457225750803733e-05, "loss": 2.9379, "step": 32428 }, { "epoch": 2.54, "learning_rate": 7.457147337881283e-05, "loss": 2.6238, "step": 32429 }, { "epoch": 2.54, "learning_rate": 7.457068924958834e-05, "loss": 2.9174, "step": 32430 }, { "epoch": 2.54, "learning_rate": 7.456990512036384e-05, "loss": 2.8852, "step": 32431 }, { "epoch": 2.54, "learning_rate": 7.456912099113935e-05, "loss": 2.9939, "step": 32432 }, { "epoch": 2.54, "learning_rate": 7.456833686191485e-05, "loss": 2.8261, "step": 32433 }, { "epoch": 2.54, "learning_rate": 7.456755273269035e-05, "loss": 2.6383, "step": 32434 }, { "epoch": 2.54, "learning_rate": 7.456676860346586e-05, "loss": 2.9651, "step": 32435 }, { "epoch": 2.54, "learning_rate": 7.456598447424136e-05, "loss": 2.8385, "step": 32436 }, { "epoch": 2.54, "learning_rate": 7.456520034501686e-05, "loss": 2.8893, "step": 32437 }, { "epoch": 2.54, "learning_rate": 7.456441621579237e-05, "loss": 2.5824, "step": 32438 }, { "epoch": 2.54, "learning_rate": 7.456363208656787e-05, "loss": 2.6249, "step": 32439 }, { "epoch": 2.54, "learning_rate": 7.456284795734338e-05, "loss": 3.0478, "step": 32440 }, { "epoch": 2.54, "learning_rate": 7.456206382811888e-05, "loss": 2.8153, "step": 32441 }, { "epoch": 2.54, "learning_rate": 7.456127969889438e-05, "loss": 2.7154, "step": 32442 }, { "epoch": 2.54, "learning_rate": 7.456049556966989e-05, "loss": 2.5609, "step": 32443 }, { "epoch": 2.54, "learning_rate": 7.455971144044539e-05, "loss": 2.6561, "step": 32444 }, { "epoch": 2.54, "learning_rate": 7.45589273112209e-05, "loss": 2.6979, "step": 32445 }, { "epoch": 2.54, "learning_rate": 7.45581431819964e-05, "loss": 2.7194, "step": 32446 }, { "epoch": 2.54, "learning_rate": 7.45573590527719e-05, "loss": 2.6891, "step": 32447 }, { "epoch": 2.54, "learning_rate": 7.45565749235474e-05, "loss": 2.727, "step": 32448 }, { "epoch": 2.54, "learning_rate": 7.455579079432291e-05, "loss": 2.7345, "step": 32449 }, { "epoch": 2.54, "learning_rate": 7.455500666509841e-05, "loss": 2.7034, "step": 32450 }, { "epoch": 2.54, "learning_rate": 7.455422253587392e-05, "loss": 2.9993, "step": 32451 }, { "epoch": 2.54, "learning_rate": 7.455343840664942e-05, "loss": 2.9481, "step": 32452 }, { "epoch": 2.54, "learning_rate": 7.455265427742492e-05, "loss": 2.8574, "step": 32453 }, { "epoch": 2.54, "learning_rate": 7.455187014820043e-05, "loss": 2.8685, "step": 32454 }, { "epoch": 2.54, "learning_rate": 7.455108601897593e-05, "loss": 2.7208, "step": 32455 }, { "epoch": 2.54, "learning_rate": 7.455030188975143e-05, "loss": 2.8173, "step": 32456 }, { "epoch": 2.55, "learning_rate": 7.454951776052694e-05, "loss": 2.7917, "step": 32457 }, { "epoch": 2.55, "learning_rate": 7.454873363130244e-05, "loss": 2.9731, "step": 32458 }, { "epoch": 2.55, "learning_rate": 7.454794950207794e-05, "loss": 2.9259, "step": 32459 }, { "epoch": 2.55, "learning_rate": 7.454716537285345e-05, "loss": 2.7356, "step": 32460 }, { "epoch": 2.55, "learning_rate": 7.454638124362895e-05, "loss": 2.7309, "step": 32461 }, { "epoch": 2.55, "learning_rate": 7.454559711440446e-05, "loss": 3.1853, "step": 32462 }, { "epoch": 2.55, "learning_rate": 7.454481298517996e-05, "loss": 2.9802, "step": 32463 }, { "epoch": 2.55, "learning_rate": 7.454402885595546e-05, "loss": 2.7727, "step": 32464 }, { "epoch": 2.55, "learning_rate": 7.454324472673097e-05, "loss": 2.7159, "step": 32465 }, { "epoch": 2.55, "learning_rate": 7.454246059750647e-05, "loss": 2.9229, "step": 32466 }, { "epoch": 2.55, "learning_rate": 7.454167646828197e-05, "loss": 2.8488, "step": 32467 }, { "epoch": 2.55, "learning_rate": 7.454089233905748e-05, "loss": 2.7632, "step": 32468 }, { "epoch": 2.55, "learning_rate": 7.454010820983298e-05, "loss": 2.9392, "step": 32469 }, { "epoch": 2.55, "learning_rate": 7.453932408060849e-05, "loss": 2.786, "step": 32470 }, { "epoch": 2.55, "learning_rate": 7.453853995138399e-05, "loss": 2.8778, "step": 32471 }, { "epoch": 2.55, "learning_rate": 7.453775582215949e-05, "loss": 2.7622, "step": 32472 }, { "epoch": 2.55, "learning_rate": 7.4536971692935e-05, "loss": 2.8071, "step": 32473 }, { "epoch": 2.55, "learning_rate": 7.45361875637105e-05, "loss": 2.7066, "step": 32474 }, { "epoch": 2.55, "learning_rate": 7.4535403434486e-05, "loss": 3.0563, "step": 32475 }, { "epoch": 2.55, "learning_rate": 7.453461930526151e-05, "loss": 2.6911, "step": 32476 }, { "epoch": 2.55, "learning_rate": 7.453383517603701e-05, "loss": 2.8421, "step": 32477 }, { "epoch": 2.55, "learning_rate": 7.453305104681251e-05, "loss": 2.7234, "step": 32478 }, { "epoch": 2.55, "learning_rate": 7.453226691758802e-05, "loss": 2.9485, "step": 32479 }, { "epoch": 2.55, "learning_rate": 7.453148278836352e-05, "loss": 2.8285, "step": 32480 }, { "epoch": 2.55, "learning_rate": 7.453069865913903e-05, "loss": 2.8695, "step": 32481 }, { "epoch": 2.55, "learning_rate": 7.452991452991453e-05, "loss": 2.6298, "step": 32482 }, { "epoch": 2.55, "learning_rate": 7.452913040069003e-05, "loss": 2.8783, "step": 32483 }, { "epoch": 2.55, "learning_rate": 7.452834627146554e-05, "loss": 3.1361, "step": 32484 }, { "epoch": 2.55, "learning_rate": 7.452756214224105e-05, "loss": 2.7514, "step": 32485 }, { "epoch": 2.55, "learning_rate": 7.452677801301656e-05, "loss": 2.9883, "step": 32486 }, { "epoch": 2.55, "learning_rate": 7.452599388379206e-05, "loss": 2.8975, "step": 32487 }, { "epoch": 2.55, "learning_rate": 7.452520975456757e-05, "loss": 2.9764, "step": 32488 }, { "epoch": 2.55, "learning_rate": 7.452442562534307e-05, "loss": 2.7388, "step": 32489 }, { "epoch": 2.55, "learning_rate": 7.452364149611857e-05, "loss": 2.7604, "step": 32490 }, { "epoch": 2.55, "learning_rate": 7.452285736689408e-05, "loss": 2.9955, "step": 32491 }, { "epoch": 2.55, "learning_rate": 7.452207323766958e-05, "loss": 2.9328, "step": 32492 }, { "epoch": 2.55, "learning_rate": 7.452128910844508e-05, "loss": 2.8405, "step": 32493 }, { "epoch": 2.55, "learning_rate": 7.452050497922059e-05, "loss": 2.5115, "step": 32494 }, { "epoch": 2.55, "learning_rate": 7.451972084999608e-05, "loss": 2.6888, "step": 32495 }, { "epoch": 2.55, "learning_rate": 7.451893672077158e-05, "loss": 2.763, "step": 32496 }, { "epoch": 2.55, "learning_rate": 7.451815259154708e-05, "loss": 2.7153, "step": 32497 }, { "epoch": 2.55, "learning_rate": 7.451736846232259e-05, "loss": 2.7513, "step": 32498 }, { "epoch": 2.55, "learning_rate": 7.451658433309809e-05, "loss": 2.6906, "step": 32499 }, { "epoch": 2.55, "learning_rate": 7.45158002038736e-05, "loss": 2.7845, "step": 32500 }, { "epoch": 2.55, "learning_rate": 7.45150160746491e-05, "loss": 2.6918, "step": 32501 }, { "epoch": 2.55, "learning_rate": 7.45142319454246e-05, "loss": 2.865, "step": 32502 }, { "epoch": 2.55, "learning_rate": 7.45134478162001e-05, "loss": 2.7743, "step": 32503 }, { "epoch": 2.55, "learning_rate": 7.451266368697561e-05, "loss": 2.8893, "step": 32504 }, { "epoch": 2.55, "learning_rate": 7.451187955775111e-05, "loss": 2.7293, "step": 32505 }, { "epoch": 2.55, "learning_rate": 7.451109542852662e-05, "loss": 2.772, "step": 32506 }, { "epoch": 2.55, "learning_rate": 7.451031129930212e-05, "loss": 2.9998, "step": 32507 }, { "epoch": 2.55, "learning_rate": 7.450952717007763e-05, "loss": 2.7638, "step": 32508 }, { "epoch": 2.55, "learning_rate": 7.450874304085313e-05, "loss": 2.719, "step": 32509 }, { "epoch": 2.55, "learning_rate": 7.450795891162863e-05, "loss": 2.6729, "step": 32510 }, { "epoch": 2.55, "learning_rate": 7.450717478240415e-05, "loss": 2.733, "step": 32511 }, { "epoch": 2.55, "learning_rate": 7.450639065317965e-05, "loss": 2.6609, "step": 32512 }, { "epoch": 2.55, "learning_rate": 7.450560652395516e-05, "loss": 2.903, "step": 32513 }, { "epoch": 2.55, "learning_rate": 7.450482239473066e-05, "loss": 2.7067, "step": 32514 }, { "epoch": 2.55, "learning_rate": 7.450403826550616e-05, "loss": 2.655, "step": 32515 }, { "epoch": 2.55, "learning_rate": 7.450325413628167e-05, "loss": 2.8341, "step": 32516 }, { "epoch": 2.55, "learning_rate": 7.450247000705717e-05, "loss": 2.6147, "step": 32517 }, { "epoch": 2.55, "learning_rate": 7.450168587783268e-05, "loss": 2.8444, "step": 32518 }, { "epoch": 2.55, "learning_rate": 7.450090174860818e-05, "loss": 2.8479, "step": 32519 }, { "epoch": 2.55, "learning_rate": 7.450011761938368e-05, "loss": 2.888, "step": 32520 }, { "epoch": 2.55, "learning_rate": 7.449933349015919e-05, "loss": 2.7231, "step": 32521 }, { "epoch": 2.55, "learning_rate": 7.449854936093469e-05, "loss": 2.9273, "step": 32522 }, { "epoch": 2.55, "learning_rate": 7.44977652317102e-05, "loss": 3.0842, "step": 32523 }, { "epoch": 2.55, "learning_rate": 7.44969811024857e-05, "loss": 2.613, "step": 32524 }, { "epoch": 2.55, "learning_rate": 7.44961969732612e-05, "loss": 2.7906, "step": 32525 }, { "epoch": 2.55, "learning_rate": 7.44954128440367e-05, "loss": 2.6902, "step": 32526 }, { "epoch": 2.55, "learning_rate": 7.449462871481221e-05, "loss": 2.735, "step": 32527 }, { "epoch": 2.55, "learning_rate": 7.449384458558771e-05, "loss": 2.8451, "step": 32528 }, { "epoch": 2.55, "learning_rate": 7.449306045636322e-05, "loss": 2.8787, "step": 32529 }, { "epoch": 2.55, "learning_rate": 7.449227632713872e-05, "loss": 2.8083, "step": 32530 }, { "epoch": 2.55, "learning_rate": 7.449149219791421e-05, "loss": 2.8815, "step": 32531 }, { "epoch": 2.55, "learning_rate": 7.449070806868971e-05, "loss": 2.8109, "step": 32532 }, { "epoch": 2.55, "learning_rate": 7.448992393946522e-05, "loss": 2.7548, "step": 32533 }, { "epoch": 2.55, "learning_rate": 7.448913981024072e-05, "loss": 2.7774, "step": 32534 }, { "epoch": 2.55, "learning_rate": 7.448835568101622e-05, "loss": 2.8271, "step": 32535 }, { "epoch": 2.55, "learning_rate": 7.448757155179174e-05, "loss": 2.8429, "step": 32536 }, { "epoch": 2.55, "learning_rate": 7.448678742256725e-05, "loss": 2.9222, "step": 32537 }, { "epoch": 2.55, "learning_rate": 7.448600329334275e-05, "loss": 2.7258, "step": 32538 }, { "epoch": 2.55, "learning_rate": 7.448521916411825e-05, "loss": 2.8684, "step": 32539 }, { "epoch": 2.55, "learning_rate": 7.448443503489376e-05, "loss": 2.8379, "step": 32540 }, { "epoch": 2.55, "learning_rate": 7.448365090566926e-05, "loss": 2.7567, "step": 32541 }, { "epoch": 2.55, "learning_rate": 7.448286677644476e-05, "loss": 2.6829, "step": 32542 }, { "epoch": 2.55, "learning_rate": 7.448208264722027e-05, "loss": 2.655, "step": 32543 }, { "epoch": 2.55, "learning_rate": 7.448129851799577e-05, "loss": 2.6789, "step": 32544 }, { "epoch": 2.55, "learning_rate": 7.448051438877127e-05, "loss": 2.9004, "step": 32545 }, { "epoch": 2.55, "learning_rate": 7.447973025954678e-05, "loss": 2.771, "step": 32546 }, { "epoch": 2.55, "learning_rate": 7.447894613032228e-05, "loss": 2.7041, "step": 32547 }, { "epoch": 2.55, "learning_rate": 7.447816200109779e-05, "loss": 2.7717, "step": 32548 }, { "epoch": 2.55, "learning_rate": 7.447737787187329e-05, "loss": 2.7417, "step": 32549 }, { "epoch": 2.55, "learning_rate": 7.447659374264879e-05, "loss": 2.9026, "step": 32550 }, { "epoch": 2.55, "learning_rate": 7.44758096134243e-05, "loss": 2.7483, "step": 32551 }, { "epoch": 2.55, "learning_rate": 7.44750254841998e-05, "loss": 2.8012, "step": 32552 }, { "epoch": 2.55, "learning_rate": 7.44742413549753e-05, "loss": 2.9457, "step": 32553 }, { "epoch": 2.55, "learning_rate": 7.447345722575081e-05, "loss": 3.0978, "step": 32554 }, { "epoch": 2.55, "learning_rate": 7.447267309652631e-05, "loss": 2.711, "step": 32555 }, { "epoch": 2.55, "learning_rate": 7.447188896730182e-05, "loss": 2.7994, "step": 32556 }, { "epoch": 2.55, "learning_rate": 7.447110483807732e-05, "loss": 2.8918, "step": 32557 }, { "epoch": 2.55, "learning_rate": 7.447032070885282e-05, "loss": 2.7287, "step": 32558 }, { "epoch": 2.55, "learning_rate": 7.446953657962833e-05, "loss": 2.6886, "step": 32559 }, { "epoch": 2.55, "learning_rate": 7.446875245040383e-05, "loss": 2.7596, "step": 32560 }, { "epoch": 2.55, "learning_rate": 7.446796832117933e-05, "loss": 2.7602, "step": 32561 }, { "epoch": 2.55, "learning_rate": 7.446718419195484e-05, "loss": 2.6997, "step": 32562 }, { "epoch": 2.55, "learning_rate": 7.446640006273034e-05, "loss": 2.9637, "step": 32563 }, { "epoch": 2.55, "learning_rate": 7.446561593350584e-05, "loss": 2.8632, "step": 32564 }, { "epoch": 2.55, "learning_rate": 7.446483180428135e-05, "loss": 2.7812, "step": 32565 }, { "epoch": 2.55, "learning_rate": 7.446404767505685e-05, "loss": 2.6927, "step": 32566 }, { "epoch": 2.55, "learning_rate": 7.446326354583236e-05, "loss": 2.709, "step": 32567 }, { "epoch": 2.55, "learning_rate": 7.446247941660786e-05, "loss": 2.7065, "step": 32568 }, { "epoch": 2.55, "learning_rate": 7.446169528738336e-05, "loss": 2.7462, "step": 32569 }, { "epoch": 2.55, "learning_rate": 7.446091115815887e-05, "loss": 2.7634, "step": 32570 }, { "epoch": 2.55, "learning_rate": 7.446012702893437e-05, "loss": 2.8231, "step": 32571 }, { "epoch": 2.55, "learning_rate": 7.445934289970987e-05, "loss": 2.9287, "step": 32572 }, { "epoch": 2.55, "learning_rate": 7.445855877048538e-05, "loss": 2.7396, "step": 32573 }, { "epoch": 2.55, "learning_rate": 7.445777464126088e-05, "loss": 2.7187, "step": 32574 }, { "epoch": 2.55, "learning_rate": 7.445699051203639e-05, "loss": 2.8803, "step": 32575 }, { "epoch": 2.55, "learning_rate": 7.445620638281189e-05, "loss": 2.6929, "step": 32576 }, { "epoch": 2.55, "learning_rate": 7.445542225358739e-05, "loss": 2.8242, "step": 32577 }, { "epoch": 2.55, "learning_rate": 7.44546381243629e-05, "loss": 2.8128, "step": 32578 }, { "epoch": 2.55, "learning_rate": 7.44538539951384e-05, "loss": 2.7885, "step": 32579 }, { "epoch": 2.55, "learning_rate": 7.44530698659139e-05, "loss": 2.8551, "step": 32580 }, { "epoch": 2.55, "learning_rate": 7.445228573668941e-05, "loss": 3.1024, "step": 32581 }, { "epoch": 2.55, "learning_rate": 7.445150160746491e-05, "loss": 2.7631, "step": 32582 }, { "epoch": 2.55, "learning_rate": 7.445071747824041e-05, "loss": 2.7207, "step": 32583 }, { "epoch": 2.56, "learning_rate": 7.444993334901592e-05, "loss": 2.8996, "step": 32584 }, { "epoch": 2.56, "learning_rate": 7.444914921979142e-05, "loss": 2.8712, "step": 32585 }, { "epoch": 2.56, "learning_rate": 7.444836509056693e-05, "loss": 2.699, "step": 32586 }, { "epoch": 2.56, "learning_rate": 7.444758096134243e-05, "loss": 2.8561, "step": 32587 }, { "epoch": 2.56, "learning_rate": 7.444679683211793e-05, "loss": 2.943, "step": 32588 }, { "epoch": 2.56, "learning_rate": 7.444601270289344e-05, "loss": 2.7734, "step": 32589 }, { "epoch": 2.56, "learning_rate": 7.444522857366894e-05, "loss": 2.6722, "step": 32590 }, { "epoch": 2.56, "learning_rate": 7.444444444444444e-05, "loss": 2.8467, "step": 32591 }, { "epoch": 2.56, "learning_rate": 7.444366031521995e-05, "loss": 2.8379, "step": 32592 }, { "epoch": 2.56, "learning_rate": 7.444287618599546e-05, "loss": 2.6468, "step": 32593 }, { "epoch": 2.56, "learning_rate": 7.444209205677097e-05, "loss": 2.7067, "step": 32594 }, { "epoch": 2.56, "learning_rate": 7.444130792754647e-05, "loss": 2.7611, "step": 32595 }, { "epoch": 2.56, "learning_rate": 7.444052379832198e-05, "loss": 2.6485, "step": 32596 }, { "epoch": 2.56, "learning_rate": 7.443973966909748e-05, "loss": 2.7204, "step": 32597 }, { "epoch": 2.56, "learning_rate": 7.443895553987298e-05, "loss": 2.6923, "step": 32598 }, { "epoch": 2.56, "learning_rate": 7.443817141064849e-05, "loss": 2.7404, "step": 32599 }, { "epoch": 2.56, "learning_rate": 7.443738728142399e-05, "loss": 3.0317, "step": 32600 }, { "epoch": 2.56, "learning_rate": 7.44366031521995e-05, "loss": 2.762, "step": 32601 }, { "epoch": 2.56, "learning_rate": 7.4435819022975e-05, "loss": 2.8251, "step": 32602 }, { "epoch": 2.56, "learning_rate": 7.443503489375049e-05, "loss": 2.7162, "step": 32603 }, { "epoch": 2.56, "learning_rate": 7.443425076452599e-05, "loss": 2.7043, "step": 32604 }, { "epoch": 2.56, "learning_rate": 7.44334666353015e-05, "loss": 2.7399, "step": 32605 }, { "epoch": 2.56, "learning_rate": 7.4432682506077e-05, "loss": 3.1201, "step": 32606 }, { "epoch": 2.56, "learning_rate": 7.44318983768525e-05, "loss": 2.7697, "step": 32607 }, { "epoch": 2.56, "learning_rate": 7.4431114247628e-05, "loss": 2.6826, "step": 32608 }, { "epoch": 2.56, "learning_rate": 7.443033011840351e-05, "loss": 2.9107, "step": 32609 }, { "epoch": 2.56, "learning_rate": 7.442954598917901e-05, "loss": 3.0347, "step": 32610 }, { "epoch": 2.56, "learning_rate": 7.442876185995452e-05, "loss": 2.6826, "step": 32611 }, { "epoch": 2.56, "learning_rate": 7.442797773073002e-05, "loss": 2.7914, "step": 32612 }, { "epoch": 2.56, "learning_rate": 7.442719360150552e-05, "loss": 2.6352, "step": 32613 }, { "epoch": 2.56, "learning_rate": 7.442640947228103e-05, "loss": 2.7768, "step": 32614 }, { "epoch": 2.56, "learning_rate": 7.442562534305653e-05, "loss": 2.7736, "step": 32615 }, { "epoch": 2.56, "learning_rate": 7.442484121383204e-05, "loss": 2.7578, "step": 32616 }, { "epoch": 2.56, "learning_rate": 7.442405708460754e-05, "loss": 2.8627, "step": 32617 }, { "epoch": 2.56, "learning_rate": 7.442327295538304e-05, "loss": 2.7184, "step": 32618 }, { "epoch": 2.56, "learning_rate": 7.442248882615856e-05, "loss": 2.7501, "step": 32619 }, { "epoch": 2.56, "learning_rate": 7.442170469693406e-05, "loss": 2.9497, "step": 32620 }, { "epoch": 2.56, "learning_rate": 7.442092056770957e-05, "loss": 2.6741, "step": 32621 }, { "epoch": 2.56, "learning_rate": 7.442013643848507e-05, "loss": 2.8119, "step": 32622 }, { "epoch": 2.56, "learning_rate": 7.441935230926058e-05, "loss": 2.7529, "step": 32623 }, { "epoch": 2.56, "learning_rate": 7.441856818003608e-05, "loss": 3.0238, "step": 32624 }, { "epoch": 2.56, "learning_rate": 7.441778405081158e-05, "loss": 2.9395, "step": 32625 }, { "epoch": 2.56, "learning_rate": 7.441699992158709e-05, "loss": 2.9098, "step": 32626 }, { "epoch": 2.56, "learning_rate": 7.441621579236259e-05, "loss": 2.7639, "step": 32627 }, { "epoch": 2.56, "learning_rate": 7.44154316631381e-05, "loss": 2.8635, "step": 32628 }, { "epoch": 2.56, "learning_rate": 7.44146475339136e-05, "loss": 2.7534, "step": 32629 }, { "epoch": 2.56, "learning_rate": 7.44138634046891e-05, "loss": 2.7995, "step": 32630 }, { "epoch": 2.56, "learning_rate": 7.44130792754646e-05, "loss": 2.893, "step": 32631 }, { "epoch": 2.56, "learning_rate": 7.441229514624011e-05, "loss": 2.6235, "step": 32632 }, { "epoch": 2.56, "learning_rate": 7.441151101701561e-05, "loss": 2.848, "step": 32633 }, { "epoch": 2.56, "learning_rate": 7.441072688779112e-05, "loss": 2.9068, "step": 32634 }, { "epoch": 2.56, "learning_rate": 7.440994275856662e-05, "loss": 2.9063, "step": 32635 }, { "epoch": 2.56, "learning_rate": 7.440915862934212e-05, "loss": 2.8809, "step": 32636 }, { "epoch": 2.56, "learning_rate": 7.440837450011763e-05, "loss": 2.7754, "step": 32637 }, { "epoch": 2.56, "learning_rate": 7.440759037089313e-05, "loss": 2.7981, "step": 32638 }, { "epoch": 2.56, "learning_rate": 7.440680624166862e-05, "loss": 2.744, "step": 32639 }, { "epoch": 2.56, "learning_rate": 7.440602211244412e-05, "loss": 2.7096, "step": 32640 }, { "epoch": 2.56, "learning_rate": 7.440523798321963e-05, "loss": 2.7217, "step": 32641 }, { "epoch": 2.56, "learning_rate": 7.440445385399513e-05, "loss": 2.8642, "step": 32642 }, { "epoch": 2.56, "learning_rate": 7.440366972477064e-05, "loss": 2.8124, "step": 32643 }, { "epoch": 2.56, "learning_rate": 7.440288559554615e-05, "loss": 2.8046, "step": 32644 }, { "epoch": 2.56, "learning_rate": 7.440210146632166e-05, "loss": 2.709, "step": 32645 }, { "epoch": 2.56, "learning_rate": 7.440131733709716e-05, "loss": 2.9405, "step": 32646 }, { "epoch": 2.56, "learning_rate": 7.440053320787266e-05, "loss": 2.8047, "step": 32647 }, { "epoch": 2.56, "learning_rate": 7.439974907864817e-05, "loss": 2.8204, "step": 32648 }, { "epoch": 2.56, "learning_rate": 7.439896494942367e-05, "loss": 2.7677, "step": 32649 }, { "epoch": 2.56, "learning_rate": 7.439818082019917e-05, "loss": 2.886, "step": 32650 }, { "epoch": 2.56, "learning_rate": 7.439739669097468e-05, "loss": 2.7283, "step": 32651 }, { "epoch": 2.56, "learning_rate": 7.439661256175018e-05, "loss": 2.6671, "step": 32652 }, { "epoch": 2.56, "learning_rate": 7.439582843252569e-05, "loss": 2.6481, "step": 32653 }, { "epoch": 2.56, "learning_rate": 7.439504430330119e-05, "loss": 2.969, "step": 32654 }, { "epoch": 2.56, "learning_rate": 7.439426017407669e-05, "loss": 2.6729, "step": 32655 }, { "epoch": 2.56, "learning_rate": 7.43934760448522e-05, "loss": 2.8755, "step": 32656 }, { "epoch": 2.56, "learning_rate": 7.43926919156277e-05, "loss": 2.7173, "step": 32657 }, { "epoch": 2.56, "learning_rate": 7.43919077864032e-05, "loss": 2.6465, "step": 32658 }, { "epoch": 2.56, "learning_rate": 7.439112365717871e-05, "loss": 3.0109, "step": 32659 }, { "epoch": 2.56, "learning_rate": 7.439033952795421e-05, "loss": 2.8568, "step": 32660 }, { "epoch": 2.56, "learning_rate": 7.438955539872971e-05, "loss": 2.7583, "step": 32661 }, { "epoch": 2.56, "learning_rate": 7.438877126950522e-05, "loss": 2.7884, "step": 32662 }, { "epoch": 2.56, "learning_rate": 7.438798714028072e-05, "loss": 2.7635, "step": 32663 }, { "epoch": 2.56, "learning_rate": 7.438720301105623e-05, "loss": 2.9313, "step": 32664 }, { "epoch": 2.56, "learning_rate": 7.438641888183173e-05, "loss": 2.7794, "step": 32665 }, { "epoch": 2.56, "learning_rate": 7.438563475260723e-05, "loss": 2.8501, "step": 32666 }, { "epoch": 2.56, "learning_rate": 7.438485062338274e-05, "loss": 2.9235, "step": 32667 }, { "epoch": 2.56, "learning_rate": 7.438406649415824e-05, "loss": 2.7067, "step": 32668 }, { "epoch": 2.56, "learning_rate": 7.438328236493374e-05, "loss": 2.8876, "step": 32669 }, { "epoch": 2.56, "learning_rate": 7.438249823570925e-05, "loss": 2.6135, "step": 32670 }, { "epoch": 2.56, "learning_rate": 7.438171410648475e-05, "loss": 2.8656, "step": 32671 }, { "epoch": 2.56, "learning_rate": 7.438092997726026e-05, "loss": 2.8432, "step": 32672 }, { "epoch": 2.56, "learning_rate": 7.438014584803576e-05, "loss": 3.1138, "step": 32673 }, { "epoch": 2.56, "learning_rate": 7.437936171881126e-05, "loss": 2.7541, "step": 32674 }, { "epoch": 2.56, "learning_rate": 7.437857758958677e-05, "loss": 2.9498, "step": 32675 }, { "epoch": 2.56, "learning_rate": 7.437779346036227e-05, "loss": 2.9593, "step": 32676 }, { "epoch": 2.56, "learning_rate": 7.437700933113777e-05, "loss": 2.7417, "step": 32677 }, { "epoch": 2.56, "learning_rate": 7.437622520191328e-05, "loss": 2.7297, "step": 32678 }, { "epoch": 2.56, "learning_rate": 7.437544107268878e-05, "loss": 2.8097, "step": 32679 }, { "epoch": 2.56, "learning_rate": 7.437465694346428e-05, "loss": 2.7075, "step": 32680 }, { "epoch": 2.56, "learning_rate": 7.437387281423979e-05, "loss": 2.7693, "step": 32681 }, { "epoch": 2.56, "learning_rate": 7.437308868501529e-05, "loss": 2.6739, "step": 32682 }, { "epoch": 2.56, "learning_rate": 7.43723045557908e-05, "loss": 2.7962, "step": 32683 }, { "epoch": 2.56, "learning_rate": 7.43715204265663e-05, "loss": 2.7437, "step": 32684 }, { "epoch": 2.56, "learning_rate": 7.43707362973418e-05, "loss": 2.9012, "step": 32685 }, { "epoch": 2.56, "learning_rate": 7.436995216811731e-05, "loss": 2.8875, "step": 32686 }, { "epoch": 2.56, "learning_rate": 7.436916803889281e-05, "loss": 2.7995, "step": 32687 }, { "epoch": 2.56, "learning_rate": 7.436838390966831e-05, "loss": 2.6832, "step": 32688 }, { "epoch": 2.56, "learning_rate": 7.436759978044382e-05, "loss": 2.9411, "step": 32689 }, { "epoch": 2.56, "learning_rate": 7.436681565121932e-05, "loss": 2.6406, "step": 32690 }, { "epoch": 2.56, "learning_rate": 7.436603152199483e-05, "loss": 2.6945, "step": 32691 }, { "epoch": 2.56, "learning_rate": 7.436524739277033e-05, "loss": 2.7976, "step": 32692 }, { "epoch": 2.56, "learning_rate": 7.436446326354583e-05, "loss": 2.8038, "step": 32693 }, { "epoch": 2.56, "learning_rate": 7.436367913432134e-05, "loss": 2.6493, "step": 32694 }, { "epoch": 2.56, "learning_rate": 7.436289500509684e-05, "loss": 2.9419, "step": 32695 }, { "epoch": 2.56, "learning_rate": 7.436211087587234e-05, "loss": 2.8075, "step": 32696 }, { "epoch": 2.56, "learning_rate": 7.436132674664785e-05, "loss": 2.7052, "step": 32697 }, { "epoch": 2.56, "learning_rate": 7.436054261742335e-05, "loss": 2.7857, "step": 32698 }, { "epoch": 2.56, "learning_rate": 7.435975848819885e-05, "loss": 2.7436, "step": 32699 }, { "epoch": 2.56, "learning_rate": 7.435897435897436e-05, "loss": 2.6917, "step": 32700 }, { "epoch": 2.56, "learning_rate": 7.435819022974988e-05, "loss": 2.9206, "step": 32701 }, { "epoch": 2.56, "learning_rate": 7.435740610052538e-05, "loss": 2.8633, "step": 32702 }, { "epoch": 2.56, "learning_rate": 7.435662197130088e-05, "loss": 2.7724, "step": 32703 }, { "epoch": 2.56, "learning_rate": 7.435583784207639e-05, "loss": 2.6036, "step": 32704 }, { "epoch": 2.56, "learning_rate": 7.435505371285189e-05, "loss": 2.8317, "step": 32705 }, { "epoch": 2.56, "learning_rate": 7.43542695836274e-05, "loss": 2.6645, "step": 32706 }, { "epoch": 2.56, "learning_rate": 7.43534854544029e-05, "loss": 2.7652, "step": 32707 }, { "epoch": 2.56, "learning_rate": 7.43527013251784e-05, "loss": 2.7104, "step": 32708 }, { "epoch": 2.56, "learning_rate": 7.43519171959539e-05, "loss": 2.7023, "step": 32709 }, { "epoch": 2.56, "learning_rate": 7.435113306672941e-05, "loss": 2.7457, "step": 32710 }, { "epoch": 2.56, "learning_rate": 7.43503489375049e-05, "loss": 2.6928, "step": 32711 }, { "epoch": 2.57, "learning_rate": 7.43495648082804e-05, "loss": 2.7574, "step": 32712 }, { "epoch": 2.57, "learning_rate": 7.43487806790559e-05, "loss": 2.6851, "step": 32713 }, { "epoch": 2.57, "learning_rate": 7.434799654983141e-05, "loss": 2.9752, "step": 32714 }, { "epoch": 2.57, "learning_rate": 7.434721242060691e-05, "loss": 2.7004, "step": 32715 }, { "epoch": 2.57, "learning_rate": 7.434642829138242e-05, "loss": 2.8049, "step": 32716 }, { "epoch": 2.57, "learning_rate": 7.434564416215792e-05, "loss": 2.9865, "step": 32717 }, { "epoch": 2.57, "learning_rate": 7.434486003293342e-05, "loss": 2.8217, "step": 32718 }, { "epoch": 2.57, "learning_rate": 7.434407590370893e-05, "loss": 2.8145, "step": 32719 }, { "epoch": 2.57, "learning_rate": 7.434329177448443e-05, "loss": 2.642, "step": 32720 }, { "epoch": 2.57, "learning_rate": 7.434250764525994e-05, "loss": 2.6554, "step": 32721 }, { "epoch": 2.57, "learning_rate": 7.434172351603544e-05, "loss": 2.8252, "step": 32722 }, { "epoch": 2.57, "learning_rate": 7.434093938681094e-05, "loss": 2.9581, "step": 32723 }, { "epoch": 2.57, "learning_rate": 7.434015525758645e-05, "loss": 2.7626, "step": 32724 }, { "epoch": 2.57, "learning_rate": 7.433937112836195e-05, "loss": 2.7513, "step": 32725 }, { "epoch": 2.57, "learning_rate": 7.433858699913745e-05, "loss": 2.9285, "step": 32726 }, { "epoch": 2.57, "learning_rate": 7.433780286991297e-05, "loss": 2.9301, "step": 32727 }, { "epoch": 2.57, "learning_rate": 7.433701874068847e-05, "loss": 2.9221, "step": 32728 }, { "epoch": 2.57, "learning_rate": 7.433623461146398e-05, "loss": 2.8158, "step": 32729 }, { "epoch": 2.57, "learning_rate": 7.433545048223948e-05, "loss": 2.7533, "step": 32730 }, { "epoch": 2.57, "learning_rate": 7.433466635301499e-05, "loss": 2.9478, "step": 32731 }, { "epoch": 2.57, "learning_rate": 7.433388222379049e-05, "loss": 2.7743, "step": 32732 }, { "epoch": 2.57, "learning_rate": 7.4333098094566e-05, "loss": 2.7761, "step": 32733 }, { "epoch": 2.57, "learning_rate": 7.43323139653415e-05, "loss": 2.7564, "step": 32734 }, { "epoch": 2.57, "learning_rate": 7.4331529836117e-05, "loss": 2.9145, "step": 32735 }, { "epoch": 2.57, "learning_rate": 7.43307457068925e-05, "loss": 2.7895, "step": 32736 }, { "epoch": 2.57, "learning_rate": 7.432996157766801e-05, "loss": 2.6597, "step": 32737 }, { "epoch": 2.57, "learning_rate": 7.432917744844351e-05, "loss": 2.8225, "step": 32738 }, { "epoch": 2.57, "learning_rate": 7.432839331921902e-05, "loss": 2.7133, "step": 32739 }, { "epoch": 2.57, "learning_rate": 7.432760918999452e-05, "loss": 2.8831, "step": 32740 }, { "epoch": 2.57, "learning_rate": 7.432682506077002e-05, "loss": 2.8038, "step": 32741 }, { "epoch": 2.57, "learning_rate": 7.432604093154553e-05, "loss": 2.6896, "step": 32742 }, { "epoch": 2.57, "learning_rate": 7.432525680232103e-05, "loss": 2.6546, "step": 32743 }, { "epoch": 2.57, "learning_rate": 7.432447267309653e-05, "loss": 2.8043, "step": 32744 }, { "epoch": 2.57, "learning_rate": 7.432368854387204e-05, "loss": 2.6688, "step": 32745 }, { "epoch": 2.57, "learning_rate": 7.432290441464754e-05, "loss": 2.9564, "step": 32746 }, { "epoch": 2.57, "learning_rate": 7.432212028542303e-05, "loss": 2.5416, "step": 32747 }, { "epoch": 2.57, "learning_rate": 7.432133615619853e-05, "loss": 2.7042, "step": 32748 }, { "epoch": 2.57, "learning_rate": 7.432055202697404e-05, "loss": 2.8004, "step": 32749 }, { "epoch": 2.57, "learning_rate": 7.431976789774954e-05, "loss": 2.6756, "step": 32750 }, { "epoch": 2.57, "learning_rate": 7.431898376852505e-05, "loss": 2.8421, "step": 32751 }, { "epoch": 2.57, "learning_rate": 7.431819963930056e-05, "loss": 2.9877, "step": 32752 }, { "epoch": 2.57, "learning_rate": 7.431741551007607e-05, "loss": 2.7632, "step": 32753 }, { "epoch": 2.57, "learning_rate": 7.431663138085157e-05, "loss": 2.8333, "step": 32754 }, { "epoch": 2.57, "learning_rate": 7.431584725162707e-05, "loss": 2.722, "step": 32755 }, { "epoch": 2.57, "learning_rate": 7.431506312240258e-05, "loss": 2.8053, "step": 32756 }, { "epoch": 2.57, "learning_rate": 7.431427899317808e-05, "loss": 2.9152, "step": 32757 }, { "epoch": 2.57, "learning_rate": 7.431349486395359e-05, "loss": 2.7428, "step": 32758 }, { "epoch": 2.57, "learning_rate": 7.431271073472909e-05, "loss": 2.7172, "step": 32759 }, { "epoch": 2.57, "learning_rate": 7.431192660550459e-05, "loss": 2.8736, "step": 32760 }, { "epoch": 2.57, "learning_rate": 7.43111424762801e-05, "loss": 2.7764, "step": 32761 }, { "epoch": 2.57, "learning_rate": 7.43103583470556e-05, "loss": 2.5924, "step": 32762 }, { "epoch": 2.57, "learning_rate": 7.43095742178311e-05, "loss": 2.8544, "step": 32763 }, { "epoch": 2.57, "learning_rate": 7.430879008860661e-05, "loss": 2.9282, "step": 32764 }, { "epoch": 2.57, "learning_rate": 7.430800595938211e-05, "loss": 2.6517, "step": 32765 }, { "epoch": 2.57, "learning_rate": 7.430722183015761e-05, "loss": 2.9828, "step": 32766 }, { "epoch": 2.57, "learning_rate": 7.430643770093312e-05, "loss": 2.5251, "step": 32767 }, { "epoch": 2.57, "learning_rate": 7.430565357170862e-05, "loss": 2.5763, "step": 32768 }, { "epoch": 2.57, "learning_rate": 7.430486944248413e-05, "loss": 2.849, "step": 32769 }, { "epoch": 2.57, "learning_rate": 7.430408531325963e-05, "loss": 2.8586, "step": 32770 }, { "epoch": 2.57, "learning_rate": 7.430330118403513e-05, "loss": 2.7164, "step": 32771 }, { "epoch": 2.57, "learning_rate": 7.430251705481064e-05, "loss": 2.6235, "step": 32772 }, { "epoch": 2.57, "learning_rate": 7.430173292558614e-05, "loss": 2.9769, "step": 32773 }, { "epoch": 2.57, "learning_rate": 7.430094879636164e-05, "loss": 2.6838, "step": 32774 }, { "epoch": 2.57, "learning_rate": 7.430016466713715e-05, "loss": 2.9105, "step": 32775 }, { "epoch": 2.57, "learning_rate": 7.429938053791265e-05, "loss": 2.7379, "step": 32776 }, { "epoch": 2.57, "learning_rate": 7.429859640868816e-05, "loss": 2.8971, "step": 32777 }, { "epoch": 2.57, "learning_rate": 7.429781227946366e-05, "loss": 2.6021, "step": 32778 }, { "epoch": 2.57, "learning_rate": 7.429702815023916e-05, "loss": 2.6338, "step": 32779 }, { "epoch": 2.57, "learning_rate": 7.429624402101467e-05, "loss": 2.793, "step": 32780 }, { "epoch": 2.57, "learning_rate": 7.429545989179017e-05, "loss": 2.833, "step": 32781 }, { "epoch": 2.57, "learning_rate": 7.429467576256567e-05, "loss": 2.6839, "step": 32782 }, { "epoch": 2.57, "learning_rate": 7.429389163334118e-05, "loss": 2.9538, "step": 32783 }, { "epoch": 2.57, "learning_rate": 7.429310750411668e-05, "loss": 2.7407, "step": 32784 }, { "epoch": 2.57, "learning_rate": 7.429232337489218e-05, "loss": 2.8478, "step": 32785 }, { "epoch": 2.57, "learning_rate": 7.429153924566769e-05, "loss": 2.8448, "step": 32786 }, { "epoch": 2.57, "learning_rate": 7.429075511644319e-05, "loss": 2.9004, "step": 32787 }, { "epoch": 2.57, "learning_rate": 7.42899709872187e-05, "loss": 2.6391, "step": 32788 }, { "epoch": 2.57, "learning_rate": 7.42891868579942e-05, "loss": 3.0025, "step": 32789 }, { "epoch": 2.57, "learning_rate": 7.42884027287697e-05, "loss": 2.6859, "step": 32790 }, { "epoch": 2.57, "learning_rate": 7.42876185995452e-05, "loss": 2.7529, "step": 32791 }, { "epoch": 2.57, "learning_rate": 7.428683447032071e-05, "loss": 2.7883, "step": 32792 }, { "epoch": 2.57, "learning_rate": 7.428605034109621e-05, "loss": 2.8698, "step": 32793 }, { "epoch": 2.57, "learning_rate": 7.428526621187172e-05, "loss": 2.8554, "step": 32794 }, { "epoch": 2.57, "learning_rate": 7.428448208264722e-05, "loss": 3.0494, "step": 32795 }, { "epoch": 2.57, "learning_rate": 7.428369795342272e-05, "loss": 2.9493, "step": 32796 }, { "epoch": 2.57, "learning_rate": 7.428291382419823e-05, "loss": 2.6235, "step": 32797 }, { "epoch": 2.57, "learning_rate": 7.428212969497373e-05, "loss": 2.9932, "step": 32798 }, { "epoch": 2.57, "learning_rate": 7.428134556574924e-05, "loss": 2.7883, "step": 32799 }, { "epoch": 2.57, "learning_rate": 7.428056143652474e-05, "loss": 2.8323, "step": 32800 }, { "epoch": 2.57, "learning_rate": 7.427977730730024e-05, "loss": 2.8792, "step": 32801 }, { "epoch": 2.57, "learning_rate": 7.427899317807575e-05, "loss": 2.8464, "step": 32802 }, { "epoch": 2.57, "learning_rate": 7.427820904885125e-05, "loss": 2.6617, "step": 32803 }, { "epoch": 2.57, "learning_rate": 7.427742491962675e-05, "loss": 2.6211, "step": 32804 }, { "epoch": 2.57, "learning_rate": 7.427664079040226e-05, "loss": 2.7869, "step": 32805 }, { "epoch": 2.57, "learning_rate": 7.427585666117776e-05, "loss": 2.6977, "step": 32806 }, { "epoch": 2.57, "learning_rate": 7.427507253195327e-05, "loss": 2.6272, "step": 32807 }, { "epoch": 2.57, "learning_rate": 7.427428840272877e-05, "loss": 2.7138, "step": 32808 }, { "epoch": 2.57, "learning_rate": 7.427350427350429e-05, "loss": 2.9675, "step": 32809 }, { "epoch": 2.57, "learning_rate": 7.427272014427979e-05, "loss": 2.7232, "step": 32810 }, { "epoch": 2.57, "learning_rate": 7.42719360150553e-05, "loss": 3.0096, "step": 32811 }, { "epoch": 2.57, "learning_rate": 7.42711518858308e-05, "loss": 2.7835, "step": 32812 }, { "epoch": 2.57, "learning_rate": 7.42703677566063e-05, "loss": 2.8655, "step": 32813 }, { "epoch": 2.57, "learning_rate": 7.42695836273818e-05, "loss": 2.8586, "step": 32814 }, { "epoch": 2.57, "learning_rate": 7.426879949815731e-05, "loss": 2.6972, "step": 32815 }, { "epoch": 2.57, "learning_rate": 7.426801536893281e-05, "loss": 2.9867, "step": 32816 }, { "epoch": 2.57, "learning_rate": 7.426723123970832e-05, "loss": 2.7955, "step": 32817 }, { "epoch": 2.57, "learning_rate": 7.426644711048382e-05, "loss": 2.8822, "step": 32818 }, { "epoch": 2.57, "learning_rate": 7.426566298125931e-05, "loss": 2.804, "step": 32819 }, { "epoch": 2.57, "learning_rate": 7.426487885203481e-05, "loss": 2.7329, "step": 32820 }, { "epoch": 2.57, "learning_rate": 7.426409472281032e-05, "loss": 2.742, "step": 32821 }, { "epoch": 2.57, "learning_rate": 7.426331059358582e-05, "loss": 2.6201, "step": 32822 }, { "epoch": 2.57, "learning_rate": 7.426252646436132e-05, "loss": 2.9777, "step": 32823 }, { "epoch": 2.57, "learning_rate": 7.426174233513683e-05, "loss": 2.6747, "step": 32824 }, { "epoch": 2.57, "learning_rate": 7.426095820591233e-05, "loss": 2.7068, "step": 32825 }, { "epoch": 2.57, "learning_rate": 7.426017407668784e-05, "loss": 2.5447, "step": 32826 }, { "epoch": 2.57, "learning_rate": 7.425938994746334e-05, "loss": 2.7466, "step": 32827 }, { "epoch": 2.57, "learning_rate": 7.425860581823884e-05, "loss": 2.7989, "step": 32828 }, { "epoch": 2.57, "learning_rate": 7.425782168901435e-05, "loss": 2.7949, "step": 32829 }, { "epoch": 2.57, "learning_rate": 7.425703755978985e-05, "loss": 2.9079, "step": 32830 }, { "epoch": 2.57, "learning_rate": 7.425625343056535e-05, "loss": 2.9195, "step": 32831 }, { "epoch": 2.57, "learning_rate": 7.425546930134086e-05, "loss": 2.9817, "step": 32832 }, { "epoch": 2.57, "learning_rate": 7.425468517211636e-05, "loss": 3.0316, "step": 32833 }, { "epoch": 2.57, "learning_rate": 7.425390104289186e-05, "loss": 2.8272, "step": 32834 }, { "epoch": 2.57, "learning_rate": 7.425311691366738e-05, "loss": 2.8537, "step": 32835 }, { "epoch": 2.57, "learning_rate": 7.425233278444289e-05, "loss": 2.8518, "step": 32836 }, { "epoch": 2.57, "learning_rate": 7.425154865521839e-05, "loss": 2.7042, "step": 32837 }, { "epoch": 2.57, "learning_rate": 7.425076452599389e-05, "loss": 2.6963, "step": 32838 }, { "epoch": 2.58, "learning_rate": 7.42499803967694e-05, "loss": 2.6319, "step": 32839 }, { "epoch": 2.58, "learning_rate": 7.42491962675449e-05, "loss": 2.8914, "step": 32840 }, { "epoch": 2.58, "learning_rate": 7.42484121383204e-05, "loss": 2.7179, "step": 32841 }, { "epoch": 2.58, "learning_rate": 7.424762800909591e-05, "loss": 2.6814, "step": 32842 }, { "epoch": 2.58, "learning_rate": 7.424684387987141e-05, "loss": 2.8179, "step": 32843 }, { "epoch": 2.58, "learning_rate": 7.424605975064692e-05, "loss": 2.7523, "step": 32844 }, { "epoch": 2.58, "learning_rate": 7.424527562142242e-05, "loss": 2.8851, "step": 32845 }, { "epoch": 2.58, "learning_rate": 7.424449149219792e-05, "loss": 2.7528, "step": 32846 }, { "epoch": 2.58, "learning_rate": 7.424370736297343e-05, "loss": 2.75, "step": 32847 }, { "epoch": 2.58, "learning_rate": 7.424292323374893e-05, "loss": 2.856, "step": 32848 }, { "epoch": 2.58, "learning_rate": 7.424213910452443e-05, "loss": 2.7844, "step": 32849 }, { "epoch": 2.58, "learning_rate": 7.424135497529994e-05, "loss": 2.7949, "step": 32850 }, { "epoch": 2.58, "learning_rate": 7.424057084607544e-05, "loss": 2.7713, "step": 32851 }, { "epoch": 2.58, "learning_rate": 7.423978671685094e-05, "loss": 2.7049, "step": 32852 }, { "epoch": 2.58, "learning_rate": 7.423900258762645e-05, "loss": 2.606, "step": 32853 }, { "epoch": 2.58, "learning_rate": 7.423821845840195e-05, "loss": 2.8328, "step": 32854 }, { "epoch": 2.58, "learning_rate": 7.423743432917744e-05, "loss": 2.6756, "step": 32855 }, { "epoch": 2.58, "learning_rate": 7.423665019995295e-05, "loss": 3.2864, "step": 32856 }, { "epoch": 2.58, "learning_rate": 7.423586607072845e-05, "loss": 2.8012, "step": 32857 }, { "epoch": 2.58, "learning_rate": 7.423508194150395e-05, "loss": 2.6923, "step": 32858 }, { "epoch": 2.58, "learning_rate": 7.423429781227946e-05, "loss": 2.7439, "step": 32859 }, { "epoch": 2.58, "learning_rate": 7.423351368305496e-05, "loss": 2.8701, "step": 32860 }, { "epoch": 2.58, "learning_rate": 7.423272955383048e-05, "loss": 2.7884, "step": 32861 }, { "epoch": 2.58, "learning_rate": 7.423194542460598e-05, "loss": 2.7373, "step": 32862 }, { "epoch": 2.58, "learning_rate": 7.423116129538148e-05, "loss": 2.7869, "step": 32863 }, { "epoch": 2.58, "learning_rate": 7.423037716615699e-05, "loss": 2.6924, "step": 32864 }, { "epoch": 2.58, "learning_rate": 7.422959303693249e-05, "loss": 2.8252, "step": 32865 }, { "epoch": 2.58, "learning_rate": 7.4228808907708e-05, "loss": 2.8112, "step": 32866 }, { "epoch": 2.58, "learning_rate": 7.42280247784835e-05, "loss": 2.7139, "step": 32867 }, { "epoch": 2.58, "learning_rate": 7.4227240649259e-05, "loss": 2.9647, "step": 32868 }, { "epoch": 2.58, "learning_rate": 7.422645652003451e-05, "loss": 2.8759, "step": 32869 }, { "epoch": 2.58, "learning_rate": 7.422567239081001e-05, "loss": 2.7109, "step": 32870 }, { "epoch": 2.58, "learning_rate": 7.422488826158551e-05, "loss": 2.7535, "step": 32871 }, { "epoch": 2.58, "learning_rate": 7.422410413236102e-05, "loss": 2.7425, "step": 32872 }, { "epoch": 2.58, "learning_rate": 7.422332000313652e-05, "loss": 2.8782, "step": 32873 }, { "epoch": 2.58, "learning_rate": 7.422253587391203e-05, "loss": 2.7792, "step": 32874 }, { "epoch": 2.58, "learning_rate": 7.422175174468753e-05, "loss": 2.7611, "step": 32875 }, { "epoch": 2.58, "learning_rate": 7.422096761546303e-05, "loss": 2.8194, "step": 32876 }, { "epoch": 2.58, "learning_rate": 7.422018348623854e-05, "loss": 2.8669, "step": 32877 }, { "epoch": 2.58, "learning_rate": 7.421939935701404e-05, "loss": 2.569, "step": 32878 }, { "epoch": 2.58, "learning_rate": 7.421861522778954e-05, "loss": 2.7567, "step": 32879 }, { "epoch": 2.58, "learning_rate": 7.421783109856505e-05, "loss": 2.7503, "step": 32880 }, { "epoch": 2.58, "learning_rate": 7.421704696934055e-05, "loss": 2.9225, "step": 32881 }, { "epoch": 2.58, "learning_rate": 7.421626284011605e-05, "loss": 2.9559, "step": 32882 }, { "epoch": 2.58, "learning_rate": 7.421547871089156e-05, "loss": 2.7242, "step": 32883 }, { "epoch": 2.58, "learning_rate": 7.421469458166706e-05, "loss": 2.9538, "step": 32884 }, { "epoch": 2.58, "learning_rate": 7.421391045244257e-05, "loss": 2.9023, "step": 32885 }, { "epoch": 2.58, "learning_rate": 7.421312632321807e-05, "loss": 2.6247, "step": 32886 }, { "epoch": 2.58, "learning_rate": 7.421234219399357e-05, "loss": 2.6067, "step": 32887 }, { "epoch": 2.58, "learning_rate": 7.421155806476908e-05, "loss": 2.7821, "step": 32888 }, { "epoch": 2.58, "learning_rate": 7.421077393554458e-05, "loss": 2.7647, "step": 32889 }, { "epoch": 2.58, "learning_rate": 7.420998980632008e-05, "loss": 2.6802, "step": 32890 }, { "epoch": 2.58, "learning_rate": 7.420920567709559e-05, "loss": 3.048, "step": 32891 }, { "epoch": 2.58, "learning_rate": 7.420842154787109e-05, "loss": 2.7178, "step": 32892 }, { "epoch": 2.58, "learning_rate": 7.42076374186466e-05, "loss": 2.6545, "step": 32893 }, { "epoch": 2.58, "learning_rate": 7.42068532894221e-05, "loss": 2.6857, "step": 32894 }, { "epoch": 2.58, "learning_rate": 7.42060691601976e-05, "loss": 2.6177, "step": 32895 }, { "epoch": 2.58, "learning_rate": 7.42052850309731e-05, "loss": 2.7304, "step": 32896 }, { "epoch": 2.58, "learning_rate": 7.420450090174861e-05, "loss": 2.7863, "step": 32897 }, { "epoch": 2.58, "learning_rate": 7.420371677252411e-05, "loss": 2.5725, "step": 32898 }, { "epoch": 2.58, "learning_rate": 7.420293264329962e-05, "loss": 2.6437, "step": 32899 }, { "epoch": 2.58, "learning_rate": 7.420214851407512e-05, "loss": 2.8353, "step": 32900 }, { "epoch": 2.58, "learning_rate": 7.420136438485062e-05, "loss": 3.0041, "step": 32901 }, { "epoch": 2.58, "learning_rate": 7.420058025562613e-05, "loss": 2.9543, "step": 32902 }, { "epoch": 2.58, "learning_rate": 7.419979612640163e-05, "loss": 2.8585, "step": 32903 }, { "epoch": 2.58, "learning_rate": 7.419901199717714e-05, "loss": 2.7988, "step": 32904 }, { "epoch": 2.58, "learning_rate": 7.419822786795264e-05, "loss": 2.7674, "step": 32905 }, { "epoch": 2.58, "learning_rate": 7.419744373872814e-05, "loss": 2.9277, "step": 32906 }, { "epoch": 2.58, "learning_rate": 7.419665960950365e-05, "loss": 2.7751, "step": 32907 }, { "epoch": 2.58, "learning_rate": 7.419587548027915e-05, "loss": 2.7064, "step": 32908 }, { "epoch": 2.58, "learning_rate": 7.419509135105465e-05, "loss": 2.7102, "step": 32909 }, { "epoch": 2.58, "learning_rate": 7.419430722183016e-05, "loss": 2.8424, "step": 32910 }, { "epoch": 2.58, "learning_rate": 7.419352309260566e-05, "loss": 2.7199, "step": 32911 }, { "epoch": 2.58, "learning_rate": 7.419273896338117e-05, "loss": 2.7181, "step": 32912 }, { "epoch": 2.58, "learning_rate": 7.419195483415667e-05, "loss": 2.7916, "step": 32913 }, { "epoch": 2.58, "learning_rate": 7.419117070493217e-05, "loss": 2.8113, "step": 32914 }, { "epoch": 2.58, "learning_rate": 7.419038657570768e-05, "loss": 2.7038, "step": 32915 }, { "epoch": 2.58, "learning_rate": 7.418960244648318e-05, "loss": 2.8436, "step": 32916 }, { "epoch": 2.58, "learning_rate": 7.41888183172587e-05, "loss": 2.5273, "step": 32917 }, { "epoch": 2.58, "learning_rate": 7.41880341880342e-05, "loss": 2.9308, "step": 32918 }, { "epoch": 2.58, "learning_rate": 7.41872500588097e-05, "loss": 2.8472, "step": 32919 }, { "epoch": 2.58, "learning_rate": 7.418646592958521e-05, "loss": 2.6857, "step": 32920 }, { "epoch": 2.58, "learning_rate": 7.418568180036071e-05, "loss": 2.8888, "step": 32921 }, { "epoch": 2.58, "learning_rate": 7.418489767113622e-05, "loss": 2.8444, "step": 32922 }, { "epoch": 2.58, "learning_rate": 7.418411354191172e-05, "loss": 2.9163, "step": 32923 }, { "epoch": 2.58, "learning_rate": 7.418332941268722e-05, "loss": 2.9045, "step": 32924 }, { "epoch": 2.58, "learning_rate": 7.418254528346273e-05, "loss": 2.7891, "step": 32925 }, { "epoch": 2.58, "learning_rate": 7.418176115423823e-05, "loss": 2.7815, "step": 32926 }, { "epoch": 2.58, "learning_rate": 7.418097702501372e-05, "loss": 2.8114, "step": 32927 }, { "epoch": 2.58, "learning_rate": 7.418019289578922e-05, "loss": 2.6487, "step": 32928 }, { "epoch": 2.58, "learning_rate": 7.417940876656473e-05, "loss": 2.7942, "step": 32929 }, { "epoch": 2.58, "learning_rate": 7.417862463734023e-05, "loss": 2.7272, "step": 32930 }, { "epoch": 2.58, "learning_rate": 7.417784050811574e-05, "loss": 2.826, "step": 32931 }, { "epoch": 2.58, "learning_rate": 7.417705637889124e-05, "loss": 2.6194, "step": 32932 }, { "epoch": 2.58, "learning_rate": 7.417627224966674e-05, "loss": 2.5896, "step": 32933 }, { "epoch": 2.58, "learning_rate": 7.417548812044225e-05, "loss": 2.8665, "step": 32934 }, { "epoch": 2.58, "learning_rate": 7.417470399121775e-05, "loss": 2.8983, "step": 32935 }, { "epoch": 2.58, "learning_rate": 7.417391986199325e-05, "loss": 2.7093, "step": 32936 }, { "epoch": 2.58, "learning_rate": 7.417313573276876e-05, "loss": 2.6778, "step": 32937 }, { "epoch": 2.58, "learning_rate": 7.417235160354426e-05, "loss": 2.9802, "step": 32938 }, { "epoch": 2.58, "learning_rate": 7.417156747431976e-05, "loss": 2.7328, "step": 32939 }, { "epoch": 2.58, "learning_rate": 7.417078334509527e-05, "loss": 2.9492, "step": 32940 }, { "epoch": 2.58, "learning_rate": 7.416999921587077e-05, "loss": 3.0284, "step": 32941 }, { "epoch": 2.58, "learning_rate": 7.416921508664628e-05, "loss": 2.806, "step": 32942 }, { "epoch": 2.58, "learning_rate": 7.416843095742179e-05, "loss": 2.7219, "step": 32943 }, { "epoch": 2.58, "learning_rate": 7.41676468281973e-05, "loss": 2.9373, "step": 32944 }, { "epoch": 2.58, "learning_rate": 7.41668626989728e-05, "loss": 2.7648, "step": 32945 }, { "epoch": 2.58, "learning_rate": 7.41660785697483e-05, "loss": 2.8509, "step": 32946 }, { "epoch": 2.58, "learning_rate": 7.416529444052381e-05, "loss": 2.7418, "step": 32947 }, { "epoch": 2.58, "learning_rate": 7.416451031129931e-05, "loss": 2.8807, "step": 32948 }, { "epoch": 2.58, "learning_rate": 7.416372618207481e-05, "loss": 2.7488, "step": 32949 }, { "epoch": 2.58, "learning_rate": 7.416294205285032e-05, "loss": 2.9839, "step": 32950 }, { "epoch": 2.58, "learning_rate": 7.416215792362582e-05, "loss": 2.6247, "step": 32951 }, { "epoch": 2.58, "learning_rate": 7.416137379440133e-05, "loss": 2.5578, "step": 32952 }, { "epoch": 2.58, "learning_rate": 7.416058966517683e-05, "loss": 2.7846, "step": 32953 }, { "epoch": 2.58, "learning_rate": 7.415980553595233e-05, "loss": 2.6801, "step": 32954 }, { "epoch": 2.58, "learning_rate": 7.415902140672784e-05, "loss": 2.9039, "step": 32955 }, { "epoch": 2.58, "learning_rate": 7.415823727750334e-05, "loss": 2.8226, "step": 32956 }, { "epoch": 2.58, "learning_rate": 7.415745314827884e-05, "loss": 2.8673, "step": 32957 }, { "epoch": 2.58, "learning_rate": 7.415666901905435e-05, "loss": 2.7407, "step": 32958 }, { "epoch": 2.58, "learning_rate": 7.415588488982985e-05, "loss": 2.8569, "step": 32959 }, { "epoch": 2.58, "learning_rate": 7.415510076060536e-05, "loss": 2.8878, "step": 32960 }, { "epoch": 2.58, "learning_rate": 7.415431663138086e-05, "loss": 2.898, "step": 32961 }, { "epoch": 2.58, "learning_rate": 7.415353250215636e-05, "loss": 2.8425, "step": 32962 }, { "epoch": 2.58, "learning_rate": 7.415274837293185e-05, "loss": 3.0827, "step": 32963 }, { "epoch": 2.58, "learning_rate": 7.415196424370736e-05, "loss": 2.9257, "step": 32964 }, { "epoch": 2.58, "learning_rate": 7.415118011448286e-05, "loss": 2.8174, "step": 32965 }, { "epoch": 2.58, "learning_rate": 7.415039598525836e-05, "loss": 2.636, "step": 32966 }, { "epoch": 2.59, "learning_rate": 7.414961185603387e-05, "loss": 2.8615, "step": 32967 }, { "epoch": 2.59, "learning_rate": 7.414882772680937e-05, "loss": 2.6724, "step": 32968 }, { "epoch": 2.59, "learning_rate": 7.414804359758489e-05, "loss": 2.843, "step": 32969 }, { "epoch": 2.59, "learning_rate": 7.414725946836039e-05, "loss": 2.8241, "step": 32970 }, { "epoch": 2.59, "learning_rate": 7.41464753391359e-05, "loss": 2.9439, "step": 32971 }, { "epoch": 2.59, "learning_rate": 7.41456912099114e-05, "loss": 2.9157, "step": 32972 }, { "epoch": 2.59, "learning_rate": 7.41449070806869e-05, "loss": 2.6665, "step": 32973 }, { "epoch": 2.59, "learning_rate": 7.41441229514624e-05, "loss": 2.6756, "step": 32974 }, { "epoch": 2.59, "learning_rate": 7.414333882223791e-05, "loss": 2.8166, "step": 32975 }, { "epoch": 2.59, "learning_rate": 7.414255469301341e-05, "loss": 2.956, "step": 32976 }, { "epoch": 2.59, "learning_rate": 7.414177056378892e-05, "loss": 3.0331, "step": 32977 }, { "epoch": 2.59, "learning_rate": 7.414098643456442e-05, "loss": 2.8047, "step": 32978 }, { "epoch": 2.59, "learning_rate": 7.414020230533993e-05, "loss": 2.6731, "step": 32979 }, { "epoch": 2.59, "learning_rate": 7.413941817611543e-05, "loss": 3.0569, "step": 32980 }, { "epoch": 2.59, "learning_rate": 7.413863404689093e-05, "loss": 3.0129, "step": 32981 }, { "epoch": 2.59, "learning_rate": 7.413784991766644e-05, "loss": 2.8887, "step": 32982 }, { "epoch": 2.59, "learning_rate": 7.413706578844194e-05, "loss": 2.7863, "step": 32983 }, { "epoch": 2.59, "learning_rate": 7.413628165921744e-05, "loss": 2.8005, "step": 32984 }, { "epoch": 2.59, "learning_rate": 7.413549752999295e-05, "loss": 2.8383, "step": 32985 }, { "epoch": 2.59, "learning_rate": 7.413471340076845e-05, "loss": 2.9223, "step": 32986 }, { "epoch": 2.59, "learning_rate": 7.413392927154395e-05, "loss": 2.6003, "step": 32987 }, { "epoch": 2.59, "learning_rate": 7.413314514231946e-05, "loss": 2.7326, "step": 32988 }, { "epoch": 2.59, "learning_rate": 7.413236101309496e-05, "loss": 2.5981, "step": 32989 }, { "epoch": 2.59, "learning_rate": 7.413157688387047e-05, "loss": 2.9352, "step": 32990 }, { "epoch": 2.59, "learning_rate": 7.413079275464597e-05, "loss": 2.8947, "step": 32991 }, { "epoch": 2.59, "learning_rate": 7.413000862542147e-05, "loss": 2.9305, "step": 32992 }, { "epoch": 2.59, "learning_rate": 7.412922449619698e-05, "loss": 2.9327, "step": 32993 }, { "epoch": 2.59, "learning_rate": 7.412844036697248e-05, "loss": 2.8558, "step": 32994 }, { "epoch": 2.59, "learning_rate": 7.412765623774798e-05, "loss": 2.7358, "step": 32995 }, { "epoch": 2.59, "learning_rate": 7.412687210852349e-05, "loss": 2.7966, "step": 32996 }, { "epoch": 2.59, "learning_rate": 7.412608797929899e-05, "loss": 2.6928, "step": 32997 }, { "epoch": 2.59, "learning_rate": 7.41253038500745e-05, "loss": 2.8328, "step": 32998 }, { "epoch": 2.59, "learning_rate": 7.412451972085e-05, "loss": 2.7867, "step": 32999 }, { "epoch": 2.59, "learning_rate": 7.41237355916255e-05, "loss": 2.5555, "step": 33000 }, { "epoch": 2.59, "eval_loss": 2.8973400592803955, "eval_runtime": 217.7885, "eval_samples_per_second": 52.032, "eval_steps_per_second": 3.255, "step": 33000 }, { "epoch": 2.59, "learning_rate": 7.4122951462401e-05, "loss": 2.8174, "step": 33001 }, { "epoch": 2.59, "learning_rate": 7.412216733317651e-05, "loss": 2.8076, "step": 33002 }, { "epoch": 2.59, "learning_rate": 7.412138320395201e-05, "loss": 2.6194, "step": 33003 }, { "epoch": 2.59, "learning_rate": 7.412059907472752e-05, "loss": 2.6008, "step": 33004 }, { "epoch": 2.59, "learning_rate": 7.411981494550302e-05, "loss": 3.0185, "step": 33005 }, { "epoch": 2.59, "learning_rate": 7.411903081627852e-05, "loss": 2.7704, "step": 33006 }, { "epoch": 2.59, "learning_rate": 7.411824668705403e-05, "loss": 2.8808, "step": 33007 }, { "epoch": 2.59, "learning_rate": 7.411746255782953e-05, "loss": 2.647, "step": 33008 }, { "epoch": 2.59, "learning_rate": 7.411667842860504e-05, "loss": 3.0892, "step": 33009 }, { "epoch": 2.59, "learning_rate": 7.411589429938054e-05, "loss": 2.826, "step": 33010 }, { "epoch": 2.59, "learning_rate": 7.411511017015604e-05, "loss": 2.738, "step": 33011 }, { "epoch": 2.59, "learning_rate": 7.411432604093155e-05, "loss": 2.7649, "step": 33012 }, { "epoch": 2.59, "learning_rate": 7.411354191170705e-05, "loss": 2.691, "step": 33013 }, { "epoch": 2.59, "learning_rate": 7.411275778248255e-05, "loss": 2.7806, "step": 33014 }, { "epoch": 2.59, "learning_rate": 7.411197365325806e-05, "loss": 2.7867, "step": 33015 }, { "epoch": 2.59, "learning_rate": 7.411118952403356e-05, "loss": 2.945, "step": 33016 }, { "epoch": 2.59, "learning_rate": 7.411040539480906e-05, "loss": 2.764, "step": 33017 }, { "epoch": 2.59, "learning_rate": 7.410962126558457e-05, "loss": 2.8223, "step": 33018 }, { "epoch": 2.59, "learning_rate": 7.410883713636007e-05, "loss": 2.8906, "step": 33019 }, { "epoch": 2.59, "learning_rate": 7.410805300713558e-05, "loss": 2.7873, "step": 33020 }, { "epoch": 2.59, "learning_rate": 7.410726887791108e-05, "loss": 2.9509, "step": 33021 }, { "epoch": 2.59, "learning_rate": 7.410648474868658e-05, "loss": 2.9079, "step": 33022 }, { "epoch": 2.59, "learning_rate": 7.410570061946209e-05, "loss": 3.1185, "step": 33023 }, { "epoch": 2.59, "learning_rate": 7.410491649023759e-05, "loss": 2.7011, "step": 33024 }, { "epoch": 2.59, "learning_rate": 7.410413236101311e-05, "loss": 3.0198, "step": 33025 }, { "epoch": 2.59, "learning_rate": 7.410334823178861e-05, "loss": 2.8617, "step": 33026 }, { "epoch": 2.59, "learning_rate": 7.410256410256412e-05, "loss": 2.91, "step": 33027 }, { "epoch": 2.59, "learning_rate": 7.410177997333962e-05, "loss": 2.9106, "step": 33028 }, { "epoch": 2.59, "learning_rate": 7.410099584411512e-05, "loss": 2.9171, "step": 33029 }, { "epoch": 2.59, "learning_rate": 7.410021171489063e-05, "loss": 2.7095, "step": 33030 }, { "epoch": 2.59, "learning_rate": 7.409942758566613e-05, "loss": 2.8285, "step": 33031 }, { "epoch": 2.59, "learning_rate": 7.409864345644163e-05, "loss": 2.7405, "step": 33032 }, { "epoch": 2.59, "learning_rate": 7.409785932721714e-05, "loss": 2.6018, "step": 33033 }, { "epoch": 2.59, "learning_rate": 7.409707519799264e-05, "loss": 2.5055, "step": 33034 }, { "epoch": 2.59, "learning_rate": 7.409629106876813e-05, "loss": 2.8394, "step": 33035 }, { "epoch": 2.59, "learning_rate": 7.409550693954363e-05, "loss": 2.6312, "step": 33036 }, { "epoch": 2.59, "learning_rate": 7.409472281031914e-05, "loss": 2.846, "step": 33037 }, { "epoch": 2.59, "learning_rate": 7.409393868109464e-05, "loss": 2.8856, "step": 33038 }, { "epoch": 2.59, "learning_rate": 7.409315455187015e-05, "loss": 2.6208, "step": 33039 }, { "epoch": 2.59, "learning_rate": 7.409237042264565e-05, "loss": 2.7532, "step": 33040 }, { "epoch": 2.59, "learning_rate": 7.409158629342115e-05, "loss": 2.974, "step": 33041 }, { "epoch": 2.59, "learning_rate": 7.409080216419666e-05, "loss": 2.973, "step": 33042 }, { "epoch": 2.59, "learning_rate": 7.409001803497216e-05, "loss": 2.6667, "step": 33043 }, { "epoch": 2.59, "learning_rate": 7.408923390574766e-05, "loss": 3.0915, "step": 33044 }, { "epoch": 2.59, "learning_rate": 7.408844977652317e-05, "loss": 3.0175, "step": 33045 }, { "epoch": 2.59, "learning_rate": 7.408766564729867e-05, "loss": 2.943, "step": 33046 }, { "epoch": 2.59, "learning_rate": 7.408688151807418e-05, "loss": 2.6579, "step": 33047 }, { "epoch": 2.59, "learning_rate": 7.408609738884968e-05, "loss": 2.7768, "step": 33048 }, { "epoch": 2.59, "learning_rate": 7.408531325962518e-05, "loss": 2.8606, "step": 33049 }, { "epoch": 2.59, "learning_rate": 7.408452913040069e-05, "loss": 3.0567, "step": 33050 }, { "epoch": 2.59, "learning_rate": 7.40837450011762e-05, "loss": 2.7858, "step": 33051 }, { "epoch": 2.59, "learning_rate": 7.408296087195171e-05, "loss": 2.802, "step": 33052 }, { "epoch": 2.59, "learning_rate": 7.408217674272721e-05, "loss": 2.792, "step": 33053 }, { "epoch": 2.59, "learning_rate": 7.408139261350271e-05, "loss": 2.6595, "step": 33054 }, { "epoch": 2.59, "learning_rate": 7.408060848427822e-05, "loss": 2.7273, "step": 33055 }, { "epoch": 2.59, "learning_rate": 7.407982435505372e-05, "loss": 2.8253, "step": 33056 }, { "epoch": 2.59, "learning_rate": 7.407904022582923e-05, "loss": 2.8299, "step": 33057 }, { "epoch": 2.59, "learning_rate": 7.407825609660473e-05, "loss": 2.8134, "step": 33058 }, { "epoch": 2.59, "learning_rate": 7.407747196738023e-05, "loss": 2.8986, "step": 33059 }, { "epoch": 2.59, "learning_rate": 7.407668783815574e-05, "loss": 2.8578, "step": 33060 }, { "epoch": 2.59, "learning_rate": 7.407590370893124e-05, "loss": 2.7495, "step": 33061 }, { "epoch": 2.59, "learning_rate": 7.407511957970674e-05, "loss": 2.772, "step": 33062 }, { "epoch": 2.59, "learning_rate": 7.407433545048225e-05, "loss": 2.9453, "step": 33063 }, { "epoch": 2.59, "learning_rate": 7.407355132125775e-05, "loss": 2.8619, "step": 33064 }, { "epoch": 2.59, "learning_rate": 7.407276719203325e-05, "loss": 2.8084, "step": 33065 }, { "epoch": 2.59, "learning_rate": 7.407198306280876e-05, "loss": 2.6157, "step": 33066 }, { "epoch": 2.59, "learning_rate": 7.407119893358426e-05, "loss": 2.9734, "step": 33067 }, { "epoch": 2.59, "learning_rate": 7.407041480435977e-05, "loss": 2.7433, "step": 33068 }, { "epoch": 2.59, "learning_rate": 7.406963067513527e-05, "loss": 2.9614, "step": 33069 }, { "epoch": 2.59, "learning_rate": 7.406884654591077e-05, "loss": 2.9466, "step": 33070 }, { "epoch": 2.59, "learning_rate": 7.406806241668626e-05, "loss": 2.8592, "step": 33071 }, { "epoch": 2.59, "learning_rate": 7.406727828746177e-05, "loss": 2.7575, "step": 33072 }, { "epoch": 2.59, "learning_rate": 7.406649415823727e-05, "loss": 2.6214, "step": 33073 }, { "epoch": 2.59, "learning_rate": 7.406571002901277e-05, "loss": 2.7118, "step": 33074 }, { "epoch": 2.59, "learning_rate": 7.406492589978828e-05, "loss": 2.7495, "step": 33075 }, { "epoch": 2.59, "learning_rate": 7.406414177056378e-05, "loss": 3.0156, "step": 33076 }, { "epoch": 2.59, "learning_rate": 7.40633576413393e-05, "loss": 2.6803, "step": 33077 }, { "epoch": 2.59, "learning_rate": 7.40625735121148e-05, "loss": 2.669, "step": 33078 }, { "epoch": 2.59, "learning_rate": 7.40617893828903e-05, "loss": 2.8274, "step": 33079 }, { "epoch": 2.59, "learning_rate": 7.406100525366581e-05, "loss": 2.6314, "step": 33080 }, { "epoch": 2.59, "learning_rate": 7.406022112444131e-05, "loss": 2.5755, "step": 33081 }, { "epoch": 2.59, "learning_rate": 7.405943699521682e-05, "loss": 2.8129, "step": 33082 }, { "epoch": 2.59, "learning_rate": 7.405865286599232e-05, "loss": 3.0035, "step": 33083 }, { "epoch": 2.59, "learning_rate": 7.405786873676782e-05, "loss": 2.8963, "step": 33084 }, { "epoch": 2.59, "learning_rate": 7.405708460754333e-05, "loss": 2.6713, "step": 33085 }, { "epoch": 2.59, "learning_rate": 7.405630047831883e-05, "loss": 3.0087, "step": 33086 }, { "epoch": 2.59, "learning_rate": 7.405551634909434e-05, "loss": 2.7533, "step": 33087 }, { "epoch": 2.59, "learning_rate": 7.405473221986984e-05, "loss": 2.8648, "step": 33088 }, { "epoch": 2.59, "learning_rate": 7.405394809064534e-05, "loss": 2.8531, "step": 33089 }, { "epoch": 2.59, "learning_rate": 7.405316396142085e-05, "loss": 2.8016, "step": 33090 }, { "epoch": 2.59, "learning_rate": 7.405237983219635e-05, "loss": 2.7416, "step": 33091 }, { "epoch": 2.59, "learning_rate": 7.405159570297185e-05, "loss": 2.6212, "step": 33092 }, { "epoch": 2.59, "learning_rate": 7.405081157374736e-05, "loss": 2.8103, "step": 33093 }, { "epoch": 2.59, "learning_rate": 7.405002744452286e-05, "loss": 2.7895, "step": 33094 }, { "epoch": 2.6, "learning_rate": 7.404924331529837e-05, "loss": 2.8341, "step": 33095 }, { "epoch": 2.6, "learning_rate": 7.404845918607387e-05, "loss": 2.7533, "step": 33096 }, { "epoch": 2.6, "learning_rate": 7.404767505684937e-05, "loss": 2.7867, "step": 33097 }, { "epoch": 2.6, "learning_rate": 7.404689092762488e-05, "loss": 2.7356, "step": 33098 }, { "epoch": 2.6, "learning_rate": 7.404610679840038e-05, "loss": 2.8509, "step": 33099 }, { "epoch": 2.6, "learning_rate": 7.404532266917588e-05, "loss": 2.6258, "step": 33100 }, { "epoch": 2.6, "learning_rate": 7.404453853995139e-05, "loss": 2.7644, "step": 33101 }, { "epoch": 2.6, "learning_rate": 7.404375441072689e-05, "loss": 2.6997, "step": 33102 }, { "epoch": 2.6, "learning_rate": 7.40429702815024e-05, "loss": 2.6335, "step": 33103 }, { "epoch": 2.6, "learning_rate": 7.40421861522779e-05, "loss": 2.7332, "step": 33104 }, { "epoch": 2.6, "learning_rate": 7.40414020230534e-05, "loss": 2.701, "step": 33105 }, { "epoch": 2.6, "learning_rate": 7.40406178938289e-05, "loss": 3.0323, "step": 33106 }, { "epoch": 2.6, "learning_rate": 7.403983376460441e-05, "loss": 2.7834, "step": 33107 }, { "epoch": 2.6, "learning_rate": 7.403904963537991e-05, "loss": 2.8579, "step": 33108 }, { "epoch": 2.6, "learning_rate": 7.403826550615542e-05, "loss": 2.9032, "step": 33109 }, { "epoch": 2.6, "learning_rate": 7.403748137693092e-05, "loss": 2.8918, "step": 33110 }, { "epoch": 2.6, "learning_rate": 7.403669724770642e-05, "loss": 2.6529, "step": 33111 }, { "epoch": 2.6, "learning_rate": 7.403591311848193e-05, "loss": 2.7039, "step": 33112 }, { "epoch": 2.6, "learning_rate": 7.403512898925743e-05, "loss": 2.856, "step": 33113 }, { "epoch": 2.6, "learning_rate": 7.403434486003294e-05, "loss": 2.829, "step": 33114 }, { "epoch": 2.6, "learning_rate": 7.403356073080844e-05, "loss": 2.7567, "step": 33115 }, { "epoch": 2.6, "learning_rate": 7.403277660158394e-05, "loss": 2.7189, "step": 33116 }, { "epoch": 2.6, "learning_rate": 7.403199247235945e-05, "loss": 3.2142, "step": 33117 }, { "epoch": 2.6, "learning_rate": 7.403120834313495e-05, "loss": 2.7512, "step": 33118 }, { "epoch": 2.6, "learning_rate": 7.403042421391045e-05, "loss": 2.8914, "step": 33119 }, { "epoch": 2.6, "learning_rate": 7.402964008468596e-05, "loss": 2.841, "step": 33120 }, { "epoch": 2.6, "learning_rate": 7.402885595546146e-05, "loss": 2.7922, "step": 33121 }, { "epoch": 2.6, "learning_rate": 7.402807182623696e-05, "loss": 2.7037, "step": 33122 }, { "epoch": 2.6, "learning_rate": 7.402728769701247e-05, "loss": 2.9799, "step": 33123 }, { "epoch": 2.6, "learning_rate": 7.402650356778797e-05, "loss": 2.9668, "step": 33124 }, { "epoch": 2.6, "learning_rate": 7.402571943856348e-05, "loss": 2.7246, "step": 33125 }, { "epoch": 2.6, "learning_rate": 7.402493530933898e-05, "loss": 3.1288, "step": 33126 }, { "epoch": 2.6, "learning_rate": 7.402415118011448e-05, "loss": 2.7751, "step": 33127 }, { "epoch": 2.6, "learning_rate": 7.402336705088999e-05, "loss": 2.6359, "step": 33128 }, { "epoch": 2.6, "learning_rate": 7.402258292166549e-05, "loss": 2.9208, "step": 33129 }, { "epoch": 2.6, "learning_rate": 7.4021798792441e-05, "loss": 2.9152, "step": 33130 }, { "epoch": 2.6, "learning_rate": 7.40210146632165e-05, "loss": 2.7649, "step": 33131 }, { "epoch": 2.6, "learning_rate": 7.4020230533992e-05, "loss": 2.7938, "step": 33132 }, { "epoch": 2.6, "learning_rate": 7.401944640476752e-05, "loss": 2.7754, "step": 33133 }, { "epoch": 2.6, "learning_rate": 7.401866227554302e-05, "loss": 2.9147, "step": 33134 }, { "epoch": 2.6, "learning_rate": 7.401787814631853e-05, "loss": 2.7862, "step": 33135 }, { "epoch": 2.6, "learning_rate": 7.401709401709403e-05, "loss": 2.5486, "step": 33136 }, { "epoch": 2.6, "learning_rate": 7.401630988786953e-05, "loss": 2.7957, "step": 33137 }, { "epoch": 2.6, "learning_rate": 7.401552575864504e-05, "loss": 2.7995, "step": 33138 }, { "epoch": 2.6, "learning_rate": 7.401474162942054e-05, "loss": 2.7223, "step": 33139 }, { "epoch": 2.6, "learning_rate": 7.401395750019604e-05, "loss": 2.6734, "step": 33140 }, { "epoch": 2.6, "learning_rate": 7.401317337097155e-05, "loss": 2.846, "step": 33141 }, { "epoch": 2.6, "learning_rate": 7.401238924174705e-05, "loss": 2.8663, "step": 33142 }, { "epoch": 2.6, "learning_rate": 7.401160511252254e-05, "loss": 2.6936, "step": 33143 }, { "epoch": 2.6, "learning_rate": 7.401082098329805e-05, "loss": 2.6576, "step": 33144 }, { "epoch": 2.6, "learning_rate": 7.401003685407355e-05, "loss": 2.9997, "step": 33145 }, { "epoch": 2.6, "learning_rate": 7.400925272484905e-05, "loss": 2.8337, "step": 33146 }, { "epoch": 2.6, "learning_rate": 7.400846859562456e-05, "loss": 2.8544, "step": 33147 }, { "epoch": 2.6, "learning_rate": 7.400768446640006e-05, "loss": 2.8276, "step": 33148 }, { "epoch": 2.6, "learning_rate": 7.400690033717556e-05, "loss": 2.7622, "step": 33149 }, { "epoch": 2.6, "learning_rate": 7.400611620795107e-05, "loss": 2.7456, "step": 33150 }, { "epoch": 2.6, "learning_rate": 7.400533207872657e-05, "loss": 2.6298, "step": 33151 }, { "epoch": 2.6, "learning_rate": 7.400454794950207e-05, "loss": 2.7551, "step": 33152 }, { "epoch": 2.6, "learning_rate": 7.400376382027758e-05, "loss": 2.4387, "step": 33153 }, { "epoch": 2.6, "learning_rate": 7.400297969105308e-05, "loss": 2.5818, "step": 33154 }, { "epoch": 2.6, "learning_rate": 7.400219556182859e-05, "loss": 2.9181, "step": 33155 }, { "epoch": 2.6, "learning_rate": 7.400141143260409e-05, "loss": 2.8754, "step": 33156 }, { "epoch": 2.6, "learning_rate": 7.40006273033796e-05, "loss": 2.6672, "step": 33157 }, { "epoch": 2.6, "learning_rate": 7.39998431741551e-05, "loss": 2.8367, "step": 33158 }, { "epoch": 2.6, "learning_rate": 7.399905904493061e-05, "loss": 2.7698, "step": 33159 }, { "epoch": 2.6, "learning_rate": 7.399827491570612e-05, "loss": 2.7847, "step": 33160 }, { "epoch": 2.6, "learning_rate": 7.399749078648162e-05, "loss": 2.7983, "step": 33161 }, { "epoch": 2.6, "learning_rate": 7.399670665725713e-05, "loss": 2.9124, "step": 33162 }, { "epoch": 2.6, "learning_rate": 7.399592252803263e-05, "loss": 2.7972, "step": 33163 }, { "epoch": 2.6, "learning_rate": 7.399513839880813e-05, "loss": 2.7556, "step": 33164 }, { "epoch": 2.6, "learning_rate": 7.399435426958364e-05, "loss": 2.7739, "step": 33165 }, { "epoch": 2.6, "learning_rate": 7.399357014035914e-05, "loss": 2.8192, "step": 33166 }, { "epoch": 2.6, "learning_rate": 7.399278601113464e-05, "loss": 2.8296, "step": 33167 }, { "epoch": 2.6, "learning_rate": 7.399200188191015e-05, "loss": 2.7491, "step": 33168 }, { "epoch": 2.6, "learning_rate": 7.399121775268565e-05, "loss": 2.7655, "step": 33169 }, { "epoch": 2.6, "learning_rate": 7.399043362346115e-05, "loss": 2.9234, "step": 33170 }, { "epoch": 2.6, "learning_rate": 7.398964949423666e-05, "loss": 2.8414, "step": 33171 }, { "epoch": 2.6, "learning_rate": 7.398886536501216e-05, "loss": 2.7201, "step": 33172 }, { "epoch": 2.6, "learning_rate": 7.398808123578767e-05, "loss": 2.9791, "step": 33173 }, { "epoch": 2.6, "learning_rate": 7.398729710656317e-05, "loss": 3.05, "step": 33174 }, { "epoch": 2.6, "learning_rate": 7.398651297733867e-05, "loss": 2.5758, "step": 33175 }, { "epoch": 2.6, "learning_rate": 7.398572884811418e-05, "loss": 2.9669, "step": 33176 }, { "epoch": 2.6, "learning_rate": 7.398494471888968e-05, "loss": 2.7704, "step": 33177 }, { "epoch": 2.6, "learning_rate": 7.398416058966518e-05, "loss": 2.7857, "step": 33178 }, { "epoch": 2.6, "learning_rate": 7.398337646044067e-05, "loss": 2.6242, "step": 33179 }, { "epoch": 2.6, "learning_rate": 7.398259233121618e-05, "loss": 2.7397, "step": 33180 }, { "epoch": 2.6, "learning_rate": 7.398180820199168e-05, "loss": 2.6329, "step": 33181 }, { "epoch": 2.6, "learning_rate": 7.398102407276719e-05, "loss": 2.8314, "step": 33182 }, { "epoch": 2.6, "learning_rate": 7.398023994354269e-05, "loss": 2.7209, "step": 33183 }, { "epoch": 2.6, "learning_rate": 7.397945581431819e-05, "loss": 2.9934, "step": 33184 }, { "epoch": 2.6, "learning_rate": 7.397867168509371e-05, "loss": 2.8035, "step": 33185 }, { "epoch": 2.6, "learning_rate": 7.397788755586921e-05, "loss": 2.7611, "step": 33186 }, { "epoch": 2.6, "learning_rate": 7.397710342664472e-05, "loss": 2.9389, "step": 33187 }, { "epoch": 2.6, "learning_rate": 7.397631929742022e-05, "loss": 2.7843, "step": 33188 }, { "epoch": 2.6, "learning_rate": 7.397553516819572e-05, "loss": 2.9218, "step": 33189 }, { "epoch": 2.6, "learning_rate": 7.397475103897123e-05, "loss": 2.7209, "step": 33190 }, { "epoch": 2.6, "learning_rate": 7.397396690974673e-05, "loss": 2.626, "step": 33191 }, { "epoch": 2.6, "learning_rate": 7.397318278052224e-05, "loss": 2.7044, "step": 33192 }, { "epoch": 2.6, "learning_rate": 7.397239865129774e-05, "loss": 2.7436, "step": 33193 }, { "epoch": 2.6, "learning_rate": 7.397161452207324e-05, "loss": 2.7491, "step": 33194 }, { "epoch": 2.6, "learning_rate": 7.397083039284875e-05, "loss": 2.9146, "step": 33195 }, { "epoch": 2.6, "learning_rate": 7.397004626362425e-05, "loss": 2.7505, "step": 33196 }, { "epoch": 2.6, "learning_rate": 7.396926213439975e-05, "loss": 2.9224, "step": 33197 }, { "epoch": 2.6, "learning_rate": 7.396847800517526e-05, "loss": 2.9317, "step": 33198 }, { "epoch": 2.6, "learning_rate": 7.396769387595076e-05, "loss": 2.7201, "step": 33199 }, { "epoch": 2.6, "learning_rate": 7.396690974672626e-05, "loss": 2.933, "step": 33200 }, { "epoch": 2.6, "learning_rate": 7.396612561750177e-05, "loss": 2.6388, "step": 33201 }, { "epoch": 2.6, "learning_rate": 7.396534148827727e-05, "loss": 2.7978, "step": 33202 }, { "epoch": 2.6, "learning_rate": 7.396455735905278e-05, "loss": 3.1322, "step": 33203 }, { "epoch": 2.6, "learning_rate": 7.396377322982828e-05, "loss": 2.8586, "step": 33204 }, { "epoch": 2.6, "learning_rate": 7.396298910060378e-05, "loss": 2.8107, "step": 33205 }, { "epoch": 2.6, "learning_rate": 7.396220497137929e-05, "loss": 2.7085, "step": 33206 }, { "epoch": 2.6, "learning_rate": 7.396142084215479e-05, "loss": 2.6921, "step": 33207 }, { "epoch": 2.6, "learning_rate": 7.39606367129303e-05, "loss": 2.7829, "step": 33208 }, { "epoch": 2.6, "learning_rate": 7.39598525837058e-05, "loss": 2.8194, "step": 33209 }, { "epoch": 2.6, "learning_rate": 7.39590684544813e-05, "loss": 2.8594, "step": 33210 }, { "epoch": 2.6, "learning_rate": 7.39582843252568e-05, "loss": 2.7304, "step": 33211 }, { "epoch": 2.6, "learning_rate": 7.395750019603231e-05, "loss": 2.7626, "step": 33212 }, { "epoch": 2.6, "learning_rate": 7.395671606680781e-05, "loss": 2.7595, "step": 33213 }, { "epoch": 2.6, "learning_rate": 7.395593193758332e-05, "loss": 2.7939, "step": 33214 }, { "epoch": 2.6, "learning_rate": 7.395514780835882e-05, "loss": 2.9012, "step": 33215 }, { "epoch": 2.6, "learning_rate": 7.395436367913432e-05, "loss": 2.7514, "step": 33216 }, { "epoch": 2.6, "learning_rate": 7.395357954990983e-05, "loss": 2.8837, "step": 33217 }, { "epoch": 2.6, "learning_rate": 7.395279542068533e-05, "loss": 3.0925, "step": 33218 }, { "epoch": 2.6, "learning_rate": 7.395201129146083e-05, "loss": 2.7859, "step": 33219 }, { "epoch": 2.6, "learning_rate": 7.395122716223634e-05, "loss": 3.059, "step": 33220 }, { "epoch": 2.6, "learning_rate": 7.395044303301184e-05, "loss": 2.6613, "step": 33221 }, { "epoch": 2.61, "learning_rate": 7.394965890378735e-05, "loss": 2.9144, "step": 33222 }, { "epoch": 2.61, "learning_rate": 7.394887477456285e-05, "loss": 2.7692, "step": 33223 }, { "epoch": 2.61, "learning_rate": 7.394809064533835e-05, "loss": 2.6322, "step": 33224 }, { "epoch": 2.61, "learning_rate": 7.394730651611386e-05, "loss": 2.7941, "step": 33225 }, { "epoch": 2.61, "learning_rate": 7.394652238688936e-05, "loss": 2.7171, "step": 33226 }, { "epoch": 2.61, "learning_rate": 7.394573825766486e-05, "loss": 2.8859, "step": 33227 }, { "epoch": 2.61, "learning_rate": 7.394495412844037e-05, "loss": 2.7921, "step": 33228 }, { "epoch": 2.61, "learning_rate": 7.394416999921587e-05, "loss": 2.7188, "step": 33229 }, { "epoch": 2.61, "learning_rate": 7.394338586999138e-05, "loss": 2.9908, "step": 33230 }, { "epoch": 2.61, "learning_rate": 7.394260174076688e-05, "loss": 2.8316, "step": 33231 }, { "epoch": 2.61, "learning_rate": 7.394181761154238e-05, "loss": 2.8762, "step": 33232 }, { "epoch": 2.61, "learning_rate": 7.394103348231789e-05, "loss": 2.9536, "step": 33233 }, { "epoch": 2.61, "learning_rate": 7.394024935309339e-05, "loss": 2.6671, "step": 33234 }, { "epoch": 2.61, "learning_rate": 7.39394652238689e-05, "loss": 2.7024, "step": 33235 }, { "epoch": 2.61, "learning_rate": 7.39386810946444e-05, "loss": 2.8237, "step": 33236 }, { "epoch": 2.61, "learning_rate": 7.39378969654199e-05, "loss": 3.0251, "step": 33237 }, { "epoch": 2.61, "learning_rate": 7.39371128361954e-05, "loss": 3.0314, "step": 33238 }, { "epoch": 2.61, "learning_rate": 7.393632870697091e-05, "loss": 2.9126, "step": 33239 }, { "epoch": 2.61, "learning_rate": 7.393554457774641e-05, "loss": 2.7626, "step": 33240 }, { "epoch": 2.61, "learning_rate": 7.393476044852192e-05, "loss": 3.0618, "step": 33241 }, { "epoch": 2.61, "learning_rate": 7.393397631929743e-05, "loss": 2.7298, "step": 33242 }, { "epoch": 2.61, "learning_rate": 7.393319219007294e-05, "loss": 3.3409, "step": 33243 }, { "epoch": 2.61, "learning_rate": 7.393240806084844e-05, "loss": 3.0052, "step": 33244 }, { "epoch": 2.61, "learning_rate": 7.393162393162394e-05, "loss": 2.9153, "step": 33245 }, { "epoch": 2.61, "learning_rate": 7.393083980239945e-05, "loss": 3.023, "step": 33246 }, { "epoch": 2.61, "learning_rate": 7.393005567317495e-05, "loss": 2.712, "step": 33247 }, { "epoch": 2.61, "learning_rate": 7.392927154395046e-05, "loss": 2.6767, "step": 33248 }, { "epoch": 2.61, "learning_rate": 7.392848741472596e-05, "loss": 2.9817, "step": 33249 }, { "epoch": 2.61, "learning_rate": 7.392770328550146e-05, "loss": 2.5346, "step": 33250 }, { "epoch": 2.61, "learning_rate": 7.392691915627695e-05, "loss": 3.1546, "step": 33251 }, { "epoch": 2.61, "learning_rate": 7.392613502705246e-05, "loss": 2.7653, "step": 33252 }, { "epoch": 2.61, "learning_rate": 7.392535089782796e-05, "loss": 2.7675, "step": 33253 }, { "epoch": 2.61, "learning_rate": 7.392456676860346e-05, "loss": 2.6258, "step": 33254 }, { "epoch": 2.61, "learning_rate": 7.392378263937897e-05, "loss": 2.7771, "step": 33255 }, { "epoch": 2.61, "learning_rate": 7.392299851015447e-05, "loss": 2.8389, "step": 33256 }, { "epoch": 2.61, "learning_rate": 7.392221438092997e-05, "loss": 2.8312, "step": 33257 }, { "epoch": 2.61, "learning_rate": 7.392143025170548e-05, "loss": 2.8395, "step": 33258 }, { "epoch": 2.61, "learning_rate": 7.392064612248098e-05, "loss": 2.7479, "step": 33259 }, { "epoch": 2.61, "learning_rate": 7.391986199325649e-05, "loss": 2.8921, "step": 33260 }, { "epoch": 2.61, "learning_rate": 7.391907786403199e-05, "loss": 2.8638, "step": 33261 }, { "epoch": 2.61, "learning_rate": 7.391829373480749e-05, "loss": 2.8806, "step": 33262 }, { "epoch": 2.61, "learning_rate": 7.3917509605583e-05, "loss": 2.6789, "step": 33263 }, { "epoch": 2.61, "learning_rate": 7.39167254763585e-05, "loss": 2.9608, "step": 33264 }, { "epoch": 2.61, "learning_rate": 7.3915941347134e-05, "loss": 2.7336, "step": 33265 }, { "epoch": 2.61, "learning_rate": 7.391515721790951e-05, "loss": 2.6745, "step": 33266 }, { "epoch": 2.61, "learning_rate": 7.391437308868502e-05, "loss": 2.8581, "step": 33267 }, { "epoch": 2.61, "learning_rate": 7.391358895946053e-05, "loss": 2.5133, "step": 33268 }, { "epoch": 2.61, "learning_rate": 7.391280483023603e-05, "loss": 2.8074, "step": 33269 }, { "epoch": 2.61, "learning_rate": 7.391202070101154e-05, "loss": 2.8849, "step": 33270 }, { "epoch": 2.61, "learning_rate": 7.391123657178704e-05, "loss": 3.0841, "step": 33271 }, { "epoch": 2.61, "learning_rate": 7.391045244256254e-05, "loss": 2.9143, "step": 33272 }, { "epoch": 2.61, "learning_rate": 7.390966831333805e-05, "loss": 2.7326, "step": 33273 }, { "epoch": 2.61, "learning_rate": 7.390888418411355e-05, "loss": 2.8542, "step": 33274 }, { "epoch": 2.61, "learning_rate": 7.390810005488905e-05, "loss": 2.7429, "step": 33275 }, { "epoch": 2.61, "learning_rate": 7.390731592566456e-05, "loss": 2.828, "step": 33276 }, { "epoch": 2.61, "learning_rate": 7.390653179644006e-05, "loss": 2.6658, "step": 33277 }, { "epoch": 2.61, "learning_rate": 7.390574766721557e-05, "loss": 2.837, "step": 33278 }, { "epoch": 2.61, "learning_rate": 7.390496353799107e-05, "loss": 2.9411, "step": 33279 }, { "epoch": 2.61, "learning_rate": 7.390417940876657e-05, "loss": 2.9015, "step": 33280 }, { "epoch": 2.61, "learning_rate": 7.390339527954208e-05, "loss": 2.9532, "step": 33281 }, { "epoch": 2.61, "learning_rate": 7.390261115031758e-05, "loss": 2.6833, "step": 33282 }, { "epoch": 2.61, "learning_rate": 7.390182702109308e-05, "loss": 3.1396, "step": 33283 }, { "epoch": 2.61, "learning_rate": 7.390104289186859e-05, "loss": 2.8748, "step": 33284 }, { "epoch": 2.61, "learning_rate": 7.390025876264409e-05, "loss": 3.1691, "step": 33285 }, { "epoch": 2.61, "learning_rate": 7.38994746334196e-05, "loss": 2.881, "step": 33286 }, { "epoch": 2.61, "learning_rate": 7.389869050419508e-05, "loss": 2.7482, "step": 33287 }, { "epoch": 2.61, "learning_rate": 7.389790637497059e-05, "loss": 2.8423, "step": 33288 }, { "epoch": 2.61, "learning_rate": 7.389712224574609e-05, "loss": 2.7172, "step": 33289 }, { "epoch": 2.61, "learning_rate": 7.38963381165216e-05, "loss": 3.0003, "step": 33290 }, { "epoch": 2.61, "learning_rate": 7.38955539872971e-05, "loss": 2.7099, "step": 33291 }, { "epoch": 2.61, "learning_rate": 7.38947698580726e-05, "loss": 2.8841, "step": 33292 }, { "epoch": 2.61, "learning_rate": 7.389398572884812e-05, "loss": 2.8134, "step": 33293 }, { "epoch": 2.61, "learning_rate": 7.389320159962362e-05, "loss": 2.6559, "step": 33294 }, { "epoch": 2.61, "learning_rate": 7.389241747039913e-05, "loss": 2.7771, "step": 33295 }, { "epoch": 2.61, "learning_rate": 7.389163334117463e-05, "loss": 2.8379, "step": 33296 }, { "epoch": 2.61, "learning_rate": 7.389084921195014e-05, "loss": 2.9101, "step": 33297 }, { "epoch": 2.61, "learning_rate": 7.389006508272564e-05, "loss": 2.8189, "step": 33298 }, { "epoch": 2.61, "learning_rate": 7.388928095350114e-05, "loss": 2.8245, "step": 33299 }, { "epoch": 2.61, "learning_rate": 7.388849682427665e-05, "loss": 2.8475, "step": 33300 }, { "epoch": 2.61, "learning_rate": 7.388771269505215e-05, "loss": 2.6514, "step": 33301 }, { "epoch": 2.61, "learning_rate": 7.388692856582765e-05, "loss": 2.8864, "step": 33302 }, { "epoch": 2.61, "learning_rate": 7.388614443660316e-05, "loss": 2.6805, "step": 33303 }, { "epoch": 2.61, "learning_rate": 7.388536030737866e-05, "loss": 2.836, "step": 33304 }, { "epoch": 2.61, "learning_rate": 7.388457617815416e-05, "loss": 2.727, "step": 33305 }, { "epoch": 2.61, "learning_rate": 7.388379204892967e-05, "loss": 2.7776, "step": 33306 }, { "epoch": 2.61, "learning_rate": 7.388300791970517e-05, "loss": 3.0854, "step": 33307 }, { "epoch": 2.61, "learning_rate": 7.388222379048068e-05, "loss": 2.8868, "step": 33308 }, { "epoch": 2.61, "learning_rate": 7.388143966125618e-05, "loss": 2.796, "step": 33309 }, { "epoch": 2.61, "learning_rate": 7.388065553203168e-05, "loss": 2.7191, "step": 33310 }, { "epoch": 2.61, "learning_rate": 7.387987140280719e-05, "loss": 2.7606, "step": 33311 }, { "epoch": 2.61, "learning_rate": 7.387908727358269e-05, "loss": 2.9816, "step": 33312 }, { "epoch": 2.61, "learning_rate": 7.38783031443582e-05, "loss": 2.697, "step": 33313 }, { "epoch": 2.61, "learning_rate": 7.38775190151337e-05, "loss": 2.6912, "step": 33314 }, { "epoch": 2.61, "learning_rate": 7.38767348859092e-05, "loss": 2.6496, "step": 33315 }, { "epoch": 2.61, "learning_rate": 7.38759507566847e-05, "loss": 2.8844, "step": 33316 }, { "epoch": 2.61, "learning_rate": 7.387516662746021e-05, "loss": 2.8104, "step": 33317 }, { "epoch": 2.61, "learning_rate": 7.387438249823571e-05, "loss": 2.9181, "step": 33318 }, { "epoch": 2.61, "learning_rate": 7.387359836901122e-05, "loss": 2.8923, "step": 33319 }, { "epoch": 2.61, "learning_rate": 7.387281423978672e-05, "loss": 2.9025, "step": 33320 }, { "epoch": 2.61, "learning_rate": 7.387203011056222e-05, "loss": 2.9172, "step": 33321 }, { "epoch": 2.61, "learning_rate": 7.387124598133773e-05, "loss": 2.731, "step": 33322 }, { "epoch": 2.61, "learning_rate": 7.387046185211323e-05, "loss": 2.7361, "step": 33323 }, { "epoch": 2.61, "learning_rate": 7.386967772288873e-05, "loss": 2.7413, "step": 33324 }, { "epoch": 2.61, "learning_rate": 7.386889359366424e-05, "loss": 2.6947, "step": 33325 }, { "epoch": 2.61, "learning_rate": 7.386810946443974e-05, "loss": 2.8702, "step": 33326 }, { "epoch": 2.61, "learning_rate": 7.386732533521525e-05, "loss": 2.7058, "step": 33327 }, { "epoch": 2.61, "learning_rate": 7.386654120599075e-05, "loss": 2.8524, "step": 33328 }, { "epoch": 2.61, "learning_rate": 7.386575707676625e-05, "loss": 2.823, "step": 33329 }, { "epoch": 2.61, "learning_rate": 7.386497294754176e-05, "loss": 2.7811, "step": 33330 }, { "epoch": 2.61, "learning_rate": 7.386418881831726e-05, "loss": 2.5969, "step": 33331 }, { "epoch": 2.61, "learning_rate": 7.386340468909276e-05, "loss": 2.9576, "step": 33332 }, { "epoch": 2.61, "learning_rate": 7.386262055986827e-05, "loss": 2.8011, "step": 33333 }, { "epoch": 2.61, "learning_rate": 7.386183643064377e-05, "loss": 2.9203, "step": 33334 }, { "epoch": 2.61, "learning_rate": 7.386105230141928e-05, "loss": 3.0249, "step": 33335 }, { "epoch": 2.61, "learning_rate": 7.386026817219478e-05, "loss": 2.696, "step": 33336 }, { "epoch": 2.61, "learning_rate": 7.385948404297028e-05, "loss": 2.8, "step": 33337 }, { "epoch": 2.61, "learning_rate": 7.385869991374579e-05, "loss": 2.8685, "step": 33338 }, { "epoch": 2.61, "learning_rate": 7.385791578452129e-05, "loss": 2.8931, "step": 33339 }, { "epoch": 2.61, "learning_rate": 7.38571316552968e-05, "loss": 2.5893, "step": 33340 }, { "epoch": 2.61, "learning_rate": 7.38563475260723e-05, "loss": 3.0147, "step": 33341 }, { "epoch": 2.61, "learning_rate": 7.38555633968478e-05, "loss": 2.8983, "step": 33342 }, { "epoch": 2.61, "learning_rate": 7.38547792676233e-05, "loss": 2.6225, "step": 33343 }, { "epoch": 2.61, "learning_rate": 7.385399513839881e-05, "loss": 2.8823, "step": 33344 }, { "epoch": 2.61, "learning_rate": 7.385321100917431e-05, "loss": 2.7318, "step": 33345 }, { "epoch": 2.61, "learning_rate": 7.385242687994982e-05, "loss": 2.8086, "step": 33346 }, { "epoch": 2.61, "learning_rate": 7.385164275072532e-05, "loss": 2.9464, "step": 33347 }, { "epoch": 2.61, "learning_rate": 7.385085862150082e-05, "loss": 2.8054, "step": 33348 }, { "epoch": 2.61, "learning_rate": 7.385007449227633e-05, "loss": 2.6558, "step": 33349 }, { "epoch": 2.62, "learning_rate": 7.384929036305184e-05, "loss": 2.9099, "step": 33350 }, { "epoch": 2.62, "learning_rate": 7.384850623382735e-05, "loss": 2.7663, "step": 33351 }, { "epoch": 2.62, "learning_rate": 7.384772210460285e-05, "loss": 2.7469, "step": 33352 }, { "epoch": 2.62, "learning_rate": 7.384693797537835e-05, "loss": 2.9131, "step": 33353 }, { "epoch": 2.62, "learning_rate": 7.384615384615386e-05, "loss": 2.7858, "step": 33354 }, { "epoch": 2.62, "learning_rate": 7.384536971692936e-05, "loss": 2.8356, "step": 33355 }, { "epoch": 2.62, "learning_rate": 7.384458558770487e-05, "loss": 2.7779, "step": 33356 }, { "epoch": 2.62, "learning_rate": 7.384380145848037e-05, "loss": 2.761, "step": 33357 }, { "epoch": 2.62, "learning_rate": 7.384301732925587e-05, "loss": 2.7759, "step": 33358 }, { "epoch": 2.62, "learning_rate": 7.384223320003136e-05, "loss": 2.9833, "step": 33359 }, { "epoch": 2.62, "learning_rate": 7.384144907080687e-05, "loss": 2.8214, "step": 33360 }, { "epoch": 2.62, "learning_rate": 7.384066494158237e-05, "loss": 2.7906, "step": 33361 }, { "epoch": 2.62, "learning_rate": 7.383988081235787e-05, "loss": 2.7308, "step": 33362 }, { "epoch": 2.62, "learning_rate": 7.383909668313338e-05, "loss": 2.9875, "step": 33363 }, { "epoch": 2.62, "learning_rate": 7.383831255390888e-05, "loss": 2.7344, "step": 33364 }, { "epoch": 2.62, "learning_rate": 7.383752842468439e-05, "loss": 3.0459, "step": 33365 }, { "epoch": 2.62, "learning_rate": 7.383674429545989e-05, "loss": 2.8082, "step": 33366 }, { "epoch": 2.62, "learning_rate": 7.383596016623539e-05, "loss": 2.7912, "step": 33367 }, { "epoch": 2.62, "learning_rate": 7.38351760370109e-05, "loss": 2.8493, "step": 33368 }, { "epoch": 2.62, "learning_rate": 7.38343919077864e-05, "loss": 2.7883, "step": 33369 }, { "epoch": 2.62, "learning_rate": 7.38336077785619e-05, "loss": 2.8994, "step": 33370 }, { "epoch": 2.62, "learning_rate": 7.383282364933741e-05, "loss": 2.832, "step": 33371 }, { "epoch": 2.62, "learning_rate": 7.383203952011291e-05, "loss": 2.7396, "step": 33372 }, { "epoch": 2.62, "learning_rate": 7.383125539088841e-05, "loss": 2.679, "step": 33373 }, { "epoch": 2.62, "learning_rate": 7.383047126166392e-05, "loss": 2.6762, "step": 33374 }, { "epoch": 2.62, "learning_rate": 7.382968713243944e-05, "loss": 2.8624, "step": 33375 }, { "epoch": 2.62, "learning_rate": 7.382890300321494e-05, "loss": 2.8041, "step": 33376 }, { "epoch": 2.62, "learning_rate": 7.382811887399044e-05, "loss": 2.8515, "step": 33377 }, { "epoch": 2.62, "learning_rate": 7.382733474476595e-05, "loss": 2.5265, "step": 33378 }, { "epoch": 2.62, "learning_rate": 7.382655061554145e-05, "loss": 2.6836, "step": 33379 }, { "epoch": 2.62, "learning_rate": 7.382576648631695e-05, "loss": 3.0067, "step": 33380 }, { "epoch": 2.62, "learning_rate": 7.382498235709246e-05, "loss": 2.519, "step": 33381 }, { "epoch": 2.62, "learning_rate": 7.382419822786796e-05, "loss": 2.8105, "step": 33382 }, { "epoch": 2.62, "learning_rate": 7.382341409864347e-05, "loss": 2.9026, "step": 33383 }, { "epoch": 2.62, "learning_rate": 7.382262996941897e-05, "loss": 2.7452, "step": 33384 }, { "epoch": 2.62, "learning_rate": 7.382184584019447e-05, "loss": 2.8742, "step": 33385 }, { "epoch": 2.62, "learning_rate": 7.382106171096998e-05, "loss": 2.8449, "step": 33386 }, { "epoch": 2.62, "learning_rate": 7.382027758174548e-05, "loss": 2.6319, "step": 33387 }, { "epoch": 2.62, "learning_rate": 7.381949345252098e-05, "loss": 2.7171, "step": 33388 }, { "epoch": 2.62, "learning_rate": 7.381870932329649e-05, "loss": 2.7298, "step": 33389 }, { "epoch": 2.62, "learning_rate": 7.381792519407199e-05, "loss": 2.7982, "step": 33390 }, { "epoch": 2.62, "learning_rate": 7.38171410648475e-05, "loss": 2.6706, "step": 33391 }, { "epoch": 2.62, "learning_rate": 7.3816356935623e-05, "loss": 2.8988, "step": 33392 }, { "epoch": 2.62, "learning_rate": 7.38155728063985e-05, "loss": 2.9016, "step": 33393 }, { "epoch": 2.62, "learning_rate": 7.3814788677174e-05, "loss": 2.6687, "step": 33394 }, { "epoch": 2.62, "learning_rate": 7.38140045479495e-05, "loss": 2.6839, "step": 33395 }, { "epoch": 2.62, "learning_rate": 7.3813220418725e-05, "loss": 2.8055, "step": 33396 }, { "epoch": 2.62, "learning_rate": 7.38124362895005e-05, "loss": 2.8943, "step": 33397 }, { "epoch": 2.62, "learning_rate": 7.3811652160276e-05, "loss": 2.6943, "step": 33398 }, { "epoch": 2.62, "learning_rate": 7.381086803105151e-05, "loss": 2.8548, "step": 33399 }, { "epoch": 2.62, "learning_rate": 7.381008390182701e-05, "loss": 2.6522, "step": 33400 }, { "epoch": 2.62, "learning_rate": 7.380929977260253e-05, "loss": 2.7822, "step": 33401 }, { "epoch": 2.62, "learning_rate": 7.380851564337803e-05, "loss": 2.6816, "step": 33402 }, { "epoch": 2.62, "learning_rate": 7.380773151415354e-05, "loss": 2.8101, "step": 33403 }, { "epoch": 2.62, "learning_rate": 7.380694738492904e-05, "loss": 2.7815, "step": 33404 }, { "epoch": 2.62, "learning_rate": 7.380616325570455e-05, "loss": 2.6844, "step": 33405 }, { "epoch": 2.62, "learning_rate": 7.380537912648005e-05, "loss": 2.8471, "step": 33406 }, { "epoch": 2.62, "learning_rate": 7.380459499725555e-05, "loss": 2.8294, "step": 33407 }, { "epoch": 2.62, "learning_rate": 7.380381086803106e-05, "loss": 2.8428, "step": 33408 }, { "epoch": 2.62, "learning_rate": 7.380302673880656e-05, "loss": 2.7544, "step": 33409 }, { "epoch": 2.62, "learning_rate": 7.380224260958206e-05, "loss": 2.6898, "step": 33410 }, { "epoch": 2.62, "learning_rate": 7.380145848035757e-05, "loss": 2.8026, "step": 33411 }, { "epoch": 2.62, "learning_rate": 7.380067435113307e-05, "loss": 2.5219, "step": 33412 }, { "epoch": 2.62, "learning_rate": 7.379989022190858e-05, "loss": 2.6815, "step": 33413 }, { "epoch": 2.62, "learning_rate": 7.379910609268408e-05, "loss": 2.7593, "step": 33414 }, { "epoch": 2.62, "learning_rate": 7.379832196345958e-05, "loss": 2.6075, "step": 33415 }, { "epoch": 2.62, "learning_rate": 7.379753783423509e-05, "loss": 2.7609, "step": 33416 }, { "epoch": 2.62, "learning_rate": 7.379675370501059e-05, "loss": 2.8526, "step": 33417 }, { "epoch": 2.62, "learning_rate": 7.37959695757861e-05, "loss": 2.5697, "step": 33418 }, { "epoch": 2.62, "learning_rate": 7.37951854465616e-05, "loss": 2.9688, "step": 33419 }, { "epoch": 2.62, "learning_rate": 7.37944013173371e-05, "loss": 2.8788, "step": 33420 }, { "epoch": 2.62, "learning_rate": 7.37936171881126e-05, "loss": 2.9688, "step": 33421 }, { "epoch": 2.62, "learning_rate": 7.379283305888811e-05, "loss": 2.9102, "step": 33422 }, { "epoch": 2.62, "learning_rate": 7.379204892966361e-05, "loss": 2.6023, "step": 33423 }, { "epoch": 2.62, "learning_rate": 7.379126480043912e-05, "loss": 2.9039, "step": 33424 }, { "epoch": 2.62, "learning_rate": 7.379048067121462e-05, "loss": 2.8001, "step": 33425 }, { "epoch": 2.62, "learning_rate": 7.378969654199012e-05, "loss": 2.7809, "step": 33426 }, { "epoch": 2.62, "learning_rate": 7.378891241276563e-05, "loss": 2.5391, "step": 33427 }, { "epoch": 2.62, "learning_rate": 7.378812828354113e-05, "loss": 2.7909, "step": 33428 }, { "epoch": 2.62, "learning_rate": 7.378734415431663e-05, "loss": 3.0999, "step": 33429 }, { "epoch": 2.62, "learning_rate": 7.378656002509214e-05, "loss": 2.9509, "step": 33430 }, { "epoch": 2.62, "learning_rate": 7.378577589586764e-05, "loss": 2.8658, "step": 33431 }, { "epoch": 2.62, "learning_rate": 7.378499176664315e-05, "loss": 2.784, "step": 33432 }, { "epoch": 2.62, "learning_rate": 7.378420763741865e-05, "loss": 2.7921, "step": 33433 }, { "epoch": 2.62, "learning_rate": 7.378342350819415e-05, "loss": 2.6208, "step": 33434 }, { "epoch": 2.62, "learning_rate": 7.378263937896966e-05, "loss": 2.6752, "step": 33435 }, { "epoch": 2.62, "learning_rate": 7.378185524974516e-05, "loss": 2.7756, "step": 33436 }, { "epoch": 2.62, "learning_rate": 7.378107112052066e-05, "loss": 2.8069, "step": 33437 }, { "epoch": 2.62, "learning_rate": 7.378028699129617e-05, "loss": 2.6334, "step": 33438 }, { "epoch": 2.62, "learning_rate": 7.377950286207167e-05, "loss": 2.9455, "step": 33439 }, { "epoch": 2.62, "learning_rate": 7.377871873284717e-05, "loss": 2.7687, "step": 33440 }, { "epoch": 2.62, "learning_rate": 7.377793460362268e-05, "loss": 2.7753, "step": 33441 }, { "epoch": 2.62, "learning_rate": 7.377715047439818e-05, "loss": 2.8413, "step": 33442 }, { "epoch": 2.62, "learning_rate": 7.377636634517369e-05, "loss": 2.8374, "step": 33443 }, { "epoch": 2.62, "learning_rate": 7.377558221594919e-05, "loss": 2.8095, "step": 33444 }, { "epoch": 2.62, "learning_rate": 7.377479808672469e-05, "loss": 2.9137, "step": 33445 }, { "epoch": 2.62, "learning_rate": 7.37740139575002e-05, "loss": 2.8955, "step": 33446 }, { "epoch": 2.62, "learning_rate": 7.37732298282757e-05, "loss": 2.8456, "step": 33447 }, { "epoch": 2.62, "learning_rate": 7.37724456990512e-05, "loss": 2.8532, "step": 33448 }, { "epoch": 2.62, "learning_rate": 7.377166156982671e-05, "loss": 2.8258, "step": 33449 }, { "epoch": 2.62, "learning_rate": 7.377087744060221e-05, "loss": 2.788, "step": 33450 }, { "epoch": 2.62, "learning_rate": 7.377009331137772e-05, "loss": 2.8324, "step": 33451 }, { "epoch": 2.62, "learning_rate": 7.376930918215322e-05, "loss": 2.9426, "step": 33452 }, { "epoch": 2.62, "learning_rate": 7.376852505292872e-05, "loss": 2.8202, "step": 33453 }, { "epoch": 2.62, "learning_rate": 7.376774092370423e-05, "loss": 2.6685, "step": 33454 }, { "epoch": 2.62, "learning_rate": 7.376695679447973e-05, "loss": 2.8891, "step": 33455 }, { "epoch": 2.62, "learning_rate": 7.376617266525523e-05, "loss": 2.8533, "step": 33456 }, { "epoch": 2.62, "learning_rate": 7.376538853603074e-05, "loss": 2.7581, "step": 33457 }, { "epoch": 2.62, "learning_rate": 7.376460440680625e-05, "loss": 2.8238, "step": 33458 }, { "epoch": 2.62, "learning_rate": 7.376382027758176e-05, "loss": 2.8609, "step": 33459 }, { "epoch": 2.62, "learning_rate": 7.376303614835726e-05, "loss": 2.883, "step": 33460 }, { "epoch": 2.62, "learning_rate": 7.376225201913277e-05, "loss": 2.7033, "step": 33461 }, { "epoch": 2.62, "learning_rate": 7.376146788990827e-05, "loss": 2.9474, "step": 33462 }, { "epoch": 2.62, "learning_rate": 7.376068376068377e-05, "loss": 2.8303, "step": 33463 }, { "epoch": 2.62, "learning_rate": 7.375989963145928e-05, "loss": 2.8238, "step": 33464 }, { "epoch": 2.62, "learning_rate": 7.375911550223478e-05, "loss": 2.7181, "step": 33465 }, { "epoch": 2.62, "learning_rate": 7.375833137301028e-05, "loss": 2.9558, "step": 33466 }, { "epoch": 2.62, "learning_rate": 7.375754724378577e-05, "loss": 2.6663, "step": 33467 }, { "epoch": 2.62, "learning_rate": 7.375676311456128e-05, "loss": 2.7595, "step": 33468 }, { "epoch": 2.62, "learning_rate": 7.375597898533678e-05, "loss": 2.787, "step": 33469 }, { "epoch": 2.62, "learning_rate": 7.375519485611229e-05, "loss": 2.7526, "step": 33470 }, { "epoch": 2.62, "learning_rate": 7.375441072688779e-05, "loss": 2.6559, "step": 33471 }, { "epoch": 2.62, "learning_rate": 7.375362659766329e-05, "loss": 2.8315, "step": 33472 }, { "epoch": 2.62, "learning_rate": 7.37528424684388e-05, "loss": 2.7471, "step": 33473 }, { "epoch": 2.62, "learning_rate": 7.37520583392143e-05, "loss": 2.7965, "step": 33474 }, { "epoch": 2.62, "learning_rate": 7.37512742099898e-05, "loss": 2.8074, "step": 33475 }, { "epoch": 2.62, "learning_rate": 7.375049008076531e-05, "loss": 2.798, "step": 33476 }, { "epoch": 2.63, "learning_rate": 7.374970595154081e-05, "loss": 2.8191, "step": 33477 }, { "epoch": 2.63, "learning_rate": 7.374892182231631e-05, "loss": 2.8361, "step": 33478 }, { "epoch": 2.63, "learning_rate": 7.374813769309182e-05, "loss": 2.9378, "step": 33479 }, { "epoch": 2.63, "learning_rate": 7.374735356386732e-05, "loss": 2.6605, "step": 33480 }, { "epoch": 2.63, "learning_rate": 7.374656943464283e-05, "loss": 2.7938, "step": 33481 }, { "epoch": 2.63, "learning_rate": 7.374578530541833e-05, "loss": 2.6793, "step": 33482 }, { "epoch": 2.63, "learning_rate": 7.374500117619385e-05, "loss": 2.7404, "step": 33483 }, { "epoch": 2.63, "learning_rate": 7.374421704696935e-05, "loss": 2.6354, "step": 33484 }, { "epoch": 2.63, "learning_rate": 7.374343291774485e-05, "loss": 2.6827, "step": 33485 }, { "epoch": 2.63, "learning_rate": 7.374264878852036e-05, "loss": 2.9163, "step": 33486 }, { "epoch": 2.63, "learning_rate": 7.374186465929586e-05, "loss": 2.7468, "step": 33487 }, { "epoch": 2.63, "learning_rate": 7.374108053007136e-05, "loss": 2.7276, "step": 33488 }, { "epoch": 2.63, "learning_rate": 7.374029640084687e-05, "loss": 2.7276, "step": 33489 }, { "epoch": 2.63, "learning_rate": 7.373951227162237e-05, "loss": 2.5877, "step": 33490 }, { "epoch": 2.63, "learning_rate": 7.373872814239788e-05, "loss": 2.6908, "step": 33491 }, { "epoch": 2.63, "learning_rate": 7.373794401317338e-05, "loss": 2.7664, "step": 33492 }, { "epoch": 2.63, "learning_rate": 7.373715988394888e-05, "loss": 2.7149, "step": 33493 }, { "epoch": 2.63, "learning_rate": 7.373637575472439e-05, "loss": 2.9606, "step": 33494 }, { "epoch": 2.63, "learning_rate": 7.373559162549989e-05, "loss": 3.0957, "step": 33495 }, { "epoch": 2.63, "learning_rate": 7.37348074962754e-05, "loss": 2.727, "step": 33496 }, { "epoch": 2.63, "learning_rate": 7.37340233670509e-05, "loss": 2.8268, "step": 33497 }, { "epoch": 2.63, "learning_rate": 7.37332392378264e-05, "loss": 2.7255, "step": 33498 }, { "epoch": 2.63, "learning_rate": 7.37324551086019e-05, "loss": 2.795, "step": 33499 }, { "epoch": 2.63, "learning_rate": 7.373167097937741e-05, "loss": 2.856, "step": 33500 }, { "epoch": 2.63, "learning_rate": 7.373088685015291e-05, "loss": 2.732, "step": 33501 }, { "epoch": 2.63, "learning_rate": 7.373010272092842e-05, "loss": 2.9844, "step": 33502 }, { "epoch": 2.63, "learning_rate": 7.37293185917039e-05, "loss": 2.8608, "step": 33503 }, { "epoch": 2.63, "learning_rate": 7.372853446247941e-05, "loss": 2.7187, "step": 33504 }, { "epoch": 2.63, "learning_rate": 7.372775033325491e-05, "loss": 2.9748, "step": 33505 }, { "epoch": 2.63, "learning_rate": 7.372696620403042e-05, "loss": 2.9416, "step": 33506 }, { "epoch": 2.63, "learning_rate": 7.372618207480592e-05, "loss": 2.7652, "step": 33507 }, { "epoch": 2.63, "learning_rate": 7.372539794558142e-05, "loss": 2.9575, "step": 33508 }, { "epoch": 2.63, "learning_rate": 7.372461381635694e-05, "loss": 2.5629, "step": 33509 }, { "epoch": 2.63, "learning_rate": 7.372382968713245e-05, "loss": 3.0048, "step": 33510 }, { "epoch": 2.63, "learning_rate": 7.372304555790795e-05, "loss": 2.7211, "step": 33511 }, { "epoch": 2.63, "learning_rate": 7.372226142868345e-05, "loss": 2.6158, "step": 33512 }, { "epoch": 2.63, "learning_rate": 7.372147729945896e-05, "loss": 3.1397, "step": 33513 }, { "epoch": 2.63, "learning_rate": 7.372069317023446e-05, "loss": 3.105, "step": 33514 }, { "epoch": 2.63, "learning_rate": 7.371990904100996e-05, "loss": 2.911, "step": 33515 }, { "epoch": 2.63, "learning_rate": 7.371912491178547e-05, "loss": 2.8116, "step": 33516 }, { "epoch": 2.63, "learning_rate": 7.371834078256097e-05, "loss": 2.7135, "step": 33517 }, { "epoch": 2.63, "learning_rate": 7.371755665333648e-05, "loss": 2.7295, "step": 33518 }, { "epoch": 2.63, "learning_rate": 7.371677252411198e-05, "loss": 2.7789, "step": 33519 }, { "epoch": 2.63, "learning_rate": 7.371598839488748e-05, "loss": 3.3135, "step": 33520 }, { "epoch": 2.63, "learning_rate": 7.371520426566299e-05, "loss": 2.7845, "step": 33521 }, { "epoch": 2.63, "learning_rate": 7.371442013643849e-05, "loss": 2.8541, "step": 33522 }, { "epoch": 2.63, "learning_rate": 7.3713636007214e-05, "loss": 2.9216, "step": 33523 }, { "epoch": 2.63, "learning_rate": 7.37128518779895e-05, "loss": 2.7429, "step": 33524 }, { "epoch": 2.63, "learning_rate": 7.3712067748765e-05, "loss": 2.7643, "step": 33525 }, { "epoch": 2.63, "learning_rate": 7.37112836195405e-05, "loss": 2.6991, "step": 33526 }, { "epoch": 2.63, "learning_rate": 7.371049949031601e-05, "loss": 2.7746, "step": 33527 }, { "epoch": 2.63, "learning_rate": 7.370971536109151e-05, "loss": 3.1032, "step": 33528 }, { "epoch": 2.63, "learning_rate": 7.370893123186702e-05, "loss": 2.7723, "step": 33529 }, { "epoch": 2.63, "learning_rate": 7.370814710264252e-05, "loss": 2.7638, "step": 33530 }, { "epoch": 2.63, "learning_rate": 7.370736297341802e-05, "loss": 2.7868, "step": 33531 }, { "epoch": 2.63, "learning_rate": 7.370657884419353e-05, "loss": 2.7752, "step": 33532 }, { "epoch": 2.63, "learning_rate": 7.370579471496903e-05, "loss": 2.5715, "step": 33533 }, { "epoch": 2.63, "learning_rate": 7.370501058574453e-05, "loss": 2.8137, "step": 33534 }, { "epoch": 2.63, "learning_rate": 7.370422645652004e-05, "loss": 2.7919, "step": 33535 }, { "epoch": 2.63, "learning_rate": 7.370344232729554e-05, "loss": 2.6838, "step": 33536 }, { "epoch": 2.63, "learning_rate": 7.370265819807105e-05, "loss": 2.6869, "step": 33537 }, { "epoch": 2.63, "learning_rate": 7.370187406884655e-05, "loss": 2.7721, "step": 33538 }, { "epoch": 2.63, "learning_rate": 7.370108993962205e-05, "loss": 2.8028, "step": 33539 }, { "epoch": 2.63, "learning_rate": 7.370030581039756e-05, "loss": 2.805, "step": 33540 }, { "epoch": 2.63, "learning_rate": 7.369952168117306e-05, "loss": 2.8065, "step": 33541 }, { "epoch": 2.63, "learning_rate": 7.369873755194856e-05, "loss": 2.6583, "step": 33542 }, { "epoch": 2.63, "learning_rate": 7.369795342272407e-05, "loss": 2.9158, "step": 33543 }, { "epoch": 2.63, "learning_rate": 7.369716929349957e-05, "loss": 3.0603, "step": 33544 }, { "epoch": 2.63, "learning_rate": 7.369638516427507e-05, "loss": 2.7869, "step": 33545 }, { "epoch": 2.63, "learning_rate": 7.369560103505058e-05, "loss": 2.8901, "step": 33546 }, { "epoch": 2.63, "learning_rate": 7.369481690582608e-05, "loss": 2.9285, "step": 33547 }, { "epoch": 2.63, "learning_rate": 7.369403277660159e-05, "loss": 2.8467, "step": 33548 }, { "epoch": 2.63, "learning_rate": 7.369324864737709e-05, "loss": 2.5611, "step": 33549 }, { "epoch": 2.63, "learning_rate": 7.369246451815259e-05, "loss": 2.6971, "step": 33550 }, { "epoch": 2.63, "learning_rate": 7.36916803889281e-05, "loss": 2.9261, "step": 33551 }, { "epoch": 2.63, "learning_rate": 7.36908962597036e-05, "loss": 2.8851, "step": 33552 }, { "epoch": 2.63, "learning_rate": 7.36901121304791e-05, "loss": 2.7689, "step": 33553 }, { "epoch": 2.63, "learning_rate": 7.368932800125461e-05, "loss": 2.688, "step": 33554 }, { "epoch": 2.63, "learning_rate": 7.368854387203011e-05, "loss": 2.9376, "step": 33555 }, { "epoch": 2.63, "learning_rate": 7.368775974280561e-05, "loss": 2.7343, "step": 33556 }, { "epoch": 2.63, "learning_rate": 7.368697561358112e-05, "loss": 2.7625, "step": 33557 }, { "epoch": 2.63, "learning_rate": 7.368619148435662e-05, "loss": 2.7556, "step": 33558 }, { "epoch": 2.63, "learning_rate": 7.368540735513213e-05, "loss": 2.8834, "step": 33559 }, { "epoch": 2.63, "learning_rate": 7.368462322590763e-05, "loss": 3.0117, "step": 33560 }, { "epoch": 2.63, "learning_rate": 7.368383909668313e-05, "loss": 2.642, "step": 33561 }, { "epoch": 2.63, "learning_rate": 7.368305496745864e-05, "loss": 2.7406, "step": 33562 }, { "epoch": 2.63, "learning_rate": 7.368227083823414e-05, "loss": 2.6842, "step": 33563 }, { "epoch": 2.63, "learning_rate": 7.368148670900964e-05, "loss": 2.9306, "step": 33564 }, { "epoch": 2.63, "learning_rate": 7.368070257978515e-05, "loss": 2.5972, "step": 33565 }, { "epoch": 2.63, "learning_rate": 7.367991845056067e-05, "loss": 2.8197, "step": 33566 }, { "epoch": 2.63, "learning_rate": 7.367913432133617e-05, "loss": 2.5216, "step": 33567 }, { "epoch": 2.63, "learning_rate": 7.367835019211167e-05, "loss": 2.728, "step": 33568 }, { "epoch": 2.63, "learning_rate": 7.367756606288718e-05, "loss": 2.7668, "step": 33569 }, { "epoch": 2.63, "learning_rate": 7.367678193366268e-05, "loss": 2.7774, "step": 33570 }, { "epoch": 2.63, "learning_rate": 7.367599780443818e-05, "loss": 2.8646, "step": 33571 }, { "epoch": 2.63, "learning_rate": 7.367521367521369e-05, "loss": 2.4533, "step": 33572 }, { "epoch": 2.63, "learning_rate": 7.367442954598919e-05, "loss": 2.8013, "step": 33573 }, { "epoch": 2.63, "learning_rate": 7.36736454167647e-05, "loss": 2.8534, "step": 33574 }, { "epoch": 2.63, "learning_rate": 7.367286128754018e-05, "loss": 2.7845, "step": 33575 }, { "epoch": 2.63, "learning_rate": 7.367207715831569e-05, "loss": 2.7658, "step": 33576 }, { "epoch": 2.63, "learning_rate": 7.367129302909119e-05, "loss": 2.8549, "step": 33577 }, { "epoch": 2.63, "learning_rate": 7.36705088998667e-05, "loss": 2.852, "step": 33578 }, { "epoch": 2.63, "learning_rate": 7.36697247706422e-05, "loss": 2.8097, "step": 33579 }, { "epoch": 2.63, "learning_rate": 7.36689406414177e-05, "loss": 2.6743, "step": 33580 }, { "epoch": 2.63, "learning_rate": 7.366815651219321e-05, "loss": 2.7034, "step": 33581 }, { "epoch": 2.63, "learning_rate": 7.366737238296871e-05, "loss": 2.7824, "step": 33582 }, { "epoch": 2.63, "learning_rate": 7.366658825374421e-05, "loss": 2.6619, "step": 33583 }, { "epoch": 2.63, "learning_rate": 7.366580412451972e-05, "loss": 2.7791, "step": 33584 }, { "epoch": 2.63, "learning_rate": 7.366501999529522e-05, "loss": 2.7472, "step": 33585 }, { "epoch": 2.63, "learning_rate": 7.366423586607073e-05, "loss": 2.916, "step": 33586 }, { "epoch": 2.63, "learning_rate": 7.366345173684623e-05, "loss": 2.7398, "step": 33587 }, { "epoch": 2.63, "learning_rate": 7.366266760762173e-05, "loss": 2.9561, "step": 33588 }, { "epoch": 2.63, "learning_rate": 7.366188347839724e-05, "loss": 2.5491, "step": 33589 }, { "epoch": 2.63, "learning_rate": 7.366109934917274e-05, "loss": 2.8647, "step": 33590 }, { "epoch": 2.63, "learning_rate": 7.366031521994826e-05, "loss": 2.6653, "step": 33591 }, { "epoch": 2.63, "learning_rate": 7.365953109072376e-05, "loss": 2.945, "step": 33592 }, { "epoch": 2.63, "learning_rate": 7.365874696149926e-05, "loss": 2.9146, "step": 33593 }, { "epoch": 2.63, "learning_rate": 7.365796283227477e-05, "loss": 2.6258, "step": 33594 }, { "epoch": 2.63, "learning_rate": 7.365717870305027e-05, "loss": 2.8473, "step": 33595 }, { "epoch": 2.63, "learning_rate": 7.365639457382578e-05, "loss": 2.848, "step": 33596 }, { "epoch": 2.63, "learning_rate": 7.365561044460128e-05, "loss": 2.9693, "step": 33597 }, { "epoch": 2.63, "learning_rate": 7.365482631537678e-05, "loss": 2.9854, "step": 33598 }, { "epoch": 2.63, "learning_rate": 7.365404218615229e-05, "loss": 2.7482, "step": 33599 }, { "epoch": 2.63, "learning_rate": 7.365325805692779e-05, "loss": 2.4897, "step": 33600 }, { "epoch": 2.63, "learning_rate": 7.36524739277033e-05, "loss": 2.9559, "step": 33601 }, { "epoch": 2.63, "learning_rate": 7.36516897984788e-05, "loss": 2.6894, "step": 33602 }, { "epoch": 2.63, "learning_rate": 7.36509056692543e-05, "loss": 2.6381, "step": 33603 }, { "epoch": 2.63, "learning_rate": 7.36501215400298e-05, "loss": 2.9005, "step": 33604 }, { "epoch": 2.64, "learning_rate": 7.364933741080531e-05, "loss": 2.828, "step": 33605 }, { "epoch": 2.64, "learning_rate": 7.364855328158081e-05, "loss": 3.0244, "step": 33606 }, { "epoch": 2.64, "learning_rate": 7.364776915235632e-05, "loss": 3.1272, "step": 33607 }, { "epoch": 2.64, "learning_rate": 7.364698502313182e-05, "loss": 2.6592, "step": 33608 }, { "epoch": 2.64, "learning_rate": 7.364620089390732e-05, "loss": 2.7639, "step": 33609 }, { "epoch": 2.64, "learning_rate": 7.364541676468283e-05, "loss": 2.5862, "step": 33610 }, { "epoch": 2.64, "learning_rate": 7.364463263545832e-05, "loss": 2.8416, "step": 33611 }, { "epoch": 2.64, "learning_rate": 7.364384850623382e-05, "loss": 2.6416, "step": 33612 }, { "epoch": 2.64, "learning_rate": 7.364306437700932e-05, "loss": 2.8041, "step": 33613 }, { "epoch": 2.64, "learning_rate": 7.364228024778483e-05, "loss": 2.7891, "step": 33614 }, { "epoch": 2.64, "learning_rate": 7.364149611856033e-05, "loss": 2.885, "step": 33615 }, { "epoch": 2.64, "learning_rate": 7.364071198933584e-05, "loss": 2.8992, "step": 33616 }, { "epoch": 2.64, "learning_rate": 7.363992786011135e-05, "loss": 2.8018, "step": 33617 }, { "epoch": 2.64, "learning_rate": 7.363914373088686e-05, "loss": 2.7648, "step": 33618 }, { "epoch": 2.64, "learning_rate": 7.363835960166236e-05, "loss": 2.8286, "step": 33619 }, { "epoch": 2.64, "learning_rate": 7.363757547243786e-05, "loss": 2.6244, "step": 33620 }, { "epoch": 2.64, "learning_rate": 7.363679134321337e-05, "loss": 3.0409, "step": 33621 }, { "epoch": 2.64, "learning_rate": 7.363600721398887e-05, "loss": 2.9178, "step": 33622 }, { "epoch": 2.64, "learning_rate": 7.363522308476437e-05, "loss": 2.7989, "step": 33623 }, { "epoch": 2.64, "learning_rate": 7.363443895553988e-05, "loss": 2.7896, "step": 33624 }, { "epoch": 2.64, "learning_rate": 7.363365482631538e-05, "loss": 2.7692, "step": 33625 }, { "epoch": 2.64, "learning_rate": 7.363287069709089e-05, "loss": 2.6408, "step": 33626 }, { "epoch": 2.64, "learning_rate": 7.363208656786639e-05, "loss": 2.7952, "step": 33627 }, { "epoch": 2.64, "learning_rate": 7.36313024386419e-05, "loss": 3.3315, "step": 33628 }, { "epoch": 2.64, "learning_rate": 7.36305183094174e-05, "loss": 2.9484, "step": 33629 }, { "epoch": 2.64, "learning_rate": 7.36297341801929e-05, "loss": 2.8779, "step": 33630 }, { "epoch": 2.64, "learning_rate": 7.36289500509684e-05, "loss": 2.7876, "step": 33631 }, { "epoch": 2.64, "learning_rate": 7.362816592174391e-05, "loss": 2.802, "step": 33632 }, { "epoch": 2.64, "learning_rate": 7.362738179251941e-05, "loss": 2.6884, "step": 33633 }, { "epoch": 2.64, "learning_rate": 7.362659766329492e-05, "loss": 2.7636, "step": 33634 }, { "epoch": 2.64, "learning_rate": 7.362581353407042e-05, "loss": 2.8329, "step": 33635 }, { "epoch": 2.64, "learning_rate": 7.362502940484592e-05, "loss": 2.7358, "step": 33636 }, { "epoch": 2.64, "learning_rate": 7.362424527562143e-05, "loss": 2.6572, "step": 33637 }, { "epoch": 2.64, "learning_rate": 7.362346114639693e-05, "loss": 2.9244, "step": 33638 }, { "epoch": 2.64, "learning_rate": 7.362267701717243e-05, "loss": 2.6521, "step": 33639 }, { "epoch": 2.64, "learning_rate": 7.362189288794794e-05, "loss": 2.6694, "step": 33640 }, { "epoch": 2.64, "learning_rate": 7.362110875872344e-05, "loss": 2.9107, "step": 33641 }, { "epoch": 2.64, "learning_rate": 7.362032462949894e-05, "loss": 2.6361, "step": 33642 }, { "epoch": 2.64, "learning_rate": 7.361954050027445e-05, "loss": 2.6482, "step": 33643 }, { "epoch": 2.64, "learning_rate": 7.361875637104995e-05, "loss": 2.7823, "step": 33644 }, { "epoch": 2.64, "learning_rate": 7.361797224182546e-05, "loss": 2.6754, "step": 33645 }, { "epoch": 2.64, "learning_rate": 7.361718811260096e-05, "loss": 2.8081, "step": 33646 }, { "epoch": 2.64, "learning_rate": 7.361640398337646e-05, "loss": 2.8083, "step": 33647 }, { "epoch": 2.64, "learning_rate": 7.361561985415197e-05, "loss": 2.8519, "step": 33648 }, { "epoch": 2.64, "learning_rate": 7.361483572492747e-05, "loss": 2.5987, "step": 33649 }, { "epoch": 2.64, "learning_rate": 7.361405159570297e-05, "loss": 2.8785, "step": 33650 }, { "epoch": 2.64, "learning_rate": 7.361326746647848e-05, "loss": 2.843, "step": 33651 }, { "epoch": 2.64, "learning_rate": 7.361248333725398e-05, "loss": 2.8154, "step": 33652 }, { "epoch": 2.64, "learning_rate": 7.361169920802949e-05, "loss": 2.9795, "step": 33653 }, { "epoch": 2.64, "learning_rate": 7.361091507880499e-05, "loss": 2.7655, "step": 33654 }, { "epoch": 2.64, "learning_rate": 7.361013094958049e-05, "loss": 2.9555, "step": 33655 }, { "epoch": 2.64, "learning_rate": 7.3609346820356e-05, "loss": 2.9911, "step": 33656 }, { "epoch": 2.64, "learning_rate": 7.36085626911315e-05, "loss": 2.583, "step": 33657 }, { "epoch": 2.64, "learning_rate": 7.3607778561907e-05, "loss": 2.8077, "step": 33658 }, { "epoch": 2.64, "learning_rate": 7.360699443268251e-05, "loss": 3.1355, "step": 33659 }, { "epoch": 2.64, "learning_rate": 7.360621030345801e-05, "loss": 2.7875, "step": 33660 }, { "epoch": 2.64, "learning_rate": 7.360542617423351e-05, "loss": 2.746, "step": 33661 }, { "epoch": 2.64, "learning_rate": 7.360464204500902e-05, "loss": 2.8347, "step": 33662 }, { "epoch": 2.64, "learning_rate": 7.360385791578452e-05, "loss": 2.7042, "step": 33663 }, { "epoch": 2.64, "learning_rate": 7.360307378656003e-05, "loss": 2.6433, "step": 33664 }, { "epoch": 2.64, "learning_rate": 7.360228965733553e-05, "loss": 2.9331, "step": 33665 }, { "epoch": 2.64, "learning_rate": 7.360150552811103e-05, "loss": 2.6721, "step": 33666 }, { "epoch": 2.64, "learning_rate": 7.360072139888654e-05, "loss": 2.8875, "step": 33667 }, { "epoch": 2.64, "learning_rate": 7.359993726966204e-05, "loss": 2.9348, "step": 33668 }, { "epoch": 2.64, "learning_rate": 7.359915314043754e-05, "loss": 2.725, "step": 33669 }, { "epoch": 2.64, "learning_rate": 7.359836901121305e-05, "loss": 2.76, "step": 33670 }, { "epoch": 2.64, "learning_rate": 7.359758488198855e-05, "loss": 2.711, "step": 33671 }, { "epoch": 2.64, "learning_rate": 7.359680075276406e-05, "loss": 2.9531, "step": 33672 }, { "epoch": 2.64, "learning_rate": 7.359601662353956e-05, "loss": 2.5978, "step": 33673 }, { "epoch": 2.64, "learning_rate": 7.359523249431508e-05, "loss": 2.6923, "step": 33674 }, { "epoch": 2.64, "learning_rate": 7.359444836509058e-05, "loss": 2.9506, "step": 33675 }, { "epoch": 2.64, "learning_rate": 7.359366423586608e-05, "loss": 2.724, "step": 33676 }, { "epoch": 2.64, "learning_rate": 7.359288010664159e-05, "loss": 2.8111, "step": 33677 }, { "epoch": 2.64, "learning_rate": 7.359209597741709e-05, "loss": 2.7487, "step": 33678 }, { "epoch": 2.64, "learning_rate": 7.35913118481926e-05, "loss": 2.6891, "step": 33679 }, { "epoch": 2.64, "learning_rate": 7.35905277189681e-05, "loss": 2.6972, "step": 33680 }, { "epoch": 2.64, "learning_rate": 7.35897435897436e-05, "loss": 3.0543, "step": 33681 }, { "epoch": 2.64, "learning_rate": 7.35889594605191e-05, "loss": 2.9684, "step": 33682 }, { "epoch": 2.64, "learning_rate": 7.35881753312946e-05, "loss": 2.8664, "step": 33683 }, { "epoch": 2.64, "learning_rate": 7.35873912020701e-05, "loss": 2.867, "step": 33684 }, { "epoch": 2.64, "learning_rate": 7.35866070728456e-05, "loss": 2.7691, "step": 33685 }, { "epoch": 2.64, "learning_rate": 7.35858229436211e-05, "loss": 2.8448, "step": 33686 }, { "epoch": 2.64, "learning_rate": 7.358503881439661e-05, "loss": 3.0447, "step": 33687 }, { "epoch": 2.64, "learning_rate": 7.358425468517211e-05, "loss": 2.7976, "step": 33688 }, { "epoch": 2.64, "learning_rate": 7.358347055594762e-05, "loss": 2.8622, "step": 33689 }, { "epoch": 2.64, "learning_rate": 7.358268642672312e-05, "loss": 2.6263, "step": 33690 }, { "epoch": 2.64, "learning_rate": 7.358190229749862e-05, "loss": 2.9322, "step": 33691 }, { "epoch": 2.64, "learning_rate": 7.358111816827413e-05, "loss": 2.8607, "step": 33692 }, { "epoch": 2.64, "learning_rate": 7.358033403904963e-05, "loss": 2.8421, "step": 33693 }, { "epoch": 2.64, "learning_rate": 7.357954990982514e-05, "loss": 2.8066, "step": 33694 }, { "epoch": 2.64, "learning_rate": 7.357876578060064e-05, "loss": 2.9768, "step": 33695 }, { "epoch": 2.64, "learning_rate": 7.357798165137614e-05, "loss": 2.7538, "step": 33696 }, { "epoch": 2.64, "learning_rate": 7.357719752215165e-05, "loss": 2.7495, "step": 33697 }, { "epoch": 2.64, "learning_rate": 7.357641339292715e-05, "loss": 2.7875, "step": 33698 }, { "epoch": 2.64, "learning_rate": 7.357562926370267e-05, "loss": 2.8407, "step": 33699 }, { "epoch": 2.64, "learning_rate": 7.357484513447817e-05, "loss": 2.8538, "step": 33700 }, { "epoch": 2.64, "learning_rate": 7.357406100525368e-05, "loss": 2.6334, "step": 33701 }, { "epoch": 2.64, "learning_rate": 7.357327687602918e-05, "loss": 2.811, "step": 33702 }, { "epoch": 2.64, "learning_rate": 7.357249274680468e-05, "loss": 2.7547, "step": 33703 }, { "epoch": 2.64, "learning_rate": 7.357170861758019e-05, "loss": 2.6793, "step": 33704 }, { "epoch": 2.64, "learning_rate": 7.357092448835569e-05, "loss": 3.0218, "step": 33705 }, { "epoch": 2.64, "learning_rate": 7.35701403591312e-05, "loss": 2.8683, "step": 33706 }, { "epoch": 2.64, "learning_rate": 7.35693562299067e-05, "loss": 2.6322, "step": 33707 }, { "epoch": 2.64, "learning_rate": 7.35685721006822e-05, "loss": 3.0706, "step": 33708 }, { "epoch": 2.64, "learning_rate": 7.35677879714577e-05, "loss": 2.6458, "step": 33709 }, { "epoch": 2.64, "learning_rate": 7.356700384223321e-05, "loss": 2.7145, "step": 33710 }, { "epoch": 2.64, "learning_rate": 7.356621971300871e-05, "loss": 2.8356, "step": 33711 }, { "epoch": 2.64, "learning_rate": 7.356543558378422e-05, "loss": 2.7067, "step": 33712 }, { "epoch": 2.64, "learning_rate": 7.356465145455972e-05, "loss": 2.9167, "step": 33713 }, { "epoch": 2.64, "learning_rate": 7.356386732533522e-05, "loss": 2.7845, "step": 33714 }, { "epoch": 2.64, "learning_rate": 7.356308319611073e-05, "loss": 2.7772, "step": 33715 }, { "epoch": 2.64, "learning_rate": 7.356229906688623e-05, "loss": 2.7973, "step": 33716 }, { "epoch": 2.64, "learning_rate": 7.356151493766173e-05, "loss": 2.7198, "step": 33717 }, { "epoch": 2.64, "learning_rate": 7.356073080843724e-05, "loss": 2.7624, "step": 33718 }, { "epoch": 2.64, "learning_rate": 7.355994667921273e-05, "loss": 2.7399, "step": 33719 }, { "epoch": 2.64, "learning_rate": 7.355916254998823e-05, "loss": 2.8185, "step": 33720 }, { "epoch": 2.64, "learning_rate": 7.355837842076374e-05, "loss": 2.8086, "step": 33721 }, { "epoch": 2.64, "learning_rate": 7.355759429153924e-05, "loss": 2.7626, "step": 33722 }, { "epoch": 2.64, "learning_rate": 7.355681016231474e-05, "loss": 2.9279, "step": 33723 }, { "epoch": 2.64, "learning_rate": 7.355602603309025e-05, "loss": 2.9156, "step": 33724 }, { "epoch": 2.64, "learning_rate": 7.355524190386576e-05, "loss": 2.7643, "step": 33725 }, { "epoch": 2.64, "learning_rate": 7.355445777464127e-05, "loss": 2.7059, "step": 33726 }, { "epoch": 2.64, "learning_rate": 7.355367364541677e-05, "loss": 2.7268, "step": 33727 }, { "epoch": 2.64, "learning_rate": 7.355288951619227e-05, "loss": 2.9913, "step": 33728 }, { "epoch": 2.64, "learning_rate": 7.355210538696778e-05, "loss": 2.8837, "step": 33729 }, { "epoch": 2.64, "learning_rate": 7.355132125774328e-05, "loss": 2.8156, "step": 33730 }, { "epoch": 2.64, "learning_rate": 7.355053712851879e-05, "loss": 2.7105, "step": 33731 }, { "epoch": 2.65, "learning_rate": 7.354975299929429e-05, "loss": 2.9859, "step": 33732 }, { "epoch": 2.65, "learning_rate": 7.354896887006979e-05, "loss": 2.675, "step": 33733 }, { "epoch": 2.65, "learning_rate": 7.35481847408453e-05, "loss": 2.834, "step": 33734 }, { "epoch": 2.65, "learning_rate": 7.35474006116208e-05, "loss": 2.5811, "step": 33735 }, { "epoch": 2.65, "learning_rate": 7.35466164823963e-05, "loss": 2.7292, "step": 33736 }, { "epoch": 2.65, "learning_rate": 7.354583235317181e-05, "loss": 2.6838, "step": 33737 }, { "epoch": 2.65, "learning_rate": 7.354504822394731e-05, "loss": 2.8673, "step": 33738 }, { "epoch": 2.65, "learning_rate": 7.354426409472282e-05, "loss": 2.7405, "step": 33739 }, { "epoch": 2.65, "learning_rate": 7.354347996549832e-05, "loss": 2.7503, "step": 33740 }, { "epoch": 2.65, "learning_rate": 7.354269583627382e-05, "loss": 2.8388, "step": 33741 }, { "epoch": 2.65, "learning_rate": 7.354191170704933e-05, "loss": 2.8828, "step": 33742 }, { "epoch": 2.65, "learning_rate": 7.354112757782483e-05, "loss": 2.7582, "step": 33743 }, { "epoch": 2.65, "learning_rate": 7.354034344860033e-05, "loss": 2.9416, "step": 33744 }, { "epoch": 2.65, "learning_rate": 7.353955931937584e-05, "loss": 2.7239, "step": 33745 }, { "epoch": 2.65, "learning_rate": 7.353877519015134e-05, "loss": 2.6856, "step": 33746 }, { "epoch": 2.65, "learning_rate": 7.353799106092684e-05, "loss": 2.997, "step": 33747 }, { "epoch": 2.65, "learning_rate": 7.353720693170235e-05, "loss": 2.8417, "step": 33748 }, { "epoch": 2.65, "learning_rate": 7.353642280247785e-05, "loss": 3.0584, "step": 33749 }, { "epoch": 2.65, "learning_rate": 7.353563867325336e-05, "loss": 2.8633, "step": 33750 }, { "epoch": 2.65, "learning_rate": 7.353485454402886e-05, "loss": 2.7947, "step": 33751 }, { "epoch": 2.65, "learning_rate": 7.353407041480436e-05, "loss": 2.8339, "step": 33752 }, { "epoch": 2.65, "learning_rate": 7.353328628557987e-05, "loss": 2.757, "step": 33753 }, { "epoch": 2.65, "learning_rate": 7.353250215635537e-05, "loss": 2.6146, "step": 33754 }, { "epoch": 2.65, "learning_rate": 7.353171802713087e-05, "loss": 2.8717, "step": 33755 }, { "epoch": 2.65, "learning_rate": 7.353093389790638e-05, "loss": 2.8128, "step": 33756 }, { "epoch": 2.65, "learning_rate": 7.353014976868188e-05, "loss": 2.7182, "step": 33757 }, { "epoch": 2.65, "learning_rate": 7.352936563945738e-05, "loss": 2.6963, "step": 33758 }, { "epoch": 2.65, "learning_rate": 7.352858151023289e-05, "loss": 2.7403, "step": 33759 }, { "epoch": 2.65, "learning_rate": 7.352779738100839e-05, "loss": 2.9673, "step": 33760 }, { "epoch": 2.65, "learning_rate": 7.35270132517839e-05, "loss": 2.6775, "step": 33761 }, { "epoch": 2.65, "learning_rate": 7.35262291225594e-05, "loss": 2.7748, "step": 33762 }, { "epoch": 2.65, "learning_rate": 7.35254449933349e-05, "loss": 2.6319, "step": 33763 }, { "epoch": 2.65, "learning_rate": 7.352466086411041e-05, "loss": 2.8385, "step": 33764 }, { "epoch": 2.65, "learning_rate": 7.352387673488591e-05, "loss": 2.7277, "step": 33765 }, { "epoch": 2.65, "learning_rate": 7.352309260566141e-05, "loss": 2.7432, "step": 33766 }, { "epoch": 2.65, "learning_rate": 7.352230847643692e-05, "loss": 2.6703, "step": 33767 }, { "epoch": 2.65, "learning_rate": 7.352152434721242e-05, "loss": 2.6903, "step": 33768 }, { "epoch": 2.65, "learning_rate": 7.352074021798793e-05, "loss": 2.6245, "step": 33769 }, { "epoch": 2.65, "learning_rate": 7.351995608876343e-05, "loss": 2.8808, "step": 33770 }, { "epoch": 2.65, "learning_rate": 7.351917195953893e-05, "loss": 2.6571, "step": 33771 }, { "epoch": 2.65, "learning_rate": 7.351838783031444e-05, "loss": 3.2869, "step": 33772 }, { "epoch": 2.65, "learning_rate": 7.351760370108994e-05, "loss": 2.9226, "step": 33773 }, { "epoch": 2.65, "learning_rate": 7.351681957186544e-05, "loss": 2.6588, "step": 33774 }, { "epoch": 2.65, "learning_rate": 7.351603544264095e-05, "loss": 2.8783, "step": 33775 }, { "epoch": 2.65, "learning_rate": 7.351525131341645e-05, "loss": 2.7972, "step": 33776 }, { "epoch": 2.65, "learning_rate": 7.351446718419195e-05, "loss": 2.8933, "step": 33777 }, { "epoch": 2.65, "learning_rate": 7.351368305496746e-05, "loss": 2.9782, "step": 33778 }, { "epoch": 2.65, "learning_rate": 7.351289892574296e-05, "loss": 2.5159, "step": 33779 }, { "epoch": 2.65, "learning_rate": 7.351211479651847e-05, "loss": 2.9419, "step": 33780 }, { "epoch": 2.65, "learning_rate": 7.351133066729397e-05, "loss": 2.7409, "step": 33781 }, { "epoch": 2.65, "learning_rate": 7.351054653806949e-05, "loss": 2.905, "step": 33782 }, { "epoch": 2.65, "learning_rate": 7.350976240884499e-05, "loss": 2.8116, "step": 33783 }, { "epoch": 2.65, "learning_rate": 7.35089782796205e-05, "loss": 2.9187, "step": 33784 }, { "epoch": 2.65, "learning_rate": 7.3508194150396e-05, "loss": 2.8774, "step": 33785 }, { "epoch": 2.65, "learning_rate": 7.35074100211715e-05, "loss": 2.8818, "step": 33786 }, { "epoch": 2.65, "learning_rate": 7.3506625891947e-05, "loss": 2.8423, "step": 33787 }, { "epoch": 2.65, "learning_rate": 7.350584176272251e-05, "loss": 2.7981, "step": 33788 }, { "epoch": 2.65, "learning_rate": 7.350505763349801e-05, "loss": 2.8215, "step": 33789 }, { "epoch": 2.65, "learning_rate": 7.350427350427352e-05, "loss": 2.9962, "step": 33790 }, { "epoch": 2.65, "learning_rate": 7.3503489375049e-05, "loss": 2.7792, "step": 33791 }, { "epoch": 2.65, "learning_rate": 7.350270524582451e-05, "loss": 2.7116, "step": 33792 }, { "epoch": 2.65, "learning_rate": 7.350192111660001e-05, "loss": 2.7996, "step": 33793 }, { "epoch": 2.65, "learning_rate": 7.350113698737552e-05, "loss": 2.6961, "step": 33794 }, { "epoch": 2.65, "learning_rate": 7.350035285815102e-05, "loss": 2.7843, "step": 33795 }, { "epoch": 2.65, "learning_rate": 7.349956872892652e-05, "loss": 2.826, "step": 33796 }, { "epoch": 2.65, "learning_rate": 7.349878459970203e-05, "loss": 2.5734, "step": 33797 }, { "epoch": 2.65, "learning_rate": 7.349800047047753e-05, "loss": 2.7351, "step": 33798 }, { "epoch": 2.65, "learning_rate": 7.349721634125304e-05, "loss": 2.9767, "step": 33799 }, { "epoch": 2.65, "learning_rate": 7.349643221202854e-05, "loss": 2.7703, "step": 33800 }, { "epoch": 2.65, "learning_rate": 7.349564808280404e-05, "loss": 3.0691, "step": 33801 }, { "epoch": 2.65, "learning_rate": 7.349486395357955e-05, "loss": 2.5439, "step": 33802 }, { "epoch": 2.65, "learning_rate": 7.349407982435505e-05, "loss": 2.8586, "step": 33803 }, { "epoch": 2.65, "learning_rate": 7.349329569513055e-05, "loss": 2.8647, "step": 33804 }, { "epoch": 2.65, "learning_rate": 7.349251156590606e-05, "loss": 2.8266, "step": 33805 }, { "epoch": 2.65, "learning_rate": 7.349172743668156e-05, "loss": 2.7608, "step": 33806 }, { "epoch": 2.65, "learning_rate": 7.349094330745708e-05, "loss": 2.9431, "step": 33807 }, { "epoch": 2.65, "learning_rate": 7.349015917823258e-05, "loss": 2.8345, "step": 33808 }, { "epoch": 2.65, "learning_rate": 7.348937504900809e-05, "loss": 2.8002, "step": 33809 }, { "epoch": 2.65, "learning_rate": 7.348859091978359e-05, "loss": 2.9922, "step": 33810 }, { "epoch": 2.65, "learning_rate": 7.34878067905591e-05, "loss": 2.5862, "step": 33811 }, { "epoch": 2.65, "learning_rate": 7.34870226613346e-05, "loss": 2.7668, "step": 33812 }, { "epoch": 2.65, "learning_rate": 7.34862385321101e-05, "loss": 2.6309, "step": 33813 }, { "epoch": 2.65, "learning_rate": 7.34854544028856e-05, "loss": 2.7046, "step": 33814 }, { "epoch": 2.65, "learning_rate": 7.348467027366111e-05, "loss": 2.976, "step": 33815 }, { "epoch": 2.65, "learning_rate": 7.348388614443661e-05, "loss": 2.7375, "step": 33816 }, { "epoch": 2.65, "learning_rate": 7.348310201521212e-05, "loss": 2.7507, "step": 33817 }, { "epoch": 2.65, "learning_rate": 7.348231788598762e-05, "loss": 2.8511, "step": 33818 }, { "epoch": 2.65, "learning_rate": 7.348153375676312e-05, "loss": 3.03, "step": 33819 }, { "epoch": 2.65, "learning_rate": 7.348074962753863e-05, "loss": 2.8471, "step": 33820 }, { "epoch": 2.65, "learning_rate": 7.347996549831413e-05, "loss": 2.9135, "step": 33821 }, { "epoch": 2.65, "learning_rate": 7.347918136908963e-05, "loss": 2.8058, "step": 33822 }, { "epoch": 2.65, "learning_rate": 7.347839723986514e-05, "loss": 2.7446, "step": 33823 }, { "epoch": 2.65, "learning_rate": 7.347761311064064e-05, "loss": 2.8517, "step": 33824 }, { "epoch": 2.65, "learning_rate": 7.347682898141614e-05, "loss": 2.5933, "step": 33825 }, { "epoch": 2.65, "learning_rate": 7.347604485219165e-05, "loss": 2.8174, "step": 33826 }, { "epoch": 2.65, "learning_rate": 7.347526072296714e-05, "loss": 2.6517, "step": 33827 }, { "epoch": 2.65, "learning_rate": 7.347447659374264e-05, "loss": 3.158, "step": 33828 }, { "epoch": 2.65, "learning_rate": 7.347369246451815e-05, "loss": 2.6542, "step": 33829 }, { "epoch": 2.65, "learning_rate": 7.347290833529365e-05, "loss": 2.7941, "step": 33830 }, { "epoch": 2.65, "learning_rate": 7.347212420606915e-05, "loss": 2.9359, "step": 33831 }, { "epoch": 2.65, "learning_rate": 7.347134007684466e-05, "loss": 2.6518, "step": 33832 }, { "epoch": 2.65, "learning_rate": 7.347055594762017e-05, "loss": 2.7172, "step": 33833 }, { "epoch": 2.65, "learning_rate": 7.346977181839568e-05, "loss": 2.7299, "step": 33834 }, { "epoch": 2.65, "learning_rate": 7.346898768917118e-05, "loss": 2.8547, "step": 33835 }, { "epoch": 2.65, "learning_rate": 7.346820355994669e-05, "loss": 2.7635, "step": 33836 }, { "epoch": 2.65, "learning_rate": 7.346741943072219e-05, "loss": 2.749, "step": 33837 }, { "epoch": 2.65, "learning_rate": 7.346663530149769e-05, "loss": 2.964, "step": 33838 }, { "epoch": 2.65, "learning_rate": 7.34658511722732e-05, "loss": 2.6891, "step": 33839 }, { "epoch": 2.65, "learning_rate": 7.34650670430487e-05, "loss": 2.6945, "step": 33840 }, { "epoch": 2.65, "learning_rate": 7.34642829138242e-05, "loss": 2.6726, "step": 33841 }, { "epoch": 2.65, "learning_rate": 7.346349878459971e-05, "loss": 2.7699, "step": 33842 }, { "epoch": 2.65, "learning_rate": 7.346271465537521e-05, "loss": 2.8388, "step": 33843 }, { "epoch": 2.65, "learning_rate": 7.346193052615071e-05, "loss": 2.6733, "step": 33844 }, { "epoch": 2.65, "learning_rate": 7.346114639692622e-05, "loss": 2.786, "step": 33845 }, { "epoch": 2.65, "learning_rate": 7.346036226770172e-05, "loss": 2.63, "step": 33846 }, { "epoch": 2.65, "learning_rate": 7.345957813847723e-05, "loss": 3.0306, "step": 33847 }, { "epoch": 2.65, "learning_rate": 7.345879400925273e-05, "loss": 2.9557, "step": 33848 }, { "epoch": 2.65, "learning_rate": 7.345800988002823e-05, "loss": 2.8961, "step": 33849 }, { "epoch": 2.65, "learning_rate": 7.345722575080374e-05, "loss": 2.7475, "step": 33850 }, { "epoch": 2.65, "learning_rate": 7.345644162157924e-05, "loss": 2.6973, "step": 33851 }, { "epoch": 2.65, "learning_rate": 7.345565749235474e-05, "loss": 2.6394, "step": 33852 }, { "epoch": 2.65, "learning_rate": 7.345487336313025e-05, "loss": 2.9258, "step": 33853 }, { "epoch": 2.65, "learning_rate": 7.345408923390575e-05, "loss": 2.6197, "step": 33854 }, { "epoch": 2.65, "learning_rate": 7.345330510468126e-05, "loss": 3.1265, "step": 33855 }, { "epoch": 2.65, "learning_rate": 7.345252097545676e-05, "loss": 2.8675, "step": 33856 }, { "epoch": 2.65, "learning_rate": 7.345173684623226e-05, "loss": 3.089, "step": 33857 }, { "epoch": 2.65, "learning_rate": 7.345095271700777e-05, "loss": 2.8402, "step": 33858 }, { "epoch": 2.65, "learning_rate": 7.345016858778327e-05, "loss": 2.6795, "step": 33859 }, { "epoch": 2.66, "learning_rate": 7.344938445855877e-05, "loss": 2.6019, "step": 33860 }, { "epoch": 2.66, "learning_rate": 7.344860032933428e-05, "loss": 2.9301, "step": 33861 }, { "epoch": 2.66, "learning_rate": 7.344781620010978e-05, "loss": 2.7254, "step": 33862 }, { "epoch": 2.66, "learning_rate": 7.344703207088528e-05, "loss": 2.6749, "step": 33863 }, { "epoch": 2.66, "learning_rate": 7.344624794166079e-05, "loss": 2.6634, "step": 33864 }, { "epoch": 2.66, "learning_rate": 7.344546381243629e-05, "loss": 2.8339, "step": 33865 }, { "epoch": 2.66, "learning_rate": 7.34446796832118e-05, "loss": 2.8279, "step": 33866 }, { "epoch": 2.66, "learning_rate": 7.34438955539873e-05, "loss": 3.1004, "step": 33867 }, { "epoch": 2.66, "learning_rate": 7.34431114247628e-05, "loss": 2.8041, "step": 33868 }, { "epoch": 2.66, "learning_rate": 7.34423272955383e-05, "loss": 2.7426, "step": 33869 }, { "epoch": 2.66, "learning_rate": 7.344154316631381e-05, "loss": 2.6369, "step": 33870 }, { "epoch": 2.66, "learning_rate": 7.344075903708931e-05, "loss": 2.9152, "step": 33871 }, { "epoch": 2.66, "learning_rate": 7.343997490786482e-05, "loss": 2.7161, "step": 33872 }, { "epoch": 2.66, "learning_rate": 7.343919077864032e-05, "loss": 2.703, "step": 33873 }, { "epoch": 2.66, "learning_rate": 7.343840664941583e-05, "loss": 2.6495, "step": 33874 }, { "epoch": 2.66, "learning_rate": 7.343762252019133e-05, "loss": 2.8641, "step": 33875 }, { "epoch": 2.66, "learning_rate": 7.343683839096683e-05, "loss": 2.8812, "step": 33876 }, { "epoch": 2.66, "learning_rate": 7.343605426174234e-05, "loss": 2.5799, "step": 33877 }, { "epoch": 2.66, "learning_rate": 7.343527013251784e-05, "loss": 2.852, "step": 33878 }, { "epoch": 2.66, "learning_rate": 7.343448600329334e-05, "loss": 2.5855, "step": 33879 }, { "epoch": 2.66, "learning_rate": 7.343370187406885e-05, "loss": 2.7434, "step": 33880 }, { "epoch": 2.66, "learning_rate": 7.343291774484435e-05, "loss": 2.8875, "step": 33881 }, { "epoch": 2.66, "learning_rate": 7.343213361561985e-05, "loss": 2.6102, "step": 33882 }, { "epoch": 2.66, "learning_rate": 7.343134948639536e-05, "loss": 3.0895, "step": 33883 }, { "epoch": 2.66, "learning_rate": 7.343056535717086e-05, "loss": 2.6553, "step": 33884 }, { "epoch": 2.66, "learning_rate": 7.342978122794637e-05, "loss": 2.6392, "step": 33885 }, { "epoch": 2.66, "learning_rate": 7.342899709872187e-05, "loss": 2.6904, "step": 33886 }, { "epoch": 2.66, "learning_rate": 7.342821296949737e-05, "loss": 2.7126, "step": 33887 }, { "epoch": 2.66, "learning_rate": 7.342742884027288e-05, "loss": 2.7458, "step": 33888 }, { "epoch": 2.66, "learning_rate": 7.342664471104838e-05, "loss": 2.7899, "step": 33889 }, { "epoch": 2.66, "learning_rate": 7.34258605818239e-05, "loss": 2.8445, "step": 33890 }, { "epoch": 2.66, "learning_rate": 7.34250764525994e-05, "loss": 2.9043, "step": 33891 }, { "epoch": 2.66, "learning_rate": 7.34242923233749e-05, "loss": 3.0194, "step": 33892 }, { "epoch": 2.66, "learning_rate": 7.342350819415041e-05, "loss": 2.877, "step": 33893 }, { "epoch": 2.66, "learning_rate": 7.342272406492591e-05, "loss": 2.8347, "step": 33894 }, { "epoch": 2.66, "learning_rate": 7.342193993570142e-05, "loss": 2.7828, "step": 33895 }, { "epoch": 2.66, "learning_rate": 7.342115580647692e-05, "loss": 3.0168, "step": 33896 }, { "epoch": 2.66, "learning_rate": 7.342037167725242e-05, "loss": 2.7203, "step": 33897 }, { "epoch": 2.66, "learning_rate": 7.341958754802793e-05, "loss": 2.6836, "step": 33898 }, { "epoch": 2.66, "learning_rate": 7.341880341880342e-05, "loss": 2.6331, "step": 33899 }, { "epoch": 2.66, "learning_rate": 7.341801928957892e-05, "loss": 2.8094, "step": 33900 }, { "epoch": 2.66, "learning_rate": 7.341723516035442e-05, "loss": 2.7404, "step": 33901 }, { "epoch": 2.66, "learning_rate": 7.341645103112993e-05, "loss": 2.6372, "step": 33902 }, { "epoch": 2.66, "learning_rate": 7.341566690190543e-05, "loss": 2.7236, "step": 33903 }, { "epoch": 2.66, "learning_rate": 7.341488277268094e-05, "loss": 2.7072, "step": 33904 }, { "epoch": 2.66, "learning_rate": 7.341409864345644e-05, "loss": 2.5267, "step": 33905 }, { "epoch": 2.66, "learning_rate": 7.341331451423194e-05, "loss": 2.9465, "step": 33906 }, { "epoch": 2.66, "learning_rate": 7.341253038500745e-05, "loss": 2.7083, "step": 33907 }, { "epoch": 2.66, "learning_rate": 7.341174625578295e-05, "loss": 2.8791, "step": 33908 }, { "epoch": 2.66, "learning_rate": 7.341096212655845e-05, "loss": 2.7273, "step": 33909 }, { "epoch": 2.66, "learning_rate": 7.341017799733396e-05, "loss": 2.6874, "step": 33910 }, { "epoch": 2.66, "learning_rate": 7.340939386810946e-05, "loss": 2.8293, "step": 33911 }, { "epoch": 2.66, "learning_rate": 7.340860973888496e-05, "loss": 3.166, "step": 33912 }, { "epoch": 2.66, "learning_rate": 7.340782560966047e-05, "loss": 2.8446, "step": 33913 }, { "epoch": 2.66, "learning_rate": 7.340704148043597e-05, "loss": 2.8187, "step": 33914 }, { "epoch": 2.66, "learning_rate": 7.340625735121148e-05, "loss": 2.756, "step": 33915 }, { "epoch": 2.66, "learning_rate": 7.340547322198699e-05, "loss": 2.8657, "step": 33916 }, { "epoch": 2.66, "learning_rate": 7.34046890927625e-05, "loss": 2.8145, "step": 33917 }, { "epoch": 2.66, "learning_rate": 7.3403904963538e-05, "loss": 2.9778, "step": 33918 }, { "epoch": 2.66, "learning_rate": 7.34031208343135e-05, "loss": 2.8725, "step": 33919 }, { "epoch": 2.66, "learning_rate": 7.340233670508901e-05, "loss": 2.6514, "step": 33920 }, { "epoch": 2.66, "learning_rate": 7.340155257586451e-05, "loss": 3.0266, "step": 33921 }, { "epoch": 2.66, "learning_rate": 7.340076844664002e-05, "loss": 2.7682, "step": 33922 }, { "epoch": 2.66, "learning_rate": 7.339998431741552e-05, "loss": 2.7653, "step": 33923 }, { "epoch": 2.66, "learning_rate": 7.339920018819102e-05, "loss": 2.9297, "step": 33924 }, { "epoch": 2.66, "learning_rate": 7.339841605896653e-05, "loss": 2.7661, "step": 33925 }, { "epoch": 2.66, "learning_rate": 7.339763192974203e-05, "loss": 2.8458, "step": 33926 }, { "epoch": 2.66, "learning_rate": 7.339684780051753e-05, "loss": 2.6342, "step": 33927 }, { "epoch": 2.66, "learning_rate": 7.339606367129304e-05, "loss": 3.143, "step": 33928 }, { "epoch": 2.66, "learning_rate": 7.339527954206854e-05, "loss": 2.6475, "step": 33929 }, { "epoch": 2.66, "learning_rate": 7.339449541284404e-05, "loss": 2.9451, "step": 33930 }, { "epoch": 2.66, "learning_rate": 7.339371128361955e-05, "loss": 2.888, "step": 33931 }, { "epoch": 2.66, "learning_rate": 7.339292715439505e-05, "loss": 2.6707, "step": 33932 }, { "epoch": 2.66, "learning_rate": 7.339214302517056e-05, "loss": 2.7186, "step": 33933 }, { "epoch": 2.66, "learning_rate": 7.339135889594606e-05, "loss": 2.9066, "step": 33934 }, { "epoch": 2.66, "learning_rate": 7.339057476672155e-05, "loss": 2.7682, "step": 33935 }, { "epoch": 2.66, "learning_rate": 7.338979063749705e-05, "loss": 2.8086, "step": 33936 }, { "epoch": 2.66, "learning_rate": 7.338900650827256e-05, "loss": 2.6909, "step": 33937 }, { "epoch": 2.66, "learning_rate": 7.338822237904806e-05, "loss": 2.6627, "step": 33938 }, { "epoch": 2.66, "learning_rate": 7.338743824982356e-05, "loss": 2.6912, "step": 33939 }, { "epoch": 2.66, "learning_rate": 7.338665412059907e-05, "loss": 2.6031, "step": 33940 }, { "epoch": 2.66, "learning_rate": 7.338586999137459e-05, "loss": 2.8152, "step": 33941 }, { "epoch": 2.66, "learning_rate": 7.338508586215009e-05, "loss": 2.847, "step": 33942 }, { "epoch": 2.66, "learning_rate": 7.338430173292559e-05, "loss": 2.8471, "step": 33943 }, { "epoch": 2.66, "learning_rate": 7.33835176037011e-05, "loss": 2.73, "step": 33944 }, { "epoch": 2.66, "learning_rate": 7.33827334744766e-05, "loss": 2.9317, "step": 33945 }, { "epoch": 2.66, "learning_rate": 7.33819493452521e-05, "loss": 2.7649, "step": 33946 }, { "epoch": 2.66, "learning_rate": 7.338116521602761e-05, "loss": 2.7308, "step": 33947 }, { "epoch": 2.66, "learning_rate": 7.338038108680311e-05, "loss": 2.7964, "step": 33948 }, { "epoch": 2.66, "learning_rate": 7.337959695757861e-05, "loss": 2.7322, "step": 33949 }, { "epoch": 2.66, "learning_rate": 7.337881282835412e-05, "loss": 2.7207, "step": 33950 }, { "epoch": 2.66, "learning_rate": 7.337802869912962e-05, "loss": 2.6252, "step": 33951 }, { "epoch": 2.66, "learning_rate": 7.337724456990513e-05, "loss": 2.8311, "step": 33952 }, { "epoch": 2.66, "learning_rate": 7.337646044068063e-05, "loss": 3.0041, "step": 33953 }, { "epoch": 2.66, "learning_rate": 7.337567631145613e-05, "loss": 2.9795, "step": 33954 }, { "epoch": 2.66, "learning_rate": 7.337489218223164e-05, "loss": 2.9733, "step": 33955 }, { "epoch": 2.66, "learning_rate": 7.337410805300714e-05, "loss": 2.7069, "step": 33956 }, { "epoch": 2.66, "learning_rate": 7.337332392378264e-05, "loss": 2.6815, "step": 33957 }, { "epoch": 2.66, "learning_rate": 7.337253979455815e-05, "loss": 3.0474, "step": 33958 }, { "epoch": 2.66, "learning_rate": 7.337175566533365e-05, "loss": 2.8002, "step": 33959 }, { "epoch": 2.66, "learning_rate": 7.337097153610915e-05, "loss": 3.101, "step": 33960 }, { "epoch": 2.66, "learning_rate": 7.337018740688466e-05, "loss": 2.9338, "step": 33961 }, { "epoch": 2.66, "learning_rate": 7.336940327766016e-05, "loss": 2.8023, "step": 33962 }, { "epoch": 2.66, "learning_rate": 7.336861914843567e-05, "loss": 2.6673, "step": 33963 }, { "epoch": 2.66, "learning_rate": 7.336783501921117e-05, "loss": 3.0206, "step": 33964 }, { "epoch": 2.66, "learning_rate": 7.336705088998667e-05, "loss": 2.9119, "step": 33965 }, { "epoch": 2.66, "learning_rate": 7.336626676076218e-05, "loss": 2.5655, "step": 33966 }, { "epoch": 2.66, "learning_rate": 7.336548263153768e-05, "loss": 2.9873, "step": 33967 }, { "epoch": 2.66, "learning_rate": 7.336469850231318e-05, "loss": 2.9861, "step": 33968 }, { "epoch": 2.66, "learning_rate": 7.336391437308869e-05, "loss": 2.6575, "step": 33969 }, { "epoch": 2.66, "learning_rate": 7.336313024386419e-05, "loss": 2.9593, "step": 33970 }, { "epoch": 2.66, "learning_rate": 7.33623461146397e-05, "loss": 2.9275, "step": 33971 }, { "epoch": 2.66, "learning_rate": 7.33615619854152e-05, "loss": 2.679, "step": 33972 }, { "epoch": 2.66, "learning_rate": 7.33607778561907e-05, "loss": 2.8504, "step": 33973 }, { "epoch": 2.66, "learning_rate": 7.33599937269662e-05, "loss": 2.6578, "step": 33974 }, { "epoch": 2.66, "learning_rate": 7.335920959774171e-05, "loss": 2.6037, "step": 33975 }, { "epoch": 2.66, "learning_rate": 7.335842546851721e-05, "loss": 2.9595, "step": 33976 }, { "epoch": 2.66, "learning_rate": 7.335764133929272e-05, "loss": 2.8268, "step": 33977 }, { "epoch": 2.66, "learning_rate": 7.335685721006822e-05, "loss": 2.8635, "step": 33978 }, { "epoch": 2.66, "learning_rate": 7.335607308084372e-05, "loss": 2.9638, "step": 33979 }, { "epoch": 2.66, "learning_rate": 7.335528895161923e-05, "loss": 2.8464, "step": 33980 }, { "epoch": 2.66, "learning_rate": 7.335450482239473e-05, "loss": 2.6243, "step": 33981 }, { "epoch": 2.66, "learning_rate": 7.335372069317024e-05, "loss": 2.86, "step": 33982 }, { "epoch": 2.66, "learning_rate": 7.335293656394574e-05, "loss": 2.7034, "step": 33983 }, { "epoch": 2.66, "learning_rate": 7.335215243472124e-05, "loss": 2.927, "step": 33984 }, { "epoch": 2.66, "learning_rate": 7.335136830549675e-05, "loss": 2.9435, "step": 33985 }, { "epoch": 2.66, "learning_rate": 7.335058417627225e-05, "loss": 2.9202, "step": 33986 }, { "epoch": 2.67, "learning_rate": 7.334980004704775e-05, "loss": 2.8162, "step": 33987 }, { "epoch": 2.67, "learning_rate": 7.334901591782326e-05, "loss": 3.0265, "step": 33988 }, { "epoch": 2.67, "learning_rate": 7.334823178859876e-05, "loss": 2.7542, "step": 33989 }, { "epoch": 2.67, "learning_rate": 7.334744765937427e-05, "loss": 2.6255, "step": 33990 }, { "epoch": 2.67, "learning_rate": 7.334666353014977e-05, "loss": 2.8009, "step": 33991 }, { "epoch": 2.67, "learning_rate": 7.334587940092527e-05, "loss": 2.7528, "step": 33992 }, { "epoch": 2.67, "learning_rate": 7.334509527170078e-05, "loss": 2.845, "step": 33993 }, { "epoch": 2.67, "learning_rate": 7.334431114247628e-05, "loss": 2.8223, "step": 33994 }, { "epoch": 2.67, "learning_rate": 7.334352701325178e-05, "loss": 2.7939, "step": 33995 }, { "epoch": 2.67, "learning_rate": 7.334274288402729e-05, "loss": 2.8887, "step": 33996 }, { "epoch": 2.67, "learning_rate": 7.334195875480279e-05, "loss": 2.7956, "step": 33997 }, { "epoch": 2.67, "learning_rate": 7.334117462557831e-05, "loss": 2.7076, "step": 33998 }, { "epoch": 2.67, "learning_rate": 7.334039049635381e-05, "loss": 2.6263, "step": 33999 }, { "epoch": 2.67, "learning_rate": 7.333960636712932e-05, "loss": 3.0262, "step": 34000 }, { "epoch": 2.67, "eval_loss": 2.8956542015075684, "eval_runtime": 217.6907, "eval_samples_per_second": 52.056, "eval_steps_per_second": 3.257, "step": 34000 }, { "epoch": 2.67, "learning_rate": 7.333882223790482e-05, "loss": 2.7091, "step": 34001 }, { "epoch": 2.67, "learning_rate": 7.333803810868032e-05, "loss": 2.8022, "step": 34002 }, { "epoch": 2.67, "learning_rate": 7.333725397945583e-05, "loss": 2.7021, "step": 34003 }, { "epoch": 2.67, "learning_rate": 7.333646985023133e-05, "loss": 2.6677, "step": 34004 }, { "epoch": 2.67, "learning_rate": 7.333568572100683e-05, "loss": 2.9181, "step": 34005 }, { "epoch": 2.67, "learning_rate": 7.333490159178234e-05, "loss": 2.6198, "step": 34006 }, { "epoch": 2.67, "learning_rate": 7.333411746255783e-05, "loss": 2.9388, "step": 34007 }, { "epoch": 2.67, "learning_rate": 7.333333333333333e-05, "loss": 2.7494, "step": 34008 }, { "epoch": 2.67, "learning_rate": 7.333254920410884e-05, "loss": 2.7858, "step": 34009 }, { "epoch": 2.67, "learning_rate": 7.333176507488434e-05, "loss": 2.862, "step": 34010 }, { "epoch": 2.67, "learning_rate": 7.333098094565984e-05, "loss": 2.7742, "step": 34011 }, { "epoch": 2.67, "learning_rate": 7.333019681643535e-05, "loss": 3.0347, "step": 34012 }, { "epoch": 2.67, "learning_rate": 7.332941268721085e-05, "loss": 2.9035, "step": 34013 }, { "epoch": 2.67, "learning_rate": 7.332862855798635e-05, "loss": 2.6347, "step": 34014 }, { "epoch": 2.67, "learning_rate": 7.332784442876186e-05, "loss": 2.6642, "step": 34015 }, { "epoch": 2.67, "learning_rate": 7.332706029953736e-05, "loss": 2.8167, "step": 34016 }, { "epoch": 2.67, "learning_rate": 7.332627617031286e-05, "loss": 2.9214, "step": 34017 }, { "epoch": 2.67, "learning_rate": 7.332549204108837e-05, "loss": 2.7539, "step": 34018 }, { "epoch": 2.67, "learning_rate": 7.332470791186387e-05, "loss": 2.9248, "step": 34019 }, { "epoch": 2.67, "learning_rate": 7.332392378263938e-05, "loss": 2.7153, "step": 34020 }, { "epoch": 2.67, "learning_rate": 7.332313965341488e-05, "loss": 2.6429, "step": 34021 }, { "epoch": 2.67, "learning_rate": 7.332235552419038e-05, "loss": 2.7875, "step": 34022 }, { "epoch": 2.67, "learning_rate": 7.332157139496589e-05, "loss": 2.6879, "step": 34023 }, { "epoch": 2.67, "learning_rate": 7.33207872657414e-05, "loss": 2.7831, "step": 34024 }, { "epoch": 2.67, "learning_rate": 7.332000313651691e-05, "loss": 3.1188, "step": 34025 }, { "epoch": 2.67, "learning_rate": 7.331921900729241e-05, "loss": 2.695, "step": 34026 }, { "epoch": 2.67, "learning_rate": 7.331843487806791e-05, "loss": 2.8243, "step": 34027 }, { "epoch": 2.67, "learning_rate": 7.331765074884342e-05, "loss": 2.6932, "step": 34028 }, { "epoch": 2.67, "learning_rate": 7.331686661961892e-05, "loss": 2.7349, "step": 34029 }, { "epoch": 2.67, "learning_rate": 7.331608249039443e-05, "loss": 2.6928, "step": 34030 }, { "epoch": 2.67, "learning_rate": 7.331529836116993e-05, "loss": 2.8976, "step": 34031 }, { "epoch": 2.67, "learning_rate": 7.331451423194543e-05, "loss": 2.5734, "step": 34032 }, { "epoch": 2.67, "learning_rate": 7.331373010272094e-05, "loss": 2.5636, "step": 34033 }, { "epoch": 2.67, "learning_rate": 7.331294597349644e-05, "loss": 2.9463, "step": 34034 }, { "epoch": 2.67, "learning_rate": 7.331216184427194e-05, "loss": 2.7721, "step": 34035 }, { "epoch": 2.67, "learning_rate": 7.331137771504745e-05, "loss": 2.8731, "step": 34036 }, { "epoch": 2.67, "learning_rate": 7.331059358582295e-05, "loss": 2.5254, "step": 34037 }, { "epoch": 2.67, "learning_rate": 7.330980945659846e-05, "loss": 2.8083, "step": 34038 }, { "epoch": 2.67, "learning_rate": 7.330902532737396e-05, "loss": 2.9469, "step": 34039 }, { "epoch": 2.67, "learning_rate": 7.330824119814946e-05, "loss": 2.6073, "step": 34040 }, { "epoch": 2.67, "learning_rate": 7.330745706892497e-05, "loss": 2.7888, "step": 34041 }, { "epoch": 2.67, "learning_rate": 7.330667293970047e-05, "loss": 2.6953, "step": 34042 }, { "epoch": 2.67, "learning_rate": 7.330588881047596e-05, "loss": 2.9721, "step": 34043 }, { "epoch": 2.67, "learning_rate": 7.330510468125146e-05, "loss": 2.777, "step": 34044 }, { "epoch": 2.67, "learning_rate": 7.330432055202697e-05, "loss": 2.8747, "step": 34045 }, { "epoch": 2.67, "learning_rate": 7.330353642280247e-05, "loss": 2.7236, "step": 34046 }, { "epoch": 2.67, "learning_rate": 7.330275229357797e-05, "loss": 2.6836, "step": 34047 }, { "epoch": 2.67, "learning_rate": 7.330196816435348e-05, "loss": 2.7579, "step": 34048 }, { "epoch": 2.67, "learning_rate": 7.3301184035129e-05, "loss": 2.7777, "step": 34049 }, { "epoch": 2.67, "learning_rate": 7.33003999059045e-05, "loss": 2.7299, "step": 34050 }, { "epoch": 2.67, "learning_rate": 7.329961577668e-05, "loss": 2.9943, "step": 34051 }, { "epoch": 2.67, "learning_rate": 7.329883164745551e-05, "loss": 2.6694, "step": 34052 }, { "epoch": 2.67, "learning_rate": 7.329804751823101e-05, "loss": 2.7584, "step": 34053 }, { "epoch": 2.67, "learning_rate": 7.329726338900651e-05, "loss": 2.9909, "step": 34054 }, { "epoch": 2.67, "learning_rate": 7.329647925978202e-05, "loss": 2.8899, "step": 34055 }, { "epoch": 2.67, "learning_rate": 7.329569513055752e-05, "loss": 2.8313, "step": 34056 }, { "epoch": 2.67, "learning_rate": 7.329491100133303e-05, "loss": 2.7957, "step": 34057 }, { "epoch": 2.67, "learning_rate": 7.329412687210853e-05, "loss": 2.9699, "step": 34058 }, { "epoch": 2.67, "learning_rate": 7.329334274288403e-05, "loss": 2.8325, "step": 34059 }, { "epoch": 2.67, "learning_rate": 7.329255861365954e-05, "loss": 3.0926, "step": 34060 }, { "epoch": 2.67, "learning_rate": 7.329177448443504e-05, "loss": 2.7396, "step": 34061 }, { "epoch": 2.67, "learning_rate": 7.329099035521054e-05, "loss": 2.9417, "step": 34062 }, { "epoch": 2.67, "learning_rate": 7.329020622598605e-05, "loss": 2.646, "step": 34063 }, { "epoch": 2.67, "learning_rate": 7.328942209676155e-05, "loss": 2.8946, "step": 34064 }, { "epoch": 2.67, "learning_rate": 7.328863796753705e-05, "loss": 2.9738, "step": 34065 }, { "epoch": 2.67, "learning_rate": 7.328785383831256e-05, "loss": 3.0073, "step": 34066 }, { "epoch": 2.67, "learning_rate": 7.328706970908806e-05, "loss": 2.8517, "step": 34067 }, { "epoch": 2.67, "learning_rate": 7.328628557986357e-05, "loss": 3.0641, "step": 34068 }, { "epoch": 2.67, "learning_rate": 7.328550145063907e-05, "loss": 3.0016, "step": 34069 }, { "epoch": 2.67, "learning_rate": 7.328471732141457e-05, "loss": 2.8643, "step": 34070 }, { "epoch": 2.67, "learning_rate": 7.328393319219008e-05, "loss": 2.6273, "step": 34071 }, { "epoch": 2.67, "learning_rate": 7.328314906296558e-05, "loss": 2.87, "step": 34072 }, { "epoch": 2.67, "learning_rate": 7.328236493374108e-05, "loss": 2.9879, "step": 34073 }, { "epoch": 2.67, "learning_rate": 7.328158080451659e-05, "loss": 2.836, "step": 34074 }, { "epoch": 2.67, "learning_rate": 7.328079667529209e-05, "loss": 3.0423, "step": 34075 }, { "epoch": 2.67, "learning_rate": 7.32800125460676e-05, "loss": 2.9137, "step": 34076 }, { "epoch": 2.67, "learning_rate": 7.32792284168431e-05, "loss": 3.08, "step": 34077 }, { "epoch": 2.67, "learning_rate": 7.32784442876186e-05, "loss": 2.8252, "step": 34078 }, { "epoch": 2.67, "learning_rate": 7.32776601583941e-05, "loss": 2.7963, "step": 34079 }, { "epoch": 2.67, "learning_rate": 7.327687602916961e-05, "loss": 2.8757, "step": 34080 }, { "epoch": 2.67, "learning_rate": 7.327609189994511e-05, "loss": 2.7085, "step": 34081 }, { "epoch": 2.67, "learning_rate": 7.327530777072062e-05, "loss": 2.8786, "step": 34082 }, { "epoch": 2.67, "learning_rate": 7.327452364149612e-05, "loss": 2.6504, "step": 34083 }, { "epoch": 2.67, "learning_rate": 7.327373951227162e-05, "loss": 2.9646, "step": 34084 }, { "epoch": 2.67, "learning_rate": 7.327295538304713e-05, "loss": 3.0319, "step": 34085 }, { "epoch": 2.67, "learning_rate": 7.327217125382263e-05, "loss": 3.0152, "step": 34086 }, { "epoch": 2.67, "learning_rate": 7.327138712459814e-05, "loss": 2.8399, "step": 34087 }, { "epoch": 2.67, "learning_rate": 7.327060299537364e-05, "loss": 2.9741, "step": 34088 }, { "epoch": 2.67, "learning_rate": 7.326981886614914e-05, "loss": 2.7869, "step": 34089 }, { "epoch": 2.67, "learning_rate": 7.326903473692465e-05, "loss": 2.8585, "step": 34090 }, { "epoch": 2.67, "learning_rate": 7.326825060770015e-05, "loss": 2.812, "step": 34091 }, { "epoch": 2.67, "learning_rate": 7.326746647847565e-05, "loss": 2.6879, "step": 34092 }, { "epoch": 2.67, "learning_rate": 7.326668234925116e-05, "loss": 2.7857, "step": 34093 }, { "epoch": 2.67, "learning_rate": 7.326589822002666e-05, "loss": 2.8559, "step": 34094 }, { "epoch": 2.67, "learning_rate": 7.326511409080216e-05, "loss": 2.9528, "step": 34095 }, { "epoch": 2.67, "learning_rate": 7.326432996157767e-05, "loss": 2.7745, "step": 34096 }, { "epoch": 2.67, "learning_rate": 7.326354583235317e-05, "loss": 2.5665, "step": 34097 }, { "epoch": 2.67, "learning_rate": 7.326276170312868e-05, "loss": 2.8992, "step": 34098 }, { "epoch": 2.67, "learning_rate": 7.326197757390418e-05, "loss": 2.7876, "step": 34099 }, { "epoch": 2.67, "learning_rate": 7.326119344467968e-05, "loss": 2.8165, "step": 34100 }, { "epoch": 2.67, "learning_rate": 7.326040931545519e-05, "loss": 2.7766, "step": 34101 }, { "epoch": 2.67, "learning_rate": 7.325962518623069e-05, "loss": 2.9978, "step": 34102 }, { "epoch": 2.67, "learning_rate": 7.32588410570062e-05, "loss": 2.8519, "step": 34103 }, { "epoch": 2.67, "learning_rate": 7.32580569277817e-05, "loss": 2.5082, "step": 34104 }, { "epoch": 2.67, "learning_rate": 7.32572727985572e-05, "loss": 2.6554, "step": 34105 }, { "epoch": 2.67, "learning_rate": 7.325648866933272e-05, "loss": 2.7487, "step": 34106 }, { "epoch": 2.67, "learning_rate": 7.325570454010822e-05, "loss": 2.945, "step": 34107 }, { "epoch": 2.67, "learning_rate": 7.325492041088373e-05, "loss": 2.9679, "step": 34108 }, { "epoch": 2.67, "learning_rate": 7.325413628165923e-05, "loss": 2.7308, "step": 34109 }, { "epoch": 2.67, "learning_rate": 7.325335215243473e-05, "loss": 2.9147, "step": 34110 }, { "epoch": 2.67, "learning_rate": 7.325256802321024e-05, "loss": 2.864, "step": 34111 }, { "epoch": 2.67, "learning_rate": 7.325178389398574e-05, "loss": 2.7588, "step": 34112 }, { "epoch": 2.67, "learning_rate": 7.325099976476124e-05, "loss": 3.2041, "step": 34113 }, { "epoch": 2.67, "learning_rate": 7.325021563553675e-05, "loss": 2.6628, "step": 34114 }, { "epoch": 2.68, "learning_rate": 7.324943150631224e-05, "loss": 2.7377, "step": 34115 }, { "epoch": 2.68, "learning_rate": 7.324864737708774e-05, "loss": 2.9461, "step": 34116 }, { "epoch": 2.68, "learning_rate": 7.324786324786325e-05, "loss": 3.016, "step": 34117 }, { "epoch": 2.68, "learning_rate": 7.324707911863875e-05, "loss": 2.7744, "step": 34118 }, { "epoch": 2.68, "learning_rate": 7.324629498941425e-05, "loss": 2.7723, "step": 34119 }, { "epoch": 2.68, "learning_rate": 7.324551086018976e-05, "loss": 2.7125, "step": 34120 }, { "epoch": 2.68, "learning_rate": 7.324472673096526e-05, "loss": 2.7832, "step": 34121 }, { "epoch": 2.68, "learning_rate": 7.324394260174076e-05, "loss": 2.9217, "step": 34122 }, { "epoch": 2.68, "learning_rate": 7.324315847251627e-05, "loss": 2.7744, "step": 34123 }, { "epoch": 2.68, "learning_rate": 7.324237434329177e-05, "loss": 3.045, "step": 34124 }, { "epoch": 2.68, "learning_rate": 7.324159021406728e-05, "loss": 2.6693, "step": 34125 }, { "epoch": 2.68, "learning_rate": 7.324080608484278e-05, "loss": 2.7054, "step": 34126 }, { "epoch": 2.68, "learning_rate": 7.324002195561828e-05, "loss": 3.0214, "step": 34127 }, { "epoch": 2.68, "learning_rate": 7.323923782639379e-05, "loss": 2.9372, "step": 34128 }, { "epoch": 2.68, "learning_rate": 7.323845369716929e-05, "loss": 2.835, "step": 34129 }, { "epoch": 2.68, "learning_rate": 7.32376695679448e-05, "loss": 3.0725, "step": 34130 }, { "epoch": 2.68, "learning_rate": 7.32368854387203e-05, "loss": 2.7249, "step": 34131 }, { "epoch": 2.68, "learning_rate": 7.323610130949581e-05, "loss": 2.883, "step": 34132 }, { "epoch": 2.68, "learning_rate": 7.323531718027132e-05, "loss": 2.7711, "step": 34133 }, { "epoch": 2.68, "learning_rate": 7.323453305104682e-05, "loss": 2.7859, "step": 34134 }, { "epoch": 2.68, "learning_rate": 7.323374892182233e-05, "loss": 3.0144, "step": 34135 }, { "epoch": 2.68, "learning_rate": 7.323296479259783e-05, "loss": 2.7467, "step": 34136 }, { "epoch": 2.68, "learning_rate": 7.323218066337333e-05, "loss": 2.9683, "step": 34137 }, { "epoch": 2.68, "learning_rate": 7.323139653414884e-05, "loss": 2.6718, "step": 34138 }, { "epoch": 2.68, "learning_rate": 7.323061240492434e-05, "loss": 2.7507, "step": 34139 }, { "epoch": 2.68, "learning_rate": 7.322982827569984e-05, "loss": 3.0457, "step": 34140 }, { "epoch": 2.68, "learning_rate": 7.322904414647535e-05, "loss": 2.8933, "step": 34141 }, { "epoch": 2.68, "learning_rate": 7.322826001725085e-05, "loss": 2.7041, "step": 34142 }, { "epoch": 2.68, "learning_rate": 7.322747588802636e-05, "loss": 2.6356, "step": 34143 }, { "epoch": 2.68, "learning_rate": 7.322669175880186e-05, "loss": 2.7561, "step": 34144 }, { "epoch": 2.68, "learning_rate": 7.322590762957736e-05, "loss": 2.6425, "step": 34145 }, { "epoch": 2.68, "learning_rate": 7.322512350035287e-05, "loss": 2.6845, "step": 34146 }, { "epoch": 2.68, "learning_rate": 7.322433937112837e-05, "loss": 2.7031, "step": 34147 }, { "epoch": 2.68, "learning_rate": 7.322355524190387e-05, "loss": 2.9401, "step": 34148 }, { "epoch": 2.68, "learning_rate": 7.322277111267938e-05, "loss": 2.9206, "step": 34149 }, { "epoch": 2.68, "learning_rate": 7.322198698345488e-05, "loss": 2.7985, "step": 34150 }, { "epoch": 2.68, "learning_rate": 7.322120285423037e-05, "loss": 2.8331, "step": 34151 }, { "epoch": 2.68, "learning_rate": 7.322041872500587e-05, "loss": 2.7643, "step": 34152 }, { "epoch": 2.68, "learning_rate": 7.321963459578138e-05, "loss": 3.0536, "step": 34153 }, { "epoch": 2.68, "learning_rate": 7.321885046655688e-05, "loss": 2.8557, "step": 34154 }, { "epoch": 2.68, "learning_rate": 7.321806633733239e-05, "loss": 2.8486, "step": 34155 }, { "epoch": 2.68, "learning_rate": 7.321728220810789e-05, "loss": 2.7416, "step": 34156 }, { "epoch": 2.68, "learning_rate": 7.32164980788834e-05, "loss": 2.9989, "step": 34157 }, { "epoch": 2.68, "learning_rate": 7.321571394965891e-05, "loss": 2.6385, "step": 34158 }, { "epoch": 2.68, "learning_rate": 7.321492982043441e-05, "loss": 2.828, "step": 34159 }, { "epoch": 2.68, "learning_rate": 7.321414569120992e-05, "loss": 2.7047, "step": 34160 }, { "epoch": 2.68, "learning_rate": 7.321336156198542e-05, "loss": 2.796, "step": 34161 }, { "epoch": 2.68, "learning_rate": 7.321257743276092e-05, "loss": 2.732, "step": 34162 }, { "epoch": 2.68, "learning_rate": 7.321179330353643e-05, "loss": 2.7746, "step": 34163 }, { "epoch": 2.68, "learning_rate": 7.321100917431193e-05, "loss": 2.8059, "step": 34164 }, { "epoch": 2.68, "learning_rate": 7.321022504508744e-05, "loss": 2.7754, "step": 34165 }, { "epoch": 2.68, "learning_rate": 7.320944091586294e-05, "loss": 2.7523, "step": 34166 }, { "epoch": 2.68, "learning_rate": 7.320865678663844e-05, "loss": 2.5219, "step": 34167 }, { "epoch": 2.68, "learning_rate": 7.320787265741395e-05, "loss": 2.7807, "step": 34168 }, { "epoch": 2.68, "learning_rate": 7.320708852818945e-05, "loss": 2.9666, "step": 34169 }, { "epoch": 2.68, "learning_rate": 7.320630439896495e-05, "loss": 2.7575, "step": 34170 }, { "epoch": 2.68, "learning_rate": 7.320552026974046e-05, "loss": 2.7652, "step": 34171 }, { "epoch": 2.68, "learning_rate": 7.320473614051596e-05, "loss": 2.7268, "step": 34172 }, { "epoch": 2.68, "learning_rate": 7.320395201129147e-05, "loss": 2.8521, "step": 34173 }, { "epoch": 2.68, "learning_rate": 7.320316788206697e-05, "loss": 2.8735, "step": 34174 }, { "epoch": 2.68, "learning_rate": 7.320238375284247e-05, "loss": 2.5313, "step": 34175 }, { "epoch": 2.68, "learning_rate": 7.320159962361798e-05, "loss": 2.9063, "step": 34176 }, { "epoch": 2.68, "learning_rate": 7.320081549439348e-05, "loss": 2.8078, "step": 34177 }, { "epoch": 2.68, "learning_rate": 7.320003136516898e-05, "loss": 2.8522, "step": 34178 }, { "epoch": 2.68, "learning_rate": 7.319924723594449e-05, "loss": 2.9682, "step": 34179 }, { "epoch": 2.68, "learning_rate": 7.319846310671999e-05, "loss": 2.6593, "step": 34180 }, { "epoch": 2.68, "learning_rate": 7.31976789774955e-05, "loss": 2.8413, "step": 34181 }, { "epoch": 2.68, "learning_rate": 7.3196894848271e-05, "loss": 2.7932, "step": 34182 }, { "epoch": 2.68, "learning_rate": 7.31961107190465e-05, "loss": 2.8139, "step": 34183 }, { "epoch": 2.68, "learning_rate": 7.3195326589822e-05, "loss": 2.7621, "step": 34184 }, { "epoch": 2.68, "learning_rate": 7.319454246059751e-05, "loss": 2.7838, "step": 34185 }, { "epoch": 2.68, "learning_rate": 7.319375833137301e-05, "loss": 2.6749, "step": 34186 }, { "epoch": 2.68, "learning_rate": 7.319297420214852e-05, "loss": 2.6831, "step": 34187 }, { "epoch": 2.68, "learning_rate": 7.319219007292402e-05, "loss": 2.8365, "step": 34188 }, { "epoch": 2.68, "learning_rate": 7.319140594369952e-05, "loss": 2.8289, "step": 34189 }, { "epoch": 2.68, "learning_rate": 7.319062181447503e-05, "loss": 2.7988, "step": 34190 }, { "epoch": 2.68, "learning_rate": 7.318983768525053e-05, "loss": 2.5625, "step": 34191 }, { "epoch": 2.68, "learning_rate": 7.318905355602604e-05, "loss": 2.924, "step": 34192 }, { "epoch": 2.68, "learning_rate": 7.318826942680154e-05, "loss": 2.9715, "step": 34193 }, { "epoch": 2.68, "learning_rate": 7.318748529757704e-05, "loss": 2.6501, "step": 34194 }, { "epoch": 2.68, "learning_rate": 7.318670116835255e-05, "loss": 2.769, "step": 34195 }, { "epoch": 2.68, "learning_rate": 7.318591703912805e-05, "loss": 2.6902, "step": 34196 }, { "epoch": 2.68, "learning_rate": 7.318513290990355e-05, "loss": 2.9227, "step": 34197 }, { "epoch": 2.68, "learning_rate": 7.318434878067906e-05, "loss": 2.7291, "step": 34198 }, { "epoch": 2.68, "learning_rate": 7.318356465145456e-05, "loss": 2.8509, "step": 34199 }, { "epoch": 2.68, "learning_rate": 7.318278052223006e-05, "loss": 2.8284, "step": 34200 }, { "epoch": 2.68, "learning_rate": 7.318199639300557e-05, "loss": 2.782, "step": 34201 }, { "epoch": 2.68, "learning_rate": 7.318121226378107e-05, "loss": 2.7191, "step": 34202 }, { "epoch": 2.68, "learning_rate": 7.318042813455658e-05, "loss": 2.918, "step": 34203 }, { "epoch": 2.68, "learning_rate": 7.317964400533208e-05, "loss": 2.6502, "step": 34204 }, { "epoch": 2.68, "learning_rate": 7.317885987610758e-05, "loss": 2.7428, "step": 34205 }, { "epoch": 2.68, "learning_rate": 7.317807574688309e-05, "loss": 2.7662, "step": 34206 }, { "epoch": 2.68, "learning_rate": 7.317729161765859e-05, "loss": 2.7498, "step": 34207 }, { "epoch": 2.68, "learning_rate": 7.31765074884341e-05, "loss": 2.7924, "step": 34208 }, { "epoch": 2.68, "learning_rate": 7.31757233592096e-05, "loss": 2.4487, "step": 34209 }, { "epoch": 2.68, "learning_rate": 7.31749392299851e-05, "loss": 3.1162, "step": 34210 }, { "epoch": 2.68, "learning_rate": 7.31741551007606e-05, "loss": 2.8138, "step": 34211 }, { "epoch": 2.68, "learning_rate": 7.317337097153611e-05, "loss": 2.8106, "step": 34212 }, { "epoch": 2.68, "learning_rate": 7.317258684231161e-05, "loss": 2.8989, "step": 34213 }, { "epoch": 2.68, "learning_rate": 7.317180271308713e-05, "loss": 2.677, "step": 34214 }, { "epoch": 2.68, "learning_rate": 7.317101858386263e-05, "loss": 2.9405, "step": 34215 }, { "epoch": 2.68, "learning_rate": 7.317023445463814e-05, "loss": 2.7492, "step": 34216 }, { "epoch": 2.68, "learning_rate": 7.316945032541364e-05, "loss": 2.8757, "step": 34217 }, { "epoch": 2.68, "learning_rate": 7.316866619618914e-05, "loss": 2.7508, "step": 34218 }, { "epoch": 2.68, "learning_rate": 7.316788206696465e-05, "loss": 2.8507, "step": 34219 }, { "epoch": 2.68, "learning_rate": 7.316709793774015e-05, "loss": 2.8909, "step": 34220 }, { "epoch": 2.68, "learning_rate": 7.316631380851566e-05, "loss": 2.887, "step": 34221 }, { "epoch": 2.68, "learning_rate": 7.316552967929116e-05, "loss": 2.8979, "step": 34222 }, { "epoch": 2.68, "learning_rate": 7.316474555006665e-05, "loss": 2.9081, "step": 34223 }, { "epoch": 2.68, "learning_rate": 7.316396142084215e-05, "loss": 2.7154, "step": 34224 }, { "epoch": 2.68, "learning_rate": 7.316317729161766e-05, "loss": 2.8136, "step": 34225 }, { "epoch": 2.68, "learning_rate": 7.316239316239316e-05, "loss": 2.6772, "step": 34226 }, { "epoch": 2.68, "learning_rate": 7.316160903316866e-05, "loss": 2.7118, "step": 34227 }, { "epoch": 2.68, "learning_rate": 7.316082490394417e-05, "loss": 2.7013, "step": 34228 }, { "epoch": 2.68, "learning_rate": 7.316004077471967e-05, "loss": 2.7807, "step": 34229 }, { "epoch": 2.68, "learning_rate": 7.315925664549517e-05, "loss": 2.6675, "step": 34230 }, { "epoch": 2.68, "learning_rate": 7.315847251627068e-05, "loss": 2.8981, "step": 34231 }, { "epoch": 2.68, "learning_rate": 7.315768838704618e-05, "loss": 2.8575, "step": 34232 }, { "epoch": 2.68, "learning_rate": 7.315690425782169e-05, "loss": 2.9369, "step": 34233 }, { "epoch": 2.68, "learning_rate": 7.315612012859719e-05, "loss": 2.7972, "step": 34234 }, { "epoch": 2.68, "learning_rate": 7.31553359993727e-05, "loss": 2.8795, "step": 34235 }, { "epoch": 2.68, "learning_rate": 7.31545518701482e-05, "loss": 2.8479, "step": 34236 }, { "epoch": 2.68, "learning_rate": 7.31537677409237e-05, "loss": 2.7925, "step": 34237 }, { "epoch": 2.68, "learning_rate": 7.31529836116992e-05, "loss": 2.5942, "step": 34238 }, { "epoch": 2.68, "learning_rate": 7.315219948247471e-05, "loss": 2.8729, "step": 34239 }, { "epoch": 2.68, "learning_rate": 7.315141535325023e-05, "loss": 2.9827, "step": 34240 }, { "epoch": 2.68, "learning_rate": 7.315063122402573e-05, "loss": 2.7553, "step": 34241 }, { "epoch": 2.69, "learning_rate": 7.314984709480123e-05, "loss": 2.734, "step": 34242 }, { "epoch": 2.69, "learning_rate": 7.314906296557674e-05, "loss": 2.7684, "step": 34243 }, { "epoch": 2.69, "learning_rate": 7.314827883635224e-05, "loss": 2.6713, "step": 34244 }, { "epoch": 2.69, "learning_rate": 7.314749470712774e-05, "loss": 2.7721, "step": 34245 }, { "epoch": 2.69, "learning_rate": 7.314671057790325e-05, "loss": 2.7554, "step": 34246 }, { "epoch": 2.69, "learning_rate": 7.314592644867875e-05, "loss": 3.2635, "step": 34247 }, { "epoch": 2.69, "learning_rate": 7.314514231945425e-05, "loss": 2.7254, "step": 34248 }, { "epoch": 2.69, "learning_rate": 7.314435819022976e-05, "loss": 2.8038, "step": 34249 }, { "epoch": 2.69, "learning_rate": 7.314357406100526e-05, "loss": 2.7558, "step": 34250 }, { "epoch": 2.69, "learning_rate": 7.314278993178077e-05, "loss": 2.6426, "step": 34251 }, { "epoch": 2.69, "learning_rate": 7.314200580255627e-05, "loss": 2.7175, "step": 34252 }, { "epoch": 2.69, "learning_rate": 7.314122167333177e-05, "loss": 2.6695, "step": 34253 }, { "epoch": 2.69, "learning_rate": 7.314043754410728e-05, "loss": 2.7833, "step": 34254 }, { "epoch": 2.69, "learning_rate": 7.313965341488278e-05, "loss": 2.8028, "step": 34255 }, { "epoch": 2.69, "learning_rate": 7.313886928565828e-05, "loss": 2.8472, "step": 34256 }, { "epoch": 2.69, "learning_rate": 7.313808515643379e-05, "loss": 2.796, "step": 34257 }, { "epoch": 2.69, "learning_rate": 7.313730102720929e-05, "loss": 2.758, "step": 34258 }, { "epoch": 2.69, "learning_rate": 7.313651689798478e-05, "loss": 2.9551, "step": 34259 }, { "epoch": 2.69, "learning_rate": 7.313573276876029e-05, "loss": 3.0842, "step": 34260 }, { "epoch": 2.69, "learning_rate": 7.313494863953579e-05, "loss": 2.5934, "step": 34261 }, { "epoch": 2.69, "learning_rate": 7.313416451031129e-05, "loss": 2.7696, "step": 34262 }, { "epoch": 2.69, "learning_rate": 7.31333803810868e-05, "loss": 2.6448, "step": 34263 }, { "epoch": 2.69, "learning_rate": 7.31325962518623e-05, "loss": 2.7831, "step": 34264 }, { "epoch": 2.69, "learning_rate": 7.313181212263782e-05, "loss": 2.9156, "step": 34265 }, { "epoch": 2.69, "learning_rate": 7.313102799341332e-05, "loss": 3.0668, "step": 34266 }, { "epoch": 2.69, "learning_rate": 7.313024386418882e-05, "loss": 2.7222, "step": 34267 }, { "epoch": 2.69, "learning_rate": 7.312945973496433e-05, "loss": 2.7002, "step": 34268 }, { "epoch": 2.69, "learning_rate": 7.312867560573983e-05, "loss": 2.6457, "step": 34269 }, { "epoch": 2.69, "learning_rate": 7.312789147651534e-05, "loss": 3.0093, "step": 34270 }, { "epoch": 2.69, "learning_rate": 7.312710734729084e-05, "loss": 2.8138, "step": 34271 }, { "epoch": 2.69, "learning_rate": 7.312632321806634e-05, "loss": 2.7379, "step": 34272 }, { "epoch": 2.69, "learning_rate": 7.312553908884185e-05, "loss": 2.7836, "step": 34273 }, { "epoch": 2.69, "learning_rate": 7.312475495961735e-05, "loss": 2.8815, "step": 34274 }, { "epoch": 2.69, "learning_rate": 7.312397083039285e-05, "loss": 2.9445, "step": 34275 }, { "epoch": 2.69, "learning_rate": 7.312318670116836e-05, "loss": 2.9293, "step": 34276 }, { "epoch": 2.69, "learning_rate": 7.312240257194386e-05, "loss": 2.8667, "step": 34277 }, { "epoch": 2.69, "learning_rate": 7.312161844271937e-05, "loss": 2.9731, "step": 34278 }, { "epoch": 2.69, "learning_rate": 7.312083431349487e-05, "loss": 2.6298, "step": 34279 }, { "epoch": 2.69, "learning_rate": 7.312005018427037e-05, "loss": 3.018, "step": 34280 }, { "epoch": 2.69, "learning_rate": 7.311926605504588e-05, "loss": 2.977, "step": 34281 }, { "epoch": 2.69, "learning_rate": 7.311848192582138e-05, "loss": 2.8086, "step": 34282 }, { "epoch": 2.69, "learning_rate": 7.311769779659688e-05, "loss": 2.7989, "step": 34283 }, { "epoch": 2.69, "learning_rate": 7.311691366737239e-05, "loss": 2.579, "step": 34284 }, { "epoch": 2.69, "learning_rate": 7.311612953814789e-05, "loss": 2.7708, "step": 34285 }, { "epoch": 2.69, "learning_rate": 7.31153454089234e-05, "loss": 2.7809, "step": 34286 }, { "epoch": 2.69, "learning_rate": 7.31145612796989e-05, "loss": 2.749, "step": 34287 }, { "epoch": 2.69, "learning_rate": 7.31137771504744e-05, "loss": 2.7391, "step": 34288 }, { "epoch": 2.69, "learning_rate": 7.31129930212499e-05, "loss": 3.0044, "step": 34289 }, { "epoch": 2.69, "learning_rate": 7.311220889202541e-05, "loss": 2.7938, "step": 34290 }, { "epoch": 2.69, "learning_rate": 7.311142476280091e-05, "loss": 2.9305, "step": 34291 }, { "epoch": 2.69, "learning_rate": 7.311064063357642e-05, "loss": 2.6596, "step": 34292 }, { "epoch": 2.69, "learning_rate": 7.310985650435192e-05, "loss": 2.8823, "step": 34293 }, { "epoch": 2.69, "learning_rate": 7.310907237512742e-05, "loss": 2.6805, "step": 34294 }, { "epoch": 2.69, "learning_rate": 7.310828824590293e-05, "loss": 2.8206, "step": 34295 }, { "epoch": 2.69, "learning_rate": 7.310750411667843e-05, "loss": 2.8871, "step": 34296 }, { "epoch": 2.69, "learning_rate": 7.310671998745393e-05, "loss": 2.6982, "step": 34297 }, { "epoch": 2.69, "learning_rate": 7.310593585822944e-05, "loss": 2.9065, "step": 34298 }, { "epoch": 2.69, "learning_rate": 7.310515172900494e-05, "loss": 2.8223, "step": 34299 }, { "epoch": 2.69, "learning_rate": 7.310436759978045e-05, "loss": 2.6506, "step": 34300 }, { "epoch": 2.69, "learning_rate": 7.310358347055595e-05, "loss": 2.9909, "step": 34301 }, { "epoch": 2.69, "learning_rate": 7.310279934133145e-05, "loss": 2.9937, "step": 34302 }, { "epoch": 2.69, "learning_rate": 7.310201521210696e-05, "loss": 2.7228, "step": 34303 }, { "epoch": 2.69, "learning_rate": 7.310123108288246e-05, "loss": 2.7816, "step": 34304 }, { "epoch": 2.69, "learning_rate": 7.310044695365796e-05, "loss": 2.7174, "step": 34305 }, { "epoch": 2.69, "learning_rate": 7.309966282443347e-05, "loss": 2.8314, "step": 34306 }, { "epoch": 2.69, "learning_rate": 7.309887869520897e-05, "loss": 2.9313, "step": 34307 }, { "epoch": 2.69, "learning_rate": 7.309809456598448e-05, "loss": 2.7458, "step": 34308 }, { "epoch": 2.69, "learning_rate": 7.309731043675998e-05, "loss": 2.9309, "step": 34309 }, { "epoch": 2.69, "learning_rate": 7.309652630753548e-05, "loss": 3.0279, "step": 34310 }, { "epoch": 2.69, "learning_rate": 7.309574217831099e-05, "loss": 2.7715, "step": 34311 }, { "epoch": 2.69, "learning_rate": 7.309495804908649e-05, "loss": 2.8341, "step": 34312 }, { "epoch": 2.69, "learning_rate": 7.3094173919862e-05, "loss": 2.6942, "step": 34313 }, { "epoch": 2.69, "learning_rate": 7.30933897906375e-05, "loss": 2.7063, "step": 34314 }, { "epoch": 2.69, "learning_rate": 7.3092605661413e-05, "loss": 2.6399, "step": 34315 }, { "epoch": 2.69, "learning_rate": 7.30918215321885e-05, "loss": 2.8881, "step": 34316 }, { "epoch": 2.69, "learning_rate": 7.309103740296401e-05, "loss": 2.7697, "step": 34317 }, { "epoch": 2.69, "learning_rate": 7.309025327373951e-05, "loss": 2.7337, "step": 34318 }, { "epoch": 2.69, "learning_rate": 7.308946914451502e-05, "loss": 2.8001, "step": 34319 }, { "epoch": 2.69, "learning_rate": 7.308868501529052e-05, "loss": 2.8369, "step": 34320 }, { "epoch": 2.69, "learning_rate": 7.308790088606602e-05, "loss": 2.976, "step": 34321 }, { "epoch": 2.69, "learning_rate": 7.308711675684154e-05, "loss": 2.8089, "step": 34322 }, { "epoch": 2.69, "learning_rate": 7.308633262761704e-05, "loss": 3.0555, "step": 34323 }, { "epoch": 2.69, "learning_rate": 7.308554849839255e-05, "loss": 2.7142, "step": 34324 }, { "epoch": 2.69, "learning_rate": 7.308476436916805e-05, "loss": 2.7884, "step": 34325 }, { "epoch": 2.69, "learning_rate": 7.308398023994356e-05, "loss": 3.0003, "step": 34326 }, { "epoch": 2.69, "learning_rate": 7.308319611071906e-05, "loss": 2.9323, "step": 34327 }, { "epoch": 2.69, "learning_rate": 7.308241198149456e-05, "loss": 2.8218, "step": 34328 }, { "epoch": 2.69, "learning_rate": 7.308162785227007e-05, "loss": 2.9273, "step": 34329 }, { "epoch": 2.69, "learning_rate": 7.308084372304557e-05, "loss": 2.8881, "step": 34330 }, { "epoch": 2.69, "learning_rate": 7.308005959382106e-05, "loss": 2.6643, "step": 34331 }, { "epoch": 2.69, "learning_rate": 7.307927546459656e-05, "loss": 2.7691, "step": 34332 }, { "epoch": 2.69, "learning_rate": 7.307849133537207e-05, "loss": 2.7272, "step": 34333 }, { "epoch": 2.69, "learning_rate": 7.307770720614757e-05, "loss": 2.9378, "step": 34334 }, { "epoch": 2.69, "learning_rate": 7.307692307692307e-05, "loss": 2.8186, "step": 34335 }, { "epoch": 2.69, "learning_rate": 7.307613894769858e-05, "loss": 2.8699, "step": 34336 }, { "epoch": 2.69, "learning_rate": 7.307535481847408e-05, "loss": 2.696, "step": 34337 }, { "epoch": 2.69, "learning_rate": 7.307457068924959e-05, "loss": 3.0051, "step": 34338 }, { "epoch": 2.69, "learning_rate": 7.307378656002509e-05, "loss": 2.7661, "step": 34339 }, { "epoch": 2.69, "learning_rate": 7.307300243080059e-05, "loss": 2.6489, "step": 34340 }, { "epoch": 2.69, "learning_rate": 7.30722183015761e-05, "loss": 2.8779, "step": 34341 }, { "epoch": 2.69, "learning_rate": 7.30714341723516e-05, "loss": 2.9394, "step": 34342 }, { "epoch": 2.69, "learning_rate": 7.30706500431271e-05, "loss": 3.2087, "step": 34343 }, { "epoch": 2.69, "learning_rate": 7.306986591390261e-05, "loss": 2.8424, "step": 34344 }, { "epoch": 2.69, "learning_rate": 7.306908178467811e-05, "loss": 2.8461, "step": 34345 }, { "epoch": 2.69, "learning_rate": 7.306829765545362e-05, "loss": 2.7964, "step": 34346 }, { "epoch": 2.69, "learning_rate": 7.306751352622912e-05, "loss": 2.7946, "step": 34347 }, { "epoch": 2.69, "learning_rate": 7.306672939700464e-05, "loss": 2.7506, "step": 34348 }, { "epoch": 2.69, "learning_rate": 7.306594526778014e-05, "loss": 2.889, "step": 34349 }, { "epoch": 2.69, "learning_rate": 7.306516113855564e-05, "loss": 2.7199, "step": 34350 }, { "epoch": 2.69, "learning_rate": 7.306437700933115e-05, "loss": 2.9331, "step": 34351 }, { "epoch": 2.69, "learning_rate": 7.306359288010665e-05, "loss": 2.6301, "step": 34352 }, { "epoch": 2.69, "learning_rate": 7.306280875088215e-05, "loss": 2.7088, "step": 34353 }, { "epoch": 2.69, "learning_rate": 7.306202462165766e-05, "loss": 2.8021, "step": 34354 }, { "epoch": 2.69, "learning_rate": 7.306124049243316e-05, "loss": 2.8153, "step": 34355 }, { "epoch": 2.69, "learning_rate": 7.306045636320867e-05, "loss": 2.7484, "step": 34356 }, { "epoch": 2.69, "learning_rate": 7.305967223398417e-05, "loss": 2.876, "step": 34357 }, { "epoch": 2.69, "learning_rate": 7.305888810475967e-05, "loss": 2.8616, "step": 34358 }, { "epoch": 2.69, "learning_rate": 7.305810397553518e-05, "loss": 2.6312, "step": 34359 }, { "epoch": 2.69, "learning_rate": 7.305731984631068e-05, "loss": 2.8799, "step": 34360 }, { "epoch": 2.69, "learning_rate": 7.305653571708618e-05, "loss": 2.6008, "step": 34361 }, { "epoch": 2.69, "learning_rate": 7.305575158786169e-05, "loss": 2.7252, "step": 34362 }, { "epoch": 2.69, "learning_rate": 7.305496745863719e-05, "loss": 2.6843, "step": 34363 }, { "epoch": 2.69, "learning_rate": 7.30541833294127e-05, "loss": 2.6422, "step": 34364 }, { "epoch": 2.69, "learning_rate": 7.30533992001882e-05, "loss": 2.85, "step": 34365 }, { "epoch": 2.69, "learning_rate": 7.30526150709637e-05, "loss": 2.8125, "step": 34366 }, { "epoch": 2.69, "learning_rate": 7.305183094173919e-05, "loss": 3.0421, "step": 34367 }, { "epoch": 2.69, "learning_rate": 7.30510468125147e-05, "loss": 2.7632, "step": 34368 }, { "epoch": 2.69, "learning_rate": 7.30502626832902e-05, "loss": 2.7234, "step": 34369 }, { "epoch": 2.7, "learning_rate": 7.30494785540657e-05, "loss": 2.8028, "step": 34370 }, { "epoch": 2.7, "learning_rate": 7.304869442484121e-05, "loss": 2.8705, "step": 34371 }, { "epoch": 2.7, "learning_rate": 7.304791029561671e-05, "loss": 2.6766, "step": 34372 }, { "epoch": 2.7, "learning_rate": 7.304712616639223e-05, "loss": 2.6862, "step": 34373 }, { "epoch": 2.7, "learning_rate": 7.304634203716773e-05, "loss": 2.9615, "step": 34374 }, { "epoch": 2.7, "learning_rate": 7.304555790794324e-05, "loss": 2.8487, "step": 34375 }, { "epoch": 2.7, "learning_rate": 7.304477377871874e-05, "loss": 2.767, "step": 34376 }, { "epoch": 2.7, "learning_rate": 7.304398964949424e-05, "loss": 2.7764, "step": 34377 }, { "epoch": 2.7, "learning_rate": 7.304320552026975e-05, "loss": 2.8033, "step": 34378 }, { "epoch": 2.7, "learning_rate": 7.304242139104525e-05, "loss": 2.8894, "step": 34379 }, { "epoch": 2.7, "learning_rate": 7.304163726182075e-05, "loss": 2.8049, "step": 34380 }, { "epoch": 2.7, "learning_rate": 7.304085313259626e-05, "loss": 2.7967, "step": 34381 }, { "epoch": 2.7, "learning_rate": 7.304006900337176e-05, "loss": 2.9302, "step": 34382 }, { "epoch": 2.7, "learning_rate": 7.303928487414726e-05, "loss": 2.6754, "step": 34383 }, { "epoch": 2.7, "learning_rate": 7.303850074492277e-05, "loss": 2.8383, "step": 34384 }, { "epoch": 2.7, "learning_rate": 7.303771661569827e-05, "loss": 2.8269, "step": 34385 }, { "epoch": 2.7, "learning_rate": 7.303693248647378e-05, "loss": 2.7764, "step": 34386 }, { "epoch": 2.7, "learning_rate": 7.303614835724928e-05, "loss": 2.782, "step": 34387 }, { "epoch": 2.7, "learning_rate": 7.303536422802478e-05, "loss": 2.6622, "step": 34388 }, { "epoch": 2.7, "learning_rate": 7.303458009880029e-05, "loss": 2.7365, "step": 34389 }, { "epoch": 2.7, "learning_rate": 7.303379596957579e-05, "loss": 2.8236, "step": 34390 }, { "epoch": 2.7, "learning_rate": 7.30330118403513e-05, "loss": 2.6928, "step": 34391 }, { "epoch": 2.7, "learning_rate": 7.30322277111268e-05, "loss": 2.8995, "step": 34392 }, { "epoch": 2.7, "learning_rate": 7.30314435819023e-05, "loss": 2.7321, "step": 34393 }, { "epoch": 2.7, "learning_rate": 7.30306594526778e-05, "loss": 2.7826, "step": 34394 }, { "epoch": 2.7, "learning_rate": 7.302987532345331e-05, "loss": 2.8122, "step": 34395 }, { "epoch": 2.7, "learning_rate": 7.302909119422881e-05, "loss": 2.7324, "step": 34396 }, { "epoch": 2.7, "learning_rate": 7.302830706500432e-05, "loss": 2.814, "step": 34397 }, { "epoch": 2.7, "learning_rate": 7.302752293577982e-05, "loss": 2.7258, "step": 34398 }, { "epoch": 2.7, "learning_rate": 7.302673880655532e-05, "loss": 2.7956, "step": 34399 }, { "epoch": 2.7, "learning_rate": 7.302595467733083e-05, "loss": 2.9147, "step": 34400 }, { "epoch": 2.7, "learning_rate": 7.302517054810633e-05, "loss": 2.8604, "step": 34401 }, { "epoch": 2.7, "learning_rate": 7.302438641888183e-05, "loss": 2.8247, "step": 34402 }, { "epoch": 2.7, "learning_rate": 7.302360228965734e-05, "loss": 2.7123, "step": 34403 }, { "epoch": 2.7, "learning_rate": 7.302281816043284e-05, "loss": 2.7597, "step": 34404 }, { "epoch": 2.7, "learning_rate": 7.302203403120835e-05, "loss": 2.756, "step": 34405 }, { "epoch": 2.7, "learning_rate": 7.302124990198385e-05, "loss": 3.086, "step": 34406 }, { "epoch": 2.7, "learning_rate": 7.302046577275935e-05, "loss": 2.802, "step": 34407 }, { "epoch": 2.7, "learning_rate": 7.301968164353486e-05, "loss": 2.6505, "step": 34408 }, { "epoch": 2.7, "learning_rate": 7.301889751431036e-05, "loss": 2.6779, "step": 34409 }, { "epoch": 2.7, "learning_rate": 7.301811338508586e-05, "loss": 2.8389, "step": 34410 }, { "epoch": 2.7, "learning_rate": 7.301732925586137e-05, "loss": 2.6556, "step": 34411 }, { "epoch": 2.7, "learning_rate": 7.301654512663687e-05, "loss": 2.7942, "step": 34412 }, { "epoch": 2.7, "learning_rate": 7.301576099741238e-05, "loss": 2.8927, "step": 34413 }, { "epoch": 2.7, "learning_rate": 7.301497686818788e-05, "loss": 2.6958, "step": 34414 }, { "epoch": 2.7, "learning_rate": 7.301419273896338e-05, "loss": 2.7253, "step": 34415 }, { "epoch": 2.7, "learning_rate": 7.301340860973889e-05, "loss": 2.6709, "step": 34416 }, { "epoch": 2.7, "learning_rate": 7.301262448051439e-05, "loss": 2.7581, "step": 34417 }, { "epoch": 2.7, "learning_rate": 7.30118403512899e-05, "loss": 2.6913, "step": 34418 }, { "epoch": 2.7, "learning_rate": 7.30110562220654e-05, "loss": 2.737, "step": 34419 }, { "epoch": 2.7, "learning_rate": 7.30102720928409e-05, "loss": 2.4938, "step": 34420 }, { "epoch": 2.7, "learning_rate": 7.30094879636164e-05, "loss": 2.9428, "step": 34421 }, { "epoch": 2.7, "learning_rate": 7.300870383439191e-05, "loss": 2.8552, "step": 34422 }, { "epoch": 2.7, "learning_rate": 7.300791970516741e-05, "loss": 2.8721, "step": 34423 }, { "epoch": 2.7, "learning_rate": 7.300713557594292e-05, "loss": 2.7995, "step": 34424 }, { "epoch": 2.7, "learning_rate": 7.300635144671842e-05, "loss": 2.9527, "step": 34425 }, { "epoch": 2.7, "learning_rate": 7.300556731749392e-05, "loss": 2.8156, "step": 34426 }, { "epoch": 2.7, "learning_rate": 7.300478318826943e-05, "loss": 2.7095, "step": 34427 }, { "epoch": 2.7, "learning_rate": 7.300399905904493e-05, "loss": 2.888, "step": 34428 }, { "epoch": 2.7, "learning_rate": 7.300321492982043e-05, "loss": 2.7362, "step": 34429 }, { "epoch": 2.7, "learning_rate": 7.300243080059595e-05, "loss": 2.7066, "step": 34430 }, { "epoch": 2.7, "learning_rate": 7.300164667137145e-05, "loss": 2.9382, "step": 34431 }, { "epoch": 2.7, "learning_rate": 7.300086254214696e-05, "loss": 2.9601, "step": 34432 }, { "epoch": 2.7, "learning_rate": 7.300007841292246e-05, "loss": 2.9733, "step": 34433 }, { "epoch": 2.7, "learning_rate": 7.299929428369797e-05, "loss": 3.0178, "step": 34434 }, { "epoch": 2.7, "learning_rate": 7.299851015447347e-05, "loss": 3.0421, "step": 34435 }, { "epoch": 2.7, "learning_rate": 7.299772602524897e-05, "loss": 2.79, "step": 34436 }, { "epoch": 2.7, "learning_rate": 7.299694189602448e-05, "loss": 2.6448, "step": 34437 }, { "epoch": 2.7, "learning_rate": 7.299615776679998e-05, "loss": 3.0407, "step": 34438 }, { "epoch": 2.7, "learning_rate": 7.299537363757547e-05, "loss": 2.6759, "step": 34439 }, { "epoch": 2.7, "learning_rate": 7.299458950835097e-05, "loss": 2.6351, "step": 34440 }, { "epoch": 2.7, "learning_rate": 7.299380537912648e-05, "loss": 3.0476, "step": 34441 }, { "epoch": 2.7, "learning_rate": 7.299302124990198e-05, "loss": 2.9059, "step": 34442 }, { "epoch": 2.7, "learning_rate": 7.299223712067749e-05, "loss": 2.8175, "step": 34443 }, { "epoch": 2.7, "learning_rate": 7.299145299145299e-05, "loss": 3.0061, "step": 34444 }, { "epoch": 2.7, "learning_rate": 7.299066886222849e-05, "loss": 2.9081, "step": 34445 }, { "epoch": 2.7, "learning_rate": 7.2989884733004e-05, "loss": 2.8532, "step": 34446 }, { "epoch": 2.7, "learning_rate": 7.29891006037795e-05, "loss": 2.7818, "step": 34447 }, { "epoch": 2.7, "learning_rate": 7.2988316474555e-05, "loss": 2.8968, "step": 34448 }, { "epoch": 2.7, "learning_rate": 7.298753234533051e-05, "loss": 2.7528, "step": 34449 }, { "epoch": 2.7, "learning_rate": 7.298674821610601e-05, "loss": 2.7518, "step": 34450 }, { "epoch": 2.7, "learning_rate": 7.298596408688151e-05, "loss": 2.7639, "step": 34451 }, { "epoch": 2.7, "learning_rate": 7.298517995765702e-05, "loss": 2.689, "step": 34452 }, { "epoch": 2.7, "learning_rate": 7.298439582843252e-05, "loss": 2.8273, "step": 34453 }, { "epoch": 2.7, "learning_rate": 7.298361169920803e-05, "loss": 2.838, "step": 34454 }, { "epoch": 2.7, "learning_rate": 7.298282756998353e-05, "loss": 2.8486, "step": 34455 }, { "epoch": 2.7, "learning_rate": 7.298204344075905e-05, "loss": 2.8627, "step": 34456 }, { "epoch": 2.7, "learning_rate": 7.298125931153455e-05, "loss": 2.6231, "step": 34457 }, { "epoch": 2.7, "learning_rate": 7.298047518231005e-05, "loss": 2.6709, "step": 34458 }, { "epoch": 2.7, "learning_rate": 7.297969105308556e-05, "loss": 2.7883, "step": 34459 }, { "epoch": 2.7, "learning_rate": 7.297890692386106e-05, "loss": 2.9189, "step": 34460 }, { "epoch": 2.7, "learning_rate": 7.297812279463657e-05, "loss": 2.7828, "step": 34461 }, { "epoch": 2.7, "learning_rate": 7.297733866541207e-05, "loss": 3.2785, "step": 34462 }, { "epoch": 2.7, "learning_rate": 7.297655453618757e-05, "loss": 2.7307, "step": 34463 }, { "epoch": 2.7, "learning_rate": 7.297577040696308e-05, "loss": 2.6688, "step": 34464 }, { "epoch": 2.7, "learning_rate": 7.297498627773858e-05, "loss": 2.6047, "step": 34465 }, { "epoch": 2.7, "learning_rate": 7.297420214851408e-05, "loss": 2.5925, "step": 34466 }, { "epoch": 2.7, "learning_rate": 7.297341801928959e-05, "loss": 2.5842, "step": 34467 }, { "epoch": 2.7, "learning_rate": 7.297263389006509e-05, "loss": 2.8114, "step": 34468 }, { "epoch": 2.7, "learning_rate": 7.29718497608406e-05, "loss": 2.9252, "step": 34469 }, { "epoch": 2.7, "learning_rate": 7.29710656316161e-05, "loss": 2.8566, "step": 34470 }, { "epoch": 2.7, "learning_rate": 7.29702815023916e-05, "loss": 2.7148, "step": 34471 }, { "epoch": 2.7, "learning_rate": 7.29694973731671e-05, "loss": 3.0588, "step": 34472 }, { "epoch": 2.7, "learning_rate": 7.296871324394261e-05, "loss": 2.8091, "step": 34473 }, { "epoch": 2.7, "learning_rate": 7.296792911471811e-05, "loss": 2.7264, "step": 34474 }, { "epoch": 2.7, "learning_rate": 7.29671449854936e-05, "loss": 2.8435, "step": 34475 }, { "epoch": 2.7, "learning_rate": 7.296636085626911e-05, "loss": 2.8887, "step": 34476 }, { "epoch": 2.7, "learning_rate": 7.296557672704461e-05, "loss": 2.8733, "step": 34477 }, { "epoch": 2.7, "learning_rate": 7.296479259782011e-05, "loss": 2.6875, "step": 34478 }, { "epoch": 2.7, "learning_rate": 7.296400846859562e-05, "loss": 2.6617, "step": 34479 }, { "epoch": 2.7, "learning_rate": 7.296322433937112e-05, "loss": 2.7074, "step": 34480 }, { "epoch": 2.7, "learning_rate": 7.296244021014664e-05, "loss": 2.8145, "step": 34481 }, { "epoch": 2.7, "learning_rate": 7.296165608092214e-05, "loss": 2.759, "step": 34482 }, { "epoch": 2.7, "learning_rate": 7.296087195169765e-05, "loss": 2.7885, "step": 34483 }, { "epoch": 2.7, "learning_rate": 7.296008782247315e-05, "loss": 2.8277, "step": 34484 }, { "epoch": 2.7, "learning_rate": 7.295930369324865e-05, "loss": 2.8082, "step": 34485 }, { "epoch": 2.7, "learning_rate": 7.295851956402416e-05, "loss": 2.6876, "step": 34486 }, { "epoch": 2.7, "learning_rate": 7.295773543479966e-05, "loss": 2.9224, "step": 34487 }, { "epoch": 2.7, "learning_rate": 7.295695130557516e-05, "loss": 2.8728, "step": 34488 }, { "epoch": 2.7, "learning_rate": 7.295616717635067e-05, "loss": 2.5878, "step": 34489 }, { "epoch": 2.7, "learning_rate": 7.295538304712617e-05, "loss": 3.0322, "step": 34490 }, { "epoch": 2.7, "learning_rate": 7.295459891790168e-05, "loss": 2.8948, "step": 34491 }, { "epoch": 2.7, "learning_rate": 7.295381478867718e-05, "loss": 2.702, "step": 34492 }, { "epoch": 2.7, "learning_rate": 7.295303065945268e-05, "loss": 2.6733, "step": 34493 }, { "epoch": 2.7, "learning_rate": 7.295224653022819e-05, "loss": 2.7719, "step": 34494 }, { "epoch": 2.7, "learning_rate": 7.295146240100369e-05, "loss": 2.8353, "step": 34495 }, { "epoch": 2.7, "learning_rate": 7.29506782717792e-05, "loss": 2.6953, "step": 34496 }, { "epoch": 2.71, "learning_rate": 7.29498941425547e-05, "loss": 2.7997, "step": 34497 }, { "epoch": 2.71, "learning_rate": 7.29491100133302e-05, "loss": 2.8125, "step": 34498 }, { "epoch": 2.71, "learning_rate": 7.29483258841057e-05, "loss": 2.8348, "step": 34499 }, { "epoch": 2.71, "learning_rate": 7.294754175488121e-05, "loss": 2.9402, "step": 34500 }, { "epoch": 2.71, "learning_rate": 7.294675762565671e-05, "loss": 3.0652, "step": 34501 }, { "epoch": 2.71, "learning_rate": 7.294597349643222e-05, "loss": 2.808, "step": 34502 }, { "epoch": 2.71, "learning_rate": 7.294518936720772e-05, "loss": 2.8656, "step": 34503 }, { "epoch": 2.71, "learning_rate": 7.294440523798322e-05, "loss": 2.964, "step": 34504 }, { "epoch": 2.71, "learning_rate": 7.294362110875873e-05, "loss": 2.7575, "step": 34505 }, { "epoch": 2.71, "learning_rate": 7.294283697953423e-05, "loss": 2.7765, "step": 34506 }, { "epoch": 2.71, "learning_rate": 7.294205285030973e-05, "loss": 2.9593, "step": 34507 }, { "epoch": 2.71, "learning_rate": 7.294126872108524e-05, "loss": 3.0533, "step": 34508 }, { "epoch": 2.71, "learning_rate": 7.294048459186074e-05, "loss": 2.6628, "step": 34509 }, { "epoch": 2.71, "learning_rate": 7.293970046263625e-05, "loss": 2.8644, "step": 34510 }, { "epoch": 2.71, "learning_rate": 7.293891633341175e-05, "loss": 2.8098, "step": 34511 }, { "epoch": 2.71, "learning_rate": 7.293813220418725e-05, "loss": 2.8476, "step": 34512 }, { "epoch": 2.71, "learning_rate": 7.293734807496276e-05, "loss": 2.7881, "step": 34513 }, { "epoch": 2.71, "learning_rate": 7.293656394573826e-05, "loss": 2.7005, "step": 34514 }, { "epoch": 2.71, "learning_rate": 7.293577981651376e-05, "loss": 2.7835, "step": 34515 }, { "epoch": 2.71, "learning_rate": 7.293499568728927e-05, "loss": 3.0373, "step": 34516 }, { "epoch": 2.71, "learning_rate": 7.293421155806477e-05, "loss": 2.6891, "step": 34517 }, { "epoch": 2.71, "learning_rate": 7.293342742884027e-05, "loss": 2.8312, "step": 34518 }, { "epoch": 2.71, "learning_rate": 7.293264329961578e-05, "loss": 2.9099, "step": 34519 }, { "epoch": 2.71, "learning_rate": 7.293185917039128e-05, "loss": 2.6808, "step": 34520 }, { "epoch": 2.71, "learning_rate": 7.293107504116679e-05, "loss": 2.8121, "step": 34521 }, { "epoch": 2.71, "learning_rate": 7.293029091194229e-05, "loss": 3.0792, "step": 34522 }, { "epoch": 2.71, "learning_rate": 7.292950678271779e-05, "loss": 2.872, "step": 34523 }, { "epoch": 2.71, "learning_rate": 7.29287226534933e-05, "loss": 2.7756, "step": 34524 }, { "epoch": 2.71, "learning_rate": 7.29279385242688e-05, "loss": 2.7262, "step": 34525 }, { "epoch": 2.71, "learning_rate": 7.29271543950443e-05, "loss": 3.0287, "step": 34526 }, { "epoch": 2.71, "learning_rate": 7.292637026581981e-05, "loss": 2.7046, "step": 34527 }, { "epoch": 2.71, "learning_rate": 7.292558613659531e-05, "loss": 2.7536, "step": 34528 }, { "epoch": 2.71, "learning_rate": 7.292480200737082e-05, "loss": 2.7671, "step": 34529 }, { "epoch": 2.71, "learning_rate": 7.292401787814632e-05, "loss": 3.1041, "step": 34530 }, { "epoch": 2.71, "learning_rate": 7.292323374892182e-05, "loss": 2.9765, "step": 34531 }, { "epoch": 2.71, "learning_rate": 7.292244961969733e-05, "loss": 2.879, "step": 34532 }, { "epoch": 2.71, "learning_rate": 7.292166549047283e-05, "loss": 2.7146, "step": 34533 }, { "epoch": 2.71, "learning_rate": 7.292088136124833e-05, "loss": 2.8987, "step": 34534 }, { "epoch": 2.71, "learning_rate": 7.292009723202384e-05, "loss": 2.7367, "step": 34535 }, { "epoch": 2.71, "learning_rate": 7.291931310279934e-05, "loss": 2.7951, "step": 34536 }, { "epoch": 2.71, "learning_rate": 7.291852897357484e-05, "loss": 2.8158, "step": 34537 }, { "epoch": 2.71, "learning_rate": 7.291774484435036e-05, "loss": 2.8425, "step": 34538 }, { "epoch": 2.71, "learning_rate": 7.291696071512587e-05, "loss": 2.9017, "step": 34539 }, { "epoch": 2.71, "learning_rate": 7.291617658590137e-05, "loss": 2.8421, "step": 34540 }, { "epoch": 2.71, "learning_rate": 7.291539245667687e-05, "loss": 2.7577, "step": 34541 }, { "epoch": 2.71, "learning_rate": 7.291460832745238e-05, "loss": 2.8291, "step": 34542 }, { "epoch": 2.71, "learning_rate": 7.291382419822788e-05, "loss": 2.8083, "step": 34543 }, { "epoch": 2.71, "learning_rate": 7.291304006900338e-05, "loss": 2.6932, "step": 34544 }, { "epoch": 2.71, "learning_rate": 7.291225593977889e-05, "loss": 2.7736, "step": 34545 }, { "epoch": 2.71, "learning_rate": 7.291147181055439e-05, "loss": 2.8863, "step": 34546 }, { "epoch": 2.71, "learning_rate": 7.291068768132988e-05, "loss": 2.7735, "step": 34547 }, { "epoch": 2.71, "learning_rate": 7.290990355210539e-05, "loss": 2.7176, "step": 34548 }, { "epoch": 2.71, "learning_rate": 7.290911942288089e-05, "loss": 2.8909, "step": 34549 }, { "epoch": 2.71, "learning_rate": 7.290833529365639e-05, "loss": 2.6886, "step": 34550 }, { "epoch": 2.71, "learning_rate": 7.29075511644319e-05, "loss": 2.9558, "step": 34551 }, { "epoch": 2.71, "learning_rate": 7.29067670352074e-05, "loss": 2.774, "step": 34552 }, { "epoch": 2.71, "learning_rate": 7.29059829059829e-05, "loss": 2.8335, "step": 34553 }, { "epoch": 2.71, "learning_rate": 7.290519877675841e-05, "loss": 3.0546, "step": 34554 }, { "epoch": 2.71, "learning_rate": 7.290441464753391e-05, "loss": 2.7348, "step": 34555 }, { "epoch": 2.71, "learning_rate": 7.290363051830941e-05, "loss": 2.5718, "step": 34556 }, { "epoch": 2.71, "learning_rate": 7.290284638908492e-05, "loss": 2.5821, "step": 34557 }, { "epoch": 2.71, "learning_rate": 7.290206225986042e-05, "loss": 2.7517, "step": 34558 }, { "epoch": 2.71, "learning_rate": 7.290127813063593e-05, "loss": 2.7019, "step": 34559 }, { "epoch": 2.71, "learning_rate": 7.290049400141143e-05, "loss": 2.7618, "step": 34560 }, { "epoch": 2.71, "learning_rate": 7.289970987218693e-05, "loss": 2.8559, "step": 34561 }, { "epoch": 2.71, "learning_rate": 7.289892574296244e-05, "loss": 2.8049, "step": 34562 }, { "epoch": 2.71, "learning_rate": 7.289814161373794e-05, "loss": 2.8216, "step": 34563 }, { "epoch": 2.71, "learning_rate": 7.289735748451346e-05, "loss": 2.5278, "step": 34564 }, { "epoch": 2.71, "learning_rate": 7.289657335528896e-05, "loss": 2.7569, "step": 34565 }, { "epoch": 2.71, "learning_rate": 7.289578922606446e-05, "loss": 2.826, "step": 34566 }, { "epoch": 2.71, "learning_rate": 7.289500509683997e-05, "loss": 2.8933, "step": 34567 }, { "epoch": 2.71, "learning_rate": 7.289422096761547e-05, "loss": 2.6717, "step": 34568 }, { "epoch": 2.71, "learning_rate": 7.289343683839098e-05, "loss": 2.8638, "step": 34569 }, { "epoch": 2.71, "learning_rate": 7.289265270916648e-05, "loss": 2.7192, "step": 34570 }, { "epoch": 2.71, "learning_rate": 7.289186857994198e-05, "loss": 2.7326, "step": 34571 }, { "epoch": 2.71, "learning_rate": 7.289108445071749e-05, "loss": 2.6871, "step": 34572 }, { "epoch": 2.71, "learning_rate": 7.289030032149299e-05, "loss": 2.6654, "step": 34573 }, { "epoch": 2.71, "learning_rate": 7.28895161922685e-05, "loss": 2.8589, "step": 34574 }, { "epoch": 2.71, "learning_rate": 7.2888732063044e-05, "loss": 2.8143, "step": 34575 }, { "epoch": 2.71, "learning_rate": 7.28879479338195e-05, "loss": 2.6897, "step": 34576 }, { "epoch": 2.71, "learning_rate": 7.2887163804595e-05, "loss": 2.6785, "step": 34577 }, { "epoch": 2.71, "learning_rate": 7.288637967537051e-05, "loss": 2.87, "step": 34578 }, { "epoch": 2.71, "learning_rate": 7.288559554614601e-05, "loss": 2.9536, "step": 34579 }, { "epoch": 2.71, "learning_rate": 7.288481141692152e-05, "loss": 2.8742, "step": 34580 }, { "epoch": 2.71, "learning_rate": 7.288402728769702e-05, "loss": 2.7082, "step": 34581 }, { "epoch": 2.71, "learning_rate": 7.288324315847252e-05, "loss": 2.7674, "step": 34582 }, { "epoch": 2.71, "learning_rate": 7.288245902924801e-05, "loss": 2.8079, "step": 34583 }, { "epoch": 2.71, "learning_rate": 7.288167490002352e-05, "loss": 2.7684, "step": 34584 }, { "epoch": 2.71, "learning_rate": 7.288089077079902e-05, "loss": 2.8646, "step": 34585 }, { "epoch": 2.71, "learning_rate": 7.288010664157452e-05, "loss": 2.8428, "step": 34586 }, { "epoch": 2.71, "learning_rate": 7.287932251235003e-05, "loss": 3.0723, "step": 34587 }, { "epoch": 2.71, "learning_rate": 7.287853838312553e-05, "loss": 2.8677, "step": 34588 }, { "epoch": 2.71, "learning_rate": 7.287775425390105e-05, "loss": 2.7061, "step": 34589 }, { "epoch": 2.71, "learning_rate": 7.287697012467655e-05, "loss": 2.806, "step": 34590 }, { "epoch": 2.71, "learning_rate": 7.287618599545206e-05, "loss": 2.6655, "step": 34591 }, { "epoch": 2.71, "learning_rate": 7.287540186622756e-05, "loss": 2.8666, "step": 34592 }, { "epoch": 2.71, "learning_rate": 7.287461773700306e-05, "loss": 2.9566, "step": 34593 }, { "epoch": 2.71, "learning_rate": 7.287383360777857e-05, "loss": 2.6624, "step": 34594 }, { "epoch": 2.71, "learning_rate": 7.287304947855407e-05, "loss": 2.9154, "step": 34595 }, { "epoch": 2.71, "learning_rate": 7.287226534932958e-05, "loss": 2.7617, "step": 34596 }, { "epoch": 2.71, "learning_rate": 7.287148122010508e-05, "loss": 3.0113, "step": 34597 }, { "epoch": 2.71, "learning_rate": 7.287069709088058e-05, "loss": 2.7625, "step": 34598 }, { "epoch": 2.71, "learning_rate": 7.286991296165609e-05, "loss": 2.7188, "step": 34599 }, { "epoch": 2.71, "learning_rate": 7.286912883243159e-05, "loss": 2.9484, "step": 34600 }, { "epoch": 2.71, "learning_rate": 7.28683447032071e-05, "loss": 2.8929, "step": 34601 }, { "epoch": 2.71, "learning_rate": 7.28675605739826e-05, "loss": 2.6526, "step": 34602 }, { "epoch": 2.71, "learning_rate": 7.28667764447581e-05, "loss": 2.8224, "step": 34603 }, { "epoch": 2.71, "learning_rate": 7.28659923155336e-05, "loss": 2.7933, "step": 34604 }, { "epoch": 2.71, "learning_rate": 7.286520818630911e-05, "loss": 2.5464, "step": 34605 }, { "epoch": 2.71, "learning_rate": 7.286442405708461e-05, "loss": 2.8671, "step": 34606 }, { "epoch": 2.71, "learning_rate": 7.286363992786012e-05, "loss": 2.8694, "step": 34607 }, { "epoch": 2.71, "learning_rate": 7.286285579863562e-05, "loss": 2.782, "step": 34608 }, { "epoch": 2.71, "learning_rate": 7.286207166941112e-05, "loss": 2.8195, "step": 34609 }, { "epoch": 2.71, "learning_rate": 7.286128754018663e-05, "loss": 2.7022, "step": 34610 }, { "epoch": 2.71, "learning_rate": 7.286050341096213e-05, "loss": 3.0259, "step": 34611 }, { "epoch": 2.71, "learning_rate": 7.285971928173763e-05, "loss": 2.9132, "step": 34612 }, { "epoch": 2.71, "learning_rate": 7.285893515251314e-05, "loss": 2.7051, "step": 34613 }, { "epoch": 2.71, "learning_rate": 7.285815102328864e-05, "loss": 2.686, "step": 34614 }, { "epoch": 2.71, "learning_rate": 7.285736689406415e-05, "loss": 2.6475, "step": 34615 }, { "epoch": 2.71, "learning_rate": 7.285658276483965e-05, "loss": 2.7786, "step": 34616 }, { "epoch": 2.71, "learning_rate": 7.285579863561515e-05, "loss": 2.7306, "step": 34617 }, { "epoch": 2.71, "learning_rate": 7.285501450639066e-05, "loss": 2.648, "step": 34618 }, { "epoch": 2.71, "learning_rate": 7.285423037716616e-05, "loss": 2.7643, "step": 34619 }, { "epoch": 2.71, "learning_rate": 7.285344624794166e-05, "loss": 2.653, "step": 34620 }, { "epoch": 2.71, "learning_rate": 7.285266211871717e-05, "loss": 2.635, "step": 34621 }, { "epoch": 2.71, "learning_rate": 7.285187798949267e-05, "loss": 2.6125, "step": 34622 }, { "epoch": 2.71, "learning_rate": 7.285109386026817e-05, "loss": 2.8025, "step": 34623 }, { "epoch": 2.71, "learning_rate": 7.285030973104368e-05, "loss": 2.7434, "step": 34624 }, { "epoch": 2.72, "learning_rate": 7.284952560181918e-05, "loss": 2.7192, "step": 34625 }, { "epoch": 2.72, "learning_rate": 7.284874147259469e-05, "loss": 2.9787, "step": 34626 }, { "epoch": 2.72, "learning_rate": 7.284795734337019e-05, "loss": 2.897, "step": 34627 }, { "epoch": 2.72, "learning_rate": 7.284717321414569e-05, "loss": 2.8743, "step": 34628 }, { "epoch": 2.72, "learning_rate": 7.28463890849212e-05, "loss": 2.9955, "step": 34629 }, { "epoch": 2.72, "learning_rate": 7.28456049556967e-05, "loss": 2.722, "step": 34630 }, { "epoch": 2.72, "learning_rate": 7.28448208264722e-05, "loss": 2.7966, "step": 34631 }, { "epoch": 2.72, "learning_rate": 7.284403669724771e-05, "loss": 3.1747, "step": 34632 }, { "epoch": 2.72, "learning_rate": 7.284325256802321e-05, "loss": 2.7027, "step": 34633 }, { "epoch": 2.72, "learning_rate": 7.284246843879871e-05, "loss": 2.6541, "step": 34634 }, { "epoch": 2.72, "learning_rate": 7.284168430957422e-05, "loss": 2.5702, "step": 34635 }, { "epoch": 2.72, "learning_rate": 7.284090018034972e-05, "loss": 2.8676, "step": 34636 }, { "epoch": 2.72, "learning_rate": 7.284011605112523e-05, "loss": 2.7689, "step": 34637 }, { "epoch": 2.72, "learning_rate": 7.283933192190073e-05, "loss": 2.8519, "step": 34638 }, { "epoch": 2.72, "learning_rate": 7.283854779267623e-05, "loss": 2.8104, "step": 34639 }, { "epoch": 2.72, "learning_rate": 7.283776366345174e-05, "loss": 2.8213, "step": 34640 }, { "epoch": 2.72, "learning_rate": 7.283697953422724e-05, "loss": 2.6392, "step": 34641 }, { "epoch": 2.72, "learning_rate": 7.283619540500274e-05, "loss": 2.876, "step": 34642 }, { "epoch": 2.72, "learning_rate": 7.283541127577825e-05, "loss": 2.7801, "step": 34643 }, { "epoch": 2.72, "learning_rate": 7.283462714655375e-05, "loss": 2.8726, "step": 34644 }, { "epoch": 2.72, "learning_rate": 7.283384301732926e-05, "loss": 2.524, "step": 34645 }, { "epoch": 2.72, "learning_rate": 7.283305888810477e-05, "loss": 2.8434, "step": 34646 }, { "epoch": 2.72, "learning_rate": 7.283227475888028e-05, "loss": 2.5482, "step": 34647 }, { "epoch": 2.72, "learning_rate": 7.283149062965578e-05, "loss": 2.7367, "step": 34648 }, { "epoch": 2.72, "learning_rate": 7.283070650043128e-05, "loss": 2.8198, "step": 34649 }, { "epoch": 2.72, "learning_rate": 7.282992237120679e-05, "loss": 2.985, "step": 34650 }, { "epoch": 2.72, "learning_rate": 7.282913824198229e-05, "loss": 2.8116, "step": 34651 }, { "epoch": 2.72, "learning_rate": 7.28283541127578e-05, "loss": 2.6323, "step": 34652 }, { "epoch": 2.72, "learning_rate": 7.28275699835333e-05, "loss": 2.7958, "step": 34653 }, { "epoch": 2.72, "learning_rate": 7.28267858543088e-05, "loss": 2.5953, "step": 34654 }, { "epoch": 2.72, "learning_rate": 7.282600172508429e-05, "loss": 2.7478, "step": 34655 }, { "epoch": 2.72, "learning_rate": 7.28252175958598e-05, "loss": 2.8704, "step": 34656 }, { "epoch": 2.72, "learning_rate": 7.28244334666353e-05, "loss": 2.6593, "step": 34657 }, { "epoch": 2.72, "learning_rate": 7.28236493374108e-05, "loss": 2.7561, "step": 34658 }, { "epoch": 2.72, "learning_rate": 7.282286520818631e-05, "loss": 2.7699, "step": 34659 }, { "epoch": 2.72, "learning_rate": 7.282208107896181e-05, "loss": 2.9582, "step": 34660 }, { "epoch": 2.72, "learning_rate": 7.282129694973731e-05, "loss": 2.7013, "step": 34661 }, { "epoch": 2.72, "learning_rate": 7.282051282051282e-05, "loss": 2.701, "step": 34662 }, { "epoch": 2.72, "learning_rate": 7.281972869128832e-05, "loss": 2.7227, "step": 34663 }, { "epoch": 2.72, "learning_rate": 7.281894456206383e-05, "loss": 2.6956, "step": 34664 }, { "epoch": 2.72, "learning_rate": 7.281816043283933e-05, "loss": 2.7722, "step": 34665 }, { "epoch": 2.72, "learning_rate": 7.281737630361483e-05, "loss": 2.8495, "step": 34666 }, { "epoch": 2.72, "learning_rate": 7.281659217439034e-05, "loss": 2.9638, "step": 34667 }, { "epoch": 2.72, "learning_rate": 7.281580804516584e-05, "loss": 2.7035, "step": 34668 }, { "epoch": 2.72, "learning_rate": 7.281502391594134e-05, "loss": 2.8583, "step": 34669 }, { "epoch": 2.72, "learning_rate": 7.281423978671685e-05, "loss": 2.5939, "step": 34670 }, { "epoch": 2.72, "learning_rate": 7.281345565749235e-05, "loss": 2.7319, "step": 34671 }, { "epoch": 2.72, "learning_rate": 7.281267152826787e-05, "loss": 2.5898, "step": 34672 }, { "epoch": 2.72, "learning_rate": 7.281188739904337e-05, "loss": 2.587, "step": 34673 }, { "epoch": 2.72, "learning_rate": 7.281110326981888e-05, "loss": 2.9743, "step": 34674 }, { "epoch": 2.72, "learning_rate": 7.281031914059438e-05, "loss": 2.6928, "step": 34675 }, { "epoch": 2.72, "learning_rate": 7.280953501136988e-05, "loss": 2.6587, "step": 34676 }, { "epoch": 2.72, "learning_rate": 7.280875088214539e-05, "loss": 3.029, "step": 34677 }, { "epoch": 2.72, "learning_rate": 7.280796675292089e-05, "loss": 2.8374, "step": 34678 }, { "epoch": 2.72, "learning_rate": 7.28071826236964e-05, "loss": 2.9492, "step": 34679 }, { "epoch": 2.72, "learning_rate": 7.28063984944719e-05, "loss": 2.7866, "step": 34680 }, { "epoch": 2.72, "learning_rate": 7.28056143652474e-05, "loss": 2.8427, "step": 34681 }, { "epoch": 2.72, "learning_rate": 7.28048302360229e-05, "loss": 2.8168, "step": 34682 }, { "epoch": 2.72, "learning_rate": 7.280404610679841e-05, "loss": 2.9882, "step": 34683 }, { "epoch": 2.72, "learning_rate": 7.280326197757391e-05, "loss": 2.68, "step": 34684 }, { "epoch": 2.72, "learning_rate": 7.280247784834942e-05, "loss": 2.7405, "step": 34685 }, { "epoch": 2.72, "learning_rate": 7.280169371912492e-05, "loss": 2.888, "step": 34686 }, { "epoch": 2.72, "learning_rate": 7.280090958990042e-05, "loss": 2.8984, "step": 34687 }, { "epoch": 2.72, "learning_rate": 7.280012546067593e-05, "loss": 2.8136, "step": 34688 }, { "epoch": 2.72, "learning_rate": 7.279934133145143e-05, "loss": 2.7234, "step": 34689 }, { "epoch": 2.72, "learning_rate": 7.279855720222693e-05, "loss": 2.7757, "step": 34690 }, { "epoch": 2.72, "learning_rate": 7.279777307300242e-05, "loss": 2.9352, "step": 34691 }, { "epoch": 2.72, "learning_rate": 7.279698894377793e-05, "loss": 3.0044, "step": 34692 }, { "epoch": 2.72, "learning_rate": 7.279620481455343e-05, "loss": 2.928, "step": 34693 }, { "epoch": 2.72, "learning_rate": 7.279542068532894e-05, "loss": 2.7848, "step": 34694 }, { "epoch": 2.72, "learning_rate": 7.279463655610444e-05, "loss": 2.9347, "step": 34695 }, { "epoch": 2.72, "learning_rate": 7.279385242687994e-05, "loss": 2.8155, "step": 34696 }, { "epoch": 2.72, "learning_rate": 7.279306829765545e-05, "loss": 3.0022, "step": 34697 }, { "epoch": 2.72, "learning_rate": 7.279228416843096e-05, "loss": 2.7634, "step": 34698 }, { "epoch": 2.72, "learning_rate": 7.279150003920647e-05, "loss": 2.725, "step": 34699 }, { "epoch": 2.72, "learning_rate": 7.279071590998197e-05, "loss": 2.8689, "step": 34700 }, { "epoch": 2.72, "learning_rate": 7.278993178075747e-05, "loss": 3.03, "step": 34701 }, { "epoch": 2.72, "learning_rate": 7.278914765153298e-05, "loss": 2.7951, "step": 34702 }, { "epoch": 2.72, "learning_rate": 7.278836352230848e-05, "loss": 2.6804, "step": 34703 }, { "epoch": 2.72, "learning_rate": 7.278757939308399e-05, "loss": 2.7923, "step": 34704 }, { "epoch": 2.72, "learning_rate": 7.278679526385949e-05, "loss": 2.7356, "step": 34705 }, { "epoch": 2.72, "learning_rate": 7.2786011134635e-05, "loss": 2.7286, "step": 34706 }, { "epoch": 2.72, "learning_rate": 7.27852270054105e-05, "loss": 2.7769, "step": 34707 }, { "epoch": 2.72, "learning_rate": 7.2784442876186e-05, "loss": 2.6863, "step": 34708 }, { "epoch": 2.72, "learning_rate": 7.27836587469615e-05, "loss": 2.7576, "step": 34709 }, { "epoch": 2.72, "learning_rate": 7.278287461773701e-05, "loss": 2.8628, "step": 34710 }, { "epoch": 2.72, "learning_rate": 7.278209048851251e-05, "loss": 2.8187, "step": 34711 }, { "epoch": 2.72, "learning_rate": 7.278130635928802e-05, "loss": 3.1107, "step": 34712 }, { "epoch": 2.72, "learning_rate": 7.278052223006352e-05, "loss": 2.6774, "step": 34713 }, { "epoch": 2.72, "learning_rate": 7.277973810083902e-05, "loss": 2.6987, "step": 34714 }, { "epoch": 2.72, "learning_rate": 7.277895397161453e-05, "loss": 2.7426, "step": 34715 }, { "epoch": 2.72, "learning_rate": 7.277816984239003e-05, "loss": 2.9337, "step": 34716 }, { "epoch": 2.72, "learning_rate": 7.277738571316553e-05, "loss": 3.1688, "step": 34717 }, { "epoch": 2.72, "learning_rate": 7.277660158394104e-05, "loss": 2.8668, "step": 34718 }, { "epoch": 2.72, "learning_rate": 7.277581745471654e-05, "loss": 2.8989, "step": 34719 }, { "epoch": 2.72, "learning_rate": 7.277503332549204e-05, "loss": 2.8762, "step": 34720 }, { "epoch": 2.72, "learning_rate": 7.277424919626755e-05, "loss": 2.6322, "step": 34721 }, { "epoch": 2.72, "learning_rate": 7.277346506704305e-05, "loss": 2.6846, "step": 34722 }, { "epoch": 2.72, "learning_rate": 7.277268093781856e-05, "loss": 2.7233, "step": 34723 }, { "epoch": 2.72, "learning_rate": 7.277189680859406e-05, "loss": 2.6824, "step": 34724 }, { "epoch": 2.72, "learning_rate": 7.277111267936956e-05, "loss": 2.8281, "step": 34725 }, { "epoch": 2.72, "learning_rate": 7.277032855014507e-05, "loss": 2.7463, "step": 34726 }, { "epoch": 2.72, "learning_rate": 7.276954442092057e-05, "loss": 2.8103, "step": 34727 }, { "epoch": 2.72, "learning_rate": 7.276876029169607e-05, "loss": 2.8883, "step": 34728 }, { "epoch": 2.72, "learning_rate": 7.276797616247158e-05, "loss": 2.9623, "step": 34729 }, { "epoch": 2.72, "learning_rate": 7.276719203324708e-05, "loss": 2.8205, "step": 34730 }, { "epoch": 2.72, "learning_rate": 7.276640790402259e-05, "loss": 2.6135, "step": 34731 }, { "epoch": 2.72, "learning_rate": 7.276562377479809e-05, "loss": 3.0508, "step": 34732 }, { "epoch": 2.72, "learning_rate": 7.276483964557359e-05, "loss": 2.8533, "step": 34733 }, { "epoch": 2.72, "learning_rate": 7.27640555163491e-05, "loss": 2.9798, "step": 34734 }, { "epoch": 2.72, "learning_rate": 7.27632713871246e-05, "loss": 2.8829, "step": 34735 }, { "epoch": 2.72, "learning_rate": 7.27624872579001e-05, "loss": 3.0506, "step": 34736 }, { "epoch": 2.72, "learning_rate": 7.276170312867561e-05, "loss": 2.7613, "step": 34737 }, { "epoch": 2.72, "learning_rate": 7.276091899945111e-05, "loss": 2.928, "step": 34738 }, { "epoch": 2.72, "learning_rate": 7.276013487022661e-05, "loss": 2.7686, "step": 34739 }, { "epoch": 2.72, "learning_rate": 7.275935074100212e-05, "loss": 2.602, "step": 34740 }, { "epoch": 2.72, "learning_rate": 7.275856661177762e-05, "loss": 2.7946, "step": 34741 }, { "epoch": 2.72, "learning_rate": 7.275778248255313e-05, "loss": 2.6433, "step": 34742 }, { "epoch": 2.72, "learning_rate": 7.275699835332863e-05, "loss": 3.0534, "step": 34743 }, { "epoch": 2.72, "learning_rate": 7.275621422410413e-05, "loss": 2.7325, "step": 34744 }, { "epoch": 2.72, "learning_rate": 7.275543009487964e-05, "loss": 2.9795, "step": 34745 }, { "epoch": 2.72, "learning_rate": 7.275464596565514e-05, "loss": 2.5842, "step": 34746 }, { "epoch": 2.72, "learning_rate": 7.275386183643064e-05, "loss": 2.6616, "step": 34747 }, { "epoch": 2.72, "learning_rate": 7.275307770720615e-05, "loss": 2.7576, "step": 34748 }, { "epoch": 2.72, "learning_rate": 7.275229357798165e-05, "loss": 2.7887, "step": 34749 }, { "epoch": 2.72, "learning_rate": 7.275150944875716e-05, "loss": 2.764, "step": 34750 }, { "epoch": 2.72, "learning_rate": 7.275072531953266e-05, "loss": 2.8757, "step": 34751 }, { "epoch": 2.73, "learning_rate": 7.274994119030816e-05, "loss": 2.806, "step": 34752 }, { "epoch": 2.73, "learning_rate": 7.274915706108367e-05, "loss": 2.9663, "step": 34753 }, { "epoch": 2.73, "learning_rate": 7.274837293185918e-05, "loss": 2.8403, "step": 34754 }, { "epoch": 2.73, "learning_rate": 7.274758880263469e-05, "loss": 2.793, "step": 34755 }, { "epoch": 2.73, "learning_rate": 7.274680467341019e-05, "loss": 2.6881, "step": 34756 }, { "epoch": 2.73, "learning_rate": 7.27460205441857e-05, "loss": 2.7349, "step": 34757 }, { "epoch": 2.73, "learning_rate": 7.27452364149612e-05, "loss": 2.8496, "step": 34758 }, { "epoch": 2.73, "learning_rate": 7.27444522857367e-05, "loss": 2.8079, "step": 34759 }, { "epoch": 2.73, "learning_rate": 7.27436681565122e-05, "loss": 2.8981, "step": 34760 }, { "epoch": 2.73, "learning_rate": 7.274288402728771e-05, "loss": 2.8153, "step": 34761 }, { "epoch": 2.73, "learning_rate": 7.274209989806321e-05, "loss": 2.6551, "step": 34762 }, { "epoch": 2.73, "learning_rate": 7.27413157688387e-05, "loss": 2.549, "step": 34763 }, { "epoch": 2.73, "learning_rate": 7.27405316396142e-05, "loss": 2.7827, "step": 34764 }, { "epoch": 2.73, "learning_rate": 7.273974751038971e-05, "loss": 2.8298, "step": 34765 }, { "epoch": 2.73, "learning_rate": 7.273896338116521e-05, "loss": 2.6922, "step": 34766 }, { "epoch": 2.73, "learning_rate": 7.273817925194072e-05, "loss": 2.7634, "step": 34767 }, { "epoch": 2.73, "learning_rate": 7.273739512271622e-05, "loss": 2.8724, "step": 34768 }, { "epoch": 2.73, "learning_rate": 7.273661099349173e-05, "loss": 2.8447, "step": 34769 }, { "epoch": 2.73, "learning_rate": 7.273582686426723e-05, "loss": 2.6629, "step": 34770 }, { "epoch": 2.73, "learning_rate": 7.273504273504273e-05, "loss": 2.926, "step": 34771 }, { "epoch": 2.73, "learning_rate": 7.273425860581824e-05, "loss": 2.9295, "step": 34772 }, { "epoch": 2.73, "learning_rate": 7.273347447659374e-05, "loss": 2.9913, "step": 34773 }, { "epoch": 2.73, "learning_rate": 7.273269034736924e-05, "loss": 2.735, "step": 34774 }, { "epoch": 2.73, "learning_rate": 7.273190621814475e-05, "loss": 2.9056, "step": 34775 }, { "epoch": 2.73, "learning_rate": 7.273112208892025e-05, "loss": 3.0425, "step": 34776 }, { "epoch": 2.73, "learning_rate": 7.273033795969575e-05, "loss": 2.7742, "step": 34777 }, { "epoch": 2.73, "learning_rate": 7.272955383047126e-05, "loss": 2.7075, "step": 34778 }, { "epoch": 2.73, "learning_rate": 7.272876970124676e-05, "loss": 2.6259, "step": 34779 }, { "epoch": 2.73, "learning_rate": 7.272798557202228e-05, "loss": 2.8778, "step": 34780 }, { "epoch": 2.73, "learning_rate": 7.272720144279778e-05, "loss": 2.7856, "step": 34781 }, { "epoch": 2.73, "learning_rate": 7.272641731357329e-05, "loss": 2.809, "step": 34782 }, { "epoch": 2.73, "learning_rate": 7.272563318434879e-05, "loss": 2.7004, "step": 34783 }, { "epoch": 2.73, "learning_rate": 7.27248490551243e-05, "loss": 2.9484, "step": 34784 }, { "epoch": 2.73, "learning_rate": 7.27240649258998e-05, "loss": 2.9483, "step": 34785 }, { "epoch": 2.73, "learning_rate": 7.27232807966753e-05, "loss": 2.8434, "step": 34786 }, { "epoch": 2.73, "learning_rate": 7.27224966674508e-05, "loss": 2.6075, "step": 34787 }, { "epoch": 2.73, "learning_rate": 7.272171253822631e-05, "loss": 2.9082, "step": 34788 }, { "epoch": 2.73, "learning_rate": 7.272092840900181e-05, "loss": 2.8185, "step": 34789 }, { "epoch": 2.73, "learning_rate": 7.272014427977732e-05, "loss": 2.7021, "step": 34790 }, { "epoch": 2.73, "learning_rate": 7.271936015055282e-05, "loss": 2.5991, "step": 34791 }, { "epoch": 2.73, "learning_rate": 7.271857602132832e-05, "loss": 2.6768, "step": 34792 }, { "epoch": 2.73, "learning_rate": 7.271779189210383e-05, "loss": 2.8017, "step": 34793 }, { "epoch": 2.73, "learning_rate": 7.271700776287933e-05, "loss": 2.6193, "step": 34794 }, { "epoch": 2.73, "learning_rate": 7.271622363365483e-05, "loss": 2.9076, "step": 34795 }, { "epoch": 2.73, "learning_rate": 7.271543950443034e-05, "loss": 2.7557, "step": 34796 }, { "epoch": 2.73, "learning_rate": 7.271465537520584e-05, "loss": 2.7625, "step": 34797 }, { "epoch": 2.73, "learning_rate": 7.271387124598135e-05, "loss": 2.6343, "step": 34798 }, { "epoch": 2.73, "learning_rate": 7.271308711675684e-05, "loss": 2.935, "step": 34799 }, { "epoch": 2.73, "learning_rate": 7.271230298753234e-05, "loss": 2.7921, "step": 34800 }, { "epoch": 2.73, "learning_rate": 7.271151885830784e-05, "loss": 3.1036, "step": 34801 }, { "epoch": 2.73, "learning_rate": 7.271073472908335e-05, "loss": 2.7153, "step": 34802 }, { "epoch": 2.73, "learning_rate": 7.270995059985885e-05, "loss": 2.8398, "step": 34803 }, { "epoch": 2.73, "learning_rate": 7.270916647063435e-05, "loss": 2.6104, "step": 34804 }, { "epoch": 2.73, "learning_rate": 7.270838234140986e-05, "loss": 2.9941, "step": 34805 }, { "epoch": 2.73, "learning_rate": 7.270759821218537e-05, "loss": 2.6855, "step": 34806 }, { "epoch": 2.73, "learning_rate": 7.270681408296088e-05, "loss": 3.1147, "step": 34807 }, { "epoch": 2.73, "learning_rate": 7.270602995373638e-05, "loss": 2.906, "step": 34808 }, { "epoch": 2.73, "learning_rate": 7.270524582451189e-05, "loss": 2.8887, "step": 34809 }, { "epoch": 2.73, "learning_rate": 7.270446169528739e-05, "loss": 2.5763, "step": 34810 }, { "epoch": 2.73, "learning_rate": 7.270367756606289e-05, "loss": 2.8656, "step": 34811 }, { "epoch": 2.73, "learning_rate": 7.27028934368384e-05, "loss": 2.8296, "step": 34812 }, { "epoch": 2.73, "learning_rate": 7.27021093076139e-05, "loss": 2.784, "step": 34813 }, { "epoch": 2.73, "learning_rate": 7.27013251783894e-05, "loss": 2.5655, "step": 34814 }, { "epoch": 2.73, "learning_rate": 7.270054104916491e-05, "loss": 2.7816, "step": 34815 }, { "epoch": 2.73, "learning_rate": 7.269975691994041e-05, "loss": 2.8849, "step": 34816 }, { "epoch": 2.73, "learning_rate": 7.269897279071592e-05, "loss": 2.7474, "step": 34817 }, { "epoch": 2.73, "learning_rate": 7.269818866149142e-05, "loss": 2.8181, "step": 34818 }, { "epoch": 2.73, "learning_rate": 7.269740453226692e-05, "loss": 2.589, "step": 34819 }, { "epoch": 2.73, "learning_rate": 7.269662040304243e-05, "loss": 2.7316, "step": 34820 }, { "epoch": 2.73, "learning_rate": 7.269583627381793e-05, "loss": 2.8805, "step": 34821 }, { "epoch": 2.73, "learning_rate": 7.269505214459343e-05, "loss": 2.9272, "step": 34822 }, { "epoch": 2.73, "learning_rate": 7.269426801536894e-05, "loss": 2.8911, "step": 34823 }, { "epoch": 2.73, "learning_rate": 7.269348388614444e-05, "loss": 2.6754, "step": 34824 }, { "epoch": 2.73, "learning_rate": 7.269269975691994e-05, "loss": 2.8123, "step": 34825 }, { "epoch": 2.73, "learning_rate": 7.269191562769545e-05, "loss": 3.0284, "step": 34826 }, { "epoch": 2.73, "learning_rate": 7.269113149847095e-05, "loss": 2.7115, "step": 34827 }, { "epoch": 2.73, "learning_rate": 7.269034736924646e-05, "loss": 2.7424, "step": 34828 }, { "epoch": 2.73, "learning_rate": 7.268956324002196e-05, "loss": 2.8907, "step": 34829 }, { "epoch": 2.73, "learning_rate": 7.268877911079746e-05, "loss": 2.9126, "step": 34830 }, { "epoch": 2.73, "learning_rate": 7.268799498157297e-05, "loss": 2.714, "step": 34831 }, { "epoch": 2.73, "learning_rate": 7.268721085234847e-05, "loss": 2.7048, "step": 34832 }, { "epoch": 2.73, "learning_rate": 7.268642672312397e-05, "loss": 2.6372, "step": 34833 }, { "epoch": 2.73, "learning_rate": 7.268564259389948e-05, "loss": 2.7593, "step": 34834 }, { "epoch": 2.73, "learning_rate": 7.268485846467498e-05, "loss": 2.8021, "step": 34835 }, { "epoch": 2.73, "learning_rate": 7.268407433545048e-05, "loss": 2.7492, "step": 34836 }, { "epoch": 2.73, "learning_rate": 7.268329020622599e-05, "loss": 2.5788, "step": 34837 }, { "epoch": 2.73, "learning_rate": 7.268250607700149e-05, "loss": 2.752, "step": 34838 }, { "epoch": 2.73, "learning_rate": 7.2681721947777e-05, "loss": 2.8076, "step": 34839 }, { "epoch": 2.73, "learning_rate": 7.26809378185525e-05, "loss": 2.9392, "step": 34840 }, { "epoch": 2.73, "learning_rate": 7.2680153689328e-05, "loss": 2.6677, "step": 34841 }, { "epoch": 2.73, "learning_rate": 7.267936956010351e-05, "loss": 2.8362, "step": 34842 }, { "epoch": 2.73, "learning_rate": 7.267858543087901e-05, "loss": 2.9056, "step": 34843 }, { "epoch": 2.73, "learning_rate": 7.267780130165451e-05, "loss": 2.6487, "step": 34844 }, { "epoch": 2.73, "learning_rate": 7.267701717243002e-05, "loss": 2.9357, "step": 34845 }, { "epoch": 2.73, "learning_rate": 7.267623304320552e-05, "loss": 2.8134, "step": 34846 }, { "epoch": 2.73, "learning_rate": 7.267544891398103e-05, "loss": 2.7059, "step": 34847 }, { "epoch": 2.73, "learning_rate": 7.267466478475653e-05, "loss": 2.5582, "step": 34848 }, { "epoch": 2.73, "learning_rate": 7.267388065553203e-05, "loss": 2.5888, "step": 34849 }, { "epoch": 2.73, "learning_rate": 7.267309652630754e-05, "loss": 2.8793, "step": 34850 }, { "epoch": 2.73, "learning_rate": 7.267231239708304e-05, "loss": 2.8413, "step": 34851 }, { "epoch": 2.73, "learning_rate": 7.267152826785854e-05, "loss": 2.8319, "step": 34852 }, { "epoch": 2.73, "learning_rate": 7.267074413863405e-05, "loss": 2.7456, "step": 34853 }, { "epoch": 2.73, "learning_rate": 7.266996000940955e-05, "loss": 2.7257, "step": 34854 }, { "epoch": 2.73, "learning_rate": 7.266917588018505e-05, "loss": 2.7857, "step": 34855 }, { "epoch": 2.73, "learning_rate": 7.266839175096056e-05, "loss": 2.831, "step": 34856 }, { "epoch": 2.73, "learning_rate": 7.266760762173606e-05, "loss": 2.7436, "step": 34857 }, { "epoch": 2.73, "learning_rate": 7.266682349251157e-05, "loss": 2.9553, "step": 34858 }, { "epoch": 2.73, "learning_rate": 7.266603936328707e-05, "loss": 2.8051, "step": 34859 }, { "epoch": 2.73, "learning_rate": 7.266525523406257e-05, "loss": 2.6393, "step": 34860 }, { "epoch": 2.73, "learning_rate": 7.266447110483808e-05, "loss": 2.7134, "step": 34861 }, { "epoch": 2.73, "learning_rate": 7.26636869756136e-05, "loss": 2.6477, "step": 34862 }, { "epoch": 2.73, "learning_rate": 7.26629028463891e-05, "loss": 2.7616, "step": 34863 }, { "epoch": 2.73, "learning_rate": 7.26621187171646e-05, "loss": 2.746, "step": 34864 }, { "epoch": 2.73, "learning_rate": 7.26613345879401e-05, "loss": 2.7664, "step": 34865 }, { "epoch": 2.73, "learning_rate": 7.266055045871561e-05, "loss": 2.8919, "step": 34866 }, { "epoch": 2.73, "learning_rate": 7.265976632949111e-05, "loss": 2.902, "step": 34867 }, { "epoch": 2.73, "learning_rate": 7.265898220026662e-05, "loss": 2.7038, "step": 34868 }, { "epoch": 2.73, "learning_rate": 7.265819807104212e-05, "loss": 2.7611, "step": 34869 }, { "epoch": 2.73, "learning_rate": 7.265741394181762e-05, "loss": 3.0204, "step": 34870 }, { "epoch": 2.73, "learning_rate": 7.265662981259311e-05, "loss": 3.1489, "step": 34871 }, { "epoch": 2.73, "learning_rate": 7.265584568336862e-05, "loss": 2.649, "step": 34872 }, { "epoch": 2.73, "learning_rate": 7.265506155414412e-05, "loss": 2.8392, "step": 34873 }, { "epoch": 2.73, "learning_rate": 7.265427742491962e-05, "loss": 2.9808, "step": 34874 }, { "epoch": 2.73, "learning_rate": 7.265349329569513e-05, "loss": 2.7343, "step": 34875 }, { "epoch": 2.73, "learning_rate": 7.265270916647063e-05, "loss": 2.9898, "step": 34876 }, { "epoch": 2.73, "learning_rate": 7.265192503724614e-05, "loss": 2.9109, "step": 34877 }, { "epoch": 2.73, "learning_rate": 7.265114090802164e-05, "loss": 2.9442, "step": 34878 }, { "epoch": 2.73, "learning_rate": 7.265035677879714e-05, "loss": 2.8988, "step": 34879 }, { "epoch": 2.74, "learning_rate": 7.264957264957265e-05, "loss": 2.5714, "step": 34880 }, { "epoch": 2.74, "learning_rate": 7.264878852034815e-05, "loss": 3.0858, "step": 34881 }, { "epoch": 2.74, "learning_rate": 7.264800439112365e-05, "loss": 2.78, "step": 34882 }, { "epoch": 2.74, "learning_rate": 7.264722026189916e-05, "loss": 2.9309, "step": 34883 }, { "epoch": 2.74, "learning_rate": 7.264643613267466e-05, "loss": 2.8502, "step": 34884 }, { "epoch": 2.74, "learning_rate": 7.264565200345017e-05, "loss": 2.8316, "step": 34885 }, { "epoch": 2.74, "learning_rate": 7.264486787422567e-05, "loss": 3.0234, "step": 34886 }, { "epoch": 2.74, "learning_rate": 7.264408374500117e-05, "loss": 2.7353, "step": 34887 }, { "epoch": 2.74, "learning_rate": 7.264329961577669e-05, "loss": 2.6842, "step": 34888 }, { "epoch": 2.74, "learning_rate": 7.26425154865522e-05, "loss": 2.7888, "step": 34889 }, { "epoch": 2.74, "learning_rate": 7.26417313573277e-05, "loss": 2.8629, "step": 34890 }, { "epoch": 2.74, "learning_rate": 7.26409472281032e-05, "loss": 2.8527, "step": 34891 }, { "epoch": 2.74, "learning_rate": 7.26401630988787e-05, "loss": 2.7859, "step": 34892 }, { "epoch": 2.74, "learning_rate": 7.263937896965421e-05, "loss": 2.6707, "step": 34893 }, { "epoch": 2.74, "learning_rate": 7.263859484042971e-05, "loss": 2.8142, "step": 34894 }, { "epoch": 2.74, "learning_rate": 7.263781071120522e-05, "loss": 2.6797, "step": 34895 }, { "epoch": 2.74, "learning_rate": 7.263702658198072e-05, "loss": 2.7762, "step": 34896 }, { "epoch": 2.74, "learning_rate": 7.263624245275622e-05, "loss": 2.706, "step": 34897 }, { "epoch": 2.74, "learning_rate": 7.263545832353173e-05, "loss": 2.7079, "step": 34898 }, { "epoch": 2.74, "learning_rate": 7.263467419430723e-05, "loss": 3.1465, "step": 34899 }, { "epoch": 2.74, "learning_rate": 7.263389006508273e-05, "loss": 2.7679, "step": 34900 }, { "epoch": 2.74, "learning_rate": 7.263310593585824e-05, "loss": 2.7023, "step": 34901 }, { "epoch": 2.74, "learning_rate": 7.263232180663374e-05, "loss": 2.8613, "step": 34902 }, { "epoch": 2.74, "learning_rate": 7.263153767740924e-05, "loss": 2.8486, "step": 34903 }, { "epoch": 2.74, "learning_rate": 7.263075354818475e-05, "loss": 2.9412, "step": 34904 }, { "epoch": 2.74, "learning_rate": 7.262996941896025e-05, "loss": 2.7531, "step": 34905 }, { "epoch": 2.74, "learning_rate": 7.262918528973576e-05, "loss": 2.8798, "step": 34906 }, { "epoch": 2.74, "learning_rate": 7.262840116051125e-05, "loss": 2.8876, "step": 34907 }, { "epoch": 2.74, "learning_rate": 7.262761703128675e-05, "loss": 2.7585, "step": 34908 }, { "epoch": 2.74, "learning_rate": 7.262683290206225e-05, "loss": 2.6611, "step": 34909 }, { "epoch": 2.74, "learning_rate": 7.262604877283776e-05, "loss": 2.6785, "step": 34910 }, { "epoch": 2.74, "learning_rate": 7.262526464361326e-05, "loss": 2.9159, "step": 34911 }, { "epoch": 2.74, "learning_rate": 7.262448051438876e-05, "loss": 2.8912, "step": 34912 }, { "epoch": 2.74, "learning_rate": 7.262369638516427e-05, "loss": 2.7981, "step": 34913 }, { "epoch": 2.74, "learning_rate": 7.262291225593979e-05, "loss": 2.7472, "step": 34914 }, { "epoch": 2.74, "learning_rate": 7.262212812671529e-05, "loss": 2.9483, "step": 34915 }, { "epoch": 2.74, "learning_rate": 7.262134399749079e-05, "loss": 2.8618, "step": 34916 }, { "epoch": 2.74, "learning_rate": 7.26205598682663e-05, "loss": 3.1987, "step": 34917 }, { "epoch": 2.74, "learning_rate": 7.26197757390418e-05, "loss": 2.7225, "step": 34918 }, { "epoch": 2.74, "learning_rate": 7.26189916098173e-05, "loss": 2.7659, "step": 34919 }, { "epoch": 2.74, "learning_rate": 7.261820748059281e-05, "loss": 2.8124, "step": 34920 }, { "epoch": 2.74, "learning_rate": 7.261742335136831e-05, "loss": 2.8583, "step": 34921 }, { "epoch": 2.74, "learning_rate": 7.261663922214381e-05, "loss": 2.7681, "step": 34922 }, { "epoch": 2.74, "learning_rate": 7.261585509291932e-05, "loss": 2.9407, "step": 34923 }, { "epoch": 2.74, "learning_rate": 7.261507096369482e-05, "loss": 3.0049, "step": 34924 }, { "epoch": 2.74, "learning_rate": 7.261428683447033e-05, "loss": 2.7039, "step": 34925 }, { "epoch": 2.74, "learning_rate": 7.261350270524583e-05, "loss": 2.7783, "step": 34926 }, { "epoch": 2.74, "learning_rate": 7.261271857602133e-05, "loss": 2.7062, "step": 34927 }, { "epoch": 2.74, "learning_rate": 7.261193444679684e-05, "loss": 2.8648, "step": 34928 }, { "epoch": 2.74, "learning_rate": 7.261115031757234e-05, "loss": 2.8193, "step": 34929 }, { "epoch": 2.74, "learning_rate": 7.261036618834784e-05, "loss": 2.6438, "step": 34930 }, { "epoch": 2.74, "learning_rate": 7.260958205912335e-05, "loss": 2.6214, "step": 34931 }, { "epoch": 2.74, "learning_rate": 7.260879792989885e-05, "loss": 2.7828, "step": 34932 }, { "epoch": 2.74, "learning_rate": 7.260801380067436e-05, "loss": 2.716, "step": 34933 }, { "epoch": 2.74, "learning_rate": 7.260722967144986e-05, "loss": 2.4899, "step": 34934 }, { "epoch": 2.74, "learning_rate": 7.260644554222536e-05, "loss": 2.6285, "step": 34935 }, { "epoch": 2.74, "learning_rate": 7.260566141300087e-05, "loss": 2.713, "step": 34936 }, { "epoch": 2.74, "learning_rate": 7.260487728377637e-05, "loss": 2.6381, "step": 34937 }, { "epoch": 2.74, "learning_rate": 7.260409315455187e-05, "loss": 2.9333, "step": 34938 }, { "epoch": 2.74, "learning_rate": 7.260330902532738e-05, "loss": 3.1273, "step": 34939 }, { "epoch": 2.74, "learning_rate": 7.260252489610288e-05, "loss": 2.7244, "step": 34940 }, { "epoch": 2.74, "learning_rate": 7.260174076687838e-05, "loss": 2.7591, "step": 34941 }, { "epoch": 2.74, "learning_rate": 7.260095663765389e-05, "loss": 2.7528, "step": 34942 }, { "epoch": 2.74, "learning_rate": 7.260017250842939e-05, "loss": 2.7409, "step": 34943 }, { "epoch": 2.74, "learning_rate": 7.25993883792049e-05, "loss": 2.6664, "step": 34944 }, { "epoch": 2.74, "learning_rate": 7.25986042499804e-05, "loss": 2.7601, "step": 34945 }, { "epoch": 2.74, "learning_rate": 7.25978201207559e-05, "loss": 2.6751, "step": 34946 }, { "epoch": 2.74, "learning_rate": 7.25970359915314e-05, "loss": 3.107, "step": 34947 }, { "epoch": 2.74, "learning_rate": 7.259625186230691e-05, "loss": 2.9409, "step": 34948 }, { "epoch": 2.74, "learning_rate": 7.259546773308241e-05, "loss": 2.7965, "step": 34949 }, { "epoch": 2.74, "learning_rate": 7.259468360385792e-05, "loss": 2.6043, "step": 34950 }, { "epoch": 2.74, "learning_rate": 7.259389947463342e-05, "loss": 3.0004, "step": 34951 }, { "epoch": 2.74, "learning_rate": 7.259311534540893e-05, "loss": 2.8534, "step": 34952 }, { "epoch": 2.74, "learning_rate": 7.259233121618443e-05, "loss": 2.6053, "step": 34953 }, { "epoch": 2.74, "learning_rate": 7.259154708695993e-05, "loss": 2.9234, "step": 34954 }, { "epoch": 2.74, "learning_rate": 7.259076295773544e-05, "loss": 2.7278, "step": 34955 }, { "epoch": 2.74, "learning_rate": 7.258997882851094e-05, "loss": 2.7279, "step": 34956 }, { "epoch": 2.74, "learning_rate": 7.258919469928644e-05, "loss": 2.8812, "step": 34957 }, { "epoch": 2.74, "learning_rate": 7.258841057006195e-05, "loss": 2.7853, "step": 34958 }, { "epoch": 2.74, "learning_rate": 7.258762644083745e-05, "loss": 2.7089, "step": 34959 }, { "epoch": 2.74, "learning_rate": 7.258684231161295e-05, "loss": 2.7898, "step": 34960 }, { "epoch": 2.74, "learning_rate": 7.258605818238846e-05, "loss": 2.6854, "step": 34961 }, { "epoch": 2.74, "learning_rate": 7.258527405316396e-05, "loss": 2.7912, "step": 34962 }, { "epoch": 2.74, "learning_rate": 7.258448992393947e-05, "loss": 3.1004, "step": 34963 }, { "epoch": 2.74, "learning_rate": 7.258370579471497e-05, "loss": 2.7514, "step": 34964 }, { "epoch": 2.74, "learning_rate": 7.258292166549047e-05, "loss": 2.8601, "step": 34965 }, { "epoch": 2.74, "learning_rate": 7.258213753626598e-05, "loss": 2.7385, "step": 34966 }, { "epoch": 2.74, "learning_rate": 7.258135340704148e-05, "loss": 2.9625, "step": 34967 }, { "epoch": 2.74, "learning_rate": 7.258056927781698e-05, "loss": 2.6291, "step": 34968 }, { "epoch": 2.74, "learning_rate": 7.257978514859249e-05, "loss": 2.6927, "step": 34969 }, { "epoch": 2.74, "learning_rate": 7.257900101936799e-05, "loss": 2.7317, "step": 34970 }, { "epoch": 2.74, "learning_rate": 7.257821689014351e-05, "loss": 2.8899, "step": 34971 }, { "epoch": 2.74, "learning_rate": 7.257743276091901e-05, "loss": 2.7662, "step": 34972 }, { "epoch": 2.74, "learning_rate": 7.257664863169452e-05, "loss": 2.5896, "step": 34973 }, { "epoch": 2.74, "learning_rate": 7.257586450247002e-05, "loss": 2.9494, "step": 34974 }, { "epoch": 2.74, "learning_rate": 7.257508037324552e-05, "loss": 2.7819, "step": 34975 }, { "epoch": 2.74, "learning_rate": 7.257429624402103e-05, "loss": 2.8529, "step": 34976 }, { "epoch": 2.74, "learning_rate": 7.257351211479653e-05, "loss": 2.6982, "step": 34977 }, { "epoch": 2.74, "learning_rate": 7.257272798557203e-05, "loss": 2.7729, "step": 34978 }, { "epoch": 2.74, "learning_rate": 7.257194385634752e-05, "loss": 2.8445, "step": 34979 }, { "epoch": 2.74, "learning_rate": 7.257115972712303e-05, "loss": 2.7122, "step": 34980 }, { "epoch": 2.74, "learning_rate": 7.257037559789853e-05, "loss": 2.7482, "step": 34981 }, { "epoch": 2.74, "learning_rate": 7.256959146867404e-05, "loss": 2.8754, "step": 34982 }, { "epoch": 2.74, "learning_rate": 7.256880733944954e-05, "loss": 2.6864, "step": 34983 }, { "epoch": 2.74, "learning_rate": 7.256802321022504e-05, "loss": 2.6353, "step": 34984 }, { "epoch": 2.74, "learning_rate": 7.256723908100055e-05, "loss": 2.7681, "step": 34985 }, { "epoch": 2.74, "learning_rate": 7.256645495177605e-05, "loss": 2.7265, "step": 34986 }, { "epoch": 2.74, "learning_rate": 7.256567082255155e-05, "loss": 2.7356, "step": 34987 }, { "epoch": 2.74, "learning_rate": 7.256488669332706e-05, "loss": 2.8244, "step": 34988 }, { "epoch": 2.74, "learning_rate": 7.256410256410256e-05, "loss": 2.7602, "step": 34989 }, { "epoch": 2.74, "learning_rate": 7.256331843487806e-05, "loss": 2.7515, "step": 34990 }, { "epoch": 2.74, "learning_rate": 7.256253430565357e-05, "loss": 2.7757, "step": 34991 }, { "epoch": 2.74, "learning_rate": 7.256175017642907e-05, "loss": 2.7431, "step": 34992 }, { "epoch": 2.74, "learning_rate": 7.256096604720458e-05, "loss": 2.7421, "step": 34993 }, { "epoch": 2.74, "learning_rate": 7.256018191798008e-05, "loss": 2.6777, "step": 34994 }, { "epoch": 2.74, "learning_rate": 7.255939778875558e-05, "loss": 2.9548, "step": 34995 }, { "epoch": 2.74, "learning_rate": 7.25586136595311e-05, "loss": 2.8765, "step": 34996 }, { "epoch": 2.74, "learning_rate": 7.25578295303066e-05, "loss": 2.8954, "step": 34997 }, { "epoch": 2.74, "learning_rate": 7.255704540108211e-05, "loss": 2.7556, "step": 34998 }, { "epoch": 2.74, "learning_rate": 7.255626127185761e-05, "loss": 2.6799, "step": 34999 }, { "epoch": 2.74, "learning_rate": 7.255547714263312e-05, "loss": 2.5067, "step": 35000 }, { "epoch": 2.74, "eval_loss": 2.893690824508667, "eval_runtime": 217.8498, "eval_samples_per_second": 52.017, "eval_steps_per_second": 3.255, "step": 35000 }, { "epoch": 2.74, "learning_rate": 7.255469301340862e-05, "loss": 3.0317, "step": 35001 }, { "epoch": 2.74, "learning_rate": 7.255390888418412e-05, "loss": 2.7113, "step": 35002 }, { "epoch": 2.74, "learning_rate": 7.255312475495963e-05, "loss": 2.8032, "step": 35003 }, { "epoch": 2.74, "learning_rate": 7.255234062573513e-05, "loss": 2.6997, "step": 35004 }, { "epoch": 2.74, "learning_rate": 7.255155649651063e-05, "loss": 3.1312, "step": 35005 }, { "epoch": 2.74, "learning_rate": 7.255077236728614e-05, "loss": 2.657, "step": 35006 }, { "epoch": 2.75, "learning_rate": 7.254998823806164e-05, "loss": 2.67, "step": 35007 }, { "epoch": 2.75, "learning_rate": 7.254920410883714e-05, "loss": 2.8263, "step": 35008 }, { "epoch": 2.75, "learning_rate": 7.254841997961265e-05, "loss": 2.5567, "step": 35009 }, { "epoch": 2.75, "learning_rate": 7.254763585038815e-05, "loss": 2.8096, "step": 35010 }, { "epoch": 2.75, "learning_rate": 7.254685172116366e-05, "loss": 2.9548, "step": 35011 }, { "epoch": 2.75, "learning_rate": 7.254606759193916e-05, "loss": 2.8665, "step": 35012 }, { "epoch": 2.75, "learning_rate": 7.254528346271466e-05, "loss": 2.7243, "step": 35013 }, { "epoch": 2.75, "learning_rate": 7.254449933349017e-05, "loss": 2.9762, "step": 35014 }, { "epoch": 2.75, "learning_rate": 7.254371520426566e-05, "loss": 2.9602, "step": 35015 }, { "epoch": 2.75, "learning_rate": 7.254293107504116e-05, "loss": 2.7228, "step": 35016 }, { "epoch": 2.75, "learning_rate": 7.254214694581666e-05, "loss": 3.0226, "step": 35017 }, { "epoch": 2.75, "learning_rate": 7.254136281659217e-05, "loss": 3.0187, "step": 35018 }, { "epoch": 2.75, "learning_rate": 7.254057868736767e-05, "loss": 2.8838, "step": 35019 }, { "epoch": 2.75, "learning_rate": 7.253979455814318e-05, "loss": 2.7951, "step": 35020 }, { "epoch": 2.75, "learning_rate": 7.253901042891868e-05, "loss": 2.8035, "step": 35021 }, { "epoch": 2.75, "learning_rate": 7.25382262996942e-05, "loss": 2.6573, "step": 35022 }, { "epoch": 2.75, "learning_rate": 7.25374421704697e-05, "loss": 2.6888, "step": 35023 }, { "epoch": 2.75, "learning_rate": 7.25366580412452e-05, "loss": 2.769, "step": 35024 }, { "epoch": 2.75, "learning_rate": 7.253587391202071e-05, "loss": 2.6835, "step": 35025 }, { "epoch": 2.75, "learning_rate": 7.253508978279621e-05, "loss": 3.0052, "step": 35026 }, { "epoch": 2.75, "learning_rate": 7.253430565357171e-05, "loss": 2.6639, "step": 35027 }, { "epoch": 2.75, "learning_rate": 7.253352152434722e-05, "loss": 2.936, "step": 35028 }, { "epoch": 2.75, "learning_rate": 7.253273739512272e-05, "loss": 2.8974, "step": 35029 }, { "epoch": 2.75, "learning_rate": 7.253195326589823e-05, "loss": 2.711, "step": 35030 }, { "epoch": 2.75, "learning_rate": 7.253116913667373e-05, "loss": 2.7825, "step": 35031 }, { "epoch": 2.75, "learning_rate": 7.253038500744923e-05, "loss": 2.7032, "step": 35032 }, { "epoch": 2.75, "learning_rate": 7.252960087822474e-05, "loss": 2.9202, "step": 35033 }, { "epoch": 2.75, "learning_rate": 7.252881674900024e-05, "loss": 2.6781, "step": 35034 }, { "epoch": 2.75, "learning_rate": 7.252803261977574e-05, "loss": 2.8807, "step": 35035 }, { "epoch": 2.75, "learning_rate": 7.252724849055125e-05, "loss": 2.7663, "step": 35036 }, { "epoch": 2.75, "learning_rate": 7.252646436132675e-05, "loss": 2.7123, "step": 35037 }, { "epoch": 2.75, "learning_rate": 7.252568023210225e-05, "loss": 2.8366, "step": 35038 }, { "epoch": 2.75, "learning_rate": 7.252489610287776e-05, "loss": 2.6138, "step": 35039 }, { "epoch": 2.75, "learning_rate": 7.252411197365326e-05, "loss": 2.6365, "step": 35040 }, { "epoch": 2.75, "learning_rate": 7.252332784442877e-05, "loss": 2.8033, "step": 35041 }, { "epoch": 2.75, "learning_rate": 7.252254371520427e-05, "loss": 2.9533, "step": 35042 }, { "epoch": 2.75, "learning_rate": 7.252175958597977e-05, "loss": 2.7739, "step": 35043 }, { "epoch": 2.75, "learning_rate": 7.252097545675528e-05, "loss": 2.7589, "step": 35044 }, { "epoch": 2.75, "learning_rate": 7.252019132753078e-05, "loss": 2.7295, "step": 35045 }, { "epoch": 2.75, "learning_rate": 7.251940719830628e-05, "loss": 3.1044, "step": 35046 }, { "epoch": 2.75, "learning_rate": 7.251862306908179e-05, "loss": 2.8974, "step": 35047 }, { "epoch": 2.75, "learning_rate": 7.251783893985729e-05, "loss": 2.7368, "step": 35048 }, { "epoch": 2.75, "learning_rate": 7.25170548106328e-05, "loss": 2.8563, "step": 35049 }, { "epoch": 2.75, "learning_rate": 7.25162706814083e-05, "loss": 2.9512, "step": 35050 }, { "epoch": 2.75, "learning_rate": 7.25154865521838e-05, "loss": 2.7691, "step": 35051 }, { "epoch": 2.75, "learning_rate": 7.25147024229593e-05, "loss": 2.8799, "step": 35052 }, { "epoch": 2.75, "learning_rate": 7.251391829373481e-05, "loss": 2.9692, "step": 35053 }, { "epoch": 2.75, "learning_rate": 7.251313416451031e-05, "loss": 2.7984, "step": 35054 }, { "epoch": 2.75, "learning_rate": 7.251235003528582e-05, "loss": 2.9262, "step": 35055 }, { "epoch": 2.75, "learning_rate": 7.251156590606132e-05, "loss": 2.7891, "step": 35056 }, { "epoch": 2.75, "learning_rate": 7.251078177683682e-05, "loss": 2.8987, "step": 35057 }, { "epoch": 2.75, "learning_rate": 7.250999764761233e-05, "loss": 2.9744, "step": 35058 }, { "epoch": 2.75, "learning_rate": 7.250921351838783e-05, "loss": 2.6406, "step": 35059 }, { "epoch": 2.75, "learning_rate": 7.250842938916334e-05, "loss": 2.8428, "step": 35060 }, { "epoch": 2.75, "learning_rate": 7.250764525993884e-05, "loss": 2.8202, "step": 35061 }, { "epoch": 2.75, "learning_rate": 7.250686113071434e-05, "loss": 2.9572, "step": 35062 }, { "epoch": 2.75, "learning_rate": 7.250607700148985e-05, "loss": 2.6882, "step": 35063 }, { "epoch": 2.75, "learning_rate": 7.250529287226535e-05, "loss": 2.7551, "step": 35064 }, { "epoch": 2.75, "learning_rate": 7.250450874304085e-05, "loss": 2.9282, "step": 35065 }, { "epoch": 2.75, "learning_rate": 7.250372461381636e-05, "loss": 2.7451, "step": 35066 }, { "epoch": 2.75, "learning_rate": 7.250294048459186e-05, "loss": 2.7819, "step": 35067 }, { "epoch": 2.75, "learning_rate": 7.250215635536737e-05, "loss": 2.8172, "step": 35068 }, { "epoch": 2.75, "learning_rate": 7.250137222614287e-05, "loss": 2.8635, "step": 35069 }, { "epoch": 2.75, "learning_rate": 7.250058809691837e-05, "loss": 2.8693, "step": 35070 }, { "epoch": 2.75, "learning_rate": 7.249980396769388e-05, "loss": 2.907, "step": 35071 }, { "epoch": 2.75, "learning_rate": 7.249901983846938e-05, "loss": 2.8231, "step": 35072 }, { "epoch": 2.75, "learning_rate": 7.249823570924488e-05, "loss": 2.7088, "step": 35073 }, { "epoch": 2.75, "learning_rate": 7.249745158002039e-05, "loss": 2.6154, "step": 35074 }, { "epoch": 2.75, "learning_rate": 7.249666745079589e-05, "loss": 2.9156, "step": 35075 }, { "epoch": 2.75, "learning_rate": 7.24958833215714e-05, "loss": 2.9743, "step": 35076 }, { "epoch": 2.75, "learning_rate": 7.24950991923469e-05, "loss": 3.1006, "step": 35077 }, { "epoch": 2.75, "learning_rate": 7.24943150631224e-05, "loss": 2.6786, "step": 35078 }, { "epoch": 2.75, "learning_rate": 7.249353093389792e-05, "loss": 2.7668, "step": 35079 }, { "epoch": 2.75, "learning_rate": 7.249274680467342e-05, "loss": 2.9469, "step": 35080 }, { "epoch": 2.75, "learning_rate": 7.249196267544893e-05, "loss": 2.6323, "step": 35081 }, { "epoch": 2.75, "learning_rate": 7.249117854622443e-05, "loss": 2.7852, "step": 35082 }, { "epoch": 2.75, "learning_rate": 7.249039441699993e-05, "loss": 2.6522, "step": 35083 }, { "epoch": 2.75, "learning_rate": 7.248961028777544e-05, "loss": 2.986, "step": 35084 }, { "epoch": 2.75, "learning_rate": 7.248882615855094e-05, "loss": 3.0198, "step": 35085 }, { "epoch": 2.75, "learning_rate": 7.248804202932645e-05, "loss": 2.6198, "step": 35086 }, { "epoch": 2.75, "learning_rate": 7.248725790010194e-05, "loss": 2.75, "step": 35087 }, { "epoch": 2.75, "learning_rate": 7.248647377087744e-05, "loss": 2.719, "step": 35088 }, { "epoch": 2.75, "learning_rate": 7.248568964165294e-05, "loss": 2.8402, "step": 35089 }, { "epoch": 2.75, "learning_rate": 7.248490551242845e-05, "loss": 2.8983, "step": 35090 }, { "epoch": 2.75, "learning_rate": 7.248412138320395e-05, "loss": 3.1716, "step": 35091 }, { "epoch": 2.75, "learning_rate": 7.248333725397945e-05, "loss": 2.8315, "step": 35092 }, { "epoch": 2.75, "learning_rate": 7.248255312475496e-05, "loss": 2.8805, "step": 35093 }, { "epoch": 2.75, "learning_rate": 7.248176899553046e-05, "loss": 2.8256, "step": 35094 }, { "epoch": 2.75, "learning_rate": 7.248098486630596e-05, "loss": 2.7378, "step": 35095 }, { "epoch": 2.75, "learning_rate": 7.248020073708147e-05, "loss": 2.6579, "step": 35096 }, { "epoch": 2.75, "learning_rate": 7.247941660785697e-05, "loss": 2.7313, "step": 35097 }, { "epoch": 2.75, "learning_rate": 7.247863247863248e-05, "loss": 2.7355, "step": 35098 }, { "epoch": 2.75, "learning_rate": 7.247784834940798e-05, "loss": 2.9499, "step": 35099 }, { "epoch": 2.75, "learning_rate": 7.247706422018348e-05, "loss": 2.8814, "step": 35100 }, { "epoch": 2.75, "learning_rate": 7.247628009095899e-05, "loss": 2.8565, "step": 35101 }, { "epoch": 2.75, "learning_rate": 7.247549596173449e-05, "loss": 2.6626, "step": 35102 }, { "epoch": 2.75, "learning_rate": 7.247471183251e-05, "loss": 2.7417, "step": 35103 }, { "epoch": 2.75, "learning_rate": 7.247392770328551e-05, "loss": 3.0608, "step": 35104 }, { "epoch": 2.75, "learning_rate": 7.247314357406101e-05, "loss": 3.0108, "step": 35105 }, { "epoch": 2.75, "learning_rate": 7.247235944483652e-05, "loss": 3.0178, "step": 35106 }, { "epoch": 2.75, "learning_rate": 7.247157531561202e-05, "loss": 2.8392, "step": 35107 }, { "epoch": 2.75, "learning_rate": 7.247079118638753e-05, "loss": 2.833, "step": 35108 }, { "epoch": 2.75, "learning_rate": 7.247000705716303e-05, "loss": 2.7885, "step": 35109 }, { "epoch": 2.75, "learning_rate": 7.246922292793853e-05, "loss": 3.0065, "step": 35110 }, { "epoch": 2.75, "learning_rate": 7.246843879871404e-05, "loss": 2.7174, "step": 35111 }, { "epoch": 2.75, "learning_rate": 7.246765466948954e-05, "loss": 2.8321, "step": 35112 }, { "epoch": 2.75, "learning_rate": 7.246687054026504e-05, "loss": 2.7221, "step": 35113 }, { "epoch": 2.75, "learning_rate": 7.246608641104055e-05, "loss": 2.8642, "step": 35114 }, { "epoch": 2.75, "learning_rate": 7.246530228181605e-05, "loss": 2.7412, "step": 35115 }, { "epoch": 2.75, "learning_rate": 7.246451815259156e-05, "loss": 2.816, "step": 35116 }, { "epoch": 2.75, "learning_rate": 7.246373402336706e-05, "loss": 2.8627, "step": 35117 }, { "epoch": 2.75, "learning_rate": 7.246294989414256e-05, "loss": 2.6665, "step": 35118 }, { "epoch": 2.75, "learning_rate": 7.246216576491807e-05, "loss": 2.8242, "step": 35119 }, { "epoch": 2.75, "learning_rate": 7.246138163569357e-05, "loss": 2.6882, "step": 35120 }, { "epoch": 2.75, "learning_rate": 7.246059750646907e-05, "loss": 2.756, "step": 35121 }, { "epoch": 2.75, "learning_rate": 7.245981337724458e-05, "loss": 2.7192, "step": 35122 }, { "epoch": 2.75, "learning_rate": 7.245902924802007e-05, "loss": 3.1871, "step": 35123 }, { "epoch": 2.75, "learning_rate": 7.245824511879557e-05, "loss": 2.7249, "step": 35124 }, { "epoch": 2.75, "learning_rate": 7.245746098957107e-05, "loss": 2.8177, "step": 35125 }, { "epoch": 2.75, "learning_rate": 7.245667686034658e-05, "loss": 2.7055, "step": 35126 }, { "epoch": 2.75, "learning_rate": 7.245589273112208e-05, "loss": 2.8186, "step": 35127 }, { "epoch": 2.75, "learning_rate": 7.245510860189759e-05, "loss": 2.798, "step": 35128 }, { "epoch": 2.75, "learning_rate": 7.245432447267309e-05, "loss": 2.7955, "step": 35129 }, { "epoch": 2.75, "learning_rate": 7.245354034344861e-05, "loss": 2.8472, "step": 35130 }, { "epoch": 2.75, "learning_rate": 7.245275621422411e-05, "loss": 3.0878, "step": 35131 }, { "epoch": 2.75, "learning_rate": 7.245197208499961e-05, "loss": 2.8651, "step": 35132 }, { "epoch": 2.75, "learning_rate": 7.245118795577512e-05, "loss": 2.8569, "step": 35133 }, { "epoch": 2.75, "learning_rate": 7.245040382655062e-05, "loss": 2.8227, "step": 35134 }, { "epoch": 2.76, "learning_rate": 7.244961969732613e-05, "loss": 2.7807, "step": 35135 }, { "epoch": 2.76, "learning_rate": 7.244883556810163e-05, "loss": 2.9334, "step": 35136 }, { "epoch": 2.76, "learning_rate": 7.244805143887713e-05, "loss": 2.646, "step": 35137 }, { "epoch": 2.76, "learning_rate": 7.244726730965264e-05, "loss": 2.6993, "step": 35138 }, { "epoch": 2.76, "learning_rate": 7.244648318042814e-05, "loss": 2.6124, "step": 35139 }, { "epoch": 2.76, "learning_rate": 7.244569905120364e-05, "loss": 2.6987, "step": 35140 }, { "epoch": 2.76, "learning_rate": 7.244491492197915e-05, "loss": 2.7621, "step": 35141 }, { "epoch": 2.76, "learning_rate": 7.244413079275465e-05, "loss": 2.8448, "step": 35142 }, { "epoch": 2.76, "learning_rate": 7.244334666353015e-05, "loss": 2.9077, "step": 35143 }, { "epoch": 2.76, "learning_rate": 7.244256253430566e-05, "loss": 2.8629, "step": 35144 }, { "epoch": 2.76, "learning_rate": 7.244177840508116e-05, "loss": 2.6676, "step": 35145 }, { "epoch": 2.76, "learning_rate": 7.244099427585667e-05, "loss": 2.6308, "step": 35146 }, { "epoch": 2.76, "learning_rate": 7.244021014663217e-05, "loss": 2.7184, "step": 35147 }, { "epoch": 2.76, "learning_rate": 7.243942601740767e-05, "loss": 2.7424, "step": 35148 }, { "epoch": 2.76, "learning_rate": 7.243864188818318e-05, "loss": 2.772, "step": 35149 }, { "epoch": 2.76, "learning_rate": 7.243785775895868e-05, "loss": 2.687, "step": 35150 }, { "epoch": 2.76, "learning_rate": 7.243707362973418e-05, "loss": 2.8545, "step": 35151 }, { "epoch": 2.76, "learning_rate": 7.243628950050969e-05, "loss": 2.7761, "step": 35152 }, { "epoch": 2.76, "learning_rate": 7.243550537128519e-05, "loss": 2.8572, "step": 35153 }, { "epoch": 2.76, "learning_rate": 7.24347212420607e-05, "loss": 2.8692, "step": 35154 }, { "epoch": 2.76, "learning_rate": 7.24339371128362e-05, "loss": 2.7692, "step": 35155 }, { "epoch": 2.76, "learning_rate": 7.24331529836117e-05, "loss": 2.6957, "step": 35156 }, { "epoch": 2.76, "learning_rate": 7.24323688543872e-05, "loss": 2.9095, "step": 35157 }, { "epoch": 2.76, "learning_rate": 7.243158472516271e-05, "loss": 2.7268, "step": 35158 }, { "epoch": 2.76, "learning_rate": 7.243080059593821e-05, "loss": 2.7256, "step": 35159 }, { "epoch": 2.76, "learning_rate": 7.243001646671372e-05, "loss": 2.9017, "step": 35160 }, { "epoch": 2.76, "learning_rate": 7.242923233748922e-05, "loss": 2.8337, "step": 35161 }, { "epoch": 2.76, "learning_rate": 7.242844820826472e-05, "loss": 2.6054, "step": 35162 }, { "epoch": 2.76, "learning_rate": 7.242766407904023e-05, "loss": 2.8799, "step": 35163 }, { "epoch": 2.76, "learning_rate": 7.242687994981573e-05, "loss": 2.8886, "step": 35164 }, { "epoch": 2.76, "learning_rate": 7.242609582059124e-05, "loss": 2.6967, "step": 35165 }, { "epoch": 2.76, "learning_rate": 7.242531169136674e-05, "loss": 3.0367, "step": 35166 }, { "epoch": 2.76, "learning_rate": 7.242452756214224e-05, "loss": 2.774, "step": 35167 }, { "epoch": 2.76, "learning_rate": 7.242374343291775e-05, "loss": 2.8582, "step": 35168 }, { "epoch": 2.76, "learning_rate": 7.242295930369325e-05, "loss": 2.6815, "step": 35169 }, { "epoch": 2.76, "learning_rate": 7.242217517446875e-05, "loss": 2.63, "step": 35170 }, { "epoch": 2.76, "learning_rate": 7.242139104524426e-05, "loss": 2.7436, "step": 35171 }, { "epoch": 2.76, "learning_rate": 7.242060691601976e-05, "loss": 2.7527, "step": 35172 }, { "epoch": 2.76, "learning_rate": 7.241982278679527e-05, "loss": 2.5928, "step": 35173 }, { "epoch": 2.76, "learning_rate": 7.241903865757077e-05, "loss": 2.8559, "step": 35174 }, { "epoch": 2.76, "learning_rate": 7.241825452834627e-05, "loss": 3.0288, "step": 35175 }, { "epoch": 2.76, "learning_rate": 7.241747039912178e-05, "loss": 2.8906, "step": 35176 }, { "epoch": 2.76, "learning_rate": 7.241668626989728e-05, "loss": 2.6254, "step": 35177 }, { "epoch": 2.76, "learning_rate": 7.241590214067278e-05, "loss": 2.6887, "step": 35178 }, { "epoch": 2.76, "learning_rate": 7.241511801144829e-05, "loss": 2.9907, "step": 35179 }, { "epoch": 2.76, "learning_rate": 7.241433388222379e-05, "loss": 2.7152, "step": 35180 }, { "epoch": 2.76, "learning_rate": 7.24135497529993e-05, "loss": 2.9007, "step": 35181 }, { "epoch": 2.76, "learning_rate": 7.24127656237748e-05, "loss": 2.8049, "step": 35182 }, { "epoch": 2.76, "learning_rate": 7.24119814945503e-05, "loss": 2.7172, "step": 35183 }, { "epoch": 2.76, "learning_rate": 7.24111973653258e-05, "loss": 2.7963, "step": 35184 }, { "epoch": 2.76, "learning_rate": 7.241041323610131e-05, "loss": 2.7107, "step": 35185 }, { "epoch": 2.76, "learning_rate": 7.240962910687681e-05, "loss": 2.833, "step": 35186 }, { "epoch": 2.76, "learning_rate": 7.240884497765233e-05, "loss": 2.7419, "step": 35187 }, { "epoch": 2.76, "learning_rate": 7.240806084842783e-05, "loss": 2.7419, "step": 35188 }, { "epoch": 2.76, "learning_rate": 7.240727671920334e-05, "loss": 2.9237, "step": 35189 }, { "epoch": 2.76, "learning_rate": 7.240649258997884e-05, "loss": 2.705, "step": 35190 }, { "epoch": 2.76, "learning_rate": 7.240570846075434e-05, "loss": 2.7709, "step": 35191 }, { "epoch": 2.76, "learning_rate": 7.240492433152985e-05, "loss": 2.8433, "step": 35192 }, { "epoch": 2.76, "learning_rate": 7.240414020230535e-05, "loss": 2.7554, "step": 35193 }, { "epoch": 2.76, "learning_rate": 7.240335607308086e-05, "loss": 2.9902, "step": 35194 }, { "epoch": 2.76, "learning_rate": 7.240257194385635e-05, "loss": 2.9473, "step": 35195 }, { "epoch": 2.76, "learning_rate": 7.240178781463185e-05, "loss": 2.5999, "step": 35196 }, { "epoch": 2.76, "learning_rate": 7.240100368540735e-05, "loss": 2.6927, "step": 35197 }, { "epoch": 2.76, "learning_rate": 7.240021955618286e-05, "loss": 2.9286, "step": 35198 }, { "epoch": 2.76, "learning_rate": 7.239943542695836e-05, "loss": 2.9063, "step": 35199 }, { "epoch": 2.76, "learning_rate": 7.239865129773386e-05, "loss": 2.869, "step": 35200 }, { "epoch": 2.76, "learning_rate": 7.239786716850937e-05, "loss": 2.8685, "step": 35201 }, { "epoch": 2.76, "learning_rate": 7.239708303928487e-05, "loss": 2.5245, "step": 35202 }, { "epoch": 2.76, "learning_rate": 7.239629891006038e-05, "loss": 2.6798, "step": 35203 }, { "epoch": 2.76, "learning_rate": 7.239551478083588e-05, "loss": 2.7603, "step": 35204 }, { "epoch": 2.76, "learning_rate": 7.239473065161138e-05, "loss": 2.8605, "step": 35205 }, { "epoch": 2.76, "learning_rate": 7.239394652238689e-05, "loss": 2.9661, "step": 35206 }, { "epoch": 2.76, "learning_rate": 7.239316239316239e-05, "loss": 2.9432, "step": 35207 }, { "epoch": 2.76, "learning_rate": 7.23923782639379e-05, "loss": 2.7283, "step": 35208 }, { "epoch": 2.76, "learning_rate": 7.23915941347134e-05, "loss": 2.8757, "step": 35209 }, { "epoch": 2.76, "learning_rate": 7.23908100054889e-05, "loss": 3.0265, "step": 35210 }, { "epoch": 2.76, "learning_rate": 7.23900258762644e-05, "loss": 2.7311, "step": 35211 }, { "epoch": 2.76, "learning_rate": 7.238924174703992e-05, "loss": 2.9429, "step": 35212 }, { "epoch": 2.76, "learning_rate": 7.238845761781543e-05, "loss": 2.6922, "step": 35213 }, { "epoch": 2.76, "learning_rate": 7.238767348859093e-05, "loss": 3.1592, "step": 35214 }, { "epoch": 2.76, "learning_rate": 7.238688935936643e-05, "loss": 2.7013, "step": 35215 }, { "epoch": 2.76, "learning_rate": 7.238610523014194e-05, "loss": 2.5903, "step": 35216 }, { "epoch": 2.76, "learning_rate": 7.238532110091744e-05, "loss": 2.6479, "step": 35217 }, { "epoch": 2.76, "learning_rate": 7.238453697169294e-05, "loss": 2.6841, "step": 35218 }, { "epoch": 2.76, "learning_rate": 7.238375284246845e-05, "loss": 2.9113, "step": 35219 }, { "epoch": 2.76, "learning_rate": 7.238296871324395e-05, "loss": 2.8056, "step": 35220 }, { "epoch": 2.76, "learning_rate": 7.238218458401946e-05, "loss": 2.6737, "step": 35221 }, { "epoch": 2.76, "learning_rate": 7.238140045479496e-05, "loss": 2.6784, "step": 35222 }, { "epoch": 2.76, "learning_rate": 7.238061632557046e-05, "loss": 2.616, "step": 35223 }, { "epoch": 2.76, "learning_rate": 7.237983219634597e-05, "loss": 2.9784, "step": 35224 }, { "epoch": 2.76, "learning_rate": 7.237904806712147e-05, "loss": 3.034, "step": 35225 }, { "epoch": 2.76, "learning_rate": 7.237826393789697e-05, "loss": 2.7111, "step": 35226 }, { "epoch": 2.76, "learning_rate": 7.237747980867248e-05, "loss": 2.8702, "step": 35227 }, { "epoch": 2.76, "learning_rate": 7.237669567944798e-05, "loss": 2.8134, "step": 35228 }, { "epoch": 2.76, "learning_rate": 7.237591155022348e-05, "loss": 2.7343, "step": 35229 }, { "epoch": 2.76, "learning_rate": 7.237512742099899e-05, "loss": 2.6885, "step": 35230 }, { "epoch": 2.76, "learning_rate": 7.237434329177448e-05, "loss": 3.0332, "step": 35231 }, { "epoch": 2.76, "learning_rate": 7.237355916254998e-05, "loss": 2.9191, "step": 35232 }, { "epoch": 2.76, "learning_rate": 7.237277503332549e-05, "loss": 2.8751, "step": 35233 }, { "epoch": 2.76, "learning_rate": 7.237199090410099e-05, "loss": 2.7867, "step": 35234 }, { "epoch": 2.76, "learning_rate": 7.237120677487649e-05, "loss": 2.7606, "step": 35235 }, { "epoch": 2.76, "learning_rate": 7.2370422645652e-05, "loss": 2.7485, "step": 35236 }, { "epoch": 2.76, "learning_rate": 7.23696385164275e-05, "loss": 2.8617, "step": 35237 }, { "epoch": 2.76, "learning_rate": 7.236885438720302e-05, "loss": 2.8079, "step": 35238 }, { "epoch": 2.76, "learning_rate": 7.236807025797852e-05, "loss": 2.8692, "step": 35239 }, { "epoch": 2.76, "learning_rate": 7.236728612875402e-05, "loss": 2.831, "step": 35240 }, { "epoch": 2.76, "learning_rate": 7.236650199952953e-05, "loss": 2.724, "step": 35241 }, { "epoch": 2.76, "learning_rate": 7.236571787030503e-05, "loss": 2.7086, "step": 35242 }, { "epoch": 2.76, "learning_rate": 7.236493374108054e-05, "loss": 2.6735, "step": 35243 }, { "epoch": 2.76, "learning_rate": 7.236414961185604e-05, "loss": 3.1041, "step": 35244 }, { "epoch": 2.76, "learning_rate": 7.236336548263154e-05, "loss": 2.8426, "step": 35245 }, { "epoch": 2.76, "learning_rate": 7.236258135340705e-05, "loss": 2.8801, "step": 35246 }, { "epoch": 2.76, "learning_rate": 7.236179722418255e-05, "loss": 2.8582, "step": 35247 }, { "epoch": 2.76, "learning_rate": 7.236101309495805e-05, "loss": 3.0347, "step": 35248 }, { "epoch": 2.76, "learning_rate": 7.236022896573356e-05, "loss": 2.8211, "step": 35249 }, { "epoch": 2.76, "learning_rate": 7.235944483650906e-05, "loss": 2.7931, "step": 35250 }, { "epoch": 2.76, "learning_rate": 7.235866070728457e-05, "loss": 2.9578, "step": 35251 }, { "epoch": 2.76, "learning_rate": 7.235787657806007e-05, "loss": 2.8592, "step": 35252 }, { "epoch": 2.76, "learning_rate": 7.235709244883557e-05, "loss": 3.1248, "step": 35253 }, { "epoch": 2.76, "learning_rate": 7.235630831961108e-05, "loss": 2.8102, "step": 35254 }, { "epoch": 2.76, "learning_rate": 7.235552419038658e-05, "loss": 2.7363, "step": 35255 }, { "epoch": 2.76, "learning_rate": 7.235474006116208e-05, "loss": 2.7931, "step": 35256 }, { "epoch": 2.76, "learning_rate": 7.235395593193759e-05, "loss": 2.6947, "step": 35257 }, { "epoch": 2.76, "learning_rate": 7.235317180271309e-05, "loss": 2.687, "step": 35258 }, { "epoch": 2.76, "learning_rate": 7.23523876734886e-05, "loss": 2.9213, "step": 35259 }, { "epoch": 2.76, "learning_rate": 7.23516035442641e-05, "loss": 2.5978, "step": 35260 }, { "epoch": 2.76, "learning_rate": 7.23508194150396e-05, "loss": 2.7455, "step": 35261 }, { "epoch": 2.76, "learning_rate": 7.23500352858151e-05, "loss": 2.6429, "step": 35262 }, { "epoch": 2.77, "learning_rate": 7.234925115659061e-05, "loss": 3.1575, "step": 35263 }, { "epoch": 2.77, "learning_rate": 7.234846702736611e-05, "loss": 2.6096, "step": 35264 }, { "epoch": 2.77, "learning_rate": 7.234768289814162e-05, "loss": 2.7772, "step": 35265 }, { "epoch": 2.77, "learning_rate": 7.234689876891712e-05, "loss": 2.6612, "step": 35266 }, { "epoch": 2.77, "learning_rate": 7.234611463969262e-05, "loss": 2.8558, "step": 35267 }, { "epoch": 2.77, "learning_rate": 7.234533051046813e-05, "loss": 2.9038, "step": 35268 }, { "epoch": 2.77, "learning_rate": 7.234454638124363e-05, "loss": 2.8733, "step": 35269 }, { "epoch": 2.77, "learning_rate": 7.234376225201914e-05, "loss": 2.9322, "step": 35270 }, { "epoch": 2.77, "learning_rate": 7.234297812279464e-05, "loss": 2.8194, "step": 35271 }, { "epoch": 2.77, "learning_rate": 7.234219399357014e-05, "loss": 2.717, "step": 35272 }, { "epoch": 2.77, "learning_rate": 7.234140986434565e-05, "loss": 2.9245, "step": 35273 }, { "epoch": 2.77, "learning_rate": 7.234062573512115e-05, "loss": 2.711, "step": 35274 }, { "epoch": 2.77, "learning_rate": 7.233984160589665e-05, "loss": 2.6711, "step": 35275 }, { "epoch": 2.77, "learning_rate": 7.233905747667216e-05, "loss": 2.7681, "step": 35276 }, { "epoch": 2.77, "learning_rate": 7.233827334744766e-05, "loss": 3.0571, "step": 35277 }, { "epoch": 2.77, "learning_rate": 7.233748921822316e-05, "loss": 2.8543, "step": 35278 }, { "epoch": 2.77, "learning_rate": 7.233670508899867e-05, "loss": 2.7922, "step": 35279 }, { "epoch": 2.77, "learning_rate": 7.233592095977417e-05, "loss": 2.9422, "step": 35280 }, { "epoch": 2.77, "learning_rate": 7.233513683054968e-05, "loss": 2.6794, "step": 35281 }, { "epoch": 2.77, "learning_rate": 7.233435270132518e-05, "loss": 2.7318, "step": 35282 }, { "epoch": 2.77, "learning_rate": 7.233356857210068e-05, "loss": 2.9417, "step": 35283 }, { "epoch": 2.77, "learning_rate": 7.233278444287619e-05, "loss": 3.0301, "step": 35284 }, { "epoch": 2.77, "learning_rate": 7.233200031365169e-05, "loss": 2.9327, "step": 35285 }, { "epoch": 2.77, "learning_rate": 7.23312161844272e-05, "loss": 2.7291, "step": 35286 }, { "epoch": 2.77, "learning_rate": 7.23304320552027e-05, "loss": 2.6434, "step": 35287 }, { "epoch": 2.77, "learning_rate": 7.23296479259782e-05, "loss": 2.7441, "step": 35288 }, { "epoch": 2.77, "learning_rate": 7.23288637967537e-05, "loss": 2.8158, "step": 35289 }, { "epoch": 2.77, "learning_rate": 7.232807966752921e-05, "loss": 2.7601, "step": 35290 }, { "epoch": 2.77, "learning_rate": 7.232729553830471e-05, "loss": 2.6867, "step": 35291 }, { "epoch": 2.77, "learning_rate": 7.232651140908022e-05, "loss": 2.7435, "step": 35292 }, { "epoch": 2.77, "learning_rate": 7.232572727985572e-05, "loss": 2.7482, "step": 35293 }, { "epoch": 2.77, "learning_rate": 7.232494315063122e-05, "loss": 2.5889, "step": 35294 }, { "epoch": 2.77, "learning_rate": 7.232415902140674e-05, "loss": 2.8613, "step": 35295 }, { "epoch": 2.77, "learning_rate": 7.232337489218224e-05, "loss": 2.9894, "step": 35296 }, { "epoch": 2.77, "learning_rate": 7.232259076295775e-05, "loss": 2.9709, "step": 35297 }, { "epoch": 2.77, "learning_rate": 7.232180663373325e-05, "loss": 2.7813, "step": 35298 }, { "epoch": 2.77, "learning_rate": 7.232102250450876e-05, "loss": 2.7508, "step": 35299 }, { "epoch": 2.77, "learning_rate": 7.232023837528426e-05, "loss": 2.7168, "step": 35300 }, { "epoch": 2.77, "learning_rate": 7.231945424605976e-05, "loss": 2.6447, "step": 35301 }, { "epoch": 2.77, "learning_rate": 7.231867011683527e-05, "loss": 3.0827, "step": 35302 }, { "epoch": 2.77, "learning_rate": 7.231788598761076e-05, "loss": 2.928, "step": 35303 }, { "epoch": 2.77, "learning_rate": 7.231710185838626e-05, "loss": 3.1733, "step": 35304 }, { "epoch": 2.77, "learning_rate": 7.231631772916176e-05, "loss": 2.7656, "step": 35305 }, { "epoch": 2.77, "learning_rate": 7.231553359993727e-05, "loss": 2.9458, "step": 35306 }, { "epoch": 2.77, "learning_rate": 7.231474947071277e-05, "loss": 2.8904, "step": 35307 }, { "epoch": 2.77, "learning_rate": 7.231396534148828e-05, "loss": 2.859, "step": 35308 }, { "epoch": 2.77, "learning_rate": 7.231318121226378e-05, "loss": 2.6032, "step": 35309 }, { "epoch": 2.77, "learning_rate": 7.231239708303928e-05, "loss": 2.8258, "step": 35310 }, { "epoch": 2.77, "learning_rate": 7.231161295381479e-05, "loss": 2.8806, "step": 35311 }, { "epoch": 2.77, "learning_rate": 7.231082882459029e-05, "loss": 2.9279, "step": 35312 }, { "epoch": 2.77, "learning_rate": 7.23100446953658e-05, "loss": 2.9932, "step": 35313 }, { "epoch": 2.77, "learning_rate": 7.23092605661413e-05, "loss": 2.7103, "step": 35314 }, { "epoch": 2.77, "learning_rate": 7.23084764369168e-05, "loss": 3.0708, "step": 35315 }, { "epoch": 2.77, "learning_rate": 7.23076923076923e-05, "loss": 3.1037, "step": 35316 }, { "epoch": 2.77, "learning_rate": 7.230690817846781e-05, "loss": 2.8122, "step": 35317 }, { "epoch": 2.77, "learning_rate": 7.230612404924331e-05, "loss": 2.9784, "step": 35318 }, { "epoch": 2.77, "learning_rate": 7.230533992001882e-05, "loss": 2.8643, "step": 35319 }, { "epoch": 2.77, "learning_rate": 7.230455579079433e-05, "loss": 2.8617, "step": 35320 }, { "epoch": 2.77, "learning_rate": 7.230377166156984e-05, "loss": 2.65, "step": 35321 }, { "epoch": 2.77, "learning_rate": 7.230298753234534e-05, "loss": 2.7529, "step": 35322 }, { "epoch": 2.77, "learning_rate": 7.230220340312084e-05, "loss": 2.7263, "step": 35323 }, { "epoch": 2.77, "learning_rate": 7.230141927389635e-05, "loss": 2.7963, "step": 35324 }, { "epoch": 2.77, "learning_rate": 7.230063514467185e-05, "loss": 3.0999, "step": 35325 }, { "epoch": 2.77, "learning_rate": 7.229985101544735e-05, "loss": 2.7304, "step": 35326 }, { "epoch": 2.77, "learning_rate": 7.229906688622286e-05, "loss": 2.7812, "step": 35327 }, { "epoch": 2.77, "learning_rate": 7.229828275699836e-05, "loss": 2.9802, "step": 35328 }, { "epoch": 2.77, "learning_rate": 7.229749862777387e-05, "loss": 2.7612, "step": 35329 }, { "epoch": 2.77, "learning_rate": 7.229671449854937e-05, "loss": 2.5967, "step": 35330 }, { "epoch": 2.77, "learning_rate": 7.229593036932487e-05, "loss": 2.7559, "step": 35331 }, { "epoch": 2.77, "learning_rate": 7.229514624010038e-05, "loss": 2.6655, "step": 35332 }, { "epoch": 2.77, "learning_rate": 7.229436211087588e-05, "loss": 2.7779, "step": 35333 }, { "epoch": 2.77, "learning_rate": 7.229357798165138e-05, "loss": 2.7754, "step": 35334 }, { "epoch": 2.77, "learning_rate": 7.229279385242689e-05, "loss": 2.9972, "step": 35335 }, { "epoch": 2.77, "learning_rate": 7.229200972320239e-05, "loss": 2.7405, "step": 35336 }, { "epoch": 2.77, "learning_rate": 7.22912255939779e-05, "loss": 2.6954, "step": 35337 }, { "epoch": 2.77, "learning_rate": 7.22904414647534e-05, "loss": 2.8006, "step": 35338 }, { "epoch": 2.77, "learning_rate": 7.228965733552889e-05, "loss": 2.8934, "step": 35339 }, { "epoch": 2.77, "learning_rate": 7.228887320630439e-05, "loss": 2.7738, "step": 35340 }, { "epoch": 2.77, "learning_rate": 7.22880890770799e-05, "loss": 2.8753, "step": 35341 }, { "epoch": 2.77, "learning_rate": 7.22873049478554e-05, "loss": 2.7188, "step": 35342 }, { "epoch": 2.77, "learning_rate": 7.22865208186309e-05, "loss": 3.0283, "step": 35343 }, { "epoch": 2.77, "learning_rate": 7.228573668940641e-05, "loss": 2.8972, "step": 35344 }, { "epoch": 2.77, "learning_rate": 7.228495256018191e-05, "loss": 2.9275, "step": 35345 }, { "epoch": 2.77, "learning_rate": 7.228416843095743e-05, "loss": 2.8301, "step": 35346 }, { "epoch": 2.77, "learning_rate": 7.228338430173293e-05, "loss": 2.9124, "step": 35347 }, { "epoch": 2.77, "learning_rate": 7.228260017250844e-05, "loss": 3.0426, "step": 35348 }, { "epoch": 2.77, "learning_rate": 7.228181604328394e-05, "loss": 2.8513, "step": 35349 }, { "epoch": 2.77, "learning_rate": 7.228103191405944e-05, "loss": 2.8809, "step": 35350 }, { "epoch": 2.77, "learning_rate": 7.228024778483495e-05, "loss": 2.9044, "step": 35351 }, { "epoch": 2.77, "learning_rate": 7.227946365561045e-05, "loss": 2.8093, "step": 35352 }, { "epoch": 2.77, "learning_rate": 7.227867952638595e-05, "loss": 2.9066, "step": 35353 }, { "epoch": 2.77, "learning_rate": 7.227789539716146e-05, "loss": 2.6854, "step": 35354 }, { "epoch": 2.77, "learning_rate": 7.227711126793696e-05, "loss": 2.6802, "step": 35355 }, { "epoch": 2.77, "learning_rate": 7.227632713871247e-05, "loss": 2.8633, "step": 35356 }, { "epoch": 2.77, "learning_rate": 7.227554300948797e-05, "loss": 2.8472, "step": 35357 }, { "epoch": 2.77, "learning_rate": 7.227475888026347e-05, "loss": 2.8259, "step": 35358 }, { "epoch": 2.77, "learning_rate": 7.227397475103898e-05, "loss": 2.8131, "step": 35359 }, { "epoch": 2.77, "learning_rate": 7.227319062181448e-05, "loss": 2.6222, "step": 35360 }, { "epoch": 2.77, "learning_rate": 7.227240649258998e-05, "loss": 2.8182, "step": 35361 }, { "epoch": 2.77, "learning_rate": 7.227162236336549e-05, "loss": 3.0694, "step": 35362 }, { "epoch": 2.77, "learning_rate": 7.227083823414099e-05, "loss": 2.613, "step": 35363 }, { "epoch": 2.77, "learning_rate": 7.22700541049165e-05, "loss": 2.9674, "step": 35364 }, { "epoch": 2.77, "learning_rate": 7.2269269975692e-05, "loss": 2.789, "step": 35365 }, { "epoch": 2.77, "learning_rate": 7.22684858464675e-05, "loss": 2.7183, "step": 35366 }, { "epoch": 2.77, "learning_rate": 7.2267701717243e-05, "loss": 2.6858, "step": 35367 }, { "epoch": 2.77, "learning_rate": 7.226691758801851e-05, "loss": 2.9769, "step": 35368 }, { "epoch": 2.77, "learning_rate": 7.226613345879401e-05, "loss": 2.8103, "step": 35369 }, { "epoch": 2.77, "learning_rate": 7.226534932956952e-05, "loss": 2.7403, "step": 35370 }, { "epoch": 2.77, "learning_rate": 7.226456520034502e-05, "loss": 2.7546, "step": 35371 }, { "epoch": 2.77, "learning_rate": 7.226378107112052e-05, "loss": 2.9178, "step": 35372 }, { "epoch": 2.77, "learning_rate": 7.226299694189603e-05, "loss": 3.0318, "step": 35373 }, { "epoch": 2.77, "learning_rate": 7.226221281267153e-05, "loss": 2.9629, "step": 35374 }, { "epoch": 2.77, "learning_rate": 7.226142868344704e-05, "loss": 2.7115, "step": 35375 }, { "epoch": 2.77, "learning_rate": 7.226064455422254e-05, "loss": 2.7969, "step": 35376 }, { "epoch": 2.77, "learning_rate": 7.225986042499804e-05, "loss": 2.734, "step": 35377 }, { "epoch": 2.77, "learning_rate": 7.225907629577355e-05, "loss": 2.7005, "step": 35378 }, { "epoch": 2.77, "learning_rate": 7.225829216654905e-05, "loss": 2.7481, "step": 35379 }, { "epoch": 2.77, "learning_rate": 7.225750803732455e-05, "loss": 2.9496, "step": 35380 }, { "epoch": 2.77, "learning_rate": 7.225672390810006e-05, "loss": 2.7575, "step": 35381 }, { "epoch": 2.77, "learning_rate": 7.225593977887556e-05, "loss": 2.7315, "step": 35382 }, { "epoch": 2.77, "learning_rate": 7.225515564965106e-05, "loss": 2.8796, "step": 35383 }, { "epoch": 2.77, "learning_rate": 7.225437152042657e-05, "loss": 2.5079, "step": 35384 }, { "epoch": 2.77, "learning_rate": 7.225358739120207e-05, "loss": 2.9052, "step": 35385 }, { "epoch": 2.77, "learning_rate": 7.225280326197758e-05, "loss": 2.7524, "step": 35386 }, { "epoch": 2.77, "learning_rate": 7.225201913275308e-05, "loss": 2.9779, "step": 35387 }, { "epoch": 2.77, "learning_rate": 7.225123500352858e-05, "loss": 3.1089, "step": 35388 }, { "epoch": 2.77, "learning_rate": 7.225045087430409e-05, "loss": 2.7664, "step": 35389 }, { "epoch": 2.78, "learning_rate": 7.224966674507959e-05, "loss": 2.9976, "step": 35390 }, { "epoch": 2.78, "learning_rate": 7.22488826158551e-05, "loss": 2.7478, "step": 35391 }, { "epoch": 2.78, "learning_rate": 7.22480984866306e-05, "loss": 2.9737, "step": 35392 }, { "epoch": 2.78, "learning_rate": 7.22473143574061e-05, "loss": 2.6341, "step": 35393 }, { "epoch": 2.78, "learning_rate": 7.22465302281816e-05, "loss": 2.7106, "step": 35394 }, { "epoch": 2.78, "learning_rate": 7.224574609895711e-05, "loss": 2.7808, "step": 35395 }, { "epoch": 2.78, "learning_rate": 7.224496196973261e-05, "loss": 2.6971, "step": 35396 }, { "epoch": 2.78, "learning_rate": 7.224417784050812e-05, "loss": 2.7275, "step": 35397 }, { "epoch": 2.78, "learning_rate": 7.224339371128362e-05, "loss": 2.8012, "step": 35398 }, { "epoch": 2.78, "learning_rate": 7.224260958205912e-05, "loss": 2.6729, "step": 35399 }, { "epoch": 2.78, "learning_rate": 7.224182545283463e-05, "loss": 2.8078, "step": 35400 }, { "epoch": 2.78, "learning_rate": 7.224104132361013e-05, "loss": 2.7332, "step": 35401 }, { "epoch": 2.78, "learning_rate": 7.224025719438563e-05, "loss": 2.7949, "step": 35402 }, { "epoch": 2.78, "learning_rate": 7.223947306516115e-05, "loss": 2.669, "step": 35403 }, { "epoch": 2.78, "learning_rate": 7.223868893593666e-05, "loss": 2.7368, "step": 35404 }, { "epoch": 2.78, "learning_rate": 7.223790480671216e-05, "loss": 2.9044, "step": 35405 }, { "epoch": 2.78, "learning_rate": 7.223712067748766e-05, "loss": 3.065, "step": 35406 }, { "epoch": 2.78, "learning_rate": 7.223633654826317e-05, "loss": 2.9265, "step": 35407 }, { "epoch": 2.78, "learning_rate": 7.223555241903867e-05, "loss": 2.8345, "step": 35408 }, { "epoch": 2.78, "learning_rate": 7.223476828981417e-05, "loss": 2.723, "step": 35409 }, { "epoch": 2.78, "learning_rate": 7.223398416058968e-05, "loss": 2.9119, "step": 35410 }, { "epoch": 2.78, "learning_rate": 7.223320003136517e-05, "loss": 2.8, "step": 35411 }, { "epoch": 2.78, "learning_rate": 7.223241590214067e-05, "loss": 2.7312, "step": 35412 }, { "epoch": 2.78, "learning_rate": 7.223163177291617e-05, "loss": 2.6828, "step": 35413 }, { "epoch": 2.78, "learning_rate": 7.223084764369168e-05, "loss": 2.9733, "step": 35414 }, { "epoch": 2.78, "learning_rate": 7.223006351446718e-05, "loss": 2.8708, "step": 35415 }, { "epoch": 2.78, "learning_rate": 7.222927938524269e-05, "loss": 2.8817, "step": 35416 }, { "epoch": 2.78, "learning_rate": 7.222849525601819e-05, "loss": 2.5624, "step": 35417 }, { "epoch": 2.78, "learning_rate": 7.222771112679369e-05, "loss": 2.6068, "step": 35418 }, { "epoch": 2.78, "learning_rate": 7.22269269975692e-05, "loss": 2.8486, "step": 35419 }, { "epoch": 2.78, "learning_rate": 7.22261428683447e-05, "loss": 2.564, "step": 35420 }, { "epoch": 2.78, "learning_rate": 7.22253587391202e-05, "loss": 2.7989, "step": 35421 }, { "epoch": 2.78, "learning_rate": 7.222457460989571e-05, "loss": 2.5986, "step": 35422 }, { "epoch": 2.78, "learning_rate": 7.222379048067121e-05, "loss": 2.5816, "step": 35423 }, { "epoch": 2.78, "learning_rate": 7.222300635144672e-05, "loss": 2.9073, "step": 35424 }, { "epoch": 2.78, "learning_rate": 7.222222222222222e-05, "loss": 2.8228, "step": 35425 }, { "epoch": 2.78, "learning_rate": 7.222143809299772e-05, "loss": 2.6584, "step": 35426 }, { "epoch": 2.78, "learning_rate": 7.222065396377323e-05, "loss": 2.8192, "step": 35427 }, { "epoch": 2.78, "learning_rate": 7.221986983454874e-05, "loss": 2.6489, "step": 35428 }, { "epoch": 2.78, "learning_rate": 7.221908570532425e-05, "loss": 2.921, "step": 35429 }, { "epoch": 2.78, "learning_rate": 7.221830157609975e-05, "loss": 2.7691, "step": 35430 }, { "epoch": 2.78, "learning_rate": 7.221751744687525e-05, "loss": 2.7129, "step": 35431 }, { "epoch": 2.78, "learning_rate": 7.221673331765076e-05, "loss": 2.6359, "step": 35432 }, { "epoch": 2.78, "learning_rate": 7.221594918842626e-05, "loss": 2.8194, "step": 35433 }, { "epoch": 2.78, "learning_rate": 7.221516505920177e-05, "loss": 2.7866, "step": 35434 }, { "epoch": 2.78, "learning_rate": 7.221438092997727e-05, "loss": 2.6924, "step": 35435 }, { "epoch": 2.78, "learning_rate": 7.221359680075277e-05, "loss": 2.7983, "step": 35436 }, { "epoch": 2.78, "learning_rate": 7.221281267152828e-05, "loss": 2.8443, "step": 35437 }, { "epoch": 2.78, "learning_rate": 7.221202854230378e-05, "loss": 2.8616, "step": 35438 }, { "epoch": 2.78, "learning_rate": 7.221124441307928e-05, "loss": 2.9475, "step": 35439 }, { "epoch": 2.78, "learning_rate": 7.221046028385479e-05, "loss": 2.7451, "step": 35440 }, { "epoch": 2.78, "learning_rate": 7.220967615463029e-05, "loss": 2.8729, "step": 35441 }, { "epoch": 2.78, "learning_rate": 7.22088920254058e-05, "loss": 2.65, "step": 35442 }, { "epoch": 2.78, "learning_rate": 7.22081078961813e-05, "loss": 2.6642, "step": 35443 }, { "epoch": 2.78, "learning_rate": 7.22073237669568e-05, "loss": 2.8632, "step": 35444 }, { "epoch": 2.78, "learning_rate": 7.22065396377323e-05, "loss": 2.8645, "step": 35445 }, { "epoch": 2.78, "learning_rate": 7.220575550850781e-05, "loss": 2.6017, "step": 35446 }, { "epoch": 2.78, "learning_rate": 7.220497137928331e-05, "loss": 2.7645, "step": 35447 }, { "epoch": 2.78, "learning_rate": 7.22041872500588e-05, "loss": 2.8038, "step": 35448 }, { "epoch": 2.78, "learning_rate": 7.220340312083431e-05, "loss": 2.8048, "step": 35449 }, { "epoch": 2.78, "learning_rate": 7.220261899160981e-05, "loss": 2.7039, "step": 35450 }, { "epoch": 2.78, "learning_rate": 7.220183486238531e-05, "loss": 2.9428, "step": 35451 }, { "epoch": 2.78, "learning_rate": 7.220105073316082e-05, "loss": 2.7724, "step": 35452 }, { "epoch": 2.78, "learning_rate": 7.220026660393632e-05, "loss": 2.8741, "step": 35453 }, { "epoch": 2.78, "learning_rate": 7.219948247471184e-05, "loss": 2.8291, "step": 35454 }, { "epoch": 2.78, "learning_rate": 7.219869834548734e-05, "loss": 2.6015, "step": 35455 }, { "epoch": 2.78, "learning_rate": 7.219791421626285e-05, "loss": 2.8246, "step": 35456 }, { "epoch": 2.78, "learning_rate": 7.219713008703835e-05, "loss": 2.7708, "step": 35457 }, { "epoch": 2.78, "learning_rate": 7.219634595781385e-05, "loss": 2.681, "step": 35458 }, { "epoch": 2.78, "learning_rate": 7.219556182858936e-05, "loss": 2.8115, "step": 35459 }, { "epoch": 2.78, "learning_rate": 7.219477769936486e-05, "loss": 2.6732, "step": 35460 }, { "epoch": 2.78, "learning_rate": 7.219399357014036e-05, "loss": 2.861, "step": 35461 }, { "epoch": 2.78, "learning_rate": 7.219320944091587e-05, "loss": 2.7951, "step": 35462 }, { "epoch": 2.78, "learning_rate": 7.219242531169137e-05, "loss": 2.8247, "step": 35463 }, { "epoch": 2.78, "learning_rate": 7.219164118246688e-05, "loss": 2.7135, "step": 35464 }, { "epoch": 2.78, "learning_rate": 7.219085705324238e-05, "loss": 2.6784, "step": 35465 }, { "epoch": 2.78, "learning_rate": 7.219007292401788e-05, "loss": 2.6865, "step": 35466 }, { "epoch": 2.78, "learning_rate": 7.218928879479339e-05, "loss": 2.8724, "step": 35467 }, { "epoch": 2.78, "learning_rate": 7.218850466556889e-05, "loss": 3.0653, "step": 35468 }, { "epoch": 2.78, "learning_rate": 7.21877205363444e-05, "loss": 3.1617, "step": 35469 }, { "epoch": 2.78, "learning_rate": 7.21869364071199e-05, "loss": 2.8638, "step": 35470 }, { "epoch": 2.78, "learning_rate": 7.21861522778954e-05, "loss": 2.9173, "step": 35471 }, { "epoch": 2.78, "learning_rate": 7.21853681486709e-05, "loss": 2.808, "step": 35472 }, { "epoch": 2.78, "learning_rate": 7.218458401944641e-05, "loss": 2.9555, "step": 35473 }, { "epoch": 2.78, "learning_rate": 7.218379989022191e-05, "loss": 2.7372, "step": 35474 }, { "epoch": 2.78, "learning_rate": 7.218301576099742e-05, "loss": 2.6488, "step": 35475 }, { "epoch": 2.78, "learning_rate": 7.218223163177292e-05, "loss": 2.6797, "step": 35476 }, { "epoch": 2.78, "learning_rate": 7.218144750254842e-05, "loss": 2.6249, "step": 35477 }, { "epoch": 2.78, "learning_rate": 7.218066337332393e-05, "loss": 2.7644, "step": 35478 }, { "epoch": 2.78, "learning_rate": 7.217987924409943e-05, "loss": 2.7652, "step": 35479 }, { "epoch": 2.78, "learning_rate": 7.217909511487493e-05, "loss": 2.828, "step": 35480 }, { "epoch": 2.78, "learning_rate": 7.217831098565044e-05, "loss": 2.7591, "step": 35481 }, { "epoch": 2.78, "learning_rate": 7.217752685642594e-05, "loss": 3.0016, "step": 35482 }, { "epoch": 2.78, "learning_rate": 7.217674272720145e-05, "loss": 2.5905, "step": 35483 }, { "epoch": 2.78, "learning_rate": 7.217595859797695e-05, "loss": 2.8921, "step": 35484 }, { "epoch": 2.78, "learning_rate": 7.217517446875245e-05, "loss": 2.8388, "step": 35485 }, { "epoch": 2.78, "learning_rate": 7.217439033952796e-05, "loss": 2.5654, "step": 35486 }, { "epoch": 2.78, "learning_rate": 7.217360621030346e-05, "loss": 2.7385, "step": 35487 }, { "epoch": 2.78, "learning_rate": 7.217282208107896e-05, "loss": 2.9406, "step": 35488 }, { "epoch": 2.78, "learning_rate": 7.217203795185447e-05, "loss": 3.0985, "step": 35489 }, { "epoch": 2.78, "learning_rate": 7.217125382262997e-05, "loss": 2.7812, "step": 35490 }, { "epoch": 2.78, "learning_rate": 7.217046969340548e-05, "loss": 3.0323, "step": 35491 }, { "epoch": 2.78, "learning_rate": 7.216968556418098e-05, "loss": 2.8992, "step": 35492 }, { "epoch": 2.78, "learning_rate": 7.216890143495648e-05, "loss": 2.811, "step": 35493 }, { "epoch": 2.78, "learning_rate": 7.216811730573199e-05, "loss": 2.7225, "step": 35494 }, { "epoch": 2.78, "learning_rate": 7.216733317650749e-05, "loss": 2.8028, "step": 35495 }, { "epoch": 2.78, "learning_rate": 7.2166549047283e-05, "loss": 2.6939, "step": 35496 }, { "epoch": 2.78, "learning_rate": 7.21657649180585e-05, "loss": 2.6355, "step": 35497 }, { "epoch": 2.78, "learning_rate": 7.2164980788834e-05, "loss": 2.8754, "step": 35498 }, { "epoch": 2.78, "learning_rate": 7.21641966596095e-05, "loss": 2.6908, "step": 35499 }, { "epoch": 2.78, "learning_rate": 7.216341253038501e-05, "loss": 2.9002, "step": 35500 }, { "epoch": 2.78, "learning_rate": 7.216262840116051e-05, "loss": 2.5671, "step": 35501 }, { "epoch": 2.78, "learning_rate": 7.216184427193602e-05, "loss": 2.8758, "step": 35502 }, { "epoch": 2.78, "learning_rate": 7.216106014271152e-05, "loss": 2.6902, "step": 35503 }, { "epoch": 2.78, "learning_rate": 7.216027601348702e-05, "loss": 2.8257, "step": 35504 }, { "epoch": 2.78, "learning_rate": 7.215949188426253e-05, "loss": 2.7608, "step": 35505 }, { "epoch": 2.78, "learning_rate": 7.215870775503803e-05, "loss": 2.6811, "step": 35506 }, { "epoch": 2.78, "learning_rate": 7.215792362581353e-05, "loss": 2.7361, "step": 35507 }, { "epoch": 2.78, "learning_rate": 7.215713949658904e-05, "loss": 2.7419, "step": 35508 }, { "epoch": 2.78, "learning_rate": 7.215635536736454e-05, "loss": 3.1445, "step": 35509 }, { "epoch": 2.78, "learning_rate": 7.215557123814005e-05, "loss": 2.8731, "step": 35510 }, { "epoch": 2.78, "learning_rate": 7.215478710891556e-05, "loss": 2.9036, "step": 35511 }, { "epoch": 2.78, "learning_rate": 7.215400297969107e-05, "loss": 2.826, "step": 35512 }, { "epoch": 2.78, "learning_rate": 7.215321885046657e-05, "loss": 2.5765, "step": 35513 }, { "epoch": 2.78, "learning_rate": 7.215243472124207e-05, "loss": 2.7977, "step": 35514 }, { "epoch": 2.78, "learning_rate": 7.215165059201758e-05, "loss": 2.8545, "step": 35515 }, { "epoch": 2.78, "learning_rate": 7.215086646279308e-05, "loss": 2.8248, "step": 35516 }, { "epoch": 2.78, "learning_rate": 7.215008233356858e-05, "loss": 2.8989, "step": 35517 }, { "epoch": 2.79, "learning_rate": 7.214929820434409e-05, "loss": 2.6047, "step": 35518 }, { "epoch": 2.79, "learning_rate": 7.214851407511959e-05, "loss": 2.82, "step": 35519 }, { "epoch": 2.79, "learning_rate": 7.214772994589508e-05, "loss": 2.9551, "step": 35520 }, { "epoch": 2.79, "learning_rate": 7.214694581667059e-05, "loss": 2.7489, "step": 35521 }, { "epoch": 2.79, "learning_rate": 7.214616168744609e-05, "loss": 2.6225, "step": 35522 }, { "epoch": 2.79, "learning_rate": 7.214537755822159e-05, "loss": 2.7205, "step": 35523 }, { "epoch": 2.79, "learning_rate": 7.21445934289971e-05, "loss": 2.7993, "step": 35524 }, { "epoch": 2.79, "learning_rate": 7.21438092997726e-05, "loss": 2.963, "step": 35525 }, { "epoch": 2.79, "learning_rate": 7.21430251705481e-05, "loss": 2.6666, "step": 35526 }, { "epoch": 2.79, "learning_rate": 7.214224104132361e-05, "loss": 2.7204, "step": 35527 }, { "epoch": 2.79, "learning_rate": 7.214145691209911e-05, "loss": 2.7945, "step": 35528 }, { "epoch": 2.79, "learning_rate": 7.214067278287461e-05, "loss": 2.7703, "step": 35529 }, { "epoch": 2.79, "learning_rate": 7.213988865365012e-05, "loss": 2.7733, "step": 35530 }, { "epoch": 2.79, "learning_rate": 7.213910452442562e-05, "loss": 2.9645, "step": 35531 }, { "epoch": 2.79, "learning_rate": 7.213832039520113e-05, "loss": 2.9513, "step": 35532 }, { "epoch": 2.79, "learning_rate": 7.213753626597663e-05, "loss": 3.0956, "step": 35533 }, { "epoch": 2.79, "learning_rate": 7.213675213675213e-05, "loss": 2.6098, "step": 35534 }, { "epoch": 2.79, "learning_rate": 7.213596800752764e-05, "loss": 2.7532, "step": 35535 }, { "epoch": 2.79, "learning_rate": 7.213518387830315e-05, "loss": 2.7148, "step": 35536 }, { "epoch": 2.79, "learning_rate": 7.213439974907866e-05, "loss": 2.687, "step": 35537 }, { "epoch": 2.79, "learning_rate": 7.213361561985416e-05, "loss": 2.9713, "step": 35538 }, { "epoch": 2.79, "learning_rate": 7.213283149062967e-05, "loss": 2.9617, "step": 35539 }, { "epoch": 2.79, "learning_rate": 7.213204736140517e-05, "loss": 2.6489, "step": 35540 }, { "epoch": 2.79, "learning_rate": 7.213126323218067e-05, "loss": 2.994, "step": 35541 }, { "epoch": 2.79, "learning_rate": 7.213047910295618e-05, "loss": 2.8861, "step": 35542 }, { "epoch": 2.79, "learning_rate": 7.212969497373168e-05, "loss": 2.7968, "step": 35543 }, { "epoch": 2.79, "learning_rate": 7.212891084450718e-05, "loss": 2.932, "step": 35544 }, { "epoch": 2.79, "learning_rate": 7.212812671528269e-05, "loss": 2.5714, "step": 35545 }, { "epoch": 2.79, "learning_rate": 7.212734258605819e-05, "loss": 2.6178, "step": 35546 }, { "epoch": 2.79, "learning_rate": 7.21265584568337e-05, "loss": 2.8972, "step": 35547 }, { "epoch": 2.79, "learning_rate": 7.21257743276092e-05, "loss": 3.0092, "step": 35548 }, { "epoch": 2.79, "learning_rate": 7.21249901983847e-05, "loss": 2.9009, "step": 35549 }, { "epoch": 2.79, "learning_rate": 7.21242060691602e-05, "loss": 2.9072, "step": 35550 }, { "epoch": 2.79, "learning_rate": 7.212342193993571e-05, "loss": 2.818, "step": 35551 }, { "epoch": 2.79, "learning_rate": 7.212263781071121e-05, "loss": 2.6857, "step": 35552 }, { "epoch": 2.79, "learning_rate": 7.212185368148672e-05, "loss": 2.6481, "step": 35553 }, { "epoch": 2.79, "learning_rate": 7.212106955226222e-05, "loss": 3.1722, "step": 35554 }, { "epoch": 2.79, "learning_rate": 7.212028542303772e-05, "loss": 2.7799, "step": 35555 }, { "epoch": 2.79, "learning_rate": 7.211950129381321e-05, "loss": 2.8273, "step": 35556 }, { "epoch": 2.79, "learning_rate": 7.211871716458872e-05, "loss": 2.7045, "step": 35557 }, { "epoch": 2.79, "learning_rate": 7.211793303536422e-05, "loss": 2.9624, "step": 35558 }, { "epoch": 2.79, "learning_rate": 7.211714890613973e-05, "loss": 2.8939, "step": 35559 }, { "epoch": 2.79, "learning_rate": 7.211636477691523e-05, "loss": 2.7427, "step": 35560 }, { "epoch": 2.79, "learning_rate": 7.211558064769073e-05, "loss": 2.8416, "step": 35561 }, { "epoch": 2.79, "learning_rate": 7.211479651846625e-05, "loss": 3.0434, "step": 35562 }, { "epoch": 2.79, "learning_rate": 7.211401238924175e-05, "loss": 2.6442, "step": 35563 }, { "epoch": 2.79, "learning_rate": 7.211322826001726e-05, "loss": 2.9161, "step": 35564 }, { "epoch": 2.79, "learning_rate": 7.211244413079276e-05, "loss": 2.8077, "step": 35565 }, { "epoch": 2.79, "learning_rate": 7.211166000156826e-05, "loss": 2.5721, "step": 35566 }, { "epoch": 2.79, "learning_rate": 7.211087587234377e-05, "loss": 2.7989, "step": 35567 }, { "epoch": 2.79, "learning_rate": 7.211009174311927e-05, "loss": 2.5908, "step": 35568 }, { "epoch": 2.79, "learning_rate": 7.210930761389478e-05, "loss": 2.7511, "step": 35569 }, { "epoch": 2.79, "learning_rate": 7.210852348467028e-05, "loss": 2.6003, "step": 35570 }, { "epoch": 2.79, "learning_rate": 7.210773935544578e-05, "loss": 2.8532, "step": 35571 }, { "epoch": 2.79, "learning_rate": 7.210695522622129e-05, "loss": 2.7611, "step": 35572 }, { "epoch": 2.79, "learning_rate": 7.210617109699679e-05, "loss": 2.665, "step": 35573 }, { "epoch": 2.79, "learning_rate": 7.21053869677723e-05, "loss": 3.0402, "step": 35574 }, { "epoch": 2.79, "learning_rate": 7.21046028385478e-05, "loss": 2.8722, "step": 35575 }, { "epoch": 2.79, "learning_rate": 7.21038187093233e-05, "loss": 2.9151, "step": 35576 }, { "epoch": 2.79, "learning_rate": 7.21030345800988e-05, "loss": 2.8211, "step": 35577 }, { "epoch": 2.79, "learning_rate": 7.210225045087431e-05, "loss": 2.646, "step": 35578 }, { "epoch": 2.79, "learning_rate": 7.210146632164981e-05, "loss": 2.8864, "step": 35579 }, { "epoch": 2.79, "learning_rate": 7.210068219242532e-05, "loss": 2.8553, "step": 35580 }, { "epoch": 2.79, "learning_rate": 7.209989806320082e-05, "loss": 2.9868, "step": 35581 }, { "epoch": 2.79, "learning_rate": 7.209911393397632e-05, "loss": 3.0128, "step": 35582 }, { "epoch": 2.79, "learning_rate": 7.209832980475183e-05, "loss": 2.7373, "step": 35583 }, { "epoch": 2.79, "learning_rate": 7.209754567552733e-05, "loss": 2.6706, "step": 35584 }, { "epoch": 2.79, "learning_rate": 7.209676154630283e-05, "loss": 2.9903, "step": 35585 }, { "epoch": 2.79, "learning_rate": 7.209597741707834e-05, "loss": 2.8203, "step": 35586 }, { "epoch": 2.79, "learning_rate": 7.209519328785384e-05, "loss": 2.7931, "step": 35587 }, { "epoch": 2.79, "learning_rate": 7.209440915862935e-05, "loss": 2.6692, "step": 35588 }, { "epoch": 2.79, "learning_rate": 7.209362502940485e-05, "loss": 2.7129, "step": 35589 }, { "epoch": 2.79, "learning_rate": 7.209284090018035e-05, "loss": 2.7843, "step": 35590 }, { "epoch": 2.79, "learning_rate": 7.209205677095586e-05, "loss": 2.7634, "step": 35591 }, { "epoch": 2.79, "learning_rate": 7.209127264173136e-05, "loss": 2.8977, "step": 35592 }, { "epoch": 2.79, "learning_rate": 7.209048851250686e-05, "loss": 2.6931, "step": 35593 }, { "epoch": 2.79, "learning_rate": 7.208970438328237e-05, "loss": 2.6799, "step": 35594 }, { "epoch": 2.79, "learning_rate": 7.208892025405787e-05, "loss": 2.7011, "step": 35595 }, { "epoch": 2.79, "learning_rate": 7.208813612483337e-05, "loss": 2.7741, "step": 35596 }, { "epoch": 2.79, "learning_rate": 7.208735199560888e-05, "loss": 2.9098, "step": 35597 }, { "epoch": 2.79, "learning_rate": 7.208656786638438e-05, "loss": 2.8454, "step": 35598 }, { "epoch": 2.79, "learning_rate": 7.208578373715989e-05, "loss": 2.5959, "step": 35599 }, { "epoch": 2.79, "learning_rate": 7.208499960793539e-05, "loss": 2.667, "step": 35600 }, { "epoch": 2.79, "learning_rate": 7.20842154787109e-05, "loss": 2.7334, "step": 35601 }, { "epoch": 2.79, "learning_rate": 7.20834313494864e-05, "loss": 2.7551, "step": 35602 }, { "epoch": 2.79, "learning_rate": 7.20826472202619e-05, "loss": 2.7605, "step": 35603 }, { "epoch": 2.79, "learning_rate": 7.20818630910374e-05, "loss": 2.6361, "step": 35604 }, { "epoch": 2.79, "learning_rate": 7.208107896181291e-05, "loss": 2.7457, "step": 35605 }, { "epoch": 2.79, "learning_rate": 7.208029483258841e-05, "loss": 2.8253, "step": 35606 }, { "epoch": 2.79, "learning_rate": 7.207951070336392e-05, "loss": 2.6951, "step": 35607 }, { "epoch": 2.79, "learning_rate": 7.207872657413942e-05, "loss": 2.9246, "step": 35608 }, { "epoch": 2.79, "learning_rate": 7.207794244491492e-05, "loss": 2.9134, "step": 35609 }, { "epoch": 2.79, "learning_rate": 7.207715831569043e-05, "loss": 2.7537, "step": 35610 }, { "epoch": 2.79, "learning_rate": 7.207637418646593e-05, "loss": 2.8851, "step": 35611 }, { "epoch": 2.79, "learning_rate": 7.207559005724143e-05, "loss": 2.6235, "step": 35612 }, { "epoch": 2.79, "learning_rate": 7.207480592801694e-05, "loss": 2.709, "step": 35613 }, { "epoch": 2.79, "learning_rate": 7.207402179879244e-05, "loss": 2.9529, "step": 35614 }, { "epoch": 2.79, "learning_rate": 7.207323766956794e-05, "loss": 2.7934, "step": 35615 }, { "epoch": 2.79, "learning_rate": 7.207245354034345e-05, "loss": 2.8803, "step": 35616 }, { "epoch": 2.79, "learning_rate": 7.207166941111895e-05, "loss": 2.7696, "step": 35617 }, { "epoch": 2.79, "learning_rate": 7.207088528189446e-05, "loss": 2.6024, "step": 35618 }, { "epoch": 2.79, "learning_rate": 7.207010115266997e-05, "loss": 2.7514, "step": 35619 }, { "epoch": 2.79, "learning_rate": 7.206931702344548e-05, "loss": 2.7739, "step": 35620 }, { "epoch": 2.79, "learning_rate": 7.206853289422098e-05, "loss": 2.8654, "step": 35621 }, { "epoch": 2.79, "learning_rate": 7.206774876499648e-05, "loss": 2.7892, "step": 35622 }, { "epoch": 2.79, "learning_rate": 7.206696463577199e-05, "loss": 2.6791, "step": 35623 }, { "epoch": 2.79, "learning_rate": 7.206618050654749e-05, "loss": 2.7843, "step": 35624 }, { "epoch": 2.79, "learning_rate": 7.2065396377323e-05, "loss": 2.8083, "step": 35625 }, { "epoch": 2.79, "learning_rate": 7.20646122480985e-05, "loss": 2.7455, "step": 35626 }, { "epoch": 2.79, "learning_rate": 7.2063828118874e-05, "loss": 2.9829, "step": 35627 }, { "epoch": 2.79, "learning_rate": 7.206304398964949e-05, "loss": 2.8142, "step": 35628 }, { "epoch": 2.79, "learning_rate": 7.2062259860425e-05, "loss": 2.712, "step": 35629 }, { "epoch": 2.79, "learning_rate": 7.20614757312005e-05, "loss": 2.7517, "step": 35630 }, { "epoch": 2.79, "learning_rate": 7.2060691601976e-05, "loss": 2.9671, "step": 35631 }, { "epoch": 2.79, "learning_rate": 7.205990747275151e-05, "loss": 2.5629, "step": 35632 }, { "epoch": 2.79, "learning_rate": 7.205912334352701e-05, "loss": 2.7924, "step": 35633 }, { "epoch": 2.79, "learning_rate": 7.205833921430251e-05, "loss": 2.7754, "step": 35634 }, { "epoch": 2.79, "learning_rate": 7.205755508507802e-05, "loss": 2.8503, "step": 35635 }, { "epoch": 2.79, "learning_rate": 7.205677095585352e-05, "loss": 2.7312, "step": 35636 }, { "epoch": 2.79, "learning_rate": 7.205598682662903e-05, "loss": 2.7067, "step": 35637 }, { "epoch": 2.79, "learning_rate": 7.205520269740453e-05, "loss": 2.733, "step": 35638 }, { "epoch": 2.79, "learning_rate": 7.205441856818003e-05, "loss": 3.041, "step": 35639 }, { "epoch": 2.79, "learning_rate": 7.205363443895554e-05, "loss": 2.8367, "step": 35640 }, { "epoch": 2.79, "learning_rate": 7.205285030973104e-05, "loss": 2.8194, "step": 35641 }, { "epoch": 2.79, "learning_rate": 7.205206618050654e-05, "loss": 2.6653, "step": 35642 }, { "epoch": 2.79, "learning_rate": 7.205128205128205e-05, "loss": 2.7855, "step": 35643 }, { "epoch": 2.79, "learning_rate": 7.205049792205756e-05, "loss": 2.9411, "step": 35644 }, { "epoch": 2.8, "learning_rate": 7.204971379283307e-05, "loss": 2.78, "step": 35645 }, { "epoch": 2.8, "learning_rate": 7.204892966360857e-05, "loss": 2.841, "step": 35646 }, { "epoch": 2.8, "learning_rate": 7.204814553438408e-05, "loss": 2.7417, "step": 35647 }, { "epoch": 2.8, "learning_rate": 7.204736140515958e-05, "loss": 2.861, "step": 35648 }, { "epoch": 2.8, "learning_rate": 7.204657727593508e-05, "loss": 2.718, "step": 35649 }, { "epoch": 2.8, "learning_rate": 7.204579314671059e-05, "loss": 2.8576, "step": 35650 }, { "epoch": 2.8, "learning_rate": 7.204500901748609e-05, "loss": 2.9348, "step": 35651 }, { "epoch": 2.8, "learning_rate": 7.20442248882616e-05, "loss": 2.7397, "step": 35652 }, { "epoch": 2.8, "learning_rate": 7.20434407590371e-05, "loss": 2.8137, "step": 35653 }, { "epoch": 2.8, "learning_rate": 7.20426566298126e-05, "loss": 2.7432, "step": 35654 }, { "epoch": 2.8, "learning_rate": 7.20418725005881e-05, "loss": 2.8199, "step": 35655 }, { "epoch": 2.8, "learning_rate": 7.204108837136361e-05, "loss": 2.6608, "step": 35656 }, { "epoch": 2.8, "learning_rate": 7.204030424213911e-05, "loss": 2.8943, "step": 35657 }, { "epoch": 2.8, "learning_rate": 7.203952011291462e-05, "loss": 2.8257, "step": 35658 }, { "epoch": 2.8, "learning_rate": 7.203873598369012e-05, "loss": 2.7411, "step": 35659 }, { "epoch": 2.8, "learning_rate": 7.203795185446562e-05, "loss": 2.8919, "step": 35660 }, { "epoch": 2.8, "learning_rate": 7.203716772524113e-05, "loss": 2.6781, "step": 35661 }, { "epoch": 2.8, "learning_rate": 7.203638359601663e-05, "loss": 2.8619, "step": 35662 }, { "epoch": 2.8, "learning_rate": 7.203559946679213e-05, "loss": 2.7903, "step": 35663 }, { "epoch": 2.8, "learning_rate": 7.203481533756762e-05, "loss": 2.7945, "step": 35664 }, { "epoch": 2.8, "learning_rate": 7.203403120834313e-05, "loss": 2.8271, "step": 35665 }, { "epoch": 2.8, "learning_rate": 7.203324707911863e-05, "loss": 2.7512, "step": 35666 }, { "epoch": 2.8, "learning_rate": 7.203246294989414e-05, "loss": 3.0563, "step": 35667 }, { "epoch": 2.8, "learning_rate": 7.203167882066964e-05, "loss": 2.7283, "step": 35668 }, { "epoch": 2.8, "learning_rate": 7.203089469144514e-05, "loss": 2.8161, "step": 35669 }, { "epoch": 2.8, "learning_rate": 7.203011056222066e-05, "loss": 2.8055, "step": 35670 }, { "epoch": 2.8, "learning_rate": 7.202932643299616e-05, "loss": 2.5312, "step": 35671 }, { "epoch": 2.8, "learning_rate": 7.202854230377167e-05, "loss": 2.7476, "step": 35672 }, { "epoch": 2.8, "learning_rate": 7.202775817454717e-05, "loss": 2.8001, "step": 35673 }, { "epoch": 2.8, "learning_rate": 7.202697404532268e-05, "loss": 3.1887, "step": 35674 }, { "epoch": 2.8, "learning_rate": 7.202618991609818e-05, "loss": 2.9278, "step": 35675 }, { "epoch": 2.8, "learning_rate": 7.202540578687368e-05, "loss": 2.6083, "step": 35676 }, { "epoch": 2.8, "learning_rate": 7.202462165764919e-05, "loss": 2.6292, "step": 35677 }, { "epoch": 2.8, "learning_rate": 7.202383752842469e-05, "loss": 2.5523, "step": 35678 }, { "epoch": 2.8, "learning_rate": 7.20230533992002e-05, "loss": 2.7986, "step": 35679 }, { "epoch": 2.8, "learning_rate": 7.20222692699757e-05, "loss": 2.775, "step": 35680 }, { "epoch": 2.8, "learning_rate": 7.20214851407512e-05, "loss": 3.389, "step": 35681 }, { "epoch": 2.8, "learning_rate": 7.20207010115267e-05, "loss": 2.8099, "step": 35682 }, { "epoch": 2.8, "learning_rate": 7.201991688230221e-05, "loss": 2.9283, "step": 35683 }, { "epoch": 2.8, "learning_rate": 7.201913275307771e-05, "loss": 2.6857, "step": 35684 }, { "epoch": 2.8, "learning_rate": 7.201834862385322e-05, "loss": 2.7107, "step": 35685 }, { "epoch": 2.8, "learning_rate": 7.201756449462872e-05, "loss": 2.6042, "step": 35686 }, { "epoch": 2.8, "learning_rate": 7.201678036540422e-05, "loss": 2.9588, "step": 35687 }, { "epoch": 2.8, "learning_rate": 7.201599623617973e-05, "loss": 2.7137, "step": 35688 }, { "epoch": 2.8, "learning_rate": 7.201521210695523e-05, "loss": 3.1338, "step": 35689 }, { "epoch": 2.8, "learning_rate": 7.201442797773073e-05, "loss": 2.8245, "step": 35690 }, { "epoch": 2.8, "learning_rate": 7.201364384850624e-05, "loss": 2.7978, "step": 35691 }, { "epoch": 2.8, "learning_rate": 7.201285971928174e-05, "loss": 2.8796, "step": 35692 }, { "epoch": 2.8, "learning_rate": 7.201207559005725e-05, "loss": 2.7387, "step": 35693 }, { "epoch": 2.8, "learning_rate": 7.201129146083275e-05, "loss": 2.874, "step": 35694 }, { "epoch": 2.8, "learning_rate": 7.201050733160825e-05, "loss": 2.7777, "step": 35695 }, { "epoch": 2.8, "learning_rate": 7.200972320238376e-05, "loss": 3.2607, "step": 35696 }, { "epoch": 2.8, "learning_rate": 7.200893907315926e-05, "loss": 2.8957, "step": 35697 }, { "epoch": 2.8, "learning_rate": 7.200815494393476e-05, "loss": 2.772, "step": 35698 }, { "epoch": 2.8, "learning_rate": 7.200737081471027e-05, "loss": 2.6449, "step": 35699 }, { "epoch": 2.8, "learning_rate": 7.200658668548577e-05, "loss": 2.6472, "step": 35700 }, { "epoch": 2.8, "learning_rate": 7.200580255626127e-05, "loss": 2.7019, "step": 35701 }, { "epoch": 2.8, "learning_rate": 7.200501842703678e-05, "loss": 2.7011, "step": 35702 }, { "epoch": 2.8, "learning_rate": 7.200423429781228e-05, "loss": 2.7474, "step": 35703 }, { "epoch": 2.8, "learning_rate": 7.200345016858779e-05, "loss": 2.8788, "step": 35704 }, { "epoch": 2.8, "learning_rate": 7.200266603936329e-05, "loss": 2.7572, "step": 35705 }, { "epoch": 2.8, "learning_rate": 7.200188191013879e-05, "loss": 2.5186, "step": 35706 }, { "epoch": 2.8, "learning_rate": 7.20010977809143e-05, "loss": 2.6647, "step": 35707 }, { "epoch": 2.8, "learning_rate": 7.20003136516898e-05, "loss": 2.5827, "step": 35708 }, { "epoch": 2.8, "learning_rate": 7.19995295224653e-05, "loss": 2.6776, "step": 35709 }, { "epoch": 2.8, "learning_rate": 7.199874539324081e-05, "loss": 2.7286, "step": 35710 }, { "epoch": 2.8, "learning_rate": 7.199796126401631e-05, "loss": 2.808, "step": 35711 }, { "epoch": 2.8, "learning_rate": 7.199717713479182e-05, "loss": 2.7919, "step": 35712 }, { "epoch": 2.8, "learning_rate": 7.199639300556732e-05, "loss": 2.8457, "step": 35713 }, { "epoch": 2.8, "learning_rate": 7.199560887634282e-05, "loss": 2.9146, "step": 35714 }, { "epoch": 2.8, "learning_rate": 7.199482474711833e-05, "loss": 3.2458, "step": 35715 }, { "epoch": 2.8, "learning_rate": 7.199404061789383e-05, "loss": 3.0345, "step": 35716 }, { "epoch": 2.8, "learning_rate": 7.199325648866933e-05, "loss": 2.7969, "step": 35717 }, { "epoch": 2.8, "learning_rate": 7.199247235944484e-05, "loss": 2.9305, "step": 35718 }, { "epoch": 2.8, "learning_rate": 7.199168823022034e-05, "loss": 2.947, "step": 35719 }, { "epoch": 2.8, "learning_rate": 7.199090410099584e-05, "loss": 2.9579, "step": 35720 }, { "epoch": 2.8, "learning_rate": 7.199011997177135e-05, "loss": 2.7812, "step": 35721 }, { "epoch": 2.8, "learning_rate": 7.198933584254685e-05, "loss": 2.8233, "step": 35722 }, { "epoch": 2.8, "learning_rate": 7.198855171332236e-05, "loss": 2.7672, "step": 35723 }, { "epoch": 2.8, "learning_rate": 7.198776758409786e-05, "loss": 2.9921, "step": 35724 }, { "epoch": 2.8, "learning_rate": 7.198698345487336e-05, "loss": 2.7655, "step": 35725 }, { "epoch": 2.8, "learning_rate": 7.198619932564887e-05, "loss": 3.1211, "step": 35726 }, { "epoch": 2.8, "learning_rate": 7.198541519642438e-05, "loss": 2.6797, "step": 35727 }, { "epoch": 2.8, "learning_rate": 7.198463106719989e-05, "loss": 2.8102, "step": 35728 }, { "epoch": 2.8, "learning_rate": 7.198384693797539e-05, "loss": 2.7903, "step": 35729 }, { "epoch": 2.8, "learning_rate": 7.19830628087509e-05, "loss": 2.5887, "step": 35730 }, { "epoch": 2.8, "learning_rate": 7.19822786795264e-05, "loss": 2.8311, "step": 35731 }, { "epoch": 2.8, "learning_rate": 7.19814945503019e-05, "loss": 2.7318, "step": 35732 }, { "epoch": 2.8, "learning_rate": 7.19807104210774e-05, "loss": 2.7484, "step": 35733 }, { "epoch": 2.8, "learning_rate": 7.197992629185291e-05, "loss": 2.7942, "step": 35734 }, { "epoch": 2.8, "learning_rate": 7.197914216262841e-05, "loss": 2.8599, "step": 35735 }, { "epoch": 2.8, "learning_rate": 7.19783580334039e-05, "loss": 2.7575, "step": 35736 }, { "epoch": 2.8, "learning_rate": 7.197757390417941e-05, "loss": 3.0041, "step": 35737 }, { "epoch": 2.8, "learning_rate": 7.197678977495491e-05, "loss": 2.7774, "step": 35738 }, { "epoch": 2.8, "learning_rate": 7.197600564573041e-05, "loss": 3.021, "step": 35739 }, { "epoch": 2.8, "learning_rate": 7.197522151650592e-05, "loss": 2.9045, "step": 35740 }, { "epoch": 2.8, "learning_rate": 7.197443738728142e-05, "loss": 2.7821, "step": 35741 }, { "epoch": 2.8, "learning_rate": 7.197365325805693e-05, "loss": 2.7474, "step": 35742 }, { "epoch": 2.8, "learning_rate": 7.197286912883243e-05, "loss": 3.1621, "step": 35743 }, { "epoch": 2.8, "learning_rate": 7.197208499960793e-05, "loss": 2.709, "step": 35744 }, { "epoch": 2.8, "learning_rate": 7.197130087038344e-05, "loss": 3.0028, "step": 35745 }, { "epoch": 2.8, "learning_rate": 7.197051674115894e-05, "loss": 2.5976, "step": 35746 }, { "epoch": 2.8, "learning_rate": 7.196973261193444e-05, "loss": 3.0522, "step": 35747 }, { "epoch": 2.8, "learning_rate": 7.196894848270995e-05, "loss": 2.8914, "step": 35748 }, { "epoch": 2.8, "learning_rate": 7.196816435348545e-05, "loss": 2.8613, "step": 35749 }, { "epoch": 2.8, "learning_rate": 7.196738022426095e-05, "loss": 3.0848, "step": 35750 }, { "epoch": 2.8, "learning_rate": 7.196659609503646e-05, "loss": 2.6721, "step": 35751 }, { "epoch": 2.8, "learning_rate": 7.196581196581196e-05, "loss": 2.8846, "step": 35752 }, { "epoch": 2.8, "learning_rate": 7.196502783658748e-05, "loss": 2.7945, "step": 35753 }, { "epoch": 2.8, "learning_rate": 7.196424370736298e-05, "loss": 2.8807, "step": 35754 }, { "epoch": 2.8, "learning_rate": 7.196345957813849e-05, "loss": 2.7835, "step": 35755 }, { "epoch": 2.8, "learning_rate": 7.196267544891399e-05, "loss": 2.8577, "step": 35756 }, { "epoch": 2.8, "learning_rate": 7.19618913196895e-05, "loss": 2.7403, "step": 35757 }, { "epoch": 2.8, "learning_rate": 7.1961107190465e-05, "loss": 2.6853, "step": 35758 }, { "epoch": 2.8, "learning_rate": 7.19603230612405e-05, "loss": 2.6901, "step": 35759 }, { "epoch": 2.8, "learning_rate": 7.1959538932016e-05, "loss": 2.7703, "step": 35760 }, { "epoch": 2.8, "learning_rate": 7.195875480279151e-05, "loss": 2.7556, "step": 35761 }, { "epoch": 2.8, "learning_rate": 7.195797067356701e-05, "loss": 2.6241, "step": 35762 }, { "epoch": 2.8, "learning_rate": 7.195718654434252e-05, "loss": 2.8591, "step": 35763 }, { "epoch": 2.8, "learning_rate": 7.195640241511802e-05, "loss": 2.8369, "step": 35764 }, { "epoch": 2.8, "learning_rate": 7.195561828589352e-05, "loss": 2.8654, "step": 35765 }, { "epoch": 2.8, "learning_rate": 7.195483415666903e-05, "loss": 2.9388, "step": 35766 }, { "epoch": 2.8, "learning_rate": 7.195405002744453e-05, "loss": 2.7701, "step": 35767 }, { "epoch": 2.8, "learning_rate": 7.195326589822003e-05, "loss": 2.8541, "step": 35768 }, { "epoch": 2.8, "learning_rate": 7.195248176899554e-05, "loss": 2.7353, "step": 35769 }, { "epoch": 2.8, "learning_rate": 7.195169763977104e-05, "loss": 2.6929, "step": 35770 }, { "epoch": 2.8, "learning_rate": 7.195091351054655e-05, "loss": 2.8047, "step": 35771 }, { "epoch": 2.8, "learning_rate": 7.195012938132204e-05, "loss": 2.5986, "step": 35772 }, { "epoch": 2.81, "learning_rate": 7.194934525209754e-05, "loss": 2.8122, "step": 35773 }, { "epoch": 2.81, "learning_rate": 7.194856112287304e-05, "loss": 2.8478, "step": 35774 }, { "epoch": 2.81, "learning_rate": 7.194777699364855e-05, "loss": 2.8178, "step": 35775 }, { "epoch": 2.81, "learning_rate": 7.194699286442405e-05, "loss": 2.7783, "step": 35776 }, { "epoch": 2.81, "learning_rate": 7.194620873519955e-05, "loss": 2.6743, "step": 35777 }, { "epoch": 2.81, "learning_rate": 7.194542460597507e-05, "loss": 2.8393, "step": 35778 }, { "epoch": 2.81, "learning_rate": 7.194464047675058e-05, "loss": 2.6518, "step": 35779 }, { "epoch": 2.81, "learning_rate": 7.194385634752608e-05, "loss": 2.8397, "step": 35780 }, { "epoch": 2.81, "learning_rate": 7.194307221830158e-05, "loss": 2.696, "step": 35781 }, { "epoch": 2.81, "learning_rate": 7.194228808907709e-05, "loss": 3.0104, "step": 35782 }, { "epoch": 2.81, "learning_rate": 7.194150395985259e-05, "loss": 2.6711, "step": 35783 }, { "epoch": 2.81, "learning_rate": 7.19407198306281e-05, "loss": 2.8411, "step": 35784 }, { "epoch": 2.81, "learning_rate": 7.19399357014036e-05, "loss": 2.8038, "step": 35785 }, { "epoch": 2.81, "learning_rate": 7.19391515721791e-05, "loss": 2.9564, "step": 35786 }, { "epoch": 2.81, "learning_rate": 7.19383674429546e-05, "loss": 2.7558, "step": 35787 }, { "epoch": 2.81, "learning_rate": 7.193758331373011e-05, "loss": 3.0246, "step": 35788 }, { "epoch": 2.81, "learning_rate": 7.193679918450561e-05, "loss": 2.7398, "step": 35789 }, { "epoch": 2.81, "learning_rate": 7.193601505528112e-05, "loss": 2.692, "step": 35790 }, { "epoch": 2.81, "learning_rate": 7.193523092605662e-05, "loss": 3.156, "step": 35791 }, { "epoch": 2.81, "learning_rate": 7.193444679683212e-05, "loss": 2.6937, "step": 35792 }, { "epoch": 2.81, "learning_rate": 7.193366266760763e-05, "loss": 2.6532, "step": 35793 }, { "epoch": 2.81, "learning_rate": 7.193287853838313e-05, "loss": 2.9971, "step": 35794 }, { "epoch": 2.81, "learning_rate": 7.193209440915863e-05, "loss": 2.7824, "step": 35795 }, { "epoch": 2.81, "learning_rate": 7.193131027993414e-05, "loss": 2.9304, "step": 35796 }, { "epoch": 2.81, "learning_rate": 7.193052615070964e-05, "loss": 2.8102, "step": 35797 }, { "epoch": 2.81, "learning_rate": 7.192974202148514e-05, "loss": 2.7333, "step": 35798 }, { "epoch": 2.81, "learning_rate": 7.192895789226065e-05, "loss": 2.7289, "step": 35799 }, { "epoch": 2.81, "learning_rate": 7.192817376303615e-05, "loss": 2.918, "step": 35800 }, { "epoch": 2.81, "learning_rate": 7.192738963381166e-05, "loss": 3.0821, "step": 35801 }, { "epoch": 2.81, "learning_rate": 7.192660550458716e-05, "loss": 3.0688, "step": 35802 }, { "epoch": 2.81, "learning_rate": 7.192582137536266e-05, "loss": 2.6536, "step": 35803 }, { "epoch": 2.81, "learning_rate": 7.192503724613817e-05, "loss": 2.6676, "step": 35804 }, { "epoch": 2.81, "learning_rate": 7.192425311691367e-05, "loss": 2.8278, "step": 35805 }, { "epoch": 2.81, "learning_rate": 7.192346898768917e-05, "loss": 2.7365, "step": 35806 }, { "epoch": 2.81, "learning_rate": 7.192268485846468e-05, "loss": 2.8142, "step": 35807 }, { "epoch": 2.81, "learning_rate": 7.192190072924018e-05, "loss": 2.7586, "step": 35808 }, { "epoch": 2.81, "learning_rate": 7.192111660001569e-05, "loss": 2.6138, "step": 35809 }, { "epoch": 2.81, "learning_rate": 7.192033247079119e-05, "loss": 2.5805, "step": 35810 }, { "epoch": 2.81, "learning_rate": 7.191954834156669e-05, "loss": 2.8766, "step": 35811 }, { "epoch": 2.81, "learning_rate": 7.19187642123422e-05, "loss": 2.8185, "step": 35812 }, { "epoch": 2.81, "learning_rate": 7.19179800831177e-05, "loss": 2.822, "step": 35813 }, { "epoch": 2.81, "learning_rate": 7.19171959538932e-05, "loss": 2.8573, "step": 35814 }, { "epoch": 2.81, "learning_rate": 7.191641182466871e-05, "loss": 2.6208, "step": 35815 }, { "epoch": 2.81, "learning_rate": 7.191562769544421e-05, "loss": 2.9674, "step": 35816 }, { "epoch": 2.81, "learning_rate": 7.191484356621971e-05, "loss": 2.6198, "step": 35817 }, { "epoch": 2.81, "learning_rate": 7.191405943699522e-05, "loss": 2.7493, "step": 35818 }, { "epoch": 2.81, "learning_rate": 7.191327530777072e-05, "loss": 2.6704, "step": 35819 }, { "epoch": 2.81, "learning_rate": 7.191249117854623e-05, "loss": 2.8455, "step": 35820 }, { "epoch": 2.81, "learning_rate": 7.191170704932173e-05, "loss": 2.7377, "step": 35821 }, { "epoch": 2.81, "learning_rate": 7.191092292009723e-05, "loss": 3.0163, "step": 35822 }, { "epoch": 2.81, "learning_rate": 7.191013879087274e-05, "loss": 2.8266, "step": 35823 }, { "epoch": 2.81, "learning_rate": 7.190935466164824e-05, "loss": 2.7856, "step": 35824 }, { "epoch": 2.81, "learning_rate": 7.190857053242374e-05, "loss": 2.7288, "step": 35825 }, { "epoch": 2.81, "learning_rate": 7.190778640319925e-05, "loss": 2.8653, "step": 35826 }, { "epoch": 2.81, "learning_rate": 7.190700227397475e-05, "loss": 2.6639, "step": 35827 }, { "epoch": 2.81, "learning_rate": 7.190621814475026e-05, "loss": 2.9859, "step": 35828 }, { "epoch": 2.81, "learning_rate": 7.190543401552576e-05, "loss": 2.8357, "step": 35829 }, { "epoch": 2.81, "learning_rate": 7.190464988630126e-05, "loss": 2.9417, "step": 35830 }, { "epoch": 2.81, "learning_rate": 7.190386575707677e-05, "loss": 2.7577, "step": 35831 }, { "epoch": 2.81, "learning_rate": 7.190308162785227e-05, "loss": 2.8434, "step": 35832 }, { "epoch": 2.81, "learning_rate": 7.190229749862777e-05, "loss": 2.7315, "step": 35833 }, { "epoch": 2.81, "learning_rate": 7.190151336940328e-05, "loss": 2.652, "step": 35834 }, { "epoch": 2.81, "learning_rate": 7.19007292401788e-05, "loss": 2.5735, "step": 35835 }, { "epoch": 2.81, "learning_rate": 7.18999451109543e-05, "loss": 2.7878, "step": 35836 }, { "epoch": 2.81, "learning_rate": 7.18991609817298e-05, "loss": 2.8437, "step": 35837 }, { "epoch": 2.81, "learning_rate": 7.18983768525053e-05, "loss": 2.9424, "step": 35838 }, { "epoch": 2.81, "learning_rate": 7.189759272328081e-05, "loss": 2.6948, "step": 35839 }, { "epoch": 2.81, "learning_rate": 7.189680859405631e-05, "loss": 2.8016, "step": 35840 }, { "epoch": 2.81, "learning_rate": 7.189602446483182e-05, "loss": 2.8277, "step": 35841 }, { "epoch": 2.81, "learning_rate": 7.189524033560732e-05, "loss": 2.758, "step": 35842 }, { "epoch": 2.81, "learning_rate": 7.189445620638282e-05, "loss": 2.6784, "step": 35843 }, { "epoch": 2.81, "learning_rate": 7.189367207715831e-05, "loss": 2.7602, "step": 35844 }, { "epoch": 2.81, "learning_rate": 7.189288794793382e-05, "loss": 2.737, "step": 35845 }, { "epoch": 2.81, "learning_rate": 7.189210381870932e-05, "loss": 2.9062, "step": 35846 }, { "epoch": 2.81, "learning_rate": 7.189131968948483e-05, "loss": 2.58, "step": 35847 }, { "epoch": 2.81, "learning_rate": 7.189053556026033e-05, "loss": 3.0025, "step": 35848 }, { "epoch": 2.81, "learning_rate": 7.188975143103583e-05, "loss": 2.6847, "step": 35849 }, { "epoch": 2.81, "learning_rate": 7.188896730181134e-05, "loss": 2.8853, "step": 35850 }, { "epoch": 2.81, "learning_rate": 7.188818317258684e-05, "loss": 3.0407, "step": 35851 }, { "epoch": 2.81, "learning_rate": 7.188739904336234e-05, "loss": 2.7443, "step": 35852 }, { "epoch": 2.81, "learning_rate": 7.188661491413785e-05, "loss": 2.7655, "step": 35853 }, { "epoch": 2.81, "learning_rate": 7.188583078491335e-05, "loss": 2.9223, "step": 35854 }, { "epoch": 2.81, "learning_rate": 7.188504665568885e-05, "loss": 2.6848, "step": 35855 }, { "epoch": 2.81, "learning_rate": 7.188426252646436e-05, "loss": 2.9547, "step": 35856 }, { "epoch": 2.81, "learning_rate": 7.188347839723986e-05, "loss": 2.6876, "step": 35857 }, { "epoch": 2.81, "learning_rate": 7.188269426801537e-05, "loss": 2.9677, "step": 35858 }, { "epoch": 2.81, "learning_rate": 7.188191013879087e-05, "loss": 2.8069, "step": 35859 }, { "epoch": 2.81, "learning_rate": 7.188112600956637e-05, "loss": 2.7809, "step": 35860 }, { "epoch": 2.81, "learning_rate": 7.188034188034189e-05, "loss": 2.7639, "step": 35861 }, { "epoch": 2.81, "learning_rate": 7.18795577511174e-05, "loss": 2.6209, "step": 35862 }, { "epoch": 2.81, "learning_rate": 7.18787736218929e-05, "loss": 2.8055, "step": 35863 }, { "epoch": 2.81, "learning_rate": 7.18779894926684e-05, "loss": 2.6003, "step": 35864 }, { "epoch": 2.81, "learning_rate": 7.18772053634439e-05, "loss": 2.7128, "step": 35865 }, { "epoch": 2.81, "learning_rate": 7.187642123421941e-05, "loss": 2.6983, "step": 35866 }, { "epoch": 2.81, "learning_rate": 7.187563710499491e-05, "loss": 2.7027, "step": 35867 }, { "epoch": 2.81, "learning_rate": 7.187485297577042e-05, "loss": 2.6872, "step": 35868 }, { "epoch": 2.81, "learning_rate": 7.187406884654592e-05, "loss": 2.9867, "step": 35869 }, { "epoch": 2.81, "learning_rate": 7.187328471732142e-05, "loss": 2.6801, "step": 35870 }, { "epoch": 2.81, "learning_rate": 7.187250058809693e-05, "loss": 2.6787, "step": 35871 }, { "epoch": 2.81, "learning_rate": 7.187171645887243e-05, "loss": 2.8364, "step": 35872 }, { "epoch": 2.81, "learning_rate": 7.187093232964793e-05, "loss": 2.7681, "step": 35873 }, { "epoch": 2.81, "learning_rate": 7.187014820042344e-05, "loss": 2.802, "step": 35874 }, { "epoch": 2.81, "learning_rate": 7.186936407119894e-05, "loss": 2.6592, "step": 35875 }, { "epoch": 2.81, "learning_rate": 7.186857994197445e-05, "loss": 2.7696, "step": 35876 }, { "epoch": 2.81, "learning_rate": 7.186779581274995e-05, "loss": 2.9118, "step": 35877 }, { "epoch": 2.81, "learning_rate": 7.186701168352545e-05, "loss": 2.8567, "step": 35878 }, { "epoch": 2.81, "learning_rate": 7.186622755430096e-05, "loss": 3.0172, "step": 35879 }, { "epoch": 2.81, "learning_rate": 7.186544342507645e-05, "loss": 2.8547, "step": 35880 }, { "epoch": 2.81, "learning_rate": 7.186465929585195e-05, "loss": 2.7671, "step": 35881 }, { "epoch": 2.81, "learning_rate": 7.186387516662745e-05, "loss": 2.7906, "step": 35882 }, { "epoch": 2.81, "learning_rate": 7.186309103740296e-05, "loss": 3.0207, "step": 35883 }, { "epoch": 2.81, "learning_rate": 7.186230690817846e-05, "loss": 2.8075, "step": 35884 }, { "epoch": 2.81, "learning_rate": 7.186152277895396e-05, "loss": 2.7588, "step": 35885 }, { "epoch": 2.81, "learning_rate": 7.186073864972948e-05, "loss": 2.9697, "step": 35886 }, { "epoch": 2.81, "learning_rate": 7.185995452050499e-05, "loss": 2.6325, "step": 35887 }, { "epoch": 2.81, "learning_rate": 7.185917039128049e-05, "loss": 3.0106, "step": 35888 }, { "epoch": 2.81, "learning_rate": 7.185838626205599e-05, "loss": 2.7262, "step": 35889 }, { "epoch": 2.81, "learning_rate": 7.18576021328315e-05, "loss": 2.9987, "step": 35890 }, { "epoch": 2.81, "learning_rate": 7.1856818003607e-05, "loss": 3.0598, "step": 35891 }, { "epoch": 2.81, "learning_rate": 7.18560338743825e-05, "loss": 2.9558, "step": 35892 }, { "epoch": 2.81, "learning_rate": 7.185524974515801e-05, "loss": 2.7044, "step": 35893 }, { "epoch": 2.81, "learning_rate": 7.185446561593351e-05, "loss": 2.964, "step": 35894 }, { "epoch": 2.81, "learning_rate": 7.185368148670902e-05, "loss": 2.6611, "step": 35895 }, { "epoch": 2.81, "learning_rate": 7.185289735748452e-05, "loss": 2.875, "step": 35896 }, { "epoch": 2.81, "learning_rate": 7.185211322826002e-05, "loss": 2.7173, "step": 35897 }, { "epoch": 2.81, "learning_rate": 7.185132909903553e-05, "loss": 2.9406, "step": 35898 }, { "epoch": 2.81, "learning_rate": 7.185054496981103e-05, "loss": 2.727, "step": 35899 }, { "epoch": 2.82, "learning_rate": 7.184976084058653e-05, "loss": 2.866, "step": 35900 }, { "epoch": 2.82, "learning_rate": 7.184897671136204e-05, "loss": 2.9945, "step": 35901 }, { "epoch": 2.82, "learning_rate": 7.184819258213754e-05, "loss": 2.7721, "step": 35902 }, { "epoch": 2.82, "learning_rate": 7.184740845291304e-05, "loss": 2.9469, "step": 35903 }, { "epoch": 2.82, "learning_rate": 7.184662432368855e-05, "loss": 3.0838, "step": 35904 }, { "epoch": 2.82, "learning_rate": 7.184584019446405e-05, "loss": 2.6936, "step": 35905 }, { "epoch": 2.82, "learning_rate": 7.184505606523956e-05, "loss": 3.0735, "step": 35906 }, { "epoch": 2.82, "learning_rate": 7.184427193601506e-05, "loss": 3.0885, "step": 35907 }, { "epoch": 2.82, "learning_rate": 7.184348780679056e-05, "loss": 2.8036, "step": 35908 }, { "epoch": 2.82, "learning_rate": 7.184270367756607e-05, "loss": 2.8153, "step": 35909 }, { "epoch": 2.82, "learning_rate": 7.184191954834157e-05, "loss": 2.7099, "step": 35910 }, { "epoch": 2.82, "learning_rate": 7.184113541911707e-05, "loss": 2.9142, "step": 35911 }, { "epoch": 2.82, "learning_rate": 7.184035128989258e-05, "loss": 2.6388, "step": 35912 }, { "epoch": 2.82, "learning_rate": 7.183956716066808e-05, "loss": 2.9659, "step": 35913 }, { "epoch": 2.82, "learning_rate": 7.183878303144359e-05, "loss": 2.8962, "step": 35914 }, { "epoch": 2.82, "learning_rate": 7.183799890221909e-05, "loss": 2.6525, "step": 35915 }, { "epoch": 2.82, "learning_rate": 7.183721477299459e-05, "loss": 2.936, "step": 35916 }, { "epoch": 2.82, "learning_rate": 7.18364306437701e-05, "loss": 2.8976, "step": 35917 }, { "epoch": 2.82, "learning_rate": 7.18356465145456e-05, "loss": 2.7471, "step": 35918 }, { "epoch": 2.82, "learning_rate": 7.18348623853211e-05, "loss": 3.0134, "step": 35919 }, { "epoch": 2.82, "learning_rate": 7.183407825609661e-05, "loss": 2.8132, "step": 35920 }, { "epoch": 2.82, "learning_rate": 7.183329412687211e-05, "loss": 3.1077, "step": 35921 }, { "epoch": 2.82, "learning_rate": 7.183250999764761e-05, "loss": 2.7067, "step": 35922 }, { "epoch": 2.82, "learning_rate": 7.183172586842312e-05, "loss": 2.9307, "step": 35923 }, { "epoch": 2.82, "learning_rate": 7.183094173919862e-05, "loss": 2.6644, "step": 35924 }, { "epoch": 2.82, "learning_rate": 7.183015760997413e-05, "loss": 2.6509, "step": 35925 }, { "epoch": 2.82, "learning_rate": 7.182937348074963e-05, "loss": 2.9106, "step": 35926 }, { "epoch": 2.82, "learning_rate": 7.182858935152513e-05, "loss": 2.6141, "step": 35927 }, { "epoch": 2.82, "learning_rate": 7.182780522230064e-05, "loss": 2.9683, "step": 35928 }, { "epoch": 2.82, "learning_rate": 7.182702109307614e-05, "loss": 2.9405, "step": 35929 }, { "epoch": 2.82, "learning_rate": 7.182623696385164e-05, "loss": 2.8067, "step": 35930 }, { "epoch": 2.82, "learning_rate": 7.182545283462715e-05, "loss": 2.6127, "step": 35931 }, { "epoch": 2.82, "learning_rate": 7.182466870540265e-05, "loss": 2.835, "step": 35932 }, { "epoch": 2.82, "learning_rate": 7.182388457617815e-05, "loss": 2.9026, "step": 35933 }, { "epoch": 2.82, "learning_rate": 7.182310044695366e-05, "loss": 2.7854, "step": 35934 }, { "epoch": 2.82, "learning_rate": 7.182231631772916e-05, "loss": 2.6793, "step": 35935 }, { "epoch": 2.82, "learning_rate": 7.182153218850467e-05, "loss": 2.8664, "step": 35936 }, { "epoch": 2.82, "learning_rate": 7.182074805928017e-05, "loss": 2.7172, "step": 35937 }, { "epoch": 2.82, "learning_rate": 7.181996393005567e-05, "loss": 2.6949, "step": 35938 }, { "epoch": 2.82, "learning_rate": 7.181917980083118e-05, "loss": 2.8132, "step": 35939 }, { "epoch": 2.82, "learning_rate": 7.181839567160668e-05, "loss": 2.6782, "step": 35940 }, { "epoch": 2.82, "learning_rate": 7.181761154238218e-05, "loss": 2.8389, "step": 35941 }, { "epoch": 2.82, "learning_rate": 7.181682741315769e-05, "loss": 2.6953, "step": 35942 }, { "epoch": 2.82, "learning_rate": 7.18160432839332e-05, "loss": 2.5875, "step": 35943 }, { "epoch": 2.82, "learning_rate": 7.181525915470871e-05, "loss": 3.1021, "step": 35944 }, { "epoch": 2.82, "learning_rate": 7.181447502548421e-05, "loss": 2.6425, "step": 35945 }, { "epoch": 2.82, "learning_rate": 7.181369089625972e-05, "loss": 2.8995, "step": 35946 }, { "epoch": 2.82, "learning_rate": 7.181290676703522e-05, "loss": 2.8692, "step": 35947 }, { "epoch": 2.82, "learning_rate": 7.181212263781072e-05, "loss": 2.732, "step": 35948 }, { "epoch": 2.82, "learning_rate": 7.181133850858623e-05, "loss": 2.6685, "step": 35949 }, { "epoch": 2.82, "learning_rate": 7.181055437936173e-05, "loss": 2.7925, "step": 35950 }, { "epoch": 2.82, "learning_rate": 7.180977025013723e-05, "loss": 2.7935, "step": 35951 }, { "epoch": 2.82, "learning_rate": 7.180898612091272e-05, "loss": 2.6711, "step": 35952 }, { "epoch": 2.82, "learning_rate": 7.180820199168823e-05, "loss": 2.8815, "step": 35953 }, { "epoch": 2.82, "learning_rate": 7.180741786246373e-05, "loss": 2.7929, "step": 35954 }, { "epoch": 2.82, "learning_rate": 7.180663373323924e-05, "loss": 2.7901, "step": 35955 }, { "epoch": 2.82, "learning_rate": 7.180584960401474e-05, "loss": 2.5872, "step": 35956 }, { "epoch": 2.82, "learning_rate": 7.180506547479024e-05, "loss": 2.9735, "step": 35957 }, { "epoch": 2.82, "learning_rate": 7.180428134556575e-05, "loss": 3.0342, "step": 35958 }, { "epoch": 2.82, "learning_rate": 7.180349721634125e-05, "loss": 2.9511, "step": 35959 }, { "epoch": 2.82, "learning_rate": 7.180271308711675e-05, "loss": 2.7382, "step": 35960 }, { "epoch": 2.82, "learning_rate": 7.180192895789226e-05, "loss": 2.6641, "step": 35961 }, { "epoch": 2.82, "learning_rate": 7.180114482866776e-05, "loss": 2.9656, "step": 35962 }, { "epoch": 2.82, "learning_rate": 7.180036069944327e-05, "loss": 2.6156, "step": 35963 }, { "epoch": 2.82, "learning_rate": 7.179957657021877e-05, "loss": 2.7509, "step": 35964 }, { "epoch": 2.82, "learning_rate": 7.179879244099427e-05, "loss": 2.6548, "step": 35965 }, { "epoch": 2.82, "learning_rate": 7.179800831176978e-05, "loss": 2.8858, "step": 35966 }, { "epoch": 2.82, "learning_rate": 7.179722418254528e-05, "loss": 2.8757, "step": 35967 }, { "epoch": 2.82, "learning_rate": 7.179644005332078e-05, "loss": 2.8605, "step": 35968 }, { "epoch": 2.82, "learning_rate": 7.17956559240963e-05, "loss": 2.6728, "step": 35969 }, { "epoch": 2.82, "learning_rate": 7.17948717948718e-05, "loss": 2.7124, "step": 35970 }, { "epoch": 2.82, "learning_rate": 7.179408766564731e-05, "loss": 2.7337, "step": 35971 }, { "epoch": 2.82, "learning_rate": 7.179330353642281e-05, "loss": 2.8981, "step": 35972 }, { "epoch": 2.82, "learning_rate": 7.179251940719832e-05, "loss": 2.9459, "step": 35973 }, { "epoch": 2.82, "learning_rate": 7.179173527797382e-05, "loss": 2.9903, "step": 35974 }, { "epoch": 2.82, "learning_rate": 7.179095114874932e-05, "loss": 2.7759, "step": 35975 }, { "epoch": 2.82, "learning_rate": 7.179016701952483e-05, "loss": 2.8906, "step": 35976 }, { "epoch": 2.82, "learning_rate": 7.178938289030033e-05, "loss": 2.8825, "step": 35977 }, { "epoch": 2.82, "learning_rate": 7.178859876107583e-05, "loss": 3.0264, "step": 35978 }, { "epoch": 2.82, "learning_rate": 7.178781463185134e-05, "loss": 2.7944, "step": 35979 }, { "epoch": 2.82, "learning_rate": 7.178703050262684e-05, "loss": 2.8987, "step": 35980 }, { "epoch": 2.82, "learning_rate": 7.178624637340235e-05, "loss": 2.6765, "step": 35981 }, { "epoch": 2.82, "learning_rate": 7.178546224417785e-05, "loss": 2.7914, "step": 35982 }, { "epoch": 2.82, "learning_rate": 7.178467811495335e-05, "loss": 2.7037, "step": 35983 }, { "epoch": 2.82, "learning_rate": 7.178389398572886e-05, "loss": 2.6651, "step": 35984 }, { "epoch": 2.82, "learning_rate": 7.178310985650436e-05, "loss": 2.9482, "step": 35985 }, { "epoch": 2.82, "learning_rate": 7.178232572727986e-05, "loss": 2.5999, "step": 35986 }, { "epoch": 2.82, "learning_rate": 7.178154159805537e-05, "loss": 2.5799, "step": 35987 }, { "epoch": 2.82, "learning_rate": 7.178075746883086e-05, "loss": 2.8551, "step": 35988 }, { "epoch": 2.82, "learning_rate": 7.177997333960636e-05, "loss": 2.7956, "step": 35989 }, { "epoch": 2.82, "learning_rate": 7.177918921038186e-05, "loss": 2.7845, "step": 35990 }, { "epoch": 2.82, "learning_rate": 7.177840508115737e-05, "loss": 2.7998, "step": 35991 }, { "epoch": 2.82, "learning_rate": 7.177762095193287e-05, "loss": 2.7905, "step": 35992 }, { "epoch": 2.82, "learning_rate": 7.177683682270838e-05, "loss": 2.866, "step": 35993 }, { "epoch": 2.82, "learning_rate": 7.177605269348389e-05, "loss": 2.8662, "step": 35994 }, { "epoch": 2.82, "learning_rate": 7.17752685642594e-05, "loss": 2.7948, "step": 35995 }, { "epoch": 2.82, "learning_rate": 7.17744844350349e-05, "loss": 2.9005, "step": 35996 }, { "epoch": 2.82, "learning_rate": 7.17737003058104e-05, "loss": 2.9105, "step": 35997 }, { "epoch": 2.82, "learning_rate": 7.177291617658591e-05, "loss": 3.0117, "step": 35998 }, { "epoch": 2.82, "learning_rate": 7.177213204736141e-05, "loss": 2.6603, "step": 35999 }, { "epoch": 2.82, "learning_rate": 7.177134791813691e-05, "loss": 2.7628, "step": 36000 }, { "epoch": 2.82, "eval_loss": 2.8916587829589844, "eval_runtime": 217.9389, "eval_samples_per_second": 51.996, "eval_steps_per_second": 3.253, "step": 36000 }, { "epoch": 2.82, "learning_rate": 7.177056378891242e-05, "loss": 2.7918, "step": 36001 }, { "epoch": 2.82, "learning_rate": 7.176977965968792e-05, "loss": 2.7574, "step": 36002 }, { "epoch": 2.82, "learning_rate": 7.176899553046343e-05, "loss": 2.7296, "step": 36003 }, { "epoch": 2.82, "learning_rate": 7.176821140123893e-05, "loss": 2.8266, "step": 36004 }, { "epoch": 2.82, "learning_rate": 7.176742727201443e-05, "loss": 2.9003, "step": 36005 }, { "epoch": 2.82, "learning_rate": 7.176664314278994e-05, "loss": 2.9339, "step": 36006 }, { "epoch": 2.82, "learning_rate": 7.176585901356544e-05, "loss": 2.7638, "step": 36007 }, { "epoch": 2.82, "learning_rate": 7.176507488434094e-05, "loss": 2.9103, "step": 36008 }, { "epoch": 2.82, "learning_rate": 7.176429075511645e-05, "loss": 2.8027, "step": 36009 }, { "epoch": 2.82, "learning_rate": 7.176350662589195e-05, "loss": 2.6888, "step": 36010 }, { "epoch": 2.82, "learning_rate": 7.176272249666746e-05, "loss": 2.809, "step": 36011 }, { "epoch": 2.82, "learning_rate": 7.176193836744296e-05, "loss": 2.9796, "step": 36012 }, { "epoch": 2.82, "learning_rate": 7.176115423821846e-05, "loss": 2.7944, "step": 36013 }, { "epoch": 2.82, "learning_rate": 7.176037010899397e-05, "loss": 2.9177, "step": 36014 }, { "epoch": 2.82, "learning_rate": 7.175958597976947e-05, "loss": 2.6331, "step": 36015 }, { "epoch": 2.82, "learning_rate": 7.175880185054497e-05, "loss": 2.8356, "step": 36016 }, { "epoch": 2.82, "learning_rate": 7.175801772132048e-05, "loss": 2.8339, "step": 36017 }, { "epoch": 2.82, "learning_rate": 7.175723359209598e-05, "loss": 2.794, "step": 36018 }, { "epoch": 2.82, "learning_rate": 7.175644946287148e-05, "loss": 3.094, "step": 36019 }, { "epoch": 2.82, "learning_rate": 7.175566533364699e-05, "loss": 2.768, "step": 36020 }, { "epoch": 2.82, "learning_rate": 7.175488120442249e-05, "loss": 2.8399, "step": 36021 }, { "epoch": 2.82, "learning_rate": 7.1754097075198e-05, "loss": 2.9417, "step": 36022 }, { "epoch": 2.82, "learning_rate": 7.17533129459735e-05, "loss": 2.8753, "step": 36023 }, { "epoch": 2.82, "learning_rate": 7.1752528816749e-05, "loss": 3.0483, "step": 36024 }, { "epoch": 2.82, "learning_rate": 7.175174468752451e-05, "loss": 2.9632, "step": 36025 }, { "epoch": 2.82, "learning_rate": 7.175096055830001e-05, "loss": 2.9219, "step": 36026 }, { "epoch": 2.82, "learning_rate": 7.175017642907551e-05, "loss": 2.7907, "step": 36027 }, { "epoch": 2.83, "learning_rate": 7.174939229985102e-05, "loss": 2.6625, "step": 36028 }, { "epoch": 2.83, "learning_rate": 7.174860817062652e-05, "loss": 2.7257, "step": 36029 }, { "epoch": 2.83, "learning_rate": 7.174782404140203e-05, "loss": 2.8227, "step": 36030 }, { "epoch": 2.83, "learning_rate": 7.174703991217753e-05, "loss": 2.6647, "step": 36031 }, { "epoch": 2.83, "learning_rate": 7.174625578295303e-05, "loss": 2.908, "step": 36032 }, { "epoch": 2.83, "learning_rate": 7.174547165372854e-05, "loss": 2.9694, "step": 36033 }, { "epoch": 2.83, "learning_rate": 7.174468752450404e-05, "loss": 2.8715, "step": 36034 }, { "epoch": 2.83, "learning_rate": 7.174390339527954e-05, "loss": 2.8099, "step": 36035 }, { "epoch": 2.83, "learning_rate": 7.174311926605505e-05, "loss": 2.9401, "step": 36036 }, { "epoch": 2.83, "learning_rate": 7.174233513683055e-05, "loss": 2.9271, "step": 36037 }, { "epoch": 2.83, "learning_rate": 7.174155100760605e-05, "loss": 2.8166, "step": 36038 }, { "epoch": 2.83, "learning_rate": 7.174076687838156e-05, "loss": 2.8783, "step": 36039 }, { "epoch": 2.83, "learning_rate": 7.173998274915706e-05, "loss": 2.7239, "step": 36040 }, { "epoch": 2.83, "learning_rate": 7.173919861993257e-05, "loss": 3.067, "step": 36041 }, { "epoch": 2.83, "learning_rate": 7.173841449070807e-05, "loss": 2.7806, "step": 36042 }, { "epoch": 2.83, "learning_rate": 7.173763036148357e-05, "loss": 2.6664, "step": 36043 }, { "epoch": 2.83, "learning_rate": 7.173684623225908e-05, "loss": 2.7445, "step": 36044 }, { "epoch": 2.83, "learning_rate": 7.173606210303458e-05, "loss": 2.6037, "step": 36045 }, { "epoch": 2.83, "learning_rate": 7.173527797381008e-05, "loss": 2.5577, "step": 36046 }, { "epoch": 2.83, "learning_rate": 7.173449384458559e-05, "loss": 2.8145, "step": 36047 }, { "epoch": 2.83, "learning_rate": 7.173370971536109e-05, "loss": 2.7955, "step": 36048 }, { "epoch": 2.83, "learning_rate": 7.17329255861366e-05, "loss": 3.0094, "step": 36049 }, { "epoch": 2.83, "learning_rate": 7.17321414569121e-05, "loss": 2.7071, "step": 36050 }, { "epoch": 2.83, "learning_rate": 7.173135732768762e-05, "loss": 2.7314, "step": 36051 }, { "epoch": 2.83, "learning_rate": 7.173057319846312e-05, "loss": 2.8291, "step": 36052 }, { "epoch": 2.83, "learning_rate": 7.172978906923862e-05, "loss": 2.9214, "step": 36053 }, { "epoch": 2.83, "learning_rate": 7.172900494001413e-05, "loss": 2.8971, "step": 36054 }, { "epoch": 2.83, "learning_rate": 7.172822081078963e-05, "loss": 3.0615, "step": 36055 }, { "epoch": 2.83, "learning_rate": 7.172743668156513e-05, "loss": 2.8671, "step": 36056 }, { "epoch": 2.83, "learning_rate": 7.172665255234064e-05, "loss": 2.8818, "step": 36057 }, { "epoch": 2.83, "learning_rate": 7.172586842311614e-05, "loss": 2.7804, "step": 36058 }, { "epoch": 2.83, "learning_rate": 7.172508429389165e-05, "loss": 2.6109, "step": 36059 }, { "epoch": 2.83, "learning_rate": 7.172430016466714e-05, "loss": 2.6949, "step": 36060 }, { "epoch": 2.83, "learning_rate": 7.172351603544264e-05, "loss": 2.6708, "step": 36061 }, { "epoch": 2.83, "learning_rate": 7.172273190621814e-05, "loss": 2.7413, "step": 36062 }, { "epoch": 2.83, "learning_rate": 7.172194777699365e-05, "loss": 2.9296, "step": 36063 }, { "epoch": 2.83, "learning_rate": 7.172116364776915e-05, "loss": 2.7052, "step": 36064 }, { "epoch": 2.83, "learning_rate": 7.172037951854465e-05, "loss": 2.6947, "step": 36065 }, { "epoch": 2.83, "learning_rate": 7.171959538932016e-05, "loss": 2.7014, "step": 36066 }, { "epoch": 2.83, "learning_rate": 7.171881126009566e-05, "loss": 2.8547, "step": 36067 }, { "epoch": 2.83, "learning_rate": 7.171802713087116e-05, "loss": 2.9202, "step": 36068 }, { "epoch": 2.83, "learning_rate": 7.171724300164667e-05, "loss": 2.8377, "step": 36069 }, { "epoch": 2.83, "learning_rate": 7.171645887242217e-05, "loss": 2.7915, "step": 36070 }, { "epoch": 2.83, "learning_rate": 7.171567474319768e-05, "loss": 2.9317, "step": 36071 }, { "epoch": 2.83, "learning_rate": 7.171489061397318e-05, "loss": 2.8499, "step": 36072 }, { "epoch": 2.83, "learning_rate": 7.171410648474868e-05, "loss": 2.6323, "step": 36073 }, { "epoch": 2.83, "learning_rate": 7.171332235552419e-05, "loss": 2.8735, "step": 36074 }, { "epoch": 2.83, "learning_rate": 7.171253822629969e-05, "loss": 2.7238, "step": 36075 }, { "epoch": 2.83, "learning_rate": 7.17117540970752e-05, "loss": 2.722, "step": 36076 }, { "epoch": 2.83, "learning_rate": 7.171096996785071e-05, "loss": 2.7619, "step": 36077 }, { "epoch": 2.83, "learning_rate": 7.171018583862622e-05, "loss": 2.7252, "step": 36078 }, { "epoch": 2.83, "learning_rate": 7.170940170940172e-05, "loss": 2.7843, "step": 36079 }, { "epoch": 2.83, "learning_rate": 7.170861758017722e-05, "loss": 2.9483, "step": 36080 }, { "epoch": 2.83, "learning_rate": 7.170783345095273e-05, "loss": 2.851, "step": 36081 }, { "epoch": 2.83, "learning_rate": 7.170704932172823e-05, "loss": 2.7085, "step": 36082 }, { "epoch": 2.83, "learning_rate": 7.170626519250373e-05, "loss": 2.9725, "step": 36083 }, { "epoch": 2.83, "learning_rate": 7.170548106327924e-05, "loss": 2.8889, "step": 36084 }, { "epoch": 2.83, "learning_rate": 7.170469693405474e-05, "loss": 2.8805, "step": 36085 }, { "epoch": 2.83, "learning_rate": 7.170391280483024e-05, "loss": 3.0017, "step": 36086 }, { "epoch": 2.83, "learning_rate": 7.170312867560575e-05, "loss": 2.7463, "step": 36087 }, { "epoch": 2.83, "learning_rate": 7.170234454638125e-05, "loss": 2.9679, "step": 36088 }, { "epoch": 2.83, "learning_rate": 7.170156041715676e-05, "loss": 2.6485, "step": 36089 }, { "epoch": 2.83, "learning_rate": 7.170077628793226e-05, "loss": 2.7335, "step": 36090 }, { "epoch": 2.83, "learning_rate": 7.169999215870776e-05, "loss": 2.7331, "step": 36091 }, { "epoch": 2.83, "learning_rate": 7.169920802948327e-05, "loss": 2.7512, "step": 36092 }, { "epoch": 2.83, "learning_rate": 7.169842390025877e-05, "loss": 2.7162, "step": 36093 }, { "epoch": 2.83, "learning_rate": 7.169763977103427e-05, "loss": 2.8145, "step": 36094 }, { "epoch": 2.83, "learning_rate": 7.169685564180978e-05, "loss": 2.9691, "step": 36095 }, { "epoch": 2.83, "learning_rate": 7.169607151258527e-05, "loss": 2.6676, "step": 36096 }, { "epoch": 2.83, "learning_rate": 7.169528738336077e-05, "loss": 2.828, "step": 36097 }, { "epoch": 2.83, "learning_rate": 7.169450325413628e-05, "loss": 2.8702, "step": 36098 }, { "epoch": 2.83, "learning_rate": 7.169371912491178e-05, "loss": 2.7103, "step": 36099 }, { "epoch": 2.83, "learning_rate": 7.169293499568728e-05, "loss": 2.6724, "step": 36100 }, { "epoch": 2.83, "learning_rate": 7.169215086646279e-05, "loss": 2.5661, "step": 36101 }, { "epoch": 2.83, "learning_rate": 7.16913667372383e-05, "loss": 2.9091, "step": 36102 }, { "epoch": 2.83, "learning_rate": 7.169058260801381e-05, "loss": 2.7646, "step": 36103 }, { "epoch": 2.83, "learning_rate": 7.168979847878931e-05, "loss": 2.7553, "step": 36104 }, { "epoch": 2.83, "learning_rate": 7.168901434956481e-05, "loss": 2.6212, "step": 36105 }, { "epoch": 2.83, "learning_rate": 7.168823022034032e-05, "loss": 2.7287, "step": 36106 }, { "epoch": 2.83, "learning_rate": 7.168744609111582e-05, "loss": 2.7271, "step": 36107 }, { "epoch": 2.83, "learning_rate": 7.168666196189133e-05, "loss": 2.6413, "step": 36108 }, { "epoch": 2.83, "learning_rate": 7.168587783266683e-05, "loss": 3.0365, "step": 36109 }, { "epoch": 2.83, "learning_rate": 7.168509370344233e-05, "loss": 2.7086, "step": 36110 }, { "epoch": 2.83, "learning_rate": 7.168430957421784e-05, "loss": 2.6821, "step": 36111 }, { "epoch": 2.83, "learning_rate": 7.168352544499334e-05, "loss": 2.9664, "step": 36112 }, { "epoch": 2.83, "learning_rate": 7.168274131576884e-05, "loss": 2.5017, "step": 36113 }, { "epoch": 2.83, "learning_rate": 7.168195718654435e-05, "loss": 2.7508, "step": 36114 }, { "epoch": 2.83, "learning_rate": 7.168117305731985e-05, "loss": 2.665, "step": 36115 }, { "epoch": 2.83, "learning_rate": 7.168038892809536e-05, "loss": 2.7432, "step": 36116 }, { "epoch": 2.83, "learning_rate": 7.167960479887086e-05, "loss": 2.6666, "step": 36117 }, { "epoch": 2.83, "learning_rate": 7.167882066964636e-05, "loss": 2.7017, "step": 36118 }, { "epoch": 2.83, "learning_rate": 7.167803654042187e-05, "loss": 3.0166, "step": 36119 }, { "epoch": 2.83, "learning_rate": 7.167725241119737e-05, "loss": 2.4925, "step": 36120 }, { "epoch": 2.83, "learning_rate": 7.167646828197287e-05, "loss": 2.5437, "step": 36121 }, { "epoch": 2.83, "learning_rate": 7.167568415274838e-05, "loss": 2.7113, "step": 36122 }, { "epoch": 2.83, "learning_rate": 7.167490002352388e-05, "loss": 2.7688, "step": 36123 }, { "epoch": 2.83, "learning_rate": 7.167411589429938e-05, "loss": 2.669, "step": 36124 }, { "epoch": 2.83, "learning_rate": 7.167333176507489e-05, "loss": 2.8721, "step": 36125 }, { "epoch": 2.83, "learning_rate": 7.167254763585039e-05, "loss": 2.9964, "step": 36126 }, { "epoch": 2.83, "learning_rate": 7.16717635066259e-05, "loss": 2.78, "step": 36127 }, { "epoch": 2.83, "learning_rate": 7.16709793774014e-05, "loss": 2.6683, "step": 36128 }, { "epoch": 2.83, "learning_rate": 7.16701952481769e-05, "loss": 2.8512, "step": 36129 }, { "epoch": 2.83, "learning_rate": 7.16694111189524e-05, "loss": 3.0115, "step": 36130 }, { "epoch": 2.83, "learning_rate": 7.166862698972791e-05, "loss": 2.9497, "step": 36131 }, { "epoch": 2.83, "learning_rate": 7.166784286050341e-05, "loss": 2.7783, "step": 36132 }, { "epoch": 2.83, "learning_rate": 7.166705873127892e-05, "loss": 2.5986, "step": 36133 }, { "epoch": 2.83, "learning_rate": 7.166627460205442e-05, "loss": 2.9827, "step": 36134 }, { "epoch": 2.83, "learning_rate": 7.166549047282992e-05, "loss": 2.6684, "step": 36135 }, { "epoch": 2.83, "learning_rate": 7.166470634360543e-05, "loss": 2.7578, "step": 36136 }, { "epoch": 2.83, "learning_rate": 7.166392221438093e-05, "loss": 2.7721, "step": 36137 }, { "epoch": 2.83, "learning_rate": 7.166313808515644e-05, "loss": 2.9661, "step": 36138 }, { "epoch": 2.83, "learning_rate": 7.166235395593194e-05, "loss": 2.6874, "step": 36139 }, { "epoch": 2.83, "learning_rate": 7.166156982670744e-05, "loss": 2.813, "step": 36140 }, { "epoch": 2.83, "learning_rate": 7.166078569748295e-05, "loss": 2.9325, "step": 36141 }, { "epoch": 2.83, "learning_rate": 7.166000156825845e-05, "loss": 2.8527, "step": 36142 }, { "epoch": 2.83, "learning_rate": 7.165921743903395e-05, "loss": 2.8051, "step": 36143 }, { "epoch": 2.83, "learning_rate": 7.165843330980946e-05, "loss": 2.9656, "step": 36144 }, { "epoch": 2.83, "learning_rate": 7.165764918058496e-05, "loss": 2.872, "step": 36145 }, { "epoch": 2.83, "learning_rate": 7.165686505136047e-05, "loss": 3.1278, "step": 36146 }, { "epoch": 2.83, "learning_rate": 7.165608092213597e-05, "loss": 2.9278, "step": 36147 }, { "epoch": 2.83, "learning_rate": 7.165529679291147e-05, "loss": 3.0182, "step": 36148 }, { "epoch": 2.83, "learning_rate": 7.165451266368698e-05, "loss": 2.8999, "step": 36149 }, { "epoch": 2.83, "learning_rate": 7.165372853446248e-05, "loss": 2.6631, "step": 36150 }, { "epoch": 2.83, "learning_rate": 7.165294440523798e-05, "loss": 2.7947, "step": 36151 }, { "epoch": 2.83, "learning_rate": 7.165216027601349e-05, "loss": 3.0646, "step": 36152 }, { "epoch": 2.83, "learning_rate": 7.165137614678899e-05, "loss": 2.9762, "step": 36153 }, { "epoch": 2.83, "learning_rate": 7.16505920175645e-05, "loss": 2.8347, "step": 36154 }, { "epoch": 2.84, "learning_rate": 7.164980788834e-05, "loss": 2.9088, "step": 36155 }, { "epoch": 2.84, "learning_rate": 7.16490237591155e-05, "loss": 2.7004, "step": 36156 }, { "epoch": 2.84, "learning_rate": 7.1648239629891e-05, "loss": 3.0437, "step": 36157 }, { "epoch": 2.84, "learning_rate": 7.164745550066651e-05, "loss": 2.7237, "step": 36158 }, { "epoch": 2.84, "learning_rate": 7.164667137144203e-05, "loss": 2.9061, "step": 36159 }, { "epoch": 2.84, "learning_rate": 7.164588724221753e-05, "loss": 2.7931, "step": 36160 }, { "epoch": 2.84, "learning_rate": 7.164510311299303e-05, "loss": 2.684, "step": 36161 }, { "epoch": 2.84, "learning_rate": 7.164431898376854e-05, "loss": 2.8637, "step": 36162 }, { "epoch": 2.84, "learning_rate": 7.164353485454404e-05, "loss": 2.7906, "step": 36163 }, { "epoch": 2.84, "learning_rate": 7.164275072531955e-05, "loss": 2.5851, "step": 36164 }, { "epoch": 2.84, "learning_rate": 7.164196659609505e-05, "loss": 2.9288, "step": 36165 }, { "epoch": 2.84, "learning_rate": 7.164118246687055e-05, "loss": 2.8646, "step": 36166 }, { "epoch": 2.84, "learning_rate": 7.164039833764606e-05, "loss": 3.0599, "step": 36167 }, { "epoch": 2.84, "learning_rate": 7.163961420842155e-05, "loss": 2.7116, "step": 36168 }, { "epoch": 2.84, "learning_rate": 7.163883007919705e-05, "loss": 2.7164, "step": 36169 }, { "epoch": 2.84, "learning_rate": 7.163804594997255e-05, "loss": 2.6582, "step": 36170 }, { "epoch": 2.84, "learning_rate": 7.163726182074806e-05, "loss": 3.0352, "step": 36171 }, { "epoch": 2.84, "learning_rate": 7.163647769152356e-05, "loss": 3.1477, "step": 36172 }, { "epoch": 2.84, "learning_rate": 7.163569356229906e-05, "loss": 2.9439, "step": 36173 }, { "epoch": 2.84, "learning_rate": 7.163490943307457e-05, "loss": 2.7209, "step": 36174 }, { "epoch": 2.84, "learning_rate": 7.163412530385007e-05, "loss": 2.7115, "step": 36175 }, { "epoch": 2.84, "learning_rate": 7.163334117462558e-05, "loss": 2.7336, "step": 36176 }, { "epoch": 2.84, "learning_rate": 7.163255704540108e-05, "loss": 2.9136, "step": 36177 }, { "epoch": 2.84, "learning_rate": 7.163177291617658e-05, "loss": 2.8081, "step": 36178 }, { "epoch": 2.84, "learning_rate": 7.163098878695209e-05, "loss": 3.1104, "step": 36179 }, { "epoch": 2.84, "learning_rate": 7.163020465772759e-05, "loss": 2.8158, "step": 36180 }, { "epoch": 2.84, "learning_rate": 7.16294205285031e-05, "loss": 2.6439, "step": 36181 }, { "epoch": 2.84, "learning_rate": 7.16286363992786e-05, "loss": 2.7918, "step": 36182 }, { "epoch": 2.84, "learning_rate": 7.16278522700541e-05, "loss": 2.7257, "step": 36183 }, { "epoch": 2.84, "learning_rate": 7.16270681408296e-05, "loss": 2.8348, "step": 36184 }, { "epoch": 2.84, "learning_rate": 7.162628401160512e-05, "loss": 2.8707, "step": 36185 }, { "epoch": 2.84, "learning_rate": 7.162549988238063e-05, "loss": 2.878, "step": 36186 }, { "epoch": 2.84, "learning_rate": 7.162471575315613e-05, "loss": 3.0805, "step": 36187 }, { "epoch": 2.84, "learning_rate": 7.162393162393163e-05, "loss": 2.7734, "step": 36188 }, { "epoch": 2.84, "learning_rate": 7.162314749470714e-05, "loss": 2.877, "step": 36189 }, { "epoch": 2.84, "learning_rate": 7.162236336548264e-05, "loss": 2.718, "step": 36190 }, { "epoch": 2.84, "learning_rate": 7.162157923625814e-05, "loss": 2.9896, "step": 36191 }, { "epoch": 2.84, "learning_rate": 7.162079510703365e-05, "loss": 2.7537, "step": 36192 }, { "epoch": 2.84, "learning_rate": 7.162001097780915e-05, "loss": 2.8552, "step": 36193 }, { "epoch": 2.84, "learning_rate": 7.161922684858466e-05, "loss": 2.7014, "step": 36194 }, { "epoch": 2.84, "learning_rate": 7.161844271936016e-05, "loss": 2.8812, "step": 36195 }, { "epoch": 2.84, "learning_rate": 7.161765859013566e-05, "loss": 2.723, "step": 36196 }, { "epoch": 2.84, "learning_rate": 7.161687446091117e-05, "loss": 3.182, "step": 36197 }, { "epoch": 2.84, "learning_rate": 7.161609033168667e-05, "loss": 2.8281, "step": 36198 }, { "epoch": 2.84, "learning_rate": 7.161530620246217e-05, "loss": 2.9342, "step": 36199 }, { "epoch": 2.84, "learning_rate": 7.161452207323768e-05, "loss": 2.856, "step": 36200 }, { "epoch": 2.84, "learning_rate": 7.161373794401318e-05, "loss": 2.5798, "step": 36201 }, { "epoch": 2.84, "learning_rate": 7.161295381478868e-05, "loss": 3.0964, "step": 36202 }, { "epoch": 2.84, "learning_rate": 7.161216968556419e-05, "loss": 2.8261, "step": 36203 }, { "epoch": 2.84, "learning_rate": 7.161138555633968e-05, "loss": 2.6876, "step": 36204 }, { "epoch": 2.84, "learning_rate": 7.161060142711518e-05, "loss": 2.9919, "step": 36205 }, { "epoch": 2.84, "learning_rate": 7.160981729789069e-05, "loss": 2.6669, "step": 36206 }, { "epoch": 2.84, "learning_rate": 7.160903316866619e-05, "loss": 2.951, "step": 36207 }, { "epoch": 2.84, "learning_rate": 7.16082490394417e-05, "loss": 2.809, "step": 36208 }, { "epoch": 2.84, "learning_rate": 7.16074649102172e-05, "loss": 2.9557, "step": 36209 }, { "epoch": 2.84, "learning_rate": 7.160668078099271e-05, "loss": 2.7171, "step": 36210 }, { "epoch": 2.84, "learning_rate": 7.160589665176822e-05, "loss": 2.6695, "step": 36211 }, { "epoch": 2.84, "learning_rate": 7.160511252254372e-05, "loss": 2.7715, "step": 36212 }, { "epoch": 2.84, "learning_rate": 7.160432839331923e-05, "loss": 2.8484, "step": 36213 }, { "epoch": 2.84, "learning_rate": 7.160354426409473e-05, "loss": 2.76, "step": 36214 }, { "epoch": 2.84, "learning_rate": 7.160276013487023e-05, "loss": 2.7488, "step": 36215 }, { "epoch": 2.84, "learning_rate": 7.160197600564574e-05, "loss": 2.6912, "step": 36216 }, { "epoch": 2.84, "learning_rate": 7.160119187642124e-05, "loss": 2.6448, "step": 36217 }, { "epoch": 2.84, "learning_rate": 7.160040774719674e-05, "loss": 2.6408, "step": 36218 }, { "epoch": 2.84, "learning_rate": 7.159962361797225e-05, "loss": 2.8006, "step": 36219 }, { "epoch": 2.84, "learning_rate": 7.159883948874775e-05, "loss": 3.0298, "step": 36220 }, { "epoch": 2.84, "learning_rate": 7.159805535952325e-05, "loss": 2.7829, "step": 36221 }, { "epoch": 2.84, "learning_rate": 7.159727123029876e-05, "loss": 2.4561, "step": 36222 }, { "epoch": 2.84, "learning_rate": 7.159648710107426e-05, "loss": 2.6589, "step": 36223 }, { "epoch": 2.84, "learning_rate": 7.159570297184977e-05, "loss": 2.74, "step": 36224 }, { "epoch": 2.84, "learning_rate": 7.159491884262527e-05, "loss": 2.7158, "step": 36225 }, { "epoch": 2.84, "learning_rate": 7.159413471340077e-05, "loss": 2.8193, "step": 36226 }, { "epoch": 2.84, "learning_rate": 7.159335058417628e-05, "loss": 2.7817, "step": 36227 }, { "epoch": 2.84, "learning_rate": 7.159256645495178e-05, "loss": 2.8122, "step": 36228 }, { "epoch": 2.84, "learning_rate": 7.159178232572728e-05, "loss": 2.7236, "step": 36229 }, { "epoch": 2.84, "learning_rate": 7.159099819650279e-05, "loss": 2.9357, "step": 36230 }, { "epoch": 2.84, "learning_rate": 7.159021406727829e-05, "loss": 2.7643, "step": 36231 }, { "epoch": 2.84, "learning_rate": 7.15894299380538e-05, "loss": 2.7693, "step": 36232 }, { "epoch": 2.84, "learning_rate": 7.15886458088293e-05, "loss": 2.8614, "step": 36233 }, { "epoch": 2.84, "learning_rate": 7.15878616796048e-05, "loss": 2.8496, "step": 36234 }, { "epoch": 2.84, "learning_rate": 7.15870775503803e-05, "loss": 2.7894, "step": 36235 }, { "epoch": 2.84, "learning_rate": 7.158629342115581e-05, "loss": 2.7973, "step": 36236 }, { "epoch": 2.84, "learning_rate": 7.158550929193131e-05, "loss": 2.836, "step": 36237 }, { "epoch": 2.84, "learning_rate": 7.158472516270682e-05, "loss": 2.6241, "step": 36238 }, { "epoch": 2.84, "learning_rate": 7.158394103348232e-05, "loss": 2.8461, "step": 36239 }, { "epoch": 2.84, "learning_rate": 7.158315690425782e-05, "loss": 2.8541, "step": 36240 }, { "epoch": 2.84, "learning_rate": 7.158237277503333e-05, "loss": 2.6795, "step": 36241 }, { "epoch": 2.84, "learning_rate": 7.158158864580883e-05, "loss": 2.6667, "step": 36242 }, { "epoch": 2.84, "learning_rate": 7.158080451658434e-05, "loss": 2.6725, "step": 36243 }, { "epoch": 2.84, "learning_rate": 7.158002038735984e-05, "loss": 2.6402, "step": 36244 }, { "epoch": 2.84, "learning_rate": 7.157923625813534e-05, "loss": 2.925, "step": 36245 }, { "epoch": 2.84, "learning_rate": 7.157845212891085e-05, "loss": 2.8691, "step": 36246 }, { "epoch": 2.84, "learning_rate": 7.157766799968635e-05, "loss": 2.8828, "step": 36247 }, { "epoch": 2.84, "learning_rate": 7.157688387046185e-05, "loss": 2.8649, "step": 36248 }, { "epoch": 2.84, "learning_rate": 7.157609974123736e-05, "loss": 2.7842, "step": 36249 }, { "epoch": 2.84, "learning_rate": 7.157531561201286e-05, "loss": 2.8375, "step": 36250 }, { "epoch": 2.84, "learning_rate": 7.157453148278837e-05, "loss": 2.8378, "step": 36251 }, { "epoch": 2.84, "learning_rate": 7.157374735356387e-05, "loss": 2.7423, "step": 36252 }, { "epoch": 2.84, "learning_rate": 7.157296322433937e-05, "loss": 2.6997, "step": 36253 }, { "epoch": 2.84, "learning_rate": 7.157217909511488e-05, "loss": 3.2203, "step": 36254 }, { "epoch": 2.84, "learning_rate": 7.157139496589038e-05, "loss": 2.9017, "step": 36255 }, { "epoch": 2.84, "learning_rate": 7.157061083666588e-05, "loss": 2.6642, "step": 36256 }, { "epoch": 2.84, "learning_rate": 7.156982670744139e-05, "loss": 2.7451, "step": 36257 }, { "epoch": 2.84, "learning_rate": 7.156904257821689e-05, "loss": 2.5366, "step": 36258 }, { "epoch": 2.84, "learning_rate": 7.15682584489924e-05, "loss": 2.5879, "step": 36259 }, { "epoch": 2.84, "learning_rate": 7.15674743197679e-05, "loss": 2.8957, "step": 36260 }, { "epoch": 2.84, "learning_rate": 7.15666901905434e-05, "loss": 2.7893, "step": 36261 }, { "epoch": 2.84, "learning_rate": 7.15659060613189e-05, "loss": 2.9581, "step": 36262 }, { "epoch": 2.84, "learning_rate": 7.156512193209441e-05, "loss": 2.8554, "step": 36263 }, { "epoch": 2.84, "learning_rate": 7.156433780286991e-05, "loss": 2.7287, "step": 36264 }, { "epoch": 2.84, "learning_rate": 7.156355367364542e-05, "loss": 3.0295, "step": 36265 }, { "epoch": 2.84, "learning_rate": 7.156276954442092e-05, "loss": 2.7122, "step": 36266 }, { "epoch": 2.84, "learning_rate": 7.156198541519644e-05, "loss": 2.7664, "step": 36267 }, { "epoch": 2.84, "learning_rate": 7.156120128597194e-05, "loss": 2.7461, "step": 36268 }, { "epoch": 2.84, "learning_rate": 7.156041715674744e-05, "loss": 2.9192, "step": 36269 }, { "epoch": 2.84, "learning_rate": 7.155963302752295e-05, "loss": 3.0795, "step": 36270 }, { "epoch": 2.84, "learning_rate": 7.155884889829845e-05, "loss": 2.7096, "step": 36271 }, { "epoch": 2.84, "learning_rate": 7.155806476907396e-05, "loss": 2.7944, "step": 36272 }, { "epoch": 2.84, "learning_rate": 7.155728063984946e-05, "loss": 2.7174, "step": 36273 }, { "epoch": 2.84, "learning_rate": 7.155649651062496e-05, "loss": 2.7986, "step": 36274 }, { "epoch": 2.84, "learning_rate": 7.155571238140047e-05, "loss": 3.0101, "step": 36275 }, { "epoch": 2.84, "learning_rate": 7.155492825217596e-05, "loss": 2.881, "step": 36276 }, { "epoch": 2.84, "learning_rate": 7.155414412295146e-05, "loss": 2.8462, "step": 36277 }, { "epoch": 2.84, "learning_rate": 7.155335999372696e-05, "loss": 2.789, "step": 36278 }, { "epoch": 2.84, "learning_rate": 7.155257586450247e-05, "loss": 2.7642, "step": 36279 }, { "epoch": 2.84, "learning_rate": 7.155179173527797e-05, "loss": 2.6525, "step": 36280 }, { "epoch": 2.84, "learning_rate": 7.155100760605348e-05, "loss": 2.7394, "step": 36281 }, { "epoch": 2.84, "learning_rate": 7.155022347682898e-05, "loss": 2.8948, "step": 36282 }, { "epoch": 2.85, "learning_rate": 7.154943934760448e-05, "loss": 2.7539, "step": 36283 }, { "epoch": 2.85, "learning_rate": 7.154865521837999e-05, "loss": 2.9694, "step": 36284 }, { "epoch": 2.85, "learning_rate": 7.154787108915549e-05, "loss": 2.8595, "step": 36285 }, { "epoch": 2.85, "learning_rate": 7.1547086959931e-05, "loss": 2.8806, "step": 36286 }, { "epoch": 2.85, "learning_rate": 7.15463028307065e-05, "loss": 2.6457, "step": 36287 }, { "epoch": 2.85, "learning_rate": 7.1545518701482e-05, "loss": 2.7128, "step": 36288 }, { "epoch": 2.85, "learning_rate": 7.15447345722575e-05, "loss": 2.6512, "step": 36289 }, { "epoch": 2.85, "learning_rate": 7.154395044303301e-05, "loss": 2.827, "step": 36290 }, { "epoch": 2.85, "learning_rate": 7.154316631380851e-05, "loss": 2.7471, "step": 36291 }, { "epoch": 2.85, "learning_rate": 7.154238218458402e-05, "loss": 2.9102, "step": 36292 }, { "epoch": 2.85, "learning_rate": 7.154159805535953e-05, "loss": 2.8784, "step": 36293 }, { "epoch": 2.85, "learning_rate": 7.154081392613504e-05, "loss": 2.6421, "step": 36294 }, { "epoch": 2.85, "learning_rate": 7.154002979691054e-05, "loss": 2.8324, "step": 36295 }, { "epoch": 2.85, "learning_rate": 7.153924566768604e-05, "loss": 2.7613, "step": 36296 }, { "epoch": 2.85, "learning_rate": 7.153846153846155e-05, "loss": 2.8294, "step": 36297 }, { "epoch": 2.85, "learning_rate": 7.153767740923705e-05, "loss": 2.6321, "step": 36298 }, { "epoch": 2.85, "learning_rate": 7.153689328001256e-05, "loss": 2.8433, "step": 36299 }, { "epoch": 2.85, "learning_rate": 7.153610915078806e-05, "loss": 2.7462, "step": 36300 }, { "epoch": 2.85, "learning_rate": 7.153532502156356e-05, "loss": 2.6399, "step": 36301 }, { "epoch": 2.85, "learning_rate": 7.153454089233907e-05, "loss": 2.7921, "step": 36302 }, { "epoch": 2.85, "learning_rate": 7.153375676311457e-05, "loss": 2.8294, "step": 36303 }, { "epoch": 2.85, "learning_rate": 7.153297263389007e-05, "loss": 3.0026, "step": 36304 }, { "epoch": 2.85, "learning_rate": 7.153218850466558e-05, "loss": 2.7308, "step": 36305 }, { "epoch": 2.85, "learning_rate": 7.153140437544108e-05, "loss": 3.0729, "step": 36306 }, { "epoch": 2.85, "learning_rate": 7.153062024621658e-05, "loss": 2.8373, "step": 36307 }, { "epoch": 2.85, "learning_rate": 7.152983611699209e-05, "loss": 2.9326, "step": 36308 }, { "epoch": 2.85, "learning_rate": 7.152905198776759e-05, "loss": 2.8079, "step": 36309 }, { "epoch": 2.85, "learning_rate": 7.15282678585431e-05, "loss": 2.8236, "step": 36310 }, { "epoch": 2.85, "learning_rate": 7.15274837293186e-05, "loss": 2.8327, "step": 36311 }, { "epoch": 2.85, "learning_rate": 7.152669960009409e-05, "loss": 2.7186, "step": 36312 }, { "epoch": 2.85, "learning_rate": 7.152591547086959e-05, "loss": 2.7977, "step": 36313 }, { "epoch": 2.85, "learning_rate": 7.15251313416451e-05, "loss": 2.7467, "step": 36314 }, { "epoch": 2.85, "learning_rate": 7.15243472124206e-05, "loss": 2.571, "step": 36315 }, { "epoch": 2.85, "learning_rate": 7.15235630831961e-05, "loss": 2.8779, "step": 36316 }, { "epoch": 2.85, "learning_rate": 7.152277895397161e-05, "loss": 2.7772, "step": 36317 }, { "epoch": 2.85, "learning_rate": 7.152199482474713e-05, "loss": 2.9517, "step": 36318 }, { "epoch": 2.85, "learning_rate": 7.152121069552263e-05, "loss": 2.7835, "step": 36319 }, { "epoch": 2.85, "learning_rate": 7.152042656629813e-05, "loss": 2.6509, "step": 36320 }, { "epoch": 2.85, "learning_rate": 7.151964243707364e-05, "loss": 2.8262, "step": 36321 }, { "epoch": 2.85, "learning_rate": 7.151885830784914e-05, "loss": 2.7342, "step": 36322 }, { "epoch": 2.85, "learning_rate": 7.151807417862464e-05, "loss": 2.803, "step": 36323 }, { "epoch": 2.85, "learning_rate": 7.151729004940015e-05, "loss": 2.7818, "step": 36324 }, { "epoch": 2.85, "learning_rate": 7.151650592017565e-05, "loss": 2.7749, "step": 36325 }, { "epoch": 2.85, "learning_rate": 7.151572179095115e-05, "loss": 2.9123, "step": 36326 }, { "epoch": 2.85, "learning_rate": 7.151493766172666e-05, "loss": 2.9143, "step": 36327 }, { "epoch": 2.85, "learning_rate": 7.151415353250216e-05, "loss": 2.7856, "step": 36328 }, { "epoch": 2.85, "learning_rate": 7.151336940327767e-05, "loss": 2.7873, "step": 36329 }, { "epoch": 2.85, "learning_rate": 7.151258527405317e-05, "loss": 2.7009, "step": 36330 }, { "epoch": 2.85, "learning_rate": 7.151180114482867e-05, "loss": 2.8453, "step": 36331 }, { "epoch": 2.85, "learning_rate": 7.151101701560418e-05, "loss": 2.9063, "step": 36332 }, { "epoch": 2.85, "learning_rate": 7.151023288637968e-05, "loss": 2.8742, "step": 36333 }, { "epoch": 2.85, "learning_rate": 7.150944875715518e-05, "loss": 2.7926, "step": 36334 }, { "epoch": 2.85, "learning_rate": 7.150866462793069e-05, "loss": 2.8895, "step": 36335 }, { "epoch": 2.85, "learning_rate": 7.150788049870619e-05, "loss": 2.7087, "step": 36336 }, { "epoch": 2.85, "learning_rate": 7.15070963694817e-05, "loss": 2.6317, "step": 36337 }, { "epoch": 2.85, "learning_rate": 7.15063122402572e-05, "loss": 3.0397, "step": 36338 }, { "epoch": 2.85, "learning_rate": 7.15055281110327e-05, "loss": 2.7186, "step": 36339 }, { "epoch": 2.85, "learning_rate": 7.15047439818082e-05, "loss": 2.6205, "step": 36340 }, { "epoch": 2.85, "learning_rate": 7.150395985258371e-05, "loss": 2.8501, "step": 36341 }, { "epoch": 2.85, "learning_rate": 7.150317572335921e-05, "loss": 2.8085, "step": 36342 }, { "epoch": 2.85, "learning_rate": 7.150239159413472e-05, "loss": 2.7512, "step": 36343 }, { "epoch": 2.85, "learning_rate": 7.150160746491022e-05, "loss": 2.6584, "step": 36344 }, { "epoch": 2.85, "learning_rate": 7.150082333568572e-05, "loss": 2.4829, "step": 36345 }, { "epoch": 2.85, "learning_rate": 7.150003920646123e-05, "loss": 2.9312, "step": 36346 }, { "epoch": 2.85, "learning_rate": 7.149925507723673e-05, "loss": 2.7206, "step": 36347 }, { "epoch": 2.85, "learning_rate": 7.149847094801224e-05, "loss": 2.6245, "step": 36348 }, { "epoch": 2.85, "learning_rate": 7.149768681878774e-05, "loss": 2.5771, "step": 36349 }, { "epoch": 2.85, "learning_rate": 7.149690268956324e-05, "loss": 2.7989, "step": 36350 }, { "epoch": 2.85, "learning_rate": 7.149611856033875e-05, "loss": 2.927, "step": 36351 }, { "epoch": 2.85, "learning_rate": 7.149533443111425e-05, "loss": 2.7363, "step": 36352 }, { "epoch": 2.85, "learning_rate": 7.149455030188975e-05, "loss": 2.4971, "step": 36353 }, { "epoch": 2.85, "learning_rate": 7.149376617266526e-05, "loss": 2.8842, "step": 36354 }, { "epoch": 2.85, "learning_rate": 7.149298204344076e-05, "loss": 2.7674, "step": 36355 }, { "epoch": 2.85, "learning_rate": 7.149219791421626e-05, "loss": 3.0404, "step": 36356 }, { "epoch": 2.85, "learning_rate": 7.149141378499177e-05, "loss": 2.7627, "step": 36357 }, { "epoch": 2.85, "learning_rate": 7.149062965576727e-05, "loss": 2.8861, "step": 36358 }, { "epoch": 2.85, "learning_rate": 7.148984552654278e-05, "loss": 2.9336, "step": 36359 }, { "epoch": 2.85, "learning_rate": 7.148906139731828e-05, "loss": 2.7471, "step": 36360 }, { "epoch": 2.85, "learning_rate": 7.148827726809378e-05, "loss": 2.752, "step": 36361 }, { "epoch": 2.85, "learning_rate": 7.148749313886929e-05, "loss": 2.7139, "step": 36362 }, { "epoch": 2.85, "learning_rate": 7.148670900964479e-05, "loss": 2.7348, "step": 36363 }, { "epoch": 2.85, "learning_rate": 7.14859248804203e-05, "loss": 2.829, "step": 36364 }, { "epoch": 2.85, "learning_rate": 7.14851407511958e-05, "loss": 2.7955, "step": 36365 }, { "epoch": 2.85, "learning_rate": 7.14843566219713e-05, "loss": 2.6595, "step": 36366 }, { "epoch": 2.85, "learning_rate": 7.14835724927468e-05, "loss": 2.9293, "step": 36367 }, { "epoch": 2.85, "learning_rate": 7.148278836352231e-05, "loss": 2.7081, "step": 36368 }, { "epoch": 2.85, "learning_rate": 7.148200423429781e-05, "loss": 2.5947, "step": 36369 }, { "epoch": 2.85, "learning_rate": 7.148122010507332e-05, "loss": 2.8042, "step": 36370 }, { "epoch": 2.85, "learning_rate": 7.148043597584882e-05, "loss": 2.6611, "step": 36371 }, { "epoch": 2.85, "learning_rate": 7.147965184662432e-05, "loss": 2.8626, "step": 36372 }, { "epoch": 2.85, "learning_rate": 7.147886771739983e-05, "loss": 2.8295, "step": 36373 }, { "epoch": 2.85, "learning_rate": 7.147808358817533e-05, "loss": 2.8717, "step": 36374 }, { "epoch": 2.85, "learning_rate": 7.147729945895085e-05, "loss": 2.779, "step": 36375 }, { "epoch": 2.85, "learning_rate": 7.147651532972635e-05, "loss": 2.9142, "step": 36376 }, { "epoch": 2.85, "learning_rate": 7.147573120050186e-05, "loss": 2.6136, "step": 36377 }, { "epoch": 2.85, "learning_rate": 7.147494707127736e-05, "loss": 2.8132, "step": 36378 }, { "epoch": 2.85, "learning_rate": 7.147416294205286e-05, "loss": 2.7959, "step": 36379 }, { "epoch": 2.85, "learning_rate": 7.147337881282837e-05, "loss": 2.818, "step": 36380 }, { "epoch": 2.85, "learning_rate": 7.147259468360387e-05, "loss": 2.966, "step": 36381 }, { "epoch": 2.85, "learning_rate": 7.147181055437937e-05, "loss": 2.7734, "step": 36382 }, { "epoch": 2.85, "learning_rate": 7.147102642515488e-05, "loss": 3.0489, "step": 36383 }, { "epoch": 2.85, "learning_rate": 7.147024229593037e-05, "loss": 2.7478, "step": 36384 }, { "epoch": 2.85, "learning_rate": 7.146945816670587e-05, "loss": 2.7085, "step": 36385 }, { "epoch": 2.85, "learning_rate": 7.146867403748138e-05, "loss": 2.8067, "step": 36386 }, { "epoch": 2.85, "learning_rate": 7.146788990825688e-05, "loss": 2.7786, "step": 36387 }, { "epoch": 2.85, "learning_rate": 7.146710577903238e-05, "loss": 2.7891, "step": 36388 }, { "epoch": 2.85, "learning_rate": 7.146632164980789e-05, "loss": 2.8089, "step": 36389 }, { "epoch": 2.85, "learning_rate": 7.146553752058339e-05, "loss": 2.6663, "step": 36390 }, { "epoch": 2.85, "learning_rate": 7.14647533913589e-05, "loss": 2.9555, "step": 36391 }, { "epoch": 2.85, "learning_rate": 7.14639692621344e-05, "loss": 2.7727, "step": 36392 }, { "epoch": 2.85, "learning_rate": 7.14631851329099e-05, "loss": 2.8822, "step": 36393 }, { "epoch": 2.85, "learning_rate": 7.14624010036854e-05, "loss": 2.68, "step": 36394 }, { "epoch": 2.85, "learning_rate": 7.146161687446091e-05, "loss": 3.2914, "step": 36395 }, { "epoch": 2.85, "learning_rate": 7.146083274523641e-05, "loss": 2.931, "step": 36396 }, { "epoch": 2.85, "learning_rate": 7.146004861601192e-05, "loss": 2.7572, "step": 36397 }, { "epoch": 2.85, "learning_rate": 7.145926448678742e-05, "loss": 2.7513, "step": 36398 }, { "epoch": 2.85, "learning_rate": 7.145848035756292e-05, "loss": 2.8701, "step": 36399 }, { "epoch": 2.85, "learning_rate": 7.145769622833843e-05, "loss": 2.894, "step": 36400 }, { "epoch": 2.85, "learning_rate": 7.145691209911394e-05, "loss": 3.1707, "step": 36401 }, { "epoch": 2.85, "learning_rate": 7.145612796988945e-05, "loss": 2.9414, "step": 36402 }, { "epoch": 2.85, "learning_rate": 7.145534384066495e-05, "loss": 2.9338, "step": 36403 }, { "epoch": 2.85, "learning_rate": 7.145455971144045e-05, "loss": 2.8106, "step": 36404 }, { "epoch": 2.85, "learning_rate": 7.145377558221596e-05, "loss": 2.6994, "step": 36405 }, { "epoch": 2.85, "learning_rate": 7.145299145299146e-05, "loss": 2.7473, "step": 36406 }, { "epoch": 2.85, "learning_rate": 7.145220732376697e-05, "loss": 2.7199, "step": 36407 }, { "epoch": 2.85, "learning_rate": 7.145142319454247e-05, "loss": 2.7042, "step": 36408 }, { "epoch": 2.85, "learning_rate": 7.145063906531797e-05, "loss": 2.7923, "step": 36409 }, { "epoch": 2.86, "learning_rate": 7.144985493609348e-05, "loss": 2.8556, "step": 36410 }, { "epoch": 2.86, "learning_rate": 7.144907080686898e-05, "loss": 2.7462, "step": 36411 }, { "epoch": 2.86, "learning_rate": 7.144828667764448e-05, "loss": 2.6968, "step": 36412 }, { "epoch": 2.86, "learning_rate": 7.144750254841999e-05, "loss": 2.8869, "step": 36413 }, { "epoch": 2.86, "learning_rate": 7.144671841919549e-05, "loss": 3.1228, "step": 36414 }, { "epoch": 2.86, "learning_rate": 7.1445934289971e-05, "loss": 2.6227, "step": 36415 }, { "epoch": 2.86, "learning_rate": 7.14451501607465e-05, "loss": 2.89, "step": 36416 }, { "epoch": 2.86, "learning_rate": 7.1444366031522e-05, "loss": 2.5492, "step": 36417 }, { "epoch": 2.86, "learning_rate": 7.14435819022975e-05, "loss": 2.8934, "step": 36418 }, { "epoch": 2.86, "learning_rate": 7.144279777307301e-05, "loss": 2.7668, "step": 36419 }, { "epoch": 2.86, "learning_rate": 7.14420136438485e-05, "loss": 2.8373, "step": 36420 }, { "epoch": 2.86, "learning_rate": 7.1441229514624e-05, "loss": 3.021, "step": 36421 }, { "epoch": 2.86, "learning_rate": 7.144044538539951e-05, "loss": 2.7436, "step": 36422 }, { "epoch": 2.86, "learning_rate": 7.143966125617501e-05, "loss": 2.7025, "step": 36423 }, { "epoch": 2.86, "learning_rate": 7.143887712695051e-05, "loss": 2.7393, "step": 36424 }, { "epoch": 2.86, "learning_rate": 7.143809299772602e-05, "loss": 2.7555, "step": 36425 }, { "epoch": 2.86, "learning_rate": 7.143730886850154e-05, "loss": 2.7524, "step": 36426 }, { "epoch": 2.86, "learning_rate": 7.143652473927704e-05, "loss": 2.8986, "step": 36427 }, { "epoch": 2.86, "learning_rate": 7.143574061005254e-05, "loss": 2.7618, "step": 36428 }, { "epoch": 2.86, "learning_rate": 7.143495648082805e-05, "loss": 2.9676, "step": 36429 }, { "epoch": 2.86, "learning_rate": 7.143417235160355e-05, "loss": 2.7889, "step": 36430 }, { "epoch": 2.86, "learning_rate": 7.143338822237905e-05, "loss": 2.7543, "step": 36431 }, { "epoch": 2.86, "learning_rate": 7.143260409315456e-05, "loss": 2.9509, "step": 36432 }, { "epoch": 2.86, "learning_rate": 7.143181996393006e-05, "loss": 2.9445, "step": 36433 }, { "epoch": 2.86, "learning_rate": 7.143103583470557e-05, "loss": 2.8974, "step": 36434 }, { "epoch": 2.86, "learning_rate": 7.143025170548107e-05, "loss": 2.8368, "step": 36435 }, { "epoch": 2.86, "learning_rate": 7.142946757625657e-05, "loss": 2.8534, "step": 36436 }, { "epoch": 2.86, "learning_rate": 7.142868344703208e-05, "loss": 2.9847, "step": 36437 }, { "epoch": 2.86, "learning_rate": 7.142789931780758e-05, "loss": 2.9155, "step": 36438 }, { "epoch": 2.86, "learning_rate": 7.142711518858308e-05, "loss": 2.8069, "step": 36439 }, { "epoch": 2.86, "learning_rate": 7.142633105935859e-05, "loss": 2.8433, "step": 36440 }, { "epoch": 2.86, "learning_rate": 7.142554693013409e-05, "loss": 2.5995, "step": 36441 }, { "epoch": 2.86, "learning_rate": 7.14247628009096e-05, "loss": 2.8818, "step": 36442 }, { "epoch": 2.86, "learning_rate": 7.14239786716851e-05, "loss": 2.8803, "step": 36443 }, { "epoch": 2.86, "learning_rate": 7.14231945424606e-05, "loss": 3.0802, "step": 36444 }, { "epoch": 2.86, "learning_rate": 7.14224104132361e-05, "loss": 2.794, "step": 36445 }, { "epoch": 2.86, "learning_rate": 7.142162628401161e-05, "loss": 2.8956, "step": 36446 }, { "epoch": 2.86, "learning_rate": 7.142084215478711e-05, "loss": 2.7207, "step": 36447 }, { "epoch": 2.86, "learning_rate": 7.142005802556262e-05, "loss": 3.075, "step": 36448 }, { "epoch": 2.86, "learning_rate": 7.141927389633812e-05, "loss": 2.919, "step": 36449 }, { "epoch": 2.86, "learning_rate": 7.141848976711362e-05, "loss": 2.7006, "step": 36450 }, { "epoch": 2.86, "learning_rate": 7.141770563788913e-05, "loss": 2.7774, "step": 36451 }, { "epoch": 2.86, "learning_rate": 7.141692150866463e-05, "loss": 2.7961, "step": 36452 }, { "epoch": 2.86, "learning_rate": 7.141613737944014e-05, "loss": 2.7006, "step": 36453 }, { "epoch": 2.86, "learning_rate": 7.141535325021564e-05, "loss": 2.9415, "step": 36454 }, { "epoch": 2.86, "learning_rate": 7.141456912099114e-05, "loss": 2.9468, "step": 36455 }, { "epoch": 2.86, "learning_rate": 7.141378499176665e-05, "loss": 2.782, "step": 36456 }, { "epoch": 2.86, "learning_rate": 7.141300086254215e-05, "loss": 2.8496, "step": 36457 }, { "epoch": 2.86, "learning_rate": 7.141221673331765e-05, "loss": 2.7339, "step": 36458 }, { "epoch": 2.86, "learning_rate": 7.141143260409316e-05, "loss": 2.8825, "step": 36459 }, { "epoch": 2.86, "learning_rate": 7.141064847486866e-05, "loss": 2.8029, "step": 36460 }, { "epoch": 2.86, "learning_rate": 7.140986434564416e-05, "loss": 2.8248, "step": 36461 }, { "epoch": 2.86, "learning_rate": 7.140908021641967e-05, "loss": 2.9708, "step": 36462 }, { "epoch": 2.86, "learning_rate": 7.140829608719517e-05, "loss": 2.9416, "step": 36463 }, { "epoch": 2.86, "learning_rate": 7.140751195797068e-05, "loss": 2.6829, "step": 36464 }, { "epoch": 2.86, "learning_rate": 7.140672782874618e-05, "loss": 2.5861, "step": 36465 }, { "epoch": 2.86, "learning_rate": 7.140594369952168e-05, "loss": 3.1318, "step": 36466 }, { "epoch": 2.86, "learning_rate": 7.140515957029719e-05, "loss": 2.9271, "step": 36467 }, { "epoch": 2.86, "learning_rate": 7.140437544107269e-05, "loss": 2.8628, "step": 36468 }, { "epoch": 2.86, "learning_rate": 7.14035913118482e-05, "loss": 2.6337, "step": 36469 }, { "epoch": 2.86, "learning_rate": 7.14028071826237e-05, "loss": 2.7784, "step": 36470 }, { "epoch": 2.86, "learning_rate": 7.14020230533992e-05, "loss": 2.7045, "step": 36471 }, { "epoch": 2.86, "learning_rate": 7.14012389241747e-05, "loss": 2.9362, "step": 36472 }, { "epoch": 2.86, "learning_rate": 7.140045479495021e-05, "loss": 2.7878, "step": 36473 }, { "epoch": 2.86, "learning_rate": 7.139967066572571e-05, "loss": 2.813, "step": 36474 }, { "epoch": 2.86, "learning_rate": 7.139888653650122e-05, "loss": 2.8629, "step": 36475 }, { "epoch": 2.86, "learning_rate": 7.139810240727672e-05, "loss": 2.7947, "step": 36476 }, { "epoch": 2.86, "learning_rate": 7.139731827805222e-05, "loss": 2.8172, "step": 36477 }, { "epoch": 2.86, "learning_rate": 7.139653414882773e-05, "loss": 3.0262, "step": 36478 }, { "epoch": 2.86, "learning_rate": 7.139575001960323e-05, "loss": 3.0991, "step": 36479 }, { "epoch": 2.86, "learning_rate": 7.139496589037873e-05, "loss": 2.8808, "step": 36480 }, { "epoch": 2.86, "learning_rate": 7.139418176115424e-05, "loss": 2.7459, "step": 36481 }, { "epoch": 2.86, "learning_rate": 7.139339763192974e-05, "loss": 2.6814, "step": 36482 }, { "epoch": 2.86, "learning_rate": 7.139261350270526e-05, "loss": 2.6292, "step": 36483 }, { "epoch": 2.86, "learning_rate": 7.139182937348076e-05, "loss": 3.0112, "step": 36484 }, { "epoch": 2.86, "learning_rate": 7.139104524425627e-05, "loss": 3.0979, "step": 36485 }, { "epoch": 2.86, "learning_rate": 7.139026111503177e-05, "loss": 2.7217, "step": 36486 }, { "epoch": 2.86, "learning_rate": 7.138947698580727e-05, "loss": 2.7851, "step": 36487 }, { "epoch": 2.86, "learning_rate": 7.138869285658278e-05, "loss": 2.7671, "step": 36488 }, { "epoch": 2.86, "learning_rate": 7.138790872735828e-05, "loss": 2.876, "step": 36489 }, { "epoch": 2.86, "learning_rate": 7.138712459813378e-05, "loss": 2.8551, "step": 36490 }, { "epoch": 2.86, "learning_rate": 7.138634046890929e-05, "loss": 2.6657, "step": 36491 }, { "epoch": 2.86, "learning_rate": 7.138555633968478e-05, "loss": 2.8126, "step": 36492 }, { "epoch": 2.86, "learning_rate": 7.138477221046028e-05, "loss": 2.8834, "step": 36493 }, { "epoch": 2.86, "learning_rate": 7.138398808123579e-05, "loss": 2.828, "step": 36494 }, { "epoch": 2.86, "learning_rate": 7.138320395201129e-05, "loss": 2.9069, "step": 36495 }, { "epoch": 2.86, "learning_rate": 7.13824198227868e-05, "loss": 2.978, "step": 36496 }, { "epoch": 2.86, "learning_rate": 7.13816356935623e-05, "loss": 3.0878, "step": 36497 }, { "epoch": 2.86, "learning_rate": 7.13808515643378e-05, "loss": 2.831, "step": 36498 }, { "epoch": 2.86, "learning_rate": 7.13800674351133e-05, "loss": 2.9056, "step": 36499 }, { "epoch": 2.86, "learning_rate": 7.137928330588881e-05, "loss": 2.731, "step": 36500 }, { "epoch": 2.86, "learning_rate": 7.137849917666431e-05, "loss": 2.7227, "step": 36501 }, { "epoch": 2.86, "learning_rate": 7.137771504743982e-05, "loss": 2.7296, "step": 36502 }, { "epoch": 2.86, "learning_rate": 7.137693091821532e-05, "loss": 2.8721, "step": 36503 }, { "epoch": 2.86, "learning_rate": 7.137614678899082e-05, "loss": 2.9558, "step": 36504 }, { "epoch": 2.86, "learning_rate": 7.137536265976633e-05, "loss": 2.5994, "step": 36505 }, { "epoch": 2.86, "learning_rate": 7.137457853054183e-05, "loss": 2.8831, "step": 36506 }, { "epoch": 2.86, "learning_rate": 7.137379440131733e-05, "loss": 2.9005, "step": 36507 }, { "epoch": 2.86, "learning_rate": 7.137301027209284e-05, "loss": 2.8158, "step": 36508 }, { "epoch": 2.86, "learning_rate": 7.137222614286835e-05, "loss": 2.6963, "step": 36509 }, { "epoch": 2.86, "learning_rate": 7.137144201364386e-05, "loss": 2.9333, "step": 36510 }, { "epoch": 2.86, "learning_rate": 7.137065788441936e-05, "loss": 2.9593, "step": 36511 }, { "epoch": 2.86, "learning_rate": 7.136987375519487e-05, "loss": 2.9381, "step": 36512 }, { "epoch": 2.86, "learning_rate": 7.136908962597037e-05, "loss": 2.6333, "step": 36513 }, { "epoch": 2.86, "learning_rate": 7.136830549674587e-05, "loss": 2.6975, "step": 36514 }, { "epoch": 2.86, "learning_rate": 7.136752136752138e-05, "loss": 2.6457, "step": 36515 }, { "epoch": 2.86, "learning_rate": 7.136673723829688e-05, "loss": 3.0756, "step": 36516 }, { "epoch": 2.86, "learning_rate": 7.136595310907238e-05, "loss": 2.6695, "step": 36517 }, { "epoch": 2.86, "learning_rate": 7.136516897984789e-05, "loss": 2.7046, "step": 36518 }, { "epoch": 2.86, "learning_rate": 7.136438485062339e-05, "loss": 2.7248, "step": 36519 }, { "epoch": 2.86, "learning_rate": 7.13636007213989e-05, "loss": 2.7185, "step": 36520 }, { "epoch": 2.86, "learning_rate": 7.13628165921744e-05, "loss": 2.8634, "step": 36521 }, { "epoch": 2.86, "learning_rate": 7.13620324629499e-05, "loss": 3.0901, "step": 36522 }, { "epoch": 2.86, "learning_rate": 7.13612483337254e-05, "loss": 2.8337, "step": 36523 }, { "epoch": 2.86, "learning_rate": 7.136046420450091e-05, "loss": 2.8803, "step": 36524 }, { "epoch": 2.86, "learning_rate": 7.135968007527641e-05, "loss": 2.7765, "step": 36525 }, { "epoch": 2.86, "learning_rate": 7.135889594605192e-05, "loss": 2.4792, "step": 36526 }, { "epoch": 2.86, "learning_rate": 7.135811181682742e-05, "loss": 2.7807, "step": 36527 }, { "epoch": 2.86, "learning_rate": 7.135732768760291e-05, "loss": 2.8101, "step": 36528 }, { "epoch": 2.86, "learning_rate": 7.135654355837841e-05, "loss": 2.7614, "step": 36529 }, { "epoch": 2.86, "learning_rate": 7.135575942915392e-05, "loss": 2.782, "step": 36530 }, { "epoch": 2.86, "learning_rate": 7.135497529992942e-05, "loss": 2.8222, "step": 36531 }, { "epoch": 2.86, "learning_rate": 7.135419117070493e-05, "loss": 2.6617, "step": 36532 }, { "epoch": 2.86, "learning_rate": 7.135340704148043e-05, "loss": 2.752, "step": 36533 }, { "epoch": 2.86, "learning_rate": 7.135262291225593e-05, "loss": 2.9402, "step": 36534 }, { "epoch": 2.86, "learning_rate": 7.135183878303145e-05, "loss": 2.8095, "step": 36535 }, { "epoch": 2.86, "learning_rate": 7.135105465380695e-05, "loss": 2.9117, "step": 36536 }, { "epoch": 2.86, "learning_rate": 7.135027052458246e-05, "loss": 2.7597, "step": 36537 }, { "epoch": 2.87, "learning_rate": 7.134948639535796e-05, "loss": 2.84, "step": 36538 }, { "epoch": 2.87, "learning_rate": 7.134870226613346e-05, "loss": 2.599, "step": 36539 }, { "epoch": 2.87, "learning_rate": 7.134791813690897e-05, "loss": 2.8373, "step": 36540 }, { "epoch": 2.87, "learning_rate": 7.134713400768447e-05, "loss": 2.8862, "step": 36541 }, { "epoch": 2.87, "learning_rate": 7.134634987845998e-05, "loss": 2.7076, "step": 36542 }, { "epoch": 2.87, "learning_rate": 7.134556574923548e-05, "loss": 2.8113, "step": 36543 }, { "epoch": 2.87, "learning_rate": 7.134478162001098e-05, "loss": 2.938, "step": 36544 }, { "epoch": 2.87, "learning_rate": 7.134399749078649e-05, "loss": 2.708, "step": 36545 }, { "epoch": 2.87, "learning_rate": 7.134321336156199e-05, "loss": 3.1099, "step": 36546 }, { "epoch": 2.87, "learning_rate": 7.13424292323375e-05, "loss": 2.5948, "step": 36547 }, { "epoch": 2.87, "learning_rate": 7.1341645103113e-05, "loss": 2.9331, "step": 36548 }, { "epoch": 2.87, "learning_rate": 7.13408609738885e-05, "loss": 3.0101, "step": 36549 }, { "epoch": 2.87, "learning_rate": 7.1340076844664e-05, "loss": 2.6208, "step": 36550 }, { "epoch": 2.87, "learning_rate": 7.133929271543951e-05, "loss": 2.5335, "step": 36551 }, { "epoch": 2.87, "learning_rate": 7.133850858621501e-05, "loss": 2.8243, "step": 36552 }, { "epoch": 2.87, "learning_rate": 7.133772445699052e-05, "loss": 2.7659, "step": 36553 }, { "epoch": 2.87, "learning_rate": 7.133694032776602e-05, "loss": 2.735, "step": 36554 }, { "epoch": 2.87, "learning_rate": 7.133615619854152e-05, "loss": 2.7772, "step": 36555 }, { "epoch": 2.87, "learning_rate": 7.133537206931703e-05, "loss": 2.9892, "step": 36556 }, { "epoch": 2.87, "learning_rate": 7.133458794009253e-05, "loss": 3.0627, "step": 36557 }, { "epoch": 2.87, "learning_rate": 7.133380381086803e-05, "loss": 3.0299, "step": 36558 }, { "epoch": 2.87, "learning_rate": 7.133301968164354e-05, "loss": 2.9577, "step": 36559 }, { "epoch": 2.87, "learning_rate": 7.133223555241904e-05, "loss": 2.6912, "step": 36560 }, { "epoch": 2.87, "learning_rate": 7.133145142319455e-05, "loss": 2.7442, "step": 36561 }, { "epoch": 2.87, "learning_rate": 7.133066729397005e-05, "loss": 2.9394, "step": 36562 }, { "epoch": 2.87, "learning_rate": 7.132988316474555e-05, "loss": 2.9519, "step": 36563 }, { "epoch": 2.87, "learning_rate": 7.132909903552106e-05, "loss": 2.7222, "step": 36564 }, { "epoch": 2.87, "learning_rate": 7.132831490629656e-05, "loss": 3.0039, "step": 36565 }, { "epoch": 2.87, "learning_rate": 7.132753077707206e-05, "loss": 2.8642, "step": 36566 }, { "epoch": 2.87, "learning_rate": 7.132674664784757e-05, "loss": 2.7128, "step": 36567 }, { "epoch": 2.87, "learning_rate": 7.132596251862307e-05, "loss": 2.9432, "step": 36568 }, { "epoch": 2.87, "learning_rate": 7.132517838939858e-05, "loss": 2.7625, "step": 36569 }, { "epoch": 2.87, "learning_rate": 7.132439426017408e-05, "loss": 2.9724, "step": 36570 }, { "epoch": 2.87, "learning_rate": 7.132361013094958e-05, "loss": 2.8513, "step": 36571 }, { "epoch": 2.87, "learning_rate": 7.132282600172509e-05, "loss": 2.6906, "step": 36572 }, { "epoch": 2.87, "learning_rate": 7.132204187250059e-05, "loss": 2.816, "step": 36573 }, { "epoch": 2.87, "learning_rate": 7.13212577432761e-05, "loss": 3.0917, "step": 36574 }, { "epoch": 2.87, "learning_rate": 7.13204736140516e-05, "loss": 3.0094, "step": 36575 }, { "epoch": 2.87, "learning_rate": 7.13196894848271e-05, "loss": 2.8848, "step": 36576 }, { "epoch": 2.87, "learning_rate": 7.13189053556026e-05, "loss": 2.7605, "step": 36577 }, { "epoch": 2.87, "learning_rate": 7.131812122637811e-05, "loss": 2.7633, "step": 36578 }, { "epoch": 2.87, "learning_rate": 7.131733709715361e-05, "loss": 2.8847, "step": 36579 }, { "epoch": 2.87, "learning_rate": 7.131655296792912e-05, "loss": 2.7656, "step": 36580 }, { "epoch": 2.87, "learning_rate": 7.131576883870462e-05, "loss": 2.8578, "step": 36581 }, { "epoch": 2.87, "learning_rate": 7.131498470948012e-05, "loss": 2.983, "step": 36582 }, { "epoch": 2.87, "learning_rate": 7.131420058025563e-05, "loss": 2.5747, "step": 36583 }, { "epoch": 2.87, "learning_rate": 7.131341645103113e-05, "loss": 2.6987, "step": 36584 }, { "epoch": 2.87, "learning_rate": 7.131263232180663e-05, "loss": 2.9824, "step": 36585 }, { "epoch": 2.87, "learning_rate": 7.131184819258214e-05, "loss": 2.9139, "step": 36586 }, { "epoch": 2.87, "learning_rate": 7.131106406335764e-05, "loss": 2.8156, "step": 36587 }, { "epoch": 2.87, "learning_rate": 7.131027993413315e-05, "loss": 2.748, "step": 36588 }, { "epoch": 2.87, "learning_rate": 7.130949580490865e-05, "loss": 2.8915, "step": 36589 }, { "epoch": 2.87, "learning_rate": 7.130871167568415e-05, "loss": 2.58, "step": 36590 }, { "epoch": 2.87, "learning_rate": 7.130792754645967e-05, "loss": 2.8174, "step": 36591 }, { "epoch": 2.87, "learning_rate": 7.130714341723517e-05, "loss": 2.7157, "step": 36592 }, { "epoch": 2.87, "learning_rate": 7.130635928801068e-05, "loss": 2.8189, "step": 36593 }, { "epoch": 2.87, "learning_rate": 7.130557515878618e-05, "loss": 2.7422, "step": 36594 }, { "epoch": 2.87, "learning_rate": 7.130479102956168e-05, "loss": 2.8955, "step": 36595 }, { "epoch": 2.87, "learning_rate": 7.130400690033719e-05, "loss": 2.8731, "step": 36596 }, { "epoch": 2.87, "learning_rate": 7.130322277111269e-05, "loss": 2.8613, "step": 36597 }, { "epoch": 2.87, "learning_rate": 7.13024386418882e-05, "loss": 2.8117, "step": 36598 }, { "epoch": 2.87, "learning_rate": 7.13016545126637e-05, "loss": 2.74, "step": 36599 }, { "epoch": 2.87, "learning_rate": 7.130087038343919e-05, "loss": 2.7059, "step": 36600 }, { "epoch": 2.87, "learning_rate": 7.130008625421469e-05, "loss": 2.7003, "step": 36601 }, { "epoch": 2.87, "learning_rate": 7.12993021249902e-05, "loss": 2.7984, "step": 36602 }, { "epoch": 2.87, "learning_rate": 7.12985179957657e-05, "loss": 2.9286, "step": 36603 }, { "epoch": 2.87, "learning_rate": 7.12977338665412e-05, "loss": 2.8311, "step": 36604 }, { "epoch": 2.87, "learning_rate": 7.129694973731671e-05, "loss": 3.0804, "step": 36605 }, { "epoch": 2.87, "learning_rate": 7.129616560809221e-05, "loss": 2.7953, "step": 36606 }, { "epoch": 2.87, "learning_rate": 7.129538147886772e-05, "loss": 2.7058, "step": 36607 }, { "epoch": 2.87, "learning_rate": 7.129459734964322e-05, "loss": 2.8731, "step": 36608 }, { "epoch": 2.87, "learning_rate": 7.129381322041872e-05, "loss": 2.8907, "step": 36609 }, { "epoch": 2.87, "learning_rate": 7.129302909119423e-05, "loss": 2.7911, "step": 36610 }, { "epoch": 2.87, "learning_rate": 7.129224496196973e-05, "loss": 2.8409, "step": 36611 }, { "epoch": 2.87, "learning_rate": 7.129146083274523e-05, "loss": 2.6653, "step": 36612 }, { "epoch": 2.87, "learning_rate": 7.129067670352074e-05, "loss": 3.0433, "step": 36613 }, { "epoch": 2.87, "learning_rate": 7.128989257429624e-05, "loss": 2.9402, "step": 36614 }, { "epoch": 2.87, "learning_rate": 7.128910844507174e-05, "loss": 2.7637, "step": 36615 }, { "epoch": 2.87, "learning_rate": 7.128832431584725e-05, "loss": 2.7439, "step": 36616 }, { "epoch": 2.87, "learning_rate": 7.128754018662277e-05, "loss": 2.8544, "step": 36617 }, { "epoch": 2.87, "learning_rate": 7.128675605739827e-05, "loss": 2.8, "step": 36618 }, { "epoch": 2.87, "learning_rate": 7.128597192817377e-05, "loss": 2.8993, "step": 36619 }, { "epoch": 2.87, "learning_rate": 7.128518779894928e-05, "loss": 2.6338, "step": 36620 }, { "epoch": 2.87, "learning_rate": 7.128440366972478e-05, "loss": 2.7085, "step": 36621 }, { "epoch": 2.87, "learning_rate": 7.128361954050028e-05, "loss": 2.9691, "step": 36622 }, { "epoch": 2.87, "learning_rate": 7.128283541127579e-05, "loss": 2.9458, "step": 36623 }, { "epoch": 2.87, "learning_rate": 7.128205128205129e-05, "loss": 2.9596, "step": 36624 }, { "epoch": 2.87, "learning_rate": 7.12812671528268e-05, "loss": 2.7905, "step": 36625 }, { "epoch": 2.87, "learning_rate": 7.12804830236023e-05, "loss": 2.9125, "step": 36626 }, { "epoch": 2.87, "learning_rate": 7.12796988943778e-05, "loss": 2.9248, "step": 36627 }, { "epoch": 2.87, "learning_rate": 7.12789147651533e-05, "loss": 2.8205, "step": 36628 }, { "epoch": 2.87, "learning_rate": 7.127813063592881e-05, "loss": 2.8365, "step": 36629 }, { "epoch": 2.87, "learning_rate": 7.127734650670431e-05, "loss": 2.782, "step": 36630 }, { "epoch": 2.87, "learning_rate": 7.127656237747982e-05, "loss": 2.6718, "step": 36631 }, { "epoch": 2.87, "learning_rate": 7.127577824825532e-05, "loss": 2.7558, "step": 36632 }, { "epoch": 2.87, "learning_rate": 7.127499411903082e-05, "loss": 3.1692, "step": 36633 }, { "epoch": 2.87, "learning_rate": 7.127420998980633e-05, "loss": 3.0172, "step": 36634 }, { "epoch": 2.87, "learning_rate": 7.127342586058183e-05, "loss": 2.6879, "step": 36635 }, { "epoch": 2.87, "learning_rate": 7.127264173135732e-05, "loss": 2.7498, "step": 36636 }, { "epoch": 2.87, "learning_rate": 7.127185760213283e-05, "loss": 2.7318, "step": 36637 }, { "epoch": 2.87, "learning_rate": 7.127107347290833e-05, "loss": 2.9581, "step": 36638 }, { "epoch": 2.87, "learning_rate": 7.127028934368383e-05, "loss": 2.6282, "step": 36639 }, { "epoch": 2.87, "learning_rate": 7.126950521445934e-05, "loss": 2.6321, "step": 36640 }, { "epoch": 2.87, "learning_rate": 7.126872108523484e-05, "loss": 2.8949, "step": 36641 }, { "epoch": 2.87, "learning_rate": 7.126793695601034e-05, "loss": 2.8805, "step": 36642 }, { "epoch": 2.87, "learning_rate": 7.126715282678586e-05, "loss": 2.6302, "step": 36643 }, { "epoch": 2.87, "learning_rate": 7.126636869756136e-05, "loss": 2.6905, "step": 36644 }, { "epoch": 2.87, "learning_rate": 7.126558456833687e-05, "loss": 2.9601, "step": 36645 }, { "epoch": 2.87, "learning_rate": 7.126480043911237e-05, "loss": 2.9334, "step": 36646 }, { "epoch": 2.87, "learning_rate": 7.126401630988788e-05, "loss": 2.7193, "step": 36647 }, { "epoch": 2.87, "learning_rate": 7.126323218066338e-05, "loss": 2.7289, "step": 36648 }, { "epoch": 2.87, "learning_rate": 7.126244805143888e-05, "loss": 2.7912, "step": 36649 }, { "epoch": 2.87, "learning_rate": 7.126166392221439e-05, "loss": 2.857, "step": 36650 }, { "epoch": 2.87, "learning_rate": 7.126087979298989e-05, "loss": 2.6503, "step": 36651 }, { "epoch": 2.87, "learning_rate": 7.12600956637654e-05, "loss": 2.8648, "step": 36652 }, { "epoch": 2.87, "learning_rate": 7.12593115345409e-05, "loss": 2.6498, "step": 36653 }, { "epoch": 2.87, "learning_rate": 7.12585274053164e-05, "loss": 2.6933, "step": 36654 }, { "epoch": 2.87, "learning_rate": 7.12577432760919e-05, "loss": 2.7288, "step": 36655 }, { "epoch": 2.87, "learning_rate": 7.125695914686741e-05, "loss": 2.7267, "step": 36656 }, { "epoch": 2.87, "learning_rate": 7.125617501764291e-05, "loss": 2.8613, "step": 36657 }, { "epoch": 2.87, "learning_rate": 7.125539088841842e-05, "loss": 2.674, "step": 36658 }, { "epoch": 2.87, "learning_rate": 7.125460675919392e-05, "loss": 2.7449, "step": 36659 }, { "epoch": 2.87, "learning_rate": 7.125382262996942e-05, "loss": 2.8663, "step": 36660 }, { "epoch": 2.87, "learning_rate": 7.125303850074493e-05, "loss": 2.8914, "step": 36661 }, { "epoch": 2.87, "learning_rate": 7.125225437152043e-05, "loss": 2.7614, "step": 36662 }, { "epoch": 2.87, "learning_rate": 7.125147024229593e-05, "loss": 2.5859, "step": 36663 }, { "epoch": 2.87, "learning_rate": 7.125068611307144e-05, "loss": 2.8739, "step": 36664 }, { "epoch": 2.88, "learning_rate": 7.124990198384694e-05, "loss": 3.0395, "step": 36665 }, { "epoch": 2.88, "learning_rate": 7.124911785462245e-05, "loss": 2.6753, "step": 36666 }, { "epoch": 2.88, "learning_rate": 7.124833372539795e-05, "loss": 2.7894, "step": 36667 }, { "epoch": 2.88, "learning_rate": 7.124754959617345e-05, "loss": 2.8087, "step": 36668 }, { "epoch": 2.88, "learning_rate": 7.124676546694896e-05, "loss": 2.8772, "step": 36669 }, { "epoch": 2.88, "learning_rate": 7.124598133772446e-05, "loss": 2.6973, "step": 36670 }, { "epoch": 2.88, "learning_rate": 7.124519720849996e-05, "loss": 2.707, "step": 36671 }, { "epoch": 2.88, "learning_rate": 7.124441307927547e-05, "loss": 2.7763, "step": 36672 }, { "epoch": 2.88, "learning_rate": 7.124362895005097e-05, "loss": 2.7787, "step": 36673 }, { "epoch": 2.88, "learning_rate": 7.124284482082647e-05, "loss": 2.936, "step": 36674 }, { "epoch": 2.88, "learning_rate": 7.124206069160198e-05, "loss": 2.8829, "step": 36675 }, { "epoch": 2.88, "learning_rate": 7.124127656237748e-05, "loss": 3.0703, "step": 36676 }, { "epoch": 2.88, "learning_rate": 7.124049243315299e-05, "loss": 2.9206, "step": 36677 }, { "epoch": 2.88, "learning_rate": 7.123970830392849e-05, "loss": 2.8917, "step": 36678 }, { "epoch": 2.88, "learning_rate": 7.1238924174704e-05, "loss": 2.8475, "step": 36679 }, { "epoch": 2.88, "learning_rate": 7.12381400454795e-05, "loss": 2.979, "step": 36680 }, { "epoch": 2.88, "learning_rate": 7.1237355916255e-05, "loss": 2.8484, "step": 36681 }, { "epoch": 2.88, "learning_rate": 7.12365717870305e-05, "loss": 2.8673, "step": 36682 }, { "epoch": 2.88, "learning_rate": 7.123578765780601e-05, "loss": 2.7963, "step": 36683 }, { "epoch": 2.88, "learning_rate": 7.123500352858151e-05, "loss": 2.8521, "step": 36684 }, { "epoch": 2.88, "learning_rate": 7.123421939935702e-05, "loss": 2.895, "step": 36685 }, { "epoch": 2.88, "learning_rate": 7.123343527013252e-05, "loss": 2.8239, "step": 36686 }, { "epoch": 2.88, "learning_rate": 7.123265114090802e-05, "loss": 2.8324, "step": 36687 }, { "epoch": 2.88, "learning_rate": 7.123186701168353e-05, "loss": 3.1085, "step": 36688 }, { "epoch": 2.88, "learning_rate": 7.123108288245903e-05, "loss": 2.6876, "step": 36689 }, { "epoch": 2.88, "learning_rate": 7.123029875323453e-05, "loss": 2.6907, "step": 36690 }, { "epoch": 2.88, "learning_rate": 7.122951462401004e-05, "loss": 2.8085, "step": 36691 }, { "epoch": 2.88, "learning_rate": 7.122873049478554e-05, "loss": 2.6695, "step": 36692 }, { "epoch": 2.88, "learning_rate": 7.122794636556104e-05, "loss": 2.792, "step": 36693 }, { "epoch": 2.88, "learning_rate": 7.122716223633655e-05, "loss": 2.9338, "step": 36694 }, { "epoch": 2.88, "learning_rate": 7.122637810711205e-05, "loss": 2.8991, "step": 36695 }, { "epoch": 2.88, "learning_rate": 7.122559397788756e-05, "loss": 2.7509, "step": 36696 }, { "epoch": 2.88, "learning_rate": 7.122480984866306e-05, "loss": 2.4694, "step": 36697 }, { "epoch": 2.88, "learning_rate": 7.122402571943856e-05, "loss": 2.7088, "step": 36698 }, { "epoch": 2.88, "learning_rate": 7.122324159021408e-05, "loss": 2.9916, "step": 36699 }, { "epoch": 2.88, "learning_rate": 7.122245746098958e-05, "loss": 2.7875, "step": 36700 }, { "epoch": 2.88, "learning_rate": 7.122167333176509e-05, "loss": 2.8995, "step": 36701 }, { "epoch": 2.88, "learning_rate": 7.122088920254059e-05, "loss": 2.8092, "step": 36702 }, { "epoch": 2.88, "learning_rate": 7.12201050733161e-05, "loss": 2.7086, "step": 36703 }, { "epoch": 2.88, "learning_rate": 7.12193209440916e-05, "loss": 2.8217, "step": 36704 }, { "epoch": 2.88, "learning_rate": 7.12185368148671e-05, "loss": 2.8355, "step": 36705 }, { "epoch": 2.88, "learning_rate": 7.12177526856426e-05, "loss": 2.8238, "step": 36706 }, { "epoch": 2.88, "learning_rate": 7.121696855641811e-05, "loss": 2.9035, "step": 36707 }, { "epoch": 2.88, "learning_rate": 7.12161844271936e-05, "loss": 2.6791, "step": 36708 }, { "epoch": 2.88, "learning_rate": 7.12154002979691e-05, "loss": 2.7052, "step": 36709 }, { "epoch": 2.88, "learning_rate": 7.121461616874461e-05, "loss": 2.6072, "step": 36710 }, { "epoch": 2.88, "learning_rate": 7.121383203952011e-05, "loss": 2.6461, "step": 36711 }, { "epoch": 2.88, "learning_rate": 7.121304791029561e-05, "loss": 2.7725, "step": 36712 }, { "epoch": 2.88, "learning_rate": 7.121226378107112e-05, "loss": 3.0386, "step": 36713 }, { "epoch": 2.88, "learning_rate": 7.121147965184662e-05, "loss": 2.7887, "step": 36714 }, { "epoch": 2.88, "learning_rate": 7.121069552262213e-05, "loss": 2.5461, "step": 36715 }, { "epoch": 2.88, "learning_rate": 7.120991139339763e-05, "loss": 2.778, "step": 36716 }, { "epoch": 2.88, "learning_rate": 7.120912726417313e-05, "loss": 2.9576, "step": 36717 }, { "epoch": 2.88, "learning_rate": 7.120834313494864e-05, "loss": 2.9961, "step": 36718 }, { "epoch": 2.88, "learning_rate": 7.120755900572414e-05, "loss": 3.0393, "step": 36719 }, { "epoch": 2.88, "learning_rate": 7.120677487649964e-05, "loss": 2.8432, "step": 36720 }, { "epoch": 2.88, "learning_rate": 7.120599074727515e-05, "loss": 2.7257, "step": 36721 }, { "epoch": 2.88, "learning_rate": 7.120520661805065e-05, "loss": 2.931, "step": 36722 }, { "epoch": 2.88, "learning_rate": 7.120442248882616e-05, "loss": 2.9444, "step": 36723 }, { "epoch": 2.88, "learning_rate": 7.120363835960166e-05, "loss": 2.7063, "step": 36724 }, { "epoch": 2.88, "learning_rate": 7.120285423037718e-05, "loss": 2.9537, "step": 36725 }, { "epoch": 2.88, "learning_rate": 7.120207010115268e-05, "loss": 2.8337, "step": 36726 }, { "epoch": 2.88, "learning_rate": 7.120128597192818e-05, "loss": 2.6873, "step": 36727 }, { "epoch": 2.88, "learning_rate": 7.120050184270369e-05, "loss": 2.7704, "step": 36728 }, { "epoch": 2.88, "learning_rate": 7.119971771347919e-05, "loss": 2.9021, "step": 36729 }, { "epoch": 2.88, "learning_rate": 7.11989335842547e-05, "loss": 2.685, "step": 36730 }, { "epoch": 2.88, "learning_rate": 7.11981494550302e-05, "loss": 2.8995, "step": 36731 }, { "epoch": 2.88, "learning_rate": 7.11973653258057e-05, "loss": 2.6991, "step": 36732 }, { "epoch": 2.88, "learning_rate": 7.11965811965812e-05, "loss": 3.0172, "step": 36733 }, { "epoch": 2.88, "learning_rate": 7.119579706735671e-05, "loss": 2.763, "step": 36734 }, { "epoch": 2.88, "learning_rate": 7.119501293813221e-05, "loss": 2.7226, "step": 36735 }, { "epoch": 2.88, "learning_rate": 7.119422880890772e-05, "loss": 3.0143, "step": 36736 }, { "epoch": 2.88, "learning_rate": 7.119344467968322e-05, "loss": 2.7988, "step": 36737 }, { "epoch": 2.88, "learning_rate": 7.119266055045872e-05, "loss": 2.7729, "step": 36738 }, { "epoch": 2.88, "learning_rate": 7.119187642123423e-05, "loss": 2.6623, "step": 36739 }, { "epoch": 2.88, "learning_rate": 7.119109229200973e-05, "loss": 2.6435, "step": 36740 }, { "epoch": 2.88, "learning_rate": 7.119030816278523e-05, "loss": 2.7427, "step": 36741 }, { "epoch": 2.88, "learning_rate": 7.118952403356074e-05, "loss": 2.8168, "step": 36742 }, { "epoch": 2.88, "learning_rate": 7.118873990433624e-05, "loss": 2.7719, "step": 36743 }, { "epoch": 2.88, "learning_rate": 7.118795577511173e-05, "loss": 2.8303, "step": 36744 }, { "epoch": 2.88, "learning_rate": 7.118717164588724e-05, "loss": 2.8022, "step": 36745 }, { "epoch": 2.88, "learning_rate": 7.118638751666274e-05, "loss": 2.6674, "step": 36746 }, { "epoch": 2.88, "learning_rate": 7.118560338743824e-05, "loss": 2.6579, "step": 36747 }, { "epoch": 2.88, "learning_rate": 7.118481925821375e-05, "loss": 2.7272, "step": 36748 }, { "epoch": 2.88, "learning_rate": 7.118403512898925e-05, "loss": 2.703, "step": 36749 }, { "epoch": 2.88, "learning_rate": 7.118325099976475e-05, "loss": 3.0264, "step": 36750 }, { "epoch": 2.88, "learning_rate": 7.118246687054027e-05, "loss": 3.0458, "step": 36751 }, { "epoch": 2.88, "learning_rate": 7.118168274131578e-05, "loss": 2.4812, "step": 36752 }, { "epoch": 2.88, "learning_rate": 7.118089861209128e-05, "loss": 2.9631, "step": 36753 }, { "epoch": 2.88, "learning_rate": 7.118011448286678e-05, "loss": 2.7967, "step": 36754 }, { "epoch": 2.88, "learning_rate": 7.117933035364229e-05, "loss": 2.8382, "step": 36755 }, { "epoch": 2.88, "learning_rate": 7.117854622441779e-05, "loss": 2.9842, "step": 36756 }, { "epoch": 2.88, "learning_rate": 7.11777620951933e-05, "loss": 2.8445, "step": 36757 }, { "epoch": 2.88, "learning_rate": 7.11769779659688e-05, "loss": 2.5287, "step": 36758 }, { "epoch": 2.88, "learning_rate": 7.11761938367443e-05, "loss": 2.9459, "step": 36759 }, { "epoch": 2.88, "learning_rate": 7.11754097075198e-05, "loss": 2.7465, "step": 36760 }, { "epoch": 2.88, "learning_rate": 7.117462557829531e-05, "loss": 2.8029, "step": 36761 }, { "epoch": 2.88, "learning_rate": 7.117384144907081e-05, "loss": 2.6616, "step": 36762 }, { "epoch": 2.88, "learning_rate": 7.117305731984632e-05, "loss": 2.9846, "step": 36763 }, { "epoch": 2.88, "learning_rate": 7.117227319062182e-05, "loss": 2.9153, "step": 36764 }, { "epoch": 2.88, "learning_rate": 7.117148906139732e-05, "loss": 2.9107, "step": 36765 }, { "epoch": 2.88, "learning_rate": 7.117070493217283e-05, "loss": 2.8645, "step": 36766 }, { "epoch": 2.88, "learning_rate": 7.116992080294833e-05, "loss": 2.7762, "step": 36767 }, { "epoch": 2.88, "learning_rate": 7.116913667372383e-05, "loss": 2.7837, "step": 36768 }, { "epoch": 2.88, "learning_rate": 7.116835254449934e-05, "loss": 2.7232, "step": 36769 }, { "epoch": 2.88, "learning_rate": 7.116756841527484e-05, "loss": 2.6469, "step": 36770 }, { "epoch": 2.88, "learning_rate": 7.116678428605035e-05, "loss": 2.8581, "step": 36771 }, { "epoch": 2.88, "learning_rate": 7.116600015682585e-05, "loss": 2.8252, "step": 36772 }, { "epoch": 2.88, "learning_rate": 7.116521602760135e-05, "loss": 2.7181, "step": 36773 }, { "epoch": 2.88, "learning_rate": 7.116443189837686e-05, "loss": 2.8009, "step": 36774 }, { "epoch": 2.88, "learning_rate": 7.116364776915236e-05, "loss": 2.84, "step": 36775 }, { "epoch": 2.88, "learning_rate": 7.116286363992786e-05, "loss": 2.7591, "step": 36776 }, { "epoch": 2.88, "learning_rate": 7.116207951070337e-05, "loss": 2.7116, "step": 36777 }, { "epoch": 2.88, "learning_rate": 7.116129538147887e-05, "loss": 2.9706, "step": 36778 }, { "epoch": 2.88, "learning_rate": 7.116051125225437e-05, "loss": 2.8714, "step": 36779 }, { "epoch": 2.88, "learning_rate": 7.115972712302988e-05, "loss": 2.9584, "step": 36780 }, { "epoch": 2.88, "learning_rate": 7.115894299380538e-05, "loss": 2.8438, "step": 36781 }, { "epoch": 2.88, "learning_rate": 7.115815886458089e-05, "loss": 2.9693, "step": 36782 }, { "epoch": 2.88, "learning_rate": 7.115737473535639e-05, "loss": 2.6508, "step": 36783 }, { "epoch": 2.88, "learning_rate": 7.115659060613189e-05, "loss": 2.7852, "step": 36784 }, { "epoch": 2.88, "learning_rate": 7.11558064769074e-05, "loss": 2.9416, "step": 36785 }, { "epoch": 2.88, "learning_rate": 7.11550223476829e-05, "loss": 2.7835, "step": 36786 }, { "epoch": 2.88, "learning_rate": 7.11542382184584e-05, "loss": 2.6451, "step": 36787 }, { "epoch": 2.88, "learning_rate": 7.115345408923391e-05, "loss": 2.726, "step": 36788 }, { "epoch": 2.88, "learning_rate": 7.115266996000941e-05, "loss": 2.8084, "step": 36789 }, { "epoch": 2.88, "learning_rate": 7.115188583078492e-05, "loss": 2.7386, "step": 36790 }, { "epoch": 2.88, "learning_rate": 7.115110170156042e-05, "loss": 2.6235, "step": 36791 }, { "epoch": 2.88, "learning_rate": 7.115031757233592e-05, "loss": 3.1165, "step": 36792 }, { "epoch": 2.89, "learning_rate": 7.114953344311143e-05, "loss": 2.897, "step": 36793 }, { "epoch": 2.89, "learning_rate": 7.114874931388693e-05, "loss": 2.8005, "step": 36794 }, { "epoch": 2.89, "learning_rate": 7.114796518466243e-05, "loss": 2.6497, "step": 36795 }, { "epoch": 2.89, "learning_rate": 7.114718105543794e-05, "loss": 2.9637, "step": 36796 }, { "epoch": 2.89, "learning_rate": 7.114639692621344e-05, "loss": 2.7659, "step": 36797 }, { "epoch": 2.89, "learning_rate": 7.114561279698894e-05, "loss": 2.7402, "step": 36798 }, { "epoch": 2.89, "learning_rate": 7.114482866776445e-05, "loss": 2.8028, "step": 36799 }, { "epoch": 2.89, "learning_rate": 7.114404453853995e-05, "loss": 2.9804, "step": 36800 }, { "epoch": 2.89, "learning_rate": 7.114326040931546e-05, "loss": 2.799, "step": 36801 }, { "epoch": 2.89, "learning_rate": 7.114247628009096e-05, "loss": 2.6854, "step": 36802 }, { "epoch": 2.89, "learning_rate": 7.114169215086646e-05, "loss": 3.0417, "step": 36803 }, { "epoch": 2.89, "learning_rate": 7.114090802164197e-05, "loss": 2.9314, "step": 36804 }, { "epoch": 2.89, "learning_rate": 7.114012389241747e-05, "loss": 3.0107, "step": 36805 }, { "epoch": 2.89, "learning_rate": 7.113933976319297e-05, "loss": 2.7459, "step": 36806 }, { "epoch": 2.89, "learning_rate": 7.113855563396848e-05, "loss": 2.794, "step": 36807 }, { "epoch": 2.89, "learning_rate": 7.1137771504744e-05, "loss": 2.8651, "step": 36808 }, { "epoch": 2.89, "learning_rate": 7.11369873755195e-05, "loss": 2.8313, "step": 36809 }, { "epoch": 2.89, "learning_rate": 7.1136203246295e-05, "loss": 2.7961, "step": 36810 }, { "epoch": 2.89, "learning_rate": 7.11354191170705e-05, "loss": 2.6345, "step": 36811 }, { "epoch": 2.89, "learning_rate": 7.113463498784601e-05, "loss": 2.6791, "step": 36812 }, { "epoch": 2.89, "learning_rate": 7.113385085862151e-05, "loss": 2.7814, "step": 36813 }, { "epoch": 2.89, "learning_rate": 7.113306672939702e-05, "loss": 3.1455, "step": 36814 }, { "epoch": 2.89, "learning_rate": 7.113228260017252e-05, "loss": 2.6644, "step": 36815 }, { "epoch": 2.89, "learning_rate": 7.113149847094801e-05, "loss": 2.7471, "step": 36816 }, { "epoch": 2.89, "learning_rate": 7.113071434172351e-05, "loss": 2.7785, "step": 36817 }, { "epoch": 2.89, "learning_rate": 7.112993021249902e-05, "loss": 2.8848, "step": 36818 }, { "epoch": 2.89, "learning_rate": 7.112914608327452e-05, "loss": 2.8148, "step": 36819 }, { "epoch": 2.89, "learning_rate": 7.112836195405003e-05, "loss": 2.7543, "step": 36820 }, { "epoch": 2.89, "learning_rate": 7.112757782482553e-05, "loss": 2.589, "step": 36821 }, { "epoch": 2.89, "learning_rate": 7.112679369560103e-05, "loss": 2.9808, "step": 36822 }, { "epoch": 2.89, "learning_rate": 7.112600956637654e-05, "loss": 2.775, "step": 36823 }, { "epoch": 2.89, "learning_rate": 7.112522543715204e-05, "loss": 2.6035, "step": 36824 }, { "epoch": 2.89, "learning_rate": 7.112444130792754e-05, "loss": 2.7686, "step": 36825 }, { "epoch": 2.89, "learning_rate": 7.112365717870305e-05, "loss": 2.8331, "step": 36826 }, { "epoch": 2.89, "learning_rate": 7.112287304947855e-05, "loss": 2.7047, "step": 36827 }, { "epoch": 2.89, "learning_rate": 7.112208892025405e-05, "loss": 2.6757, "step": 36828 }, { "epoch": 2.89, "learning_rate": 7.112130479102956e-05, "loss": 2.66, "step": 36829 }, { "epoch": 2.89, "learning_rate": 7.112052066180506e-05, "loss": 2.7277, "step": 36830 }, { "epoch": 2.89, "learning_rate": 7.111973653258057e-05, "loss": 2.7715, "step": 36831 }, { "epoch": 2.89, "learning_rate": 7.111895240335607e-05, "loss": 2.6604, "step": 36832 }, { "epoch": 2.89, "learning_rate": 7.111816827413159e-05, "loss": 2.897, "step": 36833 }, { "epoch": 2.89, "learning_rate": 7.111738414490709e-05, "loss": 2.7004, "step": 36834 }, { "epoch": 2.89, "learning_rate": 7.11166000156826e-05, "loss": 2.7526, "step": 36835 }, { "epoch": 2.89, "learning_rate": 7.11158158864581e-05, "loss": 2.766, "step": 36836 }, { "epoch": 2.89, "learning_rate": 7.11150317572336e-05, "loss": 2.7551, "step": 36837 }, { "epoch": 2.89, "learning_rate": 7.11142476280091e-05, "loss": 3.045, "step": 36838 }, { "epoch": 2.89, "learning_rate": 7.111346349878461e-05, "loss": 2.9442, "step": 36839 }, { "epoch": 2.89, "learning_rate": 7.111267936956011e-05, "loss": 2.8904, "step": 36840 }, { "epoch": 2.89, "learning_rate": 7.111189524033562e-05, "loss": 2.8829, "step": 36841 }, { "epoch": 2.89, "learning_rate": 7.111111111111112e-05, "loss": 2.6974, "step": 36842 }, { "epoch": 2.89, "learning_rate": 7.111032698188662e-05, "loss": 2.8728, "step": 36843 }, { "epoch": 2.89, "learning_rate": 7.110954285266213e-05, "loss": 3.1312, "step": 36844 }, { "epoch": 2.89, "learning_rate": 7.110875872343763e-05, "loss": 2.7309, "step": 36845 }, { "epoch": 2.89, "learning_rate": 7.110797459421313e-05, "loss": 2.8556, "step": 36846 }, { "epoch": 2.89, "learning_rate": 7.110719046498864e-05, "loss": 2.8336, "step": 36847 }, { "epoch": 2.89, "learning_rate": 7.110640633576414e-05, "loss": 2.5903, "step": 36848 }, { "epoch": 2.89, "learning_rate": 7.110562220653965e-05, "loss": 2.6831, "step": 36849 }, { "epoch": 2.89, "learning_rate": 7.110483807731515e-05, "loss": 2.8595, "step": 36850 }, { "epoch": 2.89, "learning_rate": 7.110405394809065e-05, "loss": 2.7944, "step": 36851 }, { "epoch": 2.89, "learning_rate": 7.110326981886614e-05, "loss": 2.8436, "step": 36852 }, { "epoch": 2.89, "learning_rate": 7.110248568964165e-05, "loss": 2.9024, "step": 36853 }, { "epoch": 2.89, "learning_rate": 7.110170156041715e-05, "loss": 2.7977, "step": 36854 }, { "epoch": 2.89, "learning_rate": 7.110091743119265e-05, "loss": 2.8905, "step": 36855 }, { "epoch": 2.89, "learning_rate": 7.110013330196816e-05, "loss": 2.6883, "step": 36856 }, { "epoch": 2.89, "learning_rate": 7.109934917274366e-05, "loss": 2.765, "step": 36857 }, { "epoch": 2.89, "learning_rate": 7.109856504351917e-05, "loss": 2.8287, "step": 36858 }, { "epoch": 2.89, "learning_rate": 7.109778091429468e-05, "loss": 2.9249, "step": 36859 }, { "epoch": 2.89, "learning_rate": 7.109699678507019e-05, "loss": 2.6931, "step": 36860 }, { "epoch": 2.89, "learning_rate": 7.109621265584569e-05, "loss": 2.9244, "step": 36861 }, { "epoch": 2.89, "learning_rate": 7.10954285266212e-05, "loss": 2.9875, "step": 36862 }, { "epoch": 2.89, "learning_rate": 7.10946443973967e-05, "loss": 2.7994, "step": 36863 }, { "epoch": 2.89, "learning_rate": 7.10938602681722e-05, "loss": 2.8906, "step": 36864 }, { "epoch": 2.89, "learning_rate": 7.10930761389477e-05, "loss": 2.8562, "step": 36865 }, { "epoch": 2.89, "learning_rate": 7.109229200972321e-05, "loss": 3.203, "step": 36866 }, { "epoch": 2.89, "learning_rate": 7.109150788049871e-05, "loss": 2.8298, "step": 36867 }, { "epoch": 2.89, "learning_rate": 7.109072375127422e-05, "loss": 2.9485, "step": 36868 }, { "epoch": 2.89, "learning_rate": 7.108993962204972e-05, "loss": 2.6998, "step": 36869 }, { "epoch": 2.89, "learning_rate": 7.108915549282522e-05, "loss": 2.6917, "step": 36870 }, { "epoch": 2.89, "learning_rate": 7.108837136360073e-05, "loss": 2.6575, "step": 36871 }, { "epoch": 2.89, "learning_rate": 7.108758723437623e-05, "loss": 2.7915, "step": 36872 }, { "epoch": 2.89, "learning_rate": 7.108680310515173e-05, "loss": 2.65, "step": 36873 }, { "epoch": 2.89, "learning_rate": 7.108601897592724e-05, "loss": 2.7784, "step": 36874 }, { "epoch": 2.89, "learning_rate": 7.108523484670274e-05, "loss": 2.8228, "step": 36875 }, { "epoch": 2.89, "learning_rate": 7.108445071747824e-05, "loss": 2.9554, "step": 36876 }, { "epoch": 2.89, "learning_rate": 7.108366658825375e-05, "loss": 2.7411, "step": 36877 }, { "epoch": 2.89, "learning_rate": 7.108288245902925e-05, "loss": 2.873, "step": 36878 }, { "epoch": 2.89, "learning_rate": 7.108209832980476e-05, "loss": 2.7151, "step": 36879 }, { "epoch": 2.89, "learning_rate": 7.108131420058026e-05, "loss": 2.8639, "step": 36880 }, { "epoch": 2.89, "learning_rate": 7.108053007135576e-05, "loss": 2.8468, "step": 36881 }, { "epoch": 2.89, "learning_rate": 7.107974594213127e-05, "loss": 3.1331, "step": 36882 }, { "epoch": 2.89, "learning_rate": 7.107896181290677e-05, "loss": 2.8357, "step": 36883 }, { "epoch": 2.89, "learning_rate": 7.107817768368227e-05, "loss": 2.9983, "step": 36884 }, { "epoch": 2.89, "learning_rate": 7.107739355445778e-05, "loss": 2.8415, "step": 36885 }, { "epoch": 2.89, "learning_rate": 7.107660942523328e-05, "loss": 2.859, "step": 36886 }, { "epoch": 2.89, "learning_rate": 7.107582529600879e-05, "loss": 2.7607, "step": 36887 }, { "epoch": 2.89, "learning_rate": 7.107504116678429e-05, "loss": 2.8708, "step": 36888 }, { "epoch": 2.89, "learning_rate": 7.107425703755979e-05, "loss": 2.6183, "step": 36889 }, { "epoch": 2.89, "learning_rate": 7.10734729083353e-05, "loss": 2.6992, "step": 36890 }, { "epoch": 2.89, "learning_rate": 7.10726887791108e-05, "loss": 2.7986, "step": 36891 }, { "epoch": 2.89, "learning_rate": 7.10719046498863e-05, "loss": 2.7274, "step": 36892 }, { "epoch": 2.89, "learning_rate": 7.107112052066181e-05, "loss": 2.8653, "step": 36893 }, { "epoch": 2.89, "learning_rate": 7.107033639143731e-05, "loss": 3.0378, "step": 36894 }, { "epoch": 2.89, "learning_rate": 7.106955226221281e-05, "loss": 2.7246, "step": 36895 }, { "epoch": 2.89, "learning_rate": 7.106876813298832e-05, "loss": 2.8744, "step": 36896 }, { "epoch": 2.89, "learning_rate": 7.106798400376382e-05, "loss": 2.83, "step": 36897 }, { "epoch": 2.89, "learning_rate": 7.106719987453933e-05, "loss": 2.8028, "step": 36898 }, { "epoch": 2.89, "learning_rate": 7.106641574531483e-05, "loss": 2.9681, "step": 36899 }, { "epoch": 2.89, "learning_rate": 7.106563161609033e-05, "loss": 2.7086, "step": 36900 }, { "epoch": 2.89, "learning_rate": 7.106484748686584e-05, "loss": 2.7849, "step": 36901 }, { "epoch": 2.89, "learning_rate": 7.106406335764134e-05, "loss": 2.708, "step": 36902 }, { "epoch": 2.89, "learning_rate": 7.106327922841684e-05, "loss": 2.6153, "step": 36903 }, { "epoch": 2.89, "learning_rate": 7.106249509919235e-05, "loss": 2.7005, "step": 36904 }, { "epoch": 2.89, "learning_rate": 7.106171096996785e-05, "loss": 2.8579, "step": 36905 }, { "epoch": 2.89, "learning_rate": 7.106092684074336e-05, "loss": 2.7949, "step": 36906 }, { "epoch": 2.89, "learning_rate": 7.106014271151886e-05, "loss": 2.7594, "step": 36907 }, { "epoch": 2.89, "learning_rate": 7.105935858229436e-05, "loss": 3.0303, "step": 36908 }, { "epoch": 2.89, "learning_rate": 7.105857445306987e-05, "loss": 2.8356, "step": 36909 }, { "epoch": 2.89, "learning_rate": 7.105779032384537e-05, "loss": 2.8173, "step": 36910 }, { "epoch": 2.89, "learning_rate": 7.105700619462087e-05, "loss": 2.7832, "step": 36911 }, { "epoch": 2.89, "learning_rate": 7.105622206539638e-05, "loss": 3.1736, "step": 36912 }, { "epoch": 2.89, "learning_rate": 7.105543793617188e-05, "loss": 2.745, "step": 36913 }, { "epoch": 2.89, "learning_rate": 7.105465380694738e-05, "loss": 2.9722, "step": 36914 }, { "epoch": 2.89, "learning_rate": 7.105386967772289e-05, "loss": 2.7017, "step": 36915 }, { "epoch": 2.89, "learning_rate": 7.10530855484984e-05, "loss": 2.9607, "step": 36916 }, { "epoch": 2.89, "learning_rate": 7.105230141927391e-05, "loss": 2.7806, "step": 36917 }, { "epoch": 2.89, "learning_rate": 7.105151729004941e-05, "loss": 2.7311, "step": 36918 }, { "epoch": 2.89, "learning_rate": 7.105073316082492e-05, "loss": 2.967, "step": 36919 }, { "epoch": 2.9, "learning_rate": 7.104994903160042e-05, "loss": 2.8374, "step": 36920 }, { "epoch": 2.9, "learning_rate": 7.104916490237592e-05, "loss": 2.5855, "step": 36921 }, { "epoch": 2.9, "learning_rate": 7.104838077315143e-05, "loss": 2.7594, "step": 36922 }, { "epoch": 2.9, "learning_rate": 7.104759664392693e-05, "loss": 2.8031, "step": 36923 }, { "epoch": 2.9, "learning_rate": 7.104681251470242e-05, "loss": 2.8994, "step": 36924 }, { "epoch": 2.9, "learning_rate": 7.104602838547793e-05, "loss": 2.7459, "step": 36925 }, { "epoch": 2.9, "learning_rate": 7.104524425625343e-05, "loss": 2.8239, "step": 36926 }, { "epoch": 2.9, "learning_rate": 7.104446012702893e-05, "loss": 2.7485, "step": 36927 }, { "epoch": 2.9, "learning_rate": 7.104367599780444e-05, "loss": 3.07, "step": 36928 }, { "epoch": 2.9, "learning_rate": 7.104289186857994e-05, "loss": 2.8122, "step": 36929 }, { "epoch": 2.9, "learning_rate": 7.104210773935544e-05, "loss": 2.9995, "step": 36930 }, { "epoch": 2.9, "learning_rate": 7.104132361013095e-05, "loss": 2.8812, "step": 36931 }, { "epoch": 2.9, "learning_rate": 7.104053948090645e-05, "loss": 2.9514, "step": 36932 }, { "epoch": 2.9, "learning_rate": 7.103975535168195e-05, "loss": 2.5808, "step": 36933 }, { "epoch": 2.9, "learning_rate": 7.103897122245746e-05, "loss": 2.771, "step": 36934 }, { "epoch": 2.9, "learning_rate": 7.103818709323296e-05, "loss": 3.039, "step": 36935 }, { "epoch": 2.9, "learning_rate": 7.103740296400847e-05, "loss": 2.8412, "step": 36936 }, { "epoch": 2.9, "learning_rate": 7.103661883478397e-05, "loss": 2.8445, "step": 36937 }, { "epoch": 2.9, "learning_rate": 7.103583470555947e-05, "loss": 2.7681, "step": 36938 }, { "epoch": 2.9, "learning_rate": 7.103505057633498e-05, "loss": 2.8758, "step": 36939 }, { "epoch": 2.9, "learning_rate": 7.103426644711048e-05, "loss": 2.8844, "step": 36940 }, { "epoch": 2.9, "learning_rate": 7.1033482317886e-05, "loss": 2.7963, "step": 36941 }, { "epoch": 2.9, "learning_rate": 7.10326981886615e-05, "loss": 2.566, "step": 36942 }, { "epoch": 2.9, "learning_rate": 7.1031914059437e-05, "loss": 2.7893, "step": 36943 }, { "epoch": 2.9, "learning_rate": 7.103112993021251e-05, "loss": 2.7175, "step": 36944 }, { "epoch": 2.9, "learning_rate": 7.103034580098801e-05, "loss": 2.8646, "step": 36945 }, { "epoch": 2.9, "learning_rate": 7.102956167176352e-05, "loss": 2.6918, "step": 36946 }, { "epoch": 2.9, "learning_rate": 7.102877754253902e-05, "loss": 2.7868, "step": 36947 }, { "epoch": 2.9, "learning_rate": 7.102799341331452e-05, "loss": 2.8859, "step": 36948 }, { "epoch": 2.9, "learning_rate": 7.102720928409003e-05, "loss": 2.7134, "step": 36949 }, { "epoch": 2.9, "learning_rate": 7.102642515486553e-05, "loss": 2.7745, "step": 36950 }, { "epoch": 2.9, "learning_rate": 7.102564102564103e-05, "loss": 2.5844, "step": 36951 }, { "epoch": 2.9, "learning_rate": 7.102485689641654e-05, "loss": 2.6905, "step": 36952 }, { "epoch": 2.9, "learning_rate": 7.102407276719204e-05, "loss": 2.6789, "step": 36953 }, { "epoch": 2.9, "learning_rate": 7.102328863796755e-05, "loss": 2.6485, "step": 36954 }, { "epoch": 2.9, "learning_rate": 7.102250450874305e-05, "loss": 2.6672, "step": 36955 }, { "epoch": 2.9, "learning_rate": 7.102172037951855e-05, "loss": 2.6938, "step": 36956 }, { "epoch": 2.9, "learning_rate": 7.102093625029406e-05, "loss": 2.7309, "step": 36957 }, { "epoch": 2.9, "learning_rate": 7.102015212106956e-05, "loss": 2.7491, "step": 36958 }, { "epoch": 2.9, "learning_rate": 7.101936799184506e-05, "loss": 2.9373, "step": 36959 }, { "epoch": 2.9, "learning_rate": 7.101858386262055e-05, "loss": 2.9085, "step": 36960 }, { "epoch": 2.9, "learning_rate": 7.101779973339606e-05, "loss": 2.7251, "step": 36961 }, { "epoch": 2.9, "learning_rate": 7.101701560417156e-05, "loss": 2.7145, "step": 36962 }, { "epoch": 2.9, "learning_rate": 7.101623147494706e-05, "loss": 2.7078, "step": 36963 }, { "epoch": 2.9, "learning_rate": 7.101544734572257e-05, "loss": 2.7862, "step": 36964 }, { "epoch": 2.9, "learning_rate": 7.101466321649807e-05, "loss": 2.8068, "step": 36965 }, { "epoch": 2.9, "learning_rate": 7.101387908727358e-05, "loss": 2.8006, "step": 36966 }, { "epoch": 2.9, "learning_rate": 7.101309495804909e-05, "loss": 3.0282, "step": 36967 }, { "epoch": 2.9, "learning_rate": 7.10123108288246e-05, "loss": 2.996, "step": 36968 }, { "epoch": 2.9, "learning_rate": 7.10115266996001e-05, "loss": 2.9272, "step": 36969 }, { "epoch": 2.9, "learning_rate": 7.10107425703756e-05, "loss": 2.6105, "step": 36970 }, { "epoch": 2.9, "learning_rate": 7.100995844115111e-05, "loss": 2.6534, "step": 36971 }, { "epoch": 2.9, "learning_rate": 7.100917431192661e-05, "loss": 2.7843, "step": 36972 }, { "epoch": 2.9, "learning_rate": 7.100839018270212e-05, "loss": 2.6743, "step": 36973 }, { "epoch": 2.9, "learning_rate": 7.100760605347762e-05, "loss": 2.7985, "step": 36974 }, { "epoch": 2.9, "learning_rate": 7.100682192425312e-05, "loss": 2.8594, "step": 36975 }, { "epoch": 2.9, "learning_rate": 7.100603779502863e-05, "loss": 2.9671, "step": 36976 }, { "epoch": 2.9, "learning_rate": 7.100525366580413e-05, "loss": 2.8891, "step": 36977 }, { "epoch": 2.9, "learning_rate": 7.100446953657963e-05, "loss": 2.7833, "step": 36978 }, { "epoch": 2.9, "learning_rate": 7.100368540735514e-05, "loss": 2.8524, "step": 36979 }, { "epoch": 2.9, "learning_rate": 7.100290127813064e-05, "loss": 3.0514, "step": 36980 }, { "epoch": 2.9, "learning_rate": 7.100211714890614e-05, "loss": 2.705, "step": 36981 }, { "epoch": 2.9, "learning_rate": 7.100133301968165e-05, "loss": 2.7299, "step": 36982 }, { "epoch": 2.9, "learning_rate": 7.100054889045715e-05, "loss": 3.0388, "step": 36983 }, { "epoch": 2.9, "learning_rate": 7.099976476123266e-05, "loss": 2.7724, "step": 36984 }, { "epoch": 2.9, "learning_rate": 7.099898063200816e-05, "loss": 2.6297, "step": 36985 }, { "epoch": 2.9, "learning_rate": 7.099819650278366e-05, "loss": 2.6538, "step": 36986 }, { "epoch": 2.9, "learning_rate": 7.099741237355917e-05, "loss": 2.7337, "step": 36987 }, { "epoch": 2.9, "learning_rate": 7.099662824433467e-05, "loss": 2.7586, "step": 36988 }, { "epoch": 2.9, "learning_rate": 7.099584411511017e-05, "loss": 2.8035, "step": 36989 }, { "epoch": 2.9, "learning_rate": 7.099505998588568e-05, "loss": 2.8593, "step": 36990 }, { "epoch": 2.9, "learning_rate": 7.099427585666118e-05, "loss": 2.7749, "step": 36991 }, { "epoch": 2.9, "learning_rate": 7.099349172743669e-05, "loss": 2.8651, "step": 36992 }, { "epoch": 2.9, "learning_rate": 7.099270759821219e-05, "loss": 2.594, "step": 36993 }, { "epoch": 2.9, "learning_rate": 7.099192346898769e-05, "loss": 2.8716, "step": 36994 }, { "epoch": 2.9, "learning_rate": 7.09911393397632e-05, "loss": 2.808, "step": 36995 }, { "epoch": 2.9, "learning_rate": 7.09903552105387e-05, "loss": 2.9842, "step": 36996 }, { "epoch": 2.9, "learning_rate": 7.09895710813142e-05, "loss": 2.7592, "step": 36997 }, { "epoch": 2.9, "learning_rate": 7.098878695208971e-05, "loss": 2.5994, "step": 36998 }, { "epoch": 2.9, "learning_rate": 7.098800282286521e-05, "loss": 2.6638, "step": 36999 }, { "epoch": 2.9, "learning_rate": 7.098721869364071e-05, "loss": 3.0191, "step": 37000 }, { "epoch": 2.9, "eval_loss": 2.893411874771118, "eval_runtime": 218.3199, "eval_samples_per_second": 51.905, "eval_steps_per_second": 3.248, "step": 37000 }, { "epoch": 2.9, "learning_rate": 7.098643456441622e-05, "loss": 2.7487, "step": 37001 }, { "epoch": 2.9, "learning_rate": 7.098565043519172e-05, "loss": 2.7829, "step": 37002 }, { "epoch": 2.9, "learning_rate": 7.098486630596723e-05, "loss": 2.7403, "step": 37003 }, { "epoch": 2.9, "learning_rate": 7.098408217674273e-05, "loss": 3.0598, "step": 37004 }, { "epoch": 2.9, "learning_rate": 7.098329804751823e-05, "loss": 2.9787, "step": 37005 }, { "epoch": 2.9, "learning_rate": 7.098251391829374e-05, "loss": 3.0738, "step": 37006 }, { "epoch": 2.9, "learning_rate": 7.098172978906924e-05, "loss": 2.8582, "step": 37007 }, { "epoch": 2.9, "learning_rate": 7.098094565984474e-05, "loss": 2.7299, "step": 37008 }, { "epoch": 2.9, "learning_rate": 7.098016153062025e-05, "loss": 2.8639, "step": 37009 }, { "epoch": 2.9, "learning_rate": 7.097937740139575e-05, "loss": 2.7444, "step": 37010 }, { "epoch": 2.9, "learning_rate": 7.097859327217126e-05, "loss": 3.0814, "step": 37011 }, { "epoch": 2.9, "learning_rate": 7.097780914294676e-05, "loss": 2.6024, "step": 37012 }, { "epoch": 2.9, "learning_rate": 7.097702501372226e-05, "loss": 2.7884, "step": 37013 }, { "epoch": 2.9, "learning_rate": 7.097624088449777e-05, "loss": 3.0174, "step": 37014 }, { "epoch": 2.9, "learning_rate": 7.097545675527327e-05, "loss": 2.7012, "step": 37015 }, { "epoch": 2.9, "learning_rate": 7.097467262604877e-05, "loss": 2.9131, "step": 37016 }, { "epoch": 2.9, "learning_rate": 7.097388849682428e-05, "loss": 2.8876, "step": 37017 }, { "epoch": 2.9, "learning_rate": 7.097310436759978e-05, "loss": 2.87, "step": 37018 }, { "epoch": 2.9, "learning_rate": 7.097232023837528e-05, "loss": 2.8112, "step": 37019 }, { "epoch": 2.9, "learning_rate": 7.097153610915079e-05, "loss": 2.877, "step": 37020 }, { "epoch": 2.9, "learning_rate": 7.097075197992629e-05, "loss": 2.8611, "step": 37021 }, { "epoch": 2.9, "learning_rate": 7.09699678507018e-05, "loss": 2.7539, "step": 37022 }, { "epoch": 2.9, "learning_rate": 7.09691837214773e-05, "loss": 2.7057, "step": 37023 }, { "epoch": 2.9, "learning_rate": 7.096839959225282e-05, "loss": 2.9405, "step": 37024 }, { "epoch": 2.9, "learning_rate": 7.096761546302832e-05, "loss": 2.6527, "step": 37025 }, { "epoch": 2.9, "learning_rate": 7.096683133380382e-05, "loss": 3.0446, "step": 37026 }, { "epoch": 2.9, "learning_rate": 7.096604720457933e-05, "loss": 3.0047, "step": 37027 }, { "epoch": 2.9, "learning_rate": 7.096526307535483e-05, "loss": 2.5809, "step": 37028 }, { "epoch": 2.9, "learning_rate": 7.096447894613033e-05, "loss": 3.3633, "step": 37029 }, { "epoch": 2.9, "learning_rate": 7.096369481690584e-05, "loss": 2.8067, "step": 37030 }, { "epoch": 2.9, "learning_rate": 7.096291068768134e-05, "loss": 2.6919, "step": 37031 }, { "epoch": 2.9, "learning_rate": 7.096212655845683e-05, "loss": 2.7652, "step": 37032 }, { "epoch": 2.9, "learning_rate": 7.096134242923234e-05, "loss": 2.9285, "step": 37033 }, { "epoch": 2.9, "learning_rate": 7.096055830000784e-05, "loss": 2.8959, "step": 37034 }, { "epoch": 2.9, "learning_rate": 7.095977417078334e-05, "loss": 2.601, "step": 37035 }, { "epoch": 2.9, "learning_rate": 7.095899004155885e-05, "loss": 2.6475, "step": 37036 }, { "epoch": 2.9, "learning_rate": 7.095820591233435e-05, "loss": 2.6553, "step": 37037 }, { "epoch": 2.9, "learning_rate": 7.095742178310985e-05, "loss": 2.646, "step": 37038 }, { "epoch": 2.9, "learning_rate": 7.095663765388536e-05, "loss": 2.9561, "step": 37039 }, { "epoch": 2.9, "learning_rate": 7.095585352466086e-05, "loss": 3.0853, "step": 37040 }, { "epoch": 2.9, "learning_rate": 7.095506939543637e-05, "loss": 2.8041, "step": 37041 }, { "epoch": 2.9, "learning_rate": 7.095428526621187e-05, "loss": 2.7398, "step": 37042 }, { "epoch": 2.9, "learning_rate": 7.095350113698737e-05, "loss": 2.9582, "step": 37043 }, { "epoch": 2.9, "learning_rate": 7.095271700776288e-05, "loss": 2.6518, "step": 37044 }, { "epoch": 2.9, "learning_rate": 7.095193287853838e-05, "loss": 2.8906, "step": 37045 }, { "epoch": 2.9, "learning_rate": 7.095114874931388e-05, "loss": 2.7327, "step": 37046 }, { "epoch": 2.9, "learning_rate": 7.095036462008939e-05, "loss": 2.7786, "step": 37047 }, { "epoch": 2.91, "learning_rate": 7.094958049086489e-05, "loss": 2.6624, "step": 37048 }, { "epoch": 2.91, "learning_rate": 7.094879636164041e-05, "loss": 2.7972, "step": 37049 }, { "epoch": 2.91, "learning_rate": 7.094801223241591e-05, "loss": 3.0336, "step": 37050 }, { "epoch": 2.91, "learning_rate": 7.094722810319142e-05, "loss": 2.7169, "step": 37051 }, { "epoch": 2.91, "learning_rate": 7.094644397396692e-05, "loss": 3.0458, "step": 37052 }, { "epoch": 2.91, "learning_rate": 7.094565984474242e-05, "loss": 2.7203, "step": 37053 }, { "epoch": 2.91, "learning_rate": 7.094487571551793e-05, "loss": 2.8137, "step": 37054 }, { "epoch": 2.91, "learning_rate": 7.094409158629343e-05, "loss": 2.8652, "step": 37055 }, { "epoch": 2.91, "learning_rate": 7.094330745706893e-05, "loss": 2.939, "step": 37056 }, { "epoch": 2.91, "learning_rate": 7.094252332784444e-05, "loss": 2.6549, "step": 37057 }, { "epoch": 2.91, "learning_rate": 7.094173919861994e-05, "loss": 2.927, "step": 37058 }, { "epoch": 2.91, "learning_rate": 7.094095506939545e-05, "loss": 2.6879, "step": 37059 }, { "epoch": 2.91, "learning_rate": 7.094017094017095e-05, "loss": 3.142, "step": 37060 }, { "epoch": 2.91, "learning_rate": 7.093938681094645e-05, "loss": 2.8044, "step": 37061 }, { "epoch": 2.91, "learning_rate": 7.093860268172196e-05, "loss": 2.7431, "step": 37062 }, { "epoch": 2.91, "learning_rate": 7.093781855249746e-05, "loss": 2.7421, "step": 37063 }, { "epoch": 2.91, "learning_rate": 7.093703442327296e-05, "loss": 2.7691, "step": 37064 }, { "epoch": 2.91, "learning_rate": 7.093625029404847e-05, "loss": 2.7352, "step": 37065 }, { "epoch": 2.91, "learning_rate": 7.093546616482397e-05, "loss": 2.7936, "step": 37066 }, { "epoch": 2.91, "learning_rate": 7.093468203559947e-05, "loss": 2.8341, "step": 37067 }, { "epoch": 2.91, "learning_rate": 7.093389790637496e-05, "loss": 3.0294, "step": 37068 }, { "epoch": 2.91, "learning_rate": 7.093311377715047e-05, "loss": 2.6081, "step": 37069 }, { "epoch": 2.91, "learning_rate": 7.093232964792597e-05, "loss": 2.8245, "step": 37070 }, { "epoch": 2.91, "learning_rate": 7.093154551870148e-05, "loss": 2.8957, "step": 37071 }, { "epoch": 2.91, "learning_rate": 7.093076138947698e-05, "loss": 2.7274, "step": 37072 }, { "epoch": 2.91, "learning_rate": 7.092997726025248e-05, "loss": 2.9259, "step": 37073 }, { "epoch": 2.91, "learning_rate": 7.092919313102799e-05, "loss": 2.7418, "step": 37074 }, { "epoch": 2.91, "learning_rate": 7.09284090018035e-05, "loss": 2.7844, "step": 37075 }, { "epoch": 2.91, "learning_rate": 7.092762487257901e-05, "loss": 2.8603, "step": 37076 }, { "epoch": 2.91, "learning_rate": 7.092684074335451e-05, "loss": 2.7005, "step": 37077 }, { "epoch": 2.91, "learning_rate": 7.092605661413001e-05, "loss": 3.119, "step": 37078 }, { "epoch": 2.91, "learning_rate": 7.092527248490552e-05, "loss": 2.9314, "step": 37079 }, { "epoch": 2.91, "learning_rate": 7.092448835568102e-05, "loss": 2.9591, "step": 37080 }, { "epoch": 2.91, "learning_rate": 7.092370422645653e-05, "loss": 2.6643, "step": 37081 }, { "epoch": 2.91, "learning_rate": 7.092292009723203e-05, "loss": 2.8489, "step": 37082 }, { "epoch": 2.91, "learning_rate": 7.092213596800753e-05, "loss": 2.8595, "step": 37083 }, { "epoch": 2.91, "learning_rate": 7.092135183878304e-05, "loss": 2.941, "step": 37084 }, { "epoch": 2.91, "learning_rate": 7.092056770955854e-05, "loss": 2.7127, "step": 37085 }, { "epoch": 2.91, "learning_rate": 7.091978358033404e-05, "loss": 2.992, "step": 37086 }, { "epoch": 2.91, "learning_rate": 7.091899945110955e-05, "loss": 3.0747, "step": 37087 }, { "epoch": 2.91, "learning_rate": 7.091821532188505e-05, "loss": 2.8423, "step": 37088 }, { "epoch": 2.91, "learning_rate": 7.091743119266056e-05, "loss": 2.728, "step": 37089 }, { "epoch": 2.91, "learning_rate": 7.091664706343606e-05, "loss": 2.8569, "step": 37090 }, { "epoch": 2.91, "learning_rate": 7.091586293421156e-05, "loss": 2.8882, "step": 37091 }, { "epoch": 2.91, "learning_rate": 7.091507880498707e-05, "loss": 2.7825, "step": 37092 }, { "epoch": 2.91, "learning_rate": 7.091429467576257e-05, "loss": 2.7731, "step": 37093 }, { "epoch": 2.91, "learning_rate": 7.091351054653807e-05, "loss": 2.8744, "step": 37094 }, { "epoch": 2.91, "learning_rate": 7.091272641731358e-05, "loss": 2.6618, "step": 37095 }, { "epoch": 2.91, "learning_rate": 7.091194228808908e-05, "loss": 2.661, "step": 37096 }, { "epoch": 2.91, "learning_rate": 7.091115815886458e-05, "loss": 2.7704, "step": 37097 }, { "epoch": 2.91, "learning_rate": 7.091037402964009e-05, "loss": 2.7078, "step": 37098 }, { "epoch": 2.91, "learning_rate": 7.090958990041559e-05, "loss": 2.6037, "step": 37099 }, { "epoch": 2.91, "learning_rate": 7.09088057711911e-05, "loss": 2.7653, "step": 37100 }, { "epoch": 2.91, "learning_rate": 7.09080216419666e-05, "loss": 2.9307, "step": 37101 }, { "epoch": 2.91, "learning_rate": 7.09072375127421e-05, "loss": 2.7195, "step": 37102 }, { "epoch": 2.91, "learning_rate": 7.090645338351761e-05, "loss": 2.8005, "step": 37103 }, { "epoch": 2.91, "learning_rate": 7.090566925429311e-05, "loss": 2.8278, "step": 37104 }, { "epoch": 2.91, "learning_rate": 7.090488512506861e-05, "loss": 2.7215, "step": 37105 }, { "epoch": 2.91, "learning_rate": 7.090410099584412e-05, "loss": 2.7592, "step": 37106 }, { "epoch": 2.91, "learning_rate": 7.090331686661962e-05, "loss": 2.6743, "step": 37107 }, { "epoch": 2.91, "learning_rate": 7.090253273739513e-05, "loss": 3.0041, "step": 37108 }, { "epoch": 2.91, "learning_rate": 7.090174860817063e-05, "loss": 2.6874, "step": 37109 }, { "epoch": 2.91, "learning_rate": 7.090096447894613e-05, "loss": 2.7217, "step": 37110 }, { "epoch": 2.91, "learning_rate": 7.090018034972164e-05, "loss": 3.0908, "step": 37111 }, { "epoch": 2.91, "learning_rate": 7.089939622049714e-05, "loss": 2.8788, "step": 37112 }, { "epoch": 2.91, "learning_rate": 7.089861209127264e-05, "loss": 2.8891, "step": 37113 }, { "epoch": 2.91, "learning_rate": 7.089782796204815e-05, "loss": 2.9831, "step": 37114 }, { "epoch": 2.91, "learning_rate": 7.089704383282365e-05, "loss": 2.9082, "step": 37115 }, { "epoch": 2.91, "learning_rate": 7.089625970359915e-05, "loss": 2.6713, "step": 37116 }, { "epoch": 2.91, "learning_rate": 7.089547557437466e-05, "loss": 2.6944, "step": 37117 }, { "epoch": 2.91, "learning_rate": 7.089469144515016e-05, "loss": 2.8353, "step": 37118 }, { "epoch": 2.91, "learning_rate": 7.089390731592567e-05, "loss": 2.9531, "step": 37119 }, { "epoch": 2.91, "learning_rate": 7.089312318670117e-05, "loss": 2.9148, "step": 37120 }, { "epoch": 2.91, "learning_rate": 7.089233905747667e-05, "loss": 2.8487, "step": 37121 }, { "epoch": 2.91, "learning_rate": 7.089155492825218e-05, "loss": 2.64, "step": 37122 }, { "epoch": 2.91, "learning_rate": 7.089077079902768e-05, "loss": 2.6525, "step": 37123 }, { "epoch": 2.91, "learning_rate": 7.088998666980318e-05, "loss": 3.2303, "step": 37124 }, { "epoch": 2.91, "learning_rate": 7.088920254057869e-05, "loss": 2.7944, "step": 37125 }, { "epoch": 2.91, "learning_rate": 7.088841841135419e-05, "loss": 2.6463, "step": 37126 }, { "epoch": 2.91, "learning_rate": 7.08876342821297e-05, "loss": 3.0213, "step": 37127 }, { "epoch": 2.91, "learning_rate": 7.08868501529052e-05, "loss": 2.7241, "step": 37128 }, { "epoch": 2.91, "learning_rate": 7.08860660236807e-05, "loss": 2.8232, "step": 37129 }, { "epoch": 2.91, "learning_rate": 7.08852818944562e-05, "loss": 2.8179, "step": 37130 }, { "epoch": 2.91, "learning_rate": 7.088449776523171e-05, "loss": 2.7594, "step": 37131 }, { "epoch": 2.91, "learning_rate": 7.088371363600723e-05, "loss": 2.6917, "step": 37132 }, { "epoch": 2.91, "learning_rate": 7.088292950678273e-05, "loss": 2.7303, "step": 37133 }, { "epoch": 2.91, "learning_rate": 7.088214537755823e-05, "loss": 2.8888, "step": 37134 }, { "epoch": 2.91, "learning_rate": 7.088136124833374e-05, "loss": 2.8419, "step": 37135 }, { "epoch": 2.91, "learning_rate": 7.088057711910924e-05, "loss": 2.888, "step": 37136 }, { "epoch": 2.91, "learning_rate": 7.087979298988475e-05, "loss": 3.0808, "step": 37137 }, { "epoch": 2.91, "learning_rate": 7.087900886066025e-05, "loss": 2.7622, "step": 37138 }, { "epoch": 2.91, "learning_rate": 7.087822473143575e-05, "loss": 2.7731, "step": 37139 }, { "epoch": 2.91, "learning_rate": 7.087744060221124e-05, "loss": 2.866, "step": 37140 }, { "epoch": 2.91, "learning_rate": 7.087665647298675e-05, "loss": 3.0206, "step": 37141 }, { "epoch": 2.91, "learning_rate": 7.087587234376225e-05, "loss": 2.6792, "step": 37142 }, { "epoch": 2.91, "learning_rate": 7.087508821453775e-05, "loss": 2.9021, "step": 37143 }, { "epoch": 2.91, "learning_rate": 7.087430408531326e-05, "loss": 3.0654, "step": 37144 }, { "epoch": 2.91, "learning_rate": 7.087351995608876e-05, "loss": 2.7748, "step": 37145 }, { "epoch": 2.91, "learning_rate": 7.087273582686427e-05, "loss": 2.7718, "step": 37146 }, { "epoch": 2.91, "learning_rate": 7.087195169763977e-05, "loss": 2.3783, "step": 37147 }, { "epoch": 2.91, "learning_rate": 7.087116756841527e-05, "loss": 2.5623, "step": 37148 }, { "epoch": 2.91, "learning_rate": 7.087038343919078e-05, "loss": 2.9544, "step": 37149 }, { "epoch": 2.91, "learning_rate": 7.086959930996628e-05, "loss": 2.7186, "step": 37150 }, { "epoch": 2.91, "learning_rate": 7.086881518074178e-05, "loss": 2.8542, "step": 37151 }, { "epoch": 2.91, "learning_rate": 7.086803105151729e-05, "loss": 2.9316, "step": 37152 }, { "epoch": 2.91, "learning_rate": 7.086724692229279e-05, "loss": 3.0265, "step": 37153 }, { "epoch": 2.91, "learning_rate": 7.08664627930683e-05, "loss": 2.8453, "step": 37154 }, { "epoch": 2.91, "learning_rate": 7.08656786638438e-05, "loss": 2.5941, "step": 37155 }, { "epoch": 2.91, "learning_rate": 7.08648945346193e-05, "loss": 2.7662, "step": 37156 }, { "epoch": 2.91, "learning_rate": 7.086411040539482e-05, "loss": 2.7365, "step": 37157 }, { "epoch": 2.91, "learning_rate": 7.086332627617032e-05, "loss": 2.6839, "step": 37158 }, { "epoch": 2.91, "learning_rate": 7.086254214694583e-05, "loss": 2.9486, "step": 37159 }, { "epoch": 2.91, "learning_rate": 7.086175801772133e-05, "loss": 2.8077, "step": 37160 }, { "epoch": 2.91, "learning_rate": 7.086097388849683e-05, "loss": 2.8943, "step": 37161 }, { "epoch": 2.91, "learning_rate": 7.086018975927234e-05, "loss": 2.8161, "step": 37162 }, { "epoch": 2.91, "learning_rate": 7.085940563004784e-05, "loss": 2.8142, "step": 37163 }, { "epoch": 2.91, "learning_rate": 7.085862150082334e-05, "loss": 2.9124, "step": 37164 }, { "epoch": 2.91, "learning_rate": 7.085783737159885e-05, "loss": 2.8374, "step": 37165 }, { "epoch": 2.91, "learning_rate": 7.085705324237435e-05, "loss": 2.9137, "step": 37166 }, { "epoch": 2.91, "learning_rate": 7.085626911314986e-05, "loss": 2.7865, "step": 37167 }, { "epoch": 2.91, "learning_rate": 7.085548498392536e-05, "loss": 2.8431, "step": 37168 }, { "epoch": 2.91, "learning_rate": 7.085470085470086e-05, "loss": 2.7496, "step": 37169 }, { "epoch": 2.91, "learning_rate": 7.085391672547637e-05, "loss": 2.7369, "step": 37170 }, { "epoch": 2.91, "learning_rate": 7.085313259625187e-05, "loss": 2.9045, "step": 37171 }, { "epoch": 2.91, "learning_rate": 7.085234846702737e-05, "loss": 2.8139, "step": 37172 }, { "epoch": 2.91, "learning_rate": 7.085156433780288e-05, "loss": 2.7856, "step": 37173 }, { "epoch": 2.91, "learning_rate": 7.085078020857838e-05, "loss": 2.9203, "step": 37174 }, { "epoch": 2.92, "learning_rate": 7.084999607935389e-05, "loss": 3.024, "step": 37175 }, { "epoch": 2.92, "learning_rate": 7.084921195012938e-05, "loss": 2.7717, "step": 37176 }, { "epoch": 2.92, "learning_rate": 7.084842782090488e-05, "loss": 2.5082, "step": 37177 }, { "epoch": 2.92, "learning_rate": 7.084764369168038e-05, "loss": 2.9154, "step": 37178 }, { "epoch": 2.92, "learning_rate": 7.084685956245589e-05, "loss": 3.0035, "step": 37179 }, { "epoch": 2.92, "learning_rate": 7.084607543323139e-05, "loss": 2.6447, "step": 37180 }, { "epoch": 2.92, "learning_rate": 7.08452913040069e-05, "loss": 2.7551, "step": 37181 }, { "epoch": 2.92, "learning_rate": 7.08445071747824e-05, "loss": 2.7562, "step": 37182 }, { "epoch": 2.92, "learning_rate": 7.084372304555791e-05, "loss": 2.7544, "step": 37183 }, { "epoch": 2.92, "learning_rate": 7.084293891633342e-05, "loss": 2.8526, "step": 37184 }, { "epoch": 2.92, "learning_rate": 7.084215478710892e-05, "loss": 2.7456, "step": 37185 }, { "epoch": 2.92, "learning_rate": 7.084137065788443e-05, "loss": 2.9173, "step": 37186 }, { "epoch": 2.92, "learning_rate": 7.084058652865993e-05, "loss": 2.6524, "step": 37187 }, { "epoch": 2.92, "learning_rate": 7.083980239943543e-05, "loss": 3.0269, "step": 37188 }, { "epoch": 2.92, "learning_rate": 7.083901827021094e-05, "loss": 2.6984, "step": 37189 }, { "epoch": 2.92, "learning_rate": 7.083823414098644e-05, "loss": 2.889, "step": 37190 }, { "epoch": 2.92, "learning_rate": 7.083745001176194e-05, "loss": 3.1317, "step": 37191 }, { "epoch": 2.92, "learning_rate": 7.083666588253745e-05, "loss": 2.7873, "step": 37192 }, { "epoch": 2.92, "learning_rate": 7.083588175331295e-05, "loss": 2.7294, "step": 37193 }, { "epoch": 2.92, "learning_rate": 7.083509762408846e-05, "loss": 2.7063, "step": 37194 }, { "epoch": 2.92, "learning_rate": 7.083431349486396e-05, "loss": 2.7765, "step": 37195 }, { "epoch": 2.92, "learning_rate": 7.083352936563946e-05, "loss": 2.8639, "step": 37196 }, { "epoch": 2.92, "learning_rate": 7.083274523641497e-05, "loss": 2.6889, "step": 37197 }, { "epoch": 2.92, "learning_rate": 7.083196110719047e-05, "loss": 2.8135, "step": 37198 }, { "epoch": 2.92, "learning_rate": 7.083117697796597e-05, "loss": 2.7969, "step": 37199 }, { "epoch": 2.92, "learning_rate": 7.083039284874148e-05, "loss": 2.8749, "step": 37200 }, { "epoch": 2.92, "learning_rate": 7.082960871951698e-05, "loss": 2.9465, "step": 37201 }, { "epoch": 2.92, "learning_rate": 7.082882459029248e-05, "loss": 3.0297, "step": 37202 }, { "epoch": 2.92, "learning_rate": 7.082804046106799e-05, "loss": 2.7571, "step": 37203 }, { "epoch": 2.92, "learning_rate": 7.082725633184349e-05, "loss": 2.7944, "step": 37204 }, { "epoch": 2.92, "learning_rate": 7.0826472202619e-05, "loss": 2.9433, "step": 37205 }, { "epoch": 2.92, "learning_rate": 7.08256880733945e-05, "loss": 2.8506, "step": 37206 }, { "epoch": 2.92, "learning_rate": 7.082490394417e-05, "loss": 3.0648, "step": 37207 }, { "epoch": 2.92, "learning_rate": 7.08241198149455e-05, "loss": 3.1077, "step": 37208 }, { "epoch": 2.92, "learning_rate": 7.082333568572101e-05, "loss": 3.0824, "step": 37209 }, { "epoch": 2.92, "learning_rate": 7.082255155649651e-05, "loss": 2.9213, "step": 37210 }, { "epoch": 2.92, "learning_rate": 7.082176742727202e-05, "loss": 2.559, "step": 37211 }, { "epoch": 2.92, "learning_rate": 7.082098329804752e-05, "loss": 2.8585, "step": 37212 }, { "epoch": 2.92, "learning_rate": 7.082019916882303e-05, "loss": 2.8598, "step": 37213 }, { "epoch": 2.92, "learning_rate": 7.081941503959853e-05, "loss": 2.9675, "step": 37214 }, { "epoch": 2.92, "learning_rate": 7.081863091037403e-05, "loss": 2.7505, "step": 37215 }, { "epoch": 2.92, "learning_rate": 7.081784678114954e-05, "loss": 2.5917, "step": 37216 }, { "epoch": 2.92, "learning_rate": 7.081706265192504e-05, "loss": 2.8669, "step": 37217 }, { "epoch": 2.92, "learning_rate": 7.081627852270054e-05, "loss": 2.9449, "step": 37218 }, { "epoch": 2.92, "learning_rate": 7.081549439347605e-05, "loss": 2.9068, "step": 37219 }, { "epoch": 2.92, "learning_rate": 7.081471026425155e-05, "loss": 2.986, "step": 37220 }, { "epoch": 2.92, "learning_rate": 7.081392613502705e-05, "loss": 2.8363, "step": 37221 }, { "epoch": 2.92, "learning_rate": 7.081314200580256e-05, "loss": 2.6822, "step": 37222 }, { "epoch": 2.92, "learning_rate": 7.081235787657806e-05, "loss": 2.6788, "step": 37223 }, { "epoch": 2.92, "learning_rate": 7.081157374735357e-05, "loss": 2.6679, "step": 37224 }, { "epoch": 2.92, "learning_rate": 7.081078961812907e-05, "loss": 2.7616, "step": 37225 }, { "epoch": 2.92, "learning_rate": 7.081000548890457e-05, "loss": 2.6554, "step": 37226 }, { "epoch": 2.92, "learning_rate": 7.080922135968008e-05, "loss": 2.6204, "step": 37227 }, { "epoch": 2.92, "learning_rate": 7.080843723045558e-05, "loss": 2.8203, "step": 37228 }, { "epoch": 2.92, "learning_rate": 7.080765310123108e-05, "loss": 2.5193, "step": 37229 }, { "epoch": 2.92, "learning_rate": 7.080686897200659e-05, "loss": 2.8202, "step": 37230 }, { "epoch": 2.92, "learning_rate": 7.080608484278209e-05, "loss": 2.6339, "step": 37231 }, { "epoch": 2.92, "learning_rate": 7.08053007135576e-05, "loss": 2.8829, "step": 37232 }, { "epoch": 2.92, "learning_rate": 7.08045165843331e-05, "loss": 2.8266, "step": 37233 }, { "epoch": 2.92, "learning_rate": 7.08037324551086e-05, "loss": 2.6287, "step": 37234 }, { "epoch": 2.92, "learning_rate": 7.08029483258841e-05, "loss": 2.7709, "step": 37235 }, { "epoch": 2.92, "learning_rate": 7.080216419665961e-05, "loss": 2.7428, "step": 37236 }, { "epoch": 2.92, "learning_rate": 7.080138006743511e-05, "loss": 2.8834, "step": 37237 }, { "epoch": 2.92, "learning_rate": 7.080059593821062e-05, "loss": 3.0256, "step": 37238 }, { "epoch": 2.92, "learning_rate": 7.079981180898612e-05, "loss": 2.642, "step": 37239 }, { "epoch": 2.92, "learning_rate": 7.079902767976164e-05, "loss": 2.8539, "step": 37240 }, { "epoch": 2.92, "learning_rate": 7.079824355053714e-05, "loss": 2.9396, "step": 37241 }, { "epoch": 2.92, "learning_rate": 7.079745942131265e-05, "loss": 2.942, "step": 37242 }, { "epoch": 2.92, "learning_rate": 7.079667529208815e-05, "loss": 2.6628, "step": 37243 }, { "epoch": 2.92, "learning_rate": 7.079589116286365e-05, "loss": 2.8398, "step": 37244 }, { "epoch": 2.92, "learning_rate": 7.079510703363916e-05, "loss": 2.7727, "step": 37245 }, { "epoch": 2.92, "learning_rate": 7.079432290441466e-05, "loss": 2.7781, "step": 37246 }, { "epoch": 2.92, "learning_rate": 7.079353877519016e-05, "loss": 2.8982, "step": 37247 }, { "epoch": 2.92, "learning_rate": 7.079275464596565e-05, "loss": 2.7432, "step": 37248 }, { "epoch": 2.92, "learning_rate": 7.079197051674116e-05, "loss": 2.6843, "step": 37249 }, { "epoch": 2.92, "learning_rate": 7.079118638751666e-05, "loss": 2.7375, "step": 37250 }, { "epoch": 2.92, "learning_rate": 7.079040225829216e-05, "loss": 2.5591, "step": 37251 }, { "epoch": 2.92, "learning_rate": 7.078961812906767e-05, "loss": 2.9136, "step": 37252 }, { "epoch": 2.92, "learning_rate": 7.078883399984317e-05, "loss": 2.5828, "step": 37253 }, { "epoch": 2.92, "learning_rate": 7.078804987061868e-05, "loss": 2.8893, "step": 37254 }, { "epoch": 2.92, "learning_rate": 7.078726574139418e-05, "loss": 2.695, "step": 37255 }, { "epoch": 2.92, "learning_rate": 7.078648161216968e-05, "loss": 2.7385, "step": 37256 }, { "epoch": 2.92, "learning_rate": 7.078569748294519e-05, "loss": 2.8853, "step": 37257 }, { "epoch": 2.92, "learning_rate": 7.078491335372069e-05, "loss": 2.6982, "step": 37258 }, { "epoch": 2.92, "learning_rate": 7.07841292244962e-05, "loss": 2.8355, "step": 37259 }, { "epoch": 2.92, "learning_rate": 7.07833450952717e-05, "loss": 2.613, "step": 37260 }, { "epoch": 2.92, "learning_rate": 7.07825609660472e-05, "loss": 2.7625, "step": 37261 }, { "epoch": 2.92, "learning_rate": 7.07817768368227e-05, "loss": 2.8897, "step": 37262 }, { "epoch": 2.92, "learning_rate": 7.078099270759821e-05, "loss": 2.8745, "step": 37263 }, { "epoch": 2.92, "learning_rate": 7.078020857837371e-05, "loss": 2.7109, "step": 37264 }, { "epoch": 2.92, "learning_rate": 7.077942444914923e-05, "loss": 2.8585, "step": 37265 }, { "epoch": 2.92, "learning_rate": 7.077864031992473e-05, "loss": 2.8198, "step": 37266 }, { "epoch": 2.92, "learning_rate": 7.077785619070024e-05, "loss": 2.9354, "step": 37267 }, { "epoch": 2.92, "learning_rate": 7.077707206147574e-05, "loss": 2.9149, "step": 37268 }, { "epoch": 2.92, "learning_rate": 7.077628793225124e-05, "loss": 2.8852, "step": 37269 }, { "epoch": 2.92, "learning_rate": 7.077550380302675e-05, "loss": 2.7994, "step": 37270 }, { "epoch": 2.92, "learning_rate": 7.077471967380225e-05, "loss": 2.7837, "step": 37271 }, { "epoch": 2.92, "learning_rate": 7.077393554457776e-05, "loss": 2.6671, "step": 37272 }, { "epoch": 2.92, "learning_rate": 7.077315141535326e-05, "loss": 2.7633, "step": 37273 }, { "epoch": 2.92, "learning_rate": 7.077236728612876e-05, "loss": 2.9184, "step": 37274 }, { "epoch": 2.92, "learning_rate": 7.077158315690427e-05, "loss": 2.745, "step": 37275 }, { "epoch": 2.92, "learning_rate": 7.077079902767977e-05, "loss": 2.73, "step": 37276 }, { "epoch": 2.92, "learning_rate": 7.077001489845527e-05, "loss": 2.7705, "step": 37277 }, { "epoch": 2.92, "learning_rate": 7.076923076923078e-05, "loss": 2.8864, "step": 37278 }, { "epoch": 2.92, "learning_rate": 7.076844664000628e-05, "loss": 2.8233, "step": 37279 }, { "epoch": 2.92, "learning_rate": 7.076766251078178e-05, "loss": 2.8075, "step": 37280 }, { "epoch": 2.92, "learning_rate": 7.076687838155729e-05, "loss": 3.0429, "step": 37281 }, { "epoch": 2.92, "learning_rate": 7.076609425233279e-05, "loss": 3.0627, "step": 37282 }, { "epoch": 2.92, "learning_rate": 7.07653101231083e-05, "loss": 2.7197, "step": 37283 }, { "epoch": 2.92, "learning_rate": 7.076452599388379e-05, "loss": 2.9437, "step": 37284 }, { "epoch": 2.92, "learning_rate": 7.076374186465929e-05, "loss": 2.7851, "step": 37285 }, { "epoch": 2.92, "learning_rate": 7.07629577354348e-05, "loss": 2.8905, "step": 37286 }, { "epoch": 2.92, "learning_rate": 7.07621736062103e-05, "loss": 2.9417, "step": 37287 }, { "epoch": 2.92, "learning_rate": 7.07613894769858e-05, "loss": 2.8212, "step": 37288 }, { "epoch": 2.92, "learning_rate": 7.07606053477613e-05, "loss": 2.8054, "step": 37289 }, { "epoch": 2.92, "learning_rate": 7.075982121853681e-05, "loss": 2.8702, "step": 37290 }, { "epoch": 2.92, "learning_rate": 7.075903708931233e-05, "loss": 2.6527, "step": 37291 }, { "epoch": 2.92, "learning_rate": 7.075825296008783e-05, "loss": 2.9754, "step": 37292 }, { "epoch": 2.92, "learning_rate": 7.075746883086333e-05, "loss": 3.1938, "step": 37293 }, { "epoch": 2.92, "learning_rate": 7.075668470163884e-05, "loss": 2.7444, "step": 37294 }, { "epoch": 2.92, "learning_rate": 7.075590057241434e-05, "loss": 2.9646, "step": 37295 }, { "epoch": 2.92, "learning_rate": 7.075511644318984e-05, "loss": 2.8841, "step": 37296 }, { "epoch": 2.92, "learning_rate": 7.075433231396535e-05, "loss": 2.89, "step": 37297 }, { "epoch": 2.92, "learning_rate": 7.075354818474085e-05, "loss": 2.814, "step": 37298 }, { "epoch": 2.92, "learning_rate": 7.075276405551635e-05, "loss": 2.845, "step": 37299 }, { "epoch": 2.92, "learning_rate": 7.075197992629186e-05, "loss": 2.7594, "step": 37300 }, { "epoch": 2.92, "learning_rate": 7.075119579706736e-05, "loss": 2.5525, "step": 37301 }, { "epoch": 2.92, "learning_rate": 7.075041166784287e-05, "loss": 2.6661, "step": 37302 }, { "epoch": 2.93, "learning_rate": 7.074962753861837e-05, "loss": 2.7544, "step": 37303 }, { "epoch": 2.93, "learning_rate": 7.074884340939387e-05, "loss": 2.8526, "step": 37304 }, { "epoch": 2.93, "learning_rate": 7.074805928016938e-05, "loss": 2.6112, "step": 37305 }, { "epoch": 2.93, "learning_rate": 7.074727515094488e-05, "loss": 2.8076, "step": 37306 }, { "epoch": 2.93, "learning_rate": 7.074649102172038e-05, "loss": 2.8325, "step": 37307 }, { "epoch": 2.93, "learning_rate": 7.074570689249589e-05, "loss": 2.8363, "step": 37308 }, { "epoch": 2.93, "learning_rate": 7.074492276327139e-05, "loss": 2.6792, "step": 37309 }, { "epoch": 2.93, "learning_rate": 7.07441386340469e-05, "loss": 2.79, "step": 37310 }, { "epoch": 2.93, "learning_rate": 7.07433545048224e-05, "loss": 3.0059, "step": 37311 }, { "epoch": 2.93, "learning_rate": 7.07425703755979e-05, "loss": 3.0419, "step": 37312 }, { "epoch": 2.93, "learning_rate": 7.07417862463734e-05, "loss": 2.7767, "step": 37313 }, { "epoch": 2.93, "learning_rate": 7.074100211714891e-05, "loss": 2.8728, "step": 37314 }, { "epoch": 2.93, "learning_rate": 7.074021798792441e-05, "loss": 2.7623, "step": 37315 }, { "epoch": 2.93, "learning_rate": 7.073943385869992e-05, "loss": 2.5861, "step": 37316 }, { "epoch": 2.93, "learning_rate": 7.073864972947542e-05, "loss": 2.5989, "step": 37317 }, { "epoch": 2.93, "learning_rate": 7.073786560025092e-05, "loss": 2.9916, "step": 37318 }, { "epoch": 2.93, "learning_rate": 7.073708147102643e-05, "loss": 2.6822, "step": 37319 }, { "epoch": 2.93, "learning_rate": 7.073629734180193e-05, "loss": 2.7659, "step": 37320 }, { "epoch": 2.93, "learning_rate": 7.073551321257744e-05, "loss": 2.8046, "step": 37321 }, { "epoch": 2.93, "learning_rate": 7.073472908335294e-05, "loss": 2.7809, "step": 37322 }, { "epoch": 2.93, "learning_rate": 7.073394495412844e-05, "loss": 2.8094, "step": 37323 }, { "epoch": 2.93, "learning_rate": 7.073316082490395e-05, "loss": 2.897, "step": 37324 }, { "epoch": 2.93, "learning_rate": 7.073237669567945e-05, "loss": 2.7132, "step": 37325 }, { "epoch": 2.93, "learning_rate": 7.073159256645495e-05, "loss": 2.8821, "step": 37326 }, { "epoch": 2.93, "learning_rate": 7.073080843723046e-05, "loss": 2.8463, "step": 37327 }, { "epoch": 2.93, "learning_rate": 7.073002430800596e-05, "loss": 2.6732, "step": 37328 }, { "epoch": 2.93, "learning_rate": 7.072924017878147e-05, "loss": 3.0138, "step": 37329 }, { "epoch": 2.93, "learning_rate": 7.072845604955697e-05, "loss": 2.7651, "step": 37330 }, { "epoch": 2.93, "learning_rate": 7.072767192033247e-05, "loss": 2.8595, "step": 37331 }, { "epoch": 2.93, "learning_rate": 7.072688779110798e-05, "loss": 2.8442, "step": 37332 }, { "epoch": 2.93, "learning_rate": 7.072610366188348e-05, "loss": 2.8682, "step": 37333 }, { "epoch": 2.93, "learning_rate": 7.072531953265898e-05, "loss": 2.7337, "step": 37334 }, { "epoch": 2.93, "learning_rate": 7.072453540343449e-05, "loss": 2.7034, "step": 37335 }, { "epoch": 2.93, "learning_rate": 7.072375127420999e-05, "loss": 2.8444, "step": 37336 }, { "epoch": 2.93, "learning_rate": 7.07229671449855e-05, "loss": 3.0125, "step": 37337 }, { "epoch": 2.93, "learning_rate": 7.0722183015761e-05, "loss": 2.841, "step": 37338 }, { "epoch": 2.93, "learning_rate": 7.07213988865365e-05, "loss": 2.6319, "step": 37339 }, { "epoch": 2.93, "learning_rate": 7.0720614757312e-05, "loss": 2.8368, "step": 37340 }, { "epoch": 2.93, "learning_rate": 7.071983062808751e-05, "loss": 2.7516, "step": 37341 }, { "epoch": 2.93, "learning_rate": 7.071904649886301e-05, "loss": 2.8804, "step": 37342 }, { "epoch": 2.93, "learning_rate": 7.071826236963852e-05, "loss": 2.7713, "step": 37343 }, { "epoch": 2.93, "learning_rate": 7.071747824041402e-05, "loss": 2.9406, "step": 37344 }, { "epoch": 2.93, "learning_rate": 7.071669411118952e-05, "loss": 2.9447, "step": 37345 }, { "epoch": 2.93, "learning_rate": 7.071590998196503e-05, "loss": 2.8096, "step": 37346 }, { "epoch": 2.93, "learning_rate": 7.071512585274053e-05, "loss": 2.7807, "step": 37347 }, { "epoch": 2.93, "learning_rate": 7.071434172351605e-05, "loss": 2.7934, "step": 37348 }, { "epoch": 2.93, "learning_rate": 7.071355759429155e-05, "loss": 2.7503, "step": 37349 }, { "epoch": 2.93, "learning_rate": 7.071277346506706e-05, "loss": 2.8589, "step": 37350 }, { "epoch": 2.93, "learning_rate": 7.071198933584256e-05, "loss": 2.6223, "step": 37351 }, { "epoch": 2.93, "learning_rate": 7.071120520661806e-05, "loss": 2.9905, "step": 37352 }, { "epoch": 2.93, "learning_rate": 7.071042107739357e-05, "loss": 2.8455, "step": 37353 }, { "epoch": 2.93, "learning_rate": 7.070963694816907e-05, "loss": 2.8371, "step": 37354 }, { "epoch": 2.93, "learning_rate": 7.070885281894457e-05, "loss": 2.8591, "step": 37355 }, { "epoch": 2.93, "learning_rate": 7.070806868972006e-05, "loss": 2.6463, "step": 37356 }, { "epoch": 2.93, "learning_rate": 7.070728456049557e-05, "loss": 2.9182, "step": 37357 }, { "epoch": 2.93, "learning_rate": 7.070650043127107e-05, "loss": 2.9594, "step": 37358 }, { "epoch": 2.93, "learning_rate": 7.070571630204658e-05, "loss": 2.6873, "step": 37359 }, { "epoch": 2.93, "learning_rate": 7.070493217282208e-05, "loss": 2.7799, "step": 37360 }, { "epoch": 2.93, "learning_rate": 7.070414804359758e-05, "loss": 2.8504, "step": 37361 }, { "epoch": 2.93, "learning_rate": 7.070336391437309e-05, "loss": 2.8864, "step": 37362 }, { "epoch": 2.93, "learning_rate": 7.070257978514859e-05, "loss": 2.705, "step": 37363 }, { "epoch": 2.93, "learning_rate": 7.07017956559241e-05, "loss": 2.7791, "step": 37364 }, { "epoch": 2.93, "learning_rate": 7.07010115266996e-05, "loss": 2.8756, "step": 37365 }, { "epoch": 2.93, "learning_rate": 7.07002273974751e-05, "loss": 2.9778, "step": 37366 }, { "epoch": 2.93, "learning_rate": 7.06994432682506e-05, "loss": 2.6339, "step": 37367 }, { "epoch": 2.93, "learning_rate": 7.069865913902611e-05, "loss": 2.8709, "step": 37368 }, { "epoch": 2.93, "learning_rate": 7.069787500980161e-05, "loss": 2.7044, "step": 37369 }, { "epoch": 2.93, "learning_rate": 7.069709088057712e-05, "loss": 2.839, "step": 37370 }, { "epoch": 2.93, "learning_rate": 7.069630675135262e-05, "loss": 2.8456, "step": 37371 }, { "epoch": 2.93, "learning_rate": 7.069552262212812e-05, "loss": 2.7382, "step": 37372 }, { "epoch": 2.93, "learning_rate": 7.069473849290364e-05, "loss": 2.9605, "step": 37373 }, { "epoch": 2.93, "learning_rate": 7.069395436367914e-05, "loss": 2.8995, "step": 37374 }, { "epoch": 2.93, "learning_rate": 7.069317023445465e-05, "loss": 2.8157, "step": 37375 }, { "epoch": 2.93, "learning_rate": 7.069238610523015e-05, "loss": 2.692, "step": 37376 }, { "epoch": 2.93, "learning_rate": 7.069160197600566e-05, "loss": 2.9874, "step": 37377 }, { "epoch": 2.93, "learning_rate": 7.069081784678116e-05, "loss": 2.6972, "step": 37378 }, { "epoch": 2.93, "learning_rate": 7.069003371755666e-05, "loss": 2.6546, "step": 37379 }, { "epoch": 2.93, "learning_rate": 7.068924958833217e-05, "loss": 2.8271, "step": 37380 }, { "epoch": 2.93, "learning_rate": 7.068846545910767e-05, "loss": 3.1038, "step": 37381 }, { "epoch": 2.93, "learning_rate": 7.068768132988317e-05, "loss": 2.7422, "step": 37382 }, { "epoch": 2.93, "learning_rate": 7.068689720065868e-05, "loss": 2.8972, "step": 37383 }, { "epoch": 2.93, "learning_rate": 7.068611307143418e-05, "loss": 2.6744, "step": 37384 }, { "epoch": 2.93, "learning_rate": 7.068532894220968e-05, "loss": 2.6851, "step": 37385 }, { "epoch": 2.93, "learning_rate": 7.068454481298519e-05, "loss": 2.7357, "step": 37386 }, { "epoch": 2.93, "learning_rate": 7.068376068376069e-05, "loss": 2.7433, "step": 37387 }, { "epoch": 2.93, "learning_rate": 7.06829765545362e-05, "loss": 2.9379, "step": 37388 }, { "epoch": 2.93, "learning_rate": 7.06821924253117e-05, "loss": 2.8526, "step": 37389 }, { "epoch": 2.93, "learning_rate": 7.06814082960872e-05, "loss": 2.8344, "step": 37390 }, { "epoch": 2.93, "learning_rate": 7.06806241668627e-05, "loss": 2.5841, "step": 37391 }, { "epoch": 2.93, "learning_rate": 7.06798400376382e-05, "loss": 2.7984, "step": 37392 }, { "epoch": 2.93, "learning_rate": 7.06790559084137e-05, "loss": 2.8991, "step": 37393 }, { "epoch": 2.93, "learning_rate": 7.06782717791892e-05, "loss": 2.8928, "step": 37394 }, { "epoch": 2.93, "learning_rate": 7.067748764996471e-05, "loss": 2.8984, "step": 37395 }, { "epoch": 2.93, "learning_rate": 7.067670352074021e-05, "loss": 2.7916, "step": 37396 }, { "epoch": 2.93, "learning_rate": 7.067591939151572e-05, "loss": 2.8365, "step": 37397 }, { "epoch": 2.93, "learning_rate": 7.067513526229122e-05, "loss": 2.7497, "step": 37398 }, { "epoch": 2.93, "learning_rate": 7.067435113306674e-05, "loss": 3.2133, "step": 37399 }, { "epoch": 2.93, "learning_rate": 7.067356700384224e-05, "loss": 3.0348, "step": 37400 }, { "epoch": 2.93, "learning_rate": 7.067278287461774e-05, "loss": 2.7329, "step": 37401 }, { "epoch": 2.93, "learning_rate": 7.067199874539325e-05, "loss": 2.8292, "step": 37402 }, { "epoch": 2.93, "learning_rate": 7.067121461616875e-05, "loss": 2.7835, "step": 37403 }, { "epoch": 2.93, "learning_rate": 7.067043048694425e-05, "loss": 2.9256, "step": 37404 }, { "epoch": 2.93, "learning_rate": 7.066964635771976e-05, "loss": 2.9823, "step": 37405 }, { "epoch": 2.93, "learning_rate": 7.066886222849526e-05, "loss": 2.7944, "step": 37406 }, { "epoch": 2.93, "learning_rate": 7.066807809927077e-05, "loss": 2.7121, "step": 37407 }, { "epoch": 2.93, "learning_rate": 7.066729397004627e-05, "loss": 2.8069, "step": 37408 }, { "epoch": 2.93, "learning_rate": 7.066650984082177e-05, "loss": 2.8609, "step": 37409 }, { "epoch": 2.93, "learning_rate": 7.066572571159728e-05, "loss": 2.9594, "step": 37410 }, { "epoch": 2.93, "learning_rate": 7.066494158237278e-05, "loss": 2.8348, "step": 37411 }, { "epoch": 2.93, "learning_rate": 7.066415745314828e-05, "loss": 2.7174, "step": 37412 }, { "epoch": 2.93, "learning_rate": 7.066337332392379e-05, "loss": 2.868, "step": 37413 }, { "epoch": 2.93, "learning_rate": 7.066258919469929e-05, "loss": 2.876, "step": 37414 }, { "epoch": 2.93, "learning_rate": 7.06618050654748e-05, "loss": 2.7784, "step": 37415 }, { "epoch": 2.93, "learning_rate": 7.06610209362503e-05, "loss": 2.781, "step": 37416 }, { "epoch": 2.93, "learning_rate": 7.06602368070258e-05, "loss": 2.8697, "step": 37417 }, { "epoch": 2.93, "learning_rate": 7.06594526778013e-05, "loss": 3.0673, "step": 37418 }, { "epoch": 2.93, "learning_rate": 7.065866854857681e-05, "loss": 2.6606, "step": 37419 }, { "epoch": 2.93, "learning_rate": 7.065788441935231e-05, "loss": 2.918, "step": 37420 }, { "epoch": 2.93, "learning_rate": 7.065710029012782e-05, "loss": 2.7927, "step": 37421 }, { "epoch": 2.93, "learning_rate": 7.065631616090332e-05, "loss": 2.7941, "step": 37422 }, { "epoch": 2.93, "learning_rate": 7.065553203167882e-05, "loss": 2.8967, "step": 37423 }, { "epoch": 2.93, "learning_rate": 7.065474790245433e-05, "loss": 2.9829, "step": 37424 }, { "epoch": 2.93, "learning_rate": 7.065396377322983e-05, "loss": 3.0781, "step": 37425 }, { "epoch": 2.93, "learning_rate": 7.065317964400534e-05, "loss": 2.6066, "step": 37426 }, { "epoch": 2.93, "learning_rate": 7.065239551478084e-05, "loss": 2.6371, "step": 37427 }, { "epoch": 2.93, "learning_rate": 7.065161138555634e-05, "loss": 2.6371, "step": 37428 }, { "epoch": 2.93, "learning_rate": 7.065082725633185e-05, "loss": 2.8554, "step": 37429 }, { "epoch": 2.93, "learning_rate": 7.065004312710735e-05, "loss": 2.6951, "step": 37430 }, { "epoch": 2.94, "learning_rate": 7.064925899788285e-05, "loss": 2.69, "step": 37431 }, { "epoch": 2.94, "learning_rate": 7.064847486865836e-05, "loss": 2.8457, "step": 37432 }, { "epoch": 2.94, "learning_rate": 7.064769073943386e-05, "loss": 2.6753, "step": 37433 }, { "epoch": 2.94, "learning_rate": 7.064690661020936e-05, "loss": 2.6049, "step": 37434 }, { "epoch": 2.94, "learning_rate": 7.064612248098487e-05, "loss": 2.8428, "step": 37435 }, { "epoch": 2.94, "learning_rate": 7.064533835176037e-05, "loss": 2.8169, "step": 37436 }, { "epoch": 2.94, "learning_rate": 7.064455422253588e-05, "loss": 2.8573, "step": 37437 }, { "epoch": 2.94, "learning_rate": 7.064377009331138e-05, "loss": 2.7232, "step": 37438 }, { "epoch": 2.94, "learning_rate": 7.064298596408688e-05, "loss": 2.8093, "step": 37439 }, { "epoch": 2.94, "learning_rate": 7.064220183486239e-05, "loss": 2.6894, "step": 37440 }, { "epoch": 2.94, "learning_rate": 7.064141770563789e-05, "loss": 2.7743, "step": 37441 }, { "epoch": 2.94, "learning_rate": 7.06406335764134e-05, "loss": 2.8536, "step": 37442 }, { "epoch": 2.94, "learning_rate": 7.06398494471889e-05, "loss": 2.7083, "step": 37443 }, { "epoch": 2.94, "learning_rate": 7.06390653179644e-05, "loss": 2.6823, "step": 37444 }, { "epoch": 2.94, "learning_rate": 7.06382811887399e-05, "loss": 2.7106, "step": 37445 }, { "epoch": 2.94, "learning_rate": 7.063749705951541e-05, "loss": 2.8586, "step": 37446 }, { "epoch": 2.94, "learning_rate": 7.063671293029091e-05, "loss": 2.792, "step": 37447 }, { "epoch": 2.94, "learning_rate": 7.063592880106642e-05, "loss": 2.7853, "step": 37448 }, { "epoch": 2.94, "learning_rate": 7.063514467184192e-05, "loss": 2.7433, "step": 37449 }, { "epoch": 2.94, "learning_rate": 7.063436054261742e-05, "loss": 2.6546, "step": 37450 }, { "epoch": 2.94, "learning_rate": 7.063357641339293e-05, "loss": 2.7136, "step": 37451 }, { "epoch": 2.94, "learning_rate": 7.063279228416843e-05, "loss": 2.8623, "step": 37452 }, { "epoch": 2.94, "learning_rate": 7.063200815494393e-05, "loss": 2.7912, "step": 37453 }, { "epoch": 2.94, "learning_rate": 7.063122402571944e-05, "loss": 2.6482, "step": 37454 }, { "epoch": 2.94, "learning_rate": 7.063043989649494e-05, "loss": 3.398, "step": 37455 }, { "epoch": 2.94, "learning_rate": 7.062965576727046e-05, "loss": 2.7873, "step": 37456 }, { "epoch": 2.94, "learning_rate": 7.062887163804596e-05, "loss": 2.757, "step": 37457 }, { "epoch": 2.94, "learning_rate": 7.062808750882147e-05, "loss": 2.7015, "step": 37458 }, { "epoch": 2.94, "learning_rate": 7.062730337959697e-05, "loss": 2.9905, "step": 37459 }, { "epoch": 2.94, "learning_rate": 7.062651925037247e-05, "loss": 2.9344, "step": 37460 }, { "epoch": 2.94, "learning_rate": 7.062573512114798e-05, "loss": 2.9075, "step": 37461 }, { "epoch": 2.94, "learning_rate": 7.062495099192348e-05, "loss": 3.0173, "step": 37462 }, { "epoch": 2.94, "learning_rate": 7.062416686269899e-05, "loss": 2.8233, "step": 37463 }, { "epoch": 2.94, "learning_rate": 7.062338273347448e-05, "loss": 2.5655, "step": 37464 }, { "epoch": 2.94, "learning_rate": 7.062259860424998e-05, "loss": 2.7704, "step": 37465 }, { "epoch": 2.94, "learning_rate": 7.062181447502548e-05, "loss": 2.8406, "step": 37466 }, { "epoch": 2.94, "learning_rate": 7.062103034580099e-05, "loss": 2.7883, "step": 37467 }, { "epoch": 2.94, "learning_rate": 7.062024621657649e-05, "loss": 2.5836, "step": 37468 }, { "epoch": 2.94, "learning_rate": 7.0619462087352e-05, "loss": 2.9167, "step": 37469 }, { "epoch": 2.94, "learning_rate": 7.06186779581275e-05, "loss": 2.6196, "step": 37470 }, { "epoch": 2.94, "learning_rate": 7.0617893828903e-05, "loss": 2.6344, "step": 37471 }, { "epoch": 2.94, "learning_rate": 7.06171096996785e-05, "loss": 2.625, "step": 37472 }, { "epoch": 2.94, "learning_rate": 7.061632557045401e-05, "loss": 3.125, "step": 37473 }, { "epoch": 2.94, "learning_rate": 7.061554144122951e-05, "loss": 2.6829, "step": 37474 }, { "epoch": 2.94, "learning_rate": 7.061475731200502e-05, "loss": 2.8364, "step": 37475 }, { "epoch": 2.94, "learning_rate": 7.061397318278052e-05, "loss": 2.9833, "step": 37476 }, { "epoch": 2.94, "learning_rate": 7.061318905355602e-05, "loss": 2.77, "step": 37477 }, { "epoch": 2.94, "learning_rate": 7.061240492433153e-05, "loss": 2.8801, "step": 37478 }, { "epoch": 2.94, "learning_rate": 7.061162079510703e-05, "loss": 2.8111, "step": 37479 }, { "epoch": 2.94, "learning_rate": 7.061083666588253e-05, "loss": 2.7714, "step": 37480 }, { "epoch": 2.94, "learning_rate": 7.061005253665805e-05, "loss": 2.9224, "step": 37481 }, { "epoch": 2.94, "learning_rate": 7.060926840743355e-05, "loss": 2.7484, "step": 37482 }, { "epoch": 2.94, "learning_rate": 7.060848427820906e-05, "loss": 2.7505, "step": 37483 }, { "epoch": 2.94, "learning_rate": 7.060770014898456e-05, "loss": 2.8035, "step": 37484 }, { "epoch": 2.94, "learning_rate": 7.060691601976007e-05, "loss": 2.7384, "step": 37485 }, { "epoch": 2.94, "learning_rate": 7.060613189053557e-05, "loss": 2.7806, "step": 37486 }, { "epoch": 2.94, "learning_rate": 7.060534776131107e-05, "loss": 2.8878, "step": 37487 }, { "epoch": 2.94, "learning_rate": 7.060456363208658e-05, "loss": 2.5745, "step": 37488 }, { "epoch": 2.94, "learning_rate": 7.060377950286208e-05, "loss": 2.7999, "step": 37489 }, { "epoch": 2.94, "learning_rate": 7.060299537363758e-05, "loss": 2.6602, "step": 37490 }, { "epoch": 2.94, "learning_rate": 7.060221124441309e-05, "loss": 2.6703, "step": 37491 }, { "epoch": 2.94, "learning_rate": 7.060142711518859e-05, "loss": 2.4862, "step": 37492 }, { "epoch": 2.94, "learning_rate": 7.06006429859641e-05, "loss": 2.6024, "step": 37493 }, { "epoch": 2.94, "learning_rate": 7.05998588567396e-05, "loss": 2.8405, "step": 37494 }, { "epoch": 2.94, "learning_rate": 7.05990747275151e-05, "loss": 2.9372, "step": 37495 }, { "epoch": 2.94, "learning_rate": 7.05982905982906e-05, "loss": 2.7987, "step": 37496 }, { "epoch": 2.94, "learning_rate": 7.059750646906611e-05, "loss": 2.9943, "step": 37497 }, { "epoch": 2.94, "learning_rate": 7.059672233984161e-05, "loss": 2.8044, "step": 37498 }, { "epoch": 2.94, "learning_rate": 7.059593821061712e-05, "loss": 2.9693, "step": 37499 }, { "epoch": 2.94, "learning_rate": 7.059515408139261e-05, "loss": 2.6824, "step": 37500 }, { "epoch": 2.94, "learning_rate": 7.059436995216811e-05, "loss": 2.7707, "step": 37501 }, { "epoch": 2.94, "learning_rate": 7.059358582294361e-05, "loss": 2.8794, "step": 37502 }, { "epoch": 2.94, "learning_rate": 7.059280169371912e-05, "loss": 2.7896, "step": 37503 }, { "epoch": 2.94, "learning_rate": 7.059201756449462e-05, "loss": 2.7902, "step": 37504 }, { "epoch": 2.94, "learning_rate": 7.059123343527013e-05, "loss": 2.6388, "step": 37505 }, { "epoch": 2.94, "learning_rate": 7.059044930604563e-05, "loss": 2.9814, "step": 37506 }, { "epoch": 2.94, "learning_rate": 7.058966517682115e-05, "loss": 2.856, "step": 37507 }, { "epoch": 2.94, "learning_rate": 7.058888104759665e-05, "loss": 2.8376, "step": 37508 }, { "epoch": 2.94, "learning_rate": 7.058809691837215e-05, "loss": 2.7607, "step": 37509 }, { "epoch": 2.94, "learning_rate": 7.058731278914766e-05, "loss": 2.732, "step": 37510 }, { "epoch": 2.94, "learning_rate": 7.058652865992316e-05, "loss": 2.9836, "step": 37511 }, { "epoch": 2.94, "learning_rate": 7.058574453069867e-05, "loss": 2.8674, "step": 37512 }, { "epoch": 2.94, "learning_rate": 7.058496040147417e-05, "loss": 2.8246, "step": 37513 }, { "epoch": 2.94, "learning_rate": 7.058417627224967e-05, "loss": 2.8991, "step": 37514 }, { "epoch": 2.94, "learning_rate": 7.058339214302518e-05, "loss": 3.0055, "step": 37515 }, { "epoch": 2.94, "learning_rate": 7.058260801380068e-05, "loss": 2.8368, "step": 37516 }, { "epoch": 2.94, "learning_rate": 7.058182388457618e-05, "loss": 2.8105, "step": 37517 }, { "epoch": 2.94, "learning_rate": 7.058103975535169e-05, "loss": 2.759, "step": 37518 }, { "epoch": 2.94, "learning_rate": 7.058025562612719e-05, "loss": 2.6985, "step": 37519 }, { "epoch": 2.94, "learning_rate": 7.05794714969027e-05, "loss": 2.6592, "step": 37520 }, { "epoch": 2.94, "learning_rate": 7.05786873676782e-05, "loss": 2.9094, "step": 37521 }, { "epoch": 2.94, "learning_rate": 7.05779032384537e-05, "loss": 2.9146, "step": 37522 }, { "epoch": 2.94, "learning_rate": 7.05771191092292e-05, "loss": 2.6258, "step": 37523 }, { "epoch": 2.94, "learning_rate": 7.057633498000471e-05, "loss": 2.8487, "step": 37524 }, { "epoch": 2.94, "learning_rate": 7.057555085078021e-05, "loss": 2.7822, "step": 37525 }, { "epoch": 2.94, "learning_rate": 7.057476672155572e-05, "loss": 2.7185, "step": 37526 }, { "epoch": 2.94, "learning_rate": 7.057398259233122e-05, "loss": 2.8712, "step": 37527 }, { "epoch": 2.94, "learning_rate": 7.057319846310672e-05, "loss": 2.8439, "step": 37528 }, { "epoch": 2.94, "learning_rate": 7.057241433388223e-05, "loss": 2.8507, "step": 37529 }, { "epoch": 2.94, "learning_rate": 7.057163020465773e-05, "loss": 2.9906, "step": 37530 }, { "epoch": 2.94, "learning_rate": 7.057084607543324e-05, "loss": 2.8903, "step": 37531 }, { "epoch": 2.94, "learning_rate": 7.057006194620874e-05, "loss": 2.9215, "step": 37532 }, { "epoch": 2.94, "learning_rate": 7.056927781698424e-05, "loss": 2.7738, "step": 37533 }, { "epoch": 2.94, "learning_rate": 7.056849368775975e-05, "loss": 2.8196, "step": 37534 }, { "epoch": 2.94, "learning_rate": 7.056770955853525e-05, "loss": 2.9943, "step": 37535 }, { "epoch": 2.94, "learning_rate": 7.056692542931075e-05, "loss": 2.7343, "step": 37536 }, { "epoch": 2.94, "learning_rate": 7.056614130008626e-05, "loss": 2.6467, "step": 37537 }, { "epoch": 2.94, "learning_rate": 7.056535717086176e-05, "loss": 2.8356, "step": 37538 }, { "epoch": 2.94, "learning_rate": 7.056457304163726e-05, "loss": 3.0897, "step": 37539 }, { "epoch": 2.94, "learning_rate": 7.056378891241277e-05, "loss": 2.8276, "step": 37540 }, { "epoch": 2.94, "learning_rate": 7.056300478318827e-05, "loss": 2.7436, "step": 37541 }, { "epoch": 2.94, "learning_rate": 7.056222065396378e-05, "loss": 3.0351, "step": 37542 }, { "epoch": 2.94, "learning_rate": 7.056143652473928e-05, "loss": 2.8774, "step": 37543 }, { "epoch": 2.94, "learning_rate": 7.056065239551478e-05, "loss": 2.718, "step": 37544 }, { "epoch": 2.94, "learning_rate": 7.055986826629029e-05, "loss": 2.917, "step": 37545 }, { "epoch": 2.94, "learning_rate": 7.055908413706579e-05, "loss": 2.9087, "step": 37546 }, { "epoch": 2.94, "learning_rate": 7.05583000078413e-05, "loss": 2.53, "step": 37547 }, { "epoch": 2.94, "learning_rate": 7.05575158786168e-05, "loss": 2.7139, "step": 37548 }, { "epoch": 2.94, "learning_rate": 7.05567317493923e-05, "loss": 2.7745, "step": 37549 }, { "epoch": 2.94, "learning_rate": 7.05559476201678e-05, "loss": 2.8942, "step": 37550 }, { "epoch": 2.94, "learning_rate": 7.055516349094331e-05, "loss": 2.8818, "step": 37551 }, { "epoch": 2.94, "learning_rate": 7.055437936171881e-05, "loss": 2.6562, "step": 37552 }, { "epoch": 2.94, "learning_rate": 7.055359523249432e-05, "loss": 2.7582, "step": 37553 }, { "epoch": 2.94, "learning_rate": 7.055281110326982e-05, "loss": 2.6563, "step": 37554 }, { "epoch": 2.94, "learning_rate": 7.055202697404532e-05, "loss": 2.8227, "step": 37555 }, { "epoch": 2.94, "learning_rate": 7.055124284482083e-05, "loss": 2.734, "step": 37556 }, { "epoch": 2.94, "learning_rate": 7.055045871559633e-05, "loss": 2.9842, "step": 37557 }, { "epoch": 2.95, "learning_rate": 7.054967458637183e-05, "loss": 2.6297, "step": 37558 }, { "epoch": 2.95, "learning_rate": 7.054889045714734e-05, "loss": 2.8695, "step": 37559 }, { "epoch": 2.95, "learning_rate": 7.054810632792284e-05, "loss": 2.9553, "step": 37560 }, { "epoch": 2.95, "learning_rate": 7.054732219869835e-05, "loss": 2.7007, "step": 37561 }, { "epoch": 2.95, "learning_rate": 7.054653806947385e-05, "loss": 2.8891, "step": 37562 }, { "epoch": 2.95, "learning_rate": 7.054575394024935e-05, "loss": 2.9415, "step": 37563 }, { "epoch": 2.95, "learning_rate": 7.054496981102487e-05, "loss": 2.8815, "step": 37564 }, { "epoch": 2.95, "learning_rate": 7.054418568180037e-05, "loss": 2.7717, "step": 37565 }, { "epoch": 2.95, "learning_rate": 7.054340155257588e-05, "loss": 2.688, "step": 37566 }, { "epoch": 2.95, "learning_rate": 7.054261742335138e-05, "loss": 2.769, "step": 37567 }, { "epoch": 2.95, "learning_rate": 7.054183329412688e-05, "loss": 3.0389, "step": 37568 }, { "epoch": 2.95, "learning_rate": 7.054104916490239e-05, "loss": 2.8565, "step": 37569 }, { "epoch": 2.95, "learning_rate": 7.054026503567789e-05, "loss": 2.8104, "step": 37570 }, { "epoch": 2.95, "learning_rate": 7.05394809064534e-05, "loss": 2.6672, "step": 37571 }, { "epoch": 2.95, "learning_rate": 7.053869677722889e-05, "loss": 2.8149, "step": 37572 }, { "epoch": 2.95, "learning_rate": 7.053791264800439e-05, "loss": 2.8235, "step": 37573 }, { "epoch": 2.95, "learning_rate": 7.05371285187799e-05, "loss": 2.9169, "step": 37574 }, { "epoch": 2.95, "learning_rate": 7.05363443895554e-05, "loss": 2.7577, "step": 37575 }, { "epoch": 2.95, "learning_rate": 7.05355602603309e-05, "loss": 2.741, "step": 37576 }, { "epoch": 2.95, "learning_rate": 7.05347761311064e-05, "loss": 2.7189, "step": 37577 }, { "epoch": 2.95, "learning_rate": 7.053399200188191e-05, "loss": 2.8582, "step": 37578 }, { "epoch": 2.95, "learning_rate": 7.053320787265741e-05, "loss": 2.6575, "step": 37579 }, { "epoch": 2.95, "learning_rate": 7.053242374343292e-05, "loss": 2.6095, "step": 37580 }, { "epoch": 2.95, "learning_rate": 7.053163961420842e-05, "loss": 2.8489, "step": 37581 }, { "epoch": 2.95, "learning_rate": 7.053085548498392e-05, "loss": 2.9403, "step": 37582 }, { "epoch": 2.95, "learning_rate": 7.053007135575943e-05, "loss": 2.847, "step": 37583 }, { "epoch": 2.95, "learning_rate": 7.052928722653493e-05, "loss": 2.6759, "step": 37584 }, { "epoch": 2.95, "learning_rate": 7.052850309731043e-05, "loss": 2.8384, "step": 37585 }, { "epoch": 2.95, "learning_rate": 7.052771896808594e-05, "loss": 2.9028, "step": 37586 }, { "epoch": 2.95, "learning_rate": 7.052693483886144e-05, "loss": 2.6871, "step": 37587 }, { "epoch": 2.95, "learning_rate": 7.052615070963694e-05, "loss": 2.9273, "step": 37588 }, { "epoch": 2.95, "learning_rate": 7.052536658041245e-05, "loss": 2.7591, "step": 37589 }, { "epoch": 2.95, "learning_rate": 7.052458245118797e-05, "loss": 2.8395, "step": 37590 }, { "epoch": 2.95, "learning_rate": 7.052379832196347e-05, "loss": 2.9656, "step": 37591 }, { "epoch": 2.95, "learning_rate": 7.052301419273897e-05, "loss": 2.7298, "step": 37592 }, { "epoch": 2.95, "learning_rate": 7.052223006351448e-05, "loss": 2.6145, "step": 37593 }, { "epoch": 2.95, "learning_rate": 7.052144593428998e-05, "loss": 2.6216, "step": 37594 }, { "epoch": 2.95, "learning_rate": 7.052066180506548e-05, "loss": 3.0184, "step": 37595 }, { "epoch": 2.95, "learning_rate": 7.051987767584099e-05, "loss": 2.9824, "step": 37596 }, { "epoch": 2.95, "learning_rate": 7.051909354661649e-05, "loss": 2.7774, "step": 37597 }, { "epoch": 2.95, "learning_rate": 7.0518309417392e-05, "loss": 2.9476, "step": 37598 }, { "epoch": 2.95, "learning_rate": 7.05175252881675e-05, "loss": 2.9513, "step": 37599 }, { "epoch": 2.95, "learning_rate": 7.0516741158943e-05, "loss": 2.7362, "step": 37600 }, { "epoch": 2.95, "learning_rate": 7.05159570297185e-05, "loss": 2.6786, "step": 37601 }, { "epoch": 2.95, "learning_rate": 7.051517290049401e-05, "loss": 2.7978, "step": 37602 }, { "epoch": 2.95, "learning_rate": 7.051438877126951e-05, "loss": 2.8998, "step": 37603 }, { "epoch": 2.95, "learning_rate": 7.051360464204502e-05, "loss": 2.6596, "step": 37604 }, { "epoch": 2.95, "learning_rate": 7.051282051282052e-05, "loss": 2.8467, "step": 37605 }, { "epoch": 2.95, "learning_rate": 7.051203638359602e-05, "loss": 2.742, "step": 37606 }, { "epoch": 2.95, "learning_rate": 7.051125225437153e-05, "loss": 2.7528, "step": 37607 }, { "epoch": 2.95, "learning_rate": 7.051046812514702e-05, "loss": 2.6907, "step": 37608 }, { "epoch": 2.95, "learning_rate": 7.050968399592252e-05, "loss": 2.5554, "step": 37609 }, { "epoch": 2.95, "learning_rate": 7.050889986669803e-05, "loss": 2.7483, "step": 37610 }, { "epoch": 2.95, "learning_rate": 7.050811573747353e-05, "loss": 2.6812, "step": 37611 }, { "epoch": 2.95, "learning_rate": 7.050733160824903e-05, "loss": 2.873, "step": 37612 }, { "epoch": 2.95, "learning_rate": 7.050654747902454e-05, "loss": 2.735, "step": 37613 }, { "epoch": 2.95, "learning_rate": 7.050576334980004e-05, "loss": 3.0608, "step": 37614 }, { "epoch": 2.95, "learning_rate": 7.050497922057556e-05, "loss": 2.8054, "step": 37615 }, { "epoch": 2.95, "learning_rate": 7.050419509135106e-05, "loss": 2.5901, "step": 37616 }, { "epoch": 2.95, "learning_rate": 7.050341096212657e-05, "loss": 2.599, "step": 37617 }, { "epoch": 2.95, "learning_rate": 7.050262683290207e-05, "loss": 2.7172, "step": 37618 }, { "epoch": 2.95, "learning_rate": 7.050184270367757e-05, "loss": 3.0167, "step": 37619 }, { "epoch": 2.95, "learning_rate": 7.050105857445308e-05, "loss": 2.6928, "step": 37620 }, { "epoch": 2.95, "learning_rate": 7.050027444522858e-05, "loss": 2.8139, "step": 37621 }, { "epoch": 2.95, "learning_rate": 7.049949031600408e-05, "loss": 2.6382, "step": 37622 }, { "epoch": 2.95, "learning_rate": 7.049870618677959e-05, "loss": 2.9487, "step": 37623 }, { "epoch": 2.95, "learning_rate": 7.049792205755509e-05, "loss": 2.8849, "step": 37624 }, { "epoch": 2.95, "learning_rate": 7.04971379283306e-05, "loss": 2.7191, "step": 37625 }, { "epoch": 2.95, "learning_rate": 7.04963537991061e-05, "loss": 2.7128, "step": 37626 }, { "epoch": 2.95, "learning_rate": 7.04955696698816e-05, "loss": 2.8804, "step": 37627 }, { "epoch": 2.95, "learning_rate": 7.04947855406571e-05, "loss": 2.8758, "step": 37628 }, { "epoch": 2.95, "learning_rate": 7.049400141143261e-05, "loss": 2.9976, "step": 37629 }, { "epoch": 2.95, "learning_rate": 7.049321728220811e-05, "loss": 2.6746, "step": 37630 }, { "epoch": 2.95, "learning_rate": 7.049243315298362e-05, "loss": 2.838, "step": 37631 }, { "epoch": 2.95, "learning_rate": 7.049164902375912e-05, "loss": 2.6217, "step": 37632 }, { "epoch": 2.95, "learning_rate": 7.049086489453462e-05, "loss": 2.7257, "step": 37633 }, { "epoch": 2.95, "learning_rate": 7.049008076531013e-05, "loss": 2.7144, "step": 37634 }, { "epoch": 2.95, "learning_rate": 7.048929663608563e-05, "loss": 2.7458, "step": 37635 }, { "epoch": 2.95, "learning_rate": 7.048851250686113e-05, "loss": 2.5555, "step": 37636 }, { "epoch": 2.95, "learning_rate": 7.048772837763664e-05, "loss": 2.7944, "step": 37637 }, { "epoch": 2.95, "learning_rate": 7.048694424841214e-05, "loss": 2.7384, "step": 37638 }, { "epoch": 2.95, "learning_rate": 7.048616011918765e-05, "loss": 2.8006, "step": 37639 }, { "epoch": 2.95, "learning_rate": 7.048537598996315e-05, "loss": 2.9467, "step": 37640 }, { "epoch": 2.95, "learning_rate": 7.048459186073865e-05, "loss": 2.962, "step": 37641 }, { "epoch": 2.95, "learning_rate": 7.048380773151416e-05, "loss": 2.7873, "step": 37642 }, { "epoch": 2.95, "learning_rate": 7.048302360228966e-05, "loss": 2.7454, "step": 37643 }, { "epoch": 2.95, "learning_rate": 7.048223947306516e-05, "loss": 2.6964, "step": 37644 }, { "epoch": 2.95, "learning_rate": 7.048145534384067e-05, "loss": 2.7752, "step": 37645 }, { "epoch": 2.95, "learning_rate": 7.048067121461617e-05, "loss": 3.0018, "step": 37646 }, { "epoch": 2.95, "learning_rate": 7.047988708539168e-05, "loss": 2.9575, "step": 37647 }, { "epoch": 2.95, "learning_rate": 7.047910295616718e-05, "loss": 2.6136, "step": 37648 }, { "epoch": 2.95, "learning_rate": 7.047831882694268e-05, "loss": 2.9696, "step": 37649 }, { "epoch": 2.95, "learning_rate": 7.047753469771819e-05, "loss": 2.775, "step": 37650 }, { "epoch": 2.95, "learning_rate": 7.047675056849369e-05, "loss": 2.8543, "step": 37651 }, { "epoch": 2.95, "learning_rate": 7.04759664392692e-05, "loss": 2.8979, "step": 37652 }, { "epoch": 2.95, "learning_rate": 7.04751823100447e-05, "loss": 2.5664, "step": 37653 }, { "epoch": 2.95, "learning_rate": 7.04743981808202e-05, "loss": 2.6426, "step": 37654 }, { "epoch": 2.95, "learning_rate": 7.04736140515957e-05, "loss": 2.978, "step": 37655 }, { "epoch": 2.95, "learning_rate": 7.047282992237121e-05, "loss": 2.6688, "step": 37656 }, { "epoch": 2.95, "learning_rate": 7.047204579314671e-05, "loss": 2.9212, "step": 37657 }, { "epoch": 2.95, "learning_rate": 7.047126166392222e-05, "loss": 2.8027, "step": 37658 }, { "epoch": 2.95, "learning_rate": 7.047047753469772e-05, "loss": 2.9006, "step": 37659 }, { "epoch": 2.95, "learning_rate": 7.046969340547322e-05, "loss": 2.8118, "step": 37660 }, { "epoch": 2.95, "learning_rate": 7.046890927624873e-05, "loss": 2.7892, "step": 37661 }, { "epoch": 2.95, "learning_rate": 7.046812514702423e-05, "loss": 2.8135, "step": 37662 }, { "epoch": 2.95, "learning_rate": 7.046734101779973e-05, "loss": 2.9182, "step": 37663 }, { "epoch": 2.95, "learning_rate": 7.046655688857524e-05, "loss": 2.6162, "step": 37664 }, { "epoch": 2.95, "learning_rate": 7.046577275935074e-05, "loss": 2.8786, "step": 37665 }, { "epoch": 2.95, "learning_rate": 7.046498863012625e-05, "loss": 2.9056, "step": 37666 }, { "epoch": 2.95, "learning_rate": 7.046420450090175e-05, "loss": 2.7489, "step": 37667 }, { "epoch": 2.95, "learning_rate": 7.046342037167725e-05, "loss": 2.9501, "step": 37668 }, { "epoch": 2.95, "learning_rate": 7.046263624245276e-05, "loss": 2.6335, "step": 37669 }, { "epoch": 2.95, "learning_rate": 7.046185211322826e-05, "loss": 2.8576, "step": 37670 }, { "epoch": 2.95, "learning_rate": 7.046106798400376e-05, "loss": 2.6737, "step": 37671 }, { "epoch": 2.95, "learning_rate": 7.046028385477928e-05, "loss": 2.8311, "step": 37672 }, { "epoch": 2.95, "learning_rate": 7.045949972555478e-05, "loss": 2.8555, "step": 37673 }, { "epoch": 2.95, "learning_rate": 7.045871559633029e-05, "loss": 2.7297, "step": 37674 }, { "epoch": 2.95, "learning_rate": 7.045793146710579e-05, "loss": 2.8048, "step": 37675 }, { "epoch": 2.95, "learning_rate": 7.04571473378813e-05, "loss": 2.7719, "step": 37676 }, { "epoch": 2.95, "learning_rate": 7.04563632086568e-05, "loss": 3.0025, "step": 37677 }, { "epoch": 2.95, "learning_rate": 7.04555790794323e-05, "loss": 3.153, "step": 37678 }, { "epoch": 2.95, "learning_rate": 7.04547949502078e-05, "loss": 2.8084, "step": 37679 }, { "epoch": 2.95, "learning_rate": 7.04540108209833e-05, "loss": 2.811, "step": 37680 }, { "epoch": 2.95, "learning_rate": 7.04532266917588e-05, "loss": 2.7284, "step": 37681 }, { "epoch": 2.95, "learning_rate": 7.04524425625343e-05, "loss": 2.8898, "step": 37682 }, { "epoch": 2.95, "learning_rate": 7.045165843330981e-05, "loss": 2.7447, "step": 37683 }, { "epoch": 2.95, "learning_rate": 7.045087430408531e-05, "loss": 2.8802, "step": 37684 }, { "epoch": 2.95, "learning_rate": 7.045009017486082e-05, "loss": 2.7319, "step": 37685 }, { "epoch": 2.96, "learning_rate": 7.044930604563632e-05, "loss": 2.6542, "step": 37686 }, { "epoch": 2.96, "learning_rate": 7.044852191641182e-05, "loss": 2.9801, "step": 37687 }, { "epoch": 2.96, "learning_rate": 7.044773778718733e-05, "loss": 2.754, "step": 37688 }, { "epoch": 2.96, "learning_rate": 7.044695365796283e-05, "loss": 2.8952, "step": 37689 }, { "epoch": 2.96, "learning_rate": 7.044616952873833e-05, "loss": 2.7052, "step": 37690 }, { "epoch": 2.96, "learning_rate": 7.044538539951384e-05, "loss": 2.7216, "step": 37691 }, { "epoch": 2.96, "learning_rate": 7.044460127028934e-05, "loss": 2.9493, "step": 37692 }, { "epoch": 2.96, "learning_rate": 7.044381714106484e-05, "loss": 2.699, "step": 37693 }, { "epoch": 2.96, "learning_rate": 7.044303301184035e-05, "loss": 2.7653, "step": 37694 }, { "epoch": 2.96, "learning_rate": 7.044224888261585e-05, "loss": 2.6843, "step": 37695 }, { "epoch": 2.96, "learning_rate": 7.044146475339136e-05, "loss": 2.8092, "step": 37696 }, { "epoch": 2.96, "learning_rate": 7.044068062416686e-05, "loss": 2.8712, "step": 37697 }, { "epoch": 2.96, "learning_rate": 7.043989649494238e-05, "loss": 2.5835, "step": 37698 }, { "epoch": 2.96, "learning_rate": 7.043911236571788e-05, "loss": 2.8109, "step": 37699 }, { "epoch": 2.96, "learning_rate": 7.043832823649338e-05, "loss": 2.8996, "step": 37700 }, { "epoch": 2.96, "learning_rate": 7.043754410726889e-05, "loss": 2.6563, "step": 37701 }, { "epoch": 2.96, "learning_rate": 7.043675997804439e-05, "loss": 2.7543, "step": 37702 }, { "epoch": 2.96, "learning_rate": 7.04359758488199e-05, "loss": 3.026, "step": 37703 }, { "epoch": 2.96, "learning_rate": 7.04351917195954e-05, "loss": 2.742, "step": 37704 }, { "epoch": 2.96, "learning_rate": 7.04344075903709e-05, "loss": 3.0624, "step": 37705 }, { "epoch": 2.96, "learning_rate": 7.04336234611464e-05, "loss": 2.8504, "step": 37706 }, { "epoch": 2.96, "learning_rate": 7.043283933192191e-05, "loss": 2.7499, "step": 37707 }, { "epoch": 2.96, "learning_rate": 7.043205520269741e-05, "loss": 2.8148, "step": 37708 }, { "epoch": 2.96, "learning_rate": 7.043127107347292e-05, "loss": 2.7632, "step": 37709 }, { "epoch": 2.96, "learning_rate": 7.043048694424842e-05, "loss": 2.7084, "step": 37710 }, { "epoch": 2.96, "learning_rate": 7.042970281502392e-05, "loss": 2.6353, "step": 37711 }, { "epoch": 2.96, "learning_rate": 7.042891868579943e-05, "loss": 2.7518, "step": 37712 }, { "epoch": 2.96, "learning_rate": 7.042813455657493e-05, "loss": 2.7598, "step": 37713 }, { "epoch": 2.96, "learning_rate": 7.042735042735044e-05, "loss": 2.7037, "step": 37714 }, { "epoch": 2.96, "learning_rate": 7.042656629812594e-05, "loss": 2.7161, "step": 37715 }, { "epoch": 2.96, "learning_rate": 7.042578216890143e-05, "loss": 2.9027, "step": 37716 }, { "epoch": 2.96, "learning_rate": 7.042499803967693e-05, "loss": 2.6991, "step": 37717 }, { "epoch": 2.96, "learning_rate": 7.042421391045244e-05, "loss": 2.697, "step": 37718 }, { "epoch": 2.96, "learning_rate": 7.042342978122794e-05, "loss": 2.8794, "step": 37719 }, { "epoch": 2.96, "learning_rate": 7.042264565200344e-05, "loss": 2.7257, "step": 37720 }, { "epoch": 2.96, "learning_rate": 7.042186152277895e-05, "loss": 2.6461, "step": 37721 }, { "epoch": 2.96, "learning_rate": 7.042107739355445e-05, "loss": 2.5029, "step": 37722 }, { "epoch": 2.96, "learning_rate": 7.042029326432997e-05, "loss": 2.8619, "step": 37723 }, { "epoch": 2.96, "learning_rate": 7.041950913510547e-05, "loss": 2.9079, "step": 37724 }, { "epoch": 2.96, "learning_rate": 7.041872500588098e-05, "loss": 2.7308, "step": 37725 }, { "epoch": 2.96, "learning_rate": 7.041794087665648e-05, "loss": 2.6502, "step": 37726 }, { "epoch": 2.96, "learning_rate": 7.041715674743198e-05, "loss": 2.8045, "step": 37727 }, { "epoch": 2.96, "learning_rate": 7.041637261820749e-05, "loss": 2.8256, "step": 37728 }, { "epoch": 2.96, "learning_rate": 7.041558848898299e-05, "loss": 2.8577, "step": 37729 }, { "epoch": 2.96, "learning_rate": 7.04148043597585e-05, "loss": 2.8815, "step": 37730 }, { "epoch": 2.96, "learning_rate": 7.0414020230534e-05, "loss": 2.7825, "step": 37731 }, { "epoch": 2.96, "learning_rate": 7.04132361013095e-05, "loss": 2.8204, "step": 37732 }, { "epoch": 2.96, "learning_rate": 7.0412451972085e-05, "loss": 2.9542, "step": 37733 }, { "epoch": 2.96, "learning_rate": 7.041166784286051e-05, "loss": 2.6588, "step": 37734 }, { "epoch": 2.96, "learning_rate": 7.041088371363601e-05, "loss": 2.8955, "step": 37735 }, { "epoch": 2.96, "learning_rate": 7.041009958441152e-05, "loss": 2.7951, "step": 37736 }, { "epoch": 2.96, "learning_rate": 7.040931545518702e-05, "loss": 2.7762, "step": 37737 }, { "epoch": 2.96, "learning_rate": 7.040853132596252e-05, "loss": 2.7053, "step": 37738 }, { "epoch": 2.96, "learning_rate": 7.040774719673803e-05, "loss": 2.8101, "step": 37739 }, { "epoch": 2.96, "learning_rate": 7.040696306751353e-05, "loss": 3.0476, "step": 37740 }, { "epoch": 2.96, "learning_rate": 7.040617893828903e-05, "loss": 2.724, "step": 37741 }, { "epoch": 2.96, "learning_rate": 7.040539480906454e-05, "loss": 2.7967, "step": 37742 }, { "epoch": 2.96, "learning_rate": 7.040461067984004e-05, "loss": 2.7161, "step": 37743 }, { "epoch": 2.96, "learning_rate": 7.040382655061555e-05, "loss": 2.92, "step": 37744 }, { "epoch": 2.96, "learning_rate": 7.040304242139105e-05, "loss": 3.1773, "step": 37745 }, { "epoch": 2.96, "learning_rate": 7.040225829216655e-05, "loss": 2.8619, "step": 37746 }, { "epoch": 2.96, "learning_rate": 7.040147416294206e-05, "loss": 2.6254, "step": 37747 }, { "epoch": 2.96, "learning_rate": 7.040069003371756e-05, "loss": 2.7742, "step": 37748 }, { "epoch": 2.96, "learning_rate": 7.039990590449306e-05, "loss": 2.8167, "step": 37749 }, { "epoch": 2.96, "learning_rate": 7.039912177526857e-05, "loss": 2.9729, "step": 37750 }, { "epoch": 2.96, "learning_rate": 7.039833764604407e-05, "loss": 2.6571, "step": 37751 }, { "epoch": 2.96, "learning_rate": 7.039755351681958e-05, "loss": 2.7169, "step": 37752 }, { "epoch": 2.96, "learning_rate": 7.039676938759508e-05, "loss": 2.791, "step": 37753 }, { "epoch": 2.96, "learning_rate": 7.039598525837058e-05, "loss": 2.9442, "step": 37754 }, { "epoch": 2.96, "learning_rate": 7.039520112914609e-05, "loss": 2.7013, "step": 37755 }, { "epoch": 2.96, "learning_rate": 7.039441699992159e-05, "loss": 2.7508, "step": 37756 }, { "epoch": 2.96, "learning_rate": 7.03936328706971e-05, "loss": 2.994, "step": 37757 }, { "epoch": 2.96, "learning_rate": 7.03928487414726e-05, "loss": 2.9114, "step": 37758 }, { "epoch": 2.96, "learning_rate": 7.03920646122481e-05, "loss": 2.6375, "step": 37759 }, { "epoch": 2.96, "learning_rate": 7.03912804830236e-05, "loss": 2.7797, "step": 37760 }, { "epoch": 2.96, "learning_rate": 7.039049635379911e-05, "loss": 3.0381, "step": 37761 }, { "epoch": 2.96, "learning_rate": 7.038971222457461e-05, "loss": 2.8135, "step": 37762 }, { "epoch": 2.96, "learning_rate": 7.038892809535012e-05, "loss": 3.0114, "step": 37763 }, { "epoch": 2.96, "learning_rate": 7.038814396612562e-05, "loss": 2.7379, "step": 37764 }, { "epoch": 2.96, "learning_rate": 7.038735983690112e-05, "loss": 2.7585, "step": 37765 }, { "epoch": 2.96, "learning_rate": 7.038657570767663e-05, "loss": 2.7493, "step": 37766 }, { "epoch": 2.96, "learning_rate": 7.038579157845213e-05, "loss": 2.7009, "step": 37767 }, { "epoch": 2.96, "learning_rate": 7.038500744922763e-05, "loss": 3.1816, "step": 37768 }, { "epoch": 2.96, "learning_rate": 7.038422332000314e-05, "loss": 2.8073, "step": 37769 }, { "epoch": 2.96, "learning_rate": 7.038343919077864e-05, "loss": 2.8935, "step": 37770 }, { "epoch": 2.96, "learning_rate": 7.038265506155414e-05, "loss": 2.9492, "step": 37771 }, { "epoch": 2.96, "learning_rate": 7.038187093232965e-05, "loss": 2.8419, "step": 37772 }, { "epoch": 2.96, "learning_rate": 7.038108680310515e-05, "loss": 2.6685, "step": 37773 }, { "epoch": 2.96, "learning_rate": 7.038030267388066e-05, "loss": 2.6184, "step": 37774 }, { "epoch": 2.96, "learning_rate": 7.037951854465616e-05, "loss": 3.0825, "step": 37775 }, { "epoch": 2.96, "learning_rate": 7.037873441543166e-05, "loss": 2.9209, "step": 37776 }, { "epoch": 2.96, "learning_rate": 7.037795028620717e-05, "loss": 2.6663, "step": 37777 }, { "epoch": 2.96, "learning_rate": 7.037716615698267e-05, "loss": 2.5258, "step": 37778 }, { "epoch": 2.96, "learning_rate": 7.037638202775817e-05, "loss": 2.6836, "step": 37779 }, { "epoch": 2.96, "learning_rate": 7.037559789853369e-05, "loss": 2.7369, "step": 37780 }, { "epoch": 2.96, "learning_rate": 7.03748137693092e-05, "loss": 2.7637, "step": 37781 }, { "epoch": 2.96, "learning_rate": 7.03740296400847e-05, "loss": 2.8028, "step": 37782 }, { "epoch": 2.96, "learning_rate": 7.03732455108602e-05, "loss": 2.7927, "step": 37783 }, { "epoch": 2.96, "learning_rate": 7.03724613816357e-05, "loss": 3.0357, "step": 37784 }, { "epoch": 2.96, "learning_rate": 7.037167725241121e-05, "loss": 3.0402, "step": 37785 }, { "epoch": 2.96, "learning_rate": 7.037089312318671e-05, "loss": 2.5451, "step": 37786 }, { "epoch": 2.96, "learning_rate": 7.037010899396222e-05, "loss": 2.5969, "step": 37787 }, { "epoch": 2.96, "learning_rate": 7.036932486473771e-05, "loss": 2.9846, "step": 37788 }, { "epoch": 2.96, "learning_rate": 7.036854073551321e-05, "loss": 2.8942, "step": 37789 }, { "epoch": 2.96, "learning_rate": 7.036775660628871e-05, "loss": 2.6837, "step": 37790 }, { "epoch": 2.96, "learning_rate": 7.036697247706422e-05, "loss": 2.7007, "step": 37791 }, { "epoch": 2.96, "learning_rate": 7.036618834783972e-05, "loss": 2.6152, "step": 37792 }, { "epoch": 2.96, "learning_rate": 7.036540421861523e-05, "loss": 2.9917, "step": 37793 }, { "epoch": 2.96, "learning_rate": 7.036462008939073e-05, "loss": 2.8936, "step": 37794 }, { "epoch": 2.96, "learning_rate": 7.036383596016623e-05, "loss": 2.6022, "step": 37795 }, { "epoch": 2.96, "learning_rate": 7.036305183094174e-05, "loss": 2.8025, "step": 37796 }, { "epoch": 2.96, "learning_rate": 7.036226770171724e-05, "loss": 2.8324, "step": 37797 }, { "epoch": 2.96, "learning_rate": 7.036148357249274e-05, "loss": 2.6773, "step": 37798 }, { "epoch": 2.96, "learning_rate": 7.036069944326825e-05, "loss": 2.9575, "step": 37799 }, { "epoch": 2.96, "learning_rate": 7.035991531404375e-05, "loss": 3.1151, "step": 37800 }, { "epoch": 2.96, "learning_rate": 7.035913118481926e-05, "loss": 2.7874, "step": 37801 }, { "epoch": 2.96, "learning_rate": 7.035834705559476e-05, "loss": 2.674, "step": 37802 }, { "epoch": 2.96, "learning_rate": 7.035756292637026e-05, "loss": 2.9243, "step": 37803 }, { "epoch": 2.96, "learning_rate": 7.035677879714577e-05, "loss": 2.858, "step": 37804 }, { "epoch": 2.96, "learning_rate": 7.035599466792127e-05, "loss": 2.6687, "step": 37805 }, { "epoch": 2.96, "learning_rate": 7.035521053869679e-05, "loss": 3.1277, "step": 37806 }, { "epoch": 2.96, "learning_rate": 7.035442640947229e-05, "loss": 2.7745, "step": 37807 }, { "epoch": 2.96, "learning_rate": 7.03536422802478e-05, "loss": 2.8414, "step": 37808 }, { "epoch": 2.96, "learning_rate": 7.03528581510233e-05, "loss": 2.6865, "step": 37809 }, { "epoch": 2.96, "learning_rate": 7.03520740217988e-05, "loss": 2.7842, "step": 37810 }, { "epoch": 2.96, "learning_rate": 7.03512898925743e-05, "loss": 2.5224, "step": 37811 }, { "epoch": 2.96, "learning_rate": 7.035050576334981e-05, "loss": 2.8219, "step": 37812 }, { "epoch": 2.97, "learning_rate": 7.034972163412531e-05, "loss": 2.7898, "step": 37813 }, { "epoch": 2.97, "learning_rate": 7.034893750490082e-05, "loss": 2.8766, "step": 37814 }, { "epoch": 2.97, "learning_rate": 7.034815337567632e-05, "loss": 2.699, "step": 37815 }, { "epoch": 2.97, "learning_rate": 7.034736924645182e-05, "loss": 2.7946, "step": 37816 }, { "epoch": 2.97, "learning_rate": 7.034658511722733e-05, "loss": 2.8627, "step": 37817 }, { "epoch": 2.97, "learning_rate": 7.034580098800283e-05, "loss": 3.1972, "step": 37818 }, { "epoch": 2.97, "learning_rate": 7.034501685877834e-05, "loss": 2.7458, "step": 37819 }, { "epoch": 2.97, "learning_rate": 7.034423272955384e-05, "loss": 2.6613, "step": 37820 }, { "epoch": 2.97, "learning_rate": 7.034344860032934e-05, "loss": 2.8944, "step": 37821 }, { "epoch": 2.97, "learning_rate": 7.034266447110485e-05, "loss": 2.6001, "step": 37822 }, { "epoch": 2.97, "learning_rate": 7.034188034188035e-05, "loss": 2.9086, "step": 37823 }, { "epoch": 2.97, "learning_rate": 7.034109621265584e-05, "loss": 2.8579, "step": 37824 }, { "epoch": 2.97, "learning_rate": 7.034031208343134e-05, "loss": 2.8737, "step": 37825 }, { "epoch": 2.97, "learning_rate": 7.033952795420685e-05, "loss": 2.9665, "step": 37826 }, { "epoch": 2.97, "learning_rate": 7.033874382498235e-05, "loss": 2.927, "step": 37827 }, { "epoch": 2.97, "learning_rate": 7.033795969575785e-05, "loss": 2.6253, "step": 37828 }, { "epoch": 2.97, "learning_rate": 7.033717556653336e-05, "loss": 2.8496, "step": 37829 }, { "epoch": 2.97, "learning_rate": 7.033639143730886e-05, "loss": 3.025, "step": 37830 }, { "epoch": 2.97, "learning_rate": 7.033560730808438e-05, "loss": 2.9381, "step": 37831 }, { "epoch": 2.97, "learning_rate": 7.033482317885988e-05, "loss": 2.7227, "step": 37832 }, { "epoch": 2.97, "learning_rate": 7.033403904963539e-05, "loss": 2.9064, "step": 37833 }, { "epoch": 2.97, "learning_rate": 7.033325492041089e-05, "loss": 3.076, "step": 37834 }, { "epoch": 2.97, "learning_rate": 7.03324707911864e-05, "loss": 2.6667, "step": 37835 }, { "epoch": 2.97, "learning_rate": 7.03316866619619e-05, "loss": 2.7079, "step": 37836 }, { "epoch": 2.97, "learning_rate": 7.03309025327374e-05, "loss": 2.7796, "step": 37837 }, { "epoch": 2.97, "learning_rate": 7.03301184035129e-05, "loss": 2.9851, "step": 37838 }, { "epoch": 2.97, "learning_rate": 7.032933427428841e-05, "loss": 2.9573, "step": 37839 }, { "epoch": 2.97, "learning_rate": 7.032855014506391e-05, "loss": 2.6103, "step": 37840 }, { "epoch": 2.97, "learning_rate": 7.032776601583942e-05, "loss": 2.9172, "step": 37841 }, { "epoch": 2.97, "learning_rate": 7.032698188661492e-05, "loss": 2.9812, "step": 37842 }, { "epoch": 2.97, "learning_rate": 7.032619775739042e-05, "loss": 2.9053, "step": 37843 }, { "epoch": 2.97, "learning_rate": 7.032541362816593e-05, "loss": 2.7683, "step": 37844 }, { "epoch": 2.97, "learning_rate": 7.032462949894143e-05, "loss": 2.8743, "step": 37845 }, { "epoch": 2.97, "learning_rate": 7.032384536971693e-05, "loss": 2.5262, "step": 37846 }, { "epoch": 2.97, "learning_rate": 7.032306124049244e-05, "loss": 2.7261, "step": 37847 }, { "epoch": 2.97, "learning_rate": 7.032227711126794e-05, "loss": 2.8084, "step": 37848 }, { "epoch": 2.97, "learning_rate": 7.032149298204345e-05, "loss": 2.8849, "step": 37849 }, { "epoch": 2.97, "learning_rate": 7.032070885281895e-05, "loss": 2.5907, "step": 37850 }, { "epoch": 2.97, "learning_rate": 7.031992472359445e-05, "loss": 2.9537, "step": 37851 }, { "epoch": 2.97, "learning_rate": 7.031914059436996e-05, "loss": 2.8254, "step": 37852 }, { "epoch": 2.97, "learning_rate": 7.031835646514546e-05, "loss": 2.7082, "step": 37853 }, { "epoch": 2.97, "learning_rate": 7.031757233592096e-05, "loss": 2.8737, "step": 37854 }, { "epoch": 2.97, "learning_rate": 7.031678820669647e-05, "loss": 2.8351, "step": 37855 }, { "epoch": 2.97, "learning_rate": 7.031600407747197e-05, "loss": 2.8395, "step": 37856 }, { "epoch": 2.97, "learning_rate": 7.031521994824747e-05, "loss": 2.9571, "step": 37857 }, { "epoch": 2.97, "learning_rate": 7.031443581902298e-05, "loss": 2.8732, "step": 37858 }, { "epoch": 2.97, "learning_rate": 7.031365168979848e-05, "loss": 3.0741, "step": 37859 }, { "epoch": 2.97, "learning_rate": 7.031286756057399e-05, "loss": 2.8574, "step": 37860 }, { "epoch": 2.97, "learning_rate": 7.031208343134949e-05, "loss": 2.774, "step": 37861 }, { "epoch": 2.97, "learning_rate": 7.031129930212499e-05, "loss": 2.8127, "step": 37862 }, { "epoch": 2.97, "learning_rate": 7.03105151729005e-05, "loss": 2.9308, "step": 37863 }, { "epoch": 2.97, "learning_rate": 7.0309731043676e-05, "loss": 2.9801, "step": 37864 }, { "epoch": 2.97, "learning_rate": 7.03089469144515e-05, "loss": 3.0229, "step": 37865 }, { "epoch": 2.97, "learning_rate": 7.030816278522701e-05, "loss": 2.8913, "step": 37866 }, { "epoch": 2.97, "learning_rate": 7.030737865600251e-05, "loss": 2.8197, "step": 37867 }, { "epoch": 2.97, "learning_rate": 7.030659452677802e-05, "loss": 2.8672, "step": 37868 }, { "epoch": 2.97, "learning_rate": 7.030581039755352e-05, "loss": 2.9779, "step": 37869 }, { "epoch": 2.97, "learning_rate": 7.030502626832902e-05, "loss": 2.8405, "step": 37870 }, { "epoch": 2.97, "learning_rate": 7.030424213910453e-05, "loss": 2.8203, "step": 37871 }, { "epoch": 2.97, "learning_rate": 7.030345800988003e-05, "loss": 2.8058, "step": 37872 }, { "epoch": 2.97, "learning_rate": 7.030267388065553e-05, "loss": 2.8739, "step": 37873 }, { "epoch": 2.97, "learning_rate": 7.030188975143104e-05, "loss": 2.5658, "step": 37874 }, { "epoch": 2.97, "learning_rate": 7.030110562220654e-05, "loss": 2.8076, "step": 37875 }, { "epoch": 2.97, "learning_rate": 7.030032149298204e-05, "loss": 2.8626, "step": 37876 }, { "epoch": 2.97, "learning_rate": 7.029953736375755e-05, "loss": 2.7493, "step": 37877 }, { "epoch": 2.97, "learning_rate": 7.029875323453305e-05, "loss": 2.9175, "step": 37878 }, { "epoch": 2.97, "learning_rate": 7.029796910530856e-05, "loss": 2.7492, "step": 37879 }, { "epoch": 2.97, "learning_rate": 7.029718497608406e-05, "loss": 2.7831, "step": 37880 }, { "epoch": 2.97, "learning_rate": 7.029640084685956e-05, "loss": 2.6799, "step": 37881 }, { "epoch": 2.97, "learning_rate": 7.029561671763507e-05, "loss": 2.9543, "step": 37882 }, { "epoch": 2.97, "learning_rate": 7.029483258841057e-05, "loss": 2.671, "step": 37883 }, { "epoch": 2.97, "learning_rate": 7.029404845918607e-05, "loss": 2.8715, "step": 37884 }, { "epoch": 2.97, "learning_rate": 7.029326432996158e-05, "loss": 2.965, "step": 37885 }, { "epoch": 2.97, "learning_rate": 7.029248020073708e-05, "loss": 2.9237, "step": 37886 }, { "epoch": 2.97, "learning_rate": 7.029169607151259e-05, "loss": 2.8408, "step": 37887 }, { "epoch": 2.97, "learning_rate": 7.02909119422881e-05, "loss": 2.7211, "step": 37888 }, { "epoch": 2.97, "learning_rate": 7.02901278130636e-05, "loss": 2.7281, "step": 37889 }, { "epoch": 2.97, "learning_rate": 7.028934368383911e-05, "loss": 2.8058, "step": 37890 }, { "epoch": 2.97, "learning_rate": 7.028855955461461e-05, "loss": 2.8483, "step": 37891 }, { "epoch": 2.97, "learning_rate": 7.028777542539012e-05, "loss": 2.6697, "step": 37892 }, { "epoch": 2.97, "learning_rate": 7.028699129616562e-05, "loss": 2.8396, "step": 37893 }, { "epoch": 2.97, "learning_rate": 7.028620716694112e-05, "loss": 2.9144, "step": 37894 }, { "epoch": 2.97, "learning_rate": 7.028542303771663e-05, "loss": 2.9067, "step": 37895 }, { "epoch": 2.97, "learning_rate": 7.028463890849212e-05, "loss": 2.5285, "step": 37896 }, { "epoch": 2.97, "learning_rate": 7.028385477926762e-05, "loss": 2.809, "step": 37897 }, { "epoch": 2.97, "learning_rate": 7.028307065004313e-05, "loss": 2.7922, "step": 37898 }, { "epoch": 2.97, "learning_rate": 7.028228652081863e-05, "loss": 2.8807, "step": 37899 }, { "epoch": 2.97, "learning_rate": 7.028150239159413e-05, "loss": 2.857, "step": 37900 }, { "epoch": 2.97, "learning_rate": 7.028071826236964e-05, "loss": 2.7153, "step": 37901 }, { "epoch": 2.97, "learning_rate": 7.027993413314514e-05, "loss": 2.9508, "step": 37902 }, { "epoch": 2.97, "learning_rate": 7.027915000392064e-05, "loss": 2.6978, "step": 37903 }, { "epoch": 2.97, "learning_rate": 7.027836587469615e-05, "loss": 2.6682, "step": 37904 }, { "epoch": 2.97, "learning_rate": 7.027758174547165e-05, "loss": 3.1263, "step": 37905 }, { "epoch": 2.97, "learning_rate": 7.027679761624715e-05, "loss": 2.928, "step": 37906 }, { "epoch": 2.97, "learning_rate": 7.027601348702266e-05, "loss": 2.9214, "step": 37907 }, { "epoch": 2.97, "learning_rate": 7.027522935779816e-05, "loss": 2.8862, "step": 37908 }, { "epoch": 2.97, "learning_rate": 7.027444522857367e-05, "loss": 2.7321, "step": 37909 }, { "epoch": 2.97, "learning_rate": 7.027366109934917e-05, "loss": 2.8192, "step": 37910 }, { "epoch": 2.97, "learning_rate": 7.027287697012467e-05, "loss": 2.8679, "step": 37911 }, { "epoch": 2.97, "learning_rate": 7.027209284090018e-05, "loss": 3.0493, "step": 37912 }, { "epoch": 2.97, "learning_rate": 7.027130871167568e-05, "loss": 2.7742, "step": 37913 }, { "epoch": 2.97, "learning_rate": 7.02705245824512e-05, "loss": 2.7534, "step": 37914 }, { "epoch": 2.97, "learning_rate": 7.02697404532267e-05, "loss": 2.9369, "step": 37915 }, { "epoch": 2.97, "learning_rate": 7.02689563240022e-05, "loss": 2.7593, "step": 37916 }, { "epoch": 2.97, "learning_rate": 7.026817219477771e-05, "loss": 2.8771, "step": 37917 }, { "epoch": 2.97, "learning_rate": 7.026738806555321e-05, "loss": 2.6617, "step": 37918 }, { "epoch": 2.97, "learning_rate": 7.026660393632872e-05, "loss": 2.79, "step": 37919 }, { "epoch": 2.97, "learning_rate": 7.026581980710422e-05, "loss": 2.7875, "step": 37920 }, { "epoch": 2.97, "learning_rate": 7.026503567787972e-05, "loss": 2.6288, "step": 37921 }, { "epoch": 2.97, "learning_rate": 7.026425154865523e-05, "loss": 2.9391, "step": 37922 }, { "epoch": 2.97, "learning_rate": 7.026346741943073e-05, "loss": 2.7748, "step": 37923 }, { "epoch": 2.97, "learning_rate": 7.026268329020623e-05, "loss": 2.7423, "step": 37924 }, { "epoch": 2.97, "learning_rate": 7.026189916098174e-05, "loss": 2.9632, "step": 37925 }, { "epoch": 2.97, "learning_rate": 7.026111503175724e-05, "loss": 2.7971, "step": 37926 }, { "epoch": 2.97, "learning_rate": 7.026033090253275e-05, "loss": 2.6945, "step": 37927 }, { "epoch": 2.97, "learning_rate": 7.025954677330825e-05, "loss": 2.9106, "step": 37928 }, { "epoch": 2.97, "learning_rate": 7.025876264408375e-05, "loss": 2.6904, "step": 37929 }, { "epoch": 2.97, "learning_rate": 7.025797851485926e-05, "loss": 2.9382, "step": 37930 }, { "epoch": 2.97, "learning_rate": 7.025719438563476e-05, "loss": 3.0686, "step": 37931 }, { "epoch": 2.97, "learning_rate": 7.025641025641025e-05, "loss": 2.7811, "step": 37932 }, { "epoch": 2.97, "learning_rate": 7.025562612718575e-05, "loss": 2.8419, "step": 37933 }, { "epoch": 2.97, "learning_rate": 7.025484199796126e-05, "loss": 2.6706, "step": 37934 }, { "epoch": 2.97, "learning_rate": 7.025405786873676e-05, "loss": 2.7045, "step": 37935 }, { "epoch": 2.97, "learning_rate": 7.025327373951227e-05, "loss": 2.6965, "step": 37936 }, { "epoch": 2.97, "learning_rate": 7.025248961028777e-05, "loss": 2.9815, "step": 37937 }, { "epoch": 2.97, "learning_rate": 7.025170548106327e-05, "loss": 2.7771, "step": 37938 }, { "epoch": 2.97, "learning_rate": 7.025092135183879e-05, "loss": 2.9309, "step": 37939 }, { "epoch": 2.97, "learning_rate": 7.02501372226143e-05, "loss": 2.5809, "step": 37940 }, { "epoch": 2.98, "learning_rate": 7.02493530933898e-05, "loss": 2.8764, "step": 37941 }, { "epoch": 2.98, "learning_rate": 7.02485689641653e-05, "loss": 2.7373, "step": 37942 }, { "epoch": 2.98, "learning_rate": 7.02477848349408e-05, "loss": 2.588, "step": 37943 }, { "epoch": 2.98, "learning_rate": 7.024700070571631e-05, "loss": 2.8035, "step": 37944 }, { "epoch": 2.98, "learning_rate": 7.024621657649181e-05, "loss": 2.8228, "step": 37945 }, { "epoch": 2.98, "learning_rate": 7.024543244726732e-05, "loss": 2.8339, "step": 37946 }, { "epoch": 2.98, "learning_rate": 7.024464831804282e-05, "loss": 2.9225, "step": 37947 }, { "epoch": 2.98, "learning_rate": 7.024386418881832e-05, "loss": 2.5271, "step": 37948 }, { "epoch": 2.98, "learning_rate": 7.024308005959383e-05, "loss": 2.8014, "step": 37949 }, { "epoch": 2.98, "learning_rate": 7.024229593036933e-05, "loss": 2.886, "step": 37950 }, { "epoch": 2.98, "learning_rate": 7.024151180114483e-05, "loss": 2.9045, "step": 37951 }, { "epoch": 2.98, "learning_rate": 7.024072767192034e-05, "loss": 3.0754, "step": 37952 }, { "epoch": 2.98, "learning_rate": 7.023994354269584e-05, "loss": 3.0344, "step": 37953 }, { "epoch": 2.98, "learning_rate": 7.023915941347135e-05, "loss": 2.7533, "step": 37954 }, { "epoch": 2.98, "learning_rate": 7.023837528424685e-05, "loss": 2.8135, "step": 37955 }, { "epoch": 2.98, "learning_rate": 7.023759115502235e-05, "loss": 2.8532, "step": 37956 }, { "epoch": 2.98, "learning_rate": 7.023680702579786e-05, "loss": 2.7036, "step": 37957 }, { "epoch": 2.98, "learning_rate": 7.023602289657336e-05, "loss": 2.7393, "step": 37958 }, { "epoch": 2.98, "learning_rate": 7.023523876734886e-05, "loss": 2.7173, "step": 37959 }, { "epoch": 2.98, "learning_rate": 7.023445463812437e-05, "loss": 2.8958, "step": 37960 }, { "epoch": 2.98, "learning_rate": 7.023367050889987e-05, "loss": 2.8568, "step": 37961 }, { "epoch": 2.98, "learning_rate": 7.023288637967537e-05, "loss": 2.6771, "step": 37962 }, { "epoch": 2.98, "learning_rate": 7.023210225045088e-05, "loss": 2.755, "step": 37963 }, { "epoch": 2.98, "learning_rate": 7.023131812122638e-05, "loss": 2.6989, "step": 37964 }, { "epoch": 2.98, "learning_rate": 7.023053399200189e-05, "loss": 2.8039, "step": 37965 }, { "epoch": 2.98, "learning_rate": 7.022974986277739e-05, "loss": 2.8185, "step": 37966 }, { "epoch": 2.98, "learning_rate": 7.022896573355289e-05, "loss": 3.2892, "step": 37967 }, { "epoch": 2.98, "learning_rate": 7.02281816043284e-05, "loss": 2.8127, "step": 37968 }, { "epoch": 2.98, "learning_rate": 7.02273974751039e-05, "loss": 2.6374, "step": 37969 }, { "epoch": 2.98, "learning_rate": 7.02266133458794e-05, "loss": 2.982, "step": 37970 }, { "epoch": 2.98, "learning_rate": 7.022582921665491e-05, "loss": 3.0442, "step": 37971 }, { "epoch": 2.98, "learning_rate": 7.022504508743041e-05, "loss": 2.6566, "step": 37972 }, { "epoch": 2.98, "learning_rate": 7.022426095820591e-05, "loss": 2.756, "step": 37973 }, { "epoch": 2.98, "learning_rate": 7.022347682898142e-05, "loss": 3.1115, "step": 37974 }, { "epoch": 2.98, "learning_rate": 7.022269269975692e-05, "loss": 2.6462, "step": 37975 }, { "epoch": 2.98, "learning_rate": 7.022190857053243e-05, "loss": 2.8504, "step": 37976 }, { "epoch": 2.98, "learning_rate": 7.022112444130793e-05, "loss": 2.9056, "step": 37977 }, { "epoch": 2.98, "learning_rate": 7.022034031208343e-05, "loss": 2.7615, "step": 37978 }, { "epoch": 2.98, "learning_rate": 7.021955618285894e-05, "loss": 2.844, "step": 37979 }, { "epoch": 2.98, "learning_rate": 7.021877205363444e-05, "loss": 2.8404, "step": 37980 }, { "epoch": 2.98, "learning_rate": 7.021798792440994e-05, "loss": 3.0188, "step": 37981 }, { "epoch": 2.98, "learning_rate": 7.021720379518545e-05, "loss": 2.7448, "step": 37982 }, { "epoch": 2.98, "learning_rate": 7.021641966596095e-05, "loss": 2.7291, "step": 37983 }, { "epoch": 2.98, "learning_rate": 7.021563553673646e-05, "loss": 2.7385, "step": 37984 }, { "epoch": 2.98, "learning_rate": 7.021485140751196e-05, "loss": 2.715, "step": 37985 }, { "epoch": 2.98, "learning_rate": 7.021406727828746e-05, "loss": 2.9334, "step": 37986 }, { "epoch": 2.98, "learning_rate": 7.021328314906297e-05, "loss": 3.1062, "step": 37987 }, { "epoch": 2.98, "learning_rate": 7.021249901983847e-05, "loss": 3.1211, "step": 37988 }, { "epoch": 2.98, "learning_rate": 7.021171489061397e-05, "loss": 2.8373, "step": 37989 }, { "epoch": 2.98, "learning_rate": 7.021093076138948e-05, "loss": 2.9162, "step": 37990 }, { "epoch": 2.98, "learning_rate": 7.021014663216498e-05, "loss": 2.8239, "step": 37991 }, { "epoch": 2.98, "learning_rate": 7.020936250294048e-05, "loss": 3.1185, "step": 37992 }, { "epoch": 2.98, "learning_rate": 7.020857837371599e-05, "loss": 2.7041, "step": 37993 }, { "epoch": 2.98, "learning_rate": 7.020779424449149e-05, "loss": 2.7144, "step": 37994 }, { "epoch": 2.98, "learning_rate": 7.0207010115267e-05, "loss": 2.8371, "step": 37995 }, { "epoch": 2.98, "learning_rate": 7.020622598604251e-05, "loss": 2.8179, "step": 37996 }, { "epoch": 2.98, "learning_rate": 7.020544185681802e-05, "loss": 2.7595, "step": 37997 }, { "epoch": 2.98, "learning_rate": 7.020465772759352e-05, "loss": 2.7507, "step": 37998 }, { "epoch": 2.98, "learning_rate": 7.020387359836902e-05, "loss": 2.5193, "step": 37999 }, { "epoch": 2.98, "learning_rate": 7.020308946914453e-05, "loss": 3.2266, "step": 38000 }, { "epoch": 2.98, "eval_loss": 2.8900797367095947, "eval_runtime": 218.1749, "eval_samples_per_second": 51.94, "eval_steps_per_second": 3.25, "step": 38000 }, { "epoch": 2.98, "learning_rate": 7.020230533992003e-05, "loss": 2.5755, "step": 38001 }, { "epoch": 2.98, "learning_rate": 7.020152121069554e-05, "loss": 2.7717, "step": 38002 }, { "epoch": 2.98, "learning_rate": 7.020073708147104e-05, "loss": 2.8621, "step": 38003 }, { "epoch": 2.98, "learning_rate": 7.019995295224653e-05, "loss": 2.8104, "step": 38004 }, { "epoch": 2.98, "learning_rate": 7.019916882302203e-05, "loss": 2.6917, "step": 38005 }, { "epoch": 2.98, "learning_rate": 7.019838469379754e-05, "loss": 2.6856, "step": 38006 }, { "epoch": 2.98, "learning_rate": 7.019760056457304e-05, "loss": 2.7597, "step": 38007 }, { "epoch": 2.98, "learning_rate": 7.019681643534854e-05, "loss": 2.8443, "step": 38008 }, { "epoch": 2.98, "learning_rate": 7.019603230612405e-05, "loss": 2.6774, "step": 38009 }, { "epoch": 2.98, "learning_rate": 7.019524817689955e-05, "loss": 2.7343, "step": 38010 }, { "epoch": 2.98, "learning_rate": 7.019446404767505e-05, "loss": 2.7365, "step": 38011 }, { "epoch": 2.98, "learning_rate": 7.019367991845056e-05, "loss": 2.8649, "step": 38012 }, { "epoch": 2.98, "learning_rate": 7.019289578922606e-05, "loss": 3.1083, "step": 38013 }, { "epoch": 2.98, "learning_rate": 7.019211166000157e-05, "loss": 2.9188, "step": 38014 }, { "epoch": 2.98, "learning_rate": 7.019132753077707e-05, "loss": 2.8129, "step": 38015 }, { "epoch": 2.98, "learning_rate": 7.019054340155257e-05, "loss": 2.8467, "step": 38016 }, { "epoch": 2.98, "learning_rate": 7.018975927232808e-05, "loss": 2.8998, "step": 38017 }, { "epoch": 2.98, "learning_rate": 7.018897514310358e-05, "loss": 2.7654, "step": 38018 }, { "epoch": 2.98, "learning_rate": 7.018819101387908e-05, "loss": 2.7612, "step": 38019 }, { "epoch": 2.98, "learning_rate": 7.018740688465459e-05, "loss": 2.8539, "step": 38020 }, { "epoch": 2.98, "learning_rate": 7.018662275543009e-05, "loss": 2.7643, "step": 38021 }, { "epoch": 2.98, "learning_rate": 7.018583862620561e-05, "loss": 2.8616, "step": 38022 }, { "epoch": 2.98, "learning_rate": 7.018505449698111e-05, "loss": 2.6723, "step": 38023 }, { "epoch": 2.98, "learning_rate": 7.018427036775662e-05, "loss": 3.0094, "step": 38024 }, { "epoch": 2.98, "learning_rate": 7.018348623853212e-05, "loss": 2.6371, "step": 38025 }, { "epoch": 2.98, "learning_rate": 7.018270210930762e-05, "loss": 2.7274, "step": 38026 }, { "epoch": 2.98, "learning_rate": 7.018191798008313e-05, "loss": 2.8807, "step": 38027 }, { "epoch": 2.98, "learning_rate": 7.018113385085863e-05, "loss": 2.748, "step": 38028 }, { "epoch": 2.98, "learning_rate": 7.018034972163413e-05, "loss": 2.7877, "step": 38029 }, { "epoch": 2.98, "learning_rate": 7.017956559240964e-05, "loss": 2.7354, "step": 38030 }, { "epoch": 2.98, "learning_rate": 7.017878146318514e-05, "loss": 3.0192, "step": 38031 }, { "epoch": 2.98, "learning_rate": 7.017799733396065e-05, "loss": 2.8666, "step": 38032 }, { "epoch": 2.98, "learning_rate": 7.017721320473615e-05, "loss": 2.8087, "step": 38033 }, { "epoch": 2.98, "learning_rate": 7.017642907551165e-05, "loss": 2.6305, "step": 38034 }, { "epoch": 2.98, "learning_rate": 7.017564494628716e-05, "loss": 2.7842, "step": 38035 }, { "epoch": 2.98, "learning_rate": 7.017486081706266e-05, "loss": 2.783, "step": 38036 }, { "epoch": 2.98, "learning_rate": 7.017407668783816e-05, "loss": 2.8788, "step": 38037 }, { "epoch": 2.98, "learning_rate": 7.017329255861367e-05, "loss": 2.8208, "step": 38038 }, { "epoch": 2.98, "learning_rate": 7.017250842938917e-05, "loss": 2.8667, "step": 38039 }, { "epoch": 2.98, "learning_rate": 7.017172430016466e-05, "loss": 2.6838, "step": 38040 }, { "epoch": 2.98, "learning_rate": 7.017094017094017e-05, "loss": 2.7112, "step": 38041 }, { "epoch": 2.98, "learning_rate": 7.017015604171567e-05, "loss": 2.9591, "step": 38042 }, { "epoch": 2.98, "learning_rate": 7.016937191249117e-05, "loss": 2.7739, "step": 38043 }, { "epoch": 2.98, "learning_rate": 7.016858778326668e-05, "loss": 2.8404, "step": 38044 }, { "epoch": 2.98, "learning_rate": 7.016780365404218e-05, "loss": 2.8224, "step": 38045 }, { "epoch": 2.98, "learning_rate": 7.016701952481768e-05, "loss": 2.7281, "step": 38046 }, { "epoch": 2.98, "learning_rate": 7.01662353955932e-05, "loss": 2.8239, "step": 38047 }, { "epoch": 2.98, "learning_rate": 7.01654512663687e-05, "loss": 2.7464, "step": 38048 }, { "epoch": 2.98, "learning_rate": 7.016466713714421e-05, "loss": 2.6902, "step": 38049 }, { "epoch": 2.98, "learning_rate": 7.016388300791971e-05, "loss": 2.867, "step": 38050 }, { "epoch": 2.98, "learning_rate": 7.016309887869522e-05, "loss": 2.8294, "step": 38051 }, { "epoch": 2.98, "learning_rate": 7.016231474947072e-05, "loss": 2.6661, "step": 38052 }, { "epoch": 2.98, "learning_rate": 7.016153062024622e-05, "loss": 2.8823, "step": 38053 }, { "epoch": 2.98, "learning_rate": 7.016074649102173e-05, "loss": 2.7781, "step": 38054 }, { "epoch": 2.98, "learning_rate": 7.015996236179723e-05, "loss": 2.7896, "step": 38055 }, { "epoch": 2.98, "learning_rate": 7.015917823257273e-05, "loss": 2.6646, "step": 38056 }, { "epoch": 2.98, "learning_rate": 7.015839410334824e-05, "loss": 2.7514, "step": 38057 }, { "epoch": 2.98, "learning_rate": 7.015760997412374e-05, "loss": 2.4999, "step": 38058 }, { "epoch": 2.98, "learning_rate": 7.015682584489924e-05, "loss": 2.7523, "step": 38059 }, { "epoch": 2.98, "learning_rate": 7.015604171567475e-05, "loss": 2.6798, "step": 38060 }, { "epoch": 2.98, "learning_rate": 7.015525758645025e-05, "loss": 2.6211, "step": 38061 }, { "epoch": 2.98, "learning_rate": 7.015447345722576e-05, "loss": 2.7978, "step": 38062 }, { "epoch": 2.98, "learning_rate": 7.015368932800126e-05, "loss": 2.6918, "step": 38063 }, { "epoch": 2.98, "learning_rate": 7.015290519877676e-05, "loss": 2.7209, "step": 38064 }, { "epoch": 2.98, "learning_rate": 7.015212106955227e-05, "loss": 2.7517, "step": 38065 }, { "epoch": 2.98, "learning_rate": 7.015133694032777e-05, "loss": 2.7505, "step": 38066 }, { "epoch": 2.98, "learning_rate": 7.015055281110327e-05, "loss": 2.7411, "step": 38067 }, { "epoch": 2.99, "learning_rate": 7.014976868187878e-05, "loss": 3.1566, "step": 38068 }, { "epoch": 2.99, "learning_rate": 7.014898455265428e-05, "loss": 2.5257, "step": 38069 }, { "epoch": 2.99, "learning_rate": 7.014820042342979e-05, "loss": 2.5875, "step": 38070 }, { "epoch": 2.99, "learning_rate": 7.014741629420529e-05, "loss": 2.6797, "step": 38071 }, { "epoch": 2.99, "learning_rate": 7.014663216498079e-05, "loss": 2.8886, "step": 38072 }, { "epoch": 2.99, "learning_rate": 7.01458480357563e-05, "loss": 2.8207, "step": 38073 }, { "epoch": 2.99, "learning_rate": 7.01450639065318e-05, "loss": 2.7661, "step": 38074 }, { "epoch": 2.99, "learning_rate": 7.01442797773073e-05, "loss": 2.8261, "step": 38075 }, { "epoch": 2.99, "learning_rate": 7.014349564808281e-05, "loss": 2.8272, "step": 38076 }, { "epoch": 2.99, "learning_rate": 7.014271151885831e-05, "loss": 2.7239, "step": 38077 }, { "epoch": 2.99, "learning_rate": 7.014192738963381e-05, "loss": 2.8435, "step": 38078 }, { "epoch": 2.99, "learning_rate": 7.014114326040932e-05, "loss": 2.7999, "step": 38079 }, { "epoch": 2.99, "learning_rate": 7.014035913118482e-05, "loss": 2.7032, "step": 38080 }, { "epoch": 2.99, "learning_rate": 7.013957500196033e-05, "loss": 2.6021, "step": 38081 }, { "epoch": 2.99, "learning_rate": 7.013879087273583e-05, "loss": 2.7832, "step": 38082 }, { "epoch": 2.99, "learning_rate": 7.013800674351133e-05, "loss": 2.7896, "step": 38083 }, { "epoch": 2.99, "learning_rate": 7.013722261428684e-05, "loss": 2.932, "step": 38084 }, { "epoch": 2.99, "learning_rate": 7.013643848506234e-05, "loss": 2.5824, "step": 38085 }, { "epoch": 2.99, "learning_rate": 7.013565435583784e-05, "loss": 2.8374, "step": 38086 }, { "epoch": 2.99, "learning_rate": 7.013487022661335e-05, "loss": 2.631, "step": 38087 }, { "epoch": 2.99, "learning_rate": 7.013408609738885e-05, "loss": 2.7563, "step": 38088 }, { "epoch": 2.99, "learning_rate": 7.013330196816436e-05, "loss": 3.1122, "step": 38089 }, { "epoch": 2.99, "learning_rate": 7.013251783893986e-05, "loss": 3.1676, "step": 38090 }, { "epoch": 2.99, "learning_rate": 7.013173370971536e-05, "loss": 2.6517, "step": 38091 }, { "epoch": 2.99, "learning_rate": 7.013094958049087e-05, "loss": 2.7983, "step": 38092 }, { "epoch": 2.99, "learning_rate": 7.013016545126637e-05, "loss": 2.8335, "step": 38093 }, { "epoch": 2.99, "learning_rate": 7.012938132204187e-05, "loss": 2.6165, "step": 38094 }, { "epoch": 2.99, "learning_rate": 7.012859719281738e-05, "loss": 2.877, "step": 38095 }, { "epoch": 2.99, "learning_rate": 7.012781306359288e-05, "loss": 2.9352, "step": 38096 }, { "epoch": 2.99, "learning_rate": 7.012702893436838e-05, "loss": 2.804, "step": 38097 }, { "epoch": 2.99, "learning_rate": 7.012624480514389e-05, "loss": 2.8915, "step": 38098 }, { "epoch": 2.99, "learning_rate": 7.012546067591939e-05, "loss": 2.9458, "step": 38099 }, { "epoch": 2.99, "learning_rate": 7.01246765466949e-05, "loss": 2.8375, "step": 38100 }, { "epoch": 2.99, "learning_rate": 7.01238924174704e-05, "loss": 2.7629, "step": 38101 }, { "epoch": 2.99, "learning_rate": 7.01231082882459e-05, "loss": 2.831, "step": 38102 }, { "epoch": 2.99, "learning_rate": 7.01223241590214e-05, "loss": 2.6689, "step": 38103 }, { "epoch": 2.99, "learning_rate": 7.012154002979692e-05, "loss": 2.7784, "step": 38104 }, { "epoch": 2.99, "learning_rate": 7.012075590057243e-05, "loss": 3.0262, "step": 38105 }, { "epoch": 2.99, "learning_rate": 7.011997177134793e-05, "loss": 2.8981, "step": 38106 }, { "epoch": 2.99, "learning_rate": 7.011918764212343e-05, "loss": 2.6559, "step": 38107 }, { "epoch": 2.99, "learning_rate": 7.011840351289894e-05, "loss": 2.9001, "step": 38108 }, { "epoch": 2.99, "learning_rate": 7.011761938367444e-05, "loss": 2.8989, "step": 38109 }, { "epoch": 2.99, "learning_rate": 7.011683525444995e-05, "loss": 2.8022, "step": 38110 }, { "epoch": 2.99, "learning_rate": 7.011605112522545e-05, "loss": 2.7945, "step": 38111 }, { "epoch": 2.99, "learning_rate": 7.011526699600094e-05, "loss": 2.9535, "step": 38112 }, { "epoch": 2.99, "learning_rate": 7.011448286677644e-05, "loss": 2.7919, "step": 38113 }, { "epoch": 2.99, "learning_rate": 7.011369873755195e-05, "loss": 2.883, "step": 38114 }, { "epoch": 2.99, "learning_rate": 7.011291460832745e-05, "loss": 2.7842, "step": 38115 }, { "epoch": 2.99, "learning_rate": 7.011213047910295e-05, "loss": 2.7506, "step": 38116 }, { "epoch": 2.99, "learning_rate": 7.011134634987846e-05, "loss": 2.7294, "step": 38117 }, { "epoch": 2.99, "learning_rate": 7.011056222065396e-05, "loss": 2.7531, "step": 38118 }, { "epoch": 2.99, "learning_rate": 7.010977809142947e-05, "loss": 2.6632, "step": 38119 }, { "epoch": 2.99, "learning_rate": 7.010899396220497e-05, "loss": 2.9086, "step": 38120 }, { "epoch": 2.99, "learning_rate": 7.010820983298047e-05, "loss": 3.0017, "step": 38121 }, { "epoch": 2.99, "learning_rate": 7.010742570375598e-05, "loss": 2.9189, "step": 38122 }, { "epoch": 2.99, "learning_rate": 7.010664157453148e-05, "loss": 3.0054, "step": 38123 }, { "epoch": 2.99, "learning_rate": 7.010585744530698e-05, "loss": 2.7862, "step": 38124 }, { "epoch": 2.99, "learning_rate": 7.010507331608249e-05, "loss": 2.8525, "step": 38125 }, { "epoch": 2.99, "learning_rate": 7.010428918685799e-05, "loss": 2.7773, "step": 38126 }, { "epoch": 2.99, "learning_rate": 7.01035050576335e-05, "loss": 2.7569, "step": 38127 }, { "epoch": 2.99, "learning_rate": 7.0102720928409e-05, "loss": 2.7961, "step": 38128 }, { "epoch": 2.99, "learning_rate": 7.01019367991845e-05, "loss": 2.807, "step": 38129 }, { "epoch": 2.99, "learning_rate": 7.010115266996002e-05, "loss": 2.6742, "step": 38130 }, { "epoch": 2.99, "learning_rate": 7.010036854073552e-05, "loss": 2.8189, "step": 38131 }, { "epoch": 2.99, "learning_rate": 7.009958441151103e-05, "loss": 2.5812, "step": 38132 }, { "epoch": 2.99, "learning_rate": 7.009880028228653e-05, "loss": 2.9569, "step": 38133 }, { "epoch": 2.99, "learning_rate": 7.009801615306203e-05, "loss": 2.932, "step": 38134 }, { "epoch": 2.99, "learning_rate": 7.009723202383754e-05, "loss": 2.8632, "step": 38135 }, { "epoch": 2.99, "learning_rate": 7.009644789461304e-05, "loss": 2.8427, "step": 38136 }, { "epoch": 2.99, "learning_rate": 7.009566376538855e-05, "loss": 2.7437, "step": 38137 }, { "epoch": 2.99, "learning_rate": 7.009487963616405e-05, "loss": 3.0429, "step": 38138 }, { "epoch": 2.99, "learning_rate": 7.009409550693955e-05, "loss": 2.7354, "step": 38139 }, { "epoch": 2.99, "learning_rate": 7.009331137771506e-05, "loss": 2.728, "step": 38140 }, { "epoch": 2.99, "learning_rate": 7.009252724849056e-05, "loss": 2.9556, "step": 38141 }, { "epoch": 2.99, "learning_rate": 7.009174311926606e-05, "loss": 3.1307, "step": 38142 }, { "epoch": 2.99, "learning_rate": 7.009095899004157e-05, "loss": 2.7398, "step": 38143 }, { "epoch": 2.99, "learning_rate": 7.009017486081707e-05, "loss": 2.7873, "step": 38144 }, { "epoch": 2.99, "learning_rate": 7.008939073159257e-05, "loss": 2.7239, "step": 38145 }, { "epoch": 2.99, "learning_rate": 7.008860660236808e-05, "loss": 2.7316, "step": 38146 }, { "epoch": 2.99, "learning_rate": 7.008782247314358e-05, "loss": 3.0208, "step": 38147 }, { "epoch": 2.99, "learning_rate": 7.008703834391907e-05, "loss": 2.677, "step": 38148 }, { "epoch": 2.99, "learning_rate": 7.008625421469458e-05, "loss": 2.898, "step": 38149 }, { "epoch": 2.99, "learning_rate": 7.008547008547008e-05, "loss": 2.6617, "step": 38150 }, { "epoch": 2.99, "learning_rate": 7.008468595624558e-05, "loss": 2.7635, "step": 38151 }, { "epoch": 2.99, "learning_rate": 7.008390182702109e-05, "loss": 2.8569, "step": 38152 }, { "epoch": 2.99, "learning_rate": 7.008311769779659e-05, "loss": 2.714, "step": 38153 }, { "epoch": 2.99, "learning_rate": 7.00823335685721e-05, "loss": 2.8563, "step": 38154 }, { "epoch": 2.99, "learning_rate": 7.008154943934761e-05, "loss": 2.5943, "step": 38155 }, { "epoch": 2.99, "learning_rate": 7.008076531012312e-05, "loss": 2.7673, "step": 38156 }, { "epoch": 2.99, "learning_rate": 7.007998118089862e-05, "loss": 2.7135, "step": 38157 }, { "epoch": 2.99, "learning_rate": 7.007919705167412e-05, "loss": 2.898, "step": 38158 }, { "epoch": 2.99, "learning_rate": 7.007841292244963e-05, "loss": 2.8413, "step": 38159 }, { "epoch": 2.99, "learning_rate": 7.007762879322513e-05, "loss": 2.6991, "step": 38160 }, { "epoch": 2.99, "learning_rate": 7.007684466400063e-05, "loss": 2.9213, "step": 38161 }, { "epoch": 2.99, "learning_rate": 7.007606053477614e-05, "loss": 3.0156, "step": 38162 }, { "epoch": 2.99, "learning_rate": 7.007527640555164e-05, "loss": 2.7684, "step": 38163 }, { "epoch": 2.99, "learning_rate": 7.007449227632714e-05, "loss": 2.6534, "step": 38164 }, { "epoch": 2.99, "learning_rate": 7.007370814710265e-05, "loss": 2.7919, "step": 38165 }, { "epoch": 2.99, "learning_rate": 7.007292401787815e-05, "loss": 2.8203, "step": 38166 }, { "epoch": 2.99, "learning_rate": 7.007213988865366e-05, "loss": 2.9621, "step": 38167 }, { "epoch": 2.99, "learning_rate": 7.007135575942916e-05, "loss": 2.8482, "step": 38168 }, { "epoch": 2.99, "learning_rate": 7.007057163020466e-05, "loss": 2.8807, "step": 38169 }, { "epoch": 2.99, "learning_rate": 7.006978750098017e-05, "loss": 2.8046, "step": 38170 }, { "epoch": 2.99, "learning_rate": 7.006900337175567e-05, "loss": 2.7486, "step": 38171 }, { "epoch": 2.99, "learning_rate": 7.006821924253117e-05, "loss": 2.9345, "step": 38172 }, { "epoch": 2.99, "learning_rate": 7.006743511330668e-05, "loss": 3.1022, "step": 38173 }, { "epoch": 2.99, "learning_rate": 7.006665098408218e-05, "loss": 2.8609, "step": 38174 }, { "epoch": 2.99, "learning_rate": 7.006586685485768e-05, "loss": 2.7359, "step": 38175 }, { "epoch": 2.99, "learning_rate": 7.006508272563319e-05, "loss": 2.7237, "step": 38176 }, { "epoch": 2.99, "learning_rate": 7.006429859640869e-05, "loss": 2.598, "step": 38177 }, { "epoch": 2.99, "learning_rate": 7.00635144671842e-05, "loss": 2.5798, "step": 38178 }, { "epoch": 2.99, "learning_rate": 7.00627303379597e-05, "loss": 2.8825, "step": 38179 }, { "epoch": 2.99, "learning_rate": 7.00619462087352e-05, "loss": 2.9129, "step": 38180 }, { "epoch": 2.99, "learning_rate": 7.006116207951071e-05, "loss": 3.009, "step": 38181 }, { "epoch": 2.99, "learning_rate": 7.006037795028621e-05, "loss": 2.788, "step": 38182 }, { "epoch": 2.99, "learning_rate": 7.005959382106171e-05, "loss": 2.8402, "step": 38183 }, { "epoch": 2.99, "learning_rate": 7.005880969183722e-05, "loss": 2.9005, "step": 38184 }, { "epoch": 2.99, "learning_rate": 7.005802556261272e-05, "loss": 2.8217, "step": 38185 }, { "epoch": 2.99, "learning_rate": 7.005724143338823e-05, "loss": 3.0105, "step": 38186 }, { "epoch": 2.99, "learning_rate": 7.005645730416373e-05, "loss": 2.8347, "step": 38187 }, { "epoch": 2.99, "learning_rate": 7.005567317493923e-05, "loss": 2.8894, "step": 38188 }, { "epoch": 2.99, "learning_rate": 7.005488904571474e-05, "loss": 2.7584, "step": 38189 }, { "epoch": 2.99, "learning_rate": 7.005410491649024e-05, "loss": 2.9292, "step": 38190 }, { "epoch": 2.99, "learning_rate": 7.005332078726574e-05, "loss": 2.9221, "step": 38191 }, { "epoch": 2.99, "learning_rate": 7.005253665804125e-05, "loss": 2.7656, "step": 38192 }, { "epoch": 2.99, "learning_rate": 7.005175252881675e-05, "loss": 2.841, "step": 38193 }, { "epoch": 2.99, "learning_rate": 7.005096839959225e-05, "loss": 2.763, "step": 38194 }, { "epoch": 2.99, "learning_rate": 7.005018427036776e-05, "loss": 3.1737, "step": 38195 }, { "epoch": 3.0, "learning_rate": 7.004940014114326e-05, "loss": 2.8067, "step": 38196 }, { "epoch": 3.0, "learning_rate": 7.004861601191877e-05, "loss": 2.9938, "step": 38197 }, { "epoch": 3.0, "learning_rate": 7.004783188269427e-05, "loss": 2.6951, "step": 38198 }, { "epoch": 3.0, "learning_rate": 7.004704775346977e-05, "loss": 2.9274, "step": 38199 }, { "epoch": 3.0, "learning_rate": 7.004626362424528e-05, "loss": 2.8019, "step": 38200 }, { "epoch": 3.0, "learning_rate": 7.004547949502078e-05, "loss": 2.9442, "step": 38201 }, { "epoch": 3.0, "learning_rate": 7.004469536579628e-05, "loss": 3.1628, "step": 38202 }, { "epoch": 3.0, "learning_rate": 7.004391123657179e-05, "loss": 2.9739, "step": 38203 }, { "epoch": 3.0, "learning_rate": 7.004312710734729e-05, "loss": 2.6798, "step": 38204 }, { "epoch": 3.0, "learning_rate": 7.00423429781228e-05, "loss": 2.653, "step": 38205 }, { "epoch": 3.0, "learning_rate": 7.00415588488983e-05, "loss": 2.8098, "step": 38206 }, { "epoch": 3.0, "learning_rate": 7.00407747196738e-05, "loss": 2.9281, "step": 38207 }, { "epoch": 3.0, "learning_rate": 7.00399905904493e-05, "loss": 2.8264, "step": 38208 }, { "epoch": 3.0, "learning_rate": 7.003920646122481e-05, "loss": 2.8364, "step": 38209 }, { "epoch": 3.0, "learning_rate": 7.003842233200031e-05, "loss": 2.7275, "step": 38210 }, { "epoch": 3.0, "learning_rate": 7.003763820277582e-05, "loss": 2.6879, "step": 38211 }, { "epoch": 3.0, "learning_rate": 7.003685407355133e-05, "loss": 2.8486, "step": 38212 }, { "epoch": 3.0, "learning_rate": 7.003606994432684e-05, "loss": 2.8161, "step": 38213 }, { "epoch": 3.0, "learning_rate": 7.003528581510234e-05, "loss": 2.7694, "step": 38214 }, { "epoch": 3.0, "learning_rate": 7.003450168587785e-05, "loss": 2.8381, "step": 38215 }, { "epoch": 3.0, "learning_rate": 7.003371755665335e-05, "loss": 2.79, "step": 38216 }, { "epoch": 3.0, "learning_rate": 7.003293342742885e-05, "loss": 2.5041, "step": 38217 }, { "epoch": 3.0, "learning_rate": 7.003214929820436e-05, "loss": 2.7634, "step": 38218 }, { "epoch": 3.0, "learning_rate": 7.003136516897986e-05, "loss": 2.8023, "step": 38219 }, { "epoch": 3.0, "learning_rate": 7.003058103975535e-05, "loss": 2.7953, "step": 38220 }, { "epoch": 3.0, "learning_rate": 7.002979691053085e-05, "loss": 2.5773, "step": 38221 }, { "epoch": 3.0, "learning_rate": 7.002901278130636e-05, "loss": 2.9528, "step": 38222 }, { "epoch": 3.0, "learning_rate": 7.002822865208186e-05, "loss": 2.8254, "step": 38223 }, { "epoch": 3.0, "learning_rate": 7.002744452285737e-05, "loss": 2.8111, "step": 38224 }, { "epoch": 3.0, "learning_rate": 7.002666039363287e-05, "loss": 2.876, "step": 38225 }, { "epoch": 3.0, "learning_rate": 7.002587626440837e-05, "loss": 3.0937, "step": 38226 }, { "epoch": 3.0, "learning_rate": 7.002509213518388e-05, "loss": 2.9152, "step": 38227 }, { "epoch": 3.0, "learning_rate": 7.002430800595938e-05, "loss": 2.6356, "step": 38228 }, { "epoch": 3.0, "learning_rate": 7.002352387673488e-05, "loss": 2.9102, "step": 38229 }, { "epoch": 3.0, "learning_rate": 7.002273974751039e-05, "loss": 2.9034, "step": 38230 }, { "epoch": 3.0, "learning_rate": 7.002195561828589e-05, "loss": 2.8528, "step": 38231 }, { "epoch": 3.0, "learning_rate": 7.00211714890614e-05, "loss": 2.805, "step": 38232 }, { "epoch": 3.0, "learning_rate": 7.00203873598369e-05, "loss": 2.9489, "step": 38233 }, { "epoch": 3.0, "learning_rate": 7.00196032306124e-05, "loss": 2.8258, "step": 38234 }, { "epoch": 3.0, "learning_rate": 7.00188191013879e-05, "loss": 2.6096, "step": 38235 }, { "epoch": 3.0, "learning_rate": 7.001803497216341e-05, "loss": 2.9312, "step": 38236 }, { "epoch": 3.0, "learning_rate": 7.001725084293891e-05, "loss": 2.7394, "step": 38237 }, { "epoch": 3.0, "learning_rate": 7.001646671371443e-05, "loss": 2.9621, "step": 38238 }, { "epoch": 3.0, "learning_rate": 7.001568258448993e-05, "loss": 2.6914, "step": 38239 }, { "epoch": 3.0, "learning_rate": 7.001489845526544e-05, "loss": 2.6346, "step": 38240 }, { "epoch": 3.0, "learning_rate": 7.001411432604094e-05, "loss": 2.9338, "step": 38241 }, { "epoch": 3.0, "learning_rate": 7.001333019681644e-05, "loss": 2.9634, "step": 38242 }, { "epoch": 3.0, "learning_rate": 7.001254606759195e-05, "loss": 2.9135, "step": 38243 }, { "epoch": 3.0, "learning_rate": 7.001176193836745e-05, "loss": 2.6537, "step": 38244 }, { "epoch": 3.0, "learning_rate": 7.001097780914296e-05, "loss": 3.0558, "step": 38245 }, { "epoch": 3.0, "learning_rate": 7.001019367991846e-05, "loss": 2.7852, "step": 38246 }, { "epoch": 3.0, "learning_rate": 7.000940955069396e-05, "loss": 2.9863, "step": 38247 }, { "epoch": 3.0, "learning_rate": 7.000862542146947e-05, "loss": 2.6757, "step": 38248 }, { "epoch": 3.0, "learning_rate": 7.000784129224497e-05, "loss": 2.7792, "step": 38249 }, { "epoch": 3.0, "learning_rate": 7.000705716302047e-05, "loss": 2.6739, "step": 38250 }, { "epoch": 3.0, "learning_rate": 7.000627303379598e-05, "loss": 2.5631, "step": 38251 }, { "epoch": 3.0, "learning_rate": 7.000548890457148e-05, "loss": 2.9444, "step": 38252 }, { "epoch": 3.0, "learning_rate": 7.000470477534699e-05, "loss": 2.7978, "step": 38253 }, { "epoch": 3.0, "learning_rate": 7.000392064612249e-05, "loss": 2.8797, "step": 38254 }, { "epoch": 3.0, "learning_rate": 7.000313651689799e-05, "loss": 2.8119, "step": 38255 }, { "epoch": 3.0, "learning_rate": 7.000235238767348e-05, "loss": 2.8391, "step": 38256 }, { "epoch": 3.0, "learning_rate": 7.000156825844899e-05, "loss": 2.9211, "step": 38257 }, { "epoch": 3.0, "learning_rate": 7.000078412922449e-05, "loss": 2.9472, "step": 38258 }, { "epoch": 3.0, "learning_rate": 7e-05, "loss": 2.8095, "step": 38259 }, { "epoch": 3.0, "learning_rate": 6.99992158707755e-05, "loss": 2.6506, "step": 38260 }, { "epoch": 3.0, "learning_rate": 6.9998431741551e-05, "loss": 2.7006, "step": 38261 }, { "epoch": 3.0, "learning_rate": 6.99976476123265e-05, "loss": 2.7344, "step": 38262 }, { "epoch": 3.0, "learning_rate": 6.999686348310202e-05, "loss": 2.5518, "step": 38263 }, { "epoch": 3.0, "learning_rate": 6.999607935387753e-05, "loss": 2.864, "step": 38264 }, { "epoch": 3.0, "learning_rate": 6.999529522465303e-05, "loss": 3.0586, "step": 38265 }, { "epoch": 3.0, "learning_rate": 6.999451109542853e-05, "loss": 2.8646, "step": 38266 }, { "epoch": 3.0, "learning_rate": 6.999372696620404e-05, "loss": 2.7801, "step": 38267 }, { "epoch": 3.0, "learning_rate": 6.999294283697954e-05, "loss": 2.7579, "step": 38268 }, { "epoch": 3.0, "learning_rate": 6.999215870775504e-05, "loss": 2.8683, "step": 38269 }, { "epoch": 3.0, "learning_rate": 6.999137457853055e-05, "loss": 2.6795, "step": 38270 }, { "epoch": 3.0, "learning_rate": 6.999059044930605e-05, "loss": 2.7365, "step": 38271 }, { "epoch": 3.0, "learning_rate": 6.998980632008156e-05, "loss": 2.6286, "step": 38272 }, { "epoch": 3.0, "learning_rate": 6.998902219085706e-05, "loss": 2.8752, "step": 38273 }, { "epoch": 3.0, "learning_rate": 6.998823806163256e-05, "loss": 2.5457, "step": 38274 }, { "epoch": 3.0, "learning_rate": 6.998745393240807e-05, "loss": 2.8077, "step": 38275 }, { "epoch": 3.0, "learning_rate": 6.998666980318357e-05, "loss": 2.7472, "step": 38276 }, { "epoch": 3.0, "learning_rate": 6.998588567395907e-05, "loss": 2.793, "step": 38277 }, { "epoch": 3.0, "learning_rate": 6.998510154473458e-05, "loss": 2.5836, "step": 38278 }, { "epoch": 3.0, "learning_rate": 6.998431741551008e-05, "loss": 2.5842, "step": 38279 }, { "epoch": 3.0, "learning_rate": 6.998353328628558e-05, "loss": 2.783, "step": 38280 }, { "epoch": 3.0, "learning_rate": 6.998274915706109e-05, "loss": 2.8244, "step": 38281 }, { "epoch": 3.0, "learning_rate": 6.998196502783659e-05, "loss": 2.5613, "step": 38282 }, { "epoch": 3.0, "learning_rate": 6.99811808986121e-05, "loss": 2.5782, "step": 38283 }, { "epoch": 3.0, "learning_rate": 6.99803967693876e-05, "loss": 2.6521, "step": 38284 }, { "epoch": 3.0, "learning_rate": 6.99796126401631e-05, "loss": 2.6124, "step": 38285 }, { "epoch": 3.0, "learning_rate": 6.99788285109386e-05, "loss": 2.6703, "step": 38286 }, { "epoch": 3.0, "learning_rate": 6.997804438171411e-05, "loss": 2.5031, "step": 38287 }, { "epoch": 3.0, "learning_rate": 6.997726025248961e-05, "loss": 2.6805, "step": 38288 }, { "epoch": 3.0, "learning_rate": 6.997647612326512e-05, "loss": 2.7355, "step": 38289 }, { "epoch": 3.0, "learning_rate": 6.997569199404062e-05, "loss": 2.6349, "step": 38290 }, { "epoch": 3.0, "learning_rate": 6.997490786481613e-05, "loss": 2.6679, "step": 38291 }, { "epoch": 3.0, "learning_rate": 6.997412373559163e-05, "loss": 2.6659, "step": 38292 }, { "epoch": 3.0, "learning_rate": 6.997333960636713e-05, "loss": 2.4676, "step": 38293 }, { "epoch": 3.0, "learning_rate": 6.997255547714264e-05, "loss": 2.6813, "step": 38294 }, { "epoch": 3.0, "learning_rate": 6.997177134791814e-05, "loss": 2.7793, "step": 38295 }, { "epoch": 3.0, "learning_rate": 6.997098721869364e-05, "loss": 2.7068, "step": 38296 }, { "epoch": 3.0, "learning_rate": 6.997020308946915e-05, "loss": 2.4755, "step": 38297 }, { "epoch": 3.0, "learning_rate": 6.996941896024465e-05, "loss": 2.6771, "step": 38298 }, { "epoch": 3.0, "learning_rate": 6.996863483102015e-05, "loss": 2.6575, "step": 38299 }, { "epoch": 3.0, "learning_rate": 6.996785070179566e-05, "loss": 2.782, "step": 38300 }, { "epoch": 3.0, "learning_rate": 6.996706657257116e-05, "loss": 2.5378, "step": 38301 }, { "epoch": 3.0, "learning_rate": 6.996628244334667e-05, "loss": 2.7321, "step": 38302 }, { "epoch": 3.0, "learning_rate": 6.996549831412217e-05, "loss": 2.6676, "step": 38303 }, { "epoch": 3.0, "learning_rate": 6.996471418489767e-05, "loss": 2.6448, "step": 38304 }, { "epoch": 3.0, "learning_rate": 6.996393005567318e-05, "loss": 2.774, "step": 38305 }, { "epoch": 3.0, "learning_rate": 6.996314592644868e-05, "loss": 2.8962, "step": 38306 }, { "epoch": 3.0, "learning_rate": 6.996236179722418e-05, "loss": 2.8086, "step": 38307 }, { "epoch": 3.0, "learning_rate": 6.996157766799969e-05, "loss": 2.4391, "step": 38308 }, { "epoch": 3.0, "learning_rate": 6.996079353877519e-05, "loss": 2.6346, "step": 38309 }, { "epoch": 3.0, "learning_rate": 6.99600094095507e-05, "loss": 2.8323, "step": 38310 }, { "epoch": 3.0, "learning_rate": 6.99592252803262e-05, "loss": 2.698, "step": 38311 }, { "epoch": 3.0, "learning_rate": 6.99584411511017e-05, "loss": 2.7696, "step": 38312 }, { "epoch": 3.0, "learning_rate": 6.99576570218772e-05, "loss": 2.9532, "step": 38313 }, { "epoch": 3.0, "learning_rate": 6.995687289265271e-05, "loss": 2.7513, "step": 38314 }, { "epoch": 3.0, "learning_rate": 6.995608876342821e-05, "loss": 2.7185, "step": 38315 }, { "epoch": 3.0, "learning_rate": 6.995530463420372e-05, "loss": 2.7622, "step": 38316 }, { "epoch": 3.0, "learning_rate": 6.995452050497922e-05, "loss": 2.9983, "step": 38317 }, { "epoch": 3.0, "learning_rate": 6.995373637575472e-05, "loss": 2.638, "step": 38318 }, { "epoch": 3.0, "learning_rate": 6.995295224653023e-05, "loss": 2.5961, "step": 38319 }, { "epoch": 3.0, "learning_rate": 6.995216811730575e-05, "loss": 3.1036, "step": 38320 }, { "epoch": 3.0, "learning_rate": 6.995138398808125e-05, "loss": 2.762, "step": 38321 }, { "epoch": 3.0, "learning_rate": 6.995059985885675e-05, "loss": 2.6243, "step": 38322 }, { "epoch": 3.01, "learning_rate": 6.994981572963226e-05, "loss": 2.8559, "step": 38323 }, { "epoch": 3.01, "learning_rate": 6.994903160040776e-05, "loss": 2.7801, "step": 38324 }, { "epoch": 3.01, "learning_rate": 6.994824747118326e-05, "loss": 2.5953, "step": 38325 }, { "epoch": 3.01, "learning_rate": 6.994746334195877e-05, "loss": 2.6375, "step": 38326 }, { "epoch": 3.01, "learning_rate": 6.994667921273427e-05, "loss": 2.7305, "step": 38327 }, { "epoch": 3.01, "learning_rate": 6.994589508350976e-05, "loss": 2.665, "step": 38328 }, { "epoch": 3.01, "learning_rate": 6.994511095428526e-05, "loss": 2.5784, "step": 38329 }, { "epoch": 3.01, "learning_rate": 6.994432682506077e-05, "loss": 2.4651, "step": 38330 }, { "epoch": 3.01, "learning_rate": 6.994354269583627e-05, "loss": 2.6711, "step": 38331 }, { "epoch": 3.01, "learning_rate": 6.994275856661178e-05, "loss": 2.8042, "step": 38332 }, { "epoch": 3.01, "learning_rate": 6.994197443738728e-05, "loss": 3.0404, "step": 38333 }, { "epoch": 3.01, "learning_rate": 6.994119030816278e-05, "loss": 2.6162, "step": 38334 }, { "epoch": 3.01, "learning_rate": 6.994040617893829e-05, "loss": 2.7416, "step": 38335 }, { "epoch": 3.01, "learning_rate": 6.993962204971379e-05, "loss": 2.8087, "step": 38336 }, { "epoch": 3.01, "learning_rate": 6.99388379204893e-05, "loss": 2.7107, "step": 38337 }, { "epoch": 3.01, "learning_rate": 6.99380537912648e-05, "loss": 2.6925, "step": 38338 }, { "epoch": 3.01, "learning_rate": 6.99372696620403e-05, "loss": 2.873, "step": 38339 }, { "epoch": 3.01, "learning_rate": 6.99364855328158e-05, "loss": 2.8174, "step": 38340 }, { "epoch": 3.01, "learning_rate": 6.993570140359131e-05, "loss": 2.6473, "step": 38341 }, { "epoch": 3.01, "learning_rate": 6.993491727436681e-05, "loss": 2.8281, "step": 38342 }, { "epoch": 3.01, "learning_rate": 6.993413314514232e-05, "loss": 2.5363, "step": 38343 }, { "epoch": 3.01, "learning_rate": 6.993334901591782e-05, "loss": 2.5525, "step": 38344 }, { "epoch": 3.01, "learning_rate": 6.993256488669332e-05, "loss": 2.6776, "step": 38345 }, { "epoch": 3.01, "learning_rate": 6.993178075746884e-05, "loss": 2.656, "step": 38346 }, { "epoch": 3.01, "learning_rate": 6.993099662824434e-05, "loss": 2.8068, "step": 38347 }, { "epoch": 3.01, "learning_rate": 6.993021249901985e-05, "loss": 2.7611, "step": 38348 }, { "epoch": 3.01, "learning_rate": 6.992942836979535e-05, "loss": 2.5168, "step": 38349 }, { "epoch": 3.01, "learning_rate": 6.992864424057086e-05, "loss": 2.7799, "step": 38350 }, { "epoch": 3.01, "learning_rate": 6.992786011134636e-05, "loss": 2.7034, "step": 38351 }, { "epoch": 3.01, "learning_rate": 6.992707598212186e-05, "loss": 2.6257, "step": 38352 }, { "epoch": 3.01, "learning_rate": 6.992629185289737e-05, "loss": 2.6392, "step": 38353 }, { "epoch": 3.01, "learning_rate": 6.992550772367287e-05, "loss": 2.8059, "step": 38354 }, { "epoch": 3.01, "learning_rate": 6.992472359444837e-05, "loss": 2.9567, "step": 38355 }, { "epoch": 3.01, "learning_rate": 6.992393946522388e-05, "loss": 2.7412, "step": 38356 }, { "epoch": 3.01, "learning_rate": 6.992315533599938e-05, "loss": 2.5658, "step": 38357 }, { "epoch": 3.01, "learning_rate": 6.992237120677489e-05, "loss": 3.0703, "step": 38358 }, { "epoch": 3.01, "learning_rate": 6.992158707755039e-05, "loss": 2.5779, "step": 38359 }, { "epoch": 3.01, "learning_rate": 6.992080294832589e-05, "loss": 2.714, "step": 38360 }, { "epoch": 3.01, "learning_rate": 6.99200188191014e-05, "loss": 2.9867, "step": 38361 }, { "epoch": 3.01, "learning_rate": 6.99192346898769e-05, "loss": 2.8039, "step": 38362 }, { "epoch": 3.01, "learning_rate": 6.99184505606524e-05, "loss": 2.5934, "step": 38363 }, { "epoch": 3.01, "learning_rate": 6.99176664314279e-05, "loss": 2.7862, "step": 38364 }, { "epoch": 3.01, "learning_rate": 6.99168823022034e-05, "loss": 2.756, "step": 38365 }, { "epoch": 3.01, "learning_rate": 6.99160981729789e-05, "loss": 2.7273, "step": 38366 }, { "epoch": 3.01, "learning_rate": 6.99153140437544e-05, "loss": 2.8042, "step": 38367 }, { "epoch": 3.01, "learning_rate": 6.991452991452991e-05, "loss": 2.7277, "step": 38368 }, { "epoch": 3.01, "learning_rate": 6.991374578530541e-05, "loss": 2.7497, "step": 38369 }, { "epoch": 3.01, "learning_rate": 6.991296165608092e-05, "loss": 2.668, "step": 38370 }, { "epoch": 3.01, "learning_rate": 6.991217752685642e-05, "loss": 2.5438, "step": 38371 }, { "epoch": 3.01, "learning_rate": 6.991139339763194e-05, "loss": 2.6492, "step": 38372 }, { "epoch": 3.01, "learning_rate": 6.991060926840744e-05, "loss": 2.8157, "step": 38373 }, { "epoch": 3.01, "learning_rate": 6.990982513918294e-05, "loss": 2.8234, "step": 38374 }, { "epoch": 3.01, "learning_rate": 6.990904100995845e-05, "loss": 2.8857, "step": 38375 }, { "epoch": 3.01, "learning_rate": 6.990825688073395e-05, "loss": 2.6601, "step": 38376 }, { "epoch": 3.01, "learning_rate": 6.990747275150945e-05, "loss": 2.7229, "step": 38377 }, { "epoch": 3.01, "learning_rate": 6.990668862228496e-05, "loss": 2.5634, "step": 38378 }, { "epoch": 3.01, "learning_rate": 6.990590449306046e-05, "loss": 2.7857, "step": 38379 }, { "epoch": 3.01, "learning_rate": 6.990512036383597e-05, "loss": 2.6397, "step": 38380 }, { "epoch": 3.01, "learning_rate": 6.990433623461147e-05, "loss": 2.8527, "step": 38381 }, { "epoch": 3.01, "learning_rate": 6.990355210538697e-05, "loss": 2.7627, "step": 38382 }, { "epoch": 3.01, "learning_rate": 6.990276797616248e-05, "loss": 2.9957, "step": 38383 }, { "epoch": 3.01, "learning_rate": 6.990198384693798e-05, "loss": 2.788, "step": 38384 }, { "epoch": 3.01, "learning_rate": 6.990119971771348e-05, "loss": 2.7062, "step": 38385 }, { "epoch": 3.01, "learning_rate": 6.990041558848899e-05, "loss": 2.4919, "step": 38386 }, { "epoch": 3.01, "learning_rate": 6.989963145926449e-05, "loss": 2.697, "step": 38387 }, { "epoch": 3.01, "learning_rate": 6.989884733004e-05, "loss": 2.6226, "step": 38388 }, { "epoch": 3.01, "learning_rate": 6.98980632008155e-05, "loss": 2.8947, "step": 38389 }, { "epoch": 3.01, "learning_rate": 6.9897279071591e-05, "loss": 2.6226, "step": 38390 }, { "epoch": 3.01, "learning_rate": 6.98964949423665e-05, "loss": 2.6229, "step": 38391 }, { "epoch": 3.01, "learning_rate": 6.989571081314201e-05, "loss": 2.8618, "step": 38392 }, { "epoch": 3.01, "learning_rate": 6.989492668391751e-05, "loss": 2.7158, "step": 38393 }, { "epoch": 3.01, "learning_rate": 6.989414255469302e-05, "loss": 2.6386, "step": 38394 }, { "epoch": 3.01, "learning_rate": 6.989335842546852e-05, "loss": 2.6748, "step": 38395 }, { "epoch": 3.01, "learning_rate": 6.989257429624402e-05, "loss": 2.6544, "step": 38396 }, { "epoch": 3.01, "learning_rate": 6.989179016701953e-05, "loss": 2.7733, "step": 38397 }, { "epoch": 3.01, "learning_rate": 6.989100603779503e-05, "loss": 2.7998, "step": 38398 }, { "epoch": 3.01, "learning_rate": 6.989022190857054e-05, "loss": 2.7195, "step": 38399 }, { "epoch": 3.01, "learning_rate": 6.988943777934604e-05, "loss": 2.9292, "step": 38400 }, { "epoch": 3.01, "learning_rate": 6.988865365012154e-05, "loss": 2.6745, "step": 38401 }, { "epoch": 3.01, "learning_rate": 6.988786952089705e-05, "loss": 2.7695, "step": 38402 }, { "epoch": 3.01, "learning_rate": 6.988708539167255e-05, "loss": 2.761, "step": 38403 }, { "epoch": 3.01, "learning_rate": 6.988630126244805e-05, "loss": 2.5596, "step": 38404 }, { "epoch": 3.01, "learning_rate": 6.988551713322356e-05, "loss": 2.6135, "step": 38405 }, { "epoch": 3.01, "learning_rate": 6.988473300399906e-05, "loss": 2.4283, "step": 38406 }, { "epoch": 3.01, "learning_rate": 6.988394887477457e-05, "loss": 2.6142, "step": 38407 }, { "epoch": 3.01, "learning_rate": 6.988316474555007e-05, "loss": 2.7622, "step": 38408 }, { "epoch": 3.01, "learning_rate": 6.988238061632557e-05, "loss": 2.8257, "step": 38409 }, { "epoch": 3.01, "learning_rate": 6.988159648710108e-05, "loss": 2.5364, "step": 38410 }, { "epoch": 3.01, "learning_rate": 6.988081235787658e-05, "loss": 2.528, "step": 38411 }, { "epoch": 3.01, "learning_rate": 6.988002822865208e-05, "loss": 2.7135, "step": 38412 }, { "epoch": 3.01, "learning_rate": 6.987924409942759e-05, "loss": 2.6364, "step": 38413 }, { "epoch": 3.01, "learning_rate": 6.987845997020309e-05, "loss": 2.6945, "step": 38414 }, { "epoch": 3.01, "learning_rate": 6.98776758409786e-05, "loss": 2.7882, "step": 38415 }, { "epoch": 3.01, "learning_rate": 6.98768917117541e-05, "loss": 2.7055, "step": 38416 }, { "epoch": 3.01, "learning_rate": 6.98761075825296e-05, "loss": 2.5392, "step": 38417 }, { "epoch": 3.01, "learning_rate": 6.98753234533051e-05, "loss": 2.8137, "step": 38418 }, { "epoch": 3.01, "learning_rate": 6.987453932408061e-05, "loss": 2.7207, "step": 38419 }, { "epoch": 3.01, "learning_rate": 6.987375519485611e-05, "loss": 2.5643, "step": 38420 }, { "epoch": 3.01, "learning_rate": 6.987297106563162e-05, "loss": 2.9926, "step": 38421 }, { "epoch": 3.01, "learning_rate": 6.987218693640712e-05, "loss": 2.4666, "step": 38422 }, { "epoch": 3.01, "learning_rate": 6.987140280718262e-05, "loss": 2.6203, "step": 38423 }, { "epoch": 3.01, "learning_rate": 6.987061867795813e-05, "loss": 2.6973, "step": 38424 }, { "epoch": 3.01, "learning_rate": 6.986983454873363e-05, "loss": 2.8299, "step": 38425 }, { "epoch": 3.01, "learning_rate": 6.986905041950914e-05, "loss": 2.5584, "step": 38426 }, { "epoch": 3.01, "learning_rate": 6.986826629028464e-05, "loss": 2.722, "step": 38427 }, { "epoch": 3.01, "learning_rate": 6.986748216106016e-05, "loss": 2.5723, "step": 38428 }, { "epoch": 3.01, "learning_rate": 6.986669803183566e-05, "loss": 2.8601, "step": 38429 }, { "epoch": 3.01, "learning_rate": 6.986591390261116e-05, "loss": 2.8867, "step": 38430 }, { "epoch": 3.01, "learning_rate": 6.986512977338667e-05, "loss": 2.7824, "step": 38431 }, { "epoch": 3.01, "learning_rate": 6.986434564416217e-05, "loss": 2.6192, "step": 38432 }, { "epoch": 3.01, "learning_rate": 6.986356151493767e-05, "loss": 2.6231, "step": 38433 }, { "epoch": 3.01, "learning_rate": 6.986277738571318e-05, "loss": 2.753, "step": 38434 }, { "epoch": 3.01, "learning_rate": 6.986199325648868e-05, "loss": 2.8073, "step": 38435 }, { "epoch": 3.01, "learning_rate": 6.986120912726417e-05, "loss": 2.6556, "step": 38436 }, { "epoch": 3.01, "learning_rate": 6.986042499803968e-05, "loss": 2.7323, "step": 38437 }, { "epoch": 3.01, "learning_rate": 6.985964086881518e-05, "loss": 2.6855, "step": 38438 }, { "epoch": 3.01, "learning_rate": 6.985885673959068e-05, "loss": 2.8731, "step": 38439 }, { "epoch": 3.01, "learning_rate": 6.985807261036619e-05, "loss": 2.6445, "step": 38440 }, { "epoch": 3.01, "learning_rate": 6.985728848114169e-05, "loss": 2.7049, "step": 38441 }, { "epoch": 3.01, "learning_rate": 6.98565043519172e-05, "loss": 2.7617, "step": 38442 }, { "epoch": 3.01, "learning_rate": 6.98557202226927e-05, "loss": 2.6135, "step": 38443 }, { "epoch": 3.01, "learning_rate": 6.98549360934682e-05, "loss": 2.6515, "step": 38444 }, { "epoch": 3.01, "learning_rate": 6.98541519642437e-05, "loss": 2.4954, "step": 38445 }, { "epoch": 3.01, "learning_rate": 6.985336783501921e-05, "loss": 2.7485, "step": 38446 }, { "epoch": 3.01, "learning_rate": 6.985258370579471e-05, "loss": 2.8307, "step": 38447 }, { "epoch": 3.01, "learning_rate": 6.985179957657022e-05, "loss": 2.7826, "step": 38448 }, { "epoch": 3.01, "learning_rate": 6.985101544734572e-05, "loss": 2.693, "step": 38449 }, { "epoch": 3.01, "learning_rate": 6.985023131812122e-05, "loss": 2.5903, "step": 38450 }, { "epoch": 3.02, "learning_rate": 6.984944718889673e-05, "loss": 3.0552, "step": 38451 }, { "epoch": 3.02, "learning_rate": 6.984866305967223e-05, "loss": 2.9019, "step": 38452 }, { "epoch": 3.02, "learning_rate": 6.984787893044773e-05, "loss": 2.7595, "step": 38453 }, { "epoch": 3.02, "learning_rate": 6.984709480122325e-05, "loss": 2.6526, "step": 38454 }, { "epoch": 3.02, "learning_rate": 6.984631067199876e-05, "loss": 2.8445, "step": 38455 }, { "epoch": 3.02, "learning_rate": 6.984552654277426e-05, "loss": 2.7973, "step": 38456 }, { "epoch": 3.02, "learning_rate": 6.984474241354976e-05, "loss": 2.7369, "step": 38457 }, { "epoch": 3.02, "learning_rate": 6.984395828432527e-05, "loss": 2.8028, "step": 38458 }, { "epoch": 3.02, "learning_rate": 6.984317415510077e-05, "loss": 2.7523, "step": 38459 }, { "epoch": 3.02, "learning_rate": 6.984239002587627e-05, "loss": 2.6744, "step": 38460 }, { "epoch": 3.02, "learning_rate": 6.984160589665178e-05, "loss": 2.479, "step": 38461 }, { "epoch": 3.02, "learning_rate": 6.984082176742728e-05, "loss": 2.6609, "step": 38462 }, { "epoch": 3.02, "learning_rate": 6.984003763820278e-05, "loss": 2.5612, "step": 38463 }, { "epoch": 3.02, "learning_rate": 6.983925350897829e-05, "loss": 2.6147, "step": 38464 }, { "epoch": 3.02, "learning_rate": 6.983846937975379e-05, "loss": 2.5599, "step": 38465 }, { "epoch": 3.02, "learning_rate": 6.98376852505293e-05, "loss": 2.4869, "step": 38466 }, { "epoch": 3.02, "learning_rate": 6.98369011213048e-05, "loss": 2.9938, "step": 38467 }, { "epoch": 3.02, "learning_rate": 6.98361169920803e-05, "loss": 2.7523, "step": 38468 }, { "epoch": 3.02, "learning_rate": 6.983533286285581e-05, "loss": 2.5736, "step": 38469 }, { "epoch": 3.02, "learning_rate": 6.983454873363131e-05, "loss": 2.5596, "step": 38470 }, { "epoch": 3.02, "learning_rate": 6.983376460440681e-05, "loss": 2.6964, "step": 38471 }, { "epoch": 3.02, "learning_rate": 6.98329804751823e-05, "loss": 2.9434, "step": 38472 }, { "epoch": 3.02, "learning_rate": 6.983219634595781e-05, "loss": 3.0182, "step": 38473 }, { "epoch": 3.02, "learning_rate": 6.983141221673331e-05, "loss": 2.673, "step": 38474 }, { "epoch": 3.02, "learning_rate": 6.983062808750882e-05, "loss": 2.6876, "step": 38475 }, { "epoch": 3.02, "learning_rate": 6.982984395828432e-05, "loss": 2.805, "step": 38476 }, { "epoch": 3.02, "learning_rate": 6.982905982905982e-05, "loss": 2.7402, "step": 38477 }, { "epoch": 3.02, "learning_rate": 6.982827569983533e-05, "loss": 2.8585, "step": 38478 }, { "epoch": 3.02, "learning_rate": 6.982749157061083e-05, "loss": 2.7187, "step": 38479 }, { "epoch": 3.02, "learning_rate": 6.982670744138635e-05, "loss": 2.544, "step": 38480 }, { "epoch": 3.02, "learning_rate": 6.982592331216185e-05, "loss": 2.7468, "step": 38481 }, { "epoch": 3.02, "learning_rate": 6.982513918293735e-05, "loss": 2.6445, "step": 38482 }, { "epoch": 3.02, "learning_rate": 6.982435505371286e-05, "loss": 2.6148, "step": 38483 }, { "epoch": 3.02, "learning_rate": 6.982357092448836e-05, "loss": 2.9605, "step": 38484 }, { "epoch": 3.02, "learning_rate": 6.982278679526387e-05, "loss": 2.9033, "step": 38485 }, { "epoch": 3.02, "learning_rate": 6.982200266603937e-05, "loss": 2.7723, "step": 38486 }, { "epoch": 3.02, "learning_rate": 6.982121853681487e-05, "loss": 2.6351, "step": 38487 }, { "epoch": 3.02, "learning_rate": 6.982043440759038e-05, "loss": 2.6676, "step": 38488 }, { "epoch": 3.02, "learning_rate": 6.981965027836588e-05, "loss": 2.7217, "step": 38489 }, { "epoch": 3.02, "learning_rate": 6.981886614914138e-05, "loss": 2.7452, "step": 38490 }, { "epoch": 3.02, "learning_rate": 6.981808201991689e-05, "loss": 2.7634, "step": 38491 }, { "epoch": 3.02, "learning_rate": 6.981729789069239e-05, "loss": 2.8103, "step": 38492 }, { "epoch": 3.02, "learning_rate": 6.98165137614679e-05, "loss": 2.6264, "step": 38493 }, { "epoch": 3.02, "learning_rate": 6.98157296322434e-05, "loss": 2.6646, "step": 38494 }, { "epoch": 3.02, "learning_rate": 6.98149455030189e-05, "loss": 2.6817, "step": 38495 }, { "epoch": 3.02, "learning_rate": 6.98141613737944e-05, "loss": 2.7292, "step": 38496 }, { "epoch": 3.02, "learning_rate": 6.981337724456991e-05, "loss": 2.7147, "step": 38497 }, { "epoch": 3.02, "learning_rate": 6.981259311534541e-05, "loss": 2.7876, "step": 38498 }, { "epoch": 3.02, "learning_rate": 6.981180898612092e-05, "loss": 2.6897, "step": 38499 }, { "epoch": 3.02, "learning_rate": 6.981102485689642e-05, "loss": 2.7311, "step": 38500 }, { "epoch": 3.02, "learning_rate": 6.981024072767192e-05, "loss": 2.6318, "step": 38501 }, { "epoch": 3.02, "learning_rate": 6.980945659844743e-05, "loss": 2.8219, "step": 38502 }, { "epoch": 3.02, "learning_rate": 6.980867246922293e-05, "loss": 2.7047, "step": 38503 }, { "epoch": 3.02, "learning_rate": 6.980788833999844e-05, "loss": 2.923, "step": 38504 }, { "epoch": 3.02, "learning_rate": 6.980710421077394e-05, "loss": 2.6849, "step": 38505 }, { "epoch": 3.02, "learning_rate": 6.980632008154944e-05, "loss": 2.7431, "step": 38506 }, { "epoch": 3.02, "learning_rate": 6.980553595232495e-05, "loss": 2.7129, "step": 38507 }, { "epoch": 3.02, "learning_rate": 6.980475182310045e-05, "loss": 2.7206, "step": 38508 }, { "epoch": 3.02, "learning_rate": 6.980396769387595e-05, "loss": 2.6503, "step": 38509 }, { "epoch": 3.02, "learning_rate": 6.980318356465146e-05, "loss": 2.6641, "step": 38510 }, { "epoch": 3.02, "learning_rate": 6.980239943542696e-05, "loss": 2.7471, "step": 38511 }, { "epoch": 3.02, "learning_rate": 6.980161530620246e-05, "loss": 2.9744, "step": 38512 }, { "epoch": 3.02, "learning_rate": 6.980083117697797e-05, "loss": 2.6526, "step": 38513 }, { "epoch": 3.02, "learning_rate": 6.980004704775347e-05, "loss": 2.6169, "step": 38514 }, { "epoch": 3.02, "learning_rate": 6.979926291852898e-05, "loss": 2.7722, "step": 38515 }, { "epoch": 3.02, "learning_rate": 6.979847878930448e-05, "loss": 2.5929, "step": 38516 }, { "epoch": 3.02, "learning_rate": 6.979769466007998e-05, "loss": 2.8661, "step": 38517 }, { "epoch": 3.02, "learning_rate": 6.979691053085549e-05, "loss": 2.8025, "step": 38518 }, { "epoch": 3.02, "learning_rate": 6.979612640163099e-05, "loss": 2.7689, "step": 38519 }, { "epoch": 3.02, "learning_rate": 6.97953422724065e-05, "loss": 2.5659, "step": 38520 }, { "epoch": 3.02, "learning_rate": 6.9794558143182e-05, "loss": 2.7747, "step": 38521 }, { "epoch": 3.02, "learning_rate": 6.97937740139575e-05, "loss": 2.8099, "step": 38522 }, { "epoch": 3.02, "learning_rate": 6.9792989884733e-05, "loss": 2.7957, "step": 38523 }, { "epoch": 3.02, "learning_rate": 6.979220575550851e-05, "loss": 2.706, "step": 38524 }, { "epoch": 3.02, "learning_rate": 6.979142162628401e-05, "loss": 2.7096, "step": 38525 }, { "epoch": 3.02, "learning_rate": 6.979063749705952e-05, "loss": 2.8669, "step": 38526 }, { "epoch": 3.02, "learning_rate": 6.978985336783502e-05, "loss": 2.6793, "step": 38527 }, { "epoch": 3.02, "learning_rate": 6.978906923861052e-05, "loss": 2.632, "step": 38528 }, { "epoch": 3.02, "learning_rate": 6.978828510938603e-05, "loss": 2.7171, "step": 38529 }, { "epoch": 3.02, "learning_rate": 6.978750098016153e-05, "loss": 2.5822, "step": 38530 }, { "epoch": 3.02, "learning_rate": 6.978671685093703e-05, "loss": 2.8356, "step": 38531 }, { "epoch": 3.02, "learning_rate": 6.978593272171254e-05, "loss": 2.6778, "step": 38532 }, { "epoch": 3.02, "learning_rate": 6.978514859248804e-05, "loss": 2.6027, "step": 38533 }, { "epoch": 3.02, "learning_rate": 6.978436446326355e-05, "loss": 2.6619, "step": 38534 }, { "epoch": 3.02, "learning_rate": 6.978358033403905e-05, "loss": 2.6538, "step": 38535 }, { "epoch": 3.02, "learning_rate": 6.978279620481457e-05, "loss": 2.5394, "step": 38536 }, { "epoch": 3.02, "learning_rate": 6.978201207559007e-05, "loss": 2.7314, "step": 38537 }, { "epoch": 3.02, "learning_rate": 6.978122794636557e-05, "loss": 2.5975, "step": 38538 }, { "epoch": 3.02, "learning_rate": 6.978044381714108e-05, "loss": 2.7451, "step": 38539 }, { "epoch": 3.02, "learning_rate": 6.977965968791658e-05, "loss": 2.9123, "step": 38540 }, { "epoch": 3.02, "learning_rate": 6.977887555869209e-05, "loss": 2.8159, "step": 38541 }, { "epoch": 3.02, "learning_rate": 6.977809142946759e-05, "loss": 2.6595, "step": 38542 }, { "epoch": 3.02, "learning_rate": 6.977730730024309e-05, "loss": 2.5394, "step": 38543 }, { "epoch": 3.02, "learning_rate": 6.977652317101858e-05, "loss": 2.6849, "step": 38544 }, { "epoch": 3.02, "learning_rate": 6.977573904179409e-05, "loss": 2.8629, "step": 38545 }, { "epoch": 3.02, "learning_rate": 6.977495491256959e-05, "loss": 2.8101, "step": 38546 }, { "epoch": 3.02, "learning_rate": 6.97741707833451e-05, "loss": 2.6924, "step": 38547 }, { "epoch": 3.02, "learning_rate": 6.97733866541206e-05, "loss": 2.5813, "step": 38548 }, { "epoch": 3.02, "learning_rate": 6.97726025248961e-05, "loss": 2.8618, "step": 38549 }, { "epoch": 3.02, "learning_rate": 6.97718183956716e-05, "loss": 2.6946, "step": 38550 }, { "epoch": 3.02, "learning_rate": 6.977103426644711e-05, "loss": 2.8069, "step": 38551 }, { "epoch": 3.02, "learning_rate": 6.977025013722261e-05, "loss": 2.6197, "step": 38552 }, { "epoch": 3.02, "learning_rate": 6.976946600799812e-05, "loss": 2.6805, "step": 38553 }, { "epoch": 3.02, "learning_rate": 6.976868187877362e-05, "loss": 2.595, "step": 38554 }, { "epoch": 3.02, "learning_rate": 6.976789774954912e-05, "loss": 2.598, "step": 38555 }, { "epoch": 3.02, "learning_rate": 6.976711362032463e-05, "loss": 2.7764, "step": 38556 }, { "epoch": 3.02, "learning_rate": 6.976632949110013e-05, "loss": 2.5842, "step": 38557 }, { "epoch": 3.02, "learning_rate": 6.976554536187563e-05, "loss": 2.7988, "step": 38558 }, { "epoch": 3.02, "learning_rate": 6.976476123265114e-05, "loss": 2.8761, "step": 38559 }, { "epoch": 3.02, "learning_rate": 6.976397710342664e-05, "loss": 2.6515, "step": 38560 }, { "epoch": 3.02, "learning_rate": 6.976319297420215e-05, "loss": 2.7127, "step": 38561 }, { "epoch": 3.02, "learning_rate": 6.976240884497766e-05, "loss": 2.5901, "step": 38562 }, { "epoch": 3.02, "learning_rate": 6.976162471575317e-05, "loss": 2.87, "step": 38563 }, { "epoch": 3.02, "learning_rate": 6.976084058652867e-05, "loss": 2.4777, "step": 38564 }, { "epoch": 3.02, "learning_rate": 6.976005645730417e-05, "loss": 2.5184, "step": 38565 }, { "epoch": 3.02, "learning_rate": 6.975927232807968e-05, "loss": 2.5305, "step": 38566 }, { "epoch": 3.02, "learning_rate": 6.975848819885518e-05, "loss": 2.6279, "step": 38567 }, { "epoch": 3.02, "learning_rate": 6.975770406963068e-05, "loss": 2.6633, "step": 38568 }, { "epoch": 3.02, "learning_rate": 6.975691994040619e-05, "loss": 2.8335, "step": 38569 }, { "epoch": 3.02, "learning_rate": 6.975613581118169e-05, "loss": 2.6675, "step": 38570 }, { "epoch": 3.02, "learning_rate": 6.97553516819572e-05, "loss": 2.5359, "step": 38571 }, { "epoch": 3.02, "learning_rate": 6.97545675527327e-05, "loss": 2.7253, "step": 38572 }, { "epoch": 3.02, "learning_rate": 6.97537834235082e-05, "loss": 2.9251, "step": 38573 }, { "epoch": 3.02, "learning_rate": 6.97529992942837e-05, "loss": 2.6616, "step": 38574 }, { "epoch": 3.02, "learning_rate": 6.975221516505921e-05, "loss": 2.7475, "step": 38575 }, { "epoch": 3.02, "learning_rate": 6.975143103583471e-05, "loss": 2.5614, "step": 38576 }, { "epoch": 3.02, "learning_rate": 6.975064690661022e-05, "loss": 2.6823, "step": 38577 }, { "epoch": 3.03, "learning_rate": 6.974986277738572e-05, "loss": 2.6868, "step": 38578 }, { "epoch": 3.03, "learning_rate": 6.974907864816122e-05, "loss": 2.78, "step": 38579 }, { "epoch": 3.03, "learning_rate": 6.974829451893672e-05, "loss": 2.7084, "step": 38580 }, { "epoch": 3.03, "learning_rate": 6.974751038971222e-05, "loss": 2.7436, "step": 38581 }, { "epoch": 3.03, "learning_rate": 6.974672626048772e-05, "loss": 2.7821, "step": 38582 }, { "epoch": 3.03, "learning_rate": 6.974594213126323e-05, "loss": 2.6322, "step": 38583 }, { "epoch": 3.03, "learning_rate": 6.974515800203873e-05, "loss": 2.7262, "step": 38584 }, { "epoch": 3.03, "learning_rate": 6.974437387281423e-05, "loss": 2.7335, "step": 38585 }, { "epoch": 3.03, "learning_rate": 6.974358974358974e-05, "loss": 2.516, "step": 38586 }, { "epoch": 3.03, "learning_rate": 6.974280561436524e-05, "loss": 2.7621, "step": 38587 }, { "epoch": 3.03, "learning_rate": 6.974202148514076e-05, "loss": 2.903, "step": 38588 }, { "epoch": 3.03, "learning_rate": 6.974123735591626e-05, "loss": 2.8374, "step": 38589 }, { "epoch": 3.03, "learning_rate": 6.974045322669177e-05, "loss": 2.9116, "step": 38590 }, { "epoch": 3.03, "learning_rate": 6.973966909746727e-05, "loss": 2.6594, "step": 38591 }, { "epoch": 3.03, "learning_rate": 6.973888496824277e-05, "loss": 3.1909, "step": 38592 }, { "epoch": 3.03, "learning_rate": 6.973810083901828e-05, "loss": 2.7647, "step": 38593 }, { "epoch": 3.03, "learning_rate": 6.973731670979378e-05, "loss": 2.7194, "step": 38594 }, { "epoch": 3.03, "learning_rate": 6.973653258056928e-05, "loss": 2.7039, "step": 38595 }, { "epoch": 3.03, "learning_rate": 6.973574845134479e-05, "loss": 2.9008, "step": 38596 }, { "epoch": 3.03, "learning_rate": 6.973496432212029e-05, "loss": 2.7281, "step": 38597 }, { "epoch": 3.03, "learning_rate": 6.97341801928958e-05, "loss": 2.8507, "step": 38598 }, { "epoch": 3.03, "learning_rate": 6.97333960636713e-05, "loss": 2.7239, "step": 38599 }, { "epoch": 3.03, "learning_rate": 6.97326119344468e-05, "loss": 2.9351, "step": 38600 }, { "epoch": 3.03, "learning_rate": 6.97318278052223e-05, "loss": 2.5032, "step": 38601 }, { "epoch": 3.03, "learning_rate": 6.973104367599781e-05, "loss": 2.743, "step": 38602 }, { "epoch": 3.03, "learning_rate": 6.973025954677331e-05, "loss": 2.6849, "step": 38603 }, { "epoch": 3.03, "learning_rate": 6.972947541754882e-05, "loss": 2.7879, "step": 38604 }, { "epoch": 3.03, "learning_rate": 6.972869128832432e-05, "loss": 2.597, "step": 38605 }, { "epoch": 3.03, "learning_rate": 6.972790715909982e-05, "loss": 2.5505, "step": 38606 }, { "epoch": 3.03, "learning_rate": 6.972712302987533e-05, "loss": 2.715, "step": 38607 }, { "epoch": 3.03, "learning_rate": 6.972633890065083e-05, "loss": 2.6148, "step": 38608 }, { "epoch": 3.03, "learning_rate": 6.972555477142634e-05, "loss": 2.6476, "step": 38609 }, { "epoch": 3.03, "learning_rate": 6.972477064220184e-05, "loss": 2.8182, "step": 38610 }, { "epoch": 3.03, "learning_rate": 6.972398651297734e-05, "loss": 2.7782, "step": 38611 }, { "epoch": 3.03, "learning_rate": 6.972320238375285e-05, "loss": 2.5233, "step": 38612 }, { "epoch": 3.03, "learning_rate": 6.972241825452835e-05, "loss": 2.7294, "step": 38613 }, { "epoch": 3.03, "learning_rate": 6.972163412530385e-05, "loss": 2.9165, "step": 38614 }, { "epoch": 3.03, "learning_rate": 6.972084999607936e-05, "loss": 2.8265, "step": 38615 }, { "epoch": 3.03, "learning_rate": 6.972006586685486e-05, "loss": 2.6382, "step": 38616 }, { "epoch": 3.03, "learning_rate": 6.971928173763036e-05, "loss": 2.6836, "step": 38617 }, { "epoch": 3.03, "learning_rate": 6.971849760840587e-05, "loss": 2.5899, "step": 38618 }, { "epoch": 3.03, "learning_rate": 6.971771347918137e-05, "loss": 2.6269, "step": 38619 }, { "epoch": 3.03, "learning_rate": 6.971692934995688e-05, "loss": 2.6707, "step": 38620 }, { "epoch": 3.03, "learning_rate": 6.971614522073238e-05, "loss": 2.7047, "step": 38621 }, { "epoch": 3.03, "learning_rate": 6.971536109150788e-05, "loss": 2.8852, "step": 38622 }, { "epoch": 3.03, "learning_rate": 6.971457696228339e-05, "loss": 2.5895, "step": 38623 }, { "epoch": 3.03, "learning_rate": 6.971379283305889e-05, "loss": 2.8309, "step": 38624 }, { "epoch": 3.03, "learning_rate": 6.97130087038344e-05, "loss": 2.7939, "step": 38625 }, { "epoch": 3.03, "learning_rate": 6.97122245746099e-05, "loss": 2.6012, "step": 38626 }, { "epoch": 3.03, "learning_rate": 6.97114404453854e-05, "loss": 2.7686, "step": 38627 }, { "epoch": 3.03, "learning_rate": 6.97106563161609e-05, "loss": 2.7722, "step": 38628 }, { "epoch": 3.03, "learning_rate": 6.970987218693641e-05, "loss": 2.8892, "step": 38629 }, { "epoch": 3.03, "learning_rate": 6.970908805771191e-05, "loss": 2.8214, "step": 38630 }, { "epoch": 3.03, "learning_rate": 6.970830392848742e-05, "loss": 2.7354, "step": 38631 }, { "epoch": 3.03, "learning_rate": 6.970751979926292e-05, "loss": 2.5972, "step": 38632 }, { "epoch": 3.03, "learning_rate": 6.970673567003842e-05, "loss": 2.7284, "step": 38633 }, { "epoch": 3.03, "learning_rate": 6.970595154081393e-05, "loss": 2.6987, "step": 38634 }, { "epoch": 3.03, "learning_rate": 6.970516741158943e-05, "loss": 2.6337, "step": 38635 }, { "epoch": 3.03, "learning_rate": 6.970438328236493e-05, "loss": 2.9285, "step": 38636 }, { "epoch": 3.03, "learning_rate": 6.970359915314044e-05, "loss": 2.8475, "step": 38637 }, { "epoch": 3.03, "learning_rate": 6.970281502391594e-05, "loss": 2.7278, "step": 38638 }, { "epoch": 3.03, "learning_rate": 6.970203089469145e-05, "loss": 2.8349, "step": 38639 }, { "epoch": 3.03, "learning_rate": 6.970124676546695e-05, "loss": 2.59, "step": 38640 }, { "epoch": 3.03, "learning_rate": 6.970046263624245e-05, "loss": 2.6035, "step": 38641 }, { "epoch": 3.03, "learning_rate": 6.969967850701796e-05, "loss": 2.6791, "step": 38642 }, { "epoch": 3.03, "learning_rate": 6.969889437779346e-05, "loss": 2.7835, "step": 38643 }, { "epoch": 3.03, "learning_rate": 6.969811024856896e-05, "loss": 2.5968, "step": 38644 }, { "epoch": 3.03, "learning_rate": 6.969732611934448e-05, "loss": 2.9455, "step": 38645 }, { "epoch": 3.03, "learning_rate": 6.969654199011998e-05, "loss": 2.9033, "step": 38646 }, { "epoch": 3.03, "learning_rate": 6.969575786089549e-05, "loss": 2.7608, "step": 38647 }, { "epoch": 3.03, "learning_rate": 6.969497373167099e-05, "loss": 2.7899, "step": 38648 }, { "epoch": 3.03, "learning_rate": 6.96941896024465e-05, "loss": 2.9338, "step": 38649 }, { "epoch": 3.03, "learning_rate": 6.9693405473222e-05, "loss": 2.805, "step": 38650 }, { "epoch": 3.03, "learning_rate": 6.96926213439975e-05, "loss": 2.7046, "step": 38651 }, { "epoch": 3.03, "learning_rate": 6.9691837214773e-05, "loss": 2.5888, "step": 38652 }, { "epoch": 3.03, "learning_rate": 6.96910530855485e-05, "loss": 2.8155, "step": 38653 }, { "epoch": 3.03, "learning_rate": 6.9690268956324e-05, "loss": 2.4909, "step": 38654 }, { "epoch": 3.03, "learning_rate": 6.96894848270995e-05, "loss": 2.6779, "step": 38655 }, { "epoch": 3.03, "learning_rate": 6.968870069787501e-05, "loss": 2.4012, "step": 38656 }, { "epoch": 3.03, "learning_rate": 6.968791656865051e-05, "loss": 2.9074, "step": 38657 }, { "epoch": 3.03, "learning_rate": 6.968713243942602e-05, "loss": 2.84, "step": 38658 }, { "epoch": 3.03, "learning_rate": 6.968634831020152e-05, "loss": 2.6732, "step": 38659 }, { "epoch": 3.03, "learning_rate": 6.968556418097702e-05, "loss": 2.7767, "step": 38660 }, { "epoch": 3.03, "learning_rate": 6.968478005175253e-05, "loss": 2.7864, "step": 38661 }, { "epoch": 3.03, "learning_rate": 6.968399592252803e-05, "loss": 2.8513, "step": 38662 }, { "epoch": 3.03, "learning_rate": 6.968321179330353e-05, "loss": 2.5682, "step": 38663 }, { "epoch": 3.03, "learning_rate": 6.968242766407904e-05, "loss": 2.5443, "step": 38664 }, { "epoch": 3.03, "learning_rate": 6.968164353485454e-05, "loss": 3.0154, "step": 38665 }, { "epoch": 3.03, "learning_rate": 6.968085940563004e-05, "loss": 2.6607, "step": 38666 }, { "epoch": 3.03, "learning_rate": 6.968007527640555e-05, "loss": 2.6065, "step": 38667 }, { "epoch": 3.03, "learning_rate": 6.967929114718105e-05, "loss": 2.5428, "step": 38668 }, { "epoch": 3.03, "learning_rate": 6.967850701795656e-05, "loss": 2.6494, "step": 38669 }, { "epoch": 3.03, "learning_rate": 6.967772288873207e-05, "loss": 2.8074, "step": 38670 }, { "epoch": 3.03, "learning_rate": 6.967693875950758e-05, "loss": 2.7531, "step": 38671 }, { "epoch": 3.03, "learning_rate": 6.967615463028308e-05, "loss": 2.9402, "step": 38672 }, { "epoch": 3.03, "learning_rate": 6.967537050105858e-05, "loss": 2.7269, "step": 38673 }, { "epoch": 3.03, "learning_rate": 6.967458637183409e-05, "loss": 2.7195, "step": 38674 }, { "epoch": 3.03, "learning_rate": 6.967380224260959e-05, "loss": 2.6487, "step": 38675 }, { "epoch": 3.03, "learning_rate": 6.96730181133851e-05, "loss": 2.7324, "step": 38676 }, { "epoch": 3.03, "learning_rate": 6.96722339841606e-05, "loss": 2.8933, "step": 38677 }, { "epoch": 3.03, "learning_rate": 6.96714498549361e-05, "loss": 2.658, "step": 38678 }, { "epoch": 3.03, "learning_rate": 6.96706657257116e-05, "loss": 2.645, "step": 38679 }, { "epoch": 3.03, "learning_rate": 6.966988159648711e-05, "loss": 2.7802, "step": 38680 }, { "epoch": 3.03, "learning_rate": 6.966909746726261e-05, "loss": 2.74, "step": 38681 }, { "epoch": 3.03, "learning_rate": 6.966831333803812e-05, "loss": 2.9163, "step": 38682 }, { "epoch": 3.03, "learning_rate": 6.966752920881362e-05, "loss": 3.0078, "step": 38683 }, { "epoch": 3.03, "learning_rate": 6.966674507958912e-05, "loss": 2.6459, "step": 38684 }, { "epoch": 3.03, "learning_rate": 6.966596095036463e-05, "loss": 2.8403, "step": 38685 }, { "epoch": 3.03, "learning_rate": 6.966517682114013e-05, "loss": 2.6222, "step": 38686 }, { "epoch": 3.03, "learning_rate": 6.966439269191564e-05, "loss": 2.6646, "step": 38687 }, { "epoch": 3.03, "learning_rate": 6.966360856269113e-05, "loss": 2.975, "step": 38688 }, { "epoch": 3.03, "learning_rate": 6.966282443346663e-05, "loss": 2.7798, "step": 38689 }, { "epoch": 3.03, "learning_rate": 6.966204030424213e-05, "loss": 3.0136, "step": 38690 }, { "epoch": 3.03, "learning_rate": 6.966125617501764e-05, "loss": 2.7398, "step": 38691 }, { "epoch": 3.03, "learning_rate": 6.966047204579314e-05, "loss": 2.7903, "step": 38692 }, { "epoch": 3.03, "learning_rate": 6.965968791656864e-05, "loss": 2.7369, "step": 38693 }, { "epoch": 3.03, "learning_rate": 6.965890378734415e-05, "loss": 2.6399, "step": 38694 }, { "epoch": 3.03, "learning_rate": 6.965811965811965e-05, "loss": 2.7743, "step": 38695 }, { "epoch": 3.03, "learning_rate": 6.965733552889517e-05, "loss": 2.7352, "step": 38696 }, { "epoch": 3.03, "learning_rate": 6.965655139967067e-05, "loss": 2.9489, "step": 38697 }, { "epoch": 3.03, "learning_rate": 6.965576727044618e-05, "loss": 2.7485, "step": 38698 }, { "epoch": 3.03, "learning_rate": 6.965498314122168e-05, "loss": 2.7553, "step": 38699 }, { "epoch": 3.03, "learning_rate": 6.965419901199718e-05, "loss": 2.8295, "step": 38700 }, { "epoch": 3.03, "learning_rate": 6.965341488277269e-05, "loss": 2.784, "step": 38701 }, { "epoch": 3.03, "learning_rate": 6.965263075354819e-05, "loss": 2.8404, "step": 38702 }, { "epoch": 3.03, "learning_rate": 6.96518466243237e-05, "loss": 2.6319, "step": 38703 }, { "epoch": 3.03, "learning_rate": 6.96510624950992e-05, "loss": 2.6157, "step": 38704 }, { "epoch": 3.03, "learning_rate": 6.96502783658747e-05, "loss": 2.6105, "step": 38705 }, { "epoch": 3.04, "learning_rate": 6.96494942366502e-05, "loss": 2.8977, "step": 38706 }, { "epoch": 3.04, "learning_rate": 6.964871010742571e-05, "loss": 2.6334, "step": 38707 }, { "epoch": 3.04, "learning_rate": 6.964792597820121e-05, "loss": 2.6355, "step": 38708 }, { "epoch": 3.04, "learning_rate": 6.964714184897672e-05, "loss": 2.7993, "step": 38709 }, { "epoch": 3.04, "learning_rate": 6.964635771975222e-05, "loss": 3.169, "step": 38710 }, { "epoch": 3.04, "learning_rate": 6.964557359052772e-05, "loss": 2.5744, "step": 38711 }, { "epoch": 3.04, "learning_rate": 6.964478946130323e-05, "loss": 2.8431, "step": 38712 }, { "epoch": 3.04, "learning_rate": 6.964400533207873e-05, "loss": 2.4933, "step": 38713 }, { "epoch": 3.04, "learning_rate": 6.964322120285423e-05, "loss": 2.6104, "step": 38714 }, { "epoch": 3.04, "learning_rate": 6.964243707362974e-05, "loss": 2.6566, "step": 38715 }, { "epoch": 3.04, "learning_rate": 6.964165294440524e-05, "loss": 2.8472, "step": 38716 }, { "epoch": 3.04, "learning_rate": 6.964086881518075e-05, "loss": 2.7817, "step": 38717 }, { "epoch": 3.04, "learning_rate": 6.964008468595625e-05, "loss": 2.5481, "step": 38718 }, { "epoch": 3.04, "learning_rate": 6.963930055673175e-05, "loss": 2.7645, "step": 38719 }, { "epoch": 3.04, "learning_rate": 6.963851642750726e-05, "loss": 2.7467, "step": 38720 }, { "epoch": 3.04, "learning_rate": 6.963773229828276e-05, "loss": 2.7418, "step": 38721 }, { "epoch": 3.04, "learning_rate": 6.963694816905826e-05, "loss": 2.8808, "step": 38722 }, { "epoch": 3.04, "learning_rate": 6.963616403983377e-05, "loss": 2.6966, "step": 38723 }, { "epoch": 3.04, "learning_rate": 6.963537991060927e-05, "loss": 2.8901, "step": 38724 }, { "epoch": 3.04, "learning_rate": 6.963459578138478e-05, "loss": 2.9567, "step": 38725 }, { "epoch": 3.04, "learning_rate": 6.963381165216028e-05, "loss": 2.8211, "step": 38726 }, { "epoch": 3.04, "learning_rate": 6.963302752293578e-05, "loss": 3.0319, "step": 38727 }, { "epoch": 3.04, "learning_rate": 6.963224339371129e-05, "loss": 2.8025, "step": 38728 }, { "epoch": 3.04, "learning_rate": 6.963145926448679e-05, "loss": 2.7986, "step": 38729 }, { "epoch": 3.04, "learning_rate": 6.96306751352623e-05, "loss": 2.6819, "step": 38730 }, { "epoch": 3.04, "learning_rate": 6.96298910060378e-05, "loss": 2.7599, "step": 38731 }, { "epoch": 3.04, "learning_rate": 6.96291068768133e-05, "loss": 2.7194, "step": 38732 }, { "epoch": 3.04, "learning_rate": 6.96283227475888e-05, "loss": 2.6616, "step": 38733 }, { "epoch": 3.04, "learning_rate": 6.962753861836431e-05, "loss": 2.8231, "step": 38734 }, { "epoch": 3.04, "learning_rate": 6.962675448913981e-05, "loss": 2.7193, "step": 38735 }, { "epoch": 3.04, "learning_rate": 6.962597035991532e-05, "loss": 2.6632, "step": 38736 }, { "epoch": 3.04, "learning_rate": 6.962518623069082e-05, "loss": 2.7068, "step": 38737 }, { "epoch": 3.04, "learning_rate": 6.962440210146632e-05, "loss": 2.9438, "step": 38738 }, { "epoch": 3.04, "learning_rate": 6.962361797224183e-05, "loss": 2.9444, "step": 38739 }, { "epoch": 3.04, "learning_rate": 6.962283384301733e-05, "loss": 2.7061, "step": 38740 }, { "epoch": 3.04, "learning_rate": 6.962204971379283e-05, "loss": 2.752, "step": 38741 }, { "epoch": 3.04, "learning_rate": 6.962126558456834e-05, "loss": 2.7143, "step": 38742 }, { "epoch": 3.04, "learning_rate": 6.962048145534384e-05, "loss": 2.7567, "step": 38743 }, { "epoch": 3.04, "learning_rate": 6.961969732611935e-05, "loss": 2.5432, "step": 38744 }, { "epoch": 3.04, "learning_rate": 6.961891319689485e-05, "loss": 2.6561, "step": 38745 }, { "epoch": 3.04, "learning_rate": 6.961812906767035e-05, "loss": 2.6207, "step": 38746 }, { "epoch": 3.04, "learning_rate": 6.961734493844586e-05, "loss": 2.7608, "step": 38747 }, { "epoch": 3.04, "learning_rate": 6.961656080922136e-05, "loss": 2.711, "step": 38748 }, { "epoch": 3.04, "learning_rate": 6.961577667999686e-05, "loss": 2.789, "step": 38749 }, { "epoch": 3.04, "learning_rate": 6.961499255077237e-05, "loss": 2.6226, "step": 38750 }, { "epoch": 3.04, "learning_rate": 6.961420842154787e-05, "loss": 2.5838, "step": 38751 }, { "epoch": 3.04, "learning_rate": 6.961342429232337e-05, "loss": 2.601, "step": 38752 }, { "epoch": 3.04, "learning_rate": 6.961264016309889e-05, "loss": 2.5945, "step": 38753 }, { "epoch": 3.04, "learning_rate": 6.96118560338744e-05, "loss": 2.5811, "step": 38754 }, { "epoch": 3.04, "learning_rate": 6.96110719046499e-05, "loss": 2.7359, "step": 38755 }, { "epoch": 3.04, "learning_rate": 6.96102877754254e-05, "loss": 2.735, "step": 38756 }, { "epoch": 3.04, "learning_rate": 6.96095036462009e-05, "loss": 2.8779, "step": 38757 }, { "epoch": 3.04, "learning_rate": 6.960871951697641e-05, "loss": 2.6092, "step": 38758 }, { "epoch": 3.04, "learning_rate": 6.960793538775191e-05, "loss": 2.7914, "step": 38759 }, { "epoch": 3.04, "learning_rate": 6.96071512585274e-05, "loss": 2.7561, "step": 38760 }, { "epoch": 3.04, "learning_rate": 6.960636712930291e-05, "loss": 2.488, "step": 38761 }, { "epoch": 3.04, "learning_rate": 6.960558300007841e-05, "loss": 2.5271, "step": 38762 }, { "epoch": 3.04, "learning_rate": 6.960479887085392e-05, "loss": 2.6392, "step": 38763 }, { "epoch": 3.04, "learning_rate": 6.960401474162942e-05, "loss": 2.5565, "step": 38764 }, { "epoch": 3.04, "learning_rate": 6.960323061240492e-05, "loss": 2.5999, "step": 38765 }, { "epoch": 3.04, "learning_rate": 6.960244648318043e-05, "loss": 2.6553, "step": 38766 }, { "epoch": 3.04, "learning_rate": 6.960166235395593e-05, "loss": 3.0834, "step": 38767 }, { "epoch": 3.04, "learning_rate": 6.960087822473143e-05, "loss": 2.7917, "step": 38768 }, { "epoch": 3.04, "learning_rate": 6.960009409550694e-05, "loss": 3.1142, "step": 38769 }, { "epoch": 3.04, "learning_rate": 6.959930996628244e-05, "loss": 2.8657, "step": 38770 }, { "epoch": 3.04, "learning_rate": 6.959852583705794e-05, "loss": 3.0535, "step": 38771 }, { "epoch": 3.04, "learning_rate": 6.959774170783345e-05, "loss": 2.9627, "step": 38772 }, { "epoch": 3.04, "learning_rate": 6.959695757860895e-05, "loss": 2.5907, "step": 38773 }, { "epoch": 3.04, "learning_rate": 6.959617344938446e-05, "loss": 2.7309, "step": 38774 }, { "epoch": 3.04, "learning_rate": 6.959538932015996e-05, "loss": 2.8829, "step": 38775 }, { "epoch": 3.04, "learning_rate": 6.959460519093546e-05, "loss": 2.6597, "step": 38776 }, { "epoch": 3.04, "learning_rate": 6.959382106171097e-05, "loss": 2.743, "step": 38777 }, { "epoch": 3.04, "learning_rate": 6.959303693248648e-05, "loss": 2.6225, "step": 38778 }, { "epoch": 3.04, "learning_rate": 6.959225280326199e-05, "loss": 2.6561, "step": 38779 }, { "epoch": 3.04, "learning_rate": 6.959146867403749e-05, "loss": 2.8201, "step": 38780 }, { "epoch": 3.04, "learning_rate": 6.9590684544813e-05, "loss": 2.6262, "step": 38781 }, { "epoch": 3.04, "learning_rate": 6.95899004155885e-05, "loss": 2.5944, "step": 38782 }, { "epoch": 3.04, "learning_rate": 6.9589116286364e-05, "loss": 2.9781, "step": 38783 }, { "epoch": 3.04, "learning_rate": 6.95883321571395e-05, "loss": 2.7159, "step": 38784 }, { "epoch": 3.04, "learning_rate": 6.958754802791501e-05, "loss": 2.4761, "step": 38785 }, { "epoch": 3.04, "learning_rate": 6.958676389869051e-05, "loss": 2.694, "step": 38786 }, { "epoch": 3.04, "learning_rate": 6.958597976946602e-05, "loss": 2.63, "step": 38787 }, { "epoch": 3.04, "learning_rate": 6.958519564024152e-05, "loss": 2.6449, "step": 38788 }, { "epoch": 3.04, "learning_rate": 6.958441151101702e-05, "loss": 2.8012, "step": 38789 }, { "epoch": 3.04, "learning_rate": 6.958362738179253e-05, "loss": 2.8475, "step": 38790 }, { "epoch": 3.04, "learning_rate": 6.958284325256803e-05, "loss": 3.0863, "step": 38791 }, { "epoch": 3.04, "learning_rate": 6.958205912334354e-05, "loss": 2.7004, "step": 38792 }, { "epoch": 3.04, "learning_rate": 6.958127499411904e-05, "loss": 2.675, "step": 38793 }, { "epoch": 3.04, "learning_rate": 6.958049086489454e-05, "loss": 2.767, "step": 38794 }, { "epoch": 3.04, "learning_rate": 6.957970673567005e-05, "loss": 2.6036, "step": 38795 }, { "epoch": 3.04, "learning_rate": 6.957892260644554e-05, "loss": 2.8097, "step": 38796 }, { "epoch": 3.04, "learning_rate": 6.957813847722104e-05, "loss": 2.5765, "step": 38797 }, { "epoch": 3.04, "learning_rate": 6.957735434799654e-05, "loss": 2.9698, "step": 38798 }, { "epoch": 3.04, "learning_rate": 6.957657021877205e-05, "loss": 2.7134, "step": 38799 }, { "epoch": 3.04, "learning_rate": 6.957578608954755e-05, "loss": 2.6309, "step": 38800 }, { "epoch": 3.04, "learning_rate": 6.957500196032305e-05, "loss": 2.818, "step": 38801 }, { "epoch": 3.04, "learning_rate": 6.957421783109856e-05, "loss": 2.5803, "step": 38802 }, { "epoch": 3.04, "learning_rate": 6.957343370187406e-05, "loss": 2.5195, "step": 38803 }, { "epoch": 3.04, "learning_rate": 6.957264957264958e-05, "loss": 3.0553, "step": 38804 }, { "epoch": 3.04, "learning_rate": 6.957186544342508e-05, "loss": 2.7336, "step": 38805 }, { "epoch": 3.04, "learning_rate": 6.957108131420059e-05, "loss": 2.5288, "step": 38806 }, { "epoch": 3.04, "learning_rate": 6.957029718497609e-05, "loss": 2.6077, "step": 38807 }, { "epoch": 3.04, "learning_rate": 6.95695130557516e-05, "loss": 2.655, "step": 38808 }, { "epoch": 3.04, "learning_rate": 6.95687289265271e-05, "loss": 2.6605, "step": 38809 }, { "epoch": 3.04, "learning_rate": 6.95679447973026e-05, "loss": 2.6809, "step": 38810 }, { "epoch": 3.04, "learning_rate": 6.95671606680781e-05, "loss": 2.6412, "step": 38811 }, { "epoch": 3.04, "learning_rate": 6.956637653885361e-05, "loss": 2.6711, "step": 38812 }, { "epoch": 3.04, "learning_rate": 6.956559240962911e-05, "loss": 2.8618, "step": 38813 }, { "epoch": 3.04, "learning_rate": 6.956480828040462e-05, "loss": 2.7564, "step": 38814 }, { "epoch": 3.04, "learning_rate": 6.956402415118012e-05, "loss": 2.6469, "step": 38815 }, { "epoch": 3.04, "learning_rate": 6.956324002195562e-05, "loss": 2.9151, "step": 38816 }, { "epoch": 3.04, "learning_rate": 6.956245589273113e-05, "loss": 2.5655, "step": 38817 }, { "epoch": 3.04, "learning_rate": 6.956167176350663e-05, "loss": 3.1301, "step": 38818 }, { "epoch": 3.04, "learning_rate": 6.956088763428213e-05, "loss": 2.7722, "step": 38819 }, { "epoch": 3.04, "learning_rate": 6.956010350505764e-05, "loss": 2.7008, "step": 38820 }, { "epoch": 3.04, "learning_rate": 6.955931937583314e-05, "loss": 2.8917, "step": 38821 }, { "epoch": 3.04, "learning_rate": 6.955853524660865e-05, "loss": 2.9384, "step": 38822 }, { "epoch": 3.04, "learning_rate": 6.955775111738415e-05, "loss": 2.6737, "step": 38823 }, { "epoch": 3.04, "learning_rate": 6.955696698815965e-05, "loss": 2.993, "step": 38824 }, { "epoch": 3.04, "learning_rate": 6.955618285893516e-05, "loss": 2.4461, "step": 38825 }, { "epoch": 3.04, "learning_rate": 6.955539872971066e-05, "loss": 2.612, "step": 38826 }, { "epoch": 3.04, "learning_rate": 6.955461460048616e-05, "loss": 2.7635, "step": 38827 }, { "epoch": 3.04, "learning_rate": 6.955383047126167e-05, "loss": 2.715, "step": 38828 }, { "epoch": 3.04, "learning_rate": 6.955304634203717e-05, "loss": 2.923, "step": 38829 }, { "epoch": 3.04, "learning_rate": 6.955226221281268e-05, "loss": 2.8708, "step": 38830 }, { "epoch": 3.04, "learning_rate": 6.955147808358818e-05, "loss": 2.7943, "step": 38831 }, { "epoch": 3.04, "learning_rate": 6.955069395436368e-05, "loss": 2.6616, "step": 38832 }, { "epoch": 3.05, "learning_rate": 6.954990982513919e-05, "loss": 2.6453, "step": 38833 }, { "epoch": 3.05, "learning_rate": 6.954912569591469e-05, "loss": 2.8926, "step": 38834 }, { "epoch": 3.05, "learning_rate": 6.95483415666902e-05, "loss": 2.7705, "step": 38835 }, { "epoch": 3.05, "learning_rate": 6.95475574374657e-05, "loss": 2.842, "step": 38836 }, { "epoch": 3.05, "learning_rate": 6.95467733082412e-05, "loss": 2.8012, "step": 38837 }, { "epoch": 3.05, "learning_rate": 6.95459891790167e-05, "loss": 2.8532, "step": 38838 }, { "epoch": 3.05, "learning_rate": 6.954520504979221e-05, "loss": 2.615, "step": 38839 }, { "epoch": 3.05, "learning_rate": 6.954442092056771e-05, "loss": 2.8999, "step": 38840 }, { "epoch": 3.05, "learning_rate": 6.954363679134322e-05, "loss": 2.8584, "step": 38841 }, { "epoch": 3.05, "learning_rate": 6.954285266211872e-05, "loss": 2.7414, "step": 38842 }, { "epoch": 3.05, "learning_rate": 6.954206853289422e-05, "loss": 2.5532, "step": 38843 }, { "epoch": 3.05, "learning_rate": 6.954128440366973e-05, "loss": 2.7959, "step": 38844 }, { "epoch": 3.05, "learning_rate": 6.954050027444523e-05, "loss": 2.7026, "step": 38845 }, { "epoch": 3.05, "learning_rate": 6.953971614522073e-05, "loss": 2.6276, "step": 38846 }, { "epoch": 3.05, "learning_rate": 6.953893201599624e-05, "loss": 2.4803, "step": 38847 }, { "epoch": 3.05, "learning_rate": 6.953814788677174e-05, "loss": 2.5299, "step": 38848 }, { "epoch": 3.05, "learning_rate": 6.953736375754725e-05, "loss": 2.6512, "step": 38849 }, { "epoch": 3.05, "learning_rate": 6.953657962832275e-05, "loss": 2.772, "step": 38850 }, { "epoch": 3.05, "learning_rate": 6.953579549909825e-05, "loss": 2.9422, "step": 38851 }, { "epoch": 3.05, "learning_rate": 6.953501136987376e-05, "loss": 2.6687, "step": 38852 }, { "epoch": 3.05, "learning_rate": 6.953422724064926e-05, "loss": 2.9644, "step": 38853 }, { "epoch": 3.05, "learning_rate": 6.953344311142476e-05, "loss": 2.7035, "step": 38854 }, { "epoch": 3.05, "learning_rate": 6.953265898220027e-05, "loss": 2.7511, "step": 38855 }, { "epoch": 3.05, "learning_rate": 6.953187485297577e-05, "loss": 2.7019, "step": 38856 }, { "epoch": 3.05, "learning_rate": 6.953109072375127e-05, "loss": 2.6167, "step": 38857 }, { "epoch": 3.05, "learning_rate": 6.953030659452678e-05, "loss": 2.7298, "step": 38858 }, { "epoch": 3.05, "learning_rate": 6.952952246530228e-05, "loss": 2.5468, "step": 38859 }, { "epoch": 3.05, "learning_rate": 6.952873833607779e-05, "loss": 3.0019, "step": 38860 }, { "epoch": 3.05, "learning_rate": 6.95279542068533e-05, "loss": 2.6358, "step": 38861 }, { "epoch": 3.05, "learning_rate": 6.95271700776288e-05, "loss": 2.7648, "step": 38862 }, { "epoch": 3.05, "learning_rate": 6.952638594840431e-05, "loss": 2.4897, "step": 38863 }, { "epoch": 3.05, "learning_rate": 6.952560181917981e-05, "loss": 2.6376, "step": 38864 }, { "epoch": 3.05, "learning_rate": 6.952481768995532e-05, "loss": 3.0169, "step": 38865 }, { "epoch": 3.05, "learning_rate": 6.952403356073082e-05, "loss": 2.6646, "step": 38866 }, { "epoch": 3.05, "learning_rate": 6.952324943150632e-05, "loss": 2.6185, "step": 38867 }, { "epoch": 3.05, "learning_rate": 6.952246530228181e-05, "loss": 2.541, "step": 38868 }, { "epoch": 3.05, "learning_rate": 6.952168117305732e-05, "loss": 2.5459, "step": 38869 }, { "epoch": 3.05, "learning_rate": 6.952089704383282e-05, "loss": 2.728, "step": 38870 }, { "epoch": 3.05, "learning_rate": 6.952011291460833e-05, "loss": 2.7424, "step": 38871 }, { "epoch": 3.05, "learning_rate": 6.951932878538383e-05, "loss": 2.4271, "step": 38872 }, { "epoch": 3.05, "learning_rate": 6.951854465615933e-05, "loss": 2.7678, "step": 38873 }, { "epoch": 3.05, "learning_rate": 6.951776052693484e-05, "loss": 2.7481, "step": 38874 }, { "epoch": 3.05, "learning_rate": 6.951697639771034e-05, "loss": 2.7295, "step": 38875 }, { "epoch": 3.05, "learning_rate": 6.951619226848584e-05, "loss": 2.6289, "step": 38876 }, { "epoch": 3.05, "learning_rate": 6.951540813926135e-05, "loss": 2.7185, "step": 38877 }, { "epoch": 3.05, "learning_rate": 6.951462401003685e-05, "loss": 2.6772, "step": 38878 }, { "epoch": 3.05, "learning_rate": 6.951383988081236e-05, "loss": 2.6539, "step": 38879 }, { "epoch": 3.05, "learning_rate": 6.951305575158786e-05, "loss": 2.6657, "step": 38880 }, { "epoch": 3.05, "learning_rate": 6.951227162236336e-05, "loss": 2.9704, "step": 38881 }, { "epoch": 3.05, "learning_rate": 6.951148749313887e-05, "loss": 2.6071, "step": 38882 }, { "epoch": 3.05, "learning_rate": 6.951070336391437e-05, "loss": 2.6464, "step": 38883 }, { "epoch": 3.05, "learning_rate": 6.950991923468987e-05, "loss": 2.6418, "step": 38884 }, { "epoch": 3.05, "learning_rate": 6.950913510546538e-05, "loss": 2.6596, "step": 38885 }, { "epoch": 3.05, "learning_rate": 6.95083509762409e-05, "loss": 2.6415, "step": 38886 }, { "epoch": 3.05, "learning_rate": 6.95075668470164e-05, "loss": 2.8125, "step": 38887 }, { "epoch": 3.05, "learning_rate": 6.95067827177919e-05, "loss": 2.603, "step": 38888 }, { "epoch": 3.05, "learning_rate": 6.95059985885674e-05, "loss": 2.8225, "step": 38889 }, { "epoch": 3.05, "learning_rate": 6.950521445934291e-05, "loss": 2.8004, "step": 38890 }, { "epoch": 3.05, "learning_rate": 6.950443033011841e-05, "loss": 2.8127, "step": 38891 }, { "epoch": 3.05, "learning_rate": 6.950364620089392e-05, "loss": 2.8902, "step": 38892 }, { "epoch": 3.05, "learning_rate": 6.950286207166942e-05, "loss": 2.6752, "step": 38893 }, { "epoch": 3.05, "learning_rate": 6.950207794244492e-05, "loss": 2.943, "step": 38894 }, { "epoch": 3.05, "learning_rate": 6.950129381322043e-05, "loss": 2.7008, "step": 38895 }, { "epoch": 3.05, "learning_rate": 6.950050968399593e-05, "loss": 2.6847, "step": 38896 }, { "epoch": 3.05, "learning_rate": 6.949972555477144e-05, "loss": 2.7051, "step": 38897 }, { "epoch": 3.05, "learning_rate": 6.949894142554694e-05, "loss": 2.6929, "step": 38898 }, { "epoch": 3.05, "learning_rate": 6.949815729632244e-05, "loss": 2.7738, "step": 38899 }, { "epoch": 3.05, "learning_rate": 6.949737316709795e-05, "loss": 2.6658, "step": 38900 }, { "epoch": 3.05, "learning_rate": 6.949658903787345e-05, "loss": 2.5011, "step": 38901 }, { "epoch": 3.05, "learning_rate": 6.949580490864895e-05, "loss": 2.8244, "step": 38902 }, { "epoch": 3.05, "learning_rate": 6.949502077942446e-05, "loss": 2.4793, "step": 38903 }, { "epoch": 3.05, "learning_rate": 6.949423665019995e-05, "loss": 2.6784, "step": 38904 }, { "epoch": 3.05, "learning_rate": 6.949345252097545e-05, "loss": 2.7967, "step": 38905 }, { "epoch": 3.05, "learning_rate": 6.949266839175095e-05, "loss": 2.7309, "step": 38906 }, { "epoch": 3.05, "learning_rate": 6.949188426252646e-05, "loss": 2.5811, "step": 38907 }, { "epoch": 3.05, "learning_rate": 6.949110013330196e-05, "loss": 2.8167, "step": 38908 }, { "epoch": 3.05, "learning_rate": 6.949031600407747e-05, "loss": 2.7038, "step": 38909 }, { "epoch": 3.05, "learning_rate": 6.948953187485297e-05, "loss": 2.5761, "step": 38910 }, { "epoch": 3.05, "learning_rate": 6.948874774562847e-05, "loss": 2.6041, "step": 38911 }, { "epoch": 3.05, "learning_rate": 6.948796361640399e-05, "loss": 2.5725, "step": 38912 }, { "epoch": 3.05, "learning_rate": 6.94871794871795e-05, "loss": 2.6326, "step": 38913 }, { "epoch": 3.05, "learning_rate": 6.9486395357955e-05, "loss": 2.8084, "step": 38914 }, { "epoch": 3.05, "learning_rate": 6.94856112287305e-05, "loss": 2.748, "step": 38915 }, { "epoch": 3.05, "learning_rate": 6.9484827099506e-05, "loss": 2.7115, "step": 38916 }, { "epoch": 3.05, "learning_rate": 6.948404297028151e-05, "loss": 2.6676, "step": 38917 }, { "epoch": 3.05, "learning_rate": 6.948325884105701e-05, "loss": 2.7354, "step": 38918 }, { "epoch": 3.05, "learning_rate": 6.948247471183252e-05, "loss": 2.7007, "step": 38919 }, { "epoch": 3.05, "learning_rate": 6.948169058260802e-05, "loss": 2.7354, "step": 38920 }, { "epoch": 3.05, "learning_rate": 6.948090645338352e-05, "loss": 2.7149, "step": 38921 }, { "epoch": 3.05, "learning_rate": 6.948012232415903e-05, "loss": 2.5897, "step": 38922 }, { "epoch": 3.05, "learning_rate": 6.947933819493453e-05, "loss": 2.6818, "step": 38923 }, { "epoch": 3.05, "learning_rate": 6.947855406571003e-05, "loss": 2.5206, "step": 38924 }, { "epoch": 3.05, "learning_rate": 6.947776993648554e-05, "loss": 2.8785, "step": 38925 }, { "epoch": 3.05, "learning_rate": 6.947698580726104e-05, "loss": 2.8701, "step": 38926 }, { "epoch": 3.05, "learning_rate": 6.947620167803655e-05, "loss": 2.8114, "step": 38927 }, { "epoch": 3.05, "learning_rate": 6.947541754881205e-05, "loss": 2.5316, "step": 38928 }, { "epoch": 3.05, "learning_rate": 6.947463341958755e-05, "loss": 2.8194, "step": 38929 }, { "epoch": 3.05, "learning_rate": 6.947384929036306e-05, "loss": 2.6061, "step": 38930 }, { "epoch": 3.05, "learning_rate": 6.947306516113856e-05, "loss": 2.7275, "step": 38931 }, { "epoch": 3.05, "learning_rate": 6.947228103191406e-05, "loss": 2.6206, "step": 38932 }, { "epoch": 3.05, "learning_rate": 6.947149690268957e-05, "loss": 2.8752, "step": 38933 }, { "epoch": 3.05, "learning_rate": 6.947071277346507e-05, "loss": 2.4413, "step": 38934 }, { "epoch": 3.05, "learning_rate": 6.946992864424057e-05, "loss": 2.6451, "step": 38935 }, { "epoch": 3.05, "learning_rate": 6.946914451501608e-05, "loss": 2.6666, "step": 38936 }, { "epoch": 3.05, "learning_rate": 6.946836038579158e-05, "loss": 2.6516, "step": 38937 }, { "epoch": 3.05, "learning_rate": 6.946757625656709e-05, "loss": 2.6309, "step": 38938 }, { "epoch": 3.05, "learning_rate": 6.946679212734259e-05, "loss": 2.6904, "step": 38939 }, { "epoch": 3.05, "learning_rate": 6.94660079981181e-05, "loss": 2.6878, "step": 38940 }, { "epoch": 3.05, "learning_rate": 6.94652238688936e-05, "loss": 2.6814, "step": 38941 }, { "epoch": 3.05, "learning_rate": 6.94644397396691e-05, "loss": 2.9573, "step": 38942 }, { "epoch": 3.05, "learning_rate": 6.94636556104446e-05, "loss": 2.8996, "step": 38943 }, { "epoch": 3.05, "learning_rate": 6.946287148122011e-05, "loss": 2.8296, "step": 38944 }, { "epoch": 3.05, "learning_rate": 6.946208735199561e-05, "loss": 2.7753, "step": 38945 }, { "epoch": 3.05, "learning_rate": 6.946130322277112e-05, "loss": 2.7655, "step": 38946 }, { "epoch": 3.05, "learning_rate": 6.946051909354662e-05, "loss": 2.7763, "step": 38947 }, { "epoch": 3.05, "learning_rate": 6.945973496432212e-05, "loss": 2.6243, "step": 38948 }, { "epoch": 3.05, "learning_rate": 6.945895083509763e-05, "loss": 2.6144, "step": 38949 }, { "epoch": 3.05, "learning_rate": 6.945816670587313e-05, "loss": 2.9773, "step": 38950 }, { "epoch": 3.05, "learning_rate": 6.945738257664863e-05, "loss": 2.7405, "step": 38951 }, { "epoch": 3.05, "learning_rate": 6.945659844742414e-05, "loss": 2.7859, "step": 38952 }, { "epoch": 3.05, "learning_rate": 6.945581431819964e-05, "loss": 2.7015, "step": 38953 }, { "epoch": 3.05, "learning_rate": 6.945503018897514e-05, "loss": 3.1156, "step": 38954 }, { "epoch": 3.05, "learning_rate": 6.945424605975065e-05, "loss": 2.6949, "step": 38955 }, { "epoch": 3.05, "learning_rate": 6.945346193052615e-05, "loss": 2.709, "step": 38956 }, { "epoch": 3.05, "learning_rate": 6.945267780130166e-05, "loss": 2.7942, "step": 38957 }, { "epoch": 3.05, "learning_rate": 6.945189367207716e-05, "loss": 2.7292, "step": 38958 }, { "epoch": 3.05, "learning_rate": 6.945110954285266e-05, "loss": 2.6833, "step": 38959 }, { "epoch": 3.05, "learning_rate": 6.945032541362817e-05, "loss": 2.6405, "step": 38960 }, { "epoch": 3.06, "learning_rate": 6.944954128440367e-05, "loss": 2.69, "step": 38961 }, { "epoch": 3.06, "learning_rate": 6.944875715517917e-05, "loss": 2.4698, "step": 38962 }, { "epoch": 3.06, "learning_rate": 6.944797302595468e-05, "loss": 2.6125, "step": 38963 }, { "epoch": 3.06, "learning_rate": 6.944718889673018e-05, "loss": 2.585, "step": 38964 }, { "epoch": 3.06, "learning_rate": 6.944640476750569e-05, "loss": 2.7867, "step": 38965 }, { "epoch": 3.06, "learning_rate": 6.944562063828119e-05, "loss": 2.7595, "step": 38966 }, { "epoch": 3.06, "learning_rate": 6.944483650905669e-05, "loss": 2.7693, "step": 38967 }, { "epoch": 3.06, "learning_rate": 6.94440523798322e-05, "loss": 2.9004, "step": 38968 }, { "epoch": 3.06, "learning_rate": 6.944326825060771e-05, "loss": 2.7241, "step": 38969 }, { "epoch": 3.06, "learning_rate": 6.944248412138322e-05, "loss": 2.6468, "step": 38970 }, { "epoch": 3.06, "learning_rate": 6.944169999215872e-05, "loss": 2.4702, "step": 38971 }, { "epoch": 3.06, "learning_rate": 6.944091586293422e-05, "loss": 2.8999, "step": 38972 }, { "epoch": 3.06, "learning_rate": 6.944013173370973e-05, "loss": 2.6656, "step": 38973 }, { "epoch": 3.06, "learning_rate": 6.943934760448523e-05, "loss": 2.868, "step": 38974 }, { "epoch": 3.06, "learning_rate": 6.943856347526074e-05, "loss": 2.6627, "step": 38975 }, { "epoch": 3.06, "learning_rate": 6.943777934603623e-05, "loss": 2.5414, "step": 38976 }, { "epoch": 3.06, "learning_rate": 6.943699521681173e-05, "loss": 2.8649, "step": 38977 }, { "epoch": 3.06, "learning_rate": 6.943621108758723e-05, "loss": 2.6702, "step": 38978 }, { "epoch": 3.06, "learning_rate": 6.943542695836274e-05, "loss": 2.7218, "step": 38979 }, { "epoch": 3.06, "learning_rate": 6.943464282913824e-05, "loss": 2.8023, "step": 38980 }, { "epoch": 3.06, "learning_rate": 6.943385869991374e-05, "loss": 2.6951, "step": 38981 }, { "epoch": 3.06, "learning_rate": 6.943307457068925e-05, "loss": 2.4313, "step": 38982 }, { "epoch": 3.06, "learning_rate": 6.943229044146475e-05, "loss": 2.586, "step": 38983 }, { "epoch": 3.06, "learning_rate": 6.943150631224026e-05, "loss": 2.846, "step": 38984 }, { "epoch": 3.06, "learning_rate": 6.943072218301576e-05, "loss": 2.5982, "step": 38985 }, { "epoch": 3.06, "learning_rate": 6.942993805379126e-05, "loss": 2.8288, "step": 38986 }, { "epoch": 3.06, "learning_rate": 6.942915392456677e-05, "loss": 2.8158, "step": 38987 }, { "epoch": 3.06, "learning_rate": 6.942836979534227e-05, "loss": 2.814, "step": 38988 }, { "epoch": 3.06, "learning_rate": 6.942758566611777e-05, "loss": 2.7593, "step": 38989 }, { "epoch": 3.06, "learning_rate": 6.942680153689328e-05, "loss": 2.7961, "step": 38990 }, { "epoch": 3.06, "learning_rate": 6.942601740766878e-05, "loss": 3.0448, "step": 38991 }, { "epoch": 3.06, "learning_rate": 6.942523327844428e-05, "loss": 2.5946, "step": 38992 }, { "epoch": 3.06, "learning_rate": 6.942444914921979e-05, "loss": 2.6658, "step": 38993 }, { "epoch": 3.06, "learning_rate": 6.94236650199953e-05, "loss": 2.7637, "step": 38994 }, { "epoch": 3.06, "learning_rate": 6.942288089077081e-05, "loss": 2.5816, "step": 38995 }, { "epoch": 3.06, "learning_rate": 6.942209676154631e-05, "loss": 2.9636, "step": 38996 }, { "epoch": 3.06, "learning_rate": 6.942131263232182e-05, "loss": 2.6705, "step": 38997 }, { "epoch": 3.06, "learning_rate": 6.942052850309732e-05, "loss": 2.6708, "step": 38998 }, { "epoch": 3.06, "learning_rate": 6.941974437387282e-05, "loss": 2.7394, "step": 38999 }, { "epoch": 3.06, "learning_rate": 6.941896024464833e-05, "loss": 2.8307, "step": 39000 }, { "epoch": 3.06, "eval_loss": 2.8986146450042725, "eval_runtime": 217.8272, "eval_samples_per_second": 52.023, "eval_steps_per_second": 3.255, "step": 39000 }, { "epoch": 3.06, "learning_rate": 6.941817611542383e-05, "loss": 2.6, "step": 39001 }, { "epoch": 3.06, "learning_rate": 6.941739198619933e-05, "loss": 2.9309, "step": 39002 }, { "epoch": 3.06, "learning_rate": 6.941660785697484e-05, "loss": 2.6974, "step": 39003 }, { "epoch": 3.06, "learning_rate": 6.941582372775034e-05, "loss": 2.6486, "step": 39004 }, { "epoch": 3.06, "learning_rate": 6.941503959852585e-05, "loss": 2.5347, "step": 39005 }, { "epoch": 3.06, "learning_rate": 6.941425546930135e-05, "loss": 2.8789, "step": 39006 }, { "epoch": 3.06, "learning_rate": 6.941347134007685e-05, "loss": 2.68, "step": 39007 }, { "epoch": 3.06, "learning_rate": 6.941268721085236e-05, "loss": 2.6947, "step": 39008 }, { "epoch": 3.06, "learning_rate": 6.941190308162786e-05, "loss": 2.9265, "step": 39009 }, { "epoch": 3.06, "learning_rate": 6.941111895240336e-05, "loss": 2.7397, "step": 39010 }, { "epoch": 3.06, "learning_rate": 6.941033482317887e-05, "loss": 2.554, "step": 39011 }, { "epoch": 3.06, "learning_rate": 6.940955069395436e-05, "loss": 2.5979, "step": 39012 }, { "epoch": 3.06, "learning_rate": 6.940876656472986e-05, "loss": 2.6097, "step": 39013 }, { "epoch": 3.06, "learning_rate": 6.940798243550537e-05, "loss": 2.689, "step": 39014 }, { "epoch": 3.06, "learning_rate": 6.940719830628087e-05, "loss": 2.766, "step": 39015 }, { "epoch": 3.06, "learning_rate": 6.940641417705637e-05, "loss": 2.717, "step": 39016 }, { "epoch": 3.06, "learning_rate": 6.940563004783188e-05, "loss": 2.7181, "step": 39017 }, { "epoch": 3.06, "learning_rate": 6.940484591860738e-05, "loss": 2.5698, "step": 39018 }, { "epoch": 3.06, "learning_rate": 6.940406178938288e-05, "loss": 2.7219, "step": 39019 }, { "epoch": 3.06, "learning_rate": 6.94032776601584e-05, "loss": 2.6018, "step": 39020 }, { "epoch": 3.06, "learning_rate": 6.94024935309339e-05, "loss": 2.6055, "step": 39021 }, { "epoch": 3.06, "learning_rate": 6.940170940170941e-05, "loss": 2.7328, "step": 39022 }, { "epoch": 3.06, "learning_rate": 6.940092527248491e-05, "loss": 2.6916, "step": 39023 }, { "epoch": 3.06, "learning_rate": 6.940014114326042e-05, "loss": 2.731, "step": 39024 }, { "epoch": 3.06, "learning_rate": 6.939935701403592e-05, "loss": 2.6992, "step": 39025 }, { "epoch": 3.06, "learning_rate": 6.939857288481142e-05, "loss": 2.9001, "step": 39026 }, { "epoch": 3.06, "learning_rate": 6.939778875558693e-05, "loss": 2.6942, "step": 39027 }, { "epoch": 3.06, "learning_rate": 6.939700462636243e-05, "loss": 2.6096, "step": 39028 }, { "epoch": 3.06, "learning_rate": 6.939622049713793e-05, "loss": 2.8492, "step": 39029 }, { "epoch": 3.06, "learning_rate": 6.939543636791344e-05, "loss": 2.6459, "step": 39030 }, { "epoch": 3.06, "learning_rate": 6.939465223868894e-05, "loss": 2.6918, "step": 39031 }, { "epoch": 3.06, "learning_rate": 6.939386810946445e-05, "loss": 2.6711, "step": 39032 }, { "epoch": 3.06, "learning_rate": 6.939308398023995e-05, "loss": 2.5603, "step": 39033 }, { "epoch": 3.06, "learning_rate": 6.939229985101545e-05, "loss": 2.7034, "step": 39034 }, { "epoch": 3.06, "learning_rate": 6.939151572179096e-05, "loss": 2.8349, "step": 39035 }, { "epoch": 3.06, "learning_rate": 6.939073159256646e-05, "loss": 2.8118, "step": 39036 }, { "epoch": 3.06, "learning_rate": 6.938994746334196e-05, "loss": 2.9044, "step": 39037 }, { "epoch": 3.06, "learning_rate": 6.938916333411747e-05, "loss": 2.9339, "step": 39038 }, { "epoch": 3.06, "learning_rate": 6.938837920489297e-05, "loss": 2.7979, "step": 39039 }, { "epoch": 3.06, "learning_rate": 6.938759507566847e-05, "loss": 2.6492, "step": 39040 }, { "epoch": 3.06, "learning_rate": 6.938681094644398e-05, "loss": 2.7352, "step": 39041 }, { "epoch": 3.06, "learning_rate": 6.938602681721948e-05, "loss": 2.9852, "step": 39042 }, { "epoch": 3.06, "learning_rate": 6.938524268799499e-05, "loss": 2.7538, "step": 39043 }, { "epoch": 3.06, "learning_rate": 6.938445855877049e-05, "loss": 2.8943, "step": 39044 }, { "epoch": 3.06, "learning_rate": 6.938367442954599e-05, "loss": 2.9301, "step": 39045 }, { "epoch": 3.06, "learning_rate": 6.93828903003215e-05, "loss": 2.5314, "step": 39046 }, { "epoch": 3.06, "learning_rate": 6.9382106171097e-05, "loss": 2.8869, "step": 39047 }, { "epoch": 3.06, "learning_rate": 6.93813220418725e-05, "loss": 2.6411, "step": 39048 }, { "epoch": 3.06, "learning_rate": 6.938053791264801e-05, "loss": 2.5183, "step": 39049 }, { "epoch": 3.06, "learning_rate": 6.937975378342351e-05, "loss": 2.6834, "step": 39050 }, { "epoch": 3.06, "learning_rate": 6.937896965419902e-05, "loss": 2.6704, "step": 39051 }, { "epoch": 3.06, "learning_rate": 6.937818552497452e-05, "loss": 2.7125, "step": 39052 }, { "epoch": 3.06, "learning_rate": 6.937740139575002e-05, "loss": 2.6055, "step": 39053 }, { "epoch": 3.06, "learning_rate": 6.937661726652553e-05, "loss": 3.0874, "step": 39054 }, { "epoch": 3.06, "learning_rate": 6.937583313730103e-05, "loss": 2.7004, "step": 39055 }, { "epoch": 3.06, "learning_rate": 6.937504900807653e-05, "loss": 2.5884, "step": 39056 }, { "epoch": 3.06, "learning_rate": 6.937426487885204e-05, "loss": 2.9122, "step": 39057 }, { "epoch": 3.06, "learning_rate": 6.937348074962754e-05, "loss": 2.8424, "step": 39058 }, { "epoch": 3.06, "learning_rate": 6.937269662040304e-05, "loss": 2.6457, "step": 39059 }, { "epoch": 3.06, "learning_rate": 6.937191249117855e-05, "loss": 2.5701, "step": 39060 }, { "epoch": 3.06, "learning_rate": 6.937112836195405e-05, "loss": 2.7567, "step": 39061 }, { "epoch": 3.06, "learning_rate": 6.937034423272956e-05, "loss": 2.9288, "step": 39062 }, { "epoch": 3.06, "learning_rate": 6.936956010350506e-05, "loss": 2.6005, "step": 39063 }, { "epoch": 3.06, "learning_rate": 6.936877597428056e-05, "loss": 2.6685, "step": 39064 }, { "epoch": 3.06, "learning_rate": 6.936799184505607e-05, "loss": 2.7805, "step": 39065 }, { "epoch": 3.06, "learning_rate": 6.936720771583157e-05, "loss": 2.7433, "step": 39066 }, { "epoch": 3.06, "learning_rate": 6.936642358660707e-05, "loss": 2.6285, "step": 39067 }, { "epoch": 3.06, "learning_rate": 6.936563945738258e-05, "loss": 2.8482, "step": 39068 }, { "epoch": 3.06, "learning_rate": 6.936485532815808e-05, "loss": 2.5839, "step": 39069 }, { "epoch": 3.06, "learning_rate": 6.936407119893358e-05, "loss": 2.537, "step": 39070 }, { "epoch": 3.06, "learning_rate": 6.936328706970909e-05, "loss": 2.8916, "step": 39071 }, { "epoch": 3.06, "learning_rate": 6.936250294048459e-05, "loss": 2.6172, "step": 39072 }, { "epoch": 3.06, "learning_rate": 6.93617188112601e-05, "loss": 2.7375, "step": 39073 }, { "epoch": 3.06, "learning_rate": 6.93609346820356e-05, "loss": 2.6098, "step": 39074 }, { "epoch": 3.06, "learning_rate": 6.93601505528111e-05, "loss": 2.6104, "step": 39075 }, { "epoch": 3.06, "learning_rate": 6.935936642358661e-05, "loss": 2.8201, "step": 39076 }, { "epoch": 3.06, "learning_rate": 6.935858229436212e-05, "loss": 2.6633, "step": 39077 }, { "epoch": 3.06, "learning_rate": 6.935779816513763e-05, "loss": 2.7813, "step": 39078 }, { "epoch": 3.06, "learning_rate": 6.935701403591313e-05, "loss": 2.6662, "step": 39079 }, { "epoch": 3.06, "learning_rate": 6.935622990668864e-05, "loss": 2.8167, "step": 39080 }, { "epoch": 3.06, "learning_rate": 6.935544577746414e-05, "loss": 2.7173, "step": 39081 }, { "epoch": 3.06, "learning_rate": 6.935466164823964e-05, "loss": 2.8151, "step": 39082 }, { "epoch": 3.06, "learning_rate": 6.935387751901515e-05, "loss": 2.717, "step": 39083 }, { "epoch": 3.06, "learning_rate": 6.935309338979064e-05, "loss": 2.5418, "step": 39084 }, { "epoch": 3.06, "learning_rate": 6.935230926056614e-05, "loss": 2.7939, "step": 39085 }, { "epoch": 3.06, "learning_rate": 6.935152513134164e-05, "loss": 2.6316, "step": 39086 }, { "epoch": 3.06, "learning_rate": 6.935074100211715e-05, "loss": 2.8241, "step": 39087 }, { "epoch": 3.07, "learning_rate": 6.934995687289265e-05, "loss": 3.0218, "step": 39088 }, { "epoch": 3.07, "learning_rate": 6.934917274366815e-05, "loss": 2.8102, "step": 39089 }, { "epoch": 3.07, "learning_rate": 6.934838861444366e-05, "loss": 2.7798, "step": 39090 }, { "epoch": 3.07, "learning_rate": 6.934760448521916e-05, "loss": 2.8308, "step": 39091 }, { "epoch": 3.07, "learning_rate": 6.934682035599467e-05, "loss": 2.6915, "step": 39092 }, { "epoch": 3.07, "learning_rate": 6.934603622677017e-05, "loss": 2.692, "step": 39093 }, { "epoch": 3.07, "learning_rate": 6.934525209754567e-05, "loss": 2.6089, "step": 39094 }, { "epoch": 3.07, "learning_rate": 6.934446796832118e-05, "loss": 2.6535, "step": 39095 }, { "epoch": 3.07, "learning_rate": 6.934368383909668e-05, "loss": 2.7075, "step": 39096 }, { "epoch": 3.07, "learning_rate": 6.934289970987218e-05, "loss": 2.6874, "step": 39097 }, { "epoch": 3.07, "learning_rate": 6.934211558064769e-05, "loss": 2.6823, "step": 39098 }, { "epoch": 3.07, "learning_rate": 6.934133145142319e-05, "loss": 2.9344, "step": 39099 }, { "epoch": 3.07, "learning_rate": 6.93405473221987e-05, "loss": 2.55, "step": 39100 }, { "epoch": 3.07, "learning_rate": 6.93397631929742e-05, "loss": 2.7585, "step": 39101 }, { "epoch": 3.07, "learning_rate": 6.933897906374972e-05, "loss": 2.8489, "step": 39102 }, { "epoch": 3.07, "learning_rate": 6.933819493452522e-05, "loss": 2.7169, "step": 39103 }, { "epoch": 3.07, "learning_rate": 6.933741080530072e-05, "loss": 2.5952, "step": 39104 }, { "epoch": 3.07, "learning_rate": 6.933662667607623e-05, "loss": 2.8078, "step": 39105 }, { "epoch": 3.07, "learning_rate": 6.933584254685173e-05, "loss": 2.7261, "step": 39106 }, { "epoch": 3.07, "learning_rate": 6.933505841762723e-05, "loss": 2.8132, "step": 39107 }, { "epoch": 3.07, "learning_rate": 6.933427428840274e-05, "loss": 2.6553, "step": 39108 }, { "epoch": 3.07, "learning_rate": 6.933349015917824e-05, "loss": 2.7129, "step": 39109 }, { "epoch": 3.07, "learning_rate": 6.933270602995375e-05, "loss": 2.838, "step": 39110 }, { "epoch": 3.07, "learning_rate": 6.933192190072925e-05, "loss": 2.9817, "step": 39111 }, { "epoch": 3.07, "learning_rate": 6.933113777150475e-05, "loss": 2.7616, "step": 39112 }, { "epoch": 3.07, "learning_rate": 6.933035364228026e-05, "loss": 2.7325, "step": 39113 }, { "epoch": 3.07, "learning_rate": 6.932956951305576e-05, "loss": 2.7045, "step": 39114 }, { "epoch": 3.07, "learning_rate": 6.932878538383126e-05, "loss": 2.6907, "step": 39115 }, { "epoch": 3.07, "learning_rate": 6.932800125460677e-05, "loss": 2.784, "step": 39116 }, { "epoch": 3.07, "learning_rate": 6.932721712538227e-05, "loss": 2.7022, "step": 39117 }, { "epoch": 3.07, "learning_rate": 6.932643299615777e-05, "loss": 2.6266, "step": 39118 }, { "epoch": 3.07, "learning_rate": 6.932564886693328e-05, "loss": 2.7598, "step": 39119 }, { "epoch": 3.07, "learning_rate": 6.932486473770877e-05, "loss": 2.8626, "step": 39120 }, { "epoch": 3.07, "learning_rate": 6.932408060848427e-05, "loss": 2.7678, "step": 39121 }, { "epoch": 3.07, "learning_rate": 6.932329647925978e-05, "loss": 2.6629, "step": 39122 }, { "epoch": 3.07, "learning_rate": 6.932251235003528e-05, "loss": 2.7916, "step": 39123 }, { "epoch": 3.07, "learning_rate": 6.932172822081078e-05, "loss": 2.7658, "step": 39124 }, { "epoch": 3.07, "learning_rate": 6.932094409158629e-05, "loss": 2.666, "step": 39125 }, { "epoch": 3.07, "learning_rate": 6.932015996236179e-05, "loss": 2.6482, "step": 39126 }, { "epoch": 3.07, "learning_rate": 6.93193758331373e-05, "loss": 2.7398, "step": 39127 }, { "epoch": 3.07, "learning_rate": 6.931859170391281e-05, "loss": 2.5727, "step": 39128 }, { "epoch": 3.07, "learning_rate": 6.931780757468832e-05, "loss": 2.8053, "step": 39129 }, { "epoch": 3.07, "learning_rate": 6.931702344546382e-05, "loss": 2.6808, "step": 39130 }, { "epoch": 3.07, "learning_rate": 6.931623931623932e-05, "loss": 2.7285, "step": 39131 }, { "epoch": 3.07, "learning_rate": 6.931545518701483e-05, "loss": 2.7893, "step": 39132 }, { "epoch": 3.07, "learning_rate": 6.931467105779033e-05, "loss": 2.6775, "step": 39133 }, { "epoch": 3.07, "learning_rate": 6.931388692856583e-05, "loss": 2.8589, "step": 39134 }, { "epoch": 3.07, "learning_rate": 6.931310279934134e-05, "loss": 2.5963, "step": 39135 }, { "epoch": 3.07, "learning_rate": 6.931231867011684e-05, "loss": 2.5324, "step": 39136 }, { "epoch": 3.07, "learning_rate": 6.931153454089234e-05, "loss": 3.0113, "step": 39137 }, { "epoch": 3.07, "learning_rate": 6.931075041166785e-05, "loss": 2.6958, "step": 39138 }, { "epoch": 3.07, "learning_rate": 6.930996628244335e-05, "loss": 2.6229, "step": 39139 }, { "epoch": 3.07, "learning_rate": 6.930918215321886e-05, "loss": 2.7188, "step": 39140 }, { "epoch": 3.07, "learning_rate": 6.930839802399436e-05, "loss": 2.6007, "step": 39141 }, { "epoch": 3.07, "learning_rate": 6.930761389476986e-05, "loss": 2.5812, "step": 39142 }, { "epoch": 3.07, "learning_rate": 6.930682976554537e-05, "loss": 2.7368, "step": 39143 }, { "epoch": 3.07, "learning_rate": 6.930604563632087e-05, "loss": 2.6686, "step": 39144 }, { "epoch": 3.07, "learning_rate": 6.930526150709637e-05, "loss": 2.6818, "step": 39145 }, { "epoch": 3.07, "learning_rate": 6.930447737787188e-05, "loss": 2.776, "step": 39146 }, { "epoch": 3.07, "learning_rate": 6.930369324864738e-05, "loss": 2.6549, "step": 39147 }, { "epoch": 3.07, "learning_rate": 6.930290911942289e-05, "loss": 2.7282, "step": 39148 }, { "epoch": 3.07, "learning_rate": 6.930212499019839e-05, "loss": 2.8313, "step": 39149 }, { "epoch": 3.07, "learning_rate": 6.930134086097389e-05, "loss": 2.9002, "step": 39150 }, { "epoch": 3.07, "learning_rate": 6.93005567317494e-05, "loss": 2.6593, "step": 39151 }, { "epoch": 3.07, "learning_rate": 6.92997726025249e-05, "loss": 2.7432, "step": 39152 }, { "epoch": 3.07, "learning_rate": 6.92989884733004e-05, "loss": 2.7143, "step": 39153 }, { "epoch": 3.07, "learning_rate": 6.929820434407591e-05, "loss": 2.6292, "step": 39154 }, { "epoch": 3.07, "learning_rate": 6.929742021485141e-05, "loss": 2.7156, "step": 39155 }, { "epoch": 3.07, "learning_rate": 6.929663608562691e-05, "loss": 2.7767, "step": 39156 }, { "epoch": 3.07, "learning_rate": 6.929585195640242e-05, "loss": 2.7425, "step": 39157 }, { "epoch": 3.07, "learning_rate": 6.929506782717792e-05, "loss": 2.6189, "step": 39158 }, { "epoch": 3.07, "learning_rate": 6.929428369795343e-05, "loss": 2.8869, "step": 39159 }, { "epoch": 3.07, "learning_rate": 6.929349956872893e-05, "loss": 2.7841, "step": 39160 }, { "epoch": 3.07, "learning_rate": 6.929271543950443e-05, "loss": 2.7752, "step": 39161 }, { "epoch": 3.07, "learning_rate": 6.929193131027994e-05, "loss": 2.7671, "step": 39162 }, { "epoch": 3.07, "learning_rate": 6.929114718105544e-05, "loss": 2.7651, "step": 39163 }, { "epoch": 3.07, "learning_rate": 6.929036305183094e-05, "loss": 2.5769, "step": 39164 }, { "epoch": 3.07, "learning_rate": 6.928957892260645e-05, "loss": 2.6932, "step": 39165 }, { "epoch": 3.07, "learning_rate": 6.928879479338195e-05, "loss": 2.6146, "step": 39166 }, { "epoch": 3.07, "learning_rate": 6.928801066415746e-05, "loss": 2.6675, "step": 39167 }, { "epoch": 3.07, "learning_rate": 6.928722653493296e-05, "loss": 2.7187, "step": 39168 }, { "epoch": 3.07, "learning_rate": 6.928644240570846e-05, "loss": 2.5847, "step": 39169 }, { "epoch": 3.07, "learning_rate": 6.928565827648397e-05, "loss": 2.7709, "step": 39170 }, { "epoch": 3.07, "learning_rate": 6.928487414725947e-05, "loss": 2.8914, "step": 39171 }, { "epoch": 3.07, "learning_rate": 6.928409001803497e-05, "loss": 2.7499, "step": 39172 }, { "epoch": 3.07, "learning_rate": 6.928330588881048e-05, "loss": 2.9048, "step": 39173 }, { "epoch": 3.07, "learning_rate": 6.928252175958598e-05, "loss": 2.8065, "step": 39174 }, { "epoch": 3.07, "learning_rate": 6.928173763036148e-05, "loss": 2.4524, "step": 39175 }, { "epoch": 3.07, "learning_rate": 6.928095350113699e-05, "loss": 2.6564, "step": 39176 }, { "epoch": 3.07, "learning_rate": 6.928016937191249e-05, "loss": 2.7149, "step": 39177 }, { "epoch": 3.07, "learning_rate": 6.9279385242688e-05, "loss": 2.5767, "step": 39178 }, { "epoch": 3.07, "learning_rate": 6.92786011134635e-05, "loss": 2.7931, "step": 39179 }, { "epoch": 3.07, "learning_rate": 6.9277816984239e-05, "loss": 2.8487, "step": 39180 }, { "epoch": 3.07, "learning_rate": 6.92770328550145e-05, "loss": 2.5156, "step": 39181 }, { "epoch": 3.07, "learning_rate": 6.927624872579001e-05, "loss": 2.759, "step": 39182 }, { "epoch": 3.07, "learning_rate": 6.927546459656551e-05, "loss": 2.6804, "step": 39183 }, { "epoch": 3.07, "learning_rate": 6.927468046734102e-05, "loss": 2.9531, "step": 39184 }, { "epoch": 3.07, "learning_rate": 6.927389633811653e-05, "loss": 2.4427, "step": 39185 }, { "epoch": 3.07, "learning_rate": 6.927311220889204e-05, "loss": 2.7973, "step": 39186 }, { "epoch": 3.07, "learning_rate": 6.927232807966754e-05, "loss": 2.8369, "step": 39187 }, { "epoch": 3.07, "learning_rate": 6.927154395044305e-05, "loss": 2.6663, "step": 39188 }, { "epoch": 3.07, "learning_rate": 6.927075982121855e-05, "loss": 2.5723, "step": 39189 }, { "epoch": 3.07, "learning_rate": 6.926997569199405e-05, "loss": 2.6587, "step": 39190 }, { "epoch": 3.07, "learning_rate": 6.926919156276956e-05, "loss": 2.7524, "step": 39191 }, { "epoch": 3.07, "learning_rate": 6.926840743354505e-05, "loss": 2.8261, "step": 39192 }, { "epoch": 3.07, "learning_rate": 6.926762330432055e-05, "loss": 2.8352, "step": 39193 }, { "epoch": 3.07, "learning_rate": 6.926683917509605e-05, "loss": 2.7297, "step": 39194 }, { "epoch": 3.07, "learning_rate": 6.926605504587156e-05, "loss": 2.5565, "step": 39195 }, { "epoch": 3.07, "learning_rate": 6.926527091664706e-05, "loss": 2.7877, "step": 39196 }, { "epoch": 3.07, "learning_rate": 6.926448678742257e-05, "loss": 2.7275, "step": 39197 }, { "epoch": 3.07, "learning_rate": 6.926370265819807e-05, "loss": 2.927, "step": 39198 }, { "epoch": 3.07, "learning_rate": 6.926291852897357e-05, "loss": 2.7091, "step": 39199 }, { "epoch": 3.07, "learning_rate": 6.926213439974908e-05, "loss": 2.4732, "step": 39200 }, { "epoch": 3.07, "learning_rate": 6.926135027052458e-05, "loss": 2.5407, "step": 39201 }, { "epoch": 3.07, "learning_rate": 6.926056614130008e-05, "loss": 2.8003, "step": 39202 }, { "epoch": 3.07, "learning_rate": 6.925978201207559e-05, "loss": 2.72, "step": 39203 }, { "epoch": 3.07, "learning_rate": 6.925899788285109e-05, "loss": 3.0879, "step": 39204 }, { "epoch": 3.07, "learning_rate": 6.92582137536266e-05, "loss": 2.7675, "step": 39205 }, { "epoch": 3.07, "learning_rate": 6.92574296244021e-05, "loss": 2.8236, "step": 39206 }, { "epoch": 3.07, "learning_rate": 6.92566454951776e-05, "loss": 2.6167, "step": 39207 }, { "epoch": 3.07, "learning_rate": 6.92558613659531e-05, "loss": 2.7759, "step": 39208 }, { "epoch": 3.07, "learning_rate": 6.925507723672861e-05, "loss": 2.7244, "step": 39209 }, { "epoch": 3.07, "learning_rate": 6.925429310750413e-05, "loss": 2.8283, "step": 39210 }, { "epoch": 3.07, "learning_rate": 6.925350897827963e-05, "loss": 2.799, "step": 39211 }, { "epoch": 3.07, "learning_rate": 6.925272484905513e-05, "loss": 2.524, "step": 39212 }, { "epoch": 3.07, "learning_rate": 6.925194071983064e-05, "loss": 2.6388, "step": 39213 }, { "epoch": 3.07, "learning_rate": 6.925115659060614e-05, "loss": 2.9506, "step": 39214 }, { "epoch": 3.07, "learning_rate": 6.925037246138165e-05, "loss": 2.7396, "step": 39215 }, { "epoch": 3.08, "learning_rate": 6.924958833215715e-05, "loss": 3.026, "step": 39216 }, { "epoch": 3.08, "learning_rate": 6.924880420293265e-05, "loss": 2.6838, "step": 39217 }, { "epoch": 3.08, "learning_rate": 6.924802007370816e-05, "loss": 2.7539, "step": 39218 }, { "epoch": 3.08, "learning_rate": 6.924723594448366e-05, "loss": 2.6774, "step": 39219 }, { "epoch": 3.08, "learning_rate": 6.924645181525916e-05, "loss": 2.9237, "step": 39220 }, { "epoch": 3.08, "learning_rate": 6.924566768603467e-05, "loss": 3.0188, "step": 39221 }, { "epoch": 3.08, "learning_rate": 6.924488355681017e-05, "loss": 2.6983, "step": 39222 }, { "epoch": 3.08, "learning_rate": 6.924409942758567e-05, "loss": 2.5473, "step": 39223 }, { "epoch": 3.08, "learning_rate": 6.924331529836118e-05, "loss": 2.9383, "step": 39224 }, { "epoch": 3.08, "learning_rate": 6.924253116913668e-05, "loss": 3.1133, "step": 39225 }, { "epoch": 3.08, "learning_rate": 6.924174703991219e-05, "loss": 2.7117, "step": 39226 }, { "epoch": 3.08, "learning_rate": 6.924096291068769e-05, "loss": 2.7019, "step": 39227 }, { "epoch": 3.08, "learning_rate": 6.924017878146318e-05, "loss": 2.6184, "step": 39228 }, { "epoch": 3.08, "learning_rate": 6.923939465223868e-05, "loss": 2.561, "step": 39229 }, { "epoch": 3.08, "learning_rate": 6.923861052301419e-05, "loss": 2.6111, "step": 39230 }, { "epoch": 3.08, "learning_rate": 6.923782639378969e-05, "loss": 2.6194, "step": 39231 }, { "epoch": 3.08, "learning_rate": 6.92370422645652e-05, "loss": 2.6986, "step": 39232 }, { "epoch": 3.08, "learning_rate": 6.92362581353407e-05, "loss": 2.6051, "step": 39233 }, { "epoch": 3.08, "learning_rate": 6.92354740061162e-05, "loss": 2.7757, "step": 39234 }, { "epoch": 3.08, "learning_rate": 6.92346898768917e-05, "loss": 2.6473, "step": 39235 }, { "epoch": 3.08, "learning_rate": 6.923390574766722e-05, "loss": 2.8227, "step": 39236 }, { "epoch": 3.08, "learning_rate": 6.923312161844273e-05, "loss": 2.9226, "step": 39237 }, { "epoch": 3.08, "learning_rate": 6.923233748921823e-05, "loss": 2.8786, "step": 39238 }, { "epoch": 3.08, "learning_rate": 6.923155335999373e-05, "loss": 2.8009, "step": 39239 }, { "epoch": 3.08, "learning_rate": 6.923076923076924e-05, "loss": 2.5484, "step": 39240 }, { "epoch": 3.08, "learning_rate": 6.922998510154474e-05, "loss": 2.6919, "step": 39241 }, { "epoch": 3.08, "learning_rate": 6.922920097232024e-05, "loss": 2.6578, "step": 39242 }, { "epoch": 3.08, "learning_rate": 6.922841684309575e-05, "loss": 2.8038, "step": 39243 }, { "epoch": 3.08, "learning_rate": 6.922763271387125e-05, "loss": 2.8634, "step": 39244 }, { "epoch": 3.08, "learning_rate": 6.922684858464676e-05, "loss": 2.6535, "step": 39245 }, { "epoch": 3.08, "learning_rate": 6.922606445542226e-05, "loss": 2.7413, "step": 39246 }, { "epoch": 3.08, "learning_rate": 6.922528032619776e-05, "loss": 2.8037, "step": 39247 }, { "epoch": 3.08, "learning_rate": 6.922449619697327e-05, "loss": 2.8124, "step": 39248 }, { "epoch": 3.08, "learning_rate": 6.922371206774877e-05, "loss": 2.9519, "step": 39249 }, { "epoch": 3.08, "learning_rate": 6.922292793852427e-05, "loss": 2.73, "step": 39250 }, { "epoch": 3.08, "learning_rate": 6.922214380929978e-05, "loss": 2.9495, "step": 39251 }, { "epoch": 3.08, "learning_rate": 6.922135968007528e-05, "loss": 2.5847, "step": 39252 }, { "epoch": 3.08, "learning_rate": 6.922057555085079e-05, "loss": 2.7215, "step": 39253 }, { "epoch": 3.08, "learning_rate": 6.921979142162629e-05, "loss": 2.6497, "step": 39254 }, { "epoch": 3.08, "learning_rate": 6.921900729240179e-05, "loss": 2.7746, "step": 39255 }, { "epoch": 3.08, "learning_rate": 6.92182231631773e-05, "loss": 2.8726, "step": 39256 }, { "epoch": 3.08, "learning_rate": 6.92174390339528e-05, "loss": 2.8051, "step": 39257 }, { "epoch": 3.08, "learning_rate": 6.92166549047283e-05, "loss": 3.0771, "step": 39258 }, { "epoch": 3.08, "learning_rate": 6.921587077550381e-05, "loss": 2.5852, "step": 39259 }, { "epoch": 3.08, "learning_rate": 6.921508664627931e-05, "loss": 2.6683, "step": 39260 }, { "epoch": 3.08, "learning_rate": 6.921430251705481e-05, "loss": 2.576, "step": 39261 }, { "epoch": 3.08, "learning_rate": 6.921351838783032e-05, "loss": 2.799, "step": 39262 }, { "epoch": 3.08, "learning_rate": 6.921273425860582e-05, "loss": 2.7318, "step": 39263 }, { "epoch": 3.08, "learning_rate": 6.921195012938133e-05, "loss": 2.635, "step": 39264 }, { "epoch": 3.08, "learning_rate": 6.921116600015683e-05, "loss": 2.7449, "step": 39265 }, { "epoch": 3.08, "learning_rate": 6.921038187093233e-05, "loss": 2.5063, "step": 39266 }, { "epoch": 3.08, "learning_rate": 6.920959774170784e-05, "loss": 2.6398, "step": 39267 }, { "epoch": 3.08, "learning_rate": 6.920881361248334e-05, "loss": 2.8611, "step": 39268 }, { "epoch": 3.08, "learning_rate": 6.920802948325884e-05, "loss": 2.7088, "step": 39269 }, { "epoch": 3.08, "learning_rate": 6.920724535403435e-05, "loss": 2.6657, "step": 39270 }, { "epoch": 3.08, "learning_rate": 6.920646122480985e-05, "loss": 2.5069, "step": 39271 }, { "epoch": 3.08, "learning_rate": 6.920567709558535e-05, "loss": 2.7344, "step": 39272 }, { "epoch": 3.08, "learning_rate": 6.920489296636086e-05, "loss": 2.7134, "step": 39273 }, { "epoch": 3.08, "learning_rate": 6.920410883713636e-05, "loss": 2.6536, "step": 39274 }, { "epoch": 3.08, "learning_rate": 6.920332470791187e-05, "loss": 2.8382, "step": 39275 }, { "epoch": 3.08, "learning_rate": 6.920254057868737e-05, "loss": 2.8401, "step": 39276 }, { "epoch": 3.08, "learning_rate": 6.920175644946287e-05, "loss": 2.4582, "step": 39277 }, { "epoch": 3.08, "learning_rate": 6.920097232023838e-05, "loss": 2.8199, "step": 39278 }, { "epoch": 3.08, "learning_rate": 6.920018819101388e-05, "loss": 2.7851, "step": 39279 }, { "epoch": 3.08, "learning_rate": 6.919940406178938e-05, "loss": 2.7503, "step": 39280 }, { "epoch": 3.08, "learning_rate": 6.919861993256489e-05, "loss": 2.7214, "step": 39281 }, { "epoch": 3.08, "learning_rate": 6.919783580334039e-05, "loss": 2.8379, "step": 39282 }, { "epoch": 3.08, "learning_rate": 6.91970516741159e-05, "loss": 2.5322, "step": 39283 }, { "epoch": 3.08, "learning_rate": 6.91962675448914e-05, "loss": 2.713, "step": 39284 }, { "epoch": 3.08, "learning_rate": 6.91954834156669e-05, "loss": 2.5097, "step": 39285 }, { "epoch": 3.08, "learning_rate": 6.91946992864424e-05, "loss": 2.7859, "step": 39286 }, { "epoch": 3.08, "learning_rate": 6.919391515721791e-05, "loss": 2.6138, "step": 39287 }, { "epoch": 3.08, "learning_rate": 6.919313102799341e-05, "loss": 2.5402, "step": 39288 }, { "epoch": 3.08, "learning_rate": 6.919234689876892e-05, "loss": 2.7805, "step": 39289 }, { "epoch": 3.08, "learning_rate": 6.919156276954442e-05, "loss": 2.5969, "step": 39290 }, { "epoch": 3.08, "learning_rate": 6.919077864031992e-05, "loss": 3.001, "step": 39291 }, { "epoch": 3.08, "learning_rate": 6.918999451109543e-05, "loss": 2.7489, "step": 39292 }, { "epoch": 3.08, "learning_rate": 6.918921038187095e-05, "loss": 2.5993, "step": 39293 }, { "epoch": 3.08, "learning_rate": 6.918842625264645e-05, "loss": 2.7009, "step": 39294 }, { "epoch": 3.08, "learning_rate": 6.918764212342195e-05, "loss": 2.9267, "step": 39295 }, { "epoch": 3.08, "learning_rate": 6.918685799419746e-05, "loss": 2.5537, "step": 39296 }, { "epoch": 3.08, "learning_rate": 6.918607386497296e-05, "loss": 2.6722, "step": 39297 }, { "epoch": 3.08, "learning_rate": 6.918528973574846e-05, "loss": 2.7669, "step": 39298 }, { "epoch": 3.08, "learning_rate": 6.918450560652397e-05, "loss": 2.8379, "step": 39299 }, { "epoch": 3.08, "learning_rate": 6.918372147729946e-05, "loss": 2.6739, "step": 39300 }, { "epoch": 3.08, "learning_rate": 6.918293734807496e-05, "loss": 2.5272, "step": 39301 }, { "epoch": 3.08, "learning_rate": 6.918215321885047e-05, "loss": 2.5847, "step": 39302 }, { "epoch": 3.08, "learning_rate": 6.918136908962597e-05, "loss": 2.7116, "step": 39303 }, { "epoch": 3.08, "learning_rate": 6.918058496040147e-05, "loss": 2.6771, "step": 39304 }, { "epoch": 3.08, "learning_rate": 6.917980083117698e-05, "loss": 2.7019, "step": 39305 }, { "epoch": 3.08, "learning_rate": 6.917901670195248e-05, "loss": 2.6871, "step": 39306 }, { "epoch": 3.08, "learning_rate": 6.917823257272798e-05, "loss": 2.9404, "step": 39307 }, { "epoch": 3.08, "learning_rate": 6.917744844350349e-05, "loss": 2.6513, "step": 39308 }, { "epoch": 3.08, "learning_rate": 6.917666431427899e-05, "loss": 2.9323, "step": 39309 }, { "epoch": 3.08, "learning_rate": 6.91758801850545e-05, "loss": 2.7218, "step": 39310 }, { "epoch": 3.08, "learning_rate": 6.917509605583e-05, "loss": 2.6162, "step": 39311 }, { "epoch": 3.08, "learning_rate": 6.91743119266055e-05, "loss": 2.6129, "step": 39312 }, { "epoch": 3.08, "learning_rate": 6.9173527797381e-05, "loss": 2.6224, "step": 39313 }, { "epoch": 3.08, "learning_rate": 6.917274366815651e-05, "loss": 2.8581, "step": 39314 }, { "epoch": 3.08, "learning_rate": 6.917195953893201e-05, "loss": 2.7604, "step": 39315 }, { "epoch": 3.08, "learning_rate": 6.917117540970752e-05, "loss": 2.7798, "step": 39316 }, { "epoch": 3.08, "learning_rate": 6.917039128048302e-05, "loss": 2.744, "step": 39317 }, { "epoch": 3.08, "learning_rate": 6.916960715125854e-05, "loss": 2.7265, "step": 39318 }, { "epoch": 3.08, "learning_rate": 6.916882302203404e-05, "loss": 3.1337, "step": 39319 }, { "epoch": 3.08, "learning_rate": 6.916803889280954e-05, "loss": 2.5944, "step": 39320 }, { "epoch": 3.08, "learning_rate": 6.916725476358505e-05, "loss": 2.6643, "step": 39321 }, { "epoch": 3.08, "learning_rate": 6.916647063436055e-05, "loss": 2.7545, "step": 39322 }, { "epoch": 3.08, "learning_rate": 6.916568650513606e-05, "loss": 2.6004, "step": 39323 }, { "epoch": 3.08, "learning_rate": 6.916490237591156e-05, "loss": 2.9096, "step": 39324 }, { "epoch": 3.08, "learning_rate": 6.916411824668706e-05, "loss": 2.5994, "step": 39325 }, { "epoch": 3.08, "learning_rate": 6.916333411746257e-05, "loss": 2.5327, "step": 39326 }, { "epoch": 3.08, "learning_rate": 6.916254998823807e-05, "loss": 2.58, "step": 39327 }, { "epoch": 3.08, "learning_rate": 6.916176585901357e-05, "loss": 2.4959, "step": 39328 }, { "epoch": 3.08, "learning_rate": 6.916098172978908e-05, "loss": 2.5969, "step": 39329 }, { "epoch": 3.08, "learning_rate": 6.916019760056458e-05, "loss": 2.7942, "step": 39330 }, { "epoch": 3.08, "learning_rate": 6.915941347134009e-05, "loss": 2.9957, "step": 39331 }, { "epoch": 3.08, "learning_rate": 6.915862934211559e-05, "loss": 2.6856, "step": 39332 }, { "epoch": 3.08, "learning_rate": 6.915784521289109e-05, "loss": 2.759, "step": 39333 }, { "epoch": 3.08, "learning_rate": 6.91570610836666e-05, "loss": 2.6105, "step": 39334 }, { "epoch": 3.08, "learning_rate": 6.91562769544421e-05, "loss": 2.6631, "step": 39335 }, { "epoch": 3.08, "learning_rate": 6.915549282521759e-05, "loss": 2.5594, "step": 39336 }, { "epoch": 3.08, "learning_rate": 6.91547086959931e-05, "loss": 2.6866, "step": 39337 }, { "epoch": 3.08, "learning_rate": 6.91539245667686e-05, "loss": 2.6465, "step": 39338 }, { "epoch": 3.08, "learning_rate": 6.91531404375441e-05, "loss": 2.6476, "step": 39339 }, { "epoch": 3.08, "learning_rate": 6.91523563083196e-05, "loss": 2.8186, "step": 39340 }, { "epoch": 3.08, "learning_rate": 6.915157217909511e-05, "loss": 2.7872, "step": 39341 }, { "epoch": 3.08, "learning_rate": 6.915078804987061e-05, "loss": 2.8522, "step": 39342 }, { "epoch": 3.08, "learning_rate": 6.915000392064612e-05, "loss": 2.8149, "step": 39343 }, { "epoch": 3.09, "learning_rate": 6.914921979142163e-05, "loss": 3.0342, "step": 39344 }, { "epoch": 3.09, "learning_rate": 6.914843566219714e-05, "loss": 2.7407, "step": 39345 }, { "epoch": 3.09, "learning_rate": 6.914765153297264e-05, "loss": 2.5645, "step": 39346 }, { "epoch": 3.09, "learning_rate": 6.914686740374814e-05, "loss": 2.715, "step": 39347 }, { "epoch": 3.09, "learning_rate": 6.914608327452365e-05, "loss": 2.6542, "step": 39348 }, { "epoch": 3.09, "learning_rate": 6.914529914529915e-05, "loss": 2.5173, "step": 39349 }, { "epoch": 3.09, "learning_rate": 6.914451501607466e-05, "loss": 2.6474, "step": 39350 }, { "epoch": 3.09, "learning_rate": 6.914373088685016e-05, "loss": 2.8953, "step": 39351 }, { "epoch": 3.09, "learning_rate": 6.914294675762566e-05, "loss": 2.6908, "step": 39352 }, { "epoch": 3.09, "learning_rate": 6.914216262840117e-05, "loss": 2.6592, "step": 39353 }, { "epoch": 3.09, "learning_rate": 6.914137849917667e-05, "loss": 2.6891, "step": 39354 }, { "epoch": 3.09, "learning_rate": 6.914059436995217e-05, "loss": 2.8291, "step": 39355 }, { "epoch": 3.09, "learning_rate": 6.913981024072768e-05, "loss": 2.688, "step": 39356 }, { "epoch": 3.09, "learning_rate": 6.913902611150318e-05, "loss": 2.7281, "step": 39357 }, { "epoch": 3.09, "learning_rate": 6.913824198227868e-05, "loss": 2.6084, "step": 39358 }, { "epoch": 3.09, "learning_rate": 6.913745785305419e-05, "loss": 2.5239, "step": 39359 }, { "epoch": 3.09, "learning_rate": 6.913667372382969e-05, "loss": 2.6191, "step": 39360 }, { "epoch": 3.09, "learning_rate": 6.91358895946052e-05, "loss": 2.7047, "step": 39361 }, { "epoch": 3.09, "learning_rate": 6.91351054653807e-05, "loss": 2.7765, "step": 39362 }, { "epoch": 3.09, "learning_rate": 6.91343213361562e-05, "loss": 2.8065, "step": 39363 }, { "epoch": 3.09, "learning_rate": 6.913353720693171e-05, "loss": 2.4955, "step": 39364 }, { "epoch": 3.09, "learning_rate": 6.913275307770721e-05, "loss": 2.8751, "step": 39365 }, { "epoch": 3.09, "learning_rate": 6.913196894848271e-05, "loss": 2.846, "step": 39366 }, { "epoch": 3.09, "learning_rate": 6.913118481925822e-05, "loss": 2.7494, "step": 39367 }, { "epoch": 3.09, "learning_rate": 6.913040069003372e-05, "loss": 2.7181, "step": 39368 }, { "epoch": 3.09, "learning_rate": 6.912961656080923e-05, "loss": 2.968, "step": 39369 }, { "epoch": 3.09, "learning_rate": 6.912883243158473e-05, "loss": 2.9556, "step": 39370 }, { "epoch": 3.09, "learning_rate": 6.912804830236023e-05, "loss": 2.8987, "step": 39371 }, { "epoch": 3.09, "learning_rate": 6.912726417313574e-05, "loss": 2.6794, "step": 39372 }, { "epoch": 3.09, "learning_rate": 6.912648004391124e-05, "loss": 2.5536, "step": 39373 }, { "epoch": 3.09, "learning_rate": 6.912569591468674e-05, "loss": 2.7007, "step": 39374 }, { "epoch": 3.09, "learning_rate": 6.912491178546225e-05, "loss": 2.6099, "step": 39375 }, { "epoch": 3.09, "learning_rate": 6.912412765623775e-05, "loss": 2.7264, "step": 39376 }, { "epoch": 3.09, "learning_rate": 6.912334352701325e-05, "loss": 2.7817, "step": 39377 }, { "epoch": 3.09, "learning_rate": 6.912255939778876e-05, "loss": 3.0295, "step": 39378 }, { "epoch": 3.09, "learning_rate": 6.912177526856426e-05, "loss": 2.7594, "step": 39379 }, { "epoch": 3.09, "learning_rate": 6.912099113933977e-05, "loss": 2.7285, "step": 39380 }, { "epoch": 3.09, "learning_rate": 6.912020701011527e-05, "loss": 2.8615, "step": 39381 }, { "epoch": 3.09, "learning_rate": 6.911942288089077e-05, "loss": 2.7098, "step": 39382 }, { "epoch": 3.09, "learning_rate": 6.911863875166628e-05, "loss": 2.7004, "step": 39383 }, { "epoch": 3.09, "learning_rate": 6.911785462244178e-05, "loss": 2.9697, "step": 39384 }, { "epoch": 3.09, "learning_rate": 6.911707049321728e-05, "loss": 2.9025, "step": 39385 }, { "epoch": 3.09, "learning_rate": 6.911628636399279e-05, "loss": 2.6944, "step": 39386 }, { "epoch": 3.09, "learning_rate": 6.911550223476829e-05, "loss": 2.8328, "step": 39387 }, { "epoch": 3.09, "learning_rate": 6.91147181055438e-05, "loss": 2.8032, "step": 39388 }, { "epoch": 3.09, "learning_rate": 6.91139339763193e-05, "loss": 2.7117, "step": 39389 }, { "epoch": 3.09, "learning_rate": 6.91131498470948e-05, "loss": 2.6494, "step": 39390 }, { "epoch": 3.09, "learning_rate": 6.91123657178703e-05, "loss": 2.7442, "step": 39391 }, { "epoch": 3.09, "learning_rate": 6.911158158864581e-05, "loss": 2.6821, "step": 39392 }, { "epoch": 3.09, "learning_rate": 6.911079745942131e-05, "loss": 2.7196, "step": 39393 }, { "epoch": 3.09, "learning_rate": 6.911001333019682e-05, "loss": 2.9416, "step": 39394 }, { "epoch": 3.09, "learning_rate": 6.910922920097232e-05, "loss": 2.6621, "step": 39395 }, { "epoch": 3.09, "learning_rate": 6.910844507174782e-05, "loss": 2.6948, "step": 39396 }, { "epoch": 3.09, "learning_rate": 6.910766094252333e-05, "loss": 2.6308, "step": 39397 }, { "epoch": 3.09, "learning_rate": 6.910687681329883e-05, "loss": 2.9707, "step": 39398 }, { "epoch": 3.09, "learning_rate": 6.910609268407434e-05, "loss": 2.7028, "step": 39399 }, { "epoch": 3.09, "learning_rate": 6.910530855484984e-05, "loss": 2.6957, "step": 39400 }, { "epoch": 3.09, "learning_rate": 6.910452442562536e-05, "loss": 2.643, "step": 39401 }, { "epoch": 3.09, "learning_rate": 6.910374029640086e-05, "loss": 2.9361, "step": 39402 }, { "epoch": 3.09, "learning_rate": 6.910295616717636e-05, "loss": 2.8062, "step": 39403 }, { "epoch": 3.09, "learning_rate": 6.910217203795187e-05, "loss": 2.6623, "step": 39404 }, { "epoch": 3.09, "learning_rate": 6.910138790872737e-05, "loss": 2.679, "step": 39405 }, { "epoch": 3.09, "learning_rate": 6.910060377950287e-05, "loss": 2.4742, "step": 39406 }, { "epoch": 3.09, "learning_rate": 6.909981965027838e-05, "loss": 2.7621, "step": 39407 }, { "epoch": 3.09, "learning_rate": 6.909903552105387e-05, "loss": 2.7376, "step": 39408 }, { "epoch": 3.09, "learning_rate": 6.909825139182937e-05, "loss": 2.4375, "step": 39409 }, { "epoch": 3.09, "learning_rate": 6.909746726260488e-05, "loss": 2.6812, "step": 39410 }, { "epoch": 3.09, "learning_rate": 6.909668313338038e-05, "loss": 2.7471, "step": 39411 }, { "epoch": 3.09, "learning_rate": 6.909589900415588e-05, "loss": 2.5265, "step": 39412 }, { "epoch": 3.09, "learning_rate": 6.909511487493139e-05, "loss": 3.0522, "step": 39413 }, { "epoch": 3.09, "learning_rate": 6.909433074570689e-05, "loss": 2.9638, "step": 39414 }, { "epoch": 3.09, "learning_rate": 6.90935466164824e-05, "loss": 2.687, "step": 39415 }, { "epoch": 3.09, "learning_rate": 6.90927624872579e-05, "loss": 2.65, "step": 39416 }, { "epoch": 3.09, "learning_rate": 6.90919783580334e-05, "loss": 2.7734, "step": 39417 }, { "epoch": 3.09, "learning_rate": 6.90911942288089e-05, "loss": 2.6952, "step": 39418 }, { "epoch": 3.09, "learning_rate": 6.909041009958441e-05, "loss": 2.7105, "step": 39419 }, { "epoch": 3.09, "learning_rate": 6.908962597035991e-05, "loss": 2.6556, "step": 39420 }, { "epoch": 3.09, "learning_rate": 6.908884184113542e-05, "loss": 2.6247, "step": 39421 }, { "epoch": 3.09, "learning_rate": 6.908805771191092e-05, "loss": 2.7385, "step": 39422 }, { "epoch": 3.09, "learning_rate": 6.908727358268642e-05, "loss": 2.876, "step": 39423 }, { "epoch": 3.09, "learning_rate": 6.908648945346193e-05, "loss": 2.6549, "step": 39424 }, { "epoch": 3.09, "learning_rate": 6.908570532423743e-05, "loss": 2.6728, "step": 39425 }, { "epoch": 3.09, "learning_rate": 6.908492119501293e-05, "loss": 2.8371, "step": 39426 }, { "epoch": 3.09, "learning_rate": 6.908413706578845e-05, "loss": 2.7363, "step": 39427 }, { "epoch": 3.09, "learning_rate": 6.908335293656396e-05, "loss": 2.665, "step": 39428 }, { "epoch": 3.09, "learning_rate": 6.908256880733946e-05, "loss": 2.6179, "step": 39429 }, { "epoch": 3.09, "learning_rate": 6.908178467811496e-05, "loss": 2.7903, "step": 39430 }, { "epoch": 3.09, "learning_rate": 6.908100054889047e-05, "loss": 2.7316, "step": 39431 }, { "epoch": 3.09, "learning_rate": 6.908021641966597e-05, "loss": 2.8803, "step": 39432 }, { "epoch": 3.09, "learning_rate": 6.907943229044147e-05, "loss": 2.8177, "step": 39433 }, { "epoch": 3.09, "learning_rate": 6.907864816121698e-05, "loss": 2.6092, "step": 39434 }, { "epoch": 3.09, "learning_rate": 6.907786403199248e-05, "loss": 2.9097, "step": 39435 }, { "epoch": 3.09, "learning_rate": 6.907707990276799e-05, "loss": 2.6072, "step": 39436 }, { "epoch": 3.09, "learning_rate": 6.907629577354349e-05, "loss": 2.9212, "step": 39437 }, { "epoch": 3.09, "learning_rate": 6.907551164431899e-05, "loss": 2.8938, "step": 39438 }, { "epoch": 3.09, "learning_rate": 6.90747275150945e-05, "loss": 2.6816, "step": 39439 }, { "epoch": 3.09, "learning_rate": 6.907394338587e-05, "loss": 2.8345, "step": 39440 }, { "epoch": 3.09, "learning_rate": 6.90731592566455e-05, "loss": 2.5934, "step": 39441 }, { "epoch": 3.09, "learning_rate": 6.907237512742101e-05, "loss": 2.7773, "step": 39442 }, { "epoch": 3.09, "learning_rate": 6.907159099819651e-05, "loss": 2.5711, "step": 39443 }, { "epoch": 3.09, "learning_rate": 6.9070806868972e-05, "loss": 2.705, "step": 39444 }, { "epoch": 3.09, "learning_rate": 6.90700227397475e-05, "loss": 2.8442, "step": 39445 }, { "epoch": 3.09, "learning_rate": 6.906923861052301e-05, "loss": 2.7736, "step": 39446 }, { "epoch": 3.09, "learning_rate": 6.906845448129851e-05, "loss": 2.6583, "step": 39447 }, { "epoch": 3.09, "learning_rate": 6.906767035207402e-05, "loss": 2.5506, "step": 39448 }, { "epoch": 3.09, "learning_rate": 6.906688622284952e-05, "loss": 2.554, "step": 39449 }, { "epoch": 3.09, "learning_rate": 6.906610209362502e-05, "loss": 2.6269, "step": 39450 }, { "epoch": 3.09, "learning_rate": 6.906531796440053e-05, "loss": 2.9641, "step": 39451 }, { "epoch": 3.09, "learning_rate": 6.906453383517604e-05, "loss": 2.6476, "step": 39452 }, { "epoch": 3.09, "learning_rate": 6.906374970595155e-05, "loss": 2.7261, "step": 39453 }, { "epoch": 3.09, "learning_rate": 6.906296557672705e-05, "loss": 2.5868, "step": 39454 }, { "epoch": 3.09, "learning_rate": 6.906218144750256e-05, "loss": 2.5076, "step": 39455 }, { "epoch": 3.09, "learning_rate": 6.906139731827806e-05, "loss": 2.5903, "step": 39456 }, { "epoch": 3.09, "learning_rate": 6.906061318905356e-05, "loss": 2.7709, "step": 39457 }, { "epoch": 3.09, "learning_rate": 6.905982905982907e-05, "loss": 2.9512, "step": 39458 }, { "epoch": 3.09, "learning_rate": 6.905904493060457e-05, "loss": 2.824, "step": 39459 }, { "epoch": 3.09, "learning_rate": 6.905826080138007e-05, "loss": 2.7635, "step": 39460 }, { "epoch": 3.09, "learning_rate": 6.905747667215558e-05, "loss": 2.6773, "step": 39461 }, { "epoch": 3.09, "learning_rate": 6.905669254293108e-05, "loss": 2.8658, "step": 39462 }, { "epoch": 3.09, "learning_rate": 6.905590841370658e-05, "loss": 2.7409, "step": 39463 }, { "epoch": 3.09, "learning_rate": 6.905512428448209e-05, "loss": 2.9765, "step": 39464 }, { "epoch": 3.09, "learning_rate": 6.905434015525759e-05, "loss": 2.6557, "step": 39465 }, { "epoch": 3.09, "learning_rate": 6.90535560260331e-05, "loss": 2.6576, "step": 39466 }, { "epoch": 3.09, "learning_rate": 6.90527718968086e-05, "loss": 2.8089, "step": 39467 }, { "epoch": 3.09, "learning_rate": 6.90519877675841e-05, "loss": 2.7667, "step": 39468 }, { "epoch": 3.09, "learning_rate": 6.90512036383596e-05, "loss": 2.6058, "step": 39469 }, { "epoch": 3.09, "learning_rate": 6.905041950913511e-05, "loss": 2.7416, "step": 39470 }, { "epoch": 3.1, "learning_rate": 6.904963537991061e-05, "loss": 2.6335, "step": 39471 }, { "epoch": 3.1, "learning_rate": 6.904885125068612e-05, "loss": 2.5248, "step": 39472 }, { "epoch": 3.1, "learning_rate": 6.904806712146162e-05, "loss": 2.6997, "step": 39473 }, { "epoch": 3.1, "learning_rate": 6.904728299223712e-05, "loss": 2.8077, "step": 39474 }, { "epoch": 3.1, "learning_rate": 6.904649886301263e-05, "loss": 2.7175, "step": 39475 }, { "epoch": 3.1, "learning_rate": 6.904571473378813e-05, "loss": 2.6552, "step": 39476 }, { "epoch": 3.1, "learning_rate": 6.904493060456364e-05, "loss": 2.9314, "step": 39477 }, { "epoch": 3.1, "learning_rate": 6.904414647533914e-05, "loss": 2.7439, "step": 39478 }, { "epoch": 3.1, "learning_rate": 6.904336234611464e-05, "loss": 2.8048, "step": 39479 }, { "epoch": 3.1, "learning_rate": 6.904257821689015e-05, "loss": 2.83, "step": 39480 }, { "epoch": 3.1, "learning_rate": 6.904179408766565e-05, "loss": 2.8026, "step": 39481 }, { "epoch": 3.1, "learning_rate": 6.904100995844115e-05, "loss": 2.7321, "step": 39482 }, { "epoch": 3.1, "learning_rate": 6.904022582921666e-05, "loss": 2.8748, "step": 39483 }, { "epoch": 3.1, "learning_rate": 6.903944169999216e-05, "loss": 2.7037, "step": 39484 }, { "epoch": 3.1, "learning_rate": 6.903865757076767e-05, "loss": 2.6778, "step": 39485 }, { "epoch": 3.1, "learning_rate": 6.903787344154317e-05, "loss": 2.5875, "step": 39486 }, { "epoch": 3.1, "learning_rate": 6.903708931231867e-05, "loss": 2.6939, "step": 39487 }, { "epoch": 3.1, "learning_rate": 6.903630518309418e-05, "loss": 2.7835, "step": 39488 }, { "epoch": 3.1, "learning_rate": 6.903552105386968e-05, "loss": 2.5984, "step": 39489 }, { "epoch": 3.1, "learning_rate": 6.903473692464518e-05, "loss": 2.6417, "step": 39490 }, { "epoch": 3.1, "learning_rate": 6.903395279542069e-05, "loss": 2.6642, "step": 39491 }, { "epoch": 3.1, "learning_rate": 6.903316866619619e-05, "loss": 2.8223, "step": 39492 }, { "epoch": 3.1, "learning_rate": 6.90323845369717e-05, "loss": 2.8212, "step": 39493 }, { "epoch": 3.1, "learning_rate": 6.90316004077472e-05, "loss": 2.8531, "step": 39494 }, { "epoch": 3.1, "learning_rate": 6.90308162785227e-05, "loss": 2.6251, "step": 39495 }, { "epoch": 3.1, "learning_rate": 6.90300321492982e-05, "loss": 2.6752, "step": 39496 }, { "epoch": 3.1, "learning_rate": 6.902924802007371e-05, "loss": 2.7453, "step": 39497 }, { "epoch": 3.1, "learning_rate": 6.902846389084921e-05, "loss": 2.557, "step": 39498 }, { "epoch": 3.1, "learning_rate": 6.902767976162472e-05, "loss": 2.6637, "step": 39499 }, { "epoch": 3.1, "learning_rate": 6.902689563240022e-05, "loss": 2.6665, "step": 39500 }, { "epoch": 3.1, "learning_rate": 6.902611150317572e-05, "loss": 2.8346, "step": 39501 }, { "epoch": 3.1, "learning_rate": 6.902532737395123e-05, "loss": 2.7157, "step": 39502 }, { "epoch": 3.1, "learning_rate": 6.902454324472673e-05, "loss": 2.7706, "step": 39503 }, { "epoch": 3.1, "learning_rate": 6.902375911550224e-05, "loss": 2.8414, "step": 39504 }, { "epoch": 3.1, "learning_rate": 6.902297498627774e-05, "loss": 2.6423, "step": 39505 }, { "epoch": 3.1, "learning_rate": 6.902219085705324e-05, "loss": 2.6509, "step": 39506 }, { "epoch": 3.1, "learning_rate": 6.902140672782875e-05, "loss": 2.6326, "step": 39507 }, { "epoch": 3.1, "learning_rate": 6.902062259860425e-05, "loss": 2.6077, "step": 39508 }, { "epoch": 3.1, "learning_rate": 6.901983846937977e-05, "loss": 2.925, "step": 39509 }, { "epoch": 3.1, "learning_rate": 6.901905434015527e-05, "loss": 2.8372, "step": 39510 }, { "epoch": 3.1, "learning_rate": 6.901827021093077e-05, "loss": 2.7261, "step": 39511 }, { "epoch": 3.1, "learning_rate": 6.901748608170628e-05, "loss": 2.7211, "step": 39512 }, { "epoch": 3.1, "learning_rate": 6.901670195248178e-05, "loss": 2.6925, "step": 39513 }, { "epoch": 3.1, "learning_rate": 6.901591782325729e-05, "loss": 2.749, "step": 39514 }, { "epoch": 3.1, "learning_rate": 6.901513369403279e-05, "loss": 2.6548, "step": 39515 }, { "epoch": 3.1, "learning_rate": 6.901434956480828e-05, "loss": 2.7052, "step": 39516 }, { "epoch": 3.1, "learning_rate": 6.901356543558378e-05, "loss": 2.6772, "step": 39517 }, { "epoch": 3.1, "learning_rate": 6.901278130635929e-05, "loss": 2.6763, "step": 39518 }, { "epoch": 3.1, "learning_rate": 6.901199717713479e-05, "loss": 2.7977, "step": 39519 }, { "epoch": 3.1, "learning_rate": 6.90112130479103e-05, "loss": 2.6391, "step": 39520 }, { "epoch": 3.1, "learning_rate": 6.90104289186858e-05, "loss": 2.6661, "step": 39521 }, { "epoch": 3.1, "learning_rate": 6.90096447894613e-05, "loss": 2.4842, "step": 39522 }, { "epoch": 3.1, "learning_rate": 6.90088606602368e-05, "loss": 2.7482, "step": 39523 }, { "epoch": 3.1, "learning_rate": 6.900807653101231e-05, "loss": 2.635, "step": 39524 }, { "epoch": 3.1, "learning_rate": 6.900729240178781e-05, "loss": 2.7961, "step": 39525 }, { "epoch": 3.1, "learning_rate": 6.900650827256332e-05, "loss": 2.6522, "step": 39526 }, { "epoch": 3.1, "learning_rate": 6.900572414333882e-05, "loss": 2.7322, "step": 39527 }, { "epoch": 3.1, "learning_rate": 6.900494001411432e-05, "loss": 2.8013, "step": 39528 }, { "epoch": 3.1, "learning_rate": 6.900415588488983e-05, "loss": 2.6565, "step": 39529 }, { "epoch": 3.1, "learning_rate": 6.900337175566533e-05, "loss": 2.6171, "step": 39530 }, { "epoch": 3.1, "learning_rate": 6.900258762644083e-05, "loss": 2.8489, "step": 39531 }, { "epoch": 3.1, "learning_rate": 6.900180349721634e-05, "loss": 2.8213, "step": 39532 }, { "epoch": 3.1, "learning_rate": 6.900101936799184e-05, "loss": 2.6543, "step": 39533 }, { "epoch": 3.1, "learning_rate": 6.900023523876735e-05, "loss": 2.7092, "step": 39534 }, { "epoch": 3.1, "learning_rate": 6.899945110954286e-05, "loss": 2.6077, "step": 39535 }, { "epoch": 3.1, "learning_rate": 6.899866698031837e-05, "loss": 2.81, "step": 39536 }, { "epoch": 3.1, "learning_rate": 6.899788285109387e-05, "loss": 2.7247, "step": 39537 }, { "epoch": 3.1, "learning_rate": 6.899709872186937e-05, "loss": 2.8711, "step": 39538 }, { "epoch": 3.1, "learning_rate": 6.899631459264488e-05, "loss": 2.9052, "step": 39539 }, { "epoch": 3.1, "learning_rate": 6.899553046342038e-05, "loss": 2.7419, "step": 39540 }, { "epoch": 3.1, "learning_rate": 6.899474633419588e-05, "loss": 2.6499, "step": 39541 }, { "epoch": 3.1, "learning_rate": 6.899396220497139e-05, "loss": 2.9226, "step": 39542 }, { "epoch": 3.1, "learning_rate": 6.899317807574689e-05, "loss": 2.7467, "step": 39543 }, { "epoch": 3.1, "learning_rate": 6.89923939465224e-05, "loss": 2.7304, "step": 39544 }, { "epoch": 3.1, "learning_rate": 6.89916098172979e-05, "loss": 2.7672, "step": 39545 }, { "epoch": 3.1, "learning_rate": 6.89908256880734e-05, "loss": 2.6609, "step": 39546 }, { "epoch": 3.1, "learning_rate": 6.899004155884891e-05, "loss": 2.6207, "step": 39547 }, { "epoch": 3.1, "learning_rate": 6.898925742962441e-05, "loss": 2.7404, "step": 39548 }, { "epoch": 3.1, "learning_rate": 6.898847330039991e-05, "loss": 2.6408, "step": 39549 }, { "epoch": 3.1, "learning_rate": 6.898768917117542e-05, "loss": 2.7255, "step": 39550 }, { "epoch": 3.1, "learning_rate": 6.898690504195092e-05, "loss": 2.7284, "step": 39551 }, { "epoch": 3.1, "learning_rate": 6.898612091272641e-05, "loss": 2.8432, "step": 39552 }, { "epoch": 3.1, "learning_rate": 6.898533678350192e-05, "loss": 2.6234, "step": 39553 }, { "epoch": 3.1, "learning_rate": 6.898455265427742e-05, "loss": 2.5946, "step": 39554 }, { "epoch": 3.1, "learning_rate": 6.898376852505292e-05, "loss": 2.7337, "step": 39555 }, { "epoch": 3.1, "learning_rate": 6.898298439582843e-05, "loss": 2.8395, "step": 39556 }, { "epoch": 3.1, "learning_rate": 6.898220026660393e-05, "loss": 2.5638, "step": 39557 }, { "epoch": 3.1, "learning_rate": 6.898141613737943e-05, "loss": 2.6082, "step": 39558 }, { "epoch": 3.1, "learning_rate": 6.898063200815494e-05, "loss": 2.4578, "step": 39559 }, { "epoch": 3.1, "learning_rate": 6.897984787893045e-05, "loss": 2.926, "step": 39560 }, { "epoch": 3.1, "learning_rate": 6.897906374970596e-05, "loss": 2.7301, "step": 39561 }, { "epoch": 3.1, "learning_rate": 6.897827962048146e-05, "loss": 2.9682, "step": 39562 }, { "epoch": 3.1, "learning_rate": 6.897749549125697e-05, "loss": 2.7062, "step": 39563 }, { "epoch": 3.1, "learning_rate": 6.897671136203247e-05, "loss": 2.5975, "step": 39564 }, { "epoch": 3.1, "learning_rate": 6.897592723280797e-05, "loss": 2.5988, "step": 39565 }, { "epoch": 3.1, "learning_rate": 6.897514310358348e-05, "loss": 2.8001, "step": 39566 }, { "epoch": 3.1, "learning_rate": 6.897435897435898e-05, "loss": 2.6549, "step": 39567 }, { "epoch": 3.1, "learning_rate": 6.897357484513448e-05, "loss": 2.6961, "step": 39568 }, { "epoch": 3.1, "learning_rate": 6.897279071590999e-05, "loss": 2.7617, "step": 39569 }, { "epoch": 3.1, "learning_rate": 6.897200658668549e-05, "loss": 2.7176, "step": 39570 }, { "epoch": 3.1, "learning_rate": 6.8971222457461e-05, "loss": 2.7608, "step": 39571 }, { "epoch": 3.1, "learning_rate": 6.89704383282365e-05, "loss": 2.7693, "step": 39572 }, { "epoch": 3.1, "learning_rate": 6.8969654199012e-05, "loss": 2.6723, "step": 39573 }, { "epoch": 3.1, "learning_rate": 6.89688700697875e-05, "loss": 2.6077, "step": 39574 }, { "epoch": 3.1, "learning_rate": 6.896808594056301e-05, "loss": 2.6904, "step": 39575 }, { "epoch": 3.1, "learning_rate": 6.896730181133851e-05, "loss": 2.5534, "step": 39576 }, { "epoch": 3.1, "learning_rate": 6.896651768211402e-05, "loss": 2.746, "step": 39577 }, { "epoch": 3.1, "learning_rate": 6.896573355288952e-05, "loss": 2.7156, "step": 39578 }, { "epoch": 3.1, "learning_rate": 6.896494942366502e-05, "loss": 2.7608, "step": 39579 }, { "epoch": 3.1, "learning_rate": 6.896416529444053e-05, "loss": 2.5775, "step": 39580 }, { "epoch": 3.1, "learning_rate": 6.896338116521603e-05, "loss": 2.9774, "step": 39581 }, { "epoch": 3.1, "learning_rate": 6.896259703599154e-05, "loss": 2.8292, "step": 39582 }, { "epoch": 3.1, "learning_rate": 6.896181290676704e-05, "loss": 2.5673, "step": 39583 }, { "epoch": 3.1, "learning_rate": 6.896102877754254e-05, "loss": 2.7212, "step": 39584 }, { "epoch": 3.1, "learning_rate": 6.896024464831805e-05, "loss": 3.0626, "step": 39585 }, { "epoch": 3.1, "learning_rate": 6.895946051909355e-05, "loss": 2.7876, "step": 39586 }, { "epoch": 3.1, "learning_rate": 6.895867638986905e-05, "loss": 2.8104, "step": 39587 }, { "epoch": 3.1, "learning_rate": 6.895789226064456e-05, "loss": 2.7942, "step": 39588 }, { "epoch": 3.1, "learning_rate": 6.895710813142006e-05, "loss": 2.7863, "step": 39589 }, { "epoch": 3.1, "learning_rate": 6.895632400219557e-05, "loss": 2.6782, "step": 39590 }, { "epoch": 3.1, "learning_rate": 6.895553987297107e-05, "loss": 2.6968, "step": 39591 }, { "epoch": 3.1, "learning_rate": 6.895475574374657e-05, "loss": 2.5683, "step": 39592 }, { "epoch": 3.1, "learning_rate": 6.895397161452208e-05, "loss": 2.6543, "step": 39593 }, { "epoch": 3.1, "learning_rate": 6.895318748529758e-05, "loss": 2.5998, "step": 39594 }, { "epoch": 3.1, "learning_rate": 6.895240335607308e-05, "loss": 2.6867, "step": 39595 }, { "epoch": 3.1, "learning_rate": 6.895161922684859e-05, "loss": 2.4956, "step": 39596 }, { "epoch": 3.1, "learning_rate": 6.895083509762409e-05, "loss": 2.7315, "step": 39597 }, { "epoch": 3.1, "learning_rate": 6.89500509683996e-05, "loss": 2.5978, "step": 39598 }, { "epoch": 3.11, "learning_rate": 6.89492668391751e-05, "loss": 2.8282, "step": 39599 }, { "epoch": 3.11, "learning_rate": 6.89484827099506e-05, "loss": 2.6559, "step": 39600 }, { "epoch": 3.11, "learning_rate": 6.89476985807261e-05, "loss": 2.5976, "step": 39601 }, { "epoch": 3.11, "learning_rate": 6.894691445150161e-05, "loss": 2.5899, "step": 39602 }, { "epoch": 3.11, "learning_rate": 6.894613032227711e-05, "loss": 2.593, "step": 39603 }, { "epoch": 3.11, "learning_rate": 6.894534619305262e-05, "loss": 3.0517, "step": 39604 }, { "epoch": 3.11, "learning_rate": 6.894456206382812e-05, "loss": 2.74, "step": 39605 }, { "epoch": 3.11, "learning_rate": 6.894377793460362e-05, "loss": 2.7529, "step": 39606 }, { "epoch": 3.11, "learning_rate": 6.894299380537913e-05, "loss": 2.9334, "step": 39607 }, { "epoch": 3.11, "learning_rate": 6.894220967615463e-05, "loss": 2.5964, "step": 39608 }, { "epoch": 3.11, "learning_rate": 6.894142554693013e-05, "loss": 2.4956, "step": 39609 }, { "epoch": 3.11, "learning_rate": 6.894064141770564e-05, "loss": 2.8443, "step": 39610 }, { "epoch": 3.11, "learning_rate": 6.893985728848114e-05, "loss": 2.7057, "step": 39611 }, { "epoch": 3.11, "learning_rate": 6.893907315925665e-05, "loss": 2.8416, "step": 39612 }, { "epoch": 3.11, "learning_rate": 6.893828903003215e-05, "loss": 2.6135, "step": 39613 }, { "epoch": 3.11, "learning_rate": 6.893750490080765e-05, "loss": 2.6338, "step": 39614 }, { "epoch": 3.11, "learning_rate": 6.893672077158316e-05, "loss": 2.8101, "step": 39615 }, { "epoch": 3.11, "learning_rate": 6.893593664235866e-05, "loss": 2.8077, "step": 39616 }, { "epoch": 3.11, "learning_rate": 6.893515251313418e-05, "loss": 2.6841, "step": 39617 }, { "epoch": 3.11, "learning_rate": 6.893436838390968e-05, "loss": 2.6914, "step": 39618 }, { "epoch": 3.11, "learning_rate": 6.893358425468519e-05, "loss": 2.7919, "step": 39619 }, { "epoch": 3.11, "learning_rate": 6.893280012546069e-05, "loss": 2.6524, "step": 39620 }, { "epoch": 3.11, "learning_rate": 6.893201599623619e-05, "loss": 2.5919, "step": 39621 }, { "epoch": 3.11, "learning_rate": 6.89312318670117e-05, "loss": 2.8691, "step": 39622 }, { "epoch": 3.11, "learning_rate": 6.89304477377872e-05, "loss": 2.6879, "step": 39623 }, { "epoch": 3.11, "learning_rate": 6.892966360856269e-05, "loss": 2.5025, "step": 39624 }, { "epoch": 3.11, "learning_rate": 6.89288794793382e-05, "loss": 2.9412, "step": 39625 }, { "epoch": 3.11, "learning_rate": 6.89280953501137e-05, "loss": 2.6339, "step": 39626 }, { "epoch": 3.11, "learning_rate": 6.89273112208892e-05, "loss": 2.3794, "step": 39627 }, { "epoch": 3.11, "learning_rate": 6.89265270916647e-05, "loss": 2.5914, "step": 39628 }, { "epoch": 3.11, "learning_rate": 6.892574296244021e-05, "loss": 2.5161, "step": 39629 }, { "epoch": 3.11, "learning_rate": 6.892495883321571e-05, "loss": 3.0328, "step": 39630 }, { "epoch": 3.11, "learning_rate": 6.892417470399122e-05, "loss": 2.7241, "step": 39631 }, { "epoch": 3.11, "learning_rate": 6.892339057476672e-05, "loss": 2.874, "step": 39632 }, { "epoch": 3.11, "learning_rate": 6.892260644554222e-05, "loss": 2.5935, "step": 39633 }, { "epoch": 3.11, "learning_rate": 6.892182231631773e-05, "loss": 2.764, "step": 39634 }, { "epoch": 3.11, "learning_rate": 6.892103818709323e-05, "loss": 2.6826, "step": 39635 }, { "epoch": 3.11, "learning_rate": 6.892025405786873e-05, "loss": 2.5159, "step": 39636 }, { "epoch": 3.11, "learning_rate": 6.891946992864424e-05, "loss": 2.7186, "step": 39637 }, { "epoch": 3.11, "learning_rate": 6.891868579941974e-05, "loss": 2.833, "step": 39638 }, { "epoch": 3.11, "learning_rate": 6.891790167019525e-05, "loss": 2.5188, "step": 39639 }, { "epoch": 3.11, "learning_rate": 6.891711754097075e-05, "loss": 2.8795, "step": 39640 }, { "epoch": 3.11, "learning_rate": 6.891633341174625e-05, "loss": 2.6209, "step": 39641 }, { "epoch": 3.11, "learning_rate": 6.891554928252176e-05, "loss": 2.781, "step": 39642 }, { "epoch": 3.11, "learning_rate": 6.891476515329727e-05, "loss": 3.2872, "step": 39643 }, { "epoch": 3.11, "learning_rate": 6.891398102407278e-05, "loss": 2.6095, "step": 39644 }, { "epoch": 3.11, "learning_rate": 6.891319689484828e-05, "loss": 2.8813, "step": 39645 }, { "epoch": 3.11, "learning_rate": 6.891241276562378e-05, "loss": 2.9041, "step": 39646 }, { "epoch": 3.11, "learning_rate": 6.891162863639929e-05, "loss": 2.6057, "step": 39647 }, { "epoch": 3.11, "learning_rate": 6.891084450717479e-05, "loss": 2.7551, "step": 39648 }, { "epoch": 3.11, "learning_rate": 6.89100603779503e-05, "loss": 2.7666, "step": 39649 }, { "epoch": 3.11, "learning_rate": 6.89092762487258e-05, "loss": 2.8259, "step": 39650 }, { "epoch": 3.11, "learning_rate": 6.89084921195013e-05, "loss": 2.7604, "step": 39651 }, { "epoch": 3.11, "learning_rate": 6.89077079902768e-05, "loss": 2.6912, "step": 39652 }, { "epoch": 3.11, "learning_rate": 6.890692386105231e-05, "loss": 2.7632, "step": 39653 }, { "epoch": 3.11, "learning_rate": 6.890613973182781e-05, "loss": 2.6208, "step": 39654 }, { "epoch": 3.11, "learning_rate": 6.890535560260332e-05, "loss": 2.9551, "step": 39655 }, { "epoch": 3.11, "learning_rate": 6.890457147337882e-05, "loss": 2.5154, "step": 39656 }, { "epoch": 3.11, "learning_rate": 6.890378734415433e-05, "loss": 2.6489, "step": 39657 }, { "epoch": 3.11, "learning_rate": 6.890300321492983e-05, "loss": 2.7499, "step": 39658 }, { "epoch": 3.11, "learning_rate": 6.890221908570533e-05, "loss": 2.5464, "step": 39659 }, { "epoch": 3.11, "learning_rate": 6.890143495648082e-05, "loss": 2.6827, "step": 39660 }, { "epoch": 3.11, "learning_rate": 6.890065082725633e-05, "loss": 2.7132, "step": 39661 }, { "epoch": 3.11, "learning_rate": 6.889986669803183e-05, "loss": 2.6789, "step": 39662 }, { "epoch": 3.11, "learning_rate": 6.889908256880733e-05, "loss": 2.5989, "step": 39663 }, { "epoch": 3.11, "learning_rate": 6.889829843958284e-05, "loss": 2.6717, "step": 39664 }, { "epoch": 3.11, "learning_rate": 6.889751431035834e-05, "loss": 2.85, "step": 39665 }, { "epoch": 3.11, "learning_rate": 6.889673018113384e-05, "loss": 2.8917, "step": 39666 }, { "epoch": 3.11, "learning_rate": 6.889594605190935e-05, "loss": 2.6965, "step": 39667 }, { "epoch": 3.11, "learning_rate": 6.889516192268487e-05, "loss": 2.8973, "step": 39668 }, { "epoch": 3.11, "learning_rate": 6.889437779346037e-05, "loss": 2.8938, "step": 39669 }, { "epoch": 3.11, "learning_rate": 6.889359366423587e-05, "loss": 2.8384, "step": 39670 }, { "epoch": 3.11, "learning_rate": 6.889280953501138e-05, "loss": 2.5782, "step": 39671 }, { "epoch": 3.11, "learning_rate": 6.889202540578688e-05, "loss": 2.6629, "step": 39672 }, { "epoch": 3.11, "learning_rate": 6.889124127656238e-05, "loss": 2.7103, "step": 39673 }, { "epoch": 3.11, "learning_rate": 6.889045714733789e-05, "loss": 2.6041, "step": 39674 }, { "epoch": 3.11, "learning_rate": 6.888967301811339e-05, "loss": 2.7682, "step": 39675 }, { "epoch": 3.11, "learning_rate": 6.88888888888889e-05, "loss": 2.4404, "step": 39676 }, { "epoch": 3.11, "learning_rate": 6.88881047596644e-05, "loss": 3.1441, "step": 39677 }, { "epoch": 3.11, "learning_rate": 6.88873206304399e-05, "loss": 2.5901, "step": 39678 }, { "epoch": 3.11, "learning_rate": 6.88865365012154e-05, "loss": 2.843, "step": 39679 }, { "epoch": 3.11, "learning_rate": 6.888575237199091e-05, "loss": 2.6128, "step": 39680 }, { "epoch": 3.11, "learning_rate": 6.888496824276641e-05, "loss": 2.5979, "step": 39681 }, { "epoch": 3.11, "learning_rate": 6.888418411354192e-05, "loss": 2.7901, "step": 39682 }, { "epoch": 3.11, "learning_rate": 6.888339998431742e-05, "loss": 2.825, "step": 39683 }, { "epoch": 3.11, "learning_rate": 6.888261585509292e-05, "loss": 2.7646, "step": 39684 }, { "epoch": 3.11, "learning_rate": 6.888183172586843e-05, "loss": 2.7948, "step": 39685 }, { "epoch": 3.11, "learning_rate": 6.888104759664393e-05, "loss": 2.69, "step": 39686 }, { "epoch": 3.11, "learning_rate": 6.888026346741944e-05, "loss": 2.817, "step": 39687 }, { "epoch": 3.11, "learning_rate": 6.887947933819494e-05, "loss": 2.5598, "step": 39688 }, { "epoch": 3.11, "learning_rate": 6.887869520897044e-05, "loss": 2.8005, "step": 39689 }, { "epoch": 3.11, "learning_rate": 6.887791107974595e-05, "loss": 2.6996, "step": 39690 }, { "epoch": 3.11, "learning_rate": 6.887712695052145e-05, "loss": 2.5988, "step": 39691 }, { "epoch": 3.11, "learning_rate": 6.887634282129695e-05, "loss": 2.7697, "step": 39692 }, { "epoch": 3.11, "learning_rate": 6.887555869207246e-05, "loss": 2.6317, "step": 39693 }, { "epoch": 3.11, "learning_rate": 6.887477456284796e-05, "loss": 2.6973, "step": 39694 }, { "epoch": 3.11, "learning_rate": 6.887399043362346e-05, "loss": 2.7875, "step": 39695 }, { "epoch": 3.11, "learning_rate": 6.887320630439897e-05, "loss": 2.6567, "step": 39696 }, { "epoch": 3.11, "learning_rate": 6.887242217517447e-05, "loss": 2.7692, "step": 39697 }, { "epoch": 3.11, "learning_rate": 6.887163804594998e-05, "loss": 2.7795, "step": 39698 }, { "epoch": 3.11, "learning_rate": 6.887085391672548e-05, "loss": 2.7999, "step": 39699 }, { "epoch": 3.11, "learning_rate": 6.887006978750098e-05, "loss": 2.7249, "step": 39700 }, { "epoch": 3.11, "learning_rate": 6.886928565827649e-05, "loss": 3.0545, "step": 39701 }, { "epoch": 3.11, "learning_rate": 6.886850152905199e-05, "loss": 2.7396, "step": 39702 }, { "epoch": 3.11, "learning_rate": 6.88677173998275e-05, "loss": 2.7061, "step": 39703 }, { "epoch": 3.11, "learning_rate": 6.8866933270603e-05, "loss": 2.6145, "step": 39704 }, { "epoch": 3.11, "learning_rate": 6.88661491413785e-05, "loss": 2.7841, "step": 39705 }, { "epoch": 3.11, "learning_rate": 6.8865365012154e-05, "loss": 2.8533, "step": 39706 }, { "epoch": 3.11, "learning_rate": 6.886458088292951e-05, "loss": 2.8052, "step": 39707 }, { "epoch": 3.11, "learning_rate": 6.886379675370501e-05, "loss": 2.6925, "step": 39708 }, { "epoch": 3.11, "learning_rate": 6.886301262448052e-05, "loss": 2.6331, "step": 39709 }, { "epoch": 3.11, "learning_rate": 6.886222849525602e-05, "loss": 2.7816, "step": 39710 }, { "epoch": 3.11, "learning_rate": 6.886144436603152e-05, "loss": 2.6946, "step": 39711 }, { "epoch": 3.11, "learning_rate": 6.886066023680703e-05, "loss": 2.8159, "step": 39712 }, { "epoch": 3.11, "learning_rate": 6.885987610758253e-05, "loss": 2.6353, "step": 39713 }, { "epoch": 3.11, "learning_rate": 6.885909197835803e-05, "loss": 2.6113, "step": 39714 }, { "epoch": 3.11, "learning_rate": 6.885830784913354e-05, "loss": 2.6032, "step": 39715 }, { "epoch": 3.11, "learning_rate": 6.885752371990904e-05, "loss": 2.7737, "step": 39716 }, { "epoch": 3.11, "learning_rate": 6.885673959068455e-05, "loss": 2.731, "step": 39717 }, { "epoch": 3.11, "learning_rate": 6.885595546146005e-05, "loss": 2.7761, "step": 39718 }, { "epoch": 3.11, "learning_rate": 6.885517133223555e-05, "loss": 2.7157, "step": 39719 }, { "epoch": 3.11, "learning_rate": 6.885438720301106e-05, "loss": 2.6344, "step": 39720 }, { "epoch": 3.11, "learning_rate": 6.885360307378656e-05, "loss": 2.5543, "step": 39721 }, { "epoch": 3.11, "learning_rate": 6.885281894456206e-05, "loss": 2.6734, "step": 39722 }, { "epoch": 3.11, "learning_rate": 6.885203481533757e-05, "loss": 2.6945, "step": 39723 }, { "epoch": 3.11, "learning_rate": 6.885125068611307e-05, "loss": 2.8738, "step": 39724 }, { "epoch": 3.11, "learning_rate": 6.885046655688859e-05, "loss": 2.8629, "step": 39725 }, { "epoch": 3.12, "learning_rate": 6.884968242766409e-05, "loss": 2.6858, "step": 39726 }, { "epoch": 3.12, "learning_rate": 6.88488982984396e-05, "loss": 2.5185, "step": 39727 }, { "epoch": 3.12, "learning_rate": 6.88481141692151e-05, "loss": 2.8336, "step": 39728 }, { "epoch": 3.12, "learning_rate": 6.88473300399906e-05, "loss": 2.7674, "step": 39729 }, { "epoch": 3.12, "learning_rate": 6.884654591076611e-05, "loss": 2.6556, "step": 39730 }, { "epoch": 3.12, "learning_rate": 6.884576178154161e-05, "loss": 2.7338, "step": 39731 }, { "epoch": 3.12, "learning_rate": 6.88449776523171e-05, "loss": 2.6128, "step": 39732 }, { "epoch": 3.12, "learning_rate": 6.88441935230926e-05, "loss": 2.5436, "step": 39733 }, { "epoch": 3.12, "learning_rate": 6.884340939386811e-05, "loss": 2.7071, "step": 39734 }, { "epoch": 3.12, "learning_rate": 6.884262526464361e-05, "loss": 3.1563, "step": 39735 }, { "epoch": 3.12, "learning_rate": 6.884184113541912e-05, "loss": 2.6279, "step": 39736 }, { "epoch": 3.12, "learning_rate": 6.884105700619462e-05, "loss": 2.7665, "step": 39737 }, { "epoch": 3.12, "learning_rate": 6.884027287697012e-05, "loss": 2.8674, "step": 39738 }, { "epoch": 3.12, "learning_rate": 6.883948874774563e-05, "loss": 2.5303, "step": 39739 }, { "epoch": 3.12, "learning_rate": 6.883870461852113e-05, "loss": 2.8527, "step": 39740 }, { "epoch": 3.12, "learning_rate": 6.883792048929663e-05, "loss": 2.6915, "step": 39741 }, { "epoch": 3.12, "learning_rate": 6.883713636007214e-05, "loss": 2.6524, "step": 39742 }, { "epoch": 3.12, "learning_rate": 6.883635223084764e-05, "loss": 2.672, "step": 39743 }, { "epoch": 3.12, "learning_rate": 6.883556810162314e-05, "loss": 2.5584, "step": 39744 }, { "epoch": 3.12, "learning_rate": 6.883478397239865e-05, "loss": 2.5928, "step": 39745 }, { "epoch": 3.12, "learning_rate": 6.883399984317415e-05, "loss": 2.9253, "step": 39746 }, { "epoch": 3.12, "learning_rate": 6.883321571394966e-05, "loss": 2.6762, "step": 39747 }, { "epoch": 3.12, "learning_rate": 6.883243158472516e-05, "loss": 2.6618, "step": 39748 }, { "epoch": 3.12, "learning_rate": 6.883164745550066e-05, "loss": 2.7175, "step": 39749 }, { "epoch": 3.12, "learning_rate": 6.883086332627617e-05, "loss": 2.6969, "step": 39750 }, { "epoch": 3.12, "learning_rate": 6.883007919705168e-05, "loss": 2.6362, "step": 39751 }, { "epoch": 3.12, "learning_rate": 6.882929506782719e-05, "loss": 2.6491, "step": 39752 }, { "epoch": 3.12, "learning_rate": 6.882851093860269e-05, "loss": 2.5846, "step": 39753 }, { "epoch": 3.12, "learning_rate": 6.88277268093782e-05, "loss": 2.6864, "step": 39754 }, { "epoch": 3.12, "learning_rate": 6.88269426801537e-05, "loss": 2.6842, "step": 39755 }, { "epoch": 3.12, "learning_rate": 6.88261585509292e-05, "loss": 2.6028, "step": 39756 }, { "epoch": 3.12, "learning_rate": 6.88253744217047e-05, "loss": 2.7462, "step": 39757 }, { "epoch": 3.12, "learning_rate": 6.882459029248021e-05, "loss": 2.7292, "step": 39758 }, { "epoch": 3.12, "learning_rate": 6.882380616325571e-05, "loss": 2.4765, "step": 39759 }, { "epoch": 3.12, "learning_rate": 6.882302203403122e-05, "loss": 2.6236, "step": 39760 }, { "epoch": 3.12, "learning_rate": 6.882223790480672e-05, "loss": 2.873, "step": 39761 }, { "epoch": 3.12, "learning_rate": 6.882145377558222e-05, "loss": 2.7047, "step": 39762 }, { "epoch": 3.12, "learning_rate": 6.882066964635773e-05, "loss": 2.6303, "step": 39763 }, { "epoch": 3.12, "learning_rate": 6.881988551713323e-05, "loss": 2.5972, "step": 39764 }, { "epoch": 3.12, "learning_rate": 6.881910138790874e-05, "loss": 2.741, "step": 39765 }, { "epoch": 3.12, "learning_rate": 6.881831725868424e-05, "loss": 2.7431, "step": 39766 }, { "epoch": 3.12, "learning_rate": 6.881753312945974e-05, "loss": 2.8179, "step": 39767 }, { "epoch": 3.12, "learning_rate": 6.881674900023523e-05, "loss": 2.8921, "step": 39768 }, { "epoch": 3.12, "learning_rate": 6.881596487101074e-05, "loss": 2.7918, "step": 39769 }, { "epoch": 3.12, "learning_rate": 6.881518074178624e-05, "loss": 2.8331, "step": 39770 }, { "epoch": 3.12, "learning_rate": 6.881439661256174e-05, "loss": 2.6648, "step": 39771 }, { "epoch": 3.12, "learning_rate": 6.881361248333725e-05, "loss": 3.0702, "step": 39772 }, { "epoch": 3.12, "learning_rate": 6.881282835411275e-05, "loss": 2.5883, "step": 39773 }, { "epoch": 3.12, "learning_rate": 6.881204422488826e-05, "loss": 2.8253, "step": 39774 }, { "epoch": 3.12, "learning_rate": 6.881126009566376e-05, "loss": 2.8205, "step": 39775 }, { "epoch": 3.12, "learning_rate": 6.881047596643928e-05, "loss": 2.5798, "step": 39776 }, { "epoch": 3.12, "learning_rate": 6.880969183721478e-05, "loss": 2.846, "step": 39777 }, { "epoch": 3.12, "learning_rate": 6.880890770799028e-05, "loss": 2.6399, "step": 39778 }, { "epoch": 3.12, "learning_rate": 6.880812357876579e-05, "loss": 2.7053, "step": 39779 }, { "epoch": 3.12, "learning_rate": 6.880733944954129e-05, "loss": 2.6734, "step": 39780 }, { "epoch": 3.12, "learning_rate": 6.88065553203168e-05, "loss": 2.8086, "step": 39781 }, { "epoch": 3.12, "learning_rate": 6.88057711910923e-05, "loss": 2.526, "step": 39782 }, { "epoch": 3.12, "learning_rate": 6.88049870618678e-05, "loss": 2.7409, "step": 39783 }, { "epoch": 3.12, "learning_rate": 6.88042029326433e-05, "loss": 2.4331, "step": 39784 }, { "epoch": 3.12, "learning_rate": 6.880341880341881e-05, "loss": 2.7869, "step": 39785 }, { "epoch": 3.12, "learning_rate": 6.880263467419431e-05, "loss": 2.4639, "step": 39786 }, { "epoch": 3.12, "learning_rate": 6.880185054496982e-05, "loss": 2.6291, "step": 39787 }, { "epoch": 3.12, "learning_rate": 6.880106641574532e-05, "loss": 2.6695, "step": 39788 }, { "epoch": 3.12, "learning_rate": 6.880028228652082e-05, "loss": 2.7769, "step": 39789 }, { "epoch": 3.12, "learning_rate": 6.879949815729633e-05, "loss": 2.4869, "step": 39790 }, { "epoch": 3.12, "learning_rate": 6.879871402807183e-05, "loss": 2.6024, "step": 39791 }, { "epoch": 3.12, "learning_rate": 6.879792989884734e-05, "loss": 2.5912, "step": 39792 }, { "epoch": 3.12, "learning_rate": 6.879714576962284e-05, "loss": 2.8135, "step": 39793 }, { "epoch": 3.12, "learning_rate": 6.879636164039834e-05, "loss": 2.8707, "step": 39794 }, { "epoch": 3.12, "learning_rate": 6.879557751117385e-05, "loss": 2.6592, "step": 39795 }, { "epoch": 3.12, "learning_rate": 6.879479338194935e-05, "loss": 2.6603, "step": 39796 }, { "epoch": 3.12, "learning_rate": 6.879400925272485e-05, "loss": 2.7111, "step": 39797 }, { "epoch": 3.12, "learning_rate": 6.879322512350036e-05, "loss": 2.7918, "step": 39798 }, { "epoch": 3.12, "learning_rate": 6.879244099427586e-05, "loss": 2.9301, "step": 39799 }, { "epoch": 3.12, "learning_rate": 6.879165686505136e-05, "loss": 2.6952, "step": 39800 }, { "epoch": 3.12, "learning_rate": 6.879087273582687e-05, "loss": 2.7569, "step": 39801 }, { "epoch": 3.12, "learning_rate": 6.879008860660237e-05, "loss": 2.6707, "step": 39802 }, { "epoch": 3.12, "learning_rate": 6.878930447737788e-05, "loss": 2.7681, "step": 39803 }, { "epoch": 3.12, "learning_rate": 6.878852034815338e-05, "loss": 2.7062, "step": 39804 }, { "epoch": 3.12, "learning_rate": 6.878773621892888e-05, "loss": 2.5477, "step": 39805 }, { "epoch": 3.12, "learning_rate": 6.878695208970439e-05, "loss": 2.7739, "step": 39806 }, { "epoch": 3.12, "learning_rate": 6.878616796047989e-05, "loss": 2.6735, "step": 39807 }, { "epoch": 3.12, "learning_rate": 6.87853838312554e-05, "loss": 2.5979, "step": 39808 }, { "epoch": 3.12, "learning_rate": 6.87845997020309e-05, "loss": 2.6587, "step": 39809 }, { "epoch": 3.12, "learning_rate": 6.87838155728064e-05, "loss": 2.8598, "step": 39810 }, { "epoch": 3.12, "learning_rate": 6.87830314435819e-05, "loss": 2.6973, "step": 39811 }, { "epoch": 3.12, "learning_rate": 6.878224731435741e-05, "loss": 2.783, "step": 39812 }, { "epoch": 3.12, "learning_rate": 6.878146318513291e-05, "loss": 2.6227, "step": 39813 }, { "epoch": 3.12, "learning_rate": 6.878067905590842e-05, "loss": 2.7657, "step": 39814 }, { "epoch": 3.12, "learning_rate": 6.877989492668392e-05, "loss": 2.8043, "step": 39815 }, { "epoch": 3.12, "learning_rate": 6.877911079745942e-05, "loss": 2.6604, "step": 39816 }, { "epoch": 3.12, "learning_rate": 6.877832666823493e-05, "loss": 2.749, "step": 39817 }, { "epoch": 3.12, "learning_rate": 6.877754253901043e-05, "loss": 2.8448, "step": 39818 }, { "epoch": 3.12, "learning_rate": 6.877675840978593e-05, "loss": 2.7497, "step": 39819 }, { "epoch": 3.12, "learning_rate": 6.877597428056144e-05, "loss": 2.7146, "step": 39820 }, { "epoch": 3.12, "learning_rate": 6.877519015133694e-05, "loss": 2.6886, "step": 39821 }, { "epoch": 3.12, "learning_rate": 6.877440602211245e-05, "loss": 2.7209, "step": 39822 }, { "epoch": 3.12, "learning_rate": 6.877362189288795e-05, "loss": 2.7032, "step": 39823 }, { "epoch": 3.12, "learning_rate": 6.877283776366345e-05, "loss": 2.8477, "step": 39824 }, { "epoch": 3.12, "learning_rate": 6.877205363443896e-05, "loss": 2.7121, "step": 39825 }, { "epoch": 3.12, "learning_rate": 6.877126950521446e-05, "loss": 2.8378, "step": 39826 }, { "epoch": 3.12, "learning_rate": 6.877048537598996e-05, "loss": 2.8968, "step": 39827 }, { "epoch": 3.12, "learning_rate": 6.876970124676547e-05, "loss": 2.8096, "step": 39828 }, { "epoch": 3.12, "learning_rate": 6.876891711754097e-05, "loss": 2.753, "step": 39829 }, { "epoch": 3.12, "learning_rate": 6.876813298831647e-05, "loss": 2.6577, "step": 39830 }, { "epoch": 3.12, "learning_rate": 6.876734885909198e-05, "loss": 2.8385, "step": 39831 }, { "epoch": 3.12, "learning_rate": 6.876656472986748e-05, "loss": 2.9555, "step": 39832 }, { "epoch": 3.12, "learning_rate": 6.8765780600643e-05, "loss": 2.6296, "step": 39833 }, { "epoch": 3.12, "learning_rate": 6.87649964714185e-05, "loss": 2.7523, "step": 39834 }, { "epoch": 3.12, "learning_rate": 6.8764212342194e-05, "loss": 2.6329, "step": 39835 }, { "epoch": 3.12, "learning_rate": 6.876342821296951e-05, "loss": 2.7094, "step": 39836 }, { "epoch": 3.12, "learning_rate": 6.876264408374501e-05, "loss": 2.8558, "step": 39837 }, { "epoch": 3.12, "learning_rate": 6.876185995452052e-05, "loss": 2.7999, "step": 39838 }, { "epoch": 3.12, "learning_rate": 6.876107582529602e-05, "loss": 2.6245, "step": 39839 }, { "epoch": 3.12, "learning_rate": 6.876029169607151e-05, "loss": 2.7551, "step": 39840 }, { "epoch": 3.12, "learning_rate": 6.875950756684702e-05, "loss": 2.7404, "step": 39841 }, { "epoch": 3.12, "learning_rate": 6.875872343762252e-05, "loss": 2.7185, "step": 39842 }, { "epoch": 3.12, "learning_rate": 6.875793930839802e-05, "loss": 2.7544, "step": 39843 }, { "epoch": 3.12, "learning_rate": 6.875715517917353e-05, "loss": 2.7876, "step": 39844 }, { "epoch": 3.12, "learning_rate": 6.875637104994903e-05, "loss": 2.7405, "step": 39845 }, { "epoch": 3.12, "learning_rate": 6.875558692072453e-05, "loss": 2.6264, "step": 39846 }, { "epoch": 3.12, "learning_rate": 6.875480279150004e-05, "loss": 2.6992, "step": 39847 }, { "epoch": 3.12, "learning_rate": 6.875401866227554e-05, "loss": 2.6439, "step": 39848 }, { "epoch": 3.12, "learning_rate": 6.875323453305104e-05, "loss": 2.636, "step": 39849 }, { "epoch": 3.12, "learning_rate": 6.875245040382655e-05, "loss": 2.6898, "step": 39850 }, { "epoch": 3.12, "learning_rate": 6.875166627460205e-05, "loss": 2.7169, "step": 39851 }, { "epoch": 3.12, "learning_rate": 6.875088214537756e-05, "loss": 2.6329, "step": 39852 }, { "epoch": 3.12, "learning_rate": 6.875009801615306e-05, "loss": 2.6378, "step": 39853 }, { "epoch": 3.13, "learning_rate": 6.874931388692856e-05, "loss": 2.8156, "step": 39854 }, { "epoch": 3.13, "learning_rate": 6.874852975770407e-05, "loss": 2.533, "step": 39855 }, { "epoch": 3.13, "learning_rate": 6.874774562847957e-05, "loss": 2.7689, "step": 39856 }, { "epoch": 3.13, "learning_rate": 6.874696149925507e-05, "loss": 2.8557, "step": 39857 }, { "epoch": 3.13, "learning_rate": 6.874617737003058e-05, "loss": 2.7197, "step": 39858 }, { "epoch": 3.13, "learning_rate": 6.87453932408061e-05, "loss": 2.6095, "step": 39859 }, { "epoch": 3.13, "learning_rate": 6.87446091115816e-05, "loss": 2.7102, "step": 39860 }, { "epoch": 3.13, "learning_rate": 6.87438249823571e-05, "loss": 2.5148, "step": 39861 }, { "epoch": 3.13, "learning_rate": 6.87430408531326e-05, "loss": 2.6095, "step": 39862 }, { "epoch": 3.13, "learning_rate": 6.874225672390811e-05, "loss": 2.5511, "step": 39863 }, { "epoch": 3.13, "learning_rate": 6.874147259468361e-05, "loss": 2.6166, "step": 39864 }, { "epoch": 3.13, "learning_rate": 6.874068846545912e-05, "loss": 2.9413, "step": 39865 }, { "epoch": 3.13, "learning_rate": 6.873990433623462e-05, "loss": 2.8306, "step": 39866 }, { "epoch": 3.13, "learning_rate": 6.873912020701012e-05, "loss": 3.0029, "step": 39867 }, { "epoch": 3.13, "learning_rate": 6.873833607778563e-05, "loss": 2.7455, "step": 39868 }, { "epoch": 3.13, "learning_rate": 6.873755194856113e-05, "loss": 2.7126, "step": 39869 }, { "epoch": 3.13, "learning_rate": 6.873676781933664e-05, "loss": 2.678, "step": 39870 }, { "epoch": 3.13, "learning_rate": 6.873598369011214e-05, "loss": 2.7174, "step": 39871 }, { "epoch": 3.13, "learning_rate": 6.873519956088764e-05, "loss": 2.8172, "step": 39872 }, { "epoch": 3.13, "learning_rate": 6.873441543166315e-05, "loss": 2.6156, "step": 39873 }, { "epoch": 3.13, "learning_rate": 6.873363130243865e-05, "loss": 2.6565, "step": 39874 }, { "epoch": 3.13, "learning_rate": 6.873284717321415e-05, "loss": 2.6511, "step": 39875 }, { "epoch": 3.13, "learning_rate": 6.873206304398964e-05, "loss": 2.8712, "step": 39876 }, { "epoch": 3.13, "learning_rate": 6.873127891476515e-05, "loss": 2.5497, "step": 39877 }, { "epoch": 3.13, "learning_rate": 6.873049478554065e-05, "loss": 3.2614, "step": 39878 }, { "epoch": 3.13, "learning_rate": 6.872971065631616e-05, "loss": 2.8007, "step": 39879 }, { "epoch": 3.13, "learning_rate": 6.872892652709166e-05, "loss": 2.6145, "step": 39880 }, { "epoch": 3.13, "learning_rate": 6.872814239786716e-05, "loss": 2.6346, "step": 39881 }, { "epoch": 3.13, "learning_rate": 6.872735826864267e-05, "loss": 2.5162, "step": 39882 }, { "epoch": 3.13, "learning_rate": 6.872657413941817e-05, "loss": 3.0604, "step": 39883 }, { "epoch": 3.13, "learning_rate": 6.872579001019369e-05, "loss": 2.8489, "step": 39884 }, { "epoch": 3.13, "learning_rate": 6.872500588096919e-05, "loss": 2.7016, "step": 39885 }, { "epoch": 3.13, "learning_rate": 6.87242217517447e-05, "loss": 2.7892, "step": 39886 }, { "epoch": 3.13, "learning_rate": 6.87234376225202e-05, "loss": 2.5113, "step": 39887 }, { "epoch": 3.13, "learning_rate": 6.87226534932957e-05, "loss": 2.5427, "step": 39888 }, { "epoch": 3.13, "learning_rate": 6.87218693640712e-05, "loss": 2.5495, "step": 39889 }, { "epoch": 3.13, "learning_rate": 6.872108523484671e-05, "loss": 2.8616, "step": 39890 }, { "epoch": 3.13, "learning_rate": 6.872030110562221e-05, "loss": 2.7732, "step": 39891 }, { "epoch": 3.13, "learning_rate": 6.871951697639772e-05, "loss": 2.9565, "step": 39892 }, { "epoch": 3.13, "learning_rate": 6.871873284717322e-05, "loss": 2.4655, "step": 39893 }, { "epoch": 3.13, "learning_rate": 6.871794871794872e-05, "loss": 2.672, "step": 39894 }, { "epoch": 3.13, "learning_rate": 6.871716458872423e-05, "loss": 2.6511, "step": 39895 }, { "epoch": 3.13, "learning_rate": 6.871638045949973e-05, "loss": 2.4755, "step": 39896 }, { "epoch": 3.13, "learning_rate": 6.871559633027523e-05, "loss": 2.7259, "step": 39897 }, { "epoch": 3.13, "learning_rate": 6.871481220105074e-05, "loss": 2.5998, "step": 39898 }, { "epoch": 3.13, "learning_rate": 6.871402807182624e-05, "loss": 2.8799, "step": 39899 }, { "epoch": 3.13, "learning_rate": 6.871324394260175e-05, "loss": 2.538, "step": 39900 }, { "epoch": 3.13, "learning_rate": 6.871245981337725e-05, "loss": 2.8183, "step": 39901 }, { "epoch": 3.13, "learning_rate": 6.871167568415275e-05, "loss": 2.7891, "step": 39902 }, { "epoch": 3.13, "learning_rate": 6.871089155492826e-05, "loss": 2.6021, "step": 39903 }, { "epoch": 3.13, "learning_rate": 6.871010742570376e-05, "loss": 2.7238, "step": 39904 }, { "epoch": 3.13, "learning_rate": 6.870932329647926e-05, "loss": 2.6848, "step": 39905 }, { "epoch": 3.13, "learning_rate": 6.870853916725477e-05, "loss": 2.5678, "step": 39906 }, { "epoch": 3.13, "learning_rate": 6.870775503803027e-05, "loss": 2.608, "step": 39907 }, { "epoch": 3.13, "learning_rate": 6.870697090880578e-05, "loss": 2.8003, "step": 39908 }, { "epoch": 3.13, "learning_rate": 6.870618677958128e-05, "loss": 2.7223, "step": 39909 }, { "epoch": 3.13, "learning_rate": 6.870540265035678e-05, "loss": 2.6974, "step": 39910 }, { "epoch": 3.13, "learning_rate": 6.870461852113229e-05, "loss": 2.793, "step": 39911 }, { "epoch": 3.13, "learning_rate": 6.870383439190779e-05, "loss": 2.8524, "step": 39912 }, { "epoch": 3.13, "learning_rate": 6.87030502626833e-05, "loss": 2.8433, "step": 39913 }, { "epoch": 3.13, "learning_rate": 6.87022661334588e-05, "loss": 2.698, "step": 39914 }, { "epoch": 3.13, "learning_rate": 6.87014820042343e-05, "loss": 2.6046, "step": 39915 }, { "epoch": 3.13, "learning_rate": 6.87006978750098e-05, "loss": 2.7237, "step": 39916 }, { "epoch": 3.13, "learning_rate": 6.869991374578531e-05, "loss": 2.687, "step": 39917 }, { "epoch": 3.13, "learning_rate": 6.869912961656081e-05, "loss": 2.5878, "step": 39918 }, { "epoch": 3.13, "learning_rate": 6.869834548733632e-05, "loss": 2.5976, "step": 39919 }, { "epoch": 3.13, "learning_rate": 6.869756135811182e-05, "loss": 2.8474, "step": 39920 }, { "epoch": 3.13, "learning_rate": 6.869677722888732e-05, "loss": 2.8507, "step": 39921 }, { "epoch": 3.13, "learning_rate": 6.869599309966283e-05, "loss": 2.7608, "step": 39922 }, { "epoch": 3.13, "learning_rate": 6.869520897043833e-05, "loss": 2.7252, "step": 39923 }, { "epoch": 3.13, "learning_rate": 6.869442484121383e-05, "loss": 2.7534, "step": 39924 }, { "epoch": 3.13, "learning_rate": 6.869364071198934e-05, "loss": 2.7905, "step": 39925 }, { "epoch": 3.13, "learning_rate": 6.869285658276484e-05, "loss": 2.7615, "step": 39926 }, { "epoch": 3.13, "learning_rate": 6.869207245354035e-05, "loss": 2.9584, "step": 39927 }, { "epoch": 3.13, "learning_rate": 6.869128832431585e-05, "loss": 2.7063, "step": 39928 }, { "epoch": 3.13, "learning_rate": 6.869050419509135e-05, "loss": 2.7187, "step": 39929 }, { "epoch": 3.13, "learning_rate": 6.868972006586686e-05, "loss": 2.5813, "step": 39930 }, { "epoch": 3.13, "learning_rate": 6.868893593664236e-05, "loss": 2.7615, "step": 39931 }, { "epoch": 3.13, "learning_rate": 6.868815180741786e-05, "loss": 2.7605, "step": 39932 }, { "epoch": 3.13, "learning_rate": 6.868736767819337e-05, "loss": 2.785, "step": 39933 }, { "epoch": 3.13, "learning_rate": 6.868658354896887e-05, "loss": 2.6584, "step": 39934 }, { "epoch": 3.13, "learning_rate": 6.868579941974437e-05, "loss": 2.6268, "step": 39935 }, { "epoch": 3.13, "learning_rate": 6.868501529051988e-05, "loss": 2.8644, "step": 39936 }, { "epoch": 3.13, "learning_rate": 6.868423116129538e-05, "loss": 2.796, "step": 39937 }, { "epoch": 3.13, "learning_rate": 6.868344703207089e-05, "loss": 2.8235, "step": 39938 }, { "epoch": 3.13, "learning_rate": 6.868266290284639e-05, "loss": 2.8635, "step": 39939 }, { "epoch": 3.13, "learning_rate": 6.868187877362189e-05, "loss": 2.8479, "step": 39940 }, { "epoch": 3.13, "learning_rate": 6.868109464439741e-05, "loss": 2.629, "step": 39941 }, { "epoch": 3.13, "learning_rate": 6.868031051517291e-05, "loss": 3.0566, "step": 39942 }, { "epoch": 3.13, "learning_rate": 6.867952638594842e-05, "loss": 2.7871, "step": 39943 }, { "epoch": 3.13, "learning_rate": 6.867874225672392e-05, "loss": 2.7993, "step": 39944 }, { "epoch": 3.13, "learning_rate": 6.867795812749942e-05, "loss": 2.9007, "step": 39945 }, { "epoch": 3.13, "learning_rate": 6.867717399827493e-05, "loss": 2.9117, "step": 39946 }, { "epoch": 3.13, "learning_rate": 6.867638986905043e-05, "loss": 2.7195, "step": 39947 }, { "epoch": 3.13, "learning_rate": 6.867560573982592e-05, "loss": 2.8688, "step": 39948 }, { "epoch": 3.13, "learning_rate": 6.867482161060143e-05, "loss": 2.7067, "step": 39949 }, { "epoch": 3.13, "learning_rate": 6.867403748137693e-05, "loss": 2.6986, "step": 39950 }, { "epoch": 3.13, "learning_rate": 6.867325335215243e-05, "loss": 2.6137, "step": 39951 }, { "epoch": 3.13, "learning_rate": 6.867246922292794e-05, "loss": 2.8527, "step": 39952 }, { "epoch": 3.13, "learning_rate": 6.867168509370344e-05, "loss": 2.5748, "step": 39953 }, { "epoch": 3.13, "learning_rate": 6.867090096447894e-05, "loss": 2.669, "step": 39954 }, { "epoch": 3.13, "learning_rate": 6.867011683525445e-05, "loss": 2.7947, "step": 39955 }, { "epoch": 3.13, "learning_rate": 6.866933270602995e-05, "loss": 2.6025, "step": 39956 }, { "epoch": 3.13, "learning_rate": 6.866854857680546e-05, "loss": 2.6733, "step": 39957 }, { "epoch": 3.13, "learning_rate": 6.866776444758096e-05, "loss": 2.6916, "step": 39958 }, { "epoch": 3.13, "learning_rate": 6.866698031835646e-05, "loss": 2.5243, "step": 39959 }, { "epoch": 3.13, "learning_rate": 6.866619618913197e-05, "loss": 2.7028, "step": 39960 }, { "epoch": 3.13, "learning_rate": 6.866541205990747e-05, "loss": 2.6833, "step": 39961 }, { "epoch": 3.13, "learning_rate": 6.866462793068297e-05, "loss": 2.8275, "step": 39962 }, { "epoch": 3.13, "learning_rate": 6.866384380145848e-05, "loss": 2.9258, "step": 39963 }, { "epoch": 3.13, "learning_rate": 6.866305967223398e-05, "loss": 2.8321, "step": 39964 }, { "epoch": 3.13, "learning_rate": 6.866227554300948e-05, "loss": 2.5904, "step": 39965 }, { "epoch": 3.13, "learning_rate": 6.866149141378499e-05, "loss": 2.7223, "step": 39966 }, { "epoch": 3.13, "learning_rate": 6.86607072845605e-05, "loss": 2.8335, "step": 39967 }, { "epoch": 3.13, "learning_rate": 6.865992315533601e-05, "loss": 2.7751, "step": 39968 }, { "epoch": 3.13, "learning_rate": 6.865913902611151e-05, "loss": 2.6174, "step": 39969 }, { "epoch": 3.13, "learning_rate": 6.865835489688702e-05, "loss": 2.7215, "step": 39970 }, { "epoch": 3.13, "learning_rate": 6.865757076766252e-05, "loss": 2.7871, "step": 39971 }, { "epoch": 3.13, "learning_rate": 6.865678663843802e-05, "loss": 2.9402, "step": 39972 }, { "epoch": 3.13, "learning_rate": 6.865600250921353e-05, "loss": 2.6508, "step": 39973 }, { "epoch": 3.13, "learning_rate": 6.865521837998903e-05, "loss": 2.5782, "step": 39974 }, { "epoch": 3.13, "learning_rate": 6.865443425076454e-05, "loss": 2.6899, "step": 39975 }, { "epoch": 3.13, "learning_rate": 6.865365012154004e-05, "loss": 2.7104, "step": 39976 }, { "epoch": 3.13, "learning_rate": 6.865286599231554e-05, "loss": 2.6749, "step": 39977 }, { "epoch": 3.13, "learning_rate": 6.865208186309105e-05, "loss": 2.8356, "step": 39978 }, { "epoch": 3.13, "learning_rate": 6.865129773386655e-05, "loss": 2.8216, "step": 39979 }, { "epoch": 3.13, "learning_rate": 6.865051360464205e-05, "loss": 2.583, "step": 39980 }, { "epoch": 3.14, "learning_rate": 6.864972947541756e-05, "loss": 2.9278, "step": 39981 }, { "epoch": 3.14, "learning_rate": 6.864894534619306e-05, "loss": 2.8047, "step": 39982 }, { "epoch": 3.14, "learning_rate": 6.864816121696856e-05, "loss": 2.7544, "step": 39983 }, { "epoch": 3.14, "learning_rate": 6.864737708774405e-05, "loss": 2.6371, "step": 39984 }, { "epoch": 3.14, "learning_rate": 6.864659295851956e-05, "loss": 2.6752, "step": 39985 }, { "epoch": 3.14, "learning_rate": 6.864580882929506e-05, "loss": 2.5758, "step": 39986 }, { "epoch": 3.14, "learning_rate": 6.864502470007057e-05, "loss": 2.6648, "step": 39987 }, { "epoch": 3.14, "learning_rate": 6.864424057084607e-05, "loss": 2.5903, "step": 39988 }, { "epoch": 3.14, "learning_rate": 6.864345644162157e-05, "loss": 2.7073, "step": 39989 }, { "epoch": 3.14, "learning_rate": 6.864267231239708e-05, "loss": 2.6778, "step": 39990 }, { "epoch": 3.14, "learning_rate": 6.864188818317258e-05, "loss": 2.9789, "step": 39991 }, { "epoch": 3.14, "learning_rate": 6.86411040539481e-05, "loss": 2.7345, "step": 39992 }, { "epoch": 3.14, "learning_rate": 6.86403199247236e-05, "loss": 2.8421, "step": 39993 }, { "epoch": 3.14, "learning_rate": 6.86395357954991e-05, "loss": 2.5909, "step": 39994 }, { "epoch": 3.14, "learning_rate": 6.863875166627461e-05, "loss": 2.6094, "step": 39995 }, { "epoch": 3.14, "learning_rate": 6.863796753705011e-05, "loss": 2.5704, "step": 39996 }, { "epoch": 3.14, "learning_rate": 6.863718340782562e-05, "loss": 2.7783, "step": 39997 }, { "epoch": 3.14, "learning_rate": 6.863639927860112e-05, "loss": 2.6186, "step": 39998 }, { "epoch": 3.14, "learning_rate": 6.863561514937662e-05, "loss": 2.7579, "step": 39999 }, { "epoch": 3.14, "learning_rate": 6.863483102015213e-05, "loss": 2.6327, "step": 40000 }, { "epoch": 3.14, "eval_loss": 2.898805618286133, "eval_runtime": 217.7002, "eval_samples_per_second": 52.053, "eval_steps_per_second": 3.257, "step": 40000 }, { "epoch": 3.14, "learning_rate": 6.863404689092763e-05, "loss": 2.8161, "step": 40001 }, { "epoch": 3.14, "learning_rate": 6.863326276170313e-05, "loss": 2.7021, "step": 40002 }, { "epoch": 3.14, "learning_rate": 6.863247863247864e-05, "loss": 2.7896, "step": 40003 }, { "epoch": 3.14, "learning_rate": 6.863169450325414e-05, "loss": 2.5417, "step": 40004 }, { "epoch": 3.14, "learning_rate": 6.863091037402965e-05, "loss": 2.5572, "step": 40005 }, { "epoch": 3.14, "learning_rate": 6.863012624480515e-05, "loss": 2.5617, "step": 40006 }, { "epoch": 3.14, "learning_rate": 6.862934211558065e-05, "loss": 2.7855, "step": 40007 }, { "epoch": 3.14, "learning_rate": 6.862855798635616e-05, "loss": 2.6442, "step": 40008 }, { "epoch": 3.14, "learning_rate": 6.862777385713166e-05, "loss": 2.7906, "step": 40009 }, { "epoch": 3.14, "learning_rate": 6.862698972790716e-05, "loss": 2.5944, "step": 40010 }, { "epoch": 3.14, "learning_rate": 6.862620559868267e-05, "loss": 2.6155, "step": 40011 }, { "epoch": 3.14, "learning_rate": 6.862542146945817e-05, "loss": 2.7195, "step": 40012 }, { "epoch": 3.14, "learning_rate": 6.862463734023367e-05, "loss": 2.587, "step": 40013 }, { "epoch": 3.14, "learning_rate": 6.862385321100918e-05, "loss": 2.8359, "step": 40014 }, { "epoch": 3.14, "learning_rate": 6.862306908178468e-05, "loss": 3.0096, "step": 40015 }, { "epoch": 3.14, "learning_rate": 6.862228495256019e-05, "loss": 2.6704, "step": 40016 }, { "epoch": 3.14, "learning_rate": 6.862150082333569e-05, "loss": 2.7051, "step": 40017 }, { "epoch": 3.14, "learning_rate": 6.86207166941112e-05, "loss": 2.6948, "step": 40018 }, { "epoch": 3.14, "learning_rate": 6.86199325648867e-05, "loss": 2.7112, "step": 40019 }, { "epoch": 3.14, "learning_rate": 6.86191484356622e-05, "loss": 2.832, "step": 40020 }, { "epoch": 3.14, "learning_rate": 6.86183643064377e-05, "loss": 2.8558, "step": 40021 }, { "epoch": 3.14, "learning_rate": 6.861758017721321e-05, "loss": 2.6775, "step": 40022 }, { "epoch": 3.14, "learning_rate": 6.861679604798871e-05, "loss": 2.5916, "step": 40023 }, { "epoch": 3.14, "learning_rate": 6.861601191876422e-05, "loss": 2.7367, "step": 40024 }, { "epoch": 3.14, "learning_rate": 6.861522778953972e-05, "loss": 2.8671, "step": 40025 }, { "epoch": 3.14, "learning_rate": 6.861444366031522e-05, "loss": 2.5086, "step": 40026 }, { "epoch": 3.14, "learning_rate": 6.861365953109073e-05, "loss": 2.547, "step": 40027 }, { "epoch": 3.14, "learning_rate": 6.861287540186623e-05, "loss": 2.6329, "step": 40028 }, { "epoch": 3.14, "learning_rate": 6.861209127264173e-05, "loss": 2.7604, "step": 40029 }, { "epoch": 3.14, "learning_rate": 6.861130714341724e-05, "loss": 3.0021, "step": 40030 }, { "epoch": 3.14, "learning_rate": 6.861052301419274e-05, "loss": 2.7771, "step": 40031 }, { "epoch": 3.14, "learning_rate": 6.860973888496824e-05, "loss": 2.772, "step": 40032 }, { "epoch": 3.14, "learning_rate": 6.860895475574375e-05, "loss": 2.7808, "step": 40033 }, { "epoch": 3.14, "learning_rate": 6.860817062651925e-05, "loss": 2.8773, "step": 40034 }, { "epoch": 3.14, "learning_rate": 6.860738649729476e-05, "loss": 2.719, "step": 40035 }, { "epoch": 3.14, "learning_rate": 6.860660236807026e-05, "loss": 2.5633, "step": 40036 }, { "epoch": 3.14, "learning_rate": 6.860581823884576e-05, "loss": 2.8402, "step": 40037 }, { "epoch": 3.14, "learning_rate": 6.860503410962127e-05, "loss": 2.7383, "step": 40038 }, { "epoch": 3.14, "learning_rate": 6.860424998039677e-05, "loss": 2.8147, "step": 40039 }, { "epoch": 3.14, "learning_rate": 6.860346585117227e-05, "loss": 2.6339, "step": 40040 }, { "epoch": 3.14, "learning_rate": 6.860268172194778e-05, "loss": 2.4858, "step": 40041 }, { "epoch": 3.14, "learning_rate": 6.860189759272328e-05, "loss": 2.7883, "step": 40042 }, { "epoch": 3.14, "learning_rate": 6.860111346349879e-05, "loss": 2.7467, "step": 40043 }, { "epoch": 3.14, "learning_rate": 6.860032933427429e-05, "loss": 2.7798, "step": 40044 }, { "epoch": 3.14, "learning_rate": 6.859954520504979e-05, "loss": 2.6807, "step": 40045 }, { "epoch": 3.14, "learning_rate": 6.85987610758253e-05, "loss": 2.9238, "step": 40046 }, { "epoch": 3.14, "learning_rate": 6.85979769466008e-05, "loss": 2.7778, "step": 40047 }, { "epoch": 3.14, "learning_rate": 6.85971928173763e-05, "loss": 2.4503, "step": 40048 }, { "epoch": 3.14, "learning_rate": 6.859640868815182e-05, "loss": 2.44, "step": 40049 }, { "epoch": 3.14, "learning_rate": 6.859562455892732e-05, "loss": 2.724, "step": 40050 }, { "epoch": 3.14, "learning_rate": 6.859484042970283e-05, "loss": 2.6977, "step": 40051 }, { "epoch": 3.14, "learning_rate": 6.859405630047833e-05, "loss": 2.5404, "step": 40052 }, { "epoch": 3.14, "learning_rate": 6.859327217125384e-05, "loss": 2.7769, "step": 40053 }, { "epoch": 3.14, "learning_rate": 6.859248804202934e-05, "loss": 2.6702, "step": 40054 }, { "epoch": 3.14, "learning_rate": 6.859170391280484e-05, "loss": 3.0689, "step": 40055 }, { "epoch": 3.14, "learning_rate": 6.859091978358033e-05, "loss": 2.5845, "step": 40056 }, { "epoch": 3.14, "learning_rate": 6.859013565435584e-05, "loss": 2.5097, "step": 40057 }, { "epoch": 3.14, "learning_rate": 6.858935152513134e-05, "loss": 2.704, "step": 40058 }, { "epoch": 3.14, "learning_rate": 6.858856739590684e-05, "loss": 2.7185, "step": 40059 }, { "epoch": 3.14, "learning_rate": 6.858778326668235e-05, "loss": 2.7693, "step": 40060 }, { "epoch": 3.14, "learning_rate": 6.858699913745785e-05, "loss": 2.5781, "step": 40061 }, { "epoch": 3.14, "learning_rate": 6.858621500823336e-05, "loss": 2.739, "step": 40062 }, { "epoch": 3.14, "learning_rate": 6.858543087900886e-05, "loss": 2.9588, "step": 40063 }, { "epoch": 3.14, "learning_rate": 6.858464674978436e-05, "loss": 2.9391, "step": 40064 }, { "epoch": 3.14, "learning_rate": 6.858386262055987e-05, "loss": 2.6725, "step": 40065 }, { "epoch": 3.14, "learning_rate": 6.858307849133537e-05, "loss": 2.6305, "step": 40066 }, { "epoch": 3.14, "learning_rate": 6.858229436211087e-05, "loss": 2.603, "step": 40067 }, { "epoch": 3.14, "learning_rate": 6.858151023288638e-05, "loss": 2.7719, "step": 40068 }, { "epoch": 3.14, "learning_rate": 6.858072610366188e-05, "loss": 2.7977, "step": 40069 }, { "epoch": 3.14, "learning_rate": 6.857994197443738e-05, "loss": 2.8606, "step": 40070 }, { "epoch": 3.14, "learning_rate": 6.857915784521289e-05, "loss": 2.8505, "step": 40071 }, { "epoch": 3.14, "learning_rate": 6.857837371598839e-05, "loss": 2.6197, "step": 40072 }, { "epoch": 3.14, "learning_rate": 6.85775895867639e-05, "loss": 2.7193, "step": 40073 }, { "epoch": 3.14, "learning_rate": 6.85768054575394e-05, "loss": 2.6148, "step": 40074 }, { "epoch": 3.14, "learning_rate": 6.857602132831492e-05, "loss": 2.6474, "step": 40075 }, { "epoch": 3.14, "learning_rate": 6.857523719909042e-05, "loss": 2.6947, "step": 40076 }, { "epoch": 3.14, "learning_rate": 6.857445306986592e-05, "loss": 2.6057, "step": 40077 }, { "epoch": 3.14, "learning_rate": 6.857366894064143e-05, "loss": 2.6974, "step": 40078 }, { "epoch": 3.14, "learning_rate": 6.857288481141693e-05, "loss": 2.7722, "step": 40079 }, { "epoch": 3.14, "learning_rate": 6.857210068219243e-05, "loss": 3.1348, "step": 40080 }, { "epoch": 3.14, "learning_rate": 6.857131655296794e-05, "loss": 2.6599, "step": 40081 }, { "epoch": 3.14, "learning_rate": 6.857053242374344e-05, "loss": 2.589, "step": 40082 }, { "epoch": 3.14, "learning_rate": 6.856974829451895e-05, "loss": 2.8361, "step": 40083 }, { "epoch": 3.14, "learning_rate": 6.856896416529445e-05, "loss": 2.5367, "step": 40084 }, { "epoch": 3.14, "learning_rate": 6.856818003606995e-05, "loss": 2.7325, "step": 40085 }, { "epoch": 3.14, "learning_rate": 6.856739590684546e-05, "loss": 2.4643, "step": 40086 }, { "epoch": 3.14, "learning_rate": 6.856661177762096e-05, "loss": 2.6556, "step": 40087 }, { "epoch": 3.14, "learning_rate": 6.856582764839646e-05, "loss": 2.6557, "step": 40088 }, { "epoch": 3.14, "learning_rate": 6.856504351917197e-05, "loss": 2.7719, "step": 40089 }, { "epoch": 3.14, "learning_rate": 6.856425938994747e-05, "loss": 2.7961, "step": 40090 }, { "epoch": 3.14, "learning_rate": 6.856347526072298e-05, "loss": 2.7868, "step": 40091 }, { "epoch": 3.14, "learning_rate": 6.856269113149847e-05, "loss": 2.629, "step": 40092 }, { "epoch": 3.14, "learning_rate": 6.856190700227397e-05, "loss": 2.6896, "step": 40093 }, { "epoch": 3.14, "learning_rate": 6.856112287304947e-05, "loss": 2.6894, "step": 40094 }, { "epoch": 3.14, "learning_rate": 6.856033874382498e-05, "loss": 2.8429, "step": 40095 }, { "epoch": 3.14, "learning_rate": 6.855955461460048e-05, "loss": 2.6661, "step": 40096 }, { "epoch": 3.14, "learning_rate": 6.855877048537598e-05, "loss": 2.8092, "step": 40097 }, { "epoch": 3.14, "learning_rate": 6.855798635615149e-05, "loss": 2.5602, "step": 40098 }, { "epoch": 3.14, "learning_rate": 6.855720222692699e-05, "loss": 2.624, "step": 40099 }, { "epoch": 3.14, "learning_rate": 6.855641809770251e-05, "loss": 2.8048, "step": 40100 }, { "epoch": 3.14, "learning_rate": 6.855563396847801e-05, "loss": 2.8479, "step": 40101 }, { "epoch": 3.14, "learning_rate": 6.855484983925352e-05, "loss": 2.7842, "step": 40102 }, { "epoch": 3.14, "learning_rate": 6.855406571002902e-05, "loss": 2.6043, "step": 40103 }, { "epoch": 3.14, "learning_rate": 6.855328158080452e-05, "loss": 2.7724, "step": 40104 }, { "epoch": 3.14, "learning_rate": 6.855249745158003e-05, "loss": 2.5934, "step": 40105 }, { "epoch": 3.14, "learning_rate": 6.855171332235553e-05, "loss": 2.7778, "step": 40106 }, { "epoch": 3.14, "learning_rate": 6.855092919313103e-05, "loss": 2.626, "step": 40107 }, { "epoch": 3.14, "learning_rate": 6.855014506390654e-05, "loss": 2.8625, "step": 40108 }, { "epoch": 3.15, "learning_rate": 6.854936093468204e-05, "loss": 2.7449, "step": 40109 }, { "epoch": 3.15, "learning_rate": 6.854857680545755e-05, "loss": 2.7455, "step": 40110 }, { "epoch": 3.15, "learning_rate": 6.854779267623305e-05, "loss": 2.7598, "step": 40111 }, { "epoch": 3.15, "learning_rate": 6.854700854700855e-05, "loss": 2.5699, "step": 40112 }, { "epoch": 3.15, "learning_rate": 6.854622441778406e-05, "loss": 2.5665, "step": 40113 }, { "epoch": 3.15, "learning_rate": 6.854544028855956e-05, "loss": 2.7415, "step": 40114 }, { "epoch": 3.15, "learning_rate": 6.854465615933506e-05, "loss": 2.6872, "step": 40115 }, { "epoch": 3.15, "learning_rate": 6.854387203011057e-05, "loss": 2.5727, "step": 40116 }, { "epoch": 3.15, "learning_rate": 6.854308790088607e-05, "loss": 2.6388, "step": 40117 }, { "epoch": 3.15, "learning_rate": 6.854230377166157e-05, "loss": 2.565, "step": 40118 }, { "epoch": 3.15, "learning_rate": 6.854151964243708e-05, "loss": 2.802, "step": 40119 }, { "epoch": 3.15, "learning_rate": 6.854073551321258e-05, "loss": 2.8616, "step": 40120 }, { "epoch": 3.15, "learning_rate": 6.853995138398809e-05, "loss": 2.632, "step": 40121 }, { "epoch": 3.15, "learning_rate": 6.853916725476359e-05, "loss": 2.6214, "step": 40122 }, { "epoch": 3.15, "learning_rate": 6.853838312553909e-05, "loss": 2.8664, "step": 40123 }, { "epoch": 3.15, "learning_rate": 6.85375989963146e-05, "loss": 2.5037, "step": 40124 }, { "epoch": 3.15, "learning_rate": 6.85368148670901e-05, "loss": 2.7498, "step": 40125 }, { "epoch": 3.15, "learning_rate": 6.85360307378656e-05, "loss": 2.8012, "step": 40126 }, { "epoch": 3.15, "learning_rate": 6.853524660864111e-05, "loss": 2.4482, "step": 40127 }, { "epoch": 3.15, "learning_rate": 6.853446247941661e-05, "loss": 2.8889, "step": 40128 }, { "epoch": 3.15, "learning_rate": 6.853367835019212e-05, "loss": 2.664, "step": 40129 }, { "epoch": 3.15, "learning_rate": 6.853289422096762e-05, "loss": 2.7275, "step": 40130 }, { "epoch": 3.15, "learning_rate": 6.853211009174312e-05, "loss": 2.6581, "step": 40131 }, { "epoch": 3.15, "learning_rate": 6.853132596251863e-05, "loss": 2.5925, "step": 40132 }, { "epoch": 3.15, "learning_rate": 6.853054183329413e-05, "loss": 2.6507, "step": 40133 }, { "epoch": 3.15, "learning_rate": 6.852975770406963e-05, "loss": 2.6285, "step": 40134 }, { "epoch": 3.15, "learning_rate": 6.852897357484514e-05, "loss": 2.8317, "step": 40135 }, { "epoch": 3.15, "learning_rate": 6.852818944562064e-05, "loss": 2.9036, "step": 40136 }, { "epoch": 3.15, "learning_rate": 6.852740531639614e-05, "loss": 2.9863, "step": 40137 }, { "epoch": 3.15, "learning_rate": 6.852662118717165e-05, "loss": 2.7703, "step": 40138 }, { "epoch": 3.15, "learning_rate": 6.852583705794715e-05, "loss": 2.7451, "step": 40139 }, { "epoch": 3.15, "learning_rate": 6.852505292872266e-05, "loss": 2.7487, "step": 40140 }, { "epoch": 3.15, "learning_rate": 6.852426879949816e-05, "loss": 2.5959, "step": 40141 }, { "epoch": 3.15, "learning_rate": 6.852348467027366e-05, "loss": 2.5278, "step": 40142 }, { "epoch": 3.15, "learning_rate": 6.852270054104917e-05, "loss": 2.6244, "step": 40143 }, { "epoch": 3.15, "learning_rate": 6.852191641182467e-05, "loss": 3.0011, "step": 40144 }, { "epoch": 3.15, "learning_rate": 6.852113228260017e-05, "loss": 2.534, "step": 40145 }, { "epoch": 3.15, "learning_rate": 6.852034815337568e-05, "loss": 2.6734, "step": 40146 }, { "epoch": 3.15, "learning_rate": 6.851956402415118e-05, "loss": 2.7082, "step": 40147 }, { "epoch": 3.15, "learning_rate": 6.851877989492668e-05, "loss": 2.6402, "step": 40148 }, { "epoch": 3.15, "learning_rate": 6.851799576570219e-05, "loss": 2.8538, "step": 40149 }, { "epoch": 3.15, "learning_rate": 6.851721163647769e-05, "loss": 2.8208, "step": 40150 }, { "epoch": 3.15, "learning_rate": 6.85164275072532e-05, "loss": 2.9317, "step": 40151 }, { "epoch": 3.15, "learning_rate": 6.85156433780287e-05, "loss": 2.9452, "step": 40152 }, { "epoch": 3.15, "learning_rate": 6.85148592488042e-05, "loss": 2.5684, "step": 40153 }, { "epoch": 3.15, "learning_rate": 6.851407511957971e-05, "loss": 2.6875, "step": 40154 }, { "epoch": 3.15, "learning_rate": 6.851329099035521e-05, "loss": 2.6848, "step": 40155 }, { "epoch": 3.15, "learning_rate": 6.851250686113071e-05, "loss": 2.7912, "step": 40156 }, { "epoch": 3.15, "learning_rate": 6.851172273190623e-05, "loss": 2.8733, "step": 40157 }, { "epoch": 3.15, "learning_rate": 6.851093860268174e-05, "loss": 2.607, "step": 40158 }, { "epoch": 3.15, "learning_rate": 6.851015447345724e-05, "loss": 2.9068, "step": 40159 }, { "epoch": 3.15, "learning_rate": 6.850937034423274e-05, "loss": 2.8503, "step": 40160 }, { "epoch": 3.15, "learning_rate": 6.850858621500825e-05, "loss": 2.6773, "step": 40161 }, { "epoch": 3.15, "learning_rate": 6.850780208578375e-05, "loss": 2.881, "step": 40162 }, { "epoch": 3.15, "learning_rate": 6.850701795655925e-05, "loss": 2.6247, "step": 40163 }, { "epoch": 3.15, "learning_rate": 6.850623382733474e-05, "loss": 2.6822, "step": 40164 }, { "epoch": 3.15, "learning_rate": 6.850544969811025e-05, "loss": 2.5802, "step": 40165 }, { "epoch": 3.15, "learning_rate": 6.850466556888575e-05, "loss": 2.4666, "step": 40166 }, { "epoch": 3.15, "learning_rate": 6.850388143966125e-05, "loss": 2.6612, "step": 40167 }, { "epoch": 3.15, "learning_rate": 6.850309731043676e-05, "loss": 2.7538, "step": 40168 }, { "epoch": 3.15, "learning_rate": 6.850231318121226e-05, "loss": 2.8458, "step": 40169 }, { "epoch": 3.15, "learning_rate": 6.850152905198777e-05, "loss": 2.5561, "step": 40170 }, { "epoch": 3.15, "learning_rate": 6.850074492276327e-05, "loss": 2.6054, "step": 40171 }, { "epoch": 3.15, "learning_rate": 6.849996079353877e-05, "loss": 2.8243, "step": 40172 }, { "epoch": 3.15, "learning_rate": 6.849917666431428e-05, "loss": 2.778, "step": 40173 }, { "epoch": 3.15, "learning_rate": 6.849839253508978e-05, "loss": 2.6602, "step": 40174 }, { "epoch": 3.15, "learning_rate": 6.849760840586528e-05, "loss": 2.7947, "step": 40175 }, { "epoch": 3.15, "learning_rate": 6.849682427664079e-05, "loss": 2.5629, "step": 40176 }, { "epoch": 3.15, "learning_rate": 6.849604014741629e-05, "loss": 2.7305, "step": 40177 }, { "epoch": 3.15, "learning_rate": 6.84952560181918e-05, "loss": 2.7569, "step": 40178 }, { "epoch": 3.15, "learning_rate": 6.84944718889673e-05, "loss": 2.6797, "step": 40179 }, { "epoch": 3.15, "learning_rate": 6.84936877597428e-05, "loss": 2.9109, "step": 40180 }, { "epoch": 3.15, "learning_rate": 6.84929036305183e-05, "loss": 2.7473, "step": 40181 }, { "epoch": 3.15, "learning_rate": 6.849211950129381e-05, "loss": 2.671, "step": 40182 }, { "epoch": 3.15, "learning_rate": 6.849133537206933e-05, "loss": 2.6679, "step": 40183 }, { "epoch": 3.15, "learning_rate": 6.849055124284483e-05, "loss": 2.7152, "step": 40184 }, { "epoch": 3.15, "learning_rate": 6.848976711362033e-05, "loss": 2.7183, "step": 40185 }, { "epoch": 3.15, "learning_rate": 6.848898298439584e-05, "loss": 2.7618, "step": 40186 }, { "epoch": 3.15, "learning_rate": 6.848819885517134e-05, "loss": 2.9169, "step": 40187 }, { "epoch": 3.15, "learning_rate": 6.848741472594685e-05, "loss": 2.7973, "step": 40188 }, { "epoch": 3.15, "learning_rate": 6.848663059672235e-05, "loss": 2.859, "step": 40189 }, { "epoch": 3.15, "learning_rate": 6.848584646749785e-05, "loss": 2.7122, "step": 40190 }, { "epoch": 3.15, "learning_rate": 6.848506233827336e-05, "loss": 2.7666, "step": 40191 }, { "epoch": 3.15, "learning_rate": 6.848427820904886e-05, "loss": 2.8341, "step": 40192 }, { "epoch": 3.15, "learning_rate": 6.848349407982436e-05, "loss": 2.5209, "step": 40193 }, { "epoch": 3.15, "learning_rate": 6.848270995059987e-05, "loss": 2.7865, "step": 40194 }, { "epoch": 3.15, "learning_rate": 6.848192582137537e-05, "loss": 2.5741, "step": 40195 }, { "epoch": 3.15, "learning_rate": 6.848114169215088e-05, "loss": 2.5176, "step": 40196 }, { "epoch": 3.15, "learning_rate": 6.848035756292638e-05, "loss": 2.5032, "step": 40197 }, { "epoch": 3.15, "learning_rate": 6.847957343370188e-05, "loss": 2.4854, "step": 40198 }, { "epoch": 3.15, "learning_rate": 6.847878930447739e-05, "loss": 2.5964, "step": 40199 }, { "epoch": 3.15, "learning_rate": 6.847800517525288e-05, "loss": 2.7119, "step": 40200 }, { "epoch": 3.15, "learning_rate": 6.847722104602838e-05, "loss": 2.8529, "step": 40201 }, { "epoch": 3.15, "learning_rate": 6.847643691680388e-05, "loss": 2.9904, "step": 40202 }, { "epoch": 3.15, "learning_rate": 6.847565278757939e-05, "loss": 2.6896, "step": 40203 }, { "epoch": 3.15, "learning_rate": 6.847486865835489e-05, "loss": 2.6187, "step": 40204 }, { "epoch": 3.15, "learning_rate": 6.84740845291304e-05, "loss": 2.9343, "step": 40205 }, { "epoch": 3.15, "learning_rate": 6.84733003999059e-05, "loss": 2.8325, "step": 40206 }, { "epoch": 3.15, "learning_rate": 6.84725162706814e-05, "loss": 2.8185, "step": 40207 }, { "epoch": 3.15, "learning_rate": 6.84717321414569e-05, "loss": 2.627, "step": 40208 }, { "epoch": 3.15, "learning_rate": 6.847094801223242e-05, "loss": 2.7279, "step": 40209 }, { "epoch": 3.15, "learning_rate": 6.847016388300793e-05, "loss": 2.9564, "step": 40210 }, { "epoch": 3.15, "learning_rate": 6.846937975378343e-05, "loss": 2.6429, "step": 40211 }, { "epoch": 3.15, "learning_rate": 6.846859562455893e-05, "loss": 2.6165, "step": 40212 }, { "epoch": 3.15, "learning_rate": 6.846781149533444e-05, "loss": 2.8202, "step": 40213 }, { "epoch": 3.15, "learning_rate": 6.846702736610994e-05, "loss": 2.6036, "step": 40214 }, { "epoch": 3.15, "learning_rate": 6.846624323688544e-05, "loss": 2.6244, "step": 40215 }, { "epoch": 3.15, "learning_rate": 6.846545910766095e-05, "loss": 2.6802, "step": 40216 }, { "epoch": 3.15, "learning_rate": 6.846467497843645e-05, "loss": 2.6664, "step": 40217 }, { "epoch": 3.15, "learning_rate": 6.846389084921196e-05, "loss": 2.7562, "step": 40218 }, { "epoch": 3.15, "learning_rate": 6.846310671998746e-05, "loss": 2.9052, "step": 40219 }, { "epoch": 3.15, "learning_rate": 6.846232259076296e-05, "loss": 2.6704, "step": 40220 }, { "epoch": 3.15, "learning_rate": 6.846153846153847e-05, "loss": 2.7143, "step": 40221 }, { "epoch": 3.15, "learning_rate": 6.846075433231397e-05, "loss": 2.6593, "step": 40222 }, { "epoch": 3.15, "learning_rate": 6.845997020308947e-05, "loss": 2.655, "step": 40223 }, { "epoch": 3.15, "learning_rate": 6.845918607386498e-05, "loss": 2.8466, "step": 40224 }, { "epoch": 3.15, "learning_rate": 6.845840194464048e-05, "loss": 2.856, "step": 40225 }, { "epoch": 3.15, "learning_rate": 6.845761781541599e-05, "loss": 2.6239, "step": 40226 }, { "epoch": 3.15, "learning_rate": 6.845683368619149e-05, "loss": 2.685, "step": 40227 }, { "epoch": 3.15, "learning_rate": 6.845604955696699e-05, "loss": 2.6234, "step": 40228 }, { "epoch": 3.15, "learning_rate": 6.84552654277425e-05, "loss": 2.8325, "step": 40229 }, { "epoch": 3.15, "learning_rate": 6.8454481298518e-05, "loss": 2.5062, "step": 40230 }, { "epoch": 3.15, "learning_rate": 6.84536971692935e-05, "loss": 2.7966, "step": 40231 }, { "epoch": 3.15, "learning_rate": 6.845291304006901e-05, "loss": 2.7317, "step": 40232 }, { "epoch": 3.15, "learning_rate": 6.845212891084451e-05, "loss": 2.7533, "step": 40233 }, { "epoch": 3.15, "learning_rate": 6.845134478162001e-05, "loss": 2.645, "step": 40234 }, { "epoch": 3.15, "learning_rate": 6.845056065239552e-05, "loss": 2.8471, "step": 40235 }, { "epoch": 3.16, "learning_rate": 6.844977652317102e-05, "loss": 2.6393, "step": 40236 }, { "epoch": 3.16, "learning_rate": 6.844899239394653e-05, "loss": 2.7361, "step": 40237 }, { "epoch": 3.16, "learning_rate": 6.844820826472203e-05, "loss": 2.9696, "step": 40238 }, { "epoch": 3.16, "learning_rate": 6.844742413549753e-05, "loss": 2.8837, "step": 40239 }, { "epoch": 3.16, "learning_rate": 6.844664000627304e-05, "loss": 2.5884, "step": 40240 }, { "epoch": 3.16, "learning_rate": 6.844585587704854e-05, "loss": 2.8517, "step": 40241 }, { "epoch": 3.16, "learning_rate": 6.844507174782404e-05, "loss": 2.4947, "step": 40242 }, { "epoch": 3.16, "learning_rate": 6.844428761859955e-05, "loss": 2.8489, "step": 40243 }, { "epoch": 3.16, "learning_rate": 6.844350348937505e-05, "loss": 2.6542, "step": 40244 }, { "epoch": 3.16, "learning_rate": 6.844271936015056e-05, "loss": 2.6611, "step": 40245 }, { "epoch": 3.16, "learning_rate": 6.844193523092606e-05, "loss": 2.6804, "step": 40246 }, { "epoch": 3.16, "learning_rate": 6.844115110170156e-05, "loss": 2.7156, "step": 40247 }, { "epoch": 3.16, "learning_rate": 6.844036697247707e-05, "loss": 2.6864, "step": 40248 }, { "epoch": 3.16, "learning_rate": 6.843958284325257e-05, "loss": 2.9854, "step": 40249 }, { "epoch": 3.16, "learning_rate": 6.843879871402807e-05, "loss": 2.7511, "step": 40250 }, { "epoch": 3.16, "learning_rate": 6.843801458480358e-05, "loss": 2.9247, "step": 40251 }, { "epoch": 3.16, "learning_rate": 6.843723045557908e-05, "loss": 2.8512, "step": 40252 }, { "epoch": 3.16, "learning_rate": 6.843644632635458e-05, "loss": 2.8271, "step": 40253 }, { "epoch": 3.16, "learning_rate": 6.843566219713009e-05, "loss": 2.6257, "step": 40254 }, { "epoch": 3.16, "learning_rate": 6.843487806790559e-05, "loss": 2.7446, "step": 40255 }, { "epoch": 3.16, "learning_rate": 6.84340939386811e-05, "loss": 2.6457, "step": 40256 }, { "epoch": 3.16, "learning_rate": 6.84333098094566e-05, "loss": 2.8587, "step": 40257 }, { "epoch": 3.16, "learning_rate": 6.84325256802321e-05, "loss": 2.6302, "step": 40258 }, { "epoch": 3.16, "learning_rate": 6.84317415510076e-05, "loss": 2.7214, "step": 40259 }, { "epoch": 3.16, "learning_rate": 6.843095742178311e-05, "loss": 2.6993, "step": 40260 }, { "epoch": 3.16, "learning_rate": 6.843017329255861e-05, "loss": 2.7743, "step": 40261 }, { "epoch": 3.16, "learning_rate": 6.842938916333412e-05, "loss": 2.624, "step": 40262 }, { "epoch": 3.16, "learning_rate": 6.842860503410962e-05, "loss": 2.7604, "step": 40263 }, { "epoch": 3.16, "learning_rate": 6.842782090488513e-05, "loss": 2.6608, "step": 40264 }, { "epoch": 3.16, "learning_rate": 6.842703677566064e-05, "loss": 2.6855, "step": 40265 }, { "epoch": 3.16, "learning_rate": 6.842625264643615e-05, "loss": 3.081, "step": 40266 }, { "epoch": 3.16, "learning_rate": 6.842546851721165e-05, "loss": 2.5483, "step": 40267 }, { "epoch": 3.16, "learning_rate": 6.842468438798715e-05, "loss": 2.7245, "step": 40268 }, { "epoch": 3.16, "learning_rate": 6.842390025876266e-05, "loss": 2.9816, "step": 40269 }, { "epoch": 3.16, "learning_rate": 6.842311612953816e-05, "loss": 2.8818, "step": 40270 }, { "epoch": 3.16, "learning_rate": 6.842233200031366e-05, "loss": 2.692, "step": 40271 }, { "epoch": 3.16, "learning_rate": 6.842154787108915e-05, "loss": 2.5168, "step": 40272 }, { "epoch": 3.16, "learning_rate": 6.842076374186466e-05, "loss": 2.907, "step": 40273 }, { "epoch": 3.16, "learning_rate": 6.841997961264016e-05, "loss": 2.7638, "step": 40274 }, { "epoch": 3.16, "learning_rate": 6.841919548341567e-05, "loss": 2.7611, "step": 40275 }, { "epoch": 3.16, "learning_rate": 6.841841135419117e-05, "loss": 3.005, "step": 40276 }, { "epoch": 3.16, "learning_rate": 6.841762722496667e-05, "loss": 2.6634, "step": 40277 }, { "epoch": 3.16, "learning_rate": 6.841684309574218e-05, "loss": 2.8094, "step": 40278 }, { "epoch": 3.16, "learning_rate": 6.841605896651768e-05, "loss": 2.4922, "step": 40279 }, { "epoch": 3.16, "learning_rate": 6.841527483729318e-05, "loss": 2.8088, "step": 40280 }, { "epoch": 3.16, "learning_rate": 6.841449070806869e-05, "loss": 2.6146, "step": 40281 }, { "epoch": 3.16, "learning_rate": 6.841370657884419e-05, "loss": 2.7837, "step": 40282 }, { "epoch": 3.16, "learning_rate": 6.84129224496197e-05, "loss": 2.6456, "step": 40283 }, { "epoch": 3.16, "learning_rate": 6.84121383203952e-05, "loss": 2.6337, "step": 40284 }, { "epoch": 3.16, "learning_rate": 6.84113541911707e-05, "loss": 2.7733, "step": 40285 }, { "epoch": 3.16, "learning_rate": 6.84105700619462e-05, "loss": 2.6685, "step": 40286 }, { "epoch": 3.16, "learning_rate": 6.840978593272171e-05, "loss": 2.5264, "step": 40287 }, { "epoch": 3.16, "learning_rate": 6.840900180349721e-05, "loss": 2.6381, "step": 40288 }, { "epoch": 3.16, "learning_rate": 6.840821767427272e-05, "loss": 2.6411, "step": 40289 }, { "epoch": 3.16, "learning_rate": 6.840743354504822e-05, "loss": 2.7609, "step": 40290 }, { "epoch": 3.16, "learning_rate": 6.840664941582374e-05, "loss": 2.5168, "step": 40291 }, { "epoch": 3.16, "learning_rate": 6.840586528659924e-05, "loss": 2.8085, "step": 40292 }, { "epoch": 3.16, "learning_rate": 6.840508115737475e-05, "loss": 2.7994, "step": 40293 }, { "epoch": 3.16, "learning_rate": 6.840429702815025e-05, "loss": 2.6632, "step": 40294 }, { "epoch": 3.16, "learning_rate": 6.840351289892575e-05, "loss": 2.537, "step": 40295 }, { "epoch": 3.16, "learning_rate": 6.840272876970126e-05, "loss": 2.4451, "step": 40296 }, { "epoch": 3.16, "learning_rate": 6.840194464047676e-05, "loss": 2.6348, "step": 40297 }, { "epoch": 3.16, "learning_rate": 6.840116051125226e-05, "loss": 2.5054, "step": 40298 }, { "epoch": 3.16, "learning_rate": 6.840037638202777e-05, "loss": 2.6949, "step": 40299 }, { "epoch": 3.16, "learning_rate": 6.839959225280327e-05, "loss": 2.7624, "step": 40300 }, { "epoch": 3.16, "learning_rate": 6.839880812357877e-05, "loss": 2.6879, "step": 40301 }, { "epoch": 3.16, "learning_rate": 6.839802399435428e-05, "loss": 2.5437, "step": 40302 }, { "epoch": 3.16, "learning_rate": 6.839723986512978e-05, "loss": 2.7264, "step": 40303 }, { "epoch": 3.16, "learning_rate": 6.839645573590529e-05, "loss": 2.7209, "step": 40304 }, { "epoch": 3.16, "learning_rate": 6.839567160668079e-05, "loss": 2.8152, "step": 40305 }, { "epoch": 3.16, "learning_rate": 6.839488747745629e-05, "loss": 2.6606, "step": 40306 }, { "epoch": 3.16, "learning_rate": 6.83941033482318e-05, "loss": 2.9161, "step": 40307 }, { "epoch": 3.16, "learning_rate": 6.839331921900729e-05, "loss": 2.9043, "step": 40308 }, { "epoch": 3.16, "learning_rate": 6.839253508978279e-05, "loss": 2.6263, "step": 40309 }, { "epoch": 3.16, "learning_rate": 6.83917509605583e-05, "loss": 2.6604, "step": 40310 }, { "epoch": 3.16, "learning_rate": 6.83909668313338e-05, "loss": 2.7441, "step": 40311 }, { "epoch": 3.16, "learning_rate": 6.83901827021093e-05, "loss": 2.7472, "step": 40312 }, { "epoch": 3.16, "learning_rate": 6.83893985728848e-05, "loss": 2.5993, "step": 40313 }, { "epoch": 3.16, "learning_rate": 6.838861444366031e-05, "loss": 2.8409, "step": 40314 }, { "epoch": 3.16, "learning_rate": 6.838783031443581e-05, "loss": 2.7448, "step": 40315 }, { "epoch": 3.16, "learning_rate": 6.838704618521132e-05, "loss": 2.8235, "step": 40316 }, { "epoch": 3.16, "learning_rate": 6.838626205598683e-05, "loss": 2.8091, "step": 40317 }, { "epoch": 3.16, "learning_rate": 6.838547792676234e-05, "loss": 2.5257, "step": 40318 }, { "epoch": 3.16, "learning_rate": 6.838469379753784e-05, "loss": 2.7377, "step": 40319 }, { "epoch": 3.16, "learning_rate": 6.838390966831334e-05, "loss": 2.7824, "step": 40320 }, { "epoch": 3.16, "learning_rate": 6.838312553908885e-05, "loss": 2.7691, "step": 40321 }, { "epoch": 3.16, "learning_rate": 6.838234140986435e-05, "loss": 2.6676, "step": 40322 }, { "epoch": 3.16, "learning_rate": 6.838155728063986e-05, "loss": 2.6662, "step": 40323 }, { "epoch": 3.16, "learning_rate": 6.838077315141536e-05, "loss": 2.6577, "step": 40324 }, { "epoch": 3.16, "learning_rate": 6.837998902219086e-05, "loss": 2.5556, "step": 40325 }, { "epoch": 3.16, "learning_rate": 6.837920489296637e-05, "loss": 2.9326, "step": 40326 }, { "epoch": 3.16, "learning_rate": 6.837842076374187e-05, "loss": 2.698, "step": 40327 }, { "epoch": 3.16, "learning_rate": 6.837763663451737e-05, "loss": 2.7252, "step": 40328 }, { "epoch": 3.16, "learning_rate": 6.837685250529288e-05, "loss": 2.7392, "step": 40329 }, { "epoch": 3.16, "learning_rate": 6.837606837606838e-05, "loss": 2.6032, "step": 40330 }, { "epoch": 3.16, "learning_rate": 6.837528424684389e-05, "loss": 3.0497, "step": 40331 }, { "epoch": 3.16, "learning_rate": 6.837450011761939e-05, "loss": 2.6811, "step": 40332 }, { "epoch": 3.16, "learning_rate": 6.837371598839489e-05, "loss": 2.5768, "step": 40333 }, { "epoch": 3.16, "learning_rate": 6.83729318591704e-05, "loss": 2.9004, "step": 40334 }, { "epoch": 3.16, "learning_rate": 6.83721477299459e-05, "loss": 2.4645, "step": 40335 }, { "epoch": 3.16, "learning_rate": 6.83713636007214e-05, "loss": 2.4649, "step": 40336 }, { "epoch": 3.16, "learning_rate": 6.837057947149691e-05, "loss": 2.4865, "step": 40337 }, { "epoch": 3.16, "learning_rate": 6.836979534227241e-05, "loss": 2.7838, "step": 40338 }, { "epoch": 3.16, "learning_rate": 6.836901121304791e-05, "loss": 2.5849, "step": 40339 }, { "epoch": 3.16, "learning_rate": 6.836822708382342e-05, "loss": 2.6239, "step": 40340 }, { "epoch": 3.16, "learning_rate": 6.836744295459892e-05, "loss": 2.7399, "step": 40341 }, { "epoch": 3.16, "learning_rate": 6.836665882537443e-05, "loss": 2.7936, "step": 40342 }, { "epoch": 3.16, "learning_rate": 6.836587469614993e-05, "loss": 2.7242, "step": 40343 }, { "epoch": 3.16, "learning_rate": 6.836509056692543e-05, "loss": 2.7459, "step": 40344 }, { "epoch": 3.16, "learning_rate": 6.836430643770094e-05, "loss": 2.8513, "step": 40345 }, { "epoch": 3.16, "learning_rate": 6.836352230847644e-05, "loss": 2.9158, "step": 40346 }, { "epoch": 3.16, "learning_rate": 6.836273817925194e-05, "loss": 2.8169, "step": 40347 }, { "epoch": 3.16, "learning_rate": 6.836195405002745e-05, "loss": 2.8662, "step": 40348 }, { "epoch": 3.16, "learning_rate": 6.836116992080295e-05, "loss": 2.9505, "step": 40349 }, { "epoch": 3.16, "learning_rate": 6.836038579157845e-05, "loss": 2.8043, "step": 40350 }, { "epoch": 3.16, "learning_rate": 6.835960166235396e-05, "loss": 2.717, "step": 40351 }, { "epoch": 3.16, "learning_rate": 6.835881753312946e-05, "loss": 2.7091, "step": 40352 }, { "epoch": 3.16, "learning_rate": 6.835803340390497e-05, "loss": 2.7056, "step": 40353 }, { "epoch": 3.16, "learning_rate": 6.835724927468047e-05, "loss": 2.7019, "step": 40354 }, { "epoch": 3.16, "learning_rate": 6.835646514545597e-05, "loss": 2.661, "step": 40355 }, { "epoch": 3.16, "learning_rate": 6.835568101623148e-05, "loss": 2.8183, "step": 40356 }, { "epoch": 3.16, "learning_rate": 6.835489688700698e-05, "loss": 2.7027, "step": 40357 }, { "epoch": 3.16, "learning_rate": 6.835411275778248e-05, "loss": 2.7028, "step": 40358 }, { "epoch": 3.16, "learning_rate": 6.835332862855799e-05, "loss": 2.8882, "step": 40359 }, { "epoch": 3.16, "learning_rate": 6.835254449933349e-05, "loss": 2.7296, "step": 40360 }, { "epoch": 3.16, "learning_rate": 6.8351760370109e-05, "loss": 2.6414, "step": 40361 }, { "epoch": 3.16, "learning_rate": 6.83509762408845e-05, "loss": 2.7761, "step": 40362 }, { "epoch": 3.16, "learning_rate": 6.835019211166e-05, "loss": 2.7641, "step": 40363 }, { "epoch": 3.17, "learning_rate": 6.83494079824355e-05, "loss": 2.5962, "step": 40364 }, { "epoch": 3.17, "learning_rate": 6.834862385321101e-05, "loss": 2.7791, "step": 40365 }, { "epoch": 3.17, "learning_rate": 6.834783972398651e-05, "loss": 2.809, "step": 40366 }, { "epoch": 3.17, "learning_rate": 6.834705559476202e-05, "loss": 2.7437, "step": 40367 }, { "epoch": 3.17, "learning_rate": 6.834627146553752e-05, "loss": 2.8159, "step": 40368 }, { "epoch": 3.17, "learning_rate": 6.834548733631302e-05, "loss": 2.8546, "step": 40369 }, { "epoch": 3.17, "learning_rate": 6.834470320708853e-05, "loss": 2.6808, "step": 40370 }, { "epoch": 3.17, "learning_rate": 6.834391907786403e-05, "loss": 2.8665, "step": 40371 }, { "epoch": 3.17, "learning_rate": 6.834313494863954e-05, "loss": 2.5861, "step": 40372 }, { "epoch": 3.17, "learning_rate": 6.834235081941505e-05, "loss": 2.6426, "step": 40373 }, { "epoch": 3.17, "learning_rate": 6.834156669019056e-05, "loss": 2.6847, "step": 40374 }, { "epoch": 3.17, "learning_rate": 6.834078256096606e-05, "loss": 2.7394, "step": 40375 }, { "epoch": 3.17, "learning_rate": 6.833999843174156e-05, "loss": 2.7305, "step": 40376 }, { "epoch": 3.17, "learning_rate": 6.833921430251707e-05, "loss": 2.6462, "step": 40377 }, { "epoch": 3.17, "learning_rate": 6.833843017329257e-05, "loss": 2.6181, "step": 40378 }, { "epoch": 3.17, "learning_rate": 6.833764604406808e-05, "loss": 2.9158, "step": 40379 }, { "epoch": 3.17, "learning_rate": 6.833686191484357e-05, "loss": 2.8703, "step": 40380 }, { "epoch": 3.17, "learning_rate": 6.833607778561907e-05, "loss": 2.7031, "step": 40381 }, { "epoch": 3.17, "learning_rate": 6.833529365639457e-05, "loss": 2.554, "step": 40382 }, { "epoch": 3.17, "learning_rate": 6.833450952717008e-05, "loss": 2.7638, "step": 40383 }, { "epoch": 3.17, "learning_rate": 6.833372539794558e-05, "loss": 2.6497, "step": 40384 }, { "epoch": 3.17, "learning_rate": 6.833294126872108e-05, "loss": 2.7503, "step": 40385 }, { "epoch": 3.17, "learning_rate": 6.833215713949659e-05, "loss": 2.6579, "step": 40386 }, { "epoch": 3.17, "learning_rate": 6.833137301027209e-05, "loss": 2.8508, "step": 40387 }, { "epoch": 3.17, "learning_rate": 6.83305888810476e-05, "loss": 2.7483, "step": 40388 }, { "epoch": 3.17, "learning_rate": 6.83298047518231e-05, "loss": 2.7172, "step": 40389 }, { "epoch": 3.17, "learning_rate": 6.83290206225986e-05, "loss": 2.6832, "step": 40390 }, { "epoch": 3.17, "learning_rate": 6.83282364933741e-05, "loss": 3.0974, "step": 40391 }, { "epoch": 3.17, "learning_rate": 6.832745236414961e-05, "loss": 2.7189, "step": 40392 }, { "epoch": 3.17, "learning_rate": 6.832666823492511e-05, "loss": 2.6706, "step": 40393 }, { "epoch": 3.17, "learning_rate": 6.832588410570062e-05, "loss": 2.6824, "step": 40394 }, { "epoch": 3.17, "learning_rate": 6.832509997647612e-05, "loss": 2.6458, "step": 40395 }, { "epoch": 3.17, "learning_rate": 6.832431584725162e-05, "loss": 2.8058, "step": 40396 }, { "epoch": 3.17, "learning_rate": 6.832353171802713e-05, "loss": 2.793, "step": 40397 }, { "epoch": 3.17, "learning_rate": 6.832274758880263e-05, "loss": 2.7849, "step": 40398 }, { "epoch": 3.17, "learning_rate": 6.832196345957815e-05, "loss": 2.5173, "step": 40399 }, { "epoch": 3.17, "learning_rate": 6.832117933035365e-05, "loss": 2.6495, "step": 40400 }, { "epoch": 3.17, "learning_rate": 6.832039520112916e-05, "loss": 2.9429, "step": 40401 }, { "epoch": 3.17, "learning_rate": 6.831961107190466e-05, "loss": 2.7771, "step": 40402 }, { "epoch": 3.17, "learning_rate": 6.831882694268016e-05, "loss": 2.8364, "step": 40403 }, { "epoch": 3.17, "learning_rate": 6.831804281345567e-05, "loss": 2.8196, "step": 40404 }, { "epoch": 3.17, "learning_rate": 6.831725868423117e-05, "loss": 2.8387, "step": 40405 }, { "epoch": 3.17, "learning_rate": 6.831647455500667e-05, "loss": 2.7043, "step": 40406 }, { "epoch": 3.17, "learning_rate": 6.831569042578218e-05, "loss": 2.7667, "step": 40407 }, { "epoch": 3.17, "learning_rate": 6.831490629655768e-05, "loss": 2.6864, "step": 40408 }, { "epoch": 3.17, "learning_rate": 6.831412216733319e-05, "loss": 2.7066, "step": 40409 }, { "epoch": 3.17, "learning_rate": 6.831333803810869e-05, "loss": 2.8781, "step": 40410 }, { "epoch": 3.17, "learning_rate": 6.831255390888419e-05, "loss": 2.7175, "step": 40411 }, { "epoch": 3.17, "learning_rate": 6.83117697796597e-05, "loss": 2.5863, "step": 40412 }, { "epoch": 3.17, "learning_rate": 6.83109856504352e-05, "loss": 2.6376, "step": 40413 }, { "epoch": 3.17, "learning_rate": 6.83102015212107e-05, "loss": 2.6633, "step": 40414 }, { "epoch": 3.17, "learning_rate": 6.830941739198621e-05, "loss": 2.5422, "step": 40415 }, { "epoch": 3.17, "learning_rate": 6.83086332627617e-05, "loss": 2.7562, "step": 40416 }, { "epoch": 3.17, "learning_rate": 6.83078491335372e-05, "loss": 2.7145, "step": 40417 }, { "epoch": 3.17, "learning_rate": 6.83070650043127e-05, "loss": 2.5661, "step": 40418 }, { "epoch": 3.17, "learning_rate": 6.830628087508821e-05, "loss": 2.8459, "step": 40419 }, { "epoch": 3.17, "learning_rate": 6.830549674586371e-05, "loss": 2.5222, "step": 40420 }, { "epoch": 3.17, "learning_rate": 6.830471261663922e-05, "loss": 2.87, "step": 40421 }, { "epoch": 3.17, "learning_rate": 6.830392848741472e-05, "loss": 2.6428, "step": 40422 }, { "epoch": 3.17, "learning_rate": 6.830314435819022e-05, "loss": 2.8134, "step": 40423 }, { "epoch": 3.17, "learning_rate": 6.830236022896573e-05, "loss": 2.6055, "step": 40424 }, { "epoch": 3.17, "learning_rate": 6.830157609974124e-05, "loss": 2.654, "step": 40425 }, { "epoch": 3.17, "learning_rate": 6.830079197051675e-05, "loss": 2.8126, "step": 40426 }, { "epoch": 3.17, "learning_rate": 6.830000784129225e-05, "loss": 2.7621, "step": 40427 }, { "epoch": 3.17, "learning_rate": 6.829922371206776e-05, "loss": 2.7059, "step": 40428 }, { "epoch": 3.17, "learning_rate": 6.829843958284326e-05, "loss": 2.7053, "step": 40429 }, { "epoch": 3.17, "learning_rate": 6.829765545361876e-05, "loss": 2.7551, "step": 40430 }, { "epoch": 3.17, "learning_rate": 6.829687132439427e-05, "loss": 2.7347, "step": 40431 }, { "epoch": 3.17, "learning_rate": 6.829608719516977e-05, "loss": 2.9333, "step": 40432 }, { "epoch": 3.17, "learning_rate": 6.829530306594527e-05, "loss": 2.7035, "step": 40433 }, { "epoch": 3.17, "learning_rate": 6.829451893672078e-05, "loss": 2.6291, "step": 40434 }, { "epoch": 3.17, "learning_rate": 6.829373480749628e-05, "loss": 2.6978, "step": 40435 }, { "epoch": 3.17, "learning_rate": 6.829295067827178e-05, "loss": 2.6254, "step": 40436 }, { "epoch": 3.17, "learning_rate": 6.829216654904729e-05, "loss": 2.7342, "step": 40437 }, { "epoch": 3.17, "learning_rate": 6.829138241982279e-05, "loss": 2.8988, "step": 40438 }, { "epoch": 3.17, "learning_rate": 6.82905982905983e-05, "loss": 2.6353, "step": 40439 }, { "epoch": 3.17, "learning_rate": 6.82898141613738e-05, "loss": 2.8691, "step": 40440 }, { "epoch": 3.17, "learning_rate": 6.82890300321493e-05, "loss": 2.7405, "step": 40441 }, { "epoch": 3.17, "learning_rate": 6.828824590292481e-05, "loss": 2.6969, "step": 40442 }, { "epoch": 3.17, "learning_rate": 6.828746177370031e-05, "loss": 2.6842, "step": 40443 }, { "epoch": 3.17, "learning_rate": 6.828667764447581e-05, "loss": 2.8788, "step": 40444 }, { "epoch": 3.17, "learning_rate": 6.828589351525132e-05, "loss": 3.0581, "step": 40445 }, { "epoch": 3.17, "learning_rate": 6.828510938602682e-05, "loss": 2.669, "step": 40446 }, { "epoch": 3.17, "learning_rate": 6.828432525680233e-05, "loss": 2.5408, "step": 40447 }, { "epoch": 3.17, "learning_rate": 6.828354112757783e-05, "loss": 2.6533, "step": 40448 }, { "epoch": 3.17, "learning_rate": 6.828275699835333e-05, "loss": 2.4828, "step": 40449 }, { "epoch": 3.17, "learning_rate": 6.828197286912884e-05, "loss": 2.8659, "step": 40450 }, { "epoch": 3.17, "learning_rate": 6.828118873990434e-05, "loss": 2.5105, "step": 40451 }, { "epoch": 3.17, "learning_rate": 6.828040461067984e-05, "loss": 2.724, "step": 40452 }, { "epoch": 3.17, "learning_rate": 6.827962048145535e-05, "loss": 2.5978, "step": 40453 }, { "epoch": 3.17, "learning_rate": 6.827883635223085e-05, "loss": 2.8305, "step": 40454 }, { "epoch": 3.17, "learning_rate": 6.827805222300635e-05, "loss": 2.7074, "step": 40455 }, { "epoch": 3.17, "learning_rate": 6.827726809378186e-05, "loss": 2.7927, "step": 40456 }, { "epoch": 3.17, "learning_rate": 6.827648396455736e-05, "loss": 2.4198, "step": 40457 }, { "epoch": 3.17, "learning_rate": 6.827569983533287e-05, "loss": 2.6563, "step": 40458 }, { "epoch": 3.17, "learning_rate": 6.827491570610837e-05, "loss": 2.6495, "step": 40459 }, { "epoch": 3.17, "learning_rate": 6.827413157688387e-05, "loss": 2.583, "step": 40460 }, { "epoch": 3.17, "learning_rate": 6.827334744765938e-05, "loss": 2.817, "step": 40461 }, { "epoch": 3.17, "learning_rate": 6.827256331843488e-05, "loss": 2.7103, "step": 40462 }, { "epoch": 3.17, "learning_rate": 6.827177918921038e-05, "loss": 2.7222, "step": 40463 }, { "epoch": 3.17, "learning_rate": 6.827099505998589e-05, "loss": 2.8228, "step": 40464 }, { "epoch": 3.17, "learning_rate": 6.827021093076139e-05, "loss": 2.6983, "step": 40465 }, { "epoch": 3.17, "learning_rate": 6.82694268015369e-05, "loss": 2.4978, "step": 40466 }, { "epoch": 3.17, "learning_rate": 6.82686426723124e-05, "loss": 2.9797, "step": 40467 }, { "epoch": 3.17, "learning_rate": 6.82678585430879e-05, "loss": 2.6618, "step": 40468 }, { "epoch": 3.17, "learning_rate": 6.82670744138634e-05, "loss": 2.7921, "step": 40469 }, { "epoch": 3.17, "learning_rate": 6.826629028463891e-05, "loss": 2.7097, "step": 40470 }, { "epoch": 3.17, "learning_rate": 6.826550615541441e-05, "loss": 2.8577, "step": 40471 }, { "epoch": 3.17, "learning_rate": 6.826472202618992e-05, "loss": 2.5855, "step": 40472 }, { "epoch": 3.17, "learning_rate": 6.826393789696542e-05, "loss": 2.8026, "step": 40473 }, { "epoch": 3.17, "learning_rate": 6.826315376774092e-05, "loss": 2.9886, "step": 40474 }, { "epoch": 3.17, "learning_rate": 6.826236963851643e-05, "loss": 2.4869, "step": 40475 }, { "epoch": 3.17, "learning_rate": 6.826158550929193e-05, "loss": 2.5823, "step": 40476 }, { "epoch": 3.17, "learning_rate": 6.826080138006744e-05, "loss": 2.6399, "step": 40477 }, { "epoch": 3.17, "learning_rate": 6.826001725084294e-05, "loss": 2.56, "step": 40478 }, { "epoch": 3.17, "learning_rate": 6.825923312161844e-05, "loss": 2.677, "step": 40479 }, { "epoch": 3.17, "learning_rate": 6.825844899239395e-05, "loss": 2.7524, "step": 40480 }, { "epoch": 3.17, "learning_rate": 6.825766486316945e-05, "loss": 2.735, "step": 40481 }, { "epoch": 3.17, "learning_rate": 6.825688073394497e-05, "loss": 2.8482, "step": 40482 }, { "epoch": 3.17, "learning_rate": 6.825609660472047e-05, "loss": 2.7403, "step": 40483 }, { "epoch": 3.17, "learning_rate": 6.825531247549597e-05, "loss": 2.5287, "step": 40484 }, { "epoch": 3.17, "learning_rate": 6.825452834627148e-05, "loss": 2.588, "step": 40485 }, { "epoch": 3.17, "learning_rate": 6.825374421704698e-05, "loss": 2.6871, "step": 40486 }, { "epoch": 3.17, "learning_rate": 6.825296008782249e-05, "loss": 2.7001, "step": 40487 }, { "epoch": 3.17, "learning_rate": 6.825217595859798e-05, "loss": 2.7157, "step": 40488 }, { "epoch": 3.17, "learning_rate": 6.825139182937348e-05, "loss": 2.4664, "step": 40489 }, { "epoch": 3.17, "learning_rate": 6.825060770014898e-05, "loss": 2.9431, "step": 40490 }, { "epoch": 3.18, "learning_rate": 6.824982357092449e-05, "loss": 2.7425, "step": 40491 }, { "epoch": 3.18, "learning_rate": 6.824903944169999e-05, "loss": 2.6216, "step": 40492 }, { "epoch": 3.18, "learning_rate": 6.82482553124755e-05, "loss": 2.6322, "step": 40493 }, { "epoch": 3.18, "learning_rate": 6.8247471183251e-05, "loss": 2.5811, "step": 40494 }, { "epoch": 3.18, "learning_rate": 6.82466870540265e-05, "loss": 2.7151, "step": 40495 }, { "epoch": 3.18, "learning_rate": 6.8245902924802e-05, "loss": 2.484, "step": 40496 }, { "epoch": 3.18, "learning_rate": 6.824511879557751e-05, "loss": 2.5624, "step": 40497 }, { "epoch": 3.18, "learning_rate": 6.824433466635301e-05, "loss": 2.8085, "step": 40498 }, { "epoch": 3.18, "learning_rate": 6.824355053712852e-05, "loss": 2.5883, "step": 40499 }, { "epoch": 3.18, "learning_rate": 6.824276640790402e-05, "loss": 2.684, "step": 40500 }, { "epoch": 3.18, "learning_rate": 6.824198227867952e-05, "loss": 2.7512, "step": 40501 }, { "epoch": 3.18, "learning_rate": 6.824119814945503e-05, "loss": 2.7387, "step": 40502 }, { "epoch": 3.18, "learning_rate": 6.824041402023053e-05, "loss": 2.6965, "step": 40503 }, { "epoch": 3.18, "learning_rate": 6.823962989100603e-05, "loss": 2.8549, "step": 40504 }, { "epoch": 3.18, "learning_rate": 6.823884576178154e-05, "loss": 2.8242, "step": 40505 }, { "epoch": 3.18, "learning_rate": 6.823806163255704e-05, "loss": 2.6741, "step": 40506 }, { "epoch": 3.18, "learning_rate": 6.823727750333256e-05, "loss": 2.7027, "step": 40507 }, { "epoch": 3.18, "learning_rate": 6.823649337410806e-05, "loss": 2.6181, "step": 40508 }, { "epoch": 3.18, "learning_rate": 6.823570924488357e-05, "loss": 2.6982, "step": 40509 }, { "epoch": 3.18, "learning_rate": 6.823492511565907e-05, "loss": 2.8613, "step": 40510 }, { "epoch": 3.18, "learning_rate": 6.823414098643457e-05, "loss": 2.6183, "step": 40511 }, { "epoch": 3.18, "learning_rate": 6.823335685721008e-05, "loss": 2.8721, "step": 40512 }, { "epoch": 3.18, "learning_rate": 6.823257272798558e-05, "loss": 2.8255, "step": 40513 }, { "epoch": 3.18, "learning_rate": 6.823178859876109e-05, "loss": 2.7209, "step": 40514 }, { "epoch": 3.18, "learning_rate": 6.823100446953659e-05, "loss": 2.8786, "step": 40515 }, { "epoch": 3.18, "learning_rate": 6.823022034031209e-05, "loss": 2.695, "step": 40516 }, { "epoch": 3.18, "learning_rate": 6.82294362110876e-05, "loss": 2.4981, "step": 40517 }, { "epoch": 3.18, "learning_rate": 6.82286520818631e-05, "loss": 2.7537, "step": 40518 }, { "epoch": 3.18, "learning_rate": 6.82278679526386e-05, "loss": 2.7134, "step": 40519 }, { "epoch": 3.18, "learning_rate": 6.822708382341411e-05, "loss": 2.7368, "step": 40520 }, { "epoch": 3.18, "learning_rate": 6.822629969418961e-05, "loss": 2.6749, "step": 40521 }, { "epoch": 3.18, "learning_rate": 6.822551556496511e-05, "loss": 2.7653, "step": 40522 }, { "epoch": 3.18, "learning_rate": 6.822473143574062e-05, "loss": 2.8009, "step": 40523 }, { "epoch": 3.18, "learning_rate": 6.822394730651611e-05, "loss": 2.7237, "step": 40524 }, { "epoch": 3.18, "learning_rate": 6.822316317729161e-05, "loss": 2.5615, "step": 40525 }, { "epoch": 3.18, "learning_rate": 6.822237904806712e-05, "loss": 2.6789, "step": 40526 }, { "epoch": 3.18, "learning_rate": 6.822159491884262e-05, "loss": 2.6806, "step": 40527 }, { "epoch": 3.18, "learning_rate": 6.822081078961812e-05, "loss": 2.8347, "step": 40528 }, { "epoch": 3.18, "learning_rate": 6.822002666039363e-05, "loss": 2.605, "step": 40529 }, { "epoch": 3.18, "learning_rate": 6.821924253116913e-05, "loss": 2.7436, "step": 40530 }, { "epoch": 3.18, "learning_rate": 6.821845840194463e-05, "loss": 2.6397, "step": 40531 }, { "epoch": 3.18, "learning_rate": 6.821767427272014e-05, "loss": 2.8474, "step": 40532 }, { "epoch": 3.18, "learning_rate": 6.821689014349566e-05, "loss": 2.7069, "step": 40533 }, { "epoch": 3.18, "learning_rate": 6.821610601427116e-05, "loss": 2.5368, "step": 40534 }, { "epoch": 3.18, "learning_rate": 6.821532188504666e-05, "loss": 2.8829, "step": 40535 }, { "epoch": 3.18, "learning_rate": 6.821453775582217e-05, "loss": 2.8337, "step": 40536 }, { "epoch": 3.18, "learning_rate": 6.821375362659767e-05, "loss": 2.7709, "step": 40537 }, { "epoch": 3.18, "learning_rate": 6.821296949737317e-05, "loss": 2.6214, "step": 40538 }, { "epoch": 3.18, "learning_rate": 6.821218536814868e-05, "loss": 2.8279, "step": 40539 }, { "epoch": 3.18, "learning_rate": 6.821140123892418e-05, "loss": 2.4711, "step": 40540 }, { "epoch": 3.18, "learning_rate": 6.821061710969968e-05, "loss": 2.6218, "step": 40541 }, { "epoch": 3.18, "learning_rate": 6.820983298047519e-05, "loss": 2.4073, "step": 40542 }, { "epoch": 3.18, "learning_rate": 6.820904885125069e-05, "loss": 2.6062, "step": 40543 }, { "epoch": 3.18, "learning_rate": 6.82082647220262e-05, "loss": 2.8819, "step": 40544 }, { "epoch": 3.18, "learning_rate": 6.82074805928017e-05, "loss": 2.6628, "step": 40545 }, { "epoch": 3.18, "learning_rate": 6.82066964635772e-05, "loss": 2.7252, "step": 40546 }, { "epoch": 3.18, "learning_rate": 6.82059123343527e-05, "loss": 2.8169, "step": 40547 }, { "epoch": 3.18, "learning_rate": 6.820512820512821e-05, "loss": 2.7128, "step": 40548 }, { "epoch": 3.18, "learning_rate": 6.820434407590371e-05, "loss": 2.7017, "step": 40549 }, { "epoch": 3.18, "learning_rate": 6.820355994667922e-05, "loss": 2.8042, "step": 40550 }, { "epoch": 3.18, "learning_rate": 6.820277581745472e-05, "loss": 2.8012, "step": 40551 }, { "epoch": 3.18, "learning_rate": 6.820199168823022e-05, "loss": 2.597, "step": 40552 }, { "epoch": 3.18, "learning_rate": 6.820120755900573e-05, "loss": 2.6596, "step": 40553 }, { "epoch": 3.18, "learning_rate": 6.820042342978123e-05, "loss": 2.7902, "step": 40554 }, { "epoch": 3.18, "learning_rate": 6.819963930055674e-05, "loss": 2.7997, "step": 40555 }, { "epoch": 3.18, "learning_rate": 6.819885517133224e-05, "loss": 2.6594, "step": 40556 }, { "epoch": 3.18, "learning_rate": 6.819807104210774e-05, "loss": 2.9221, "step": 40557 }, { "epoch": 3.18, "learning_rate": 6.819728691288325e-05, "loss": 2.829, "step": 40558 }, { "epoch": 3.18, "learning_rate": 6.819650278365875e-05, "loss": 2.9116, "step": 40559 }, { "epoch": 3.18, "learning_rate": 6.819571865443425e-05, "loss": 2.6252, "step": 40560 }, { "epoch": 3.18, "learning_rate": 6.819493452520976e-05, "loss": 2.8518, "step": 40561 }, { "epoch": 3.18, "learning_rate": 6.819415039598526e-05, "loss": 2.6471, "step": 40562 }, { "epoch": 3.18, "learning_rate": 6.819336626676077e-05, "loss": 2.5264, "step": 40563 }, { "epoch": 3.18, "learning_rate": 6.819258213753627e-05, "loss": 2.7485, "step": 40564 }, { "epoch": 3.18, "learning_rate": 6.819179800831177e-05, "loss": 2.7975, "step": 40565 }, { "epoch": 3.18, "learning_rate": 6.819101387908728e-05, "loss": 2.6474, "step": 40566 }, { "epoch": 3.18, "learning_rate": 6.819022974986278e-05, "loss": 2.8326, "step": 40567 }, { "epoch": 3.18, "learning_rate": 6.818944562063828e-05, "loss": 2.731, "step": 40568 }, { "epoch": 3.18, "learning_rate": 6.818866149141379e-05, "loss": 2.9785, "step": 40569 }, { "epoch": 3.18, "learning_rate": 6.818787736218929e-05, "loss": 2.4698, "step": 40570 }, { "epoch": 3.18, "learning_rate": 6.81870932329648e-05, "loss": 2.8437, "step": 40571 }, { "epoch": 3.18, "learning_rate": 6.81863091037403e-05, "loss": 2.899, "step": 40572 }, { "epoch": 3.18, "learning_rate": 6.81855249745158e-05, "loss": 2.5823, "step": 40573 }, { "epoch": 3.18, "learning_rate": 6.81847408452913e-05, "loss": 2.9745, "step": 40574 }, { "epoch": 3.18, "learning_rate": 6.818395671606681e-05, "loss": 2.7177, "step": 40575 }, { "epoch": 3.18, "learning_rate": 6.818317258684231e-05, "loss": 2.706, "step": 40576 }, { "epoch": 3.18, "learning_rate": 6.818238845761782e-05, "loss": 2.7296, "step": 40577 }, { "epoch": 3.18, "learning_rate": 6.818160432839332e-05, "loss": 2.7485, "step": 40578 }, { "epoch": 3.18, "learning_rate": 6.818082019916882e-05, "loss": 2.662, "step": 40579 }, { "epoch": 3.18, "learning_rate": 6.818003606994433e-05, "loss": 2.8116, "step": 40580 }, { "epoch": 3.18, "learning_rate": 6.817925194071983e-05, "loss": 2.6491, "step": 40581 }, { "epoch": 3.18, "learning_rate": 6.817846781149534e-05, "loss": 2.4601, "step": 40582 }, { "epoch": 3.18, "learning_rate": 6.817768368227084e-05, "loss": 2.6687, "step": 40583 }, { "epoch": 3.18, "learning_rate": 6.817689955304634e-05, "loss": 2.6573, "step": 40584 }, { "epoch": 3.18, "learning_rate": 6.817611542382185e-05, "loss": 2.7099, "step": 40585 }, { "epoch": 3.18, "learning_rate": 6.817533129459735e-05, "loss": 2.6594, "step": 40586 }, { "epoch": 3.18, "learning_rate": 6.817454716537285e-05, "loss": 2.6326, "step": 40587 }, { "epoch": 3.18, "learning_rate": 6.817376303614836e-05, "loss": 2.6749, "step": 40588 }, { "epoch": 3.18, "learning_rate": 6.817297890692386e-05, "loss": 2.5754, "step": 40589 }, { "epoch": 3.18, "learning_rate": 6.817219477769938e-05, "loss": 2.5308, "step": 40590 }, { "epoch": 3.18, "learning_rate": 6.817141064847488e-05, "loss": 2.864, "step": 40591 }, { "epoch": 3.18, "learning_rate": 6.817062651925039e-05, "loss": 2.884, "step": 40592 }, { "epoch": 3.18, "learning_rate": 6.816984239002589e-05, "loss": 2.8805, "step": 40593 }, { "epoch": 3.18, "learning_rate": 6.816905826080139e-05, "loss": 2.5925, "step": 40594 }, { "epoch": 3.18, "learning_rate": 6.81682741315769e-05, "loss": 2.7347, "step": 40595 }, { "epoch": 3.18, "learning_rate": 6.816749000235239e-05, "loss": 2.7527, "step": 40596 }, { "epoch": 3.18, "learning_rate": 6.816670587312789e-05, "loss": 2.6352, "step": 40597 }, { "epoch": 3.18, "learning_rate": 6.81659217439034e-05, "loss": 2.6398, "step": 40598 }, { "epoch": 3.18, "learning_rate": 6.81651376146789e-05, "loss": 2.7663, "step": 40599 }, { "epoch": 3.18, "learning_rate": 6.81643534854544e-05, "loss": 2.6885, "step": 40600 }, { "epoch": 3.18, "learning_rate": 6.81635693562299e-05, "loss": 2.5044, "step": 40601 }, { "epoch": 3.18, "learning_rate": 6.816278522700541e-05, "loss": 2.6376, "step": 40602 }, { "epoch": 3.18, "learning_rate": 6.816200109778091e-05, "loss": 2.6707, "step": 40603 }, { "epoch": 3.18, "learning_rate": 6.816121696855642e-05, "loss": 2.6963, "step": 40604 }, { "epoch": 3.18, "learning_rate": 6.816043283933192e-05, "loss": 2.7457, "step": 40605 }, { "epoch": 3.18, "learning_rate": 6.815964871010742e-05, "loss": 2.5727, "step": 40606 }, { "epoch": 3.18, "learning_rate": 6.815886458088293e-05, "loss": 2.6356, "step": 40607 }, { "epoch": 3.18, "learning_rate": 6.815808045165843e-05, "loss": 2.7792, "step": 40608 }, { "epoch": 3.18, "learning_rate": 6.815729632243393e-05, "loss": 2.8559, "step": 40609 }, { "epoch": 3.18, "learning_rate": 6.815651219320944e-05, "loss": 2.5648, "step": 40610 }, { "epoch": 3.18, "learning_rate": 6.815572806398494e-05, "loss": 2.6476, "step": 40611 }, { "epoch": 3.18, "learning_rate": 6.815494393476045e-05, "loss": 2.7803, "step": 40612 }, { "epoch": 3.18, "learning_rate": 6.815415980553595e-05, "loss": 2.7165, "step": 40613 }, { "epoch": 3.18, "learning_rate": 6.815337567631145e-05, "loss": 2.7144, "step": 40614 }, { "epoch": 3.18, "learning_rate": 6.815259154708697e-05, "loss": 2.8896, "step": 40615 }, { "epoch": 3.18, "learning_rate": 6.815180741786247e-05, "loss": 2.7875, "step": 40616 }, { "epoch": 3.18, "learning_rate": 6.815102328863798e-05, "loss": 2.9077, "step": 40617 }, { "epoch": 3.18, "learning_rate": 6.815023915941348e-05, "loss": 2.7204, "step": 40618 }, { "epoch": 3.19, "learning_rate": 6.814945503018898e-05, "loss": 2.7803, "step": 40619 }, { "epoch": 3.19, "learning_rate": 6.814867090096449e-05, "loss": 2.6185, "step": 40620 }, { "epoch": 3.19, "learning_rate": 6.814788677173999e-05, "loss": 2.6252, "step": 40621 }, { "epoch": 3.19, "learning_rate": 6.81471026425155e-05, "loss": 2.7476, "step": 40622 }, { "epoch": 3.19, "learning_rate": 6.8146318513291e-05, "loss": 2.8375, "step": 40623 }, { "epoch": 3.19, "learning_rate": 6.81455343840665e-05, "loss": 2.7668, "step": 40624 }, { "epoch": 3.19, "learning_rate": 6.814475025484201e-05, "loss": 2.5718, "step": 40625 }, { "epoch": 3.19, "learning_rate": 6.814396612561751e-05, "loss": 2.6229, "step": 40626 }, { "epoch": 3.19, "learning_rate": 6.814318199639301e-05, "loss": 2.6187, "step": 40627 }, { "epoch": 3.19, "learning_rate": 6.814239786716852e-05, "loss": 2.6906, "step": 40628 }, { "epoch": 3.19, "learning_rate": 6.814161373794402e-05, "loss": 2.767, "step": 40629 }, { "epoch": 3.19, "learning_rate": 6.814082960871953e-05, "loss": 2.7163, "step": 40630 }, { "epoch": 3.19, "learning_rate": 6.814004547949503e-05, "loss": 2.7684, "step": 40631 }, { "epoch": 3.19, "learning_rate": 6.813926135027052e-05, "loss": 2.7875, "step": 40632 }, { "epoch": 3.19, "learning_rate": 6.813847722104602e-05, "loss": 2.8218, "step": 40633 }, { "epoch": 3.19, "learning_rate": 6.813769309182153e-05, "loss": 2.8692, "step": 40634 }, { "epoch": 3.19, "learning_rate": 6.813690896259703e-05, "loss": 2.8627, "step": 40635 }, { "epoch": 3.19, "learning_rate": 6.813612483337253e-05, "loss": 2.7492, "step": 40636 }, { "epoch": 3.19, "learning_rate": 6.813534070414804e-05, "loss": 2.7765, "step": 40637 }, { "epoch": 3.19, "learning_rate": 6.813455657492354e-05, "loss": 2.8449, "step": 40638 }, { "epoch": 3.19, "learning_rate": 6.813377244569904e-05, "loss": 2.6532, "step": 40639 }, { "epoch": 3.19, "learning_rate": 6.813298831647455e-05, "loss": 2.619, "step": 40640 }, { "epoch": 3.19, "learning_rate": 6.813220418725007e-05, "loss": 2.7848, "step": 40641 }, { "epoch": 3.19, "learning_rate": 6.813142005802557e-05, "loss": 2.767, "step": 40642 }, { "epoch": 3.19, "learning_rate": 6.813063592880107e-05, "loss": 2.5618, "step": 40643 }, { "epoch": 3.19, "learning_rate": 6.812985179957658e-05, "loss": 2.688, "step": 40644 }, { "epoch": 3.19, "learning_rate": 6.812906767035208e-05, "loss": 2.8756, "step": 40645 }, { "epoch": 3.19, "learning_rate": 6.812828354112758e-05, "loss": 2.7846, "step": 40646 }, { "epoch": 3.19, "learning_rate": 6.812749941190309e-05, "loss": 2.7593, "step": 40647 }, { "epoch": 3.19, "learning_rate": 6.812671528267859e-05, "loss": 2.7221, "step": 40648 }, { "epoch": 3.19, "learning_rate": 6.81259311534541e-05, "loss": 2.8647, "step": 40649 }, { "epoch": 3.19, "learning_rate": 6.81251470242296e-05, "loss": 2.7899, "step": 40650 }, { "epoch": 3.19, "learning_rate": 6.81243628950051e-05, "loss": 2.5537, "step": 40651 }, { "epoch": 3.19, "learning_rate": 6.81235787657806e-05, "loss": 2.8614, "step": 40652 }, { "epoch": 3.19, "learning_rate": 6.812279463655611e-05, "loss": 2.5371, "step": 40653 }, { "epoch": 3.19, "learning_rate": 6.812201050733161e-05, "loss": 2.923, "step": 40654 }, { "epoch": 3.19, "learning_rate": 6.812122637810712e-05, "loss": 2.7913, "step": 40655 }, { "epoch": 3.19, "learning_rate": 6.812044224888262e-05, "loss": 2.6388, "step": 40656 }, { "epoch": 3.19, "learning_rate": 6.811965811965812e-05, "loss": 2.548, "step": 40657 }, { "epoch": 3.19, "learning_rate": 6.811887399043363e-05, "loss": 2.717, "step": 40658 }, { "epoch": 3.19, "learning_rate": 6.811808986120913e-05, "loss": 2.6687, "step": 40659 }, { "epoch": 3.19, "learning_rate": 6.811730573198464e-05, "loss": 2.555, "step": 40660 }, { "epoch": 3.19, "learning_rate": 6.811652160276014e-05, "loss": 2.7685, "step": 40661 }, { "epoch": 3.19, "learning_rate": 6.811573747353564e-05, "loss": 2.6816, "step": 40662 }, { "epoch": 3.19, "learning_rate": 6.811495334431115e-05, "loss": 2.8378, "step": 40663 }, { "epoch": 3.19, "learning_rate": 6.811416921508665e-05, "loss": 2.6291, "step": 40664 }, { "epoch": 3.19, "learning_rate": 6.811338508586215e-05, "loss": 2.6456, "step": 40665 }, { "epoch": 3.19, "learning_rate": 6.811260095663766e-05, "loss": 3.0714, "step": 40666 }, { "epoch": 3.19, "learning_rate": 6.811181682741316e-05, "loss": 2.7656, "step": 40667 }, { "epoch": 3.19, "learning_rate": 6.811103269818867e-05, "loss": 2.6955, "step": 40668 }, { "epoch": 3.19, "learning_rate": 6.811024856896417e-05, "loss": 2.6648, "step": 40669 }, { "epoch": 3.19, "learning_rate": 6.810946443973967e-05, "loss": 2.7671, "step": 40670 }, { "epoch": 3.19, "learning_rate": 6.810868031051518e-05, "loss": 2.7122, "step": 40671 }, { "epoch": 3.19, "learning_rate": 6.810789618129068e-05, "loss": 2.8162, "step": 40672 }, { "epoch": 3.19, "learning_rate": 6.810711205206618e-05, "loss": 2.8527, "step": 40673 }, { "epoch": 3.19, "learning_rate": 6.810632792284169e-05, "loss": 2.8494, "step": 40674 }, { "epoch": 3.19, "learning_rate": 6.810554379361719e-05, "loss": 2.5982, "step": 40675 }, { "epoch": 3.19, "learning_rate": 6.81047596643927e-05, "loss": 2.6813, "step": 40676 }, { "epoch": 3.19, "learning_rate": 6.81039755351682e-05, "loss": 2.7644, "step": 40677 }, { "epoch": 3.19, "learning_rate": 6.81031914059437e-05, "loss": 2.7022, "step": 40678 }, { "epoch": 3.19, "learning_rate": 6.81024072767192e-05, "loss": 2.7273, "step": 40679 }, { "epoch": 3.19, "learning_rate": 6.810162314749471e-05, "loss": 2.6891, "step": 40680 }, { "epoch": 3.19, "learning_rate": 6.810083901827021e-05, "loss": 2.7442, "step": 40681 }, { "epoch": 3.19, "learning_rate": 6.810005488904572e-05, "loss": 2.6598, "step": 40682 }, { "epoch": 3.19, "learning_rate": 6.809927075982122e-05, "loss": 2.6483, "step": 40683 }, { "epoch": 3.19, "learning_rate": 6.809848663059672e-05, "loss": 2.782, "step": 40684 }, { "epoch": 3.19, "learning_rate": 6.809770250137223e-05, "loss": 2.7178, "step": 40685 }, { "epoch": 3.19, "learning_rate": 6.809691837214773e-05, "loss": 2.7287, "step": 40686 }, { "epoch": 3.19, "learning_rate": 6.809613424292324e-05, "loss": 2.733, "step": 40687 }, { "epoch": 3.19, "learning_rate": 6.809535011369874e-05, "loss": 2.5889, "step": 40688 }, { "epoch": 3.19, "learning_rate": 6.809456598447424e-05, "loss": 2.7911, "step": 40689 }, { "epoch": 3.19, "learning_rate": 6.809378185524975e-05, "loss": 2.7416, "step": 40690 }, { "epoch": 3.19, "learning_rate": 6.809299772602525e-05, "loss": 2.5711, "step": 40691 }, { "epoch": 3.19, "learning_rate": 6.809221359680075e-05, "loss": 2.8501, "step": 40692 }, { "epoch": 3.19, "learning_rate": 6.809142946757626e-05, "loss": 2.9767, "step": 40693 }, { "epoch": 3.19, "learning_rate": 6.809064533835176e-05, "loss": 2.6213, "step": 40694 }, { "epoch": 3.19, "learning_rate": 6.808986120912726e-05, "loss": 2.6447, "step": 40695 }, { "epoch": 3.19, "learning_rate": 6.808907707990277e-05, "loss": 2.7413, "step": 40696 }, { "epoch": 3.19, "learning_rate": 6.808829295067827e-05, "loss": 2.7005, "step": 40697 }, { "epoch": 3.19, "learning_rate": 6.808750882145379e-05, "loss": 2.6697, "step": 40698 }, { "epoch": 3.19, "learning_rate": 6.808672469222929e-05, "loss": 2.9034, "step": 40699 }, { "epoch": 3.19, "learning_rate": 6.80859405630048e-05, "loss": 2.7417, "step": 40700 }, { "epoch": 3.19, "learning_rate": 6.80851564337803e-05, "loss": 2.7084, "step": 40701 }, { "epoch": 3.19, "learning_rate": 6.80843723045558e-05, "loss": 2.8224, "step": 40702 }, { "epoch": 3.19, "learning_rate": 6.808358817533131e-05, "loss": 2.6597, "step": 40703 }, { "epoch": 3.19, "learning_rate": 6.80828040461068e-05, "loss": 2.7295, "step": 40704 }, { "epoch": 3.19, "learning_rate": 6.80820199168823e-05, "loss": 2.7197, "step": 40705 }, { "epoch": 3.19, "learning_rate": 6.80812357876578e-05, "loss": 2.7643, "step": 40706 }, { "epoch": 3.19, "learning_rate": 6.808045165843331e-05, "loss": 2.7093, "step": 40707 }, { "epoch": 3.19, "learning_rate": 6.807966752920881e-05, "loss": 2.719, "step": 40708 }, { "epoch": 3.19, "learning_rate": 6.807888339998432e-05, "loss": 2.8261, "step": 40709 }, { "epoch": 3.19, "learning_rate": 6.807809927075982e-05, "loss": 2.7133, "step": 40710 }, { "epoch": 3.19, "learning_rate": 6.807731514153532e-05, "loss": 2.7398, "step": 40711 }, { "epoch": 3.19, "learning_rate": 6.807653101231083e-05, "loss": 2.5116, "step": 40712 }, { "epoch": 3.19, "learning_rate": 6.807574688308633e-05, "loss": 2.8611, "step": 40713 }, { "epoch": 3.19, "learning_rate": 6.807496275386183e-05, "loss": 2.6818, "step": 40714 }, { "epoch": 3.19, "learning_rate": 6.807417862463734e-05, "loss": 2.7045, "step": 40715 }, { "epoch": 3.19, "learning_rate": 6.807339449541284e-05, "loss": 2.8914, "step": 40716 }, { "epoch": 3.19, "learning_rate": 6.807261036618835e-05, "loss": 2.6259, "step": 40717 }, { "epoch": 3.19, "learning_rate": 6.807182623696385e-05, "loss": 2.6448, "step": 40718 }, { "epoch": 3.19, "learning_rate": 6.807104210773935e-05, "loss": 2.7725, "step": 40719 }, { "epoch": 3.19, "learning_rate": 6.807025797851486e-05, "loss": 2.7896, "step": 40720 }, { "epoch": 3.19, "learning_rate": 6.806947384929036e-05, "loss": 2.7222, "step": 40721 }, { "epoch": 3.19, "learning_rate": 6.806868972006586e-05, "loss": 2.8282, "step": 40722 }, { "epoch": 3.19, "learning_rate": 6.806790559084138e-05, "loss": 2.6062, "step": 40723 }, { "epoch": 3.19, "learning_rate": 6.806712146161688e-05, "loss": 2.5936, "step": 40724 }, { "epoch": 3.19, "learning_rate": 6.806633733239239e-05, "loss": 2.9049, "step": 40725 }, { "epoch": 3.19, "learning_rate": 6.806555320316789e-05, "loss": 2.7617, "step": 40726 }, { "epoch": 3.19, "learning_rate": 6.80647690739434e-05, "loss": 2.8359, "step": 40727 }, { "epoch": 3.19, "learning_rate": 6.80639849447189e-05, "loss": 2.6939, "step": 40728 }, { "epoch": 3.19, "learning_rate": 6.80632008154944e-05, "loss": 2.7793, "step": 40729 }, { "epoch": 3.19, "learning_rate": 6.80624166862699e-05, "loss": 2.876, "step": 40730 }, { "epoch": 3.19, "learning_rate": 6.806163255704541e-05, "loss": 2.7289, "step": 40731 }, { "epoch": 3.19, "learning_rate": 6.806084842782091e-05, "loss": 2.7681, "step": 40732 }, { "epoch": 3.19, "learning_rate": 6.806006429859642e-05, "loss": 2.7774, "step": 40733 }, { "epoch": 3.19, "learning_rate": 6.805928016937192e-05, "loss": 2.8304, "step": 40734 }, { "epoch": 3.19, "learning_rate": 6.805849604014743e-05, "loss": 2.589, "step": 40735 }, { "epoch": 3.19, "learning_rate": 6.805771191092293e-05, "loss": 2.6953, "step": 40736 }, { "epoch": 3.19, "learning_rate": 6.805692778169843e-05, "loss": 2.622, "step": 40737 }, { "epoch": 3.19, "learning_rate": 6.805614365247394e-05, "loss": 2.6566, "step": 40738 }, { "epoch": 3.19, "learning_rate": 6.805535952324944e-05, "loss": 2.849, "step": 40739 }, { "epoch": 3.19, "learning_rate": 6.805457539402493e-05, "loss": 2.6972, "step": 40740 }, { "epoch": 3.19, "learning_rate": 6.805379126480043e-05, "loss": 2.8065, "step": 40741 }, { "epoch": 3.19, "learning_rate": 6.805300713557594e-05, "loss": 2.6931, "step": 40742 }, { "epoch": 3.19, "learning_rate": 6.805222300635144e-05, "loss": 2.5348, "step": 40743 }, { "epoch": 3.19, "learning_rate": 6.805143887712694e-05, "loss": 2.6856, "step": 40744 }, { "epoch": 3.19, "learning_rate": 6.805065474790245e-05, "loss": 2.6462, "step": 40745 }, { "epoch": 3.2, "learning_rate": 6.804987061867795e-05, "loss": 2.6269, "step": 40746 }, { "epoch": 3.2, "learning_rate": 6.804908648945346e-05, "loss": 2.7244, "step": 40747 }, { "epoch": 3.2, "learning_rate": 6.804830236022896e-05, "loss": 2.8303, "step": 40748 }, { "epoch": 3.2, "learning_rate": 6.804751823100448e-05, "loss": 2.735, "step": 40749 }, { "epoch": 3.2, "learning_rate": 6.804673410177998e-05, "loss": 2.655, "step": 40750 }, { "epoch": 3.2, "learning_rate": 6.804594997255548e-05, "loss": 2.8503, "step": 40751 }, { "epoch": 3.2, "learning_rate": 6.804516584333099e-05, "loss": 2.6903, "step": 40752 }, { "epoch": 3.2, "learning_rate": 6.804438171410649e-05, "loss": 2.4955, "step": 40753 }, { "epoch": 3.2, "learning_rate": 6.8043597584882e-05, "loss": 2.6109, "step": 40754 }, { "epoch": 3.2, "learning_rate": 6.80428134556575e-05, "loss": 2.5859, "step": 40755 }, { "epoch": 3.2, "learning_rate": 6.8042029326433e-05, "loss": 2.5733, "step": 40756 }, { "epoch": 3.2, "learning_rate": 6.80412451972085e-05, "loss": 2.8045, "step": 40757 }, { "epoch": 3.2, "learning_rate": 6.804046106798401e-05, "loss": 2.6904, "step": 40758 }, { "epoch": 3.2, "learning_rate": 6.803967693875951e-05, "loss": 2.5388, "step": 40759 }, { "epoch": 3.2, "learning_rate": 6.803889280953502e-05, "loss": 2.7974, "step": 40760 }, { "epoch": 3.2, "learning_rate": 6.803810868031052e-05, "loss": 2.7387, "step": 40761 }, { "epoch": 3.2, "learning_rate": 6.803732455108602e-05, "loss": 2.8038, "step": 40762 }, { "epoch": 3.2, "learning_rate": 6.803654042186153e-05, "loss": 2.7753, "step": 40763 }, { "epoch": 3.2, "learning_rate": 6.803575629263703e-05, "loss": 2.6759, "step": 40764 }, { "epoch": 3.2, "learning_rate": 6.803497216341254e-05, "loss": 2.8264, "step": 40765 }, { "epoch": 3.2, "learning_rate": 6.803418803418804e-05, "loss": 2.677, "step": 40766 }, { "epoch": 3.2, "learning_rate": 6.803340390496354e-05, "loss": 2.7327, "step": 40767 }, { "epoch": 3.2, "learning_rate": 6.803261977573905e-05, "loss": 2.687, "step": 40768 }, { "epoch": 3.2, "learning_rate": 6.803183564651455e-05, "loss": 2.6217, "step": 40769 }, { "epoch": 3.2, "learning_rate": 6.803105151729005e-05, "loss": 2.6121, "step": 40770 }, { "epoch": 3.2, "learning_rate": 6.803026738806556e-05, "loss": 2.7764, "step": 40771 }, { "epoch": 3.2, "learning_rate": 6.802948325884106e-05, "loss": 3.0511, "step": 40772 }, { "epoch": 3.2, "learning_rate": 6.802869912961656e-05, "loss": 2.8225, "step": 40773 }, { "epoch": 3.2, "learning_rate": 6.802791500039207e-05, "loss": 2.6007, "step": 40774 }, { "epoch": 3.2, "learning_rate": 6.802713087116757e-05, "loss": 2.8265, "step": 40775 }, { "epoch": 3.2, "learning_rate": 6.802634674194308e-05, "loss": 2.6666, "step": 40776 }, { "epoch": 3.2, "learning_rate": 6.802556261271858e-05, "loss": 2.7502, "step": 40777 }, { "epoch": 3.2, "learning_rate": 6.802477848349408e-05, "loss": 2.669, "step": 40778 }, { "epoch": 3.2, "learning_rate": 6.802399435426959e-05, "loss": 2.5124, "step": 40779 }, { "epoch": 3.2, "learning_rate": 6.802321022504509e-05, "loss": 2.9029, "step": 40780 }, { "epoch": 3.2, "learning_rate": 6.80224260958206e-05, "loss": 2.5493, "step": 40781 }, { "epoch": 3.2, "learning_rate": 6.80216419665961e-05, "loss": 2.7798, "step": 40782 }, { "epoch": 3.2, "learning_rate": 6.80208578373716e-05, "loss": 2.882, "step": 40783 }, { "epoch": 3.2, "learning_rate": 6.80200737081471e-05, "loss": 2.7406, "step": 40784 }, { "epoch": 3.2, "learning_rate": 6.801928957892261e-05, "loss": 2.6245, "step": 40785 }, { "epoch": 3.2, "learning_rate": 6.801850544969811e-05, "loss": 2.797, "step": 40786 }, { "epoch": 3.2, "learning_rate": 6.801772132047362e-05, "loss": 2.7126, "step": 40787 }, { "epoch": 3.2, "learning_rate": 6.801693719124912e-05, "loss": 2.9222, "step": 40788 }, { "epoch": 3.2, "learning_rate": 6.801615306202462e-05, "loss": 2.6771, "step": 40789 }, { "epoch": 3.2, "learning_rate": 6.801536893280013e-05, "loss": 2.7561, "step": 40790 }, { "epoch": 3.2, "learning_rate": 6.801458480357563e-05, "loss": 2.7546, "step": 40791 }, { "epoch": 3.2, "learning_rate": 6.801380067435113e-05, "loss": 2.6727, "step": 40792 }, { "epoch": 3.2, "learning_rate": 6.801301654512664e-05, "loss": 2.771, "step": 40793 }, { "epoch": 3.2, "learning_rate": 6.801223241590214e-05, "loss": 2.7607, "step": 40794 }, { "epoch": 3.2, "learning_rate": 6.801144828667765e-05, "loss": 2.8787, "step": 40795 }, { "epoch": 3.2, "learning_rate": 6.801066415745315e-05, "loss": 2.6957, "step": 40796 }, { "epoch": 3.2, "learning_rate": 6.800988002822865e-05, "loss": 2.5791, "step": 40797 }, { "epoch": 3.2, "learning_rate": 6.800909589900416e-05, "loss": 2.5107, "step": 40798 }, { "epoch": 3.2, "learning_rate": 6.800831176977966e-05, "loss": 2.7507, "step": 40799 }, { "epoch": 3.2, "learning_rate": 6.800752764055516e-05, "loss": 2.6829, "step": 40800 }, { "epoch": 3.2, "learning_rate": 6.800674351133067e-05, "loss": 2.6744, "step": 40801 }, { "epoch": 3.2, "learning_rate": 6.800595938210617e-05, "loss": 2.5593, "step": 40802 }, { "epoch": 3.2, "learning_rate": 6.800517525288168e-05, "loss": 2.572, "step": 40803 }, { "epoch": 3.2, "learning_rate": 6.800439112365718e-05, "loss": 2.8177, "step": 40804 }, { "epoch": 3.2, "learning_rate": 6.800360699443268e-05, "loss": 2.7748, "step": 40805 }, { "epoch": 3.2, "learning_rate": 6.80028228652082e-05, "loss": 2.8471, "step": 40806 }, { "epoch": 3.2, "learning_rate": 6.80020387359837e-05, "loss": 2.6981, "step": 40807 }, { "epoch": 3.2, "learning_rate": 6.800125460675921e-05, "loss": 2.7882, "step": 40808 }, { "epoch": 3.2, "learning_rate": 6.800047047753471e-05, "loss": 2.6314, "step": 40809 }, { "epoch": 3.2, "learning_rate": 6.799968634831021e-05, "loss": 2.4726, "step": 40810 }, { "epoch": 3.2, "learning_rate": 6.799890221908572e-05, "loss": 2.8585, "step": 40811 }, { "epoch": 3.2, "learning_rate": 6.799811808986121e-05, "loss": 2.6792, "step": 40812 }, { "epoch": 3.2, "learning_rate": 6.799733396063671e-05, "loss": 2.7597, "step": 40813 }, { "epoch": 3.2, "learning_rate": 6.799654983141222e-05, "loss": 2.9288, "step": 40814 }, { "epoch": 3.2, "learning_rate": 6.799576570218772e-05, "loss": 2.5961, "step": 40815 }, { "epoch": 3.2, "learning_rate": 6.799498157296322e-05, "loss": 2.6696, "step": 40816 }, { "epoch": 3.2, "learning_rate": 6.799419744373873e-05, "loss": 2.5624, "step": 40817 }, { "epoch": 3.2, "learning_rate": 6.799341331451423e-05, "loss": 2.7013, "step": 40818 }, { "epoch": 3.2, "learning_rate": 6.799262918528973e-05, "loss": 2.6814, "step": 40819 }, { "epoch": 3.2, "learning_rate": 6.799184505606524e-05, "loss": 2.7345, "step": 40820 }, { "epoch": 3.2, "learning_rate": 6.799106092684074e-05, "loss": 2.7427, "step": 40821 }, { "epoch": 3.2, "learning_rate": 6.799027679761625e-05, "loss": 2.5556, "step": 40822 }, { "epoch": 3.2, "learning_rate": 6.798949266839175e-05, "loss": 2.7956, "step": 40823 }, { "epoch": 3.2, "learning_rate": 6.798870853916725e-05, "loss": 2.7374, "step": 40824 }, { "epoch": 3.2, "learning_rate": 6.798792440994276e-05, "loss": 2.8275, "step": 40825 }, { "epoch": 3.2, "learning_rate": 6.798714028071826e-05, "loss": 2.7965, "step": 40826 }, { "epoch": 3.2, "learning_rate": 6.798635615149376e-05, "loss": 2.6613, "step": 40827 }, { "epoch": 3.2, "learning_rate": 6.798557202226927e-05, "loss": 2.4777, "step": 40828 }, { "epoch": 3.2, "learning_rate": 6.798478789304477e-05, "loss": 2.678, "step": 40829 }, { "epoch": 3.2, "learning_rate": 6.798400376382027e-05, "loss": 2.6753, "step": 40830 }, { "epoch": 3.2, "learning_rate": 6.798321963459579e-05, "loss": 2.7746, "step": 40831 }, { "epoch": 3.2, "learning_rate": 6.79824355053713e-05, "loss": 2.664, "step": 40832 }, { "epoch": 3.2, "learning_rate": 6.79816513761468e-05, "loss": 2.6653, "step": 40833 }, { "epoch": 3.2, "learning_rate": 6.79808672469223e-05, "loss": 2.5077, "step": 40834 }, { "epoch": 3.2, "learning_rate": 6.79800831176978e-05, "loss": 2.7624, "step": 40835 }, { "epoch": 3.2, "learning_rate": 6.797929898847331e-05, "loss": 2.7502, "step": 40836 }, { "epoch": 3.2, "learning_rate": 6.797851485924881e-05, "loss": 2.604, "step": 40837 }, { "epoch": 3.2, "learning_rate": 6.797773073002432e-05, "loss": 3.0279, "step": 40838 }, { "epoch": 3.2, "learning_rate": 6.797694660079982e-05, "loss": 2.9095, "step": 40839 }, { "epoch": 3.2, "learning_rate": 6.797616247157532e-05, "loss": 2.525, "step": 40840 }, { "epoch": 3.2, "learning_rate": 6.797537834235083e-05, "loss": 2.8716, "step": 40841 }, { "epoch": 3.2, "learning_rate": 6.797459421312633e-05, "loss": 2.8187, "step": 40842 }, { "epoch": 3.2, "learning_rate": 6.797381008390184e-05, "loss": 2.7731, "step": 40843 }, { "epoch": 3.2, "learning_rate": 6.797302595467734e-05, "loss": 2.7051, "step": 40844 }, { "epoch": 3.2, "learning_rate": 6.797224182545284e-05, "loss": 2.586, "step": 40845 }, { "epoch": 3.2, "learning_rate": 6.797145769622835e-05, "loss": 2.8122, "step": 40846 }, { "epoch": 3.2, "learning_rate": 6.797067356700385e-05, "loss": 2.6688, "step": 40847 }, { "epoch": 3.2, "learning_rate": 6.796988943777934e-05, "loss": 2.6393, "step": 40848 }, { "epoch": 3.2, "learning_rate": 6.796910530855484e-05, "loss": 2.7761, "step": 40849 }, { "epoch": 3.2, "learning_rate": 6.796832117933035e-05, "loss": 2.6796, "step": 40850 }, { "epoch": 3.2, "learning_rate": 6.796753705010585e-05, "loss": 2.5386, "step": 40851 }, { "epoch": 3.2, "learning_rate": 6.796675292088136e-05, "loss": 2.6128, "step": 40852 }, { "epoch": 3.2, "learning_rate": 6.796596879165686e-05, "loss": 2.6595, "step": 40853 }, { "epoch": 3.2, "learning_rate": 6.796518466243236e-05, "loss": 2.8503, "step": 40854 }, { "epoch": 3.2, "learning_rate": 6.796440053320787e-05, "loss": 2.945, "step": 40855 }, { "epoch": 3.2, "learning_rate": 6.796361640398337e-05, "loss": 2.6866, "step": 40856 }, { "epoch": 3.2, "learning_rate": 6.796283227475889e-05, "loss": 2.678, "step": 40857 }, { "epoch": 3.2, "learning_rate": 6.796204814553439e-05, "loss": 2.7582, "step": 40858 }, { "epoch": 3.2, "learning_rate": 6.79612640163099e-05, "loss": 2.9686, "step": 40859 }, { "epoch": 3.2, "learning_rate": 6.79604798870854e-05, "loss": 2.6528, "step": 40860 }, { "epoch": 3.2, "learning_rate": 6.79596957578609e-05, "loss": 2.7987, "step": 40861 }, { "epoch": 3.2, "learning_rate": 6.79589116286364e-05, "loss": 2.7676, "step": 40862 }, { "epoch": 3.2, "learning_rate": 6.795812749941191e-05, "loss": 2.6251, "step": 40863 }, { "epoch": 3.2, "learning_rate": 6.795734337018741e-05, "loss": 2.5588, "step": 40864 }, { "epoch": 3.2, "learning_rate": 6.795655924096292e-05, "loss": 2.6535, "step": 40865 }, { "epoch": 3.2, "learning_rate": 6.795577511173842e-05, "loss": 2.7309, "step": 40866 }, { "epoch": 3.2, "learning_rate": 6.795499098251392e-05, "loss": 2.6366, "step": 40867 }, { "epoch": 3.2, "learning_rate": 6.795420685328943e-05, "loss": 2.6309, "step": 40868 }, { "epoch": 3.2, "learning_rate": 6.795342272406493e-05, "loss": 2.71, "step": 40869 }, { "epoch": 3.2, "learning_rate": 6.795263859484044e-05, "loss": 2.6771, "step": 40870 }, { "epoch": 3.2, "learning_rate": 6.795185446561594e-05, "loss": 2.5716, "step": 40871 }, { "epoch": 3.2, "learning_rate": 6.795107033639144e-05, "loss": 2.6181, "step": 40872 }, { "epoch": 3.2, "learning_rate": 6.795028620716695e-05, "loss": 2.7993, "step": 40873 }, { "epoch": 3.21, "learning_rate": 6.794950207794245e-05, "loss": 2.7839, "step": 40874 }, { "epoch": 3.21, "learning_rate": 6.794871794871795e-05, "loss": 2.685, "step": 40875 }, { "epoch": 3.21, "learning_rate": 6.794793381949346e-05, "loss": 2.6888, "step": 40876 }, { "epoch": 3.21, "learning_rate": 6.794714969026896e-05, "loss": 2.6629, "step": 40877 }, { "epoch": 3.21, "learning_rate": 6.794636556104446e-05, "loss": 2.8197, "step": 40878 }, { "epoch": 3.21, "learning_rate": 6.794558143181997e-05, "loss": 2.9065, "step": 40879 }, { "epoch": 3.21, "learning_rate": 6.794479730259547e-05, "loss": 2.8124, "step": 40880 }, { "epoch": 3.21, "learning_rate": 6.794401317337098e-05, "loss": 2.6702, "step": 40881 }, { "epoch": 3.21, "learning_rate": 6.794322904414648e-05, "loss": 2.7771, "step": 40882 }, { "epoch": 3.21, "learning_rate": 6.794244491492198e-05, "loss": 2.5791, "step": 40883 }, { "epoch": 3.21, "learning_rate": 6.794166078569749e-05, "loss": 2.6048, "step": 40884 }, { "epoch": 3.21, "learning_rate": 6.794087665647299e-05, "loss": 2.5791, "step": 40885 }, { "epoch": 3.21, "learning_rate": 6.79400925272485e-05, "loss": 2.8701, "step": 40886 }, { "epoch": 3.21, "learning_rate": 6.7939308398024e-05, "loss": 2.7881, "step": 40887 }, { "epoch": 3.21, "learning_rate": 6.79385242687995e-05, "loss": 2.6546, "step": 40888 }, { "epoch": 3.21, "learning_rate": 6.7937740139575e-05, "loss": 2.8695, "step": 40889 }, { "epoch": 3.21, "learning_rate": 6.793695601035051e-05, "loss": 2.7453, "step": 40890 }, { "epoch": 3.21, "learning_rate": 6.793617188112601e-05, "loss": 2.6184, "step": 40891 }, { "epoch": 3.21, "learning_rate": 6.793538775190152e-05, "loss": 2.815, "step": 40892 }, { "epoch": 3.21, "learning_rate": 6.793460362267702e-05, "loss": 2.9125, "step": 40893 }, { "epoch": 3.21, "learning_rate": 6.793381949345252e-05, "loss": 2.6611, "step": 40894 }, { "epoch": 3.21, "learning_rate": 6.793303536422803e-05, "loss": 2.6145, "step": 40895 }, { "epoch": 3.21, "learning_rate": 6.793225123500353e-05, "loss": 2.7946, "step": 40896 }, { "epoch": 3.21, "learning_rate": 6.793146710577903e-05, "loss": 2.649, "step": 40897 }, { "epoch": 3.21, "learning_rate": 6.793068297655454e-05, "loss": 2.5699, "step": 40898 }, { "epoch": 3.21, "learning_rate": 6.792989884733004e-05, "loss": 2.688, "step": 40899 }, { "epoch": 3.21, "learning_rate": 6.792911471810555e-05, "loss": 2.6693, "step": 40900 }, { "epoch": 3.21, "learning_rate": 6.792833058888105e-05, "loss": 2.7125, "step": 40901 }, { "epoch": 3.21, "learning_rate": 6.792754645965655e-05, "loss": 2.6108, "step": 40902 }, { "epoch": 3.21, "learning_rate": 6.792676233043206e-05, "loss": 3.0237, "step": 40903 }, { "epoch": 3.21, "learning_rate": 6.792597820120756e-05, "loss": 2.8847, "step": 40904 }, { "epoch": 3.21, "learning_rate": 6.792519407198306e-05, "loss": 2.5814, "step": 40905 }, { "epoch": 3.21, "learning_rate": 6.792440994275857e-05, "loss": 2.6549, "step": 40906 }, { "epoch": 3.21, "learning_rate": 6.792362581353407e-05, "loss": 3.163, "step": 40907 }, { "epoch": 3.21, "learning_rate": 6.792284168430957e-05, "loss": 2.6405, "step": 40908 }, { "epoch": 3.21, "learning_rate": 6.792205755508508e-05, "loss": 2.6945, "step": 40909 }, { "epoch": 3.21, "learning_rate": 6.792127342586058e-05, "loss": 2.7284, "step": 40910 }, { "epoch": 3.21, "learning_rate": 6.792048929663609e-05, "loss": 2.7083, "step": 40911 }, { "epoch": 3.21, "learning_rate": 6.791970516741159e-05, "loss": 2.6902, "step": 40912 }, { "epoch": 3.21, "learning_rate": 6.79189210381871e-05, "loss": 2.7131, "step": 40913 }, { "epoch": 3.21, "learning_rate": 6.791813690896261e-05, "loss": 2.6483, "step": 40914 }, { "epoch": 3.21, "learning_rate": 6.791735277973811e-05, "loss": 2.818, "step": 40915 }, { "epoch": 3.21, "learning_rate": 6.791656865051362e-05, "loss": 2.8969, "step": 40916 }, { "epoch": 3.21, "learning_rate": 6.791578452128912e-05, "loss": 2.6009, "step": 40917 }, { "epoch": 3.21, "learning_rate": 6.791500039206463e-05, "loss": 2.8119, "step": 40918 }, { "epoch": 3.21, "learning_rate": 6.791421626284013e-05, "loss": 2.7934, "step": 40919 }, { "epoch": 3.21, "learning_rate": 6.791343213361562e-05, "loss": 2.5999, "step": 40920 }, { "epoch": 3.21, "learning_rate": 6.791264800439112e-05, "loss": 2.8668, "step": 40921 }, { "epoch": 3.21, "learning_rate": 6.791186387516663e-05, "loss": 2.7624, "step": 40922 }, { "epoch": 3.21, "learning_rate": 6.791107974594213e-05, "loss": 2.8412, "step": 40923 }, { "epoch": 3.21, "learning_rate": 6.791029561671763e-05, "loss": 2.6713, "step": 40924 }, { "epoch": 3.21, "learning_rate": 6.790951148749314e-05, "loss": 2.909, "step": 40925 }, { "epoch": 3.21, "learning_rate": 6.790872735826864e-05, "loss": 2.6523, "step": 40926 }, { "epoch": 3.21, "learning_rate": 6.790794322904414e-05, "loss": 2.5324, "step": 40927 }, { "epoch": 3.21, "learning_rate": 6.790715909981965e-05, "loss": 2.6522, "step": 40928 }, { "epoch": 3.21, "learning_rate": 6.790637497059515e-05, "loss": 2.6497, "step": 40929 }, { "epoch": 3.21, "learning_rate": 6.790559084137066e-05, "loss": 2.7456, "step": 40930 }, { "epoch": 3.21, "learning_rate": 6.790480671214616e-05, "loss": 2.6053, "step": 40931 }, { "epoch": 3.21, "learning_rate": 6.790402258292166e-05, "loss": 2.6401, "step": 40932 }, { "epoch": 3.21, "learning_rate": 6.790323845369717e-05, "loss": 2.6397, "step": 40933 }, { "epoch": 3.21, "learning_rate": 6.790245432447267e-05, "loss": 2.6537, "step": 40934 }, { "epoch": 3.21, "learning_rate": 6.790167019524817e-05, "loss": 2.416, "step": 40935 }, { "epoch": 3.21, "learning_rate": 6.790088606602368e-05, "loss": 2.6751, "step": 40936 }, { "epoch": 3.21, "learning_rate": 6.790010193679918e-05, "loss": 2.5411, "step": 40937 }, { "epoch": 3.21, "learning_rate": 6.789931780757469e-05, "loss": 2.6029, "step": 40938 }, { "epoch": 3.21, "learning_rate": 6.78985336783502e-05, "loss": 2.5783, "step": 40939 }, { "epoch": 3.21, "learning_rate": 6.78977495491257e-05, "loss": 2.6333, "step": 40940 }, { "epoch": 3.21, "learning_rate": 6.789696541990121e-05, "loss": 2.7539, "step": 40941 }, { "epoch": 3.21, "learning_rate": 6.789618129067671e-05, "loss": 2.8116, "step": 40942 }, { "epoch": 3.21, "learning_rate": 6.789539716145222e-05, "loss": 2.7342, "step": 40943 }, { "epoch": 3.21, "learning_rate": 6.789461303222772e-05, "loss": 2.67, "step": 40944 }, { "epoch": 3.21, "learning_rate": 6.789382890300322e-05, "loss": 2.651, "step": 40945 }, { "epoch": 3.21, "learning_rate": 6.789304477377873e-05, "loss": 2.6406, "step": 40946 }, { "epoch": 3.21, "learning_rate": 6.789226064455423e-05, "loss": 2.8384, "step": 40947 }, { "epoch": 3.21, "learning_rate": 6.789147651532974e-05, "loss": 2.7338, "step": 40948 }, { "epoch": 3.21, "learning_rate": 6.789069238610524e-05, "loss": 2.6638, "step": 40949 }, { "epoch": 3.21, "learning_rate": 6.788990825688074e-05, "loss": 2.7542, "step": 40950 }, { "epoch": 3.21, "learning_rate": 6.788912412765625e-05, "loss": 2.8833, "step": 40951 }, { "epoch": 3.21, "learning_rate": 6.788833999843175e-05, "loss": 2.7878, "step": 40952 }, { "epoch": 3.21, "learning_rate": 6.788755586920725e-05, "loss": 2.6002, "step": 40953 }, { "epoch": 3.21, "learning_rate": 6.788677173998276e-05, "loss": 2.7362, "step": 40954 }, { "epoch": 3.21, "learning_rate": 6.788598761075826e-05, "loss": 3.0963, "step": 40955 }, { "epoch": 3.21, "learning_rate": 6.788520348153375e-05, "loss": 2.8988, "step": 40956 }, { "epoch": 3.21, "learning_rate": 6.788441935230926e-05, "loss": 2.8399, "step": 40957 }, { "epoch": 3.21, "learning_rate": 6.788363522308476e-05, "loss": 2.5828, "step": 40958 }, { "epoch": 3.21, "learning_rate": 6.788285109386026e-05, "loss": 2.5596, "step": 40959 }, { "epoch": 3.21, "learning_rate": 6.788206696463577e-05, "loss": 2.8248, "step": 40960 }, { "epoch": 3.21, "learning_rate": 6.788128283541127e-05, "loss": 2.933, "step": 40961 }, { "epoch": 3.21, "learning_rate": 6.788049870618677e-05, "loss": 2.7993, "step": 40962 }, { "epoch": 3.21, "learning_rate": 6.787971457696228e-05, "loss": 2.548, "step": 40963 }, { "epoch": 3.21, "learning_rate": 6.787893044773778e-05, "loss": 2.7359, "step": 40964 }, { "epoch": 3.21, "learning_rate": 6.78781463185133e-05, "loss": 2.6768, "step": 40965 }, { "epoch": 3.21, "learning_rate": 6.78773621892888e-05, "loss": 2.659, "step": 40966 }, { "epoch": 3.21, "learning_rate": 6.78765780600643e-05, "loss": 2.622, "step": 40967 }, { "epoch": 3.21, "learning_rate": 6.787579393083981e-05, "loss": 2.7376, "step": 40968 }, { "epoch": 3.21, "learning_rate": 6.787500980161531e-05, "loss": 2.805, "step": 40969 }, { "epoch": 3.21, "learning_rate": 6.787422567239082e-05, "loss": 2.6061, "step": 40970 }, { "epoch": 3.21, "learning_rate": 6.787344154316632e-05, "loss": 2.6963, "step": 40971 }, { "epoch": 3.21, "learning_rate": 6.787265741394182e-05, "loss": 3.0667, "step": 40972 }, { "epoch": 3.21, "learning_rate": 6.787187328471733e-05, "loss": 2.6353, "step": 40973 }, { "epoch": 3.21, "learning_rate": 6.787108915549283e-05, "loss": 2.6187, "step": 40974 }, { "epoch": 3.21, "learning_rate": 6.787030502626833e-05, "loss": 2.7292, "step": 40975 }, { "epoch": 3.21, "learning_rate": 6.786952089704384e-05, "loss": 2.5413, "step": 40976 }, { "epoch": 3.21, "learning_rate": 6.786873676781934e-05, "loss": 2.8023, "step": 40977 }, { "epoch": 3.21, "learning_rate": 6.786795263859485e-05, "loss": 2.6184, "step": 40978 }, { "epoch": 3.21, "learning_rate": 6.786716850937035e-05, "loss": 2.6787, "step": 40979 }, { "epoch": 3.21, "learning_rate": 6.786638438014585e-05, "loss": 2.9574, "step": 40980 }, { "epoch": 3.21, "learning_rate": 6.786560025092136e-05, "loss": 2.8835, "step": 40981 }, { "epoch": 3.21, "learning_rate": 6.786481612169686e-05, "loss": 2.6255, "step": 40982 }, { "epoch": 3.21, "learning_rate": 6.786403199247236e-05, "loss": 2.6732, "step": 40983 }, { "epoch": 3.21, "learning_rate": 6.786324786324787e-05, "loss": 2.6408, "step": 40984 }, { "epoch": 3.21, "learning_rate": 6.786246373402337e-05, "loss": 2.5332, "step": 40985 }, { "epoch": 3.21, "learning_rate": 6.786167960479888e-05, "loss": 2.7766, "step": 40986 }, { "epoch": 3.21, "learning_rate": 6.786089547557438e-05, "loss": 2.5441, "step": 40987 }, { "epoch": 3.21, "learning_rate": 6.786011134634988e-05, "loss": 2.8323, "step": 40988 }, { "epoch": 3.21, "learning_rate": 6.785932721712539e-05, "loss": 2.615, "step": 40989 }, { "epoch": 3.21, "learning_rate": 6.785854308790089e-05, "loss": 2.81, "step": 40990 }, { "epoch": 3.21, "learning_rate": 6.78577589586764e-05, "loss": 2.7405, "step": 40991 }, { "epoch": 3.21, "learning_rate": 6.78569748294519e-05, "loss": 2.6429, "step": 40992 }, { "epoch": 3.21, "learning_rate": 6.78561907002274e-05, "loss": 2.5123, "step": 40993 }, { "epoch": 3.21, "learning_rate": 6.78554065710029e-05, "loss": 2.8138, "step": 40994 }, { "epoch": 3.21, "learning_rate": 6.785462244177841e-05, "loss": 2.537, "step": 40995 }, { "epoch": 3.21, "learning_rate": 6.785383831255391e-05, "loss": 2.5823, "step": 40996 }, { "epoch": 3.21, "learning_rate": 6.785305418332942e-05, "loss": 2.9791, "step": 40997 }, { "epoch": 3.21, "learning_rate": 6.785227005410492e-05, "loss": 2.8033, "step": 40998 }, { "epoch": 3.21, "learning_rate": 6.785148592488042e-05, "loss": 2.4898, "step": 40999 }, { "epoch": 3.21, "learning_rate": 6.785070179565593e-05, "loss": 2.6465, "step": 41000 }, { "epoch": 3.21, "eval_loss": 2.898484706878662, "eval_runtime": 217.8091, "eval_samples_per_second": 52.027, "eval_steps_per_second": 3.255, "step": 41000 }, { "epoch": 3.22, "learning_rate": 6.784991766643143e-05, "loss": 2.6304, "step": 41001 }, { "epoch": 3.22, "learning_rate": 6.784913353720693e-05, "loss": 2.6838, "step": 41002 }, { "epoch": 3.22, "learning_rate": 6.784834940798244e-05, "loss": 2.5068, "step": 41003 }, { "epoch": 3.22, "learning_rate": 6.784756527875794e-05, "loss": 2.7529, "step": 41004 }, { "epoch": 3.22, "learning_rate": 6.784678114953345e-05, "loss": 2.9783, "step": 41005 }, { "epoch": 3.22, "learning_rate": 6.784599702030895e-05, "loss": 2.8194, "step": 41006 }, { "epoch": 3.22, "learning_rate": 6.784521289108445e-05, "loss": 2.6664, "step": 41007 }, { "epoch": 3.22, "learning_rate": 6.784442876185996e-05, "loss": 2.7168, "step": 41008 }, { "epoch": 3.22, "learning_rate": 6.784364463263546e-05, "loss": 2.8128, "step": 41009 }, { "epoch": 3.22, "learning_rate": 6.784286050341096e-05, "loss": 2.6533, "step": 41010 }, { "epoch": 3.22, "learning_rate": 6.784207637418647e-05, "loss": 2.8114, "step": 41011 }, { "epoch": 3.22, "learning_rate": 6.784129224496197e-05, "loss": 2.7863, "step": 41012 }, { "epoch": 3.22, "learning_rate": 6.784050811573747e-05, "loss": 2.7991, "step": 41013 }, { "epoch": 3.22, "learning_rate": 6.783972398651298e-05, "loss": 2.5852, "step": 41014 }, { "epoch": 3.22, "learning_rate": 6.783893985728848e-05, "loss": 2.6688, "step": 41015 }, { "epoch": 3.22, "learning_rate": 6.783815572806399e-05, "loss": 2.6813, "step": 41016 }, { "epoch": 3.22, "learning_rate": 6.783737159883949e-05, "loss": 2.7974, "step": 41017 }, { "epoch": 3.22, "learning_rate": 6.783658746961499e-05, "loss": 2.7071, "step": 41018 }, { "epoch": 3.22, "learning_rate": 6.78358033403905e-05, "loss": 2.8571, "step": 41019 }, { "epoch": 3.22, "learning_rate": 6.7835019211166e-05, "loss": 2.6637, "step": 41020 }, { "epoch": 3.22, "learning_rate": 6.78342350819415e-05, "loss": 2.6487, "step": 41021 }, { "epoch": 3.22, "learning_rate": 6.783345095271702e-05, "loss": 2.6979, "step": 41022 }, { "epoch": 3.22, "learning_rate": 6.783266682349252e-05, "loss": 2.559, "step": 41023 }, { "epoch": 3.22, "learning_rate": 6.783188269426803e-05, "loss": 2.6992, "step": 41024 }, { "epoch": 3.22, "learning_rate": 6.783109856504353e-05, "loss": 2.7203, "step": 41025 }, { "epoch": 3.22, "learning_rate": 6.783031443581904e-05, "loss": 2.5229, "step": 41026 }, { "epoch": 3.22, "learning_rate": 6.782953030659454e-05, "loss": 2.7311, "step": 41027 }, { "epoch": 3.22, "learning_rate": 6.782874617737003e-05, "loss": 2.6181, "step": 41028 }, { "epoch": 3.22, "learning_rate": 6.782796204814553e-05, "loss": 2.8182, "step": 41029 }, { "epoch": 3.22, "learning_rate": 6.782717791892104e-05, "loss": 2.6447, "step": 41030 }, { "epoch": 3.22, "learning_rate": 6.782639378969654e-05, "loss": 2.66, "step": 41031 }, { "epoch": 3.22, "learning_rate": 6.782560966047204e-05, "loss": 2.6888, "step": 41032 }, { "epoch": 3.22, "learning_rate": 6.782482553124755e-05, "loss": 2.8811, "step": 41033 }, { "epoch": 3.22, "learning_rate": 6.782404140202305e-05, "loss": 2.8278, "step": 41034 }, { "epoch": 3.22, "learning_rate": 6.782325727279856e-05, "loss": 2.7508, "step": 41035 }, { "epoch": 3.22, "learning_rate": 6.782247314357406e-05, "loss": 2.8325, "step": 41036 }, { "epoch": 3.22, "learning_rate": 6.782168901434956e-05, "loss": 2.4629, "step": 41037 }, { "epoch": 3.22, "learning_rate": 6.782090488512507e-05, "loss": 2.8202, "step": 41038 }, { "epoch": 3.22, "learning_rate": 6.782012075590057e-05, "loss": 2.822, "step": 41039 }, { "epoch": 3.22, "learning_rate": 6.781933662667607e-05, "loss": 2.5095, "step": 41040 }, { "epoch": 3.22, "learning_rate": 6.781855249745158e-05, "loss": 2.7422, "step": 41041 }, { "epoch": 3.22, "learning_rate": 6.781776836822708e-05, "loss": 2.7832, "step": 41042 }, { "epoch": 3.22, "learning_rate": 6.781698423900258e-05, "loss": 2.7743, "step": 41043 }, { "epoch": 3.22, "learning_rate": 6.781620010977809e-05, "loss": 2.6742, "step": 41044 }, { "epoch": 3.22, "learning_rate": 6.781541598055359e-05, "loss": 2.5807, "step": 41045 }, { "epoch": 3.22, "learning_rate": 6.78146318513291e-05, "loss": 2.6281, "step": 41046 }, { "epoch": 3.22, "learning_rate": 6.781384772210461e-05, "loss": 2.698, "step": 41047 }, { "epoch": 3.22, "learning_rate": 6.781306359288012e-05, "loss": 2.6388, "step": 41048 }, { "epoch": 3.22, "learning_rate": 6.781227946365562e-05, "loss": 2.7177, "step": 41049 }, { "epoch": 3.22, "learning_rate": 6.781149533443112e-05, "loss": 3.012, "step": 41050 }, { "epoch": 3.22, "learning_rate": 6.781071120520663e-05, "loss": 2.7568, "step": 41051 }, { "epoch": 3.22, "learning_rate": 6.780992707598213e-05, "loss": 2.8172, "step": 41052 }, { "epoch": 3.22, "learning_rate": 6.780914294675764e-05, "loss": 2.52, "step": 41053 }, { "epoch": 3.22, "learning_rate": 6.780835881753314e-05, "loss": 2.6249, "step": 41054 }, { "epoch": 3.22, "learning_rate": 6.780757468830864e-05, "loss": 2.4771, "step": 41055 }, { "epoch": 3.22, "learning_rate": 6.780679055908415e-05, "loss": 2.8544, "step": 41056 }, { "epoch": 3.22, "learning_rate": 6.780600642985965e-05, "loss": 2.6819, "step": 41057 }, { "epoch": 3.22, "learning_rate": 6.780522230063515e-05, "loss": 2.7543, "step": 41058 }, { "epoch": 3.22, "learning_rate": 6.780443817141066e-05, "loss": 2.6613, "step": 41059 }, { "epoch": 3.22, "learning_rate": 6.780365404218616e-05, "loss": 2.7045, "step": 41060 }, { "epoch": 3.22, "learning_rate": 6.780286991296166e-05, "loss": 2.6715, "step": 41061 }, { "epoch": 3.22, "learning_rate": 6.780208578373717e-05, "loss": 2.6325, "step": 41062 }, { "epoch": 3.22, "learning_rate": 6.780130165451267e-05, "loss": 2.857, "step": 41063 }, { "epoch": 3.22, "learning_rate": 6.780051752528816e-05, "loss": 2.7592, "step": 41064 }, { "epoch": 3.22, "learning_rate": 6.779973339606367e-05, "loss": 2.7146, "step": 41065 }, { "epoch": 3.22, "learning_rate": 6.779894926683917e-05, "loss": 2.8466, "step": 41066 }, { "epoch": 3.22, "learning_rate": 6.779816513761467e-05, "loss": 2.6487, "step": 41067 }, { "epoch": 3.22, "learning_rate": 6.779738100839018e-05, "loss": 2.9966, "step": 41068 }, { "epoch": 3.22, "learning_rate": 6.779659687916568e-05, "loss": 2.6284, "step": 41069 }, { "epoch": 3.22, "learning_rate": 6.779581274994118e-05, "loss": 2.6397, "step": 41070 }, { "epoch": 3.22, "learning_rate": 6.779502862071669e-05, "loss": 2.7365, "step": 41071 }, { "epoch": 3.22, "learning_rate": 6.779424449149219e-05, "loss": 2.7406, "step": 41072 }, { "epoch": 3.22, "learning_rate": 6.779346036226771e-05, "loss": 2.692, "step": 41073 }, { "epoch": 3.22, "learning_rate": 6.779267623304321e-05, "loss": 2.7609, "step": 41074 }, { "epoch": 3.22, "learning_rate": 6.779189210381872e-05, "loss": 2.5908, "step": 41075 }, { "epoch": 3.22, "learning_rate": 6.779110797459422e-05, "loss": 2.8666, "step": 41076 }, { "epoch": 3.22, "learning_rate": 6.779032384536972e-05, "loss": 2.7239, "step": 41077 }, { "epoch": 3.22, "learning_rate": 6.778953971614523e-05, "loss": 2.7079, "step": 41078 }, { "epoch": 3.22, "learning_rate": 6.778875558692073e-05, "loss": 2.6727, "step": 41079 }, { "epoch": 3.22, "learning_rate": 6.778797145769623e-05, "loss": 2.6302, "step": 41080 }, { "epoch": 3.22, "learning_rate": 6.778718732847174e-05, "loss": 3.072, "step": 41081 }, { "epoch": 3.22, "learning_rate": 6.778640319924724e-05, "loss": 2.8892, "step": 41082 }, { "epoch": 3.22, "learning_rate": 6.778561907002275e-05, "loss": 2.7616, "step": 41083 }, { "epoch": 3.22, "learning_rate": 6.778483494079825e-05, "loss": 2.7236, "step": 41084 }, { "epoch": 3.22, "learning_rate": 6.778405081157375e-05, "loss": 2.6387, "step": 41085 }, { "epoch": 3.22, "learning_rate": 6.778326668234926e-05, "loss": 2.8973, "step": 41086 }, { "epoch": 3.22, "learning_rate": 6.778248255312476e-05, "loss": 2.7332, "step": 41087 }, { "epoch": 3.22, "learning_rate": 6.778169842390026e-05, "loss": 2.6553, "step": 41088 }, { "epoch": 3.22, "learning_rate": 6.778091429467577e-05, "loss": 2.7016, "step": 41089 }, { "epoch": 3.22, "learning_rate": 6.778013016545127e-05, "loss": 2.7462, "step": 41090 }, { "epoch": 3.22, "learning_rate": 6.777934603622678e-05, "loss": 2.6786, "step": 41091 }, { "epoch": 3.22, "learning_rate": 6.777856190700228e-05, "loss": 2.8066, "step": 41092 }, { "epoch": 3.22, "learning_rate": 6.777777777777778e-05, "loss": 2.8283, "step": 41093 }, { "epoch": 3.22, "learning_rate": 6.777699364855329e-05, "loss": 2.6042, "step": 41094 }, { "epoch": 3.22, "learning_rate": 6.777620951932879e-05, "loss": 2.9584, "step": 41095 }, { "epoch": 3.22, "learning_rate": 6.77754253901043e-05, "loss": 2.8957, "step": 41096 }, { "epoch": 3.22, "learning_rate": 6.77746412608798e-05, "loss": 2.637, "step": 41097 }, { "epoch": 3.22, "learning_rate": 6.77738571316553e-05, "loss": 2.7627, "step": 41098 }, { "epoch": 3.22, "learning_rate": 6.77730730024308e-05, "loss": 2.7398, "step": 41099 }, { "epoch": 3.22, "learning_rate": 6.777228887320631e-05, "loss": 2.8164, "step": 41100 }, { "epoch": 3.22, "learning_rate": 6.777150474398181e-05, "loss": 2.5416, "step": 41101 }, { "epoch": 3.22, "learning_rate": 6.777072061475732e-05, "loss": 2.7012, "step": 41102 }, { "epoch": 3.22, "learning_rate": 6.776993648553282e-05, "loss": 2.6856, "step": 41103 }, { "epoch": 3.22, "learning_rate": 6.776915235630832e-05, "loss": 2.854, "step": 41104 }, { "epoch": 3.22, "learning_rate": 6.776836822708383e-05, "loss": 2.7219, "step": 41105 }, { "epoch": 3.22, "learning_rate": 6.776758409785933e-05, "loss": 2.5355, "step": 41106 }, { "epoch": 3.22, "learning_rate": 6.776679996863483e-05, "loss": 2.765, "step": 41107 }, { "epoch": 3.22, "learning_rate": 6.776601583941034e-05, "loss": 2.6952, "step": 41108 }, { "epoch": 3.22, "learning_rate": 6.776523171018584e-05, "loss": 2.7114, "step": 41109 }, { "epoch": 3.22, "learning_rate": 6.776444758096134e-05, "loss": 2.7201, "step": 41110 }, { "epoch": 3.22, "learning_rate": 6.776366345173685e-05, "loss": 2.6697, "step": 41111 }, { "epoch": 3.22, "learning_rate": 6.776287932251235e-05, "loss": 2.6428, "step": 41112 }, { "epoch": 3.22, "learning_rate": 6.776209519328786e-05, "loss": 2.5452, "step": 41113 }, { "epoch": 3.22, "learning_rate": 6.776131106406336e-05, "loss": 2.6101, "step": 41114 }, { "epoch": 3.22, "learning_rate": 6.776052693483886e-05, "loss": 2.6495, "step": 41115 }, { "epoch": 3.22, "learning_rate": 6.775974280561437e-05, "loss": 2.6715, "step": 41116 }, { "epoch": 3.22, "learning_rate": 6.775895867638987e-05, "loss": 2.7871, "step": 41117 }, { "epoch": 3.22, "learning_rate": 6.775817454716537e-05, "loss": 2.7319, "step": 41118 }, { "epoch": 3.22, "learning_rate": 6.775739041794088e-05, "loss": 2.657, "step": 41119 }, { "epoch": 3.22, "learning_rate": 6.775660628871638e-05, "loss": 2.6818, "step": 41120 }, { "epoch": 3.22, "learning_rate": 6.775582215949189e-05, "loss": 2.663, "step": 41121 }, { "epoch": 3.22, "learning_rate": 6.775503803026739e-05, "loss": 2.6633, "step": 41122 }, { "epoch": 3.22, "learning_rate": 6.775425390104289e-05, "loss": 2.766, "step": 41123 }, { "epoch": 3.22, "learning_rate": 6.77534697718184e-05, "loss": 2.5609, "step": 41124 }, { "epoch": 3.22, "learning_rate": 6.77526856425939e-05, "loss": 2.5809, "step": 41125 }, { "epoch": 3.22, "learning_rate": 6.77519015133694e-05, "loss": 2.6603, "step": 41126 }, { "epoch": 3.22, "learning_rate": 6.775111738414491e-05, "loss": 2.5716, "step": 41127 }, { "epoch": 3.22, "learning_rate": 6.775033325492041e-05, "loss": 2.8332, "step": 41128 }, { "epoch": 3.23, "learning_rate": 6.774954912569591e-05, "loss": 2.8261, "step": 41129 }, { "epoch": 3.23, "learning_rate": 6.774876499647143e-05, "loss": 2.6253, "step": 41130 }, { "epoch": 3.23, "learning_rate": 6.774798086724694e-05, "loss": 2.8002, "step": 41131 }, { "epoch": 3.23, "learning_rate": 6.774719673802244e-05, "loss": 2.6045, "step": 41132 }, { "epoch": 3.23, "learning_rate": 6.774641260879794e-05, "loss": 2.6221, "step": 41133 }, { "epoch": 3.23, "learning_rate": 6.774562847957345e-05, "loss": 2.7129, "step": 41134 }, { "epoch": 3.23, "learning_rate": 6.774484435034895e-05, "loss": 2.8082, "step": 41135 }, { "epoch": 3.23, "learning_rate": 6.774406022112444e-05, "loss": 2.5587, "step": 41136 }, { "epoch": 3.23, "learning_rate": 6.774327609189994e-05, "loss": 2.6639, "step": 41137 }, { "epoch": 3.23, "learning_rate": 6.774249196267545e-05, "loss": 2.6382, "step": 41138 }, { "epoch": 3.23, "learning_rate": 6.774170783345095e-05, "loss": 2.4992, "step": 41139 }, { "epoch": 3.23, "learning_rate": 6.774092370422646e-05, "loss": 2.675, "step": 41140 }, { "epoch": 3.23, "learning_rate": 6.774013957500196e-05, "loss": 2.7201, "step": 41141 }, { "epoch": 3.23, "learning_rate": 6.773935544577746e-05, "loss": 2.8522, "step": 41142 }, { "epoch": 3.23, "learning_rate": 6.773857131655297e-05, "loss": 2.6849, "step": 41143 }, { "epoch": 3.23, "learning_rate": 6.773778718732847e-05, "loss": 3.0017, "step": 41144 }, { "epoch": 3.23, "learning_rate": 6.773700305810397e-05, "loss": 2.8427, "step": 41145 }, { "epoch": 3.23, "learning_rate": 6.773621892887948e-05, "loss": 2.6533, "step": 41146 }, { "epoch": 3.23, "learning_rate": 6.773543479965498e-05, "loss": 2.6932, "step": 41147 }, { "epoch": 3.23, "learning_rate": 6.773465067043048e-05, "loss": 2.7291, "step": 41148 }, { "epoch": 3.23, "learning_rate": 6.773386654120599e-05, "loss": 2.7256, "step": 41149 }, { "epoch": 3.23, "learning_rate": 6.773308241198149e-05, "loss": 2.9342, "step": 41150 }, { "epoch": 3.23, "learning_rate": 6.7732298282757e-05, "loss": 2.7279, "step": 41151 }, { "epoch": 3.23, "learning_rate": 6.77315141535325e-05, "loss": 2.546, "step": 41152 }, { "epoch": 3.23, "learning_rate": 6.7730730024308e-05, "loss": 2.7122, "step": 41153 }, { "epoch": 3.23, "learning_rate": 6.77299458950835e-05, "loss": 2.7671, "step": 41154 }, { "epoch": 3.23, "learning_rate": 6.772916176585902e-05, "loss": 2.8032, "step": 41155 }, { "epoch": 3.23, "learning_rate": 6.772837763663453e-05, "loss": 2.8215, "step": 41156 }, { "epoch": 3.23, "learning_rate": 6.772759350741003e-05, "loss": 2.6122, "step": 41157 }, { "epoch": 3.23, "learning_rate": 6.772680937818553e-05, "loss": 2.5612, "step": 41158 }, { "epoch": 3.23, "learning_rate": 6.772602524896104e-05, "loss": 2.7595, "step": 41159 }, { "epoch": 3.23, "learning_rate": 6.772524111973654e-05, "loss": 2.7153, "step": 41160 }, { "epoch": 3.23, "learning_rate": 6.772445699051205e-05, "loss": 2.8635, "step": 41161 }, { "epoch": 3.23, "learning_rate": 6.772367286128755e-05, "loss": 2.735, "step": 41162 }, { "epoch": 3.23, "learning_rate": 6.772288873206305e-05, "loss": 2.7713, "step": 41163 }, { "epoch": 3.23, "learning_rate": 6.772210460283856e-05, "loss": 2.9414, "step": 41164 }, { "epoch": 3.23, "learning_rate": 6.772132047361406e-05, "loss": 2.6733, "step": 41165 }, { "epoch": 3.23, "learning_rate": 6.772053634438956e-05, "loss": 2.5475, "step": 41166 }, { "epoch": 3.23, "learning_rate": 6.771975221516507e-05, "loss": 2.8278, "step": 41167 }, { "epoch": 3.23, "learning_rate": 6.771896808594057e-05, "loss": 2.8147, "step": 41168 }, { "epoch": 3.23, "learning_rate": 6.771818395671608e-05, "loss": 2.9047, "step": 41169 }, { "epoch": 3.23, "learning_rate": 6.771739982749158e-05, "loss": 2.6557, "step": 41170 }, { "epoch": 3.23, "learning_rate": 6.771661569826708e-05, "loss": 3.177, "step": 41171 }, { "epoch": 3.23, "learning_rate": 6.771583156904257e-05, "loss": 2.7602, "step": 41172 }, { "epoch": 3.23, "learning_rate": 6.771504743981808e-05, "loss": 2.8214, "step": 41173 }, { "epoch": 3.23, "learning_rate": 6.771426331059358e-05, "loss": 2.8228, "step": 41174 }, { "epoch": 3.23, "learning_rate": 6.771347918136908e-05, "loss": 2.7624, "step": 41175 }, { "epoch": 3.23, "learning_rate": 6.771269505214459e-05, "loss": 2.5997, "step": 41176 }, { "epoch": 3.23, "learning_rate": 6.771191092292009e-05, "loss": 2.8339, "step": 41177 }, { "epoch": 3.23, "learning_rate": 6.77111267936956e-05, "loss": 2.7557, "step": 41178 }, { "epoch": 3.23, "learning_rate": 6.77103426644711e-05, "loss": 2.6445, "step": 41179 }, { "epoch": 3.23, "learning_rate": 6.77095585352466e-05, "loss": 2.5484, "step": 41180 }, { "epoch": 3.23, "learning_rate": 6.770877440602212e-05, "loss": 2.9282, "step": 41181 }, { "epoch": 3.23, "learning_rate": 6.770799027679762e-05, "loss": 2.6308, "step": 41182 }, { "epoch": 3.23, "learning_rate": 6.770720614757313e-05, "loss": 2.7703, "step": 41183 }, { "epoch": 3.23, "learning_rate": 6.770642201834863e-05, "loss": 2.7044, "step": 41184 }, { "epoch": 3.23, "learning_rate": 6.770563788912413e-05, "loss": 2.6696, "step": 41185 }, { "epoch": 3.23, "learning_rate": 6.770485375989964e-05, "loss": 2.6107, "step": 41186 }, { "epoch": 3.23, "learning_rate": 6.770406963067514e-05, "loss": 2.6929, "step": 41187 }, { "epoch": 3.23, "learning_rate": 6.770328550145065e-05, "loss": 2.5302, "step": 41188 }, { "epoch": 3.23, "learning_rate": 6.770250137222615e-05, "loss": 2.7411, "step": 41189 }, { "epoch": 3.23, "learning_rate": 6.770171724300165e-05, "loss": 2.8171, "step": 41190 }, { "epoch": 3.23, "learning_rate": 6.770093311377716e-05, "loss": 2.784, "step": 41191 }, { "epoch": 3.23, "learning_rate": 6.770014898455266e-05, "loss": 2.7556, "step": 41192 }, { "epoch": 3.23, "learning_rate": 6.769936485532816e-05, "loss": 2.5072, "step": 41193 }, { "epoch": 3.23, "learning_rate": 6.769858072610367e-05, "loss": 2.8355, "step": 41194 }, { "epoch": 3.23, "learning_rate": 6.769779659687917e-05, "loss": 2.5718, "step": 41195 }, { "epoch": 3.23, "learning_rate": 6.769701246765467e-05, "loss": 2.8003, "step": 41196 }, { "epoch": 3.23, "learning_rate": 6.769622833843018e-05, "loss": 2.785, "step": 41197 }, { "epoch": 3.23, "learning_rate": 6.769544420920568e-05, "loss": 2.7048, "step": 41198 }, { "epoch": 3.23, "learning_rate": 6.769466007998119e-05, "loss": 2.5461, "step": 41199 }, { "epoch": 3.23, "learning_rate": 6.769387595075669e-05, "loss": 2.606, "step": 41200 }, { "epoch": 3.23, "learning_rate": 6.769309182153219e-05, "loss": 2.943, "step": 41201 }, { "epoch": 3.23, "learning_rate": 6.76923076923077e-05, "loss": 2.6681, "step": 41202 }, { "epoch": 3.23, "learning_rate": 6.76915235630832e-05, "loss": 2.7358, "step": 41203 }, { "epoch": 3.23, "learning_rate": 6.76907394338587e-05, "loss": 2.4983, "step": 41204 }, { "epoch": 3.23, "learning_rate": 6.768995530463421e-05, "loss": 2.6764, "step": 41205 }, { "epoch": 3.23, "learning_rate": 6.768917117540971e-05, "loss": 2.6937, "step": 41206 }, { "epoch": 3.23, "learning_rate": 6.768838704618522e-05, "loss": 2.6581, "step": 41207 }, { "epoch": 3.23, "learning_rate": 6.768760291696072e-05, "loss": 2.7131, "step": 41208 }, { "epoch": 3.23, "learning_rate": 6.768681878773622e-05, "loss": 2.7062, "step": 41209 }, { "epoch": 3.23, "learning_rate": 6.768603465851173e-05, "loss": 2.839, "step": 41210 }, { "epoch": 3.23, "learning_rate": 6.768525052928723e-05, "loss": 2.7765, "step": 41211 }, { "epoch": 3.23, "learning_rate": 6.768446640006273e-05, "loss": 2.8034, "step": 41212 }, { "epoch": 3.23, "learning_rate": 6.768368227083824e-05, "loss": 2.5753, "step": 41213 }, { "epoch": 3.23, "learning_rate": 6.768289814161374e-05, "loss": 2.6169, "step": 41214 }, { "epoch": 3.23, "learning_rate": 6.768211401238924e-05, "loss": 2.8491, "step": 41215 }, { "epoch": 3.23, "learning_rate": 6.768132988316475e-05, "loss": 2.6454, "step": 41216 }, { "epoch": 3.23, "learning_rate": 6.768054575394025e-05, "loss": 2.7459, "step": 41217 }, { "epoch": 3.23, "learning_rate": 6.767976162471576e-05, "loss": 2.6775, "step": 41218 }, { "epoch": 3.23, "learning_rate": 6.767897749549126e-05, "loss": 2.7801, "step": 41219 }, { "epoch": 3.23, "learning_rate": 6.767819336626676e-05, "loss": 2.6922, "step": 41220 }, { "epoch": 3.23, "learning_rate": 6.767740923704227e-05, "loss": 2.7694, "step": 41221 }, { "epoch": 3.23, "learning_rate": 6.767662510781777e-05, "loss": 2.7292, "step": 41222 }, { "epoch": 3.23, "learning_rate": 6.767584097859327e-05, "loss": 2.7787, "step": 41223 }, { "epoch": 3.23, "learning_rate": 6.767505684936878e-05, "loss": 2.7083, "step": 41224 }, { "epoch": 3.23, "learning_rate": 6.767427272014428e-05, "loss": 2.7407, "step": 41225 }, { "epoch": 3.23, "learning_rate": 6.767348859091979e-05, "loss": 2.7593, "step": 41226 }, { "epoch": 3.23, "learning_rate": 6.767270446169529e-05, "loss": 2.7831, "step": 41227 }, { "epoch": 3.23, "learning_rate": 6.767192033247079e-05, "loss": 2.6981, "step": 41228 }, { "epoch": 3.23, "learning_rate": 6.76711362032463e-05, "loss": 2.6736, "step": 41229 }, { "epoch": 3.23, "learning_rate": 6.76703520740218e-05, "loss": 2.8413, "step": 41230 }, { "epoch": 3.23, "learning_rate": 6.76695679447973e-05, "loss": 2.6399, "step": 41231 }, { "epoch": 3.23, "learning_rate": 6.766878381557281e-05, "loss": 2.9084, "step": 41232 }, { "epoch": 3.23, "learning_rate": 6.766799968634831e-05, "loss": 2.7725, "step": 41233 }, { "epoch": 3.23, "learning_rate": 6.766721555712381e-05, "loss": 2.7034, "step": 41234 }, { "epoch": 3.23, "learning_rate": 6.766643142789932e-05, "loss": 2.9564, "step": 41235 }, { "epoch": 3.23, "learning_rate": 6.766564729867482e-05, "loss": 2.7148, "step": 41236 }, { "epoch": 3.23, "learning_rate": 6.766486316945033e-05, "loss": 2.6776, "step": 41237 }, { "epoch": 3.23, "learning_rate": 6.766407904022584e-05, "loss": 2.6137, "step": 41238 }, { "epoch": 3.23, "learning_rate": 6.766329491100135e-05, "loss": 2.5984, "step": 41239 }, { "epoch": 3.23, "learning_rate": 6.766251078177685e-05, "loss": 2.8356, "step": 41240 }, { "epoch": 3.23, "learning_rate": 6.766172665255235e-05, "loss": 2.6843, "step": 41241 }, { "epoch": 3.23, "learning_rate": 6.766094252332786e-05, "loss": 2.6985, "step": 41242 }, { "epoch": 3.23, "learning_rate": 6.766015839410336e-05, "loss": 2.8178, "step": 41243 }, { "epoch": 3.23, "learning_rate": 6.765937426487885e-05, "loss": 2.7183, "step": 41244 }, { "epoch": 3.23, "learning_rate": 6.765859013565435e-05, "loss": 2.6209, "step": 41245 }, { "epoch": 3.23, "learning_rate": 6.765780600642986e-05, "loss": 2.4754, "step": 41246 }, { "epoch": 3.23, "learning_rate": 6.765702187720536e-05, "loss": 2.5758, "step": 41247 }, { "epoch": 3.23, "learning_rate": 6.765623774798087e-05, "loss": 2.6775, "step": 41248 }, { "epoch": 3.23, "learning_rate": 6.765545361875637e-05, "loss": 2.6237, "step": 41249 }, { "epoch": 3.23, "learning_rate": 6.765466948953187e-05, "loss": 2.6493, "step": 41250 }, { "epoch": 3.23, "learning_rate": 6.765388536030738e-05, "loss": 2.8724, "step": 41251 }, { "epoch": 3.23, "learning_rate": 6.765310123108288e-05, "loss": 3.0993, "step": 41252 }, { "epoch": 3.23, "learning_rate": 6.765231710185838e-05, "loss": 2.6976, "step": 41253 }, { "epoch": 3.23, "learning_rate": 6.765153297263389e-05, "loss": 2.5923, "step": 41254 }, { "epoch": 3.23, "learning_rate": 6.765074884340939e-05, "loss": 2.5857, "step": 41255 }, { "epoch": 3.24, "learning_rate": 6.76499647141849e-05, "loss": 2.546, "step": 41256 }, { "epoch": 3.24, "learning_rate": 6.76491805849604e-05, "loss": 2.678, "step": 41257 }, { "epoch": 3.24, "learning_rate": 6.76483964557359e-05, "loss": 2.6519, "step": 41258 }, { "epoch": 3.24, "learning_rate": 6.76476123265114e-05, "loss": 2.5502, "step": 41259 }, { "epoch": 3.24, "learning_rate": 6.764682819728691e-05, "loss": 2.9952, "step": 41260 }, { "epoch": 3.24, "learning_rate": 6.764604406806241e-05, "loss": 2.5353, "step": 41261 }, { "epoch": 3.24, "learning_rate": 6.764525993883792e-05, "loss": 2.7369, "step": 41262 }, { "epoch": 3.24, "learning_rate": 6.764447580961342e-05, "loss": 2.584, "step": 41263 }, { "epoch": 3.24, "learning_rate": 6.764369168038894e-05, "loss": 2.7767, "step": 41264 }, { "epoch": 3.24, "learning_rate": 6.764290755116444e-05, "loss": 2.6539, "step": 41265 }, { "epoch": 3.24, "learning_rate": 6.764212342193995e-05, "loss": 2.6284, "step": 41266 }, { "epoch": 3.24, "learning_rate": 6.764133929271545e-05, "loss": 2.7979, "step": 41267 }, { "epoch": 3.24, "learning_rate": 6.764055516349095e-05, "loss": 2.8633, "step": 41268 }, { "epoch": 3.24, "learning_rate": 6.763977103426646e-05, "loss": 2.7014, "step": 41269 }, { "epoch": 3.24, "learning_rate": 6.763898690504196e-05, "loss": 2.6281, "step": 41270 }, { "epoch": 3.24, "learning_rate": 6.763820277581746e-05, "loss": 2.9505, "step": 41271 }, { "epoch": 3.24, "learning_rate": 6.763741864659297e-05, "loss": 2.6929, "step": 41272 }, { "epoch": 3.24, "learning_rate": 6.763663451736847e-05, "loss": 2.6953, "step": 41273 }, { "epoch": 3.24, "learning_rate": 6.763585038814398e-05, "loss": 2.7491, "step": 41274 }, { "epoch": 3.24, "learning_rate": 6.763506625891948e-05, "loss": 2.7878, "step": 41275 }, { "epoch": 3.24, "learning_rate": 6.763428212969498e-05, "loss": 2.798, "step": 41276 }, { "epoch": 3.24, "learning_rate": 6.763349800047049e-05, "loss": 2.7281, "step": 41277 }, { "epoch": 3.24, "learning_rate": 6.763271387124599e-05, "loss": 2.7199, "step": 41278 }, { "epoch": 3.24, "learning_rate": 6.76319297420215e-05, "loss": 2.7654, "step": 41279 }, { "epoch": 3.24, "learning_rate": 6.763114561279698e-05, "loss": 2.7608, "step": 41280 }, { "epoch": 3.24, "learning_rate": 6.763036148357249e-05, "loss": 2.7265, "step": 41281 }, { "epoch": 3.24, "learning_rate": 6.762957735434799e-05, "loss": 2.7844, "step": 41282 }, { "epoch": 3.24, "learning_rate": 6.76287932251235e-05, "loss": 2.8464, "step": 41283 }, { "epoch": 3.24, "learning_rate": 6.7628009095899e-05, "loss": 2.6748, "step": 41284 }, { "epoch": 3.24, "learning_rate": 6.76272249666745e-05, "loss": 2.5059, "step": 41285 }, { "epoch": 3.24, "learning_rate": 6.762644083745e-05, "loss": 2.7542, "step": 41286 }, { "epoch": 3.24, "learning_rate": 6.762565670822551e-05, "loss": 2.506, "step": 41287 }, { "epoch": 3.24, "learning_rate": 6.762487257900101e-05, "loss": 2.705, "step": 41288 }, { "epoch": 3.24, "learning_rate": 6.762408844977653e-05, "loss": 2.7793, "step": 41289 }, { "epoch": 3.24, "learning_rate": 6.762330432055203e-05, "loss": 2.6837, "step": 41290 }, { "epoch": 3.24, "learning_rate": 6.762252019132754e-05, "loss": 2.8203, "step": 41291 }, { "epoch": 3.24, "learning_rate": 6.762173606210304e-05, "loss": 2.8581, "step": 41292 }, { "epoch": 3.24, "learning_rate": 6.762095193287855e-05, "loss": 2.7187, "step": 41293 }, { "epoch": 3.24, "learning_rate": 6.762016780365405e-05, "loss": 2.966, "step": 41294 }, { "epoch": 3.24, "learning_rate": 6.761938367442955e-05, "loss": 2.5765, "step": 41295 }, { "epoch": 3.24, "learning_rate": 6.761859954520506e-05, "loss": 2.9339, "step": 41296 }, { "epoch": 3.24, "learning_rate": 6.761781541598056e-05, "loss": 2.672, "step": 41297 }, { "epoch": 3.24, "learning_rate": 6.761703128675606e-05, "loss": 2.7843, "step": 41298 }, { "epoch": 3.24, "learning_rate": 6.761624715753157e-05, "loss": 2.8477, "step": 41299 }, { "epoch": 3.24, "learning_rate": 6.761546302830707e-05, "loss": 2.6249, "step": 41300 }, { "epoch": 3.24, "learning_rate": 6.761467889908257e-05, "loss": 2.7346, "step": 41301 }, { "epoch": 3.24, "learning_rate": 6.761389476985808e-05, "loss": 2.7825, "step": 41302 }, { "epoch": 3.24, "learning_rate": 6.761311064063358e-05, "loss": 2.7907, "step": 41303 }, { "epoch": 3.24, "learning_rate": 6.761232651140909e-05, "loss": 2.537, "step": 41304 }, { "epoch": 3.24, "learning_rate": 6.761154238218459e-05, "loss": 2.7604, "step": 41305 }, { "epoch": 3.24, "learning_rate": 6.761075825296009e-05, "loss": 2.6593, "step": 41306 }, { "epoch": 3.24, "learning_rate": 6.76099741237356e-05, "loss": 2.7743, "step": 41307 }, { "epoch": 3.24, "learning_rate": 6.76091899945111e-05, "loss": 2.8101, "step": 41308 }, { "epoch": 3.24, "learning_rate": 6.76084058652866e-05, "loss": 2.6948, "step": 41309 }, { "epoch": 3.24, "learning_rate": 6.760762173606211e-05, "loss": 2.5938, "step": 41310 }, { "epoch": 3.24, "learning_rate": 6.760683760683761e-05, "loss": 2.7605, "step": 41311 }, { "epoch": 3.24, "learning_rate": 6.760605347761311e-05, "loss": 2.6415, "step": 41312 }, { "epoch": 3.24, "learning_rate": 6.760526934838862e-05, "loss": 2.6399, "step": 41313 }, { "epoch": 3.24, "learning_rate": 6.760448521916412e-05, "loss": 2.7352, "step": 41314 }, { "epoch": 3.24, "learning_rate": 6.760370108993963e-05, "loss": 2.6391, "step": 41315 }, { "epoch": 3.24, "learning_rate": 6.760291696071513e-05, "loss": 2.8384, "step": 41316 }, { "epoch": 3.24, "learning_rate": 6.760213283149063e-05, "loss": 2.9987, "step": 41317 }, { "epoch": 3.24, "learning_rate": 6.760134870226614e-05, "loss": 2.7766, "step": 41318 }, { "epoch": 3.24, "learning_rate": 6.760056457304164e-05, "loss": 2.6913, "step": 41319 }, { "epoch": 3.24, "learning_rate": 6.759978044381714e-05, "loss": 2.6013, "step": 41320 }, { "epoch": 3.24, "learning_rate": 6.759899631459265e-05, "loss": 2.7955, "step": 41321 }, { "epoch": 3.24, "learning_rate": 6.759821218536815e-05, "loss": 2.6649, "step": 41322 }, { "epoch": 3.24, "learning_rate": 6.759742805614366e-05, "loss": 2.6716, "step": 41323 }, { "epoch": 3.24, "learning_rate": 6.759664392691916e-05, "loss": 2.7357, "step": 41324 }, { "epoch": 3.24, "learning_rate": 6.759585979769466e-05, "loss": 2.8483, "step": 41325 }, { "epoch": 3.24, "learning_rate": 6.759507566847017e-05, "loss": 2.576, "step": 41326 }, { "epoch": 3.24, "learning_rate": 6.759429153924567e-05, "loss": 2.6465, "step": 41327 }, { "epoch": 3.24, "learning_rate": 6.759350741002117e-05, "loss": 2.6827, "step": 41328 }, { "epoch": 3.24, "learning_rate": 6.759272328079668e-05, "loss": 2.6064, "step": 41329 }, { "epoch": 3.24, "learning_rate": 6.759193915157218e-05, "loss": 2.7506, "step": 41330 }, { "epoch": 3.24, "learning_rate": 6.759115502234768e-05, "loss": 2.7063, "step": 41331 }, { "epoch": 3.24, "learning_rate": 6.759037089312319e-05, "loss": 2.6566, "step": 41332 }, { "epoch": 3.24, "learning_rate": 6.758958676389869e-05, "loss": 2.6754, "step": 41333 }, { "epoch": 3.24, "learning_rate": 6.75888026346742e-05, "loss": 2.9088, "step": 41334 }, { "epoch": 3.24, "learning_rate": 6.75880185054497e-05, "loss": 2.9043, "step": 41335 }, { "epoch": 3.24, "learning_rate": 6.75872343762252e-05, "loss": 2.8985, "step": 41336 }, { "epoch": 3.24, "learning_rate": 6.758645024700071e-05, "loss": 2.683, "step": 41337 }, { "epoch": 3.24, "learning_rate": 6.758566611777621e-05, "loss": 2.8081, "step": 41338 }, { "epoch": 3.24, "learning_rate": 6.758488198855171e-05, "loss": 2.6729, "step": 41339 }, { "epoch": 3.24, "learning_rate": 6.758409785932722e-05, "loss": 2.8327, "step": 41340 }, { "epoch": 3.24, "learning_rate": 6.758331373010272e-05, "loss": 2.6124, "step": 41341 }, { "epoch": 3.24, "learning_rate": 6.758252960087823e-05, "loss": 2.6251, "step": 41342 }, { "epoch": 3.24, "learning_rate": 6.758174547165373e-05, "loss": 2.9573, "step": 41343 }, { "epoch": 3.24, "learning_rate": 6.758096134242923e-05, "loss": 2.6198, "step": 41344 }, { "epoch": 3.24, "learning_rate": 6.758017721320474e-05, "loss": 2.5619, "step": 41345 }, { "epoch": 3.24, "learning_rate": 6.757939308398025e-05, "loss": 2.7173, "step": 41346 }, { "epoch": 3.24, "learning_rate": 6.757860895475576e-05, "loss": 2.8159, "step": 41347 }, { "epoch": 3.24, "learning_rate": 6.757782482553126e-05, "loss": 2.8455, "step": 41348 }, { "epoch": 3.24, "learning_rate": 6.757704069630676e-05, "loss": 2.8047, "step": 41349 }, { "epoch": 3.24, "learning_rate": 6.757625656708227e-05, "loss": 2.8905, "step": 41350 }, { "epoch": 3.24, "learning_rate": 6.757547243785777e-05, "loss": 2.8731, "step": 41351 }, { "epoch": 3.24, "learning_rate": 6.757468830863326e-05, "loss": 2.792, "step": 41352 }, { "epoch": 3.24, "learning_rate": 6.757390417940877e-05, "loss": 2.6377, "step": 41353 }, { "epoch": 3.24, "learning_rate": 6.757312005018427e-05, "loss": 2.508, "step": 41354 }, { "epoch": 3.24, "learning_rate": 6.757233592095977e-05, "loss": 2.6622, "step": 41355 }, { "epoch": 3.24, "learning_rate": 6.757155179173528e-05, "loss": 2.5473, "step": 41356 }, { "epoch": 3.24, "learning_rate": 6.757076766251078e-05, "loss": 2.9891, "step": 41357 }, { "epoch": 3.24, "learning_rate": 6.756998353328628e-05, "loss": 2.9508, "step": 41358 }, { "epoch": 3.24, "learning_rate": 6.756919940406179e-05, "loss": 2.7161, "step": 41359 }, { "epoch": 3.24, "learning_rate": 6.756841527483729e-05, "loss": 2.8339, "step": 41360 }, { "epoch": 3.24, "learning_rate": 6.75676311456128e-05, "loss": 2.6244, "step": 41361 }, { "epoch": 3.24, "learning_rate": 6.75668470163883e-05, "loss": 2.7463, "step": 41362 }, { "epoch": 3.24, "learning_rate": 6.75660628871638e-05, "loss": 2.75, "step": 41363 }, { "epoch": 3.24, "learning_rate": 6.75652787579393e-05, "loss": 2.73, "step": 41364 }, { "epoch": 3.24, "learning_rate": 6.756449462871481e-05, "loss": 2.7213, "step": 41365 }, { "epoch": 3.24, "learning_rate": 6.756371049949031e-05, "loss": 2.7053, "step": 41366 }, { "epoch": 3.24, "learning_rate": 6.756292637026582e-05, "loss": 2.6453, "step": 41367 }, { "epoch": 3.24, "learning_rate": 6.756214224104132e-05, "loss": 2.7045, "step": 41368 }, { "epoch": 3.24, "learning_rate": 6.756135811181682e-05, "loss": 2.6819, "step": 41369 }, { "epoch": 3.24, "learning_rate": 6.756057398259233e-05, "loss": 2.7529, "step": 41370 }, { "epoch": 3.24, "learning_rate": 6.755978985336783e-05, "loss": 2.559, "step": 41371 }, { "epoch": 3.24, "learning_rate": 6.755900572414335e-05, "loss": 2.7068, "step": 41372 }, { "epoch": 3.24, "learning_rate": 6.755822159491885e-05, "loss": 2.5552, "step": 41373 }, { "epoch": 3.24, "learning_rate": 6.755743746569436e-05, "loss": 2.7487, "step": 41374 }, { "epoch": 3.24, "learning_rate": 6.755665333646986e-05, "loss": 2.7533, "step": 41375 }, { "epoch": 3.24, "learning_rate": 6.755586920724536e-05, "loss": 2.7693, "step": 41376 }, { "epoch": 3.24, "learning_rate": 6.755508507802087e-05, "loss": 2.9092, "step": 41377 }, { "epoch": 3.24, "learning_rate": 6.755430094879637e-05, "loss": 2.7671, "step": 41378 }, { "epoch": 3.24, "learning_rate": 6.755351681957187e-05, "loss": 2.4924, "step": 41379 }, { "epoch": 3.24, "learning_rate": 6.755273269034738e-05, "loss": 2.7365, "step": 41380 }, { "epoch": 3.24, "learning_rate": 6.755194856112288e-05, "loss": 2.8518, "step": 41381 }, { "epoch": 3.24, "learning_rate": 6.755116443189839e-05, "loss": 2.5775, "step": 41382 }, { "epoch": 3.24, "learning_rate": 6.755038030267389e-05, "loss": 2.7649, "step": 41383 }, { "epoch": 3.25, "learning_rate": 6.75495961734494e-05, "loss": 2.6865, "step": 41384 }, { "epoch": 3.25, "learning_rate": 6.75488120442249e-05, "loss": 2.6619, "step": 41385 }, { "epoch": 3.25, "learning_rate": 6.75480279150004e-05, "loss": 2.7668, "step": 41386 }, { "epoch": 3.25, "learning_rate": 6.75472437857759e-05, "loss": 2.6838, "step": 41387 }, { "epoch": 3.25, "learning_rate": 6.75464596565514e-05, "loss": 2.8506, "step": 41388 }, { "epoch": 3.25, "learning_rate": 6.75456755273269e-05, "loss": 2.6961, "step": 41389 }, { "epoch": 3.25, "learning_rate": 6.75448913981024e-05, "loss": 2.7471, "step": 41390 }, { "epoch": 3.25, "learning_rate": 6.75441072688779e-05, "loss": 2.8046, "step": 41391 }, { "epoch": 3.25, "learning_rate": 6.754332313965341e-05, "loss": 2.4725, "step": 41392 }, { "epoch": 3.25, "learning_rate": 6.754253901042891e-05, "loss": 2.6021, "step": 41393 }, { "epoch": 3.25, "learning_rate": 6.754175488120442e-05, "loss": 2.5535, "step": 41394 }, { "epoch": 3.25, "learning_rate": 6.754097075197992e-05, "loss": 2.5407, "step": 41395 }, { "epoch": 3.25, "learning_rate": 6.754018662275542e-05, "loss": 2.5126, "step": 41396 }, { "epoch": 3.25, "learning_rate": 6.753940249353094e-05, "loss": 2.6036, "step": 41397 }, { "epoch": 3.25, "learning_rate": 6.753861836430644e-05, "loss": 2.7198, "step": 41398 }, { "epoch": 3.25, "learning_rate": 6.753783423508195e-05, "loss": 2.6366, "step": 41399 }, { "epoch": 3.25, "learning_rate": 6.753705010585745e-05, "loss": 2.5331, "step": 41400 }, { "epoch": 3.25, "learning_rate": 6.753626597663296e-05, "loss": 2.6639, "step": 41401 }, { "epoch": 3.25, "learning_rate": 6.753548184740846e-05, "loss": 2.5676, "step": 41402 }, { "epoch": 3.25, "learning_rate": 6.753469771818396e-05, "loss": 2.6536, "step": 41403 }, { "epoch": 3.25, "learning_rate": 6.753391358895947e-05, "loss": 2.6545, "step": 41404 }, { "epoch": 3.25, "learning_rate": 6.753312945973497e-05, "loss": 2.629, "step": 41405 }, { "epoch": 3.25, "learning_rate": 6.753234533051047e-05, "loss": 2.6743, "step": 41406 }, { "epoch": 3.25, "learning_rate": 6.753156120128598e-05, "loss": 2.7392, "step": 41407 }, { "epoch": 3.25, "learning_rate": 6.753077707206148e-05, "loss": 2.52, "step": 41408 }, { "epoch": 3.25, "learning_rate": 6.752999294283699e-05, "loss": 2.7927, "step": 41409 }, { "epoch": 3.25, "learning_rate": 6.752920881361249e-05, "loss": 2.7576, "step": 41410 }, { "epoch": 3.25, "learning_rate": 6.752842468438799e-05, "loss": 2.6079, "step": 41411 }, { "epoch": 3.25, "learning_rate": 6.75276405551635e-05, "loss": 2.669, "step": 41412 }, { "epoch": 3.25, "learning_rate": 6.7526856425939e-05, "loss": 2.7859, "step": 41413 }, { "epoch": 3.25, "learning_rate": 6.75260722967145e-05, "loss": 2.3958, "step": 41414 }, { "epoch": 3.25, "learning_rate": 6.752528816749001e-05, "loss": 2.8286, "step": 41415 }, { "epoch": 3.25, "learning_rate": 6.752450403826551e-05, "loss": 2.7962, "step": 41416 }, { "epoch": 3.25, "learning_rate": 6.752371990904101e-05, "loss": 2.7144, "step": 41417 }, { "epoch": 3.25, "learning_rate": 6.752293577981652e-05, "loss": 2.4956, "step": 41418 }, { "epoch": 3.25, "learning_rate": 6.752215165059202e-05, "loss": 2.5672, "step": 41419 }, { "epoch": 3.25, "learning_rate": 6.752136752136753e-05, "loss": 2.6882, "step": 41420 }, { "epoch": 3.25, "learning_rate": 6.752058339214303e-05, "loss": 2.6765, "step": 41421 }, { "epoch": 3.25, "learning_rate": 6.751979926291853e-05, "loss": 2.747, "step": 41422 }, { "epoch": 3.25, "learning_rate": 6.751901513369404e-05, "loss": 2.5511, "step": 41423 }, { "epoch": 3.25, "learning_rate": 6.751823100446954e-05, "loss": 2.6934, "step": 41424 }, { "epoch": 3.25, "learning_rate": 6.751744687524504e-05, "loss": 2.6958, "step": 41425 }, { "epoch": 3.25, "learning_rate": 6.751666274602055e-05, "loss": 2.5857, "step": 41426 }, { "epoch": 3.25, "learning_rate": 6.751587861679605e-05, "loss": 2.6465, "step": 41427 }, { "epoch": 3.25, "learning_rate": 6.751509448757156e-05, "loss": 2.765, "step": 41428 }, { "epoch": 3.25, "learning_rate": 6.751431035834706e-05, "loss": 2.6752, "step": 41429 }, { "epoch": 3.25, "learning_rate": 6.751352622912256e-05, "loss": 2.5831, "step": 41430 }, { "epoch": 3.25, "learning_rate": 6.751274209989807e-05, "loss": 2.7808, "step": 41431 }, { "epoch": 3.25, "learning_rate": 6.751195797067357e-05, "loss": 2.8299, "step": 41432 }, { "epoch": 3.25, "learning_rate": 6.751117384144907e-05, "loss": 2.9231, "step": 41433 }, { "epoch": 3.25, "learning_rate": 6.751038971222458e-05, "loss": 2.899, "step": 41434 }, { "epoch": 3.25, "learning_rate": 6.750960558300008e-05, "loss": 2.7598, "step": 41435 }, { "epoch": 3.25, "learning_rate": 6.750882145377558e-05, "loss": 2.8111, "step": 41436 }, { "epoch": 3.25, "learning_rate": 6.750803732455109e-05, "loss": 2.6838, "step": 41437 }, { "epoch": 3.25, "learning_rate": 6.750725319532659e-05, "loss": 2.6278, "step": 41438 }, { "epoch": 3.25, "learning_rate": 6.75064690661021e-05, "loss": 2.8499, "step": 41439 }, { "epoch": 3.25, "learning_rate": 6.75056849368776e-05, "loss": 2.6668, "step": 41440 }, { "epoch": 3.25, "learning_rate": 6.75049008076531e-05, "loss": 2.6637, "step": 41441 }, { "epoch": 3.25, "learning_rate": 6.75041166784286e-05, "loss": 2.6274, "step": 41442 }, { "epoch": 3.25, "learning_rate": 6.750333254920411e-05, "loss": 2.8298, "step": 41443 }, { "epoch": 3.25, "learning_rate": 6.750254841997961e-05, "loss": 2.9822, "step": 41444 }, { "epoch": 3.25, "learning_rate": 6.750176429075512e-05, "loss": 2.7956, "step": 41445 }, { "epoch": 3.25, "learning_rate": 6.750098016153062e-05, "loss": 2.7876, "step": 41446 }, { "epoch": 3.25, "learning_rate": 6.750019603230612e-05, "loss": 2.5077, "step": 41447 }, { "epoch": 3.25, "learning_rate": 6.749941190308163e-05, "loss": 2.5843, "step": 41448 }, { "epoch": 3.25, "learning_rate": 6.749862777385713e-05, "loss": 2.8683, "step": 41449 }, { "epoch": 3.25, "learning_rate": 6.749784364463264e-05, "loss": 2.4817, "step": 41450 }, { "epoch": 3.25, "learning_rate": 6.749705951540814e-05, "loss": 2.7164, "step": 41451 }, { "epoch": 3.25, "learning_rate": 6.749627538618364e-05, "loss": 2.8341, "step": 41452 }, { "epoch": 3.25, "learning_rate": 6.749549125695915e-05, "loss": 2.564, "step": 41453 }, { "epoch": 3.25, "learning_rate": 6.749470712773466e-05, "loss": 2.6453, "step": 41454 }, { "epoch": 3.25, "learning_rate": 6.749392299851017e-05, "loss": 2.9044, "step": 41455 }, { "epoch": 3.25, "learning_rate": 6.749313886928567e-05, "loss": 2.586, "step": 41456 }, { "epoch": 3.25, "learning_rate": 6.749235474006118e-05, "loss": 2.96, "step": 41457 }, { "epoch": 3.25, "learning_rate": 6.749157061083668e-05, "loss": 2.7317, "step": 41458 }, { "epoch": 3.25, "learning_rate": 6.749078648161218e-05, "loss": 2.7789, "step": 41459 }, { "epoch": 3.25, "learning_rate": 6.749000235238767e-05, "loss": 2.8185, "step": 41460 }, { "epoch": 3.25, "learning_rate": 6.748921822316318e-05, "loss": 2.6726, "step": 41461 }, { "epoch": 3.25, "learning_rate": 6.748843409393868e-05, "loss": 2.4745, "step": 41462 }, { "epoch": 3.25, "learning_rate": 6.748764996471418e-05, "loss": 2.6789, "step": 41463 }, { "epoch": 3.25, "learning_rate": 6.748686583548969e-05, "loss": 2.5943, "step": 41464 }, { "epoch": 3.25, "learning_rate": 6.748608170626519e-05, "loss": 2.7989, "step": 41465 }, { "epoch": 3.25, "learning_rate": 6.74852975770407e-05, "loss": 2.7756, "step": 41466 }, { "epoch": 3.25, "learning_rate": 6.74845134478162e-05, "loss": 2.7437, "step": 41467 }, { "epoch": 3.25, "learning_rate": 6.74837293185917e-05, "loss": 2.8678, "step": 41468 }, { "epoch": 3.25, "learning_rate": 6.74829451893672e-05, "loss": 2.9516, "step": 41469 }, { "epoch": 3.25, "learning_rate": 6.748216106014271e-05, "loss": 2.6699, "step": 41470 }, { "epoch": 3.25, "learning_rate": 6.748137693091821e-05, "loss": 2.6671, "step": 41471 }, { "epoch": 3.25, "learning_rate": 6.748059280169372e-05, "loss": 2.7255, "step": 41472 }, { "epoch": 3.25, "learning_rate": 6.747980867246922e-05, "loss": 2.5918, "step": 41473 }, { "epoch": 3.25, "learning_rate": 6.747902454324472e-05, "loss": 2.746, "step": 41474 }, { "epoch": 3.25, "learning_rate": 6.747824041402023e-05, "loss": 2.753, "step": 41475 }, { "epoch": 3.25, "learning_rate": 6.747745628479573e-05, "loss": 2.6602, "step": 41476 }, { "epoch": 3.25, "learning_rate": 6.747667215557124e-05, "loss": 2.8586, "step": 41477 }, { "epoch": 3.25, "learning_rate": 6.747588802634674e-05, "loss": 2.6163, "step": 41478 }, { "epoch": 3.25, "learning_rate": 6.747510389712224e-05, "loss": 2.6976, "step": 41479 }, { "epoch": 3.25, "learning_rate": 6.747431976789776e-05, "loss": 2.9589, "step": 41480 }, { "epoch": 3.25, "learning_rate": 6.747353563867326e-05, "loss": 2.6208, "step": 41481 }, { "epoch": 3.25, "learning_rate": 6.747275150944877e-05, "loss": 2.5922, "step": 41482 }, { "epoch": 3.25, "learning_rate": 6.747196738022427e-05, "loss": 2.8033, "step": 41483 }, { "epoch": 3.25, "learning_rate": 6.747118325099977e-05, "loss": 3.0327, "step": 41484 }, { "epoch": 3.25, "learning_rate": 6.747039912177528e-05, "loss": 2.873, "step": 41485 }, { "epoch": 3.25, "learning_rate": 6.746961499255078e-05, "loss": 2.5288, "step": 41486 }, { "epoch": 3.25, "learning_rate": 6.746883086332629e-05, "loss": 2.8233, "step": 41487 }, { "epoch": 3.25, "learning_rate": 6.746804673410179e-05, "loss": 2.7334, "step": 41488 }, { "epoch": 3.25, "learning_rate": 6.746726260487729e-05, "loss": 2.8803, "step": 41489 }, { "epoch": 3.25, "learning_rate": 6.74664784756528e-05, "loss": 2.7201, "step": 41490 }, { "epoch": 3.25, "learning_rate": 6.74656943464283e-05, "loss": 2.8592, "step": 41491 }, { "epoch": 3.25, "learning_rate": 6.74649102172038e-05, "loss": 2.6414, "step": 41492 }, { "epoch": 3.25, "learning_rate": 6.746412608797931e-05, "loss": 2.8043, "step": 41493 }, { "epoch": 3.25, "learning_rate": 6.746334195875481e-05, "loss": 2.7578, "step": 41494 }, { "epoch": 3.25, "learning_rate": 6.746255782953032e-05, "loss": 2.5724, "step": 41495 }, { "epoch": 3.25, "learning_rate": 6.74617737003058e-05, "loss": 2.4806, "step": 41496 }, { "epoch": 3.25, "learning_rate": 6.746098957108131e-05, "loss": 2.9185, "step": 41497 }, { "epoch": 3.25, "learning_rate": 6.746020544185681e-05, "loss": 2.7515, "step": 41498 }, { "epoch": 3.25, "learning_rate": 6.745942131263232e-05, "loss": 2.8014, "step": 41499 }, { "epoch": 3.25, "learning_rate": 6.745863718340782e-05, "loss": 2.6394, "step": 41500 }, { "epoch": 3.25, "learning_rate": 6.745785305418332e-05, "loss": 2.6642, "step": 41501 }, { "epoch": 3.25, "learning_rate": 6.745706892495883e-05, "loss": 2.652, "step": 41502 }, { "epoch": 3.25, "learning_rate": 6.745628479573433e-05, "loss": 2.6943, "step": 41503 }, { "epoch": 3.25, "learning_rate": 6.745550066650983e-05, "loss": 2.7324, "step": 41504 }, { "epoch": 3.25, "learning_rate": 6.745471653728535e-05, "loss": 2.7103, "step": 41505 }, { "epoch": 3.25, "learning_rate": 6.745393240806086e-05, "loss": 2.679, "step": 41506 }, { "epoch": 3.25, "learning_rate": 6.745314827883636e-05, "loss": 2.5877, "step": 41507 }, { "epoch": 3.25, "learning_rate": 6.745236414961186e-05, "loss": 2.5786, "step": 41508 }, { "epoch": 3.25, "learning_rate": 6.745158002038737e-05, "loss": 2.8052, "step": 41509 }, { "epoch": 3.25, "learning_rate": 6.745079589116287e-05, "loss": 2.8412, "step": 41510 }, { "epoch": 3.25, "learning_rate": 6.745001176193837e-05, "loss": 2.6848, "step": 41511 }, { "epoch": 3.26, "learning_rate": 6.744922763271388e-05, "loss": 2.7304, "step": 41512 }, { "epoch": 3.26, "learning_rate": 6.744844350348938e-05, "loss": 2.6145, "step": 41513 }, { "epoch": 3.26, "learning_rate": 6.744765937426488e-05, "loss": 2.6769, "step": 41514 }, { "epoch": 3.26, "learning_rate": 6.744687524504039e-05, "loss": 2.9028, "step": 41515 }, { "epoch": 3.26, "learning_rate": 6.744609111581589e-05, "loss": 2.7437, "step": 41516 }, { "epoch": 3.26, "learning_rate": 6.74453069865914e-05, "loss": 2.9786, "step": 41517 }, { "epoch": 3.26, "learning_rate": 6.74445228573669e-05, "loss": 2.6334, "step": 41518 }, { "epoch": 3.26, "learning_rate": 6.74437387281424e-05, "loss": 2.9006, "step": 41519 }, { "epoch": 3.26, "learning_rate": 6.744295459891791e-05, "loss": 2.9777, "step": 41520 }, { "epoch": 3.26, "learning_rate": 6.744217046969341e-05, "loss": 2.7065, "step": 41521 }, { "epoch": 3.26, "learning_rate": 6.744138634046891e-05, "loss": 2.7762, "step": 41522 }, { "epoch": 3.26, "learning_rate": 6.744060221124442e-05, "loss": 2.8352, "step": 41523 }, { "epoch": 3.26, "learning_rate": 6.743981808201992e-05, "loss": 2.874, "step": 41524 }, { "epoch": 3.26, "learning_rate": 6.743903395279543e-05, "loss": 2.8893, "step": 41525 }, { "epoch": 3.26, "learning_rate": 6.743824982357093e-05, "loss": 2.539, "step": 41526 }, { "epoch": 3.26, "learning_rate": 6.743746569434643e-05, "loss": 2.9234, "step": 41527 }, { "epoch": 3.26, "learning_rate": 6.743668156512194e-05, "loss": 2.6349, "step": 41528 }, { "epoch": 3.26, "learning_rate": 6.743589743589744e-05, "loss": 2.6458, "step": 41529 }, { "epoch": 3.26, "learning_rate": 6.743511330667294e-05, "loss": 2.6296, "step": 41530 }, { "epoch": 3.26, "learning_rate": 6.743432917744845e-05, "loss": 2.8246, "step": 41531 }, { "epoch": 3.26, "learning_rate": 6.743354504822395e-05, "loss": 2.6484, "step": 41532 }, { "epoch": 3.26, "learning_rate": 6.743276091899945e-05, "loss": 2.6569, "step": 41533 }, { "epoch": 3.26, "learning_rate": 6.743197678977496e-05, "loss": 2.7397, "step": 41534 }, { "epoch": 3.26, "learning_rate": 6.743119266055046e-05, "loss": 2.8358, "step": 41535 }, { "epoch": 3.26, "learning_rate": 6.743040853132597e-05, "loss": 2.8409, "step": 41536 }, { "epoch": 3.26, "learning_rate": 6.742962440210147e-05, "loss": 2.5781, "step": 41537 }, { "epoch": 3.26, "learning_rate": 6.742884027287697e-05, "loss": 2.662, "step": 41538 }, { "epoch": 3.26, "learning_rate": 6.742805614365248e-05, "loss": 2.7675, "step": 41539 }, { "epoch": 3.26, "learning_rate": 6.742727201442798e-05, "loss": 2.8459, "step": 41540 }, { "epoch": 3.26, "learning_rate": 6.742648788520348e-05, "loss": 2.6187, "step": 41541 }, { "epoch": 3.26, "learning_rate": 6.742570375597899e-05, "loss": 2.8919, "step": 41542 }, { "epoch": 3.26, "learning_rate": 6.742491962675449e-05, "loss": 2.9059, "step": 41543 }, { "epoch": 3.26, "learning_rate": 6.742413549753e-05, "loss": 2.9709, "step": 41544 }, { "epoch": 3.26, "learning_rate": 6.74233513683055e-05, "loss": 2.6607, "step": 41545 }, { "epoch": 3.26, "learning_rate": 6.7422567239081e-05, "loss": 2.8533, "step": 41546 }, { "epoch": 3.26, "learning_rate": 6.74217831098565e-05, "loss": 2.8373, "step": 41547 }, { "epoch": 3.26, "learning_rate": 6.742099898063201e-05, "loss": 2.9699, "step": 41548 }, { "epoch": 3.26, "learning_rate": 6.742021485140751e-05, "loss": 2.6102, "step": 41549 }, { "epoch": 3.26, "learning_rate": 6.741943072218302e-05, "loss": 2.6117, "step": 41550 }, { "epoch": 3.26, "learning_rate": 6.741864659295852e-05, "loss": 2.7764, "step": 41551 }, { "epoch": 3.26, "learning_rate": 6.741786246373402e-05, "loss": 2.7108, "step": 41552 }, { "epoch": 3.26, "learning_rate": 6.741707833450953e-05, "loss": 2.7069, "step": 41553 }, { "epoch": 3.26, "learning_rate": 6.741629420528503e-05, "loss": 2.7037, "step": 41554 }, { "epoch": 3.26, "learning_rate": 6.741551007606054e-05, "loss": 2.572, "step": 41555 }, { "epoch": 3.26, "learning_rate": 6.741472594683604e-05, "loss": 2.6128, "step": 41556 }, { "epoch": 3.26, "learning_rate": 6.741394181761154e-05, "loss": 2.8909, "step": 41557 }, { "epoch": 3.26, "learning_rate": 6.741315768838705e-05, "loss": 2.7384, "step": 41558 }, { "epoch": 3.26, "learning_rate": 6.741237355916255e-05, "loss": 2.716, "step": 41559 }, { "epoch": 3.26, "learning_rate": 6.741158942993805e-05, "loss": 2.6422, "step": 41560 }, { "epoch": 3.26, "learning_rate": 6.741080530071356e-05, "loss": 2.6791, "step": 41561 }, { "epoch": 3.26, "learning_rate": 6.741002117148907e-05, "loss": 2.7496, "step": 41562 }, { "epoch": 3.26, "learning_rate": 6.740923704226458e-05, "loss": 2.9536, "step": 41563 }, { "epoch": 3.26, "learning_rate": 6.740845291304008e-05, "loss": 2.9928, "step": 41564 }, { "epoch": 3.26, "learning_rate": 6.740766878381559e-05, "loss": 2.9114, "step": 41565 }, { "epoch": 3.26, "learning_rate": 6.740688465459109e-05, "loss": 2.5685, "step": 41566 }, { "epoch": 3.26, "learning_rate": 6.74061005253666e-05, "loss": 2.6038, "step": 41567 }, { "epoch": 3.26, "learning_rate": 6.740531639614208e-05, "loss": 3.1847, "step": 41568 }, { "epoch": 3.26, "learning_rate": 6.740453226691759e-05, "loss": 2.5266, "step": 41569 }, { "epoch": 3.26, "learning_rate": 6.740374813769309e-05, "loss": 2.7991, "step": 41570 }, { "epoch": 3.26, "learning_rate": 6.74029640084686e-05, "loss": 2.7478, "step": 41571 }, { "epoch": 3.26, "learning_rate": 6.74021798792441e-05, "loss": 3.0936, "step": 41572 }, { "epoch": 3.26, "learning_rate": 6.74013957500196e-05, "loss": 2.4983, "step": 41573 }, { "epoch": 3.26, "learning_rate": 6.74006116207951e-05, "loss": 2.7479, "step": 41574 }, { "epoch": 3.26, "learning_rate": 6.739982749157061e-05, "loss": 2.7392, "step": 41575 }, { "epoch": 3.26, "learning_rate": 6.739904336234611e-05, "loss": 2.8655, "step": 41576 }, { "epoch": 3.26, "learning_rate": 6.739825923312162e-05, "loss": 2.6732, "step": 41577 }, { "epoch": 3.26, "learning_rate": 6.739747510389712e-05, "loss": 2.8254, "step": 41578 }, { "epoch": 3.26, "learning_rate": 6.739669097467262e-05, "loss": 2.708, "step": 41579 }, { "epoch": 3.26, "learning_rate": 6.739590684544813e-05, "loss": 2.8081, "step": 41580 }, { "epoch": 3.26, "learning_rate": 6.739512271622363e-05, "loss": 3.0558, "step": 41581 }, { "epoch": 3.26, "learning_rate": 6.739433858699913e-05, "loss": 2.6998, "step": 41582 }, { "epoch": 3.26, "learning_rate": 6.739355445777464e-05, "loss": 2.7526, "step": 41583 }, { "epoch": 3.26, "learning_rate": 6.739277032855014e-05, "loss": 2.7018, "step": 41584 }, { "epoch": 3.26, "learning_rate": 6.739198619932565e-05, "loss": 2.8488, "step": 41585 }, { "epoch": 3.26, "learning_rate": 6.739120207010115e-05, "loss": 2.7287, "step": 41586 }, { "epoch": 3.26, "learning_rate": 6.739041794087665e-05, "loss": 2.6311, "step": 41587 }, { "epoch": 3.26, "learning_rate": 6.738963381165217e-05, "loss": 2.7589, "step": 41588 }, { "epoch": 3.26, "learning_rate": 6.738884968242767e-05, "loss": 2.6196, "step": 41589 }, { "epoch": 3.26, "learning_rate": 6.738806555320318e-05, "loss": 2.6181, "step": 41590 }, { "epoch": 3.26, "learning_rate": 6.738728142397868e-05, "loss": 2.7267, "step": 41591 }, { "epoch": 3.26, "learning_rate": 6.738649729475419e-05, "loss": 2.5607, "step": 41592 }, { "epoch": 3.26, "learning_rate": 6.738571316552969e-05, "loss": 2.6932, "step": 41593 }, { "epoch": 3.26, "learning_rate": 6.738492903630519e-05, "loss": 2.688, "step": 41594 }, { "epoch": 3.26, "learning_rate": 6.73841449070807e-05, "loss": 2.6409, "step": 41595 }, { "epoch": 3.26, "learning_rate": 6.73833607778562e-05, "loss": 2.5784, "step": 41596 }, { "epoch": 3.26, "learning_rate": 6.73825766486317e-05, "loss": 3.0245, "step": 41597 }, { "epoch": 3.26, "learning_rate": 6.738179251940721e-05, "loss": 2.5102, "step": 41598 }, { "epoch": 3.26, "learning_rate": 6.738100839018271e-05, "loss": 2.983, "step": 41599 }, { "epoch": 3.26, "learning_rate": 6.738022426095821e-05, "loss": 2.8306, "step": 41600 }, { "epoch": 3.26, "learning_rate": 6.737944013173372e-05, "loss": 2.7711, "step": 41601 }, { "epoch": 3.26, "learning_rate": 6.737865600250922e-05, "loss": 2.7005, "step": 41602 }, { "epoch": 3.26, "learning_rate": 6.737787187328473e-05, "loss": 2.7195, "step": 41603 }, { "epoch": 3.26, "learning_rate": 6.737708774406022e-05, "loss": 2.7226, "step": 41604 }, { "epoch": 3.26, "learning_rate": 6.737630361483572e-05, "loss": 2.6976, "step": 41605 }, { "epoch": 3.26, "learning_rate": 6.737551948561122e-05, "loss": 2.692, "step": 41606 }, { "epoch": 3.26, "learning_rate": 6.737473535638673e-05, "loss": 2.9388, "step": 41607 }, { "epoch": 3.26, "learning_rate": 6.737395122716223e-05, "loss": 2.7898, "step": 41608 }, { "epoch": 3.26, "learning_rate": 6.737316709793773e-05, "loss": 2.8386, "step": 41609 }, { "epoch": 3.26, "learning_rate": 6.737238296871324e-05, "loss": 2.751, "step": 41610 }, { "epoch": 3.26, "learning_rate": 6.737159883948874e-05, "loss": 2.8352, "step": 41611 }, { "epoch": 3.26, "learning_rate": 6.737081471026425e-05, "loss": 2.5737, "step": 41612 }, { "epoch": 3.26, "learning_rate": 6.737003058103976e-05, "loss": 2.7462, "step": 41613 }, { "epoch": 3.26, "learning_rate": 6.736924645181527e-05, "loss": 2.8303, "step": 41614 }, { "epoch": 3.26, "learning_rate": 6.736846232259077e-05, "loss": 2.8428, "step": 41615 }, { "epoch": 3.26, "learning_rate": 6.736767819336627e-05, "loss": 2.6066, "step": 41616 }, { "epoch": 3.26, "learning_rate": 6.736689406414178e-05, "loss": 2.7585, "step": 41617 }, { "epoch": 3.26, "learning_rate": 6.736610993491728e-05, "loss": 2.9766, "step": 41618 }, { "epoch": 3.26, "learning_rate": 6.736532580569278e-05, "loss": 2.617, "step": 41619 }, { "epoch": 3.26, "learning_rate": 6.736454167646829e-05, "loss": 2.7009, "step": 41620 }, { "epoch": 3.26, "learning_rate": 6.736375754724379e-05, "loss": 2.6471, "step": 41621 }, { "epoch": 3.26, "learning_rate": 6.73629734180193e-05, "loss": 2.9366, "step": 41622 }, { "epoch": 3.26, "learning_rate": 6.73621892887948e-05, "loss": 2.7153, "step": 41623 }, { "epoch": 3.26, "learning_rate": 6.73614051595703e-05, "loss": 2.8339, "step": 41624 }, { "epoch": 3.26, "learning_rate": 6.73606210303458e-05, "loss": 2.7117, "step": 41625 }, { "epoch": 3.26, "learning_rate": 6.735983690112131e-05, "loss": 2.6274, "step": 41626 }, { "epoch": 3.26, "learning_rate": 6.735905277189681e-05, "loss": 2.7581, "step": 41627 }, { "epoch": 3.26, "learning_rate": 6.735826864267232e-05, "loss": 2.8214, "step": 41628 }, { "epoch": 3.26, "learning_rate": 6.735748451344782e-05, "loss": 2.6668, "step": 41629 }, { "epoch": 3.26, "learning_rate": 6.735670038422333e-05, "loss": 2.676, "step": 41630 }, { "epoch": 3.26, "learning_rate": 6.735591625499883e-05, "loss": 2.8637, "step": 41631 }, { "epoch": 3.26, "learning_rate": 6.735513212577433e-05, "loss": 2.7513, "step": 41632 }, { "epoch": 3.26, "learning_rate": 6.735434799654984e-05, "loss": 2.8432, "step": 41633 }, { "epoch": 3.26, "learning_rate": 6.735356386732534e-05, "loss": 2.7424, "step": 41634 }, { "epoch": 3.26, "learning_rate": 6.735277973810084e-05, "loss": 2.7219, "step": 41635 }, { "epoch": 3.26, "learning_rate": 6.735199560887635e-05, "loss": 2.7459, "step": 41636 }, { "epoch": 3.26, "learning_rate": 6.735121147965185e-05, "loss": 2.626, "step": 41637 }, { "epoch": 3.26, "learning_rate": 6.735042735042735e-05, "loss": 2.8292, "step": 41638 }, { "epoch": 3.27, "learning_rate": 6.734964322120286e-05, "loss": 2.8526, "step": 41639 }, { "epoch": 3.27, "learning_rate": 6.734885909197836e-05, "loss": 2.77, "step": 41640 }, { "epoch": 3.27, "learning_rate": 6.734807496275387e-05, "loss": 2.4625, "step": 41641 }, { "epoch": 3.27, "learning_rate": 6.734729083352937e-05, "loss": 2.5228, "step": 41642 }, { "epoch": 3.27, "learning_rate": 6.734650670430487e-05, "loss": 3.1652, "step": 41643 }, { "epoch": 3.27, "learning_rate": 6.734572257508038e-05, "loss": 2.6815, "step": 41644 }, { "epoch": 3.27, "learning_rate": 6.734493844585588e-05, "loss": 2.5957, "step": 41645 }, { "epoch": 3.27, "learning_rate": 6.734415431663138e-05, "loss": 2.7786, "step": 41646 }, { "epoch": 3.27, "learning_rate": 6.734337018740689e-05, "loss": 2.73, "step": 41647 }, { "epoch": 3.27, "learning_rate": 6.734258605818239e-05, "loss": 2.5272, "step": 41648 }, { "epoch": 3.27, "learning_rate": 6.73418019289579e-05, "loss": 2.7524, "step": 41649 }, { "epoch": 3.27, "learning_rate": 6.73410177997334e-05, "loss": 2.5061, "step": 41650 }, { "epoch": 3.27, "learning_rate": 6.73402336705089e-05, "loss": 2.8313, "step": 41651 }, { "epoch": 3.27, "learning_rate": 6.73394495412844e-05, "loss": 2.7761, "step": 41652 }, { "epoch": 3.27, "learning_rate": 6.733866541205991e-05, "loss": 2.6019, "step": 41653 }, { "epoch": 3.27, "learning_rate": 6.733788128283541e-05, "loss": 2.621, "step": 41654 }, { "epoch": 3.27, "learning_rate": 6.733709715361092e-05, "loss": 2.7203, "step": 41655 }, { "epoch": 3.27, "learning_rate": 6.733631302438642e-05, "loss": 2.9535, "step": 41656 }, { "epoch": 3.27, "learning_rate": 6.733552889516192e-05, "loss": 2.6592, "step": 41657 }, { "epoch": 3.27, "learning_rate": 6.733474476593743e-05, "loss": 2.9171, "step": 41658 }, { "epoch": 3.27, "learning_rate": 6.733396063671293e-05, "loss": 2.8832, "step": 41659 }, { "epoch": 3.27, "learning_rate": 6.733317650748844e-05, "loss": 2.6, "step": 41660 }, { "epoch": 3.27, "learning_rate": 6.733239237826394e-05, "loss": 2.7172, "step": 41661 }, { "epoch": 3.27, "learning_rate": 6.733160824903944e-05, "loss": 2.8267, "step": 41662 }, { "epoch": 3.27, "learning_rate": 6.733082411981495e-05, "loss": 2.8178, "step": 41663 }, { "epoch": 3.27, "learning_rate": 6.733003999059045e-05, "loss": 2.673, "step": 41664 }, { "epoch": 3.27, "learning_rate": 6.732925586136595e-05, "loss": 2.7031, "step": 41665 }, { "epoch": 3.27, "learning_rate": 6.732847173214146e-05, "loss": 2.9245, "step": 41666 }, { "epoch": 3.27, "learning_rate": 6.732768760291696e-05, "loss": 2.837, "step": 41667 }, { "epoch": 3.27, "learning_rate": 6.732690347369246e-05, "loss": 2.6461, "step": 41668 }, { "epoch": 3.27, "learning_rate": 6.732611934446797e-05, "loss": 2.7715, "step": 41669 }, { "epoch": 3.27, "learning_rate": 6.732533521524349e-05, "loss": 2.6677, "step": 41670 }, { "epoch": 3.27, "learning_rate": 6.732455108601899e-05, "loss": 2.6764, "step": 41671 }, { "epoch": 3.27, "learning_rate": 6.732376695679449e-05, "loss": 2.5589, "step": 41672 }, { "epoch": 3.27, "learning_rate": 6.732298282757e-05, "loss": 2.9013, "step": 41673 }, { "epoch": 3.27, "learning_rate": 6.73221986983455e-05, "loss": 2.9669, "step": 41674 }, { "epoch": 3.27, "learning_rate": 6.7321414569121e-05, "loss": 2.9948, "step": 41675 }, { "epoch": 3.27, "learning_rate": 6.73206304398965e-05, "loss": 2.7941, "step": 41676 }, { "epoch": 3.27, "learning_rate": 6.7319846310672e-05, "loss": 2.5983, "step": 41677 }, { "epoch": 3.27, "learning_rate": 6.73190621814475e-05, "loss": 2.9563, "step": 41678 }, { "epoch": 3.27, "learning_rate": 6.7318278052223e-05, "loss": 2.5486, "step": 41679 }, { "epoch": 3.27, "learning_rate": 6.731749392299851e-05, "loss": 2.6819, "step": 41680 }, { "epoch": 3.27, "learning_rate": 6.731670979377401e-05, "loss": 2.8612, "step": 41681 }, { "epoch": 3.27, "learning_rate": 6.731592566454952e-05, "loss": 2.6998, "step": 41682 }, { "epoch": 3.27, "learning_rate": 6.731514153532502e-05, "loss": 2.6064, "step": 41683 }, { "epoch": 3.27, "learning_rate": 6.731435740610052e-05, "loss": 2.6461, "step": 41684 }, { "epoch": 3.27, "learning_rate": 6.731357327687603e-05, "loss": 2.7771, "step": 41685 }, { "epoch": 3.27, "learning_rate": 6.731278914765153e-05, "loss": 2.7771, "step": 41686 }, { "epoch": 3.27, "learning_rate": 6.731200501842703e-05, "loss": 2.7166, "step": 41687 }, { "epoch": 3.27, "learning_rate": 6.731122088920254e-05, "loss": 2.6679, "step": 41688 }, { "epoch": 3.27, "learning_rate": 6.731043675997804e-05, "loss": 2.7568, "step": 41689 }, { "epoch": 3.27, "learning_rate": 6.730965263075355e-05, "loss": 2.8422, "step": 41690 }, { "epoch": 3.27, "learning_rate": 6.730886850152905e-05, "loss": 2.7858, "step": 41691 }, { "epoch": 3.27, "learning_rate": 6.730808437230455e-05, "loss": 2.7051, "step": 41692 }, { "epoch": 3.27, "learning_rate": 6.730730024308006e-05, "loss": 2.7645, "step": 41693 }, { "epoch": 3.27, "learning_rate": 6.730651611385556e-05, "loss": 2.9262, "step": 41694 }, { "epoch": 3.27, "learning_rate": 6.730573198463106e-05, "loss": 2.7726, "step": 41695 }, { "epoch": 3.27, "learning_rate": 6.730494785540658e-05, "loss": 2.6297, "step": 41696 }, { "epoch": 3.27, "learning_rate": 6.730416372618209e-05, "loss": 2.682, "step": 41697 }, { "epoch": 3.27, "learning_rate": 6.730337959695759e-05, "loss": 2.8149, "step": 41698 }, { "epoch": 3.27, "learning_rate": 6.730259546773309e-05, "loss": 2.5335, "step": 41699 }, { "epoch": 3.27, "learning_rate": 6.73018113385086e-05, "loss": 2.5487, "step": 41700 }, { "epoch": 3.27, "learning_rate": 6.73010272092841e-05, "loss": 2.7972, "step": 41701 }, { "epoch": 3.27, "learning_rate": 6.73002430800596e-05, "loss": 2.5978, "step": 41702 }, { "epoch": 3.27, "learning_rate": 6.729945895083511e-05, "loss": 2.8149, "step": 41703 }, { "epoch": 3.27, "learning_rate": 6.729867482161061e-05, "loss": 2.6064, "step": 41704 }, { "epoch": 3.27, "learning_rate": 6.729789069238611e-05, "loss": 2.6577, "step": 41705 }, { "epoch": 3.27, "learning_rate": 6.729710656316162e-05, "loss": 2.5739, "step": 41706 }, { "epoch": 3.27, "learning_rate": 6.729632243393712e-05, "loss": 2.6695, "step": 41707 }, { "epoch": 3.27, "learning_rate": 6.729553830471263e-05, "loss": 2.7582, "step": 41708 }, { "epoch": 3.27, "learning_rate": 6.729475417548813e-05, "loss": 2.7183, "step": 41709 }, { "epoch": 3.27, "learning_rate": 6.729397004626363e-05, "loss": 2.6215, "step": 41710 }, { "epoch": 3.27, "learning_rate": 6.729318591703914e-05, "loss": 2.5413, "step": 41711 }, { "epoch": 3.27, "learning_rate": 6.729240178781463e-05, "loss": 2.926, "step": 41712 }, { "epoch": 3.27, "learning_rate": 6.729161765859013e-05, "loss": 2.8007, "step": 41713 }, { "epoch": 3.27, "learning_rate": 6.729083352936563e-05, "loss": 2.5261, "step": 41714 }, { "epoch": 3.27, "learning_rate": 6.729004940014114e-05, "loss": 2.7085, "step": 41715 }, { "epoch": 3.27, "learning_rate": 6.728926527091664e-05, "loss": 2.6376, "step": 41716 }, { "epoch": 3.27, "learning_rate": 6.728848114169215e-05, "loss": 2.7572, "step": 41717 }, { "epoch": 3.27, "learning_rate": 6.728769701246765e-05, "loss": 2.9999, "step": 41718 }, { "epoch": 3.27, "learning_rate": 6.728691288324315e-05, "loss": 2.7594, "step": 41719 }, { "epoch": 3.27, "learning_rate": 6.728612875401866e-05, "loss": 2.8105, "step": 41720 }, { "epoch": 3.27, "learning_rate": 6.728534462479417e-05, "loss": 2.7568, "step": 41721 }, { "epoch": 3.27, "learning_rate": 6.728456049556968e-05, "loss": 2.6972, "step": 41722 }, { "epoch": 3.27, "learning_rate": 6.728377636634518e-05, "loss": 2.6682, "step": 41723 }, { "epoch": 3.27, "learning_rate": 6.728299223712068e-05, "loss": 2.617, "step": 41724 }, { "epoch": 3.27, "learning_rate": 6.728220810789619e-05, "loss": 2.9263, "step": 41725 }, { "epoch": 3.27, "learning_rate": 6.728142397867169e-05, "loss": 2.6351, "step": 41726 }, { "epoch": 3.27, "learning_rate": 6.72806398494472e-05, "loss": 2.6858, "step": 41727 }, { "epoch": 3.27, "learning_rate": 6.72798557202227e-05, "loss": 2.6137, "step": 41728 }, { "epoch": 3.27, "learning_rate": 6.72790715909982e-05, "loss": 2.6141, "step": 41729 }, { "epoch": 3.27, "learning_rate": 6.72782874617737e-05, "loss": 2.8276, "step": 41730 }, { "epoch": 3.27, "learning_rate": 6.727750333254921e-05, "loss": 2.6914, "step": 41731 }, { "epoch": 3.27, "learning_rate": 6.727671920332471e-05, "loss": 2.6564, "step": 41732 }, { "epoch": 3.27, "learning_rate": 6.727593507410022e-05, "loss": 2.6816, "step": 41733 }, { "epoch": 3.27, "learning_rate": 6.727515094487572e-05, "loss": 2.6625, "step": 41734 }, { "epoch": 3.27, "learning_rate": 6.727436681565122e-05, "loss": 2.5556, "step": 41735 }, { "epoch": 3.27, "learning_rate": 6.727358268642673e-05, "loss": 2.6592, "step": 41736 }, { "epoch": 3.27, "learning_rate": 6.727279855720223e-05, "loss": 2.7239, "step": 41737 }, { "epoch": 3.27, "learning_rate": 6.727201442797774e-05, "loss": 2.5695, "step": 41738 }, { "epoch": 3.27, "learning_rate": 6.727123029875324e-05, "loss": 2.6765, "step": 41739 }, { "epoch": 3.27, "learning_rate": 6.727044616952874e-05, "loss": 2.7098, "step": 41740 }, { "epoch": 3.27, "learning_rate": 6.726966204030425e-05, "loss": 2.8521, "step": 41741 }, { "epoch": 3.27, "learning_rate": 6.726887791107975e-05, "loss": 2.9269, "step": 41742 }, { "epoch": 3.27, "learning_rate": 6.726809378185525e-05, "loss": 2.6406, "step": 41743 }, { "epoch": 3.27, "learning_rate": 6.726730965263076e-05, "loss": 2.7541, "step": 41744 }, { "epoch": 3.27, "learning_rate": 6.726652552340626e-05, "loss": 2.7131, "step": 41745 }, { "epoch": 3.27, "learning_rate": 6.726574139418177e-05, "loss": 2.6591, "step": 41746 }, { "epoch": 3.27, "learning_rate": 6.726495726495727e-05, "loss": 2.7496, "step": 41747 }, { "epoch": 3.27, "learning_rate": 6.726417313573277e-05, "loss": 2.7689, "step": 41748 }, { "epoch": 3.27, "learning_rate": 6.726338900650828e-05, "loss": 2.6981, "step": 41749 }, { "epoch": 3.27, "learning_rate": 6.726260487728378e-05, "loss": 2.756, "step": 41750 }, { "epoch": 3.27, "learning_rate": 6.726182074805928e-05, "loss": 2.7609, "step": 41751 }, { "epoch": 3.27, "learning_rate": 6.726103661883479e-05, "loss": 2.682, "step": 41752 }, { "epoch": 3.27, "learning_rate": 6.726025248961029e-05, "loss": 2.8218, "step": 41753 }, { "epoch": 3.27, "learning_rate": 6.72594683603858e-05, "loss": 2.6762, "step": 41754 }, { "epoch": 3.27, "learning_rate": 6.72586842311613e-05, "loss": 2.7866, "step": 41755 }, { "epoch": 3.27, "learning_rate": 6.72579001019368e-05, "loss": 2.6777, "step": 41756 }, { "epoch": 3.27, "learning_rate": 6.72571159727123e-05, "loss": 2.9015, "step": 41757 }, { "epoch": 3.27, "learning_rate": 6.725633184348781e-05, "loss": 2.6354, "step": 41758 }, { "epoch": 3.27, "learning_rate": 6.725554771426331e-05, "loss": 2.8089, "step": 41759 }, { "epoch": 3.27, "learning_rate": 6.725476358503882e-05, "loss": 2.6416, "step": 41760 }, { "epoch": 3.27, "learning_rate": 6.725397945581432e-05, "loss": 2.7501, "step": 41761 }, { "epoch": 3.27, "learning_rate": 6.725319532658982e-05, "loss": 2.5826, "step": 41762 }, { "epoch": 3.27, "learning_rate": 6.725241119736533e-05, "loss": 2.7338, "step": 41763 }, { "epoch": 3.27, "learning_rate": 6.725162706814083e-05, "loss": 2.6642, "step": 41764 }, { "epoch": 3.27, "learning_rate": 6.725084293891634e-05, "loss": 2.6274, "step": 41765 }, { "epoch": 3.27, "learning_rate": 6.725005880969184e-05, "loss": 2.7357, "step": 41766 }, { "epoch": 3.28, "learning_rate": 6.724927468046734e-05, "loss": 2.5808, "step": 41767 }, { "epoch": 3.28, "learning_rate": 6.724849055124285e-05, "loss": 2.8213, "step": 41768 }, { "epoch": 3.28, "learning_rate": 6.724770642201835e-05, "loss": 2.8748, "step": 41769 }, { "epoch": 3.28, "learning_rate": 6.724692229279385e-05, "loss": 2.8724, "step": 41770 }, { "epoch": 3.28, "learning_rate": 6.724613816356936e-05, "loss": 2.6141, "step": 41771 }, { "epoch": 3.28, "learning_rate": 6.724535403434486e-05, "loss": 2.7202, "step": 41772 }, { "epoch": 3.28, "learning_rate": 6.724456990512036e-05, "loss": 2.7365, "step": 41773 }, { "epoch": 3.28, "learning_rate": 6.724378577589587e-05, "loss": 2.5511, "step": 41774 }, { "epoch": 3.28, "learning_rate": 6.724300164667137e-05, "loss": 2.7251, "step": 41775 }, { "epoch": 3.28, "learning_rate": 6.724221751744688e-05, "loss": 2.6851, "step": 41776 }, { "epoch": 3.28, "learning_rate": 6.724143338822238e-05, "loss": 2.9318, "step": 41777 }, { "epoch": 3.28, "learning_rate": 6.72406492589979e-05, "loss": 2.7068, "step": 41778 }, { "epoch": 3.28, "learning_rate": 6.72398651297734e-05, "loss": 2.7654, "step": 41779 }, { "epoch": 3.28, "learning_rate": 6.72390810005489e-05, "loss": 2.6562, "step": 41780 }, { "epoch": 3.28, "learning_rate": 6.723829687132441e-05, "loss": 2.7052, "step": 41781 }, { "epoch": 3.28, "learning_rate": 6.723751274209991e-05, "loss": 2.7957, "step": 41782 }, { "epoch": 3.28, "learning_rate": 6.723672861287541e-05, "loss": 2.8361, "step": 41783 }, { "epoch": 3.28, "learning_rate": 6.72359444836509e-05, "loss": 2.5344, "step": 41784 }, { "epoch": 3.28, "learning_rate": 6.723516035442641e-05, "loss": 2.6608, "step": 41785 }, { "epoch": 3.28, "learning_rate": 6.723437622520191e-05, "loss": 2.5543, "step": 41786 }, { "epoch": 3.28, "learning_rate": 6.723359209597742e-05, "loss": 2.9308, "step": 41787 }, { "epoch": 3.28, "learning_rate": 6.723280796675292e-05, "loss": 2.7503, "step": 41788 }, { "epoch": 3.28, "learning_rate": 6.723202383752842e-05, "loss": 2.9835, "step": 41789 }, { "epoch": 3.28, "learning_rate": 6.723123970830393e-05, "loss": 2.7326, "step": 41790 }, { "epoch": 3.28, "learning_rate": 6.723045557907943e-05, "loss": 2.5626, "step": 41791 }, { "epoch": 3.28, "learning_rate": 6.722967144985493e-05, "loss": 2.6826, "step": 41792 }, { "epoch": 3.28, "learning_rate": 6.722888732063044e-05, "loss": 2.7047, "step": 41793 }, { "epoch": 3.28, "learning_rate": 6.722810319140594e-05, "loss": 2.7264, "step": 41794 }, { "epoch": 3.28, "learning_rate": 6.722731906218145e-05, "loss": 2.7864, "step": 41795 }, { "epoch": 3.28, "learning_rate": 6.722653493295695e-05, "loss": 2.8024, "step": 41796 }, { "epoch": 3.28, "learning_rate": 6.722575080373245e-05, "loss": 2.5104, "step": 41797 }, { "epoch": 3.28, "learning_rate": 6.722496667450796e-05, "loss": 2.717, "step": 41798 }, { "epoch": 3.28, "learning_rate": 6.722418254528346e-05, "loss": 2.6799, "step": 41799 }, { "epoch": 3.28, "learning_rate": 6.722339841605896e-05, "loss": 3.0583, "step": 41800 }, { "epoch": 3.28, "learning_rate": 6.722261428683447e-05, "loss": 2.76, "step": 41801 }, { "epoch": 3.28, "learning_rate": 6.722183015760997e-05, "loss": 2.8192, "step": 41802 }, { "epoch": 3.28, "learning_rate": 6.722104602838547e-05, "loss": 2.7266, "step": 41803 }, { "epoch": 3.28, "learning_rate": 6.722026189916099e-05, "loss": 2.7925, "step": 41804 }, { "epoch": 3.28, "learning_rate": 6.72194777699365e-05, "loss": 2.6137, "step": 41805 }, { "epoch": 3.28, "learning_rate": 6.7218693640712e-05, "loss": 2.7045, "step": 41806 }, { "epoch": 3.28, "learning_rate": 6.72179095114875e-05, "loss": 2.7107, "step": 41807 }, { "epoch": 3.28, "learning_rate": 6.721712538226301e-05, "loss": 2.6382, "step": 41808 }, { "epoch": 3.28, "learning_rate": 6.721634125303851e-05, "loss": 2.7081, "step": 41809 }, { "epoch": 3.28, "learning_rate": 6.721555712381401e-05, "loss": 2.7138, "step": 41810 }, { "epoch": 3.28, "learning_rate": 6.721477299458952e-05, "loss": 2.6676, "step": 41811 }, { "epoch": 3.28, "learning_rate": 6.721398886536502e-05, "loss": 2.9008, "step": 41812 }, { "epoch": 3.28, "learning_rate": 6.721320473614053e-05, "loss": 2.7154, "step": 41813 }, { "epoch": 3.28, "learning_rate": 6.721242060691603e-05, "loss": 2.7747, "step": 41814 }, { "epoch": 3.28, "learning_rate": 6.721163647769153e-05, "loss": 2.7244, "step": 41815 }, { "epoch": 3.28, "learning_rate": 6.721085234846704e-05, "loss": 2.6493, "step": 41816 }, { "epoch": 3.28, "learning_rate": 6.721006821924254e-05, "loss": 2.704, "step": 41817 }, { "epoch": 3.28, "learning_rate": 6.720928409001804e-05, "loss": 2.6768, "step": 41818 }, { "epoch": 3.28, "learning_rate": 6.720849996079355e-05, "loss": 2.7169, "step": 41819 }, { "epoch": 3.28, "learning_rate": 6.720771583156904e-05, "loss": 2.8393, "step": 41820 }, { "epoch": 3.28, "learning_rate": 6.720693170234454e-05, "loss": 2.6066, "step": 41821 }, { "epoch": 3.28, "learning_rate": 6.720614757312004e-05, "loss": 2.7214, "step": 41822 }, { "epoch": 3.28, "learning_rate": 6.720536344389555e-05, "loss": 2.6325, "step": 41823 }, { "epoch": 3.28, "learning_rate": 6.720457931467105e-05, "loss": 2.6433, "step": 41824 }, { "epoch": 3.28, "learning_rate": 6.720379518544656e-05, "loss": 2.6895, "step": 41825 }, { "epoch": 3.28, "learning_rate": 6.720301105622206e-05, "loss": 2.7719, "step": 41826 }, { "epoch": 3.28, "learning_rate": 6.720222692699756e-05, "loss": 2.8656, "step": 41827 }, { "epoch": 3.28, "learning_rate": 6.720144279777307e-05, "loss": 2.8918, "step": 41828 }, { "epoch": 3.28, "learning_rate": 6.720065866854858e-05, "loss": 2.8093, "step": 41829 }, { "epoch": 3.28, "learning_rate": 6.719987453932409e-05, "loss": 2.5485, "step": 41830 }, { "epoch": 3.28, "learning_rate": 6.719909041009959e-05, "loss": 2.8124, "step": 41831 }, { "epoch": 3.28, "learning_rate": 6.71983062808751e-05, "loss": 2.7529, "step": 41832 }, { "epoch": 3.28, "learning_rate": 6.71975221516506e-05, "loss": 2.5562, "step": 41833 }, { "epoch": 3.28, "learning_rate": 6.71967380224261e-05, "loss": 2.5958, "step": 41834 }, { "epoch": 3.28, "learning_rate": 6.71959538932016e-05, "loss": 2.9128, "step": 41835 }, { "epoch": 3.28, "learning_rate": 6.719516976397711e-05, "loss": 2.7655, "step": 41836 }, { "epoch": 3.28, "learning_rate": 6.719438563475261e-05, "loss": 2.6917, "step": 41837 }, { "epoch": 3.28, "learning_rate": 6.719360150552812e-05, "loss": 2.8953, "step": 41838 }, { "epoch": 3.28, "learning_rate": 6.719281737630362e-05, "loss": 2.5558, "step": 41839 }, { "epoch": 3.28, "learning_rate": 6.719203324707912e-05, "loss": 2.8288, "step": 41840 }, { "epoch": 3.28, "learning_rate": 6.719124911785463e-05, "loss": 2.7422, "step": 41841 }, { "epoch": 3.28, "learning_rate": 6.719046498863013e-05, "loss": 2.7571, "step": 41842 }, { "epoch": 3.28, "learning_rate": 6.718968085940564e-05, "loss": 2.6598, "step": 41843 }, { "epoch": 3.28, "learning_rate": 6.718889673018114e-05, "loss": 2.6376, "step": 41844 }, { "epoch": 3.28, "learning_rate": 6.718811260095664e-05, "loss": 2.846, "step": 41845 }, { "epoch": 3.28, "learning_rate": 6.718732847173215e-05, "loss": 2.7466, "step": 41846 }, { "epoch": 3.28, "learning_rate": 6.718654434250765e-05, "loss": 2.6823, "step": 41847 }, { "epoch": 3.28, "learning_rate": 6.718576021328315e-05, "loss": 3.1145, "step": 41848 }, { "epoch": 3.28, "learning_rate": 6.718497608405866e-05, "loss": 2.6272, "step": 41849 }, { "epoch": 3.28, "learning_rate": 6.718419195483416e-05, "loss": 2.6429, "step": 41850 }, { "epoch": 3.28, "learning_rate": 6.718340782560966e-05, "loss": 2.7159, "step": 41851 }, { "epoch": 3.28, "learning_rate": 6.718262369638517e-05, "loss": 2.5672, "step": 41852 }, { "epoch": 3.28, "learning_rate": 6.718183956716067e-05, "loss": 2.705, "step": 41853 }, { "epoch": 3.28, "learning_rate": 6.718105543793618e-05, "loss": 2.6007, "step": 41854 }, { "epoch": 3.28, "learning_rate": 6.718027130871168e-05, "loss": 2.7647, "step": 41855 }, { "epoch": 3.28, "learning_rate": 6.717948717948718e-05, "loss": 2.8071, "step": 41856 }, { "epoch": 3.28, "learning_rate": 6.717870305026269e-05, "loss": 2.6066, "step": 41857 }, { "epoch": 3.28, "learning_rate": 6.717791892103819e-05, "loss": 2.8141, "step": 41858 }, { "epoch": 3.28, "learning_rate": 6.71771347918137e-05, "loss": 2.6931, "step": 41859 }, { "epoch": 3.28, "learning_rate": 6.71763506625892e-05, "loss": 2.7877, "step": 41860 }, { "epoch": 3.28, "learning_rate": 6.71755665333647e-05, "loss": 2.7348, "step": 41861 }, { "epoch": 3.28, "learning_rate": 6.71747824041402e-05, "loss": 2.8029, "step": 41862 }, { "epoch": 3.28, "learning_rate": 6.717399827491571e-05, "loss": 2.7422, "step": 41863 }, { "epoch": 3.28, "learning_rate": 6.717321414569121e-05, "loss": 2.6608, "step": 41864 }, { "epoch": 3.28, "learning_rate": 6.717243001646672e-05, "loss": 2.7617, "step": 41865 }, { "epoch": 3.28, "learning_rate": 6.717164588724222e-05, "loss": 2.8181, "step": 41866 }, { "epoch": 3.28, "learning_rate": 6.717086175801772e-05, "loss": 2.7535, "step": 41867 }, { "epoch": 3.28, "learning_rate": 6.717007762879323e-05, "loss": 2.6727, "step": 41868 }, { "epoch": 3.28, "learning_rate": 6.716929349956873e-05, "loss": 2.6966, "step": 41869 }, { "epoch": 3.28, "learning_rate": 6.716850937034423e-05, "loss": 2.61, "step": 41870 }, { "epoch": 3.28, "learning_rate": 6.716772524111974e-05, "loss": 2.7244, "step": 41871 }, { "epoch": 3.28, "learning_rate": 6.716694111189524e-05, "loss": 2.5099, "step": 41872 }, { "epoch": 3.28, "learning_rate": 6.716615698267075e-05, "loss": 2.7926, "step": 41873 }, { "epoch": 3.28, "learning_rate": 6.716537285344625e-05, "loss": 2.6612, "step": 41874 }, { "epoch": 3.28, "learning_rate": 6.716458872422175e-05, "loss": 2.6521, "step": 41875 }, { "epoch": 3.28, "learning_rate": 6.716380459499726e-05, "loss": 2.7168, "step": 41876 }, { "epoch": 3.28, "learning_rate": 6.716302046577276e-05, "loss": 2.7659, "step": 41877 }, { "epoch": 3.28, "learning_rate": 6.716223633654826e-05, "loss": 2.4201, "step": 41878 }, { "epoch": 3.28, "learning_rate": 6.716145220732377e-05, "loss": 3.0697, "step": 41879 }, { "epoch": 3.28, "learning_rate": 6.716066807809927e-05, "loss": 2.9979, "step": 41880 }, { "epoch": 3.28, "learning_rate": 6.715988394887478e-05, "loss": 2.7155, "step": 41881 }, { "epoch": 3.28, "learning_rate": 6.715909981965028e-05, "loss": 2.7257, "step": 41882 }, { "epoch": 3.28, "learning_rate": 6.715831569042578e-05, "loss": 3.0278, "step": 41883 }, { "epoch": 3.28, "learning_rate": 6.715753156120129e-05, "loss": 2.7342, "step": 41884 }, { "epoch": 3.28, "learning_rate": 6.715674743197679e-05, "loss": 2.7495, "step": 41885 }, { "epoch": 3.28, "learning_rate": 6.715596330275231e-05, "loss": 3.0177, "step": 41886 }, { "epoch": 3.28, "learning_rate": 6.715517917352781e-05, "loss": 2.8993, "step": 41887 }, { "epoch": 3.28, "learning_rate": 6.715439504430331e-05, "loss": 2.7891, "step": 41888 }, { "epoch": 3.28, "learning_rate": 6.715361091507882e-05, "loss": 2.7428, "step": 41889 }, { "epoch": 3.28, "learning_rate": 6.715282678585432e-05, "loss": 2.6671, "step": 41890 }, { "epoch": 3.28, "learning_rate": 6.715204265662983e-05, "loss": 2.6197, "step": 41891 }, { "epoch": 3.28, "learning_rate": 6.715125852740532e-05, "loss": 2.6169, "step": 41892 }, { "epoch": 3.28, "learning_rate": 6.715047439818082e-05, "loss": 2.7615, "step": 41893 }, { "epoch": 3.29, "learning_rate": 6.714969026895632e-05, "loss": 2.5796, "step": 41894 }, { "epoch": 3.29, "learning_rate": 6.714890613973183e-05, "loss": 2.8224, "step": 41895 }, { "epoch": 3.29, "learning_rate": 6.714812201050733e-05, "loss": 2.5619, "step": 41896 }, { "epoch": 3.29, "learning_rate": 6.714733788128283e-05, "loss": 2.9138, "step": 41897 }, { "epoch": 3.29, "learning_rate": 6.714655375205834e-05, "loss": 2.7728, "step": 41898 }, { "epoch": 3.29, "learning_rate": 6.714576962283384e-05, "loss": 2.8606, "step": 41899 }, { "epoch": 3.29, "learning_rate": 6.714498549360935e-05, "loss": 2.5334, "step": 41900 }, { "epoch": 3.29, "learning_rate": 6.714420136438485e-05, "loss": 2.6623, "step": 41901 }, { "epoch": 3.29, "learning_rate": 6.714341723516035e-05, "loss": 2.6109, "step": 41902 }, { "epoch": 3.29, "learning_rate": 6.714263310593586e-05, "loss": 2.739, "step": 41903 }, { "epoch": 3.29, "learning_rate": 6.714184897671136e-05, "loss": 2.5792, "step": 41904 }, { "epoch": 3.29, "learning_rate": 6.714106484748686e-05, "loss": 2.7435, "step": 41905 }, { "epoch": 3.29, "learning_rate": 6.714028071826237e-05, "loss": 2.9397, "step": 41906 }, { "epoch": 3.29, "learning_rate": 6.713949658903787e-05, "loss": 2.9121, "step": 41907 }, { "epoch": 3.29, "learning_rate": 6.713871245981337e-05, "loss": 2.9796, "step": 41908 }, { "epoch": 3.29, "learning_rate": 6.713792833058888e-05, "loss": 2.6546, "step": 41909 }, { "epoch": 3.29, "learning_rate": 6.713714420136438e-05, "loss": 2.6661, "step": 41910 }, { "epoch": 3.29, "learning_rate": 6.713636007213989e-05, "loss": 2.6385, "step": 41911 }, { "epoch": 3.29, "learning_rate": 6.71355759429154e-05, "loss": 2.7183, "step": 41912 }, { "epoch": 3.29, "learning_rate": 6.71347918136909e-05, "loss": 2.8867, "step": 41913 }, { "epoch": 3.29, "learning_rate": 6.713400768446641e-05, "loss": 2.8241, "step": 41914 }, { "epoch": 3.29, "learning_rate": 6.713322355524191e-05, "loss": 2.6162, "step": 41915 }, { "epoch": 3.29, "learning_rate": 6.713243942601742e-05, "loss": 2.6598, "step": 41916 }, { "epoch": 3.29, "learning_rate": 6.713165529679292e-05, "loss": 2.803, "step": 41917 }, { "epoch": 3.29, "learning_rate": 6.713087116756842e-05, "loss": 2.8711, "step": 41918 }, { "epoch": 3.29, "learning_rate": 6.713008703834393e-05, "loss": 2.8197, "step": 41919 }, { "epoch": 3.29, "learning_rate": 6.712930290911943e-05, "loss": 2.8826, "step": 41920 }, { "epoch": 3.29, "learning_rate": 6.712851877989494e-05, "loss": 2.5613, "step": 41921 }, { "epoch": 3.29, "learning_rate": 6.712773465067044e-05, "loss": 2.8328, "step": 41922 }, { "epoch": 3.29, "learning_rate": 6.712695052144594e-05, "loss": 2.639, "step": 41923 }, { "epoch": 3.29, "learning_rate": 6.712616639222145e-05, "loss": 2.8857, "step": 41924 }, { "epoch": 3.29, "learning_rate": 6.712538226299695e-05, "loss": 2.4398, "step": 41925 }, { "epoch": 3.29, "learning_rate": 6.712459813377245e-05, "loss": 2.9907, "step": 41926 }, { "epoch": 3.29, "learning_rate": 6.712381400454796e-05, "loss": 2.857, "step": 41927 }, { "epoch": 3.29, "learning_rate": 6.712302987532345e-05, "loss": 2.5739, "step": 41928 }, { "epoch": 3.29, "learning_rate": 6.712224574609895e-05, "loss": 2.6436, "step": 41929 }, { "epoch": 3.29, "learning_rate": 6.712146161687446e-05, "loss": 2.7549, "step": 41930 }, { "epoch": 3.29, "learning_rate": 6.712067748764996e-05, "loss": 2.6506, "step": 41931 }, { "epoch": 3.29, "learning_rate": 6.711989335842546e-05, "loss": 2.6341, "step": 41932 }, { "epoch": 3.29, "learning_rate": 6.711910922920097e-05, "loss": 2.6555, "step": 41933 }, { "epoch": 3.29, "learning_rate": 6.711832509997647e-05, "loss": 2.6876, "step": 41934 }, { "epoch": 3.29, "learning_rate": 6.711754097075197e-05, "loss": 2.7042, "step": 41935 }, { "epoch": 3.29, "learning_rate": 6.711675684152748e-05, "loss": 2.6713, "step": 41936 }, { "epoch": 3.29, "learning_rate": 6.711597271230298e-05, "loss": 2.6428, "step": 41937 }, { "epoch": 3.29, "learning_rate": 6.71151885830785e-05, "loss": 2.7414, "step": 41938 }, { "epoch": 3.29, "learning_rate": 6.7114404453854e-05, "loss": 2.8497, "step": 41939 }, { "epoch": 3.29, "learning_rate": 6.71136203246295e-05, "loss": 2.8321, "step": 41940 }, { "epoch": 3.29, "learning_rate": 6.711283619540501e-05, "loss": 2.711, "step": 41941 }, { "epoch": 3.29, "learning_rate": 6.711205206618051e-05, "loss": 2.8416, "step": 41942 }, { "epoch": 3.29, "learning_rate": 6.711126793695602e-05, "loss": 2.718, "step": 41943 }, { "epoch": 3.29, "learning_rate": 6.711048380773152e-05, "loss": 2.8922, "step": 41944 }, { "epoch": 3.29, "learning_rate": 6.710969967850702e-05, "loss": 2.7358, "step": 41945 }, { "epoch": 3.29, "learning_rate": 6.710891554928253e-05, "loss": 2.7112, "step": 41946 }, { "epoch": 3.29, "learning_rate": 6.710813142005803e-05, "loss": 2.6915, "step": 41947 }, { "epoch": 3.29, "learning_rate": 6.710734729083354e-05, "loss": 2.512, "step": 41948 }, { "epoch": 3.29, "learning_rate": 6.710656316160904e-05, "loss": 2.622, "step": 41949 }, { "epoch": 3.29, "learning_rate": 6.710577903238454e-05, "loss": 2.8721, "step": 41950 }, { "epoch": 3.29, "learning_rate": 6.710499490316005e-05, "loss": 2.5781, "step": 41951 }, { "epoch": 3.29, "learning_rate": 6.710421077393555e-05, "loss": 2.6581, "step": 41952 }, { "epoch": 3.29, "learning_rate": 6.710342664471105e-05, "loss": 2.5096, "step": 41953 }, { "epoch": 3.29, "learning_rate": 6.710264251548656e-05, "loss": 2.6358, "step": 41954 }, { "epoch": 3.29, "learning_rate": 6.710185838626206e-05, "loss": 2.833, "step": 41955 }, { "epoch": 3.29, "learning_rate": 6.710107425703756e-05, "loss": 2.8603, "step": 41956 }, { "epoch": 3.29, "learning_rate": 6.710029012781307e-05, "loss": 2.6626, "step": 41957 }, { "epoch": 3.29, "learning_rate": 6.709950599858857e-05, "loss": 2.5371, "step": 41958 }, { "epoch": 3.29, "learning_rate": 6.709872186936408e-05, "loss": 2.6203, "step": 41959 }, { "epoch": 3.29, "learning_rate": 6.709793774013958e-05, "loss": 2.6672, "step": 41960 }, { "epoch": 3.29, "learning_rate": 6.709715361091508e-05, "loss": 2.5414, "step": 41961 }, { "epoch": 3.29, "learning_rate": 6.709636948169059e-05, "loss": 2.6399, "step": 41962 }, { "epoch": 3.29, "learning_rate": 6.709558535246609e-05, "loss": 2.6, "step": 41963 }, { "epoch": 3.29, "learning_rate": 6.70948012232416e-05, "loss": 2.5679, "step": 41964 }, { "epoch": 3.29, "learning_rate": 6.70940170940171e-05, "loss": 2.5398, "step": 41965 }, { "epoch": 3.29, "learning_rate": 6.70932329647926e-05, "loss": 2.693, "step": 41966 }, { "epoch": 3.29, "learning_rate": 6.70924488355681e-05, "loss": 2.5384, "step": 41967 }, { "epoch": 3.29, "learning_rate": 6.709166470634361e-05, "loss": 2.6323, "step": 41968 }, { "epoch": 3.29, "learning_rate": 6.709088057711911e-05, "loss": 2.7933, "step": 41969 }, { "epoch": 3.29, "learning_rate": 6.709009644789462e-05, "loss": 2.4651, "step": 41970 }, { "epoch": 3.29, "learning_rate": 6.708931231867012e-05, "loss": 2.616, "step": 41971 }, { "epoch": 3.29, "learning_rate": 6.708852818944562e-05, "loss": 2.8394, "step": 41972 }, { "epoch": 3.29, "learning_rate": 6.708774406022113e-05, "loss": 2.7047, "step": 41973 }, { "epoch": 3.29, "learning_rate": 6.708695993099663e-05, "loss": 2.8934, "step": 41974 }, { "epoch": 3.29, "learning_rate": 6.708617580177213e-05, "loss": 2.629, "step": 41975 }, { "epoch": 3.29, "learning_rate": 6.708539167254764e-05, "loss": 2.6381, "step": 41976 }, { "epoch": 3.29, "learning_rate": 6.708460754332314e-05, "loss": 2.6385, "step": 41977 }, { "epoch": 3.29, "learning_rate": 6.708382341409865e-05, "loss": 2.719, "step": 41978 }, { "epoch": 3.29, "learning_rate": 6.708303928487415e-05, "loss": 2.6163, "step": 41979 }, { "epoch": 3.29, "learning_rate": 6.708225515564965e-05, "loss": 2.6677, "step": 41980 }, { "epoch": 3.29, "learning_rate": 6.708147102642516e-05, "loss": 2.609, "step": 41981 }, { "epoch": 3.29, "learning_rate": 6.708068689720066e-05, "loss": 2.5923, "step": 41982 }, { "epoch": 3.29, "learning_rate": 6.707990276797616e-05, "loss": 2.724, "step": 41983 }, { "epoch": 3.29, "learning_rate": 6.707911863875167e-05, "loss": 2.5061, "step": 41984 }, { "epoch": 3.29, "learning_rate": 6.707833450952717e-05, "loss": 2.7407, "step": 41985 }, { "epoch": 3.29, "learning_rate": 6.707755038030267e-05, "loss": 2.5741, "step": 41986 }, { "epoch": 3.29, "learning_rate": 6.707676625107818e-05, "loss": 2.7085, "step": 41987 }, { "epoch": 3.29, "learning_rate": 6.707598212185368e-05, "loss": 2.8145, "step": 41988 }, { "epoch": 3.29, "learning_rate": 6.707519799262919e-05, "loss": 2.7825, "step": 41989 }, { "epoch": 3.29, "learning_rate": 6.707441386340469e-05, "loss": 2.7812, "step": 41990 }, { "epoch": 3.29, "learning_rate": 6.70736297341802e-05, "loss": 2.8067, "step": 41991 }, { "epoch": 3.29, "learning_rate": 6.70728456049557e-05, "loss": 2.6066, "step": 41992 }, { "epoch": 3.29, "learning_rate": 6.70720614757312e-05, "loss": 2.9485, "step": 41993 }, { "epoch": 3.29, "learning_rate": 6.707127734650672e-05, "loss": 2.7298, "step": 41994 }, { "epoch": 3.29, "learning_rate": 6.707049321728222e-05, "loss": 2.8654, "step": 41995 }, { "epoch": 3.29, "learning_rate": 6.706970908805773e-05, "loss": 2.7615, "step": 41996 }, { "epoch": 3.29, "learning_rate": 6.706892495883323e-05, "loss": 2.6931, "step": 41997 }, { "epoch": 3.29, "learning_rate": 6.706814082960873e-05, "loss": 2.6621, "step": 41998 }, { "epoch": 3.29, "learning_rate": 6.706735670038424e-05, "loss": 2.6517, "step": 41999 }, { "epoch": 3.29, "learning_rate": 6.706657257115973e-05, "loss": 2.8005, "step": 42000 }, { "epoch": 3.29, "eval_loss": 2.899847984313965, "eval_runtime": 217.8423, "eval_samples_per_second": 52.019, "eval_steps_per_second": 3.255, "step": 42000 }, { "epoch": 3.29, "learning_rate": 6.706578844193523e-05, "loss": 2.5358, "step": 42001 }, { "epoch": 3.29, "learning_rate": 6.706500431271073e-05, "loss": 2.6124, "step": 42002 }, { "epoch": 3.29, "learning_rate": 6.706422018348624e-05, "loss": 2.5603, "step": 42003 }, { "epoch": 3.29, "learning_rate": 6.706343605426174e-05, "loss": 2.7666, "step": 42004 }, { "epoch": 3.29, "learning_rate": 6.706265192503724e-05, "loss": 2.7888, "step": 42005 }, { "epoch": 3.29, "learning_rate": 6.706186779581275e-05, "loss": 2.5855, "step": 42006 }, { "epoch": 3.29, "learning_rate": 6.706108366658825e-05, "loss": 2.7492, "step": 42007 }, { "epoch": 3.29, "learning_rate": 6.706029953736376e-05, "loss": 2.6218, "step": 42008 }, { "epoch": 3.29, "learning_rate": 6.705951540813926e-05, "loss": 2.6686, "step": 42009 }, { "epoch": 3.29, "learning_rate": 6.705873127891476e-05, "loss": 2.6926, "step": 42010 }, { "epoch": 3.29, "learning_rate": 6.705794714969027e-05, "loss": 2.5715, "step": 42011 }, { "epoch": 3.29, "learning_rate": 6.705716302046577e-05, "loss": 2.6933, "step": 42012 }, { "epoch": 3.29, "learning_rate": 6.705637889124127e-05, "loss": 2.5239, "step": 42013 }, { "epoch": 3.29, "learning_rate": 6.705559476201678e-05, "loss": 2.6273, "step": 42014 }, { "epoch": 3.29, "learning_rate": 6.705481063279228e-05, "loss": 2.868, "step": 42015 }, { "epoch": 3.29, "learning_rate": 6.705402650356779e-05, "loss": 2.6271, "step": 42016 }, { "epoch": 3.29, "learning_rate": 6.705324237434329e-05, "loss": 2.7065, "step": 42017 }, { "epoch": 3.29, "learning_rate": 6.705245824511879e-05, "loss": 2.731, "step": 42018 }, { "epoch": 3.29, "learning_rate": 6.70516741158943e-05, "loss": 2.5831, "step": 42019 }, { "epoch": 3.29, "learning_rate": 6.705088998666981e-05, "loss": 2.6819, "step": 42020 }, { "epoch": 3.29, "learning_rate": 6.705010585744532e-05, "loss": 3.0277, "step": 42021 }, { "epoch": 3.3, "learning_rate": 6.704932172822082e-05, "loss": 2.6385, "step": 42022 }, { "epoch": 3.3, "learning_rate": 6.704853759899632e-05, "loss": 2.6191, "step": 42023 }, { "epoch": 3.3, "learning_rate": 6.704775346977183e-05, "loss": 2.8226, "step": 42024 }, { "epoch": 3.3, "learning_rate": 6.704696934054733e-05, "loss": 2.7274, "step": 42025 }, { "epoch": 3.3, "learning_rate": 6.704618521132284e-05, "loss": 2.6806, "step": 42026 }, { "epoch": 3.3, "learning_rate": 6.704540108209834e-05, "loss": 2.8015, "step": 42027 }, { "epoch": 3.3, "learning_rate": 6.704461695287384e-05, "loss": 3.0485, "step": 42028 }, { "epoch": 3.3, "learning_rate": 6.704383282364935e-05, "loss": 2.8362, "step": 42029 }, { "epoch": 3.3, "learning_rate": 6.704304869442485e-05, "loss": 2.8461, "step": 42030 }, { "epoch": 3.3, "learning_rate": 6.704226456520035e-05, "loss": 2.7797, "step": 42031 }, { "epoch": 3.3, "learning_rate": 6.704148043597586e-05, "loss": 2.9955, "step": 42032 }, { "epoch": 3.3, "learning_rate": 6.704069630675136e-05, "loss": 2.6804, "step": 42033 }, { "epoch": 3.3, "learning_rate": 6.703991217752687e-05, "loss": 2.617, "step": 42034 }, { "epoch": 3.3, "learning_rate": 6.703912804830237e-05, "loss": 2.5664, "step": 42035 }, { "epoch": 3.3, "learning_rate": 6.703834391907786e-05, "loss": 2.6053, "step": 42036 }, { "epoch": 3.3, "learning_rate": 6.703755978985336e-05, "loss": 2.7089, "step": 42037 }, { "epoch": 3.3, "learning_rate": 6.703677566062887e-05, "loss": 2.7679, "step": 42038 }, { "epoch": 3.3, "learning_rate": 6.703599153140437e-05, "loss": 2.5733, "step": 42039 }, { "epoch": 3.3, "learning_rate": 6.703520740217987e-05, "loss": 2.6608, "step": 42040 }, { "epoch": 3.3, "learning_rate": 6.703442327295538e-05, "loss": 2.6516, "step": 42041 }, { "epoch": 3.3, "learning_rate": 6.703363914373088e-05, "loss": 2.8614, "step": 42042 }, { "epoch": 3.3, "learning_rate": 6.703285501450638e-05, "loss": 2.7355, "step": 42043 }, { "epoch": 3.3, "learning_rate": 6.703207088528189e-05, "loss": 2.8023, "step": 42044 }, { "epoch": 3.3, "learning_rate": 6.703128675605739e-05, "loss": 2.7376, "step": 42045 }, { "epoch": 3.3, "learning_rate": 6.703050262683291e-05, "loss": 2.5803, "step": 42046 }, { "epoch": 3.3, "learning_rate": 6.702971849760841e-05, "loss": 2.4637, "step": 42047 }, { "epoch": 3.3, "learning_rate": 6.702893436838392e-05, "loss": 2.4732, "step": 42048 }, { "epoch": 3.3, "learning_rate": 6.702815023915942e-05, "loss": 2.9512, "step": 42049 }, { "epoch": 3.3, "learning_rate": 6.702736610993492e-05, "loss": 2.8329, "step": 42050 }, { "epoch": 3.3, "learning_rate": 6.702658198071043e-05, "loss": 2.7297, "step": 42051 }, { "epoch": 3.3, "learning_rate": 6.702579785148593e-05, "loss": 2.8842, "step": 42052 }, { "epoch": 3.3, "learning_rate": 6.702501372226143e-05, "loss": 2.7306, "step": 42053 }, { "epoch": 3.3, "learning_rate": 6.702422959303694e-05, "loss": 2.664, "step": 42054 }, { "epoch": 3.3, "learning_rate": 6.702344546381244e-05, "loss": 2.763, "step": 42055 }, { "epoch": 3.3, "learning_rate": 6.702266133458795e-05, "loss": 2.7809, "step": 42056 }, { "epoch": 3.3, "learning_rate": 6.702187720536345e-05, "loss": 2.7903, "step": 42057 }, { "epoch": 3.3, "learning_rate": 6.702109307613895e-05, "loss": 2.7658, "step": 42058 }, { "epoch": 3.3, "learning_rate": 6.702030894691446e-05, "loss": 2.8035, "step": 42059 }, { "epoch": 3.3, "learning_rate": 6.701952481768996e-05, "loss": 2.6693, "step": 42060 }, { "epoch": 3.3, "learning_rate": 6.701874068846546e-05, "loss": 2.655, "step": 42061 }, { "epoch": 3.3, "learning_rate": 6.701795655924097e-05, "loss": 2.627, "step": 42062 }, { "epoch": 3.3, "learning_rate": 6.701717243001647e-05, "loss": 2.6338, "step": 42063 }, { "epoch": 3.3, "learning_rate": 6.701638830079198e-05, "loss": 3.0062, "step": 42064 }, { "epoch": 3.3, "learning_rate": 6.701560417156748e-05, "loss": 2.73, "step": 42065 }, { "epoch": 3.3, "learning_rate": 6.701482004234298e-05, "loss": 2.6238, "step": 42066 }, { "epoch": 3.3, "learning_rate": 6.701403591311849e-05, "loss": 2.7324, "step": 42067 }, { "epoch": 3.3, "learning_rate": 6.701325178389399e-05, "loss": 2.7078, "step": 42068 }, { "epoch": 3.3, "learning_rate": 6.70124676546695e-05, "loss": 2.7259, "step": 42069 }, { "epoch": 3.3, "learning_rate": 6.7011683525445e-05, "loss": 2.5448, "step": 42070 }, { "epoch": 3.3, "learning_rate": 6.70108993962205e-05, "loss": 2.9974, "step": 42071 }, { "epoch": 3.3, "learning_rate": 6.7010115266996e-05, "loss": 2.8414, "step": 42072 }, { "epoch": 3.3, "learning_rate": 6.700933113777151e-05, "loss": 2.7055, "step": 42073 }, { "epoch": 3.3, "learning_rate": 6.700854700854701e-05, "loss": 2.6951, "step": 42074 }, { "epoch": 3.3, "learning_rate": 6.700776287932252e-05, "loss": 2.9256, "step": 42075 }, { "epoch": 3.3, "learning_rate": 6.700697875009802e-05, "loss": 2.8318, "step": 42076 }, { "epoch": 3.3, "learning_rate": 6.700619462087352e-05, "loss": 2.6875, "step": 42077 }, { "epoch": 3.3, "learning_rate": 6.700541049164903e-05, "loss": 2.8805, "step": 42078 }, { "epoch": 3.3, "learning_rate": 6.700462636242453e-05, "loss": 2.5166, "step": 42079 }, { "epoch": 3.3, "learning_rate": 6.700384223320003e-05, "loss": 2.7836, "step": 42080 }, { "epoch": 3.3, "learning_rate": 6.700305810397554e-05, "loss": 2.6666, "step": 42081 }, { "epoch": 3.3, "learning_rate": 6.700227397475104e-05, "loss": 2.6273, "step": 42082 }, { "epoch": 3.3, "learning_rate": 6.700148984552655e-05, "loss": 2.5017, "step": 42083 }, { "epoch": 3.3, "learning_rate": 6.700070571630205e-05, "loss": 2.9494, "step": 42084 }, { "epoch": 3.3, "learning_rate": 6.699992158707755e-05, "loss": 2.6667, "step": 42085 }, { "epoch": 3.3, "learning_rate": 6.699913745785306e-05, "loss": 2.8402, "step": 42086 }, { "epoch": 3.3, "learning_rate": 6.699835332862856e-05, "loss": 2.592, "step": 42087 }, { "epoch": 3.3, "learning_rate": 6.699756919940406e-05, "loss": 2.8121, "step": 42088 }, { "epoch": 3.3, "learning_rate": 6.699678507017957e-05, "loss": 2.8661, "step": 42089 }, { "epoch": 3.3, "learning_rate": 6.699600094095507e-05, "loss": 2.5755, "step": 42090 }, { "epoch": 3.3, "learning_rate": 6.699521681173057e-05, "loss": 2.7318, "step": 42091 }, { "epoch": 3.3, "learning_rate": 6.699443268250608e-05, "loss": 2.645, "step": 42092 }, { "epoch": 3.3, "learning_rate": 6.699364855328158e-05, "loss": 2.7586, "step": 42093 }, { "epoch": 3.3, "learning_rate": 6.699286442405709e-05, "loss": 2.8503, "step": 42094 }, { "epoch": 3.3, "learning_rate": 6.699208029483259e-05, "loss": 2.7235, "step": 42095 }, { "epoch": 3.3, "learning_rate": 6.699129616560809e-05, "loss": 2.7686, "step": 42096 }, { "epoch": 3.3, "learning_rate": 6.69905120363836e-05, "loss": 2.4611, "step": 42097 }, { "epoch": 3.3, "learning_rate": 6.69897279071591e-05, "loss": 2.6018, "step": 42098 }, { "epoch": 3.3, "learning_rate": 6.69889437779346e-05, "loss": 2.8194, "step": 42099 }, { "epoch": 3.3, "learning_rate": 6.698815964871011e-05, "loss": 2.8009, "step": 42100 }, { "epoch": 3.3, "learning_rate": 6.698737551948561e-05, "loss": 2.6236, "step": 42101 }, { "epoch": 3.3, "learning_rate": 6.698659139026113e-05, "loss": 2.6991, "step": 42102 }, { "epoch": 3.3, "learning_rate": 6.698580726103663e-05, "loss": 2.5166, "step": 42103 }, { "epoch": 3.3, "learning_rate": 6.698502313181214e-05, "loss": 2.6823, "step": 42104 }, { "epoch": 3.3, "learning_rate": 6.698423900258764e-05, "loss": 2.6429, "step": 42105 }, { "epoch": 3.3, "learning_rate": 6.698345487336314e-05, "loss": 2.9221, "step": 42106 }, { "epoch": 3.3, "learning_rate": 6.698267074413865e-05, "loss": 2.7211, "step": 42107 }, { "epoch": 3.3, "learning_rate": 6.698188661491414e-05, "loss": 2.98, "step": 42108 }, { "epoch": 3.3, "learning_rate": 6.698110248568964e-05, "loss": 2.627, "step": 42109 }, { "epoch": 3.3, "learning_rate": 6.698031835646514e-05, "loss": 2.6789, "step": 42110 }, { "epoch": 3.3, "learning_rate": 6.697953422724065e-05, "loss": 2.9585, "step": 42111 }, { "epoch": 3.3, "learning_rate": 6.697875009801615e-05, "loss": 2.6698, "step": 42112 }, { "epoch": 3.3, "learning_rate": 6.697796596879166e-05, "loss": 2.6895, "step": 42113 }, { "epoch": 3.3, "learning_rate": 6.697718183956716e-05, "loss": 2.669, "step": 42114 }, { "epoch": 3.3, "learning_rate": 6.697639771034266e-05, "loss": 2.6975, "step": 42115 }, { "epoch": 3.3, "learning_rate": 6.697561358111817e-05, "loss": 2.7592, "step": 42116 }, { "epoch": 3.3, "learning_rate": 6.697482945189367e-05, "loss": 2.6433, "step": 42117 }, { "epoch": 3.3, "learning_rate": 6.697404532266917e-05, "loss": 2.9111, "step": 42118 }, { "epoch": 3.3, "learning_rate": 6.697326119344468e-05, "loss": 2.6798, "step": 42119 }, { "epoch": 3.3, "learning_rate": 6.697247706422018e-05, "loss": 2.5882, "step": 42120 }, { "epoch": 3.3, "learning_rate": 6.697169293499569e-05, "loss": 2.7513, "step": 42121 }, { "epoch": 3.3, "learning_rate": 6.697090880577119e-05, "loss": 2.7164, "step": 42122 }, { "epoch": 3.3, "learning_rate": 6.697012467654669e-05, "loss": 2.9209, "step": 42123 }, { "epoch": 3.3, "learning_rate": 6.69693405473222e-05, "loss": 2.9353, "step": 42124 }, { "epoch": 3.3, "learning_rate": 6.69685564180977e-05, "loss": 2.8012, "step": 42125 }, { "epoch": 3.3, "learning_rate": 6.69677722888732e-05, "loss": 2.6817, "step": 42126 }, { "epoch": 3.3, "learning_rate": 6.696698815964871e-05, "loss": 2.6658, "step": 42127 }, { "epoch": 3.3, "learning_rate": 6.696620403042422e-05, "loss": 2.5458, "step": 42128 }, { "epoch": 3.3, "learning_rate": 6.696541990119973e-05, "loss": 2.622, "step": 42129 }, { "epoch": 3.3, "learning_rate": 6.696463577197523e-05, "loss": 2.7292, "step": 42130 }, { "epoch": 3.3, "learning_rate": 6.696385164275074e-05, "loss": 3.0265, "step": 42131 }, { "epoch": 3.3, "learning_rate": 6.696306751352624e-05, "loss": 2.7089, "step": 42132 }, { "epoch": 3.3, "learning_rate": 6.696228338430174e-05, "loss": 2.667, "step": 42133 }, { "epoch": 3.3, "learning_rate": 6.696149925507725e-05, "loss": 2.8362, "step": 42134 }, { "epoch": 3.3, "learning_rate": 6.696071512585275e-05, "loss": 2.7956, "step": 42135 }, { "epoch": 3.3, "learning_rate": 6.695993099662825e-05, "loss": 2.5524, "step": 42136 }, { "epoch": 3.3, "learning_rate": 6.695914686740376e-05, "loss": 2.7622, "step": 42137 }, { "epoch": 3.3, "learning_rate": 6.695836273817926e-05, "loss": 2.8657, "step": 42138 }, { "epoch": 3.3, "learning_rate": 6.695757860895476e-05, "loss": 2.6694, "step": 42139 }, { "epoch": 3.3, "learning_rate": 6.695679447973027e-05, "loss": 2.5602, "step": 42140 }, { "epoch": 3.3, "learning_rate": 6.695601035050577e-05, "loss": 3.0235, "step": 42141 }, { "epoch": 3.3, "learning_rate": 6.695522622128128e-05, "loss": 2.9352, "step": 42142 }, { "epoch": 3.3, "learning_rate": 6.695444209205678e-05, "loss": 2.7672, "step": 42143 }, { "epoch": 3.3, "learning_rate": 6.695365796283227e-05, "loss": 2.6705, "step": 42144 }, { "epoch": 3.3, "learning_rate": 6.695287383360777e-05, "loss": 3.1305, "step": 42145 }, { "epoch": 3.3, "learning_rate": 6.695208970438328e-05, "loss": 2.6023, "step": 42146 }, { "epoch": 3.3, "learning_rate": 6.695130557515878e-05, "loss": 2.7731, "step": 42147 }, { "epoch": 3.3, "learning_rate": 6.695052144593428e-05, "loss": 2.8937, "step": 42148 }, { "epoch": 3.31, "learning_rate": 6.694973731670979e-05, "loss": 2.6561, "step": 42149 }, { "epoch": 3.31, "learning_rate": 6.694895318748529e-05, "loss": 2.4871, "step": 42150 }, { "epoch": 3.31, "learning_rate": 6.69481690582608e-05, "loss": 2.5051, "step": 42151 }, { "epoch": 3.31, "learning_rate": 6.69473849290363e-05, "loss": 2.6802, "step": 42152 }, { "epoch": 3.31, "learning_rate": 6.69466007998118e-05, "loss": 2.7585, "step": 42153 }, { "epoch": 3.31, "learning_rate": 6.694581667058732e-05, "loss": 2.9847, "step": 42154 }, { "epoch": 3.31, "learning_rate": 6.694503254136282e-05, "loss": 2.7308, "step": 42155 }, { "epoch": 3.31, "learning_rate": 6.694424841213833e-05, "loss": 2.6772, "step": 42156 }, { "epoch": 3.31, "learning_rate": 6.694346428291383e-05, "loss": 2.6232, "step": 42157 }, { "epoch": 3.31, "learning_rate": 6.694268015368933e-05, "loss": 2.6328, "step": 42158 }, { "epoch": 3.31, "learning_rate": 6.694189602446484e-05, "loss": 2.9228, "step": 42159 }, { "epoch": 3.31, "learning_rate": 6.694111189524034e-05, "loss": 2.5722, "step": 42160 }, { "epoch": 3.31, "learning_rate": 6.694032776601585e-05, "loss": 2.6308, "step": 42161 }, { "epoch": 3.31, "learning_rate": 6.693954363679135e-05, "loss": 2.5812, "step": 42162 }, { "epoch": 3.31, "learning_rate": 6.693875950756685e-05, "loss": 2.6813, "step": 42163 }, { "epoch": 3.31, "learning_rate": 6.693797537834236e-05, "loss": 2.674, "step": 42164 }, { "epoch": 3.31, "learning_rate": 6.693719124911786e-05, "loss": 2.8679, "step": 42165 }, { "epoch": 3.31, "learning_rate": 6.693640711989336e-05, "loss": 2.8683, "step": 42166 }, { "epoch": 3.31, "learning_rate": 6.693562299066887e-05, "loss": 2.6752, "step": 42167 }, { "epoch": 3.31, "learning_rate": 6.693483886144437e-05, "loss": 2.8402, "step": 42168 }, { "epoch": 3.31, "learning_rate": 6.693405473221988e-05, "loss": 2.7631, "step": 42169 }, { "epoch": 3.31, "learning_rate": 6.693327060299538e-05, "loss": 2.8013, "step": 42170 }, { "epoch": 3.31, "learning_rate": 6.693248647377088e-05, "loss": 2.7396, "step": 42171 }, { "epoch": 3.31, "learning_rate": 6.693170234454639e-05, "loss": 2.7648, "step": 42172 }, { "epoch": 3.31, "learning_rate": 6.693091821532189e-05, "loss": 2.6229, "step": 42173 }, { "epoch": 3.31, "learning_rate": 6.69301340860974e-05, "loss": 2.9386, "step": 42174 }, { "epoch": 3.31, "learning_rate": 6.69293499568729e-05, "loss": 2.9655, "step": 42175 }, { "epoch": 3.31, "learning_rate": 6.69285658276484e-05, "loss": 2.672, "step": 42176 }, { "epoch": 3.31, "learning_rate": 6.69277816984239e-05, "loss": 2.73, "step": 42177 }, { "epoch": 3.31, "learning_rate": 6.692699756919941e-05, "loss": 2.7936, "step": 42178 }, { "epoch": 3.31, "learning_rate": 6.692621343997491e-05, "loss": 2.7033, "step": 42179 }, { "epoch": 3.31, "learning_rate": 6.692542931075042e-05, "loss": 2.7744, "step": 42180 }, { "epoch": 3.31, "learning_rate": 6.692464518152592e-05, "loss": 2.9166, "step": 42181 }, { "epoch": 3.31, "learning_rate": 6.692386105230142e-05, "loss": 2.6924, "step": 42182 }, { "epoch": 3.31, "learning_rate": 6.692307692307693e-05, "loss": 2.7219, "step": 42183 }, { "epoch": 3.31, "learning_rate": 6.692229279385243e-05, "loss": 2.7369, "step": 42184 }, { "epoch": 3.31, "learning_rate": 6.692150866462793e-05, "loss": 2.7351, "step": 42185 }, { "epoch": 3.31, "learning_rate": 6.692072453540344e-05, "loss": 2.6782, "step": 42186 }, { "epoch": 3.31, "learning_rate": 6.691994040617894e-05, "loss": 2.9809, "step": 42187 }, { "epoch": 3.31, "learning_rate": 6.691915627695444e-05, "loss": 2.6755, "step": 42188 }, { "epoch": 3.31, "learning_rate": 6.691837214772995e-05, "loss": 2.4919, "step": 42189 }, { "epoch": 3.31, "learning_rate": 6.691758801850545e-05, "loss": 2.5444, "step": 42190 }, { "epoch": 3.31, "learning_rate": 6.691680388928096e-05, "loss": 2.8677, "step": 42191 }, { "epoch": 3.31, "learning_rate": 6.691601976005646e-05, "loss": 2.6808, "step": 42192 }, { "epoch": 3.31, "learning_rate": 6.691523563083196e-05, "loss": 2.6108, "step": 42193 }, { "epoch": 3.31, "learning_rate": 6.691445150160747e-05, "loss": 2.9106, "step": 42194 }, { "epoch": 3.31, "learning_rate": 6.691366737238297e-05, "loss": 2.7004, "step": 42195 }, { "epoch": 3.31, "learning_rate": 6.691288324315847e-05, "loss": 2.6777, "step": 42196 }, { "epoch": 3.31, "learning_rate": 6.691209911393398e-05, "loss": 2.5159, "step": 42197 }, { "epoch": 3.31, "learning_rate": 6.691131498470948e-05, "loss": 2.692, "step": 42198 }, { "epoch": 3.31, "learning_rate": 6.691053085548499e-05, "loss": 2.7331, "step": 42199 }, { "epoch": 3.31, "learning_rate": 6.690974672626049e-05, "loss": 2.5441, "step": 42200 }, { "epoch": 3.31, "learning_rate": 6.690896259703599e-05, "loss": 2.8161, "step": 42201 }, { "epoch": 3.31, "learning_rate": 6.69081784678115e-05, "loss": 2.7209, "step": 42202 }, { "epoch": 3.31, "learning_rate": 6.6907394338587e-05, "loss": 2.8688, "step": 42203 }, { "epoch": 3.31, "learning_rate": 6.69066102093625e-05, "loss": 2.6858, "step": 42204 }, { "epoch": 3.31, "learning_rate": 6.690582608013801e-05, "loss": 2.8976, "step": 42205 }, { "epoch": 3.31, "learning_rate": 6.690504195091351e-05, "loss": 2.8158, "step": 42206 }, { "epoch": 3.31, "learning_rate": 6.690425782168901e-05, "loss": 2.5857, "step": 42207 }, { "epoch": 3.31, "learning_rate": 6.690347369246452e-05, "loss": 2.6554, "step": 42208 }, { "epoch": 3.31, "learning_rate": 6.690268956324002e-05, "loss": 2.7386, "step": 42209 }, { "epoch": 3.31, "learning_rate": 6.690190543401554e-05, "loss": 2.44, "step": 42210 }, { "epoch": 3.31, "learning_rate": 6.690112130479104e-05, "loss": 2.9038, "step": 42211 }, { "epoch": 3.31, "learning_rate": 6.690033717556655e-05, "loss": 2.7795, "step": 42212 }, { "epoch": 3.31, "learning_rate": 6.689955304634205e-05, "loss": 2.6276, "step": 42213 }, { "epoch": 3.31, "learning_rate": 6.689876891711755e-05, "loss": 2.7245, "step": 42214 }, { "epoch": 3.31, "learning_rate": 6.689798478789306e-05, "loss": 2.7942, "step": 42215 }, { "epoch": 3.31, "learning_rate": 6.689720065866855e-05, "loss": 2.6795, "step": 42216 }, { "epoch": 3.31, "learning_rate": 6.689641652944405e-05, "loss": 2.6294, "step": 42217 }, { "epoch": 3.31, "learning_rate": 6.689563240021956e-05, "loss": 2.6552, "step": 42218 }, { "epoch": 3.31, "learning_rate": 6.689484827099506e-05, "loss": 2.5026, "step": 42219 }, { "epoch": 3.31, "learning_rate": 6.689406414177056e-05, "loss": 2.6668, "step": 42220 }, { "epoch": 3.31, "learning_rate": 6.689328001254607e-05, "loss": 2.7785, "step": 42221 }, { "epoch": 3.31, "learning_rate": 6.689249588332157e-05, "loss": 2.9122, "step": 42222 }, { "epoch": 3.31, "learning_rate": 6.689171175409707e-05, "loss": 2.8436, "step": 42223 }, { "epoch": 3.31, "learning_rate": 6.689092762487258e-05, "loss": 2.785, "step": 42224 }, { "epoch": 3.31, "learning_rate": 6.689014349564808e-05, "loss": 2.6895, "step": 42225 }, { "epoch": 3.31, "learning_rate": 6.688935936642358e-05, "loss": 2.7727, "step": 42226 }, { "epoch": 3.31, "learning_rate": 6.688857523719909e-05, "loss": 2.669, "step": 42227 }, { "epoch": 3.31, "learning_rate": 6.688779110797459e-05, "loss": 2.6124, "step": 42228 }, { "epoch": 3.31, "learning_rate": 6.68870069787501e-05, "loss": 2.7591, "step": 42229 }, { "epoch": 3.31, "learning_rate": 6.68862228495256e-05, "loss": 2.9722, "step": 42230 }, { "epoch": 3.31, "learning_rate": 6.68854387203011e-05, "loss": 2.7004, "step": 42231 }, { "epoch": 3.31, "learning_rate": 6.688465459107661e-05, "loss": 2.914, "step": 42232 }, { "epoch": 3.31, "learning_rate": 6.688387046185211e-05, "loss": 2.6653, "step": 42233 }, { "epoch": 3.31, "learning_rate": 6.688308633262761e-05, "loss": 2.5732, "step": 42234 }, { "epoch": 3.31, "learning_rate": 6.688230220340312e-05, "loss": 2.7476, "step": 42235 }, { "epoch": 3.31, "learning_rate": 6.688151807417864e-05, "loss": 2.7581, "step": 42236 }, { "epoch": 3.31, "learning_rate": 6.688073394495414e-05, "loss": 2.7925, "step": 42237 }, { "epoch": 3.31, "learning_rate": 6.687994981572964e-05, "loss": 2.8092, "step": 42238 }, { "epoch": 3.31, "learning_rate": 6.687916568650515e-05, "loss": 2.445, "step": 42239 }, { "epoch": 3.31, "learning_rate": 6.687838155728065e-05, "loss": 2.5882, "step": 42240 }, { "epoch": 3.31, "learning_rate": 6.687759742805615e-05, "loss": 2.6223, "step": 42241 }, { "epoch": 3.31, "learning_rate": 6.687681329883166e-05, "loss": 2.9079, "step": 42242 }, { "epoch": 3.31, "learning_rate": 6.687602916960716e-05, "loss": 2.763, "step": 42243 }, { "epoch": 3.31, "learning_rate": 6.687524504038266e-05, "loss": 2.7562, "step": 42244 }, { "epoch": 3.31, "learning_rate": 6.687446091115817e-05, "loss": 2.7855, "step": 42245 }, { "epoch": 3.31, "learning_rate": 6.687367678193367e-05, "loss": 2.7282, "step": 42246 }, { "epoch": 3.31, "learning_rate": 6.687289265270918e-05, "loss": 2.8011, "step": 42247 }, { "epoch": 3.31, "learning_rate": 6.687210852348468e-05, "loss": 2.7743, "step": 42248 }, { "epoch": 3.31, "learning_rate": 6.687132439426018e-05, "loss": 2.7271, "step": 42249 }, { "epoch": 3.31, "learning_rate": 6.687054026503569e-05, "loss": 2.7222, "step": 42250 }, { "epoch": 3.31, "learning_rate": 6.686975613581119e-05, "loss": 2.6434, "step": 42251 }, { "epoch": 3.31, "learning_rate": 6.686897200658668e-05, "loss": 2.6379, "step": 42252 }, { "epoch": 3.31, "learning_rate": 6.686818787736218e-05, "loss": 2.955, "step": 42253 }, { "epoch": 3.31, "learning_rate": 6.686740374813769e-05, "loss": 2.677, "step": 42254 }, { "epoch": 3.31, "learning_rate": 6.686661961891319e-05, "loss": 2.6772, "step": 42255 }, { "epoch": 3.31, "learning_rate": 6.68658354896887e-05, "loss": 2.8717, "step": 42256 }, { "epoch": 3.31, "learning_rate": 6.68650513604642e-05, "loss": 2.8233, "step": 42257 }, { "epoch": 3.31, "learning_rate": 6.68642672312397e-05, "loss": 2.6683, "step": 42258 }, { "epoch": 3.31, "learning_rate": 6.68634831020152e-05, "loss": 2.6864, "step": 42259 }, { "epoch": 3.31, "learning_rate": 6.686269897279071e-05, "loss": 2.8847, "step": 42260 }, { "epoch": 3.31, "learning_rate": 6.686191484356621e-05, "loss": 2.709, "step": 42261 }, { "epoch": 3.31, "learning_rate": 6.686113071434173e-05, "loss": 2.6396, "step": 42262 }, { "epoch": 3.31, "learning_rate": 6.686034658511723e-05, "loss": 2.7162, "step": 42263 }, { "epoch": 3.31, "learning_rate": 6.685956245589274e-05, "loss": 2.6927, "step": 42264 }, { "epoch": 3.31, "learning_rate": 6.685877832666824e-05, "loss": 2.6738, "step": 42265 }, { "epoch": 3.31, "learning_rate": 6.685799419744375e-05, "loss": 2.8766, "step": 42266 }, { "epoch": 3.31, "learning_rate": 6.685721006821925e-05, "loss": 2.9232, "step": 42267 }, { "epoch": 3.31, "learning_rate": 6.685642593899475e-05, "loss": 2.8908, "step": 42268 }, { "epoch": 3.31, "learning_rate": 6.685564180977026e-05, "loss": 2.8279, "step": 42269 }, { "epoch": 3.31, "learning_rate": 6.685485768054576e-05, "loss": 2.8714, "step": 42270 }, { "epoch": 3.31, "learning_rate": 6.685407355132126e-05, "loss": 2.607, "step": 42271 }, { "epoch": 3.31, "learning_rate": 6.685328942209677e-05, "loss": 2.7416, "step": 42272 }, { "epoch": 3.31, "learning_rate": 6.685250529287227e-05, "loss": 2.5713, "step": 42273 }, { "epoch": 3.31, "learning_rate": 6.685172116364777e-05, "loss": 2.6259, "step": 42274 }, { "epoch": 3.31, "learning_rate": 6.685093703442328e-05, "loss": 2.7235, "step": 42275 }, { "epoch": 3.31, "learning_rate": 6.685015290519878e-05, "loss": 2.3955, "step": 42276 }, { "epoch": 3.32, "learning_rate": 6.684936877597429e-05, "loss": 2.6756, "step": 42277 }, { "epoch": 3.32, "learning_rate": 6.684858464674979e-05, "loss": 2.6174, "step": 42278 }, { "epoch": 3.32, "learning_rate": 6.684780051752529e-05, "loss": 2.9439, "step": 42279 }, { "epoch": 3.32, "learning_rate": 6.68470163883008e-05, "loss": 2.756, "step": 42280 }, { "epoch": 3.32, "learning_rate": 6.68462322590763e-05, "loss": 2.7816, "step": 42281 }, { "epoch": 3.32, "learning_rate": 6.68454481298518e-05, "loss": 2.5773, "step": 42282 }, { "epoch": 3.32, "learning_rate": 6.684466400062731e-05, "loss": 2.7398, "step": 42283 }, { "epoch": 3.32, "learning_rate": 6.684387987140281e-05, "loss": 2.7477, "step": 42284 }, { "epoch": 3.32, "learning_rate": 6.684309574217832e-05, "loss": 2.5068, "step": 42285 }, { "epoch": 3.32, "learning_rate": 6.684231161295382e-05, "loss": 2.7736, "step": 42286 }, { "epoch": 3.32, "learning_rate": 6.684152748372932e-05, "loss": 2.6216, "step": 42287 }, { "epoch": 3.32, "learning_rate": 6.684074335450483e-05, "loss": 2.8436, "step": 42288 }, { "epoch": 3.32, "learning_rate": 6.683995922528033e-05, "loss": 2.709, "step": 42289 }, { "epoch": 3.32, "learning_rate": 6.683917509605583e-05, "loss": 2.8648, "step": 42290 }, { "epoch": 3.32, "learning_rate": 6.683839096683134e-05, "loss": 2.6632, "step": 42291 }, { "epoch": 3.32, "learning_rate": 6.683760683760684e-05, "loss": 2.6036, "step": 42292 }, { "epoch": 3.32, "learning_rate": 6.683682270838234e-05, "loss": 2.9712, "step": 42293 }, { "epoch": 3.32, "learning_rate": 6.683603857915785e-05, "loss": 2.8296, "step": 42294 }, { "epoch": 3.32, "learning_rate": 6.683525444993335e-05, "loss": 2.7043, "step": 42295 }, { "epoch": 3.32, "learning_rate": 6.683447032070886e-05, "loss": 2.7343, "step": 42296 }, { "epoch": 3.32, "learning_rate": 6.683368619148436e-05, "loss": 2.651, "step": 42297 }, { "epoch": 3.32, "learning_rate": 6.683290206225986e-05, "loss": 2.6602, "step": 42298 }, { "epoch": 3.32, "learning_rate": 6.683211793303537e-05, "loss": 2.5997, "step": 42299 }, { "epoch": 3.32, "learning_rate": 6.683133380381087e-05, "loss": 2.8976, "step": 42300 }, { "epoch": 3.32, "learning_rate": 6.683054967458637e-05, "loss": 2.7384, "step": 42301 }, { "epoch": 3.32, "learning_rate": 6.682976554536188e-05, "loss": 2.7766, "step": 42302 }, { "epoch": 3.32, "learning_rate": 6.682898141613738e-05, "loss": 2.701, "step": 42303 }, { "epoch": 3.32, "learning_rate": 6.682819728691289e-05, "loss": 2.6218, "step": 42304 }, { "epoch": 3.32, "learning_rate": 6.682741315768839e-05, "loss": 2.7113, "step": 42305 }, { "epoch": 3.32, "learning_rate": 6.682662902846389e-05, "loss": 2.7945, "step": 42306 }, { "epoch": 3.32, "learning_rate": 6.68258448992394e-05, "loss": 2.9885, "step": 42307 }, { "epoch": 3.32, "learning_rate": 6.68250607700149e-05, "loss": 2.6327, "step": 42308 }, { "epoch": 3.32, "learning_rate": 6.68242766407904e-05, "loss": 2.7781, "step": 42309 }, { "epoch": 3.32, "learning_rate": 6.682349251156591e-05, "loss": 2.8372, "step": 42310 }, { "epoch": 3.32, "learning_rate": 6.682270838234141e-05, "loss": 2.7077, "step": 42311 }, { "epoch": 3.32, "learning_rate": 6.682192425311691e-05, "loss": 2.8286, "step": 42312 }, { "epoch": 3.32, "learning_rate": 6.682114012389242e-05, "loss": 2.8812, "step": 42313 }, { "epoch": 3.32, "learning_rate": 6.682035599466792e-05, "loss": 2.657, "step": 42314 }, { "epoch": 3.32, "learning_rate": 6.681957186544343e-05, "loss": 2.7991, "step": 42315 }, { "epoch": 3.32, "learning_rate": 6.681878773621893e-05, "loss": 2.6935, "step": 42316 }, { "epoch": 3.32, "learning_rate": 6.681800360699443e-05, "loss": 2.9691, "step": 42317 }, { "epoch": 3.32, "learning_rate": 6.681721947776994e-05, "loss": 2.7863, "step": 42318 }, { "epoch": 3.32, "learning_rate": 6.681643534854545e-05, "loss": 2.817, "step": 42319 }, { "epoch": 3.32, "learning_rate": 6.681565121932096e-05, "loss": 2.8441, "step": 42320 }, { "epoch": 3.32, "learning_rate": 6.681486709009646e-05, "loss": 2.7896, "step": 42321 }, { "epoch": 3.32, "learning_rate": 6.681408296087196e-05, "loss": 2.5934, "step": 42322 }, { "epoch": 3.32, "learning_rate": 6.681329883164747e-05, "loss": 2.8001, "step": 42323 }, { "epoch": 3.32, "learning_rate": 6.681251470242296e-05, "loss": 2.6474, "step": 42324 }, { "epoch": 3.32, "learning_rate": 6.681173057319846e-05, "loss": 2.7475, "step": 42325 }, { "epoch": 3.32, "learning_rate": 6.681094644397397e-05, "loss": 2.7395, "step": 42326 }, { "epoch": 3.32, "learning_rate": 6.681016231474947e-05, "loss": 2.5675, "step": 42327 }, { "epoch": 3.32, "learning_rate": 6.680937818552497e-05, "loss": 2.6526, "step": 42328 }, { "epoch": 3.32, "learning_rate": 6.680859405630048e-05, "loss": 2.7434, "step": 42329 }, { "epoch": 3.32, "learning_rate": 6.680780992707598e-05, "loss": 2.6501, "step": 42330 }, { "epoch": 3.32, "learning_rate": 6.680702579785148e-05, "loss": 2.7079, "step": 42331 }, { "epoch": 3.32, "learning_rate": 6.680624166862699e-05, "loss": 2.6283, "step": 42332 }, { "epoch": 3.32, "learning_rate": 6.680545753940249e-05, "loss": 2.7239, "step": 42333 }, { "epoch": 3.32, "learning_rate": 6.6804673410178e-05, "loss": 2.7413, "step": 42334 }, { "epoch": 3.32, "learning_rate": 6.68038892809535e-05, "loss": 2.5007, "step": 42335 }, { "epoch": 3.32, "learning_rate": 6.6803105151729e-05, "loss": 2.6872, "step": 42336 }, { "epoch": 3.32, "learning_rate": 6.68023210225045e-05, "loss": 2.664, "step": 42337 }, { "epoch": 3.32, "learning_rate": 6.680153689328001e-05, "loss": 2.6627, "step": 42338 }, { "epoch": 3.32, "learning_rate": 6.680075276405551e-05, "loss": 2.7908, "step": 42339 }, { "epoch": 3.32, "learning_rate": 6.679996863483102e-05, "loss": 3.1034, "step": 42340 }, { "epoch": 3.32, "learning_rate": 6.679918450560652e-05, "loss": 2.6452, "step": 42341 }, { "epoch": 3.32, "learning_rate": 6.679840037638202e-05, "loss": 2.5283, "step": 42342 }, { "epoch": 3.32, "learning_rate": 6.679761624715753e-05, "loss": 2.8004, "step": 42343 }, { "epoch": 3.32, "learning_rate": 6.679683211793305e-05, "loss": 2.9803, "step": 42344 }, { "epoch": 3.32, "learning_rate": 6.679604798870855e-05, "loss": 2.7387, "step": 42345 }, { "epoch": 3.32, "learning_rate": 6.679526385948405e-05, "loss": 2.887, "step": 42346 }, { "epoch": 3.32, "learning_rate": 6.679447973025956e-05, "loss": 2.6613, "step": 42347 }, { "epoch": 3.32, "learning_rate": 6.679369560103506e-05, "loss": 2.4282, "step": 42348 }, { "epoch": 3.32, "learning_rate": 6.679291147181056e-05, "loss": 2.5673, "step": 42349 }, { "epoch": 3.32, "learning_rate": 6.679212734258607e-05, "loss": 2.6002, "step": 42350 }, { "epoch": 3.32, "learning_rate": 6.679134321336157e-05, "loss": 2.6488, "step": 42351 }, { "epoch": 3.32, "learning_rate": 6.679055908413708e-05, "loss": 2.7402, "step": 42352 }, { "epoch": 3.32, "learning_rate": 6.678977495491258e-05, "loss": 2.7799, "step": 42353 }, { "epoch": 3.32, "learning_rate": 6.678899082568808e-05, "loss": 2.6032, "step": 42354 }, { "epoch": 3.32, "learning_rate": 6.678820669646359e-05, "loss": 2.8563, "step": 42355 }, { "epoch": 3.32, "learning_rate": 6.678742256723909e-05, "loss": 2.858, "step": 42356 }, { "epoch": 3.32, "learning_rate": 6.67866384380146e-05, "loss": 2.7321, "step": 42357 }, { "epoch": 3.32, "learning_rate": 6.67858543087901e-05, "loss": 2.9406, "step": 42358 }, { "epoch": 3.32, "learning_rate": 6.67850701795656e-05, "loss": 2.8485, "step": 42359 }, { "epoch": 3.32, "learning_rate": 6.678428605034109e-05, "loss": 2.8247, "step": 42360 }, { "epoch": 3.32, "learning_rate": 6.67835019211166e-05, "loss": 2.7635, "step": 42361 }, { "epoch": 3.32, "learning_rate": 6.67827177918921e-05, "loss": 2.6657, "step": 42362 }, { "epoch": 3.32, "learning_rate": 6.67819336626676e-05, "loss": 2.7771, "step": 42363 }, { "epoch": 3.32, "learning_rate": 6.67811495334431e-05, "loss": 2.603, "step": 42364 }, { "epoch": 3.32, "learning_rate": 6.678036540421861e-05, "loss": 2.8922, "step": 42365 }, { "epoch": 3.32, "learning_rate": 6.677958127499411e-05, "loss": 2.7791, "step": 42366 }, { "epoch": 3.32, "learning_rate": 6.677879714576962e-05, "loss": 2.7595, "step": 42367 }, { "epoch": 3.32, "learning_rate": 6.677801301654512e-05, "loss": 2.712, "step": 42368 }, { "epoch": 3.32, "learning_rate": 6.677722888732062e-05, "loss": 2.7673, "step": 42369 }, { "epoch": 3.32, "learning_rate": 6.677644475809614e-05, "loss": 2.6197, "step": 42370 }, { "epoch": 3.32, "learning_rate": 6.677566062887165e-05, "loss": 2.7904, "step": 42371 }, { "epoch": 3.32, "learning_rate": 6.677487649964715e-05, "loss": 2.66, "step": 42372 }, { "epoch": 3.32, "learning_rate": 6.677409237042265e-05, "loss": 2.6256, "step": 42373 }, { "epoch": 3.32, "learning_rate": 6.677330824119816e-05, "loss": 2.7197, "step": 42374 }, { "epoch": 3.32, "learning_rate": 6.677252411197366e-05, "loss": 2.8162, "step": 42375 }, { "epoch": 3.32, "learning_rate": 6.677173998274916e-05, "loss": 2.5181, "step": 42376 }, { "epoch": 3.32, "learning_rate": 6.677095585352467e-05, "loss": 3.0067, "step": 42377 }, { "epoch": 3.32, "learning_rate": 6.677017172430017e-05, "loss": 2.6963, "step": 42378 }, { "epoch": 3.32, "learning_rate": 6.676938759507567e-05, "loss": 2.6449, "step": 42379 }, { "epoch": 3.32, "learning_rate": 6.676860346585118e-05, "loss": 2.6101, "step": 42380 }, { "epoch": 3.32, "learning_rate": 6.676781933662668e-05, "loss": 2.9119, "step": 42381 }, { "epoch": 3.32, "learning_rate": 6.676703520740219e-05, "loss": 2.7187, "step": 42382 }, { "epoch": 3.32, "learning_rate": 6.676625107817769e-05, "loss": 2.5337, "step": 42383 }, { "epoch": 3.32, "learning_rate": 6.676546694895319e-05, "loss": 2.7539, "step": 42384 }, { "epoch": 3.32, "learning_rate": 6.67646828197287e-05, "loss": 2.7817, "step": 42385 }, { "epoch": 3.32, "learning_rate": 6.67638986905042e-05, "loss": 2.7071, "step": 42386 }, { "epoch": 3.32, "learning_rate": 6.67631145612797e-05, "loss": 2.5111, "step": 42387 }, { "epoch": 3.32, "learning_rate": 6.676233043205521e-05, "loss": 2.6931, "step": 42388 }, { "epoch": 3.32, "learning_rate": 6.676154630283071e-05, "loss": 2.875, "step": 42389 }, { "epoch": 3.32, "learning_rate": 6.676076217360621e-05, "loss": 2.7362, "step": 42390 }, { "epoch": 3.32, "learning_rate": 6.675997804438172e-05, "loss": 2.8442, "step": 42391 }, { "epoch": 3.32, "learning_rate": 6.675919391515722e-05, "loss": 2.7529, "step": 42392 }, { "epoch": 3.32, "learning_rate": 6.675840978593273e-05, "loss": 2.9337, "step": 42393 }, { "epoch": 3.32, "learning_rate": 6.675762565670823e-05, "loss": 2.6848, "step": 42394 }, { "epoch": 3.32, "learning_rate": 6.675684152748373e-05, "loss": 2.6187, "step": 42395 }, { "epoch": 3.32, "learning_rate": 6.675605739825924e-05, "loss": 2.6727, "step": 42396 }, { "epoch": 3.32, "learning_rate": 6.675527326903474e-05, "loss": 2.7698, "step": 42397 }, { "epoch": 3.32, "learning_rate": 6.675448913981024e-05, "loss": 2.9866, "step": 42398 }, { "epoch": 3.32, "learning_rate": 6.675370501058575e-05, "loss": 2.63, "step": 42399 }, { "epoch": 3.32, "learning_rate": 6.675292088136125e-05, "loss": 2.8067, "step": 42400 }, { "epoch": 3.32, "learning_rate": 6.675213675213676e-05, "loss": 2.5203, "step": 42401 }, { "epoch": 3.32, "learning_rate": 6.675135262291226e-05, "loss": 2.7786, "step": 42402 }, { "epoch": 3.32, "learning_rate": 6.675056849368776e-05, "loss": 2.6123, "step": 42403 }, { "epoch": 3.33, "learning_rate": 6.674978436446327e-05, "loss": 2.7538, "step": 42404 }, { "epoch": 3.33, "learning_rate": 6.674900023523877e-05, "loss": 2.8009, "step": 42405 }, { "epoch": 3.33, "learning_rate": 6.674821610601427e-05, "loss": 2.759, "step": 42406 }, { "epoch": 3.33, "learning_rate": 6.674743197678978e-05, "loss": 3.0041, "step": 42407 }, { "epoch": 3.33, "learning_rate": 6.674664784756528e-05, "loss": 2.6551, "step": 42408 }, { "epoch": 3.33, "learning_rate": 6.674586371834078e-05, "loss": 2.8254, "step": 42409 }, { "epoch": 3.33, "learning_rate": 6.674507958911629e-05, "loss": 2.7816, "step": 42410 }, { "epoch": 3.33, "learning_rate": 6.674429545989179e-05, "loss": 2.9424, "step": 42411 }, { "epoch": 3.33, "learning_rate": 6.67435113306673e-05, "loss": 2.6076, "step": 42412 }, { "epoch": 3.33, "learning_rate": 6.67427272014428e-05, "loss": 2.7289, "step": 42413 }, { "epoch": 3.33, "learning_rate": 6.67419430722183e-05, "loss": 2.5705, "step": 42414 }, { "epoch": 3.33, "learning_rate": 6.674115894299381e-05, "loss": 2.7962, "step": 42415 }, { "epoch": 3.33, "learning_rate": 6.674037481376931e-05, "loss": 2.7745, "step": 42416 }, { "epoch": 3.33, "learning_rate": 6.673959068454481e-05, "loss": 2.7599, "step": 42417 }, { "epoch": 3.33, "learning_rate": 6.673880655532032e-05, "loss": 2.4755, "step": 42418 }, { "epoch": 3.33, "learning_rate": 6.673802242609582e-05, "loss": 2.6053, "step": 42419 }, { "epoch": 3.33, "learning_rate": 6.673723829687133e-05, "loss": 2.7297, "step": 42420 }, { "epoch": 3.33, "learning_rate": 6.673645416764683e-05, "loss": 2.6983, "step": 42421 }, { "epoch": 3.33, "learning_rate": 6.673567003842233e-05, "loss": 2.8618, "step": 42422 }, { "epoch": 3.33, "learning_rate": 6.673488590919784e-05, "loss": 2.7469, "step": 42423 }, { "epoch": 3.33, "learning_rate": 6.673410177997334e-05, "loss": 2.6045, "step": 42424 }, { "epoch": 3.33, "learning_rate": 6.673331765074884e-05, "loss": 2.659, "step": 42425 }, { "epoch": 3.33, "learning_rate": 6.673253352152435e-05, "loss": 2.6509, "step": 42426 }, { "epoch": 3.33, "learning_rate": 6.673174939229986e-05, "loss": 2.7273, "step": 42427 }, { "epoch": 3.33, "learning_rate": 6.673096526307537e-05, "loss": 2.727, "step": 42428 }, { "epoch": 3.33, "learning_rate": 6.673018113385087e-05, "loss": 2.583, "step": 42429 }, { "epoch": 3.33, "learning_rate": 6.672939700462638e-05, "loss": 2.6676, "step": 42430 }, { "epoch": 3.33, "learning_rate": 6.672861287540188e-05, "loss": 2.5601, "step": 42431 }, { "epoch": 3.33, "learning_rate": 6.672782874617737e-05, "loss": 2.6788, "step": 42432 }, { "epoch": 3.33, "learning_rate": 6.672704461695287e-05, "loss": 2.5997, "step": 42433 }, { "epoch": 3.33, "learning_rate": 6.672626048772838e-05, "loss": 2.821, "step": 42434 }, { "epoch": 3.33, "learning_rate": 6.672547635850388e-05, "loss": 2.5271, "step": 42435 }, { "epoch": 3.33, "learning_rate": 6.672469222927938e-05, "loss": 2.7704, "step": 42436 }, { "epoch": 3.33, "learning_rate": 6.672390810005489e-05, "loss": 2.5583, "step": 42437 }, { "epoch": 3.33, "learning_rate": 6.672312397083039e-05, "loss": 2.6391, "step": 42438 }, { "epoch": 3.33, "learning_rate": 6.67223398416059e-05, "loss": 2.675, "step": 42439 }, { "epoch": 3.33, "learning_rate": 6.67215557123814e-05, "loss": 2.6559, "step": 42440 }, { "epoch": 3.33, "learning_rate": 6.67207715831569e-05, "loss": 2.5379, "step": 42441 }, { "epoch": 3.33, "learning_rate": 6.67199874539324e-05, "loss": 2.437, "step": 42442 }, { "epoch": 3.33, "learning_rate": 6.671920332470791e-05, "loss": 2.7215, "step": 42443 }, { "epoch": 3.33, "learning_rate": 6.671841919548341e-05, "loss": 2.6794, "step": 42444 }, { "epoch": 3.33, "learning_rate": 6.671763506625892e-05, "loss": 2.6699, "step": 42445 }, { "epoch": 3.33, "learning_rate": 6.671685093703442e-05, "loss": 2.5903, "step": 42446 }, { "epoch": 3.33, "learning_rate": 6.671606680780992e-05, "loss": 2.6158, "step": 42447 }, { "epoch": 3.33, "learning_rate": 6.671528267858543e-05, "loss": 2.8458, "step": 42448 }, { "epoch": 3.33, "learning_rate": 6.671449854936093e-05, "loss": 2.7837, "step": 42449 }, { "epoch": 3.33, "learning_rate": 6.671371442013644e-05, "loss": 2.6301, "step": 42450 }, { "epoch": 3.33, "learning_rate": 6.671293029091194e-05, "loss": 2.8949, "step": 42451 }, { "epoch": 3.33, "learning_rate": 6.671214616168746e-05, "loss": 2.899, "step": 42452 }, { "epoch": 3.33, "learning_rate": 6.671136203246296e-05, "loss": 3.009, "step": 42453 }, { "epoch": 3.33, "learning_rate": 6.671057790323846e-05, "loss": 2.6645, "step": 42454 }, { "epoch": 3.33, "learning_rate": 6.670979377401397e-05, "loss": 2.8195, "step": 42455 }, { "epoch": 3.33, "learning_rate": 6.670900964478947e-05, "loss": 2.8524, "step": 42456 }, { "epoch": 3.33, "learning_rate": 6.670822551556497e-05, "loss": 2.886, "step": 42457 }, { "epoch": 3.33, "learning_rate": 6.670744138634048e-05, "loss": 2.7356, "step": 42458 }, { "epoch": 3.33, "learning_rate": 6.670665725711598e-05, "loss": 2.739, "step": 42459 }, { "epoch": 3.33, "learning_rate": 6.670587312789149e-05, "loss": 2.748, "step": 42460 }, { "epoch": 3.33, "learning_rate": 6.670508899866699e-05, "loss": 2.9908, "step": 42461 }, { "epoch": 3.33, "learning_rate": 6.67043048694425e-05, "loss": 2.8739, "step": 42462 }, { "epoch": 3.33, "learning_rate": 6.6703520740218e-05, "loss": 2.7137, "step": 42463 }, { "epoch": 3.33, "learning_rate": 6.67027366109935e-05, "loss": 2.7003, "step": 42464 }, { "epoch": 3.33, "learning_rate": 6.6701952481769e-05, "loss": 2.569, "step": 42465 }, { "epoch": 3.33, "learning_rate": 6.670116835254451e-05, "loss": 2.8187, "step": 42466 }, { "epoch": 3.33, "learning_rate": 6.670038422332001e-05, "loss": 2.6028, "step": 42467 }, { "epoch": 3.33, "learning_rate": 6.66996000940955e-05, "loss": 2.7078, "step": 42468 }, { "epoch": 3.33, "learning_rate": 6.6698815964871e-05, "loss": 2.7916, "step": 42469 }, { "epoch": 3.33, "learning_rate": 6.669803183564651e-05, "loss": 2.7088, "step": 42470 }, { "epoch": 3.33, "learning_rate": 6.669724770642201e-05, "loss": 2.7187, "step": 42471 }, { "epoch": 3.33, "learning_rate": 6.669646357719752e-05, "loss": 2.6777, "step": 42472 }, { "epoch": 3.33, "learning_rate": 6.669567944797302e-05, "loss": 2.8346, "step": 42473 }, { "epoch": 3.33, "learning_rate": 6.669489531874852e-05, "loss": 2.8614, "step": 42474 }, { "epoch": 3.33, "learning_rate": 6.669411118952403e-05, "loss": 2.8747, "step": 42475 }, { "epoch": 3.33, "learning_rate": 6.669332706029953e-05, "loss": 2.7328, "step": 42476 }, { "epoch": 3.33, "learning_rate": 6.669254293107503e-05, "loss": 2.6061, "step": 42477 }, { "epoch": 3.33, "learning_rate": 6.669175880185055e-05, "loss": 2.7058, "step": 42478 }, { "epoch": 3.33, "learning_rate": 6.669097467262606e-05, "loss": 2.6895, "step": 42479 }, { "epoch": 3.33, "learning_rate": 6.669019054340156e-05, "loss": 2.835, "step": 42480 }, { "epoch": 3.33, "learning_rate": 6.668940641417706e-05, "loss": 2.7217, "step": 42481 }, { "epoch": 3.33, "learning_rate": 6.668862228495257e-05, "loss": 2.617, "step": 42482 }, { "epoch": 3.33, "learning_rate": 6.668783815572807e-05, "loss": 2.7063, "step": 42483 }, { "epoch": 3.33, "learning_rate": 6.668705402650357e-05, "loss": 2.9211, "step": 42484 }, { "epoch": 3.33, "learning_rate": 6.668626989727908e-05, "loss": 2.7043, "step": 42485 }, { "epoch": 3.33, "learning_rate": 6.668548576805458e-05, "loss": 2.7863, "step": 42486 }, { "epoch": 3.33, "learning_rate": 6.668470163883009e-05, "loss": 2.6458, "step": 42487 }, { "epoch": 3.33, "learning_rate": 6.668391750960559e-05, "loss": 2.7933, "step": 42488 }, { "epoch": 3.33, "learning_rate": 6.668313338038109e-05, "loss": 2.791, "step": 42489 }, { "epoch": 3.33, "learning_rate": 6.66823492511566e-05, "loss": 2.6276, "step": 42490 }, { "epoch": 3.33, "learning_rate": 6.66815651219321e-05, "loss": 2.7235, "step": 42491 }, { "epoch": 3.33, "learning_rate": 6.66807809927076e-05, "loss": 2.797, "step": 42492 }, { "epoch": 3.33, "learning_rate": 6.667999686348311e-05, "loss": 2.7245, "step": 42493 }, { "epoch": 3.33, "learning_rate": 6.667921273425861e-05, "loss": 2.5833, "step": 42494 }, { "epoch": 3.33, "learning_rate": 6.667842860503411e-05, "loss": 2.8086, "step": 42495 }, { "epoch": 3.33, "learning_rate": 6.667764447580962e-05, "loss": 2.6817, "step": 42496 }, { "epoch": 3.33, "learning_rate": 6.667686034658512e-05, "loss": 2.6082, "step": 42497 }, { "epoch": 3.33, "learning_rate": 6.667607621736063e-05, "loss": 2.6051, "step": 42498 }, { "epoch": 3.33, "learning_rate": 6.667529208813613e-05, "loss": 2.6691, "step": 42499 }, { "epoch": 3.33, "learning_rate": 6.667450795891163e-05, "loss": 2.8208, "step": 42500 }, { "epoch": 3.33, "learning_rate": 6.667372382968714e-05, "loss": 2.9157, "step": 42501 }, { "epoch": 3.33, "learning_rate": 6.667293970046264e-05, "loss": 2.7834, "step": 42502 }, { "epoch": 3.33, "learning_rate": 6.667215557123814e-05, "loss": 2.7723, "step": 42503 }, { "epoch": 3.33, "learning_rate": 6.667137144201365e-05, "loss": 2.9679, "step": 42504 }, { "epoch": 3.33, "learning_rate": 6.667058731278915e-05, "loss": 2.7494, "step": 42505 }, { "epoch": 3.33, "learning_rate": 6.666980318356466e-05, "loss": 2.7439, "step": 42506 }, { "epoch": 3.33, "learning_rate": 6.666901905434016e-05, "loss": 2.6441, "step": 42507 }, { "epoch": 3.33, "learning_rate": 6.666823492511566e-05, "loss": 2.8436, "step": 42508 }, { "epoch": 3.33, "learning_rate": 6.666745079589117e-05, "loss": 2.9343, "step": 42509 }, { "epoch": 3.33, "learning_rate": 6.666666666666667e-05, "loss": 2.7969, "step": 42510 }, { "epoch": 3.33, "learning_rate": 6.666588253744217e-05, "loss": 2.6928, "step": 42511 }, { "epoch": 3.33, "learning_rate": 6.666509840821768e-05, "loss": 2.5111, "step": 42512 }, { "epoch": 3.33, "learning_rate": 6.666431427899318e-05, "loss": 2.6177, "step": 42513 }, { "epoch": 3.33, "learning_rate": 6.666353014976868e-05, "loss": 2.641, "step": 42514 }, { "epoch": 3.33, "learning_rate": 6.666274602054419e-05, "loss": 2.6152, "step": 42515 }, { "epoch": 3.33, "learning_rate": 6.666196189131969e-05, "loss": 2.5837, "step": 42516 }, { "epoch": 3.33, "learning_rate": 6.66611777620952e-05, "loss": 2.6166, "step": 42517 }, { "epoch": 3.33, "learning_rate": 6.66603936328707e-05, "loss": 2.6887, "step": 42518 }, { "epoch": 3.33, "learning_rate": 6.66596095036462e-05, "loss": 2.7893, "step": 42519 }, { "epoch": 3.33, "learning_rate": 6.66588253744217e-05, "loss": 2.7643, "step": 42520 }, { "epoch": 3.33, "learning_rate": 6.665804124519721e-05, "loss": 2.5322, "step": 42521 }, { "epoch": 3.33, "learning_rate": 6.665725711597271e-05, "loss": 2.6366, "step": 42522 }, { "epoch": 3.33, "learning_rate": 6.665647298674822e-05, "loss": 2.9478, "step": 42523 }, { "epoch": 3.33, "learning_rate": 6.665568885752372e-05, "loss": 2.8129, "step": 42524 }, { "epoch": 3.33, "learning_rate": 6.665490472829923e-05, "loss": 2.7528, "step": 42525 }, { "epoch": 3.33, "learning_rate": 6.665412059907473e-05, "loss": 2.8973, "step": 42526 }, { "epoch": 3.33, "learning_rate": 6.665333646985023e-05, "loss": 2.7684, "step": 42527 }, { "epoch": 3.33, "learning_rate": 6.665255234062574e-05, "loss": 2.7206, "step": 42528 }, { "epoch": 3.33, "learning_rate": 6.665176821140124e-05, "loss": 2.7289, "step": 42529 }, { "epoch": 3.33, "learning_rate": 6.665098408217674e-05, "loss": 3.0491, "step": 42530 }, { "epoch": 3.33, "learning_rate": 6.665019995295225e-05, "loss": 2.6775, "step": 42531 }, { "epoch": 3.34, "learning_rate": 6.664941582372775e-05, "loss": 2.7863, "step": 42532 }, { "epoch": 3.34, "learning_rate": 6.664863169450325e-05, "loss": 2.6923, "step": 42533 }, { "epoch": 3.34, "learning_rate": 6.664784756527876e-05, "loss": 2.6623, "step": 42534 }, { "epoch": 3.34, "learning_rate": 6.664706343605428e-05, "loss": 2.8433, "step": 42535 }, { "epoch": 3.34, "learning_rate": 6.664627930682978e-05, "loss": 2.722, "step": 42536 }, { "epoch": 3.34, "learning_rate": 6.664549517760528e-05, "loss": 2.7789, "step": 42537 }, { "epoch": 3.34, "learning_rate": 6.664471104838079e-05, "loss": 2.4561, "step": 42538 }, { "epoch": 3.34, "learning_rate": 6.664392691915629e-05, "loss": 2.6901, "step": 42539 }, { "epoch": 3.34, "learning_rate": 6.66431427899318e-05, "loss": 2.6718, "step": 42540 }, { "epoch": 3.34, "learning_rate": 6.664235866070728e-05, "loss": 2.8169, "step": 42541 }, { "epoch": 3.34, "learning_rate": 6.664157453148279e-05, "loss": 2.5706, "step": 42542 }, { "epoch": 3.34, "learning_rate": 6.664079040225829e-05, "loss": 2.8924, "step": 42543 }, { "epoch": 3.34, "learning_rate": 6.66400062730338e-05, "loss": 2.813, "step": 42544 }, { "epoch": 3.34, "learning_rate": 6.66392221438093e-05, "loss": 2.6633, "step": 42545 }, { "epoch": 3.34, "learning_rate": 6.66384380145848e-05, "loss": 2.8378, "step": 42546 }, { "epoch": 3.34, "learning_rate": 6.66376538853603e-05, "loss": 2.5825, "step": 42547 }, { "epoch": 3.34, "learning_rate": 6.663686975613581e-05, "loss": 2.6324, "step": 42548 }, { "epoch": 3.34, "learning_rate": 6.663608562691131e-05, "loss": 2.6576, "step": 42549 }, { "epoch": 3.34, "learning_rate": 6.663530149768682e-05, "loss": 2.7372, "step": 42550 }, { "epoch": 3.34, "learning_rate": 6.663451736846232e-05, "loss": 2.5106, "step": 42551 }, { "epoch": 3.34, "learning_rate": 6.663373323923782e-05, "loss": 2.6459, "step": 42552 }, { "epoch": 3.34, "learning_rate": 6.663294911001333e-05, "loss": 2.8516, "step": 42553 }, { "epoch": 3.34, "learning_rate": 6.663216498078883e-05, "loss": 2.6198, "step": 42554 }, { "epoch": 3.34, "learning_rate": 6.663138085156434e-05, "loss": 2.708, "step": 42555 }, { "epoch": 3.34, "learning_rate": 6.663059672233984e-05, "loss": 2.8504, "step": 42556 }, { "epoch": 3.34, "learning_rate": 6.662981259311534e-05, "loss": 2.596, "step": 42557 }, { "epoch": 3.34, "learning_rate": 6.662902846389085e-05, "loss": 2.7774, "step": 42558 }, { "epoch": 3.34, "learning_rate": 6.662824433466635e-05, "loss": 2.5673, "step": 42559 }, { "epoch": 3.34, "learning_rate": 6.662746020544187e-05, "loss": 2.7424, "step": 42560 }, { "epoch": 3.34, "learning_rate": 6.662667607621737e-05, "loss": 2.5346, "step": 42561 }, { "epoch": 3.34, "learning_rate": 6.662589194699287e-05, "loss": 2.6009, "step": 42562 }, { "epoch": 3.34, "learning_rate": 6.662510781776838e-05, "loss": 2.8688, "step": 42563 }, { "epoch": 3.34, "learning_rate": 6.662432368854388e-05, "loss": 2.8949, "step": 42564 }, { "epoch": 3.34, "learning_rate": 6.662353955931939e-05, "loss": 2.6477, "step": 42565 }, { "epoch": 3.34, "learning_rate": 6.662275543009489e-05, "loss": 2.7001, "step": 42566 }, { "epoch": 3.34, "learning_rate": 6.662197130087039e-05, "loss": 2.6484, "step": 42567 }, { "epoch": 3.34, "learning_rate": 6.66211871716459e-05, "loss": 2.6078, "step": 42568 }, { "epoch": 3.34, "learning_rate": 6.66204030424214e-05, "loss": 2.7357, "step": 42569 }, { "epoch": 3.34, "learning_rate": 6.66196189131969e-05, "loss": 2.7529, "step": 42570 }, { "epoch": 3.34, "learning_rate": 6.661883478397241e-05, "loss": 2.7229, "step": 42571 }, { "epoch": 3.34, "learning_rate": 6.661805065474791e-05, "loss": 2.7865, "step": 42572 }, { "epoch": 3.34, "learning_rate": 6.661726652552342e-05, "loss": 2.8829, "step": 42573 }, { "epoch": 3.34, "learning_rate": 6.661648239629892e-05, "loss": 2.9612, "step": 42574 }, { "epoch": 3.34, "learning_rate": 6.661569826707442e-05, "loss": 2.6821, "step": 42575 }, { "epoch": 3.34, "learning_rate": 6.661491413784993e-05, "loss": 2.731, "step": 42576 }, { "epoch": 3.34, "learning_rate": 6.661413000862542e-05, "loss": 2.7176, "step": 42577 }, { "epoch": 3.34, "learning_rate": 6.661334587940092e-05, "loss": 2.698, "step": 42578 }, { "epoch": 3.34, "learning_rate": 6.661256175017642e-05, "loss": 2.9364, "step": 42579 }, { "epoch": 3.34, "learning_rate": 6.661177762095193e-05, "loss": 2.5876, "step": 42580 }, { "epoch": 3.34, "learning_rate": 6.661099349172743e-05, "loss": 2.813, "step": 42581 }, { "epoch": 3.34, "learning_rate": 6.661020936250293e-05, "loss": 2.4614, "step": 42582 }, { "epoch": 3.34, "learning_rate": 6.660942523327844e-05, "loss": 2.983, "step": 42583 }, { "epoch": 3.34, "learning_rate": 6.660864110405394e-05, "loss": 2.7128, "step": 42584 }, { "epoch": 3.34, "learning_rate": 6.660785697482945e-05, "loss": 2.7548, "step": 42585 }, { "epoch": 3.34, "learning_rate": 6.660707284560496e-05, "loss": 2.8073, "step": 42586 }, { "epoch": 3.34, "learning_rate": 6.660628871638047e-05, "loss": 2.659, "step": 42587 }, { "epoch": 3.34, "learning_rate": 6.660550458715597e-05, "loss": 2.8509, "step": 42588 }, { "epoch": 3.34, "learning_rate": 6.660472045793147e-05, "loss": 2.8887, "step": 42589 }, { "epoch": 3.34, "learning_rate": 6.660393632870698e-05, "loss": 2.6639, "step": 42590 }, { "epoch": 3.34, "learning_rate": 6.660315219948248e-05, "loss": 2.6804, "step": 42591 }, { "epoch": 3.34, "learning_rate": 6.660236807025798e-05, "loss": 2.6293, "step": 42592 }, { "epoch": 3.34, "learning_rate": 6.660158394103349e-05, "loss": 2.6859, "step": 42593 }, { "epoch": 3.34, "learning_rate": 6.660079981180899e-05, "loss": 2.6878, "step": 42594 }, { "epoch": 3.34, "learning_rate": 6.66000156825845e-05, "loss": 2.7752, "step": 42595 }, { "epoch": 3.34, "learning_rate": 6.659923155336e-05, "loss": 2.6998, "step": 42596 }, { "epoch": 3.34, "learning_rate": 6.65984474241355e-05, "loss": 2.9734, "step": 42597 }, { "epoch": 3.34, "learning_rate": 6.659766329491101e-05, "loss": 2.459, "step": 42598 }, { "epoch": 3.34, "learning_rate": 6.659687916568651e-05, "loss": 2.7093, "step": 42599 }, { "epoch": 3.34, "learning_rate": 6.659609503646201e-05, "loss": 2.6028, "step": 42600 }, { "epoch": 3.34, "learning_rate": 6.659531090723752e-05, "loss": 3.0344, "step": 42601 }, { "epoch": 3.34, "learning_rate": 6.659452677801302e-05, "loss": 2.5917, "step": 42602 }, { "epoch": 3.34, "learning_rate": 6.659374264878853e-05, "loss": 2.7075, "step": 42603 }, { "epoch": 3.34, "learning_rate": 6.659295851956403e-05, "loss": 2.7937, "step": 42604 }, { "epoch": 3.34, "learning_rate": 6.659217439033953e-05, "loss": 2.7484, "step": 42605 }, { "epoch": 3.34, "learning_rate": 6.659139026111504e-05, "loss": 2.5187, "step": 42606 }, { "epoch": 3.34, "learning_rate": 6.659060613189054e-05, "loss": 2.6944, "step": 42607 }, { "epoch": 3.34, "learning_rate": 6.658982200266604e-05, "loss": 2.7304, "step": 42608 }, { "epoch": 3.34, "learning_rate": 6.658903787344155e-05, "loss": 2.8791, "step": 42609 }, { "epoch": 3.34, "learning_rate": 6.658825374421705e-05, "loss": 2.6526, "step": 42610 }, { "epoch": 3.34, "learning_rate": 6.658746961499255e-05, "loss": 2.5459, "step": 42611 }, { "epoch": 3.34, "learning_rate": 6.658668548576806e-05, "loss": 2.6733, "step": 42612 }, { "epoch": 3.34, "learning_rate": 6.658590135654356e-05, "loss": 2.8107, "step": 42613 }, { "epoch": 3.34, "learning_rate": 6.658511722731907e-05, "loss": 2.7994, "step": 42614 }, { "epoch": 3.34, "learning_rate": 6.658433309809457e-05, "loss": 2.9113, "step": 42615 }, { "epoch": 3.34, "learning_rate": 6.658354896887007e-05, "loss": 2.7634, "step": 42616 }, { "epoch": 3.34, "learning_rate": 6.658276483964558e-05, "loss": 2.7174, "step": 42617 }, { "epoch": 3.34, "learning_rate": 6.658198071042108e-05, "loss": 2.8291, "step": 42618 }, { "epoch": 3.34, "learning_rate": 6.658119658119658e-05, "loss": 2.6453, "step": 42619 }, { "epoch": 3.34, "learning_rate": 6.658041245197209e-05, "loss": 2.788, "step": 42620 }, { "epoch": 3.34, "learning_rate": 6.657962832274759e-05, "loss": 2.7005, "step": 42621 }, { "epoch": 3.34, "learning_rate": 6.65788441935231e-05, "loss": 2.6404, "step": 42622 }, { "epoch": 3.34, "learning_rate": 6.65780600642986e-05, "loss": 2.9241, "step": 42623 }, { "epoch": 3.34, "learning_rate": 6.65772759350741e-05, "loss": 2.5342, "step": 42624 }, { "epoch": 3.34, "learning_rate": 6.65764918058496e-05, "loss": 2.7547, "step": 42625 }, { "epoch": 3.34, "learning_rate": 6.657570767662511e-05, "loss": 2.812, "step": 42626 }, { "epoch": 3.34, "learning_rate": 6.657492354740061e-05, "loss": 2.9153, "step": 42627 }, { "epoch": 3.34, "learning_rate": 6.657413941817612e-05, "loss": 2.6885, "step": 42628 }, { "epoch": 3.34, "learning_rate": 6.657335528895162e-05, "loss": 2.799, "step": 42629 }, { "epoch": 3.34, "learning_rate": 6.657257115972712e-05, "loss": 2.838, "step": 42630 }, { "epoch": 3.34, "learning_rate": 6.657178703050263e-05, "loss": 2.616, "step": 42631 }, { "epoch": 3.34, "learning_rate": 6.657100290127813e-05, "loss": 2.7023, "step": 42632 }, { "epoch": 3.34, "learning_rate": 6.657021877205364e-05, "loss": 2.9812, "step": 42633 }, { "epoch": 3.34, "learning_rate": 6.656943464282914e-05, "loss": 2.61, "step": 42634 }, { "epoch": 3.34, "learning_rate": 6.656865051360464e-05, "loss": 2.7885, "step": 42635 }, { "epoch": 3.34, "learning_rate": 6.656786638438015e-05, "loss": 2.9772, "step": 42636 }, { "epoch": 3.34, "learning_rate": 6.656708225515565e-05, "loss": 3.027, "step": 42637 }, { "epoch": 3.34, "learning_rate": 6.656629812593115e-05, "loss": 2.5744, "step": 42638 }, { "epoch": 3.34, "learning_rate": 6.656551399670666e-05, "loss": 2.6704, "step": 42639 }, { "epoch": 3.34, "learning_rate": 6.656472986748216e-05, "loss": 2.9734, "step": 42640 }, { "epoch": 3.34, "learning_rate": 6.656394573825767e-05, "loss": 2.7442, "step": 42641 }, { "epoch": 3.34, "learning_rate": 6.656316160903317e-05, "loss": 2.6004, "step": 42642 }, { "epoch": 3.34, "learning_rate": 6.656237747980869e-05, "loss": 2.5099, "step": 42643 }, { "epoch": 3.34, "learning_rate": 6.656159335058419e-05, "loss": 2.5387, "step": 42644 }, { "epoch": 3.34, "learning_rate": 6.65608092213597e-05, "loss": 2.8036, "step": 42645 }, { "epoch": 3.34, "learning_rate": 6.65600250921352e-05, "loss": 2.8511, "step": 42646 }, { "epoch": 3.34, "learning_rate": 6.65592409629107e-05, "loss": 2.7729, "step": 42647 }, { "epoch": 3.34, "learning_rate": 6.65584568336862e-05, "loss": 2.9205, "step": 42648 }, { "epoch": 3.34, "learning_rate": 6.65576727044617e-05, "loss": 2.7122, "step": 42649 }, { "epoch": 3.34, "learning_rate": 6.65568885752372e-05, "loss": 2.7779, "step": 42650 }, { "epoch": 3.34, "learning_rate": 6.65561044460127e-05, "loss": 2.7313, "step": 42651 }, { "epoch": 3.34, "learning_rate": 6.65553203167882e-05, "loss": 2.6873, "step": 42652 }, { "epoch": 3.34, "learning_rate": 6.655453618756371e-05, "loss": 2.7352, "step": 42653 }, { "epoch": 3.34, "learning_rate": 6.655375205833921e-05, "loss": 2.8146, "step": 42654 }, { "epoch": 3.34, "learning_rate": 6.655296792911472e-05, "loss": 2.8548, "step": 42655 }, { "epoch": 3.34, "learning_rate": 6.655218379989022e-05, "loss": 2.7039, "step": 42656 }, { "epoch": 3.34, "learning_rate": 6.655139967066572e-05, "loss": 2.8465, "step": 42657 }, { "epoch": 3.34, "learning_rate": 6.655061554144123e-05, "loss": 2.861, "step": 42658 }, { "epoch": 3.35, "learning_rate": 6.654983141221673e-05, "loss": 2.8655, "step": 42659 }, { "epoch": 3.35, "learning_rate": 6.654904728299224e-05, "loss": 2.7207, "step": 42660 }, { "epoch": 3.35, "learning_rate": 6.654826315376774e-05, "loss": 2.6814, "step": 42661 }, { "epoch": 3.35, "learning_rate": 6.654747902454324e-05, "loss": 2.7868, "step": 42662 }, { "epoch": 3.35, "learning_rate": 6.654669489531875e-05, "loss": 2.6133, "step": 42663 }, { "epoch": 3.35, "learning_rate": 6.654591076609425e-05, "loss": 2.5838, "step": 42664 }, { "epoch": 3.35, "learning_rate": 6.654512663686975e-05, "loss": 2.6485, "step": 42665 }, { "epoch": 3.35, "learning_rate": 6.654434250764526e-05, "loss": 2.6148, "step": 42666 }, { "epoch": 3.35, "learning_rate": 6.654355837842076e-05, "loss": 2.63, "step": 42667 }, { "epoch": 3.35, "learning_rate": 6.654277424919628e-05, "loss": 2.5036, "step": 42668 }, { "epoch": 3.35, "learning_rate": 6.654199011997178e-05, "loss": 2.5258, "step": 42669 }, { "epoch": 3.35, "learning_rate": 6.654120599074729e-05, "loss": 2.5934, "step": 42670 }, { "epoch": 3.35, "learning_rate": 6.654042186152279e-05, "loss": 2.8085, "step": 42671 }, { "epoch": 3.35, "learning_rate": 6.653963773229829e-05, "loss": 2.7173, "step": 42672 }, { "epoch": 3.35, "learning_rate": 6.65388536030738e-05, "loss": 2.6479, "step": 42673 }, { "epoch": 3.35, "learning_rate": 6.65380694738493e-05, "loss": 2.7046, "step": 42674 }, { "epoch": 3.35, "learning_rate": 6.65372853446248e-05, "loss": 2.8008, "step": 42675 }, { "epoch": 3.35, "learning_rate": 6.653650121540031e-05, "loss": 2.5875, "step": 42676 }, { "epoch": 3.35, "learning_rate": 6.653571708617581e-05, "loss": 2.6552, "step": 42677 }, { "epoch": 3.35, "learning_rate": 6.653493295695131e-05, "loss": 2.7127, "step": 42678 }, { "epoch": 3.35, "learning_rate": 6.653414882772682e-05, "loss": 2.5622, "step": 42679 }, { "epoch": 3.35, "learning_rate": 6.653336469850232e-05, "loss": 2.9893, "step": 42680 }, { "epoch": 3.35, "learning_rate": 6.653258056927783e-05, "loss": 2.7714, "step": 42681 }, { "epoch": 3.35, "learning_rate": 6.653179644005333e-05, "loss": 2.4635, "step": 42682 }, { "epoch": 3.35, "learning_rate": 6.653101231082883e-05, "loss": 2.7575, "step": 42683 }, { "epoch": 3.35, "learning_rate": 6.653022818160434e-05, "loss": 2.6753, "step": 42684 }, { "epoch": 3.35, "learning_rate": 6.652944405237983e-05, "loss": 2.6562, "step": 42685 }, { "epoch": 3.35, "learning_rate": 6.652865992315533e-05, "loss": 2.8958, "step": 42686 }, { "epoch": 3.35, "learning_rate": 6.652787579393083e-05, "loss": 2.803, "step": 42687 }, { "epoch": 3.35, "learning_rate": 6.652709166470634e-05, "loss": 3.0784, "step": 42688 }, { "epoch": 3.35, "learning_rate": 6.652630753548184e-05, "loss": 2.8327, "step": 42689 }, { "epoch": 3.35, "learning_rate": 6.652552340625735e-05, "loss": 2.9178, "step": 42690 }, { "epoch": 3.35, "learning_rate": 6.652473927703285e-05, "loss": 2.8367, "step": 42691 }, { "epoch": 3.35, "learning_rate": 6.652395514780835e-05, "loss": 2.8172, "step": 42692 }, { "epoch": 3.35, "learning_rate": 6.652317101858386e-05, "loss": 2.6452, "step": 42693 }, { "epoch": 3.35, "learning_rate": 6.652238688935937e-05, "loss": 2.6869, "step": 42694 }, { "epoch": 3.35, "learning_rate": 6.652160276013488e-05, "loss": 2.7395, "step": 42695 }, { "epoch": 3.35, "learning_rate": 6.652081863091038e-05, "loss": 2.6955, "step": 42696 }, { "epoch": 3.35, "learning_rate": 6.652003450168588e-05, "loss": 2.9507, "step": 42697 }, { "epoch": 3.35, "learning_rate": 6.651925037246139e-05, "loss": 2.745, "step": 42698 }, { "epoch": 3.35, "learning_rate": 6.651846624323689e-05, "loss": 2.8362, "step": 42699 }, { "epoch": 3.35, "learning_rate": 6.65176821140124e-05, "loss": 2.6708, "step": 42700 }, { "epoch": 3.35, "learning_rate": 6.65168979847879e-05, "loss": 2.6228, "step": 42701 }, { "epoch": 3.35, "learning_rate": 6.65161138555634e-05, "loss": 2.7498, "step": 42702 }, { "epoch": 3.35, "learning_rate": 6.65153297263389e-05, "loss": 2.9537, "step": 42703 }, { "epoch": 3.35, "learning_rate": 6.651454559711441e-05, "loss": 2.777, "step": 42704 }, { "epoch": 3.35, "learning_rate": 6.651376146788991e-05, "loss": 2.7468, "step": 42705 }, { "epoch": 3.35, "learning_rate": 6.651297733866542e-05, "loss": 2.8519, "step": 42706 }, { "epoch": 3.35, "learning_rate": 6.651219320944092e-05, "loss": 2.667, "step": 42707 }, { "epoch": 3.35, "learning_rate": 6.651140908021643e-05, "loss": 2.8428, "step": 42708 }, { "epoch": 3.35, "learning_rate": 6.651062495099193e-05, "loss": 2.7557, "step": 42709 }, { "epoch": 3.35, "learning_rate": 6.650984082176743e-05, "loss": 2.7763, "step": 42710 }, { "epoch": 3.35, "learning_rate": 6.650905669254294e-05, "loss": 2.7143, "step": 42711 }, { "epoch": 3.35, "learning_rate": 6.650827256331844e-05, "loss": 2.8127, "step": 42712 }, { "epoch": 3.35, "learning_rate": 6.650748843409394e-05, "loss": 2.928, "step": 42713 }, { "epoch": 3.35, "learning_rate": 6.650670430486945e-05, "loss": 2.844, "step": 42714 }, { "epoch": 3.35, "learning_rate": 6.650592017564495e-05, "loss": 2.7027, "step": 42715 }, { "epoch": 3.35, "learning_rate": 6.650513604642045e-05, "loss": 2.8516, "step": 42716 }, { "epoch": 3.35, "learning_rate": 6.650435191719596e-05, "loss": 2.9761, "step": 42717 }, { "epoch": 3.35, "learning_rate": 6.650356778797146e-05, "loss": 2.7411, "step": 42718 }, { "epoch": 3.35, "learning_rate": 6.650278365874697e-05, "loss": 2.6145, "step": 42719 }, { "epoch": 3.35, "learning_rate": 6.650199952952247e-05, "loss": 2.7395, "step": 42720 }, { "epoch": 3.35, "learning_rate": 6.650121540029797e-05, "loss": 2.8042, "step": 42721 }, { "epoch": 3.35, "learning_rate": 6.650043127107348e-05, "loss": 2.8558, "step": 42722 }, { "epoch": 3.35, "learning_rate": 6.649964714184898e-05, "loss": 2.6966, "step": 42723 }, { "epoch": 3.35, "learning_rate": 6.649886301262448e-05, "loss": 2.6224, "step": 42724 }, { "epoch": 3.35, "learning_rate": 6.649807888339999e-05, "loss": 2.607, "step": 42725 }, { "epoch": 3.35, "learning_rate": 6.649729475417549e-05, "loss": 2.7171, "step": 42726 }, { "epoch": 3.35, "learning_rate": 6.6496510624951e-05, "loss": 2.7955, "step": 42727 }, { "epoch": 3.35, "learning_rate": 6.64957264957265e-05, "loss": 2.6964, "step": 42728 }, { "epoch": 3.35, "learning_rate": 6.6494942366502e-05, "loss": 2.701, "step": 42729 }, { "epoch": 3.35, "learning_rate": 6.64941582372775e-05, "loss": 2.65, "step": 42730 }, { "epoch": 3.35, "learning_rate": 6.649337410805301e-05, "loss": 2.7278, "step": 42731 }, { "epoch": 3.35, "learning_rate": 6.649258997882851e-05, "loss": 2.6661, "step": 42732 }, { "epoch": 3.35, "learning_rate": 6.649180584960402e-05, "loss": 2.5454, "step": 42733 }, { "epoch": 3.35, "learning_rate": 6.649102172037952e-05, "loss": 2.5713, "step": 42734 }, { "epoch": 3.35, "learning_rate": 6.649023759115502e-05, "loss": 3.1158, "step": 42735 }, { "epoch": 3.35, "learning_rate": 6.648945346193053e-05, "loss": 2.6797, "step": 42736 }, { "epoch": 3.35, "learning_rate": 6.648866933270603e-05, "loss": 2.7568, "step": 42737 }, { "epoch": 3.35, "learning_rate": 6.648788520348154e-05, "loss": 2.5169, "step": 42738 }, { "epoch": 3.35, "learning_rate": 6.648710107425704e-05, "loss": 2.7154, "step": 42739 }, { "epoch": 3.35, "learning_rate": 6.648631694503254e-05, "loss": 2.6088, "step": 42740 }, { "epoch": 3.35, "learning_rate": 6.648553281580805e-05, "loss": 2.6291, "step": 42741 }, { "epoch": 3.35, "learning_rate": 6.648474868658355e-05, "loss": 2.6973, "step": 42742 }, { "epoch": 3.35, "learning_rate": 6.648396455735905e-05, "loss": 2.655, "step": 42743 }, { "epoch": 3.35, "learning_rate": 6.648318042813456e-05, "loss": 2.7646, "step": 42744 }, { "epoch": 3.35, "learning_rate": 6.648239629891006e-05, "loss": 2.8958, "step": 42745 }, { "epoch": 3.35, "learning_rate": 6.648161216968556e-05, "loss": 2.6598, "step": 42746 }, { "epoch": 3.35, "learning_rate": 6.648082804046107e-05, "loss": 2.7054, "step": 42747 }, { "epoch": 3.35, "learning_rate": 6.648004391123657e-05, "loss": 2.8321, "step": 42748 }, { "epoch": 3.35, "learning_rate": 6.647925978201208e-05, "loss": 2.6614, "step": 42749 }, { "epoch": 3.35, "learning_rate": 6.647847565278758e-05, "loss": 2.7892, "step": 42750 }, { "epoch": 3.35, "learning_rate": 6.64776915235631e-05, "loss": 2.5592, "step": 42751 }, { "epoch": 3.35, "learning_rate": 6.64769073943386e-05, "loss": 2.6699, "step": 42752 }, { "epoch": 3.35, "learning_rate": 6.64761232651141e-05, "loss": 2.6574, "step": 42753 }, { "epoch": 3.35, "learning_rate": 6.647533913588961e-05, "loss": 2.8407, "step": 42754 }, { "epoch": 3.35, "learning_rate": 6.647455500666511e-05, "loss": 2.6937, "step": 42755 }, { "epoch": 3.35, "learning_rate": 6.647377087744062e-05, "loss": 2.5777, "step": 42756 }, { "epoch": 3.35, "learning_rate": 6.64729867482161e-05, "loss": 2.7116, "step": 42757 }, { "epoch": 3.35, "learning_rate": 6.647220261899161e-05, "loss": 2.686, "step": 42758 }, { "epoch": 3.35, "learning_rate": 6.647141848976711e-05, "loss": 2.6132, "step": 42759 }, { "epoch": 3.35, "learning_rate": 6.647063436054262e-05, "loss": 2.7948, "step": 42760 }, { "epoch": 3.35, "learning_rate": 6.646985023131812e-05, "loss": 2.812, "step": 42761 }, { "epoch": 3.35, "learning_rate": 6.646906610209362e-05, "loss": 2.7746, "step": 42762 }, { "epoch": 3.35, "learning_rate": 6.646828197286913e-05, "loss": 2.7032, "step": 42763 }, { "epoch": 3.35, "learning_rate": 6.646749784364463e-05, "loss": 2.7824, "step": 42764 }, { "epoch": 3.35, "learning_rate": 6.646671371442013e-05, "loss": 2.7921, "step": 42765 }, { "epoch": 3.35, "learning_rate": 6.646592958519564e-05, "loss": 2.8973, "step": 42766 }, { "epoch": 3.35, "learning_rate": 6.646514545597114e-05, "loss": 2.7024, "step": 42767 }, { "epoch": 3.35, "learning_rate": 6.646436132674665e-05, "loss": 2.7215, "step": 42768 }, { "epoch": 3.35, "learning_rate": 6.646357719752215e-05, "loss": 2.6633, "step": 42769 }, { "epoch": 3.35, "learning_rate": 6.646279306829765e-05, "loss": 2.6152, "step": 42770 }, { "epoch": 3.35, "learning_rate": 6.646200893907316e-05, "loss": 2.7116, "step": 42771 }, { "epoch": 3.35, "learning_rate": 6.646122480984866e-05, "loss": 2.7401, "step": 42772 }, { "epoch": 3.35, "learning_rate": 6.646044068062416e-05, "loss": 2.6548, "step": 42773 }, { "epoch": 3.35, "learning_rate": 6.645965655139967e-05, "loss": 2.7903, "step": 42774 }, { "epoch": 3.35, "learning_rate": 6.645887242217517e-05, "loss": 2.6818, "step": 42775 }, { "epoch": 3.35, "learning_rate": 6.645808829295069e-05, "loss": 2.6253, "step": 42776 }, { "epoch": 3.35, "learning_rate": 6.645730416372619e-05, "loss": 2.8681, "step": 42777 }, { "epoch": 3.35, "learning_rate": 6.64565200345017e-05, "loss": 2.6458, "step": 42778 }, { "epoch": 3.35, "learning_rate": 6.64557359052772e-05, "loss": 2.8251, "step": 42779 }, { "epoch": 3.35, "learning_rate": 6.64549517760527e-05, "loss": 2.5051, "step": 42780 }, { "epoch": 3.35, "learning_rate": 6.645416764682821e-05, "loss": 2.5888, "step": 42781 }, { "epoch": 3.35, "learning_rate": 6.645338351760371e-05, "loss": 2.7646, "step": 42782 }, { "epoch": 3.35, "learning_rate": 6.645259938837921e-05, "loss": 2.686, "step": 42783 }, { "epoch": 3.35, "learning_rate": 6.645181525915472e-05, "loss": 2.7183, "step": 42784 }, { "epoch": 3.35, "learning_rate": 6.645103112993022e-05, "loss": 2.9657, "step": 42785 }, { "epoch": 3.35, "learning_rate": 6.645024700070573e-05, "loss": 2.5728, "step": 42786 }, { "epoch": 3.36, "learning_rate": 6.644946287148123e-05, "loss": 2.4789, "step": 42787 }, { "epoch": 3.36, "learning_rate": 6.644867874225673e-05, "loss": 2.8436, "step": 42788 }, { "epoch": 3.36, "learning_rate": 6.644789461303224e-05, "loss": 2.8238, "step": 42789 }, { "epoch": 3.36, "learning_rate": 6.644711048380774e-05, "loss": 2.7294, "step": 42790 }, { "epoch": 3.36, "learning_rate": 6.644632635458324e-05, "loss": 2.8777, "step": 42791 }, { "epoch": 3.36, "learning_rate": 6.644554222535875e-05, "loss": 2.6638, "step": 42792 }, { "epoch": 3.36, "learning_rate": 6.644475809613424e-05, "loss": 2.8121, "step": 42793 }, { "epoch": 3.36, "learning_rate": 6.644397396690974e-05, "loss": 2.6858, "step": 42794 }, { "epoch": 3.36, "learning_rate": 6.644318983768525e-05, "loss": 2.6383, "step": 42795 }, { "epoch": 3.36, "learning_rate": 6.644240570846075e-05, "loss": 2.5891, "step": 42796 }, { "epoch": 3.36, "learning_rate": 6.644162157923625e-05, "loss": 2.7394, "step": 42797 }, { "epoch": 3.36, "learning_rate": 6.644083745001176e-05, "loss": 2.7341, "step": 42798 }, { "epoch": 3.36, "learning_rate": 6.644005332078726e-05, "loss": 2.8233, "step": 42799 }, { "epoch": 3.36, "learning_rate": 6.643926919156276e-05, "loss": 2.87, "step": 42800 }, { "epoch": 3.36, "learning_rate": 6.643848506233827e-05, "loss": 2.7878, "step": 42801 }, { "epoch": 3.36, "learning_rate": 6.643770093311378e-05, "loss": 2.6278, "step": 42802 }, { "epoch": 3.36, "learning_rate": 6.643691680388929e-05, "loss": 2.7341, "step": 42803 }, { "epoch": 3.36, "learning_rate": 6.643613267466479e-05, "loss": 2.6986, "step": 42804 }, { "epoch": 3.36, "learning_rate": 6.64353485454403e-05, "loss": 2.7106, "step": 42805 }, { "epoch": 3.36, "learning_rate": 6.64345644162158e-05, "loss": 2.4783, "step": 42806 }, { "epoch": 3.36, "learning_rate": 6.64337802869913e-05, "loss": 2.637, "step": 42807 }, { "epoch": 3.36, "learning_rate": 6.64329961577668e-05, "loss": 2.6429, "step": 42808 }, { "epoch": 3.36, "learning_rate": 6.643221202854231e-05, "loss": 2.6334, "step": 42809 }, { "epoch": 3.36, "learning_rate": 6.643142789931781e-05, "loss": 2.6964, "step": 42810 }, { "epoch": 3.36, "learning_rate": 6.643064377009332e-05, "loss": 2.7613, "step": 42811 }, { "epoch": 3.36, "learning_rate": 6.642985964086882e-05, "loss": 2.7368, "step": 42812 }, { "epoch": 3.36, "learning_rate": 6.642907551164432e-05, "loss": 2.7331, "step": 42813 }, { "epoch": 3.36, "learning_rate": 6.642829138241983e-05, "loss": 2.5295, "step": 42814 }, { "epoch": 3.36, "learning_rate": 6.642750725319533e-05, "loss": 2.7349, "step": 42815 }, { "epoch": 3.36, "learning_rate": 6.642672312397084e-05, "loss": 2.71, "step": 42816 }, { "epoch": 3.36, "learning_rate": 6.642593899474634e-05, "loss": 2.5496, "step": 42817 }, { "epoch": 3.36, "learning_rate": 6.642515486552184e-05, "loss": 2.7028, "step": 42818 }, { "epoch": 3.36, "learning_rate": 6.642437073629735e-05, "loss": 2.7556, "step": 42819 }, { "epoch": 3.36, "learning_rate": 6.642358660707285e-05, "loss": 2.6309, "step": 42820 }, { "epoch": 3.36, "learning_rate": 6.642280247784835e-05, "loss": 2.6319, "step": 42821 }, { "epoch": 3.36, "learning_rate": 6.642201834862386e-05, "loss": 2.5568, "step": 42822 }, { "epoch": 3.36, "learning_rate": 6.642123421939936e-05, "loss": 2.5745, "step": 42823 }, { "epoch": 3.36, "learning_rate": 6.642045009017487e-05, "loss": 2.6371, "step": 42824 }, { "epoch": 3.36, "learning_rate": 6.641966596095037e-05, "loss": 2.7578, "step": 42825 }, { "epoch": 3.36, "learning_rate": 6.641888183172587e-05, "loss": 2.6373, "step": 42826 }, { "epoch": 3.36, "learning_rate": 6.641809770250138e-05, "loss": 2.5482, "step": 42827 }, { "epoch": 3.36, "learning_rate": 6.641731357327688e-05, "loss": 2.5605, "step": 42828 }, { "epoch": 3.36, "learning_rate": 6.641652944405238e-05, "loss": 2.6119, "step": 42829 }, { "epoch": 3.36, "learning_rate": 6.641574531482789e-05, "loss": 2.8241, "step": 42830 }, { "epoch": 3.36, "learning_rate": 6.641496118560339e-05, "loss": 2.7294, "step": 42831 }, { "epoch": 3.36, "learning_rate": 6.64141770563789e-05, "loss": 2.5443, "step": 42832 }, { "epoch": 3.36, "learning_rate": 6.64133929271544e-05, "loss": 2.9247, "step": 42833 }, { "epoch": 3.36, "learning_rate": 6.64126087979299e-05, "loss": 2.5811, "step": 42834 }, { "epoch": 3.36, "learning_rate": 6.64118246687054e-05, "loss": 2.8139, "step": 42835 }, { "epoch": 3.36, "learning_rate": 6.641104053948091e-05, "loss": 2.4828, "step": 42836 }, { "epoch": 3.36, "learning_rate": 6.641025641025641e-05, "loss": 2.603, "step": 42837 }, { "epoch": 3.36, "learning_rate": 6.640947228103192e-05, "loss": 2.9008, "step": 42838 }, { "epoch": 3.36, "learning_rate": 6.640868815180742e-05, "loss": 2.7352, "step": 42839 }, { "epoch": 3.36, "learning_rate": 6.640790402258292e-05, "loss": 2.8821, "step": 42840 }, { "epoch": 3.36, "learning_rate": 6.640711989335843e-05, "loss": 2.7761, "step": 42841 }, { "epoch": 3.36, "learning_rate": 6.640633576413393e-05, "loss": 2.608, "step": 42842 }, { "epoch": 3.36, "learning_rate": 6.640555163490944e-05, "loss": 2.7485, "step": 42843 }, { "epoch": 3.36, "learning_rate": 6.640476750568494e-05, "loss": 2.7152, "step": 42844 }, { "epoch": 3.36, "learning_rate": 6.640398337646044e-05, "loss": 2.6733, "step": 42845 }, { "epoch": 3.36, "learning_rate": 6.640319924723595e-05, "loss": 2.8359, "step": 42846 }, { "epoch": 3.36, "learning_rate": 6.640241511801145e-05, "loss": 2.9397, "step": 42847 }, { "epoch": 3.36, "learning_rate": 6.640163098878695e-05, "loss": 2.8856, "step": 42848 }, { "epoch": 3.36, "learning_rate": 6.640084685956246e-05, "loss": 2.8276, "step": 42849 }, { "epoch": 3.36, "learning_rate": 6.640006273033796e-05, "loss": 2.8308, "step": 42850 }, { "epoch": 3.36, "learning_rate": 6.639927860111346e-05, "loss": 2.8915, "step": 42851 }, { "epoch": 3.36, "learning_rate": 6.639849447188897e-05, "loss": 2.8331, "step": 42852 }, { "epoch": 3.36, "learning_rate": 6.639771034266447e-05, "loss": 2.831, "step": 42853 }, { "epoch": 3.36, "learning_rate": 6.639692621343998e-05, "loss": 2.5932, "step": 42854 }, { "epoch": 3.36, "learning_rate": 6.639614208421548e-05, "loss": 3.0548, "step": 42855 }, { "epoch": 3.36, "learning_rate": 6.639535795499098e-05, "loss": 2.621, "step": 42856 }, { "epoch": 3.36, "learning_rate": 6.639457382576649e-05, "loss": 2.8498, "step": 42857 }, { "epoch": 3.36, "learning_rate": 6.639378969654199e-05, "loss": 2.8483, "step": 42858 }, { "epoch": 3.36, "learning_rate": 6.639300556731751e-05, "loss": 2.6972, "step": 42859 }, { "epoch": 3.36, "learning_rate": 6.639222143809301e-05, "loss": 2.524, "step": 42860 }, { "epoch": 3.36, "learning_rate": 6.639143730886851e-05, "loss": 2.826, "step": 42861 }, { "epoch": 3.36, "learning_rate": 6.639065317964402e-05, "loss": 2.7632, "step": 42862 }, { "epoch": 3.36, "learning_rate": 6.638986905041952e-05, "loss": 2.7375, "step": 42863 }, { "epoch": 3.36, "learning_rate": 6.638908492119503e-05, "loss": 2.7284, "step": 42864 }, { "epoch": 3.36, "learning_rate": 6.638830079197052e-05, "loss": 2.7056, "step": 42865 }, { "epoch": 3.36, "learning_rate": 6.638751666274602e-05, "loss": 2.6766, "step": 42866 }, { "epoch": 3.36, "learning_rate": 6.638673253352152e-05, "loss": 2.9204, "step": 42867 }, { "epoch": 3.36, "learning_rate": 6.638594840429703e-05, "loss": 2.9177, "step": 42868 }, { "epoch": 3.36, "learning_rate": 6.638516427507253e-05, "loss": 2.707, "step": 42869 }, { "epoch": 3.36, "learning_rate": 6.638438014584803e-05, "loss": 2.3954, "step": 42870 }, { "epoch": 3.36, "learning_rate": 6.638359601662354e-05, "loss": 2.5445, "step": 42871 }, { "epoch": 3.36, "learning_rate": 6.638281188739904e-05, "loss": 2.8255, "step": 42872 }, { "epoch": 3.36, "learning_rate": 6.638202775817455e-05, "loss": 2.7163, "step": 42873 }, { "epoch": 3.36, "learning_rate": 6.638124362895005e-05, "loss": 2.924, "step": 42874 }, { "epoch": 3.36, "learning_rate": 6.638045949972555e-05, "loss": 2.8421, "step": 42875 }, { "epoch": 3.36, "learning_rate": 6.637967537050106e-05, "loss": 2.7112, "step": 42876 }, { "epoch": 3.36, "learning_rate": 6.637889124127656e-05, "loss": 2.6104, "step": 42877 }, { "epoch": 3.36, "learning_rate": 6.637810711205206e-05, "loss": 2.7045, "step": 42878 }, { "epoch": 3.36, "learning_rate": 6.637732298282757e-05, "loss": 2.6783, "step": 42879 }, { "epoch": 3.36, "learning_rate": 6.637653885360307e-05, "loss": 2.8304, "step": 42880 }, { "epoch": 3.36, "learning_rate": 6.637575472437857e-05, "loss": 2.6771, "step": 42881 }, { "epoch": 3.36, "learning_rate": 6.637497059515408e-05, "loss": 2.6364, "step": 42882 }, { "epoch": 3.36, "learning_rate": 6.637418646592958e-05, "loss": 2.5934, "step": 42883 }, { "epoch": 3.36, "learning_rate": 6.63734023367051e-05, "loss": 2.8502, "step": 42884 }, { "epoch": 3.36, "learning_rate": 6.63726182074806e-05, "loss": 2.6833, "step": 42885 }, { "epoch": 3.36, "learning_rate": 6.637183407825611e-05, "loss": 2.674, "step": 42886 }, { "epoch": 3.36, "learning_rate": 6.637104994903161e-05, "loss": 2.6786, "step": 42887 }, { "epoch": 3.36, "learning_rate": 6.637026581980711e-05, "loss": 2.7158, "step": 42888 }, { "epoch": 3.36, "learning_rate": 6.636948169058262e-05, "loss": 2.8365, "step": 42889 }, { "epoch": 3.36, "learning_rate": 6.636869756135812e-05, "loss": 2.6189, "step": 42890 }, { "epoch": 3.36, "learning_rate": 6.636791343213363e-05, "loss": 2.6702, "step": 42891 }, { "epoch": 3.36, "learning_rate": 6.636712930290913e-05, "loss": 2.69, "step": 42892 }, { "epoch": 3.36, "learning_rate": 6.636634517368463e-05, "loss": 2.9033, "step": 42893 }, { "epoch": 3.36, "learning_rate": 6.636556104446014e-05, "loss": 2.9865, "step": 42894 }, { "epoch": 3.36, "learning_rate": 6.636477691523564e-05, "loss": 2.7098, "step": 42895 }, { "epoch": 3.36, "learning_rate": 6.636399278601114e-05, "loss": 2.8035, "step": 42896 }, { "epoch": 3.36, "learning_rate": 6.636320865678665e-05, "loss": 2.6105, "step": 42897 }, { "epoch": 3.36, "learning_rate": 6.636242452756215e-05, "loss": 2.6869, "step": 42898 }, { "epoch": 3.36, "learning_rate": 6.636164039833765e-05, "loss": 2.7342, "step": 42899 }, { "epoch": 3.36, "learning_rate": 6.636085626911316e-05, "loss": 2.8626, "step": 42900 }, { "epoch": 3.36, "learning_rate": 6.636007213988865e-05, "loss": 2.7593, "step": 42901 }, { "epoch": 3.36, "learning_rate": 6.635928801066415e-05, "loss": 2.7217, "step": 42902 }, { "epoch": 3.36, "learning_rate": 6.635850388143966e-05, "loss": 2.8282, "step": 42903 }, { "epoch": 3.36, "learning_rate": 6.635771975221516e-05, "loss": 2.8515, "step": 42904 }, { "epoch": 3.36, "learning_rate": 6.635693562299066e-05, "loss": 2.7213, "step": 42905 }, { "epoch": 3.36, "learning_rate": 6.635615149376617e-05, "loss": 2.8141, "step": 42906 }, { "epoch": 3.36, "learning_rate": 6.635536736454167e-05, "loss": 2.8092, "step": 42907 }, { "epoch": 3.36, "learning_rate": 6.635458323531717e-05, "loss": 2.6406, "step": 42908 }, { "epoch": 3.36, "learning_rate": 6.635379910609268e-05, "loss": 2.6948, "step": 42909 }, { "epoch": 3.36, "learning_rate": 6.63530149768682e-05, "loss": 2.979, "step": 42910 }, { "epoch": 3.36, "learning_rate": 6.63522308476437e-05, "loss": 2.9461, "step": 42911 }, { "epoch": 3.36, "learning_rate": 6.63514467184192e-05, "loss": 2.727, "step": 42912 }, { "epoch": 3.36, "learning_rate": 6.63506625891947e-05, "loss": 2.7075, "step": 42913 }, { "epoch": 3.37, "learning_rate": 6.634987845997021e-05, "loss": 2.6638, "step": 42914 }, { "epoch": 3.37, "learning_rate": 6.634909433074571e-05, "loss": 2.9354, "step": 42915 }, { "epoch": 3.37, "learning_rate": 6.634831020152122e-05, "loss": 2.6493, "step": 42916 }, { "epoch": 3.37, "learning_rate": 6.634752607229672e-05, "loss": 2.6884, "step": 42917 }, { "epoch": 3.37, "learning_rate": 6.634674194307222e-05, "loss": 2.7134, "step": 42918 }, { "epoch": 3.37, "learning_rate": 6.634595781384773e-05, "loss": 2.8357, "step": 42919 }, { "epoch": 3.37, "learning_rate": 6.634517368462323e-05, "loss": 2.6905, "step": 42920 }, { "epoch": 3.37, "learning_rate": 6.634438955539874e-05, "loss": 2.8874, "step": 42921 }, { "epoch": 3.37, "learning_rate": 6.634360542617424e-05, "loss": 2.7995, "step": 42922 }, { "epoch": 3.37, "learning_rate": 6.634282129694974e-05, "loss": 2.5807, "step": 42923 }, { "epoch": 3.37, "learning_rate": 6.634203716772525e-05, "loss": 2.8464, "step": 42924 }, { "epoch": 3.37, "learning_rate": 6.634125303850075e-05, "loss": 2.7366, "step": 42925 }, { "epoch": 3.37, "learning_rate": 6.634046890927625e-05, "loss": 2.6695, "step": 42926 }, { "epoch": 3.37, "learning_rate": 6.633968478005176e-05, "loss": 2.8345, "step": 42927 }, { "epoch": 3.37, "learning_rate": 6.633890065082726e-05, "loss": 2.614, "step": 42928 }, { "epoch": 3.37, "learning_rate": 6.633811652160277e-05, "loss": 2.8091, "step": 42929 }, { "epoch": 3.37, "learning_rate": 6.633733239237827e-05, "loss": 2.6507, "step": 42930 }, { "epoch": 3.37, "learning_rate": 6.633654826315377e-05, "loss": 2.6055, "step": 42931 }, { "epoch": 3.37, "learning_rate": 6.633576413392928e-05, "loss": 2.7353, "step": 42932 }, { "epoch": 3.37, "learning_rate": 6.633498000470478e-05, "loss": 2.8647, "step": 42933 }, { "epoch": 3.37, "learning_rate": 6.633419587548028e-05, "loss": 2.7455, "step": 42934 }, { "epoch": 3.37, "learning_rate": 6.633341174625579e-05, "loss": 2.7374, "step": 42935 }, { "epoch": 3.37, "learning_rate": 6.633262761703129e-05, "loss": 2.6259, "step": 42936 }, { "epoch": 3.37, "learning_rate": 6.63318434878068e-05, "loss": 2.8227, "step": 42937 }, { "epoch": 3.37, "learning_rate": 6.63310593585823e-05, "loss": 2.8323, "step": 42938 }, { "epoch": 3.37, "learning_rate": 6.63302752293578e-05, "loss": 2.6275, "step": 42939 }, { "epoch": 3.37, "learning_rate": 6.63294911001333e-05, "loss": 2.6033, "step": 42940 }, { "epoch": 3.37, "learning_rate": 6.632870697090881e-05, "loss": 2.66, "step": 42941 }, { "epoch": 3.37, "learning_rate": 6.632792284168431e-05, "loss": 2.9517, "step": 42942 }, { "epoch": 3.37, "learning_rate": 6.632713871245982e-05, "loss": 2.7403, "step": 42943 }, { "epoch": 3.37, "learning_rate": 6.632635458323532e-05, "loss": 2.7446, "step": 42944 }, { "epoch": 3.37, "learning_rate": 6.632557045401082e-05, "loss": 2.7146, "step": 42945 }, { "epoch": 3.37, "learning_rate": 6.632478632478633e-05, "loss": 2.7417, "step": 42946 }, { "epoch": 3.37, "learning_rate": 6.632400219556183e-05, "loss": 2.7704, "step": 42947 }, { "epoch": 3.37, "learning_rate": 6.632321806633733e-05, "loss": 2.7462, "step": 42948 }, { "epoch": 3.37, "learning_rate": 6.632243393711284e-05, "loss": 2.7422, "step": 42949 }, { "epoch": 3.37, "learning_rate": 6.632164980788834e-05, "loss": 2.6389, "step": 42950 }, { "epoch": 3.37, "learning_rate": 6.632086567866385e-05, "loss": 2.7278, "step": 42951 }, { "epoch": 3.37, "learning_rate": 6.632008154943935e-05, "loss": 2.7987, "step": 42952 }, { "epoch": 3.37, "learning_rate": 6.631929742021485e-05, "loss": 2.7063, "step": 42953 }, { "epoch": 3.37, "learning_rate": 6.631851329099036e-05, "loss": 2.6839, "step": 42954 }, { "epoch": 3.37, "learning_rate": 6.631772916176586e-05, "loss": 2.6597, "step": 42955 }, { "epoch": 3.37, "learning_rate": 6.631694503254136e-05, "loss": 2.6337, "step": 42956 }, { "epoch": 3.37, "learning_rate": 6.631616090331687e-05, "loss": 2.6125, "step": 42957 }, { "epoch": 3.37, "learning_rate": 6.631537677409237e-05, "loss": 2.762, "step": 42958 }, { "epoch": 3.37, "learning_rate": 6.631459264486788e-05, "loss": 2.6746, "step": 42959 }, { "epoch": 3.37, "learning_rate": 6.631380851564338e-05, "loss": 2.9033, "step": 42960 }, { "epoch": 3.37, "learning_rate": 6.631302438641888e-05, "loss": 2.7633, "step": 42961 }, { "epoch": 3.37, "learning_rate": 6.631224025719439e-05, "loss": 2.6757, "step": 42962 }, { "epoch": 3.37, "learning_rate": 6.631145612796989e-05, "loss": 2.8274, "step": 42963 }, { "epoch": 3.37, "learning_rate": 6.63106719987454e-05, "loss": 2.9088, "step": 42964 }, { "epoch": 3.37, "learning_rate": 6.63098878695209e-05, "loss": 2.799, "step": 42965 }, { "epoch": 3.37, "learning_rate": 6.63091037402964e-05, "loss": 2.6683, "step": 42966 }, { "epoch": 3.37, "learning_rate": 6.630831961107192e-05, "loss": 2.7897, "step": 42967 }, { "epoch": 3.37, "learning_rate": 6.630753548184742e-05, "loss": 2.6448, "step": 42968 }, { "epoch": 3.37, "learning_rate": 6.630675135262293e-05, "loss": 2.8662, "step": 42969 }, { "epoch": 3.37, "learning_rate": 6.630596722339843e-05, "loss": 2.757, "step": 42970 }, { "epoch": 3.37, "learning_rate": 6.630518309417393e-05, "loss": 2.8047, "step": 42971 }, { "epoch": 3.37, "learning_rate": 6.630439896494944e-05, "loss": 2.9841, "step": 42972 }, { "epoch": 3.37, "learning_rate": 6.630361483572493e-05, "loss": 2.6728, "step": 42973 }, { "epoch": 3.37, "learning_rate": 6.630283070650043e-05, "loss": 2.7404, "step": 42974 }, { "epoch": 3.37, "learning_rate": 6.630204657727593e-05, "loss": 3.0538, "step": 42975 }, { "epoch": 3.37, "learning_rate": 6.630126244805144e-05, "loss": 2.7953, "step": 42976 }, { "epoch": 3.37, "learning_rate": 6.630047831882694e-05, "loss": 2.7616, "step": 42977 }, { "epoch": 3.37, "learning_rate": 6.629969418960245e-05, "loss": 2.6639, "step": 42978 }, { "epoch": 3.37, "learning_rate": 6.629891006037795e-05, "loss": 2.6641, "step": 42979 }, { "epoch": 3.37, "learning_rate": 6.629812593115345e-05, "loss": 2.9375, "step": 42980 }, { "epoch": 3.37, "learning_rate": 6.629734180192896e-05, "loss": 2.8031, "step": 42981 }, { "epoch": 3.37, "learning_rate": 6.629655767270446e-05, "loss": 2.7348, "step": 42982 }, { "epoch": 3.37, "learning_rate": 6.629577354347996e-05, "loss": 2.7039, "step": 42983 }, { "epoch": 3.37, "learning_rate": 6.629498941425547e-05, "loss": 2.651, "step": 42984 }, { "epoch": 3.37, "learning_rate": 6.629420528503097e-05, "loss": 2.7362, "step": 42985 }, { "epoch": 3.37, "learning_rate": 6.629342115580647e-05, "loss": 2.6304, "step": 42986 }, { "epoch": 3.37, "learning_rate": 6.629263702658198e-05, "loss": 2.7498, "step": 42987 }, { "epoch": 3.37, "learning_rate": 6.629185289735748e-05, "loss": 2.632, "step": 42988 }, { "epoch": 3.37, "learning_rate": 6.629106876813299e-05, "loss": 2.6363, "step": 42989 }, { "epoch": 3.37, "learning_rate": 6.629028463890849e-05, "loss": 2.6461, "step": 42990 }, { "epoch": 3.37, "learning_rate": 6.628950050968399e-05, "loss": 2.6104, "step": 42991 }, { "epoch": 3.37, "learning_rate": 6.62887163804595e-05, "loss": 2.7761, "step": 42992 }, { "epoch": 3.37, "learning_rate": 6.628793225123501e-05, "loss": 3.0892, "step": 42993 }, { "epoch": 3.37, "learning_rate": 6.628714812201052e-05, "loss": 2.9639, "step": 42994 }, { "epoch": 3.37, "learning_rate": 6.628636399278602e-05, "loss": 2.6738, "step": 42995 }, { "epoch": 3.37, "learning_rate": 6.628557986356152e-05, "loss": 2.6373, "step": 42996 }, { "epoch": 3.37, "learning_rate": 6.628479573433703e-05, "loss": 2.761, "step": 42997 }, { "epoch": 3.37, "learning_rate": 6.628401160511253e-05, "loss": 2.6999, "step": 42998 }, { "epoch": 3.37, "learning_rate": 6.628322747588804e-05, "loss": 2.8836, "step": 42999 }, { "epoch": 3.37, "learning_rate": 6.628244334666354e-05, "loss": 2.6941, "step": 43000 }, { "epoch": 3.37, "eval_loss": 2.8968517780303955, "eval_runtime": 217.8957, "eval_samples_per_second": 52.007, "eval_steps_per_second": 3.254, "step": 43000 }, { "epoch": 3.37, "learning_rate": 6.628165921743904e-05, "loss": 2.7849, "step": 43001 }, { "epoch": 3.37, "learning_rate": 6.628087508821455e-05, "loss": 2.5361, "step": 43002 }, { "epoch": 3.37, "learning_rate": 6.628009095899005e-05, "loss": 2.7011, "step": 43003 }, { "epoch": 3.37, "learning_rate": 6.627930682976555e-05, "loss": 2.6747, "step": 43004 }, { "epoch": 3.37, "learning_rate": 6.627852270054106e-05, "loss": 2.7378, "step": 43005 }, { "epoch": 3.37, "learning_rate": 6.627773857131656e-05, "loss": 2.6589, "step": 43006 }, { "epoch": 3.37, "learning_rate": 6.627695444209207e-05, "loss": 2.4649, "step": 43007 }, { "epoch": 3.37, "learning_rate": 6.627617031286757e-05, "loss": 2.6256, "step": 43008 }, { "epoch": 3.37, "learning_rate": 6.627538618364306e-05, "loss": 2.6536, "step": 43009 }, { "epoch": 3.37, "learning_rate": 6.627460205441856e-05, "loss": 2.8515, "step": 43010 }, { "epoch": 3.37, "learning_rate": 6.627381792519407e-05, "loss": 2.7025, "step": 43011 }, { "epoch": 3.37, "learning_rate": 6.627303379596957e-05, "loss": 2.8144, "step": 43012 }, { "epoch": 3.37, "learning_rate": 6.627224966674507e-05, "loss": 2.8746, "step": 43013 }, { "epoch": 3.37, "learning_rate": 6.627146553752058e-05, "loss": 2.9433, "step": 43014 }, { "epoch": 3.37, "learning_rate": 6.627068140829608e-05, "loss": 2.633, "step": 43015 }, { "epoch": 3.37, "learning_rate": 6.626989727907158e-05, "loss": 2.5338, "step": 43016 }, { "epoch": 3.37, "learning_rate": 6.626911314984709e-05, "loss": 2.7903, "step": 43017 }, { "epoch": 3.37, "learning_rate": 6.62683290206226e-05, "loss": 2.9036, "step": 43018 }, { "epoch": 3.37, "learning_rate": 6.626754489139811e-05, "loss": 2.6021, "step": 43019 }, { "epoch": 3.37, "learning_rate": 6.626676076217361e-05, "loss": 2.6397, "step": 43020 }, { "epoch": 3.37, "learning_rate": 6.626597663294912e-05, "loss": 2.6962, "step": 43021 }, { "epoch": 3.37, "learning_rate": 6.626519250372462e-05, "loss": 2.7071, "step": 43022 }, { "epoch": 3.37, "learning_rate": 6.626440837450012e-05, "loss": 2.6154, "step": 43023 }, { "epoch": 3.37, "learning_rate": 6.626362424527563e-05, "loss": 2.8847, "step": 43024 }, { "epoch": 3.37, "learning_rate": 6.626284011605113e-05, "loss": 2.7859, "step": 43025 }, { "epoch": 3.37, "learning_rate": 6.626205598682664e-05, "loss": 2.6048, "step": 43026 }, { "epoch": 3.37, "learning_rate": 6.626127185760214e-05, "loss": 2.7972, "step": 43027 }, { "epoch": 3.37, "learning_rate": 6.626048772837764e-05, "loss": 2.7589, "step": 43028 }, { "epoch": 3.37, "learning_rate": 6.625970359915315e-05, "loss": 2.8126, "step": 43029 }, { "epoch": 3.37, "learning_rate": 6.625891946992865e-05, "loss": 2.7359, "step": 43030 }, { "epoch": 3.37, "learning_rate": 6.625813534070415e-05, "loss": 2.8312, "step": 43031 }, { "epoch": 3.37, "learning_rate": 6.625735121147966e-05, "loss": 3.1019, "step": 43032 }, { "epoch": 3.37, "learning_rate": 6.625656708225516e-05, "loss": 2.6652, "step": 43033 }, { "epoch": 3.37, "learning_rate": 6.625578295303066e-05, "loss": 2.7402, "step": 43034 }, { "epoch": 3.37, "learning_rate": 6.625499882380617e-05, "loss": 2.4954, "step": 43035 }, { "epoch": 3.37, "learning_rate": 6.625421469458167e-05, "loss": 2.6998, "step": 43036 }, { "epoch": 3.37, "learning_rate": 6.625343056535718e-05, "loss": 2.9651, "step": 43037 }, { "epoch": 3.37, "learning_rate": 6.625264643613268e-05, "loss": 2.6592, "step": 43038 }, { "epoch": 3.37, "learning_rate": 6.625186230690818e-05, "loss": 2.6036, "step": 43039 }, { "epoch": 3.37, "learning_rate": 6.625107817768369e-05, "loss": 2.765, "step": 43040 }, { "epoch": 3.37, "learning_rate": 6.625029404845919e-05, "loss": 2.8512, "step": 43041 }, { "epoch": 3.38, "learning_rate": 6.62495099192347e-05, "loss": 2.8756, "step": 43042 }, { "epoch": 3.38, "learning_rate": 6.62487257900102e-05, "loss": 2.7286, "step": 43043 }, { "epoch": 3.38, "learning_rate": 6.62479416607857e-05, "loss": 3.0526, "step": 43044 }, { "epoch": 3.38, "learning_rate": 6.62471575315612e-05, "loss": 2.9124, "step": 43045 }, { "epoch": 3.38, "learning_rate": 6.624637340233671e-05, "loss": 2.7173, "step": 43046 }, { "epoch": 3.38, "learning_rate": 6.624558927311221e-05, "loss": 2.7529, "step": 43047 }, { "epoch": 3.38, "learning_rate": 6.624480514388772e-05, "loss": 2.6796, "step": 43048 }, { "epoch": 3.38, "learning_rate": 6.624402101466322e-05, "loss": 2.5434, "step": 43049 }, { "epoch": 3.38, "learning_rate": 6.624323688543872e-05, "loss": 2.4937, "step": 43050 }, { "epoch": 3.38, "learning_rate": 6.624245275621423e-05, "loss": 2.7775, "step": 43051 }, { "epoch": 3.38, "learning_rate": 6.624166862698973e-05, "loss": 2.8557, "step": 43052 }, { "epoch": 3.38, "learning_rate": 6.624088449776523e-05, "loss": 2.5303, "step": 43053 }, { "epoch": 3.38, "learning_rate": 6.624010036854074e-05, "loss": 2.7974, "step": 43054 }, { "epoch": 3.38, "learning_rate": 6.623931623931624e-05, "loss": 2.6972, "step": 43055 }, { "epoch": 3.38, "learning_rate": 6.623853211009175e-05, "loss": 2.6811, "step": 43056 }, { "epoch": 3.38, "learning_rate": 6.623774798086725e-05, "loss": 2.9622, "step": 43057 }, { "epoch": 3.38, "learning_rate": 6.623696385164275e-05, "loss": 2.6779, "step": 43058 }, { "epoch": 3.38, "learning_rate": 6.623617972241826e-05, "loss": 2.786, "step": 43059 }, { "epoch": 3.38, "learning_rate": 6.623539559319376e-05, "loss": 2.785, "step": 43060 }, { "epoch": 3.38, "learning_rate": 6.623461146396926e-05, "loss": 2.7111, "step": 43061 }, { "epoch": 3.38, "learning_rate": 6.623382733474477e-05, "loss": 2.7961, "step": 43062 }, { "epoch": 3.38, "learning_rate": 6.623304320552027e-05, "loss": 2.8996, "step": 43063 }, { "epoch": 3.38, "learning_rate": 6.623225907629578e-05, "loss": 2.8194, "step": 43064 }, { "epoch": 3.38, "learning_rate": 6.623147494707128e-05, "loss": 2.6809, "step": 43065 }, { "epoch": 3.38, "learning_rate": 6.623069081784678e-05, "loss": 2.609, "step": 43066 }, { "epoch": 3.38, "learning_rate": 6.622990668862229e-05, "loss": 2.721, "step": 43067 }, { "epoch": 3.38, "learning_rate": 6.622912255939779e-05, "loss": 2.8671, "step": 43068 }, { "epoch": 3.38, "learning_rate": 6.62283384301733e-05, "loss": 2.4703, "step": 43069 }, { "epoch": 3.38, "learning_rate": 6.62275543009488e-05, "loss": 2.619, "step": 43070 }, { "epoch": 3.38, "learning_rate": 6.62267701717243e-05, "loss": 2.596, "step": 43071 }, { "epoch": 3.38, "learning_rate": 6.62259860424998e-05, "loss": 2.487, "step": 43072 }, { "epoch": 3.38, "learning_rate": 6.622520191327531e-05, "loss": 3.044, "step": 43073 }, { "epoch": 3.38, "learning_rate": 6.622441778405081e-05, "loss": 2.6669, "step": 43074 }, { "epoch": 3.38, "learning_rate": 6.622363365482633e-05, "loss": 2.6704, "step": 43075 }, { "epoch": 3.38, "learning_rate": 6.622284952560183e-05, "loss": 2.69, "step": 43076 }, { "epoch": 3.38, "learning_rate": 6.622206539637734e-05, "loss": 2.6645, "step": 43077 }, { "epoch": 3.38, "learning_rate": 6.622128126715284e-05, "loss": 2.8019, "step": 43078 }, { "epoch": 3.38, "learning_rate": 6.622049713792834e-05, "loss": 2.9308, "step": 43079 }, { "epoch": 3.38, "learning_rate": 6.621971300870385e-05, "loss": 2.793, "step": 43080 }, { "epoch": 3.38, "learning_rate": 6.621892887947934e-05, "loss": 3.015, "step": 43081 }, { "epoch": 3.38, "learning_rate": 6.621814475025484e-05, "loss": 2.7205, "step": 43082 }, { "epoch": 3.38, "learning_rate": 6.621736062103034e-05, "loss": 2.6548, "step": 43083 }, { "epoch": 3.38, "learning_rate": 6.621657649180585e-05, "loss": 2.7295, "step": 43084 }, { "epoch": 3.38, "learning_rate": 6.621579236258135e-05, "loss": 2.5748, "step": 43085 }, { "epoch": 3.38, "learning_rate": 6.621500823335686e-05, "loss": 2.6263, "step": 43086 }, { "epoch": 3.38, "learning_rate": 6.621422410413236e-05, "loss": 2.6361, "step": 43087 }, { "epoch": 3.38, "learning_rate": 6.621343997490786e-05, "loss": 2.6373, "step": 43088 }, { "epoch": 3.38, "learning_rate": 6.621265584568337e-05, "loss": 2.5092, "step": 43089 }, { "epoch": 3.38, "learning_rate": 6.621187171645887e-05, "loss": 2.6085, "step": 43090 }, { "epoch": 3.38, "learning_rate": 6.621108758723437e-05, "loss": 2.8721, "step": 43091 }, { "epoch": 3.38, "learning_rate": 6.621030345800988e-05, "loss": 2.6833, "step": 43092 }, { "epoch": 3.38, "learning_rate": 6.620951932878538e-05, "loss": 2.9792, "step": 43093 }, { "epoch": 3.38, "learning_rate": 6.620873519956089e-05, "loss": 3.0376, "step": 43094 }, { "epoch": 3.38, "learning_rate": 6.620795107033639e-05, "loss": 2.7544, "step": 43095 }, { "epoch": 3.38, "learning_rate": 6.620716694111189e-05, "loss": 2.7608, "step": 43096 }, { "epoch": 3.38, "learning_rate": 6.62063828118874e-05, "loss": 2.6692, "step": 43097 }, { "epoch": 3.38, "learning_rate": 6.62055986826629e-05, "loss": 2.7346, "step": 43098 }, { "epoch": 3.38, "learning_rate": 6.62048145534384e-05, "loss": 2.796, "step": 43099 }, { "epoch": 3.38, "learning_rate": 6.620403042421391e-05, "loss": 2.8337, "step": 43100 }, { "epoch": 3.38, "learning_rate": 6.620324629498942e-05, "loss": 2.6961, "step": 43101 }, { "epoch": 3.38, "learning_rate": 6.620246216576493e-05, "loss": 2.6765, "step": 43102 }, { "epoch": 3.38, "learning_rate": 6.620167803654043e-05, "loss": 2.6869, "step": 43103 }, { "epoch": 3.38, "learning_rate": 6.620089390731594e-05, "loss": 2.4819, "step": 43104 }, { "epoch": 3.38, "learning_rate": 6.620010977809144e-05, "loss": 2.7209, "step": 43105 }, { "epoch": 3.38, "learning_rate": 6.619932564886694e-05, "loss": 2.7557, "step": 43106 }, { "epoch": 3.38, "learning_rate": 6.619854151964245e-05, "loss": 2.8393, "step": 43107 }, { "epoch": 3.38, "learning_rate": 6.619775739041795e-05, "loss": 2.5178, "step": 43108 }, { "epoch": 3.38, "learning_rate": 6.619697326119345e-05, "loss": 2.941, "step": 43109 }, { "epoch": 3.38, "learning_rate": 6.619618913196896e-05, "loss": 2.7083, "step": 43110 }, { "epoch": 3.38, "learning_rate": 6.619540500274446e-05, "loss": 2.6924, "step": 43111 }, { "epoch": 3.38, "learning_rate": 6.619462087351997e-05, "loss": 2.5773, "step": 43112 }, { "epoch": 3.38, "learning_rate": 6.619383674429547e-05, "loss": 2.6116, "step": 43113 }, { "epoch": 3.38, "learning_rate": 6.619305261507097e-05, "loss": 2.693, "step": 43114 }, { "epoch": 3.38, "learning_rate": 6.619226848584648e-05, "loss": 2.6005, "step": 43115 }, { "epoch": 3.38, "learning_rate": 6.619148435662198e-05, "loss": 2.8584, "step": 43116 }, { "epoch": 3.38, "learning_rate": 6.619070022739747e-05, "loss": 2.9, "step": 43117 }, { "epoch": 3.38, "learning_rate": 6.618991609817297e-05, "loss": 2.7676, "step": 43118 }, { "epoch": 3.38, "learning_rate": 6.618913196894848e-05, "loss": 2.8033, "step": 43119 }, { "epoch": 3.38, "learning_rate": 6.618834783972398e-05, "loss": 2.7622, "step": 43120 }, { "epoch": 3.38, "learning_rate": 6.618756371049948e-05, "loss": 2.8363, "step": 43121 }, { "epoch": 3.38, "learning_rate": 6.618677958127499e-05, "loss": 2.6583, "step": 43122 }, { "epoch": 3.38, "learning_rate": 6.618599545205049e-05, "loss": 2.6689, "step": 43123 }, { "epoch": 3.38, "learning_rate": 6.6185211322826e-05, "loss": 2.8577, "step": 43124 }, { "epoch": 3.38, "learning_rate": 6.61844271936015e-05, "loss": 2.8018, "step": 43125 }, { "epoch": 3.38, "learning_rate": 6.618364306437702e-05, "loss": 2.8624, "step": 43126 }, { "epoch": 3.38, "learning_rate": 6.618285893515252e-05, "loss": 2.7667, "step": 43127 }, { "epoch": 3.38, "learning_rate": 6.618207480592802e-05, "loss": 2.4675, "step": 43128 }, { "epoch": 3.38, "learning_rate": 6.618129067670353e-05, "loss": 2.6618, "step": 43129 }, { "epoch": 3.38, "learning_rate": 6.618050654747903e-05, "loss": 2.9098, "step": 43130 }, { "epoch": 3.38, "learning_rate": 6.617972241825454e-05, "loss": 2.8396, "step": 43131 }, { "epoch": 3.38, "learning_rate": 6.617893828903004e-05, "loss": 2.7418, "step": 43132 }, { "epoch": 3.38, "learning_rate": 6.617815415980554e-05, "loss": 2.6618, "step": 43133 }, { "epoch": 3.38, "learning_rate": 6.617737003058105e-05, "loss": 2.797, "step": 43134 }, { "epoch": 3.38, "learning_rate": 6.617658590135655e-05, "loss": 2.6983, "step": 43135 }, { "epoch": 3.38, "learning_rate": 6.617580177213205e-05, "loss": 2.7009, "step": 43136 }, { "epoch": 3.38, "learning_rate": 6.617501764290756e-05, "loss": 2.5923, "step": 43137 }, { "epoch": 3.38, "learning_rate": 6.617423351368306e-05, "loss": 2.7064, "step": 43138 }, { "epoch": 3.38, "learning_rate": 6.617344938445856e-05, "loss": 2.5642, "step": 43139 }, { "epoch": 3.38, "learning_rate": 6.617266525523407e-05, "loss": 2.7415, "step": 43140 }, { "epoch": 3.38, "learning_rate": 6.617188112600957e-05, "loss": 2.417, "step": 43141 }, { "epoch": 3.38, "learning_rate": 6.617109699678508e-05, "loss": 2.7707, "step": 43142 }, { "epoch": 3.38, "learning_rate": 6.617031286756058e-05, "loss": 2.8036, "step": 43143 }, { "epoch": 3.38, "learning_rate": 6.616952873833608e-05, "loss": 2.6123, "step": 43144 }, { "epoch": 3.38, "learning_rate": 6.616874460911159e-05, "loss": 2.6772, "step": 43145 }, { "epoch": 3.38, "learning_rate": 6.616796047988709e-05, "loss": 2.6803, "step": 43146 }, { "epoch": 3.38, "learning_rate": 6.61671763506626e-05, "loss": 2.9093, "step": 43147 }, { "epoch": 3.38, "learning_rate": 6.61663922214381e-05, "loss": 2.7152, "step": 43148 }, { "epoch": 3.38, "learning_rate": 6.61656080922136e-05, "loss": 2.6484, "step": 43149 }, { "epoch": 3.38, "learning_rate": 6.61648239629891e-05, "loss": 2.6326, "step": 43150 }, { "epoch": 3.38, "learning_rate": 6.616403983376461e-05, "loss": 2.6985, "step": 43151 }, { "epoch": 3.38, "learning_rate": 6.616325570454011e-05, "loss": 2.7508, "step": 43152 }, { "epoch": 3.38, "learning_rate": 6.616247157531562e-05, "loss": 2.4774, "step": 43153 }, { "epoch": 3.38, "learning_rate": 6.616168744609112e-05, "loss": 2.7483, "step": 43154 }, { "epoch": 3.38, "learning_rate": 6.616090331686662e-05, "loss": 2.7464, "step": 43155 }, { "epoch": 3.38, "learning_rate": 6.616011918764213e-05, "loss": 2.5782, "step": 43156 }, { "epoch": 3.38, "learning_rate": 6.615933505841763e-05, "loss": 2.8164, "step": 43157 }, { "epoch": 3.38, "learning_rate": 6.615855092919313e-05, "loss": 2.6112, "step": 43158 }, { "epoch": 3.38, "learning_rate": 6.615776679996864e-05, "loss": 2.7967, "step": 43159 }, { "epoch": 3.38, "learning_rate": 6.615698267074414e-05, "loss": 2.6968, "step": 43160 }, { "epoch": 3.38, "learning_rate": 6.615619854151965e-05, "loss": 3.0195, "step": 43161 }, { "epoch": 3.38, "learning_rate": 6.615541441229515e-05, "loss": 2.5106, "step": 43162 }, { "epoch": 3.38, "learning_rate": 6.615463028307065e-05, "loss": 2.4701, "step": 43163 }, { "epoch": 3.38, "learning_rate": 6.615384615384616e-05, "loss": 2.7734, "step": 43164 }, { "epoch": 3.38, "learning_rate": 6.615306202462166e-05, "loss": 2.631, "step": 43165 }, { "epoch": 3.38, "learning_rate": 6.615227789539716e-05, "loss": 2.783, "step": 43166 }, { "epoch": 3.38, "learning_rate": 6.615149376617267e-05, "loss": 2.7529, "step": 43167 }, { "epoch": 3.38, "learning_rate": 6.615070963694817e-05, "loss": 2.5475, "step": 43168 }, { "epoch": 3.39, "learning_rate": 6.614992550772367e-05, "loss": 2.7035, "step": 43169 }, { "epoch": 3.39, "learning_rate": 6.614914137849918e-05, "loss": 2.8423, "step": 43170 }, { "epoch": 3.39, "learning_rate": 6.614835724927468e-05, "loss": 2.7954, "step": 43171 }, { "epoch": 3.39, "learning_rate": 6.614757312005019e-05, "loss": 2.6812, "step": 43172 }, { "epoch": 3.39, "learning_rate": 6.614678899082569e-05, "loss": 2.7902, "step": 43173 }, { "epoch": 3.39, "learning_rate": 6.614600486160119e-05, "loss": 2.9199, "step": 43174 }, { "epoch": 3.39, "learning_rate": 6.61452207323767e-05, "loss": 2.6559, "step": 43175 }, { "epoch": 3.39, "learning_rate": 6.61444366031522e-05, "loss": 2.7213, "step": 43176 }, { "epoch": 3.39, "learning_rate": 6.61436524739277e-05, "loss": 2.6158, "step": 43177 }, { "epoch": 3.39, "learning_rate": 6.614286834470321e-05, "loss": 2.9024, "step": 43178 }, { "epoch": 3.39, "learning_rate": 6.614208421547871e-05, "loss": 2.5545, "step": 43179 }, { "epoch": 3.39, "learning_rate": 6.614130008625422e-05, "loss": 2.5345, "step": 43180 }, { "epoch": 3.39, "learning_rate": 6.614051595702972e-05, "loss": 2.9003, "step": 43181 }, { "epoch": 3.39, "learning_rate": 6.613973182780522e-05, "loss": 2.6804, "step": 43182 }, { "epoch": 3.39, "learning_rate": 6.613894769858074e-05, "loss": 2.5907, "step": 43183 }, { "epoch": 3.39, "learning_rate": 6.613816356935624e-05, "loss": 2.9292, "step": 43184 }, { "epoch": 3.39, "learning_rate": 6.613737944013175e-05, "loss": 2.769, "step": 43185 }, { "epoch": 3.39, "learning_rate": 6.613659531090725e-05, "loss": 2.7006, "step": 43186 }, { "epoch": 3.39, "learning_rate": 6.613581118168275e-05, "loss": 2.8139, "step": 43187 }, { "epoch": 3.39, "learning_rate": 6.613502705245826e-05, "loss": 2.6223, "step": 43188 }, { "epoch": 3.39, "learning_rate": 6.613424292323375e-05, "loss": 2.8028, "step": 43189 }, { "epoch": 3.39, "learning_rate": 6.613345879400925e-05, "loss": 2.6417, "step": 43190 }, { "epoch": 3.39, "learning_rate": 6.613267466478476e-05, "loss": 2.6597, "step": 43191 }, { "epoch": 3.39, "learning_rate": 6.613189053556026e-05, "loss": 2.7456, "step": 43192 }, { "epoch": 3.39, "learning_rate": 6.613110640633576e-05, "loss": 2.6868, "step": 43193 }, { "epoch": 3.39, "learning_rate": 6.613032227711127e-05, "loss": 2.6163, "step": 43194 }, { "epoch": 3.39, "learning_rate": 6.612953814788677e-05, "loss": 2.6717, "step": 43195 }, { "epoch": 3.39, "learning_rate": 6.612875401866227e-05, "loss": 2.7322, "step": 43196 }, { "epoch": 3.39, "learning_rate": 6.612796988943778e-05, "loss": 2.8589, "step": 43197 }, { "epoch": 3.39, "learning_rate": 6.612718576021328e-05, "loss": 2.7112, "step": 43198 }, { "epoch": 3.39, "learning_rate": 6.612640163098879e-05, "loss": 2.9407, "step": 43199 }, { "epoch": 3.39, "learning_rate": 6.612561750176429e-05, "loss": 2.7274, "step": 43200 }, { "epoch": 3.39, "learning_rate": 6.612483337253979e-05, "loss": 2.8296, "step": 43201 }, { "epoch": 3.39, "learning_rate": 6.61240492433153e-05, "loss": 2.6357, "step": 43202 }, { "epoch": 3.39, "learning_rate": 6.61232651140908e-05, "loss": 2.8556, "step": 43203 }, { "epoch": 3.39, "learning_rate": 6.61224809848663e-05, "loss": 2.5127, "step": 43204 }, { "epoch": 3.39, "learning_rate": 6.612169685564181e-05, "loss": 2.8882, "step": 43205 }, { "epoch": 3.39, "learning_rate": 6.612091272641731e-05, "loss": 2.6575, "step": 43206 }, { "epoch": 3.39, "learning_rate": 6.612012859719281e-05, "loss": 2.4957, "step": 43207 }, { "epoch": 3.39, "learning_rate": 6.611934446796832e-05, "loss": 2.8197, "step": 43208 }, { "epoch": 3.39, "learning_rate": 6.611856033874384e-05, "loss": 2.8848, "step": 43209 }, { "epoch": 3.39, "learning_rate": 6.611777620951934e-05, "loss": 2.7205, "step": 43210 }, { "epoch": 3.39, "learning_rate": 6.611699208029484e-05, "loss": 2.6051, "step": 43211 }, { "epoch": 3.39, "learning_rate": 6.611620795107035e-05, "loss": 2.7677, "step": 43212 }, { "epoch": 3.39, "learning_rate": 6.611542382184585e-05, "loss": 2.8923, "step": 43213 }, { "epoch": 3.39, "learning_rate": 6.611463969262135e-05, "loss": 2.5939, "step": 43214 }, { "epoch": 3.39, "learning_rate": 6.611385556339686e-05, "loss": 2.8959, "step": 43215 }, { "epoch": 3.39, "learning_rate": 6.611307143417236e-05, "loss": 2.8556, "step": 43216 }, { "epoch": 3.39, "learning_rate": 6.611228730494786e-05, "loss": 2.5878, "step": 43217 }, { "epoch": 3.39, "learning_rate": 6.611150317572337e-05, "loss": 2.7112, "step": 43218 }, { "epoch": 3.39, "learning_rate": 6.611071904649887e-05, "loss": 2.7618, "step": 43219 }, { "epoch": 3.39, "learning_rate": 6.610993491727438e-05, "loss": 2.4107, "step": 43220 }, { "epoch": 3.39, "learning_rate": 6.610915078804988e-05, "loss": 2.5919, "step": 43221 }, { "epoch": 3.39, "learning_rate": 6.610836665882538e-05, "loss": 3.0199, "step": 43222 }, { "epoch": 3.39, "learning_rate": 6.610758252960089e-05, "loss": 2.6563, "step": 43223 }, { "epoch": 3.39, "learning_rate": 6.610679840037639e-05, "loss": 2.6452, "step": 43224 }, { "epoch": 3.39, "learning_rate": 6.610601427115188e-05, "loss": 3.0205, "step": 43225 }, { "epoch": 3.39, "learning_rate": 6.610523014192738e-05, "loss": 3.1494, "step": 43226 }, { "epoch": 3.39, "learning_rate": 6.610444601270289e-05, "loss": 2.7087, "step": 43227 }, { "epoch": 3.39, "learning_rate": 6.610366188347839e-05, "loss": 2.702, "step": 43228 }, { "epoch": 3.39, "learning_rate": 6.61028777542539e-05, "loss": 2.6474, "step": 43229 }, { "epoch": 3.39, "learning_rate": 6.61020936250294e-05, "loss": 2.7293, "step": 43230 }, { "epoch": 3.39, "learning_rate": 6.61013094958049e-05, "loss": 2.8489, "step": 43231 }, { "epoch": 3.39, "learning_rate": 6.61005253665804e-05, "loss": 2.7365, "step": 43232 }, { "epoch": 3.39, "learning_rate": 6.609974123735591e-05, "loss": 3.0255, "step": 43233 }, { "epoch": 3.39, "learning_rate": 6.609895710813143e-05, "loss": 2.7043, "step": 43234 }, { "epoch": 3.39, "learning_rate": 6.609817297890693e-05, "loss": 2.6953, "step": 43235 }, { "epoch": 3.39, "learning_rate": 6.609738884968243e-05, "loss": 2.8457, "step": 43236 }, { "epoch": 3.39, "learning_rate": 6.609660472045794e-05, "loss": 2.578, "step": 43237 }, { "epoch": 3.39, "learning_rate": 6.609582059123344e-05, "loss": 2.7071, "step": 43238 }, { "epoch": 3.39, "learning_rate": 6.609503646200895e-05, "loss": 2.8743, "step": 43239 }, { "epoch": 3.39, "learning_rate": 6.609425233278445e-05, "loss": 2.7008, "step": 43240 }, { "epoch": 3.39, "learning_rate": 6.609346820355995e-05, "loss": 2.775, "step": 43241 }, { "epoch": 3.39, "learning_rate": 6.609268407433546e-05, "loss": 2.8765, "step": 43242 }, { "epoch": 3.39, "learning_rate": 6.609189994511096e-05, "loss": 2.6764, "step": 43243 }, { "epoch": 3.39, "learning_rate": 6.609111581588646e-05, "loss": 2.7327, "step": 43244 }, { "epoch": 3.39, "learning_rate": 6.609033168666197e-05, "loss": 2.5626, "step": 43245 }, { "epoch": 3.39, "learning_rate": 6.608954755743747e-05, "loss": 2.697, "step": 43246 }, { "epoch": 3.39, "learning_rate": 6.608876342821298e-05, "loss": 2.8687, "step": 43247 }, { "epoch": 3.39, "learning_rate": 6.608797929898848e-05, "loss": 2.661, "step": 43248 }, { "epoch": 3.39, "learning_rate": 6.608719516976398e-05, "loss": 2.8288, "step": 43249 }, { "epoch": 3.39, "learning_rate": 6.608641104053949e-05, "loss": 3.0023, "step": 43250 }, { "epoch": 3.39, "learning_rate": 6.608562691131499e-05, "loss": 2.7046, "step": 43251 }, { "epoch": 3.39, "learning_rate": 6.60848427820905e-05, "loss": 3.0025, "step": 43252 }, { "epoch": 3.39, "learning_rate": 6.6084058652866e-05, "loss": 2.7203, "step": 43253 }, { "epoch": 3.39, "learning_rate": 6.60832745236415e-05, "loss": 2.9579, "step": 43254 }, { "epoch": 3.39, "learning_rate": 6.6082490394417e-05, "loss": 2.6734, "step": 43255 }, { "epoch": 3.39, "learning_rate": 6.608170626519251e-05, "loss": 2.6723, "step": 43256 }, { "epoch": 3.39, "learning_rate": 6.608092213596801e-05, "loss": 2.7784, "step": 43257 }, { "epoch": 3.39, "learning_rate": 6.608013800674352e-05, "loss": 2.7012, "step": 43258 }, { "epoch": 3.39, "learning_rate": 6.607935387751902e-05, "loss": 2.5502, "step": 43259 }, { "epoch": 3.39, "learning_rate": 6.607856974829452e-05, "loss": 2.8503, "step": 43260 }, { "epoch": 3.39, "learning_rate": 6.607778561907003e-05, "loss": 2.6397, "step": 43261 }, { "epoch": 3.39, "learning_rate": 6.607700148984553e-05, "loss": 2.6401, "step": 43262 }, { "epoch": 3.39, "learning_rate": 6.607621736062103e-05, "loss": 2.6536, "step": 43263 }, { "epoch": 3.39, "learning_rate": 6.607543323139654e-05, "loss": 2.8453, "step": 43264 }, { "epoch": 3.39, "learning_rate": 6.607464910217204e-05, "loss": 2.7888, "step": 43265 }, { "epoch": 3.39, "learning_rate": 6.607386497294755e-05, "loss": 2.5879, "step": 43266 }, { "epoch": 3.39, "learning_rate": 6.607308084372305e-05, "loss": 2.7966, "step": 43267 }, { "epoch": 3.39, "learning_rate": 6.607229671449855e-05, "loss": 2.6838, "step": 43268 }, { "epoch": 3.39, "learning_rate": 6.607151258527406e-05, "loss": 2.7112, "step": 43269 }, { "epoch": 3.39, "learning_rate": 6.607072845604956e-05, "loss": 2.6207, "step": 43270 }, { "epoch": 3.39, "learning_rate": 6.606994432682506e-05, "loss": 2.5816, "step": 43271 }, { "epoch": 3.39, "learning_rate": 6.606916019760057e-05, "loss": 2.7899, "step": 43272 }, { "epoch": 3.39, "learning_rate": 6.606837606837607e-05, "loss": 2.5918, "step": 43273 }, { "epoch": 3.39, "learning_rate": 6.606759193915157e-05, "loss": 2.8546, "step": 43274 }, { "epoch": 3.39, "learning_rate": 6.606680780992708e-05, "loss": 2.8272, "step": 43275 }, { "epoch": 3.39, "learning_rate": 6.606602368070258e-05, "loss": 2.6767, "step": 43276 }, { "epoch": 3.39, "learning_rate": 6.606523955147809e-05, "loss": 2.7872, "step": 43277 }, { "epoch": 3.39, "learning_rate": 6.606445542225359e-05, "loss": 2.7916, "step": 43278 }, { "epoch": 3.39, "learning_rate": 6.606367129302909e-05, "loss": 3.0689, "step": 43279 }, { "epoch": 3.39, "learning_rate": 6.60628871638046e-05, "loss": 2.4994, "step": 43280 }, { "epoch": 3.39, "learning_rate": 6.60621030345801e-05, "loss": 2.6866, "step": 43281 }, { "epoch": 3.39, "learning_rate": 6.60613189053556e-05, "loss": 2.7264, "step": 43282 }, { "epoch": 3.39, "learning_rate": 6.606053477613111e-05, "loss": 2.6338, "step": 43283 }, { "epoch": 3.39, "learning_rate": 6.605975064690661e-05, "loss": 2.6897, "step": 43284 }, { "epoch": 3.39, "learning_rate": 6.605896651768211e-05, "loss": 2.9095, "step": 43285 }, { "epoch": 3.39, "learning_rate": 6.605818238845762e-05, "loss": 2.6121, "step": 43286 }, { "epoch": 3.39, "learning_rate": 6.605739825923312e-05, "loss": 2.5607, "step": 43287 }, { "epoch": 3.39, "learning_rate": 6.605661413000863e-05, "loss": 2.4997, "step": 43288 }, { "epoch": 3.39, "learning_rate": 6.605583000078413e-05, "loss": 2.7807, "step": 43289 }, { "epoch": 3.39, "learning_rate": 6.605504587155963e-05, "loss": 3.0264, "step": 43290 }, { "epoch": 3.39, "learning_rate": 6.605426174233515e-05, "loss": 2.553, "step": 43291 }, { "epoch": 3.39, "learning_rate": 6.605347761311065e-05, "loss": 2.7542, "step": 43292 }, { "epoch": 3.39, "learning_rate": 6.605269348388616e-05, "loss": 2.8163, "step": 43293 }, { "epoch": 3.39, "learning_rate": 6.605190935466166e-05, "loss": 2.7423, "step": 43294 }, { "epoch": 3.39, "learning_rate": 6.605112522543717e-05, "loss": 2.5308, "step": 43295 }, { "epoch": 3.39, "learning_rate": 6.605034109621267e-05, "loss": 2.7855, "step": 43296 }, { "epoch": 3.4, "learning_rate": 6.604955696698816e-05, "loss": 2.7905, "step": 43297 }, { "epoch": 3.4, "learning_rate": 6.604877283776366e-05, "loss": 2.818, "step": 43298 }, { "epoch": 3.4, "learning_rate": 6.604798870853917e-05, "loss": 2.5848, "step": 43299 }, { "epoch": 3.4, "learning_rate": 6.604720457931467e-05, "loss": 2.8728, "step": 43300 }, { "epoch": 3.4, "learning_rate": 6.604642045009017e-05, "loss": 2.5533, "step": 43301 }, { "epoch": 3.4, "learning_rate": 6.604563632086568e-05, "loss": 2.6167, "step": 43302 }, { "epoch": 3.4, "learning_rate": 6.604485219164118e-05, "loss": 2.5983, "step": 43303 }, { "epoch": 3.4, "learning_rate": 6.604406806241668e-05, "loss": 2.5581, "step": 43304 }, { "epoch": 3.4, "learning_rate": 6.604328393319219e-05, "loss": 2.663, "step": 43305 }, { "epoch": 3.4, "learning_rate": 6.604249980396769e-05, "loss": 2.6622, "step": 43306 }, { "epoch": 3.4, "learning_rate": 6.60417156747432e-05, "loss": 2.7909, "step": 43307 }, { "epoch": 3.4, "learning_rate": 6.60409315455187e-05, "loss": 2.8045, "step": 43308 }, { "epoch": 3.4, "learning_rate": 6.60401474162942e-05, "loss": 2.8177, "step": 43309 }, { "epoch": 3.4, "learning_rate": 6.603936328706971e-05, "loss": 2.7218, "step": 43310 }, { "epoch": 3.4, "learning_rate": 6.603857915784521e-05, "loss": 2.9178, "step": 43311 }, { "epoch": 3.4, "learning_rate": 6.603779502862071e-05, "loss": 2.6162, "step": 43312 }, { "epoch": 3.4, "learning_rate": 6.603701089939622e-05, "loss": 2.6741, "step": 43313 }, { "epoch": 3.4, "learning_rate": 6.603622677017172e-05, "loss": 2.6728, "step": 43314 }, { "epoch": 3.4, "learning_rate": 6.603544264094723e-05, "loss": 2.8332, "step": 43315 }, { "epoch": 3.4, "learning_rate": 6.603465851172273e-05, "loss": 2.4695, "step": 43316 }, { "epoch": 3.4, "learning_rate": 6.603387438249825e-05, "loss": 2.8746, "step": 43317 }, { "epoch": 3.4, "learning_rate": 6.603309025327375e-05, "loss": 2.5064, "step": 43318 }, { "epoch": 3.4, "learning_rate": 6.603230612404925e-05, "loss": 2.7534, "step": 43319 }, { "epoch": 3.4, "learning_rate": 6.603152199482476e-05, "loss": 2.464, "step": 43320 }, { "epoch": 3.4, "learning_rate": 6.603073786560026e-05, "loss": 2.6872, "step": 43321 }, { "epoch": 3.4, "learning_rate": 6.602995373637576e-05, "loss": 2.765, "step": 43322 }, { "epoch": 3.4, "learning_rate": 6.602916960715127e-05, "loss": 2.7896, "step": 43323 }, { "epoch": 3.4, "learning_rate": 6.602838547792677e-05, "loss": 2.7184, "step": 43324 }, { "epoch": 3.4, "learning_rate": 6.602760134870228e-05, "loss": 2.6409, "step": 43325 }, { "epoch": 3.4, "learning_rate": 6.602681721947778e-05, "loss": 2.7438, "step": 43326 }, { "epoch": 3.4, "learning_rate": 6.602603309025328e-05, "loss": 2.8292, "step": 43327 }, { "epoch": 3.4, "learning_rate": 6.602524896102879e-05, "loss": 2.8918, "step": 43328 }, { "epoch": 3.4, "learning_rate": 6.602446483180429e-05, "loss": 2.7777, "step": 43329 }, { "epoch": 3.4, "learning_rate": 6.60236807025798e-05, "loss": 2.8482, "step": 43330 }, { "epoch": 3.4, "learning_rate": 6.60228965733553e-05, "loss": 2.7456, "step": 43331 }, { "epoch": 3.4, "learning_rate": 6.60221124441308e-05, "loss": 2.9403, "step": 43332 }, { "epoch": 3.4, "learning_rate": 6.602132831490629e-05, "loss": 2.6701, "step": 43333 }, { "epoch": 3.4, "learning_rate": 6.60205441856818e-05, "loss": 2.6456, "step": 43334 }, { "epoch": 3.4, "learning_rate": 6.60197600564573e-05, "loss": 2.7491, "step": 43335 }, { "epoch": 3.4, "learning_rate": 6.60189759272328e-05, "loss": 2.781, "step": 43336 }, { "epoch": 3.4, "learning_rate": 6.60181917980083e-05, "loss": 2.7964, "step": 43337 }, { "epoch": 3.4, "learning_rate": 6.601740766878381e-05, "loss": 2.8733, "step": 43338 }, { "epoch": 3.4, "learning_rate": 6.601662353955931e-05, "loss": 2.6484, "step": 43339 }, { "epoch": 3.4, "learning_rate": 6.601583941033482e-05, "loss": 2.7342, "step": 43340 }, { "epoch": 3.4, "learning_rate": 6.601505528111032e-05, "loss": 2.5045, "step": 43341 }, { "epoch": 3.4, "learning_rate": 6.601427115188584e-05, "loss": 2.5524, "step": 43342 }, { "epoch": 3.4, "learning_rate": 6.601348702266134e-05, "loss": 2.6942, "step": 43343 }, { "epoch": 3.4, "learning_rate": 6.601270289343685e-05, "loss": 3.057, "step": 43344 }, { "epoch": 3.4, "learning_rate": 6.601191876421235e-05, "loss": 2.5247, "step": 43345 }, { "epoch": 3.4, "learning_rate": 6.601113463498785e-05, "loss": 2.8091, "step": 43346 }, { "epoch": 3.4, "learning_rate": 6.601035050576336e-05, "loss": 3.0358, "step": 43347 }, { "epoch": 3.4, "learning_rate": 6.600956637653886e-05, "loss": 2.6883, "step": 43348 }, { "epoch": 3.4, "learning_rate": 6.600878224731436e-05, "loss": 2.8165, "step": 43349 }, { "epoch": 3.4, "learning_rate": 6.600799811808987e-05, "loss": 2.9109, "step": 43350 }, { "epoch": 3.4, "learning_rate": 6.600721398886537e-05, "loss": 2.8061, "step": 43351 }, { "epoch": 3.4, "learning_rate": 6.600642985964087e-05, "loss": 2.7145, "step": 43352 }, { "epoch": 3.4, "learning_rate": 6.600564573041638e-05, "loss": 2.7609, "step": 43353 }, { "epoch": 3.4, "learning_rate": 6.600486160119188e-05, "loss": 2.7318, "step": 43354 }, { "epoch": 3.4, "learning_rate": 6.600407747196739e-05, "loss": 2.7428, "step": 43355 }, { "epoch": 3.4, "learning_rate": 6.600329334274289e-05, "loss": 2.7482, "step": 43356 }, { "epoch": 3.4, "learning_rate": 6.60025092135184e-05, "loss": 2.8083, "step": 43357 }, { "epoch": 3.4, "learning_rate": 6.60017250842939e-05, "loss": 2.7259, "step": 43358 }, { "epoch": 3.4, "learning_rate": 6.60009409550694e-05, "loss": 2.7269, "step": 43359 }, { "epoch": 3.4, "learning_rate": 6.60001568258449e-05, "loss": 2.8613, "step": 43360 }, { "epoch": 3.4, "learning_rate": 6.599937269662041e-05, "loss": 2.8822, "step": 43361 }, { "epoch": 3.4, "learning_rate": 6.599858856739591e-05, "loss": 2.7016, "step": 43362 }, { "epoch": 3.4, "learning_rate": 6.599780443817142e-05, "loss": 2.6579, "step": 43363 }, { "epoch": 3.4, "learning_rate": 6.599702030894692e-05, "loss": 2.7462, "step": 43364 }, { "epoch": 3.4, "learning_rate": 6.599623617972242e-05, "loss": 2.9279, "step": 43365 }, { "epoch": 3.4, "learning_rate": 6.599545205049793e-05, "loss": 2.5652, "step": 43366 }, { "epoch": 3.4, "learning_rate": 6.599466792127343e-05, "loss": 2.7512, "step": 43367 }, { "epoch": 3.4, "learning_rate": 6.599388379204893e-05, "loss": 2.7834, "step": 43368 }, { "epoch": 3.4, "learning_rate": 6.599309966282444e-05, "loss": 2.8714, "step": 43369 }, { "epoch": 3.4, "learning_rate": 6.599231553359994e-05, "loss": 2.7141, "step": 43370 }, { "epoch": 3.4, "learning_rate": 6.599153140437544e-05, "loss": 2.6992, "step": 43371 }, { "epoch": 3.4, "learning_rate": 6.599074727515095e-05, "loss": 2.7858, "step": 43372 }, { "epoch": 3.4, "learning_rate": 6.598996314592645e-05, "loss": 2.7413, "step": 43373 }, { "epoch": 3.4, "learning_rate": 6.598917901670196e-05, "loss": 2.7762, "step": 43374 }, { "epoch": 3.4, "learning_rate": 6.598839488747746e-05, "loss": 2.5684, "step": 43375 }, { "epoch": 3.4, "learning_rate": 6.598761075825296e-05, "loss": 2.6898, "step": 43376 }, { "epoch": 3.4, "learning_rate": 6.598682662902847e-05, "loss": 2.7519, "step": 43377 }, { "epoch": 3.4, "learning_rate": 6.598604249980397e-05, "loss": 2.7997, "step": 43378 }, { "epoch": 3.4, "learning_rate": 6.598525837057947e-05, "loss": 2.9447, "step": 43379 }, { "epoch": 3.4, "learning_rate": 6.598447424135498e-05, "loss": 2.8579, "step": 43380 }, { "epoch": 3.4, "learning_rate": 6.598369011213048e-05, "loss": 2.8399, "step": 43381 }, { "epoch": 3.4, "learning_rate": 6.598290598290599e-05, "loss": 2.845, "step": 43382 }, { "epoch": 3.4, "learning_rate": 6.598212185368149e-05, "loss": 2.5761, "step": 43383 }, { "epoch": 3.4, "learning_rate": 6.598133772445699e-05, "loss": 2.7434, "step": 43384 }, { "epoch": 3.4, "learning_rate": 6.59805535952325e-05, "loss": 2.7039, "step": 43385 }, { "epoch": 3.4, "learning_rate": 6.5979769466008e-05, "loss": 2.9222, "step": 43386 }, { "epoch": 3.4, "learning_rate": 6.59789853367835e-05, "loss": 2.7935, "step": 43387 }, { "epoch": 3.4, "learning_rate": 6.597820120755901e-05, "loss": 2.8319, "step": 43388 }, { "epoch": 3.4, "learning_rate": 6.597741707833451e-05, "loss": 2.8336, "step": 43389 }, { "epoch": 3.4, "learning_rate": 6.597663294911001e-05, "loss": 2.648, "step": 43390 }, { "epoch": 3.4, "learning_rate": 6.597584881988552e-05, "loss": 2.6711, "step": 43391 }, { "epoch": 3.4, "learning_rate": 6.597506469066102e-05, "loss": 2.573, "step": 43392 }, { "epoch": 3.4, "learning_rate": 6.597428056143653e-05, "loss": 2.6867, "step": 43393 }, { "epoch": 3.4, "learning_rate": 6.597349643221203e-05, "loss": 2.7668, "step": 43394 }, { "epoch": 3.4, "learning_rate": 6.597271230298753e-05, "loss": 2.6132, "step": 43395 }, { "epoch": 3.4, "learning_rate": 6.597192817376304e-05, "loss": 2.8502, "step": 43396 }, { "epoch": 3.4, "learning_rate": 6.597114404453854e-05, "loss": 2.9968, "step": 43397 }, { "epoch": 3.4, "learning_rate": 6.597035991531404e-05, "loss": 2.7636, "step": 43398 }, { "epoch": 3.4, "learning_rate": 6.596957578608956e-05, "loss": 2.577, "step": 43399 }, { "epoch": 3.4, "learning_rate": 6.596879165686506e-05, "loss": 2.5709, "step": 43400 }, { "epoch": 3.4, "learning_rate": 6.596800752764057e-05, "loss": 2.8449, "step": 43401 }, { "epoch": 3.4, "learning_rate": 6.596722339841607e-05, "loss": 2.6699, "step": 43402 }, { "epoch": 3.4, "learning_rate": 6.596643926919158e-05, "loss": 2.6778, "step": 43403 }, { "epoch": 3.4, "learning_rate": 6.596565513996708e-05, "loss": 2.7018, "step": 43404 }, { "epoch": 3.4, "learning_rate": 6.596487101074257e-05, "loss": 2.8832, "step": 43405 }, { "epoch": 3.4, "learning_rate": 6.596408688151807e-05, "loss": 2.5884, "step": 43406 }, { "epoch": 3.4, "learning_rate": 6.596330275229358e-05, "loss": 2.6671, "step": 43407 }, { "epoch": 3.4, "learning_rate": 6.596251862306908e-05, "loss": 2.6769, "step": 43408 }, { "epoch": 3.4, "learning_rate": 6.596173449384458e-05, "loss": 2.7703, "step": 43409 }, { "epoch": 3.4, "learning_rate": 6.596095036462009e-05, "loss": 2.8959, "step": 43410 }, { "epoch": 3.4, "learning_rate": 6.596016623539559e-05, "loss": 2.935, "step": 43411 }, { "epoch": 3.4, "learning_rate": 6.59593821061711e-05, "loss": 2.7831, "step": 43412 }, { "epoch": 3.4, "learning_rate": 6.59585979769466e-05, "loss": 2.7642, "step": 43413 }, { "epoch": 3.4, "learning_rate": 6.59578138477221e-05, "loss": 2.7691, "step": 43414 }, { "epoch": 3.4, "learning_rate": 6.59570297184976e-05, "loss": 2.6173, "step": 43415 }, { "epoch": 3.4, "learning_rate": 6.595624558927311e-05, "loss": 2.5128, "step": 43416 }, { "epoch": 3.4, "learning_rate": 6.595546146004861e-05, "loss": 2.8068, "step": 43417 }, { "epoch": 3.4, "learning_rate": 6.595467733082412e-05, "loss": 2.7517, "step": 43418 }, { "epoch": 3.4, "learning_rate": 6.595389320159962e-05, "loss": 2.6629, "step": 43419 }, { "epoch": 3.4, "learning_rate": 6.595310907237512e-05, "loss": 2.5994, "step": 43420 }, { "epoch": 3.4, "learning_rate": 6.595232494315063e-05, "loss": 2.6898, "step": 43421 }, { "epoch": 3.4, "learning_rate": 6.595154081392613e-05, "loss": 2.7066, "step": 43422 }, { "epoch": 3.4, "learning_rate": 6.595075668470164e-05, "loss": 2.7337, "step": 43423 }, { "epoch": 3.41, "learning_rate": 6.594997255547714e-05, "loss": 2.5127, "step": 43424 }, { "epoch": 3.41, "learning_rate": 6.594918842625266e-05, "loss": 2.7468, "step": 43425 }, { "epoch": 3.41, "learning_rate": 6.594840429702816e-05, "loss": 2.6069, "step": 43426 }, { "epoch": 3.41, "learning_rate": 6.594762016780366e-05, "loss": 2.6884, "step": 43427 }, { "epoch": 3.41, "learning_rate": 6.594683603857917e-05, "loss": 2.6931, "step": 43428 }, { "epoch": 3.41, "learning_rate": 6.594605190935467e-05, "loss": 2.7799, "step": 43429 }, { "epoch": 3.41, "learning_rate": 6.594526778013018e-05, "loss": 2.7549, "step": 43430 }, { "epoch": 3.41, "learning_rate": 6.594448365090568e-05, "loss": 2.6307, "step": 43431 }, { "epoch": 3.41, "learning_rate": 6.594369952168118e-05, "loss": 2.7077, "step": 43432 }, { "epoch": 3.41, "learning_rate": 6.594291539245669e-05, "loss": 2.8099, "step": 43433 }, { "epoch": 3.41, "learning_rate": 6.594213126323219e-05, "loss": 2.6119, "step": 43434 }, { "epoch": 3.41, "learning_rate": 6.59413471340077e-05, "loss": 2.688, "step": 43435 }, { "epoch": 3.41, "learning_rate": 6.59405630047832e-05, "loss": 3.1118, "step": 43436 }, { "epoch": 3.41, "learning_rate": 6.59397788755587e-05, "loss": 2.6407, "step": 43437 }, { "epoch": 3.41, "learning_rate": 6.59389947463342e-05, "loss": 2.9093, "step": 43438 }, { "epoch": 3.41, "learning_rate": 6.593821061710971e-05, "loss": 2.9494, "step": 43439 }, { "epoch": 3.41, "learning_rate": 6.593742648788521e-05, "loss": 2.7789, "step": 43440 }, { "epoch": 3.41, "learning_rate": 6.59366423586607e-05, "loss": 2.8016, "step": 43441 }, { "epoch": 3.41, "learning_rate": 6.59358582294362e-05, "loss": 2.9042, "step": 43442 }, { "epoch": 3.41, "learning_rate": 6.593507410021171e-05, "loss": 2.6953, "step": 43443 }, { "epoch": 3.41, "learning_rate": 6.593428997098721e-05, "loss": 2.6504, "step": 43444 }, { "epoch": 3.41, "learning_rate": 6.593350584176272e-05, "loss": 2.7985, "step": 43445 }, { "epoch": 3.41, "learning_rate": 6.593272171253822e-05, "loss": 2.785, "step": 43446 }, { "epoch": 3.41, "learning_rate": 6.593193758331372e-05, "loss": 2.4761, "step": 43447 }, { "epoch": 3.41, "learning_rate": 6.593115345408923e-05, "loss": 2.7889, "step": 43448 }, { "epoch": 3.41, "learning_rate": 6.593036932486473e-05, "loss": 2.9368, "step": 43449 }, { "epoch": 3.41, "learning_rate": 6.592958519564025e-05, "loss": 2.606, "step": 43450 }, { "epoch": 3.41, "learning_rate": 6.592880106641575e-05, "loss": 2.6553, "step": 43451 }, { "epoch": 3.41, "learning_rate": 6.592801693719126e-05, "loss": 2.5416, "step": 43452 }, { "epoch": 3.41, "learning_rate": 6.592723280796676e-05, "loss": 2.8678, "step": 43453 }, { "epoch": 3.41, "learning_rate": 6.592644867874226e-05, "loss": 2.8845, "step": 43454 }, { "epoch": 3.41, "learning_rate": 6.592566454951777e-05, "loss": 2.7605, "step": 43455 }, { "epoch": 3.41, "learning_rate": 6.592488042029327e-05, "loss": 2.7953, "step": 43456 }, { "epoch": 3.41, "learning_rate": 6.592409629106877e-05, "loss": 2.6699, "step": 43457 }, { "epoch": 3.41, "learning_rate": 6.592331216184428e-05, "loss": 2.9475, "step": 43458 }, { "epoch": 3.41, "learning_rate": 6.592252803261978e-05, "loss": 2.6281, "step": 43459 }, { "epoch": 3.41, "learning_rate": 6.592174390339529e-05, "loss": 2.864, "step": 43460 }, { "epoch": 3.41, "learning_rate": 6.592095977417079e-05, "loss": 2.6765, "step": 43461 }, { "epoch": 3.41, "learning_rate": 6.592017564494629e-05, "loss": 2.7001, "step": 43462 }, { "epoch": 3.41, "learning_rate": 6.59193915157218e-05, "loss": 2.7332, "step": 43463 }, { "epoch": 3.41, "learning_rate": 6.59186073864973e-05, "loss": 2.845, "step": 43464 }, { "epoch": 3.41, "learning_rate": 6.59178232572728e-05, "loss": 2.6989, "step": 43465 }, { "epoch": 3.41, "learning_rate": 6.591703912804831e-05, "loss": 2.7855, "step": 43466 }, { "epoch": 3.41, "learning_rate": 6.591625499882381e-05, "loss": 2.6121, "step": 43467 }, { "epoch": 3.41, "learning_rate": 6.591547086959932e-05, "loss": 2.7059, "step": 43468 }, { "epoch": 3.41, "learning_rate": 6.591468674037482e-05, "loss": 2.6231, "step": 43469 }, { "epoch": 3.41, "learning_rate": 6.591390261115032e-05, "loss": 2.6947, "step": 43470 }, { "epoch": 3.41, "learning_rate": 6.591311848192583e-05, "loss": 2.8503, "step": 43471 }, { "epoch": 3.41, "learning_rate": 6.591233435270133e-05, "loss": 2.8203, "step": 43472 }, { "epoch": 3.41, "learning_rate": 6.591155022347683e-05, "loss": 2.6388, "step": 43473 }, { "epoch": 3.41, "learning_rate": 6.591076609425234e-05, "loss": 2.6845, "step": 43474 }, { "epoch": 3.41, "learning_rate": 6.590998196502784e-05, "loss": 2.9119, "step": 43475 }, { "epoch": 3.41, "learning_rate": 6.590919783580334e-05, "loss": 2.7802, "step": 43476 }, { "epoch": 3.41, "learning_rate": 6.590841370657885e-05, "loss": 2.771, "step": 43477 }, { "epoch": 3.41, "learning_rate": 6.590762957735435e-05, "loss": 2.9181, "step": 43478 }, { "epoch": 3.41, "learning_rate": 6.590684544812986e-05, "loss": 2.7098, "step": 43479 }, { "epoch": 3.41, "learning_rate": 6.590606131890536e-05, "loss": 2.5119, "step": 43480 }, { "epoch": 3.41, "learning_rate": 6.590527718968086e-05, "loss": 2.8631, "step": 43481 }, { "epoch": 3.41, "learning_rate": 6.590449306045637e-05, "loss": 2.6424, "step": 43482 }, { "epoch": 3.41, "learning_rate": 6.590370893123187e-05, "loss": 2.6341, "step": 43483 }, { "epoch": 3.41, "learning_rate": 6.590292480200737e-05, "loss": 2.776, "step": 43484 }, { "epoch": 3.41, "learning_rate": 6.590214067278288e-05, "loss": 2.574, "step": 43485 }, { "epoch": 3.41, "learning_rate": 6.590135654355838e-05, "loss": 2.9788, "step": 43486 }, { "epoch": 3.41, "learning_rate": 6.590057241433388e-05, "loss": 2.6187, "step": 43487 }, { "epoch": 3.41, "learning_rate": 6.589978828510939e-05, "loss": 2.7061, "step": 43488 }, { "epoch": 3.41, "learning_rate": 6.589900415588489e-05, "loss": 2.8221, "step": 43489 }, { "epoch": 3.41, "learning_rate": 6.58982200266604e-05, "loss": 2.8567, "step": 43490 }, { "epoch": 3.41, "learning_rate": 6.58974358974359e-05, "loss": 2.643, "step": 43491 }, { "epoch": 3.41, "learning_rate": 6.58966517682114e-05, "loss": 2.7942, "step": 43492 }, { "epoch": 3.41, "learning_rate": 6.589586763898691e-05, "loss": 3.0162, "step": 43493 }, { "epoch": 3.41, "learning_rate": 6.589508350976241e-05, "loss": 2.5528, "step": 43494 }, { "epoch": 3.41, "learning_rate": 6.589429938053791e-05, "loss": 2.6636, "step": 43495 }, { "epoch": 3.41, "learning_rate": 6.589351525131342e-05, "loss": 2.7127, "step": 43496 }, { "epoch": 3.41, "learning_rate": 6.589273112208892e-05, "loss": 2.6183, "step": 43497 }, { "epoch": 3.41, "learning_rate": 6.589194699286443e-05, "loss": 2.7684, "step": 43498 }, { "epoch": 3.41, "learning_rate": 6.589116286363993e-05, "loss": 2.7028, "step": 43499 }, { "epoch": 3.41, "learning_rate": 6.589037873441543e-05, "loss": 2.6772, "step": 43500 }, { "epoch": 3.41, "learning_rate": 6.588959460519094e-05, "loss": 2.5899, "step": 43501 }, { "epoch": 3.41, "learning_rate": 6.588881047596644e-05, "loss": 2.7878, "step": 43502 }, { "epoch": 3.41, "learning_rate": 6.588802634674194e-05, "loss": 2.6551, "step": 43503 }, { "epoch": 3.41, "learning_rate": 6.588724221751745e-05, "loss": 2.7735, "step": 43504 }, { "epoch": 3.41, "learning_rate": 6.588645808829295e-05, "loss": 3.1027, "step": 43505 }, { "epoch": 3.41, "learning_rate": 6.588567395906845e-05, "loss": 2.7607, "step": 43506 }, { "epoch": 3.41, "learning_rate": 6.588488982984397e-05, "loss": 2.6669, "step": 43507 }, { "epoch": 3.41, "learning_rate": 6.588410570061948e-05, "loss": 2.7372, "step": 43508 }, { "epoch": 3.41, "learning_rate": 6.588332157139498e-05, "loss": 3.2975, "step": 43509 }, { "epoch": 3.41, "learning_rate": 6.588253744217048e-05, "loss": 2.9085, "step": 43510 }, { "epoch": 3.41, "learning_rate": 6.588175331294599e-05, "loss": 2.891, "step": 43511 }, { "epoch": 3.41, "learning_rate": 6.588096918372149e-05, "loss": 2.4672, "step": 43512 }, { "epoch": 3.41, "learning_rate": 6.588018505449698e-05, "loss": 2.6437, "step": 43513 }, { "epoch": 3.41, "learning_rate": 6.587940092527248e-05, "loss": 2.754, "step": 43514 }, { "epoch": 3.41, "learning_rate": 6.587861679604799e-05, "loss": 2.7453, "step": 43515 }, { "epoch": 3.41, "learning_rate": 6.587783266682349e-05, "loss": 2.7317, "step": 43516 }, { "epoch": 3.41, "learning_rate": 6.5877048537599e-05, "loss": 2.76, "step": 43517 }, { "epoch": 3.41, "learning_rate": 6.58762644083745e-05, "loss": 2.6862, "step": 43518 }, { "epoch": 3.41, "learning_rate": 6.587548027915e-05, "loss": 2.7949, "step": 43519 }, { "epoch": 3.41, "learning_rate": 6.58746961499255e-05, "loss": 2.4677, "step": 43520 }, { "epoch": 3.41, "learning_rate": 6.587391202070101e-05, "loss": 2.8653, "step": 43521 }, { "epoch": 3.41, "learning_rate": 6.587312789147651e-05, "loss": 2.6879, "step": 43522 }, { "epoch": 3.41, "learning_rate": 6.587234376225202e-05, "loss": 2.5603, "step": 43523 }, { "epoch": 3.41, "learning_rate": 6.587155963302752e-05, "loss": 2.7192, "step": 43524 }, { "epoch": 3.41, "learning_rate": 6.587077550380302e-05, "loss": 2.6753, "step": 43525 }, { "epoch": 3.41, "learning_rate": 6.586999137457853e-05, "loss": 2.795, "step": 43526 }, { "epoch": 3.41, "learning_rate": 6.586920724535403e-05, "loss": 2.8537, "step": 43527 }, { "epoch": 3.41, "learning_rate": 6.586842311612954e-05, "loss": 2.7079, "step": 43528 }, { "epoch": 3.41, "learning_rate": 6.586763898690504e-05, "loss": 2.7058, "step": 43529 }, { "epoch": 3.41, "learning_rate": 6.586685485768054e-05, "loss": 2.9227, "step": 43530 }, { "epoch": 3.41, "learning_rate": 6.586607072845605e-05, "loss": 2.8611, "step": 43531 }, { "epoch": 3.41, "learning_rate": 6.586528659923155e-05, "loss": 2.6361, "step": 43532 }, { "epoch": 3.41, "learning_rate": 6.586450247000707e-05, "loss": 2.7814, "step": 43533 }, { "epoch": 3.41, "learning_rate": 6.586371834078257e-05, "loss": 2.7273, "step": 43534 }, { "epoch": 3.41, "learning_rate": 6.586293421155808e-05, "loss": 2.6991, "step": 43535 }, { "epoch": 3.41, "learning_rate": 6.586215008233358e-05, "loss": 2.5757, "step": 43536 }, { "epoch": 3.41, "learning_rate": 6.586136595310908e-05, "loss": 2.5817, "step": 43537 }, { "epoch": 3.41, "learning_rate": 6.586058182388459e-05, "loss": 2.6452, "step": 43538 }, { "epoch": 3.41, "learning_rate": 6.585979769466009e-05, "loss": 2.6702, "step": 43539 }, { "epoch": 3.41, "learning_rate": 6.58590135654356e-05, "loss": 2.8618, "step": 43540 }, { "epoch": 3.41, "learning_rate": 6.58582294362111e-05, "loss": 2.8726, "step": 43541 }, { "epoch": 3.41, "learning_rate": 6.58574453069866e-05, "loss": 2.5361, "step": 43542 }, { "epoch": 3.41, "learning_rate": 6.58566611777621e-05, "loss": 2.6672, "step": 43543 }, { "epoch": 3.41, "learning_rate": 6.585587704853761e-05, "loss": 2.7585, "step": 43544 }, { "epoch": 3.41, "learning_rate": 6.585509291931311e-05, "loss": 2.8558, "step": 43545 }, { "epoch": 3.41, "learning_rate": 6.585430879008862e-05, "loss": 2.7866, "step": 43546 }, { "epoch": 3.41, "learning_rate": 6.585352466086412e-05, "loss": 2.6491, "step": 43547 }, { "epoch": 3.41, "learning_rate": 6.585274053163962e-05, "loss": 2.7614, "step": 43548 }, { "epoch": 3.41, "learning_rate": 6.585195640241511e-05, "loss": 2.7314, "step": 43549 }, { "epoch": 3.41, "learning_rate": 6.585117227319062e-05, "loss": 2.7, "step": 43550 }, { "epoch": 3.41, "learning_rate": 6.585038814396612e-05, "loss": 2.5454, "step": 43551 }, { "epoch": 3.42, "learning_rate": 6.584960401474162e-05, "loss": 2.8982, "step": 43552 }, { "epoch": 3.42, "learning_rate": 6.584881988551713e-05, "loss": 2.8451, "step": 43553 }, { "epoch": 3.42, "learning_rate": 6.584803575629263e-05, "loss": 2.6396, "step": 43554 }, { "epoch": 3.42, "learning_rate": 6.584725162706814e-05, "loss": 2.7812, "step": 43555 }, { "epoch": 3.42, "learning_rate": 6.584646749784364e-05, "loss": 2.6923, "step": 43556 }, { "epoch": 3.42, "learning_rate": 6.584568336861914e-05, "loss": 2.7241, "step": 43557 }, { "epoch": 3.42, "learning_rate": 6.584489923939466e-05, "loss": 2.645, "step": 43558 }, { "epoch": 3.42, "learning_rate": 6.584411511017016e-05, "loss": 2.798, "step": 43559 }, { "epoch": 3.42, "learning_rate": 6.584333098094567e-05, "loss": 2.8561, "step": 43560 }, { "epoch": 3.42, "learning_rate": 6.584254685172117e-05, "loss": 2.7293, "step": 43561 }, { "epoch": 3.42, "learning_rate": 6.584176272249667e-05, "loss": 3.0503, "step": 43562 }, { "epoch": 3.42, "learning_rate": 6.584097859327218e-05, "loss": 2.9166, "step": 43563 }, { "epoch": 3.42, "learning_rate": 6.584019446404768e-05, "loss": 2.4675, "step": 43564 }, { "epoch": 3.42, "learning_rate": 6.583941033482319e-05, "loss": 2.6845, "step": 43565 }, { "epoch": 3.42, "learning_rate": 6.583862620559869e-05, "loss": 2.8474, "step": 43566 }, { "epoch": 3.42, "learning_rate": 6.583784207637419e-05, "loss": 2.7905, "step": 43567 }, { "epoch": 3.42, "learning_rate": 6.58370579471497e-05, "loss": 2.6652, "step": 43568 }, { "epoch": 3.42, "learning_rate": 6.58362738179252e-05, "loss": 2.6187, "step": 43569 }, { "epoch": 3.42, "learning_rate": 6.58354896887007e-05, "loss": 2.5676, "step": 43570 }, { "epoch": 3.42, "learning_rate": 6.583470555947621e-05, "loss": 2.6345, "step": 43571 }, { "epoch": 3.42, "learning_rate": 6.583392143025171e-05, "loss": 2.567, "step": 43572 }, { "epoch": 3.42, "learning_rate": 6.583313730102721e-05, "loss": 2.7382, "step": 43573 }, { "epoch": 3.42, "learning_rate": 6.583235317180272e-05, "loss": 2.8767, "step": 43574 }, { "epoch": 3.42, "learning_rate": 6.583156904257822e-05, "loss": 2.827, "step": 43575 }, { "epoch": 3.42, "learning_rate": 6.583078491335373e-05, "loss": 2.5384, "step": 43576 }, { "epoch": 3.42, "learning_rate": 6.583000078412923e-05, "loss": 2.816, "step": 43577 }, { "epoch": 3.42, "learning_rate": 6.582921665490473e-05, "loss": 2.5235, "step": 43578 }, { "epoch": 3.42, "learning_rate": 6.582843252568024e-05, "loss": 2.8188, "step": 43579 }, { "epoch": 3.42, "learning_rate": 6.582764839645574e-05, "loss": 2.5075, "step": 43580 }, { "epoch": 3.42, "learning_rate": 6.582686426723124e-05, "loss": 2.5324, "step": 43581 }, { "epoch": 3.42, "learning_rate": 6.582608013800675e-05, "loss": 2.6604, "step": 43582 }, { "epoch": 3.42, "learning_rate": 6.582529600878225e-05, "loss": 2.5808, "step": 43583 }, { "epoch": 3.42, "learning_rate": 6.582451187955776e-05, "loss": 2.6229, "step": 43584 }, { "epoch": 3.42, "learning_rate": 6.582372775033326e-05, "loss": 2.9107, "step": 43585 }, { "epoch": 3.42, "learning_rate": 6.582294362110876e-05, "loss": 2.5195, "step": 43586 }, { "epoch": 3.42, "learning_rate": 6.582215949188427e-05, "loss": 2.716, "step": 43587 }, { "epoch": 3.42, "learning_rate": 6.582137536265977e-05, "loss": 2.6321, "step": 43588 }, { "epoch": 3.42, "learning_rate": 6.582059123343527e-05, "loss": 2.5856, "step": 43589 }, { "epoch": 3.42, "learning_rate": 6.581980710421078e-05, "loss": 2.5347, "step": 43590 }, { "epoch": 3.42, "learning_rate": 6.581902297498628e-05, "loss": 2.7848, "step": 43591 }, { "epoch": 3.42, "learning_rate": 6.581823884576178e-05, "loss": 2.5425, "step": 43592 }, { "epoch": 3.42, "learning_rate": 6.581745471653729e-05, "loss": 2.7299, "step": 43593 }, { "epoch": 3.42, "learning_rate": 6.581667058731279e-05, "loss": 2.8863, "step": 43594 }, { "epoch": 3.42, "learning_rate": 6.58158864580883e-05, "loss": 2.7589, "step": 43595 }, { "epoch": 3.42, "learning_rate": 6.58151023288638e-05, "loss": 2.7747, "step": 43596 }, { "epoch": 3.42, "learning_rate": 6.58143181996393e-05, "loss": 2.6973, "step": 43597 }, { "epoch": 3.42, "learning_rate": 6.58135340704148e-05, "loss": 2.3356, "step": 43598 }, { "epoch": 3.42, "learning_rate": 6.581274994119031e-05, "loss": 2.6418, "step": 43599 }, { "epoch": 3.42, "learning_rate": 6.581196581196581e-05, "loss": 2.8988, "step": 43600 }, { "epoch": 3.42, "learning_rate": 6.581118168274132e-05, "loss": 2.7173, "step": 43601 }, { "epoch": 3.42, "learning_rate": 6.581039755351682e-05, "loss": 2.6741, "step": 43602 }, { "epoch": 3.42, "learning_rate": 6.580961342429233e-05, "loss": 2.7174, "step": 43603 }, { "epoch": 3.42, "learning_rate": 6.580882929506783e-05, "loss": 2.5622, "step": 43604 }, { "epoch": 3.42, "learning_rate": 6.580804516584333e-05, "loss": 2.7524, "step": 43605 }, { "epoch": 3.42, "learning_rate": 6.580726103661884e-05, "loss": 2.7562, "step": 43606 }, { "epoch": 3.42, "learning_rate": 6.580647690739434e-05, "loss": 2.7159, "step": 43607 }, { "epoch": 3.42, "learning_rate": 6.580569277816984e-05, "loss": 2.8868, "step": 43608 }, { "epoch": 3.42, "learning_rate": 6.580490864894535e-05, "loss": 2.6675, "step": 43609 }, { "epoch": 3.42, "learning_rate": 6.580412451972085e-05, "loss": 2.8083, "step": 43610 }, { "epoch": 3.42, "learning_rate": 6.580334039049635e-05, "loss": 2.6856, "step": 43611 }, { "epoch": 3.42, "learning_rate": 6.580255626127186e-05, "loss": 2.6408, "step": 43612 }, { "epoch": 3.42, "learning_rate": 6.580177213204736e-05, "loss": 2.6456, "step": 43613 }, { "epoch": 3.42, "learning_rate": 6.580098800282287e-05, "loss": 2.6576, "step": 43614 }, { "epoch": 3.42, "learning_rate": 6.580020387359838e-05, "loss": 2.7207, "step": 43615 }, { "epoch": 3.42, "learning_rate": 6.579941974437389e-05, "loss": 2.5618, "step": 43616 }, { "epoch": 3.42, "learning_rate": 6.579863561514939e-05, "loss": 2.7301, "step": 43617 }, { "epoch": 3.42, "learning_rate": 6.57978514859249e-05, "loss": 2.9277, "step": 43618 }, { "epoch": 3.42, "learning_rate": 6.57970673567004e-05, "loss": 2.7379, "step": 43619 }, { "epoch": 3.42, "learning_rate": 6.57962832274759e-05, "loss": 2.6359, "step": 43620 }, { "epoch": 3.42, "learning_rate": 6.579549909825139e-05, "loss": 2.7879, "step": 43621 }, { "epoch": 3.42, "learning_rate": 6.57947149690269e-05, "loss": 2.5513, "step": 43622 }, { "epoch": 3.42, "learning_rate": 6.57939308398024e-05, "loss": 2.7961, "step": 43623 }, { "epoch": 3.42, "learning_rate": 6.57931467105779e-05, "loss": 2.6379, "step": 43624 }, { "epoch": 3.42, "learning_rate": 6.57923625813534e-05, "loss": 2.699, "step": 43625 }, { "epoch": 3.42, "learning_rate": 6.579157845212891e-05, "loss": 2.7986, "step": 43626 }, { "epoch": 3.42, "learning_rate": 6.579079432290441e-05, "loss": 2.6542, "step": 43627 }, { "epoch": 3.42, "learning_rate": 6.579001019367992e-05, "loss": 2.7676, "step": 43628 }, { "epoch": 3.42, "learning_rate": 6.578922606445542e-05, "loss": 2.5062, "step": 43629 }, { "epoch": 3.42, "learning_rate": 6.578844193523092e-05, "loss": 2.7254, "step": 43630 }, { "epoch": 3.42, "learning_rate": 6.578765780600643e-05, "loss": 2.7454, "step": 43631 }, { "epoch": 3.42, "learning_rate": 6.578687367678193e-05, "loss": 2.5501, "step": 43632 }, { "epoch": 3.42, "learning_rate": 6.578608954755744e-05, "loss": 2.8091, "step": 43633 }, { "epoch": 3.42, "learning_rate": 6.578530541833294e-05, "loss": 2.6973, "step": 43634 }, { "epoch": 3.42, "learning_rate": 6.578452128910844e-05, "loss": 2.5659, "step": 43635 }, { "epoch": 3.42, "learning_rate": 6.578373715988395e-05, "loss": 2.7677, "step": 43636 }, { "epoch": 3.42, "learning_rate": 6.578295303065945e-05, "loss": 2.6096, "step": 43637 }, { "epoch": 3.42, "learning_rate": 6.578216890143495e-05, "loss": 2.604, "step": 43638 }, { "epoch": 3.42, "learning_rate": 6.578138477221046e-05, "loss": 2.5693, "step": 43639 }, { "epoch": 3.42, "learning_rate": 6.578060064298596e-05, "loss": 2.7723, "step": 43640 }, { "epoch": 3.42, "learning_rate": 6.577981651376148e-05, "loss": 2.7475, "step": 43641 }, { "epoch": 3.42, "learning_rate": 6.577903238453698e-05, "loss": 2.6688, "step": 43642 }, { "epoch": 3.42, "learning_rate": 6.577824825531249e-05, "loss": 2.3898, "step": 43643 }, { "epoch": 3.42, "learning_rate": 6.577746412608799e-05, "loss": 2.5976, "step": 43644 }, { "epoch": 3.42, "learning_rate": 6.577667999686349e-05, "loss": 2.8481, "step": 43645 }, { "epoch": 3.42, "learning_rate": 6.5775895867639e-05, "loss": 2.5657, "step": 43646 }, { "epoch": 3.42, "learning_rate": 6.57751117384145e-05, "loss": 2.7793, "step": 43647 }, { "epoch": 3.42, "learning_rate": 6.577432760919e-05, "loss": 2.6303, "step": 43648 }, { "epoch": 3.42, "learning_rate": 6.577354347996551e-05, "loss": 2.723, "step": 43649 }, { "epoch": 3.42, "learning_rate": 6.577275935074101e-05, "loss": 2.705, "step": 43650 }, { "epoch": 3.42, "learning_rate": 6.577197522151652e-05, "loss": 2.9059, "step": 43651 }, { "epoch": 3.42, "learning_rate": 6.577119109229202e-05, "loss": 2.7754, "step": 43652 }, { "epoch": 3.42, "learning_rate": 6.577040696306752e-05, "loss": 2.708, "step": 43653 }, { "epoch": 3.42, "learning_rate": 6.576962283384303e-05, "loss": 2.5953, "step": 43654 }, { "epoch": 3.42, "learning_rate": 6.576883870461853e-05, "loss": 2.6608, "step": 43655 }, { "epoch": 3.42, "learning_rate": 6.576805457539403e-05, "loss": 2.6244, "step": 43656 }, { "epoch": 3.42, "learning_rate": 6.576727044616952e-05, "loss": 3.0374, "step": 43657 }, { "epoch": 3.42, "learning_rate": 6.576648631694503e-05, "loss": 2.7657, "step": 43658 }, { "epoch": 3.42, "learning_rate": 6.576570218772053e-05, "loss": 2.8049, "step": 43659 }, { "epoch": 3.42, "learning_rate": 6.576491805849603e-05, "loss": 2.6842, "step": 43660 }, { "epoch": 3.42, "learning_rate": 6.576413392927154e-05, "loss": 2.6638, "step": 43661 }, { "epoch": 3.42, "learning_rate": 6.576334980004704e-05, "loss": 2.7682, "step": 43662 }, { "epoch": 3.42, "learning_rate": 6.576256567082255e-05, "loss": 2.7687, "step": 43663 }, { "epoch": 3.42, "learning_rate": 6.576178154159805e-05, "loss": 2.9683, "step": 43664 }, { "epoch": 3.42, "learning_rate": 6.576099741237355e-05, "loss": 2.6683, "step": 43665 }, { "epoch": 3.42, "learning_rate": 6.576021328314907e-05, "loss": 2.7328, "step": 43666 }, { "epoch": 3.42, "learning_rate": 6.575942915392457e-05, "loss": 2.9244, "step": 43667 }, { "epoch": 3.42, "learning_rate": 6.575864502470008e-05, "loss": 2.7758, "step": 43668 }, { "epoch": 3.42, "learning_rate": 6.575786089547558e-05, "loss": 2.6354, "step": 43669 }, { "epoch": 3.42, "learning_rate": 6.575707676625109e-05, "loss": 2.7452, "step": 43670 }, { "epoch": 3.42, "learning_rate": 6.575629263702659e-05, "loss": 2.6206, "step": 43671 }, { "epoch": 3.42, "learning_rate": 6.575550850780209e-05, "loss": 2.7957, "step": 43672 }, { "epoch": 3.42, "learning_rate": 6.57547243785776e-05, "loss": 2.8214, "step": 43673 }, { "epoch": 3.42, "learning_rate": 6.57539402493531e-05, "loss": 2.9452, "step": 43674 }, { "epoch": 3.42, "learning_rate": 6.57531561201286e-05, "loss": 2.6995, "step": 43675 }, { "epoch": 3.42, "learning_rate": 6.575237199090411e-05, "loss": 2.4091, "step": 43676 }, { "epoch": 3.42, "learning_rate": 6.575158786167961e-05, "loss": 2.769, "step": 43677 }, { "epoch": 3.42, "learning_rate": 6.575080373245511e-05, "loss": 2.782, "step": 43678 }, { "epoch": 3.42, "learning_rate": 6.575001960323062e-05, "loss": 2.7121, "step": 43679 }, { "epoch": 3.43, "learning_rate": 6.574923547400612e-05, "loss": 2.6208, "step": 43680 }, { "epoch": 3.43, "learning_rate": 6.574845134478163e-05, "loss": 2.694, "step": 43681 }, { "epoch": 3.43, "learning_rate": 6.574766721555713e-05, "loss": 2.6778, "step": 43682 }, { "epoch": 3.43, "learning_rate": 6.574688308633263e-05, "loss": 2.7065, "step": 43683 }, { "epoch": 3.43, "learning_rate": 6.574609895710814e-05, "loss": 2.9233, "step": 43684 }, { "epoch": 3.43, "learning_rate": 6.574531482788364e-05, "loss": 2.791, "step": 43685 }, { "epoch": 3.43, "learning_rate": 6.574453069865914e-05, "loss": 2.9283, "step": 43686 }, { "epoch": 3.43, "learning_rate": 6.574374656943465e-05, "loss": 2.6169, "step": 43687 }, { "epoch": 3.43, "learning_rate": 6.574296244021015e-05, "loss": 2.6421, "step": 43688 }, { "epoch": 3.43, "learning_rate": 6.574217831098565e-05, "loss": 2.7402, "step": 43689 }, { "epoch": 3.43, "learning_rate": 6.574139418176116e-05, "loss": 2.8302, "step": 43690 }, { "epoch": 3.43, "learning_rate": 6.574061005253666e-05, "loss": 2.6526, "step": 43691 }, { "epoch": 3.43, "learning_rate": 6.573982592331217e-05, "loss": 2.7568, "step": 43692 }, { "epoch": 3.43, "learning_rate": 6.573904179408767e-05, "loss": 2.7593, "step": 43693 }, { "epoch": 3.43, "learning_rate": 6.573825766486317e-05, "loss": 2.5231, "step": 43694 }, { "epoch": 3.43, "learning_rate": 6.573747353563868e-05, "loss": 2.7558, "step": 43695 }, { "epoch": 3.43, "learning_rate": 6.573668940641418e-05, "loss": 2.6764, "step": 43696 }, { "epoch": 3.43, "learning_rate": 6.573590527718968e-05, "loss": 2.8046, "step": 43697 }, { "epoch": 3.43, "learning_rate": 6.573512114796519e-05, "loss": 2.6892, "step": 43698 }, { "epoch": 3.43, "learning_rate": 6.573433701874069e-05, "loss": 2.6297, "step": 43699 }, { "epoch": 3.43, "learning_rate": 6.57335528895162e-05, "loss": 2.8766, "step": 43700 }, { "epoch": 3.43, "learning_rate": 6.57327687602917e-05, "loss": 2.8359, "step": 43701 }, { "epoch": 3.43, "learning_rate": 6.57319846310672e-05, "loss": 2.6459, "step": 43702 }, { "epoch": 3.43, "learning_rate": 6.57312005018427e-05, "loss": 2.6788, "step": 43703 }, { "epoch": 3.43, "learning_rate": 6.573041637261821e-05, "loss": 2.5763, "step": 43704 }, { "epoch": 3.43, "learning_rate": 6.572963224339371e-05, "loss": 2.6413, "step": 43705 }, { "epoch": 3.43, "learning_rate": 6.572884811416922e-05, "loss": 2.6979, "step": 43706 }, { "epoch": 3.43, "learning_rate": 6.572806398494472e-05, "loss": 2.7184, "step": 43707 }, { "epoch": 3.43, "learning_rate": 6.572727985572022e-05, "loss": 2.6851, "step": 43708 }, { "epoch": 3.43, "learning_rate": 6.572649572649573e-05, "loss": 2.9633, "step": 43709 }, { "epoch": 3.43, "learning_rate": 6.572571159727123e-05, "loss": 2.5831, "step": 43710 }, { "epoch": 3.43, "learning_rate": 6.572492746804674e-05, "loss": 2.7235, "step": 43711 }, { "epoch": 3.43, "learning_rate": 6.572414333882224e-05, "loss": 2.7258, "step": 43712 }, { "epoch": 3.43, "learning_rate": 6.572335920959774e-05, "loss": 2.9978, "step": 43713 }, { "epoch": 3.43, "learning_rate": 6.572257508037325e-05, "loss": 2.8108, "step": 43714 }, { "epoch": 3.43, "learning_rate": 6.572179095114875e-05, "loss": 2.7132, "step": 43715 }, { "epoch": 3.43, "learning_rate": 6.572100682192425e-05, "loss": 2.537, "step": 43716 }, { "epoch": 3.43, "learning_rate": 6.572022269269976e-05, "loss": 2.8825, "step": 43717 }, { "epoch": 3.43, "learning_rate": 6.571943856347526e-05, "loss": 2.7542, "step": 43718 }, { "epoch": 3.43, "learning_rate": 6.571865443425077e-05, "loss": 2.7916, "step": 43719 }, { "epoch": 3.43, "learning_rate": 6.571787030502627e-05, "loss": 2.799, "step": 43720 }, { "epoch": 3.43, "learning_rate": 6.571708617580177e-05, "loss": 2.6258, "step": 43721 }, { "epoch": 3.43, "learning_rate": 6.571630204657728e-05, "loss": 2.8493, "step": 43722 }, { "epoch": 3.43, "learning_rate": 6.57155179173528e-05, "loss": 2.7591, "step": 43723 }, { "epoch": 3.43, "learning_rate": 6.57147337881283e-05, "loss": 2.5886, "step": 43724 }, { "epoch": 3.43, "learning_rate": 6.57139496589038e-05, "loss": 2.6893, "step": 43725 }, { "epoch": 3.43, "learning_rate": 6.57131655296793e-05, "loss": 2.8587, "step": 43726 }, { "epoch": 3.43, "learning_rate": 6.571238140045481e-05, "loss": 2.7614, "step": 43727 }, { "epoch": 3.43, "learning_rate": 6.571159727123031e-05, "loss": 2.6732, "step": 43728 }, { "epoch": 3.43, "learning_rate": 6.57108131420058e-05, "loss": 2.8062, "step": 43729 }, { "epoch": 3.43, "learning_rate": 6.57100290127813e-05, "loss": 2.7411, "step": 43730 }, { "epoch": 3.43, "learning_rate": 6.570924488355681e-05, "loss": 2.9888, "step": 43731 }, { "epoch": 3.43, "learning_rate": 6.570846075433231e-05, "loss": 2.6658, "step": 43732 }, { "epoch": 3.43, "learning_rate": 6.570767662510782e-05, "loss": 2.7932, "step": 43733 }, { "epoch": 3.43, "learning_rate": 6.570689249588332e-05, "loss": 2.8262, "step": 43734 }, { "epoch": 3.43, "learning_rate": 6.570610836665882e-05, "loss": 2.8656, "step": 43735 }, { "epoch": 3.43, "learning_rate": 6.570532423743433e-05, "loss": 2.8245, "step": 43736 }, { "epoch": 3.43, "learning_rate": 6.570454010820983e-05, "loss": 2.6398, "step": 43737 }, { "epoch": 3.43, "learning_rate": 6.570375597898534e-05, "loss": 2.6461, "step": 43738 }, { "epoch": 3.43, "learning_rate": 6.570297184976084e-05, "loss": 2.5484, "step": 43739 }, { "epoch": 3.43, "learning_rate": 6.570218772053634e-05, "loss": 2.7088, "step": 43740 }, { "epoch": 3.43, "learning_rate": 6.570140359131185e-05, "loss": 2.4526, "step": 43741 }, { "epoch": 3.43, "learning_rate": 6.570061946208735e-05, "loss": 2.6375, "step": 43742 }, { "epoch": 3.43, "learning_rate": 6.569983533286285e-05, "loss": 2.7054, "step": 43743 }, { "epoch": 3.43, "learning_rate": 6.569905120363836e-05, "loss": 2.4944, "step": 43744 }, { "epoch": 3.43, "learning_rate": 6.569826707441386e-05, "loss": 2.8575, "step": 43745 }, { "epoch": 3.43, "learning_rate": 6.569748294518936e-05, "loss": 2.8132, "step": 43746 }, { "epoch": 3.43, "learning_rate": 6.569669881596487e-05, "loss": 2.9139, "step": 43747 }, { "epoch": 3.43, "learning_rate": 6.569591468674037e-05, "loss": 2.63, "step": 43748 }, { "epoch": 3.43, "learning_rate": 6.569513055751589e-05, "loss": 2.57, "step": 43749 }, { "epoch": 3.43, "learning_rate": 6.569434642829139e-05, "loss": 2.7578, "step": 43750 }, { "epoch": 3.43, "learning_rate": 6.56935622990669e-05, "loss": 2.9399, "step": 43751 }, { "epoch": 3.43, "learning_rate": 6.56927781698424e-05, "loss": 2.5988, "step": 43752 }, { "epoch": 3.43, "learning_rate": 6.56919940406179e-05, "loss": 2.863, "step": 43753 }, { "epoch": 3.43, "learning_rate": 6.569120991139341e-05, "loss": 2.7943, "step": 43754 }, { "epoch": 3.43, "learning_rate": 6.569042578216891e-05, "loss": 2.9102, "step": 43755 }, { "epoch": 3.43, "learning_rate": 6.568964165294441e-05, "loss": 2.6111, "step": 43756 }, { "epoch": 3.43, "learning_rate": 6.568885752371992e-05, "loss": 2.7518, "step": 43757 }, { "epoch": 3.43, "learning_rate": 6.568807339449542e-05, "loss": 2.9661, "step": 43758 }, { "epoch": 3.43, "learning_rate": 6.568728926527093e-05, "loss": 2.6317, "step": 43759 }, { "epoch": 3.43, "learning_rate": 6.568650513604643e-05, "loss": 2.6155, "step": 43760 }, { "epoch": 3.43, "learning_rate": 6.568572100682193e-05, "loss": 2.7689, "step": 43761 }, { "epoch": 3.43, "learning_rate": 6.568493687759744e-05, "loss": 2.7325, "step": 43762 }, { "epoch": 3.43, "learning_rate": 6.568415274837294e-05, "loss": 2.6627, "step": 43763 }, { "epoch": 3.43, "learning_rate": 6.568336861914844e-05, "loss": 2.8355, "step": 43764 }, { "epoch": 3.43, "learning_rate": 6.568258448992393e-05, "loss": 2.5699, "step": 43765 }, { "epoch": 3.43, "learning_rate": 6.568180036069944e-05, "loss": 2.6471, "step": 43766 }, { "epoch": 3.43, "learning_rate": 6.568101623147494e-05, "loss": 2.5213, "step": 43767 }, { "epoch": 3.43, "learning_rate": 6.568023210225045e-05, "loss": 2.5488, "step": 43768 }, { "epoch": 3.43, "learning_rate": 6.567944797302595e-05, "loss": 2.9108, "step": 43769 }, { "epoch": 3.43, "learning_rate": 6.567866384380145e-05, "loss": 2.6523, "step": 43770 }, { "epoch": 3.43, "learning_rate": 6.567787971457696e-05, "loss": 2.5528, "step": 43771 }, { "epoch": 3.43, "learning_rate": 6.567709558535246e-05, "loss": 2.74, "step": 43772 }, { "epoch": 3.43, "learning_rate": 6.567631145612796e-05, "loss": 2.9473, "step": 43773 }, { "epoch": 3.43, "learning_rate": 6.567552732690347e-05, "loss": 2.7092, "step": 43774 }, { "epoch": 3.43, "learning_rate": 6.567474319767898e-05, "loss": 2.6615, "step": 43775 }, { "epoch": 3.43, "learning_rate": 6.567395906845449e-05, "loss": 2.6812, "step": 43776 }, { "epoch": 3.43, "learning_rate": 6.567317493922999e-05, "loss": 2.5443, "step": 43777 }, { "epoch": 3.43, "learning_rate": 6.56723908100055e-05, "loss": 2.836, "step": 43778 }, { "epoch": 3.43, "learning_rate": 6.5671606680781e-05, "loss": 2.4575, "step": 43779 }, { "epoch": 3.43, "learning_rate": 6.56708225515565e-05, "loss": 3.1389, "step": 43780 }, { "epoch": 3.43, "learning_rate": 6.567003842233201e-05, "loss": 2.7821, "step": 43781 }, { "epoch": 3.43, "learning_rate": 6.566925429310751e-05, "loss": 2.7818, "step": 43782 }, { "epoch": 3.43, "learning_rate": 6.566847016388301e-05, "loss": 2.6687, "step": 43783 }, { "epoch": 3.43, "learning_rate": 6.566768603465852e-05, "loss": 2.6602, "step": 43784 }, { "epoch": 3.43, "learning_rate": 6.566690190543402e-05, "loss": 2.4995, "step": 43785 }, { "epoch": 3.43, "learning_rate": 6.566611777620953e-05, "loss": 3.1499, "step": 43786 }, { "epoch": 3.43, "learning_rate": 6.566533364698503e-05, "loss": 2.7757, "step": 43787 }, { "epoch": 3.43, "learning_rate": 6.566454951776053e-05, "loss": 2.6138, "step": 43788 }, { "epoch": 3.43, "learning_rate": 6.566376538853604e-05, "loss": 2.8726, "step": 43789 }, { "epoch": 3.43, "learning_rate": 6.566298125931154e-05, "loss": 2.6895, "step": 43790 }, { "epoch": 3.43, "learning_rate": 6.566219713008704e-05, "loss": 2.6774, "step": 43791 }, { "epoch": 3.43, "learning_rate": 6.566141300086255e-05, "loss": 2.7801, "step": 43792 }, { "epoch": 3.43, "learning_rate": 6.566062887163805e-05, "loss": 2.6985, "step": 43793 }, { "epoch": 3.43, "learning_rate": 6.565984474241355e-05, "loss": 2.8738, "step": 43794 }, { "epoch": 3.43, "learning_rate": 6.565906061318906e-05, "loss": 2.6802, "step": 43795 }, { "epoch": 3.43, "learning_rate": 6.565827648396456e-05, "loss": 2.5052, "step": 43796 }, { "epoch": 3.43, "learning_rate": 6.565749235474007e-05, "loss": 2.7012, "step": 43797 }, { "epoch": 3.43, "learning_rate": 6.565670822551557e-05, "loss": 2.8918, "step": 43798 }, { "epoch": 3.43, "learning_rate": 6.565592409629107e-05, "loss": 2.6473, "step": 43799 }, { "epoch": 3.43, "learning_rate": 6.565513996706658e-05, "loss": 2.7807, "step": 43800 }, { "epoch": 3.43, "learning_rate": 6.565435583784208e-05, "loss": 2.7786, "step": 43801 }, { "epoch": 3.43, "learning_rate": 6.565357170861758e-05, "loss": 2.7023, "step": 43802 }, { "epoch": 3.43, "learning_rate": 6.565278757939309e-05, "loss": 2.783, "step": 43803 }, { "epoch": 3.43, "learning_rate": 6.565200345016859e-05, "loss": 2.7699, "step": 43804 }, { "epoch": 3.43, "learning_rate": 6.56512193209441e-05, "loss": 2.7209, "step": 43805 }, { "epoch": 3.43, "learning_rate": 6.56504351917196e-05, "loss": 2.8384, "step": 43806 }, { "epoch": 3.44, "learning_rate": 6.56496510624951e-05, "loss": 2.7945, "step": 43807 }, { "epoch": 3.44, "learning_rate": 6.56488669332706e-05, "loss": 2.9534, "step": 43808 }, { "epoch": 3.44, "learning_rate": 6.564808280404611e-05, "loss": 2.8422, "step": 43809 }, { "epoch": 3.44, "learning_rate": 6.564729867482161e-05, "loss": 3.0412, "step": 43810 }, { "epoch": 3.44, "learning_rate": 6.564651454559712e-05, "loss": 2.912, "step": 43811 }, { "epoch": 3.44, "learning_rate": 6.564573041637262e-05, "loss": 2.703, "step": 43812 }, { "epoch": 3.44, "learning_rate": 6.564494628714812e-05, "loss": 2.8319, "step": 43813 }, { "epoch": 3.44, "learning_rate": 6.564416215792363e-05, "loss": 2.6858, "step": 43814 }, { "epoch": 3.44, "learning_rate": 6.564337802869913e-05, "loss": 2.6298, "step": 43815 }, { "epoch": 3.44, "learning_rate": 6.564259389947464e-05, "loss": 2.6822, "step": 43816 }, { "epoch": 3.44, "learning_rate": 6.564180977025014e-05, "loss": 2.6681, "step": 43817 }, { "epoch": 3.44, "learning_rate": 6.564102564102564e-05, "loss": 2.7642, "step": 43818 }, { "epoch": 3.44, "learning_rate": 6.564024151180115e-05, "loss": 2.9259, "step": 43819 }, { "epoch": 3.44, "learning_rate": 6.563945738257665e-05, "loss": 2.6695, "step": 43820 }, { "epoch": 3.44, "learning_rate": 6.563867325335215e-05, "loss": 2.7733, "step": 43821 }, { "epoch": 3.44, "learning_rate": 6.563788912412766e-05, "loss": 2.721, "step": 43822 }, { "epoch": 3.44, "learning_rate": 6.563710499490316e-05, "loss": 2.5688, "step": 43823 }, { "epoch": 3.44, "learning_rate": 6.563632086567866e-05, "loss": 2.7492, "step": 43824 }, { "epoch": 3.44, "learning_rate": 6.563553673645417e-05, "loss": 2.8012, "step": 43825 }, { "epoch": 3.44, "learning_rate": 6.563475260722967e-05, "loss": 2.6312, "step": 43826 }, { "epoch": 3.44, "learning_rate": 6.563396847800518e-05, "loss": 2.6429, "step": 43827 }, { "epoch": 3.44, "learning_rate": 6.563318434878068e-05, "loss": 3.0799, "step": 43828 }, { "epoch": 3.44, "learning_rate": 6.563240021955618e-05, "loss": 2.6479, "step": 43829 }, { "epoch": 3.44, "learning_rate": 6.563161609033169e-05, "loss": 2.5477, "step": 43830 }, { "epoch": 3.44, "learning_rate": 6.56308319611072e-05, "loss": 2.6844, "step": 43831 }, { "epoch": 3.44, "learning_rate": 6.563004783188271e-05, "loss": 3.0974, "step": 43832 }, { "epoch": 3.44, "learning_rate": 6.562926370265821e-05, "loss": 2.8847, "step": 43833 }, { "epoch": 3.44, "learning_rate": 6.562847957343372e-05, "loss": 2.8604, "step": 43834 }, { "epoch": 3.44, "learning_rate": 6.562769544420922e-05, "loss": 2.7277, "step": 43835 }, { "epoch": 3.44, "learning_rate": 6.562691131498472e-05, "loss": 2.9015, "step": 43836 }, { "epoch": 3.44, "learning_rate": 6.562612718576021e-05, "loss": 2.7467, "step": 43837 }, { "epoch": 3.44, "learning_rate": 6.562534305653572e-05, "loss": 3.1226, "step": 43838 }, { "epoch": 3.44, "learning_rate": 6.562455892731122e-05, "loss": 2.7566, "step": 43839 }, { "epoch": 3.44, "learning_rate": 6.562377479808672e-05, "loss": 2.6313, "step": 43840 }, { "epoch": 3.44, "learning_rate": 6.562299066886223e-05, "loss": 2.7708, "step": 43841 }, { "epoch": 3.44, "learning_rate": 6.562220653963773e-05, "loss": 2.7904, "step": 43842 }, { "epoch": 3.44, "learning_rate": 6.562142241041323e-05, "loss": 2.8607, "step": 43843 }, { "epoch": 3.44, "learning_rate": 6.562063828118874e-05, "loss": 2.7555, "step": 43844 }, { "epoch": 3.44, "learning_rate": 6.561985415196424e-05, "loss": 2.4643, "step": 43845 }, { "epoch": 3.44, "learning_rate": 6.561907002273975e-05, "loss": 2.6925, "step": 43846 }, { "epoch": 3.44, "learning_rate": 6.561828589351525e-05, "loss": 3.014, "step": 43847 }, { "epoch": 3.44, "learning_rate": 6.561750176429075e-05, "loss": 2.7903, "step": 43848 }, { "epoch": 3.44, "learning_rate": 6.561671763506626e-05, "loss": 2.5395, "step": 43849 }, { "epoch": 3.44, "learning_rate": 6.561593350584176e-05, "loss": 2.6556, "step": 43850 }, { "epoch": 3.44, "learning_rate": 6.561514937661726e-05, "loss": 2.7377, "step": 43851 }, { "epoch": 3.44, "learning_rate": 6.561436524739277e-05, "loss": 2.6601, "step": 43852 }, { "epoch": 3.44, "learning_rate": 6.561358111816827e-05, "loss": 2.7977, "step": 43853 }, { "epoch": 3.44, "learning_rate": 6.561279698894378e-05, "loss": 2.764, "step": 43854 }, { "epoch": 3.44, "learning_rate": 6.561201285971928e-05, "loss": 2.5572, "step": 43855 }, { "epoch": 3.44, "learning_rate": 6.561122873049478e-05, "loss": 2.555, "step": 43856 }, { "epoch": 3.44, "learning_rate": 6.56104446012703e-05, "loss": 2.7854, "step": 43857 }, { "epoch": 3.44, "learning_rate": 6.56096604720458e-05, "loss": 2.7256, "step": 43858 }, { "epoch": 3.44, "learning_rate": 6.560887634282131e-05, "loss": 2.5633, "step": 43859 }, { "epoch": 3.44, "learning_rate": 6.560809221359681e-05, "loss": 2.7931, "step": 43860 }, { "epoch": 3.44, "learning_rate": 6.560730808437231e-05, "loss": 2.6815, "step": 43861 }, { "epoch": 3.44, "learning_rate": 6.560652395514782e-05, "loss": 2.7554, "step": 43862 }, { "epoch": 3.44, "learning_rate": 6.560573982592332e-05, "loss": 2.7274, "step": 43863 }, { "epoch": 3.44, "learning_rate": 6.560495569669883e-05, "loss": 2.936, "step": 43864 }, { "epoch": 3.44, "learning_rate": 6.560417156747433e-05, "loss": 2.8022, "step": 43865 }, { "epoch": 3.44, "learning_rate": 6.560338743824983e-05, "loss": 2.8969, "step": 43866 }, { "epoch": 3.44, "learning_rate": 6.560260330902534e-05, "loss": 2.7351, "step": 43867 }, { "epoch": 3.44, "learning_rate": 6.560181917980084e-05, "loss": 2.546, "step": 43868 }, { "epoch": 3.44, "learning_rate": 6.560103505057634e-05, "loss": 2.6747, "step": 43869 }, { "epoch": 3.44, "learning_rate": 6.560025092135185e-05, "loss": 2.8273, "step": 43870 }, { "epoch": 3.44, "learning_rate": 6.559946679212735e-05, "loss": 2.6213, "step": 43871 }, { "epoch": 3.44, "learning_rate": 6.559868266290286e-05, "loss": 2.7581, "step": 43872 }, { "epoch": 3.44, "learning_rate": 6.559789853367835e-05, "loss": 2.7604, "step": 43873 }, { "epoch": 3.44, "learning_rate": 6.559711440445385e-05, "loss": 2.6084, "step": 43874 }, { "epoch": 3.44, "learning_rate": 6.559633027522935e-05, "loss": 2.6116, "step": 43875 }, { "epoch": 3.44, "learning_rate": 6.559554614600486e-05, "loss": 2.7252, "step": 43876 }, { "epoch": 3.44, "learning_rate": 6.559476201678036e-05, "loss": 2.6574, "step": 43877 }, { "epoch": 3.44, "learning_rate": 6.559397788755586e-05, "loss": 2.6147, "step": 43878 }, { "epoch": 3.44, "learning_rate": 6.559319375833137e-05, "loss": 2.6962, "step": 43879 }, { "epoch": 3.44, "learning_rate": 6.559240962910687e-05, "loss": 2.6873, "step": 43880 }, { "epoch": 3.44, "learning_rate": 6.559162549988237e-05, "loss": 2.7204, "step": 43881 }, { "epoch": 3.44, "learning_rate": 6.559084137065788e-05, "loss": 2.83, "step": 43882 }, { "epoch": 3.44, "learning_rate": 6.55900572414334e-05, "loss": 2.8734, "step": 43883 }, { "epoch": 3.44, "learning_rate": 6.55892731122089e-05, "loss": 2.6446, "step": 43884 }, { "epoch": 3.44, "learning_rate": 6.55884889829844e-05, "loss": 2.6005, "step": 43885 }, { "epoch": 3.44, "learning_rate": 6.55877048537599e-05, "loss": 2.7083, "step": 43886 }, { "epoch": 3.44, "learning_rate": 6.558692072453541e-05, "loss": 2.7286, "step": 43887 }, { "epoch": 3.44, "learning_rate": 6.558613659531091e-05, "loss": 2.8934, "step": 43888 }, { "epoch": 3.44, "learning_rate": 6.558535246608642e-05, "loss": 2.4841, "step": 43889 }, { "epoch": 3.44, "learning_rate": 6.558456833686192e-05, "loss": 2.7524, "step": 43890 }, { "epoch": 3.44, "learning_rate": 6.558378420763742e-05, "loss": 2.6408, "step": 43891 }, { "epoch": 3.44, "learning_rate": 6.558300007841293e-05, "loss": 3.0157, "step": 43892 }, { "epoch": 3.44, "learning_rate": 6.558221594918843e-05, "loss": 2.9309, "step": 43893 }, { "epoch": 3.44, "learning_rate": 6.558143181996394e-05, "loss": 2.8903, "step": 43894 }, { "epoch": 3.44, "learning_rate": 6.558064769073944e-05, "loss": 2.8214, "step": 43895 }, { "epoch": 3.44, "learning_rate": 6.557986356151494e-05, "loss": 2.6055, "step": 43896 }, { "epoch": 3.44, "learning_rate": 6.557907943229045e-05, "loss": 2.7351, "step": 43897 }, { "epoch": 3.44, "learning_rate": 6.557829530306595e-05, "loss": 2.8282, "step": 43898 }, { "epoch": 3.44, "learning_rate": 6.557751117384145e-05, "loss": 2.9283, "step": 43899 }, { "epoch": 3.44, "learning_rate": 6.557672704461696e-05, "loss": 2.6024, "step": 43900 }, { "epoch": 3.44, "learning_rate": 6.557594291539246e-05, "loss": 2.9095, "step": 43901 }, { "epoch": 3.44, "learning_rate": 6.557515878616797e-05, "loss": 2.648, "step": 43902 }, { "epoch": 3.44, "learning_rate": 6.557437465694347e-05, "loss": 2.741, "step": 43903 }, { "epoch": 3.44, "learning_rate": 6.557359052771897e-05, "loss": 2.8011, "step": 43904 }, { "epoch": 3.44, "learning_rate": 6.557280639849448e-05, "loss": 2.9529, "step": 43905 }, { "epoch": 3.44, "learning_rate": 6.557202226926998e-05, "loss": 2.7362, "step": 43906 }, { "epoch": 3.44, "learning_rate": 6.557123814004548e-05, "loss": 2.7804, "step": 43907 }, { "epoch": 3.44, "learning_rate": 6.557045401082099e-05, "loss": 2.7496, "step": 43908 }, { "epoch": 3.44, "learning_rate": 6.556966988159649e-05, "loss": 2.5045, "step": 43909 }, { "epoch": 3.44, "learning_rate": 6.5568885752372e-05, "loss": 2.7401, "step": 43910 }, { "epoch": 3.44, "learning_rate": 6.55681016231475e-05, "loss": 2.5736, "step": 43911 }, { "epoch": 3.44, "learning_rate": 6.5567317493923e-05, "loss": 2.563, "step": 43912 }, { "epoch": 3.44, "learning_rate": 6.55665333646985e-05, "loss": 2.7969, "step": 43913 }, { "epoch": 3.44, "learning_rate": 6.556574923547401e-05, "loss": 2.7567, "step": 43914 }, { "epoch": 3.44, "learning_rate": 6.556496510624951e-05, "loss": 2.753, "step": 43915 }, { "epoch": 3.44, "learning_rate": 6.556418097702502e-05, "loss": 2.865, "step": 43916 }, { "epoch": 3.44, "learning_rate": 6.556339684780052e-05, "loss": 2.9555, "step": 43917 }, { "epoch": 3.44, "learning_rate": 6.556261271857602e-05, "loss": 2.7697, "step": 43918 }, { "epoch": 3.44, "learning_rate": 6.556182858935153e-05, "loss": 2.891, "step": 43919 }, { "epoch": 3.44, "learning_rate": 6.556104446012703e-05, "loss": 2.6364, "step": 43920 }, { "epoch": 3.44, "learning_rate": 6.556026033090254e-05, "loss": 2.7628, "step": 43921 }, { "epoch": 3.44, "learning_rate": 6.555947620167804e-05, "loss": 2.7968, "step": 43922 }, { "epoch": 3.44, "learning_rate": 6.555869207245354e-05, "loss": 2.9177, "step": 43923 }, { "epoch": 3.44, "learning_rate": 6.555790794322905e-05, "loss": 2.9973, "step": 43924 }, { "epoch": 3.44, "learning_rate": 6.555712381400455e-05, "loss": 2.72, "step": 43925 }, { "epoch": 3.44, "learning_rate": 6.555633968478005e-05, "loss": 2.7858, "step": 43926 }, { "epoch": 3.44, "learning_rate": 6.555555555555556e-05, "loss": 2.7268, "step": 43927 }, { "epoch": 3.44, "learning_rate": 6.555477142633106e-05, "loss": 3.1326, "step": 43928 }, { "epoch": 3.44, "learning_rate": 6.555398729710656e-05, "loss": 2.5146, "step": 43929 }, { "epoch": 3.44, "learning_rate": 6.555320316788207e-05, "loss": 2.7394, "step": 43930 }, { "epoch": 3.44, "learning_rate": 6.555241903865757e-05, "loss": 2.8629, "step": 43931 }, { "epoch": 3.44, "learning_rate": 6.555163490943308e-05, "loss": 2.6875, "step": 43932 }, { "epoch": 3.44, "learning_rate": 6.555085078020858e-05, "loss": 2.7403, "step": 43933 }, { "epoch": 3.44, "learning_rate": 6.555006665098408e-05, "loss": 2.8055, "step": 43934 }, { "epoch": 3.45, "learning_rate": 6.554928252175959e-05, "loss": 2.6164, "step": 43935 }, { "epoch": 3.45, "learning_rate": 6.554849839253509e-05, "loss": 2.5665, "step": 43936 }, { "epoch": 3.45, "learning_rate": 6.55477142633106e-05, "loss": 2.7061, "step": 43937 }, { "epoch": 3.45, "learning_rate": 6.55469301340861e-05, "loss": 2.6652, "step": 43938 }, { "epoch": 3.45, "learning_rate": 6.554614600486162e-05, "loss": 2.5912, "step": 43939 }, { "epoch": 3.45, "learning_rate": 6.554536187563712e-05, "loss": 2.6243, "step": 43940 }, { "epoch": 3.45, "learning_rate": 6.554457774641262e-05, "loss": 2.758, "step": 43941 }, { "epoch": 3.45, "learning_rate": 6.554379361718813e-05, "loss": 2.7161, "step": 43942 }, { "epoch": 3.45, "learning_rate": 6.554300948796363e-05, "loss": 2.7599, "step": 43943 }, { "epoch": 3.45, "learning_rate": 6.554222535873913e-05, "loss": 2.7883, "step": 43944 }, { "epoch": 3.45, "learning_rate": 6.554144122951462e-05, "loss": 2.7123, "step": 43945 }, { "epoch": 3.45, "learning_rate": 6.554065710029013e-05, "loss": 2.9451, "step": 43946 }, { "epoch": 3.45, "learning_rate": 6.553987297106563e-05, "loss": 2.6087, "step": 43947 }, { "epoch": 3.45, "learning_rate": 6.553908884184113e-05, "loss": 2.8775, "step": 43948 }, { "epoch": 3.45, "learning_rate": 6.553830471261664e-05, "loss": 2.8545, "step": 43949 }, { "epoch": 3.45, "learning_rate": 6.553752058339214e-05, "loss": 2.6451, "step": 43950 }, { "epoch": 3.45, "learning_rate": 6.553673645416765e-05, "loss": 2.9042, "step": 43951 }, { "epoch": 3.45, "learning_rate": 6.553595232494315e-05, "loss": 2.7423, "step": 43952 }, { "epoch": 3.45, "learning_rate": 6.553516819571865e-05, "loss": 2.3817, "step": 43953 }, { "epoch": 3.45, "learning_rate": 6.553438406649416e-05, "loss": 2.8822, "step": 43954 }, { "epoch": 3.45, "learning_rate": 6.553359993726966e-05, "loss": 2.6299, "step": 43955 }, { "epoch": 3.45, "learning_rate": 6.553281580804516e-05, "loss": 2.6574, "step": 43956 }, { "epoch": 3.45, "learning_rate": 6.553203167882067e-05, "loss": 2.4244, "step": 43957 }, { "epoch": 3.45, "learning_rate": 6.553124754959617e-05, "loss": 2.6728, "step": 43958 }, { "epoch": 3.45, "learning_rate": 6.553046342037168e-05, "loss": 2.5479, "step": 43959 }, { "epoch": 3.45, "learning_rate": 6.552967929114718e-05, "loss": 2.8726, "step": 43960 }, { "epoch": 3.45, "learning_rate": 6.552889516192268e-05, "loss": 2.9726, "step": 43961 }, { "epoch": 3.45, "learning_rate": 6.552811103269819e-05, "loss": 2.8046, "step": 43962 }, { "epoch": 3.45, "learning_rate": 6.552732690347369e-05, "loss": 2.8077, "step": 43963 }, { "epoch": 3.45, "learning_rate": 6.55265427742492e-05, "loss": 2.7939, "step": 43964 }, { "epoch": 3.45, "learning_rate": 6.552575864502471e-05, "loss": 2.7867, "step": 43965 }, { "epoch": 3.45, "learning_rate": 6.552497451580021e-05, "loss": 2.6359, "step": 43966 }, { "epoch": 3.45, "learning_rate": 6.552419038657572e-05, "loss": 2.6621, "step": 43967 }, { "epoch": 3.45, "learning_rate": 6.552340625735122e-05, "loss": 2.5344, "step": 43968 }, { "epoch": 3.45, "learning_rate": 6.552262212812673e-05, "loss": 2.9476, "step": 43969 }, { "epoch": 3.45, "learning_rate": 6.552183799890223e-05, "loss": 2.6328, "step": 43970 }, { "epoch": 3.45, "learning_rate": 6.552105386967773e-05, "loss": 2.6551, "step": 43971 }, { "epoch": 3.45, "learning_rate": 6.552026974045324e-05, "loss": 2.7418, "step": 43972 }, { "epoch": 3.45, "learning_rate": 6.551948561122874e-05, "loss": 2.7986, "step": 43973 }, { "epoch": 3.45, "learning_rate": 6.551870148200424e-05, "loss": 2.8749, "step": 43974 }, { "epoch": 3.45, "learning_rate": 6.551791735277975e-05, "loss": 2.7436, "step": 43975 }, { "epoch": 3.45, "learning_rate": 6.551713322355525e-05, "loss": 2.6336, "step": 43976 }, { "epoch": 3.45, "learning_rate": 6.551634909433075e-05, "loss": 2.7422, "step": 43977 }, { "epoch": 3.45, "learning_rate": 6.551556496510626e-05, "loss": 2.7948, "step": 43978 }, { "epoch": 3.45, "learning_rate": 6.551478083588176e-05, "loss": 2.6796, "step": 43979 }, { "epoch": 3.45, "learning_rate": 6.551399670665727e-05, "loss": 2.7446, "step": 43980 }, { "epoch": 3.45, "learning_rate": 6.551321257743276e-05, "loss": 2.7914, "step": 43981 }, { "epoch": 3.45, "learning_rate": 6.551242844820826e-05, "loss": 2.801, "step": 43982 }, { "epoch": 3.45, "learning_rate": 6.551164431898376e-05, "loss": 2.7973, "step": 43983 }, { "epoch": 3.45, "learning_rate": 6.551086018975927e-05, "loss": 2.6717, "step": 43984 }, { "epoch": 3.45, "learning_rate": 6.551007606053477e-05, "loss": 2.6371, "step": 43985 }, { "epoch": 3.45, "learning_rate": 6.550929193131027e-05, "loss": 2.8153, "step": 43986 }, { "epoch": 3.45, "learning_rate": 6.550850780208578e-05, "loss": 2.6692, "step": 43987 }, { "epoch": 3.45, "learning_rate": 6.550772367286128e-05, "loss": 2.8717, "step": 43988 }, { "epoch": 3.45, "learning_rate": 6.550693954363679e-05, "loss": 2.6926, "step": 43989 }, { "epoch": 3.45, "learning_rate": 6.550615541441229e-05, "loss": 2.724, "step": 43990 }, { "epoch": 3.45, "learning_rate": 6.55053712851878e-05, "loss": 2.6453, "step": 43991 }, { "epoch": 3.45, "learning_rate": 6.550458715596331e-05, "loss": 2.7173, "step": 43992 }, { "epoch": 3.45, "learning_rate": 6.550380302673881e-05, "loss": 2.6716, "step": 43993 }, { "epoch": 3.45, "learning_rate": 6.550301889751432e-05, "loss": 2.8146, "step": 43994 }, { "epoch": 3.45, "learning_rate": 6.550223476828982e-05, "loss": 2.6705, "step": 43995 }, { "epoch": 3.45, "learning_rate": 6.550145063906532e-05, "loss": 2.5993, "step": 43996 }, { "epoch": 3.45, "learning_rate": 6.550066650984083e-05, "loss": 2.5448, "step": 43997 }, { "epoch": 3.45, "learning_rate": 6.549988238061633e-05, "loss": 2.6111, "step": 43998 }, { "epoch": 3.45, "learning_rate": 6.549909825139184e-05, "loss": 2.7249, "step": 43999 }, { "epoch": 3.45, "learning_rate": 6.549831412216734e-05, "loss": 2.6462, "step": 44000 }, { "epoch": 3.45, "eval_loss": 2.8957736492156982, "eval_runtime": 217.9759, "eval_samples_per_second": 51.987, "eval_steps_per_second": 3.253, "step": 44000 }, { "epoch": 3.45, "learning_rate": 6.549752999294284e-05, "loss": 2.718, "step": 44001 }, { "epoch": 3.45, "learning_rate": 6.549674586371835e-05, "loss": 2.8487, "step": 44002 }, { "epoch": 3.45, "learning_rate": 6.549596173449385e-05, "loss": 2.8122, "step": 44003 }, { "epoch": 3.45, "learning_rate": 6.549517760526935e-05, "loss": 2.6683, "step": 44004 }, { "epoch": 3.45, "learning_rate": 6.549439347604486e-05, "loss": 2.6939, "step": 44005 }, { "epoch": 3.45, "learning_rate": 6.549360934682036e-05, "loss": 2.7082, "step": 44006 }, { "epoch": 3.45, "learning_rate": 6.549282521759587e-05, "loss": 2.7184, "step": 44007 }, { "epoch": 3.45, "learning_rate": 6.549204108837137e-05, "loss": 2.6989, "step": 44008 }, { "epoch": 3.45, "learning_rate": 6.549125695914687e-05, "loss": 2.7301, "step": 44009 }, { "epoch": 3.45, "learning_rate": 6.549047282992238e-05, "loss": 2.9248, "step": 44010 }, { "epoch": 3.45, "learning_rate": 6.548968870069788e-05, "loss": 2.8166, "step": 44011 }, { "epoch": 3.45, "learning_rate": 6.548890457147338e-05, "loss": 2.7157, "step": 44012 }, { "epoch": 3.45, "learning_rate": 6.548812044224889e-05, "loss": 2.7511, "step": 44013 }, { "epoch": 3.45, "learning_rate": 6.548733631302439e-05, "loss": 2.7626, "step": 44014 }, { "epoch": 3.45, "learning_rate": 6.54865521837999e-05, "loss": 2.6771, "step": 44015 }, { "epoch": 3.45, "learning_rate": 6.54857680545754e-05, "loss": 2.7571, "step": 44016 }, { "epoch": 3.45, "learning_rate": 6.54849839253509e-05, "loss": 2.8185, "step": 44017 }, { "epoch": 3.45, "learning_rate": 6.54841997961264e-05, "loss": 2.7824, "step": 44018 }, { "epoch": 3.45, "learning_rate": 6.548341566690191e-05, "loss": 2.8074, "step": 44019 }, { "epoch": 3.45, "learning_rate": 6.548263153767741e-05, "loss": 2.8495, "step": 44020 }, { "epoch": 3.45, "learning_rate": 6.548184740845292e-05, "loss": 2.6586, "step": 44021 }, { "epoch": 3.45, "learning_rate": 6.548106327922842e-05, "loss": 2.7554, "step": 44022 }, { "epoch": 3.45, "learning_rate": 6.548027915000392e-05, "loss": 2.7839, "step": 44023 }, { "epoch": 3.45, "learning_rate": 6.547949502077943e-05, "loss": 2.8028, "step": 44024 }, { "epoch": 3.45, "learning_rate": 6.547871089155493e-05, "loss": 2.5625, "step": 44025 }, { "epoch": 3.45, "learning_rate": 6.547792676233043e-05, "loss": 2.5701, "step": 44026 }, { "epoch": 3.45, "learning_rate": 6.547714263310594e-05, "loss": 2.8638, "step": 44027 }, { "epoch": 3.45, "learning_rate": 6.547635850388144e-05, "loss": 2.4954, "step": 44028 }, { "epoch": 3.45, "learning_rate": 6.547557437465695e-05, "loss": 3.0957, "step": 44029 }, { "epoch": 3.45, "learning_rate": 6.547479024543245e-05, "loss": 2.7362, "step": 44030 }, { "epoch": 3.45, "learning_rate": 6.547400611620795e-05, "loss": 2.8905, "step": 44031 }, { "epoch": 3.45, "learning_rate": 6.547322198698346e-05, "loss": 2.747, "step": 44032 }, { "epoch": 3.45, "learning_rate": 6.547243785775896e-05, "loss": 2.5703, "step": 44033 }, { "epoch": 3.45, "learning_rate": 6.547165372853446e-05, "loss": 2.5266, "step": 44034 }, { "epoch": 3.45, "learning_rate": 6.547086959930997e-05, "loss": 2.7671, "step": 44035 }, { "epoch": 3.45, "learning_rate": 6.547008547008547e-05, "loss": 2.6829, "step": 44036 }, { "epoch": 3.45, "learning_rate": 6.546930134086098e-05, "loss": 2.8744, "step": 44037 }, { "epoch": 3.45, "learning_rate": 6.546851721163648e-05, "loss": 2.7342, "step": 44038 }, { "epoch": 3.45, "learning_rate": 6.546773308241198e-05, "loss": 2.5687, "step": 44039 }, { "epoch": 3.45, "learning_rate": 6.546694895318749e-05, "loss": 2.9548, "step": 44040 }, { "epoch": 3.45, "learning_rate": 6.546616482396299e-05, "loss": 2.6411, "step": 44041 }, { "epoch": 3.45, "learning_rate": 6.54653806947385e-05, "loss": 2.798, "step": 44042 }, { "epoch": 3.45, "learning_rate": 6.5464596565514e-05, "loss": 2.4765, "step": 44043 }, { "epoch": 3.45, "learning_rate": 6.54638124362895e-05, "loss": 2.6404, "step": 44044 }, { "epoch": 3.45, "learning_rate": 6.5463028307065e-05, "loss": 2.4941, "step": 44045 }, { "epoch": 3.45, "learning_rate": 6.546224417784051e-05, "loss": 2.7731, "step": 44046 }, { "epoch": 3.45, "learning_rate": 6.546146004861603e-05, "loss": 2.7252, "step": 44047 }, { "epoch": 3.45, "learning_rate": 6.546067591939153e-05, "loss": 2.831, "step": 44048 }, { "epoch": 3.45, "learning_rate": 6.545989179016703e-05, "loss": 2.6935, "step": 44049 }, { "epoch": 3.45, "learning_rate": 6.545910766094254e-05, "loss": 2.7173, "step": 44050 }, { "epoch": 3.45, "learning_rate": 6.545832353171804e-05, "loss": 2.7124, "step": 44051 }, { "epoch": 3.45, "learning_rate": 6.545753940249354e-05, "loss": 2.7064, "step": 44052 }, { "epoch": 3.45, "learning_rate": 6.545675527326903e-05, "loss": 2.7254, "step": 44053 }, { "epoch": 3.45, "learning_rate": 6.545597114404454e-05, "loss": 2.8168, "step": 44054 }, { "epoch": 3.45, "learning_rate": 6.545518701482004e-05, "loss": 2.7281, "step": 44055 }, { "epoch": 3.45, "learning_rate": 6.545440288559555e-05, "loss": 2.6093, "step": 44056 }, { "epoch": 3.45, "learning_rate": 6.545361875637105e-05, "loss": 2.7439, "step": 44057 }, { "epoch": 3.45, "learning_rate": 6.545283462714655e-05, "loss": 2.5629, "step": 44058 }, { "epoch": 3.45, "learning_rate": 6.545205049792206e-05, "loss": 2.7475, "step": 44059 }, { "epoch": 3.45, "learning_rate": 6.545126636869756e-05, "loss": 2.7889, "step": 44060 }, { "epoch": 3.45, "learning_rate": 6.545048223947306e-05, "loss": 2.7381, "step": 44061 }, { "epoch": 3.46, "learning_rate": 6.544969811024857e-05, "loss": 2.7451, "step": 44062 }, { "epoch": 3.46, "learning_rate": 6.544891398102407e-05, "loss": 2.569, "step": 44063 }, { "epoch": 3.46, "learning_rate": 6.544812985179957e-05, "loss": 2.6657, "step": 44064 }, { "epoch": 3.46, "learning_rate": 6.544734572257508e-05, "loss": 2.7295, "step": 44065 }, { "epoch": 3.46, "learning_rate": 6.544656159335058e-05, "loss": 2.7356, "step": 44066 }, { "epoch": 3.46, "learning_rate": 6.544577746412609e-05, "loss": 2.6775, "step": 44067 }, { "epoch": 3.46, "learning_rate": 6.544499333490159e-05, "loss": 2.8602, "step": 44068 }, { "epoch": 3.46, "learning_rate": 6.544420920567709e-05, "loss": 2.6825, "step": 44069 }, { "epoch": 3.46, "learning_rate": 6.54434250764526e-05, "loss": 2.6747, "step": 44070 }, { "epoch": 3.46, "learning_rate": 6.54426409472281e-05, "loss": 2.6882, "step": 44071 }, { "epoch": 3.46, "learning_rate": 6.54418568180036e-05, "loss": 2.7751, "step": 44072 }, { "epoch": 3.46, "learning_rate": 6.544107268877912e-05, "loss": 2.7234, "step": 44073 }, { "epoch": 3.46, "learning_rate": 6.544028855955463e-05, "loss": 2.8271, "step": 44074 }, { "epoch": 3.46, "learning_rate": 6.543950443033013e-05, "loss": 2.7357, "step": 44075 }, { "epoch": 3.46, "learning_rate": 6.543872030110563e-05, "loss": 2.7906, "step": 44076 }, { "epoch": 3.46, "learning_rate": 6.543793617188114e-05, "loss": 2.7382, "step": 44077 }, { "epoch": 3.46, "learning_rate": 6.543715204265664e-05, "loss": 2.5671, "step": 44078 }, { "epoch": 3.46, "learning_rate": 6.543636791343214e-05, "loss": 2.9366, "step": 44079 }, { "epoch": 3.46, "learning_rate": 6.543558378420765e-05, "loss": 2.6537, "step": 44080 }, { "epoch": 3.46, "learning_rate": 6.543479965498315e-05, "loss": 2.851, "step": 44081 }, { "epoch": 3.46, "learning_rate": 6.543401552575865e-05, "loss": 2.8782, "step": 44082 }, { "epoch": 3.46, "learning_rate": 6.543323139653416e-05, "loss": 2.6003, "step": 44083 }, { "epoch": 3.46, "learning_rate": 6.543244726730966e-05, "loss": 2.5717, "step": 44084 }, { "epoch": 3.46, "learning_rate": 6.543166313808517e-05, "loss": 2.8994, "step": 44085 }, { "epoch": 3.46, "learning_rate": 6.543087900886067e-05, "loss": 2.8679, "step": 44086 }, { "epoch": 3.46, "learning_rate": 6.543009487963617e-05, "loss": 2.7293, "step": 44087 }, { "epoch": 3.46, "learning_rate": 6.542931075041168e-05, "loss": 2.6818, "step": 44088 }, { "epoch": 3.46, "learning_rate": 6.542852662118717e-05, "loss": 2.9955, "step": 44089 }, { "epoch": 3.46, "learning_rate": 6.542774249196267e-05, "loss": 2.7796, "step": 44090 }, { "epoch": 3.46, "learning_rate": 6.542695836273817e-05, "loss": 2.3407, "step": 44091 }, { "epoch": 3.46, "learning_rate": 6.542617423351368e-05, "loss": 2.7676, "step": 44092 }, { "epoch": 3.46, "learning_rate": 6.542539010428918e-05, "loss": 2.6707, "step": 44093 }, { "epoch": 3.46, "learning_rate": 6.542460597506469e-05, "loss": 2.9701, "step": 44094 }, { "epoch": 3.46, "learning_rate": 6.542382184584019e-05, "loss": 2.7859, "step": 44095 }, { "epoch": 3.46, "learning_rate": 6.542303771661569e-05, "loss": 2.5211, "step": 44096 }, { "epoch": 3.46, "learning_rate": 6.54222535873912e-05, "loss": 2.7915, "step": 44097 }, { "epoch": 3.46, "learning_rate": 6.54214694581667e-05, "loss": 2.6351, "step": 44098 }, { "epoch": 3.46, "learning_rate": 6.542068532894222e-05, "loss": 2.7131, "step": 44099 }, { "epoch": 3.46, "learning_rate": 6.541990119971772e-05, "loss": 2.6587, "step": 44100 }, { "epoch": 3.46, "learning_rate": 6.541911707049322e-05, "loss": 3.0085, "step": 44101 }, { "epoch": 3.46, "learning_rate": 6.541833294126873e-05, "loss": 2.8169, "step": 44102 }, { "epoch": 3.46, "learning_rate": 6.541754881204423e-05, "loss": 2.492, "step": 44103 }, { "epoch": 3.46, "learning_rate": 6.541676468281974e-05, "loss": 3.1859, "step": 44104 }, { "epoch": 3.46, "learning_rate": 6.541598055359524e-05, "loss": 2.5603, "step": 44105 }, { "epoch": 3.46, "learning_rate": 6.541519642437074e-05, "loss": 2.8608, "step": 44106 }, { "epoch": 3.46, "learning_rate": 6.541441229514625e-05, "loss": 2.5366, "step": 44107 }, { "epoch": 3.46, "learning_rate": 6.541362816592175e-05, "loss": 2.7007, "step": 44108 }, { "epoch": 3.46, "learning_rate": 6.541284403669725e-05, "loss": 2.7611, "step": 44109 }, { "epoch": 3.46, "learning_rate": 6.541205990747276e-05, "loss": 2.6382, "step": 44110 }, { "epoch": 3.46, "learning_rate": 6.541127577824826e-05, "loss": 2.6742, "step": 44111 }, { "epoch": 3.46, "learning_rate": 6.541049164902376e-05, "loss": 2.6514, "step": 44112 }, { "epoch": 3.46, "learning_rate": 6.540970751979927e-05, "loss": 2.895, "step": 44113 }, { "epoch": 3.46, "learning_rate": 6.540892339057477e-05, "loss": 2.8355, "step": 44114 }, { "epoch": 3.46, "learning_rate": 6.540813926135028e-05, "loss": 2.5901, "step": 44115 }, { "epoch": 3.46, "learning_rate": 6.540735513212578e-05, "loss": 2.5453, "step": 44116 }, { "epoch": 3.46, "learning_rate": 6.540657100290128e-05, "loss": 2.6855, "step": 44117 }, { "epoch": 3.46, "learning_rate": 6.540578687367679e-05, "loss": 2.6862, "step": 44118 }, { "epoch": 3.46, "learning_rate": 6.540500274445229e-05, "loss": 2.7388, "step": 44119 }, { "epoch": 3.46, "learning_rate": 6.54042186152278e-05, "loss": 2.6833, "step": 44120 }, { "epoch": 3.46, "learning_rate": 6.54034344860033e-05, "loss": 2.662, "step": 44121 }, { "epoch": 3.46, "learning_rate": 6.54026503567788e-05, "loss": 2.6204, "step": 44122 }, { "epoch": 3.46, "learning_rate": 6.54018662275543e-05, "loss": 2.6468, "step": 44123 }, { "epoch": 3.46, "learning_rate": 6.540108209832981e-05, "loss": 2.6026, "step": 44124 }, { "epoch": 3.46, "learning_rate": 6.540029796910531e-05, "loss": 2.7242, "step": 44125 }, { "epoch": 3.46, "learning_rate": 6.539951383988082e-05, "loss": 2.5446, "step": 44126 }, { "epoch": 3.46, "learning_rate": 6.539872971065632e-05, "loss": 2.707, "step": 44127 }, { "epoch": 3.46, "learning_rate": 6.539794558143182e-05, "loss": 2.6036, "step": 44128 }, { "epoch": 3.46, "learning_rate": 6.539716145220733e-05, "loss": 2.6492, "step": 44129 }, { "epoch": 3.46, "learning_rate": 6.539637732298283e-05, "loss": 2.7942, "step": 44130 }, { "epoch": 3.46, "learning_rate": 6.539559319375833e-05, "loss": 2.6016, "step": 44131 }, { "epoch": 3.46, "learning_rate": 6.539480906453384e-05, "loss": 2.864, "step": 44132 }, { "epoch": 3.46, "learning_rate": 6.539402493530934e-05, "loss": 2.723, "step": 44133 }, { "epoch": 3.46, "learning_rate": 6.539324080608485e-05, "loss": 2.5794, "step": 44134 }, { "epoch": 3.46, "learning_rate": 6.539245667686035e-05, "loss": 2.7438, "step": 44135 }, { "epoch": 3.46, "learning_rate": 6.539167254763585e-05, "loss": 2.7162, "step": 44136 }, { "epoch": 3.46, "learning_rate": 6.539088841841136e-05, "loss": 2.7851, "step": 44137 }, { "epoch": 3.46, "learning_rate": 6.539010428918686e-05, "loss": 2.6016, "step": 44138 }, { "epoch": 3.46, "learning_rate": 6.538932015996236e-05, "loss": 2.7091, "step": 44139 }, { "epoch": 3.46, "learning_rate": 6.538853603073787e-05, "loss": 2.7193, "step": 44140 }, { "epoch": 3.46, "learning_rate": 6.538775190151337e-05, "loss": 2.8902, "step": 44141 }, { "epoch": 3.46, "learning_rate": 6.538696777228888e-05, "loss": 2.8189, "step": 44142 }, { "epoch": 3.46, "learning_rate": 6.538618364306438e-05, "loss": 2.4895, "step": 44143 }, { "epoch": 3.46, "learning_rate": 6.538539951383988e-05, "loss": 2.5177, "step": 44144 }, { "epoch": 3.46, "learning_rate": 6.538461538461539e-05, "loss": 2.5564, "step": 44145 }, { "epoch": 3.46, "learning_rate": 6.538383125539089e-05, "loss": 2.8357, "step": 44146 }, { "epoch": 3.46, "learning_rate": 6.53830471261664e-05, "loss": 2.5076, "step": 44147 }, { "epoch": 3.46, "learning_rate": 6.53822629969419e-05, "loss": 2.6846, "step": 44148 }, { "epoch": 3.46, "learning_rate": 6.53814788677174e-05, "loss": 3.2109, "step": 44149 }, { "epoch": 3.46, "learning_rate": 6.53806947384929e-05, "loss": 2.5483, "step": 44150 }, { "epoch": 3.46, "learning_rate": 6.537991060926841e-05, "loss": 2.8921, "step": 44151 }, { "epoch": 3.46, "learning_rate": 6.537912648004391e-05, "loss": 2.8028, "step": 44152 }, { "epoch": 3.46, "learning_rate": 6.537834235081942e-05, "loss": 2.6114, "step": 44153 }, { "epoch": 3.46, "learning_rate": 6.537755822159492e-05, "loss": 2.8329, "step": 44154 }, { "epoch": 3.46, "learning_rate": 6.537677409237042e-05, "loss": 2.8128, "step": 44155 }, { "epoch": 3.46, "learning_rate": 6.537598996314594e-05, "loss": 2.5381, "step": 44156 }, { "epoch": 3.46, "learning_rate": 6.537520583392144e-05, "loss": 2.7436, "step": 44157 }, { "epoch": 3.46, "learning_rate": 6.537442170469695e-05, "loss": 2.6994, "step": 44158 }, { "epoch": 3.46, "learning_rate": 6.537363757547245e-05, "loss": 2.6725, "step": 44159 }, { "epoch": 3.46, "learning_rate": 6.537285344624795e-05, "loss": 2.6029, "step": 44160 }, { "epoch": 3.46, "learning_rate": 6.537206931702345e-05, "loss": 2.7323, "step": 44161 }, { "epoch": 3.46, "learning_rate": 6.537128518779895e-05, "loss": 3.0081, "step": 44162 }, { "epoch": 3.46, "learning_rate": 6.537050105857445e-05, "loss": 2.5872, "step": 44163 }, { "epoch": 3.46, "learning_rate": 6.536971692934996e-05, "loss": 2.6565, "step": 44164 }, { "epoch": 3.46, "learning_rate": 6.536893280012546e-05, "loss": 2.7983, "step": 44165 }, { "epoch": 3.46, "learning_rate": 6.536814867090096e-05, "loss": 2.7816, "step": 44166 }, { "epoch": 3.46, "learning_rate": 6.536736454167647e-05, "loss": 2.6554, "step": 44167 }, { "epoch": 3.46, "learning_rate": 6.536658041245197e-05, "loss": 2.7193, "step": 44168 }, { "epoch": 3.46, "learning_rate": 6.536579628322747e-05, "loss": 2.4299, "step": 44169 }, { "epoch": 3.46, "learning_rate": 6.536501215400298e-05, "loss": 2.7091, "step": 44170 }, { "epoch": 3.46, "learning_rate": 6.536422802477848e-05, "loss": 2.9224, "step": 44171 }, { "epoch": 3.46, "learning_rate": 6.536344389555399e-05, "loss": 2.5527, "step": 44172 }, { "epoch": 3.46, "learning_rate": 6.536265976632949e-05, "loss": 2.6043, "step": 44173 }, { "epoch": 3.46, "learning_rate": 6.536187563710499e-05, "loss": 2.6754, "step": 44174 }, { "epoch": 3.46, "learning_rate": 6.53610915078805e-05, "loss": 2.7589, "step": 44175 }, { "epoch": 3.46, "learning_rate": 6.5360307378656e-05, "loss": 2.5681, "step": 44176 }, { "epoch": 3.46, "learning_rate": 6.53595232494315e-05, "loss": 2.7557, "step": 44177 }, { "epoch": 3.46, "learning_rate": 6.535873912020701e-05, "loss": 2.8683, "step": 44178 }, { "epoch": 3.46, "learning_rate": 6.535795499098251e-05, "loss": 2.7567, "step": 44179 }, { "epoch": 3.46, "learning_rate": 6.535717086175801e-05, "loss": 3.005, "step": 44180 }, { "epoch": 3.46, "learning_rate": 6.535638673253353e-05, "loss": 2.6445, "step": 44181 }, { "epoch": 3.46, "learning_rate": 6.535560260330904e-05, "loss": 2.7425, "step": 44182 }, { "epoch": 3.46, "learning_rate": 6.535481847408454e-05, "loss": 2.5216, "step": 44183 }, { "epoch": 3.46, "learning_rate": 6.535403434486004e-05, "loss": 2.8328, "step": 44184 }, { "epoch": 3.46, "learning_rate": 6.535325021563555e-05, "loss": 2.9625, "step": 44185 }, { "epoch": 3.46, "learning_rate": 6.535246608641105e-05, "loss": 2.7401, "step": 44186 }, { "epoch": 3.46, "learning_rate": 6.535168195718655e-05, "loss": 2.7056, "step": 44187 }, { "epoch": 3.46, "learning_rate": 6.535089782796206e-05, "loss": 2.5716, "step": 44188 }, { "epoch": 3.46, "learning_rate": 6.535011369873756e-05, "loss": 2.6893, "step": 44189 }, { "epoch": 3.47, "learning_rate": 6.534932956951307e-05, "loss": 2.6925, "step": 44190 }, { "epoch": 3.47, "learning_rate": 6.534854544028857e-05, "loss": 2.8673, "step": 44191 }, { "epoch": 3.47, "learning_rate": 6.534776131106407e-05, "loss": 2.6069, "step": 44192 }, { "epoch": 3.47, "learning_rate": 6.534697718183958e-05, "loss": 2.7902, "step": 44193 }, { "epoch": 3.47, "learning_rate": 6.534619305261508e-05, "loss": 2.6826, "step": 44194 }, { "epoch": 3.47, "learning_rate": 6.534540892339058e-05, "loss": 2.6974, "step": 44195 }, { "epoch": 3.47, "learning_rate": 6.534462479416609e-05, "loss": 2.6447, "step": 44196 }, { "epoch": 3.47, "learning_rate": 6.534384066494158e-05, "loss": 2.7774, "step": 44197 }, { "epoch": 3.47, "learning_rate": 6.534305653571708e-05, "loss": 2.5949, "step": 44198 }, { "epoch": 3.47, "learning_rate": 6.534227240649258e-05, "loss": 2.8979, "step": 44199 }, { "epoch": 3.47, "learning_rate": 6.534148827726809e-05, "loss": 2.5191, "step": 44200 }, { "epoch": 3.47, "learning_rate": 6.534070414804359e-05, "loss": 2.7055, "step": 44201 }, { "epoch": 3.47, "learning_rate": 6.53399200188191e-05, "loss": 2.8413, "step": 44202 }, { "epoch": 3.47, "learning_rate": 6.53391358895946e-05, "loss": 2.6382, "step": 44203 }, { "epoch": 3.47, "learning_rate": 6.53383517603701e-05, "loss": 3.04, "step": 44204 }, { "epoch": 3.47, "learning_rate": 6.533756763114561e-05, "loss": 2.8944, "step": 44205 }, { "epoch": 3.47, "learning_rate": 6.533678350192111e-05, "loss": 2.671, "step": 44206 }, { "epoch": 3.47, "learning_rate": 6.533599937269663e-05, "loss": 2.5696, "step": 44207 }, { "epoch": 3.47, "learning_rate": 6.533521524347213e-05, "loss": 2.6831, "step": 44208 }, { "epoch": 3.47, "learning_rate": 6.533443111424764e-05, "loss": 2.8115, "step": 44209 }, { "epoch": 3.47, "learning_rate": 6.533364698502314e-05, "loss": 2.7691, "step": 44210 }, { "epoch": 3.47, "learning_rate": 6.533286285579864e-05, "loss": 2.7929, "step": 44211 }, { "epoch": 3.47, "learning_rate": 6.533207872657415e-05, "loss": 2.7528, "step": 44212 }, { "epoch": 3.47, "learning_rate": 6.533129459734965e-05, "loss": 2.6645, "step": 44213 }, { "epoch": 3.47, "learning_rate": 6.533051046812515e-05, "loss": 2.6385, "step": 44214 }, { "epoch": 3.47, "learning_rate": 6.532972633890066e-05, "loss": 2.7868, "step": 44215 }, { "epoch": 3.47, "learning_rate": 6.532894220967616e-05, "loss": 2.6836, "step": 44216 }, { "epoch": 3.47, "learning_rate": 6.532815808045166e-05, "loss": 2.765, "step": 44217 }, { "epoch": 3.47, "learning_rate": 6.532737395122717e-05, "loss": 2.8585, "step": 44218 }, { "epoch": 3.47, "learning_rate": 6.532658982200267e-05, "loss": 2.7631, "step": 44219 }, { "epoch": 3.47, "learning_rate": 6.532580569277818e-05, "loss": 2.656, "step": 44220 }, { "epoch": 3.47, "learning_rate": 6.532502156355368e-05, "loss": 2.8931, "step": 44221 }, { "epoch": 3.47, "learning_rate": 6.532423743432918e-05, "loss": 2.6267, "step": 44222 }, { "epoch": 3.47, "learning_rate": 6.532345330510469e-05, "loss": 2.4947, "step": 44223 }, { "epoch": 3.47, "learning_rate": 6.532266917588019e-05, "loss": 2.7859, "step": 44224 }, { "epoch": 3.47, "learning_rate": 6.53218850466557e-05, "loss": 2.6767, "step": 44225 }, { "epoch": 3.47, "learning_rate": 6.53211009174312e-05, "loss": 2.655, "step": 44226 }, { "epoch": 3.47, "learning_rate": 6.53203167882067e-05, "loss": 2.7513, "step": 44227 }, { "epoch": 3.47, "learning_rate": 6.53195326589822e-05, "loss": 2.7847, "step": 44228 }, { "epoch": 3.47, "learning_rate": 6.531874852975771e-05, "loss": 2.643, "step": 44229 }, { "epoch": 3.47, "learning_rate": 6.531796440053321e-05, "loss": 3.1131, "step": 44230 }, { "epoch": 3.47, "learning_rate": 6.531718027130872e-05, "loss": 2.647, "step": 44231 }, { "epoch": 3.47, "learning_rate": 6.531639614208422e-05, "loss": 2.8808, "step": 44232 }, { "epoch": 3.47, "learning_rate": 6.531561201285972e-05, "loss": 2.6571, "step": 44233 }, { "epoch": 3.47, "learning_rate": 6.531482788363523e-05, "loss": 2.6118, "step": 44234 }, { "epoch": 3.47, "learning_rate": 6.531404375441073e-05, "loss": 2.8861, "step": 44235 }, { "epoch": 3.47, "learning_rate": 6.531325962518623e-05, "loss": 2.6971, "step": 44236 }, { "epoch": 3.47, "learning_rate": 6.531247549596174e-05, "loss": 2.5903, "step": 44237 }, { "epoch": 3.47, "learning_rate": 6.531169136673724e-05, "loss": 2.8171, "step": 44238 }, { "epoch": 3.47, "learning_rate": 6.531090723751275e-05, "loss": 2.7776, "step": 44239 }, { "epoch": 3.47, "learning_rate": 6.531012310828825e-05, "loss": 2.7166, "step": 44240 }, { "epoch": 3.47, "learning_rate": 6.530933897906375e-05, "loss": 2.5939, "step": 44241 }, { "epoch": 3.47, "learning_rate": 6.530855484983926e-05, "loss": 2.8314, "step": 44242 }, { "epoch": 3.47, "learning_rate": 6.530777072061476e-05, "loss": 2.8699, "step": 44243 }, { "epoch": 3.47, "learning_rate": 6.530698659139026e-05, "loss": 2.527, "step": 44244 }, { "epoch": 3.47, "learning_rate": 6.530620246216577e-05, "loss": 2.838, "step": 44245 }, { "epoch": 3.47, "learning_rate": 6.530541833294127e-05, "loss": 2.8079, "step": 44246 }, { "epoch": 3.47, "learning_rate": 6.530463420371677e-05, "loss": 2.5971, "step": 44247 }, { "epoch": 3.47, "learning_rate": 6.530385007449228e-05, "loss": 3.0797, "step": 44248 }, { "epoch": 3.47, "learning_rate": 6.530306594526778e-05, "loss": 2.7765, "step": 44249 }, { "epoch": 3.47, "learning_rate": 6.530228181604329e-05, "loss": 2.5591, "step": 44250 }, { "epoch": 3.47, "learning_rate": 6.530149768681879e-05, "loss": 2.6299, "step": 44251 }, { "epoch": 3.47, "learning_rate": 6.53007135575943e-05, "loss": 2.5822, "step": 44252 }, { "epoch": 3.47, "learning_rate": 6.52999294283698e-05, "loss": 2.952, "step": 44253 }, { "epoch": 3.47, "learning_rate": 6.52991452991453e-05, "loss": 2.6883, "step": 44254 }, { "epoch": 3.47, "learning_rate": 6.52983611699208e-05, "loss": 2.8426, "step": 44255 }, { "epoch": 3.47, "learning_rate": 6.529757704069631e-05, "loss": 2.7747, "step": 44256 }, { "epoch": 3.47, "learning_rate": 6.529679291147181e-05, "loss": 2.5215, "step": 44257 }, { "epoch": 3.47, "learning_rate": 6.529600878224732e-05, "loss": 2.7094, "step": 44258 }, { "epoch": 3.47, "learning_rate": 6.529522465302282e-05, "loss": 2.5914, "step": 44259 }, { "epoch": 3.47, "learning_rate": 6.529444052379832e-05, "loss": 2.7192, "step": 44260 }, { "epoch": 3.47, "learning_rate": 6.529365639457383e-05, "loss": 2.816, "step": 44261 }, { "epoch": 3.47, "learning_rate": 6.529287226534933e-05, "loss": 2.549, "step": 44262 }, { "epoch": 3.47, "learning_rate": 6.529208813612483e-05, "loss": 2.6629, "step": 44263 }, { "epoch": 3.47, "learning_rate": 6.529130400690035e-05, "loss": 2.7379, "step": 44264 }, { "epoch": 3.47, "learning_rate": 6.529051987767585e-05, "loss": 2.7505, "step": 44265 }, { "epoch": 3.47, "learning_rate": 6.528973574845136e-05, "loss": 2.6939, "step": 44266 }, { "epoch": 3.47, "learning_rate": 6.528895161922686e-05, "loss": 2.8401, "step": 44267 }, { "epoch": 3.47, "learning_rate": 6.528816749000237e-05, "loss": 2.7678, "step": 44268 }, { "epoch": 3.47, "learning_rate": 6.528738336077786e-05, "loss": 2.8132, "step": 44269 }, { "epoch": 3.47, "learning_rate": 6.528659923155336e-05, "loss": 2.7193, "step": 44270 }, { "epoch": 3.47, "learning_rate": 6.528581510232886e-05, "loss": 3.0408, "step": 44271 }, { "epoch": 3.47, "learning_rate": 6.528503097310437e-05, "loss": 2.6467, "step": 44272 }, { "epoch": 3.47, "learning_rate": 6.528424684387987e-05, "loss": 2.6126, "step": 44273 }, { "epoch": 3.47, "learning_rate": 6.528346271465537e-05, "loss": 2.7341, "step": 44274 }, { "epoch": 3.47, "learning_rate": 6.528267858543088e-05, "loss": 2.6384, "step": 44275 }, { "epoch": 3.47, "learning_rate": 6.528189445620638e-05, "loss": 2.574, "step": 44276 }, { "epoch": 3.47, "learning_rate": 6.528111032698189e-05, "loss": 2.849, "step": 44277 }, { "epoch": 3.47, "learning_rate": 6.528032619775739e-05, "loss": 2.6903, "step": 44278 }, { "epoch": 3.47, "learning_rate": 6.527954206853289e-05, "loss": 2.8624, "step": 44279 }, { "epoch": 3.47, "learning_rate": 6.52787579393084e-05, "loss": 2.8032, "step": 44280 }, { "epoch": 3.47, "learning_rate": 6.52779738100839e-05, "loss": 2.8766, "step": 44281 }, { "epoch": 3.47, "learning_rate": 6.52771896808594e-05, "loss": 2.8042, "step": 44282 }, { "epoch": 3.47, "learning_rate": 6.527640555163491e-05, "loss": 2.5427, "step": 44283 }, { "epoch": 3.47, "learning_rate": 6.527562142241041e-05, "loss": 2.7165, "step": 44284 }, { "epoch": 3.47, "learning_rate": 6.527483729318591e-05, "loss": 2.7466, "step": 44285 }, { "epoch": 3.47, "learning_rate": 6.527405316396142e-05, "loss": 2.8143, "step": 44286 }, { "epoch": 3.47, "learning_rate": 6.527326903473692e-05, "loss": 2.6811, "step": 44287 }, { "epoch": 3.47, "learning_rate": 6.527248490551243e-05, "loss": 2.7665, "step": 44288 }, { "epoch": 3.47, "learning_rate": 6.527170077628794e-05, "loss": 2.507, "step": 44289 }, { "epoch": 3.47, "learning_rate": 6.527091664706345e-05, "loss": 2.792, "step": 44290 }, { "epoch": 3.47, "learning_rate": 6.527013251783895e-05, "loss": 2.8063, "step": 44291 }, { "epoch": 3.47, "learning_rate": 6.526934838861445e-05, "loss": 2.448, "step": 44292 }, { "epoch": 3.47, "learning_rate": 6.526856425938996e-05, "loss": 2.7145, "step": 44293 }, { "epoch": 3.47, "learning_rate": 6.526778013016546e-05, "loss": 2.7626, "step": 44294 }, { "epoch": 3.47, "learning_rate": 6.526699600094096e-05, "loss": 2.622, "step": 44295 }, { "epoch": 3.47, "learning_rate": 6.526621187171647e-05, "loss": 2.6875, "step": 44296 }, { "epoch": 3.47, "learning_rate": 6.526542774249197e-05, "loss": 2.7537, "step": 44297 }, { "epoch": 3.47, "learning_rate": 6.526464361326748e-05, "loss": 2.6121, "step": 44298 }, { "epoch": 3.47, "learning_rate": 6.526385948404298e-05, "loss": 2.7368, "step": 44299 }, { "epoch": 3.47, "learning_rate": 6.526307535481848e-05, "loss": 2.7472, "step": 44300 }, { "epoch": 3.47, "learning_rate": 6.526229122559399e-05, "loss": 2.7031, "step": 44301 }, { "epoch": 3.47, "learning_rate": 6.526150709636949e-05, "loss": 2.8429, "step": 44302 }, { "epoch": 3.47, "learning_rate": 6.5260722967145e-05, "loss": 2.5826, "step": 44303 }, { "epoch": 3.47, "learning_rate": 6.52599388379205e-05, "loss": 2.6571, "step": 44304 }, { "epoch": 3.47, "learning_rate": 6.525915470869599e-05, "loss": 3.0023, "step": 44305 }, { "epoch": 3.47, "learning_rate": 6.525837057947149e-05, "loss": 2.8386, "step": 44306 }, { "epoch": 3.47, "learning_rate": 6.5257586450247e-05, "loss": 2.8458, "step": 44307 }, { "epoch": 3.47, "learning_rate": 6.52568023210225e-05, "loss": 2.7614, "step": 44308 }, { "epoch": 3.47, "learning_rate": 6.5256018191798e-05, "loss": 2.7368, "step": 44309 }, { "epoch": 3.47, "learning_rate": 6.52552340625735e-05, "loss": 2.6257, "step": 44310 }, { "epoch": 3.47, "learning_rate": 6.525444993334901e-05, "loss": 2.7315, "step": 44311 }, { "epoch": 3.47, "learning_rate": 6.525366580412451e-05, "loss": 2.5523, "step": 44312 }, { "epoch": 3.47, "learning_rate": 6.525288167490002e-05, "loss": 2.8798, "step": 44313 }, { "epoch": 3.47, "learning_rate": 6.525209754567552e-05, "loss": 2.86, "step": 44314 }, { "epoch": 3.47, "learning_rate": 6.525131341645104e-05, "loss": 2.7552, "step": 44315 }, { "epoch": 3.47, "learning_rate": 6.525052928722654e-05, "loss": 2.7852, "step": 44316 }, { "epoch": 3.48, "learning_rate": 6.524974515800205e-05, "loss": 2.8381, "step": 44317 }, { "epoch": 3.48, "learning_rate": 6.524896102877755e-05, "loss": 2.6646, "step": 44318 }, { "epoch": 3.48, "learning_rate": 6.524817689955305e-05, "loss": 2.8163, "step": 44319 }, { "epoch": 3.48, "learning_rate": 6.524739277032856e-05, "loss": 2.5046, "step": 44320 }, { "epoch": 3.48, "learning_rate": 6.524660864110406e-05, "loss": 2.6393, "step": 44321 }, { "epoch": 3.48, "learning_rate": 6.524582451187956e-05, "loss": 2.7577, "step": 44322 }, { "epoch": 3.48, "learning_rate": 6.524504038265507e-05, "loss": 2.9248, "step": 44323 }, { "epoch": 3.48, "learning_rate": 6.524425625343057e-05, "loss": 2.6089, "step": 44324 }, { "epoch": 3.48, "learning_rate": 6.524347212420608e-05, "loss": 2.9215, "step": 44325 }, { "epoch": 3.48, "learning_rate": 6.524268799498158e-05, "loss": 2.6339, "step": 44326 }, { "epoch": 3.48, "learning_rate": 6.524190386575708e-05, "loss": 2.5988, "step": 44327 }, { "epoch": 3.48, "learning_rate": 6.524111973653259e-05, "loss": 2.6545, "step": 44328 }, { "epoch": 3.48, "learning_rate": 6.524033560730809e-05, "loss": 2.742, "step": 44329 }, { "epoch": 3.48, "learning_rate": 6.52395514780836e-05, "loss": 2.7615, "step": 44330 }, { "epoch": 3.48, "learning_rate": 6.52387673488591e-05, "loss": 2.8122, "step": 44331 }, { "epoch": 3.48, "learning_rate": 6.52379832196346e-05, "loss": 2.5642, "step": 44332 }, { "epoch": 3.48, "learning_rate": 6.52371990904101e-05, "loss": 2.5744, "step": 44333 }, { "epoch": 3.48, "learning_rate": 6.523641496118561e-05, "loss": 2.6457, "step": 44334 }, { "epoch": 3.48, "learning_rate": 6.523563083196111e-05, "loss": 2.6974, "step": 44335 }, { "epoch": 3.48, "learning_rate": 6.523484670273662e-05, "loss": 2.8216, "step": 44336 }, { "epoch": 3.48, "learning_rate": 6.523406257351212e-05, "loss": 2.7216, "step": 44337 }, { "epoch": 3.48, "learning_rate": 6.523327844428762e-05, "loss": 2.8685, "step": 44338 }, { "epoch": 3.48, "learning_rate": 6.523249431506313e-05, "loss": 2.7629, "step": 44339 }, { "epoch": 3.48, "learning_rate": 6.523171018583863e-05, "loss": 2.6049, "step": 44340 }, { "epoch": 3.48, "learning_rate": 6.523092605661413e-05, "loss": 2.5543, "step": 44341 }, { "epoch": 3.48, "learning_rate": 6.523014192738964e-05, "loss": 2.7758, "step": 44342 }, { "epoch": 3.48, "learning_rate": 6.522935779816514e-05, "loss": 2.9759, "step": 44343 }, { "epoch": 3.48, "learning_rate": 6.522857366894065e-05, "loss": 2.7884, "step": 44344 }, { "epoch": 3.48, "learning_rate": 6.522778953971615e-05, "loss": 2.7896, "step": 44345 }, { "epoch": 3.48, "learning_rate": 6.522700541049165e-05, "loss": 2.6746, "step": 44346 }, { "epoch": 3.48, "learning_rate": 6.522622128126716e-05, "loss": 3.0071, "step": 44347 }, { "epoch": 3.48, "learning_rate": 6.522543715204266e-05, "loss": 2.778, "step": 44348 }, { "epoch": 3.48, "learning_rate": 6.522465302281816e-05, "loss": 2.7413, "step": 44349 }, { "epoch": 3.48, "learning_rate": 6.522386889359367e-05, "loss": 2.6291, "step": 44350 }, { "epoch": 3.48, "learning_rate": 6.522308476436917e-05, "loss": 2.7809, "step": 44351 }, { "epoch": 3.48, "learning_rate": 6.522230063514467e-05, "loss": 2.5737, "step": 44352 }, { "epoch": 3.48, "learning_rate": 6.522151650592018e-05, "loss": 2.6677, "step": 44353 }, { "epoch": 3.48, "learning_rate": 6.522073237669568e-05, "loss": 2.7042, "step": 44354 }, { "epoch": 3.48, "learning_rate": 6.521994824747119e-05, "loss": 2.692, "step": 44355 }, { "epoch": 3.48, "learning_rate": 6.521916411824669e-05, "loss": 2.7311, "step": 44356 }, { "epoch": 3.48, "learning_rate": 6.521837998902219e-05, "loss": 2.8621, "step": 44357 }, { "epoch": 3.48, "learning_rate": 6.52175958597977e-05, "loss": 2.9383, "step": 44358 }, { "epoch": 3.48, "learning_rate": 6.52168117305732e-05, "loss": 2.9368, "step": 44359 }, { "epoch": 3.48, "learning_rate": 6.52160276013487e-05, "loss": 2.4069, "step": 44360 }, { "epoch": 3.48, "learning_rate": 6.521524347212421e-05, "loss": 2.7135, "step": 44361 }, { "epoch": 3.48, "learning_rate": 6.521445934289971e-05, "loss": 2.8285, "step": 44362 }, { "epoch": 3.48, "learning_rate": 6.521367521367522e-05, "loss": 2.5839, "step": 44363 }, { "epoch": 3.48, "learning_rate": 6.521289108445072e-05, "loss": 2.6516, "step": 44364 }, { "epoch": 3.48, "learning_rate": 6.521210695522622e-05, "loss": 2.7803, "step": 44365 }, { "epoch": 3.48, "learning_rate": 6.521132282600173e-05, "loss": 2.9089, "step": 44366 }, { "epoch": 3.48, "learning_rate": 6.521053869677723e-05, "loss": 3.0439, "step": 44367 }, { "epoch": 3.48, "learning_rate": 6.520975456755273e-05, "loss": 2.4281, "step": 44368 }, { "epoch": 3.48, "learning_rate": 6.520897043832824e-05, "loss": 2.6061, "step": 44369 }, { "epoch": 3.48, "learning_rate": 6.520818630910374e-05, "loss": 2.554, "step": 44370 }, { "epoch": 3.48, "learning_rate": 6.520740217987924e-05, "loss": 2.8944, "step": 44371 }, { "epoch": 3.48, "learning_rate": 6.520661805065476e-05, "loss": 2.8921, "step": 44372 }, { "epoch": 3.48, "learning_rate": 6.520583392143027e-05, "loss": 2.586, "step": 44373 }, { "epoch": 3.48, "learning_rate": 6.520504979220577e-05, "loss": 2.809, "step": 44374 }, { "epoch": 3.48, "learning_rate": 6.520426566298127e-05, "loss": 2.7734, "step": 44375 }, { "epoch": 3.48, "learning_rate": 6.520348153375678e-05, "loss": 2.5656, "step": 44376 }, { "epoch": 3.48, "learning_rate": 6.520269740453227e-05, "loss": 2.6722, "step": 44377 }, { "epoch": 3.48, "learning_rate": 6.520191327530777e-05, "loss": 2.783, "step": 44378 }, { "epoch": 3.48, "learning_rate": 6.520112914608327e-05, "loss": 2.8072, "step": 44379 }, { "epoch": 3.48, "learning_rate": 6.520034501685878e-05, "loss": 2.6712, "step": 44380 }, { "epoch": 3.48, "learning_rate": 6.519956088763428e-05, "loss": 2.7628, "step": 44381 }, { "epoch": 3.48, "learning_rate": 6.519877675840978e-05, "loss": 2.8915, "step": 44382 }, { "epoch": 3.48, "learning_rate": 6.519799262918529e-05, "loss": 2.7996, "step": 44383 }, { "epoch": 3.48, "learning_rate": 6.519720849996079e-05, "loss": 2.62, "step": 44384 }, { "epoch": 3.48, "learning_rate": 6.51964243707363e-05, "loss": 2.7753, "step": 44385 }, { "epoch": 3.48, "learning_rate": 6.51956402415118e-05, "loss": 2.7513, "step": 44386 }, { "epoch": 3.48, "learning_rate": 6.51948561122873e-05, "loss": 2.7561, "step": 44387 }, { "epoch": 3.48, "learning_rate": 6.519407198306281e-05, "loss": 2.51, "step": 44388 }, { "epoch": 3.48, "learning_rate": 6.519328785383831e-05, "loss": 2.7318, "step": 44389 }, { "epoch": 3.48, "learning_rate": 6.519250372461381e-05, "loss": 2.5676, "step": 44390 }, { "epoch": 3.48, "learning_rate": 6.519171959538932e-05, "loss": 2.64, "step": 44391 }, { "epoch": 3.48, "learning_rate": 6.519093546616482e-05, "loss": 2.6629, "step": 44392 }, { "epoch": 3.48, "learning_rate": 6.519015133694033e-05, "loss": 2.5129, "step": 44393 }, { "epoch": 3.48, "learning_rate": 6.518936720771583e-05, "loss": 2.7727, "step": 44394 }, { "epoch": 3.48, "learning_rate": 6.518858307849133e-05, "loss": 2.6838, "step": 44395 }, { "epoch": 3.48, "learning_rate": 6.518779894926684e-05, "loss": 2.8835, "step": 44396 }, { "epoch": 3.48, "learning_rate": 6.518701482004235e-05, "loss": 2.6838, "step": 44397 }, { "epoch": 3.48, "learning_rate": 6.518623069081786e-05, "loss": 2.8216, "step": 44398 }, { "epoch": 3.48, "learning_rate": 6.518544656159336e-05, "loss": 2.9266, "step": 44399 }, { "epoch": 3.48, "learning_rate": 6.518466243236886e-05, "loss": 2.711, "step": 44400 }, { "epoch": 3.48, "learning_rate": 6.518387830314437e-05, "loss": 2.699, "step": 44401 }, { "epoch": 3.48, "learning_rate": 6.518309417391987e-05, "loss": 2.8171, "step": 44402 }, { "epoch": 3.48, "learning_rate": 6.518231004469538e-05, "loss": 2.633, "step": 44403 }, { "epoch": 3.48, "learning_rate": 6.518152591547088e-05, "loss": 2.6507, "step": 44404 }, { "epoch": 3.48, "learning_rate": 6.518074178624638e-05, "loss": 2.824, "step": 44405 }, { "epoch": 3.48, "learning_rate": 6.517995765702189e-05, "loss": 2.8086, "step": 44406 }, { "epoch": 3.48, "learning_rate": 6.517917352779739e-05, "loss": 2.8563, "step": 44407 }, { "epoch": 3.48, "learning_rate": 6.51783893985729e-05, "loss": 2.789, "step": 44408 }, { "epoch": 3.48, "learning_rate": 6.51776052693484e-05, "loss": 2.6689, "step": 44409 }, { "epoch": 3.48, "learning_rate": 6.51768211401239e-05, "loss": 2.613, "step": 44410 }, { "epoch": 3.48, "learning_rate": 6.51760370108994e-05, "loss": 2.8815, "step": 44411 }, { "epoch": 3.48, "learning_rate": 6.517525288167491e-05, "loss": 2.7364, "step": 44412 }, { "epoch": 3.48, "learning_rate": 6.51744687524504e-05, "loss": 2.5326, "step": 44413 }, { "epoch": 3.48, "learning_rate": 6.51736846232259e-05, "loss": 2.7295, "step": 44414 }, { "epoch": 3.48, "learning_rate": 6.51729004940014e-05, "loss": 2.7451, "step": 44415 }, { "epoch": 3.48, "learning_rate": 6.517211636477691e-05, "loss": 2.6576, "step": 44416 }, { "epoch": 3.48, "learning_rate": 6.517133223555241e-05, "loss": 2.6199, "step": 44417 }, { "epoch": 3.48, "learning_rate": 6.517054810632792e-05, "loss": 2.533, "step": 44418 }, { "epoch": 3.48, "learning_rate": 6.516976397710342e-05, "loss": 2.9793, "step": 44419 }, { "epoch": 3.48, "learning_rate": 6.516897984787892e-05, "loss": 2.7582, "step": 44420 }, { "epoch": 3.48, "learning_rate": 6.516819571865443e-05, "loss": 2.7298, "step": 44421 }, { "epoch": 3.48, "learning_rate": 6.516741158942993e-05, "loss": 2.7386, "step": 44422 }, { "epoch": 3.48, "learning_rate": 6.516662746020545e-05, "loss": 2.6253, "step": 44423 }, { "epoch": 3.48, "learning_rate": 6.516584333098095e-05, "loss": 2.6195, "step": 44424 }, { "epoch": 3.48, "learning_rate": 6.516505920175646e-05, "loss": 2.5675, "step": 44425 }, { "epoch": 3.48, "learning_rate": 6.516427507253196e-05, "loss": 2.7294, "step": 44426 }, { "epoch": 3.48, "learning_rate": 6.516349094330746e-05, "loss": 2.7068, "step": 44427 }, { "epoch": 3.48, "learning_rate": 6.516270681408297e-05, "loss": 2.657, "step": 44428 }, { "epoch": 3.48, "learning_rate": 6.516192268485847e-05, "loss": 2.6595, "step": 44429 }, { "epoch": 3.48, "learning_rate": 6.516113855563397e-05, "loss": 2.7098, "step": 44430 }, { "epoch": 3.48, "learning_rate": 6.516035442640948e-05, "loss": 2.6775, "step": 44431 }, { "epoch": 3.48, "learning_rate": 6.515957029718498e-05, "loss": 2.6528, "step": 44432 }, { "epoch": 3.48, "learning_rate": 6.515878616796049e-05, "loss": 2.6875, "step": 44433 }, { "epoch": 3.48, "learning_rate": 6.515800203873599e-05, "loss": 2.5806, "step": 44434 }, { "epoch": 3.48, "learning_rate": 6.51572179095115e-05, "loss": 2.7583, "step": 44435 }, { "epoch": 3.48, "learning_rate": 6.5156433780287e-05, "loss": 2.9036, "step": 44436 }, { "epoch": 3.48, "learning_rate": 6.51556496510625e-05, "loss": 2.8043, "step": 44437 }, { "epoch": 3.48, "learning_rate": 6.5154865521838e-05, "loss": 2.6789, "step": 44438 }, { "epoch": 3.48, "learning_rate": 6.515408139261351e-05, "loss": 2.7504, "step": 44439 }, { "epoch": 3.48, "learning_rate": 6.515329726338901e-05, "loss": 2.8151, "step": 44440 }, { "epoch": 3.48, "learning_rate": 6.515251313416452e-05, "loss": 2.6884, "step": 44441 }, { "epoch": 3.48, "learning_rate": 6.515172900494002e-05, "loss": 2.743, "step": 44442 }, { "epoch": 3.48, "learning_rate": 6.515094487571552e-05, "loss": 2.5248, "step": 44443 }, { "epoch": 3.48, "learning_rate": 6.515016074649103e-05, "loss": 2.8245, "step": 44444 }, { "epoch": 3.49, "learning_rate": 6.514937661726653e-05, "loss": 2.8026, "step": 44445 }, { "epoch": 3.49, "learning_rate": 6.514859248804203e-05, "loss": 2.6266, "step": 44446 }, { "epoch": 3.49, "learning_rate": 6.514780835881754e-05, "loss": 2.6169, "step": 44447 }, { "epoch": 3.49, "learning_rate": 6.514702422959304e-05, "loss": 2.6371, "step": 44448 }, { "epoch": 3.49, "learning_rate": 6.514624010036854e-05, "loss": 2.5061, "step": 44449 }, { "epoch": 3.49, "learning_rate": 6.514545597114405e-05, "loss": 2.8327, "step": 44450 }, { "epoch": 3.49, "learning_rate": 6.514467184191955e-05, "loss": 2.5733, "step": 44451 }, { "epoch": 3.49, "learning_rate": 6.514388771269506e-05, "loss": 2.6298, "step": 44452 }, { "epoch": 3.49, "learning_rate": 6.514310358347056e-05, "loss": 2.8237, "step": 44453 }, { "epoch": 3.49, "learning_rate": 6.514231945424606e-05, "loss": 2.6493, "step": 44454 }, { "epoch": 3.49, "learning_rate": 6.514153532502157e-05, "loss": 2.7528, "step": 44455 }, { "epoch": 3.49, "learning_rate": 6.514075119579707e-05, "loss": 2.6734, "step": 44456 }, { "epoch": 3.49, "learning_rate": 6.513996706657257e-05, "loss": 2.8268, "step": 44457 }, { "epoch": 3.49, "learning_rate": 6.513918293734808e-05, "loss": 2.5122, "step": 44458 }, { "epoch": 3.49, "learning_rate": 6.513839880812358e-05, "loss": 2.6784, "step": 44459 }, { "epoch": 3.49, "learning_rate": 6.513761467889909e-05, "loss": 2.7691, "step": 44460 }, { "epoch": 3.49, "learning_rate": 6.513683054967459e-05, "loss": 2.7096, "step": 44461 }, { "epoch": 3.49, "learning_rate": 6.513604642045009e-05, "loss": 2.7313, "step": 44462 }, { "epoch": 3.49, "learning_rate": 6.51352622912256e-05, "loss": 2.7779, "step": 44463 }, { "epoch": 3.49, "learning_rate": 6.51344781620011e-05, "loss": 2.7187, "step": 44464 }, { "epoch": 3.49, "learning_rate": 6.51336940327766e-05, "loss": 2.8207, "step": 44465 }, { "epoch": 3.49, "learning_rate": 6.513290990355211e-05, "loss": 2.831, "step": 44466 }, { "epoch": 3.49, "learning_rate": 6.513212577432761e-05, "loss": 2.6181, "step": 44467 }, { "epoch": 3.49, "learning_rate": 6.513134164510311e-05, "loss": 2.6746, "step": 44468 }, { "epoch": 3.49, "learning_rate": 6.513055751587862e-05, "loss": 2.753, "step": 44469 }, { "epoch": 3.49, "learning_rate": 6.512977338665412e-05, "loss": 2.6887, "step": 44470 }, { "epoch": 3.49, "learning_rate": 6.512898925742963e-05, "loss": 2.6433, "step": 44471 }, { "epoch": 3.49, "learning_rate": 6.512820512820513e-05, "loss": 2.7147, "step": 44472 }, { "epoch": 3.49, "learning_rate": 6.512742099898063e-05, "loss": 2.8308, "step": 44473 }, { "epoch": 3.49, "learning_rate": 6.512663686975614e-05, "loss": 2.7102, "step": 44474 }, { "epoch": 3.49, "learning_rate": 6.512585274053164e-05, "loss": 2.8907, "step": 44475 }, { "epoch": 3.49, "learning_rate": 6.512506861130714e-05, "loss": 2.6504, "step": 44476 }, { "epoch": 3.49, "learning_rate": 6.512428448208265e-05, "loss": 2.87, "step": 44477 }, { "epoch": 3.49, "learning_rate": 6.512350035285815e-05, "loss": 2.7073, "step": 44478 }, { "epoch": 3.49, "learning_rate": 6.512271622363366e-05, "loss": 2.9498, "step": 44479 }, { "epoch": 3.49, "learning_rate": 6.512193209440917e-05, "loss": 2.7251, "step": 44480 }, { "epoch": 3.49, "learning_rate": 6.512114796518468e-05, "loss": 2.5917, "step": 44481 }, { "epoch": 3.49, "learning_rate": 6.512036383596018e-05, "loss": 2.7506, "step": 44482 }, { "epoch": 3.49, "learning_rate": 6.511957970673568e-05, "loss": 2.5393, "step": 44483 }, { "epoch": 3.49, "learning_rate": 6.511879557751119e-05, "loss": 2.6889, "step": 44484 }, { "epoch": 3.49, "learning_rate": 6.511801144828668e-05, "loss": 2.7875, "step": 44485 }, { "epoch": 3.49, "learning_rate": 6.511722731906218e-05, "loss": 2.8712, "step": 44486 }, { "epoch": 3.49, "learning_rate": 6.511644318983768e-05, "loss": 2.9364, "step": 44487 }, { "epoch": 3.49, "learning_rate": 6.511565906061319e-05, "loss": 2.8122, "step": 44488 }, { "epoch": 3.49, "learning_rate": 6.511487493138869e-05, "loss": 2.7942, "step": 44489 }, { "epoch": 3.49, "learning_rate": 6.51140908021642e-05, "loss": 2.7321, "step": 44490 }, { "epoch": 3.49, "learning_rate": 6.51133066729397e-05, "loss": 2.8329, "step": 44491 }, { "epoch": 3.49, "learning_rate": 6.51125225437152e-05, "loss": 2.585, "step": 44492 }, { "epoch": 3.49, "learning_rate": 6.51117384144907e-05, "loss": 2.7925, "step": 44493 }, { "epoch": 3.49, "learning_rate": 6.511095428526621e-05, "loss": 2.7672, "step": 44494 }, { "epoch": 3.49, "learning_rate": 6.511017015604171e-05, "loss": 2.5588, "step": 44495 }, { "epoch": 3.49, "learning_rate": 6.510938602681722e-05, "loss": 2.9257, "step": 44496 }, { "epoch": 3.49, "learning_rate": 6.510860189759272e-05, "loss": 2.7595, "step": 44497 }, { "epoch": 3.49, "learning_rate": 6.510781776836823e-05, "loss": 2.8343, "step": 44498 }, { "epoch": 3.49, "learning_rate": 6.510703363914373e-05, "loss": 2.8617, "step": 44499 }, { "epoch": 3.49, "learning_rate": 6.510624950991923e-05, "loss": 2.655, "step": 44500 }, { "epoch": 3.49, "learning_rate": 6.510546538069474e-05, "loss": 2.7693, "step": 44501 }, { "epoch": 3.49, "learning_rate": 6.510468125147024e-05, "loss": 2.6307, "step": 44502 }, { "epoch": 3.49, "learning_rate": 6.510389712224574e-05, "loss": 2.7728, "step": 44503 }, { "epoch": 3.49, "learning_rate": 6.510311299302125e-05, "loss": 2.7906, "step": 44504 }, { "epoch": 3.49, "learning_rate": 6.510232886379676e-05, "loss": 2.7087, "step": 44505 }, { "epoch": 3.49, "learning_rate": 6.510154473457227e-05, "loss": 2.7697, "step": 44506 }, { "epoch": 3.49, "learning_rate": 6.510076060534777e-05, "loss": 2.7399, "step": 44507 }, { "epoch": 3.49, "learning_rate": 6.509997647612328e-05, "loss": 2.5732, "step": 44508 }, { "epoch": 3.49, "learning_rate": 6.509919234689878e-05, "loss": 2.7101, "step": 44509 }, { "epoch": 3.49, "learning_rate": 6.509840821767428e-05, "loss": 2.5835, "step": 44510 }, { "epoch": 3.49, "learning_rate": 6.509762408844979e-05, "loss": 2.5459, "step": 44511 }, { "epoch": 3.49, "learning_rate": 6.509683995922529e-05, "loss": 2.8155, "step": 44512 }, { "epoch": 3.49, "learning_rate": 6.50960558300008e-05, "loss": 2.6103, "step": 44513 }, { "epoch": 3.49, "learning_rate": 6.50952717007763e-05, "loss": 2.7865, "step": 44514 }, { "epoch": 3.49, "learning_rate": 6.50944875715518e-05, "loss": 2.6348, "step": 44515 }, { "epoch": 3.49, "learning_rate": 6.50937034423273e-05, "loss": 2.6487, "step": 44516 }, { "epoch": 3.49, "learning_rate": 6.509291931310281e-05, "loss": 2.7373, "step": 44517 }, { "epoch": 3.49, "learning_rate": 6.509213518387831e-05, "loss": 2.6191, "step": 44518 }, { "epoch": 3.49, "learning_rate": 6.509135105465382e-05, "loss": 2.8605, "step": 44519 }, { "epoch": 3.49, "learning_rate": 6.509056692542932e-05, "loss": 2.6883, "step": 44520 }, { "epoch": 3.49, "learning_rate": 6.508978279620481e-05, "loss": 2.7927, "step": 44521 }, { "epoch": 3.49, "learning_rate": 6.508899866698031e-05, "loss": 2.8505, "step": 44522 }, { "epoch": 3.49, "learning_rate": 6.508821453775582e-05, "loss": 2.6342, "step": 44523 }, { "epoch": 3.49, "learning_rate": 6.508743040853132e-05, "loss": 2.7178, "step": 44524 }, { "epoch": 3.49, "learning_rate": 6.508664627930682e-05, "loss": 2.7494, "step": 44525 }, { "epoch": 3.49, "learning_rate": 6.508586215008233e-05, "loss": 2.5458, "step": 44526 }, { "epoch": 3.49, "learning_rate": 6.508507802085783e-05, "loss": 2.7495, "step": 44527 }, { "epoch": 3.49, "learning_rate": 6.508429389163334e-05, "loss": 2.8372, "step": 44528 }, { "epoch": 3.49, "learning_rate": 6.508350976240884e-05, "loss": 2.6452, "step": 44529 }, { "epoch": 3.49, "learning_rate": 6.508272563318434e-05, "loss": 2.442, "step": 44530 }, { "epoch": 3.49, "learning_rate": 6.508194150395986e-05, "loss": 2.6193, "step": 44531 }, { "epoch": 3.49, "learning_rate": 6.508115737473536e-05, "loss": 2.8129, "step": 44532 }, { "epoch": 3.49, "learning_rate": 6.508037324551087e-05, "loss": 2.5631, "step": 44533 }, { "epoch": 3.49, "learning_rate": 6.507958911628637e-05, "loss": 2.6375, "step": 44534 }, { "epoch": 3.49, "learning_rate": 6.507880498706187e-05, "loss": 2.6625, "step": 44535 }, { "epoch": 3.49, "learning_rate": 6.507802085783738e-05, "loss": 2.7125, "step": 44536 }, { "epoch": 3.49, "learning_rate": 6.507723672861288e-05, "loss": 2.8848, "step": 44537 }, { "epoch": 3.49, "learning_rate": 6.507645259938839e-05, "loss": 2.774, "step": 44538 }, { "epoch": 3.49, "learning_rate": 6.507566847016389e-05, "loss": 2.7145, "step": 44539 }, { "epoch": 3.49, "learning_rate": 6.507488434093939e-05, "loss": 2.6427, "step": 44540 }, { "epoch": 3.49, "learning_rate": 6.50741002117149e-05, "loss": 2.8666, "step": 44541 }, { "epoch": 3.49, "learning_rate": 6.50733160824904e-05, "loss": 2.8079, "step": 44542 }, { "epoch": 3.49, "learning_rate": 6.50725319532659e-05, "loss": 2.8887, "step": 44543 }, { "epoch": 3.49, "learning_rate": 6.507174782404141e-05, "loss": 2.5022, "step": 44544 }, { "epoch": 3.49, "learning_rate": 6.507096369481691e-05, "loss": 2.8244, "step": 44545 }, { "epoch": 3.49, "learning_rate": 6.507017956559242e-05, "loss": 2.7053, "step": 44546 }, { "epoch": 3.49, "learning_rate": 6.506939543636792e-05, "loss": 2.743, "step": 44547 }, { "epoch": 3.49, "learning_rate": 6.506861130714342e-05, "loss": 2.6007, "step": 44548 }, { "epoch": 3.49, "learning_rate": 6.506782717791893e-05, "loss": 2.6099, "step": 44549 }, { "epoch": 3.49, "learning_rate": 6.506704304869443e-05, "loss": 2.7586, "step": 44550 }, { "epoch": 3.49, "learning_rate": 6.506625891946993e-05, "loss": 2.6814, "step": 44551 }, { "epoch": 3.49, "learning_rate": 6.506547479024544e-05, "loss": 2.6592, "step": 44552 }, { "epoch": 3.49, "learning_rate": 6.506469066102094e-05, "loss": 2.7506, "step": 44553 }, { "epoch": 3.49, "learning_rate": 6.506390653179644e-05, "loss": 2.6856, "step": 44554 }, { "epoch": 3.49, "learning_rate": 6.506312240257195e-05, "loss": 2.8252, "step": 44555 }, { "epoch": 3.49, "learning_rate": 6.506233827334745e-05, "loss": 2.7368, "step": 44556 }, { "epoch": 3.49, "learning_rate": 6.506155414412296e-05, "loss": 2.9175, "step": 44557 }, { "epoch": 3.49, "learning_rate": 6.506077001489846e-05, "loss": 2.6652, "step": 44558 }, { "epoch": 3.49, "learning_rate": 6.505998588567396e-05, "loss": 2.7462, "step": 44559 }, { "epoch": 3.49, "learning_rate": 6.505920175644947e-05, "loss": 2.7676, "step": 44560 }, { "epoch": 3.49, "learning_rate": 6.505841762722497e-05, "loss": 2.7216, "step": 44561 }, { "epoch": 3.49, "learning_rate": 6.505763349800047e-05, "loss": 2.5662, "step": 44562 }, { "epoch": 3.49, "learning_rate": 6.505684936877598e-05, "loss": 2.6967, "step": 44563 }, { "epoch": 3.49, "learning_rate": 6.505606523955148e-05, "loss": 2.6196, "step": 44564 }, { "epoch": 3.49, "learning_rate": 6.505528111032699e-05, "loss": 2.8143, "step": 44565 }, { "epoch": 3.49, "learning_rate": 6.505449698110249e-05, "loss": 2.6503, "step": 44566 }, { "epoch": 3.49, "learning_rate": 6.505371285187799e-05, "loss": 2.6686, "step": 44567 }, { "epoch": 3.49, "learning_rate": 6.50529287226535e-05, "loss": 2.7799, "step": 44568 }, { "epoch": 3.49, "learning_rate": 6.5052144593429e-05, "loss": 2.7366, "step": 44569 }, { "epoch": 3.49, "learning_rate": 6.50513604642045e-05, "loss": 2.7465, "step": 44570 }, { "epoch": 3.49, "learning_rate": 6.505057633498001e-05, "loss": 2.6633, "step": 44571 }, { "epoch": 3.5, "learning_rate": 6.504979220575551e-05, "loss": 2.8326, "step": 44572 }, { "epoch": 3.5, "learning_rate": 6.504900807653101e-05, "loss": 2.7015, "step": 44573 }, { "epoch": 3.5, "learning_rate": 6.504822394730652e-05, "loss": 2.9073, "step": 44574 }, { "epoch": 3.5, "learning_rate": 6.504743981808202e-05, "loss": 2.7202, "step": 44575 }, { "epoch": 3.5, "learning_rate": 6.504665568885753e-05, "loss": 2.8342, "step": 44576 }, { "epoch": 3.5, "learning_rate": 6.504587155963303e-05, "loss": 2.7328, "step": 44577 }, { "epoch": 3.5, "learning_rate": 6.504508743040853e-05, "loss": 2.6687, "step": 44578 }, { "epoch": 3.5, "learning_rate": 6.504430330118404e-05, "loss": 2.7043, "step": 44579 }, { "epoch": 3.5, "learning_rate": 6.504351917195954e-05, "loss": 2.7614, "step": 44580 }, { "epoch": 3.5, "learning_rate": 6.504273504273504e-05, "loss": 2.7494, "step": 44581 }, { "epoch": 3.5, "learning_rate": 6.504195091351055e-05, "loss": 2.7151, "step": 44582 }, { "epoch": 3.5, "learning_rate": 6.504116678428605e-05, "loss": 2.7715, "step": 44583 }, { "epoch": 3.5, "learning_rate": 6.504038265506155e-05, "loss": 2.6789, "step": 44584 }, { "epoch": 3.5, "learning_rate": 6.503959852583706e-05, "loss": 2.5584, "step": 44585 }, { "epoch": 3.5, "learning_rate": 6.503881439661256e-05, "loss": 2.7311, "step": 44586 }, { "epoch": 3.5, "learning_rate": 6.503803026738807e-05, "loss": 2.8057, "step": 44587 }, { "epoch": 3.5, "learning_rate": 6.503724613816358e-05, "loss": 2.6718, "step": 44588 }, { "epoch": 3.5, "learning_rate": 6.503646200893909e-05, "loss": 2.7394, "step": 44589 }, { "epoch": 3.5, "learning_rate": 6.503567787971459e-05, "loss": 2.7139, "step": 44590 }, { "epoch": 3.5, "learning_rate": 6.50348937504901e-05, "loss": 2.763, "step": 44591 }, { "epoch": 3.5, "learning_rate": 6.50341096212656e-05, "loss": 2.9862, "step": 44592 }, { "epoch": 3.5, "learning_rate": 6.503332549204109e-05, "loss": 2.9091, "step": 44593 }, { "epoch": 3.5, "learning_rate": 6.503254136281659e-05, "loss": 2.7811, "step": 44594 }, { "epoch": 3.5, "learning_rate": 6.50317572335921e-05, "loss": 2.5237, "step": 44595 }, { "epoch": 3.5, "learning_rate": 6.50309731043676e-05, "loss": 2.6296, "step": 44596 }, { "epoch": 3.5, "learning_rate": 6.50301889751431e-05, "loss": 2.7801, "step": 44597 }, { "epoch": 3.5, "learning_rate": 6.50294048459186e-05, "loss": 2.7735, "step": 44598 }, { "epoch": 3.5, "learning_rate": 6.502862071669411e-05, "loss": 2.9927, "step": 44599 }, { "epoch": 3.5, "learning_rate": 6.502783658746961e-05, "loss": 2.8414, "step": 44600 }, { "epoch": 3.5, "learning_rate": 6.502705245824512e-05, "loss": 2.6298, "step": 44601 }, { "epoch": 3.5, "learning_rate": 6.502626832902062e-05, "loss": 2.7462, "step": 44602 }, { "epoch": 3.5, "learning_rate": 6.502548419979612e-05, "loss": 2.9179, "step": 44603 }, { "epoch": 3.5, "learning_rate": 6.502470007057163e-05, "loss": 2.6362, "step": 44604 }, { "epoch": 3.5, "learning_rate": 6.502391594134713e-05, "loss": 2.5507, "step": 44605 }, { "epoch": 3.5, "learning_rate": 6.502313181212264e-05, "loss": 2.5964, "step": 44606 }, { "epoch": 3.5, "learning_rate": 6.502234768289814e-05, "loss": 2.8274, "step": 44607 }, { "epoch": 3.5, "learning_rate": 6.502156355367364e-05, "loss": 2.7136, "step": 44608 }, { "epoch": 3.5, "learning_rate": 6.502077942444915e-05, "loss": 2.8258, "step": 44609 }, { "epoch": 3.5, "learning_rate": 6.501999529522465e-05, "loss": 2.6921, "step": 44610 }, { "epoch": 3.5, "learning_rate": 6.501921116600015e-05, "loss": 2.9077, "step": 44611 }, { "epoch": 3.5, "learning_rate": 6.501842703677566e-05, "loss": 3.0085, "step": 44612 }, { "epoch": 3.5, "learning_rate": 6.501764290755118e-05, "loss": 2.7638, "step": 44613 }, { "epoch": 3.5, "learning_rate": 6.501685877832668e-05, "loss": 2.7263, "step": 44614 }, { "epoch": 3.5, "learning_rate": 6.501607464910218e-05, "loss": 2.6421, "step": 44615 }, { "epoch": 3.5, "learning_rate": 6.501529051987769e-05, "loss": 2.8365, "step": 44616 }, { "epoch": 3.5, "learning_rate": 6.501450639065319e-05, "loss": 2.6905, "step": 44617 }, { "epoch": 3.5, "learning_rate": 6.50137222614287e-05, "loss": 2.7137, "step": 44618 }, { "epoch": 3.5, "learning_rate": 6.50129381322042e-05, "loss": 2.7218, "step": 44619 }, { "epoch": 3.5, "learning_rate": 6.50121540029797e-05, "loss": 2.645, "step": 44620 }, { "epoch": 3.5, "learning_rate": 6.50113698737552e-05, "loss": 2.7092, "step": 44621 }, { "epoch": 3.5, "learning_rate": 6.501058574453071e-05, "loss": 2.6803, "step": 44622 }, { "epoch": 3.5, "learning_rate": 6.500980161530621e-05, "loss": 2.9172, "step": 44623 }, { "epoch": 3.5, "learning_rate": 6.500901748608172e-05, "loss": 2.7525, "step": 44624 }, { "epoch": 3.5, "learning_rate": 6.500823335685722e-05, "loss": 2.4929, "step": 44625 }, { "epoch": 3.5, "learning_rate": 6.500744922763272e-05, "loss": 2.9098, "step": 44626 }, { "epoch": 3.5, "learning_rate": 6.500666509840823e-05, "loss": 2.8657, "step": 44627 }, { "epoch": 3.5, "learning_rate": 6.500588096918373e-05, "loss": 2.5903, "step": 44628 }, { "epoch": 3.5, "learning_rate": 6.500509683995922e-05, "loss": 2.8879, "step": 44629 }, { "epoch": 3.5, "learning_rate": 6.500431271073472e-05, "loss": 2.7502, "step": 44630 }, { "epoch": 3.5, "learning_rate": 6.500352858151023e-05, "loss": 2.7258, "step": 44631 }, { "epoch": 3.5, "learning_rate": 6.500274445228573e-05, "loss": 2.6982, "step": 44632 }, { "epoch": 3.5, "learning_rate": 6.500196032306124e-05, "loss": 2.5012, "step": 44633 }, { "epoch": 3.5, "learning_rate": 6.500117619383674e-05, "loss": 2.5852, "step": 44634 }, { "epoch": 3.5, "learning_rate": 6.500039206461224e-05, "loss": 2.818, "step": 44635 }, { "epoch": 3.5, "learning_rate": 6.499960793538775e-05, "loss": 2.8226, "step": 44636 }, { "epoch": 3.5, "learning_rate": 6.499882380616325e-05, "loss": 2.9495, "step": 44637 }, { "epoch": 3.5, "learning_rate": 6.499803967693875e-05, "loss": 2.6939, "step": 44638 }, { "epoch": 3.5, "learning_rate": 6.499725554771427e-05, "loss": 2.6171, "step": 44639 }, { "epoch": 3.5, "learning_rate": 6.499647141848977e-05, "loss": 2.7942, "step": 44640 }, { "epoch": 3.5, "learning_rate": 6.499568728926528e-05, "loss": 2.7958, "step": 44641 }, { "epoch": 3.5, "learning_rate": 6.499490316004078e-05, "loss": 2.6201, "step": 44642 }, { "epoch": 3.5, "learning_rate": 6.499411903081629e-05, "loss": 2.9833, "step": 44643 }, { "epoch": 3.5, "learning_rate": 6.499333490159179e-05, "loss": 2.864, "step": 44644 }, { "epoch": 3.5, "learning_rate": 6.499255077236729e-05, "loss": 2.9309, "step": 44645 }, { "epoch": 3.5, "learning_rate": 6.49917666431428e-05, "loss": 2.5452, "step": 44646 }, { "epoch": 3.5, "learning_rate": 6.49909825139183e-05, "loss": 2.756, "step": 44647 }, { "epoch": 3.5, "learning_rate": 6.49901983846938e-05, "loss": 2.7373, "step": 44648 }, { "epoch": 3.5, "learning_rate": 6.498941425546931e-05, "loss": 2.6351, "step": 44649 }, { "epoch": 3.5, "learning_rate": 6.498863012624481e-05, "loss": 2.6465, "step": 44650 }, { "epoch": 3.5, "learning_rate": 6.498784599702031e-05, "loss": 2.657, "step": 44651 }, { "epoch": 3.5, "learning_rate": 6.498706186779582e-05, "loss": 2.7815, "step": 44652 }, { "epoch": 3.5, "learning_rate": 6.498627773857132e-05, "loss": 2.8867, "step": 44653 }, { "epoch": 3.5, "learning_rate": 6.498549360934683e-05, "loss": 3.0224, "step": 44654 }, { "epoch": 3.5, "learning_rate": 6.498470948012233e-05, "loss": 2.5959, "step": 44655 }, { "epoch": 3.5, "learning_rate": 6.498392535089783e-05, "loss": 2.8657, "step": 44656 }, { "epoch": 3.5, "learning_rate": 6.498314122167334e-05, "loss": 2.6289, "step": 44657 }, { "epoch": 3.5, "learning_rate": 6.498235709244884e-05, "loss": 2.804, "step": 44658 }, { "epoch": 3.5, "learning_rate": 6.498157296322434e-05, "loss": 2.8014, "step": 44659 }, { "epoch": 3.5, "learning_rate": 6.498078883399985e-05, "loss": 2.7293, "step": 44660 }, { "epoch": 3.5, "learning_rate": 6.498000470477535e-05, "loss": 2.6248, "step": 44661 }, { "epoch": 3.5, "learning_rate": 6.497922057555086e-05, "loss": 2.6677, "step": 44662 }, { "epoch": 3.5, "learning_rate": 6.497843644632636e-05, "loss": 2.6091, "step": 44663 }, { "epoch": 3.5, "learning_rate": 6.497765231710186e-05, "loss": 2.5523, "step": 44664 }, { "epoch": 3.5, "learning_rate": 6.497686818787737e-05, "loss": 2.5948, "step": 44665 }, { "epoch": 3.5, "learning_rate": 6.497608405865287e-05, "loss": 2.6371, "step": 44666 }, { "epoch": 3.5, "learning_rate": 6.497529992942837e-05, "loss": 2.8309, "step": 44667 }, { "epoch": 3.5, "learning_rate": 6.497451580020388e-05, "loss": 2.7223, "step": 44668 }, { "epoch": 3.5, "learning_rate": 6.497373167097938e-05, "loss": 3.0385, "step": 44669 }, { "epoch": 3.5, "learning_rate": 6.497294754175488e-05, "loss": 2.7024, "step": 44670 }, { "epoch": 3.5, "learning_rate": 6.497216341253039e-05, "loss": 2.7267, "step": 44671 }, { "epoch": 3.5, "learning_rate": 6.497137928330589e-05, "loss": 2.9518, "step": 44672 }, { "epoch": 3.5, "learning_rate": 6.49705951540814e-05, "loss": 2.7612, "step": 44673 }, { "epoch": 3.5, "learning_rate": 6.49698110248569e-05, "loss": 2.7134, "step": 44674 }, { "epoch": 3.5, "learning_rate": 6.49690268956324e-05, "loss": 2.5475, "step": 44675 }, { "epoch": 3.5, "learning_rate": 6.496824276640791e-05, "loss": 2.8335, "step": 44676 }, { "epoch": 3.5, "learning_rate": 6.496745863718341e-05, "loss": 2.9739, "step": 44677 }, { "epoch": 3.5, "learning_rate": 6.496667450795891e-05, "loss": 2.6511, "step": 44678 }, { "epoch": 3.5, "learning_rate": 6.496589037873442e-05, "loss": 2.8415, "step": 44679 }, { "epoch": 3.5, "learning_rate": 6.496510624950992e-05, "loss": 2.7882, "step": 44680 }, { "epoch": 3.5, "learning_rate": 6.496432212028543e-05, "loss": 2.6805, "step": 44681 }, { "epoch": 3.5, "learning_rate": 6.496353799106093e-05, "loss": 2.7315, "step": 44682 }, { "epoch": 3.5, "learning_rate": 6.496275386183643e-05, "loss": 2.9012, "step": 44683 }, { "epoch": 3.5, "learning_rate": 6.496196973261194e-05, "loss": 2.7701, "step": 44684 }, { "epoch": 3.5, "learning_rate": 6.496118560338744e-05, "loss": 2.751, "step": 44685 }, { "epoch": 3.5, "learning_rate": 6.496040147416294e-05, "loss": 2.8162, "step": 44686 }, { "epoch": 3.5, "learning_rate": 6.495961734493845e-05, "loss": 2.6634, "step": 44687 }, { "epoch": 3.5, "learning_rate": 6.495883321571395e-05, "loss": 2.7064, "step": 44688 }, { "epoch": 3.5, "learning_rate": 6.495804908648945e-05, "loss": 2.7831, "step": 44689 }, { "epoch": 3.5, "learning_rate": 6.495726495726496e-05, "loss": 2.6392, "step": 44690 }, { "epoch": 3.5, "learning_rate": 6.495648082804046e-05, "loss": 2.951, "step": 44691 }, { "epoch": 3.5, "learning_rate": 6.495569669881597e-05, "loss": 2.7436, "step": 44692 }, { "epoch": 3.5, "learning_rate": 6.495491256959147e-05, "loss": 2.8894, "step": 44693 }, { "epoch": 3.5, "learning_rate": 6.495412844036697e-05, "loss": 2.7522, "step": 44694 }, { "epoch": 3.5, "learning_rate": 6.495334431114248e-05, "loss": 2.5855, "step": 44695 }, { "epoch": 3.5, "learning_rate": 6.4952560181918e-05, "loss": 2.7623, "step": 44696 }, { "epoch": 3.5, "learning_rate": 6.49517760526935e-05, "loss": 2.7359, "step": 44697 }, { "epoch": 3.5, "learning_rate": 6.4950991923469e-05, "loss": 2.7999, "step": 44698 }, { "epoch": 3.5, "learning_rate": 6.49502077942445e-05, "loss": 2.6348, "step": 44699 }, { "epoch": 3.51, "learning_rate": 6.494942366502001e-05, "loss": 2.8731, "step": 44700 }, { "epoch": 3.51, "learning_rate": 6.49486395357955e-05, "loss": 2.8006, "step": 44701 }, { "epoch": 3.51, "learning_rate": 6.4947855406571e-05, "loss": 2.7579, "step": 44702 }, { "epoch": 3.51, "learning_rate": 6.49470712773465e-05, "loss": 2.7341, "step": 44703 }, { "epoch": 3.51, "learning_rate": 6.494628714812201e-05, "loss": 2.7205, "step": 44704 }, { "epoch": 3.51, "learning_rate": 6.494550301889751e-05, "loss": 2.9056, "step": 44705 }, { "epoch": 3.51, "learning_rate": 6.494471888967302e-05, "loss": 2.6921, "step": 44706 }, { "epoch": 3.51, "learning_rate": 6.494393476044852e-05, "loss": 2.6923, "step": 44707 }, { "epoch": 3.51, "learning_rate": 6.494315063122402e-05, "loss": 2.7369, "step": 44708 }, { "epoch": 3.51, "learning_rate": 6.494236650199953e-05, "loss": 2.7388, "step": 44709 }, { "epoch": 3.51, "learning_rate": 6.494158237277503e-05, "loss": 2.7502, "step": 44710 }, { "epoch": 3.51, "learning_rate": 6.494079824355054e-05, "loss": 2.6101, "step": 44711 }, { "epoch": 3.51, "learning_rate": 6.494001411432604e-05, "loss": 2.6013, "step": 44712 }, { "epoch": 3.51, "learning_rate": 6.493922998510154e-05, "loss": 2.8132, "step": 44713 }, { "epoch": 3.51, "learning_rate": 6.493844585587705e-05, "loss": 2.975, "step": 44714 }, { "epoch": 3.51, "learning_rate": 6.493766172665255e-05, "loss": 2.6526, "step": 44715 }, { "epoch": 3.51, "learning_rate": 6.493687759742805e-05, "loss": 3.0027, "step": 44716 }, { "epoch": 3.51, "learning_rate": 6.493609346820356e-05, "loss": 2.5737, "step": 44717 }, { "epoch": 3.51, "learning_rate": 6.493530933897906e-05, "loss": 2.7063, "step": 44718 }, { "epoch": 3.51, "learning_rate": 6.493452520975456e-05, "loss": 2.6418, "step": 44719 }, { "epoch": 3.51, "learning_rate": 6.493374108053007e-05, "loss": 2.8086, "step": 44720 }, { "epoch": 3.51, "learning_rate": 6.493295695130559e-05, "loss": 2.7317, "step": 44721 }, { "epoch": 3.51, "learning_rate": 6.493217282208109e-05, "loss": 2.8935, "step": 44722 }, { "epoch": 3.51, "learning_rate": 6.493138869285659e-05, "loss": 2.7494, "step": 44723 }, { "epoch": 3.51, "learning_rate": 6.49306045636321e-05, "loss": 2.7005, "step": 44724 }, { "epoch": 3.51, "learning_rate": 6.49298204344076e-05, "loss": 2.6878, "step": 44725 }, { "epoch": 3.51, "learning_rate": 6.49290363051831e-05, "loss": 2.7431, "step": 44726 }, { "epoch": 3.51, "learning_rate": 6.492825217595861e-05, "loss": 2.8729, "step": 44727 }, { "epoch": 3.51, "learning_rate": 6.492746804673411e-05, "loss": 2.98, "step": 44728 }, { "epoch": 3.51, "learning_rate": 6.492668391750962e-05, "loss": 2.7005, "step": 44729 }, { "epoch": 3.51, "learning_rate": 6.492589978828512e-05, "loss": 2.9845, "step": 44730 }, { "epoch": 3.51, "learning_rate": 6.492511565906062e-05, "loss": 2.619, "step": 44731 }, { "epoch": 3.51, "learning_rate": 6.492433152983613e-05, "loss": 2.8873, "step": 44732 }, { "epoch": 3.51, "learning_rate": 6.492354740061163e-05, "loss": 2.8607, "step": 44733 }, { "epoch": 3.51, "learning_rate": 6.492276327138713e-05, "loss": 2.8639, "step": 44734 }, { "epoch": 3.51, "learning_rate": 6.492197914216264e-05, "loss": 2.9026, "step": 44735 }, { "epoch": 3.51, "learning_rate": 6.492119501293814e-05, "loss": 2.8832, "step": 44736 }, { "epoch": 3.51, "learning_rate": 6.492041088371363e-05, "loss": 2.6378, "step": 44737 }, { "epoch": 3.51, "learning_rate": 6.491962675448913e-05, "loss": 2.5961, "step": 44738 }, { "epoch": 3.51, "learning_rate": 6.491884262526464e-05, "loss": 2.6858, "step": 44739 }, { "epoch": 3.51, "learning_rate": 6.491805849604014e-05, "loss": 2.7144, "step": 44740 }, { "epoch": 3.51, "learning_rate": 6.491727436681565e-05, "loss": 2.5334, "step": 44741 }, { "epoch": 3.51, "learning_rate": 6.491649023759115e-05, "loss": 3.0056, "step": 44742 }, { "epoch": 3.51, "learning_rate": 6.491570610836665e-05, "loss": 2.8125, "step": 44743 }, { "epoch": 3.51, "learning_rate": 6.491492197914216e-05, "loss": 3.1164, "step": 44744 }, { "epoch": 3.51, "learning_rate": 6.491413784991766e-05, "loss": 2.5596, "step": 44745 }, { "epoch": 3.51, "learning_rate": 6.491335372069316e-05, "loss": 2.5597, "step": 44746 }, { "epoch": 3.51, "learning_rate": 6.491256959146868e-05, "loss": 2.7994, "step": 44747 }, { "epoch": 3.51, "learning_rate": 6.491178546224419e-05, "loss": 2.7416, "step": 44748 }, { "epoch": 3.51, "learning_rate": 6.491100133301969e-05, "loss": 2.7474, "step": 44749 }, { "epoch": 3.51, "learning_rate": 6.491021720379519e-05, "loss": 2.6071, "step": 44750 }, { "epoch": 3.51, "learning_rate": 6.49094330745707e-05, "loss": 2.7446, "step": 44751 }, { "epoch": 3.51, "learning_rate": 6.49086489453462e-05, "loss": 2.8245, "step": 44752 }, { "epoch": 3.51, "learning_rate": 6.49078648161217e-05, "loss": 2.9005, "step": 44753 }, { "epoch": 3.51, "learning_rate": 6.490708068689721e-05, "loss": 2.6125, "step": 44754 }, { "epoch": 3.51, "learning_rate": 6.490629655767271e-05, "loss": 2.8001, "step": 44755 }, { "epoch": 3.51, "learning_rate": 6.490551242844821e-05, "loss": 2.7811, "step": 44756 }, { "epoch": 3.51, "learning_rate": 6.490472829922372e-05, "loss": 2.6094, "step": 44757 }, { "epoch": 3.51, "learning_rate": 6.490394416999922e-05, "loss": 2.5536, "step": 44758 }, { "epoch": 3.51, "learning_rate": 6.490316004077473e-05, "loss": 2.7348, "step": 44759 }, { "epoch": 3.51, "learning_rate": 6.490237591155023e-05, "loss": 2.7026, "step": 44760 }, { "epoch": 3.51, "learning_rate": 6.490159178232573e-05, "loss": 2.8113, "step": 44761 }, { "epoch": 3.51, "learning_rate": 6.490080765310124e-05, "loss": 2.6909, "step": 44762 }, { "epoch": 3.51, "learning_rate": 6.490002352387674e-05, "loss": 2.9554, "step": 44763 }, { "epoch": 3.51, "learning_rate": 6.489923939465224e-05, "loss": 2.8855, "step": 44764 }, { "epoch": 3.51, "learning_rate": 6.489845526542775e-05, "loss": 2.8482, "step": 44765 }, { "epoch": 3.51, "learning_rate": 6.489767113620325e-05, "loss": 2.6539, "step": 44766 }, { "epoch": 3.51, "learning_rate": 6.489688700697876e-05, "loss": 2.6227, "step": 44767 }, { "epoch": 3.51, "learning_rate": 6.489610287775426e-05, "loss": 2.8635, "step": 44768 }, { "epoch": 3.51, "learning_rate": 6.489531874852976e-05, "loss": 2.7347, "step": 44769 }, { "epoch": 3.51, "learning_rate": 6.489453461930527e-05, "loss": 2.6733, "step": 44770 }, { "epoch": 3.51, "learning_rate": 6.489375049008077e-05, "loss": 2.6919, "step": 44771 }, { "epoch": 3.51, "learning_rate": 6.489296636085627e-05, "loss": 2.7816, "step": 44772 }, { "epoch": 3.51, "learning_rate": 6.489218223163178e-05, "loss": 2.5931, "step": 44773 }, { "epoch": 3.51, "learning_rate": 6.489139810240728e-05, "loss": 2.64, "step": 44774 }, { "epoch": 3.51, "learning_rate": 6.489061397318278e-05, "loss": 2.7701, "step": 44775 }, { "epoch": 3.51, "learning_rate": 6.488982984395829e-05, "loss": 2.6627, "step": 44776 }, { "epoch": 3.51, "learning_rate": 6.488904571473379e-05, "loss": 2.5598, "step": 44777 }, { "epoch": 3.51, "learning_rate": 6.48882615855093e-05, "loss": 2.719, "step": 44778 }, { "epoch": 3.51, "learning_rate": 6.48874774562848e-05, "loss": 2.8551, "step": 44779 }, { "epoch": 3.51, "learning_rate": 6.48866933270603e-05, "loss": 2.5054, "step": 44780 }, { "epoch": 3.51, "learning_rate": 6.48859091978358e-05, "loss": 2.8169, "step": 44781 }, { "epoch": 3.51, "learning_rate": 6.488512506861131e-05, "loss": 2.884, "step": 44782 }, { "epoch": 3.51, "learning_rate": 6.488434093938681e-05, "loss": 2.7291, "step": 44783 }, { "epoch": 3.51, "learning_rate": 6.488355681016232e-05, "loss": 2.8516, "step": 44784 }, { "epoch": 3.51, "learning_rate": 6.488277268093782e-05, "loss": 2.7329, "step": 44785 }, { "epoch": 3.51, "learning_rate": 6.488198855171332e-05, "loss": 2.8119, "step": 44786 }, { "epoch": 3.51, "learning_rate": 6.488120442248883e-05, "loss": 2.3982, "step": 44787 }, { "epoch": 3.51, "learning_rate": 6.488042029326433e-05, "loss": 2.6824, "step": 44788 }, { "epoch": 3.51, "learning_rate": 6.487963616403984e-05, "loss": 2.837, "step": 44789 }, { "epoch": 3.51, "learning_rate": 6.487885203481534e-05, "loss": 2.5338, "step": 44790 }, { "epoch": 3.51, "learning_rate": 6.487806790559084e-05, "loss": 2.7381, "step": 44791 }, { "epoch": 3.51, "learning_rate": 6.487728377636635e-05, "loss": 2.9445, "step": 44792 }, { "epoch": 3.51, "learning_rate": 6.487649964714185e-05, "loss": 2.6278, "step": 44793 }, { "epoch": 3.51, "learning_rate": 6.487571551791735e-05, "loss": 2.4942, "step": 44794 }, { "epoch": 3.51, "learning_rate": 6.487493138869286e-05, "loss": 2.6223, "step": 44795 }, { "epoch": 3.51, "learning_rate": 6.487414725946836e-05, "loss": 3.0772, "step": 44796 }, { "epoch": 3.51, "learning_rate": 6.487336313024387e-05, "loss": 2.6917, "step": 44797 }, { "epoch": 3.51, "learning_rate": 6.487257900101937e-05, "loss": 2.6863, "step": 44798 }, { "epoch": 3.51, "learning_rate": 6.487179487179487e-05, "loss": 2.5018, "step": 44799 }, { "epoch": 3.51, "learning_rate": 6.487101074257038e-05, "loss": 2.8339, "step": 44800 }, { "epoch": 3.51, "learning_rate": 6.487022661334588e-05, "loss": 2.8502, "step": 44801 }, { "epoch": 3.51, "learning_rate": 6.486944248412138e-05, "loss": 2.6328, "step": 44802 }, { "epoch": 3.51, "learning_rate": 6.486865835489689e-05, "loss": 2.8846, "step": 44803 }, { "epoch": 3.51, "learning_rate": 6.48678742256724e-05, "loss": 2.5499, "step": 44804 }, { "epoch": 3.51, "learning_rate": 6.486709009644791e-05, "loss": 2.6869, "step": 44805 }, { "epoch": 3.51, "learning_rate": 6.486630596722341e-05, "loss": 2.9479, "step": 44806 }, { "epoch": 3.51, "learning_rate": 6.486552183799892e-05, "loss": 2.6706, "step": 44807 }, { "epoch": 3.51, "learning_rate": 6.486473770877442e-05, "loss": 2.8283, "step": 44808 }, { "epoch": 3.51, "learning_rate": 6.486395357954991e-05, "loss": 2.75, "step": 44809 }, { "epoch": 3.51, "learning_rate": 6.486316945032541e-05, "loss": 2.7169, "step": 44810 }, { "epoch": 3.51, "learning_rate": 6.486238532110092e-05, "loss": 2.7058, "step": 44811 }, { "epoch": 3.51, "learning_rate": 6.486160119187642e-05, "loss": 2.7934, "step": 44812 }, { "epoch": 3.51, "learning_rate": 6.486081706265192e-05, "loss": 2.6554, "step": 44813 }, { "epoch": 3.51, "learning_rate": 6.486003293342743e-05, "loss": 2.7501, "step": 44814 }, { "epoch": 3.51, "learning_rate": 6.485924880420293e-05, "loss": 2.8288, "step": 44815 }, { "epoch": 3.51, "learning_rate": 6.485846467497844e-05, "loss": 2.8783, "step": 44816 }, { "epoch": 3.51, "learning_rate": 6.485768054575394e-05, "loss": 2.5041, "step": 44817 }, { "epoch": 3.51, "learning_rate": 6.485689641652944e-05, "loss": 2.7313, "step": 44818 }, { "epoch": 3.51, "learning_rate": 6.485611228730495e-05, "loss": 2.8053, "step": 44819 }, { "epoch": 3.51, "learning_rate": 6.485532815808045e-05, "loss": 2.8412, "step": 44820 }, { "epoch": 3.51, "learning_rate": 6.485454402885595e-05, "loss": 2.7763, "step": 44821 }, { "epoch": 3.51, "learning_rate": 6.485375989963146e-05, "loss": 2.8672, "step": 44822 }, { "epoch": 3.51, "learning_rate": 6.485297577040696e-05, "loss": 2.7432, "step": 44823 }, { "epoch": 3.51, "learning_rate": 6.485219164118246e-05, "loss": 2.881, "step": 44824 }, { "epoch": 3.51, "learning_rate": 6.485140751195797e-05, "loss": 2.6582, "step": 44825 }, { "epoch": 3.51, "learning_rate": 6.485062338273347e-05, "loss": 2.7291, "step": 44826 }, { "epoch": 3.52, "learning_rate": 6.484983925350898e-05, "loss": 2.716, "step": 44827 }, { "epoch": 3.52, "learning_rate": 6.484905512428448e-05, "loss": 2.7226, "step": 44828 }, { "epoch": 3.52, "learning_rate": 6.484827099505998e-05, "loss": 2.7852, "step": 44829 }, { "epoch": 3.52, "learning_rate": 6.48474868658355e-05, "loss": 2.7782, "step": 44830 }, { "epoch": 3.52, "learning_rate": 6.4846702736611e-05, "loss": 2.987, "step": 44831 }, { "epoch": 3.52, "learning_rate": 6.484591860738651e-05, "loss": 2.7333, "step": 44832 }, { "epoch": 3.52, "learning_rate": 6.484513447816201e-05, "loss": 2.7152, "step": 44833 }, { "epoch": 3.52, "learning_rate": 6.484435034893751e-05, "loss": 2.8462, "step": 44834 }, { "epoch": 3.52, "learning_rate": 6.484356621971302e-05, "loss": 2.6449, "step": 44835 }, { "epoch": 3.52, "learning_rate": 6.484278209048852e-05, "loss": 2.7635, "step": 44836 }, { "epoch": 3.52, "learning_rate": 6.484199796126403e-05, "loss": 2.8731, "step": 44837 }, { "epoch": 3.52, "learning_rate": 6.484121383203953e-05, "loss": 2.6216, "step": 44838 }, { "epoch": 3.52, "learning_rate": 6.484042970281503e-05, "loss": 2.5696, "step": 44839 }, { "epoch": 3.52, "learning_rate": 6.483964557359054e-05, "loss": 2.583, "step": 44840 }, { "epoch": 3.52, "learning_rate": 6.483886144436604e-05, "loss": 2.9086, "step": 44841 }, { "epoch": 3.52, "learning_rate": 6.483807731514154e-05, "loss": 2.7017, "step": 44842 }, { "epoch": 3.52, "learning_rate": 6.483729318591705e-05, "loss": 2.6962, "step": 44843 }, { "epoch": 3.52, "learning_rate": 6.483650905669255e-05, "loss": 2.7821, "step": 44844 }, { "epoch": 3.52, "learning_rate": 6.483572492746804e-05, "loss": 2.7273, "step": 44845 }, { "epoch": 3.52, "learning_rate": 6.483494079824355e-05, "loss": 2.8588, "step": 44846 }, { "epoch": 3.52, "learning_rate": 6.483415666901905e-05, "loss": 2.6874, "step": 44847 }, { "epoch": 3.52, "learning_rate": 6.483337253979455e-05, "loss": 2.8356, "step": 44848 }, { "epoch": 3.52, "learning_rate": 6.483258841057006e-05, "loss": 2.7289, "step": 44849 }, { "epoch": 3.52, "learning_rate": 6.483180428134556e-05, "loss": 2.8271, "step": 44850 }, { "epoch": 3.52, "learning_rate": 6.483102015212106e-05, "loss": 2.6714, "step": 44851 }, { "epoch": 3.52, "learning_rate": 6.483023602289657e-05, "loss": 2.7422, "step": 44852 }, { "epoch": 3.52, "learning_rate": 6.482945189367207e-05, "loss": 2.5476, "step": 44853 }, { "epoch": 3.52, "learning_rate": 6.482866776444757e-05, "loss": 2.6316, "step": 44854 }, { "epoch": 3.52, "learning_rate": 6.482788363522309e-05, "loss": 2.697, "step": 44855 }, { "epoch": 3.52, "learning_rate": 6.48270995059986e-05, "loss": 2.8482, "step": 44856 }, { "epoch": 3.52, "learning_rate": 6.48263153767741e-05, "loss": 2.6145, "step": 44857 }, { "epoch": 3.52, "learning_rate": 6.48255312475496e-05, "loss": 2.767, "step": 44858 }, { "epoch": 3.52, "learning_rate": 6.482474711832511e-05, "loss": 2.7661, "step": 44859 }, { "epoch": 3.52, "learning_rate": 6.482396298910061e-05, "loss": 2.7711, "step": 44860 }, { "epoch": 3.52, "learning_rate": 6.482317885987611e-05, "loss": 2.648, "step": 44861 }, { "epoch": 3.52, "learning_rate": 6.482239473065162e-05, "loss": 2.6654, "step": 44862 }, { "epoch": 3.52, "learning_rate": 6.482161060142712e-05, "loss": 2.624, "step": 44863 }, { "epoch": 3.52, "learning_rate": 6.482082647220263e-05, "loss": 2.5819, "step": 44864 }, { "epoch": 3.52, "learning_rate": 6.482004234297813e-05, "loss": 2.7104, "step": 44865 }, { "epoch": 3.52, "learning_rate": 6.481925821375363e-05, "loss": 2.4327, "step": 44866 }, { "epoch": 3.52, "learning_rate": 6.481847408452914e-05, "loss": 2.6469, "step": 44867 }, { "epoch": 3.52, "learning_rate": 6.481768995530464e-05, "loss": 2.6496, "step": 44868 }, { "epoch": 3.52, "learning_rate": 6.481690582608014e-05, "loss": 2.6882, "step": 44869 }, { "epoch": 3.52, "learning_rate": 6.481612169685565e-05, "loss": 2.9301, "step": 44870 }, { "epoch": 3.52, "learning_rate": 6.481533756763115e-05, "loss": 2.6871, "step": 44871 }, { "epoch": 3.52, "learning_rate": 6.481455343840665e-05, "loss": 2.5967, "step": 44872 }, { "epoch": 3.52, "learning_rate": 6.481376930918216e-05, "loss": 2.6802, "step": 44873 }, { "epoch": 3.52, "learning_rate": 6.481298517995766e-05, "loss": 2.8582, "step": 44874 }, { "epoch": 3.52, "learning_rate": 6.481220105073317e-05, "loss": 2.5304, "step": 44875 }, { "epoch": 3.52, "learning_rate": 6.481141692150867e-05, "loss": 2.6859, "step": 44876 }, { "epoch": 3.52, "learning_rate": 6.481063279228417e-05, "loss": 2.5833, "step": 44877 }, { "epoch": 3.52, "learning_rate": 6.480984866305968e-05, "loss": 2.7113, "step": 44878 }, { "epoch": 3.52, "learning_rate": 6.480906453383518e-05, "loss": 2.815, "step": 44879 }, { "epoch": 3.52, "learning_rate": 6.480828040461068e-05, "loss": 2.6933, "step": 44880 }, { "epoch": 3.52, "learning_rate": 6.480749627538619e-05, "loss": 2.7257, "step": 44881 }, { "epoch": 3.52, "learning_rate": 6.480671214616169e-05, "loss": 2.5986, "step": 44882 }, { "epoch": 3.52, "learning_rate": 6.48059280169372e-05, "loss": 2.5496, "step": 44883 }, { "epoch": 3.52, "learning_rate": 6.48051438877127e-05, "loss": 2.8609, "step": 44884 }, { "epoch": 3.52, "learning_rate": 6.48043597584882e-05, "loss": 2.9099, "step": 44885 }, { "epoch": 3.52, "learning_rate": 6.48035756292637e-05, "loss": 2.7388, "step": 44886 }, { "epoch": 3.52, "learning_rate": 6.480279150003921e-05, "loss": 2.8673, "step": 44887 }, { "epoch": 3.52, "learning_rate": 6.480200737081471e-05, "loss": 2.8164, "step": 44888 }, { "epoch": 3.52, "learning_rate": 6.480122324159022e-05, "loss": 2.7312, "step": 44889 }, { "epoch": 3.52, "learning_rate": 6.480043911236572e-05, "loss": 2.8895, "step": 44890 }, { "epoch": 3.52, "learning_rate": 6.479965498314122e-05, "loss": 2.6578, "step": 44891 }, { "epoch": 3.52, "learning_rate": 6.479887085391673e-05, "loss": 2.5725, "step": 44892 }, { "epoch": 3.52, "learning_rate": 6.479808672469223e-05, "loss": 2.6846, "step": 44893 }, { "epoch": 3.52, "learning_rate": 6.479730259546774e-05, "loss": 2.5399, "step": 44894 }, { "epoch": 3.52, "learning_rate": 6.479651846624324e-05, "loss": 2.7401, "step": 44895 }, { "epoch": 3.52, "learning_rate": 6.479573433701874e-05, "loss": 2.7332, "step": 44896 }, { "epoch": 3.52, "learning_rate": 6.479495020779425e-05, "loss": 2.6722, "step": 44897 }, { "epoch": 3.52, "learning_rate": 6.479416607856975e-05, "loss": 2.5708, "step": 44898 }, { "epoch": 3.52, "learning_rate": 6.479338194934525e-05, "loss": 2.6969, "step": 44899 }, { "epoch": 3.52, "learning_rate": 6.479259782012076e-05, "loss": 2.6373, "step": 44900 }, { "epoch": 3.52, "learning_rate": 6.479181369089626e-05, "loss": 2.5941, "step": 44901 }, { "epoch": 3.52, "learning_rate": 6.479102956167177e-05, "loss": 2.8146, "step": 44902 }, { "epoch": 3.52, "learning_rate": 6.479024543244727e-05, "loss": 2.9982, "step": 44903 }, { "epoch": 3.52, "learning_rate": 6.478946130322277e-05, "loss": 2.8903, "step": 44904 }, { "epoch": 3.52, "learning_rate": 6.478867717399828e-05, "loss": 2.7687, "step": 44905 }, { "epoch": 3.52, "learning_rate": 6.478789304477378e-05, "loss": 3.0812, "step": 44906 }, { "epoch": 3.52, "learning_rate": 6.478710891554928e-05, "loss": 2.9076, "step": 44907 }, { "epoch": 3.52, "learning_rate": 6.478632478632479e-05, "loss": 2.7221, "step": 44908 }, { "epoch": 3.52, "learning_rate": 6.478554065710029e-05, "loss": 2.7692, "step": 44909 }, { "epoch": 3.52, "learning_rate": 6.47847565278758e-05, "loss": 2.9564, "step": 44910 }, { "epoch": 3.52, "learning_rate": 6.47839723986513e-05, "loss": 2.7709, "step": 44911 }, { "epoch": 3.52, "learning_rate": 6.478318826942682e-05, "loss": 2.8276, "step": 44912 }, { "epoch": 3.52, "learning_rate": 6.478240414020232e-05, "loss": 2.5813, "step": 44913 }, { "epoch": 3.52, "learning_rate": 6.478162001097782e-05, "loss": 2.7146, "step": 44914 }, { "epoch": 3.52, "learning_rate": 6.478083588175333e-05, "loss": 2.7095, "step": 44915 }, { "epoch": 3.52, "learning_rate": 6.478005175252883e-05, "loss": 2.8497, "step": 44916 }, { "epoch": 3.52, "learning_rate": 6.477926762330432e-05, "loss": 2.5332, "step": 44917 }, { "epoch": 3.52, "learning_rate": 6.477848349407982e-05, "loss": 2.7462, "step": 44918 }, { "epoch": 3.52, "learning_rate": 6.477769936485533e-05, "loss": 2.8018, "step": 44919 }, { "epoch": 3.52, "learning_rate": 6.477691523563083e-05, "loss": 2.6722, "step": 44920 }, { "epoch": 3.52, "learning_rate": 6.477613110640633e-05, "loss": 2.7316, "step": 44921 }, { "epoch": 3.52, "learning_rate": 6.477534697718184e-05, "loss": 2.7021, "step": 44922 }, { "epoch": 3.52, "learning_rate": 6.477456284795734e-05, "loss": 2.8881, "step": 44923 }, { "epoch": 3.52, "learning_rate": 6.477377871873285e-05, "loss": 2.6889, "step": 44924 }, { "epoch": 3.52, "learning_rate": 6.477299458950835e-05, "loss": 2.5533, "step": 44925 }, { "epoch": 3.52, "learning_rate": 6.477221046028385e-05, "loss": 2.6497, "step": 44926 }, { "epoch": 3.52, "learning_rate": 6.477142633105936e-05, "loss": 2.7002, "step": 44927 }, { "epoch": 3.52, "learning_rate": 6.477064220183486e-05, "loss": 2.7269, "step": 44928 }, { "epoch": 3.52, "learning_rate": 6.476985807261036e-05, "loss": 2.65, "step": 44929 }, { "epoch": 3.52, "learning_rate": 6.476907394338587e-05, "loss": 2.8526, "step": 44930 }, { "epoch": 3.52, "learning_rate": 6.476828981416137e-05, "loss": 2.7661, "step": 44931 }, { "epoch": 3.52, "learning_rate": 6.476750568493688e-05, "loss": 2.9184, "step": 44932 }, { "epoch": 3.52, "learning_rate": 6.476672155571238e-05, "loss": 2.6439, "step": 44933 }, { "epoch": 3.52, "learning_rate": 6.476593742648788e-05, "loss": 2.6864, "step": 44934 }, { "epoch": 3.52, "learning_rate": 6.476515329726339e-05, "loss": 2.6952, "step": 44935 }, { "epoch": 3.52, "learning_rate": 6.476436916803889e-05, "loss": 2.7379, "step": 44936 }, { "epoch": 3.52, "learning_rate": 6.47635850388144e-05, "loss": 2.8582, "step": 44937 }, { "epoch": 3.52, "learning_rate": 6.476280090958991e-05, "loss": 2.4759, "step": 44938 }, { "epoch": 3.52, "learning_rate": 6.476201678036541e-05, "loss": 2.6147, "step": 44939 }, { "epoch": 3.52, "learning_rate": 6.476123265114092e-05, "loss": 2.7576, "step": 44940 }, { "epoch": 3.52, "learning_rate": 6.476044852191642e-05, "loss": 2.8305, "step": 44941 }, { "epoch": 3.52, "learning_rate": 6.475966439269193e-05, "loss": 2.7388, "step": 44942 }, { "epoch": 3.52, "learning_rate": 6.475888026346743e-05, "loss": 2.6912, "step": 44943 }, { "epoch": 3.52, "learning_rate": 6.475809613424293e-05, "loss": 2.8018, "step": 44944 }, { "epoch": 3.52, "learning_rate": 6.475731200501844e-05, "loss": 3.0931, "step": 44945 }, { "epoch": 3.52, "learning_rate": 6.475652787579394e-05, "loss": 2.7016, "step": 44946 }, { "epoch": 3.52, "learning_rate": 6.475574374656944e-05, "loss": 2.8765, "step": 44947 }, { "epoch": 3.52, "learning_rate": 6.475495961734495e-05, "loss": 2.7125, "step": 44948 }, { "epoch": 3.52, "learning_rate": 6.475417548812045e-05, "loss": 2.6061, "step": 44949 }, { "epoch": 3.52, "learning_rate": 6.475339135889596e-05, "loss": 2.5373, "step": 44950 }, { "epoch": 3.52, "learning_rate": 6.475260722967146e-05, "loss": 2.5872, "step": 44951 }, { "epoch": 3.52, "learning_rate": 6.475182310044696e-05, "loss": 2.7247, "step": 44952 }, { "epoch": 3.52, "learning_rate": 6.475103897122245e-05, "loss": 2.7274, "step": 44953 }, { "epoch": 3.52, "learning_rate": 6.475025484199796e-05, "loss": 2.5123, "step": 44954 }, { "epoch": 3.53, "learning_rate": 6.474947071277346e-05, "loss": 2.5776, "step": 44955 }, { "epoch": 3.53, "learning_rate": 6.474868658354896e-05, "loss": 2.7314, "step": 44956 }, { "epoch": 3.53, "learning_rate": 6.474790245432447e-05, "loss": 2.9507, "step": 44957 }, { "epoch": 3.53, "learning_rate": 6.474711832509997e-05, "loss": 2.7938, "step": 44958 }, { "epoch": 3.53, "learning_rate": 6.474633419587547e-05, "loss": 2.7387, "step": 44959 }, { "epoch": 3.53, "learning_rate": 6.474555006665098e-05, "loss": 2.6661, "step": 44960 }, { "epoch": 3.53, "learning_rate": 6.474476593742648e-05, "loss": 2.6963, "step": 44961 }, { "epoch": 3.53, "learning_rate": 6.474398180820199e-05, "loss": 2.6049, "step": 44962 }, { "epoch": 3.53, "learning_rate": 6.47431976789775e-05, "loss": 2.7029, "step": 44963 }, { "epoch": 3.53, "learning_rate": 6.4742413549753e-05, "loss": 2.5682, "step": 44964 }, { "epoch": 3.53, "learning_rate": 6.474162942052851e-05, "loss": 2.8107, "step": 44965 }, { "epoch": 3.53, "learning_rate": 6.474084529130401e-05, "loss": 2.6889, "step": 44966 }, { "epoch": 3.53, "learning_rate": 6.474006116207952e-05, "loss": 2.6867, "step": 44967 }, { "epoch": 3.53, "learning_rate": 6.473927703285502e-05, "loss": 2.8412, "step": 44968 }, { "epoch": 3.53, "learning_rate": 6.473849290363053e-05, "loss": 2.6632, "step": 44969 }, { "epoch": 3.53, "learning_rate": 6.473770877440603e-05, "loss": 2.685, "step": 44970 }, { "epoch": 3.53, "learning_rate": 6.473692464518153e-05, "loss": 2.6807, "step": 44971 }, { "epoch": 3.53, "learning_rate": 6.473614051595704e-05, "loss": 2.7182, "step": 44972 }, { "epoch": 3.53, "learning_rate": 6.473535638673254e-05, "loss": 2.6367, "step": 44973 }, { "epoch": 3.53, "learning_rate": 6.473457225750804e-05, "loss": 2.7491, "step": 44974 }, { "epoch": 3.53, "learning_rate": 6.473378812828355e-05, "loss": 2.6313, "step": 44975 }, { "epoch": 3.53, "learning_rate": 6.473300399905905e-05, "loss": 2.7522, "step": 44976 }, { "epoch": 3.53, "learning_rate": 6.473221986983455e-05, "loss": 2.6622, "step": 44977 }, { "epoch": 3.53, "learning_rate": 6.473143574061006e-05, "loss": 2.6633, "step": 44978 }, { "epoch": 3.53, "learning_rate": 6.473065161138556e-05, "loss": 3.1014, "step": 44979 }, { "epoch": 3.53, "learning_rate": 6.472986748216107e-05, "loss": 2.5026, "step": 44980 }, { "epoch": 3.53, "learning_rate": 6.472908335293657e-05, "loss": 2.7076, "step": 44981 }, { "epoch": 3.53, "learning_rate": 6.472829922371207e-05, "loss": 2.6355, "step": 44982 }, { "epoch": 3.53, "learning_rate": 6.472751509448758e-05, "loss": 2.6674, "step": 44983 }, { "epoch": 3.53, "learning_rate": 6.472673096526308e-05, "loss": 2.5196, "step": 44984 }, { "epoch": 3.53, "learning_rate": 6.472594683603858e-05, "loss": 2.7574, "step": 44985 }, { "epoch": 3.53, "learning_rate": 6.472516270681409e-05, "loss": 2.8636, "step": 44986 }, { "epoch": 3.53, "learning_rate": 6.472437857758959e-05, "loss": 2.5615, "step": 44987 }, { "epoch": 3.53, "learning_rate": 6.47235944483651e-05, "loss": 2.8669, "step": 44988 }, { "epoch": 3.53, "learning_rate": 6.47228103191406e-05, "loss": 2.6142, "step": 44989 }, { "epoch": 3.53, "learning_rate": 6.47220261899161e-05, "loss": 2.9539, "step": 44990 }, { "epoch": 3.53, "learning_rate": 6.47212420606916e-05, "loss": 2.5675, "step": 44991 }, { "epoch": 3.53, "learning_rate": 6.472045793146711e-05, "loss": 2.6914, "step": 44992 }, { "epoch": 3.53, "learning_rate": 6.471967380224261e-05, "loss": 2.713, "step": 44993 }, { "epoch": 3.53, "learning_rate": 6.471888967301812e-05, "loss": 2.7231, "step": 44994 }, { "epoch": 3.53, "learning_rate": 6.471810554379362e-05, "loss": 2.6004, "step": 44995 }, { "epoch": 3.53, "learning_rate": 6.471732141456912e-05, "loss": 3.0482, "step": 44996 }, { "epoch": 3.53, "learning_rate": 6.471653728534463e-05, "loss": 2.7621, "step": 44997 }, { "epoch": 3.53, "learning_rate": 6.471575315612013e-05, "loss": 2.6354, "step": 44998 }, { "epoch": 3.53, "learning_rate": 6.471496902689564e-05, "loss": 2.6783, "step": 44999 }, { "epoch": 3.53, "learning_rate": 6.471418489767114e-05, "loss": 2.9731, "step": 45000 }, { "epoch": 3.53, "eval_loss": 2.8939766883850098, "eval_runtime": 218.0025, "eval_samples_per_second": 51.981, "eval_steps_per_second": 3.252, "step": 45000 }, { "epoch": 3.53, "learning_rate": 6.471340076844664e-05, "loss": 2.7703, "step": 45001 }, { "epoch": 3.53, "learning_rate": 6.471261663922215e-05, "loss": 2.6364, "step": 45002 }, { "epoch": 3.53, "learning_rate": 6.471183250999765e-05, "loss": 2.7352, "step": 45003 }, { "epoch": 3.53, "learning_rate": 6.471104838077315e-05, "loss": 2.9038, "step": 45004 }, { "epoch": 3.53, "learning_rate": 6.471026425154866e-05, "loss": 2.6987, "step": 45005 }, { "epoch": 3.53, "learning_rate": 6.470948012232416e-05, "loss": 2.9006, "step": 45006 }, { "epoch": 3.53, "learning_rate": 6.470869599309966e-05, "loss": 2.6986, "step": 45007 }, { "epoch": 3.53, "learning_rate": 6.470791186387517e-05, "loss": 2.6287, "step": 45008 }, { "epoch": 3.53, "learning_rate": 6.470712773465067e-05, "loss": 2.8018, "step": 45009 }, { "epoch": 3.53, "learning_rate": 6.470634360542618e-05, "loss": 2.6902, "step": 45010 }, { "epoch": 3.53, "learning_rate": 6.470555947620168e-05, "loss": 2.7592, "step": 45011 }, { "epoch": 3.53, "learning_rate": 6.470477534697718e-05, "loss": 2.7065, "step": 45012 }, { "epoch": 3.53, "learning_rate": 6.470399121775269e-05, "loss": 2.9621, "step": 45013 }, { "epoch": 3.53, "learning_rate": 6.470320708852819e-05, "loss": 2.6676, "step": 45014 }, { "epoch": 3.53, "learning_rate": 6.47024229593037e-05, "loss": 2.8523, "step": 45015 }, { "epoch": 3.53, "learning_rate": 6.47016388300792e-05, "loss": 2.8051, "step": 45016 }, { "epoch": 3.53, "learning_rate": 6.47008547008547e-05, "loss": 2.8046, "step": 45017 }, { "epoch": 3.53, "learning_rate": 6.47000705716302e-05, "loss": 2.7247, "step": 45018 }, { "epoch": 3.53, "learning_rate": 6.469928644240571e-05, "loss": 2.7588, "step": 45019 }, { "epoch": 3.53, "learning_rate": 6.469850231318123e-05, "loss": 2.9367, "step": 45020 }, { "epoch": 3.53, "learning_rate": 6.469771818395673e-05, "loss": 2.5578, "step": 45021 }, { "epoch": 3.53, "learning_rate": 6.469693405473223e-05, "loss": 2.7624, "step": 45022 }, { "epoch": 3.53, "learning_rate": 6.469614992550774e-05, "loss": 2.6864, "step": 45023 }, { "epoch": 3.53, "learning_rate": 6.469536579628324e-05, "loss": 2.8902, "step": 45024 }, { "epoch": 3.53, "learning_rate": 6.469458166705873e-05, "loss": 2.6325, "step": 45025 }, { "epoch": 3.53, "learning_rate": 6.469379753783423e-05, "loss": 2.8329, "step": 45026 }, { "epoch": 3.53, "learning_rate": 6.469301340860974e-05, "loss": 2.864, "step": 45027 }, { "epoch": 3.53, "learning_rate": 6.469222927938524e-05, "loss": 3.0125, "step": 45028 }, { "epoch": 3.53, "learning_rate": 6.469144515016075e-05, "loss": 2.7957, "step": 45029 }, { "epoch": 3.53, "learning_rate": 6.469066102093625e-05, "loss": 2.9804, "step": 45030 }, { "epoch": 3.53, "learning_rate": 6.468987689171175e-05, "loss": 2.5143, "step": 45031 }, { "epoch": 3.53, "learning_rate": 6.468909276248726e-05, "loss": 2.6523, "step": 45032 }, { "epoch": 3.53, "learning_rate": 6.468830863326276e-05, "loss": 2.8688, "step": 45033 }, { "epoch": 3.53, "learning_rate": 6.468752450403826e-05, "loss": 2.6778, "step": 45034 }, { "epoch": 3.53, "learning_rate": 6.468674037481377e-05, "loss": 2.9717, "step": 45035 }, { "epoch": 3.53, "learning_rate": 6.468595624558927e-05, "loss": 2.6938, "step": 45036 }, { "epoch": 3.53, "learning_rate": 6.468517211636478e-05, "loss": 2.878, "step": 45037 }, { "epoch": 3.53, "learning_rate": 6.468438798714028e-05, "loss": 2.6782, "step": 45038 }, { "epoch": 3.53, "learning_rate": 6.468360385791578e-05, "loss": 2.8421, "step": 45039 }, { "epoch": 3.53, "learning_rate": 6.468281972869129e-05, "loss": 2.9195, "step": 45040 }, { "epoch": 3.53, "learning_rate": 6.468203559946679e-05, "loss": 2.8528, "step": 45041 }, { "epoch": 3.53, "learning_rate": 6.46812514702423e-05, "loss": 2.474, "step": 45042 }, { "epoch": 3.53, "learning_rate": 6.46804673410178e-05, "loss": 2.8442, "step": 45043 }, { "epoch": 3.53, "learning_rate": 6.46796832117933e-05, "loss": 2.6126, "step": 45044 }, { "epoch": 3.53, "learning_rate": 6.46788990825688e-05, "loss": 2.5947, "step": 45045 }, { "epoch": 3.53, "learning_rate": 6.467811495334432e-05, "loss": 2.7181, "step": 45046 }, { "epoch": 3.53, "learning_rate": 6.467733082411983e-05, "loss": 2.64, "step": 45047 }, { "epoch": 3.53, "learning_rate": 6.467654669489533e-05, "loss": 2.7283, "step": 45048 }, { "epoch": 3.53, "learning_rate": 6.467576256567083e-05, "loss": 2.6857, "step": 45049 }, { "epoch": 3.53, "learning_rate": 6.467497843644634e-05, "loss": 2.7812, "step": 45050 }, { "epoch": 3.53, "learning_rate": 6.467419430722184e-05, "loss": 2.7012, "step": 45051 }, { "epoch": 3.53, "learning_rate": 6.467341017799734e-05, "loss": 2.512, "step": 45052 }, { "epoch": 3.53, "learning_rate": 6.467262604877285e-05, "loss": 2.6633, "step": 45053 }, { "epoch": 3.53, "learning_rate": 6.467184191954835e-05, "loss": 2.7381, "step": 45054 }, { "epoch": 3.53, "learning_rate": 6.467105779032385e-05, "loss": 2.5391, "step": 45055 }, { "epoch": 3.53, "learning_rate": 6.467027366109936e-05, "loss": 2.621, "step": 45056 }, { "epoch": 3.53, "learning_rate": 6.466948953187486e-05, "loss": 2.6338, "step": 45057 }, { "epoch": 3.53, "learning_rate": 6.466870540265037e-05, "loss": 2.6592, "step": 45058 }, { "epoch": 3.53, "learning_rate": 6.466792127342587e-05, "loss": 2.7764, "step": 45059 }, { "epoch": 3.53, "learning_rate": 6.466713714420137e-05, "loss": 2.677, "step": 45060 }, { "epoch": 3.53, "learning_rate": 6.466635301497686e-05, "loss": 2.7845, "step": 45061 }, { "epoch": 3.53, "learning_rate": 6.466556888575237e-05, "loss": 2.8251, "step": 45062 }, { "epoch": 3.53, "learning_rate": 6.466478475652787e-05, "loss": 2.7525, "step": 45063 }, { "epoch": 3.53, "learning_rate": 6.466400062730337e-05, "loss": 2.699, "step": 45064 }, { "epoch": 3.53, "learning_rate": 6.466321649807888e-05, "loss": 2.4295, "step": 45065 }, { "epoch": 3.53, "learning_rate": 6.466243236885438e-05, "loss": 2.5216, "step": 45066 }, { "epoch": 3.53, "learning_rate": 6.466164823962989e-05, "loss": 2.7083, "step": 45067 }, { "epoch": 3.53, "learning_rate": 6.466086411040539e-05, "loss": 2.7018, "step": 45068 }, { "epoch": 3.53, "learning_rate": 6.466007998118089e-05, "loss": 2.9056, "step": 45069 }, { "epoch": 3.53, "learning_rate": 6.46592958519564e-05, "loss": 2.6592, "step": 45070 }, { "epoch": 3.53, "learning_rate": 6.465851172273191e-05, "loss": 2.9342, "step": 45071 }, { "epoch": 3.53, "learning_rate": 6.465772759350742e-05, "loss": 2.7207, "step": 45072 }, { "epoch": 3.53, "learning_rate": 6.465694346428292e-05, "loss": 2.7932, "step": 45073 }, { "epoch": 3.53, "learning_rate": 6.465615933505842e-05, "loss": 2.518, "step": 45074 }, { "epoch": 3.53, "learning_rate": 6.465537520583393e-05, "loss": 2.6076, "step": 45075 }, { "epoch": 3.53, "learning_rate": 6.465459107660943e-05, "loss": 2.6236, "step": 45076 }, { "epoch": 3.53, "learning_rate": 6.465380694738494e-05, "loss": 2.6271, "step": 45077 }, { "epoch": 3.53, "learning_rate": 6.465302281816044e-05, "loss": 2.5878, "step": 45078 }, { "epoch": 3.53, "learning_rate": 6.465223868893594e-05, "loss": 2.7382, "step": 45079 }, { "epoch": 3.53, "learning_rate": 6.465145455971145e-05, "loss": 2.9888, "step": 45080 }, { "epoch": 3.53, "learning_rate": 6.465067043048695e-05, "loss": 2.6811, "step": 45081 }, { "epoch": 3.54, "learning_rate": 6.464988630126245e-05, "loss": 2.6646, "step": 45082 }, { "epoch": 3.54, "learning_rate": 6.464910217203796e-05, "loss": 2.6817, "step": 45083 }, { "epoch": 3.54, "learning_rate": 6.464831804281346e-05, "loss": 2.802, "step": 45084 }, { "epoch": 3.54, "learning_rate": 6.464753391358897e-05, "loss": 2.5684, "step": 45085 }, { "epoch": 3.54, "learning_rate": 6.464674978436447e-05, "loss": 2.6862, "step": 45086 }, { "epoch": 3.54, "learning_rate": 6.464596565513997e-05, "loss": 2.5843, "step": 45087 }, { "epoch": 3.54, "learning_rate": 6.464518152591548e-05, "loss": 2.6737, "step": 45088 }, { "epoch": 3.54, "learning_rate": 6.464439739669098e-05, "loss": 2.7593, "step": 45089 }, { "epoch": 3.54, "learning_rate": 6.464361326746648e-05, "loss": 2.825, "step": 45090 }, { "epoch": 3.54, "learning_rate": 6.464282913824199e-05, "loss": 2.9706, "step": 45091 }, { "epoch": 3.54, "learning_rate": 6.464204500901749e-05, "loss": 2.7815, "step": 45092 }, { "epoch": 3.54, "learning_rate": 6.4641260879793e-05, "loss": 2.6118, "step": 45093 }, { "epoch": 3.54, "learning_rate": 6.46404767505685e-05, "loss": 2.7858, "step": 45094 }, { "epoch": 3.54, "learning_rate": 6.4639692621344e-05, "loss": 2.8135, "step": 45095 }, { "epoch": 3.54, "learning_rate": 6.46389084921195e-05, "loss": 2.7604, "step": 45096 }, { "epoch": 3.54, "learning_rate": 6.463812436289501e-05, "loss": 2.5606, "step": 45097 }, { "epoch": 3.54, "learning_rate": 6.463734023367051e-05, "loss": 2.5972, "step": 45098 }, { "epoch": 3.54, "learning_rate": 6.463655610444602e-05, "loss": 2.7472, "step": 45099 }, { "epoch": 3.54, "learning_rate": 6.463577197522152e-05, "loss": 2.6067, "step": 45100 }, { "epoch": 3.54, "learning_rate": 6.463498784599702e-05, "loss": 2.9528, "step": 45101 }, { "epoch": 3.54, "learning_rate": 6.463420371677253e-05, "loss": 2.6468, "step": 45102 }, { "epoch": 3.54, "learning_rate": 6.463341958754803e-05, "loss": 2.7417, "step": 45103 }, { "epoch": 3.54, "learning_rate": 6.463263545832354e-05, "loss": 2.6377, "step": 45104 }, { "epoch": 3.54, "learning_rate": 6.463185132909904e-05, "loss": 2.6966, "step": 45105 }, { "epoch": 3.54, "learning_rate": 6.463106719987454e-05, "loss": 2.7596, "step": 45106 }, { "epoch": 3.54, "learning_rate": 6.463028307065005e-05, "loss": 2.6974, "step": 45107 }, { "epoch": 3.54, "learning_rate": 6.462949894142555e-05, "loss": 2.5618, "step": 45108 }, { "epoch": 3.54, "learning_rate": 6.462871481220105e-05, "loss": 2.7219, "step": 45109 }, { "epoch": 3.54, "learning_rate": 6.462793068297656e-05, "loss": 2.9555, "step": 45110 }, { "epoch": 3.54, "learning_rate": 6.462714655375206e-05, "loss": 2.8234, "step": 45111 }, { "epoch": 3.54, "learning_rate": 6.462636242452756e-05, "loss": 2.8577, "step": 45112 }, { "epoch": 3.54, "learning_rate": 6.462557829530307e-05, "loss": 2.8552, "step": 45113 }, { "epoch": 3.54, "learning_rate": 6.462479416607857e-05, "loss": 2.692, "step": 45114 }, { "epoch": 3.54, "learning_rate": 6.462401003685408e-05, "loss": 2.8793, "step": 45115 }, { "epoch": 3.54, "learning_rate": 6.462322590762958e-05, "loss": 2.7973, "step": 45116 }, { "epoch": 3.54, "learning_rate": 6.462244177840508e-05, "loss": 2.7989, "step": 45117 }, { "epoch": 3.54, "learning_rate": 6.462165764918059e-05, "loss": 2.7735, "step": 45118 }, { "epoch": 3.54, "learning_rate": 6.462087351995609e-05, "loss": 2.89, "step": 45119 }, { "epoch": 3.54, "learning_rate": 6.46200893907316e-05, "loss": 2.7704, "step": 45120 }, { "epoch": 3.54, "learning_rate": 6.46193052615071e-05, "loss": 2.8336, "step": 45121 }, { "epoch": 3.54, "learning_rate": 6.46185211322826e-05, "loss": 2.8126, "step": 45122 }, { "epoch": 3.54, "learning_rate": 6.46177370030581e-05, "loss": 2.9235, "step": 45123 }, { "epoch": 3.54, "learning_rate": 6.461695287383361e-05, "loss": 2.6454, "step": 45124 }, { "epoch": 3.54, "learning_rate": 6.461616874460911e-05, "loss": 2.8743, "step": 45125 }, { "epoch": 3.54, "learning_rate": 6.461538461538462e-05, "loss": 2.6194, "step": 45126 }, { "epoch": 3.54, "learning_rate": 6.461460048616012e-05, "loss": 2.8151, "step": 45127 }, { "epoch": 3.54, "learning_rate": 6.461381635693564e-05, "loss": 2.8156, "step": 45128 }, { "epoch": 3.54, "learning_rate": 6.461303222771114e-05, "loss": 2.7364, "step": 45129 }, { "epoch": 3.54, "learning_rate": 6.461224809848664e-05, "loss": 2.731, "step": 45130 }, { "epoch": 3.54, "learning_rate": 6.461146396926215e-05, "loss": 2.7238, "step": 45131 }, { "epoch": 3.54, "learning_rate": 6.461067984003765e-05, "loss": 3.0321, "step": 45132 }, { "epoch": 3.54, "learning_rate": 6.460989571081314e-05, "loss": 2.597, "step": 45133 }, { "epoch": 3.54, "learning_rate": 6.460911158158865e-05, "loss": 2.5835, "step": 45134 }, { "epoch": 3.54, "learning_rate": 6.460832745236415e-05, "loss": 2.8963, "step": 45135 }, { "epoch": 3.54, "learning_rate": 6.460754332313965e-05, "loss": 2.6332, "step": 45136 }, { "epoch": 3.54, "learning_rate": 6.460675919391516e-05, "loss": 2.5576, "step": 45137 }, { "epoch": 3.54, "learning_rate": 6.460597506469066e-05, "loss": 2.5597, "step": 45138 }, { "epoch": 3.54, "learning_rate": 6.460519093546616e-05, "loss": 2.8121, "step": 45139 }, { "epoch": 3.54, "learning_rate": 6.460440680624167e-05, "loss": 2.7956, "step": 45140 }, { "epoch": 3.54, "learning_rate": 6.460362267701717e-05, "loss": 2.6409, "step": 45141 }, { "epoch": 3.54, "learning_rate": 6.460283854779267e-05, "loss": 2.871, "step": 45142 }, { "epoch": 3.54, "learning_rate": 6.460205441856818e-05, "loss": 2.8023, "step": 45143 }, { "epoch": 3.54, "learning_rate": 6.460127028934368e-05, "loss": 2.608, "step": 45144 }, { "epoch": 3.54, "learning_rate": 6.460048616011919e-05, "loss": 2.8696, "step": 45145 }, { "epoch": 3.54, "learning_rate": 6.459970203089469e-05, "loss": 2.7307, "step": 45146 }, { "epoch": 3.54, "learning_rate": 6.459891790167019e-05, "loss": 2.637, "step": 45147 }, { "epoch": 3.54, "learning_rate": 6.45981337724457e-05, "loss": 2.7194, "step": 45148 }, { "epoch": 3.54, "learning_rate": 6.45973496432212e-05, "loss": 2.5826, "step": 45149 }, { "epoch": 3.54, "learning_rate": 6.45965655139967e-05, "loss": 2.6395, "step": 45150 }, { "epoch": 3.54, "learning_rate": 6.459578138477221e-05, "loss": 2.6968, "step": 45151 }, { "epoch": 3.54, "learning_rate": 6.459499725554771e-05, "loss": 2.5768, "step": 45152 }, { "epoch": 3.54, "learning_rate": 6.459421312632322e-05, "loss": 2.7135, "step": 45153 }, { "epoch": 3.54, "learning_rate": 6.459342899709873e-05, "loss": 2.6575, "step": 45154 }, { "epoch": 3.54, "learning_rate": 6.459264486787424e-05, "loss": 2.7222, "step": 45155 }, { "epoch": 3.54, "learning_rate": 6.459186073864974e-05, "loss": 2.7134, "step": 45156 }, { "epoch": 3.54, "learning_rate": 6.459107660942524e-05, "loss": 2.6321, "step": 45157 }, { "epoch": 3.54, "learning_rate": 6.459029248020075e-05, "loss": 2.7858, "step": 45158 }, { "epoch": 3.54, "learning_rate": 6.458950835097625e-05, "loss": 2.8023, "step": 45159 }, { "epoch": 3.54, "learning_rate": 6.458872422175175e-05, "loss": 2.6614, "step": 45160 }, { "epoch": 3.54, "learning_rate": 6.458794009252726e-05, "loss": 2.5522, "step": 45161 }, { "epoch": 3.54, "learning_rate": 6.458715596330276e-05, "loss": 2.8426, "step": 45162 }, { "epoch": 3.54, "learning_rate": 6.458637183407827e-05, "loss": 2.8374, "step": 45163 }, { "epoch": 3.54, "learning_rate": 6.458558770485377e-05, "loss": 2.8228, "step": 45164 }, { "epoch": 3.54, "learning_rate": 6.458480357562927e-05, "loss": 2.7831, "step": 45165 }, { "epoch": 3.54, "learning_rate": 6.458401944640478e-05, "loss": 2.7474, "step": 45166 }, { "epoch": 3.54, "learning_rate": 6.458323531718028e-05, "loss": 2.448, "step": 45167 }, { "epoch": 3.54, "learning_rate": 6.458245118795578e-05, "loss": 2.8857, "step": 45168 }, { "epoch": 3.54, "learning_rate": 6.458166705873127e-05, "loss": 2.7724, "step": 45169 }, { "epoch": 3.54, "learning_rate": 6.458088292950678e-05, "loss": 2.5658, "step": 45170 }, { "epoch": 3.54, "learning_rate": 6.458009880028228e-05, "loss": 2.7519, "step": 45171 }, { "epoch": 3.54, "learning_rate": 6.457931467105779e-05, "loss": 2.7068, "step": 45172 }, { "epoch": 3.54, "learning_rate": 6.457853054183329e-05, "loss": 2.6335, "step": 45173 }, { "epoch": 3.54, "learning_rate": 6.457774641260879e-05, "loss": 2.5693, "step": 45174 }, { "epoch": 3.54, "learning_rate": 6.45769622833843e-05, "loss": 3.0182, "step": 45175 }, { "epoch": 3.54, "learning_rate": 6.45761781541598e-05, "loss": 2.632, "step": 45176 }, { "epoch": 3.54, "learning_rate": 6.45753940249353e-05, "loss": 2.9524, "step": 45177 }, { "epoch": 3.54, "learning_rate": 6.457460989571081e-05, "loss": 2.6508, "step": 45178 }, { "epoch": 3.54, "learning_rate": 6.457382576648632e-05, "loss": 2.6421, "step": 45179 }, { "epoch": 3.54, "learning_rate": 6.457304163726183e-05, "loss": 2.8054, "step": 45180 }, { "epoch": 3.54, "learning_rate": 6.457225750803733e-05, "loss": 2.7934, "step": 45181 }, { "epoch": 3.54, "learning_rate": 6.457147337881284e-05, "loss": 2.7243, "step": 45182 }, { "epoch": 3.54, "learning_rate": 6.457068924958834e-05, "loss": 2.4923, "step": 45183 }, { "epoch": 3.54, "learning_rate": 6.456990512036384e-05, "loss": 2.6095, "step": 45184 }, { "epoch": 3.54, "learning_rate": 6.456912099113935e-05, "loss": 2.7801, "step": 45185 }, { "epoch": 3.54, "learning_rate": 6.456833686191485e-05, "loss": 2.8053, "step": 45186 }, { "epoch": 3.54, "learning_rate": 6.456755273269035e-05, "loss": 2.8906, "step": 45187 }, { "epoch": 3.54, "learning_rate": 6.456676860346586e-05, "loss": 2.8148, "step": 45188 }, { "epoch": 3.54, "learning_rate": 6.456598447424136e-05, "loss": 2.8118, "step": 45189 }, { "epoch": 3.54, "learning_rate": 6.456520034501686e-05, "loss": 2.7647, "step": 45190 }, { "epoch": 3.54, "learning_rate": 6.456441621579237e-05, "loss": 2.8776, "step": 45191 }, { "epoch": 3.54, "learning_rate": 6.456363208656787e-05, "loss": 2.6679, "step": 45192 }, { "epoch": 3.54, "learning_rate": 6.456284795734338e-05, "loss": 2.7009, "step": 45193 }, { "epoch": 3.54, "learning_rate": 6.456206382811888e-05, "loss": 2.7153, "step": 45194 }, { "epoch": 3.54, "learning_rate": 6.456127969889438e-05, "loss": 2.9012, "step": 45195 }, { "epoch": 3.54, "learning_rate": 6.456049556966989e-05, "loss": 2.667, "step": 45196 }, { "epoch": 3.54, "learning_rate": 6.455971144044539e-05, "loss": 2.8377, "step": 45197 }, { "epoch": 3.54, "learning_rate": 6.45589273112209e-05, "loss": 2.6985, "step": 45198 }, { "epoch": 3.54, "learning_rate": 6.45581431819964e-05, "loss": 2.4749, "step": 45199 }, { "epoch": 3.54, "learning_rate": 6.45573590527719e-05, "loss": 2.8545, "step": 45200 }, { "epoch": 3.54, "learning_rate": 6.45565749235474e-05, "loss": 2.6841, "step": 45201 }, { "epoch": 3.54, "learning_rate": 6.455579079432291e-05, "loss": 2.6864, "step": 45202 }, { "epoch": 3.54, "learning_rate": 6.455500666509841e-05, "loss": 2.7461, "step": 45203 }, { "epoch": 3.54, "learning_rate": 6.455422253587392e-05, "loss": 2.7992, "step": 45204 }, { "epoch": 3.54, "learning_rate": 6.455343840664942e-05, "loss": 2.4573, "step": 45205 }, { "epoch": 3.54, "learning_rate": 6.455265427742492e-05, "loss": 2.8201, "step": 45206 }, { "epoch": 3.54, "learning_rate": 6.455187014820043e-05, "loss": 2.7565, "step": 45207 }, { "epoch": 3.54, "learning_rate": 6.455108601897593e-05, "loss": 2.6717, "step": 45208 }, { "epoch": 3.54, "learning_rate": 6.455030188975143e-05, "loss": 2.6365, "step": 45209 }, { "epoch": 3.55, "learning_rate": 6.454951776052694e-05, "loss": 2.6817, "step": 45210 }, { "epoch": 3.55, "learning_rate": 6.454873363130244e-05, "loss": 2.728, "step": 45211 }, { "epoch": 3.55, "learning_rate": 6.454794950207795e-05, "loss": 2.6975, "step": 45212 }, { "epoch": 3.55, "learning_rate": 6.454716537285345e-05, "loss": 2.6069, "step": 45213 }, { "epoch": 3.55, "learning_rate": 6.454638124362895e-05, "loss": 2.7016, "step": 45214 }, { "epoch": 3.55, "learning_rate": 6.454559711440446e-05, "loss": 2.6267, "step": 45215 }, { "epoch": 3.55, "learning_rate": 6.454481298517996e-05, "loss": 2.7791, "step": 45216 }, { "epoch": 3.55, "learning_rate": 6.454402885595546e-05, "loss": 2.7494, "step": 45217 }, { "epoch": 3.55, "learning_rate": 6.454324472673097e-05, "loss": 2.7992, "step": 45218 }, { "epoch": 3.55, "learning_rate": 6.454246059750647e-05, "loss": 2.7032, "step": 45219 }, { "epoch": 3.55, "learning_rate": 6.454167646828198e-05, "loss": 2.6572, "step": 45220 }, { "epoch": 3.55, "learning_rate": 6.454089233905748e-05, "loss": 2.5613, "step": 45221 }, { "epoch": 3.55, "learning_rate": 6.454010820983298e-05, "loss": 2.6557, "step": 45222 }, { "epoch": 3.55, "learning_rate": 6.453932408060849e-05, "loss": 2.67, "step": 45223 }, { "epoch": 3.55, "learning_rate": 6.453853995138399e-05, "loss": 2.7175, "step": 45224 }, { "epoch": 3.55, "learning_rate": 6.45377558221595e-05, "loss": 2.6471, "step": 45225 }, { "epoch": 3.55, "learning_rate": 6.4536971692935e-05, "loss": 2.712, "step": 45226 }, { "epoch": 3.55, "learning_rate": 6.45361875637105e-05, "loss": 2.7345, "step": 45227 }, { "epoch": 3.55, "learning_rate": 6.4535403434486e-05, "loss": 3.0016, "step": 45228 }, { "epoch": 3.55, "learning_rate": 6.453461930526151e-05, "loss": 2.6627, "step": 45229 }, { "epoch": 3.55, "learning_rate": 6.453383517603701e-05, "loss": 2.6426, "step": 45230 }, { "epoch": 3.55, "learning_rate": 6.453305104681252e-05, "loss": 2.8837, "step": 45231 }, { "epoch": 3.55, "learning_rate": 6.453226691758802e-05, "loss": 2.5997, "step": 45232 }, { "epoch": 3.55, "learning_rate": 6.453148278836352e-05, "loss": 2.6663, "step": 45233 }, { "epoch": 3.55, "learning_rate": 6.453069865913903e-05, "loss": 2.5928, "step": 45234 }, { "epoch": 3.55, "learning_rate": 6.452991452991453e-05, "loss": 2.7323, "step": 45235 }, { "epoch": 3.55, "learning_rate": 6.452913040069005e-05, "loss": 2.6292, "step": 45236 }, { "epoch": 3.55, "learning_rate": 6.452834627146555e-05, "loss": 2.7027, "step": 45237 }, { "epoch": 3.55, "learning_rate": 6.452756214224105e-05, "loss": 2.6916, "step": 45238 }, { "epoch": 3.55, "learning_rate": 6.452677801301656e-05, "loss": 2.5292, "step": 45239 }, { "epoch": 3.55, "learning_rate": 6.452599388379206e-05, "loss": 2.6497, "step": 45240 }, { "epoch": 3.55, "learning_rate": 6.452520975456755e-05, "loss": 2.8704, "step": 45241 }, { "epoch": 3.55, "learning_rate": 6.452442562534306e-05, "loss": 2.7975, "step": 45242 }, { "epoch": 3.55, "learning_rate": 6.452364149611856e-05, "loss": 2.6903, "step": 45243 }, { "epoch": 3.55, "learning_rate": 6.452285736689406e-05, "loss": 2.5232, "step": 45244 }, { "epoch": 3.55, "learning_rate": 6.452207323766957e-05, "loss": 2.5928, "step": 45245 }, { "epoch": 3.55, "learning_rate": 6.452128910844507e-05, "loss": 2.8812, "step": 45246 }, { "epoch": 3.55, "learning_rate": 6.452050497922057e-05, "loss": 2.5172, "step": 45247 }, { "epoch": 3.55, "learning_rate": 6.451972084999608e-05, "loss": 2.7291, "step": 45248 }, { "epoch": 3.55, "learning_rate": 6.451893672077158e-05, "loss": 2.9931, "step": 45249 }, { "epoch": 3.55, "learning_rate": 6.451815259154709e-05, "loss": 2.5407, "step": 45250 }, { "epoch": 3.55, "learning_rate": 6.451736846232259e-05, "loss": 2.8273, "step": 45251 }, { "epoch": 3.55, "learning_rate": 6.451658433309809e-05, "loss": 2.7921, "step": 45252 }, { "epoch": 3.55, "learning_rate": 6.45158002038736e-05, "loss": 2.5527, "step": 45253 }, { "epoch": 3.55, "learning_rate": 6.45150160746491e-05, "loss": 2.9275, "step": 45254 }, { "epoch": 3.55, "learning_rate": 6.45142319454246e-05, "loss": 2.736, "step": 45255 }, { "epoch": 3.55, "learning_rate": 6.451344781620011e-05, "loss": 2.8117, "step": 45256 }, { "epoch": 3.55, "learning_rate": 6.451266368697561e-05, "loss": 2.8546, "step": 45257 }, { "epoch": 3.55, "learning_rate": 6.451187955775111e-05, "loss": 2.7072, "step": 45258 }, { "epoch": 3.55, "learning_rate": 6.451109542852662e-05, "loss": 2.7003, "step": 45259 }, { "epoch": 3.55, "learning_rate": 6.451031129930212e-05, "loss": 2.7927, "step": 45260 }, { "epoch": 3.55, "learning_rate": 6.450952717007763e-05, "loss": 2.4686, "step": 45261 }, { "epoch": 3.55, "learning_rate": 6.450874304085314e-05, "loss": 2.6386, "step": 45262 }, { "epoch": 3.55, "learning_rate": 6.450795891162865e-05, "loss": 2.7962, "step": 45263 }, { "epoch": 3.55, "learning_rate": 6.450717478240415e-05, "loss": 2.6324, "step": 45264 }, { "epoch": 3.55, "learning_rate": 6.450639065317965e-05, "loss": 2.524, "step": 45265 }, { "epoch": 3.55, "learning_rate": 6.450560652395516e-05, "loss": 2.7176, "step": 45266 }, { "epoch": 3.55, "learning_rate": 6.450482239473066e-05, "loss": 2.6716, "step": 45267 }, { "epoch": 3.55, "learning_rate": 6.450403826550617e-05, "loss": 2.8556, "step": 45268 }, { "epoch": 3.55, "learning_rate": 6.450325413628167e-05, "loss": 2.9231, "step": 45269 }, { "epoch": 3.55, "learning_rate": 6.450247000705717e-05, "loss": 2.9559, "step": 45270 }, { "epoch": 3.55, "learning_rate": 6.450168587783268e-05, "loss": 2.8263, "step": 45271 }, { "epoch": 3.55, "learning_rate": 6.450090174860818e-05, "loss": 2.6146, "step": 45272 }, { "epoch": 3.55, "learning_rate": 6.450011761938368e-05, "loss": 2.7742, "step": 45273 }, { "epoch": 3.55, "learning_rate": 6.449933349015919e-05, "loss": 2.6753, "step": 45274 }, { "epoch": 3.55, "learning_rate": 6.449854936093469e-05, "loss": 2.8589, "step": 45275 }, { "epoch": 3.55, "learning_rate": 6.44977652317102e-05, "loss": 2.8223, "step": 45276 }, { "epoch": 3.55, "learning_rate": 6.449698110248568e-05, "loss": 2.7788, "step": 45277 }, { "epoch": 3.55, "learning_rate": 6.449619697326119e-05, "loss": 2.6954, "step": 45278 }, { "epoch": 3.55, "learning_rate": 6.449541284403669e-05, "loss": 2.5622, "step": 45279 }, { "epoch": 3.55, "learning_rate": 6.44946287148122e-05, "loss": 2.5755, "step": 45280 }, { "epoch": 3.55, "learning_rate": 6.44938445855877e-05, "loss": 2.7376, "step": 45281 }, { "epoch": 3.55, "learning_rate": 6.44930604563632e-05, "loss": 2.8232, "step": 45282 }, { "epoch": 3.55, "learning_rate": 6.449227632713871e-05, "loss": 2.7525, "step": 45283 }, { "epoch": 3.55, "learning_rate": 6.449149219791421e-05, "loss": 2.9021, "step": 45284 }, { "epoch": 3.55, "learning_rate": 6.449070806868971e-05, "loss": 2.9606, "step": 45285 }, { "epoch": 3.55, "learning_rate": 6.448992393946522e-05, "loss": 2.583, "step": 45286 }, { "epoch": 3.55, "learning_rate": 6.448913981024074e-05, "loss": 2.8158, "step": 45287 }, { "epoch": 3.55, "learning_rate": 6.448835568101624e-05, "loss": 2.8378, "step": 45288 }, { "epoch": 3.55, "learning_rate": 6.448757155179174e-05, "loss": 2.7878, "step": 45289 }, { "epoch": 3.55, "learning_rate": 6.448678742256725e-05, "loss": 2.7918, "step": 45290 }, { "epoch": 3.55, "learning_rate": 6.448600329334275e-05, "loss": 2.7251, "step": 45291 }, { "epoch": 3.55, "learning_rate": 6.448521916411825e-05, "loss": 2.5247, "step": 45292 }, { "epoch": 3.55, "learning_rate": 6.448443503489376e-05, "loss": 2.6262, "step": 45293 }, { "epoch": 3.55, "learning_rate": 6.448365090566926e-05, "loss": 2.6592, "step": 45294 }, { "epoch": 3.55, "learning_rate": 6.448286677644476e-05, "loss": 2.6888, "step": 45295 }, { "epoch": 3.55, "learning_rate": 6.448208264722027e-05, "loss": 2.6797, "step": 45296 }, { "epoch": 3.55, "learning_rate": 6.448129851799577e-05, "loss": 2.6892, "step": 45297 }, { "epoch": 3.55, "learning_rate": 6.448051438877128e-05, "loss": 2.9216, "step": 45298 }, { "epoch": 3.55, "learning_rate": 6.447973025954678e-05, "loss": 2.6146, "step": 45299 }, { "epoch": 3.55, "learning_rate": 6.447894613032228e-05, "loss": 2.7065, "step": 45300 }, { "epoch": 3.55, "learning_rate": 6.447816200109779e-05, "loss": 2.8993, "step": 45301 }, { "epoch": 3.55, "learning_rate": 6.447737787187329e-05, "loss": 2.827, "step": 45302 }, { "epoch": 3.55, "learning_rate": 6.44765937426488e-05, "loss": 3.0218, "step": 45303 }, { "epoch": 3.55, "learning_rate": 6.44758096134243e-05, "loss": 2.7444, "step": 45304 }, { "epoch": 3.55, "learning_rate": 6.44750254841998e-05, "loss": 2.8141, "step": 45305 }, { "epoch": 3.55, "learning_rate": 6.44742413549753e-05, "loss": 2.7759, "step": 45306 }, { "epoch": 3.55, "learning_rate": 6.447345722575081e-05, "loss": 2.7762, "step": 45307 }, { "epoch": 3.55, "learning_rate": 6.447267309652631e-05, "loss": 2.7132, "step": 45308 }, { "epoch": 3.55, "learning_rate": 6.447188896730182e-05, "loss": 2.7547, "step": 45309 }, { "epoch": 3.55, "learning_rate": 6.447110483807732e-05, "loss": 2.8345, "step": 45310 }, { "epoch": 3.55, "learning_rate": 6.447032070885282e-05, "loss": 2.5893, "step": 45311 }, { "epoch": 3.55, "learning_rate": 6.446953657962833e-05, "loss": 2.705, "step": 45312 }, { "epoch": 3.55, "learning_rate": 6.446875245040383e-05, "loss": 2.6093, "step": 45313 }, { "epoch": 3.55, "learning_rate": 6.446796832117933e-05, "loss": 2.6783, "step": 45314 }, { "epoch": 3.55, "learning_rate": 6.446718419195484e-05, "loss": 2.7772, "step": 45315 }, { "epoch": 3.55, "learning_rate": 6.446640006273034e-05, "loss": 2.7969, "step": 45316 }, { "epoch": 3.55, "learning_rate": 6.446561593350585e-05, "loss": 2.6831, "step": 45317 }, { "epoch": 3.55, "learning_rate": 6.446483180428135e-05, "loss": 2.8509, "step": 45318 }, { "epoch": 3.55, "learning_rate": 6.446404767505685e-05, "loss": 2.8136, "step": 45319 }, { "epoch": 3.55, "learning_rate": 6.446326354583236e-05, "loss": 2.68, "step": 45320 }, { "epoch": 3.55, "learning_rate": 6.446247941660786e-05, "loss": 2.5945, "step": 45321 }, { "epoch": 3.55, "learning_rate": 6.446169528738336e-05, "loss": 2.7918, "step": 45322 }, { "epoch": 3.55, "learning_rate": 6.446091115815887e-05, "loss": 2.8051, "step": 45323 }, { "epoch": 3.55, "learning_rate": 6.446012702893437e-05, "loss": 2.869, "step": 45324 }, { "epoch": 3.55, "learning_rate": 6.445934289970987e-05, "loss": 2.6581, "step": 45325 }, { "epoch": 3.55, "learning_rate": 6.445855877048538e-05, "loss": 2.6666, "step": 45326 }, { "epoch": 3.55, "learning_rate": 6.445777464126088e-05, "loss": 2.8235, "step": 45327 }, { "epoch": 3.55, "learning_rate": 6.445699051203639e-05, "loss": 2.6655, "step": 45328 }, { "epoch": 3.55, "learning_rate": 6.445620638281189e-05, "loss": 2.6555, "step": 45329 }, { "epoch": 3.55, "learning_rate": 6.44554222535874e-05, "loss": 2.7094, "step": 45330 }, { "epoch": 3.55, "learning_rate": 6.44546381243629e-05, "loss": 2.7928, "step": 45331 }, { "epoch": 3.55, "learning_rate": 6.44538539951384e-05, "loss": 2.7198, "step": 45332 }, { "epoch": 3.55, "learning_rate": 6.44530698659139e-05, "loss": 2.7997, "step": 45333 }, { "epoch": 3.55, "learning_rate": 6.445228573668941e-05, "loss": 2.7037, "step": 45334 }, { "epoch": 3.55, "learning_rate": 6.445150160746491e-05, "loss": 2.7954, "step": 45335 }, { "epoch": 3.55, "learning_rate": 6.445071747824042e-05, "loss": 2.9444, "step": 45336 }, { "epoch": 3.56, "learning_rate": 6.444993334901592e-05, "loss": 2.7369, "step": 45337 }, { "epoch": 3.56, "learning_rate": 6.444914921979142e-05, "loss": 2.5784, "step": 45338 }, { "epoch": 3.56, "learning_rate": 6.444836509056693e-05, "loss": 2.5814, "step": 45339 }, { "epoch": 3.56, "learning_rate": 6.444758096134243e-05, "loss": 2.537, "step": 45340 }, { "epoch": 3.56, "learning_rate": 6.444679683211793e-05, "loss": 2.7048, "step": 45341 }, { "epoch": 3.56, "learning_rate": 6.444601270289344e-05, "loss": 2.5596, "step": 45342 }, { "epoch": 3.56, "learning_rate": 6.444522857366894e-05, "loss": 2.6702, "step": 45343 }, { "epoch": 3.56, "learning_rate": 6.444444444444446e-05, "loss": 2.8772, "step": 45344 }, { "epoch": 3.56, "learning_rate": 6.444366031521996e-05, "loss": 3.014, "step": 45345 }, { "epoch": 3.56, "learning_rate": 6.444287618599547e-05, "loss": 2.7549, "step": 45346 }, { "epoch": 3.56, "learning_rate": 6.444209205677097e-05, "loss": 2.7311, "step": 45347 }, { "epoch": 3.56, "learning_rate": 6.444130792754647e-05, "loss": 2.7325, "step": 45348 }, { "epoch": 3.56, "learning_rate": 6.444052379832196e-05, "loss": 2.8092, "step": 45349 }, { "epoch": 3.56, "learning_rate": 6.443973966909747e-05, "loss": 2.9109, "step": 45350 }, { "epoch": 3.56, "learning_rate": 6.443895553987297e-05, "loss": 2.6991, "step": 45351 }, { "epoch": 3.56, "learning_rate": 6.443817141064847e-05, "loss": 2.7154, "step": 45352 }, { "epoch": 3.56, "learning_rate": 6.443738728142398e-05, "loss": 2.6534, "step": 45353 }, { "epoch": 3.56, "learning_rate": 6.443660315219948e-05, "loss": 2.8357, "step": 45354 }, { "epoch": 3.56, "learning_rate": 6.443581902297499e-05, "loss": 2.7517, "step": 45355 }, { "epoch": 3.56, "learning_rate": 6.443503489375049e-05, "loss": 2.6573, "step": 45356 }, { "epoch": 3.56, "learning_rate": 6.443425076452599e-05, "loss": 2.6881, "step": 45357 }, { "epoch": 3.56, "learning_rate": 6.44334666353015e-05, "loss": 2.9558, "step": 45358 }, { "epoch": 3.56, "learning_rate": 6.4432682506077e-05, "loss": 2.861, "step": 45359 }, { "epoch": 3.56, "learning_rate": 6.44318983768525e-05, "loss": 2.6711, "step": 45360 }, { "epoch": 3.56, "learning_rate": 6.443111424762801e-05, "loss": 2.8117, "step": 45361 }, { "epoch": 3.56, "learning_rate": 6.443033011840351e-05, "loss": 2.5477, "step": 45362 }, { "epoch": 3.56, "learning_rate": 6.442954598917901e-05, "loss": 2.6798, "step": 45363 }, { "epoch": 3.56, "learning_rate": 6.442876185995452e-05, "loss": 2.6519, "step": 45364 }, { "epoch": 3.56, "learning_rate": 6.442797773073002e-05, "loss": 2.6764, "step": 45365 }, { "epoch": 3.56, "learning_rate": 6.442719360150553e-05, "loss": 2.5987, "step": 45366 }, { "epoch": 3.56, "learning_rate": 6.442640947228103e-05, "loss": 2.7501, "step": 45367 }, { "epoch": 3.56, "learning_rate": 6.442562534305653e-05, "loss": 2.6411, "step": 45368 }, { "epoch": 3.56, "learning_rate": 6.442484121383204e-05, "loss": 2.6989, "step": 45369 }, { "epoch": 3.56, "learning_rate": 6.442405708460755e-05, "loss": 2.8299, "step": 45370 }, { "epoch": 3.56, "learning_rate": 6.442327295538306e-05, "loss": 2.8092, "step": 45371 }, { "epoch": 3.56, "learning_rate": 6.442248882615856e-05, "loss": 2.7646, "step": 45372 }, { "epoch": 3.56, "learning_rate": 6.442170469693407e-05, "loss": 2.6779, "step": 45373 }, { "epoch": 3.56, "learning_rate": 6.442092056770957e-05, "loss": 2.7075, "step": 45374 }, { "epoch": 3.56, "learning_rate": 6.442013643848507e-05, "loss": 2.6927, "step": 45375 }, { "epoch": 3.56, "learning_rate": 6.441935230926058e-05, "loss": 2.8144, "step": 45376 }, { "epoch": 3.56, "learning_rate": 6.441856818003608e-05, "loss": 3.0344, "step": 45377 }, { "epoch": 3.56, "learning_rate": 6.441778405081158e-05, "loss": 2.7019, "step": 45378 }, { "epoch": 3.56, "learning_rate": 6.441699992158709e-05, "loss": 2.5281, "step": 45379 }, { "epoch": 3.56, "learning_rate": 6.441621579236259e-05, "loss": 2.6144, "step": 45380 }, { "epoch": 3.56, "learning_rate": 6.44154316631381e-05, "loss": 2.9032, "step": 45381 }, { "epoch": 3.56, "learning_rate": 6.44146475339136e-05, "loss": 2.6852, "step": 45382 }, { "epoch": 3.56, "learning_rate": 6.44138634046891e-05, "loss": 2.5829, "step": 45383 }, { "epoch": 3.56, "learning_rate": 6.44130792754646e-05, "loss": 2.9646, "step": 45384 }, { "epoch": 3.56, "learning_rate": 6.44122951462401e-05, "loss": 2.7645, "step": 45385 }, { "epoch": 3.56, "learning_rate": 6.44115110170156e-05, "loss": 2.6438, "step": 45386 }, { "epoch": 3.56, "learning_rate": 6.44107268877911e-05, "loss": 2.828, "step": 45387 }, { "epoch": 3.56, "learning_rate": 6.44099427585666e-05, "loss": 2.8332, "step": 45388 }, { "epoch": 3.56, "learning_rate": 6.440915862934211e-05, "loss": 2.6881, "step": 45389 }, { "epoch": 3.56, "learning_rate": 6.440837450011761e-05, "loss": 2.929, "step": 45390 }, { "epoch": 3.56, "learning_rate": 6.440759037089312e-05, "loss": 2.8029, "step": 45391 }, { "epoch": 3.56, "learning_rate": 6.440680624166862e-05, "loss": 2.6058, "step": 45392 }, { "epoch": 3.56, "learning_rate": 6.440602211244413e-05, "loss": 2.7317, "step": 45393 }, { "epoch": 3.56, "learning_rate": 6.440523798321963e-05, "loss": 2.7413, "step": 45394 }, { "epoch": 3.56, "learning_rate": 6.440445385399515e-05, "loss": 2.8304, "step": 45395 }, { "epoch": 3.56, "learning_rate": 6.440366972477065e-05, "loss": 2.6199, "step": 45396 }, { "epoch": 3.56, "learning_rate": 6.440288559554615e-05, "loss": 2.8782, "step": 45397 }, { "epoch": 3.56, "learning_rate": 6.440210146632166e-05, "loss": 2.7469, "step": 45398 }, { "epoch": 3.56, "learning_rate": 6.440131733709716e-05, "loss": 2.9169, "step": 45399 }, { "epoch": 3.56, "learning_rate": 6.440053320787266e-05, "loss": 2.6051, "step": 45400 }, { "epoch": 3.56, "learning_rate": 6.439974907864817e-05, "loss": 3.0795, "step": 45401 }, { "epoch": 3.56, "learning_rate": 6.439896494942367e-05, "loss": 2.8286, "step": 45402 }, { "epoch": 3.56, "learning_rate": 6.439818082019918e-05, "loss": 2.7645, "step": 45403 }, { "epoch": 3.56, "learning_rate": 6.439739669097468e-05, "loss": 2.5945, "step": 45404 }, { "epoch": 3.56, "learning_rate": 6.439661256175018e-05, "loss": 2.7485, "step": 45405 }, { "epoch": 3.56, "learning_rate": 6.439582843252569e-05, "loss": 2.8179, "step": 45406 }, { "epoch": 3.56, "learning_rate": 6.439504430330119e-05, "loss": 2.6316, "step": 45407 }, { "epoch": 3.56, "learning_rate": 6.43942601740767e-05, "loss": 2.9584, "step": 45408 }, { "epoch": 3.56, "learning_rate": 6.43934760448522e-05, "loss": 2.6305, "step": 45409 }, { "epoch": 3.56, "learning_rate": 6.43926919156277e-05, "loss": 2.5377, "step": 45410 }, { "epoch": 3.56, "learning_rate": 6.43919077864032e-05, "loss": 2.7349, "step": 45411 }, { "epoch": 3.56, "learning_rate": 6.439112365717871e-05, "loss": 2.5981, "step": 45412 }, { "epoch": 3.56, "learning_rate": 6.439033952795421e-05, "loss": 2.7384, "step": 45413 }, { "epoch": 3.56, "learning_rate": 6.438955539872972e-05, "loss": 2.6949, "step": 45414 }, { "epoch": 3.56, "learning_rate": 6.438877126950522e-05, "loss": 2.5308, "step": 45415 }, { "epoch": 3.56, "learning_rate": 6.438798714028072e-05, "loss": 3.2093, "step": 45416 }, { "epoch": 3.56, "learning_rate": 6.438720301105623e-05, "loss": 2.652, "step": 45417 }, { "epoch": 3.56, "learning_rate": 6.438641888183173e-05, "loss": 2.623, "step": 45418 }, { "epoch": 3.56, "learning_rate": 6.438563475260723e-05, "loss": 2.6013, "step": 45419 }, { "epoch": 3.56, "learning_rate": 6.438485062338274e-05, "loss": 2.719, "step": 45420 }, { "epoch": 3.56, "learning_rate": 6.438406649415824e-05, "loss": 2.6517, "step": 45421 }, { "epoch": 3.56, "learning_rate": 6.438328236493375e-05, "loss": 2.632, "step": 45422 }, { "epoch": 3.56, "learning_rate": 6.438249823570925e-05, "loss": 2.742, "step": 45423 }, { "epoch": 3.56, "learning_rate": 6.438171410648475e-05, "loss": 2.8136, "step": 45424 }, { "epoch": 3.56, "learning_rate": 6.438092997726026e-05, "loss": 2.8542, "step": 45425 }, { "epoch": 3.56, "learning_rate": 6.438014584803576e-05, "loss": 2.7595, "step": 45426 }, { "epoch": 3.56, "learning_rate": 6.437936171881126e-05, "loss": 2.7669, "step": 45427 }, { "epoch": 3.56, "learning_rate": 6.437857758958677e-05, "loss": 2.6441, "step": 45428 }, { "epoch": 3.56, "learning_rate": 6.437779346036227e-05, "loss": 2.6904, "step": 45429 }, { "epoch": 3.56, "learning_rate": 6.437700933113777e-05, "loss": 2.7074, "step": 45430 }, { "epoch": 3.56, "learning_rate": 6.437622520191328e-05, "loss": 2.603, "step": 45431 }, { "epoch": 3.56, "learning_rate": 6.437544107268878e-05, "loss": 2.7174, "step": 45432 }, { "epoch": 3.56, "learning_rate": 6.437465694346429e-05, "loss": 2.687, "step": 45433 }, { "epoch": 3.56, "learning_rate": 6.437387281423979e-05, "loss": 2.7989, "step": 45434 }, { "epoch": 3.56, "learning_rate": 6.437308868501529e-05, "loss": 2.6201, "step": 45435 }, { "epoch": 3.56, "learning_rate": 6.43723045557908e-05, "loss": 2.6, "step": 45436 }, { "epoch": 3.56, "learning_rate": 6.43715204265663e-05, "loss": 2.9103, "step": 45437 }, { "epoch": 3.56, "learning_rate": 6.43707362973418e-05, "loss": 2.7951, "step": 45438 }, { "epoch": 3.56, "learning_rate": 6.436995216811731e-05, "loss": 2.7081, "step": 45439 }, { "epoch": 3.56, "learning_rate": 6.436916803889281e-05, "loss": 2.7548, "step": 45440 }, { "epoch": 3.56, "learning_rate": 6.436838390966832e-05, "loss": 2.6959, "step": 45441 }, { "epoch": 3.56, "learning_rate": 6.436759978044382e-05, "loss": 2.6018, "step": 45442 }, { "epoch": 3.56, "learning_rate": 6.436681565121932e-05, "loss": 2.7025, "step": 45443 }, { "epoch": 3.56, "learning_rate": 6.436603152199483e-05, "loss": 2.5441, "step": 45444 }, { "epoch": 3.56, "learning_rate": 6.436524739277033e-05, "loss": 2.786, "step": 45445 }, { "epoch": 3.56, "learning_rate": 6.436446326354583e-05, "loss": 2.5512, "step": 45446 }, { "epoch": 3.56, "learning_rate": 6.436367913432134e-05, "loss": 3.0447, "step": 45447 }, { "epoch": 3.56, "learning_rate": 6.436289500509684e-05, "loss": 2.9794, "step": 45448 }, { "epoch": 3.56, "learning_rate": 6.436211087587234e-05, "loss": 2.6654, "step": 45449 }, { "epoch": 3.56, "learning_rate": 6.436132674664785e-05, "loss": 2.4778, "step": 45450 }, { "epoch": 3.56, "learning_rate": 6.436054261742335e-05, "loss": 2.7771, "step": 45451 }, { "epoch": 3.56, "learning_rate": 6.435975848819887e-05, "loss": 2.6739, "step": 45452 }, { "epoch": 3.56, "learning_rate": 6.435897435897437e-05, "loss": 2.7115, "step": 45453 }, { "epoch": 3.56, "learning_rate": 6.435819022974988e-05, "loss": 3.0228, "step": 45454 }, { "epoch": 3.56, "learning_rate": 6.435740610052538e-05, "loss": 2.7919, "step": 45455 }, { "epoch": 3.56, "learning_rate": 6.435662197130088e-05, "loss": 2.766, "step": 45456 }, { "epoch": 3.56, "learning_rate": 6.435583784207637e-05, "loss": 2.7403, "step": 45457 }, { "epoch": 3.56, "learning_rate": 6.435505371285188e-05, "loss": 2.8644, "step": 45458 }, { "epoch": 3.56, "learning_rate": 6.435426958362738e-05, "loss": 2.6375, "step": 45459 }, { "epoch": 3.56, "learning_rate": 6.435348545440288e-05, "loss": 2.7629, "step": 45460 }, { "epoch": 3.56, "learning_rate": 6.435270132517839e-05, "loss": 2.711, "step": 45461 }, { "epoch": 3.56, "learning_rate": 6.435191719595389e-05, "loss": 2.7311, "step": 45462 }, { "epoch": 3.56, "learning_rate": 6.43511330667294e-05, "loss": 2.7022, "step": 45463 }, { "epoch": 3.56, "learning_rate": 6.43503489375049e-05, "loss": 2.6748, "step": 45464 }, { "epoch": 3.57, "learning_rate": 6.43495648082804e-05, "loss": 2.6146, "step": 45465 }, { "epoch": 3.57, "learning_rate": 6.434878067905591e-05, "loss": 2.824, "step": 45466 }, { "epoch": 3.57, "learning_rate": 6.434799654983141e-05, "loss": 2.7238, "step": 45467 }, { "epoch": 3.57, "learning_rate": 6.434721242060691e-05, "loss": 2.6406, "step": 45468 }, { "epoch": 3.57, "learning_rate": 6.434642829138242e-05, "loss": 2.6919, "step": 45469 }, { "epoch": 3.57, "learning_rate": 6.434564416215792e-05, "loss": 2.6398, "step": 45470 }, { "epoch": 3.57, "learning_rate": 6.434486003293343e-05, "loss": 2.6029, "step": 45471 }, { "epoch": 3.57, "learning_rate": 6.434407590370893e-05, "loss": 2.6879, "step": 45472 }, { "epoch": 3.57, "learning_rate": 6.434329177448443e-05, "loss": 2.7527, "step": 45473 }, { "epoch": 3.57, "learning_rate": 6.434250764525994e-05, "loss": 2.6485, "step": 45474 }, { "epoch": 3.57, "learning_rate": 6.434172351603544e-05, "loss": 2.8533, "step": 45475 }, { "epoch": 3.57, "learning_rate": 6.434093938681094e-05, "loss": 2.7475, "step": 45476 }, { "epoch": 3.57, "learning_rate": 6.434015525758645e-05, "loss": 2.7414, "step": 45477 }, { "epoch": 3.57, "learning_rate": 6.433937112836196e-05, "loss": 2.8114, "step": 45478 }, { "epoch": 3.57, "learning_rate": 6.433858699913747e-05, "loss": 2.6707, "step": 45479 }, { "epoch": 3.57, "learning_rate": 6.433780286991297e-05, "loss": 2.6919, "step": 45480 }, { "epoch": 3.57, "learning_rate": 6.433701874068848e-05, "loss": 2.8539, "step": 45481 }, { "epoch": 3.57, "learning_rate": 6.433623461146398e-05, "loss": 2.8195, "step": 45482 }, { "epoch": 3.57, "learning_rate": 6.433545048223948e-05, "loss": 2.7111, "step": 45483 }, { "epoch": 3.57, "learning_rate": 6.433466635301499e-05, "loss": 2.8123, "step": 45484 }, { "epoch": 3.57, "learning_rate": 6.433388222379049e-05, "loss": 2.9232, "step": 45485 }, { "epoch": 3.57, "learning_rate": 6.4333098094566e-05, "loss": 2.5618, "step": 45486 }, { "epoch": 3.57, "learning_rate": 6.43323139653415e-05, "loss": 2.8048, "step": 45487 }, { "epoch": 3.57, "learning_rate": 6.4331529836117e-05, "loss": 2.539, "step": 45488 }, { "epoch": 3.57, "learning_rate": 6.43307457068925e-05, "loss": 2.6584, "step": 45489 }, { "epoch": 3.57, "learning_rate": 6.432996157766801e-05, "loss": 2.7794, "step": 45490 }, { "epoch": 3.57, "learning_rate": 6.432917744844351e-05, "loss": 2.7701, "step": 45491 }, { "epoch": 3.57, "learning_rate": 6.432839331921902e-05, "loss": 2.7527, "step": 45492 }, { "epoch": 3.57, "learning_rate": 6.43276091899945e-05, "loss": 2.7458, "step": 45493 }, { "epoch": 3.57, "learning_rate": 6.432682506077001e-05, "loss": 2.8882, "step": 45494 }, { "epoch": 3.57, "learning_rate": 6.432604093154551e-05, "loss": 2.7875, "step": 45495 }, { "epoch": 3.57, "learning_rate": 6.432525680232102e-05, "loss": 2.7204, "step": 45496 }, { "epoch": 3.57, "learning_rate": 6.432447267309652e-05, "loss": 2.698, "step": 45497 }, { "epoch": 3.57, "learning_rate": 6.432368854387202e-05, "loss": 2.805, "step": 45498 }, { "epoch": 3.57, "learning_rate": 6.432290441464753e-05, "loss": 2.8854, "step": 45499 }, { "epoch": 3.57, "learning_rate": 6.432212028542303e-05, "loss": 2.9283, "step": 45500 }, { "epoch": 3.57, "learning_rate": 6.432133615619854e-05, "loss": 3.0169, "step": 45501 }, { "epoch": 3.57, "learning_rate": 6.432055202697404e-05, "loss": 2.6387, "step": 45502 }, { "epoch": 3.57, "learning_rate": 6.431976789774956e-05, "loss": 2.7224, "step": 45503 }, { "epoch": 3.57, "learning_rate": 6.431898376852506e-05, "loss": 2.6563, "step": 45504 }, { "epoch": 3.57, "learning_rate": 6.431819963930056e-05, "loss": 2.7832, "step": 45505 }, { "epoch": 3.57, "learning_rate": 6.431741551007607e-05, "loss": 2.4758, "step": 45506 }, { "epoch": 3.57, "learning_rate": 6.431663138085157e-05, "loss": 2.653, "step": 45507 }, { "epoch": 3.57, "learning_rate": 6.431584725162708e-05, "loss": 2.9055, "step": 45508 }, { "epoch": 3.57, "learning_rate": 6.431506312240258e-05, "loss": 2.6535, "step": 45509 }, { "epoch": 3.57, "learning_rate": 6.431427899317808e-05, "loss": 2.7247, "step": 45510 }, { "epoch": 3.57, "learning_rate": 6.431349486395359e-05, "loss": 2.916, "step": 45511 }, { "epoch": 3.57, "learning_rate": 6.431271073472909e-05, "loss": 2.8736, "step": 45512 }, { "epoch": 3.57, "learning_rate": 6.43119266055046e-05, "loss": 2.6154, "step": 45513 }, { "epoch": 3.57, "learning_rate": 6.43111424762801e-05, "loss": 2.5524, "step": 45514 }, { "epoch": 3.57, "learning_rate": 6.43103583470556e-05, "loss": 2.7277, "step": 45515 }, { "epoch": 3.57, "learning_rate": 6.43095742178311e-05, "loss": 2.6977, "step": 45516 }, { "epoch": 3.57, "learning_rate": 6.430879008860661e-05, "loss": 2.7898, "step": 45517 }, { "epoch": 3.57, "learning_rate": 6.430800595938211e-05, "loss": 2.627, "step": 45518 }, { "epoch": 3.57, "learning_rate": 6.430722183015762e-05, "loss": 2.7847, "step": 45519 }, { "epoch": 3.57, "learning_rate": 6.430643770093312e-05, "loss": 2.6016, "step": 45520 }, { "epoch": 3.57, "learning_rate": 6.430565357170862e-05, "loss": 2.93, "step": 45521 }, { "epoch": 3.57, "learning_rate": 6.430486944248413e-05, "loss": 2.8216, "step": 45522 }, { "epoch": 3.57, "learning_rate": 6.430408531325963e-05, "loss": 2.8896, "step": 45523 }, { "epoch": 3.57, "learning_rate": 6.430330118403513e-05, "loss": 2.7922, "step": 45524 }, { "epoch": 3.57, "learning_rate": 6.430251705481064e-05, "loss": 2.7439, "step": 45525 }, { "epoch": 3.57, "learning_rate": 6.430173292558614e-05, "loss": 2.5731, "step": 45526 }, { "epoch": 3.57, "learning_rate": 6.430094879636164e-05, "loss": 2.9528, "step": 45527 }, { "epoch": 3.57, "learning_rate": 6.430016466713715e-05, "loss": 2.7183, "step": 45528 }, { "epoch": 3.57, "learning_rate": 6.429938053791265e-05, "loss": 2.7821, "step": 45529 }, { "epoch": 3.57, "learning_rate": 6.429859640868816e-05, "loss": 2.861, "step": 45530 }, { "epoch": 3.57, "learning_rate": 6.429781227946366e-05, "loss": 2.7928, "step": 45531 }, { "epoch": 3.57, "learning_rate": 6.429702815023916e-05, "loss": 2.7248, "step": 45532 }, { "epoch": 3.57, "learning_rate": 6.429624402101467e-05, "loss": 2.7807, "step": 45533 }, { "epoch": 3.57, "learning_rate": 6.429545989179017e-05, "loss": 2.8418, "step": 45534 }, { "epoch": 3.57, "learning_rate": 6.429467576256567e-05, "loss": 2.5643, "step": 45535 }, { "epoch": 3.57, "learning_rate": 6.429389163334118e-05, "loss": 2.7434, "step": 45536 }, { "epoch": 3.57, "learning_rate": 6.429310750411668e-05, "loss": 2.6418, "step": 45537 }, { "epoch": 3.57, "learning_rate": 6.429232337489219e-05, "loss": 2.6184, "step": 45538 }, { "epoch": 3.57, "learning_rate": 6.429153924566769e-05, "loss": 2.8532, "step": 45539 }, { "epoch": 3.57, "learning_rate": 6.429075511644319e-05, "loss": 2.6611, "step": 45540 }, { "epoch": 3.57, "learning_rate": 6.42899709872187e-05, "loss": 2.844, "step": 45541 }, { "epoch": 3.57, "learning_rate": 6.42891868579942e-05, "loss": 2.6452, "step": 45542 }, { "epoch": 3.57, "learning_rate": 6.42884027287697e-05, "loss": 2.9028, "step": 45543 }, { "epoch": 3.57, "learning_rate": 6.428761859954521e-05, "loss": 2.6396, "step": 45544 }, { "epoch": 3.57, "learning_rate": 6.428683447032071e-05, "loss": 2.5316, "step": 45545 }, { "epoch": 3.57, "learning_rate": 6.428605034109621e-05, "loss": 2.9009, "step": 45546 }, { "epoch": 3.57, "learning_rate": 6.428526621187172e-05, "loss": 2.6432, "step": 45547 }, { "epoch": 3.57, "learning_rate": 6.428448208264722e-05, "loss": 2.5843, "step": 45548 }, { "epoch": 3.57, "learning_rate": 6.428369795342273e-05, "loss": 2.8987, "step": 45549 }, { "epoch": 3.57, "learning_rate": 6.428291382419823e-05, "loss": 2.7476, "step": 45550 }, { "epoch": 3.57, "learning_rate": 6.428212969497373e-05, "loss": 2.7462, "step": 45551 }, { "epoch": 3.57, "learning_rate": 6.428134556574924e-05, "loss": 2.8046, "step": 45552 }, { "epoch": 3.57, "learning_rate": 6.428056143652474e-05, "loss": 2.9319, "step": 45553 }, { "epoch": 3.57, "learning_rate": 6.427977730730024e-05, "loss": 2.5332, "step": 45554 }, { "epoch": 3.57, "learning_rate": 6.427899317807575e-05, "loss": 2.7383, "step": 45555 }, { "epoch": 3.57, "learning_rate": 6.427820904885125e-05, "loss": 2.9188, "step": 45556 }, { "epoch": 3.57, "learning_rate": 6.427742491962676e-05, "loss": 2.7214, "step": 45557 }, { "epoch": 3.57, "learning_rate": 6.427664079040226e-05, "loss": 2.6233, "step": 45558 }, { "epoch": 3.57, "learning_rate": 6.427585666117776e-05, "loss": 2.8878, "step": 45559 }, { "epoch": 3.57, "learning_rate": 6.427507253195328e-05, "loss": 2.5521, "step": 45560 }, { "epoch": 3.57, "learning_rate": 6.427428840272878e-05, "loss": 2.9322, "step": 45561 }, { "epoch": 3.57, "learning_rate": 6.427350427350429e-05, "loss": 2.8187, "step": 45562 }, { "epoch": 3.57, "learning_rate": 6.427272014427979e-05, "loss": 2.5817, "step": 45563 }, { "epoch": 3.57, "learning_rate": 6.42719360150553e-05, "loss": 2.703, "step": 45564 }, { "epoch": 3.57, "learning_rate": 6.427115188583078e-05, "loss": 2.7939, "step": 45565 }, { "epoch": 3.57, "learning_rate": 6.427036775660629e-05, "loss": 3.0241, "step": 45566 }, { "epoch": 3.57, "learning_rate": 6.426958362738179e-05, "loss": 2.851, "step": 45567 }, { "epoch": 3.57, "learning_rate": 6.42687994981573e-05, "loss": 2.6702, "step": 45568 }, { "epoch": 3.57, "learning_rate": 6.42680153689328e-05, "loss": 2.8397, "step": 45569 }, { "epoch": 3.57, "learning_rate": 6.42672312397083e-05, "loss": 2.5542, "step": 45570 }, { "epoch": 3.57, "learning_rate": 6.42664471104838e-05, "loss": 2.627, "step": 45571 }, { "epoch": 3.57, "learning_rate": 6.426566298125931e-05, "loss": 2.839, "step": 45572 }, { "epoch": 3.57, "learning_rate": 6.426487885203481e-05, "loss": 2.7577, "step": 45573 }, { "epoch": 3.57, "learning_rate": 6.426409472281032e-05, "loss": 2.6173, "step": 45574 }, { "epoch": 3.57, "learning_rate": 6.426331059358582e-05, "loss": 2.8148, "step": 45575 }, { "epoch": 3.57, "learning_rate": 6.426252646436133e-05, "loss": 2.6622, "step": 45576 }, { "epoch": 3.57, "learning_rate": 6.426174233513683e-05, "loss": 2.7932, "step": 45577 }, { "epoch": 3.57, "learning_rate": 6.426095820591233e-05, "loss": 2.8884, "step": 45578 }, { "epoch": 3.57, "learning_rate": 6.426017407668784e-05, "loss": 2.6362, "step": 45579 }, { "epoch": 3.57, "learning_rate": 6.425938994746334e-05, "loss": 2.5787, "step": 45580 }, { "epoch": 3.57, "learning_rate": 6.425860581823884e-05, "loss": 2.9252, "step": 45581 }, { "epoch": 3.57, "learning_rate": 6.425782168901435e-05, "loss": 2.5786, "step": 45582 }, { "epoch": 3.57, "learning_rate": 6.425703755978985e-05, "loss": 2.9785, "step": 45583 }, { "epoch": 3.57, "learning_rate": 6.425625343056535e-05, "loss": 2.6219, "step": 45584 }, { "epoch": 3.57, "learning_rate": 6.425546930134086e-05, "loss": 3.1977, "step": 45585 }, { "epoch": 3.57, "learning_rate": 6.425468517211638e-05, "loss": 2.8555, "step": 45586 }, { "epoch": 3.57, "learning_rate": 6.425390104289188e-05, "loss": 2.961, "step": 45587 }, { "epoch": 3.57, "learning_rate": 6.425311691366738e-05, "loss": 2.6785, "step": 45588 }, { "epoch": 3.57, "learning_rate": 6.425233278444289e-05, "loss": 3.0087, "step": 45589 }, { "epoch": 3.57, "learning_rate": 6.425154865521839e-05, "loss": 2.5899, "step": 45590 }, { "epoch": 3.57, "learning_rate": 6.42507645259939e-05, "loss": 2.9961, "step": 45591 }, { "epoch": 3.58, "learning_rate": 6.42499803967694e-05, "loss": 2.9834, "step": 45592 }, { "epoch": 3.58, "learning_rate": 6.42491962675449e-05, "loss": 2.7188, "step": 45593 }, { "epoch": 3.58, "learning_rate": 6.42484121383204e-05, "loss": 2.6116, "step": 45594 }, { "epoch": 3.58, "learning_rate": 6.424762800909591e-05, "loss": 2.6372, "step": 45595 }, { "epoch": 3.58, "learning_rate": 6.424684387987141e-05, "loss": 2.8656, "step": 45596 }, { "epoch": 3.58, "learning_rate": 6.424605975064692e-05, "loss": 2.9581, "step": 45597 }, { "epoch": 3.58, "learning_rate": 6.424527562142242e-05, "loss": 2.7665, "step": 45598 }, { "epoch": 3.58, "learning_rate": 6.424449149219792e-05, "loss": 2.5422, "step": 45599 }, { "epoch": 3.58, "learning_rate": 6.424370736297343e-05, "loss": 2.8015, "step": 45600 }, { "epoch": 3.58, "learning_rate": 6.424292323374892e-05, "loss": 2.495, "step": 45601 }, { "epoch": 3.58, "learning_rate": 6.424213910452442e-05, "loss": 2.7235, "step": 45602 }, { "epoch": 3.58, "learning_rate": 6.424135497529992e-05, "loss": 2.6615, "step": 45603 }, { "epoch": 3.58, "learning_rate": 6.424057084607543e-05, "loss": 2.562, "step": 45604 }, { "epoch": 3.58, "learning_rate": 6.423978671685093e-05, "loss": 2.6414, "step": 45605 }, { "epoch": 3.58, "learning_rate": 6.423900258762644e-05, "loss": 2.9185, "step": 45606 }, { "epoch": 3.58, "learning_rate": 6.423821845840194e-05, "loss": 2.799, "step": 45607 }, { "epoch": 3.58, "learning_rate": 6.423743432917744e-05, "loss": 2.6634, "step": 45608 }, { "epoch": 3.58, "learning_rate": 6.423665019995295e-05, "loss": 2.6015, "step": 45609 }, { "epoch": 3.58, "learning_rate": 6.423586607072845e-05, "loss": 2.8873, "step": 45610 }, { "epoch": 3.58, "learning_rate": 6.423508194150395e-05, "loss": 2.7447, "step": 45611 }, { "epoch": 3.58, "learning_rate": 6.423429781227947e-05, "loss": 2.8545, "step": 45612 }, { "epoch": 3.58, "learning_rate": 6.423351368305497e-05, "loss": 2.7108, "step": 45613 }, { "epoch": 3.58, "learning_rate": 6.423272955383048e-05, "loss": 2.7567, "step": 45614 }, { "epoch": 3.58, "learning_rate": 6.423194542460598e-05, "loss": 2.8244, "step": 45615 }, { "epoch": 3.58, "learning_rate": 6.423116129538149e-05, "loss": 2.4383, "step": 45616 }, { "epoch": 3.58, "learning_rate": 6.423037716615699e-05, "loss": 2.6666, "step": 45617 }, { "epoch": 3.58, "learning_rate": 6.422959303693249e-05, "loss": 2.6996, "step": 45618 }, { "epoch": 3.58, "learning_rate": 6.4228808907708e-05, "loss": 2.4766, "step": 45619 }, { "epoch": 3.58, "learning_rate": 6.42280247784835e-05, "loss": 2.6417, "step": 45620 }, { "epoch": 3.58, "learning_rate": 6.4227240649259e-05, "loss": 2.8974, "step": 45621 }, { "epoch": 3.58, "learning_rate": 6.422645652003451e-05, "loss": 2.6497, "step": 45622 }, { "epoch": 3.58, "learning_rate": 6.422567239081001e-05, "loss": 2.8999, "step": 45623 }, { "epoch": 3.58, "learning_rate": 6.422488826158552e-05, "loss": 2.7033, "step": 45624 }, { "epoch": 3.58, "learning_rate": 6.422410413236102e-05, "loss": 2.5487, "step": 45625 }, { "epoch": 3.58, "learning_rate": 6.422332000313652e-05, "loss": 2.8676, "step": 45626 }, { "epoch": 3.58, "learning_rate": 6.422253587391203e-05, "loss": 2.7961, "step": 45627 }, { "epoch": 3.58, "learning_rate": 6.422175174468753e-05, "loss": 2.715, "step": 45628 }, { "epoch": 3.58, "learning_rate": 6.422096761546303e-05, "loss": 2.752, "step": 45629 }, { "epoch": 3.58, "learning_rate": 6.422018348623854e-05, "loss": 2.8365, "step": 45630 }, { "epoch": 3.58, "learning_rate": 6.421939935701404e-05, "loss": 2.6739, "step": 45631 }, { "epoch": 3.58, "learning_rate": 6.421861522778954e-05, "loss": 2.5823, "step": 45632 }, { "epoch": 3.58, "learning_rate": 6.421783109856505e-05, "loss": 2.7752, "step": 45633 }, { "epoch": 3.58, "learning_rate": 6.421704696934055e-05, "loss": 2.7798, "step": 45634 }, { "epoch": 3.58, "learning_rate": 6.421626284011606e-05, "loss": 2.6954, "step": 45635 }, { "epoch": 3.58, "learning_rate": 6.421547871089156e-05, "loss": 2.5245, "step": 45636 }, { "epoch": 3.58, "learning_rate": 6.421469458166706e-05, "loss": 2.8183, "step": 45637 }, { "epoch": 3.58, "learning_rate": 6.421391045244257e-05, "loss": 3.0252, "step": 45638 }, { "epoch": 3.58, "learning_rate": 6.421312632321807e-05, "loss": 2.7546, "step": 45639 }, { "epoch": 3.58, "learning_rate": 6.421234219399357e-05, "loss": 2.7761, "step": 45640 }, { "epoch": 3.58, "learning_rate": 6.421155806476908e-05, "loss": 2.8894, "step": 45641 }, { "epoch": 3.58, "learning_rate": 6.421077393554458e-05, "loss": 2.6856, "step": 45642 }, { "epoch": 3.58, "learning_rate": 6.420998980632009e-05, "loss": 2.6842, "step": 45643 }, { "epoch": 3.58, "learning_rate": 6.420920567709559e-05, "loss": 2.8202, "step": 45644 }, { "epoch": 3.58, "learning_rate": 6.420842154787109e-05, "loss": 2.7021, "step": 45645 }, { "epoch": 3.58, "learning_rate": 6.42076374186466e-05, "loss": 2.7844, "step": 45646 }, { "epoch": 3.58, "learning_rate": 6.42068532894221e-05, "loss": 2.8027, "step": 45647 }, { "epoch": 3.58, "learning_rate": 6.42060691601976e-05, "loss": 2.5389, "step": 45648 }, { "epoch": 3.58, "learning_rate": 6.420528503097311e-05, "loss": 2.7542, "step": 45649 }, { "epoch": 3.58, "learning_rate": 6.420450090174861e-05, "loss": 2.8782, "step": 45650 }, { "epoch": 3.58, "learning_rate": 6.420371677252411e-05, "loss": 2.6989, "step": 45651 }, { "epoch": 3.58, "learning_rate": 6.420293264329962e-05, "loss": 2.7427, "step": 45652 }, { "epoch": 3.58, "learning_rate": 6.420214851407512e-05, "loss": 2.7319, "step": 45653 }, { "epoch": 3.58, "learning_rate": 6.420136438485063e-05, "loss": 2.8837, "step": 45654 }, { "epoch": 3.58, "learning_rate": 6.420058025562613e-05, "loss": 2.8827, "step": 45655 }, { "epoch": 3.58, "learning_rate": 6.419979612640163e-05, "loss": 2.7124, "step": 45656 }, { "epoch": 3.58, "learning_rate": 6.419901199717714e-05, "loss": 2.7413, "step": 45657 }, { "epoch": 3.58, "learning_rate": 6.419822786795264e-05, "loss": 2.828, "step": 45658 }, { "epoch": 3.58, "learning_rate": 6.419744373872814e-05, "loss": 2.919, "step": 45659 }, { "epoch": 3.58, "learning_rate": 6.419665960950365e-05, "loss": 2.6199, "step": 45660 }, { "epoch": 3.58, "learning_rate": 6.419587548027915e-05, "loss": 2.8826, "step": 45661 }, { "epoch": 3.58, "learning_rate": 6.419509135105465e-05, "loss": 2.7116, "step": 45662 }, { "epoch": 3.58, "learning_rate": 6.419430722183016e-05, "loss": 2.6153, "step": 45663 }, { "epoch": 3.58, "learning_rate": 6.419352309260566e-05, "loss": 2.5647, "step": 45664 }, { "epoch": 3.58, "learning_rate": 6.419273896338117e-05, "loss": 2.8235, "step": 45665 }, { "epoch": 3.58, "learning_rate": 6.419195483415667e-05, "loss": 2.7537, "step": 45666 }, { "epoch": 3.58, "learning_rate": 6.419117070493217e-05, "loss": 2.7819, "step": 45667 }, { "epoch": 3.58, "learning_rate": 6.419038657570769e-05, "loss": 2.7166, "step": 45668 }, { "epoch": 3.58, "learning_rate": 6.41896024464832e-05, "loss": 2.6581, "step": 45669 }, { "epoch": 3.58, "learning_rate": 6.41888183172587e-05, "loss": 2.6523, "step": 45670 }, { "epoch": 3.58, "learning_rate": 6.41880341880342e-05, "loss": 2.5447, "step": 45671 }, { "epoch": 3.58, "learning_rate": 6.41872500588097e-05, "loss": 2.6441, "step": 45672 }, { "epoch": 3.58, "learning_rate": 6.41864659295852e-05, "loss": 2.732, "step": 45673 }, { "epoch": 3.58, "learning_rate": 6.41856818003607e-05, "loss": 2.5505, "step": 45674 }, { "epoch": 3.58, "learning_rate": 6.41848976711362e-05, "loss": 2.6852, "step": 45675 }, { "epoch": 3.58, "learning_rate": 6.41841135419117e-05, "loss": 2.6464, "step": 45676 }, { "epoch": 3.58, "learning_rate": 6.418332941268721e-05, "loss": 2.9017, "step": 45677 }, { "epoch": 3.58, "learning_rate": 6.418254528346271e-05, "loss": 2.9369, "step": 45678 }, { "epoch": 3.58, "learning_rate": 6.418176115423822e-05, "loss": 2.7352, "step": 45679 }, { "epoch": 3.58, "learning_rate": 6.418097702501372e-05, "loss": 2.6961, "step": 45680 }, { "epoch": 3.58, "learning_rate": 6.418019289578922e-05, "loss": 2.7808, "step": 45681 }, { "epoch": 3.58, "learning_rate": 6.417940876656473e-05, "loss": 2.7574, "step": 45682 }, { "epoch": 3.58, "learning_rate": 6.417862463734023e-05, "loss": 2.553, "step": 45683 }, { "epoch": 3.58, "learning_rate": 6.417784050811574e-05, "loss": 2.7221, "step": 45684 }, { "epoch": 3.58, "learning_rate": 6.417705637889124e-05, "loss": 2.6377, "step": 45685 }, { "epoch": 3.58, "learning_rate": 6.417627224966674e-05, "loss": 2.6583, "step": 45686 }, { "epoch": 3.58, "learning_rate": 6.417548812044225e-05, "loss": 2.53, "step": 45687 }, { "epoch": 3.58, "learning_rate": 6.417470399121775e-05, "loss": 2.6698, "step": 45688 }, { "epoch": 3.58, "learning_rate": 6.417391986199325e-05, "loss": 2.5836, "step": 45689 }, { "epoch": 3.58, "learning_rate": 6.417313573276876e-05, "loss": 2.7238, "step": 45690 }, { "epoch": 3.58, "learning_rate": 6.417235160354426e-05, "loss": 2.9116, "step": 45691 }, { "epoch": 3.58, "learning_rate": 6.417156747431977e-05, "loss": 2.7272, "step": 45692 }, { "epoch": 3.58, "learning_rate": 6.417078334509527e-05, "loss": 2.7542, "step": 45693 }, { "epoch": 3.58, "learning_rate": 6.416999921587079e-05, "loss": 2.6902, "step": 45694 }, { "epoch": 3.58, "learning_rate": 6.416921508664629e-05, "loss": 2.6378, "step": 45695 }, { "epoch": 3.58, "learning_rate": 6.41684309574218e-05, "loss": 2.904, "step": 45696 }, { "epoch": 3.58, "learning_rate": 6.41676468281973e-05, "loss": 2.6718, "step": 45697 }, { "epoch": 3.58, "learning_rate": 6.41668626989728e-05, "loss": 2.9208, "step": 45698 }, { "epoch": 3.58, "learning_rate": 6.41660785697483e-05, "loss": 2.9953, "step": 45699 }, { "epoch": 3.58, "learning_rate": 6.416529444052381e-05, "loss": 2.6679, "step": 45700 }, { "epoch": 3.58, "learning_rate": 6.416451031129931e-05, "loss": 2.8886, "step": 45701 }, { "epoch": 3.58, "learning_rate": 6.416372618207482e-05, "loss": 2.7214, "step": 45702 }, { "epoch": 3.58, "learning_rate": 6.416294205285032e-05, "loss": 2.8427, "step": 45703 }, { "epoch": 3.58, "learning_rate": 6.416215792362582e-05, "loss": 2.7825, "step": 45704 }, { "epoch": 3.58, "learning_rate": 6.416137379440133e-05, "loss": 2.8561, "step": 45705 }, { "epoch": 3.58, "learning_rate": 6.416058966517683e-05, "loss": 2.6734, "step": 45706 }, { "epoch": 3.58, "learning_rate": 6.415980553595233e-05, "loss": 2.8057, "step": 45707 }, { "epoch": 3.58, "learning_rate": 6.415902140672784e-05, "loss": 2.8401, "step": 45708 }, { "epoch": 3.58, "learning_rate": 6.415823727750333e-05, "loss": 2.6595, "step": 45709 }, { "epoch": 3.58, "learning_rate": 6.415745314827883e-05, "loss": 2.7349, "step": 45710 }, { "epoch": 3.58, "learning_rate": 6.415666901905434e-05, "loss": 2.9263, "step": 45711 }, { "epoch": 3.58, "learning_rate": 6.415588488982984e-05, "loss": 2.8332, "step": 45712 }, { "epoch": 3.58, "learning_rate": 6.415510076060534e-05, "loss": 2.5503, "step": 45713 }, { "epoch": 3.58, "learning_rate": 6.415431663138085e-05, "loss": 2.6384, "step": 45714 }, { "epoch": 3.58, "learning_rate": 6.415353250215635e-05, "loss": 2.6136, "step": 45715 }, { "epoch": 3.58, "learning_rate": 6.415274837293185e-05, "loss": 2.8382, "step": 45716 }, { "epoch": 3.58, "learning_rate": 6.415196424370736e-05, "loss": 2.6182, "step": 45717 }, { "epoch": 3.58, "learning_rate": 6.415118011448286e-05, "loss": 2.7749, "step": 45718 }, { "epoch": 3.58, "learning_rate": 6.415039598525836e-05, "loss": 2.9571, "step": 45719 }, { "epoch": 3.59, "learning_rate": 6.414961185603388e-05, "loss": 2.5808, "step": 45720 }, { "epoch": 3.59, "learning_rate": 6.414882772680939e-05, "loss": 2.5322, "step": 45721 }, { "epoch": 3.59, "learning_rate": 6.414804359758489e-05, "loss": 2.8125, "step": 45722 }, { "epoch": 3.59, "learning_rate": 6.414725946836039e-05, "loss": 2.8079, "step": 45723 }, { "epoch": 3.59, "learning_rate": 6.41464753391359e-05, "loss": 3.1123, "step": 45724 }, { "epoch": 3.59, "learning_rate": 6.41456912099114e-05, "loss": 2.6964, "step": 45725 }, { "epoch": 3.59, "learning_rate": 6.41449070806869e-05, "loss": 2.7745, "step": 45726 }, { "epoch": 3.59, "learning_rate": 6.414412295146241e-05, "loss": 2.7734, "step": 45727 }, { "epoch": 3.59, "learning_rate": 6.414333882223791e-05, "loss": 2.6877, "step": 45728 }, { "epoch": 3.59, "learning_rate": 6.414255469301341e-05, "loss": 2.753, "step": 45729 }, { "epoch": 3.59, "learning_rate": 6.414177056378892e-05, "loss": 2.8419, "step": 45730 }, { "epoch": 3.59, "learning_rate": 6.414098643456442e-05, "loss": 2.9023, "step": 45731 }, { "epoch": 3.59, "learning_rate": 6.414020230533993e-05, "loss": 2.8393, "step": 45732 }, { "epoch": 3.59, "learning_rate": 6.413941817611543e-05, "loss": 2.3931, "step": 45733 }, { "epoch": 3.59, "learning_rate": 6.413863404689093e-05, "loss": 2.9509, "step": 45734 }, { "epoch": 3.59, "learning_rate": 6.413784991766644e-05, "loss": 2.8957, "step": 45735 }, { "epoch": 3.59, "learning_rate": 6.413706578844194e-05, "loss": 2.8633, "step": 45736 }, { "epoch": 3.59, "learning_rate": 6.413628165921744e-05, "loss": 2.606, "step": 45737 }, { "epoch": 3.59, "learning_rate": 6.413549752999295e-05, "loss": 2.6085, "step": 45738 }, { "epoch": 3.59, "learning_rate": 6.413471340076845e-05, "loss": 2.7247, "step": 45739 }, { "epoch": 3.59, "learning_rate": 6.413392927154396e-05, "loss": 2.8221, "step": 45740 }, { "epoch": 3.59, "learning_rate": 6.413314514231946e-05, "loss": 2.6663, "step": 45741 }, { "epoch": 3.59, "learning_rate": 6.413236101309496e-05, "loss": 2.7567, "step": 45742 }, { "epoch": 3.59, "learning_rate": 6.413157688387047e-05, "loss": 2.6854, "step": 45743 }, { "epoch": 3.59, "learning_rate": 6.413079275464597e-05, "loss": 2.7072, "step": 45744 }, { "epoch": 3.59, "learning_rate": 6.413000862542147e-05, "loss": 2.6642, "step": 45745 }, { "epoch": 3.59, "learning_rate": 6.412922449619698e-05, "loss": 2.7368, "step": 45746 }, { "epoch": 3.59, "learning_rate": 6.412844036697248e-05, "loss": 2.6941, "step": 45747 }, { "epoch": 3.59, "learning_rate": 6.412765623774798e-05, "loss": 2.6648, "step": 45748 }, { "epoch": 3.59, "learning_rate": 6.412687210852349e-05, "loss": 2.9998, "step": 45749 }, { "epoch": 3.59, "learning_rate": 6.412608797929899e-05, "loss": 2.8795, "step": 45750 }, { "epoch": 3.59, "learning_rate": 6.41253038500745e-05, "loss": 2.6988, "step": 45751 }, { "epoch": 3.59, "learning_rate": 6.412451972085e-05, "loss": 2.8004, "step": 45752 }, { "epoch": 3.59, "learning_rate": 6.41237355916255e-05, "loss": 2.6436, "step": 45753 }, { "epoch": 3.59, "learning_rate": 6.412295146240101e-05, "loss": 2.8476, "step": 45754 }, { "epoch": 3.59, "learning_rate": 6.412216733317651e-05, "loss": 2.9271, "step": 45755 }, { "epoch": 3.59, "learning_rate": 6.412138320395201e-05, "loss": 2.7187, "step": 45756 }, { "epoch": 3.59, "learning_rate": 6.412059907472752e-05, "loss": 2.8431, "step": 45757 }, { "epoch": 3.59, "learning_rate": 6.411981494550302e-05, "loss": 2.9113, "step": 45758 }, { "epoch": 3.59, "learning_rate": 6.411903081627853e-05, "loss": 3.2007, "step": 45759 }, { "epoch": 3.59, "learning_rate": 6.411824668705403e-05, "loss": 2.7328, "step": 45760 }, { "epoch": 3.59, "learning_rate": 6.411746255782953e-05, "loss": 2.8972, "step": 45761 }, { "epoch": 3.59, "learning_rate": 6.411667842860504e-05, "loss": 2.7233, "step": 45762 }, { "epoch": 3.59, "learning_rate": 6.411589429938054e-05, "loss": 2.7172, "step": 45763 }, { "epoch": 3.59, "learning_rate": 6.411511017015604e-05, "loss": 2.7139, "step": 45764 }, { "epoch": 3.59, "learning_rate": 6.411432604093155e-05, "loss": 2.8345, "step": 45765 }, { "epoch": 3.59, "learning_rate": 6.411354191170705e-05, "loss": 2.5903, "step": 45766 }, { "epoch": 3.59, "learning_rate": 6.411275778248255e-05, "loss": 2.8646, "step": 45767 }, { "epoch": 3.59, "learning_rate": 6.411197365325806e-05, "loss": 2.5417, "step": 45768 }, { "epoch": 3.59, "learning_rate": 6.411118952403356e-05, "loss": 2.7552, "step": 45769 }, { "epoch": 3.59, "learning_rate": 6.411040539480907e-05, "loss": 2.8575, "step": 45770 }, { "epoch": 3.59, "learning_rate": 6.410962126558457e-05, "loss": 2.8222, "step": 45771 }, { "epoch": 3.59, "learning_rate": 6.410883713636007e-05, "loss": 2.7196, "step": 45772 }, { "epoch": 3.59, "learning_rate": 6.410805300713558e-05, "loss": 2.7212, "step": 45773 }, { "epoch": 3.59, "learning_rate": 6.410726887791108e-05, "loss": 2.7266, "step": 45774 }, { "epoch": 3.59, "learning_rate": 6.410648474868658e-05, "loss": 2.9235, "step": 45775 }, { "epoch": 3.59, "learning_rate": 6.41057006194621e-05, "loss": 2.8773, "step": 45776 }, { "epoch": 3.59, "learning_rate": 6.41049164902376e-05, "loss": 2.6261, "step": 45777 }, { "epoch": 3.59, "learning_rate": 6.410413236101311e-05, "loss": 2.6504, "step": 45778 }, { "epoch": 3.59, "learning_rate": 6.410334823178861e-05, "loss": 2.706, "step": 45779 }, { "epoch": 3.59, "learning_rate": 6.410256410256412e-05, "loss": 2.7185, "step": 45780 }, { "epoch": 3.59, "learning_rate": 6.41017799733396e-05, "loss": 2.8706, "step": 45781 }, { "epoch": 3.59, "learning_rate": 6.410099584411511e-05, "loss": 2.8437, "step": 45782 }, { "epoch": 3.59, "learning_rate": 6.410021171489061e-05, "loss": 2.7311, "step": 45783 }, { "epoch": 3.59, "learning_rate": 6.409942758566612e-05, "loss": 2.7682, "step": 45784 }, { "epoch": 3.59, "learning_rate": 6.409864345644162e-05, "loss": 2.7761, "step": 45785 }, { "epoch": 3.59, "learning_rate": 6.409785932721712e-05, "loss": 2.9508, "step": 45786 }, { "epoch": 3.59, "learning_rate": 6.409707519799263e-05, "loss": 2.7003, "step": 45787 }, { "epoch": 3.59, "learning_rate": 6.409629106876813e-05, "loss": 2.7394, "step": 45788 }, { "epoch": 3.59, "learning_rate": 6.409550693954364e-05, "loss": 3.0253, "step": 45789 }, { "epoch": 3.59, "learning_rate": 6.409472281031914e-05, "loss": 2.526, "step": 45790 }, { "epoch": 3.59, "learning_rate": 6.409393868109464e-05, "loss": 2.7275, "step": 45791 }, { "epoch": 3.59, "learning_rate": 6.409315455187015e-05, "loss": 2.4996, "step": 45792 }, { "epoch": 3.59, "learning_rate": 6.409237042264565e-05, "loss": 2.6726, "step": 45793 }, { "epoch": 3.59, "learning_rate": 6.409158629342115e-05, "loss": 2.6134, "step": 45794 }, { "epoch": 3.59, "learning_rate": 6.409080216419666e-05, "loss": 2.8153, "step": 45795 }, { "epoch": 3.59, "learning_rate": 6.409001803497216e-05, "loss": 2.6984, "step": 45796 }, { "epoch": 3.59, "learning_rate": 6.408923390574767e-05, "loss": 2.7275, "step": 45797 }, { "epoch": 3.59, "learning_rate": 6.408844977652317e-05, "loss": 2.9328, "step": 45798 }, { "epoch": 3.59, "learning_rate": 6.408766564729867e-05, "loss": 2.8994, "step": 45799 }, { "epoch": 3.59, "learning_rate": 6.408688151807418e-05, "loss": 2.8119, "step": 45800 }, { "epoch": 3.59, "learning_rate": 6.408609738884968e-05, "loss": 2.8855, "step": 45801 }, { "epoch": 3.59, "learning_rate": 6.40853132596252e-05, "loss": 2.4723, "step": 45802 }, { "epoch": 3.59, "learning_rate": 6.40845291304007e-05, "loss": 2.8848, "step": 45803 }, { "epoch": 3.59, "learning_rate": 6.40837450011762e-05, "loss": 2.7775, "step": 45804 }, { "epoch": 3.59, "learning_rate": 6.408296087195171e-05, "loss": 2.7658, "step": 45805 }, { "epoch": 3.59, "learning_rate": 6.408217674272721e-05, "loss": 2.618, "step": 45806 }, { "epoch": 3.59, "learning_rate": 6.408139261350272e-05, "loss": 2.806, "step": 45807 }, { "epoch": 3.59, "learning_rate": 6.408060848427822e-05, "loss": 2.9495, "step": 45808 }, { "epoch": 3.59, "learning_rate": 6.407982435505372e-05, "loss": 2.7546, "step": 45809 }, { "epoch": 3.59, "learning_rate": 6.407904022582923e-05, "loss": 2.8826, "step": 45810 }, { "epoch": 3.59, "learning_rate": 6.407825609660473e-05, "loss": 2.7555, "step": 45811 }, { "epoch": 3.59, "learning_rate": 6.407747196738023e-05, "loss": 2.8231, "step": 45812 }, { "epoch": 3.59, "learning_rate": 6.407668783815574e-05, "loss": 2.7148, "step": 45813 }, { "epoch": 3.59, "learning_rate": 6.407590370893124e-05, "loss": 2.8402, "step": 45814 }, { "epoch": 3.59, "learning_rate": 6.407511957970674e-05, "loss": 2.9451, "step": 45815 }, { "epoch": 3.59, "learning_rate": 6.407433545048225e-05, "loss": 2.8733, "step": 45816 }, { "epoch": 3.59, "learning_rate": 6.407355132125774e-05, "loss": 2.7536, "step": 45817 }, { "epoch": 3.59, "learning_rate": 6.407276719203324e-05, "loss": 2.6681, "step": 45818 }, { "epoch": 3.59, "learning_rate": 6.407198306280875e-05, "loss": 2.8098, "step": 45819 }, { "epoch": 3.59, "learning_rate": 6.407119893358425e-05, "loss": 2.659, "step": 45820 }, { "epoch": 3.59, "learning_rate": 6.407041480435975e-05, "loss": 2.8066, "step": 45821 }, { "epoch": 3.59, "learning_rate": 6.406963067513526e-05, "loss": 2.5902, "step": 45822 }, { "epoch": 3.59, "learning_rate": 6.406884654591076e-05, "loss": 2.7002, "step": 45823 }, { "epoch": 3.59, "learning_rate": 6.406806241668626e-05, "loss": 2.6673, "step": 45824 }, { "epoch": 3.59, "learning_rate": 6.406727828746177e-05, "loss": 2.6864, "step": 45825 }, { "epoch": 3.59, "learning_rate": 6.406649415823727e-05, "loss": 2.678, "step": 45826 }, { "epoch": 3.59, "learning_rate": 6.406571002901278e-05, "loss": 2.698, "step": 45827 }, { "epoch": 3.59, "learning_rate": 6.406492589978829e-05, "loss": 2.6898, "step": 45828 }, { "epoch": 3.59, "learning_rate": 6.40641417705638e-05, "loss": 2.8043, "step": 45829 }, { "epoch": 3.59, "learning_rate": 6.40633576413393e-05, "loss": 2.6862, "step": 45830 }, { "epoch": 3.59, "learning_rate": 6.40625735121148e-05, "loss": 2.5723, "step": 45831 }, { "epoch": 3.59, "learning_rate": 6.406178938289031e-05, "loss": 2.7181, "step": 45832 }, { "epoch": 3.59, "learning_rate": 6.406100525366581e-05, "loss": 2.7768, "step": 45833 }, { "epoch": 3.59, "learning_rate": 6.406022112444131e-05, "loss": 2.6947, "step": 45834 }, { "epoch": 3.59, "learning_rate": 6.405943699521682e-05, "loss": 2.8561, "step": 45835 }, { "epoch": 3.59, "learning_rate": 6.405865286599232e-05, "loss": 2.8225, "step": 45836 }, { "epoch": 3.59, "learning_rate": 6.405786873676783e-05, "loss": 2.6669, "step": 45837 }, { "epoch": 3.59, "learning_rate": 6.405708460754333e-05, "loss": 2.8322, "step": 45838 }, { "epoch": 3.59, "learning_rate": 6.405630047831883e-05, "loss": 2.6166, "step": 45839 }, { "epoch": 3.59, "learning_rate": 6.405551634909434e-05, "loss": 2.7787, "step": 45840 }, { "epoch": 3.59, "learning_rate": 6.405473221986984e-05, "loss": 2.8886, "step": 45841 }, { "epoch": 3.59, "learning_rate": 6.405394809064534e-05, "loss": 2.7605, "step": 45842 }, { "epoch": 3.59, "learning_rate": 6.405316396142085e-05, "loss": 2.6426, "step": 45843 }, { "epoch": 3.59, "learning_rate": 6.405237983219635e-05, "loss": 2.8934, "step": 45844 }, { "epoch": 3.59, "learning_rate": 6.405159570297186e-05, "loss": 2.657, "step": 45845 }, { "epoch": 3.59, "learning_rate": 6.405081157374736e-05, "loss": 2.8157, "step": 45846 }, { "epoch": 3.59, "learning_rate": 6.405002744452286e-05, "loss": 2.7982, "step": 45847 }, { "epoch": 3.6, "learning_rate": 6.404924331529837e-05, "loss": 2.6339, "step": 45848 }, { "epoch": 3.6, "learning_rate": 6.404845918607387e-05, "loss": 2.6121, "step": 45849 }, { "epoch": 3.6, "learning_rate": 6.404767505684937e-05, "loss": 2.8033, "step": 45850 }, { "epoch": 3.6, "learning_rate": 6.404689092762488e-05, "loss": 2.7539, "step": 45851 }, { "epoch": 3.6, "learning_rate": 6.404610679840038e-05, "loss": 2.8335, "step": 45852 }, { "epoch": 3.6, "learning_rate": 6.404532266917588e-05, "loss": 2.7924, "step": 45853 }, { "epoch": 3.6, "learning_rate": 6.404453853995139e-05, "loss": 2.6975, "step": 45854 }, { "epoch": 3.6, "learning_rate": 6.404375441072689e-05, "loss": 2.8678, "step": 45855 }, { "epoch": 3.6, "learning_rate": 6.40429702815024e-05, "loss": 2.6528, "step": 45856 }, { "epoch": 3.6, "learning_rate": 6.40421861522779e-05, "loss": 2.7081, "step": 45857 }, { "epoch": 3.6, "learning_rate": 6.40414020230534e-05, "loss": 2.7648, "step": 45858 }, { "epoch": 3.6, "learning_rate": 6.40406178938289e-05, "loss": 2.6446, "step": 45859 }, { "epoch": 3.6, "learning_rate": 6.403983376460441e-05, "loss": 2.9304, "step": 45860 }, { "epoch": 3.6, "learning_rate": 6.403904963537991e-05, "loss": 2.5423, "step": 45861 }, { "epoch": 3.6, "learning_rate": 6.403826550615542e-05, "loss": 2.6146, "step": 45862 }, { "epoch": 3.6, "learning_rate": 6.403748137693092e-05, "loss": 2.884, "step": 45863 }, { "epoch": 3.6, "learning_rate": 6.403669724770642e-05, "loss": 2.7239, "step": 45864 }, { "epoch": 3.6, "learning_rate": 6.403591311848193e-05, "loss": 2.5452, "step": 45865 }, { "epoch": 3.6, "learning_rate": 6.403512898925743e-05, "loss": 2.9695, "step": 45866 }, { "epoch": 3.6, "learning_rate": 6.403434486003294e-05, "loss": 2.6107, "step": 45867 }, { "epoch": 3.6, "learning_rate": 6.403356073080844e-05, "loss": 2.6618, "step": 45868 }, { "epoch": 3.6, "learning_rate": 6.403277660158394e-05, "loss": 2.7204, "step": 45869 }, { "epoch": 3.6, "learning_rate": 6.403199247235945e-05, "loss": 2.8968, "step": 45870 }, { "epoch": 3.6, "learning_rate": 6.403120834313495e-05, "loss": 2.5689, "step": 45871 }, { "epoch": 3.6, "learning_rate": 6.403042421391045e-05, "loss": 2.7599, "step": 45872 }, { "epoch": 3.6, "learning_rate": 6.402964008468596e-05, "loss": 2.7773, "step": 45873 }, { "epoch": 3.6, "learning_rate": 6.402885595546146e-05, "loss": 2.6379, "step": 45874 }, { "epoch": 3.6, "learning_rate": 6.402807182623697e-05, "loss": 2.7593, "step": 45875 }, { "epoch": 3.6, "learning_rate": 6.402728769701247e-05, "loss": 2.8303, "step": 45876 }, { "epoch": 3.6, "learning_rate": 6.402650356778797e-05, "loss": 2.6543, "step": 45877 }, { "epoch": 3.6, "learning_rate": 6.402571943856348e-05, "loss": 2.7472, "step": 45878 }, { "epoch": 3.6, "learning_rate": 6.402493530933898e-05, "loss": 2.8119, "step": 45879 }, { "epoch": 3.6, "learning_rate": 6.402415118011448e-05, "loss": 2.7153, "step": 45880 }, { "epoch": 3.6, "learning_rate": 6.402336705088999e-05, "loss": 2.9896, "step": 45881 }, { "epoch": 3.6, "learning_rate": 6.402258292166549e-05, "loss": 2.5927, "step": 45882 }, { "epoch": 3.6, "learning_rate": 6.4021798792441e-05, "loss": 2.5663, "step": 45883 }, { "epoch": 3.6, "learning_rate": 6.40210146632165e-05, "loss": 2.8724, "step": 45884 }, { "epoch": 3.6, "learning_rate": 6.402023053399202e-05, "loss": 2.5678, "step": 45885 }, { "epoch": 3.6, "learning_rate": 6.401944640476752e-05, "loss": 2.7744, "step": 45886 }, { "epoch": 3.6, "learning_rate": 6.401866227554302e-05, "loss": 2.7951, "step": 45887 }, { "epoch": 3.6, "learning_rate": 6.401787814631853e-05, "loss": 3.0108, "step": 45888 }, { "epoch": 3.6, "learning_rate": 6.401709401709402e-05, "loss": 2.7713, "step": 45889 }, { "epoch": 3.6, "learning_rate": 6.401630988786952e-05, "loss": 2.706, "step": 45890 }, { "epoch": 3.6, "learning_rate": 6.401552575864502e-05, "loss": 2.5958, "step": 45891 }, { "epoch": 3.6, "learning_rate": 6.401474162942053e-05, "loss": 2.5768, "step": 45892 }, { "epoch": 3.6, "learning_rate": 6.401395750019603e-05, "loss": 2.7478, "step": 45893 }, { "epoch": 3.6, "learning_rate": 6.401317337097154e-05, "loss": 2.8976, "step": 45894 }, { "epoch": 3.6, "learning_rate": 6.401238924174704e-05, "loss": 2.721, "step": 45895 }, { "epoch": 3.6, "learning_rate": 6.401160511252254e-05, "loss": 2.4746, "step": 45896 }, { "epoch": 3.6, "learning_rate": 6.401082098329805e-05, "loss": 2.7158, "step": 45897 }, { "epoch": 3.6, "learning_rate": 6.401003685407355e-05, "loss": 2.7078, "step": 45898 }, { "epoch": 3.6, "learning_rate": 6.400925272484905e-05, "loss": 2.6104, "step": 45899 }, { "epoch": 3.6, "learning_rate": 6.400846859562456e-05, "loss": 2.7142, "step": 45900 }, { "epoch": 3.6, "learning_rate": 6.400768446640006e-05, "loss": 2.796, "step": 45901 }, { "epoch": 3.6, "learning_rate": 6.400690033717556e-05, "loss": 2.5617, "step": 45902 }, { "epoch": 3.6, "learning_rate": 6.400611620795107e-05, "loss": 2.7068, "step": 45903 }, { "epoch": 3.6, "learning_rate": 6.400533207872657e-05, "loss": 2.9967, "step": 45904 }, { "epoch": 3.6, "learning_rate": 6.400454794950208e-05, "loss": 2.541, "step": 45905 }, { "epoch": 3.6, "learning_rate": 6.400376382027758e-05, "loss": 2.8244, "step": 45906 }, { "epoch": 3.6, "learning_rate": 6.400297969105308e-05, "loss": 2.7647, "step": 45907 }, { "epoch": 3.6, "learning_rate": 6.400219556182859e-05, "loss": 2.6708, "step": 45908 }, { "epoch": 3.6, "learning_rate": 6.400141143260409e-05, "loss": 2.6635, "step": 45909 }, { "epoch": 3.6, "learning_rate": 6.400062730337961e-05, "loss": 2.7678, "step": 45910 }, { "epoch": 3.6, "learning_rate": 6.399984317415511e-05, "loss": 2.8683, "step": 45911 }, { "epoch": 3.6, "learning_rate": 6.399905904493062e-05, "loss": 2.688, "step": 45912 }, { "epoch": 3.6, "learning_rate": 6.399827491570612e-05, "loss": 2.6221, "step": 45913 }, { "epoch": 3.6, "learning_rate": 6.399749078648162e-05, "loss": 2.752, "step": 45914 }, { "epoch": 3.6, "learning_rate": 6.399670665725713e-05, "loss": 2.9351, "step": 45915 }, { "epoch": 3.6, "learning_rate": 6.399592252803263e-05, "loss": 2.5998, "step": 45916 }, { "epoch": 3.6, "learning_rate": 6.399513839880813e-05, "loss": 2.8849, "step": 45917 }, { "epoch": 3.6, "learning_rate": 6.399435426958364e-05, "loss": 2.7816, "step": 45918 }, { "epoch": 3.6, "learning_rate": 6.399357014035914e-05, "loss": 2.9579, "step": 45919 }, { "epoch": 3.6, "learning_rate": 6.399278601113464e-05, "loss": 2.7541, "step": 45920 }, { "epoch": 3.6, "learning_rate": 6.399200188191015e-05, "loss": 2.6371, "step": 45921 }, { "epoch": 3.6, "learning_rate": 6.399121775268565e-05, "loss": 2.8186, "step": 45922 }, { "epoch": 3.6, "learning_rate": 6.399043362346116e-05, "loss": 2.8062, "step": 45923 }, { "epoch": 3.6, "learning_rate": 6.398964949423666e-05, "loss": 2.5528, "step": 45924 }, { "epoch": 3.6, "learning_rate": 6.398886536501215e-05, "loss": 2.9933, "step": 45925 }, { "epoch": 3.6, "learning_rate": 6.398808123578765e-05, "loss": 2.9132, "step": 45926 }, { "epoch": 3.6, "learning_rate": 6.398729710656316e-05, "loss": 2.8282, "step": 45927 }, { "epoch": 3.6, "learning_rate": 6.398651297733866e-05, "loss": 2.5804, "step": 45928 }, { "epoch": 3.6, "learning_rate": 6.398572884811416e-05, "loss": 2.5358, "step": 45929 }, { "epoch": 3.6, "learning_rate": 6.398494471888967e-05, "loss": 2.6892, "step": 45930 }, { "epoch": 3.6, "learning_rate": 6.398416058966517e-05, "loss": 2.7534, "step": 45931 }, { "epoch": 3.6, "learning_rate": 6.398337646044068e-05, "loss": 2.7442, "step": 45932 }, { "epoch": 3.6, "learning_rate": 6.398259233121618e-05, "loss": 2.5531, "step": 45933 }, { "epoch": 3.6, "learning_rate": 6.398180820199168e-05, "loss": 2.602, "step": 45934 }, { "epoch": 3.6, "learning_rate": 6.398102407276719e-05, "loss": 2.8846, "step": 45935 }, { "epoch": 3.6, "learning_rate": 6.39802399435427e-05, "loss": 2.6638, "step": 45936 }, { "epoch": 3.6, "learning_rate": 6.397945581431821e-05, "loss": 2.651, "step": 45937 }, { "epoch": 3.6, "learning_rate": 6.397867168509371e-05, "loss": 2.8119, "step": 45938 }, { "epoch": 3.6, "learning_rate": 6.397788755586921e-05, "loss": 2.5589, "step": 45939 }, { "epoch": 3.6, "learning_rate": 6.397710342664472e-05, "loss": 2.6683, "step": 45940 }, { "epoch": 3.6, "learning_rate": 6.397631929742022e-05, "loss": 2.4969, "step": 45941 }, { "epoch": 3.6, "learning_rate": 6.397553516819573e-05, "loss": 2.8544, "step": 45942 }, { "epoch": 3.6, "learning_rate": 6.397475103897123e-05, "loss": 2.6686, "step": 45943 }, { "epoch": 3.6, "learning_rate": 6.397396690974673e-05, "loss": 2.6959, "step": 45944 }, { "epoch": 3.6, "learning_rate": 6.397318278052224e-05, "loss": 2.6775, "step": 45945 }, { "epoch": 3.6, "learning_rate": 6.397239865129774e-05, "loss": 2.873, "step": 45946 }, { "epoch": 3.6, "learning_rate": 6.397161452207324e-05, "loss": 2.5895, "step": 45947 }, { "epoch": 3.6, "learning_rate": 6.397083039284875e-05, "loss": 2.4769, "step": 45948 }, { "epoch": 3.6, "learning_rate": 6.397004626362425e-05, "loss": 2.9287, "step": 45949 }, { "epoch": 3.6, "learning_rate": 6.396926213439975e-05, "loss": 2.8738, "step": 45950 }, { "epoch": 3.6, "learning_rate": 6.396847800517526e-05, "loss": 2.6412, "step": 45951 }, { "epoch": 3.6, "learning_rate": 6.396769387595076e-05, "loss": 2.848, "step": 45952 }, { "epoch": 3.6, "learning_rate": 6.396690974672627e-05, "loss": 2.6521, "step": 45953 }, { "epoch": 3.6, "learning_rate": 6.396612561750177e-05, "loss": 2.7764, "step": 45954 }, { "epoch": 3.6, "learning_rate": 6.396534148827727e-05, "loss": 2.887, "step": 45955 }, { "epoch": 3.6, "learning_rate": 6.396455735905278e-05, "loss": 2.8918, "step": 45956 }, { "epoch": 3.6, "learning_rate": 6.396377322982828e-05, "loss": 2.58, "step": 45957 }, { "epoch": 3.6, "learning_rate": 6.396298910060378e-05, "loss": 2.619, "step": 45958 }, { "epoch": 3.6, "learning_rate": 6.396220497137929e-05, "loss": 2.7884, "step": 45959 }, { "epoch": 3.6, "learning_rate": 6.396142084215479e-05, "loss": 2.838, "step": 45960 }, { "epoch": 3.6, "learning_rate": 6.39606367129303e-05, "loss": 2.6477, "step": 45961 }, { "epoch": 3.6, "learning_rate": 6.39598525837058e-05, "loss": 2.5195, "step": 45962 }, { "epoch": 3.6, "learning_rate": 6.39590684544813e-05, "loss": 2.5593, "step": 45963 }, { "epoch": 3.6, "learning_rate": 6.39582843252568e-05, "loss": 2.6499, "step": 45964 }, { "epoch": 3.6, "learning_rate": 6.395750019603231e-05, "loss": 2.7047, "step": 45965 }, { "epoch": 3.6, "learning_rate": 6.395671606680781e-05, "loss": 2.9259, "step": 45966 }, { "epoch": 3.6, "learning_rate": 6.395593193758332e-05, "loss": 2.7884, "step": 45967 }, { "epoch": 3.6, "learning_rate": 6.395514780835882e-05, "loss": 2.8064, "step": 45968 }, { "epoch": 3.6, "learning_rate": 6.395436367913432e-05, "loss": 2.754, "step": 45969 }, { "epoch": 3.6, "learning_rate": 6.395357954990983e-05, "loss": 2.8143, "step": 45970 }, { "epoch": 3.6, "learning_rate": 6.395279542068533e-05, "loss": 2.7094, "step": 45971 }, { "epoch": 3.6, "learning_rate": 6.395201129146084e-05, "loss": 2.7269, "step": 45972 }, { "epoch": 3.6, "learning_rate": 6.395122716223634e-05, "loss": 2.9303, "step": 45973 }, { "epoch": 3.6, "learning_rate": 6.395044303301184e-05, "loss": 2.6354, "step": 45974 }, { "epoch": 3.61, "learning_rate": 6.394965890378735e-05, "loss": 2.7579, "step": 45975 }, { "epoch": 3.61, "learning_rate": 6.394887477456285e-05, "loss": 2.7546, "step": 45976 }, { "epoch": 3.61, "learning_rate": 6.394809064533835e-05, "loss": 2.7106, "step": 45977 }, { "epoch": 3.61, "learning_rate": 6.394730651611386e-05, "loss": 2.8024, "step": 45978 }, { "epoch": 3.61, "learning_rate": 6.394652238688936e-05, "loss": 2.6779, "step": 45979 }, { "epoch": 3.61, "learning_rate": 6.394573825766487e-05, "loss": 2.6072, "step": 45980 }, { "epoch": 3.61, "learning_rate": 6.394495412844037e-05, "loss": 2.5528, "step": 45981 }, { "epoch": 3.61, "learning_rate": 6.394416999921587e-05, "loss": 2.7896, "step": 45982 }, { "epoch": 3.61, "learning_rate": 6.394338586999138e-05, "loss": 2.5866, "step": 45983 }, { "epoch": 3.61, "learning_rate": 6.394260174076688e-05, "loss": 2.6504, "step": 45984 }, { "epoch": 3.61, "learning_rate": 6.394181761154238e-05, "loss": 2.7424, "step": 45985 }, { "epoch": 3.61, "learning_rate": 6.394103348231789e-05, "loss": 2.7342, "step": 45986 }, { "epoch": 3.61, "learning_rate": 6.394024935309339e-05, "loss": 2.9249, "step": 45987 }, { "epoch": 3.61, "learning_rate": 6.39394652238689e-05, "loss": 2.792, "step": 45988 }, { "epoch": 3.61, "learning_rate": 6.39386810946444e-05, "loss": 2.629, "step": 45989 }, { "epoch": 3.61, "learning_rate": 6.39378969654199e-05, "loss": 2.9538, "step": 45990 }, { "epoch": 3.61, "learning_rate": 6.39371128361954e-05, "loss": 2.5956, "step": 45991 }, { "epoch": 3.61, "learning_rate": 6.393632870697091e-05, "loss": 2.8648, "step": 45992 }, { "epoch": 3.61, "learning_rate": 6.393554457774643e-05, "loss": 2.6902, "step": 45993 }, { "epoch": 3.61, "learning_rate": 6.393476044852193e-05, "loss": 2.9105, "step": 45994 }, { "epoch": 3.61, "learning_rate": 6.393397631929743e-05, "loss": 2.8215, "step": 45995 }, { "epoch": 3.61, "learning_rate": 6.393319219007294e-05, "loss": 2.713, "step": 45996 }, { "epoch": 3.61, "learning_rate": 6.393240806084843e-05, "loss": 2.6574, "step": 45997 }, { "epoch": 3.61, "learning_rate": 6.393162393162393e-05, "loss": 2.7591, "step": 45998 }, { "epoch": 3.61, "learning_rate": 6.393083980239944e-05, "loss": 2.7976, "step": 45999 }, { "epoch": 3.61, "learning_rate": 6.393005567317494e-05, "loss": 2.844, "step": 46000 }, { "epoch": 3.61, "eval_loss": 2.8943471908569336, "eval_runtime": 217.8393, "eval_samples_per_second": 52.02, "eval_steps_per_second": 3.255, "step": 46000 }, { "epoch": 3.61, "learning_rate": 6.392927154395044e-05, "loss": 2.708, "step": 46001 }, { "epoch": 3.61, "learning_rate": 6.392848741472595e-05, "loss": 2.6538, "step": 46002 }, { "epoch": 3.61, "learning_rate": 6.392770328550145e-05, "loss": 2.8091, "step": 46003 }, { "epoch": 3.61, "learning_rate": 6.392691915627695e-05, "loss": 2.6425, "step": 46004 }, { "epoch": 3.61, "learning_rate": 6.392613502705246e-05, "loss": 2.8597, "step": 46005 }, { "epoch": 3.61, "learning_rate": 6.392535089782796e-05, "loss": 2.7738, "step": 46006 }, { "epoch": 3.61, "learning_rate": 6.392456676860346e-05, "loss": 2.7908, "step": 46007 }, { "epoch": 3.61, "learning_rate": 6.392378263937897e-05, "loss": 2.6976, "step": 46008 }, { "epoch": 3.61, "learning_rate": 6.392299851015447e-05, "loss": 2.8284, "step": 46009 }, { "epoch": 3.61, "learning_rate": 6.392221438092998e-05, "loss": 2.8113, "step": 46010 }, { "epoch": 3.61, "learning_rate": 6.392143025170548e-05, "loss": 2.5744, "step": 46011 }, { "epoch": 3.61, "learning_rate": 6.392064612248098e-05, "loss": 2.6165, "step": 46012 }, { "epoch": 3.61, "learning_rate": 6.391986199325649e-05, "loss": 2.6236, "step": 46013 }, { "epoch": 3.61, "learning_rate": 6.391907786403199e-05, "loss": 2.9338, "step": 46014 }, { "epoch": 3.61, "learning_rate": 6.39182937348075e-05, "loss": 2.6761, "step": 46015 }, { "epoch": 3.61, "learning_rate": 6.3917509605583e-05, "loss": 2.7919, "step": 46016 }, { "epoch": 3.61, "learning_rate": 6.39167254763585e-05, "loss": 2.7864, "step": 46017 }, { "epoch": 3.61, "learning_rate": 6.391594134713402e-05, "loss": 2.4114, "step": 46018 }, { "epoch": 3.61, "learning_rate": 6.391515721790952e-05, "loss": 2.7679, "step": 46019 }, { "epoch": 3.61, "learning_rate": 6.391437308868503e-05, "loss": 2.8049, "step": 46020 }, { "epoch": 3.61, "learning_rate": 6.391358895946053e-05, "loss": 2.6851, "step": 46021 }, { "epoch": 3.61, "learning_rate": 6.391280483023603e-05, "loss": 2.5551, "step": 46022 }, { "epoch": 3.61, "learning_rate": 6.391202070101154e-05, "loss": 2.5768, "step": 46023 }, { "epoch": 3.61, "learning_rate": 6.391123657178704e-05, "loss": 2.9243, "step": 46024 }, { "epoch": 3.61, "learning_rate": 6.391045244256254e-05, "loss": 2.6907, "step": 46025 }, { "epoch": 3.61, "learning_rate": 6.390966831333805e-05, "loss": 2.6468, "step": 46026 }, { "epoch": 3.61, "learning_rate": 6.390888418411355e-05, "loss": 2.7505, "step": 46027 }, { "epoch": 3.61, "learning_rate": 6.390810005488906e-05, "loss": 2.7818, "step": 46028 }, { "epoch": 3.61, "learning_rate": 6.390731592566456e-05, "loss": 2.7082, "step": 46029 }, { "epoch": 3.61, "learning_rate": 6.390653179644006e-05, "loss": 2.8369, "step": 46030 }, { "epoch": 3.61, "learning_rate": 6.390574766721557e-05, "loss": 2.7362, "step": 46031 }, { "epoch": 3.61, "learning_rate": 6.390496353799107e-05, "loss": 2.673, "step": 46032 }, { "epoch": 3.61, "learning_rate": 6.390417940876656e-05, "loss": 2.553, "step": 46033 }, { "epoch": 3.61, "learning_rate": 6.390339527954206e-05, "loss": 2.8754, "step": 46034 }, { "epoch": 3.61, "learning_rate": 6.390261115031757e-05, "loss": 3.0006, "step": 46035 }, { "epoch": 3.61, "learning_rate": 6.390182702109307e-05, "loss": 2.6411, "step": 46036 }, { "epoch": 3.61, "learning_rate": 6.390104289186857e-05, "loss": 2.8407, "step": 46037 }, { "epoch": 3.61, "learning_rate": 6.390025876264408e-05, "loss": 2.9091, "step": 46038 }, { "epoch": 3.61, "learning_rate": 6.389947463341958e-05, "loss": 2.6438, "step": 46039 }, { "epoch": 3.61, "learning_rate": 6.389869050419509e-05, "loss": 3.0216, "step": 46040 }, { "epoch": 3.61, "learning_rate": 6.389790637497059e-05, "loss": 2.6026, "step": 46041 }, { "epoch": 3.61, "learning_rate": 6.389712224574609e-05, "loss": 2.6749, "step": 46042 }, { "epoch": 3.61, "learning_rate": 6.38963381165216e-05, "loss": 2.6577, "step": 46043 }, { "epoch": 3.61, "learning_rate": 6.389555398729711e-05, "loss": 2.6131, "step": 46044 }, { "epoch": 3.61, "learning_rate": 6.389476985807262e-05, "loss": 2.9228, "step": 46045 }, { "epoch": 3.61, "learning_rate": 6.389398572884812e-05, "loss": 2.9004, "step": 46046 }, { "epoch": 3.61, "learning_rate": 6.389320159962363e-05, "loss": 2.5966, "step": 46047 }, { "epoch": 3.61, "learning_rate": 6.389241747039913e-05, "loss": 2.6736, "step": 46048 }, { "epoch": 3.61, "learning_rate": 6.389163334117463e-05, "loss": 2.6874, "step": 46049 }, { "epoch": 3.61, "learning_rate": 6.389084921195014e-05, "loss": 2.791, "step": 46050 }, { "epoch": 3.61, "learning_rate": 6.389006508272564e-05, "loss": 2.8086, "step": 46051 }, { "epoch": 3.61, "learning_rate": 6.388928095350114e-05, "loss": 2.6178, "step": 46052 }, { "epoch": 3.61, "learning_rate": 6.388849682427665e-05, "loss": 2.6583, "step": 46053 }, { "epoch": 3.61, "learning_rate": 6.388771269505215e-05, "loss": 3.0574, "step": 46054 }, { "epoch": 3.61, "learning_rate": 6.388692856582765e-05, "loss": 2.7781, "step": 46055 }, { "epoch": 3.61, "learning_rate": 6.388614443660316e-05, "loss": 2.6368, "step": 46056 }, { "epoch": 3.61, "learning_rate": 6.388536030737866e-05, "loss": 2.6308, "step": 46057 }, { "epoch": 3.61, "learning_rate": 6.388457617815417e-05, "loss": 2.7848, "step": 46058 }, { "epoch": 3.61, "learning_rate": 6.388379204892967e-05, "loss": 2.6537, "step": 46059 }, { "epoch": 3.61, "learning_rate": 6.388300791970517e-05, "loss": 2.6825, "step": 46060 }, { "epoch": 3.61, "learning_rate": 6.388222379048068e-05, "loss": 2.5523, "step": 46061 }, { "epoch": 3.61, "learning_rate": 6.388143966125618e-05, "loss": 2.79, "step": 46062 }, { "epoch": 3.61, "learning_rate": 6.388065553203168e-05, "loss": 2.8244, "step": 46063 }, { "epoch": 3.61, "learning_rate": 6.387987140280719e-05, "loss": 2.721, "step": 46064 }, { "epoch": 3.61, "learning_rate": 6.387908727358269e-05, "loss": 2.7771, "step": 46065 }, { "epoch": 3.61, "learning_rate": 6.38783031443582e-05, "loss": 2.6102, "step": 46066 }, { "epoch": 3.61, "learning_rate": 6.38775190151337e-05, "loss": 2.7795, "step": 46067 }, { "epoch": 3.61, "learning_rate": 6.38767348859092e-05, "loss": 2.6438, "step": 46068 }, { "epoch": 3.61, "learning_rate": 6.38759507566847e-05, "loss": 2.6611, "step": 46069 }, { "epoch": 3.61, "learning_rate": 6.387516662746021e-05, "loss": 2.7498, "step": 46070 }, { "epoch": 3.61, "learning_rate": 6.387438249823571e-05, "loss": 2.8126, "step": 46071 }, { "epoch": 3.61, "learning_rate": 6.387359836901122e-05, "loss": 2.8815, "step": 46072 }, { "epoch": 3.61, "learning_rate": 6.387281423978672e-05, "loss": 2.7402, "step": 46073 }, { "epoch": 3.61, "learning_rate": 6.387203011056222e-05, "loss": 2.7017, "step": 46074 }, { "epoch": 3.61, "learning_rate": 6.387124598133773e-05, "loss": 2.6627, "step": 46075 }, { "epoch": 3.61, "learning_rate": 6.387046185211323e-05, "loss": 2.5117, "step": 46076 }, { "epoch": 3.61, "learning_rate": 6.386967772288874e-05, "loss": 2.5912, "step": 46077 }, { "epoch": 3.61, "learning_rate": 6.386889359366424e-05, "loss": 2.7357, "step": 46078 }, { "epoch": 3.61, "learning_rate": 6.386810946443974e-05, "loss": 2.7021, "step": 46079 }, { "epoch": 3.61, "learning_rate": 6.386732533521525e-05, "loss": 2.6766, "step": 46080 }, { "epoch": 3.61, "learning_rate": 6.386654120599075e-05, "loss": 2.9351, "step": 46081 }, { "epoch": 3.61, "learning_rate": 6.386575707676625e-05, "loss": 2.6739, "step": 46082 }, { "epoch": 3.61, "learning_rate": 6.386497294754176e-05, "loss": 2.7724, "step": 46083 }, { "epoch": 3.61, "learning_rate": 6.386418881831726e-05, "loss": 2.8087, "step": 46084 }, { "epoch": 3.61, "learning_rate": 6.386340468909276e-05, "loss": 3.0138, "step": 46085 }, { "epoch": 3.61, "learning_rate": 6.386262055986827e-05, "loss": 2.7634, "step": 46086 }, { "epoch": 3.61, "learning_rate": 6.386183643064377e-05, "loss": 2.7193, "step": 46087 }, { "epoch": 3.61, "learning_rate": 6.386105230141928e-05, "loss": 2.8735, "step": 46088 }, { "epoch": 3.61, "learning_rate": 6.386026817219478e-05, "loss": 2.631, "step": 46089 }, { "epoch": 3.61, "learning_rate": 6.385948404297028e-05, "loss": 2.4629, "step": 46090 }, { "epoch": 3.61, "learning_rate": 6.385869991374579e-05, "loss": 2.9722, "step": 46091 }, { "epoch": 3.61, "learning_rate": 6.385791578452129e-05, "loss": 2.9084, "step": 46092 }, { "epoch": 3.61, "learning_rate": 6.38571316552968e-05, "loss": 2.6269, "step": 46093 }, { "epoch": 3.61, "learning_rate": 6.38563475260723e-05, "loss": 2.8502, "step": 46094 }, { "epoch": 3.61, "learning_rate": 6.38555633968478e-05, "loss": 2.8645, "step": 46095 }, { "epoch": 3.61, "learning_rate": 6.38547792676233e-05, "loss": 2.5513, "step": 46096 }, { "epoch": 3.61, "learning_rate": 6.385399513839881e-05, "loss": 2.7998, "step": 46097 }, { "epoch": 3.61, "learning_rate": 6.385321100917431e-05, "loss": 2.7949, "step": 46098 }, { "epoch": 3.61, "learning_rate": 6.385242687994982e-05, "loss": 2.8043, "step": 46099 }, { "epoch": 3.61, "learning_rate": 6.385164275072532e-05, "loss": 2.6663, "step": 46100 }, { "epoch": 3.61, "learning_rate": 6.385085862150084e-05, "loss": 3.0493, "step": 46101 }, { "epoch": 3.61, "learning_rate": 6.385007449227634e-05, "loss": 2.5612, "step": 46102 }, { "epoch": 3.62, "learning_rate": 6.384929036305184e-05, "loss": 2.7534, "step": 46103 }, { "epoch": 3.62, "learning_rate": 6.384850623382735e-05, "loss": 2.6983, "step": 46104 }, { "epoch": 3.62, "learning_rate": 6.384772210460284e-05, "loss": 2.7288, "step": 46105 }, { "epoch": 3.62, "learning_rate": 6.384693797537834e-05, "loss": 2.6675, "step": 46106 }, { "epoch": 3.62, "learning_rate": 6.384615384615385e-05, "loss": 2.6692, "step": 46107 }, { "epoch": 3.62, "learning_rate": 6.384536971692935e-05, "loss": 2.8015, "step": 46108 }, { "epoch": 3.62, "learning_rate": 6.384458558770485e-05, "loss": 2.6836, "step": 46109 }, { "epoch": 3.62, "learning_rate": 6.384380145848036e-05, "loss": 2.7293, "step": 46110 }, { "epoch": 3.62, "learning_rate": 6.384301732925586e-05, "loss": 2.8944, "step": 46111 }, { "epoch": 3.62, "learning_rate": 6.384223320003136e-05, "loss": 2.7941, "step": 46112 }, { "epoch": 3.62, "learning_rate": 6.384144907080687e-05, "loss": 2.9675, "step": 46113 }, { "epoch": 3.62, "learning_rate": 6.384066494158237e-05, "loss": 2.8073, "step": 46114 }, { "epoch": 3.62, "learning_rate": 6.383988081235788e-05, "loss": 2.659, "step": 46115 }, { "epoch": 3.62, "learning_rate": 6.383909668313338e-05, "loss": 2.8835, "step": 46116 }, { "epoch": 3.62, "learning_rate": 6.383831255390888e-05, "loss": 2.6658, "step": 46117 }, { "epoch": 3.62, "learning_rate": 6.383752842468439e-05, "loss": 2.75, "step": 46118 }, { "epoch": 3.62, "learning_rate": 6.383674429545989e-05, "loss": 2.6103, "step": 46119 }, { "epoch": 3.62, "learning_rate": 6.38359601662354e-05, "loss": 3.0669, "step": 46120 }, { "epoch": 3.62, "learning_rate": 6.38351760370109e-05, "loss": 2.8029, "step": 46121 }, { "epoch": 3.62, "learning_rate": 6.38343919077864e-05, "loss": 2.6864, "step": 46122 }, { "epoch": 3.62, "learning_rate": 6.38336077785619e-05, "loss": 2.824, "step": 46123 }, { "epoch": 3.62, "learning_rate": 6.383282364933741e-05, "loss": 2.6165, "step": 46124 }, { "epoch": 3.62, "learning_rate": 6.383203952011291e-05, "loss": 2.6156, "step": 46125 }, { "epoch": 3.62, "learning_rate": 6.383125539088843e-05, "loss": 2.4797, "step": 46126 }, { "epoch": 3.62, "learning_rate": 6.383047126166393e-05, "loss": 2.7385, "step": 46127 }, { "epoch": 3.62, "learning_rate": 6.382968713243944e-05, "loss": 2.8606, "step": 46128 }, { "epoch": 3.62, "learning_rate": 6.382890300321494e-05, "loss": 2.6045, "step": 46129 }, { "epoch": 3.62, "learning_rate": 6.382811887399044e-05, "loss": 2.684, "step": 46130 }, { "epoch": 3.62, "learning_rate": 6.382733474476595e-05, "loss": 2.5988, "step": 46131 }, { "epoch": 3.62, "learning_rate": 6.382655061554145e-05, "loss": 2.7499, "step": 46132 }, { "epoch": 3.62, "learning_rate": 6.382576648631695e-05, "loss": 2.8188, "step": 46133 }, { "epoch": 3.62, "learning_rate": 6.382498235709246e-05, "loss": 2.8119, "step": 46134 }, { "epoch": 3.62, "learning_rate": 6.382419822786796e-05, "loss": 2.7322, "step": 46135 }, { "epoch": 3.62, "learning_rate": 6.382341409864347e-05, "loss": 2.6457, "step": 46136 }, { "epoch": 3.62, "learning_rate": 6.382262996941897e-05, "loss": 2.7825, "step": 46137 }, { "epoch": 3.62, "learning_rate": 6.382184584019447e-05, "loss": 2.9066, "step": 46138 }, { "epoch": 3.62, "learning_rate": 6.382106171096998e-05, "loss": 2.724, "step": 46139 }, { "epoch": 3.62, "learning_rate": 6.382027758174548e-05, "loss": 2.818, "step": 46140 }, { "epoch": 3.62, "learning_rate": 6.381949345252097e-05, "loss": 2.7644, "step": 46141 }, { "epoch": 3.62, "learning_rate": 6.381870932329647e-05, "loss": 2.7586, "step": 46142 }, { "epoch": 3.62, "learning_rate": 6.381792519407198e-05, "loss": 2.9541, "step": 46143 }, { "epoch": 3.62, "learning_rate": 6.381714106484748e-05, "loss": 2.4551, "step": 46144 }, { "epoch": 3.62, "learning_rate": 6.381635693562299e-05, "loss": 2.6263, "step": 46145 }, { "epoch": 3.62, "learning_rate": 6.381557280639849e-05, "loss": 2.7817, "step": 46146 }, { "epoch": 3.62, "learning_rate": 6.381478867717399e-05, "loss": 2.6945, "step": 46147 }, { "epoch": 3.62, "learning_rate": 6.38140045479495e-05, "loss": 2.4429, "step": 46148 }, { "epoch": 3.62, "learning_rate": 6.3813220418725e-05, "loss": 2.7634, "step": 46149 }, { "epoch": 3.62, "learning_rate": 6.38124362895005e-05, "loss": 2.7206, "step": 46150 }, { "epoch": 3.62, "learning_rate": 6.381165216027601e-05, "loss": 2.5364, "step": 46151 }, { "epoch": 3.62, "learning_rate": 6.381086803105152e-05, "loss": 2.6173, "step": 46152 }, { "epoch": 3.62, "learning_rate": 6.381008390182703e-05, "loss": 2.8793, "step": 46153 }, { "epoch": 3.62, "learning_rate": 6.380929977260253e-05, "loss": 2.5417, "step": 46154 }, { "epoch": 3.62, "learning_rate": 6.380851564337804e-05, "loss": 2.8142, "step": 46155 }, { "epoch": 3.62, "learning_rate": 6.380773151415354e-05, "loss": 2.6927, "step": 46156 }, { "epoch": 3.62, "learning_rate": 6.380694738492904e-05, "loss": 2.8423, "step": 46157 }, { "epoch": 3.62, "learning_rate": 6.380616325570455e-05, "loss": 2.9147, "step": 46158 }, { "epoch": 3.62, "learning_rate": 6.380537912648005e-05, "loss": 2.601, "step": 46159 }, { "epoch": 3.62, "learning_rate": 6.380459499725555e-05, "loss": 2.6057, "step": 46160 }, { "epoch": 3.62, "learning_rate": 6.380381086803106e-05, "loss": 2.7886, "step": 46161 }, { "epoch": 3.62, "learning_rate": 6.380302673880656e-05, "loss": 2.7172, "step": 46162 }, { "epoch": 3.62, "learning_rate": 6.380224260958207e-05, "loss": 2.7794, "step": 46163 }, { "epoch": 3.62, "learning_rate": 6.380145848035757e-05, "loss": 2.623, "step": 46164 }, { "epoch": 3.62, "learning_rate": 6.380067435113307e-05, "loss": 2.8027, "step": 46165 }, { "epoch": 3.62, "learning_rate": 6.379989022190858e-05, "loss": 2.6422, "step": 46166 }, { "epoch": 3.62, "learning_rate": 6.379910609268408e-05, "loss": 2.7958, "step": 46167 }, { "epoch": 3.62, "learning_rate": 6.379832196345958e-05, "loss": 2.6077, "step": 46168 }, { "epoch": 3.62, "learning_rate": 6.379753783423509e-05, "loss": 2.7056, "step": 46169 }, { "epoch": 3.62, "learning_rate": 6.379675370501059e-05, "loss": 2.7301, "step": 46170 }, { "epoch": 3.62, "learning_rate": 6.37959695757861e-05, "loss": 2.5836, "step": 46171 }, { "epoch": 3.62, "learning_rate": 6.37951854465616e-05, "loss": 3.0116, "step": 46172 }, { "epoch": 3.62, "learning_rate": 6.37944013173371e-05, "loss": 2.678, "step": 46173 }, { "epoch": 3.62, "learning_rate": 6.37936171881126e-05, "loss": 2.8313, "step": 46174 }, { "epoch": 3.62, "learning_rate": 6.379283305888811e-05, "loss": 2.7311, "step": 46175 }, { "epoch": 3.62, "learning_rate": 6.379204892966361e-05, "loss": 2.8568, "step": 46176 }, { "epoch": 3.62, "learning_rate": 6.379126480043912e-05, "loss": 3.0172, "step": 46177 }, { "epoch": 3.62, "learning_rate": 6.379048067121462e-05, "loss": 2.5204, "step": 46178 }, { "epoch": 3.62, "learning_rate": 6.378969654199012e-05, "loss": 2.8924, "step": 46179 }, { "epoch": 3.62, "learning_rate": 6.378891241276563e-05, "loss": 2.7055, "step": 46180 }, { "epoch": 3.62, "learning_rate": 6.378812828354113e-05, "loss": 2.8036, "step": 46181 }, { "epoch": 3.62, "learning_rate": 6.378734415431664e-05, "loss": 2.5878, "step": 46182 }, { "epoch": 3.62, "learning_rate": 6.378656002509214e-05, "loss": 2.666, "step": 46183 }, { "epoch": 3.62, "learning_rate": 6.378577589586764e-05, "loss": 2.7275, "step": 46184 }, { "epoch": 3.62, "learning_rate": 6.378499176664315e-05, "loss": 2.739, "step": 46185 }, { "epoch": 3.62, "learning_rate": 6.378420763741865e-05, "loss": 2.6324, "step": 46186 }, { "epoch": 3.62, "learning_rate": 6.378342350819415e-05, "loss": 2.7733, "step": 46187 }, { "epoch": 3.62, "learning_rate": 6.378263937896966e-05, "loss": 2.8701, "step": 46188 }, { "epoch": 3.62, "learning_rate": 6.378185524974516e-05, "loss": 2.7549, "step": 46189 }, { "epoch": 3.62, "learning_rate": 6.378107112052066e-05, "loss": 2.7336, "step": 46190 }, { "epoch": 3.62, "learning_rate": 6.378028699129617e-05, "loss": 2.5437, "step": 46191 }, { "epoch": 3.62, "learning_rate": 6.377950286207167e-05, "loss": 2.7809, "step": 46192 }, { "epoch": 3.62, "learning_rate": 6.377871873284718e-05, "loss": 2.6694, "step": 46193 }, { "epoch": 3.62, "learning_rate": 6.377793460362268e-05, "loss": 2.7364, "step": 46194 }, { "epoch": 3.62, "learning_rate": 6.377715047439818e-05, "loss": 2.5549, "step": 46195 }, { "epoch": 3.62, "learning_rate": 6.377636634517369e-05, "loss": 2.7943, "step": 46196 }, { "epoch": 3.62, "learning_rate": 6.377558221594919e-05, "loss": 2.7013, "step": 46197 }, { "epoch": 3.62, "learning_rate": 6.37747980867247e-05, "loss": 2.5721, "step": 46198 }, { "epoch": 3.62, "learning_rate": 6.37740139575002e-05, "loss": 2.7807, "step": 46199 }, { "epoch": 3.62, "learning_rate": 6.37732298282757e-05, "loss": 2.8045, "step": 46200 }, { "epoch": 3.62, "learning_rate": 6.37724456990512e-05, "loss": 2.6016, "step": 46201 }, { "epoch": 3.62, "learning_rate": 6.377166156982671e-05, "loss": 2.6224, "step": 46202 }, { "epoch": 3.62, "learning_rate": 6.377087744060221e-05, "loss": 2.7113, "step": 46203 }, { "epoch": 3.62, "learning_rate": 6.377009331137772e-05, "loss": 2.4925, "step": 46204 }, { "epoch": 3.62, "learning_rate": 6.376930918215322e-05, "loss": 2.803, "step": 46205 }, { "epoch": 3.62, "learning_rate": 6.376852505292872e-05, "loss": 2.7165, "step": 46206 }, { "epoch": 3.62, "learning_rate": 6.376774092370423e-05, "loss": 2.796, "step": 46207 }, { "epoch": 3.62, "learning_rate": 6.376695679447973e-05, "loss": 2.7769, "step": 46208 }, { "epoch": 3.62, "learning_rate": 6.376617266525525e-05, "loss": 2.9088, "step": 46209 }, { "epoch": 3.62, "learning_rate": 6.376538853603075e-05, "loss": 2.7196, "step": 46210 }, { "epoch": 3.62, "learning_rate": 6.376460440680626e-05, "loss": 2.7689, "step": 46211 }, { "epoch": 3.62, "learning_rate": 6.376382027758176e-05, "loss": 2.4242, "step": 46212 }, { "epoch": 3.62, "learning_rate": 6.376303614835725e-05, "loss": 2.829, "step": 46213 }, { "epoch": 3.62, "learning_rate": 6.376225201913275e-05, "loss": 2.7478, "step": 46214 }, { "epoch": 3.62, "learning_rate": 6.376146788990826e-05, "loss": 2.5342, "step": 46215 }, { "epoch": 3.62, "learning_rate": 6.376068376068376e-05, "loss": 2.6653, "step": 46216 }, { "epoch": 3.62, "learning_rate": 6.375989963145926e-05, "loss": 2.716, "step": 46217 }, { "epoch": 3.62, "learning_rate": 6.375911550223477e-05, "loss": 2.7447, "step": 46218 }, { "epoch": 3.62, "learning_rate": 6.375833137301027e-05, "loss": 2.7538, "step": 46219 }, { "epoch": 3.62, "learning_rate": 6.375754724378577e-05, "loss": 2.766, "step": 46220 }, { "epoch": 3.62, "learning_rate": 6.375676311456128e-05, "loss": 2.9265, "step": 46221 }, { "epoch": 3.62, "learning_rate": 6.375597898533678e-05, "loss": 2.884, "step": 46222 }, { "epoch": 3.62, "learning_rate": 6.375519485611229e-05, "loss": 2.689, "step": 46223 }, { "epoch": 3.62, "learning_rate": 6.375441072688779e-05, "loss": 2.5627, "step": 46224 }, { "epoch": 3.62, "learning_rate": 6.37536265976633e-05, "loss": 2.8267, "step": 46225 }, { "epoch": 3.62, "learning_rate": 6.37528424684388e-05, "loss": 2.6558, "step": 46226 }, { "epoch": 3.62, "learning_rate": 6.37520583392143e-05, "loss": 2.7233, "step": 46227 }, { "epoch": 3.62, "learning_rate": 6.37512742099898e-05, "loss": 2.5696, "step": 46228 }, { "epoch": 3.62, "learning_rate": 6.375049008076531e-05, "loss": 2.7484, "step": 46229 }, { "epoch": 3.63, "learning_rate": 6.374970595154081e-05, "loss": 2.6494, "step": 46230 }, { "epoch": 3.63, "learning_rate": 6.374892182231632e-05, "loss": 2.5331, "step": 46231 }, { "epoch": 3.63, "learning_rate": 6.374813769309182e-05, "loss": 2.6835, "step": 46232 }, { "epoch": 3.63, "learning_rate": 6.374735356386732e-05, "loss": 2.7384, "step": 46233 }, { "epoch": 3.63, "learning_rate": 6.374656943464284e-05, "loss": 2.7216, "step": 46234 }, { "epoch": 3.63, "learning_rate": 6.374578530541834e-05, "loss": 2.6663, "step": 46235 }, { "epoch": 3.63, "learning_rate": 6.374500117619385e-05, "loss": 2.7038, "step": 46236 }, { "epoch": 3.63, "learning_rate": 6.374421704696935e-05, "loss": 2.781, "step": 46237 }, { "epoch": 3.63, "learning_rate": 6.374343291774485e-05, "loss": 2.6913, "step": 46238 }, { "epoch": 3.63, "learning_rate": 6.374264878852036e-05, "loss": 2.8777, "step": 46239 }, { "epoch": 3.63, "learning_rate": 6.374186465929586e-05, "loss": 2.611, "step": 46240 }, { "epoch": 3.63, "learning_rate": 6.374108053007137e-05, "loss": 2.6732, "step": 46241 }, { "epoch": 3.63, "learning_rate": 6.374029640084687e-05, "loss": 2.6795, "step": 46242 }, { "epoch": 3.63, "learning_rate": 6.373951227162237e-05, "loss": 2.961, "step": 46243 }, { "epoch": 3.63, "learning_rate": 6.373872814239788e-05, "loss": 2.913, "step": 46244 }, { "epoch": 3.63, "learning_rate": 6.373794401317338e-05, "loss": 2.7381, "step": 46245 }, { "epoch": 3.63, "learning_rate": 6.373715988394888e-05, "loss": 2.6808, "step": 46246 }, { "epoch": 3.63, "learning_rate": 6.373637575472439e-05, "loss": 2.8197, "step": 46247 }, { "epoch": 3.63, "learning_rate": 6.373559162549989e-05, "loss": 2.8905, "step": 46248 }, { "epoch": 3.63, "learning_rate": 6.373480749627538e-05, "loss": 2.7171, "step": 46249 }, { "epoch": 3.63, "learning_rate": 6.373402336705089e-05, "loss": 2.6939, "step": 46250 }, { "epoch": 3.63, "learning_rate": 6.373323923782639e-05, "loss": 2.6671, "step": 46251 }, { "epoch": 3.63, "learning_rate": 6.373245510860189e-05, "loss": 2.8215, "step": 46252 }, { "epoch": 3.63, "learning_rate": 6.37316709793774e-05, "loss": 2.5914, "step": 46253 }, { "epoch": 3.63, "learning_rate": 6.37308868501529e-05, "loss": 2.7915, "step": 46254 }, { "epoch": 3.63, "learning_rate": 6.37301027209284e-05, "loss": 2.8434, "step": 46255 }, { "epoch": 3.63, "learning_rate": 6.372931859170391e-05, "loss": 2.6713, "step": 46256 }, { "epoch": 3.63, "learning_rate": 6.372853446247941e-05, "loss": 2.6954, "step": 46257 }, { "epoch": 3.63, "learning_rate": 6.372775033325491e-05, "loss": 2.6861, "step": 46258 }, { "epoch": 3.63, "learning_rate": 6.372696620403042e-05, "loss": 2.6238, "step": 46259 }, { "epoch": 3.63, "learning_rate": 6.372618207480594e-05, "loss": 2.5434, "step": 46260 }, { "epoch": 3.63, "learning_rate": 6.372539794558144e-05, "loss": 2.5892, "step": 46261 }, { "epoch": 3.63, "learning_rate": 6.372461381635694e-05, "loss": 2.6244, "step": 46262 }, { "epoch": 3.63, "learning_rate": 6.372382968713245e-05, "loss": 2.6511, "step": 46263 }, { "epoch": 3.63, "learning_rate": 6.372304555790795e-05, "loss": 2.6431, "step": 46264 }, { "epoch": 3.63, "learning_rate": 6.372226142868345e-05, "loss": 2.8115, "step": 46265 }, { "epoch": 3.63, "learning_rate": 6.372147729945896e-05, "loss": 2.7587, "step": 46266 }, { "epoch": 3.63, "learning_rate": 6.372069317023446e-05, "loss": 2.7188, "step": 46267 }, { "epoch": 3.63, "learning_rate": 6.371990904100996e-05, "loss": 2.8248, "step": 46268 }, { "epoch": 3.63, "learning_rate": 6.371912491178547e-05, "loss": 2.6478, "step": 46269 }, { "epoch": 3.63, "learning_rate": 6.371834078256097e-05, "loss": 2.8165, "step": 46270 }, { "epoch": 3.63, "learning_rate": 6.371755665333648e-05, "loss": 2.646, "step": 46271 }, { "epoch": 3.63, "learning_rate": 6.371677252411198e-05, "loss": 2.9964, "step": 46272 }, { "epoch": 3.63, "learning_rate": 6.371598839488748e-05, "loss": 2.6464, "step": 46273 }, { "epoch": 3.63, "learning_rate": 6.371520426566299e-05, "loss": 2.8575, "step": 46274 }, { "epoch": 3.63, "learning_rate": 6.371442013643849e-05, "loss": 2.7876, "step": 46275 }, { "epoch": 3.63, "learning_rate": 6.3713636007214e-05, "loss": 2.8648, "step": 46276 }, { "epoch": 3.63, "learning_rate": 6.37128518779895e-05, "loss": 2.8768, "step": 46277 }, { "epoch": 3.63, "learning_rate": 6.3712067748765e-05, "loss": 2.5098, "step": 46278 }, { "epoch": 3.63, "learning_rate": 6.37112836195405e-05, "loss": 2.6546, "step": 46279 }, { "epoch": 3.63, "learning_rate": 6.371049949031601e-05, "loss": 2.7952, "step": 46280 }, { "epoch": 3.63, "learning_rate": 6.370971536109151e-05, "loss": 2.8008, "step": 46281 }, { "epoch": 3.63, "learning_rate": 6.370893123186702e-05, "loss": 2.6212, "step": 46282 }, { "epoch": 3.63, "learning_rate": 6.370814710264252e-05, "loss": 2.741, "step": 46283 }, { "epoch": 3.63, "learning_rate": 6.370736297341802e-05, "loss": 2.7219, "step": 46284 }, { "epoch": 3.63, "learning_rate": 6.370657884419353e-05, "loss": 2.6177, "step": 46285 }, { "epoch": 3.63, "learning_rate": 6.370579471496903e-05, "loss": 2.7041, "step": 46286 }, { "epoch": 3.63, "learning_rate": 6.370501058574453e-05, "loss": 2.6415, "step": 46287 }, { "epoch": 3.63, "learning_rate": 6.370422645652004e-05, "loss": 2.7499, "step": 46288 }, { "epoch": 3.63, "learning_rate": 6.370344232729554e-05, "loss": 2.5919, "step": 46289 }, { "epoch": 3.63, "learning_rate": 6.370265819807105e-05, "loss": 2.6013, "step": 46290 }, { "epoch": 3.63, "learning_rate": 6.370187406884655e-05, "loss": 2.679, "step": 46291 }, { "epoch": 3.63, "learning_rate": 6.370108993962205e-05, "loss": 2.7302, "step": 46292 }, { "epoch": 3.63, "learning_rate": 6.370030581039756e-05, "loss": 2.9059, "step": 46293 }, { "epoch": 3.63, "learning_rate": 6.369952168117306e-05, "loss": 2.972, "step": 46294 }, { "epoch": 3.63, "learning_rate": 6.369873755194856e-05, "loss": 2.6951, "step": 46295 }, { "epoch": 3.63, "learning_rate": 6.369795342272407e-05, "loss": 2.7189, "step": 46296 }, { "epoch": 3.63, "learning_rate": 6.369716929349957e-05, "loss": 2.768, "step": 46297 }, { "epoch": 3.63, "learning_rate": 6.369638516427508e-05, "loss": 2.5912, "step": 46298 }, { "epoch": 3.63, "learning_rate": 6.369560103505058e-05, "loss": 2.6372, "step": 46299 }, { "epoch": 3.63, "learning_rate": 6.369481690582608e-05, "loss": 2.7392, "step": 46300 }, { "epoch": 3.63, "learning_rate": 6.369403277660159e-05, "loss": 2.6478, "step": 46301 }, { "epoch": 3.63, "learning_rate": 6.369324864737709e-05, "loss": 2.9696, "step": 46302 }, { "epoch": 3.63, "learning_rate": 6.36924645181526e-05, "loss": 2.864, "step": 46303 }, { "epoch": 3.63, "learning_rate": 6.36916803889281e-05, "loss": 2.6256, "step": 46304 }, { "epoch": 3.63, "learning_rate": 6.36908962597036e-05, "loss": 2.6601, "step": 46305 }, { "epoch": 3.63, "learning_rate": 6.36901121304791e-05, "loss": 2.6426, "step": 46306 }, { "epoch": 3.63, "learning_rate": 6.368932800125461e-05, "loss": 3.0504, "step": 46307 }, { "epoch": 3.63, "learning_rate": 6.368854387203011e-05, "loss": 3.3928, "step": 46308 }, { "epoch": 3.63, "learning_rate": 6.368775974280562e-05, "loss": 2.6919, "step": 46309 }, { "epoch": 3.63, "learning_rate": 6.368697561358112e-05, "loss": 2.6931, "step": 46310 }, { "epoch": 3.63, "learning_rate": 6.368619148435662e-05, "loss": 2.754, "step": 46311 }, { "epoch": 3.63, "learning_rate": 6.368540735513213e-05, "loss": 2.603, "step": 46312 }, { "epoch": 3.63, "learning_rate": 6.368462322590763e-05, "loss": 2.8479, "step": 46313 }, { "epoch": 3.63, "learning_rate": 6.368383909668313e-05, "loss": 2.7467, "step": 46314 }, { "epoch": 3.63, "learning_rate": 6.368305496745864e-05, "loss": 2.9506, "step": 46315 }, { "epoch": 3.63, "learning_rate": 6.368227083823414e-05, "loss": 2.637, "step": 46316 }, { "epoch": 3.63, "learning_rate": 6.368148670900966e-05, "loss": 2.7981, "step": 46317 }, { "epoch": 3.63, "learning_rate": 6.368070257978516e-05, "loss": 2.7143, "step": 46318 }, { "epoch": 3.63, "learning_rate": 6.367991845056067e-05, "loss": 2.541, "step": 46319 }, { "epoch": 3.63, "learning_rate": 6.367913432133617e-05, "loss": 2.9224, "step": 46320 }, { "epoch": 3.63, "learning_rate": 6.367835019211166e-05, "loss": 2.6274, "step": 46321 }, { "epoch": 3.63, "learning_rate": 6.367756606288716e-05, "loss": 2.8623, "step": 46322 }, { "epoch": 3.63, "learning_rate": 6.367678193366267e-05, "loss": 2.6734, "step": 46323 }, { "epoch": 3.63, "learning_rate": 6.367599780443817e-05, "loss": 2.7623, "step": 46324 }, { "epoch": 3.63, "learning_rate": 6.367521367521367e-05, "loss": 2.6227, "step": 46325 }, { "epoch": 3.63, "learning_rate": 6.367442954598918e-05, "loss": 2.7365, "step": 46326 }, { "epoch": 3.63, "learning_rate": 6.367364541676468e-05, "loss": 2.691, "step": 46327 }, { "epoch": 3.63, "learning_rate": 6.367286128754019e-05, "loss": 2.7065, "step": 46328 }, { "epoch": 3.63, "learning_rate": 6.367207715831569e-05, "loss": 2.9918, "step": 46329 }, { "epoch": 3.63, "learning_rate": 6.367129302909119e-05, "loss": 2.5183, "step": 46330 }, { "epoch": 3.63, "learning_rate": 6.36705088998667e-05, "loss": 2.7637, "step": 46331 }, { "epoch": 3.63, "learning_rate": 6.36697247706422e-05, "loss": 2.7982, "step": 46332 }, { "epoch": 3.63, "learning_rate": 6.36689406414177e-05, "loss": 2.8103, "step": 46333 }, { "epoch": 3.63, "learning_rate": 6.366815651219321e-05, "loss": 2.5789, "step": 46334 }, { "epoch": 3.63, "learning_rate": 6.366737238296871e-05, "loss": 2.8883, "step": 46335 }, { "epoch": 3.63, "learning_rate": 6.366658825374422e-05, "loss": 2.7147, "step": 46336 }, { "epoch": 3.63, "learning_rate": 6.366580412451972e-05, "loss": 2.6228, "step": 46337 }, { "epoch": 3.63, "learning_rate": 6.366501999529522e-05, "loss": 2.7412, "step": 46338 }, { "epoch": 3.63, "learning_rate": 6.366423586607073e-05, "loss": 2.6399, "step": 46339 }, { "epoch": 3.63, "learning_rate": 6.366345173684623e-05, "loss": 2.5882, "step": 46340 }, { "epoch": 3.63, "learning_rate": 6.366266760762173e-05, "loss": 2.8255, "step": 46341 }, { "epoch": 3.63, "learning_rate": 6.366188347839725e-05, "loss": 3.0611, "step": 46342 }, { "epoch": 3.63, "learning_rate": 6.366109934917275e-05, "loss": 2.6488, "step": 46343 }, { "epoch": 3.63, "learning_rate": 6.366031521994826e-05, "loss": 2.6199, "step": 46344 }, { "epoch": 3.63, "learning_rate": 6.365953109072376e-05, "loss": 2.6471, "step": 46345 }, { "epoch": 3.63, "learning_rate": 6.365874696149927e-05, "loss": 2.7236, "step": 46346 }, { "epoch": 3.63, "learning_rate": 6.365796283227477e-05, "loss": 2.9368, "step": 46347 }, { "epoch": 3.63, "learning_rate": 6.365717870305027e-05, "loss": 2.8839, "step": 46348 }, { "epoch": 3.63, "learning_rate": 6.365639457382578e-05, "loss": 2.8499, "step": 46349 }, { "epoch": 3.63, "learning_rate": 6.365561044460128e-05, "loss": 2.6157, "step": 46350 }, { "epoch": 3.63, "learning_rate": 6.365482631537678e-05, "loss": 3.0701, "step": 46351 }, { "epoch": 3.63, "learning_rate": 6.365404218615229e-05, "loss": 2.6897, "step": 46352 }, { "epoch": 3.63, "learning_rate": 6.365325805692779e-05, "loss": 2.844, "step": 46353 }, { "epoch": 3.63, "learning_rate": 6.36524739277033e-05, "loss": 2.5896, "step": 46354 }, { "epoch": 3.63, "learning_rate": 6.36516897984788e-05, "loss": 2.7189, "step": 46355 }, { "epoch": 3.63, "learning_rate": 6.36509056692543e-05, "loss": 2.8634, "step": 46356 }, { "epoch": 3.63, "learning_rate": 6.365012154002979e-05, "loss": 2.7206, "step": 46357 }, { "epoch": 3.64, "learning_rate": 6.36493374108053e-05, "loss": 2.6296, "step": 46358 }, { "epoch": 3.64, "learning_rate": 6.36485532815808e-05, "loss": 2.6504, "step": 46359 }, { "epoch": 3.64, "learning_rate": 6.36477691523563e-05, "loss": 2.7005, "step": 46360 }, { "epoch": 3.64, "learning_rate": 6.364698502313181e-05, "loss": 2.705, "step": 46361 }, { "epoch": 3.64, "learning_rate": 6.364620089390731e-05, "loss": 2.8173, "step": 46362 }, { "epoch": 3.64, "learning_rate": 6.364541676468281e-05, "loss": 2.6968, "step": 46363 }, { "epoch": 3.64, "learning_rate": 6.364463263545832e-05, "loss": 2.6839, "step": 46364 }, { "epoch": 3.64, "learning_rate": 6.364384850623382e-05, "loss": 2.7491, "step": 46365 }, { "epoch": 3.64, "learning_rate": 6.364306437700933e-05, "loss": 2.5219, "step": 46366 }, { "epoch": 3.64, "learning_rate": 6.364228024778483e-05, "loss": 2.7454, "step": 46367 }, { "epoch": 3.64, "learning_rate": 6.364149611856035e-05, "loss": 2.5579, "step": 46368 }, { "epoch": 3.64, "learning_rate": 6.364071198933585e-05, "loss": 2.7268, "step": 46369 }, { "epoch": 3.64, "learning_rate": 6.363992786011135e-05, "loss": 2.7268, "step": 46370 }, { "epoch": 3.64, "learning_rate": 6.363914373088686e-05, "loss": 2.7758, "step": 46371 }, { "epoch": 3.64, "learning_rate": 6.363835960166236e-05, "loss": 3.04, "step": 46372 }, { "epoch": 3.64, "learning_rate": 6.363757547243786e-05, "loss": 2.9742, "step": 46373 }, { "epoch": 3.64, "learning_rate": 6.363679134321337e-05, "loss": 2.6897, "step": 46374 }, { "epoch": 3.64, "learning_rate": 6.363600721398887e-05, "loss": 2.6669, "step": 46375 }, { "epoch": 3.64, "learning_rate": 6.363522308476438e-05, "loss": 2.7534, "step": 46376 }, { "epoch": 3.64, "learning_rate": 6.363443895553988e-05, "loss": 2.6782, "step": 46377 }, { "epoch": 3.64, "learning_rate": 6.363365482631538e-05, "loss": 2.6415, "step": 46378 }, { "epoch": 3.64, "learning_rate": 6.363287069709089e-05, "loss": 2.629, "step": 46379 }, { "epoch": 3.64, "learning_rate": 6.363208656786639e-05, "loss": 2.9418, "step": 46380 }, { "epoch": 3.64, "learning_rate": 6.36313024386419e-05, "loss": 2.8977, "step": 46381 }, { "epoch": 3.64, "learning_rate": 6.36305183094174e-05, "loss": 2.807, "step": 46382 }, { "epoch": 3.64, "learning_rate": 6.36297341801929e-05, "loss": 2.762, "step": 46383 }, { "epoch": 3.64, "learning_rate": 6.36289500509684e-05, "loss": 2.6914, "step": 46384 }, { "epoch": 3.64, "learning_rate": 6.362816592174391e-05, "loss": 2.9314, "step": 46385 }, { "epoch": 3.64, "learning_rate": 6.362738179251941e-05, "loss": 2.8227, "step": 46386 }, { "epoch": 3.64, "learning_rate": 6.362659766329492e-05, "loss": 2.9407, "step": 46387 }, { "epoch": 3.64, "learning_rate": 6.362581353407042e-05, "loss": 2.7825, "step": 46388 }, { "epoch": 3.64, "learning_rate": 6.362502940484592e-05, "loss": 2.6896, "step": 46389 }, { "epoch": 3.64, "learning_rate": 6.362424527562143e-05, "loss": 2.6097, "step": 46390 }, { "epoch": 3.64, "learning_rate": 6.362346114639693e-05, "loss": 2.5523, "step": 46391 }, { "epoch": 3.64, "learning_rate": 6.362267701717243e-05, "loss": 2.6346, "step": 46392 }, { "epoch": 3.64, "learning_rate": 6.362189288794792e-05, "loss": 2.6383, "step": 46393 }, { "epoch": 3.64, "learning_rate": 6.362110875872344e-05, "loss": 2.6565, "step": 46394 }, { "epoch": 3.64, "learning_rate": 6.362032462949895e-05, "loss": 2.7608, "step": 46395 }, { "epoch": 3.64, "learning_rate": 6.361954050027445e-05, "loss": 2.7476, "step": 46396 }, { "epoch": 3.64, "learning_rate": 6.361875637104995e-05, "loss": 2.6248, "step": 46397 }, { "epoch": 3.64, "learning_rate": 6.361797224182546e-05, "loss": 2.661, "step": 46398 }, { "epoch": 3.64, "learning_rate": 6.361718811260096e-05, "loss": 2.5357, "step": 46399 }, { "epoch": 3.64, "learning_rate": 6.361640398337646e-05, "loss": 2.7091, "step": 46400 }, { "epoch": 3.64, "learning_rate": 6.361561985415197e-05, "loss": 2.8472, "step": 46401 }, { "epoch": 3.64, "learning_rate": 6.361483572492747e-05, "loss": 2.7729, "step": 46402 }, { "epoch": 3.64, "learning_rate": 6.361405159570298e-05, "loss": 2.913, "step": 46403 }, { "epoch": 3.64, "learning_rate": 6.361326746647848e-05, "loss": 2.8927, "step": 46404 }, { "epoch": 3.64, "learning_rate": 6.361248333725398e-05, "loss": 2.6737, "step": 46405 }, { "epoch": 3.64, "learning_rate": 6.361169920802949e-05, "loss": 2.6134, "step": 46406 }, { "epoch": 3.64, "learning_rate": 6.361091507880499e-05, "loss": 2.6208, "step": 46407 }, { "epoch": 3.64, "learning_rate": 6.36101309495805e-05, "loss": 2.7295, "step": 46408 }, { "epoch": 3.64, "learning_rate": 6.3609346820356e-05, "loss": 2.6914, "step": 46409 }, { "epoch": 3.64, "learning_rate": 6.36085626911315e-05, "loss": 2.7653, "step": 46410 }, { "epoch": 3.64, "learning_rate": 6.3607778561907e-05, "loss": 2.8587, "step": 46411 }, { "epoch": 3.64, "learning_rate": 6.360699443268251e-05, "loss": 2.6116, "step": 46412 }, { "epoch": 3.64, "learning_rate": 6.360621030345801e-05, "loss": 2.7348, "step": 46413 }, { "epoch": 3.64, "learning_rate": 6.360542617423352e-05, "loss": 2.7734, "step": 46414 }, { "epoch": 3.64, "learning_rate": 6.360464204500902e-05, "loss": 2.5009, "step": 46415 }, { "epoch": 3.64, "learning_rate": 6.360385791578452e-05, "loss": 2.6558, "step": 46416 }, { "epoch": 3.64, "learning_rate": 6.360307378656003e-05, "loss": 2.7668, "step": 46417 }, { "epoch": 3.64, "learning_rate": 6.360228965733553e-05, "loss": 2.9662, "step": 46418 }, { "epoch": 3.64, "learning_rate": 6.360150552811103e-05, "loss": 2.7606, "step": 46419 }, { "epoch": 3.64, "learning_rate": 6.360072139888654e-05, "loss": 2.8278, "step": 46420 }, { "epoch": 3.64, "learning_rate": 6.359993726966204e-05, "loss": 2.6943, "step": 46421 }, { "epoch": 3.64, "learning_rate": 6.359915314043754e-05, "loss": 2.6896, "step": 46422 }, { "epoch": 3.64, "learning_rate": 6.359836901121305e-05, "loss": 2.8608, "step": 46423 }, { "epoch": 3.64, "learning_rate": 6.359758488198855e-05, "loss": 2.8442, "step": 46424 }, { "epoch": 3.64, "learning_rate": 6.359680075276407e-05, "loss": 2.71, "step": 46425 }, { "epoch": 3.64, "learning_rate": 6.359601662353957e-05, "loss": 2.7457, "step": 46426 }, { "epoch": 3.64, "learning_rate": 6.359523249431508e-05, "loss": 2.8836, "step": 46427 }, { "epoch": 3.64, "learning_rate": 6.359444836509058e-05, "loss": 2.7267, "step": 46428 }, { "epoch": 3.64, "learning_rate": 6.359366423586607e-05, "loss": 2.7362, "step": 46429 }, { "epoch": 3.64, "learning_rate": 6.359288010664157e-05, "loss": 2.7583, "step": 46430 }, { "epoch": 3.64, "learning_rate": 6.359209597741708e-05, "loss": 2.7245, "step": 46431 }, { "epoch": 3.64, "learning_rate": 6.359131184819258e-05, "loss": 2.5817, "step": 46432 }, { "epoch": 3.64, "learning_rate": 6.359052771896809e-05, "loss": 2.758, "step": 46433 }, { "epoch": 3.64, "learning_rate": 6.358974358974359e-05, "loss": 2.885, "step": 46434 }, { "epoch": 3.64, "learning_rate": 6.358895946051909e-05, "loss": 2.6994, "step": 46435 }, { "epoch": 3.64, "learning_rate": 6.35881753312946e-05, "loss": 2.7154, "step": 46436 }, { "epoch": 3.64, "learning_rate": 6.35873912020701e-05, "loss": 2.6589, "step": 46437 }, { "epoch": 3.64, "learning_rate": 6.35866070728456e-05, "loss": 2.6131, "step": 46438 }, { "epoch": 3.64, "learning_rate": 6.358582294362111e-05, "loss": 2.6718, "step": 46439 }, { "epoch": 3.64, "learning_rate": 6.358503881439661e-05, "loss": 2.9077, "step": 46440 }, { "epoch": 3.64, "learning_rate": 6.358425468517211e-05, "loss": 2.7338, "step": 46441 }, { "epoch": 3.64, "learning_rate": 6.358347055594762e-05, "loss": 2.8191, "step": 46442 }, { "epoch": 3.64, "learning_rate": 6.358268642672312e-05, "loss": 2.6009, "step": 46443 }, { "epoch": 3.64, "learning_rate": 6.358190229749863e-05, "loss": 2.8082, "step": 46444 }, { "epoch": 3.64, "learning_rate": 6.358111816827413e-05, "loss": 2.8664, "step": 46445 }, { "epoch": 3.64, "learning_rate": 6.358033403904963e-05, "loss": 2.7909, "step": 46446 }, { "epoch": 3.64, "learning_rate": 6.357954990982514e-05, "loss": 2.8453, "step": 46447 }, { "epoch": 3.64, "learning_rate": 6.357876578060064e-05, "loss": 2.8254, "step": 46448 }, { "epoch": 3.64, "learning_rate": 6.357798165137614e-05, "loss": 2.7357, "step": 46449 }, { "epoch": 3.64, "learning_rate": 6.357719752215166e-05, "loss": 2.5597, "step": 46450 }, { "epoch": 3.64, "learning_rate": 6.357641339292717e-05, "loss": 2.7276, "step": 46451 }, { "epoch": 3.64, "learning_rate": 6.357562926370267e-05, "loss": 2.7763, "step": 46452 }, { "epoch": 3.64, "learning_rate": 6.357484513447817e-05, "loss": 2.6064, "step": 46453 }, { "epoch": 3.64, "learning_rate": 6.357406100525368e-05, "loss": 2.6339, "step": 46454 }, { "epoch": 3.64, "learning_rate": 6.357327687602918e-05, "loss": 2.6977, "step": 46455 }, { "epoch": 3.64, "learning_rate": 6.357249274680468e-05, "loss": 2.761, "step": 46456 }, { "epoch": 3.64, "learning_rate": 6.357170861758019e-05, "loss": 2.7141, "step": 46457 }, { "epoch": 3.64, "learning_rate": 6.357092448835569e-05, "loss": 2.77, "step": 46458 }, { "epoch": 3.64, "learning_rate": 6.35701403591312e-05, "loss": 2.4982, "step": 46459 }, { "epoch": 3.64, "learning_rate": 6.35693562299067e-05, "loss": 2.8086, "step": 46460 }, { "epoch": 3.64, "learning_rate": 6.35685721006822e-05, "loss": 2.6342, "step": 46461 }, { "epoch": 3.64, "learning_rate": 6.35677879714577e-05, "loss": 2.8244, "step": 46462 }, { "epoch": 3.64, "learning_rate": 6.356700384223321e-05, "loss": 2.6532, "step": 46463 }, { "epoch": 3.64, "learning_rate": 6.356621971300871e-05, "loss": 2.713, "step": 46464 }, { "epoch": 3.64, "learning_rate": 6.35654355837842e-05, "loss": 2.6545, "step": 46465 }, { "epoch": 3.64, "learning_rate": 6.35646514545597e-05, "loss": 2.5408, "step": 46466 }, { "epoch": 3.64, "learning_rate": 6.356386732533521e-05, "loss": 3.0098, "step": 46467 }, { "epoch": 3.64, "learning_rate": 6.356308319611071e-05, "loss": 2.7847, "step": 46468 }, { "epoch": 3.64, "learning_rate": 6.356229906688622e-05, "loss": 2.5634, "step": 46469 }, { "epoch": 3.64, "learning_rate": 6.356151493766172e-05, "loss": 2.6732, "step": 46470 }, { "epoch": 3.64, "learning_rate": 6.356073080843723e-05, "loss": 2.5966, "step": 46471 }, { "epoch": 3.64, "learning_rate": 6.355994667921273e-05, "loss": 2.7064, "step": 46472 }, { "epoch": 3.64, "learning_rate": 6.355916254998823e-05, "loss": 2.676, "step": 46473 }, { "epoch": 3.64, "learning_rate": 6.355837842076374e-05, "loss": 2.8016, "step": 46474 }, { "epoch": 3.64, "learning_rate": 6.355759429153924e-05, "loss": 2.7405, "step": 46475 }, { "epoch": 3.64, "learning_rate": 6.355681016231476e-05, "loss": 3.02, "step": 46476 }, { "epoch": 3.64, "learning_rate": 6.355602603309026e-05, "loss": 2.8048, "step": 46477 }, { "epoch": 3.64, "learning_rate": 6.355524190386576e-05, "loss": 2.5621, "step": 46478 }, { "epoch": 3.64, "learning_rate": 6.355445777464127e-05, "loss": 2.8521, "step": 46479 }, { "epoch": 3.64, "learning_rate": 6.355367364541677e-05, "loss": 2.7451, "step": 46480 }, { "epoch": 3.64, "learning_rate": 6.355288951619228e-05, "loss": 2.5481, "step": 46481 }, { "epoch": 3.64, "learning_rate": 6.355210538696778e-05, "loss": 2.8478, "step": 46482 }, { "epoch": 3.64, "learning_rate": 6.355132125774328e-05, "loss": 2.6018, "step": 46483 }, { "epoch": 3.64, "learning_rate": 6.355053712851879e-05, "loss": 2.7097, "step": 46484 }, { "epoch": 3.65, "learning_rate": 6.354975299929429e-05, "loss": 2.6582, "step": 46485 }, { "epoch": 3.65, "learning_rate": 6.35489688700698e-05, "loss": 2.6523, "step": 46486 }, { "epoch": 3.65, "learning_rate": 6.35481847408453e-05, "loss": 2.8303, "step": 46487 }, { "epoch": 3.65, "learning_rate": 6.35474006116208e-05, "loss": 2.7752, "step": 46488 }, { "epoch": 3.65, "learning_rate": 6.35466164823963e-05, "loss": 2.8932, "step": 46489 }, { "epoch": 3.65, "learning_rate": 6.354583235317181e-05, "loss": 2.9851, "step": 46490 }, { "epoch": 3.65, "learning_rate": 6.354504822394731e-05, "loss": 2.8476, "step": 46491 }, { "epoch": 3.65, "learning_rate": 6.354426409472282e-05, "loss": 2.7491, "step": 46492 }, { "epoch": 3.65, "learning_rate": 6.354347996549832e-05, "loss": 2.7111, "step": 46493 }, { "epoch": 3.65, "learning_rate": 6.354269583627382e-05, "loss": 2.6821, "step": 46494 }, { "epoch": 3.65, "learning_rate": 6.354191170704933e-05, "loss": 2.6935, "step": 46495 }, { "epoch": 3.65, "learning_rate": 6.354112757782483e-05, "loss": 2.6905, "step": 46496 }, { "epoch": 3.65, "learning_rate": 6.354034344860033e-05, "loss": 2.7613, "step": 46497 }, { "epoch": 3.65, "learning_rate": 6.353955931937584e-05, "loss": 2.7172, "step": 46498 }, { "epoch": 3.65, "learning_rate": 6.353877519015134e-05, "loss": 2.7919, "step": 46499 }, { "epoch": 3.65, "learning_rate": 6.353799106092685e-05, "loss": 2.6386, "step": 46500 }, { "epoch": 3.65, "learning_rate": 6.353720693170234e-05, "loss": 2.9869, "step": 46501 }, { "epoch": 3.65, "learning_rate": 6.353642280247785e-05, "loss": 2.6706, "step": 46502 }, { "epoch": 3.65, "learning_rate": 6.353563867325336e-05, "loss": 2.7957, "step": 46503 }, { "epoch": 3.65, "learning_rate": 6.353485454402886e-05, "loss": 2.5632, "step": 46504 }, { "epoch": 3.65, "learning_rate": 6.353407041480436e-05, "loss": 2.8944, "step": 46505 }, { "epoch": 3.65, "learning_rate": 6.353328628557987e-05, "loss": 2.8372, "step": 46506 }, { "epoch": 3.65, "learning_rate": 6.353250215635537e-05, "loss": 2.6364, "step": 46507 }, { "epoch": 3.65, "learning_rate": 6.353171802713087e-05, "loss": 2.7006, "step": 46508 }, { "epoch": 3.65, "learning_rate": 6.353093389790638e-05, "loss": 2.7164, "step": 46509 }, { "epoch": 3.65, "learning_rate": 6.353014976868188e-05, "loss": 2.5182, "step": 46510 }, { "epoch": 3.65, "learning_rate": 6.352936563945739e-05, "loss": 2.93, "step": 46511 }, { "epoch": 3.65, "learning_rate": 6.352858151023289e-05, "loss": 2.8163, "step": 46512 }, { "epoch": 3.65, "learning_rate": 6.352779738100839e-05, "loss": 2.7829, "step": 46513 }, { "epoch": 3.65, "learning_rate": 6.35270132517839e-05, "loss": 2.7231, "step": 46514 }, { "epoch": 3.65, "learning_rate": 6.35262291225594e-05, "loss": 2.6235, "step": 46515 }, { "epoch": 3.65, "learning_rate": 6.35254449933349e-05, "loss": 2.8369, "step": 46516 }, { "epoch": 3.65, "learning_rate": 6.352466086411041e-05, "loss": 2.7885, "step": 46517 }, { "epoch": 3.65, "learning_rate": 6.352387673488591e-05, "loss": 2.6533, "step": 46518 }, { "epoch": 3.65, "learning_rate": 6.352309260566142e-05, "loss": 2.5943, "step": 46519 }, { "epoch": 3.65, "learning_rate": 6.352230847643692e-05, "loss": 2.8547, "step": 46520 }, { "epoch": 3.65, "learning_rate": 6.352152434721242e-05, "loss": 2.7024, "step": 46521 }, { "epoch": 3.65, "learning_rate": 6.352074021798793e-05, "loss": 2.7854, "step": 46522 }, { "epoch": 3.65, "learning_rate": 6.351995608876343e-05, "loss": 2.6991, "step": 46523 }, { "epoch": 3.65, "learning_rate": 6.351917195953893e-05, "loss": 2.9431, "step": 46524 }, { "epoch": 3.65, "learning_rate": 6.351838783031444e-05, "loss": 2.6497, "step": 46525 }, { "epoch": 3.65, "learning_rate": 6.351760370108994e-05, "loss": 2.6913, "step": 46526 }, { "epoch": 3.65, "learning_rate": 6.351681957186544e-05, "loss": 2.5764, "step": 46527 }, { "epoch": 3.65, "learning_rate": 6.351603544264095e-05, "loss": 2.5943, "step": 46528 }, { "epoch": 3.65, "learning_rate": 6.351525131341645e-05, "loss": 2.7528, "step": 46529 }, { "epoch": 3.65, "learning_rate": 6.351446718419196e-05, "loss": 2.6072, "step": 46530 }, { "epoch": 3.65, "learning_rate": 6.351368305496746e-05, "loss": 2.9974, "step": 46531 }, { "epoch": 3.65, "learning_rate": 6.351289892574296e-05, "loss": 2.692, "step": 46532 }, { "epoch": 3.65, "learning_rate": 6.351211479651848e-05, "loss": 3.1295, "step": 46533 }, { "epoch": 3.65, "learning_rate": 6.351133066729398e-05, "loss": 2.9611, "step": 46534 }, { "epoch": 3.65, "learning_rate": 6.351054653806949e-05, "loss": 2.6845, "step": 46535 }, { "epoch": 3.65, "learning_rate": 6.350976240884499e-05, "loss": 2.712, "step": 46536 }, { "epoch": 3.65, "learning_rate": 6.350897827962048e-05, "loss": 2.735, "step": 46537 }, { "epoch": 3.65, "learning_rate": 6.350819415039599e-05, "loss": 2.6485, "step": 46538 }, { "epoch": 3.65, "learning_rate": 6.350741002117149e-05, "loss": 3.0021, "step": 46539 }, { "epoch": 3.65, "learning_rate": 6.350662589194699e-05, "loss": 2.8174, "step": 46540 }, { "epoch": 3.65, "learning_rate": 6.35058417627225e-05, "loss": 2.6657, "step": 46541 }, { "epoch": 3.65, "learning_rate": 6.3505057633498e-05, "loss": 2.7813, "step": 46542 }, { "epoch": 3.65, "learning_rate": 6.35042735042735e-05, "loss": 2.8589, "step": 46543 }, { "epoch": 3.65, "learning_rate": 6.350348937504901e-05, "loss": 2.7987, "step": 46544 }, { "epoch": 3.65, "learning_rate": 6.350270524582451e-05, "loss": 2.8356, "step": 46545 }, { "epoch": 3.65, "learning_rate": 6.350192111660001e-05, "loss": 2.7778, "step": 46546 }, { "epoch": 3.65, "learning_rate": 6.350113698737552e-05, "loss": 2.8301, "step": 46547 }, { "epoch": 3.65, "learning_rate": 6.350035285815102e-05, "loss": 2.7639, "step": 46548 }, { "epoch": 3.65, "learning_rate": 6.349956872892653e-05, "loss": 2.6447, "step": 46549 }, { "epoch": 3.65, "learning_rate": 6.349878459970203e-05, "loss": 2.9711, "step": 46550 }, { "epoch": 3.65, "learning_rate": 6.349800047047753e-05, "loss": 2.9423, "step": 46551 }, { "epoch": 3.65, "learning_rate": 6.349721634125304e-05, "loss": 2.7374, "step": 46552 }, { "epoch": 3.65, "learning_rate": 6.349643221202854e-05, "loss": 2.5376, "step": 46553 }, { "epoch": 3.65, "learning_rate": 6.349564808280404e-05, "loss": 2.9195, "step": 46554 }, { "epoch": 3.65, "learning_rate": 6.349486395357955e-05, "loss": 2.9708, "step": 46555 }, { "epoch": 3.65, "learning_rate": 6.349407982435505e-05, "loss": 2.7231, "step": 46556 }, { "epoch": 3.65, "learning_rate": 6.349329569513055e-05, "loss": 2.8567, "step": 46557 }, { "epoch": 3.65, "learning_rate": 6.349251156590607e-05, "loss": 2.5889, "step": 46558 }, { "epoch": 3.65, "learning_rate": 6.349172743668158e-05, "loss": 2.7624, "step": 46559 }, { "epoch": 3.65, "learning_rate": 6.349094330745708e-05, "loss": 2.5864, "step": 46560 }, { "epoch": 3.65, "learning_rate": 6.349015917823258e-05, "loss": 2.6649, "step": 46561 }, { "epoch": 3.65, "learning_rate": 6.348937504900809e-05, "loss": 2.4645, "step": 46562 }, { "epoch": 3.65, "learning_rate": 6.348859091978359e-05, "loss": 2.6233, "step": 46563 }, { "epoch": 3.65, "learning_rate": 6.34878067905591e-05, "loss": 3.0231, "step": 46564 }, { "epoch": 3.65, "learning_rate": 6.34870226613346e-05, "loss": 2.6175, "step": 46565 }, { "epoch": 3.65, "learning_rate": 6.34862385321101e-05, "loss": 2.7428, "step": 46566 }, { "epoch": 3.65, "learning_rate": 6.34854544028856e-05, "loss": 2.5562, "step": 46567 }, { "epoch": 3.65, "learning_rate": 6.348467027366111e-05, "loss": 2.9363, "step": 46568 }, { "epoch": 3.65, "learning_rate": 6.348388614443661e-05, "loss": 2.5934, "step": 46569 }, { "epoch": 3.65, "learning_rate": 6.348310201521212e-05, "loss": 2.8166, "step": 46570 }, { "epoch": 3.65, "learning_rate": 6.348231788598762e-05, "loss": 2.6572, "step": 46571 }, { "epoch": 3.65, "learning_rate": 6.348153375676312e-05, "loss": 2.7147, "step": 46572 }, { "epoch": 3.65, "learning_rate": 6.348074962753861e-05, "loss": 2.9836, "step": 46573 }, { "epoch": 3.65, "learning_rate": 6.347996549831412e-05, "loss": 2.6721, "step": 46574 }, { "epoch": 3.65, "learning_rate": 6.347918136908962e-05, "loss": 2.849, "step": 46575 }, { "epoch": 3.65, "learning_rate": 6.347839723986512e-05, "loss": 2.8176, "step": 46576 }, { "epoch": 3.65, "learning_rate": 6.347761311064063e-05, "loss": 2.6665, "step": 46577 }, { "epoch": 3.65, "learning_rate": 6.347682898141613e-05, "loss": 2.738, "step": 46578 }, { "epoch": 3.65, "learning_rate": 6.347604485219164e-05, "loss": 2.8092, "step": 46579 }, { "epoch": 3.65, "learning_rate": 6.347526072296714e-05, "loss": 2.8195, "step": 46580 }, { "epoch": 3.65, "learning_rate": 6.347447659374264e-05, "loss": 2.7213, "step": 46581 }, { "epoch": 3.65, "learning_rate": 6.347369246451815e-05, "loss": 2.729, "step": 46582 }, { "epoch": 3.65, "learning_rate": 6.347290833529365e-05, "loss": 2.5506, "step": 46583 }, { "epoch": 3.65, "learning_rate": 6.347212420606917e-05, "loss": 2.6484, "step": 46584 }, { "epoch": 3.65, "learning_rate": 6.347134007684467e-05, "loss": 2.6859, "step": 46585 }, { "epoch": 3.65, "learning_rate": 6.347055594762018e-05, "loss": 2.5395, "step": 46586 }, { "epoch": 3.65, "learning_rate": 6.346977181839568e-05, "loss": 2.775, "step": 46587 }, { "epoch": 3.65, "learning_rate": 6.346898768917118e-05, "loss": 2.6085, "step": 46588 }, { "epoch": 3.65, "learning_rate": 6.346820355994669e-05, "loss": 2.7667, "step": 46589 }, { "epoch": 3.65, "learning_rate": 6.346741943072219e-05, "loss": 2.815, "step": 46590 }, { "epoch": 3.65, "learning_rate": 6.34666353014977e-05, "loss": 2.8818, "step": 46591 }, { "epoch": 3.65, "learning_rate": 6.34658511722732e-05, "loss": 2.6283, "step": 46592 }, { "epoch": 3.65, "learning_rate": 6.34650670430487e-05, "loss": 2.6857, "step": 46593 }, { "epoch": 3.65, "learning_rate": 6.34642829138242e-05, "loss": 2.6578, "step": 46594 }, { "epoch": 3.65, "learning_rate": 6.346349878459971e-05, "loss": 2.788, "step": 46595 }, { "epoch": 3.65, "learning_rate": 6.346271465537521e-05, "loss": 3.0813, "step": 46596 }, { "epoch": 3.65, "learning_rate": 6.346193052615072e-05, "loss": 2.6855, "step": 46597 }, { "epoch": 3.65, "learning_rate": 6.346114639692622e-05, "loss": 2.7006, "step": 46598 }, { "epoch": 3.65, "learning_rate": 6.346036226770172e-05, "loss": 2.8528, "step": 46599 }, { "epoch": 3.65, "learning_rate": 6.345957813847723e-05, "loss": 2.6123, "step": 46600 }, { "epoch": 3.65, "learning_rate": 6.345879400925273e-05, "loss": 2.7964, "step": 46601 }, { "epoch": 3.65, "learning_rate": 6.345800988002823e-05, "loss": 2.8505, "step": 46602 }, { "epoch": 3.65, "learning_rate": 6.345722575080374e-05, "loss": 2.717, "step": 46603 }, { "epoch": 3.65, "learning_rate": 6.345644162157924e-05, "loss": 2.7853, "step": 46604 }, { "epoch": 3.65, "learning_rate": 6.345565749235475e-05, "loss": 2.662, "step": 46605 }, { "epoch": 3.65, "learning_rate": 6.345487336313025e-05, "loss": 2.7487, "step": 46606 }, { "epoch": 3.65, "learning_rate": 6.345408923390575e-05, "loss": 2.7247, "step": 46607 }, { "epoch": 3.65, "learning_rate": 6.345330510468126e-05, "loss": 2.8784, "step": 46608 }, { "epoch": 3.65, "learning_rate": 6.345252097545675e-05, "loss": 2.7802, "step": 46609 }, { "epoch": 3.65, "learning_rate": 6.345173684623226e-05, "loss": 2.7927, "step": 46610 }, { "epoch": 3.65, "learning_rate": 6.345095271700777e-05, "loss": 2.6727, "step": 46611 }, { "epoch": 3.65, "learning_rate": 6.345016858778327e-05, "loss": 2.8005, "step": 46612 }, { "epoch": 3.66, "learning_rate": 6.344938445855877e-05, "loss": 2.6939, "step": 46613 }, { "epoch": 3.66, "learning_rate": 6.344860032933428e-05, "loss": 2.8803, "step": 46614 }, { "epoch": 3.66, "learning_rate": 6.344781620010978e-05, "loss": 2.6112, "step": 46615 }, { "epoch": 3.66, "learning_rate": 6.344703207088529e-05, "loss": 2.8108, "step": 46616 }, { "epoch": 3.66, "learning_rate": 6.344624794166079e-05, "loss": 2.8086, "step": 46617 }, { "epoch": 3.66, "learning_rate": 6.344546381243629e-05, "loss": 2.6646, "step": 46618 }, { "epoch": 3.66, "learning_rate": 6.34446796832118e-05, "loss": 2.7962, "step": 46619 }, { "epoch": 3.66, "learning_rate": 6.34438955539873e-05, "loss": 2.6714, "step": 46620 }, { "epoch": 3.66, "learning_rate": 6.34431114247628e-05, "loss": 2.8691, "step": 46621 }, { "epoch": 3.66, "learning_rate": 6.344232729553831e-05, "loss": 2.8224, "step": 46622 }, { "epoch": 3.66, "learning_rate": 6.344154316631381e-05, "loss": 2.6635, "step": 46623 }, { "epoch": 3.66, "learning_rate": 6.344075903708931e-05, "loss": 2.6208, "step": 46624 }, { "epoch": 3.66, "learning_rate": 6.343997490786482e-05, "loss": 2.9766, "step": 46625 }, { "epoch": 3.66, "learning_rate": 6.343919077864032e-05, "loss": 2.7649, "step": 46626 }, { "epoch": 3.66, "learning_rate": 6.343840664941583e-05, "loss": 2.8197, "step": 46627 }, { "epoch": 3.66, "learning_rate": 6.343762252019133e-05, "loss": 2.9986, "step": 46628 }, { "epoch": 3.66, "learning_rate": 6.343683839096683e-05, "loss": 2.6382, "step": 46629 }, { "epoch": 3.66, "learning_rate": 6.343605426174234e-05, "loss": 2.6384, "step": 46630 }, { "epoch": 3.66, "learning_rate": 6.343527013251784e-05, "loss": 2.549, "step": 46631 }, { "epoch": 3.66, "learning_rate": 6.343448600329334e-05, "loss": 2.5619, "step": 46632 }, { "epoch": 3.66, "learning_rate": 6.343370187406885e-05, "loss": 2.6546, "step": 46633 }, { "epoch": 3.66, "learning_rate": 6.343291774484435e-05, "loss": 2.8241, "step": 46634 }, { "epoch": 3.66, "learning_rate": 6.343213361561986e-05, "loss": 2.6842, "step": 46635 }, { "epoch": 3.66, "learning_rate": 6.343134948639536e-05, "loss": 2.8493, "step": 46636 }, { "epoch": 3.66, "learning_rate": 6.343056535717086e-05, "loss": 2.7613, "step": 46637 }, { "epoch": 3.66, "learning_rate": 6.342978122794637e-05, "loss": 2.7577, "step": 46638 }, { "epoch": 3.66, "learning_rate": 6.342899709872187e-05, "loss": 2.5905, "step": 46639 }, { "epoch": 3.66, "learning_rate": 6.342821296949737e-05, "loss": 2.8257, "step": 46640 }, { "epoch": 3.66, "learning_rate": 6.342742884027289e-05, "loss": 2.6858, "step": 46641 }, { "epoch": 3.66, "learning_rate": 6.34266447110484e-05, "loss": 2.5757, "step": 46642 }, { "epoch": 3.66, "learning_rate": 6.34258605818239e-05, "loss": 2.8456, "step": 46643 }, { "epoch": 3.66, "learning_rate": 6.34250764525994e-05, "loss": 2.6787, "step": 46644 }, { "epoch": 3.66, "learning_rate": 6.342429232337489e-05, "loss": 2.6633, "step": 46645 }, { "epoch": 3.66, "learning_rate": 6.34235081941504e-05, "loss": 3.1058, "step": 46646 }, { "epoch": 3.66, "learning_rate": 6.34227240649259e-05, "loss": 2.8851, "step": 46647 }, { "epoch": 3.66, "learning_rate": 6.34219399357014e-05, "loss": 2.698, "step": 46648 }, { "epoch": 3.66, "learning_rate": 6.342115580647691e-05, "loss": 2.5937, "step": 46649 }, { "epoch": 3.66, "learning_rate": 6.342037167725241e-05, "loss": 2.5578, "step": 46650 }, { "epoch": 3.66, "learning_rate": 6.341958754802791e-05, "loss": 2.5555, "step": 46651 }, { "epoch": 3.66, "learning_rate": 6.341880341880342e-05, "loss": 2.7623, "step": 46652 }, { "epoch": 3.66, "learning_rate": 6.341801928957892e-05, "loss": 2.7796, "step": 46653 }, { "epoch": 3.66, "learning_rate": 6.341723516035443e-05, "loss": 2.759, "step": 46654 }, { "epoch": 3.66, "learning_rate": 6.341645103112993e-05, "loss": 2.8151, "step": 46655 }, { "epoch": 3.66, "learning_rate": 6.341566690190543e-05, "loss": 2.7741, "step": 46656 }, { "epoch": 3.66, "learning_rate": 6.341488277268094e-05, "loss": 2.7255, "step": 46657 }, { "epoch": 3.66, "learning_rate": 6.341409864345644e-05, "loss": 2.9135, "step": 46658 }, { "epoch": 3.66, "learning_rate": 6.341331451423194e-05, "loss": 2.6751, "step": 46659 }, { "epoch": 3.66, "learning_rate": 6.341253038500745e-05, "loss": 2.8137, "step": 46660 }, { "epoch": 3.66, "learning_rate": 6.341174625578295e-05, "loss": 2.6597, "step": 46661 }, { "epoch": 3.66, "learning_rate": 6.341096212655845e-05, "loss": 2.7247, "step": 46662 }, { "epoch": 3.66, "learning_rate": 6.341017799733396e-05, "loss": 2.6149, "step": 46663 }, { "epoch": 3.66, "learning_rate": 6.340939386810946e-05, "loss": 2.8324, "step": 46664 }, { "epoch": 3.66, "learning_rate": 6.340860973888497e-05, "loss": 2.6674, "step": 46665 }, { "epoch": 3.66, "learning_rate": 6.340782560966047e-05, "loss": 2.541, "step": 46666 }, { "epoch": 3.66, "learning_rate": 6.340704148043599e-05, "loss": 2.6752, "step": 46667 }, { "epoch": 3.66, "learning_rate": 6.340625735121149e-05, "loss": 2.6938, "step": 46668 }, { "epoch": 3.66, "learning_rate": 6.3405473221987e-05, "loss": 2.6751, "step": 46669 }, { "epoch": 3.66, "learning_rate": 6.34046890927625e-05, "loss": 2.6173, "step": 46670 }, { "epoch": 3.66, "learning_rate": 6.3403904963538e-05, "loss": 2.8535, "step": 46671 }, { "epoch": 3.66, "learning_rate": 6.34031208343135e-05, "loss": 2.703, "step": 46672 }, { "epoch": 3.66, "learning_rate": 6.340233670508901e-05, "loss": 2.7842, "step": 46673 }, { "epoch": 3.66, "learning_rate": 6.340155257586451e-05, "loss": 2.5562, "step": 46674 }, { "epoch": 3.66, "learning_rate": 6.340076844664002e-05, "loss": 2.8899, "step": 46675 }, { "epoch": 3.66, "learning_rate": 6.339998431741552e-05, "loss": 2.5864, "step": 46676 }, { "epoch": 3.66, "learning_rate": 6.339920018819102e-05, "loss": 3.0988, "step": 46677 }, { "epoch": 3.66, "learning_rate": 6.339841605896653e-05, "loss": 2.772, "step": 46678 }, { "epoch": 3.66, "learning_rate": 6.339763192974203e-05, "loss": 2.7867, "step": 46679 }, { "epoch": 3.66, "learning_rate": 6.339684780051753e-05, "loss": 2.7227, "step": 46680 }, { "epoch": 3.66, "learning_rate": 6.339606367129302e-05, "loss": 2.8246, "step": 46681 }, { "epoch": 3.66, "learning_rate": 6.339527954206853e-05, "loss": 2.68, "step": 46682 }, { "epoch": 3.66, "learning_rate": 6.339449541284403e-05, "loss": 2.8485, "step": 46683 }, { "epoch": 3.66, "learning_rate": 6.339371128361954e-05, "loss": 2.7988, "step": 46684 }, { "epoch": 3.66, "learning_rate": 6.339292715439504e-05, "loss": 2.7068, "step": 46685 }, { "epoch": 3.66, "learning_rate": 6.339214302517054e-05, "loss": 2.7379, "step": 46686 }, { "epoch": 3.66, "learning_rate": 6.339135889594605e-05, "loss": 2.6885, "step": 46687 }, { "epoch": 3.66, "learning_rate": 6.339057476672155e-05, "loss": 2.7828, "step": 46688 }, { "epoch": 3.66, "learning_rate": 6.338979063749705e-05, "loss": 2.6715, "step": 46689 }, { "epoch": 3.66, "learning_rate": 6.338900650827256e-05, "loss": 2.9052, "step": 46690 }, { "epoch": 3.66, "learning_rate": 6.338822237904806e-05, "loss": 2.6778, "step": 46691 }, { "epoch": 3.66, "learning_rate": 6.338743824982358e-05, "loss": 2.8066, "step": 46692 }, { "epoch": 3.66, "learning_rate": 6.338665412059908e-05, "loss": 2.8464, "step": 46693 }, { "epoch": 3.66, "learning_rate": 6.338586999137459e-05, "loss": 2.8161, "step": 46694 }, { "epoch": 3.66, "learning_rate": 6.338508586215009e-05, "loss": 2.8736, "step": 46695 }, { "epoch": 3.66, "learning_rate": 6.33843017329256e-05, "loss": 2.7552, "step": 46696 }, { "epoch": 3.66, "learning_rate": 6.33835176037011e-05, "loss": 2.6393, "step": 46697 }, { "epoch": 3.66, "learning_rate": 6.33827334744766e-05, "loss": 2.6326, "step": 46698 }, { "epoch": 3.66, "learning_rate": 6.33819493452521e-05, "loss": 2.7386, "step": 46699 }, { "epoch": 3.66, "learning_rate": 6.338116521602761e-05, "loss": 2.8931, "step": 46700 }, { "epoch": 3.66, "learning_rate": 6.338038108680311e-05, "loss": 2.5475, "step": 46701 }, { "epoch": 3.66, "learning_rate": 6.337959695757862e-05, "loss": 2.9813, "step": 46702 }, { "epoch": 3.66, "learning_rate": 6.337881282835412e-05, "loss": 2.619, "step": 46703 }, { "epoch": 3.66, "learning_rate": 6.337802869912962e-05, "loss": 2.9053, "step": 46704 }, { "epoch": 3.66, "learning_rate": 6.337724456990513e-05, "loss": 2.815, "step": 46705 }, { "epoch": 3.66, "learning_rate": 6.337646044068063e-05, "loss": 2.5197, "step": 46706 }, { "epoch": 3.66, "learning_rate": 6.337567631145613e-05, "loss": 2.8333, "step": 46707 }, { "epoch": 3.66, "learning_rate": 6.337489218223164e-05, "loss": 2.6833, "step": 46708 }, { "epoch": 3.66, "learning_rate": 6.337410805300714e-05, "loss": 2.6269, "step": 46709 }, { "epoch": 3.66, "learning_rate": 6.337332392378264e-05, "loss": 2.7751, "step": 46710 }, { "epoch": 3.66, "learning_rate": 6.337253979455815e-05, "loss": 2.7145, "step": 46711 }, { "epoch": 3.66, "learning_rate": 6.337175566533365e-05, "loss": 2.8071, "step": 46712 }, { "epoch": 3.66, "learning_rate": 6.337097153610916e-05, "loss": 2.8096, "step": 46713 }, { "epoch": 3.66, "learning_rate": 6.337018740688466e-05, "loss": 2.5658, "step": 46714 }, { "epoch": 3.66, "learning_rate": 6.336940327766016e-05, "loss": 2.8007, "step": 46715 }, { "epoch": 3.66, "learning_rate": 6.336861914843567e-05, "loss": 2.8241, "step": 46716 }, { "epoch": 3.66, "learning_rate": 6.336783501921116e-05, "loss": 2.779, "step": 46717 }, { "epoch": 3.66, "learning_rate": 6.336705088998667e-05, "loss": 2.6181, "step": 46718 }, { "epoch": 3.66, "learning_rate": 6.336626676076218e-05, "loss": 2.6183, "step": 46719 }, { "epoch": 3.66, "learning_rate": 6.336548263153768e-05, "loss": 2.5768, "step": 46720 }, { "epoch": 3.66, "learning_rate": 6.336469850231319e-05, "loss": 2.8007, "step": 46721 }, { "epoch": 3.66, "learning_rate": 6.336391437308869e-05, "loss": 2.7046, "step": 46722 }, { "epoch": 3.66, "learning_rate": 6.336313024386419e-05, "loss": 2.6134, "step": 46723 }, { "epoch": 3.66, "learning_rate": 6.33623461146397e-05, "loss": 2.9721, "step": 46724 }, { "epoch": 3.66, "learning_rate": 6.33615619854152e-05, "loss": 2.6658, "step": 46725 }, { "epoch": 3.66, "learning_rate": 6.33607778561907e-05, "loss": 2.6858, "step": 46726 }, { "epoch": 3.66, "learning_rate": 6.335999372696621e-05, "loss": 2.7136, "step": 46727 }, { "epoch": 3.66, "learning_rate": 6.335920959774171e-05, "loss": 2.8127, "step": 46728 }, { "epoch": 3.66, "learning_rate": 6.335842546851721e-05, "loss": 2.8759, "step": 46729 }, { "epoch": 3.66, "learning_rate": 6.335764133929272e-05, "loss": 2.6726, "step": 46730 }, { "epoch": 3.66, "learning_rate": 6.335685721006822e-05, "loss": 2.5779, "step": 46731 }, { "epoch": 3.66, "learning_rate": 6.335607308084373e-05, "loss": 2.6958, "step": 46732 }, { "epoch": 3.66, "learning_rate": 6.335528895161923e-05, "loss": 2.587, "step": 46733 }, { "epoch": 3.66, "learning_rate": 6.335450482239473e-05, "loss": 2.3882, "step": 46734 }, { "epoch": 3.66, "learning_rate": 6.335372069317024e-05, "loss": 2.8676, "step": 46735 }, { "epoch": 3.66, "learning_rate": 6.335293656394574e-05, "loss": 2.7753, "step": 46736 }, { "epoch": 3.66, "learning_rate": 6.335215243472124e-05, "loss": 2.6111, "step": 46737 }, { "epoch": 3.66, "learning_rate": 6.335136830549675e-05, "loss": 2.8611, "step": 46738 }, { "epoch": 3.66, "learning_rate": 6.335058417627225e-05, "loss": 2.6585, "step": 46739 }, { "epoch": 3.67, "learning_rate": 6.334980004704776e-05, "loss": 2.7302, "step": 46740 }, { "epoch": 3.67, "learning_rate": 6.334901591782326e-05, "loss": 2.4654, "step": 46741 }, { "epoch": 3.67, "learning_rate": 6.334823178859876e-05, "loss": 2.8295, "step": 46742 }, { "epoch": 3.67, "learning_rate": 6.334744765937427e-05, "loss": 2.6828, "step": 46743 }, { "epoch": 3.67, "learning_rate": 6.334666353014977e-05, "loss": 2.7844, "step": 46744 }, { "epoch": 3.67, "learning_rate": 6.334587940092527e-05, "loss": 3.2894, "step": 46745 }, { "epoch": 3.67, "learning_rate": 6.334509527170078e-05, "loss": 2.468, "step": 46746 }, { "epoch": 3.67, "learning_rate": 6.334431114247628e-05, "loss": 2.9104, "step": 46747 }, { "epoch": 3.67, "learning_rate": 6.334352701325178e-05, "loss": 2.7434, "step": 46748 }, { "epoch": 3.67, "learning_rate": 6.33427428840273e-05, "loss": 2.8052, "step": 46749 }, { "epoch": 3.67, "learning_rate": 6.33419587548028e-05, "loss": 2.8287, "step": 46750 }, { "epoch": 3.67, "learning_rate": 6.334117462557831e-05, "loss": 2.737, "step": 46751 }, { "epoch": 3.67, "learning_rate": 6.334039049635381e-05, "loss": 2.6336, "step": 46752 }, { "epoch": 3.67, "learning_rate": 6.33396063671293e-05, "loss": 2.6611, "step": 46753 }, { "epoch": 3.67, "learning_rate": 6.33388222379048e-05, "loss": 2.9984, "step": 46754 }, { "epoch": 3.67, "learning_rate": 6.333803810868031e-05, "loss": 2.7258, "step": 46755 }, { "epoch": 3.67, "learning_rate": 6.333725397945581e-05, "loss": 2.6877, "step": 46756 }, { "epoch": 3.67, "learning_rate": 6.333646985023132e-05, "loss": 2.6628, "step": 46757 }, { "epoch": 3.67, "learning_rate": 6.333568572100682e-05, "loss": 2.6264, "step": 46758 }, { "epoch": 3.67, "learning_rate": 6.333490159178232e-05, "loss": 2.8306, "step": 46759 }, { "epoch": 3.67, "learning_rate": 6.333411746255783e-05, "loss": 2.7646, "step": 46760 }, { "epoch": 3.67, "learning_rate": 6.333333333333333e-05, "loss": 2.7045, "step": 46761 }, { "epoch": 3.67, "learning_rate": 6.333254920410884e-05, "loss": 2.6798, "step": 46762 }, { "epoch": 3.67, "learning_rate": 6.333176507488434e-05, "loss": 2.5259, "step": 46763 }, { "epoch": 3.67, "learning_rate": 6.333098094565984e-05, "loss": 2.5319, "step": 46764 }, { "epoch": 3.67, "learning_rate": 6.333019681643535e-05, "loss": 2.5991, "step": 46765 }, { "epoch": 3.67, "learning_rate": 6.332941268721085e-05, "loss": 2.6413, "step": 46766 }, { "epoch": 3.67, "learning_rate": 6.332862855798635e-05, "loss": 2.5235, "step": 46767 }, { "epoch": 3.67, "learning_rate": 6.332784442876186e-05, "loss": 2.6625, "step": 46768 }, { "epoch": 3.67, "learning_rate": 6.332706029953736e-05, "loss": 2.8964, "step": 46769 }, { "epoch": 3.67, "learning_rate": 6.332627617031287e-05, "loss": 2.8727, "step": 46770 }, { "epoch": 3.67, "learning_rate": 6.332549204108837e-05, "loss": 2.6808, "step": 46771 }, { "epoch": 3.67, "learning_rate": 6.332470791186387e-05, "loss": 2.7309, "step": 46772 }, { "epoch": 3.67, "learning_rate": 6.332392378263938e-05, "loss": 2.4452, "step": 46773 }, { "epoch": 3.67, "learning_rate": 6.332313965341488e-05, "loss": 2.5652, "step": 46774 }, { "epoch": 3.67, "learning_rate": 6.33223555241904e-05, "loss": 2.8656, "step": 46775 }, { "epoch": 3.67, "learning_rate": 6.33215713949659e-05, "loss": 2.6909, "step": 46776 }, { "epoch": 3.67, "learning_rate": 6.33207872657414e-05, "loss": 2.6546, "step": 46777 }, { "epoch": 3.67, "learning_rate": 6.332000313651691e-05, "loss": 2.5402, "step": 46778 }, { "epoch": 3.67, "learning_rate": 6.331921900729241e-05, "loss": 2.7169, "step": 46779 }, { "epoch": 3.67, "learning_rate": 6.331843487806792e-05, "loss": 2.7367, "step": 46780 }, { "epoch": 3.67, "learning_rate": 6.331765074884342e-05, "loss": 2.4452, "step": 46781 }, { "epoch": 3.67, "learning_rate": 6.331686661961892e-05, "loss": 2.7184, "step": 46782 }, { "epoch": 3.67, "learning_rate": 6.331608249039443e-05, "loss": 2.7227, "step": 46783 }, { "epoch": 3.67, "learning_rate": 6.331529836116993e-05, "loss": 2.7504, "step": 46784 }, { "epoch": 3.67, "learning_rate": 6.331451423194543e-05, "loss": 2.71, "step": 46785 }, { "epoch": 3.67, "learning_rate": 6.331373010272094e-05, "loss": 2.5662, "step": 46786 }, { "epoch": 3.67, "learning_rate": 6.331294597349644e-05, "loss": 3.1293, "step": 46787 }, { "epoch": 3.67, "learning_rate": 6.331216184427195e-05, "loss": 2.6218, "step": 46788 }, { "epoch": 3.67, "learning_rate": 6.331137771504744e-05, "loss": 2.5198, "step": 46789 }, { "epoch": 3.67, "learning_rate": 6.331059358582294e-05, "loss": 2.8986, "step": 46790 }, { "epoch": 3.67, "learning_rate": 6.330980945659844e-05, "loss": 2.6127, "step": 46791 }, { "epoch": 3.67, "learning_rate": 6.330902532737395e-05, "loss": 2.8369, "step": 46792 }, { "epoch": 3.67, "learning_rate": 6.330824119814945e-05, "loss": 2.7061, "step": 46793 }, { "epoch": 3.67, "learning_rate": 6.330745706892495e-05, "loss": 2.5947, "step": 46794 }, { "epoch": 3.67, "learning_rate": 6.330667293970046e-05, "loss": 2.6142, "step": 46795 }, { "epoch": 3.67, "learning_rate": 6.330588881047596e-05, "loss": 2.8931, "step": 46796 }, { "epoch": 3.67, "learning_rate": 6.330510468125146e-05, "loss": 2.848, "step": 46797 }, { "epoch": 3.67, "learning_rate": 6.330432055202697e-05, "loss": 2.6789, "step": 46798 }, { "epoch": 3.67, "learning_rate": 6.330353642280247e-05, "loss": 2.6212, "step": 46799 }, { "epoch": 3.67, "learning_rate": 6.330275229357799e-05, "loss": 2.7198, "step": 46800 }, { "epoch": 3.67, "learning_rate": 6.330196816435349e-05, "loss": 2.8226, "step": 46801 }, { "epoch": 3.67, "learning_rate": 6.3301184035129e-05, "loss": 2.9593, "step": 46802 }, { "epoch": 3.67, "learning_rate": 6.33003999059045e-05, "loss": 2.7347, "step": 46803 }, { "epoch": 3.67, "learning_rate": 6.329961577668e-05, "loss": 2.7285, "step": 46804 }, { "epoch": 3.67, "learning_rate": 6.329883164745551e-05, "loss": 2.6505, "step": 46805 }, { "epoch": 3.67, "learning_rate": 6.329804751823101e-05, "loss": 2.6806, "step": 46806 }, { "epoch": 3.67, "learning_rate": 6.329726338900652e-05, "loss": 2.8937, "step": 46807 }, { "epoch": 3.67, "learning_rate": 6.329647925978202e-05, "loss": 2.729, "step": 46808 }, { "epoch": 3.67, "learning_rate": 6.329569513055752e-05, "loss": 2.9026, "step": 46809 }, { "epoch": 3.67, "learning_rate": 6.329491100133303e-05, "loss": 2.704, "step": 46810 }, { "epoch": 3.67, "learning_rate": 6.329412687210853e-05, "loss": 2.8064, "step": 46811 }, { "epoch": 3.67, "learning_rate": 6.329334274288403e-05, "loss": 2.8051, "step": 46812 }, { "epoch": 3.67, "learning_rate": 6.329255861365954e-05, "loss": 2.7908, "step": 46813 }, { "epoch": 3.67, "learning_rate": 6.329177448443504e-05, "loss": 2.7706, "step": 46814 }, { "epoch": 3.67, "learning_rate": 6.329099035521054e-05, "loss": 2.6581, "step": 46815 }, { "epoch": 3.67, "learning_rate": 6.329020622598605e-05, "loss": 2.8425, "step": 46816 }, { "epoch": 3.67, "learning_rate": 6.328942209676155e-05, "loss": 2.9654, "step": 46817 }, { "epoch": 3.67, "learning_rate": 6.328863796753706e-05, "loss": 2.7587, "step": 46818 }, { "epoch": 3.67, "learning_rate": 6.328785383831256e-05, "loss": 2.9171, "step": 46819 }, { "epoch": 3.67, "learning_rate": 6.328706970908806e-05, "loss": 2.641, "step": 46820 }, { "epoch": 3.67, "learning_rate": 6.328628557986357e-05, "loss": 2.8516, "step": 46821 }, { "epoch": 3.67, "learning_rate": 6.328550145063907e-05, "loss": 2.8874, "step": 46822 }, { "epoch": 3.67, "learning_rate": 6.328471732141457e-05, "loss": 2.8477, "step": 46823 }, { "epoch": 3.67, "learning_rate": 6.328393319219008e-05, "loss": 3.0165, "step": 46824 }, { "epoch": 3.67, "learning_rate": 6.328314906296557e-05, "loss": 2.6715, "step": 46825 }, { "epoch": 3.67, "learning_rate": 6.328236493374108e-05, "loss": 2.6444, "step": 46826 }, { "epoch": 3.67, "learning_rate": 6.328158080451659e-05, "loss": 2.6927, "step": 46827 }, { "epoch": 3.67, "learning_rate": 6.328079667529209e-05, "loss": 2.6913, "step": 46828 }, { "epoch": 3.67, "learning_rate": 6.32800125460676e-05, "loss": 2.6514, "step": 46829 }, { "epoch": 3.67, "learning_rate": 6.32792284168431e-05, "loss": 2.6688, "step": 46830 }, { "epoch": 3.67, "learning_rate": 6.32784442876186e-05, "loss": 2.6198, "step": 46831 }, { "epoch": 3.67, "learning_rate": 6.327766015839411e-05, "loss": 2.9878, "step": 46832 }, { "epoch": 3.67, "learning_rate": 6.327687602916961e-05, "loss": 2.8731, "step": 46833 }, { "epoch": 3.67, "learning_rate": 6.327609189994511e-05, "loss": 2.7035, "step": 46834 }, { "epoch": 3.67, "learning_rate": 6.327530777072062e-05, "loss": 2.7775, "step": 46835 }, { "epoch": 3.67, "learning_rate": 6.327452364149612e-05, "loss": 2.7542, "step": 46836 }, { "epoch": 3.67, "learning_rate": 6.327373951227163e-05, "loss": 2.5678, "step": 46837 }, { "epoch": 3.67, "learning_rate": 6.327295538304713e-05, "loss": 2.7977, "step": 46838 }, { "epoch": 3.67, "learning_rate": 6.327217125382263e-05, "loss": 2.7768, "step": 46839 }, { "epoch": 3.67, "learning_rate": 6.327138712459814e-05, "loss": 2.7786, "step": 46840 }, { "epoch": 3.67, "learning_rate": 6.327060299537364e-05, "loss": 2.7027, "step": 46841 }, { "epoch": 3.67, "learning_rate": 6.326981886614914e-05, "loss": 2.816, "step": 46842 }, { "epoch": 3.67, "learning_rate": 6.326903473692465e-05, "loss": 2.5811, "step": 46843 }, { "epoch": 3.67, "learning_rate": 6.326825060770015e-05, "loss": 2.7132, "step": 46844 }, { "epoch": 3.67, "learning_rate": 6.326746647847565e-05, "loss": 2.8118, "step": 46845 }, { "epoch": 3.67, "learning_rate": 6.326668234925116e-05, "loss": 2.602, "step": 46846 }, { "epoch": 3.67, "learning_rate": 6.326589822002666e-05, "loss": 2.7119, "step": 46847 }, { "epoch": 3.67, "learning_rate": 6.326511409080217e-05, "loss": 2.9727, "step": 46848 }, { "epoch": 3.67, "learning_rate": 6.326432996157767e-05, "loss": 2.7727, "step": 46849 }, { "epoch": 3.67, "learning_rate": 6.326354583235317e-05, "loss": 2.8427, "step": 46850 }, { "epoch": 3.67, "learning_rate": 6.326276170312868e-05, "loss": 2.5762, "step": 46851 }, { "epoch": 3.67, "learning_rate": 6.326197757390418e-05, "loss": 3.0011, "step": 46852 }, { "epoch": 3.67, "learning_rate": 6.326119344467968e-05, "loss": 2.6779, "step": 46853 }, { "epoch": 3.67, "learning_rate": 6.326040931545519e-05, "loss": 2.6868, "step": 46854 }, { "epoch": 3.67, "learning_rate": 6.325962518623069e-05, "loss": 2.8108, "step": 46855 }, { "epoch": 3.67, "learning_rate": 6.32588410570062e-05, "loss": 2.7201, "step": 46856 }, { "epoch": 3.67, "learning_rate": 6.325805692778171e-05, "loss": 2.7635, "step": 46857 }, { "epoch": 3.67, "learning_rate": 6.325727279855722e-05, "loss": 2.779, "step": 46858 }, { "epoch": 3.67, "learning_rate": 6.325648866933272e-05, "loss": 2.7509, "step": 46859 }, { "epoch": 3.67, "learning_rate": 6.325570454010822e-05, "loss": 2.7874, "step": 46860 }, { "epoch": 3.67, "learning_rate": 6.325492041088371e-05, "loss": 2.6618, "step": 46861 }, { "epoch": 3.67, "learning_rate": 6.325413628165922e-05, "loss": 2.9096, "step": 46862 }, { "epoch": 3.67, "learning_rate": 6.325335215243472e-05, "loss": 2.6533, "step": 46863 }, { "epoch": 3.67, "learning_rate": 6.325256802321022e-05, "loss": 2.7176, "step": 46864 }, { "epoch": 3.67, "learning_rate": 6.325178389398573e-05, "loss": 2.7095, "step": 46865 }, { "epoch": 3.67, "learning_rate": 6.325099976476123e-05, "loss": 2.6574, "step": 46866 }, { "epoch": 3.67, "learning_rate": 6.325021563553674e-05, "loss": 2.5472, "step": 46867 }, { "epoch": 3.68, "learning_rate": 6.324943150631224e-05, "loss": 2.9249, "step": 46868 }, { "epoch": 3.68, "learning_rate": 6.324864737708774e-05, "loss": 2.7496, "step": 46869 }, { "epoch": 3.68, "learning_rate": 6.324786324786325e-05, "loss": 2.7767, "step": 46870 }, { "epoch": 3.68, "learning_rate": 6.324707911863875e-05, "loss": 2.6131, "step": 46871 }, { "epoch": 3.68, "learning_rate": 6.324629498941425e-05, "loss": 2.8146, "step": 46872 }, { "epoch": 3.68, "learning_rate": 6.324551086018976e-05, "loss": 2.5584, "step": 46873 }, { "epoch": 3.68, "learning_rate": 6.324472673096526e-05, "loss": 2.7266, "step": 46874 }, { "epoch": 3.68, "learning_rate": 6.324394260174077e-05, "loss": 2.5649, "step": 46875 }, { "epoch": 3.68, "learning_rate": 6.324315847251627e-05, "loss": 2.6993, "step": 46876 }, { "epoch": 3.68, "learning_rate": 6.324237434329177e-05, "loss": 2.7226, "step": 46877 }, { "epoch": 3.68, "learning_rate": 6.324159021406728e-05, "loss": 2.5665, "step": 46878 }, { "epoch": 3.68, "learning_rate": 6.324080608484278e-05, "loss": 2.7948, "step": 46879 }, { "epoch": 3.68, "learning_rate": 6.324002195561828e-05, "loss": 2.8024, "step": 46880 }, { "epoch": 3.68, "learning_rate": 6.323923782639379e-05, "loss": 2.6939, "step": 46881 }, { "epoch": 3.68, "learning_rate": 6.323845369716929e-05, "loss": 3.0098, "step": 46882 }, { "epoch": 3.68, "learning_rate": 6.323766956794481e-05, "loss": 2.6134, "step": 46883 }, { "epoch": 3.68, "learning_rate": 6.323688543872031e-05, "loss": 2.7545, "step": 46884 }, { "epoch": 3.68, "learning_rate": 6.323610130949582e-05, "loss": 2.6687, "step": 46885 }, { "epoch": 3.68, "learning_rate": 6.323531718027132e-05, "loss": 2.6591, "step": 46886 }, { "epoch": 3.68, "learning_rate": 6.323453305104682e-05, "loss": 2.7329, "step": 46887 }, { "epoch": 3.68, "learning_rate": 6.323374892182233e-05, "loss": 2.7281, "step": 46888 }, { "epoch": 3.68, "learning_rate": 6.323296479259783e-05, "loss": 2.6857, "step": 46889 }, { "epoch": 3.68, "learning_rate": 6.323218066337333e-05, "loss": 2.6567, "step": 46890 }, { "epoch": 3.68, "learning_rate": 6.323139653414884e-05, "loss": 2.8986, "step": 46891 }, { "epoch": 3.68, "learning_rate": 6.323061240492434e-05, "loss": 2.9957, "step": 46892 }, { "epoch": 3.68, "learning_rate": 6.322982827569984e-05, "loss": 2.8091, "step": 46893 }, { "epoch": 3.68, "learning_rate": 6.322904414647535e-05, "loss": 2.7268, "step": 46894 }, { "epoch": 3.68, "learning_rate": 6.322826001725085e-05, "loss": 2.8605, "step": 46895 }, { "epoch": 3.68, "learning_rate": 6.322747588802636e-05, "loss": 2.8986, "step": 46896 }, { "epoch": 3.68, "learning_rate": 6.322669175880185e-05, "loss": 2.8067, "step": 46897 }, { "epoch": 3.68, "learning_rate": 6.322590762957735e-05, "loss": 2.6418, "step": 46898 }, { "epoch": 3.68, "learning_rate": 6.322512350035285e-05, "loss": 2.8531, "step": 46899 }, { "epoch": 3.68, "learning_rate": 6.322433937112836e-05, "loss": 2.6157, "step": 46900 }, { "epoch": 3.68, "learning_rate": 6.322355524190386e-05, "loss": 2.9247, "step": 46901 }, { "epoch": 3.68, "learning_rate": 6.322277111267936e-05, "loss": 2.8148, "step": 46902 }, { "epoch": 3.68, "learning_rate": 6.322198698345487e-05, "loss": 2.6514, "step": 46903 }, { "epoch": 3.68, "learning_rate": 6.322120285423037e-05, "loss": 2.6188, "step": 46904 }, { "epoch": 3.68, "learning_rate": 6.322041872500588e-05, "loss": 2.6226, "step": 46905 }, { "epoch": 3.68, "learning_rate": 6.321963459578138e-05, "loss": 2.758, "step": 46906 }, { "epoch": 3.68, "learning_rate": 6.321885046655688e-05, "loss": 2.6297, "step": 46907 }, { "epoch": 3.68, "learning_rate": 6.32180663373324e-05, "loss": 2.638, "step": 46908 }, { "epoch": 3.68, "learning_rate": 6.32172822081079e-05, "loss": 2.7455, "step": 46909 }, { "epoch": 3.68, "learning_rate": 6.321649807888341e-05, "loss": 2.6051, "step": 46910 }, { "epoch": 3.68, "learning_rate": 6.321571394965891e-05, "loss": 2.7028, "step": 46911 }, { "epoch": 3.68, "learning_rate": 6.321492982043441e-05, "loss": 2.7305, "step": 46912 }, { "epoch": 3.68, "learning_rate": 6.321414569120992e-05, "loss": 2.7618, "step": 46913 }, { "epoch": 3.68, "learning_rate": 6.321336156198542e-05, "loss": 2.7434, "step": 46914 }, { "epoch": 3.68, "learning_rate": 6.321257743276093e-05, "loss": 2.6212, "step": 46915 }, { "epoch": 3.68, "learning_rate": 6.321179330353643e-05, "loss": 2.9034, "step": 46916 }, { "epoch": 3.68, "learning_rate": 6.321100917431193e-05, "loss": 2.7462, "step": 46917 }, { "epoch": 3.68, "learning_rate": 6.321022504508744e-05, "loss": 2.7316, "step": 46918 }, { "epoch": 3.68, "learning_rate": 6.320944091586294e-05, "loss": 2.6503, "step": 46919 }, { "epoch": 3.68, "learning_rate": 6.320865678663844e-05, "loss": 2.776, "step": 46920 }, { "epoch": 3.68, "learning_rate": 6.320787265741395e-05, "loss": 2.8239, "step": 46921 }, { "epoch": 3.68, "learning_rate": 6.320708852818945e-05, "loss": 2.5429, "step": 46922 }, { "epoch": 3.68, "learning_rate": 6.320630439896496e-05, "loss": 2.7427, "step": 46923 }, { "epoch": 3.68, "learning_rate": 6.320552026974046e-05, "loss": 2.5721, "step": 46924 }, { "epoch": 3.68, "learning_rate": 6.320473614051596e-05, "loss": 2.6787, "step": 46925 }, { "epoch": 3.68, "learning_rate": 6.320395201129147e-05, "loss": 2.7135, "step": 46926 }, { "epoch": 3.68, "learning_rate": 6.320316788206697e-05, "loss": 2.7406, "step": 46927 }, { "epoch": 3.68, "learning_rate": 6.320238375284247e-05, "loss": 2.6601, "step": 46928 }, { "epoch": 3.68, "learning_rate": 6.320159962361798e-05, "loss": 2.7382, "step": 46929 }, { "epoch": 3.68, "learning_rate": 6.320081549439348e-05, "loss": 3.0892, "step": 46930 }, { "epoch": 3.68, "learning_rate": 6.320003136516898e-05, "loss": 2.6126, "step": 46931 }, { "epoch": 3.68, "learning_rate": 6.319924723594449e-05, "loss": 2.7657, "step": 46932 }, { "epoch": 3.68, "learning_rate": 6.319846310671998e-05, "loss": 2.7125, "step": 46933 }, { "epoch": 3.68, "learning_rate": 6.31976789774955e-05, "loss": 2.757, "step": 46934 }, { "epoch": 3.68, "learning_rate": 6.3196894848271e-05, "loss": 2.7426, "step": 46935 }, { "epoch": 3.68, "learning_rate": 6.31961107190465e-05, "loss": 2.7686, "step": 46936 }, { "epoch": 3.68, "learning_rate": 6.3195326589822e-05, "loss": 2.6704, "step": 46937 }, { "epoch": 3.68, "learning_rate": 6.319454246059751e-05, "loss": 2.7544, "step": 46938 }, { "epoch": 3.68, "learning_rate": 6.319375833137301e-05, "loss": 2.7493, "step": 46939 }, { "epoch": 3.68, "learning_rate": 6.319297420214852e-05, "loss": 2.7071, "step": 46940 }, { "epoch": 3.68, "learning_rate": 6.319219007292402e-05, "loss": 2.7609, "step": 46941 }, { "epoch": 3.68, "learning_rate": 6.319140594369953e-05, "loss": 2.7718, "step": 46942 }, { "epoch": 3.68, "learning_rate": 6.319062181447503e-05, "loss": 2.7356, "step": 46943 }, { "epoch": 3.68, "learning_rate": 6.318983768525053e-05, "loss": 2.5419, "step": 46944 }, { "epoch": 3.68, "learning_rate": 6.318905355602604e-05, "loss": 2.7117, "step": 46945 }, { "epoch": 3.68, "learning_rate": 6.318826942680154e-05, "loss": 2.7581, "step": 46946 }, { "epoch": 3.68, "learning_rate": 6.318748529757704e-05, "loss": 2.7671, "step": 46947 }, { "epoch": 3.68, "learning_rate": 6.318670116835255e-05, "loss": 2.6762, "step": 46948 }, { "epoch": 3.68, "learning_rate": 6.318591703912805e-05, "loss": 2.7949, "step": 46949 }, { "epoch": 3.68, "learning_rate": 6.318513290990355e-05, "loss": 2.807, "step": 46950 }, { "epoch": 3.68, "learning_rate": 6.318434878067906e-05, "loss": 2.5592, "step": 46951 }, { "epoch": 3.68, "learning_rate": 6.318356465145456e-05, "loss": 2.8573, "step": 46952 }, { "epoch": 3.68, "learning_rate": 6.318278052223007e-05, "loss": 2.7042, "step": 46953 }, { "epoch": 3.68, "learning_rate": 6.318199639300557e-05, "loss": 2.8142, "step": 46954 }, { "epoch": 3.68, "learning_rate": 6.318121226378107e-05, "loss": 2.8339, "step": 46955 }, { "epoch": 3.68, "learning_rate": 6.318042813455658e-05, "loss": 2.7522, "step": 46956 }, { "epoch": 3.68, "learning_rate": 6.317964400533208e-05, "loss": 2.7537, "step": 46957 }, { "epoch": 3.68, "learning_rate": 6.317885987610758e-05, "loss": 2.8888, "step": 46958 }, { "epoch": 3.68, "learning_rate": 6.317807574688309e-05, "loss": 2.6657, "step": 46959 }, { "epoch": 3.68, "learning_rate": 6.317729161765859e-05, "loss": 2.6394, "step": 46960 }, { "epoch": 3.68, "learning_rate": 6.31765074884341e-05, "loss": 2.7505, "step": 46961 }, { "epoch": 3.68, "learning_rate": 6.31757233592096e-05, "loss": 2.6545, "step": 46962 }, { "epoch": 3.68, "learning_rate": 6.31749392299851e-05, "loss": 2.7291, "step": 46963 }, { "epoch": 3.68, "learning_rate": 6.31741551007606e-05, "loss": 2.6933, "step": 46964 }, { "epoch": 3.68, "learning_rate": 6.317337097153612e-05, "loss": 2.6323, "step": 46965 }, { "epoch": 3.68, "learning_rate": 6.317258684231163e-05, "loss": 2.7866, "step": 46966 }, { "epoch": 3.68, "learning_rate": 6.317180271308713e-05, "loss": 2.7238, "step": 46967 }, { "epoch": 3.68, "learning_rate": 6.317101858386263e-05, "loss": 2.573, "step": 46968 }, { "epoch": 3.68, "learning_rate": 6.317023445463812e-05, "loss": 2.6267, "step": 46969 }, { "epoch": 3.68, "learning_rate": 6.316945032541363e-05, "loss": 2.7467, "step": 46970 }, { "epoch": 3.68, "learning_rate": 6.316866619618913e-05, "loss": 2.8328, "step": 46971 }, { "epoch": 3.68, "learning_rate": 6.316788206696464e-05, "loss": 2.9134, "step": 46972 }, { "epoch": 3.68, "learning_rate": 6.316709793774014e-05, "loss": 2.5833, "step": 46973 }, { "epoch": 3.68, "learning_rate": 6.316631380851564e-05, "loss": 2.6191, "step": 46974 }, { "epoch": 3.68, "learning_rate": 6.316552967929115e-05, "loss": 2.7086, "step": 46975 }, { "epoch": 3.68, "learning_rate": 6.316474555006665e-05, "loss": 2.7083, "step": 46976 }, { "epoch": 3.68, "learning_rate": 6.316396142084215e-05, "loss": 2.7002, "step": 46977 }, { "epoch": 3.68, "learning_rate": 6.316317729161766e-05, "loss": 2.5988, "step": 46978 }, { "epoch": 3.68, "learning_rate": 6.316239316239316e-05, "loss": 2.7153, "step": 46979 }, { "epoch": 3.68, "learning_rate": 6.316160903316866e-05, "loss": 2.6976, "step": 46980 }, { "epoch": 3.68, "learning_rate": 6.316082490394417e-05, "loss": 2.7176, "step": 46981 }, { "epoch": 3.68, "learning_rate": 6.316004077471967e-05, "loss": 2.8244, "step": 46982 }, { "epoch": 3.68, "learning_rate": 6.315925664549518e-05, "loss": 2.7973, "step": 46983 }, { "epoch": 3.68, "learning_rate": 6.315847251627068e-05, "loss": 2.65, "step": 46984 }, { "epoch": 3.68, "learning_rate": 6.315768838704618e-05, "loss": 2.6502, "step": 46985 }, { "epoch": 3.68, "learning_rate": 6.315690425782169e-05, "loss": 2.7994, "step": 46986 }, { "epoch": 3.68, "learning_rate": 6.315612012859719e-05, "loss": 2.6722, "step": 46987 }, { "epoch": 3.68, "learning_rate": 6.31553359993727e-05, "loss": 2.8098, "step": 46988 }, { "epoch": 3.68, "learning_rate": 6.31545518701482e-05, "loss": 2.8917, "step": 46989 }, { "epoch": 3.68, "learning_rate": 6.31537677409237e-05, "loss": 2.6403, "step": 46990 }, { "epoch": 3.68, "learning_rate": 6.315298361169922e-05, "loss": 2.7053, "step": 46991 }, { "epoch": 3.68, "learning_rate": 6.315219948247472e-05, "loss": 2.5709, "step": 46992 }, { "epoch": 3.68, "learning_rate": 6.315141535325023e-05, "loss": 2.6455, "step": 46993 }, { "epoch": 3.68, "learning_rate": 6.315063122402573e-05, "loss": 2.4959, "step": 46994 }, { "epoch": 3.69, "learning_rate": 6.314984709480123e-05, "loss": 2.39, "step": 46995 }, { "epoch": 3.69, "learning_rate": 6.314906296557674e-05, "loss": 2.7159, "step": 46996 }, { "epoch": 3.69, "learning_rate": 6.314827883635224e-05, "loss": 2.6159, "step": 46997 }, { "epoch": 3.69, "learning_rate": 6.314749470712774e-05, "loss": 2.6133, "step": 46998 }, { "epoch": 3.69, "learning_rate": 6.314671057790325e-05, "loss": 2.7278, "step": 46999 }, { "epoch": 3.69, "learning_rate": 6.314592644867875e-05, "loss": 2.7141, "step": 47000 }, { "epoch": 3.69, "eval_loss": 2.8924410343170166, "eval_runtime": 217.7767, "eval_samples_per_second": 52.035, "eval_steps_per_second": 3.256, "step": 47000 }, { "epoch": 3.69, "learning_rate": 6.314514231945426e-05, "loss": 2.6168, "step": 47001 }, { "epoch": 3.69, "learning_rate": 6.314435819022976e-05, "loss": 2.9128, "step": 47002 }, { "epoch": 3.69, "learning_rate": 6.314357406100526e-05, "loss": 2.4492, "step": 47003 }, { "epoch": 3.69, "learning_rate": 6.314278993178077e-05, "loss": 2.9882, "step": 47004 }, { "epoch": 3.69, "learning_rate": 6.314200580255626e-05, "loss": 2.7364, "step": 47005 }, { "epoch": 3.69, "learning_rate": 6.314122167333176e-05, "loss": 2.4951, "step": 47006 }, { "epoch": 3.69, "learning_rate": 6.314043754410726e-05, "loss": 2.7209, "step": 47007 }, { "epoch": 3.69, "learning_rate": 6.313965341488277e-05, "loss": 2.8358, "step": 47008 }, { "epoch": 3.69, "learning_rate": 6.313886928565827e-05, "loss": 2.96, "step": 47009 }, { "epoch": 3.69, "learning_rate": 6.313808515643378e-05, "loss": 2.8147, "step": 47010 }, { "epoch": 3.69, "learning_rate": 6.313730102720928e-05, "loss": 2.733, "step": 47011 }, { "epoch": 3.69, "learning_rate": 6.313651689798478e-05, "loss": 3.0429, "step": 47012 }, { "epoch": 3.69, "learning_rate": 6.313573276876029e-05, "loss": 2.5146, "step": 47013 }, { "epoch": 3.69, "learning_rate": 6.313494863953579e-05, "loss": 3.1445, "step": 47014 }, { "epoch": 3.69, "learning_rate": 6.31341645103113e-05, "loss": 2.8211, "step": 47015 }, { "epoch": 3.69, "learning_rate": 6.313338038108681e-05, "loss": 2.7534, "step": 47016 }, { "epoch": 3.69, "learning_rate": 6.313259625186231e-05, "loss": 2.8277, "step": 47017 }, { "epoch": 3.69, "learning_rate": 6.313181212263782e-05, "loss": 2.7198, "step": 47018 }, { "epoch": 3.69, "learning_rate": 6.313102799341332e-05, "loss": 2.631, "step": 47019 }, { "epoch": 3.69, "learning_rate": 6.313024386418883e-05, "loss": 2.5947, "step": 47020 }, { "epoch": 3.69, "learning_rate": 6.312945973496433e-05, "loss": 2.8167, "step": 47021 }, { "epoch": 3.69, "learning_rate": 6.312867560573983e-05, "loss": 2.812, "step": 47022 }, { "epoch": 3.69, "learning_rate": 6.312789147651534e-05, "loss": 2.5947, "step": 47023 }, { "epoch": 3.69, "learning_rate": 6.312710734729084e-05, "loss": 2.4943, "step": 47024 }, { "epoch": 3.69, "learning_rate": 6.312632321806634e-05, "loss": 2.5929, "step": 47025 }, { "epoch": 3.69, "learning_rate": 6.312553908884185e-05, "loss": 2.7002, "step": 47026 }, { "epoch": 3.69, "learning_rate": 6.312475495961735e-05, "loss": 2.7731, "step": 47027 }, { "epoch": 3.69, "learning_rate": 6.312397083039285e-05, "loss": 2.6462, "step": 47028 }, { "epoch": 3.69, "learning_rate": 6.312318670116836e-05, "loss": 2.7639, "step": 47029 }, { "epoch": 3.69, "learning_rate": 6.312240257194386e-05, "loss": 2.6849, "step": 47030 }, { "epoch": 3.69, "learning_rate": 6.312161844271937e-05, "loss": 2.8762, "step": 47031 }, { "epoch": 3.69, "learning_rate": 6.312083431349487e-05, "loss": 2.8581, "step": 47032 }, { "epoch": 3.69, "learning_rate": 6.312005018427037e-05, "loss": 2.8026, "step": 47033 }, { "epoch": 3.69, "learning_rate": 6.311926605504588e-05, "loss": 2.9127, "step": 47034 }, { "epoch": 3.69, "learning_rate": 6.311848192582138e-05, "loss": 2.8429, "step": 47035 }, { "epoch": 3.69, "learning_rate": 6.311769779659688e-05, "loss": 2.8212, "step": 47036 }, { "epoch": 3.69, "learning_rate": 6.311691366737239e-05, "loss": 2.9625, "step": 47037 }, { "epoch": 3.69, "learning_rate": 6.311612953814789e-05, "loss": 2.8669, "step": 47038 }, { "epoch": 3.69, "learning_rate": 6.31153454089234e-05, "loss": 2.6903, "step": 47039 }, { "epoch": 3.69, "learning_rate": 6.31145612796989e-05, "loss": 2.7211, "step": 47040 }, { "epoch": 3.69, "learning_rate": 6.311377715047439e-05, "loss": 2.8886, "step": 47041 }, { "epoch": 3.69, "learning_rate": 6.31129930212499e-05, "loss": 2.8363, "step": 47042 }, { "epoch": 3.69, "learning_rate": 6.311220889202541e-05, "loss": 2.7465, "step": 47043 }, { "epoch": 3.69, "learning_rate": 6.311142476280091e-05, "loss": 2.6203, "step": 47044 }, { "epoch": 3.69, "learning_rate": 6.311064063357642e-05, "loss": 2.7527, "step": 47045 }, { "epoch": 3.69, "learning_rate": 6.310985650435192e-05, "loss": 2.4177, "step": 47046 }, { "epoch": 3.69, "learning_rate": 6.310907237512742e-05, "loss": 2.9462, "step": 47047 }, { "epoch": 3.69, "learning_rate": 6.310828824590293e-05, "loss": 2.8098, "step": 47048 }, { "epoch": 3.69, "learning_rate": 6.310750411667843e-05, "loss": 2.7543, "step": 47049 }, { "epoch": 3.69, "learning_rate": 6.310671998745394e-05, "loss": 2.564, "step": 47050 }, { "epoch": 3.69, "learning_rate": 6.310593585822944e-05, "loss": 2.8405, "step": 47051 }, { "epoch": 3.69, "learning_rate": 6.310515172900494e-05, "loss": 2.7371, "step": 47052 }, { "epoch": 3.69, "learning_rate": 6.310436759978045e-05, "loss": 2.818, "step": 47053 }, { "epoch": 3.69, "learning_rate": 6.310358347055595e-05, "loss": 2.6734, "step": 47054 }, { "epoch": 3.69, "learning_rate": 6.310279934133145e-05, "loss": 2.6566, "step": 47055 }, { "epoch": 3.69, "learning_rate": 6.310201521210696e-05, "loss": 2.8394, "step": 47056 }, { "epoch": 3.69, "learning_rate": 6.310123108288246e-05, "loss": 2.7725, "step": 47057 }, { "epoch": 3.69, "learning_rate": 6.310044695365797e-05, "loss": 2.531, "step": 47058 }, { "epoch": 3.69, "learning_rate": 6.309966282443347e-05, "loss": 2.7104, "step": 47059 }, { "epoch": 3.69, "learning_rate": 6.309887869520897e-05, "loss": 2.8789, "step": 47060 }, { "epoch": 3.69, "learning_rate": 6.309809456598448e-05, "loss": 3.0705, "step": 47061 }, { "epoch": 3.69, "learning_rate": 6.309731043675998e-05, "loss": 2.5579, "step": 47062 }, { "epoch": 3.69, "learning_rate": 6.309652630753548e-05, "loss": 2.9364, "step": 47063 }, { "epoch": 3.69, "learning_rate": 6.309574217831099e-05, "loss": 2.7104, "step": 47064 }, { "epoch": 3.69, "learning_rate": 6.309495804908649e-05, "loss": 2.7278, "step": 47065 }, { "epoch": 3.69, "learning_rate": 6.3094173919862e-05, "loss": 2.6689, "step": 47066 }, { "epoch": 3.69, "learning_rate": 6.30933897906375e-05, "loss": 2.9443, "step": 47067 }, { "epoch": 3.69, "learning_rate": 6.3092605661413e-05, "loss": 2.7891, "step": 47068 }, { "epoch": 3.69, "learning_rate": 6.30918215321885e-05, "loss": 2.6436, "step": 47069 }, { "epoch": 3.69, "learning_rate": 6.309103740296401e-05, "loss": 2.8658, "step": 47070 }, { "epoch": 3.69, "learning_rate": 6.309025327373951e-05, "loss": 3.1101, "step": 47071 }, { "epoch": 3.69, "learning_rate": 6.308946914451502e-05, "loss": 2.7164, "step": 47072 }, { "epoch": 3.69, "learning_rate": 6.308868501529053e-05, "loss": 2.6339, "step": 47073 }, { "epoch": 3.69, "learning_rate": 6.308790088606604e-05, "loss": 2.7205, "step": 47074 }, { "epoch": 3.69, "learning_rate": 6.308711675684154e-05, "loss": 2.8602, "step": 47075 }, { "epoch": 3.69, "learning_rate": 6.308633262761704e-05, "loss": 2.9912, "step": 47076 }, { "epoch": 3.69, "learning_rate": 6.308554849839254e-05, "loss": 2.7958, "step": 47077 }, { "epoch": 3.69, "learning_rate": 6.308476436916804e-05, "loss": 2.9185, "step": 47078 }, { "epoch": 3.69, "learning_rate": 6.308398023994354e-05, "loss": 2.6505, "step": 47079 }, { "epoch": 3.69, "learning_rate": 6.308319611071905e-05, "loss": 2.8251, "step": 47080 }, { "epoch": 3.69, "learning_rate": 6.308241198149455e-05, "loss": 2.6322, "step": 47081 }, { "epoch": 3.69, "learning_rate": 6.308162785227005e-05, "loss": 2.645, "step": 47082 }, { "epoch": 3.69, "learning_rate": 6.308084372304556e-05, "loss": 2.7552, "step": 47083 }, { "epoch": 3.69, "learning_rate": 6.308005959382106e-05, "loss": 2.7579, "step": 47084 }, { "epoch": 3.69, "learning_rate": 6.307927546459656e-05, "loss": 2.5565, "step": 47085 }, { "epoch": 3.69, "learning_rate": 6.307849133537207e-05, "loss": 2.5631, "step": 47086 }, { "epoch": 3.69, "learning_rate": 6.307770720614757e-05, "loss": 2.6299, "step": 47087 }, { "epoch": 3.69, "learning_rate": 6.307692307692308e-05, "loss": 2.4751, "step": 47088 }, { "epoch": 3.69, "learning_rate": 6.307613894769858e-05, "loss": 2.7349, "step": 47089 }, { "epoch": 3.69, "learning_rate": 6.307535481847408e-05, "loss": 2.5522, "step": 47090 }, { "epoch": 3.69, "learning_rate": 6.307457068924959e-05, "loss": 2.7155, "step": 47091 }, { "epoch": 3.69, "learning_rate": 6.307378656002509e-05, "loss": 2.6176, "step": 47092 }, { "epoch": 3.69, "learning_rate": 6.30730024308006e-05, "loss": 2.9918, "step": 47093 }, { "epoch": 3.69, "learning_rate": 6.30722183015761e-05, "loss": 2.7183, "step": 47094 }, { "epoch": 3.69, "learning_rate": 6.30714341723516e-05, "loss": 2.5748, "step": 47095 }, { "epoch": 3.69, "learning_rate": 6.30706500431271e-05, "loss": 2.8642, "step": 47096 }, { "epoch": 3.69, "learning_rate": 6.306986591390261e-05, "loss": 2.6633, "step": 47097 }, { "epoch": 3.69, "learning_rate": 6.306908178467811e-05, "loss": 2.7391, "step": 47098 }, { "epoch": 3.69, "learning_rate": 6.306829765545363e-05, "loss": 2.8906, "step": 47099 }, { "epoch": 3.69, "learning_rate": 6.306751352622913e-05, "loss": 2.7367, "step": 47100 }, { "epoch": 3.69, "learning_rate": 6.306672939700464e-05, "loss": 2.8206, "step": 47101 }, { "epoch": 3.69, "learning_rate": 6.306594526778014e-05, "loss": 2.6319, "step": 47102 }, { "epoch": 3.69, "learning_rate": 6.306516113855564e-05, "loss": 2.6737, "step": 47103 }, { "epoch": 3.69, "learning_rate": 6.306437700933115e-05, "loss": 2.6044, "step": 47104 }, { "epoch": 3.69, "learning_rate": 6.306359288010665e-05, "loss": 2.7274, "step": 47105 }, { "epoch": 3.69, "learning_rate": 6.306280875088216e-05, "loss": 2.6553, "step": 47106 }, { "epoch": 3.69, "learning_rate": 6.306202462165766e-05, "loss": 2.5832, "step": 47107 }, { "epoch": 3.69, "learning_rate": 6.306124049243316e-05, "loss": 2.611, "step": 47108 }, { "epoch": 3.69, "learning_rate": 6.306045636320867e-05, "loss": 3.0103, "step": 47109 }, { "epoch": 3.69, "learning_rate": 6.305967223398417e-05, "loss": 2.7006, "step": 47110 }, { "epoch": 3.69, "learning_rate": 6.305888810475967e-05, "loss": 2.8302, "step": 47111 }, { "epoch": 3.69, "learning_rate": 6.305810397553518e-05, "loss": 2.6238, "step": 47112 }, { "epoch": 3.69, "learning_rate": 6.305731984631067e-05, "loss": 2.7431, "step": 47113 }, { "epoch": 3.69, "learning_rate": 6.305653571708617e-05, "loss": 2.8986, "step": 47114 }, { "epoch": 3.69, "learning_rate": 6.305575158786167e-05, "loss": 2.6453, "step": 47115 }, { "epoch": 3.69, "learning_rate": 6.305496745863718e-05, "loss": 2.6667, "step": 47116 }, { "epoch": 3.69, "learning_rate": 6.305418332941268e-05, "loss": 2.699, "step": 47117 }, { "epoch": 3.69, "learning_rate": 6.305339920018819e-05, "loss": 2.6293, "step": 47118 }, { "epoch": 3.69, "learning_rate": 6.305261507096369e-05, "loss": 2.8533, "step": 47119 }, { "epoch": 3.69, "learning_rate": 6.30518309417392e-05, "loss": 2.5226, "step": 47120 }, { "epoch": 3.69, "learning_rate": 6.30510468125147e-05, "loss": 2.6129, "step": 47121 }, { "epoch": 3.69, "learning_rate": 6.30502626832902e-05, "loss": 2.6781, "step": 47122 }, { "epoch": 3.7, "learning_rate": 6.30494785540657e-05, "loss": 2.7483, "step": 47123 }, { "epoch": 3.7, "learning_rate": 6.304869442484122e-05, "loss": 2.5333, "step": 47124 }, { "epoch": 3.7, "learning_rate": 6.304791029561673e-05, "loss": 2.7388, "step": 47125 }, { "epoch": 3.7, "learning_rate": 6.304712616639223e-05, "loss": 2.7686, "step": 47126 }, { "epoch": 3.7, "learning_rate": 6.304634203716773e-05, "loss": 2.7113, "step": 47127 }, { "epoch": 3.7, "learning_rate": 6.304555790794324e-05, "loss": 2.5421, "step": 47128 }, { "epoch": 3.7, "learning_rate": 6.304477377871874e-05, "loss": 2.4976, "step": 47129 }, { "epoch": 3.7, "learning_rate": 6.304398964949424e-05, "loss": 2.7353, "step": 47130 }, { "epoch": 3.7, "learning_rate": 6.304320552026975e-05, "loss": 2.8125, "step": 47131 }, { "epoch": 3.7, "learning_rate": 6.304242139104525e-05, "loss": 2.68, "step": 47132 }, { "epoch": 3.7, "learning_rate": 6.304163726182075e-05, "loss": 2.8092, "step": 47133 }, { "epoch": 3.7, "learning_rate": 6.304085313259626e-05, "loss": 2.4923, "step": 47134 }, { "epoch": 3.7, "learning_rate": 6.304006900337176e-05, "loss": 2.935, "step": 47135 }, { "epoch": 3.7, "learning_rate": 6.303928487414727e-05, "loss": 2.8398, "step": 47136 }, { "epoch": 3.7, "learning_rate": 6.303850074492277e-05, "loss": 2.663, "step": 47137 }, { "epoch": 3.7, "learning_rate": 6.303771661569827e-05, "loss": 2.7114, "step": 47138 }, { "epoch": 3.7, "learning_rate": 6.303693248647378e-05, "loss": 2.7221, "step": 47139 }, { "epoch": 3.7, "learning_rate": 6.303614835724928e-05, "loss": 2.7616, "step": 47140 }, { "epoch": 3.7, "learning_rate": 6.303536422802478e-05, "loss": 2.6653, "step": 47141 }, { "epoch": 3.7, "learning_rate": 6.303458009880029e-05, "loss": 2.6869, "step": 47142 }, { "epoch": 3.7, "learning_rate": 6.303379596957579e-05, "loss": 2.5524, "step": 47143 }, { "epoch": 3.7, "learning_rate": 6.30330118403513e-05, "loss": 2.7387, "step": 47144 }, { "epoch": 3.7, "learning_rate": 6.30322277111268e-05, "loss": 2.6942, "step": 47145 }, { "epoch": 3.7, "learning_rate": 6.30314435819023e-05, "loss": 2.7404, "step": 47146 }, { "epoch": 3.7, "learning_rate": 6.30306594526778e-05, "loss": 2.6152, "step": 47147 }, { "epoch": 3.7, "learning_rate": 6.302987532345331e-05, "loss": 2.7088, "step": 47148 }, { "epoch": 3.7, "learning_rate": 6.30290911942288e-05, "loss": 2.5401, "step": 47149 }, { "epoch": 3.7, "learning_rate": 6.302830706500432e-05, "loss": 2.6486, "step": 47150 }, { "epoch": 3.7, "learning_rate": 6.302752293577982e-05, "loss": 2.6015, "step": 47151 }, { "epoch": 3.7, "learning_rate": 6.302673880655532e-05, "loss": 2.7997, "step": 47152 }, { "epoch": 3.7, "learning_rate": 6.302595467733083e-05, "loss": 2.7593, "step": 47153 }, { "epoch": 3.7, "learning_rate": 6.302517054810633e-05, "loss": 2.6227, "step": 47154 }, { "epoch": 3.7, "learning_rate": 6.302438641888184e-05, "loss": 2.8524, "step": 47155 }, { "epoch": 3.7, "learning_rate": 6.302360228965734e-05, "loss": 2.9248, "step": 47156 }, { "epoch": 3.7, "learning_rate": 6.302281816043284e-05, "loss": 2.6494, "step": 47157 }, { "epoch": 3.7, "learning_rate": 6.302203403120835e-05, "loss": 2.7508, "step": 47158 }, { "epoch": 3.7, "learning_rate": 6.302124990198385e-05, "loss": 2.6921, "step": 47159 }, { "epoch": 3.7, "learning_rate": 6.302046577275935e-05, "loss": 2.694, "step": 47160 }, { "epoch": 3.7, "learning_rate": 6.301968164353486e-05, "loss": 2.5922, "step": 47161 }, { "epoch": 3.7, "learning_rate": 6.301889751431036e-05, "loss": 2.8112, "step": 47162 }, { "epoch": 3.7, "learning_rate": 6.301811338508586e-05, "loss": 2.7407, "step": 47163 }, { "epoch": 3.7, "learning_rate": 6.301732925586137e-05, "loss": 2.6655, "step": 47164 }, { "epoch": 3.7, "learning_rate": 6.301654512663687e-05, "loss": 2.5826, "step": 47165 }, { "epoch": 3.7, "learning_rate": 6.301576099741238e-05, "loss": 2.558, "step": 47166 }, { "epoch": 3.7, "learning_rate": 6.301497686818788e-05, "loss": 2.9155, "step": 47167 }, { "epoch": 3.7, "learning_rate": 6.301419273896338e-05, "loss": 2.838, "step": 47168 }, { "epoch": 3.7, "learning_rate": 6.301340860973889e-05, "loss": 2.6955, "step": 47169 }, { "epoch": 3.7, "learning_rate": 6.301262448051439e-05, "loss": 2.9629, "step": 47170 }, { "epoch": 3.7, "learning_rate": 6.30118403512899e-05, "loss": 2.4582, "step": 47171 }, { "epoch": 3.7, "learning_rate": 6.30110562220654e-05, "loss": 2.7393, "step": 47172 }, { "epoch": 3.7, "learning_rate": 6.30102720928409e-05, "loss": 2.7447, "step": 47173 }, { "epoch": 3.7, "learning_rate": 6.30094879636164e-05, "loss": 2.9133, "step": 47174 }, { "epoch": 3.7, "learning_rate": 6.300870383439191e-05, "loss": 2.8419, "step": 47175 }, { "epoch": 3.7, "learning_rate": 6.300791970516741e-05, "loss": 2.5532, "step": 47176 }, { "epoch": 3.7, "learning_rate": 6.300713557594292e-05, "loss": 2.5734, "step": 47177 }, { "epoch": 3.7, "learning_rate": 6.300635144671842e-05, "loss": 2.8742, "step": 47178 }, { "epoch": 3.7, "learning_rate": 6.300556731749392e-05, "loss": 2.8905, "step": 47179 }, { "epoch": 3.7, "learning_rate": 6.300478318826943e-05, "loss": 2.898, "step": 47180 }, { "epoch": 3.7, "learning_rate": 6.300399905904494e-05, "loss": 2.7011, "step": 47181 }, { "epoch": 3.7, "learning_rate": 6.300321492982045e-05, "loss": 2.7649, "step": 47182 }, { "epoch": 3.7, "learning_rate": 6.300243080059595e-05, "loss": 2.7125, "step": 47183 }, { "epoch": 3.7, "learning_rate": 6.300164667137146e-05, "loss": 2.6673, "step": 47184 }, { "epoch": 3.7, "learning_rate": 6.300086254214695e-05, "loss": 2.4844, "step": 47185 }, { "epoch": 3.7, "learning_rate": 6.300007841292245e-05, "loss": 2.631, "step": 47186 }, { "epoch": 3.7, "learning_rate": 6.299929428369795e-05, "loss": 2.4895, "step": 47187 }, { "epoch": 3.7, "learning_rate": 6.299851015447346e-05, "loss": 2.8134, "step": 47188 }, { "epoch": 3.7, "learning_rate": 6.299772602524896e-05, "loss": 2.8668, "step": 47189 }, { "epoch": 3.7, "learning_rate": 6.299694189602446e-05, "loss": 2.4619, "step": 47190 }, { "epoch": 3.7, "learning_rate": 6.299615776679997e-05, "loss": 2.9216, "step": 47191 }, { "epoch": 3.7, "learning_rate": 6.299537363757547e-05, "loss": 2.8059, "step": 47192 }, { "epoch": 3.7, "learning_rate": 6.299458950835098e-05, "loss": 2.7304, "step": 47193 }, { "epoch": 3.7, "learning_rate": 6.299380537912648e-05, "loss": 2.6138, "step": 47194 }, { "epoch": 3.7, "learning_rate": 6.299302124990198e-05, "loss": 2.6962, "step": 47195 }, { "epoch": 3.7, "learning_rate": 6.299223712067749e-05, "loss": 2.8109, "step": 47196 }, { "epoch": 3.7, "learning_rate": 6.299145299145299e-05, "loss": 2.8091, "step": 47197 }, { "epoch": 3.7, "learning_rate": 6.29906688622285e-05, "loss": 2.7254, "step": 47198 }, { "epoch": 3.7, "learning_rate": 6.2989884733004e-05, "loss": 2.9339, "step": 47199 }, { "epoch": 3.7, "learning_rate": 6.29891006037795e-05, "loss": 2.7418, "step": 47200 }, { "epoch": 3.7, "learning_rate": 6.2988316474555e-05, "loss": 3.1726, "step": 47201 }, { "epoch": 3.7, "learning_rate": 6.298753234533051e-05, "loss": 2.932, "step": 47202 }, { "epoch": 3.7, "learning_rate": 6.298674821610601e-05, "loss": 2.6994, "step": 47203 }, { "epoch": 3.7, "learning_rate": 6.298596408688152e-05, "loss": 2.7795, "step": 47204 }, { "epoch": 3.7, "learning_rate": 6.298517995765702e-05, "loss": 2.6638, "step": 47205 }, { "epoch": 3.7, "learning_rate": 6.298439582843252e-05, "loss": 2.8304, "step": 47206 }, { "epoch": 3.7, "learning_rate": 6.298361169920804e-05, "loss": 2.6332, "step": 47207 }, { "epoch": 3.7, "learning_rate": 6.298282756998354e-05, "loss": 2.5674, "step": 47208 }, { "epoch": 3.7, "learning_rate": 6.298204344075905e-05, "loss": 2.9023, "step": 47209 }, { "epoch": 3.7, "learning_rate": 6.298125931153455e-05, "loss": 2.7263, "step": 47210 }, { "epoch": 3.7, "learning_rate": 6.298047518231006e-05, "loss": 2.6783, "step": 47211 }, { "epoch": 3.7, "learning_rate": 6.297969105308556e-05, "loss": 2.6335, "step": 47212 }, { "epoch": 3.7, "learning_rate": 6.297890692386106e-05, "loss": 2.5741, "step": 47213 }, { "epoch": 3.7, "learning_rate": 6.297812279463657e-05, "loss": 2.6881, "step": 47214 }, { "epoch": 3.7, "learning_rate": 6.297733866541207e-05, "loss": 2.5978, "step": 47215 }, { "epoch": 3.7, "learning_rate": 6.297655453618757e-05, "loss": 2.7, "step": 47216 }, { "epoch": 3.7, "learning_rate": 6.297577040696308e-05, "loss": 2.6801, "step": 47217 }, { "epoch": 3.7, "learning_rate": 6.297498627773858e-05, "loss": 2.7435, "step": 47218 }, { "epoch": 3.7, "learning_rate": 6.297420214851408e-05, "loss": 2.6581, "step": 47219 }, { "epoch": 3.7, "learning_rate": 6.297341801928959e-05, "loss": 2.434, "step": 47220 }, { "epoch": 3.7, "learning_rate": 6.297263389006508e-05, "loss": 2.6263, "step": 47221 }, { "epoch": 3.7, "learning_rate": 6.297184976084058e-05, "loss": 2.7666, "step": 47222 }, { "epoch": 3.7, "learning_rate": 6.297106563161609e-05, "loss": 2.5983, "step": 47223 }, { "epoch": 3.7, "learning_rate": 6.297028150239159e-05, "loss": 2.6234, "step": 47224 }, { "epoch": 3.7, "learning_rate": 6.296949737316709e-05, "loss": 2.8616, "step": 47225 }, { "epoch": 3.7, "learning_rate": 6.29687132439426e-05, "loss": 2.9379, "step": 47226 }, { "epoch": 3.7, "learning_rate": 6.29679291147181e-05, "loss": 2.6923, "step": 47227 }, { "epoch": 3.7, "learning_rate": 6.29671449854936e-05, "loss": 2.8025, "step": 47228 }, { "epoch": 3.7, "learning_rate": 6.296636085626911e-05, "loss": 2.7664, "step": 47229 }, { "epoch": 3.7, "learning_rate": 6.296557672704461e-05, "loss": 2.8405, "step": 47230 }, { "epoch": 3.7, "learning_rate": 6.296479259782012e-05, "loss": 2.7661, "step": 47231 }, { "epoch": 3.7, "learning_rate": 6.296400846859563e-05, "loss": 2.8385, "step": 47232 }, { "epoch": 3.7, "learning_rate": 6.296322433937114e-05, "loss": 2.6913, "step": 47233 }, { "epoch": 3.7, "learning_rate": 6.296244021014664e-05, "loss": 2.635, "step": 47234 }, { "epoch": 3.7, "learning_rate": 6.296165608092214e-05, "loss": 2.7525, "step": 47235 }, { "epoch": 3.7, "learning_rate": 6.296087195169765e-05, "loss": 2.5082, "step": 47236 }, { "epoch": 3.7, "learning_rate": 6.296008782247315e-05, "loss": 2.825, "step": 47237 }, { "epoch": 3.7, "learning_rate": 6.295930369324865e-05, "loss": 2.8731, "step": 47238 }, { "epoch": 3.7, "learning_rate": 6.295851956402416e-05, "loss": 2.6736, "step": 47239 }, { "epoch": 3.7, "learning_rate": 6.295773543479966e-05, "loss": 2.8731, "step": 47240 }, { "epoch": 3.7, "learning_rate": 6.295695130557517e-05, "loss": 2.7282, "step": 47241 }, { "epoch": 3.7, "learning_rate": 6.295616717635067e-05, "loss": 2.5656, "step": 47242 }, { "epoch": 3.7, "learning_rate": 6.295538304712617e-05, "loss": 2.7298, "step": 47243 }, { "epoch": 3.7, "learning_rate": 6.295459891790168e-05, "loss": 2.7897, "step": 47244 }, { "epoch": 3.7, "learning_rate": 6.295381478867718e-05, "loss": 2.6597, "step": 47245 }, { "epoch": 3.7, "learning_rate": 6.295303065945268e-05, "loss": 2.8278, "step": 47246 }, { "epoch": 3.7, "learning_rate": 6.295224653022819e-05, "loss": 2.9747, "step": 47247 }, { "epoch": 3.7, "learning_rate": 6.295146240100369e-05, "loss": 2.7554, "step": 47248 }, { "epoch": 3.7, "learning_rate": 6.29506782717792e-05, "loss": 2.725, "step": 47249 }, { "epoch": 3.71, "learning_rate": 6.29498941425547e-05, "loss": 2.7279, "step": 47250 }, { "epoch": 3.71, "learning_rate": 6.29491100133302e-05, "loss": 2.7797, "step": 47251 }, { "epoch": 3.71, "learning_rate": 6.29483258841057e-05, "loss": 2.7131, "step": 47252 }, { "epoch": 3.71, "learning_rate": 6.294754175488121e-05, "loss": 2.7143, "step": 47253 }, { "epoch": 3.71, "learning_rate": 6.294675762565671e-05, "loss": 2.7811, "step": 47254 }, { "epoch": 3.71, "learning_rate": 6.294597349643222e-05, "loss": 2.6077, "step": 47255 }, { "epoch": 3.71, "learning_rate": 6.294518936720772e-05, "loss": 2.4882, "step": 47256 }, { "epoch": 3.71, "learning_rate": 6.294440523798321e-05, "loss": 2.7535, "step": 47257 }, { "epoch": 3.71, "learning_rate": 6.294362110875873e-05, "loss": 2.8716, "step": 47258 }, { "epoch": 3.71, "learning_rate": 6.294283697953423e-05, "loss": 2.6367, "step": 47259 }, { "epoch": 3.71, "learning_rate": 6.294205285030974e-05, "loss": 2.6981, "step": 47260 }, { "epoch": 3.71, "learning_rate": 6.294126872108524e-05, "loss": 2.5457, "step": 47261 }, { "epoch": 3.71, "learning_rate": 6.294048459186074e-05, "loss": 2.6738, "step": 47262 }, { "epoch": 3.71, "learning_rate": 6.293970046263625e-05, "loss": 2.6223, "step": 47263 }, { "epoch": 3.71, "learning_rate": 6.293891633341175e-05, "loss": 2.8373, "step": 47264 }, { "epoch": 3.71, "learning_rate": 6.293813220418725e-05, "loss": 2.816, "step": 47265 }, { "epoch": 3.71, "learning_rate": 6.293734807496276e-05, "loss": 2.811, "step": 47266 }, { "epoch": 3.71, "learning_rate": 6.293656394573826e-05, "loss": 2.7956, "step": 47267 }, { "epoch": 3.71, "learning_rate": 6.293577981651376e-05, "loss": 3.0832, "step": 47268 }, { "epoch": 3.71, "learning_rate": 6.293499568728927e-05, "loss": 2.6767, "step": 47269 }, { "epoch": 3.71, "learning_rate": 6.293421155806477e-05, "loss": 2.743, "step": 47270 }, { "epoch": 3.71, "learning_rate": 6.293342742884028e-05, "loss": 2.7079, "step": 47271 }, { "epoch": 3.71, "learning_rate": 6.293264329961578e-05, "loss": 2.7578, "step": 47272 }, { "epoch": 3.71, "learning_rate": 6.293185917039128e-05, "loss": 2.8129, "step": 47273 }, { "epoch": 3.71, "learning_rate": 6.293107504116679e-05, "loss": 2.6808, "step": 47274 }, { "epoch": 3.71, "learning_rate": 6.293029091194229e-05, "loss": 2.7702, "step": 47275 }, { "epoch": 3.71, "learning_rate": 6.29295067827178e-05, "loss": 2.7358, "step": 47276 }, { "epoch": 3.71, "learning_rate": 6.29287226534933e-05, "loss": 2.5478, "step": 47277 }, { "epoch": 3.71, "learning_rate": 6.29279385242688e-05, "loss": 2.6805, "step": 47278 }, { "epoch": 3.71, "learning_rate": 6.29271543950443e-05, "loss": 2.6031, "step": 47279 }, { "epoch": 3.71, "learning_rate": 6.292637026581981e-05, "loss": 2.7169, "step": 47280 }, { "epoch": 3.71, "learning_rate": 6.292558613659531e-05, "loss": 2.7037, "step": 47281 }, { "epoch": 3.71, "learning_rate": 6.292480200737082e-05, "loss": 2.6862, "step": 47282 }, { "epoch": 3.71, "learning_rate": 6.292401787814632e-05, "loss": 2.6838, "step": 47283 }, { "epoch": 3.71, "learning_rate": 6.292323374892182e-05, "loss": 2.9168, "step": 47284 }, { "epoch": 3.71, "learning_rate": 6.292244961969733e-05, "loss": 2.6272, "step": 47285 }, { "epoch": 3.71, "learning_rate": 6.292166549047283e-05, "loss": 2.7544, "step": 47286 }, { "epoch": 3.71, "learning_rate": 6.292088136124833e-05, "loss": 2.8957, "step": 47287 }, { "epoch": 3.71, "learning_rate": 6.292009723202384e-05, "loss": 2.709, "step": 47288 }, { "epoch": 3.71, "learning_rate": 6.291931310279936e-05, "loss": 2.5668, "step": 47289 }, { "epoch": 3.71, "learning_rate": 6.291852897357486e-05, "loss": 2.9566, "step": 47290 }, { "epoch": 3.71, "learning_rate": 6.291774484435036e-05, "loss": 2.5838, "step": 47291 }, { "epoch": 3.71, "learning_rate": 6.291696071512587e-05, "loss": 2.882, "step": 47292 }, { "epoch": 3.71, "learning_rate": 6.291617658590136e-05, "loss": 2.8253, "step": 47293 }, { "epoch": 3.71, "learning_rate": 6.291539245667686e-05, "loss": 2.868, "step": 47294 }, { "epoch": 3.71, "learning_rate": 6.291460832745236e-05, "loss": 2.6324, "step": 47295 }, { "epoch": 3.71, "learning_rate": 6.291382419822787e-05, "loss": 2.6227, "step": 47296 }, { "epoch": 3.71, "learning_rate": 6.291304006900337e-05, "loss": 2.5759, "step": 47297 }, { "epoch": 3.71, "learning_rate": 6.291225593977887e-05, "loss": 2.667, "step": 47298 }, { "epoch": 3.71, "learning_rate": 6.291147181055438e-05, "loss": 2.5089, "step": 47299 }, { "epoch": 3.71, "learning_rate": 6.291068768132988e-05, "loss": 2.8597, "step": 47300 }, { "epoch": 3.71, "learning_rate": 6.290990355210539e-05, "loss": 2.6875, "step": 47301 }, { "epoch": 3.71, "learning_rate": 6.290911942288089e-05, "loss": 2.5944, "step": 47302 }, { "epoch": 3.71, "learning_rate": 6.29083352936564e-05, "loss": 2.7075, "step": 47303 }, { "epoch": 3.71, "learning_rate": 6.29075511644319e-05, "loss": 2.8605, "step": 47304 }, { "epoch": 3.71, "learning_rate": 6.29067670352074e-05, "loss": 2.6051, "step": 47305 }, { "epoch": 3.71, "learning_rate": 6.29059829059829e-05, "loss": 2.6568, "step": 47306 }, { "epoch": 3.71, "learning_rate": 6.290519877675841e-05, "loss": 2.728, "step": 47307 }, { "epoch": 3.71, "learning_rate": 6.290441464753391e-05, "loss": 2.6194, "step": 47308 }, { "epoch": 3.71, "learning_rate": 6.290363051830942e-05, "loss": 2.7245, "step": 47309 }, { "epoch": 3.71, "learning_rate": 6.290284638908492e-05, "loss": 2.5616, "step": 47310 }, { "epoch": 3.71, "learning_rate": 6.290206225986042e-05, "loss": 3.0517, "step": 47311 }, { "epoch": 3.71, "learning_rate": 6.290127813063593e-05, "loss": 2.567, "step": 47312 }, { "epoch": 3.71, "learning_rate": 6.290049400141143e-05, "loss": 2.709, "step": 47313 }, { "epoch": 3.71, "learning_rate": 6.289970987218693e-05, "loss": 2.6322, "step": 47314 }, { "epoch": 3.71, "learning_rate": 6.289892574296245e-05, "loss": 2.8164, "step": 47315 }, { "epoch": 3.71, "learning_rate": 6.289814161373795e-05, "loss": 2.7952, "step": 47316 }, { "epoch": 3.71, "learning_rate": 6.289735748451346e-05, "loss": 2.6409, "step": 47317 }, { "epoch": 3.71, "learning_rate": 6.289657335528896e-05, "loss": 2.7483, "step": 47318 }, { "epoch": 3.71, "learning_rate": 6.289578922606447e-05, "loss": 2.8386, "step": 47319 }, { "epoch": 3.71, "learning_rate": 6.289500509683997e-05, "loss": 2.8909, "step": 47320 }, { "epoch": 3.71, "learning_rate": 6.289422096761547e-05, "loss": 2.7208, "step": 47321 }, { "epoch": 3.71, "learning_rate": 6.289343683839098e-05, "loss": 2.9501, "step": 47322 }, { "epoch": 3.71, "learning_rate": 6.289265270916648e-05, "loss": 2.712, "step": 47323 }, { "epoch": 3.71, "learning_rate": 6.289186857994198e-05, "loss": 2.8153, "step": 47324 }, { "epoch": 3.71, "learning_rate": 6.289108445071749e-05, "loss": 2.7105, "step": 47325 }, { "epoch": 3.71, "learning_rate": 6.289030032149299e-05, "loss": 2.5538, "step": 47326 }, { "epoch": 3.71, "learning_rate": 6.28895161922685e-05, "loss": 2.621, "step": 47327 }, { "epoch": 3.71, "learning_rate": 6.2888732063044e-05, "loss": 2.6486, "step": 47328 }, { "epoch": 3.71, "learning_rate": 6.288794793381949e-05, "loss": 2.7577, "step": 47329 }, { "epoch": 3.71, "learning_rate": 6.288716380459499e-05, "loss": 2.7365, "step": 47330 }, { "epoch": 3.71, "learning_rate": 6.28863796753705e-05, "loss": 2.7602, "step": 47331 }, { "epoch": 3.71, "learning_rate": 6.2885595546146e-05, "loss": 2.6713, "step": 47332 }, { "epoch": 3.71, "learning_rate": 6.28848114169215e-05, "loss": 2.605, "step": 47333 }, { "epoch": 3.71, "learning_rate": 6.288402728769701e-05, "loss": 2.6157, "step": 47334 }, { "epoch": 3.71, "learning_rate": 6.288324315847251e-05, "loss": 2.9053, "step": 47335 }, { "epoch": 3.71, "learning_rate": 6.288245902924801e-05, "loss": 2.6717, "step": 47336 }, { "epoch": 3.71, "learning_rate": 6.288167490002352e-05, "loss": 2.5751, "step": 47337 }, { "epoch": 3.71, "learning_rate": 6.288089077079902e-05, "loss": 2.8749, "step": 47338 }, { "epoch": 3.71, "learning_rate": 6.288010664157453e-05, "loss": 2.5707, "step": 47339 }, { "epoch": 3.71, "learning_rate": 6.287932251235004e-05, "loss": 2.7894, "step": 47340 }, { "epoch": 3.71, "learning_rate": 6.287853838312555e-05, "loss": 2.6462, "step": 47341 }, { "epoch": 3.71, "learning_rate": 6.287775425390105e-05, "loss": 2.6424, "step": 47342 }, { "epoch": 3.71, "learning_rate": 6.287697012467655e-05, "loss": 2.8228, "step": 47343 }, { "epoch": 3.71, "learning_rate": 6.287618599545206e-05, "loss": 2.7607, "step": 47344 }, { "epoch": 3.71, "learning_rate": 6.287540186622756e-05, "loss": 2.8263, "step": 47345 }, { "epoch": 3.71, "learning_rate": 6.287461773700307e-05, "loss": 2.8364, "step": 47346 }, { "epoch": 3.71, "learning_rate": 6.287383360777857e-05, "loss": 2.6627, "step": 47347 }, { "epoch": 3.71, "learning_rate": 6.287304947855407e-05, "loss": 3.1487, "step": 47348 }, { "epoch": 3.71, "learning_rate": 6.287226534932958e-05, "loss": 2.8673, "step": 47349 }, { "epoch": 3.71, "learning_rate": 6.287148122010508e-05, "loss": 2.8796, "step": 47350 }, { "epoch": 3.71, "learning_rate": 6.287069709088058e-05, "loss": 2.7035, "step": 47351 }, { "epoch": 3.71, "learning_rate": 6.286991296165609e-05, "loss": 2.6971, "step": 47352 }, { "epoch": 3.71, "learning_rate": 6.286912883243159e-05, "loss": 2.7197, "step": 47353 }, { "epoch": 3.71, "learning_rate": 6.28683447032071e-05, "loss": 2.9274, "step": 47354 }, { "epoch": 3.71, "learning_rate": 6.28675605739826e-05, "loss": 2.8519, "step": 47355 }, { "epoch": 3.71, "learning_rate": 6.28667764447581e-05, "loss": 2.61, "step": 47356 }, { "epoch": 3.71, "learning_rate": 6.28659923155336e-05, "loss": 2.7277, "step": 47357 }, { "epoch": 3.71, "learning_rate": 6.286520818630911e-05, "loss": 2.8349, "step": 47358 }, { "epoch": 3.71, "learning_rate": 6.286442405708461e-05, "loss": 2.5822, "step": 47359 }, { "epoch": 3.71, "learning_rate": 6.286363992786012e-05, "loss": 2.7812, "step": 47360 }, { "epoch": 3.71, "learning_rate": 6.286285579863562e-05, "loss": 2.6732, "step": 47361 }, { "epoch": 3.71, "learning_rate": 6.286207166941112e-05, "loss": 2.8546, "step": 47362 }, { "epoch": 3.71, "learning_rate": 6.286128754018663e-05, "loss": 2.7475, "step": 47363 }, { "epoch": 3.71, "learning_rate": 6.286050341096213e-05, "loss": 2.6466, "step": 47364 }, { "epoch": 3.71, "learning_rate": 6.285971928173762e-05, "loss": 3.0735, "step": 47365 }, { "epoch": 3.71, "learning_rate": 6.285893515251314e-05, "loss": 2.5756, "step": 47366 }, { "epoch": 3.71, "learning_rate": 6.285815102328864e-05, "loss": 2.7298, "step": 47367 }, { "epoch": 3.71, "learning_rate": 6.285736689406415e-05, "loss": 2.7441, "step": 47368 }, { "epoch": 3.71, "learning_rate": 6.285658276483965e-05, "loss": 2.6412, "step": 47369 }, { "epoch": 3.71, "learning_rate": 6.285579863561515e-05, "loss": 2.5906, "step": 47370 }, { "epoch": 3.71, "learning_rate": 6.285501450639066e-05, "loss": 2.7097, "step": 47371 }, { "epoch": 3.71, "learning_rate": 6.285423037716616e-05, "loss": 2.8949, "step": 47372 }, { "epoch": 3.71, "learning_rate": 6.285344624794166e-05, "loss": 2.6998, "step": 47373 }, { "epoch": 3.71, "learning_rate": 6.285266211871717e-05, "loss": 2.7407, "step": 47374 }, { "epoch": 3.71, "learning_rate": 6.285187798949267e-05, "loss": 2.7924, "step": 47375 }, { "epoch": 3.71, "learning_rate": 6.285109386026818e-05, "loss": 2.6307, "step": 47376 }, { "epoch": 3.71, "learning_rate": 6.285030973104368e-05, "loss": 2.5248, "step": 47377 }, { "epoch": 3.72, "learning_rate": 6.284952560181918e-05, "loss": 2.8404, "step": 47378 }, { "epoch": 3.72, "learning_rate": 6.284874147259469e-05, "loss": 2.605, "step": 47379 }, { "epoch": 3.72, "learning_rate": 6.284795734337019e-05, "loss": 2.6324, "step": 47380 }, { "epoch": 3.72, "learning_rate": 6.28471732141457e-05, "loss": 2.6674, "step": 47381 }, { "epoch": 3.72, "learning_rate": 6.28463890849212e-05, "loss": 2.4921, "step": 47382 }, { "epoch": 3.72, "learning_rate": 6.28456049556967e-05, "loss": 2.6188, "step": 47383 }, { "epoch": 3.72, "learning_rate": 6.28448208264722e-05, "loss": 2.7396, "step": 47384 }, { "epoch": 3.72, "learning_rate": 6.284403669724771e-05, "loss": 2.7503, "step": 47385 }, { "epoch": 3.72, "learning_rate": 6.284325256802321e-05, "loss": 2.789, "step": 47386 }, { "epoch": 3.72, "learning_rate": 6.284246843879872e-05, "loss": 2.8798, "step": 47387 }, { "epoch": 3.72, "learning_rate": 6.284168430957422e-05, "loss": 2.5365, "step": 47388 }, { "epoch": 3.72, "learning_rate": 6.284090018034972e-05, "loss": 2.6112, "step": 47389 }, { "epoch": 3.72, "learning_rate": 6.284011605112523e-05, "loss": 2.6746, "step": 47390 }, { "epoch": 3.72, "learning_rate": 6.283933192190073e-05, "loss": 2.5985, "step": 47391 }, { "epoch": 3.72, "learning_rate": 6.283854779267623e-05, "loss": 2.801, "step": 47392 }, { "epoch": 3.72, "learning_rate": 6.283776366345174e-05, "loss": 2.8942, "step": 47393 }, { "epoch": 3.72, "learning_rate": 6.283697953422724e-05, "loss": 2.6957, "step": 47394 }, { "epoch": 3.72, "learning_rate": 6.283619540500275e-05, "loss": 2.4897, "step": 47395 }, { "epoch": 3.72, "learning_rate": 6.283541127577825e-05, "loss": 2.6645, "step": 47396 }, { "epoch": 3.72, "learning_rate": 6.283462714655377e-05, "loss": 2.7616, "step": 47397 }, { "epoch": 3.72, "learning_rate": 6.283384301732927e-05, "loss": 2.8461, "step": 47398 }, { "epoch": 3.72, "learning_rate": 6.283305888810477e-05, "loss": 2.6713, "step": 47399 }, { "epoch": 3.72, "learning_rate": 6.283227475888028e-05, "loss": 2.8771, "step": 47400 }, { "epoch": 3.72, "learning_rate": 6.283149062965577e-05, "loss": 2.6757, "step": 47401 }, { "epoch": 3.72, "learning_rate": 6.283070650043127e-05, "loss": 2.6466, "step": 47402 }, { "epoch": 3.72, "learning_rate": 6.282992237120677e-05, "loss": 2.6855, "step": 47403 }, { "epoch": 3.72, "learning_rate": 6.282913824198228e-05, "loss": 2.9346, "step": 47404 }, { "epoch": 3.72, "learning_rate": 6.282835411275778e-05, "loss": 2.6292, "step": 47405 }, { "epoch": 3.72, "learning_rate": 6.282756998353329e-05, "loss": 2.6796, "step": 47406 }, { "epoch": 3.72, "learning_rate": 6.282678585430879e-05, "loss": 2.7717, "step": 47407 }, { "epoch": 3.72, "learning_rate": 6.282600172508429e-05, "loss": 2.5561, "step": 47408 }, { "epoch": 3.72, "learning_rate": 6.28252175958598e-05, "loss": 2.7111, "step": 47409 }, { "epoch": 3.72, "learning_rate": 6.28244334666353e-05, "loss": 2.8241, "step": 47410 }, { "epoch": 3.72, "learning_rate": 6.28236493374108e-05, "loss": 2.6748, "step": 47411 }, { "epoch": 3.72, "learning_rate": 6.282286520818631e-05, "loss": 2.6625, "step": 47412 }, { "epoch": 3.72, "learning_rate": 6.282208107896181e-05, "loss": 2.8521, "step": 47413 }, { "epoch": 3.72, "learning_rate": 6.282129694973732e-05, "loss": 2.5767, "step": 47414 }, { "epoch": 3.72, "learning_rate": 6.282051282051282e-05, "loss": 2.7021, "step": 47415 }, { "epoch": 3.72, "learning_rate": 6.281972869128832e-05, "loss": 2.8215, "step": 47416 }, { "epoch": 3.72, "learning_rate": 6.281894456206383e-05, "loss": 2.7016, "step": 47417 }, { "epoch": 3.72, "learning_rate": 6.281816043283933e-05, "loss": 2.7913, "step": 47418 }, { "epoch": 3.72, "learning_rate": 6.281737630361483e-05, "loss": 2.8071, "step": 47419 }, { "epoch": 3.72, "learning_rate": 6.281659217439034e-05, "loss": 2.5051, "step": 47420 }, { "epoch": 3.72, "learning_rate": 6.281580804516584e-05, "loss": 2.5992, "step": 47421 }, { "epoch": 3.72, "learning_rate": 6.281502391594134e-05, "loss": 2.7012, "step": 47422 }, { "epoch": 3.72, "learning_rate": 6.281423978671686e-05, "loss": 2.8, "step": 47423 }, { "epoch": 3.72, "learning_rate": 6.281345565749237e-05, "loss": 2.8133, "step": 47424 }, { "epoch": 3.72, "learning_rate": 6.281267152826787e-05, "loss": 2.6946, "step": 47425 }, { "epoch": 3.72, "learning_rate": 6.281188739904337e-05, "loss": 2.564, "step": 47426 }, { "epoch": 3.72, "learning_rate": 6.281110326981888e-05, "loss": 2.9619, "step": 47427 }, { "epoch": 3.72, "learning_rate": 6.281031914059438e-05, "loss": 2.7462, "step": 47428 }, { "epoch": 3.72, "learning_rate": 6.280953501136988e-05, "loss": 2.8001, "step": 47429 }, { "epoch": 3.72, "learning_rate": 6.280875088214539e-05, "loss": 2.6478, "step": 47430 }, { "epoch": 3.72, "learning_rate": 6.280796675292089e-05, "loss": 2.6445, "step": 47431 }, { "epoch": 3.72, "learning_rate": 6.28071826236964e-05, "loss": 2.452, "step": 47432 }, { "epoch": 3.72, "learning_rate": 6.28063984944719e-05, "loss": 2.6223, "step": 47433 }, { "epoch": 3.72, "learning_rate": 6.28056143652474e-05, "loss": 2.6928, "step": 47434 }, { "epoch": 3.72, "learning_rate": 6.28048302360229e-05, "loss": 2.7, "step": 47435 }, { "epoch": 3.72, "learning_rate": 6.280404610679841e-05, "loss": 2.6846, "step": 47436 }, { "epoch": 3.72, "learning_rate": 6.28032619775739e-05, "loss": 2.9292, "step": 47437 }, { "epoch": 3.72, "learning_rate": 6.28024778483494e-05, "loss": 2.8462, "step": 47438 }, { "epoch": 3.72, "learning_rate": 6.280169371912491e-05, "loss": 2.9715, "step": 47439 }, { "epoch": 3.72, "learning_rate": 6.280090958990041e-05, "loss": 2.6125, "step": 47440 }, { "epoch": 3.72, "learning_rate": 6.280012546067591e-05, "loss": 2.8275, "step": 47441 }, { "epoch": 3.72, "learning_rate": 6.279934133145142e-05, "loss": 2.5663, "step": 47442 }, { "epoch": 3.72, "learning_rate": 6.279855720222692e-05, "loss": 2.8157, "step": 47443 }, { "epoch": 3.72, "learning_rate": 6.279777307300243e-05, "loss": 2.803, "step": 47444 }, { "epoch": 3.72, "learning_rate": 6.279698894377793e-05, "loss": 2.5767, "step": 47445 }, { "epoch": 3.72, "learning_rate": 6.279620481455343e-05, "loss": 2.5785, "step": 47446 }, { "epoch": 3.72, "learning_rate": 6.279542068532894e-05, "loss": 3.0431, "step": 47447 }, { "epoch": 3.72, "learning_rate": 6.279463655610444e-05, "loss": 2.5263, "step": 47448 }, { "epoch": 3.72, "learning_rate": 6.279385242687996e-05, "loss": 2.8212, "step": 47449 }, { "epoch": 3.72, "learning_rate": 6.279306829765546e-05, "loss": 2.6324, "step": 47450 }, { "epoch": 3.72, "learning_rate": 6.279228416843096e-05, "loss": 2.941, "step": 47451 }, { "epoch": 3.72, "learning_rate": 6.279150003920647e-05, "loss": 2.7546, "step": 47452 }, { "epoch": 3.72, "learning_rate": 6.279071590998197e-05, "loss": 2.9239, "step": 47453 }, { "epoch": 3.72, "learning_rate": 6.278993178075748e-05, "loss": 2.6852, "step": 47454 }, { "epoch": 3.72, "learning_rate": 6.278914765153298e-05, "loss": 2.5811, "step": 47455 }, { "epoch": 3.72, "learning_rate": 6.278836352230848e-05, "loss": 2.7019, "step": 47456 }, { "epoch": 3.72, "learning_rate": 6.278757939308399e-05, "loss": 2.6432, "step": 47457 }, { "epoch": 3.72, "learning_rate": 6.278679526385949e-05, "loss": 2.6533, "step": 47458 }, { "epoch": 3.72, "learning_rate": 6.2786011134635e-05, "loss": 2.8091, "step": 47459 }, { "epoch": 3.72, "learning_rate": 6.27852270054105e-05, "loss": 2.7559, "step": 47460 }, { "epoch": 3.72, "learning_rate": 6.2784442876186e-05, "loss": 2.9004, "step": 47461 }, { "epoch": 3.72, "learning_rate": 6.27836587469615e-05, "loss": 2.6431, "step": 47462 }, { "epoch": 3.72, "learning_rate": 6.278287461773701e-05, "loss": 2.7447, "step": 47463 }, { "epoch": 3.72, "learning_rate": 6.278209048851251e-05, "loss": 3.0048, "step": 47464 }, { "epoch": 3.72, "learning_rate": 6.278130635928802e-05, "loss": 2.7169, "step": 47465 }, { "epoch": 3.72, "learning_rate": 6.278052223006352e-05, "loss": 2.7301, "step": 47466 }, { "epoch": 3.72, "learning_rate": 6.277973810083902e-05, "loss": 2.5649, "step": 47467 }, { "epoch": 3.72, "learning_rate": 6.277895397161453e-05, "loss": 2.6156, "step": 47468 }, { "epoch": 3.72, "learning_rate": 6.277816984239003e-05, "loss": 2.6402, "step": 47469 }, { "epoch": 3.72, "learning_rate": 6.277738571316553e-05, "loss": 2.9093, "step": 47470 }, { "epoch": 3.72, "learning_rate": 6.277660158394104e-05, "loss": 2.7488, "step": 47471 }, { "epoch": 3.72, "learning_rate": 6.277581745471654e-05, "loss": 2.6592, "step": 47472 }, { "epoch": 3.72, "learning_rate": 6.277503332549203e-05, "loss": 2.5528, "step": 47473 }, { "epoch": 3.72, "learning_rate": 6.277424919626755e-05, "loss": 2.7211, "step": 47474 }, { "epoch": 3.72, "learning_rate": 6.277346506704305e-05, "loss": 2.8197, "step": 47475 }, { "epoch": 3.72, "learning_rate": 6.277268093781856e-05, "loss": 2.7897, "step": 47476 }, { "epoch": 3.72, "learning_rate": 6.277189680859406e-05, "loss": 2.5849, "step": 47477 }, { "epoch": 3.72, "learning_rate": 6.277111267936956e-05, "loss": 2.8, "step": 47478 }, { "epoch": 3.72, "learning_rate": 6.277032855014507e-05, "loss": 2.6954, "step": 47479 }, { "epoch": 3.72, "learning_rate": 6.276954442092057e-05, "loss": 2.5442, "step": 47480 }, { "epoch": 3.72, "learning_rate": 6.276876029169608e-05, "loss": 2.5788, "step": 47481 }, { "epoch": 3.72, "learning_rate": 6.276797616247158e-05, "loss": 2.674, "step": 47482 }, { "epoch": 3.72, "learning_rate": 6.276719203324708e-05, "loss": 2.8337, "step": 47483 }, { "epoch": 3.72, "learning_rate": 6.276640790402259e-05, "loss": 2.5476, "step": 47484 }, { "epoch": 3.72, "learning_rate": 6.276562377479809e-05, "loss": 2.5621, "step": 47485 }, { "epoch": 3.72, "learning_rate": 6.27648396455736e-05, "loss": 2.6082, "step": 47486 }, { "epoch": 3.72, "learning_rate": 6.27640555163491e-05, "loss": 2.6416, "step": 47487 }, { "epoch": 3.72, "learning_rate": 6.27632713871246e-05, "loss": 2.7774, "step": 47488 }, { "epoch": 3.72, "learning_rate": 6.27624872579001e-05, "loss": 2.6571, "step": 47489 }, { "epoch": 3.72, "learning_rate": 6.276170312867561e-05, "loss": 2.8508, "step": 47490 }, { "epoch": 3.72, "learning_rate": 6.276091899945111e-05, "loss": 2.7691, "step": 47491 }, { "epoch": 3.72, "learning_rate": 6.276013487022662e-05, "loss": 2.62, "step": 47492 }, { "epoch": 3.72, "learning_rate": 6.275935074100212e-05, "loss": 2.4342, "step": 47493 }, { "epoch": 3.72, "learning_rate": 6.275856661177762e-05, "loss": 2.6905, "step": 47494 }, { "epoch": 3.72, "learning_rate": 6.275778248255313e-05, "loss": 2.6408, "step": 47495 }, { "epoch": 3.72, "learning_rate": 6.275699835332863e-05, "loss": 2.8422, "step": 47496 }, { "epoch": 3.72, "learning_rate": 6.275621422410413e-05, "loss": 2.5594, "step": 47497 }, { "epoch": 3.72, "learning_rate": 6.275543009487964e-05, "loss": 2.8822, "step": 47498 }, { "epoch": 3.72, "learning_rate": 6.275464596565514e-05, "loss": 2.6593, "step": 47499 }, { "epoch": 3.72, "learning_rate": 6.275386183643064e-05, "loss": 2.7896, "step": 47500 }, { "epoch": 3.72, "learning_rate": 6.275307770720615e-05, "loss": 2.7884, "step": 47501 }, { "epoch": 3.72, "learning_rate": 6.275229357798165e-05, "loss": 2.5178, "step": 47502 }, { "epoch": 3.72, "learning_rate": 6.275150944875716e-05, "loss": 2.7407, "step": 47503 }, { "epoch": 3.72, "learning_rate": 6.275072531953266e-05, "loss": 2.636, "step": 47504 }, { "epoch": 3.73, "learning_rate": 6.274994119030818e-05, "loss": 2.8424, "step": 47505 }, { "epoch": 3.73, "learning_rate": 6.274915706108368e-05, "loss": 2.5722, "step": 47506 }, { "epoch": 3.73, "learning_rate": 6.274837293185918e-05, "loss": 2.6806, "step": 47507 }, { "epoch": 3.73, "learning_rate": 6.274758880263469e-05, "loss": 2.475, "step": 47508 }, { "epoch": 3.73, "learning_rate": 6.274680467341018e-05, "loss": 2.7403, "step": 47509 }, { "epoch": 3.73, "learning_rate": 6.274602054418568e-05, "loss": 2.6194, "step": 47510 }, { "epoch": 3.73, "learning_rate": 6.274523641496119e-05, "loss": 2.5815, "step": 47511 }, { "epoch": 3.73, "learning_rate": 6.274445228573669e-05, "loss": 2.5775, "step": 47512 }, { "epoch": 3.73, "learning_rate": 6.274366815651219e-05, "loss": 2.7457, "step": 47513 }, { "epoch": 3.73, "learning_rate": 6.27428840272877e-05, "loss": 2.7368, "step": 47514 }, { "epoch": 3.73, "learning_rate": 6.27420998980632e-05, "loss": 2.7533, "step": 47515 }, { "epoch": 3.73, "learning_rate": 6.27413157688387e-05, "loss": 2.6997, "step": 47516 }, { "epoch": 3.73, "learning_rate": 6.274053163961421e-05, "loss": 2.5911, "step": 47517 }, { "epoch": 3.73, "learning_rate": 6.273974751038971e-05, "loss": 2.7164, "step": 47518 }, { "epoch": 3.73, "learning_rate": 6.273896338116521e-05, "loss": 2.8103, "step": 47519 }, { "epoch": 3.73, "learning_rate": 6.273817925194072e-05, "loss": 2.4835, "step": 47520 }, { "epoch": 3.73, "learning_rate": 6.273739512271622e-05, "loss": 2.7476, "step": 47521 }, { "epoch": 3.73, "learning_rate": 6.273661099349173e-05, "loss": 2.8192, "step": 47522 }, { "epoch": 3.73, "learning_rate": 6.273582686426723e-05, "loss": 2.5942, "step": 47523 }, { "epoch": 3.73, "learning_rate": 6.273504273504273e-05, "loss": 2.6615, "step": 47524 }, { "epoch": 3.73, "learning_rate": 6.273425860581824e-05, "loss": 2.6275, "step": 47525 }, { "epoch": 3.73, "learning_rate": 6.273347447659374e-05, "loss": 2.7525, "step": 47526 }, { "epoch": 3.73, "learning_rate": 6.273269034736924e-05, "loss": 2.8099, "step": 47527 }, { "epoch": 3.73, "learning_rate": 6.273190621814475e-05, "loss": 2.5315, "step": 47528 }, { "epoch": 3.73, "learning_rate": 6.273112208892025e-05, "loss": 2.8016, "step": 47529 }, { "epoch": 3.73, "learning_rate": 6.273033795969576e-05, "loss": 2.9451, "step": 47530 }, { "epoch": 3.73, "learning_rate": 6.272955383047127e-05, "loss": 2.6531, "step": 47531 }, { "epoch": 3.73, "learning_rate": 6.272876970124678e-05, "loss": 2.8399, "step": 47532 }, { "epoch": 3.73, "learning_rate": 6.272798557202228e-05, "loss": 2.9815, "step": 47533 }, { "epoch": 3.73, "learning_rate": 6.272720144279778e-05, "loss": 2.7718, "step": 47534 }, { "epoch": 3.73, "learning_rate": 6.272641731357329e-05, "loss": 2.5907, "step": 47535 }, { "epoch": 3.73, "learning_rate": 6.272563318434879e-05, "loss": 2.5927, "step": 47536 }, { "epoch": 3.73, "learning_rate": 6.27248490551243e-05, "loss": 2.6421, "step": 47537 }, { "epoch": 3.73, "learning_rate": 6.27240649258998e-05, "loss": 2.5382, "step": 47538 }, { "epoch": 3.73, "learning_rate": 6.27232807966753e-05, "loss": 2.7138, "step": 47539 }, { "epoch": 3.73, "learning_rate": 6.27224966674508e-05, "loss": 2.8354, "step": 47540 }, { "epoch": 3.73, "learning_rate": 6.272171253822631e-05, "loss": 2.8052, "step": 47541 }, { "epoch": 3.73, "learning_rate": 6.272092840900181e-05, "loss": 2.6939, "step": 47542 }, { "epoch": 3.73, "learning_rate": 6.272014427977732e-05, "loss": 2.9731, "step": 47543 }, { "epoch": 3.73, "learning_rate": 6.271936015055282e-05, "loss": 2.6743, "step": 47544 }, { "epoch": 3.73, "learning_rate": 6.271857602132831e-05, "loss": 2.6908, "step": 47545 }, { "epoch": 3.73, "learning_rate": 6.271779189210381e-05, "loss": 2.6785, "step": 47546 }, { "epoch": 3.73, "learning_rate": 6.271700776287932e-05, "loss": 2.5408, "step": 47547 }, { "epoch": 3.73, "learning_rate": 6.271622363365482e-05, "loss": 2.6203, "step": 47548 }, { "epoch": 3.73, "learning_rate": 6.271543950443033e-05, "loss": 2.6035, "step": 47549 }, { "epoch": 3.73, "learning_rate": 6.271465537520583e-05, "loss": 2.6778, "step": 47550 }, { "epoch": 3.73, "learning_rate": 6.271387124598133e-05, "loss": 2.7378, "step": 47551 }, { "epoch": 3.73, "learning_rate": 6.271308711675684e-05, "loss": 2.6763, "step": 47552 }, { "epoch": 3.73, "learning_rate": 6.271230298753234e-05, "loss": 2.7851, "step": 47553 }, { "epoch": 3.73, "learning_rate": 6.271151885830784e-05, "loss": 2.562, "step": 47554 }, { "epoch": 3.73, "learning_rate": 6.271073472908335e-05, "loss": 2.9468, "step": 47555 }, { "epoch": 3.73, "learning_rate": 6.270995059985885e-05, "loss": 2.6143, "step": 47556 }, { "epoch": 3.73, "learning_rate": 6.270916647063437e-05, "loss": 2.8066, "step": 47557 }, { "epoch": 3.73, "learning_rate": 6.270838234140987e-05, "loss": 2.5707, "step": 47558 }, { "epoch": 3.73, "learning_rate": 6.270759821218538e-05, "loss": 2.7406, "step": 47559 }, { "epoch": 3.73, "learning_rate": 6.270681408296088e-05, "loss": 2.6813, "step": 47560 }, { "epoch": 3.73, "learning_rate": 6.270602995373638e-05, "loss": 2.8609, "step": 47561 }, { "epoch": 3.73, "learning_rate": 6.270524582451189e-05, "loss": 2.9038, "step": 47562 }, { "epoch": 3.73, "learning_rate": 6.270446169528739e-05, "loss": 2.6585, "step": 47563 }, { "epoch": 3.73, "learning_rate": 6.27036775660629e-05, "loss": 2.8936, "step": 47564 }, { "epoch": 3.73, "learning_rate": 6.27028934368384e-05, "loss": 2.5944, "step": 47565 }, { "epoch": 3.73, "learning_rate": 6.27021093076139e-05, "loss": 2.4552, "step": 47566 }, { "epoch": 3.73, "learning_rate": 6.27013251783894e-05, "loss": 2.8259, "step": 47567 }, { "epoch": 3.73, "learning_rate": 6.270054104916491e-05, "loss": 2.998, "step": 47568 }, { "epoch": 3.73, "learning_rate": 6.269975691994041e-05, "loss": 2.6696, "step": 47569 }, { "epoch": 3.73, "learning_rate": 6.269897279071592e-05, "loss": 2.597, "step": 47570 }, { "epoch": 3.73, "learning_rate": 6.269818866149142e-05, "loss": 2.7387, "step": 47571 }, { "epoch": 3.73, "learning_rate": 6.269740453226692e-05, "loss": 2.5528, "step": 47572 }, { "epoch": 3.73, "learning_rate": 6.269662040304243e-05, "loss": 2.6435, "step": 47573 }, { "epoch": 3.73, "learning_rate": 6.269583627381793e-05, "loss": 2.7689, "step": 47574 }, { "epoch": 3.73, "learning_rate": 6.269505214459343e-05, "loss": 2.57, "step": 47575 }, { "epoch": 3.73, "learning_rate": 6.269426801536894e-05, "loss": 2.7422, "step": 47576 }, { "epoch": 3.73, "learning_rate": 6.269348388614444e-05, "loss": 2.6026, "step": 47577 }, { "epoch": 3.73, "learning_rate": 6.269269975691995e-05, "loss": 2.9601, "step": 47578 }, { "epoch": 3.73, "learning_rate": 6.269191562769545e-05, "loss": 2.6145, "step": 47579 }, { "epoch": 3.73, "learning_rate": 6.269113149847095e-05, "loss": 2.6745, "step": 47580 }, { "epoch": 3.73, "learning_rate": 6.269034736924644e-05, "loss": 2.6545, "step": 47581 }, { "epoch": 3.73, "learning_rate": 6.268956324002196e-05, "loss": 2.6994, "step": 47582 }, { "epoch": 3.73, "learning_rate": 6.268877911079746e-05, "loss": 2.6755, "step": 47583 }, { "epoch": 3.73, "learning_rate": 6.268799498157297e-05, "loss": 2.8426, "step": 47584 }, { "epoch": 3.73, "learning_rate": 6.268721085234847e-05, "loss": 2.5612, "step": 47585 }, { "epoch": 3.73, "learning_rate": 6.268642672312397e-05, "loss": 2.8208, "step": 47586 }, { "epoch": 3.73, "learning_rate": 6.268564259389948e-05, "loss": 2.6256, "step": 47587 }, { "epoch": 3.73, "learning_rate": 6.268485846467498e-05, "loss": 2.6219, "step": 47588 }, { "epoch": 3.73, "learning_rate": 6.268407433545049e-05, "loss": 2.8927, "step": 47589 }, { "epoch": 3.73, "learning_rate": 6.268329020622599e-05, "loss": 2.6653, "step": 47590 }, { "epoch": 3.73, "learning_rate": 6.268250607700149e-05, "loss": 2.7334, "step": 47591 }, { "epoch": 3.73, "learning_rate": 6.2681721947777e-05, "loss": 2.6623, "step": 47592 }, { "epoch": 3.73, "learning_rate": 6.26809378185525e-05, "loss": 2.9176, "step": 47593 }, { "epoch": 3.73, "learning_rate": 6.2680153689328e-05, "loss": 2.7278, "step": 47594 }, { "epoch": 3.73, "learning_rate": 6.267936956010351e-05, "loss": 2.6486, "step": 47595 }, { "epoch": 3.73, "learning_rate": 6.267858543087901e-05, "loss": 2.8754, "step": 47596 }, { "epoch": 3.73, "learning_rate": 6.267780130165452e-05, "loss": 2.6447, "step": 47597 }, { "epoch": 3.73, "learning_rate": 6.267701717243002e-05, "loss": 2.5558, "step": 47598 }, { "epoch": 3.73, "learning_rate": 6.267623304320552e-05, "loss": 2.5887, "step": 47599 }, { "epoch": 3.73, "learning_rate": 6.267544891398103e-05, "loss": 2.6289, "step": 47600 }, { "epoch": 3.73, "learning_rate": 6.267466478475653e-05, "loss": 2.6368, "step": 47601 }, { "epoch": 3.73, "learning_rate": 6.267388065553203e-05, "loss": 2.7594, "step": 47602 }, { "epoch": 3.73, "learning_rate": 6.267309652630754e-05, "loss": 2.8092, "step": 47603 }, { "epoch": 3.73, "learning_rate": 6.267231239708304e-05, "loss": 2.6669, "step": 47604 }, { "epoch": 3.73, "learning_rate": 6.267152826785854e-05, "loss": 2.778, "step": 47605 }, { "epoch": 3.73, "learning_rate": 6.267074413863405e-05, "loss": 2.6562, "step": 47606 }, { "epoch": 3.73, "learning_rate": 6.266996000940955e-05, "loss": 2.7033, "step": 47607 }, { "epoch": 3.73, "learning_rate": 6.266917588018506e-05, "loss": 2.6322, "step": 47608 }, { "epoch": 3.73, "learning_rate": 6.266839175096056e-05, "loss": 2.7339, "step": 47609 }, { "epoch": 3.73, "learning_rate": 6.266760762173606e-05, "loss": 2.5988, "step": 47610 }, { "epoch": 3.73, "learning_rate": 6.266682349251157e-05, "loss": 2.6751, "step": 47611 }, { "epoch": 3.73, "learning_rate": 6.266603936328707e-05, "loss": 2.6512, "step": 47612 }, { "epoch": 3.73, "learning_rate": 6.266525523406259e-05, "loss": 2.7673, "step": 47613 }, { "epoch": 3.73, "learning_rate": 6.266447110483809e-05, "loss": 2.7757, "step": 47614 }, { "epoch": 3.73, "learning_rate": 6.26636869756136e-05, "loss": 2.6476, "step": 47615 }, { "epoch": 3.73, "learning_rate": 6.26629028463891e-05, "loss": 2.5408, "step": 47616 }, { "epoch": 3.73, "learning_rate": 6.266211871716459e-05, "loss": 2.7819, "step": 47617 }, { "epoch": 3.73, "learning_rate": 6.266133458794009e-05, "loss": 2.7353, "step": 47618 }, { "epoch": 3.73, "learning_rate": 6.26605504587156e-05, "loss": 2.6397, "step": 47619 }, { "epoch": 3.73, "learning_rate": 6.26597663294911e-05, "loss": 2.8022, "step": 47620 }, { "epoch": 3.73, "learning_rate": 6.26589822002666e-05, "loss": 2.9143, "step": 47621 }, { "epoch": 3.73, "learning_rate": 6.265819807104211e-05, "loss": 2.5799, "step": 47622 }, { "epoch": 3.73, "learning_rate": 6.265741394181761e-05, "loss": 2.629, "step": 47623 }, { "epoch": 3.73, "learning_rate": 6.265662981259311e-05, "loss": 2.8252, "step": 47624 }, { "epoch": 3.73, "learning_rate": 6.265584568336862e-05, "loss": 2.6946, "step": 47625 }, { "epoch": 3.73, "learning_rate": 6.265506155414412e-05, "loss": 2.8484, "step": 47626 }, { "epoch": 3.73, "learning_rate": 6.265427742491963e-05, "loss": 2.703, "step": 47627 }, { "epoch": 3.73, "learning_rate": 6.265349329569513e-05, "loss": 2.8707, "step": 47628 }, { "epoch": 3.73, "learning_rate": 6.265270916647063e-05, "loss": 2.903, "step": 47629 }, { "epoch": 3.73, "learning_rate": 6.265192503724614e-05, "loss": 2.9016, "step": 47630 }, { "epoch": 3.73, "learning_rate": 6.265114090802164e-05, "loss": 2.7175, "step": 47631 }, { "epoch": 3.73, "learning_rate": 6.265035677879714e-05, "loss": 2.9895, "step": 47632 }, { "epoch": 3.74, "learning_rate": 6.264957264957265e-05, "loss": 2.6707, "step": 47633 }, { "epoch": 3.74, "learning_rate": 6.264878852034815e-05, "loss": 2.8972, "step": 47634 }, { "epoch": 3.74, "learning_rate": 6.264800439112366e-05, "loss": 2.7487, "step": 47635 }, { "epoch": 3.74, "learning_rate": 6.264722026189916e-05, "loss": 2.7235, "step": 47636 }, { "epoch": 3.74, "learning_rate": 6.264643613267466e-05, "loss": 2.7263, "step": 47637 }, { "epoch": 3.74, "learning_rate": 6.264565200345017e-05, "loss": 2.7444, "step": 47638 }, { "epoch": 3.74, "learning_rate": 6.264486787422568e-05, "loss": 2.7329, "step": 47639 }, { "epoch": 3.74, "learning_rate": 6.264408374500119e-05, "loss": 2.8255, "step": 47640 }, { "epoch": 3.74, "learning_rate": 6.264329961577669e-05, "loss": 2.6386, "step": 47641 }, { "epoch": 3.74, "learning_rate": 6.26425154865522e-05, "loss": 2.6788, "step": 47642 }, { "epoch": 3.74, "learning_rate": 6.26417313573277e-05, "loss": 2.7819, "step": 47643 }, { "epoch": 3.74, "learning_rate": 6.26409472281032e-05, "loss": 2.7999, "step": 47644 }, { "epoch": 3.74, "learning_rate": 6.26401630988787e-05, "loss": 2.5356, "step": 47645 }, { "epoch": 3.74, "learning_rate": 6.263937896965421e-05, "loss": 2.6613, "step": 47646 }, { "epoch": 3.74, "learning_rate": 6.263859484042971e-05, "loss": 2.8323, "step": 47647 }, { "epoch": 3.74, "learning_rate": 6.263781071120522e-05, "loss": 2.5796, "step": 47648 }, { "epoch": 3.74, "learning_rate": 6.263702658198072e-05, "loss": 2.6347, "step": 47649 }, { "epoch": 3.74, "learning_rate": 6.263624245275622e-05, "loss": 2.8876, "step": 47650 }, { "epoch": 3.74, "learning_rate": 6.263545832353173e-05, "loss": 2.4929, "step": 47651 }, { "epoch": 3.74, "learning_rate": 6.263467419430723e-05, "loss": 2.6039, "step": 47652 }, { "epoch": 3.74, "learning_rate": 6.263389006508272e-05, "loss": 2.8066, "step": 47653 }, { "epoch": 3.74, "learning_rate": 6.263310593585822e-05, "loss": 2.6162, "step": 47654 }, { "epoch": 3.74, "learning_rate": 6.263232180663373e-05, "loss": 2.9026, "step": 47655 }, { "epoch": 3.74, "learning_rate": 6.263153767740923e-05, "loss": 2.8797, "step": 47656 }, { "epoch": 3.74, "learning_rate": 6.263075354818474e-05, "loss": 2.7514, "step": 47657 }, { "epoch": 3.74, "learning_rate": 6.262996941896024e-05, "loss": 2.7436, "step": 47658 }, { "epoch": 3.74, "learning_rate": 6.262918528973574e-05, "loss": 2.5484, "step": 47659 }, { "epoch": 3.74, "learning_rate": 6.262840116051125e-05, "loss": 2.8091, "step": 47660 }, { "epoch": 3.74, "learning_rate": 6.262761703128675e-05, "loss": 2.6022, "step": 47661 }, { "epoch": 3.74, "learning_rate": 6.262683290206225e-05, "loss": 2.8528, "step": 47662 }, { "epoch": 3.74, "learning_rate": 6.262604877283776e-05, "loss": 2.6243, "step": 47663 }, { "epoch": 3.74, "learning_rate": 6.262526464361326e-05, "loss": 2.7, "step": 47664 }, { "epoch": 3.74, "learning_rate": 6.262448051438878e-05, "loss": 2.539, "step": 47665 }, { "epoch": 3.74, "learning_rate": 6.262369638516428e-05, "loss": 2.5945, "step": 47666 }, { "epoch": 3.74, "learning_rate": 6.262291225593979e-05, "loss": 2.6887, "step": 47667 }, { "epoch": 3.74, "learning_rate": 6.262212812671529e-05, "loss": 2.6035, "step": 47668 }, { "epoch": 3.74, "learning_rate": 6.26213439974908e-05, "loss": 2.8075, "step": 47669 }, { "epoch": 3.74, "learning_rate": 6.26205598682663e-05, "loss": 2.7088, "step": 47670 }, { "epoch": 3.74, "learning_rate": 6.26197757390418e-05, "loss": 2.7206, "step": 47671 }, { "epoch": 3.74, "learning_rate": 6.26189916098173e-05, "loss": 2.5291, "step": 47672 }, { "epoch": 3.74, "learning_rate": 6.261820748059281e-05, "loss": 2.7151, "step": 47673 }, { "epoch": 3.74, "learning_rate": 6.261742335136831e-05, "loss": 2.613, "step": 47674 }, { "epoch": 3.74, "learning_rate": 6.261663922214382e-05, "loss": 2.5377, "step": 47675 }, { "epoch": 3.74, "learning_rate": 6.261585509291932e-05, "loss": 2.5643, "step": 47676 }, { "epoch": 3.74, "learning_rate": 6.261507096369482e-05, "loss": 2.7963, "step": 47677 }, { "epoch": 3.74, "learning_rate": 6.261428683447033e-05, "loss": 2.8052, "step": 47678 }, { "epoch": 3.74, "learning_rate": 6.261350270524583e-05, "loss": 2.8819, "step": 47679 }, { "epoch": 3.74, "learning_rate": 6.261271857602133e-05, "loss": 2.7291, "step": 47680 }, { "epoch": 3.74, "learning_rate": 6.261193444679684e-05, "loss": 2.7022, "step": 47681 }, { "epoch": 3.74, "learning_rate": 6.261115031757234e-05, "loss": 2.784, "step": 47682 }, { "epoch": 3.74, "learning_rate": 6.261036618834785e-05, "loss": 2.7396, "step": 47683 }, { "epoch": 3.74, "learning_rate": 6.260958205912335e-05, "loss": 2.7393, "step": 47684 }, { "epoch": 3.74, "learning_rate": 6.260879792989885e-05, "loss": 2.7233, "step": 47685 }, { "epoch": 3.74, "learning_rate": 6.260801380067436e-05, "loss": 2.4397, "step": 47686 }, { "epoch": 3.74, "learning_rate": 6.260722967144986e-05, "loss": 2.8103, "step": 47687 }, { "epoch": 3.74, "learning_rate": 6.260644554222536e-05, "loss": 2.7274, "step": 47688 }, { "epoch": 3.74, "learning_rate": 6.260566141300085e-05, "loss": 2.6054, "step": 47689 }, { "epoch": 3.74, "learning_rate": 6.260487728377637e-05, "loss": 2.7071, "step": 47690 }, { "epoch": 3.74, "learning_rate": 6.260409315455187e-05, "loss": 2.7322, "step": 47691 }, { "epoch": 3.74, "learning_rate": 6.260330902532738e-05, "loss": 2.6087, "step": 47692 }, { "epoch": 3.74, "learning_rate": 6.260252489610288e-05, "loss": 2.7351, "step": 47693 }, { "epoch": 3.74, "learning_rate": 6.260174076687839e-05, "loss": 2.7604, "step": 47694 }, { "epoch": 3.74, "learning_rate": 6.260095663765389e-05, "loss": 2.6165, "step": 47695 }, { "epoch": 3.74, "learning_rate": 6.260017250842939e-05, "loss": 2.9373, "step": 47696 }, { "epoch": 3.74, "learning_rate": 6.25993883792049e-05, "loss": 2.6864, "step": 47697 }, { "epoch": 3.74, "learning_rate": 6.25986042499804e-05, "loss": 2.6926, "step": 47698 }, { "epoch": 3.74, "learning_rate": 6.25978201207559e-05, "loss": 2.9499, "step": 47699 }, { "epoch": 3.74, "learning_rate": 6.259703599153141e-05, "loss": 2.576, "step": 47700 }, { "epoch": 3.74, "learning_rate": 6.259625186230691e-05, "loss": 2.7474, "step": 47701 }, { "epoch": 3.74, "learning_rate": 6.259546773308241e-05, "loss": 2.6133, "step": 47702 }, { "epoch": 3.74, "learning_rate": 6.259468360385792e-05, "loss": 2.7306, "step": 47703 }, { "epoch": 3.74, "learning_rate": 6.259389947463342e-05, "loss": 2.5828, "step": 47704 }, { "epoch": 3.74, "learning_rate": 6.259311534540893e-05, "loss": 2.6938, "step": 47705 }, { "epoch": 3.74, "learning_rate": 6.259233121618443e-05, "loss": 2.5666, "step": 47706 }, { "epoch": 3.74, "learning_rate": 6.259154708695993e-05, "loss": 2.8451, "step": 47707 }, { "epoch": 3.74, "learning_rate": 6.259076295773544e-05, "loss": 2.8098, "step": 47708 }, { "epoch": 3.74, "learning_rate": 6.258997882851094e-05, "loss": 2.6432, "step": 47709 }, { "epoch": 3.74, "learning_rate": 6.258919469928644e-05, "loss": 2.814, "step": 47710 }, { "epoch": 3.74, "learning_rate": 6.258841057006195e-05, "loss": 2.6347, "step": 47711 }, { "epoch": 3.74, "learning_rate": 6.258762644083745e-05, "loss": 2.8756, "step": 47712 }, { "epoch": 3.74, "learning_rate": 6.258684231161296e-05, "loss": 2.7345, "step": 47713 }, { "epoch": 3.74, "learning_rate": 6.258605818238846e-05, "loss": 2.7373, "step": 47714 }, { "epoch": 3.74, "learning_rate": 6.258527405316396e-05, "loss": 2.6565, "step": 47715 }, { "epoch": 3.74, "learning_rate": 6.258448992393947e-05, "loss": 2.7777, "step": 47716 }, { "epoch": 3.74, "learning_rate": 6.258370579471497e-05, "loss": 3.1105, "step": 47717 }, { "epoch": 3.74, "learning_rate": 6.258292166549047e-05, "loss": 2.4414, "step": 47718 }, { "epoch": 3.74, "learning_rate": 6.258213753626598e-05, "loss": 2.7962, "step": 47719 }, { "epoch": 3.74, "learning_rate": 6.258135340704148e-05, "loss": 2.8139, "step": 47720 }, { "epoch": 3.74, "learning_rate": 6.258056927781698e-05, "loss": 2.7815, "step": 47721 }, { "epoch": 3.74, "learning_rate": 6.25797851485925e-05, "loss": 2.8385, "step": 47722 }, { "epoch": 3.74, "learning_rate": 6.2579001019368e-05, "loss": 2.6941, "step": 47723 }, { "epoch": 3.74, "learning_rate": 6.257821689014351e-05, "loss": 2.5058, "step": 47724 }, { "epoch": 3.74, "learning_rate": 6.2577432760919e-05, "loss": 2.7164, "step": 47725 }, { "epoch": 3.74, "learning_rate": 6.25766486316945e-05, "loss": 2.7212, "step": 47726 }, { "epoch": 3.74, "learning_rate": 6.257586450247001e-05, "loss": 2.7049, "step": 47727 }, { "epoch": 3.74, "learning_rate": 6.257508037324551e-05, "loss": 2.8157, "step": 47728 }, { "epoch": 3.74, "learning_rate": 6.257429624402101e-05, "loss": 2.7164, "step": 47729 }, { "epoch": 3.74, "learning_rate": 6.257351211479652e-05, "loss": 2.5665, "step": 47730 }, { "epoch": 3.74, "learning_rate": 6.257272798557202e-05, "loss": 2.9893, "step": 47731 }, { "epoch": 3.74, "learning_rate": 6.257194385634753e-05, "loss": 2.6516, "step": 47732 }, { "epoch": 3.74, "learning_rate": 6.257115972712303e-05, "loss": 2.7787, "step": 47733 }, { "epoch": 3.74, "learning_rate": 6.257037559789853e-05, "loss": 2.6884, "step": 47734 }, { "epoch": 3.74, "learning_rate": 6.256959146867404e-05, "loss": 2.6541, "step": 47735 }, { "epoch": 3.74, "learning_rate": 6.256880733944954e-05, "loss": 3.1126, "step": 47736 }, { "epoch": 3.74, "learning_rate": 6.256802321022504e-05, "loss": 2.6873, "step": 47737 }, { "epoch": 3.74, "learning_rate": 6.256723908100055e-05, "loss": 2.6465, "step": 47738 }, { "epoch": 3.74, "learning_rate": 6.256645495177605e-05, "loss": 2.8981, "step": 47739 }, { "epoch": 3.74, "learning_rate": 6.256567082255155e-05, "loss": 2.7323, "step": 47740 }, { "epoch": 3.74, "learning_rate": 6.256488669332706e-05, "loss": 2.7871, "step": 47741 }, { "epoch": 3.74, "learning_rate": 6.256410256410256e-05, "loss": 2.7006, "step": 47742 }, { "epoch": 3.74, "learning_rate": 6.256331843487807e-05, "loss": 2.7744, "step": 47743 }, { "epoch": 3.74, "learning_rate": 6.256253430565357e-05, "loss": 2.6282, "step": 47744 }, { "epoch": 3.74, "learning_rate": 6.256175017642907e-05, "loss": 2.6378, "step": 47745 }, { "epoch": 3.74, "learning_rate": 6.256096604720458e-05, "loss": 2.7304, "step": 47746 }, { "epoch": 3.74, "learning_rate": 6.25601819179801e-05, "loss": 2.6148, "step": 47747 }, { "epoch": 3.74, "learning_rate": 6.25593977887556e-05, "loss": 2.6573, "step": 47748 }, { "epoch": 3.74, "learning_rate": 6.25586136595311e-05, "loss": 2.7569, "step": 47749 }, { "epoch": 3.74, "learning_rate": 6.25578295303066e-05, "loss": 2.5381, "step": 47750 }, { "epoch": 3.74, "learning_rate": 6.255704540108211e-05, "loss": 2.7606, "step": 47751 }, { "epoch": 3.74, "learning_rate": 6.255626127185761e-05, "loss": 2.7724, "step": 47752 }, { "epoch": 3.74, "learning_rate": 6.255547714263312e-05, "loss": 2.7875, "step": 47753 }, { "epoch": 3.74, "learning_rate": 6.255469301340862e-05, "loss": 2.7531, "step": 47754 }, { "epoch": 3.74, "learning_rate": 6.255390888418412e-05, "loss": 2.7495, "step": 47755 }, { "epoch": 3.74, "learning_rate": 6.255312475495963e-05, "loss": 2.7962, "step": 47756 }, { "epoch": 3.74, "learning_rate": 6.255234062573513e-05, "loss": 2.7307, "step": 47757 }, { "epoch": 3.74, "learning_rate": 6.255155649651063e-05, "loss": 2.7481, "step": 47758 }, { "epoch": 3.74, "learning_rate": 6.255077236728614e-05, "loss": 2.7599, "step": 47759 }, { "epoch": 3.75, "learning_rate": 6.254998823806164e-05, "loss": 2.7736, "step": 47760 }, { "epoch": 3.75, "learning_rate": 6.254920410883713e-05, "loss": 2.8814, "step": 47761 }, { "epoch": 3.75, "learning_rate": 6.254841997961264e-05, "loss": 2.6903, "step": 47762 }, { "epoch": 3.75, "learning_rate": 6.254763585038814e-05, "loss": 2.5817, "step": 47763 }, { "epoch": 3.75, "learning_rate": 6.254685172116364e-05, "loss": 2.6803, "step": 47764 }, { "epoch": 3.75, "learning_rate": 6.254606759193915e-05, "loss": 2.8644, "step": 47765 }, { "epoch": 3.75, "learning_rate": 6.254528346271465e-05, "loss": 3.0652, "step": 47766 }, { "epoch": 3.75, "learning_rate": 6.254449933349015e-05, "loss": 2.9331, "step": 47767 }, { "epoch": 3.75, "learning_rate": 6.254371520426566e-05, "loss": 2.615, "step": 47768 }, { "epoch": 3.75, "learning_rate": 6.254293107504116e-05, "loss": 2.7179, "step": 47769 }, { "epoch": 3.75, "learning_rate": 6.254214694581667e-05, "loss": 2.7521, "step": 47770 }, { "epoch": 3.75, "learning_rate": 6.254136281659217e-05, "loss": 2.9116, "step": 47771 }, { "epoch": 3.75, "learning_rate": 6.254057868736767e-05, "loss": 2.8461, "step": 47772 }, { "epoch": 3.75, "learning_rate": 6.253979455814319e-05, "loss": 2.9135, "step": 47773 }, { "epoch": 3.75, "learning_rate": 6.25390104289187e-05, "loss": 2.3987, "step": 47774 }, { "epoch": 3.75, "learning_rate": 6.25382262996942e-05, "loss": 2.7226, "step": 47775 }, { "epoch": 3.75, "learning_rate": 6.25374421704697e-05, "loss": 2.7067, "step": 47776 }, { "epoch": 3.75, "learning_rate": 6.25366580412452e-05, "loss": 2.6131, "step": 47777 }, { "epoch": 3.75, "learning_rate": 6.253587391202071e-05, "loss": 2.5503, "step": 47778 }, { "epoch": 3.75, "learning_rate": 6.253508978279621e-05, "loss": 2.6725, "step": 47779 }, { "epoch": 3.75, "learning_rate": 6.253430565357172e-05, "loss": 2.8136, "step": 47780 }, { "epoch": 3.75, "learning_rate": 6.253352152434722e-05, "loss": 2.7567, "step": 47781 }, { "epoch": 3.75, "learning_rate": 6.253273739512272e-05, "loss": 2.8085, "step": 47782 }, { "epoch": 3.75, "learning_rate": 6.253195326589823e-05, "loss": 2.7697, "step": 47783 }, { "epoch": 3.75, "learning_rate": 6.253116913667373e-05, "loss": 2.7554, "step": 47784 }, { "epoch": 3.75, "learning_rate": 6.253038500744923e-05, "loss": 2.6495, "step": 47785 }, { "epoch": 3.75, "learning_rate": 6.252960087822474e-05, "loss": 2.9347, "step": 47786 }, { "epoch": 3.75, "learning_rate": 6.252881674900024e-05, "loss": 2.7653, "step": 47787 }, { "epoch": 3.75, "learning_rate": 6.252803261977574e-05, "loss": 2.986, "step": 47788 }, { "epoch": 3.75, "learning_rate": 6.252724849055125e-05, "loss": 2.8467, "step": 47789 }, { "epoch": 3.75, "learning_rate": 6.252646436132675e-05, "loss": 2.8881, "step": 47790 }, { "epoch": 3.75, "learning_rate": 6.252568023210226e-05, "loss": 2.6045, "step": 47791 }, { "epoch": 3.75, "learning_rate": 6.252489610287776e-05, "loss": 2.7186, "step": 47792 }, { "epoch": 3.75, "learning_rate": 6.252411197365326e-05, "loss": 2.714, "step": 47793 }, { "epoch": 3.75, "learning_rate": 6.252332784442877e-05, "loss": 2.8746, "step": 47794 }, { "epoch": 3.75, "learning_rate": 6.252254371520427e-05, "loss": 2.8186, "step": 47795 }, { "epoch": 3.75, "learning_rate": 6.252175958597977e-05, "loss": 2.5069, "step": 47796 }, { "epoch": 3.75, "learning_rate": 6.252097545675526e-05, "loss": 2.9255, "step": 47797 }, { "epoch": 3.75, "learning_rate": 6.252019132753078e-05, "loss": 2.9455, "step": 47798 }, { "epoch": 3.75, "learning_rate": 6.251940719830629e-05, "loss": 2.8462, "step": 47799 }, { "epoch": 3.75, "learning_rate": 6.251862306908179e-05, "loss": 2.9269, "step": 47800 }, { "epoch": 3.75, "learning_rate": 6.251783893985729e-05, "loss": 2.7958, "step": 47801 }, { "epoch": 3.75, "learning_rate": 6.25170548106328e-05, "loss": 2.7908, "step": 47802 }, { "epoch": 3.75, "learning_rate": 6.25162706814083e-05, "loss": 2.9147, "step": 47803 }, { "epoch": 3.75, "learning_rate": 6.25154865521838e-05, "loss": 2.6891, "step": 47804 }, { "epoch": 3.75, "learning_rate": 6.251470242295931e-05, "loss": 2.7441, "step": 47805 }, { "epoch": 3.75, "learning_rate": 6.251391829373481e-05, "loss": 2.6218, "step": 47806 }, { "epoch": 3.75, "learning_rate": 6.251313416451031e-05, "loss": 2.7012, "step": 47807 }, { "epoch": 3.75, "learning_rate": 6.251235003528582e-05, "loss": 2.7044, "step": 47808 }, { "epoch": 3.75, "learning_rate": 6.251156590606132e-05, "loss": 2.8359, "step": 47809 }, { "epoch": 3.75, "learning_rate": 6.251078177683683e-05, "loss": 2.5319, "step": 47810 }, { "epoch": 3.75, "learning_rate": 6.250999764761233e-05, "loss": 2.6698, "step": 47811 }, { "epoch": 3.75, "learning_rate": 6.250921351838783e-05, "loss": 2.7447, "step": 47812 }, { "epoch": 3.75, "learning_rate": 6.250842938916334e-05, "loss": 2.6163, "step": 47813 }, { "epoch": 3.75, "learning_rate": 6.250764525993884e-05, "loss": 2.7123, "step": 47814 }, { "epoch": 3.75, "learning_rate": 6.250686113071434e-05, "loss": 2.6739, "step": 47815 }, { "epoch": 3.75, "learning_rate": 6.250607700148985e-05, "loss": 2.6961, "step": 47816 }, { "epoch": 3.75, "learning_rate": 6.250529287226535e-05, "loss": 2.6497, "step": 47817 }, { "epoch": 3.75, "learning_rate": 6.250450874304086e-05, "loss": 2.5714, "step": 47818 }, { "epoch": 3.75, "learning_rate": 6.250372461381636e-05, "loss": 2.9756, "step": 47819 }, { "epoch": 3.75, "learning_rate": 6.250294048459186e-05, "loss": 2.7282, "step": 47820 }, { "epoch": 3.75, "learning_rate": 6.250215635536737e-05, "loss": 2.6769, "step": 47821 }, { "epoch": 3.75, "learning_rate": 6.250137222614287e-05, "loss": 2.7665, "step": 47822 }, { "epoch": 3.75, "learning_rate": 6.250058809691837e-05, "loss": 2.7437, "step": 47823 }, { "epoch": 3.75, "learning_rate": 6.249980396769388e-05, "loss": 2.6168, "step": 47824 }, { "epoch": 3.75, "learning_rate": 6.249901983846938e-05, "loss": 2.9017, "step": 47825 }, { "epoch": 3.75, "learning_rate": 6.249823570924488e-05, "loss": 2.7999, "step": 47826 }, { "epoch": 3.75, "learning_rate": 6.249745158002039e-05, "loss": 2.8493, "step": 47827 }, { "epoch": 3.75, "learning_rate": 6.249666745079589e-05, "loss": 2.7685, "step": 47828 }, { "epoch": 3.75, "learning_rate": 6.24958833215714e-05, "loss": 2.8123, "step": 47829 }, { "epoch": 3.75, "learning_rate": 6.249509919234691e-05, "loss": 2.709, "step": 47830 }, { "epoch": 3.75, "learning_rate": 6.249431506312242e-05, "loss": 2.8569, "step": 47831 }, { "epoch": 3.75, "learning_rate": 6.249353093389792e-05, "loss": 2.7577, "step": 47832 }, { "epoch": 3.75, "learning_rate": 6.249274680467341e-05, "loss": 2.6574, "step": 47833 }, { "epoch": 3.75, "learning_rate": 6.249196267544891e-05, "loss": 2.5398, "step": 47834 }, { "epoch": 3.75, "learning_rate": 6.249117854622442e-05, "loss": 2.9507, "step": 47835 }, { "epoch": 3.75, "learning_rate": 6.249039441699992e-05, "loss": 2.67, "step": 47836 }, { "epoch": 3.75, "learning_rate": 6.248961028777543e-05, "loss": 2.769, "step": 47837 }, { "epoch": 3.75, "learning_rate": 6.248882615855093e-05, "loss": 2.4947, "step": 47838 }, { "epoch": 3.75, "learning_rate": 6.248804202932643e-05, "loss": 2.5249, "step": 47839 }, { "epoch": 3.75, "learning_rate": 6.248725790010194e-05, "loss": 2.8195, "step": 47840 }, { "epoch": 3.75, "learning_rate": 6.248647377087744e-05, "loss": 2.8081, "step": 47841 }, { "epoch": 3.75, "learning_rate": 6.248568964165294e-05, "loss": 3.0577, "step": 47842 }, { "epoch": 3.75, "learning_rate": 6.248490551242845e-05, "loss": 2.7864, "step": 47843 }, { "epoch": 3.75, "learning_rate": 6.248412138320395e-05, "loss": 2.7859, "step": 47844 }, { "epoch": 3.75, "learning_rate": 6.248333725397945e-05, "loss": 2.773, "step": 47845 }, { "epoch": 3.75, "learning_rate": 6.248255312475496e-05, "loss": 2.5253, "step": 47846 }, { "epoch": 3.75, "learning_rate": 6.248176899553046e-05, "loss": 2.6886, "step": 47847 }, { "epoch": 3.75, "learning_rate": 6.248098486630597e-05, "loss": 2.7384, "step": 47848 }, { "epoch": 3.75, "learning_rate": 6.248020073708147e-05, "loss": 2.8128, "step": 47849 }, { "epoch": 3.75, "learning_rate": 6.247941660785697e-05, "loss": 2.8339, "step": 47850 }, { "epoch": 3.75, "learning_rate": 6.247863247863248e-05, "loss": 2.6075, "step": 47851 }, { "epoch": 3.75, "learning_rate": 6.247784834940798e-05, "loss": 2.5054, "step": 47852 }, { "epoch": 3.75, "learning_rate": 6.247706422018348e-05, "loss": 2.6115, "step": 47853 }, { "epoch": 3.75, "learning_rate": 6.247628009095899e-05, "loss": 2.7344, "step": 47854 }, { "epoch": 3.75, "learning_rate": 6.24754959617345e-05, "loss": 2.5402, "step": 47855 }, { "epoch": 3.75, "learning_rate": 6.247471183251001e-05, "loss": 2.7234, "step": 47856 }, { "epoch": 3.75, "learning_rate": 6.247392770328551e-05, "loss": 2.7686, "step": 47857 }, { "epoch": 3.75, "learning_rate": 6.247314357406102e-05, "loss": 2.9657, "step": 47858 }, { "epoch": 3.75, "learning_rate": 6.247235944483652e-05, "loss": 2.7413, "step": 47859 }, { "epoch": 3.75, "learning_rate": 6.247157531561202e-05, "loss": 2.7428, "step": 47860 }, { "epoch": 3.75, "learning_rate": 6.247079118638753e-05, "loss": 2.6579, "step": 47861 }, { "epoch": 3.75, "learning_rate": 6.247000705716303e-05, "loss": 2.9477, "step": 47862 }, { "epoch": 3.75, "learning_rate": 6.246922292793853e-05, "loss": 2.7399, "step": 47863 }, { "epoch": 3.75, "learning_rate": 6.246843879871404e-05, "loss": 2.5223, "step": 47864 }, { "epoch": 3.75, "learning_rate": 6.246765466948954e-05, "loss": 2.7324, "step": 47865 }, { "epoch": 3.75, "learning_rate": 6.246687054026505e-05, "loss": 3.1927, "step": 47866 }, { "epoch": 3.75, "learning_rate": 6.246608641104055e-05, "loss": 2.7907, "step": 47867 }, { "epoch": 3.75, "learning_rate": 6.246530228181605e-05, "loss": 2.638, "step": 47868 }, { "epoch": 3.75, "learning_rate": 6.246451815259154e-05, "loss": 2.8706, "step": 47869 }, { "epoch": 3.75, "learning_rate": 6.246373402336705e-05, "loss": 2.6122, "step": 47870 }, { "epoch": 3.75, "learning_rate": 6.246294989414255e-05, "loss": 2.7611, "step": 47871 }, { "epoch": 3.75, "learning_rate": 6.246216576491805e-05, "loss": 2.7231, "step": 47872 }, { "epoch": 3.75, "learning_rate": 6.246138163569356e-05, "loss": 2.705, "step": 47873 }, { "epoch": 3.75, "learning_rate": 6.246059750646906e-05, "loss": 2.3894, "step": 47874 }, { "epoch": 3.75, "learning_rate": 6.245981337724456e-05, "loss": 2.5716, "step": 47875 }, { "epoch": 3.75, "learning_rate": 6.245902924802007e-05, "loss": 2.549, "step": 47876 }, { "epoch": 3.75, "learning_rate": 6.245824511879557e-05, "loss": 2.9227, "step": 47877 }, { "epoch": 3.75, "learning_rate": 6.245746098957108e-05, "loss": 2.7782, "step": 47878 }, { "epoch": 3.75, "learning_rate": 6.245667686034658e-05, "loss": 2.8208, "step": 47879 }, { "epoch": 3.75, "learning_rate": 6.245589273112208e-05, "loss": 2.9477, "step": 47880 }, { "epoch": 3.75, "learning_rate": 6.24551086018976e-05, "loss": 2.9691, "step": 47881 }, { "epoch": 3.75, "learning_rate": 6.24543244726731e-05, "loss": 2.7168, "step": 47882 }, { "epoch": 3.75, "learning_rate": 6.245354034344861e-05, "loss": 2.762, "step": 47883 }, { "epoch": 3.75, "learning_rate": 6.245275621422411e-05, "loss": 2.7995, "step": 47884 }, { "epoch": 3.75, "learning_rate": 6.245197208499962e-05, "loss": 2.5766, "step": 47885 }, { "epoch": 3.75, "learning_rate": 6.245118795577512e-05, "loss": 2.5169, "step": 47886 }, { "epoch": 3.75, "learning_rate": 6.245040382655062e-05, "loss": 2.5761, "step": 47887 }, { "epoch": 3.76, "learning_rate": 6.244961969732613e-05, "loss": 2.8441, "step": 47888 }, { "epoch": 3.76, "learning_rate": 6.244883556810163e-05, "loss": 2.7001, "step": 47889 }, { "epoch": 3.76, "learning_rate": 6.244805143887713e-05, "loss": 2.7261, "step": 47890 }, { "epoch": 3.76, "learning_rate": 6.244726730965264e-05, "loss": 2.7673, "step": 47891 }, { "epoch": 3.76, "learning_rate": 6.244648318042814e-05, "loss": 2.9013, "step": 47892 }, { "epoch": 3.76, "learning_rate": 6.244569905120364e-05, "loss": 2.8037, "step": 47893 }, { "epoch": 3.76, "learning_rate": 6.244491492197915e-05, "loss": 2.5633, "step": 47894 }, { "epoch": 3.76, "learning_rate": 6.244413079275465e-05, "loss": 2.6749, "step": 47895 }, { "epoch": 3.76, "learning_rate": 6.244334666353016e-05, "loss": 2.795, "step": 47896 }, { "epoch": 3.76, "learning_rate": 6.244256253430566e-05, "loss": 2.8096, "step": 47897 }, { "epoch": 3.76, "learning_rate": 6.244177840508116e-05, "loss": 2.7425, "step": 47898 }, { "epoch": 3.76, "learning_rate": 6.244099427585667e-05, "loss": 2.8226, "step": 47899 }, { "epoch": 3.76, "learning_rate": 6.244021014663217e-05, "loss": 2.7689, "step": 47900 }, { "epoch": 3.76, "learning_rate": 6.243942601740767e-05, "loss": 2.6833, "step": 47901 }, { "epoch": 3.76, "learning_rate": 6.243864188818318e-05, "loss": 2.9013, "step": 47902 }, { "epoch": 3.76, "learning_rate": 6.243785775895868e-05, "loss": 2.7529, "step": 47903 }, { "epoch": 3.76, "learning_rate": 6.243707362973418e-05, "loss": 2.7056, "step": 47904 }, { "epoch": 3.76, "learning_rate": 6.243628950050968e-05, "loss": 2.6298, "step": 47905 }, { "epoch": 3.76, "learning_rate": 6.243550537128519e-05, "loss": 2.5945, "step": 47906 }, { "epoch": 3.76, "learning_rate": 6.24347212420607e-05, "loss": 2.606, "step": 47907 }, { "epoch": 3.76, "learning_rate": 6.24339371128362e-05, "loss": 2.8763, "step": 47908 }, { "epoch": 3.76, "learning_rate": 6.24331529836117e-05, "loss": 2.734, "step": 47909 }, { "epoch": 3.76, "learning_rate": 6.243236885438721e-05, "loss": 2.8669, "step": 47910 }, { "epoch": 3.76, "learning_rate": 6.243158472516271e-05, "loss": 2.6437, "step": 47911 }, { "epoch": 3.76, "learning_rate": 6.243080059593821e-05, "loss": 2.6267, "step": 47912 }, { "epoch": 3.76, "learning_rate": 6.243001646671372e-05, "loss": 2.5768, "step": 47913 }, { "epoch": 3.76, "learning_rate": 6.242923233748922e-05, "loss": 2.6488, "step": 47914 }, { "epoch": 3.76, "learning_rate": 6.242844820826473e-05, "loss": 2.6702, "step": 47915 }, { "epoch": 3.76, "learning_rate": 6.242766407904023e-05, "loss": 2.5903, "step": 47916 }, { "epoch": 3.76, "learning_rate": 6.242687994981573e-05, "loss": 2.6381, "step": 47917 }, { "epoch": 3.76, "learning_rate": 6.242609582059124e-05, "loss": 2.8189, "step": 47918 }, { "epoch": 3.76, "learning_rate": 6.242531169136674e-05, "loss": 2.8811, "step": 47919 }, { "epoch": 3.76, "learning_rate": 6.242452756214224e-05, "loss": 2.5875, "step": 47920 }, { "epoch": 3.76, "learning_rate": 6.242374343291775e-05, "loss": 2.7326, "step": 47921 }, { "epoch": 3.76, "learning_rate": 6.242295930369325e-05, "loss": 2.5107, "step": 47922 }, { "epoch": 3.76, "learning_rate": 6.242217517446875e-05, "loss": 2.7981, "step": 47923 }, { "epoch": 3.76, "learning_rate": 6.242139104524426e-05, "loss": 2.752, "step": 47924 }, { "epoch": 3.76, "learning_rate": 6.242060691601976e-05, "loss": 2.6268, "step": 47925 }, { "epoch": 3.76, "learning_rate": 6.241982278679527e-05, "loss": 2.5509, "step": 47926 }, { "epoch": 3.76, "learning_rate": 6.241903865757077e-05, "loss": 2.7947, "step": 47927 }, { "epoch": 3.76, "learning_rate": 6.241825452834627e-05, "loss": 2.6539, "step": 47928 }, { "epoch": 3.76, "learning_rate": 6.241747039912178e-05, "loss": 2.8327, "step": 47929 }, { "epoch": 3.76, "learning_rate": 6.241668626989728e-05, "loss": 2.9254, "step": 47930 }, { "epoch": 3.76, "learning_rate": 6.241590214067278e-05, "loss": 2.6715, "step": 47931 }, { "epoch": 3.76, "learning_rate": 6.241511801144829e-05, "loss": 2.7651, "step": 47932 }, { "epoch": 3.76, "learning_rate": 6.241433388222379e-05, "loss": 2.5995, "step": 47933 }, { "epoch": 3.76, "learning_rate": 6.24135497529993e-05, "loss": 2.5642, "step": 47934 }, { "epoch": 3.76, "learning_rate": 6.24127656237748e-05, "loss": 2.7914, "step": 47935 }, { "epoch": 3.76, "learning_rate": 6.24119814945503e-05, "loss": 2.6306, "step": 47936 }, { "epoch": 3.76, "learning_rate": 6.24111973653258e-05, "loss": 2.5949, "step": 47937 }, { "epoch": 3.76, "learning_rate": 6.241041323610132e-05, "loss": 2.8164, "step": 47938 }, { "epoch": 3.76, "learning_rate": 6.240962910687683e-05, "loss": 2.7715, "step": 47939 }, { "epoch": 3.76, "learning_rate": 6.240884497765233e-05, "loss": 2.7999, "step": 47940 }, { "epoch": 3.76, "learning_rate": 6.240806084842782e-05, "loss": 2.5633, "step": 47941 }, { "epoch": 3.76, "learning_rate": 6.240727671920332e-05, "loss": 2.7261, "step": 47942 }, { "epoch": 3.76, "learning_rate": 6.240649258997883e-05, "loss": 2.6843, "step": 47943 }, { "epoch": 3.76, "learning_rate": 6.240570846075433e-05, "loss": 2.6078, "step": 47944 }, { "epoch": 3.76, "learning_rate": 6.240492433152984e-05, "loss": 2.8965, "step": 47945 }, { "epoch": 3.76, "learning_rate": 6.240414020230534e-05, "loss": 2.7466, "step": 47946 }, { "epoch": 3.76, "learning_rate": 6.240335607308084e-05, "loss": 2.9028, "step": 47947 }, { "epoch": 3.76, "learning_rate": 6.240257194385635e-05, "loss": 2.9764, "step": 47948 }, { "epoch": 3.76, "learning_rate": 6.240178781463185e-05, "loss": 2.7654, "step": 47949 }, { "epoch": 3.76, "learning_rate": 6.240100368540735e-05, "loss": 2.5587, "step": 47950 }, { "epoch": 3.76, "learning_rate": 6.240021955618286e-05, "loss": 2.7993, "step": 47951 }, { "epoch": 3.76, "learning_rate": 6.239943542695836e-05, "loss": 2.5889, "step": 47952 }, { "epoch": 3.76, "learning_rate": 6.239865129773387e-05, "loss": 2.6682, "step": 47953 }, { "epoch": 3.76, "learning_rate": 6.239786716850937e-05, "loss": 3.0086, "step": 47954 }, { "epoch": 3.76, "learning_rate": 6.239708303928487e-05, "loss": 2.9242, "step": 47955 }, { "epoch": 3.76, "learning_rate": 6.239629891006038e-05, "loss": 2.7115, "step": 47956 }, { "epoch": 3.76, "learning_rate": 6.239551478083588e-05, "loss": 2.603, "step": 47957 }, { "epoch": 3.76, "learning_rate": 6.239473065161138e-05, "loss": 2.5735, "step": 47958 }, { "epoch": 3.76, "learning_rate": 6.239394652238689e-05, "loss": 2.7269, "step": 47959 }, { "epoch": 3.76, "learning_rate": 6.239316239316239e-05, "loss": 2.5777, "step": 47960 }, { "epoch": 3.76, "learning_rate": 6.23923782639379e-05, "loss": 2.7069, "step": 47961 }, { "epoch": 3.76, "learning_rate": 6.23915941347134e-05, "loss": 2.6822, "step": 47962 }, { "epoch": 3.76, "learning_rate": 6.239081000548892e-05, "loss": 2.6525, "step": 47963 }, { "epoch": 3.76, "learning_rate": 6.239002587626442e-05, "loss": 2.7973, "step": 47964 }, { "epoch": 3.76, "learning_rate": 6.238924174703992e-05, "loss": 2.7914, "step": 47965 }, { "epoch": 3.76, "learning_rate": 6.238845761781543e-05, "loss": 2.8001, "step": 47966 }, { "epoch": 3.76, "learning_rate": 6.238767348859093e-05, "loss": 2.9209, "step": 47967 }, { "epoch": 3.76, "learning_rate": 6.238688935936643e-05, "loss": 2.7603, "step": 47968 }, { "epoch": 3.76, "learning_rate": 6.238610523014194e-05, "loss": 2.8042, "step": 47969 }, { "epoch": 3.76, "learning_rate": 6.238532110091744e-05, "loss": 2.7665, "step": 47970 }, { "epoch": 3.76, "learning_rate": 6.238453697169294e-05, "loss": 2.9299, "step": 47971 }, { "epoch": 3.76, "learning_rate": 6.238375284246845e-05, "loss": 2.7924, "step": 47972 }, { "epoch": 3.76, "learning_rate": 6.238296871324395e-05, "loss": 2.8051, "step": 47973 }, { "epoch": 3.76, "learning_rate": 6.238218458401946e-05, "loss": 2.697, "step": 47974 }, { "epoch": 3.76, "learning_rate": 6.238140045479496e-05, "loss": 2.7763, "step": 47975 }, { "epoch": 3.76, "learning_rate": 6.238061632557046e-05, "loss": 2.7624, "step": 47976 }, { "epoch": 3.76, "learning_rate": 6.237983219634595e-05, "loss": 2.5002, "step": 47977 }, { "epoch": 3.76, "learning_rate": 6.237904806712146e-05, "loss": 2.6172, "step": 47978 }, { "epoch": 3.76, "learning_rate": 6.237826393789696e-05, "loss": 2.8081, "step": 47979 }, { "epoch": 3.76, "learning_rate": 6.237747980867246e-05, "loss": 2.8497, "step": 47980 }, { "epoch": 3.76, "learning_rate": 6.237669567944797e-05, "loss": 2.8896, "step": 47981 }, { "epoch": 3.76, "learning_rate": 6.237591155022347e-05, "loss": 2.6818, "step": 47982 }, { "epoch": 3.76, "learning_rate": 6.237512742099898e-05, "loss": 2.7912, "step": 47983 }, { "epoch": 3.76, "learning_rate": 6.237434329177448e-05, "loss": 2.686, "step": 47984 }, { "epoch": 3.76, "learning_rate": 6.237355916254998e-05, "loss": 2.7382, "step": 47985 }, { "epoch": 3.76, "learning_rate": 6.237277503332549e-05, "loss": 3.0312, "step": 47986 }, { "epoch": 3.76, "learning_rate": 6.237199090410099e-05, "loss": 2.9127, "step": 47987 }, { "epoch": 3.76, "learning_rate": 6.23712067748765e-05, "loss": 2.737, "step": 47988 }, { "epoch": 3.76, "learning_rate": 6.237042264565201e-05, "loss": 3.1113, "step": 47989 }, { "epoch": 3.76, "learning_rate": 6.236963851642751e-05, "loss": 2.7077, "step": 47990 }, { "epoch": 3.76, "learning_rate": 6.236885438720302e-05, "loss": 2.7639, "step": 47991 }, { "epoch": 3.76, "learning_rate": 6.236807025797852e-05, "loss": 2.6687, "step": 47992 }, { "epoch": 3.76, "learning_rate": 6.236728612875403e-05, "loss": 2.8419, "step": 47993 }, { "epoch": 3.76, "learning_rate": 6.236650199952953e-05, "loss": 2.6327, "step": 47994 }, { "epoch": 3.76, "learning_rate": 6.236571787030503e-05, "loss": 2.7593, "step": 47995 }, { "epoch": 3.76, "learning_rate": 6.236493374108054e-05, "loss": 2.5872, "step": 47996 }, { "epoch": 3.76, "learning_rate": 6.236414961185604e-05, "loss": 2.8558, "step": 47997 }, { "epoch": 3.76, "learning_rate": 6.236336548263154e-05, "loss": 2.603, "step": 47998 }, { "epoch": 3.76, "learning_rate": 6.236258135340705e-05, "loss": 2.6627, "step": 47999 }, { "epoch": 3.76, "learning_rate": 6.236179722418255e-05, "loss": 2.7137, "step": 48000 }, { "epoch": 3.76, "eval_loss": 2.8905482292175293, "eval_runtime": 217.8592, "eval_samples_per_second": 52.015, "eval_steps_per_second": 3.254, "step": 48000 }, { "epoch": 3.76, "learning_rate": 6.236101309495806e-05, "loss": 2.9173, "step": 48001 }, { "epoch": 3.76, "learning_rate": 6.236022896573356e-05, "loss": 2.5659, "step": 48002 }, { "epoch": 3.76, "learning_rate": 6.235944483650906e-05, "loss": 2.7014, "step": 48003 }, { "epoch": 3.76, "learning_rate": 6.235866070728457e-05, "loss": 2.6299, "step": 48004 }, { "epoch": 3.76, "learning_rate": 6.235787657806007e-05, "loss": 2.7862, "step": 48005 }, { "epoch": 3.76, "learning_rate": 6.235709244883557e-05, "loss": 2.6499, "step": 48006 }, { "epoch": 3.76, "learning_rate": 6.235630831961108e-05, "loss": 2.7903, "step": 48007 }, { "epoch": 3.76, "learning_rate": 6.235552419038658e-05, "loss": 2.6498, "step": 48008 }, { "epoch": 3.76, "learning_rate": 6.235474006116208e-05, "loss": 2.6283, "step": 48009 }, { "epoch": 3.76, "learning_rate": 6.235395593193759e-05, "loss": 2.8744, "step": 48010 }, { "epoch": 3.76, "learning_rate": 6.235317180271309e-05, "loss": 2.8524, "step": 48011 }, { "epoch": 3.76, "learning_rate": 6.23523876734886e-05, "loss": 2.8225, "step": 48012 }, { "epoch": 3.76, "learning_rate": 6.235160354426409e-05, "loss": 2.6959, "step": 48013 }, { "epoch": 3.76, "learning_rate": 6.23508194150396e-05, "loss": 2.8807, "step": 48014 }, { "epoch": 3.76, "learning_rate": 6.23500352858151e-05, "loss": 2.7208, "step": 48015 }, { "epoch": 3.77, "learning_rate": 6.234925115659061e-05, "loss": 2.7255, "step": 48016 }, { "epoch": 3.77, "learning_rate": 6.234846702736611e-05, "loss": 2.6424, "step": 48017 }, { "epoch": 3.77, "learning_rate": 6.234768289814162e-05, "loss": 2.6867, "step": 48018 }, { "epoch": 3.77, "learning_rate": 6.234689876891712e-05, "loss": 2.7813, "step": 48019 }, { "epoch": 3.77, "learning_rate": 6.234611463969263e-05, "loss": 2.8408, "step": 48020 }, { "epoch": 3.77, "learning_rate": 6.234533051046813e-05, "loss": 2.8667, "step": 48021 }, { "epoch": 3.77, "learning_rate": 6.234454638124363e-05, "loss": 2.6491, "step": 48022 }, { "epoch": 3.77, "learning_rate": 6.234376225201914e-05, "loss": 2.7162, "step": 48023 }, { "epoch": 3.77, "learning_rate": 6.234297812279464e-05, "loss": 2.7566, "step": 48024 }, { "epoch": 3.77, "learning_rate": 6.234219399357014e-05, "loss": 2.5946, "step": 48025 }, { "epoch": 3.77, "learning_rate": 6.234140986434565e-05, "loss": 2.8061, "step": 48026 }, { "epoch": 3.77, "learning_rate": 6.234062573512115e-05, "loss": 2.8093, "step": 48027 }, { "epoch": 3.77, "learning_rate": 6.233984160589665e-05, "loss": 2.7235, "step": 48028 }, { "epoch": 3.77, "learning_rate": 6.233905747667216e-05, "loss": 2.8421, "step": 48029 }, { "epoch": 3.77, "learning_rate": 6.233827334744766e-05, "loss": 2.7623, "step": 48030 }, { "epoch": 3.77, "learning_rate": 6.233748921822317e-05, "loss": 2.6856, "step": 48031 }, { "epoch": 3.77, "learning_rate": 6.233670508899867e-05, "loss": 2.7301, "step": 48032 }, { "epoch": 3.77, "learning_rate": 6.233592095977417e-05, "loss": 2.7659, "step": 48033 }, { "epoch": 3.77, "learning_rate": 6.233513683054968e-05, "loss": 2.7908, "step": 48034 }, { "epoch": 3.77, "learning_rate": 6.233435270132518e-05, "loss": 2.7834, "step": 48035 }, { "epoch": 3.77, "learning_rate": 6.233356857210068e-05, "loss": 2.7618, "step": 48036 }, { "epoch": 3.77, "learning_rate": 6.233278444287619e-05, "loss": 2.6783, "step": 48037 }, { "epoch": 3.77, "learning_rate": 6.233200031365169e-05, "loss": 2.7623, "step": 48038 }, { "epoch": 3.77, "learning_rate": 6.23312161844272e-05, "loss": 2.7364, "step": 48039 }, { "epoch": 3.77, "learning_rate": 6.23304320552027e-05, "loss": 2.7526, "step": 48040 }, { "epoch": 3.77, "learning_rate": 6.23296479259782e-05, "loss": 2.6456, "step": 48041 }, { "epoch": 3.77, "learning_rate": 6.23288637967537e-05, "loss": 2.6215, "step": 48042 }, { "epoch": 3.77, "learning_rate": 6.232807966752921e-05, "loss": 2.7958, "step": 48043 }, { "epoch": 3.77, "learning_rate": 6.232729553830471e-05, "loss": 2.5962, "step": 48044 }, { "epoch": 3.77, "learning_rate": 6.232651140908022e-05, "loss": 2.8272, "step": 48045 }, { "epoch": 3.77, "learning_rate": 6.232572727985573e-05, "loss": 2.7656, "step": 48046 }, { "epoch": 3.77, "learning_rate": 6.232494315063124e-05, "loss": 2.6238, "step": 48047 }, { "epoch": 3.77, "learning_rate": 6.232415902140674e-05, "loss": 2.7289, "step": 48048 }, { "epoch": 3.77, "learning_rate": 6.232337489218223e-05, "loss": 2.5014, "step": 48049 }, { "epoch": 3.77, "learning_rate": 6.232259076295774e-05, "loss": 2.779, "step": 48050 }, { "epoch": 3.77, "learning_rate": 6.232180663373324e-05, "loss": 2.5306, "step": 48051 }, { "epoch": 3.77, "learning_rate": 6.232102250450874e-05, "loss": 2.864, "step": 48052 }, { "epoch": 3.77, "learning_rate": 6.232023837528425e-05, "loss": 2.8017, "step": 48053 }, { "epoch": 3.77, "learning_rate": 6.231945424605975e-05, "loss": 2.7528, "step": 48054 }, { "epoch": 3.77, "learning_rate": 6.231867011683525e-05, "loss": 2.6714, "step": 48055 }, { "epoch": 3.77, "learning_rate": 6.231788598761076e-05, "loss": 2.607, "step": 48056 }, { "epoch": 3.77, "learning_rate": 6.231710185838626e-05, "loss": 2.8378, "step": 48057 }, { "epoch": 3.77, "learning_rate": 6.231631772916176e-05, "loss": 2.6367, "step": 48058 }, { "epoch": 3.77, "learning_rate": 6.231553359993727e-05, "loss": 2.6285, "step": 48059 }, { "epoch": 3.77, "learning_rate": 6.231474947071277e-05, "loss": 2.6877, "step": 48060 }, { "epoch": 3.77, "learning_rate": 6.231396534148828e-05, "loss": 2.5231, "step": 48061 }, { "epoch": 3.77, "learning_rate": 6.231318121226378e-05, "loss": 2.8704, "step": 48062 }, { "epoch": 3.77, "learning_rate": 6.231239708303928e-05, "loss": 2.9565, "step": 48063 }, { "epoch": 3.77, "learning_rate": 6.231161295381479e-05, "loss": 2.6931, "step": 48064 }, { "epoch": 3.77, "learning_rate": 6.231082882459029e-05, "loss": 2.7832, "step": 48065 }, { "epoch": 3.77, "learning_rate": 6.23100446953658e-05, "loss": 2.6003, "step": 48066 }, { "epoch": 3.77, "learning_rate": 6.23092605661413e-05, "loss": 2.5425, "step": 48067 }, { "epoch": 3.77, "learning_rate": 6.23084764369168e-05, "loss": 2.7133, "step": 48068 }, { "epoch": 3.77, "learning_rate": 6.23076923076923e-05, "loss": 2.8647, "step": 48069 }, { "epoch": 3.77, "learning_rate": 6.230690817846781e-05, "loss": 2.7805, "step": 48070 }, { "epoch": 3.77, "learning_rate": 6.230612404924333e-05, "loss": 2.8693, "step": 48071 }, { "epoch": 3.77, "learning_rate": 6.230533992001883e-05, "loss": 2.6802, "step": 48072 }, { "epoch": 3.77, "learning_rate": 6.230455579079433e-05, "loss": 2.8728, "step": 48073 }, { "epoch": 3.77, "learning_rate": 6.230377166156984e-05, "loss": 2.6538, "step": 48074 }, { "epoch": 3.77, "learning_rate": 6.230298753234534e-05, "loss": 2.7782, "step": 48075 }, { "epoch": 3.77, "learning_rate": 6.230220340312084e-05, "loss": 2.6526, "step": 48076 }, { "epoch": 3.77, "learning_rate": 6.230141927389635e-05, "loss": 2.7948, "step": 48077 }, { "epoch": 3.77, "learning_rate": 6.230063514467185e-05, "loss": 2.7747, "step": 48078 }, { "epoch": 3.77, "learning_rate": 6.229985101544736e-05, "loss": 2.7193, "step": 48079 }, { "epoch": 3.77, "learning_rate": 6.229906688622286e-05, "loss": 2.8808, "step": 48080 }, { "epoch": 3.77, "learning_rate": 6.229828275699836e-05, "loss": 2.8605, "step": 48081 }, { "epoch": 3.77, "learning_rate": 6.229749862777387e-05, "loss": 2.4692, "step": 48082 }, { "epoch": 3.77, "learning_rate": 6.229671449854937e-05, "loss": 2.8178, "step": 48083 }, { "epoch": 3.77, "learning_rate": 6.229593036932487e-05, "loss": 2.7179, "step": 48084 }, { "epoch": 3.77, "learning_rate": 6.229514624010036e-05, "loss": 2.8048, "step": 48085 }, { "epoch": 3.77, "learning_rate": 6.229436211087587e-05, "loss": 2.6843, "step": 48086 }, { "epoch": 3.77, "learning_rate": 6.229357798165137e-05, "loss": 2.8509, "step": 48087 }, { "epoch": 3.77, "learning_rate": 6.229279385242688e-05, "loss": 2.802, "step": 48088 }, { "epoch": 3.77, "learning_rate": 6.229200972320238e-05, "loss": 2.8016, "step": 48089 }, { "epoch": 3.77, "learning_rate": 6.229122559397788e-05, "loss": 2.8678, "step": 48090 }, { "epoch": 3.77, "learning_rate": 6.229044146475339e-05, "loss": 2.7133, "step": 48091 }, { "epoch": 3.77, "learning_rate": 6.228965733552889e-05, "loss": 2.6738, "step": 48092 }, { "epoch": 3.77, "learning_rate": 6.22888732063044e-05, "loss": 2.6558, "step": 48093 }, { "epoch": 3.77, "learning_rate": 6.22880890770799e-05, "loss": 2.5687, "step": 48094 }, { "epoch": 3.77, "learning_rate": 6.22873049478554e-05, "loss": 2.6251, "step": 48095 }, { "epoch": 3.77, "learning_rate": 6.22865208186309e-05, "loss": 2.6531, "step": 48096 }, { "epoch": 3.77, "learning_rate": 6.228573668940642e-05, "loss": 2.8817, "step": 48097 }, { "epoch": 3.77, "learning_rate": 6.228495256018193e-05, "loss": 2.5922, "step": 48098 }, { "epoch": 3.77, "learning_rate": 6.228416843095743e-05, "loss": 2.5952, "step": 48099 }, { "epoch": 3.77, "learning_rate": 6.228338430173293e-05, "loss": 2.8662, "step": 48100 }, { "epoch": 3.77, "learning_rate": 6.228260017250844e-05, "loss": 2.6935, "step": 48101 }, { "epoch": 3.77, "learning_rate": 6.228181604328394e-05, "loss": 2.9457, "step": 48102 }, { "epoch": 3.77, "learning_rate": 6.228103191405944e-05, "loss": 2.6714, "step": 48103 }, { "epoch": 3.77, "learning_rate": 6.228024778483495e-05, "loss": 2.6529, "step": 48104 }, { "epoch": 3.77, "learning_rate": 6.227946365561045e-05, "loss": 2.6781, "step": 48105 }, { "epoch": 3.77, "learning_rate": 6.227867952638595e-05, "loss": 2.5946, "step": 48106 }, { "epoch": 3.77, "learning_rate": 6.227789539716146e-05, "loss": 2.7672, "step": 48107 }, { "epoch": 3.77, "learning_rate": 6.227711126793696e-05, "loss": 2.8149, "step": 48108 }, { "epoch": 3.77, "learning_rate": 6.227632713871247e-05, "loss": 2.7472, "step": 48109 }, { "epoch": 3.77, "learning_rate": 6.227554300948797e-05, "loss": 2.9148, "step": 48110 }, { "epoch": 3.77, "learning_rate": 6.227475888026347e-05, "loss": 2.6583, "step": 48111 }, { "epoch": 3.77, "learning_rate": 6.227397475103898e-05, "loss": 2.9707, "step": 48112 }, { "epoch": 3.77, "learning_rate": 6.227319062181448e-05, "loss": 2.7288, "step": 48113 }, { "epoch": 3.77, "learning_rate": 6.227240649258998e-05, "loss": 2.6178, "step": 48114 }, { "epoch": 3.77, "learning_rate": 6.227162236336549e-05, "loss": 2.754, "step": 48115 }, { "epoch": 3.77, "learning_rate": 6.227083823414099e-05, "loss": 2.88, "step": 48116 }, { "epoch": 3.77, "learning_rate": 6.22700541049165e-05, "loss": 3.1959, "step": 48117 }, { "epoch": 3.77, "learning_rate": 6.2269269975692e-05, "loss": 2.6941, "step": 48118 }, { "epoch": 3.77, "learning_rate": 6.22684858464675e-05, "loss": 2.4912, "step": 48119 }, { "epoch": 3.77, "learning_rate": 6.2267701717243e-05, "loss": 2.8634, "step": 48120 }, { "epoch": 3.77, "learning_rate": 6.22669175880185e-05, "loss": 2.8272, "step": 48121 }, { "epoch": 3.77, "learning_rate": 6.226613345879401e-05, "loss": 2.9116, "step": 48122 }, { "epoch": 3.77, "learning_rate": 6.226534932956952e-05, "loss": 2.9964, "step": 48123 }, { "epoch": 3.77, "learning_rate": 6.226456520034502e-05, "loss": 2.6465, "step": 48124 }, { "epoch": 3.77, "learning_rate": 6.226378107112052e-05, "loss": 2.7637, "step": 48125 }, { "epoch": 3.77, "learning_rate": 6.226299694189603e-05, "loss": 2.544, "step": 48126 }, { "epoch": 3.77, "learning_rate": 6.226221281267153e-05, "loss": 2.7831, "step": 48127 }, { "epoch": 3.77, "learning_rate": 6.226142868344704e-05, "loss": 2.7725, "step": 48128 }, { "epoch": 3.77, "learning_rate": 6.226064455422254e-05, "loss": 2.5949, "step": 48129 }, { "epoch": 3.77, "learning_rate": 6.225986042499804e-05, "loss": 2.5793, "step": 48130 }, { "epoch": 3.77, "learning_rate": 6.225907629577355e-05, "loss": 2.5575, "step": 48131 }, { "epoch": 3.77, "learning_rate": 6.225829216654905e-05, "loss": 2.606, "step": 48132 }, { "epoch": 3.77, "learning_rate": 6.225750803732455e-05, "loss": 2.8393, "step": 48133 }, { "epoch": 3.77, "learning_rate": 6.225672390810006e-05, "loss": 2.7421, "step": 48134 }, { "epoch": 3.77, "learning_rate": 6.225593977887556e-05, "loss": 2.823, "step": 48135 }, { "epoch": 3.77, "learning_rate": 6.225515564965107e-05, "loss": 2.5414, "step": 48136 }, { "epoch": 3.77, "learning_rate": 6.225437152042657e-05, "loss": 2.7642, "step": 48137 }, { "epoch": 3.77, "learning_rate": 6.225358739120207e-05, "loss": 2.5844, "step": 48138 }, { "epoch": 3.77, "learning_rate": 6.225280326197758e-05, "loss": 2.6819, "step": 48139 }, { "epoch": 3.77, "learning_rate": 6.225201913275308e-05, "loss": 2.6216, "step": 48140 }, { "epoch": 3.77, "learning_rate": 6.225123500352858e-05, "loss": 2.7536, "step": 48141 }, { "epoch": 3.77, "learning_rate": 6.225045087430409e-05, "loss": 2.9902, "step": 48142 }, { "epoch": 3.78, "learning_rate": 6.224966674507959e-05, "loss": 2.6498, "step": 48143 }, { "epoch": 3.78, "learning_rate": 6.22488826158551e-05, "loss": 2.5472, "step": 48144 }, { "epoch": 3.78, "learning_rate": 6.22480984866306e-05, "loss": 2.6446, "step": 48145 }, { "epoch": 3.78, "learning_rate": 6.22473143574061e-05, "loss": 2.8442, "step": 48146 }, { "epoch": 3.78, "learning_rate": 6.22465302281816e-05, "loss": 2.5939, "step": 48147 }, { "epoch": 3.78, "learning_rate": 6.224574609895711e-05, "loss": 2.7046, "step": 48148 }, { "epoch": 3.78, "learning_rate": 6.224496196973261e-05, "loss": 2.7973, "step": 48149 }, { "epoch": 3.78, "learning_rate": 6.224417784050812e-05, "loss": 2.8526, "step": 48150 }, { "epoch": 3.78, "learning_rate": 6.224339371128362e-05, "loss": 2.5802, "step": 48151 }, { "epoch": 3.78, "learning_rate": 6.224260958205912e-05, "loss": 3.0224, "step": 48152 }, { "epoch": 3.78, "learning_rate": 6.224182545283463e-05, "loss": 2.7574, "step": 48153 }, { "epoch": 3.78, "learning_rate": 6.224104132361015e-05, "loss": 2.8643, "step": 48154 }, { "epoch": 3.78, "learning_rate": 6.224025719438565e-05, "loss": 2.7397, "step": 48155 }, { "epoch": 3.78, "learning_rate": 6.223947306516115e-05, "loss": 2.6182, "step": 48156 }, { "epoch": 3.78, "learning_rate": 6.223868893593664e-05, "loss": 2.822, "step": 48157 }, { "epoch": 3.78, "learning_rate": 6.223790480671215e-05, "loss": 2.7935, "step": 48158 }, { "epoch": 3.78, "learning_rate": 6.223712067748765e-05, "loss": 2.6337, "step": 48159 }, { "epoch": 3.78, "learning_rate": 6.223633654826315e-05, "loss": 2.7403, "step": 48160 }, { "epoch": 3.78, "learning_rate": 6.223555241903866e-05, "loss": 2.6343, "step": 48161 }, { "epoch": 3.78, "learning_rate": 6.223476828981416e-05, "loss": 2.5963, "step": 48162 }, { "epoch": 3.78, "learning_rate": 6.223398416058966e-05, "loss": 2.7435, "step": 48163 }, { "epoch": 3.78, "learning_rate": 6.223320003136517e-05, "loss": 2.9045, "step": 48164 }, { "epoch": 3.78, "learning_rate": 6.223241590214067e-05, "loss": 2.7658, "step": 48165 }, { "epoch": 3.78, "learning_rate": 6.223163177291618e-05, "loss": 2.6731, "step": 48166 }, { "epoch": 3.78, "learning_rate": 6.223084764369168e-05, "loss": 2.7272, "step": 48167 }, { "epoch": 3.78, "learning_rate": 6.223006351446718e-05, "loss": 2.7756, "step": 48168 }, { "epoch": 3.78, "learning_rate": 6.222927938524269e-05, "loss": 2.6144, "step": 48169 }, { "epoch": 3.78, "learning_rate": 6.222849525601819e-05, "loss": 2.6481, "step": 48170 }, { "epoch": 3.78, "learning_rate": 6.22277111267937e-05, "loss": 2.4857, "step": 48171 }, { "epoch": 3.78, "learning_rate": 6.22269269975692e-05, "loss": 2.6534, "step": 48172 }, { "epoch": 3.78, "learning_rate": 6.22261428683447e-05, "loss": 2.5885, "step": 48173 }, { "epoch": 3.78, "learning_rate": 6.22253587391202e-05, "loss": 2.9917, "step": 48174 }, { "epoch": 3.78, "learning_rate": 6.222457460989571e-05, "loss": 2.9635, "step": 48175 }, { "epoch": 3.78, "learning_rate": 6.222379048067121e-05, "loss": 2.6736, "step": 48176 }, { "epoch": 3.78, "learning_rate": 6.222300635144672e-05, "loss": 2.8283, "step": 48177 }, { "epoch": 3.78, "learning_rate": 6.222222222222222e-05, "loss": 2.6681, "step": 48178 }, { "epoch": 3.78, "learning_rate": 6.222143809299774e-05, "loss": 2.7879, "step": 48179 }, { "epoch": 3.78, "learning_rate": 6.222065396377324e-05, "loss": 2.7174, "step": 48180 }, { "epoch": 3.78, "learning_rate": 6.221986983454874e-05, "loss": 2.7767, "step": 48181 }, { "epoch": 3.78, "learning_rate": 6.221908570532425e-05, "loss": 2.773, "step": 48182 }, { "epoch": 3.78, "learning_rate": 6.221830157609975e-05, "loss": 2.5577, "step": 48183 }, { "epoch": 3.78, "learning_rate": 6.221751744687526e-05, "loss": 2.8316, "step": 48184 }, { "epoch": 3.78, "learning_rate": 6.221673331765076e-05, "loss": 2.7526, "step": 48185 }, { "epoch": 3.78, "learning_rate": 6.221594918842626e-05, "loss": 2.7094, "step": 48186 }, { "epoch": 3.78, "learning_rate": 6.221516505920177e-05, "loss": 2.6921, "step": 48187 }, { "epoch": 3.78, "learning_rate": 6.221438092997727e-05, "loss": 2.7813, "step": 48188 }, { "epoch": 3.78, "learning_rate": 6.221359680075277e-05, "loss": 2.6531, "step": 48189 }, { "epoch": 3.78, "learning_rate": 6.221281267152828e-05, "loss": 2.9576, "step": 48190 }, { "epoch": 3.78, "learning_rate": 6.221202854230378e-05, "loss": 2.6795, "step": 48191 }, { "epoch": 3.78, "learning_rate": 6.221124441307928e-05, "loss": 2.847, "step": 48192 }, { "epoch": 3.78, "learning_rate": 6.221046028385477e-05, "loss": 2.9311, "step": 48193 }, { "epoch": 3.78, "learning_rate": 6.220967615463028e-05, "loss": 2.7143, "step": 48194 }, { "epoch": 3.78, "learning_rate": 6.220889202540578e-05, "loss": 2.7103, "step": 48195 }, { "epoch": 3.78, "learning_rate": 6.220810789618129e-05, "loss": 2.6164, "step": 48196 }, { "epoch": 3.78, "learning_rate": 6.220732376695679e-05, "loss": 2.9646, "step": 48197 }, { "epoch": 3.78, "learning_rate": 6.22065396377323e-05, "loss": 2.8978, "step": 48198 }, { "epoch": 3.78, "learning_rate": 6.22057555085078e-05, "loss": 2.6542, "step": 48199 }, { "epoch": 3.78, "learning_rate": 6.22049713792833e-05, "loss": 2.6855, "step": 48200 }, { "epoch": 3.78, "learning_rate": 6.22041872500588e-05, "loss": 2.6759, "step": 48201 }, { "epoch": 3.78, "learning_rate": 6.220340312083431e-05, "loss": 2.5526, "step": 48202 }, { "epoch": 3.78, "learning_rate": 6.220261899160981e-05, "loss": 2.7427, "step": 48203 }, { "epoch": 3.78, "learning_rate": 6.220183486238532e-05, "loss": 3.1408, "step": 48204 }, { "epoch": 3.78, "learning_rate": 6.220105073316083e-05, "loss": 2.5146, "step": 48205 }, { "epoch": 3.78, "learning_rate": 6.220026660393634e-05, "loss": 2.5401, "step": 48206 }, { "epoch": 3.78, "learning_rate": 6.219948247471184e-05, "loss": 2.7343, "step": 48207 }, { "epoch": 3.78, "learning_rate": 6.219869834548734e-05, "loss": 2.6249, "step": 48208 }, { "epoch": 3.78, "learning_rate": 6.219791421626285e-05, "loss": 2.72, "step": 48209 }, { "epoch": 3.78, "learning_rate": 6.219713008703835e-05, "loss": 2.6798, "step": 48210 }, { "epoch": 3.78, "learning_rate": 6.219634595781385e-05, "loss": 2.4896, "step": 48211 }, { "epoch": 3.78, "learning_rate": 6.219556182858936e-05, "loss": 2.6229, "step": 48212 }, { "epoch": 3.78, "learning_rate": 6.219477769936486e-05, "loss": 2.755, "step": 48213 }, { "epoch": 3.78, "learning_rate": 6.219399357014037e-05, "loss": 2.6853, "step": 48214 }, { "epoch": 3.78, "learning_rate": 6.219320944091587e-05, "loss": 2.6514, "step": 48215 }, { "epoch": 3.78, "learning_rate": 6.219242531169137e-05, "loss": 2.7261, "step": 48216 }, { "epoch": 3.78, "learning_rate": 6.219164118246688e-05, "loss": 2.762, "step": 48217 }, { "epoch": 3.78, "learning_rate": 6.219085705324238e-05, "loss": 2.8003, "step": 48218 }, { "epoch": 3.78, "learning_rate": 6.219007292401788e-05, "loss": 2.6927, "step": 48219 }, { "epoch": 3.78, "learning_rate": 6.218928879479339e-05, "loss": 2.832, "step": 48220 }, { "epoch": 3.78, "learning_rate": 6.218850466556889e-05, "loss": 2.7804, "step": 48221 }, { "epoch": 3.78, "learning_rate": 6.21877205363444e-05, "loss": 2.6721, "step": 48222 }, { "epoch": 3.78, "learning_rate": 6.21869364071199e-05, "loss": 2.5836, "step": 48223 }, { "epoch": 3.78, "learning_rate": 6.21861522778954e-05, "loss": 2.6802, "step": 48224 }, { "epoch": 3.78, "learning_rate": 6.21853681486709e-05, "loss": 2.7356, "step": 48225 }, { "epoch": 3.78, "learning_rate": 6.218458401944641e-05, "loss": 2.5705, "step": 48226 }, { "epoch": 3.78, "learning_rate": 6.218379989022191e-05, "loss": 2.8942, "step": 48227 }, { "epoch": 3.78, "learning_rate": 6.218301576099742e-05, "loss": 2.7331, "step": 48228 }, { "epoch": 3.78, "learning_rate": 6.218223163177291e-05, "loss": 2.5352, "step": 48229 }, { "epoch": 3.78, "learning_rate": 6.218144750254841e-05, "loss": 2.8479, "step": 48230 }, { "epoch": 3.78, "learning_rate": 6.218066337332393e-05, "loss": 2.6861, "step": 48231 }, { "epoch": 3.78, "learning_rate": 6.217987924409943e-05, "loss": 2.7567, "step": 48232 }, { "epoch": 3.78, "learning_rate": 6.217909511487494e-05, "loss": 2.9283, "step": 48233 }, { "epoch": 3.78, "learning_rate": 6.217831098565044e-05, "loss": 2.7085, "step": 48234 }, { "epoch": 3.78, "learning_rate": 6.217752685642594e-05, "loss": 2.7441, "step": 48235 }, { "epoch": 3.78, "learning_rate": 6.217674272720145e-05, "loss": 2.9039, "step": 48236 }, { "epoch": 3.78, "learning_rate": 6.217595859797695e-05, "loss": 2.7906, "step": 48237 }, { "epoch": 3.78, "learning_rate": 6.217517446875245e-05, "loss": 2.7891, "step": 48238 }, { "epoch": 3.78, "learning_rate": 6.217439033952796e-05, "loss": 2.684, "step": 48239 }, { "epoch": 3.78, "learning_rate": 6.217360621030346e-05, "loss": 2.648, "step": 48240 }, { "epoch": 3.78, "learning_rate": 6.217282208107897e-05, "loss": 2.8241, "step": 48241 }, { "epoch": 3.78, "learning_rate": 6.217203795185447e-05, "loss": 2.6503, "step": 48242 }, { "epoch": 3.78, "learning_rate": 6.217125382262997e-05, "loss": 2.6098, "step": 48243 }, { "epoch": 3.78, "learning_rate": 6.217046969340548e-05, "loss": 2.674, "step": 48244 }, { "epoch": 3.78, "learning_rate": 6.216968556418098e-05, "loss": 2.7318, "step": 48245 }, { "epoch": 3.78, "learning_rate": 6.216890143495648e-05, "loss": 2.6104, "step": 48246 }, { "epoch": 3.78, "learning_rate": 6.216811730573199e-05, "loss": 2.8206, "step": 48247 }, { "epoch": 3.78, "learning_rate": 6.216733317650749e-05, "loss": 2.7518, "step": 48248 }, { "epoch": 3.78, "learning_rate": 6.2166549047283e-05, "loss": 2.9117, "step": 48249 }, { "epoch": 3.78, "learning_rate": 6.21657649180585e-05, "loss": 2.5804, "step": 48250 }, { "epoch": 3.78, "learning_rate": 6.2164980788834e-05, "loss": 2.5616, "step": 48251 }, { "epoch": 3.78, "learning_rate": 6.21641966596095e-05, "loss": 2.7612, "step": 48252 }, { "epoch": 3.78, "learning_rate": 6.216341253038501e-05, "loss": 2.729, "step": 48253 }, { "epoch": 3.78, "learning_rate": 6.216262840116051e-05, "loss": 2.9213, "step": 48254 }, { "epoch": 3.78, "learning_rate": 6.216184427193602e-05, "loss": 2.6514, "step": 48255 }, { "epoch": 3.78, "learning_rate": 6.216106014271152e-05, "loss": 2.7913, "step": 48256 }, { "epoch": 3.78, "learning_rate": 6.216027601348702e-05, "loss": 2.6036, "step": 48257 }, { "epoch": 3.78, "learning_rate": 6.215949188426253e-05, "loss": 2.6035, "step": 48258 }, { "epoch": 3.78, "learning_rate": 6.215870775503803e-05, "loss": 2.6549, "step": 48259 }, { "epoch": 3.78, "learning_rate": 6.215792362581353e-05, "loss": 2.8432, "step": 48260 }, { "epoch": 3.78, "learning_rate": 6.215713949658904e-05, "loss": 2.7731, "step": 48261 }, { "epoch": 3.78, "learning_rate": 6.215635536736456e-05, "loss": 2.7611, "step": 48262 }, { "epoch": 3.78, "learning_rate": 6.215557123814006e-05, "loss": 2.6019, "step": 48263 }, { "epoch": 3.78, "learning_rate": 6.215478710891556e-05, "loss": 2.619, "step": 48264 }, { "epoch": 3.78, "learning_rate": 6.215400297969105e-05, "loss": 2.9289, "step": 48265 }, { "epoch": 3.78, "learning_rate": 6.215321885046656e-05, "loss": 2.6919, "step": 48266 }, { "epoch": 3.78, "learning_rate": 6.215243472124206e-05, "loss": 2.8612, "step": 48267 }, { "epoch": 3.78, "learning_rate": 6.215165059201756e-05, "loss": 2.6753, "step": 48268 }, { "epoch": 3.78, "learning_rate": 6.215086646279307e-05, "loss": 2.6226, "step": 48269 }, { "epoch": 3.78, "learning_rate": 6.215008233356857e-05, "loss": 2.743, "step": 48270 }, { "epoch": 3.79, "learning_rate": 6.214929820434408e-05, "loss": 2.6708, "step": 48271 }, { "epoch": 3.79, "learning_rate": 6.214851407511958e-05, "loss": 2.8758, "step": 48272 }, { "epoch": 3.79, "learning_rate": 6.214772994589508e-05, "loss": 2.7455, "step": 48273 }, { "epoch": 3.79, "learning_rate": 6.214694581667059e-05, "loss": 2.6497, "step": 48274 }, { "epoch": 3.79, "learning_rate": 6.214616168744609e-05, "loss": 2.799, "step": 48275 }, { "epoch": 3.79, "learning_rate": 6.21453775582216e-05, "loss": 2.6264, "step": 48276 }, { "epoch": 3.79, "learning_rate": 6.21445934289971e-05, "loss": 2.5976, "step": 48277 }, { "epoch": 3.79, "learning_rate": 6.21438092997726e-05, "loss": 2.7247, "step": 48278 }, { "epoch": 3.79, "learning_rate": 6.21430251705481e-05, "loss": 2.7708, "step": 48279 }, { "epoch": 3.79, "learning_rate": 6.214224104132361e-05, "loss": 2.6387, "step": 48280 }, { "epoch": 3.79, "learning_rate": 6.214145691209911e-05, "loss": 2.6619, "step": 48281 }, { "epoch": 3.79, "learning_rate": 6.214067278287462e-05, "loss": 2.8634, "step": 48282 }, { "epoch": 3.79, "learning_rate": 6.213988865365012e-05, "loss": 2.7335, "step": 48283 }, { "epoch": 3.79, "learning_rate": 6.213910452442562e-05, "loss": 2.715, "step": 48284 }, { "epoch": 3.79, "learning_rate": 6.213832039520113e-05, "loss": 2.7369, "step": 48285 }, { "epoch": 3.79, "learning_rate": 6.213753626597663e-05, "loss": 2.7497, "step": 48286 }, { "epoch": 3.79, "learning_rate": 6.213675213675215e-05, "loss": 2.5647, "step": 48287 }, { "epoch": 3.79, "learning_rate": 6.213596800752765e-05, "loss": 2.7074, "step": 48288 }, { "epoch": 3.79, "learning_rate": 6.213518387830316e-05, "loss": 2.7026, "step": 48289 }, { "epoch": 3.79, "learning_rate": 6.213439974907866e-05, "loss": 2.8389, "step": 48290 }, { "epoch": 3.79, "learning_rate": 6.213361561985416e-05, "loss": 2.7004, "step": 48291 }, { "epoch": 3.79, "learning_rate": 6.213283149062967e-05, "loss": 2.8837, "step": 48292 }, { "epoch": 3.79, "learning_rate": 6.213204736140517e-05, "loss": 2.7421, "step": 48293 }, { "epoch": 3.79, "learning_rate": 6.213126323218067e-05, "loss": 2.7395, "step": 48294 }, { "epoch": 3.79, "learning_rate": 6.213047910295618e-05, "loss": 2.7604, "step": 48295 }, { "epoch": 3.79, "learning_rate": 6.212969497373168e-05, "loss": 2.8181, "step": 48296 }, { "epoch": 3.79, "learning_rate": 6.212891084450718e-05, "loss": 2.7541, "step": 48297 }, { "epoch": 3.79, "learning_rate": 6.212812671528269e-05, "loss": 2.6602, "step": 48298 }, { "epoch": 3.79, "learning_rate": 6.212734258605819e-05, "loss": 2.6854, "step": 48299 }, { "epoch": 3.79, "learning_rate": 6.21265584568337e-05, "loss": 2.5626, "step": 48300 }, { "epoch": 3.79, "learning_rate": 6.212577432760919e-05, "loss": 2.8272, "step": 48301 }, { "epoch": 3.79, "learning_rate": 6.212499019838469e-05, "loss": 3.0445, "step": 48302 }, { "epoch": 3.79, "learning_rate": 6.212420606916019e-05, "loss": 2.8135, "step": 48303 }, { "epoch": 3.79, "learning_rate": 6.21234219399357e-05, "loss": 2.9695, "step": 48304 }, { "epoch": 3.79, "learning_rate": 6.21226378107112e-05, "loss": 2.6164, "step": 48305 }, { "epoch": 3.79, "learning_rate": 6.21218536814867e-05, "loss": 2.5683, "step": 48306 }, { "epoch": 3.79, "learning_rate": 6.212106955226221e-05, "loss": 2.7534, "step": 48307 }, { "epoch": 3.79, "learning_rate": 6.212028542303771e-05, "loss": 2.8195, "step": 48308 }, { "epoch": 3.79, "learning_rate": 6.211950129381322e-05, "loss": 2.7496, "step": 48309 }, { "epoch": 3.79, "learning_rate": 6.211871716458872e-05, "loss": 2.6548, "step": 48310 }, { "epoch": 3.79, "learning_rate": 6.211793303536422e-05, "loss": 2.9752, "step": 48311 }, { "epoch": 3.79, "learning_rate": 6.211714890613973e-05, "loss": 2.8019, "step": 48312 }, { "epoch": 3.79, "learning_rate": 6.211636477691524e-05, "loss": 2.9053, "step": 48313 }, { "epoch": 3.79, "learning_rate": 6.211558064769075e-05, "loss": 2.6897, "step": 48314 }, { "epoch": 3.79, "learning_rate": 6.211479651846625e-05, "loss": 2.6861, "step": 48315 }, { "epoch": 3.79, "learning_rate": 6.211401238924175e-05, "loss": 2.7879, "step": 48316 }, { "epoch": 3.79, "learning_rate": 6.211322826001726e-05, "loss": 2.6749, "step": 48317 }, { "epoch": 3.79, "learning_rate": 6.211244413079276e-05, "loss": 2.759, "step": 48318 }, { "epoch": 3.79, "learning_rate": 6.211166000156827e-05, "loss": 2.6557, "step": 48319 }, { "epoch": 3.79, "learning_rate": 6.211087587234377e-05, "loss": 2.796, "step": 48320 }, { "epoch": 3.79, "learning_rate": 6.211009174311927e-05, "loss": 2.6334, "step": 48321 }, { "epoch": 3.79, "learning_rate": 6.210930761389478e-05, "loss": 2.5071, "step": 48322 }, { "epoch": 3.79, "learning_rate": 6.210852348467028e-05, "loss": 2.62, "step": 48323 }, { "epoch": 3.79, "learning_rate": 6.210773935544578e-05, "loss": 2.664, "step": 48324 }, { "epoch": 3.79, "learning_rate": 6.210695522622129e-05, "loss": 2.6599, "step": 48325 }, { "epoch": 3.79, "learning_rate": 6.210617109699679e-05, "loss": 2.7174, "step": 48326 }, { "epoch": 3.79, "learning_rate": 6.21053869677723e-05, "loss": 2.624, "step": 48327 }, { "epoch": 3.79, "learning_rate": 6.21046028385478e-05, "loss": 2.5791, "step": 48328 }, { "epoch": 3.79, "learning_rate": 6.21038187093233e-05, "loss": 2.518, "step": 48329 }, { "epoch": 3.79, "learning_rate": 6.21030345800988e-05, "loss": 2.764, "step": 48330 }, { "epoch": 3.79, "learning_rate": 6.210225045087431e-05, "loss": 2.5971, "step": 48331 }, { "epoch": 3.79, "learning_rate": 6.210146632164981e-05, "loss": 2.6533, "step": 48332 }, { "epoch": 3.79, "learning_rate": 6.210068219242532e-05, "loss": 2.7206, "step": 48333 }, { "epoch": 3.79, "learning_rate": 6.209989806320082e-05, "loss": 2.7533, "step": 48334 }, { "epoch": 3.79, "learning_rate": 6.209911393397632e-05, "loss": 2.7057, "step": 48335 }, { "epoch": 3.79, "learning_rate": 6.209832980475183e-05, "loss": 2.705, "step": 48336 }, { "epoch": 3.79, "learning_rate": 6.209754567552732e-05, "loss": 2.7608, "step": 48337 }, { "epoch": 3.79, "learning_rate": 6.209676154630282e-05, "loss": 2.643, "step": 48338 }, { "epoch": 3.79, "learning_rate": 6.209597741707834e-05, "loss": 2.9079, "step": 48339 }, { "epoch": 3.79, "learning_rate": 6.209519328785384e-05, "loss": 2.6467, "step": 48340 }, { "epoch": 3.79, "learning_rate": 6.209440915862935e-05, "loss": 2.7578, "step": 48341 }, { "epoch": 3.79, "learning_rate": 6.209362502940485e-05, "loss": 2.6365, "step": 48342 }, { "epoch": 3.79, "learning_rate": 6.209284090018035e-05, "loss": 2.7129, "step": 48343 }, { "epoch": 3.79, "learning_rate": 6.209205677095586e-05, "loss": 2.6303, "step": 48344 }, { "epoch": 3.79, "learning_rate": 6.209127264173136e-05, "loss": 2.5725, "step": 48345 }, { "epoch": 3.79, "learning_rate": 6.209048851250686e-05, "loss": 2.7313, "step": 48346 }, { "epoch": 3.79, "learning_rate": 6.208970438328237e-05, "loss": 2.7111, "step": 48347 }, { "epoch": 3.79, "learning_rate": 6.208892025405787e-05, "loss": 2.6927, "step": 48348 }, { "epoch": 3.79, "learning_rate": 6.208813612483338e-05, "loss": 2.7199, "step": 48349 }, { "epoch": 3.79, "learning_rate": 6.208735199560888e-05, "loss": 2.8089, "step": 48350 }, { "epoch": 3.79, "learning_rate": 6.208656786638438e-05, "loss": 2.5231, "step": 48351 }, { "epoch": 3.79, "learning_rate": 6.208578373715989e-05, "loss": 3.0343, "step": 48352 }, { "epoch": 3.79, "learning_rate": 6.208499960793539e-05, "loss": 2.7355, "step": 48353 }, { "epoch": 3.79, "learning_rate": 6.20842154787109e-05, "loss": 2.5337, "step": 48354 }, { "epoch": 3.79, "learning_rate": 6.20834313494864e-05, "loss": 2.5944, "step": 48355 }, { "epoch": 3.79, "learning_rate": 6.20826472202619e-05, "loss": 2.8982, "step": 48356 }, { "epoch": 3.79, "learning_rate": 6.20818630910374e-05, "loss": 2.7625, "step": 48357 }, { "epoch": 3.79, "learning_rate": 6.208107896181291e-05, "loss": 2.6368, "step": 48358 }, { "epoch": 3.79, "learning_rate": 6.208029483258841e-05, "loss": 2.682, "step": 48359 }, { "epoch": 3.79, "learning_rate": 6.207951070336392e-05, "loss": 2.8379, "step": 48360 }, { "epoch": 3.79, "learning_rate": 6.207872657413942e-05, "loss": 2.8514, "step": 48361 }, { "epoch": 3.79, "learning_rate": 6.207794244491492e-05, "loss": 3.0243, "step": 48362 }, { "epoch": 3.79, "learning_rate": 6.207715831569043e-05, "loss": 2.8437, "step": 48363 }, { "epoch": 3.79, "learning_rate": 6.207637418646593e-05, "loss": 2.622, "step": 48364 }, { "epoch": 3.79, "learning_rate": 6.207559005724143e-05, "loss": 2.7387, "step": 48365 }, { "epoch": 3.79, "learning_rate": 6.207480592801694e-05, "loss": 2.7544, "step": 48366 }, { "epoch": 3.79, "learning_rate": 6.207402179879244e-05, "loss": 2.6054, "step": 48367 }, { "epoch": 3.79, "learning_rate": 6.207323766956795e-05, "loss": 2.6667, "step": 48368 }, { "epoch": 3.79, "learning_rate": 6.207245354034345e-05, "loss": 2.8029, "step": 48369 }, { "epoch": 3.79, "learning_rate": 6.207166941111897e-05, "loss": 2.8161, "step": 48370 }, { "epoch": 3.79, "learning_rate": 6.207088528189447e-05, "loss": 2.6641, "step": 48371 }, { "epoch": 3.79, "learning_rate": 6.207010115266997e-05, "loss": 2.664, "step": 48372 }, { "epoch": 3.79, "learning_rate": 6.206931702344546e-05, "loss": 2.5163, "step": 48373 }, { "epoch": 3.79, "learning_rate": 6.206853289422097e-05, "loss": 2.6726, "step": 48374 }, { "epoch": 3.79, "learning_rate": 6.206774876499647e-05, "loss": 2.8399, "step": 48375 }, { "epoch": 3.79, "learning_rate": 6.206696463577198e-05, "loss": 2.8313, "step": 48376 }, { "epoch": 3.79, "learning_rate": 6.206618050654748e-05, "loss": 2.726, "step": 48377 }, { "epoch": 3.79, "learning_rate": 6.206539637732298e-05, "loss": 2.5766, "step": 48378 }, { "epoch": 3.79, "learning_rate": 6.206461224809849e-05, "loss": 3.1407, "step": 48379 }, { "epoch": 3.79, "learning_rate": 6.206382811887399e-05, "loss": 2.5923, "step": 48380 }, { "epoch": 3.79, "learning_rate": 6.20630439896495e-05, "loss": 2.7145, "step": 48381 }, { "epoch": 3.79, "learning_rate": 6.2062259860425e-05, "loss": 2.6195, "step": 48382 }, { "epoch": 3.79, "learning_rate": 6.20614757312005e-05, "loss": 2.6813, "step": 48383 }, { "epoch": 3.79, "learning_rate": 6.2060691601976e-05, "loss": 2.7409, "step": 48384 }, { "epoch": 3.79, "learning_rate": 6.205990747275151e-05, "loss": 2.8676, "step": 48385 }, { "epoch": 3.79, "learning_rate": 6.205912334352701e-05, "loss": 2.6891, "step": 48386 }, { "epoch": 3.79, "learning_rate": 6.205833921430252e-05, "loss": 2.5856, "step": 48387 }, { "epoch": 3.79, "learning_rate": 6.205755508507802e-05, "loss": 2.9332, "step": 48388 }, { "epoch": 3.79, "learning_rate": 6.205677095585352e-05, "loss": 2.7605, "step": 48389 }, { "epoch": 3.79, "learning_rate": 6.205598682662903e-05, "loss": 2.655, "step": 48390 }, { "epoch": 3.79, "learning_rate": 6.205520269740453e-05, "loss": 2.8347, "step": 48391 }, { "epoch": 3.79, "learning_rate": 6.205441856818003e-05, "loss": 2.7412, "step": 48392 }, { "epoch": 3.79, "learning_rate": 6.205363443895554e-05, "loss": 2.69, "step": 48393 }, { "epoch": 3.79, "learning_rate": 6.205285030973104e-05, "loss": 2.5809, "step": 48394 }, { "epoch": 3.79, "learning_rate": 6.205206618050656e-05, "loss": 2.9425, "step": 48395 }, { "epoch": 3.79, "learning_rate": 6.205128205128206e-05, "loss": 2.6577, "step": 48396 }, { "epoch": 3.79, "learning_rate": 6.205049792205757e-05, "loss": 2.7409, "step": 48397 }, { "epoch": 3.8, "learning_rate": 6.204971379283307e-05, "loss": 2.7089, "step": 48398 }, { "epoch": 3.8, "learning_rate": 6.204892966360857e-05, "loss": 2.8009, "step": 48399 }, { "epoch": 3.8, "learning_rate": 6.204814553438408e-05, "loss": 2.6307, "step": 48400 }, { "epoch": 3.8, "learning_rate": 6.204736140515958e-05, "loss": 2.5597, "step": 48401 }, { "epoch": 3.8, "learning_rate": 6.204657727593508e-05, "loss": 2.793, "step": 48402 }, { "epoch": 3.8, "learning_rate": 6.204579314671059e-05, "loss": 2.7634, "step": 48403 }, { "epoch": 3.8, "learning_rate": 6.204500901748609e-05, "loss": 2.8204, "step": 48404 }, { "epoch": 3.8, "learning_rate": 6.20442248882616e-05, "loss": 2.649, "step": 48405 }, { "epoch": 3.8, "learning_rate": 6.20434407590371e-05, "loss": 2.62, "step": 48406 }, { "epoch": 3.8, "learning_rate": 6.20426566298126e-05, "loss": 2.6538, "step": 48407 }, { "epoch": 3.8, "learning_rate": 6.20418725005881e-05, "loss": 2.5728, "step": 48408 }, { "epoch": 3.8, "learning_rate": 6.20410883713636e-05, "loss": 2.8139, "step": 48409 }, { "epoch": 3.8, "learning_rate": 6.20403042421391e-05, "loss": 2.7592, "step": 48410 }, { "epoch": 3.8, "learning_rate": 6.20395201129146e-05, "loss": 2.6171, "step": 48411 }, { "epoch": 3.8, "learning_rate": 6.203873598369011e-05, "loss": 2.8002, "step": 48412 }, { "epoch": 3.8, "learning_rate": 6.203795185446561e-05, "loss": 2.5846, "step": 48413 }, { "epoch": 3.8, "learning_rate": 6.203716772524111e-05, "loss": 2.6629, "step": 48414 }, { "epoch": 3.8, "learning_rate": 6.203638359601662e-05, "loss": 2.6204, "step": 48415 }, { "epoch": 3.8, "learning_rate": 6.203559946679212e-05, "loss": 2.7386, "step": 48416 }, { "epoch": 3.8, "learning_rate": 6.203481533756763e-05, "loss": 2.5619, "step": 48417 }, { "epoch": 3.8, "learning_rate": 6.203403120834313e-05, "loss": 2.7732, "step": 48418 }, { "epoch": 3.8, "learning_rate": 6.203324707911863e-05, "loss": 2.954, "step": 48419 }, { "epoch": 3.8, "learning_rate": 6.203246294989414e-05, "loss": 2.8499, "step": 48420 }, { "epoch": 3.8, "learning_rate": 6.203167882066965e-05, "loss": 2.5197, "step": 48421 }, { "epoch": 3.8, "learning_rate": 6.203089469144516e-05, "loss": 2.6385, "step": 48422 }, { "epoch": 3.8, "learning_rate": 6.203011056222066e-05, "loss": 2.8501, "step": 48423 }, { "epoch": 3.8, "learning_rate": 6.202932643299617e-05, "loss": 2.9131, "step": 48424 }, { "epoch": 3.8, "learning_rate": 6.202854230377167e-05, "loss": 2.8376, "step": 48425 }, { "epoch": 3.8, "learning_rate": 6.202775817454717e-05, "loss": 2.6508, "step": 48426 }, { "epoch": 3.8, "learning_rate": 6.202697404532268e-05, "loss": 2.7056, "step": 48427 }, { "epoch": 3.8, "learning_rate": 6.202618991609818e-05, "loss": 2.7866, "step": 48428 }, { "epoch": 3.8, "learning_rate": 6.202540578687368e-05, "loss": 2.7725, "step": 48429 }, { "epoch": 3.8, "learning_rate": 6.202462165764919e-05, "loss": 2.6072, "step": 48430 }, { "epoch": 3.8, "learning_rate": 6.202383752842469e-05, "loss": 2.6359, "step": 48431 }, { "epoch": 3.8, "learning_rate": 6.20230533992002e-05, "loss": 2.7466, "step": 48432 }, { "epoch": 3.8, "learning_rate": 6.20222692699757e-05, "loss": 2.776, "step": 48433 }, { "epoch": 3.8, "learning_rate": 6.20214851407512e-05, "loss": 2.5577, "step": 48434 }, { "epoch": 3.8, "learning_rate": 6.20207010115267e-05, "loss": 2.6713, "step": 48435 }, { "epoch": 3.8, "learning_rate": 6.201991688230221e-05, "loss": 2.8825, "step": 48436 }, { "epoch": 3.8, "learning_rate": 6.201913275307771e-05, "loss": 2.9394, "step": 48437 }, { "epoch": 3.8, "learning_rate": 6.201834862385322e-05, "loss": 2.6737, "step": 48438 }, { "epoch": 3.8, "learning_rate": 6.201756449462872e-05, "loss": 2.7671, "step": 48439 }, { "epoch": 3.8, "learning_rate": 6.201678036540422e-05, "loss": 2.9062, "step": 48440 }, { "epoch": 3.8, "learning_rate": 6.201599623617973e-05, "loss": 2.8557, "step": 48441 }, { "epoch": 3.8, "learning_rate": 6.201521210695523e-05, "loss": 2.6772, "step": 48442 }, { "epoch": 3.8, "learning_rate": 6.201442797773074e-05, "loss": 2.6106, "step": 48443 }, { "epoch": 3.8, "learning_rate": 6.201364384850624e-05, "loss": 2.7299, "step": 48444 }, { "epoch": 3.8, "learning_rate": 6.201285971928173e-05, "loss": 2.7593, "step": 48445 }, { "epoch": 3.8, "learning_rate": 6.201207559005723e-05, "loss": 2.7592, "step": 48446 }, { "epoch": 3.8, "learning_rate": 6.201129146083275e-05, "loss": 2.7241, "step": 48447 }, { "epoch": 3.8, "learning_rate": 6.201050733160825e-05, "loss": 2.8404, "step": 48448 }, { "epoch": 3.8, "learning_rate": 6.200972320238376e-05, "loss": 2.6305, "step": 48449 }, { "epoch": 3.8, "learning_rate": 6.200893907315926e-05, "loss": 2.6929, "step": 48450 }, { "epoch": 3.8, "learning_rate": 6.200815494393476e-05, "loss": 2.6416, "step": 48451 }, { "epoch": 3.8, "learning_rate": 6.200737081471027e-05, "loss": 2.7369, "step": 48452 }, { "epoch": 3.8, "learning_rate": 6.200658668548577e-05, "loss": 2.7424, "step": 48453 }, { "epoch": 3.8, "learning_rate": 6.200580255626128e-05, "loss": 2.8222, "step": 48454 }, { "epoch": 3.8, "learning_rate": 6.200501842703678e-05, "loss": 2.6734, "step": 48455 }, { "epoch": 3.8, "learning_rate": 6.200423429781228e-05, "loss": 2.7725, "step": 48456 }, { "epoch": 3.8, "learning_rate": 6.200345016858779e-05, "loss": 2.6266, "step": 48457 }, { "epoch": 3.8, "learning_rate": 6.200266603936329e-05, "loss": 2.9647, "step": 48458 }, { "epoch": 3.8, "learning_rate": 6.20018819101388e-05, "loss": 2.8388, "step": 48459 }, { "epoch": 3.8, "learning_rate": 6.20010977809143e-05, "loss": 2.7515, "step": 48460 }, { "epoch": 3.8, "learning_rate": 6.20003136516898e-05, "loss": 2.9301, "step": 48461 }, { "epoch": 3.8, "learning_rate": 6.19995295224653e-05, "loss": 2.5814, "step": 48462 }, { "epoch": 3.8, "learning_rate": 6.199874539324081e-05, "loss": 2.6927, "step": 48463 }, { "epoch": 3.8, "learning_rate": 6.199796126401631e-05, "loss": 2.6355, "step": 48464 }, { "epoch": 3.8, "learning_rate": 6.199717713479182e-05, "loss": 2.7699, "step": 48465 }, { "epoch": 3.8, "learning_rate": 6.199639300556732e-05, "loss": 2.8363, "step": 48466 }, { "epoch": 3.8, "learning_rate": 6.199560887634282e-05, "loss": 2.7762, "step": 48467 }, { "epoch": 3.8, "learning_rate": 6.199482474711833e-05, "loss": 2.9308, "step": 48468 }, { "epoch": 3.8, "learning_rate": 6.199404061789383e-05, "loss": 2.8181, "step": 48469 }, { "epoch": 3.8, "learning_rate": 6.199325648866933e-05, "loss": 2.7606, "step": 48470 }, { "epoch": 3.8, "learning_rate": 6.199247235944484e-05, "loss": 2.7897, "step": 48471 }, { "epoch": 3.8, "learning_rate": 6.199168823022034e-05, "loss": 2.7235, "step": 48472 }, { "epoch": 3.8, "learning_rate": 6.199090410099585e-05, "loss": 2.5365, "step": 48473 }, { "epoch": 3.8, "learning_rate": 6.199011997177135e-05, "loss": 2.7544, "step": 48474 }, { "epoch": 3.8, "learning_rate": 6.198933584254685e-05, "loss": 2.7856, "step": 48475 }, { "epoch": 3.8, "learning_rate": 6.198855171332236e-05, "loss": 2.8131, "step": 48476 }, { "epoch": 3.8, "learning_rate": 6.198776758409786e-05, "loss": 2.6447, "step": 48477 }, { "epoch": 3.8, "learning_rate": 6.198698345487338e-05, "loss": 2.7412, "step": 48478 }, { "epoch": 3.8, "learning_rate": 6.198619932564888e-05, "loss": 2.7772, "step": 48479 }, { "epoch": 3.8, "learning_rate": 6.198541519642438e-05, "loss": 2.7984, "step": 48480 }, { "epoch": 3.8, "learning_rate": 6.198463106719987e-05, "loss": 2.653, "step": 48481 }, { "epoch": 3.8, "learning_rate": 6.198384693797538e-05, "loss": 2.6546, "step": 48482 }, { "epoch": 3.8, "learning_rate": 6.198306280875088e-05, "loss": 2.676, "step": 48483 }, { "epoch": 3.8, "learning_rate": 6.198227867952639e-05, "loss": 2.6003, "step": 48484 }, { "epoch": 3.8, "learning_rate": 6.198149455030189e-05, "loss": 2.7037, "step": 48485 }, { "epoch": 3.8, "learning_rate": 6.198071042107739e-05, "loss": 2.9205, "step": 48486 }, { "epoch": 3.8, "learning_rate": 6.19799262918529e-05, "loss": 2.629, "step": 48487 }, { "epoch": 3.8, "learning_rate": 6.19791421626284e-05, "loss": 2.7319, "step": 48488 }, { "epoch": 3.8, "learning_rate": 6.19783580334039e-05, "loss": 2.7058, "step": 48489 }, { "epoch": 3.8, "learning_rate": 6.197757390417941e-05, "loss": 2.5914, "step": 48490 }, { "epoch": 3.8, "learning_rate": 6.197678977495491e-05, "loss": 2.6889, "step": 48491 }, { "epoch": 3.8, "learning_rate": 6.197600564573042e-05, "loss": 2.7271, "step": 48492 }, { "epoch": 3.8, "learning_rate": 6.197522151650592e-05, "loss": 2.8625, "step": 48493 }, { "epoch": 3.8, "learning_rate": 6.197443738728142e-05, "loss": 2.6096, "step": 48494 }, { "epoch": 3.8, "learning_rate": 6.197365325805693e-05, "loss": 2.5872, "step": 48495 }, { "epoch": 3.8, "learning_rate": 6.197286912883243e-05, "loss": 2.6161, "step": 48496 }, { "epoch": 3.8, "learning_rate": 6.197208499960793e-05, "loss": 2.8843, "step": 48497 }, { "epoch": 3.8, "learning_rate": 6.197130087038344e-05, "loss": 2.8931, "step": 48498 }, { "epoch": 3.8, "learning_rate": 6.197051674115894e-05, "loss": 2.8358, "step": 48499 }, { "epoch": 3.8, "learning_rate": 6.196973261193444e-05, "loss": 2.7694, "step": 48500 }, { "epoch": 3.8, "learning_rate": 6.196894848270995e-05, "loss": 2.6357, "step": 48501 }, { "epoch": 3.8, "learning_rate": 6.196816435348545e-05, "loss": 2.7926, "step": 48502 }, { "epoch": 3.8, "learning_rate": 6.196738022426096e-05, "loss": 2.6371, "step": 48503 }, { "epoch": 3.8, "learning_rate": 6.196659609503647e-05, "loss": 2.6441, "step": 48504 }, { "epoch": 3.8, "learning_rate": 6.196581196581198e-05, "loss": 2.5781, "step": 48505 }, { "epoch": 3.8, "learning_rate": 6.196502783658748e-05, "loss": 2.978, "step": 48506 }, { "epoch": 3.8, "learning_rate": 6.196424370736298e-05, "loss": 2.7604, "step": 48507 }, { "epoch": 3.8, "learning_rate": 6.196345957813849e-05, "loss": 2.7749, "step": 48508 }, { "epoch": 3.8, "learning_rate": 6.196267544891399e-05, "loss": 2.7043, "step": 48509 }, { "epoch": 3.8, "learning_rate": 6.19618913196895e-05, "loss": 2.7349, "step": 48510 }, { "epoch": 3.8, "learning_rate": 6.1961107190465e-05, "loss": 2.7592, "step": 48511 }, { "epoch": 3.8, "learning_rate": 6.19603230612405e-05, "loss": 2.7383, "step": 48512 }, { "epoch": 3.8, "learning_rate": 6.1959538932016e-05, "loss": 2.5347, "step": 48513 }, { "epoch": 3.8, "learning_rate": 6.195875480279151e-05, "loss": 2.7027, "step": 48514 }, { "epoch": 3.8, "learning_rate": 6.195797067356701e-05, "loss": 2.8407, "step": 48515 }, { "epoch": 3.8, "learning_rate": 6.195718654434252e-05, "loss": 2.6823, "step": 48516 }, { "epoch": 3.8, "learning_rate": 6.195640241511801e-05, "loss": 2.6857, "step": 48517 }, { "epoch": 3.8, "learning_rate": 6.195561828589351e-05, "loss": 2.6975, "step": 48518 }, { "epoch": 3.8, "learning_rate": 6.195483415666901e-05, "loss": 2.7164, "step": 48519 }, { "epoch": 3.8, "learning_rate": 6.195405002744452e-05, "loss": 2.7501, "step": 48520 }, { "epoch": 3.8, "learning_rate": 6.195326589822002e-05, "loss": 2.6369, "step": 48521 }, { "epoch": 3.8, "learning_rate": 6.195248176899553e-05, "loss": 2.9017, "step": 48522 }, { "epoch": 3.8, "learning_rate": 6.195169763977103e-05, "loss": 2.4136, "step": 48523 }, { "epoch": 3.8, "learning_rate": 6.195091351054653e-05, "loss": 2.6893, "step": 48524 }, { "epoch": 3.8, "learning_rate": 6.195012938132204e-05, "loss": 2.7934, "step": 48525 }, { "epoch": 3.81, "learning_rate": 6.194934525209754e-05, "loss": 2.6962, "step": 48526 }, { "epoch": 3.81, "learning_rate": 6.194856112287304e-05, "loss": 2.6924, "step": 48527 }, { "epoch": 3.81, "learning_rate": 6.194777699364855e-05, "loss": 2.8028, "step": 48528 }, { "epoch": 3.81, "learning_rate": 6.194699286442406e-05, "loss": 2.7921, "step": 48529 }, { "epoch": 3.81, "learning_rate": 6.194620873519957e-05, "loss": 2.7553, "step": 48530 }, { "epoch": 3.81, "learning_rate": 6.194542460597507e-05, "loss": 2.8644, "step": 48531 }, { "epoch": 3.81, "learning_rate": 6.194464047675058e-05, "loss": 2.799, "step": 48532 }, { "epoch": 3.81, "learning_rate": 6.194385634752608e-05, "loss": 2.864, "step": 48533 }, { "epoch": 3.81, "learning_rate": 6.194307221830158e-05, "loss": 2.7728, "step": 48534 }, { "epoch": 3.81, "learning_rate": 6.194228808907709e-05, "loss": 2.7887, "step": 48535 }, { "epoch": 3.81, "learning_rate": 6.194150395985259e-05, "loss": 2.6556, "step": 48536 }, { "epoch": 3.81, "learning_rate": 6.19407198306281e-05, "loss": 2.6282, "step": 48537 }, { "epoch": 3.81, "learning_rate": 6.19399357014036e-05, "loss": 2.656, "step": 48538 }, { "epoch": 3.81, "learning_rate": 6.19391515721791e-05, "loss": 2.7524, "step": 48539 }, { "epoch": 3.81, "learning_rate": 6.19383674429546e-05, "loss": 2.809, "step": 48540 }, { "epoch": 3.81, "learning_rate": 6.193758331373011e-05, "loss": 2.6281, "step": 48541 }, { "epoch": 3.81, "learning_rate": 6.193679918450561e-05, "loss": 2.8945, "step": 48542 }, { "epoch": 3.81, "learning_rate": 6.193601505528112e-05, "loss": 2.6394, "step": 48543 }, { "epoch": 3.81, "learning_rate": 6.193523092605662e-05, "loss": 2.7764, "step": 48544 }, { "epoch": 3.81, "learning_rate": 6.193444679683212e-05, "loss": 2.7193, "step": 48545 }, { "epoch": 3.81, "learning_rate": 6.193366266760763e-05, "loss": 2.6567, "step": 48546 }, { "epoch": 3.81, "learning_rate": 6.193287853838313e-05, "loss": 3.0066, "step": 48547 }, { "epoch": 3.81, "learning_rate": 6.193209440915863e-05, "loss": 2.6091, "step": 48548 }, { "epoch": 3.81, "learning_rate": 6.193131027993414e-05, "loss": 2.6376, "step": 48549 }, { "epoch": 3.81, "learning_rate": 6.193052615070964e-05, "loss": 2.7047, "step": 48550 }, { "epoch": 3.81, "learning_rate": 6.192974202148515e-05, "loss": 2.8594, "step": 48551 }, { "epoch": 3.81, "learning_rate": 6.192895789226065e-05, "loss": 2.6563, "step": 48552 }, { "epoch": 3.81, "learning_rate": 6.192817376303614e-05, "loss": 2.8839, "step": 48553 }, { "epoch": 3.81, "learning_rate": 6.192738963381164e-05, "loss": 2.5154, "step": 48554 }, { "epoch": 3.81, "learning_rate": 6.192660550458716e-05, "loss": 2.7329, "step": 48555 }, { "epoch": 3.81, "learning_rate": 6.192582137536266e-05, "loss": 2.8107, "step": 48556 }, { "epoch": 3.81, "learning_rate": 6.192503724613817e-05, "loss": 2.6142, "step": 48557 }, { "epoch": 3.81, "learning_rate": 6.192425311691367e-05, "loss": 2.7702, "step": 48558 }, { "epoch": 3.81, "learning_rate": 6.192346898768918e-05, "loss": 2.6625, "step": 48559 }, { "epoch": 3.81, "learning_rate": 6.192268485846468e-05, "loss": 2.8784, "step": 48560 }, { "epoch": 3.81, "learning_rate": 6.192190072924018e-05, "loss": 2.6891, "step": 48561 }, { "epoch": 3.81, "learning_rate": 6.192111660001569e-05, "loss": 2.9006, "step": 48562 }, { "epoch": 3.81, "learning_rate": 6.192033247079119e-05, "loss": 2.6695, "step": 48563 }, { "epoch": 3.81, "learning_rate": 6.19195483415667e-05, "loss": 2.7026, "step": 48564 }, { "epoch": 3.81, "learning_rate": 6.19187642123422e-05, "loss": 2.5917, "step": 48565 }, { "epoch": 3.81, "learning_rate": 6.19179800831177e-05, "loss": 2.7678, "step": 48566 }, { "epoch": 3.81, "learning_rate": 6.19171959538932e-05, "loss": 2.9495, "step": 48567 }, { "epoch": 3.81, "learning_rate": 6.191641182466871e-05, "loss": 2.6727, "step": 48568 }, { "epoch": 3.81, "learning_rate": 6.191562769544421e-05, "loss": 2.8254, "step": 48569 }, { "epoch": 3.81, "learning_rate": 6.191484356621972e-05, "loss": 2.559, "step": 48570 }, { "epoch": 3.81, "learning_rate": 6.191405943699522e-05, "loss": 2.6803, "step": 48571 }, { "epoch": 3.81, "learning_rate": 6.191327530777072e-05, "loss": 2.9407, "step": 48572 }, { "epoch": 3.81, "learning_rate": 6.191249117854623e-05, "loss": 2.7561, "step": 48573 }, { "epoch": 3.81, "learning_rate": 6.191170704932173e-05, "loss": 2.7668, "step": 48574 }, { "epoch": 3.81, "learning_rate": 6.191092292009723e-05, "loss": 2.79, "step": 48575 }, { "epoch": 3.81, "learning_rate": 6.191013879087274e-05, "loss": 2.8259, "step": 48576 }, { "epoch": 3.81, "learning_rate": 6.190935466164824e-05, "loss": 2.6404, "step": 48577 }, { "epoch": 3.81, "learning_rate": 6.190857053242375e-05, "loss": 2.609, "step": 48578 }, { "epoch": 3.81, "learning_rate": 6.190778640319925e-05, "loss": 2.9629, "step": 48579 }, { "epoch": 3.81, "learning_rate": 6.190700227397475e-05, "loss": 2.7662, "step": 48580 }, { "epoch": 3.81, "learning_rate": 6.190621814475026e-05, "loss": 2.7764, "step": 48581 }, { "epoch": 3.81, "learning_rate": 6.190543401552576e-05, "loss": 2.6286, "step": 48582 }, { "epoch": 3.81, "learning_rate": 6.190464988630126e-05, "loss": 2.6792, "step": 48583 }, { "epoch": 3.81, "learning_rate": 6.190386575707677e-05, "loss": 2.6009, "step": 48584 }, { "epoch": 3.81, "learning_rate": 6.190308162785227e-05, "loss": 2.8378, "step": 48585 }, { "epoch": 3.81, "learning_rate": 6.190229749862779e-05, "loss": 2.8034, "step": 48586 }, { "epoch": 3.81, "learning_rate": 6.190151336940329e-05, "loss": 2.7288, "step": 48587 }, { "epoch": 3.81, "learning_rate": 6.19007292401788e-05, "loss": 2.7011, "step": 48588 }, { "epoch": 3.81, "learning_rate": 6.189994511095429e-05, "loss": 2.6221, "step": 48589 }, { "epoch": 3.81, "learning_rate": 6.189916098172979e-05, "loss": 2.6525, "step": 48590 }, { "epoch": 3.81, "learning_rate": 6.189837685250529e-05, "loss": 2.5176, "step": 48591 }, { "epoch": 3.81, "learning_rate": 6.18975927232808e-05, "loss": 2.6838, "step": 48592 }, { "epoch": 3.81, "learning_rate": 6.18968085940563e-05, "loss": 2.6859, "step": 48593 }, { "epoch": 3.81, "learning_rate": 6.18960244648318e-05, "loss": 3.0054, "step": 48594 }, { "epoch": 3.81, "learning_rate": 6.189524033560731e-05, "loss": 2.7357, "step": 48595 }, { "epoch": 3.81, "learning_rate": 6.189445620638281e-05, "loss": 2.8107, "step": 48596 }, { "epoch": 3.81, "learning_rate": 6.189367207715831e-05, "loss": 2.7216, "step": 48597 }, { "epoch": 3.81, "learning_rate": 6.189288794793382e-05, "loss": 2.7467, "step": 48598 }, { "epoch": 3.81, "learning_rate": 6.189210381870932e-05, "loss": 2.5534, "step": 48599 }, { "epoch": 3.81, "learning_rate": 6.189131968948483e-05, "loss": 2.5916, "step": 48600 }, { "epoch": 3.81, "learning_rate": 6.189053556026033e-05, "loss": 2.6528, "step": 48601 }, { "epoch": 3.81, "learning_rate": 6.188975143103583e-05, "loss": 2.9994, "step": 48602 }, { "epoch": 3.81, "learning_rate": 6.188896730181134e-05, "loss": 2.7358, "step": 48603 }, { "epoch": 3.81, "learning_rate": 6.188818317258684e-05, "loss": 2.6321, "step": 48604 }, { "epoch": 3.81, "learning_rate": 6.188739904336234e-05, "loss": 2.6527, "step": 48605 }, { "epoch": 3.81, "learning_rate": 6.188661491413785e-05, "loss": 2.8272, "step": 48606 }, { "epoch": 3.81, "learning_rate": 6.188583078491335e-05, "loss": 2.9103, "step": 48607 }, { "epoch": 3.81, "learning_rate": 6.188504665568886e-05, "loss": 2.7079, "step": 48608 }, { "epoch": 3.81, "learning_rate": 6.188426252646436e-05, "loss": 2.9128, "step": 48609 }, { "epoch": 3.81, "learning_rate": 6.188347839723986e-05, "loss": 2.6485, "step": 48610 }, { "epoch": 3.81, "learning_rate": 6.188269426801537e-05, "loss": 2.7726, "step": 48611 }, { "epoch": 3.81, "learning_rate": 6.188191013879088e-05, "loss": 2.6525, "step": 48612 }, { "epoch": 3.81, "learning_rate": 6.188112600956639e-05, "loss": 2.9792, "step": 48613 }, { "epoch": 3.81, "learning_rate": 6.188034188034189e-05, "loss": 2.6159, "step": 48614 }, { "epoch": 3.81, "learning_rate": 6.18795577511174e-05, "loss": 2.7123, "step": 48615 }, { "epoch": 3.81, "learning_rate": 6.18787736218929e-05, "loss": 2.7775, "step": 48616 }, { "epoch": 3.81, "learning_rate": 6.18779894926684e-05, "loss": 2.8173, "step": 48617 }, { "epoch": 3.81, "learning_rate": 6.18772053634439e-05, "loss": 2.673, "step": 48618 }, { "epoch": 3.81, "learning_rate": 6.187642123421941e-05, "loss": 2.6536, "step": 48619 }, { "epoch": 3.81, "learning_rate": 6.187563710499491e-05, "loss": 2.8358, "step": 48620 }, { "epoch": 3.81, "learning_rate": 6.187485297577042e-05, "loss": 2.8667, "step": 48621 }, { "epoch": 3.81, "learning_rate": 6.187406884654592e-05, "loss": 2.6667, "step": 48622 }, { "epoch": 3.81, "learning_rate": 6.187328471732142e-05, "loss": 3.0311, "step": 48623 }, { "epoch": 3.81, "learning_rate": 6.187250058809693e-05, "loss": 2.6518, "step": 48624 }, { "epoch": 3.81, "learning_rate": 6.187171645887242e-05, "loss": 2.4833, "step": 48625 }, { "epoch": 3.81, "learning_rate": 6.187093232964792e-05, "loss": 2.8207, "step": 48626 }, { "epoch": 3.81, "learning_rate": 6.187014820042343e-05, "loss": 2.6675, "step": 48627 }, { "epoch": 3.81, "learning_rate": 6.186936407119893e-05, "loss": 3.0458, "step": 48628 }, { "epoch": 3.81, "learning_rate": 6.186857994197443e-05, "loss": 2.8064, "step": 48629 }, { "epoch": 3.81, "learning_rate": 6.186779581274994e-05, "loss": 2.8307, "step": 48630 }, { "epoch": 3.81, "learning_rate": 6.186701168352544e-05, "loss": 2.811, "step": 48631 }, { "epoch": 3.81, "learning_rate": 6.186622755430094e-05, "loss": 2.6551, "step": 48632 }, { "epoch": 3.81, "learning_rate": 6.186544342507645e-05, "loss": 2.7914, "step": 48633 }, { "epoch": 3.81, "learning_rate": 6.186465929585195e-05, "loss": 2.8162, "step": 48634 }, { "epoch": 3.81, "learning_rate": 6.186387516662745e-05, "loss": 2.6413, "step": 48635 }, { "epoch": 3.81, "learning_rate": 6.186309103740296e-05, "loss": 2.6414, "step": 48636 }, { "epoch": 3.81, "learning_rate": 6.186230690817848e-05, "loss": 2.6015, "step": 48637 }, { "epoch": 3.81, "learning_rate": 6.186152277895398e-05, "loss": 2.8968, "step": 48638 }, { "epoch": 3.81, "learning_rate": 6.186073864972948e-05, "loss": 2.6256, "step": 48639 }, { "epoch": 3.81, "learning_rate": 6.185995452050499e-05, "loss": 2.6798, "step": 48640 }, { "epoch": 3.81, "learning_rate": 6.185917039128049e-05, "loss": 2.6218, "step": 48641 }, { "epoch": 3.81, "learning_rate": 6.1858386262056e-05, "loss": 2.5622, "step": 48642 }, { "epoch": 3.81, "learning_rate": 6.18576021328315e-05, "loss": 2.6765, "step": 48643 }, { "epoch": 3.81, "learning_rate": 6.1856818003607e-05, "loss": 2.8199, "step": 48644 }, { "epoch": 3.81, "learning_rate": 6.18560338743825e-05, "loss": 2.9012, "step": 48645 }, { "epoch": 3.81, "learning_rate": 6.185524974515801e-05, "loss": 2.6679, "step": 48646 }, { "epoch": 3.81, "learning_rate": 6.185446561593351e-05, "loss": 2.6737, "step": 48647 }, { "epoch": 3.81, "learning_rate": 6.185368148670902e-05, "loss": 2.6219, "step": 48648 }, { "epoch": 3.81, "learning_rate": 6.185289735748452e-05, "loss": 2.5674, "step": 48649 }, { "epoch": 3.81, "learning_rate": 6.185211322826002e-05, "loss": 2.672, "step": 48650 }, { "epoch": 3.81, "learning_rate": 6.185132909903553e-05, "loss": 2.7922, "step": 48651 }, { "epoch": 3.81, "learning_rate": 6.185054496981103e-05, "loss": 2.8852, "step": 48652 }, { "epoch": 3.82, "learning_rate": 6.184976084058653e-05, "loss": 3.0242, "step": 48653 }, { "epoch": 3.82, "learning_rate": 6.184897671136204e-05, "loss": 2.6604, "step": 48654 }, { "epoch": 3.82, "learning_rate": 6.184819258213754e-05, "loss": 2.907, "step": 48655 }, { "epoch": 3.82, "learning_rate": 6.184740845291305e-05, "loss": 2.6861, "step": 48656 }, { "epoch": 3.82, "learning_rate": 6.184662432368855e-05, "loss": 2.7296, "step": 48657 }, { "epoch": 3.82, "learning_rate": 6.184584019446405e-05, "loss": 2.8555, "step": 48658 }, { "epoch": 3.82, "learning_rate": 6.184505606523956e-05, "loss": 2.7274, "step": 48659 }, { "epoch": 3.82, "learning_rate": 6.184427193601506e-05, "loss": 2.6558, "step": 48660 }, { "epoch": 3.82, "learning_rate": 6.184348780679055e-05, "loss": 2.6834, "step": 48661 }, { "epoch": 3.82, "learning_rate": 6.184270367756605e-05, "loss": 2.6403, "step": 48662 }, { "epoch": 3.82, "learning_rate": 6.184191954834157e-05, "loss": 2.7711, "step": 48663 }, { "epoch": 3.82, "learning_rate": 6.184113541911707e-05, "loss": 2.6866, "step": 48664 }, { "epoch": 3.82, "learning_rate": 6.184035128989258e-05, "loss": 2.6936, "step": 48665 }, { "epoch": 3.82, "learning_rate": 6.183956716066808e-05, "loss": 2.6646, "step": 48666 }, { "epoch": 3.82, "learning_rate": 6.183878303144359e-05, "loss": 2.7094, "step": 48667 }, { "epoch": 3.82, "learning_rate": 6.183799890221909e-05, "loss": 2.861, "step": 48668 }, { "epoch": 3.82, "learning_rate": 6.18372147729946e-05, "loss": 2.6056, "step": 48669 }, { "epoch": 3.82, "learning_rate": 6.18364306437701e-05, "loss": 2.8303, "step": 48670 }, { "epoch": 3.82, "learning_rate": 6.18356465145456e-05, "loss": 2.7798, "step": 48671 }, { "epoch": 3.82, "learning_rate": 6.18348623853211e-05, "loss": 2.8395, "step": 48672 }, { "epoch": 3.82, "learning_rate": 6.183407825609661e-05, "loss": 2.7421, "step": 48673 }, { "epoch": 3.82, "learning_rate": 6.183329412687211e-05, "loss": 2.7847, "step": 48674 }, { "epoch": 3.82, "learning_rate": 6.183250999764762e-05, "loss": 2.706, "step": 48675 }, { "epoch": 3.82, "learning_rate": 6.183172586842312e-05, "loss": 2.8133, "step": 48676 }, { "epoch": 3.82, "learning_rate": 6.183094173919862e-05, "loss": 2.7575, "step": 48677 }, { "epoch": 3.82, "learning_rate": 6.183015760997413e-05, "loss": 2.9223, "step": 48678 }, { "epoch": 3.82, "learning_rate": 6.182937348074963e-05, "loss": 3.0328, "step": 48679 }, { "epoch": 3.82, "learning_rate": 6.182858935152513e-05, "loss": 2.8285, "step": 48680 }, { "epoch": 3.82, "learning_rate": 6.182780522230064e-05, "loss": 2.714, "step": 48681 }, { "epoch": 3.82, "learning_rate": 6.182702109307614e-05, "loss": 2.7201, "step": 48682 }, { "epoch": 3.82, "learning_rate": 6.182623696385164e-05, "loss": 2.7159, "step": 48683 }, { "epoch": 3.82, "learning_rate": 6.182545283462715e-05, "loss": 2.6885, "step": 48684 }, { "epoch": 3.82, "learning_rate": 6.182466870540265e-05, "loss": 2.8121, "step": 48685 }, { "epoch": 3.82, "learning_rate": 6.182388457617816e-05, "loss": 2.7942, "step": 48686 }, { "epoch": 3.82, "learning_rate": 6.182310044695366e-05, "loss": 2.6138, "step": 48687 }, { "epoch": 3.82, "learning_rate": 6.182231631772916e-05, "loss": 2.7568, "step": 48688 }, { "epoch": 3.82, "learning_rate": 6.182153218850467e-05, "loss": 2.7369, "step": 48689 }, { "epoch": 3.82, "learning_rate": 6.182074805928017e-05, "loss": 2.704, "step": 48690 }, { "epoch": 3.82, "learning_rate": 6.181996393005567e-05, "loss": 2.8505, "step": 48691 }, { "epoch": 3.82, "learning_rate": 6.181917980083118e-05, "loss": 2.6739, "step": 48692 }, { "epoch": 3.82, "learning_rate": 6.181839567160668e-05, "loss": 2.562, "step": 48693 }, { "epoch": 3.82, "learning_rate": 6.18176115423822e-05, "loss": 2.9636, "step": 48694 }, { "epoch": 3.82, "learning_rate": 6.18168274131577e-05, "loss": 2.6673, "step": 48695 }, { "epoch": 3.82, "learning_rate": 6.18160432839332e-05, "loss": 2.771, "step": 48696 }, { "epoch": 3.82, "learning_rate": 6.18152591547087e-05, "loss": 2.4645, "step": 48697 }, { "epoch": 3.82, "learning_rate": 6.18144750254842e-05, "loss": 2.9291, "step": 48698 }, { "epoch": 3.82, "learning_rate": 6.18136908962597e-05, "loss": 2.797, "step": 48699 }, { "epoch": 3.82, "learning_rate": 6.181290676703521e-05, "loss": 2.6796, "step": 48700 }, { "epoch": 3.82, "learning_rate": 6.181212263781071e-05, "loss": 2.6696, "step": 48701 }, { "epoch": 3.82, "learning_rate": 6.181133850858621e-05, "loss": 3.0962, "step": 48702 }, { "epoch": 3.82, "learning_rate": 6.181055437936172e-05, "loss": 2.8372, "step": 48703 }, { "epoch": 3.82, "learning_rate": 6.180977025013722e-05, "loss": 2.6212, "step": 48704 }, { "epoch": 3.82, "learning_rate": 6.180898612091273e-05, "loss": 2.7787, "step": 48705 }, { "epoch": 3.82, "learning_rate": 6.180820199168823e-05, "loss": 2.9863, "step": 48706 }, { "epoch": 3.82, "learning_rate": 6.180741786246373e-05, "loss": 2.6592, "step": 48707 }, { "epoch": 3.82, "learning_rate": 6.180663373323924e-05, "loss": 2.9217, "step": 48708 }, { "epoch": 3.82, "learning_rate": 6.180584960401474e-05, "loss": 2.9031, "step": 48709 }, { "epoch": 3.82, "learning_rate": 6.180506547479024e-05, "loss": 2.7214, "step": 48710 }, { "epoch": 3.82, "learning_rate": 6.180428134556575e-05, "loss": 2.7595, "step": 48711 }, { "epoch": 3.82, "learning_rate": 6.180349721634125e-05, "loss": 2.6583, "step": 48712 }, { "epoch": 3.82, "learning_rate": 6.180271308711676e-05, "loss": 2.8412, "step": 48713 }, { "epoch": 3.82, "learning_rate": 6.180192895789226e-05, "loss": 2.6615, "step": 48714 }, { "epoch": 3.82, "learning_rate": 6.180114482866776e-05, "loss": 2.6954, "step": 48715 }, { "epoch": 3.82, "learning_rate": 6.180036069944327e-05, "loss": 2.5022, "step": 48716 }, { "epoch": 3.82, "learning_rate": 6.179957657021877e-05, "loss": 2.633, "step": 48717 }, { "epoch": 3.82, "learning_rate": 6.179879244099427e-05, "loss": 2.8566, "step": 48718 }, { "epoch": 3.82, "learning_rate": 6.179800831176978e-05, "loss": 2.7622, "step": 48719 }, { "epoch": 3.82, "learning_rate": 6.17972241825453e-05, "loss": 2.7307, "step": 48720 }, { "epoch": 3.82, "learning_rate": 6.17964400533208e-05, "loss": 2.6197, "step": 48721 }, { "epoch": 3.82, "learning_rate": 6.17956559240963e-05, "loss": 2.9117, "step": 48722 }, { "epoch": 3.82, "learning_rate": 6.17948717948718e-05, "loss": 2.6321, "step": 48723 }, { "epoch": 3.82, "learning_rate": 6.179408766564731e-05, "loss": 2.848, "step": 48724 }, { "epoch": 3.82, "learning_rate": 6.179330353642281e-05, "loss": 2.5257, "step": 48725 }, { "epoch": 3.82, "learning_rate": 6.179251940719832e-05, "loss": 2.6573, "step": 48726 }, { "epoch": 3.82, "learning_rate": 6.179173527797382e-05, "loss": 2.8275, "step": 48727 }, { "epoch": 3.82, "learning_rate": 6.179095114874932e-05, "loss": 2.8444, "step": 48728 }, { "epoch": 3.82, "learning_rate": 6.179016701952483e-05, "loss": 2.5735, "step": 48729 }, { "epoch": 3.82, "learning_rate": 6.178938289030033e-05, "loss": 2.8681, "step": 48730 }, { "epoch": 3.82, "learning_rate": 6.178859876107583e-05, "loss": 2.7211, "step": 48731 }, { "epoch": 3.82, "learning_rate": 6.178781463185134e-05, "loss": 2.711, "step": 48732 }, { "epoch": 3.82, "learning_rate": 6.178703050262683e-05, "loss": 2.7101, "step": 48733 }, { "epoch": 3.82, "learning_rate": 6.178624637340233e-05, "loss": 2.5408, "step": 48734 }, { "epoch": 3.82, "learning_rate": 6.178546224417784e-05, "loss": 2.6507, "step": 48735 }, { "epoch": 3.82, "learning_rate": 6.178467811495334e-05, "loss": 2.9739, "step": 48736 }, { "epoch": 3.82, "learning_rate": 6.178389398572884e-05, "loss": 2.8886, "step": 48737 }, { "epoch": 3.82, "learning_rate": 6.178310985650435e-05, "loss": 2.8447, "step": 48738 }, { "epoch": 3.82, "learning_rate": 6.178232572727985e-05, "loss": 2.5609, "step": 48739 }, { "epoch": 3.82, "learning_rate": 6.178154159805535e-05, "loss": 2.756, "step": 48740 }, { "epoch": 3.82, "learning_rate": 6.178075746883086e-05, "loss": 2.9178, "step": 48741 }, { "epoch": 3.82, "learning_rate": 6.177997333960636e-05, "loss": 2.8244, "step": 48742 }, { "epoch": 3.82, "learning_rate": 6.177918921038187e-05, "loss": 2.8619, "step": 48743 }, { "epoch": 3.82, "learning_rate": 6.177840508115737e-05, "loss": 2.6483, "step": 48744 }, { "epoch": 3.82, "learning_rate": 6.177762095193289e-05, "loss": 3.0603, "step": 48745 }, { "epoch": 3.82, "learning_rate": 6.177683682270839e-05, "loss": 2.7231, "step": 48746 }, { "epoch": 3.82, "learning_rate": 6.17760526934839e-05, "loss": 2.7434, "step": 48747 }, { "epoch": 3.82, "learning_rate": 6.17752685642594e-05, "loss": 2.8451, "step": 48748 }, { "epoch": 3.82, "learning_rate": 6.17744844350349e-05, "loss": 2.8148, "step": 48749 }, { "epoch": 3.82, "learning_rate": 6.17737003058104e-05, "loss": 2.7796, "step": 48750 }, { "epoch": 3.82, "learning_rate": 6.177291617658591e-05, "loss": 2.576, "step": 48751 }, { "epoch": 3.82, "learning_rate": 6.177213204736141e-05, "loss": 2.6018, "step": 48752 }, { "epoch": 3.82, "learning_rate": 6.177134791813692e-05, "loss": 2.5755, "step": 48753 }, { "epoch": 3.82, "learning_rate": 6.177056378891242e-05, "loss": 2.7543, "step": 48754 }, { "epoch": 3.82, "learning_rate": 6.176977965968792e-05, "loss": 2.7233, "step": 48755 }, { "epoch": 3.82, "learning_rate": 6.176899553046343e-05, "loss": 2.7242, "step": 48756 }, { "epoch": 3.82, "learning_rate": 6.176821140123893e-05, "loss": 2.5728, "step": 48757 }, { "epoch": 3.82, "learning_rate": 6.176742727201443e-05, "loss": 2.7424, "step": 48758 }, { "epoch": 3.82, "learning_rate": 6.176664314278994e-05, "loss": 2.6731, "step": 48759 }, { "epoch": 3.82, "learning_rate": 6.176585901356544e-05, "loss": 2.864, "step": 48760 }, { "epoch": 3.82, "learning_rate": 6.176507488434095e-05, "loss": 2.7814, "step": 48761 }, { "epoch": 3.82, "learning_rate": 6.176429075511645e-05, "loss": 2.6391, "step": 48762 }, { "epoch": 3.82, "learning_rate": 6.176350662589195e-05, "loss": 2.6879, "step": 48763 }, { "epoch": 3.82, "learning_rate": 6.176272249666746e-05, "loss": 2.6779, "step": 48764 }, { "epoch": 3.82, "learning_rate": 6.176193836744296e-05, "loss": 2.7642, "step": 48765 }, { "epoch": 3.82, "learning_rate": 6.176115423821846e-05, "loss": 2.6381, "step": 48766 }, { "epoch": 3.82, "learning_rate": 6.176037010899397e-05, "loss": 3.0692, "step": 48767 }, { "epoch": 3.82, "learning_rate": 6.175958597976947e-05, "loss": 2.6999, "step": 48768 }, { "epoch": 3.82, "learning_rate": 6.175880185054496e-05, "loss": 2.7887, "step": 48769 }, { "epoch": 3.82, "learning_rate": 6.175801772132046e-05, "loss": 2.6807, "step": 48770 }, { "epoch": 3.82, "learning_rate": 6.175723359209598e-05, "loss": 2.7427, "step": 48771 }, { "epoch": 3.82, "learning_rate": 6.175644946287149e-05, "loss": 2.8685, "step": 48772 }, { "epoch": 3.82, "learning_rate": 6.175566533364699e-05, "loss": 2.6446, "step": 48773 }, { "epoch": 3.82, "learning_rate": 6.175488120442249e-05, "loss": 2.8078, "step": 48774 }, { "epoch": 3.82, "learning_rate": 6.1754097075198e-05, "loss": 2.8156, "step": 48775 }, { "epoch": 3.82, "learning_rate": 6.17533129459735e-05, "loss": 2.6846, "step": 48776 }, { "epoch": 3.82, "learning_rate": 6.1752528816749e-05, "loss": 2.794, "step": 48777 }, { "epoch": 3.82, "learning_rate": 6.175174468752451e-05, "loss": 2.8377, "step": 48778 }, { "epoch": 3.82, "learning_rate": 6.175096055830001e-05, "loss": 2.6942, "step": 48779 }, { "epoch": 3.82, "learning_rate": 6.175017642907552e-05, "loss": 3.0147, "step": 48780 }, { "epoch": 3.83, "learning_rate": 6.174939229985102e-05, "loss": 2.8924, "step": 48781 }, { "epoch": 3.83, "learning_rate": 6.174860817062652e-05, "loss": 2.6672, "step": 48782 }, { "epoch": 3.83, "learning_rate": 6.174782404140203e-05, "loss": 2.7408, "step": 48783 }, { "epoch": 3.83, "learning_rate": 6.174703991217753e-05, "loss": 2.8515, "step": 48784 }, { "epoch": 3.83, "learning_rate": 6.174625578295303e-05, "loss": 2.9102, "step": 48785 }, { "epoch": 3.83, "learning_rate": 6.174547165372854e-05, "loss": 2.6256, "step": 48786 }, { "epoch": 3.83, "learning_rate": 6.174468752450404e-05, "loss": 2.7894, "step": 48787 }, { "epoch": 3.83, "learning_rate": 6.174390339527954e-05, "loss": 2.5425, "step": 48788 }, { "epoch": 3.83, "learning_rate": 6.174311926605505e-05, "loss": 2.7475, "step": 48789 }, { "epoch": 3.83, "learning_rate": 6.174233513683055e-05, "loss": 2.8184, "step": 48790 }, { "epoch": 3.83, "learning_rate": 6.174155100760606e-05, "loss": 2.8537, "step": 48791 }, { "epoch": 3.83, "learning_rate": 6.174076687838156e-05, "loss": 2.6704, "step": 48792 }, { "epoch": 3.83, "learning_rate": 6.173998274915706e-05, "loss": 2.5518, "step": 48793 }, { "epoch": 3.83, "learning_rate": 6.173919861993257e-05, "loss": 2.8395, "step": 48794 }, { "epoch": 3.83, "learning_rate": 6.173841449070807e-05, "loss": 2.7048, "step": 48795 }, { "epoch": 3.83, "learning_rate": 6.173763036148357e-05, "loss": 2.7189, "step": 48796 }, { "epoch": 3.83, "learning_rate": 6.173684623225908e-05, "loss": 2.7486, "step": 48797 }, { "epoch": 3.83, "learning_rate": 6.173606210303458e-05, "loss": 3.0009, "step": 48798 }, { "epoch": 3.83, "learning_rate": 6.173527797381008e-05, "loss": 2.7411, "step": 48799 }, { "epoch": 3.83, "learning_rate": 6.173449384458559e-05, "loss": 2.6765, "step": 48800 }, { "epoch": 3.83, "learning_rate": 6.173370971536109e-05, "loss": 2.8335, "step": 48801 }, { "epoch": 3.83, "learning_rate": 6.173292558613661e-05, "loss": 2.7303, "step": 48802 }, { "epoch": 3.83, "learning_rate": 6.173214145691211e-05, "loss": 2.8442, "step": 48803 }, { "epoch": 3.83, "learning_rate": 6.173135732768762e-05, "loss": 2.7439, "step": 48804 }, { "epoch": 3.83, "learning_rate": 6.173057319846311e-05, "loss": 2.7803, "step": 48805 }, { "epoch": 3.83, "learning_rate": 6.172978906923861e-05, "loss": 2.7841, "step": 48806 }, { "epoch": 3.83, "learning_rate": 6.172900494001411e-05, "loss": 2.8766, "step": 48807 }, { "epoch": 3.83, "learning_rate": 6.172822081078962e-05, "loss": 2.6833, "step": 48808 }, { "epoch": 3.83, "learning_rate": 6.172743668156512e-05, "loss": 2.8395, "step": 48809 }, { "epoch": 3.83, "learning_rate": 6.172665255234063e-05, "loss": 2.518, "step": 48810 }, { "epoch": 3.83, "learning_rate": 6.172586842311613e-05, "loss": 2.6356, "step": 48811 }, { "epoch": 3.83, "learning_rate": 6.172508429389163e-05, "loss": 2.9404, "step": 48812 }, { "epoch": 3.83, "learning_rate": 6.172430016466714e-05, "loss": 2.6748, "step": 48813 }, { "epoch": 3.83, "learning_rate": 6.172351603544264e-05, "loss": 2.8988, "step": 48814 }, { "epoch": 3.83, "learning_rate": 6.172273190621814e-05, "loss": 2.9759, "step": 48815 }, { "epoch": 3.83, "learning_rate": 6.172194777699365e-05, "loss": 2.7015, "step": 48816 }, { "epoch": 3.83, "learning_rate": 6.172116364776915e-05, "loss": 2.6803, "step": 48817 }, { "epoch": 3.83, "learning_rate": 6.172037951854465e-05, "loss": 2.7281, "step": 48818 }, { "epoch": 3.83, "learning_rate": 6.171959538932016e-05, "loss": 2.5628, "step": 48819 }, { "epoch": 3.83, "learning_rate": 6.171881126009566e-05, "loss": 2.7121, "step": 48820 }, { "epoch": 3.83, "learning_rate": 6.171802713087117e-05, "loss": 2.7752, "step": 48821 }, { "epoch": 3.83, "learning_rate": 6.171724300164667e-05, "loss": 2.8154, "step": 48822 }, { "epoch": 3.83, "learning_rate": 6.171645887242217e-05, "loss": 2.665, "step": 48823 }, { "epoch": 3.83, "learning_rate": 6.171567474319768e-05, "loss": 2.6318, "step": 48824 }, { "epoch": 3.83, "learning_rate": 6.171489061397318e-05, "loss": 2.7558, "step": 48825 }, { "epoch": 3.83, "learning_rate": 6.171410648474868e-05, "loss": 2.7155, "step": 48826 }, { "epoch": 3.83, "learning_rate": 6.171332235552419e-05, "loss": 2.9016, "step": 48827 }, { "epoch": 3.83, "learning_rate": 6.17125382262997e-05, "loss": 2.9363, "step": 48828 }, { "epoch": 3.83, "learning_rate": 6.171175409707521e-05, "loss": 2.4644, "step": 48829 }, { "epoch": 3.83, "learning_rate": 6.171096996785071e-05, "loss": 2.4804, "step": 48830 }, { "epoch": 3.83, "learning_rate": 6.171018583862622e-05, "loss": 2.9066, "step": 48831 }, { "epoch": 3.83, "learning_rate": 6.170940170940172e-05, "loss": 2.6225, "step": 48832 }, { "epoch": 3.83, "learning_rate": 6.170861758017722e-05, "loss": 2.7992, "step": 48833 }, { "epoch": 3.83, "learning_rate": 6.170783345095273e-05, "loss": 2.7032, "step": 48834 }, { "epoch": 3.83, "learning_rate": 6.170704932172823e-05, "loss": 2.9358, "step": 48835 }, { "epoch": 3.83, "learning_rate": 6.170626519250373e-05, "loss": 2.65, "step": 48836 }, { "epoch": 3.83, "learning_rate": 6.170548106327924e-05, "loss": 2.7463, "step": 48837 }, { "epoch": 3.83, "learning_rate": 6.170469693405474e-05, "loss": 2.6901, "step": 48838 }, { "epoch": 3.83, "learning_rate": 6.170391280483025e-05, "loss": 2.9736, "step": 48839 }, { "epoch": 3.83, "learning_rate": 6.170312867560575e-05, "loss": 2.7142, "step": 48840 }, { "epoch": 3.83, "learning_rate": 6.170234454638124e-05, "loss": 2.9159, "step": 48841 }, { "epoch": 3.83, "learning_rate": 6.170156041715674e-05, "loss": 2.7601, "step": 48842 }, { "epoch": 3.83, "learning_rate": 6.170077628793225e-05, "loss": 2.712, "step": 48843 }, { "epoch": 3.83, "learning_rate": 6.169999215870775e-05, "loss": 2.5888, "step": 48844 }, { "epoch": 3.83, "learning_rate": 6.169920802948325e-05, "loss": 2.6955, "step": 48845 }, { "epoch": 3.83, "learning_rate": 6.169842390025876e-05, "loss": 2.5911, "step": 48846 }, { "epoch": 3.83, "learning_rate": 6.169763977103426e-05, "loss": 2.7182, "step": 48847 }, { "epoch": 3.83, "learning_rate": 6.169685564180977e-05, "loss": 2.6217, "step": 48848 }, { "epoch": 3.83, "learning_rate": 6.169607151258527e-05, "loss": 2.7449, "step": 48849 }, { "epoch": 3.83, "learning_rate": 6.169528738336077e-05, "loss": 2.5857, "step": 48850 }, { "epoch": 3.83, "learning_rate": 6.169450325413628e-05, "loss": 2.6977, "step": 48851 }, { "epoch": 3.83, "learning_rate": 6.169371912491178e-05, "loss": 2.6628, "step": 48852 }, { "epoch": 3.83, "learning_rate": 6.16929349956873e-05, "loss": 2.7055, "step": 48853 }, { "epoch": 3.83, "learning_rate": 6.16921508664628e-05, "loss": 2.7796, "step": 48854 }, { "epoch": 3.83, "learning_rate": 6.16913667372383e-05, "loss": 2.5679, "step": 48855 }, { "epoch": 3.83, "learning_rate": 6.169058260801381e-05, "loss": 2.7224, "step": 48856 }, { "epoch": 3.83, "learning_rate": 6.168979847878931e-05, "loss": 2.8082, "step": 48857 }, { "epoch": 3.83, "learning_rate": 6.168901434956482e-05, "loss": 2.8618, "step": 48858 }, { "epoch": 3.83, "learning_rate": 6.168823022034032e-05, "loss": 2.5485, "step": 48859 }, { "epoch": 3.83, "learning_rate": 6.168744609111582e-05, "loss": 2.6925, "step": 48860 }, { "epoch": 3.83, "learning_rate": 6.168666196189133e-05, "loss": 2.6775, "step": 48861 }, { "epoch": 3.83, "learning_rate": 6.168587783266683e-05, "loss": 2.659, "step": 48862 }, { "epoch": 3.83, "learning_rate": 6.168509370344233e-05, "loss": 2.4887, "step": 48863 }, { "epoch": 3.83, "learning_rate": 6.168430957421784e-05, "loss": 2.8208, "step": 48864 }, { "epoch": 3.83, "learning_rate": 6.168352544499334e-05, "loss": 2.6049, "step": 48865 }, { "epoch": 3.83, "learning_rate": 6.168274131576884e-05, "loss": 2.6522, "step": 48866 }, { "epoch": 3.83, "learning_rate": 6.168195718654435e-05, "loss": 2.6416, "step": 48867 }, { "epoch": 3.83, "learning_rate": 6.168117305731985e-05, "loss": 2.8128, "step": 48868 }, { "epoch": 3.83, "learning_rate": 6.168038892809536e-05, "loss": 2.7333, "step": 48869 }, { "epoch": 3.83, "learning_rate": 6.167960479887086e-05, "loss": 2.8795, "step": 48870 }, { "epoch": 3.83, "learning_rate": 6.167882066964636e-05, "loss": 2.712, "step": 48871 }, { "epoch": 3.83, "learning_rate": 6.167803654042187e-05, "loss": 2.8439, "step": 48872 }, { "epoch": 3.83, "learning_rate": 6.167725241119737e-05, "loss": 2.826, "step": 48873 }, { "epoch": 3.83, "learning_rate": 6.167646828197287e-05, "loss": 2.8347, "step": 48874 }, { "epoch": 3.83, "learning_rate": 6.167568415274838e-05, "loss": 2.929, "step": 48875 }, { "epoch": 3.83, "learning_rate": 6.167490002352388e-05, "loss": 2.8536, "step": 48876 }, { "epoch": 3.83, "learning_rate": 6.167411589429937e-05, "loss": 2.6659, "step": 48877 }, { "epoch": 3.83, "learning_rate": 6.167333176507488e-05, "loss": 2.7188, "step": 48878 }, { "epoch": 3.83, "learning_rate": 6.167254763585039e-05, "loss": 2.627, "step": 48879 }, { "epoch": 3.83, "learning_rate": 6.16717635066259e-05, "loss": 2.839, "step": 48880 }, { "epoch": 3.83, "learning_rate": 6.16709793774014e-05, "loss": 2.6241, "step": 48881 }, { "epoch": 3.83, "learning_rate": 6.16701952481769e-05, "loss": 2.734, "step": 48882 }, { "epoch": 3.83, "learning_rate": 6.166941111895241e-05, "loss": 2.58, "step": 48883 }, { "epoch": 3.83, "learning_rate": 6.166862698972791e-05, "loss": 2.7591, "step": 48884 }, { "epoch": 3.83, "learning_rate": 6.166784286050341e-05, "loss": 2.5905, "step": 48885 }, { "epoch": 3.83, "learning_rate": 6.166705873127892e-05, "loss": 2.7271, "step": 48886 }, { "epoch": 3.83, "learning_rate": 6.166627460205442e-05, "loss": 2.7113, "step": 48887 }, { "epoch": 3.83, "learning_rate": 6.166549047282993e-05, "loss": 2.8284, "step": 48888 }, { "epoch": 3.83, "learning_rate": 6.166470634360543e-05, "loss": 2.7889, "step": 48889 }, { "epoch": 3.83, "learning_rate": 6.166392221438093e-05, "loss": 2.5255, "step": 48890 }, { "epoch": 3.83, "learning_rate": 6.166313808515644e-05, "loss": 2.7612, "step": 48891 }, { "epoch": 3.83, "learning_rate": 6.166235395593194e-05, "loss": 2.6202, "step": 48892 }, { "epoch": 3.83, "learning_rate": 6.166156982670744e-05, "loss": 2.5088, "step": 48893 }, { "epoch": 3.83, "learning_rate": 6.166078569748295e-05, "loss": 2.7675, "step": 48894 }, { "epoch": 3.83, "learning_rate": 6.166000156825845e-05, "loss": 2.5649, "step": 48895 }, { "epoch": 3.83, "learning_rate": 6.165921743903396e-05, "loss": 2.7352, "step": 48896 }, { "epoch": 3.83, "learning_rate": 6.165843330980946e-05, "loss": 2.5869, "step": 48897 }, { "epoch": 3.83, "learning_rate": 6.165764918058496e-05, "loss": 2.7883, "step": 48898 }, { "epoch": 3.83, "learning_rate": 6.165686505136047e-05, "loss": 2.7033, "step": 48899 }, { "epoch": 3.83, "learning_rate": 6.165608092213597e-05, "loss": 2.7319, "step": 48900 }, { "epoch": 3.83, "learning_rate": 6.165529679291147e-05, "loss": 2.6687, "step": 48901 }, { "epoch": 3.83, "learning_rate": 6.165451266368698e-05, "loss": 2.677, "step": 48902 }, { "epoch": 3.83, "learning_rate": 6.165372853446248e-05, "loss": 2.6574, "step": 48903 }, { "epoch": 3.83, "learning_rate": 6.165294440523798e-05, "loss": 2.6983, "step": 48904 }, { "epoch": 3.83, "learning_rate": 6.165216027601349e-05, "loss": 2.7502, "step": 48905 }, { "epoch": 3.83, "learning_rate": 6.165137614678899e-05, "loss": 2.8638, "step": 48906 }, { "epoch": 3.83, "learning_rate": 6.16505920175645e-05, "loss": 2.7202, "step": 48907 }, { "epoch": 3.84, "learning_rate": 6.164980788834e-05, "loss": 2.8182, "step": 48908 }, { "epoch": 3.84, "learning_rate": 6.16490237591155e-05, "loss": 2.7588, "step": 48909 }, { "epoch": 3.84, "learning_rate": 6.164823962989102e-05, "loss": 2.7039, "step": 48910 }, { "epoch": 3.84, "learning_rate": 6.164745550066652e-05, "loss": 2.7034, "step": 48911 }, { "epoch": 3.84, "learning_rate": 6.164667137144203e-05, "loss": 2.7038, "step": 48912 }, { "epoch": 3.84, "learning_rate": 6.164588724221752e-05, "loss": 2.6847, "step": 48913 }, { "epoch": 3.84, "learning_rate": 6.164510311299302e-05, "loss": 2.8933, "step": 48914 }, { "epoch": 3.84, "learning_rate": 6.164431898376853e-05, "loss": 3.03, "step": 48915 }, { "epoch": 3.84, "learning_rate": 6.164353485454403e-05, "loss": 2.7311, "step": 48916 }, { "epoch": 3.84, "learning_rate": 6.164275072531953e-05, "loss": 2.6483, "step": 48917 }, { "epoch": 3.84, "learning_rate": 6.164196659609504e-05, "loss": 2.6079, "step": 48918 }, { "epoch": 3.84, "learning_rate": 6.164118246687054e-05, "loss": 2.6485, "step": 48919 }, { "epoch": 3.84, "learning_rate": 6.164039833764604e-05, "loss": 2.6316, "step": 48920 }, { "epoch": 3.84, "learning_rate": 6.163961420842155e-05, "loss": 2.7828, "step": 48921 }, { "epoch": 3.84, "learning_rate": 6.163883007919705e-05, "loss": 2.863, "step": 48922 }, { "epoch": 3.84, "learning_rate": 6.163804594997255e-05, "loss": 2.5261, "step": 48923 }, { "epoch": 3.84, "learning_rate": 6.163726182074806e-05, "loss": 2.8694, "step": 48924 }, { "epoch": 3.84, "learning_rate": 6.163647769152356e-05, "loss": 2.9153, "step": 48925 }, { "epoch": 3.84, "learning_rate": 6.163569356229907e-05, "loss": 2.788, "step": 48926 }, { "epoch": 3.84, "learning_rate": 6.163490943307457e-05, "loss": 2.8949, "step": 48927 }, { "epoch": 3.84, "learning_rate": 6.163412530385007e-05, "loss": 2.7456, "step": 48928 }, { "epoch": 3.84, "learning_rate": 6.163334117462558e-05, "loss": 2.5898, "step": 48929 }, { "epoch": 3.84, "learning_rate": 6.163255704540108e-05, "loss": 2.8526, "step": 48930 }, { "epoch": 3.84, "learning_rate": 6.163177291617658e-05, "loss": 2.6923, "step": 48931 }, { "epoch": 3.84, "learning_rate": 6.163098878695209e-05, "loss": 2.8783, "step": 48932 }, { "epoch": 3.84, "learning_rate": 6.163020465772759e-05, "loss": 2.518, "step": 48933 }, { "epoch": 3.84, "learning_rate": 6.16294205285031e-05, "loss": 2.7967, "step": 48934 }, { "epoch": 3.84, "learning_rate": 6.16286363992786e-05, "loss": 2.7097, "step": 48935 }, { "epoch": 3.84, "learning_rate": 6.162785227005412e-05, "loss": 2.8965, "step": 48936 }, { "epoch": 3.84, "learning_rate": 6.162706814082962e-05, "loss": 2.756, "step": 48937 }, { "epoch": 3.84, "learning_rate": 6.162628401160512e-05, "loss": 2.8093, "step": 48938 }, { "epoch": 3.84, "learning_rate": 6.162549988238063e-05, "loss": 2.7026, "step": 48939 }, { "epoch": 3.84, "learning_rate": 6.162471575315613e-05, "loss": 2.7354, "step": 48940 }, { "epoch": 3.84, "learning_rate": 6.162393162393163e-05, "loss": 2.3875, "step": 48941 }, { "epoch": 3.84, "learning_rate": 6.162314749470714e-05, "loss": 2.8095, "step": 48942 }, { "epoch": 3.84, "learning_rate": 6.162236336548264e-05, "loss": 2.7802, "step": 48943 }, { "epoch": 3.84, "learning_rate": 6.162157923625815e-05, "loss": 2.7982, "step": 48944 }, { "epoch": 3.84, "learning_rate": 6.162079510703365e-05, "loss": 2.8552, "step": 48945 }, { "epoch": 3.84, "learning_rate": 6.162001097780915e-05, "loss": 2.912, "step": 48946 }, { "epoch": 3.84, "learning_rate": 6.161922684858466e-05, "loss": 2.6445, "step": 48947 }, { "epoch": 3.84, "learning_rate": 6.161844271936016e-05, "loss": 2.5483, "step": 48948 }, { "epoch": 3.84, "learning_rate": 6.161765859013565e-05, "loss": 2.5204, "step": 48949 }, { "epoch": 3.84, "learning_rate": 6.161687446091115e-05, "loss": 2.8358, "step": 48950 }, { "epoch": 3.84, "learning_rate": 6.161609033168666e-05, "loss": 2.8379, "step": 48951 }, { "epoch": 3.84, "learning_rate": 6.161530620246216e-05, "loss": 2.7728, "step": 48952 }, { "epoch": 3.84, "learning_rate": 6.161452207323766e-05, "loss": 2.9273, "step": 48953 }, { "epoch": 3.84, "learning_rate": 6.161373794401317e-05, "loss": 2.818, "step": 48954 }, { "epoch": 3.84, "learning_rate": 6.161295381478867e-05, "loss": 2.7688, "step": 48955 }, { "epoch": 3.84, "learning_rate": 6.161216968556418e-05, "loss": 2.6189, "step": 48956 }, { "epoch": 3.84, "learning_rate": 6.161138555633968e-05, "loss": 2.6603, "step": 48957 }, { "epoch": 3.84, "learning_rate": 6.161060142711518e-05, "loss": 3.0415, "step": 48958 }, { "epoch": 3.84, "learning_rate": 6.160981729789069e-05, "loss": 2.6725, "step": 48959 }, { "epoch": 3.84, "learning_rate": 6.160903316866619e-05, "loss": 2.8546, "step": 48960 }, { "epoch": 3.84, "learning_rate": 6.160824903944171e-05, "loss": 2.9654, "step": 48961 }, { "epoch": 3.84, "learning_rate": 6.160746491021721e-05, "loss": 2.8169, "step": 48962 }, { "epoch": 3.84, "learning_rate": 6.160668078099272e-05, "loss": 2.8249, "step": 48963 }, { "epoch": 3.84, "learning_rate": 6.160589665176822e-05, "loss": 2.686, "step": 48964 }, { "epoch": 3.84, "learning_rate": 6.160511252254372e-05, "loss": 2.8727, "step": 48965 }, { "epoch": 3.84, "learning_rate": 6.160432839331923e-05, "loss": 2.5368, "step": 48966 }, { "epoch": 3.84, "learning_rate": 6.160354426409473e-05, "loss": 2.7123, "step": 48967 }, { "epoch": 3.84, "learning_rate": 6.160276013487023e-05, "loss": 2.8122, "step": 48968 }, { "epoch": 3.84, "learning_rate": 6.160197600564574e-05, "loss": 2.6351, "step": 48969 }, { "epoch": 3.84, "learning_rate": 6.160119187642124e-05, "loss": 2.7861, "step": 48970 }, { "epoch": 3.84, "learning_rate": 6.160040774719674e-05, "loss": 2.9039, "step": 48971 }, { "epoch": 3.84, "learning_rate": 6.159962361797225e-05, "loss": 2.6755, "step": 48972 }, { "epoch": 3.84, "learning_rate": 6.159883948874775e-05, "loss": 2.618, "step": 48973 }, { "epoch": 3.84, "learning_rate": 6.159805535952326e-05, "loss": 2.7034, "step": 48974 }, { "epoch": 3.84, "learning_rate": 6.159727123029876e-05, "loss": 2.8521, "step": 48975 }, { "epoch": 3.84, "learning_rate": 6.159648710107426e-05, "loss": 2.7305, "step": 48976 }, { "epoch": 3.84, "learning_rate": 6.159570297184977e-05, "loss": 2.4612, "step": 48977 }, { "epoch": 3.84, "learning_rate": 6.159491884262527e-05, "loss": 2.8048, "step": 48978 }, { "epoch": 3.84, "learning_rate": 6.159413471340077e-05, "loss": 2.7327, "step": 48979 }, { "epoch": 3.84, "learning_rate": 6.159335058417628e-05, "loss": 2.7991, "step": 48980 }, { "epoch": 3.84, "learning_rate": 6.159256645495178e-05, "loss": 2.5059, "step": 48981 }, { "epoch": 3.84, "learning_rate": 6.159178232572729e-05, "loss": 2.709, "step": 48982 }, { "epoch": 3.84, "learning_rate": 6.159099819650279e-05, "loss": 2.7839, "step": 48983 }, { "epoch": 3.84, "learning_rate": 6.159021406727829e-05, "loss": 2.7488, "step": 48984 }, { "epoch": 3.84, "learning_rate": 6.158942993805378e-05, "loss": 2.5497, "step": 48985 }, { "epoch": 3.84, "learning_rate": 6.158864580882929e-05, "loss": 2.5503, "step": 48986 }, { "epoch": 3.84, "learning_rate": 6.15878616796048e-05, "loss": 2.7356, "step": 48987 }, { "epoch": 3.84, "learning_rate": 6.158707755038031e-05, "loss": 2.7245, "step": 48988 }, { "epoch": 3.84, "learning_rate": 6.158629342115581e-05, "loss": 2.8087, "step": 48989 }, { "epoch": 3.84, "learning_rate": 6.158550929193131e-05, "loss": 2.7783, "step": 48990 }, { "epoch": 3.84, "learning_rate": 6.158472516270682e-05, "loss": 2.8981, "step": 48991 }, { "epoch": 3.84, "learning_rate": 6.158394103348232e-05, "loss": 2.7871, "step": 48992 }, { "epoch": 3.84, "learning_rate": 6.158315690425783e-05, "loss": 2.7628, "step": 48993 }, { "epoch": 3.84, "learning_rate": 6.158237277503333e-05, "loss": 2.5792, "step": 48994 }, { "epoch": 3.84, "learning_rate": 6.158158864580883e-05, "loss": 2.7628, "step": 48995 }, { "epoch": 3.84, "learning_rate": 6.158080451658434e-05, "loss": 2.7065, "step": 48996 }, { "epoch": 3.84, "learning_rate": 6.158002038735984e-05, "loss": 2.7624, "step": 48997 }, { "epoch": 3.84, "learning_rate": 6.157923625813534e-05, "loss": 2.6318, "step": 48998 }, { "epoch": 3.84, "learning_rate": 6.157845212891085e-05, "loss": 2.684, "step": 48999 }, { "epoch": 3.84, "learning_rate": 6.157766799968635e-05, "loss": 2.6075, "step": 49000 }, { "epoch": 3.84, "eval_loss": 2.889347553253174, "eval_runtime": 217.9105, "eval_samples_per_second": 52.003, "eval_steps_per_second": 3.254, "step": 49000 }, { "epoch": 3.84, "learning_rate": 6.157688387046185e-05, "loss": 2.8564, "step": 49001 }, { "epoch": 3.84, "learning_rate": 6.157609974123736e-05, "loss": 2.7882, "step": 49002 }, { "epoch": 3.84, "learning_rate": 6.157531561201286e-05, "loss": 2.5841, "step": 49003 }, { "epoch": 3.84, "learning_rate": 6.157453148278837e-05, "loss": 2.751, "step": 49004 }, { "epoch": 3.84, "learning_rate": 6.157374735356387e-05, "loss": 2.7978, "step": 49005 }, { "epoch": 3.84, "learning_rate": 6.157296322433937e-05, "loss": 2.7326, "step": 49006 }, { "epoch": 3.84, "learning_rate": 6.157217909511488e-05, "loss": 2.6883, "step": 49007 }, { "epoch": 3.84, "learning_rate": 6.157139496589038e-05, "loss": 2.6079, "step": 49008 }, { "epoch": 3.84, "learning_rate": 6.157061083666588e-05, "loss": 2.6858, "step": 49009 }, { "epoch": 3.84, "learning_rate": 6.156982670744139e-05, "loss": 3.1619, "step": 49010 }, { "epoch": 3.84, "learning_rate": 6.156904257821689e-05, "loss": 2.5617, "step": 49011 }, { "epoch": 3.84, "learning_rate": 6.15682584489924e-05, "loss": 2.7578, "step": 49012 }, { "epoch": 3.84, "learning_rate": 6.15674743197679e-05, "loss": 2.958, "step": 49013 }, { "epoch": 3.84, "learning_rate": 6.15666901905434e-05, "loss": 2.5814, "step": 49014 }, { "epoch": 3.84, "learning_rate": 6.15659060613189e-05, "loss": 2.5933, "step": 49015 }, { "epoch": 3.84, "learning_rate": 6.156512193209441e-05, "loss": 2.6673, "step": 49016 }, { "epoch": 3.84, "learning_rate": 6.156433780286991e-05, "loss": 2.6333, "step": 49017 }, { "epoch": 3.84, "learning_rate": 6.156355367364543e-05, "loss": 2.7133, "step": 49018 }, { "epoch": 3.84, "learning_rate": 6.156276954442093e-05, "loss": 2.855, "step": 49019 }, { "epoch": 3.84, "learning_rate": 6.156198541519644e-05, "loss": 2.9771, "step": 49020 }, { "epoch": 3.84, "learning_rate": 6.156120128597193e-05, "loss": 2.6032, "step": 49021 }, { "epoch": 3.84, "learning_rate": 6.156041715674743e-05, "loss": 2.7997, "step": 49022 }, { "epoch": 3.84, "learning_rate": 6.155963302752294e-05, "loss": 2.7413, "step": 49023 }, { "epoch": 3.84, "learning_rate": 6.155884889829844e-05, "loss": 2.6389, "step": 49024 }, { "epoch": 3.84, "learning_rate": 6.155806476907394e-05, "loss": 2.5872, "step": 49025 }, { "epoch": 3.84, "learning_rate": 6.155728063984945e-05, "loss": 2.7115, "step": 49026 }, { "epoch": 3.84, "learning_rate": 6.155649651062495e-05, "loss": 2.8143, "step": 49027 }, { "epoch": 3.84, "learning_rate": 6.155571238140045e-05, "loss": 2.7326, "step": 49028 }, { "epoch": 3.84, "learning_rate": 6.155492825217596e-05, "loss": 2.7747, "step": 49029 }, { "epoch": 3.84, "learning_rate": 6.155414412295146e-05, "loss": 2.7443, "step": 49030 }, { "epoch": 3.84, "learning_rate": 6.155335999372697e-05, "loss": 2.6771, "step": 49031 }, { "epoch": 3.84, "learning_rate": 6.155257586450247e-05, "loss": 2.6409, "step": 49032 }, { "epoch": 3.84, "learning_rate": 6.155179173527797e-05, "loss": 2.7327, "step": 49033 }, { "epoch": 3.84, "learning_rate": 6.155100760605348e-05, "loss": 2.9056, "step": 49034 }, { "epoch": 3.84, "learning_rate": 6.155022347682898e-05, "loss": 2.641, "step": 49035 }, { "epoch": 3.85, "learning_rate": 6.154943934760448e-05, "loss": 2.6612, "step": 49036 }, { "epoch": 3.85, "learning_rate": 6.154865521837999e-05, "loss": 2.7613, "step": 49037 }, { "epoch": 3.85, "learning_rate": 6.154787108915549e-05, "loss": 2.8252, "step": 49038 }, { "epoch": 3.85, "learning_rate": 6.1547086959931e-05, "loss": 2.6881, "step": 49039 }, { "epoch": 3.85, "learning_rate": 6.15463028307065e-05, "loss": 3.0097, "step": 49040 }, { "epoch": 3.85, "learning_rate": 6.1545518701482e-05, "loss": 2.8498, "step": 49041 }, { "epoch": 3.85, "learning_rate": 6.15447345722575e-05, "loss": 2.8277, "step": 49042 }, { "epoch": 3.85, "learning_rate": 6.154395044303301e-05, "loss": 2.6662, "step": 49043 }, { "epoch": 3.85, "learning_rate": 6.154316631380853e-05, "loss": 2.6607, "step": 49044 }, { "epoch": 3.85, "learning_rate": 6.154238218458403e-05, "loss": 2.5627, "step": 49045 }, { "epoch": 3.85, "learning_rate": 6.154159805535953e-05, "loss": 2.5769, "step": 49046 }, { "epoch": 3.85, "learning_rate": 6.154081392613504e-05, "loss": 2.8762, "step": 49047 }, { "epoch": 3.85, "learning_rate": 6.154002979691054e-05, "loss": 2.6604, "step": 49048 }, { "epoch": 3.85, "learning_rate": 6.153924566768605e-05, "loss": 2.7185, "step": 49049 }, { "epoch": 3.85, "learning_rate": 6.153846153846155e-05, "loss": 2.6402, "step": 49050 }, { "epoch": 3.85, "learning_rate": 6.153767740923705e-05, "loss": 2.8081, "step": 49051 }, { "epoch": 3.85, "learning_rate": 6.153689328001256e-05, "loss": 2.8026, "step": 49052 }, { "epoch": 3.85, "learning_rate": 6.153610915078806e-05, "loss": 2.8082, "step": 49053 }, { "epoch": 3.85, "learning_rate": 6.153532502156356e-05, "loss": 2.6884, "step": 49054 }, { "epoch": 3.85, "learning_rate": 6.153454089233907e-05, "loss": 2.5862, "step": 49055 }, { "epoch": 3.85, "learning_rate": 6.153375676311457e-05, "loss": 2.78, "step": 49056 }, { "epoch": 3.85, "learning_rate": 6.153297263389006e-05, "loss": 2.7295, "step": 49057 }, { "epoch": 3.85, "learning_rate": 6.153218850466556e-05, "loss": 2.7058, "step": 49058 }, { "epoch": 3.85, "learning_rate": 6.153140437544107e-05, "loss": 2.6197, "step": 49059 }, { "epoch": 3.85, "learning_rate": 6.153062024621657e-05, "loss": 2.8198, "step": 49060 }, { "epoch": 3.85, "learning_rate": 6.152983611699208e-05, "loss": 2.7902, "step": 49061 }, { "epoch": 3.85, "learning_rate": 6.152905198776758e-05, "loss": 2.6878, "step": 49062 }, { "epoch": 3.85, "learning_rate": 6.152826785854308e-05, "loss": 2.7072, "step": 49063 }, { "epoch": 3.85, "learning_rate": 6.152748372931859e-05, "loss": 2.6507, "step": 49064 }, { "epoch": 3.85, "learning_rate": 6.152669960009409e-05, "loss": 2.5621, "step": 49065 }, { "epoch": 3.85, "learning_rate": 6.15259154708696e-05, "loss": 2.5974, "step": 49066 }, { "epoch": 3.85, "learning_rate": 6.15251313416451e-05, "loss": 2.8474, "step": 49067 }, { "epoch": 3.85, "learning_rate": 6.15243472124206e-05, "loss": 2.6968, "step": 49068 }, { "epoch": 3.85, "learning_rate": 6.152356308319612e-05, "loss": 2.6096, "step": 49069 }, { "epoch": 3.85, "learning_rate": 6.152277895397162e-05, "loss": 2.9137, "step": 49070 }, { "epoch": 3.85, "learning_rate": 6.152199482474713e-05, "loss": 2.7149, "step": 49071 }, { "epoch": 3.85, "learning_rate": 6.152121069552263e-05, "loss": 2.9604, "step": 49072 }, { "epoch": 3.85, "learning_rate": 6.152042656629813e-05, "loss": 2.8708, "step": 49073 }, { "epoch": 3.85, "learning_rate": 6.151964243707364e-05, "loss": 2.6723, "step": 49074 }, { "epoch": 3.85, "learning_rate": 6.151885830784914e-05, "loss": 2.7534, "step": 49075 }, { "epoch": 3.85, "learning_rate": 6.151807417862464e-05, "loss": 2.7778, "step": 49076 }, { "epoch": 3.85, "learning_rate": 6.151729004940015e-05, "loss": 2.6748, "step": 49077 }, { "epoch": 3.85, "learning_rate": 6.151650592017565e-05, "loss": 2.7747, "step": 49078 }, { "epoch": 3.85, "learning_rate": 6.151572179095116e-05, "loss": 2.6316, "step": 49079 }, { "epoch": 3.85, "learning_rate": 6.151493766172666e-05, "loss": 2.6155, "step": 49080 }, { "epoch": 3.85, "learning_rate": 6.151415353250216e-05, "loss": 2.706, "step": 49081 }, { "epoch": 3.85, "learning_rate": 6.151336940327767e-05, "loss": 2.747, "step": 49082 }, { "epoch": 3.85, "learning_rate": 6.151258527405317e-05, "loss": 2.9011, "step": 49083 }, { "epoch": 3.85, "learning_rate": 6.151180114482867e-05, "loss": 3.0717, "step": 49084 }, { "epoch": 3.85, "learning_rate": 6.151101701560418e-05, "loss": 2.5928, "step": 49085 }, { "epoch": 3.85, "learning_rate": 6.151023288637968e-05, "loss": 3.1736, "step": 49086 }, { "epoch": 3.85, "learning_rate": 6.150944875715518e-05, "loss": 2.6914, "step": 49087 }, { "epoch": 3.85, "learning_rate": 6.150866462793069e-05, "loss": 2.75, "step": 49088 }, { "epoch": 3.85, "learning_rate": 6.150788049870619e-05, "loss": 2.7293, "step": 49089 }, { "epoch": 3.85, "learning_rate": 6.15070963694817e-05, "loss": 2.656, "step": 49090 }, { "epoch": 3.85, "learning_rate": 6.15063122402572e-05, "loss": 2.642, "step": 49091 }, { "epoch": 3.85, "learning_rate": 6.15055281110327e-05, "loss": 2.7541, "step": 49092 }, { "epoch": 3.85, "learning_rate": 6.15047439818082e-05, "loss": 2.9094, "step": 49093 }, { "epoch": 3.85, "learning_rate": 6.15039598525837e-05, "loss": 3.1874, "step": 49094 }, { "epoch": 3.85, "learning_rate": 6.150317572335921e-05, "loss": 2.5485, "step": 49095 }, { "epoch": 3.85, "learning_rate": 6.150239159413472e-05, "loss": 2.8725, "step": 49096 }, { "epoch": 3.85, "learning_rate": 6.150160746491022e-05, "loss": 2.7, "step": 49097 }, { "epoch": 3.85, "learning_rate": 6.150082333568573e-05, "loss": 2.6903, "step": 49098 }, { "epoch": 3.85, "learning_rate": 6.150003920646123e-05, "loss": 2.7514, "step": 49099 }, { "epoch": 3.85, "learning_rate": 6.149925507723673e-05, "loss": 2.7388, "step": 49100 }, { "epoch": 3.85, "learning_rate": 6.149847094801224e-05, "loss": 2.6645, "step": 49101 }, { "epoch": 3.85, "learning_rate": 6.149768681878774e-05, "loss": 2.7304, "step": 49102 }, { "epoch": 3.85, "learning_rate": 6.149690268956324e-05, "loss": 2.7851, "step": 49103 }, { "epoch": 3.85, "learning_rate": 6.149611856033875e-05, "loss": 2.6043, "step": 49104 }, { "epoch": 3.85, "learning_rate": 6.149533443111425e-05, "loss": 2.7192, "step": 49105 }, { "epoch": 3.85, "learning_rate": 6.149455030188975e-05, "loss": 2.7214, "step": 49106 }, { "epoch": 3.85, "learning_rate": 6.149376617266526e-05, "loss": 2.7259, "step": 49107 }, { "epoch": 3.85, "learning_rate": 6.149298204344076e-05, "loss": 2.6337, "step": 49108 }, { "epoch": 3.85, "learning_rate": 6.149219791421627e-05, "loss": 2.7747, "step": 49109 }, { "epoch": 3.85, "learning_rate": 6.149141378499177e-05, "loss": 2.7516, "step": 49110 }, { "epoch": 3.85, "learning_rate": 6.149062965576727e-05, "loss": 2.7748, "step": 49111 }, { "epoch": 3.85, "learning_rate": 6.148984552654278e-05, "loss": 2.6729, "step": 49112 }, { "epoch": 3.85, "learning_rate": 6.148906139731828e-05, "loss": 3.033, "step": 49113 }, { "epoch": 3.85, "learning_rate": 6.148827726809378e-05, "loss": 2.743, "step": 49114 }, { "epoch": 3.85, "learning_rate": 6.148749313886929e-05, "loss": 2.5732, "step": 49115 }, { "epoch": 3.85, "learning_rate": 6.148670900964479e-05, "loss": 2.8042, "step": 49116 }, { "epoch": 3.85, "learning_rate": 6.14859248804203e-05, "loss": 2.7909, "step": 49117 }, { "epoch": 3.85, "learning_rate": 6.14851407511958e-05, "loss": 2.9482, "step": 49118 }, { "epoch": 3.85, "learning_rate": 6.14843566219713e-05, "loss": 2.7323, "step": 49119 }, { "epoch": 3.85, "learning_rate": 6.14835724927468e-05, "loss": 2.6813, "step": 49120 }, { "epoch": 3.85, "learning_rate": 6.148278836352231e-05, "loss": 2.9227, "step": 49121 }, { "epoch": 3.85, "learning_rate": 6.148200423429781e-05, "loss": 2.6996, "step": 49122 }, { "epoch": 3.85, "learning_rate": 6.148122010507332e-05, "loss": 2.8444, "step": 49123 }, { "epoch": 3.85, "learning_rate": 6.148043597584882e-05, "loss": 2.9468, "step": 49124 }, { "epoch": 3.85, "learning_rate": 6.147965184662432e-05, "loss": 2.824, "step": 49125 }, { "epoch": 3.85, "learning_rate": 6.147886771739984e-05, "loss": 2.6608, "step": 49126 }, { "epoch": 3.85, "learning_rate": 6.147808358817535e-05, "loss": 2.7728, "step": 49127 }, { "epoch": 3.85, "learning_rate": 6.147729945895085e-05, "loss": 2.634, "step": 49128 }, { "epoch": 3.85, "learning_rate": 6.147651532972634e-05, "loss": 2.7364, "step": 49129 }, { "epoch": 3.85, "learning_rate": 6.147573120050184e-05, "loss": 2.8821, "step": 49130 }, { "epoch": 3.85, "learning_rate": 6.147494707127735e-05, "loss": 2.714, "step": 49131 }, { "epoch": 3.85, "learning_rate": 6.147416294205285e-05, "loss": 2.6454, "step": 49132 }, { "epoch": 3.85, "learning_rate": 6.147337881282835e-05, "loss": 2.8716, "step": 49133 }, { "epoch": 3.85, "learning_rate": 6.147259468360386e-05, "loss": 2.8323, "step": 49134 }, { "epoch": 3.85, "learning_rate": 6.147181055437936e-05, "loss": 2.4959, "step": 49135 }, { "epoch": 3.85, "learning_rate": 6.147102642515486e-05, "loss": 2.7793, "step": 49136 }, { "epoch": 3.85, "learning_rate": 6.147024229593037e-05, "loss": 2.584, "step": 49137 }, { "epoch": 3.85, "learning_rate": 6.146945816670587e-05, "loss": 2.5499, "step": 49138 }, { "epoch": 3.85, "learning_rate": 6.146867403748138e-05, "loss": 2.562, "step": 49139 }, { "epoch": 3.85, "learning_rate": 6.146788990825688e-05, "loss": 2.6617, "step": 49140 }, { "epoch": 3.85, "learning_rate": 6.146710577903238e-05, "loss": 2.5288, "step": 49141 }, { "epoch": 3.85, "learning_rate": 6.146632164980789e-05, "loss": 2.8084, "step": 49142 }, { "epoch": 3.85, "learning_rate": 6.146553752058339e-05, "loss": 2.6704, "step": 49143 }, { "epoch": 3.85, "learning_rate": 6.14647533913589e-05, "loss": 2.9563, "step": 49144 }, { "epoch": 3.85, "learning_rate": 6.14639692621344e-05, "loss": 2.6178, "step": 49145 }, { "epoch": 3.85, "learning_rate": 6.14631851329099e-05, "loss": 2.4357, "step": 49146 }, { "epoch": 3.85, "learning_rate": 6.14624010036854e-05, "loss": 2.9118, "step": 49147 }, { "epoch": 3.85, "learning_rate": 6.146161687446091e-05, "loss": 2.7536, "step": 49148 }, { "epoch": 3.85, "learning_rate": 6.146083274523641e-05, "loss": 2.6916, "step": 49149 }, { "epoch": 3.85, "learning_rate": 6.146004861601192e-05, "loss": 2.5981, "step": 49150 }, { "epoch": 3.85, "learning_rate": 6.145926448678742e-05, "loss": 2.8686, "step": 49151 }, { "epoch": 3.85, "learning_rate": 6.145848035756294e-05, "loss": 2.7864, "step": 49152 }, { "epoch": 3.85, "learning_rate": 6.145769622833844e-05, "loss": 2.7025, "step": 49153 }, { "epoch": 3.85, "learning_rate": 6.145691209911394e-05, "loss": 2.8737, "step": 49154 }, { "epoch": 3.85, "learning_rate": 6.145612796988945e-05, "loss": 2.792, "step": 49155 }, { "epoch": 3.85, "learning_rate": 6.145534384066495e-05, "loss": 2.6212, "step": 49156 }, { "epoch": 3.85, "learning_rate": 6.145455971144046e-05, "loss": 2.7546, "step": 49157 }, { "epoch": 3.85, "learning_rate": 6.145377558221596e-05, "loss": 2.7294, "step": 49158 }, { "epoch": 3.85, "learning_rate": 6.145299145299146e-05, "loss": 2.8197, "step": 49159 }, { "epoch": 3.85, "learning_rate": 6.145220732376697e-05, "loss": 2.9325, "step": 49160 }, { "epoch": 3.85, "learning_rate": 6.145142319454247e-05, "loss": 2.6364, "step": 49161 }, { "epoch": 3.85, "learning_rate": 6.145063906531797e-05, "loss": 2.6916, "step": 49162 }, { "epoch": 3.86, "learning_rate": 6.144985493609348e-05, "loss": 2.7269, "step": 49163 }, { "epoch": 3.86, "learning_rate": 6.144907080686898e-05, "loss": 2.8439, "step": 49164 }, { "epoch": 3.86, "learning_rate": 6.144828667764447e-05, "loss": 2.9208, "step": 49165 }, { "epoch": 3.86, "learning_rate": 6.144750254841998e-05, "loss": 2.8677, "step": 49166 }, { "epoch": 3.86, "learning_rate": 6.144671841919548e-05, "loss": 2.6288, "step": 49167 }, { "epoch": 3.86, "learning_rate": 6.144593428997098e-05, "loss": 2.9027, "step": 49168 }, { "epoch": 3.86, "learning_rate": 6.144515016074649e-05, "loss": 2.7825, "step": 49169 }, { "epoch": 3.86, "learning_rate": 6.144436603152199e-05, "loss": 2.6391, "step": 49170 }, { "epoch": 3.86, "learning_rate": 6.14435819022975e-05, "loss": 2.5799, "step": 49171 }, { "epoch": 3.86, "learning_rate": 6.1442797773073e-05, "loss": 2.6205, "step": 49172 }, { "epoch": 3.86, "learning_rate": 6.14420136438485e-05, "loss": 2.6464, "step": 49173 }, { "epoch": 3.86, "learning_rate": 6.1441229514624e-05, "loss": 2.7328, "step": 49174 }, { "epoch": 3.86, "learning_rate": 6.144044538539951e-05, "loss": 3.0565, "step": 49175 }, { "epoch": 3.86, "learning_rate": 6.143966125617501e-05, "loss": 2.621, "step": 49176 }, { "epoch": 3.86, "learning_rate": 6.143887712695053e-05, "loss": 2.6171, "step": 49177 }, { "epoch": 3.86, "learning_rate": 6.143809299772603e-05, "loss": 2.6718, "step": 49178 }, { "epoch": 3.86, "learning_rate": 6.143730886850154e-05, "loss": 2.7725, "step": 49179 }, { "epoch": 3.86, "learning_rate": 6.143652473927704e-05, "loss": 3.085, "step": 49180 }, { "epoch": 3.86, "learning_rate": 6.143574061005254e-05, "loss": 2.523, "step": 49181 }, { "epoch": 3.86, "learning_rate": 6.143495648082805e-05, "loss": 2.638, "step": 49182 }, { "epoch": 3.86, "learning_rate": 6.143417235160355e-05, "loss": 2.4689, "step": 49183 }, { "epoch": 3.86, "learning_rate": 6.143338822237906e-05, "loss": 2.7459, "step": 49184 }, { "epoch": 3.86, "learning_rate": 6.143260409315456e-05, "loss": 2.731, "step": 49185 }, { "epoch": 3.86, "learning_rate": 6.143181996393006e-05, "loss": 2.6188, "step": 49186 }, { "epoch": 3.86, "learning_rate": 6.143103583470557e-05, "loss": 2.677, "step": 49187 }, { "epoch": 3.86, "learning_rate": 6.143025170548107e-05, "loss": 2.413, "step": 49188 }, { "epoch": 3.86, "learning_rate": 6.142946757625657e-05, "loss": 2.6688, "step": 49189 }, { "epoch": 3.86, "learning_rate": 6.142868344703208e-05, "loss": 2.9927, "step": 49190 }, { "epoch": 3.86, "learning_rate": 6.142789931780758e-05, "loss": 2.7375, "step": 49191 }, { "epoch": 3.86, "learning_rate": 6.142711518858308e-05, "loss": 2.7782, "step": 49192 }, { "epoch": 3.86, "learning_rate": 6.142633105935859e-05, "loss": 2.655, "step": 49193 }, { "epoch": 3.86, "learning_rate": 6.142554693013409e-05, "loss": 2.7471, "step": 49194 }, { "epoch": 3.86, "learning_rate": 6.14247628009096e-05, "loss": 2.7733, "step": 49195 }, { "epoch": 3.86, "learning_rate": 6.14239786716851e-05, "loss": 2.6422, "step": 49196 }, { "epoch": 3.86, "learning_rate": 6.14231945424606e-05, "loss": 2.6429, "step": 49197 }, { "epoch": 3.86, "learning_rate": 6.14224104132361e-05, "loss": 2.723, "step": 49198 }, { "epoch": 3.86, "learning_rate": 6.142162628401161e-05, "loss": 2.4803, "step": 49199 }, { "epoch": 3.86, "learning_rate": 6.142084215478711e-05, "loss": 2.7137, "step": 49200 }, { "epoch": 3.86, "learning_rate": 6.14200580255626e-05, "loss": 2.8343, "step": 49201 }, { "epoch": 3.86, "learning_rate": 6.141927389633811e-05, "loss": 2.9311, "step": 49202 }, { "epoch": 3.86, "learning_rate": 6.141848976711362e-05, "loss": 2.7496, "step": 49203 }, { "epoch": 3.86, "learning_rate": 6.141770563788913e-05, "loss": 2.7371, "step": 49204 }, { "epoch": 3.86, "learning_rate": 6.141692150866463e-05, "loss": 2.7027, "step": 49205 }, { "epoch": 3.86, "learning_rate": 6.141613737944014e-05, "loss": 2.8116, "step": 49206 }, { "epoch": 3.86, "learning_rate": 6.141535325021564e-05, "loss": 2.7187, "step": 49207 }, { "epoch": 3.86, "learning_rate": 6.141456912099114e-05, "loss": 2.8101, "step": 49208 }, { "epoch": 3.86, "learning_rate": 6.141378499176665e-05, "loss": 2.8317, "step": 49209 }, { "epoch": 3.86, "learning_rate": 6.141300086254215e-05, "loss": 2.7992, "step": 49210 }, { "epoch": 3.86, "learning_rate": 6.141221673331765e-05, "loss": 2.7631, "step": 49211 }, { "epoch": 3.86, "learning_rate": 6.141143260409316e-05, "loss": 2.853, "step": 49212 }, { "epoch": 3.86, "learning_rate": 6.141064847486866e-05, "loss": 2.8427, "step": 49213 }, { "epoch": 3.86, "learning_rate": 6.140986434564417e-05, "loss": 2.8021, "step": 49214 }, { "epoch": 3.86, "learning_rate": 6.140908021641967e-05, "loss": 2.745, "step": 49215 }, { "epoch": 3.86, "learning_rate": 6.140829608719517e-05, "loss": 3.1141, "step": 49216 }, { "epoch": 3.86, "learning_rate": 6.140751195797068e-05, "loss": 2.9298, "step": 49217 }, { "epoch": 3.86, "learning_rate": 6.140672782874618e-05, "loss": 2.6772, "step": 49218 }, { "epoch": 3.86, "learning_rate": 6.140594369952168e-05, "loss": 2.8636, "step": 49219 }, { "epoch": 3.86, "learning_rate": 6.140515957029719e-05, "loss": 2.5727, "step": 49220 }, { "epoch": 3.86, "learning_rate": 6.140437544107269e-05, "loss": 2.7505, "step": 49221 }, { "epoch": 3.86, "learning_rate": 6.14035913118482e-05, "loss": 2.8153, "step": 49222 }, { "epoch": 3.86, "learning_rate": 6.14028071826237e-05, "loss": 2.6322, "step": 49223 }, { "epoch": 3.86, "learning_rate": 6.14020230533992e-05, "loss": 2.5908, "step": 49224 }, { "epoch": 3.86, "learning_rate": 6.14012389241747e-05, "loss": 2.6457, "step": 49225 }, { "epoch": 3.86, "learning_rate": 6.140045479495021e-05, "loss": 2.7042, "step": 49226 }, { "epoch": 3.86, "learning_rate": 6.139967066572571e-05, "loss": 2.7592, "step": 49227 }, { "epoch": 3.86, "learning_rate": 6.139888653650122e-05, "loss": 2.6345, "step": 49228 }, { "epoch": 3.86, "learning_rate": 6.139810240727672e-05, "loss": 2.8535, "step": 49229 }, { "epoch": 3.86, "learning_rate": 6.139731827805222e-05, "loss": 2.6307, "step": 49230 }, { "epoch": 3.86, "learning_rate": 6.139653414882773e-05, "loss": 2.8074, "step": 49231 }, { "epoch": 3.86, "learning_rate": 6.139575001960323e-05, "loss": 2.7535, "step": 49232 }, { "epoch": 3.86, "learning_rate": 6.139496589037874e-05, "loss": 3.0797, "step": 49233 }, { "epoch": 3.86, "learning_rate": 6.139418176115425e-05, "loss": 2.7732, "step": 49234 }, { "epoch": 3.86, "learning_rate": 6.139339763192976e-05, "loss": 2.7476, "step": 49235 }, { "epoch": 3.86, "learning_rate": 6.139261350270526e-05, "loss": 2.7417, "step": 49236 }, { "epoch": 3.86, "learning_rate": 6.139182937348075e-05, "loss": 2.6957, "step": 49237 }, { "epoch": 3.86, "learning_rate": 6.139104524425625e-05, "loss": 2.7223, "step": 49238 }, { "epoch": 3.86, "learning_rate": 6.139026111503176e-05, "loss": 2.8794, "step": 49239 }, { "epoch": 3.86, "learning_rate": 6.138947698580726e-05, "loss": 2.7849, "step": 49240 }, { "epoch": 3.86, "learning_rate": 6.138869285658276e-05, "loss": 2.6576, "step": 49241 }, { "epoch": 3.86, "learning_rate": 6.138790872735827e-05, "loss": 2.7569, "step": 49242 }, { "epoch": 3.86, "learning_rate": 6.138712459813377e-05, "loss": 2.672, "step": 49243 }, { "epoch": 3.86, "learning_rate": 6.138634046890928e-05, "loss": 2.608, "step": 49244 }, { "epoch": 3.86, "learning_rate": 6.138555633968478e-05, "loss": 2.6179, "step": 49245 }, { "epoch": 3.86, "learning_rate": 6.138477221046028e-05, "loss": 2.6909, "step": 49246 }, { "epoch": 3.86, "learning_rate": 6.138398808123579e-05, "loss": 2.8955, "step": 49247 }, { "epoch": 3.86, "learning_rate": 6.138320395201129e-05, "loss": 2.9159, "step": 49248 }, { "epoch": 3.86, "learning_rate": 6.13824198227868e-05, "loss": 2.7364, "step": 49249 }, { "epoch": 3.86, "learning_rate": 6.13816356935623e-05, "loss": 2.7335, "step": 49250 }, { "epoch": 3.86, "learning_rate": 6.13808515643378e-05, "loss": 2.6658, "step": 49251 }, { "epoch": 3.86, "learning_rate": 6.13800674351133e-05, "loss": 2.6359, "step": 49252 }, { "epoch": 3.86, "learning_rate": 6.137928330588881e-05, "loss": 2.864, "step": 49253 }, { "epoch": 3.86, "learning_rate": 6.137849917666431e-05, "loss": 2.6445, "step": 49254 }, { "epoch": 3.86, "learning_rate": 6.137771504743982e-05, "loss": 2.7838, "step": 49255 }, { "epoch": 3.86, "learning_rate": 6.137693091821532e-05, "loss": 2.7535, "step": 49256 }, { "epoch": 3.86, "learning_rate": 6.137614678899082e-05, "loss": 2.6627, "step": 49257 }, { "epoch": 3.86, "learning_rate": 6.137536265976633e-05, "loss": 2.8038, "step": 49258 }, { "epoch": 3.86, "learning_rate": 6.137457853054183e-05, "loss": 2.8598, "step": 49259 }, { "epoch": 3.86, "learning_rate": 6.137379440131735e-05, "loss": 2.8106, "step": 49260 }, { "epoch": 3.86, "learning_rate": 6.137301027209285e-05, "loss": 2.7148, "step": 49261 }, { "epoch": 3.86, "learning_rate": 6.137222614286836e-05, "loss": 2.7051, "step": 49262 }, { "epoch": 3.86, "learning_rate": 6.137144201364386e-05, "loss": 2.6158, "step": 49263 }, { "epoch": 3.86, "learning_rate": 6.137065788441936e-05, "loss": 2.5284, "step": 49264 }, { "epoch": 3.86, "learning_rate": 6.136987375519487e-05, "loss": 2.8141, "step": 49265 }, { "epoch": 3.86, "learning_rate": 6.136908962597037e-05, "loss": 2.8956, "step": 49266 }, { "epoch": 3.86, "learning_rate": 6.136830549674587e-05, "loss": 2.7324, "step": 49267 }, { "epoch": 3.86, "learning_rate": 6.136752136752138e-05, "loss": 2.6303, "step": 49268 }, { "epoch": 3.86, "learning_rate": 6.136673723829688e-05, "loss": 2.8101, "step": 49269 }, { "epoch": 3.86, "learning_rate": 6.136595310907238e-05, "loss": 2.6123, "step": 49270 }, { "epoch": 3.86, "learning_rate": 6.136516897984789e-05, "loss": 2.8337, "step": 49271 }, { "epoch": 3.86, "learning_rate": 6.136438485062339e-05, "loss": 2.5604, "step": 49272 }, { "epoch": 3.86, "learning_rate": 6.136360072139888e-05, "loss": 2.6143, "step": 49273 }, { "epoch": 3.86, "learning_rate": 6.136281659217439e-05, "loss": 2.7355, "step": 49274 }, { "epoch": 3.86, "learning_rate": 6.136203246294989e-05, "loss": 2.7725, "step": 49275 }, { "epoch": 3.86, "learning_rate": 6.13612483337254e-05, "loss": 2.7628, "step": 49276 }, { "epoch": 3.86, "learning_rate": 6.13604642045009e-05, "loss": 2.6518, "step": 49277 }, { "epoch": 3.86, "learning_rate": 6.13596800752764e-05, "loss": 2.5914, "step": 49278 }, { "epoch": 3.86, "learning_rate": 6.13588959460519e-05, "loss": 2.8777, "step": 49279 }, { "epoch": 3.86, "learning_rate": 6.135811181682741e-05, "loss": 2.6022, "step": 49280 }, { "epoch": 3.86, "learning_rate": 6.135732768760291e-05, "loss": 2.5634, "step": 49281 }, { "epoch": 3.86, "learning_rate": 6.135654355837842e-05, "loss": 2.6176, "step": 49282 }, { "epoch": 3.86, "learning_rate": 6.135575942915392e-05, "loss": 2.5857, "step": 49283 }, { "epoch": 3.86, "learning_rate": 6.135497529992942e-05, "loss": 2.869, "step": 49284 }, { "epoch": 3.86, "learning_rate": 6.135419117070493e-05, "loss": 2.5064, "step": 49285 }, { "epoch": 3.86, "learning_rate": 6.135340704148044e-05, "loss": 2.6554, "step": 49286 }, { "epoch": 3.86, "learning_rate": 6.135262291225595e-05, "loss": 2.9734, "step": 49287 }, { "epoch": 3.86, "learning_rate": 6.135183878303145e-05, "loss": 2.6365, "step": 49288 }, { "epoch": 3.86, "learning_rate": 6.135105465380695e-05, "loss": 2.6571, "step": 49289 }, { "epoch": 3.86, "learning_rate": 6.135027052458246e-05, "loss": 2.6909, "step": 49290 }, { "epoch": 3.87, "learning_rate": 6.134948639535796e-05, "loss": 2.7448, "step": 49291 }, { "epoch": 3.87, "learning_rate": 6.134870226613347e-05, "loss": 2.6832, "step": 49292 }, { "epoch": 3.87, "learning_rate": 6.134791813690897e-05, "loss": 2.7093, "step": 49293 }, { "epoch": 3.87, "learning_rate": 6.134713400768447e-05, "loss": 2.9444, "step": 49294 }, { "epoch": 3.87, "learning_rate": 6.134634987845998e-05, "loss": 2.5983, "step": 49295 }, { "epoch": 3.87, "learning_rate": 6.134556574923548e-05, "loss": 2.9586, "step": 49296 }, { "epoch": 3.87, "learning_rate": 6.134478162001098e-05, "loss": 3.0644, "step": 49297 }, { "epoch": 3.87, "learning_rate": 6.134399749078649e-05, "loss": 2.7625, "step": 49298 }, { "epoch": 3.87, "learning_rate": 6.134321336156199e-05, "loss": 2.8058, "step": 49299 }, { "epoch": 3.87, "learning_rate": 6.13424292323375e-05, "loss": 2.9646, "step": 49300 }, { "epoch": 3.87, "learning_rate": 6.1341645103113e-05, "loss": 3.0331, "step": 49301 }, { "epoch": 3.87, "learning_rate": 6.13408609738885e-05, "loss": 2.815, "step": 49302 }, { "epoch": 3.87, "learning_rate": 6.1340076844664e-05, "loss": 2.7024, "step": 49303 }, { "epoch": 3.87, "learning_rate": 6.133929271543951e-05, "loss": 2.6747, "step": 49304 }, { "epoch": 3.87, "learning_rate": 6.133850858621501e-05, "loss": 2.7408, "step": 49305 }, { "epoch": 3.87, "learning_rate": 6.133772445699052e-05, "loss": 2.7673, "step": 49306 }, { "epoch": 3.87, "learning_rate": 6.133694032776602e-05, "loss": 2.5231, "step": 49307 }, { "epoch": 3.87, "learning_rate": 6.133615619854152e-05, "loss": 2.5969, "step": 49308 }, { "epoch": 3.87, "learning_rate": 6.133537206931701e-05, "loss": 2.7039, "step": 49309 }, { "epoch": 3.87, "learning_rate": 6.133458794009252e-05, "loss": 2.7463, "step": 49310 }, { "epoch": 3.87, "learning_rate": 6.133380381086804e-05, "loss": 2.6872, "step": 49311 }, { "epoch": 3.87, "learning_rate": 6.133301968164354e-05, "loss": 2.9476, "step": 49312 }, { "epoch": 3.87, "learning_rate": 6.133223555241904e-05, "loss": 3.1061, "step": 49313 }, { "epoch": 3.87, "learning_rate": 6.133145142319455e-05, "loss": 2.8338, "step": 49314 }, { "epoch": 3.87, "learning_rate": 6.133066729397005e-05, "loss": 2.9737, "step": 49315 }, { "epoch": 3.87, "learning_rate": 6.132988316474555e-05, "loss": 2.6866, "step": 49316 }, { "epoch": 3.87, "learning_rate": 6.132909903552106e-05, "loss": 2.736, "step": 49317 }, { "epoch": 3.87, "learning_rate": 6.132831490629656e-05, "loss": 2.839, "step": 49318 }, { "epoch": 3.87, "learning_rate": 6.132753077707207e-05, "loss": 2.7729, "step": 49319 }, { "epoch": 3.87, "learning_rate": 6.132674664784757e-05, "loss": 2.6705, "step": 49320 }, { "epoch": 3.87, "learning_rate": 6.132596251862307e-05, "loss": 2.5065, "step": 49321 }, { "epoch": 3.87, "learning_rate": 6.132517838939858e-05, "loss": 2.5826, "step": 49322 }, { "epoch": 3.87, "learning_rate": 6.132439426017408e-05, "loss": 2.9107, "step": 49323 }, { "epoch": 3.87, "learning_rate": 6.132361013094958e-05, "loss": 2.7704, "step": 49324 }, { "epoch": 3.87, "learning_rate": 6.132282600172509e-05, "loss": 2.7161, "step": 49325 }, { "epoch": 3.87, "learning_rate": 6.132204187250059e-05, "loss": 2.8363, "step": 49326 }, { "epoch": 3.87, "learning_rate": 6.13212577432761e-05, "loss": 2.6466, "step": 49327 }, { "epoch": 3.87, "learning_rate": 6.13204736140516e-05, "loss": 2.5948, "step": 49328 }, { "epoch": 3.87, "learning_rate": 6.13196894848271e-05, "loss": 2.7637, "step": 49329 }, { "epoch": 3.87, "learning_rate": 6.13189053556026e-05, "loss": 2.6254, "step": 49330 }, { "epoch": 3.87, "learning_rate": 6.131812122637811e-05, "loss": 2.6657, "step": 49331 }, { "epoch": 3.87, "learning_rate": 6.131733709715361e-05, "loss": 2.8189, "step": 49332 }, { "epoch": 3.87, "learning_rate": 6.131655296792912e-05, "loss": 2.9518, "step": 49333 }, { "epoch": 3.87, "learning_rate": 6.131576883870462e-05, "loss": 2.7301, "step": 49334 }, { "epoch": 3.87, "learning_rate": 6.131498470948012e-05, "loss": 2.5672, "step": 49335 }, { "epoch": 3.87, "learning_rate": 6.131420058025563e-05, "loss": 2.6751, "step": 49336 }, { "epoch": 3.87, "learning_rate": 6.131341645103113e-05, "loss": 2.5077, "step": 49337 }, { "epoch": 3.87, "learning_rate": 6.131263232180663e-05, "loss": 2.8174, "step": 49338 }, { "epoch": 3.87, "learning_rate": 6.131184819258214e-05, "loss": 2.7465, "step": 49339 }, { "epoch": 3.87, "learning_rate": 6.131106406335764e-05, "loss": 3.0065, "step": 49340 }, { "epoch": 3.87, "learning_rate": 6.131027993413315e-05, "loss": 2.7626, "step": 49341 }, { "epoch": 3.87, "learning_rate": 6.130949580490866e-05, "loss": 2.675, "step": 49342 }, { "epoch": 3.87, "learning_rate": 6.130871167568417e-05, "loss": 2.9091, "step": 49343 }, { "epoch": 3.87, "learning_rate": 6.130792754645967e-05, "loss": 2.5623, "step": 49344 }, { "epoch": 3.87, "learning_rate": 6.130714341723516e-05, "loss": 2.6913, "step": 49345 }, { "epoch": 3.87, "learning_rate": 6.130635928801066e-05, "loss": 2.7894, "step": 49346 }, { "epoch": 3.87, "learning_rate": 6.130557515878617e-05, "loss": 2.9757, "step": 49347 }, { "epoch": 3.87, "learning_rate": 6.130479102956167e-05, "loss": 2.7235, "step": 49348 }, { "epoch": 3.87, "learning_rate": 6.130400690033718e-05, "loss": 2.6779, "step": 49349 }, { "epoch": 3.87, "learning_rate": 6.130322277111268e-05, "loss": 2.7338, "step": 49350 }, { "epoch": 3.87, "learning_rate": 6.130243864188818e-05, "loss": 3.0385, "step": 49351 }, { "epoch": 3.87, "learning_rate": 6.130165451266369e-05, "loss": 2.7942, "step": 49352 }, { "epoch": 3.87, "learning_rate": 6.130087038343919e-05, "loss": 2.6704, "step": 49353 }, { "epoch": 3.87, "learning_rate": 6.13000862542147e-05, "loss": 2.6857, "step": 49354 }, { "epoch": 3.87, "learning_rate": 6.12993021249902e-05, "loss": 2.8268, "step": 49355 }, { "epoch": 3.87, "learning_rate": 6.12985179957657e-05, "loss": 2.4728, "step": 49356 }, { "epoch": 3.87, "learning_rate": 6.12977338665412e-05, "loss": 2.5946, "step": 49357 }, { "epoch": 3.87, "learning_rate": 6.129694973731671e-05, "loss": 2.899, "step": 49358 }, { "epoch": 3.87, "learning_rate": 6.129616560809221e-05, "loss": 2.7438, "step": 49359 }, { "epoch": 3.87, "learning_rate": 6.129538147886772e-05, "loss": 2.888, "step": 49360 }, { "epoch": 3.87, "learning_rate": 6.129459734964322e-05, "loss": 2.7709, "step": 49361 }, { "epoch": 3.87, "learning_rate": 6.129381322041872e-05, "loss": 2.7366, "step": 49362 }, { "epoch": 3.87, "learning_rate": 6.129302909119423e-05, "loss": 2.6923, "step": 49363 }, { "epoch": 3.87, "learning_rate": 6.129224496196973e-05, "loss": 2.7701, "step": 49364 }, { "epoch": 3.87, "learning_rate": 6.129146083274523e-05, "loss": 2.807, "step": 49365 }, { "epoch": 3.87, "learning_rate": 6.129067670352074e-05, "loss": 2.8005, "step": 49366 }, { "epoch": 3.87, "learning_rate": 6.128989257429624e-05, "loss": 2.5654, "step": 49367 }, { "epoch": 3.87, "learning_rate": 6.128910844507176e-05, "loss": 2.7205, "step": 49368 }, { "epoch": 3.87, "learning_rate": 6.128832431584726e-05, "loss": 2.7277, "step": 49369 }, { "epoch": 3.87, "learning_rate": 6.128754018662277e-05, "loss": 2.6606, "step": 49370 }, { "epoch": 3.87, "learning_rate": 6.128675605739827e-05, "loss": 2.9416, "step": 49371 }, { "epoch": 3.87, "learning_rate": 6.128597192817377e-05, "loss": 2.75, "step": 49372 }, { "epoch": 3.87, "learning_rate": 6.128518779894928e-05, "loss": 2.7203, "step": 49373 }, { "epoch": 3.87, "learning_rate": 6.128440366972478e-05, "loss": 2.9337, "step": 49374 }, { "epoch": 3.87, "learning_rate": 6.128361954050028e-05, "loss": 2.9353, "step": 49375 }, { "epoch": 3.87, "learning_rate": 6.128283541127579e-05, "loss": 2.638, "step": 49376 }, { "epoch": 3.87, "learning_rate": 6.128205128205129e-05, "loss": 2.6431, "step": 49377 }, { "epoch": 3.87, "learning_rate": 6.12812671528268e-05, "loss": 2.6932, "step": 49378 }, { "epoch": 3.87, "learning_rate": 6.12804830236023e-05, "loss": 2.6221, "step": 49379 }, { "epoch": 3.87, "learning_rate": 6.12796988943778e-05, "loss": 2.809, "step": 49380 }, { "epoch": 3.87, "learning_rate": 6.127891476515329e-05, "loss": 2.7967, "step": 49381 }, { "epoch": 3.87, "learning_rate": 6.12781306359288e-05, "loss": 2.6415, "step": 49382 }, { "epoch": 3.87, "learning_rate": 6.12773465067043e-05, "loss": 2.8309, "step": 49383 }, { "epoch": 3.87, "learning_rate": 6.12765623774798e-05, "loss": 2.691, "step": 49384 }, { "epoch": 3.87, "learning_rate": 6.127577824825531e-05, "loss": 2.745, "step": 49385 }, { "epoch": 3.87, "learning_rate": 6.127499411903081e-05, "loss": 2.6684, "step": 49386 }, { "epoch": 3.87, "learning_rate": 6.127420998980632e-05, "loss": 2.6474, "step": 49387 }, { "epoch": 3.87, "learning_rate": 6.127342586058182e-05, "loss": 2.7438, "step": 49388 }, { "epoch": 3.87, "learning_rate": 6.127264173135732e-05, "loss": 2.7144, "step": 49389 }, { "epoch": 3.87, "learning_rate": 6.127185760213283e-05, "loss": 2.4997, "step": 49390 }, { "epoch": 3.87, "learning_rate": 6.127107347290833e-05, "loss": 2.7508, "step": 49391 }, { "epoch": 3.87, "learning_rate": 6.127028934368383e-05, "loss": 2.6432, "step": 49392 }, { "epoch": 3.87, "learning_rate": 6.126950521445934e-05, "loss": 2.6404, "step": 49393 }, { "epoch": 3.87, "learning_rate": 6.126872108523485e-05, "loss": 2.6958, "step": 49394 }, { "epoch": 3.87, "learning_rate": 6.126793695601036e-05, "loss": 2.9363, "step": 49395 }, { "epoch": 3.87, "learning_rate": 6.126715282678586e-05, "loss": 2.7523, "step": 49396 }, { "epoch": 3.87, "learning_rate": 6.126636869756137e-05, "loss": 2.6429, "step": 49397 }, { "epoch": 3.87, "learning_rate": 6.126558456833687e-05, "loss": 2.9401, "step": 49398 }, { "epoch": 3.87, "learning_rate": 6.126480043911237e-05, "loss": 2.6236, "step": 49399 }, { "epoch": 3.87, "learning_rate": 6.126401630988788e-05, "loss": 2.7076, "step": 49400 }, { "epoch": 3.87, "learning_rate": 6.126323218066338e-05, "loss": 2.8643, "step": 49401 }, { "epoch": 3.87, "learning_rate": 6.126244805143888e-05, "loss": 2.7464, "step": 49402 }, { "epoch": 3.87, "learning_rate": 6.126166392221439e-05, "loss": 2.7776, "step": 49403 }, { "epoch": 3.87, "learning_rate": 6.126087979298989e-05, "loss": 2.6642, "step": 49404 }, { "epoch": 3.87, "learning_rate": 6.12600956637654e-05, "loss": 2.7972, "step": 49405 }, { "epoch": 3.87, "learning_rate": 6.12593115345409e-05, "loss": 2.8105, "step": 49406 }, { "epoch": 3.87, "learning_rate": 6.12585274053164e-05, "loss": 2.7991, "step": 49407 }, { "epoch": 3.87, "learning_rate": 6.12577432760919e-05, "loss": 2.6635, "step": 49408 }, { "epoch": 3.87, "learning_rate": 6.125695914686741e-05, "loss": 3.0058, "step": 49409 }, { "epoch": 3.87, "learning_rate": 6.125617501764291e-05, "loss": 2.6493, "step": 49410 }, { "epoch": 3.87, "learning_rate": 6.125539088841842e-05, "loss": 2.808, "step": 49411 }, { "epoch": 3.87, "learning_rate": 6.125460675919392e-05, "loss": 3.0916, "step": 49412 }, { "epoch": 3.87, "learning_rate": 6.125382262996942e-05, "loss": 2.7945, "step": 49413 }, { "epoch": 3.87, "learning_rate": 6.125303850074493e-05, "loss": 2.7327, "step": 49414 }, { "epoch": 3.87, "learning_rate": 6.125225437152043e-05, "loss": 2.8989, "step": 49415 }, { "epoch": 3.87, "learning_rate": 6.125147024229594e-05, "loss": 2.7619, "step": 49416 }, { "epoch": 3.87, "learning_rate": 6.125068611307143e-05, "loss": 2.7606, "step": 49417 }, { "epoch": 3.88, "learning_rate": 6.124990198384693e-05, "loss": 2.8283, "step": 49418 }, { "epoch": 3.88, "learning_rate": 6.124911785462245e-05, "loss": 2.7914, "step": 49419 }, { "epoch": 3.88, "learning_rate": 6.124833372539795e-05, "loss": 2.6334, "step": 49420 }, { "epoch": 3.88, "learning_rate": 6.124754959617345e-05, "loss": 2.546, "step": 49421 }, { "epoch": 3.88, "learning_rate": 6.124676546694896e-05, "loss": 2.8756, "step": 49422 }, { "epoch": 3.88, "learning_rate": 6.124598133772446e-05, "loss": 2.5749, "step": 49423 }, { "epoch": 3.88, "learning_rate": 6.124519720849996e-05, "loss": 2.5549, "step": 49424 }, { "epoch": 3.88, "learning_rate": 6.124441307927547e-05, "loss": 2.802, "step": 49425 }, { "epoch": 3.88, "learning_rate": 6.124362895005097e-05, "loss": 2.763, "step": 49426 }, { "epoch": 3.88, "learning_rate": 6.124284482082648e-05, "loss": 2.6727, "step": 49427 }, { "epoch": 3.88, "learning_rate": 6.124206069160198e-05, "loss": 3.1858, "step": 49428 }, { "epoch": 3.88, "learning_rate": 6.124127656237748e-05, "loss": 2.6463, "step": 49429 }, { "epoch": 3.88, "learning_rate": 6.124049243315299e-05, "loss": 2.6827, "step": 49430 }, { "epoch": 3.88, "learning_rate": 6.123970830392849e-05, "loss": 2.9159, "step": 49431 }, { "epoch": 3.88, "learning_rate": 6.1238924174704e-05, "loss": 2.5962, "step": 49432 }, { "epoch": 3.88, "learning_rate": 6.12381400454795e-05, "loss": 2.7537, "step": 49433 }, { "epoch": 3.88, "learning_rate": 6.1237355916255e-05, "loss": 2.6862, "step": 49434 }, { "epoch": 3.88, "learning_rate": 6.12365717870305e-05, "loss": 2.5915, "step": 49435 }, { "epoch": 3.88, "learning_rate": 6.123578765780601e-05, "loss": 2.9143, "step": 49436 }, { "epoch": 3.88, "learning_rate": 6.123500352858151e-05, "loss": 2.7118, "step": 49437 }, { "epoch": 3.88, "learning_rate": 6.123421939935702e-05, "loss": 2.8137, "step": 49438 }, { "epoch": 3.88, "learning_rate": 6.123343527013252e-05, "loss": 2.6322, "step": 49439 }, { "epoch": 3.88, "learning_rate": 6.123265114090802e-05, "loss": 2.74, "step": 49440 }, { "epoch": 3.88, "learning_rate": 6.123186701168353e-05, "loss": 2.8561, "step": 49441 }, { "epoch": 3.88, "learning_rate": 6.123108288245903e-05, "loss": 2.8422, "step": 49442 }, { "epoch": 3.88, "learning_rate": 6.123029875323453e-05, "loss": 2.6113, "step": 49443 }, { "epoch": 3.88, "learning_rate": 6.122951462401004e-05, "loss": 2.7445, "step": 49444 }, { "epoch": 3.88, "learning_rate": 6.122873049478554e-05, "loss": 2.8267, "step": 49445 }, { "epoch": 3.88, "learning_rate": 6.122794636556105e-05, "loss": 2.7342, "step": 49446 }, { "epoch": 3.88, "learning_rate": 6.122716223633655e-05, "loss": 2.6704, "step": 49447 }, { "epoch": 3.88, "learning_rate": 6.122637810711205e-05, "loss": 2.8001, "step": 49448 }, { "epoch": 3.88, "learning_rate": 6.122559397788756e-05, "loss": 2.6272, "step": 49449 }, { "epoch": 3.88, "learning_rate": 6.122480984866307e-05, "loss": 2.8251, "step": 49450 }, { "epoch": 3.88, "learning_rate": 6.122402571943858e-05, "loss": 2.5966, "step": 49451 }, { "epoch": 3.88, "learning_rate": 6.122324159021408e-05, "loss": 2.6725, "step": 49452 }, { "epoch": 3.88, "learning_rate": 6.122245746098957e-05, "loss": 2.811, "step": 49453 }, { "epoch": 3.88, "learning_rate": 6.122167333176508e-05, "loss": 2.7662, "step": 49454 }, { "epoch": 3.88, "learning_rate": 6.122088920254058e-05, "loss": 2.7317, "step": 49455 }, { "epoch": 3.88, "learning_rate": 6.122010507331608e-05, "loss": 2.7881, "step": 49456 }, { "epoch": 3.88, "learning_rate": 6.121932094409159e-05, "loss": 2.6482, "step": 49457 }, { "epoch": 3.88, "learning_rate": 6.121853681486709e-05, "loss": 2.9063, "step": 49458 }, { "epoch": 3.88, "learning_rate": 6.12177526856426e-05, "loss": 2.5501, "step": 49459 }, { "epoch": 3.88, "learning_rate": 6.12169685564181e-05, "loss": 2.7353, "step": 49460 }, { "epoch": 3.88, "learning_rate": 6.12161844271936e-05, "loss": 2.9475, "step": 49461 }, { "epoch": 3.88, "learning_rate": 6.12154002979691e-05, "loss": 2.5764, "step": 49462 }, { "epoch": 3.88, "learning_rate": 6.121461616874461e-05, "loss": 2.6959, "step": 49463 }, { "epoch": 3.88, "learning_rate": 6.121383203952011e-05, "loss": 2.929, "step": 49464 }, { "epoch": 3.88, "learning_rate": 6.121304791029562e-05, "loss": 2.7905, "step": 49465 }, { "epoch": 3.88, "learning_rate": 6.121226378107112e-05, "loss": 2.7243, "step": 49466 }, { "epoch": 3.88, "learning_rate": 6.121147965184662e-05, "loss": 2.6352, "step": 49467 }, { "epoch": 3.88, "learning_rate": 6.121069552262213e-05, "loss": 2.9105, "step": 49468 }, { "epoch": 3.88, "learning_rate": 6.120991139339763e-05, "loss": 3.0004, "step": 49469 }, { "epoch": 3.88, "learning_rate": 6.120912726417313e-05, "loss": 2.7647, "step": 49470 }, { "epoch": 3.88, "learning_rate": 6.120834313494864e-05, "loss": 2.9146, "step": 49471 }, { "epoch": 3.88, "learning_rate": 6.120755900572414e-05, "loss": 2.722, "step": 49472 }, { "epoch": 3.88, "learning_rate": 6.120677487649965e-05, "loss": 2.6431, "step": 49473 }, { "epoch": 3.88, "learning_rate": 6.120599074727515e-05, "loss": 2.8423, "step": 49474 }, { "epoch": 3.88, "learning_rate": 6.120520661805065e-05, "loss": 2.698, "step": 49475 }, { "epoch": 3.88, "learning_rate": 6.120442248882617e-05, "loss": 2.63, "step": 49476 }, { "epoch": 3.88, "learning_rate": 6.120363835960167e-05, "loss": 2.914, "step": 49477 }, { "epoch": 3.88, "learning_rate": 6.120285423037718e-05, "loss": 2.8128, "step": 49478 }, { "epoch": 3.88, "learning_rate": 6.120207010115268e-05, "loss": 2.634, "step": 49479 }, { "epoch": 3.88, "learning_rate": 6.120128597192818e-05, "loss": 2.5865, "step": 49480 }, { "epoch": 3.88, "learning_rate": 6.120050184270369e-05, "loss": 2.897, "step": 49481 }, { "epoch": 3.88, "learning_rate": 6.119971771347919e-05, "loss": 2.5419, "step": 49482 }, { "epoch": 3.88, "learning_rate": 6.11989335842547e-05, "loss": 2.7725, "step": 49483 }, { "epoch": 3.88, "learning_rate": 6.11981494550302e-05, "loss": 2.6647, "step": 49484 }, { "epoch": 3.88, "learning_rate": 6.11973653258057e-05, "loss": 2.8749, "step": 49485 }, { "epoch": 3.88, "learning_rate": 6.11965811965812e-05, "loss": 2.8456, "step": 49486 }, { "epoch": 3.88, "learning_rate": 6.119579706735671e-05, "loss": 2.4774, "step": 49487 }, { "epoch": 3.88, "learning_rate": 6.119501293813221e-05, "loss": 2.6759, "step": 49488 }, { "epoch": 3.88, "learning_rate": 6.11942288089077e-05, "loss": 2.8775, "step": 49489 }, { "epoch": 3.88, "learning_rate": 6.119344467968321e-05, "loss": 2.7586, "step": 49490 }, { "epoch": 3.88, "learning_rate": 6.119266055045871e-05, "loss": 2.6185, "step": 49491 }, { "epoch": 3.88, "learning_rate": 6.119187642123421e-05, "loss": 2.6296, "step": 49492 }, { "epoch": 3.88, "learning_rate": 6.119109229200972e-05, "loss": 2.744, "step": 49493 }, { "epoch": 3.88, "learning_rate": 6.119030816278522e-05, "loss": 2.7544, "step": 49494 }, { "epoch": 3.88, "learning_rate": 6.118952403356073e-05, "loss": 2.6093, "step": 49495 }, { "epoch": 3.88, "learning_rate": 6.118873990433623e-05, "loss": 2.797, "step": 49496 }, { "epoch": 3.88, "learning_rate": 6.118795577511173e-05, "loss": 2.9143, "step": 49497 }, { "epoch": 3.88, "learning_rate": 6.118717164588724e-05, "loss": 2.7353, "step": 49498 }, { "epoch": 3.88, "learning_rate": 6.118638751666274e-05, "loss": 2.8581, "step": 49499 }, { "epoch": 3.88, "learning_rate": 6.118560338743824e-05, "loss": 2.5904, "step": 49500 }, { "epoch": 3.88, "learning_rate": 6.118481925821375e-05, "loss": 2.9278, "step": 49501 }, { "epoch": 3.88, "learning_rate": 6.118403512898927e-05, "loss": 2.6892, "step": 49502 }, { "epoch": 3.88, "learning_rate": 6.118325099976477e-05, "loss": 2.6403, "step": 49503 }, { "epoch": 3.88, "learning_rate": 6.118246687054027e-05, "loss": 2.5004, "step": 49504 }, { "epoch": 3.88, "learning_rate": 6.118168274131578e-05, "loss": 2.6488, "step": 49505 }, { "epoch": 3.88, "learning_rate": 6.118089861209128e-05, "loss": 2.7187, "step": 49506 }, { "epoch": 3.88, "learning_rate": 6.118011448286678e-05, "loss": 2.9062, "step": 49507 }, { "epoch": 3.88, "learning_rate": 6.117933035364229e-05, "loss": 2.8231, "step": 49508 }, { "epoch": 3.88, "learning_rate": 6.117854622441779e-05, "loss": 2.694, "step": 49509 }, { "epoch": 3.88, "learning_rate": 6.11777620951933e-05, "loss": 2.7626, "step": 49510 }, { "epoch": 3.88, "learning_rate": 6.11769779659688e-05, "loss": 2.7292, "step": 49511 }, { "epoch": 3.88, "learning_rate": 6.11761938367443e-05, "loss": 2.8342, "step": 49512 }, { "epoch": 3.88, "learning_rate": 6.11754097075198e-05, "loss": 2.741, "step": 49513 }, { "epoch": 3.88, "learning_rate": 6.117462557829531e-05, "loss": 2.6649, "step": 49514 }, { "epoch": 3.88, "learning_rate": 6.117384144907081e-05, "loss": 2.7197, "step": 49515 }, { "epoch": 3.88, "learning_rate": 6.117305731984632e-05, "loss": 2.7846, "step": 49516 }, { "epoch": 3.88, "learning_rate": 6.117227319062182e-05, "loss": 2.8118, "step": 49517 }, { "epoch": 3.88, "learning_rate": 6.117148906139732e-05, "loss": 2.6428, "step": 49518 }, { "epoch": 3.88, "learning_rate": 6.117070493217283e-05, "loss": 2.7467, "step": 49519 }, { "epoch": 3.88, "learning_rate": 6.116992080294833e-05, "loss": 2.813, "step": 49520 }, { "epoch": 3.88, "learning_rate": 6.116913667372384e-05, "loss": 2.6638, "step": 49521 }, { "epoch": 3.88, "learning_rate": 6.116835254449934e-05, "loss": 2.6674, "step": 49522 }, { "epoch": 3.88, "learning_rate": 6.116756841527484e-05, "loss": 2.6994, "step": 49523 }, { "epoch": 3.88, "learning_rate": 6.116678428605035e-05, "loss": 2.6198, "step": 49524 }, { "epoch": 3.88, "learning_rate": 6.116600015682584e-05, "loss": 2.8752, "step": 49525 }, { "epoch": 3.88, "learning_rate": 6.116521602760134e-05, "loss": 2.738, "step": 49526 }, { "epoch": 3.88, "learning_rate": 6.116443189837686e-05, "loss": 2.6772, "step": 49527 }, { "epoch": 3.88, "learning_rate": 6.116364776915236e-05, "loss": 2.6886, "step": 49528 }, { "epoch": 3.88, "learning_rate": 6.116286363992786e-05, "loss": 2.7286, "step": 49529 }, { "epoch": 3.88, "learning_rate": 6.116207951070337e-05, "loss": 2.9245, "step": 49530 }, { "epoch": 3.88, "learning_rate": 6.116129538147887e-05, "loss": 2.889, "step": 49531 }, { "epoch": 3.88, "learning_rate": 6.116051125225438e-05, "loss": 2.738, "step": 49532 }, { "epoch": 3.88, "learning_rate": 6.115972712302988e-05, "loss": 2.6837, "step": 49533 }, { "epoch": 3.88, "learning_rate": 6.115894299380538e-05, "loss": 2.5315, "step": 49534 }, { "epoch": 3.88, "learning_rate": 6.115815886458089e-05, "loss": 2.6754, "step": 49535 }, { "epoch": 3.88, "learning_rate": 6.115737473535639e-05, "loss": 2.5182, "step": 49536 }, { "epoch": 3.88, "learning_rate": 6.11565906061319e-05, "loss": 2.6847, "step": 49537 }, { "epoch": 3.88, "learning_rate": 6.11558064769074e-05, "loss": 2.8632, "step": 49538 }, { "epoch": 3.88, "learning_rate": 6.11550223476829e-05, "loss": 2.7227, "step": 49539 }, { "epoch": 3.88, "learning_rate": 6.11542382184584e-05, "loss": 2.6302, "step": 49540 }, { "epoch": 3.88, "learning_rate": 6.115345408923391e-05, "loss": 2.7691, "step": 49541 }, { "epoch": 3.88, "learning_rate": 6.115266996000941e-05, "loss": 2.8758, "step": 49542 }, { "epoch": 3.88, "learning_rate": 6.115188583078492e-05, "loss": 2.7473, "step": 49543 }, { "epoch": 3.88, "learning_rate": 6.115110170156042e-05, "loss": 2.8025, "step": 49544 }, { "epoch": 3.88, "learning_rate": 6.115031757233592e-05, "loss": 2.5457, "step": 49545 }, { "epoch": 3.89, "learning_rate": 6.114953344311143e-05, "loss": 2.8511, "step": 49546 }, { "epoch": 3.89, "learning_rate": 6.114874931388693e-05, "loss": 2.6603, "step": 49547 }, { "epoch": 3.89, "learning_rate": 6.114796518466243e-05, "loss": 2.564, "step": 49548 }, { "epoch": 3.89, "learning_rate": 6.114718105543794e-05, "loss": 2.5178, "step": 49549 }, { "epoch": 3.89, "learning_rate": 6.114639692621344e-05, "loss": 2.6643, "step": 49550 }, { "epoch": 3.89, "learning_rate": 6.114561279698895e-05, "loss": 2.8709, "step": 49551 }, { "epoch": 3.89, "learning_rate": 6.114482866776445e-05, "loss": 2.7355, "step": 49552 }, { "epoch": 3.89, "learning_rate": 6.114404453853995e-05, "loss": 2.7576, "step": 49553 }, { "epoch": 3.89, "learning_rate": 6.114326040931546e-05, "loss": 2.8639, "step": 49554 }, { "epoch": 3.89, "learning_rate": 6.114247628009096e-05, "loss": 2.6868, "step": 49555 }, { "epoch": 3.89, "learning_rate": 6.114169215086646e-05, "loss": 2.7146, "step": 49556 }, { "epoch": 3.89, "learning_rate": 6.114090802164197e-05, "loss": 2.5239, "step": 49557 }, { "epoch": 3.89, "learning_rate": 6.114012389241747e-05, "loss": 2.7925, "step": 49558 }, { "epoch": 3.89, "learning_rate": 6.113933976319299e-05, "loss": 2.7055, "step": 49559 }, { "epoch": 3.89, "learning_rate": 6.113855563396849e-05, "loss": 2.7257, "step": 49560 }, { "epoch": 3.89, "learning_rate": 6.113777150474398e-05, "loss": 2.824, "step": 49561 }, { "epoch": 3.89, "learning_rate": 6.113698737551949e-05, "loss": 2.8271, "step": 49562 }, { "epoch": 3.89, "learning_rate": 6.113620324629499e-05, "loss": 2.512, "step": 49563 }, { "epoch": 3.89, "learning_rate": 6.11354191170705e-05, "loss": 2.9625, "step": 49564 }, { "epoch": 3.89, "learning_rate": 6.1134634987846e-05, "loss": 2.7145, "step": 49565 }, { "epoch": 3.89, "learning_rate": 6.11338508586215e-05, "loss": 2.8013, "step": 49566 }, { "epoch": 3.89, "learning_rate": 6.1133066729397e-05, "loss": 2.6464, "step": 49567 }, { "epoch": 3.89, "learning_rate": 6.113228260017251e-05, "loss": 2.7409, "step": 49568 }, { "epoch": 3.89, "learning_rate": 6.113149847094801e-05, "loss": 2.7384, "step": 49569 }, { "epoch": 3.89, "learning_rate": 6.113071434172352e-05, "loss": 2.7483, "step": 49570 }, { "epoch": 3.89, "learning_rate": 6.112993021249902e-05, "loss": 2.995, "step": 49571 }, { "epoch": 3.89, "learning_rate": 6.112914608327452e-05, "loss": 2.8527, "step": 49572 }, { "epoch": 3.89, "learning_rate": 6.112836195405003e-05, "loss": 2.6238, "step": 49573 }, { "epoch": 3.89, "learning_rate": 6.112757782482553e-05, "loss": 2.655, "step": 49574 }, { "epoch": 3.89, "learning_rate": 6.112679369560103e-05, "loss": 2.6051, "step": 49575 }, { "epoch": 3.89, "learning_rate": 6.112600956637654e-05, "loss": 2.6842, "step": 49576 }, { "epoch": 3.89, "learning_rate": 6.112522543715204e-05, "loss": 2.7357, "step": 49577 }, { "epoch": 3.89, "learning_rate": 6.112444130792754e-05, "loss": 2.5558, "step": 49578 }, { "epoch": 3.89, "learning_rate": 6.112365717870305e-05, "loss": 2.7592, "step": 49579 }, { "epoch": 3.89, "learning_rate": 6.112287304947855e-05, "loss": 2.7061, "step": 49580 }, { "epoch": 3.89, "learning_rate": 6.112208892025406e-05, "loss": 2.7046, "step": 49581 }, { "epoch": 3.89, "learning_rate": 6.112130479102956e-05, "loss": 2.6666, "step": 49582 }, { "epoch": 3.89, "learning_rate": 6.112052066180506e-05, "loss": 2.6051, "step": 49583 }, { "epoch": 3.89, "learning_rate": 6.111973653258058e-05, "loss": 2.6082, "step": 49584 }, { "epoch": 3.89, "learning_rate": 6.111895240335608e-05, "loss": 2.6577, "step": 49585 }, { "epoch": 3.89, "learning_rate": 6.111816827413159e-05, "loss": 2.7636, "step": 49586 }, { "epoch": 3.89, "learning_rate": 6.111738414490709e-05, "loss": 2.7282, "step": 49587 }, { "epoch": 3.89, "learning_rate": 6.11166000156826e-05, "loss": 2.7263, "step": 49588 }, { "epoch": 3.89, "learning_rate": 6.11158158864581e-05, "loss": 2.7947, "step": 49589 }, { "epoch": 3.89, "learning_rate": 6.11150317572336e-05, "loss": 2.7338, "step": 49590 }, { "epoch": 3.89, "learning_rate": 6.11142476280091e-05, "loss": 2.7442, "step": 49591 }, { "epoch": 3.89, "learning_rate": 6.111346349878461e-05, "loss": 2.5638, "step": 49592 }, { "epoch": 3.89, "learning_rate": 6.111267936956011e-05, "loss": 2.5828, "step": 49593 }, { "epoch": 3.89, "learning_rate": 6.111189524033562e-05, "loss": 2.8723, "step": 49594 }, { "epoch": 3.89, "learning_rate": 6.111111111111112e-05, "loss": 2.8857, "step": 49595 }, { "epoch": 3.89, "learning_rate": 6.111032698188662e-05, "loss": 2.6675, "step": 49596 }, { "epoch": 3.89, "learning_rate": 6.110954285266213e-05, "loss": 2.9048, "step": 49597 }, { "epoch": 3.89, "learning_rate": 6.110875872343762e-05, "loss": 2.7469, "step": 49598 }, { "epoch": 3.89, "learning_rate": 6.110797459421312e-05, "loss": 2.6895, "step": 49599 }, { "epoch": 3.89, "learning_rate": 6.110719046498863e-05, "loss": 2.7109, "step": 49600 }, { "epoch": 3.89, "learning_rate": 6.110640633576413e-05, "loss": 2.7174, "step": 49601 }, { "epoch": 3.89, "learning_rate": 6.110562220653963e-05, "loss": 3.0095, "step": 49602 }, { "epoch": 3.89, "learning_rate": 6.110483807731514e-05, "loss": 2.8169, "step": 49603 }, { "epoch": 3.89, "learning_rate": 6.110405394809064e-05, "loss": 2.9944, "step": 49604 }, { "epoch": 3.89, "learning_rate": 6.110326981886614e-05, "loss": 2.6108, "step": 49605 }, { "epoch": 3.89, "learning_rate": 6.110248568964165e-05, "loss": 2.7505, "step": 49606 }, { "epoch": 3.89, "learning_rate": 6.110170156041715e-05, "loss": 2.502, "step": 49607 }, { "epoch": 3.89, "learning_rate": 6.110091743119266e-05, "loss": 2.5658, "step": 49608 }, { "epoch": 3.89, "learning_rate": 6.110013330196816e-05, "loss": 2.5051, "step": 49609 }, { "epoch": 3.89, "learning_rate": 6.109934917274368e-05, "loss": 2.844, "step": 49610 }, { "epoch": 3.89, "learning_rate": 6.109856504351918e-05, "loss": 2.5936, "step": 49611 }, { "epoch": 3.89, "learning_rate": 6.109778091429468e-05, "loss": 2.8552, "step": 49612 }, { "epoch": 3.89, "learning_rate": 6.109699678507019e-05, "loss": 2.5461, "step": 49613 }, { "epoch": 3.89, "learning_rate": 6.109621265584569e-05, "loss": 2.7498, "step": 49614 }, { "epoch": 3.89, "learning_rate": 6.10954285266212e-05, "loss": 2.6572, "step": 49615 }, { "epoch": 3.89, "learning_rate": 6.10946443973967e-05, "loss": 2.9377, "step": 49616 }, { "epoch": 3.89, "learning_rate": 6.10938602681722e-05, "loss": 2.8162, "step": 49617 }, { "epoch": 3.89, "learning_rate": 6.10930761389477e-05, "loss": 2.9085, "step": 49618 }, { "epoch": 3.89, "learning_rate": 6.109229200972321e-05, "loss": 2.6365, "step": 49619 }, { "epoch": 3.89, "learning_rate": 6.109150788049871e-05, "loss": 2.5841, "step": 49620 }, { "epoch": 3.89, "learning_rate": 6.109072375127422e-05, "loss": 2.6491, "step": 49621 }, { "epoch": 3.89, "learning_rate": 6.108993962204972e-05, "loss": 2.7683, "step": 49622 }, { "epoch": 3.89, "learning_rate": 6.108915549282522e-05, "loss": 2.5348, "step": 49623 }, { "epoch": 3.89, "learning_rate": 6.108837136360073e-05, "loss": 2.9351, "step": 49624 }, { "epoch": 3.89, "learning_rate": 6.108758723437623e-05, "loss": 2.8919, "step": 49625 }, { "epoch": 3.89, "learning_rate": 6.108680310515173e-05, "loss": 2.5578, "step": 49626 }, { "epoch": 3.89, "learning_rate": 6.108601897592724e-05, "loss": 2.9999, "step": 49627 }, { "epoch": 3.89, "learning_rate": 6.108523484670274e-05, "loss": 2.6661, "step": 49628 }, { "epoch": 3.89, "learning_rate": 6.108445071747825e-05, "loss": 2.8694, "step": 49629 }, { "epoch": 3.89, "learning_rate": 6.108366658825375e-05, "loss": 2.745, "step": 49630 }, { "epoch": 3.89, "learning_rate": 6.108288245902925e-05, "loss": 2.6418, "step": 49631 }, { "epoch": 3.89, "learning_rate": 6.108209832980476e-05, "loss": 2.6727, "step": 49632 }, { "epoch": 3.89, "learning_rate": 6.108131420058026e-05, "loss": 2.7702, "step": 49633 }, { "epoch": 3.89, "learning_rate": 6.108053007135575e-05, "loss": 2.8622, "step": 49634 }, { "epoch": 3.89, "learning_rate": 6.107974594213127e-05, "loss": 2.7239, "step": 49635 }, { "epoch": 3.89, "learning_rate": 6.107896181290677e-05, "loss": 2.6014, "step": 49636 }, { "epoch": 3.89, "learning_rate": 6.107817768368228e-05, "loss": 2.6564, "step": 49637 }, { "epoch": 3.89, "learning_rate": 6.107739355445778e-05, "loss": 2.7792, "step": 49638 }, { "epoch": 3.89, "learning_rate": 6.107660942523328e-05, "loss": 2.6821, "step": 49639 }, { "epoch": 3.89, "learning_rate": 6.107582529600879e-05, "loss": 2.6634, "step": 49640 }, { "epoch": 3.89, "learning_rate": 6.107504116678429e-05, "loss": 2.7675, "step": 49641 }, { "epoch": 3.89, "learning_rate": 6.10742570375598e-05, "loss": 2.9006, "step": 49642 }, { "epoch": 3.89, "learning_rate": 6.10734729083353e-05, "loss": 2.593, "step": 49643 }, { "epoch": 3.89, "learning_rate": 6.10726887791108e-05, "loss": 2.7155, "step": 49644 }, { "epoch": 3.89, "learning_rate": 6.10719046498863e-05, "loss": 2.8123, "step": 49645 }, { "epoch": 3.89, "learning_rate": 6.107112052066181e-05, "loss": 2.5092, "step": 49646 }, { "epoch": 3.89, "learning_rate": 6.107033639143731e-05, "loss": 2.6817, "step": 49647 }, { "epoch": 3.89, "learning_rate": 6.106955226221282e-05, "loss": 2.6539, "step": 49648 }, { "epoch": 3.89, "learning_rate": 6.106876813298832e-05, "loss": 2.7111, "step": 49649 }, { "epoch": 3.89, "learning_rate": 6.106798400376382e-05, "loss": 2.5489, "step": 49650 }, { "epoch": 3.89, "learning_rate": 6.106719987453933e-05, "loss": 2.5827, "step": 49651 }, { "epoch": 3.89, "learning_rate": 6.106641574531483e-05, "loss": 2.5781, "step": 49652 }, { "epoch": 3.89, "learning_rate": 6.106563161609033e-05, "loss": 2.7376, "step": 49653 }, { "epoch": 3.89, "learning_rate": 6.106484748686584e-05, "loss": 2.8297, "step": 49654 }, { "epoch": 3.89, "learning_rate": 6.106406335764134e-05, "loss": 3.063, "step": 49655 }, { "epoch": 3.89, "learning_rate": 6.106327922841685e-05, "loss": 2.7171, "step": 49656 }, { "epoch": 3.89, "learning_rate": 6.106249509919235e-05, "loss": 2.8178, "step": 49657 }, { "epoch": 3.89, "learning_rate": 6.106171096996785e-05, "loss": 2.5503, "step": 49658 }, { "epoch": 3.89, "learning_rate": 6.106092684074336e-05, "loss": 2.6787, "step": 49659 }, { "epoch": 3.89, "learning_rate": 6.106014271151886e-05, "loss": 2.8374, "step": 49660 }, { "epoch": 3.89, "learning_rate": 6.105935858229436e-05, "loss": 2.696, "step": 49661 }, { "epoch": 3.89, "learning_rate": 6.105857445306987e-05, "loss": 2.8039, "step": 49662 }, { "epoch": 3.89, "learning_rate": 6.105779032384537e-05, "loss": 2.7267, "step": 49663 }, { "epoch": 3.89, "learning_rate": 6.105700619462087e-05, "loss": 2.6309, "step": 49664 }, { "epoch": 3.89, "learning_rate": 6.105622206539638e-05, "loss": 2.7247, "step": 49665 }, { "epoch": 3.89, "learning_rate": 6.105543793617188e-05, "loss": 2.7093, "step": 49666 }, { "epoch": 3.89, "learning_rate": 6.10546538069474e-05, "loss": 2.7721, "step": 49667 }, { "epoch": 3.89, "learning_rate": 6.10538696777229e-05, "loss": 2.7631, "step": 49668 }, { "epoch": 3.89, "learning_rate": 6.10530855484984e-05, "loss": 2.7141, "step": 49669 }, { "epoch": 3.89, "learning_rate": 6.10523014192739e-05, "loss": 2.7645, "step": 49670 }, { "epoch": 3.89, "learning_rate": 6.10515172900494e-05, "loss": 2.7452, "step": 49671 }, { "epoch": 3.89, "learning_rate": 6.10507331608249e-05, "loss": 2.98, "step": 49672 }, { "epoch": 3.9, "learning_rate": 6.104994903160041e-05, "loss": 2.7143, "step": 49673 }, { "epoch": 3.9, "learning_rate": 6.104916490237591e-05, "loss": 2.7581, "step": 49674 }, { "epoch": 3.9, "learning_rate": 6.104838077315142e-05, "loss": 2.7051, "step": 49675 }, { "epoch": 3.9, "learning_rate": 6.104759664392692e-05, "loss": 2.8523, "step": 49676 }, { "epoch": 3.9, "learning_rate": 6.104681251470242e-05, "loss": 2.6354, "step": 49677 }, { "epoch": 3.9, "learning_rate": 6.104602838547793e-05, "loss": 2.4212, "step": 49678 }, { "epoch": 3.9, "learning_rate": 6.104524425625343e-05, "loss": 2.7203, "step": 49679 }, { "epoch": 3.9, "learning_rate": 6.104446012702893e-05, "loss": 2.717, "step": 49680 }, { "epoch": 3.9, "learning_rate": 6.104367599780444e-05, "loss": 2.7695, "step": 49681 }, { "epoch": 3.9, "learning_rate": 6.104289186857994e-05, "loss": 2.585, "step": 49682 }, { "epoch": 3.9, "learning_rate": 6.104210773935544e-05, "loss": 2.5137, "step": 49683 }, { "epoch": 3.9, "learning_rate": 6.104132361013095e-05, "loss": 2.5375, "step": 49684 }, { "epoch": 3.9, "learning_rate": 6.104053948090645e-05, "loss": 2.5641, "step": 49685 }, { "epoch": 3.9, "learning_rate": 6.103975535168196e-05, "loss": 2.6401, "step": 49686 }, { "epoch": 3.9, "learning_rate": 6.103897122245746e-05, "loss": 2.8189, "step": 49687 }, { "epoch": 3.9, "learning_rate": 6.103818709323296e-05, "loss": 2.721, "step": 49688 }, { "epoch": 3.9, "learning_rate": 6.103740296400847e-05, "loss": 2.5782, "step": 49689 }, { "epoch": 3.9, "learning_rate": 6.103661883478397e-05, "loss": 2.7436, "step": 49690 }, { "epoch": 3.9, "learning_rate": 6.103583470555947e-05, "loss": 2.6454, "step": 49691 }, { "epoch": 3.9, "learning_rate": 6.1035050576334984e-05, "loss": 2.7732, "step": 49692 }, { "epoch": 3.9, "learning_rate": 6.103426644711049e-05, "loss": 2.9794, "step": 49693 }, { "epoch": 3.9, "learning_rate": 6.103348231788599e-05, "loss": 2.8505, "step": 49694 }, { "epoch": 3.9, "learning_rate": 6.1032698188661495e-05, "loss": 2.5577, "step": 49695 }, { "epoch": 3.9, "learning_rate": 6.1031914059437e-05, "loss": 2.7495, "step": 49696 }, { "epoch": 3.9, "learning_rate": 6.10311299302125e-05, "loss": 2.6651, "step": 49697 }, { "epoch": 3.9, "learning_rate": 6.1030345800988006e-05, "loss": 2.7755, "step": 49698 }, { "epoch": 3.9, "learning_rate": 6.102956167176351e-05, "loss": 2.5191, "step": 49699 }, { "epoch": 3.9, "learning_rate": 6.102877754253902e-05, "loss": 2.8033, "step": 49700 }, { "epoch": 3.9, "learning_rate": 6.1027993413314524e-05, "loss": 2.8232, "step": 49701 }, { "epoch": 3.9, "learning_rate": 6.102720928409003e-05, "loss": 2.7687, "step": 49702 }, { "epoch": 3.9, "learning_rate": 6.102642515486553e-05, "loss": 2.6987, "step": 49703 }, { "epoch": 3.9, "learning_rate": 6.1025641025641035e-05, "loss": 2.9232, "step": 49704 }, { "epoch": 3.9, "learning_rate": 6.102485689641654e-05, "loss": 2.858, "step": 49705 }, { "epoch": 3.9, "learning_rate": 6.102407276719203e-05, "loss": 2.8295, "step": 49706 }, { "epoch": 3.9, "learning_rate": 6.102328863796753e-05, "loss": 2.7138, "step": 49707 }, { "epoch": 3.9, "learning_rate": 6.1022504508743036e-05, "loss": 2.8721, "step": 49708 }, { "epoch": 3.9, "learning_rate": 6.102172037951854e-05, "loss": 2.7549, "step": 49709 }, { "epoch": 3.9, "learning_rate": 6.102093625029405e-05, "loss": 2.6874, "step": 49710 }, { "epoch": 3.9, "learning_rate": 6.1020152121069554e-05, "loss": 2.7545, "step": 49711 }, { "epoch": 3.9, "learning_rate": 6.101936799184506e-05, "loss": 2.7837, "step": 49712 }, { "epoch": 3.9, "learning_rate": 6.101858386262056e-05, "loss": 2.778, "step": 49713 }, { "epoch": 3.9, "learning_rate": 6.1017799733396065e-05, "loss": 2.7507, "step": 49714 }, { "epoch": 3.9, "learning_rate": 6.101701560417157e-05, "loss": 2.7673, "step": 49715 }, { "epoch": 3.9, "learning_rate": 6.101623147494707e-05, "loss": 2.6243, "step": 49716 }, { "epoch": 3.9, "learning_rate": 6.1015447345722576e-05, "loss": 2.8392, "step": 49717 }, { "epoch": 3.9, "learning_rate": 6.101466321649808e-05, "loss": 2.846, "step": 49718 }, { "epoch": 3.9, "learning_rate": 6.1013879087273584e-05, "loss": 2.5226, "step": 49719 }, { "epoch": 3.9, "learning_rate": 6.101309495804909e-05, "loss": 2.6191, "step": 49720 }, { "epoch": 3.9, "learning_rate": 6.101231082882459e-05, "loss": 2.6002, "step": 49721 }, { "epoch": 3.9, "learning_rate": 6.1011526699600095e-05, "loss": 2.6955, "step": 49722 }, { "epoch": 3.9, "learning_rate": 6.10107425703756e-05, "loss": 2.7702, "step": 49723 }, { "epoch": 3.9, "learning_rate": 6.10099584411511e-05, "loss": 2.9542, "step": 49724 }, { "epoch": 3.9, "learning_rate": 6.1009174311926606e-05, "loss": 2.433, "step": 49725 }, { "epoch": 3.9, "learning_rate": 6.1008390182702116e-05, "loss": 2.8948, "step": 49726 }, { "epoch": 3.9, "learning_rate": 6.100760605347762e-05, "loss": 2.9083, "step": 49727 }, { "epoch": 3.9, "learning_rate": 6.1006821924253124e-05, "loss": 2.7868, "step": 49728 }, { "epoch": 3.9, "learning_rate": 6.100603779502863e-05, "loss": 2.6491, "step": 49729 }, { "epoch": 3.9, "learning_rate": 6.100525366580413e-05, "loss": 2.7854, "step": 49730 }, { "epoch": 3.9, "learning_rate": 6.1004469536579635e-05, "loss": 2.6256, "step": 49731 }, { "epoch": 3.9, "learning_rate": 6.100368540735514e-05, "loss": 2.7135, "step": 49732 }, { "epoch": 3.9, "learning_rate": 6.100290127813064e-05, "loss": 2.6161, "step": 49733 }, { "epoch": 3.9, "learning_rate": 6.1002117148906146e-05, "loss": 2.5413, "step": 49734 }, { "epoch": 3.9, "learning_rate": 6.100133301968165e-05, "loss": 2.6465, "step": 49735 }, { "epoch": 3.9, "learning_rate": 6.100054889045715e-05, "loss": 2.6512, "step": 49736 }, { "epoch": 3.9, "learning_rate": 6.0999764761232657e-05, "loss": 3.1839, "step": 49737 }, { "epoch": 3.9, "learning_rate": 6.099898063200816e-05, "loss": 3.0504, "step": 49738 }, { "epoch": 3.9, "learning_rate": 6.0998196502783664e-05, "loss": 2.7192, "step": 49739 }, { "epoch": 3.9, "learning_rate": 6.099741237355917e-05, "loss": 2.9461, "step": 49740 }, { "epoch": 3.9, "learning_rate": 6.099662824433468e-05, "loss": 2.7336, "step": 49741 }, { "epoch": 3.9, "learning_rate": 6.099584411511017e-05, "loss": 2.8055, "step": 49742 }, { "epoch": 3.9, "learning_rate": 6.099505998588567e-05, "loss": 2.7127, "step": 49743 }, { "epoch": 3.9, "learning_rate": 6.0994275856661176e-05, "loss": 2.6518, "step": 49744 }, { "epoch": 3.9, "learning_rate": 6.099349172743668e-05, "loss": 2.8607, "step": 49745 }, { "epoch": 3.9, "learning_rate": 6.099270759821218e-05, "loss": 2.5454, "step": 49746 }, { "epoch": 3.9, "learning_rate": 6.0991923468987687e-05, "loss": 2.828, "step": 49747 }, { "epoch": 3.9, "learning_rate": 6.099113933976319e-05, "loss": 3.257, "step": 49748 }, { "epoch": 3.9, "learning_rate": 6.0990355210538694e-05, "loss": 2.5573, "step": 49749 }, { "epoch": 3.9, "learning_rate": 6.09895710813142e-05, "loss": 2.8402, "step": 49750 }, { "epoch": 3.9, "learning_rate": 6.098878695208971e-05, "loss": 2.6729, "step": 49751 }, { "epoch": 3.9, "learning_rate": 6.098800282286521e-05, "loss": 2.7067, "step": 49752 }, { "epoch": 3.9, "learning_rate": 6.0987218693640715e-05, "loss": 2.8303, "step": 49753 }, { "epoch": 3.9, "learning_rate": 6.098643456441622e-05, "loss": 2.7479, "step": 49754 }, { "epoch": 3.9, "learning_rate": 6.098565043519172e-05, "loss": 2.4897, "step": 49755 }, { "epoch": 3.9, "learning_rate": 6.0984866305967226e-05, "loss": 2.9287, "step": 49756 }, { "epoch": 3.9, "learning_rate": 6.098408217674273e-05, "loss": 2.6669, "step": 49757 }, { "epoch": 3.9, "learning_rate": 6.0983298047518234e-05, "loss": 2.7597, "step": 49758 }, { "epoch": 3.9, "learning_rate": 6.098251391829374e-05, "loss": 2.6691, "step": 49759 }, { "epoch": 3.9, "learning_rate": 6.098172978906924e-05, "loss": 2.6034, "step": 49760 }, { "epoch": 3.9, "learning_rate": 6.0980945659844745e-05, "loss": 2.7016, "step": 49761 }, { "epoch": 3.9, "learning_rate": 6.098016153062025e-05, "loss": 2.7351, "step": 49762 }, { "epoch": 3.9, "learning_rate": 6.097937740139575e-05, "loss": 2.9007, "step": 49763 }, { "epoch": 3.9, "learning_rate": 6.0978593272171256e-05, "loss": 2.6385, "step": 49764 }, { "epoch": 3.9, "learning_rate": 6.097780914294676e-05, "loss": 2.7458, "step": 49765 }, { "epoch": 3.9, "learning_rate": 6.097702501372226e-05, "loss": 2.951, "step": 49766 }, { "epoch": 3.9, "learning_rate": 6.0976240884497774e-05, "loss": 2.9485, "step": 49767 }, { "epoch": 3.9, "learning_rate": 6.097545675527328e-05, "loss": 2.8648, "step": 49768 }, { "epoch": 3.9, "learning_rate": 6.097467262604878e-05, "loss": 2.9297, "step": 49769 }, { "epoch": 3.9, "learning_rate": 6.0973888496824285e-05, "loss": 2.6446, "step": 49770 }, { "epoch": 3.9, "learning_rate": 6.097310436759979e-05, "loss": 2.6871, "step": 49771 }, { "epoch": 3.9, "learning_rate": 6.097232023837529e-05, "loss": 2.8012, "step": 49772 }, { "epoch": 3.9, "learning_rate": 6.0971536109150796e-05, "loss": 2.7562, "step": 49773 }, { "epoch": 3.9, "learning_rate": 6.09707519799263e-05, "loss": 2.8545, "step": 49774 }, { "epoch": 3.9, "learning_rate": 6.09699678507018e-05, "loss": 2.9094, "step": 49775 }, { "epoch": 3.9, "learning_rate": 6.096918372147731e-05, "loss": 2.6433, "step": 49776 }, { "epoch": 3.9, "learning_rate": 6.096839959225281e-05, "loss": 3.1193, "step": 49777 }, { "epoch": 3.9, "learning_rate": 6.096761546302831e-05, "loss": 2.7256, "step": 49778 }, { "epoch": 3.9, "learning_rate": 6.096683133380381e-05, "loss": 2.7622, "step": 49779 }, { "epoch": 3.9, "learning_rate": 6.0966047204579315e-05, "loss": 2.8277, "step": 49780 }, { "epoch": 3.9, "learning_rate": 6.096526307535482e-05, "loss": 2.7385, "step": 49781 }, { "epoch": 3.9, "learning_rate": 6.096447894613032e-05, "loss": 2.7434, "step": 49782 }, { "epoch": 3.9, "learning_rate": 6.0963694816905826e-05, "loss": 2.8743, "step": 49783 }, { "epoch": 3.9, "learning_rate": 6.096291068768133e-05, "loss": 2.5138, "step": 49784 }, { "epoch": 3.9, "learning_rate": 6.096212655845683e-05, "loss": 2.8327, "step": 49785 }, { "epoch": 3.9, "learning_rate": 6.096134242923234e-05, "loss": 2.6212, "step": 49786 }, { "epoch": 3.9, "learning_rate": 6.096055830000784e-05, "loss": 2.736, "step": 49787 }, { "epoch": 3.9, "learning_rate": 6.0959774170783344e-05, "loss": 2.7596, "step": 49788 }, { "epoch": 3.9, "learning_rate": 6.095899004155885e-05, "loss": 2.4189, "step": 49789 }, { "epoch": 3.9, "learning_rate": 6.095820591233435e-05, "loss": 3.1426, "step": 49790 }, { "epoch": 3.9, "learning_rate": 6.0957421783109855e-05, "loss": 3.1548, "step": 49791 }, { "epoch": 3.9, "learning_rate": 6.0956637653885366e-05, "loss": 2.7206, "step": 49792 }, { "epoch": 3.9, "learning_rate": 6.095585352466087e-05, "loss": 2.6228, "step": 49793 }, { "epoch": 3.9, "learning_rate": 6.095506939543637e-05, "loss": 2.8396, "step": 49794 }, { "epoch": 3.9, "learning_rate": 6.095428526621188e-05, "loss": 2.8155, "step": 49795 }, { "epoch": 3.9, "learning_rate": 6.095350113698738e-05, "loss": 2.7766, "step": 49796 }, { "epoch": 3.9, "learning_rate": 6.0952717007762884e-05, "loss": 2.7995, "step": 49797 }, { "epoch": 3.9, "learning_rate": 6.095193287853839e-05, "loss": 2.6332, "step": 49798 }, { "epoch": 3.9, "learning_rate": 6.095114874931389e-05, "loss": 2.5868, "step": 49799 }, { "epoch": 3.9, "learning_rate": 6.0950364620089395e-05, "loss": 2.6507, "step": 49800 }, { "epoch": 3.91, "learning_rate": 6.09495804908649e-05, "loss": 2.7113, "step": 49801 }, { "epoch": 3.91, "learning_rate": 6.09487963616404e-05, "loss": 2.921, "step": 49802 }, { "epoch": 3.91, "learning_rate": 6.0948012232415906e-05, "loss": 2.8249, "step": 49803 }, { "epoch": 3.91, "learning_rate": 6.094722810319141e-05, "loss": 2.7984, "step": 49804 }, { "epoch": 3.91, "learning_rate": 6.0946443973966913e-05, "loss": 2.6247, "step": 49805 }, { "epoch": 3.91, "learning_rate": 6.094565984474242e-05, "loss": 2.632, "step": 49806 }, { "epoch": 3.91, "learning_rate": 6.094487571551792e-05, "loss": 2.9711, "step": 49807 }, { "epoch": 3.91, "learning_rate": 6.094409158629343e-05, "loss": 2.7048, "step": 49808 }, { "epoch": 3.91, "learning_rate": 6.0943307457068935e-05, "loss": 3.1428, "step": 49809 }, { "epoch": 3.91, "learning_rate": 6.094252332784444e-05, "loss": 2.8202, "step": 49810 }, { "epoch": 3.91, "learning_rate": 6.094173919861994e-05, "loss": 2.5874, "step": 49811 }, { "epoch": 3.91, "learning_rate": 6.0940955069395446e-05, "loss": 2.8058, "step": 49812 }, { "epoch": 3.91, "learning_rate": 6.094017094017095e-05, "loss": 2.7516, "step": 49813 }, { "epoch": 3.91, "learning_rate": 6.093938681094644e-05, "loss": 2.7683, "step": 49814 }, { "epoch": 3.91, "learning_rate": 6.0938602681721943e-05, "loss": 2.7851, "step": 49815 }, { "epoch": 3.91, "learning_rate": 6.093781855249745e-05, "loss": 2.8721, "step": 49816 }, { "epoch": 3.91, "learning_rate": 6.093703442327295e-05, "loss": 2.6293, "step": 49817 }, { "epoch": 3.91, "learning_rate": 6.093625029404846e-05, "loss": 2.7814, "step": 49818 }, { "epoch": 3.91, "learning_rate": 6.0935466164823965e-05, "loss": 2.5983, "step": 49819 }, { "epoch": 3.91, "learning_rate": 6.093468203559947e-05, "loss": 3.1158, "step": 49820 }, { "epoch": 3.91, "learning_rate": 6.093389790637497e-05, "loss": 2.5575, "step": 49821 }, { "epoch": 3.91, "learning_rate": 6.0933113777150476e-05, "loss": 2.7129, "step": 49822 }, { "epoch": 3.91, "learning_rate": 6.093232964792598e-05, "loss": 2.8285, "step": 49823 }, { "epoch": 3.91, "learning_rate": 6.093154551870148e-05, "loss": 2.5645, "step": 49824 }, { "epoch": 3.91, "learning_rate": 6.093076138947699e-05, "loss": 2.6401, "step": 49825 }, { "epoch": 3.91, "learning_rate": 6.092997726025249e-05, "loss": 2.6226, "step": 49826 }, { "epoch": 3.91, "learning_rate": 6.0929193131027994e-05, "loss": 2.8889, "step": 49827 }, { "epoch": 3.91, "learning_rate": 6.09284090018035e-05, "loss": 2.7506, "step": 49828 }, { "epoch": 3.91, "learning_rate": 6.0927624872579e-05, "loss": 2.6903, "step": 49829 }, { "epoch": 3.91, "learning_rate": 6.0926840743354505e-05, "loss": 2.5645, "step": 49830 }, { "epoch": 3.91, "learning_rate": 6.092605661413001e-05, "loss": 2.7387, "step": 49831 }, { "epoch": 3.91, "learning_rate": 6.092527248490551e-05, "loss": 2.7105, "step": 49832 }, { "epoch": 3.91, "learning_rate": 6.0924488355681016e-05, "loss": 2.6782, "step": 49833 }, { "epoch": 3.91, "learning_rate": 6.092370422645653e-05, "loss": 2.5106, "step": 49834 }, { "epoch": 3.91, "learning_rate": 6.092292009723203e-05, "loss": 2.8938, "step": 49835 }, { "epoch": 3.91, "learning_rate": 6.0922135968007534e-05, "loss": 2.6636, "step": 49836 }, { "epoch": 3.91, "learning_rate": 6.092135183878304e-05, "loss": 2.5708, "step": 49837 }, { "epoch": 3.91, "learning_rate": 6.092056770955854e-05, "loss": 2.7428, "step": 49838 }, { "epoch": 3.91, "learning_rate": 6.0919783580334045e-05, "loss": 2.6346, "step": 49839 }, { "epoch": 3.91, "learning_rate": 6.091899945110955e-05, "loss": 2.8544, "step": 49840 }, { "epoch": 3.91, "learning_rate": 6.091821532188505e-05, "loss": 2.6598, "step": 49841 }, { "epoch": 3.91, "learning_rate": 6.0917431192660556e-05, "loss": 2.935, "step": 49842 }, { "epoch": 3.91, "learning_rate": 6.091664706343606e-05, "loss": 2.6373, "step": 49843 }, { "epoch": 3.91, "learning_rate": 6.0915862934211564e-05, "loss": 2.7147, "step": 49844 }, { "epoch": 3.91, "learning_rate": 6.091507880498707e-05, "loss": 2.7076, "step": 49845 }, { "epoch": 3.91, "learning_rate": 6.091429467576257e-05, "loss": 2.8493, "step": 49846 }, { "epoch": 3.91, "learning_rate": 6.0913510546538075e-05, "loss": 2.5516, "step": 49847 }, { "epoch": 3.91, "learning_rate": 6.091272641731358e-05, "loss": 2.6958, "step": 49848 }, { "epoch": 3.91, "learning_rate": 6.091194228808909e-05, "loss": 2.6344, "step": 49849 }, { "epoch": 3.91, "learning_rate": 6.091115815886458e-05, "loss": 2.7258, "step": 49850 }, { "epoch": 3.91, "learning_rate": 6.091037402964008e-05, "loss": 2.764, "step": 49851 }, { "epoch": 3.91, "learning_rate": 6.0909589900415586e-05, "loss": 2.5569, "step": 49852 }, { "epoch": 3.91, "learning_rate": 6.090880577119109e-05, "loss": 2.5988, "step": 49853 }, { "epoch": 3.91, "learning_rate": 6.0908021641966594e-05, "loss": 2.7388, "step": 49854 }, { "epoch": 3.91, "learning_rate": 6.09072375127421e-05, "loss": 2.8774, "step": 49855 }, { "epoch": 3.91, "learning_rate": 6.09064533835176e-05, "loss": 2.5984, "step": 49856 }, { "epoch": 3.91, "learning_rate": 6.0905669254293105e-05, "loss": 2.6712, "step": 49857 }, { "epoch": 3.91, "learning_rate": 6.090488512506861e-05, "loss": 2.8375, "step": 49858 }, { "epoch": 3.91, "learning_rate": 6.090410099584412e-05, "loss": 2.5316, "step": 49859 }, { "epoch": 3.91, "learning_rate": 6.090331686661962e-05, "loss": 2.8947, "step": 49860 }, { "epoch": 3.91, "learning_rate": 6.0902532737395126e-05, "loss": 2.8207, "step": 49861 }, { "epoch": 3.91, "learning_rate": 6.090174860817063e-05, "loss": 3.1832, "step": 49862 }, { "epoch": 3.91, "learning_rate": 6.0900964478946134e-05, "loss": 2.8601, "step": 49863 }, { "epoch": 3.91, "learning_rate": 6.090018034972164e-05, "loss": 2.8322, "step": 49864 }, { "epoch": 3.91, "learning_rate": 6.089939622049714e-05, "loss": 2.6806, "step": 49865 }, { "epoch": 3.91, "learning_rate": 6.0898612091272645e-05, "loss": 2.8255, "step": 49866 }, { "epoch": 3.91, "learning_rate": 6.089782796204815e-05, "loss": 2.6837, "step": 49867 }, { "epoch": 3.91, "learning_rate": 6.089704383282365e-05, "loss": 2.6428, "step": 49868 }, { "epoch": 3.91, "learning_rate": 6.0896259703599156e-05, "loss": 2.7407, "step": 49869 }, { "epoch": 3.91, "learning_rate": 6.089547557437466e-05, "loss": 2.8105, "step": 49870 }, { "epoch": 3.91, "learning_rate": 6.089469144515016e-05, "loss": 2.7452, "step": 49871 }, { "epoch": 3.91, "learning_rate": 6.089390731592567e-05, "loss": 3.0101, "step": 49872 }, { "epoch": 3.91, "learning_rate": 6.089312318670117e-05, "loss": 2.6326, "step": 49873 }, { "epoch": 3.91, "learning_rate": 6.0892339057476674e-05, "loss": 2.7809, "step": 49874 }, { "epoch": 3.91, "learning_rate": 6.0891554928252184e-05, "loss": 2.6886, "step": 49875 }, { "epoch": 3.91, "learning_rate": 6.089077079902769e-05, "loss": 2.7965, "step": 49876 }, { "epoch": 3.91, "learning_rate": 6.088998666980319e-05, "loss": 2.6404, "step": 49877 }, { "epoch": 3.91, "learning_rate": 6.0889202540578695e-05, "loss": 2.7106, "step": 49878 }, { "epoch": 3.91, "learning_rate": 6.08884184113542e-05, "loss": 2.6435, "step": 49879 }, { "epoch": 3.91, "learning_rate": 6.08876342821297e-05, "loss": 2.709, "step": 49880 }, { "epoch": 3.91, "learning_rate": 6.0886850152905207e-05, "loss": 2.7398, "step": 49881 }, { "epoch": 3.91, "learning_rate": 6.088606602368071e-05, "loss": 2.8363, "step": 49882 }, { "epoch": 3.91, "learning_rate": 6.0885281894456214e-05, "loss": 2.685, "step": 49883 }, { "epoch": 3.91, "learning_rate": 6.088449776523172e-05, "loss": 2.8697, "step": 49884 }, { "epoch": 3.91, "learning_rate": 6.088371363600722e-05, "loss": 2.9552, "step": 49885 }, { "epoch": 3.91, "learning_rate": 6.088292950678272e-05, "loss": 2.7906, "step": 49886 }, { "epoch": 3.91, "learning_rate": 6.088214537755822e-05, "loss": 2.8595, "step": 49887 }, { "epoch": 3.91, "learning_rate": 6.0881361248333725e-05, "loss": 2.7705, "step": 49888 }, { "epoch": 3.91, "learning_rate": 6.088057711910923e-05, "loss": 2.8524, "step": 49889 }, { "epoch": 3.91, "learning_rate": 6.087979298988473e-05, "loss": 2.6599, "step": 49890 }, { "epoch": 3.91, "learning_rate": 6.0879008860660237e-05, "loss": 2.8062, "step": 49891 }, { "epoch": 3.91, "learning_rate": 6.087822473143574e-05, "loss": 2.9272, "step": 49892 }, { "epoch": 3.91, "learning_rate": 6.0877440602211244e-05, "loss": 2.7335, "step": 49893 }, { "epoch": 3.91, "learning_rate": 6.087665647298675e-05, "loss": 2.6984, "step": 49894 }, { "epoch": 3.91, "learning_rate": 6.087587234376225e-05, "loss": 2.7613, "step": 49895 }, { "epoch": 3.91, "learning_rate": 6.0875088214537755e-05, "loss": 2.6937, "step": 49896 }, { "epoch": 3.91, "learning_rate": 6.087430408531326e-05, "loss": 2.5762, "step": 49897 }, { "epoch": 3.91, "learning_rate": 6.087351995608876e-05, "loss": 2.7381, "step": 49898 }, { "epoch": 3.91, "learning_rate": 6.0872735826864266e-05, "loss": 3.0246, "step": 49899 }, { "epoch": 3.91, "learning_rate": 6.0871951697639776e-05, "loss": 2.7301, "step": 49900 }, { "epoch": 3.91, "learning_rate": 6.087116756841528e-05, "loss": 2.6758, "step": 49901 }, { "epoch": 3.91, "learning_rate": 6.0870383439190784e-05, "loss": 2.7206, "step": 49902 }, { "epoch": 3.91, "learning_rate": 6.086959930996629e-05, "loss": 2.7771, "step": 49903 }, { "epoch": 3.91, "learning_rate": 6.086881518074179e-05, "loss": 2.4602, "step": 49904 }, { "epoch": 3.91, "learning_rate": 6.0868031051517295e-05, "loss": 2.7018, "step": 49905 }, { "epoch": 3.91, "learning_rate": 6.08672469222928e-05, "loss": 2.6571, "step": 49906 }, { "epoch": 3.91, "learning_rate": 6.08664627930683e-05, "loss": 2.8854, "step": 49907 }, { "epoch": 3.91, "learning_rate": 6.0865678663843806e-05, "loss": 2.6519, "step": 49908 }, { "epoch": 3.91, "learning_rate": 6.086489453461931e-05, "loss": 2.5819, "step": 49909 }, { "epoch": 3.91, "learning_rate": 6.086411040539481e-05, "loss": 2.8118, "step": 49910 }, { "epoch": 3.91, "learning_rate": 6.086332627617032e-05, "loss": 2.8814, "step": 49911 }, { "epoch": 3.91, "learning_rate": 6.086254214694582e-05, "loss": 2.6686, "step": 49912 }, { "epoch": 3.91, "learning_rate": 6.0861758017721324e-05, "loss": 2.6067, "step": 49913 }, { "epoch": 3.91, "learning_rate": 6.086097388849683e-05, "loss": 2.8999, "step": 49914 }, { "epoch": 3.91, "learning_rate": 6.086018975927233e-05, "loss": 2.7165, "step": 49915 }, { "epoch": 3.91, "learning_rate": 6.085940563004784e-05, "loss": 2.6676, "step": 49916 }, { "epoch": 3.91, "learning_rate": 6.0858621500823346e-05, "loss": 2.7484, "step": 49917 }, { "epoch": 3.91, "learning_rate": 6.085783737159885e-05, "loss": 2.8176, "step": 49918 }, { "epoch": 3.91, "learning_rate": 6.085705324237435e-05, "loss": 3.0705, "step": 49919 }, { "epoch": 3.91, "learning_rate": 6.085626911314986e-05, "loss": 2.6103, "step": 49920 }, { "epoch": 3.91, "learning_rate": 6.085548498392536e-05, "loss": 2.6751, "step": 49921 }, { "epoch": 3.91, "learning_rate": 6.085470085470085e-05, "loss": 2.6277, "step": 49922 }, { "epoch": 3.91, "learning_rate": 6.0853916725476354e-05, "loss": 2.6407, "step": 49923 }, { "epoch": 3.91, "learning_rate": 6.085313259625186e-05, "loss": 2.6374, "step": 49924 }, { "epoch": 3.91, "learning_rate": 6.085234846702736e-05, "loss": 2.8094, "step": 49925 }, { "epoch": 3.91, "learning_rate": 6.085156433780287e-05, "loss": 2.5379, "step": 49926 }, { "epoch": 3.91, "learning_rate": 6.0850780208578376e-05, "loss": 2.5896, "step": 49927 }, { "epoch": 3.92, "learning_rate": 6.084999607935388e-05, "loss": 2.8521, "step": 49928 }, { "epoch": 3.92, "learning_rate": 6.084921195012938e-05, "loss": 2.6805, "step": 49929 }, { "epoch": 3.92, "learning_rate": 6.084842782090489e-05, "loss": 2.7625, "step": 49930 }, { "epoch": 3.92, "learning_rate": 6.084764369168039e-05, "loss": 2.901, "step": 49931 }, { "epoch": 3.92, "learning_rate": 6.0846859562455894e-05, "loss": 2.5844, "step": 49932 }, { "epoch": 3.92, "learning_rate": 6.08460754332314e-05, "loss": 2.8437, "step": 49933 }, { "epoch": 3.92, "learning_rate": 6.08452913040069e-05, "loss": 2.8226, "step": 49934 }, { "epoch": 3.92, "learning_rate": 6.0844507174782405e-05, "loss": 2.6749, "step": 49935 }, { "epoch": 3.92, "learning_rate": 6.084372304555791e-05, "loss": 2.6866, "step": 49936 }, { "epoch": 3.92, "learning_rate": 6.084293891633341e-05, "loss": 2.7129, "step": 49937 }, { "epoch": 3.92, "learning_rate": 6.0842154787108916e-05, "loss": 2.5633, "step": 49938 }, { "epoch": 3.92, "learning_rate": 6.084137065788442e-05, "loss": 2.6933, "step": 49939 }, { "epoch": 3.92, "learning_rate": 6.0840586528659924e-05, "loss": 2.8408, "step": 49940 }, { "epoch": 3.92, "learning_rate": 6.083980239943543e-05, "loss": 2.5631, "step": 49941 }, { "epoch": 3.92, "learning_rate": 6.083901827021094e-05, "loss": 2.7751, "step": 49942 }, { "epoch": 3.92, "learning_rate": 6.083823414098644e-05, "loss": 2.9652, "step": 49943 }, { "epoch": 3.92, "learning_rate": 6.0837450011761945e-05, "loss": 2.6583, "step": 49944 }, { "epoch": 3.92, "learning_rate": 6.083666588253745e-05, "loss": 2.6547, "step": 49945 }, { "epoch": 3.92, "learning_rate": 6.083588175331295e-05, "loss": 2.7982, "step": 49946 }, { "epoch": 3.92, "learning_rate": 6.0835097624088456e-05, "loss": 2.6938, "step": 49947 }, { "epoch": 3.92, "learning_rate": 6.083431349486396e-05, "loss": 2.867, "step": 49948 }, { "epoch": 3.92, "learning_rate": 6.083352936563946e-05, "loss": 2.6442, "step": 49949 }, { "epoch": 3.92, "learning_rate": 6.083274523641497e-05, "loss": 2.8816, "step": 49950 }, { "epoch": 3.92, "learning_rate": 6.083196110719047e-05, "loss": 2.7622, "step": 49951 }, { "epoch": 3.92, "learning_rate": 6.0831176977965974e-05, "loss": 2.7014, "step": 49952 }, { "epoch": 3.92, "learning_rate": 6.083039284874148e-05, "loss": 2.7495, "step": 49953 }, { "epoch": 3.92, "learning_rate": 6.082960871951698e-05, "loss": 2.7544, "step": 49954 }, { "epoch": 3.92, "learning_rate": 6.0828824590292485e-05, "loss": 2.7611, "step": 49955 }, { "epoch": 3.92, "learning_rate": 6.082804046106799e-05, "loss": 2.7643, "step": 49956 }, { "epoch": 3.92, "learning_rate": 6.08272563318435e-05, "loss": 2.5258, "step": 49957 }, { "epoch": 3.92, "learning_rate": 6.082647220261899e-05, "loss": 2.5933, "step": 49958 }, { "epoch": 3.92, "learning_rate": 6.082568807339449e-05, "loss": 2.5132, "step": 49959 }, { "epoch": 3.92, "learning_rate": 6.082490394417e-05, "loss": 2.8204, "step": 49960 }, { "epoch": 3.92, "learning_rate": 6.08241198149455e-05, "loss": 2.8317, "step": 49961 }, { "epoch": 3.92, "learning_rate": 6.0823335685721004e-05, "loss": 2.8303, "step": 49962 }, { "epoch": 3.92, "learning_rate": 6.082255155649651e-05, "loss": 2.6744, "step": 49963 }, { "epoch": 3.92, "learning_rate": 6.082176742727201e-05, "loss": 2.5958, "step": 49964 }, { "epoch": 3.92, "learning_rate": 6.0820983298047515e-05, "loss": 2.5127, "step": 49965 }, { "epoch": 3.92, "learning_rate": 6.082019916882302e-05, "loss": 2.5688, "step": 49966 }, { "epoch": 3.92, "learning_rate": 6.081941503959853e-05, "loss": 2.5907, "step": 49967 }, { "epoch": 3.92, "learning_rate": 6.081863091037403e-05, "loss": 2.6615, "step": 49968 }, { "epoch": 3.92, "learning_rate": 6.081784678114954e-05, "loss": 2.6581, "step": 49969 }, { "epoch": 3.92, "learning_rate": 6.081706265192504e-05, "loss": 2.721, "step": 49970 }, { "epoch": 3.92, "learning_rate": 6.0816278522700544e-05, "loss": 2.8321, "step": 49971 }, { "epoch": 3.92, "learning_rate": 6.081549439347605e-05, "loss": 2.9795, "step": 49972 }, { "epoch": 3.92, "learning_rate": 6.081471026425155e-05, "loss": 2.8744, "step": 49973 }, { "epoch": 3.92, "learning_rate": 6.0813926135027055e-05, "loss": 2.7091, "step": 49974 }, { "epoch": 3.92, "learning_rate": 6.081314200580256e-05, "loss": 2.7142, "step": 49975 }, { "epoch": 3.92, "learning_rate": 6.081235787657806e-05, "loss": 2.89, "step": 49976 }, { "epoch": 3.92, "learning_rate": 6.0811573747353566e-05, "loss": 2.8346, "step": 49977 }, { "epoch": 3.92, "learning_rate": 6.081078961812907e-05, "loss": 2.9426, "step": 49978 }, { "epoch": 3.92, "learning_rate": 6.0810005488904574e-05, "loss": 2.8044, "step": 49979 }, { "epoch": 3.92, "learning_rate": 6.080922135968008e-05, "loss": 2.749, "step": 49980 }, { "epoch": 3.92, "learning_rate": 6.080843723045558e-05, "loss": 2.8108, "step": 49981 }, { "epoch": 3.92, "learning_rate": 6.0807653101231085e-05, "loss": 2.5303, "step": 49982 }, { "epoch": 3.92, "learning_rate": 6.0806868972006595e-05, "loss": 2.7517, "step": 49983 }, { "epoch": 3.92, "learning_rate": 6.08060848427821e-05, "loss": 2.5076, "step": 49984 }, { "epoch": 3.92, "learning_rate": 6.08053007135576e-05, "loss": 2.7956, "step": 49985 }, { "epoch": 3.92, "learning_rate": 6.0804516584333106e-05, "loss": 3.0054, "step": 49986 }, { "epoch": 3.92, "learning_rate": 6.080373245510861e-05, "loss": 2.5963, "step": 49987 }, { "epoch": 3.92, "learning_rate": 6.0802948325884114e-05, "loss": 2.5816, "step": 49988 }, { "epoch": 3.92, "learning_rate": 6.080216419665962e-05, "loss": 2.7255, "step": 49989 }, { "epoch": 3.92, "learning_rate": 6.080138006743512e-05, "loss": 3.0292, "step": 49990 }, { "epoch": 3.92, "learning_rate": 6.0800595938210625e-05, "loss": 2.6953, "step": 49991 }, { "epoch": 3.92, "learning_rate": 6.079981180898613e-05, "loss": 2.6812, "step": 49992 }, { "epoch": 3.92, "learning_rate": 6.079902767976163e-05, "loss": 2.7009, "step": 49993 }, { "epoch": 3.92, "learning_rate": 6.079824355053713e-05, "loss": 3.0815, "step": 49994 }, { "epoch": 3.92, "learning_rate": 6.079745942131263e-05, "loss": 2.6398, "step": 49995 }, { "epoch": 3.92, "learning_rate": 6.0796675292088136e-05, "loss": 2.7873, "step": 49996 }, { "epoch": 3.92, "learning_rate": 6.079589116286364e-05, "loss": 2.507, "step": 49997 }, { "epoch": 3.92, "learning_rate": 6.0795107033639144e-05, "loss": 2.6646, "step": 49998 }, { "epoch": 3.92, "learning_rate": 6.079432290441465e-05, "loss": 2.6889, "step": 49999 }, { "epoch": 3.92, "learning_rate": 6.079353877519015e-05, "loss": 2.6515, "step": 50000 }, { "epoch": 3.92, "eval_loss": 2.889965534210205, "eval_runtime": 217.9068, "eval_samples_per_second": 52.004, "eval_steps_per_second": 3.254, "step": 50000 }, { "epoch": 3.92, "learning_rate": 6.0792754645965655e-05, "loss": 2.8593, "step": 50001 }, { "epoch": 3.92, "learning_rate": 6.079197051674116e-05, "loss": 2.7483, "step": 50002 }, { "epoch": 3.92, "learning_rate": 6.079118638751666e-05, "loss": 2.7312, "step": 50003 }, { "epoch": 3.92, "learning_rate": 6.0790402258292166e-05, "loss": 2.8438, "step": 50004 }, { "epoch": 3.92, "learning_rate": 6.078961812906767e-05, "loss": 2.7224, "step": 50005 }, { "epoch": 3.92, "learning_rate": 6.078883399984317e-05, "loss": 2.7318, "step": 50006 }, { "epoch": 3.92, "learning_rate": 6.078804987061868e-05, "loss": 2.6358, "step": 50007 }, { "epoch": 3.92, "learning_rate": 6.078726574139418e-05, "loss": 2.7992, "step": 50008 }, { "epoch": 3.92, "learning_rate": 6.078648161216969e-05, "loss": 2.5939, "step": 50009 }, { "epoch": 3.92, "learning_rate": 6.0785697482945194e-05, "loss": 2.618, "step": 50010 }, { "epoch": 3.92, "learning_rate": 6.07849133537207e-05, "loss": 2.9211, "step": 50011 }, { "epoch": 3.92, "learning_rate": 6.07841292244962e-05, "loss": 2.5749, "step": 50012 }, { "epoch": 3.92, "learning_rate": 6.0783345095271706e-05, "loss": 2.6659, "step": 50013 }, { "epoch": 3.92, "learning_rate": 6.078256096604721e-05, "loss": 2.5429, "step": 50014 }, { "epoch": 3.92, "learning_rate": 6.078177683682271e-05, "loss": 2.7504, "step": 50015 }, { "epoch": 3.92, "learning_rate": 6.0780992707598217e-05, "loss": 2.7923, "step": 50016 }, { "epoch": 3.92, "learning_rate": 6.078020857837372e-05, "loss": 2.777, "step": 50017 }, { "epoch": 3.92, "learning_rate": 6.0779424449149224e-05, "loss": 2.693, "step": 50018 }, { "epoch": 3.92, "learning_rate": 6.077864031992473e-05, "loss": 2.6955, "step": 50019 }, { "epoch": 3.92, "learning_rate": 6.077785619070023e-05, "loss": 2.7406, "step": 50020 }, { "epoch": 3.92, "learning_rate": 6.0777072061475735e-05, "loss": 2.6232, "step": 50021 }, { "epoch": 3.92, "learning_rate": 6.077628793225124e-05, "loss": 2.608, "step": 50022 }, { "epoch": 3.92, "learning_rate": 6.077550380302674e-05, "loss": 2.6839, "step": 50023 }, { "epoch": 3.92, "learning_rate": 6.077471967380225e-05, "loss": 2.7304, "step": 50024 }, { "epoch": 3.92, "learning_rate": 6.0773935544577756e-05, "loss": 2.594, "step": 50025 }, { "epoch": 3.92, "learning_rate": 6.077315141535326e-05, "loss": 3.0031, "step": 50026 }, { "epoch": 3.92, "learning_rate": 6.0772367286128764e-05, "loss": 2.6427, "step": 50027 }, { "epoch": 3.92, "learning_rate": 6.077158315690427e-05, "loss": 2.7158, "step": 50028 }, { "epoch": 3.92, "learning_rate": 6.077079902767977e-05, "loss": 2.6712, "step": 50029 }, { "epoch": 3.92, "learning_rate": 6.077001489845526e-05, "loss": 2.6633, "step": 50030 }, { "epoch": 3.92, "learning_rate": 6.0769230769230765e-05, "loss": 2.9967, "step": 50031 }, { "epoch": 3.92, "learning_rate": 6.076844664000627e-05, "loss": 2.8632, "step": 50032 }, { "epoch": 3.92, "learning_rate": 6.076766251078177e-05, "loss": 2.6526, "step": 50033 }, { "epoch": 3.92, "learning_rate": 6.076687838155728e-05, "loss": 2.7206, "step": 50034 }, { "epoch": 3.92, "learning_rate": 6.0766094252332786e-05, "loss": 2.6698, "step": 50035 }, { "epoch": 3.92, "learning_rate": 6.076531012310829e-05, "loss": 2.7907, "step": 50036 }, { "epoch": 3.92, "learning_rate": 6.0764525993883794e-05, "loss": 2.7484, "step": 50037 }, { "epoch": 3.92, "learning_rate": 6.07637418646593e-05, "loss": 2.7214, "step": 50038 }, { "epoch": 3.92, "learning_rate": 6.07629577354348e-05, "loss": 2.6417, "step": 50039 }, { "epoch": 3.92, "learning_rate": 6.0762173606210305e-05, "loss": 2.6541, "step": 50040 }, { "epoch": 3.92, "learning_rate": 6.076138947698581e-05, "loss": 2.7472, "step": 50041 }, { "epoch": 3.92, "learning_rate": 6.076060534776131e-05, "loss": 2.7214, "step": 50042 }, { "epoch": 3.92, "learning_rate": 6.0759821218536816e-05, "loss": 2.8458, "step": 50043 }, { "epoch": 3.92, "learning_rate": 6.075903708931232e-05, "loss": 2.765, "step": 50044 }, { "epoch": 3.92, "learning_rate": 6.075825296008782e-05, "loss": 2.7336, "step": 50045 }, { "epoch": 3.92, "learning_rate": 6.075746883086333e-05, "loss": 2.865, "step": 50046 }, { "epoch": 3.92, "learning_rate": 6.075668470163883e-05, "loss": 2.6575, "step": 50047 }, { "epoch": 3.92, "learning_rate": 6.0755900572414334e-05, "loss": 2.7139, "step": 50048 }, { "epoch": 3.92, "learning_rate": 6.075511644318984e-05, "loss": 2.7301, "step": 50049 }, { "epoch": 3.92, "learning_rate": 6.075433231396535e-05, "loss": 2.5498, "step": 50050 }, { "epoch": 3.92, "learning_rate": 6.075354818474085e-05, "loss": 2.6991, "step": 50051 }, { "epoch": 3.92, "learning_rate": 6.0752764055516356e-05, "loss": 2.6428, "step": 50052 }, { "epoch": 3.92, "learning_rate": 6.075197992629186e-05, "loss": 2.6651, "step": 50053 }, { "epoch": 3.92, "learning_rate": 6.075119579706736e-05, "loss": 2.6727, "step": 50054 }, { "epoch": 3.92, "learning_rate": 6.075041166784287e-05, "loss": 2.7066, "step": 50055 }, { "epoch": 3.93, "learning_rate": 6.074962753861837e-05, "loss": 2.5571, "step": 50056 }, { "epoch": 3.93, "learning_rate": 6.0748843409393874e-05, "loss": 2.8551, "step": 50057 }, { "epoch": 3.93, "learning_rate": 6.074805928016938e-05, "loss": 2.6602, "step": 50058 }, { "epoch": 3.93, "learning_rate": 6.074727515094488e-05, "loss": 2.7118, "step": 50059 }, { "epoch": 3.93, "learning_rate": 6.0746491021720385e-05, "loss": 2.729, "step": 50060 }, { "epoch": 3.93, "learning_rate": 6.074570689249589e-05, "loss": 2.6963, "step": 50061 }, { "epoch": 3.93, "learning_rate": 6.074492276327139e-05, "loss": 2.6508, "step": 50062 }, { "epoch": 3.93, "learning_rate": 6.0744138634046896e-05, "loss": 2.7291, "step": 50063 }, { "epoch": 3.93, "learning_rate": 6.07433545048224e-05, "loss": 2.9236, "step": 50064 }, { "epoch": 3.93, "learning_rate": 6.074257037559791e-05, "loss": 2.6957, "step": 50065 }, { "epoch": 3.93, "learning_rate": 6.07417862463734e-05, "loss": 2.6433, "step": 50066 }, { "epoch": 3.93, "learning_rate": 6.0741002117148904e-05, "loss": 3.0536, "step": 50067 }, { "epoch": 3.93, "learning_rate": 6.074021798792441e-05, "loss": 2.9, "step": 50068 }, { "epoch": 3.93, "learning_rate": 6.073943385869991e-05, "loss": 2.8046, "step": 50069 }, { "epoch": 3.93, "learning_rate": 6.0738649729475415e-05, "loss": 2.5937, "step": 50070 }, { "epoch": 3.93, "learning_rate": 6.073786560025092e-05, "loss": 2.8098, "step": 50071 }, { "epoch": 3.93, "learning_rate": 6.073708147102642e-05, "loss": 2.7699, "step": 50072 }, { "epoch": 3.93, "learning_rate": 6.0736297341801926e-05, "loss": 2.5917, "step": 50073 }, { "epoch": 3.93, "learning_rate": 6.073551321257743e-05, "loss": 2.9649, "step": 50074 }, { "epoch": 3.93, "learning_rate": 6.073472908335294e-05, "loss": 2.7386, "step": 50075 }, { "epoch": 3.93, "learning_rate": 6.0733944954128444e-05, "loss": 2.5903, "step": 50076 }, { "epoch": 3.93, "learning_rate": 6.073316082490395e-05, "loss": 2.6877, "step": 50077 }, { "epoch": 3.93, "learning_rate": 6.073237669567945e-05, "loss": 2.8587, "step": 50078 }, { "epoch": 3.93, "learning_rate": 6.0731592566454955e-05, "loss": 2.8472, "step": 50079 }, { "epoch": 3.93, "learning_rate": 6.073080843723046e-05, "loss": 2.8404, "step": 50080 }, { "epoch": 3.93, "learning_rate": 6.073002430800596e-05, "loss": 2.7836, "step": 50081 }, { "epoch": 3.93, "learning_rate": 6.0729240178781466e-05, "loss": 2.5829, "step": 50082 }, { "epoch": 3.93, "learning_rate": 6.072845604955697e-05, "loss": 2.585, "step": 50083 }, { "epoch": 3.93, "learning_rate": 6.0727671920332473e-05, "loss": 2.8586, "step": 50084 }, { "epoch": 3.93, "learning_rate": 6.072688779110798e-05, "loss": 2.8849, "step": 50085 }, { "epoch": 3.93, "learning_rate": 6.072610366188348e-05, "loss": 2.8015, "step": 50086 }, { "epoch": 3.93, "learning_rate": 6.0725319532658984e-05, "loss": 2.5886, "step": 50087 }, { "epoch": 3.93, "learning_rate": 6.072453540343449e-05, "loss": 2.7875, "step": 50088 }, { "epoch": 3.93, "learning_rate": 6.072375127420999e-05, "loss": 2.821, "step": 50089 }, { "epoch": 3.93, "learning_rate": 6.0722967144985495e-05, "loss": 2.7567, "step": 50090 }, { "epoch": 3.93, "learning_rate": 6.0722183015761006e-05, "loss": 2.7202, "step": 50091 }, { "epoch": 3.93, "learning_rate": 6.072139888653651e-05, "loss": 2.8237, "step": 50092 }, { "epoch": 3.93, "learning_rate": 6.072061475731201e-05, "loss": 2.7339, "step": 50093 }, { "epoch": 3.93, "learning_rate": 6.071983062808752e-05, "loss": 2.6626, "step": 50094 }, { "epoch": 3.93, "learning_rate": 6.071904649886302e-05, "loss": 2.7226, "step": 50095 }, { "epoch": 3.93, "learning_rate": 6.0718262369638524e-05, "loss": 2.5875, "step": 50096 }, { "epoch": 3.93, "learning_rate": 6.071747824041403e-05, "loss": 2.5185, "step": 50097 }, { "epoch": 3.93, "learning_rate": 6.071669411118953e-05, "loss": 2.8774, "step": 50098 }, { "epoch": 3.93, "learning_rate": 6.0715909981965035e-05, "loss": 2.6322, "step": 50099 }, { "epoch": 3.93, "learning_rate": 6.071512585274054e-05, "loss": 2.5898, "step": 50100 }, { "epoch": 3.93, "learning_rate": 6.071434172351604e-05, "loss": 2.6194, "step": 50101 }, { "epoch": 3.93, "learning_rate": 6.071355759429154e-05, "loss": 2.7133, "step": 50102 }, { "epoch": 3.93, "learning_rate": 6.071277346506704e-05, "loss": 2.5811, "step": 50103 }, { "epoch": 3.93, "learning_rate": 6.071198933584255e-05, "loss": 2.855, "step": 50104 }, { "epoch": 3.93, "learning_rate": 6.071120520661805e-05, "loss": 2.7067, "step": 50105 }, { "epoch": 3.93, "learning_rate": 6.0710421077393554e-05, "loss": 2.7392, "step": 50106 }, { "epoch": 3.93, "learning_rate": 6.070963694816906e-05, "loss": 2.6962, "step": 50107 }, { "epoch": 3.93, "learning_rate": 6.070885281894456e-05, "loss": 2.7176, "step": 50108 }, { "epoch": 3.93, "learning_rate": 6.0708068689720065e-05, "loss": 2.9424, "step": 50109 }, { "epoch": 3.93, "learning_rate": 6.070728456049557e-05, "loss": 2.8506, "step": 50110 }, { "epoch": 3.93, "learning_rate": 6.070650043127107e-05, "loss": 2.7121, "step": 50111 }, { "epoch": 3.93, "learning_rate": 6.0705716302046576e-05, "loss": 2.6681, "step": 50112 }, { "epoch": 3.93, "learning_rate": 6.070493217282208e-05, "loss": 2.5933, "step": 50113 }, { "epoch": 3.93, "learning_rate": 6.0704148043597584e-05, "loss": 2.9177, "step": 50114 }, { "epoch": 3.93, "learning_rate": 6.070336391437309e-05, "loss": 2.5795, "step": 50115 }, { "epoch": 3.93, "learning_rate": 6.070257978514859e-05, "loss": 2.7082, "step": 50116 }, { "epoch": 3.93, "learning_rate": 6.07017956559241e-05, "loss": 2.7088, "step": 50117 }, { "epoch": 3.93, "learning_rate": 6.0701011526699605e-05, "loss": 2.6243, "step": 50118 }, { "epoch": 3.93, "learning_rate": 6.070022739747511e-05, "loss": 2.663, "step": 50119 }, { "epoch": 3.93, "learning_rate": 6.069944326825061e-05, "loss": 2.654, "step": 50120 }, { "epoch": 3.93, "learning_rate": 6.0698659139026116e-05, "loss": 2.6716, "step": 50121 }, { "epoch": 3.93, "learning_rate": 6.069787500980162e-05, "loss": 2.8053, "step": 50122 }, { "epoch": 3.93, "learning_rate": 6.0697090880577124e-05, "loss": 2.6223, "step": 50123 }, { "epoch": 3.93, "learning_rate": 6.069630675135263e-05, "loss": 2.6966, "step": 50124 }, { "epoch": 3.93, "learning_rate": 6.069552262212813e-05, "loss": 2.3799, "step": 50125 }, { "epoch": 3.93, "learning_rate": 6.0694738492903635e-05, "loss": 2.7036, "step": 50126 }, { "epoch": 3.93, "learning_rate": 6.069395436367914e-05, "loss": 2.8108, "step": 50127 }, { "epoch": 3.93, "learning_rate": 6.069317023445464e-05, "loss": 2.7378, "step": 50128 }, { "epoch": 3.93, "learning_rate": 6.0692386105230146e-05, "loss": 2.9539, "step": 50129 }, { "epoch": 3.93, "learning_rate": 6.069160197600565e-05, "loss": 2.581, "step": 50130 }, { "epoch": 3.93, "learning_rate": 6.069081784678115e-05, "loss": 2.7465, "step": 50131 }, { "epoch": 3.93, "learning_rate": 6.0690033717556664e-05, "loss": 2.5907, "step": 50132 }, { "epoch": 3.93, "learning_rate": 6.068924958833217e-05, "loss": 2.7139, "step": 50133 }, { "epoch": 3.93, "learning_rate": 6.068846545910767e-05, "loss": 2.6705, "step": 50134 }, { "epoch": 3.93, "learning_rate": 6.0687681329883175e-05, "loss": 2.7932, "step": 50135 }, { "epoch": 3.93, "learning_rate": 6.068689720065868e-05, "loss": 2.7628, "step": 50136 }, { "epoch": 3.93, "learning_rate": 6.068611307143418e-05, "loss": 2.7651, "step": 50137 }, { "epoch": 3.93, "learning_rate": 6.068532894220967e-05, "loss": 2.7418, "step": 50138 }, { "epoch": 3.93, "learning_rate": 6.0684544812985176e-05, "loss": 2.6683, "step": 50139 }, { "epoch": 3.93, "learning_rate": 6.068376068376068e-05, "loss": 2.6651, "step": 50140 }, { "epoch": 3.93, "learning_rate": 6.068297655453618e-05, "loss": 2.7186, "step": 50141 }, { "epoch": 3.93, "learning_rate": 6.0682192425311694e-05, "loss": 2.8284, "step": 50142 }, { "epoch": 3.93, "learning_rate": 6.06814082960872e-05, "loss": 2.7395, "step": 50143 }, { "epoch": 3.93, "learning_rate": 6.06806241668627e-05, "loss": 2.6854, "step": 50144 }, { "epoch": 3.93, "learning_rate": 6.0679840037638205e-05, "loss": 2.6916, "step": 50145 }, { "epoch": 3.93, "learning_rate": 6.067905590841371e-05, "loss": 2.7545, "step": 50146 }, { "epoch": 3.93, "learning_rate": 6.067827177918921e-05, "loss": 2.6789, "step": 50147 }, { "epoch": 3.93, "learning_rate": 6.0677487649964716e-05, "loss": 2.6302, "step": 50148 }, { "epoch": 3.93, "learning_rate": 6.067670352074022e-05, "loss": 2.563, "step": 50149 }, { "epoch": 3.93, "learning_rate": 6.067591939151572e-05, "loss": 2.7589, "step": 50150 }, { "epoch": 3.93, "learning_rate": 6.0675135262291227e-05, "loss": 2.747, "step": 50151 }, { "epoch": 3.93, "learning_rate": 6.067435113306673e-05, "loss": 2.9088, "step": 50152 }, { "epoch": 3.93, "learning_rate": 6.0673567003842234e-05, "loss": 2.9297, "step": 50153 }, { "epoch": 3.93, "learning_rate": 6.067278287461774e-05, "loss": 2.7163, "step": 50154 }, { "epoch": 3.93, "learning_rate": 6.067199874539324e-05, "loss": 2.8735, "step": 50155 }, { "epoch": 3.93, "learning_rate": 6.0671214616168745e-05, "loss": 2.5693, "step": 50156 }, { "epoch": 3.93, "learning_rate": 6.067043048694425e-05, "loss": 2.8903, "step": 50157 }, { "epoch": 3.93, "learning_rate": 6.066964635771976e-05, "loss": 2.7291, "step": 50158 }, { "epoch": 3.93, "learning_rate": 6.066886222849526e-05, "loss": 2.7366, "step": 50159 }, { "epoch": 3.93, "learning_rate": 6.0668078099270766e-05, "loss": 2.7653, "step": 50160 }, { "epoch": 3.93, "learning_rate": 6.066729397004627e-05, "loss": 2.8235, "step": 50161 }, { "epoch": 3.93, "learning_rate": 6.0666509840821774e-05, "loss": 2.9796, "step": 50162 }, { "epoch": 3.93, "learning_rate": 6.066572571159728e-05, "loss": 2.8151, "step": 50163 }, { "epoch": 3.93, "learning_rate": 6.066494158237278e-05, "loss": 2.6525, "step": 50164 }, { "epoch": 3.93, "learning_rate": 6.0664157453148285e-05, "loss": 2.8089, "step": 50165 }, { "epoch": 3.93, "learning_rate": 6.066337332392379e-05, "loss": 2.8195, "step": 50166 }, { "epoch": 3.93, "learning_rate": 6.066258919469929e-05, "loss": 2.7916, "step": 50167 }, { "epoch": 3.93, "learning_rate": 6.0661805065474796e-05, "loss": 2.6898, "step": 50168 }, { "epoch": 3.93, "learning_rate": 6.06610209362503e-05, "loss": 2.7318, "step": 50169 }, { "epoch": 3.93, "learning_rate": 6.06602368070258e-05, "loss": 2.8779, "step": 50170 }, { "epoch": 3.93, "learning_rate": 6.065945267780131e-05, "loss": 2.6838, "step": 50171 }, { "epoch": 3.93, "learning_rate": 6.065866854857681e-05, "loss": 2.6453, "step": 50172 }, { "epoch": 3.93, "learning_rate": 6.065788441935232e-05, "loss": 2.7979, "step": 50173 }, { "epoch": 3.93, "learning_rate": 6.065710029012781e-05, "loss": 2.7139, "step": 50174 }, { "epoch": 3.93, "learning_rate": 6.0656316160903315e-05, "loss": 2.8907, "step": 50175 }, { "epoch": 3.93, "learning_rate": 6.065553203167882e-05, "loss": 2.7995, "step": 50176 }, { "epoch": 3.93, "learning_rate": 6.065474790245432e-05, "loss": 2.5282, "step": 50177 }, { "epoch": 3.93, "learning_rate": 6.0653963773229826e-05, "loss": 2.6975, "step": 50178 }, { "epoch": 3.93, "learning_rate": 6.065317964400533e-05, "loss": 2.5282, "step": 50179 }, { "epoch": 3.93, "learning_rate": 6.065239551478083e-05, "loss": 2.8003, "step": 50180 }, { "epoch": 3.93, "learning_rate": 6.065161138555634e-05, "loss": 2.6683, "step": 50181 }, { "epoch": 3.93, "learning_rate": 6.065082725633184e-05, "loss": 2.9156, "step": 50182 }, { "epoch": 3.93, "learning_rate": 6.065004312710735e-05, "loss": 2.9, "step": 50183 }, { "epoch": 3.94, "learning_rate": 6.0649258997882855e-05, "loss": 2.7034, "step": 50184 }, { "epoch": 3.94, "learning_rate": 6.064847486865836e-05, "loss": 2.7679, "step": 50185 }, { "epoch": 3.94, "learning_rate": 6.064769073943386e-05, "loss": 2.6896, "step": 50186 }, { "epoch": 3.94, "learning_rate": 6.0646906610209366e-05, "loss": 2.7265, "step": 50187 }, { "epoch": 3.94, "learning_rate": 6.064612248098487e-05, "loss": 2.683, "step": 50188 }, { "epoch": 3.94, "learning_rate": 6.064533835176037e-05, "loss": 2.7173, "step": 50189 }, { "epoch": 3.94, "learning_rate": 6.064455422253588e-05, "loss": 2.8523, "step": 50190 }, { "epoch": 3.94, "learning_rate": 6.064377009331138e-05, "loss": 2.7203, "step": 50191 }, { "epoch": 3.94, "learning_rate": 6.0642985964086884e-05, "loss": 2.6393, "step": 50192 }, { "epoch": 3.94, "learning_rate": 6.064220183486239e-05, "loss": 2.8767, "step": 50193 }, { "epoch": 3.94, "learning_rate": 6.064141770563789e-05, "loss": 2.6371, "step": 50194 }, { "epoch": 3.94, "learning_rate": 6.0640633576413395e-05, "loss": 2.7281, "step": 50195 }, { "epoch": 3.94, "learning_rate": 6.06398494471889e-05, "loss": 2.9655, "step": 50196 }, { "epoch": 3.94, "learning_rate": 6.06390653179644e-05, "loss": 2.6592, "step": 50197 }, { "epoch": 3.94, "learning_rate": 6.0638281188739906e-05, "loss": 2.8057, "step": 50198 }, { "epoch": 3.94, "learning_rate": 6.063749705951542e-05, "loss": 2.8291, "step": 50199 }, { "epoch": 3.94, "learning_rate": 6.063671293029092e-05, "loss": 2.703, "step": 50200 }, { "epoch": 3.94, "learning_rate": 6.0635928801066424e-05, "loss": 2.9234, "step": 50201 }, { "epoch": 3.94, "learning_rate": 6.063514467184193e-05, "loss": 2.6248, "step": 50202 }, { "epoch": 3.94, "learning_rate": 6.063436054261743e-05, "loss": 2.8078, "step": 50203 }, { "epoch": 3.94, "learning_rate": 6.0633576413392935e-05, "loss": 2.5754, "step": 50204 }, { "epoch": 3.94, "learning_rate": 6.063279228416844e-05, "loss": 2.6678, "step": 50205 }, { "epoch": 3.94, "learning_rate": 6.063200815494394e-05, "loss": 2.646, "step": 50206 }, { "epoch": 3.94, "learning_rate": 6.0631224025719446e-05, "loss": 2.7448, "step": 50207 }, { "epoch": 3.94, "learning_rate": 6.063043989649495e-05, "loss": 2.6312, "step": 50208 }, { "epoch": 3.94, "learning_rate": 6.0629655767270453e-05, "loss": 2.7123, "step": 50209 }, { "epoch": 3.94, "learning_rate": 6.062887163804595e-05, "loss": 2.7384, "step": 50210 }, { "epoch": 3.94, "learning_rate": 6.0628087508821454e-05, "loss": 2.5241, "step": 50211 }, { "epoch": 3.94, "learning_rate": 6.062730337959696e-05, "loss": 2.6756, "step": 50212 }, { "epoch": 3.94, "learning_rate": 6.062651925037246e-05, "loss": 2.702, "step": 50213 }, { "epoch": 3.94, "learning_rate": 6.0625735121147965e-05, "loss": 2.7139, "step": 50214 }, { "epoch": 3.94, "learning_rate": 6.062495099192347e-05, "loss": 2.6368, "step": 50215 }, { "epoch": 3.94, "learning_rate": 6.062416686269897e-05, "loss": 2.7129, "step": 50216 }, { "epoch": 3.94, "learning_rate": 6.0623382733474476e-05, "loss": 2.6926, "step": 50217 }, { "epoch": 3.94, "learning_rate": 6.062259860424998e-05, "loss": 2.8084, "step": 50218 }, { "epoch": 3.94, "learning_rate": 6.0621814475025483e-05, "loss": 2.6803, "step": 50219 }, { "epoch": 3.94, "learning_rate": 6.062103034580099e-05, "loss": 2.8647, "step": 50220 }, { "epoch": 3.94, "learning_rate": 6.062024621657649e-05, "loss": 2.5849, "step": 50221 }, { "epoch": 3.94, "learning_rate": 6.0619462087351994e-05, "loss": 2.6554, "step": 50222 }, { "epoch": 3.94, "learning_rate": 6.06186779581275e-05, "loss": 2.7048, "step": 50223 }, { "epoch": 3.94, "learning_rate": 6.0617893828903e-05, "loss": 2.7437, "step": 50224 }, { "epoch": 3.94, "learning_rate": 6.061710969967851e-05, "loss": 2.9716, "step": 50225 }, { "epoch": 3.94, "learning_rate": 6.0616325570454016e-05, "loss": 2.6216, "step": 50226 }, { "epoch": 3.94, "learning_rate": 6.061554144122952e-05, "loss": 2.6013, "step": 50227 }, { "epoch": 3.94, "learning_rate": 6.061475731200502e-05, "loss": 2.5641, "step": 50228 }, { "epoch": 3.94, "learning_rate": 6.061397318278053e-05, "loss": 2.7761, "step": 50229 }, { "epoch": 3.94, "learning_rate": 6.061318905355603e-05, "loss": 2.6472, "step": 50230 }, { "epoch": 3.94, "learning_rate": 6.0612404924331534e-05, "loss": 2.8268, "step": 50231 }, { "epoch": 3.94, "learning_rate": 6.061162079510704e-05, "loss": 2.6592, "step": 50232 }, { "epoch": 3.94, "learning_rate": 6.061083666588254e-05, "loss": 2.5233, "step": 50233 }, { "epoch": 3.94, "learning_rate": 6.0610052536658045e-05, "loss": 2.963, "step": 50234 }, { "epoch": 3.94, "learning_rate": 6.060926840743355e-05, "loss": 2.8167, "step": 50235 }, { "epoch": 3.94, "learning_rate": 6.060848427820905e-05, "loss": 2.6351, "step": 50236 }, { "epoch": 3.94, "learning_rate": 6.0607700148984556e-05, "loss": 2.8007, "step": 50237 }, { "epoch": 3.94, "learning_rate": 6.060691601976006e-05, "loss": 2.9115, "step": 50238 }, { "epoch": 3.94, "learning_rate": 6.0606131890535564e-05, "loss": 2.862, "step": 50239 }, { "epoch": 3.94, "learning_rate": 6.0605347761311074e-05, "loss": 2.8027, "step": 50240 }, { "epoch": 3.94, "learning_rate": 6.060456363208658e-05, "loss": 2.5684, "step": 50241 }, { "epoch": 3.94, "learning_rate": 6.060377950286208e-05, "loss": 2.9167, "step": 50242 }, { "epoch": 3.94, "learning_rate": 6.0602995373637585e-05, "loss": 2.6046, "step": 50243 }, { "epoch": 3.94, "learning_rate": 6.060221124441309e-05, "loss": 2.7457, "step": 50244 }, { "epoch": 3.94, "learning_rate": 6.060142711518859e-05, "loss": 2.5916, "step": 50245 }, { "epoch": 3.94, "learning_rate": 6.060064298596408e-05, "loss": 2.7995, "step": 50246 }, { "epoch": 3.94, "learning_rate": 6.0599858856739586e-05, "loss": 2.7214, "step": 50247 }, { "epoch": 3.94, "learning_rate": 6.059907472751509e-05, "loss": 2.6702, "step": 50248 }, { "epoch": 3.94, "learning_rate": 6.0598290598290594e-05, "loss": 2.7281, "step": 50249 }, { "epoch": 3.94, "learning_rate": 6.0597506469066104e-05, "loss": 2.7478, "step": 50250 }, { "epoch": 3.94, "learning_rate": 6.059672233984161e-05, "loss": 2.7867, "step": 50251 }, { "epoch": 3.94, "learning_rate": 6.059593821061711e-05, "loss": 2.706, "step": 50252 }, { "epoch": 3.94, "learning_rate": 6.0595154081392615e-05, "loss": 2.7988, "step": 50253 }, { "epoch": 3.94, "learning_rate": 6.059436995216812e-05, "loss": 2.8023, "step": 50254 }, { "epoch": 3.94, "learning_rate": 6.059358582294362e-05, "loss": 3.0196, "step": 50255 }, { "epoch": 3.94, "learning_rate": 6.0592801693719126e-05, "loss": 2.7519, "step": 50256 }, { "epoch": 3.94, "learning_rate": 6.059201756449463e-05, "loss": 2.7381, "step": 50257 }, { "epoch": 3.94, "learning_rate": 6.0591233435270134e-05, "loss": 2.6336, "step": 50258 }, { "epoch": 3.94, "learning_rate": 6.059044930604564e-05, "loss": 2.8348, "step": 50259 }, { "epoch": 3.94, "learning_rate": 6.058966517682114e-05, "loss": 2.8038, "step": 50260 }, { "epoch": 3.94, "learning_rate": 6.0588881047596645e-05, "loss": 2.795, "step": 50261 }, { "epoch": 3.94, "learning_rate": 6.058809691837215e-05, "loss": 2.6101, "step": 50262 }, { "epoch": 3.94, "learning_rate": 6.058731278914765e-05, "loss": 2.5662, "step": 50263 }, { "epoch": 3.94, "learning_rate": 6.0586528659923156e-05, "loss": 2.7696, "step": 50264 }, { "epoch": 3.94, "learning_rate": 6.058574453069866e-05, "loss": 2.6955, "step": 50265 }, { "epoch": 3.94, "learning_rate": 6.058496040147417e-05, "loss": 2.8055, "step": 50266 }, { "epoch": 3.94, "learning_rate": 6.0584176272249674e-05, "loss": 2.7635, "step": 50267 }, { "epoch": 3.94, "learning_rate": 6.058339214302518e-05, "loss": 2.8469, "step": 50268 }, { "epoch": 3.94, "learning_rate": 6.058260801380068e-05, "loss": 2.7379, "step": 50269 }, { "epoch": 3.94, "learning_rate": 6.0581823884576185e-05, "loss": 2.7468, "step": 50270 }, { "epoch": 3.94, "learning_rate": 6.058103975535169e-05, "loss": 2.6702, "step": 50271 }, { "epoch": 3.94, "learning_rate": 6.058025562612719e-05, "loss": 2.4912, "step": 50272 }, { "epoch": 3.94, "learning_rate": 6.0579471496902696e-05, "loss": 2.7054, "step": 50273 }, { "epoch": 3.94, "learning_rate": 6.05786873676782e-05, "loss": 2.5891, "step": 50274 }, { "epoch": 3.94, "learning_rate": 6.05779032384537e-05, "loss": 2.9082, "step": 50275 }, { "epoch": 3.94, "learning_rate": 6.057711910922921e-05, "loss": 2.8198, "step": 50276 }, { "epoch": 3.94, "learning_rate": 6.057633498000471e-05, "loss": 2.664, "step": 50277 }, { "epoch": 3.94, "learning_rate": 6.0575550850780214e-05, "loss": 2.5384, "step": 50278 }, { "epoch": 3.94, "learning_rate": 6.057476672155572e-05, "loss": 2.7654, "step": 50279 }, { "epoch": 3.94, "learning_rate": 6.057398259233122e-05, "loss": 2.67, "step": 50280 }, { "epoch": 3.94, "learning_rate": 6.0573198463106725e-05, "loss": 2.8842, "step": 50281 }, { "epoch": 3.94, "learning_rate": 6.057241433388222e-05, "loss": 2.736, "step": 50282 }, { "epoch": 3.94, "learning_rate": 6.0571630204657726e-05, "loss": 2.9188, "step": 50283 }, { "epoch": 3.94, "learning_rate": 6.057084607543323e-05, "loss": 2.6395, "step": 50284 }, { "epoch": 3.94, "learning_rate": 6.057006194620873e-05, "loss": 2.6513, "step": 50285 }, { "epoch": 3.94, "learning_rate": 6.056927781698424e-05, "loss": 2.5669, "step": 50286 }, { "epoch": 3.94, "learning_rate": 6.056849368775974e-05, "loss": 2.7705, "step": 50287 }, { "epoch": 3.94, "learning_rate": 6.0567709558535244e-05, "loss": 2.6706, "step": 50288 }, { "epoch": 3.94, "learning_rate": 6.056692542931075e-05, "loss": 2.92, "step": 50289 }, { "epoch": 3.94, "learning_rate": 6.056614130008625e-05, "loss": 2.6619, "step": 50290 }, { "epoch": 3.94, "learning_rate": 6.056535717086176e-05, "loss": 2.687, "step": 50291 }, { "epoch": 3.94, "learning_rate": 6.0564573041637265e-05, "loss": 2.9002, "step": 50292 }, { "epoch": 3.94, "learning_rate": 6.056378891241277e-05, "loss": 2.7302, "step": 50293 }, { "epoch": 3.94, "learning_rate": 6.056300478318827e-05, "loss": 3.0694, "step": 50294 }, { "epoch": 3.94, "learning_rate": 6.0562220653963777e-05, "loss": 2.736, "step": 50295 }, { "epoch": 3.94, "learning_rate": 6.056143652473928e-05, "loss": 2.5659, "step": 50296 }, { "epoch": 3.94, "learning_rate": 6.0560652395514784e-05, "loss": 2.6536, "step": 50297 }, { "epoch": 3.94, "learning_rate": 6.055986826629029e-05, "loss": 2.6097, "step": 50298 }, { "epoch": 3.94, "learning_rate": 6.055908413706579e-05, "loss": 2.9127, "step": 50299 }, { "epoch": 3.94, "learning_rate": 6.0558300007841295e-05, "loss": 2.5484, "step": 50300 }, { "epoch": 3.94, "learning_rate": 6.05575158786168e-05, "loss": 2.6782, "step": 50301 }, { "epoch": 3.94, "learning_rate": 6.05567317493923e-05, "loss": 2.7195, "step": 50302 }, { "epoch": 3.94, "learning_rate": 6.0555947620167806e-05, "loss": 2.8412, "step": 50303 }, { "epoch": 3.94, "learning_rate": 6.055516349094331e-05, "loss": 2.7681, "step": 50304 }, { "epoch": 3.94, "learning_rate": 6.055437936171881e-05, "loss": 2.6288, "step": 50305 }, { "epoch": 3.94, "learning_rate": 6.055359523249432e-05, "loss": 2.9136, "step": 50306 }, { "epoch": 3.94, "learning_rate": 6.055281110326983e-05, "loss": 2.6124, "step": 50307 }, { "epoch": 3.94, "learning_rate": 6.055202697404533e-05, "loss": 2.8702, "step": 50308 }, { "epoch": 3.94, "learning_rate": 6.0551242844820835e-05, "loss": 2.7652, "step": 50309 }, { "epoch": 3.94, "learning_rate": 6.055045871559634e-05, "loss": 2.7682, "step": 50310 }, { "epoch": 3.95, "learning_rate": 6.054967458637184e-05, "loss": 2.6568, "step": 50311 }, { "epoch": 3.95, "learning_rate": 6.0548890457147346e-05, "loss": 2.697, "step": 50312 }, { "epoch": 3.95, "learning_rate": 6.054810632792285e-05, "loss": 2.8409, "step": 50313 }, { "epoch": 3.95, "learning_rate": 6.054732219869835e-05, "loss": 2.5251, "step": 50314 }, { "epoch": 3.95, "learning_rate": 6.054653806947386e-05, "loss": 2.8305, "step": 50315 }, { "epoch": 3.95, "learning_rate": 6.054575394024936e-05, "loss": 2.82, "step": 50316 }, { "epoch": 3.95, "learning_rate": 6.0544969811024864e-05, "loss": 2.8113, "step": 50317 }, { "epoch": 3.95, "learning_rate": 6.054418568180036e-05, "loss": 2.8337, "step": 50318 }, { "epoch": 3.95, "learning_rate": 6.0543401552575865e-05, "loss": 2.8211, "step": 50319 }, { "epoch": 3.95, "learning_rate": 6.054261742335137e-05, "loss": 2.6079, "step": 50320 }, { "epoch": 3.95, "learning_rate": 6.054183329412687e-05, "loss": 2.9298, "step": 50321 }, { "epoch": 3.95, "learning_rate": 6.0541049164902376e-05, "loss": 2.8247, "step": 50322 }, { "epoch": 3.95, "learning_rate": 6.054026503567788e-05, "loss": 2.6935, "step": 50323 }, { "epoch": 3.95, "learning_rate": 6.053948090645338e-05, "loss": 2.6242, "step": 50324 }, { "epoch": 3.95, "learning_rate": 6.053869677722889e-05, "loss": 2.8357, "step": 50325 }, { "epoch": 3.95, "learning_rate": 6.053791264800439e-05, "loss": 2.5075, "step": 50326 }, { "epoch": 3.95, "learning_rate": 6.0537128518779894e-05, "loss": 3.1666, "step": 50327 }, { "epoch": 3.95, "learning_rate": 6.05363443895554e-05, "loss": 2.5077, "step": 50328 }, { "epoch": 3.95, "learning_rate": 6.05355602603309e-05, "loss": 2.6432, "step": 50329 }, { "epoch": 3.95, "learning_rate": 6.0534776131106405e-05, "loss": 2.7728, "step": 50330 }, { "epoch": 3.95, "learning_rate": 6.053399200188191e-05, "loss": 2.8349, "step": 50331 }, { "epoch": 3.95, "learning_rate": 6.053320787265741e-05, "loss": 2.6108, "step": 50332 }, { "epoch": 3.95, "learning_rate": 6.053242374343292e-05, "loss": 2.5889, "step": 50333 }, { "epoch": 3.95, "learning_rate": 6.053163961420843e-05, "loss": 2.6185, "step": 50334 }, { "epoch": 3.95, "learning_rate": 6.053085548498393e-05, "loss": 2.6744, "step": 50335 }, { "epoch": 3.95, "learning_rate": 6.0530071355759434e-05, "loss": 2.9544, "step": 50336 }, { "epoch": 3.95, "learning_rate": 6.052928722653494e-05, "loss": 2.7636, "step": 50337 }, { "epoch": 3.95, "learning_rate": 6.052850309731044e-05, "loss": 2.6881, "step": 50338 }, { "epoch": 3.95, "learning_rate": 6.0527718968085945e-05, "loss": 2.8734, "step": 50339 }, { "epoch": 3.95, "learning_rate": 6.052693483886145e-05, "loss": 2.7714, "step": 50340 }, { "epoch": 3.95, "learning_rate": 6.052615070963695e-05, "loss": 2.6648, "step": 50341 }, { "epoch": 3.95, "learning_rate": 6.0525366580412456e-05, "loss": 2.7981, "step": 50342 }, { "epoch": 3.95, "learning_rate": 6.052458245118796e-05, "loss": 2.7279, "step": 50343 }, { "epoch": 3.95, "learning_rate": 6.0523798321963464e-05, "loss": 2.6145, "step": 50344 }, { "epoch": 3.95, "learning_rate": 6.052301419273897e-05, "loss": 2.6463, "step": 50345 }, { "epoch": 3.95, "learning_rate": 6.052223006351447e-05, "loss": 2.7128, "step": 50346 }, { "epoch": 3.95, "learning_rate": 6.0521445934289975e-05, "loss": 2.4843, "step": 50347 }, { "epoch": 3.95, "learning_rate": 6.0520661805065485e-05, "loss": 2.9889, "step": 50348 }, { "epoch": 3.95, "learning_rate": 6.051987767584099e-05, "loss": 2.7231, "step": 50349 }, { "epoch": 3.95, "learning_rate": 6.051909354661649e-05, "loss": 2.6162, "step": 50350 }, { "epoch": 3.95, "learning_rate": 6.0518309417391996e-05, "loss": 2.5149, "step": 50351 }, { "epoch": 3.95, "learning_rate": 6.05175252881675e-05, "loss": 2.7995, "step": 50352 }, { "epoch": 3.95, "learning_rate": 6.0516741158943e-05, "loss": 2.6351, "step": 50353 }, { "epoch": 3.95, "learning_rate": 6.0515957029718494e-05, "loss": 2.6763, "step": 50354 }, { "epoch": 3.95, "learning_rate": 6.0515172900494e-05, "loss": 2.7286, "step": 50355 }, { "epoch": 3.95, "learning_rate": 6.05143887712695e-05, "loss": 2.6172, "step": 50356 }, { "epoch": 3.95, "learning_rate": 6.0513604642045005e-05, "loss": 2.5719, "step": 50357 }, { "epoch": 3.95, "learning_rate": 6.0512820512820515e-05, "loss": 2.6627, "step": 50358 }, { "epoch": 3.95, "learning_rate": 6.051203638359602e-05, "loss": 2.7226, "step": 50359 }, { "epoch": 3.95, "learning_rate": 6.051125225437152e-05, "loss": 2.6209, "step": 50360 }, { "epoch": 3.95, "learning_rate": 6.0510468125147026e-05, "loss": 2.8898, "step": 50361 }, { "epoch": 3.95, "learning_rate": 6.050968399592253e-05, "loss": 2.6643, "step": 50362 }, { "epoch": 3.95, "learning_rate": 6.050889986669803e-05, "loss": 2.5761, "step": 50363 }, { "epoch": 3.95, "learning_rate": 6.050811573747354e-05, "loss": 2.5369, "step": 50364 }, { "epoch": 3.95, "learning_rate": 6.050733160824904e-05, "loss": 2.8985, "step": 50365 }, { "epoch": 3.95, "learning_rate": 6.0506547479024544e-05, "loss": 2.5061, "step": 50366 }, { "epoch": 3.95, "learning_rate": 6.050576334980005e-05, "loss": 2.6649, "step": 50367 }, { "epoch": 3.95, "learning_rate": 6.050497922057555e-05, "loss": 2.8621, "step": 50368 }, { "epoch": 3.95, "learning_rate": 6.0504195091351055e-05, "loss": 2.6513, "step": 50369 }, { "epoch": 3.95, "learning_rate": 6.050341096212656e-05, "loss": 2.7002, "step": 50370 }, { "epoch": 3.95, "learning_rate": 6.050262683290206e-05, "loss": 2.6242, "step": 50371 }, { "epoch": 3.95, "learning_rate": 6.0501842703677566e-05, "loss": 2.8776, "step": 50372 }, { "epoch": 3.95, "learning_rate": 6.050105857445307e-05, "loss": 2.6269, "step": 50373 }, { "epoch": 3.95, "learning_rate": 6.050027444522858e-05, "loss": 2.798, "step": 50374 }, { "epoch": 3.95, "learning_rate": 6.0499490316004084e-05, "loss": 2.6374, "step": 50375 }, { "epoch": 3.95, "learning_rate": 6.049870618677959e-05, "loss": 2.5575, "step": 50376 }, { "epoch": 3.95, "learning_rate": 6.049792205755509e-05, "loss": 2.6034, "step": 50377 }, { "epoch": 3.95, "learning_rate": 6.0497137928330595e-05, "loss": 2.7233, "step": 50378 }, { "epoch": 3.95, "learning_rate": 6.04963537991061e-05, "loss": 2.7873, "step": 50379 }, { "epoch": 3.95, "learning_rate": 6.04955696698816e-05, "loss": 2.8253, "step": 50380 }, { "epoch": 3.95, "learning_rate": 6.0494785540657106e-05, "loss": 2.6738, "step": 50381 }, { "epoch": 3.95, "learning_rate": 6.049400141143261e-05, "loss": 2.8133, "step": 50382 }, { "epoch": 3.95, "learning_rate": 6.0493217282208114e-05, "loss": 2.6414, "step": 50383 }, { "epoch": 3.95, "learning_rate": 6.049243315298362e-05, "loss": 2.8622, "step": 50384 }, { "epoch": 3.95, "learning_rate": 6.049164902375912e-05, "loss": 2.7411, "step": 50385 }, { "epoch": 3.95, "learning_rate": 6.0490864894534625e-05, "loss": 2.7131, "step": 50386 }, { "epoch": 3.95, "learning_rate": 6.049008076531013e-05, "loss": 2.8382, "step": 50387 }, { "epoch": 3.95, "learning_rate": 6.048929663608563e-05, "loss": 2.7576, "step": 50388 }, { "epoch": 3.95, "learning_rate": 6.0488512506861136e-05, "loss": 2.6838, "step": 50389 }, { "epoch": 3.95, "learning_rate": 6.048772837763663e-05, "loss": 2.7182, "step": 50390 }, { "epoch": 3.95, "learning_rate": 6.0486944248412136e-05, "loss": 2.7614, "step": 50391 }, { "epoch": 3.95, "learning_rate": 6.048616011918764e-05, "loss": 2.5317, "step": 50392 }, { "epoch": 3.95, "learning_rate": 6.0485375989963144e-05, "loss": 2.7637, "step": 50393 }, { "epoch": 3.95, "learning_rate": 6.048459186073865e-05, "loss": 2.6834, "step": 50394 }, { "epoch": 3.95, "learning_rate": 6.048380773151415e-05, "loss": 2.8448, "step": 50395 }, { "epoch": 3.95, "learning_rate": 6.0483023602289655e-05, "loss": 2.8683, "step": 50396 }, { "epoch": 3.95, "learning_rate": 6.048223947306516e-05, "loss": 2.7798, "step": 50397 }, { "epoch": 3.95, "learning_rate": 6.048145534384066e-05, "loss": 2.7514, "step": 50398 }, { "epoch": 3.95, "learning_rate": 6.0480671214616166e-05, "loss": 2.6319, "step": 50399 }, { "epoch": 3.95, "learning_rate": 6.0479887085391676e-05, "loss": 2.4887, "step": 50400 }, { "epoch": 3.95, "learning_rate": 6.047910295616718e-05, "loss": 2.6668, "step": 50401 }, { "epoch": 3.95, "learning_rate": 6.0478318826942684e-05, "loss": 2.7043, "step": 50402 }, { "epoch": 3.95, "learning_rate": 6.047753469771819e-05, "loss": 2.7048, "step": 50403 }, { "epoch": 3.95, "learning_rate": 6.047675056849369e-05, "loss": 2.6952, "step": 50404 }, { "epoch": 3.95, "learning_rate": 6.0475966439269195e-05, "loss": 2.6083, "step": 50405 }, { "epoch": 3.95, "learning_rate": 6.04751823100447e-05, "loss": 2.679, "step": 50406 }, { "epoch": 3.95, "learning_rate": 6.04743981808202e-05, "loss": 2.771, "step": 50407 }, { "epoch": 3.95, "learning_rate": 6.0473614051595706e-05, "loss": 2.6534, "step": 50408 }, { "epoch": 3.95, "learning_rate": 6.047282992237121e-05, "loss": 2.8069, "step": 50409 }, { "epoch": 3.95, "learning_rate": 6.047204579314671e-05, "loss": 2.571, "step": 50410 }, { "epoch": 3.95, "learning_rate": 6.047126166392222e-05, "loss": 2.8149, "step": 50411 }, { "epoch": 3.95, "learning_rate": 6.047047753469772e-05, "loss": 2.8137, "step": 50412 }, { "epoch": 3.95, "learning_rate": 6.0469693405473224e-05, "loss": 2.6155, "step": 50413 }, { "epoch": 3.95, "learning_rate": 6.046890927624873e-05, "loss": 2.6621, "step": 50414 }, { "epoch": 3.95, "learning_rate": 6.046812514702424e-05, "loss": 2.6166, "step": 50415 }, { "epoch": 3.95, "learning_rate": 6.046734101779974e-05, "loss": 2.95, "step": 50416 }, { "epoch": 3.95, "learning_rate": 6.0466556888575246e-05, "loss": 2.7309, "step": 50417 }, { "epoch": 3.95, "learning_rate": 6.046577275935075e-05, "loss": 2.7619, "step": 50418 }, { "epoch": 3.95, "learning_rate": 6.046498863012625e-05, "loss": 2.5424, "step": 50419 }, { "epoch": 3.95, "learning_rate": 6.0464204500901757e-05, "loss": 2.7999, "step": 50420 }, { "epoch": 3.95, "learning_rate": 6.046342037167726e-05, "loss": 2.7364, "step": 50421 }, { "epoch": 3.95, "learning_rate": 6.0462636242452764e-05, "loss": 2.6078, "step": 50422 }, { "epoch": 3.95, "learning_rate": 6.046185211322827e-05, "loss": 2.8355, "step": 50423 }, { "epoch": 3.95, "learning_rate": 6.046106798400377e-05, "loss": 3.1255, "step": 50424 }, { "epoch": 3.95, "learning_rate": 6.0460283854779275e-05, "loss": 2.6379, "step": 50425 }, { "epoch": 3.95, "learning_rate": 6.045949972555477e-05, "loss": 2.5981, "step": 50426 }, { "epoch": 3.95, "learning_rate": 6.0458715596330276e-05, "loss": 2.7669, "step": 50427 }, { "epoch": 3.95, "learning_rate": 6.045793146710578e-05, "loss": 2.8228, "step": 50428 }, { "epoch": 3.95, "learning_rate": 6.045714733788128e-05, "loss": 2.7413, "step": 50429 }, { "epoch": 3.95, "learning_rate": 6.0456363208656787e-05, "loss": 2.7776, "step": 50430 }, { "epoch": 3.95, "learning_rate": 6.045557907943229e-05, "loss": 2.8251, "step": 50431 }, { "epoch": 3.95, "learning_rate": 6.0454794950207794e-05, "loss": 2.8074, "step": 50432 }, { "epoch": 3.95, "learning_rate": 6.04540108209833e-05, "loss": 2.7391, "step": 50433 }, { "epoch": 3.95, "learning_rate": 6.04532266917588e-05, "loss": 2.5627, "step": 50434 }, { "epoch": 3.95, "learning_rate": 6.0452442562534305e-05, "loss": 2.8338, "step": 50435 }, { "epoch": 3.95, "learning_rate": 6.045165843330981e-05, "loss": 2.7838, "step": 50436 }, { "epoch": 3.95, "learning_rate": 6.045087430408531e-05, "loss": 2.5766, "step": 50437 }, { "epoch": 3.95, "learning_rate": 6.0450090174860816e-05, "loss": 2.7346, "step": 50438 }, { "epoch": 3.96, "learning_rate": 6.044930604563632e-05, "loss": 2.7563, "step": 50439 }, { "epoch": 3.96, "learning_rate": 6.044852191641182e-05, "loss": 2.9558, "step": 50440 }, { "epoch": 3.96, "learning_rate": 6.0447737787187334e-05, "loss": 2.6257, "step": 50441 }, { "epoch": 3.96, "learning_rate": 6.044695365796284e-05, "loss": 2.7638, "step": 50442 }, { "epoch": 3.96, "learning_rate": 6.044616952873834e-05, "loss": 2.7209, "step": 50443 }, { "epoch": 3.96, "learning_rate": 6.0445385399513845e-05, "loss": 2.7691, "step": 50444 }, { "epoch": 3.96, "learning_rate": 6.044460127028935e-05, "loss": 2.6687, "step": 50445 }, { "epoch": 3.96, "learning_rate": 6.044381714106485e-05, "loss": 2.8149, "step": 50446 }, { "epoch": 3.96, "learning_rate": 6.0443033011840356e-05, "loss": 2.9729, "step": 50447 }, { "epoch": 3.96, "learning_rate": 6.044224888261586e-05, "loss": 2.5479, "step": 50448 }, { "epoch": 3.96, "learning_rate": 6.044146475339136e-05, "loss": 2.8908, "step": 50449 }, { "epoch": 3.96, "learning_rate": 6.044068062416687e-05, "loss": 2.7057, "step": 50450 }, { "epoch": 3.96, "learning_rate": 6.043989649494237e-05, "loss": 2.521, "step": 50451 }, { "epoch": 3.96, "learning_rate": 6.0439112365717874e-05, "loss": 2.6136, "step": 50452 }, { "epoch": 3.96, "learning_rate": 6.043832823649338e-05, "loss": 2.8293, "step": 50453 }, { "epoch": 3.96, "learning_rate": 6.043754410726888e-05, "loss": 2.7333, "step": 50454 }, { "epoch": 3.96, "learning_rate": 6.0436759978044385e-05, "loss": 2.7425, "step": 50455 }, { "epoch": 3.96, "learning_rate": 6.0435975848819896e-05, "loss": 2.5715, "step": 50456 }, { "epoch": 3.96, "learning_rate": 6.04351917195954e-05, "loss": 2.864, "step": 50457 }, { "epoch": 3.96, "learning_rate": 6.04344075903709e-05, "loss": 2.7877, "step": 50458 }, { "epoch": 3.96, "learning_rate": 6.043362346114641e-05, "loss": 2.6616, "step": 50459 }, { "epoch": 3.96, "learning_rate": 6.043283933192191e-05, "loss": 3.0756, "step": 50460 }, { "epoch": 3.96, "learning_rate": 6.0432055202697414e-05, "loss": 2.7076, "step": 50461 }, { "epoch": 3.96, "learning_rate": 6.0431271073472904e-05, "loss": 2.6508, "step": 50462 }, { "epoch": 3.96, "learning_rate": 6.043048694424841e-05, "loss": 2.898, "step": 50463 }, { "epoch": 3.96, "learning_rate": 6.042970281502391e-05, "loss": 2.6463, "step": 50464 }, { "epoch": 3.96, "learning_rate": 6.0428918685799415e-05, "loss": 2.8852, "step": 50465 }, { "epoch": 3.96, "learning_rate": 6.0428134556574926e-05, "loss": 2.6195, "step": 50466 }, { "epoch": 3.96, "learning_rate": 6.042735042735043e-05, "loss": 2.9665, "step": 50467 }, { "epoch": 3.96, "learning_rate": 6.042656629812593e-05, "loss": 2.7326, "step": 50468 }, { "epoch": 3.96, "learning_rate": 6.042578216890144e-05, "loss": 2.5662, "step": 50469 }, { "epoch": 3.96, "learning_rate": 6.042499803967694e-05, "loss": 2.7829, "step": 50470 }, { "epoch": 3.96, "learning_rate": 6.0424213910452444e-05, "loss": 2.7141, "step": 50471 }, { "epoch": 3.96, "learning_rate": 6.042342978122795e-05, "loss": 2.6097, "step": 50472 }, { "epoch": 3.96, "learning_rate": 6.042264565200345e-05, "loss": 2.8323, "step": 50473 }, { "epoch": 3.96, "learning_rate": 6.0421861522778955e-05, "loss": 2.8308, "step": 50474 }, { "epoch": 3.96, "learning_rate": 6.042107739355446e-05, "loss": 2.9047, "step": 50475 }, { "epoch": 3.96, "learning_rate": 6.042029326432996e-05, "loss": 2.8633, "step": 50476 }, { "epoch": 3.96, "learning_rate": 6.0419509135105466e-05, "loss": 2.7211, "step": 50477 }, { "epoch": 3.96, "learning_rate": 6.041872500588097e-05, "loss": 2.5901, "step": 50478 }, { "epoch": 3.96, "learning_rate": 6.0417940876656474e-05, "loss": 2.82, "step": 50479 }, { "epoch": 3.96, "learning_rate": 6.041715674743198e-05, "loss": 2.4724, "step": 50480 }, { "epoch": 3.96, "learning_rate": 6.041637261820748e-05, "loss": 2.7659, "step": 50481 }, { "epoch": 3.96, "learning_rate": 6.041558848898299e-05, "loss": 2.6618, "step": 50482 }, { "epoch": 3.96, "learning_rate": 6.0414804359758495e-05, "loss": 2.7677, "step": 50483 }, { "epoch": 3.96, "learning_rate": 6.0414020230534e-05, "loss": 2.7943, "step": 50484 }, { "epoch": 3.96, "learning_rate": 6.04132361013095e-05, "loss": 2.8952, "step": 50485 }, { "epoch": 3.96, "learning_rate": 6.0412451972085006e-05, "loss": 2.8432, "step": 50486 }, { "epoch": 3.96, "learning_rate": 6.041166784286051e-05, "loss": 2.5929, "step": 50487 }, { "epoch": 3.96, "learning_rate": 6.0410883713636013e-05, "loss": 2.7494, "step": 50488 }, { "epoch": 3.96, "learning_rate": 6.041009958441152e-05, "loss": 2.7277, "step": 50489 }, { "epoch": 3.96, "learning_rate": 6.040931545518702e-05, "loss": 2.8701, "step": 50490 }, { "epoch": 3.96, "learning_rate": 6.0408531325962524e-05, "loss": 2.6338, "step": 50491 }, { "epoch": 3.96, "learning_rate": 6.040774719673803e-05, "loss": 2.6869, "step": 50492 }, { "epoch": 3.96, "learning_rate": 6.040696306751353e-05, "loss": 2.7932, "step": 50493 }, { "epoch": 3.96, "learning_rate": 6.0406178938289035e-05, "loss": 2.8366, "step": 50494 }, { "epoch": 3.96, "learning_rate": 6.040539480906454e-05, "loss": 2.7194, "step": 50495 }, { "epoch": 3.96, "learning_rate": 6.040461067984004e-05, "loss": 2.8587, "step": 50496 }, { "epoch": 3.96, "learning_rate": 6.0403826550615547e-05, "loss": 2.6509, "step": 50497 }, { "epoch": 3.96, "learning_rate": 6.0403042421391043e-05, "loss": 2.9392, "step": 50498 }, { "epoch": 3.96, "learning_rate": 6.040225829216655e-05, "loss": 2.7025, "step": 50499 }, { "epoch": 3.96, "learning_rate": 6.040147416294205e-05, "loss": 2.6417, "step": 50500 }, { "epoch": 3.96, "learning_rate": 6.0400690033717554e-05, "loss": 2.828, "step": 50501 }, { "epoch": 3.96, "learning_rate": 6.039990590449306e-05, "loss": 2.8649, "step": 50502 }, { "epoch": 3.96, "learning_rate": 6.039912177526856e-05, "loss": 2.5684, "step": 50503 }, { "epoch": 3.96, "learning_rate": 6.0398337646044065e-05, "loss": 2.7927, "step": 50504 }, { "epoch": 3.96, "learning_rate": 6.039755351681957e-05, "loss": 2.6086, "step": 50505 }, { "epoch": 3.96, "learning_rate": 6.039676938759507e-05, "loss": 2.6731, "step": 50506 }, { "epoch": 3.96, "learning_rate": 6.0395985258370577e-05, "loss": 2.6408, "step": 50507 }, { "epoch": 3.96, "learning_rate": 6.039520112914609e-05, "loss": 2.573, "step": 50508 }, { "epoch": 3.96, "learning_rate": 6.039441699992159e-05, "loss": 2.5917, "step": 50509 }, { "epoch": 3.96, "learning_rate": 6.0393632870697094e-05, "loss": 2.7976, "step": 50510 }, { "epoch": 3.96, "learning_rate": 6.03928487414726e-05, "loss": 2.6426, "step": 50511 }, { "epoch": 3.96, "learning_rate": 6.03920646122481e-05, "loss": 2.7732, "step": 50512 }, { "epoch": 3.96, "learning_rate": 6.0391280483023605e-05, "loss": 2.7427, "step": 50513 }, { "epoch": 3.96, "learning_rate": 6.039049635379911e-05, "loss": 2.6821, "step": 50514 }, { "epoch": 3.96, "learning_rate": 6.038971222457461e-05, "loss": 2.7537, "step": 50515 }, { "epoch": 3.96, "learning_rate": 6.0388928095350116e-05, "loss": 2.8358, "step": 50516 }, { "epoch": 3.96, "learning_rate": 6.038814396612562e-05, "loss": 3.2179, "step": 50517 }, { "epoch": 3.96, "learning_rate": 6.0387359836901124e-05, "loss": 2.6384, "step": 50518 }, { "epoch": 3.96, "learning_rate": 6.038657570767663e-05, "loss": 2.7523, "step": 50519 }, { "epoch": 3.96, "learning_rate": 6.038579157845213e-05, "loss": 2.8816, "step": 50520 }, { "epoch": 3.96, "learning_rate": 6.0385007449227635e-05, "loss": 2.7547, "step": 50521 }, { "epoch": 3.96, "learning_rate": 6.038422332000314e-05, "loss": 2.7212, "step": 50522 }, { "epoch": 3.96, "learning_rate": 6.038343919077865e-05, "loss": 2.8247, "step": 50523 }, { "epoch": 3.96, "learning_rate": 6.038265506155415e-05, "loss": 2.666, "step": 50524 }, { "epoch": 3.96, "learning_rate": 6.0381870932329656e-05, "loss": 2.8682, "step": 50525 }, { "epoch": 3.96, "learning_rate": 6.038108680310516e-05, "loss": 2.499, "step": 50526 }, { "epoch": 3.96, "learning_rate": 6.0380302673880664e-05, "loss": 2.6578, "step": 50527 }, { "epoch": 3.96, "learning_rate": 6.037951854465617e-05, "loss": 2.8722, "step": 50528 }, { "epoch": 3.96, "learning_rate": 6.037873441543167e-05, "loss": 2.8666, "step": 50529 }, { "epoch": 3.96, "learning_rate": 6.0377950286207175e-05, "loss": 2.716, "step": 50530 }, { "epoch": 3.96, "learning_rate": 6.037716615698268e-05, "loss": 2.7935, "step": 50531 }, { "epoch": 3.96, "learning_rate": 6.037638202775818e-05, "loss": 2.6895, "step": 50532 }, { "epoch": 3.96, "learning_rate": 6.0375597898533686e-05, "loss": 2.5681, "step": 50533 }, { "epoch": 3.96, "learning_rate": 6.037481376930918e-05, "loss": 2.6351, "step": 50534 }, { "epoch": 3.96, "learning_rate": 6.0374029640084686e-05, "loss": 2.5695, "step": 50535 }, { "epoch": 3.96, "learning_rate": 6.037324551086019e-05, "loss": 2.778, "step": 50536 }, { "epoch": 3.96, "learning_rate": 6.0372461381635694e-05, "loss": 2.8187, "step": 50537 }, { "epoch": 3.96, "learning_rate": 6.03716772524112e-05, "loss": 2.7924, "step": 50538 }, { "epoch": 3.96, "learning_rate": 6.03708931231867e-05, "loss": 2.7725, "step": 50539 }, { "epoch": 3.96, "learning_rate": 6.0370108993962205e-05, "loss": 2.6963, "step": 50540 }, { "epoch": 3.96, "learning_rate": 6.036932486473771e-05, "loss": 2.7695, "step": 50541 }, { "epoch": 3.96, "learning_rate": 6.036854073551321e-05, "loss": 2.7065, "step": 50542 }, { "epoch": 3.96, "learning_rate": 6.0367756606288716e-05, "loss": 2.6572, "step": 50543 }, { "epoch": 3.96, "learning_rate": 6.036697247706422e-05, "loss": 2.6607, "step": 50544 }, { "epoch": 3.96, "learning_rate": 6.036618834783972e-05, "loss": 2.6166, "step": 50545 }, { "epoch": 3.96, "learning_rate": 6.036540421861523e-05, "loss": 2.5669, "step": 50546 }, { "epoch": 3.96, "learning_rate": 6.036462008939073e-05, "loss": 2.7257, "step": 50547 }, { "epoch": 3.96, "learning_rate": 6.0363835960166234e-05, "loss": 2.7223, "step": 50548 }, { "epoch": 3.96, "learning_rate": 6.0363051830941745e-05, "loss": 2.8085, "step": 50549 }, { "epoch": 3.96, "learning_rate": 6.036226770171725e-05, "loss": 2.8254, "step": 50550 }, { "epoch": 3.96, "learning_rate": 6.036148357249275e-05, "loss": 2.8246, "step": 50551 }, { "epoch": 3.96, "learning_rate": 6.0360699443268256e-05, "loss": 2.5387, "step": 50552 }, { "epoch": 3.96, "learning_rate": 6.035991531404376e-05, "loss": 2.5296, "step": 50553 }, { "epoch": 3.96, "learning_rate": 6.035913118481926e-05, "loss": 2.7303, "step": 50554 }, { "epoch": 3.96, "learning_rate": 6.0358347055594767e-05, "loss": 2.5636, "step": 50555 }, { "epoch": 3.96, "learning_rate": 6.035756292637027e-05, "loss": 2.71, "step": 50556 }, { "epoch": 3.96, "learning_rate": 6.0356778797145774e-05, "loss": 2.6893, "step": 50557 }, { "epoch": 3.96, "learning_rate": 6.035599466792128e-05, "loss": 2.7032, "step": 50558 }, { "epoch": 3.96, "learning_rate": 6.035521053869678e-05, "loss": 2.5316, "step": 50559 }, { "epoch": 3.96, "learning_rate": 6.0354426409472285e-05, "loss": 2.5013, "step": 50560 }, { "epoch": 3.96, "learning_rate": 6.035364228024779e-05, "loss": 2.8402, "step": 50561 }, { "epoch": 3.96, "learning_rate": 6.035285815102329e-05, "loss": 2.6172, "step": 50562 }, { "epoch": 3.96, "learning_rate": 6.0352074021798796e-05, "loss": 2.7228, "step": 50563 }, { "epoch": 3.96, "learning_rate": 6.0351289892574306e-05, "loss": 2.7317, "step": 50564 }, { "epoch": 3.96, "learning_rate": 6.035050576334981e-05, "loss": 2.5832, "step": 50565 }, { "epoch": 3.97, "learning_rate": 6.0349721634125314e-05, "loss": 2.7829, "step": 50566 }, { "epoch": 3.97, "learning_rate": 6.034893750490082e-05, "loss": 2.7605, "step": 50567 }, { "epoch": 3.97, "learning_rate": 6.034815337567632e-05, "loss": 2.6686, "step": 50568 }, { "epoch": 3.97, "learning_rate": 6.0347369246451825e-05, "loss": 2.6964, "step": 50569 }, { "epoch": 3.97, "learning_rate": 6.0346585117227315e-05, "loss": 2.819, "step": 50570 }, { "epoch": 3.97, "learning_rate": 6.034580098800282e-05, "loss": 2.6917, "step": 50571 }, { "epoch": 3.97, "learning_rate": 6.034501685877832e-05, "loss": 2.8881, "step": 50572 }, { "epoch": 3.97, "learning_rate": 6.0344232729553826e-05, "loss": 2.9903, "step": 50573 }, { "epoch": 3.97, "learning_rate": 6.0343448600329336e-05, "loss": 2.835, "step": 50574 }, { "epoch": 3.97, "learning_rate": 6.034266447110484e-05, "loss": 3.0946, "step": 50575 }, { "epoch": 3.97, "learning_rate": 6.0341880341880344e-05, "loss": 2.7354, "step": 50576 }, { "epoch": 3.97, "learning_rate": 6.034109621265585e-05, "loss": 2.6299, "step": 50577 }, { "epoch": 3.97, "learning_rate": 6.034031208343135e-05, "loss": 2.6516, "step": 50578 }, { "epoch": 3.97, "learning_rate": 6.0339527954206855e-05, "loss": 2.7876, "step": 50579 }, { "epoch": 3.97, "learning_rate": 6.033874382498236e-05, "loss": 2.5883, "step": 50580 }, { "epoch": 3.97, "learning_rate": 6.033795969575786e-05, "loss": 2.7099, "step": 50581 }, { "epoch": 3.97, "learning_rate": 6.0337175566533366e-05, "loss": 2.8622, "step": 50582 }, { "epoch": 3.97, "learning_rate": 6.033639143730887e-05, "loss": 3.1269, "step": 50583 }, { "epoch": 3.97, "learning_rate": 6.033560730808437e-05, "loss": 2.7779, "step": 50584 }, { "epoch": 3.97, "learning_rate": 6.033482317885988e-05, "loss": 2.7848, "step": 50585 }, { "epoch": 3.97, "learning_rate": 6.033403904963538e-05, "loss": 2.5531, "step": 50586 }, { "epoch": 3.97, "learning_rate": 6.0333254920410884e-05, "loss": 3.0602, "step": 50587 }, { "epoch": 3.97, "learning_rate": 6.033247079118639e-05, "loss": 2.6811, "step": 50588 }, { "epoch": 3.97, "learning_rate": 6.033168666196189e-05, "loss": 2.551, "step": 50589 }, { "epoch": 3.97, "learning_rate": 6.03309025327374e-05, "loss": 2.7862, "step": 50590 }, { "epoch": 3.97, "learning_rate": 6.0330118403512906e-05, "loss": 2.7087, "step": 50591 }, { "epoch": 3.97, "learning_rate": 6.032933427428841e-05, "loss": 2.8386, "step": 50592 }, { "epoch": 3.97, "learning_rate": 6.032855014506391e-05, "loss": 2.7389, "step": 50593 }, { "epoch": 3.97, "learning_rate": 6.032776601583942e-05, "loss": 2.7768, "step": 50594 }, { "epoch": 3.97, "learning_rate": 6.032698188661492e-05, "loss": 2.7364, "step": 50595 }, { "epoch": 3.97, "learning_rate": 6.0326197757390424e-05, "loss": 2.726, "step": 50596 }, { "epoch": 3.97, "learning_rate": 6.032541362816593e-05, "loss": 2.8479, "step": 50597 }, { "epoch": 3.97, "learning_rate": 6.032462949894143e-05, "loss": 2.8742, "step": 50598 }, { "epoch": 3.97, "learning_rate": 6.0323845369716935e-05, "loss": 2.5483, "step": 50599 }, { "epoch": 3.97, "learning_rate": 6.032306124049244e-05, "loss": 2.873, "step": 50600 }, { "epoch": 3.97, "learning_rate": 6.032227711126794e-05, "loss": 2.7249, "step": 50601 }, { "epoch": 3.97, "learning_rate": 6.0321492982043446e-05, "loss": 2.5342, "step": 50602 }, { "epoch": 3.97, "learning_rate": 6.032070885281895e-05, "loss": 2.6804, "step": 50603 }, { "epoch": 3.97, "learning_rate": 6.0319924723594454e-05, "loss": 2.6197, "step": 50604 }, { "epoch": 3.97, "learning_rate": 6.031914059436996e-05, "loss": 2.6902, "step": 50605 }, { "epoch": 3.97, "learning_rate": 6.0318356465145454e-05, "loss": 2.7896, "step": 50606 }, { "epoch": 3.97, "learning_rate": 6.031757233592096e-05, "loss": 2.6478, "step": 50607 }, { "epoch": 3.97, "learning_rate": 6.031678820669646e-05, "loss": 2.9036, "step": 50608 }, { "epoch": 3.97, "learning_rate": 6.0316004077471965e-05, "loss": 2.6487, "step": 50609 }, { "epoch": 3.97, "learning_rate": 6.031521994824747e-05, "loss": 2.7445, "step": 50610 }, { "epoch": 3.97, "learning_rate": 6.031443581902297e-05, "loss": 2.7728, "step": 50611 }, { "epoch": 3.97, "learning_rate": 6.0313651689798476e-05, "loss": 2.7073, "step": 50612 }, { "epoch": 3.97, "learning_rate": 6.031286756057398e-05, "loss": 2.8164, "step": 50613 }, { "epoch": 3.97, "learning_rate": 6.0312083431349484e-05, "loss": 2.7767, "step": 50614 }, { "epoch": 3.97, "learning_rate": 6.031129930212499e-05, "loss": 2.7377, "step": 50615 }, { "epoch": 3.97, "learning_rate": 6.03105151729005e-05, "loss": 2.9842, "step": 50616 }, { "epoch": 3.97, "learning_rate": 6.0309731043676e-05, "loss": 2.6344, "step": 50617 }, { "epoch": 3.97, "learning_rate": 6.0308946914451505e-05, "loss": 2.7252, "step": 50618 }, { "epoch": 3.97, "learning_rate": 6.030816278522701e-05, "loss": 2.6502, "step": 50619 }, { "epoch": 3.97, "learning_rate": 6.030737865600251e-05, "loss": 2.6585, "step": 50620 }, { "epoch": 3.97, "learning_rate": 6.0306594526778016e-05, "loss": 2.7154, "step": 50621 }, { "epoch": 3.97, "learning_rate": 6.030581039755352e-05, "loss": 2.5932, "step": 50622 }, { "epoch": 3.97, "learning_rate": 6.0305026268329023e-05, "loss": 2.7627, "step": 50623 }, { "epoch": 3.97, "learning_rate": 6.030424213910453e-05, "loss": 2.9641, "step": 50624 }, { "epoch": 3.97, "learning_rate": 6.030345800988003e-05, "loss": 2.7218, "step": 50625 }, { "epoch": 3.97, "learning_rate": 6.0302673880655534e-05, "loss": 2.7106, "step": 50626 }, { "epoch": 3.97, "learning_rate": 6.030188975143104e-05, "loss": 2.7353, "step": 50627 }, { "epoch": 3.97, "learning_rate": 6.030110562220654e-05, "loss": 2.6481, "step": 50628 }, { "epoch": 3.97, "learning_rate": 6.0300321492982046e-05, "loss": 2.7588, "step": 50629 }, { "epoch": 3.97, "learning_rate": 6.029953736375755e-05, "loss": 2.6048, "step": 50630 }, { "epoch": 3.97, "learning_rate": 6.029875323453306e-05, "loss": 2.8131, "step": 50631 }, { "epoch": 3.97, "learning_rate": 6.029796910530856e-05, "loss": 2.658, "step": 50632 }, { "epoch": 3.97, "learning_rate": 6.029718497608407e-05, "loss": 2.7438, "step": 50633 }, { "epoch": 3.97, "learning_rate": 6.029640084685957e-05, "loss": 2.6708, "step": 50634 }, { "epoch": 3.97, "learning_rate": 6.0295616717635074e-05, "loss": 2.5248, "step": 50635 }, { "epoch": 3.97, "learning_rate": 6.029483258841058e-05, "loss": 2.6012, "step": 50636 }, { "epoch": 3.97, "learning_rate": 6.029404845918608e-05, "loss": 2.8825, "step": 50637 }, { "epoch": 3.97, "learning_rate": 6.0293264329961585e-05, "loss": 2.614, "step": 50638 }, { "epoch": 3.97, "learning_rate": 6.029248020073709e-05, "loss": 2.5639, "step": 50639 }, { "epoch": 3.97, "learning_rate": 6.029169607151259e-05, "loss": 2.8117, "step": 50640 }, { "epoch": 3.97, "learning_rate": 6.0290911942288096e-05, "loss": 2.8368, "step": 50641 }, { "epoch": 3.97, "learning_rate": 6.029012781306359e-05, "loss": 2.6869, "step": 50642 }, { "epoch": 3.97, "learning_rate": 6.02893436838391e-05, "loss": 2.6341, "step": 50643 }, { "epoch": 3.97, "learning_rate": 6.02885595546146e-05, "loss": 2.871, "step": 50644 }, { "epoch": 3.97, "learning_rate": 6.0287775425390104e-05, "loss": 2.5082, "step": 50645 }, { "epoch": 3.97, "learning_rate": 6.028699129616561e-05, "loss": 2.7778, "step": 50646 }, { "epoch": 3.97, "learning_rate": 6.028620716694111e-05, "loss": 2.5564, "step": 50647 }, { "epoch": 3.97, "learning_rate": 6.0285423037716615e-05, "loss": 2.6535, "step": 50648 }, { "epoch": 3.97, "learning_rate": 6.028463890849212e-05, "loss": 2.5596, "step": 50649 }, { "epoch": 3.97, "learning_rate": 6.028385477926762e-05, "loss": 2.8076, "step": 50650 }, { "epoch": 3.97, "learning_rate": 6.0283070650043126e-05, "loss": 2.8174, "step": 50651 }, { "epoch": 3.97, "learning_rate": 6.028228652081863e-05, "loss": 2.7921, "step": 50652 }, { "epoch": 3.97, "learning_rate": 6.0281502391594134e-05, "loss": 2.6306, "step": 50653 }, { "epoch": 3.97, "learning_rate": 6.028071826236964e-05, "loss": 2.8186, "step": 50654 }, { "epoch": 3.97, "learning_rate": 6.027993413314514e-05, "loss": 2.6043, "step": 50655 }, { "epoch": 3.97, "learning_rate": 6.0279150003920645e-05, "loss": 2.6647, "step": 50656 }, { "epoch": 3.97, "learning_rate": 6.0278365874696155e-05, "loss": 2.7382, "step": 50657 }, { "epoch": 3.97, "learning_rate": 6.027758174547166e-05, "loss": 2.6273, "step": 50658 }, { "epoch": 3.97, "learning_rate": 6.027679761624716e-05, "loss": 2.6374, "step": 50659 }, { "epoch": 3.97, "learning_rate": 6.0276013487022666e-05, "loss": 2.6676, "step": 50660 }, { "epoch": 3.97, "learning_rate": 6.027522935779817e-05, "loss": 2.6419, "step": 50661 }, { "epoch": 3.97, "learning_rate": 6.0274445228573674e-05, "loss": 2.7461, "step": 50662 }, { "epoch": 3.97, "learning_rate": 6.027366109934918e-05, "loss": 2.9774, "step": 50663 }, { "epoch": 3.97, "learning_rate": 6.027287697012468e-05, "loss": 2.9168, "step": 50664 }, { "epoch": 3.97, "learning_rate": 6.0272092840900185e-05, "loss": 2.83, "step": 50665 }, { "epoch": 3.97, "learning_rate": 6.027130871167569e-05, "loss": 2.7856, "step": 50666 }, { "epoch": 3.97, "learning_rate": 6.027052458245119e-05, "loss": 2.8008, "step": 50667 }, { "epoch": 3.97, "learning_rate": 6.0269740453226696e-05, "loss": 2.6908, "step": 50668 }, { "epoch": 3.97, "learning_rate": 6.02689563240022e-05, "loss": 2.8594, "step": 50669 }, { "epoch": 3.97, "learning_rate": 6.02681721947777e-05, "loss": 2.7956, "step": 50670 }, { "epoch": 3.97, "learning_rate": 6.026738806555321e-05, "loss": 2.7026, "step": 50671 }, { "epoch": 3.97, "learning_rate": 6.026660393632871e-05, "loss": 2.7159, "step": 50672 }, { "epoch": 3.97, "learning_rate": 6.026581980710422e-05, "loss": 2.8712, "step": 50673 }, { "epoch": 3.97, "learning_rate": 6.0265035677879725e-05, "loss": 2.7092, "step": 50674 }, { "epoch": 3.97, "learning_rate": 6.026425154865523e-05, "loss": 2.7985, "step": 50675 }, { "epoch": 3.97, "learning_rate": 6.026346741943073e-05, "loss": 2.6035, "step": 50676 }, { "epoch": 3.97, "learning_rate": 6.0262683290206236e-05, "loss": 2.6886, "step": 50677 }, { "epoch": 3.97, "learning_rate": 6.0261899160981726e-05, "loss": 2.5184, "step": 50678 }, { "epoch": 3.97, "learning_rate": 6.026111503175723e-05, "loss": 2.7217, "step": 50679 }, { "epoch": 3.97, "learning_rate": 6.026033090253273e-05, "loss": 2.7153, "step": 50680 }, { "epoch": 3.97, "learning_rate": 6.025954677330824e-05, "loss": 2.9378, "step": 50681 }, { "epoch": 3.97, "learning_rate": 6.025876264408374e-05, "loss": 2.7401, "step": 50682 }, { "epoch": 3.97, "learning_rate": 6.025797851485925e-05, "loss": 2.7484, "step": 50683 }, { "epoch": 3.97, "learning_rate": 6.0257194385634755e-05, "loss": 2.6381, "step": 50684 }, { "epoch": 3.97, "learning_rate": 6.025641025641026e-05, "loss": 3.0581, "step": 50685 }, { "epoch": 3.97, "learning_rate": 6.025562612718576e-05, "loss": 2.6873, "step": 50686 }, { "epoch": 3.97, "learning_rate": 6.0254841997961266e-05, "loss": 2.8487, "step": 50687 }, { "epoch": 3.97, "learning_rate": 6.025405786873677e-05, "loss": 2.5041, "step": 50688 }, { "epoch": 3.97, "learning_rate": 6.025327373951227e-05, "loss": 2.6468, "step": 50689 }, { "epoch": 3.97, "learning_rate": 6.025248961028778e-05, "loss": 2.8916, "step": 50690 }, { "epoch": 3.97, "learning_rate": 6.025170548106328e-05, "loss": 2.7835, "step": 50691 }, { "epoch": 3.97, "learning_rate": 6.0250921351838784e-05, "loss": 2.6344, "step": 50692 }, { "epoch": 3.97, "learning_rate": 6.025013722261429e-05, "loss": 2.6936, "step": 50693 }, { "epoch": 3.98, "learning_rate": 6.024935309338979e-05, "loss": 2.6512, "step": 50694 }, { "epoch": 3.98, "learning_rate": 6.0248568964165295e-05, "loss": 2.8249, "step": 50695 }, { "epoch": 3.98, "learning_rate": 6.02477848349408e-05, "loss": 2.5391, "step": 50696 }, { "epoch": 3.98, "learning_rate": 6.02470007057163e-05, "loss": 2.8536, "step": 50697 }, { "epoch": 3.98, "learning_rate": 6.024621657649181e-05, "loss": 2.6156, "step": 50698 }, { "epoch": 3.98, "learning_rate": 6.0245432447267317e-05, "loss": 2.919, "step": 50699 }, { "epoch": 3.98, "learning_rate": 6.024464831804282e-05, "loss": 2.5886, "step": 50700 }, { "epoch": 3.98, "learning_rate": 6.0243864188818324e-05, "loss": 2.7084, "step": 50701 }, { "epoch": 3.98, "learning_rate": 6.024308005959383e-05, "loss": 2.9105, "step": 50702 }, { "epoch": 3.98, "learning_rate": 6.024229593036933e-05, "loss": 2.6045, "step": 50703 }, { "epoch": 3.98, "learning_rate": 6.0241511801144835e-05, "loss": 2.7521, "step": 50704 }, { "epoch": 3.98, "learning_rate": 6.024072767192034e-05, "loss": 2.522, "step": 50705 }, { "epoch": 3.98, "learning_rate": 6.023994354269584e-05, "loss": 2.9926, "step": 50706 }, { "epoch": 3.98, "learning_rate": 6.0239159413471346e-05, "loss": 2.6682, "step": 50707 }, { "epoch": 3.98, "learning_rate": 6.023837528424685e-05, "loss": 2.7584, "step": 50708 }, { "epoch": 3.98, "learning_rate": 6.023759115502235e-05, "loss": 2.6465, "step": 50709 }, { "epoch": 3.98, "learning_rate": 6.023680702579786e-05, "loss": 2.6545, "step": 50710 }, { "epoch": 3.98, "learning_rate": 6.023602289657336e-05, "loss": 2.5374, "step": 50711 }, { "epoch": 3.98, "learning_rate": 6.0235238767348864e-05, "loss": 2.6441, "step": 50712 }, { "epoch": 3.98, "learning_rate": 6.023445463812437e-05, "loss": 2.9318, "step": 50713 }, { "epoch": 3.98, "learning_rate": 6.0233670508899865e-05, "loss": 2.7513, "step": 50714 }, { "epoch": 3.98, "learning_rate": 6.023288637967537e-05, "loss": 2.6717, "step": 50715 }, { "epoch": 3.98, "learning_rate": 6.023210225045087e-05, "loss": 2.7016, "step": 50716 }, { "epoch": 3.98, "learning_rate": 6.0231318121226376e-05, "loss": 2.6632, "step": 50717 }, { "epoch": 3.98, "learning_rate": 6.023053399200188e-05, "loss": 2.5774, "step": 50718 }, { "epoch": 3.98, "learning_rate": 6.022974986277738e-05, "loss": 2.713, "step": 50719 }, { "epoch": 3.98, "learning_rate": 6.022896573355289e-05, "loss": 2.6724, "step": 50720 }, { "epoch": 3.98, "learning_rate": 6.022818160432839e-05, "loss": 2.8206, "step": 50721 }, { "epoch": 3.98, "learning_rate": 6.0227397475103894e-05, "loss": 2.7102, "step": 50722 }, { "epoch": 3.98, "learning_rate": 6.02266133458794e-05, "loss": 2.3865, "step": 50723 }, { "epoch": 3.98, "learning_rate": 6.022582921665491e-05, "loss": 2.7691, "step": 50724 }, { "epoch": 3.98, "learning_rate": 6.022504508743041e-05, "loss": 2.7063, "step": 50725 }, { "epoch": 3.98, "learning_rate": 6.0224260958205916e-05, "loss": 2.8956, "step": 50726 }, { "epoch": 3.98, "learning_rate": 6.022347682898142e-05, "loss": 2.5835, "step": 50727 }, { "epoch": 3.98, "learning_rate": 6.022269269975692e-05, "loss": 2.884, "step": 50728 }, { "epoch": 3.98, "learning_rate": 6.022190857053243e-05, "loss": 2.7824, "step": 50729 }, { "epoch": 3.98, "learning_rate": 6.022112444130793e-05, "loss": 2.9582, "step": 50730 }, { "epoch": 3.98, "learning_rate": 6.0220340312083434e-05, "loss": 2.6604, "step": 50731 }, { "epoch": 3.98, "learning_rate": 6.021955618285894e-05, "loss": 2.8752, "step": 50732 }, { "epoch": 3.98, "learning_rate": 6.021877205363444e-05, "loss": 2.8128, "step": 50733 }, { "epoch": 3.98, "learning_rate": 6.0217987924409945e-05, "loss": 3.1339, "step": 50734 }, { "epoch": 3.98, "learning_rate": 6.021720379518545e-05, "loss": 2.7425, "step": 50735 }, { "epoch": 3.98, "learning_rate": 6.021641966596095e-05, "loss": 2.6147, "step": 50736 }, { "epoch": 3.98, "learning_rate": 6.0215635536736456e-05, "loss": 2.5527, "step": 50737 }, { "epoch": 3.98, "learning_rate": 6.021485140751196e-05, "loss": 2.7049, "step": 50738 }, { "epoch": 3.98, "learning_rate": 6.021406727828747e-05, "loss": 2.8412, "step": 50739 }, { "epoch": 3.98, "learning_rate": 6.0213283149062974e-05, "loss": 2.895, "step": 50740 }, { "epoch": 3.98, "learning_rate": 6.021249901983848e-05, "loss": 2.5674, "step": 50741 }, { "epoch": 3.98, "learning_rate": 6.021171489061398e-05, "loss": 2.8019, "step": 50742 }, { "epoch": 3.98, "learning_rate": 6.0210930761389485e-05, "loss": 2.6489, "step": 50743 }, { "epoch": 3.98, "learning_rate": 6.021014663216499e-05, "loss": 2.6454, "step": 50744 }, { "epoch": 3.98, "learning_rate": 6.020936250294049e-05, "loss": 2.6335, "step": 50745 }, { "epoch": 3.98, "learning_rate": 6.0208578373715996e-05, "loss": 2.5429, "step": 50746 }, { "epoch": 3.98, "learning_rate": 6.02077942444915e-05, "loss": 2.6897, "step": 50747 }, { "epoch": 3.98, "learning_rate": 6.0207010115267004e-05, "loss": 2.7636, "step": 50748 }, { "epoch": 3.98, "learning_rate": 6.020622598604251e-05, "loss": 2.8843, "step": 50749 }, { "epoch": 3.98, "learning_rate": 6.0205441856818004e-05, "loss": 2.8846, "step": 50750 }, { "epoch": 3.98, "learning_rate": 6.020465772759351e-05, "loss": 2.7719, "step": 50751 }, { "epoch": 3.98, "learning_rate": 6.020387359836901e-05, "loss": 2.5242, "step": 50752 }, { "epoch": 3.98, "learning_rate": 6.0203089469144515e-05, "loss": 2.8037, "step": 50753 }, { "epoch": 3.98, "learning_rate": 6.020230533992002e-05, "loss": 2.6842, "step": 50754 }, { "epoch": 3.98, "learning_rate": 6.020152121069552e-05, "loss": 2.7265, "step": 50755 }, { "epoch": 3.98, "learning_rate": 6.0200737081471026e-05, "loss": 2.9842, "step": 50756 }, { "epoch": 3.98, "learning_rate": 6.019995295224653e-05, "loss": 2.8114, "step": 50757 }, { "epoch": 3.98, "learning_rate": 6.0199168823022034e-05, "loss": 2.718, "step": 50758 }, { "epoch": 3.98, "learning_rate": 6.019838469379754e-05, "loss": 2.6056, "step": 50759 }, { "epoch": 3.98, "learning_rate": 6.019760056457304e-05, "loss": 2.647, "step": 50760 }, { "epoch": 3.98, "learning_rate": 6.0196816435348545e-05, "loss": 3.0102, "step": 50761 }, { "epoch": 3.98, "learning_rate": 6.019603230612405e-05, "loss": 2.8073, "step": 50762 }, { "epoch": 3.98, "learning_rate": 6.019524817689955e-05, "loss": 2.8757, "step": 50763 }, { "epoch": 3.98, "learning_rate": 6.0194464047675056e-05, "loss": 2.6424, "step": 50764 }, { "epoch": 3.98, "learning_rate": 6.0193679918450566e-05, "loss": 2.7851, "step": 50765 }, { "epoch": 3.98, "learning_rate": 6.019289578922607e-05, "loss": 2.7097, "step": 50766 }, { "epoch": 3.98, "learning_rate": 6.019211166000157e-05, "loss": 2.6239, "step": 50767 }, { "epoch": 3.98, "learning_rate": 6.019132753077708e-05, "loss": 2.9114, "step": 50768 }, { "epoch": 3.98, "learning_rate": 6.019054340155258e-05, "loss": 2.9782, "step": 50769 }, { "epoch": 3.98, "learning_rate": 6.0189759272328084e-05, "loss": 2.6288, "step": 50770 }, { "epoch": 3.98, "learning_rate": 6.018897514310359e-05, "loss": 2.7461, "step": 50771 }, { "epoch": 3.98, "learning_rate": 6.018819101387909e-05, "loss": 2.714, "step": 50772 }, { "epoch": 3.98, "learning_rate": 6.0187406884654595e-05, "loss": 2.6347, "step": 50773 }, { "epoch": 3.98, "learning_rate": 6.01866227554301e-05, "loss": 3.0076, "step": 50774 }, { "epoch": 3.98, "learning_rate": 6.01858386262056e-05, "loss": 2.6533, "step": 50775 }, { "epoch": 3.98, "learning_rate": 6.0185054496981106e-05, "loss": 2.7317, "step": 50776 }, { "epoch": 3.98, "learning_rate": 6.018427036775661e-05, "loss": 2.8885, "step": 50777 }, { "epoch": 3.98, "learning_rate": 6.0183486238532114e-05, "loss": 2.6706, "step": 50778 }, { "epoch": 3.98, "learning_rate": 6.018270210930762e-05, "loss": 2.7942, "step": 50779 }, { "epoch": 3.98, "learning_rate": 6.018191798008312e-05, "loss": 2.8375, "step": 50780 }, { "epoch": 3.98, "learning_rate": 6.018113385085863e-05, "loss": 2.6806, "step": 50781 }, { "epoch": 3.98, "learning_rate": 6.0180349721634135e-05, "loss": 2.7349, "step": 50782 }, { "epoch": 3.98, "learning_rate": 6.017956559240964e-05, "loss": 2.8311, "step": 50783 }, { "epoch": 3.98, "learning_rate": 6.017878146318514e-05, "loss": 2.7344, "step": 50784 }, { "epoch": 3.98, "learning_rate": 6.0177997333960646e-05, "loss": 2.8931, "step": 50785 }, { "epoch": 3.98, "learning_rate": 6.0177213204736136e-05, "loss": 2.6386, "step": 50786 }, { "epoch": 3.98, "learning_rate": 6.017642907551164e-05, "loss": 2.6584, "step": 50787 }, { "epoch": 3.98, "learning_rate": 6.0175644946287144e-05, "loss": 2.4933, "step": 50788 }, { "epoch": 3.98, "learning_rate": 6.017486081706265e-05, "loss": 2.7158, "step": 50789 }, { "epoch": 3.98, "learning_rate": 6.017407668783815e-05, "loss": 2.5582, "step": 50790 }, { "epoch": 3.98, "learning_rate": 6.017329255861366e-05, "loss": 2.8857, "step": 50791 }, { "epoch": 3.98, "learning_rate": 6.0172508429389165e-05, "loss": 2.6877, "step": 50792 }, { "epoch": 3.98, "learning_rate": 6.017172430016467e-05, "loss": 2.8952, "step": 50793 }, { "epoch": 3.98, "learning_rate": 6.017094017094017e-05, "loss": 2.8292, "step": 50794 }, { "epoch": 3.98, "learning_rate": 6.0170156041715676e-05, "loss": 2.7805, "step": 50795 }, { "epoch": 3.98, "learning_rate": 6.016937191249118e-05, "loss": 2.6516, "step": 50796 }, { "epoch": 3.98, "learning_rate": 6.0168587783266684e-05, "loss": 2.7196, "step": 50797 }, { "epoch": 3.98, "learning_rate": 6.016780365404219e-05, "loss": 2.9175, "step": 50798 }, { "epoch": 3.98, "learning_rate": 6.016701952481769e-05, "loss": 2.945, "step": 50799 }, { "epoch": 3.98, "learning_rate": 6.0166235395593195e-05, "loss": 2.6654, "step": 50800 }, { "epoch": 3.98, "learning_rate": 6.01654512663687e-05, "loss": 2.6966, "step": 50801 }, { "epoch": 3.98, "learning_rate": 6.01646671371442e-05, "loss": 2.6938, "step": 50802 }, { "epoch": 3.98, "learning_rate": 6.0163883007919706e-05, "loss": 2.7648, "step": 50803 }, { "epoch": 3.98, "learning_rate": 6.016309887869521e-05, "loss": 2.7172, "step": 50804 }, { "epoch": 3.98, "learning_rate": 6.016231474947071e-05, "loss": 2.5347, "step": 50805 }, { "epoch": 3.98, "learning_rate": 6.0161530620246224e-05, "loss": 2.8087, "step": 50806 }, { "epoch": 3.98, "learning_rate": 6.016074649102173e-05, "loss": 2.6604, "step": 50807 }, { "epoch": 3.98, "learning_rate": 6.015996236179723e-05, "loss": 2.7424, "step": 50808 }, { "epoch": 3.98, "learning_rate": 6.0159178232572735e-05, "loss": 2.8804, "step": 50809 }, { "epoch": 3.98, "learning_rate": 6.015839410334824e-05, "loss": 2.5894, "step": 50810 }, { "epoch": 3.98, "learning_rate": 6.015760997412374e-05, "loss": 2.5686, "step": 50811 }, { "epoch": 3.98, "learning_rate": 6.0156825844899246e-05, "loss": 2.6296, "step": 50812 }, { "epoch": 3.98, "learning_rate": 6.015604171567475e-05, "loss": 2.8558, "step": 50813 }, { "epoch": 3.98, "learning_rate": 6.015525758645025e-05, "loss": 2.8317, "step": 50814 }, { "epoch": 3.98, "learning_rate": 6.015447345722576e-05, "loss": 2.5335, "step": 50815 }, { "epoch": 3.98, "learning_rate": 6.015368932800126e-05, "loss": 2.6774, "step": 50816 }, { "epoch": 3.98, "learning_rate": 6.0152905198776764e-05, "loss": 2.8332, "step": 50817 }, { "epoch": 3.98, "learning_rate": 6.015212106955227e-05, "loss": 2.595, "step": 50818 }, { "epoch": 3.98, "learning_rate": 6.015133694032777e-05, "loss": 2.5918, "step": 50819 }, { "epoch": 3.98, "learning_rate": 6.0150552811103275e-05, "loss": 2.8264, "step": 50820 }, { "epoch": 3.99, "learning_rate": 6.014976868187878e-05, "loss": 2.8567, "step": 50821 }, { "epoch": 3.99, "learning_rate": 6.0148984552654276e-05, "loss": 2.9109, "step": 50822 }, { "epoch": 3.99, "learning_rate": 6.014820042342978e-05, "loss": 3.0054, "step": 50823 }, { "epoch": 3.99, "learning_rate": 6.014741629420528e-05, "loss": 2.5304, "step": 50824 }, { "epoch": 3.99, "learning_rate": 6.014663216498079e-05, "loss": 2.8559, "step": 50825 }, { "epoch": 3.99, "learning_rate": 6.014584803575629e-05, "loss": 2.7522, "step": 50826 }, { "epoch": 3.99, "learning_rate": 6.0145063906531794e-05, "loss": 2.8252, "step": 50827 }, { "epoch": 3.99, "learning_rate": 6.01442797773073e-05, "loss": 2.6076, "step": 50828 }, { "epoch": 3.99, "learning_rate": 6.01434956480828e-05, "loss": 2.8725, "step": 50829 }, { "epoch": 3.99, "learning_rate": 6.0142711518858305e-05, "loss": 2.8894, "step": 50830 }, { "epoch": 3.99, "learning_rate": 6.014192738963381e-05, "loss": 2.7392, "step": 50831 }, { "epoch": 3.99, "learning_rate": 6.014114326040932e-05, "loss": 2.7739, "step": 50832 }, { "epoch": 3.99, "learning_rate": 6.014035913118482e-05, "loss": 2.7714, "step": 50833 }, { "epoch": 3.99, "learning_rate": 6.0139575001960327e-05, "loss": 2.7596, "step": 50834 }, { "epoch": 3.99, "learning_rate": 6.013879087273583e-05, "loss": 2.9284, "step": 50835 }, { "epoch": 3.99, "learning_rate": 6.0138006743511334e-05, "loss": 2.8635, "step": 50836 }, { "epoch": 3.99, "learning_rate": 6.013722261428684e-05, "loss": 2.7314, "step": 50837 }, { "epoch": 3.99, "learning_rate": 6.013643848506234e-05, "loss": 2.7751, "step": 50838 }, { "epoch": 3.99, "learning_rate": 6.0135654355837845e-05, "loss": 2.6093, "step": 50839 }, { "epoch": 3.99, "learning_rate": 6.013487022661335e-05, "loss": 2.6435, "step": 50840 }, { "epoch": 3.99, "learning_rate": 6.013408609738885e-05, "loss": 2.6439, "step": 50841 }, { "epoch": 3.99, "learning_rate": 6.0133301968164356e-05, "loss": 2.5744, "step": 50842 }, { "epoch": 3.99, "learning_rate": 6.013251783893986e-05, "loss": 2.934, "step": 50843 }, { "epoch": 3.99, "learning_rate": 6.013173370971536e-05, "loss": 3.0713, "step": 50844 }, { "epoch": 3.99, "learning_rate": 6.013094958049087e-05, "loss": 2.508, "step": 50845 }, { "epoch": 3.99, "learning_rate": 6.013016545126637e-05, "loss": 2.9599, "step": 50846 }, { "epoch": 3.99, "learning_rate": 6.012938132204188e-05, "loss": 2.7156, "step": 50847 }, { "epoch": 3.99, "learning_rate": 6.0128597192817385e-05, "loss": 2.5252, "step": 50848 }, { "epoch": 3.99, "learning_rate": 6.012781306359289e-05, "loss": 2.6745, "step": 50849 }, { "epoch": 3.99, "learning_rate": 6.012702893436839e-05, "loss": 2.581, "step": 50850 }, { "epoch": 3.99, "learning_rate": 6.0126244805143896e-05, "loss": 2.6287, "step": 50851 }, { "epoch": 3.99, "learning_rate": 6.01254606759194e-05, "loss": 2.65, "step": 50852 }, { "epoch": 3.99, "learning_rate": 6.01246765466949e-05, "loss": 2.8692, "step": 50853 }, { "epoch": 3.99, "learning_rate": 6.012389241747041e-05, "loss": 2.7578, "step": 50854 }, { "epoch": 3.99, "learning_rate": 6.012310828824591e-05, "loss": 2.6438, "step": 50855 }, { "epoch": 3.99, "learning_rate": 6.0122324159021414e-05, "loss": 2.9305, "step": 50856 }, { "epoch": 3.99, "learning_rate": 6.012154002979692e-05, "loss": 2.618, "step": 50857 }, { "epoch": 3.99, "learning_rate": 6.0120755900572415e-05, "loss": 2.818, "step": 50858 }, { "epoch": 3.99, "learning_rate": 6.011997177134792e-05, "loss": 3.0395, "step": 50859 }, { "epoch": 3.99, "learning_rate": 6.011918764212342e-05, "loss": 2.875, "step": 50860 }, { "epoch": 3.99, "learning_rate": 6.0118403512898926e-05, "loss": 2.6855, "step": 50861 }, { "epoch": 3.99, "learning_rate": 6.011761938367443e-05, "loss": 2.8966, "step": 50862 }, { "epoch": 3.99, "learning_rate": 6.011683525444993e-05, "loss": 2.5559, "step": 50863 }, { "epoch": 3.99, "learning_rate": 6.011605112522544e-05, "loss": 2.842, "step": 50864 }, { "epoch": 3.99, "learning_rate": 6.011526699600094e-05, "loss": 2.8676, "step": 50865 }, { "epoch": 3.99, "learning_rate": 6.0114482866776444e-05, "loss": 2.6436, "step": 50866 }, { "epoch": 3.99, "learning_rate": 6.011369873755195e-05, "loss": 2.7784, "step": 50867 }, { "epoch": 3.99, "learning_rate": 6.011291460832745e-05, "loss": 2.6613, "step": 50868 }, { "epoch": 3.99, "learning_rate": 6.0112130479102955e-05, "loss": 2.7202, "step": 50869 }, { "epoch": 3.99, "learning_rate": 6.011134634987846e-05, "loss": 2.7831, "step": 50870 }, { "epoch": 3.99, "learning_rate": 6.011056222065396e-05, "loss": 2.696, "step": 50871 }, { "epoch": 3.99, "learning_rate": 6.0109778091429466e-05, "loss": 2.7443, "step": 50872 }, { "epoch": 3.99, "learning_rate": 6.010899396220498e-05, "loss": 2.9491, "step": 50873 }, { "epoch": 3.99, "learning_rate": 6.010820983298048e-05, "loss": 2.648, "step": 50874 }, { "epoch": 3.99, "learning_rate": 6.0107425703755984e-05, "loss": 2.742, "step": 50875 }, { "epoch": 3.99, "learning_rate": 6.010664157453149e-05, "loss": 2.5808, "step": 50876 }, { "epoch": 3.99, "learning_rate": 6.010585744530699e-05, "loss": 2.8437, "step": 50877 }, { "epoch": 3.99, "learning_rate": 6.0105073316082495e-05, "loss": 2.8274, "step": 50878 }, { "epoch": 3.99, "learning_rate": 6.0104289186858e-05, "loss": 2.629, "step": 50879 }, { "epoch": 3.99, "learning_rate": 6.01035050576335e-05, "loss": 2.643, "step": 50880 }, { "epoch": 3.99, "learning_rate": 6.0102720928409006e-05, "loss": 2.6429, "step": 50881 }, { "epoch": 3.99, "learning_rate": 6.010193679918451e-05, "loss": 2.598, "step": 50882 }, { "epoch": 3.99, "learning_rate": 6.0101152669960014e-05, "loss": 2.7968, "step": 50883 }, { "epoch": 3.99, "learning_rate": 6.010036854073552e-05, "loss": 2.6782, "step": 50884 }, { "epoch": 3.99, "learning_rate": 6.009958441151102e-05, "loss": 2.7608, "step": 50885 }, { "epoch": 3.99, "learning_rate": 6.0098800282286525e-05, "loss": 2.8821, "step": 50886 }, { "epoch": 3.99, "learning_rate": 6.009801615306203e-05, "loss": 2.7872, "step": 50887 }, { "epoch": 3.99, "learning_rate": 6.009723202383753e-05, "loss": 2.6604, "step": 50888 }, { "epoch": 3.99, "learning_rate": 6.009644789461304e-05, "loss": 2.6333, "step": 50889 }, { "epoch": 3.99, "learning_rate": 6.0095663765388546e-05, "loss": 2.4969, "step": 50890 }, { "epoch": 3.99, "learning_rate": 6.009487963616405e-05, "loss": 2.714, "step": 50891 }, { "epoch": 3.99, "learning_rate": 6.0094095506939553e-05, "loss": 2.6434, "step": 50892 }, { "epoch": 3.99, "learning_rate": 6.009331137771506e-05, "loss": 2.7938, "step": 50893 }, { "epoch": 3.99, "learning_rate": 6.009252724849055e-05, "loss": 2.5047, "step": 50894 }, { "epoch": 3.99, "learning_rate": 6.009174311926605e-05, "loss": 2.5064, "step": 50895 }, { "epoch": 3.99, "learning_rate": 6.0090958990041555e-05, "loss": 2.6828, "step": 50896 }, { "epoch": 3.99, "learning_rate": 6.009017486081706e-05, "loss": 2.7147, "step": 50897 }, { "epoch": 3.99, "learning_rate": 6.008939073159256e-05, "loss": 2.9464, "step": 50898 }, { "epoch": 3.99, "learning_rate": 6.008860660236807e-05, "loss": 2.7038, "step": 50899 }, { "epoch": 3.99, "learning_rate": 6.0087822473143576e-05, "loss": 2.8073, "step": 50900 }, { "epoch": 3.99, "learning_rate": 6.008703834391908e-05, "loss": 2.6301, "step": 50901 }, { "epoch": 3.99, "learning_rate": 6.0086254214694583e-05, "loss": 2.7566, "step": 50902 }, { "epoch": 3.99, "learning_rate": 6.008547008547009e-05, "loss": 2.7582, "step": 50903 }, { "epoch": 3.99, "learning_rate": 6.008468595624559e-05, "loss": 2.6587, "step": 50904 }, { "epoch": 3.99, "learning_rate": 6.0083901827021094e-05, "loss": 2.9477, "step": 50905 }, { "epoch": 3.99, "learning_rate": 6.00831176977966e-05, "loss": 2.8364, "step": 50906 }, { "epoch": 3.99, "learning_rate": 6.00823335685721e-05, "loss": 2.6162, "step": 50907 }, { "epoch": 3.99, "learning_rate": 6.0081549439347605e-05, "loss": 2.6989, "step": 50908 }, { "epoch": 3.99, "learning_rate": 6.008076531012311e-05, "loss": 2.9893, "step": 50909 }, { "epoch": 3.99, "learning_rate": 6.007998118089861e-05, "loss": 2.8286, "step": 50910 }, { "epoch": 3.99, "learning_rate": 6.0079197051674117e-05, "loss": 2.6704, "step": 50911 }, { "epoch": 3.99, "learning_rate": 6.007841292244962e-05, "loss": 2.5638, "step": 50912 }, { "epoch": 3.99, "learning_rate": 6.0077628793225124e-05, "loss": 2.6006, "step": 50913 }, { "epoch": 3.99, "learning_rate": 6.0076844664000634e-05, "loss": 2.8966, "step": 50914 }, { "epoch": 3.99, "learning_rate": 6.007606053477614e-05, "loss": 2.8021, "step": 50915 }, { "epoch": 3.99, "learning_rate": 6.007527640555164e-05, "loss": 2.6203, "step": 50916 }, { "epoch": 3.99, "learning_rate": 6.0074492276327145e-05, "loss": 2.7358, "step": 50917 }, { "epoch": 3.99, "learning_rate": 6.007370814710265e-05, "loss": 2.6446, "step": 50918 }, { "epoch": 3.99, "learning_rate": 6.007292401787815e-05, "loss": 2.8326, "step": 50919 }, { "epoch": 3.99, "learning_rate": 6.0072139888653656e-05, "loss": 3.2375, "step": 50920 }, { "epoch": 3.99, "learning_rate": 6.007135575942916e-05, "loss": 2.6682, "step": 50921 }, { "epoch": 3.99, "learning_rate": 6.0070571630204664e-05, "loss": 2.9847, "step": 50922 }, { "epoch": 3.99, "learning_rate": 6.006978750098017e-05, "loss": 2.9603, "step": 50923 }, { "epoch": 3.99, "learning_rate": 6.006900337175567e-05, "loss": 2.6022, "step": 50924 }, { "epoch": 3.99, "learning_rate": 6.0068219242531175e-05, "loss": 2.6807, "step": 50925 }, { "epoch": 3.99, "learning_rate": 6.006743511330668e-05, "loss": 2.6268, "step": 50926 }, { "epoch": 3.99, "learning_rate": 6.006665098408218e-05, "loss": 2.8812, "step": 50927 }, { "epoch": 3.99, "learning_rate": 6.0065866854857686e-05, "loss": 2.7258, "step": 50928 }, { "epoch": 3.99, "learning_rate": 6.006508272563319e-05, "loss": 2.6964, "step": 50929 }, { "epoch": 3.99, "learning_rate": 6.0064298596408686e-05, "loss": 2.6745, "step": 50930 }, { "epoch": 3.99, "learning_rate": 6.006351446718419e-05, "loss": 2.8138, "step": 50931 }, { "epoch": 3.99, "learning_rate": 6.0062730337959694e-05, "loss": 2.909, "step": 50932 }, { "epoch": 3.99, "learning_rate": 6.00619462087352e-05, "loss": 2.8848, "step": 50933 }, { "epoch": 3.99, "learning_rate": 6.00611620795107e-05, "loss": 2.7105, "step": 50934 }, { "epoch": 3.99, "learning_rate": 6.0060377950286205e-05, "loss": 2.8168, "step": 50935 }, { "epoch": 3.99, "learning_rate": 6.005959382106171e-05, "loss": 2.686, "step": 50936 }, { "epoch": 3.99, "learning_rate": 6.005880969183721e-05, "loss": 2.7185, "step": 50937 }, { "epoch": 3.99, "learning_rate": 6.0058025562612716e-05, "loss": 2.7998, "step": 50938 }, { "epoch": 3.99, "learning_rate": 6.005724143338822e-05, "loss": 2.7088, "step": 50939 }, { "epoch": 3.99, "learning_rate": 6.005645730416373e-05, "loss": 2.7085, "step": 50940 }, { "epoch": 3.99, "learning_rate": 6.0055673174939234e-05, "loss": 2.674, "step": 50941 }, { "epoch": 3.99, "learning_rate": 6.005488904571474e-05, "loss": 2.8148, "step": 50942 }, { "epoch": 3.99, "learning_rate": 6.005410491649024e-05, "loss": 2.9829, "step": 50943 }, { "epoch": 3.99, "learning_rate": 6.0053320787265745e-05, "loss": 2.6589, "step": 50944 }, { "epoch": 3.99, "learning_rate": 6.005253665804125e-05, "loss": 2.8354, "step": 50945 }, { "epoch": 3.99, "learning_rate": 6.005175252881675e-05, "loss": 2.7524, "step": 50946 }, { "epoch": 3.99, "learning_rate": 6.0050968399592256e-05, "loss": 2.9708, "step": 50947 }, { "epoch": 3.99, "learning_rate": 6.005018427036776e-05, "loss": 2.6045, "step": 50948 }, { "epoch": 4.0, "learning_rate": 6.004940014114326e-05, "loss": 3.1907, "step": 50949 }, { "epoch": 4.0, "learning_rate": 6.004861601191877e-05, "loss": 2.6074, "step": 50950 }, { "epoch": 4.0, "learning_rate": 6.004783188269427e-05, "loss": 2.5483, "step": 50951 }, { "epoch": 4.0, "learning_rate": 6.0047047753469774e-05, "loss": 2.7087, "step": 50952 }, { "epoch": 4.0, "learning_rate": 6.004626362424528e-05, "loss": 2.6312, "step": 50953 }, { "epoch": 4.0, "learning_rate": 6.004547949502078e-05, "loss": 2.7316, "step": 50954 }, { "epoch": 4.0, "learning_rate": 6.004469536579629e-05, "loss": 2.6963, "step": 50955 }, { "epoch": 4.0, "learning_rate": 6.0043911236571796e-05, "loss": 2.7958, "step": 50956 }, { "epoch": 4.0, "learning_rate": 6.00431271073473e-05, "loss": 2.6674, "step": 50957 }, { "epoch": 4.0, "learning_rate": 6.00423429781228e-05, "loss": 3.1264, "step": 50958 }, { "epoch": 4.0, "learning_rate": 6.0041558848898307e-05, "loss": 2.7409, "step": 50959 }, { "epoch": 4.0, "learning_rate": 6.004077471967381e-05, "loss": 2.7137, "step": 50960 }, { "epoch": 4.0, "learning_rate": 6.0039990590449314e-05, "loss": 2.7308, "step": 50961 }, { "epoch": 4.0, "learning_rate": 6.003920646122482e-05, "loss": 2.8124, "step": 50962 }, { "epoch": 4.0, "learning_rate": 6.003842233200032e-05, "loss": 2.6952, "step": 50963 }, { "epoch": 4.0, "learning_rate": 6.0037638202775825e-05, "loss": 2.6076, "step": 50964 }, { "epoch": 4.0, "learning_rate": 6.003685407355133e-05, "loss": 2.586, "step": 50965 }, { "epoch": 4.0, "learning_rate": 6.0036069944326826e-05, "loss": 2.7757, "step": 50966 }, { "epoch": 4.0, "learning_rate": 6.003528581510233e-05, "loss": 2.5651, "step": 50967 }, { "epoch": 4.0, "learning_rate": 6.003450168587783e-05, "loss": 2.7172, "step": 50968 }, { "epoch": 4.0, "learning_rate": 6.0033717556653337e-05, "loss": 2.5759, "step": 50969 }, { "epoch": 4.0, "learning_rate": 6.003293342742884e-05, "loss": 2.7539, "step": 50970 }, { "epoch": 4.0, "learning_rate": 6.0032149298204344e-05, "loss": 2.794, "step": 50971 }, { "epoch": 4.0, "learning_rate": 6.003136516897985e-05, "loss": 2.6544, "step": 50972 }, { "epoch": 4.0, "learning_rate": 6.003058103975535e-05, "loss": 2.8418, "step": 50973 }, { "epoch": 4.0, "learning_rate": 6.0029796910530855e-05, "loss": 2.8213, "step": 50974 }, { "epoch": 4.0, "learning_rate": 6.002901278130636e-05, "loss": 2.8629, "step": 50975 }, { "epoch": 4.0, "learning_rate": 6.002822865208186e-05, "loss": 2.7556, "step": 50976 }, { "epoch": 4.0, "learning_rate": 6.0027444522857366e-05, "loss": 2.6018, "step": 50977 }, { "epoch": 4.0, "learning_rate": 6.002666039363287e-05, "loss": 2.821, "step": 50978 }, { "epoch": 4.0, "learning_rate": 6.002587626440837e-05, "loss": 2.7209, "step": 50979 }, { "epoch": 4.0, "learning_rate": 6.002509213518388e-05, "loss": 2.8027, "step": 50980 }, { "epoch": 4.0, "learning_rate": 6.002430800595939e-05, "loss": 2.6319, "step": 50981 }, { "epoch": 4.0, "learning_rate": 6.002352387673489e-05, "loss": 2.8527, "step": 50982 }, { "epoch": 4.0, "learning_rate": 6.0022739747510395e-05, "loss": 2.9151, "step": 50983 }, { "epoch": 4.0, "learning_rate": 6.00219556182859e-05, "loss": 2.7851, "step": 50984 }, { "epoch": 4.0, "learning_rate": 6.00211714890614e-05, "loss": 2.7797, "step": 50985 }, { "epoch": 4.0, "learning_rate": 6.0020387359836906e-05, "loss": 2.8081, "step": 50986 }, { "epoch": 4.0, "learning_rate": 6.001960323061241e-05, "loss": 2.653, "step": 50987 }, { "epoch": 4.0, "learning_rate": 6.001881910138791e-05, "loss": 2.7661, "step": 50988 }, { "epoch": 4.0, "learning_rate": 6.001803497216342e-05, "loss": 2.658, "step": 50989 }, { "epoch": 4.0, "learning_rate": 6.001725084293892e-05, "loss": 2.7365, "step": 50990 }, { "epoch": 4.0, "learning_rate": 6.0016466713714424e-05, "loss": 2.7321, "step": 50991 }, { "epoch": 4.0, "learning_rate": 6.001568258448993e-05, "loss": 2.702, "step": 50992 }, { "epoch": 4.0, "learning_rate": 6.001489845526543e-05, "loss": 2.6819, "step": 50993 }, { "epoch": 4.0, "learning_rate": 6.0014114326040935e-05, "loss": 2.7115, "step": 50994 }, { "epoch": 4.0, "learning_rate": 6.001333019681644e-05, "loss": 2.9129, "step": 50995 }, { "epoch": 4.0, "learning_rate": 6.001254606759194e-05, "loss": 2.6402, "step": 50996 }, { "epoch": 4.0, "learning_rate": 6.001176193836745e-05, "loss": 2.5669, "step": 50997 }, { "epoch": 4.0, "learning_rate": 6.001097780914296e-05, "loss": 2.6363, "step": 50998 }, { "epoch": 4.0, "learning_rate": 6.001019367991846e-05, "loss": 2.811, "step": 50999 }, { "epoch": 4.0, "learning_rate": 6.0009409550693964e-05, "loss": 2.737, "step": 51000 }, { "epoch": 4.0, "eval_loss": 2.8893203735351562, "eval_runtime": 217.9299, "eval_samples_per_second": 51.998, "eval_steps_per_second": 3.253, "step": 51000 }, { "epoch": 4.0, "learning_rate": 6.000862542146947e-05, "loss": 2.8528, "step": 51001 }, { "epoch": 4.0, "learning_rate": 6.000784129224496e-05, "loss": 2.9407, "step": 51002 }, { "epoch": 4.0, "learning_rate": 6.000705716302046e-05, "loss": 2.7071, "step": 51003 }, { "epoch": 4.0, "learning_rate": 6.0006273033795965e-05, "loss": 2.7554, "step": 51004 }, { "epoch": 4.0, "learning_rate": 6.000548890457147e-05, "loss": 2.6551, "step": 51005 }, { "epoch": 4.0, "learning_rate": 6.000470477534697e-05, "loss": 2.8037, "step": 51006 }, { "epoch": 4.0, "learning_rate": 6.000392064612248e-05, "loss": 2.8033, "step": 51007 }, { "epoch": 4.0, "learning_rate": 6.000313651689799e-05, "loss": 2.6906, "step": 51008 }, { "epoch": 4.0, "learning_rate": 6.000235238767349e-05, "loss": 2.6404, "step": 51009 }, { "epoch": 4.0, "learning_rate": 6.0001568258448994e-05, "loss": 2.6755, "step": 51010 }, { "epoch": 4.0, "learning_rate": 6.00007841292245e-05, "loss": 2.7837, "step": 51011 }, { "epoch": 4.0, "learning_rate": 6e-05, "loss": 2.8095, "step": 51012 }, { "epoch": 4.0, "learning_rate": 5.9999215870775505e-05, "loss": 2.662, "step": 51013 }, { "epoch": 4.0, "learning_rate": 5.999843174155101e-05, "loss": 2.7289, "step": 51014 }, { "epoch": 4.0, "learning_rate": 5.999764761232651e-05, "loss": 2.6257, "step": 51015 }, { "epoch": 4.0, "learning_rate": 5.9996863483102016e-05, "loss": 2.6622, "step": 51016 }, { "epoch": 4.0, "learning_rate": 5.999607935387752e-05, "loss": 2.5658, "step": 51017 }, { "epoch": 4.0, "learning_rate": 5.9995295224653024e-05, "loss": 2.8587, "step": 51018 }, { "epoch": 4.0, "learning_rate": 5.999451109542853e-05, "loss": 2.6908, "step": 51019 }, { "epoch": 4.0, "learning_rate": 5.999372696620403e-05, "loss": 2.5653, "step": 51020 }, { "epoch": 4.0, "learning_rate": 5.9992942836979535e-05, "loss": 2.5827, "step": 51021 }, { "epoch": 4.0, "learning_rate": 5.9992158707755045e-05, "loss": 2.6825, "step": 51022 }, { "epoch": 4.0, "learning_rate": 5.999137457853055e-05, "loss": 2.6348, "step": 51023 }, { "epoch": 4.0, "learning_rate": 5.999059044930605e-05, "loss": 2.6472, "step": 51024 }, { "epoch": 4.0, "learning_rate": 5.9989806320081556e-05, "loss": 2.4987, "step": 51025 }, { "epoch": 4.0, "learning_rate": 5.998902219085706e-05, "loss": 2.5241, "step": 51026 }, { "epoch": 4.0, "learning_rate": 5.9988238061632563e-05, "loss": 2.5901, "step": 51027 }, { "epoch": 4.0, "learning_rate": 5.998745393240807e-05, "loss": 2.6104, "step": 51028 }, { "epoch": 4.0, "learning_rate": 5.998666980318357e-05, "loss": 2.5576, "step": 51029 }, { "epoch": 4.0, "learning_rate": 5.9985885673959074e-05, "loss": 2.4344, "step": 51030 }, { "epoch": 4.0, "learning_rate": 5.998510154473458e-05, "loss": 2.7918, "step": 51031 }, { "epoch": 4.0, "learning_rate": 5.998431741551008e-05, "loss": 2.5507, "step": 51032 }, { "epoch": 4.0, "learning_rate": 5.9983533286285586e-05, "loss": 2.8359, "step": 51033 }, { "epoch": 4.0, "learning_rate": 5.998274915706109e-05, "loss": 2.6686, "step": 51034 }, { "epoch": 4.0, "learning_rate": 5.998196502783659e-05, "loss": 2.6929, "step": 51035 }, { "epoch": 4.0, "learning_rate": 5.9981180898612097e-05, "loss": 2.6143, "step": 51036 }, { "epoch": 4.0, "learning_rate": 5.99803967693876e-05, "loss": 2.7302, "step": 51037 }, { "epoch": 4.0, "learning_rate": 5.99796126401631e-05, "loss": 2.8126, "step": 51038 }, { "epoch": 4.0, "learning_rate": 5.99788285109386e-05, "loss": 2.6949, "step": 51039 }, { "epoch": 4.0, "learning_rate": 5.9978044381714104e-05, "loss": 2.4869, "step": 51040 }, { "epoch": 4.0, "learning_rate": 5.997726025248961e-05, "loss": 2.6564, "step": 51041 }, { "epoch": 4.0, "learning_rate": 5.997647612326511e-05, "loss": 2.5702, "step": 51042 }, { "epoch": 4.0, "learning_rate": 5.9975691994040616e-05, "loss": 2.4438, "step": 51043 }, { "epoch": 4.0, "learning_rate": 5.997490786481612e-05, "loss": 2.5174, "step": 51044 }, { "epoch": 4.0, "learning_rate": 5.997412373559162e-05, "loss": 2.7749, "step": 51045 }, { "epoch": 4.0, "learning_rate": 5.9973339606367127e-05, "loss": 2.3917, "step": 51046 }, { "epoch": 4.0, "learning_rate": 5.997255547714263e-05, "loss": 2.6614, "step": 51047 }, { "epoch": 4.0, "learning_rate": 5.997177134791814e-05, "loss": 2.5359, "step": 51048 }, { "epoch": 4.0, "learning_rate": 5.9970987218693644e-05, "loss": 2.5626, "step": 51049 }, { "epoch": 4.0, "learning_rate": 5.997020308946915e-05, "loss": 2.5935, "step": 51050 }, { "epoch": 4.0, "learning_rate": 5.996941896024465e-05, "loss": 2.7274, "step": 51051 }, { "epoch": 4.0, "learning_rate": 5.9968634831020155e-05, "loss": 2.6865, "step": 51052 }, { "epoch": 4.0, "learning_rate": 5.996785070179566e-05, "loss": 2.8483, "step": 51053 }, { "epoch": 4.0, "learning_rate": 5.996706657257116e-05, "loss": 2.6137, "step": 51054 }, { "epoch": 4.0, "learning_rate": 5.9966282443346666e-05, "loss": 2.6954, "step": 51055 }, { "epoch": 4.0, "learning_rate": 5.996549831412217e-05, "loss": 2.5943, "step": 51056 }, { "epoch": 4.0, "learning_rate": 5.9964714184897674e-05, "loss": 2.525, "step": 51057 }, { "epoch": 4.0, "learning_rate": 5.996393005567318e-05, "loss": 2.5648, "step": 51058 }, { "epoch": 4.0, "learning_rate": 5.996314592644868e-05, "loss": 2.6246, "step": 51059 }, { "epoch": 4.0, "learning_rate": 5.9962361797224185e-05, "loss": 2.7527, "step": 51060 }, { "epoch": 4.0, "learning_rate": 5.996157766799969e-05, "loss": 2.6645, "step": 51061 }, { "epoch": 4.0, "learning_rate": 5.996079353877519e-05, "loss": 2.6655, "step": 51062 }, { "epoch": 4.0, "learning_rate": 5.9960009409550696e-05, "loss": 2.7334, "step": 51063 }, { "epoch": 4.0, "learning_rate": 5.9959225280326206e-05, "loss": 2.4883, "step": 51064 }, { "epoch": 4.0, "learning_rate": 5.995844115110171e-05, "loss": 2.6788, "step": 51065 }, { "epoch": 4.0, "learning_rate": 5.9957657021877214e-05, "loss": 2.5689, "step": 51066 }, { "epoch": 4.0, "learning_rate": 5.995687289265272e-05, "loss": 2.7817, "step": 51067 }, { "epoch": 4.0, "learning_rate": 5.995608876342822e-05, "loss": 2.5797, "step": 51068 }, { "epoch": 4.0, "learning_rate": 5.9955304634203725e-05, "loss": 2.7781, "step": 51069 }, { "epoch": 4.0, "learning_rate": 5.995452050497923e-05, "loss": 2.5237, "step": 51070 }, { "epoch": 4.0, "learning_rate": 5.995373637575473e-05, "loss": 2.7023, "step": 51071 }, { "epoch": 4.0, "learning_rate": 5.9952952246530236e-05, "loss": 2.8218, "step": 51072 }, { "epoch": 4.0, "learning_rate": 5.995216811730574e-05, "loss": 2.5935, "step": 51073 }, { "epoch": 4.0, "learning_rate": 5.9951383988081236e-05, "loss": 2.494, "step": 51074 }, { "epoch": 4.0, "learning_rate": 5.995059985885674e-05, "loss": 2.7025, "step": 51075 }, { "epoch": 4.01, "learning_rate": 5.9949815729632244e-05, "loss": 2.7496, "step": 51076 }, { "epoch": 4.01, "learning_rate": 5.994903160040775e-05, "loss": 2.7987, "step": 51077 }, { "epoch": 4.01, "learning_rate": 5.994824747118325e-05, "loss": 2.5698, "step": 51078 }, { "epoch": 4.01, "learning_rate": 5.9947463341958755e-05, "loss": 2.8205, "step": 51079 }, { "epoch": 4.01, "learning_rate": 5.994667921273426e-05, "loss": 2.6184, "step": 51080 }, { "epoch": 4.01, "learning_rate": 5.994589508350976e-05, "loss": 2.7491, "step": 51081 }, { "epoch": 4.01, "learning_rate": 5.9945110954285266e-05, "loss": 2.5518, "step": 51082 }, { "epoch": 4.01, "learning_rate": 5.994432682506077e-05, "loss": 2.7244, "step": 51083 }, { "epoch": 4.01, "learning_rate": 5.994354269583627e-05, "loss": 2.4989, "step": 51084 }, { "epoch": 4.01, "learning_rate": 5.994275856661178e-05, "loss": 2.5792, "step": 51085 }, { "epoch": 4.01, "learning_rate": 5.994197443738728e-05, "loss": 2.7719, "step": 51086 }, { "epoch": 4.01, "learning_rate": 5.9941190308162784e-05, "loss": 2.4693, "step": 51087 }, { "epoch": 4.01, "learning_rate": 5.994040617893829e-05, "loss": 2.627, "step": 51088 }, { "epoch": 4.01, "learning_rate": 5.99396220497138e-05, "loss": 2.4341, "step": 51089 }, { "epoch": 4.01, "learning_rate": 5.99388379204893e-05, "loss": 2.5282, "step": 51090 }, { "epoch": 4.01, "learning_rate": 5.9938053791264806e-05, "loss": 2.5739, "step": 51091 }, { "epoch": 4.01, "learning_rate": 5.993726966204031e-05, "loss": 2.7725, "step": 51092 }, { "epoch": 4.01, "learning_rate": 5.993648553281581e-05, "loss": 2.8038, "step": 51093 }, { "epoch": 4.01, "learning_rate": 5.993570140359132e-05, "loss": 2.6466, "step": 51094 }, { "epoch": 4.01, "learning_rate": 5.993491727436682e-05, "loss": 2.7295, "step": 51095 }, { "epoch": 4.01, "learning_rate": 5.9934133145142324e-05, "loss": 2.7608, "step": 51096 }, { "epoch": 4.01, "learning_rate": 5.993334901591783e-05, "loss": 2.7815, "step": 51097 }, { "epoch": 4.01, "learning_rate": 5.993256488669333e-05, "loss": 2.7331, "step": 51098 }, { "epoch": 4.01, "learning_rate": 5.9931780757468835e-05, "loss": 2.5402, "step": 51099 }, { "epoch": 4.01, "learning_rate": 5.993099662824434e-05, "loss": 2.6372, "step": 51100 }, { "epoch": 4.01, "learning_rate": 5.993021249901984e-05, "loss": 2.8013, "step": 51101 }, { "epoch": 4.01, "learning_rate": 5.9929428369795346e-05, "loss": 2.626, "step": 51102 }, { "epoch": 4.01, "learning_rate": 5.992864424057085e-05, "loss": 2.6411, "step": 51103 }, { "epoch": 4.01, "learning_rate": 5.9927860111346353e-05, "loss": 2.765, "step": 51104 }, { "epoch": 4.01, "learning_rate": 5.9927075982121864e-05, "loss": 2.5669, "step": 51105 }, { "epoch": 4.01, "learning_rate": 5.992629185289737e-05, "loss": 2.6625, "step": 51106 }, { "epoch": 4.01, "learning_rate": 5.992550772367287e-05, "loss": 2.6307, "step": 51107 }, { "epoch": 4.01, "learning_rate": 5.9924723594448375e-05, "loss": 2.7137, "step": 51108 }, { "epoch": 4.01, "learning_rate": 5.992393946522388e-05, "loss": 2.6159, "step": 51109 }, { "epoch": 4.01, "learning_rate": 5.992315533599937e-05, "loss": 2.4663, "step": 51110 }, { "epoch": 4.01, "learning_rate": 5.992237120677487e-05, "loss": 2.7328, "step": 51111 }, { "epoch": 4.01, "learning_rate": 5.9921587077550376e-05, "loss": 2.7016, "step": 51112 }, { "epoch": 4.01, "learning_rate": 5.992080294832588e-05, "loss": 2.8119, "step": 51113 }, { "epoch": 4.01, "learning_rate": 5.9920018819101383e-05, "loss": 2.5975, "step": 51114 }, { "epoch": 4.01, "learning_rate": 5.9919234689876894e-05, "loss": 2.6878, "step": 51115 }, { "epoch": 4.01, "learning_rate": 5.99184505606524e-05, "loss": 2.7079, "step": 51116 }, { "epoch": 4.01, "learning_rate": 5.99176664314279e-05, "loss": 2.5153, "step": 51117 }, { "epoch": 4.01, "learning_rate": 5.9916882302203405e-05, "loss": 2.5401, "step": 51118 }, { "epoch": 4.01, "learning_rate": 5.991609817297891e-05, "loss": 2.7079, "step": 51119 }, { "epoch": 4.01, "learning_rate": 5.991531404375441e-05, "loss": 2.6622, "step": 51120 }, { "epoch": 4.01, "learning_rate": 5.9914529914529916e-05, "loss": 2.8075, "step": 51121 }, { "epoch": 4.01, "learning_rate": 5.991374578530542e-05, "loss": 2.7491, "step": 51122 }, { "epoch": 4.01, "learning_rate": 5.991296165608092e-05, "loss": 2.6683, "step": 51123 }, { "epoch": 4.01, "learning_rate": 5.991217752685643e-05, "loss": 2.6749, "step": 51124 }, { "epoch": 4.01, "learning_rate": 5.991139339763193e-05, "loss": 2.5575, "step": 51125 }, { "epoch": 4.01, "learning_rate": 5.9910609268407434e-05, "loss": 2.7043, "step": 51126 }, { "epoch": 4.01, "learning_rate": 5.990982513918294e-05, "loss": 2.7746, "step": 51127 }, { "epoch": 4.01, "learning_rate": 5.990904100995844e-05, "loss": 2.5097, "step": 51128 }, { "epoch": 4.01, "learning_rate": 5.9908256880733945e-05, "loss": 2.8047, "step": 51129 }, { "epoch": 4.01, "learning_rate": 5.9907472751509456e-05, "loss": 2.8733, "step": 51130 }, { "epoch": 4.01, "learning_rate": 5.990668862228496e-05, "loss": 2.5136, "step": 51131 }, { "epoch": 4.01, "learning_rate": 5.990590449306046e-05, "loss": 2.5682, "step": 51132 }, { "epoch": 4.01, "learning_rate": 5.990512036383597e-05, "loss": 2.6932, "step": 51133 }, { "epoch": 4.01, "learning_rate": 5.990433623461147e-05, "loss": 2.7697, "step": 51134 }, { "epoch": 4.01, "learning_rate": 5.9903552105386974e-05, "loss": 2.6281, "step": 51135 }, { "epoch": 4.01, "learning_rate": 5.990276797616248e-05, "loss": 2.6993, "step": 51136 }, { "epoch": 4.01, "learning_rate": 5.990198384693798e-05, "loss": 2.6296, "step": 51137 }, { "epoch": 4.01, "learning_rate": 5.9901199717713485e-05, "loss": 2.7711, "step": 51138 }, { "epoch": 4.01, "learning_rate": 5.990041558848899e-05, "loss": 2.4733, "step": 51139 }, { "epoch": 4.01, "learning_rate": 5.989963145926449e-05, "loss": 2.958, "step": 51140 }, { "epoch": 4.01, "learning_rate": 5.9898847330039996e-05, "loss": 2.7896, "step": 51141 }, { "epoch": 4.01, "learning_rate": 5.98980632008155e-05, "loss": 2.5688, "step": 51142 }, { "epoch": 4.01, "learning_rate": 5.9897279071591004e-05, "loss": 2.4713, "step": 51143 }, { "epoch": 4.01, "learning_rate": 5.989649494236651e-05, "loss": 2.4709, "step": 51144 }, { "epoch": 4.01, "learning_rate": 5.989571081314201e-05, "loss": 2.6297, "step": 51145 }, { "epoch": 4.01, "learning_rate": 5.989492668391751e-05, "loss": 2.6171, "step": 51146 }, { "epoch": 4.01, "learning_rate": 5.989414255469301e-05, "loss": 2.6909, "step": 51147 }, { "epoch": 4.01, "learning_rate": 5.9893358425468515e-05, "loss": 2.5403, "step": 51148 }, { "epoch": 4.01, "learning_rate": 5.989257429624402e-05, "loss": 2.556, "step": 51149 }, { "epoch": 4.01, "learning_rate": 5.989179016701952e-05, "loss": 2.9345, "step": 51150 }, { "epoch": 4.01, "learning_rate": 5.9891006037795026e-05, "loss": 2.6418, "step": 51151 }, { "epoch": 4.01, "learning_rate": 5.989022190857053e-05, "loss": 2.563, "step": 51152 }, { "epoch": 4.01, "learning_rate": 5.9889437779346034e-05, "loss": 2.9462, "step": 51153 }, { "epoch": 4.01, "learning_rate": 5.988865365012154e-05, "loss": 2.4614, "step": 51154 }, { "epoch": 4.01, "learning_rate": 5.988786952089704e-05, "loss": 2.8242, "step": 51155 }, { "epoch": 4.01, "learning_rate": 5.988708539167255e-05, "loss": 2.629, "step": 51156 }, { "epoch": 4.01, "learning_rate": 5.9886301262448055e-05, "loss": 2.5661, "step": 51157 }, { "epoch": 4.01, "learning_rate": 5.988551713322356e-05, "loss": 2.7309, "step": 51158 }, { "epoch": 4.01, "learning_rate": 5.988473300399906e-05, "loss": 2.5822, "step": 51159 }, { "epoch": 4.01, "learning_rate": 5.9883948874774566e-05, "loss": 2.5796, "step": 51160 }, { "epoch": 4.01, "learning_rate": 5.988316474555007e-05, "loss": 2.6442, "step": 51161 }, { "epoch": 4.01, "learning_rate": 5.9882380616325574e-05, "loss": 2.5171, "step": 51162 }, { "epoch": 4.01, "learning_rate": 5.988159648710108e-05, "loss": 2.7105, "step": 51163 }, { "epoch": 4.01, "learning_rate": 5.988081235787658e-05, "loss": 2.5435, "step": 51164 }, { "epoch": 4.01, "learning_rate": 5.9880028228652085e-05, "loss": 2.7636, "step": 51165 }, { "epoch": 4.01, "learning_rate": 5.987924409942759e-05, "loss": 2.4049, "step": 51166 }, { "epoch": 4.01, "learning_rate": 5.987845997020309e-05, "loss": 2.7276, "step": 51167 }, { "epoch": 4.01, "learning_rate": 5.9877675840978596e-05, "loss": 2.754, "step": 51168 }, { "epoch": 4.01, "learning_rate": 5.98768917117541e-05, "loss": 2.4208, "step": 51169 }, { "epoch": 4.01, "learning_rate": 5.98761075825296e-05, "loss": 2.7006, "step": 51170 }, { "epoch": 4.01, "learning_rate": 5.9875323453305107e-05, "loss": 2.5408, "step": 51171 }, { "epoch": 4.01, "learning_rate": 5.987453932408062e-05, "loss": 2.7982, "step": 51172 }, { "epoch": 4.01, "learning_rate": 5.987375519485612e-05, "loss": 2.6797, "step": 51173 }, { "epoch": 4.01, "learning_rate": 5.9872971065631624e-05, "loss": 2.6473, "step": 51174 }, { "epoch": 4.01, "learning_rate": 5.987218693640713e-05, "loss": 2.6675, "step": 51175 }, { "epoch": 4.01, "learning_rate": 5.987140280718263e-05, "loss": 2.6055, "step": 51176 }, { "epoch": 4.01, "learning_rate": 5.9870618677958135e-05, "loss": 2.6877, "step": 51177 }, { "epoch": 4.01, "learning_rate": 5.986983454873364e-05, "loss": 2.6285, "step": 51178 }, { "epoch": 4.01, "learning_rate": 5.986905041950914e-05, "loss": 2.6663, "step": 51179 }, { "epoch": 4.01, "learning_rate": 5.9868266290284646e-05, "loss": 2.5755, "step": 51180 }, { "epoch": 4.01, "learning_rate": 5.986748216106015e-05, "loss": 2.7931, "step": 51181 }, { "epoch": 4.01, "learning_rate": 5.986669803183565e-05, "loss": 2.4741, "step": 51182 }, { "epoch": 4.01, "learning_rate": 5.986591390261115e-05, "loss": 2.4925, "step": 51183 }, { "epoch": 4.01, "learning_rate": 5.9865129773386654e-05, "loss": 2.7646, "step": 51184 }, { "epoch": 4.01, "learning_rate": 5.986434564416216e-05, "loss": 2.6926, "step": 51185 }, { "epoch": 4.01, "learning_rate": 5.986356151493766e-05, "loss": 2.748, "step": 51186 }, { "epoch": 4.01, "learning_rate": 5.9862777385713165e-05, "loss": 2.6203, "step": 51187 }, { "epoch": 4.01, "learning_rate": 5.986199325648867e-05, "loss": 2.7573, "step": 51188 }, { "epoch": 4.01, "learning_rate": 5.986120912726417e-05, "loss": 2.7555, "step": 51189 }, { "epoch": 4.01, "learning_rate": 5.9860424998039676e-05, "loss": 2.6027, "step": 51190 }, { "epoch": 4.01, "learning_rate": 5.985964086881518e-05, "loss": 2.631, "step": 51191 }, { "epoch": 4.01, "learning_rate": 5.9858856739590684e-05, "loss": 2.5345, "step": 51192 }, { "epoch": 4.01, "learning_rate": 5.985807261036619e-05, "loss": 2.7162, "step": 51193 }, { "epoch": 4.01, "learning_rate": 5.985728848114169e-05, "loss": 2.6674, "step": 51194 }, { "epoch": 4.01, "learning_rate": 5.9856504351917195e-05, "loss": 2.6538, "step": 51195 }, { "epoch": 4.01, "learning_rate": 5.98557202226927e-05, "loss": 2.5664, "step": 51196 }, { "epoch": 4.01, "learning_rate": 5.985493609346821e-05, "loss": 2.541, "step": 51197 }, { "epoch": 4.01, "learning_rate": 5.985415196424371e-05, "loss": 2.8075, "step": 51198 }, { "epoch": 4.01, "learning_rate": 5.9853367835019216e-05, "loss": 2.6595, "step": 51199 }, { "epoch": 4.01, "learning_rate": 5.985258370579472e-05, "loss": 2.3797, "step": 51200 }, { "epoch": 4.01, "learning_rate": 5.9851799576570224e-05, "loss": 2.6661, "step": 51201 }, { "epoch": 4.01, "learning_rate": 5.985101544734573e-05, "loss": 2.8416, "step": 51202 }, { "epoch": 4.01, "learning_rate": 5.985023131812123e-05, "loss": 2.7584, "step": 51203 }, { "epoch": 4.02, "learning_rate": 5.9849447188896735e-05, "loss": 2.599, "step": 51204 }, { "epoch": 4.02, "learning_rate": 5.984866305967224e-05, "loss": 2.7089, "step": 51205 }, { "epoch": 4.02, "learning_rate": 5.984787893044774e-05, "loss": 2.5422, "step": 51206 }, { "epoch": 4.02, "learning_rate": 5.9847094801223246e-05, "loss": 2.6262, "step": 51207 }, { "epoch": 4.02, "learning_rate": 5.984631067199875e-05, "loss": 2.7056, "step": 51208 }, { "epoch": 4.02, "learning_rate": 5.984552654277425e-05, "loss": 2.5538, "step": 51209 }, { "epoch": 4.02, "learning_rate": 5.984474241354976e-05, "loss": 2.807, "step": 51210 }, { "epoch": 4.02, "learning_rate": 5.984395828432526e-05, "loss": 2.8314, "step": 51211 }, { "epoch": 4.02, "learning_rate": 5.9843174155100764e-05, "loss": 2.7019, "step": 51212 }, { "epoch": 4.02, "learning_rate": 5.9842390025876275e-05, "loss": 2.6741, "step": 51213 }, { "epoch": 4.02, "learning_rate": 5.984160589665178e-05, "loss": 2.5661, "step": 51214 }, { "epoch": 4.02, "learning_rate": 5.984082176742728e-05, "loss": 2.6289, "step": 51215 }, { "epoch": 4.02, "learning_rate": 5.9840037638202786e-05, "loss": 2.6549, "step": 51216 }, { "epoch": 4.02, "learning_rate": 5.983925350897829e-05, "loss": 2.745, "step": 51217 }, { "epoch": 4.02, "learning_rate": 5.983846937975378e-05, "loss": 2.5099, "step": 51218 }, { "epoch": 4.02, "learning_rate": 5.983768525052928e-05, "loss": 2.6382, "step": 51219 }, { "epoch": 4.02, "learning_rate": 5.983690112130479e-05, "loss": 2.7613, "step": 51220 }, { "epoch": 4.02, "learning_rate": 5.983611699208029e-05, "loss": 2.598, "step": 51221 }, { "epoch": 4.02, "learning_rate": 5.9835332862855794e-05, "loss": 2.6092, "step": 51222 }, { "epoch": 4.02, "learning_rate": 5.9834548733631305e-05, "loss": 2.6151, "step": 51223 }, { "epoch": 4.02, "learning_rate": 5.983376460440681e-05, "loss": 2.678, "step": 51224 }, { "epoch": 4.02, "learning_rate": 5.983298047518231e-05, "loss": 2.679, "step": 51225 }, { "epoch": 4.02, "learning_rate": 5.9832196345957816e-05, "loss": 2.7312, "step": 51226 }, { "epoch": 4.02, "learning_rate": 5.983141221673332e-05, "loss": 2.5031, "step": 51227 }, { "epoch": 4.02, "learning_rate": 5.983062808750882e-05, "loss": 2.754, "step": 51228 }, { "epoch": 4.02, "learning_rate": 5.982984395828433e-05, "loss": 2.5363, "step": 51229 }, { "epoch": 4.02, "learning_rate": 5.982905982905983e-05, "loss": 2.6654, "step": 51230 }, { "epoch": 4.02, "learning_rate": 5.9828275699835334e-05, "loss": 2.5011, "step": 51231 }, { "epoch": 4.02, "learning_rate": 5.982749157061084e-05, "loss": 2.5679, "step": 51232 }, { "epoch": 4.02, "learning_rate": 5.982670744138634e-05, "loss": 2.6406, "step": 51233 }, { "epoch": 4.02, "learning_rate": 5.9825923312161845e-05, "loss": 2.653, "step": 51234 }, { "epoch": 4.02, "learning_rate": 5.982513918293735e-05, "loss": 2.4882, "step": 51235 }, { "epoch": 4.02, "learning_rate": 5.982435505371285e-05, "loss": 2.7274, "step": 51236 }, { "epoch": 4.02, "learning_rate": 5.9823570924488356e-05, "loss": 2.7147, "step": 51237 }, { "epoch": 4.02, "learning_rate": 5.9822786795263867e-05, "loss": 2.6905, "step": 51238 }, { "epoch": 4.02, "learning_rate": 5.982200266603937e-05, "loss": 2.6383, "step": 51239 }, { "epoch": 4.02, "learning_rate": 5.9821218536814874e-05, "loss": 2.5567, "step": 51240 }, { "epoch": 4.02, "learning_rate": 5.982043440759038e-05, "loss": 2.6217, "step": 51241 }, { "epoch": 4.02, "learning_rate": 5.981965027836588e-05, "loss": 2.8619, "step": 51242 }, { "epoch": 4.02, "learning_rate": 5.9818866149141385e-05, "loss": 2.5508, "step": 51243 }, { "epoch": 4.02, "learning_rate": 5.981808201991689e-05, "loss": 2.6037, "step": 51244 }, { "epoch": 4.02, "learning_rate": 5.981729789069239e-05, "loss": 2.6186, "step": 51245 }, { "epoch": 4.02, "learning_rate": 5.9816513761467896e-05, "loss": 2.6489, "step": 51246 }, { "epoch": 4.02, "learning_rate": 5.98157296322434e-05, "loss": 2.4927, "step": 51247 }, { "epoch": 4.02, "learning_rate": 5.98149455030189e-05, "loss": 2.7843, "step": 51248 }, { "epoch": 4.02, "learning_rate": 5.981416137379441e-05, "loss": 2.6553, "step": 51249 }, { "epoch": 4.02, "learning_rate": 5.981337724456991e-05, "loss": 2.6517, "step": 51250 }, { "epoch": 4.02, "learning_rate": 5.9812593115345414e-05, "loss": 2.5818, "step": 51251 }, { "epoch": 4.02, "learning_rate": 5.981180898612092e-05, "loss": 2.6596, "step": 51252 }, { "epoch": 4.02, "learning_rate": 5.981102485689642e-05, "loss": 2.553, "step": 51253 }, { "epoch": 4.02, "learning_rate": 5.981024072767192e-05, "loss": 2.5396, "step": 51254 }, { "epoch": 4.02, "learning_rate": 5.980945659844742e-05, "loss": 2.6395, "step": 51255 }, { "epoch": 4.02, "learning_rate": 5.9808672469222926e-05, "loss": 2.4283, "step": 51256 }, { "epoch": 4.02, "learning_rate": 5.980788833999843e-05, "loss": 2.6825, "step": 51257 }, { "epoch": 4.02, "learning_rate": 5.980710421077393e-05, "loss": 2.7493, "step": 51258 }, { "epoch": 4.02, "learning_rate": 5.980632008154944e-05, "loss": 2.6375, "step": 51259 }, { "epoch": 4.02, "learning_rate": 5.980553595232494e-05, "loss": 2.7425, "step": 51260 }, { "epoch": 4.02, "learning_rate": 5.9804751823100444e-05, "loss": 2.612, "step": 51261 }, { "epoch": 4.02, "learning_rate": 5.980396769387595e-05, "loss": 2.5115, "step": 51262 }, { "epoch": 4.02, "learning_rate": 5.980318356465145e-05, "loss": 2.5016, "step": 51263 }, { "epoch": 4.02, "learning_rate": 5.980239943542696e-05, "loss": 2.5509, "step": 51264 }, { "epoch": 4.02, "learning_rate": 5.9801615306202466e-05, "loss": 2.6064, "step": 51265 }, { "epoch": 4.02, "learning_rate": 5.980083117697797e-05, "loss": 2.6505, "step": 51266 }, { "epoch": 4.02, "learning_rate": 5.980004704775347e-05, "loss": 2.7453, "step": 51267 }, { "epoch": 4.02, "learning_rate": 5.979926291852898e-05, "loss": 2.5686, "step": 51268 }, { "epoch": 4.02, "learning_rate": 5.979847878930448e-05, "loss": 2.8002, "step": 51269 }, { "epoch": 4.02, "learning_rate": 5.9797694660079984e-05, "loss": 2.6618, "step": 51270 }, { "epoch": 4.02, "learning_rate": 5.979691053085549e-05, "loss": 2.8423, "step": 51271 }, { "epoch": 4.02, "learning_rate": 5.979612640163099e-05, "loss": 2.7007, "step": 51272 }, { "epoch": 4.02, "learning_rate": 5.9795342272406495e-05, "loss": 2.6602, "step": 51273 }, { "epoch": 4.02, "learning_rate": 5.9794558143182e-05, "loss": 2.906, "step": 51274 }, { "epoch": 4.02, "learning_rate": 5.97937740139575e-05, "loss": 2.7737, "step": 51275 }, { "epoch": 4.02, "learning_rate": 5.9792989884733006e-05, "loss": 2.851, "step": 51276 }, { "epoch": 4.02, "learning_rate": 5.979220575550851e-05, "loss": 2.6541, "step": 51277 }, { "epoch": 4.02, "learning_rate": 5.9791421626284014e-05, "loss": 2.7262, "step": 51278 }, { "epoch": 4.02, "learning_rate": 5.979063749705952e-05, "loss": 2.5104, "step": 51279 }, { "epoch": 4.02, "learning_rate": 5.978985336783503e-05, "loss": 2.4627, "step": 51280 }, { "epoch": 4.02, "learning_rate": 5.978906923861053e-05, "loss": 2.8777, "step": 51281 }, { "epoch": 4.02, "learning_rate": 5.9788285109386035e-05, "loss": 2.6784, "step": 51282 }, { "epoch": 4.02, "learning_rate": 5.978750098016154e-05, "loss": 2.6096, "step": 51283 }, { "epoch": 4.02, "learning_rate": 5.978671685093704e-05, "loss": 2.6416, "step": 51284 }, { "epoch": 4.02, "learning_rate": 5.9785932721712546e-05, "loss": 2.5628, "step": 51285 }, { "epoch": 4.02, "learning_rate": 5.978514859248805e-05, "loss": 2.6779, "step": 51286 }, { "epoch": 4.02, "learning_rate": 5.9784364463263554e-05, "loss": 2.7975, "step": 51287 }, { "epoch": 4.02, "learning_rate": 5.978358033403906e-05, "loss": 2.7524, "step": 51288 }, { "epoch": 4.02, "learning_rate": 5.978279620481456e-05, "loss": 2.6657, "step": 51289 }, { "epoch": 4.02, "learning_rate": 5.978201207559006e-05, "loss": 2.6075, "step": 51290 }, { "epoch": 4.02, "learning_rate": 5.978122794636556e-05, "loss": 2.7246, "step": 51291 }, { "epoch": 4.02, "learning_rate": 5.9780443817141065e-05, "loss": 2.5003, "step": 51292 }, { "epoch": 4.02, "learning_rate": 5.977965968791657e-05, "loss": 2.6751, "step": 51293 }, { "epoch": 4.02, "learning_rate": 5.977887555869207e-05, "loss": 2.5798, "step": 51294 }, { "epoch": 4.02, "learning_rate": 5.9778091429467576e-05, "loss": 2.6869, "step": 51295 }, { "epoch": 4.02, "learning_rate": 5.977730730024308e-05, "loss": 2.6156, "step": 51296 }, { "epoch": 4.02, "learning_rate": 5.9776523171018584e-05, "loss": 2.7116, "step": 51297 }, { "epoch": 4.02, "learning_rate": 5.977573904179409e-05, "loss": 2.9105, "step": 51298 }, { "epoch": 4.02, "learning_rate": 5.977495491256959e-05, "loss": 2.8737, "step": 51299 }, { "epoch": 4.02, "learning_rate": 5.9774170783345095e-05, "loss": 2.6963, "step": 51300 }, { "epoch": 4.02, "learning_rate": 5.97733866541206e-05, "loss": 2.7662, "step": 51301 }, { "epoch": 4.02, "learning_rate": 5.97726025248961e-05, "loss": 2.5418, "step": 51302 }, { "epoch": 4.02, "learning_rate": 5.9771818395671606e-05, "loss": 2.6919, "step": 51303 }, { "epoch": 4.02, "learning_rate": 5.977103426644711e-05, "loss": 2.6241, "step": 51304 }, { "epoch": 4.02, "learning_rate": 5.977025013722262e-05, "loss": 2.4687, "step": 51305 }, { "epoch": 4.02, "learning_rate": 5.9769466007998123e-05, "loss": 2.5884, "step": 51306 }, { "epoch": 4.02, "learning_rate": 5.976868187877363e-05, "loss": 2.6595, "step": 51307 }, { "epoch": 4.02, "learning_rate": 5.976789774954913e-05, "loss": 2.5584, "step": 51308 }, { "epoch": 4.02, "learning_rate": 5.9767113620324634e-05, "loss": 2.9032, "step": 51309 }, { "epoch": 4.02, "learning_rate": 5.976632949110014e-05, "loss": 2.4662, "step": 51310 }, { "epoch": 4.02, "learning_rate": 5.976554536187564e-05, "loss": 2.5928, "step": 51311 }, { "epoch": 4.02, "learning_rate": 5.9764761232651145e-05, "loss": 2.8151, "step": 51312 }, { "epoch": 4.02, "learning_rate": 5.976397710342665e-05, "loss": 2.6341, "step": 51313 }, { "epoch": 4.02, "learning_rate": 5.976319297420215e-05, "loss": 2.5853, "step": 51314 }, { "epoch": 4.02, "learning_rate": 5.9762408844977657e-05, "loss": 2.753, "step": 51315 }, { "epoch": 4.02, "learning_rate": 5.976162471575316e-05, "loss": 2.5333, "step": 51316 }, { "epoch": 4.02, "learning_rate": 5.9760840586528664e-05, "loss": 2.6843, "step": 51317 }, { "epoch": 4.02, "learning_rate": 5.976005645730417e-05, "loss": 2.6051, "step": 51318 }, { "epoch": 4.02, "learning_rate": 5.975927232807967e-05, "loss": 2.6117, "step": 51319 }, { "epoch": 4.02, "learning_rate": 5.9758488198855175e-05, "loss": 2.5047, "step": 51320 }, { "epoch": 4.02, "learning_rate": 5.9757704069630685e-05, "loss": 2.4907, "step": 51321 }, { "epoch": 4.02, "learning_rate": 5.975691994040619e-05, "loss": 2.5148, "step": 51322 }, { "epoch": 4.02, "learning_rate": 5.975613581118169e-05, "loss": 2.667, "step": 51323 }, { "epoch": 4.02, "learning_rate": 5.9755351681957196e-05, "loss": 2.9073, "step": 51324 }, { "epoch": 4.02, "learning_rate": 5.97545675527327e-05, "loss": 2.6081, "step": 51325 }, { "epoch": 4.02, "learning_rate": 5.975378342350819e-05, "loss": 2.9209, "step": 51326 }, { "epoch": 4.02, "learning_rate": 5.9752999294283694e-05, "loss": 2.5584, "step": 51327 }, { "epoch": 4.02, "learning_rate": 5.97522151650592e-05, "loss": 2.7749, "step": 51328 }, { "epoch": 4.02, "learning_rate": 5.97514310358347e-05, "loss": 2.6262, "step": 51329 }, { "epoch": 4.02, "learning_rate": 5.9750646906610205e-05, "loss": 2.5845, "step": 51330 }, { "epoch": 4.03, "learning_rate": 5.9749862777385715e-05, "loss": 2.5476, "step": 51331 }, { "epoch": 4.03, "learning_rate": 5.974907864816122e-05, "loss": 2.6156, "step": 51332 }, { "epoch": 4.03, "learning_rate": 5.974829451893672e-05, "loss": 2.6603, "step": 51333 }, { "epoch": 4.03, "learning_rate": 5.9747510389712226e-05, "loss": 2.5823, "step": 51334 }, { "epoch": 4.03, "learning_rate": 5.974672626048773e-05, "loss": 2.4886, "step": 51335 }, { "epoch": 4.03, "learning_rate": 5.9745942131263234e-05, "loss": 2.7141, "step": 51336 }, { "epoch": 4.03, "learning_rate": 5.974515800203874e-05, "loss": 2.8473, "step": 51337 }, { "epoch": 4.03, "learning_rate": 5.974437387281424e-05, "loss": 2.5903, "step": 51338 }, { "epoch": 4.03, "learning_rate": 5.9743589743589745e-05, "loss": 2.5929, "step": 51339 }, { "epoch": 4.03, "learning_rate": 5.974280561436525e-05, "loss": 2.9129, "step": 51340 }, { "epoch": 4.03, "learning_rate": 5.974202148514075e-05, "loss": 2.6346, "step": 51341 }, { "epoch": 4.03, "learning_rate": 5.9741237355916256e-05, "loss": 2.7999, "step": 51342 }, { "epoch": 4.03, "learning_rate": 5.974045322669176e-05, "loss": 2.6866, "step": 51343 }, { "epoch": 4.03, "learning_rate": 5.973966909746726e-05, "loss": 2.6736, "step": 51344 }, { "epoch": 4.03, "learning_rate": 5.973888496824277e-05, "loss": 2.5128, "step": 51345 }, { "epoch": 4.03, "learning_rate": 5.973810083901828e-05, "loss": 2.5803, "step": 51346 }, { "epoch": 4.03, "learning_rate": 5.973731670979378e-05, "loss": 2.9073, "step": 51347 }, { "epoch": 4.03, "learning_rate": 5.9736532580569285e-05, "loss": 2.6363, "step": 51348 }, { "epoch": 4.03, "learning_rate": 5.973574845134479e-05, "loss": 2.6644, "step": 51349 }, { "epoch": 4.03, "learning_rate": 5.973496432212029e-05, "loss": 2.49, "step": 51350 }, { "epoch": 4.03, "learning_rate": 5.9734180192895796e-05, "loss": 2.6876, "step": 51351 }, { "epoch": 4.03, "learning_rate": 5.97333960636713e-05, "loss": 2.6712, "step": 51352 }, { "epoch": 4.03, "learning_rate": 5.97326119344468e-05, "loss": 2.6992, "step": 51353 }, { "epoch": 4.03, "learning_rate": 5.973182780522231e-05, "loss": 2.7617, "step": 51354 }, { "epoch": 4.03, "learning_rate": 5.973104367599781e-05, "loss": 2.6718, "step": 51355 }, { "epoch": 4.03, "learning_rate": 5.9730259546773314e-05, "loss": 2.6664, "step": 51356 }, { "epoch": 4.03, "learning_rate": 5.972947541754882e-05, "loss": 2.726, "step": 51357 }, { "epoch": 4.03, "learning_rate": 5.972869128832432e-05, "loss": 2.5855, "step": 51358 }, { "epoch": 4.03, "learning_rate": 5.9727907159099825e-05, "loss": 2.6368, "step": 51359 }, { "epoch": 4.03, "learning_rate": 5.972712302987533e-05, "loss": 2.641, "step": 51360 }, { "epoch": 4.03, "learning_rate": 5.972633890065083e-05, "loss": 2.7116, "step": 51361 }, { "epoch": 4.03, "learning_rate": 5.972555477142633e-05, "loss": 2.729, "step": 51362 }, { "epoch": 4.03, "learning_rate": 5.972477064220183e-05, "loss": 2.5782, "step": 51363 }, { "epoch": 4.03, "learning_rate": 5.972398651297734e-05, "loss": 2.7594, "step": 51364 }, { "epoch": 4.03, "learning_rate": 5.972320238375284e-05, "loss": 2.5995, "step": 51365 }, { "epoch": 4.03, "learning_rate": 5.9722418254528344e-05, "loss": 2.5417, "step": 51366 }, { "epoch": 4.03, "learning_rate": 5.972163412530385e-05, "loss": 2.5918, "step": 51367 }, { "epoch": 4.03, "learning_rate": 5.972084999607935e-05, "loss": 2.6455, "step": 51368 }, { "epoch": 4.03, "learning_rate": 5.9720065866854855e-05, "loss": 2.5582, "step": 51369 }, { "epoch": 4.03, "learning_rate": 5.971928173763036e-05, "loss": 2.7548, "step": 51370 }, { "epoch": 4.03, "learning_rate": 5.971849760840586e-05, "loss": 2.8313, "step": 51371 }, { "epoch": 4.03, "learning_rate": 5.971771347918137e-05, "loss": 2.5815, "step": 51372 }, { "epoch": 4.03, "learning_rate": 5.9716929349956877e-05, "loss": 2.6044, "step": 51373 }, { "epoch": 4.03, "learning_rate": 5.971614522073238e-05, "loss": 2.676, "step": 51374 }, { "epoch": 4.03, "learning_rate": 5.9715361091507884e-05, "loss": 2.8192, "step": 51375 }, { "epoch": 4.03, "learning_rate": 5.971457696228339e-05, "loss": 2.5289, "step": 51376 }, { "epoch": 4.03, "learning_rate": 5.971379283305889e-05, "loss": 2.5234, "step": 51377 }, { "epoch": 4.03, "learning_rate": 5.9713008703834395e-05, "loss": 2.6528, "step": 51378 }, { "epoch": 4.03, "learning_rate": 5.97122245746099e-05, "loss": 2.6687, "step": 51379 }, { "epoch": 4.03, "learning_rate": 5.97114404453854e-05, "loss": 2.5777, "step": 51380 }, { "epoch": 4.03, "learning_rate": 5.9710656316160906e-05, "loss": 2.5988, "step": 51381 }, { "epoch": 4.03, "learning_rate": 5.970987218693641e-05, "loss": 2.9784, "step": 51382 }, { "epoch": 4.03, "learning_rate": 5.970908805771191e-05, "loss": 2.8254, "step": 51383 }, { "epoch": 4.03, "learning_rate": 5.970830392848742e-05, "loss": 2.6462, "step": 51384 }, { "epoch": 4.03, "learning_rate": 5.970751979926292e-05, "loss": 2.5105, "step": 51385 }, { "epoch": 4.03, "learning_rate": 5.9706735670038424e-05, "loss": 2.7045, "step": 51386 }, { "epoch": 4.03, "learning_rate": 5.970595154081393e-05, "loss": 2.6865, "step": 51387 }, { "epoch": 4.03, "learning_rate": 5.970516741158944e-05, "loss": 2.6191, "step": 51388 }, { "epoch": 4.03, "learning_rate": 5.970438328236494e-05, "loss": 2.6707, "step": 51389 }, { "epoch": 4.03, "learning_rate": 5.9703599153140446e-05, "loss": 2.8169, "step": 51390 }, { "epoch": 4.03, "learning_rate": 5.970281502391595e-05, "loss": 2.7362, "step": 51391 }, { "epoch": 4.03, "learning_rate": 5.970203089469145e-05, "loss": 2.4681, "step": 51392 }, { "epoch": 4.03, "learning_rate": 5.970124676546696e-05, "loss": 2.5746, "step": 51393 }, { "epoch": 4.03, "learning_rate": 5.970046263624246e-05, "loss": 2.5012, "step": 51394 }, { "epoch": 4.03, "learning_rate": 5.9699678507017964e-05, "loss": 2.3881, "step": 51395 }, { "epoch": 4.03, "learning_rate": 5.969889437779347e-05, "loss": 2.6982, "step": 51396 }, { "epoch": 4.03, "learning_rate": 5.969811024856897e-05, "loss": 2.3778, "step": 51397 }, { "epoch": 4.03, "learning_rate": 5.969732611934447e-05, "loss": 2.4835, "step": 51398 }, { "epoch": 4.03, "learning_rate": 5.969654199011997e-05, "loss": 2.585, "step": 51399 }, { "epoch": 4.03, "learning_rate": 5.9695757860895476e-05, "loss": 2.551, "step": 51400 }, { "epoch": 4.03, "learning_rate": 5.969497373167098e-05, "loss": 2.5751, "step": 51401 }, { "epoch": 4.03, "learning_rate": 5.969418960244648e-05, "loss": 2.5722, "step": 51402 }, { "epoch": 4.03, "learning_rate": 5.969340547322199e-05, "loss": 2.5764, "step": 51403 }, { "epoch": 4.03, "learning_rate": 5.969262134399749e-05, "loss": 2.4378, "step": 51404 }, { "epoch": 4.03, "learning_rate": 5.9691837214772994e-05, "loss": 2.6353, "step": 51405 }, { "epoch": 4.03, "learning_rate": 5.96910530855485e-05, "loss": 2.558, "step": 51406 }, { "epoch": 4.03, "learning_rate": 5.9690268956324e-05, "loss": 2.5034, "step": 51407 }, { "epoch": 4.03, "learning_rate": 5.9689484827099505e-05, "loss": 2.7237, "step": 51408 }, { "epoch": 4.03, "learning_rate": 5.968870069787501e-05, "loss": 2.7355, "step": 51409 }, { "epoch": 4.03, "learning_rate": 5.968791656865051e-05, "loss": 2.6339, "step": 51410 }, { "epoch": 4.03, "learning_rate": 5.9687132439426016e-05, "loss": 2.7469, "step": 51411 }, { "epoch": 4.03, "learning_rate": 5.968634831020152e-05, "loss": 2.5802, "step": 51412 }, { "epoch": 4.03, "learning_rate": 5.968556418097703e-05, "loss": 2.6031, "step": 51413 }, { "epoch": 4.03, "learning_rate": 5.9684780051752534e-05, "loss": 2.6008, "step": 51414 }, { "epoch": 4.03, "learning_rate": 5.968399592252804e-05, "loss": 2.7712, "step": 51415 }, { "epoch": 4.03, "learning_rate": 5.968321179330354e-05, "loss": 2.7995, "step": 51416 }, { "epoch": 4.03, "learning_rate": 5.9682427664079045e-05, "loss": 2.6725, "step": 51417 }, { "epoch": 4.03, "learning_rate": 5.968164353485455e-05, "loss": 2.691, "step": 51418 }, { "epoch": 4.03, "learning_rate": 5.968085940563005e-05, "loss": 2.633, "step": 51419 }, { "epoch": 4.03, "learning_rate": 5.9680075276405556e-05, "loss": 2.4927, "step": 51420 }, { "epoch": 4.03, "learning_rate": 5.967929114718106e-05, "loss": 2.6755, "step": 51421 }, { "epoch": 4.03, "learning_rate": 5.9678507017956564e-05, "loss": 2.563, "step": 51422 }, { "epoch": 4.03, "learning_rate": 5.967772288873207e-05, "loss": 2.6183, "step": 51423 }, { "epoch": 4.03, "learning_rate": 5.967693875950757e-05, "loss": 2.6693, "step": 51424 }, { "epoch": 4.03, "learning_rate": 5.9676154630283075e-05, "loss": 2.5546, "step": 51425 }, { "epoch": 4.03, "learning_rate": 5.967537050105858e-05, "loss": 2.537, "step": 51426 }, { "epoch": 4.03, "learning_rate": 5.967458637183408e-05, "loss": 2.586, "step": 51427 }, { "epoch": 4.03, "learning_rate": 5.9673802242609586e-05, "loss": 2.5505, "step": 51428 }, { "epoch": 4.03, "learning_rate": 5.9673018113385096e-05, "loss": 2.7196, "step": 51429 }, { "epoch": 4.03, "learning_rate": 5.96722339841606e-05, "loss": 2.6568, "step": 51430 }, { "epoch": 4.03, "learning_rate": 5.9671449854936103e-05, "loss": 2.5023, "step": 51431 }, { "epoch": 4.03, "learning_rate": 5.967066572571161e-05, "loss": 2.5184, "step": 51432 }, { "epoch": 4.03, "learning_rate": 5.966988159648711e-05, "loss": 2.6325, "step": 51433 }, { "epoch": 4.03, "learning_rate": 5.96690974672626e-05, "loss": 2.543, "step": 51434 }, { "epoch": 4.03, "learning_rate": 5.9668313338038105e-05, "loss": 2.5532, "step": 51435 }, { "epoch": 4.03, "learning_rate": 5.966752920881361e-05, "loss": 2.7831, "step": 51436 }, { "epoch": 4.03, "learning_rate": 5.966674507958911e-05, "loss": 2.6566, "step": 51437 }, { "epoch": 4.03, "learning_rate": 5.9665960950364616e-05, "loss": 2.5527, "step": 51438 }, { "epoch": 4.03, "learning_rate": 5.9665176821140126e-05, "loss": 2.6919, "step": 51439 }, { "epoch": 4.03, "learning_rate": 5.966439269191563e-05, "loss": 2.6472, "step": 51440 }, { "epoch": 4.03, "learning_rate": 5.9663608562691133e-05, "loss": 2.5368, "step": 51441 }, { "epoch": 4.03, "learning_rate": 5.966282443346664e-05, "loss": 2.6557, "step": 51442 }, { "epoch": 4.03, "learning_rate": 5.966204030424214e-05, "loss": 2.624, "step": 51443 }, { "epoch": 4.03, "learning_rate": 5.9661256175017644e-05, "loss": 2.6442, "step": 51444 }, { "epoch": 4.03, "learning_rate": 5.966047204579315e-05, "loss": 2.6223, "step": 51445 }, { "epoch": 4.03, "learning_rate": 5.965968791656865e-05, "loss": 2.6387, "step": 51446 }, { "epoch": 4.03, "learning_rate": 5.9658903787344156e-05, "loss": 2.8434, "step": 51447 }, { "epoch": 4.03, "learning_rate": 5.965811965811966e-05, "loss": 2.4762, "step": 51448 }, { "epoch": 4.03, "learning_rate": 5.965733552889516e-05, "loss": 2.7811, "step": 51449 }, { "epoch": 4.03, "learning_rate": 5.9656551399670667e-05, "loss": 2.5577, "step": 51450 }, { "epoch": 4.03, "learning_rate": 5.965576727044617e-05, "loss": 2.6144, "step": 51451 }, { "epoch": 4.03, "learning_rate": 5.9654983141221674e-05, "loss": 2.6093, "step": 51452 }, { "epoch": 4.03, "learning_rate": 5.965419901199718e-05, "loss": 2.7239, "step": 51453 }, { "epoch": 4.03, "learning_rate": 5.965341488277268e-05, "loss": 2.4344, "step": 51454 }, { "epoch": 4.03, "learning_rate": 5.965263075354819e-05, "loss": 2.8911, "step": 51455 }, { "epoch": 4.03, "learning_rate": 5.9651846624323695e-05, "loss": 2.7303, "step": 51456 }, { "epoch": 4.03, "learning_rate": 5.96510624950992e-05, "loss": 2.5981, "step": 51457 }, { "epoch": 4.03, "learning_rate": 5.96502783658747e-05, "loss": 2.8084, "step": 51458 }, { "epoch": 4.04, "learning_rate": 5.9649494236650206e-05, "loss": 2.5995, "step": 51459 }, { "epoch": 4.04, "learning_rate": 5.964871010742571e-05, "loss": 2.7084, "step": 51460 }, { "epoch": 4.04, "learning_rate": 5.9647925978201214e-05, "loss": 2.6726, "step": 51461 }, { "epoch": 4.04, "learning_rate": 5.964714184897672e-05, "loss": 2.9434, "step": 51462 }, { "epoch": 4.04, "learning_rate": 5.964635771975222e-05, "loss": 2.4955, "step": 51463 }, { "epoch": 4.04, "learning_rate": 5.9645573590527725e-05, "loss": 2.5629, "step": 51464 }, { "epoch": 4.04, "learning_rate": 5.964478946130323e-05, "loss": 2.5915, "step": 51465 }, { "epoch": 4.04, "learning_rate": 5.964400533207873e-05, "loss": 2.8404, "step": 51466 }, { "epoch": 4.04, "learning_rate": 5.9643221202854236e-05, "loss": 2.6742, "step": 51467 }, { "epoch": 4.04, "learning_rate": 5.964243707362974e-05, "loss": 2.6448, "step": 51468 }, { "epoch": 4.04, "learning_rate": 5.964165294440524e-05, "loss": 2.6028, "step": 51469 }, { "epoch": 4.04, "learning_rate": 5.964086881518074e-05, "loss": 2.7605, "step": 51470 }, { "epoch": 4.04, "learning_rate": 5.9640084685956244e-05, "loss": 2.4704, "step": 51471 }, { "epoch": 4.04, "learning_rate": 5.963930055673175e-05, "loss": 2.6717, "step": 51472 }, { "epoch": 4.04, "learning_rate": 5.963851642750725e-05, "loss": 3.1462, "step": 51473 }, { "epoch": 4.04, "learning_rate": 5.9637732298282755e-05, "loss": 2.6295, "step": 51474 }, { "epoch": 4.04, "learning_rate": 5.963694816905826e-05, "loss": 2.5691, "step": 51475 }, { "epoch": 4.04, "learning_rate": 5.963616403983376e-05, "loss": 2.6103, "step": 51476 }, { "epoch": 4.04, "learning_rate": 5.9635379910609266e-05, "loss": 2.5374, "step": 51477 }, { "epoch": 4.04, "learning_rate": 5.963459578138477e-05, "loss": 2.6126, "step": 51478 }, { "epoch": 4.04, "learning_rate": 5.963381165216027e-05, "loss": 2.7231, "step": 51479 }, { "epoch": 4.04, "learning_rate": 5.9633027522935784e-05, "loss": 2.5423, "step": 51480 }, { "epoch": 4.04, "learning_rate": 5.963224339371129e-05, "loss": 2.5897, "step": 51481 }, { "epoch": 4.04, "learning_rate": 5.963145926448679e-05, "loss": 2.8596, "step": 51482 }, { "epoch": 4.04, "learning_rate": 5.9630675135262295e-05, "loss": 2.5942, "step": 51483 }, { "epoch": 4.04, "learning_rate": 5.96298910060378e-05, "loss": 2.5226, "step": 51484 }, { "epoch": 4.04, "learning_rate": 5.96291068768133e-05, "loss": 2.7652, "step": 51485 }, { "epoch": 4.04, "learning_rate": 5.9628322747588806e-05, "loss": 2.7766, "step": 51486 }, { "epoch": 4.04, "learning_rate": 5.962753861836431e-05, "loss": 2.7353, "step": 51487 }, { "epoch": 4.04, "learning_rate": 5.962675448913981e-05, "loss": 2.6884, "step": 51488 }, { "epoch": 4.04, "learning_rate": 5.962597035991532e-05, "loss": 2.5218, "step": 51489 }, { "epoch": 4.04, "learning_rate": 5.962518623069082e-05, "loss": 2.8372, "step": 51490 }, { "epoch": 4.04, "learning_rate": 5.9624402101466324e-05, "loss": 2.4705, "step": 51491 }, { "epoch": 4.04, "learning_rate": 5.962361797224183e-05, "loss": 2.6687, "step": 51492 }, { "epoch": 4.04, "learning_rate": 5.962283384301733e-05, "loss": 2.8169, "step": 51493 }, { "epoch": 4.04, "learning_rate": 5.9622049713792835e-05, "loss": 2.611, "step": 51494 }, { "epoch": 4.04, "learning_rate": 5.962126558456834e-05, "loss": 2.7014, "step": 51495 }, { "epoch": 4.04, "learning_rate": 5.962048145534385e-05, "loss": 2.868, "step": 51496 }, { "epoch": 4.04, "learning_rate": 5.961969732611935e-05, "loss": 2.5138, "step": 51497 }, { "epoch": 4.04, "learning_rate": 5.961891319689486e-05, "loss": 2.3963, "step": 51498 }, { "epoch": 4.04, "learning_rate": 5.961812906767036e-05, "loss": 2.5577, "step": 51499 }, { "epoch": 4.04, "learning_rate": 5.9617344938445864e-05, "loss": 2.5841, "step": 51500 }, { "epoch": 4.04, "learning_rate": 5.961656080922137e-05, "loss": 2.616, "step": 51501 }, { "epoch": 4.04, "learning_rate": 5.961577667999687e-05, "loss": 2.6852, "step": 51502 }, { "epoch": 4.04, "learning_rate": 5.9614992550772375e-05, "loss": 2.7713, "step": 51503 }, { "epoch": 4.04, "learning_rate": 5.961420842154788e-05, "loss": 2.5858, "step": 51504 }, { "epoch": 4.04, "learning_rate": 5.961342429232338e-05, "loss": 2.4509, "step": 51505 }, { "epoch": 4.04, "learning_rate": 5.961264016309888e-05, "loss": 2.7343, "step": 51506 }, { "epoch": 4.04, "learning_rate": 5.961185603387438e-05, "loss": 2.7456, "step": 51507 }, { "epoch": 4.04, "learning_rate": 5.961107190464989e-05, "loss": 2.647, "step": 51508 }, { "epoch": 4.04, "learning_rate": 5.961028777542539e-05, "loss": 2.8341, "step": 51509 }, { "epoch": 4.04, "learning_rate": 5.9609503646200894e-05, "loss": 2.4875, "step": 51510 }, { "epoch": 4.04, "learning_rate": 5.96087195169764e-05, "loss": 2.6918, "step": 51511 }, { "epoch": 4.04, "learning_rate": 5.96079353877519e-05, "loss": 2.5121, "step": 51512 }, { "epoch": 4.04, "learning_rate": 5.9607151258527405e-05, "loss": 2.7707, "step": 51513 }, { "epoch": 4.04, "learning_rate": 5.960636712930291e-05, "loss": 2.7652, "step": 51514 }, { "epoch": 4.04, "learning_rate": 5.960558300007841e-05, "loss": 2.6021, "step": 51515 }, { "epoch": 4.04, "learning_rate": 5.9604798870853916e-05, "loss": 2.5318, "step": 51516 }, { "epoch": 4.04, "learning_rate": 5.960401474162942e-05, "loss": 2.5808, "step": 51517 }, { "epoch": 4.04, "learning_rate": 5.9603230612404923e-05, "loss": 2.5395, "step": 51518 }, { "epoch": 4.04, "learning_rate": 5.960244648318043e-05, "loss": 2.5559, "step": 51519 }, { "epoch": 4.04, "learning_rate": 5.960166235395593e-05, "loss": 2.8897, "step": 51520 }, { "epoch": 4.04, "learning_rate": 5.960087822473144e-05, "loss": 2.697, "step": 51521 }, { "epoch": 4.04, "learning_rate": 5.9600094095506945e-05, "loss": 2.9761, "step": 51522 }, { "epoch": 4.04, "learning_rate": 5.959930996628245e-05, "loss": 2.515, "step": 51523 }, { "epoch": 4.04, "learning_rate": 5.959852583705795e-05, "loss": 2.7446, "step": 51524 }, { "epoch": 4.04, "learning_rate": 5.9597741707833456e-05, "loss": 2.7885, "step": 51525 }, { "epoch": 4.04, "learning_rate": 5.959695757860896e-05, "loss": 2.6122, "step": 51526 }, { "epoch": 4.04, "learning_rate": 5.959617344938446e-05, "loss": 2.5991, "step": 51527 }, { "epoch": 4.04, "learning_rate": 5.959538932015997e-05, "loss": 2.7264, "step": 51528 }, { "epoch": 4.04, "learning_rate": 5.959460519093547e-05, "loss": 2.724, "step": 51529 }, { "epoch": 4.04, "learning_rate": 5.9593821061710974e-05, "loss": 3.0121, "step": 51530 }, { "epoch": 4.04, "learning_rate": 5.959303693248648e-05, "loss": 2.5953, "step": 51531 }, { "epoch": 4.04, "learning_rate": 5.959225280326198e-05, "loss": 2.6627, "step": 51532 }, { "epoch": 4.04, "learning_rate": 5.9591468674037485e-05, "loss": 2.8352, "step": 51533 }, { "epoch": 4.04, "learning_rate": 5.959068454481299e-05, "loss": 2.4269, "step": 51534 }, { "epoch": 4.04, "learning_rate": 5.958990041558849e-05, "loss": 2.5477, "step": 51535 }, { "epoch": 4.04, "learning_rate": 5.9589116286363996e-05, "loss": 2.7051, "step": 51536 }, { "epoch": 4.04, "learning_rate": 5.958833215713951e-05, "loss": 2.6367, "step": 51537 }, { "epoch": 4.04, "learning_rate": 5.958754802791501e-05, "loss": 2.7636, "step": 51538 }, { "epoch": 4.04, "learning_rate": 5.9586763898690514e-05, "loss": 2.7493, "step": 51539 }, { "epoch": 4.04, "learning_rate": 5.958597976946602e-05, "loss": 2.7176, "step": 51540 }, { "epoch": 4.04, "learning_rate": 5.958519564024152e-05, "loss": 2.7132, "step": 51541 }, { "epoch": 4.04, "learning_rate": 5.958441151101701e-05, "loss": 2.4191, "step": 51542 }, { "epoch": 4.04, "learning_rate": 5.9583627381792515e-05, "loss": 2.7213, "step": 51543 }, { "epoch": 4.04, "learning_rate": 5.958284325256802e-05, "loss": 2.6766, "step": 51544 }, { "epoch": 4.04, "learning_rate": 5.958205912334352e-05, "loss": 2.726, "step": 51545 }, { "epoch": 4.04, "learning_rate": 5.9581274994119026e-05, "loss": 2.5899, "step": 51546 }, { "epoch": 4.04, "learning_rate": 5.958049086489454e-05, "loss": 2.6044, "step": 51547 }, { "epoch": 4.04, "learning_rate": 5.957970673567004e-05, "loss": 2.6376, "step": 51548 }, { "epoch": 4.04, "learning_rate": 5.9578922606445544e-05, "loss": 2.6035, "step": 51549 }, { "epoch": 4.04, "learning_rate": 5.957813847722105e-05, "loss": 2.4923, "step": 51550 }, { "epoch": 4.04, "learning_rate": 5.957735434799655e-05, "loss": 2.5357, "step": 51551 }, { "epoch": 4.04, "learning_rate": 5.9576570218772055e-05, "loss": 2.7047, "step": 51552 }, { "epoch": 4.04, "learning_rate": 5.957578608954756e-05, "loss": 2.6876, "step": 51553 }, { "epoch": 4.04, "learning_rate": 5.957500196032306e-05, "loss": 2.5684, "step": 51554 }, { "epoch": 4.04, "learning_rate": 5.9574217831098566e-05, "loss": 2.9055, "step": 51555 }, { "epoch": 4.04, "learning_rate": 5.957343370187407e-05, "loss": 2.6599, "step": 51556 }, { "epoch": 4.04, "learning_rate": 5.9572649572649574e-05, "loss": 2.4845, "step": 51557 }, { "epoch": 4.04, "learning_rate": 5.957186544342508e-05, "loss": 2.6295, "step": 51558 }, { "epoch": 4.04, "learning_rate": 5.957108131420058e-05, "loss": 2.5804, "step": 51559 }, { "epoch": 4.04, "learning_rate": 5.9570297184976085e-05, "loss": 2.7172, "step": 51560 }, { "epoch": 4.04, "learning_rate": 5.956951305575159e-05, "loss": 2.532, "step": 51561 }, { "epoch": 4.04, "learning_rate": 5.956872892652709e-05, "loss": 2.5628, "step": 51562 }, { "epoch": 4.04, "learning_rate": 5.95679447973026e-05, "loss": 2.5074, "step": 51563 }, { "epoch": 4.04, "learning_rate": 5.9567160668078106e-05, "loss": 2.687, "step": 51564 }, { "epoch": 4.04, "learning_rate": 5.956637653885361e-05, "loss": 2.826, "step": 51565 }, { "epoch": 4.04, "learning_rate": 5.9565592409629114e-05, "loss": 2.5963, "step": 51566 }, { "epoch": 4.04, "learning_rate": 5.956480828040462e-05, "loss": 2.6618, "step": 51567 }, { "epoch": 4.04, "learning_rate": 5.956402415118012e-05, "loss": 2.648, "step": 51568 }, { "epoch": 4.04, "learning_rate": 5.9563240021955625e-05, "loss": 2.4791, "step": 51569 }, { "epoch": 4.04, "learning_rate": 5.956245589273113e-05, "loss": 2.5864, "step": 51570 }, { "epoch": 4.04, "learning_rate": 5.956167176350663e-05, "loss": 2.7626, "step": 51571 }, { "epoch": 4.04, "learning_rate": 5.9560887634282136e-05, "loss": 2.6663, "step": 51572 }, { "epoch": 4.04, "learning_rate": 5.956010350505764e-05, "loss": 2.6087, "step": 51573 }, { "epoch": 4.04, "learning_rate": 5.955931937583314e-05, "loss": 2.8078, "step": 51574 }, { "epoch": 4.04, "learning_rate": 5.9558535246608647e-05, "loss": 2.4969, "step": 51575 }, { "epoch": 4.04, "learning_rate": 5.955775111738415e-05, "loss": 2.7886, "step": 51576 }, { "epoch": 4.04, "learning_rate": 5.9556966988159654e-05, "loss": 2.7314, "step": 51577 }, { "epoch": 4.04, "learning_rate": 5.955618285893515e-05, "loss": 2.5112, "step": 51578 }, { "epoch": 4.04, "learning_rate": 5.9555398729710655e-05, "loss": 2.4943, "step": 51579 }, { "epoch": 4.04, "learning_rate": 5.955461460048616e-05, "loss": 2.516, "step": 51580 }, { "epoch": 4.04, "learning_rate": 5.955383047126166e-05, "loss": 2.7566, "step": 51581 }, { "epoch": 4.04, "learning_rate": 5.9553046342037166e-05, "loss": 2.513, "step": 51582 }, { "epoch": 4.04, "learning_rate": 5.955226221281267e-05, "loss": 2.7215, "step": 51583 }, { "epoch": 4.04, "learning_rate": 5.955147808358817e-05, "loss": 2.7465, "step": 51584 }, { "epoch": 4.04, "learning_rate": 5.9550693954363677e-05, "loss": 2.5952, "step": 51585 }, { "epoch": 4.05, "learning_rate": 5.954990982513918e-05, "loss": 2.7817, "step": 51586 }, { "epoch": 4.05, "learning_rate": 5.9549125695914684e-05, "loss": 2.6133, "step": 51587 }, { "epoch": 4.05, "learning_rate": 5.9548341566690194e-05, "loss": 2.7959, "step": 51588 }, { "epoch": 4.05, "learning_rate": 5.95475574374657e-05, "loss": 2.5736, "step": 51589 }, { "epoch": 4.05, "learning_rate": 5.95467733082412e-05, "loss": 2.6291, "step": 51590 }, { "epoch": 4.05, "learning_rate": 5.9545989179016705e-05, "loss": 2.8134, "step": 51591 }, { "epoch": 4.05, "learning_rate": 5.954520504979221e-05, "loss": 2.6944, "step": 51592 }, { "epoch": 4.05, "learning_rate": 5.954442092056771e-05, "loss": 2.6581, "step": 51593 }, { "epoch": 4.05, "learning_rate": 5.9543636791343216e-05, "loss": 2.5929, "step": 51594 }, { "epoch": 4.05, "learning_rate": 5.954285266211872e-05, "loss": 2.4762, "step": 51595 }, { "epoch": 4.05, "learning_rate": 5.9542068532894224e-05, "loss": 2.4414, "step": 51596 }, { "epoch": 4.05, "learning_rate": 5.954128440366973e-05, "loss": 2.7277, "step": 51597 }, { "epoch": 4.05, "learning_rate": 5.954050027444523e-05, "loss": 2.6176, "step": 51598 }, { "epoch": 4.05, "learning_rate": 5.9539716145220735e-05, "loss": 2.6784, "step": 51599 }, { "epoch": 4.05, "learning_rate": 5.953893201599624e-05, "loss": 2.6793, "step": 51600 }, { "epoch": 4.05, "learning_rate": 5.953814788677174e-05, "loss": 2.6401, "step": 51601 }, { "epoch": 4.05, "learning_rate": 5.9537363757547246e-05, "loss": 2.632, "step": 51602 }, { "epoch": 4.05, "learning_rate": 5.953657962832275e-05, "loss": 2.5034, "step": 51603 }, { "epoch": 4.05, "learning_rate": 5.953579549909826e-05, "loss": 2.5717, "step": 51604 }, { "epoch": 4.05, "learning_rate": 5.9535011369873764e-05, "loss": 2.6209, "step": 51605 }, { "epoch": 4.05, "learning_rate": 5.953422724064927e-05, "loss": 2.6321, "step": 51606 }, { "epoch": 4.05, "learning_rate": 5.953344311142477e-05, "loss": 2.628, "step": 51607 }, { "epoch": 4.05, "learning_rate": 5.9532658982200275e-05, "loss": 2.8244, "step": 51608 }, { "epoch": 4.05, "learning_rate": 5.953187485297578e-05, "loss": 2.7839, "step": 51609 }, { "epoch": 4.05, "learning_rate": 5.953109072375128e-05, "loss": 2.5752, "step": 51610 }, { "epoch": 4.05, "learning_rate": 5.9530306594526786e-05, "loss": 2.7419, "step": 51611 }, { "epoch": 4.05, "learning_rate": 5.952952246530229e-05, "loss": 2.4965, "step": 51612 }, { "epoch": 4.05, "learning_rate": 5.952873833607779e-05, "loss": 2.492, "step": 51613 }, { "epoch": 4.05, "learning_rate": 5.952795420685329e-05, "loss": 2.957, "step": 51614 }, { "epoch": 4.05, "learning_rate": 5.9527170077628794e-05, "loss": 2.8042, "step": 51615 }, { "epoch": 4.05, "learning_rate": 5.95263859484043e-05, "loss": 2.5293, "step": 51616 }, { "epoch": 4.05, "learning_rate": 5.95256018191798e-05, "loss": 2.5631, "step": 51617 }, { "epoch": 4.05, "learning_rate": 5.9524817689955305e-05, "loss": 2.7961, "step": 51618 }, { "epoch": 4.05, "learning_rate": 5.952403356073081e-05, "loss": 2.763, "step": 51619 }, { "epoch": 4.05, "learning_rate": 5.952324943150631e-05, "loss": 2.8334, "step": 51620 }, { "epoch": 4.05, "learning_rate": 5.9522465302281816e-05, "loss": 2.6225, "step": 51621 }, { "epoch": 4.05, "learning_rate": 5.952168117305732e-05, "loss": 2.8035, "step": 51622 }, { "epoch": 4.05, "learning_rate": 5.952089704383282e-05, "loss": 2.6278, "step": 51623 }, { "epoch": 4.05, "learning_rate": 5.952011291460833e-05, "loss": 2.5931, "step": 51624 }, { "epoch": 4.05, "learning_rate": 5.951932878538383e-05, "loss": 2.7075, "step": 51625 }, { "epoch": 4.05, "learning_rate": 5.9518544656159334e-05, "loss": 2.6559, "step": 51626 }, { "epoch": 4.05, "learning_rate": 5.951776052693484e-05, "loss": 2.6072, "step": 51627 }, { "epoch": 4.05, "learning_rate": 5.951697639771034e-05, "loss": 2.7388, "step": 51628 }, { "epoch": 4.05, "learning_rate": 5.951619226848585e-05, "loss": 2.8058, "step": 51629 }, { "epoch": 4.05, "learning_rate": 5.9515408139261356e-05, "loss": 2.7287, "step": 51630 }, { "epoch": 4.05, "learning_rate": 5.951462401003686e-05, "loss": 2.6746, "step": 51631 }, { "epoch": 4.05, "learning_rate": 5.951383988081236e-05, "loss": 2.5826, "step": 51632 }, { "epoch": 4.05, "learning_rate": 5.951305575158787e-05, "loss": 2.866, "step": 51633 }, { "epoch": 4.05, "learning_rate": 5.951227162236337e-05, "loss": 2.7609, "step": 51634 }, { "epoch": 4.05, "learning_rate": 5.9511487493138874e-05, "loss": 2.3874, "step": 51635 }, { "epoch": 4.05, "learning_rate": 5.951070336391438e-05, "loss": 2.7262, "step": 51636 }, { "epoch": 4.05, "learning_rate": 5.950991923468988e-05, "loss": 2.7003, "step": 51637 }, { "epoch": 4.05, "learning_rate": 5.9509135105465385e-05, "loss": 2.7588, "step": 51638 }, { "epoch": 4.05, "learning_rate": 5.950835097624089e-05, "loss": 2.6471, "step": 51639 }, { "epoch": 4.05, "learning_rate": 5.950756684701639e-05, "loss": 2.7608, "step": 51640 }, { "epoch": 4.05, "learning_rate": 5.9506782717791896e-05, "loss": 2.6967, "step": 51641 }, { "epoch": 4.05, "learning_rate": 5.95059985885674e-05, "loss": 2.5479, "step": 51642 }, { "epoch": 4.05, "learning_rate": 5.9505214459342903e-05, "loss": 2.6333, "step": 51643 }, { "epoch": 4.05, "learning_rate": 5.950443033011841e-05, "loss": 2.5688, "step": 51644 }, { "epoch": 4.05, "learning_rate": 5.950364620089392e-05, "loss": 2.6915, "step": 51645 }, { "epoch": 4.05, "learning_rate": 5.950286207166942e-05, "loss": 2.5005, "step": 51646 }, { "epoch": 4.05, "learning_rate": 5.9502077942444925e-05, "loss": 2.6397, "step": 51647 }, { "epoch": 4.05, "learning_rate": 5.950129381322043e-05, "loss": 2.4745, "step": 51648 }, { "epoch": 4.05, "learning_rate": 5.950050968399593e-05, "loss": 2.775, "step": 51649 }, { "epoch": 4.05, "learning_rate": 5.949972555477142e-05, "loss": 2.5426, "step": 51650 }, { "epoch": 4.05, "learning_rate": 5.9498941425546926e-05, "loss": 2.6168, "step": 51651 }, { "epoch": 4.05, "learning_rate": 5.949815729632243e-05, "loss": 2.5975, "step": 51652 }, { "epoch": 4.05, "learning_rate": 5.9497373167097933e-05, "loss": 2.7064, "step": 51653 }, { "epoch": 4.05, "learning_rate": 5.949658903787344e-05, "loss": 2.7601, "step": 51654 }, { "epoch": 4.05, "learning_rate": 5.949580490864895e-05, "loss": 2.5584, "step": 51655 }, { "epoch": 4.05, "learning_rate": 5.949502077942445e-05, "loss": 2.4484, "step": 51656 }, { "epoch": 4.05, "learning_rate": 5.9494236650199955e-05, "loss": 2.5696, "step": 51657 }, { "epoch": 4.05, "learning_rate": 5.949345252097546e-05, "loss": 2.642, "step": 51658 }, { "epoch": 4.05, "learning_rate": 5.949266839175096e-05, "loss": 2.4857, "step": 51659 }, { "epoch": 4.05, "learning_rate": 5.9491884262526466e-05, "loss": 2.7408, "step": 51660 }, { "epoch": 4.05, "learning_rate": 5.949110013330197e-05, "loss": 2.7697, "step": 51661 }, { "epoch": 4.05, "learning_rate": 5.949031600407747e-05, "loss": 2.6203, "step": 51662 }, { "epoch": 4.05, "learning_rate": 5.948953187485298e-05, "loss": 2.8735, "step": 51663 }, { "epoch": 4.05, "learning_rate": 5.948874774562848e-05, "loss": 2.7611, "step": 51664 }, { "epoch": 4.05, "learning_rate": 5.9487963616403984e-05, "loss": 2.6992, "step": 51665 }, { "epoch": 4.05, "learning_rate": 5.948717948717949e-05, "loss": 2.5997, "step": 51666 }, { "epoch": 4.05, "learning_rate": 5.948639535795499e-05, "loss": 2.7071, "step": 51667 }, { "epoch": 4.05, "learning_rate": 5.9485611228730495e-05, "loss": 2.5571, "step": 51668 }, { "epoch": 4.05, "learning_rate": 5.9484827099506e-05, "loss": 2.5487, "step": 51669 }, { "epoch": 4.05, "learning_rate": 5.94840429702815e-05, "loss": 2.4915, "step": 51670 }, { "epoch": 4.05, "learning_rate": 5.948325884105701e-05, "loss": 2.9532, "step": 51671 }, { "epoch": 4.05, "learning_rate": 5.948247471183252e-05, "loss": 2.8028, "step": 51672 }, { "epoch": 4.05, "learning_rate": 5.948169058260802e-05, "loss": 2.9382, "step": 51673 }, { "epoch": 4.05, "learning_rate": 5.9480906453383524e-05, "loss": 2.6348, "step": 51674 }, { "epoch": 4.05, "learning_rate": 5.948012232415903e-05, "loss": 2.7659, "step": 51675 }, { "epoch": 4.05, "learning_rate": 5.947933819493453e-05, "loss": 2.6584, "step": 51676 }, { "epoch": 4.05, "learning_rate": 5.9478554065710035e-05, "loss": 2.6438, "step": 51677 }, { "epoch": 4.05, "learning_rate": 5.947776993648554e-05, "loss": 2.719, "step": 51678 }, { "epoch": 4.05, "learning_rate": 5.947698580726104e-05, "loss": 2.5777, "step": 51679 }, { "epoch": 4.05, "learning_rate": 5.9476201678036546e-05, "loss": 2.6558, "step": 51680 }, { "epoch": 4.05, "learning_rate": 5.947541754881205e-05, "loss": 2.9708, "step": 51681 }, { "epoch": 4.05, "learning_rate": 5.9474633419587554e-05, "loss": 2.6751, "step": 51682 }, { "epoch": 4.05, "learning_rate": 5.947384929036306e-05, "loss": 2.6492, "step": 51683 }, { "epoch": 4.05, "learning_rate": 5.947306516113856e-05, "loss": 2.836, "step": 51684 }, { "epoch": 4.05, "learning_rate": 5.9472281031914065e-05, "loss": 2.5086, "step": 51685 }, { "epoch": 4.05, "learning_rate": 5.947149690268956e-05, "loss": 2.5635, "step": 51686 }, { "epoch": 4.05, "learning_rate": 5.9470712773465065e-05, "loss": 2.7565, "step": 51687 }, { "epoch": 4.05, "learning_rate": 5.946992864424057e-05, "loss": 2.8809, "step": 51688 }, { "epoch": 4.05, "learning_rate": 5.946914451501607e-05, "loss": 2.5623, "step": 51689 }, { "epoch": 4.05, "learning_rate": 5.9468360385791576e-05, "loss": 2.8404, "step": 51690 }, { "epoch": 4.05, "learning_rate": 5.946757625656708e-05, "loss": 2.7975, "step": 51691 }, { "epoch": 4.05, "learning_rate": 5.9466792127342584e-05, "loss": 2.8005, "step": 51692 }, { "epoch": 4.05, "learning_rate": 5.946600799811809e-05, "loss": 2.5656, "step": 51693 }, { "epoch": 4.05, "learning_rate": 5.946522386889359e-05, "loss": 2.8566, "step": 51694 }, { "epoch": 4.05, "learning_rate": 5.9464439739669095e-05, "loss": 2.5612, "step": 51695 }, { "epoch": 4.05, "learning_rate": 5.9463655610444605e-05, "loss": 2.8953, "step": 51696 }, { "epoch": 4.05, "learning_rate": 5.946287148122011e-05, "loss": 2.8082, "step": 51697 }, { "epoch": 4.05, "learning_rate": 5.946208735199561e-05, "loss": 2.8306, "step": 51698 }, { "epoch": 4.05, "learning_rate": 5.9461303222771116e-05, "loss": 2.9026, "step": 51699 }, { "epoch": 4.05, "learning_rate": 5.946051909354662e-05, "loss": 2.6863, "step": 51700 }, { "epoch": 4.05, "learning_rate": 5.9459734964322124e-05, "loss": 2.7623, "step": 51701 }, { "epoch": 4.05, "learning_rate": 5.945895083509763e-05, "loss": 2.7226, "step": 51702 }, { "epoch": 4.05, "learning_rate": 5.945816670587313e-05, "loss": 2.6118, "step": 51703 }, { "epoch": 4.05, "learning_rate": 5.9457382576648635e-05, "loss": 2.6586, "step": 51704 }, { "epoch": 4.05, "learning_rate": 5.945659844742414e-05, "loss": 2.8032, "step": 51705 }, { "epoch": 4.05, "learning_rate": 5.945581431819964e-05, "loss": 2.5875, "step": 51706 }, { "epoch": 4.05, "learning_rate": 5.9455030188975146e-05, "loss": 2.6235, "step": 51707 }, { "epoch": 4.05, "learning_rate": 5.945424605975065e-05, "loss": 2.9287, "step": 51708 }, { "epoch": 4.05, "learning_rate": 5.945346193052615e-05, "loss": 2.5596, "step": 51709 }, { "epoch": 4.05, "learning_rate": 5.945267780130166e-05, "loss": 2.6996, "step": 51710 }, { "epoch": 4.05, "learning_rate": 5.945189367207716e-05, "loss": 2.7295, "step": 51711 }, { "epoch": 4.05, "learning_rate": 5.945110954285267e-05, "loss": 2.683, "step": 51712 }, { "epoch": 4.05, "learning_rate": 5.9450325413628174e-05, "loss": 2.5712, "step": 51713 }, { "epoch": 4.06, "learning_rate": 5.944954128440368e-05, "loss": 2.4834, "step": 51714 }, { "epoch": 4.06, "learning_rate": 5.944875715517918e-05, "loss": 2.7239, "step": 51715 }, { "epoch": 4.06, "learning_rate": 5.9447973025954685e-05, "loss": 2.8265, "step": 51716 }, { "epoch": 4.06, "learning_rate": 5.944718889673019e-05, "loss": 2.6652, "step": 51717 }, { "epoch": 4.06, "learning_rate": 5.944640476750569e-05, "loss": 2.5891, "step": 51718 }, { "epoch": 4.06, "learning_rate": 5.9445620638281197e-05, "loss": 2.9492, "step": 51719 }, { "epoch": 4.06, "learning_rate": 5.94448365090567e-05, "loss": 2.5772, "step": 51720 }, { "epoch": 4.06, "learning_rate": 5.9444052379832204e-05, "loss": 2.6123, "step": 51721 }, { "epoch": 4.06, "learning_rate": 5.94432682506077e-05, "loss": 2.5581, "step": 51722 }, { "epoch": 4.06, "learning_rate": 5.9442484121383204e-05, "loss": 2.6106, "step": 51723 }, { "epoch": 4.06, "learning_rate": 5.944169999215871e-05, "loss": 2.8813, "step": 51724 }, { "epoch": 4.06, "learning_rate": 5.944091586293421e-05, "loss": 2.5567, "step": 51725 }, { "epoch": 4.06, "learning_rate": 5.9440131733709715e-05, "loss": 2.8272, "step": 51726 }, { "epoch": 4.06, "learning_rate": 5.943934760448522e-05, "loss": 2.8, "step": 51727 }, { "epoch": 4.06, "learning_rate": 5.943856347526072e-05, "loss": 2.7846, "step": 51728 }, { "epoch": 4.06, "learning_rate": 5.9437779346036227e-05, "loss": 2.6784, "step": 51729 }, { "epoch": 4.06, "learning_rate": 5.943699521681173e-05, "loss": 2.717, "step": 51730 }, { "epoch": 4.06, "learning_rate": 5.9436211087587234e-05, "loss": 2.5782, "step": 51731 }, { "epoch": 4.06, "learning_rate": 5.943542695836274e-05, "loss": 2.7695, "step": 51732 }, { "epoch": 4.06, "learning_rate": 5.943464282913824e-05, "loss": 2.6271, "step": 51733 }, { "epoch": 4.06, "learning_rate": 5.9433858699913745e-05, "loss": 2.6527, "step": 51734 }, { "epoch": 4.06, "learning_rate": 5.943307457068925e-05, "loss": 2.6348, "step": 51735 }, { "epoch": 4.06, "learning_rate": 5.943229044146475e-05, "loss": 2.6473, "step": 51736 }, { "epoch": 4.06, "learning_rate": 5.943150631224026e-05, "loss": 2.7569, "step": 51737 }, { "epoch": 4.06, "learning_rate": 5.9430722183015766e-05, "loss": 2.513, "step": 51738 }, { "epoch": 4.06, "learning_rate": 5.942993805379127e-05, "loss": 2.7201, "step": 51739 }, { "epoch": 4.06, "learning_rate": 5.9429153924566774e-05, "loss": 2.816, "step": 51740 }, { "epoch": 4.06, "learning_rate": 5.942836979534228e-05, "loss": 2.6795, "step": 51741 }, { "epoch": 4.06, "learning_rate": 5.942758566611778e-05, "loss": 2.5686, "step": 51742 }, { "epoch": 4.06, "learning_rate": 5.9426801536893285e-05, "loss": 2.5638, "step": 51743 }, { "epoch": 4.06, "learning_rate": 5.942601740766879e-05, "loss": 2.7645, "step": 51744 }, { "epoch": 4.06, "learning_rate": 5.942523327844429e-05, "loss": 2.7926, "step": 51745 }, { "epoch": 4.06, "learning_rate": 5.9424449149219796e-05, "loss": 2.6315, "step": 51746 }, { "epoch": 4.06, "learning_rate": 5.94236650199953e-05, "loss": 2.543, "step": 51747 }, { "epoch": 4.06, "learning_rate": 5.94228808907708e-05, "loss": 2.596, "step": 51748 }, { "epoch": 4.06, "learning_rate": 5.942209676154631e-05, "loss": 2.5522, "step": 51749 }, { "epoch": 4.06, "learning_rate": 5.942131263232181e-05, "loss": 2.6824, "step": 51750 }, { "epoch": 4.06, "learning_rate": 5.9420528503097314e-05, "loss": 2.6249, "step": 51751 }, { "epoch": 4.06, "learning_rate": 5.941974437387282e-05, "loss": 2.7139, "step": 51752 }, { "epoch": 4.06, "learning_rate": 5.941896024464833e-05, "loss": 2.7193, "step": 51753 }, { "epoch": 4.06, "learning_rate": 5.941817611542383e-05, "loss": 2.6198, "step": 51754 }, { "epoch": 4.06, "learning_rate": 5.9417391986199336e-05, "loss": 2.6277, "step": 51755 }, { "epoch": 4.06, "learning_rate": 5.941660785697484e-05, "loss": 2.6306, "step": 51756 }, { "epoch": 4.06, "learning_rate": 5.941582372775034e-05, "loss": 2.6326, "step": 51757 }, { "epoch": 4.06, "learning_rate": 5.941503959852583e-05, "loss": 2.4848, "step": 51758 }, { "epoch": 4.06, "learning_rate": 5.941425546930134e-05, "loss": 2.6934, "step": 51759 }, { "epoch": 4.06, "learning_rate": 5.941347134007684e-05, "loss": 2.6699, "step": 51760 }, { "epoch": 4.06, "learning_rate": 5.9412687210852344e-05, "loss": 2.7275, "step": 51761 }, { "epoch": 4.06, "learning_rate": 5.941190308162785e-05, "loss": 2.6709, "step": 51762 }, { "epoch": 4.06, "learning_rate": 5.941111895240336e-05, "loss": 2.8388, "step": 51763 }, { "epoch": 4.06, "learning_rate": 5.941033482317886e-05, "loss": 2.7026, "step": 51764 }, { "epoch": 4.06, "learning_rate": 5.9409550693954366e-05, "loss": 2.8321, "step": 51765 }, { "epoch": 4.06, "learning_rate": 5.940876656472987e-05, "loss": 2.671, "step": 51766 }, { "epoch": 4.06, "learning_rate": 5.940798243550537e-05, "loss": 2.6289, "step": 51767 }, { "epoch": 4.06, "learning_rate": 5.940719830628088e-05, "loss": 2.7621, "step": 51768 }, { "epoch": 4.06, "learning_rate": 5.940641417705638e-05, "loss": 2.7102, "step": 51769 }, { "epoch": 4.06, "learning_rate": 5.9405630047831884e-05, "loss": 2.7068, "step": 51770 }, { "epoch": 4.06, "learning_rate": 5.940484591860739e-05, "loss": 2.6021, "step": 51771 }, { "epoch": 4.06, "learning_rate": 5.940406178938289e-05, "loss": 2.6846, "step": 51772 }, { "epoch": 4.06, "learning_rate": 5.9403277660158395e-05, "loss": 2.5285, "step": 51773 }, { "epoch": 4.06, "learning_rate": 5.94024935309339e-05, "loss": 2.6585, "step": 51774 }, { "epoch": 4.06, "learning_rate": 5.94017094017094e-05, "loss": 2.5245, "step": 51775 }, { "epoch": 4.06, "learning_rate": 5.9400925272484906e-05, "loss": 2.7497, "step": 51776 }, { "epoch": 4.06, "learning_rate": 5.940014114326041e-05, "loss": 2.6606, "step": 51777 }, { "epoch": 4.06, "learning_rate": 5.9399357014035914e-05, "loss": 2.7166, "step": 51778 }, { "epoch": 4.06, "learning_rate": 5.9398572884811424e-05, "loss": 2.5208, "step": 51779 }, { "epoch": 4.06, "learning_rate": 5.939778875558693e-05, "loss": 2.6553, "step": 51780 }, { "epoch": 4.06, "learning_rate": 5.939700462636243e-05, "loss": 2.7976, "step": 51781 }, { "epoch": 4.06, "learning_rate": 5.9396220497137935e-05, "loss": 2.6162, "step": 51782 }, { "epoch": 4.06, "learning_rate": 5.939543636791344e-05, "loss": 2.5823, "step": 51783 }, { "epoch": 4.06, "learning_rate": 5.939465223868894e-05, "loss": 2.6638, "step": 51784 }, { "epoch": 4.06, "learning_rate": 5.9393868109464446e-05, "loss": 2.6587, "step": 51785 }, { "epoch": 4.06, "learning_rate": 5.939308398023995e-05, "loss": 2.7337, "step": 51786 }, { "epoch": 4.06, "learning_rate": 5.939229985101545e-05, "loss": 2.6248, "step": 51787 }, { "epoch": 4.06, "learning_rate": 5.939151572179096e-05, "loss": 3.0353, "step": 51788 }, { "epoch": 4.06, "learning_rate": 5.939073159256646e-05, "loss": 2.5982, "step": 51789 }, { "epoch": 4.06, "learning_rate": 5.9389947463341964e-05, "loss": 2.8578, "step": 51790 }, { "epoch": 4.06, "learning_rate": 5.938916333411747e-05, "loss": 2.5026, "step": 51791 }, { "epoch": 4.06, "learning_rate": 5.938837920489297e-05, "loss": 2.829, "step": 51792 }, { "epoch": 4.06, "learning_rate": 5.9387595075668475e-05, "loss": 2.7018, "step": 51793 }, { "epoch": 4.06, "learning_rate": 5.938681094644397e-05, "loss": 2.8002, "step": 51794 }, { "epoch": 4.06, "learning_rate": 5.9386026817219476e-05, "loss": 2.5788, "step": 51795 }, { "epoch": 4.06, "learning_rate": 5.938524268799498e-05, "loss": 2.7813, "step": 51796 }, { "epoch": 4.06, "learning_rate": 5.938445855877048e-05, "loss": 2.6092, "step": 51797 }, { "epoch": 4.06, "learning_rate": 5.938367442954599e-05, "loss": 2.483, "step": 51798 }, { "epoch": 4.06, "learning_rate": 5.938289030032149e-05, "loss": 2.589, "step": 51799 }, { "epoch": 4.06, "learning_rate": 5.9382106171096994e-05, "loss": 2.4419, "step": 51800 }, { "epoch": 4.06, "learning_rate": 5.93813220418725e-05, "loss": 2.4628, "step": 51801 }, { "epoch": 4.06, "learning_rate": 5.9380537912648e-05, "loss": 2.7034, "step": 51802 }, { "epoch": 4.06, "learning_rate": 5.9379753783423505e-05, "loss": 2.7574, "step": 51803 }, { "epoch": 4.06, "learning_rate": 5.9378969654199016e-05, "loss": 2.5448, "step": 51804 }, { "epoch": 4.06, "learning_rate": 5.937818552497452e-05, "loss": 2.5064, "step": 51805 }, { "epoch": 4.06, "learning_rate": 5.937740139575002e-05, "loss": 2.599, "step": 51806 }, { "epoch": 4.06, "learning_rate": 5.937661726652553e-05, "loss": 2.6285, "step": 51807 }, { "epoch": 4.06, "learning_rate": 5.937583313730103e-05, "loss": 2.681, "step": 51808 }, { "epoch": 4.06, "learning_rate": 5.9375049008076534e-05, "loss": 2.7356, "step": 51809 }, { "epoch": 4.06, "learning_rate": 5.937426487885204e-05, "loss": 2.607, "step": 51810 }, { "epoch": 4.06, "learning_rate": 5.937348074962754e-05, "loss": 2.6175, "step": 51811 }, { "epoch": 4.06, "learning_rate": 5.9372696620403045e-05, "loss": 2.7916, "step": 51812 }, { "epoch": 4.06, "learning_rate": 5.937191249117855e-05, "loss": 2.8307, "step": 51813 }, { "epoch": 4.06, "learning_rate": 5.937112836195405e-05, "loss": 2.5244, "step": 51814 }, { "epoch": 4.06, "learning_rate": 5.9370344232729556e-05, "loss": 2.5362, "step": 51815 }, { "epoch": 4.06, "learning_rate": 5.936956010350506e-05, "loss": 2.5379, "step": 51816 }, { "epoch": 4.06, "learning_rate": 5.9368775974280564e-05, "loss": 2.5398, "step": 51817 }, { "epoch": 4.06, "learning_rate": 5.936799184505607e-05, "loss": 2.7579, "step": 51818 }, { "epoch": 4.06, "learning_rate": 5.936720771583157e-05, "loss": 2.8571, "step": 51819 }, { "epoch": 4.06, "learning_rate": 5.936642358660708e-05, "loss": 2.4686, "step": 51820 }, { "epoch": 4.06, "learning_rate": 5.9365639457382585e-05, "loss": 2.6594, "step": 51821 }, { "epoch": 4.06, "learning_rate": 5.936485532815809e-05, "loss": 2.6829, "step": 51822 }, { "epoch": 4.06, "learning_rate": 5.936407119893359e-05, "loss": 2.6377, "step": 51823 }, { "epoch": 4.06, "learning_rate": 5.9363287069709096e-05, "loss": 2.7042, "step": 51824 }, { "epoch": 4.06, "learning_rate": 5.93625029404846e-05, "loss": 2.6831, "step": 51825 }, { "epoch": 4.06, "learning_rate": 5.9361718811260104e-05, "loss": 2.4199, "step": 51826 }, { "epoch": 4.06, "learning_rate": 5.936093468203561e-05, "loss": 2.8458, "step": 51827 }, { "epoch": 4.06, "learning_rate": 5.936015055281111e-05, "loss": 2.5159, "step": 51828 }, { "epoch": 4.06, "learning_rate": 5.9359366423586615e-05, "loss": 2.755, "step": 51829 }, { "epoch": 4.06, "learning_rate": 5.935858229436211e-05, "loss": 2.7609, "step": 51830 }, { "epoch": 4.06, "learning_rate": 5.9357798165137615e-05, "loss": 2.8532, "step": 51831 }, { "epoch": 4.06, "learning_rate": 5.935701403591312e-05, "loss": 2.71, "step": 51832 }, { "epoch": 4.06, "learning_rate": 5.935622990668862e-05, "loss": 2.5881, "step": 51833 }, { "epoch": 4.06, "learning_rate": 5.9355445777464126e-05, "loss": 2.5968, "step": 51834 }, { "epoch": 4.06, "learning_rate": 5.935466164823963e-05, "loss": 2.6981, "step": 51835 }, { "epoch": 4.06, "learning_rate": 5.9353877519015134e-05, "loss": 2.5767, "step": 51836 }, { "epoch": 4.06, "learning_rate": 5.935309338979064e-05, "loss": 2.5295, "step": 51837 }, { "epoch": 4.06, "learning_rate": 5.935230926056614e-05, "loss": 2.5913, "step": 51838 }, { "epoch": 4.06, "learning_rate": 5.9351525131341645e-05, "loss": 2.6715, "step": 51839 }, { "epoch": 4.06, "learning_rate": 5.935074100211715e-05, "loss": 2.5882, "step": 51840 }, { "epoch": 4.07, "learning_rate": 5.934995687289265e-05, "loss": 2.5106, "step": 51841 }, { "epoch": 4.07, "learning_rate": 5.9349172743668156e-05, "loss": 2.6092, "step": 51842 }, { "epoch": 4.07, "learning_rate": 5.934838861444366e-05, "loss": 2.5329, "step": 51843 }, { "epoch": 4.07, "learning_rate": 5.934760448521916e-05, "loss": 2.5244, "step": 51844 }, { "epoch": 4.07, "learning_rate": 5.934682035599467e-05, "loss": 2.9633, "step": 51845 }, { "epoch": 4.07, "learning_rate": 5.934603622677018e-05, "loss": 2.7395, "step": 51846 }, { "epoch": 4.07, "learning_rate": 5.934525209754568e-05, "loss": 2.5873, "step": 51847 }, { "epoch": 4.07, "learning_rate": 5.9344467968321184e-05, "loss": 2.7761, "step": 51848 }, { "epoch": 4.07, "learning_rate": 5.934368383909669e-05, "loss": 2.4447, "step": 51849 }, { "epoch": 4.07, "learning_rate": 5.934289970987219e-05, "loss": 2.5774, "step": 51850 }, { "epoch": 4.07, "learning_rate": 5.9342115580647696e-05, "loss": 2.7886, "step": 51851 }, { "epoch": 4.07, "learning_rate": 5.93413314514232e-05, "loss": 2.6496, "step": 51852 }, { "epoch": 4.07, "learning_rate": 5.93405473221987e-05, "loss": 2.6084, "step": 51853 }, { "epoch": 4.07, "learning_rate": 5.9339763192974207e-05, "loss": 2.6001, "step": 51854 }, { "epoch": 4.07, "learning_rate": 5.933897906374971e-05, "loss": 2.7027, "step": 51855 }, { "epoch": 4.07, "learning_rate": 5.9338194934525214e-05, "loss": 2.5831, "step": 51856 }, { "epoch": 4.07, "learning_rate": 5.933741080530072e-05, "loss": 2.5707, "step": 51857 }, { "epoch": 4.07, "learning_rate": 5.933662667607622e-05, "loss": 2.6927, "step": 51858 }, { "epoch": 4.07, "learning_rate": 5.9335842546851725e-05, "loss": 2.6742, "step": 51859 }, { "epoch": 4.07, "learning_rate": 5.933505841762723e-05, "loss": 2.6839, "step": 51860 }, { "epoch": 4.07, "learning_rate": 5.933427428840274e-05, "loss": 2.7952, "step": 51861 }, { "epoch": 4.07, "learning_rate": 5.933349015917824e-05, "loss": 2.711, "step": 51862 }, { "epoch": 4.07, "learning_rate": 5.9332706029953746e-05, "loss": 2.6561, "step": 51863 }, { "epoch": 4.07, "learning_rate": 5.933192190072925e-05, "loss": 2.4791, "step": 51864 }, { "epoch": 4.07, "learning_rate": 5.9331137771504754e-05, "loss": 2.5702, "step": 51865 }, { "epoch": 4.07, "learning_rate": 5.9330353642280244e-05, "loss": 2.6562, "step": 51866 }, { "epoch": 4.07, "learning_rate": 5.932956951305575e-05, "loss": 2.7032, "step": 51867 }, { "epoch": 4.07, "learning_rate": 5.932878538383125e-05, "loss": 2.7018, "step": 51868 }, { "epoch": 4.07, "learning_rate": 5.9328001254606755e-05, "loss": 2.7063, "step": 51869 }, { "epoch": 4.07, "learning_rate": 5.932721712538226e-05, "loss": 2.7417, "step": 51870 }, { "epoch": 4.07, "learning_rate": 5.932643299615777e-05, "loss": 2.5529, "step": 51871 }, { "epoch": 4.07, "learning_rate": 5.932564886693327e-05, "loss": 2.8112, "step": 51872 }, { "epoch": 4.07, "learning_rate": 5.9324864737708776e-05, "loss": 2.6281, "step": 51873 }, { "epoch": 4.07, "learning_rate": 5.932408060848428e-05, "loss": 2.6172, "step": 51874 }, { "epoch": 4.07, "learning_rate": 5.9323296479259784e-05, "loss": 2.5856, "step": 51875 }, { "epoch": 4.07, "learning_rate": 5.932251235003529e-05, "loss": 2.5727, "step": 51876 }, { "epoch": 4.07, "learning_rate": 5.932172822081079e-05, "loss": 2.4615, "step": 51877 }, { "epoch": 4.07, "learning_rate": 5.9320944091586295e-05, "loss": 2.6676, "step": 51878 }, { "epoch": 4.07, "learning_rate": 5.93201599623618e-05, "loss": 2.6118, "step": 51879 }, { "epoch": 4.07, "learning_rate": 5.93193758331373e-05, "loss": 2.7845, "step": 51880 }, { "epoch": 4.07, "learning_rate": 5.9318591703912806e-05, "loss": 2.8131, "step": 51881 }, { "epoch": 4.07, "learning_rate": 5.931780757468831e-05, "loss": 2.8723, "step": 51882 }, { "epoch": 4.07, "learning_rate": 5.931702344546381e-05, "loss": 2.8206, "step": 51883 }, { "epoch": 4.07, "learning_rate": 5.931623931623932e-05, "loss": 2.7065, "step": 51884 }, { "epoch": 4.07, "learning_rate": 5.931545518701482e-05, "loss": 2.8007, "step": 51885 }, { "epoch": 4.07, "learning_rate": 5.9314671057790324e-05, "loss": 2.7613, "step": 51886 }, { "epoch": 4.07, "learning_rate": 5.9313886928565835e-05, "loss": 2.5434, "step": 51887 }, { "epoch": 4.07, "learning_rate": 5.931310279934134e-05, "loss": 2.7485, "step": 51888 }, { "epoch": 4.07, "learning_rate": 5.931231867011684e-05, "loss": 2.5444, "step": 51889 }, { "epoch": 4.07, "learning_rate": 5.9311534540892346e-05, "loss": 2.6924, "step": 51890 }, { "epoch": 4.07, "learning_rate": 5.931075041166785e-05, "loss": 2.588, "step": 51891 }, { "epoch": 4.07, "learning_rate": 5.930996628244335e-05, "loss": 2.5105, "step": 51892 }, { "epoch": 4.07, "learning_rate": 5.930918215321886e-05, "loss": 2.7273, "step": 51893 }, { "epoch": 4.07, "learning_rate": 5.930839802399436e-05, "loss": 2.7753, "step": 51894 }, { "epoch": 4.07, "learning_rate": 5.9307613894769864e-05, "loss": 2.9066, "step": 51895 }, { "epoch": 4.07, "learning_rate": 5.930682976554537e-05, "loss": 2.7089, "step": 51896 }, { "epoch": 4.07, "learning_rate": 5.930604563632087e-05, "loss": 2.6396, "step": 51897 }, { "epoch": 4.07, "learning_rate": 5.9305261507096375e-05, "loss": 2.7129, "step": 51898 }, { "epoch": 4.07, "learning_rate": 5.930447737787188e-05, "loss": 2.7187, "step": 51899 }, { "epoch": 4.07, "learning_rate": 5.930369324864738e-05, "loss": 2.9104, "step": 51900 }, { "epoch": 4.07, "learning_rate": 5.9302909119422886e-05, "loss": 2.5569, "step": 51901 }, { "epoch": 4.07, "learning_rate": 5.930212499019838e-05, "loss": 2.6957, "step": 51902 }, { "epoch": 4.07, "learning_rate": 5.930134086097389e-05, "loss": 2.5603, "step": 51903 }, { "epoch": 4.07, "learning_rate": 5.930055673174939e-05, "loss": 2.5474, "step": 51904 }, { "epoch": 4.07, "learning_rate": 5.9299772602524894e-05, "loss": 2.7479, "step": 51905 }, { "epoch": 4.07, "learning_rate": 5.92989884733004e-05, "loss": 2.599, "step": 51906 }, { "epoch": 4.07, "learning_rate": 5.92982043440759e-05, "loss": 2.7011, "step": 51907 }, { "epoch": 4.07, "learning_rate": 5.9297420214851405e-05, "loss": 2.533, "step": 51908 }, { "epoch": 4.07, "learning_rate": 5.929663608562691e-05, "loss": 2.6748, "step": 51909 }, { "epoch": 4.07, "learning_rate": 5.929585195640241e-05, "loss": 2.6241, "step": 51910 }, { "epoch": 4.07, "learning_rate": 5.9295067827177916e-05, "loss": 2.5562, "step": 51911 }, { "epoch": 4.07, "learning_rate": 5.929428369795343e-05, "loss": 2.5478, "step": 51912 }, { "epoch": 4.07, "learning_rate": 5.929349956872893e-05, "loss": 2.5938, "step": 51913 }, { "epoch": 4.07, "learning_rate": 5.9292715439504434e-05, "loss": 2.6505, "step": 51914 }, { "epoch": 4.07, "learning_rate": 5.929193131027994e-05, "loss": 2.7116, "step": 51915 }, { "epoch": 4.07, "learning_rate": 5.929114718105544e-05, "loss": 2.5781, "step": 51916 }, { "epoch": 4.07, "learning_rate": 5.9290363051830945e-05, "loss": 2.7796, "step": 51917 }, { "epoch": 4.07, "learning_rate": 5.928957892260645e-05, "loss": 2.6028, "step": 51918 }, { "epoch": 4.07, "learning_rate": 5.928879479338195e-05, "loss": 2.5979, "step": 51919 }, { "epoch": 4.07, "learning_rate": 5.9288010664157456e-05, "loss": 2.68, "step": 51920 }, { "epoch": 4.07, "learning_rate": 5.928722653493296e-05, "loss": 2.6687, "step": 51921 }, { "epoch": 4.07, "learning_rate": 5.9286442405708463e-05, "loss": 2.6644, "step": 51922 }, { "epoch": 4.07, "learning_rate": 5.928565827648397e-05, "loss": 2.5751, "step": 51923 }, { "epoch": 4.07, "learning_rate": 5.928487414725947e-05, "loss": 2.8336, "step": 51924 }, { "epoch": 4.07, "learning_rate": 5.9284090018034974e-05, "loss": 2.6963, "step": 51925 }, { "epoch": 4.07, "learning_rate": 5.928330588881048e-05, "loss": 2.5194, "step": 51926 }, { "epoch": 4.07, "learning_rate": 5.928252175958598e-05, "loss": 2.578, "step": 51927 }, { "epoch": 4.07, "learning_rate": 5.928173763036149e-05, "loss": 2.8552, "step": 51928 }, { "epoch": 4.07, "learning_rate": 5.9280953501136996e-05, "loss": 2.5932, "step": 51929 }, { "epoch": 4.07, "learning_rate": 5.92801693719125e-05, "loss": 2.6, "step": 51930 }, { "epoch": 4.07, "learning_rate": 5.9279385242688e-05, "loss": 2.5933, "step": 51931 }, { "epoch": 4.07, "learning_rate": 5.927860111346351e-05, "loss": 2.5253, "step": 51932 }, { "epoch": 4.07, "learning_rate": 5.927781698423901e-05, "loss": 2.6403, "step": 51933 }, { "epoch": 4.07, "learning_rate": 5.9277032855014514e-05, "loss": 2.5319, "step": 51934 }, { "epoch": 4.07, "learning_rate": 5.927624872579002e-05, "loss": 2.8683, "step": 51935 }, { "epoch": 4.07, "learning_rate": 5.927546459656552e-05, "loss": 2.6447, "step": 51936 }, { "epoch": 4.07, "learning_rate": 5.9274680467341025e-05, "loss": 2.6293, "step": 51937 }, { "epoch": 4.07, "learning_rate": 5.927389633811652e-05, "loss": 2.5481, "step": 51938 }, { "epoch": 4.07, "learning_rate": 5.9273112208892026e-05, "loss": 2.6848, "step": 51939 }, { "epoch": 4.07, "learning_rate": 5.927232807966753e-05, "loss": 2.4776, "step": 51940 }, { "epoch": 4.07, "learning_rate": 5.927154395044303e-05, "loss": 2.4826, "step": 51941 }, { "epoch": 4.07, "learning_rate": 5.927075982121854e-05, "loss": 2.5635, "step": 51942 }, { "epoch": 4.07, "learning_rate": 5.926997569199404e-05, "loss": 2.8854, "step": 51943 }, { "epoch": 4.07, "learning_rate": 5.9269191562769544e-05, "loss": 2.8583, "step": 51944 }, { "epoch": 4.07, "learning_rate": 5.926840743354505e-05, "loss": 2.5894, "step": 51945 }, { "epoch": 4.07, "learning_rate": 5.926762330432055e-05, "loss": 2.6804, "step": 51946 }, { "epoch": 4.07, "learning_rate": 5.9266839175096055e-05, "loss": 2.7755, "step": 51947 }, { "epoch": 4.07, "learning_rate": 5.926605504587156e-05, "loss": 2.6296, "step": 51948 }, { "epoch": 4.07, "learning_rate": 5.926527091664706e-05, "loss": 2.5777, "step": 51949 }, { "epoch": 4.07, "learning_rate": 5.9264486787422566e-05, "loss": 2.6888, "step": 51950 }, { "epoch": 4.07, "learning_rate": 5.926370265819807e-05, "loss": 2.5102, "step": 51951 }, { "epoch": 4.07, "learning_rate": 5.9262918528973574e-05, "loss": 2.5317, "step": 51952 }, { "epoch": 4.07, "learning_rate": 5.926213439974908e-05, "loss": 2.7057, "step": 51953 }, { "epoch": 4.07, "learning_rate": 5.926135027052459e-05, "loss": 2.5979, "step": 51954 }, { "epoch": 4.07, "learning_rate": 5.926056614130009e-05, "loss": 2.6206, "step": 51955 }, { "epoch": 4.07, "learning_rate": 5.9259782012075595e-05, "loss": 2.5737, "step": 51956 }, { "epoch": 4.07, "learning_rate": 5.92589978828511e-05, "loss": 2.6203, "step": 51957 }, { "epoch": 4.07, "learning_rate": 5.92582137536266e-05, "loss": 2.5278, "step": 51958 }, { "epoch": 4.07, "learning_rate": 5.9257429624402106e-05, "loss": 2.6979, "step": 51959 }, { "epoch": 4.07, "learning_rate": 5.925664549517761e-05, "loss": 2.6358, "step": 51960 }, { "epoch": 4.07, "learning_rate": 5.9255861365953114e-05, "loss": 2.5351, "step": 51961 }, { "epoch": 4.07, "learning_rate": 5.925507723672862e-05, "loss": 2.722, "step": 51962 }, { "epoch": 4.07, "learning_rate": 5.925429310750412e-05, "loss": 2.4822, "step": 51963 }, { "epoch": 4.07, "learning_rate": 5.9253508978279625e-05, "loss": 2.7219, "step": 51964 }, { "epoch": 4.07, "learning_rate": 5.925272484905513e-05, "loss": 2.6876, "step": 51965 }, { "epoch": 4.07, "learning_rate": 5.925194071983063e-05, "loss": 2.7534, "step": 51966 }, { "epoch": 4.07, "learning_rate": 5.9251156590606136e-05, "loss": 2.5883, "step": 51967 }, { "epoch": 4.07, "learning_rate": 5.925037246138164e-05, "loss": 2.5544, "step": 51968 }, { "epoch": 4.08, "learning_rate": 5.924958833215715e-05, "loss": 2.5468, "step": 51969 }, { "epoch": 4.08, "learning_rate": 5.9248804202932654e-05, "loss": 2.5544, "step": 51970 }, { "epoch": 4.08, "learning_rate": 5.924802007370816e-05, "loss": 2.7168, "step": 51971 }, { "epoch": 4.08, "learning_rate": 5.924723594448366e-05, "loss": 2.5146, "step": 51972 }, { "epoch": 4.08, "learning_rate": 5.9246451815259165e-05, "loss": 2.7498, "step": 51973 }, { "epoch": 4.08, "learning_rate": 5.9245667686034655e-05, "loss": 2.6129, "step": 51974 }, { "epoch": 4.08, "learning_rate": 5.924488355681016e-05, "loss": 2.5322, "step": 51975 }, { "epoch": 4.08, "learning_rate": 5.924409942758566e-05, "loss": 2.7235, "step": 51976 }, { "epoch": 4.08, "learning_rate": 5.9243315298361166e-05, "loss": 2.8023, "step": 51977 }, { "epoch": 4.08, "learning_rate": 5.924253116913667e-05, "loss": 2.4554, "step": 51978 }, { "epoch": 4.08, "learning_rate": 5.924174703991218e-05, "loss": 2.6152, "step": 51979 }, { "epoch": 4.08, "learning_rate": 5.9240962910687684e-05, "loss": 2.4573, "step": 51980 }, { "epoch": 4.08, "learning_rate": 5.924017878146319e-05, "loss": 2.6357, "step": 51981 }, { "epoch": 4.08, "learning_rate": 5.923939465223869e-05, "loss": 2.8412, "step": 51982 }, { "epoch": 4.08, "learning_rate": 5.9238610523014195e-05, "loss": 2.5353, "step": 51983 }, { "epoch": 4.08, "learning_rate": 5.92378263937897e-05, "loss": 2.7278, "step": 51984 }, { "epoch": 4.08, "learning_rate": 5.92370422645652e-05, "loss": 2.5977, "step": 51985 }, { "epoch": 4.08, "learning_rate": 5.9236258135340706e-05, "loss": 2.5611, "step": 51986 }, { "epoch": 4.08, "learning_rate": 5.923547400611621e-05, "loss": 2.6785, "step": 51987 }, { "epoch": 4.08, "learning_rate": 5.923468987689171e-05, "loss": 2.6921, "step": 51988 }, { "epoch": 4.08, "learning_rate": 5.9233905747667217e-05, "loss": 2.6187, "step": 51989 }, { "epoch": 4.08, "learning_rate": 5.923312161844272e-05, "loss": 2.4978, "step": 51990 }, { "epoch": 4.08, "learning_rate": 5.9232337489218224e-05, "loss": 2.5025, "step": 51991 }, { "epoch": 4.08, "learning_rate": 5.923155335999373e-05, "loss": 2.5005, "step": 51992 }, { "epoch": 4.08, "learning_rate": 5.923076923076923e-05, "loss": 2.7776, "step": 51993 }, { "epoch": 4.08, "learning_rate": 5.9229985101544735e-05, "loss": 2.769, "step": 51994 }, { "epoch": 4.08, "learning_rate": 5.9229200972320245e-05, "loss": 2.5212, "step": 51995 }, { "epoch": 4.08, "learning_rate": 5.922841684309575e-05, "loss": 2.6959, "step": 51996 }, { "epoch": 4.08, "learning_rate": 5.922763271387125e-05, "loss": 2.5506, "step": 51997 }, { "epoch": 4.08, "learning_rate": 5.9226848584646756e-05, "loss": 2.7116, "step": 51998 }, { "epoch": 4.08, "learning_rate": 5.922606445542226e-05, "loss": 2.7079, "step": 51999 }, { "epoch": 4.08, "learning_rate": 5.9225280326197764e-05, "loss": 2.6167, "step": 52000 }, { "epoch": 4.08, "eval_loss": 2.8992769718170166, "eval_runtime": 217.9916, "eval_samples_per_second": 51.984, "eval_steps_per_second": 3.252, "step": 52000 }, { "epoch": 4.08, "learning_rate": 5.922449619697327e-05, "loss": 2.701, "step": 52001 }, { "epoch": 4.08, "learning_rate": 5.922371206774877e-05, "loss": 2.7614, "step": 52002 }, { "epoch": 4.08, "learning_rate": 5.9222927938524275e-05, "loss": 2.6358, "step": 52003 }, { "epoch": 4.08, "learning_rate": 5.922214380929978e-05, "loss": 2.7379, "step": 52004 }, { "epoch": 4.08, "learning_rate": 5.922135968007528e-05, "loss": 2.7635, "step": 52005 }, { "epoch": 4.08, "learning_rate": 5.9220575550850786e-05, "loss": 2.5899, "step": 52006 }, { "epoch": 4.08, "learning_rate": 5.921979142162629e-05, "loss": 2.4767, "step": 52007 }, { "epoch": 4.08, "learning_rate": 5.921900729240179e-05, "loss": 2.7218, "step": 52008 }, { "epoch": 4.08, "learning_rate": 5.92182231631773e-05, "loss": 2.6639, "step": 52009 }, { "epoch": 4.08, "learning_rate": 5.9217439033952794e-05, "loss": 2.6201, "step": 52010 }, { "epoch": 4.08, "learning_rate": 5.92166549047283e-05, "loss": 2.8716, "step": 52011 }, { "epoch": 4.08, "learning_rate": 5.92158707755038e-05, "loss": 2.8139, "step": 52012 }, { "epoch": 4.08, "learning_rate": 5.9215086646279305e-05, "loss": 2.4954, "step": 52013 }, { "epoch": 4.08, "learning_rate": 5.921430251705481e-05, "loss": 2.5507, "step": 52014 }, { "epoch": 4.08, "learning_rate": 5.921351838783031e-05, "loss": 2.5169, "step": 52015 }, { "epoch": 4.08, "learning_rate": 5.9212734258605816e-05, "loss": 2.6602, "step": 52016 }, { "epoch": 4.08, "learning_rate": 5.921195012938132e-05, "loss": 2.6776, "step": 52017 }, { "epoch": 4.08, "learning_rate": 5.921116600015682e-05, "loss": 2.5667, "step": 52018 }, { "epoch": 4.08, "learning_rate": 5.921038187093233e-05, "loss": 2.7348, "step": 52019 }, { "epoch": 4.08, "learning_rate": 5.920959774170784e-05, "loss": 2.6654, "step": 52020 }, { "epoch": 4.08, "learning_rate": 5.920881361248334e-05, "loss": 2.6402, "step": 52021 }, { "epoch": 4.08, "learning_rate": 5.9208029483258845e-05, "loss": 2.6574, "step": 52022 }, { "epoch": 4.08, "learning_rate": 5.920724535403435e-05, "loss": 2.7059, "step": 52023 }, { "epoch": 4.08, "learning_rate": 5.920646122480985e-05, "loss": 2.5822, "step": 52024 }, { "epoch": 4.08, "learning_rate": 5.9205677095585356e-05, "loss": 2.8352, "step": 52025 }, { "epoch": 4.08, "learning_rate": 5.920489296636086e-05, "loss": 2.8342, "step": 52026 }, { "epoch": 4.08, "learning_rate": 5.920410883713636e-05, "loss": 2.7475, "step": 52027 }, { "epoch": 4.08, "learning_rate": 5.920332470791187e-05, "loss": 2.6125, "step": 52028 }, { "epoch": 4.08, "learning_rate": 5.920254057868737e-05, "loss": 2.7145, "step": 52029 }, { "epoch": 4.08, "learning_rate": 5.9201756449462874e-05, "loss": 2.5495, "step": 52030 }, { "epoch": 4.08, "learning_rate": 5.920097232023838e-05, "loss": 2.6459, "step": 52031 }, { "epoch": 4.08, "learning_rate": 5.920018819101388e-05, "loss": 2.5157, "step": 52032 }, { "epoch": 4.08, "learning_rate": 5.9199404061789385e-05, "loss": 2.5403, "step": 52033 }, { "epoch": 4.08, "learning_rate": 5.919861993256489e-05, "loss": 2.6061, "step": 52034 }, { "epoch": 4.08, "learning_rate": 5.919783580334039e-05, "loss": 2.6716, "step": 52035 }, { "epoch": 4.08, "learning_rate": 5.91970516741159e-05, "loss": 2.6446, "step": 52036 }, { "epoch": 4.08, "learning_rate": 5.919626754489141e-05, "loss": 2.5628, "step": 52037 }, { "epoch": 4.08, "learning_rate": 5.919548341566691e-05, "loss": 2.5722, "step": 52038 }, { "epoch": 4.08, "learning_rate": 5.9194699286442414e-05, "loss": 2.6635, "step": 52039 }, { "epoch": 4.08, "learning_rate": 5.919391515721792e-05, "loss": 2.621, "step": 52040 }, { "epoch": 4.08, "learning_rate": 5.919313102799342e-05, "loss": 2.5885, "step": 52041 }, { "epoch": 4.08, "learning_rate": 5.9192346898768925e-05, "loss": 2.7398, "step": 52042 }, { "epoch": 4.08, "learning_rate": 5.919156276954443e-05, "loss": 2.5825, "step": 52043 }, { "epoch": 4.08, "learning_rate": 5.919077864031993e-05, "loss": 2.5185, "step": 52044 }, { "epoch": 4.08, "learning_rate": 5.9189994511095436e-05, "loss": 2.5443, "step": 52045 }, { "epoch": 4.08, "learning_rate": 5.918921038187093e-05, "loss": 2.5455, "step": 52046 }, { "epoch": 4.08, "learning_rate": 5.918842625264644e-05, "loss": 2.5917, "step": 52047 }, { "epoch": 4.08, "learning_rate": 5.918764212342194e-05, "loss": 2.659, "step": 52048 }, { "epoch": 4.08, "learning_rate": 5.9186857994197444e-05, "loss": 2.7292, "step": 52049 }, { "epoch": 4.08, "learning_rate": 5.918607386497295e-05, "loss": 2.7876, "step": 52050 }, { "epoch": 4.08, "learning_rate": 5.918528973574845e-05, "loss": 2.8304, "step": 52051 }, { "epoch": 4.08, "learning_rate": 5.9184505606523955e-05, "loss": 2.6809, "step": 52052 }, { "epoch": 4.08, "learning_rate": 5.918372147729946e-05, "loss": 2.6641, "step": 52053 }, { "epoch": 4.08, "learning_rate": 5.918293734807496e-05, "loss": 2.6593, "step": 52054 }, { "epoch": 4.08, "learning_rate": 5.9182153218850466e-05, "loss": 2.6575, "step": 52055 }, { "epoch": 4.08, "learning_rate": 5.918136908962597e-05, "loss": 2.7412, "step": 52056 }, { "epoch": 4.08, "learning_rate": 5.9180584960401473e-05, "loss": 2.7402, "step": 52057 }, { "epoch": 4.08, "learning_rate": 5.917980083117698e-05, "loss": 2.7679, "step": 52058 }, { "epoch": 4.08, "learning_rate": 5.917901670195248e-05, "loss": 2.624, "step": 52059 }, { "epoch": 4.08, "learning_rate": 5.9178232572727984e-05, "loss": 2.6964, "step": 52060 }, { "epoch": 4.08, "learning_rate": 5.917744844350349e-05, "loss": 2.6261, "step": 52061 }, { "epoch": 4.08, "learning_rate": 5.9176664314279e-05, "loss": 2.9731, "step": 52062 }, { "epoch": 4.08, "learning_rate": 5.91758801850545e-05, "loss": 2.6432, "step": 52063 }, { "epoch": 4.08, "learning_rate": 5.9175096055830006e-05, "loss": 2.8672, "step": 52064 }, { "epoch": 4.08, "learning_rate": 5.917431192660551e-05, "loss": 2.7401, "step": 52065 }, { "epoch": 4.08, "learning_rate": 5.917352779738101e-05, "loss": 2.7034, "step": 52066 }, { "epoch": 4.08, "learning_rate": 5.917274366815652e-05, "loss": 2.6538, "step": 52067 }, { "epoch": 4.08, "learning_rate": 5.917195953893202e-05, "loss": 2.5633, "step": 52068 }, { "epoch": 4.08, "learning_rate": 5.9171175409707524e-05, "loss": 2.6444, "step": 52069 }, { "epoch": 4.08, "learning_rate": 5.917039128048303e-05, "loss": 2.5206, "step": 52070 }, { "epoch": 4.08, "learning_rate": 5.916960715125853e-05, "loss": 2.7762, "step": 52071 }, { "epoch": 4.08, "learning_rate": 5.9168823022034035e-05, "loss": 2.7367, "step": 52072 }, { "epoch": 4.08, "learning_rate": 5.916803889280954e-05, "loss": 2.8161, "step": 52073 }, { "epoch": 4.08, "learning_rate": 5.916725476358504e-05, "loss": 2.5825, "step": 52074 }, { "epoch": 4.08, "learning_rate": 5.9166470634360546e-05, "loss": 2.8926, "step": 52075 }, { "epoch": 4.08, "learning_rate": 5.916568650513605e-05, "loss": 2.6769, "step": 52076 }, { "epoch": 4.08, "learning_rate": 5.916490237591156e-05, "loss": 2.7287, "step": 52077 }, { "epoch": 4.08, "learning_rate": 5.9164118246687064e-05, "loss": 2.8278, "step": 52078 }, { "epoch": 4.08, "learning_rate": 5.916333411746257e-05, "loss": 2.5131, "step": 52079 }, { "epoch": 4.08, "learning_rate": 5.916254998823807e-05, "loss": 2.5905, "step": 52080 }, { "epoch": 4.08, "learning_rate": 5.9161765859013575e-05, "loss": 2.5928, "step": 52081 }, { "epoch": 4.08, "learning_rate": 5.9160981729789065e-05, "loss": 2.5997, "step": 52082 }, { "epoch": 4.08, "learning_rate": 5.916019760056457e-05, "loss": 2.5617, "step": 52083 }, { "epoch": 4.08, "learning_rate": 5.915941347134007e-05, "loss": 2.5372, "step": 52084 }, { "epoch": 4.08, "learning_rate": 5.9158629342115576e-05, "loss": 2.5614, "step": 52085 }, { "epoch": 4.08, "learning_rate": 5.915784521289108e-05, "loss": 2.662, "step": 52086 }, { "epoch": 4.08, "learning_rate": 5.915706108366659e-05, "loss": 2.5794, "step": 52087 }, { "epoch": 4.08, "learning_rate": 5.9156276954442094e-05, "loss": 2.471, "step": 52088 }, { "epoch": 4.08, "learning_rate": 5.91554928252176e-05, "loss": 2.6851, "step": 52089 }, { "epoch": 4.08, "learning_rate": 5.91547086959931e-05, "loss": 2.7074, "step": 52090 }, { "epoch": 4.08, "learning_rate": 5.9153924566768605e-05, "loss": 2.5047, "step": 52091 }, { "epoch": 4.08, "learning_rate": 5.915314043754411e-05, "loss": 2.6002, "step": 52092 }, { "epoch": 4.08, "learning_rate": 5.915235630831961e-05, "loss": 2.5806, "step": 52093 }, { "epoch": 4.08, "learning_rate": 5.9151572179095116e-05, "loss": 2.8083, "step": 52094 }, { "epoch": 4.08, "learning_rate": 5.915078804987062e-05, "loss": 2.6661, "step": 52095 }, { "epoch": 4.08, "learning_rate": 5.9150003920646124e-05, "loss": 2.5287, "step": 52096 }, { "epoch": 4.09, "learning_rate": 5.914921979142163e-05, "loss": 2.6224, "step": 52097 }, { "epoch": 4.09, "learning_rate": 5.914843566219713e-05, "loss": 2.6386, "step": 52098 }, { "epoch": 4.09, "learning_rate": 5.9147651532972635e-05, "loss": 2.6851, "step": 52099 }, { "epoch": 4.09, "learning_rate": 5.914686740374814e-05, "loss": 2.8757, "step": 52100 }, { "epoch": 4.09, "learning_rate": 5.914608327452364e-05, "loss": 2.4403, "step": 52101 }, { "epoch": 4.09, "learning_rate": 5.9145299145299146e-05, "loss": 2.5202, "step": 52102 }, { "epoch": 4.09, "learning_rate": 5.9144515016074656e-05, "loss": 2.7219, "step": 52103 }, { "epoch": 4.09, "learning_rate": 5.914373088685016e-05, "loss": 2.6621, "step": 52104 }, { "epoch": 4.09, "learning_rate": 5.9142946757625664e-05, "loss": 2.7064, "step": 52105 }, { "epoch": 4.09, "learning_rate": 5.914216262840117e-05, "loss": 2.555, "step": 52106 }, { "epoch": 4.09, "learning_rate": 5.914137849917667e-05, "loss": 2.9553, "step": 52107 }, { "epoch": 4.09, "learning_rate": 5.9140594369952175e-05, "loss": 2.4887, "step": 52108 }, { "epoch": 4.09, "learning_rate": 5.913981024072768e-05, "loss": 2.5806, "step": 52109 }, { "epoch": 4.09, "learning_rate": 5.913902611150318e-05, "loss": 2.7539, "step": 52110 }, { "epoch": 4.09, "learning_rate": 5.9138241982278686e-05, "loss": 2.7566, "step": 52111 }, { "epoch": 4.09, "learning_rate": 5.913745785305419e-05, "loss": 2.4875, "step": 52112 }, { "epoch": 4.09, "learning_rate": 5.913667372382969e-05, "loss": 2.7898, "step": 52113 }, { "epoch": 4.09, "learning_rate": 5.91358895946052e-05, "loss": 2.7326, "step": 52114 }, { "epoch": 4.09, "learning_rate": 5.91351054653807e-05, "loss": 2.9069, "step": 52115 }, { "epoch": 4.09, "learning_rate": 5.9134321336156204e-05, "loss": 2.5818, "step": 52116 }, { "epoch": 4.09, "learning_rate": 5.913353720693171e-05, "loss": 2.8347, "step": 52117 }, { "epoch": 4.09, "learning_rate": 5.9132753077707205e-05, "loss": 2.7738, "step": 52118 }, { "epoch": 4.09, "learning_rate": 5.913196894848271e-05, "loss": 2.7182, "step": 52119 }, { "epoch": 4.09, "learning_rate": 5.913118481925821e-05, "loss": 2.5306, "step": 52120 }, { "epoch": 4.09, "learning_rate": 5.9130400690033716e-05, "loss": 2.6426, "step": 52121 }, { "epoch": 4.09, "learning_rate": 5.912961656080922e-05, "loss": 2.5713, "step": 52122 }, { "epoch": 4.09, "learning_rate": 5.912883243158472e-05, "loss": 2.6723, "step": 52123 }, { "epoch": 4.09, "learning_rate": 5.912804830236023e-05, "loss": 2.5086, "step": 52124 }, { "epoch": 4.09, "learning_rate": 5.912726417313573e-05, "loss": 2.6689, "step": 52125 }, { "epoch": 4.09, "learning_rate": 5.9126480043911234e-05, "loss": 2.7526, "step": 52126 }, { "epoch": 4.09, "learning_rate": 5.912569591468674e-05, "loss": 2.6561, "step": 52127 }, { "epoch": 4.09, "learning_rate": 5.912491178546224e-05, "loss": 2.8146, "step": 52128 }, { "epoch": 4.09, "learning_rate": 5.912412765623775e-05, "loss": 2.6642, "step": 52129 }, { "epoch": 4.09, "learning_rate": 5.9123343527013255e-05, "loss": 2.7417, "step": 52130 }, { "epoch": 4.09, "learning_rate": 5.912255939778876e-05, "loss": 2.5374, "step": 52131 }, { "epoch": 4.09, "learning_rate": 5.912177526856426e-05, "loss": 2.6171, "step": 52132 }, { "epoch": 4.09, "learning_rate": 5.9120991139339767e-05, "loss": 2.6167, "step": 52133 }, { "epoch": 4.09, "learning_rate": 5.912020701011527e-05, "loss": 2.6535, "step": 52134 }, { "epoch": 4.09, "learning_rate": 5.9119422880890774e-05, "loss": 2.6756, "step": 52135 }, { "epoch": 4.09, "learning_rate": 5.911863875166628e-05, "loss": 2.7527, "step": 52136 }, { "epoch": 4.09, "learning_rate": 5.911785462244178e-05, "loss": 2.5722, "step": 52137 }, { "epoch": 4.09, "learning_rate": 5.9117070493217285e-05, "loss": 2.7075, "step": 52138 }, { "epoch": 4.09, "learning_rate": 5.911628636399279e-05, "loss": 2.5353, "step": 52139 }, { "epoch": 4.09, "learning_rate": 5.911550223476829e-05, "loss": 2.7323, "step": 52140 }, { "epoch": 4.09, "learning_rate": 5.9114718105543796e-05, "loss": 2.9344, "step": 52141 }, { "epoch": 4.09, "learning_rate": 5.91139339763193e-05, "loss": 2.6086, "step": 52142 }, { "epoch": 4.09, "learning_rate": 5.91131498470948e-05, "loss": 2.6288, "step": 52143 }, { "epoch": 4.09, "learning_rate": 5.9112365717870314e-05, "loss": 2.6987, "step": 52144 }, { "epoch": 4.09, "learning_rate": 5.911158158864582e-05, "loss": 2.9184, "step": 52145 }, { "epoch": 4.09, "learning_rate": 5.911079745942132e-05, "loss": 2.6179, "step": 52146 }, { "epoch": 4.09, "learning_rate": 5.9110013330196825e-05, "loss": 2.5753, "step": 52147 }, { "epoch": 4.09, "learning_rate": 5.910922920097233e-05, "loss": 2.6416, "step": 52148 }, { "epoch": 4.09, "learning_rate": 5.910844507174783e-05, "loss": 2.6457, "step": 52149 }, { "epoch": 4.09, "learning_rate": 5.9107660942523336e-05, "loss": 2.6059, "step": 52150 }, { "epoch": 4.09, "learning_rate": 5.910687681329884e-05, "loss": 2.6979, "step": 52151 }, { "epoch": 4.09, "learning_rate": 5.910609268407434e-05, "loss": 2.5067, "step": 52152 }, { "epoch": 4.09, "learning_rate": 5.910530855484985e-05, "loss": 2.6218, "step": 52153 }, { "epoch": 4.09, "learning_rate": 5.9104524425625344e-05, "loss": 2.6418, "step": 52154 }, { "epoch": 4.09, "learning_rate": 5.910374029640085e-05, "loss": 2.8072, "step": 52155 }, { "epoch": 4.09, "learning_rate": 5.910295616717635e-05, "loss": 2.6573, "step": 52156 }, { "epoch": 4.09, "learning_rate": 5.9102172037951855e-05, "loss": 2.5746, "step": 52157 }, { "epoch": 4.09, "learning_rate": 5.910138790872736e-05, "loss": 2.631, "step": 52158 }, { "epoch": 4.09, "learning_rate": 5.910060377950286e-05, "loss": 2.7076, "step": 52159 }, { "epoch": 4.09, "learning_rate": 5.9099819650278366e-05, "loss": 2.5218, "step": 52160 }, { "epoch": 4.09, "learning_rate": 5.909903552105387e-05, "loss": 2.5632, "step": 52161 }, { "epoch": 4.09, "learning_rate": 5.909825139182937e-05, "loss": 2.5687, "step": 52162 }, { "epoch": 4.09, "learning_rate": 5.909746726260488e-05, "loss": 2.5955, "step": 52163 }, { "epoch": 4.09, "learning_rate": 5.909668313338038e-05, "loss": 2.6717, "step": 52164 }, { "epoch": 4.09, "learning_rate": 5.9095899004155884e-05, "loss": 2.8157, "step": 52165 }, { "epoch": 4.09, "learning_rate": 5.909511487493139e-05, "loss": 2.4294, "step": 52166 }, { "epoch": 4.09, "learning_rate": 5.909433074570689e-05, "loss": 2.6303, "step": 52167 }, { "epoch": 4.09, "learning_rate": 5.9093546616482395e-05, "loss": 2.685, "step": 52168 }, { "epoch": 4.09, "learning_rate": 5.90927624872579e-05, "loss": 2.9825, "step": 52169 }, { "epoch": 4.09, "learning_rate": 5.909197835803341e-05, "loss": 2.7032, "step": 52170 }, { "epoch": 4.09, "learning_rate": 5.909119422880891e-05, "loss": 2.4956, "step": 52171 }, { "epoch": 4.09, "learning_rate": 5.909041009958442e-05, "loss": 2.7273, "step": 52172 }, { "epoch": 4.09, "learning_rate": 5.908962597035992e-05, "loss": 2.6608, "step": 52173 }, { "epoch": 4.09, "learning_rate": 5.9088841841135424e-05, "loss": 2.7868, "step": 52174 }, { "epoch": 4.09, "learning_rate": 5.908805771191093e-05, "loss": 2.6601, "step": 52175 }, { "epoch": 4.09, "learning_rate": 5.908727358268643e-05, "loss": 2.634, "step": 52176 }, { "epoch": 4.09, "learning_rate": 5.9086489453461935e-05, "loss": 3.0298, "step": 52177 }, { "epoch": 4.09, "learning_rate": 5.908570532423744e-05, "loss": 2.7471, "step": 52178 }, { "epoch": 4.09, "learning_rate": 5.908492119501294e-05, "loss": 2.5447, "step": 52179 }, { "epoch": 4.09, "learning_rate": 5.9084137065788446e-05, "loss": 2.78, "step": 52180 }, { "epoch": 4.09, "learning_rate": 5.908335293656395e-05, "loss": 2.7192, "step": 52181 }, { "epoch": 4.09, "learning_rate": 5.9082568807339454e-05, "loss": 2.5425, "step": 52182 }, { "epoch": 4.09, "learning_rate": 5.908178467811496e-05, "loss": 2.6636, "step": 52183 }, { "epoch": 4.09, "learning_rate": 5.908100054889046e-05, "loss": 2.3914, "step": 52184 }, { "epoch": 4.09, "learning_rate": 5.908021641966597e-05, "loss": 2.5059, "step": 52185 }, { "epoch": 4.09, "learning_rate": 5.9079432290441475e-05, "loss": 2.7427, "step": 52186 }, { "epoch": 4.09, "learning_rate": 5.907864816121698e-05, "loss": 2.5205, "step": 52187 }, { "epoch": 4.09, "learning_rate": 5.907786403199248e-05, "loss": 2.6543, "step": 52188 }, { "epoch": 4.09, "learning_rate": 5.9077079902767986e-05, "loss": 2.6126, "step": 52189 }, { "epoch": 4.09, "learning_rate": 5.9076295773543476e-05, "loss": 2.7174, "step": 52190 }, { "epoch": 4.09, "learning_rate": 5.907551164431898e-05, "loss": 2.7087, "step": 52191 }, { "epoch": 4.09, "learning_rate": 5.9074727515094484e-05, "loss": 2.6094, "step": 52192 }, { "epoch": 4.09, "learning_rate": 5.907394338586999e-05, "loss": 2.6871, "step": 52193 }, { "epoch": 4.09, "learning_rate": 5.907315925664549e-05, "loss": 2.7529, "step": 52194 }, { "epoch": 4.09, "learning_rate": 5.9072375127421e-05, "loss": 2.5862, "step": 52195 }, { "epoch": 4.09, "learning_rate": 5.9071590998196505e-05, "loss": 2.5859, "step": 52196 }, { "epoch": 4.09, "learning_rate": 5.907080686897201e-05, "loss": 2.7471, "step": 52197 }, { "epoch": 4.09, "learning_rate": 5.907002273974751e-05, "loss": 2.7196, "step": 52198 }, { "epoch": 4.09, "learning_rate": 5.9069238610523016e-05, "loss": 2.6556, "step": 52199 }, { "epoch": 4.09, "learning_rate": 5.906845448129852e-05, "loss": 2.725, "step": 52200 }, { "epoch": 4.09, "learning_rate": 5.906767035207402e-05, "loss": 2.6034, "step": 52201 }, { "epoch": 4.09, "learning_rate": 5.906688622284953e-05, "loss": 2.542, "step": 52202 }, { "epoch": 4.09, "learning_rate": 5.906610209362503e-05, "loss": 2.6035, "step": 52203 }, { "epoch": 4.09, "learning_rate": 5.9065317964400534e-05, "loss": 2.6023, "step": 52204 }, { "epoch": 4.09, "learning_rate": 5.906453383517604e-05, "loss": 2.7171, "step": 52205 }, { "epoch": 4.09, "learning_rate": 5.906374970595154e-05, "loss": 2.9186, "step": 52206 }, { "epoch": 4.09, "learning_rate": 5.9062965576727045e-05, "loss": 2.6229, "step": 52207 }, { "epoch": 4.09, "learning_rate": 5.906218144750255e-05, "loss": 2.7323, "step": 52208 }, { "epoch": 4.09, "learning_rate": 5.906139731827805e-05, "loss": 2.7103, "step": 52209 }, { "epoch": 4.09, "learning_rate": 5.9060613189053556e-05, "loss": 2.804, "step": 52210 }, { "epoch": 4.09, "learning_rate": 5.905982905982907e-05, "loss": 2.5087, "step": 52211 }, { "epoch": 4.09, "learning_rate": 5.905904493060457e-05, "loss": 2.6564, "step": 52212 }, { "epoch": 4.09, "learning_rate": 5.9058260801380074e-05, "loss": 2.6373, "step": 52213 }, { "epoch": 4.09, "learning_rate": 5.905747667215558e-05, "loss": 2.5898, "step": 52214 }, { "epoch": 4.09, "learning_rate": 5.905669254293108e-05, "loss": 2.7261, "step": 52215 }, { "epoch": 4.09, "learning_rate": 5.9055908413706585e-05, "loss": 2.6732, "step": 52216 }, { "epoch": 4.09, "learning_rate": 5.905512428448209e-05, "loss": 2.7179, "step": 52217 }, { "epoch": 4.09, "learning_rate": 5.905434015525759e-05, "loss": 2.5874, "step": 52218 }, { "epoch": 4.09, "learning_rate": 5.9053556026033096e-05, "loss": 2.846, "step": 52219 }, { "epoch": 4.09, "learning_rate": 5.90527718968086e-05, "loss": 2.4578, "step": 52220 }, { "epoch": 4.09, "learning_rate": 5.9051987767584104e-05, "loss": 2.5427, "step": 52221 }, { "epoch": 4.09, "learning_rate": 5.905120363835961e-05, "loss": 2.6843, "step": 52222 }, { "epoch": 4.09, "learning_rate": 5.905041950913511e-05, "loss": 2.6782, "step": 52223 }, { "epoch": 4.1, "learning_rate": 5.9049635379910615e-05, "loss": 3.0035, "step": 52224 }, { "epoch": 4.1, "learning_rate": 5.904885125068612e-05, "loss": 2.6012, "step": 52225 }, { "epoch": 4.1, "learning_rate": 5.9048067121461615e-05, "loss": 2.6403, "step": 52226 }, { "epoch": 4.1, "learning_rate": 5.904728299223712e-05, "loss": 2.5685, "step": 52227 }, { "epoch": 4.1, "learning_rate": 5.904649886301262e-05, "loss": 2.787, "step": 52228 }, { "epoch": 4.1, "learning_rate": 5.9045714733788126e-05, "loss": 2.6845, "step": 52229 }, { "epoch": 4.1, "learning_rate": 5.904493060456363e-05, "loss": 2.6798, "step": 52230 }, { "epoch": 4.1, "learning_rate": 5.9044146475339134e-05, "loss": 2.6336, "step": 52231 }, { "epoch": 4.1, "learning_rate": 5.904336234611464e-05, "loss": 2.8383, "step": 52232 }, { "epoch": 4.1, "learning_rate": 5.904257821689014e-05, "loss": 2.4754, "step": 52233 }, { "epoch": 4.1, "learning_rate": 5.9041794087665645e-05, "loss": 2.5921, "step": 52234 }, { "epoch": 4.1, "learning_rate": 5.904100995844115e-05, "loss": 2.6482, "step": 52235 }, { "epoch": 4.1, "learning_rate": 5.904022582921665e-05, "loss": 2.6995, "step": 52236 }, { "epoch": 4.1, "learning_rate": 5.903944169999216e-05, "loss": 2.8588, "step": 52237 }, { "epoch": 4.1, "learning_rate": 5.9038657570767666e-05, "loss": 2.6361, "step": 52238 }, { "epoch": 4.1, "learning_rate": 5.903787344154317e-05, "loss": 2.6755, "step": 52239 }, { "epoch": 4.1, "learning_rate": 5.9037089312318674e-05, "loss": 2.6821, "step": 52240 }, { "epoch": 4.1, "learning_rate": 5.903630518309418e-05, "loss": 2.7326, "step": 52241 }, { "epoch": 4.1, "learning_rate": 5.903552105386968e-05, "loss": 2.7569, "step": 52242 }, { "epoch": 4.1, "learning_rate": 5.9034736924645185e-05, "loss": 2.696, "step": 52243 }, { "epoch": 4.1, "learning_rate": 5.903395279542069e-05, "loss": 2.6024, "step": 52244 }, { "epoch": 4.1, "learning_rate": 5.903316866619619e-05, "loss": 2.6943, "step": 52245 }, { "epoch": 4.1, "learning_rate": 5.9032384536971696e-05, "loss": 2.6435, "step": 52246 }, { "epoch": 4.1, "learning_rate": 5.90316004077472e-05, "loss": 2.7269, "step": 52247 }, { "epoch": 4.1, "learning_rate": 5.90308162785227e-05, "loss": 2.4377, "step": 52248 }, { "epoch": 4.1, "learning_rate": 5.903003214929821e-05, "loss": 2.7942, "step": 52249 }, { "epoch": 4.1, "learning_rate": 5.902924802007371e-05, "loss": 2.5458, "step": 52250 }, { "epoch": 4.1, "learning_rate": 5.9028463890849214e-05, "loss": 2.5941, "step": 52251 }, { "epoch": 4.1, "learning_rate": 5.9027679761624724e-05, "loss": 2.452, "step": 52252 }, { "epoch": 4.1, "learning_rate": 5.902689563240023e-05, "loss": 2.8626, "step": 52253 }, { "epoch": 4.1, "learning_rate": 5.902611150317573e-05, "loss": 2.4669, "step": 52254 }, { "epoch": 4.1, "learning_rate": 5.9025327373951236e-05, "loss": 2.8211, "step": 52255 }, { "epoch": 4.1, "learning_rate": 5.902454324472674e-05, "loss": 2.6964, "step": 52256 }, { "epoch": 4.1, "learning_rate": 5.902375911550224e-05, "loss": 2.7088, "step": 52257 }, { "epoch": 4.1, "learning_rate": 5.9022974986277747e-05, "loss": 2.5874, "step": 52258 }, { "epoch": 4.1, "learning_rate": 5.902219085705325e-05, "loss": 2.7583, "step": 52259 }, { "epoch": 4.1, "learning_rate": 5.9021406727828754e-05, "loss": 2.5104, "step": 52260 }, { "epoch": 4.1, "learning_rate": 5.902062259860426e-05, "loss": 2.6247, "step": 52261 }, { "epoch": 4.1, "learning_rate": 5.9019838469379754e-05, "loss": 2.5583, "step": 52262 }, { "epoch": 4.1, "learning_rate": 5.901905434015526e-05, "loss": 2.7856, "step": 52263 }, { "epoch": 4.1, "learning_rate": 5.901827021093076e-05, "loss": 2.6183, "step": 52264 }, { "epoch": 4.1, "learning_rate": 5.9017486081706266e-05, "loss": 2.5425, "step": 52265 }, { "epoch": 4.1, "learning_rate": 5.901670195248177e-05, "loss": 2.6384, "step": 52266 }, { "epoch": 4.1, "learning_rate": 5.901591782325727e-05, "loss": 2.6261, "step": 52267 }, { "epoch": 4.1, "learning_rate": 5.9015133694032777e-05, "loss": 2.5857, "step": 52268 }, { "epoch": 4.1, "learning_rate": 5.901434956480828e-05, "loss": 2.6286, "step": 52269 }, { "epoch": 4.1, "learning_rate": 5.9013565435583784e-05, "loss": 2.5749, "step": 52270 }, { "epoch": 4.1, "learning_rate": 5.901278130635929e-05, "loss": 2.7071, "step": 52271 }, { "epoch": 4.1, "learning_rate": 5.901199717713479e-05, "loss": 2.7415, "step": 52272 }, { "epoch": 4.1, "learning_rate": 5.9011213047910295e-05, "loss": 2.5617, "step": 52273 }, { "epoch": 4.1, "learning_rate": 5.90104289186858e-05, "loss": 2.5167, "step": 52274 }, { "epoch": 4.1, "learning_rate": 5.90096447894613e-05, "loss": 2.6928, "step": 52275 }, { "epoch": 4.1, "learning_rate": 5.9008860660236806e-05, "loss": 2.7196, "step": 52276 }, { "epoch": 4.1, "learning_rate": 5.900807653101231e-05, "loss": 2.6553, "step": 52277 }, { "epoch": 4.1, "learning_rate": 5.900729240178782e-05, "loss": 2.7028, "step": 52278 }, { "epoch": 4.1, "learning_rate": 5.9006508272563324e-05, "loss": 2.6823, "step": 52279 }, { "epoch": 4.1, "learning_rate": 5.900572414333883e-05, "loss": 2.6466, "step": 52280 }, { "epoch": 4.1, "learning_rate": 5.900494001411433e-05, "loss": 2.631, "step": 52281 }, { "epoch": 4.1, "learning_rate": 5.9004155884889835e-05, "loss": 2.8145, "step": 52282 }, { "epoch": 4.1, "learning_rate": 5.900337175566534e-05, "loss": 2.5834, "step": 52283 }, { "epoch": 4.1, "learning_rate": 5.900258762644084e-05, "loss": 2.6355, "step": 52284 }, { "epoch": 4.1, "learning_rate": 5.9001803497216346e-05, "loss": 2.6883, "step": 52285 }, { "epoch": 4.1, "learning_rate": 5.900101936799185e-05, "loss": 2.4978, "step": 52286 }, { "epoch": 4.1, "learning_rate": 5.900023523876735e-05, "loss": 2.5994, "step": 52287 }, { "epoch": 4.1, "learning_rate": 5.899945110954286e-05, "loss": 2.5502, "step": 52288 }, { "epoch": 4.1, "learning_rate": 5.899866698031836e-05, "loss": 2.8709, "step": 52289 }, { "epoch": 4.1, "learning_rate": 5.8997882851093864e-05, "loss": 2.5481, "step": 52290 }, { "epoch": 4.1, "learning_rate": 5.899709872186937e-05, "loss": 2.7707, "step": 52291 }, { "epoch": 4.1, "learning_rate": 5.899631459264487e-05, "loss": 2.7126, "step": 52292 }, { "epoch": 4.1, "learning_rate": 5.899553046342038e-05, "loss": 2.6147, "step": 52293 }, { "epoch": 4.1, "learning_rate": 5.8994746334195886e-05, "loss": 2.6982, "step": 52294 }, { "epoch": 4.1, "learning_rate": 5.899396220497139e-05, "loss": 2.6443, "step": 52295 }, { "epoch": 4.1, "learning_rate": 5.899317807574689e-05, "loss": 2.7204, "step": 52296 }, { "epoch": 4.1, "learning_rate": 5.89923939465224e-05, "loss": 2.6524, "step": 52297 }, { "epoch": 4.1, "learning_rate": 5.899160981729789e-05, "loss": 2.4955, "step": 52298 }, { "epoch": 4.1, "learning_rate": 5.899082568807339e-05, "loss": 2.7701, "step": 52299 }, { "epoch": 4.1, "learning_rate": 5.8990041558848894e-05, "loss": 2.5267, "step": 52300 }, { "epoch": 4.1, "learning_rate": 5.89892574296244e-05, "loss": 2.6088, "step": 52301 }, { "epoch": 4.1, "learning_rate": 5.89884733003999e-05, "loss": 2.5328, "step": 52302 }, { "epoch": 4.1, "learning_rate": 5.898768917117541e-05, "loss": 2.5618, "step": 52303 }, { "epoch": 4.1, "learning_rate": 5.8986905041950916e-05, "loss": 2.6045, "step": 52304 }, { "epoch": 4.1, "learning_rate": 5.898612091272642e-05, "loss": 2.7264, "step": 52305 }, { "epoch": 4.1, "learning_rate": 5.898533678350192e-05, "loss": 2.469, "step": 52306 }, { "epoch": 4.1, "learning_rate": 5.898455265427743e-05, "loss": 2.4563, "step": 52307 }, { "epoch": 4.1, "learning_rate": 5.898376852505293e-05, "loss": 2.7337, "step": 52308 }, { "epoch": 4.1, "learning_rate": 5.8982984395828434e-05, "loss": 2.4847, "step": 52309 }, { "epoch": 4.1, "learning_rate": 5.898220026660394e-05, "loss": 2.6966, "step": 52310 }, { "epoch": 4.1, "learning_rate": 5.898141613737944e-05, "loss": 2.6852, "step": 52311 }, { "epoch": 4.1, "learning_rate": 5.8980632008154945e-05, "loss": 2.5326, "step": 52312 }, { "epoch": 4.1, "learning_rate": 5.897984787893045e-05, "loss": 2.645, "step": 52313 }, { "epoch": 4.1, "learning_rate": 5.897906374970595e-05, "loss": 2.661, "step": 52314 }, { "epoch": 4.1, "learning_rate": 5.8978279620481456e-05, "loss": 2.7663, "step": 52315 }, { "epoch": 4.1, "learning_rate": 5.897749549125696e-05, "loss": 2.5597, "step": 52316 }, { "epoch": 4.1, "learning_rate": 5.8976711362032464e-05, "loss": 2.5787, "step": 52317 }, { "epoch": 4.1, "learning_rate": 5.897592723280797e-05, "loss": 2.7259, "step": 52318 }, { "epoch": 4.1, "learning_rate": 5.897514310358348e-05, "loss": 2.8177, "step": 52319 }, { "epoch": 4.1, "learning_rate": 5.897435897435898e-05, "loss": 2.5397, "step": 52320 }, { "epoch": 4.1, "learning_rate": 5.8973574845134485e-05, "loss": 2.7621, "step": 52321 }, { "epoch": 4.1, "learning_rate": 5.897279071590999e-05, "loss": 2.461, "step": 52322 }, { "epoch": 4.1, "learning_rate": 5.897200658668549e-05, "loss": 2.6289, "step": 52323 }, { "epoch": 4.1, "learning_rate": 5.8971222457460996e-05, "loss": 2.6505, "step": 52324 }, { "epoch": 4.1, "learning_rate": 5.89704383282365e-05, "loss": 2.7797, "step": 52325 }, { "epoch": 4.1, "learning_rate": 5.8969654199012003e-05, "loss": 2.661, "step": 52326 }, { "epoch": 4.1, "learning_rate": 5.896887006978751e-05, "loss": 2.9273, "step": 52327 }, { "epoch": 4.1, "learning_rate": 5.896808594056301e-05, "loss": 2.5247, "step": 52328 }, { "epoch": 4.1, "learning_rate": 5.8967301811338514e-05, "loss": 2.8116, "step": 52329 }, { "epoch": 4.1, "learning_rate": 5.896651768211402e-05, "loss": 2.6466, "step": 52330 }, { "epoch": 4.1, "learning_rate": 5.896573355288952e-05, "loss": 2.7225, "step": 52331 }, { "epoch": 4.1, "learning_rate": 5.8964949423665025e-05, "loss": 2.7208, "step": 52332 }, { "epoch": 4.1, "learning_rate": 5.896416529444053e-05, "loss": 2.6168, "step": 52333 }, { "epoch": 4.1, "learning_rate": 5.8963381165216026e-05, "loss": 2.5777, "step": 52334 }, { "epoch": 4.1, "learning_rate": 5.896259703599153e-05, "loss": 2.5024, "step": 52335 }, { "epoch": 4.1, "learning_rate": 5.8961812906767033e-05, "loss": 2.8636, "step": 52336 }, { "epoch": 4.1, "learning_rate": 5.896102877754254e-05, "loss": 2.6836, "step": 52337 }, { "epoch": 4.1, "learning_rate": 5.896024464831804e-05, "loss": 2.7557, "step": 52338 }, { "epoch": 4.1, "learning_rate": 5.8959460519093544e-05, "loss": 2.6576, "step": 52339 }, { "epoch": 4.1, "learning_rate": 5.895867638986905e-05, "loss": 2.6228, "step": 52340 }, { "epoch": 4.1, "learning_rate": 5.895789226064455e-05, "loss": 2.7203, "step": 52341 }, { "epoch": 4.1, "learning_rate": 5.8957108131420055e-05, "loss": 2.6332, "step": 52342 }, { "epoch": 4.1, "learning_rate": 5.895632400219556e-05, "loss": 2.9906, "step": 52343 }, { "epoch": 4.1, "learning_rate": 5.895553987297106e-05, "loss": 2.5534, "step": 52344 }, { "epoch": 4.1, "learning_rate": 5.895475574374657e-05, "loss": 2.5118, "step": 52345 }, { "epoch": 4.1, "learning_rate": 5.895397161452208e-05, "loss": 2.6089, "step": 52346 }, { "epoch": 4.1, "learning_rate": 5.895318748529758e-05, "loss": 2.7805, "step": 52347 }, { "epoch": 4.1, "learning_rate": 5.8952403356073084e-05, "loss": 2.798, "step": 52348 }, { "epoch": 4.1, "learning_rate": 5.895161922684859e-05, "loss": 2.6724, "step": 52349 }, { "epoch": 4.1, "learning_rate": 5.895083509762409e-05, "loss": 2.688, "step": 52350 }, { "epoch": 4.1, "learning_rate": 5.8950050968399595e-05, "loss": 2.6318, "step": 52351 }, { "epoch": 4.11, "learning_rate": 5.89492668391751e-05, "loss": 2.6023, "step": 52352 }, { "epoch": 4.11, "learning_rate": 5.89484827099506e-05, "loss": 2.6936, "step": 52353 }, { "epoch": 4.11, "learning_rate": 5.8947698580726106e-05, "loss": 2.5848, "step": 52354 }, { "epoch": 4.11, "learning_rate": 5.894691445150161e-05, "loss": 2.8837, "step": 52355 }, { "epoch": 4.11, "learning_rate": 5.8946130322277114e-05, "loss": 2.767, "step": 52356 }, { "epoch": 4.11, "learning_rate": 5.894534619305262e-05, "loss": 2.6868, "step": 52357 }, { "epoch": 4.11, "learning_rate": 5.894456206382812e-05, "loss": 2.5212, "step": 52358 }, { "epoch": 4.11, "learning_rate": 5.8943777934603625e-05, "loss": 2.4621, "step": 52359 }, { "epoch": 4.11, "learning_rate": 5.8942993805379135e-05, "loss": 2.9234, "step": 52360 }, { "epoch": 4.11, "learning_rate": 5.894220967615464e-05, "loss": 2.7055, "step": 52361 }, { "epoch": 4.11, "learning_rate": 5.894142554693014e-05, "loss": 2.5483, "step": 52362 }, { "epoch": 4.11, "learning_rate": 5.8940641417705646e-05, "loss": 2.5044, "step": 52363 }, { "epoch": 4.11, "learning_rate": 5.893985728848115e-05, "loss": 2.7059, "step": 52364 }, { "epoch": 4.11, "learning_rate": 5.8939073159256654e-05, "loss": 2.8305, "step": 52365 }, { "epoch": 4.11, "learning_rate": 5.893828903003216e-05, "loss": 2.61, "step": 52366 }, { "epoch": 4.11, "learning_rate": 5.893750490080766e-05, "loss": 2.7696, "step": 52367 }, { "epoch": 4.11, "learning_rate": 5.8936720771583165e-05, "loss": 2.6759, "step": 52368 }, { "epoch": 4.11, "learning_rate": 5.893593664235867e-05, "loss": 2.6199, "step": 52369 }, { "epoch": 4.11, "learning_rate": 5.8935152513134165e-05, "loss": 2.5618, "step": 52370 }, { "epoch": 4.11, "learning_rate": 5.893436838390967e-05, "loss": 2.7129, "step": 52371 }, { "epoch": 4.11, "learning_rate": 5.893358425468517e-05, "loss": 2.7287, "step": 52372 }, { "epoch": 4.11, "learning_rate": 5.8932800125460676e-05, "loss": 2.6178, "step": 52373 }, { "epoch": 4.11, "learning_rate": 5.893201599623618e-05, "loss": 2.7537, "step": 52374 }, { "epoch": 4.11, "learning_rate": 5.8931231867011684e-05, "loss": 2.6035, "step": 52375 }, { "epoch": 4.11, "learning_rate": 5.893044773778719e-05, "loss": 2.6458, "step": 52376 }, { "epoch": 4.11, "learning_rate": 5.892966360856269e-05, "loss": 2.7791, "step": 52377 }, { "epoch": 4.11, "learning_rate": 5.8928879479338195e-05, "loss": 2.72, "step": 52378 }, { "epoch": 4.11, "learning_rate": 5.89280953501137e-05, "loss": 2.508, "step": 52379 }, { "epoch": 4.11, "learning_rate": 5.89273112208892e-05, "loss": 2.6619, "step": 52380 }, { "epoch": 4.11, "learning_rate": 5.8926527091664706e-05, "loss": 2.5286, "step": 52381 }, { "epoch": 4.11, "learning_rate": 5.892574296244021e-05, "loss": 2.7402, "step": 52382 }, { "epoch": 4.11, "learning_rate": 5.892495883321571e-05, "loss": 2.8838, "step": 52383 }, { "epoch": 4.11, "learning_rate": 5.892417470399122e-05, "loss": 2.5445, "step": 52384 }, { "epoch": 4.11, "learning_rate": 5.892339057476672e-05, "loss": 2.6476, "step": 52385 }, { "epoch": 4.11, "learning_rate": 5.892260644554223e-05, "loss": 2.6409, "step": 52386 }, { "epoch": 4.11, "learning_rate": 5.8921822316317735e-05, "loss": 2.6365, "step": 52387 }, { "epoch": 4.11, "learning_rate": 5.892103818709324e-05, "loss": 2.6112, "step": 52388 }, { "epoch": 4.11, "learning_rate": 5.892025405786874e-05, "loss": 2.6742, "step": 52389 }, { "epoch": 4.11, "learning_rate": 5.8919469928644246e-05, "loss": 2.6862, "step": 52390 }, { "epoch": 4.11, "learning_rate": 5.891868579941975e-05, "loss": 2.6272, "step": 52391 }, { "epoch": 4.11, "learning_rate": 5.891790167019525e-05, "loss": 2.6706, "step": 52392 }, { "epoch": 4.11, "learning_rate": 5.8917117540970757e-05, "loss": 2.6276, "step": 52393 }, { "epoch": 4.11, "learning_rate": 5.891633341174626e-05, "loss": 2.4184, "step": 52394 }, { "epoch": 4.11, "learning_rate": 5.8915549282521764e-05, "loss": 2.6681, "step": 52395 }, { "epoch": 4.11, "learning_rate": 5.891476515329727e-05, "loss": 2.7391, "step": 52396 }, { "epoch": 4.11, "learning_rate": 5.891398102407277e-05, "loss": 2.6469, "step": 52397 }, { "epoch": 4.11, "learning_rate": 5.8913196894848275e-05, "loss": 2.5609, "step": 52398 }, { "epoch": 4.11, "learning_rate": 5.891241276562378e-05, "loss": 2.5355, "step": 52399 }, { "epoch": 4.11, "learning_rate": 5.891162863639928e-05, "loss": 2.6879, "step": 52400 }, { "epoch": 4.11, "learning_rate": 5.891084450717479e-05, "loss": 2.7616, "step": 52401 }, { "epoch": 4.11, "learning_rate": 5.8910060377950296e-05, "loss": 2.6333, "step": 52402 }, { "epoch": 4.11, "learning_rate": 5.89092762487258e-05, "loss": 2.684, "step": 52403 }, { "epoch": 4.11, "learning_rate": 5.8908492119501304e-05, "loss": 2.8657, "step": 52404 }, { "epoch": 4.11, "learning_rate": 5.890770799027681e-05, "loss": 2.8433, "step": 52405 }, { "epoch": 4.11, "learning_rate": 5.89069238610523e-05, "loss": 2.6375, "step": 52406 }, { "epoch": 4.11, "learning_rate": 5.89061397318278e-05, "loss": 2.7739, "step": 52407 }, { "epoch": 4.11, "learning_rate": 5.8905355602603305e-05, "loss": 2.5517, "step": 52408 }, { "epoch": 4.11, "learning_rate": 5.890457147337881e-05, "loss": 2.6887, "step": 52409 }, { "epoch": 4.11, "learning_rate": 5.890378734415431e-05, "loss": 2.696, "step": 52410 }, { "epoch": 4.11, "learning_rate": 5.890300321492982e-05, "loss": 2.5734, "step": 52411 }, { "epoch": 4.11, "learning_rate": 5.8902219085705326e-05, "loss": 2.5872, "step": 52412 }, { "epoch": 4.11, "learning_rate": 5.890143495648083e-05, "loss": 2.5077, "step": 52413 }, { "epoch": 4.11, "learning_rate": 5.8900650827256334e-05, "loss": 2.7917, "step": 52414 }, { "epoch": 4.11, "learning_rate": 5.889986669803184e-05, "loss": 2.5461, "step": 52415 }, { "epoch": 4.11, "learning_rate": 5.889908256880734e-05, "loss": 2.7132, "step": 52416 }, { "epoch": 4.11, "learning_rate": 5.8898298439582845e-05, "loss": 2.6274, "step": 52417 }, { "epoch": 4.11, "learning_rate": 5.889751431035835e-05, "loss": 2.5765, "step": 52418 }, { "epoch": 4.11, "learning_rate": 5.889673018113385e-05, "loss": 2.6822, "step": 52419 }, { "epoch": 4.11, "learning_rate": 5.8895946051909356e-05, "loss": 2.8229, "step": 52420 }, { "epoch": 4.11, "learning_rate": 5.889516192268486e-05, "loss": 2.8462, "step": 52421 }, { "epoch": 4.11, "learning_rate": 5.889437779346036e-05, "loss": 2.5734, "step": 52422 }, { "epoch": 4.11, "learning_rate": 5.889359366423587e-05, "loss": 2.6327, "step": 52423 }, { "epoch": 4.11, "learning_rate": 5.889280953501137e-05, "loss": 2.7155, "step": 52424 }, { "epoch": 4.11, "learning_rate": 5.8892025405786874e-05, "loss": 2.5289, "step": 52425 }, { "epoch": 4.11, "learning_rate": 5.889124127656238e-05, "loss": 2.7747, "step": 52426 }, { "epoch": 4.11, "learning_rate": 5.889045714733789e-05, "loss": 2.5446, "step": 52427 }, { "epoch": 4.11, "learning_rate": 5.888967301811339e-05, "loss": 2.578, "step": 52428 }, { "epoch": 4.11, "learning_rate": 5.8888888888888896e-05, "loss": 2.6415, "step": 52429 }, { "epoch": 4.11, "learning_rate": 5.88881047596644e-05, "loss": 2.6658, "step": 52430 }, { "epoch": 4.11, "learning_rate": 5.88873206304399e-05, "loss": 2.8767, "step": 52431 }, { "epoch": 4.11, "learning_rate": 5.888653650121541e-05, "loss": 2.7586, "step": 52432 }, { "epoch": 4.11, "learning_rate": 5.888575237199091e-05, "loss": 2.9451, "step": 52433 }, { "epoch": 4.11, "learning_rate": 5.8884968242766414e-05, "loss": 2.6076, "step": 52434 }, { "epoch": 4.11, "learning_rate": 5.888418411354192e-05, "loss": 2.7977, "step": 52435 }, { "epoch": 4.11, "learning_rate": 5.888339998431742e-05, "loss": 2.4917, "step": 52436 }, { "epoch": 4.11, "learning_rate": 5.8882615855092925e-05, "loss": 2.4821, "step": 52437 }, { "epoch": 4.11, "learning_rate": 5.888183172586843e-05, "loss": 2.8094, "step": 52438 }, { "epoch": 4.11, "learning_rate": 5.888104759664393e-05, "loss": 2.5895, "step": 52439 }, { "epoch": 4.11, "learning_rate": 5.8880263467419436e-05, "loss": 2.3615, "step": 52440 }, { "epoch": 4.11, "learning_rate": 5.887947933819494e-05, "loss": 2.6128, "step": 52441 }, { "epoch": 4.11, "learning_rate": 5.887869520897044e-05, "loss": 2.7191, "step": 52442 }, { "epoch": 4.11, "learning_rate": 5.887791107974594e-05, "loss": 2.8414, "step": 52443 }, { "epoch": 4.11, "learning_rate": 5.8877126950521444e-05, "loss": 2.7592, "step": 52444 }, { "epoch": 4.11, "learning_rate": 5.887634282129695e-05, "loss": 2.7229, "step": 52445 }, { "epoch": 4.11, "learning_rate": 5.887555869207245e-05, "loss": 2.6267, "step": 52446 }, { "epoch": 4.11, "learning_rate": 5.8874774562847955e-05, "loss": 2.7595, "step": 52447 }, { "epoch": 4.11, "learning_rate": 5.887399043362346e-05, "loss": 2.5769, "step": 52448 }, { "epoch": 4.11, "learning_rate": 5.887320630439896e-05, "loss": 2.6991, "step": 52449 }, { "epoch": 4.11, "learning_rate": 5.8872422175174466e-05, "loss": 2.5136, "step": 52450 }, { "epoch": 4.11, "learning_rate": 5.887163804594997e-05, "loss": 2.6994, "step": 52451 }, { "epoch": 4.11, "learning_rate": 5.8870853916725474e-05, "loss": 2.6127, "step": 52452 }, { "epoch": 4.11, "learning_rate": 5.8870069787500984e-05, "loss": 2.5754, "step": 52453 }, { "epoch": 4.11, "learning_rate": 5.886928565827649e-05, "loss": 2.4854, "step": 52454 }, { "epoch": 4.11, "learning_rate": 5.886850152905199e-05, "loss": 2.6574, "step": 52455 }, { "epoch": 4.11, "learning_rate": 5.8867717399827495e-05, "loss": 2.8951, "step": 52456 }, { "epoch": 4.11, "learning_rate": 5.8866933270603e-05, "loss": 2.5953, "step": 52457 }, { "epoch": 4.11, "learning_rate": 5.88661491413785e-05, "loss": 2.6049, "step": 52458 }, { "epoch": 4.11, "learning_rate": 5.8865365012154006e-05, "loss": 2.7351, "step": 52459 }, { "epoch": 4.11, "learning_rate": 5.886458088292951e-05, "loss": 2.591, "step": 52460 }, { "epoch": 4.11, "learning_rate": 5.8863796753705013e-05, "loss": 2.744, "step": 52461 }, { "epoch": 4.11, "learning_rate": 5.886301262448052e-05, "loss": 2.5938, "step": 52462 }, { "epoch": 4.11, "learning_rate": 5.886222849525602e-05, "loss": 2.5923, "step": 52463 }, { "epoch": 4.11, "learning_rate": 5.8861444366031524e-05, "loss": 2.7546, "step": 52464 }, { "epoch": 4.11, "learning_rate": 5.886066023680703e-05, "loss": 2.5968, "step": 52465 }, { "epoch": 4.11, "learning_rate": 5.885987610758253e-05, "loss": 2.7373, "step": 52466 }, { "epoch": 4.11, "learning_rate": 5.8859091978358036e-05, "loss": 2.7292, "step": 52467 }, { "epoch": 4.11, "learning_rate": 5.8858307849133546e-05, "loss": 2.5164, "step": 52468 }, { "epoch": 4.11, "learning_rate": 5.885752371990905e-05, "loss": 2.6405, "step": 52469 }, { "epoch": 4.11, "learning_rate": 5.885673959068455e-05, "loss": 2.6444, "step": 52470 }, { "epoch": 4.11, "learning_rate": 5.885595546146006e-05, "loss": 2.6595, "step": 52471 }, { "epoch": 4.11, "learning_rate": 5.885517133223556e-05, "loss": 2.8842, "step": 52472 }, { "epoch": 4.11, "learning_rate": 5.8854387203011064e-05, "loss": 2.5205, "step": 52473 }, { "epoch": 4.11, "learning_rate": 5.885360307378657e-05, "loss": 2.8089, "step": 52474 }, { "epoch": 4.11, "learning_rate": 5.885281894456207e-05, "loss": 2.9625, "step": 52475 }, { "epoch": 4.11, "learning_rate": 5.8852034815337575e-05, "loss": 2.5642, "step": 52476 }, { "epoch": 4.11, "learning_rate": 5.885125068611308e-05, "loss": 2.734, "step": 52477 }, { "epoch": 4.11, "learning_rate": 5.8850466556888576e-05, "loss": 2.5208, "step": 52478 }, { "epoch": 4.12, "learning_rate": 5.884968242766408e-05, "loss": 2.76, "step": 52479 }, { "epoch": 4.12, "learning_rate": 5.884889829843958e-05, "loss": 2.5962, "step": 52480 }, { "epoch": 4.12, "learning_rate": 5.884811416921509e-05, "loss": 2.5045, "step": 52481 }, { "epoch": 4.12, "learning_rate": 5.884733003999059e-05, "loss": 2.4231, "step": 52482 }, { "epoch": 4.12, "learning_rate": 5.8846545910766094e-05, "loss": 2.6782, "step": 52483 }, { "epoch": 4.12, "learning_rate": 5.88457617815416e-05, "loss": 2.6392, "step": 52484 }, { "epoch": 4.12, "learning_rate": 5.88449776523171e-05, "loss": 2.821, "step": 52485 }, { "epoch": 4.12, "learning_rate": 5.8844193523092605e-05, "loss": 2.6717, "step": 52486 }, { "epoch": 4.12, "learning_rate": 5.884340939386811e-05, "loss": 2.6885, "step": 52487 }, { "epoch": 4.12, "learning_rate": 5.884262526464361e-05, "loss": 2.5747, "step": 52488 }, { "epoch": 4.12, "learning_rate": 5.8841841135419116e-05, "loss": 2.5624, "step": 52489 }, { "epoch": 4.12, "learning_rate": 5.884105700619462e-05, "loss": 2.6097, "step": 52490 }, { "epoch": 4.12, "learning_rate": 5.8840272876970124e-05, "loss": 2.5459, "step": 52491 }, { "epoch": 4.12, "learning_rate": 5.883948874774563e-05, "loss": 2.6458, "step": 52492 }, { "epoch": 4.12, "learning_rate": 5.883870461852113e-05, "loss": 2.5339, "step": 52493 }, { "epoch": 4.12, "learning_rate": 5.883792048929664e-05, "loss": 2.4448, "step": 52494 }, { "epoch": 4.12, "learning_rate": 5.8837136360072145e-05, "loss": 2.5776, "step": 52495 }, { "epoch": 4.12, "learning_rate": 5.883635223084765e-05, "loss": 2.6647, "step": 52496 }, { "epoch": 4.12, "learning_rate": 5.883556810162315e-05, "loss": 2.564, "step": 52497 }, { "epoch": 4.12, "learning_rate": 5.8834783972398656e-05, "loss": 2.4771, "step": 52498 }, { "epoch": 4.12, "learning_rate": 5.883399984317416e-05, "loss": 2.6053, "step": 52499 }, { "epoch": 4.12, "learning_rate": 5.8833215713949664e-05, "loss": 2.8125, "step": 52500 }, { "epoch": 4.12, "learning_rate": 5.883243158472517e-05, "loss": 2.6146, "step": 52501 }, { "epoch": 4.12, "learning_rate": 5.883164745550067e-05, "loss": 2.664, "step": 52502 }, { "epoch": 4.12, "learning_rate": 5.8830863326276175e-05, "loss": 3.0155, "step": 52503 }, { "epoch": 4.12, "learning_rate": 5.883007919705168e-05, "loss": 2.7039, "step": 52504 }, { "epoch": 4.12, "learning_rate": 5.882929506782718e-05, "loss": 2.3466, "step": 52505 }, { "epoch": 4.12, "learning_rate": 5.8828510938602686e-05, "loss": 2.5446, "step": 52506 }, { "epoch": 4.12, "learning_rate": 5.882772680937819e-05, "loss": 2.6866, "step": 52507 }, { "epoch": 4.12, "learning_rate": 5.882694268015369e-05, "loss": 2.7117, "step": 52508 }, { "epoch": 4.12, "learning_rate": 5.88261585509292e-05, "loss": 2.4249, "step": 52509 }, { "epoch": 4.12, "learning_rate": 5.882537442170471e-05, "loss": 2.6388, "step": 52510 }, { "epoch": 4.12, "learning_rate": 5.882459029248021e-05, "loss": 2.6931, "step": 52511 }, { "epoch": 4.12, "learning_rate": 5.8823806163255715e-05, "loss": 2.4897, "step": 52512 }, { "epoch": 4.12, "learning_rate": 5.882302203403122e-05, "loss": 2.5832, "step": 52513 }, { "epoch": 4.12, "learning_rate": 5.882223790480671e-05, "loss": 2.938, "step": 52514 }, { "epoch": 4.12, "learning_rate": 5.882145377558221e-05, "loss": 2.6471, "step": 52515 }, { "epoch": 4.12, "learning_rate": 5.8820669646357716e-05, "loss": 2.7042, "step": 52516 }, { "epoch": 4.12, "learning_rate": 5.881988551713322e-05, "loss": 2.7316, "step": 52517 }, { "epoch": 4.12, "learning_rate": 5.881910138790872e-05, "loss": 2.6703, "step": 52518 }, { "epoch": 4.12, "learning_rate": 5.881831725868423e-05, "loss": 2.4864, "step": 52519 }, { "epoch": 4.12, "learning_rate": 5.881753312945974e-05, "loss": 2.8506, "step": 52520 }, { "epoch": 4.12, "learning_rate": 5.881674900023524e-05, "loss": 2.6927, "step": 52521 }, { "epoch": 4.12, "learning_rate": 5.8815964871010745e-05, "loss": 2.7487, "step": 52522 }, { "epoch": 4.12, "learning_rate": 5.881518074178625e-05, "loss": 2.7227, "step": 52523 }, { "epoch": 4.12, "learning_rate": 5.881439661256175e-05, "loss": 2.5151, "step": 52524 }, { "epoch": 4.12, "learning_rate": 5.8813612483337256e-05, "loss": 2.6079, "step": 52525 }, { "epoch": 4.12, "learning_rate": 5.881282835411276e-05, "loss": 2.4779, "step": 52526 }, { "epoch": 4.12, "learning_rate": 5.881204422488826e-05, "loss": 2.6622, "step": 52527 }, { "epoch": 4.12, "learning_rate": 5.881126009566377e-05, "loss": 2.5347, "step": 52528 }, { "epoch": 4.12, "learning_rate": 5.881047596643927e-05, "loss": 2.7027, "step": 52529 }, { "epoch": 4.12, "learning_rate": 5.8809691837214774e-05, "loss": 2.6284, "step": 52530 }, { "epoch": 4.12, "learning_rate": 5.880890770799028e-05, "loss": 2.6922, "step": 52531 }, { "epoch": 4.12, "learning_rate": 5.880812357876578e-05, "loss": 2.7689, "step": 52532 }, { "epoch": 4.12, "learning_rate": 5.8807339449541285e-05, "loss": 2.5461, "step": 52533 }, { "epoch": 4.12, "learning_rate": 5.880655532031679e-05, "loss": 2.4963, "step": 52534 }, { "epoch": 4.12, "learning_rate": 5.88057711910923e-05, "loss": 2.5589, "step": 52535 }, { "epoch": 4.12, "learning_rate": 5.88049870618678e-05, "loss": 2.775, "step": 52536 }, { "epoch": 4.12, "learning_rate": 5.8804202932643307e-05, "loss": 2.5048, "step": 52537 }, { "epoch": 4.12, "learning_rate": 5.880341880341881e-05, "loss": 2.5866, "step": 52538 }, { "epoch": 4.12, "learning_rate": 5.8802634674194314e-05, "loss": 2.4609, "step": 52539 }, { "epoch": 4.12, "learning_rate": 5.880185054496982e-05, "loss": 2.5444, "step": 52540 }, { "epoch": 4.12, "learning_rate": 5.880106641574532e-05, "loss": 2.7934, "step": 52541 }, { "epoch": 4.12, "learning_rate": 5.8800282286520825e-05, "loss": 2.6559, "step": 52542 }, { "epoch": 4.12, "learning_rate": 5.879949815729633e-05, "loss": 2.6836, "step": 52543 }, { "epoch": 4.12, "learning_rate": 5.879871402807183e-05, "loss": 2.6338, "step": 52544 }, { "epoch": 4.12, "learning_rate": 5.8797929898847336e-05, "loss": 2.7839, "step": 52545 }, { "epoch": 4.12, "learning_rate": 5.879714576962284e-05, "loss": 2.6785, "step": 52546 }, { "epoch": 4.12, "learning_rate": 5.879636164039834e-05, "loss": 2.3779, "step": 52547 }, { "epoch": 4.12, "learning_rate": 5.879557751117385e-05, "loss": 2.6855, "step": 52548 }, { "epoch": 4.12, "learning_rate": 5.879479338194935e-05, "loss": 2.5936, "step": 52549 }, { "epoch": 4.12, "learning_rate": 5.879400925272485e-05, "loss": 2.6758, "step": 52550 }, { "epoch": 4.12, "learning_rate": 5.879322512350035e-05, "loss": 2.4102, "step": 52551 }, { "epoch": 4.12, "learning_rate": 5.8792440994275855e-05, "loss": 2.7515, "step": 52552 }, { "epoch": 4.12, "learning_rate": 5.879165686505136e-05, "loss": 2.6347, "step": 52553 }, { "epoch": 4.12, "learning_rate": 5.879087273582686e-05, "loss": 2.7629, "step": 52554 }, { "epoch": 4.12, "learning_rate": 5.8790088606602366e-05, "loss": 2.742, "step": 52555 }, { "epoch": 4.12, "learning_rate": 5.878930447737787e-05, "loss": 2.7196, "step": 52556 }, { "epoch": 4.12, "learning_rate": 5.878852034815337e-05, "loss": 2.9652, "step": 52557 }, { "epoch": 4.12, "learning_rate": 5.878773621892888e-05, "loss": 2.7085, "step": 52558 }, { "epoch": 4.12, "learning_rate": 5.878695208970438e-05, "loss": 2.6746, "step": 52559 }, { "epoch": 4.12, "learning_rate": 5.8786167960479884e-05, "loss": 2.7063, "step": 52560 }, { "epoch": 4.12, "learning_rate": 5.8785383831255395e-05, "loss": 2.8535, "step": 52561 }, { "epoch": 4.12, "learning_rate": 5.87845997020309e-05, "loss": 2.6227, "step": 52562 }, { "epoch": 4.12, "learning_rate": 5.87838155728064e-05, "loss": 2.627, "step": 52563 }, { "epoch": 4.12, "learning_rate": 5.8783031443581906e-05, "loss": 2.6818, "step": 52564 }, { "epoch": 4.12, "learning_rate": 5.878224731435741e-05, "loss": 2.7319, "step": 52565 }, { "epoch": 4.12, "learning_rate": 5.878146318513291e-05, "loss": 2.7756, "step": 52566 }, { "epoch": 4.12, "learning_rate": 5.878067905590842e-05, "loss": 2.6569, "step": 52567 }, { "epoch": 4.12, "learning_rate": 5.877989492668392e-05, "loss": 2.5295, "step": 52568 }, { "epoch": 4.12, "learning_rate": 5.8779110797459424e-05, "loss": 2.6377, "step": 52569 }, { "epoch": 4.12, "learning_rate": 5.877832666823493e-05, "loss": 2.5617, "step": 52570 }, { "epoch": 4.12, "learning_rate": 5.877754253901043e-05, "loss": 2.6457, "step": 52571 }, { "epoch": 4.12, "learning_rate": 5.8776758409785935e-05, "loss": 2.9299, "step": 52572 }, { "epoch": 4.12, "learning_rate": 5.877597428056144e-05, "loss": 2.5871, "step": 52573 }, { "epoch": 4.12, "learning_rate": 5.877519015133694e-05, "loss": 2.5243, "step": 52574 }, { "epoch": 4.12, "learning_rate": 5.8774406022112446e-05, "loss": 2.503, "step": 52575 }, { "epoch": 4.12, "learning_rate": 5.877362189288796e-05, "loss": 2.6205, "step": 52576 }, { "epoch": 4.12, "learning_rate": 5.877283776366346e-05, "loss": 2.8601, "step": 52577 }, { "epoch": 4.12, "learning_rate": 5.8772053634438964e-05, "loss": 2.5569, "step": 52578 }, { "epoch": 4.12, "learning_rate": 5.877126950521447e-05, "loss": 2.6879, "step": 52579 }, { "epoch": 4.12, "learning_rate": 5.877048537598997e-05, "loss": 2.7334, "step": 52580 }, { "epoch": 4.12, "learning_rate": 5.8769701246765475e-05, "loss": 2.7938, "step": 52581 }, { "epoch": 4.12, "learning_rate": 5.876891711754098e-05, "loss": 2.6555, "step": 52582 }, { "epoch": 4.12, "learning_rate": 5.876813298831648e-05, "loss": 2.7268, "step": 52583 }, { "epoch": 4.12, "learning_rate": 5.8767348859091986e-05, "loss": 2.6436, "step": 52584 }, { "epoch": 4.12, "learning_rate": 5.876656472986749e-05, "loss": 2.6303, "step": 52585 }, { "epoch": 4.12, "learning_rate": 5.876578060064299e-05, "loss": 2.6841, "step": 52586 }, { "epoch": 4.12, "learning_rate": 5.876499647141849e-05, "loss": 2.516, "step": 52587 }, { "epoch": 4.12, "learning_rate": 5.8764212342193994e-05, "loss": 2.6393, "step": 52588 }, { "epoch": 4.12, "learning_rate": 5.87634282129695e-05, "loss": 2.8152, "step": 52589 }, { "epoch": 4.12, "learning_rate": 5.8762644083745e-05, "loss": 2.4851, "step": 52590 }, { "epoch": 4.12, "learning_rate": 5.8761859954520505e-05, "loss": 2.6682, "step": 52591 }, { "epoch": 4.12, "learning_rate": 5.876107582529601e-05, "loss": 2.5677, "step": 52592 }, { "epoch": 4.12, "learning_rate": 5.876029169607151e-05, "loss": 2.6703, "step": 52593 }, { "epoch": 4.12, "learning_rate": 5.8759507566847016e-05, "loss": 2.5776, "step": 52594 }, { "epoch": 4.12, "learning_rate": 5.875872343762252e-05, "loss": 2.7417, "step": 52595 }, { "epoch": 4.12, "learning_rate": 5.8757939308398024e-05, "loss": 2.4738, "step": 52596 }, { "epoch": 4.12, "learning_rate": 5.875715517917353e-05, "loss": 2.5093, "step": 52597 }, { "epoch": 4.12, "learning_rate": 5.875637104994903e-05, "loss": 2.6963, "step": 52598 }, { "epoch": 4.12, "learning_rate": 5.8755586920724535e-05, "loss": 2.3973, "step": 52599 }, { "epoch": 4.12, "learning_rate": 5.875480279150004e-05, "loss": 2.5747, "step": 52600 }, { "epoch": 4.12, "learning_rate": 5.875401866227554e-05, "loss": 2.6772, "step": 52601 }, { "epoch": 4.12, "learning_rate": 5.875323453305105e-05, "loss": 2.8385, "step": 52602 }, { "epoch": 4.12, "learning_rate": 5.8752450403826556e-05, "loss": 2.5236, "step": 52603 }, { "epoch": 4.12, "learning_rate": 5.875166627460206e-05, "loss": 2.7128, "step": 52604 }, { "epoch": 4.12, "learning_rate": 5.875088214537756e-05, "loss": 2.5618, "step": 52605 }, { "epoch": 4.12, "learning_rate": 5.875009801615307e-05, "loss": 2.7028, "step": 52606 }, { "epoch": 4.13, "learning_rate": 5.874931388692857e-05, "loss": 2.8068, "step": 52607 }, { "epoch": 4.13, "learning_rate": 5.8748529757704074e-05, "loss": 3.1067, "step": 52608 }, { "epoch": 4.13, "learning_rate": 5.874774562847958e-05, "loss": 2.5596, "step": 52609 }, { "epoch": 4.13, "learning_rate": 5.874696149925508e-05, "loss": 2.6236, "step": 52610 }, { "epoch": 4.13, "learning_rate": 5.8746177370030585e-05, "loss": 2.6337, "step": 52611 }, { "epoch": 4.13, "learning_rate": 5.874539324080609e-05, "loss": 2.5647, "step": 52612 }, { "epoch": 4.13, "learning_rate": 5.874460911158159e-05, "loss": 2.6978, "step": 52613 }, { "epoch": 4.13, "learning_rate": 5.8743824982357096e-05, "loss": 2.6363, "step": 52614 }, { "epoch": 4.13, "learning_rate": 5.87430408531326e-05, "loss": 2.6008, "step": 52615 }, { "epoch": 4.13, "learning_rate": 5.8742256723908104e-05, "loss": 2.6968, "step": 52616 }, { "epoch": 4.13, "learning_rate": 5.874147259468361e-05, "loss": 2.6041, "step": 52617 }, { "epoch": 4.13, "learning_rate": 5.874068846545912e-05, "loss": 2.6226, "step": 52618 }, { "epoch": 4.13, "learning_rate": 5.873990433623462e-05, "loss": 2.7961, "step": 52619 }, { "epoch": 4.13, "learning_rate": 5.8739120207010125e-05, "loss": 2.6323, "step": 52620 }, { "epoch": 4.13, "learning_rate": 5.873833607778563e-05, "loss": 2.6585, "step": 52621 }, { "epoch": 4.13, "learning_rate": 5.873755194856112e-05, "loss": 2.6878, "step": 52622 }, { "epoch": 4.13, "learning_rate": 5.873676781933662e-05, "loss": 2.3964, "step": 52623 }, { "epoch": 4.13, "learning_rate": 5.8735983690112126e-05, "loss": 2.6928, "step": 52624 }, { "epoch": 4.13, "learning_rate": 5.873519956088763e-05, "loss": 2.8172, "step": 52625 }, { "epoch": 4.13, "learning_rate": 5.8734415431663134e-05, "loss": 2.6211, "step": 52626 }, { "epoch": 4.13, "learning_rate": 5.873363130243864e-05, "loss": 2.7686, "step": 52627 }, { "epoch": 4.13, "learning_rate": 5.873284717321415e-05, "loss": 2.8141, "step": 52628 }, { "epoch": 4.13, "learning_rate": 5.873206304398965e-05, "loss": 2.6179, "step": 52629 }, { "epoch": 4.13, "learning_rate": 5.8731278914765155e-05, "loss": 2.5897, "step": 52630 }, { "epoch": 4.13, "learning_rate": 5.873049478554066e-05, "loss": 2.7061, "step": 52631 }, { "epoch": 4.13, "learning_rate": 5.872971065631616e-05, "loss": 2.6215, "step": 52632 }, { "epoch": 4.13, "learning_rate": 5.8728926527091666e-05, "loss": 2.6937, "step": 52633 }, { "epoch": 4.13, "learning_rate": 5.872814239786717e-05, "loss": 2.5331, "step": 52634 }, { "epoch": 4.13, "learning_rate": 5.8727358268642674e-05, "loss": 2.5136, "step": 52635 }, { "epoch": 4.13, "learning_rate": 5.872657413941818e-05, "loss": 2.6023, "step": 52636 }, { "epoch": 4.13, "learning_rate": 5.872579001019368e-05, "loss": 2.7257, "step": 52637 }, { "epoch": 4.13, "learning_rate": 5.8725005880969185e-05, "loss": 2.6114, "step": 52638 }, { "epoch": 4.13, "learning_rate": 5.872422175174469e-05, "loss": 2.6524, "step": 52639 }, { "epoch": 4.13, "learning_rate": 5.872343762252019e-05, "loss": 2.5043, "step": 52640 }, { "epoch": 4.13, "learning_rate": 5.8722653493295696e-05, "loss": 2.6845, "step": 52641 }, { "epoch": 4.13, "learning_rate": 5.87218693640712e-05, "loss": 2.7096, "step": 52642 }, { "epoch": 4.13, "learning_rate": 5.872108523484671e-05, "loss": 2.4691, "step": 52643 }, { "epoch": 4.13, "learning_rate": 5.8720301105622214e-05, "loss": 2.755, "step": 52644 }, { "epoch": 4.13, "learning_rate": 5.871951697639772e-05, "loss": 2.5329, "step": 52645 }, { "epoch": 4.13, "learning_rate": 5.871873284717322e-05, "loss": 2.7452, "step": 52646 }, { "epoch": 4.13, "learning_rate": 5.8717948717948725e-05, "loss": 2.6589, "step": 52647 }, { "epoch": 4.13, "learning_rate": 5.871716458872423e-05, "loss": 2.5976, "step": 52648 }, { "epoch": 4.13, "learning_rate": 5.871638045949973e-05, "loss": 2.6593, "step": 52649 }, { "epoch": 4.13, "learning_rate": 5.8715596330275236e-05, "loss": 2.5835, "step": 52650 }, { "epoch": 4.13, "learning_rate": 5.871481220105074e-05, "loss": 2.6184, "step": 52651 }, { "epoch": 4.13, "learning_rate": 5.871402807182624e-05, "loss": 2.6205, "step": 52652 }, { "epoch": 4.13, "learning_rate": 5.871324394260175e-05, "loss": 2.6992, "step": 52653 }, { "epoch": 4.13, "learning_rate": 5.871245981337725e-05, "loss": 2.8677, "step": 52654 }, { "epoch": 4.13, "learning_rate": 5.8711675684152754e-05, "loss": 2.6807, "step": 52655 }, { "epoch": 4.13, "learning_rate": 5.871089155492826e-05, "loss": 2.6525, "step": 52656 }, { "epoch": 4.13, "learning_rate": 5.871010742570376e-05, "loss": 2.6091, "step": 52657 }, { "epoch": 4.13, "learning_rate": 5.870932329647926e-05, "loss": 2.576, "step": 52658 }, { "epoch": 4.13, "learning_rate": 5.870853916725476e-05, "loss": 2.6873, "step": 52659 }, { "epoch": 4.13, "learning_rate": 5.8707755038030266e-05, "loss": 2.7932, "step": 52660 }, { "epoch": 4.13, "learning_rate": 5.870697090880577e-05, "loss": 2.6262, "step": 52661 }, { "epoch": 4.13, "learning_rate": 5.870618677958127e-05, "loss": 2.6621, "step": 52662 }, { "epoch": 4.13, "learning_rate": 5.870540265035678e-05, "loss": 2.7538, "step": 52663 }, { "epoch": 4.13, "learning_rate": 5.870461852113228e-05, "loss": 2.6071, "step": 52664 }, { "epoch": 4.13, "learning_rate": 5.8703834391907784e-05, "loss": 2.6125, "step": 52665 }, { "epoch": 4.13, "learning_rate": 5.870305026268329e-05, "loss": 2.7273, "step": 52666 }, { "epoch": 4.13, "learning_rate": 5.870226613345879e-05, "loss": 2.859, "step": 52667 }, { "epoch": 4.13, "learning_rate": 5.8701482004234295e-05, "loss": 2.5713, "step": 52668 }, { "epoch": 4.13, "learning_rate": 5.8700697875009806e-05, "loss": 2.4481, "step": 52669 }, { "epoch": 4.13, "learning_rate": 5.869991374578531e-05, "loss": 2.6892, "step": 52670 }, { "epoch": 4.13, "learning_rate": 5.869912961656081e-05, "loss": 2.6056, "step": 52671 }, { "epoch": 4.13, "learning_rate": 5.8698345487336317e-05, "loss": 2.7256, "step": 52672 }, { "epoch": 4.13, "learning_rate": 5.869756135811182e-05, "loss": 2.5118, "step": 52673 }, { "epoch": 4.13, "learning_rate": 5.8696777228887324e-05, "loss": 2.8104, "step": 52674 }, { "epoch": 4.13, "learning_rate": 5.869599309966283e-05, "loss": 2.7052, "step": 52675 }, { "epoch": 4.13, "learning_rate": 5.869520897043833e-05, "loss": 2.5831, "step": 52676 }, { "epoch": 4.13, "learning_rate": 5.8694424841213835e-05, "loss": 2.5585, "step": 52677 }, { "epoch": 4.13, "learning_rate": 5.869364071198934e-05, "loss": 2.6327, "step": 52678 }, { "epoch": 4.13, "learning_rate": 5.869285658276484e-05, "loss": 2.5262, "step": 52679 }, { "epoch": 4.13, "learning_rate": 5.8692072453540346e-05, "loss": 2.6171, "step": 52680 }, { "epoch": 4.13, "learning_rate": 5.869128832431585e-05, "loss": 2.5958, "step": 52681 }, { "epoch": 4.13, "learning_rate": 5.869050419509135e-05, "loss": 2.5465, "step": 52682 }, { "epoch": 4.13, "learning_rate": 5.868972006586686e-05, "loss": 2.7084, "step": 52683 }, { "epoch": 4.13, "learning_rate": 5.868893593664237e-05, "loss": 2.7825, "step": 52684 }, { "epoch": 4.13, "learning_rate": 5.868815180741787e-05, "loss": 2.7512, "step": 52685 }, { "epoch": 4.13, "learning_rate": 5.8687367678193375e-05, "loss": 2.6595, "step": 52686 }, { "epoch": 4.13, "learning_rate": 5.868658354896888e-05, "loss": 2.7755, "step": 52687 }, { "epoch": 4.13, "learning_rate": 5.868579941974438e-05, "loss": 2.4526, "step": 52688 }, { "epoch": 4.13, "learning_rate": 5.8685015290519886e-05, "loss": 2.7049, "step": 52689 }, { "epoch": 4.13, "learning_rate": 5.868423116129539e-05, "loss": 2.5061, "step": 52690 }, { "epoch": 4.13, "learning_rate": 5.868344703207089e-05, "loss": 2.6614, "step": 52691 }, { "epoch": 4.13, "learning_rate": 5.86826629028464e-05, "loss": 2.7212, "step": 52692 }, { "epoch": 4.13, "learning_rate": 5.86818787736219e-05, "loss": 2.5601, "step": 52693 }, { "epoch": 4.13, "learning_rate": 5.86810946443974e-05, "loss": 2.5541, "step": 52694 }, { "epoch": 4.13, "learning_rate": 5.86803105151729e-05, "loss": 2.6071, "step": 52695 }, { "epoch": 4.13, "learning_rate": 5.8679526385948405e-05, "loss": 2.7985, "step": 52696 }, { "epoch": 4.13, "learning_rate": 5.867874225672391e-05, "loss": 2.6988, "step": 52697 }, { "epoch": 4.13, "learning_rate": 5.867795812749941e-05, "loss": 2.7421, "step": 52698 }, { "epoch": 4.13, "learning_rate": 5.8677173998274916e-05, "loss": 2.532, "step": 52699 }, { "epoch": 4.13, "learning_rate": 5.867638986905042e-05, "loss": 2.422, "step": 52700 }, { "epoch": 4.13, "learning_rate": 5.867560573982592e-05, "loss": 2.6111, "step": 52701 }, { "epoch": 4.13, "learning_rate": 5.867482161060143e-05, "loss": 2.8354, "step": 52702 }, { "epoch": 4.13, "learning_rate": 5.867403748137693e-05, "loss": 2.4839, "step": 52703 }, { "epoch": 4.13, "learning_rate": 5.8673253352152434e-05, "loss": 2.5163, "step": 52704 }, { "epoch": 4.13, "learning_rate": 5.867246922292794e-05, "loss": 2.5539, "step": 52705 }, { "epoch": 4.13, "learning_rate": 5.867168509370344e-05, "loss": 3.0353, "step": 52706 }, { "epoch": 4.13, "learning_rate": 5.8670900964478945e-05, "loss": 2.6565, "step": 52707 }, { "epoch": 4.13, "learning_rate": 5.867011683525445e-05, "loss": 2.7573, "step": 52708 }, { "epoch": 4.13, "learning_rate": 5.866933270602995e-05, "loss": 2.5673, "step": 52709 }, { "epoch": 4.13, "learning_rate": 5.866854857680546e-05, "loss": 2.6911, "step": 52710 }, { "epoch": 4.13, "learning_rate": 5.866776444758097e-05, "loss": 2.7511, "step": 52711 }, { "epoch": 4.13, "learning_rate": 5.866698031835647e-05, "loss": 2.6099, "step": 52712 }, { "epoch": 4.13, "learning_rate": 5.8666196189131974e-05, "loss": 2.658, "step": 52713 }, { "epoch": 4.13, "learning_rate": 5.866541205990748e-05, "loss": 2.7386, "step": 52714 }, { "epoch": 4.13, "learning_rate": 5.866462793068298e-05, "loss": 2.7038, "step": 52715 }, { "epoch": 4.13, "learning_rate": 5.8663843801458485e-05, "loss": 2.5977, "step": 52716 }, { "epoch": 4.13, "learning_rate": 5.866305967223399e-05, "loss": 2.6554, "step": 52717 }, { "epoch": 4.13, "learning_rate": 5.866227554300949e-05, "loss": 2.7692, "step": 52718 }, { "epoch": 4.13, "learning_rate": 5.8661491413784996e-05, "loss": 2.6087, "step": 52719 }, { "epoch": 4.13, "learning_rate": 5.86607072845605e-05, "loss": 2.7765, "step": 52720 }, { "epoch": 4.13, "learning_rate": 5.8659923155336004e-05, "loss": 2.6327, "step": 52721 }, { "epoch": 4.13, "learning_rate": 5.865913902611151e-05, "loss": 2.848, "step": 52722 }, { "epoch": 4.13, "learning_rate": 5.865835489688701e-05, "loss": 2.6874, "step": 52723 }, { "epoch": 4.13, "learning_rate": 5.8657570767662515e-05, "loss": 2.5931, "step": 52724 }, { "epoch": 4.13, "learning_rate": 5.865678663843802e-05, "loss": 2.8453, "step": 52725 }, { "epoch": 4.13, "learning_rate": 5.865600250921353e-05, "loss": 2.4573, "step": 52726 }, { "epoch": 4.13, "learning_rate": 5.865521837998903e-05, "loss": 2.4555, "step": 52727 }, { "epoch": 4.13, "learning_rate": 5.8654434250764536e-05, "loss": 2.5331, "step": 52728 }, { "epoch": 4.13, "learning_rate": 5.865365012154004e-05, "loss": 2.7523, "step": 52729 }, { "epoch": 4.13, "learning_rate": 5.865286599231553e-05, "loss": 2.731, "step": 52730 }, { "epoch": 4.13, "learning_rate": 5.8652081863091034e-05, "loss": 2.6717, "step": 52731 }, { "epoch": 4.13, "learning_rate": 5.865129773386654e-05, "loss": 2.6121, "step": 52732 }, { "epoch": 4.13, "learning_rate": 5.865051360464204e-05, "loss": 2.5944, "step": 52733 }, { "epoch": 4.14, "learning_rate": 5.8649729475417545e-05, "loss": 2.7022, "step": 52734 }, { "epoch": 4.14, "learning_rate": 5.864894534619305e-05, "loss": 2.6801, "step": 52735 }, { "epoch": 4.14, "learning_rate": 5.864816121696856e-05, "loss": 2.5356, "step": 52736 }, { "epoch": 4.14, "learning_rate": 5.864737708774406e-05, "loss": 2.6693, "step": 52737 }, { "epoch": 4.14, "learning_rate": 5.8646592958519566e-05, "loss": 2.4418, "step": 52738 }, { "epoch": 4.14, "learning_rate": 5.864580882929507e-05, "loss": 2.5669, "step": 52739 }, { "epoch": 4.14, "learning_rate": 5.8645024700070573e-05, "loss": 2.7564, "step": 52740 }, { "epoch": 4.14, "learning_rate": 5.864424057084608e-05, "loss": 2.8185, "step": 52741 }, { "epoch": 4.14, "learning_rate": 5.864345644162158e-05, "loss": 2.6419, "step": 52742 }, { "epoch": 4.14, "learning_rate": 5.8642672312397084e-05, "loss": 2.8536, "step": 52743 }, { "epoch": 4.14, "learning_rate": 5.864188818317259e-05, "loss": 2.7068, "step": 52744 }, { "epoch": 4.14, "learning_rate": 5.864110405394809e-05, "loss": 2.6289, "step": 52745 }, { "epoch": 4.14, "learning_rate": 5.8640319924723595e-05, "loss": 2.6035, "step": 52746 }, { "epoch": 4.14, "learning_rate": 5.86395357954991e-05, "loss": 2.9746, "step": 52747 }, { "epoch": 4.14, "learning_rate": 5.86387516662746e-05, "loss": 2.6936, "step": 52748 }, { "epoch": 4.14, "learning_rate": 5.8637967537050107e-05, "loss": 2.6778, "step": 52749 }, { "epoch": 4.14, "learning_rate": 5.863718340782561e-05, "loss": 2.7644, "step": 52750 }, { "epoch": 4.14, "learning_rate": 5.863639927860112e-05, "loss": 2.5851, "step": 52751 }, { "epoch": 4.14, "learning_rate": 5.8635615149376624e-05, "loss": 2.5632, "step": 52752 }, { "epoch": 4.14, "learning_rate": 5.863483102015213e-05, "loss": 2.7024, "step": 52753 }, { "epoch": 4.14, "learning_rate": 5.863404689092763e-05, "loss": 2.6172, "step": 52754 }, { "epoch": 4.14, "learning_rate": 5.8633262761703135e-05, "loss": 2.6348, "step": 52755 }, { "epoch": 4.14, "learning_rate": 5.863247863247864e-05, "loss": 2.4467, "step": 52756 }, { "epoch": 4.14, "learning_rate": 5.863169450325414e-05, "loss": 2.6898, "step": 52757 }, { "epoch": 4.14, "learning_rate": 5.8630910374029646e-05, "loss": 2.7398, "step": 52758 }, { "epoch": 4.14, "learning_rate": 5.863012624480515e-05, "loss": 2.4323, "step": 52759 }, { "epoch": 4.14, "learning_rate": 5.8629342115580654e-05, "loss": 2.4349, "step": 52760 }, { "epoch": 4.14, "learning_rate": 5.862855798635616e-05, "loss": 2.699, "step": 52761 }, { "epoch": 4.14, "learning_rate": 5.862777385713166e-05, "loss": 2.6235, "step": 52762 }, { "epoch": 4.14, "learning_rate": 5.8626989727907165e-05, "loss": 2.4825, "step": 52763 }, { "epoch": 4.14, "learning_rate": 5.862620559868267e-05, "loss": 2.6786, "step": 52764 }, { "epoch": 4.14, "learning_rate": 5.862542146945817e-05, "loss": 2.4699, "step": 52765 }, { "epoch": 4.14, "learning_rate": 5.862463734023367e-05, "loss": 2.5209, "step": 52766 }, { "epoch": 4.14, "learning_rate": 5.862385321100917e-05, "loss": 2.6353, "step": 52767 }, { "epoch": 4.14, "learning_rate": 5.8623069081784676e-05, "loss": 2.5591, "step": 52768 }, { "epoch": 4.14, "learning_rate": 5.862228495256018e-05, "loss": 2.7093, "step": 52769 }, { "epoch": 4.14, "learning_rate": 5.8621500823335684e-05, "loss": 3.0362, "step": 52770 }, { "epoch": 4.14, "learning_rate": 5.862071669411119e-05, "loss": 2.6965, "step": 52771 }, { "epoch": 4.14, "learning_rate": 5.861993256488669e-05, "loss": 2.4961, "step": 52772 }, { "epoch": 4.14, "learning_rate": 5.8619148435662195e-05, "loss": 2.7033, "step": 52773 }, { "epoch": 4.14, "learning_rate": 5.86183643064377e-05, "loss": 2.7643, "step": 52774 }, { "epoch": 4.14, "learning_rate": 5.86175801772132e-05, "loss": 2.6163, "step": 52775 }, { "epoch": 4.14, "learning_rate": 5.8616796047988706e-05, "loss": 2.7692, "step": 52776 }, { "epoch": 4.14, "learning_rate": 5.8616011918764216e-05, "loss": 2.5727, "step": 52777 }, { "epoch": 4.14, "learning_rate": 5.861522778953972e-05, "loss": 2.7756, "step": 52778 }, { "epoch": 4.14, "learning_rate": 5.8614443660315224e-05, "loss": 2.6532, "step": 52779 }, { "epoch": 4.14, "learning_rate": 5.861365953109073e-05, "loss": 2.758, "step": 52780 }, { "epoch": 4.14, "learning_rate": 5.861287540186623e-05, "loss": 2.5797, "step": 52781 }, { "epoch": 4.14, "learning_rate": 5.8612091272641735e-05, "loss": 2.7592, "step": 52782 }, { "epoch": 4.14, "learning_rate": 5.861130714341724e-05, "loss": 2.6607, "step": 52783 }, { "epoch": 4.14, "learning_rate": 5.861052301419274e-05, "loss": 2.813, "step": 52784 }, { "epoch": 4.14, "learning_rate": 5.8609738884968246e-05, "loss": 2.7992, "step": 52785 }, { "epoch": 4.14, "learning_rate": 5.860895475574375e-05, "loss": 2.7316, "step": 52786 }, { "epoch": 4.14, "learning_rate": 5.860817062651925e-05, "loss": 2.5641, "step": 52787 }, { "epoch": 4.14, "learning_rate": 5.860738649729476e-05, "loss": 2.7188, "step": 52788 }, { "epoch": 4.14, "learning_rate": 5.860660236807026e-05, "loss": 2.6326, "step": 52789 }, { "epoch": 4.14, "learning_rate": 5.8605818238845764e-05, "loss": 2.6228, "step": 52790 }, { "epoch": 4.14, "learning_rate": 5.860503410962127e-05, "loss": 2.547, "step": 52791 }, { "epoch": 4.14, "learning_rate": 5.860424998039678e-05, "loss": 2.7395, "step": 52792 }, { "epoch": 4.14, "learning_rate": 5.860346585117228e-05, "loss": 2.5681, "step": 52793 }, { "epoch": 4.14, "learning_rate": 5.8602681721947786e-05, "loss": 2.6529, "step": 52794 }, { "epoch": 4.14, "learning_rate": 5.860189759272329e-05, "loss": 2.5909, "step": 52795 }, { "epoch": 4.14, "learning_rate": 5.860111346349879e-05, "loss": 2.6302, "step": 52796 }, { "epoch": 4.14, "learning_rate": 5.8600329334274297e-05, "loss": 2.5647, "step": 52797 }, { "epoch": 4.14, "learning_rate": 5.85995452050498e-05, "loss": 2.7257, "step": 52798 }, { "epoch": 4.14, "learning_rate": 5.8598761075825304e-05, "loss": 2.8082, "step": 52799 }, { "epoch": 4.14, "learning_rate": 5.859797694660081e-05, "loss": 2.7404, "step": 52800 }, { "epoch": 4.14, "learning_rate": 5.859719281737631e-05, "loss": 2.5651, "step": 52801 }, { "epoch": 4.14, "learning_rate": 5.859640868815181e-05, "loss": 2.6036, "step": 52802 }, { "epoch": 4.14, "learning_rate": 5.859562455892731e-05, "loss": 2.5758, "step": 52803 }, { "epoch": 4.14, "learning_rate": 5.8594840429702816e-05, "loss": 2.5366, "step": 52804 }, { "epoch": 4.14, "learning_rate": 5.859405630047832e-05, "loss": 2.681, "step": 52805 }, { "epoch": 4.14, "learning_rate": 5.859327217125382e-05, "loss": 2.6112, "step": 52806 }, { "epoch": 4.14, "learning_rate": 5.8592488042029327e-05, "loss": 2.7429, "step": 52807 }, { "epoch": 4.14, "learning_rate": 5.859170391280483e-05, "loss": 2.5713, "step": 52808 }, { "epoch": 4.14, "learning_rate": 5.8590919783580334e-05, "loss": 2.7357, "step": 52809 }, { "epoch": 4.14, "learning_rate": 5.859013565435584e-05, "loss": 2.5392, "step": 52810 }, { "epoch": 4.14, "learning_rate": 5.858935152513134e-05, "loss": 2.7896, "step": 52811 }, { "epoch": 4.14, "learning_rate": 5.8588567395906845e-05, "loss": 2.6887, "step": 52812 }, { "epoch": 4.14, "learning_rate": 5.858778326668235e-05, "loss": 2.7518, "step": 52813 }, { "epoch": 4.14, "learning_rate": 5.858699913745785e-05, "loss": 2.8956, "step": 52814 }, { "epoch": 4.14, "learning_rate": 5.8586215008233356e-05, "loss": 2.5666, "step": 52815 }, { "epoch": 4.14, "learning_rate": 5.858543087900886e-05, "loss": 2.6246, "step": 52816 }, { "epoch": 4.14, "learning_rate": 5.858464674978436e-05, "loss": 2.8424, "step": 52817 }, { "epoch": 4.14, "learning_rate": 5.8583862620559874e-05, "loss": 2.6856, "step": 52818 }, { "epoch": 4.14, "learning_rate": 5.858307849133538e-05, "loss": 2.7508, "step": 52819 }, { "epoch": 4.14, "learning_rate": 5.858229436211088e-05, "loss": 2.6393, "step": 52820 }, { "epoch": 4.14, "learning_rate": 5.8581510232886385e-05, "loss": 2.5146, "step": 52821 }, { "epoch": 4.14, "learning_rate": 5.858072610366189e-05, "loss": 2.6651, "step": 52822 }, { "epoch": 4.14, "learning_rate": 5.857994197443739e-05, "loss": 2.3822, "step": 52823 }, { "epoch": 4.14, "learning_rate": 5.8579157845212896e-05, "loss": 2.8065, "step": 52824 }, { "epoch": 4.14, "learning_rate": 5.85783737159884e-05, "loss": 2.6407, "step": 52825 }, { "epoch": 4.14, "learning_rate": 5.85775895867639e-05, "loss": 2.6174, "step": 52826 }, { "epoch": 4.14, "learning_rate": 5.857680545753941e-05, "loss": 2.6875, "step": 52827 }, { "epoch": 4.14, "learning_rate": 5.857602132831491e-05, "loss": 2.61, "step": 52828 }, { "epoch": 4.14, "learning_rate": 5.8575237199090414e-05, "loss": 2.7474, "step": 52829 }, { "epoch": 4.14, "learning_rate": 5.857445306986592e-05, "loss": 2.7119, "step": 52830 }, { "epoch": 4.14, "learning_rate": 5.857366894064142e-05, "loss": 2.5738, "step": 52831 }, { "epoch": 4.14, "learning_rate": 5.8572884811416925e-05, "loss": 2.625, "step": 52832 }, { "epoch": 4.14, "learning_rate": 5.857210068219243e-05, "loss": 2.6779, "step": 52833 }, { "epoch": 4.14, "learning_rate": 5.857131655296794e-05, "loss": 2.5765, "step": 52834 }, { "epoch": 4.14, "learning_rate": 5.857053242374344e-05, "loss": 2.6198, "step": 52835 }, { "epoch": 4.14, "learning_rate": 5.856974829451895e-05, "loss": 2.5165, "step": 52836 }, { "epoch": 4.14, "learning_rate": 5.856896416529445e-05, "loss": 2.5639, "step": 52837 }, { "epoch": 4.14, "learning_rate": 5.856818003606994e-05, "loss": 2.844, "step": 52838 }, { "epoch": 4.14, "learning_rate": 5.8567395906845444e-05, "loss": 2.6915, "step": 52839 }, { "epoch": 4.14, "learning_rate": 5.856661177762095e-05, "loss": 2.7276, "step": 52840 }, { "epoch": 4.14, "learning_rate": 5.856582764839645e-05, "loss": 2.6279, "step": 52841 }, { "epoch": 4.14, "learning_rate": 5.8565043519171955e-05, "loss": 2.7941, "step": 52842 }, { "epoch": 4.14, "learning_rate": 5.856425938994746e-05, "loss": 2.5639, "step": 52843 }, { "epoch": 4.14, "learning_rate": 5.856347526072297e-05, "loss": 2.6667, "step": 52844 }, { "epoch": 4.14, "learning_rate": 5.856269113149847e-05, "loss": 2.5489, "step": 52845 }, { "epoch": 4.14, "learning_rate": 5.856190700227398e-05, "loss": 2.6346, "step": 52846 }, { "epoch": 4.14, "learning_rate": 5.856112287304948e-05, "loss": 2.805, "step": 52847 }, { "epoch": 4.14, "learning_rate": 5.8560338743824984e-05, "loss": 2.6359, "step": 52848 }, { "epoch": 4.14, "learning_rate": 5.855955461460049e-05, "loss": 2.7151, "step": 52849 }, { "epoch": 4.14, "learning_rate": 5.855877048537599e-05, "loss": 2.7672, "step": 52850 }, { "epoch": 4.14, "learning_rate": 5.8557986356151495e-05, "loss": 2.9554, "step": 52851 }, { "epoch": 4.14, "learning_rate": 5.8557202226927e-05, "loss": 2.5923, "step": 52852 }, { "epoch": 4.14, "learning_rate": 5.85564180977025e-05, "loss": 2.5649, "step": 52853 }, { "epoch": 4.14, "learning_rate": 5.8555633968478006e-05, "loss": 2.737, "step": 52854 }, { "epoch": 4.14, "learning_rate": 5.855484983925351e-05, "loss": 2.5693, "step": 52855 }, { "epoch": 4.14, "learning_rate": 5.8554065710029014e-05, "loss": 2.6008, "step": 52856 }, { "epoch": 4.14, "learning_rate": 5.855328158080452e-05, "loss": 2.5993, "step": 52857 }, { "epoch": 4.14, "learning_rate": 5.855249745158002e-05, "loss": 2.7038, "step": 52858 }, { "epoch": 4.14, "learning_rate": 5.855171332235553e-05, "loss": 2.7755, "step": 52859 }, { "epoch": 4.14, "learning_rate": 5.8550929193131035e-05, "loss": 2.4629, "step": 52860 }, { "epoch": 4.14, "learning_rate": 5.855014506390654e-05, "loss": 2.8287, "step": 52861 }, { "epoch": 4.15, "learning_rate": 5.854936093468204e-05, "loss": 2.5973, "step": 52862 }, { "epoch": 4.15, "learning_rate": 5.8548576805457546e-05, "loss": 2.6309, "step": 52863 }, { "epoch": 4.15, "learning_rate": 5.854779267623305e-05, "loss": 2.4092, "step": 52864 }, { "epoch": 4.15, "learning_rate": 5.8547008547008553e-05, "loss": 2.3983, "step": 52865 }, { "epoch": 4.15, "learning_rate": 5.854622441778406e-05, "loss": 2.6843, "step": 52866 }, { "epoch": 4.15, "learning_rate": 5.854544028855956e-05, "loss": 2.6272, "step": 52867 }, { "epoch": 4.15, "learning_rate": 5.8544656159335064e-05, "loss": 2.7891, "step": 52868 }, { "epoch": 4.15, "learning_rate": 5.854387203011057e-05, "loss": 2.6063, "step": 52869 }, { "epoch": 4.15, "learning_rate": 5.854308790088607e-05, "loss": 2.5728, "step": 52870 }, { "epoch": 4.15, "learning_rate": 5.8542303771661576e-05, "loss": 2.7031, "step": 52871 }, { "epoch": 4.15, "learning_rate": 5.854151964243708e-05, "loss": 2.784, "step": 52872 }, { "epoch": 4.15, "learning_rate": 5.854073551321258e-05, "loss": 2.8814, "step": 52873 }, { "epoch": 4.15, "learning_rate": 5.853995138398808e-05, "loss": 2.4621, "step": 52874 }, { "epoch": 4.15, "learning_rate": 5.8539167254763583e-05, "loss": 2.5273, "step": 52875 }, { "epoch": 4.15, "learning_rate": 5.853838312553909e-05, "loss": 2.5204, "step": 52876 }, { "epoch": 4.15, "learning_rate": 5.853759899631459e-05, "loss": 2.5945, "step": 52877 }, { "epoch": 4.15, "learning_rate": 5.8536814867090094e-05, "loss": 2.7356, "step": 52878 }, { "epoch": 4.15, "learning_rate": 5.85360307378656e-05, "loss": 2.8215, "step": 52879 }, { "epoch": 4.15, "learning_rate": 5.85352466086411e-05, "loss": 2.6254, "step": 52880 }, { "epoch": 4.15, "learning_rate": 5.8534462479416606e-05, "loss": 2.8551, "step": 52881 }, { "epoch": 4.15, "learning_rate": 5.853367835019211e-05, "loss": 2.6918, "step": 52882 }, { "epoch": 4.15, "learning_rate": 5.853289422096761e-05, "loss": 2.6373, "step": 52883 }, { "epoch": 4.15, "learning_rate": 5.8532110091743117e-05, "loss": 2.8033, "step": 52884 }, { "epoch": 4.15, "learning_rate": 5.853132596251863e-05, "loss": 2.5509, "step": 52885 }, { "epoch": 4.15, "learning_rate": 5.853054183329413e-05, "loss": 2.6689, "step": 52886 }, { "epoch": 4.15, "learning_rate": 5.8529757704069634e-05, "loss": 2.8425, "step": 52887 }, { "epoch": 4.15, "learning_rate": 5.852897357484514e-05, "loss": 2.4925, "step": 52888 }, { "epoch": 4.15, "learning_rate": 5.852818944562064e-05, "loss": 2.6562, "step": 52889 }, { "epoch": 4.15, "learning_rate": 5.8527405316396145e-05, "loss": 2.5068, "step": 52890 }, { "epoch": 4.15, "learning_rate": 5.852662118717165e-05, "loss": 2.6412, "step": 52891 }, { "epoch": 4.15, "learning_rate": 5.852583705794715e-05, "loss": 2.7544, "step": 52892 }, { "epoch": 4.15, "learning_rate": 5.8525052928722656e-05, "loss": 2.5959, "step": 52893 }, { "epoch": 4.15, "learning_rate": 5.852426879949816e-05, "loss": 2.9459, "step": 52894 }, { "epoch": 4.15, "learning_rate": 5.8523484670273664e-05, "loss": 2.5231, "step": 52895 }, { "epoch": 4.15, "learning_rate": 5.852270054104917e-05, "loss": 2.5587, "step": 52896 }, { "epoch": 4.15, "learning_rate": 5.852191641182467e-05, "loss": 2.4364, "step": 52897 }, { "epoch": 4.15, "learning_rate": 5.8521132282600175e-05, "loss": 2.6747, "step": 52898 }, { "epoch": 4.15, "learning_rate": 5.852034815337568e-05, "loss": 2.5191, "step": 52899 }, { "epoch": 4.15, "learning_rate": 5.851956402415118e-05, "loss": 2.747, "step": 52900 }, { "epoch": 4.15, "learning_rate": 5.851877989492669e-05, "loss": 2.6093, "step": 52901 }, { "epoch": 4.15, "learning_rate": 5.8517995765702196e-05, "loss": 2.7093, "step": 52902 }, { "epoch": 4.15, "learning_rate": 5.85172116364777e-05, "loss": 2.7891, "step": 52903 }, { "epoch": 4.15, "learning_rate": 5.8516427507253204e-05, "loss": 2.7185, "step": 52904 }, { "epoch": 4.15, "learning_rate": 5.851564337802871e-05, "loss": 2.7352, "step": 52905 }, { "epoch": 4.15, "learning_rate": 5.851485924880421e-05, "loss": 2.4786, "step": 52906 }, { "epoch": 4.15, "learning_rate": 5.8514075119579715e-05, "loss": 2.4664, "step": 52907 }, { "epoch": 4.15, "learning_rate": 5.851329099035522e-05, "loss": 2.6007, "step": 52908 }, { "epoch": 4.15, "learning_rate": 5.851250686113072e-05, "loss": 2.5854, "step": 52909 }, { "epoch": 4.15, "learning_rate": 5.851172273190621e-05, "loss": 2.4913, "step": 52910 }, { "epoch": 4.15, "learning_rate": 5.851093860268172e-05, "loss": 2.7127, "step": 52911 }, { "epoch": 4.15, "learning_rate": 5.8510154473457226e-05, "loss": 2.6419, "step": 52912 }, { "epoch": 4.15, "learning_rate": 5.850937034423273e-05, "loss": 2.5553, "step": 52913 }, { "epoch": 4.15, "learning_rate": 5.8508586215008234e-05, "loss": 2.7005, "step": 52914 }, { "epoch": 4.15, "learning_rate": 5.850780208578374e-05, "loss": 2.7553, "step": 52915 }, { "epoch": 4.15, "learning_rate": 5.850701795655924e-05, "loss": 2.7131, "step": 52916 }, { "epoch": 4.15, "learning_rate": 5.8506233827334745e-05, "loss": 2.5588, "step": 52917 }, { "epoch": 4.15, "learning_rate": 5.850544969811025e-05, "loss": 2.6385, "step": 52918 }, { "epoch": 4.15, "learning_rate": 5.850466556888575e-05, "loss": 2.6964, "step": 52919 }, { "epoch": 4.15, "learning_rate": 5.8503881439661256e-05, "loss": 2.5595, "step": 52920 }, { "epoch": 4.15, "learning_rate": 5.850309731043676e-05, "loss": 2.7787, "step": 52921 }, { "epoch": 4.15, "learning_rate": 5.850231318121226e-05, "loss": 2.5289, "step": 52922 }, { "epoch": 4.15, "learning_rate": 5.850152905198777e-05, "loss": 2.5911, "step": 52923 }, { "epoch": 4.15, "learning_rate": 5.850074492276327e-05, "loss": 2.6511, "step": 52924 }, { "epoch": 4.15, "learning_rate": 5.8499960793538774e-05, "loss": 2.6298, "step": 52925 }, { "epoch": 4.15, "learning_rate": 5.8499176664314285e-05, "loss": 2.4053, "step": 52926 }, { "epoch": 4.15, "learning_rate": 5.849839253508979e-05, "loss": 2.7274, "step": 52927 }, { "epoch": 4.15, "learning_rate": 5.849760840586529e-05, "loss": 2.8433, "step": 52928 }, { "epoch": 4.15, "learning_rate": 5.8496824276640796e-05, "loss": 2.57, "step": 52929 }, { "epoch": 4.15, "learning_rate": 5.84960401474163e-05, "loss": 2.8612, "step": 52930 }, { "epoch": 4.15, "learning_rate": 5.84952560181918e-05, "loss": 2.6386, "step": 52931 }, { "epoch": 4.15, "learning_rate": 5.849447188896731e-05, "loss": 2.7869, "step": 52932 }, { "epoch": 4.15, "learning_rate": 5.849368775974281e-05, "loss": 2.6859, "step": 52933 }, { "epoch": 4.15, "learning_rate": 5.8492903630518314e-05, "loss": 2.8392, "step": 52934 }, { "epoch": 4.15, "learning_rate": 5.849211950129382e-05, "loss": 2.7315, "step": 52935 }, { "epoch": 4.15, "learning_rate": 5.849133537206932e-05, "loss": 2.7681, "step": 52936 }, { "epoch": 4.15, "learning_rate": 5.8490551242844825e-05, "loss": 2.6247, "step": 52937 }, { "epoch": 4.15, "learning_rate": 5.848976711362033e-05, "loss": 2.6485, "step": 52938 }, { "epoch": 4.15, "learning_rate": 5.848898298439583e-05, "loss": 2.5576, "step": 52939 }, { "epoch": 4.15, "learning_rate": 5.8488198855171336e-05, "loss": 2.8725, "step": 52940 }, { "epoch": 4.15, "learning_rate": 5.848741472594684e-05, "loss": 2.5307, "step": 52941 }, { "epoch": 4.15, "learning_rate": 5.848663059672235e-05, "loss": 2.629, "step": 52942 }, { "epoch": 4.15, "learning_rate": 5.8485846467497854e-05, "loss": 2.8249, "step": 52943 }, { "epoch": 4.15, "learning_rate": 5.848506233827336e-05, "loss": 2.6802, "step": 52944 }, { "epoch": 4.15, "learning_rate": 5.848427820904886e-05, "loss": 2.563, "step": 52945 }, { "epoch": 4.15, "learning_rate": 5.848349407982435e-05, "loss": 2.7552, "step": 52946 }, { "epoch": 4.15, "learning_rate": 5.8482709950599855e-05, "loss": 2.7222, "step": 52947 }, { "epoch": 4.15, "learning_rate": 5.848192582137536e-05, "loss": 2.7744, "step": 52948 }, { "epoch": 4.15, "learning_rate": 5.848114169215086e-05, "loss": 2.523, "step": 52949 }, { "epoch": 4.15, "learning_rate": 5.8480357562926366e-05, "loss": 2.5875, "step": 52950 }, { "epoch": 4.15, "learning_rate": 5.847957343370187e-05, "loss": 2.544, "step": 52951 }, { "epoch": 4.15, "learning_rate": 5.847878930447738e-05, "loss": 2.5559, "step": 52952 }, { "epoch": 4.15, "learning_rate": 5.8478005175252884e-05, "loss": 2.6015, "step": 52953 }, { "epoch": 4.15, "learning_rate": 5.847722104602839e-05, "loss": 2.6472, "step": 52954 }, { "epoch": 4.15, "learning_rate": 5.847643691680389e-05, "loss": 2.5296, "step": 52955 }, { "epoch": 4.15, "learning_rate": 5.8475652787579395e-05, "loss": 2.5292, "step": 52956 }, { "epoch": 4.15, "learning_rate": 5.84748686583549e-05, "loss": 2.6481, "step": 52957 }, { "epoch": 4.15, "learning_rate": 5.84740845291304e-05, "loss": 2.8209, "step": 52958 }, { "epoch": 4.15, "learning_rate": 5.8473300399905906e-05, "loss": 2.7335, "step": 52959 }, { "epoch": 4.15, "learning_rate": 5.847251627068141e-05, "loss": 2.5381, "step": 52960 }, { "epoch": 4.15, "learning_rate": 5.847173214145691e-05, "loss": 2.6165, "step": 52961 }, { "epoch": 4.15, "learning_rate": 5.847094801223242e-05, "loss": 2.5063, "step": 52962 }, { "epoch": 4.15, "learning_rate": 5.847016388300792e-05, "loss": 2.6906, "step": 52963 }, { "epoch": 4.15, "learning_rate": 5.8469379753783424e-05, "loss": 2.5274, "step": 52964 }, { "epoch": 4.15, "learning_rate": 5.846859562455893e-05, "loss": 2.5005, "step": 52965 }, { "epoch": 4.15, "learning_rate": 5.846781149533443e-05, "loss": 2.5635, "step": 52966 }, { "epoch": 4.15, "learning_rate": 5.846702736610994e-05, "loss": 2.7779, "step": 52967 }, { "epoch": 4.15, "learning_rate": 5.8466243236885446e-05, "loss": 2.6146, "step": 52968 }, { "epoch": 4.15, "learning_rate": 5.846545910766095e-05, "loss": 2.6792, "step": 52969 }, { "epoch": 4.15, "learning_rate": 5.846467497843645e-05, "loss": 2.4812, "step": 52970 }, { "epoch": 4.15, "learning_rate": 5.846389084921196e-05, "loss": 2.5425, "step": 52971 }, { "epoch": 4.15, "learning_rate": 5.846310671998746e-05, "loss": 2.671, "step": 52972 }, { "epoch": 4.15, "learning_rate": 5.8462322590762964e-05, "loss": 2.5314, "step": 52973 }, { "epoch": 4.15, "learning_rate": 5.846153846153847e-05, "loss": 2.671, "step": 52974 }, { "epoch": 4.15, "learning_rate": 5.846075433231397e-05, "loss": 2.6487, "step": 52975 }, { "epoch": 4.15, "learning_rate": 5.8459970203089475e-05, "loss": 2.6803, "step": 52976 }, { "epoch": 4.15, "learning_rate": 5.845918607386498e-05, "loss": 2.5052, "step": 52977 }, { "epoch": 4.15, "learning_rate": 5.845840194464048e-05, "loss": 2.6966, "step": 52978 }, { "epoch": 4.15, "learning_rate": 5.8457617815415986e-05, "loss": 2.6016, "step": 52979 }, { "epoch": 4.15, "learning_rate": 5.845683368619149e-05, "loss": 2.805, "step": 52980 }, { "epoch": 4.15, "learning_rate": 5.8456049556966994e-05, "loss": 2.6599, "step": 52981 }, { "epoch": 4.15, "learning_rate": 5.845526542774249e-05, "loss": 2.6808, "step": 52982 }, { "epoch": 4.15, "learning_rate": 5.8454481298517994e-05, "loss": 2.7088, "step": 52983 }, { "epoch": 4.15, "learning_rate": 5.84536971692935e-05, "loss": 2.9034, "step": 52984 }, { "epoch": 4.15, "learning_rate": 5.8452913040069e-05, "loss": 2.7158, "step": 52985 }, { "epoch": 4.15, "learning_rate": 5.8452128910844505e-05, "loss": 2.9054, "step": 52986 }, { "epoch": 4.15, "learning_rate": 5.845134478162001e-05, "loss": 2.5494, "step": 52987 }, { "epoch": 4.15, "learning_rate": 5.845056065239551e-05, "loss": 2.913, "step": 52988 }, { "epoch": 4.16, "learning_rate": 5.8449776523171016e-05, "loss": 2.6686, "step": 52989 }, { "epoch": 4.16, "learning_rate": 5.844899239394652e-05, "loss": 2.6986, "step": 52990 }, { "epoch": 4.16, "learning_rate": 5.8448208264722024e-05, "loss": 2.6597, "step": 52991 }, { "epoch": 4.16, "learning_rate": 5.844742413549753e-05, "loss": 2.5303, "step": 52992 }, { "epoch": 4.16, "learning_rate": 5.844664000627304e-05, "loss": 2.5571, "step": 52993 }, { "epoch": 4.16, "learning_rate": 5.844585587704854e-05, "loss": 2.6387, "step": 52994 }, { "epoch": 4.16, "learning_rate": 5.8445071747824045e-05, "loss": 2.6662, "step": 52995 }, { "epoch": 4.16, "learning_rate": 5.844428761859955e-05, "loss": 2.8299, "step": 52996 }, { "epoch": 4.16, "learning_rate": 5.844350348937505e-05, "loss": 2.6607, "step": 52997 }, { "epoch": 4.16, "learning_rate": 5.8442719360150556e-05, "loss": 2.8019, "step": 52998 }, { "epoch": 4.16, "learning_rate": 5.844193523092606e-05, "loss": 2.7304, "step": 52999 }, { "epoch": 4.16, "learning_rate": 5.8441151101701564e-05, "loss": 2.6548, "step": 53000 }, { "epoch": 4.16, "eval_loss": 2.899749517440796, "eval_runtime": 218.0315, "eval_samples_per_second": 51.974, "eval_steps_per_second": 3.252, "step": 53000 }, { "epoch": 4.16, "learning_rate": 5.844036697247707e-05, "loss": 2.5789, "step": 53001 }, { "epoch": 4.16, "learning_rate": 5.843958284325257e-05, "loss": 2.4862, "step": 53002 }, { "epoch": 4.16, "learning_rate": 5.8438798714028075e-05, "loss": 2.7293, "step": 53003 }, { "epoch": 4.16, "learning_rate": 5.843801458480358e-05, "loss": 2.6321, "step": 53004 }, { "epoch": 4.16, "learning_rate": 5.843723045557908e-05, "loss": 2.6204, "step": 53005 }, { "epoch": 4.16, "learning_rate": 5.8436446326354586e-05, "loss": 2.5961, "step": 53006 }, { "epoch": 4.16, "learning_rate": 5.843566219713009e-05, "loss": 2.6398, "step": 53007 }, { "epoch": 4.16, "learning_rate": 5.843487806790559e-05, "loss": 2.8641, "step": 53008 }, { "epoch": 4.16, "learning_rate": 5.84340939386811e-05, "loss": 2.5688, "step": 53009 }, { "epoch": 4.16, "learning_rate": 5.843330980945661e-05, "loss": 2.7167, "step": 53010 }, { "epoch": 4.16, "learning_rate": 5.843252568023211e-05, "loss": 2.59, "step": 53011 }, { "epoch": 4.16, "learning_rate": 5.8431741551007614e-05, "loss": 2.6794, "step": 53012 }, { "epoch": 4.16, "learning_rate": 5.843095742178312e-05, "loss": 2.6071, "step": 53013 }, { "epoch": 4.16, "learning_rate": 5.843017329255862e-05, "loss": 2.6845, "step": 53014 }, { "epoch": 4.16, "learning_rate": 5.8429389163334125e-05, "loss": 2.8788, "step": 53015 }, { "epoch": 4.16, "learning_rate": 5.842860503410963e-05, "loss": 2.5895, "step": 53016 }, { "epoch": 4.16, "learning_rate": 5.842782090488513e-05, "loss": 2.6128, "step": 53017 }, { "epoch": 4.16, "learning_rate": 5.842703677566062e-05, "loss": 2.6306, "step": 53018 }, { "epoch": 4.16, "learning_rate": 5.842625264643613e-05, "loss": 2.5147, "step": 53019 }, { "epoch": 4.16, "learning_rate": 5.842546851721164e-05, "loss": 2.615, "step": 53020 }, { "epoch": 4.16, "learning_rate": 5.842468438798714e-05, "loss": 2.7992, "step": 53021 }, { "epoch": 4.16, "learning_rate": 5.8423900258762644e-05, "loss": 2.7272, "step": 53022 }, { "epoch": 4.16, "learning_rate": 5.842311612953815e-05, "loss": 2.5522, "step": 53023 }, { "epoch": 4.16, "learning_rate": 5.842233200031365e-05, "loss": 2.7762, "step": 53024 }, { "epoch": 4.16, "learning_rate": 5.8421547871089155e-05, "loss": 2.8251, "step": 53025 }, { "epoch": 4.16, "learning_rate": 5.842076374186466e-05, "loss": 2.6541, "step": 53026 }, { "epoch": 4.16, "learning_rate": 5.841997961264016e-05, "loss": 2.5941, "step": 53027 }, { "epoch": 4.16, "learning_rate": 5.8419195483415666e-05, "loss": 2.4826, "step": 53028 }, { "epoch": 4.16, "learning_rate": 5.841841135419117e-05, "loss": 2.629, "step": 53029 }, { "epoch": 4.16, "learning_rate": 5.8417627224966674e-05, "loss": 2.6797, "step": 53030 }, { "epoch": 4.16, "learning_rate": 5.841684309574218e-05, "loss": 2.5939, "step": 53031 }, { "epoch": 4.16, "learning_rate": 5.841605896651768e-05, "loss": 2.6494, "step": 53032 }, { "epoch": 4.16, "learning_rate": 5.8415274837293185e-05, "loss": 2.7338, "step": 53033 }, { "epoch": 4.16, "learning_rate": 5.8414490708068695e-05, "loss": 2.6108, "step": 53034 }, { "epoch": 4.16, "learning_rate": 5.84137065788442e-05, "loss": 2.5313, "step": 53035 }, { "epoch": 4.16, "learning_rate": 5.84129224496197e-05, "loss": 2.4573, "step": 53036 }, { "epoch": 4.16, "learning_rate": 5.8412138320395206e-05, "loss": 2.9053, "step": 53037 }, { "epoch": 4.16, "learning_rate": 5.841135419117071e-05, "loss": 2.7859, "step": 53038 }, { "epoch": 4.16, "learning_rate": 5.8410570061946214e-05, "loss": 2.6143, "step": 53039 }, { "epoch": 4.16, "learning_rate": 5.840978593272172e-05, "loss": 2.7738, "step": 53040 }, { "epoch": 4.16, "learning_rate": 5.840900180349722e-05, "loss": 2.6382, "step": 53041 }, { "epoch": 4.16, "learning_rate": 5.8408217674272725e-05, "loss": 2.5236, "step": 53042 }, { "epoch": 4.16, "learning_rate": 5.840743354504823e-05, "loss": 2.7737, "step": 53043 }, { "epoch": 4.16, "learning_rate": 5.840664941582373e-05, "loss": 2.6034, "step": 53044 }, { "epoch": 4.16, "learning_rate": 5.8405865286599236e-05, "loss": 2.7623, "step": 53045 }, { "epoch": 4.16, "learning_rate": 5.840508115737474e-05, "loss": 2.5297, "step": 53046 }, { "epoch": 4.16, "learning_rate": 5.840429702815024e-05, "loss": 2.8929, "step": 53047 }, { "epoch": 4.16, "learning_rate": 5.840351289892575e-05, "loss": 2.4158, "step": 53048 }, { "epoch": 4.16, "learning_rate": 5.840272876970125e-05, "loss": 2.8036, "step": 53049 }, { "epoch": 4.16, "learning_rate": 5.840194464047676e-05, "loss": 2.6695, "step": 53050 }, { "epoch": 4.16, "learning_rate": 5.8401160511252265e-05, "loss": 2.7309, "step": 53051 }, { "epoch": 4.16, "learning_rate": 5.840037638202777e-05, "loss": 2.7797, "step": 53052 }, { "epoch": 4.16, "learning_rate": 5.839959225280327e-05, "loss": 2.6686, "step": 53053 }, { "epoch": 4.16, "learning_rate": 5.839880812357876e-05, "loss": 2.6065, "step": 53054 }, { "epoch": 4.16, "learning_rate": 5.8398023994354266e-05, "loss": 2.7423, "step": 53055 }, { "epoch": 4.16, "learning_rate": 5.839723986512977e-05, "loss": 2.5761, "step": 53056 }, { "epoch": 4.16, "learning_rate": 5.839645573590527e-05, "loss": 2.6695, "step": 53057 }, { "epoch": 4.16, "learning_rate": 5.839567160668078e-05, "loss": 2.3904, "step": 53058 }, { "epoch": 4.16, "learning_rate": 5.839488747745628e-05, "loss": 2.6911, "step": 53059 }, { "epoch": 4.16, "learning_rate": 5.839410334823179e-05, "loss": 2.624, "step": 53060 }, { "epoch": 4.16, "learning_rate": 5.8393319219007295e-05, "loss": 2.8044, "step": 53061 }, { "epoch": 4.16, "learning_rate": 5.83925350897828e-05, "loss": 2.7768, "step": 53062 }, { "epoch": 4.16, "learning_rate": 5.83917509605583e-05, "loss": 2.7622, "step": 53063 }, { "epoch": 4.16, "learning_rate": 5.8390966831333806e-05, "loss": 2.7294, "step": 53064 }, { "epoch": 4.16, "learning_rate": 5.839018270210931e-05, "loss": 2.5343, "step": 53065 }, { "epoch": 4.16, "learning_rate": 5.838939857288481e-05, "loss": 2.8314, "step": 53066 }, { "epoch": 4.16, "learning_rate": 5.838861444366032e-05, "loss": 2.541, "step": 53067 }, { "epoch": 4.16, "learning_rate": 5.838783031443582e-05, "loss": 2.6448, "step": 53068 }, { "epoch": 4.16, "learning_rate": 5.8387046185211324e-05, "loss": 2.6568, "step": 53069 }, { "epoch": 4.16, "learning_rate": 5.838626205598683e-05, "loss": 2.6355, "step": 53070 }, { "epoch": 4.16, "learning_rate": 5.838547792676233e-05, "loss": 2.6803, "step": 53071 }, { "epoch": 4.16, "learning_rate": 5.8384693797537835e-05, "loss": 2.7752, "step": 53072 }, { "epoch": 4.16, "learning_rate": 5.838390966831334e-05, "loss": 2.67, "step": 53073 }, { "epoch": 4.16, "learning_rate": 5.838312553908884e-05, "loss": 2.7132, "step": 53074 }, { "epoch": 4.16, "learning_rate": 5.838234140986435e-05, "loss": 2.6662, "step": 53075 }, { "epoch": 4.16, "learning_rate": 5.8381557280639857e-05, "loss": 2.696, "step": 53076 }, { "epoch": 4.16, "learning_rate": 5.838077315141536e-05, "loss": 2.6001, "step": 53077 }, { "epoch": 4.16, "learning_rate": 5.8379989022190864e-05, "loss": 2.7813, "step": 53078 }, { "epoch": 4.16, "learning_rate": 5.837920489296637e-05, "loss": 2.5219, "step": 53079 }, { "epoch": 4.16, "learning_rate": 5.837842076374187e-05, "loss": 2.7386, "step": 53080 }, { "epoch": 4.16, "learning_rate": 5.8377636634517375e-05, "loss": 2.7255, "step": 53081 }, { "epoch": 4.16, "learning_rate": 5.837685250529288e-05, "loss": 2.6563, "step": 53082 }, { "epoch": 4.16, "learning_rate": 5.837606837606838e-05, "loss": 2.7722, "step": 53083 }, { "epoch": 4.16, "learning_rate": 5.8375284246843886e-05, "loss": 2.6504, "step": 53084 }, { "epoch": 4.16, "learning_rate": 5.837450011761939e-05, "loss": 2.7348, "step": 53085 }, { "epoch": 4.16, "learning_rate": 5.837371598839489e-05, "loss": 2.7824, "step": 53086 }, { "epoch": 4.16, "learning_rate": 5.83729318591704e-05, "loss": 2.6335, "step": 53087 }, { "epoch": 4.16, "learning_rate": 5.83721477299459e-05, "loss": 2.6096, "step": 53088 }, { "epoch": 4.16, "learning_rate": 5.8371363600721404e-05, "loss": 2.6726, "step": 53089 }, { "epoch": 4.16, "learning_rate": 5.83705794714969e-05, "loss": 2.6435, "step": 53090 }, { "epoch": 4.16, "learning_rate": 5.8369795342272405e-05, "loss": 2.6797, "step": 53091 }, { "epoch": 4.16, "learning_rate": 5.836901121304791e-05, "loss": 2.5865, "step": 53092 }, { "epoch": 4.16, "learning_rate": 5.836822708382341e-05, "loss": 2.6179, "step": 53093 }, { "epoch": 4.16, "learning_rate": 5.8367442954598916e-05, "loss": 2.7108, "step": 53094 }, { "epoch": 4.16, "learning_rate": 5.836665882537442e-05, "loss": 2.7316, "step": 53095 }, { "epoch": 4.16, "learning_rate": 5.836587469614992e-05, "loss": 2.7798, "step": 53096 }, { "epoch": 4.16, "learning_rate": 5.836509056692543e-05, "loss": 2.588, "step": 53097 }, { "epoch": 4.16, "learning_rate": 5.836430643770093e-05, "loss": 2.7575, "step": 53098 }, { "epoch": 4.16, "learning_rate": 5.8363522308476434e-05, "loss": 2.6239, "step": 53099 }, { "epoch": 4.16, "learning_rate": 5.836273817925194e-05, "loss": 2.7563, "step": 53100 }, { "epoch": 4.16, "learning_rate": 5.836195405002745e-05, "loss": 2.6185, "step": 53101 }, { "epoch": 4.16, "learning_rate": 5.836116992080295e-05, "loss": 2.4256, "step": 53102 }, { "epoch": 4.16, "learning_rate": 5.8360385791578456e-05, "loss": 2.7094, "step": 53103 }, { "epoch": 4.16, "learning_rate": 5.835960166235396e-05, "loss": 2.5878, "step": 53104 }, { "epoch": 4.16, "learning_rate": 5.835881753312946e-05, "loss": 2.5729, "step": 53105 }, { "epoch": 4.16, "learning_rate": 5.835803340390497e-05, "loss": 2.5988, "step": 53106 }, { "epoch": 4.16, "learning_rate": 5.835724927468047e-05, "loss": 2.6874, "step": 53107 }, { "epoch": 4.16, "learning_rate": 5.8356465145455974e-05, "loss": 2.6462, "step": 53108 }, { "epoch": 4.16, "learning_rate": 5.835568101623148e-05, "loss": 2.7945, "step": 53109 }, { "epoch": 4.16, "learning_rate": 5.835489688700698e-05, "loss": 2.719, "step": 53110 }, { "epoch": 4.16, "learning_rate": 5.8354112757782485e-05, "loss": 2.5687, "step": 53111 }, { "epoch": 4.16, "learning_rate": 5.835332862855799e-05, "loss": 2.6041, "step": 53112 }, { "epoch": 4.16, "learning_rate": 5.835254449933349e-05, "loss": 2.7216, "step": 53113 }, { "epoch": 4.16, "learning_rate": 5.8351760370108996e-05, "loss": 2.7696, "step": 53114 }, { "epoch": 4.16, "learning_rate": 5.83509762408845e-05, "loss": 2.5818, "step": 53115 }, { "epoch": 4.16, "learning_rate": 5.8350192111660004e-05, "loss": 2.6665, "step": 53116 }, { "epoch": 4.17, "learning_rate": 5.8349407982435514e-05, "loss": 2.7004, "step": 53117 }, { "epoch": 4.17, "learning_rate": 5.834862385321102e-05, "loss": 2.7081, "step": 53118 }, { "epoch": 4.17, "learning_rate": 5.834783972398652e-05, "loss": 2.4897, "step": 53119 }, { "epoch": 4.17, "learning_rate": 5.8347055594762025e-05, "loss": 2.7038, "step": 53120 }, { "epoch": 4.17, "learning_rate": 5.834627146553753e-05, "loss": 2.6212, "step": 53121 }, { "epoch": 4.17, "learning_rate": 5.834548733631303e-05, "loss": 2.8304, "step": 53122 }, { "epoch": 4.17, "learning_rate": 5.8344703207088536e-05, "loss": 2.6316, "step": 53123 }, { "epoch": 4.17, "learning_rate": 5.834391907786404e-05, "loss": 2.6081, "step": 53124 }, { "epoch": 4.17, "learning_rate": 5.8343134948639544e-05, "loss": 2.6325, "step": 53125 }, { "epoch": 4.17, "learning_rate": 5.8342350819415034e-05, "loss": 2.5068, "step": 53126 }, { "epoch": 4.17, "learning_rate": 5.8341566690190544e-05, "loss": 2.6631, "step": 53127 }, { "epoch": 4.17, "learning_rate": 5.834078256096605e-05, "loss": 2.6507, "step": 53128 }, { "epoch": 4.17, "learning_rate": 5.833999843174155e-05, "loss": 2.5998, "step": 53129 }, { "epoch": 4.17, "learning_rate": 5.8339214302517055e-05, "loss": 2.5874, "step": 53130 }, { "epoch": 4.17, "learning_rate": 5.833843017329256e-05, "loss": 2.6175, "step": 53131 }, { "epoch": 4.17, "learning_rate": 5.833764604406806e-05, "loss": 2.8856, "step": 53132 }, { "epoch": 4.17, "learning_rate": 5.8336861914843566e-05, "loss": 2.6915, "step": 53133 }, { "epoch": 4.17, "learning_rate": 5.833607778561907e-05, "loss": 2.6331, "step": 53134 }, { "epoch": 4.17, "learning_rate": 5.8335293656394574e-05, "loss": 2.5898, "step": 53135 }, { "epoch": 4.17, "learning_rate": 5.833450952717008e-05, "loss": 2.5431, "step": 53136 }, { "epoch": 4.17, "learning_rate": 5.833372539794558e-05, "loss": 2.416, "step": 53137 }, { "epoch": 4.17, "learning_rate": 5.8332941268721085e-05, "loss": 2.6382, "step": 53138 }, { "epoch": 4.17, "learning_rate": 5.833215713949659e-05, "loss": 2.6106, "step": 53139 }, { "epoch": 4.17, "learning_rate": 5.833137301027209e-05, "loss": 2.5815, "step": 53140 }, { "epoch": 4.17, "learning_rate": 5.8330588881047596e-05, "loss": 2.7546, "step": 53141 }, { "epoch": 4.17, "learning_rate": 5.8329804751823106e-05, "loss": 2.7963, "step": 53142 }, { "epoch": 4.17, "learning_rate": 5.832902062259861e-05, "loss": 2.7231, "step": 53143 }, { "epoch": 4.17, "learning_rate": 5.8328236493374113e-05, "loss": 2.5004, "step": 53144 }, { "epoch": 4.17, "learning_rate": 5.832745236414962e-05, "loss": 2.5754, "step": 53145 }, { "epoch": 4.17, "learning_rate": 5.832666823492512e-05, "loss": 2.4904, "step": 53146 }, { "epoch": 4.17, "learning_rate": 5.8325884105700624e-05, "loss": 2.6919, "step": 53147 }, { "epoch": 4.17, "learning_rate": 5.832509997647613e-05, "loss": 2.8544, "step": 53148 }, { "epoch": 4.17, "learning_rate": 5.832431584725163e-05, "loss": 2.8282, "step": 53149 }, { "epoch": 4.17, "learning_rate": 5.8323531718027135e-05, "loss": 2.7261, "step": 53150 }, { "epoch": 4.17, "learning_rate": 5.832274758880264e-05, "loss": 2.5583, "step": 53151 }, { "epoch": 4.17, "learning_rate": 5.832196345957814e-05, "loss": 2.9398, "step": 53152 }, { "epoch": 4.17, "learning_rate": 5.8321179330353647e-05, "loss": 2.5806, "step": 53153 }, { "epoch": 4.17, "learning_rate": 5.832039520112915e-05, "loss": 2.8618, "step": 53154 }, { "epoch": 4.17, "learning_rate": 5.8319611071904654e-05, "loss": 2.7075, "step": 53155 }, { "epoch": 4.17, "learning_rate": 5.831882694268016e-05, "loss": 2.7361, "step": 53156 }, { "epoch": 4.17, "learning_rate": 5.831804281345566e-05, "loss": 2.6026, "step": 53157 }, { "epoch": 4.17, "learning_rate": 5.831725868423117e-05, "loss": 2.4747, "step": 53158 }, { "epoch": 4.17, "learning_rate": 5.8316474555006675e-05, "loss": 2.5377, "step": 53159 }, { "epoch": 4.17, "learning_rate": 5.831569042578218e-05, "loss": 2.6738, "step": 53160 }, { "epoch": 4.17, "learning_rate": 5.831490629655768e-05, "loss": 2.4332, "step": 53161 }, { "epoch": 4.17, "learning_rate": 5.831412216733317e-05, "loss": 2.659, "step": 53162 }, { "epoch": 4.17, "learning_rate": 5.8313338038108677e-05, "loss": 2.5501, "step": 53163 }, { "epoch": 4.17, "learning_rate": 5.831255390888418e-05, "loss": 2.6401, "step": 53164 }, { "epoch": 4.17, "learning_rate": 5.8311769779659684e-05, "loss": 2.6108, "step": 53165 }, { "epoch": 4.17, "learning_rate": 5.831098565043519e-05, "loss": 2.8654, "step": 53166 }, { "epoch": 4.17, "learning_rate": 5.831020152121069e-05, "loss": 2.6355, "step": 53167 }, { "epoch": 4.17, "learning_rate": 5.83094173919862e-05, "loss": 2.4988, "step": 53168 }, { "epoch": 4.17, "learning_rate": 5.8308633262761705e-05, "loss": 2.5885, "step": 53169 }, { "epoch": 4.17, "learning_rate": 5.830784913353721e-05, "loss": 2.6688, "step": 53170 }, { "epoch": 4.17, "learning_rate": 5.830706500431271e-05, "loss": 2.7229, "step": 53171 }, { "epoch": 4.17, "learning_rate": 5.8306280875088216e-05, "loss": 2.5844, "step": 53172 }, { "epoch": 4.17, "learning_rate": 5.830549674586372e-05, "loss": 2.4429, "step": 53173 }, { "epoch": 4.17, "learning_rate": 5.8304712616639224e-05, "loss": 2.6166, "step": 53174 }, { "epoch": 4.17, "learning_rate": 5.830392848741473e-05, "loss": 2.5788, "step": 53175 }, { "epoch": 4.17, "learning_rate": 5.830314435819023e-05, "loss": 2.486, "step": 53176 }, { "epoch": 4.17, "learning_rate": 5.8302360228965735e-05, "loss": 2.6947, "step": 53177 }, { "epoch": 4.17, "learning_rate": 5.830157609974124e-05, "loss": 2.646, "step": 53178 }, { "epoch": 4.17, "learning_rate": 5.830079197051674e-05, "loss": 2.4842, "step": 53179 }, { "epoch": 4.17, "learning_rate": 5.8300007841292246e-05, "loss": 2.5481, "step": 53180 }, { "epoch": 4.17, "learning_rate": 5.829922371206775e-05, "loss": 2.42, "step": 53181 }, { "epoch": 4.17, "learning_rate": 5.829843958284325e-05, "loss": 2.6739, "step": 53182 }, { "epoch": 4.17, "learning_rate": 5.8297655453618764e-05, "loss": 2.8459, "step": 53183 }, { "epoch": 4.17, "learning_rate": 5.829687132439427e-05, "loss": 2.5441, "step": 53184 }, { "epoch": 4.17, "learning_rate": 5.829608719516977e-05, "loss": 2.667, "step": 53185 }, { "epoch": 4.17, "learning_rate": 5.8295303065945275e-05, "loss": 2.8794, "step": 53186 }, { "epoch": 4.17, "learning_rate": 5.829451893672078e-05, "loss": 2.5185, "step": 53187 }, { "epoch": 4.17, "learning_rate": 5.829373480749628e-05, "loss": 2.7591, "step": 53188 }, { "epoch": 4.17, "learning_rate": 5.8292950678271786e-05, "loss": 2.6738, "step": 53189 }, { "epoch": 4.17, "learning_rate": 5.829216654904729e-05, "loss": 2.4816, "step": 53190 }, { "epoch": 4.17, "learning_rate": 5.829138241982279e-05, "loss": 2.6836, "step": 53191 }, { "epoch": 4.17, "learning_rate": 5.82905982905983e-05, "loss": 2.5746, "step": 53192 }, { "epoch": 4.17, "learning_rate": 5.82898141613738e-05, "loss": 2.7888, "step": 53193 }, { "epoch": 4.17, "learning_rate": 5.8289030032149304e-05, "loss": 2.5875, "step": 53194 }, { "epoch": 4.17, "learning_rate": 5.828824590292481e-05, "loss": 2.6704, "step": 53195 }, { "epoch": 4.17, "learning_rate": 5.828746177370031e-05, "loss": 2.6571, "step": 53196 }, { "epoch": 4.17, "learning_rate": 5.8286677644475815e-05, "loss": 2.877, "step": 53197 }, { "epoch": 4.17, "learning_rate": 5.828589351525131e-05, "loss": 2.7908, "step": 53198 }, { "epoch": 4.17, "learning_rate": 5.8285109386026816e-05, "loss": 2.6849, "step": 53199 }, { "epoch": 4.17, "learning_rate": 5.828432525680232e-05, "loss": 2.6953, "step": 53200 }, { "epoch": 4.17, "learning_rate": 5.828354112757782e-05, "loss": 2.4605, "step": 53201 }, { "epoch": 4.17, "learning_rate": 5.828275699835333e-05, "loss": 2.6406, "step": 53202 }, { "epoch": 4.17, "learning_rate": 5.828197286912883e-05, "loss": 2.509, "step": 53203 }, { "epoch": 4.17, "learning_rate": 5.8281188739904334e-05, "loss": 2.73, "step": 53204 }, { "epoch": 4.17, "learning_rate": 5.828040461067984e-05, "loss": 2.8482, "step": 53205 }, { "epoch": 4.17, "learning_rate": 5.827962048145534e-05, "loss": 2.6339, "step": 53206 }, { "epoch": 4.17, "learning_rate": 5.8278836352230845e-05, "loss": 2.5776, "step": 53207 }, { "epoch": 4.17, "learning_rate": 5.827805222300635e-05, "loss": 2.6209, "step": 53208 }, { "epoch": 4.17, "learning_rate": 5.827726809378186e-05, "loss": 2.6485, "step": 53209 }, { "epoch": 4.17, "learning_rate": 5.827648396455736e-05, "loss": 2.8382, "step": 53210 }, { "epoch": 4.17, "learning_rate": 5.8275699835332867e-05, "loss": 2.5294, "step": 53211 }, { "epoch": 4.17, "learning_rate": 5.827491570610837e-05, "loss": 2.584, "step": 53212 }, { "epoch": 4.17, "learning_rate": 5.8274131576883874e-05, "loss": 2.5344, "step": 53213 }, { "epoch": 4.17, "learning_rate": 5.827334744765938e-05, "loss": 2.9185, "step": 53214 }, { "epoch": 4.17, "learning_rate": 5.827256331843488e-05, "loss": 2.6718, "step": 53215 }, { "epoch": 4.17, "learning_rate": 5.8271779189210385e-05, "loss": 2.5041, "step": 53216 }, { "epoch": 4.17, "learning_rate": 5.827099505998589e-05, "loss": 2.6978, "step": 53217 }, { "epoch": 4.17, "learning_rate": 5.827021093076139e-05, "loss": 2.94, "step": 53218 }, { "epoch": 4.17, "learning_rate": 5.8269426801536896e-05, "loss": 2.6878, "step": 53219 }, { "epoch": 4.17, "learning_rate": 5.82686426723124e-05, "loss": 2.8792, "step": 53220 }, { "epoch": 4.17, "learning_rate": 5.82678585430879e-05, "loss": 2.6298, "step": 53221 }, { "epoch": 4.17, "learning_rate": 5.826707441386341e-05, "loss": 2.5381, "step": 53222 }, { "epoch": 4.17, "learning_rate": 5.826629028463891e-05, "loss": 2.6287, "step": 53223 }, { "epoch": 4.17, "learning_rate": 5.8265506155414414e-05, "loss": 2.8543, "step": 53224 }, { "epoch": 4.17, "learning_rate": 5.8264722026189925e-05, "loss": 2.7201, "step": 53225 }, { "epoch": 4.17, "learning_rate": 5.826393789696543e-05, "loss": 2.5927, "step": 53226 }, { "epoch": 4.17, "learning_rate": 5.826315376774093e-05, "loss": 2.7342, "step": 53227 }, { "epoch": 4.17, "learning_rate": 5.8262369638516436e-05, "loss": 2.5057, "step": 53228 }, { "epoch": 4.17, "learning_rate": 5.826158550929194e-05, "loss": 2.5786, "step": 53229 }, { "epoch": 4.17, "learning_rate": 5.826080138006744e-05, "loss": 2.8029, "step": 53230 }, { "epoch": 4.17, "learning_rate": 5.826001725084295e-05, "loss": 2.4611, "step": 53231 }, { "epoch": 4.17, "learning_rate": 5.825923312161845e-05, "loss": 2.5027, "step": 53232 }, { "epoch": 4.17, "learning_rate": 5.8258448992393954e-05, "loss": 2.6562, "step": 53233 }, { "epoch": 4.17, "learning_rate": 5.8257664863169444e-05, "loss": 2.6046, "step": 53234 }, { "epoch": 4.17, "learning_rate": 5.8256880733944955e-05, "loss": 2.4735, "step": 53235 }, { "epoch": 4.17, "learning_rate": 5.825609660472046e-05, "loss": 2.573, "step": 53236 }, { "epoch": 4.17, "learning_rate": 5.825531247549596e-05, "loss": 2.3782, "step": 53237 }, { "epoch": 4.17, "learning_rate": 5.8254528346271466e-05, "loss": 2.6291, "step": 53238 }, { "epoch": 4.17, "learning_rate": 5.825374421704697e-05, "loss": 2.7745, "step": 53239 }, { "epoch": 4.17, "learning_rate": 5.825296008782247e-05, "loss": 2.7459, "step": 53240 }, { "epoch": 4.17, "learning_rate": 5.825217595859798e-05, "loss": 2.4622, "step": 53241 }, { "epoch": 4.17, "learning_rate": 5.825139182937348e-05, "loss": 2.6673, "step": 53242 }, { "epoch": 4.17, "learning_rate": 5.8250607700148984e-05, "loss": 2.5328, "step": 53243 }, { "epoch": 4.18, "learning_rate": 5.824982357092449e-05, "loss": 2.5317, "step": 53244 }, { "epoch": 4.18, "learning_rate": 5.824903944169999e-05, "loss": 2.6815, "step": 53245 }, { "epoch": 4.18, "learning_rate": 5.8248255312475495e-05, "loss": 2.6562, "step": 53246 }, { "epoch": 4.18, "learning_rate": 5.8247471183251e-05, "loss": 2.8491, "step": 53247 }, { "epoch": 4.18, "learning_rate": 5.82466870540265e-05, "loss": 2.6822, "step": 53248 }, { "epoch": 4.18, "learning_rate": 5.8245902924802006e-05, "loss": 2.5487, "step": 53249 }, { "epoch": 4.18, "learning_rate": 5.824511879557752e-05, "loss": 2.4309, "step": 53250 }, { "epoch": 4.18, "learning_rate": 5.824433466635302e-05, "loss": 2.7461, "step": 53251 }, { "epoch": 4.18, "learning_rate": 5.8243550537128524e-05, "loss": 2.8465, "step": 53252 }, { "epoch": 4.18, "learning_rate": 5.824276640790403e-05, "loss": 2.6174, "step": 53253 }, { "epoch": 4.18, "learning_rate": 5.824198227867953e-05, "loss": 2.8073, "step": 53254 }, { "epoch": 4.18, "learning_rate": 5.8241198149455035e-05, "loss": 2.6211, "step": 53255 }, { "epoch": 4.18, "learning_rate": 5.824041402023054e-05, "loss": 2.6641, "step": 53256 }, { "epoch": 4.18, "learning_rate": 5.823962989100604e-05, "loss": 2.6031, "step": 53257 }, { "epoch": 4.18, "learning_rate": 5.8238845761781546e-05, "loss": 2.7984, "step": 53258 }, { "epoch": 4.18, "learning_rate": 5.823806163255705e-05, "loss": 2.5687, "step": 53259 }, { "epoch": 4.18, "learning_rate": 5.8237277503332554e-05, "loss": 2.7131, "step": 53260 }, { "epoch": 4.18, "learning_rate": 5.823649337410806e-05, "loss": 2.6374, "step": 53261 }, { "epoch": 4.18, "learning_rate": 5.823570924488356e-05, "loss": 2.8213, "step": 53262 }, { "epoch": 4.18, "learning_rate": 5.8234925115659065e-05, "loss": 2.7431, "step": 53263 }, { "epoch": 4.18, "learning_rate": 5.823414098643457e-05, "loss": 2.7814, "step": 53264 }, { "epoch": 4.18, "learning_rate": 5.823335685721007e-05, "loss": 2.6195, "step": 53265 }, { "epoch": 4.18, "learning_rate": 5.823257272798558e-05, "loss": 2.708, "step": 53266 }, { "epoch": 4.18, "learning_rate": 5.8231788598761086e-05, "loss": 2.5503, "step": 53267 }, { "epoch": 4.18, "learning_rate": 5.823100446953659e-05, "loss": 2.7874, "step": 53268 }, { "epoch": 4.18, "learning_rate": 5.8230220340312093e-05, "loss": 2.538, "step": 53269 }, { "epoch": 4.18, "learning_rate": 5.8229436211087584e-05, "loss": 2.6889, "step": 53270 }, { "epoch": 4.18, "learning_rate": 5.822865208186309e-05, "loss": 2.5713, "step": 53271 }, { "epoch": 4.18, "learning_rate": 5.822786795263859e-05, "loss": 2.6737, "step": 53272 }, { "epoch": 4.18, "learning_rate": 5.8227083823414095e-05, "loss": 2.5562, "step": 53273 }, { "epoch": 4.18, "learning_rate": 5.82262996941896e-05, "loss": 2.6064, "step": 53274 }, { "epoch": 4.18, "learning_rate": 5.82255155649651e-05, "loss": 2.69, "step": 53275 }, { "epoch": 4.18, "learning_rate": 5.822473143574061e-05, "loss": 2.8993, "step": 53276 }, { "epoch": 4.18, "learning_rate": 5.8223947306516116e-05, "loss": 2.8256, "step": 53277 }, { "epoch": 4.18, "learning_rate": 5.822316317729162e-05, "loss": 2.6075, "step": 53278 }, { "epoch": 4.18, "learning_rate": 5.8222379048067123e-05, "loss": 2.4842, "step": 53279 }, { "epoch": 4.18, "learning_rate": 5.822159491884263e-05, "loss": 2.7146, "step": 53280 }, { "epoch": 4.18, "learning_rate": 5.822081078961813e-05, "loss": 2.5452, "step": 53281 }, { "epoch": 4.18, "learning_rate": 5.8220026660393634e-05, "loss": 2.6185, "step": 53282 }, { "epoch": 4.18, "learning_rate": 5.821924253116914e-05, "loss": 2.816, "step": 53283 }, { "epoch": 4.18, "learning_rate": 5.821845840194464e-05, "loss": 2.6697, "step": 53284 }, { "epoch": 4.18, "learning_rate": 5.8217674272720146e-05, "loss": 2.7963, "step": 53285 }, { "epoch": 4.18, "learning_rate": 5.821689014349565e-05, "loss": 2.5538, "step": 53286 }, { "epoch": 4.18, "learning_rate": 5.821610601427115e-05, "loss": 2.8253, "step": 53287 }, { "epoch": 4.18, "learning_rate": 5.8215321885046657e-05, "loss": 2.6597, "step": 53288 }, { "epoch": 4.18, "learning_rate": 5.821453775582216e-05, "loss": 2.5587, "step": 53289 }, { "epoch": 4.18, "learning_rate": 5.8213753626597664e-05, "loss": 2.6924, "step": 53290 }, { "epoch": 4.18, "learning_rate": 5.821296949737317e-05, "loss": 2.644, "step": 53291 }, { "epoch": 4.18, "learning_rate": 5.821218536814868e-05, "loss": 2.5612, "step": 53292 }, { "epoch": 4.18, "learning_rate": 5.821140123892418e-05, "loss": 2.6748, "step": 53293 }, { "epoch": 4.18, "learning_rate": 5.8210617109699685e-05, "loss": 2.672, "step": 53294 }, { "epoch": 4.18, "learning_rate": 5.820983298047519e-05, "loss": 2.7751, "step": 53295 }, { "epoch": 4.18, "learning_rate": 5.820904885125069e-05, "loss": 2.6478, "step": 53296 }, { "epoch": 4.18, "learning_rate": 5.8208264722026196e-05, "loss": 2.7379, "step": 53297 }, { "epoch": 4.18, "learning_rate": 5.82074805928017e-05, "loss": 2.4988, "step": 53298 }, { "epoch": 4.18, "learning_rate": 5.8206696463577204e-05, "loss": 2.5698, "step": 53299 }, { "epoch": 4.18, "learning_rate": 5.820591233435271e-05, "loss": 2.5572, "step": 53300 }, { "epoch": 4.18, "learning_rate": 5.820512820512821e-05, "loss": 2.5326, "step": 53301 }, { "epoch": 4.18, "learning_rate": 5.8204344075903715e-05, "loss": 2.7432, "step": 53302 }, { "epoch": 4.18, "learning_rate": 5.820355994667922e-05, "loss": 2.5809, "step": 53303 }, { "epoch": 4.18, "learning_rate": 5.820277581745472e-05, "loss": 2.7202, "step": 53304 }, { "epoch": 4.18, "learning_rate": 5.8201991688230226e-05, "loss": 2.6919, "step": 53305 }, { "epoch": 4.18, "learning_rate": 5.820120755900572e-05, "loss": 2.6146, "step": 53306 }, { "epoch": 4.18, "learning_rate": 5.8200423429781226e-05, "loss": 2.667, "step": 53307 }, { "epoch": 4.18, "learning_rate": 5.819963930055673e-05, "loss": 2.7765, "step": 53308 }, { "epoch": 4.18, "learning_rate": 5.8198855171332234e-05, "loss": 2.5916, "step": 53309 }, { "epoch": 4.18, "learning_rate": 5.819807104210774e-05, "loss": 2.7905, "step": 53310 }, { "epoch": 4.18, "learning_rate": 5.819728691288324e-05, "loss": 2.5585, "step": 53311 }, { "epoch": 4.18, "learning_rate": 5.8196502783658745e-05, "loss": 2.589, "step": 53312 }, { "epoch": 4.18, "learning_rate": 5.819571865443425e-05, "loss": 2.9042, "step": 53313 }, { "epoch": 4.18, "learning_rate": 5.819493452520975e-05, "loss": 2.5199, "step": 53314 }, { "epoch": 4.18, "learning_rate": 5.8194150395985256e-05, "loss": 2.6725, "step": 53315 }, { "epoch": 4.18, "learning_rate": 5.819336626676076e-05, "loss": 2.6218, "step": 53316 }, { "epoch": 4.18, "learning_rate": 5.819258213753627e-05, "loss": 2.6804, "step": 53317 }, { "epoch": 4.18, "learning_rate": 5.8191798008311774e-05, "loss": 2.7863, "step": 53318 }, { "epoch": 4.18, "learning_rate": 5.819101387908728e-05, "loss": 2.6569, "step": 53319 }, { "epoch": 4.18, "learning_rate": 5.819022974986278e-05, "loss": 2.8032, "step": 53320 }, { "epoch": 4.18, "learning_rate": 5.8189445620638285e-05, "loss": 2.6627, "step": 53321 }, { "epoch": 4.18, "learning_rate": 5.818866149141379e-05, "loss": 2.8151, "step": 53322 }, { "epoch": 4.18, "learning_rate": 5.818787736218929e-05, "loss": 2.8067, "step": 53323 }, { "epoch": 4.18, "learning_rate": 5.8187093232964796e-05, "loss": 2.9116, "step": 53324 }, { "epoch": 4.18, "learning_rate": 5.81863091037403e-05, "loss": 2.6411, "step": 53325 }, { "epoch": 4.18, "learning_rate": 5.81855249745158e-05, "loss": 2.8291, "step": 53326 }, { "epoch": 4.18, "learning_rate": 5.818474084529131e-05, "loss": 2.7535, "step": 53327 }, { "epoch": 4.18, "learning_rate": 5.818395671606681e-05, "loss": 2.7837, "step": 53328 }, { "epoch": 4.18, "learning_rate": 5.8183172586842314e-05, "loss": 2.6865, "step": 53329 }, { "epoch": 4.18, "learning_rate": 5.818238845761782e-05, "loss": 2.5824, "step": 53330 }, { "epoch": 4.18, "learning_rate": 5.818160432839332e-05, "loss": 2.5598, "step": 53331 }, { "epoch": 4.18, "learning_rate": 5.8180820199168825e-05, "loss": 2.5697, "step": 53332 }, { "epoch": 4.18, "learning_rate": 5.8180036069944336e-05, "loss": 2.6847, "step": 53333 }, { "epoch": 4.18, "learning_rate": 5.817925194071984e-05, "loss": 2.6651, "step": 53334 }, { "epoch": 4.18, "learning_rate": 5.817846781149534e-05, "loss": 2.5826, "step": 53335 }, { "epoch": 4.18, "learning_rate": 5.817768368227085e-05, "loss": 2.5975, "step": 53336 }, { "epoch": 4.18, "learning_rate": 5.817689955304635e-05, "loss": 2.6632, "step": 53337 }, { "epoch": 4.18, "learning_rate": 5.8176115423821854e-05, "loss": 2.9654, "step": 53338 }, { "epoch": 4.18, "learning_rate": 5.817533129459736e-05, "loss": 2.5444, "step": 53339 }, { "epoch": 4.18, "learning_rate": 5.817454716537286e-05, "loss": 2.6743, "step": 53340 }, { "epoch": 4.18, "learning_rate": 5.8173763036148365e-05, "loss": 2.7961, "step": 53341 }, { "epoch": 4.18, "learning_rate": 5.8172978906923855e-05, "loss": 2.6911, "step": 53342 }, { "epoch": 4.18, "learning_rate": 5.8172194777699366e-05, "loss": 2.8323, "step": 53343 }, { "epoch": 4.18, "learning_rate": 5.817141064847487e-05, "loss": 2.6991, "step": 53344 }, { "epoch": 4.18, "learning_rate": 5.817062651925037e-05, "loss": 2.6604, "step": 53345 }, { "epoch": 4.18, "learning_rate": 5.816984239002588e-05, "loss": 2.7341, "step": 53346 }, { "epoch": 4.18, "learning_rate": 5.816905826080138e-05, "loss": 2.6427, "step": 53347 }, { "epoch": 4.18, "learning_rate": 5.8168274131576884e-05, "loss": 2.5305, "step": 53348 }, { "epoch": 4.18, "learning_rate": 5.816749000235239e-05, "loss": 2.5765, "step": 53349 }, { "epoch": 4.18, "learning_rate": 5.816670587312789e-05, "loss": 2.5887, "step": 53350 }, { "epoch": 4.18, "learning_rate": 5.8165921743903395e-05, "loss": 2.6681, "step": 53351 }, { "epoch": 4.18, "learning_rate": 5.81651376146789e-05, "loss": 2.6681, "step": 53352 }, { "epoch": 4.18, "learning_rate": 5.81643534854544e-05, "loss": 2.5336, "step": 53353 }, { "epoch": 4.18, "learning_rate": 5.8163569356229906e-05, "loss": 2.6015, "step": 53354 }, { "epoch": 4.18, "learning_rate": 5.816278522700541e-05, "loss": 2.5408, "step": 53355 }, { "epoch": 4.18, "learning_rate": 5.8162001097780913e-05, "loss": 2.7485, "step": 53356 }, { "epoch": 4.18, "learning_rate": 5.816121696855642e-05, "loss": 2.7379, "step": 53357 }, { "epoch": 4.18, "learning_rate": 5.816043283933193e-05, "loss": 2.765, "step": 53358 }, { "epoch": 4.18, "learning_rate": 5.815964871010743e-05, "loss": 2.5718, "step": 53359 }, { "epoch": 4.18, "learning_rate": 5.8158864580882935e-05, "loss": 2.8263, "step": 53360 }, { "epoch": 4.18, "learning_rate": 5.815808045165844e-05, "loss": 2.5992, "step": 53361 }, { "epoch": 4.18, "learning_rate": 5.815729632243394e-05, "loss": 2.5471, "step": 53362 }, { "epoch": 4.18, "learning_rate": 5.8156512193209446e-05, "loss": 2.6078, "step": 53363 }, { "epoch": 4.18, "learning_rate": 5.815572806398495e-05, "loss": 2.5358, "step": 53364 }, { "epoch": 4.18, "learning_rate": 5.815494393476045e-05, "loss": 2.7702, "step": 53365 }, { "epoch": 4.18, "learning_rate": 5.815415980553596e-05, "loss": 2.8817, "step": 53366 }, { "epoch": 4.18, "learning_rate": 5.815337567631146e-05, "loss": 2.8558, "step": 53367 }, { "epoch": 4.18, "learning_rate": 5.8152591547086964e-05, "loss": 2.9246, "step": 53368 }, { "epoch": 4.18, "learning_rate": 5.815180741786247e-05, "loss": 2.5202, "step": 53369 }, { "epoch": 4.18, "learning_rate": 5.815102328863797e-05, "loss": 2.5446, "step": 53370 }, { "epoch": 4.18, "learning_rate": 5.8150239159413475e-05, "loss": 2.9244, "step": 53371 }, { "epoch": 4.19, "learning_rate": 5.814945503018898e-05, "loss": 2.813, "step": 53372 }, { "epoch": 4.19, "learning_rate": 5.814867090096448e-05, "loss": 2.5463, "step": 53373 }, { "epoch": 4.19, "learning_rate": 5.814788677173999e-05, "loss": 2.6504, "step": 53374 }, { "epoch": 4.19, "learning_rate": 5.81471026425155e-05, "loss": 2.6972, "step": 53375 }, { "epoch": 4.19, "learning_rate": 5.8146318513291e-05, "loss": 2.5537, "step": 53376 }, { "epoch": 4.19, "learning_rate": 5.8145534384066504e-05, "loss": 2.6665, "step": 53377 }, { "epoch": 4.19, "learning_rate": 5.8144750254841994e-05, "loss": 2.6561, "step": 53378 }, { "epoch": 4.19, "learning_rate": 5.81439661256175e-05, "loss": 2.588, "step": 53379 }, { "epoch": 4.19, "learning_rate": 5.8143181996393e-05, "loss": 2.7265, "step": 53380 }, { "epoch": 4.19, "learning_rate": 5.8142397867168505e-05, "loss": 2.6932, "step": 53381 }, { "epoch": 4.19, "learning_rate": 5.814161373794401e-05, "loss": 2.622, "step": 53382 }, { "epoch": 4.19, "learning_rate": 5.814082960871951e-05, "loss": 2.7076, "step": 53383 }, { "epoch": 4.19, "learning_rate": 5.814004547949502e-05, "loss": 2.7829, "step": 53384 }, { "epoch": 4.19, "learning_rate": 5.813926135027053e-05, "loss": 2.8095, "step": 53385 }, { "epoch": 4.19, "learning_rate": 5.813847722104603e-05, "loss": 2.6422, "step": 53386 }, { "epoch": 4.19, "learning_rate": 5.8137693091821534e-05, "loss": 2.7642, "step": 53387 }, { "epoch": 4.19, "learning_rate": 5.813690896259704e-05, "loss": 2.7023, "step": 53388 }, { "epoch": 4.19, "learning_rate": 5.813612483337254e-05, "loss": 2.7785, "step": 53389 }, { "epoch": 4.19, "learning_rate": 5.8135340704148045e-05, "loss": 2.7015, "step": 53390 }, { "epoch": 4.19, "learning_rate": 5.813455657492355e-05, "loss": 2.7018, "step": 53391 }, { "epoch": 4.19, "learning_rate": 5.813377244569905e-05, "loss": 2.6212, "step": 53392 }, { "epoch": 4.19, "learning_rate": 5.8132988316474556e-05, "loss": 2.7634, "step": 53393 }, { "epoch": 4.19, "learning_rate": 5.813220418725006e-05, "loss": 2.6267, "step": 53394 }, { "epoch": 4.19, "learning_rate": 5.8131420058025564e-05, "loss": 2.7614, "step": 53395 }, { "epoch": 4.19, "learning_rate": 5.813063592880107e-05, "loss": 2.7615, "step": 53396 }, { "epoch": 4.19, "learning_rate": 5.812985179957657e-05, "loss": 2.6488, "step": 53397 }, { "epoch": 4.19, "learning_rate": 5.8129067670352075e-05, "loss": 2.7457, "step": 53398 }, { "epoch": 4.19, "learning_rate": 5.812828354112758e-05, "loss": 2.7795, "step": 53399 }, { "epoch": 4.19, "learning_rate": 5.812749941190309e-05, "loss": 2.6046, "step": 53400 }, { "epoch": 4.19, "learning_rate": 5.812671528267859e-05, "loss": 2.951, "step": 53401 }, { "epoch": 4.19, "learning_rate": 5.8125931153454096e-05, "loss": 2.6494, "step": 53402 }, { "epoch": 4.19, "learning_rate": 5.81251470242296e-05, "loss": 2.5944, "step": 53403 }, { "epoch": 4.19, "learning_rate": 5.8124362895005104e-05, "loss": 2.5768, "step": 53404 }, { "epoch": 4.19, "learning_rate": 5.812357876578061e-05, "loss": 2.519, "step": 53405 }, { "epoch": 4.19, "learning_rate": 5.812279463655611e-05, "loss": 2.6954, "step": 53406 }, { "epoch": 4.19, "learning_rate": 5.8122010507331615e-05, "loss": 2.6071, "step": 53407 }, { "epoch": 4.19, "learning_rate": 5.812122637810712e-05, "loss": 2.5925, "step": 53408 }, { "epoch": 4.19, "learning_rate": 5.812044224888262e-05, "loss": 2.6127, "step": 53409 }, { "epoch": 4.19, "learning_rate": 5.8119658119658126e-05, "loss": 2.5648, "step": 53410 }, { "epoch": 4.19, "learning_rate": 5.811887399043363e-05, "loss": 2.7837, "step": 53411 }, { "epoch": 4.19, "learning_rate": 5.811808986120913e-05, "loss": 2.9501, "step": 53412 }, { "epoch": 4.19, "learning_rate": 5.8117305731984637e-05, "loss": 2.7519, "step": 53413 }, { "epoch": 4.19, "learning_rate": 5.8116521602760134e-05, "loss": 2.674, "step": 53414 }, { "epoch": 4.19, "learning_rate": 5.811573747353564e-05, "loss": 2.4679, "step": 53415 }, { "epoch": 4.19, "learning_rate": 5.811495334431114e-05, "loss": 2.6302, "step": 53416 }, { "epoch": 4.19, "learning_rate": 5.8114169215086645e-05, "loss": 2.6895, "step": 53417 }, { "epoch": 4.19, "learning_rate": 5.811338508586215e-05, "loss": 2.6635, "step": 53418 }, { "epoch": 4.19, "learning_rate": 5.811260095663765e-05, "loss": 2.844, "step": 53419 }, { "epoch": 4.19, "learning_rate": 5.8111816827413156e-05, "loss": 2.6315, "step": 53420 }, { "epoch": 4.19, "learning_rate": 5.811103269818866e-05, "loss": 2.4182, "step": 53421 }, { "epoch": 4.19, "learning_rate": 5.811024856896416e-05, "loss": 2.7293, "step": 53422 }, { "epoch": 4.19, "learning_rate": 5.8109464439739667e-05, "loss": 2.5792, "step": 53423 }, { "epoch": 4.19, "learning_rate": 5.810868031051517e-05, "loss": 2.7092, "step": 53424 }, { "epoch": 4.19, "learning_rate": 5.810789618129068e-05, "loss": 2.6791, "step": 53425 }, { "epoch": 4.19, "learning_rate": 5.8107112052066184e-05, "loss": 2.7972, "step": 53426 }, { "epoch": 4.19, "learning_rate": 5.810632792284169e-05, "loss": 2.5235, "step": 53427 }, { "epoch": 4.19, "learning_rate": 5.810554379361719e-05, "loss": 2.5102, "step": 53428 }, { "epoch": 4.19, "learning_rate": 5.8104759664392695e-05, "loss": 2.5384, "step": 53429 }, { "epoch": 4.19, "learning_rate": 5.81039755351682e-05, "loss": 2.5725, "step": 53430 }, { "epoch": 4.19, "learning_rate": 5.81031914059437e-05, "loss": 2.5851, "step": 53431 }, { "epoch": 4.19, "learning_rate": 5.8102407276719206e-05, "loss": 2.663, "step": 53432 }, { "epoch": 4.19, "learning_rate": 5.810162314749471e-05, "loss": 2.5974, "step": 53433 }, { "epoch": 4.19, "learning_rate": 5.8100839018270214e-05, "loss": 2.6447, "step": 53434 }, { "epoch": 4.19, "learning_rate": 5.810005488904572e-05, "loss": 2.7879, "step": 53435 }, { "epoch": 4.19, "learning_rate": 5.809927075982122e-05, "loss": 2.7528, "step": 53436 }, { "epoch": 4.19, "learning_rate": 5.8098486630596725e-05, "loss": 2.8642, "step": 53437 }, { "epoch": 4.19, "learning_rate": 5.809770250137223e-05, "loss": 2.6398, "step": 53438 }, { "epoch": 4.19, "learning_rate": 5.809691837214773e-05, "loss": 2.5718, "step": 53439 }, { "epoch": 4.19, "learning_rate": 5.8096134242923236e-05, "loss": 2.5673, "step": 53440 }, { "epoch": 4.19, "learning_rate": 5.8095350113698746e-05, "loss": 2.6459, "step": 53441 }, { "epoch": 4.19, "learning_rate": 5.809456598447425e-05, "loss": 2.8338, "step": 53442 }, { "epoch": 4.19, "learning_rate": 5.8093781855249754e-05, "loss": 2.6016, "step": 53443 }, { "epoch": 4.19, "learning_rate": 5.809299772602526e-05, "loss": 2.5722, "step": 53444 }, { "epoch": 4.19, "learning_rate": 5.809221359680076e-05, "loss": 2.6733, "step": 53445 }, { "epoch": 4.19, "learning_rate": 5.8091429467576265e-05, "loss": 2.7314, "step": 53446 }, { "epoch": 4.19, "learning_rate": 5.809064533835177e-05, "loss": 2.5553, "step": 53447 }, { "epoch": 4.19, "learning_rate": 5.808986120912727e-05, "loss": 2.5525, "step": 53448 }, { "epoch": 4.19, "learning_rate": 5.8089077079902776e-05, "loss": 2.7277, "step": 53449 }, { "epoch": 4.19, "learning_rate": 5.8088292950678266e-05, "loss": 2.6943, "step": 53450 }, { "epoch": 4.19, "learning_rate": 5.8087508821453776e-05, "loss": 2.5878, "step": 53451 }, { "epoch": 4.19, "learning_rate": 5.808672469222928e-05, "loss": 2.6327, "step": 53452 }, { "epoch": 4.19, "learning_rate": 5.8085940563004784e-05, "loss": 2.5236, "step": 53453 }, { "epoch": 4.19, "learning_rate": 5.808515643378029e-05, "loss": 2.7048, "step": 53454 }, { "epoch": 4.19, "learning_rate": 5.808437230455579e-05, "loss": 2.6684, "step": 53455 }, { "epoch": 4.19, "learning_rate": 5.8083588175331295e-05, "loss": 2.5689, "step": 53456 }, { "epoch": 4.19, "learning_rate": 5.80828040461068e-05, "loss": 2.6483, "step": 53457 }, { "epoch": 4.19, "learning_rate": 5.80820199168823e-05, "loss": 2.9106, "step": 53458 }, { "epoch": 4.19, "learning_rate": 5.8081235787657806e-05, "loss": 2.7358, "step": 53459 }, { "epoch": 4.19, "learning_rate": 5.808045165843331e-05, "loss": 2.5474, "step": 53460 }, { "epoch": 4.19, "learning_rate": 5.807966752920881e-05, "loss": 2.6437, "step": 53461 }, { "epoch": 4.19, "learning_rate": 5.807888339998432e-05, "loss": 2.6096, "step": 53462 }, { "epoch": 4.19, "learning_rate": 5.807809927075982e-05, "loss": 2.6752, "step": 53463 }, { "epoch": 4.19, "learning_rate": 5.8077315141535324e-05, "loss": 2.518, "step": 53464 }, { "epoch": 4.19, "learning_rate": 5.807653101231083e-05, "loss": 2.7276, "step": 53465 }, { "epoch": 4.19, "learning_rate": 5.807574688308634e-05, "loss": 2.5464, "step": 53466 }, { "epoch": 4.19, "learning_rate": 5.807496275386184e-05, "loss": 2.6575, "step": 53467 }, { "epoch": 4.19, "learning_rate": 5.8074178624637346e-05, "loss": 2.5658, "step": 53468 }, { "epoch": 4.19, "learning_rate": 5.807339449541285e-05, "loss": 2.7992, "step": 53469 }, { "epoch": 4.19, "learning_rate": 5.807261036618835e-05, "loss": 2.6325, "step": 53470 }, { "epoch": 4.19, "learning_rate": 5.807182623696386e-05, "loss": 2.528, "step": 53471 }, { "epoch": 4.19, "learning_rate": 5.807104210773936e-05, "loss": 2.502, "step": 53472 }, { "epoch": 4.19, "learning_rate": 5.8070257978514864e-05, "loss": 2.6338, "step": 53473 }, { "epoch": 4.19, "learning_rate": 5.806947384929037e-05, "loss": 2.6128, "step": 53474 }, { "epoch": 4.19, "learning_rate": 5.806868972006587e-05, "loss": 2.6227, "step": 53475 }, { "epoch": 4.19, "learning_rate": 5.8067905590841375e-05, "loss": 2.5882, "step": 53476 }, { "epoch": 4.19, "learning_rate": 5.806712146161688e-05, "loss": 2.8737, "step": 53477 }, { "epoch": 4.19, "learning_rate": 5.806633733239238e-05, "loss": 2.6048, "step": 53478 }, { "epoch": 4.19, "learning_rate": 5.8065553203167886e-05, "loss": 2.5049, "step": 53479 }, { "epoch": 4.19, "learning_rate": 5.806476907394339e-05, "loss": 2.8455, "step": 53480 }, { "epoch": 4.19, "learning_rate": 5.8063984944718893e-05, "loss": 2.8842, "step": 53481 }, { "epoch": 4.19, "learning_rate": 5.8063200815494404e-05, "loss": 2.5752, "step": 53482 }, { "epoch": 4.19, "learning_rate": 5.806241668626991e-05, "loss": 2.6728, "step": 53483 }, { "epoch": 4.19, "learning_rate": 5.806163255704541e-05, "loss": 2.6278, "step": 53484 }, { "epoch": 4.19, "learning_rate": 5.8060848427820915e-05, "loss": 2.5059, "step": 53485 }, { "epoch": 4.19, "learning_rate": 5.8060064298596405e-05, "loss": 2.5469, "step": 53486 }, { "epoch": 4.19, "learning_rate": 5.805928016937191e-05, "loss": 2.8996, "step": 53487 }, { "epoch": 4.19, "learning_rate": 5.805849604014741e-05, "loss": 2.8537, "step": 53488 }, { "epoch": 4.19, "learning_rate": 5.8057711910922916e-05, "loss": 2.6907, "step": 53489 }, { "epoch": 4.19, "learning_rate": 5.805692778169842e-05, "loss": 2.7706, "step": 53490 }, { "epoch": 4.19, "learning_rate": 5.8056143652473923e-05, "loss": 2.5561, "step": 53491 }, { "epoch": 4.19, "learning_rate": 5.8055359523249434e-05, "loss": 2.5168, "step": 53492 }, { "epoch": 4.19, "learning_rate": 5.805457539402494e-05, "loss": 2.685, "step": 53493 }, { "epoch": 4.19, "learning_rate": 5.805379126480044e-05, "loss": 2.6324, "step": 53494 }, { "epoch": 4.19, "learning_rate": 5.8053007135575945e-05, "loss": 2.7364, "step": 53495 }, { "epoch": 4.19, "learning_rate": 5.805222300635145e-05, "loss": 2.4994, "step": 53496 }, { "epoch": 4.19, "learning_rate": 5.805143887712695e-05, "loss": 2.9409, "step": 53497 }, { "epoch": 4.19, "learning_rate": 5.8050654747902456e-05, "loss": 2.7976, "step": 53498 }, { "epoch": 4.2, "learning_rate": 5.804987061867796e-05, "loss": 2.7325, "step": 53499 }, { "epoch": 4.2, "learning_rate": 5.804908648945346e-05, "loss": 2.6937, "step": 53500 }, { "epoch": 4.2, "learning_rate": 5.804830236022897e-05, "loss": 2.7219, "step": 53501 }, { "epoch": 4.2, "learning_rate": 5.804751823100447e-05, "loss": 2.6824, "step": 53502 }, { "epoch": 4.2, "learning_rate": 5.8046734101779974e-05, "loss": 2.6418, "step": 53503 }, { "epoch": 4.2, "learning_rate": 5.804594997255548e-05, "loss": 2.5492, "step": 53504 }, { "epoch": 4.2, "learning_rate": 5.804516584333098e-05, "loss": 2.4515, "step": 53505 }, { "epoch": 4.2, "learning_rate": 5.8044381714106485e-05, "loss": 2.4969, "step": 53506 }, { "epoch": 4.2, "learning_rate": 5.804359758488199e-05, "loss": 2.6225, "step": 53507 }, { "epoch": 4.2, "learning_rate": 5.80428134556575e-05, "loss": 2.8354, "step": 53508 }, { "epoch": 4.2, "learning_rate": 5.8042029326433e-05, "loss": 2.7564, "step": 53509 }, { "epoch": 4.2, "learning_rate": 5.804124519720851e-05, "loss": 3.0116, "step": 53510 }, { "epoch": 4.2, "learning_rate": 5.804046106798401e-05, "loss": 2.755, "step": 53511 }, { "epoch": 4.2, "learning_rate": 5.8039676938759514e-05, "loss": 2.5781, "step": 53512 }, { "epoch": 4.2, "learning_rate": 5.803889280953502e-05, "loss": 2.5567, "step": 53513 }, { "epoch": 4.2, "learning_rate": 5.803810868031052e-05, "loss": 2.5808, "step": 53514 }, { "epoch": 4.2, "learning_rate": 5.8037324551086025e-05, "loss": 2.5069, "step": 53515 }, { "epoch": 4.2, "learning_rate": 5.803654042186153e-05, "loss": 2.7969, "step": 53516 }, { "epoch": 4.2, "learning_rate": 5.803575629263703e-05, "loss": 2.6221, "step": 53517 }, { "epoch": 4.2, "learning_rate": 5.8034972163412536e-05, "loss": 2.5406, "step": 53518 }, { "epoch": 4.2, "learning_rate": 5.803418803418804e-05, "loss": 2.7407, "step": 53519 }, { "epoch": 4.2, "learning_rate": 5.8033403904963544e-05, "loss": 2.5404, "step": 53520 }, { "epoch": 4.2, "learning_rate": 5.803261977573905e-05, "loss": 2.6426, "step": 53521 }, { "epoch": 4.2, "learning_rate": 5.8031835646514544e-05, "loss": 2.6046, "step": 53522 }, { "epoch": 4.2, "learning_rate": 5.803105151729005e-05, "loss": 2.7414, "step": 53523 }, { "epoch": 4.2, "learning_rate": 5.803026738806555e-05, "loss": 2.5293, "step": 53524 }, { "epoch": 4.2, "learning_rate": 5.8029483258841055e-05, "loss": 2.6175, "step": 53525 }, { "epoch": 4.2, "learning_rate": 5.802869912961656e-05, "loss": 2.615, "step": 53526 }, { "epoch": 4.2, "learning_rate": 5.802791500039206e-05, "loss": 2.6712, "step": 53527 }, { "epoch": 4.2, "learning_rate": 5.8027130871167566e-05, "loss": 2.5622, "step": 53528 }, { "epoch": 4.2, "learning_rate": 5.802634674194307e-05, "loss": 2.7504, "step": 53529 }, { "epoch": 4.2, "learning_rate": 5.8025562612718574e-05, "loss": 2.6118, "step": 53530 }, { "epoch": 4.2, "learning_rate": 5.802477848349408e-05, "loss": 2.6015, "step": 53531 }, { "epoch": 4.2, "learning_rate": 5.802399435426958e-05, "loss": 2.6607, "step": 53532 }, { "epoch": 4.2, "learning_rate": 5.802321022504509e-05, "loss": 2.9736, "step": 53533 }, { "epoch": 4.2, "learning_rate": 5.8022426095820595e-05, "loss": 2.6387, "step": 53534 }, { "epoch": 4.2, "learning_rate": 5.80216419665961e-05, "loss": 2.8597, "step": 53535 }, { "epoch": 4.2, "learning_rate": 5.80208578373716e-05, "loss": 2.8797, "step": 53536 }, { "epoch": 4.2, "learning_rate": 5.8020073708147106e-05, "loss": 2.7276, "step": 53537 }, { "epoch": 4.2, "learning_rate": 5.801928957892261e-05, "loss": 2.4597, "step": 53538 }, { "epoch": 4.2, "learning_rate": 5.8018505449698114e-05, "loss": 2.5378, "step": 53539 }, { "epoch": 4.2, "learning_rate": 5.801772132047362e-05, "loss": 2.6398, "step": 53540 }, { "epoch": 4.2, "learning_rate": 5.801693719124912e-05, "loss": 2.7519, "step": 53541 }, { "epoch": 4.2, "learning_rate": 5.8016153062024625e-05, "loss": 2.8403, "step": 53542 }, { "epoch": 4.2, "learning_rate": 5.801536893280013e-05, "loss": 2.6867, "step": 53543 }, { "epoch": 4.2, "learning_rate": 5.801458480357563e-05, "loss": 2.5691, "step": 53544 }, { "epoch": 4.2, "learning_rate": 5.8013800674351136e-05, "loss": 2.6005, "step": 53545 }, { "epoch": 4.2, "learning_rate": 5.801301654512664e-05, "loss": 2.4501, "step": 53546 }, { "epoch": 4.2, "learning_rate": 5.801223241590214e-05, "loss": 2.5142, "step": 53547 }, { "epoch": 4.2, "learning_rate": 5.801144828667765e-05, "loss": 2.7102, "step": 53548 }, { "epoch": 4.2, "learning_rate": 5.801066415745316e-05, "loss": 2.4695, "step": 53549 }, { "epoch": 4.2, "learning_rate": 5.800988002822866e-05, "loss": 2.6778, "step": 53550 }, { "epoch": 4.2, "learning_rate": 5.8009095899004164e-05, "loss": 2.7842, "step": 53551 }, { "epoch": 4.2, "learning_rate": 5.800831176977967e-05, "loss": 2.537, "step": 53552 }, { "epoch": 4.2, "learning_rate": 5.800752764055517e-05, "loss": 2.5396, "step": 53553 }, { "epoch": 4.2, "learning_rate": 5.8006743511330675e-05, "loss": 2.9682, "step": 53554 }, { "epoch": 4.2, "learning_rate": 5.800595938210618e-05, "loss": 2.6777, "step": 53555 }, { "epoch": 4.2, "learning_rate": 5.800517525288168e-05, "loss": 2.8132, "step": 53556 }, { "epoch": 4.2, "learning_rate": 5.8004391123657187e-05, "loss": 2.9694, "step": 53557 }, { "epoch": 4.2, "learning_rate": 5.800360699443268e-05, "loss": 2.7544, "step": 53558 }, { "epoch": 4.2, "learning_rate": 5.800282286520819e-05, "loss": 2.5957, "step": 53559 }, { "epoch": 4.2, "learning_rate": 5.800203873598369e-05, "loss": 2.5783, "step": 53560 }, { "epoch": 4.2, "learning_rate": 5.8001254606759194e-05, "loss": 2.6723, "step": 53561 }, { "epoch": 4.2, "learning_rate": 5.80004704775347e-05, "loss": 2.5124, "step": 53562 }, { "epoch": 4.2, "learning_rate": 5.79996863483102e-05, "loss": 2.4862, "step": 53563 }, { "epoch": 4.2, "learning_rate": 5.7998902219085705e-05, "loss": 2.8072, "step": 53564 }, { "epoch": 4.2, "learning_rate": 5.799811808986121e-05, "loss": 2.7474, "step": 53565 }, { "epoch": 4.2, "learning_rate": 5.799733396063671e-05, "loss": 2.7087, "step": 53566 }, { "epoch": 4.2, "learning_rate": 5.7996549831412217e-05, "loss": 2.6466, "step": 53567 }, { "epoch": 4.2, "learning_rate": 5.799576570218772e-05, "loss": 2.7602, "step": 53568 }, { "epoch": 4.2, "learning_rate": 5.7994981572963224e-05, "loss": 2.7422, "step": 53569 }, { "epoch": 4.2, "learning_rate": 5.799419744373873e-05, "loss": 2.6935, "step": 53570 }, { "epoch": 4.2, "learning_rate": 5.799341331451423e-05, "loss": 2.6196, "step": 53571 }, { "epoch": 4.2, "learning_rate": 5.7992629185289735e-05, "loss": 2.6057, "step": 53572 }, { "epoch": 4.2, "learning_rate": 5.799184505606524e-05, "loss": 2.5231, "step": 53573 }, { "epoch": 4.2, "learning_rate": 5.799106092684074e-05, "loss": 2.5993, "step": 53574 }, { "epoch": 4.2, "learning_rate": 5.799027679761625e-05, "loss": 2.5719, "step": 53575 }, { "epoch": 4.2, "learning_rate": 5.7989492668391756e-05, "loss": 2.7352, "step": 53576 }, { "epoch": 4.2, "learning_rate": 5.798870853916726e-05, "loss": 2.705, "step": 53577 }, { "epoch": 4.2, "learning_rate": 5.7987924409942764e-05, "loss": 2.6999, "step": 53578 }, { "epoch": 4.2, "learning_rate": 5.798714028071827e-05, "loss": 2.5869, "step": 53579 }, { "epoch": 4.2, "learning_rate": 5.798635615149377e-05, "loss": 2.6177, "step": 53580 }, { "epoch": 4.2, "learning_rate": 5.7985572022269275e-05, "loss": 2.689, "step": 53581 }, { "epoch": 4.2, "learning_rate": 5.798478789304478e-05, "loss": 2.5431, "step": 53582 }, { "epoch": 4.2, "learning_rate": 5.798400376382028e-05, "loss": 2.6027, "step": 53583 }, { "epoch": 4.2, "learning_rate": 5.7983219634595786e-05, "loss": 2.9361, "step": 53584 }, { "epoch": 4.2, "learning_rate": 5.798243550537129e-05, "loss": 2.9148, "step": 53585 }, { "epoch": 4.2, "learning_rate": 5.798165137614679e-05, "loss": 2.8744, "step": 53586 }, { "epoch": 4.2, "learning_rate": 5.79808672469223e-05, "loss": 2.6797, "step": 53587 }, { "epoch": 4.2, "learning_rate": 5.79800831176978e-05, "loss": 2.6996, "step": 53588 }, { "epoch": 4.2, "learning_rate": 5.7979298988473304e-05, "loss": 2.5565, "step": 53589 }, { "epoch": 4.2, "learning_rate": 5.7978514859248815e-05, "loss": 2.7212, "step": 53590 }, { "epoch": 4.2, "learning_rate": 5.797773073002432e-05, "loss": 2.6861, "step": 53591 }, { "epoch": 4.2, "learning_rate": 5.797694660079982e-05, "loss": 2.703, "step": 53592 }, { "epoch": 4.2, "learning_rate": 5.7976162471575326e-05, "loss": 2.7374, "step": 53593 }, { "epoch": 4.2, "learning_rate": 5.7975378342350816e-05, "loss": 2.5778, "step": 53594 }, { "epoch": 4.2, "learning_rate": 5.797459421312632e-05, "loss": 2.5255, "step": 53595 }, { "epoch": 4.2, "learning_rate": 5.797381008390182e-05, "loss": 2.7629, "step": 53596 }, { "epoch": 4.2, "learning_rate": 5.797302595467733e-05, "loss": 2.5487, "step": 53597 }, { "epoch": 4.2, "learning_rate": 5.797224182545283e-05, "loss": 2.6291, "step": 53598 }, { "epoch": 4.2, "learning_rate": 5.7971457696228334e-05, "loss": 2.5626, "step": 53599 }, { "epoch": 4.2, "learning_rate": 5.7970673567003845e-05, "loss": 2.4134, "step": 53600 }, { "epoch": 4.2, "learning_rate": 5.796988943777935e-05, "loss": 2.7573, "step": 53601 }, { "epoch": 4.2, "learning_rate": 5.796910530855485e-05, "loss": 2.4941, "step": 53602 }, { "epoch": 4.2, "learning_rate": 5.7968321179330356e-05, "loss": 2.549, "step": 53603 }, { "epoch": 4.2, "learning_rate": 5.796753705010586e-05, "loss": 2.6465, "step": 53604 }, { "epoch": 4.2, "learning_rate": 5.796675292088136e-05, "loss": 2.6732, "step": 53605 }, { "epoch": 4.2, "learning_rate": 5.796596879165687e-05, "loss": 2.7769, "step": 53606 }, { "epoch": 4.2, "learning_rate": 5.796518466243237e-05, "loss": 2.7088, "step": 53607 }, { "epoch": 4.2, "learning_rate": 5.7964400533207874e-05, "loss": 2.7966, "step": 53608 }, { "epoch": 4.2, "learning_rate": 5.796361640398338e-05, "loss": 2.7359, "step": 53609 }, { "epoch": 4.2, "learning_rate": 5.796283227475888e-05, "loss": 2.476, "step": 53610 }, { "epoch": 4.2, "learning_rate": 5.7962048145534385e-05, "loss": 2.7778, "step": 53611 }, { "epoch": 4.2, "learning_rate": 5.796126401630989e-05, "loss": 2.7511, "step": 53612 }, { "epoch": 4.2, "learning_rate": 5.796047988708539e-05, "loss": 2.7369, "step": 53613 }, { "epoch": 4.2, "learning_rate": 5.7959695757860896e-05, "loss": 2.775, "step": 53614 }, { "epoch": 4.2, "learning_rate": 5.79589116286364e-05, "loss": 2.6624, "step": 53615 }, { "epoch": 4.2, "learning_rate": 5.795812749941191e-05, "loss": 3.0108, "step": 53616 }, { "epoch": 4.2, "learning_rate": 5.7957343370187414e-05, "loss": 2.8599, "step": 53617 }, { "epoch": 4.2, "learning_rate": 5.795655924096292e-05, "loss": 2.7497, "step": 53618 }, { "epoch": 4.2, "learning_rate": 5.795577511173842e-05, "loss": 2.8711, "step": 53619 }, { "epoch": 4.2, "learning_rate": 5.7954990982513925e-05, "loss": 2.7907, "step": 53620 }, { "epoch": 4.2, "learning_rate": 5.795420685328943e-05, "loss": 2.6565, "step": 53621 }, { "epoch": 4.2, "learning_rate": 5.795342272406493e-05, "loss": 2.6041, "step": 53622 }, { "epoch": 4.2, "learning_rate": 5.7952638594840436e-05, "loss": 2.6006, "step": 53623 }, { "epoch": 4.2, "learning_rate": 5.795185446561594e-05, "loss": 2.7056, "step": 53624 }, { "epoch": 4.2, "learning_rate": 5.795107033639144e-05, "loss": 2.6987, "step": 53625 }, { "epoch": 4.2, "learning_rate": 5.795028620716695e-05, "loss": 2.5557, "step": 53626 }, { "epoch": 4.21, "learning_rate": 5.794950207794245e-05, "loss": 2.5952, "step": 53627 }, { "epoch": 4.21, "learning_rate": 5.7948717948717954e-05, "loss": 2.504, "step": 53628 }, { "epoch": 4.21, "learning_rate": 5.794793381949346e-05, "loss": 2.6371, "step": 53629 }, { "epoch": 4.21, "learning_rate": 5.7947149690268955e-05, "loss": 2.7613, "step": 53630 }, { "epoch": 4.21, "learning_rate": 5.794636556104446e-05, "loss": 2.6993, "step": 53631 }, { "epoch": 4.21, "learning_rate": 5.794558143181996e-05, "loss": 2.9167, "step": 53632 }, { "epoch": 4.21, "learning_rate": 5.7944797302595466e-05, "loss": 2.6602, "step": 53633 }, { "epoch": 4.21, "learning_rate": 5.794401317337097e-05, "loss": 2.4941, "step": 53634 }, { "epoch": 4.21, "learning_rate": 5.794322904414647e-05, "loss": 2.7592, "step": 53635 }, { "epoch": 4.21, "learning_rate": 5.794244491492198e-05, "loss": 2.8122, "step": 53636 }, { "epoch": 4.21, "learning_rate": 5.794166078569748e-05, "loss": 2.6801, "step": 53637 }, { "epoch": 4.21, "learning_rate": 5.7940876656472984e-05, "loss": 2.7594, "step": 53638 }, { "epoch": 4.21, "learning_rate": 5.794009252724849e-05, "loss": 2.61, "step": 53639 }, { "epoch": 4.21, "learning_rate": 5.793930839802399e-05, "loss": 2.3565, "step": 53640 }, { "epoch": 4.21, "learning_rate": 5.79385242687995e-05, "loss": 2.7585, "step": 53641 }, { "epoch": 4.21, "learning_rate": 5.7937740139575006e-05, "loss": 2.7836, "step": 53642 }, { "epoch": 4.21, "learning_rate": 5.793695601035051e-05, "loss": 2.5177, "step": 53643 }, { "epoch": 4.21, "learning_rate": 5.793617188112601e-05, "loss": 2.61, "step": 53644 }, { "epoch": 4.21, "learning_rate": 5.793538775190152e-05, "loss": 2.8519, "step": 53645 }, { "epoch": 4.21, "learning_rate": 5.793460362267702e-05, "loss": 2.7303, "step": 53646 }, { "epoch": 4.21, "learning_rate": 5.7933819493452524e-05, "loss": 2.6264, "step": 53647 }, { "epoch": 4.21, "learning_rate": 5.793303536422803e-05, "loss": 2.5211, "step": 53648 }, { "epoch": 4.21, "learning_rate": 5.793225123500353e-05, "loss": 2.7617, "step": 53649 }, { "epoch": 4.21, "learning_rate": 5.7931467105779035e-05, "loss": 2.6177, "step": 53650 }, { "epoch": 4.21, "learning_rate": 5.793068297655454e-05, "loss": 2.7461, "step": 53651 }, { "epoch": 4.21, "learning_rate": 5.792989884733004e-05, "loss": 2.6939, "step": 53652 }, { "epoch": 4.21, "learning_rate": 5.7929114718105546e-05, "loss": 2.6372, "step": 53653 }, { "epoch": 4.21, "learning_rate": 5.792833058888105e-05, "loss": 2.6035, "step": 53654 }, { "epoch": 4.21, "learning_rate": 5.7927546459656554e-05, "loss": 2.5734, "step": 53655 }, { "epoch": 4.21, "learning_rate": 5.792676233043206e-05, "loss": 2.6709, "step": 53656 }, { "epoch": 4.21, "learning_rate": 5.792597820120757e-05, "loss": 2.7692, "step": 53657 }, { "epoch": 4.21, "learning_rate": 5.792519407198307e-05, "loss": 2.4541, "step": 53658 }, { "epoch": 4.21, "learning_rate": 5.7924409942758575e-05, "loss": 2.8916, "step": 53659 }, { "epoch": 4.21, "learning_rate": 5.792362581353408e-05, "loss": 2.6974, "step": 53660 }, { "epoch": 4.21, "learning_rate": 5.792284168430958e-05, "loss": 2.6131, "step": 53661 }, { "epoch": 4.21, "learning_rate": 5.7922057555085086e-05, "loss": 2.5532, "step": 53662 }, { "epoch": 4.21, "learning_rate": 5.792127342586059e-05, "loss": 2.5209, "step": 53663 }, { "epoch": 4.21, "learning_rate": 5.7920489296636094e-05, "loss": 2.6471, "step": 53664 }, { "epoch": 4.21, "learning_rate": 5.79197051674116e-05, "loss": 2.6391, "step": 53665 }, { "epoch": 4.21, "learning_rate": 5.791892103818709e-05, "loss": 2.6311, "step": 53666 }, { "epoch": 4.21, "learning_rate": 5.79181369089626e-05, "loss": 2.4707, "step": 53667 }, { "epoch": 4.21, "learning_rate": 5.79173527797381e-05, "loss": 2.6172, "step": 53668 }, { "epoch": 4.21, "learning_rate": 5.7916568650513605e-05, "loss": 2.5584, "step": 53669 }, { "epoch": 4.21, "learning_rate": 5.791578452128911e-05, "loss": 2.5108, "step": 53670 }, { "epoch": 4.21, "learning_rate": 5.791500039206461e-05, "loss": 2.5838, "step": 53671 }, { "epoch": 4.21, "learning_rate": 5.7914216262840116e-05, "loss": 2.5417, "step": 53672 }, { "epoch": 4.21, "learning_rate": 5.791343213361562e-05, "loss": 2.6075, "step": 53673 }, { "epoch": 4.21, "learning_rate": 5.7912648004391124e-05, "loss": 2.6062, "step": 53674 }, { "epoch": 4.21, "learning_rate": 5.791186387516663e-05, "loss": 2.6728, "step": 53675 }, { "epoch": 4.21, "learning_rate": 5.791107974594213e-05, "loss": 2.6029, "step": 53676 }, { "epoch": 4.21, "learning_rate": 5.7910295616717635e-05, "loss": 2.6121, "step": 53677 }, { "epoch": 4.21, "learning_rate": 5.790951148749314e-05, "loss": 2.5821, "step": 53678 }, { "epoch": 4.21, "learning_rate": 5.790872735826864e-05, "loss": 2.5805, "step": 53679 }, { "epoch": 4.21, "learning_rate": 5.7907943229044146e-05, "loss": 2.5035, "step": 53680 }, { "epoch": 4.21, "learning_rate": 5.790715909981965e-05, "loss": 2.6615, "step": 53681 }, { "epoch": 4.21, "learning_rate": 5.790637497059515e-05, "loss": 2.5058, "step": 53682 }, { "epoch": 4.21, "learning_rate": 5.7905590841370663e-05, "loss": 2.7081, "step": 53683 }, { "epoch": 4.21, "learning_rate": 5.790480671214617e-05, "loss": 2.5565, "step": 53684 }, { "epoch": 4.21, "learning_rate": 5.790402258292167e-05, "loss": 2.5799, "step": 53685 }, { "epoch": 4.21, "learning_rate": 5.7903238453697174e-05, "loss": 2.6129, "step": 53686 }, { "epoch": 4.21, "learning_rate": 5.790245432447268e-05, "loss": 2.6788, "step": 53687 }, { "epoch": 4.21, "learning_rate": 5.790167019524818e-05, "loss": 2.8236, "step": 53688 }, { "epoch": 4.21, "learning_rate": 5.7900886066023686e-05, "loss": 2.63, "step": 53689 }, { "epoch": 4.21, "learning_rate": 5.790010193679919e-05, "loss": 2.6424, "step": 53690 }, { "epoch": 4.21, "learning_rate": 5.789931780757469e-05, "loss": 2.6491, "step": 53691 }, { "epoch": 4.21, "learning_rate": 5.7898533678350197e-05, "loss": 2.8097, "step": 53692 }, { "epoch": 4.21, "learning_rate": 5.78977495491257e-05, "loss": 2.5794, "step": 53693 }, { "epoch": 4.21, "learning_rate": 5.7896965419901204e-05, "loss": 2.5064, "step": 53694 }, { "epoch": 4.21, "learning_rate": 5.789618129067671e-05, "loss": 2.673, "step": 53695 }, { "epoch": 4.21, "learning_rate": 5.789539716145221e-05, "loss": 2.5591, "step": 53696 }, { "epoch": 4.21, "learning_rate": 5.7894613032227715e-05, "loss": 2.5795, "step": 53697 }, { "epoch": 4.21, "learning_rate": 5.7893828903003225e-05, "loss": 2.6689, "step": 53698 }, { "epoch": 4.21, "learning_rate": 5.789304477377873e-05, "loss": 2.7853, "step": 53699 }, { "epoch": 4.21, "learning_rate": 5.789226064455423e-05, "loss": 2.8207, "step": 53700 }, { "epoch": 4.21, "learning_rate": 5.7891476515329736e-05, "loss": 2.6759, "step": 53701 }, { "epoch": 4.21, "learning_rate": 5.7890692386105227e-05, "loss": 2.7517, "step": 53702 }, { "epoch": 4.21, "learning_rate": 5.788990825688073e-05, "loss": 2.65, "step": 53703 }, { "epoch": 4.21, "learning_rate": 5.7889124127656234e-05, "loss": 2.729, "step": 53704 }, { "epoch": 4.21, "learning_rate": 5.788833999843174e-05, "loss": 2.5955, "step": 53705 }, { "epoch": 4.21, "learning_rate": 5.788755586920724e-05, "loss": 2.6157, "step": 53706 }, { "epoch": 4.21, "learning_rate": 5.7886771739982745e-05, "loss": 2.5809, "step": 53707 }, { "epoch": 4.21, "learning_rate": 5.7885987610758255e-05, "loss": 2.5881, "step": 53708 }, { "epoch": 4.21, "learning_rate": 5.788520348153376e-05, "loss": 2.5761, "step": 53709 }, { "epoch": 4.21, "learning_rate": 5.788441935230926e-05, "loss": 2.6924, "step": 53710 }, { "epoch": 4.21, "learning_rate": 5.7883635223084766e-05, "loss": 2.7069, "step": 53711 }, { "epoch": 4.21, "learning_rate": 5.788285109386027e-05, "loss": 2.5179, "step": 53712 }, { "epoch": 4.21, "learning_rate": 5.7882066964635774e-05, "loss": 2.6648, "step": 53713 }, { "epoch": 4.21, "learning_rate": 5.788128283541128e-05, "loss": 2.6402, "step": 53714 }, { "epoch": 4.21, "learning_rate": 5.788049870618678e-05, "loss": 2.4794, "step": 53715 }, { "epoch": 4.21, "learning_rate": 5.7879714576962285e-05, "loss": 2.6586, "step": 53716 }, { "epoch": 4.21, "learning_rate": 5.787893044773779e-05, "loss": 2.6909, "step": 53717 }, { "epoch": 4.21, "learning_rate": 5.787814631851329e-05, "loss": 2.563, "step": 53718 }, { "epoch": 4.21, "learning_rate": 5.7877362189288796e-05, "loss": 2.6795, "step": 53719 }, { "epoch": 4.21, "learning_rate": 5.78765780600643e-05, "loss": 2.5684, "step": 53720 }, { "epoch": 4.21, "learning_rate": 5.78757939308398e-05, "loss": 2.7054, "step": 53721 }, { "epoch": 4.21, "learning_rate": 5.787500980161531e-05, "loss": 3.0959, "step": 53722 }, { "epoch": 4.21, "learning_rate": 5.787422567239081e-05, "loss": 2.7264, "step": 53723 }, { "epoch": 4.21, "learning_rate": 5.787344154316632e-05, "loss": 2.6121, "step": 53724 }, { "epoch": 4.21, "learning_rate": 5.7872657413941825e-05, "loss": 2.5389, "step": 53725 }, { "epoch": 4.21, "learning_rate": 5.787187328471733e-05, "loss": 2.5476, "step": 53726 }, { "epoch": 4.21, "learning_rate": 5.787108915549283e-05, "loss": 2.8687, "step": 53727 }, { "epoch": 4.21, "learning_rate": 5.7870305026268336e-05, "loss": 2.6303, "step": 53728 }, { "epoch": 4.21, "learning_rate": 5.786952089704384e-05, "loss": 2.6302, "step": 53729 }, { "epoch": 4.21, "learning_rate": 5.786873676781934e-05, "loss": 2.584, "step": 53730 }, { "epoch": 4.21, "learning_rate": 5.786795263859485e-05, "loss": 2.9912, "step": 53731 }, { "epoch": 4.21, "learning_rate": 5.786716850937035e-05, "loss": 2.7063, "step": 53732 }, { "epoch": 4.21, "learning_rate": 5.7866384380145854e-05, "loss": 2.7776, "step": 53733 }, { "epoch": 4.21, "learning_rate": 5.786560025092136e-05, "loss": 2.5864, "step": 53734 }, { "epoch": 4.21, "learning_rate": 5.786481612169686e-05, "loss": 2.5922, "step": 53735 }, { "epoch": 4.21, "learning_rate": 5.7864031992472365e-05, "loss": 2.4857, "step": 53736 }, { "epoch": 4.21, "learning_rate": 5.786324786324787e-05, "loss": 2.6847, "step": 53737 }, { "epoch": 4.21, "learning_rate": 5.7862463734023366e-05, "loss": 2.6497, "step": 53738 }, { "epoch": 4.21, "learning_rate": 5.786167960479887e-05, "loss": 2.5703, "step": 53739 }, { "epoch": 4.21, "learning_rate": 5.786089547557437e-05, "loss": 2.7751, "step": 53740 }, { "epoch": 4.21, "learning_rate": 5.786011134634988e-05, "loss": 2.3834, "step": 53741 }, { "epoch": 4.21, "learning_rate": 5.785932721712538e-05, "loss": 2.6558, "step": 53742 }, { "epoch": 4.21, "learning_rate": 5.7858543087900884e-05, "loss": 2.6208, "step": 53743 }, { "epoch": 4.21, "learning_rate": 5.785775895867639e-05, "loss": 2.6966, "step": 53744 }, { "epoch": 4.21, "learning_rate": 5.785697482945189e-05, "loss": 2.6488, "step": 53745 }, { "epoch": 4.21, "learning_rate": 5.7856190700227395e-05, "loss": 2.711, "step": 53746 }, { "epoch": 4.21, "learning_rate": 5.78554065710029e-05, "loss": 2.6697, "step": 53747 }, { "epoch": 4.21, "learning_rate": 5.78546224417784e-05, "loss": 2.5805, "step": 53748 }, { "epoch": 4.21, "learning_rate": 5.785383831255391e-05, "loss": 2.643, "step": 53749 }, { "epoch": 4.21, "learning_rate": 5.785305418332942e-05, "loss": 2.7759, "step": 53750 }, { "epoch": 4.21, "learning_rate": 5.785227005410492e-05, "loss": 2.5327, "step": 53751 }, { "epoch": 4.21, "learning_rate": 5.7851485924880424e-05, "loss": 2.9743, "step": 53752 }, { "epoch": 4.21, "learning_rate": 5.785070179565593e-05, "loss": 2.3954, "step": 53753 }, { "epoch": 4.22, "learning_rate": 5.784991766643143e-05, "loss": 2.5215, "step": 53754 }, { "epoch": 4.22, "learning_rate": 5.7849133537206935e-05, "loss": 2.5124, "step": 53755 }, { "epoch": 4.22, "learning_rate": 5.784834940798244e-05, "loss": 2.5326, "step": 53756 }, { "epoch": 4.22, "learning_rate": 5.784756527875794e-05, "loss": 2.5482, "step": 53757 }, { "epoch": 4.22, "learning_rate": 5.7846781149533446e-05, "loss": 2.7302, "step": 53758 }, { "epoch": 4.22, "learning_rate": 5.784599702030895e-05, "loss": 2.7378, "step": 53759 }, { "epoch": 4.22, "learning_rate": 5.7845212891084453e-05, "loss": 2.5792, "step": 53760 }, { "epoch": 4.22, "learning_rate": 5.784442876185996e-05, "loss": 2.4679, "step": 53761 }, { "epoch": 4.22, "learning_rate": 5.784364463263546e-05, "loss": 2.6102, "step": 53762 }, { "epoch": 4.22, "learning_rate": 5.7842860503410964e-05, "loss": 2.5959, "step": 53763 }, { "epoch": 4.22, "learning_rate": 5.784207637418647e-05, "loss": 2.525, "step": 53764 }, { "epoch": 4.22, "learning_rate": 5.784129224496198e-05, "loss": 2.8653, "step": 53765 }, { "epoch": 4.22, "learning_rate": 5.784050811573748e-05, "loss": 2.5357, "step": 53766 }, { "epoch": 4.22, "learning_rate": 5.7839723986512986e-05, "loss": 2.6676, "step": 53767 }, { "epoch": 4.22, "learning_rate": 5.783893985728849e-05, "loss": 2.7116, "step": 53768 }, { "epoch": 4.22, "learning_rate": 5.783815572806399e-05, "loss": 2.7585, "step": 53769 }, { "epoch": 4.22, "learning_rate": 5.78373715988395e-05, "loss": 2.6507, "step": 53770 }, { "epoch": 4.22, "learning_rate": 5.7836587469615e-05, "loss": 2.6062, "step": 53771 }, { "epoch": 4.22, "learning_rate": 5.7835803340390504e-05, "loss": 2.6597, "step": 53772 }, { "epoch": 4.22, "learning_rate": 5.783501921116601e-05, "loss": 2.6779, "step": 53773 }, { "epoch": 4.22, "learning_rate": 5.78342350819415e-05, "loss": 2.6446, "step": 53774 }, { "epoch": 4.22, "learning_rate": 5.783345095271701e-05, "loss": 2.9265, "step": 53775 }, { "epoch": 4.22, "learning_rate": 5.783266682349251e-05, "loss": 2.5369, "step": 53776 }, { "epoch": 4.22, "learning_rate": 5.7831882694268016e-05, "loss": 2.6153, "step": 53777 }, { "epoch": 4.22, "learning_rate": 5.783109856504352e-05, "loss": 2.6828, "step": 53778 }, { "epoch": 4.22, "learning_rate": 5.783031443581902e-05, "loss": 2.5649, "step": 53779 }, { "epoch": 4.22, "learning_rate": 5.782953030659453e-05, "loss": 2.5246, "step": 53780 }, { "epoch": 4.22, "learning_rate": 5.782874617737003e-05, "loss": 2.6156, "step": 53781 }, { "epoch": 4.22, "learning_rate": 5.7827962048145534e-05, "loss": 2.6386, "step": 53782 }, { "epoch": 4.22, "learning_rate": 5.782717791892104e-05, "loss": 2.5323, "step": 53783 }, { "epoch": 4.22, "learning_rate": 5.782639378969654e-05, "loss": 2.9276, "step": 53784 }, { "epoch": 4.22, "learning_rate": 5.7825609660472045e-05, "loss": 2.8432, "step": 53785 }, { "epoch": 4.22, "learning_rate": 5.782482553124755e-05, "loss": 2.7254, "step": 53786 }, { "epoch": 4.22, "learning_rate": 5.782404140202305e-05, "loss": 2.9476, "step": 53787 }, { "epoch": 4.22, "learning_rate": 5.7823257272798556e-05, "loss": 2.5598, "step": 53788 }, { "epoch": 4.22, "learning_rate": 5.782247314357406e-05, "loss": 2.7363, "step": 53789 }, { "epoch": 4.22, "learning_rate": 5.7821689014349564e-05, "loss": 2.532, "step": 53790 }, { "epoch": 4.22, "learning_rate": 5.7820904885125074e-05, "loss": 2.7874, "step": 53791 }, { "epoch": 4.22, "learning_rate": 5.782012075590058e-05, "loss": 2.5292, "step": 53792 }, { "epoch": 4.22, "learning_rate": 5.781933662667608e-05, "loss": 2.6797, "step": 53793 }, { "epoch": 4.22, "learning_rate": 5.7818552497451585e-05, "loss": 2.6856, "step": 53794 }, { "epoch": 4.22, "learning_rate": 5.781776836822709e-05, "loss": 2.6777, "step": 53795 }, { "epoch": 4.22, "learning_rate": 5.781698423900259e-05, "loss": 2.7936, "step": 53796 }, { "epoch": 4.22, "learning_rate": 5.7816200109778096e-05, "loss": 2.7204, "step": 53797 }, { "epoch": 4.22, "learning_rate": 5.78154159805536e-05, "loss": 2.8142, "step": 53798 }, { "epoch": 4.22, "learning_rate": 5.7814631851329104e-05, "loss": 2.5878, "step": 53799 }, { "epoch": 4.22, "learning_rate": 5.781384772210461e-05, "loss": 2.6802, "step": 53800 }, { "epoch": 4.22, "learning_rate": 5.781306359288011e-05, "loss": 2.46, "step": 53801 }, { "epoch": 4.22, "learning_rate": 5.7812279463655615e-05, "loss": 2.9845, "step": 53802 }, { "epoch": 4.22, "learning_rate": 5.781149533443112e-05, "loss": 2.6245, "step": 53803 }, { "epoch": 4.22, "learning_rate": 5.781071120520662e-05, "loss": 2.7009, "step": 53804 }, { "epoch": 4.22, "learning_rate": 5.7809927075982126e-05, "loss": 2.642, "step": 53805 }, { "epoch": 4.22, "learning_rate": 5.7809142946757636e-05, "loss": 2.559, "step": 53806 }, { "epoch": 4.22, "learning_rate": 5.780835881753314e-05, "loss": 2.6412, "step": 53807 }, { "epoch": 4.22, "learning_rate": 5.7807574688308644e-05, "loss": 2.7252, "step": 53808 }, { "epoch": 4.22, "learning_rate": 5.780679055908415e-05, "loss": 2.5848, "step": 53809 }, { "epoch": 4.22, "learning_rate": 5.780600642985964e-05, "loss": 2.7303, "step": 53810 }, { "epoch": 4.22, "learning_rate": 5.780522230063514e-05, "loss": 2.4179, "step": 53811 }, { "epoch": 4.22, "learning_rate": 5.7804438171410645e-05, "loss": 2.5623, "step": 53812 }, { "epoch": 4.22, "learning_rate": 5.780365404218615e-05, "loss": 2.619, "step": 53813 }, { "epoch": 4.22, "learning_rate": 5.780286991296165e-05, "loss": 2.5131, "step": 53814 }, { "epoch": 4.22, "learning_rate": 5.7802085783737156e-05, "loss": 2.5337, "step": 53815 }, { "epoch": 4.22, "learning_rate": 5.7801301654512666e-05, "loss": 2.7408, "step": 53816 }, { "epoch": 4.22, "learning_rate": 5.780051752528817e-05, "loss": 2.6524, "step": 53817 }, { "epoch": 4.22, "learning_rate": 5.7799733396063674e-05, "loss": 2.6111, "step": 53818 }, { "epoch": 4.22, "learning_rate": 5.779894926683918e-05, "loss": 2.5653, "step": 53819 }, { "epoch": 4.22, "learning_rate": 5.779816513761468e-05, "loss": 2.7278, "step": 53820 }, { "epoch": 4.22, "learning_rate": 5.7797381008390185e-05, "loss": 2.7213, "step": 53821 }, { "epoch": 4.22, "learning_rate": 5.779659687916569e-05, "loss": 2.6358, "step": 53822 }, { "epoch": 4.22, "learning_rate": 5.779581274994119e-05, "loss": 2.855, "step": 53823 }, { "epoch": 4.22, "learning_rate": 5.7795028620716696e-05, "loss": 2.694, "step": 53824 }, { "epoch": 4.22, "learning_rate": 5.77942444914922e-05, "loss": 2.8556, "step": 53825 }, { "epoch": 4.22, "learning_rate": 5.77934603622677e-05, "loss": 2.8518, "step": 53826 }, { "epoch": 4.22, "learning_rate": 5.7792676233043207e-05, "loss": 2.6149, "step": 53827 }, { "epoch": 4.22, "learning_rate": 5.779189210381871e-05, "loss": 2.6176, "step": 53828 }, { "epoch": 4.22, "learning_rate": 5.7791107974594214e-05, "loss": 2.6655, "step": 53829 }, { "epoch": 4.22, "learning_rate": 5.779032384536972e-05, "loss": 2.6855, "step": 53830 }, { "epoch": 4.22, "learning_rate": 5.778953971614522e-05, "loss": 2.4927, "step": 53831 }, { "epoch": 4.22, "learning_rate": 5.778875558692073e-05, "loss": 2.7836, "step": 53832 }, { "epoch": 4.22, "learning_rate": 5.7787971457696235e-05, "loss": 2.6911, "step": 53833 }, { "epoch": 4.22, "learning_rate": 5.778718732847174e-05, "loss": 2.5884, "step": 53834 }, { "epoch": 4.22, "learning_rate": 5.778640319924724e-05, "loss": 2.8017, "step": 53835 }, { "epoch": 4.22, "learning_rate": 5.7785619070022746e-05, "loss": 2.6029, "step": 53836 }, { "epoch": 4.22, "learning_rate": 5.778483494079825e-05, "loss": 2.6952, "step": 53837 }, { "epoch": 4.22, "learning_rate": 5.7784050811573754e-05, "loss": 2.4791, "step": 53838 }, { "epoch": 4.22, "learning_rate": 5.778326668234926e-05, "loss": 2.6248, "step": 53839 }, { "epoch": 4.22, "learning_rate": 5.778248255312476e-05, "loss": 2.7363, "step": 53840 }, { "epoch": 4.22, "learning_rate": 5.7781698423900265e-05, "loss": 2.544, "step": 53841 }, { "epoch": 4.22, "learning_rate": 5.778091429467577e-05, "loss": 2.494, "step": 53842 }, { "epoch": 4.22, "learning_rate": 5.778013016545127e-05, "loss": 2.8307, "step": 53843 }, { "epoch": 4.22, "learning_rate": 5.7779346036226776e-05, "loss": 2.8718, "step": 53844 }, { "epoch": 4.22, "learning_rate": 5.777856190700228e-05, "loss": 2.6933, "step": 53845 }, { "epoch": 4.22, "learning_rate": 5.7777777777777776e-05, "loss": 2.6671, "step": 53846 }, { "epoch": 4.22, "learning_rate": 5.777699364855328e-05, "loss": 2.7026, "step": 53847 }, { "epoch": 4.22, "learning_rate": 5.7776209519328784e-05, "loss": 2.6709, "step": 53848 }, { "epoch": 4.22, "learning_rate": 5.777542539010429e-05, "loss": 2.6684, "step": 53849 }, { "epoch": 4.22, "learning_rate": 5.777464126087979e-05, "loss": 2.6983, "step": 53850 }, { "epoch": 4.22, "learning_rate": 5.7773857131655295e-05, "loss": 2.5671, "step": 53851 }, { "epoch": 4.22, "learning_rate": 5.77730730024308e-05, "loss": 2.7509, "step": 53852 }, { "epoch": 4.22, "learning_rate": 5.77722888732063e-05, "loss": 2.765, "step": 53853 }, { "epoch": 4.22, "learning_rate": 5.7771504743981806e-05, "loss": 2.7721, "step": 53854 }, { "epoch": 4.22, "learning_rate": 5.777072061475731e-05, "loss": 2.7635, "step": 53855 }, { "epoch": 4.22, "learning_rate": 5.776993648553281e-05, "loss": 2.6635, "step": 53856 }, { "epoch": 4.22, "learning_rate": 5.7769152356308324e-05, "loss": 2.4882, "step": 53857 }, { "epoch": 4.22, "learning_rate": 5.776836822708383e-05, "loss": 2.5865, "step": 53858 }, { "epoch": 4.22, "learning_rate": 5.776758409785933e-05, "loss": 2.7118, "step": 53859 }, { "epoch": 4.22, "learning_rate": 5.7766799968634835e-05, "loss": 2.6362, "step": 53860 }, { "epoch": 4.22, "learning_rate": 5.776601583941034e-05, "loss": 2.6584, "step": 53861 }, { "epoch": 4.22, "learning_rate": 5.776523171018584e-05, "loss": 2.6273, "step": 53862 }, { "epoch": 4.22, "learning_rate": 5.7764447580961346e-05, "loss": 2.5738, "step": 53863 }, { "epoch": 4.22, "learning_rate": 5.776366345173685e-05, "loss": 2.5298, "step": 53864 }, { "epoch": 4.22, "learning_rate": 5.776287932251235e-05, "loss": 2.6643, "step": 53865 }, { "epoch": 4.22, "learning_rate": 5.776209519328786e-05, "loss": 2.6255, "step": 53866 }, { "epoch": 4.22, "learning_rate": 5.776131106406336e-05, "loss": 2.5987, "step": 53867 }, { "epoch": 4.22, "learning_rate": 5.7760526934838864e-05, "loss": 2.9421, "step": 53868 }, { "epoch": 4.22, "learning_rate": 5.775974280561437e-05, "loss": 2.6291, "step": 53869 }, { "epoch": 4.22, "learning_rate": 5.775895867638987e-05, "loss": 2.6766, "step": 53870 }, { "epoch": 4.22, "learning_rate": 5.7758174547165375e-05, "loss": 2.6989, "step": 53871 }, { "epoch": 4.22, "learning_rate": 5.775739041794088e-05, "loss": 2.7345, "step": 53872 }, { "epoch": 4.22, "learning_rate": 5.775660628871639e-05, "loss": 2.6816, "step": 53873 }, { "epoch": 4.22, "learning_rate": 5.775582215949189e-05, "loss": 2.6059, "step": 53874 }, { "epoch": 4.22, "learning_rate": 5.77550380302674e-05, "loss": 2.6789, "step": 53875 }, { "epoch": 4.22, "learning_rate": 5.77542539010429e-05, "loss": 2.6305, "step": 53876 }, { "epoch": 4.22, "learning_rate": 5.7753469771818404e-05, "loss": 2.711, "step": 53877 }, { "epoch": 4.22, "learning_rate": 5.775268564259391e-05, "loss": 2.5448, "step": 53878 }, { "epoch": 4.22, "learning_rate": 5.775190151336941e-05, "loss": 2.6827, "step": 53879 }, { "epoch": 4.22, "learning_rate": 5.7751117384144915e-05, "loss": 2.6902, "step": 53880 }, { "epoch": 4.22, "learning_rate": 5.775033325492042e-05, "loss": 2.5557, "step": 53881 }, { "epoch": 4.23, "learning_rate": 5.774954912569591e-05, "loss": 2.6628, "step": 53882 }, { "epoch": 4.23, "learning_rate": 5.774876499647142e-05, "loss": 2.7335, "step": 53883 }, { "epoch": 4.23, "learning_rate": 5.774798086724692e-05, "loss": 2.5946, "step": 53884 }, { "epoch": 4.23, "learning_rate": 5.774719673802243e-05, "loss": 2.4921, "step": 53885 }, { "epoch": 4.23, "learning_rate": 5.774641260879793e-05, "loss": 2.6063, "step": 53886 }, { "epoch": 4.23, "learning_rate": 5.7745628479573434e-05, "loss": 2.5777, "step": 53887 }, { "epoch": 4.23, "learning_rate": 5.774484435034894e-05, "loss": 2.5129, "step": 53888 }, { "epoch": 4.23, "learning_rate": 5.774406022112444e-05, "loss": 2.6414, "step": 53889 }, { "epoch": 4.23, "learning_rate": 5.7743276091899945e-05, "loss": 2.7218, "step": 53890 }, { "epoch": 4.23, "learning_rate": 5.774249196267545e-05, "loss": 2.7176, "step": 53891 }, { "epoch": 4.23, "learning_rate": 5.774170783345095e-05, "loss": 2.4912, "step": 53892 }, { "epoch": 4.23, "learning_rate": 5.7740923704226456e-05, "loss": 2.5509, "step": 53893 }, { "epoch": 4.23, "learning_rate": 5.774013957500196e-05, "loss": 2.6345, "step": 53894 }, { "epoch": 4.23, "learning_rate": 5.7739355445777463e-05, "loss": 2.8151, "step": 53895 }, { "epoch": 4.23, "learning_rate": 5.773857131655297e-05, "loss": 2.5299, "step": 53896 }, { "epoch": 4.23, "learning_rate": 5.773778718732847e-05, "loss": 2.4944, "step": 53897 }, { "epoch": 4.23, "learning_rate": 5.7737003058103974e-05, "loss": 2.4947, "step": 53898 }, { "epoch": 4.23, "learning_rate": 5.7736218928879485e-05, "loss": 2.432, "step": 53899 }, { "epoch": 4.23, "learning_rate": 5.773543479965499e-05, "loss": 2.6955, "step": 53900 }, { "epoch": 4.23, "learning_rate": 5.773465067043049e-05, "loss": 2.7256, "step": 53901 }, { "epoch": 4.23, "learning_rate": 5.7733866541205996e-05, "loss": 2.6439, "step": 53902 }, { "epoch": 4.23, "learning_rate": 5.77330824119815e-05, "loss": 2.4857, "step": 53903 }, { "epoch": 4.23, "learning_rate": 5.7732298282757e-05, "loss": 2.6824, "step": 53904 }, { "epoch": 4.23, "learning_rate": 5.773151415353251e-05, "loss": 2.7998, "step": 53905 }, { "epoch": 4.23, "learning_rate": 5.773073002430801e-05, "loss": 2.659, "step": 53906 }, { "epoch": 4.23, "learning_rate": 5.7729945895083514e-05, "loss": 2.6585, "step": 53907 }, { "epoch": 4.23, "learning_rate": 5.772916176585902e-05, "loss": 2.6598, "step": 53908 }, { "epoch": 4.23, "learning_rate": 5.772837763663452e-05, "loss": 2.7738, "step": 53909 }, { "epoch": 4.23, "learning_rate": 5.7727593507410025e-05, "loss": 2.7169, "step": 53910 }, { "epoch": 4.23, "learning_rate": 5.772680937818553e-05, "loss": 2.5015, "step": 53911 }, { "epoch": 4.23, "learning_rate": 5.772602524896103e-05, "loss": 2.6704, "step": 53912 }, { "epoch": 4.23, "learning_rate": 5.7725241119736536e-05, "loss": 2.7181, "step": 53913 }, { "epoch": 4.23, "learning_rate": 5.772445699051205e-05, "loss": 2.6333, "step": 53914 }, { "epoch": 4.23, "learning_rate": 5.772367286128755e-05, "loss": 2.688, "step": 53915 }, { "epoch": 4.23, "learning_rate": 5.7722888732063054e-05, "loss": 2.635, "step": 53916 }, { "epoch": 4.23, "learning_rate": 5.772210460283856e-05, "loss": 2.7244, "step": 53917 }, { "epoch": 4.23, "learning_rate": 5.772132047361405e-05, "loss": 2.6893, "step": 53918 }, { "epoch": 4.23, "learning_rate": 5.772053634438955e-05, "loss": 2.6125, "step": 53919 }, { "epoch": 4.23, "learning_rate": 5.7719752215165055e-05, "loss": 2.4769, "step": 53920 }, { "epoch": 4.23, "learning_rate": 5.771896808594056e-05, "loss": 2.6422, "step": 53921 }, { "epoch": 4.23, "learning_rate": 5.771818395671606e-05, "loss": 2.6551, "step": 53922 }, { "epoch": 4.23, "learning_rate": 5.7717399827491566e-05, "loss": 2.622, "step": 53923 }, { "epoch": 4.23, "learning_rate": 5.771661569826708e-05, "loss": 2.6789, "step": 53924 }, { "epoch": 4.23, "learning_rate": 5.771583156904258e-05, "loss": 2.7626, "step": 53925 }, { "epoch": 4.23, "learning_rate": 5.7715047439818084e-05, "loss": 2.5363, "step": 53926 }, { "epoch": 4.23, "learning_rate": 5.771426331059359e-05, "loss": 2.8057, "step": 53927 }, { "epoch": 4.23, "learning_rate": 5.771347918136909e-05, "loss": 2.8086, "step": 53928 }, { "epoch": 4.23, "learning_rate": 5.7712695052144595e-05, "loss": 2.5084, "step": 53929 }, { "epoch": 4.23, "learning_rate": 5.77119109229201e-05, "loss": 2.5686, "step": 53930 }, { "epoch": 4.23, "learning_rate": 5.77111267936956e-05, "loss": 2.7845, "step": 53931 }, { "epoch": 4.23, "learning_rate": 5.7710342664471106e-05, "loss": 2.8319, "step": 53932 }, { "epoch": 4.23, "learning_rate": 5.770955853524661e-05, "loss": 2.801, "step": 53933 }, { "epoch": 4.23, "learning_rate": 5.7708774406022114e-05, "loss": 2.6293, "step": 53934 }, { "epoch": 4.23, "learning_rate": 5.770799027679762e-05, "loss": 2.5523, "step": 53935 }, { "epoch": 4.23, "learning_rate": 5.770720614757312e-05, "loss": 2.4639, "step": 53936 }, { "epoch": 4.23, "learning_rate": 5.7706422018348625e-05, "loss": 2.8797, "step": 53937 }, { "epoch": 4.23, "learning_rate": 5.770563788912413e-05, "loss": 2.7256, "step": 53938 }, { "epoch": 4.23, "learning_rate": 5.770485375989963e-05, "loss": 2.8004, "step": 53939 }, { "epoch": 4.23, "learning_rate": 5.770406963067514e-05, "loss": 2.4406, "step": 53940 }, { "epoch": 4.23, "learning_rate": 5.7703285501450646e-05, "loss": 2.5459, "step": 53941 }, { "epoch": 4.23, "learning_rate": 5.770250137222615e-05, "loss": 2.7392, "step": 53942 }, { "epoch": 4.23, "learning_rate": 5.7701717243001654e-05, "loss": 2.6146, "step": 53943 }, { "epoch": 4.23, "learning_rate": 5.770093311377716e-05, "loss": 2.7313, "step": 53944 }, { "epoch": 4.23, "learning_rate": 5.770014898455266e-05, "loss": 2.7024, "step": 53945 }, { "epoch": 4.23, "learning_rate": 5.7699364855328165e-05, "loss": 2.7278, "step": 53946 }, { "epoch": 4.23, "learning_rate": 5.769858072610367e-05, "loss": 2.6569, "step": 53947 }, { "epoch": 4.23, "learning_rate": 5.769779659687917e-05, "loss": 2.6566, "step": 53948 }, { "epoch": 4.23, "learning_rate": 5.7697012467654676e-05, "loss": 2.4426, "step": 53949 }, { "epoch": 4.23, "learning_rate": 5.769622833843018e-05, "loss": 2.8243, "step": 53950 }, { "epoch": 4.23, "learning_rate": 5.769544420920568e-05, "loss": 2.7272, "step": 53951 }, { "epoch": 4.23, "learning_rate": 5.769466007998119e-05, "loss": 2.6023, "step": 53952 }, { "epoch": 4.23, "learning_rate": 5.769387595075669e-05, "loss": 2.8479, "step": 53953 }, { "epoch": 4.23, "learning_rate": 5.769309182153219e-05, "loss": 2.8069, "step": 53954 }, { "epoch": 4.23, "learning_rate": 5.769230769230769e-05, "loss": 2.5778, "step": 53955 }, { "epoch": 4.23, "learning_rate": 5.7691523563083195e-05, "loss": 2.7286, "step": 53956 }, { "epoch": 4.23, "learning_rate": 5.76907394338587e-05, "loss": 2.5852, "step": 53957 }, { "epoch": 4.23, "learning_rate": 5.76899553046342e-05, "loss": 2.6643, "step": 53958 }, { "epoch": 4.23, "learning_rate": 5.7689171175409706e-05, "loss": 2.7007, "step": 53959 }, { "epoch": 4.23, "learning_rate": 5.768838704618521e-05, "loss": 2.5769, "step": 53960 }, { "epoch": 4.23, "learning_rate": 5.768760291696071e-05, "loss": 2.4845, "step": 53961 }, { "epoch": 4.23, "learning_rate": 5.768681878773622e-05, "loss": 2.5561, "step": 53962 }, { "epoch": 4.23, "learning_rate": 5.768603465851172e-05, "loss": 2.6818, "step": 53963 }, { "epoch": 4.23, "learning_rate": 5.7685250529287224e-05, "loss": 2.5503, "step": 53964 }, { "epoch": 4.23, "learning_rate": 5.768446640006273e-05, "loss": 2.6561, "step": 53965 }, { "epoch": 4.23, "learning_rate": 5.768368227083824e-05, "loss": 2.7551, "step": 53966 }, { "epoch": 4.23, "learning_rate": 5.768289814161374e-05, "loss": 2.5658, "step": 53967 }, { "epoch": 4.23, "learning_rate": 5.7682114012389245e-05, "loss": 2.5763, "step": 53968 }, { "epoch": 4.23, "learning_rate": 5.768132988316475e-05, "loss": 2.6392, "step": 53969 }, { "epoch": 4.23, "learning_rate": 5.768054575394025e-05, "loss": 2.7718, "step": 53970 }, { "epoch": 4.23, "learning_rate": 5.7679761624715757e-05, "loss": 2.5619, "step": 53971 }, { "epoch": 4.23, "learning_rate": 5.767897749549126e-05, "loss": 2.7689, "step": 53972 }, { "epoch": 4.23, "learning_rate": 5.7678193366266764e-05, "loss": 2.7159, "step": 53973 }, { "epoch": 4.23, "learning_rate": 5.767740923704227e-05, "loss": 2.7655, "step": 53974 }, { "epoch": 4.23, "learning_rate": 5.767662510781777e-05, "loss": 2.7384, "step": 53975 }, { "epoch": 4.23, "learning_rate": 5.7675840978593275e-05, "loss": 2.6907, "step": 53976 }, { "epoch": 4.23, "learning_rate": 5.767505684936878e-05, "loss": 2.6781, "step": 53977 }, { "epoch": 4.23, "learning_rate": 5.767427272014428e-05, "loss": 2.5862, "step": 53978 }, { "epoch": 4.23, "learning_rate": 5.7673488590919786e-05, "loss": 2.5724, "step": 53979 }, { "epoch": 4.23, "learning_rate": 5.767270446169529e-05, "loss": 2.5015, "step": 53980 }, { "epoch": 4.23, "learning_rate": 5.76719203324708e-05, "loss": 2.6395, "step": 53981 }, { "epoch": 4.23, "learning_rate": 5.7671136203246304e-05, "loss": 2.5323, "step": 53982 }, { "epoch": 4.23, "learning_rate": 5.767035207402181e-05, "loss": 2.5264, "step": 53983 }, { "epoch": 4.23, "learning_rate": 5.766956794479731e-05, "loss": 2.4708, "step": 53984 }, { "epoch": 4.23, "learning_rate": 5.7668783815572815e-05, "loss": 2.5572, "step": 53985 }, { "epoch": 4.23, "learning_rate": 5.766799968634832e-05, "loss": 2.5445, "step": 53986 }, { "epoch": 4.23, "learning_rate": 5.766721555712382e-05, "loss": 3.039, "step": 53987 }, { "epoch": 4.23, "learning_rate": 5.7666431427899326e-05, "loss": 2.5185, "step": 53988 }, { "epoch": 4.23, "learning_rate": 5.766564729867483e-05, "loss": 2.6632, "step": 53989 }, { "epoch": 4.23, "learning_rate": 5.766486316945032e-05, "loss": 2.6185, "step": 53990 }, { "epoch": 4.23, "learning_rate": 5.766407904022583e-05, "loss": 2.5124, "step": 53991 }, { "epoch": 4.23, "learning_rate": 5.7663294911001334e-05, "loss": 2.8017, "step": 53992 }, { "epoch": 4.23, "learning_rate": 5.766251078177684e-05, "loss": 2.6171, "step": 53993 }, { "epoch": 4.23, "learning_rate": 5.766172665255234e-05, "loss": 2.646, "step": 53994 }, { "epoch": 4.23, "learning_rate": 5.7660942523327845e-05, "loss": 2.7939, "step": 53995 }, { "epoch": 4.23, "learning_rate": 5.766015839410335e-05, "loss": 2.6432, "step": 53996 }, { "epoch": 4.23, "learning_rate": 5.765937426487885e-05, "loss": 2.6359, "step": 53997 }, { "epoch": 4.23, "learning_rate": 5.7658590135654356e-05, "loss": 2.7914, "step": 53998 }, { "epoch": 4.23, "learning_rate": 5.765780600642986e-05, "loss": 2.9024, "step": 53999 }, { "epoch": 4.23, "learning_rate": 5.765702187720536e-05, "loss": 2.8282, "step": 54000 }, { "epoch": 4.23, "eval_loss": 2.8997983932495117, "eval_runtime": 217.9763, "eval_samples_per_second": 51.987, "eval_steps_per_second": 3.253, "step": 54000 }, { "epoch": 4.23, "learning_rate": 5.765623774798087e-05, "loss": 2.5572, "step": 54001 }, { "epoch": 4.23, "learning_rate": 5.765545361875637e-05, "loss": 2.6508, "step": 54002 }, { "epoch": 4.23, "learning_rate": 5.7654669489531874e-05, "loss": 2.6806, "step": 54003 }, { "epoch": 4.23, "learning_rate": 5.765388536030738e-05, "loss": 2.7488, "step": 54004 }, { "epoch": 4.23, "learning_rate": 5.765310123108288e-05, "loss": 2.7675, "step": 54005 }, { "epoch": 4.23, "learning_rate": 5.7652317101858385e-05, "loss": 2.5157, "step": 54006 }, { "epoch": 4.23, "learning_rate": 5.7651532972633896e-05, "loss": 2.5807, "step": 54007 }, { "epoch": 4.23, "learning_rate": 5.76507488434094e-05, "loss": 2.7409, "step": 54008 }, { "epoch": 4.24, "learning_rate": 5.76499647141849e-05, "loss": 2.8087, "step": 54009 }, { "epoch": 4.24, "learning_rate": 5.764918058496041e-05, "loss": 2.7374, "step": 54010 }, { "epoch": 4.24, "learning_rate": 5.764839645573591e-05, "loss": 2.466, "step": 54011 }, { "epoch": 4.24, "learning_rate": 5.7647612326511414e-05, "loss": 2.4505, "step": 54012 }, { "epoch": 4.24, "learning_rate": 5.764682819728692e-05, "loss": 2.4358, "step": 54013 }, { "epoch": 4.24, "learning_rate": 5.764604406806242e-05, "loss": 2.4594, "step": 54014 }, { "epoch": 4.24, "learning_rate": 5.7645259938837925e-05, "loss": 2.6922, "step": 54015 }, { "epoch": 4.24, "learning_rate": 5.764447580961343e-05, "loss": 2.829, "step": 54016 }, { "epoch": 4.24, "learning_rate": 5.764369168038893e-05, "loss": 2.8626, "step": 54017 }, { "epoch": 4.24, "learning_rate": 5.7642907551164436e-05, "loss": 2.6649, "step": 54018 }, { "epoch": 4.24, "learning_rate": 5.764212342193994e-05, "loss": 2.7216, "step": 54019 }, { "epoch": 4.24, "learning_rate": 5.7641339292715444e-05, "loss": 2.6821, "step": 54020 }, { "epoch": 4.24, "learning_rate": 5.764055516349095e-05, "loss": 2.5684, "step": 54021 }, { "epoch": 4.24, "learning_rate": 5.763977103426646e-05, "loss": 2.8025, "step": 54022 }, { "epoch": 4.24, "learning_rate": 5.763898690504196e-05, "loss": 2.6129, "step": 54023 }, { "epoch": 4.24, "learning_rate": 5.7638202775817465e-05, "loss": 2.5648, "step": 54024 }, { "epoch": 4.24, "learning_rate": 5.763741864659297e-05, "loss": 2.7627, "step": 54025 }, { "epoch": 4.24, "learning_rate": 5.763663451736846e-05, "loss": 2.7731, "step": 54026 }, { "epoch": 4.24, "learning_rate": 5.763585038814396e-05, "loss": 2.7193, "step": 54027 }, { "epoch": 4.24, "learning_rate": 5.7635066258919466e-05, "loss": 2.7394, "step": 54028 }, { "epoch": 4.24, "learning_rate": 5.763428212969497e-05, "loss": 2.7378, "step": 54029 }, { "epoch": 4.24, "learning_rate": 5.7633498000470474e-05, "loss": 2.7557, "step": 54030 }, { "epoch": 4.24, "learning_rate": 5.763271387124598e-05, "loss": 2.7784, "step": 54031 }, { "epoch": 4.24, "learning_rate": 5.763192974202149e-05, "loss": 2.7155, "step": 54032 }, { "epoch": 4.24, "learning_rate": 5.763114561279699e-05, "loss": 2.5684, "step": 54033 }, { "epoch": 4.24, "learning_rate": 5.7630361483572495e-05, "loss": 2.859, "step": 54034 }, { "epoch": 4.24, "learning_rate": 5.7629577354348e-05, "loss": 2.6528, "step": 54035 }, { "epoch": 4.24, "learning_rate": 5.76287932251235e-05, "loss": 2.8743, "step": 54036 }, { "epoch": 4.24, "learning_rate": 5.7628009095899006e-05, "loss": 2.6166, "step": 54037 }, { "epoch": 4.24, "learning_rate": 5.762722496667451e-05, "loss": 2.6041, "step": 54038 }, { "epoch": 4.24, "learning_rate": 5.762644083745001e-05, "loss": 2.5981, "step": 54039 }, { "epoch": 4.24, "learning_rate": 5.762565670822552e-05, "loss": 2.7469, "step": 54040 }, { "epoch": 4.24, "learning_rate": 5.762487257900102e-05, "loss": 2.7108, "step": 54041 }, { "epoch": 4.24, "learning_rate": 5.7624088449776524e-05, "loss": 2.7671, "step": 54042 }, { "epoch": 4.24, "learning_rate": 5.762330432055203e-05, "loss": 2.7674, "step": 54043 }, { "epoch": 4.24, "learning_rate": 5.762252019132753e-05, "loss": 2.7008, "step": 54044 }, { "epoch": 4.24, "learning_rate": 5.7621736062103035e-05, "loss": 2.5651, "step": 54045 }, { "epoch": 4.24, "learning_rate": 5.762095193287854e-05, "loss": 2.5692, "step": 54046 }, { "epoch": 4.24, "learning_rate": 5.762016780365404e-05, "loss": 2.7906, "step": 54047 }, { "epoch": 4.24, "learning_rate": 5.761938367442955e-05, "loss": 2.6512, "step": 54048 }, { "epoch": 4.24, "learning_rate": 5.761859954520506e-05, "loss": 2.5449, "step": 54049 }, { "epoch": 4.24, "learning_rate": 5.761781541598056e-05, "loss": 2.5271, "step": 54050 }, { "epoch": 4.24, "learning_rate": 5.7617031286756064e-05, "loss": 2.5886, "step": 54051 }, { "epoch": 4.24, "learning_rate": 5.761624715753157e-05, "loss": 2.6353, "step": 54052 }, { "epoch": 4.24, "learning_rate": 5.761546302830707e-05, "loss": 2.8568, "step": 54053 }, { "epoch": 4.24, "learning_rate": 5.7614678899082575e-05, "loss": 2.7772, "step": 54054 }, { "epoch": 4.24, "learning_rate": 5.761389476985808e-05, "loss": 2.7101, "step": 54055 }, { "epoch": 4.24, "learning_rate": 5.761311064063358e-05, "loss": 2.8514, "step": 54056 }, { "epoch": 4.24, "learning_rate": 5.7612326511409086e-05, "loss": 2.8324, "step": 54057 }, { "epoch": 4.24, "learning_rate": 5.761154238218459e-05, "loss": 2.7077, "step": 54058 }, { "epoch": 4.24, "learning_rate": 5.7610758252960094e-05, "loss": 2.6986, "step": 54059 }, { "epoch": 4.24, "learning_rate": 5.76099741237356e-05, "loss": 2.7154, "step": 54060 }, { "epoch": 4.24, "learning_rate": 5.76091899945111e-05, "loss": 2.6407, "step": 54061 }, { "epoch": 4.24, "learning_rate": 5.76084058652866e-05, "loss": 2.6953, "step": 54062 }, { "epoch": 4.24, "learning_rate": 5.76076217360621e-05, "loss": 2.5908, "step": 54063 }, { "epoch": 4.24, "learning_rate": 5.7606837606837605e-05, "loss": 2.6803, "step": 54064 }, { "epoch": 4.24, "learning_rate": 5.760605347761311e-05, "loss": 2.4582, "step": 54065 }, { "epoch": 4.24, "learning_rate": 5.760526934838861e-05, "loss": 2.8387, "step": 54066 }, { "epoch": 4.24, "learning_rate": 5.7604485219164116e-05, "loss": 2.5862, "step": 54067 }, { "epoch": 4.24, "learning_rate": 5.760370108993962e-05, "loss": 2.7224, "step": 54068 }, { "epoch": 4.24, "learning_rate": 5.7602916960715124e-05, "loss": 2.6429, "step": 54069 }, { "epoch": 4.24, "learning_rate": 5.760213283149063e-05, "loss": 2.6015, "step": 54070 }, { "epoch": 4.24, "learning_rate": 5.760134870226613e-05, "loss": 2.8699, "step": 54071 }, { "epoch": 4.24, "learning_rate": 5.7600564573041635e-05, "loss": 2.7826, "step": 54072 }, { "epoch": 4.24, "learning_rate": 5.759978044381714e-05, "loss": 2.5338, "step": 54073 }, { "epoch": 4.24, "learning_rate": 5.759899631459265e-05, "loss": 2.5411, "step": 54074 }, { "epoch": 4.24, "learning_rate": 5.759821218536815e-05, "loss": 2.5612, "step": 54075 }, { "epoch": 4.24, "learning_rate": 5.7597428056143656e-05, "loss": 2.6332, "step": 54076 }, { "epoch": 4.24, "learning_rate": 5.759664392691916e-05, "loss": 2.6393, "step": 54077 }, { "epoch": 4.24, "learning_rate": 5.7595859797694664e-05, "loss": 2.6222, "step": 54078 }, { "epoch": 4.24, "learning_rate": 5.759507566847017e-05, "loss": 2.8459, "step": 54079 }, { "epoch": 4.24, "learning_rate": 5.759429153924567e-05, "loss": 2.7141, "step": 54080 }, { "epoch": 4.24, "learning_rate": 5.7593507410021175e-05, "loss": 2.5954, "step": 54081 }, { "epoch": 4.24, "learning_rate": 5.759272328079668e-05, "loss": 2.636, "step": 54082 }, { "epoch": 4.24, "learning_rate": 5.759193915157218e-05, "loss": 2.5099, "step": 54083 }, { "epoch": 4.24, "learning_rate": 5.7591155022347686e-05, "loss": 2.5593, "step": 54084 }, { "epoch": 4.24, "learning_rate": 5.759037089312319e-05, "loss": 2.6575, "step": 54085 }, { "epoch": 4.24, "learning_rate": 5.758958676389869e-05, "loss": 2.5745, "step": 54086 }, { "epoch": 4.24, "learning_rate": 5.75888026346742e-05, "loss": 2.4304, "step": 54087 }, { "epoch": 4.24, "learning_rate": 5.75880185054497e-05, "loss": 2.6634, "step": 54088 }, { "epoch": 4.24, "learning_rate": 5.758723437622521e-05, "loss": 2.5811, "step": 54089 }, { "epoch": 4.24, "learning_rate": 5.7586450247000714e-05, "loss": 2.5961, "step": 54090 }, { "epoch": 4.24, "learning_rate": 5.758566611777622e-05, "loss": 2.7797, "step": 54091 }, { "epoch": 4.24, "learning_rate": 5.758488198855172e-05, "loss": 2.6245, "step": 54092 }, { "epoch": 4.24, "learning_rate": 5.7584097859327226e-05, "loss": 2.5775, "step": 54093 }, { "epoch": 4.24, "learning_rate": 5.758331373010273e-05, "loss": 2.7703, "step": 54094 }, { "epoch": 4.24, "learning_rate": 5.758252960087823e-05, "loss": 2.5908, "step": 54095 }, { "epoch": 4.24, "learning_rate": 5.7581745471653737e-05, "loss": 2.7225, "step": 54096 }, { "epoch": 4.24, "learning_rate": 5.758096134242924e-05, "loss": 2.584, "step": 54097 }, { "epoch": 4.24, "learning_rate": 5.758017721320473e-05, "loss": 2.8365, "step": 54098 }, { "epoch": 4.24, "learning_rate": 5.757939308398024e-05, "loss": 2.7597, "step": 54099 }, { "epoch": 4.24, "learning_rate": 5.7578608954755744e-05, "loss": 2.6501, "step": 54100 }, { "epoch": 4.24, "learning_rate": 5.757782482553125e-05, "loss": 2.5341, "step": 54101 }, { "epoch": 4.24, "learning_rate": 5.757704069630675e-05, "loss": 2.7033, "step": 54102 }, { "epoch": 4.24, "learning_rate": 5.7576256567082256e-05, "loss": 2.8123, "step": 54103 }, { "epoch": 4.24, "learning_rate": 5.757547243785776e-05, "loss": 2.501, "step": 54104 }, { "epoch": 4.24, "learning_rate": 5.757468830863326e-05, "loss": 2.633, "step": 54105 }, { "epoch": 4.24, "learning_rate": 5.7573904179408767e-05, "loss": 2.7825, "step": 54106 }, { "epoch": 4.24, "learning_rate": 5.757312005018427e-05, "loss": 2.6515, "step": 54107 }, { "epoch": 4.24, "learning_rate": 5.7572335920959774e-05, "loss": 2.9718, "step": 54108 }, { "epoch": 4.24, "learning_rate": 5.757155179173528e-05, "loss": 2.6199, "step": 54109 }, { "epoch": 4.24, "learning_rate": 5.757076766251078e-05, "loss": 2.6587, "step": 54110 }, { "epoch": 4.24, "learning_rate": 5.7569983533286285e-05, "loss": 2.6191, "step": 54111 }, { "epoch": 4.24, "learning_rate": 5.756919940406179e-05, "loss": 2.6391, "step": 54112 }, { "epoch": 4.24, "learning_rate": 5.756841527483729e-05, "loss": 2.5974, "step": 54113 }, { "epoch": 4.24, "learning_rate": 5.7567631145612796e-05, "loss": 2.6097, "step": 54114 }, { "epoch": 4.24, "learning_rate": 5.7566847016388306e-05, "loss": 2.5887, "step": 54115 }, { "epoch": 4.24, "learning_rate": 5.756606288716381e-05, "loss": 2.7364, "step": 54116 }, { "epoch": 4.24, "learning_rate": 5.7565278757939314e-05, "loss": 2.7402, "step": 54117 }, { "epoch": 4.24, "learning_rate": 5.756449462871482e-05, "loss": 2.4523, "step": 54118 }, { "epoch": 4.24, "learning_rate": 5.756371049949032e-05, "loss": 2.6121, "step": 54119 }, { "epoch": 4.24, "learning_rate": 5.7562926370265825e-05, "loss": 2.5479, "step": 54120 }, { "epoch": 4.24, "learning_rate": 5.756214224104133e-05, "loss": 2.7789, "step": 54121 }, { "epoch": 4.24, "learning_rate": 5.756135811181683e-05, "loss": 2.5026, "step": 54122 }, { "epoch": 4.24, "learning_rate": 5.7560573982592336e-05, "loss": 2.5458, "step": 54123 }, { "epoch": 4.24, "learning_rate": 5.755978985336784e-05, "loss": 2.6877, "step": 54124 }, { "epoch": 4.24, "learning_rate": 5.755900572414334e-05, "loss": 2.5913, "step": 54125 }, { "epoch": 4.24, "learning_rate": 5.755822159491885e-05, "loss": 2.7384, "step": 54126 }, { "epoch": 4.24, "learning_rate": 5.755743746569435e-05, "loss": 2.9077, "step": 54127 }, { "epoch": 4.24, "learning_rate": 5.7556653336469854e-05, "loss": 2.7891, "step": 54128 }, { "epoch": 4.24, "learning_rate": 5.755586920724536e-05, "loss": 2.6346, "step": 54129 }, { "epoch": 4.24, "learning_rate": 5.755508507802087e-05, "loss": 2.7301, "step": 54130 }, { "epoch": 4.24, "learning_rate": 5.755430094879637e-05, "loss": 2.5801, "step": 54131 }, { "epoch": 4.24, "learning_rate": 5.7553516819571876e-05, "loss": 2.5821, "step": 54132 }, { "epoch": 4.24, "learning_rate": 5.755273269034738e-05, "loss": 2.5644, "step": 54133 }, { "epoch": 4.24, "learning_rate": 5.755194856112287e-05, "loss": 2.8429, "step": 54134 }, { "epoch": 4.24, "learning_rate": 5.755116443189837e-05, "loss": 2.7988, "step": 54135 }, { "epoch": 4.24, "learning_rate": 5.755038030267388e-05, "loss": 2.7198, "step": 54136 }, { "epoch": 4.25, "learning_rate": 5.754959617344938e-05, "loss": 2.7897, "step": 54137 }, { "epoch": 4.25, "learning_rate": 5.7548812044224884e-05, "loss": 2.5862, "step": 54138 }, { "epoch": 4.25, "learning_rate": 5.754802791500039e-05, "loss": 2.6546, "step": 54139 }, { "epoch": 4.25, "learning_rate": 5.75472437857759e-05, "loss": 2.7196, "step": 54140 }, { "epoch": 4.25, "learning_rate": 5.75464596565514e-05, "loss": 2.6473, "step": 54141 }, { "epoch": 4.25, "learning_rate": 5.7545675527326906e-05, "loss": 2.6963, "step": 54142 }, { "epoch": 4.25, "learning_rate": 5.754489139810241e-05, "loss": 2.5057, "step": 54143 }, { "epoch": 4.25, "learning_rate": 5.754410726887791e-05, "loss": 2.6949, "step": 54144 }, { "epoch": 4.25, "learning_rate": 5.754332313965342e-05, "loss": 2.6785, "step": 54145 }, { "epoch": 4.25, "learning_rate": 5.754253901042892e-05, "loss": 2.4727, "step": 54146 }, { "epoch": 4.25, "learning_rate": 5.7541754881204424e-05, "loss": 2.5106, "step": 54147 }, { "epoch": 4.25, "learning_rate": 5.754097075197993e-05, "loss": 2.7703, "step": 54148 }, { "epoch": 4.25, "learning_rate": 5.754018662275543e-05, "loss": 2.5869, "step": 54149 }, { "epoch": 4.25, "learning_rate": 5.7539402493530935e-05, "loss": 2.5753, "step": 54150 }, { "epoch": 4.25, "learning_rate": 5.753861836430644e-05, "loss": 2.7045, "step": 54151 }, { "epoch": 4.25, "learning_rate": 5.753783423508194e-05, "loss": 2.5486, "step": 54152 }, { "epoch": 4.25, "learning_rate": 5.7537050105857446e-05, "loss": 2.5188, "step": 54153 }, { "epoch": 4.25, "learning_rate": 5.753626597663295e-05, "loss": 2.5552, "step": 54154 }, { "epoch": 4.25, "learning_rate": 5.7535481847408454e-05, "loss": 2.6542, "step": 54155 }, { "epoch": 4.25, "learning_rate": 5.7534697718183964e-05, "loss": 2.8669, "step": 54156 }, { "epoch": 4.25, "learning_rate": 5.753391358895947e-05, "loss": 2.77, "step": 54157 }, { "epoch": 4.25, "learning_rate": 5.753312945973497e-05, "loss": 2.7219, "step": 54158 }, { "epoch": 4.25, "learning_rate": 5.7532345330510475e-05, "loss": 2.9665, "step": 54159 }, { "epoch": 4.25, "learning_rate": 5.753156120128598e-05, "loss": 2.6747, "step": 54160 }, { "epoch": 4.25, "learning_rate": 5.753077707206148e-05, "loss": 2.596, "step": 54161 }, { "epoch": 4.25, "learning_rate": 5.7529992942836986e-05, "loss": 2.6734, "step": 54162 }, { "epoch": 4.25, "learning_rate": 5.752920881361249e-05, "loss": 2.6748, "step": 54163 }, { "epoch": 4.25, "learning_rate": 5.7528424684387993e-05, "loss": 2.849, "step": 54164 }, { "epoch": 4.25, "learning_rate": 5.75276405551635e-05, "loss": 2.6822, "step": 54165 }, { "epoch": 4.25, "learning_rate": 5.7526856425939e-05, "loss": 2.5332, "step": 54166 }, { "epoch": 4.25, "learning_rate": 5.7526072296714504e-05, "loss": 2.7249, "step": 54167 }, { "epoch": 4.25, "learning_rate": 5.752528816749001e-05, "loss": 2.784, "step": 54168 }, { "epoch": 4.25, "learning_rate": 5.752450403826551e-05, "loss": 2.5676, "step": 54169 }, { "epoch": 4.25, "learning_rate": 5.752371990904101e-05, "loss": 2.4905, "step": 54170 }, { "epoch": 4.25, "learning_rate": 5.752293577981651e-05, "loss": 2.656, "step": 54171 }, { "epoch": 4.25, "learning_rate": 5.7522151650592016e-05, "loss": 2.5669, "step": 54172 }, { "epoch": 4.25, "learning_rate": 5.752136752136752e-05, "loss": 2.655, "step": 54173 }, { "epoch": 4.25, "learning_rate": 5.7520583392143023e-05, "loss": 2.6671, "step": 54174 }, { "epoch": 4.25, "learning_rate": 5.751979926291853e-05, "loss": 2.4856, "step": 54175 }, { "epoch": 4.25, "learning_rate": 5.751901513369403e-05, "loss": 2.8861, "step": 54176 }, { "epoch": 4.25, "learning_rate": 5.7518231004469534e-05, "loss": 2.6414, "step": 54177 }, { "epoch": 4.25, "learning_rate": 5.751744687524504e-05, "loss": 2.5512, "step": 54178 }, { "epoch": 4.25, "learning_rate": 5.751666274602054e-05, "loss": 2.4868, "step": 54179 }, { "epoch": 4.25, "learning_rate": 5.7515878616796045e-05, "loss": 2.6121, "step": 54180 }, { "epoch": 4.25, "learning_rate": 5.751509448757155e-05, "loss": 2.8001, "step": 54181 }, { "epoch": 4.25, "learning_rate": 5.751431035834706e-05, "loss": 2.5729, "step": 54182 }, { "epoch": 4.25, "learning_rate": 5.751352622912256e-05, "loss": 2.8781, "step": 54183 }, { "epoch": 4.25, "learning_rate": 5.751274209989807e-05, "loss": 2.6579, "step": 54184 }, { "epoch": 4.25, "learning_rate": 5.751195797067357e-05, "loss": 2.6638, "step": 54185 }, { "epoch": 4.25, "learning_rate": 5.7511173841449074e-05, "loss": 2.675, "step": 54186 }, { "epoch": 4.25, "learning_rate": 5.751038971222458e-05, "loss": 2.5299, "step": 54187 }, { "epoch": 4.25, "learning_rate": 5.750960558300008e-05, "loss": 2.7493, "step": 54188 }, { "epoch": 4.25, "learning_rate": 5.7508821453775585e-05, "loss": 2.5809, "step": 54189 }, { "epoch": 4.25, "learning_rate": 5.750803732455109e-05, "loss": 2.6819, "step": 54190 }, { "epoch": 4.25, "learning_rate": 5.750725319532659e-05, "loss": 2.7988, "step": 54191 }, { "epoch": 4.25, "learning_rate": 5.7506469066102096e-05, "loss": 2.6314, "step": 54192 }, { "epoch": 4.25, "learning_rate": 5.75056849368776e-05, "loss": 2.7783, "step": 54193 }, { "epoch": 4.25, "learning_rate": 5.7504900807653104e-05, "loss": 2.4956, "step": 54194 }, { "epoch": 4.25, "learning_rate": 5.750411667842861e-05, "loss": 2.9022, "step": 54195 }, { "epoch": 4.25, "learning_rate": 5.750333254920411e-05, "loss": 2.5875, "step": 54196 }, { "epoch": 4.25, "learning_rate": 5.750254841997962e-05, "loss": 2.7698, "step": 54197 }, { "epoch": 4.25, "learning_rate": 5.7501764290755125e-05, "loss": 2.5329, "step": 54198 }, { "epoch": 4.25, "learning_rate": 5.750098016153063e-05, "loss": 2.6922, "step": 54199 }, { "epoch": 4.25, "learning_rate": 5.750019603230613e-05, "loss": 2.7058, "step": 54200 }, { "epoch": 4.25, "learning_rate": 5.7499411903081636e-05, "loss": 2.5747, "step": 54201 }, { "epoch": 4.25, "learning_rate": 5.749862777385714e-05, "loss": 2.4698, "step": 54202 }, { "epoch": 4.25, "learning_rate": 5.7497843644632644e-05, "loss": 2.6887, "step": 54203 }, { "epoch": 4.25, "learning_rate": 5.749705951540815e-05, "loss": 2.8076, "step": 54204 }, { "epoch": 4.25, "learning_rate": 5.749627538618365e-05, "loss": 2.5264, "step": 54205 }, { "epoch": 4.25, "learning_rate": 5.749549125695914e-05, "loss": 2.6855, "step": 54206 }, { "epoch": 4.25, "learning_rate": 5.749470712773465e-05, "loss": 2.7106, "step": 54207 }, { "epoch": 4.25, "learning_rate": 5.7493922998510155e-05, "loss": 2.5572, "step": 54208 }, { "epoch": 4.25, "learning_rate": 5.749313886928566e-05, "loss": 2.6029, "step": 54209 }, { "epoch": 4.25, "learning_rate": 5.749235474006116e-05, "loss": 2.6525, "step": 54210 }, { "epoch": 4.25, "learning_rate": 5.7491570610836666e-05, "loss": 2.554, "step": 54211 }, { "epoch": 4.25, "learning_rate": 5.749078648161217e-05, "loss": 2.5187, "step": 54212 }, { "epoch": 4.25, "learning_rate": 5.7490002352387674e-05, "loss": 2.6821, "step": 54213 }, { "epoch": 4.25, "learning_rate": 5.748921822316318e-05, "loss": 2.7101, "step": 54214 }, { "epoch": 4.25, "learning_rate": 5.748843409393868e-05, "loss": 2.805, "step": 54215 }, { "epoch": 4.25, "learning_rate": 5.7487649964714185e-05, "loss": 2.624, "step": 54216 }, { "epoch": 4.25, "learning_rate": 5.748686583548969e-05, "loss": 2.7447, "step": 54217 }, { "epoch": 4.25, "learning_rate": 5.748608170626519e-05, "loss": 2.6378, "step": 54218 }, { "epoch": 4.25, "learning_rate": 5.7485297577040696e-05, "loss": 2.5428, "step": 54219 }, { "epoch": 4.25, "learning_rate": 5.74845134478162e-05, "loss": 2.5818, "step": 54220 }, { "epoch": 4.25, "learning_rate": 5.74837293185917e-05, "loss": 2.8573, "step": 54221 }, { "epoch": 4.25, "learning_rate": 5.748294518936721e-05, "loss": 2.6547, "step": 54222 }, { "epoch": 4.25, "learning_rate": 5.748216106014272e-05, "loss": 2.7733, "step": 54223 }, { "epoch": 4.25, "learning_rate": 5.748137693091822e-05, "loss": 2.5721, "step": 54224 }, { "epoch": 4.25, "learning_rate": 5.7480592801693725e-05, "loss": 2.6424, "step": 54225 }, { "epoch": 4.25, "learning_rate": 5.747980867246923e-05, "loss": 2.7384, "step": 54226 }, { "epoch": 4.25, "learning_rate": 5.747902454324473e-05, "loss": 2.8959, "step": 54227 }, { "epoch": 4.25, "learning_rate": 5.7478240414020236e-05, "loss": 2.7146, "step": 54228 }, { "epoch": 4.25, "learning_rate": 5.747745628479574e-05, "loss": 2.8745, "step": 54229 }, { "epoch": 4.25, "learning_rate": 5.747667215557124e-05, "loss": 2.7203, "step": 54230 }, { "epoch": 4.25, "learning_rate": 5.7475888026346747e-05, "loss": 2.6884, "step": 54231 }, { "epoch": 4.25, "learning_rate": 5.747510389712225e-05, "loss": 2.8277, "step": 54232 }, { "epoch": 4.25, "learning_rate": 5.7474319767897754e-05, "loss": 2.5719, "step": 54233 }, { "epoch": 4.25, "learning_rate": 5.747353563867326e-05, "loss": 2.9586, "step": 54234 }, { "epoch": 4.25, "learning_rate": 5.747275150944876e-05, "loss": 2.5179, "step": 54235 }, { "epoch": 4.25, "learning_rate": 5.7471967380224265e-05, "loss": 2.816, "step": 54236 }, { "epoch": 4.25, "learning_rate": 5.747118325099977e-05, "loss": 2.5695, "step": 54237 }, { "epoch": 4.25, "learning_rate": 5.747039912177528e-05, "loss": 2.8549, "step": 54238 }, { "epoch": 4.25, "learning_rate": 5.746961499255078e-05, "loss": 2.7172, "step": 54239 }, { "epoch": 4.25, "learning_rate": 5.7468830863326286e-05, "loss": 2.8444, "step": 54240 }, { "epoch": 4.25, "learning_rate": 5.746804673410179e-05, "loss": 2.5566, "step": 54241 }, { "epoch": 4.25, "learning_rate": 5.746726260487728e-05, "loss": 2.956, "step": 54242 }, { "epoch": 4.25, "learning_rate": 5.7466478475652784e-05, "loss": 2.6041, "step": 54243 }, { "epoch": 4.25, "learning_rate": 5.746569434642829e-05, "loss": 2.826, "step": 54244 }, { "epoch": 4.25, "learning_rate": 5.746491021720379e-05, "loss": 2.5044, "step": 54245 }, { "epoch": 4.25, "learning_rate": 5.7464126087979295e-05, "loss": 2.7644, "step": 54246 }, { "epoch": 4.25, "learning_rate": 5.74633419587548e-05, "loss": 2.7688, "step": 54247 }, { "epoch": 4.25, "learning_rate": 5.746255782953031e-05, "loss": 2.8191, "step": 54248 }, { "epoch": 4.25, "learning_rate": 5.746177370030581e-05, "loss": 2.9223, "step": 54249 }, { "epoch": 4.25, "learning_rate": 5.7460989571081316e-05, "loss": 2.7515, "step": 54250 }, { "epoch": 4.25, "learning_rate": 5.746020544185682e-05, "loss": 2.7647, "step": 54251 }, { "epoch": 4.25, "learning_rate": 5.7459421312632324e-05, "loss": 2.561, "step": 54252 }, { "epoch": 4.25, "learning_rate": 5.745863718340783e-05, "loss": 2.7737, "step": 54253 }, { "epoch": 4.25, "learning_rate": 5.745785305418333e-05, "loss": 2.5826, "step": 54254 }, { "epoch": 4.25, "learning_rate": 5.7457068924958835e-05, "loss": 2.649, "step": 54255 }, { "epoch": 4.25, "learning_rate": 5.745628479573434e-05, "loss": 2.6258, "step": 54256 }, { "epoch": 4.25, "learning_rate": 5.745550066650984e-05, "loss": 2.5821, "step": 54257 }, { "epoch": 4.25, "learning_rate": 5.7454716537285346e-05, "loss": 2.8486, "step": 54258 }, { "epoch": 4.25, "learning_rate": 5.745393240806085e-05, "loss": 2.6446, "step": 54259 }, { "epoch": 4.25, "learning_rate": 5.745314827883635e-05, "loss": 2.8984, "step": 54260 }, { "epoch": 4.25, "learning_rate": 5.745236414961186e-05, "loss": 2.4935, "step": 54261 }, { "epoch": 4.25, "learning_rate": 5.745158002038736e-05, "loss": 2.5925, "step": 54262 }, { "epoch": 4.25, "learning_rate": 5.7450795891162864e-05, "loss": 2.7149, "step": 54263 }, { "epoch": 4.25, "learning_rate": 5.7450011761938375e-05, "loss": 2.6458, "step": 54264 }, { "epoch": 4.26, "learning_rate": 5.744922763271388e-05, "loss": 2.8126, "step": 54265 }, { "epoch": 4.26, "learning_rate": 5.744844350348938e-05, "loss": 2.5166, "step": 54266 }, { "epoch": 4.26, "learning_rate": 5.7447659374264886e-05, "loss": 2.6862, "step": 54267 }, { "epoch": 4.26, "learning_rate": 5.744687524504039e-05, "loss": 2.6576, "step": 54268 }, { "epoch": 4.26, "learning_rate": 5.744609111581589e-05, "loss": 2.5663, "step": 54269 }, { "epoch": 4.26, "learning_rate": 5.74453069865914e-05, "loss": 2.5687, "step": 54270 }, { "epoch": 4.26, "learning_rate": 5.74445228573669e-05, "loss": 2.751, "step": 54271 }, { "epoch": 4.26, "learning_rate": 5.7443738728142404e-05, "loss": 2.5378, "step": 54272 }, { "epoch": 4.26, "learning_rate": 5.744295459891791e-05, "loss": 2.6085, "step": 54273 }, { "epoch": 4.26, "learning_rate": 5.744217046969341e-05, "loss": 2.6283, "step": 54274 }, { "epoch": 4.26, "learning_rate": 5.7441386340468915e-05, "loss": 2.6027, "step": 54275 }, { "epoch": 4.26, "learning_rate": 5.744060221124442e-05, "loss": 2.6232, "step": 54276 }, { "epoch": 4.26, "learning_rate": 5.743981808201992e-05, "loss": 2.7789, "step": 54277 }, { "epoch": 4.26, "learning_rate": 5.743903395279542e-05, "loss": 2.6464, "step": 54278 }, { "epoch": 4.26, "learning_rate": 5.743824982357092e-05, "loss": 2.5918, "step": 54279 }, { "epoch": 4.26, "learning_rate": 5.743746569434643e-05, "loss": 2.5226, "step": 54280 }, { "epoch": 4.26, "learning_rate": 5.743668156512193e-05, "loss": 2.6953, "step": 54281 }, { "epoch": 4.26, "learning_rate": 5.7435897435897434e-05, "loss": 2.4971, "step": 54282 }, { "epoch": 4.26, "learning_rate": 5.743511330667294e-05, "loss": 2.5872, "step": 54283 }, { "epoch": 4.26, "learning_rate": 5.743432917744844e-05, "loss": 2.7269, "step": 54284 }, { "epoch": 4.26, "learning_rate": 5.7433545048223945e-05, "loss": 2.72, "step": 54285 }, { "epoch": 4.26, "learning_rate": 5.743276091899945e-05, "loss": 2.8775, "step": 54286 }, { "epoch": 4.26, "learning_rate": 5.743197678977495e-05, "loss": 2.6615, "step": 54287 }, { "epoch": 4.26, "learning_rate": 5.7431192660550456e-05, "loss": 2.7295, "step": 54288 }, { "epoch": 4.26, "learning_rate": 5.743040853132596e-05, "loss": 2.7405, "step": 54289 }, { "epoch": 4.26, "learning_rate": 5.742962440210147e-05, "loss": 2.5794, "step": 54290 }, { "epoch": 4.26, "learning_rate": 5.7428840272876974e-05, "loss": 2.4544, "step": 54291 }, { "epoch": 4.26, "learning_rate": 5.742805614365248e-05, "loss": 2.6565, "step": 54292 }, { "epoch": 4.26, "learning_rate": 5.742727201442798e-05, "loss": 2.7998, "step": 54293 }, { "epoch": 4.26, "learning_rate": 5.7426487885203485e-05, "loss": 2.8994, "step": 54294 }, { "epoch": 4.26, "learning_rate": 5.742570375597899e-05, "loss": 2.8457, "step": 54295 }, { "epoch": 4.26, "learning_rate": 5.742491962675449e-05, "loss": 2.6792, "step": 54296 }, { "epoch": 4.26, "learning_rate": 5.7424135497529996e-05, "loss": 2.589, "step": 54297 }, { "epoch": 4.26, "learning_rate": 5.74233513683055e-05, "loss": 2.651, "step": 54298 }, { "epoch": 4.26, "learning_rate": 5.7422567239081003e-05, "loss": 2.756, "step": 54299 }, { "epoch": 4.26, "learning_rate": 5.742178310985651e-05, "loss": 2.49, "step": 54300 }, { "epoch": 4.26, "learning_rate": 5.742099898063201e-05, "loss": 2.7082, "step": 54301 }, { "epoch": 4.26, "learning_rate": 5.7420214851407514e-05, "loss": 2.4933, "step": 54302 }, { "epoch": 4.26, "learning_rate": 5.741943072218302e-05, "loss": 2.6847, "step": 54303 }, { "epoch": 4.26, "learning_rate": 5.741864659295852e-05, "loss": 2.6006, "step": 54304 }, { "epoch": 4.26, "learning_rate": 5.741786246373403e-05, "loss": 2.7182, "step": 54305 }, { "epoch": 4.26, "learning_rate": 5.7417078334509536e-05, "loss": 2.6315, "step": 54306 }, { "epoch": 4.26, "learning_rate": 5.741629420528504e-05, "loss": 2.5965, "step": 54307 }, { "epoch": 4.26, "learning_rate": 5.741551007606054e-05, "loss": 2.5352, "step": 54308 }, { "epoch": 4.26, "learning_rate": 5.741472594683605e-05, "loss": 2.5741, "step": 54309 }, { "epoch": 4.26, "learning_rate": 5.741394181761155e-05, "loss": 2.6132, "step": 54310 }, { "epoch": 4.26, "learning_rate": 5.7413157688387054e-05, "loss": 2.6279, "step": 54311 }, { "epoch": 4.26, "learning_rate": 5.741237355916256e-05, "loss": 2.7543, "step": 54312 }, { "epoch": 4.26, "learning_rate": 5.741158942993806e-05, "loss": 2.7107, "step": 54313 }, { "epoch": 4.26, "learning_rate": 5.741080530071355e-05, "loss": 2.6874, "step": 54314 }, { "epoch": 4.26, "learning_rate": 5.741002117148906e-05, "loss": 2.7111, "step": 54315 }, { "epoch": 4.26, "learning_rate": 5.7409237042264566e-05, "loss": 2.5628, "step": 54316 }, { "epoch": 4.26, "learning_rate": 5.740845291304007e-05, "loss": 2.5589, "step": 54317 }, { "epoch": 4.26, "learning_rate": 5.740766878381557e-05, "loss": 2.6038, "step": 54318 }, { "epoch": 4.26, "learning_rate": 5.740688465459108e-05, "loss": 2.7242, "step": 54319 }, { "epoch": 4.26, "learning_rate": 5.740610052536658e-05, "loss": 2.8049, "step": 54320 }, { "epoch": 4.26, "learning_rate": 5.7405316396142084e-05, "loss": 2.7068, "step": 54321 }, { "epoch": 4.26, "learning_rate": 5.740453226691759e-05, "loss": 2.6703, "step": 54322 }, { "epoch": 4.26, "learning_rate": 5.740374813769309e-05, "loss": 2.7729, "step": 54323 }, { "epoch": 4.26, "learning_rate": 5.7402964008468595e-05, "loss": 2.6104, "step": 54324 }, { "epoch": 4.26, "learning_rate": 5.74021798792441e-05, "loss": 2.496, "step": 54325 }, { "epoch": 4.26, "learning_rate": 5.74013957500196e-05, "loss": 2.5739, "step": 54326 }, { "epoch": 4.26, "learning_rate": 5.7400611620795106e-05, "loss": 2.8506, "step": 54327 }, { "epoch": 4.26, "learning_rate": 5.739982749157061e-05, "loss": 2.6181, "step": 54328 }, { "epoch": 4.26, "learning_rate": 5.7399043362346114e-05, "loss": 2.814, "step": 54329 }, { "epoch": 4.26, "learning_rate": 5.739825923312162e-05, "loss": 2.804, "step": 54330 }, { "epoch": 4.26, "learning_rate": 5.739747510389713e-05, "loss": 2.6563, "step": 54331 }, { "epoch": 4.26, "learning_rate": 5.739669097467263e-05, "loss": 2.8119, "step": 54332 }, { "epoch": 4.26, "learning_rate": 5.7395906845448135e-05, "loss": 2.8446, "step": 54333 }, { "epoch": 4.26, "learning_rate": 5.739512271622364e-05, "loss": 2.6427, "step": 54334 }, { "epoch": 4.26, "learning_rate": 5.739433858699914e-05, "loss": 2.6222, "step": 54335 }, { "epoch": 4.26, "learning_rate": 5.7393554457774646e-05, "loss": 2.842, "step": 54336 }, { "epoch": 4.26, "learning_rate": 5.739277032855015e-05, "loss": 2.6593, "step": 54337 }, { "epoch": 4.26, "learning_rate": 5.7391986199325654e-05, "loss": 2.6398, "step": 54338 }, { "epoch": 4.26, "learning_rate": 5.739120207010116e-05, "loss": 2.7912, "step": 54339 }, { "epoch": 4.26, "learning_rate": 5.739041794087666e-05, "loss": 2.6561, "step": 54340 }, { "epoch": 4.26, "learning_rate": 5.7389633811652165e-05, "loss": 2.6077, "step": 54341 }, { "epoch": 4.26, "learning_rate": 5.738884968242767e-05, "loss": 2.6387, "step": 54342 }, { "epoch": 4.26, "learning_rate": 5.738806555320317e-05, "loss": 2.5943, "step": 54343 }, { "epoch": 4.26, "learning_rate": 5.7387281423978676e-05, "loss": 2.7131, "step": 54344 }, { "epoch": 4.26, "learning_rate": 5.738649729475418e-05, "loss": 2.647, "step": 54345 }, { "epoch": 4.26, "learning_rate": 5.738571316552968e-05, "loss": 2.4441, "step": 54346 }, { "epoch": 4.26, "learning_rate": 5.7384929036305194e-05, "loss": 2.6892, "step": 54347 }, { "epoch": 4.26, "learning_rate": 5.73841449070807e-05, "loss": 2.5178, "step": 54348 }, { "epoch": 4.26, "learning_rate": 5.73833607778562e-05, "loss": 2.489, "step": 54349 }, { "epoch": 4.26, "learning_rate": 5.738257664863169e-05, "loss": 2.4368, "step": 54350 }, { "epoch": 4.26, "learning_rate": 5.7381792519407195e-05, "loss": 2.5114, "step": 54351 }, { "epoch": 4.26, "learning_rate": 5.73810083901827e-05, "loss": 2.4293, "step": 54352 }, { "epoch": 4.26, "learning_rate": 5.73802242609582e-05, "loss": 2.9157, "step": 54353 }, { "epoch": 4.26, "learning_rate": 5.7379440131733706e-05, "loss": 3.0155, "step": 54354 }, { "epoch": 4.26, "learning_rate": 5.737865600250921e-05, "loss": 2.7344, "step": 54355 }, { "epoch": 4.26, "learning_rate": 5.737787187328471e-05, "loss": 3.0809, "step": 54356 }, { "epoch": 4.26, "learning_rate": 5.7377087744060224e-05, "loss": 2.5828, "step": 54357 }, { "epoch": 4.26, "learning_rate": 5.737630361483573e-05, "loss": 2.9206, "step": 54358 }, { "epoch": 4.26, "learning_rate": 5.737551948561123e-05, "loss": 2.6432, "step": 54359 }, { "epoch": 4.26, "learning_rate": 5.7374735356386735e-05, "loss": 2.5226, "step": 54360 }, { "epoch": 4.26, "learning_rate": 5.737395122716224e-05, "loss": 2.5795, "step": 54361 }, { "epoch": 4.26, "learning_rate": 5.737316709793774e-05, "loss": 2.6722, "step": 54362 }, { "epoch": 4.26, "learning_rate": 5.7372382968713246e-05, "loss": 2.776, "step": 54363 }, { "epoch": 4.26, "learning_rate": 5.737159883948875e-05, "loss": 2.6232, "step": 54364 }, { "epoch": 4.26, "learning_rate": 5.737081471026425e-05, "loss": 2.6154, "step": 54365 }, { "epoch": 4.26, "learning_rate": 5.737003058103976e-05, "loss": 2.6129, "step": 54366 }, { "epoch": 4.26, "learning_rate": 5.736924645181526e-05, "loss": 2.6498, "step": 54367 }, { "epoch": 4.26, "learning_rate": 5.7368462322590764e-05, "loss": 2.8059, "step": 54368 }, { "epoch": 4.26, "learning_rate": 5.736767819336627e-05, "loss": 2.5678, "step": 54369 }, { "epoch": 4.26, "learning_rate": 5.736689406414177e-05, "loss": 2.6868, "step": 54370 }, { "epoch": 4.26, "learning_rate": 5.7366109934917275e-05, "loss": 2.4926, "step": 54371 }, { "epoch": 4.26, "learning_rate": 5.7365325805692785e-05, "loss": 2.4582, "step": 54372 }, { "epoch": 4.26, "learning_rate": 5.736454167646829e-05, "loss": 2.5939, "step": 54373 }, { "epoch": 4.26, "learning_rate": 5.736375754724379e-05, "loss": 2.7301, "step": 54374 }, { "epoch": 4.26, "learning_rate": 5.7362973418019297e-05, "loss": 2.7245, "step": 54375 }, { "epoch": 4.26, "learning_rate": 5.73621892887948e-05, "loss": 2.7087, "step": 54376 }, { "epoch": 4.26, "learning_rate": 5.7361405159570304e-05, "loss": 2.6923, "step": 54377 }, { "epoch": 4.26, "learning_rate": 5.736062103034581e-05, "loss": 2.6712, "step": 54378 }, { "epoch": 4.26, "learning_rate": 5.735983690112131e-05, "loss": 2.6974, "step": 54379 }, { "epoch": 4.26, "learning_rate": 5.7359052771896815e-05, "loss": 2.6266, "step": 54380 }, { "epoch": 4.26, "learning_rate": 5.735826864267232e-05, "loss": 2.6199, "step": 54381 }, { "epoch": 4.26, "learning_rate": 5.735748451344782e-05, "loss": 2.6643, "step": 54382 }, { "epoch": 4.26, "learning_rate": 5.7356700384223326e-05, "loss": 2.6917, "step": 54383 }, { "epoch": 4.26, "learning_rate": 5.735591625499883e-05, "loss": 2.8418, "step": 54384 }, { "epoch": 4.26, "learning_rate": 5.735513212577433e-05, "loss": 2.899, "step": 54385 }, { "epoch": 4.26, "learning_rate": 5.735434799654983e-05, "loss": 2.8704, "step": 54386 }, { "epoch": 4.26, "learning_rate": 5.7353563867325334e-05, "loss": 2.5772, "step": 54387 }, { "epoch": 4.26, "learning_rate": 5.735277973810084e-05, "loss": 2.6884, "step": 54388 }, { "epoch": 4.26, "learning_rate": 5.735199560887634e-05, "loss": 2.8087, "step": 54389 }, { "epoch": 4.26, "learning_rate": 5.7351211479651845e-05, "loss": 2.6689, "step": 54390 }, { "epoch": 4.26, "learning_rate": 5.735042735042735e-05, "loss": 2.6574, "step": 54391 }, { "epoch": 4.27, "learning_rate": 5.734964322120285e-05, "loss": 2.6847, "step": 54392 }, { "epoch": 4.27, "learning_rate": 5.7348859091978356e-05, "loss": 2.7564, "step": 54393 }, { "epoch": 4.27, "learning_rate": 5.734807496275386e-05, "loss": 2.4569, "step": 54394 }, { "epoch": 4.27, "learning_rate": 5.734729083352936e-05, "loss": 2.6113, "step": 54395 }, { "epoch": 4.27, "learning_rate": 5.734650670430487e-05, "loss": 2.7194, "step": 54396 }, { "epoch": 4.27, "learning_rate": 5.734572257508037e-05, "loss": 2.7173, "step": 54397 }, { "epoch": 4.27, "learning_rate": 5.734493844585588e-05, "loss": 2.6623, "step": 54398 }, { "epoch": 4.27, "learning_rate": 5.7344154316631385e-05, "loss": 2.6079, "step": 54399 }, { "epoch": 4.27, "learning_rate": 5.734337018740689e-05, "loss": 2.5996, "step": 54400 }, { "epoch": 4.27, "learning_rate": 5.734258605818239e-05, "loss": 2.68, "step": 54401 }, { "epoch": 4.27, "learning_rate": 5.7341801928957896e-05, "loss": 2.6205, "step": 54402 }, { "epoch": 4.27, "learning_rate": 5.73410177997334e-05, "loss": 2.6504, "step": 54403 }, { "epoch": 4.27, "learning_rate": 5.73402336705089e-05, "loss": 2.6632, "step": 54404 }, { "epoch": 4.27, "learning_rate": 5.733944954128441e-05, "loss": 2.8605, "step": 54405 }, { "epoch": 4.27, "learning_rate": 5.733866541205991e-05, "loss": 2.6075, "step": 54406 }, { "epoch": 4.27, "learning_rate": 5.7337881282835414e-05, "loss": 2.8134, "step": 54407 }, { "epoch": 4.27, "learning_rate": 5.733709715361092e-05, "loss": 2.5014, "step": 54408 }, { "epoch": 4.27, "learning_rate": 5.733631302438642e-05, "loss": 2.559, "step": 54409 }, { "epoch": 4.27, "learning_rate": 5.7335528895161925e-05, "loss": 2.6072, "step": 54410 }, { "epoch": 4.27, "learning_rate": 5.733474476593743e-05, "loss": 2.5585, "step": 54411 }, { "epoch": 4.27, "learning_rate": 5.733396063671293e-05, "loss": 2.5221, "step": 54412 }, { "epoch": 4.27, "learning_rate": 5.733317650748844e-05, "loss": 2.8385, "step": 54413 }, { "epoch": 4.27, "learning_rate": 5.733239237826395e-05, "loss": 2.7858, "step": 54414 }, { "epoch": 4.27, "learning_rate": 5.733160824903945e-05, "loss": 2.7149, "step": 54415 }, { "epoch": 4.27, "learning_rate": 5.7330824119814954e-05, "loss": 2.72, "step": 54416 }, { "epoch": 4.27, "learning_rate": 5.733003999059046e-05, "loss": 2.5702, "step": 54417 }, { "epoch": 4.27, "learning_rate": 5.732925586136596e-05, "loss": 2.6258, "step": 54418 }, { "epoch": 4.27, "learning_rate": 5.7328471732141465e-05, "loss": 2.6074, "step": 54419 }, { "epoch": 4.27, "learning_rate": 5.732768760291697e-05, "loss": 2.9231, "step": 54420 }, { "epoch": 4.27, "learning_rate": 5.732690347369247e-05, "loss": 2.6879, "step": 54421 }, { "epoch": 4.27, "learning_rate": 5.732611934446796e-05, "loss": 2.703, "step": 54422 }, { "epoch": 4.27, "learning_rate": 5.732533521524347e-05, "loss": 2.7404, "step": 54423 }, { "epoch": 4.27, "learning_rate": 5.732455108601898e-05, "loss": 2.6642, "step": 54424 }, { "epoch": 4.27, "learning_rate": 5.732376695679448e-05, "loss": 2.7571, "step": 54425 }, { "epoch": 4.27, "learning_rate": 5.7322982827569984e-05, "loss": 2.6735, "step": 54426 }, { "epoch": 4.27, "learning_rate": 5.732219869834549e-05, "loss": 2.8327, "step": 54427 }, { "epoch": 4.27, "learning_rate": 5.732141456912099e-05, "loss": 2.5655, "step": 54428 }, { "epoch": 4.27, "learning_rate": 5.7320630439896495e-05, "loss": 2.6767, "step": 54429 }, { "epoch": 4.27, "learning_rate": 5.7319846310672e-05, "loss": 2.6292, "step": 54430 }, { "epoch": 4.27, "learning_rate": 5.73190621814475e-05, "loss": 2.7486, "step": 54431 }, { "epoch": 4.27, "learning_rate": 5.7318278052223006e-05, "loss": 2.7313, "step": 54432 }, { "epoch": 4.27, "learning_rate": 5.731749392299851e-05, "loss": 2.7815, "step": 54433 }, { "epoch": 4.27, "learning_rate": 5.7316709793774014e-05, "loss": 2.4738, "step": 54434 }, { "epoch": 4.27, "learning_rate": 5.731592566454952e-05, "loss": 2.5187, "step": 54435 }, { "epoch": 4.27, "learning_rate": 5.731514153532502e-05, "loss": 2.8335, "step": 54436 }, { "epoch": 4.27, "learning_rate": 5.7314357406100525e-05, "loss": 2.798, "step": 54437 }, { "epoch": 4.27, "learning_rate": 5.731357327687603e-05, "loss": 2.7089, "step": 54438 }, { "epoch": 4.27, "learning_rate": 5.731278914765154e-05, "loss": 2.648, "step": 54439 }, { "epoch": 4.27, "learning_rate": 5.731200501842704e-05, "loss": 2.5592, "step": 54440 }, { "epoch": 4.27, "learning_rate": 5.7311220889202546e-05, "loss": 2.5854, "step": 54441 }, { "epoch": 4.27, "learning_rate": 5.731043675997805e-05, "loss": 2.6261, "step": 54442 }, { "epoch": 4.27, "learning_rate": 5.730965263075355e-05, "loss": 2.6877, "step": 54443 }, { "epoch": 4.27, "learning_rate": 5.730886850152906e-05, "loss": 2.7798, "step": 54444 }, { "epoch": 4.27, "learning_rate": 5.730808437230456e-05, "loss": 2.9035, "step": 54445 }, { "epoch": 4.27, "learning_rate": 5.7307300243080064e-05, "loss": 2.6504, "step": 54446 }, { "epoch": 4.27, "learning_rate": 5.730651611385557e-05, "loss": 2.5213, "step": 54447 }, { "epoch": 4.27, "learning_rate": 5.730573198463107e-05, "loss": 2.5431, "step": 54448 }, { "epoch": 4.27, "learning_rate": 5.7304947855406575e-05, "loss": 2.6293, "step": 54449 }, { "epoch": 4.27, "learning_rate": 5.730416372618208e-05, "loss": 3.1479, "step": 54450 }, { "epoch": 4.27, "learning_rate": 5.730337959695758e-05, "loss": 2.7581, "step": 54451 }, { "epoch": 4.27, "learning_rate": 5.7302595467733086e-05, "loss": 2.5501, "step": 54452 }, { "epoch": 4.27, "learning_rate": 5.730181133850859e-05, "loss": 2.5939, "step": 54453 }, { "epoch": 4.27, "learning_rate": 5.7301027209284094e-05, "loss": 2.6729, "step": 54454 }, { "epoch": 4.27, "learning_rate": 5.7300243080059604e-05, "loss": 2.5383, "step": 54455 }, { "epoch": 4.27, "learning_rate": 5.729945895083511e-05, "loss": 2.7908, "step": 54456 }, { "epoch": 4.27, "learning_rate": 5.729867482161061e-05, "loss": 2.5863, "step": 54457 }, { "epoch": 4.27, "learning_rate": 5.72978906923861e-05, "loss": 2.5528, "step": 54458 }, { "epoch": 4.27, "learning_rate": 5.7297106563161605e-05, "loss": 2.5198, "step": 54459 }, { "epoch": 4.27, "learning_rate": 5.729632243393711e-05, "loss": 2.6567, "step": 54460 }, { "epoch": 4.27, "learning_rate": 5.729553830471261e-05, "loss": 2.9175, "step": 54461 }, { "epoch": 4.27, "learning_rate": 5.7294754175488116e-05, "loss": 2.4471, "step": 54462 }, { "epoch": 4.27, "learning_rate": 5.729397004626362e-05, "loss": 2.6837, "step": 54463 }, { "epoch": 4.27, "learning_rate": 5.7293185917039124e-05, "loss": 2.7307, "step": 54464 }, { "epoch": 4.27, "learning_rate": 5.7292401787814634e-05, "loss": 2.7016, "step": 54465 }, { "epoch": 4.27, "learning_rate": 5.729161765859014e-05, "loss": 2.5315, "step": 54466 }, { "epoch": 4.27, "learning_rate": 5.729083352936564e-05, "loss": 2.4824, "step": 54467 }, { "epoch": 4.27, "learning_rate": 5.7290049400141145e-05, "loss": 2.6625, "step": 54468 }, { "epoch": 4.27, "learning_rate": 5.728926527091665e-05, "loss": 2.7419, "step": 54469 }, { "epoch": 4.27, "learning_rate": 5.728848114169215e-05, "loss": 2.7323, "step": 54470 }, { "epoch": 4.27, "learning_rate": 5.7287697012467656e-05, "loss": 2.4954, "step": 54471 }, { "epoch": 4.27, "learning_rate": 5.728691288324316e-05, "loss": 2.6273, "step": 54472 }, { "epoch": 4.27, "learning_rate": 5.7286128754018664e-05, "loss": 2.6425, "step": 54473 }, { "epoch": 4.27, "learning_rate": 5.728534462479417e-05, "loss": 2.6586, "step": 54474 }, { "epoch": 4.27, "learning_rate": 5.728456049556967e-05, "loss": 2.6817, "step": 54475 }, { "epoch": 4.27, "learning_rate": 5.7283776366345175e-05, "loss": 2.8803, "step": 54476 }, { "epoch": 4.27, "learning_rate": 5.728299223712068e-05, "loss": 2.6449, "step": 54477 }, { "epoch": 4.27, "learning_rate": 5.728220810789618e-05, "loss": 2.7626, "step": 54478 }, { "epoch": 4.27, "learning_rate": 5.7281423978671686e-05, "loss": 2.5658, "step": 54479 }, { "epoch": 4.27, "learning_rate": 5.7280639849447196e-05, "loss": 2.6511, "step": 54480 }, { "epoch": 4.27, "learning_rate": 5.72798557202227e-05, "loss": 2.7322, "step": 54481 }, { "epoch": 4.27, "learning_rate": 5.7279071590998204e-05, "loss": 2.6206, "step": 54482 }, { "epoch": 4.27, "learning_rate": 5.727828746177371e-05, "loss": 2.8487, "step": 54483 }, { "epoch": 4.27, "learning_rate": 5.727750333254921e-05, "loss": 2.7796, "step": 54484 }, { "epoch": 4.27, "learning_rate": 5.7276719203324715e-05, "loss": 2.4754, "step": 54485 }, { "epoch": 4.27, "learning_rate": 5.727593507410022e-05, "loss": 2.5961, "step": 54486 }, { "epoch": 4.27, "learning_rate": 5.727515094487572e-05, "loss": 2.5877, "step": 54487 }, { "epoch": 4.27, "learning_rate": 5.7274366815651226e-05, "loss": 2.7105, "step": 54488 }, { "epoch": 4.27, "learning_rate": 5.727358268642673e-05, "loss": 2.6405, "step": 54489 }, { "epoch": 4.27, "learning_rate": 5.727279855720223e-05, "loss": 2.7244, "step": 54490 }, { "epoch": 4.27, "learning_rate": 5.727201442797774e-05, "loss": 2.7199, "step": 54491 }, { "epoch": 4.27, "learning_rate": 5.727123029875324e-05, "loss": 2.6702, "step": 54492 }, { "epoch": 4.27, "learning_rate": 5.7270446169528744e-05, "loss": 2.6412, "step": 54493 }, { "epoch": 4.27, "learning_rate": 5.726966204030424e-05, "loss": 2.6747, "step": 54494 }, { "epoch": 4.27, "learning_rate": 5.7268877911079745e-05, "loss": 2.9276, "step": 54495 }, { "epoch": 4.27, "learning_rate": 5.726809378185525e-05, "loss": 2.7521, "step": 54496 }, { "epoch": 4.27, "learning_rate": 5.726730965263075e-05, "loss": 2.7779, "step": 54497 }, { "epoch": 4.27, "learning_rate": 5.7266525523406256e-05, "loss": 2.499, "step": 54498 }, { "epoch": 4.27, "learning_rate": 5.726574139418176e-05, "loss": 2.6126, "step": 54499 }, { "epoch": 4.27, "learning_rate": 5.726495726495726e-05, "loss": 2.7483, "step": 54500 }, { "epoch": 4.27, "learning_rate": 5.726417313573277e-05, "loss": 2.5416, "step": 54501 }, { "epoch": 4.27, "learning_rate": 5.726338900650827e-05, "loss": 2.839, "step": 54502 }, { "epoch": 4.27, "learning_rate": 5.7262604877283774e-05, "loss": 2.767, "step": 54503 }, { "epoch": 4.27, "learning_rate": 5.726182074805928e-05, "loss": 2.8702, "step": 54504 }, { "epoch": 4.27, "learning_rate": 5.726103661883478e-05, "loss": 2.5326, "step": 54505 }, { "epoch": 4.27, "learning_rate": 5.726025248961029e-05, "loss": 2.625, "step": 54506 }, { "epoch": 4.27, "learning_rate": 5.7259468360385796e-05, "loss": 2.7124, "step": 54507 }, { "epoch": 4.27, "learning_rate": 5.72586842311613e-05, "loss": 2.9023, "step": 54508 }, { "epoch": 4.27, "learning_rate": 5.72579001019368e-05, "loss": 2.6875, "step": 54509 }, { "epoch": 4.27, "learning_rate": 5.7257115972712307e-05, "loss": 2.6367, "step": 54510 }, { "epoch": 4.27, "learning_rate": 5.725633184348781e-05, "loss": 2.7788, "step": 54511 }, { "epoch": 4.27, "learning_rate": 5.7255547714263314e-05, "loss": 2.6538, "step": 54512 }, { "epoch": 4.27, "learning_rate": 5.725476358503882e-05, "loss": 2.8305, "step": 54513 }, { "epoch": 4.27, "learning_rate": 5.725397945581432e-05, "loss": 2.4967, "step": 54514 }, { "epoch": 4.27, "learning_rate": 5.7253195326589825e-05, "loss": 2.9267, "step": 54515 }, { "epoch": 4.27, "learning_rate": 5.725241119736533e-05, "loss": 2.6873, "step": 54516 }, { "epoch": 4.27, "learning_rate": 5.725162706814083e-05, "loss": 2.5879, "step": 54517 }, { "epoch": 4.27, "learning_rate": 5.7250842938916336e-05, "loss": 2.8184, "step": 54518 }, { "epoch": 4.27, "learning_rate": 5.725005880969184e-05, "loss": 2.5828, "step": 54519 }, { "epoch": 4.28, "learning_rate": 5.724927468046734e-05, "loss": 2.5785, "step": 54520 }, { "epoch": 4.28, "learning_rate": 5.7248490551242854e-05, "loss": 2.8884, "step": 54521 }, { "epoch": 4.28, "learning_rate": 5.724770642201836e-05, "loss": 2.7019, "step": 54522 }, { "epoch": 4.28, "learning_rate": 5.724692229279386e-05, "loss": 2.743, "step": 54523 }, { "epoch": 4.28, "learning_rate": 5.7246138163569365e-05, "loss": 2.633, "step": 54524 }, { "epoch": 4.28, "learning_rate": 5.724535403434487e-05, "loss": 2.5548, "step": 54525 }, { "epoch": 4.28, "learning_rate": 5.724456990512037e-05, "loss": 2.5732, "step": 54526 }, { "epoch": 4.28, "learning_rate": 5.7243785775895876e-05, "loss": 2.613, "step": 54527 }, { "epoch": 4.28, "learning_rate": 5.724300164667138e-05, "loss": 2.5256, "step": 54528 }, { "epoch": 4.28, "learning_rate": 5.724221751744688e-05, "loss": 2.6577, "step": 54529 }, { "epoch": 4.28, "learning_rate": 5.724143338822237e-05, "loss": 2.6756, "step": 54530 }, { "epoch": 4.28, "learning_rate": 5.7240649258997884e-05, "loss": 2.6514, "step": 54531 }, { "epoch": 4.28, "learning_rate": 5.723986512977339e-05, "loss": 2.6642, "step": 54532 }, { "epoch": 4.28, "learning_rate": 5.723908100054889e-05, "loss": 2.6467, "step": 54533 }, { "epoch": 4.28, "learning_rate": 5.7238296871324395e-05, "loss": 2.6878, "step": 54534 }, { "epoch": 4.28, "learning_rate": 5.72375127420999e-05, "loss": 3.0044, "step": 54535 }, { "epoch": 4.28, "learning_rate": 5.72367286128754e-05, "loss": 2.6393, "step": 54536 }, { "epoch": 4.28, "learning_rate": 5.7235944483650906e-05, "loss": 2.6906, "step": 54537 }, { "epoch": 4.28, "learning_rate": 5.723516035442641e-05, "loss": 2.5863, "step": 54538 }, { "epoch": 4.28, "learning_rate": 5.723437622520191e-05, "loss": 2.8285, "step": 54539 }, { "epoch": 4.28, "learning_rate": 5.723359209597742e-05, "loss": 2.5023, "step": 54540 }, { "epoch": 4.28, "learning_rate": 5.723280796675292e-05, "loss": 2.7098, "step": 54541 }, { "epoch": 4.28, "learning_rate": 5.7232023837528424e-05, "loss": 2.672, "step": 54542 }, { "epoch": 4.28, "learning_rate": 5.723123970830393e-05, "loss": 2.4851, "step": 54543 }, { "epoch": 4.28, "learning_rate": 5.723045557907943e-05, "loss": 2.7294, "step": 54544 }, { "epoch": 4.28, "learning_rate": 5.7229671449854935e-05, "loss": 2.5767, "step": 54545 }, { "epoch": 4.28, "learning_rate": 5.722888732063044e-05, "loss": 2.5005, "step": 54546 }, { "epoch": 4.28, "learning_rate": 5.722810319140595e-05, "loss": 2.5928, "step": 54547 }, { "epoch": 4.28, "learning_rate": 5.722731906218145e-05, "loss": 2.7089, "step": 54548 }, { "epoch": 4.28, "learning_rate": 5.722653493295696e-05, "loss": 2.7158, "step": 54549 }, { "epoch": 4.28, "learning_rate": 5.722575080373246e-05, "loss": 2.7978, "step": 54550 }, { "epoch": 4.28, "learning_rate": 5.7224966674507964e-05, "loss": 2.5994, "step": 54551 }, { "epoch": 4.28, "learning_rate": 5.722418254528347e-05, "loss": 2.719, "step": 54552 }, { "epoch": 4.28, "learning_rate": 5.722339841605897e-05, "loss": 2.942, "step": 54553 }, { "epoch": 4.28, "learning_rate": 5.7222614286834475e-05, "loss": 2.6079, "step": 54554 }, { "epoch": 4.28, "learning_rate": 5.722183015760998e-05, "loss": 2.7155, "step": 54555 }, { "epoch": 4.28, "learning_rate": 5.722104602838548e-05, "loss": 2.4868, "step": 54556 }, { "epoch": 4.28, "learning_rate": 5.7220261899160986e-05, "loss": 2.695, "step": 54557 }, { "epoch": 4.28, "learning_rate": 5.721947776993649e-05, "loss": 2.826, "step": 54558 }, { "epoch": 4.28, "learning_rate": 5.7218693640711994e-05, "loss": 2.6925, "step": 54559 }, { "epoch": 4.28, "learning_rate": 5.72179095114875e-05, "loss": 2.481, "step": 54560 }, { "epoch": 4.28, "learning_rate": 5.7217125382263e-05, "loss": 2.5816, "step": 54561 }, { "epoch": 4.28, "learning_rate": 5.7216341253038505e-05, "loss": 2.4679, "step": 54562 }, { "epoch": 4.28, "learning_rate": 5.7215557123814015e-05, "loss": 2.8436, "step": 54563 }, { "epoch": 4.28, "learning_rate": 5.721477299458952e-05, "loss": 2.7033, "step": 54564 }, { "epoch": 4.28, "learning_rate": 5.721398886536502e-05, "loss": 2.5708, "step": 54565 }, { "epoch": 4.28, "learning_rate": 5.721320473614051e-05, "loss": 2.5682, "step": 54566 }, { "epoch": 4.28, "learning_rate": 5.7212420606916016e-05, "loss": 2.6412, "step": 54567 }, { "epoch": 4.28, "learning_rate": 5.721163647769152e-05, "loss": 2.7041, "step": 54568 }, { "epoch": 4.28, "learning_rate": 5.7210852348467024e-05, "loss": 2.5663, "step": 54569 }, { "epoch": 4.28, "learning_rate": 5.721006821924253e-05, "loss": 2.6455, "step": 54570 }, { "epoch": 4.28, "learning_rate": 5.720928409001803e-05, "loss": 2.5448, "step": 54571 }, { "epoch": 4.28, "learning_rate": 5.7208499960793535e-05, "loss": 2.6658, "step": 54572 }, { "epoch": 4.28, "learning_rate": 5.7207715831569045e-05, "loss": 2.6473, "step": 54573 }, { "epoch": 4.28, "learning_rate": 5.720693170234455e-05, "loss": 2.7966, "step": 54574 }, { "epoch": 4.28, "learning_rate": 5.720614757312005e-05, "loss": 2.574, "step": 54575 }, { "epoch": 4.28, "learning_rate": 5.7205363443895556e-05, "loss": 2.6111, "step": 54576 }, { "epoch": 4.28, "learning_rate": 5.720457931467106e-05, "loss": 2.788, "step": 54577 }, { "epoch": 4.28, "learning_rate": 5.7203795185446563e-05, "loss": 2.602, "step": 54578 }, { "epoch": 4.28, "learning_rate": 5.720301105622207e-05, "loss": 2.547, "step": 54579 }, { "epoch": 4.28, "learning_rate": 5.720222692699757e-05, "loss": 2.4832, "step": 54580 }, { "epoch": 4.28, "learning_rate": 5.7201442797773074e-05, "loss": 2.6673, "step": 54581 }, { "epoch": 4.28, "learning_rate": 5.720065866854858e-05, "loss": 2.7267, "step": 54582 }, { "epoch": 4.28, "learning_rate": 5.719987453932408e-05, "loss": 2.5179, "step": 54583 }, { "epoch": 4.28, "learning_rate": 5.7199090410099585e-05, "loss": 2.4934, "step": 54584 }, { "epoch": 4.28, "learning_rate": 5.719830628087509e-05, "loss": 2.7321, "step": 54585 }, { "epoch": 4.28, "learning_rate": 5.719752215165059e-05, "loss": 2.6408, "step": 54586 }, { "epoch": 4.28, "learning_rate": 5.7196738022426097e-05, "loss": 2.721, "step": 54587 }, { "epoch": 4.28, "learning_rate": 5.719595389320161e-05, "loss": 2.8326, "step": 54588 }, { "epoch": 4.28, "learning_rate": 5.719516976397711e-05, "loss": 2.646, "step": 54589 }, { "epoch": 4.28, "learning_rate": 5.7194385634752614e-05, "loss": 2.7434, "step": 54590 }, { "epoch": 4.28, "learning_rate": 5.719360150552812e-05, "loss": 2.651, "step": 54591 }, { "epoch": 4.28, "learning_rate": 5.719281737630362e-05, "loss": 2.7317, "step": 54592 }, { "epoch": 4.28, "learning_rate": 5.7192033247079125e-05, "loss": 2.7797, "step": 54593 }, { "epoch": 4.28, "learning_rate": 5.719124911785463e-05, "loss": 2.5957, "step": 54594 }, { "epoch": 4.28, "learning_rate": 5.719046498863013e-05, "loss": 2.6785, "step": 54595 }, { "epoch": 4.28, "learning_rate": 5.7189680859405636e-05, "loss": 2.7495, "step": 54596 }, { "epoch": 4.28, "learning_rate": 5.718889673018114e-05, "loss": 2.7594, "step": 54597 }, { "epoch": 4.28, "learning_rate": 5.7188112600956644e-05, "loss": 2.6547, "step": 54598 }, { "epoch": 4.28, "learning_rate": 5.718732847173215e-05, "loss": 2.6438, "step": 54599 }, { "epoch": 4.28, "learning_rate": 5.718654434250765e-05, "loss": 2.777, "step": 54600 }, { "epoch": 4.28, "learning_rate": 5.7185760213283155e-05, "loss": 2.5276, "step": 54601 }, { "epoch": 4.28, "learning_rate": 5.718497608405865e-05, "loss": 2.7174, "step": 54602 }, { "epoch": 4.28, "learning_rate": 5.7184191954834155e-05, "loss": 2.5801, "step": 54603 }, { "epoch": 4.28, "learning_rate": 5.718340782560966e-05, "loss": 2.7161, "step": 54604 }, { "epoch": 4.28, "learning_rate": 5.718262369638516e-05, "loss": 2.6665, "step": 54605 }, { "epoch": 4.28, "learning_rate": 5.7181839567160666e-05, "loss": 2.8563, "step": 54606 }, { "epoch": 4.28, "learning_rate": 5.718105543793617e-05, "loss": 2.5751, "step": 54607 }, { "epoch": 4.28, "learning_rate": 5.7180271308711674e-05, "loss": 3.0265, "step": 54608 }, { "epoch": 4.28, "learning_rate": 5.717948717948718e-05, "loss": 2.6832, "step": 54609 }, { "epoch": 4.28, "learning_rate": 5.717870305026268e-05, "loss": 2.6132, "step": 54610 }, { "epoch": 4.28, "learning_rate": 5.7177918921038185e-05, "loss": 2.6132, "step": 54611 }, { "epoch": 4.28, "learning_rate": 5.717713479181369e-05, "loss": 2.7586, "step": 54612 }, { "epoch": 4.28, "learning_rate": 5.717635066258919e-05, "loss": 2.653, "step": 54613 }, { "epoch": 4.28, "learning_rate": 5.71755665333647e-05, "loss": 2.7272, "step": 54614 }, { "epoch": 4.28, "learning_rate": 5.7174782404140206e-05, "loss": 2.4974, "step": 54615 }, { "epoch": 4.28, "learning_rate": 5.717399827491571e-05, "loss": 2.8266, "step": 54616 }, { "epoch": 4.28, "learning_rate": 5.7173214145691214e-05, "loss": 2.7337, "step": 54617 }, { "epoch": 4.28, "learning_rate": 5.717243001646672e-05, "loss": 2.5647, "step": 54618 }, { "epoch": 4.28, "learning_rate": 5.717164588724222e-05, "loss": 2.7973, "step": 54619 }, { "epoch": 4.28, "learning_rate": 5.7170861758017725e-05, "loss": 2.5155, "step": 54620 }, { "epoch": 4.28, "learning_rate": 5.717007762879323e-05, "loss": 2.6035, "step": 54621 }, { "epoch": 4.28, "learning_rate": 5.716929349956873e-05, "loss": 2.5871, "step": 54622 }, { "epoch": 4.28, "learning_rate": 5.7168509370344236e-05, "loss": 2.7432, "step": 54623 }, { "epoch": 4.28, "learning_rate": 5.716772524111974e-05, "loss": 2.6664, "step": 54624 }, { "epoch": 4.28, "learning_rate": 5.716694111189524e-05, "loss": 2.5028, "step": 54625 }, { "epoch": 4.28, "learning_rate": 5.716615698267075e-05, "loss": 2.738, "step": 54626 }, { "epoch": 4.28, "learning_rate": 5.716537285344625e-05, "loss": 2.9902, "step": 54627 }, { "epoch": 4.28, "learning_rate": 5.7164588724221754e-05, "loss": 2.7405, "step": 54628 }, { "epoch": 4.28, "learning_rate": 5.7163804594997265e-05, "loss": 2.6056, "step": 54629 }, { "epoch": 4.28, "learning_rate": 5.716302046577277e-05, "loss": 2.6273, "step": 54630 }, { "epoch": 4.28, "learning_rate": 5.716223633654827e-05, "loss": 2.54, "step": 54631 }, { "epoch": 4.28, "learning_rate": 5.7161452207323776e-05, "loss": 2.4638, "step": 54632 }, { "epoch": 4.28, "learning_rate": 5.716066807809928e-05, "loss": 2.7801, "step": 54633 }, { "epoch": 4.28, "learning_rate": 5.715988394887478e-05, "loss": 2.713, "step": 54634 }, { "epoch": 4.28, "learning_rate": 5.7159099819650287e-05, "loss": 2.6631, "step": 54635 }, { "epoch": 4.28, "learning_rate": 5.715831569042579e-05, "loss": 2.5757, "step": 54636 }, { "epoch": 4.28, "learning_rate": 5.7157531561201294e-05, "loss": 2.682, "step": 54637 }, { "epoch": 4.28, "learning_rate": 5.7156747431976784e-05, "loss": 2.6958, "step": 54638 }, { "epoch": 4.28, "learning_rate": 5.7155963302752295e-05, "loss": 2.6402, "step": 54639 }, { "epoch": 4.28, "learning_rate": 5.71551791735278e-05, "loss": 2.7844, "step": 54640 }, { "epoch": 4.28, "learning_rate": 5.71543950443033e-05, "loss": 2.6413, "step": 54641 }, { "epoch": 4.28, "learning_rate": 5.7153610915078806e-05, "loss": 2.7223, "step": 54642 }, { "epoch": 4.28, "learning_rate": 5.715282678585431e-05, "loss": 2.5501, "step": 54643 }, { "epoch": 4.28, "learning_rate": 5.715204265662981e-05, "loss": 2.682, "step": 54644 }, { "epoch": 4.28, "learning_rate": 5.7151258527405317e-05, "loss": 2.4746, "step": 54645 }, { "epoch": 4.28, "learning_rate": 5.715047439818082e-05, "loss": 2.4696, "step": 54646 }, { "epoch": 4.29, "learning_rate": 5.7149690268956324e-05, "loss": 2.7225, "step": 54647 }, { "epoch": 4.29, "learning_rate": 5.714890613973183e-05, "loss": 2.6268, "step": 54648 }, { "epoch": 4.29, "learning_rate": 5.714812201050733e-05, "loss": 2.5302, "step": 54649 }, { "epoch": 4.29, "learning_rate": 5.7147337881282835e-05, "loss": 2.7557, "step": 54650 }, { "epoch": 4.29, "learning_rate": 5.714655375205834e-05, "loss": 2.7636, "step": 54651 }, { "epoch": 4.29, "learning_rate": 5.714576962283384e-05, "loss": 2.5671, "step": 54652 }, { "epoch": 4.29, "learning_rate": 5.7144985493609346e-05, "loss": 2.6328, "step": 54653 }, { "epoch": 4.29, "learning_rate": 5.714420136438485e-05, "loss": 2.6532, "step": 54654 }, { "epoch": 4.29, "learning_rate": 5.714341723516036e-05, "loss": 2.7325, "step": 54655 }, { "epoch": 4.29, "learning_rate": 5.7142633105935864e-05, "loss": 2.6708, "step": 54656 }, { "epoch": 4.29, "learning_rate": 5.714184897671137e-05, "loss": 2.5999, "step": 54657 }, { "epoch": 4.29, "learning_rate": 5.714106484748687e-05, "loss": 2.6405, "step": 54658 }, { "epoch": 4.29, "learning_rate": 5.7140280718262375e-05, "loss": 2.7745, "step": 54659 }, { "epoch": 4.29, "learning_rate": 5.713949658903788e-05, "loss": 2.6, "step": 54660 }, { "epoch": 4.29, "learning_rate": 5.713871245981338e-05, "loss": 2.9413, "step": 54661 }, { "epoch": 4.29, "learning_rate": 5.7137928330588886e-05, "loss": 2.5642, "step": 54662 }, { "epoch": 4.29, "learning_rate": 5.713714420136439e-05, "loss": 2.7436, "step": 54663 }, { "epoch": 4.29, "learning_rate": 5.713636007213989e-05, "loss": 2.5875, "step": 54664 }, { "epoch": 4.29, "learning_rate": 5.71355759429154e-05, "loss": 2.7015, "step": 54665 }, { "epoch": 4.29, "learning_rate": 5.71347918136909e-05, "loss": 2.5626, "step": 54666 }, { "epoch": 4.29, "learning_rate": 5.7134007684466404e-05, "loss": 2.7188, "step": 54667 }, { "epoch": 4.29, "learning_rate": 5.713322355524191e-05, "loss": 2.5003, "step": 54668 }, { "epoch": 4.29, "learning_rate": 5.713243942601741e-05, "loss": 2.8937, "step": 54669 }, { "epoch": 4.29, "learning_rate": 5.7131655296792915e-05, "loss": 2.6344, "step": 54670 }, { "epoch": 4.29, "learning_rate": 5.7130871167568426e-05, "loss": 2.8189, "step": 54671 }, { "epoch": 4.29, "learning_rate": 5.713008703834393e-05, "loss": 2.7259, "step": 54672 }, { "epoch": 4.29, "learning_rate": 5.712930290911943e-05, "loss": 2.598, "step": 54673 }, { "epoch": 4.29, "learning_rate": 5.712851877989492e-05, "loss": 2.8361, "step": 54674 }, { "epoch": 4.29, "learning_rate": 5.712773465067043e-05, "loss": 2.6874, "step": 54675 }, { "epoch": 4.29, "learning_rate": 5.712695052144593e-05, "loss": 2.6811, "step": 54676 }, { "epoch": 4.29, "learning_rate": 5.7126166392221434e-05, "loss": 2.8169, "step": 54677 }, { "epoch": 4.29, "learning_rate": 5.712538226299694e-05, "loss": 2.666, "step": 54678 }, { "epoch": 4.29, "learning_rate": 5.712459813377244e-05, "loss": 2.474, "step": 54679 }, { "epoch": 4.29, "learning_rate": 5.7123814004547945e-05, "loss": 2.6175, "step": 54680 }, { "epoch": 4.29, "learning_rate": 5.7123029875323456e-05, "loss": 2.652, "step": 54681 }, { "epoch": 4.29, "learning_rate": 5.712224574609896e-05, "loss": 2.4559, "step": 54682 }, { "epoch": 4.29, "learning_rate": 5.712146161687446e-05, "loss": 2.5974, "step": 54683 }, { "epoch": 4.29, "learning_rate": 5.712067748764997e-05, "loss": 2.89, "step": 54684 }, { "epoch": 4.29, "learning_rate": 5.711989335842547e-05, "loss": 2.7233, "step": 54685 }, { "epoch": 4.29, "learning_rate": 5.7119109229200974e-05, "loss": 2.6525, "step": 54686 }, { "epoch": 4.29, "learning_rate": 5.711832509997648e-05, "loss": 2.5446, "step": 54687 }, { "epoch": 4.29, "learning_rate": 5.711754097075198e-05, "loss": 2.6802, "step": 54688 }, { "epoch": 4.29, "learning_rate": 5.7116756841527485e-05, "loss": 2.7901, "step": 54689 }, { "epoch": 4.29, "learning_rate": 5.711597271230299e-05, "loss": 2.6066, "step": 54690 }, { "epoch": 4.29, "learning_rate": 5.711518858307849e-05, "loss": 2.8743, "step": 54691 }, { "epoch": 4.29, "learning_rate": 5.7114404453853996e-05, "loss": 2.6204, "step": 54692 }, { "epoch": 4.29, "learning_rate": 5.71136203246295e-05, "loss": 2.5406, "step": 54693 }, { "epoch": 4.29, "learning_rate": 5.7112836195405004e-05, "loss": 2.7681, "step": 54694 }, { "epoch": 4.29, "learning_rate": 5.711205206618051e-05, "loss": 2.5547, "step": 54695 }, { "epoch": 4.29, "learning_rate": 5.711126793695602e-05, "loss": 2.5613, "step": 54696 }, { "epoch": 4.29, "learning_rate": 5.711048380773152e-05, "loss": 2.7314, "step": 54697 }, { "epoch": 4.29, "learning_rate": 5.7109699678507025e-05, "loss": 2.6963, "step": 54698 }, { "epoch": 4.29, "learning_rate": 5.710891554928253e-05, "loss": 2.649, "step": 54699 }, { "epoch": 4.29, "learning_rate": 5.710813142005803e-05, "loss": 2.5916, "step": 54700 }, { "epoch": 4.29, "learning_rate": 5.7107347290833536e-05, "loss": 2.8225, "step": 54701 }, { "epoch": 4.29, "learning_rate": 5.710656316160904e-05, "loss": 2.7701, "step": 54702 }, { "epoch": 4.29, "learning_rate": 5.7105779032384543e-05, "loss": 2.6144, "step": 54703 }, { "epoch": 4.29, "learning_rate": 5.710499490316005e-05, "loss": 2.7216, "step": 54704 }, { "epoch": 4.29, "learning_rate": 5.710421077393555e-05, "loss": 2.5836, "step": 54705 }, { "epoch": 4.29, "learning_rate": 5.7103426644711054e-05, "loss": 2.6443, "step": 54706 }, { "epoch": 4.29, "learning_rate": 5.710264251548656e-05, "loss": 2.5553, "step": 54707 }, { "epoch": 4.29, "learning_rate": 5.710185838626206e-05, "loss": 2.5079, "step": 54708 }, { "epoch": 4.29, "learning_rate": 5.7101074257037566e-05, "loss": 2.6715, "step": 54709 }, { "epoch": 4.29, "learning_rate": 5.710029012781306e-05, "loss": 2.5721, "step": 54710 }, { "epoch": 4.29, "learning_rate": 5.7099505998588566e-05, "loss": 2.5168, "step": 54711 }, { "epoch": 4.29, "learning_rate": 5.709872186936407e-05, "loss": 2.5559, "step": 54712 }, { "epoch": 4.29, "learning_rate": 5.7097937740139573e-05, "loss": 2.6851, "step": 54713 }, { "epoch": 4.29, "learning_rate": 5.709715361091508e-05, "loss": 2.5954, "step": 54714 }, { "epoch": 4.29, "learning_rate": 5.709636948169058e-05, "loss": 2.5392, "step": 54715 }, { "epoch": 4.29, "learning_rate": 5.7095585352466084e-05, "loss": 2.8548, "step": 54716 }, { "epoch": 4.29, "learning_rate": 5.709480122324159e-05, "loss": 2.5708, "step": 54717 }, { "epoch": 4.29, "learning_rate": 5.709401709401709e-05, "loss": 2.6199, "step": 54718 }, { "epoch": 4.29, "learning_rate": 5.7093232964792596e-05, "loss": 2.808, "step": 54719 }, { "epoch": 4.29, "learning_rate": 5.70924488355681e-05, "loss": 2.5432, "step": 54720 }, { "epoch": 4.29, "learning_rate": 5.70916647063436e-05, "loss": 2.6932, "step": 54721 }, { "epoch": 4.29, "learning_rate": 5.709088057711911e-05, "loss": 2.5072, "step": 54722 }, { "epoch": 4.29, "learning_rate": 5.709009644789462e-05, "loss": 2.81, "step": 54723 }, { "epoch": 4.29, "learning_rate": 5.708931231867012e-05, "loss": 2.7764, "step": 54724 }, { "epoch": 4.29, "learning_rate": 5.7088528189445624e-05, "loss": 2.7052, "step": 54725 }, { "epoch": 4.29, "learning_rate": 5.708774406022113e-05, "loss": 2.6113, "step": 54726 }, { "epoch": 4.29, "learning_rate": 5.708695993099663e-05, "loss": 2.632, "step": 54727 }, { "epoch": 4.29, "learning_rate": 5.7086175801772135e-05, "loss": 2.6728, "step": 54728 }, { "epoch": 4.29, "learning_rate": 5.708539167254764e-05, "loss": 2.5766, "step": 54729 }, { "epoch": 4.29, "learning_rate": 5.708460754332314e-05, "loss": 2.7289, "step": 54730 }, { "epoch": 4.29, "learning_rate": 5.7083823414098646e-05, "loss": 2.6821, "step": 54731 }, { "epoch": 4.29, "learning_rate": 5.708303928487415e-05, "loss": 2.9061, "step": 54732 }, { "epoch": 4.29, "learning_rate": 5.7082255155649654e-05, "loss": 2.6541, "step": 54733 }, { "epoch": 4.29, "learning_rate": 5.708147102642516e-05, "loss": 2.5033, "step": 54734 }, { "epoch": 4.29, "learning_rate": 5.708068689720066e-05, "loss": 2.6656, "step": 54735 }, { "epoch": 4.29, "learning_rate": 5.7079902767976165e-05, "loss": 2.7865, "step": 54736 }, { "epoch": 4.29, "learning_rate": 5.707911863875167e-05, "loss": 2.6048, "step": 54737 }, { "epoch": 4.29, "learning_rate": 5.707833450952718e-05, "loss": 2.6898, "step": 54738 }, { "epoch": 4.29, "learning_rate": 5.707755038030268e-05, "loss": 2.6594, "step": 54739 }, { "epoch": 4.29, "learning_rate": 5.7076766251078186e-05, "loss": 2.6587, "step": 54740 }, { "epoch": 4.29, "learning_rate": 5.707598212185369e-05, "loss": 2.7097, "step": 54741 }, { "epoch": 4.29, "learning_rate": 5.7075197992629194e-05, "loss": 2.589, "step": 54742 }, { "epoch": 4.29, "learning_rate": 5.70744138634047e-05, "loss": 2.7082, "step": 54743 }, { "epoch": 4.29, "learning_rate": 5.70736297341802e-05, "loss": 2.6589, "step": 54744 }, { "epoch": 4.29, "learning_rate": 5.7072845604955705e-05, "loss": 2.6223, "step": 54745 }, { "epoch": 4.29, "learning_rate": 5.7072061475731195e-05, "loss": 2.7123, "step": 54746 }, { "epoch": 4.29, "learning_rate": 5.70712773465067e-05, "loss": 2.7355, "step": 54747 }, { "epoch": 4.29, "learning_rate": 5.707049321728221e-05, "loss": 2.8048, "step": 54748 }, { "epoch": 4.29, "learning_rate": 5.706970908805771e-05, "loss": 2.5569, "step": 54749 }, { "epoch": 4.29, "learning_rate": 5.7068924958833216e-05, "loss": 2.6492, "step": 54750 }, { "epoch": 4.29, "learning_rate": 5.706814082960872e-05, "loss": 2.6786, "step": 54751 }, { "epoch": 4.29, "learning_rate": 5.7067356700384224e-05, "loss": 2.632, "step": 54752 }, { "epoch": 4.29, "learning_rate": 5.706657257115973e-05, "loss": 2.6804, "step": 54753 }, { "epoch": 4.29, "learning_rate": 5.706578844193523e-05, "loss": 2.7696, "step": 54754 }, { "epoch": 4.29, "learning_rate": 5.7065004312710735e-05, "loss": 2.658, "step": 54755 }, { "epoch": 4.29, "learning_rate": 5.706422018348624e-05, "loss": 2.7802, "step": 54756 }, { "epoch": 4.29, "learning_rate": 5.706343605426174e-05, "loss": 2.7203, "step": 54757 }, { "epoch": 4.29, "learning_rate": 5.7062651925037246e-05, "loss": 2.7507, "step": 54758 }, { "epoch": 4.29, "learning_rate": 5.706186779581275e-05, "loss": 2.6555, "step": 54759 }, { "epoch": 4.29, "learning_rate": 5.706108366658825e-05, "loss": 2.5626, "step": 54760 }, { "epoch": 4.29, "learning_rate": 5.706029953736376e-05, "loss": 2.8858, "step": 54761 }, { "epoch": 4.29, "learning_rate": 5.705951540813926e-05, "loss": 2.6634, "step": 54762 }, { "epoch": 4.29, "learning_rate": 5.705873127891477e-05, "loss": 2.645, "step": 54763 }, { "epoch": 4.29, "learning_rate": 5.7057947149690275e-05, "loss": 2.4633, "step": 54764 }, { "epoch": 4.29, "learning_rate": 5.705716302046578e-05, "loss": 2.5538, "step": 54765 }, { "epoch": 4.29, "learning_rate": 5.705637889124128e-05, "loss": 2.5857, "step": 54766 }, { "epoch": 4.29, "learning_rate": 5.7055594762016786e-05, "loss": 3.0645, "step": 54767 }, { "epoch": 4.29, "learning_rate": 5.705481063279229e-05, "loss": 2.7016, "step": 54768 }, { "epoch": 4.29, "learning_rate": 5.705402650356779e-05, "loss": 2.5105, "step": 54769 }, { "epoch": 4.29, "learning_rate": 5.70532423743433e-05, "loss": 2.8814, "step": 54770 }, { "epoch": 4.29, "learning_rate": 5.70524582451188e-05, "loss": 2.6383, "step": 54771 }, { "epoch": 4.29, "learning_rate": 5.7051674115894304e-05, "loss": 2.4954, "step": 54772 }, { "epoch": 4.29, "learning_rate": 5.705088998666981e-05, "loss": 2.6584, "step": 54773 }, { "epoch": 4.29, "learning_rate": 5.705010585744531e-05, "loss": 2.662, "step": 54774 }, { "epoch": 4.3, "learning_rate": 5.7049321728220815e-05, "loss": 2.6736, "step": 54775 }, { "epoch": 4.3, "learning_rate": 5.704853759899632e-05, "loss": 2.6001, "step": 54776 }, { "epoch": 4.3, "learning_rate": 5.704775346977182e-05, "loss": 2.6384, "step": 54777 }, { "epoch": 4.3, "learning_rate": 5.7046969340547326e-05, "loss": 2.9335, "step": 54778 }, { "epoch": 4.3, "learning_rate": 5.7046185211322837e-05, "loss": 2.6092, "step": 54779 }, { "epoch": 4.3, "learning_rate": 5.704540108209834e-05, "loss": 2.5894, "step": 54780 }, { "epoch": 4.3, "learning_rate": 5.7044616952873844e-05, "loss": 2.5159, "step": 54781 }, { "epoch": 4.3, "learning_rate": 5.7043832823649334e-05, "loss": 2.7703, "step": 54782 }, { "epoch": 4.3, "learning_rate": 5.704304869442484e-05, "loss": 2.6337, "step": 54783 }, { "epoch": 4.3, "learning_rate": 5.704226456520034e-05, "loss": 2.5159, "step": 54784 }, { "epoch": 4.3, "learning_rate": 5.7041480435975845e-05, "loss": 2.5403, "step": 54785 }, { "epoch": 4.3, "learning_rate": 5.704069630675135e-05, "loss": 2.5297, "step": 54786 }, { "epoch": 4.3, "learning_rate": 5.703991217752685e-05, "loss": 2.8884, "step": 54787 }, { "epoch": 4.3, "learning_rate": 5.7039128048302356e-05, "loss": 2.7656, "step": 54788 }, { "epoch": 4.3, "learning_rate": 5.7038343919077867e-05, "loss": 2.5377, "step": 54789 }, { "epoch": 4.3, "learning_rate": 5.703755978985337e-05, "loss": 2.7959, "step": 54790 }, { "epoch": 4.3, "learning_rate": 5.7036775660628874e-05, "loss": 2.6718, "step": 54791 }, { "epoch": 4.3, "learning_rate": 5.703599153140438e-05, "loss": 2.6208, "step": 54792 }, { "epoch": 4.3, "learning_rate": 5.703520740217988e-05, "loss": 2.9783, "step": 54793 }, { "epoch": 4.3, "learning_rate": 5.7034423272955385e-05, "loss": 2.5797, "step": 54794 }, { "epoch": 4.3, "learning_rate": 5.703363914373089e-05, "loss": 2.5725, "step": 54795 }, { "epoch": 4.3, "learning_rate": 5.703285501450639e-05, "loss": 2.5736, "step": 54796 }, { "epoch": 4.3, "learning_rate": 5.7032070885281896e-05, "loss": 2.5429, "step": 54797 }, { "epoch": 4.3, "learning_rate": 5.70312867560574e-05, "loss": 2.7643, "step": 54798 }, { "epoch": 4.3, "learning_rate": 5.70305026268329e-05, "loss": 2.492, "step": 54799 }, { "epoch": 4.3, "learning_rate": 5.702971849760841e-05, "loss": 2.7293, "step": 54800 }, { "epoch": 4.3, "learning_rate": 5.702893436838391e-05, "loss": 2.7885, "step": 54801 }, { "epoch": 4.3, "learning_rate": 5.7028150239159414e-05, "loss": 2.5416, "step": 54802 }, { "epoch": 4.3, "learning_rate": 5.702736610993492e-05, "loss": 2.7279, "step": 54803 }, { "epoch": 4.3, "learning_rate": 5.702658198071043e-05, "loss": 2.5959, "step": 54804 }, { "epoch": 4.3, "learning_rate": 5.702579785148593e-05, "loss": 2.7083, "step": 54805 }, { "epoch": 4.3, "learning_rate": 5.7025013722261436e-05, "loss": 2.6978, "step": 54806 }, { "epoch": 4.3, "learning_rate": 5.702422959303694e-05, "loss": 2.7036, "step": 54807 }, { "epoch": 4.3, "learning_rate": 5.702344546381244e-05, "loss": 2.5997, "step": 54808 }, { "epoch": 4.3, "learning_rate": 5.702266133458795e-05, "loss": 2.5125, "step": 54809 }, { "epoch": 4.3, "learning_rate": 5.702187720536345e-05, "loss": 2.8685, "step": 54810 }, { "epoch": 4.3, "learning_rate": 5.7021093076138954e-05, "loss": 2.4961, "step": 54811 }, { "epoch": 4.3, "learning_rate": 5.702030894691446e-05, "loss": 2.7075, "step": 54812 }, { "epoch": 4.3, "learning_rate": 5.701952481768996e-05, "loss": 2.7696, "step": 54813 }, { "epoch": 4.3, "learning_rate": 5.7018740688465465e-05, "loss": 2.4641, "step": 54814 }, { "epoch": 4.3, "learning_rate": 5.701795655924097e-05, "loss": 2.5074, "step": 54815 }, { "epoch": 4.3, "learning_rate": 5.701717243001647e-05, "loss": 2.8594, "step": 54816 }, { "epoch": 4.3, "learning_rate": 5.7016388300791976e-05, "loss": 2.6081, "step": 54817 }, { "epoch": 4.3, "learning_rate": 5.701560417156747e-05, "loss": 2.6439, "step": 54818 }, { "epoch": 4.3, "learning_rate": 5.701482004234298e-05, "loss": 2.6833, "step": 54819 }, { "epoch": 4.3, "learning_rate": 5.701403591311848e-05, "loss": 2.788, "step": 54820 }, { "epoch": 4.3, "learning_rate": 5.7013251783893984e-05, "loss": 2.5278, "step": 54821 }, { "epoch": 4.3, "learning_rate": 5.701246765466949e-05, "loss": 2.8166, "step": 54822 }, { "epoch": 4.3, "learning_rate": 5.701168352544499e-05, "loss": 2.8849, "step": 54823 }, { "epoch": 4.3, "learning_rate": 5.7010899396220495e-05, "loss": 2.4101, "step": 54824 }, { "epoch": 4.3, "learning_rate": 5.7010115266996e-05, "loss": 2.712, "step": 54825 }, { "epoch": 4.3, "learning_rate": 5.70093311377715e-05, "loss": 2.5702, "step": 54826 }, { "epoch": 4.3, "learning_rate": 5.7008547008547006e-05, "loss": 2.742, "step": 54827 }, { "epoch": 4.3, "learning_rate": 5.700776287932251e-05, "loss": 2.6334, "step": 54828 }, { "epoch": 4.3, "learning_rate": 5.7006978750098014e-05, "loss": 2.4625, "step": 54829 }, { "epoch": 4.3, "learning_rate": 5.7006194620873524e-05, "loss": 2.6972, "step": 54830 }, { "epoch": 4.3, "learning_rate": 5.700541049164903e-05, "loss": 2.7249, "step": 54831 }, { "epoch": 4.3, "learning_rate": 5.700462636242453e-05, "loss": 2.5733, "step": 54832 }, { "epoch": 4.3, "learning_rate": 5.7003842233200035e-05, "loss": 2.5975, "step": 54833 }, { "epoch": 4.3, "learning_rate": 5.700305810397554e-05, "loss": 2.7704, "step": 54834 }, { "epoch": 4.3, "learning_rate": 5.700227397475104e-05, "loss": 2.5164, "step": 54835 }, { "epoch": 4.3, "learning_rate": 5.7001489845526546e-05, "loss": 2.6456, "step": 54836 }, { "epoch": 4.3, "learning_rate": 5.700070571630205e-05, "loss": 2.5202, "step": 54837 }, { "epoch": 4.3, "learning_rate": 5.6999921587077554e-05, "loss": 2.6364, "step": 54838 }, { "epoch": 4.3, "learning_rate": 5.699913745785306e-05, "loss": 2.6945, "step": 54839 }, { "epoch": 4.3, "learning_rate": 5.699835332862856e-05, "loss": 2.5738, "step": 54840 }, { "epoch": 4.3, "learning_rate": 5.6997569199404065e-05, "loss": 2.5605, "step": 54841 }, { "epoch": 4.3, "learning_rate": 5.699678507017957e-05, "loss": 2.6211, "step": 54842 }, { "epoch": 4.3, "learning_rate": 5.699600094095507e-05, "loss": 2.891, "step": 54843 }, { "epoch": 4.3, "learning_rate": 5.6995216811730576e-05, "loss": 2.6185, "step": 54844 }, { "epoch": 4.3, "learning_rate": 5.699443268250608e-05, "loss": 2.5707, "step": 54845 }, { "epoch": 4.3, "learning_rate": 5.699364855328159e-05, "loss": 2.5314, "step": 54846 }, { "epoch": 4.3, "learning_rate": 5.699286442405709e-05, "loss": 2.8319, "step": 54847 }, { "epoch": 4.3, "learning_rate": 5.69920802948326e-05, "loss": 2.78, "step": 54848 }, { "epoch": 4.3, "learning_rate": 5.69912961656081e-05, "loss": 2.554, "step": 54849 }, { "epoch": 4.3, "learning_rate": 5.6990512036383604e-05, "loss": 2.5959, "step": 54850 }, { "epoch": 4.3, "learning_rate": 5.698972790715911e-05, "loss": 2.7363, "step": 54851 }, { "epoch": 4.3, "learning_rate": 5.698894377793461e-05, "loss": 2.6218, "step": 54852 }, { "epoch": 4.3, "learning_rate": 5.6988159648710115e-05, "loss": 2.4821, "step": 54853 }, { "epoch": 4.3, "learning_rate": 5.6987375519485606e-05, "loss": 2.7776, "step": 54854 }, { "epoch": 4.3, "learning_rate": 5.698659139026111e-05, "loss": 2.5193, "step": 54855 }, { "epoch": 4.3, "learning_rate": 5.698580726103662e-05, "loss": 2.5654, "step": 54856 }, { "epoch": 4.3, "learning_rate": 5.698502313181212e-05, "loss": 2.4986, "step": 54857 }, { "epoch": 4.3, "learning_rate": 5.698423900258763e-05, "loss": 2.5393, "step": 54858 }, { "epoch": 4.3, "learning_rate": 5.698345487336313e-05, "loss": 2.7781, "step": 54859 }, { "epoch": 4.3, "learning_rate": 5.6982670744138634e-05, "loss": 2.6407, "step": 54860 }, { "epoch": 4.3, "learning_rate": 5.698188661491414e-05, "loss": 2.8069, "step": 54861 }, { "epoch": 4.3, "learning_rate": 5.698110248568964e-05, "loss": 2.6268, "step": 54862 }, { "epoch": 4.3, "learning_rate": 5.6980318356465145e-05, "loss": 2.8066, "step": 54863 }, { "epoch": 4.3, "learning_rate": 5.697953422724065e-05, "loss": 2.5679, "step": 54864 }, { "epoch": 4.3, "learning_rate": 5.697875009801615e-05, "loss": 2.6691, "step": 54865 }, { "epoch": 4.3, "learning_rate": 5.6977965968791656e-05, "loss": 2.5603, "step": 54866 }, { "epoch": 4.3, "learning_rate": 5.697718183956716e-05, "loss": 2.9549, "step": 54867 }, { "epoch": 4.3, "learning_rate": 5.6976397710342664e-05, "loss": 2.5907, "step": 54868 }, { "epoch": 4.3, "learning_rate": 5.697561358111817e-05, "loss": 2.484, "step": 54869 }, { "epoch": 4.3, "learning_rate": 5.697482945189367e-05, "loss": 2.7731, "step": 54870 }, { "epoch": 4.3, "learning_rate": 5.697404532266918e-05, "loss": 2.709, "step": 54871 }, { "epoch": 4.3, "learning_rate": 5.6973261193444685e-05, "loss": 2.7301, "step": 54872 }, { "epoch": 4.3, "learning_rate": 5.697247706422019e-05, "loss": 2.7858, "step": 54873 }, { "epoch": 4.3, "learning_rate": 5.697169293499569e-05, "loss": 2.5162, "step": 54874 }, { "epoch": 4.3, "learning_rate": 5.6970908805771196e-05, "loss": 2.6212, "step": 54875 }, { "epoch": 4.3, "learning_rate": 5.69701246765467e-05, "loss": 2.4864, "step": 54876 }, { "epoch": 4.3, "learning_rate": 5.6969340547322204e-05, "loss": 2.8818, "step": 54877 }, { "epoch": 4.3, "learning_rate": 5.696855641809771e-05, "loss": 2.6654, "step": 54878 }, { "epoch": 4.3, "learning_rate": 5.696777228887321e-05, "loss": 2.5679, "step": 54879 }, { "epoch": 4.3, "learning_rate": 5.6966988159648715e-05, "loss": 2.7265, "step": 54880 }, { "epoch": 4.3, "learning_rate": 5.696620403042422e-05, "loss": 2.3981, "step": 54881 }, { "epoch": 4.3, "learning_rate": 5.696541990119972e-05, "loss": 2.6435, "step": 54882 }, { "epoch": 4.3, "learning_rate": 5.6964635771975226e-05, "loss": 2.5886, "step": 54883 }, { "epoch": 4.3, "learning_rate": 5.696385164275073e-05, "loss": 2.9035, "step": 54884 }, { "epoch": 4.3, "learning_rate": 5.696306751352623e-05, "loss": 2.5787, "step": 54885 }, { "epoch": 4.3, "learning_rate": 5.696228338430174e-05, "loss": 2.3832, "step": 54886 }, { "epoch": 4.3, "learning_rate": 5.696149925507725e-05, "loss": 2.6764, "step": 54887 }, { "epoch": 4.3, "learning_rate": 5.696071512585275e-05, "loss": 2.8595, "step": 54888 }, { "epoch": 4.3, "learning_rate": 5.6959930996628255e-05, "loss": 2.7264, "step": 54889 }, { "epoch": 4.3, "learning_rate": 5.6959146867403745e-05, "loss": 2.5979, "step": 54890 }, { "epoch": 4.3, "learning_rate": 5.695836273817925e-05, "loss": 2.5791, "step": 54891 }, { "epoch": 4.3, "learning_rate": 5.695757860895475e-05, "loss": 2.7226, "step": 54892 }, { "epoch": 4.3, "learning_rate": 5.6956794479730256e-05, "loss": 2.6618, "step": 54893 }, { "epoch": 4.3, "learning_rate": 5.695601035050576e-05, "loss": 2.7417, "step": 54894 }, { "epoch": 4.3, "learning_rate": 5.695522622128126e-05, "loss": 2.7213, "step": 54895 }, { "epoch": 4.3, "learning_rate": 5.695444209205677e-05, "loss": 2.5326, "step": 54896 }, { "epoch": 4.3, "learning_rate": 5.695365796283228e-05, "loss": 2.7339, "step": 54897 }, { "epoch": 4.3, "learning_rate": 5.695287383360778e-05, "loss": 2.7709, "step": 54898 }, { "epoch": 4.3, "learning_rate": 5.6952089704383285e-05, "loss": 2.7322, "step": 54899 }, { "epoch": 4.3, "learning_rate": 5.695130557515879e-05, "loss": 2.6044, "step": 54900 }, { "epoch": 4.3, "learning_rate": 5.695052144593429e-05, "loss": 2.624, "step": 54901 }, { "epoch": 4.31, "learning_rate": 5.6949737316709796e-05, "loss": 2.7688, "step": 54902 }, { "epoch": 4.31, "learning_rate": 5.69489531874853e-05, "loss": 2.6548, "step": 54903 }, { "epoch": 4.31, "learning_rate": 5.69481690582608e-05, "loss": 2.5469, "step": 54904 }, { "epoch": 4.31, "learning_rate": 5.694738492903631e-05, "loss": 2.7535, "step": 54905 }, { "epoch": 4.31, "learning_rate": 5.694660079981181e-05, "loss": 2.6708, "step": 54906 }, { "epoch": 4.31, "learning_rate": 5.6945816670587314e-05, "loss": 2.9367, "step": 54907 }, { "epoch": 4.31, "learning_rate": 5.694503254136282e-05, "loss": 2.9144, "step": 54908 }, { "epoch": 4.31, "learning_rate": 5.694424841213832e-05, "loss": 2.4751, "step": 54909 }, { "epoch": 4.31, "learning_rate": 5.6943464282913825e-05, "loss": 2.6541, "step": 54910 }, { "epoch": 4.31, "learning_rate": 5.694268015368933e-05, "loss": 2.7307, "step": 54911 }, { "epoch": 4.31, "learning_rate": 5.694189602446484e-05, "loss": 2.6737, "step": 54912 }, { "epoch": 4.31, "learning_rate": 5.694111189524034e-05, "loss": 2.6471, "step": 54913 }, { "epoch": 4.31, "learning_rate": 5.6940327766015847e-05, "loss": 2.6658, "step": 54914 }, { "epoch": 4.31, "learning_rate": 5.693954363679135e-05, "loss": 2.804, "step": 54915 }, { "epoch": 4.31, "learning_rate": 5.6938759507566854e-05, "loss": 2.6619, "step": 54916 }, { "epoch": 4.31, "learning_rate": 5.693797537834236e-05, "loss": 2.585, "step": 54917 }, { "epoch": 4.31, "learning_rate": 5.693719124911786e-05, "loss": 2.4713, "step": 54918 }, { "epoch": 4.31, "learning_rate": 5.6936407119893365e-05, "loss": 2.591, "step": 54919 }, { "epoch": 4.31, "learning_rate": 5.693562299066887e-05, "loss": 2.5823, "step": 54920 }, { "epoch": 4.31, "learning_rate": 5.693483886144437e-05, "loss": 2.4687, "step": 54921 }, { "epoch": 4.31, "learning_rate": 5.6934054732219876e-05, "loss": 2.5058, "step": 54922 }, { "epoch": 4.31, "learning_rate": 5.693327060299538e-05, "loss": 2.8074, "step": 54923 }, { "epoch": 4.31, "learning_rate": 5.693248647377088e-05, "loss": 2.635, "step": 54924 }, { "epoch": 4.31, "learning_rate": 5.693170234454639e-05, "loss": 2.6442, "step": 54925 }, { "epoch": 4.31, "learning_rate": 5.6930918215321884e-05, "loss": 2.914, "step": 54926 }, { "epoch": 4.31, "learning_rate": 5.693013408609739e-05, "loss": 2.5489, "step": 54927 }, { "epoch": 4.31, "learning_rate": 5.692934995687289e-05, "loss": 2.5833, "step": 54928 }, { "epoch": 4.31, "learning_rate": 5.6928565827648395e-05, "loss": 2.5286, "step": 54929 }, { "epoch": 4.31, "learning_rate": 5.69277816984239e-05, "loss": 2.635, "step": 54930 }, { "epoch": 4.31, "learning_rate": 5.69269975691994e-05, "loss": 2.8532, "step": 54931 }, { "epoch": 4.31, "learning_rate": 5.6926213439974906e-05, "loss": 2.4289, "step": 54932 }, { "epoch": 4.31, "learning_rate": 5.692542931075041e-05, "loss": 2.6086, "step": 54933 }, { "epoch": 4.31, "learning_rate": 5.692464518152591e-05, "loss": 2.5803, "step": 54934 }, { "epoch": 4.31, "learning_rate": 5.692386105230142e-05, "loss": 2.7282, "step": 54935 }, { "epoch": 4.31, "learning_rate": 5.692307692307692e-05, "loss": 2.5343, "step": 54936 }, { "epoch": 4.31, "learning_rate": 5.6922292793852424e-05, "loss": 2.6644, "step": 54937 }, { "epoch": 4.31, "learning_rate": 5.6921508664627935e-05, "loss": 2.5269, "step": 54938 }, { "epoch": 4.31, "learning_rate": 5.692072453540344e-05, "loss": 2.7135, "step": 54939 }, { "epoch": 4.31, "learning_rate": 5.691994040617894e-05, "loss": 2.5892, "step": 54940 }, { "epoch": 4.31, "learning_rate": 5.6919156276954446e-05, "loss": 2.442, "step": 54941 }, { "epoch": 4.31, "learning_rate": 5.691837214772995e-05, "loss": 2.9595, "step": 54942 }, { "epoch": 4.31, "learning_rate": 5.691758801850545e-05, "loss": 2.6218, "step": 54943 }, { "epoch": 4.31, "learning_rate": 5.691680388928096e-05, "loss": 2.7027, "step": 54944 }, { "epoch": 4.31, "learning_rate": 5.691601976005646e-05, "loss": 2.7299, "step": 54945 }, { "epoch": 4.31, "learning_rate": 5.6915235630831964e-05, "loss": 2.6045, "step": 54946 }, { "epoch": 4.31, "learning_rate": 5.691445150160747e-05, "loss": 2.6054, "step": 54947 }, { "epoch": 4.31, "learning_rate": 5.691366737238297e-05, "loss": 2.5846, "step": 54948 }, { "epoch": 4.31, "learning_rate": 5.6912883243158475e-05, "loss": 2.6928, "step": 54949 }, { "epoch": 4.31, "learning_rate": 5.691209911393398e-05, "loss": 2.8361, "step": 54950 }, { "epoch": 4.31, "learning_rate": 5.691131498470948e-05, "loss": 2.7185, "step": 54951 }, { "epoch": 4.31, "learning_rate": 5.6910530855484986e-05, "loss": 2.8981, "step": 54952 }, { "epoch": 4.31, "learning_rate": 5.690974672626049e-05, "loss": 2.6096, "step": 54953 }, { "epoch": 4.31, "learning_rate": 5.6908962597036e-05, "loss": 2.6893, "step": 54954 }, { "epoch": 4.31, "learning_rate": 5.6908178467811504e-05, "loss": 2.5371, "step": 54955 }, { "epoch": 4.31, "learning_rate": 5.690739433858701e-05, "loss": 2.5489, "step": 54956 }, { "epoch": 4.31, "learning_rate": 5.690661020936251e-05, "loss": 2.6229, "step": 54957 }, { "epoch": 4.31, "learning_rate": 5.6905826080138015e-05, "loss": 2.6476, "step": 54958 }, { "epoch": 4.31, "learning_rate": 5.690504195091352e-05, "loss": 2.5604, "step": 54959 }, { "epoch": 4.31, "learning_rate": 5.690425782168902e-05, "loss": 2.6208, "step": 54960 }, { "epoch": 4.31, "learning_rate": 5.6903473692464526e-05, "loss": 2.5872, "step": 54961 }, { "epoch": 4.31, "learning_rate": 5.6902689563240016e-05, "loss": 2.7279, "step": 54962 }, { "epoch": 4.31, "learning_rate": 5.690190543401552e-05, "loss": 2.7482, "step": 54963 }, { "epoch": 4.31, "learning_rate": 5.690112130479103e-05, "loss": 2.6157, "step": 54964 }, { "epoch": 4.31, "learning_rate": 5.6900337175566534e-05, "loss": 2.8728, "step": 54965 }, { "epoch": 4.31, "learning_rate": 5.689955304634204e-05, "loss": 2.8141, "step": 54966 }, { "epoch": 4.31, "learning_rate": 5.689876891711754e-05, "loss": 2.7448, "step": 54967 }, { "epoch": 4.31, "learning_rate": 5.6897984787893045e-05, "loss": 2.7477, "step": 54968 }, { "epoch": 4.31, "learning_rate": 5.689720065866855e-05, "loss": 2.7443, "step": 54969 }, { "epoch": 4.31, "learning_rate": 5.689641652944405e-05, "loss": 2.7219, "step": 54970 }, { "epoch": 4.31, "learning_rate": 5.6895632400219556e-05, "loss": 2.9339, "step": 54971 }, { "epoch": 4.31, "learning_rate": 5.689484827099506e-05, "loss": 2.7038, "step": 54972 }, { "epoch": 4.31, "learning_rate": 5.6894064141770564e-05, "loss": 2.5039, "step": 54973 }, { "epoch": 4.31, "learning_rate": 5.689328001254607e-05, "loss": 2.742, "step": 54974 }, { "epoch": 4.31, "learning_rate": 5.689249588332157e-05, "loss": 2.7023, "step": 54975 }, { "epoch": 4.31, "learning_rate": 5.6891711754097075e-05, "loss": 2.6302, "step": 54976 }, { "epoch": 4.31, "learning_rate": 5.689092762487258e-05, "loss": 2.7539, "step": 54977 }, { "epoch": 4.31, "learning_rate": 5.689014349564808e-05, "loss": 2.5067, "step": 54978 }, { "epoch": 4.31, "learning_rate": 5.688935936642359e-05, "loss": 2.5992, "step": 54979 }, { "epoch": 4.31, "learning_rate": 5.6888575237199096e-05, "loss": 2.7407, "step": 54980 }, { "epoch": 4.31, "learning_rate": 5.68877911079746e-05, "loss": 2.7577, "step": 54981 }, { "epoch": 4.31, "learning_rate": 5.6887006978750103e-05, "loss": 2.735, "step": 54982 }, { "epoch": 4.31, "learning_rate": 5.688622284952561e-05, "loss": 2.4932, "step": 54983 }, { "epoch": 4.31, "learning_rate": 5.688543872030111e-05, "loss": 2.6189, "step": 54984 }, { "epoch": 4.31, "learning_rate": 5.6884654591076614e-05, "loss": 2.6307, "step": 54985 }, { "epoch": 4.31, "learning_rate": 5.688387046185212e-05, "loss": 2.793, "step": 54986 }, { "epoch": 4.31, "learning_rate": 5.688308633262762e-05, "loss": 2.7233, "step": 54987 }, { "epoch": 4.31, "learning_rate": 5.6882302203403125e-05, "loss": 2.7407, "step": 54988 }, { "epoch": 4.31, "learning_rate": 5.688151807417863e-05, "loss": 2.6405, "step": 54989 }, { "epoch": 4.31, "learning_rate": 5.688073394495413e-05, "loss": 2.7318, "step": 54990 }, { "epoch": 4.31, "learning_rate": 5.6879949815729637e-05, "loss": 2.5986, "step": 54991 }, { "epoch": 4.31, "learning_rate": 5.687916568650514e-05, "loss": 2.3937, "step": 54992 }, { "epoch": 4.31, "learning_rate": 5.6878381557280644e-05, "loss": 2.5976, "step": 54993 }, { "epoch": 4.31, "learning_rate": 5.687759742805615e-05, "loss": 2.4654, "step": 54994 }, { "epoch": 4.31, "learning_rate": 5.687681329883166e-05, "loss": 2.6743, "step": 54995 }, { "epoch": 4.31, "learning_rate": 5.687602916960716e-05, "loss": 2.5054, "step": 54996 }, { "epoch": 4.31, "learning_rate": 5.6875245040382665e-05, "loss": 2.6587, "step": 54997 }, { "epoch": 4.31, "learning_rate": 5.6874460911158155e-05, "loss": 2.6295, "step": 54998 }, { "epoch": 4.31, "learning_rate": 5.687367678193366e-05, "loss": 2.5764, "step": 54999 }, { "epoch": 4.31, "learning_rate": 5.687289265270916e-05, "loss": 2.7469, "step": 55000 }, { "epoch": 4.31, "eval_loss": 2.8985750675201416, "eval_runtime": 217.7076, "eval_samples_per_second": 52.051, "eval_steps_per_second": 3.257, "step": 55000 }, { "epoch": 4.31, "learning_rate": 5.6872108523484667e-05, "loss": 2.7738, "step": 55001 }, { "epoch": 4.31, "learning_rate": 5.687132439426017e-05, "loss": 2.6795, "step": 55002 }, { "epoch": 4.31, "learning_rate": 5.6870540265035674e-05, "loss": 2.6624, "step": 55003 }, { "epoch": 4.31, "learning_rate": 5.686975613581118e-05, "loss": 2.5853, "step": 55004 }, { "epoch": 4.31, "learning_rate": 5.686897200658669e-05, "loss": 2.7221, "step": 55005 }, { "epoch": 4.31, "learning_rate": 5.686818787736219e-05, "loss": 2.7018, "step": 55006 }, { "epoch": 4.31, "learning_rate": 5.6867403748137695e-05, "loss": 2.7821, "step": 55007 }, { "epoch": 4.31, "learning_rate": 5.68666196189132e-05, "loss": 2.5667, "step": 55008 }, { "epoch": 4.31, "learning_rate": 5.68658354896887e-05, "loss": 2.3975, "step": 55009 }, { "epoch": 4.31, "learning_rate": 5.6865051360464206e-05, "loss": 2.6421, "step": 55010 }, { "epoch": 4.31, "learning_rate": 5.686426723123971e-05, "loss": 2.5607, "step": 55011 }, { "epoch": 4.31, "learning_rate": 5.6863483102015214e-05, "loss": 2.5993, "step": 55012 }, { "epoch": 4.31, "learning_rate": 5.686269897279072e-05, "loss": 2.8851, "step": 55013 }, { "epoch": 4.31, "learning_rate": 5.686191484356622e-05, "loss": 2.4813, "step": 55014 }, { "epoch": 4.31, "learning_rate": 5.6861130714341725e-05, "loss": 2.6051, "step": 55015 }, { "epoch": 4.31, "learning_rate": 5.686034658511723e-05, "loss": 2.7423, "step": 55016 }, { "epoch": 4.31, "learning_rate": 5.685956245589273e-05, "loss": 2.6055, "step": 55017 }, { "epoch": 4.31, "learning_rate": 5.6858778326668236e-05, "loss": 2.5651, "step": 55018 }, { "epoch": 4.31, "learning_rate": 5.685799419744374e-05, "loss": 2.4864, "step": 55019 }, { "epoch": 4.31, "learning_rate": 5.685721006821924e-05, "loss": 2.4606, "step": 55020 }, { "epoch": 4.31, "learning_rate": 5.6856425938994754e-05, "loss": 2.6184, "step": 55021 }, { "epoch": 4.31, "learning_rate": 5.685564180977026e-05, "loss": 2.711, "step": 55022 }, { "epoch": 4.31, "learning_rate": 5.685485768054576e-05, "loss": 2.6957, "step": 55023 }, { "epoch": 4.31, "learning_rate": 5.6854073551321265e-05, "loss": 2.5829, "step": 55024 }, { "epoch": 4.31, "learning_rate": 5.685328942209677e-05, "loss": 2.6314, "step": 55025 }, { "epoch": 4.31, "learning_rate": 5.685250529287227e-05, "loss": 2.6636, "step": 55026 }, { "epoch": 4.31, "learning_rate": 5.6851721163647776e-05, "loss": 2.8156, "step": 55027 }, { "epoch": 4.31, "learning_rate": 5.685093703442328e-05, "loss": 2.7279, "step": 55028 }, { "epoch": 4.31, "learning_rate": 5.685015290519878e-05, "loss": 2.5347, "step": 55029 }, { "epoch": 4.32, "learning_rate": 5.684936877597429e-05, "loss": 2.5581, "step": 55030 }, { "epoch": 4.32, "learning_rate": 5.684858464674979e-05, "loss": 2.8717, "step": 55031 }, { "epoch": 4.32, "learning_rate": 5.6847800517525294e-05, "loss": 2.5096, "step": 55032 }, { "epoch": 4.32, "learning_rate": 5.68470163883008e-05, "loss": 2.7085, "step": 55033 }, { "epoch": 4.32, "learning_rate": 5.6846232259076295e-05, "loss": 2.8601, "step": 55034 }, { "epoch": 4.32, "learning_rate": 5.68454481298518e-05, "loss": 2.6241, "step": 55035 }, { "epoch": 4.32, "learning_rate": 5.68446640006273e-05, "loss": 2.6819, "step": 55036 }, { "epoch": 4.32, "learning_rate": 5.6843879871402806e-05, "loss": 2.5527, "step": 55037 }, { "epoch": 4.32, "learning_rate": 5.684309574217831e-05, "loss": 2.6704, "step": 55038 }, { "epoch": 4.32, "learning_rate": 5.684231161295381e-05, "loss": 2.7681, "step": 55039 }, { "epoch": 4.32, "learning_rate": 5.684152748372932e-05, "loss": 2.6939, "step": 55040 }, { "epoch": 4.32, "learning_rate": 5.684074335450482e-05, "loss": 2.6063, "step": 55041 }, { "epoch": 4.32, "learning_rate": 5.6839959225280324e-05, "loss": 2.515, "step": 55042 }, { "epoch": 4.32, "learning_rate": 5.683917509605583e-05, "loss": 2.923, "step": 55043 }, { "epoch": 4.32, "learning_rate": 5.683839096683133e-05, "loss": 2.5612, "step": 55044 }, { "epoch": 4.32, "learning_rate": 5.6837606837606835e-05, "loss": 2.6929, "step": 55045 }, { "epoch": 4.32, "learning_rate": 5.6836822708382346e-05, "loss": 2.6948, "step": 55046 }, { "epoch": 4.32, "learning_rate": 5.683603857915785e-05, "loss": 2.7842, "step": 55047 }, { "epoch": 4.32, "learning_rate": 5.683525444993335e-05, "loss": 2.5248, "step": 55048 }, { "epoch": 4.32, "learning_rate": 5.6834470320708857e-05, "loss": 2.7199, "step": 55049 }, { "epoch": 4.32, "learning_rate": 5.683368619148436e-05, "loss": 2.8118, "step": 55050 }, { "epoch": 4.32, "learning_rate": 5.6832902062259864e-05, "loss": 2.7666, "step": 55051 }, { "epoch": 4.32, "learning_rate": 5.683211793303537e-05, "loss": 2.5498, "step": 55052 }, { "epoch": 4.32, "learning_rate": 5.683133380381087e-05, "loss": 2.6564, "step": 55053 }, { "epoch": 4.32, "learning_rate": 5.6830549674586375e-05, "loss": 2.6072, "step": 55054 }, { "epoch": 4.32, "learning_rate": 5.682976554536188e-05, "loss": 2.6755, "step": 55055 }, { "epoch": 4.32, "learning_rate": 5.682898141613738e-05, "loss": 2.5853, "step": 55056 }, { "epoch": 4.32, "learning_rate": 5.6828197286912886e-05, "loss": 2.7523, "step": 55057 }, { "epoch": 4.32, "learning_rate": 5.682741315768839e-05, "loss": 2.4397, "step": 55058 }, { "epoch": 4.32, "learning_rate": 5.682662902846389e-05, "loss": 2.9407, "step": 55059 }, { "epoch": 4.32, "learning_rate": 5.68258448992394e-05, "loss": 2.5882, "step": 55060 }, { "epoch": 4.32, "learning_rate": 5.68250607700149e-05, "loss": 2.8237, "step": 55061 }, { "epoch": 4.32, "learning_rate": 5.682427664079041e-05, "loss": 2.6752, "step": 55062 }, { "epoch": 4.32, "learning_rate": 5.6823492511565915e-05, "loss": 2.7188, "step": 55063 }, { "epoch": 4.32, "learning_rate": 5.682270838234142e-05, "loss": 2.781, "step": 55064 }, { "epoch": 4.32, "learning_rate": 5.682192425311692e-05, "loss": 2.6961, "step": 55065 }, { "epoch": 4.32, "learning_rate": 5.6821140123892426e-05, "loss": 2.6969, "step": 55066 }, { "epoch": 4.32, "learning_rate": 5.682035599466793e-05, "loss": 2.6627, "step": 55067 }, { "epoch": 4.32, "learning_rate": 5.681957186544343e-05, "loss": 2.7101, "step": 55068 }, { "epoch": 4.32, "learning_rate": 5.681878773621894e-05, "loss": 2.552, "step": 55069 }, { "epoch": 4.32, "learning_rate": 5.681800360699443e-05, "loss": 2.5094, "step": 55070 }, { "epoch": 4.32, "learning_rate": 5.681721947776993e-05, "loss": 2.8084, "step": 55071 }, { "epoch": 4.32, "learning_rate": 5.681643534854544e-05, "loss": 2.7777, "step": 55072 }, { "epoch": 4.32, "learning_rate": 5.6815651219320945e-05, "loss": 2.7289, "step": 55073 }, { "epoch": 4.32, "learning_rate": 5.681486709009645e-05, "loss": 2.677, "step": 55074 }, { "epoch": 4.32, "learning_rate": 5.681408296087195e-05, "loss": 2.7344, "step": 55075 }, { "epoch": 4.32, "learning_rate": 5.6813298831647456e-05, "loss": 2.5617, "step": 55076 }, { "epoch": 4.32, "learning_rate": 5.681251470242296e-05, "loss": 2.6482, "step": 55077 }, { "epoch": 4.32, "learning_rate": 5.681173057319846e-05, "loss": 2.7076, "step": 55078 }, { "epoch": 4.32, "learning_rate": 5.681094644397397e-05, "loss": 2.613, "step": 55079 }, { "epoch": 4.32, "learning_rate": 5.681016231474947e-05, "loss": 2.8417, "step": 55080 }, { "epoch": 4.32, "learning_rate": 5.6809378185524974e-05, "loss": 2.5024, "step": 55081 }, { "epoch": 4.32, "learning_rate": 5.680859405630048e-05, "loss": 2.6063, "step": 55082 }, { "epoch": 4.32, "learning_rate": 5.680780992707598e-05, "loss": 2.4747, "step": 55083 }, { "epoch": 4.32, "learning_rate": 5.6807025797851485e-05, "loss": 2.5158, "step": 55084 }, { "epoch": 4.32, "learning_rate": 5.680624166862699e-05, "loss": 2.746, "step": 55085 }, { "epoch": 4.32, "learning_rate": 5.680545753940249e-05, "loss": 2.6915, "step": 55086 }, { "epoch": 4.32, "learning_rate": 5.6804673410178e-05, "loss": 2.8016, "step": 55087 }, { "epoch": 4.32, "learning_rate": 5.680388928095351e-05, "loss": 2.6568, "step": 55088 }, { "epoch": 4.32, "learning_rate": 5.680310515172901e-05, "loss": 2.5736, "step": 55089 }, { "epoch": 4.32, "learning_rate": 5.6802321022504514e-05, "loss": 2.8015, "step": 55090 }, { "epoch": 4.32, "learning_rate": 5.680153689328002e-05, "loss": 2.6889, "step": 55091 }, { "epoch": 4.32, "learning_rate": 5.680075276405552e-05, "loss": 2.5999, "step": 55092 }, { "epoch": 4.32, "learning_rate": 5.6799968634831025e-05, "loss": 2.4049, "step": 55093 }, { "epoch": 4.32, "learning_rate": 5.679918450560653e-05, "loss": 2.4424, "step": 55094 }, { "epoch": 4.32, "learning_rate": 5.679840037638203e-05, "loss": 2.6448, "step": 55095 }, { "epoch": 4.32, "learning_rate": 5.6797616247157536e-05, "loss": 2.7516, "step": 55096 }, { "epoch": 4.32, "learning_rate": 5.679683211793304e-05, "loss": 2.7763, "step": 55097 }, { "epoch": 4.32, "learning_rate": 5.6796047988708544e-05, "loss": 2.6232, "step": 55098 }, { "epoch": 4.32, "learning_rate": 5.679526385948405e-05, "loss": 2.6469, "step": 55099 }, { "epoch": 4.32, "learning_rate": 5.679447973025955e-05, "loss": 2.4522, "step": 55100 }, { "epoch": 4.32, "learning_rate": 5.6793695601035055e-05, "loss": 2.6477, "step": 55101 }, { "epoch": 4.32, "learning_rate": 5.679291147181056e-05, "loss": 2.695, "step": 55102 }, { "epoch": 4.32, "learning_rate": 5.679212734258607e-05, "loss": 2.5359, "step": 55103 }, { "epoch": 4.32, "learning_rate": 5.679134321336157e-05, "loss": 2.5655, "step": 55104 }, { "epoch": 4.32, "learning_rate": 5.6790559084137076e-05, "loss": 2.6549, "step": 55105 }, { "epoch": 4.32, "learning_rate": 5.6789774954912566e-05, "loss": 2.7273, "step": 55106 }, { "epoch": 4.32, "learning_rate": 5.678899082568807e-05, "loss": 2.3943, "step": 55107 }, { "epoch": 4.32, "learning_rate": 5.6788206696463574e-05, "loss": 2.6092, "step": 55108 }, { "epoch": 4.32, "learning_rate": 5.678742256723908e-05, "loss": 2.646, "step": 55109 }, { "epoch": 4.32, "learning_rate": 5.678663843801458e-05, "loss": 2.6271, "step": 55110 }, { "epoch": 4.32, "learning_rate": 5.6785854308790085e-05, "loss": 2.6233, "step": 55111 }, { "epoch": 4.32, "learning_rate": 5.678507017956559e-05, "loss": 2.6747, "step": 55112 }, { "epoch": 4.32, "learning_rate": 5.67842860503411e-05, "loss": 2.7137, "step": 55113 }, { "epoch": 4.32, "learning_rate": 5.67835019211166e-05, "loss": 2.6007, "step": 55114 }, { "epoch": 4.32, "learning_rate": 5.6782717791892106e-05, "loss": 2.6027, "step": 55115 }, { "epoch": 4.32, "learning_rate": 5.678193366266761e-05, "loss": 2.776, "step": 55116 }, { "epoch": 4.32, "learning_rate": 5.6781149533443113e-05, "loss": 2.4397, "step": 55117 }, { "epoch": 4.32, "learning_rate": 5.678036540421862e-05, "loss": 2.7434, "step": 55118 }, { "epoch": 4.32, "learning_rate": 5.677958127499412e-05, "loss": 2.6246, "step": 55119 }, { "epoch": 4.32, "learning_rate": 5.6778797145769624e-05, "loss": 2.6751, "step": 55120 }, { "epoch": 4.32, "learning_rate": 5.677801301654513e-05, "loss": 2.7878, "step": 55121 }, { "epoch": 4.32, "learning_rate": 5.677722888732063e-05, "loss": 2.6876, "step": 55122 }, { "epoch": 4.32, "learning_rate": 5.6776444758096136e-05, "loss": 2.8221, "step": 55123 }, { "epoch": 4.32, "learning_rate": 5.677566062887164e-05, "loss": 2.8264, "step": 55124 }, { "epoch": 4.32, "learning_rate": 5.677487649964714e-05, "loss": 2.7162, "step": 55125 }, { "epoch": 4.32, "learning_rate": 5.6774092370422647e-05, "loss": 2.8243, "step": 55126 }, { "epoch": 4.32, "learning_rate": 5.677330824119815e-05, "loss": 2.5777, "step": 55127 }, { "epoch": 4.32, "learning_rate": 5.6772524111973654e-05, "loss": 2.5561, "step": 55128 }, { "epoch": 4.32, "learning_rate": 5.6771739982749164e-05, "loss": 2.5578, "step": 55129 }, { "epoch": 4.32, "learning_rate": 5.677095585352467e-05, "loss": 2.5959, "step": 55130 }, { "epoch": 4.32, "learning_rate": 5.677017172430017e-05, "loss": 2.6946, "step": 55131 }, { "epoch": 4.32, "learning_rate": 5.6769387595075675e-05, "loss": 2.5642, "step": 55132 }, { "epoch": 4.32, "learning_rate": 5.676860346585118e-05, "loss": 2.6783, "step": 55133 }, { "epoch": 4.32, "learning_rate": 5.676781933662668e-05, "loss": 2.6918, "step": 55134 }, { "epoch": 4.32, "learning_rate": 5.6767035207402186e-05, "loss": 2.6342, "step": 55135 }, { "epoch": 4.32, "learning_rate": 5.676625107817769e-05, "loss": 2.6416, "step": 55136 }, { "epoch": 4.32, "learning_rate": 5.6765466948953194e-05, "loss": 2.7547, "step": 55137 }, { "epoch": 4.32, "learning_rate": 5.67646828197287e-05, "loss": 2.6525, "step": 55138 }, { "epoch": 4.32, "learning_rate": 5.67638986905042e-05, "loss": 2.7192, "step": 55139 }, { "epoch": 4.32, "learning_rate": 5.6763114561279705e-05, "loss": 2.7771, "step": 55140 }, { "epoch": 4.32, "learning_rate": 5.676233043205521e-05, "loss": 2.5781, "step": 55141 }, { "epoch": 4.32, "learning_rate": 5.6761546302830705e-05, "loss": 2.9203, "step": 55142 }, { "epoch": 4.32, "learning_rate": 5.676076217360621e-05, "loss": 2.3306, "step": 55143 }, { "epoch": 4.32, "learning_rate": 5.675997804438171e-05, "loss": 2.5169, "step": 55144 }, { "epoch": 4.32, "learning_rate": 5.6759193915157216e-05, "loss": 2.6094, "step": 55145 }, { "epoch": 4.32, "learning_rate": 5.675840978593272e-05, "loss": 2.7761, "step": 55146 }, { "epoch": 4.32, "learning_rate": 5.6757625656708224e-05, "loss": 2.6142, "step": 55147 }, { "epoch": 4.32, "learning_rate": 5.675684152748373e-05, "loss": 2.6462, "step": 55148 }, { "epoch": 4.32, "learning_rate": 5.675605739825923e-05, "loss": 2.6081, "step": 55149 }, { "epoch": 4.32, "learning_rate": 5.6755273269034735e-05, "loss": 2.5709, "step": 55150 }, { "epoch": 4.32, "learning_rate": 5.675448913981024e-05, "loss": 2.6351, "step": 55151 }, { "epoch": 4.32, "learning_rate": 5.675370501058574e-05, "loss": 2.77, "step": 55152 }, { "epoch": 4.32, "learning_rate": 5.6752920881361246e-05, "loss": 2.6621, "step": 55153 }, { "epoch": 4.32, "learning_rate": 5.6752136752136756e-05, "loss": 2.6931, "step": 55154 }, { "epoch": 4.32, "learning_rate": 5.675135262291226e-05, "loss": 2.7723, "step": 55155 }, { "epoch": 4.32, "learning_rate": 5.6750568493687764e-05, "loss": 2.6906, "step": 55156 }, { "epoch": 4.33, "learning_rate": 5.674978436446327e-05, "loss": 2.6835, "step": 55157 }, { "epoch": 4.33, "learning_rate": 5.674900023523877e-05, "loss": 2.6469, "step": 55158 }, { "epoch": 4.33, "learning_rate": 5.6748216106014275e-05, "loss": 2.6222, "step": 55159 }, { "epoch": 4.33, "learning_rate": 5.674743197678978e-05, "loss": 2.5773, "step": 55160 }, { "epoch": 4.33, "learning_rate": 5.674664784756528e-05, "loss": 2.5833, "step": 55161 }, { "epoch": 4.33, "learning_rate": 5.6745863718340786e-05, "loss": 2.669, "step": 55162 }, { "epoch": 4.33, "learning_rate": 5.674507958911629e-05, "loss": 2.6808, "step": 55163 }, { "epoch": 4.33, "learning_rate": 5.674429545989179e-05, "loss": 2.5817, "step": 55164 }, { "epoch": 4.33, "learning_rate": 5.67435113306673e-05, "loss": 2.6178, "step": 55165 }, { "epoch": 4.33, "learning_rate": 5.67427272014428e-05, "loss": 2.7427, "step": 55166 }, { "epoch": 4.33, "learning_rate": 5.6741943072218304e-05, "loss": 2.7192, "step": 55167 }, { "epoch": 4.33, "learning_rate": 5.674115894299381e-05, "loss": 2.6946, "step": 55168 }, { "epoch": 4.33, "learning_rate": 5.674037481376931e-05, "loss": 2.9545, "step": 55169 }, { "epoch": 4.33, "learning_rate": 5.673959068454482e-05, "loss": 2.6283, "step": 55170 }, { "epoch": 4.33, "learning_rate": 5.6738806555320326e-05, "loss": 2.5047, "step": 55171 }, { "epoch": 4.33, "learning_rate": 5.673802242609583e-05, "loss": 2.6741, "step": 55172 }, { "epoch": 4.33, "learning_rate": 5.673723829687133e-05, "loss": 2.5907, "step": 55173 }, { "epoch": 4.33, "learning_rate": 5.673645416764684e-05, "loss": 2.7296, "step": 55174 }, { "epoch": 4.33, "learning_rate": 5.673567003842234e-05, "loss": 2.7628, "step": 55175 }, { "epoch": 4.33, "learning_rate": 5.6734885909197844e-05, "loss": 2.5633, "step": 55176 }, { "epoch": 4.33, "learning_rate": 5.673410177997335e-05, "loss": 2.7266, "step": 55177 }, { "epoch": 4.33, "learning_rate": 5.673331765074884e-05, "loss": 2.5872, "step": 55178 }, { "epoch": 4.33, "learning_rate": 5.673253352152434e-05, "loss": 2.7655, "step": 55179 }, { "epoch": 4.33, "learning_rate": 5.673174939229985e-05, "loss": 2.6368, "step": 55180 }, { "epoch": 4.33, "learning_rate": 5.6730965263075356e-05, "loss": 2.7157, "step": 55181 }, { "epoch": 4.33, "learning_rate": 5.673018113385086e-05, "loss": 2.5254, "step": 55182 }, { "epoch": 4.33, "learning_rate": 5.672939700462636e-05, "loss": 2.7039, "step": 55183 }, { "epoch": 4.33, "learning_rate": 5.672861287540187e-05, "loss": 2.6759, "step": 55184 }, { "epoch": 4.33, "learning_rate": 5.672782874617737e-05, "loss": 2.7268, "step": 55185 }, { "epoch": 4.33, "learning_rate": 5.6727044616952874e-05, "loss": 2.5977, "step": 55186 }, { "epoch": 4.33, "learning_rate": 5.672626048772838e-05, "loss": 2.7387, "step": 55187 }, { "epoch": 4.33, "learning_rate": 5.672547635850388e-05, "loss": 2.7096, "step": 55188 }, { "epoch": 4.33, "learning_rate": 5.6724692229279385e-05, "loss": 2.7363, "step": 55189 }, { "epoch": 4.33, "learning_rate": 5.672390810005489e-05, "loss": 2.7806, "step": 55190 }, { "epoch": 4.33, "learning_rate": 5.672312397083039e-05, "loss": 2.7782, "step": 55191 }, { "epoch": 4.33, "learning_rate": 5.6722339841605896e-05, "loss": 2.503, "step": 55192 }, { "epoch": 4.33, "learning_rate": 5.67215557123814e-05, "loss": 2.7836, "step": 55193 }, { "epoch": 4.33, "learning_rate": 5.6720771583156903e-05, "loss": 2.559, "step": 55194 }, { "epoch": 4.33, "learning_rate": 5.6719987453932414e-05, "loss": 2.5894, "step": 55195 }, { "epoch": 4.33, "learning_rate": 5.671920332470792e-05, "loss": 2.7384, "step": 55196 }, { "epoch": 4.33, "learning_rate": 5.671841919548342e-05, "loss": 2.7948, "step": 55197 }, { "epoch": 4.33, "learning_rate": 5.6717635066258925e-05, "loss": 2.7765, "step": 55198 }, { "epoch": 4.33, "learning_rate": 5.671685093703443e-05, "loss": 2.7149, "step": 55199 }, { "epoch": 4.33, "learning_rate": 5.671606680780993e-05, "loss": 2.6941, "step": 55200 }, { "epoch": 4.33, "learning_rate": 5.6715282678585436e-05, "loss": 2.5, "step": 55201 }, { "epoch": 4.33, "learning_rate": 5.671449854936094e-05, "loss": 2.7063, "step": 55202 }, { "epoch": 4.33, "learning_rate": 5.671371442013644e-05, "loss": 2.4377, "step": 55203 }, { "epoch": 4.33, "learning_rate": 5.671293029091195e-05, "loss": 2.6293, "step": 55204 }, { "epoch": 4.33, "learning_rate": 5.671214616168745e-05, "loss": 2.7087, "step": 55205 }, { "epoch": 4.33, "learning_rate": 5.6711362032462954e-05, "loss": 2.6926, "step": 55206 }, { "epoch": 4.33, "learning_rate": 5.671057790323846e-05, "loss": 2.6836, "step": 55207 }, { "epoch": 4.33, "learning_rate": 5.670979377401396e-05, "loss": 2.576, "step": 55208 }, { "epoch": 4.33, "learning_rate": 5.6709009644789465e-05, "loss": 2.6673, "step": 55209 }, { "epoch": 4.33, "learning_rate": 5.670822551556497e-05, "loss": 2.6502, "step": 55210 }, { "epoch": 4.33, "learning_rate": 5.670744138634048e-05, "loss": 2.7196, "step": 55211 }, { "epoch": 4.33, "learning_rate": 5.670665725711598e-05, "loss": 2.8513, "step": 55212 }, { "epoch": 4.33, "learning_rate": 5.670587312789149e-05, "loss": 2.625, "step": 55213 }, { "epoch": 4.33, "learning_rate": 5.670508899866698e-05, "loss": 2.5436, "step": 55214 }, { "epoch": 4.33, "learning_rate": 5.670430486944248e-05, "loss": 2.6083, "step": 55215 }, { "epoch": 4.33, "learning_rate": 5.6703520740217984e-05, "loss": 2.5805, "step": 55216 }, { "epoch": 4.33, "learning_rate": 5.670273661099349e-05, "loss": 2.7786, "step": 55217 }, { "epoch": 4.33, "learning_rate": 5.670195248176899e-05, "loss": 2.8864, "step": 55218 }, { "epoch": 4.33, "learning_rate": 5.6701168352544495e-05, "loss": 2.78, "step": 55219 }, { "epoch": 4.33, "learning_rate": 5.670038422332e-05, "loss": 2.6468, "step": 55220 }, { "epoch": 4.33, "learning_rate": 5.669960009409551e-05, "loss": 2.6455, "step": 55221 }, { "epoch": 4.33, "learning_rate": 5.669881596487101e-05, "loss": 2.6701, "step": 55222 }, { "epoch": 4.33, "learning_rate": 5.669803183564652e-05, "loss": 2.4568, "step": 55223 }, { "epoch": 4.33, "learning_rate": 5.669724770642202e-05, "loss": 2.4397, "step": 55224 }, { "epoch": 4.33, "learning_rate": 5.6696463577197524e-05, "loss": 2.7678, "step": 55225 }, { "epoch": 4.33, "learning_rate": 5.669567944797303e-05, "loss": 2.775, "step": 55226 }, { "epoch": 4.33, "learning_rate": 5.669489531874853e-05, "loss": 2.7534, "step": 55227 }, { "epoch": 4.33, "learning_rate": 5.6694111189524035e-05, "loss": 2.5966, "step": 55228 }, { "epoch": 4.33, "learning_rate": 5.669332706029954e-05, "loss": 2.7584, "step": 55229 }, { "epoch": 4.33, "learning_rate": 5.669254293107504e-05, "loss": 2.7911, "step": 55230 }, { "epoch": 4.33, "learning_rate": 5.6691758801850546e-05, "loss": 2.5624, "step": 55231 }, { "epoch": 4.33, "learning_rate": 5.669097467262605e-05, "loss": 2.6242, "step": 55232 }, { "epoch": 4.33, "learning_rate": 5.6690190543401554e-05, "loss": 2.4978, "step": 55233 }, { "epoch": 4.33, "learning_rate": 5.668940641417706e-05, "loss": 2.597, "step": 55234 }, { "epoch": 4.33, "learning_rate": 5.668862228495256e-05, "loss": 2.6362, "step": 55235 }, { "epoch": 4.33, "learning_rate": 5.6687838155728065e-05, "loss": 2.6558, "step": 55236 }, { "epoch": 4.33, "learning_rate": 5.6687054026503575e-05, "loss": 2.5694, "step": 55237 }, { "epoch": 4.33, "learning_rate": 5.668626989727908e-05, "loss": 2.526, "step": 55238 }, { "epoch": 4.33, "learning_rate": 5.668548576805458e-05, "loss": 2.5261, "step": 55239 }, { "epoch": 4.33, "learning_rate": 5.6684701638830086e-05, "loss": 2.6514, "step": 55240 }, { "epoch": 4.33, "learning_rate": 5.668391750960559e-05, "loss": 2.7557, "step": 55241 }, { "epoch": 4.33, "learning_rate": 5.6683133380381094e-05, "loss": 2.5924, "step": 55242 }, { "epoch": 4.33, "learning_rate": 5.66823492511566e-05, "loss": 2.7775, "step": 55243 }, { "epoch": 4.33, "learning_rate": 5.66815651219321e-05, "loss": 2.7177, "step": 55244 }, { "epoch": 4.33, "learning_rate": 5.6680780992707605e-05, "loss": 2.6867, "step": 55245 }, { "epoch": 4.33, "learning_rate": 5.667999686348311e-05, "loss": 2.6227, "step": 55246 }, { "epoch": 4.33, "learning_rate": 5.667921273425861e-05, "loss": 2.5341, "step": 55247 }, { "epoch": 4.33, "learning_rate": 5.6678428605034116e-05, "loss": 2.6207, "step": 55248 }, { "epoch": 4.33, "learning_rate": 5.667764447580962e-05, "loss": 2.5611, "step": 55249 }, { "epoch": 4.33, "learning_rate": 5.6676860346585116e-05, "loss": 2.4843, "step": 55250 }, { "epoch": 4.33, "learning_rate": 5.667607621736062e-05, "loss": 2.8584, "step": 55251 }, { "epoch": 4.33, "learning_rate": 5.6675292088136124e-05, "loss": 2.7909, "step": 55252 }, { "epoch": 4.33, "learning_rate": 5.667450795891163e-05, "loss": 2.5855, "step": 55253 }, { "epoch": 4.33, "learning_rate": 5.667372382968713e-05, "loss": 2.8999, "step": 55254 }, { "epoch": 4.33, "learning_rate": 5.6672939700462635e-05, "loss": 2.8666, "step": 55255 }, { "epoch": 4.33, "learning_rate": 5.667215557123814e-05, "loss": 2.7606, "step": 55256 }, { "epoch": 4.33, "learning_rate": 5.667137144201364e-05, "loss": 2.5138, "step": 55257 }, { "epoch": 4.33, "learning_rate": 5.6670587312789146e-05, "loss": 2.429, "step": 55258 }, { "epoch": 4.33, "learning_rate": 5.666980318356465e-05, "loss": 2.7376, "step": 55259 }, { "epoch": 4.33, "learning_rate": 5.666901905434015e-05, "loss": 2.5645, "step": 55260 }, { "epoch": 4.33, "learning_rate": 5.6668234925115657e-05, "loss": 2.455, "step": 55261 }, { "epoch": 4.33, "learning_rate": 5.666745079589117e-05, "loss": 2.7869, "step": 55262 }, { "epoch": 4.33, "learning_rate": 5.666666666666667e-05, "loss": 3.0023, "step": 55263 }, { "epoch": 4.33, "learning_rate": 5.6665882537442174e-05, "loss": 2.7151, "step": 55264 }, { "epoch": 4.33, "learning_rate": 5.666509840821768e-05, "loss": 2.5518, "step": 55265 }, { "epoch": 4.33, "learning_rate": 5.666431427899318e-05, "loss": 2.6472, "step": 55266 }, { "epoch": 4.33, "learning_rate": 5.6663530149768685e-05, "loss": 2.5947, "step": 55267 }, { "epoch": 4.33, "learning_rate": 5.666274602054419e-05, "loss": 2.5973, "step": 55268 }, { "epoch": 4.33, "learning_rate": 5.666196189131969e-05, "loss": 2.6693, "step": 55269 }, { "epoch": 4.33, "learning_rate": 5.6661177762095196e-05, "loss": 2.6583, "step": 55270 }, { "epoch": 4.33, "learning_rate": 5.66603936328707e-05, "loss": 2.5796, "step": 55271 }, { "epoch": 4.33, "learning_rate": 5.6659609503646204e-05, "loss": 2.5579, "step": 55272 }, { "epoch": 4.33, "learning_rate": 5.665882537442171e-05, "loss": 2.4953, "step": 55273 }, { "epoch": 4.33, "learning_rate": 5.665804124519721e-05, "loss": 2.6317, "step": 55274 }, { "epoch": 4.33, "learning_rate": 5.6657257115972715e-05, "loss": 2.5333, "step": 55275 }, { "epoch": 4.33, "learning_rate": 5.665647298674822e-05, "loss": 2.8057, "step": 55276 }, { "epoch": 4.33, "learning_rate": 5.665568885752372e-05, "loss": 2.7371, "step": 55277 }, { "epoch": 4.33, "learning_rate": 5.665490472829923e-05, "loss": 2.7401, "step": 55278 }, { "epoch": 4.33, "learning_rate": 5.6654120599074736e-05, "loss": 2.9615, "step": 55279 }, { "epoch": 4.33, "learning_rate": 5.665333646985024e-05, "loss": 2.6024, "step": 55280 }, { "epoch": 4.33, "learning_rate": 5.6652552340625744e-05, "loss": 2.5066, "step": 55281 }, { "epoch": 4.33, "learning_rate": 5.665176821140125e-05, "loss": 2.7063, "step": 55282 }, { "epoch": 4.33, "learning_rate": 5.665098408217675e-05, "loss": 2.5797, "step": 55283 }, { "epoch": 4.33, "learning_rate": 5.6650199952952255e-05, "loss": 2.9127, "step": 55284 }, { "epoch": 4.34, "learning_rate": 5.664941582372776e-05, "loss": 2.6682, "step": 55285 }, { "epoch": 4.34, "learning_rate": 5.664863169450325e-05, "loss": 2.6944, "step": 55286 }, { "epoch": 4.34, "learning_rate": 5.664784756527875e-05, "loss": 2.4979, "step": 55287 }, { "epoch": 4.34, "learning_rate": 5.664706343605426e-05, "loss": 2.6713, "step": 55288 }, { "epoch": 4.34, "learning_rate": 5.6646279306829766e-05, "loss": 2.6056, "step": 55289 }, { "epoch": 4.34, "learning_rate": 5.664549517760527e-05, "loss": 2.7863, "step": 55290 }, { "epoch": 4.34, "learning_rate": 5.6644711048380774e-05, "loss": 2.4961, "step": 55291 }, { "epoch": 4.34, "learning_rate": 5.664392691915628e-05, "loss": 2.781, "step": 55292 }, { "epoch": 4.34, "learning_rate": 5.664314278993178e-05, "loss": 2.8069, "step": 55293 }, { "epoch": 4.34, "learning_rate": 5.6642358660707285e-05, "loss": 2.7039, "step": 55294 }, { "epoch": 4.34, "learning_rate": 5.664157453148279e-05, "loss": 2.5716, "step": 55295 }, { "epoch": 4.34, "learning_rate": 5.664079040225829e-05, "loss": 2.6548, "step": 55296 }, { "epoch": 4.34, "learning_rate": 5.6640006273033796e-05, "loss": 2.6259, "step": 55297 }, { "epoch": 4.34, "learning_rate": 5.66392221438093e-05, "loss": 2.5078, "step": 55298 }, { "epoch": 4.34, "learning_rate": 5.66384380145848e-05, "loss": 2.771, "step": 55299 }, { "epoch": 4.34, "learning_rate": 5.663765388536031e-05, "loss": 2.5299, "step": 55300 }, { "epoch": 4.34, "learning_rate": 5.663686975613581e-05, "loss": 2.5946, "step": 55301 }, { "epoch": 4.34, "learning_rate": 5.6636085626911314e-05, "loss": 2.5696, "step": 55302 }, { "epoch": 4.34, "learning_rate": 5.6635301497686825e-05, "loss": 2.8236, "step": 55303 }, { "epoch": 4.34, "learning_rate": 5.663451736846233e-05, "loss": 2.6527, "step": 55304 }, { "epoch": 4.34, "learning_rate": 5.663373323923783e-05, "loss": 2.6141, "step": 55305 }, { "epoch": 4.34, "learning_rate": 5.6632949110013336e-05, "loss": 2.6394, "step": 55306 }, { "epoch": 4.34, "learning_rate": 5.663216498078884e-05, "loss": 2.7786, "step": 55307 }, { "epoch": 4.34, "learning_rate": 5.663138085156434e-05, "loss": 2.7563, "step": 55308 }, { "epoch": 4.34, "learning_rate": 5.663059672233985e-05, "loss": 2.6955, "step": 55309 }, { "epoch": 4.34, "learning_rate": 5.662981259311535e-05, "loss": 2.7626, "step": 55310 }, { "epoch": 4.34, "learning_rate": 5.6629028463890854e-05, "loss": 2.7505, "step": 55311 }, { "epoch": 4.34, "learning_rate": 5.662824433466636e-05, "loss": 2.5992, "step": 55312 }, { "epoch": 4.34, "learning_rate": 5.662746020544186e-05, "loss": 2.5147, "step": 55313 }, { "epoch": 4.34, "learning_rate": 5.6626676076217365e-05, "loss": 2.6607, "step": 55314 }, { "epoch": 4.34, "learning_rate": 5.662589194699287e-05, "loss": 2.4283, "step": 55315 }, { "epoch": 4.34, "learning_rate": 5.662510781776837e-05, "loss": 2.6703, "step": 55316 }, { "epoch": 4.34, "learning_rate": 5.6624323688543876e-05, "loss": 2.82, "step": 55317 }, { "epoch": 4.34, "learning_rate": 5.662353955931938e-05, "loss": 2.5824, "step": 55318 }, { "epoch": 4.34, "learning_rate": 5.662275543009489e-05, "loss": 2.5721, "step": 55319 }, { "epoch": 4.34, "learning_rate": 5.6621971300870394e-05, "loss": 2.5693, "step": 55320 }, { "epoch": 4.34, "learning_rate": 5.66211871716459e-05, "loss": 2.8314, "step": 55321 }, { "epoch": 4.34, "learning_rate": 5.662040304242139e-05, "loss": 3.1215, "step": 55322 }, { "epoch": 4.34, "learning_rate": 5.661961891319689e-05, "loss": 2.999, "step": 55323 }, { "epoch": 4.34, "learning_rate": 5.6618834783972395e-05, "loss": 2.6837, "step": 55324 }, { "epoch": 4.34, "learning_rate": 5.66180506547479e-05, "loss": 2.6312, "step": 55325 }, { "epoch": 4.34, "learning_rate": 5.66172665255234e-05, "loss": 2.5628, "step": 55326 }, { "epoch": 4.34, "learning_rate": 5.6616482396298906e-05, "loss": 2.9651, "step": 55327 }, { "epoch": 4.34, "learning_rate": 5.661569826707441e-05, "loss": 2.8931, "step": 55328 }, { "epoch": 4.34, "learning_rate": 5.661491413784992e-05, "loss": 2.5485, "step": 55329 }, { "epoch": 4.34, "learning_rate": 5.6614130008625424e-05, "loss": 2.9234, "step": 55330 }, { "epoch": 4.34, "learning_rate": 5.661334587940093e-05, "loss": 2.7697, "step": 55331 }, { "epoch": 4.34, "learning_rate": 5.661256175017643e-05, "loss": 2.6923, "step": 55332 }, { "epoch": 4.34, "learning_rate": 5.6611777620951935e-05, "loss": 2.6532, "step": 55333 }, { "epoch": 4.34, "learning_rate": 5.661099349172744e-05, "loss": 2.5802, "step": 55334 }, { "epoch": 4.34, "learning_rate": 5.661020936250294e-05, "loss": 2.8963, "step": 55335 }, { "epoch": 4.34, "learning_rate": 5.6609425233278446e-05, "loss": 2.6031, "step": 55336 }, { "epoch": 4.34, "learning_rate": 5.660864110405395e-05, "loss": 2.6789, "step": 55337 }, { "epoch": 4.34, "learning_rate": 5.660785697482945e-05, "loss": 2.7798, "step": 55338 }, { "epoch": 4.34, "learning_rate": 5.660707284560496e-05, "loss": 2.7755, "step": 55339 }, { "epoch": 4.34, "learning_rate": 5.660628871638046e-05, "loss": 2.7214, "step": 55340 }, { "epoch": 4.34, "learning_rate": 5.6605504587155964e-05, "loss": 2.8586, "step": 55341 }, { "epoch": 4.34, "learning_rate": 5.660472045793147e-05, "loss": 2.5593, "step": 55342 }, { "epoch": 4.34, "learning_rate": 5.660393632870697e-05, "loss": 2.6074, "step": 55343 }, { "epoch": 4.34, "learning_rate": 5.6603152199482475e-05, "loss": 2.5897, "step": 55344 }, { "epoch": 4.34, "learning_rate": 5.6602368070257986e-05, "loss": 2.7862, "step": 55345 }, { "epoch": 4.34, "learning_rate": 5.660158394103349e-05, "loss": 2.6706, "step": 55346 }, { "epoch": 4.34, "learning_rate": 5.660079981180899e-05, "loss": 2.9134, "step": 55347 }, { "epoch": 4.34, "learning_rate": 5.66000156825845e-05, "loss": 2.6313, "step": 55348 }, { "epoch": 4.34, "learning_rate": 5.659923155336e-05, "loss": 2.7821, "step": 55349 }, { "epoch": 4.34, "learning_rate": 5.6598447424135504e-05, "loss": 2.9552, "step": 55350 }, { "epoch": 4.34, "learning_rate": 5.659766329491101e-05, "loss": 3.0366, "step": 55351 }, { "epoch": 4.34, "learning_rate": 5.659687916568651e-05, "loss": 2.4625, "step": 55352 }, { "epoch": 4.34, "learning_rate": 5.6596095036462015e-05, "loss": 2.6209, "step": 55353 }, { "epoch": 4.34, "learning_rate": 5.659531090723752e-05, "loss": 2.5722, "step": 55354 }, { "epoch": 4.34, "learning_rate": 5.659452677801302e-05, "loss": 2.792, "step": 55355 }, { "epoch": 4.34, "learning_rate": 5.6593742648788526e-05, "loss": 2.4983, "step": 55356 }, { "epoch": 4.34, "learning_rate": 5.659295851956403e-05, "loss": 2.6756, "step": 55357 }, { "epoch": 4.34, "learning_rate": 5.659217439033953e-05, "loss": 2.6353, "step": 55358 }, { "epoch": 4.34, "learning_rate": 5.659139026111503e-05, "loss": 2.8643, "step": 55359 }, { "epoch": 4.34, "learning_rate": 5.6590606131890534e-05, "loss": 2.6592, "step": 55360 }, { "epoch": 4.34, "learning_rate": 5.658982200266604e-05, "loss": 2.6602, "step": 55361 }, { "epoch": 4.34, "learning_rate": 5.658903787344154e-05, "loss": 2.6732, "step": 55362 }, { "epoch": 4.34, "learning_rate": 5.6588253744217045e-05, "loss": 2.6013, "step": 55363 }, { "epoch": 4.34, "learning_rate": 5.658746961499255e-05, "loss": 2.7119, "step": 55364 }, { "epoch": 4.34, "learning_rate": 5.658668548576805e-05, "loss": 2.7228, "step": 55365 }, { "epoch": 4.34, "learning_rate": 5.6585901356543556e-05, "loss": 2.8119, "step": 55366 }, { "epoch": 4.34, "learning_rate": 5.658511722731906e-05, "loss": 2.6714, "step": 55367 }, { "epoch": 4.34, "learning_rate": 5.6584333098094564e-05, "loss": 2.4597, "step": 55368 }, { "epoch": 4.34, "learning_rate": 5.658354896887007e-05, "loss": 2.6462, "step": 55369 }, { "epoch": 4.34, "learning_rate": 5.658276483964558e-05, "loss": 2.7338, "step": 55370 }, { "epoch": 4.34, "learning_rate": 5.658198071042108e-05, "loss": 2.5182, "step": 55371 }, { "epoch": 4.34, "learning_rate": 5.6581196581196585e-05, "loss": 2.6097, "step": 55372 }, { "epoch": 4.34, "learning_rate": 5.658041245197209e-05, "loss": 2.5985, "step": 55373 }, { "epoch": 4.34, "learning_rate": 5.657962832274759e-05, "loss": 2.623, "step": 55374 }, { "epoch": 4.34, "learning_rate": 5.6578844193523096e-05, "loss": 2.6543, "step": 55375 }, { "epoch": 4.34, "learning_rate": 5.65780600642986e-05, "loss": 2.8751, "step": 55376 }, { "epoch": 4.34, "learning_rate": 5.6577275935074104e-05, "loss": 2.7031, "step": 55377 }, { "epoch": 4.34, "learning_rate": 5.657649180584961e-05, "loss": 2.5213, "step": 55378 }, { "epoch": 4.34, "learning_rate": 5.657570767662511e-05, "loss": 2.4961, "step": 55379 }, { "epoch": 4.34, "learning_rate": 5.6574923547400615e-05, "loss": 2.6572, "step": 55380 }, { "epoch": 4.34, "learning_rate": 5.657413941817612e-05, "loss": 2.6692, "step": 55381 }, { "epoch": 4.34, "learning_rate": 5.657335528895162e-05, "loss": 2.784, "step": 55382 }, { "epoch": 4.34, "learning_rate": 5.6572571159727126e-05, "loss": 2.4162, "step": 55383 }, { "epoch": 4.34, "learning_rate": 5.657178703050263e-05, "loss": 2.831, "step": 55384 }, { "epoch": 4.34, "learning_rate": 5.657100290127813e-05, "loss": 2.8615, "step": 55385 }, { "epoch": 4.34, "learning_rate": 5.6570218772053643e-05, "loss": 2.7444, "step": 55386 }, { "epoch": 4.34, "learning_rate": 5.656943464282915e-05, "loss": 2.7325, "step": 55387 }, { "epoch": 4.34, "learning_rate": 5.656865051360465e-05, "loss": 2.7833, "step": 55388 }, { "epoch": 4.34, "learning_rate": 5.6567866384380154e-05, "loss": 2.6172, "step": 55389 }, { "epoch": 4.34, "learning_rate": 5.656708225515566e-05, "loss": 2.6179, "step": 55390 }, { "epoch": 4.34, "learning_rate": 5.656629812593116e-05, "loss": 2.5326, "step": 55391 }, { "epoch": 4.34, "learning_rate": 5.6565513996706665e-05, "loss": 2.493, "step": 55392 }, { "epoch": 4.34, "learning_rate": 5.656472986748217e-05, "loss": 2.625, "step": 55393 }, { "epoch": 4.34, "learning_rate": 5.656394573825766e-05, "loss": 2.5014, "step": 55394 }, { "epoch": 4.34, "learning_rate": 5.656316160903316e-05, "loss": 2.8139, "step": 55395 }, { "epoch": 4.34, "learning_rate": 5.6562377479808673e-05, "loss": 2.9459, "step": 55396 }, { "epoch": 4.34, "learning_rate": 5.656159335058418e-05, "loss": 2.8412, "step": 55397 }, { "epoch": 4.34, "learning_rate": 5.656080922135968e-05, "loss": 2.7134, "step": 55398 }, { "epoch": 4.34, "learning_rate": 5.6560025092135184e-05, "loss": 2.5229, "step": 55399 }, { "epoch": 4.34, "learning_rate": 5.655924096291069e-05, "loss": 2.7409, "step": 55400 }, { "epoch": 4.34, "learning_rate": 5.655845683368619e-05, "loss": 2.7873, "step": 55401 }, { "epoch": 4.34, "learning_rate": 5.6557672704461695e-05, "loss": 2.7741, "step": 55402 }, { "epoch": 4.34, "learning_rate": 5.65568885752372e-05, "loss": 2.7462, "step": 55403 }, { "epoch": 4.34, "learning_rate": 5.65561044460127e-05, "loss": 2.6164, "step": 55404 }, { "epoch": 4.34, "learning_rate": 5.6555320316788207e-05, "loss": 2.534, "step": 55405 }, { "epoch": 4.34, "learning_rate": 5.655453618756371e-05, "loss": 2.618, "step": 55406 }, { "epoch": 4.34, "learning_rate": 5.6553752058339214e-05, "loss": 2.6578, "step": 55407 }, { "epoch": 4.34, "learning_rate": 5.655296792911472e-05, "loss": 2.8835, "step": 55408 }, { "epoch": 4.34, "learning_rate": 5.655218379989022e-05, "loss": 2.6584, "step": 55409 }, { "epoch": 4.34, "learning_rate": 5.6551399670665725e-05, "loss": 2.481, "step": 55410 }, { "epoch": 4.34, "learning_rate": 5.655061554144123e-05, "loss": 2.6645, "step": 55411 }, { "epoch": 4.35, "learning_rate": 5.654983141221674e-05, "loss": 2.6622, "step": 55412 }, { "epoch": 4.35, "learning_rate": 5.654904728299224e-05, "loss": 2.6024, "step": 55413 }, { "epoch": 4.35, "learning_rate": 5.6548263153767746e-05, "loss": 2.4947, "step": 55414 }, { "epoch": 4.35, "learning_rate": 5.654747902454325e-05, "loss": 2.66, "step": 55415 }, { "epoch": 4.35, "learning_rate": 5.6546694895318754e-05, "loss": 2.64, "step": 55416 }, { "epoch": 4.35, "learning_rate": 5.654591076609426e-05, "loss": 2.8606, "step": 55417 }, { "epoch": 4.35, "learning_rate": 5.654512663686976e-05, "loss": 2.6878, "step": 55418 }, { "epoch": 4.35, "learning_rate": 5.6544342507645265e-05, "loss": 2.8476, "step": 55419 }, { "epoch": 4.35, "learning_rate": 5.654355837842077e-05, "loss": 2.6628, "step": 55420 }, { "epoch": 4.35, "learning_rate": 5.654277424919627e-05, "loss": 2.6712, "step": 55421 }, { "epoch": 4.35, "learning_rate": 5.6541990119971776e-05, "loss": 2.827, "step": 55422 }, { "epoch": 4.35, "learning_rate": 5.654120599074728e-05, "loss": 2.5271, "step": 55423 }, { "epoch": 4.35, "learning_rate": 5.654042186152278e-05, "loss": 2.8332, "step": 55424 }, { "epoch": 4.35, "learning_rate": 5.653963773229829e-05, "loss": 2.7394, "step": 55425 }, { "epoch": 4.35, "learning_rate": 5.653885360307379e-05, "loss": 2.5299, "step": 55426 }, { "epoch": 4.35, "learning_rate": 5.65380694738493e-05, "loss": 2.6613, "step": 55427 }, { "epoch": 4.35, "learning_rate": 5.6537285344624805e-05, "loss": 2.6312, "step": 55428 }, { "epoch": 4.35, "learning_rate": 5.653650121540031e-05, "loss": 2.6507, "step": 55429 }, { "epoch": 4.35, "learning_rate": 5.65357170861758e-05, "loss": 2.5439, "step": 55430 }, { "epoch": 4.35, "learning_rate": 5.65349329569513e-05, "loss": 2.5809, "step": 55431 }, { "epoch": 4.35, "learning_rate": 5.6534148827726806e-05, "loss": 2.6294, "step": 55432 }, { "epoch": 4.35, "learning_rate": 5.653336469850231e-05, "loss": 2.6271, "step": 55433 }, { "epoch": 4.35, "learning_rate": 5.653258056927781e-05, "loss": 2.795, "step": 55434 }, { "epoch": 4.35, "learning_rate": 5.653179644005332e-05, "loss": 2.8229, "step": 55435 }, { "epoch": 4.35, "learning_rate": 5.653101231082882e-05, "loss": 2.5925, "step": 55436 }, { "epoch": 4.35, "learning_rate": 5.653022818160433e-05, "loss": 2.4707, "step": 55437 }, { "epoch": 4.35, "learning_rate": 5.6529444052379835e-05, "loss": 2.6229, "step": 55438 }, { "epoch": 4.35, "learning_rate": 5.652865992315534e-05, "loss": 2.7182, "step": 55439 }, { "epoch": 4.35, "learning_rate": 5.652787579393084e-05, "loss": 2.7898, "step": 55440 }, { "epoch": 4.35, "learning_rate": 5.6527091664706346e-05, "loss": 2.7284, "step": 55441 }, { "epoch": 4.35, "learning_rate": 5.652630753548185e-05, "loss": 2.3457, "step": 55442 }, { "epoch": 4.35, "learning_rate": 5.652552340625735e-05, "loss": 2.6465, "step": 55443 }, { "epoch": 4.35, "learning_rate": 5.652473927703286e-05, "loss": 2.5903, "step": 55444 }, { "epoch": 4.35, "learning_rate": 5.652395514780836e-05, "loss": 2.7484, "step": 55445 }, { "epoch": 4.35, "learning_rate": 5.6523171018583864e-05, "loss": 2.8238, "step": 55446 }, { "epoch": 4.35, "learning_rate": 5.652238688935937e-05, "loss": 2.5061, "step": 55447 }, { "epoch": 4.35, "learning_rate": 5.652160276013487e-05, "loss": 2.7266, "step": 55448 }, { "epoch": 4.35, "learning_rate": 5.6520818630910375e-05, "loss": 2.6093, "step": 55449 }, { "epoch": 4.35, "learning_rate": 5.652003450168588e-05, "loss": 2.5139, "step": 55450 }, { "epoch": 4.35, "learning_rate": 5.651925037246138e-05, "loss": 2.6531, "step": 55451 }, { "epoch": 4.35, "learning_rate": 5.6518466243236886e-05, "loss": 2.6403, "step": 55452 }, { "epoch": 4.35, "learning_rate": 5.6517682114012397e-05, "loss": 2.6638, "step": 55453 }, { "epoch": 4.35, "learning_rate": 5.65168979847879e-05, "loss": 2.6432, "step": 55454 }, { "epoch": 4.35, "learning_rate": 5.6516113855563404e-05, "loss": 2.7707, "step": 55455 }, { "epoch": 4.35, "learning_rate": 5.651532972633891e-05, "loss": 2.673, "step": 55456 }, { "epoch": 4.35, "learning_rate": 5.651454559711441e-05, "loss": 2.6926, "step": 55457 }, { "epoch": 4.35, "learning_rate": 5.6513761467889915e-05, "loss": 2.4948, "step": 55458 }, { "epoch": 4.35, "learning_rate": 5.651297733866542e-05, "loss": 2.5194, "step": 55459 }, { "epoch": 4.35, "learning_rate": 5.651219320944092e-05, "loss": 2.7696, "step": 55460 }, { "epoch": 4.35, "learning_rate": 5.6511409080216426e-05, "loss": 2.7162, "step": 55461 }, { "epoch": 4.35, "learning_rate": 5.651062495099193e-05, "loss": 2.7956, "step": 55462 }, { "epoch": 4.35, "learning_rate": 5.650984082176743e-05, "loss": 2.7251, "step": 55463 }, { "epoch": 4.35, "learning_rate": 5.650905669254294e-05, "loss": 2.7376, "step": 55464 }, { "epoch": 4.35, "learning_rate": 5.650827256331844e-05, "loss": 2.6188, "step": 55465 }, { "epoch": 4.35, "learning_rate": 5.650748843409394e-05, "loss": 2.7033, "step": 55466 }, { "epoch": 4.35, "learning_rate": 5.650670430486944e-05, "loss": 2.7505, "step": 55467 }, { "epoch": 4.35, "learning_rate": 5.6505920175644945e-05, "loss": 2.6518, "step": 55468 }, { "epoch": 4.35, "learning_rate": 5.650513604642045e-05, "loss": 2.7943, "step": 55469 }, { "epoch": 4.35, "learning_rate": 5.650435191719595e-05, "loss": 2.5741, "step": 55470 }, { "epoch": 4.35, "learning_rate": 5.6503567787971456e-05, "loss": 2.693, "step": 55471 }, { "epoch": 4.35, "learning_rate": 5.650278365874696e-05, "loss": 2.5383, "step": 55472 }, { "epoch": 4.35, "learning_rate": 5.650199952952246e-05, "loss": 2.6938, "step": 55473 }, { "epoch": 4.35, "learning_rate": 5.650121540029797e-05, "loss": 2.6049, "step": 55474 }, { "epoch": 4.35, "learning_rate": 5.650043127107347e-05, "loss": 2.9405, "step": 55475 }, { "epoch": 4.35, "learning_rate": 5.6499647141848974e-05, "loss": 2.949, "step": 55476 }, { "epoch": 4.35, "learning_rate": 5.649886301262448e-05, "loss": 2.4154, "step": 55477 }, { "epoch": 4.35, "learning_rate": 5.649807888339999e-05, "loss": 2.5585, "step": 55478 }, { "epoch": 4.35, "learning_rate": 5.649729475417549e-05, "loss": 2.7765, "step": 55479 }, { "epoch": 4.35, "learning_rate": 5.6496510624950996e-05, "loss": 2.7021, "step": 55480 }, { "epoch": 4.35, "learning_rate": 5.64957264957265e-05, "loss": 2.3368, "step": 55481 }, { "epoch": 4.35, "learning_rate": 5.6494942366502e-05, "loss": 2.7834, "step": 55482 }, { "epoch": 4.35, "learning_rate": 5.649415823727751e-05, "loss": 2.7059, "step": 55483 }, { "epoch": 4.35, "learning_rate": 5.649337410805301e-05, "loss": 2.66, "step": 55484 }, { "epoch": 4.35, "learning_rate": 5.6492589978828514e-05, "loss": 2.7172, "step": 55485 }, { "epoch": 4.35, "learning_rate": 5.649180584960402e-05, "loss": 2.6169, "step": 55486 }, { "epoch": 4.35, "learning_rate": 5.649102172037952e-05, "loss": 2.7952, "step": 55487 }, { "epoch": 4.35, "learning_rate": 5.6490237591155025e-05, "loss": 2.6345, "step": 55488 }, { "epoch": 4.35, "learning_rate": 5.648945346193053e-05, "loss": 2.9182, "step": 55489 }, { "epoch": 4.35, "learning_rate": 5.648866933270603e-05, "loss": 2.6249, "step": 55490 }, { "epoch": 4.35, "learning_rate": 5.6487885203481536e-05, "loss": 2.9843, "step": 55491 }, { "epoch": 4.35, "learning_rate": 5.648710107425704e-05, "loss": 2.878, "step": 55492 }, { "epoch": 4.35, "learning_rate": 5.6486316945032544e-05, "loss": 2.4373, "step": 55493 }, { "epoch": 4.35, "learning_rate": 5.6485532815808054e-05, "loss": 2.42, "step": 55494 }, { "epoch": 4.35, "learning_rate": 5.648474868658356e-05, "loss": 2.9871, "step": 55495 }, { "epoch": 4.35, "learning_rate": 5.648396455735906e-05, "loss": 2.6949, "step": 55496 }, { "epoch": 4.35, "learning_rate": 5.6483180428134565e-05, "loss": 2.8594, "step": 55497 }, { "epoch": 4.35, "learning_rate": 5.648239629891007e-05, "loss": 2.7373, "step": 55498 }, { "epoch": 4.35, "learning_rate": 5.648161216968557e-05, "loss": 2.8502, "step": 55499 }, { "epoch": 4.35, "learning_rate": 5.6480828040461076e-05, "loss": 2.5657, "step": 55500 }, { "epoch": 4.35, "learning_rate": 5.648004391123658e-05, "loss": 2.79, "step": 55501 }, { "epoch": 4.35, "learning_rate": 5.647925978201207e-05, "loss": 2.5901, "step": 55502 }, { "epoch": 4.35, "learning_rate": 5.6478475652787574e-05, "loss": 2.5945, "step": 55503 }, { "epoch": 4.35, "learning_rate": 5.6477691523563084e-05, "loss": 2.7577, "step": 55504 }, { "epoch": 4.35, "learning_rate": 5.647690739433859e-05, "loss": 2.5551, "step": 55505 }, { "epoch": 4.35, "learning_rate": 5.647612326511409e-05, "loss": 2.7837, "step": 55506 }, { "epoch": 4.35, "learning_rate": 5.6475339135889595e-05, "loss": 2.5698, "step": 55507 }, { "epoch": 4.35, "learning_rate": 5.64745550066651e-05, "loss": 2.5747, "step": 55508 }, { "epoch": 4.35, "learning_rate": 5.64737708774406e-05, "loss": 2.7026, "step": 55509 }, { "epoch": 4.35, "learning_rate": 5.6472986748216106e-05, "loss": 2.561, "step": 55510 }, { "epoch": 4.35, "learning_rate": 5.647220261899161e-05, "loss": 2.7057, "step": 55511 }, { "epoch": 4.35, "learning_rate": 5.6471418489767114e-05, "loss": 2.6316, "step": 55512 }, { "epoch": 4.35, "learning_rate": 5.647063436054262e-05, "loss": 2.6331, "step": 55513 }, { "epoch": 4.35, "learning_rate": 5.646985023131812e-05, "loss": 2.4762, "step": 55514 }, { "epoch": 4.35, "learning_rate": 5.6469066102093625e-05, "loss": 3.0878, "step": 55515 }, { "epoch": 4.35, "learning_rate": 5.646828197286913e-05, "loss": 2.5418, "step": 55516 }, { "epoch": 4.35, "learning_rate": 5.646749784364463e-05, "loss": 2.7799, "step": 55517 }, { "epoch": 4.35, "learning_rate": 5.6466713714420136e-05, "loss": 2.617, "step": 55518 }, { "epoch": 4.35, "learning_rate": 5.646592958519564e-05, "loss": 2.5819, "step": 55519 }, { "epoch": 4.35, "learning_rate": 5.646514545597115e-05, "loss": 2.6831, "step": 55520 }, { "epoch": 4.35, "learning_rate": 5.6464361326746653e-05, "loss": 2.6967, "step": 55521 }, { "epoch": 4.35, "learning_rate": 5.646357719752216e-05, "loss": 2.6483, "step": 55522 }, { "epoch": 4.35, "learning_rate": 5.646279306829766e-05, "loss": 2.7983, "step": 55523 }, { "epoch": 4.35, "learning_rate": 5.6462008939073164e-05, "loss": 2.6893, "step": 55524 }, { "epoch": 4.35, "learning_rate": 5.646122480984867e-05, "loss": 2.7898, "step": 55525 }, { "epoch": 4.35, "learning_rate": 5.646044068062417e-05, "loss": 2.5212, "step": 55526 }, { "epoch": 4.35, "learning_rate": 5.6459656551399676e-05, "loss": 2.8097, "step": 55527 }, { "epoch": 4.35, "learning_rate": 5.645887242217518e-05, "loss": 2.6039, "step": 55528 }, { "epoch": 4.35, "learning_rate": 5.645808829295068e-05, "loss": 2.6469, "step": 55529 }, { "epoch": 4.35, "learning_rate": 5.6457304163726187e-05, "loss": 2.6637, "step": 55530 }, { "epoch": 4.35, "learning_rate": 5.645652003450169e-05, "loss": 2.6939, "step": 55531 }, { "epoch": 4.35, "learning_rate": 5.6455735905277194e-05, "loss": 2.503, "step": 55532 }, { "epoch": 4.35, "learning_rate": 5.64549517760527e-05, "loss": 2.6649, "step": 55533 }, { "epoch": 4.35, "learning_rate": 5.64541676468282e-05, "loss": 2.624, "step": 55534 }, { "epoch": 4.35, "learning_rate": 5.645338351760371e-05, "loss": 2.7914, "step": 55535 }, { "epoch": 4.35, "learning_rate": 5.6452599388379215e-05, "loss": 2.6723, "step": 55536 }, { "epoch": 4.35, "learning_rate": 5.645181525915472e-05, "loss": 2.5426, "step": 55537 }, { "epoch": 4.35, "learning_rate": 5.645103112993021e-05, "loss": 2.6113, "step": 55538 }, { "epoch": 4.35, "learning_rate": 5.645024700070571e-05, "loss": 2.7426, "step": 55539 }, { "epoch": 4.36, "learning_rate": 5.6449462871481217e-05, "loss": 2.5092, "step": 55540 }, { "epoch": 4.36, "learning_rate": 5.644867874225672e-05, "loss": 2.7436, "step": 55541 }, { "epoch": 4.36, "learning_rate": 5.6447894613032224e-05, "loss": 2.6891, "step": 55542 }, { "epoch": 4.36, "learning_rate": 5.644711048380773e-05, "loss": 2.4607, "step": 55543 }, { "epoch": 4.36, "learning_rate": 5.644632635458323e-05, "loss": 2.718, "step": 55544 }, { "epoch": 4.36, "learning_rate": 5.644554222535874e-05, "loss": 2.6068, "step": 55545 }, { "epoch": 4.36, "learning_rate": 5.6444758096134245e-05, "loss": 2.7779, "step": 55546 }, { "epoch": 4.36, "learning_rate": 5.644397396690975e-05, "loss": 2.8268, "step": 55547 }, { "epoch": 4.36, "learning_rate": 5.644318983768525e-05, "loss": 2.779, "step": 55548 }, { "epoch": 4.36, "learning_rate": 5.6442405708460756e-05, "loss": 2.6159, "step": 55549 }, { "epoch": 4.36, "learning_rate": 5.644162157923626e-05, "loss": 2.8499, "step": 55550 }, { "epoch": 4.36, "learning_rate": 5.6440837450011764e-05, "loss": 2.5869, "step": 55551 }, { "epoch": 4.36, "learning_rate": 5.644005332078727e-05, "loss": 2.6344, "step": 55552 }, { "epoch": 4.36, "learning_rate": 5.643926919156277e-05, "loss": 2.6195, "step": 55553 }, { "epoch": 4.36, "learning_rate": 5.6438485062338275e-05, "loss": 2.6927, "step": 55554 }, { "epoch": 4.36, "learning_rate": 5.643770093311378e-05, "loss": 2.6799, "step": 55555 }, { "epoch": 4.36, "learning_rate": 5.643691680388928e-05, "loss": 2.6436, "step": 55556 }, { "epoch": 4.36, "learning_rate": 5.6436132674664786e-05, "loss": 2.4539, "step": 55557 }, { "epoch": 4.36, "learning_rate": 5.643534854544029e-05, "loss": 2.5595, "step": 55558 }, { "epoch": 4.36, "learning_rate": 5.643456441621579e-05, "loss": 2.6406, "step": 55559 }, { "epoch": 4.36, "learning_rate": 5.64337802869913e-05, "loss": 2.7136, "step": 55560 }, { "epoch": 4.36, "learning_rate": 5.643299615776681e-05, "loss": 2.7142, "step": 55561 }, { "epoch": 4.36, "learning_rate": 5.643221202854231e-05, "loss": 2.8007, "step": 55562 }, { "epoch": 4.36, "learning_rate": 5.6431427899317815e-05, "loss": 2.6989, "step": 55563 }, { "epoch": 4.36, "learning_rate": 5.643064377009332e-05, "loss": 2.6384, "step": 55564 }, { "epoch": 4.36, "learning_rate": 5.642985964086882e-05, "loss": 2.5763, "step": 55565 }, { "epoch": 4.36, "learning_rate": 5.6429075511644326e-05, "loss": 2.758, "step": 55566 }, { "epoch": 4.36, "learning_rate": 5.642829138241983e-05, "loss": 2.5416, "step": 55567 }, { "epoch": 4.36, "learning_rate": 5.642750725319533e-05, "loss": 2.8001, "step": 55568 }, { "epoch": 4.36, "learning_rate": 5.642672312397084e-05, "loss": 2.7861, "step": 55569 }, { "epoch": 4.36, "learning_rate": 5.642593899474634e-05, "loss": 2.5717, "step": 55570 }, { "epoch": 4.36, "learning_rate": 5.6425154865521844e-05, "loss": 2.6642, "step": 55571 }, { "epoch": 4.36, "learning_rate": 5.642437073629735e-05, "loss": 2.7177, "step": 55572 }, { "epoch": 4.36, "learning_rate": 5.642358660707285e-05, "loss": 2.6632, "step": 55573 }, { "epoch": 4.36, "learning_rate": 5.642280247784835e-05, "loss": 2.5246, "step": 55574 }, { "epoch": 4.36, "learning_rate": 5.642201834862385e-05, "loss": 2.7171, "step": 55575 }, { "epoch": 4.36, "learning_rate": 5.6421234219399356e-05, "loss": 2.5725, "step": 55576 }, { "epoch": 4.36, "learning_rate": 5.642045009017486e-05, "loss": 2.7131, "step": 55577 }, { "epoch": 4.36, "learning_rate": 5.641966596095036e-05, "loss": 2.6436, "step": 55578 }, { "epoch": 4.36, "learning_rate": 5.641888183172587e-05, "loss": 2.4643, "step": 55579 }, { "epoch": 4.36, "learning_rate": 5.641809770250137e-05, "loss": 2.6055, "step": 55580 }, { "epoch": 4.36, "learning_rate": 5.6417313573276874e-05, "loss": 2.7046, "step": 55581 }, { "epoch": 4.36, "learning_rate": 5.641652944405238e-05, "loss": 2.5988, "step": 55582 }, { "epoch": 4.36, "learning_rate": 5.641574531482788e-05, "loss": 2.7714, "step": 55583 }, { "epoch": 4.36, "learning_rate": 5.6414961185603385e-05, "loss": 2.4776, "step": 55584 }, { "epoch": 4.36, "learning_rate": 5.641417705637889e-05, "loss": 2.6931, "step": 55585 }, { "epoch": 4.36, "learning_rate": 5.64133929271544e-05, "loss": 2.7665, "step": 55586 }, { "epoch": 4.36, "learning_rate": 5.64126087979299e-05, "loss": 2.6446, "step": 55587 }, { "epoch": 4.36, "learning_rate": 5.641182466870541e-05, "loss": 2.7916, "step": 55588 }, { "epoch": 4.36, "learning_rate": 5.641104053948091e-05, "loss": 2.6683, "step": 55589 }, { "epoch": 4.36, "learning_rate": 5.6410256410256414e-05, "loss": 2.5222, "step": 55590 }, { "epoch": 4.36, "learning_rate": 5.640947228103192e-05, "loss": 2.6762, "step": 55591 }, { "epoch": 4.36, "learning_rate": 5.640868815180742e-05, "loss": 2.7295, "step": 55592 }, { "epoch": 4.36, "learning_rate": 5.6407904022582925e-05, "loss": 2.6198, "step": 55593 }, { "epoch": 4.36, "learning_rate": 5.640711989335843e-05, "loss": 2.5949, "step": 55594 }, { "epoch": 4.36, "learning_rate": 5.640633576413393e-05, "loss": 2.9288, "step": 55595 }, { "epoch": 4.36, "learning_rate": 5.6405551634909436e-05, "loss": 2.6015, "step": 55596 }, { "epoch": 4.36, "learning_rate": 5.640476750568494e-05, "loss": 2.6852, "step": 55597 }, { "epoch": 4.36, "learning_rate": 5.6403983376460443e-05, "loss": 2.6193, "step": 55598 }, { "epoch": 4.36, "learning_rate": 5.640319924723595e-05, "loss": 2.5817, "step": 55599 }, { "epoch": 4.36, "learning_rate": 5.640241511801145e-05, "loss": 2.8509, "step": 55600 }, { "epoch": 4.36, "learning_rate": 5.6401630988786954e-05, "loss": 2.8608, "step": 55601 }, { "epoch": 4.36, "learning_rate": 5.6400846859562465e-05, "loss": 2.8297, "step": 55602 }, { "epoch": 4.36, "learning_rate": 5.640006273033797e-05, "loss": 2.7599, "step": 55603 }, { "epoch": 4.36, "learning_rate": 5.639927860111347e-05, "loss": 2.7024, "step": 55604 }, { "epoch": 4.36, "learning_rate": 5.6398494471888976e-05, "loss": 2.7435, "step": 55605 }, { "epoch": 4.36, "learning_rate": 5.639771034266448e-05, "loss": 2.7289, "step": 55606 }, { "epoch": 4.36, "learning_rate": 5.639692621343998e-05, "loss": 2.5948, "step": 55607 }, { "epoch": 4.36, "learning_rate": 5.639614208421549e-05, "loss": 2.5575, "step": 55608 }, { "epoch": 4.36, "learning_rate": 5.639535795499099e-05, "loss": 2.7796, "step": 55609 }, { "epoch": 4.36, "learning_rate": 5.639457382576648e-05, "loss": 2.7327, "step": 55610 }, { "epoch": 4.36, "learning_rate": 5.6393789696541984e-05, "loss": 2.7602, "step": 55611 }, { "epoch": 4.36, "learning_rate": 5.6393005567317495e-05, "loss": 2.758, "step": 55612 }, { "epoch": 4.36, "learning_rate": 5.6392221438093e-05, "loss": 2.654, "step": 55613 }, { "epoch": 4.36, "learning_rate": 5.63914373088685e-05, "loss": 2.9943, "step": 55614 }, { "epoch": 4.36, "learning_rate": 5.6390653179644006e-05, "loss": 2.745, "step": 55615 }, { "epoch": 4.36, "learning_rate": 5.638986905041951e-05, "loss": 2.5207, "step": 55616 }, { "epoch": 4.36, "learning_rate": 5.638908492119501e-05, "loss": 2.7725, "step": 55617 }, { "epoch": 4.36, "learning_rate": 5.638830079197052e-05, "loss": 2.4155, "step": 55618 }, { "epoch": 4.36, "learning_rate": 5.638751666274602e-05, "loss": 2.8854, "step": 55619 }, { "epoch": 4.36, "learning_rate": 5.6386732533521524e-05, "loss": 2.7928, "step": 55620 }, { "epoch": 4.36, "learning_rate": 5.638594840429703e-05, "loss": 2.5767, "step": 55621 }, { "epoch": 4.36, "learning_rate": 5.638516427507253e-05, "loss": 2.6202, "step": 55622 }, { "epoch": 4.36, "learning_rate": 5.6384380145848035e-05, "loss": 2.4676, "step": 55623 }, { "epoch": 4.36, "learning_rate": 5.638359601662354e-05, "loss": 2.7053, "step": 55624 }, { "epoch": 4.36, "learning_rate": 5.638281188739904e-05, "loss": 2.5708, "step": 55625 }, { "epoch": 4.36, "learning_rate": 5.6382027758174546e-05, "loss": 2.5735, "step": 55626 }, { "epoch": 4.36, "learning_rate": 5.638124362895005e-05, "loss": 2.6827, "step": 55627 }, { "epoch": 4.36, "learning_rate": 5.638045949972556e-05, "loss": 2.6515, "step": 55628 }, { "epoch": 4.36, "learning_rate": 5.6379675370501064e-05, "loss": 2.7193, "step": 55629 }, { "epoch": 4.36, "learning_rate": 5.637889124127657e-05, "loss": 2.4426, "step": 55630 }, { "epoch": 4.36, "learning_rate": 5.637810711205207e-05, "loss": 2.7143, "step": 55631 }, { "epoch": 4.36, "learning_rate": 5.6377322982827575e-05, "loss": 2.8688, "step": 55632 }, { "epoch": 4.36, "learning_rate": 5.637653885360308e-05, "loss": 2.5618, "step": 55633 }, { "epoch": 4.36, "learning_rate": 5.637575472437858e-05, "loss": 2.6704, "step": 55634 }, { "epoch": 4.36, "learning_rate": 5.6374970595154086e-05, "loss": 2.5689, "step": 55635 }, { "epoch": 4.36, "learning_rate": 5.637418646592959e-05, "loss": 2.6068, "step": 55636 }, { "epoch": 4.36, "learning_rate": 5.6373402336705094e-05, "loss": 2.7151, "step": 55637 }, { "epoch": 4.36, "learning_rate": 5.63726182074806e-05, "loss": 2.522, "step": 55638 }, { "epoch": 4.36, "learning_rate": 5.63718340782561e-05, "loss": 2.5933, "step": 55639 }, { "epoch": 4.36, "learning_rate": 5.6371049949031605e-05, "loss": 2.673, "step": 55640 }, { "epoch": 4.36, "learning_rate": 5.637026581980711e-05, "loss": 2.6357, "step": 55641 }, { "epoch": 4.36, "learning_rate": 5.636948169058261e-05, "loss": 2.6227, "step": 55642 }, { "epoch": 4.36, "learning_rate": 5.636869756135812e-05, "loss": 2.5927, "step": 55643 }, { "epoch": 4.36, "learning_rate": 5.6367913432133626e-05, "loss": 2.4721, "step": 55644 }, { "epoch": 4.36, "learning_rate": 5.636712930290913e-05, "loss": 2.7133, "step": 55645 }, { "epoch": 4.36, "learning_rate": 5.636634517368462e-05, "loss": 2.7383, "step": 55646 }, { "epoch": 4.36, "learning_rate": 5.6365561044460124e-05, "loss": 2.5493, "step": 55647 }, { "epoch": 4.36, "learning_rate": 5.636477691523563e-05, "loss": 2.6756, "step": 55648 }, { "epoch": 4.36, "learning_rate": 5.636399278601113e-05, "loss": 2.6733, "step": 55649 }, { "epoch": 4.36, "learning_rate": 5.6363208656786635e-05, "loss": 2.6425, "step": 55650 }, { "epoch": 4.36, "learning_rate": 5.636242452756214e-05, "loss": 2.4672, "step": 55651 }, { "epoch": 4.36, "learning_rate": 5.636164039833764e-05, "loss": 2.6837, "step": 55652 }, { "epoch": 4.36, "learning_rate": 5.636085626911315e-05, "loss": 2.5407, "step": 55653 }, { "epoch": 4.36, "learning_rate": 5.6360072139888656e-05, "loss": 2.8526, "step": 55654 }, { "epoch": 4.36, "learning_rate": 5.635928801066416e-05, "loss": 2.709, "step": 55655 }, { "epoch": 4.36, "learning_rate": 5.6358503881439664e-05, "loss": 2.6658, "step": 55656 }, { "epoch": 4.36, "learning_rate": 5.635771975221517e-05, "loss": 2.6198, "step": 55657 }, { "epoch": 4.36, "learning_rate": 5.635693562299067e-05, "loss": 2.736, "step": 55658 }, { "epoch": 4.36, "learning_rate": 5.6356151493766175e-05, "loss": 2.8633, "step": 55659 }, { "epoch": 4.36, "learning_rate": 5.635536736454168e-05, "loss": 2.6374, "step": 55660 }, { "epoch": 4.36, "learning_rate": 5.635458323531718e-05, "loss": 2.6353, "step": 55661 }, { "epoch": 4.36, "learning_rate": 5.6353799106092686e-05, "loss": 2.7837, "step": 55662 }, { "epoch": 4.36, "learning_rate": 5.635301497686819e-05, "loss": 2.802, "step": 55663 }, { "epoch": 4.36, "learning_rate": 5.635223084764369e-05, "loss": 2.5456, "step": 55664 }, { "epoch": 4.36, "learning_rate": 5.6351446718419197e-05, "loss": 2.7207, "step": 55665 }, { "epoch": 4.36, "learning_rate": 5.63506625891947e-05, "loss": 2.4301, "step": 55666 }, { "epoch": 4.37, "learning_rate": 5.6349878459970204e-05, "loss": 2.5159, "step": 55667 }, { "epoch": 4.37, "learning_rate": 5.634909433074571e-05, "loss": 2.7843, "step": 55668 }, { "epoch": 4.37, "learning_rate": 5.634831020152122e-05, "loss": 2.841, "step": 55669 }, { "epoch": 4.37, "learning_rate": 5.634752607229672e-05, "loss": 2.5403, "step": 55670 }, { "epoch": 4.37, "learning_rate": 5.6346741943072225e-05, "loss": 2.5637, "step": 55671 }, { "epoch": 4.37, "learning_rate": 5.634595781384773e-05, "loss": 2.7538, "step": 55672 }, { "epoch": 4.37, "learning_rate": 5.634517368462323e-05, "loss": 2.7752, "step": 55673 }, { "epoch": 4.37, "learning_rate": 5.6344389555398736e-05, "loss": 2.5972, "step": 55674 }, { "epoch": 4.37, "learning_rate": 5.634360542617424e-05, "loss": 2.8495, "step": 55675 }, { "epoch": 4.37, "learning_rate": 5.6342821296949744e-05, "loss": 2.8781, "step": 55676 }, { "epoch": 4.37, "learning_rate": 5.634203716772525e-05, "loss": 2.7574, "step": 55677 }, { "epoch": 4.37, "learning_rate": 5.634125303850075e-05, "loss": 2.6287, "step": 55678 }, { "epoch": 4.37, "learning_rate": 5.6340468909276255e-05, "loss": 2.6783, "step": 55679 }, { "epoch": 4.37, "learning_rate": 5.633968478005176e-05, "loss": 2.5965, "step": 55680 }, { "epoch": 4.37, "learning_rate": 5.633890065082726e-05, "loss": 2.5247, "step": 55681 }, { "epoch": 4.37, "learning_rate": 5.633811652160276e-05, "loss": 2.4748, "step": 55682 }, { "epoch": 4.37, "learning_rate": 5.633733239237826e-05, "loss": 2.931, "step": 55683 }, { "epoch": 4.37, "learning_rate": 5.6336548263153766e-05, "loss": 2.6604, "step": 55684 }, { "epoch": 4.37, "learning_rate": 5.633576413392927e-05, "loss": 2.7316, "step": 55685 }, { "epoch": 4.37, "learning_rate": 5.6334980004704774e-05, "loss": 2.6609, "step": 55686 }, { "epoch": 4.37, "learning_rate": 5.633419587548028e-05, "loss": 2.5647, "step": 55687 }, { "epoch": 4.37, "learning_rate": 5.633341174625578e-05, "loss": 2.7091, "step": 55688 }, { "epoch": 4.37, "learning_rate": 5.6332627617031285e-05, "loss": 2.5989, "step": 55689 }, { "epoch": 4.37, "learning_rate": 5.633184348780679e-05, "loss": 2.7854, "step": 55690 }, { "epoch": 4.37, "learning_rate": 5.633105935858229e-05, "loss": 2.672, "step": 55691 }, { "epoch": 4.37, "learning_rate": 5.6330275229357796e-05, "loss": 2.5403, "step": 55692 }, { "epoch": 4.37, "learning_rate": 5.63294911001333e-05, "loss": 2.7615, "step": 55693 }, { "epoch": 4.37, "learning_rate": 5.632870697090881e-05, "loss": 2.6284, "step": 55694 }, { "epoch": 4.37, "learning_rate": 5.6327922841684314e-05, "loss": 2.7649, "step": 55695 }, { "epoch": 4.37, "learning_rate": 5.632713871245982e-05, "loss": 2.8254, "step": 55696 }, { "epoch": 4.37, "learning_rate": 5.632635458323532e-05, "loss": 2.6028, "step": 55697 }, { "epoch": 4.37, "learning_rate": 5.6325570454010825e-05, "loss": 2.6939, "step": 55698 }, { "epoch": 4.37, "learning_rate": 5.632478632478633e-05, "loss": 2.7789, "step": 55699 }, { "epoch": 4.37, "learning_rate": 5.632400219556183e-05, "loss": 2.723, "step": 55700 }, { "epoch": 4.37, "learning_rate": 5.6323218066337336e-05, "loss": 2.6614, "step": 55701 }, { "epoch": 4.37, "learning_rate": 5.632243393711284e-05, "loss": 2.6201, "step": 55702 }, { "epoch": 4.37, "learning_rate": 5.632164980788834e-05, "loss": 2.5258, "step": 55703 }, { "epoch": 4.37, "learning_rate": 5.632086567866385e-05, "loss": 2.66, "step": 55704 }, { "epoch": 4.37, "learning_rate": 5.632008154943935e-05, "loss": 2.7116, "step": 55705 }, { "epoch": 4.37, "learning_rate": 5.6319297420214854e-05, "loss": 2.8142, "step": 55706 }, { "epoch": 4.37, "learning_rate": 5.631851329099036e-05, "loss": 2.6464, "step": 55707 }, { "epoch": 4.37, "learning_rate": 5.631772916176586e-05, "loss": 2.57, "step": 55708 }, { "epoch": 4.37, "learning_rate": 5.6316945032541365e-05, "loss": 2.6411, "step": 55709 }, { "epoch": 4.37, "learning_rate": 5.6316160903316876e-05, "loss": 2.6637, "step": 55710 }, { "epoch": 4.37, "learning_rate": 5.631537677409238e-05, "loss": 2.8437, "step": 55711 }, { "epoch": 4.37, "learning_rate": 5.631459264486788e-05, "loss": 2.5961, "step": 55712 }, { "epoch": 4.37, "learning_rate": 5.631380851564339e-05, "loss": 2.5947, "step": 55713 }, { "epoch": 4.37, "learning_rate": 5.631302438641889e-05, "loss": 2.815, "step": 55714 }, { "epoch": 4.37, "learning_rate": 5.6312240257194394e-05, "loss": 2.5974, "step": 55715 }, { "epoch": 4.37, "learning_rate": 5.63114561279699e-05, "loss": 2.64, "step": 55716 }, { "epoch": 4.37, "learning_rate": 5.63106719987454e-05, "loss": 2.9008, "step": 55717 }, { "epoch": 4.37, "learning_rate": 5.630988786952089e-05, "loss": 2.6158, "step": 55718 }, { "epoch": 4.37, "learning_rate": 5.6309103740296395e-05, "loss": 2.8021, "step": 55719 }, { "epoch": 4.37, "learning_rate": 5.6308319611071906e-05, "loss": 2.5603, "step": 55720 }, { "epoch": 4.37, "learning_rate": 5.630753548184741e-05, "loss": 2.4534, "step": 55721 }, { "epoch": 4.37, "learning_rate": 5.630675135262291e-05, "loss": 2.6286, "step": 55722 }, { "epoch": 4.37, "learning_rate": 5.630596722339842e-05, "loss": 2.6221, "step": 55723 }, { "epoch": 4.37, "learning_rate": 5.630518309417392e-05, "loss": 2.669, "step": 55724 }, { "epoch": 4.37, "learning_rate": 5.6304398964949424e-05, "loss": 2.6029, "step": 55725 }, { "epoch": 4.37, "learning_rate": 5.630361483572493e-05, "loss": 2.6248, "step": 55726 }, { "epoch": 4.37, "learning_rate": 5.630283070650043e-05, "loss": 2.6028, "step": 55727 }, { "epoch": 4.37, "learning_rate": 5.6302046577275935e-05, "loss": 2.7863, "step": 55728 }, { "epoch": 4.37, "learning_rate": 5.630126244805144e-05, "loss": 2.511, "step": 55729 }, { "epoch": 4.37, "learning_rate": 5.630047831882694e-05, "loss": 2.7665, "step": 55730 }, { "epoch": 4.37, "learning_rate": 5.6299694189602446e-05, "loss": 2.5603, "step": 55731 }, { "epoch": 4.37, "learning_rate": 5.629891006037795e-05, "loss": 2.5546, "step": 55732 }, { "epoch": 4.37, "learning_rate": 5.6298125931153453e-05, "loss": 2.732, "step": 55733 }, { "epoch": 4.37, "learning_rate": 5.629734180192896e-05, "loss": 2.6298, "step": 55734 }, { "epoch": 4.37, "learning_rate": 5.629655767270446e-05, "loss": 2.9729, "step": 55735 }, { "epoch": 4.37, "learning_rate": 5.629577354347997e-05, "loss": 2.5375, "step": 55736 }, { "epoch": 4.37, "learning_rate": 5.6294989414255475e-05, "loss": 2.6796, "step": 55737 }, { "epoch": 4.37, "learning_rate": 5.629420528503098e-05, "loss": 2.4472, "step": 55738 }, { "epoch": 4.37, "learning_rate": 5.629342115580648e-05, "loss": 2.8183, "step": 55739 }, { "epoch": 4.37, "learning_rate": 5.6292637026581986e-05, "loss": 2.5484, "step": 55740 }, { "epoch": 4.37, "learning_rate": 5.629185289735749e-05, "loss": 2.515, "step": 55741 }, { "epoch": 4.37, "learning_rate": 5.629106876813299e-05, "loss": 2.9115, "step": 55742 }, { "epoch": 4.37, "learning_rate": 5.62902846389085e-05, "loss": 2.765, "step": 55743 }, { "epoch": 4.37, "learning_rate": 5.6289500509684e-05, "loss": 2.6276, "step": 55744 }, { "epoch": 4.37, "learning_rate": 5.6288716380459504e-05, "loss": 2.8319, "step": 55745 }, { "epoch": 4.37, "learning_rate": 5.628793225123501e-05, "loss": 2.7546, "step": 55746 }, { "epoch": 4.37, "learning_rate": 5.628714812201051e-05, "loss": 2.557, "step": 55747 }, { "epoch": 4.37, "learning_rate": 5.6286363992786015e-05, "loss": 2.7804, "step": 55748 }, { "epoch": 4.37, "learning_rate": 5.628557986356152e-05, "loss": 2.7303, "step": 55749 }, { "epoch": 4.37, "learning_rate": 5.628479573433702e-05, "loss": 2.8526, "step": 55750 }, { "epoch": 4.37, "learning_rate": 5.628401160511253e-05, "loss": 2.8736, "step": 55751 }, { "epoch": 4.37, "learning_rate": 5.628322747588804e-05, "loss": 2.5865, "step": 55752 }, { "epoch": 4.37, "learning_rate": 5.628244334666354e-05, "loss": 2.7248, "step": 55753 }, { "epoch": 4.37, "learning_rate": 5.628165921743903e-05, "loss": 2.8358, "step": 55754 }, { "epoch": 4.37, "learning_rate": 5.6280875088214534e-05, "loss": 2.7168, "step": 55755 }, { "epoch": 4.37, "learning_rate": 5.628009095899004e-05, "loss": 2.6345, "step": 55756 }, { "epoch": 4.37, "learning_rate": 5.627930682976554e-05, "loss": 2.579, "step": 55757 }, { "epoch": 4.37, "learning_rate": 5.6278522700541045e-05, "loss": 2.5593, "step": 55758 }, { "epoch": 4.37, "learning_rate": 5.627773857131655e-05, "loss": 2.608, "step": 55759 }, { "epoch": 4.37, "learning_rate": 5.627695444209205e-05, "loss": 2.5396, "step": 55760 }, { "epoch": 4.37, "learning_rate": 5.627617031286756e-05, "loss": 2.5684, "step": 55761 }, { "epoch": 4.37, "learning_rate": 5.627538618364307e-05, "loss": 2.596, "step": 55762 }, { "epoch": 4.37, "learning_rate": 5.627460205441857e-05, "loss": 2.8211, "step": 55763 }, { "epoch": 4.37, "learning_rate": 5.6273817925194074e-05, "loss": 2.5557, "step": 55764 }, { "epoch": 4.37, "learning_rate": 5.627303379596958e-05, "loss": 2.8596, "step": 55765 }, { "epoch": 4.37, "learning_rate": 5.627224966674508e-05, "loss": 2.7026, "step": 55766 }, { "epoch": 4.37, "learning_rate": 5.6271465537520585e-05, "loss": 2.6879, "step": 55767 }, { "epoch": 4.37, "learning_rate": 5.627068140829609e-05, "loss": 2.7721, "step": 55768 }, { "epoch": 4.37, "learning_rate": 5.626989727907159e-05, "loss": 2.7422, "step": 55769 }, { "epoch": 4.37, "learning_rate": 5.6269113149847096e-05, "loss": 2.7634, "step": 55770 }, { "epoch": 4.37, "learning_rate": 5.62683290206226e-05, "loss": 2.5875, "step": 55771 }, { "epoch": 4.37, "learning_rate": 5.6267544891398104e-05, "loss": 2.5741, "step": 55772 }, { "epoch": 4.37, "learning_rate": 5.626676076217361e-05, "loss": 2.7348, "step": 55773 }, { "epoch": 4.37, "learning_rate": 5.626597663294911e-05, "loss": 2.8909, "step": 55774 }, { "epoch": 4.37, "learning_rate": 5.6265192503724615e-05, "loss": 2.6248, "step": 55775 }, { "epoch": 4.37, "learning_rate": 5.626440837450012e-05, "loss": 2.5467, "step": 55776 }, { "epoch": 4.37, "learning_rate": 5.626362424527563e-05, "loss": 2.5318, "step": 55777 }, { "epoch": 4.37, "learning_rate": 5.626284011605113e-05, "loss": 2.7823, "step": 55778 }, { "epoch": 4.37, "learning_rate": 5.6262055986826636e-05, "loss": 2.4686, "step": 55779 }, { "epoch": 4.37, "learning_rate": 5.626127185760214e-05, "loss": 2.6288, "step": 55780 }, { "epoch": 4.37, "learning_rate": 5.6260487728377644e-05, "loss": 2.6297, "step": 55781 }, { "epoch": 4.37, "learning_rate": 5.625970359915315e-05, "loss": 2.5301, "step": 55782 }, { "epoch": 4.37, "learning_rate": 5.625891946992865e-05, "loss": 2.7018, "step": 55783 }, { "epoch": 4.37, "learning_rate": 5.6258135340704155e-05, "loss": 2.6603, "step": 55784 }, { "epoch": 4.37, "learning_rate": 5.625735121147966e-05, "loss": 2.9127, "step": 55785 }, { "epoch": 4.37, "learning_rate": 5.625656708225516e-05, "loss": 2.591, "step": 55786 }, { "epoch": 4.37, "learning_rate": 5.6255782953030666e-05, "loss": 2.6906, "step": 55787 }, { "epoch": 4.37, "learning_rate": 5.625499882380617e-05, "loss": 2.7575, "step": 55788 }, { "epoch": 4.37, "learning_rate": 5.625421469458167e-05, "loss": 2.6501, "step": 55789 }, { "epoch": 4.37, "learning_rate": 5.625343056535717e-05, "loss": 2.7022, "step": 55790 }, { "epoch": 4.37, "learning_rate": 5.6252646436132674e-05, "loss": 2.5679, "step": 55791 }, { "epoch": 4.37, "learning_rate": 5.625186230690818e-05, "loss": 2.7933, "step": 55792 }, { "epoch": 4.37, "learning_rate": 5.625107817768368e-05, "loss": 2.7357, "step": 55793 }, { "epoch": 4.37, "learning_rate": 5.6250294048459185e-05, "loss": 2.7936, "step": 55794 }, { "epoch": 4.38, "learning_rate": 5.624950991923469e-05, "loss": 2.6684, "step": 55795 }, { "epoch": 4.38, "learning_rate": 5.624872579001019e-05, "loss": 2.6652, "step": 55796 }, { "epoch": 4.38, "learning_rate": 5.6247941660785696e-05, "loss": 2.6956, "step": 55797 }, { "epoch": 4.38, "learning_rate": 5.62471575315612e-05, "loss": 2.701, "step": 55798 }, { "epoch": 4.38, "learning_rate": 5.62463734023367e-05, "loss": 2.6115, "step": 55799 }, { "epoch": 4.38, "learning_rate": 5.624558927311221e-05, "loss": 2.7552, "step": 55800 }, { "epoch": 4.38, "learning_rate": 5.624480514388771e-05, "loss": 2.6134, "step": 55801 }, { "epoch": 4.38, "learning_rate": 5.6244021014663214e-05, "loss": 2.6156, "step": 55802 }, { "epoch": 4.38, "learning_rate": 5.6243236885438724e-05, "loss": 2.5673, "step": 55803 }, { "epoch": 4.38, "learning_rate": 5.624245275621423e-05, "loss": 2.6507, "step": 55804 }, { "epoch": 4.38, "learning_rate": 5.624166862698973e-05, "loss": 2.8847, "step": 55805 }, { "epoch": 4.38, "learning_rate": 5.6240884497765235e-05, "loss": 2.7533, "step": 55806 }, { "epoch": 4.38, "learning_rate": 5.624010036854074e-05, "loss": 3.0614, "step": 55807 }, { "epoch": 4.38, "learning_rate": 5.623931623931624e-05, "loss": 2.7969, "step": 55808 }, { "epoch": 4.38, "learning_rate": 5.6238532110091747e-05, "loss": 2.7179, "step": 55809 }, { "epoch": 4.38, "learning_rate": 5.623774798086725e-05, "loss": 2.6297, "step": 55810 }, { "epoch": 4.38, "learning_rate": 5.6236963851642754e-05, "loss": 2.6, "step": 55811 }, { "epoch": 4.38, "learning_rate": 5.623617972241826e-05, "loss": 2.7087, "step": 55812 }, { "epoch": 4.38, "learning_rate": 5.623539559319376e-05, "loss": 2.9065, "step": 55813 }, { "epoch": 4.38, "learning_rate": 5.6234611463969265e-05, "loss": 2.5966, "step": 55814 }, { "epoch": 4.38, "learning_rate": 5.623382733474477e-05, "loss": 2.5334, "step": 55815 }, { "epoch": 4.38, "learning_rate": 5.623304320552027e-05, "loss": 2.6155, "step": 55816 }, { "epoch": 4.38, "learning_rate": 5.6232259076295776e-05, "loss": 2.3195, "step": 55817 }, { "epoch": 4.38, "learning_rate": 5.6231474947071286e-05, "loss": 2.4375, "step": 55818 }, { "epoch": 4.38, "learning_rate": 5.623069081784679e-05, "loss": 2.6991, "step": 55819 }, { "epoch": 4.38, "learning_rate": 5.6229906688622294e-05, "loss": 2.5812, "step": 55820 }, { "epoch": 4.38, "learning_rate": 5.62291225593978e-05, "loss": 2.6093, "step": 55821 }, { "epoch": 4.38, "learning_rate": 5.62283384301733e-05, "loss": 2.7942, "step": 55822 }, { "epoch": 4.38, "learning_rate": 5.6227554300948805e-05, "loss": 2.7102, "step": 55823 }, { "epoch": 4.38, "learning_rate": 5.622677017172431e-05, "loss": 2.6281, "step": 55824 }, { "epoch": 4.38, "learning_rate": 5.622598604249981e-05, "loss": 2.6535, "step": 55825 }, { "epoch": 4.38, "learning_rate": 5.62252019132753e-05, "loss": 2.7094, "step": 55826 }, { "epoch": 4.38, "learning_rate": 5.6224417784050806e-05, "loss": 2.5367, "step": 55827 }, { "epoch": 4.38, "learning_rate": 5.6223633654826316e-05, "loss": 2.6639, "step": 55828 }, { "epoch": 4.38, "learning_rate": 5.622284952560182e-05, "loss": 2.5439, "step": 55829 }, { "epoch": 4.38, "learning_rate": 5.6222065396377324e-05, "loss": 2.7374, "step": 55830 }, { "epoch": 4.38, "learning_rate": 5.622128126715283e-05, "loss": 2.7639, "step": 55831 }, { "epoch": 4.38, "learning_rate": 5.622049713792833e-05, "loss": 2.3906, "step": 55832 }, { "epoch": 4.38, "learning_rate": 5.6219713008703835e-05, "loss": 2.6711, "step": 55833 }, { "epoch": 4.38, "learning_rate": 5.621892887947934e-05, "loss": 2.655, "step": 55834 }, { "epoch": 4.38, "learning_rate": 5.621814475025484e-05, "loss": 2.4794, "step": 55835 }, { "epoch": 4.38, "learning_rate": 5.6217360621030346e-05, "loss": 2.6404, "step": 55836 }, { "epoch": 4.38, "learning_rate": 5.621657649180585e-05, "loss": 2.729, "step": 55837 }, { "epoch": 4.38, "learning_rate": 5.621579236258135e-05, "loss": 2.5303, "step": 55838 }, { "epoch": 4.38, "learning_rate": 5.621500823335686e-05, "loss": 2.5741, "step": 55839 }, { "epoch": 4.38, "learning_rate": 5.621422410413236e-05, "loss": 2.8771, "step": 55840 }, { "epoch": 4.38, "learning_rate": 5.6213439974907864e-05, "loss": 2.5684, "step": 55841 }, { "epoch": 4.38, "learning_rate": 5.621265584568337e-05, "loss": 2.7673, "step": 55842 }, { "epoch": 4.38, "learning_rate": 5.621187171645887e-05, "loss": 2.5766, "step": 55843 }, { "epoch": 4.38, "learning_rate": 5.621108758723438e-05, "loss": 2.706, "step": 55844 }, { "epoch": 4.38, "learning_rate": 5.6210303458009886e-05, "loss": 2.6884, "step": 55845 }, { "epoch": 4.38, "learning_rate": 5.620951932878539e-05, "loss": 2.9572, "step": 55846 }, { "epoch": 4.38, "learning_rate": 5.620873519956089e-05, "loss": 2.7328, "step": 55847 }, { "epoch": 4.38, "learning_rate": 5.62079510703364e-05, "loss": 2.3961, "step": 55848 }, { "epoch": 4.38, "learning_rate": 5.62071669411119e-05, "loss": 2.5765, "step": 55849 }, { "epoch": 4.38, "learning_rate": 5.6206382811887404e-05, "loss": 2.812, "step": 55850 }, { "epoch": 4.38, "learning_rate": 5.620559868266291e-05, "loss": 2.7451, "step": 55851 }, { "epoch": 4.38, "learning_rate": 5.620481455343841e-05, "loss": 2.5365, "step": 55852 }, { "epoch": 4.38, "learning_rate": 5.6204030424213915e-05, "loss": 2.5998, "step": 55853 }, { "epoch": 4.38, "learning_rate": 5.620324629498942e-05, "loss": 2.7399, "step": 55854 }, { "epoch": 4.38, "learning_rate": 5.620246216576492e-05, "loss": 2.6605, "step": 55855 }, { "epoch": 4.38, "learning_rate": 5.6201678036540426e-05, "loss": 2.5581, "step": 55856 }, { "epoch": 4.38, "learning_rate": 5.620089390731593e-05, "loss": 2.8677, "step": 55857 }, { "epoch": 4.38, "learning_rate": 5.6200109778091434e-05, "loss": 2.6218, "step": 55858 }, { "epoch": 4.38, "learning_rate": 5.6199325648866944e-05, "loss": 2.5603, "step": 55859 }, { "epoch": 4.38, "learning_rate": 5.619854151964245e-05, "loss": 2.6085, "step": 55860 }, { "epoch": 4.38, "learning_rate": 5.619775739041795e-05, "loss": 2.6856, "step": 55861 }, { "epoch": 4.38, "learning_rate": 5.619697326119344e-05, "loss": 2.6299, "step": 55862 }, { "epoch": 4.38, "learning_rate": 5.6196189131968945e-05, "loss": 2.6542, "step": 55863 }, { "epoch": 4.38, "learning_rate": 5.619540500274445e-05, "loss": 2.5814, "step": 55864 }, { "epoch": 4.38, "learning_rate": 5.619462087351995e-05, "loss": 2.6046, "step": 55865 }, { "epoch": 4.38, "learning_rate": 5.6193836744295456e-05, "loss": 2.6048, "step": 55866 }, { "epoch": 4.38, "learning_rate": 5.619305261507096e-05, "loss": 2.5801, "step": 55867 }, { "epoch": 4.38, "learning_rate": 5.6192268485846464e-05, "loss": 2.5245, "step": 55868 }, { "epoch": 4.38, "learning_rate": 5.6191484356621974e-05, "loss": 2.6322, "step": 55869 }, { "epoch": 4.38, "learning_rate": 5.619070022739748e-05, "loss": 2.7944, "step": 55870 }, { "epoch": 4.38, "learning_rate": 5.618991609817298e-05, "loss": 2.725, "step": 55871 }, { "epoch": 4.38, "learning_rate": 5.6189131968948485e-05, "loss": 2.6148, "step": 55872 }, { "epoch": 4.38, "learning_rate": 5.618834783972399e-05, "loss": 2.8499, "step": 55873 }, { "epoch": 4.38, "learning_rate": 5.618756371049949e-05, "loss": 2.7338, "step": 55874 }, { "epoch": 4.38, "learning_rate": 5.6186779581274996e-05, "loss": 2.8808, "step": 55875 }, { "epoch": 4.38, "learning_rate": 5.61859954520505e-05, "loss": 3.0125, "step": 55876 }, { "epoch": 4.38, "learning_rate": 5.6185211322826e-05, "loss": 2.6494, "step": 55877 }, { "epoch": 4.38, "learning_rate": 5.618442719360151e-05, "loss": 2.505, "step": 55878 }, { "epoch": 4.38, "learning_rate": 5.618364306437701e-05, "loss": 2.682, "step": 55879 }, { "epoch": 4.38, "learning_rate": 5.6182858935152514e-05, "loss": 2.7038, "step": 55880 }, { "epoch": 4.38, "learning_rate": 5.618207480592802e-05, "loss": 2.5766, "step": 55881 }, { "epoch": 4.38, "learning_rate": 5.618129067670352e-05, "loss": 2.5829, "step": 55882 }, { "epoch": 4.38, "learning_rate": 5.6180506547479025e-05, "loss": 2.7257, "step": 55883 }, { "epoch": 4.38, "learning_rate": 5.617972241825453e-05, "loss": 2.6703, "step": 55884 }, { "epoch": 4.38, "learning_rate": 5.617893828903004e-05, "loss": 2.6045, "step": 55885 }, { "epoch": 4.38, "learning_rate": 5.617815415980554e-05, "loss": 2.5108, "step": 55886 }, { "epoch": 4.38, "learning_rate": 5.617737003058105e-05, "loss": 2.6325, "step": 55887 }, { "epoch": 4.38, "learning_rate": 5.617658590135655e-05, "loss": 2.6655, "step": 55888 }, { "epoch": 4.38, "learning_rate": 5.6175801772132054e-05, "loss": 2.8714, "step": 55889 }, { "epoch": 4.38, "learning_rate": 5.617501764290756e-05, "loss": 2.5427, "step": 55890 }, { "epoch": 4.38, "learning_rate": 5.617423351368306e-05, "loss": 2.6053, "step": 55891 }, { "epoch": 4.38, "learning_rate": 5.6173449384458565e-05, "loss": 2.6505, "step": 55892 }, { "epoch": 4.38, "learning_rate": 5.617266525523407e-05, "loss": 2.8218, "step": 55893 }, { "epoch": 4.38, "learning_rate": 5.617188112600957e-05, "loss": 2.8011, "step": 55894 }, { "epoch": 4.38, "learning_rate": 5.6171096996785076e-05, "loss": 2.8483, "step": 55895 }, { "epoch": 4.38, "learning_rate": 5.617031286756058e-05, "loss": 2.5514, "step": 55896 }, { "epoch": 4.38, "learning_rate": 5.6169528738336084e-05, "loss": 2.6545, "step": 55897 }, { "epoch": 4.38, "learning_rate": 5.616874460911158e-05, "loss": 2.7987, "step": 55898 }, { "epoch": 4.38, "learning_rate": 5.6167960479887084e-05, "loss": 2.5415, "step": 55899 }, { "epoch": 4.38, "learning_rate": 5.616717635066259e-05, "loss": 2.7433, "step": 55900 }, { "epoch": 4.38, "learning_rate": 5.616639222143809e-05, "loss": 2.5782, "step": 55901 }, { "epoch": 4.38, "learning_rate": 5.6165608092213595e-05, "loss": 2.5558, "step": 55902 }, { "epoch": 4.38, "learning_rate": 5.61648239629891e-05, "loss": 2.4797, "step": 55903 }, { "epoch": 4.38, "learning_rate": 5.61640398337646e-05, "loss": 2.73, "step": 55904 }, { "epoch": 4.38, "learning_rate": 5.6163255704540106e-05, "loss": 2.7674, "step": 55905 }, { "epoch": 4.38, "learning_rate": 5.616247157531561e-05, "loss": 2.7442, "step": 55906 }, { "epoch": 4.38, "learning_rate": 5.6161687446091114e-05, "loss": 2.6749, "step": 55907 }, { "epoch": 4.38, "learning_rate": 5.616090331686662e-05, "loss": 2.4609, "step": 55908 }, { "epoch": 4.38, "learning_rate": 5.616011918764212e-05, "loss": 2.9012, "step": 55909 }, { "epoch": 4.38, "learning_rate": 5.6159335058417625e-05, "loss": 2.5287, "step": 55910 }, { "epoch": 4.38, "learning_rate": 5.6158550929193135e-05, "loss": 2.7488, "step": 55911 }, { "epoch": 4.38, "learning_rate": 5.615776679996864e-05, "loss": 2.5081, "step": 55912 }, { "epoch": 4.38, "learning_rate": 5.615698267074414e-05, "loss": 2.6554, "step": 55913 }, { "epoch": 4.38, "learning_rate": 5.6156198541519646e-05, "loss": 2.8381, "step": 55914 }, { "epoch": 4.38, "learning_rate": 5.615541441229515e-05, "loss": 2.5998, "step": 55915 }, { "epoch": 4.38, "learning_rate": 5.6154630283070654e-05, "loss": 2.5935, "step": 55916 }, { "epoch": 4.38, "learning_rate": 5.615384615384616e-05, "loss": 2.664, "step": 55917 }, { "epoch": 4.38, "learning_rate": 5.615306202462166e-05, "loss": 2.5696, "step": 55918 }, { "epoch": 4.38, "learning_rate": 5.6152277895397165e-05, "loss": 2.648, "step": 55919 }, { "epoch": 4.38, "learning_rate": 5.615149376617267e-05, "loss": 2.6033, "step": 55920 }, { "epoch": 4.38, "learning_rate": 5.615070963694817e-05, "loss": 2.5854, "step": 55921 }, { "epoch": 4.39, "learning_rate": 5.6149925507723676e-05, "loss": 2.5708, "step": 55922 }, { "epoch": 4.39, "learning_rate": 5.614914137849918e-05, "loss": 2.6307, "step": 55923 }, { "epoch": 4.39, "learning_rate": 5.614835724927468e-05, "loss": 2.7857, "step": 55924 }, { "epoch": 4.39, "learning_rate": 5.614757312005019e-05, "loss": 2.7079, "step": 55925 }, { "epoch": 4.39, "learning_rate": 5.61467889908257e-05, "loss": 2.7032, "step": 55926 }, { "epoch": 4.39, "learning_rate": 5.61460048616012e-05, "loss": 2.4841, "step": 55927 }, { "epoch": 4.39, "learning_rate": 5.6145220732376704e-05, "loss": 2.7038, "step": 55928 }, { "epoch": 4.39, "learning_rate": 5.614443660315221e-05, "loss": 2.5538, "step": 55929 }, { "epoch": 4.39, "learning_rate": 5.614365247392771e-05, "loss": 2.7283, "step": 55930 }, { "epoch": 4.39, "learning_rate": 5.6142868344703216e-05, "loss": 2.6389, "step": 55931 }, { "epoch": 4.39, "learning_rate": 5.614208421547872e-05, "loss": 2.7142, "step": 55932 }, { "epoch": 4.39, "learning_rate": 5.614130008625422e-05, "loss": 2.6499, "step": 55933 }, { "epoch": 4.39, "learning_rate": 5.614051595702971e-05, "loss": 2.6942, "step": 55934 }, { "epoch": 4.39, "learning_rate": 5.613973182780522e-05, "loss": 2.4611, "step": 55935 }, { "epoch": 4.39, "learning_rate": 5.613894769858073e-05, "loss": 2.663, "step": 55936 }, { "epoch": 4.39, "learning_rate": 5.613816356935623e-05, "loss": 2.6094, "step": 55937 }, { "epoch": 4.39, "learning_rate": 5.6137379440131734e-05, "loss": 2.7308, "step": 55938 }, { "epoch": 4.39, "learning_rate": 5.613659531090724e-05, "loss": 2.5614, "step": 55939 }, { "epoch": 4.39, "learning_rate": 5.613581118168274e-05, "loss": 2.7444, "step": 55940 }, { "epoch": 4.39, "learning_rate": 5.6135027052458246e-05, "loss": 2.5711, "step": 55941 }, { "epoch": 4.39, "learning_rate": 5.613424292323375e-05, "loss": 2.6588, "step": 55942 }, { "epoch": 4.39, "learning_rate": 5.613345879400925e-05, "loss": 2.5482, "step": 55943 }, { "epoch": 4.39, "learning_rate": 5.6132674664784757e-05, "loss": 2.7862, "step": 55944 }, { "epoch": 4.39, "learning_rate": 5.613189053556026e-05, "loss": 2.5072, "step": 55945 }, { "epoch": 4.39, "learning_rate": 5.6131106406335764e-05, "loss": 2.6574, "step": 55946 }, { "epoch": 4.39, "learning_rate": 5.613032227711127e-05, "loss": 2.7982, "step": 55947 }, { "epoch": 4.39, "learning_rate": 5.612953814788677e-05, "loss": 2.5866, "step": 55948 }, { "epoch": 4.39, "learning_rate": 5.6128754018662275e-05, "loss": 2.5952, "step": 55949 }, { "epoch": 4.39, "learning_rate": 5.612796988943778e-05, "loss": 2.5546, "step": 55950 }, { "epoch": 4.39, "learning_rate": 5.612718576021328e-05, "loss": 2.6592, "step": 55951 }, { "epoch": 4.39, "learning_rate": 5.612640163098879e-05, "loss": 2.5913, "step": 55952 }, { "epoch": 4.39, "learning_rate": 5.6125617501764296e-05, "loss": 2.5665, "step": 55953 }, { "epoch": 4.39, "learning_rate": 5.61248333725398e-05, "loss": 2.5999, "step": 55954 }, { "epoch": 4.39, "learning_rate": 5.6124049243315304e-05, "loss": 2.55, "step": 55955 }, { "epoch": 4.39, "learning_rate": 5.612326511409081e-05, "loss": 2.7172, "step": 55956 }, { "epoch": 4.39, "learning_rate": 5.612248098486631e-05, "loss": 2.7732, "step": 55957 }, { "epoch": 4.39, "learning_rate": 5.6121696855641815e-05, "loss": 2.7486, "step": 55958 }, { "epoch": 4.39, "learning_rate": 5.612091272641732e-05, "loss": 2.7529, "step": 55959 }, { "epoch": 4.39, "learning_rate": 5.612012859719282e-05, "loss": 2.7091, "step": 55960 }, { "epoch": 4.39, "learning_rate": 5.6119344467968326e-05, "loss": 2.6322, "step": 55961 }, { "epoch": 4.39, "learning_rate": 5.611856033874383e-05, "loss": 2.8539, "step": 55962 }, { "epoch": 4.39, "learning_rate": 5.611777620951933e-05, "loss": 2.6729, "step": 55963 }, { "epoch": 4.39, "learning_rate": 5.611699208029484e-05, "loss": 2.7492, "step": 55964 }, { "epoch": 4.39, "learning_rate": 5.611620795107034e-05, "loss": 2.4764, "step": 55965 }, { "epoch": 4.39, "learning_rate": 5.6115423821845844e-05, "loss": 2.9497, "step": 55966 }, { "epoch": 4.39, "learning_rate": 5.6114639692621355e-05, "loss": 2.5026, "step": 55967 }, { "epoch": 4.39, "learning_rate": 5.611385556339686e-05, "loss": 2.7579, "step": 55968 }, { "epoch": 4.39, "learning_rate": 5.611307143417236e-05, "loss": 2.5042, "step": 55969 }, { "epoch": 4.39, "learning_rate": 5.611228730494785e-05, "loss": 2.5913, "step": 55970 }, { "epoch": 4.39, "learning_rate": 5.6111503175723356e-05, "loss": 2.6356, "step": 55971 }, { "epoch": 4.39, "learning_rate": 5.611071904649886e-05, "loss": 2.6475, "step": 55972 }, { "epoch": 4.39, "learning_rate": 5.610993491727436e-05, "loss": 2.5773, "step": 55973 }, { "epoch": 4.39, "learning_rate": 5.610915078804987e-05, "loss": 2.7214, "step": 55974 }, { "epoch": 4.39, "learning_rate": 5.610836665882537e-05, "loss": 2.5893, "step": 55975 }, { "epoch": 4.39, "learning_rate": 5.6107582529600874e-05, "loss": 2.7705, "step": 55976 }, { "epoch": 4.39, "learning_rate": 5.6106798400376385e-05, "loss": 2.6735, "step": 55977 }, { "epoch": 4.39, "learning_rate": 5.610601427115189e-05, "loss": 2.8695, "step": 55978 }, { "epoch": 4.39, "learning_rate": 5.610523014192739e-05, "loss": 2.7259, "step": 55979 }, { "epoch": 4.39, "learning_rate": 5.6104446012702896e-05, "loss": 2.7119, "step": 55980 }, { "epoch": 4.39, "learning_rate": 5.61036618834784e-05, "loss": 2.9817, "step": 55981 }, { "epoch": 4.39, "learning_rate": 5.61028777542539e-05, "loss": 2.6059, "step": 55982 }, { "epoch": 4.39, "learning_rate": 5.610209362502941e-05, "loss": 2.65, "step": 55983 }, { "epoch": 4.39, "learning_rate": 5.610130949580491e-05, "loss": 2.4456, "step": 55984 }, { "epoch": 4.39, "learning_rate": 5.6100525366580414e-05, "loss": 2.6819, "step": 55985 }, { "epoch": 4.39, "learning_rate": 5.609974123735592e-05, "loss": 2.5966, "step": 55986 }, { "epoch": 4.39, "learning_rate": 5.609895710813142e-05, "loss": 2.7126, "step": 55987 }, { "epoch": 4.39, "learning_rate": 5.6098172978906925e-05, "loss": 2.4667, "step": 55988 }, { "epoch": 4.39, "learning_rate": 5.609738884968243e-05, "loss": 2.7473, "step": 55989 }, { "epoch": 4.39, "learning_rate": 5.609660472045793e-05, "loss": 2.7368, "step": 55990 }, { "epoch": 4.39, "learning_rate": 5.6095820591233436e-05, "loss": 2.73, "step": 55991 }, { "epoch": 4.39, "learning_rate": 5.609503646200894e-05, "loss": 2.7736, "step": 55992 }, { "epoch": 4.39, "learning_rate": 5.609425233278445e-05, "loss": 2.5976, "step": 55993 }, { "epoch": 4.39, "learning_rate": 5.6093468203559954e-05, "loss": 2.6205, "step": 55994 }, { "epoch": 4.39, "learning_rate": 5.609268407433546e-05, "loss": 2.5463, "step": 55995 }, { "epoch": 4.39, "learning_rate": 5.609189994511096e-05, "loss": 2.6458, "step": 55996 }, { "epoch": 4.39, "learning_rate": 5.6091115815886465e-05, "loss": 2.6285, "step": 55997 }, { "epoch": 4.39, "learning_rate": 5.609033168666197e-05, "loss": 2.7476, "step": 55998 }, { "epoch": 4.39, "learning_rate": 5.608954755743747e-05, "loss": 2.6752, "step": 55999 }, { "epoch": 4.39, "learning_rate": 5.6088763428212976e-05, "loss": 2.7136, "step": 56000 }, { "epoch": 4.39, "eval_loss": 2.8992557525634766, "eval_runtime": 217.6695, "eval_samples_per_second": 52.061, "eval_steps_per_second": 3.257, "step": 56000 }, { "epoch": 4.39, "learning_rate": 5.608797929898848e-05, "loss": 2.7495, "step": 56001 }, { "epoch": 4.39, "learning_rate": 5.6087195169763983e-05, "loss": 2.6006, "step": 56002 }, { "epoch": 4.39, "learning_rate": 5.608641104053949e-05, "loss": 2.7565, "step": 56003 }, { "epoch": 4.39, "learning_rate": 5.608562691131499e-05, "loss": 2.6932, "step": 56004 }, { "epoch": 4.39, "learning_rate": 5.6084842782090494e-05, "loss": 2.5533, "step": 56005 }, { "epoch": 4.39, "learning_rate": 5.608405865286599e-05, "loss": 2.7259, "step": 56006 }, { "epoch": 4.39, "learning_rate": 5.6083274523641495e-05, "loss": 2.6189, "step": 56007 }, { "epoch": 4.39, "learning_rate": 5.6082490394417e-05, "loss": 2.6201, "step": 56008 }, { "epoch": 4.39, "learning_rate": 5.60817062651925e-05, "loss": 2.6926, "step": 56009 }, { "epoch": 4.39, "learning_rate": 5.6080922135968006e-05, "loss": 2.6418, "step": 56010 }, { "epoch": 4.39, "learning_rate": 5.608013800674351e-05, "loss": 2.7041, "step": 56011 }, { "epoch": 4.39, "learning_rate": 5.6079353877519013e-05, "loss": 2.5044, "step": 56012 }, { "epoch": 4.39, "learning_rate": 5.607856974829452e-05, "loss": 2.5887, "step": 56013 }, { "epoch": 4.39, "learning_rate": 5.607778561907002e-05, "loss": 2.4894, "step": 56014 }, { "epoch": 4.39, "learning_rate": 5.6077001489845524e-05, "loss": 2.7595, "step": 56015 }, { "epoch": 4.39, "learning_rate": 5.607621736062103e-05, "loss": 2.58, "step": 56016 }, { "epoch": 4.39, "learning_rate": 5.607543323139653e-05, "loss": 2.7223, "step": 56017 }, { "epoch": 4.39, "learning_rate": 5.6074649102172035e-05, "loss": 2.6044, "step": 56018 }, { "epoch": 4.39, "learning_rate": 5.6073864972947546e-05, "loss": 2.6039, "step": 56019 }, { "epoch": 4.39, "learning_rate": 5.607308084372305e-05, "loss": 2.4668, "step": 56020 }, { "epoch": 4.39, "learning_rate": 5.607229671449855e-05, "loss": 2.6855, "step": 56021 }, { "epoch": 4.39, "learning_rate": 5.607151258527406e-05, "loss": 2.6913, "step": 56022 }, { "epoch": 4.39, "learning_rate": 5.607072845604956e-05, "loss": 2.6438, "step": 56023 }, { "epoch": 4.39, "learning_rate": 5.6069944326825064e-05, "loss": 2.5891, "step": 56024 }, { "epoch": 4.39, "learning_rate": 5.606916019760057e-05, "loss": 2.6687, "step": 56025 }, { "epoch": 4.39, "learning_rate": 5.606837606837607e-05, "loss": 2.7351, "step": 56026 }, { "epoch": 4.39, "learning_rate": 5.6067591939151575e-05, "loss": 2.5371, "step": 56027 }, { "epoch": 4.39, "learning_rate": 5.606680780992708e-05, "loss": 2.582, "step": 56028 }, { "epoch": 4.39, "learning_rate": 5.606602368070258e-05, "loss": 2.6282, "step": 56029 }, { "epoch": 4.39, "learning_rate": 5.6065239551478086e-05, "loss": 2.6848, "step": 56030 }, { "epoch": 4.39, "learning_rate": 5.606445542225359e-05, "loss": 2.5474, "step": 56031 }, { "epoch": 4.39, "learning_rate": 5.6063671293029094e-05, "loss": 2.866, "step": 56032 }, { "epoch": 4.39, "learning_rate": 5.60628871638046e-05, "loss": 2.6846, "step": 56033 }, { "epoch": 4.39, "learning_rate": 5.606210303458011e-05, "loss": 2.812, "step": 56034 }, { "epoch": 4.39, "learning_rate": 5.606131890535561e-05, "loss": 2.6695, "step": 56035 }, { "epoch": 4.39, "learning_rate": 5.6060534776131115e-05, "loss": 2.6568, "step": 56036 }, { "epoch": 4.39, "learning_rate": 5.605975064690662e-05, "loss": 2.5801, "step": 56037 }, { "epoch": 4.39, "learning_rate": 5.605896651768212e-05, "loss": 2.5225, "step": 56038 }, { "epoch": 4.39, "learning_rate": 5.6058182388457626e-05, "loss": 2.5097, "step": 56039 }, { "epoch": 4.39, "learning_rate": 5.605739825923313e-05, "loss": 2.7874, "step": 56040 }, { "epoch": 4.39, "learning_rate": 5.6056614130008634e-05, "loss": 2.4525, "step": 56041 }, { "epoch": 4.39, "learning_rate": 5.6055830000784124e-05, "loss": 2.8267, "step": 56042 }, { "epoch": 4.39, "learning_rate": 5.605504587155963e-05, "loss": 2.5915, "step": 56043 }, { "epoch": 4.39, "learning_rate": 5.605426174233514e-05, "loss": 2.685, "step": 56044 }, { "epoch": 4.39, "learning_rate": 5.605347761311064e-05, "loss": 2.5589, "step": 56045 }, { "epoch": 4.39, "learning_rate": 5.6052693483886145e-05, "loss": 2.6047, "step": 56046 }, { "epoch": 4.39, "learning_rate": 5.605190935466165e-05, "loss": 2.7846, "step": 56047 }, { "epoch": 4.39, "learning_rate": 5.605112522543715e-05, "loss": 2.7205, "step": 56048 }, { "epoch": 4.39, "learning_rate": 5.6050341096212656e-05, "loss": 2.7278, "step": 56049 }, { "epoch": 4.4, "learning_rate": 5.604955696698816e-05, "loss": 2.6933, "step": 56050 }, { "epoch": 4.4, "learning_rate": 5.6048772837763664e-05, "loss": 2.5372, "step": 56051 }, { "epoch": 4.4, "learning_rate": 5.604798870853917e-05, "loss": 2.6024, "step": 56052 }, { "epoch": 4.4, "learning_rate": 5.604720457931467e-05, "loss": 2.7023, "step": 56053 }, { "epoch": 4.4, "learning_rate": 5.6046420450090175e-05, "loss": 2.7501, "step": 56054 }, { "epoch": 4.4, "learning_rate": 5.604563632086568e-05, "loss": 2.6581, "step": 56055 }, { "epoch": 4.4, "learning_rate": 5.604485219164118e-05, "loss": 2.6845, "step": 56056 }, { "epoch": 4.4, "learning_rate": 5.6044068062416686e-05, "loss": 2.6858, "step": 56057 }, { "epoch": 4.4, "learning_rate": 5.604328393319219e-05, "loss": 2.7625, "step": 56058 }, { "epoch": 4.4, "learning_rate": 5.604249980396769e-05, "loss": 2.6672, "step": 56059 }, { "epoch": 4.4, "learning_rate": 5.6041715674743204e-05, "loss": 2.6162, "step": 56060 }, { "epoch": 4.4, "learning_rate": 5.604093154551871e-05, "loss": 2.585, "step": 56061 }, { "epoch": 4.4, "learning_rate": 5.604014741629421e-05, "loss": 2.576, "step": 56062 }, { "epoch": 4.4, "learning_rate": 5.6039363287069715e-05, "loss": 2.6217, "step": 56063 }, { "epoch": 4.4, "learning_rate": 5.603857915784522e-05, "loss": 2.5887, "step": 56064 }, { "epoch": 4.4, "learning_rate": 5.603779502862072e-05, "loss": 2.6575, "step": 56065 }, { "epoch": 4.4, "learning_rate": 5.6037010899396226e-05, "loss": 2.571, "step": 56066 }, { "epoch": 4.4, "learning_rate": 5.603622677017173e-05, "loss": 2.7233, "step": 56067 }, { "epoch": 4.4, "learning_rate": 5.603544264094723e-05, "loss": 2.6242, "step": 56068 }, { "epoch": 4.4, "learning_rate": 5.6034658511722737e-05, "loss": 2.7861, "step": 56069 }, { "epoch": 4.4, "learning_rate": 5.603387438249824e-05, "loss": 2.6876, "step": 56070 }, { "epoch": 4.4, "learning_rate": 5.6033090253273744e-05, "loss": 2.6955, "step": 56071 }, { "epoch": 4.4, "learning_rate": 5.603230612404925e-05, "loss": 2.6659, "step": 56072 }, { "epoch": 4.4, "learning_rate": 5.603152199482475e-05, "loss": 2.8298, "step": 56073 }, { "epoch": 4.4, "learning_rate": 5.6030737865600255e-05, "loss": 2.6569, "step": 56074 }, { "epoch": 4.4, "learning_rate": 5.6029953736375765e-05, "loss": 2.6276, "step": 56075 }, { "epoch": 4.4, "learning_rate": 5.602916960715127e-05, "loss": 2.8358, "step": 56076 }, { "epoch": 4.4, "learning_rate": 5.602838547792677e-05, "loss": 2.5547, "step": 56077 }, { "epoch": 4.4, "learning_rate": 5.602760134870226e-05, "loss": 2.7045, "step": 56078 }, { "epoch": 4.4, "learning_rate": 5.6026817219477767e-05, "loss": 2.6158, "step": 56079 }, { "epoch": 4.4, "learning_rate": 5.602603309025327e-05, "loss": 2.5317, "step": 56080 }, { "epoch": 4.4, "learning_rate": 5.6025248961028774e-05, "loss": 2.5488, "step": 56081 }, { "epoch": 4.4, "learning_rate": 5.602446483180428e-05, "loss": 2.6576, "step": 56082 }, { "epoch": 4.4, "learning_rate": 5.602368070257978e-05, "loss": 2.7586, "step": 56083 }, { "epoch": 4.4, "learning_rate": 5.6022896573355285e-05, "loss": 2.5584, "step": 56084 }, { "epoch": 4.4, "learning_rate": 5.6022112444130795e-05, "loss": 2.641, "step": 56085 }, { "epoch": 4.4, "learning_rate": 5.60213283149063e-05, "loss": 2.5592, "step": 56086 }, { "epoch": 4.4, "learning_rate": 5.60205441856818e-05, "loss": 2.7395, "step": 56087 }, { "epoch": 4.4, "learning_rate": 5.6019760056457306e-05, "loss": 2.6601, "step": 56088 }, { "epoch": 4.4, "learning_rate": 5.601897592723281e-05, "loss": 2.8257, "step": 56089 }, { "epoch": 4.4, "learning_rate": 5.6018191798008314e-05, "loss": 2.747, "step": 56090 }, { "epoch": 4.4, "learning_rate": 5.601740766878382e-05, "loss": 2.7585, "step": 56091 }, { "epoch": 4.4, "learning_rate": 5.601662353955932e-05, "loss": 2.6405, "step": 56092 }, { "epoch": 4.4, "learning_rate": 5.6015839410334825e-05, "loss": 2.9998, "step": 56093 }, { "epoch": 4.4, "learning_rate": 5.601505528111033e-05, "loss": 2.7074, "step": 56094 }, { "epoch": 4.4, "learning_rate": 5.601427115188583e-05, "loss": 2.5621, "step": 56095 }, { "epoch": 4.4, "learning_rate": 5.6013487022661336e-05, "loss": 2.6332, "step": 56096 }, { "epoch": 4.4, "learning_rate": 5.601270289343684e-05, "loss": 2.8021, "step": 56097 }, { "epoch": 4.4, "learning_rate": 5.601191876421234e-05, "loss": 2.594, "step": 56098 }, { "epoch": 4.4, "learning_rate": 5.601113463498785e-05, "loss": 2.6099, "step": 56099 }, { "epoch": 4.4, "learning_rate": 5.601035050576335e-05, "loss": 2.6927, "step": 56100 }, { "epoch": 4.4, "learning_rate": 5.600956637653886e-05, "loss": 2.6876, "step": 56101 }, { "epoch": 4.4, "learning_rate": 5.6008782247314365e-05, "loss": 2.6173, "step": 56102 }, { "epoch": 4.4, "learning_rate": 5.600799811808987e-05, "loss": 2.5577, "step": 56103 }, { "epoch": 4.4, "learning_rate": 5.600721398886537e-05, "loss": 2.8293, "step": 56104 }, { "epoch": 4.4, "learning_rate": 5.6006429859640876e-05, "loss": 2.523, "step": 56105 }, { "epoch": 4.4, "learning_rate": 5.600564573041638e-05, "loss": 2.7262, "step": 56106 }, { "epoch": 4.4, "learning_rate": 5.600486160119188e-05, "loss": 2.5253, "step": 56107 }, { "epoch": 4.4, "learning_rate": 5.600407747196739e-05, "loss": 2.634, "step": 56108 }, { "epoch": 4.4, "learning_rate": 5.600329334274289e-05, "loss": 2.6706, "step": 56109 }, { "epoch": 4.4, "learning_rate": 5.6002509213518394e-05, "loss": 2.6331, "step": 56110 }, { "epoch": 4.4, "learning_rate": 5.60017250842939e-05, "loss": 2.7334, "step": 56111 }, { "epoch": 4.4, "learning_rate": 5.60009409550694e-05, "loss": 2.8741, "step": 56112 }, { "epoch": 4.4, "learning_rate": 5.6000156825844905e-05, "loss": 2.655, "step": 56113 }, { "epoch": 4.4, "learning_rate": 5.59993726966204e-05, "loss": 2.6735, "step": 56114 }, { "epoch": 4.4, "learning_rate": 5.5998588567395906e-05, "loss": 2.8189, "step": 56115 }, { "epoch": 4.4, "learning_rate": 5.599780443817141e-05, "loss": 2.6514, "step": 56116 }, { "epoch": 4.4, "learning_rate": 5.599702030894691e-05, "loss": 2.8629, "step": 56117 }, { "epoch": 4.4, "learning_rate": 5.599623617972242e-05, "loss": 2.5889, "step": 56118 }, { "epoch": 4.4, "learning_rate": 5.599545205049792e-05, "loss": 2.7304, "step": 56119 }, { "epoch": 4.4, "learning_rate": 5.5994667921273424e-05, "loss": 2.6132, "step": 56120 }, { "epoch": 4.4, "learning_rate": 5.599388379204893e-05, "loss": 2.6098, "step": 56121 }, { "epoch": 4.4, "learning_rate": 5.599309966282443e-05, "loss": 2.6454, "step": 56122 }, { "epoch": 4.4, "learning_rate": 5.5992315533599935e-05, "loss": 2.6444, "step": 56123 }, { "epoch": 4.4, "learning_rate": 5.599153140437544e-05, "loss": 2.583, "step": 56124 }, { "epoch": 4.4, "learning_rate": 5.599074727515094e-05, "loss": 2.5883, "step": 56125 }, { "epoch": 4.4, "learning_rate": 5.5989963145926446e-05, "loss": 2.5556, "step": 56126 }, { "epoch": 4.4, "learning_rate": 5.598917901670196e-05, "loss": 2.7241, "step": 56127 }, { "epoch": 4.4, "learning_rate": 5.598839488747746e-05, "loss": 2.5943, "step": 56128 }, { "epoch": 4.4, "learning_rate": 5.5987610758252964e-05, "loss": 2.7948, "step": 56129 }, { "epoch": 4.4, "learning_rate": 5.598682662902847e-05, "loss": 2.5403, "step": 56130 }, { "epoch": 4.4, "learning_rate": 5.598604249980397e-05, "loss": 2.7366, "step": 56131 }, { "epoch": 4.4, "learning_rate": 5.5985258370579475e-05, "loss": 2.6952, "step": 56132 }, { "epoch": 4.4, "learning_rate": 5.598447424135498e-05, "loss": 2.5935, "step": 56133 }, { "epoch": 4.4, "learning_rate": 5.598369011213048e-05, "loss": 2.5883, "step": 56134 }, { "epoch": 4.4, "learning_rate": 5.5982905982905986e-05, "loss": 2.6859, "step": 56135 }, { "epoch": 4.4, "learning_rate": 5.598212185368149e-05, "loss": 2.7292, "step": 56136 }, { "epoch": 4.4, "learning_rate": 5.5981337724456993e-05, "loss": 2.5737, "step": 56137 }, { "epoch": 4.4, "learning_rate": 5.59805535952325e-05, "loss": 2.6561, "step": 56138 }, { "epoch": 4.4, "learning_rate": 5.5979769466008e-05, "loss": 2.7817, "step": 56139 }, { "epoch": 4.4, "learning_rate": 5.5978985336783504e-05, "loss": 2.5223, "step": 56140 }, { "epoch": 4.4, "learning_rate": 5.597820120755901e-05, "loss": 2.7551, "step": 56141 }, { "epoch": 4.4, "learning_rate": 5.597741707833452e-05, "loss": 2.6556, "step": 56142 }, { "epoch": 4.4, "learning_rate": 5.597663294911002e-05, "loss": 2.6464, "step": 56143 }, { "epoch": 4.4, "learning_rate": 5.5975848819885526e-05, "loss": 2.6291, "step": 56144 }, { "epoch": 4.4, "learning_rate": 5.597506469066103e-05, "loss": 2.5283, "step": 56145 }, { "epoch": 4.4, "learning_rate": 5.597428056143653e-05, "loss": 2.5593, "step": 56146 }, { "epoch": 4.4, "learning_rate": 5.597349643221204e-05, "loss": 2.6357, "step": 56147 }, { "epoch": 4.4, "learning_rate": 5.597271230298754e-05, "loss": 2.9267, "step": 56148 }, { "epoch": 4.4, "learning_rate": 5.5971928173763044e-05, "loss": 2.6607, "step": 56149 }, { "epoch": 4.4, "learning_rate": 5.5971144044538534e-05, "loss": 2.706, "step": 56150 }, { "epoch": 4.4, "learning_rate": 5.597035991531404e-05, "loss": 2.6293, "step": 56151 }, { "epoch": 4.4, "learning_rate": 5.596957578608955e-05, "loss": 2.6886, "step": 56152 }, { "epoch": 4.4, "learning_rate": 5.596879165686505e-05, "loss": 2.6386, "step": 56153 }, { "epoch": 4.4, "learning_rate": 5.5968007527640556e-05, "loss": 2.9098, "step": 56154 }, { "epoch": 4.4, "learning_rate": 5.596722339841606e-05, "loss": 2.7092, "step": 56155 }, { "epoch": 4.4, "learning_rate": 5.596643926919156e-05, "loss": 2.7407, "step": 56156 }, { "epoch": 4.4, "learning_rate": 5.596565513996707e-05, "loss": 2.827, "step": 56157 }, { "epoch": 4.4, "learning_rate": 5.596487101074257e-05, "loss": 2.5995, "step": 56158 }, { "epoch": 4.4, "learning_rate": 5.5964086881518074e-05, "loss": 2.8346, "step": 56159 }, { "epoch": 4.4, "learning_rate": 5.596330275229358e-05, "loss": 2.5364, "step": 56160 }, { "epoch": 4.4, "learning_rate": 5.596251862306908e-05, "loss": 2.7149, "step": 56161 }, { "epoch": 4.4, "learning_rate": 5.5961734493844585e-05, "loss": 2.7107, "step": 56162 }, { "epoch": 4.4, "learning_rate": 5.596095036462009e-05, "loss": 2.6106, "step": 56163 }, { "epoch": 4.4, "learning_rate": 5.596016623539559e-05, "loss": 2.7467, "step": 56164 }, { "epoch": 4.4, "learning_rate": 5.5959382106171096e-05, "loss": 2.6215, "step": 56165 }, { "epoch": 4.4, "learning_rate": 5.59585979769466e-05, "loss": 2.8152, "step": 56166 }, { "epoch": 4.4, "learning_rate": 5.5957813847722104e-05, "loss": 2.6985, "step": 56167 }, { "epoch": 4.4, "learning_rate": 5.5957029718497614e-05, "loss": 2.7591, "step": 56168 }, { "epoch": 4.4, "learning_rate": 5.595624558927312e-05, "loss": 2.6373, "step": 56169 }, { "epoch": 4.4, "learning_rate": 5.595546146004862e-05, "loss": 2.6385, "step": 56170 }, { "epoch": 4.4, "learning_rate": 5.5954677330824125e-05, "loss": 2.7522, "step": 56171 }, { "epoch": 4.4, "learning_rate": 5.595389320159963e-05, "loss": 2.64, "step": 56172 }, { "epoch": 4.4, "learning_rate": 5.595310907237513e-05, "loss": 2.6522, "step": 56173 }, { "epoch": 4.4, "learning_rate": 5.5952324943150636e-05, "loss": 2.675, "step": 56174 }, { "epoch": 4.4, "learning_rate": 5.595154081392614e-05, "loss": 2.5021, "step": 56175 }, { "epoch": 4.4, "learning_rate": 5.5950756684701644e-05, "loss": 2.7551, "step": 56176 }, { "epoch": 4.41, "learning_rate": 5.594997255547715e-05, "loss": 2.5816, "step": 56177 }, { "epoch": 4.41, "learning_rate": 5.594918842625265e-05, "loss": 2.6435, "step": 56178 }, { "epoch": 4.41, "learning_rate": 5.5948404297028155e-05, "loss": 2.5508, "step": 56179 }, { "epoch": 4.41, "learning_rate": 5.594762016780366e-05, "loss": 2.5639, "step": 56180 }, { "epoch": 4.41, "learning_rate": 5.594683603857916e-05, "loss": 2.6844, "step": 56181 }, { "epoch": 4.41, "learning_rate": 5.5946051909354666e-05, "loss": 2.5806, "step": 56182 }, { "epoch": 4.41, "learning_rate": 5.594526778013017e-05, "loss": 2.8365, "step": 56183 }, { "epoch": 4.41, "learning_rate": 5.594448365090568e-05, "loss": 2.7525, "step": 56184 }, { "epoch": 4.41, "learning_rate": 5.5943699521681184e-05, "loss": 2.6909, "step": 56185 }, { "epoch": 4.41, "learning_rate": 5.5942915392456674e-05, "loss": 2.5949, "step": 56186 }, { "epoch": 4.41, "learning_rate": 5.594213126323218e-05, "loss": 2.5174, "step": 56187 }, { "epoch": 4.41, "learning_rate": 5.594134713400768e-05, "loss": 2.8774, "step": 56188 }, { "epoch": 4.41, "learning_rate": 5.5940563004783185e-05, "loss": 2.6695, "step": 56189 }, { "epoch": 4.41, "learning_rate": 5.593977887555869e-05, "loss": 2.6231, "step": 56190 }, { "epoch": 4.41, "learning_rate": 5.593899474633419e-05, "loss": 2.6579, "step": 56191 }, { "epoch": 4.41, "learning_rate": 5.5938210617109696e-05, "loss": 2.6216, "step": 56192 }, { "epoch": 4.41, "learning_rate": 5.59374264878852e-05, "loss": 2.4362, "step": 56193 }, { "epoch": 4.41, "learning_rate": 5.593664235866071e-05, "loss": 2.6815, "step": 56194 }, { "epoch": 4.41, "learning_rate": 5.5935858229436214e-05, "loss": 2.516, "step": 56195 }, { "epoch": 4.41, "learning_rate": 5.593507410021172e-05, "loss": 2.5909, "step": 56196 }, { "epoch": 4.41, "learning_rate": 5.593428997098722e-05, "loss": 2.6177, "step": 56197 }, { "epoch": 4.41, "learning_rate": 5.5933505841762725e-05, "loss": 2.6386, "step": 56198 }, { "epoch": 4.41, "learning_rate": 5.593272171253823e-05, "loss": 2.6254, "step": 56199 }, { "epoch": 4.41, "learning_rate": 5.593193758331373e-05, "loss": 2.729, "step": 56200 }, { "epoch": 4.41, "learning_rate": 5.5931153454089236e-05, "loss": 2.7793, "step": 56201 }, { "epoch": 4.41, "learning_rate": 5.593036932486474e-05, "loss": 2.8132, "step": 56202 }, { "epoch": 4.41, "learning_rate": 5.592958519564024e-05, "loss": 2.6523, "step": 56203 }, { "epoch": 4.41, "learning_rate": 5.592880106641575e-05, "loss": 2.6431, "step": 56204 }, { "epoch": 4.41, "learning_rate": 5.592801693719125e-05, "loss": 2.5157, "step": 56205 }, { "epoch": 4.41, "learning_rate": 5.5927232807966754e-05, "loss": 2.7078, "step": 56206 }, { "epoch": 4.41, "learning_rate": 5.592644867874226e-05, "loss": 2.697, "step": 56207 }, { "epoch": 4.41, "learning_rate": 5.592566454951776e-05, "loss": 2.5244, "step": 56208 }, { "epoch": 4.41, "learning_rate": 5.592488042029327e-05, "loss": 2.8096, "step": 56209 }, { "epoch": 4.41, "learning_rate": 5.5924096291068775e-05, "loss": 2.7185, "step": 56210 }, { "epoch": 4.41, "learning_rate": 5.592331216184428e-05, "loss": 2.5133, "step": 56211 }, { "epoch": 4.41, "learning_rate": 5.592252803261978e-05, "loss": 2.7339, "step": 56212 }, { "epoch": 4.41, "learning_rate": 5.5921743903395287e-05, "loss": 2.5799, "step": 56213 }, { "epoch": 4.41, "learning_rate": 5.592095977417079e-05, "loss": 2.5838, "step": 56214 }, { "epoch": 4.41, "learning_rate": 5.5920175644946294e-05, "loss": 2.5687, "step": 56215 }, { "epoch": 4.41, "learning_rate": 5.59193915157218e-05, "loss": 2.753, "step": 56216 }, { "epoch": 4.41, "learning_rate": 5.59186073864973e-05, "loss": 2.6485, "step": 56217 }, { "epoch": 4.41, "learning_rate": 5.5917823257272805e-05, "loss": 2.7281, "step": 56218 }, { "epoch": 4.41, "learning_rate": 5.591703912804831e-05, "loss": 2.6313, "step": 56219 }, { "epoch": 4.41, "learning_rate": 5.591625499882381e-05, "loss": 2.9499, "step": 56220 }, { "epoch": 4.41, "learning_rate": 5.5915470869599316e-05, "loss": 2.7126, "step": 56221 }, { "epoch": 4.41, "learning_rate": 5.591468674037481e-05, "loss": 2.5881, "step": 56222 }, { "epoch": 4.41, "learning_rate": 5.5913902611150317e-05, "loss": 2.6806, "step": 56223 }, { "epoch": 4.41, "learning_rate": 5.591311848192582e-05, "loss": 2.7478, "step": 56224 }, { "epoch": 4.41, "learning_rate": 5.5912334352701324e-05, "loss": 2.5763, "step": 56225 }, { "epoch": 4.41, "learning_rate": 5.591155022347683e-05, "loss": 2.8233, "step": 56226 }, { "epoch": 4.41, "learning_rate": 5.591076609425233e-05, "loss": 2.6183, "step": 56227 }, { "epoch": 4.41, "learning_rate": 5.5909981965027835e-05, "loss": 2.548, "step": 56228 }, { "epoch": 4.41, "learning_rate": 5.590919783580334e-05, "loss": 2.9523, "step": 56229 }, { "epoch": 4.41, "learning_rate": 5.590841370657884e-05, "loss": 2.6483, "step": 56230 }, { "epoch": 4.41, "learning_rate": 5.5907629577354346e-05, "loss": 2.7044, "step": 56231 }, { "epoch": 4.41, "learning_rate": 5.590684544812985e-05, "loss": 2.8046, "step": 56232 }, { "epoch": 4.41, "learning_rate": 5.590606131890535e-05, "loss": 2.7806, "step": 56233 }, { "epoch": 4.41, "learning_rate": 5.590527718968086e-05, "loss": 2.4265, "step": 56234 }, { "epoch": 4.41, "learning_rate": 5.590449306045637e-05, "loss": 2.6778, "step": 56235 }, { "epoch": 4.41, "learning_rate": 5.590370893123187e-05, "loss": 2.7976, "step": 56236 }, { "epoch": 4.41, "learning_rate": 5.5902924802007375e-05, "loss": 2.6899, "step": 56237 }, { "epoch": 4.41, "learning_rate": 5.590214067278288e-05, "loss": 2.5417, "step": 56238 }, { "epoch": 4.41, "learning_rate": 5.590135654355838e-05, "loss": 2.7182, "step": 56239 }, { "epoch": 4.41, "learning_rate": 5.5900572414333886e-05, "loss": 2.8075, "step": 56240 }, { "epoch": 4.41, "learning_rate": 5.589978828510939e-05, "loss": 2.69, "step": 56241 }, { "epoch": 4.41, "learning_rate": 5.589900415588489e-05, "loss": 2.6069, "step": 56242 }, { "epoch": 4.41, "learning_rate": 5.58982200266604e-05, "loss": 2.6882, "step": 56243 }, { "epoch": 4.41, "learning_rate": 5.58974358974359e-05, "loss": 2.6661, "step": 56244 }, { "epoch": 4.41, "learning_rate": 5.5896651768211404e-05, "loss": 2.6007, "step": 56245 }, { "epoch": 4.41, "learning_rate": 5.589586763898691e-05, "loss": 2.8934, "step": 56246 }, { "epoch": 4.41, "learning_rate": 5.589508350976241e-05, "loss": 2.7494, "step": 56247 }, { "epoch": 4.41, "learning_rate": 5.5894299380537915e-05, "loss": 2.4734, "step": 56248 }, { "epoch": 4.41, "learning_rate": 5.589351525131342e-05, "loss": 2.5146, "step": 56249 }, { "epoch": 4.41, "learning_rate": 5.589273112208893e-05, "loss": 2.5465, "step": 56250 }, { "epoch": 4.41, "learning_rate": 5.589194699286443e-05, "loss": 2.5309, "step": 56251 }, { "epoch": 4.41, "learning_rate": 5.589116286363994e-05, "loss": 2.9429, "step": 56252 }, { "epoch": 4.41, "learning_rate": 5.589037873441544e-05, "loss": 2.6202, "step": 56253 }, { "epoch": 4.41, "learning_rate": 5.5889594605190944e-05, "loss": 2.7258, "step": 56254 }, { "epoch": 4.41, "learning_rate": 5.588881047596645e-05, "loss": 2.6118, "step": 56255 }, { "epoch": 4.41, "learning_rate": 5.588802634674195e-05, "loss": 2.6963, "step": 56256 }, { "epoch": 4.41, "learning_rate": 5.5887242217517455e-05, "loss": 2.7624, "step": 56257 }, { "epoch": 4.41, "learning_rate": 5.5886458088292945e-05, "loss": 2.7786, "step": 56258 }, { "epoch": 4.41, "learning_rate": 5.588567395906845e-05, "loss": 2.4813, "step": 56259 }, { "epoch": 4.41, "learning_rate": 5.588488982984396e-05, "loss": 2.7701, "step": 56260 }, { "epoch": 4.41, "learning_rate": 5.588410570061946e-05, "loss": 2.7597, "step": 56261 }, { "epoch": 4.41, "learning_rate": 5.588332157139497e-05, "loss": 2.6023, "step": 56262 }, { "epoch": 4.41, "learning_rate": 5.588253744217047e-05, "loss": 2.9407, "step": 56263 }, { "epoch": 4.41, "learning_rate": 5.5881753312945974e-05, "loss": 2.6368, "step": 56264 }, { "epoch": 4.41, "learning_rate": 5.588096918372148e-05, "loss": 2.7666, "step": 56265 }, { "epoch": 4.41, "learning_rate": 5.588018505449698e-05, "loss": 2.5586, "step": 56266 }, { "epoch": 4.41, "learning_rate": 5.5879400925272485e-05, "loss": 2.6784, "step": 56267 }, { "epoch": 4.41, "learning_rate": 5.587861679604799e-05, "loss": 2.5005, "step": 56268 }, { "epoch": 4.41, "learning_rate": 5.587783266682349e-05, "loss": 2.6027, "step": 56269 }, { "epoch": 4.41, "learning_rate": 5.5877048537598996e-05, "loss": 2.6999, "step": 56270 }, { "epoch": 4.41, "learning_rate": 5.58762644083745e-05, "loss": 2.9542, "step": 56271 }, { "epoch": 4.41, "learning_rate": 5.5875480279150004e-05, "loss": 2.6877, "step": 56272 }, { "epoch": 4.41, "learning_rate": 5.587469614992551e-05, "loss": 2.751, "step": 56273 }, { "epoch": 4.41, "learning_rate": 5.587391202070101e-05, "loss": 2.6651, "step": 56274 }, { "epoch": 4.41, "learning_rate": 5.5873127891476515e-05, "loss": 2.4841, "step": 56275 }, { "epoch": 4.41, "learning_rate": 5.5872343762252025e-05, "loss": 2.7868, "step": 56276 }, { "epoch": 4.41, "learning_rate": 5.587155963302753e-05, "loss": 2.5771, "step": 56277 }, { "epoch": 4.41, "learning_rate": 5.587077550380303e-05, "loss": 2.7675, "step": 56278 }, { "epoch": 4.41, "learning_rate": 5.5869991374578536e-05, "loss": 2.6553, "step": 56279 }, { "epoch": 4.41, "learning_rate": 5.586920724535404e-05, "loss": 2.6738, "step": 56280 }, { "epoch": 4.41, "learning_rate": 5.586842311612954e-05, "loss": 2.8203, "step": 56281 }, { "epoch": 4.41, "learning_rate": 5.586763898690505e-05, "loss": 2.5449, "step": 56282 }, { "epoch": 4.41, "learning_rate": 5.586685485768055e-05, "loss": 2.9307, "step": 56283 }, { "epoch": 4.41, "learning_rate": 5.5866070728456054e-05, "loss": 2.5973, "step": 56284 }, { "epoch": 4.41, "learning_rate": 5.586528659923156e-05, "loss": 2.6933, "step": 56285 }, { "epoch": 4.41, "learning_rate": 5.586450247000706e-05, "loss": 2.602, "step": 56286 }, { "epoch": 4.41, "learning_rate": 5.5863718340782565e-05, "loss": 2.5962, "step": 56287 }, { "epoch": 4.41, "learning_rate": 5.586293421155807e-05, "loss": 2.6658, "step": 56288 }, { "epoch": 4.41, "learning_rate": 5.586215008233357e-05, "loss": 2.6702, "step": 56289 }, { "epoch": 4.41, "learning_rate": 5.5861365953109076e-05, "loss": 2.6026, "step": 56290 }, { "epoch": 4.41, "learning_rate": 5.586058182388458e-05, "loss": 2.8533, "step": 56291 }, { "epoch": 4.41, "learning_rate": 5.585979769466009e-05, "loss": 2.9008, "step": 56292 }, { "epoch": 4.41, "learning_rate": 5.5859013565435594e-05, "loss": 2.8258, "step": 56293 }, { "epoch": 4.41, "learning_rate": 5.5858229436211084e-05, "loss": 2.7099, "step": 56294 }, { "epoch": 4.41, "learning_rate": 5.585744530698659e-05, "loss": 2.5553, "step": 56295 }, { "epoch": 4.41, "learning_rate": 5.585666117776209e-05, "loss": 2.6777, "step": 56296 }, { "epoch": 4.41, "learning_rate": 5.5855877048537595e-05, "loss": 2.6982, "step": 56297 }, { "epoch": 4.41, "learning_rate": 5.58550929193131e-05, "loss": 2.9333, "step": 56298 }, { "epoch": 4.41, "learning_rate": 5.58543087900886e-05, "loss": 2.5131, "step": 56299 }, { "epoch": 4.41, "learning_rate": 5.5853524660864106e-05, "loss": 2.5496, "step": 56300 }, { "epoch": 4.41, "learning_rate": 5.585274053163961e-05, "loss": 2.6574, "step": 56301 }, { "epoch": 4.41, "learning_rate": 5.585195640241512e-05, "loss": 2.5783, "step": 56302 }, { "epoch": 4.41, "learning_rate": 5.5851172273190624e-05, "loss": 2.7083, "step": 56303 }, { "epoch": 4.41, "learning_rate": 5.585038814396613e-05, "loss": 2.5901, "step": 56304 }, { "epoch": 4.42, "learning_rate": 5.584960401474163e-05, "loss": 2.5953, "step": 56305 }, { "epoch": 4.42, "learning_rate": 5.5848819885517135e-05, "loss": 2.7277, "step": 56306 }, { "epoch": 4.42, "learning_rate": 5.584803575629264e-05, "loss": 2.5358, "step": 56307 }, { "epoch": 4.42, "learning_rate": 5.584725162706814e-05, "loss": 2.5882, "step": 56308 }, { "epoch": 4.42, "learning_rate": 5.5846467497843646e-05, "loss": 2.8411, "step": 56309 }, { "epoch": 4.42, "learning_rate": 5.584568336861915e-05, "loss": 2.6541, "step": 56310 }, { "epoch": 4.42, "learning_rate": 5.5844899239394654e-05, "loss": 2.6474, "step": 56311 }, { "epoch": 4.42, "learning_rate": 5.584411511017016e-05, "loss": 2.3667, "step": 56312 }, { "epoch": 4.42, "learning_rate": 5.584333098094566e-05, "loss": 2.5826, "step": 56313 }, { "epoch": 4.42, "learning_rate": 5.5842546851721165e-05, "loss": 2.7112, "step": 56314 }, { "epoch": 4.42, "learning_rate": 5.584176272249667e-05, "loss": 2.6941, "step": 56315 }, { "epoch": 4.42, "learning_rate": 5.584097859327217e-05, "loss": 2.6389, "step": 56316 }, { "epoch": 4.42, "learning_rate": 5.584019446404768e-05, "loss": 2.6246, "step": 56317 }, { "epoch": 4.42, "learning_rate": 5.5839410334823186e-05, "loss": 2.5524, "step": 56318 }, { "epoch": 4.42, "learning_rate": 5.583862620559869e-05, "loss": 2.6735, "step": 56319 }, { "epoch": 4.42, "learning_rate": 5.5837842076374194e-05, "loss": 2.7336, "step": 56320 }, { "epoch": 4.42, "learning_rate": 5.58370579471497e-05, "loss": 2.6765, "step": 56321 }, { "epoch": 4.42, "learning_rate": 5.58362738179252e-05, "loss": 2.3393, "step": 56322 }, { "epoch": 4.42, "learning_rate": 5.5835489688700705e-05, "loss": 2.5812, "step": 56323 }, { "epoch": 4.42, "learning_rate": 5.583470555947621e-05, "loss": 2.6177, "step": 56324 }, { "epoch": 4.42, "learning_rate": 5.583392143025171e-05, "loss": 2.6179, "step": 56325 }, { "epoch": 4.42, "learning_rate": 5.5833137301027216e-05, "loss": 2.6276, "step": 56326 }, { "epoch": 4.42, "learning_rate": 5.583235317180272e-05, "loss": 2.5644, "step": 56327 }, { "epoch": 4.42, "learning_rate": 5.583156904257822e-05, "loss": 2.7201, "step": 56328 }, { "epoch": 4.42, "learning_rate": 5.583078491335373e-05, "loss": 2.6251, "step": 56329 }, { "epoch": 4.42, "learning_rate": 5.5830000784129224e-05, "loss": 2.5973, "step": 56330 }, { "epoch": 4.42, "learning_rate": 5.582921665490473e-05, "loss": 2.6817, "step": 56331 }, { "epoch": 4.42, "learning_rate": 5.582843252568023e-05, "loss": 2.7543, "step": 56332 }, { "epoch": 4.42, "learning_rate": 5.5827648396455735e-05, "loss": 2.8597, "step": 56333 }, { "epoch": 4.42, "learning_rate": 5.582686426723124e-05, "loss": 2.8249, "step": 56334 }, { "epoch": 4.42, "learning_rate": 5.582608013800674e-05, "loss": 2.7652, "step": 56335 }, { "epoch": 4.42, "learning_rate": 5.5825296008782246e-05, "loss": 2.7329, "step": 56336 }, { "epoch": 4.42, "learning_rate": 5.582451187955775e-05, "loss": 2.7819, "step": 56337 }, { "epoch": 4.42, "learning_rate": 5.582372775033325e-05, "loss": 2.9559, "step": 56338 }, { "epoch": 4.42, "learning_rate": 5.582294362110876e-05, "loss": 2.6433, "step": 56339 }, { "epoch": 4.42, "learning_rate": 5.582215949188426e-05, "loss": 2.6454, "step": 56340 }, { "epoch": 4.42, "learning_rate": 5.5821375362659764e-05, "loss": 2.85, "step": 56341 }, { "epoch": 4.42, "learning_rate": 5.582059123343527e-05, "loss": 2.7832, "step": 56342 }, { "epoch": 4.42, "learning_rate": 5.581980710421078e-05, "loss": 2.7418, "step": 56343 }, { "epoch": 4.42, "learning_rate": 5.581902297498628e-05, "loss": 2.7177, "step": 56344 }, { "epoch": 4.42, "learning_rate": 5.5818238845761786e-05, "loss": 2.5321, "step": 56345 }, { "epoch": 4.42, "learning_rate": 5.581745471653729e-05, "loss": 2.6363, "step": 56346 }, { "epoch": 4.42, "learning_rate": 5.581667058731279e-05, "loss": 2.8381, "step": 56347 }, { "epoch": 4.42, "learning_rate": 5.5815886458088297e-05, "loss": 2.8388, "step": 56348 }, { "epoch": 4.42, "learning_rate": 5.58151023288638e-05, "loss": 2.487, "step": 56349 }, { "epoch": 4.42, "learning_rate": 5.5814318199639304e-05, "loss": 2.8596, "step": 56350 }, { "epoch": 4.42, "learning_rate": 5.581353407041481e-05, "loss": 2.564, "step": 56351 }, { "epoch": 4.42, "learning_rate": 5.581274994119031e-05, "loss": 2.7342, "step": 56352 }, { "epoch": 4.42, "learning_rate": 5.5811965811965815e-05, "loss": 2.6075, "step": 56353 }, { "epoch": 4.42, "learning_rate": 5.581118168274132e-05, "loss": 2.5956, "step": 56354 }, { "epoch": 4.42, "learning_rate": 5.581039755351682e-05, "loss": 2.4274, "step": 56355 }, { "epoch": 4.42, "learning_rate": 5.5809613424292326e-05, "loss": 2.5976, "step": 56356 }, { "epoch": 4.42, "learning_rate": 5.580882929506783e-05, "loss": 2.9358, "step": 56357 }, { "epoch": 4.42, "learning_rate": 5.580804516584334e-05, "loss": 2.8751, "step": 56358 }, { "epoch": 4.42, "learning_rate": 5.5807261036618844e-05, "loss": 2.7964, "step": 56359 }, { "epoch": 4.42, "learning_rate": 5.580647690739435e-05, "loss": 2.6443, "step": 56360 }, { "epoch": 4.42, "learning_rate": 5.580569277816985e-05, "loss": 2.6488, "step": 56361 }, { "epoch": 4.42, "learning_rate": 5.5804908648945355e-05, "loss": 2.6371, "step": 56362 }, { "epoch": 4.42, "learning_rate": 5.580412451972086e-05, "loss": 2.5106, "step": 56363 }, { "epoch": 4.42, "learning_rate": 5.580334039049636e-05, "loss": 2.7284, "step": 56364 }, { "epoch": 4.42, "learning_rate": 5.5802556261271866e-05, "loss": 2.4504, "step": 56365 }, { "epoch": 4.42, "learning_rate": 5.5801772132047356e-05, "loss": 2.5761, "step": 56366 }, { "epoch": 4.42, "learning_rate": 5.580098800282286e-05, "loss": 2.7847, "step": 56367 }, { "epoch": 4.42, "learning_rate": 5.580020387359837e-05, "loss": 2.6927, "step": 56368 }, { "epoch": 4.42, "learning_rate": 5.5799419744373874e-05, "loss": 2.5818, "step": 56369 }, { "epoch": 4.42, "learning_rate": 5.579863561514938e-05, "loss": 2.7247, "step": 56370 }, { "epoch": 4.42, "learning_rate": 5.579785148592488e-05, "loss": 2.527, "step": 56371 }, { "epoch": 4.42, "learning_rate": 5.5797067356700385e-05, "loss": 2.5833, "step": 56372 }, { "epoch": 4.42, "learning_rate": 5.579628322747589e-05, "loss": 2.6696, "step": 56373 }, { "epoch": 4.42, "learning_rate": 5.579549909825139e-05, "loss": 2.7103, "step": 56374 }, { "epoch": 4.42, "learning_rate": 5.5794714969026896e-05, "loss": 2.6082, "step": 56375 }, { "epoch": 4.42, "learning_rate": 5.57939308398024e-05, "loss": 2.7357, "step": 56376 }, { "epoch": 4.42, "learning_rate": 5.57931467105779e-05, "loss": 2.71, "step": 56377 }, { "epoch": 4.42, "learning_rate": 5.579236258135341e-05, "loss": 2.6797, "step": 56378 }, { "epoch": 4.42, "learning_rate": 5.579157845212891e-05, "loss": 2.421, "step": 56379 }, { "epoch": 4.42, "learning_rate": 5.5790794322904414e-05, "loss": 2.6629, "step": 56380 }, { "epoch": 4.42, "learning_rate": 5.579001019367992e-05, "loss": 2.7385, "step": 56381 }, { "epoch": 4.42, "learning_rate": 5.578922606445542e-05, "loss": 2.5662, "step": 56382 }, { "epoch": 4.42, "learning_rate": 5.5788441935230925e-05, "loss": 2.9116, "step": 56383 }, { "epoch": 4.42, "learning_rate": 5.5787657806006436e-05, "loss": 2.6613, "step": 56384 }, { "epoch": 4.42, "learning_rate": 5.578687367678194e-05, "loss": 2.6466, "step": 56385 }, { "epoch": 4.42, "learning_rate": 5.578608954755744e-05, "loss": 2.6463, "step": 56386 }, { "epoch": 4.42, "learning_rate": 5.578530541833295e-05, "loss": 2.5352, "step": 56387 }, { "epoch": 4.42, "learning_rate": 5.578452128910845e-05, "loss": 2.5016, "step": 56388 }, { "epoch": 4.42, "learning_rate": 5.5783737159883954e-05, "loss": 2.7987, "step": 56389 }, { "epoch": 4.42, "learning_rate": 5.578295303065946e-05, "loss": 2.7889, "step": 56390 }, { "epoch": 4.42, "learning_rate": 5.578216890143496e-05, "loss": 2.8229, "step": 56391 }, { "epoch": 4.42, "learning_rate": 5.5781384772210465e-05, "loss": 2.655, "step": 56392 }, { "epoch": 4.42, "learning_rate": 5.578060064298597e-05, "loss": 2.4843, "step": 56393 }, { "epoch": 4.42, "learning_rate": 5.577981651376147e-05, "loss": 2.701, "step": 56394 }, { "epoch": 4.42, "learning_rate": 5.5779032384536976e-05, "loss": 2.6156, "step": 56395 }, { "epoch": 4.42, "learning_rate": 5.577824825531248e-05, "loss": 2.6052, "step": 56396 }, { "epoch": 4.42, "learning_rate": 5.5777464126087984e-05, "loss": 2.857, "step": 56397 }, { "epoch": 4.42, "learning_rate": 5.577667999686349e-05, "loss": 2.5116, "step": 56398 }, { "epoch": 4.42, "learning_rate": 5.577589586763899e-05, "loss": 2.7518, "step": 56399 }, { "epoch": 4.42, "learning_rate": 5.57751117384145e-05, "loss": 2.5793, "step": 56400 }, { "epoch": 4.42, "learning_rate": 5.5774327609190005e-05, "loss": 2.5718, "step": 56401 }, { "epoch": 4.42, "learning_rate": 5.5773543479965495e-05, "loss": 2.7225, "step": 56402 }, { "epoch": 4.42, "learning_rate": 5.5772759350741e-05, "loss": 2.8554, "step": 56403 }, { "epoch": 4.42, "learning_rate": 5.57719752215165e-05, "loss": 2.9761, "step": 56404 }, { "epoch": 4.42, "learning_rate": 5.5771191092292006e-05, "loss": 2.6132, "step": 56405 }, { "epoch": 4.42, "learning_rate": 5.577040696306751e-05, "loss": 2.5121, "step": 56406 }, { "epoch": 4.42, "learning_rate": 5.5769622833843014e-05, "loss": 2.6633, "step": 56407 }, { "epoch": 4.42, "learning_rate": 5.576883870461852e-05, "loss": 2.6863, "step": 56408 }, { "epoch": 4.42, "learning_rate": 5.576805457539402e-05, "loss": 2.8446, "step": 56409 }, { "epoch": 4.42, "learning_rate": 5.576727044616953e-05, "loss": 2.6297, "step": 56410 }, { "epoch": 4.42, "learning_rate": 5.5766486316945035e-05, "loss": 2.6575, "step": 56411 }, { "epoch": 4.42, "learning_rate": 5.576570218772054e-05, "loss": 2.5993, "step": 56412 }, { "epoch": 4.42, "learning_rate": 5.576491805849604e-05, "loss": 2.7582, "step": 56413 }, { "epoch": 4.42, "learning_rate": 5.5764133929271546e-05, "loss": 2.677, "step": 56414 }, { "epoch": 4.42, "learning_rate": 5.576334980004705e-05, "loss": 2.5339, "step": 56415 }, { "epoch": 4.42, "learning_rate": 5.5762565670822553e-05, "loss": 2.6968, "step": 56416 }, { "epoch": 4.42, "learning_rate": 5.576178154159806e-05, "loss": 2.6641, "step": 56417 }, { "epoch": 4.42, "learning_rate": 5.576099741237356e-05, "loss": 2.5332, "step": 56418 }, { "epoch": 4.42, "learning_rate": 5.5760213283149064e-05, "loss": 2.666, "step": 56419 }, { "epoch": 4.42, "learning_rate": 5.575942915392457e-05, "loss": 2.8776, "step": 56420 }, { "epoch": 4.42, "learning_rate": 5.575864502470007e-05, "loss": 2.8088, "step": 56421 }, { "epoch": 4.42, "learning_rate": 5.5757860895475575e-05, "loss": 2.8413, "step": 56422 }, { "epoch": 4.42, "learning_rate": 5.575707676625108e-05, "loss": 2.7011, "step": 56423 }, { "epoch": 4.42, "learning_rate": 5.575629263702658e-05, "loss": 2.6488, "step": 56424 }, { "epoch": 4.42, "learning_rate": 5.575550850780209e-05, "loss": 2.8592, "step": 56425 }, { "epoch": 4.42, "learning_rate": 5.57547243785776e-05, "loss": 2.4707, "step": 56426 }, { "epoch": 4.42, "learning_rate": 5.57539402493531e-05, "loss": 2.6645, "step": 56427 }, { "epoch": 4.42, "learning_rate": 5.5753156120128604e-05, "loss": 2.5235, "step": 56428 }, { "epoch": 4.42, "learning_rate": 5.575237199090411e-05, "loss": 2.5202, "step": 56429 }, { "epoch": 4.42, "learning_rate": 5.575158786167961e-05, "loss": 2.6107, "step": 56430 }, { "epoch": 4.42, "learning_rate": 5.5750803732455115e-05, "loss": 2.8621, "step": 56431 }, { "epoch": 4.42, "learning_rate": 5.575001960323062e-05, "loss": 2.6537, "step": 56432 }, { "epoch": 4.43, "learning_rate": 5.574923547400612e-05, "loss": 2.6717, "step": 56433 }, { "epoch": 4.43, "learning_rate": 5.5748451344781626e-05, "loss": 2.6251, "step": 56434 }, { "epoch": 4.43, "learning_rate": 5.574766721555713e-05, "loss": 2.3899, "step": 56435 }, { "epoch": 4.43, "learning_rate": 5.5746883086332634e-05, "loss": 2.8137, "step": 56436 }, { "epoch": 4.43, "learning_rate": 5.574609895710814e-05, "loss": 2.6244, "step": 56437 }, { "epoch": 4.43, "learning_rate": 5.5745314827883634e-05, "loss": 2.537, "step": 56438 }, { "epoch": 4.43, "learning_rate": 5.574453069865914e-05, "loss": 2.6921, "step": 56439 }, { "epoch": 4.43, "learning_rate": 5.574374656943464e-05, "loss": 2.5465, "step": 56440 }, { "epoch": 4.43, "learning_rate": 5.5742962440210145e-05, "loss": 2.6047, "step": 56441 }, { "epoch": 4.43, "learning_rate": 5.574217831098565e-05, "loss": 2.5788, "step": 56442 }, { "epoch": 4.43, "learning_rate": 5.574139418176115e-05, "loss": 2.7128, "step": 56443 }, { "epoch": 4.43, "learning_rate": 5.5740610052536656e-05, "loss": 2.4784, "step": 56444 }, { "epoch": 4.43, "learning_rate": 5.573982592331216e-05, "loss": 2.6754, "step": 56445 }, { "epoch": 4.43, "learning_rate": 5.5739041794087664e-05, "loss": 2.6936, "step": 56446 }, { "epoch": 4.43, "learning_rate": 5.573825766486317e-05, "loss": 2.7715, "step": 56447 }, { "epoch": 4.43, "learning_rate": 5.573747353563867e-05, "loss": 2.6403, "step": 56448 }, { "epoch": 4.43, "learning_rate": 5.5736689406414175e-05, "loss": 2.6146, "step": 56449 }, { "epoch": 4.43, "learning_rate": 5.573590527718968e-05, "loss": 2.6712, "step": 56450 }, { "epoch": 4.43, "learning_rate": 5.573512114796519e-05, "loss": 2.6342, "step": 56451 }, { "epoch": 4.43, "learning_rate": 5.573433701874069e-05, "loss": 2.8206, "step": 56452 }, { "epoch": 4.43, "learning_rate": 5.5733552889516196e-05, "loss": 2.73, "step": 56453 }, { "epoch": 4.43, "learning_rate": 5.57327687602917e-05, "loss": 2.7568, "step": 56454 }, { "epoch": 4.43, "learning_rate": 5.5731984631067204e-05, "loss": 2.6278, "step": 56455 }, { "epoch": 4.43, "learning_rate": 5.573120050184271e-05, "loss": 2.7279, "step": 56456 }, { "epoch": 4.43, "learning_rate": 5.573041637261821e-05, "loss": 3.0055, "step": 56457 }, { "epoch": 4.43, "learning_rate": 5.5729632243393715e-05, "loss": 2.5334, "step": 56458 }, { "epoch": 4.43, "learning_rate": 5.572884811416922e-05, "loss": 2.6172, "step": 56459 }, { "epoch": 4.43, "learning_rate": 5.572806398494472e-05, "loss": 2.6931, "step": 56460 }, { "epoch": 4.43, "learning_rate": 5.5727279855720226e-05, "loss": 2.4109, "step": 56461 }, { "epoch": 4.43, "learning_rate": 5.572649572649573e-05, "loss": 2.544, "step": 56462 }, { "epoch": 4.43, "learning_rate": 5.572571159727123e-05, "loss": 2.8205, "step": 56463 }, { "epoch": 4.43, "learning_rate": 5.572492746804674e-05, "loss": 2.6113, "step": 56464 }, { "epoch": 4.43, "learning_rate": 5.572414333882224e-05, "loss": 2.6905, "step": 56465 }, { "epoch": 4.43, "learning_rate": 5.5723359209597744e-05, "loss": 2.6766, "step": 56466 }, { "epoch": 4.43, "learning_rate": 5.5722575080373255e-05, "loss": 2.5907, "step": 56467 }, { "epoch": 4.43, "learning_rate": 5.572179095114876e-05, "loss": 2.5541, "step": 56468 }, { "epoch": 4.43, "learning_rate": 5.572100682192426e-05, "loss": 2.5249, "step": 56469 }, { "epoch": 4.43, "learning_rate": 5.5720222692699766e-05, "loss": 2.5921, "step": 56470 }, { "epoch": 4.43, "learning_rate": 5.571943856347527e-05, "loss": 2.9824, "step": 56471 }, { "epoch": 4.43, "learning_rate": 5.571865443425077e-05, "loss": 2.672, "step": 56472 }, { "epoch": 4.43, "learning_rate": 5.5717870305026277e-05, "loss": 2.7309, "step": 56473 }, { "epoch": 4.43, "learning_rate": 5.571708617580177e-05, "loss": 2.6015, "step": 56474 }, { "epoch": 4.43, "learning_rate": 5.571630204657727e-05, "loss": 2.6833, "step": 56475 }, { "epoch": 4.43, "learning_rate": 5.571551791735278e-05, "loss": 2.768, "step": 56476 }, { "epoch": 4.43, "learning_rate": 5.5714733788128285e-05, "loss": 2.6806, "step": 56477 }, { "epoch": 4.43, "learning_rate": 5.571394965890379e-05, "loss": 2.4977, "step": 56478 }, { "epoch": 4.43, "learning_rate": 5.571316552967929e-05, "loss": 2.5521, "step": 56479 }, { "epoch": 4.43, "learning_rate": 5.5712381400454796e-05, "loss": 2.9019, "step": 56480 }, { "epoch": 4.43, "learning_rate": 5.57115972712303e-05, "loss": 2.4697, "step": 56481 }, { "epoch": 4.43, "learning_rate": 5.57108131420058e-05, "loss": 2.7644, "step": 56482 }, { "epoch": 4.43, "learning_rate": 5.5710029012781307e-05, "loss": 2.5592, "step": 56483 }, { "epoch": 4.43, "learning_rate": 5.570924488355681e-05, "loss": 2.7129, "step": 56484 }, { "epoch": 4.43, "learning_rate": 5.5708460754332314e-05, "loss": 2.7635, "step": 56485 }, { "epoch": 4.43, "learning_rate": 5.570767662510782e-05, "loss": 2.5699, "step": 56486 }, { "epoch": 4.43, "learning_rate": 5.570689249588332e-05, "loss": 2.5472, "step": 56487 }, { "epoch": 4.43, "learning_rate": 5.5706108366658825e-05, "loss": 2.7544, "step": 56488 }, { "epoch": 4.43, "learning_rate": 5.570532423743433e-05, "loss": 2.6745, "step": 56489 }, { "epoch": 4.43, "learning_rate": 5.570454010820983e-05, "loss": 2.6354, "step": 56490 }, { "epoch": 4.43, "learning_rate": 5.5703755978985336e-05, "loss": 2.5583, "step": 56491 }, { "epoch": 4.43, "learning_rate": 5.5702971849760846e-05, "loss": 2.4678, "step": 56492 }, { "epoch": 4.43, "learning_rate": 5.570218772053635e-05, "loss": 2.7668, "step": 56493 }, { "epoch": 4.43, "learning_rate": 5.5701403591311854e-05, "loss": 2.9915, "step": 56494 }, { "epoch": 4.43, "learning_rate": 5.570061946208736e-05, "loss": 2.6322, "step": 56495 }, { "epoch": 4.43, "learning_rate": 5.569983533286286e-05, "loss": 2.579, "step": 56496 }, { "epoch": 4.43, "learning_rate": 5.5699051203638365e-05, "loss": 2.4437, "step": 56497 }, { "epoch": 4.43, "learning_rate": 5.569826707441387e-05, "loss": 2.7917, "step": 56498 }, { "epoch": 4.43, "learning_rate": 5.569748294518937e-05, "loss": 2.6225, "step": 56499 }, { "epoch": 4.43, "learning_rate": 5.5696698815964876e-05, "loss": 2.6179, "step": 56500 }, { "epoch": 4.43, "learning_rate": 5.569591468674038e-05, "loss": 2.417, "step": 56501 }, { "epoch": 4.43, "learning_rate": 5.569513055751588e-05, "loss": 2.7109, "step": 56502 }, { "epoch": 4.43, "learning_rate": 5.569434642829139e-05, "loss": 2.6433, "step": 56503 }, { "epoch": 4.43, "learning_rate": 5.569356229906689e-05, "loss": 2.6266, "step": 56504 }, { "epoch": 4.43, "learning_rate": 5.5692778169842394e-05, "loss": 2.7616, "step": 56505 }, { "epoch": 4.43, "learning_rate": 5.56919940406179e-05, "loss": 2.4536, "step": 56506 }, { "epoch": 4.43, "learning_rate": 5.56912099113934e-05, "loss": 2.8421, "step": 56507 }, { "epoch": 4.43, "learning_rate": 5.569042578216891e-05, "loss": 2.7197, "step": 56508 }, { "epoch": 4.43, "learning_rate": 5.5689641652944416e-05, "loss": 2.8788, "step": 56509 }, { "epoch": 4.43, "learning_rate": 5.5688857523719906e-05, "loss": 2.6831, "step": 56510 }, { "epoch": 4.43, "learning_rate": 5.568807339449541e-05, "loss": 2.8388, "step": 56511 }, { "epoch": 4.43, "learning_rate": 5.568728926527091e-05, "loss": 2.7589, "step": 56512 }, { "epoch": 4.43, "learning_rate": 5.568650513604642e-05, "loss": 2.5706, "step": 56513 }, { "epoch": 4.43, "learning_rate": 5.568572100682192e-05, "loss": 2.5621, "step": 56514 }, { "epoch": 4.43, "learning_rate": 5.5684936877597424e-05, "loss": 2.68, "step": 56515 }, { "epoch": 4.43, "learning_rate": 5.568415274837293e-05, "loss": 2.696, "step": 56516 }, { "epoch": 4.43, "learning_rate": 5.568336861914843e-05, "loss": 2.6426, "step": 56517 }, { "epoch": 4.43, "learning_rate": 5.568258448992394e-05, "loss": 2.5309, "step": 56518 }, { "epoch": 4.43, "learning_rate": 5.5681800360699446e-05, "loss": 2.4765, "step": 56519 }, { "epoch": 4.43, "learning_rate": 5.568101623147495e-05, "loss": 2.5607, "step": 56520 }, { "epoch": 4.43, "learning_rate": 5.568023210225045e-05, "loss": 2.639, "step": 56521 }, { "epoch": 4.43, "learning_rate": 5.567944797302596e-05, "loss": 2.5743, "step": 56522 }, { "epoch": 4.43, "learning_rate": 5.567866384380146e-05, "loss": 2.5394, "step": 56523 }, { "epoch": 4.43, "learning_rate": 5.5677879714576964e-05, "loss": 2.5722, "step": 56524 }, { "epoch": 4.43, "learning_rate": 5.567709558535247e-05, "loss": 2.6095, "step": 56525 }, { "epoch": 4.43, "learning_rate": 5.567631145612797e-05, "loss": 2.6517, "step": 56526 }, { "epoch": 4.43, "learning_rate": 5.5675527326903475e-05, "loss": 2.5236, "step": 56527 }, { "epoch": 4.43, "learning_rate": 5.567474319767898e-05, "loss": 2.615, "step": 56528 }, { "epoch": 4.43, "learning_rate": 5.567395906845448e-05, "loss": 2.6925, "step": 56529 }, { "epoch": 4.43, "learning_rate": 5.5673174939229986e-05, "loss": 2.7413, "step": 56530 }, { "epoch": 4.43, "learning_rate": 5.567239081000549e-05, "loss": 2.7211, "step": 56531 }, { "epoch": 4.43, "learning_rate": 5.5671606680780994e-05, "loss": 2.5167, "step": 56532 }, { "epoch": 4.43, "learning_rate": 5.5670822551556504e-05, "loss": 2.7426, "step": 56533 }, { "epoch": 4.43, "learning_rate": 5.567003842233201e-05, "loss": 2.7834, "step": 56534 }, { "epoch": 4.43, "learning_rate": 5.566925429310751e-05, "loss": 2.5822, "step": 56535 }, { "epoch": 4.43, "learning_rate": 5.5668470163883015e-05, "loss": 2.5894, "step": 56536 }, { "epoch": 4.43, "learning_rate": 5.566768603465852e-05, "loss": 2.5329, "step": 56537 }, { "epoch": 4.43, "learning_rate": 5.566690190543402e-05, "loss": 2.7453, "step": 56538 }, { "epoch": 4.43, "learning_rate": 5.5666117776209526e-05, "loss": 2.6802, "step": 56539 }, { "epoch": 4.43, "learning_rate": 5.566533364698503e-05, "loss": 2.4322, "step": 56540 }, { "epoch": 4.43, "learning_rate": 5.5664549517760533e-05, "loss": 2.4855, "step": 56541 }, { "epoch": 4.43, "learning_rate": 5.566376538853604e-05, "loss": 2.6555, "step": 56542 }, { "epoch": 4.43, "learning_rate": 5.566298125931154e-05, "loss": 2.6589, "step": 56543 }, { "epoch": 4.43, "learning_rate": 5.5662197130087044e-05, "loss": 2.6353, "step": 56544 }, { "epoch": 4.43, "learning_rate": 5.566141300086255e-05, "loss": 2.8158, "step": 56545 }, { "epoch": 4.43, "learning_rate": 5.5660628871638045e-05, "loss": 2.6642, "step": 56546 }, { "epoch": 4.43, "learning_rate": 5.565984474241355e-05, "loss": 2.6423, "step": 56547 }, { "epoch": 4.43, "learning_rate": 5.565906061318905e-05, "loss": 2.6416, "step": 56548 }, { "epoch": 4.43, "learning_rate": 5.5658276483964556e-05, "loss": 2.6315, "step": 56549 }, { "epoch": 4.43, "learning_rate": 5.565749235474006e-05, "loss": 2.803, "step": 56550 }, { "epoch": 4.43, "learning_rate": 5.5656708225515563e-05, "loss": 2.7147, "step": 56551 }, { "epoch": 4.43, "learning_rate": 5.565592409629107e-05, "loss": 2.5496, "step": 56552 }, { "epoch": 4.43, "learning_rate": 5.565513996706657e-05, "loss": 2.7225, "step": 56553 }, { "epoch": 4.43, "learning_rate": 5.5654355837842074e-05, "loss": 2.5586, "step": 56554 }, { "epoch": 4.43, "learning_rate": 5.565357170861758e-05, "loss": 2.6585, "step": 56555 }, { "epoch": 4.43, "learning_rate": 5.565278757939308e-05, "loss": 3.0073, "step": 56556 }, { "epoch": 4.43, "learning_rate": 5.5652003450168586e-05, "loss": 2.6873, "step": 56557 }, { "epoch": 4.43, "learning_rate": 5.565121932094409e-05, "loss": 2.4839, "step": 56558 }, { "epoch": 4.43, "learning_rate": 5.56504351917196e-05, "loss": 2.6096, "step": 56559 }, { "epoch": 4.44, "learning_rate": 5.56496510624951e-05, "loss": 2.7435, "step": 56560 }, { "epoch": 4.44, "learning_rate": 5.564886693327061e-05, "loss": 2.9772, "step": 56561 }, { "epoch": 4.44, "learning_rate": 5.564808280404611e-05, "loss": 2.6092, "step": 56562 }, { "epoch": 4.44, "learning_rate": 5.5647298674821614e-05, "loss": 2.6094, "step": 56563 }, { "epoch": 4.44, "learning_rate": 5.564651454559712e-05, "loss": 2.5487, "step": 56564 }, { "epoch": 4.44, "learning_rate": 5.564573041637262e-05, "loss": 2.5113, "step": 56565 }, { "epoch": 4.44, "learning_rate": 5.5644946287148125e-05, "loss": 2.5275, "step": 56566 }, { "epoch": 4.44, "learning_rate": 5.564416215792363e-05, "loss": 2.8007, "step": 56567 }, { "epoch": 4.44, "learning_rate": 5.564337802869913e-05, "loss": 2.5584, "step": 56568 }, { "epoch": 4.44, "learning_rate": 5.5642593899474636e-05, "loss": 2.7569, "step": 56569 }, { "epoch": 4.44, "learning_rate": 5.564180977025014e-05, "loss": 2.6187, "step": 56570 }, { "epoch": 4.44, "learning_rate": 5.5641025641025644e-05, "loss": 2.6369, "step": 56571 }, { "epoch": 4.44, "learning_rate": 5.564024151180115e-05, "loss": 2.6457, "step": 56572 }, { "epoch": 4.44, "learning_rate": 5.563945738257665e-05, "loss": 2.6574, "step": 56573 }, { "epoch": 4.44, "learning_rate": 5.5638673253352155e-05, "loss": 2.6672, "step": 56574 }, { "epoch": 4.44, "learning_rate": 5.5637889124127665e-05, "loss": 2.6618, "step": 56575 }, { "epoch": 4.44, "learning_rate": 5.563710499490317e-05, "loss": 2.6516, "step": 56576 }, { "epoch": 4.44, "learning_rate": 5.563632086567867e-05, "loss": 2.7615, "step": 56577 }, { "epoch": 4.44, "learning_rate": 5.5635536736454176e-05, "loss": 2.6764, "step": 56578 }, { "epoch": 4.44, "learning_rate": 5.563475260722968e-05, "loss": 2.6082, "step": 56579 }, { "epoch": 4.44, "learning_rate": 5.5633968478005184e-05, "loss": 2.5982, "step": 56580 }, { "epoch": 4.44, "learning_rate": 5.563318434878069e-05, "loss": 2.6564, "step": 56581 }, { "epoch": 4.44, "learning_rate": 5.563240021955618e-05, "loss": 2.6958, "step": 56582 }, { "epoch": 4.44, "learning_rate": 5.563161609033168e-05, "loss": 2.5068, "step": 56583 }, { "epoch": 4.44, "learning_rate": 5.5630831961107185e-05, "loss": 2.7274, "step": 56584 }, { "epoch": 4.44, "learning_rate": 5.5630047831882695e-05, "loss": 2.6199, "step": 56585 }, { "epoch": 4.44, "learning_rate": 5.56292637026582e-05, "loss": 2.6688, "step": 56586 }, { "epoch": 4.44, "learning_rate": 5.56284795734337e-05, "loss": 2.7575, "step": 56587 }, { "epoch": 4.44, "learning_rate": 5.5627695444209206e-05, "loss": 2.5372, "step": 56588 }, { "epoch": 4.44, "learning_rate": 5.562691131498471e-05, "loss": 2.6272, "step": 56589 }, { "epoch": 4.44, "learning_rate": 5.5626127185760214e-05, "loss": 2.6691, "step": 56590 }, { "epoch": 4.44, "learning_rate": 5.562534305653572e-05, "loss": 2.598, "step": 56591 }, { "epoch": 4.44, "learning_rate": 5.562455892731122e-05, "loss": 2.9036, "step": 56592 }, { "epoch": 4.44, "learning_rate": 5.5623774798086725e-05, "loss": 2.4915, "step": 56593 }, { "epoch": 4.44, "learning_rate": 5.562299066886223e-05, "loss": 2.6352, "step": 56594 }, { "epoch": 4.44, "learning_rate": 5.562220653963773e-05, "loss": 2.5391, "step": 56595 }, { "epoch": 4.44, "learning_rate": 5.5621422410413236e-05, "loss": 2.6563, "step": 56596 }, { "epoch": 4.44, "learning_rate": 5.562063828118874e-05, "loss": 2.9903, "step": 56597 }, { "epoch": 4.44, "learning_rate": 5.561985415196424e-05, "loss": 2.7538, "step": 56598 }, { "epoch": 4.44, "learning_rate": 5.561907002273975e-05, "loss": 2.6241, "step": 56599 }, { "epoch": 4.44, "learning_rate": 5.561828589351526e-05, "loss": 2.5858, "step": 56600 }, { "epoch": 4.44, "learning_rate": 5.561750176429076e-05, "loss": 2.6904, "step": 56601 }, { "epoch": 4.44, "learning_rate": 5.5616717635066265e-05, "loss": 2.5228, "step": 56602 }, { "epoch": 4.44, "learning_rate": 5.561593350584177e-05, "loss": 2.754, "step": 56603 }, { "epoch": 4.44, "learning_rate": 5.561514937661727e-05, "loss": 2.4885, "step": 56604 }, { "epoch": 4.44, "learning_rate": 5.5614365247392776e-05, "loss": 2.8075, "step": 56605 }, { "epoch": 4.44, "learning_rate": 5.561358111816828e-05, "loss": 2.6929, "step": 56606 }, { "epoch": 4.44, "learning_rate": 5.561279698894378e-05, "loss": 2.6616, "step": 56607 }, { "epoch": 4.44, "learning_rate": 5.561201285971929e-05, "loss": 2.8054, "step": 56608 }, { "epoch": 4.44, "learning_rate": 5.561122873049479e-05, "loss": 2.6124, "step": 56609 }, { "epoch": 4.44, "learning_rate": 5.5610444601270294e-05, "loss": 2.7181, "step": 56610 }, { "epoch": 4.44, "learning_rate": 5.56096604720458e-05, "loss": 2.5729, "step": 56611 }, { "epoch": 4.44, "learning_rate": 5.56088763428213e-05, "loss": 2.9972, "step": 56612 }, { "epoch": 4.44, "learning_rate": 5.5608092213596805e-05, "loss": 2.7252, "step": 56613 }, { "epoch": 4.44, "learning_rate": 5.560730808437231e-05, "loss": 2.6038, "step": 56614 }, { "epoch": 4.44, "learning_rate": 5.560652395514781e-05, "loss": 2.4577, "step": 56615 }, { "epoch": 4.44, "learning_rate": 5.560573982592332e-05, "loss": 2.5776, "step": 56616 }, { "epoch": 4.44, "learning_rate": 5.5604955696698827e-05, "loss": 2.5836, "step": 56617 }, { "epoch": 4.44, "learning_rate": 5.560417156747432e-05, "loss": 2.7196, "step": 56618 }, { "epoch": 4.44, "learning_rate": 5.560338743824982e-05, "loss": 2.6069, "step": 56619 }, { "epoch": 4.44, "learning_rate": 5.5602603309025324e-05, "loss": 2.5742, "step": 56620 }, { "epoch": 4.44, "learning_rate": 5.560181917980083e-05, "loss": 2.7211, "step": 56621 }, { "epoch": 4.44, "learning_rate": 5.560103505057633e-05, "loss": 2.528, "step": 56622 }, { "epoch": 4.44, "learning_rate": 5.5600250921351835e-05, "loss": 2.82, "step": 56623 }, { "epoch": 4.44, "learning_rate": 5.559946679212734e-05, "loss": 2.7164, "step": 56624 }, { "epoch": 4.44, "learning_rate": 5.559868266290284e-05, "loss": 2.63, "step": 56625 }, { "epoch": 4.44, "learning_rate": 5.559789853367835e-05, "loss": 2.8764, "step": 56626 }, { "epoch": 4.44, "learning_rate": 5.5597114404453857e-05, "loss": 2.6654, "step": 56627 }, { "epoch": 4.44, "learning_rate": 5.559633027522936e-05, "loss": 2.7207, "step": 56628 }, { "epoch": 4.44, "learning_rate": 5.5595546146004864e-05, "loss": 2.6012, "step": 56629 }, { "epoch": 4.44, "learning_rate": 5.559476201678037e-05, "loss": 2.7229, "step": 56630 }, { "epoch": 4.44, "learning_rate": 5.559397788755587e-05, "loss": 2.603, "step": 56631 }, { "epoch": 4.44, "learning_rate": 5.5593193758331375e-05, "loss": 2.7445, "step": 56632 }, { "epoch": 4.44, "learning_rate": 5.559240962910688e-05, "loss": 2.9534, "step": 56633 }, { "epoch": 4.44, "learning_rate": 5.559162549988238e-05, "loss": 2.7383, "step": 56634 }, { "epoch": 4.44, "learning_rate": 5.5590841370657886e-05, "loss": 2.7217, "step": 56635 }, { "epoch": 4.44, "learning_rate": 5.559005724143339e-05, "loss": 2.4643, "step": 56636 }, { "epoch": 4.44, "learning_rate": 5.558927311220889e-05, "loss": 2.6685, "step": 56637 }, { "epoch": 4.44, "learning_rate": 5.55884889829844e-05, "loss": 2.6106, "step": 56638 }, { "epoch": 4.44, "learning_rate": 5.55877048537599e-05, "loss": 2.6359, "step": 56639 }, { "epoch": 4.44, "learning_rate": 5.5586920724535404e-05, "loss": 2.5991, "step": 56640 }, { "epoch": 4.44, "learning_rate": 5.5586136595310915e-05, "loss": 2.6262, "step": 56641 }, { "epoch": 4.44, "learning_rate": 5.558535246608642e-05, "loss": 2.6114, "step": 56642 }, { "epoch": 4.44, "learning_rate": 5.558456833686192e-05, "loss": 2.6914, "step": 56643 }, { "epoch": 4.44, "learning_rate": 5.5583784207637426e-05, "loss": 2.6455, "step": 56644 }, { "epoch": 4.44, "learning_rate": 5.558300007841293e-05, "loss": 2.7074, "step": 56645 }, { "epoch": 4.44, "learning_rate": 5.558221594918843e-05, "loss": 2.6948, "step": 56646 }, { "epoch": 4.44, "learning_rate": 5.558143181996394e-05, "loss": 2.827, "step": 56647 }, { "epoch": 4.44, "learning_rate": 5.558064769073944e-05, "loss": 2.6813, "step": 56648 }, { "epoch": 4.44, "learning_rate": 5.5579863561514944e-05, "loss": 2.6037, "step": 56649 }, { "epoch": 4.44, "learning_rate": 5.557907943229045e-05, "loss": 2.6635, "step": 56650 }, { "epoch": 4.44, "learning_rate": 5.557829530306595e-05, "loss": 2.5564, "step": 56651 }, { "epoch": 4.44, "learning_rate": 5.5577511173841455e-05, "loss": 2.78, "step": 56652 }, { "epoch": 4.44, "learning_rate": 5.557672704461696e-05, "loss": 2.8924, "step": 56653 }, { "epoch": 4.44, "learning_rate": 5.5575942915392456e-05, "loss": 2.7896, "step": 56654 }, { "epoch": 4.44, "learning_rate": 5.557515878616796e-05, "loss": 2.5509, "step": 56655 }, { "epoch": 4.44, "learning_rate": 5.557437465694346e-05, "loss": 2.5335, "step": 56656 }, { "epoch": 4.44, "learning_rate": 5.557359052771897e-05, "loss": 2.5135, "step": 56657 }, { "epoch": 4.44, "learning_rate": 5.557280639849447e-05, "loss": 2.6768, "step": 56658 }, { "epoch": 4.44, "learning_rate": 5.5572022269269974e-05, "loss": 2.7428, "step": 56659 }, { "epoch": 4.44, "learning_rate": 5.557123814004548e-05, "loss": 2.7941, "step": 56660 }, { "epoch": 4.44, "learning_rate": 5.557045401082098e-05, "loss": 2.7205, "step": 56661 }, { "epoch": 4.44, "learning_rate": 5.5569669881596485e-05, "loss": 2.6546, "step": 56662 }, { "epoch": 4.44, "learning_rate": 5.556888575237199e-05, "loss": 2.6133, "step": 56663 }, { "epoch": 4.44, "learning_rate": 5.556810162314749e-05, "loss": 2.5601, "step": 56664 }, { "epoch": 4.44, "learning_rate": 5.5567317493922996e-05, "loss": 2.7039, "step": 56665 }, { "epoch": 4.44, "learning_rate": 5.55665333646985e-05, "loss": 2.6033, "step": 56666 }, { "epoch": 4.44, "learning_rate": 5.556574923547401e-05, "loss": 2.516, "step": 56667 }, { "epoch": 4.44, "learning_rate": 5.5564965106249514e-05, "loss": 2.5184, "step": 56668 }, { "epoch": 4.44, "learning_rate": 5.556418097702502e-05, "loss": 2.6345, "step": 56669 }, { "epoch": 4.44, "learning_rate": 5.556339684780052e-05, "loss": 2.6734, "step": 56670 }, { "epoch": 4.44, "learning_rate": 5.5562612718576025e-05, "loss": 2.634, "step": 56671 }, { "epoch": 4.44, "learning_rate": 5.556182858935153e-05, "loss": 2.8236, "step": 56672 }, { "epoch": 4.44, "learning_rate": 5.556104446012703e-05, "loss": 2.5382, "step": 56673 }, { "epoch": 4.44, "learning_rate": 5.5560260330902536e-05, "loss": 2.8127, "step": 56674 }, { "epoch": 4.44, "learning_rate": 5.555947620167804e-05, "loss": 2.5627, "step": 56675 }, { "epoch": 4.44, "learning_rate": 5.5558692072453544e-05, "loss": 2.5234, "step": 56676 }, { "epoch": 4.44, "learning_rate": 5.555790794322905e-05, "loss": 2.7918, "step": 56677 }, { "epoch": 4.44, "learning_rate": 5.555712381400455e-05, "loss": 2.5664, "step": 56678 }, { "epoch": 4.44, "learning_rate": 5.5556339684780055e-05, "loss": 2.6048, "step": 56679 }, { "epoch": 4.44, "learning_rate": 5.555555555555556e-05, "loss": 2.8106, "step": 56680 }, { "epoch": 4.44, "learning_rate": 5.555477142633106e-05, "loss": 2.7556, "step": 56681 }, { "epoch": 4.44, "learning_rate": 5.5553987297106566e-05, "loss": 2.6191, "step": 56682 }, { "epoch": 4.44, "learning_rate": 5.5553203167882076e-05, "loss": 2.5899, "step": 56683 }, { "epoch": 4.44, "learning_rate": 5.555241903865758e-05, "loss": 2.7771, "step": 56684 }, { "epoch": 4.44, "learning_rate": 5.555163490943308e-05, "loss": 2.5842, "step": 56685 }, { "epoch": 4.44, "learning_rate": 5.555085078020859e-05, "loss": 2.6418, "step": 56686 }, { "epoch": 4.44, "learning_rate": 5.555006665098409e-05, "loss": 2.5898, "step": 56687 }, { "epoch": 4.45, "learning_rate": 5.5549282521759594e-05, "loss": 2.4611, "step": 56688 }, { "epoch": 4.45, "learning_rate": 5.55484983925351e-05, "loss": 2.3528, "step": 56689 }, { "epoch": 4.45, "learning_rate": 5.55477142633106e-05, "loss": 2.4597, "step": 56690 }, { "epoch": 4.45, "learning_rate": 5.554693013408609e-05, "loss": 2.6152, "step": 56691 }, { "epoch": 4.45, "learning_rate": 5.5546146004861596e-05, "loss": 2.6116, "step": 56692 }, { "epoch": 4.45, "learning_rate": 5.5545361875637106e-05, "loss": 2.6189, "step": 56693 }, { "epoch": 4.45, "learning_rate": 5.554457774641261e-05, "loss": 2.5159, "step": 56694 }, { "epoch": 4.45, "learning_rate": 5.554379361718811e-05, "loss": 2.6393, "step": 56695 }, { "epoch": 4.45, "learning_rate": 5.554300948796362e-05, "loss": 2.5499, "step": 56696 }, { "epoch": 4.45, "learning_rate": 5.554222535873912e-05, "loss": 2.7455, "step": 56697 }, { "epoch": 4.45, "learning_rate": 5.5541441229514624e-05, "loss": 2.6831, "step": 56698 }, { "epoch": 4.45, "learning_rate": 5.554065710029013e-05, "loss": 2.7276, "step": 56699 }, { "epoch": 4.45, "learning_rate": 5.553987297106563e-05, "loss": 2.5449, "step": 56700 }, { "epoch": 4.45, "learning_rate": 5.5539088841841135e-05, "loss": 2.7375, "step": 56701 }, { "epoch": 4.45, "learning_rate": 5.553830471261664e-05, "loss": 2.5271, "step": 56702 }, { "epoch": 4.45, "learning_rate": 5.553752058339214e-05, "loss": 2.8084, "step": 56703 }, { "epoch": 4.45, "learning_rate": 5.5536736454167646e-05, "loss": 2.613, "step": 56704 }, { "epoch": 4.45, "learning_rate": 5.553595232494315e-05, "loss": 2.6116, "step": 56705 }, { "epoch": 4.45, "learning_rate": 5.5535168195718654e-05, "loss": 2.7066, "step": 56706 }, { "epoch": 4.45, "learning_rate": 5.553438406649416e-05, "loss": 2.8043, "step": 56707 }, { "epoch": 4.45, "learning_rate": 5.553359993726967e-05, "loss": 2.7341, "step": 56708 }, { "epoch": 4.45, "learning_rate": 5.553281580804517e-05, "loss": 2.7247, "step": 56709 }, { "epoch": 4.45, "learning_rate": 5.5532031678820675e-05, "loss": 2.575, "step": 56710 }, { "epoch": 4.45, "learning_rate": 5.553124754959618e-05, "loss": 2.7276, "step": 56711 }, { "epoch": 4.45, "learning_rate": 5.553046342037168e-05, "loss": 2.6754, "step": 56712 }, { "epoch": 4.45, "learning_rate": 5.5529679291147186e-05, "loss": 2.7186, "step": 56713 }, { "epoch": 4.45, "learning_rate": 5.552889516192269e-05, "loss": 2.9947, "step": 56714 }, { "epoch": 4.45, "learning_rate": 5.5528111032698194e-05, "loss": 2.6941, "step": 56715 }, { "epoch": 4.45, "learning_rate": 5.55273269034737e-05, "loss": 2.6461, "step": 56716 }, { "epoch": 4.45, "learning_rate": 5.55265427742492e-05, "loss": 2.6402, "step": 56717 }, { "epoch": 4.45, "learning_rate": 5.5525758645024705e-05, "loss": 2.5518, "step": 56718 }, { "epoch": 4.45, "learning_rate": 5.552497451580021e-05, "loss": 2.5677, "step": 56719 }, { "epoch": 4.45, "learning_rate": 5.552419038657571e-05, "loss": 2.6125, "step": 56720 }, { "epoch": 4.45, "learning_rate": 5.5523406257351216e-05, "loss": 2.6371, "step": 56721 }, { "epoch": 4.45, "learning_rate": 5.552262212812672e-05, "loss": 2.7031, "step": 56722 }, { "epoch": 4.45, "learning_rate": 5.552183799890222e-05, "loss": 2.6449, "step": 56723 }, { "epoch": 4.45, "learning_rate": 5.5521053869677734e-05, "loss": 2.5199, "step": 56724 }, { "epoch": 4.45, "learning_rate": 5.552026974045324e-05, "loss": 2.5836, "step": 56725 }, { "epoch": 4.45, "learning_rate": 5.551948561122874e-05, "loss": 2.6639, "step": 56726 }, { "epoch": 4.45, "learning_rate": 5.551870148200423e-05, "loss": 2.5951, "step": 56727 }, { "epoch": 4.45, "learning_rate": 5.5517917352779735e-05, "loss": 2.6007, "step": 56728 }, { "epoch": 4.45, "learning_rate": 5.551713322355524e-05, "loss": 2.5846, "step": 56729 }, { "epoch": 4.45, "learning_rate": 5.551634909433074e-05, "loss": 2.5998, "step": 56730 }, { "epoch": 4.45, "learning_rate": 5.5515564965106246e-05, "loss": 2.6321, "step": 56731 }, { "epoch": 4.45, "learning_rate": 5.551478083588175e-05, "loss": 2.8013, "step": 56732 }, { "epoch": 4.45, "learning_rate": 5.551399670665725e-05, "loss": 2.8306, "step": 56733 }, { "epoch": 4.45, "learning_rate": 5.5513212577432764e-05, "loss": 2.6667, "step": 56734 }, { "epoch": 4.45, "learning_rate": 5.551242844820827e-05, "loss": 3.1497, "step": 56735 }, { "epoch": 4.45, "learning_rate": 5.551164431898377e-05, "loss": 2.4815, "step": 56736 }, { "epoch": 4.45, "learning_rate": 5.5510860189759275e-05, "loss": 2.7073, "step": 56737 }, { "epoch": 4.45, "learning_rate": 5.551007606053478e-05, "loss": 2.5011, "step": 56738 }, { "epoch": 4.45, "learning_rate": 5.550929193131028e-05, "loss": 2.6617, "step": 56739 }, { "epoch": 4.45, "learning_rate": 5.5508507802085786e-05, "loss": 2.6888, "step": 56740 }, { "epoch": 4.45, "learning_rate": 5.550772367286129e-05, "loss": 2.6113, "step": 56741 }, { "epoch": 4.45, "learning_rate": 5.550693954363679e-05, "loss": 2.6229, "step": 56742 }, { "epoch": 4.45, "learning_rate": 5.55061554144123e-05, "loss": 2.6317, "step": 56743 }, { "epoch": 4.45, "learning_rate": 5.55053712851878e-05, "loss": 2.7385, "step": 56744 }, { "epoch": 4.45, "learning_rate": 5.5504587155963304e-05, "loss": 2.6816, "step": 56745 }, { "epoch": 4.45, "learning_rate": 5.550380302673881e-05, "loss": 2.6825, "step": 56746 }, { "epoch": 4.45, "learning_rate": 5.550301889751431e-05, "loss": 2.4587, "step": 56747 }, { "epoch": 4.45, "learning_rate": 5.5502234768289815e-05, "loss": 2.7404, "step": 56748 }, { "epoch": 4.45, "learning_rate": 5.5501450639065326e-05, "loss": 2.6728, "step": 56749 }, { "epoch": 4.45, "learning_rate": 5.550066650984083e-05, "loss": 2.5151, "step": 56750 }, { "epoch": 4.45, "learning_rate": 5.549988238061633e-05, "loss": 2.7601, "step": 56751 }, { "epoch": 4.45, "learning_rate": 5.5499098251391837e-05, "loss": 2.7576, "step": 56752 }, { "epoch": 4.45, "learning_rate": 5.549831412216734e-05, "loss": 2.7328, "step": 56753 }, { "epoch": 4.45, "learning_rate": 5.5497529992942844e-05, "loss": 2.5997, "step": 56754 }, { "epoch": 4.45, "learning_rate": 5.549674586371835e-05, "loss": 2.6381, "step": 56755 }, { "epoch": 4.45, "learning_rate": 5.549596173449385e-05, "loss": 2.539, "step": 56756 }, { "epoch": 4.45, "learning_rate": 5.5495177605269355e-05, "loss": 2.7429, "step": 56757 }, { "epoch": 4.45, "learning_rate": 5.549439347604486e-05, "loss": 2.787, "step": 56758 }, { "epoch": 4.45, "learning_rate": 5.549360934682036e-05, "loss": 2.6269, "step": 56759 }, { "epoch": 4.45, "learning_rate": 5.5492825217595866e-05, "loss": 2.6785, "step": 56760 }, { "epoch": 4.45, "learning_rate": 5.549204108837137e-05, "loss": 2.6869, "step": 56761 }, { "epoch": 4.45, "learning_rate": 5.549125695914687e-05, "loss": 2.7618, "step": 56762 }, { "epoch": 4.45, "learning_rate": 5.549047282992237e-05, "loss": 2.5721, "step": 56763 }, { "epoch": 4.45, "learning_rate": 5.5489688700697874e-05, "loss": 2.4118, "step": 56764 }, { "epoch": 4.45, "learning_rate": 5.548890457147338e-05, "loss": 2.8468, "step": 56765 }, { "epoch": 4.45, "learning_rate": 5.548812044224888e-05, "loss": 2.7272, "step": 56766 }, { "epoch": 4.45, "learning_rate": 5.5487336313024385e-05, "loss": 2.7051, "step": 56767 }, { "epoch": 4.45, "learning_rate": 5.548655218379989e-05, "loss": 2.8189, "step": 56768 }, { "epoch": 4.45, "learning_rate": 5.548576805457539e-05, "loss": 2.7824, "step": 56769 }, { "epoch": 4.45, "learning_rate": 5.5484983925350896e-05, "loss": 2.7271, "step": 56770 }, { "epoch": 4.45, "learning_rate": 5.54841997961264e-05, "loss": 2.6174, "step": 56771 }, { "epoch": 4.45, "learning_rate": 5.54834156669019e-05, "loss": 2.8099, "step": 56772 }, { "epoch": 4.45, "learning_rate": 5.548263153767741e-05, "loss": 2.6207, "step": 56773 }, { "epoch": 4.45, "learning_rate": 5.548184740845291e-05, "loss": 2.712, "step": 56774 }, { "epoch": 4.45, "learning_rate": 5.548106327922842e-05, "loss": 2.8286, "step": 56775 }, { "epoch": 4.45, "learning_rate": 5.5480279150003925e-05, "loss": 2.6726, "step": 56776 }, { "epoch": 4.45, "learning_rate": 5.547949502077943e-05, "loss": 2.7592, "step": 56777 }, { "epoch": 4.45, "learning_rate": 5.547871089155493e-05, "loss": 2.5546, "step": 56778 }, { "epoch": 4.45, "learning_rate": 5.5477926762330436e-05, "loss": 2.5355, "step": 56779 }, { "epoch": 4.45, "learning_rate": 5.547714263310594e-05, "loss": 2.8753, "step": 56780 }, { "epoch": 4.45, "learning_rate": 5.547635850388144e-05, "loss": 2.6976, "step": 56781 }, { "epoch": 4.45, "learning_rate": 5.547557437465695e-05, "loss": 2.689, "step": 56782 }, { "epoch": 4.45, "learning_rate": 5.547479024543245e-05, "loss": 2.7355, "step": 56783 }, { "epoch": 4.45, "learning_rate": 5.5474006116207954e-05, "loss": 2.6152, "step": 56784 }, { "epoch": 4.45, "learning_rate": 5.547322198698346e-05, "loss": 2.5859, "step": 56785 }, { "epoch": 4.45, "learning_rate": 5.547243785775896e-05, "loss": 2.5878, "step": 56786 }, { "epoch": 4.45, "learning_rate": 5.5471653728534465e-05, "loss": 2.6758, "step": 56787 }, { "epoch": 4.45, "learning_rate": 5.547086959930997e-05, "loss": 2.8949, "step": 56788 }, { "epoch": 4.45, "learning_rate": 5.547008547008547e-05, "loss": 2.5083, "step": 56789 }, { "epoch": 4.45, "learning_rate": 5.5469301340860976e-05, "loss": 2.6173, "step": 56790 }, { "epoch": 4.45, "learning_rate": 5.546851721163649e-05, "loss": 2.6269, "step": 56791 }, { "epoch": 4.45, "learning_rate": 5.546773308241199e-05, "loss": 2.6897, "step": 56792 }, { "epoch": 4.45, "learning_rate": 5.5466948953187494e-05, "loss": 2.5847, "step": 56793 }, { "epoch": 4.45, "learning_rate": 5.5466164823963e-05, "loss": 2.5397, "step": 56794 }, { "epoch": 4.45, "learning_rate": 5.54653806947385e-05, "loss": 2.7643, "step": 56795 }, { "epoch": 4.45, "learning_rate": 5.5464596565514005e-05, "loss": 2.7444, "step": 56796 }, { "epoch": 4.45, "learning_rate": 5.546381243628951e-05, "loss": 2.7974, "step": 56797 }, { "epoch": 4.45, "learning_rate": 5.546302830706501e-05, "loss": 2.7418, "step": 56798 }, { "epoch": 4.45, "learning_rate": 5.54622441778405e-05, "loss": 2.5555, "step": 56799 }, { "epoch": 4.45, "learning_rate": 5.5461460048616006e-05, "loss": 2.4488, "step": 56800 }, { "epoch": 4.45, "learning_rate": 5.546067591939152e-05, "loss": 2.7396, "step": 56801 }, { "epoch": 4.45, "learning_rate": 5.545989179016702e-05, "loss": 2.5746, "step": 56802 }, { "epoch": 4.45, "learning_rate": 5.5459107660942524e-05, "loss": 2.666, "step": 56803 }, { "epoch": 4.45, "learning_rate": 5.545832353171803e-05, "loss": 2.5729, "step": 56804 }, { "epoch": 4.45, "learning_rate": 5.545753940249353e-05, "loss": 2.7458, "step": 56805 }, { "epoch": 4.45, "learning_rate": 5.5456755273269035e-05, "loss": 2.6934, "step": 56806 }, { "epoch": 4.45, "learning_rate": 5.545597114404454e-05, "loss": 2.6066, "step": 56807 }, { "epoch": 4.45, "learning_rate": 5.545518701482004e-05, "loss": 2.6761, "step": 56808 }, { "epoch": 4.45, "learning_rate": 5.5454402885595546e-05, "loss": 2.6036, "step": 56809 }, { "epoch": 4.45, "learning_rate": 5.545361875637105e-05, "loss": 2.6003, "step": 56810 }, { "epoch": 4.45, "learning_rate": 5.5452834627146554e-05, "loss": 2.6754, "step": 56811 }, { "epoch": 4.45, "learning_rate": 5.545205049792206e-05, "loss": 2.7062, "step": 56812 }, { "epoch": 4.45, "learning_rate": 5.545126636869756e-05, "loss": 2.8317, "step": 56813 }, { "epoch": 4.45, "learning_rate": 5.5450482239473065e-05, "loss": 2.7015, "step": 56814 }, { "epoch": 4.46, "learning_rate": 5.544969811024857e-05, "loss": 2.8026, "step": 56815 }, { "epoch": 4.46, "learning_rate": 5.544891398102408e-05, "loss": 2.7041, "step": 56816 }, { "epoch": 4.46, "learning_rate": 5.544812985179958e-05, "loss": 2.7605, "step": 56817 }, { "epoch": 4.46, "learning_rate": 5.5447345722575086e-05, "loss": 2.5643, "step": 56818 }, { "epoch": 4.46, "learning_rate": 5.544656159335059e-05, "loss": 2.6414, "step": 56819 }, { "epoch": 4.46, "learning_rate": 5.5445777464126093e-05, "loss": 2.7439, "step": 56820 }, { "epoch": 4.46, "learning_rate": 5.54449933349016e-05, "loss": 2.5899, "step": 56821 }, { "epoch": 4.46, "learning_rate": 5.54442092056771e-05, "loss": 2.5925, "step": 56822 }, { "epoch": 4.46, "learning_rate": 5.5443425076452604e-05, "loss": 2.4611, "step": 56823 }, { "epoch": 4.46, "learning_rate": 5.544264094722811e-05, "loss": 2.6442, "step": 56824 }, { "epoch": 4.46, "learning_rate": 5.544185681800361e-05, "loss": 2.8991, "step": 56825 }, { "epoch": 4.46, "learning_rate": 5.5441072688779115e-05, "loss": 2.7628, "step": 56826 }, { "epoch": 4.46, "learning_rate": 5.544028855955462e-05, "loss": 2.7286, "step": 56827 }, { "epoch": 4.46, "learning_rate": 5.543950443033012e-05, "loss": 2.6974, "step": 56828 }, { "epoch": 4.46, "learning_rate": 5.5438720301105627e-05, "loss": 2.6409, "step": 56829 }, { "epoch": 4.46, "learning_rate": 5.543793617188113e-05, "loss": 2.5404, "step": 56830 }, { "epoch": 4.46, "learning_rate": 5.5437152042656634e-05, "loss": 2.8717, "step": 56831 }, { "epoch": 4.46, "learning_rate": 5.5436367913432144e-05, "loss": 2.8695, "step": 56832 }, { "epoch": 4.46, "learning_rate": 5.543558378420765e-05, "loss": 3.18, "step": 56833 }, { "epoch": 4.46, "learning_rate": 5.543479965498315e-05, "loss": 2.6848, "step": 56834 }, { "epoch": 4.46, "learning_rate": 5.543401552575864e-05, "loss": 2.6788, "step": 56835 }, { "epoch": 4.46, "learning_rate": 5.5433231396534145e-05, "loss": 2.4389, "step": 56836 }, { "epoch": 4.46, "learning_rate": 5.543244726730965e-05, "loss": 2.5874, "step": 56837 }, { "epoch": 4.46, "learning_rate": 5.543166313808515e-05, "loss": 2.7179, "step": 56838 }, { "epoch": 4.46, "learning_rate": 5.5430879008860657e-05, "loss": 2.8054, "step": 56839 }, { "epoch": 4.46, "learning_rate": 5.543009487963616e-05, "loss": 2.6098, "step": 56840 }, { "epoch": 4.46, "learning_rate": 5.5429310750411664e-05, "loss": 2.7393, "step": 56841 }, { "epoch": 4.46, "learning_rate": 5.5428526621187174e-05, "loss": 2.5495, "step": 56842 }, { "epoch": 4.46, "learning_rate": 5.542774249196268e-05, "loss": 2.5416, "step": 56843 }, { "epoch": 4.46, "learning_rate": 5.542695836273818e-05, "loss": 2.6926, "step": 56844 }, { "epoch": 4.46, "learning_rate": 5.5426174233513685e-05, "loss": 2.5671, "step": 56845 }, { "epoch": 4.46, "learning_rate": 5.542539010428919e-05, "loss": 2.6009, "step": 56846 }, { "epoch": 4.46, "learning_rate": 5.542460597506469e-05, "loss": 2.6951, "step": 56847 }, { "epoch": 4.46, "learning_rate": 5.5423821845840196e-05, "loss": 2.6862, "step": 56848 }, { "epoch": 4.46, "learning_rate": 5.54230377166157e-05, "loss": 2.8222, "step": 56849 }, { "epoch": 4.46, "learning_rate": 5.5422253587391204e-05, "loss": 2.7759, "step": 56850 }, { "epoch": 4.46, "learning_rate": 5.542146945816671e-05, "loss": 2.5805, "step": 56851 }, { "epoch": 4.46, "learning_rate": 5.542068532894221e-05, "loss": 2.6731, "step": 56852 }, { "epoch": 4.46, "learning_rate": 5.5419901199717715e-05, "loss": 2.5003, "step": 56853 }, { "epoch": 4.46, "learning_rate": 5.541911707049322e-05, "loss": 2.5804, "step": 56854 }, { "epoch": 4.46, "learning_rate": 5.541833294126872e-05, "loss": 2.6979, "step": 56855 }, { "epoch": 4.46, "learning_rate": 5.5417548812044226e-05, "loss": 2.6894, "step": 56856 }, { "epoch": 4.46, "learning_rate": 5.541676468281973e-05, "loss": 2.7401, "step": 56857 }, { "epoch": 4.46, "learning_rate": 5.541598055359524e-05, "loss": 2.7631, "step": 56858 }, { "epoch": 4.46, "learning_rate": 5.5415196424370744e-05, "loss": 2.6392, "step": 56859 }, { "epoch": 4.46, "learning_rate": 5.541441229514625e-05, "loss": 2.577, "step": 56860 }, { "epoch": 4.46, "learning_rate": 5.541362816592175e-05, "loss": 2.6224, "step": 56861 }, { "epoch": 4.46, "learning_rate": 5.5412844036697255e-05, "loss": 2.608, "step": 56862 }, { "epoch": 4.46, "learning_rate": 5.541205990747276e-05, "loss": 2.5203, "step": 56863 }, { "epoch": 4.46, "learning_rate": 5.541127577824826e-05, "loss": 2.6492, "step": 56864 }, { "epoch": 4.46, "learning_rate": 5.5410491649023766e-05, "loss": 2.7414, "step": 56865 }, { "epoch": 4.46, "learning_rate": 5.540970751979927e-05, "loss": 2.5846, "step": 56866 }, { "epoch": 4.46, "learning_rate": 5.540892339057477e-05, "loss": 2.7011, "step": 56867 }, { "epoch": 4.46, "learning_rate": 5.540813926135028e-05, "loss": 2.6798, "step": 56868 }, { "epoch": 4.46, "learning_rate": 5.540735513212578e-05, "loss": 2.86, "step": 56869 }, { "epoch": 4.46, "learning_rate": 5.5406571002901284e-05, "loss": 2.5893, "step": 56870 }, { "epoch": 4.46, "learning_rate": 5.540578687367678e-05, "loss": 2.4584, "step": 56871 }, { "epoch": 4.46, "learning_rate": 5.5405002744452285e-05, "loss": 2.6505, "step": 56872 }, { "epoch": 4.46, "learning_rate": 5.540421861522779e-05, "loss": 2.6919, "step": 56873 }, { "epoch": 4.46, "learning_rate": 5.540343448600329e-05, "loss": 2.6694, "step": 56874 }, { "epoch": 4.46, "learning_rate": 5.5402650356778796e-05, "loss": 2.5557, "step": 56875 }, { "epoch": 4.46, "learning_rate": 5.54018662275543e-05, "loss": 2.8211, "step": 56876 }, { "epoch": 4.46, "learning_rate": 5.54010820983298e-05, "loss": 2.4902, "step": 56877 }, { "epoch": 4.46, "learning_rate": 5.540029796910531e-05, "loss": 2.5716, "step": 56878 }, { "epoch": 4.46, "learning_rate": 5.539951383988081e-05, "loss": 2.6858, "step": 56879 }, { "epoch": 4.46, "learning_rate": 5.5398729710656314e-05, "loss": 2.6855, "step": 56880 }, { "epoch": 4.46, "learning_rate": 5.539794558143182e-05, "loss": 2.7833, "step": 56881 }, { "epoch": 4.46, "learning_rate": 5.539716145220732e-05, "loss": 2.6242, "step": 56882 }, { "epoch": 4.46, "learning_rate": 5.539637732298283e-05, "loss": 2.5383, "step": 56883 }, { "epoch": 4.46, "learning_rate": 5.5395593193758336e-05, "loss": 2.7834, "step": 56884 }, { "epoch": 4.46, "learning_rate": 5.539480906453384e-05, "loss": 2.6066, "step": 56885 }, { "epoch": 4.46, "learning_rate": 5.539402493530934e-05, "loss": 2.6237, "step": 56886 }, { "epoch": 4.46, "learning_rate": 5.5393240806084847e-05, "loss": 2.8531, "step": 56887 }, { "epoch": 4.46, "learning_rate": 5.539245667686035e-05, "loss": 2.8959, "step": 56888 }, { "epoch": 4.46, "learning_rate": 5.5391672547635854e-05, "loss": 2.806, "step": 56889 }, { "epoch": 4.46, "learning_rate": 5.539088841841136e-05, "loss": 2.6725, "step": 56890 }, { "epoch": 4.46, "learning_rate": 5.539010428918686e-05, "loss": 2.6563, "step": 56891 }, { "epoch": 4.46, "learning_rate": 5.5389320159962365e-05, "loss": 2.6767, "step": 56892 }, { "epoch": 4.46, "learning_rate": 5.538853603073787e-05, "loss": 2.6743, "step": 56893 }, { "epoch": 4.46, "learning_rate": 5.538775190151337e-05, "loss": 2.9598, "step": 56894 }, { "epoch": 4.46, "learning_rate": 5.5386967772288876e-05, "loss": 2.7227, "step": 56895 }, { "epoch": 4.46, "learning_rate": 5.538618364306438e-05, "loss": 2.647, "step": 56896 }, { "epoch": 4.46, "learning_rate": 5.538539951383988e-05, "loss": 2.763, "step": 56897 }, { "epoch": 4.46, "learning_rate": 5.538461538461539e-05, "loss": 2.7449, "step": 56898 }, { "epoch": 4.46, "learning_rate": 5.53838312553909e-05, "loss": 2.5215, "step": 56899 }, { "epoch": 4.46, "learning_rate": 5.53830471261664e-05, "loss": 2.6554, "step": 56900 }, { "epoch": 4.46, "learning_rate": 5.5382262996941905e-05, "loss": 2.5804, "step": 56901 }, { "epoch": 4.46, "learning_rate": 5.538147886771741e-05, "loss": 2.7399, "step": 56902 }, { "epoch": 4.46, "learning_rate": 5.538069473849291e-05, "loss": 2.6639, "step": 56903 }, { "epoch": 4.46, "learning_rate": 5.5379910609268416e-05, "loss": 2.717, "step": 56904 }, { "epoch": 4.46, "learning_rate": 5.537912648004392e-05, "loss": 2.5206, "step": 56905 }, { "epoch": 4.46, "learning_rate": 5.537834235081942e-05, "loss": 2.5111, "step": 56906 }, { "epoch": 4.46, "learning_rate": 5.537755822159491e-05, "loss": 2.6265, "step": 56907 }, { "epoch": 4.46, "learning_rate": 5.537677409237042e-05, "loss": 2.5299, "step": 56908 }, { "epoch": 4.46, "learning_rate": 5.537598996314593e-05, "loss": 3.1198, "step": 56909 }, { "epoch": 4.46, "learning_rate": 5.537520583392143e-05, "loss": 3.0685, "step": 56910 }, { "epoch": 4.46, "learning_rate": 5.5374421704696935e-05, "loss": 2.5563, "step": 56911 }, { "epoch": 4.46, "learning_rate": 5.537363757547244e-05, "loss": 2.6261, "step": 56912 }, { "epoch": 4.46, "learning_rate": 5.537285344624794e-05, "loss": 2.6468, "step": 56913 }, { "epoch": 4.46, "learning_rate": 5.5372069317023446e-05, "loss": 2.763, "step": 56914 }, { "epoch": 4.46, "learning_rate": 5.537128518779895e-05, "loss": 2.5788, "step": 56915 }, { "epoch": 4.46, "learning_rate": 5.537050105857445e-05, "loss": 2.5782, "step": 56916 }, { "epoch": 4.46, "learning_rate": 5.536971692934996e-05, "loss": 2.5813, "step": 56917 }, { "epoch": 4.46, "learning_rate": 5.536893280012546e-05, "loss": 2.5345, "step": 56918 }, { "epoch": 4.46, "learning_rate": 5.5368148670900964e-05, "loss": 2.5805, "step": 56919 }, { "epoch": 4.46, "learning_rate": 5.536736454167647e-05, "loss": 2.6331, "step": 56920 }, { "epoch": 4.46, "learning_rate": 5.536658041245197e-05, "loss": 2.7789, "step": 56921 }, { "epoch": 4.46, "learning_rate": 5.5365796283227475e-05, "loss": 2.7854, "step": 56922 }, { "epoch": 4.46, "learning_rate": 5.536501215400298e-05, "loss": 2.6347, "step": 56923 }, { "epoch": 4.46, "learning_rate": 5.536422802477849e-05, "loss": 2.6194, "step": 56924 }, { "epoch": 4.46, "learning_rate": 5.536344389555399e-05, "loss": 2.476, "step": 56925 }, { "epoch": 4.46, "learning_rate": 5.53626597663295e-05, "loss": 2.6827, "step": 56926 }, { "epoch": 4.46, "learning_rate": 5.5361875637105e-05, "loss": 2.7955, "step": 56927 }, { "epoch": 4.46, "learning_rate": 5.5361091507880504e-05, "loss": 2.5808, "step": 56928 }, { "epoch": 4.46, "learning_rate": 5.536030737865601e-05, "loss": 2.6267, "step": 56929 }, { "epoch": 4.46, "learning_rate": 5.535952324943151e-05, "loss": 2.825, "step": 56930 }, { "epoch": 4.46, "learning_rate": 5.5358739120207015e-05, "loss": 2.8826, "step": 56931 }, { "epoch": 4.46, "learning_rate": 5.535795499098252e-05, "loss": 2.8692, "step": 56932 }, { "epoch": 4.46, "learning_rate": 5.535717086175802e-05, "loss": 2.586, "step": 56933 }, { "epoch": 4.46, "learning_rate": 5.5356386732533526e-05, "loss": 2.685, "step": 56934 }, { "epoch": 4.46, "learning_rate": 5.535560260330903e-05, "loss": 2.6946, "step": 56935 }, { "epoch": 4.46, "learning_rate": 5.5354818474084534e-05, "loss": 2.6767, "step": 56936 }, { "epoch": 4.46, "learning_rate": 5.535403434486004e-05, "loss": 2.6303, "step": 56937 }, { "epoch": 4.46, "learning_rate": 5.535325021563554e-05, "loss": 2.7698, "step": 56938 }, { "epoch": 4.46, "learning_rate": 5.5352466086411045e-05, "loss": 2.7439, "step": 56939 }, { "epoch": 4.46, "learning_rate": 5.5351681957186555e-05, "loss": 2.547, "step": 56940 }, { "epoch": 4.46, "learning_rate": 5.535089782796206e-05, "loss": 2.659, "step": 56941 }, { "epoch": 4.46, "learning_rate": 5.535011369873756e-05, "loss": 2.6449, "step": 56942 }, { "epoch": 4.47, "learning_rate": 5.534932956951305e-05, "loss": 2.6454, "step": 56943 }, { "epoch": 4.47, "learning_rate": 5.5348545440288556e-05, "loss": 2.7626, "step": 56944 }, { "epoch": 4.47, "learning_rate": 5.534776131106406e-05, "loss": 2.7763, "step": 56945 }, { "epoch": 4.47, "learning_rate": 5.5346977181839564e-05, "loss": 2.5071, "step": 56946 }, { "epoch": 4.47, "learning_rate": 5.534619305261507e-05, "loss": 2.6599, "step": 56947 }, { "epoch": 4.47, "learning_rate": 5.534540892339057e-05, "loss": 2.5493, "step": 56948 }, { "epoch": 4.47, "learning_rate": 5.5344624794166075e-05, "loss": 2.6427, "step": 56949 }, { "epoch": 4.47, "learning_rate": 5.5343840664941585e-05, "loss": 2.5546, "step": 56950 }, { "epoch": 4.47, "learning_rate": 5.534305653571709e-05, "loss": 2.568, "step": 56951 }, { "epoch": 4.47, "learning_rate": 5.534227240649259e-05, "loss": 2.7944, "step": 56952 }, { "epoch": 4.47, "learning_rate": 5.5341488277268096e-05, "loss": 2.7702, "step": 56953 }, { "epoch": 4.47, "learning_rate": 5.53407041480436e-05, "loss": 2.9463, "step": 56954 }, { "epoch": 4.47, "learning_rate": 5.5339920018819103e-05, "loss": 2.6227, "step": 56955 }, { "epoch": 4.47, "learning_rate": 5.533913588959461e-05, "loss": 2.5962, "step": 56956 }, { "epoch": 4.47, "learning_rate": 5.533835176037011e-05, "loss": 2.6021, "step": 56957 }, { "epoch": 4.47, "learning_rate": 5.5337567631145614e-05, "loss": 2.7169, "step": 56958 }, { "epoch": 4.47, "learning_rate": 5.533678350192112e-05, "loss": 2.6208, "step": 56959 }, { "epoch": 4.47, "learning_rate": 5.533599937269662e-05, "loss": 2.8932, "step": 56960 }, { "epoch": 4.47, "learning_rate": 5.5335215243472126e-05, "loss": 2.6524, "step": 56961 }, { "epoch": 4.47, "learning_rate": 5.533443111424763e-05, "loss": 2.6524, "step": 56962 }, { "epoch": 4.47, "learning_rate": 5.533364698502313e-05, "loss": 2.634, "step": 56963 }, { "epoch": 4.47, "learning_rate": 5.5332862855798637e-05, "loss": 2.6531, "step": 56964 }, { "epoch": 4.47, "learning_rate": 5.533207872657414e-05, "loss": 2.6996, "step": 56965 }, { "epoch": 4.47, "learning_rate": 5.533129459734965e-05, "loss": 2.5637, "step": 56966 }, { "epoch": 4.47, "learning_rate": 5.5330510468125154e-05, "loss": 2.8379, "step": 56967 }, { "epoch": 4.47, "learning_rate": 5.532972633890066e-05, "loss": 2.9032, "step": 56968 }, { "epoch": 4.47, "learning_rate": 5.532894220967616e-05, "loss": 2.6646, "step": 56969 }, { "epoch": 4.47, "learning_rate": 5.5328158080451665e-05, "loss": 2.6984, "step": 56970 }, { "epoch": 4.47, "learning_rate": 5.532737395122717e-05, "loss": 2.5182, "step": 56971 }, { "epoch": 4.47, "learning_rate": 5.532658982200267e-05, "loss": 2.8019, "step": 56972 }, { "epoch": 4.47, "learning_rate": 5.5325805692778176e-05, "loss": 2.6067, "step": 56973 }, { "epoch": 4.47, "learning_rate": 5.532502156355368e-05, "loss": 2.8612, "step": 56974 }, { "epoch": 4.47, "learning_rate": 5.5324237434329184e-05, "loss": 2.7411, "step": 56975 }, { "epoch": 4.47, "learning_rate": 5.532345330510469e-05, "loss": 2.9438, "step": 56976 }, { "epoch": 4.47, "learning_rate": 5.532266917588019e-05, "loss": 2.6603, "step": 56977 }, { "epoch": 4.47, "learning_rate": 5.5321885046655695e-05, "loss": 2.6783, "step": 56978 }, { "epoch": 4.47, "learning_rate": 5.532110091743119e-05, "loss": 2.6056, "step": 56979 }, { "epoch": 4.47, "learning_rate": 5.5320316788206695e-05, "loss": 2.7299, "step": 56980 }, { "epoch": 4.47, "learning_rate": 5.53195326589822e-05, "loss": 2.7228, "step": 56981 }, { "epoch": 4.47, "learning_rate": 5.53187485297577e-05, "loss": 2.7034, "step": 56982 }, { "epoch": 4.47, "learning_rate": 5.5317964400533206e-05, "loss": 3.0658, "step": 56983 }, { "epoch": 4.47, "learning_rate": 5.531718027130871e-05, "loss": 2.5419, "step": 56984 }, { "epoch": 4.47, "learning_rate": 5.5316396142084214e-05, "loss": 2.7079, "step": 56985 }, { "epoch": 4.47, "learning_rate": 5.531561201285972e-05, "loss": 2.3708, "step": 56986 }, { "epoch": 4.47, "learning_rate": 5.531482788363522e-05, "loss": 2.5312, "step": 56987 }, { "epoch": 4.47, "learning_rate": 5.5314043754410725e-05, "loss": 2.662, "step": 56988 }, { "epoch": 4.47, "learning_rate": 5.531325962518623e-05, "loss": 2.6137, "step": 56989 }, { "epoch": 4.47, "learning_rate": 5.531247549596173e-05, "loss": 2.4353, "step": 56990 }, { "epoch": 4.47, "learning_rate": 5.531169136673724e-05, "loss": 2.6223, "step": 56991 }, { "epoch": 4.47, "learning_rate": 5.5310907237512746e-05, "loss": 2.754, "step": 56992 }, { "epoch": 4.47, "learning_rate": 5.531012310828825e-05, "loss": 2.6604, "step": 56993 }, { "epoch": 4.47, "learning_rate": 5.5309338979063754e-05, "loss": 2.7623, "step": 56994 }, { "epoch": 4.47, "learning_rate": 5.530855484983926e-05, "loss": 2.5499, "step": 56995 }, { "epoch": 4.47, "learning_rate": 5.530777072061476e-05, "loss": 2.5536, "step": 56996 }, { "epoch": 4.47, "learning_rate": 5.5306986591390265e-05, "loss": 2.5267, "step": 56997 }, { "epoch": 4.47, "learning_rate": 5.530620246216577e-05, "loss": 2.7499, "step": 56998 }, { "epoch": 4.47, "learning_rate": 5.530541833294127e-05, "loss": 2.5424, "step": 56999 }, { "epoch": 4.47, "learning_rate": 5.5304634203716776e-05, "loss": 2.613, "step": 57000 }, { "epoch": 4.47, "eval_loss": 2.8975040912628174, "eval_runtime": 217.7863, "eval_samples_per_second": 52.033, "eval_steps_per_second": 3.255, "step": 57000 }, { "epoch": 4.47, "learning_rate": 5.530385007449228e-05, "loss": 2.5985, "step": 57001 }, { "epoch": 4.47, "learning_rate": 5.530306594526778e-05, "loss": 2.4976, "step": 57002 }, { "epoch": 4.47, "learning_rate": 5.530228181604329e-05, "loss": 2.669, "step": 57003 }, { "epoch": 4.47, "learning_rate": 5.530149768681879e-05, "loss": 2.6109, "step": 57004 }, { "epoch": 4.47, "learning_rate": 5.5300713557594294e-05, "loss": 2.7429, "step": 57005 }, { "epoch": 4.47, "learning_rate": 5.52999294283698e-05, "loss": 2.5339, "step": 57006 }, { "epoch": 4.47, "learning_rate": 5.529914529914531e-05, "loss": 2.794, "step": 57007 }, { "epoch": 4.47, "learning_rate": 5.529836116992081e-05, "loss": 2.6457, "step": 57008 }, { "epoch": 4.47, "learning_rate": 5.5297577040696316e-05, "loss": 2.8624, "step": 57009 }, { "epoch": 4.47, "learning_rate": 5.529679291147182e-05, "loss": 2.7128, "step": 57010 }, { "epoch": 4.47, "learning_rate": 5.529600878224732e-05, "loss": 2.9255, "step": 57011 }, { "epoch": 4.47, "learning_rate": 5.529522465302283e-05, "loss": 2.6428, "step": 57012 }, { "epoch": 4.47, "learning_rate": 5.529444052379833e-05, "loss": 2.7343, "step": 57013 }, { "epoch": 4.47, "learning_rate": 5.5293656394573834e-05, "loss": 2.4127, "step": 57014 }, { "epoch": 4.47, "learning_rate": 5.5292872265349324e-05, "loss": 2.5678, "step": 57015 }, { "epoch": 4.47, "learning_rate": 5.529208813612483e-05, "loss": 2.7463, "step": 57016 }, { "epoch": 4.47, "learning_rate": 5.529130400690034e-05, "loss": 2.7658, "step": 57017 }, { "epoch": 4.47, "learning_rate": 5.529051987767584e-05, "loss": 2.6272, "step": 57018 }, { "epoch": 4.47, "learning_rate": 5.5289735748451346e-05, "loss": 2.8408, "step": 57019 }, { "epoch": 4.47, "learning_rate": 5.528895161922685e-05, "loss": 2.7102, "step": 57020 }, { "epoch": 4.47, "learning_rate": 5.528816749000235e-05, "loss": 2.6994, "step": 57021 }, { "epoch": 4.47, "learning_rate": 5.528738336077786e-05, "loss": 2.6278, "step": 57022 }, { "epoch": 4.47, "learning_rate": 5.528659923155336e-05, "loss": 2.7596, "step": 57023 }, { "epoch": 4.47, "learning_rate": 5.5285815102328864e-05, "loss": 2.7381, "step": 57024 }, { "epoch": 4.47, "learning_rate": 5.528503097310437e-05, "loss": 2.7931, "step": 57025 }, { "epoch": 4.47, "learning_rate": 5.528424684387987e-05, "loss": 2.6914, "step": 57026 }, { "epoch": 4.47, "learning_rate": 5.5283462714655375e-05, "loss": 2.5995, "step": 57027 }, { "epoch": 4.47, "learning_rate": 5.528267858543088e-05, "loss": 2.5561, "step": 57028 }, { "epoch": 4.47, "learning_rate": 5.528189445620638e-05, "loss": 2.5973, "step": 57029 }, { "epoch": 4.47, "learning_rate": 5.5281110326981886e-05, "loss": 2.6696, "step": 57030 }, { "epoch": 4.47, "learning_rate": 5.528032619775739e-05, "loss": 2.6126, "step": 57031 }, { "epoch": 4.47, "learning_rate": 5.52795420685329e-05, "loss": 2.6772, "step": 57032 }, { "epoch": 4.47, "learning_rate": 5.5278757939308404e-05, "loss": 2.7378, "step": 57033 }, { "epoch": 4.47, "learning_rate": 5.527797381008391e-05, "loss": 2.649, "step": 57034 }, { "epoch": 4.47, "learning_rate": 5.527718968085941e-05, "loss": 2.7621, "step": 57035 }, { "epoch": 4.47, "learning_rate": 5.5276405551634915e-05, "loss": 2.6189, "step": 57036 }, { "epoch": 4.47, "learning_rate": 5.527562142241042e-05, "loss": 2.4955, "step": 57037 }, { "epoch": 4.47, "learning_rate": 5.527483729318592e-05, "loss": 2.8203, "step": 57038 }, { "epoch": 4.47, "learning_rate": 5.5274053163961426e-05, "loss": 2.727, "step": 57039 }, { "epoch": 4.47, "learning_rate": 5.527326903473693e-05, "loss": 2.4492, "step": 57040 }, { "epoch": 4.47, "learning_rate": 5.527248490551243e-05, "loss": 2.6112, "step": 57041 }, { "epoch": 4.47, "learning_rate": 5.527170077628794e-05, "loss": 2.7455, "step": 57042 }, { "epoch": 4.47, "learning_rate": 5.527091664706344e-05, "loss": 2.5912, "step": 57043 }, { "epoch": 4.47, "learning_rate": 5.5270132517838944e-05, "loss": 2.519, "step": 57044 }, { "epoch": 4.47, "learning_rate": 5.526934838861445e-05, "loss": 2.544, "step": 57045 }, { "epoch": 4.47, "learning_rate": 5.526856425938995e-05, "loss": 2.6822, "step": 57046 }, { "epoch": 4.47, "learning_rate": 5.5267780130165455e-05, "loss": 2.8711, "step": 57047 }, { "epoch": 4.47, "learning_rate": 5.5266996000940966e-05, "loss": 2.5815, "step": 57048 }, { "epoch": 4.47, "learning_rate": 5.526621187171647e-05, "loss": 2.618, "step": 57049 }, { "epoch": 4.47, "learning_rate": 5.526542774249197e-05, "loss": 2.6181, "step": 57050 }, { "epoch": 4.47, "learning_rate": 5.526464361326746e-05, "loss": 2.5035, "step": 57051 }, { "epoch": 4.47, "learning_rate": 5.526385948404297e-05, "loss": 2.4045, "step": 57052 }, { "epoch": 4.47, "learning_rate": 5.526307535481847e-05, "loss": 2.6065, "step": 57053 }, { "epoch": 4.47, "learning_rate": 5.5262291225593974e-05, "loss": 2.7368, "step": 57054 }, { "epoch": 4.47, "learning_rate": 5.526150709636948e-05, "loss": 2.532, "step": 57055 }, { "epoch": 4.47, "learning_rate": 5.526072296714498e-05, "loss": 2.7034, "step": 57056 }, { "epoch": 4.47, "learning_rate": 5.5259938837920485e-05, "loss": 2.8859, "step": 57057 }, { "epoch": 4.47, "learning_rate": 5.5259154708695996e-05, "loss": 2.653, "step": 57058 }, { "epoch": 4.47, "learning_rate": 5.52583705794715e-05, "loss": 2.7201, "step": 57059 }, { "epoch": 4.47, "learning_rate": 5.5257586450247e-05, "loss": 2.5952, "step": 57060 }, { "epoch": 4.47, "learning_rate": 5.525680232102251e-05, "loss": 2.4222, "step": 57061 }, { "epoch": 4.47, "learning_rate": 5.525601819179801e-05, "loss": 2.5091, "step": 57062 }, { "epoch": 4.47, "learning_rate": 5.5255234062573514e-05, "loss": 2.791, "step": 57063 }, { "epoch": 4.47, "learning_rate": 5.525444993334902e-05, "loss": 2.4691, "step": 57064 }, { "epoch": 4.47, "learning_rate": 5.525366580412452e-05, "loss": 2.636, "step": 57065 }, { "epoch": 4.47, "learning_rate": 5.5252881674900025e-05, "loss": 2.7919, "step": 57066 }, { "epoch": 4.47, "learning_rate": 5.525209754567553e-05, "loss": 2.5174, "step": 57067 }, { "epoch": 4.47, "learning_rate": 5.525131341645103e-05, "loss": 2.7586, "step": 57068 }, { "epoch": 4.47, "learning_rate": 5.5250529287226536e-05, "loss": 2.5561, "step": 57069 }, { "epoch": 4.48, "learning_rate": 5.524974515800204e-05, "loss": 2.5043, "step": 57070 }, { "epoch": 4.48, "learning_rate": 5.5248961028777544e-05, "loss": 2.6455, "step": 57071 }, { "epoch": 4.48, "learning_rate": 5.524817689955305e-05, "loss": 2.5604, "step": 57072 }, { "epoch": 4.48, "learning_rate": 5.524739277032855e-05, "loss": 2.8549, "step": 57073 }, { "epoch": 4.48, "learning_rate": 5.524660864110406e-05, "loss": 2.5567, "step": 57074 }, { "epoch": 4.48, "learning_rate": 5.5245824511879565e-05, "loss": 2.8065, "step": 57075 }, { "epoch": 4.48, "learning_rate": 5.524504038265507e-05, "loss": 2.7622, "step": 57076 }, { "epoch": 4.48, "learning_rate": 5.524425625343057e-05, "loss": 2.6636, "step": 57077 }, { "epoch": 4.48, "learning_rate": 5.5243472124206076e-05, "loss": 2.5522, "step": 57078 }, { "epoch": 4.48, "learning_rate": 5.524268799498158e-05, "loss": 2.7112, "step": 57079 }, { "epoch": 4.48, "learning_rate": 5.5241903865757084e-05, "loss": 2.7885, "step": 57080 }, { "epoch": 4.48, "learning_rate": 5.524111973653259e-05, "loss": 2.5956, "step": 57081 }, { "epoch": 4.48, "learning_rate": 5.524033560730809e-05, "loss": 2.7045, "step": 57082 }, { "epoch": 4.48, "learning_rate": 5.5239551478083595e-05, "loss": 2.6487, "step": 57083 }, { "epoch": 4.48, "learning_rate": 5.52387673488591e-05, "loss": 2.5603, "step": 57084 }, { "epoch": 4.48, "learning_rate": 5.52379832196346e-05, "loss": 2.6837, "step": 57085 }, { "epoch": 4.48, "learning_rate": 5.5237199090410106e-05, "loss": 2.4388, "step": 57086 }, { "epoch": 4.48, "learning_rate": 5.52364149611856e-05, "loss": 2.5777, "step": 57087 }, { "epoch": 4.48, "learning_rate": 5.5235630831961106e-05, "loss": 2.6316, "step": 57088 }, { "epoch": 4.48, "learning_rate": 5.523484670273661e-05, "loss": 2.6487, "step": 57089 }, { "epoch": 4.48, "learning_rate": 5.5234062573512114e-05, "loss": 2.7474, "step": 57090 }, { "epoch": 4.48, "learning_rate": 5.523327844428762e-05, "loss": 2.6099, "step": 57091 }, { "epoch": 4.48, "learning_rate": 5.523249431506312e-05, "loss": 2.5317, "step": 57092 }, { "epoch": 4.48, "learning_rate": 5.5231710185838625e-05, "loss": 2.7523, "step": 57093 }, { "epoch": 4.48, "learning_rate": 5.523092605661413e-05, "loss": 2.8152, "step": 57094 }, { "epoch": 4.48, "learning_rate": 5.523014192738963e-05, "loss": 2.5808, "step": 57095 }, { "epoch": 4.48, "learning_rate": 5.5229357798165136e-05, "loss": 2.6493, "step": 57096 }, { "epoch": 4.48, "learning_rate": 5.522857366894064e-05, "loss": 2.5935, "step": 57097 }, { "epoch": 4.48, "learning_rate": 5.522778953971614e-05, "loss": 2.7327, "step": 57098 }, { "epoch": 4.48, "learning_rate": 5.522700541049165e-05, "loss": 2.8764, "step": 57099 }, { "epoch": 4.48, "learning_rate": 5.522622128126716e-05, "loss": 2.6274, "step": 57100 }, { "epoch": 4.48, "learning_rate": 5.522543715204266e-05, "loss": 2.9676, "step": 57101 }, { "epoch": 4.48, "learning_rate": 5.5224653022818164e-05, "loss": 2.7788, "step": 57102 }, { "epoch": 4.48, "learning_rate": 5.522386889359367e-05, "loss": 2.749, "step": 57103 }, { "epoch": 4.48, "learning_rate": 5.522308476436917e-05, "loss": 2.6996, "step": 57104 }, { "epoch": 4.48, "learning_rate": 5.5222300635144675e-05, "loss": 2.641, "step": 57105 }, { "epoch": 4.48, "learning_rate": 5.522151650592018e-05, "loss": 2.6755, "step": 57106 }, { "epoch": 4.48, "learning_rate": 5.522073237669568e-05, "loss": 2.9414, "step": 57107 }, { "epoch": 4.48, "learning_rate": 5.5219948247471186e-05, "loss": 2.7922, "step": 57108 }, { "epoch": 4.48, "learning_rate": 5.521916411824669e-05, "loss": 2.6978, "step": 57109 }, { "epoch": 4.48, "learning_rate": 5.5218379989022194e-05, "loss": 2.6725, "step": 57110 }, { "epoch": 4.48, "learning_rate": 5.52175958597977e-05, "loss": 2.4398, "step": 57111 }, { "epoch": 4.48, "learning_rate": 5.52168117305732e-05, "loss": 2.6779, "step": 57112 }, { "epoch": 4.48, "learning_rate": 5.5216027601348705e-05, "loss": 2.6512, "step": 57113 }, { "epoch": 4.48, "learning_rate": 5.521524347212421e-05, "loss": 2.677, "step": 57114 }, { "epoch": 4.48, "learning_rate": 5.521445934289972e-05, "loss": 2.7021, "step": 57115 }, { "epoch": 4.48, "learning_rate": 5.521367521367522e-05, "loss": 2.8031, "step": 57116 }, { "epoch": 4.48, "learning_rate": 5.5212891084450726e-05, "loss": 2.5427, "step": 57117 }, { "epoch": 4.48, "learning_rate": 5.521210695522623e-05, "loss": 2.6111, "step": 57118 }, { "epoch": 4.48, "learning_rate": 5.5211322826001734e-05, "loss": 2.5611, "step": 57119 }, { "epoch": 4.48, "learning_rate": 5.521053869677724e-05, "loss": 2.7866, "step": 57120 }, { "epoch": 4.48, "learning_rate": 5.520975456755274e-05, "loss": 2.645, "step": 57121 }, { "epoch": 4.48, "learning_rate": 5.5208970438328245e-05, "loss": 2.585, "step": 57122 }, { "epoch": 4.48, "learning_rate": 5.5208186309103735e-05, "loss": 2.571, "step": 57123 }, { "epoch": 4.48, "learning_rate": 5.520740217987924e-05, "loss": 2.5815, "step": 57124 }, { "epoch": 4.48, "learning_rate": 5.520661805065475e-05, "loss": 2.6819, "step": 57125 }, { "epoch": 4.48, "learning_rate": 5.520583392143025e-05, "loss": 2.5684, "step": 57126 }, { "epoch": 4.48, "learning_rate": 5.5205049792205756e-05, "loss": 2.7011, "step": 57127 }, { "epoch": 4.48, "learning_rate": 5.520426566298126e-05, "loss": 2.644, "step": 57128 }, { "epoch": 4.48, "learning_rate": 5.5203481533756764e-05, "loss": 2.5883, "step": 57129 }, { "epoch": 4.48, "learning_rate": 5.520269740453227e-05, "loss": 2.5636, "step": 57130 }, { "epoch": 4.48, "learning_rate": 5.520191327530777e-05, "loss": 2.6607, "step": 57131 }, { "epoch": 4.48, "learning_rate": 5.5201129146083275e-05, "loss": 2.6459, "step": 57132 }, { "epoch": 4.48, "learning_rate": 5.520034501685878e-05, "loss": 2.984, "step": 57133 }, { "epoch": 4.48, "learning_rate": 5.519956088763428e-05, "loss": 2.5119, "step": 57134 }, { "epoch": 4.48, "learning_rate": 5.5198776758409786e-05, "loss": 2.4091, "step": 57135 }, { "epoch": 4.48, "learning_rate": 5.519799262918529e-05, "loss": 2.5991, "step": 57136 }, { "epoch": 4.48, "learning_rate": 5.519720849996079e-05, "loss": 2.6212, "step": 57137 }, { "epoch": 4.48, "learning_rate": 5.51964243707363e-05, "loss": 2.8988, "step": 57138 }, { "epoch": 4.48, "learning_rate": 5.51956402415118e-05, "loss": 2.6528, "step": 57139 }, { "epoch": 4.48, "learning_rate": 5.519485611228731e-05, "loss": 2.5965, "step": 57140 }, { "epoch": 4.48, "learning_rate": 5.5194071983062815e-05, "loss": 2.7543, "step": 57141 }, { "epoch": 4.48, "learning_rate": 5.519328785383832e-05, "loss": 2.7347, "step": 57142 }, { "epoch": 4.48, "learning_rate": 5.519250372461382e-05, "loss": 2.7797, "step": 57143 }, { "epoch": 4.48, "learning_rate": 5.5191719595389326e-05, "loss": 2.7216, "step": 57144 }, { "epoch": 4.48, "learning_rate": 5.519093546616483e-05, "loss": 2.556, "step": 57145 }, { "epoch": 4.48, "learning_rate": 5.519015133694033e-05, "loss": 2.8062, "step": 57146 }, { "epoch": 4.48, "learning_rate": 5.518936720771584e-05, "loss": 2.6711, "step": 57147 }, { "epoch": 4.48, "learning_rate": 5.518858307849134e-05, "loss": 2.6782, "step": 57148 }, { "epoch": 4.48, "learning_rate": 5.5187798949266844e-05, "loss": 2.777, "step": 57149 }, { "epoch": 4.48, "learning_rate": 5.518701482004235e-05, "loss": 2.8458, "step": 57150 }, { "epoch": 4.48, "learning_rate": 5.518623069081785e-05, "loss": 2.7972, "step": 57151 }, { "epoch": 4.48, "learning_rate": 5.5185446561593355e-05, "loss": 2.7284, "step": 57152 }, { "epoch": 4.48, "learning_rate": 5.518466243236886e-05, "loss": 2.699, "step": 57153 }, { "epoch": 4.48, "learning_rate": 5.518387830314436e-05, "loss": 2.9532, "step": 57154 }, { "epoch": 4.48, "learning_rate": 5.5183094173919866e-05, "loss": 2.539, "step": 57155 }, { "epoch": 4.48, "learning_rate": 5.5182310044695377e-05, "loss": 2.7215, "step": 57156 }, { "epoch": 4.48, "learning_rate": 5.518152591547088e-05, "loss": 2.6047, "step": 57157 }, { "epoch": 4.48, "learning_rate": 5.5180741786246384e-05, "loss": 2.6038, "step": 57158 }, { "epoch": 4.48, "learning_rate": 5.5179957657021874e-05, "loss": 2.7502, "step": 57159 }, { "epoch": 4.48, "learning_rate": 5.517917352779738e-05, "loss": 2.7082, "step": 57160 }, { "epoch": 4.48, "learning_rate": 5.517838939857288e-05, "loss": 2.7282, "step": 57161 }, { "epoch": 4.48, "learning_rate": 5.5177605269348385e-05, "loss": 2.5, "step": 57162 }, { "epoch": 4.48, "learning_rate": 5.517682114012389e-05, "loss": 2.5078, "step": 57163 }, { "epoch": 4.48, "learning_rate": 5.517603701089939e-05, "loss": 2.8436, "step": 57164 }, { "epoch": 4.48, "learning_rate": 5.5175252881674896e-05, "loss": 2.7769, "step": 57165 }, { "epoch": 4.48, "learning_rate": 5.5174468752450407e-05, "loss": 2.5972, "step": 57166 }, { "epoch": 4.48, "learning_rate": 5.517368462322591e-05, "loss": 2.7115, "step": 57167 }, { "epoch": 4.48, "learning_rate": 5.5172900494001414e-05, "loss": 2.752, "step": 57168 }, { "epoch": 4.48, "learning_rate": 5.517211636477692e-05, "loss": 2.5854, "step": 57169 }, { "epoch": 4.48, "learning_rate": 5.517133223555242e-05, "loss": 2.7049, "step": 57170 }, { "epoch": 4.48, "learning_rate": 5.5170548106327925e-05, "loss": 2.5799, "step": 57171 }, { "epoch": 4.48, "learning_rate": 5.516976397710343e-05, "loss": 2.664, "step": 57172 }, { "epoch": 4.48, "learning_rate": 5.516897984787893e-05, "loss": 3.0202, "step": 57173 }, { "epoch": 4.48, "learning_rate": 5.5168195718654436e-05, "loss": 2.5959, "step": 57174 }, { "epoch": 4.48, "learning_rate": 5.516741158942994e-05, "loss": 2.5708, "step": 57175 }, { "epoch": 4.48, "learning_rate": 5.516662746020544e-05, "loss": 2.5175, "step": 57176 }, { "epoch": 4.48, "learning_rate": 5.516584333098095e-05, "loss": 2.7731, "step": 57177 }, { "epoch": 4.48, "learning_rate": 5.516505920175645e-05, "loss": 2.6553, "step": 57178 }, { "epoch": 4.48, "learning_rate": 5.5164275072531954e-05, "loss": 2.6577, "step": 57179 }, { "epoch": 4.48, "learning_rate": 5.516349094330746e-05, "loss": 2.5652, "step": 57180 }, { "epoch": 4.48, "learning_rate": 5.516270681408296e-05, "loss": 2.6679, "step": 57181 }, { "epoch": 4.48, "learning_rate": 5.516192268485847e-05, "loss": 2.7465, "step": 57182 }, { "epoch": 4.48, "learning_rate": 5.5161138555633976e-05, "loss": 2.6494, "step": 57183 }, { "epoch": 4.48, "learning_rate": 5.516035442640948e-05, "loss": 2.8068, "step": 57184 }, { "epoch": 4.48, "learning_rate": 5.515957029718498e-05, "loss": 2.7484, "step": 57185 }, { "epoch": 4.48, "learning_rate": 5.515878616796049e-05, "loss": 2.828, "step": 57186 }, { "epoch": 4.48, "learning_rate": 5.515800203873599e-05, "loss": 2.6657, "step": 57187 }, { "epoch": 4.48, "learning_rate": 5.5157217909511494e-05, "loss": 2.4859, "step": 57188 }, { "epoch": 4.48, "learning_rate": 5.5156433780287e-05, "loss": 2.9595, "step": 57189 }, { "epoch": 4.48, "learning_rate": 5.51556496510625e-05, "loss": 2.5357, "step": 57190 }, { "epoch": 4.48, "learning_rate": 5.5154865521838005e-05, "loss": 2.7358, "step": 57191 }, { "epoch": 4.48, "learning_rate": 5.515408139261351e-05, "loss": 2.514, "step": 57192 }, { "epoch": 4.48, "learning_rate": 5.515329726338901e-05, "loss": 2.5446, "step": 57193 }, { "epoch": 4.48, "learning_rate": 5.5152513134164516e-05, "loss": 2.6567, "step": 57194 }, { "epoch": 4.48, "learning_rate": 5.515172900494001e-05, "loss": 2.5416, "step": 57195 }, { "epoch": 4.48, "learning_rate": 5.515094487571552e-05, "loss": 2.6586, "step": 57196 }, { "epoch": 4.48, "learning_rate": 5.515016074649102e-05, "loss": 2.7184, "step": 57197 }, { "epoch": 4.49, "learning_rate": 5.5149376617266524e-05, "loss": 2.5537, "step": 57198 }, { "epoch": 4.49, "learning_rate": 5.514859248804203e-05, "loss": 2.6456, "step": 57199 }, { "epoch": 4.49, "learning_rate": 5.514780835881753e-05, "loss": 2.7585, "step": 57200 }, { "epoch": 4.49, "learning_rate": 5.5147024229593035e-05, "loss": 2.897, "step": 57201 }, { "epoch": 4.49, "learning_rate": 5.514624010036854e-05, "loss": 2.7388, "step": 57202 }, { "epoch": 4.49, "learning_rate": 5.514545597114404e-05, "loss": 2.5796, "step": 57203 }, { "epoch": 4.49, "learning_rate": 5.5144671841919546e-05, "loss": 2.803, "step": 57204 }, { "epoch": 4.49, "learning_rate": 5.514388771269505e-05, "loss": 2.6018, "step": 57205 }, { "epoch": 4.49, "learning_rate": 5.5143103583470554e-05, "loss": 2.6583, "step": 57206 }, { "epoch": 4.49, "learning_rate": 5.5142319454246064e-05, "loss": 2.6482, "step": 57207 }, { "epoch": 4.49, "learning_rate": 5.514153532502157e-05, "loss": 2.7187, "step": 57208 }, { "epoch": 4.49, "learning_rate": 5.514075119579707e-05, "loss": 2.5723, "step": 57209 }, { "epoch": 4.49, "learning_rate": 5.5139967066572575e-05, "loss": 2.808, "step": 57210 }, { "epoch": 4.49, "learning_rate": 5.513918293734808e-05, "loss": 2.6382, "step": 57211 }, { "epoch": 4.49, "learning_rate": 5.513839880812358e-05, "loss": 2.7922, "step": 57212 }, { "epoch": 4.49, "learning_rate": 5.5137614678899086e-05, "loss": 2.7367, "step": 57213 }, { "epoch": 4.49, "learning_rate": 5.513683054967459e-05, "loss": 2.6809, "step": 57214 }, { "epoch": 4.49, "learning_rate": 5.5136046420450094e-05, "loss": 2.5862, "step": 57215 }, { "epoch": 4.49, "learning_rate": 5.51352622912256e-05, "loss": 2.8602, "step": 57216 }, { "epoch": 4.49, "learning_rate": 5.51344781620011e-05, "loss": 2.5045, "step": 57217 }, { "epoch": 4.49, "learning_rate": 5.5133694032776605e-05, "loss": 2.65, "step": 57218 }, { "epoch": 4.49, "learning_rate": 5.513290990355211e-05, "loss": 2.7154, "step": 57219 }, { "epoch": 4.49, "learning_rate": 5.513212577432761e-05, "loss": 2.8015, "step": 57220 }, { "epoch": 4.49, "learning_rate": 5.5131341645103116e-05, "loss": 2.8211, "step": 57221 }, { "epoch": 4.49, "learning_rate": 5.513055751587862e-05, "loss": 2.7175, "step": 57222 }, { "epoch": 4.49, "learning_rate": 5.512977338665413e-05, "loss": 2.7752, "step": 57223 }, { "epoch": 4.49, "learning_rate": 5.5128989257429633e-05, "loss": 2.8611, "step": 57224 }, { "epoch": 4.49, "learning_rate": 5.512820512820514e-05, "loss": 2.6148, "step": 57225 }, { "epoch": 4.49, "learning_rate": 5.512742099898064e-05, "loss": 2.5832, "step": 57226 }, { "epoch": 4.49, "learning_rate": 5.5126636869756144e-05, "loss": 2.8825, "step": 57227 }, { "epoch": 4.49, "learning_rate": 5.512585274053165e-05, "loss": 2.4721, "step": 57228 }, { "epoch": 4.49, "learning_rate": 5.512506861130715e-05, "loss": 2.653, "step": 57229 }, { "epoch": 4.49, "learning_rate": 5.5124284482082655e-05, "loss": 2.7381, "step": 57230 }, { "epoch": 4.49, "learning_rate": 5.5123500352858146e-05, "loss": 2.7773, "step": 57231 }, { "epoch": 4.49, "learning_rate": 5.512271622363365e-05, "loss": 2.9514, "step": 57232 }, { "epoch": 4.49, "learning_rate": 5.512193209440916e-05, "loss": 2.8865, "step": 57233 }, { "epoch": 4.49, "learning_rate": 5.5121147965184663e-05, "loss": 2.6581, "step": 57234 }, { "epoch": 4.49, "learning_rate": 5.512036383596017e-05, "loss": 2.896, "step": 57235 }, { "epoch": 4.49, "learning_rate": 5.511957970673567e-05, "loss": 2.6656, "step": 57236 }, { "epoch": 4.49, "learning_rate": 5.5118795577511174e-05, "loss": 2.7872, "step": 57237 }, { "epoch": 4.49, "learning_rate": 5.511801144828668e-05, "loss": 2.7324, "step": 57238 }, { "epoch": 4.49, "learning_rate": 5.511722731906218e-05, "loss": 2.8906, "step": 57239 }, { "epoch": 4.49, "learning_rate": 5.5116443189837685e-05, "loss": 2.5817, "step": 57240 }, { "epoch": 4.49, "learning_rate": 5.511565906061319e-05, "loss": 2.8105, "step": 57241 }, { "epoch": 4.49, "learning_rate": 5.511487493138869e-05, "loss": 2.7535, "step": 57242 }, { "epoch": 4.49, "learning_rate": 5.5114090802164197e-05, "loss": 2.7419, "step": 57243 }, { "epoch": 4.49, "learning_rate": 5.51133066729397e-05, "loss": 2.6359, "step": 57244 }, { "epoch": 4.49, "learning_rate": 5.5112522543715204e-05, "loss": 2.6955, "step": 57245 }, { "epoch": 4.49, "learning_rate": 5.511173841449071e-05, "loss": 2.6493, "step": 57246 }, { "epoch": 4.49, "learning_rate": 5.511095428526621e-05, "loss": 2.4956, "step": 57247 }, { "epoch": 4.49, "learning_rate": 5.5110170156041715e-05, "loss": 2.7637, "step": 57248 }, { "epoch": 4.49, "learning_rate": 5.5109386026817225e-05, "loss": 2.6466, "step": 57249 }, { "epoch": 4.49, "learning_rate": 5.510860189759273e-05, "loss": 2.8233, "step": 57250 }, { "epoch": 4.49, "learning_rate": 5.510781776836823e-05, "loss": 2.7242, "step": 57251 }, { "epoch": 4.49, "learning_rate": 5.5107033639143736e-05, "loss": 2.5477, "step": 57252 }, { "epoch": 4.49, "learning_rate": 5.510624950991924e-05, "loss": 2.6167, "step": 57253 }, { "epoch": 4.49, "learning_rate": 5.5105465380694744e-05, "loss": 2.4899, "step": 57254 }, { "epoch": 4.49, "learning_rate": 5.510468125147025e-05, "loss": 2.6353, "step": 57255 }, { "epoch": 4.49, "learning_rate": 5.510389712224575e-05, "loss": 2.5418, "step": 57256 }, { "epoch": 4.49, "learning_rate": 5.5103112993021255e-05, "loss": 2.8461, "step": 57257 }, { "epoch": 4.49, "learning_rate": 5.510232886379676e-05, "loss": 2.7748, "step": 57258 }, { "epoch": 4.49, "learning_rate": 5.510154473457226e-05, "loss": 2.7778, "step": 57259 }, { "epoch": 4.49, "learning_rate": 5.5100760605347766e-05, "loss": 2.6821, "step": 57260 }, { "epoch": 4.49, "learning_rate": 5.509997647612327e-05, "loss": 2.6949, "step": 57261 }, { "epoch": 4.49, "learning_rate": 5.509919234689877e-05, "loss": 2.8701, "step": 57262 }, { "epoch": 4.49, "learning_rate": 5.509840821767428e-05, "loss": 2.5895, "step": 57263 }, { "epoch": 4.49, "learning_rate": 5.509762408844979e-05, "loss": 2.5848, "step": 57264 }, { "epoch": 4.49, "learning_rate": 5.509683995922529e-05, "loss": 2.6377, "step": 57265 }, { "epoch": 4.49, "learning_rate": 5.5096055830000795e-05, "loss": 2.5353, "step": 57266 }, { "epoch": 4.49, "learning_rate": 5.5095271700776285e-05, "loss": 2.5928, "step": 57267 }, { "epoch": 4.49, "learning_rate": 5.509448757155179e-05, "loss": 2.6779, "step": 57268 }, { "epoch": 4.49, "learning_rate": 5.509370344232729e-05, "loss": 2.7675, "step": 57269 }, { "epoch": 4.49, "learning_rate": 5.5092919313102796e-05, "loss": 2.5137, "step": 57270 }, { "epoch": 4.49, "learning_rate": 5.50921351838783e-05, "loss": 2.5934, "step": 57271 }, { "epoch": 4.49, "learning_rate": 5.50913510546538e-05, "loss": 2.5541, "step": 57272 }, { "epoch": 4.49, "learning_rate": 5.509056692542931e-05, "loss": 2.8647, "step": 57273 }, { "epoch": 4.49, "learning_rate": 5.508978279620482e-05, "loss": 2.5318, "step": 57274 }, { "epoch": 4.49, "learning_rate": 5.508899866698032e-05, "loss": 2.5616, "step": 57275 }, { "epoch": 4.49, "learning_rate": 5.5088214537755825e-05, "loss": 2.638, "step": 57276 }, { "epoch": 4.49, "learning_rate": 5.508743040853133e-05, "loss": 2.6406, "step": 57277 }, { "epoch": 4.49, "learning_rate": 5.508664627930683e-05, "loss": 2.6226, "step": 57278 }, { "epoch": 4.49, "learning_rate": 5.5085862150082336e-05, "loss": 2.6454, "step": 57279 }, { "epoch": 4.49, "learning_rate": 5.508507802085784e-05, "loss": 2.7255, "step": 57280 }, { "epoch": 4.49, "learning_rate": 5.508429389163334e-05, "loss": 2.6214, "step": 57281 }, { "epoch": 4.49, "learning_rate": 5.508350976240885e-05, "loss": 2.6512, "step": 57282 }, { "epoch": 4.49, "learning_rate": 5.508272563318435e-05, "loss": 2.5429, "step": 57283 }, { "epoch": 4.49, "learning_rate": 5.5081941503959854e-05, "loss": 2.5188, "step": 57284 }, { "epoch": 4.49, "learning_rate": 5.508115737473536e-05, "loss": 2.6305, "step": 57285 }, { "epoch": 4.49, "learning_rate": 5.508037324551086e-05, "loss": 2.6657, "step": 57286 }, { "epoch": 4.49, "learning_rate": 5.5079589116286365e-05, "loss": 2.8641, "step": 57287 }, { "epoch": 4.49, "learning_rate": 5.507880498706187e-05, "loss": 2.792, "step": 57288 }, { "epoch": 4.49, "learning_rate": 5.507802085783737e-05, "loss": 2.7523, "step": 57289 }, { "epoch": 4.49, "learning_rate": 5.507723672861288e-05, "loss": 2.6123, "step": 57290 }, { "epoch": 4.49, "learning_rate": 5.5076452599388387e-05, "loss": 2.7913, "step": 57291 }, { "epoch": 4.49, "learning_rate": 5.507566847016389e-05, "loss": 2.7838, "step": 57292 }, { "epoch": 4.49, "learning_rate": 5.5074884340939394e-05, "loss": 2.6884, "step": 57293 }, { "epoch": 4.49, "learning_rate": 5.50741002117149e-05, "loss": 2.7192, "step": 57294 }, { "epoch": 4.49, "learning_rate": 5.50733160824904e-05, "loss": 2.9018, "step": 57295 }, { "epoch": 4.49, "learning_rate": 5.5072531953265905e-05, "loss": 2.7144, "step": 57296 }, { "epoch": 4.49, "learning_rate": 5.507174782404141e-05, "loss": 2.7523, "step": 57297 }, { "epoch": 4.49, "learning_rate": 5.507096369481691e-05, "loss": 2.654, "step": 57298 }, { "epoch": 4.49, "learning_rate": 5.5070179565592416e-05, "loss": 2.6294, "step": 57299 }, { "epoch": 4.49, "learning_rate": 5.506939543636792e-05, "loss": 2.7516, "step": 57300 }, { "epoch": 4.49, "learning_rate": 5.506861130714342e-05, "loss": 2.4973, "step": 57301 }, { "epoch": 4.49, "learning_rate": 5.506782717791893e-05, "loss": 2.5636, "step": 57302 }, { "epoch": 4.49, "learning_rate": 5.5067043048694424e-05, "loss": 2.646, "step": 57303 }, { "epoch": 4.49, "learning_rate": 5.506625891946993e-05, "loss": 2.8211, "step": 57304 }, { "epoch": 4.49, "learning_rate": 5.506547479024543e-05, "loss": 2.6832, "step": 57305 }, { "epoch": 4.49, "learning_rate": 5.5064690661020935e-05, "loss": 2.7165, "step": 57306 }, { "epoch": 4.49, "learning_rate": 5.506390653179644e-05, "loss": 2.7916, "step": 57307 }, { "epoch": 4.49, "learning_rate": 5.506312240257194e-05, "loss": 2.7656, "step": 57308 }, { "epoch": 4.49, "learning_rate": 5.5062338273347446e-05, "loss": 2.6401, "step": 57309 }, { "epoch": 4.49, "learning_rate": 5.506155414412295e-05, "loss": 3.0748, "step": 57310 }, { "epoch": 4.49, "learning_rate": 5.506077001489845e-05, "loss": 2.6623, "step": 57311 }, { "epoch": 4.49, "learning_rate": 5.505998588567396e-05, "loss": 2.5187, "step": 57312 }, { "epoch": 4.49, "learning_rate": 5.505920175644946e-05, "loss": 2.6929, "step": 57313 }, { "epoch": 4.49, "learning_rate": 5.5058417627224964e-05, "loss": 2.8106, "step": 57314 }, { "epoch": 4.49, "learning_rate": 5.5057633498000475e-05, "loss": 2.5159, "step": 57315 }, { "epoch": 4.49, "learning_rate": 5.505684936877598e-05, "loss": 2.5681, "step": 57316 }, { "epoch": 4.49, "learning_rate": 5.505606523955148e-05, "loss": 2.6583, "step": 57317 }, { "epoch": 4.49, "learning_rate": 5.5055281110326986e-05, "loss": 2.6341, "step": 57318 }, { "epoch": 4.49, "learning_rate": 5.505449698110249e-05, "loss": 2.8815, "step": 57319 }, { "epoch": 4.49, "learning_rate": 5.505371285187799e-05, "loss": 2.6702, "step": 57320 }, { "epoch": 4.49, "learning_rate": 5.50529287226535e-05, "loss": 2.5905, "step": 57321 }, { "epoch": 4.49, "learning_rate": 5.5052144593429e-05, "loss": 2.5647, "step": 57322 }, { "epoch": 4.49, "learning_rate": 5.5051360464204504e-05, "loss": 2.8023, "step": 57323 }, { "epoch": 4.49, "learning_rate": 5.505057633498001e-05, "loss": 2.671, "step": 57324 }, { "epoch": 4.5, "learning_rate": 5.504979220575551e-05, "loss": 2.7116, "step": 57325 }, { "epoch": 4.5, "learning_rate": 5.5049008076531015e-05, "loss": 2.6276, "step": 57326 }, { "epoch": 4.5, "learning_rate": 5.504822394730652e-05, "loss": 2.7875, "step": 57327 }, { "epoch": 4.5, "learning_rate": 5.504743981808202e-05, "loss": 2.6198, "step": 57328 }, { "epoch": 4.5, "learning_rate": 5.5046655688857526e-05, "loss": 2.6593, "step": 57329 }, { "epoch": 4.5, "learning_rate": 5.504587155963303e-05, "loss": 2.6901, "step": 57330 }, { "epoch": 4.5, "learning_rate": 5.504508743040854e-05, "loss": 2.4169, "step": 57331 }, { "epoch": 4.5, "learning_rate": 5.5044303301184044e-05, "loss": 2.7068, "step": 57332 }, { "epoch": 4.5, "learning_rate": 5.504351917195955e-05, "loss": 2.6317, "step": 57333 }, { "epoch": 4.5, "learning_rate": 5.504273504273505e-05, "loss": 2.7821, "step": 57334 }, { "epoch": 4.5, "learning_rate": 5.5041950913510555e-05, "loss": 2.7045, "step": 57335 }, { "epoch": 4.5, "learning_rate": 5.504116678428606e-05, "loss": 2.6639, "step": 57336 }, { "epoch": 4.5, "learning_rate": 5.504038265506156e-05, "loss": 2.4642, "step": 57337 }, { "epoch": 4.5, "learning_rate": 5.5039598525837066e-05, "loss": 2.6066, "step": 57338 }, { "epoch": 4.5, "learning_rate": 5.5038814396612556e-05, "loss": 2.7235, "step": 57339 }, { "epoch": 4.5, "learning_rate": 5.503803026738806e-05, "loss": 2.6958, "step": 57340 }, { "epoch": 4.5, "learning_rate": 5.503724613816357e-05, "loss": 2.8208, "step": 57341 }, { "epoch": 4.5, "learning_rate": 5.5036462008939074e-05, "loss": 2.5168, "step": 57342 }, { "epoch": 4.5, "learning_rate": 5.503567787971458e-05, "loss": 2.5849, "step": 57343 }, { "epoch": 4.5, "learning_rate": 5.503489375049008e-05, "loss": 2.5225, "step": 57344 }, { "epoch": 4.5, "learning_rate": 5.5034109621265585e-05, "loss": 2.5429, "step": 57345 }, { "epoch": 4.5, "learning_rate": 5.503332549204109e-05, "loss": 2.706, "step": 57346 }, { "epoch": 4.5, "learning_rate": 5.503254136281659e-05, "loss": 2.6946, "step": 57347 }, { "epoch": 4.5, "learning_rate": 5.5031757233592096e-05, "loss": 2.6375, "step": 57348 }, { "epoch": 4.5, "learning_rate": 5.50309731043676e-05, "loss": 2.5898, "step": 57349 }, { "epoch": 4.5, "learning_rate": 5.5030188975143104e-05, "loss": 2.7047, "step": 57350 }, { "epoch": 4.5, "learning_rate": 5.502940484591861e-05, "loss": 2.5657, "step": 57351 }, { "epoch": 4.5, "learning_rate": 5.502862071669411e-05, "loss": 2.6759, "step": 57352 }, { "epoch": 4.5, "learning_rate": 5.5027836587469615e-05, "loss": 2.6097, "step": 57353 }, { "epoch": 4.5, "learning_rate": 5.502705245824512e-05, "loss": 2.6212, "step": 57354 }, { "epoch": 4.5, "learning_rate": 5.502626832902062e-05, "loss": 2.5899, "step": 57355 }, { "epoch": 4.5, "learning_rate": 5.5025484199796126e-05, "loss": 2.7346, "step": 57356 }, { "epoch": 4.5, "learning_rate": 5.5024700070571636e-05, "loss": 2.6698, "step": 57357 }, { "epoch": 4.5, "learning_rate": 5.502391594134714e-05, "loss": 2.7208, "step": 57358 }, { "epoch": 4.5, "learning_rate": 5.5023131812122643e-05, "loss": 2.6852, "step": 57359 }, { "epoch": 4.5, "learning_rate": 5.502234768289815e-05, "loss": 2.6206, "step": 57360 }, { "epoch": 4.5, "learning_rate": 5.502156355367365e-05, "loss": 2.6158, "step": 57361 }, { "epoch": 4.5, "learning_rate": 5.5020779424449154e-05, "loss": 2.6555, "step": 57362 }, { "epoch": 4.5, "learning_rate": 5.501999529522466e-05, "loss": 2.9954, "step": 57363 }, { "epoch": 4.5, "learning_rate": 5.501921116600016e-05, "loss": 2.6179, "step": 57364 }, { "epoch": 4.5, "learning_rate": 5.5018427036775666e-05, "loss": 2.682, "step": 57365 }, { "epoch": 4.5, "learning_rate": 5.501764290755117e-05, "loss": 2.7947, "step": 57366 }, { "epoch": 4.5, "learning_rate": 5.501685877832667e-05, "loss": 2.8591, "step": 57367 }, { "epoch": 4.5, "learning_rate": 5.5016074649102177e-05, "loss": 2.6588, "step": 57368 }, { "epoch": 4.5, "learning_rate": 5.501529051987768e-05, "loss": 2.6517, "step": 57369 }, { "epoch": 4.5, "learning_rate": 5.5014506390653184e-05, "loss": 2.6938, "step": 57370 }, { "epoch": 4.5, "learning_rate": 5.501372226142869e-05, "loss": 2.6113, "step": 57371 }, { "epoch": 4.5, "learning_rate": 5.50129381322042e-05, "loss": 2.6818, "step": 57372 }, { "epoch": 4.5, "learning_rate": 5.50121540029797e-05, "loss": 2.8571, "step": 57373 }, { "epoch": 4.5, "learning_rate": 5.5011369873755205e-05, "loss": 2.7371, "step": 57374 }, { "epoch": 4.5, "learning_rate": 5.5010585744530696e-05, "loss": 2.809, "step": 57375 }, { "epoch": 4.5, "learning_rate": 5.50098016153062e-05, "loss": 2.7058, "step": 57376 }, { "epoch": 4.5, "learning_rate": 5.50090174860817e-05, "loss": 2.703, "step": 57377 }, { "epoch": 4.5, "learning_rate": 5.5008233356857207e-05, "loss": 2.7089, "step": 57378 }, { "epoch": 4.5, "learning_rate": 5.500744922763271e-05, "loss": 2.6059, "step": 57379 }, { "epoch": 4.5, "learning_rate": 5.5006665098408214e-05, "loss": 2.736, "step": 57380 }, { "epoch": 4.5, "learning_rate": 5.500588096918372e-05, "loss": 2.722, "step": 57381 }, { "epoch": 4.5, "learning_rate": 5.500509683995923e-05, "loss": 2.4451, "step": 57382 }, { "epoch": 4.5, "learning_rate": 5.500431271073473e-05, "loss": 2.6164, "step": 57383 }, { "epoch": 4.5, "learning_rate": 5.5003528581510235e-05, "loss": 2.5695, "step": 57384 }, { "epoch": 4.5, "learning_rate": 5.500274445228574e-05, "loss": 2.6376, "step": 57385 }, { "epoch": 4.5, "learning_rate": 5.500196032306124e-05, "loss": 2.5811, "step": 57386 }, { "epoch": 4.5, "learning_rate": 5.5001176193836746e-05, "loss": 2.5288, "step": 57387 }, { "epoch": 4.5, "learning_rate": 5.500039206461225e-05, "loss": 2.5678, "step": 57388 }, { "epoch": 4.5, "learning_rate": 5.4999607935387754e-05, "loss": 2.6049, "step": 57389 }, { "epoch": 4.5, "learning_rate": 5.499882380616326e-05, "loss": 2.7712, "step": 57390 }, { "epoch": 4.5, "learning_rate": 5.499803967693876e-05, "loss": 2.6928, "step": 57391 }, { "epoch": 4.5, "learning_rate": 5.4997255547714265e-05, "loss": 2.6818, "step": 57392 }, { "epoch": 4.5, "learning_rate": 5.499647141848977e-05, "loss": 2.5477, "step": 57393 }, { "epoch": 4.5, "learning_rate": 5.499568728926527e-05, "loss": 2.4934, "step": 57394 }, { "epoch": 4.5, "learning_rate": 5.4994903160040776e-05, "loss": 2.5338, "step": 57395 }, { "epoch": 4.5, "learning_rate": 5.499411903081628e-05, "loss": 2.6192, "step": 57396 }, { "epoch": 4.5, "learning_rate": 5.499333490159178e-05, "loss": 2.603, "step": 57397 }, { "epoch": 4.5, "learning_rate": 5.4992550772367294e-05, "loss": 2.6341, "step": 57398 }, { "epoch": 4.5, "learning_rate": 5.49917666431428e-05, "loss": 2.9167, "step": 57399 }, { "epoch": 4.5, "learning_rate": 5.49909825139183e-05, "loss": 2.7604, "step": 57400 }, { "epoch": 4.5, "learning_rate": 5.4990198384693805e-05, "loss": 2.5974, "step": 57401 }, { "epoch": 4.5, "learning_rate": 5.498941425546931e-05, "loss": 2.8966, "step": 57402 }, { "epoch": 4.5, "learning_rate": 5.498863012624481e-05, "loss": 2.6206, "step": 57403 }, { "epoch": 4.5, "learning_rate": 5.4987845997020316e-05, "loss": 2.626, "step": 57404 }, { "epoch": 4.5, "learning_rate": 5.498706186779582e-05, "loss": 3.0196, "step": 57405 }, { "epoch": 4.5, "learning_rate": 5.498627773857132e-05, "loss": 2.4909, "step": 57406 }, { "epoch": 4.5, "learning_rate": 5.498549360934683e-05, "loss": 2.6771, "step": 57407 }, { "epoch": 4.5, "learning_rate": 5.498470948012233e-05, "loss": 2.4604, "step": 57408 }, { "epoch": 4.5, "learning_rate": 5.4983925350897834e-05, "loss": 2.6082, "step": 57409 }, { "epoch": 4.5, "learning_rate": 5.498314122167334e-05, "loss": 2.7066, "step": 57410 }, { "epoch": 4.5, "learning_rate": 5.4982357092448835e-05, "loss": 2.6765, "step": 57411 }, { "epoch": 4.5, "learning_rate": 5.498157296322434e-05, "loss": 2.536, "step": 57412 }, { "epoch": 4.5, "learning_rate": 5.498078883399984e-05, "loss": 2.6336, "step": 57413 }, { "epoch": 4.5, "learning_rate": 5.4980004704775346e-05, "loss": 2.4034, "step": 57414 }, { "epoch": 4.5, "learning_rate": 5.497922057555085e-05, "loss": 2.8276, "step": 57415 }, { "epoch": 4.5, "learning_rate": 5.497843644632635e-05, "loss": 2.6878, "step": 57416 }, { "epoch": 4.5, "learning_rate": 5.497765231710186e-05, "loss": 2.8823, "step": 57417 }, { "epoch": 4.5, "learning_rate": 5.497686818787736e-05, "loss": 2.6976, "step": 57418 }, { "epoch": 4.5, "learning_rate": 5.4976084058652864e-05, "loss": 2.5234, "step": 57419 }, { "epoch": 4.5, "learning_rate": 5.497529992942837e-05, "loss": 2.9068, "step": 57420 }, { "epoch": 4.5, "learning_rate": 5.497451580020387e-05, "loss": 2.8119, "step": 57421 }, { "epoch": 4.5, "learning_rate": 5.4973731670979375e-05, "loss": 2.4988, "step": 57422 }, { "epoch": 4.5, "learning_rate": 5.4972947541754886e-05, "loss": 2.5663, "step": 57423 }, { "epoch": 4.5, "learning_rate": 5.497216341253039e-05, "loss": 2.7369, "step": 57424 }, { "epoch": 4.5, "learning_rate": 5.497137928330589e-05, "loss": 2.687, "step": 57425 }, { "epoch": 4.5, "learning_rate": 5.49705951540814e-05, "loss": 2.6089, "step": 57426 }, { "epoch": 4.5, "learning_rate": 5.49698110248569e-05, "loss": 2.7395, "step": 57427 }, { "epoch": 4.5, "learning_rate": 5.4969026895632404e-05, "loss": 2.429, "step": 57428 }, { "epoch": 4.5, "learning_rate": 5.496824276640791e-05, "loss": 2.5869, "step": 57429 }, { "epoch": 4.5, "learning_rate": 5.496745863718341e-05, "loss": 2.4349, "step": 57430 }, { "epoch": 4.5, "learning_rate": 5.4966674507958915e-05, "loss": 2.6788, "step": 57431 }, { "epoch": 4.5, "learning_rate": 5.496589037873442e-05, "loss": 2.8346, "step": 57432 }, { "epoch": 4.5, "learning_rate": 5.496510624950992e-05, "loss": 2.5223, "step": 57433 }, { "epoch": 4.5, "learning_rate": 5.4964322120285426e-05, "loss": 2.607, "step": 57434 }, { "epoch": 4.5, "learning_rate": 5.496353799106093e-05, "loss": 2.6332, "step": 57435 }, { "epoch": 4.5, "learning_rate": 5.4962753861836433e-05, "loss": 2.6984, "step": 57436 }, { "epoch": 4.5, "learning_rate": 5.496196973261194e-05, "loss": 2.8122, "step": 57437 }, { "epoch": 4.5, "learning_rate": 5.496118560338744e-05, "loss": 2.5931, "step": 57438 }, { "epoch": 4.5, "learning_rate": 5.496040147416295e-05, "loss": 2.7487, "step": 57439 }, { "epoch": 4.5, "learning_rate": 5.4959617344938455e-05, "loss": 3.1187, "step": 57440 }, { "epoch": 4.5, "learning_rate": 5.495883321571396e-05, "loss": 2.6423, "step": 57441 }, { "epoch": 4.5, "learning_rate": 5.495804908648946e-05, "loss": 2.6054, "step": 57442 }, { "epoch": 4.5, "learning_rate": 5.4957264957264966e-05, "loss": 2.8697, "step": 57443 }, { "epoch": 4.5, "learning_rate": 5.495648082804047e-05, "loss": 2.5775, "step": 57444 }, { "epoch": 4.5, "learning_rate": 5.495569669881597e-05, "loss": 2.7147, "step": 57445 }, { "epoch": 4.5, "learning_rate": 5.495491256959148e-05, "loss": 2.636, "step": 57446 }, { "epoch": 4.5, "learning_rate": 5.495412844036697e-05, "loss": 2.7511, "step": 57447 }, { "epoch": 4.5, "learning_rate": 5.495334431114247e-05, "loss": 2.6644, "step": 57448 }, { "epoch": 4.5, "learning_rate": 5.495256018191798e-05, "loss": 2.5216, "step": 57449 }, { "epoch": 4.5, "learning_rate": 5.4951776052693485e-05, "loss": 2.7922, "step": 57450 }, { "epoch": 4.5, "learning_rate": 5.495099192346899e-05, "loss": 2.6798, "step": 57451 }, { "epoch": 4.5, "learning_rate": 5.495020779424449e-05, "loss": 2.6461, "step": 57452 }, { "epoch": 4.51, "learning_rate": 5.4949423665019996e-05, "loss": 2.7452, "step": 57453 }, { "epoch": 4.51, "learning_rate": 5.49486395357955e-05, "loss": 2.8446, "step": 57454 }, { "epoch": 4.51, "learning_rate": 5.4947855406571e-05, "loss": 2.6929, "step": 57455 }, { "epoch": 4.51, "learning_rate": 5.494707127734651e-05, "loss": 2.6109, "step": 57456 }, { "epoch": 4.51, "learning_rate": 5.494628714812201e-05, "loss": 2.8072, "step": 57457 }, { "epoch": 4.51, "learning_rate": 5.4945503018897514e-05, "loss": 2.5413, "step": 57458 }, { "epoch": 4.51, "learning_rate": 5.494471888967302e-05, "loss": 2.5939, "step": 57459 }, { "epoch": 4.51, "learning_rate": 5.494393476044852e-05, "loss": 2.6132, "step": 57460 }, { "epoch": 4.51, "learning_rate": 5.4943150631224025e-05, "loss": 2.7115, "step": 57461 }, { "epoch": 4.51, "learning_rate": 5.494236650199953e-05, "loss": 2.6289, "step": 57462 }, { "epoch": 4.51, "learning_rate": 5.494158237277503e-05, "loss": 2.5618, "step": 57463 }, { "epoch": 4.51, "learning_rate": 5.4940798243550536e-05, "loss": 2.6352, "step": 57464 }, { "epoch": 4.51, "learning_rate": 5.494001411432605e-05, "loss": 2.8592, "step": 57465 }, { "epoch": 4.51, "learning_rate": 5.493922998510155e-05, "loss": 2.6421, "step": 57466 }, { "epoch": 4.51, "learning_rate": 5.4938445855877054e-05, "loss": 2.8083, "step": 57467 }, { "epoch": 4.51, "learning_rate": 5.493766172665256e-05, "loss": 2.5952, "step": 57468 }, { "epoch": 4.51, "learning_rate": 5.493687759742806e-05, "loss": 2.4297, "step": 57469 }, { "epoch": 4.51, "learning_rate": 5.4936093468203565e-05, "loss": 2.6071, "step": 57470 }, { "epoch": 4.51, "learning_rate": 5.493530933897907e-05, "loss": 2.6538, "step": 57471 }, { "epoch": 4.51, "learning_rate": 5.493452520975457e-05, "loss": 2.6136, "step": 57472 }, { "epoch": 4.51, "learning_rate": 5.4933741080530076e-05, "loss": 2.4518, "step": 57473 }, { "epoch": 4.51, "learning_rate": 5.493295695130558e-05, "loss": 2.8374, "step": 57474 }, { "epoch": 4.51, "learning_rate": 5.4932172822081084e-05, "loss": 2.7566, "step": 57475 }, { "epoch": 4.51, "learning_rate": 5.493138869285659e-05, "loss": 2.6062, "step": 57476 }, { "epoch": 4.51, "learning_rate": 5.493060456363209e-05, "loss": 2.6881, "step": 57477 }, { "epoch": 4.51, "learning_rate": 5.4929820434407595e-05, "loss": 2.5411, "step": 57478 }, { "epoch": 4.51, "learning_rate": 5.49290363051831e-05, "loss": 2.8835, "step": 57479 }, { "epoch": 4.51, "learning_rate": 5.492825217595861e-05, "loss": 2.5898, "step": 57480 }, { "epoch": 4.51, "learning_rate": 5.492746804673411e-05, "loss": 2.6681, "step": 57481 }, { "epoch": 4.51, "learning_rate": 5.4926683917509616e-05, "loss": 2.8125, "step": 57482 }, { "epoch": 4.51, "learning_rate": 5.4925899788285106e-05, "loss": 2.5342, "step": 57483 }, { "epoch": 4.51, "learning_rate": 5.492511565906061e-05, "loss": 2.6732, "step": 57484 }, { "epoch": 4.51, "learning_rate": 5.4924331529836114e-05, "loss": 2.5264, "step": 57485 }, { "epoch": 4.51, "learning_rate": 5.492354740061162e-05, "loss": 2.6256, "step": 57486 }, { "epoch": 4.51, "learning_rate": 5.492276327138712e-05, "loss": 2.8186, "step": 57487 }, { "epoch": 4.51, "learning_rate": 5.4921979142162625e-05, "loss": 2.8445, "step": 57488 }, { "epoch": 4.51, "learning_rate": 5.492119501293813e-05, "loss": 2.736, "step": 57489 }, { "epoch": 4.51, "learning_rate": 5.492041088371364e-05, "loss": 2.7699, "step": 57490 }, { "epoch": 4.51, "learning_rate": 5.491962675448914e-05, "loss": 2.5844, "step": 57491 }, { "epoch": 4.51, "learning_rate": 5.4918842625264646e-05, "loss": 2.6373, "step": 57492 }, { "epoch": 4.51, "learning_rate": 5.491805849604015e-05, "loss": 2.7151, "step": 57493 }, { "epoch": 4.51, "learning_rate": 5.4917274366815654e-05, "loss": 2.6025, "step": 57494 }, { "epoch": 4.51, "learning_rate": 5.491649023759116e-05, "loss": 2.4828, "step": 57495 }, { "epoch": 4.51, "learning_rate": 5.491570610836666e-05, "loss": 2.658, "step": 57496 }, { "epoch": 4.51, "learning_rate": 5.4914921979142165e-05, "loss": 2.5855, "step": 57497 }, { "epoch": 4.51, "learning_rate": 5.491413784991767e-05, "loss": 2.6727, "step": 57498 }, { "epoch": 4.51, "learning_rate": 5.491335372069317e-05, "loss": 2.6175, "step": 57499 }, { "epoch": 4.51, "learning_rate": 5.4912569591468676e-05, "loss": 2.7173, "step": 57500 }, { "epoch": 4.51, "learning_rate": 5.491178546224418e-05, "loss": 2.7379, "step": 57501 }, { "epoch": 4.51, "learning_rate": 5.491100133301968e-05, "loss": 2.6434, "step": 57502 }, { "epoch": 4.51, "learning_rate": 5.4910217203795187e-05, "loss": 2.6287, "step": 57503 }, { "epoch": 4.51, "learning_rate": 5.490943307457069e-05, "loss": 2.5041, "step": 57504 }, { "epoch": 4.51, "learning_rate": 5.4908648945346194e-05, "loss": 2.5148, "step": 57505 }, { "epoch": 4.51, "learning_rate": 5.4907864816121704e-05, "loss": 2.8185, "step": 57506 }, { "epoch": 4.51, "learning_rate": 5.490708068689721e-05, "loss": 2.5714, "step": 57507 }, { "epoch": 4.51, "learning_rate": 5.490629655767271e-05, "loss": 2.464, "step": 57508 }, { "epoch": 4.51, "learning_rate": 5.4905512428448215e-05, "loss": 2.7841, "step": 57509 }, { "epoch": 4.51, "learning_rate": 5.490472829922372e-05, "loss": 2.6538, "step": 57510 }, { "epoch": 4.51, "learning_rate": 5.490394416999922e-05, "loss": 2.5365, "step": 57511 }, { "epoch": 4.51, "learning_rate": 5.4903160040774726e-05, "loss": 2.6186, "step": 57512 }, { "epoch": 4.51, "learning_rate": 5.490237591155023e-05, "loss": 3.0867, "step": 57513 }, { "epoch": 4.51, "learning_rate": 5.4901591782325734e-05, "loss": 2.8272, "step": 57514 }, { "epoch": 4.51, "learning_rate": 5.490080765310124e-05, "loss": 2.9576, "step": 57515 }, { "epoch": 4.51, "learning_rate": 5.490002352387674e-05, "loss": 2.3901, "step": 57516 }, { "epoch": 4.51, "learning_rate": 5.4899239394652245e-05, "loss": 2.6666, "step": 57517 }, { "epoch": 4.51, "learning_rate": 5.489845526542775e-05, "loss": 2.6071, "step": 57518 }, { "epoch": 4.51, "learning_rate": 5.4897671136203245e-05, "loss": 2.7513, "step": 57519 }, { "epoch": 4.51, "learning_rate": 5.489688700697875e-05, "loss": 2.6482, "step": 57520 }, { "epoch": 4.51, "learning_rate": 5.489610287775425e-05, "loss": 2.7017, "step": 57521 }, { "epoch": 4.51, "learning_rate": 5.4895318748529756e-05, "loss": 2.6711, "step": 57522 }, { "epoch": 4.51, "learning_rate": 5.489453461930526e-05, "loss": 2.5497, "step": 57523 }, { "epoch": 4.51, "learning_rate": 5.4893750490080764e-05, "loss": 2.5897, "step": 57524 }, { "epoch": 4.51, "learning_rate": 5.489296636085627e-05, "loss": 2.7233, "step": 57525 }, { "epoch": 4.51, "learning_rate": 5.489218223163177e-05, "loss": 2.5699, "step": 57526 }, { "epoch": 4.51, "learning_rate": 5.4891398102407275e-05, "loss": 2.6545, "step": 57527 }, { "epoch": 4.51, "learning_rate": 5.489061397318278e-05, "loss": 2.743, "step": 57528 }, { "epoch": 4.51, "learning_rate": 5.488982984395828e-05, "loss": 2.6845, "step": 57529 }, { "epoch": 4.51, "learning_rate": 5.4889045714733786e-05, "loss": 2.6377, "step": 57530 }, { "epoch": 4.51, "learning_rate": 5.4888261585509296e-05, "loss": 2.6594, "step": 57531 }, { "epoch": 4.51, "learning_rate": 5.48874774562848e-05, "loss": 2.5592, "step": 57532 }, { "epoch": 4.51, "learning_rate": 5.4886693327060304e-05, "loss": 2.5184, "step": 57533 }, { "epoch": 4.51, "learning_rate": 5.488590919783581e-05, "loss": 2.7666, "step": 57534 }, { "epoch": 4.51, "learning_rate": 5.488512506861131e-05, "loss": 2.6135, "step": 57535 }, { "epoch": 4.51, "learning_rate": 5.4884340939386815e-05, "loss": 2.7948, "step": 57536 }, { "epoch": 4.51, "learning_rate": 5.488355681016232e-05, "loss": 2.5369, "step": 57537 }, { "epoch": 4.51, "learning_rate": 5.488277268093782e-05, "loss": 2.7423, "step": 57538 }, { "epoch": 4.51, "learning_rate": 5.4881988551713326e-05, "loss": 2.6633, "step": 57539 }, { "epoch": 4.51, "learning_rate": 5.488120442248883e-05, "loss": 2.6025, "step": 57540 }, { "epoch": 4.51, "learning_rate": 5.488042029326433e-05, "loss": 2.6975, "step": 57541 }, { "epoch": 4.51, "learning_rate": 5.487963616403984e-05, "loss": 2.834, "step": 57542 }, { "epoch": 4.51, "learning_rate": 5.487885203481534e-05, "loss": 2.8102, "step": 57543 }, { "epoch": 4.51, "learning_rate": 5.4878067905590844e-05, "loss": 2.6091, "step": 57544 }, { "epoch": 4.51, "learning_rate": 5.487728377636635e-05, "loss": 2.6643, "step": 57545 }, { "epoch": 4.51, "learning_rate": 5.487649964714185e-05, "loss": 2.6645, "step": 57546 }, { "epoch": 4.51, "learning_rate": 5.487571551791736e-05, "loss": 2.776, "step": 57547 }, { "epoch": 4.51, "learning_rate": 5.4874931388692866e-05, "loss": 2.4706, "step": 57548 }, { "epoch": 4.51, "learning_rate": 5.487414725946837e-05, "loss": 2.7114, "step": 57549 }, { "epoch": 4.51, "learning_rate": 5.487336313024387e-05, "loss": 2.7178, "step": 57550 }, { "epoch": 4.51, "learning_rate": 5.487257900101938e-05, "loss": 2.6877, "step": 57551 }, { "epoch": 4.51, "learning_rate": 5.487179487179488e-05, "loss": 2.796, "step": 57552 }, { "epoch": 4.51, "learning_rate": 5.4871010742570384e-05, "loss": 2.5571, "step": 57553 }, { "epoch": 4.51, "learning_rate": 5.487022661334589e-05, "loss": 2.3862, "step": 57554 }, { "epoch": 4.51, "learning_rate": 5.486944248412138e-05, "loss": 2.5752, "step": 57555 }, { "epoch": 4.51, "learning_rate": 5.486865835489688e-05, "loss": 2.7188, "step": 57556 }, { "epoch": 4.51, "learning_rate": 5.486787422567239e-05, "loss": 2.726, "step": 57557 }, { "epoch": 4.51, "learning_rate": 5.4867090096447896e-05, "loss": 2.5568, "step": 57558 }, { "epoch": 4.51, "learning_rate": 5.48663059672234e-05, "loss": 2.7312, "step": 57559 }, { "epoch": 4.51, "learning_rate": 5.48655218379989e-05, "loss": 2.6245, "step": 57560 }, { "epoch": 4.51, "learning_rate": 5.486473770877441e-05, "loss": 2.7815, "step": 57561 }, { "epoch": 4.51, "learning_rate": 5.486395357954991e-05, "loss": 2.6793, "step": 57562 }, { "epoch": 4.51, "learning_rate": 5.4863169450325414e-05, "loss": 2.4703, "step": 57563 }, { "epoch": 4.51, "learning_rate": 5.486238532110092e-05, "loss": 2.5254, "step": 57564 }, { "epoch": 4.51, "learning_rate": 5.486160119187642e-05, "loss": 2.6561, "step": 57565 }, { "epoch": 4.51, "learning_rate": 5.4860817062651925e-05, "loss": 2.691, "step": 57566 }, { "epoch": 4.51, "learning_rate": 5.486003293342743e-05, "loss": 2.84, "step": 57567 }, { "epoch": 4.51, "learning_rate": 5.485924880420293e-05, "loss": 2.6201, "step": 57568 }, { "epoch": 4.51, "learning_rate": 5.4858464674978436e-05, "loss": 2.724, "step": 57569 }, { "epoch": 4.51, "learning_rate": 5.485768054575394e-05, "loss": 2.5425, "step": 57570 }, { "epoch": 4.51, "learning_rate": 5.4856896416529443e-05, "loss": 2.6278, "step": 57571 }, { "epoch": 4.51, "learning_rate": 5.485611228730495e-05, "loss": 2.7089, "step": 57572 }, { "epoch": 4.51, "learning_rate": 5.485532815808046e-05, "loss": 2.6586, "step": 57573 }, { "epoch": 4.51, "learning_rate": 5.485454402885596e-05, "loss": 2.6672, "step": 57574 }, { "epoch": 4.51, "learning_rate": 5.4853759899631465e-05, "loss": 2.6543, "step": 57575 }, { "epoch": 4.51, "learning_rate": 5.485297577040697e-05, "loss": 2.5732, "step": 57576 }, { "epoch": 4.51, "learning_rate": 5.485219164118247e-05, "loss": 2.5865, "step": 57577 }, { "epoch": 4.51, "learning_rate": 5.4851407511957976e-05, "loss": 2.6322, "step": 57578 }, { "epoch": 4.51, "learning_rate": 5.485062338273348e-05, "loss": 2.5195, "step": 57579 }, { "epoch": 4.52, "learning_rate": 5.484983925350898e-05, "loss": 2.5203, "step": 57580 }, { "epoch": 4.52, "learning_rate": 5.484905512428449e-05, "loss": 2.6394, "step": 57581 }, { "epoch": 4.52, "learning_rate": 5.484827099505999e-05, "loss": 2.8744, "step": 57582 }, { "epoch": 4.52, "learning_rate": 5.4847486865835494e-05, "loss": 2.4519, "step": 57583 }, { "epoch": 4.52, "learning_rate": 5.4846702736611e-05, "loss": 2.437, "step": 57584 }, { "epoch": 4.52, "learning_rate": 5.48459186073865e-05, "loss": 2.8231, "step": 57585 }, { "epoch": 4.52, "learning_rate": 5.4845134478162005e-05, "loss": 2.4458, "step": 57586 }, { "epoch": 4.52, "learning_rate": 5.484435034893751e-05, "loss": 2.7581, "step": 57587 }, { "epoch": 4.52, "learning_rate": 5.484356621971302e-05, "loss": 2.5645, "step": 57588 }, { "epoch": 4.52, "learning_rate": 5.484278209048852e-05, "loss": 2.5972, "step": 57589 }, { "epoch": 4.52, "learning_rate": 5.484199796126403e-05, "loss": 2.6077, "step": 57590 }, { "epoch": 4.52, "learning_rate": 5.484121383203952e-05, "loss": 2.7351, "step": 57591 }, { "epoch": 4.52, "learning_rate": 5.484042970281502e-05, "loss": 2.7587, "step": 57592 }, { "epoch": 4.52, "learning_rate": 5.4839645573590524e-05, "loss": 2.7589, "step": 57593 }, { "epoch": 4.52, "learning_rate": 5.483886144436603e-05, "loss": 2.6748, "step": 57594 }, { "epoch": 4.52, "learning_rate": 5.483807731514153e-05, "loss": 2.5555, "step": 57595 }, { "epoch": 4.52, "learning_rate": 5.4837293185917035e-05, "loss": 2.5564, "step": 57596 }, { "epoch": 4.52, "learning_rate": 5.483650905669254e-05, "loss": 2.7877, "step": 57597 }, { "epoch": 4.52, "learning_rate": 5.483572492746805e-05, "loss": 2.5321, "step": 57598 }, { "epoch": 4.52, "learning_rate": 5.483494079824355e-05, "loss": 2.6251, "step": 57599 }, { "epoch": 4.52, "learning_rate": 5.483415666901906e-05, "loss": 2.6393, "step": 57600 }, { "epoch": 4.52, "learning_rate": 5.483337253979456e-05, "loss": 2.5214, "step": 57601 }, { "epoch": 4.52, "learning_rate": 5.4832588410570064e-05, "loss": 2.6454, "step": 57602 }, { "epoch": 4.52, "learning_rate": 5.483180428134557e-05, "loss": 2.7103, "step": 57603 }, { "epoch": 4.52, "learning_rate": 5.483102015212107e-05, "loss": 2.5978, "step": 57604 }, { "epoch": 4.52, "learning_rate": 5.4830236022896575e-05, "loss": 2.6992, "step": 57605 }, { "epoch": 4.52, "learning_rate": 5.482945189367208e-05, "loss": 2.7993, "step": 57606 }, { "epoch": 4.52, "learning_rate": 5.482866776444758e-05, "loss": 2.5748, "step": 57607 }, { "epoch": 4.52, "learning_rate": 5.4827883635223086e-05, "loss": 2.6265, "step": 57608 }, { "epoch": 4.52, "learning_rate": 5.482709950599859e-05, "loss": 2.6526, "step": 57609 }, { "epoch": 4.52, "learning_rate": 5.4826315376774094e-05, "loss": 2.6446, "step": 57610 }, { "epoch": 4.52, "learning_rate": 5.48255312475496e-05, "loss": 2.6805, "step": 57611 }, { "epoch": 4.52, "learning_rate": 5.48247471183251e-05, "loss": 2.753, "step": 57612 }, { "epoch": 4.52, "learning_rate": 5.4823962989100605e-05, "loss": 2.5575, "step": 57613 }, { "epoch": 4.52, "learning_rate": 5.4823178859876115e-05, "loss": 2.4987, "step": 57614 }, { "epoch": 4.52, "learning_rate": 5.482239473065162e-05, "loss": 2.8271, "step": 57615 }, { "epoch": 4.52, "learning_rate": 5.482161060142712e-05, "loss": 2.5998, "step": 57616 }, { "epoch": 4.52, "learning_rate": 5.4820826472202626e-05, "loss": 2.6883, "step": 57617 }, { "epoch": 4.52, "learning_rate": 5.482004234297813e-05, "loss": 2.6209, "step": 57618 }, { "epoch": 4.52, "learning_rate": 5.4819258213753634e-05, "loss": 2.5326, "step": 57619 }, { "epoch": 4.52, "learning_rate": 5.481847408452914e-05, "loss": 2.7043, "step": 57620 }, { "epoch": 4.52, "learning_rate": 5.481768995530464e-05, "loss": 2.5778, "step": 57621 }, { "epoch": 4.52, "learning_rate": 5.4816905826080145e-05, "loss": 2.7413, "step": 57622 }, { "epoch": 4.52, "learning_rate": 5.481612169685565e-05, "loss": 2.8803, "step": 57623 }, { "epoch": 4.52, "learning_rate": 5.481533756763115e-05, "loss": 2.5457, "step": 57624 }, { "epoch": 4.52, "learning_rate": 5.4814553438406656e-05, "loss": 2.666, "step": 57625 }, { "epoch": 4.52, "learning_rate": 5.481376930918216e-05, "loss": 2.7147, "step": 57626 }, { "epoch": 4.52, "learning_rate": 5.4812985179957656e-05, "loss": 2.7462, "step": 57627 }, { "epoch": 4.52, "learning_rate": 5.481220105073316e-05, "loss": 2.5566, "step": 57628 }, { "epoch": 4.52, "learning_rate": 5.4811416921508664e-05, "loss": 2.7535, "step": 57629 }, { "epoch": 4.52, "learning_rate": 5.481063279228417e-05, "loss": 2.5657, "step": 57630 }, { "epoch": 4.52, "learning_rate": 5.480984866305967e-05, "loss": 2.5604, "step": 57631 }, { "epoch": 4.52, "learning_rate": 5.4809064533835175e-05, "loss": 2.5725, "step": 57632 }, { "epoch": 4.52, "learning_rate": 5.480828040461068e-05, "loss": 2.7658, "step": 57633 }, { "epoch": 4.52, "learning_rate": 5.480749627538618e-05, "loss": 2.667, "step": 57634 }, { "epoch": 4.52, "learning_rate": 5.4806712146161686e-05, "loss": 2.7537, "step": 57635 }, { "epoch": 4.52, "learning_rate": 5.480592801693719e-05, "loss": 2.5178, "step": 57636 }, { "epoch": 4.52, "learning_rate": 5.480514388771269e-05, "loss": 2.9605, "step": 57637 }, { "epoch": 4.52, "learning_rate": 5.48043597584882e-05, "loss": 2.8195, "step": 57638 }, { "epoch": 4.52, "learning_rate": 5.48035756292637e-05, "loss": 2.6746, "step": 57639 }, { "epoch": 4.52, "learning_rate": 5.480279150003921e-05, "loss": 3.0492, "step": 57640 }, { "epoch": 4.52, "learning_rate": 5.4802007370814714e-05, "loss": 2.5158, "step": 57641 }, { "epoch": 4.52, "learning_rate": 5.480122324159022e-05, "loss": 2.7275, "step": 57642 }, { "epoch": 4.52, "learning_rate": 5.480043911236572e-05, "loss": 2.6614, "step": 57643 }, { "epoch": 4.52, "learning_rate": 5.4799654983141225e-05, "loss": 2.5853, "step": 57644 }, { "epoch": 4.52, "learning_rate": 5.479887085391673e-05, "loss": 2.6412, "step": 57645 }, { "epoch": 4.52, "learning_rate": 5.479808672469223e-05, "loss": 2.7211, "step": 57646 }, { "epoch": 4.52, "learning_rate": 5.4797302595467737e-05, "loss": 2.6398, "step": 57647 }, { "epoch": 4.52, "learning_rate": 5.479651846624324e-05, "loss": 2.7806, "step": 57648 }, { "epoch": 4.52, "learning_rate": 5.4795734337018744e-05, "loss": 2.561, "step": 57649 }, { "epoch": 4.52, "learning_rate": 5.479495020779425e-05, "loss": 2.5884, "step": 57650 }, { "epoch": 4.52, "learning_rate": 5.479416607856975e-05, "loss": 2.8467, "step": 57651 }, { "epoch": 4.52, "learning_rate": 5.4793381949345255e-05, "loss": 2.7426, "step": 57652 }, { "epoch": 4.52, "learning_rate": 5.479259782012076e-05, "loss": 2.6289, "step": 57653 }, { "epoch": 4.52, "learning_rate": 5.479181369089626e-05, "loss": 2.5338, "step": 57654 }, { "epoch": 4.52, "learning_rate": 5.479102956167177e-05, "loss": 2.7796, "step": 57655 }, { "epoch": 4.52, "learning_rate": 5.4790245432447276e-05, "loss": 2.9474, "step": 57656 }, { "epoch": 4.52, "learning_rate": 5.478946130322278e-05, "loss": 2.5487, "step": 57657 }, { "epoch": 4.52, "learning_rate": 5.4788677173998284e-05, "loss": 2.7277, "step": 57658 }, { "epoch": 4.52, "learning_rate": 5.478789304477379e-05, "loss": 2.5804, "step": 57659 }, { "epoch": 4.52, "learning_rate": 5.478710891554929e-05, "loss": 2.5788, "step": 57660 }, { "epoch": 4.52, "learning_rate": 5.4786324786324795e-05, "loss": 2.5387, "step": 57661 }, { "epoch": 4.52, "learning_rate": 5.47855406571003e-05, "loss": 2.5975, "step": 57662 }, { "epoch": 4.52, "learning_rate": 5.478475652787579e-05, "loss": 2.5105, "step": 57663 }, { "epoch": 4.52, "learning_rate": 5.478397239865129e-05, "loss": 2.7192, "step": 57664 }, { "epoch": 4.52, "learning_rate": 5.47831882694268e-05, "loss": 2.64, "step": 57665 }, { "epoch": 4.52, "learning_rate": 5.4782404140202306e-05, "loss": 2.7145, "step": 57666 }, { "epoch": 4.52, "learning_rate": 5.478162001097781e-05, "loss": 2.7306, "step": 57667 }, { "epoch": 4.52, "learning_rate": 5.4780835881753314e-05, "loss": 2.5511, "step": 57668 }, { "epoch": 4.52, "learning_rate": 5.478005175252882e-05, "loss": 2.5525, "step": 57669 }, { "epoch": 4.52, "learning_rate": 5.477926762330432e-05, "loss": 2.5351, "step": 57670 }, { "epoch": 4.52, "learning_rate": 5.4778483494079825e-05, "loss": 3.0474, "step": 57671 }, { "epoch": 4.52, "learning_rate": 5.477769936485533e-05, "loss": 2.8055, "step": 57672 }, { "epoch": 4.52, "learning_rate": 5.477691523563083e-05, "loss": 2.5988, "step": 57673 }, { "epoch": 4.52, "learning_rate": 5.4776131106406336e-05, "loss": 2.4684, "step": 57674 }, { "epoch": 4.52, "learning_rate": 5.477534697718184e-05, "loss": 2.4685, "step": 57675 }, { "epoch": 4.52, "learning_rate": 5.477456284795734e-05, "loss": 2.7151, "step": 57676 }, { "epoch": 4.52, "learning_rate": 5.477377871873285e-05, "loss": 2.8377, "step": 57677 }, { "epoch": 4.52, "learning_rate": 5.477299458950835e-05, "loss": 2.651, "step": 57678 }, { "epoch": 4.52, "learning_rate": 5.4772210460283854e-05, "loss": 2.4589, "step": 57679 }, { "epoch": 4.52, "learning_rate": 5.477142633105936e-05, "loss": 2.6506, "step": 57680 }, { "epoch": 4.52, "learning_rate": 5.477064220183487e-05, "loss": 2.5711, "step": 57681 }, { "epoch": 4.52, "learning_rate": 5.476985807261037e-05, "loss": 2.5101, "step": 57682 }, { "epoch": 4.52, "learning_rate": 5.4769073943385876e-05, "loss": 2.6761, "step": 57683 }, { "epoch": 4.52, "learning_rate": 5.476828981416138e-05, "loss": 2.5167, "step": 57684 }, { "epoch": 4.52, "learning_rate": 5.476750568493688e-05, "loss": 2.6628, "step": 57685 }, { "epoch": 4.52, "learning_rate": 5.476672155571239e-05, "loss": 2.7439, "step": 57686 }, { "epoch": 4.52, "learning_rate": 5.476593742648789e-05, "loss": 2.6376, "step": 57687 }, { "epoch": 4.52, "learning_rate": 5.4765153297263394e-05, "loss": 2.7741, "step": 57688 }, { "epoch": 4.52, "learning_rate": 5.47643691680389e-05, "loss": 2.5907, "step": 57689 }, { "epoch": 4.52, "learning_rate": 5.47635850388144e-05, "loss": 2.7294, "step": 57690 }, { "epoch": 4.52, "learning_rate": 5.4762800909589905e-05, "loss": 2.6775, "step": 57691 }, { "epoch": 4.52, "learning_rate": 5.476201678036541e-05, "loss": 2.75, "step": 57692 }, { "epoch": 4.52, "learning_rate": 5.476123265114091e-05, "loss": 2.6934, "step": 57693 }, { "epoch": 4.52, "learning_rate": 5.4760448521916416e-05, "loss": 2.6046, "step": 57694 }, { "epoch": 4.52, "learning_rate": 5.475966439269192e-05, "loss": 2.5403, "step": 57695 }, { "epoch": 4.52, "learning_rate": 5.475888026346743e-05, "loss": 2.5808, "step": 57696 }, { "epoch": 4.52, "learning_rate": 5.4758096134242934e-05, "loss": 2.7456, "step": 57697 }, { "epoch": 4.52, "learning_rate": 5.475731200501844e-05, "loss": 2.4869, "step": 57698 }, { "epoch": 4.52, "learning_rate": 5.475652787579393e-05, "loss": 2.8361, "step": 57699 }, { "epoch": 4.52, "learning_rate": 5.475574374656943e-05, "loss": 2.7012, "step": 57700 }, { "epoch": 4.52, "learning_rate": 5.4754959617344935e-05, "loss": 2.612, "step": 57701 }, { "epoch": 4.52, "learning_rate": 5.475417548812044e-05, "loss": 2.7412, "step": 57702 }, { "epoch": 4.52, "learning_rate": 5.475339135889594e-05, "loss": 2.4998, "step": 57703 }, { "epoch": 4.52, "learning_rate": 5.4752607229671446e-05, "loss": 2.6862, "step": 57704 }, { "epoch": 4.52, "learning_rate": 5.475182310044695e-05, "loss": 2.5174, "step": 57705 }, { "epoch": 4.52, "learning_rate": 5.475103897122246e-05, "loss": 2.657, "step": 57706 }, { "epoch": 4.52, "learning_rate": 5.4750254841997964e-05, "loss": 2.5652, "step": 57707 }, { "epoch": 4.53, "learning_rate": 5.474947071277347e-05, "loss": 2.5505, "step": 57708 }, { "epoch": 4.53, "learning_rate": 5.474868658354897e-05, "loss": 2.6731, "step": 57709 }, { "epoch": 4.53, "learning_rate": 5.4747902454324475e-05, "loss": 2.5754, "step": 57710 }, { "epoch": 4.53, "learning_rate": 5.474711832509998e-05, "loss": 2.7668, "step": 57711 }, { "epoch": 4.53, "learning_rate": 5.474633419587548e-05, "loss": 2.5969, "step": 57712 }, { "epoch": 4.53, "learning_rate": 5.4745550066650986e-05, "loss": 2.7996, "step": 57713 }, { "epoch": 4.53, "learning_rate": 5.474476593742649e-05, "loss": 2.7398, "step": 57714 }, { "epoch": 4.53, "learning_rate": 5.474398180820199e-05, "loss": 2.4994, "step": 57715 }, { "epoch": 4.53, "learning_rate": 5.47431976789775e-05, "loss": 2.6857, "step": 57716 }, { "epoch": 4.53, "learning_rate": 5.4742413549753e-05, "loss": 2.619, "step": 57717 }, { "epoch": 4.53, "learning_rate": 5.4741629420528504e-05, "loss": 2.7792, "step": 57718 }, { "epoch": 4.53, "learning_rate": 5.474084529130401e-05, "loss": 2.6259, "step": 57719 }, { "epoch": 4.53, "learning_rate": 5.474006116207951e-05, "loss": 2.5811, "step": 57720 }, { "epoch": 4.53, "learning_rate": 5.4739277032855015e-05, "loss": 2.7126, "step": 57721 }, { "epoch": 4.53, "learning_rate": 5.4738492903630526e-05, "loss": 2.5722, "step": 57722 }, { "epoch": 4.53, "learning_rate": 5.473770877440603e-05, "loss": 2.6536, "step": 57723 }, { "epoch": 4.53, "learning_rate": 5.473692464518153e-05, "loss": 2.8167, "step": 57724 }, { "epoch": 4.53, "learning_rate": 5.473614051595704e-05, "loss": 2.5183, "step": 57725 }, { "epoch": 4.53, "learning_rate": 5.473535638673254e-05, "loss": 2.9314, "step": 57726 }, { "epoch": 4.53, "learning_rate": 5.4734572257508044e-05, "loss": 2.63, "step": 57727 }, { "epoch": 4.53, "learning_rate": 5.473378812828355e-05, "loss": 2.6625, "step": 57728 }, { "epoch": 4.53, "learning_rate": 5.473300399905905e-05, "loss": 2.5572, "step": 57729 }, { "epoch": 4.53, "learning_rate": 5.4732219869834555e-05, "loss": 2.7855, "step": 57730 }, { "epoch": 4.53, "learning_rate": 5.473143574061006e-05, "loss": 2.7336, "step": 57731 }, { "epoch": 4.53, "learning_rate": 5.473065161138556e-05, "loss": 2.8175, "step": 57732 }, { "epoch": 4.53, "learning_rate": 5.4729867482161066e-05, "loss": 2.6183, "step": 57733 }, { "epoch": 4.53, "learning_rate": 5.472908335293657e-05, "loss": 2.6994, "step": 57734 }, { "epoch": 4.53, "learning_rate": 5.472829922371207e-05, "loss": 2.5995, "step": 57735 }, { "epoch": 4.53, "learning_rate": 5.472751509448757e-05, "loss": 2.705, "step": 57736 }, { "epoch": 4.53, "learning_rate": 5.4726730965263074e-05, "loss": 2.6032, "step": 57737 }, { "epoch": 4.53, "learning_rate": 5.472594683603858e-05, "loss": 2.8075, "step": 57738 }, { "epoch": 4.53, "learning_rate": 5.472516270681408e-05, "loss": 2.6198, "step": 57739 }, { "epoch": 4.53, "learning_rate": 5.4724378577589585e-05, "loss": 2.4918, "step": 57740 }, { "epoch": 4.53, "learning_rate": 5.472359444836509e-05, "loss": 2.6289, "step": 57741 }, { "epoch": 4.53, "learning_rate": 5.472281031914059e-05, "loss": 2.6145, "step": 57742 }, { "epoch": 4.53, "learning_rate": 5.4722026189916096e-05, "loss": 2.6451, "step": 57743 }, { "epoch": 4.53, "learning_rate": 5.47212420606916e-05, "loss": 2.5023, "step": 57744 }, { "epoch": 4.53, "learning_rate": 5.4720457931467104e-05, "loss": 2.5305, "step": 57745 }, { "epoch": 4.53, "learning_rate": 5.471967380224261e-05, "loss": 2.5039, "step": 57746 }, { "epoch": 4.53, "learning_rate": 5.471888967301811e-05, "loss": 2.5887, "step": 57747 }, { "epoch": 4.53, "learning_rate": 5.471810554379362e-05, "loss": 2.5388, "step": 57748 }, { "epoch": 4.53, "learning_rate": 5.4717321414569125e-05, "loss": 2.5892, "step": 57749 }, { "epoch": 4.53, "learning_rate": 5.471653728534463e-05, "loss": 2.6436, "step": 57750 }, { "epoch": 4.53, "learning_rate": 5.471575315612013e-05, "loss": 2.7484, "step": 57751 }, { "epoch": 4.53, "learning_rate": 5.4714969026895636e-05, "loss": 2.5274, "step": 57752 }, { "epoch": 4.53, "learning_rate": 5.471418489767114e-05, "loss": 2.552, "step": 57753 }, { "epoch": 4.53, "learning_rate": 5.4713400768446644e-05, "loss": 2.7391, "step": 57754 }, { "epoch": 4.53, "learning_rate": 5.471261663922215e-05, "loss": 2.694, "step": 57755 }, { "epoch": 4.53, "learning_rate": 5.471183250999765e-05, "loss": 2.5901, "step": 57756 }, { "epoch": 4.53, "learning_rate": 5.4711048380773155e-05, "loss": 2.7565, "step": 57757 }, { "epoch": 4.53, "learning_rate": 5.471026425154866e-05, "loss": 2.5567, "step": 57758 }, { "epoch": 4.53, "learning_rate": 5.470948012232416e-05, "loss": 2.538, "step": 57759 }, { "epoch": 4.53, "learning_rate": 5.4708695993099666e-05, "loss": 2.7687, "step": 57760 }, { "epoch": 4.53, "learning_rate": 5.470791186387517e-05, "loss": 2.673, "step": 57761 }, { "epoch": 4.53, "learning_rate": 5.470712773465067e-05, "loss": 2.7328, "step": 57762 }, { "epoch": 4.53, "learning_rate": 5.4706343605426183e-05, "loss": 2.7287, "step": 57763 }, { "epoch": 4.53, "learning_rate": 5.470555947620169e-05, "loss": 2.5439, "step": 57764 }, { "epoch": 4.53, "learning_rate": 5.470477534697719e-05, "loss": 2.7601, "step": 57765 }, { "epoch": 4.53, "learning_rate": 5.4703991217752694e-05, "loss": 2.5726, "step": 57766 }, { "epoch": 4.53, "learning_rate": 5.47032070885282e-05, "loss": 2.6291, "step": 57767 }, { "epoch": 4.53, "learning_rate": 5.47024229593037e-05, "loss": 2.6736, "step": 57768 }, { "epoch": 4.53, "learning_rate": 5.4701638830079206e-05, "loss": 2.5811, "step": 57769 }, { "epoch": 4.53, "learning_rate": 5.470085470085471e-05, "loss": 2.5846, "step": 57770 }, { "epoch": 4.53, "learning_rate": 5.47000705716302e-05, "loss": 2.7332, "step": 57771 }, { "epoch": 4.53, "learning_rate": 5.46992864424057e-05, "loss": 2.7445, "step": 57772 }, { "epoch": 4.53, "learning_rate": 5.4698502313181213e-05, "loss": 2.6216, "step": 57773 }, { "epoch": 4.53, "learning_rate": 5.469771818395672e-05, "loss": 2.852, "step": 57774 }, { "epoch": 4.53, "learning_rate": 5.469693405473222e-05, "loss": 2.7512, "step": 57775 }, { "epoch": 4.53, "learning_rate": 5.4696149925507724e-05, "loss": 2.6151, "step": 57776 }, { "epoch": 4.53, "learning_rate": 5.469536579628323e-05, "loss": 2.8239, "step": 57777 }, { "epoch": 4.53, "learning_rate": 5.469458166705873e-05, "loss": 2.6768, "step": 57778 }, { "epoch": 4.53, "learning_rate": 5.4693797537834236e-05, "loss": 2.6683, "step": 57779 }, { "epoch": 4.53, "learning_rate": 5.469301340860974e-05, "loss": 2.7087, "step": 57780 }, { "epoch": 4.53, "learning_rate": 5.469222927938524e-05, "loss": 2.5968, "step": 57781 }, { "epoch": 4.53, "learning_rate": 5.4691445150160747e-05, "loss": 2.6635, "step": 57782 }, { "epoch": 4.53, "learning_rate": 5.469066102093625e-05, "loss": 2.7633, "step": 57783 }, { "epoch": 4.53, "learning_rate": 5.4689876891711754e-05, "loss": 2.726, "step": 57784 }, { "epoch": 4.53, "learning_rate": 5.468909276248726e-05, "loss": 2.6304, "step": 57785 }, { "epoch": 4.53, "learning_rate": 5.468830863326276e-05, "loss": 2.6022, "step": 57786 }, { "epoch": 4.53, "learning_rate": 5.4687524504038265e-05, "loss": 2.6931, "step": 57787 }, { "epoch": 4.53, "learning_rate": 5.468674037481377e-05, "loss": 2.6004, "step": 57788 }, { "epoch": 4.53, "learning_rate": 5.468595624558928e-05, "loss": 2.5876, "step": 57789 }, { "epoch": 4.53, "learning_rate": 5.468517211636478e-05, "loss": 2.8954, "step": 57790 }, { "epoch": 4.53, "learning_rate": 5.4684387987140286e-05, "loss": 2.8627, "step": 57791 }, { "epoch": 4.53, "learning_rate": 5.468360385791579e-05, "loss": 2.7924, "step": 57792 }, { "epoch": 4.53, "learning_rate": 5.4682819728691294e-05, "loss": 2.7408, "step": 57793 }, { "epoch": 4.53, "learning_rate": 5.46820355994668e-05, "loss": 2.5786, "step": 57794 }, { "epoch": 4.53, "learning_rate": 5.46812514702423e-05, "loss": 2.659, "step": 57795 }, { "epoch": 4.53, "learning_rate": 5.4680467341017805e-05, "loss": 2.7056, "step": 57796 }, { "epoch": 4.53, "learning_rate": 5.467968321179331e-05, "loss": 2.4251, "step": 57797 }, { "epoch": 4.53, "learning_rate": 5.467889908256881e-05, "loss": 2.5208, "step": 57798 }, { "epoch": 4.53, "learning_rate": 5.4678114953344316e-05, "loss": 2.7052, "step": 57799 }, { "epoch": 4.53, "learning_rate": 5.467733082411982e-05, "loss": 2.7501, "step": 57800 }, { "epoch": 4.53, "learning_rate": 5.467654669489532e-05, "loss": 2.7362, "step": 57801 }, { "epoch": 4.53, "learning_rate": 5.467576256567083e-05, "loss": 2.8575, "step": 57802 }, { "epoch": 4.53, "learning_rate": 5.467497843644633e-05, "loss": 2.6466, "step": 57803 }, { "epoch": 4.53, "learning_rate": 5.467419430722184e-05, "loss": 2.5436, "step": 57804 }, { "epoch": 4.53, "learning_rate": 5.4673410177997345e-05, "loss": 2.4614, "step": 57805 }, { "epoch": 4.53, "learning_rate": 5.467262604877285e-05, "loss": 2.7106, "step": 57806 }, { "epoch": 4.53, "learning_rate": 5.467184191954834e-05, "loss": 2.6753, "step": 57807 }, { "epoch": 4.53, "learning_rate": 5.467105779032384e-05, "loss": 2.6642, "step": 57808 }, { "epoch": 4.53, "learning_rate": 5.4670273661099346e-05, "loss": 2.5656, "step": 57809 }, { "epoch": 4.53, "learning_rate": 5.466948953187485e-05, "loss": 2.8657, "step": 57810 }, { "epoch": 4.53, "learning_rate": 5.466870540265035e-05, "loss": 2.532, "step": 57811 }, { "epoch": 4.53, "learning_rate": 5.466792127342586e-05, "loss": 2.7214, "step": 57812 }, { "epoch": 4.53, "learning_rate": 5.466713714420136e-05, "loss": 2.6578, "step": 57813 }, { "epoch": 4.53, "learning_rate": 5.466635301497687e-05, "loss": 3.0494, "step": 57814 }, { "epoch": 4.53, "learning_rate": 5.4665568885752375e-05, "loss": 2.5066, "step": 57815 }, { "epoch": 4.53, "learning_rate": 5.466478475652788e-05, "loss": 2.8113, "step": 57816 }, { "epoch": 4.53, "learning_rate": 5.466400062730338e-05, "loss": 2.6476, "step": 57817 }, { "epoch": 4.53, "learning_rate": 5.4663216498078886e-05, "loss": 2.7882, "step": 57818 }, { "epoch": 4.53, "learning_rate": 5.466243236885439e-05, "loss": 2.7938, "step": 57819 }, { "epoch": 4.53, "learning_rate": 5.466164823962989e-05, "loss": 2.6903, "step": 57820 }, { "epoch": 4.53, "learning_rate": 5.46608641104054e-05, "loss": 2.6012, "step": 57821 }, { "epoch": 4.53, "learning_rate": 5.46600799811809e-05, "loss": 2.927, "step": 57822 }, { "epoch": 4.53, "learning_rate": 5.4659295851956404e-05, "loss": 2.8479, "step": 57823 }, { "epoch": 4.53, "learning_rate": 5.465851172273191e-05, "loss": 2.4913, "step": 57824 }, { "epoch": 4.53, "learning_rate": 5.465772759350741e-05, "loss": 2.4984, "step": 57825 }, { "epoch": 4.53, "learning_rate": 5.4656943464282915e-05, "loss": 2.503, "step": 57826 }, { "epoch": 4.53, "learning_rate": 5.465615933505842e-05, "loss": 2.6952, "step": 57827 }, { "epoch": 4.53, "learning_rate": 5.465537520583392e-05, "loss": 2.6871, "step": 57828 }, { "epoch": 4.53, "learning_rate": 5.4654591076609426e-05, "loss": 2.4982, "step": 57829 }, { "epoch": 4.53, "learning_rate": 5.465380694738494e-05, "loss": 2.9356, "step": 57830 }, { "epoch": 4.53, "learning_rate": 5.465302281816044e-05, "loss": 2.7157, "step": 57831 }, { "epoch": 4.53, "learning_rate": 5.4652238688935944e-05, "loss": 2.7013, "step": 57832 }, { "epoch": 4.53, "learning_rate": 5.465145455971145e-05, "loss": 2.617, "step": 57833 }, { "epoch": 4.53, "learning_rate": 5.465067043048695e-05, "loss": 2.6418, "step": 57834 }, { "epoch": 4.54, "learning_rate": 5.4649886301262455e-05, "loss": 2.7291, "step": 57835 }, { "epoch": 4.54, "learning_rate": 5.464910217203796e-05, "loss": 2.7043, "step": 57836 }, { "epoch": 4.54, "learning_rate": 5.464831804281346e-05, "loss": 2.6594, "step": 57837 }, { "epoch": 4.54, "learning_rate": 5.4647533913588966e-05, "loss": 2.5348, "step": 57838 }, { "epoch": 4.54, "learning_rate": 5.464674978436447e-05, "loss": 2.6268, "step": 57839 }, { "epoch": 4.54, "learning_rate": 5.4645965655139973e-05, "loss": 2.9001, "step": 57840 }, { "epoch": 4.54, "learning_rate": 5.464518152591548e-05, "loss": 2.5166, "step": 57841 }, { "epoch": 4.54, "learning_rate": 5.464439739669098e-05, "loss": 2.6383, "step": 57842 }, { "epoch": 4.54, "learning_rate": 5.464361326746648e-05, "loss": 2.746, "step": 57843 }, { "epoch": 4.54, "learning_rate": 5.464282913824198e-05, "loss": 2.8947, "step": 57844 }, { "epoch": 4.54, "learning_rate": 5.4642045009017485e-05, "loss": 2.5998, "step": 57845 }, { "epoch": 4.54, "learning_rate": 5.464126087979299e-05, "loss": 2.6177, "step": 57846 }, { "epoch": 4.54, "learning_rate": 5.464047675056849e-05, "loss": 2.6471, "step": 57847 }, { "epoch": 4.54, "learning_rate": 5.4639692621343996e-05, "loss": 2.6849, "step": 57848 }, { "epoch": 4.54, "learning_rate": 5.46389084921195e-05, "loss": 2.6132, "step": 57849 }, { "epoch": 4.54, "learning_rate": 5.4638124362895003e-05, "loss": 2.5491, "step": 57850 }, { "epoch": 4.54, "learning_rate": 5.463734023367051e-05, "loss": 2.6069, "step": 57851 }, { "epoch": 4.54, "learning_rate": 5.463655610444601e-05, "loss": 2.7217, "step": 57852 }, { "epoch": 4.54, "learning_rate": 5.4635771975221514e-05, "loss": 2.6563, "step": 57853 }, { "epoch": 4.54, "learning_rate": 5.463498784599702e-05, "loss": 2.6733, "step": 57854 }, { "epoch": 4.54, "learning_rate": 5.463420371677252e-05, "loss": 2.5264, "step": 57855 }, { "epoch": 4.54, "learning_rate": 5.463341958754803e-05, "loss": 2.3995, "step": 57856 }, { "epoch": 4.54, "learning_rate": 5.4632635458323536e-05, "loss": 2.8622, "step": 57857 }, { "epoch": 4.54, "learning_rate": 5.463185132909904e-05, "loss": 2.8536, "step": 57858 }, { "epoch": 4.54, "learning_rate": 5.463106719987454e-05, "loss": 2.5168, "step": 57859 }, { "epoch": 4.54, "learning_rate": 5.463028307065005e-05, "loss": 2.7312, "step": 57860 }, { "epoch": 4.54, "learning_rate": 5.462949894142555e-05, "loss": 2.5134, "step": 57861 }, { "epoch": 4.54, "learning_rate": 5.4628714812201054e-05, "loss": 2.5729, "step": 57862 }, { "epoch": 4.54, "learning_rate": 5.462793068297656e-05, "loss": 2.712, "step": 57863 }, { "epoch": 4.54, "learning_rate": 5.462714655375206e-05, "loss": 2.6111, "step": 57864 }, { "epoch": 4.54, "learning_rate": 5.4626362424527565e-05, "loss": 2.7268, "step": 57865 }, { "epoch": 4.54, "learning_rate": 5.462557829530307e-05, "loss": 2.7438, "step": 57866 }, { "epoch": 4.54, "learning_rate": 5.462479416607857e-05, "loss": 2.4425, "step": 57867 }, { "epoch": 4.54, "learning_rate": 5.4624010036854076e-05, "loss": 2.5224, "step": 57868 }, { "epoch": 4.54, "learning_rate": 5.462322590762958e-05, "loss": 2.641, "step": 57869 }, { "epoch": 4.54, "learning_rate": 5.4622441778405084e-05, "loss": 2.5308, "step": 57870 }, { "epoch": 4.54, "learning_rate": 5.4621657649180594e-05, "loss": 2.7364, "step": 57871 }, { "epoch": 4.54, "learning_rate": 5.46208735199561e-05, "loss": 2.5398, "step": 57872 }, { "epoch": 4.54, "learning_rate": 5.46200893907316e-05, "loss": 2.7369, "step": 57873 }, { "epoch": 4.54, "learning_rate": 5.4619305261507105e-05, "loss": 2.5693, "step": 57874 }, { "epoch": 4.54, "learning_rate": 5.461852113228261e-05, "loss": 2.7578, "step": 57875 }, { "epoch": 4.54, "learning_rate": 5.461773700305811e-05, "loss": 2.4763, "step": 57876 }, { "epoch": 4.54, "learning_rate": 5.4616952873833616e-05, "loss": 2.8431, "step": 57877 }, { "epoch": 4.54, "learning_rate": 5.461616874460912e-05, "loss": 3.1085, "step": 57878 }, { "epoch": 4.54, "learning_rate": 5.461538461538461e-05, "loss": 2.6168, "step": 57879 }, { "epoch": 4.54, "learning_rate": 5.4614600486160114e-05, "loss": 2.7526, "step": 57880 }, { "epoch": 4.54, "learning_rate": 5.4613816356935624e-05, "loss": 2.566, "step": 57881 }, { "epoch": 4.54, "learning_rate": 5.461303222771113e-05, "loss": 2.7472, "step": 57882 }, { "epoch": 4.54, "learning_rate": 5.461224809848663e-05, "loss": 2.5967, "step": 57883 }, { "epoch": 4.54, "learning_rate": 5.4611463969262135e-05, "loss": 2.6596, "step": 57884 }, { "epoch": 4.54, "learning_rate": 5.461067984003764e-05, "loss": 2.8025, "step": 57885 }, { "epoch": 4.54, "learning_rate": 5.460989571081314e-05, "loss": 2.4946, "step": 57886 }, { "epoch": 4.54, "learning_rate": 5.4609111581588646e-05, "loss": 2.6828, "step": 57887 }, { "epoch": 4.54, "learning_rate": 5.460832745236415e-05, "loss": 2.8033, "step": 57888 }, { "epoch": 4.54, "learning_rate": 5.4607543323139654e-05, "loss": 2.5891, "step": 57889 }, { "epoch": 4.54, "learning_rate": 5.460675919391516e-05, "loss": 2.5862, "step": 57890 }, { "epoch": 4.54, "learning_rate": 5.460597506469066e-05, "loss": 2.4853, "step": 57891 }, { "epoch": 4.54, "learning_rate": 5.4605190935466165e-05, "loss": 2.6754, "step": 57892 }, { "epoch": 4.54, "learning_rate": 5.460440680624167e-05, "loss": 2.6982, "step": 57893 }, { "epoch": 4.54, "learning_rate": 5.460362267701717e-05, "loss": 2.4941, "step": 57894 }, { "epoch": 4.54, "learning_rate": 5.4602838547792676e-05, "loss": 2.6697, "step": 57895 }, { "epoch": 4.54, "learning_rate": 5.460205441856818e-05, "loss": 2.6217, "step": 57896 }, { "epoch": 4.54, "learning_rate": 5.460127028934369e-05, "loss": 2.7267, "step": 57897 }, { "epoch": 4.54, "learning_rate": 5.4600486160119194e-05, "loss": 2.5713, "step": 57898 }, { "epoch": 4.54, "learning_rate": 5.45997020308947e-05, "loss": 2.7007, "step": 57899 }, { "epoch": 4.54, "learning_rate": 5.45989179016702e-05, "loss": 2.7395, "step": 57900 }, { "epoch": 4.54, "learning_rate": 5.4598133772445705e-05, "loss": 2.5514, "step": 57901 }, { "epoch": 4.54, "learning_rate": 5.459734964322121e-05, "loss": 2.939, "step": 57902 }, { "epoch": 4.54, "learning_rate": 5.459656551399671e-05, "loss": 2.7035, "step": 57903 }, { "epoch": 4.54, "learning_rate": 5.4595781384772216e-05, "loss": 2.5773, "step": 57904 }, { "epoch": 4.54, "learning_rate": 5.459499725554772e-05, "loss": 2.7447, "step": 57905 }, { "epoch": 4.54, "learning_rate": 5.459421312632322e-05, "loss": 2.6091, "step": 57906 }, { "epoch": 4.54, "learning_rate": 5.4593428997098727e-05, "loss": 2.6863, "step": 57907 }, { "epoch": 4.54, "learning_rate": 5.459264486787423e-05, "loss": 2.7465, "step": 57908 }, { "epoch": 4.54, "learning_rate": 5.4591860738649734e-05, "loss": 2.8063, "step": 57909 }, { "epoch": 4.54, "learning_rate": 5.459107660942524e-05, "loss": 2.777, "step": 57910 }, { "epoch": 4.54, "learning_rate": 5.459029248020074e-05, "loss": 2.4669, "step": 57911 }, { "epoch": 4.54, "learning_rate": 5.4589508350976245e-05, "loss": 2.6704, "step": 57912 }, { "epoch": 4.54, "learning_rate": 5.4588724221751755e-05, "loss": 2.8524, "step": 57913 }, { "epoch": 4.54, "learning_rate": 5.458794009252726e-05, "loss": 2.689, "step": 57914 }, { "epoch": 4.54, "learning_rate": 5.458715596330275e-05, "loss": 2.9124, "step": 57915 }, { "epoch": 4.54, "learning_rate": 5.458637183407825e-05, "loss": 2.7439, "step": 57916 }, { "epoch": 4.54, "learning_rate": 5.4585587704853757e-05, "loss": 2.7297, "step": 57917 }, { "epoch": 4.54, "learning_rate": 5.458480357562926e-05, "loss": 2.6432, "step": 57918 }, { "epoch": 4.54, "learning_rate": 5.4584019446404764e-05, "loss": 2.8961, "step": 57919 }, { "epoch": 4.54, "learning_rate": 5.458323531718027e-05, "loss": 2.7729, "step": 57920 }, { "epoch": 4.54, "learning_rate": 5.458245118795577e-05, "loss": 2.5321, "step": 57921 }, { "epoch": 4.54, "learning_rate": 5.458166705873128e-05, "loss": 2.7414, "step": 57922 }, { "epoch": 4.54, "learning_rate": 5.4580882929506785e-05, "loss": 2.579, "step": 57923 }, { "epoch": 4.54, "learning_rate": 5.458009880028229e-05, "loss": 2.7893, "step": 57924 }, { "epoch": 4.54, "learning_rate": 5.457931467105779e-05, "loss": 2.6298, "step": 57925 }, { "epoch": 4.54, "learning_rate": 5.4578530541833296e-05, "loss": 2.8168, "step": 57926 }, { "epoch": 4.54, "learning_rate": 5.45777464126088e-05, "loss": 2.7453, "step": 57927 }, { "epoch": 4.54, "learning_rate": 5.4576962283384304e-05, "loss": 2.5081, "step": 57928 }, { "epoch": 4.54, "learning_rate": 5.457617815415981e-05, "loss": 2.7195, "step": 57929 }, { "epoch": 4.54, "learning_rate": 5.457539402493531e-05, "loss": 2.594, "step": 57930 }, { "epoch": 4.54, "learning_rate": 5.4574609895710815e-05, "loss": 2.7218, "step": 57931 }, { "epoch": 4.54, "learning_rate": 5.457382576648632e-05, "loss": 2.5173, "step": 57932 }, { "epoch": 4.54, "learning_rate": 5.457304163726182e-05, "loss": 2.7391, "step": 57933 }, { "epoch": 4.54, "learning_rate": 5.4572257508037326e-05, "loss": 2.7369, "step": 57934 }, { "epoch": 4.54, "learning_rate": 5.457147337881283e-05, "loss": 2.5737, "step": 57935 }, { "epoch": 4.54, "learning_rate": 5.457068924958833e-05, "loss": 2.4562, "step": 57936 }, { "epoch": 4.54, "learning_rate": 5.456990512036384e-05, "loss": 2.667, "step": 57937 }, { "epoch": 4.54, "learning_rate": 5.456912099113935e-05, "loss": 2.7028, "step": 57938 }, { "epoch": 4.54, "learning_rate": 5.456833686191485e-05, "loss": 2.5802, "step": 57939 }, { "epoch": 4.54, "learning_rate": 5.4567552732690355e-05, "loss": 2.659, "step": 57940 }, { "epoch": 4.54, "learning_rate": 5.456676860346586e-05, "loss": 2.7006, "step": 57941 }, { "epoch": 4.54, "learning_rate": 5.456598447424136e-05, "loss": 2.6752, "step": 57942 }, { "epoch": 4.54, "learning_rate": 5.4565200345016866e-05, "loss": 2.8649, "step": 57943 }, { "epoch": 4.54, "learning_rate": 5.456441621579237e-05, "loss": 2.6665, "step": 57944 }, { "epoch": 4.54, "learning_rate": 5.456363208656787e-05, "loss": 2.6345, "step": 57945 }, { "epoch": 4.54, "learning_rate": 5.456284795734338e-05, "loss": 2.5646, "step": 57946 }, { "epoch": 4.54, "learning_rate": 5.456206382811888e-05, "loss": 2.6266, "step": 57947 }, { "epoch": 4.54, "learning_rate": 5.4561279698894384e-05, "loss": 2.499, "step": 57948 }, { "epoch": 4.54, "learning_rate": 5.456049556966989e-05, "loss": 2.6023, "step": 57949 }, { "epoch": 4.54, "learning_rate": 5.455971144044539e-05, "loss": 2.7131, "step": 57950 }, { "epoch": 4.54, "learning_rate": 5.455892731122089e-05, "loss": 2.5469, "step": 57951 }, { "epoch": 4.54, "learning_rate": 5.455814318199639e-05, "loss": 2.7546, "step": 57952 }, { "epoch": 4.54, "learning_rate": 5.4557359052771896e-05, "loss": 2.6347, "step": 57953 }, { "epoch": 4.54, "learning_rate": 5.45565749235474e-05, "loss": 2.5773, "step": 57954 }, { "epoch": 4.54, "learning_rate": 5.45557907943229e-05, "loss": 2.9411, "step": 57955 }, { "epoch": 4.54, "learning_rate": 5.455500666509841e-05, "loss": 2.7396, "step": 57956 }, { "epoch": 4.54, "learning_rate": 5.455422253587391e-05, "loss": 2.8835, "step": 57957 }, { "epoch": 4.54, "learning_rate": 5.4553438406649414e-05, "loss": 2.6365, "step": 57958 }, { "epoch": 4.54, "learning_rate": 5.455265427742492e-05, "loss": 2.6926, "step": 57959 }, { "epoch": 4.54, "learning_rate": 5.455187014820042e-05, "loss": 2.6882, "step": 57960 }, { "epoch": 4.54, "learning_rate": 5.4551086018975925e-05, "loss": 2.6271, "step": 57961 }, { "epoch": 4.54, "learning_rate": 5.455030188975143e-05, "loss": 2.5798, "step": 57962 }, { "epoch": 4.55, "learning_rate": 5.454951776052693e-05, "loss": 2.706, "step": 57963 }, { "epoch": 4.55, "learning_rate": 5.454873363130244e-05, "loss": 2.8616, "step": 57964 }, { "epoch": 4.55, "learning_rate": 5.454794950207795e-05, "loss": 2.6297, "step": 57965 }, { "epoch": 4.55, "learning_rate": 5.454716537285345e-05, "loss": 2.8234, "step": 57966 }, { "epoch": 4.55, "learning_rate": 5.4546381243628954e-05, "loss": 2.7469, "step": 57967 }, { "epoch": 4.55, "learning_rate": 5.454559711440446e-05, "loss": 2.5692, "step": 57968 }, { "epoch": 4.55, "learning_rate": 5.454481298517996e-05, "loss": 2.751, "step": 57969 }, { "epoch": 4.55, "learning_rate": 5.4544028855955465e-05, "loss": 2.792, "step": 57970 }, { "epoch": 4.55, "learning_rate": 5.454324472673097e-05, "loss": 2.4052, "step": 57971 }, { "epoch": 4.55, "learning_rate": 5.454246059750647e-05, "loss": 2.7754, "step": 57972 }, { "epoch": 4.55, "learning_rate": 5.4541676468281976e-05, "loss": 2.5597, "step": 57973 }, { "epoch": 4.55, "learning_rate": 5.454089233905748e-05, "loss": 2.4181, "step": 57974 }, { "epoch": 4.55, "learning_rate": 5.4540108209832983e-05, "loss": 2.6203, "step": 57975 }, { "epoch": 4.55, "learning_rate": 5.453932408060849e-05, "loss": 2.7292, "step": 57976 }, { "epoch": 4.55, "learning_rate": 5.453853995138399e-05, "loss": 2.4388, "step": 57977 }, { "epoch": 4.55, "learning_rate": 5.4537755822159494e-05, "loss": 2.6132, "step": 57978 }, { "epoch": 4.55, "learning_rate": 5.4536971692935005e-05, "loss": 2.6876, "step": 57979 }, { "epoch": 4.55, "learning_rate": 5.453618756371051e-05, "loss": 2.9316, "step": 57980 }, { "epoch": 4.55, "learning_rate": 5.453540343448601e-05, "loss": 2.6332, "step": 57981 }, { "epoch": 4.55, "learning_rate": 5.4534619305261516e-05, "loss": 2.708, "step": 57982 }, { "epoch": 4.55, "learning_rate": 5.453383517603702e-05, "loss": 2.7362, "step": 57983 }, { "epoch": 4.55, "learning_rate": 5.453305104681252e-05, "loss": 2.6443, "step": 57984 }, { "epoch": 4.55, "learning_rate": 5.453226691758803e-05, "loss": 2.7147, "step": 57985 }, { "epoch": 4.55, "learning_rate": 5.453148278836353e-05, "loss": 2.7777, "step": 57986 }, { "epoch": 4.55, "learning_rate": 5.453069865913902e-05, "loss": 2.5609, "step": 57987 }, { "epoch": 4.55, "learning_rate": 5.4529914529914525e-05, "loss": 2.6912, "step": 57988 }, { "epoch": 4.55, "learning_rate": 5.4529130400690035e-05, "loss": 2.701, "step": 57989 }, { "epoch": 4.55, "learning_rate": 5.452834627146554e-05, "loss": 2.593, "step": 57990 }, { "epoch": 4.55, "learning_rate": 5.452756214224104e-05, "loss": 2.7549, "step": 57991 }, { "epoch": 4.55, "learning_rate": 5.4526778013016546e-05, "loss": 2.4886, "step": 57992 }, { "epoch": 4.55, "learning_rate": 5.452599388379205e-05, "loss": 2.5449, "step": 57993 }, { "epoch": 4.55, "learning_rate": 5.452520975456755e-05, "loss": 2.4873, "step": 57994 }, { "epoch": 4.55, "learning_rate": 5.452442562534306e-05, "loss": 2.8006, "step": 57995 }, { "epoch": 4.55, "learning_rate": 5.452364149611856e-05, "loss": 2.8749, "step": 57996 }, { "epoch": 4.55, "learning_rate": 5.4522857366894064e-05, "loss": 2.7289, "step": 57997 }, { "epoch": 4.55, "learning_rate": 5.452207323766957e-05, "loss": 2.5925, "step": 57998 }, { "epoch": 4.55, "learning_rate": 5.452128910844507e-05, "loss": 2.8463, "step": 57999 }, { "epoch": 4.55, "learning_rate": 5.4520504979220575e-05, "loss": 2.7972, "step": 58000 }, { "epoch": 4.55, "eval_loss": 2.896052122116089, "eval_runtime": 217.6504, "eval_samples_per_second": 52.065, "eval_steps_per_second": 3.258, "step": 58000 }, { "epoch": 4.55, "learning_rate": 5.451972084999608e-05, "loss": 2.5963, "step": 58001 }, { "epoch": 4.55, "learning_rate": 5.451893672077158e-05, "loss": 2.605, "step": 58002 }, { "epoch": 4.55, "learning_rate": 5.4518152591547086e-05, "loss": 2.8381, "step": 58003 }, { "epoch": 4.55, "learning_rate": 5.451736846232259e-05, "loss": 2.6435, "step": 58004 }, { "epoch": 4.55, "learning_rate": 5.45165843330981e-05, "loss": 2.5687, "step": 58005 }, { "epoch": 4.55, "learning_rate": 5.4515800203873604e-05, "loss": 2.4614, "step": 58006 }, { "epoch": 4.55, "learning_rate": 5.451501607464911e-05, "loss": 2.4699, "step": 58007 }, { "epoch": 4.55, "learning_rate": 5.451423194542461e-05, "loss": 2.8666, "step": 58008 }, { "epoch": 4.55, "learning_rate": 5.4513447816200115e-05, "loss": 2.5399, "step": 58009 }, { "epoch": 4.55, "learning_rate": 5.451266368697562e-05, "loss": 2.8108, "step": 58010 }, { "epoch": 4.55, "learning_rate": 5.451187955775112e-05, "loss": 2.6962, "step": 58011 }, { "epoch": 4.55, "learning_rate": 5.4511095428526626e-05, "loss": 2.5613, "step": 58012 }, { "epoch": 4.55, "learning_rate": 5.451031129930213e-05, "loss": 2.8433, "step": 58013 }, { "epoch": 4.55, "learning_rate": 5.4509527170077634e-05, "loss": 2.3422, "step": 58014 }, { "epoch": 4.55, "learning_rate": 5.450874304085314e-05, "loss": 2.5946, "step": 58015 }, { "epoch": 4.55, "learning_rate": 5.450795891162864e-05, "loss": 2.4653, "step": 58016 }, { "epoch": 4.55, "learning_rate": 5.4507174782404145e-05, "loss": 2.6489, "step": 58017 }, { "epoch": 4.55, "learning_rate": 5.450639065317965e-05, "loss": 2.5773, "step": 58018 }, { "epoch": 4.55, "learning_rate": 5.450560652395515e-05, "loss": 2.7335, "step": 58019 }, { "epoch": 4.55, "learning_rate": 5.4504822394730656e-05, "loss": 2.5732, "step": 58020 }, { "epoch": 4.55, "learning_rate": 5.4504038265506166e-05, "loss": 2.6614, "step": 58021 }, { "epoch": 4.55, "learning_rate": 5.450325413628167e-05, "loss": 2.5777, "step": 58022 }, { "epoch": 4.55, "learning_rate": 5.450247000705716e-05, "loss": 2.5841, "step": 58023 }, { "epoch": 4.55, "learning_rate": 5.4501685877832664e-05, "loss": 2.4993, "step": 58024 }, { "epoch": 4.55, "learning_rate": 5.450090174860817e-05, "loss": 2.6299, "step": 58025 }, { "epoch": 4.55, "learning_rate": 5.450011761938367e-05, "loss": 2.636, "step": 58026 }, { "epoch": 4.55, "learning_rate": 5.4499333490159175e-05, "loss": 2.4033, "step": 58027 }, { "epoch": 4.55, "learning_rate": 5.449854936093468e-05, "loss": 2.6715, "step": 58028 }, { "epoch": 4.55, "learning_rate": 5.449776523171018e-05, "loss": 2.5633, "step": 58029 }, { "epoch": 4.55, "learning_rate": 5.4496981102485686e-05, "loss": 2.7531, "step": 58030 }, { "epoch": 4.55, "learning_rate": 5.4496196973261196e-05, "loss": 2.9378, "step": 58031 }, { "epoch": 4.55, "learning_rate": 5.44954128440367e-05, "loss": 2.4794, "step": 58032 }, { "epoch": 4.55, "learning_rate": 5.4494628714812204e-05, "loss": 2.7775, "step": 58033 }, { "epoch": 4.55, "learning_rate": 5.449384458558771e-05, "loss": 2.5683, "step": 58034 }, { "epoch": 4.55, "learning_rate": 5.449306045636321e-05, "loss": 2.7899, "step": 58035 }, { "epoch": 4.55, "learning_rate": 5.4492276327138715e-05, "loss": 2.9545, "step": 58036 }, { "epoch": 4.55, "learning_rate": 5.449149219791422e-05, "loss": 2.6095, "step": 58037 }, { "epoch": 4.55, "learning_rate": 5.449070806868972e-05, "loss": 2.6102, "step": 58038 }, { "epoch": 4.55, "learning_rate": 5.4489923939465226e-05, "loss": 2.7835, "step": 58039 }, { "epoch": 4.55, "learning_rate": 5.448913981024073e-05, "loss": 2.5608, "step": 58040 }, { "epoch": 4.55, "learning_rate": 5.448835568101623e-05, "loss": 2.5234, "step": 58041 }, { "epoch": 4.55, "learning_rate": 5.448757155179174e-05, "loss": 2.494, "step": 58042 }, { "epoch": 4.55, "learning_rate": 5.448678742256724e-05, "loss": 2.5707, "step": 58043 }, { "epoch": 4.55, "learning_rate": 5.4486003293342744e-05, "loss": 2.6534, "step": 58044 }, { "epoch": 4.55, "learning_rate": 5.448521916411825e-05, "loss": 2.6232, "step": 58045 }, { "epoch": 4.55, "learning_rate": 5.448443503489376e-05, "loss": 2.7067, "step": 58046 }, { "epoch": 4.55, "learning_rate": 5.448365090566926e-05, "loss": 2.5483, "step": 58047 }, { "epoch": 4.55, "learning_rate": 5.4482866776444765e-05, "loss": 2.7351, "step": 58048 }, { "epoch": 4.55, "learning_rate": 5.448208264722027e-05, "loss": 2.7839, "step": 58049 }, { "epoch": 4.55, "learning_rate": 5.448129851799577e-05, "loss": 2.7814, "step": 58050 }, { "epoch": 4.55, "learning_rate": 5.4480514388771277e-05, "loss": 2.639, "step": 58051 }, { "epoch": 4.55, "learning_rate": 5.447973025954678e-05, "loss": 2.9231, "step": 58052 }, { "epoch": 4.55, "learning_rate": 5.4478946130322284e-05, "loss": 2.4366, "step": 58053 }, { "epoch": 4.55, "learning_rate": 5.447816200109779e-05, "loss": 2.5313, "step": 58054 }, { "epoch": 4.55, "learning_rate": 5.447737787187329e-05, "loss": 2.6009, "step": 58055 }, { "epoch": 4.55, "learning_rate": 5.4476593742648795e-05, "loss": 2.6988, "step": 58056 }, { "epoch": 4.55, "learning_rate": 5.44758096134243e-05, "loss": 2.4864, "step": 58057 }, { "epoch": 4.55, "learning_rate": 5.44750254841998e-05, "loss": 2.6862, "step": 58058 }, { "epoch": 4.55, "learning_rate": 5.44742413549753e-05, "loss": 2.6718, "step": 58059 }, { "epoch": 4.55, "learning_rate": 5.44734572257508e-05, "loss": 2.5551, "step": 58060 }, { "epoch": 4.55, "learning_rate": 5.4472673096526307e-05, "loss": 2.8556, "step": 58061 }, { "epoch": 4.55, "learning_rate": 5.447188896730181e-05, "loss": 2.6638, "step": 58062 }, { "epoch": 4.55, "learning_rate": 5.4471104838077314e-05, "loss": 2.5432, "step": 58063 }, { "epoch": 4.55, "learning_rate": 5.447032070885282e-05, "loss": 2.6162, "step": 58064 }, { "epoch": 4.55, "learning_rate": 5.446953657962832e-05, "loss": 2.6911, "step": 58065 }, { "epoch": 4.55, "learning_rate": 5.4468752450403825e-05, "loss": 2.6564, "step": 58066 }, { "epoch": 4.55, "learning_rate": 5.446796832117933e-05, "loss": 2.7294, "step": 58067 }, { "epoch": 4.55, "learning_rate": 5.446718419195483e-05, "loss": 2.7414, "step": 58068 }, { "epoch": 4.55, "learning_rate": 5.4466400062730336e-05, "loss": 2.7047, "step": 58069 }, { "epoch": 4.55, "learning_rate": 5.446561593350584e-05, "loss": 2.8578, "step": 58070 }, { "epoch": 4.55, "learning_rate": 5.446483180428134e-05, "loss": 2.6616, "step": 58071 }, { "epoch": 4.55, "learning_rate": 5.4464047675056854e-05, "loss": 2.6278, "step": 58072 }, { "epoch": 4.55, "learning_rate": 5.446326354583236e-05, "loss": 2.4872, "step": 58073 }, { "epoch": 4.55, "learning_rate": 5.446247941660786e-05, "loss": 2.8021, "step": 58074 }, { "epoch": 4.55, "learning_rate": 5.4461695287383365e-05, "loss": 2.4626, "step": 58075 }, { "epoch": 4.55, "learning_rate": 5.446091115815887e-05, "loss": 2.602, "step": 58076 }, { "epoch": 4.55, "learning_rate": 5.446012702893437e-05, "loss": 2.516, "step": 58077 }, { "epoch": 4.55, "learning_rate": 5.4459342899709876e-05, "loss": 2.6409, "step": 58078 }, { "epoch": 4.55, "learning_rate": 5.445855877048538e-05, "loss": 2.6398, "step": 58079 }, { "epoch": 4.55, "learning_rate": 5.445777464126088e-05, "loss": 2.5154, "step": 58080 }, { "epoch": 4.55, "learning_rate": 5.445699051203639e-05, "loss": 2.5707, "step": 58081 }, { "epoch": 4.55, "learning_rate": 5.445620638281189e-05, "loss": 2.6628, "step": 58082 }, { "epoch": 4.55, "learning_rate": 5.4455422253587394e-05, "loss": 2.4563, "step": 58083 }, { "epoch": 4.55, "learning_rate": 5.44546381243629e-05, "loss": 2.8972, "step": 58084 }, { "epoch": 4.55, "learning_rate": 5.44538539951384e-05, "loss": 2.563, "step": 58085 }, { "epoch": 4.55, "learning_rate": 5.4453069865913905e-05, "loss": 2.6181, "step": 58086 }, { "epoch": 4.55, "learning_rate": 5.4452285736689416e-05, "loss": 2.8952, "step": 58087 }, { "epoch": 4.55, "learning_rate": 5.445150160746492e-05, "loss": 2.5708, "step": 58088 }, { "epoch": 4.55, "learning_rate": 5.445071747824042e-05, "loss": 2.4993, "step": 58089 }, { "epoch": 4.56, "learning_rate": 5.444993334901593e-05, "loss": 2.6756, "step": 58090 }, { "epoch": 4.56, "learning_rate": 5.444914921979143e-05, "loss": 2.53, "step": 58091 }, { "epoch": 4.56, "learning_rate": 5.4448365090566934e-05, "loss": 2.6384, "step": 58092 }, { "epoch": 4.56, "learning_rate": 5.444758096134244e-05, "loss": 2.6066, "step": 58093 }, { "epoch": 4.56, "learning_rate": 5.444679683211794e-05, "loss": 2.5294, "step": 58094 }, { "epoch": 4.56, "learning_rate": 5.444601270289343e-05, "loss": 2.7503, "step": 58095 }, { "epoch": 4.56, "learning_rate": 5.4445228573668935e-05, "loss": 2.6907, "step": 58096 }, { "epoch": 4.56, "learning_rate": 5.4444444444444446e-05, "loss": 2.6972, "step": 58097 }, { "epoch": 4.56, "learning_rate": 5.444366031521995e-05, "loss": 2.5965, "step": 58098 }, { "epoch": 4.56, "learning_rate": 5.444287618599545e-05, "loss": 2.5094, "step": 58099 }, { "epoch": 4.56, "learning_rate": 5.444209205677096e-05, "loss": 2.6055, "step": 58100 }, { "epoch": 4.56, "learning_rate": 5.444130792754646e-05, "loss": 2.5598, "step": 58101 }, { "epoch": 4.56, "learning_rate": 5.4440523798321964e-05, "loss": 2.6657, "step": 58102 }, { "epoch": 4.56, "learning_rate": 5.443973966909747e-05, "loss": 2.6881, "step": 58103 }, { "epoch": 4.56, "learning_rate": 5.443895553987297e-05, "loss": 2.5829, "step": 58104 }, { "epoch": 4.56, "learning_rate": 5.4438171410648475e-05, "loss": 2.5658, "step": 58105 }, { "epoch": 4.56, "learning_rate": 5.443738728142398e-05, "loss": 2.4662, "step": 58106 }, { "epoch": 4.56, "learning_rate": 5.443660315219948e-05, "loss": 2.4627, "step": 58107 }, { "epoch": 4.56, "learning_rate": 5.4435819022974986e-05, "loss": 2.6579, "step": 58108 }, { "epoch": 4.56, "learning_rate": 5.443503489375049e-05, "loss": 2.7415, "step": 58109 }, { "epoch": 4.56, "learning_rate": 5.4434250764525994e-05, "loss": 2.4587, "step": 58110 }, { "epoch": 4.56, "learning_rate": 5.44334666353015e-05, "loss": 2.8343, "step": 58111 }, { "epoch": 4.56, "learning_rate": 5.4432682506077e-05, "loss": 2.5001, "step": 58112 }, { "epoch": 4.56, "learning_rate": 5.443189837685251e-05, "loss": 2.5, "step": 58113 }, { "epoch": 4.56, "learning_rate": 5.4431114247628015e-05, "loss": 2.377, "step": 58114 }, { "epoch": 4.56, "learning_rate": 5.443033011840352e-05, "loss": 2.6279, "step": 58115 }, { "epoch": 4.56, "learning_rate": 5.442954598917902e-05, "loss": 2.796, "step": 58116 }, { "epoch": 4.56, "learning_rate": 5.4428761859954526e-05, "loss": 2.6021, "step": 58117 }, { "epoch": 4.56, "learning_rate": 5.442797773073003e-05, "loss": 2.6918, "step": 58118 }, { "epoch": 4.56, "learning_rate": 5.442719360150553e-05, "loss": 2.6537, "step": 58119 }, { "epoch": 4.56, "learning_rate": 5.442640947228104e-05, "loss": 2.6333, "step": 58120 }, { "epoch": 4.56, "learning_rate": 5.442562534305654e-05, "loss": 2.6667, "step": 58121 }, { "epoch": 4.56, "learning_rate": 5.4424841213832044e-05, "loss": 2.6267, "step": 58122 }, { "epoch": 4.56, "learning_rate": 5.442405708460755e-05, "loss": 2.6889, "step": 58123 }, { "epoch": 4.56, "learning_rate": 5.442327295538305e-05, "loss": 2.8289, "step": 58124 }, { "epoch": 4.56, "learning_rate": 5.4422488826158555e-05, "loss": 2.8194, "step": 58125 }, { "epoch": 4.56, "learning_rate": 5.442170469693406e-05, "loss": 2.8982, "step": 58126 }, { "epoch": 4.56, "learning_rate": 5.442092056770956e-05, "loss": 2.7822, "step": 58127 }, { "epoch": 4.56, "learning_rate": 5.4420136438485066e-05, "loss": 2.6173, "step": 58128 }, { "epoch": 4.56, "learning_rate": 5.441935230926058e-05, "loss": 2.5669, "step": 58129 }, { "epoch": 4.56, "learning_rate": 5.441856818003608e-05, "loss": 2.6902, "step": 58130 }, { "epoch": 4.56, "learning_rate": 5.441778405081157e-05, "loss": 2.5797, "step": 58131 }, { "epoch": 4.56, "learning_rate": 5.4416999921587074e-05, "loss": 2.7422, "step": 58132 }, { "epoch": 4.56, "learning_rate": 5.441621579236258e-05, "loss": 2.8505, "step": 58133 }, { "epoch": 4.56, "learning_rate": 5.441543166313808e-05, "loss": 2.5467, "step": 58134 }, { "epoch": 4.56, "learning_rate": 5.4414647533913585e-05, "loss": 2.4913, "step": 58135 }, { "epoch": 4.56, "learning_rate": 5.441386340468909e-05, "loss": 2.643, "step": 58136 }, { "epoch": 4.56, "learning_rate": 5.441307927546459e-05, "loss": 2.6553, "step": 58137 }, { "epoch": 4.56, "learning_rate": 5.4412295146240096e-05, "loss": 2.7682, "step": 58138 }, { "epoch": 4.56, "learning_rate": 5.441151101701561e-05, "loss": 2.6596, "step": 58139 }, { "epoch": 4.56, "learning_rate": 5.441072688779111e-05, "loss": 2.6365, "step": 58140 }, { "epoch": 4.56, "learning_rate": 5.4409942758566614e-05, "loss": 2.5571, "step": 58141 }, { "epoch": 4.56, "learning_rate": 5.440915862934212e-05, "loss": 2.9245, "step": 58142 }, { "epoch": 4.56, "learning_rate": 5.440837450011762e-05, "loss": 2.7538, "step": 58143 }, { "epoch": 4.56, "learning_rate": 5.4407590370893125e-05, "loss": 2.8175, "step": 58144 }, { "epoch": 4.56, "learning_rate": 5.440680624166863e-05, "loss": 2.571, "step": 58145 }, { "epoch": 4.56, "learning_rate": 5.440602211244413e-05, "loss": 2.722, "step": 58146 }, { "epoch": 4.56, "learning_rate": 5.4405237983219636e-05, "loss": 2.4636, "step": 58147 }, { "epoch": 4.56, "learning_rate": 5.440445385399514e-05, "loss": 2.5742, "step": 58148 }, { "epoch": 4.56, "learning_rate": 5.4403669724770644e-05, "loss": 2.5334, "step": 58149 }, { "epoch": 4.56, "learning_rate": 5.440288559554615e-05, "loss": 2.7815, "step": 58150 }, { "epoch": 4.56, "learning_rate": 5.440210146632165e-05, "loss": 2.7848, "step": 58151 }, { "epoch": 4.56, "learning_rate": 5.4401317337097155e-05, "loss": 2.5504, "step": 58152 }, { "epoch": 4.56, "learning_rate": 5.440053320787266e-05, "loss": 2.5614, "step": 58153 }, { "epoch": 4.56, "learning_rate": 5.439974907864817e-05, "loss": 2.6837, "step": 58154 }, { "epoch": 4.56, "learning_rate": 5.439896494942367e-05, "loss": 2.6617, "step": 58155 }, { "epoch": 4.56, "learning_rate": 5.4398180820199176e-05, "loss": 3.0609, "step": 58156 }, { "epoch": 4.56, "learning_rate": 5.439739669097468e-05, "loss": 2.5044, "step": 58157 }, { "epoch": 4.56, "learning_rate": 5.4396612561750184e-05, "loss": 2.7322, "step": 58158 }, { "epoch": 4.56, "learning_rate": 5.439582843252569e-05, "loss": 2.5996, "step": 58159 }, { "epoch": 4.56, "learning_rate": 5.439504430330119e-05, "loss": 2.6791, "step": 58160 }, { "epoch": 4.56, "learning_rate": 5.4394260174076695e-05, "loss": 2.6979, "step": 58161 }, { "epoch": 4.56, "learning_rate": 5.43934760448522e-05, "loss": 2.8163, "step": 58162 }, { "epoch": 4.56, "learning_rate": 5.43926919156277e-05, "loss": 2.9974, "step": 58163 }, { "epoch": 4.56, "learning_rate": 5.4391907786403206e-05, "loss": 2.5706, "step": 58164 }, { "epoch": 4.56, "learning_rate": 5.439112365717871e-05, "loss": 2.581, "step": 58165 }, { "epoch": 4.56, "learning_rate": 5.439033952795421e-05, "loss": 2.7237, "step": 58166 }, { "epoch": 4.56, "learning_rate": 5.438955539872971e-05, "loss": 2.5646, "step": 58167 }, { "epoch": 4.56, "learning_rate": 5.4388771269505214e-05, "loss": 2.6362, "step": 58168 }, { "epoch": 4.56, "learning_rate": 5.438798714028072e-05, "loss": 2.6692, "step": 58169 }, { "epoch": 4.56, "learning_rate": 5.438720301105622e-05, "loss": 2.757, "step": 58170 }, { "epoch": 4.56, "learning_rate": 5.4386418881831725e-05, "loss": 2.5556, "step": 58171 }, { "epoch": 4.56, "learning_rate": 5.438563475260723e-05, "loss": 2.7348, "step": 58172 }, { "epoch": 4.56, "learning_rate": 5.438485062338273e-05, "loss": 2.6507, "step": 58173 }, { "epoch": 4.56, "learning_rate": 5.4384066494158236e-05, "loss": 2.5489, "step": 58174 }, { "epoch": 4.56, "learning_rate": 5.438328236493374e-05, "loss": 2.8262, "step": 58175 }, { "epoch": 4.56, "learning_rate": 5.438249823570924e-05, "loss": 2.6244, "step": 58176 }, { "epoch": 4.56, "learning_rate": 5.438171410648475e-05, "loss": 2.5846, "step": 58177 }, { "epoch": 4.56, "learning_rate": 5.438092997726025e-05, "loss": 2.6416, "step": 58178 }, { "epoch": 4.56, "learning_rate": 5.4380145848035754e-05, "loss": 2.7177, "step": 58179 }, { "epoch": 4.56, "learning_rate": 5.4379361718811264e-05, "loss": 2.6927, "step": 58180 }, { "epoch": 4.56, "learning_rate": 5.437857758958677e-05, "loss": 2.5829, "step": 58181 }, { "epoch": 4.56, "learning_rate": 5.437779346036227e-05, "loss": 2.6518, "step": 58182 }, { "epoch": 4.56, "learning_rate": 5.4377009331137776e-05, "loss": 2.6226, "step": 58183 }, { "epoch": 4.56, "learning_rate": 5.437622520191328e-05, "loss": 2.6709, "step": 58184 }, { "epoch": 4.56, "learning_rate": 5.437544107268878e-05, "loss": 2.5351, "step": 58185 }, { "epoch": 4.56, "learning_rate": 5.4374656943464287e-05, "loss": 2.5571, "step": 58186 }, { "epoch": 4.56, "learning_rate": 5.437387281423979e-05, "loss": 2.4712, "step": 58187 }, { "epoch": 4.56, "learning_rate": 5.4373088685015294e-05, "loss": 2.6416, "step": 58188 }, { "epoch": 4.56, "learning_rate": 5.43723045557908e-05, "loss": 2.4926, "step": 58189 }, { "epoch": 4.56, "learning_rate": 5.43715204265663e-05, "loss": 2.6208, "step": 58190 }, { "epoch": 4.56, "learning_rate": 5.4370736297341805e-05, "loss": 2.7678, "step": 58191 }, { "epoch": 4.56, "learning_rate": 5.436995216811731e-05, "loss": 2.7576, "step": 58192 }, { "epoch": 4.56, "learning_rate": 5.436916803889281e-05, "loss": 2.7266, "step": 58193 }, { "epoch": 4.56, "learning_rate": 5.4368383909668316e-05, "loss": 2.6648, "step": 58194 }, { "epoch": 4.56, "learning_rate": 5.4367599780443826e-05, "loss": 2.6804, "step": 58195 }, { "epoch": 4.56, "learning_rate": 5.436681565121933e-05, "loss": 2.7243, "step": 58196 }, { "epoch": 4.56, "learning_rate": 5.4366031521994834e-05, "loss": 2.436, "step": 58197 }, { "epoch": 4.56, "learning_rate": 5.436524739277034e-05, "loss": 2.6907, "step": 58198 }, { "epoch": 4.56, "learning_rate": 5.436446326354584e-05, "loss": 2.5875, "step": 58199 }, { "epoch": 4.56, "learning_rate": 5.4363679134321345e-05, "loss": 2.6246, "step": 58200 }, { "epoch": 4.56, "learning_rate": 5.436289500509685e-05, "loss": 2.5355, "step": 58201 }, { "epoch": 4.56, "learning_rate": 5.436211087587235e-05, "loss": 2.6288, "step": 58202 }, { "epoch": 4.56, "learning_rate": 5.436132674664784e-05, "loss": 2.6006, "step": 58203 }, { "epoch": 4.56, "learning_rate": 5.4360542617423346e-05, "loss": 2.803, "step": 58204 }, { "epoch": 4.56, "learning_rate": 5.4359758488198856e-05, "loss": 2.7646, "step": 58205 }, { "epoch": 4.56, "learning_rate": 5.435897435897436e-05, "loss": 2.8262, "step": 58206 }, { "epoch": 4.56, "learning_rate": 5.4358190229749864e-05, "loss": 2.6808, "step": 58207 }, { "epoch": 4.56, "learning_rate": 5.435740610052537e-05, "loss": 2.6118, "step": 58208 }, { "epoch": 4.56, "learning_rate": 5.435662197130087e-05, "loss": 2.4155, "step": 58209 }, { "epoch": 4.56, "learning_rate": 5.4355837842076375e-05, "loss": 2.7641, "step": 58210 }, { "epoch": 4.56, "learning_rate": 5.435505371285188e-05, "loss": 2.5077, "step": 58211 }, { "epoch": 4.56, "learning_rate": 5.435426958362738e-05, "loss": 2.7696, "step": 58212 }, { "epoch": 4.56, "learning_rate": 5.4353485454402886e-05, "loss": 2.4778, "step": 58213 }, { "epoch": 4.56, "learning_rate": 5.435270132517839e-05, "loss": 2.5828, "step": 58214 }, { "epoch": 4.56, "learning_rate": 5.435191719595389e-05, "loss": 2.773, "step": 58215 }, { "epoch": 4.56, "learning_rate": 5.43511330667294e-05, "loss": 2.4356, "step": 58216 }, { "epoch": 4.56, "learning_rate": 5.43503489375049e-05, "loss": 2.7158, "step": 58217 }, { "epoch": 4.57, "learning_rate": 5.4349564808280404e-05, "loss": 2.7603, "step": 58218 }, { "epoch": 4.57, "learning_rate": 5.434878067905591e-05, "loss": 2.762, "step": 58219 }, { "epoch": 4.57, "learning_rate": 5.434799654983141e-05, "loss": 2.6141, "step": 58220 }, { "epoch": 4.57, "learning_rate": 5.434721242060692e-05, "loss": 2.6275, "step": 58221 }, { "epoch": 4.57, "learning_rate": 5.4346428291382426e-05, "loss": 2.5576, "step": 58222 }, { "epoch": 4.57, "learning_rate": 5.434564416215793e-05, "loss": 2.5848, "step": 58223 }, { "epoch": 4.57, "learning_rate": 5.434486003293343e-05, "loss": 2.5991, "step": 58224 }, { "epoch": 4.57, "learning_rate": 5.434407590370894e-05, "loss": 2.4713, "step": 58225 }, { "epoch": 4.57, "learning_rate": 5.434329177448444e-05, "loss": 2.4986, "step": 58226 }, { "epoch": 4.57, "learning_rate": 5.4342507645259944e-05, "loss": 2.5804, "step": 58227 }, { "epoch": 4.57, "learning_rate": 5.434172351603545e-05, "loss": 2.5689, "step": 58228 }, { "epoch": 4.57, "learning_rate": 5.434093938681095e-05, "loss": 2.7873, "step": 58229 }, { "epoch": 4.57, "learning_rate": 5.4340155257586455e-05, "loss": 2.6574, "step": 58230 }, { "epoch": 4.57, "learning_rate": 5.433937112836196e-05, "loss": 2.7184, "step": 58231 }, { "epoch": 4.57, "learning_rate": 5.433858699913746e-05, "loss": 2.7405, "step": 58232 }, { "epoch": 4.57, "learning_rate": 5.4337802869912966e-05, "loss": 2.5544, "step": 58233 }, { "epoch": 4.57, "learning_rate": 5.433701874068847e-05, "loss": 2.6432, "step": 58234 }, { "epoch": 4.57, "learning_rate": 5.4336234611463974e-05, "loss": 2.8281, "step": 58235 }, { "epoch": 4.57, "learning_rate": 5.433545048223948e-05, "loss": 2.676, "step": 58236 }, { "epoch": 4.57, "learning_rate": 5.433466635301499e-05, "loss": 2.602, "step": 58237 }, { "epoch": 4.57, "learning_rate": 5.433388222379049e-05, "loss": 2.667, "step": 58238 }, { "epoch": 4.57, "learning_rate": 5.433309809456598e-05, "loss": 2.5592, "step": 58239 }, { "epoch": 4.57, "learning_rate": 5.4332313965341485e-05, "loss": 2.6081, "step": 58240 }, { "epoch": 4.57, "learning_rate": 5.433152983611699e-05, "loss": 2.5778, "step": 58241 }, { "epoch": 4.57, "learning_rate": 5.433074570689249e-05, "loss": 2.8924, "step": 58242 }, { "epoch": 4.57, "learning_rate": 5.4329961577667996e-05, "loss": 2.5305, "step": 58243 }, { "epoch": 4.57, "learning_rate": 5.43291774484435e-05, "loss": 2.5501, "step": 58244 }, { "epoch": 4.57, "learning_rate": 5.4328393319219004e-05, "loss": 2.582, "step": 58245 }, { "epoch": 4.57, "learning_rate": 5.432760918999451e-05, "loss": 2.6306, "step": 58246 }, { "epoch": 4.57, "learning_rate": 5.432682506077002e-05, "loss": 2.8572, "step": 58247 }, { "epoch": 4.57, "learning_rate": 5.432604093154552e-05, "loss": 2.6818, "step": 58248 }, { "epoch": 4.57, "learning_rate": 5.4325256802321025e-05, "loss": 2.7731, "step": 58249 }, { "epoch": 4.57, "learning_rate": 5.432447267309653e-05, "loss": 2.6413, "step": 58250 }, { "epoch": 4.57, "learning_rate": 5.432368854387203e-05, "loss": 2.7898, "step": 58251 }, { "epoch": 4.57, "learning_rate": 5.4322904414647536e-05, "loss": 2.5088, "step": 58252 }, { "epoch": 4.57, "learning_rate": 5.432212028542304e-05, "loss": 2.4021, "step": 58253 }, { "epoch": 4.57, "learning_rate": 5.4321336156198543e-05, "loss": 2.7674, "step": 58254 }, { "epoch": 4.57, "learning_rate": 5.432055202697405e-05, "loss": 2.5906, "step": 58255 }, { "epoch": 4.57, "learning_rate": 5.431976789774955e-05, "loss": 2.631, "step": 58256 }, { "epoch": 4.57, "learning_rate": 5.4318983768525054e-05, "loss": 2.6503, "step": 58257 }, { "epoch": 4.57, "learning_rate": 5.431819963930056e-05, "loss": 2.5448, "step": 58258 }, { "epoch": 4.57, "learning_rate": 5.431741551007606e-05, "loss": 2.583, "step": 58259 }, { "epoch": 4.57, "learning_rate": 5.4316631380851565e-05, "loss": 2.5707, "step": 58260 }, { "epoch": 4.57, "learning_rate": 5.431584725162707e-05, "loss": 2.5884, "step": 58261 }, { "epoch": 4.57, "learning_rate": 5.431506312240258e-05, "loss": 2.842, "step": 58262 }, { "epoch": 4.57, "learning_rate": 5.431427899317808e-05, "loss": 2.5956, "step": 58263 }, { "epoch": 4.57, "learning_rate": 5.431349486395359e-05, "loss": 2.4968, "step": 58264 }, { "epoch": 4.57, "learning_rate": 5.431271073472909e-05, "loss": 2.605, "step": 58265 }, { "epoch": 4.57, "learning_rate": 5.4311926605504594e-05, "loss": 2.6513, "step": 58266 }, { "epoch": 4.57, "learning_rate": 5.43111424762801e-05, "loss": 2.8301, "step": 58267 }, { "epoch": 4.57, "learning_rate": 5.43103583470556e-05, "loss": 2.5836, "step": 58268 }, { "epoch": 4.57, "learning_rate": 5.4309574217831105e-05, "loss": 2.769, "step": 58269 }, { "epoch": 4.57, "learning_rate": 5.430879008860661e-05, "loss": 2.8485, "step": 58270 }, { "epoch": 4.57, "learning_rate": 5.430800595938211e-05, "loss": 2.6517, "step": 58271 }, { "epoch": 4.57, "learning_rate": 5.4307221830157616e-05, "loss": 2.7107, "step": 58272 }, { "epoch": 4.57, "learning_rate": 5.430643770093312e-05, "loss": 2.6037, "step": 58273 }, { "epoch": 4.57, "learning_rate": 5.4305653571708624e-05, "loss": 2.5684, "step": 58274 }, { "epoch": 4.57, "learning_rate": 5.430486944248412e-05, "loss": 2.6902, "step": 58275 }, { "epoch": 4.57, "learning_rate": 5.4304085313259624e-05, "loss": 2.5299, "step": 58276 }, { "epoch": 4.57, "learning_rate": 5.430330118403513e-05, "loss": 2.7118, "step": 58277 }, { "epoch": 4.57, "learning_rate": 5.430251705481063e-05, "loss": 2.7376, "step": 58278 }, { "epoch": 4.57, "learning_rate": 5.4301732925586135e-05, "loss": 2.4834, "step": 58279 }, { "epoch": 4.57, "learning_rate": 5.430094879636164e-05, "loss": 2.5468, "step": 58280 }, { "epoch": 4.57, "learning_rate": 5.430016466713714e-05, "loss": 2.7472, "step": 58281 }, { "epoch": 4.57, "learning_rate": 5.4299380537912646e-05, "loss": 2.7295, "step": 58282 }, { "epoch": 4.57, "learning_rate": 5.429859640868815e-05, "loss": 2.5802, "step": 58283 }, { "epoch": 4.57, "learning_rate": 5.4297812279463654e-05, "loss": 2.5812, "step": 58284 }, { "epoch": 4.57, "learning_rate": 5.429702815023916e-05, "loss": 2.6048, "step": 58285 }, { "epoch": 4.57, "learning_rate": 5.429624402101466e-05, "loss": 2.533, "step": 58286 }, { "epoch": 4.57, "learning_rate": 5.4295459891790165e-05, "loss": 2.6958, "step": 58287 }, { "epoch": 4.57, "learning_rate": 5.4294675762565675e-05, "loss": 2.6236, "step": 58288 }, { "epoch": 4.57, "learning_rate": 5.429389163334118e-05, "loss": 2.6928, "step": 58289 }, { "epoch": 4.57, "learning_rate": 5.429310750411668e-05, "loss": 2.6406, "step": 58290 }, { "epoch": 4.57, "learning_rate": 5.4292323374892186e-05, "loss": 2.6558, "step": 58291 }, { "epoch": 4.57, "learning_rate": 5.429153924566769e-05, "loss": 2.7166, "step": 58292 }, { "epoch": 4.57, "learning_rate": 5.4290755116443194e-05, "loss": 2.5656, "step": 58293 }, { "epoch": 4.57, "learning_rate": 5.42899709872187e-05, "loss": 2.7025, "step": 58294 }, { "epoch": 4.57, "learning_rate": 5.42891868579942e-05, "loss": 2.6623, "step": 58295 }, { "epoch": 4.57, "learning_rate": 5.4288402728769705e-05, "loss": 2.6734, "step": 58296 }, { "epoch": 4.57, "learning_rate": 5.428761859954521e-05, "loss": 2.7445, "step": 58297 }, { "epoch": 4.57, "learning_rate": 5.428683447032071e-05, "loss": 2.3782, "step": 58298 }, { "epoch": 4.57, "learning_rate": 5.4286050341096216e-05, "loss": 2.6023, "step": 58299 }, { "epoch": 4.57, "learning_rate": 5.428526621187172e-05, "loss": 2.6401, "step": 58300 }, { "epoch": 4.57, "learning_rate": 5.428448208264722e-05, "loss": 2.8024, "step": 58301 }, { "epoch": 4.57, "learning_rate": 5.428369795342273e-05, "loss": 2.6926, "step": 58302 }, { "epoch": 4.57, "learning_rate": 5.428291382419823e-05, "loss": 2.6087, "step": 58303 }, { "epoch": 4.57, "learning_rate": 5.428212969497374e-05, "loss": 2.8219, "step": 58304 }, { "epoch": 4.57, "learning_rate": 5.4281345565749245e-05, "loss": 2.4753, "step": 58305 }, { "epoch": 4.57, "learning_rate": 5.428056143652475e-05, "loss": 3.1948, "step": 58306 }, { "epoch": 4.57, "learning_rate": 5.427977730730025e-05, "loss": 2.6975, "step": 58307 }, { "epoch": 4.57, "learning_rate": 5.4278993178075756e-05, "loss": 2.5525, "step": 58308 }, { "epoch": 4.57, "learning_rate": 5.427820904885126e-05, "loss": 2.6839, "step": 58309 }, { "epoch": 4.57, "learning_rate": 5.427742491962676e-05, "loss": 2.8464, "step": 58310 }, { "epoch": 4.57, "learning_rate": 5.427664079040225e-05, "loss": 2.9611, "step": 58311 }, { "epoch": 4.57, "learning_rate": 5.427585666117776e-05, "loss": 2.6768, "step": 58312 }, { "epoch": 4.57, "learning_rate": 5.427507253195327e-05, "loss": 2.5502, "step": 58313 }, { "epoch": 4.57, "learning_rate": 5.427428840272877e-05, "loss": 2.7252, "step": 58314 }, { "epoch": 4.57, "learning_rate": 5.4273504273504275e-05, "loss": 2.7309, "step": 58315 }, { "epoch": 4.57, "learning_rate": 5.427272014427978e-05, "loss": 2.7352, "step": 58316 }, { "epoch": 4.57, "learning_rate": 5.427193601505528e-05, "loss": 2.9031, "step": 58317 }, { "epoch": 4.57, "learning_rate": 5.4271151885830786e-05, "loss": 2.4259, "step": 58318 }, { "epoch": 4.57, "learning_rate": 5.427036775660629e-05, "loss": 2.8548, "step": 58319 }, { "epoch": 4.57, "learning_rate": 5.426958362738179e-05, "loss": 2.6069, "step": 58320 }, { "epoch": 4.57, "learning_rate": 5.4268799498157297e-05, "loss": 2.6708, "step": 58321 }, { "epoch": 4.57, "learning_rate": 5.42680153689328e-05, "loss": 2.7631, "step": 58322 }, { "epoch": 4.57, "learning_rate": 5.4267231239708304e-05, "loss": 2.5766, "step": 58323 }, { "epoch": 4.57, "learning_rate": 5.426644711048381e-05, "loss": 2.6053, "step": 58324 }, { "epoch": 4.57, "learning_rate": 5.426566298125931e-05, "loss": 2.5732, "step": 58325 }, { "epoch": 4.57, "learning_rate": 5.4264878852034815e-05, "loss": 2.6082, "step": 58326 }, { "epoch": 4.57, "learning_rate": 5.426409472281032e-05, "loss": 2.8026, "step": 58327 }, { "epoch": 4.57, "learning_rate": 5.426331059358582e-05, "loss": 2.7278, "step": 58328 }, { "epoch": 4.57, "learning_rate": 5.426252646436133e-05, "loss": 2.672, "step": 58329 }, { "epoch": 4.57, "learning_rate": 5.4261742335136836e-05, "loss": 2.6946, "step": 58330 }, { "epoch": 4.57, "learning_rate": 5.426095820591234e-05, "loss": 2.6905, "step": 58331 }, { "epoch": 4.57, "learning_rate": 5.4260174076687844e-05, "loss": 2.6644, "step": 58332 }, { "epoch": 4.57, "learning_rate": 5.425938994746335e-05, "loss": 2.7958, "step": 58333 }, { "epoch": 4.57, "learning_rate": 5.425860581823885e-05, "loss": 2.7455, "step": 58334 }, { "epoch": 4.57, "learning_rate": 5.4257821689014355e-05, "loss": 2.4099, "step": 58335 }, { "epoch": 4.57, "learning_rate": 5.425703755978986e-05, "loss": 2.6728, "step": 58336 }, { "epoch": 4.57, "learning_rate": 5.425625343056536e-05, "loss": 2.6099, "step": 58337 }, { "epoch": 4.57, "learning_rate": 5.4255469301340866e-05, "loss": 2.6652, "step": 58338 }, { "epoch": 4.57, "learning_rate": 5.425468517211637e-05, "loss": 2.7015, "step": 58339 }, { "epoch": 4.57, "learning_rate": 5.425390104289187e-05, "loss": 2.6474, "step": 58340 }, { "epoch": 4.57, "learning_rate": 5.425311691366738e-05, "loss": 2.6042, "step": 58341 }, { "epoch": 4.57, "learning_rate": 5.425233278444288e-05, "loss": 2.7029, "step": 58342 }, { "epoch": 4.57, "learning_rate": 5.4251548655218384e-05, "loss": 2.5438, "step": 58343 }, { "epoch": 4.57, "learning_rate": 5.425076452599389e-05, "loss": 2.6885, "step": 58344 }, { "epoch": 4.58, "learning_rate": 5.42499803967694e-05, "loss": 2.5268, "step": 58345 }, { "epoch": 4.58, "learning_rate": 5.42491962675449e-05, "loss": 2.6385, "step": 58346 }, { "epoch": 4.58, "learning_rate": 5.424841213832039e-05, "loss": 2.6177, "step": 58347 }, { "epoch": 4.58, "learning_rate": 5.4247628009095896e-05, "loss": 2.7055, "step": 58348 }, { "epoch": 4.58, "learning_rate": 5.42468438798714e-05, "loss": 2.6723, "step": 58349 }, { "epoch": 4.58, "learning_rate": 5.42460597506469e-05, "loss": 2.7462, "step": 58350 }, { "epoch": 4.58, "learning_rate": 5.424527562142241e-05, "loss": 2.6008, "step": 58351 }, { "epoch": 4.58, "learning_rate": 5.424449149219791e-05, "loss": 2.8511, "step": 58352 }, { "epoch": 4.58, "learning_rate": 5.4243707362973414e-05, "loss": 2.599, "step": 58353 }, { "epoch": 4.58, "learning_rate": 5.424292323374892e-05, "loss": 2.7808, "step": 58354 }, { "epoch": 4.58, "learning_rate": 5.424213910452443e-05, "loss": 2.784, "step": 58355 }, { "epoch": 4.58, "learning_rate": 5.424135497529993e-05, "loss": 2.7898, "step": 58356 }, { "epoch": 4.58, "learning_rate": 5.4240570846075436e-05, "loss": 2.74, "step": 58357 }, { "epoch": 4.58, "learning_rate": 5.423978671685094e-05, "loss": 2.5746, "step": 58358 }, { "epoch": 4.58, "learning_rate": 5.423900258762644e-05, "loss": 2.8556, "step": 58359 }, { "epoch": 4.58, "learning_rate": 5.423821845840195e-05, "loss": 2.6913, "step": 58360 }, { "epoch": 4.58, "learning_rate": 5.423743432917745e-05, "loss": 2.8007, "step": 58361 }, { "epoch": 4.58, "learning_rate": 5.4236650199952954e-05, "loss": 2.5238, "step": 58362 }, { "epoch": 4.58, "learning_rate": 5.423586607072846e-05, "loss": 2.7801, "step": 58363 }, { "epoch": 4.58, "learning_rate": 5.423508194150396e-05, "loss": 2.6025, "step": 58364 }, { "epoch": 4.58, "learning_rate": 5.4234297812279465e-05, "loss": 2.529, "step": 58365 }, { "epoch": 4.58, "learning_rate": 5.423351368305497e-05, "loss": 2.8654, "step": 58366 }, { "epoch": 4.58, "learning_rate": 5.423272955383047e-05, "loss": 2.5544, "step": 58367 }, { "epoch": 4.58, "learning_rate": 5.4231945424605976e-05, "loss": 2.5545, "step": 58368 }, { "epoch": 4.58, "learning_rate": 5.423116129538148e-05, "loss": 2.6814, "step": 58369 }, { "epoch": 4.58, "learning_rate": 5.423037716615699e-05, "loss": 2.805, "step": 58370 }, { "epoch": 4.58, "learning_rate": 5.4229593036932494e-05, "loss": 2.5499, "step": 58371 }, { "epoch": 4.58, "learning_rate": 5.4228808907708e-05, "loss": 2.7252, "step": 58372 }, { "epoch": 4.58, "learning_rate": 5.42280247784835e-05, "loss": 2.6361, "step": 58373 }, { "epoch": 4.58, "learning_rate": 5.4227240649259005e-05, "loss": 2.6072, "step": 58374 }, { "epoch": 4.58, "learning_rate": 5.422645652003451e-05, "loss": 2.8584, "step": 58375 }, { "epoch": 4.58, "learning_rate": 5.422567239081001e-05, "loss": 2.5784, "step": 58376 }, { "epoch": 4.58, "learning_rate": 5.4224888261585516e-05, "loss": 2.7968, "step": 58377 }, { "epoch": 4.58, "learning_rate": 5.422410413236102e-05, "loss": 2.5786, "step": 58378 }, { "epoch": 4.58, "learning_rate": 5.4223320003136523e-05, "loss": 2.5114, "step": 58379 }, { "epoch": 4.58, "learning_rate": 5.422253587391203e-05, "loss": 2.4684, "step": 58380 }, { "epoch": 4.58, "learning_rate": 5.422175174468753e-05, "loss": 2.7159, "step": 58381 }, { "epoch": 4.58, "learning_rate": 5.4220967615463034e-05, "loss": 2.8706, "step": 58382 }, { "epoch": 4.58, "learning_rate": 5.422018348623853e-05, "loss": 2.4501, "step": 58383 }, { "epoch": 4.58, "learning_rate": 5.4219399357014035e-05, "loss": 2.6266, "step": 58384 }, { "epoch": 4.58, "learning_rate": 5.421861522778954e-05, "loss": 2.5871, "step": 58385 }, { "epoch": 4.58, "learning_rate": 5.421783109856504e-05, "loss": 2.6682, "step": 58386 }, { "epoch": 4.58, "learning_rate": 5.4217046969340546e-05, "loss": 2.5453, "step": 58387 }, { "epoch": 4.58, "learning_rate": 5.421626284011605e-05, "loss": 2.6258, "step": 58388 }, { "epoch": 4.58, "learning_rate": 5.4215478710891553e-05, "loss": 2.784, "step": 58389 }, { "epoch": 4.58, "learning_rate": 5.421469458166706e-05, "loss": 2.6795, "step": 58390 }, { "epoch": 4.58, "learning_rate": 5.421391045244256e-05, "loss": 2.5848, "step": 58391 }, { "epoch": 4.58, "learning_rate": 5.4213126323218065e-05, "loss": 2.6438, "step": 58392 }, { "epoch": 4.58, "learning_rate": 5.421234219399357e-05, "loss": 2.5489, "step": 58393 }, { "epoch": 4.58, "learning_rate": 5.421155806476907e-05, "loss": 2.5634, "step": 58394 }, { "epoch": 4.58, "learning_rate": 5.4210773935544576e-05, "loss": 2.7118, "step": 58395 }, { "epoch": 4.58, "learning_rate": 5.4209989806320086e-05, "loss": 2.6225, "step": 58396 }, { "epoch": 4.58, "learning_rate": 5.420920567709559e-05, "loss": 2.4503, "step": 58397 }, { "epoch": 4.58, "learning_rate": 5.420842154787109e-05, "loss": 2.6103, "step": 58398 }, { "epoch": 4.58, "learning_rate": 5.42076374186466e-05, "loss": 2.6789, "step": 58399 }, { "epoch": 4.58, "learning_rate": 5.42068532894221e-05, "loss": 2.6097, "step": 58400 }, { "epoch": 4.58, "learning_rate": 5.4206069160197604e-05, "loss": 2.7716, "step": 58401 }, { "epoch": 4.58, "learning_rate": 5.420528503097311e-05, "loss": 2.6327, "step": 58402 }, { "epoch": 4.58, "learning_rate": 5.420450090174861e-05, "loss": 2.5542, "step": 58403 }, { "epoch": 4.58, "learning_rate": 5.4203716772524115e-05, "loss": 2.4643, "step": 58404 }, { "epoch": 4.58, "learning_rate": 5.420293264329962e-05, "loss": 2.7532, "step": 58405 }, { "epoch": 4.58, "learning_rate": 5.420214851407512e-05, "loss": 2.7822, "step": 58406 }, { "epoch": 4.58, "learning_rate": 5.4201364384850626e-05, "loss": 2.5397, "step": 58407 }, { "epoch": 4.58, "learning_rate": 5.420058025562613e-05, "loss": 2.6309, "step": 58408 }, { "epoch": 4.58, "learning_rate": 5.4199796126401634e-05, "loss": 2.5527, "step": 58409 }, { "epoch": 4.58, "learning_rate": 5.419901199717714e-05, "loss": 2.6965, "step": 58410 }, { "epoch": 4.58, "learning_rate": 5.419822786795264e-05, "loss": 2.4347, "step": 58411 }, { "epoch": 4.58, "learning_rate": 5.419744373872815e-05, "loss": 2.8617, "step": 58412 }, { "epoch": 4.58, "learning_rate": 5.4196659609503655e-05, "loss": 2.7926, "step": 58413 }, { "epoch": 4.58, "learning_rate": 5.419587548027916e-05, "loss": 2.7484, "step": 58414 }, { "epoch": 4.58, "learning_rate": 5.419509135105466e-05, "loss": 2.6586, "step": 58415 }, { "epoch": 4.58, "learning_rate": 5.4194307221830166e-05, "loss": 2.5927, "step": 58416 }, { "epoch": 4.58, "learning_rate": 5.419352309260567e-05, "loss": 2.6827, "step": 58417 }, { "epoch": 4.58, "learning_rate": 5.4192738963381174e-05, "loss": 2.6383, "step": 58418 }, { "epoch": 4.58, "learning_rate": 5.4191954834156664e-05, "loss": 2.8021, "step": 58419 }, { "epoch": 4.58, "learning_rate": 5.419117070493217e-05, "loss": 2.6577, "step": 58420 }, { "epoch": 4.58, "learning_rate": 5.419038657570767e-05, "loss": 2.7617, "step": 58421 }, { "epoch": 4.58, "learning_rate": 5.418960244648318e-05, "loss": 2.6985, "step": 58422 }, { "epoch": 4.58, "learning_rate": 5.4188818317258685e-05, "loss": 2.7335, "step": 58423 }, { "epoch": 4.58, "learning_rate": 5.418803418803419e-05, "loss": 2.5781, "step": 58424 }, { "epoch": 4.58, "learning_rate": 5.418725005880969e-05, "loss": 2.5163, "step": 58425 }, { "epoch": 4.58, "learning_rate": 5.4186465929585196e-05, "loss": 2.7057, "step": 58426 }, { "epoch": 4.58, "learning_rate": 5.41856818003607e-05, "loss": 2.6367, "step": 58427 }, { "epoch": 4.58, "learning_rate": 5.4184897671136204e-05, "loss": 2.7273, "step": 58428 }, { "epoch": 4.58, "learning_rate": 5.418411354191171e-05, "loss": 2.7583, "step": 58429 }, { "epoch": 4.58, "learning_rate": 5.418332941268721e-05, "loss": 2.7653, "step": 58430 }, { "epoch": 4.58, "learning_rate": 5.4182545283462715e-05, "loss": 2.6269, "step": 58431 }, { "epoch": 4.58, "learning_rate": 5.418176115423822e-05, "loss": 2.6174, "step": 58432 }, { "epoch": 4.58, "learning_rate": 5.418097702501372e-05, "loss": 2.683, "step": 58433 }, { "epoch": 4.58, "learning_rate": 5.4180192895789226e-05, "loss": 2.5498, "step": 58434 }, { "epoch": 4.58, "learning_rate": 5.417940876656473e-05, "loss": 2.66, "step": 58435 }, { "epoch": 4.58, "learning_rate": 5.417862463734023e-05, "loss": 2.6985, "step": 58436 }, { "epoch": 4.58, "learning_rate": 5.4177840508115744e-05, "loss": 2.5337, "step": 58437 }, { "epoch": 4.58, "learning_rate": 5.417705637889125e-05, "loss": 2.5617, "step": 58438 }, { "epoch": 4.58, "learning_rate": 5.417627224966675e-05, "loss": 2.7633, "step": 58439 }, { "epoch": 4.58, "learning_rate": 5.4175488120442255e-05, "loss": 2.7912, "step": 58440 }, { "epoch": 4.58, "learning_rate": 5.417470399121776e-05, "loss": 2.5767, "step": 58441 }, { "epoch": 4.58, "learning_rate": 5.417391986199326e-05, "loss": 2.595, "step": 58442 }, { "epoch": 4.58, "learning_rate": 5.4173135732768766e-05, "loss": 2.7397, "step": 58443 }, { "epoch": 4.58, "learning_rate": 5.417235160354427e-05, "loss": 2.4579, "step": 58444 }, { "epoch": 4.58, "learning_rate": 5.417156747431977e-05, "loss": 2.5258, "step": 58445 }, { "epoch": 4.58, "learning_rate": 5.417078334509528e-05, "loss": 2.6194, "step": 58446 }, { "epoch": 4.58, "learning_rate": 5.416999921587078e-05, "loss": 2.7968, "step": 58447 }, { "epoch": 4.58, "learning_rate": 5.4169215086646284e-05, "loss": 2.7941, "step": 58448 }, { "epoch": 4.58, "learning_rate": 5.416843095742179e-05, "loss": 2.7762, "step": 58449 }, { "epoch": 4.58, "learning_rate": 5.416764682819729e-05, "loss": 2.6165, "step": 58450 }, { "epoch": 4.58, "learning_rate": 5.4166862698972795e-05, "loss": 2.6627, "step": 58451 }, { "epoch": 4.58, "learning_rate": 5.41660785697483e-05, "loss": 2.4877, "step": 58452 }, { "epoch": 4.58, "learning_rate": 5.416529444052381e-05, "loss": 2.6654, "step": 58453 }, { "epoch": 4.58, "learning_rate": 5.416451031129931e-05, "loss": 2.6086, "step": 58454 }, { "epoch": 4.58, "learning_rate": 5.41637261820748e-05, "loss": 2.791, "step": 58455 }, { "epoch": 4.58, "learning_rate": 5.416294205285031e-05, "loss": 2.6575, "step": 58456 }, { "epoch": 4.58, "learning_rate": 5.416215792362581e-05, "loss": 2.6468, "step": 58457 }, { "epoch": 4.58, "learning_rate": 5.4161373794401314e-05, "loss": 2.8328, "step": 58458 }, { "epoch": 4.58, "learning_rate": 5.416058966517682e-05, "loss": 2.5132, "step": 58459 }, { "epoch": 4.58, "learning_rate": 5.415980553595232e-05, "loss": 2.5522, "step": 58460 }, { "epoch": 4.58, "learning_rate": 5.4159021406727825e-05, "loss": 2.6004, "step": 58461 }, { "epoch": 4.58, "learning_rate": 5.415823727750333e-05, "loss": 2.8045, "step": 58462 }, { "epoch": 4.58, "learning_rate": 5.415745314827884e-05, "loss": 2.676, "step": 58463 }, { "epoch": 4.58, "learning_rate": 5.415666901905434e-05, "loss": 2.5649, "step": 58464 }, { "epoch": 4.58, "learning_rate": 5.4155884889829847e-05, "loss": 2.678, "step": 58465 }, { "epoch": 4.58, "learning_rate": 5.415510076060535e-05, "loss": 2.7802, "step": 58466 }, { "epoch": 4.58, "learning_rate": 5.4154316631380854e-05, "loss": 2.5488, "step": 58467 }, { "epoch": 4.58, "learning_rate": 5.415353250215636e-05, "loss": 2.5614, "step": 58468 }, { "epoch": 4.58, "learning_rate": 5.415274837293186e-05, "loss": 2.5805, "step": 58469 }, { "epoch": 4.58, "learning_rate": 5.4151964243707365e-05, "loss": 2.5718, "step": 58470 }, { "epoch": 4.58, "learning_rate": 5.415118011448287e-05, "loss": 2.6551, "step": 58471 }, { "epoch": 4.58, "learning_rate": 5.415039598525837e-05, "loss": 2.8724, "step": 58472 }, { "epoch": 4.59, "learning_rate": 5.4149611856033876e-05, "loss": 2.7857, "step": 58473 }, { "epoch": 4.59, "learning_rate": 5.414882772680938e-05, "loss": 2.9236, "step": 58474 }, { "epoch": 4.59, "learning_rate": 5.414804359758488e-05, "loss": 2.8175, "step": 58475 }, { "epoch": 4.59, "learning_rate": 5.414725946836039e-05, "loss": 2.7929, "step": 58476 }, { "epoch": 4.59, "learning_rate": 5.414647533913589e-05, "loss": 2.7378, "step": 58477 }, { "epoch": 4.59, "learning_rate": 5.41456912099114e-05, "loss": 2.5856, "step": 58478 }, { "epoch": 4.59, "learning_rate": 5.4144907080686905e-05, "loss": 2.6444, "step": 58479 }, { "epoch": 4.59, "learning_rate": 5.414412295146241e-05, "loss": 2.5826, "step": 58480 }, { "epoch": 4.59, "learning_rate": 5.414333882223791e-05, "loss": 2.8457, "step": 58481 }, { "epoch": 4.59, "learning_rate": 5.4142554693013416e-05, "loss": 2.6629, "step": 58482 }, { "epoch": 4.59, "learning_rate": 5.414177056378892e-05, "loss": 2.8386, "step": 58483 }, { "epoch": 4.59, "learning_rate": 5.414098643456442e-05, "loss": 2.7384, "step": 58484 }, { "epoch": 4.59, "learning_rate": 5.414020230533993e-05, "loss": 2.5387, "step": 58485 }, { "epoch": 4.59, "learning_rate": 5.413941817611543e-05, "loss": 2.4647, "step": 58486 }, { "epoch": 4.59, "learning_rate": 5.4138634046890934e-05, "loss": 2.4391, "step": 58487 }, { "epoch": 4.59, "learning_rate": 5.413784991766644e-05, "loss": 2.5222, "step": 58488 }, { "epoch": 4.59, "learning_rate": 5.413706578844194e-05, "loss": 2.5962, "step": 58489 }, { "epoch": 4.59, "learning_rate": 5.4136281659217445e-05, "loss": 2.7899, "step": 58490 }, { "epoch": 4.59, "learning_rate": 5.413549752999294e-05, "loss": 2.6945, "step": 58491 }, { "epoch": 4.59, "learning_rate": 5.4134713400768446e-05, "loss": 2.7674, "step": 58492 }, { "epoch": 4.59, "learning_rate": 5.413392927154395e-05, "loss": 2.5143, "step": 58493 }, { "epoch": 4.59, "learning_rate": 5.413314514231945e-05, "loss": 2.7061, "step": 58494 }, { "epoch": 4.59, "learning_rate": 5.413236101309496e-05, "loss": 2.5954, "step": 58495 }, { "epoch": 4.59, "learning_rate": 5.413157688387046e-05, "loss": 2.6752, "step": 58496 }, { "epoch": 4.59, "learning_rate": 5.4130792754645964e-05, "loss": 2.5018, "step": 58497 }, { "epoch": 4.59, "learning_rate": 5.413000862542147e-05, "loss": 2.8017, "step": 58498 }, { "epoch": 4.59, "learning_rate": 5.412922449619697e-05, "loss": 2.6126, "step": 58499 }, { "epoch": 4.59, "learning_rate": 5.4128440366972475e-05, "loss": 2.6398, "step": 58500 }, { "epoch": 4.59, "learning_rate": 5.412765623774798e-05, "loss": 2.6713, "step": 58501 }, { "epoch": 4.59, "learning_rate": 5.412687210852348e-05, "loss": 2.7373, "step": 58502 }, { "epoch": 4.59, "learning_rate": 5.4126087979298986e-05, "loss": 2.823, "step": 58503 }, { "epoch": 4.59, "learning_rate": 5.41253038500745e-05, "loss": 2.7077, "step": 58504 }, { "epoch": 4.59, "learning_rate": 5.412451972085e-05, "loss": 2.6967, "step": 58505 }, { "epoch": 4.59, "learning_rate": 5.4123735591625504e-05, "loss": 2.6433, "step": 58506 }, { "epoch": 4.59, "learning_rate": 5.412295146240101e-05, "loss": 2.5628, "step": 58507 }, { "epoch": 4.59, "learning_rate": 5.412216733317651e-05, "loss": 2.9205, "step": 58508 }, { "epoch": 4.59, "learning_rate": 5.4121383203952015e-05, "loss": 2.7259, "step": 58509 }, { "epoch": 4.59, "learning_rate": 5.412059907472752e-05, "loss": 2.6433, "step": 58510 }, { "epoch": 4.59, "learning_rate": 5.411981494550302e-05, "loss": 2.8181, "step": 58511 }, { "epoch": 4.59, "learning_rate": 5.4119030816278526e-05, "loss": 2.5986, "step": 58512 }, { "epoch": 4.59, "learning_rate": 5.411824668705403e-05, "loss": 2.6649, "step": 58513 }, { "epoch": 4.59, "learning_rate": 5.4117462557829534e-05, "loss": 2.6524, "step": 58514 }, { "epoch": 4.59, "learning_rate": 5.411667842860504e-05, "loss": 2.8105, "step": 58515 }, { "epoch": 4.59, "learning_rate": 5.411589429938054e-05, "loss": 2.6811, "step": 58516 }, { "epoch": 4.59, "learning_rate": 5.4115110170156045e-05, "loss": 2.7409, "step": 58517 }, { "epoch": 4.59, "learning_rate": 5.411432604093155e-05, "loss": 2.5402, "step": 58518 }, { "epoch": 4.59, "learning_rate": 5.411354191170705e-05, "loss": 2.4943, "step": 58519 }, { "epoch": 4.59, "learning_rate": 5.411275778248256e-05, "loss": 2.6194, "step": 58520 }, { "epoch": 4.59, "learning_rate": 5.4111973653258066e-05, "loss": 2.687, "step": 58521 }, { "epoch": 4.59, "learning_rate": 5.411118952403357e-05, "loss": 2.6496, "step": 58522 }, { "epoch": 4.59, "learning_rate": 5.411040539480907e-05, "loss": 2.7517, "step": 58523 }, { "epoch": 4.59, "learning_rate": 5.410962126558458e-05, "loss": 2.8835, "step": 58524 }, { "epoch": 4.59, "learning_rate": 5.410883713636008e-05, "loss": 2.616, "step": 58525 }, { "epoch": 4.59, "learning_rate": 5.4108053007135584e-05, "loss": 2.6146, "step": 58526 }, { "epoch": 4.59, "learning_rate": 5.4107268877911075e-05, "loss": 2.5706, "step": 58527 }, { "epoch": 4.59, "learning_rate": 5.410648474868658e-05, "loss": 2.5555, "step": 58528 }, { "epoch": 4.59, "learning_rate": 5.410570061946208e-05, "loss": 2.783, "step": 58529 }, { "epoch": 4.59, "learning_rate": 5.410491649023759e-05, "loss": 2.6501, "step": 58530 }, { "epoch": 4.59, "learning_rate": 5.4104132361013096e-05, "loss": 2.6322, "step": 58531 }, { "epoch": 4.59, "learning_rate": 5.41033482317886e-05, "loss": 2.7419, "step": 58532 }, { "epoch": 4.59, "learning_rate": 5.41025641025641e-05, "loss": 2.8975, "step": 58533 }, { "epoch": 4.59, "learning_rate": 5.410177997333961e-05, "loss": 2.6437, "step": 58534 }, { "epoch": 4.59, "learning_rate": 5.410099584411511e-05, "loss": 2.6581, "step": 58535 }, { "epoch": 4.59, "learning_rate": 5.4100211714890614e-05, "loss": 2.5953, "step": 58536 }, { "epoch": 4.59, "learning_rate": 5.409942758566612e-05, "loss": 2.5312, "step": 58537 }, { "epoch": 4.59, "learning_rate": 5.409864345644162e-05, "loss": 2.6098, "step": 58538 }, { "epoch": 4.59, "learning_rate": 5.4097859327217125e-05, "loss": 2.7678, "step": 58539 }, { "epoch": 4.59, "learning_rate": 5.409707519799263e-05, "loss": 2.6928, "step": 58540 }, { "epoch": 4.59, "learning_rate": 5.409629106876813e-05, "loss": 2.7076, "step": 58541 }, { "epoch": 4.59, "learning_rate": 5.4095506939543636e-05, "loss": 2.6762, "step": 58542 }, { "epoch": 4.59, "learning_rate": 5.409472281031914e-05, "loss": 2.6163, "step": 58543 }, { "epoch": 4.59, "learning_rate": 5.4093938681094644e-05, "loss": 2.7344, "step": 58544 }, { "epoch": 4.59, "learning_rate": 5.4093154551870154e-05, "loss": 2.5267, "step": 58545 }, { "epoch": 4.59, "learning_rate": 5.409237042264566e-05, "loss": 2.8505, "step": 58546 }, { "epoch": 4.59, "learning_rate": 5.409158629342116e-05, "loss": 2.6542, "step": 58547 }, { "epoch": 4.59, "learning_rate": 5.4090802164196665e-05, "loss": 2.7344, "step": 58548 }, { "epoch": 4.59, "learning_rate": 5.409001803497217e-05, "loss": 2.5109, "step": 58549 }, { "epoch": 4.59, "learning_rate": 5.408923390574767e-05, "loss": 2.5831, "step": 58550 }, { "epoch": 4.59, "learning_rate": 5.4088449776523176e-05, "loss": 2.4778, "step": 58551 }, { "epoch": 4.59, "learning_rate": 5.408766564729868e-05, "loss": 2.8453, "step": 58552 }, { "epoch": 4.59, "learning_rate": 5.4086881518074184e-05, "loss": 2.7444, "step": 58553 }, { "epoch": 4.59, "learning_rate": 5.408609738884969e-05, "loss": 2.5132, "step": 58554 }, { "epoch": 4.59, "learning_rate": 5.408531325962519e-05, "loss": 2.7582, "step": 58555 }, { "epoch": 4.59, "learning_rate": 5.4084529130400695e-05, "loss": 2.6073, "step": 58556 }, { "epoch": 4.59, "learning_rate": 5.40837450011762e-05, "loss": 2.5171, "step": 58557 }, { "epoch": 4.59, "learning_rate": 5.40829608719517e-05, "loss": 2.6016, "step": 58558 }, { "epoch": 4.59, "learning_rate": 5.4082176742727206e-05, "loss": 2.7231, "step": 58559 }, { "epoch": 4.59, "learning_rate": 5.408139261350271e-05, "loss": 2.6284, "step": 58560 }, { "epoch": 4.59, "learning_rate": 5.408060848427822e-05, "loss": 2.7328, "step": 58561 }, { "epoch": 4.59, "learning_rate": 5.4079824355053724e-05, "loss": 2.7409, "step": 58562 }, { "epoch": 4.59, "learning_rate": 5.4079040225829214e-05, "loss": 2.7025, "step": 58563 }, { "epoch": 4.59, "learning_rate": 5.407825609660472e-05, "loss": 2.5908, "step": 58564 }, { "epoch": 4.59, "learning_rate": 5.407747196738022e-05, "loss": 2.7623, "step": 58565 }, { "epoch": 4.59, "learning_rate": 5.4076687838155725e-05, "loss": 2.6758, "step": 58566 }, { "epoch": 4.59, "learning_rate": 5.407590370893123e-05, "loss": 2.6322, "step": 58567 }, { "epoch": 4.59, "learning_rate": 5.407511957970673e-05, "loss": 2.654, "step": 58568 }, { "epoch": 4.59, "learning_rate": 5.4074335450482236e-05, "loss": 2.8094, "step": 58569 }, { "epoch": 4.59, "learning_rate": 5.407355132125774e-05, "loss": 2.7132, "step": 58570 }, { "epoch": 4.59, "learning_rate": 5.407276719203325e-05, "loss": 2.7751, "step": 58571 }, { "epoch": 4.59, "learning_rate": 5.4071983062808754e-05, "loss": 2.5728, "step": 58572 }, { "epoch": 4.59, "learning_rate": 5.407119893358426e-05, "loss": 2.6205, "step": 58573 }, { "epoch": 4.59, "learning_rate": 5.407041480435976e-05, "loss": 2.8318, "step": 58574 }, { "epoch": 4.59, "learning_rate": 5.4069630675135265e-05, "loss": 2.7125, "step": 58575 }, { "epoch": 4.59, "learning_rate": 5.406884654591077e-05, "loss": 2.4922, "step": 58576 }, { "epoch": 4.59, "learning_rate": 5.406806241668627e-05, "loss": 2.7602, "step": 58577 }, { "epoch": 4.59, "learning_rate": 5.4067278287461776e-05, "loss": 2.5123, "step": 58578 }, { "epoch": 4.59, "learning_rate": 5.406649415823728e-05, "loss": 2.7148, "step": 58579 }, { "epoch": 4.59, "learning_rate": 5.406571002901278e-05, "loss": 2.8038, "step": 58580 }, { "epoch": 4.59, "learning_rate": 5.406492589978829e-05, "loss": 2.6274, "step": 58581 }, { "epoch": 4.59, "learning_rate": 5.406414177056379e-05, "loss": 2.6312, "step": 58582 }, { "epoch": 4.59, "learning_rate": 5.4063357641339294e-05, "loss": 2.6516, "step": 58583 }, { "epoch": 4.59, "learning_rate": 5.40625735121148e-05, "loss": 2.5789, "step": 58584 }, { "epoch": 4.59, "learning_rate": 5.40617893828903e-05, "loss": 2.6568, "step": 58585 }, { "epoch": 4.59, "learning_rate": 5.406100525366581e-05, "loss": 2.6017, "step": 58586 }, { "epoch": 4.59, "learning_rate": 5.4060221124441316e-05, "loss": 2.8164, "step": 58587 }, { "epoch": 4.59, "learning_rate": 5.405943699521682e-05, "loss": 2.6356, "step": 58588 }, { "epoch": 4.59, "learning_rate": 5.405865286599232e-05, "loss": 2.5886, "step": 58589 }, { "epoch": 4.59, "learning_rate": 5.4057868736767827e-05, "loss": 2.6123, "step": 58590 }, { "epoch": 4.59, "learning_rate": 5.405708460754333e-05, "loss": 2.6324, "step": 58591 }, { "epoch": 4.59, "learning_rate": 5.4056300478318834e-05, "loss": 2.4663, "step": 58592 }, { "epoch": 4.59, "learning_rate": 5.405551634909434e-05, "loss": 2.8273, "step": 58593 }, { "epoch": 4.59, "learning_rate": 5.405473221986984e-05, "loss": 2.6116, "step": 58594 }, { "epoch": 4.59, "learning_rate": 5.4053948090645345e-05, "loss": 2.6416, "step": 58595 }, { "epoch": 4.59, "learning_rate": 5.405316396142085e-05, "loss": 2.7313, "step": 58596 }, { "epoch": 4.59, "learning_rate": 5.405237983219635e-05, "loss": 2.7248, "step": 58597 }, { "epoch": 4.59, "learning_rate": 5.4051595702971856e-05, "loss": 2.6602, "step": 58598 }, { "epoch": 4.59, "learning_rate": 5.405081157374735e-05, "loss": 2.6331, "step": 58599 }, { "epoch": 4.59, "learning_rate": 5.4050027444522857e-05, "loss": 2.5596, "step": 58600 }, { "epoch": 4.6, "learning_rate": 5.404924331529836e-05, "loss": 2.5397, "step": 58601 }, { "epoch": 4.6, "learning_rate": 5.4048459186073864e-05, "loss": 2.54, "step": 58602 }, { "epoch": 4.6, "learning_rate": 5.404767505684937e-05, "loss": 2.624, "step": 58603 }, { "epoch": 4.6, "learning_rate": 5.404689092762487e-05, "loss": 2.6272, "step": 58604 }, { "epoch": 4.6, "learning_rate": 5.4046106798400375e-05, "loss": 2.7789, "step": 58605 }, { "epoch": 4.6, "learning_rate": 5.404532266917588e-05, "loss": 2.5617, "step": 58606 }, { "epoch": 4.6, "learning_rate": 5.404453853995138e-05, "loss": 2.7536, "step": 58607 }, { "epoch": 4.6, "learning_rate": 5.4043754410726886e-05, "loss": 2.6886, "step": 58608 }, { "epoch": 4.6, "learning_rate": 5.404297028150239e-05, "loss": 2.6696, "step": 58609 }, { "epoch": 4.6, "learning_rate": 5.404218615227789e-05, "loss": 2.7231, "step": 58610 }, { "epoch": 4.6, "learning_rate": 5.40414020230534e-05, "loss": 2.5394, "step": 58611 }, { "epoch": 4.6, "learning_rate": 5.404061789382891e-05, "loss": 2.9199, "step": 58612 }, { "epoch": 4.6, "learning_rate": 5.403983376460441e-05, "loss": 2.8127, "step": 58613 }, { "epoch": 4.6, "learning_rate": 5.4039049635379915e-05, "loss": 2.7209, "step": 58614 }, { "epoch": 4.6, "learning_rate": 5.403826550615542e-05, "loss": 2.6335, "step": 58615 }, { "epoch": 4.6, "learning_rate": 5.403748137693092e-05, "loss": 2.6282, "step": 58616 }, { "epoch": 4.6, "learning_rate": 5.4036697247706426e-05, "loss": 2.653, "step": 58617 }, { "epoch": 4.6, "learning_rate": 5.403591311848193e-05, "loss": 2.446, "step": 58618 }, { "epoch": 4.6, "learning_rate": 5.403512898925743e-05, "loss": 2.7612, "step": 58619 }, { "epoch": 4.6, "learning_rate": 5.403434486003294e-05, "loss": 2.6489, "step": 58620 }, { "epoch": 4.6, "learning_rate": 5.403356073080844e-05, "loss": 2.633, "step": 58621 }, { "epoch": 4.6, "learning_rate": 5.4032776601583944e-05, "loss": 2.6298, "step": 58622 }, { "epoch": 4.6, "learning_rate": 5.403199247235945e-05, "loss": 2.6923, "step": 58623 }, { "epoch": 4.6, "learning_rate": 5.403120834313495e-05, "loss": 2.3961, "step": 58624 }, { "epoch": 4.6, "learning_rate": 5.4030424213910455e-05, "loss": 2.5878, "step": 58625 }, { "epoch": 4.6, "learning_rate": 5.402964008468596e-05, "loss": 2.6297, "step": 58626 }, { "epoch": 4.6, "learning_rate": 5.402885595546146e-05, "loss": 2.5386, "step": 58627 }, { "epoch": 4.6, "learning_rate": 5.402807182623697e-05, "loss": 2.4904, "step": 58628 }, { "epoch": 4.6, "learning_rate": 5.402728769701248e-05, "loss": 2.6208, "step": 58629 }, { "epoch": 4.6, "learning_rate": 5.402650356778798e-05, "loss": 2.6388, "step": 58630 }, { "epoch": 4.6, "learning_rate": 5.4025719438563484e-05, "loss": 2.9366, "step": 58631 }, { "epoch": 4.6, "learning_rate": 5.402493530933899e-05, "loss": 2.4286, "step": 58632 }, { "epoch": 4.6, "learning_rate": 5.402415118011449e-05, "loss": 2.9139, "step": 58633 }, { "epoch": 4.6, "learning_rate": 5.4023367050889995e-05, "loss": 2.7647, "step": 58634 }, { "epoch": 4.6, "learning_rate": 5.4022582921665485e-05, "loss": 2.6989, "step": 58635 }, { "epoch": 4.6, "learning_rate": 5.402179879244099e-05, "loss": 2.7416, "step": 58636 }, { "epoch": 4.6, "learning_rate": 5.402101466321649e-05, "loss": 2.6652, "step": 58637 }, { "epoch": 4.6, "learning_rate": 5.4020230533992e-05, "loss": 2.6514, "step": 58638 }, { "epoch": 4.6, "learning_rate": 5.401944640476751e-05, "loss": 2.6029, "step": 58639 }, { "epoch": 4.6, "learning_rate": 5.401866227554301e-05, "loss": 2.6533, "step": 58640 }, { "epoch": 4.6, "learning_rate": 5.4017878146318514e-05, "loss": 2.581, "step": 58641 }, { "epoch": 4.6, "learning_rate": 5.401709401709402e-05, "loss": 2.7913, "step": 58642 }, { "epoch": 4.6, "learning_rate": 5.401630988786952e-05, "loss": 2.6354, "step": 58643 }, { "epoch": 4.6, "learning_rate": 5.4015525758645025e-05, "loss": 2.6181, "step": 58644 }, { "epoch": 4.6, "learning_rate": 5.401474162942053e-05, "loss": 2.5598, "step": 58645 }, { "epoch": 4.6, "learning_rate": 5.401395750019603e-05, "loss": 2.7124, "step": 58646 }, { "epoch": 4.6, "learning_rate": 5.4013173370971536e-05, "loss": 2.6584, "step": 58647 }, { "epoch": 4.6, "learning_rate": 5.401238924174704e-05, "loss": 2.7918, "step": 58648 }, { "epoch": 4.6, "learning_rate": 5.4011605112522544e-05, "loss": 2.724, "step": 58649 }, { "epoch": 4.6, "learning_rate": 5.401082098329805e-05, "loss": 2.7515, "step": 58650 }, { "epoch": 4.6, "learning_rate": 5.401003685407355e-05, "loss": 2.5017, "step": 58651 }, { "epoch": 4.6, "learning_rate": 5.4009252724849055e-05, "loss": 2.6331, "step": 58652 }, { "epoch": 4.6, "learning_rate": 5.4008468595624565e-05, "loss": 2.7297, "step": 58653 }, { "epoch": 4.6, "learning_rate": 5.400768446640007e-05, "loss": 2.7212, "step": 58654 }, { "epoch": 4.6, "learning_rate": 5.400690033717557e-05, "loss": 2.6278, "step": 58655 }, { "epoch": 4.6, "learning_rate": 5.4006116207951076e-05, "loss": 2.8517, "step": 58656 }, { "epoch": 4.6, "learning_rate": 5.400533207872658e-05, "loss": 2.4784, "step": 58657 }, { "epoch": 4.6, "learning_rate": 5.4004547949502083e-05, "loss": 2.8132, "step": 58658 }, { "epoch": 4.6, "learning_rate": 5.400376382027759e-05, "loss": 2.5789, "step": 58659 }, { "epoch": 4.6, "learning_rate": 5.400297969105309e-05, "loss": 2.5577, "step": 58660 }, { "epoch": 4.6, "learning_rate": 5.4002195561828594e-05, "loss": 2.7524, "step": 58661 }, { "epoch": 4.6, "learning_rate": 5.40014114326041e-05, "loss": 2.5866, "step": 58662 }, { "epoch": 4.6, "learning_rate": 5.40006273033796e-05, "loss": 2.7216, "step": 58663 }, { "epoch": 4.6, "learning_rate": 5.3999843174155105e-05, "loss": 2.8192, "step": 58664 }, { "epoch": 4.6, "learning_rate": 5.399905904493061e-05, "loss": 2.6077, "step": 58665 }, { "epoch": 4.6, "learning_rate": 5.399827491570611e-05, "loss": 2.6655, "step": 58666 }, { "epoch": 4.6, "learning_rate": 5.3997490786481617e-05, "loss": 2.779, "step": 58667 }, { "epoch": 4.6, "learning_rate": 5.399670665725712e-05, "loss": 2.6882, "step": 58668 }, { "epoch": 4.6, "learning_rate": 5.399592252803263e-05, "loss": 2.6276, "step": 58669 }, { "epoch": 4.6, "learning_rate": 5.3995138398808134e-05, "loss": 2.6814, "step": 58670 }, { "epoch": 4.6, "learning_rate": 5.3994354269583624e-05, "loss": 2.6149, "step": 58671 }, { "epoch": 4.6, "learning_rate": 5.399357014035913e-05, "loss": 2.8226, "step": 58672 }, { "epoch": 4.6, "learning_rate": 5.399278601113463e-05, "loss": 2.7227, "step": 58673 }, { "epoch": 4.6, "learning_rate": 5.3992001881910135e-05, "loss": 2.6825, "step": 58674 }, { "epoch": 4.6, "learning_rate": 5.399121775268564e-05, "loss": 2.7059, "step": 58675 }, { "epoch": 4.6, "learning_rate": 5.399043362346114e-05, "loss": 2.5292, "step": 58676 }, { "epoch": 4.6, "learning_rate": 5.3989649494236647e-05, "loss": 2.5591, "step": 58677 }, { "epoch": 4.6, "learning_rate": 5.398886536501215e-05, "loss": 2.7327, "step": 58678 }, { "epoch": 4.6, "learning_rate": 5.398808123578766e-05, "loss": 2.5839, "step": 58679 }, { "epoch": 4.6, "learning_rate": 5.3987297106563164e-05, "loss": 2.5211, "step": 58680 }, { "epoch": 4.6, "learning_rate": 5.398651297733867e-05, "loss": 2.7127, "step": 58681 }, { "epoch": 4.6, "learning_rate": 5.398572884811417e-05, "loss": 2.6231, "step": 58682 }, { "epoch": 4.6, "learning_rate": 5.3984944718889675e-05, "loss": 2.7331, "step": 58683 }, { "epoch": 4.6, "learning_rate": 5.398416058966518e-05, "loss": 2.7092, "step": 58684 }, { "epoch": 4.6, "learning_rate": 5.398337646044068e-05, "loss": 2.5728, "step": 58685 }, { "epoch": 4.6, "learning_rate": 5.3982592331216186e-05, "loss": 2.8065, "step": 58686 }, { "epoch": 4.6, "learning_rate": 5.398180820199169e-05, "loss": 2.6093, "step": 58687 }, { "epoch": 4.6, "learning_rate": 5.3981024072767194e-05, "loss": 2.6733, "step": 58688 }, { "epoch": 4.6, "learning_rate": 5.39802399435427e-05, "loss": 2.7297, "step": 58689 }, { "epoch": 4.6, "learning_rate": 5.39794558143182e-05, "loss": 2.6027, "step": 58690 }, { "epoch": 4.6, "learning_rate": 5.3978671685093705e-05, "loss": 2.7268, "step": 58691 }, { "epoch": 4.6, "learning_rate": 5.397788755586921e-05, "loss": 2.6561, "step": 58692 }, { "epoch": 4.6, "learning_rate": 5.397710342664471e-05, "loss": 2.6587, "step": 58693 }, { "epoch": 4.6, "learning_rate": 5.3976319297420216e-05, "loss": 2.7847, "step": 58694 }, { "epoch": 4.6, "learning_rate": 5.3975535168195726e-05, "loss": 2.6907, "step": 58695 }, { "epoch": 4.6, "learning_rate": 5.397475103897123e-05, "loss": 2.7575, "step": 58696 }, { "epoch": 4.6, "learning_rate": 5.3973966909746734e-05, "loss": 2.5655, "step": 58697 }, { "epoch": 4.6, "learning_rate": 5.397318278052224e-05, "loss": 2.7034, "step": 58698 }, { "epoch": 4.6, "learning_rate": 5.397239865129774e-05, "loss": 2.5454, "step": 58699 }, { "epoch": 4.6, "learning_rate": 5.3971614522073245e-05, "loss": 2.7374, "step": 58700 }, { "epoch": 4.6, "learning_rate": 5.397083039284875e-05, "loss": 2.8002, "step": 58701 }, { "epoch": 4.6, "learning_rate": 5.397004626362425e-05, "loss": 2.7423, "step": 58702 }, { "epoch": 4.6, "learning_rate": 5.3969262134399756e-05, "loss": 2.5126, "step": 58703 }, { "epoch": 4.6, "learning_rate": 5.396847800517526e-05, "loss": 2.5706, "step": 58704 }, { "epoch": 4.6, "learning_rate": 5.396769387595076e-05, "loss": 2.5701, "step": 58705 }, { "epoch": 4.6, "learning_rate": 5.396690974672627e-05, "loss": 2.5401, "step": 58706 }, { "epoch": 4.6, "learning_rate": 5.3966125617501764e-05, "loss": 2.6489, "step": 58707 }, { "epoch": 4.6, "learning_rate": 5.396534148827727e-05, "loss": 2.7093, "step": 58708 }, { "epoch": 4.6, "learning_rate": 5.396455735905277e-05, "loss": 2.6829, "step": 58709 }, { "epoch": 4.6, "learning_rate": 5.3963773229828275e-05, "loss": 2.7232, "step": 58710 }, { "epoch": 4.6, "learning_rate": 5.396298910060378e-05, "loss": 2.7046, "step": 58711 }, { "epoch": 4.6, "learning_rate": 5.396220497137928e-05, "loss": 2.66, "step": 58712 }, { "epoch": 4.6, "learning_rate": 5.3961420842154786e-05, "loss": 2.5863, "step": 58713 }, { "epoch": 4.6, "learning_rate": 5.396063671293029e-05, "loss": 2.6623, "step": 58714 }, { "epoch": 4.6, "learning_rate": 5.395985258370579e-05, "loss": 2.5655, "step": 58715 }, { "epoch": 4.6, "learning_rate": 5.39590684544813e-05, "loss": 2.6302, "step": 58716 }, { "epoch": 4.6, "learning_rate": 5.39582843252568e-05, "loss": 2.6454, "step": 58717 }, { "epoch": 4.6, "learning_rate": 5.3957500196032304e-05, "loss": 2.6466, "step": 58718 }, { "epoch": 4.6, "learning_rate": 5.395671606680781e-05, "loss": 2.6227, "step": 58719 }, { "epoch": 4.6, "learning_rate": 5.395593193758332e-05, "loss": 2.6595, "step": 58720 }, { "epoch": 4.6, "learning_rate": 5.395514780835882e-05, "loss": 2.7587, "step": 58721 }, { "epoch": 4.6, "learning_rate": 5.3954363679134326e-05, "loss": 2.9222, "step": 58722 }, { "epoch": 4.6, "learning_rate": 5.395357954990983e-05, "loss": 2.7767, "step": 58723 }, { "epoch": 4.6, "learning_rate": 5.395279542068533e-05, "loss": 2.6553, "step": 58724 }, { "epoch": 4.6, "learning_rate": 5.3952011291460837e-05, "loss": 2.5683, "step": 58725 }, { "epoch": 4.6, "learning_rate": 5.395122716223634e-05, "loss": 2.6182, "step": 58726 }, { "epoch": 4.6, "learning_rate": 5.3950443033011844e-05, "loss": 2.623, "step": 58727 }, { "epoch": 4.61, "learning_rate": 5.394965890378735e-05, "loss": 2.6321, "step": 58728 }, { "epoch": 4.61, "learning_rate": 5.394887477456285e-05, "loss": 2.6041, "step": 58729 }, { "epoch": 4.61, "learning_rate": 5.3948090645338355e-05, "loss": 2.7131, "step": 58730 }, { "epoch": 4.61, "learning_rate": 5.394730651611386e-05, "loss": 2.4996, "step": 58731 }, { "epoch": 4.61, "learning_rate": 5.394652238688936e-05, "loss": 2.7509, "step": 58732 }, { "epoch": 4.61, "learning_rate": 5.3945738257664866e-05, "loss": 2.659, "step": 58733 }, { "epoch": 4.61, "learning_rate": 5.394495412844037e-05, "loss": 2.5553, "step": 58734 }, { "epoch": 4.61, "learning_rate": 5.394416999921587e-05, "loss": 2.6219, "step": 58735 }, { "epoch": 4.61, "learning_rate": 5.3943385869991384e-05, "loss": 2.7742, "step": 58736 }, { "epoch": 4.61, "learning_rate": 5.394260174076689e-05, "loss": 2.6403, "step": 58737 }, { "epoch": 4.61, "learning_rate": 5.394181761154239e-05, "loss": 2.7148, "step": 58738 }, { "epoch": 4.61, "learning_rate": 5.3941033482317895e-05, "loss": 2.5658, "step": 58739 }, { "epoch": 4.61, "learning_rate": 5.39402493530934e-05, "loss": 2.5739, "step": 58740 }, { "epoch": 4.61, "learning_rate": 5.39394652238689e-05, "loss": 2.5241, "step": 58741 }, { "epoch": 4.61, "learning_rate": 5.3938681094644406e-05, "loss": 2.7729, "step": 58742 }, { "epoch": 4.61, "learning_rate": 5.3937896965419896e-05, "loss": 2.8005, "step": 58743 }, { "epoch": 4.61, "learning_rate": 5.39371128361954e-05, "loss": 2.6738, "step": 58744 }, { "epoch": 4.61, "learning_rate": 5.39363287069709e-05, "loss": 2.8109, "step": 58745 }, { "epoch": 4.61, "learning_rate": 5.3935544577746414e-05, "loss": 2.6215, "step": 58746 }, { "epoch": 4.61, "learning_rate": 5.393476044852192e-05, "loss": 2.6328, "step": 58747 }, { "epoch": 4.61, "learning_rate": 5.393397631929742e-05, "loss": 2.472, "step": 58748 }, { "epoch": 4.61, "learning_rate": 5.3933192190072925e-05, "loss": 2.544, "step": 58749 }, { "epoch": 4.61, "learning_rate": 5.393240806084843e-05, "loss": 2.8354, "step": 58750 }, { "epoch": 4.61, "learning_rate": 5.393162393162393e-05, "loss": 2.5924, "step": 58751 }, { "epoch": 4.61, "learning_rate": 5.3930839802399436e-05, "loss": 2.5535, "step": 58752 }, { "epoch": 4.61, "learning_rate": 5.393005567317494e-05, "loss": 2.7307, "step": 58753 }, { "epoch": 4.61, "learning_rate": 5.392927154395044e-05, "loss": 2.5443, "step": 58754 }, { "epoch": 4.61, "learning_rate": 5.392848741472595e-05, "loss": 2.7526, "step": 58755 }, { "epoch": 4.61, "learning_rate": 5.392770328550145e-05, "loss": 2.6617, "step": 58756 }, { "epoch": 4.61, "learning_rate": 5.3926919156276954e-05, "loss": 2.7298, "step": 58757 }, { "epoch": 4.61, "learning_rate": 5.392613502705246e-05, "loss": 2.4957, "step": 58758 }, { "epoch": 4.61, "learning_rate": 5.392535089782796e-05, "loss": 2.5108, "step": 58759 }, { "epoch": 4.61, "learning_rate": 5.3924566768603465e-05, "loss": 2.727, "step": 58760 }, { "epoch": 4.61, "learning_rate": 5.3923782639378976e-05, "loss": 2.5398, "step": 58761 }, { "epoch": 4.61, "learning_rate": 5.392299851015448e-05, "loss": 2.681, "step": 58762 }, { "epoch": 4.61, "learning_rate": 5.392221438092998e-05, "loss": 2.6775, "step": 58763 }, { "epoch": 4.61, "learning_rate": 5.392143025170549e-05, "loss": 2.8093, "step": 58764 }, { "epoch": 4.61, "learning_rate": 5.392064612248099e-05, "loss": 2.5516, "step": 58765 }, { "epoch": 4.61, "learning_rate": 5.3919861993256494e-05, "loss": 2.8122, "step": 58766 }, { "epoch": 4.61, "learning_rate": 5.3919077864032e-05, "loss": 2.7407, "step": 58767 }, { "epoch": 4.61, "learning_rate": 5.39182937348075e-05, "loss": 2.6095, "step": 58768 }, { "epoch": 4.61, "learning_rate": 5.3917509605583005e-05, "loss": 2.3968, "step": 58769 }, { "epoch": 4.61, "learning_rate": 5.391672547635851e-05, "loss": 2.7413, "step": 58770 }, { "epoch": 4.61, "learning_rate": 5.391594134713401e-05, "loss": 2.9896, "step": 58771 }, { "epoch": 4.61, "learning_rate": 5.3915157217909516e-05, "loss": 2.5958, "step": 58772 }, { "epoch": 4.61, "learning_rate": 5.391437308868502e-05, "loss": 2.6952, "step": 58773 }, { "epoch": 4.61, "learning_rate": 5.3913588959460524e-05, "loss": 2.8384, "step": 58774 }, { "epoch": 4.61, "learning_rate": 5.391280483023603e-05, "loss": 2.6829, "step": 58775 }, { "epoch": 4.61, "learning_rate": 5.391202070101153e-05, "loss": 2.6535, "step": 58776 }, { "epoch": 4.61, "learning_rate": 5.391123657178704e-05, "loss": 2.659, "step": 58777 }, { "epoch": 4.61, "learning_rate": 5.3910452442562545e-05, "loss": 2.5378, "step": 58778 }, { "epoch": 4.61, "learning_rate": 5.3909668313338035e-05, "loss": 2.6016, "step": 58779 }, { "epoch": 4.61, "learning_rate": 5.390888418411354e-05, "loss": 2.5839, "step": 58780 }, { "epoch": 4.61, "learning_rate": 5.390810005488904e-05, "loss": 2.6489, "step": 58781 }, { "epoch": 4.61, "learning_rate": 5.3907315925664546e-05, "loss": 2.5293, "step": 58782 }, { "epoch": 4.61, "learning_rate": 5.390653179644005e-05, "loss": 2.889, "step": 58783 }, { "epoch": 4.61, "learning_rate": 5.3905747667215554e-05, "loss": 2.5806, "step": 58784 }, { "epoch": 4.61, "learning_rate": 5.390496353799106e-05, "loss": 2.5138, "step": 58785 }, { "epoch": 4.61, "learning_rate": 5.390417940876656e-05, "loss": 2.9441, "step": 58786 }, { "epoch": 4.61, "learning_rate": 5.390339527954207e-05, "loss": 2.6098, "step": 58787 }, { "epoch": 4.61, "learning_rate": 5.3902611150317575e-05, "loss": 2.5108, "step": 58788 }, { "epoch": 4.61, "learning_rate": 5.390182702109308e-05, "loss": 2.5208, "step": 58789 }, { "epoch": 4.61, "learning_rate": 5.390104289186858e-05, "loss": 2.5815, "step": 58790 }, { "epoch": 4.61, "learning_rate": 5.3900258762644086e-05, "loss": 2.6248, "step": 58791 }, { "epoch": 4.61, "learning_rate": 5.389947463341959e-05, "loss": 2.5793, "step": 58792 }, { "epoch": 4.61, "learning_rate": 5.3898690504195093e-05, "loss": 2.8703, "step": 58793 }, { "epoch": 4.61, "learning_rate": 5.38979063749706e-05, "loss": 2.6769, "step": 58794 }, { "epoch": 4.61, "learning_rate": 5.38971222457461e-05, "loss": 2.6775, "step": 58795 }, { "epoch": 4.61, "learning_rate": 5.3896338116521605e-05, "loss": 2.4165, "step": 58796 }, { "epoch": 4.61, "learning_rate": 5.389555398729711e-05, "loss": 2.8225, "step": 58797 }, { "epoch": 4.61, "learning_rate": 5.389476985807261e-05, "loss": 2.5866, "step": 58798 }, { "epoch": 4.61, "learning_rate": 5.3893985728848116e-05, "loss": 2.6882, "step": 58799 }, { "epoch": 4.61, "learning_rate": 5.389320159962362e-05, "loss": 2.8527, "step": 58800 }, { "epoch": 4.61, "learning_rate": 5.389241747039912e-05, "loss": 2.6586, "step": 58801 }, { "epoch": 4.61, "learning_rate": 5.3891633341174627e-05, "loss": 2.9681, "step": 58802 }, { "epoch": 4.61, "learning_rate": 5.389084921195014e-05, "loss": 2.6799, "step": 58803 }, { "epoch": 4.61, "learning_rate": 5.389006508272564e-05, "loss": 2.5812, "step": 58804 }, { "epoch": 4.61, "learning_rate": 5.3889280953501144e-05, "loss": 2.4374, "step": 58805 }, { "epoch": 4.61, "learning_rate": 5.388849682427665e-05, "loss": 2.6789, "step": 58806 }, { "epoch": 4.61, "learning_rate": 5.388771269505215e-05, "loss": 2.6043, "step": 58807 }, { "epoch": 4.61, "learning_rate": 5.3886928565827655e-05, "loss": 2.4633, "step": 58808 }, { "epoch": 4.61, "learning_rate": 5.388614443660316e-05, "loss": 2.5634, "step": 58809 }, { "epoch": 4.61, "learning_rate": 5.388536030737866e-05, "loss": 2.6659, "step": 58810 }, { "epoch": 4.61, "learning_rate": 5.3884576178154166e-05, "loss": 2.7173, "step": 58811 }, { "epoch": 4.61, "learning_rate": 5.388379204892967e-05, "loss": 2.7212, "step": 58812 }, { "epoch": 4.61, "learning_rate": 5.3883007919705174e-05, "loss": 2.6899, "step": 58813 }, { "epoch": 4.61, "learning_rate": 5.388222379048068e-05, "loss": 2.534, "step": 58814 }, { "epoch": 4.61, "learning_rate": 5.3881439661256174e-05, "loss": 2.5409, "step": 58815 }, { "epoch": 4.61, "learning_rate": 5.388065553203168e-05, "loss": 2.7764, "step": 58816 }, { "epoch": 4.61, "learning_rate": 5.387987140280718e-05, "loss": 2.7618, "step": 58817 }, { "epoch": 4.61, "learning_rate": 5.3879087273582685e-05, "loss": 2.6357, "step": 58818 }, { "epoch": 4.61, "learning_rate": 5.387830314435819e-05, "loss": 2.7175, "step": 58819 }, { "epoch": 4.61, "learning_rate": 5.387751901513369e-05, "loss": 2.6826, "step": 58820 }, { "epoch": 4.61, "learning_rate": 5.3876734885909196e-05, "loss": 2.6385, "step": 58821 }, { "epoch": 4.61, "learning_rate": 5.38759507566847e-05, "loss": 2.6439, "step": 58822 }, { "epoch": 4.61, "learning_rate": 5.3875166627460204e-05, "loss": 2.6845, "step": 58823 }, { "epoch": 4.61, "learning_rate": 5.387438249823571e-05, "loss": 2.6241, "step": 58824 }, { "epoch": 4.61, "learning_rate": 5.387359836901121e-05, "loss": 2.6084, "step": 58825 }, { "epoch": 4.61, "learning_rate": 5.3872814239786715e-05, "loss": 2.7334, "step": 58826 }, { "epoch": 4.61, "learning_rate": 5.387203011056222e-05, "loss": 2.8463, "step": 58827 }, { "epoch": 4.61, "learning_rate": 5.387124598133773e-05, "loss": 2.4779, "step": 58828 }, { "epoch": 4.61, "learning_rate": 5.387046185211323e-05, "loss": 2.6094, "step": 58829 }, { "epoch": 4.61, "learning_rate": 5.3869677722888736e-05, "loss": 2.9168, "step": 58830 }, { "epoch": 4.61, "learning_rate": 5.386889359366424e-05, "loss": 2.6741, "step": 58831 }, { "epoch": 4.61, "learning_rate": 5.3868109464439744e-05, "loss": 2.647, "step": 58832 }, { "epoch": 4.61, "learning_rate": 5.386732533521525e-05, "loss": 2.7238, "step": 58833 }, { "epoch": 4.61, "learning_rate": 5.386654120599075e-05, "loss": 2.6511, "step": 58834 }, { "epoch": 4.61, "learning_rate": 5.3865757076766255e-05, "loss": 2.7031, "step": 58835 }, { "epoch": 4.61, "learning_rate": 5.386497294754176e-05, "loss": 2.5681, "step": 58836 }, { "epoch": 4.61, "learning_rate": 5.386418881831726e-05, "loss": 2.8652, "step": 58837 }, { "epoch": 4.61, "learning_rate": 5.3863404689092766e-05, "loss": 2.7964, "step": 58838 }, { "epoch": 4.61, "learning_rate": 5.386262055986827e-05, "loss": 2.7169, "step": 58839 }, { "epoch": 4.61, "learning_rate": 5.386183643064377e-05, "loss": 2.7856, "step": 58840 }, { "epoch": 4.61, "learning_rate": 5.386105230141928e-05, "loss": 2.6627, "step": 58841 }, { "epoch": 4.61, "learning_rate": 5.386026817219478e-05, "loss": 2.7128, "step": 58842 }, { "epoch": 4.61, "learning_rate": 5.3859484042970284e-05, "loss": 2.6344, "step": 58843 }, { "epoch": 4.61, "learning_rate": 5.3858699913745795e-05, "loss": 2.6161, "step": 58844 }, { "epoch": 4.61, "learning_rate": 5.38579157845213e-05, "loss": 2.4989, "step": 58845 }, { "epoch": 4.61, "learning_rate": 5.38571316552968e-05, "loss": 2.889, "step": 58846 }, { "epoch": 4.61, "learning_rate": 5.3856347526072306e-05, "loss": 2.7974, "step": 58847 }, { "epoch": 4.61, "learning_rate": 5.385556339684781e-05, "loss": 2.7709, "step": 58848 }, { "epoch": 4.61, "learning_rate": 5.385477926762331e-05, "loss": 2.5368, "step": 58849 }, { "epoch": 4.61, "learning_rate": 5.385399513839882e-05, "loss": 2.7415, "step": 58850 }, { "epoch": 4.61, "learning_rate": 5.385321100917431e-05, "loss": 2.6358, "step": 58851 }, { "epoch": 4.61, "learning_rate": 5.385242687994981e-05, "loss": 2.5518, "step": 58852 }, { "epoch": 4.61, "learning_rate": 5.3851642750725314e-05, "loss": 2.718, "step": 58853 }, { "epoch": 4.61, "learning_rate": 5.3850858621500825e-05, "loss": 2.7726, "step": 58854 }, { "epoch": 4.61, "learning_rate": 5.385007449227633e-05, "loss": 2.6314, "step": 58855 }, { "epoch": 4.62, "learning_rate": 5.384929036305183e-05, "loss": 2.5409, "step": 58856 }, { "epoch": 4.62, "learning_rate": 5.3848506233827336e-05, "loss": 2.5045, "step": 58857 }, { "epoch": 4.62, "learning_rate": 5.384772210460284e-05, "loss": 2.7054, "step": 58858 }, { "epoch": 4.62, "learning_rate": 5.384693797537834e-05, "loss": 2.7493, "step": 58859 }, { "epoch": 4.62, "learning_rate": 5.384615384615385e-05, "loss": 2.6259, "step": 58860 }, { "epoch": 4.62, "learning_rate": 5.384536971692935e-05, "loss": 2.6407, "step": 58861 }, { "epoch": 4.62, "learning_rate": 5.3844585587704854e-05, "loss": 2.8991, "step": 58862 }, { "epoch": 4.62, "learning_rate": 5.384380145848036e-05, "loss": 2.5823, "step": 58863 }, { "epoch": 4.62, "learning_rate": 5.384301732925586e-05, "loss": 2.568, "step": 58864 }, { "epoch": 4.62, "learning_rate": 5.3842233200031365e-05, "loss": 3.0788, "step": 58865 }, { "epoch": 4.62, "learning_rate": 5.384144907080687e-05, "loss": 2.7576, "step": 58866 }, { "epoch": 4.62, "learning_rate": 5.384066494158237e-05, "loss": 2.662, "step": 58867 }, { "epoch": 4.62, "learning_rate": 5.3839880812357876e-05, "loss": 2.7292, "step": 58868 }, { "epoch": 4.62, "learning_rate": 5.3839096683133387e-05, "loss": 2.7854, "step": 58869 }, { "epoch": 4.62, "learning_rate": 5.383831255390889e-05, "loss": 2.7344, "step": 58870 }, { "epoch": 4.62, "learning_rate": 5.3837528424684394e-05, "loss": 2.8292, "step": 58871 }, { "epoch": 4.62, "learning_rate": 5.38367442954599e-05, "loss": 2.6256, "step": 58872 }, { "epoch": 4.62, "learning_rate": 5.38359601662354e-05, "loss": 2.568, "step": 58873 }, { "epoch": 4.62, "learning_rate": 5.3835176037010905e-05, "loss": 2.7225, "step": 58874 }, { "epoch": 4.62, "learning_rate": 5.383439190778641e-05, "loss": 2.5799, "step": 58875 }, { "epoch": 4.62, "learning_rate": 5.383360777856191e-05, "loss": 2.6283, "step": 58876 }, { "epoch": 4.62, "learning_rate": 5.3832823649337416e-05, "loss": 2.5196, "step": 58877 }, { "epoch": 4.62, "learning_rate": 5.383203952011292e-05, "loss": 2.5447, "step": 58878 }, { "epoch": 4.62, "learning_rate": 5.383125539088842e-05, "loss": 2.5055, "step": 58879 }, { "epoch": 4.62, "learning_rate": 5.383047126166393e-05, "loss": 2.7545, "step": 58880 }, { "epoch": 4.62, "learning_rate": 5.382968713243943e-05, "loss": 2.5258, "step": 58881 }, { "epoch": 4.62, "learning_rate": 5.3828903003214934e-05, "loss": 2.6784, "step": 58882 }, { "epoch": 4.62, "learning_rate": 5.382811887399044e-05, "loss": 2.7331, "step": 58883 }, { "epoch": 4.62, "learning_rate": 5.382733474476594e-05, "loss": 2.7245, "step": 58884 }, { "epoch": 4.62, "learning_rate": 5.382655061554145e-05, "loss": 2.5892, "step": 58885 }, { "epoch": 4.62, "learning_rate": 5.3825766486316956e-05, "loss": 2.496, "step": 58886 }, { "epoch": 4.62, "learning_rate": 5.3824982357092446e-05, "loss": 2.9067, "step": 58887 }, { "epoch": 4.62, "learning_rate": 5.382419822786795e-05, "loss": 2.6033, "step": 58888 }, { "epoch": 4.62, "learning_rate": 5.382341409864345e-05, "loss": 2.6412, "step": 58889 }, { "epoch": 4.62, "learning_rate": 5.382262996941896e-05, "loss": 2.8714, "step": 58890 }, { "epoch": 4.62, "learning_rate": 5.382184584019446e-05, "loss": 2.8278, "step": 58891 }, { "epoch": 4.62, "learning_rate": 5.3821061710969964e-05, "loss": 2.826, "step": 58892 }, { "epoch": 4.62, "learning_rate": 5.382027758174547e-05, "loss": 2.7905, "step": 58893 }, { "epoch": 4.62, "learning_rate": 5.381949345252097e-05, "loss": 2.7293, "step": 58894 }, { "epoch": 4.62, "learning_rate": 5.381870932329648e-05, "loss": 2.6103, "step": 58895 }, { "epoch": 4.62, "learning_rate": 5.3817925194071986e-05, "loss": 2.7923, "step": 58896 }, { "epoch": 4.62, "learning_rate": 5.381714106484749e-05, "loss": 2.8275, "step": 58897 }, { "epoch": 4.62, "learning_rate": 5.381635693562299e-05, "loss": 2.7128, "step": 58898 }, { "epoch": 4.62, "learning_rate": 5.38155728063985e-05, "loss": 2.6644, "step": 58899 }, { "epoch": 4.62, "learning_rate": 5.3814788677174e-05, "loss": 2.7248, "step": 58900 }, { "epoch": 4.62, "learning_rate": 5.3814004547949504e-05, "loss": 2.8902, "step": 58901 }, { "epoch": 4.62, "learning_rate": 5.381322041872501e-05, "loss": 2.6648, "step": 58902 }, { "epoch": 4.62, "learning_rate": 5.381243628950051e-05, "loss": 2.647, "step": 58903 }, { "epoch": 4.62, "learning_rate": 5.3811652160276015e-05, "loss": 2.8041, "step": 58904 }, { "epoch": 4.62, "learning_rate": 5.381086803105152e-05, "loss": 2.5373, "step": 58905 }, { "epoch": 4.62, "learning_rate": 5.381008390182702e-05, "loss": 2.6906, "step": 58906 }, { "epoch": 4.62, "learning_rate": 5.3809299772602526e-05, "loss": 2.7337, "step": 58907 }, { "epoch": 4.62, "learning_rate": 5.380851564337803e-05, "loss": 2.7021, "step": 58908 }, { "epoch": 4.62, "learning_rate": 5.3807731514153534e-05, "loss": 2.982, "step": 58909 }, { "epoch": 4.62, "learning_rate": 5.380694738492904e-05, "loss": 2.9208, "step": 58910 }, { "epoch": 4.62, "learning_rate": 5.380616325570455e-05, "loss": 2.6125, "step": 58911 }, { "epoch": 4.62, "learning_rate": 5.380537912648005e-05, "loss": 2.5915, "step": 58912 }, { "epoch": 4.62, "learning_rate": 5.3804594997255555e-05, "loss": 2.7485, "step": 58913 }, { "epoch": 4.62, "learning_rate": 5.380381086803106e-05, "loss": 2.785, "step": 58914 }, { "epoch": 4.62, "learning_rate": 5.380302673880656e-05, "loss": 2.4893, "step": 58915 }, { "epoch": 4.62, "learning_rate": 5.3802242609582066e-05, "loss": 2.7121, "step": 58916 }, { "epoch": 4.62, "learning_rate": 5.380145848035757e-05, "loss": 2.5894, "step": 58917 }, { "epoch": 4.62, "learning_rate": 5.3800674351133074e-05, "loss": 2.7104, "step": 58918 }, { "epoch": 4.62, "learning_rate": 5.379989022190858e-05, "loss": 2.9164, "step": 58919 }, { "epoch": 4.62, "learning_rate": 5.379910609268408e-05, "loss": 2.6022, "step": 58920 }, { "epoch": 4.62, "learning_rate": 5.3798321963459585e-05, "loss": 2.7099, "step": 58921 }, { "epoch": 4.62, "learning_rate": 5.379753783423509e-05, "loss": 2.5084, "step": 58922 }, { "epoch": 4.62, "learning_rate": 5.3796753705010585e-05, "loss": 2.6841, "step": 58923 }, { "epoch": 4.62, "learning_rate": 5.379596957578609e-05, "loss": 2.7401, "step": 58924 }, { "epoch": 4.62, "learning_rate": 5.379518544656159e-05, "loss": 2.4898, "step": 58925 }, { "epoch": 4.62, "learning_rate": 5.3794401317337096e-05, "loss": 2.6186, "step": 58926 }, { "epoch": 4.62, "learning_rate": 5.37936171881126e-05, "loss": 2.6719, "step": 58927 }, { "epoch": 4.62, "learning_rate": 5.3792833058888104e-05, "loss": 2.6477, "step": 58928 }, { "epoch": 4.62, "learning_rate": 5.379204892966361e-05, "loss": 2.6928, "step": 58929 }, { "epoch": 4.62, "learning_rate": 5.379126480043911e-05, "loss": 2.504, "step": 58930 }, { "epoch": 4.62, "learning_rate": 5.3790480671214615e-05, "loss": 2.5363, "step": 58931 }, { "epoch": 4.62, "learning_rate": 5.378969654199012e-05, "loss": 2.6763, "step": 58932 }, { "epoch": 4.62, "learning_rate": 5.378891241276562e-05, "loss": 2.9889, "step": 58933 }, { "epoch": 4.62, "learning_rate": 5.3788128283541126e-05, "loss": 2.6965, "step": 58934 }, { "epoch": 4.62, "learning_rate": 5.378734415431663e-05, "loss": 2.679, "step": 58935 }, { "epoch": 4.62, "learning_rate": 5.378656002509214e-05, "loss": 2.5224, "step": 58936 }, { "epoch": 4.62, "learning_rate": 5.378577589586764e-05, "loss": 2.6465, "step": 58937 }, { "epoch": 4.62, "learning_rate": 5.378499176664315e-05, "loss": 2.7282, "step": 58938 }, { "epoch": 4.62, "learning_rate": 5.378420763741865e-05, "loss": 2.7325, "step": 58939 }, { "epoch": 4.62, "learning_rate": 5.3783423508194154e-05, "loss": 2.6823, "step": 58940 }, { "epoch": 4.62, "learning_rate": 5.378263937896966e-05, "loss": 2.7991, "step": 58941 }, { "epoch": 4.62, "learning_rate": 5.378185524974516e-05, "loss": 2.9826, "step": 58942 }, { "epoch": 4.62, "learning_rate": 5.3781071120520665e-05, "loss": 2.5006, "step": 58943 }, { "epoch": 4.62, "learning_rate": 5.378028699129617e-05, "loss": 2.787, "step": 58944 }, { "epoch": 4.62, "learning_rate": 5.377950286207167e-05, "loss": 2.5697, "step": 58945 }, { "epoch": 4.62, "learning_rate": 5.3778718732847176e-05, "loss": 2.7384, "step": 58946 }, { "epoch": 4.62, "learning_rate": 5.377793460362268e-05, "loss": 2.4615, "step": 58947 }, { "epoch": 4.62, "learning_rate": 5.3777150474398184e-05, "loss": 2.7167, "step": 58948 }, { "epoch": 4.62, "learning_rate": 5.377636634517369e-05, "loss": 2.6228, "step": 58949 }, { "epoch": 4.62, "learning_rate": 5.377558221594919e-05, "loss": 2.4592, "step": 58950 }, { "epoch": 4.62, "learning_rate": 5.3774798086724695e-05, "loss": 2.8723, "step": 58951 }, { "epoch": 4.62, "learning_rate": 5.3774013957500205e-05, "loss": 2.6079, "step": 58952 }, { "epoch": 4.62, "learning_rate": 5.377322982827571e-05, "loss": 2.6357, "step": 58953 }, { "epoch": 4.62, "learning_rate": 5.377244569905121e-05, "loss": 2.8245, "step": 58954 }, { "epoch": 4.62, "learning_rate": 5.3771661569826716e-05, "loss": 2.8347, "step": 58955 }, { "epoch": 4.62, "learning_rate": 5.377087744060222e-05, "loss": 2.6642, "step": 58956 }, { "epoch": 4.62, "learning_rate": 5.3770093311377724e-05, "loss": 2.6482, "step": 58957 }, { "epoch": 4.62, "learning_rate": 5.376930918215323e-05, "loss": 2.6779, "step": 58958 }, { "epoch": 4.62, "learning_rate": 5.376852505292872e-05, "loss": 2.7883, "step": 58959 }, { "epoch": 4.62, "learning_rate": 5.376774092370422e-05, "loss": 2.4739, "step": 58960 }, { "epoch": 4.62, "learning_rate": 5.3766956794479725e-05, "loss": 2.5192, "step": 58961 }, { "epoch": 4.62, "learning_rate": 5.3766172665255235e-05, "loss": 2.6885, "step": 58962 }, { "epoch": 4.62, "learning_rate": 5.376538853603074e-05, "loss": 2.5765, "step": 58963 }, { "epoch": 4.62, "learning_rate": 5.376460440680624e-05, "loss": 2.5754, "step": 58964 }, { "epoch": 4.62, "learning_rate": 5.3763820277581746e-05, "loss": 2.5239, "step": 58965 }, { "epoch": 4.62, "learning_rate": 5.376303614835725e-05, "loss": 2.6071, "step": 58966 }, { "epoch": 4.62, "learning_rate": 5.3762252019132754e-05, "loss": 2.8863, "step": 58967 }, { "epoch": 4.62, "learning_rate": 5.376146788990826e-05, "loss": 2.791, "step": 58968 }, { "epoch": 4.62, "learning_rate": 5.376068376068376e-05, "loss": 2.5903, "step": 58969 }, { "epoch": 4.62, "learning_rate": 5.3759899631459265e-05, "loss": 2.7277, "step": 58970 }, { "epoch": 4.62, "learning_rate": 5.375911550223477e-05, "loss": 2.6759, "step": 58971 }, { "epoch": 4.62, "learning_rate": 5.375833137301027e-05, "loss": 2.6087, "step": 58972 }, { "epoch": 4.62, "learning_rate": 5.3757547243785776e-05, "loss": 2.6613, "step": 58973 }, { "epoch": 4.62, "learning_rate": 5.375676311456128e-05, "loss": 2.7787, "step": 58974 }, { "epoch": 4.62, "learning_rate": 5.375597898533678e-05, "loss": 2.592, "step": 58975 }, { "epoch": 4.62, "learning_rate": 5.375519485611229e-05, "loss": 2.5813, "step": 58976 }, { "epoch": 4.62, "learning_rate": 5.37544107268878e-05, "loss": 2.7728, "step": 58977 }, { "epoch": 4.62, "learning_rate": 5.37536265976633e-05, "loss": 2.7084, "step": 58978 }, { "epoch": 4.62, "learning_rate": 5.3752842468438805e-05, "loss": 2.5955, "step": 58979 }, { "epoch": 4.62, "learning_rate": 5.375205833921431e-05, "loss": 2.7252, "step": 58980 }, { "epoch": 4.62, "learning_rate": 5.375127420998981e-05, "loss": 2.4911, "step": 58981 }, { "epoch": 4.62, "learning_rate": 5.3750490080765316e-05, "loss": 2.6671, "step": 58982 }, { "epoch": 4.63, "learning_rate": 5.374970595154082e-05, "loss": 2.677, "step": 58983 }, { "epoch": 4.63, "learning_rate": 5.374892182231632e-05, "loss": 2.6504, "step": 58984 }, { "epoch": 4.63, "learning_rate": 5.374813769309183e-05, "loss": 2.6195, "step": 58985 }, { "epoch": 4.63, "learning_rate": 5.374735356386733e-05, "loss": 2.999, "step": 58986 }, { "epoch": 4.63, "learning_rate": 5.3746569434642834e-05, "loss": 2.7742, "step": 58987 }, { "epoch": 4.63, "learning_rate": 5.374578530541834e-05, "loss": 2.7421, "step": 58988 }, { "epoch": 4.63, "learning_rate": 5.374500117619384e-05, "loss": 2.6285, "step": 58989 }, { "epoch": 4.63, "learning_rate": 5.3744217046969345e-05, "loss": 2.723, "step": 58990 }, { "epoch": 4.63, "learning_rate": 5.374343291774485e-05, "loss": 2.5861, "step": 58991 }, { "epoch": 4.63, "learning_rate": 5.374264878852035e-05, "loss": 2.5767, "step": 58992 }, { "epoch": 4.63, "learning_rate": 5.374186465929586e-05, "loss": 2.5998, "step": 58993 }, { "epoch": 4.63, "learning_rate": 5.3741080530071367e-05, "loss": 2.8387, "step": 58994 }, { "epoch": 4.63, "learning_rate": 5.374029640084686e-05, "loss": 2.7729, "step": 58995 }, { "epoch": 4.63, "learning_rate": 5.373951227162236e-05, "loss": 2.7979, "step": 58996 }, { "epoch": 4.63, "learning_rate": 5.3738728142397864e-05, "loss": 2.6174, "step": 58997 }, { "epoch": 4.63, "learning_rate": 5.373794401317337e-05, "loss": 2.797, "step": 58998 }, { "epoch": 4.63, "learning_rate": 5.373715988394887e-05, "loss": 2.7225, "step": 58999 }, { "epoch": 4.63, "learning_rate": 5.3736375754724375e-05, "loss": 2.6955, "step": 59000 }, { "epoch": 4.63, "eval_loss": 2.896040439605713, "eval_runtime": 217.5802, "eval_samples_per_second": 52.082, "eval_steps_per_second": 3.259, "step": 59000 }, { "epoch": 4.63, "learning_rate": 5.373559162549988e-05, "loss": 2.7196, "step": 59001 }, { "epoch": 4.63, "learning_rate": 5.373480749627538e-05, "loss": 2.6838, "step": 59002 }, { "epoch": 4.63, "learning_rate": 5.373402336705089e-05, "loss": 2.7911, "step": 59003 }, { "epoch": 4.63, "learning_rate": 5.3733239237826397e-05, "loss": 2.7435, "step": 59004 }, { "epoch": 4.63, "learning_rate": 5.37324551086019e-05, "loss": 2.7669, "step": 59005 }, { "epoch": 4.63, "learning_rate": 5.3731670979377404e-05, "loss": 2.657, "step": 59006 }, { "epoch": 4.63, "learning_rate": 5.373088685015291e-05, "loss": 2.5696, "step": 59007 }, { "epoch": 4.63, "learning_rate": 5.373010272092841e-05, "loss": 2.7079, "step": 59008 }, { "epoch": 4.63, "learning_rate": 5.3729318591703915e-05, "loss": 2.551, "step": 59009 }, { "epoch": 4.63, "learning_rate": 5.372853446247942e-05, "loss": 2.5998, "step": 59010 }, { "epoch": 4.63, "learning_rate": 5.372775033325492e-05, "loss": 2.6031, "step": 59011 }, { "epoch": 4.63, "learning_rate": 5.3726966204030426e-05, "loss": 2.7415, "step": 59012 }, { "epoch": 4.63, "learning_rate": 5.372618207480593e-05, "loss": 2.723, "step": 59013 }, { "epoch": 4.63, "learning_rate": 5.372539794558143e-05, "loss": 2.5129, "step": 59014 }, { "epoch": 4.63, "learning_rate": 5.372461381635694e-05, "loss": 2.5696, "step": 59015 }, { "epoch": 4.63, "learning_rate": 5.372382968713244e-05, "loss": 2.6122, "step": 59016 }, { "epoch": 4.63, "learning_rate": 5.3723045557907944e-05, "loss": 2.6804, "step": 59017 }, { "epoch": 4.63, "learning_rate": 5.372226142868345e-05, "loss": 2.6292, "step": 59018 }, { "epoch": 4.63, "learning_rate": 5.372147729945896e-05, "loss": 2.6669, "step": 59019 }, { "epoch": 4.63, "learning_rate": 5.372069317023446e-05, "loss": 2.5873, "step": 59020 }, { "epoch": 4.63, "learning_rate": 5.3719909041009966e-05, "loss": 2.6301, "step": 59021 }, { "epoch": 4.63, "learning_rate": 5.371912491178547e-05, "loss": 2.5267, "step": 59022 }, { "epoch": 4.63, "learning_rate": 5.371834078256097e-05, "loss": 2.5334, "step": 59023 }, { "epoch": 4.63, "learning_rate": 5.371755665333648e-05, "loss": 2.7256, "step": 59024 }, { "epoch": 4.63, "learning_rate": 5.371677252411198e-05, "loss": 2.685, "step": 59025 }, { "epoch": 4.63, "learning_rate": 5.3715988394887484e-05, "loss": 2.6098, "step": 59026 }, { "epoch": 4.63, "learning_rate": 5.371520426566299e-05, "loss": 2.7381, "step": 59027 }, { "epoch": 4.63, "learning_rate": 5.371442013643849e-05, "loss": 2.8651, "step": 59028 }, { "epoch": 4.63, "learning_rate": 5.3713636007213995e-05, "loss": 2.4671, "step": 59029 }, { "epoch": 4.63, "learning_rate": 5.37128518779895e-05, "loss": 2.6655, "step": 59030 }, { "epoch": 4.63, "learning_rate": 5.3712067748764996e-05, "loss": 2.5442, "step": 59031 }, { "epoch": 4.63, "learning_rate": 5.37112836195405e-05, "loss": 2.4901, "step": 59032 }, { "epoch": 4.63, "learning_rate": 5.3710499490316e-05, "loss": 2.7671, "step": 59033 }, { "epoch": 4.63, "learning_rate": 5.370971536109151e-05, "loss": 2.7681, "step": 59034 }, { "epoch": 4.63, "learning_rate": 5.370893123186701e-05, "loss": 3.0045, "step": 59035 }, { "epoch": 4.63, "learning_rate": 5.3708147102642514e-05, "loss": 2.6201, "step": 59036 }, { "epoch": 4.63, "learning_rate": 5.370736297341802e-05, "loss": 2.5069, "step": 59037 }, { "epoch": 4.63, "learning_rate": 5.370657884419352e-05, "loss": 2.8521, "step": 59038 }, { "epoch": 4.63, "learning_rate": 5.3705794714969025e-05, "loss": 2.7578, "step": 59039 }, { "epoch": 4.63, "learning_rate": 5.370501058574453e-05, "loss": 2.577, "step": 59040 }, { "epoch": 4.63, "learning_rate": 5.370422645652003e-05, "loss": 2.6475, "step": 59041 }, { "epoch": 4.63, "learning_rate": 5.3703442327295536e-05, "loss": 2.669, "step": 59042 }, { "epoch": 4.63, "learning_rate": 5.370265819807104e-05, "loss": 2.6509, "step": 59043 }, { "epoch": 4.63, "learning_rate": 5.370187406884655e-05, "loss": 2.8379, "step": 59044 }, { "epoch": 4.63, "learning_rate": 5.3701089939622054e-05, "loss": 2.7503, "step": 59045 }, { "epoch": 4.63, "learning_rate": 5.370030581039756e-05, "loss": 2.572, "step": 59046 }, { "epoch": 4.63, "learning_rate": 5.369952168117306e-05, "loss": 2.6444, "step": 59047 }, { "epoch": 4.63, "learning_rate": 5.3698737551948565e-05, "loss": 2.8412, "step": 59048 }, { "epoch": 4.63, "learning_rate": 5.369795342272407e-05, "loss": 2.552, "step": 59049 }, { "epoch": 4.63, "learning_rate": 5.369716929349957e-05, "loss": 2.6264, "step": 59050 }, { "epoch": 4.63, "learning_rate": 5.3696385164275076e-05, "loss": 2.9731, "step": 59051 }, { "epoch": 4.63, "learning_rate": 5.369560103505058e-05, "loss": 2.6133, "step": 59052 }, { "epoch": 4.63, "learning_rate": 5.3694816905826084e-05, "loss": 2.8244, "step": 59053 }, { "epoch": 4.63, "learning_rate": 5.369403277660159e-05, "loss": 2.5706, "step": 59054 }, { "epoch": 4.63, "learning_rate": 5.369324864737709e-05, "loss": 2.7665, "step": 59055 }, { "epoch": 4.63, "learning_rate": 5.3692464518152595e-05, "loss": 2.629, "step": 59056 }, { "epoch": 4.63, "learning_rate": 5.36916803889281e-05, "loss": 2.5791, "step": 59057 }, { "epoch": 4.63, "learning_rate": 5.36908962597036e-05, "loss": 2.4337, "step": 59058 }, { "epoch": 4.63, "learning_rate": 5.3690112130479106e-05, "loss": 2.504, "step": 59059 }, { "epoch": 4.63, "learning_rate": 5.3689328001254616e-05, "loss": 2.5851, "step": 59060 }, { "epoch": 4.63, "learning_rate": 5.368854387203012e-05, "loss": 2.5, "step": 59061 }, { "epoch": 4.63, "learning_rate": 5.3687759742805623e-05, "loss": 2.8775, "step": 59062 }, { "epoch": 4.63, "learning_rate": 5.368697561358113e-05, "loss": 2.7206, "step": 59063 }, { "epoch": 4.63, "learning_rate": 5.368619148435663e-05, "loss": 2.5651, "step": 59064 }, { "epoch": 4.63, "learning_rate": 5.3685407355132134e-05, "loss": 2.631, "step": 59065 }, { "epoch": 4.63, "learning_rate": 5.368462322590764e-05, "loss": 2.4935, "step": 59066 }, { "epoch": 4.63, "learning_rate": 5.368383909668313e-05, "loss": 2.6425, "step": 59067 }, { "epoch": 4.63, "learning_rate": 5.368305496745863e-05, "loss": 2.7135, "step": 59068 }, { "epoch": 4.63, "learning_rate": 5.3682270838234136e-05, "loss": 2.5134, "step": 59069 }, { "epoch": 4.63, "learning_rate": 5.3681486709009646e-05, "loss": 2.6047, "step": 59070 }, { "epoch": 4.63, "learning_rate": 5.368070257978515e-05, "loss": 2.8275, "step": 59071 }, { "epoch": 4.63, "learning_rate": 5.3679918450560653e-05, "loss": 2.6685, "step": 59072 }, { "epoch": 4.63, "learning_rate": 5.367913432133616e-05, "loss": 2.6264, "step": 59073 }, { "epoch": 4.63, "learning_rate": 5.367835019211166e-05, "loss": 2.6229, "step": 59074 }, { "epoch": 4.63, "learning_rate": 5.3677566062887164e-05, "loss": 2.6199, "step": 59075 }, { "epoch": 4.63, "learning_rate": 5.367678193366267e-05, "loss": 2.5372, "step": 59076 }, { "epoch": 4.63, "learning_rate": 5.367599780443817e-05, "loss": 2.5114, "step": 59077 }, { "epoch": 4.63, "learning_rate": 5.3675213675213675e-05, "loss": 2.8469, "step": 59078 }, { "epoch": 4.63, "learning_rate": 5.367442954598918e-05, "loss": 2.6143, "step": 59079 }, { "epoch": 4.63, "learning_rate": 5.367364541676468e-05, "loss": 3.08, "step": 59080 }, { "epoch": 4.63, "learning_rate": 5.3672861287540187e-05, "loss": 2.4704, "step": 59081 }, { "epoch": 4.63, "learning_rate": 5.367207715831569e-05, "loss": 2.5141, "step": 59082 }, { "epoch": 4.63, "learning_rate": 5.3671293029091194e-05, "loss": 2.5206, "step": 59083 }, { "epoch": 4.63, "learning_rate": 5.36705088998667e-05, "loss": 2.6275, "step": 59084 }, { "epoch": 4.63, "learning_rate": 5.36697247706422e-05, "loss": 2.7812, "step": 59085 }, { "epoch": 4.63, "learning_rate": 5.366894064141771e-05, "loss": 2.8751, "step": 59086 }, { "epoch": 4.63, "learning_rate": 5.3668156512193215e-05, "loss": 2.6106, "step": 59087 }, { "epoch": 4.63, "learning_rate": 5.366737238296872e-05, "loss": 2.6967, "step": 59088 }, { "epoch": 4.63, "learning_rate": 5.366658825374422e-05, "loss": 2.5585, "step": 59089 }, { "epoch": 4.63, "learning_rate": 5.3665804124519726e-05, "loss": 2.4126, "step": 59090 }, { "epoch": 4.63, "learning_rate": 5.366501999529523e-05, "loss": 2.5892, "step": 59091 }, { "epoch": 4.63, "learning_rate": 5.3664235866070734e-05, "loss": 2.4964, "step": 59092 }, { "epoch": 4.63, "learning_rate": 5.366345173684624e-05, "loss": 2.6143, "step": 59093 }, { "epoch": 4.63, "learning_rate": 5.366266760762174e-05, "loss": 2.8196, "step": 59094 }, { "epoch": 4.63, "learning_rate": 5.3661883478397245e-05, "loss": 2.5778, "step": 59095 }, { "epoch": 4.63, "learning_rate": 5.366109934917275e-05, "loss": 2.8793, "step": 59096 }, { "epoch": 4.63, "learning_rate": 5.366031521994825e-05, "loss": 2.8346, "step": 59097 }, { "epoch": 4.63, "learning_rate": 5.3659531090723756e-05, "loss": 2.6139, "step": 59098 }, { "epoch": 4.63, "learning_rate": 5.365874696149926e-05, "loss": 2.7274, "step": 59099 }, { "epoch": 4.63, "learning_rate": 5.365796283227476e-05, "loss": 2.5839, "step": 59100 }, { "epoch": 4.63, "learning_rate": 5.3657178703050274e-05, "loss": 2.5618, "step": 59101 }, { "epoch": 4.63, "learning_rate": 5.365639457382578e-05, "loss": 2.7984, "step": 59102 }, { "epoch": 4.63, "learning_rate": 5.365561044460127e-05, "loss": 2.8344, "step": 59103 }, { "epoch": 4.63, "learning_rate": 5.365482631537677e-05, "loss": 2.5461, "step": 59104 }, { "epoch": 4.63, "learning_rate": 5.3654042186152275e-05, "loss": 2.7136, "step": 59105 }, { "epoch": 4.63, "learning_rate": 5.365325805692778e-05, "loss": 2.679, "step": 59106 }, { "epoch": 4.63, "learning_rate": 5.365247392770328e-05, "loss": 2.7143, "step": 59107 }, { "epoch": 4.63, "learning_rate": 5.3651689798478786e-05, "loss": 2.702, "step": 59108 }, { "epoch": 4.63, "learning_rate": 5.365090566925429e-05, "loss": 2.8153, "step": 59109 }, { "epoch": 4.63, "learning_rate": 5.365012154002979e-05, "loss": 2.6208, "step": 59110 }, { "epoch": 4.64, "learning_rate": 5.3649337410805304e-05, "loss": 2.6864, "step": 59111 }, { "epoch": 4.64, "learning_rate": 5.364855328158081e-05, "loss": 2.5011, "step": 59112 }, { "epoch": 4.64, "learning_rate": 5.364776915235631e-05, "loss": 2.7551, "step": 59113 }, { "epoch": 4.64, "learning_rate": 5.3646985023131815e-05, "loss": 2.7445, "step": 59114 }, { "epoch": 4.64, "learning_rate": 5.364620089390732e-05, "loss": 2.7216, "step": 59115 }, { "epoch": 4.64, "learning_rate": 5.364541676468282e-05, "loss": 2.609, "step": 59116 }, { "epoch": 4.64, "learning_rate": 5.3644632635458326e-05, "loss": 2.6624, "step": 59117 }, { "epoch": 4.64, "learning_rate": 5.364384850623383e-05, "loss": 2.7377, "step": 59118 }, { "epoch": 4.64, "learning_rate": 5.364306437700933e-05, "loss": 2.5459, "step": 59119 }, { "epoch": 4.64, "learning_rate": 5.364228024778484e-05, "loss": 2.6792, "step": 59120 }, { "epoch": 4.64, "learning_rate": 5.364149611856034e-05, "loss": 2.5926, "step": 59121 }, { "epoch": 4.64, "learning_rate": 5.3640711989335844e-05, "loss": 2.6349, "step": 59122 }, { "epoch": 4.64, "learning_rate": 5.363992786011135e-05, "loss": 2.494, "step": 59123 }, { "epoch": 4.64, "learning_rate": 5.363914373088685e-05, "loss": 2.5908, "step": 59124 }, { "epoch": 4.64, "learning_rate": 5.3638359601662355e-05, "loss": 2.6184, "step": 59125 }, { "epoch": 4.64, "learning_rate": 5.363757547243786e-05, "loss": 2.729, "step": 59126 }, { "epoch": 4.64, "learning_rate": 5.363679134321337e-05, "loss": 2.7143, "step": 59127 }, { "epoch": 4.64, "learning_rate": 5.363600721398887e-05, "loss": 2.785, "step": 59128 }, { "epoch": 4.64, "learning_rate": 5.3635223084764377e-05, "loss": 2.7279, "step": 59129 }, { "epoch": 4.64, "learning_rate": 5.363443895553988e-05, "loss": 2.5088, "step": 59130 }, { "epoch": 4.64, "learning_rate": 5.3633654826315384e-05, "loss": 2.7034, "step": 59131 }, { "epoch": 4.64, "learning_rate": 5.363287069709089e-05, "loss": 2.6522, "step": 59132 }, { "epoch": 4.64, "learning_rate": 5.363208656786639e-05, "loss": 2.7646, "step": 59133 }, { "epoch": 4.64, "learning_rate": 5.3631302438641895e-05, "loss": 2.6062, "step": 59134 }, { "epoch": 4.64, "learning_rate": 5.36305183094174e-05, "loss": 2.5617, "step": 59135 }, { "epoch": 4.64, "learning_rate": 5.36297341801929e-05, "loss": 2.5941, "step": 59136 }, { "epoch": 4.64, "learning_rate": 5.3628950050968406e-05, "loss": 2.5689, "step": 59137 }, { "epoch": 4.64, "learning_rate": 5.362816592174391e-05, "loss": 2.6445, "step": 59138 }, { "epoch": 4.64, "learning_rate": 5.3627381792519407e-05, "loss": 2.5474, "step": 59139 }, { "epoch": 4.64, "learning_rate": 5.362659766329491e-05, "loss": 2.4654, "step": 59140 }, { "epoch": 4.64, "learning_rate": 5.3625813534070414e-05, "loss": 2.851, "step": 59141 }, { "epoch": 4.64, "learning_rate": 5.362502940484592e-05, "loss": 2.6658, "step": 59142 }, { "epoch": 4.64, "learning_rate": 5.362424527562142e-05, "loss": 2.6857, "step": 59143 }, { "epoch": 4.64, "learning_rate": 5.3623461146396925e-05, "loss": 2.6373, "step": 59144 }, { "epoch": 4.64, "learning_rate": 5.362267701717243e-05, "loss": 2.8638, "step": 59145 }, { "epoch": 4.64, "learning_rate": 5.362189288794793e-05, "loss": 2.5954, "step": 59146 }, { "epoch": 4.64, "learning_rate": 5.3621108758723436e-05, "loss": 2.7753, "step": 59147 }, { "epoch": 4.64, "learning_rate": 5.362032462949894e-05, "loss": 2.6144, "step": 59148 }, { "epoch": 4.64, "learning_rate": 5.361954050027444e-05, "loss": 2.7022, "step": 59149 }, { "epoch": 4.64, "learning_rate": 5.361875637104995e-05, "loss": 2.8446, "step": 59150 }, { "epoch": 4.64, "learning_rate": 5.361797224182545e-05, "loss": 2.4246, "step": 59151 }, { "epoch": 4.64, "learning_rate": 5.361718811260096e-05, "loss": 2.8343, "step": 59152 }, { "epoch": 4.64, "learning_rate": 5.3616403983376465e-05, "loss": 2.5075, "step": 59153 }, { "epoch": 4.64, "learning_rate": 5.361561985415197e-05, "loss": 2.8111, "step": 59154 }, { "epoch": 4.64, "learning_rate": 5.361483572492747e-05, "loss": 2.628, "step": 59155 }, { "epoch": 4.64, "learning_rate": 5.3614051595702976e-05, "loss": 2.8425, "step": 59156 }, { "epoch": 4.64, "learning_rate": 5.361326746647848e-05, "loss": 2.5359, "step": 59157 }, { "epoch": 4.64, "learning_rate": 5.361248333725398e-05, "loss": 2.6254, "step": 59158 }, { "epoch": 4.64, "learning_rate": 5.361169920802949e-05, "loss": 2.615, "step": 59159 }, { "epoch": 4.64, "learning_rate": 5.361091507880499e-05, "loss": 2.5236, "step": 59160 }, { "epoch": 4.64, "learning_rate": 5.3610130949580494e-05, "loss": 2.6262, "step": 59161 }, { "epoch": 4.64, "learning_rate": 5.3609346820356e-05, "loss": 2.4941, "step": 59162 }, { "epoch": 4.64, "learning_rate": 5.36085626911315e-05, "loss": 2.5778, "step": 59163 }, { "epoch": 4.64, "learning_rate": 5.3607778561907005e-05, "loss": 2.7886, "step": 59164 }, { "epoch": 4.64, "learning_rate": 5.360699443268251e-05, "loss": 2.6324, "step": 59165 }, { "epoch": 4.64, "learning_rate": 5.360621030345801e-05, "loss": 2.5915, "step": 59166 }, { "epoch": 4.64, "learning_rate": 5.3605426174233516e-05, "loss": 2.6989, "step": 59167 }, { "epoch": 4.64, "learning_rate": 5.360464204500903e-05, "loss": 2.586, "step": 59168 }, { "epoch": 4.64, "learning_rate": 5.360385791578453e-05, "loss": 2.5081, "step": 59169 }, { "epoch": 4.64, "learning_rate": 5.3603073786560034e-05, "loss": 2.7087, "step": 59170 }, { "epoch": 4.64, "learning_rate": 5.360228965733554e-05, "loss": 2.6116, "step": 59171 }, { "epoch": 4.64, "learning_rate": 5.360150552811104e-05, "loss": 2.6009, "step": 59172 }, { "epoch": 4.64, "learning_rate": 5.3600721398886545e-05, "loss": 2.3862, "step": 59173 }, { "epoch": 4.64, "learning_rate": 5.359993726966205e-05, "loss": 2.816, "step": 59174 }, { "epoch": 4.64, "learning_rate": 5.359915314043754e-05, "loss": 2.577, "step": 59175 }, { "epoch": 4.64, "learning_rate": 5.359836901121304e-05, "loss": 2.6098, "step": 59176 }, { "epoch": 4.64, "learning_rate": 5.3597584881988546e-05, "loss": 2.7963, "step": 59177 }, { "epoch": 4.64, "learning_rate": 5.359680075276406e-05, "loss": 2.8004, "step": 59178 }, { "epoch": 4.64, "learning_rate": 5.359601662353956e-05, "loss": 2.6119, "step": 59179 }, { "epoch": 4.64, "learning_rate": 5.3595232494315064e-05, "loss": 2.5923, "step": 59180 }, { "epoch": 4.64, "learning_rate": 5.359444836509057e-05, "loss": 2.497, "step": 59181 }, { "epoch": 4.64, "learning_rate": 5.359366423586607e-05, "loss": 2.6814, "step": 59182 }, { "epoch": 4.64, "learning_rate": 5.3592880106641575e-05, "loss": 2.8299, "step": 59183 }, { "epoch": 4.64, "learning_rate": 5.359209597741708e-05, "loss": 2.7057, "step": 59184 }, { "epoch": 4.64, "learning_rate": 5.359131184819258e-05, "loss": 2.6479, "step": 59185 }, { "epoch": 4.64, "learning_rate": 5.3590527718968086e-05, "loss": 2.568, "step": 59186 }, { "epoch": 4.64, "learning_rate": 5.358974358974359e-05, "loss": 2.6737, "step": 59187 }, { "epoch": 4.64, "learning_rate": 5.3588959460519094e-05, "loss": 2.6835, "step": 59188 }, { "epoch": 4.64, "learning_rate": 5.35881753312946e-05, "loss": 2.6963, "step": 59189 }, { "epoch": 4.64, "learning_rate": 5.35873912020701e-05, "loss": 2.8225, "step": 59190 }, { "epoch": 4.64, "learning_rate": 5.3586607072845605e-05, "loss": 2.8166, "step": 59191 }, { "epoch": 4.64, "learning_rate": 5.358582294362111e-05, "loss": 2.6355, "step": 59192 }, { "epoch": 4.64, "learning_rate": 5.358503881439661e-05, "loss": 2.6644, "step": 59193 }, { "epoch": 4.64, "learning_rate": 5.358425468517212e-05, "loss": 2.6814, "step": 59194 }, { "epoch": 4.64, "learning_rate": 5.3583470555947626e-05, "loss": 2.5579, "step": 59195 }, { "epoch": 4.64, "learning_rate": 5.358268642672313e-05, "loss": 2.625, "step": 59196 }, { "epoch": 4.64, "learning_rate": 5.3581902297498633e-05, "loss": 2.5344, "step": 59197 }, { "epoch": 4.64, "learning_rate": 5.358111816827414e-05, "loss": 2.6595, "step": 59198 }, { "epoch": 4.64, "learning_rate": 5.358033403904964e-05, "loss": 2.7603, "step": 59199 }, { "epoch": 4.64, "learning_rate": 5.3579549909825145e-05, "loss": 2.5756, "step": 59200 }, { "epoch": 4.64, "learning_rate": 5.357876578060065e-05, "loss": 2.5471, "step": 59201 }, { "epoch": 4.64, "learning_rate": 5.357798165137615e-05, "loss": 2.5728, "step": 59202 }, { "epoch": 4.64, "learning_rate": 5.3577197522151656e-05, "loss": 2.5092, "step": 59203 }, { "epoch": 4.64, "learning_rate": 5.357641339292716e-05, "loss": 2.7154, "step": 59204 }, { "epoch": 4.64, "learning_rate": 5.357562926370266e-05, "loss": 2.7559, "step": 59205 }, { "epoch": 4.64, "learning_rate": 5.3574845134478167e-05, "loss": 2.8171, "step": 59206 }, { "epoch": 4.64, "learning_rate": 5.357406100525367e-05, "loss": 2.9092, "step": 59207 }, { "epoch": 4.64, "learning_rate": 5.3573276876029174e-05, "loss": 2.7156, "step": 59208 }, { "epoch": 4.64, "learning_rate": 5.3572492746804684e-05, "loss": 2.7442, "step": 59209 }, { "epoch": 4.64, "learning_rate": 5.357170861758019e-05, "loss": 2.554, "step": 59210 }, { "epoch": 4.64, "learning_rate": 5.357092448835568e-05, "loss": 2.5445, "step": 59211 }, { "epoch": 4.64, "learning_rate": 5.357014035913118e-05, "loss": 2.5591, "step": 59212 }, { "epoch": 4.64, "learning_rate": 5.3569356229906686e-05, "loss": 2.4897, "step": 59213 }, { "epoch": 4.64, "learning_rate": 5.356857210068219e-05, "loss": 2.6511, "step": 59214 }, { "epoch": 4.64, "learning_rate": 5.356778797145769e-05, "loss": 2.6657, "step": 59215 }, { "epoch": 4.64, "learning_rate": 5.3567003842233197e-05, "loss": 2.6374, "step": 59216 }, { "epoch": 4.64, "learning_rate": 5.35662197130087e-05, "loss": 2.8357, "step": 59217 }, { "epoch": 4.64, "learning_rate": 5.3565435583784204e-05, "loss": 2.6993, "step": 59218 }, { "epoch": 4.64, "learning_rate": 5.3564651454559714e-05, "loss": 2.5733, "step": 59219 }, { "epoch": 4.64, "learning_rate": 5.356386732533522e-05, "loss": 2.6505, "step": 59220 }, { "epoch": 4.64, "learning_rate": 5.356308319611072e-05, "loss": 2.709, "step": 59221 }, { "epoch": 4.64, "learning_rate": 5.3562299066886225e-05, "loss": 2.6174, "step": 59222 }, { "epoch": 4.64, "learning_rate": 5.356151493766173e-05, "loss": 2.6499, "step": 59223 }, { "epoch": 4.64, "learning_rate": 5.356073080843723e-05, "loss": 2.7312, "step": 59224 }, { "epoch": 4.64, "learning_rate": 5.3559946679212736e-05, "loss": 2.5403, "step": 59225 }, { "epoch": 4.64, "learning_rate": 5.355916254998824e-05, "loss": 2.5035, "step": 59226 }, { "epoch": 4.64, "learning_rate": 5.3558378420763744e-05, "loss": 2.6068, "step": 59227 }, { "epoch": 4.64, "learning_rate": 5.355759429153925e-05, "loss": 2.6904, "step": 59228 }, { "epoch": 4.64, "learning_rate": 5.355681016231475e-05, "loss": 2.5819, "step": 59229 }, { "epoch": 4.64, "learning_rate": 5.3556026033090255e-05, "loss": 2.8935, "step": 59230 }, { "epoch": 4.64, "learning_rate": 5.355524190386576e-05, "loss": 2.842, "step": 59231 }, { "epoch": 4.64, "learning_rate": 5.355445777464126e-05, "loss": 2.5828, "step": 59232 }, { "epoch": 4.64, "learning_rate": 5.3553673645416766e-05, "loss": 2.539, "step": 59233 }, { "epoch": 4.64, "learning_rate": 5.355288951619227e-05, "loss": 2.5821, "step": 59234 }, { "epoch": 4.64, "learning_rate": 5.355210538696778e-05, "loss": 2.5571, "step": 59235 }, { "epoch": 4.64, "learning_rate": 5.3551321257743284e-05, "loss": 2.8865, "step": 59236 }, { "epoch": 4.64, "learning_rate": 5.355053712851879e-05, "loss": 2.5894, "step": 59237 }, { "epoch": 4.65, "learning_rate": 5.354975299929429e-05, "loss": 2.6652, "step": 59238 }, { "epoch": 4.65, "learning_rate": 5.3548968870069795e-05, "loss": 2.5668, "step": 59239 }, { "epoch": 4.65, "learning_rate": 5.35481847408453e-05, "loss": 2.7376, "step": 59240 }, { "epoch": 4.65, "learning_rate": 5.35474006116208e-05, "loss": 2.5045, "step": 59241 }, { "epoch": 4.65, "learning_rate": 5.3546616482396306e-05, "loss": 2.8266, "step": 59242 }, { "epoch": 4.65, "learning_rate": 5.354583235317181e-05, "loss": 2.6169, "step": 59243 }, { "epoch": 4.65, "learning_rate": 5.354504822394731e-05, "loss": 2.7241, "step": 59244 }, { "epoch": 4.65, "learning_rate": 5.354426409472282e-05, "loss": 2.6863, "step": 59245 }, { "epoch": 4.65, "learning_rate": 5.354347996549832e-05, "loss": 2.6384, "step": 59246 }, { "epoch": 4.65, "learning_rate": 5.354269583627382e-05, "loss": 2.6284, "step": 59247 }, { "epoch": 4.65, "learning_rate": 5.354191170704932e-05, "loss": 2.8284, "step": 59248 }, { "epoch": 4.65, "learning_rate": 5.3541127577824825e-05, "loss": 2.7118, "step": 59249 }, { "epoch": 4.65, "learning_rate": 5.354034344860033e-05, "loss": 2.5095, "step": 59250 }, { "epoch": 4.65, "learning_rate": 5.353955931937583e-05, "loss": 2.6908, "step": 59251 }, { "epoch": 4.65, "learning_rate": 5.3538775190151336e-05, "loss": 2.8288, "step": 59252 }, { "epoch": 4.65, "learning_rate": 5.353799106092684e-05, "loss": 2.5452, "step": 59253 }, { "epoch": 4.65, "learning_rate": 5.353720693170234e-05, "loss": 2.684, "step": 59254 }, { "epoch": 4.65, "learning_rate": 5.353642280247785e-05, "loss": 2.7448, "step": 59255 }, { "epoch": 4.65, "learning_rate": 5.353563867325335e-05, "loss": 2.5546, "step": 59256 }, { "epoch": 4.65, "learning_rate": 5.3534854544028854e-05, "loss": 2.6202, "step": 59257 }, { "epoch": 4.65, "learning_rate": 5.353407041480436e-05, "loss": 2.5139, "step": 59258 }, { "epoch": 4.65, "learning_rate": 5.353328628557986e-05, "loss": 2.7916, "step": 59259 }, { "epoch": 4.65, "learning_rate": 5.353250215635537e-05, "loss": 2.7838, "step": 59260 }, { "epoch": 4.65, "learning_rate": 5.3531718027130876e-05, "loss": 2.5883, "step": 59261 }, { "epoch": 4.65, "learning_rate": 5.353093389790638e-05, "loss": 2.7856, "step": 59262 }, { "epoch": 4.65, "learning_rate": 5.353014976868188e-05, "loss": 2.508, "step": 59263 }, { "epoch": 4.65, "learning_rate": 5.352936563945739e-05, "loss": 2.6283, "step": 59264 }, { "epoch": 4.65, "learning_rate": 5.352858151023289e-05, "loss": 2.4761, "step": 59265 }, { "epoch": 4.65, "learning_rate": 5.3527797381008394e-05, "loss": 2.6431, "step": 59266 }, { "epoch": 4.65, "learning_rate": 5.35270132517839e-05, "loss": 2.6588, "step": 59267 }, { "epoch": 4.65, "learning_rate": 5.35262291225594e-05, "loss": 2.7205, "step": 59268 }, { "epoch": 4.65, "learning_rate": 5.3525444993334905e-05, "loss": 2.6089, "step": 59269 }, { "epoch": 4.65, "learning_rate": 5.352466086411041e-05, "loss": 2.6174, "step": 59270 }, { "epoch": 4.65, "learning_rate": 5.352387673488591e-05, "loss": 2.7675, "step": 59271 }, { "epoch": 4.65, "learning_rate": 5.3523092605661416e-05, "loss": 2.5856, "step": 59272 }, { "epoch": 4.65, "learning_rate": 5.352230847643692e-05, "loss": 2.5909, "step": 59273 }, { "epoch": 4.65, "learning_rate": 5.3521524347212423e-05, "loss": 2.573, "step": 59274 }, { "epoch": 4.65, "learning_rate": 5.352074021798793e-05, "loss": 2.5061, "step": 59275 }, { "epoch": 4.65, "learning_rate": 5.351995608876344e-05, "loss": 2.6995, "step": 59276 }, { "epoch": 4.65, "learning_rate": 5.351917195953894e-05, "loss": 2.7734, "step": 59277 }, { "epoch": 4.65, "learning_rate": 5.3518387830314445e-05, "loss": 2.7023, "step": 59278 }, { "epoch": 4.65, "learning_rate": 5.351760370108995e-05, "loss": 2.6902, "step": 59279 }, { "epoch": 4.65, "learning_rate": 5.351681957186545e-05, "loss": 2.6226, "step": 59280 }, { "epoch": 4.65, "learning_rate": 5.3516035442640956e-05, "loss": 2.6145, "step": 59281 }, { "epoch": 4.65, "learning_rate": 5.351525131341646e-05, "loss": 2.5869, "step": 59282 }, { "epoch": 4.65, "learning_rate": 5.351446718419195e-05, "loss": 2.8481, "step": 59283 }, { "epoch": 4.65, "learning_rate": 5.3513683054967453e-05, "loss": 2.5477, "step": 59284 }, { "epoch": 4.65, "learning_rate": 5.351289892574296e-05, "loss": 2.5671, "step": 59285 }, { "epoch": 4.65, "learning_rate": 5.351211479651847e-05, "loss": 2.6323, "step": 59286 }, { "epoch": 4.65, "learning_rate": 5.351133066729397e-05, "loss": 2.7378, "step": 59287 }, { "epoch": 4.65, "learning_rate": 5.3510546538069475e-05, "loss": 2.5602, "step": 59288 }, { "epoch": 4.65, "learning_rate": 5.350976240884498e-05, "loss": 2.6422, "step": 59289 }, { "epoch": 4.65, "learning_rate": 5.350897827962048e-05, "loss": 2.7547, "step": 59290 }, { "epoch": 4.65, "learning_rate": 5.3508194150395986e-05, "loss": 2.6883, "step": 59291 }, { "epoch": 4.65, "learning_rate": 5.350741002117149e-05, "loss": 2.7318, "step": 59292 }, { "epoch": 4.65, "learning_rate": 5.350662589194699e-05, "loss": 2.8189, "step": 59293 }, { "epoch": 4.65, "learning_rate": 5.35058417627225e-05, "loss": 2.7598, "step": 59294 }, { "epoch": 4.65, "learning_rate": 5.3505057633498e-05, "loss": 2.838, "step": 59295 }, { "epoch": 4.65, "learning_rate": 5.3504273504273504e-05, "loss": 2.7493, "step": 59296 }, { "epoch": 4.65, "learning_rate": 5.350348937504901e-05, "loss": 2.6598, "step": 59297 }, { "epoch": 4.65, "learning_rate": 5.350270524582451e-05, "loss": 2.6147, "step": 59298 }, { "epoch": 4.65, "learning_rate": 5.3501921116600015e-05, "loss": 2.7802, "step": 59299 }, { "epoch": 4.65, "learning_rate": 5.350113698737552e-05, "loss": 2.5692, "step": 59300 }, { "epoch": 4.65, "learning_rate": 5.350035285815102e-05, "loss": 2.6561, "step": 59301 }, { "epoch": 4.65, "learning_rate": 5.349956872892653e-05, "loss": 2.7243, "step": 59302 }, { "epoch": 4.65, "learning_rate": 5.349878459970204e-05, "loss": 2.6585, "step": 59303 }, { "epoch": 4.65, "learning_rate": 5.349800047047754e-05, "loss": 2.8477, "step": 59304 }, { "epoch": 4.65, "learning_rate": 5.3497216341253044e-05, "loss": 2.8342, "step": 59305 }, { "epoch": 4.65, "learning_rate": 5.349643221202855e-05, "loss": 2.6243, "step": 59306 }, { "epoch": 4.65, "learning_rate": 5.349564808280405e-05, "loss": 2.6238, "step": 59307 }, { "epoch": 4.65, "learning_rate": 5.3494863953579555e-05, "loss": 2.8531, "step": 59308 }, { "epoch": 4.65, "learning_rate": 5.349407982435506e-05, "loss": 2.5249, "step": 59309 }, { "epoch": 4.65, "learning_rate": 5.349329569513056e-05, "loss": 2.809, "step": 59310 }, { "epoch": 4.65, "learning_rate": 5.3492511565906066e-05, "loss": 2.7292, "step": 59311 }, { "epoch": 4.65, "learning_rate": 5.349172743668157e-05, "loss": 2.618, "step": 59312 }, { "epoch": 4.65, "learning_rate": 5.3490943307457074e-05, "loss": 2.5107, "step": 59313 }, { "epoch": 4.65, "learning_rate": 5.349015917823258e-05, "loss": 2.4892, "step": 59314 }, { "epoch": 4.65, "learning_rate": 5.348937504900808e-05, "loss": 2.6178, "step": 59315 }, { "epoch": 4.65, "learning_rate": 5.3488590919783585e-05, "loss": 2.5748, "step": 59316 }, { "epoch": 4.65, "learning_rate": 5.3487806790559095e-05, "loss": 2.5123, "step": 59317 }, { "epoch": 4.65, "learning_rate": 5.34870226613346e-05, "loss": 2.4474, "step": 59318 }, { "epoch": 4.65, "learning_rate": 5.348623853211009e-05, "loss": 2.6489, "step": 59319 }, { "epoch": 4.65, "learning_rate": 5.348545440288559e-05, "loss": 2.8031, "step": 59320 }, { "epoch": 4.65, "learning_rate": 5.3484670273661096e-05, "loss": 3.0505, "step": 59321 }, { "epoch": 4.65, "learning_rate": 5.34838861444366e-05, "loss": 2.7016, "step": 59322 }, { "epoch": 4.65, "learning_rate": 5.3483102015212104e-05, "loss": 2.6025, "step": 59323 }, { "epoch": 4.65, "learning_rate": 5.348231788598761e-05, "loss": 2.6503, "step": 59324 }, { "epoch": 4.65, "learning_rate": 5.348153375676311e-05, "loss": 2.6367, "step": 59325 }, { "epoch": 4.65, "learning_rate": 5.3480749627538615e-05, "loss": 2.5812, "step": 59326 }, { "epoch": 4.65, "learning_rate": 5.3479965498314125e-05, "loss": 2.6062, "step": 59327 }, { "epoch": 4.65, "learning_rate": 5.347918136908963e-05, "loss": 2.6526, "step": 59328 }, { "epoch": 4.65, "learning_rate": 5.347839723986513e-05, "loss": 2.8154, "step": 59329 }, { "epoch": 4.65, "learning_rate": 5.3477613110640636e-05, "loss": 2.8951, "step": 59330 }, { "epoch": 4.65, "learning_rate": 5.347682898141614e-05, "loss": 2.6819, "step": 59331 }, { "epoch": 4.65, "learning_rate": 5.3476044852191644e-05, "loss": 2.9164, "step": 59332 }, { "epoch": 4.65, "learning_rate": 5.347526072296715e-05, "loss": 2.743, "step": 59333 }, { "epoch": 4.65, "learning_rate": 5.347447659374265e-05, "loss": 2.5752, "step": 59334 }, { "epoch": 4.65, "learning_rate": 5.3473692464518155e-05, "loss": 2.6354, "step": 59335 }, { "epoch": 4.65, "learning_rate": 5.347290833529366e-05, "loss": 2.7108, "step": 59336 }, { "epoch": 4.65, "learning_rate": 5.347212420606916e-05, "loss": 2.6509, "step": 59337 }, { "epoch": 4.65, "learning_rate": 5.3471340076844666e-05, "loss": 2.6539, "step": 59338 }, { "epoch": 4.65, "learning_rate": 5.347055594762017e-05, "loss": 2.6454, "step": 59339 }, { "epoch": 4.65, "learning_rate": 5.346977181839567e-05, "loss": 2.8501, "step": 59340 }, { "epoch": 4.65, "learning_rate": 5.3468987689171177e-05, "loss": 2.5126, "step": 59341 }, { "epoch": 4.65, "learning_rate": 5.346820355994668e-05, "loss": 2.6123, "step": 59342 }, { "epoch": 4.65, "learning_rate": 5.346741943072219e-05, "loss": 2.5999, "step": 59343 }, { "epoch": 4.65, "learning_rate": 5.3466635301497694e-05, "loss": 2.5371, "step": 59344 }, { "epoch": 4.65, "learning_rate": 5.34658511722732e-05, "loss": 2.723, "step": 59345 }, { "epoch": 4.65, "learning_rate": 5.34650670430487e-05, "loss": 2.5097, "step": 59346 }, { "epoch": 4.65, "learning_rate": 5.3464282913824205e-05, "loss": 2.747, "step": 59347 }, { "epoch": 4.65, "learning_rate": 5.346349878459971e-05, "loss": 2.6582, "step": 59348 }, { "epoch": 4.65, "learning_rate": 5.346271465537521e-05, "loss": 2.7795, "step": 59349 }, { "epoch": 4.65, "learning_rate": 5.3461930526150716e-05, "loss": 2.5374, "step": 59350 }, { "epoch": 4.65, "learning_rate": 5.346114639692622e-05, "loss": 2.5112, "step": 59351 }, { "epoch": 4.65, "learning_rate": 5.3460362267701724e-05, "loss": 2.651, "step": 59352 }, { "epoch": 4.65, "learning_rate": 5.345957813847723e-05, "loss": 2.9096, "step": 59353 }, { "epoch": 4.65, "learning_rate": 5.345879400925273e-05, "loss": 2.7369, "step": 59354 }, { "epoch": 4.65, "learning_rate": 5.345800988002823e-05, "loss": 2.6745, "step": 59355 }, { "epoch": 4.65, "learning_rate": 5.345722575080373e-05, "loss": 2.6797, "step": 59356 }, { "epoch": 4.65, "learning_rate": 5.3456441621579235e-05, "loss": 2.6905, "step": 59357 }, { "epoch": 4.65, "learning_rate": 5.345565749235474e-05, "loss": 2.6335, "step": 59358 }, { "epoch": 4.65, "learning_rate": 5.345487336313024e-05, "loss": 2.7506, "step": 59359 }, { "epoch": 4.65, "learning_rate": 5.3454089233905746e-05, "loss": 2.6603, "step": 59360 }, { "epoch": 4.65, "learning_rate": 5.345330510468125e-05, "loss": 2.6699, "step": 59361 }, { "epoch": 4.65, "learning_rate": 5.3452520975456754e-05, "loss": 2.6482, "step": 59362 }, { "epoch": 4.65, "learning_rate": 5.345173684623226e-05, "loss": 2.888, "step": 59363 }, { "epoch": 4.65, "learning_rate": 5.345095271700776e-05, "loss": 2.5509, "step": 59364 }, { "epoch": 4.65, "learning_rate": 5.3450168587783265e-05, "loss": 2.8578, "step": 59365 }, { "epoch": 4.66, "learning_rate": 5.344938445855877e-05, "loss": 2.7344, "step": 59366 }, { "epoch": 4.66, "learning_rate": 5.344860032933427e-05, "loss": 2.6817, "step": 59367 }, { "epoch": 4.66, "learning_rate": 5.344781620010978e-05, "loss": 2.7021, "step": 59368 }, { "epoch": 4.66, "learning_rate": 5.3447032070885286e-05, "loss": 2.6045, "step": 59369 }, { "epoch": 4.66, "learning_rate": 5.344624794166079e-05, "loss": 2.7426, "step": 59370 }, { "epoch": 4.66, "learning_rate": 5.3445463812436294e-05, "loss": 2.4908, "step": 59371 }, { "epoch": 4.66, "learning_rate": 5.34446796832118e-05, "loss": 2.7657, "step": 59372 }, { "epoch": 4.66, "learning_rate": 5.34438955539873e-05, "loss": 2.8062, "step": 59373 }, { "epoch": 4.66, "learning_rate": 5.3443111424762805e-05, "loss": 2.5281, "step": 59374 }, { "epoch": 4.66, "learning_rate": 5.344232729553831e-05, "loss": 2.5285, "step": 59375 }, { "epoch": 4.66, "learning_rate": 5.344154316631381e-05, "loss": 2.5519, "step": 59376 }, { "epoch": 4.66, "learning_rate": 5.3440759037089316e-05, "loss": 2.5544, "step": 59377 }, { "epoch": 4.66, "learning_rate": 5.343997490786482e-05, "loss": 2.6816, "step": 59378 }, { "epoch": 4.66, "learning_rate": 5.343919077864032e-05, "loss": 2.6105, "step": 59379 }, { "epoch": 4.66, "learning_rate": 5.343840664941583e-05, "loss": 2.5975, "step": 59380 }, { "epoch": 4.66, "learning_rate": 5.343762252019133e-05, "loss": 2.6709, "step": 59381 }, { "epoch": 4.66, "learning_rate": 5.3436838390966834e-05, "loss": 2.7719, "step": 59382 }, { "epoch": 4.66, "learning_rate": 5.343605426174234e-05, "loss": 2.7034, "step": 59383 }, { "epoch": 4.66, "learning_rate": 5.343527013251785e-05, "loss": 2.598, "step": 59384 }, { "epoch": 4.66, "learning_rate": 5.343448600329335e-05, "loss": 2.6714, "step": 59385 }, { "epoch": 4.66, "learning_rate": 5.3433701874068856e-05, "loss": 2.7201, "step": 59386 }, { "epoch": 4.66, "learning_rate": 5.343291774484436e-05, "loss": 2.6332, "step": 59387 }, { "epoch": 4.66, "learning_rate": 5.343213361561986e-05, "loss": 2.6999, "step": 59388 }, { "epoch": 4.66, "learning_rate": 5.343134948639537e-05, "loss": 2.6472, "step": 59389 }, { "epoch": 4.66, "learning_rate": 5.343056535717087e-05, "loss": 2.6706, "step": 59390 }, { "epoch": 4.66, "learning_rate": 5.342978122794636e-05, "loss": 2.6584, "step": 59391 }, { "epoch": 4.66, "learning_rate": 5.3428997098721864e-05, "loss": 2.5386, "step": 59392 }, { "epoch": 4.66, "learning_rate": 5.342821296949737e-05, "loss": 2.6822, "step": 59393 }, { "epoch": 4.66, "learning_rate": 5.342742884027288e-05, "loss": 2.7073, "step": 59394 }, { "epoch": 4.66, "learning_rate": 5.342664471104838e-05, "loss": 2.6863, "step": 59395 }, { "epoch": 4.66, "learning_rate": 5.3425860581823886e-05, "loss": 2.5634, "step": 59396 }, { "epoch": 4.66, "learning_rate": 5.342507645259939e-05, "loss": 2.6869, "step": 59397 }, { "epoch": 4.66, "learning_rate": 5.342429232337489e-05, "loss": 2.615, "step": 59398 }, { "epoch": 4.66, "learning_rate": 5.34235081941504e-05, "loss": 2.6491, "step": 59399 }, { "epoch": 4.66, "learning_rate": 5.34227240649259e-05, "loss": 2.6602, "step": 59400 }, { "epoch": 4.66, "learning_rate": 5.3421939935701404e-05, "loss": 2.6924, "step": 59401 }, { "epoch": 4.66, "learning_rate": 5.342115580647691e-05, "loss": 2.7622, "step": 59402 }, { "epoch": 4.66, "learning_rate": 5.342037167725241e-05, "loss": 2.6389, "step": 59403 }, { "epoch": 4.66, "learning_rate": 5.3419587548027915e-05, "loss": 2.6783, "step": 59404 }, { "epoch": 4.66, "learning_rate": 5.341880341880342e-05, "loss": 2.5039, "step": 59405 }, { "epoch": 4.66, "learning_rate": 5.341801928957892e-05, "loss": 2.6128, "step": 59406 }, { "epoch": 4.66, "learning_rate": 5.3417235160354426e-05, "loss": 2.7138, "step": 59407 }, { "epoch": 4.66, "learning_rate": 5.341645103112993e-05, "loss": 2.7963, "step": 59408 }, { "epoch": 4.66, "learning_rate": 5.3415666901905433e-05, "loss": 2.7215, "step": 59409 }, { "epoch": 4.66, "learning_rate": 5.3414882772680944e-05, "loss": 2.7841, "step": 59410 }, { "epoch": 4.66, "learning_rate": 5.341409864345645e-05, "loss": 2.8175, "step": 59411 }, { "epoch": 4.66, "learning_rate": 5.341331451423195e-05, "loss": 2.6183, "step": 59412 }, { "epoch": 4.66, "learning_rate": 5.3412530385007455e-05, "loss": 2.6661, "step": 59413 }, { "epoch": 4.66, "learning_rate": 5.341174625578296e-05, "loss": 2.6217, "step": 59414 }, { "epoch": 4.66, "learning_rate": 5.341096212655846e-05, "loss": 2.766, "step": 59415 }, { "epoch": 4.66, "learning_rate": 5.3410177997333966e-05, "loss": 2.533, "step": 59416 }, { "epoch": 4.66, "learning_rate": 5.340939386810947e-05, "loss": 2.4955, "step": 59417 }, { "epoch": 4.66, "learning_rate": 5.340860973888497e-05, "loss": 2.6141, "step": 59418 }, { "epoch": 4.66, "learning_rate": 5.340782560966048e-05, "loss": 2.6168, "step": 59419 }, { "epoch": 4.66, "learning_rate": 5.340704148043598e-05, "loss": 2.6557, "step": 59420 }, { "epoch": 4.66, "learning_rate": 5.3406257351211484e-05, "loss": 2.6631, "step": 59421 }, { "epoch": 4.66, "learning_rate": 5.340547322198699e-05, "loss": 2.6633, "step": 59422 }, { "epoch": 4.66, "learning_rate": 5.340468909276249e-05, "loss": 2.4114, "step": 59423 }, { "epoch": 4.66, "learning_rate": 5.3403904963537995e-05, "loss": 2.4173, "step": 59424 }, { "epoch": 4.66, "learning_rate": 5.3403120834313506e-05, "loss": 2.6139, "step": 59425 }, { "epoch": 4.66, "learning_rate": 5.340233670508901e-05, "loss": 2.564, "step": 59426 }, { "epoch": 4.66, "learning_rate": 5.34015525758645e-05, "loss": 2.6831, "step": 59427 }, { "epoch": 4.66, "learning_rate": 5.340076844664e-05, "loss": 2.6286, "step": 59428 }, { "epoch": 4.66, "learning_rate": 5.339998431741551e-05, "loss": 2.6174, "step": 59429 }, { "epoch": 4.66, "learning_rate": 5.339920018819101e-05, "loss": 2.6035, "step": 59430 }, { "epoch": 4.66, "learning_rate": 5.3398416058966514e-05, "loss": 2.8541, "step": 59431 }, { "epoch": 4.66, "learning_rate": 5.339763192974202e-05, "loss": 2.5102, "step": 59432 }, { "epoch": 4.66, "learning_rate": 5.339684780051752e-05, "loss": 2.5293, "step": 59433 }, { "epoch": 4.66, "learning_rate": 5.3396063671293025e-05, "loss": 2.5579, "step": 59434 }, { "epoch": 4.66, "learning_rate": 5.3395279542068536e-05, "loss": 2.89, "step": 59435 }, { "epoch": 4.66, "learning_rate": 5.339449541284404e-05, "loss": 2.8896, "step": 59436 }, { "epoch": 4.66, "learning_rate": 5.339371128361954e-05, "loss": 2.793, "step": 59437 }, { "epoch": 4.66, "learning_rate": 5.339292715439505e-05, "loss": 2.5302, "step": 59438 }, { "epoch": 4.66, "learning_rate": 5.339214302517055e-05, "loss": 2.6558, "step": 59439 }, { "epoch": 4.66, "learning_rate": 5.3391358895946054e-05, "loss": 2.4908, "step": 59440 }, { "epoch": 4.66, "learning_rate": 5.339057476672156e-05, "loss": 2.8678, "step": 59441 }, { "epoch": 4.66, "learning_rate": 5.338979063749706e-05, "loss": 2.6306, "step": 59442 }, { "epoch": 4.66, "learning_rate": 5.3389006508272565e-05, "loss": 2.7665, "step": 59443 }, { "epoch": 4.66, "learning_rate": 5.338822237904807e-05, "loss": 2.7363, "step": 59444 }, { "epoch": 4.66, "learning_rate": 5.338743824982357e-05, "loss": 2.7875, "step": 59445 }, { "epoch": 4.66, "learning_rate": 5.3386654120599076e-05, "loss": 2.6713, "step": 59446 }, { "epoch": 4.66, "learning_rate": 5.338586999137458e-05, "loss": 2.5497, "step": 59447 }, { "epoch": 4.66, "learning_rate": 5.3385085862150084e-05, "loss": 2.5296, "step": 59448 }, { "epoch": 4.66, "learning_rate": 5.338430173292559e-05, "loss": 2.7301, "step": 59449 }, { "epoch": 4.66, "learning_rate": 5.338351760370109e-05, "loss": 2.5523, "step": 59450 }, { "epoch": 4.66, "learning_rate": 5.33827334744766e-05, "loss": 2.6226, "step": 59451 }, { "epoch": 4.66, "learning_rate": 5.3381949345252105e-05, "loss": 2.6959, "step": 59452 }, { "epoch": 4.66, "learning_rate": 5.338116521602761e-05, "loss": 2.4995, "step": 59453 }, { "epoch": 4.66, "learning_rate": 5.338038108680311e-05, "loss": 2.8313, "step": 59454 }, { "epoch": 4.66, "learning_rate": 5.3379596957578616e-05, "loss": 2.9621, "step": 59455 }, { "epoch": 4.66, "learning_rate": 5.337881282835412e-05, "loss": 2.7059, "step": 59456 }, { "epoch": 4.66, "learning_rate": 5.3378028699129624e-05, "loss": 2.9003, "step": 59457 }, { "epoch": 4.66, "learning_rate": 5.337724456990513e-05, "loss": 2.7536, "step": 59458 }, { "epoch": 4.66, "learning_rate": 5.337646044068063e-05, "loss": 2.629, "step": 59459 }, { "epoch": 4.66, "learning_rate": 5.3375676311456135e-05, "loss": 2.5865, "step": 59460 }, { "epoch": 4.66, "learning_rate": 5.337489218223164e-05, "loss": 2.665, "step": 59461 }, { "epoch": 4.66, "learning_rate": 5.337410805300714e-05, "loss": 2.7118, "step": 59462 }, { "epoch": 4.66, "learning_rate": 5.337332392378264e-05, "loss": 2.8923, "step": 59463 }, { "epoch": 4.66, "learning_rate": 5.337253979455814e-05, "loss": 2.7068, "step": 59464 }, { "epoch": 4.66, "learning_rate": 5.3371755665333646e-05, "loss": 3.0165, "step": 59465 }, { "epoch": 4.66, "learning_rate": 5.337097153610915e-05, "loss": 2.951, "step": 59466 }, { "epoch": 4.66, "learning_rate": 5.3370187406884654e-05, "loss": 2.3777, "step": 59467 }, { "epoch": 4.66, "learning_rate": 5.336940327766016e-05, "loss": 2.5496, "step": 59468 }, { "epoch": 4.66, "learning_rate": 5.336861914843566e-05, "loss": 2.5664, "step": 59469 }, { "epoch": 4.66, "learning_rate": 5.3367835019211165e-05, "loss": 2.5865, "step": 59470 }, { "epoch": 4.66, "learning_rate": 5.336705088998667e-05, "loss": 2.9251, "step": 59471 }, { "epoch": 4.66, "learning_rate": 5.336626676076217e-05, "loss": 2.739, "step": 59472 }, { "epoch": 4.66, "learning_rate": 5.3365482631537676e-05, "loss": 2.5081, "step": 59473 }, { "epoch": 4.66, "learning_rate": 5.336469850231318e-05, "loss": 2.6848, "step": 59474 }, { "epoch": 4.66, "learning_rate": 5.336391437308868e-05, "loss": 2.6762, "step": 59475 }, { "epoch": 4.66, "learning_rate": 5.336313024386419e-05, "loss": 2.4468, "step": 59476 }, { "epoch": 4.66, "learning_rate": 5.33623461146397e-05, "loss": 2.736, "step": 59477 }, { "epoch": 4.66, "learning_rate": 5.33615619854152e-05, "loss": 2.7867, "step": 59478 }, { "epoch": 4.66, "learning_rate": 5.3360777856190704e-05, "loss": 2.86, "step": 59479 }, { "epoch": 4.66, "learning_rate": 5.335999372696621e-05, "loss": 2.737, "step": 59480 }, { "epoch": 4.66, "learning_rate": 5.335920959774171e-05, "loss": 2.6053, "step": 59481 }, { "epoch": 4.66, "learning_rate": 5.3358425468517215e-05, "loss": 2.6753, "step": 59482 }, { "epoch": 4.66, "learning_rate": 5.335764133929272e-05, "loss": 2.5338, "step": 59483 }, { "epoch": 4.66, "learning_rate": 5.335685721006822e-05, "loss": 2.6099, "step": 59484 }, { "epoch": 4.66, "learning_rate": 5.3356073080843727e-05, "loss": 2.471, "step": 59485 }, { "epoch": 4.66, "learning_rate": 5.335528895161923e-05, "loss": 2.6079, "step": 59486 }, { "epoch": 4.66, "learning_rate": 5.3354504822394734e-05, "loss": 2.7479, "step": 59487 }, { "epoch": 4.66, "learning_rate": 5.335372069317024e-05, "loss": 2.5008, "step": 59488 }, { "epoch": 4.66, "learning_rate": 5.335293656394574e-05, "loss": 2.7361, "step": 59489 }, { "epoch": 4.66, "learning_rate": 5.3352152434721245e-05, "loss": 2.7141, "step": 59490 }, { "epoch": 4.66, "learning_rate": 5.335136830549675e-05, "loss": 2.557, "step": 59491 }, { "epoch": 4.66, "learning_rate": 5.335058417627226e-05, "loss": 2.6596, "step": 59492 }, { "epoch": 4.67, "learning_rate": 5.334980004704776e-05, "loss": 2.5953, "step": 59493 }, { "epoch": 4.67, "learning_rate": 5.3349015917823266e-05, "loss": 2.758, "step": 59494 }, { "epoch": 4.67, "learning_rate": 5.334823178859877e-05, "loss": 2.4742, "step": 59495 }, { "epoch": 4.67, "learning_rate": 5.3347447659374274e-05, "loss": 2.6794, "step": 59496 }, { "epoch": 4.67, "learning_rate": 5.334666353014978e-05, "loss": 2.5745, "step": 59497 }, { "epoch": 4.67, "learning_rate": 5.334587940092528e-05, "loss": 2.7125, "step": 59498 }, { "epoch": 4.67, "learning_rate": 5.334509527170077e-05, "loss": 2.7842, "step": 59499 }, { "epoch": 4.67, "learning_rate": 5.3344311142476275e-05, "loss": 2.7623, "step": 59500 }, { "epoch": 4.67, "learning_rate": 5.334352701325178e-05, "loss": 2.6676, "step": 59501 }, { "epoch": 4.67, "learning_rate": 5.334274288402729e-05, "loss": 2.8434, "step": 59502 }, { "epoch": 4.67, "learning_rate": 5.334195875480279e-05, "loss": 2.5581, "step": 59503 }, { "epoch": 4.67, "learning_rate": 5.3341174625578296e-05, "loss": 2.7831, "step": 59504 }, { "epoch": 4.67, "learning_rate": 5.33403904963538e-05, "loss": 2.5348, "step": 59505 }, { "epoch": 4.67, "learning_rate": 5.3339606367129304e-05, "loss": 2.5044, "step": 59506 }, { "epoch": 4.67, "learning_rate": 5.333882223790481e-05, "loss": 2.5013, "step": 59507 }, { "epoch": 4.67, "learning_rate": 5.333803810868031e-05, "loss": 2.6072, "step": 59508 }, { "epoch": 4.67, "learning_rate": 5.3337253979455815e-05, "loss": 2.7557, "step": 59509 }, { "epoch": 4.67, "learning_rate": 5.333646985023132e-05, "loss": 2.4189, "step": 59510 }, { "epoch": 4.67, "learning_rate": 5.333568572100682e-05, "loss": 2.6186, "step": 59511 }, { "epoch": 4.67, "learning_rate": 5.3334901591782326e-05, "loss": 2.7388, "step": 59512 }, { "epoch": 4.67, "learning_rate": 5.333411746255783e-05, "loss": 2.5558, "step": 59513 }, { "epoch": 4.67, "learning_rate": 5.333333333333333e-05, "loss": 2.7638, "step": 59514 }, { "epoch": 4.67, "learning_rate": 5.333254920410884e-05, "loss": 2.7236, "step": 59515 }, { "epoch": 4.67, "learning_rate": 5.333176507488434e-05, "loss": 2.5886, "step": 59516 }, { "epoch": 4.67, "learning_rate": 5.3330980945659844e-05, "loss": 2.4998, "step": 59517 }, { "epoch": 4.67, "learning_rate": 5.3330196816435355e-05, "loss": 2.7959, "step": 59518 }, { "epoch": 4.67, "learning_rate": 5.332941268721086e-05, "loss": 2.5519, "step": 59519 }, { "epoch": 4.67, "learning_rate": 5.332862855798636e-05, "loss": 2.5202, "step": 59520 }, { "epoch": 4.67, "learning_rate": 5.3327844428761866e-05, "loss": 2.6318, "step": 59521 }, { "epoch": 4.67, "learning_rate": 5.332706029953737e-05, "loss": 2.6405, "step": 59522 }, { "epoch": 4.67, "learning_rate": 5.332627617031287e-05, "loss": 2.6361, "step": 59523 }, { "epoch": 4.67, "learning_rate": 5.332549204108838e-05, "loss": 2.6636, "step": 59524 }, { "epoch": 4.67, "learning_rate": 5.332470791186388e-05, "loss": 2.7675, "step": 59525 }, { "epoch": 4.67, "learning_rate": 5.3323923782639384e-05, "loss": 2.6701, "step": 59526 }, { "epoch": 4.67, "learning_rate": 5.332313965341489e-05, "loss": 2.7438, "step": 59527 }, { "epoch": 4.67, "learning_rate": 5.332235552419039e-05, "loss": 2.5275, "step": 59528 }, { "epoch": 4.67, "learning_rate": 5.3321571394965895e-05, "loss": 2.5991, "step": 59529 }, { "epoch": 4.67, "learning_rate": 5.33207872657414e-05, "loss": 3.0399, "step": 59530 }, { "epoch": 4.67, "learning_rate": 5.33200031365169e-05, "loss": 2.6907, "step": 59531 }, { "epoch": 4.67, "learning_rate": 5.3319219007292406e-05, "loss": 2.6121, "step": 59532 }, { "epoch": 4.67, "learning_rate": 5.3318434878067917e-05, "loss": 2.6885, "step": 59533 }, { "epoch": 4.67, "learning_rate": 5.331765074884342e-05, "loss": 2.7667, "step": 59534 }, { "epoch": 4.67, "learning_rate": 5.331686661961891e-05, "loss": 2.8158, "step": 59535 }, { "epoch": 4.67, "learning_rate": 5.3316082490394414e-05, "loss": 2.4889, "step": 59536 }, { "epoch": 4.67, "learning_rate": 5.331529836116992e-05, "loss": 2.801, "step": 59537 }, { "epoch": 4.67, "learning_rate": 5.331451423194542e-05, "loss": 2.742, "step": 59538 }, { "epoch": 4.67, "learning_rate": 5.3313730102720925e-05, "loss": 2.5818, "step": 59539 }, { "epoch": 4.67, "learning_rate": 5.331294597349643e-05, "loss": 2.9057, "step": 59540 }, { "epoch": 4.67, "learning_rate": 5.331216184427193e-05, "loss": 2.5771, "step": 59541 }, { "epoch": 4.67, "learning_rate": 5.3311377715047436e-05, "loss": 2.859, "step": 59542 }, { "epoch": 4.67, "learning_rate": 5.3310593585822947e-05, "loss": 2.6726, "step": 59543 }, { "epoch": 4.67, "learning_rate": 5.330980945659845e-05, "loss": 2.829, "step": 59544 }, { "epoch": 4.67, "learning_rate": 5.3309025327373954e-05, "loss": 2.8264, "step": 59545 }, { "epoch": 4.67, "learning_rate": 5.330824119814946e-05, "loss": 2.7207, "step": 59546 }, { "epoch": 4.67, "learning_rate": 5.330745706892496e-05, "loss": 2.5895, "step": 59547 }, { "epoch": 4.67, "learning_rate": 5.3306672939700465e-05, "loss": 2.8832, "step": 59548 }, { "epoch": 4.67, "learning_rate": 5.330588881047597e-05, "loss": 2.6427, "step": 59549 }, { "epoch": 4.67, "learning_rate": 5.330510468125147e-05, "loss": 2.6326, "step": 59550 }, { "epoch": 4.67, "learning_rate": 5.3304320552026976e-05, "loss": 2.6154, "step": 59551 }, { "epoch": 4.67, "learning_rate": 5.330353642280248e-05, "loss": 2.5135, "step": 59552 }, { "epoch": 4.67, "learning_rate": 5.330275229357798e-05, "loss": 2.531, "step": 59553 }, { "epoch": 4.67, "learning_rate": 5.330196816435349e-05, "loss": 2.837, "step": 59554 }, { "epoch": 4.67, "learning_rate": 5.330118403512899e-05, "loss": 2.7291, "step": 59555 }, { "epoch": 4.67, "learning_rate": 5.3300399905904494e-05, "loss": 2.8701, "step": 59556 }, { "epoch": 4.67, "learning_rate": 5.329961577668e-05, "loss": 2.968, "step": 59557 }, { "epoch": 4.67, "learning_rate": 5.32988316474555e-05, "loss": 2.7709, "step": 59558 }, { "epoch": 4.67, "learning_rate": 5.329804751823101e-05, "loss": 2.6862, "step": 59559 }, { "epoch": 4.67, "learning_rate": 5.3297263389006516e-05, "loss": 2.4884, "step": 59560 }, { "epoch": 4.67, "learning_rate": 5.329647925978202e-05, "loss": 2.7152, "step": 59561 }, { "epoch": 4.67, "learning_rate": 5.329569513055752e-05, "loss": 2.6264, "step": 59562 }, { "epoch": 4.67, "learning_rate": 5.329491100133303e-05, "loss": 2.7457, "step": 59563 }, { "epoch": 4.67, "learning_rate": 5.329412687210853e-05, "loss": 2.7455, "step": 59564 }, { "epoch": 4.67, "learning_rate": 5.3293342742884034e-05, "loss": 2.589, "step": 59565 }, { "epoch": 4.67, "learning_rate": 5.329255861365954e-05, "loss": 3.2089, "step": 59566 }, { "epoch": 4.67, "learning_rate": 5.329177448443504e-05, "loss": 2.5491, "step": 59567 }, { "epoch": 4.67, "learning_rate": 5.3290990355210545e-05, "loss": 2.7526, "step": 59568 }, { "epoch": 4.67, "learning_rate": 5.329020622598605e-05, "loss": 2.7822, "step": 59569 }, { "epoch": 4.67, "learning_rate": 5.328942209676155e-05, "loss": 2.6654, "step": 59570 }, { "epoch": 4.67, "learning_rate": 5.328863796753705e-05, "loss": 2.5544, "step": 59571 }, { "epoch": 4.67, "learning_rate": 5.328785383831255e-05, "loss": 2.7748, "step": 59572 }, { "epoch": 4.67, "learning_rate": 5.328706970908806e-05, "loss": 2.5774, "step": 59573 }, { "epoch": 4.67, "learning_rate": 5.328628557986356e-05, "loss": 2.5473, "step": 59574 }, { "epoch": 4.67, "learning_rate": 5.3285501450639064e-05, "loss": 2.5992, "step": 59575 }, { "epoch": 4.67, "learning_rate": 5.328471732141457e-05, "loss": 2.7948, "step": 59576 }, { "epoch": 4.67, "learning_rate": 5.328393319219007e-05, "loss": 2.4188, "step": 59577 }, { "epoch": 4.67, "learning_rate": 5.3283149062965575e-05, "loss": 2.7066, "step": 59578 }, { "epoch": 4.67, "learning_rate": 5.328236493374108e-05, "loss": 2.816, "step": 59579 }, { "epoch": 4.67, "learning_rate": 5.328158080451658e-05, "loss": 2.6754, "step": 59580 }, { "epoch": 4.67, "learning_rate": 5.3280796675292086e-05, "loss": 2.7469, "step": 59581 }, { "epoch": 4.67, "learning_rate": 5.328001254606759e-05, "loss": 2.7043, "step": 59582 }, { "epoch": 4.67, "learning_rate": 5.3279228416843094e-05, "loss": 2.7523, "step": 59583 }, { "epoch": 4.67, "learning_rate": 5.32784442876186e-05, "loss": 2.6232, "step": 59584 }, { "epoch": 4.67, "learning_rate": 5.327766015839411e-05, "loss": 2.6812, "step": 59585 }, { "epoch": 4.67, "learning_rate": 5.327687602916961e-05, "loss": 2.9137, "step": 59586 }, { "epoch": 4.67, "learning_rate": 5.3276091899945115e-05, "loss": 2.6673, "step": 59587 }, { "epoch": 4.67, "learning_rate": 5.327530777072062e-05, "loss": 2.7953, "step": 59588 }, { "epoch": 4.67, "learning_rate": 5.327452364149612e-05, "loss": 2.5957, "step": 59589 }, { "epoch": 4.67, "learning_rate": 5.3273739512271626e-05, "loss": 2.5749, "step": 59590 }, { "epoch": 4.67, "learning_rate": 5.327295538304713e-05, "loss": 2.6504, "step": 59591 }, { "epoch": 4.67, "learning_rate": 5.3272171253822634e-05, "loss": 2.7709, "step": 59592 }, { "epoch": 4.67, "learning_rate": 5.327138712459814e-05, "loss": 2.662, "step": 59593 }, { "epoch": 4.67, "learning_rate": 5.327060299537364e-05, "loss": 2.7032, "step": 59594 }, { "epoch": 4.67, "learning_rate": 5.3269818866149145e-05, "loss": 2.7066, "step": 59595 }, { "epoch": 4.67, "learning_rate": 5.326903473692465e-05, "loss": 2.683, "step": 59596 }, { "epoch": 4.67, "learning_rate": 5.326825060770015e-05, "loss": 2.6117, "step": 59597 }, { "epoch": 4.67, "learning_rate": 5.3267466478475656e-05, "loss": 2.8625, "step": 59598 }, { "epoch": 4.67, "learning_rate": 5.326668234925116e-05, "loss": 2.9882, "step": 59599 }, { "epoch": 4.67, "learning_rate": 5.326589822002667e-05, "loss": 2.6372, "step": 59600 }, { "epoch": 4.67, "learning_rate": 5.3265114090802173e-05, "loss": 2.69, "step": 59601 }, { "epoch": 4.67, "learning_rate": 5.326432996157768e-05, "loss": 2.674, "step": 59602 }, { "epoch": 4.67, "learning_rate": 5.326354583235318e-05, "loss": 2.6448, "step": 59603 }, { "epoch": 4.67, "learning_rate": 5.3262761703128685e-05, "loss": 2.7678, "step": 59604 }, { "epoch": 4.67, "learning_rate": 5.326197757390419e-05, "loss": 3.0292, "step": 59605 }, { "epoch": 4.67, "learning_rate": 5.326119344467969e-05, "loss": 2.5942, "step": 59606 }, { "epoch": 4.67, "learning_rate": 5.326040931545518e-05, "loss": 2.5905, "step": 59607 }, { "epoch": 4.67, "learning_rate": 5.3259625186230686e-05, "loss": 2.5245, "step": 59608 }, { "epoch": 4.67, "learning_rate": 5.325884105700619e-05, "loss": 2.741, "step": 59609 }, { "epoch": 4.67, "learning_rate": 5.32580569277817e-05, "loss": 2.6864, "step": 59610 }, { "epoch": 4.67, "learning_rate": 5.3257272798557203e-05, "loss": 2.8944, "step": 59611 }, { "epoch": 4.67, "learning_rate": 5.325648866933271e-05, "loss": 2.6373, "step": 59612 }, { "epoch": 4.67, "learning_rate": 5.325570454010821e-05, "loss": 2.7307, "step": 59613 }, { "epoch": 4.67, "learning_rate": 5.3254920410883715e-05, "loss": 2.6298, "step": 59614 }, { "epoch": 4.67, "learning_rate": 5.325413628165922e-05, "loss": 2.6652, "step": 59615 }, { "epoch": 4.67, "learning_rate": 5.325335215243472e-05, "loss": 2.5879, "step": 59616 }, { "epoch": 4.67, "learning_rate": 5.3252568023210226e-05, "loss": 2.7084, "step": 59617 }, { "epoch": 4.67, "learning_rate": 5.325178389398573e-05, "loss": 2.7337, "step": 59618 }, { "epoch": 4.67, "learning_rate": 5.325099976476123e-05, "loss": 2.6888, "step": 59619 }, { "epoch": 4.67, "learning_rate": 5.3250215635536737e-05, "loss": 2.5804, "step": 59620 }, { "epoch": 4.68, "learning_rate": 5.324943150631224e-05, "loss": 2.8191, "step": 59621 }, { "epoch": 4.68, "learning_rate": 5.3248647377087744e-05, "loss": 2.6739, "step": 59622 }, { "epoch": 4.68, "learning_rate": 5.324786324786325e-05, "loss": 2.4994, "step": 59623 }, { "epoch": 4.68, "learning_rate": 5.324707911863875e-05, "loss": 2.819, "step": 59624 }, { "epoch": 4.68, "learning_rate": 5.3246294989414255e-05, "loss": 2.4481, "step": 59625 }, { "epoch": 4.68, "learning_rate": 5.3245510860189765e-05, "loss": 2.7785, "step": 59626 }, { "epoch": 4.68, "learning_rate": 5.324472673096527e-05, "loss": 2.6873, "step": 59627 }, { "epoch": 4.68, "learning_rate": 5.324394260174077e-05, "loss": 2.6019, "step": 59628 }, { "epoch": 4.68, "learning_rate": 5.3243158472516276e-05, "loss": 2.6191, "step": 59629 }, { "epoch": 4.68, "learning_rate": 5.324237434329178e-05, "loss": 2.6035, "step": 59630 }, { "epoch": 4.68, "learning_rate": 5.3241590214067284e-05, "loss": 2.7822, "step": 59631 }, { "epoch": 4.68, "learning_rate": 5.324080608484279e-05, "loss": 2.6016, "step": 59632 }, { "epoch": 4.68, "learning_rate": 5.324002195561829e-05, "loss": 2.565, "step": 59633 }, { "epoch": 4.68, "learning_rate": 5.3239237826393795e-05, "loss": 2.7202, "step": 59634 }, { "epoch": 4.68, "learning_rate": 5.32384536971693e-05, "loss": 2.849, "step": 59635 }, { "epoch": 4.68, "learning_rate": 5.32376695679448e-05, "loss": 2.6073, "step": 59636 }, { "epoch": 4.68, "learning_rate": 5.3236885438720306e-05, "loss": 2.8772, "step": 59637 }, { "epoch": 4.68, "learning_rate": 5.323610130949581e-05, "loss": 2.7949, "step": 59638 }, { "epoch": 4.68, "learning_rate": 5.323531718027131e-05, "loss": 2.7643, "step": 59639 }, { "epoch": 4.68, "learning_rate": 5.323453305104682e-05, "loss": 2.7775, "step": 59640 }, { "epoch": 4.68, "learning_rate": 5.323374892182233e-05, "loss": 2.4388, "step": 59641 }, { "epoch": 4.68, "learning_rate": 5.323296479259783e-05, "loss": 2.6167, "step": 59642 }, { "epoch": 4.68, "learning_rate": 5.323218066337332e-05, "loss": 2.7397, "step": 59643 }, { "epoch": 4.68, "learning_rate": 5.3231396534148825e-05, "loss": 2.5315, "step": 59644 }, { "epoch": 4.68, "learning_rate": 5.323061240492433e-05, "loss": 2.819, "step": 59645 }, { "epoch": 4.68, "learning_rate": 5.322982827569983e-05, "loss": 2.5918, "step": 59646 }, { "epoch": 4.68, "learning_rate": 5.3229044146475336e-05, "loss": 2.7492, "step": 59647 }, { "epoch": 4.68, "learning_rate": 5.322826001725084e-05, "loss": 2.8765, "step": 59648 }, { "epoch": 4.68, "learning_rate": 5.322747588802634e-05, "loss": 2.9559, "step": 59649 }, { "epoch": 4.68, "learning_rate": 5.322669175880185e-05, "loss": 2.5407, "step": 59650 }, { "epoch": 4.68, "learning_rate": 5.322590762957736e-05, "loss": 2.7773, "step": 59651 }, { "epoch": 4.68, "learning_rate": 5.322512350035286e-05, "loss": 2.6515, "step": 59652 }, { "epoch": 4.68, "learning_rate": 5.3224339371128365e-05, "loss": 2.6176, "step": 59653 }, { "epoch": 4.68, "learning_rate": 5.322355524190387e-05, "loss": 2.7191, "step": 59654 }, { "epoch": 4.68, "learning_rate": 5.322277111267937e-05, "loss": 2.9039, "step": 59655 }, { "epoch": 4.68, "learning_rate": 5.3221986983454876e-05, "loss": 2.6438, "step": 59656 }, { "epoch": 4.68, "learning_rate": 5.322120285423038e-05, "loss": 2.7757, "step": 59657 }, { "epoch": 4.68, "learning_rate": 5.322041872500588e-05, "loss": 2.6544, "step": 59658 }, { "epoch": 4.68, "learning_rate": 5.321963459578139e-05, "loss": 2.7108, "step": 59659 }, { "epoch": 4.68, "learning_rate": 5.321885046655689e-05, "loss": 2.7633, "step": 59660 }, { "epoch": 4.68, "learning_rate": 5.3218066337332394e-05, "loss": 2.7091, "step": 59661 }, { "epoch": 4.68, "learning_rate": 5.32172822081079e-05, "loss": 2.4349, "step": 59662 }, { "epoch": 4.68, "learning_rate": 5.32164980788834e-05, "loss": 2.7777, "step": 59663 }, { "epoch": 4.68, "learning_rate": 5.3215713949658905e-05, "loss": 2.4584, "step": 59664 }, { "epoch": 4.68, "learning_rate": 5.321492982043441e-05, "loss": 2.7376, "step": 59665 }, { "epoch": 4.68, "learning_rate": 5.321414569120991e-05, "loss": 2.542, "step": 59666 }, { "epoch": 4.68, "learning_rate": 5.321336156198542e-05, "loss": 2.7995, "step": 59667 }, { "epoch": 4.68, "learning_rate": 5.321257743276093e-05, "loss": 2.7122, "step": 59668 }, { "epoch": 4.68, "learning_rate": 5.321179330353643e-05, "loss": 2.5382, "step": 59669 }, { "epoch": 4.68, "learning_rate": 5.3211009174311934e-05, "loss": 2.7551, "step": 59670 }, { "epoch": 4.68, "learning_rate": 5.321022504508744e-05, "loss": 2.6448, "step": 59671 }, { "epoch": 4.68, "learning_rate": 5.320944091586294e-05, "loss": 2.6786, "step": 59672 }, { "epoch": 4.68, "learning_rate": 5.3208656786638445e-05, "loss": 2.6233, "step": 59673 }, { "epoch": 4.68, "learning_rate": 5.320787265741395e-05, "loss": 2.7309, "step": 59674 }, { "epoch": 4.68, "learning_rate": 5.320708852818945e-05, "loss": 2.7945, "step": 59675 }, { "epoch": 4.68, "learning_rate": 5.3206304398964956e-05, "loss": 2.5565, "step": 59676 }, { "epoch": 4.68, "learning_rate": 5.320552026974046e-05, "loss": 2.5357, "step": 59677 }, { "epoch": 4.68, "learning_rate": 5.3204736140515963e-05, "loss": 2.6084, "step": 59678 }, { "epoch": 4.68, "learning_rate": 5.320395201129146e-05, "loss": 2.6563, "step": 59679 }, { "epoch": 4.68, "learning_rate": 5.3203167882066964e-05, "loss": 2.7851, "step": 59680 }, { "epoch": 4.68, "learning_rate": 5.320238375284247e-05, "loss": 2.788, "step": 59681 }, { "epoch": 4.68, "learning_rate": 5.320159962361797e-05, "loss": 2.6661, "step": 59682 }, { "epoch": 4.68, "learning_rate": 5.3200815494393475e-05, "loss": 2.7726, "step": 59683 }, { "epoch": 4.68, "learning_rate": 5.320003136516898e-05, "loss": 2.6925, "step": 59684 }, { "epoch": 4.68, "learning_rate": 5.319924723594448e-05, "loss": 2.5867, "step": 59685 }, { "epoch": 4.68, "learning_rate": 5.3198463106719986e-05, "loss": 2.6727, "step": 59686 }, { "epoch": 4.68, "learning_rate": 5.319767897749549e-05, "loss": 2.9373, "step": 59687 }, { "epoch": 4.68, "learning_rate": 5.3196894848270993e-05, "loss": 2.632, "step": 59688 }, { "epoch": 4.68, "learning_rate": 5.31961107190465e-05, "loss": 2.5377, "step": 59689 }, { "epoch": 4.68, "learning_rate": 5.3195326589822e-05, "loss": 2.9139, "step": 59690 }, { "epoch": 4.68, "learning_rate": 5.3194542460597504e-05, "loss": 2.5195, "step": 59691 }, { "epoch": 4.68, "learning_rate": 5.319375833137301e-05, "loss": 2.7343, "step": 59692 }, { "epoch": 4.68, "learning_rate": 5.319297420214852e-05, "loss": 2.7809, "step": 59693 }, { "epoch": 4.68, "learning_rate": 5.319219007292402e-05, "loss": 2.6763, "step": 59694 }, { "epoch": 4.68, "learning_rate": 5.3191405943699526e-05, "loss": 2.7551, "step": 59695 }, { "epoch": 4.68, "learning_rate": 5.319062181447503e-05, "loss": 2.5351, "step": 59696 }, { "epoch": 4.68, "learning_rate": 5.318983768525053e-05, "loss": 2.6097, "step": 59697 }, { "epoch": 4.68, "learning_rate": 5.318905355602604e-05, "loss": 2.439, "step": 59698 }, { "epoch": 4.68, "learning_rate": 5.318826942680154e-05, "loss": 2.6844, "step": 59699 }, { "epoch": 4.68, "learning_rate": 5.3187485297577044e-05, "loss": 2.7133, "step": 59700 }, { "epoch": 4.68, "learning_rate": 5.318670116835255e-05, "loss": 2.6958, "step": 59701 }, { "epoch": 4.68, "learning_rate": 5.318591703912805e-05, "loss": 2.5609, "step": 59702 }, { "epoch": 4.68, "learning_rate": 5.3185132909903555e-05, "loss": 2.7054, "step": 59703 }, { "epoch": 4.68, "learning_rate": 5.318434878067906e-05, "loss": 2.6484, "step": 59704 }, { "epoch": 4.68, "learning_rate": 5.318356465145456e-05, "loss": 2.5588, "step": 59705 }, { "epoch": 4.68, "learning_rate": 5.3182780522230066e-05, "loss": 2.7643, "step": 59706 }, { "epoch": 4.68, "learning_rate": 5.318199639300557e-05, "loss": 2.6031, "step": 59707 }, { "epoch": 4.68, "learning_rate": 5.318121226378108e-05, "loss": 2.7466, "step": 59708 }, { "epoch": 4.68, "learning_rate": 5.3180428134556584e-05, "loss": 2.4913, "step": 59709 }, { "epoch": 4.68, "learning_rate": 5.317964400533209e-05, "loss": 2.6824, "step": 59710 }, { "epoch": 4.68, "learning_rate": 5.317885987610759e-05, "loss": 2.5776, "step": 59711 }, { "epoch": 4.68, "learning_rate": 5.3178075746883095e-05, "loss": 2.6855, "step": 59712 }, { "epoch": 4.68, "learning_rate": 5.31772916176586e-05, "loss": 2.4693, "step": 59713 }, { "epoch": 4.68, "learning_rate": 5.31765074884341e-05, "loss": 2.7873, "step": 59714 }, { "epoch": 4.68, "learning_rate": 5.317572335920959e-05, "loss": 2.668, "step": 59715 }, { "epoch": 4.68, "learning_rate": 5.3174939229985096e-05, "loss": 2.7977, "step": 59716 }, { "epoch": 4.68, "learning_rate": 5.31741551007606e-05, "loss": 2.6484, "step": 59717 }, { "epoch": 4.68, "learning_rate": 5.317337097153611e-05, "loss": 2.6422, "step": 59718 }, { "epoch": 4.68, "learning_rate": 5.3172586842311614e-05, "loss": 2.742, "step": 59719 }, { "epoch": 4.68, "learning_rate": 5.317180271308712e-05, "loss": 2.8471, "step": 59720 }, { "epoch": 4.68, "learning_rate": 5.317101858386262e-05, "loss": 2.8451, "step": 59721 }, { "epoch": 4.68, "learning_rate": 5.3170234454638125e-05, "loss": 2.7824, "step": 59722 }, { "epoch": 4.68, "learning_rate": 5.316945032541363e-05, "loss": 2.8187, "step": 59723 }, { "epoch": 4.68, "learning_rate": 5.316866619618913e-05, "loss": 2.8213, "step": 59724 }, { "epoch": 4.68, "learning_rate": 5.3167882066964636e-05, "loss": 2.5459, "step": 59725 }, { "epoch": 4.68, "learning_rate": 5.316709793774014e-05, "loss": 2.72, "step": 59726 }, { "epoch": 4.68, "learning_rate": 5.3166313808515644e-05, "loss": 2.4842, "step": 59727 }, { "epoch": 4.68, "learning_rate": 5.316552967929115e-05, "loss": 2.6151, "step": 59728 }, { "epoch": 4.68, "learning_rate": 5.316474555006665e-05, "loss": 2.5673, "step": 59729 }, { "epoch": 4.68, "learning_rate": 5.3163961420842155e-05, "loss": 2.7743, "step": 59730 }, { "epoch": 4.68, "learning_rate": 5.316317729161766e-05, "loss": 2.6341, "step": 59731 }, { "epoch": 4.68, "learning_rate": 5.316239316239316e-05, "loss": 2.6276, "step": 59732 }, { "epoch": 4.68, "learning_rate": 5.3161609033168666e-05, "loss": 2.9195, "step": 59733 }, { "epoch": 4.68, "learning_rate": 5.3160824903944176e-05, "loss": 2.6339, "step": 59734 }, { "epoch": 4.68, "learning_rate": 5.316004077471968e-05, "loss": 2.7355, "step": 59735 }, { "epoch": 4.68, "learning_rate": 5.3159256645495184e-05, "loss": 2.5911, "step": 59736 }, { "epoch": 4.68, "learning_rate": 5.315847251627069e-05, "loss": 2.6013, "step": 59737 }, { "epoch": 4.68, "learning_rate": 5.315768838704619e-05, "loss": 2.7379, "step": 59738 }, { "epoch": 4.68, "learning_rate": 5.3156904257821695e-05, "loss": 2.5005, "step": 59739 }, { "epoch": 4.68, "learning_rate": 5.31561201285972e-05, "loss": 2.7599, "step": 59740 }, { "epoch": 4.68, "learning_rate": 5.31553359993727e-05, "loss": 2.738, "step": 59741 }, { "epoch": 4.68, "learning_rate": 5.3154551870148206e-05, "loss": 2.4133, "step": 59742 }, { "epoch": 4.68, "learning_rate": 5.315376774092371e-05, "loss": 2.5548, "step": 59743 }, { "epoch": 4.68, "learning_rate": 5.315298361169921e-05, "loss": 2.5613, "step": 59744 }, { "epoch": 4.68, "learning_rate": 5.3152199482474717e-05, "loss": 2.5632, "step": 59745 }, { "epoch": 4.68, "learning_rate": 5.315141535325022e-05, "loss": 2.6046, "step": 59746 }, { "epoch": 4.68, "learning_rate": 5.3150631224025724e-05, "loss": 2.5138, "step": 59747 }, { "epoch": 4.69, "learning_rate": 5.314984709480123e-05, "loss": 2.6879, "step": 59748 }, { "epoch": 4.69, "learning_rate": 5.314906296557673e-05, "loss": 2.7043, "step": 59749 }, { "epoch": 4.69, "learning_rate": 5.314827883635224e-05, "loss": 2.6668, "step": 59750 }, { "epoch": 4.69, "learning_rate": 5.314749470712773e-05, "loss": 2.7552, "step": 59751 }, { "epoch": 4.69, "learning_rate": 5.3146710577903236e-05, "loss": 2.589, "step": 59752 }, { "epoch": 4.69, "learning_rate": 5.314592644867874e-05, "loss": 2.5473, "step": 59753 }, { "epoch": 4.69, "learning_rate": 5.314514231945424e-05, "loss": 2.5876, "step": 59754 }, { "epoch": 4.69, "learning_rate": 5.3144358190229747e-05, "loss": 2.9439, "step": 59755 }, { "epoch": 4.69, "learning_rate": 5.314357406100525e-05, "loss": 2.6722, "step": 59756 }, { "epoch": 4.69, "learning_rate": 5.3142789931780754e-05, "loss": 2.637, "step": 59757 }, { "epoch": 4.69, "learning_rate": 5.314200580255626e-05, "loss": 2.6202, "step": 59758 }, { "epoch": 4.69, "learning_rate": 5.314122167333177e-05, "loss": 2.6639, "step": 59759 }, { "epoch": 4.69, "learning_rate": 5.314043754410727e-05, "loss": 2.7026, "step": 59760 }, { "epoch": 4.69, "learning_rate": 5.3139653414882775e-05, "loss": 2.5982, "step": 59761 }, { "epoch": 4.69, "learning_rate": 5.313886928565828e-05, "loss": 2.7713, "step": 59762 }, { "epoch": 4.69, "learning_rate": 5.313808515643378e-05, "loss": 2.7292, "step": 59763 }, { "epoch": 4.69, "learning_rate": 5.3137301027209286e-05, "loss": 2.6797, "step": 59764 }, { "epoch": 4.69, "learning_rate": 5.313651689798479e-05, "loss": 2.6076, "step": 59765 }, { "epoch": 4.69, "learning_rate": 5.3135732768760294e-05, "loss": 2.8733, "step": 59766 }, { "epoch": 4.69, "learning_rate": 5.31349486395358e-05, "loss": 2.644, "step": 59767 }, { "epoch": 4.69, "learning_rate": 5.31341645103113e-05, "loss": 2.5643, "step": 59768 }, { "epoch": 4.69, "learning_rate": 5.3133380381086805e-05, "loss": 2.6051, "step": 59769 }, { "epoch": 4.69, "learning_rate": 5.313259625186231e-05, "loss": 2.764, "step": 59770 }, { "epoch": 4.69, "learning_rate": 5.313181212263781e-05, "loss": 2.7596, "step": 59771 }, { "epoch": 4.69, "learning_rate": 5.3131027993413316e-05, "loss": 2.7253, "step": 59772 }, { "epoch": 4.69, "learning_rate": 5.313024386418882e-05, "loss": 2.9063, "step": 59773 }, { "epoch": 4.69, "learning_rate": 5.312945973496432e-05, "loss": 2.4066, "step": 59774 }, { "epoch": 4.69, "learning_rate": 5.3128675605739834e-05, "loss": 2.7251, "step": 59775 }, { "epoch": 4.69, "learning_rate": 5.312789147651534e-05, "loss": 2.6056, "step": 59776 }, { "epoch": 4.69, "learning_rate": 5.312710734729084e-05, "loss": 2.7995, "step": 59777 }, { "epoch": 4.69, "learning_rate": 5.3126323218066345e-05, "loss": 2.5023, "step": 59778 }, { "epoch": 4.69, "learning_rate": 5.312553908884185e-05, "loss": 2.5847, "step": 59779 }, { "epoch": 4.69, "learning_rate": 5.312475495961735e-05, "loss": 2.634, "step": 59780 }, { "epoch": 4.69, "learning_rate": 5.3123970830392856e-05, "loss": 2.6235, "step": 59781 }, { "epoch": 4.69, "learning_rate": 5.312318670116836e-05, "loss": 2.5088, "step": 59782 }, { "epoch": 4.69, "learning_rate": 5.312240257194386e-05, "loss": 2.6594, "step": 59783 }, { "epoch": 4.69, "learning_rate": 5.312161844271937e-05, "loss": 2.5518, "step": 59784 }, { "epoch": 4.69, "learning_rate": 5.312083431349487e-05, "loss": 2.6227, "step": 59785 }, { "epoch": 4.69, "learning_rate": 5.3120050184270374e-05, "loss": 2.4904, "step": 59786 }, { "epoch": 4.69, "learning_rate": 5.311926605504587e-05, "loss": 2.7331, "step": 59787 }, { "epoch": 4.69, "learning_rate": 5.3118481925821375e-05, "loss": 2.6967, "step": 59788 }, { "epoch": 4.69, "learning_rate": 5.311769779659688e-05, "loss": 2.5718, "step": 59789 }, { "epoch": 4.69, "learning_rate": 5.311691366737238e-05, "loss": 2.6024, "step": 59790 }, { "epoch": 4.69, "learning_rate": 5.3116129538147886e-05, "loss": 2.5185, "step": 59791 }, { "epoch": 4.69, "learning_rate": 5.311534540892339e-05, "loss": 2.6117, "step": 59792 }, { "epoch": 4.69, "learning_rate": 5.311456127969889e-05, "loss": 2.649, "step": 59793 }, { "epoch": 4.69, "learning_rate": 5.31137771504744e-05, "loss": 2.511, "step": 59794 }, { "epoch": 4.69, "learning_rate": 5.31129930212499e-05, "loss": 2.7177, "step": 59795 }, { "epoch": 4.69, "learning_rate": 5.3112208892025404e-05, "loss": 2.5581, "step": 59796 }, { "epoch": 4.69, "learning_rate": 5.311142476280091e-05, "loss": 2.6587, "step": 59797 }, { "epoch": 4.69, "learning_rate": 5.311064063357641e-05, "loss": 2.5434, "step": 59798 }, { "epoch": 4.69, "learning_rate": 5.3109856504351915e-05, "loss": 2.6672, "step": 59799 }, { "epoch": 4.69, "learning_rate": 5.310907237512742e-05, "loss": 2.6469, "step": 59800 }, { "epoch": 4.69, "learning_rate": 5.310828824590293e-05, "loss": 2.6266, "step": 59801 }, { "epoch": 4.69, "learning_rate": 5.310750411667843e-05, "loss": 2.7919, "step": 59802 }, { "epoch": 4.69, "learning_rate": 5.310671998745394e-05, "loss": 2.7569, "step": 59803 }, { "epoch": 4.69, "learning_rate": 5.310593585822944e-05, "loss": 2.6873, "step": 59804 }, { "epoch": 4.69, "learning_rate": 5.3105151729004944e-05, "loss": 2.6, "step": 59805 }, { "epoch": 4.69, "learning_rate": 5.310436759978045e-05, "loss": 2.7698, "step": 59806 }, { "epoch": 4.69, "learning_rate": 5.310358347055595e-05, "loss": 2.7072, "step": 59807 }, { "epoch": 4.69, "learning_rate": 5.3102799341331455e-05, "loss": 2.7759, "step": 59808 }, { "epoch": 4.69, "learning_rate": 5.310201521210696e-05, "loss": 2.7383, "step": 59809 }, { "epoch": 4.69, "learning_rate": 5.310123108288246e-05, "loss": 2.7402, "step": 59810 }, { "epoch": 4.69, "learning_rate": 5.3100446953657966e-05, "loss": 2.4858, "step": 59811 }, { "epoch": 4.69, "learning_rate": 5.309966282443347e-05, "loss": 2.5604, "step": 59812 }, { "epoch": 4.69, "learning_rate": 5.3098878695208973e-05, "loss": 2.8451, "step": 59813 }, { "epoch": 4.69, "learning_rate": 5.309809456598448e-05, "loss": 2.6556, "step": 59814 }, { "epoch": 4.69, "learning_rate": 5.309731043675998e-05, "loss": 2.5815, "step": 59815 }, { "epoch": 4.69, "learning_rate": 5.309652630753549e-05, "loss": 2.5282, "step": 59816 }, { "epoch": 4.69, "learning_rate": 5.3095742178310995e-05, "loss": 3.123, "step": 59817 }, { "epoch": 4.69, "learning_rate": 5.30949580490865e-05, "loss": 2.8243, "step": 59818 }, { "epoch": 4.69, "learning_rate": 5.3094173919862e-05, "loss": 2.8693, "step": 59819 }, { "epoch": 4.69, "learning_rate": 5.3093389790637506e-05, "loss": 2.5458, "step": 59820 }, { "epoch": 4.69, "learning_rate": 5.309260566141301e-05, "loss": 2.6615, "step": 59821 }, { "epoch": 4.69, "learning_rate": 5.309182153218851e-05, "loss": 2.5969, "step": 59822 }, { "epoch": 4.69, "learning_rate": 5.3091037402964003e-05, "loss": 2.5926, "step": 59823 }, { "epoch": 4.69, "learning_rate": 5.309025327373951e-05, "loss": 2.7386, "step": 59824 }, { "epoch": 4.69, "learning_rate": 5.308946914451501e-05, "loss": 2.6694, "step": 59825 }, { "epoch": 4.69, "learning_rate": 5.308868501529052e-05, "loss": 2.8163, "step": 59826 }, { "epoch": 4.69, "learning_rate": 5.3087900886066025e-05, "loss": 2.7473, "step": 59827 }, { "epoch": 4.69, "learning_rate": 5.308711675684153e-05, "loss": 2.6069, "step": 59828 }, { "epoch": 4.69, "learning_rate": 5.308633262761703e-05, "loss": 2.616, "step": 59829 }, { "epoch": 4.69, "learning_rate": 5.3085548498392536e-05, "loss": 2.6796, "step": 59830 }, { "epoch": 4.69, "learning_rate": 5.308476436916804e-05, "loss": 2.7448, "step": 59831 }, { "epoch": 4.69, "learning_rate": 5.308398023994354e-05, "loss": 2.7416, "step": 59832 }, { "epoch": 4.69, "learning_rate": 5.308319611071905e-05, "loss": 2.6813, "step": 59833 }, { "epoch": 4.69, "learning_rate": 5.308241198149455e-05, "loss": 2.5195, "step": 59834 }, { "epoch": 4.69, "learning_rate": 5.3081627852270054e-05, "loss": 2.6384, "step": 59835 }, { "epoch": 4.69, "learning_rate": 5.308084372304556e-05, "loss": 2.6221, "step": 59836 }, { "epoch": 4.69, "learning_rate": 5.308005959382106e-05, "loss": 2.51, "step": 59837 }, { "epoch": 4.69, "learning_rate": 5.3079275464596565e-05, "loss": 2.6132, "step": 59838 }, { "epoch": 4.69, "learning_rate": 5.307849133537207e-05, "loss": 2.695, "step": 59839 }, { "epoch": 4.69, "learning_rate": 5.307770720614757e-05, "loss": 2.6532, "step": 59840 }, { "epoch": 4.69, "learning_rate": 5.3076923076923076e-05, "loss": 2.6815, "step": 59841 }, { "epoch": 4.69, "learning_rate": 5.307613894769859e-05, "loss": 2.6421, "step": 59842 }, { "epoch": 4.69, "learning_rate": 5.307535481847409e-05, "loss": 2.7745, "step": 59843 }, { "epoch": 4.69, "learning_rate": 5.3074570689249594e-05, "loss": 2.5641, "step": 59844 }, { "epoch": 4.69, "learning_rate": 5.30737865600251e-05, "loss": 2.5917, "step": 59845 }, { "epoch": 4.69, "learning_rate": 5.30730024308006e-05, "loss": 2.6858, "step": 59846 }, { "epoch": 4.69, "learning_rate": 5.3072218301576105e-05, "loss": 2.6757, "step": 59847 }, { "epoch": 4.69, "learning_rate": 5.307143417235161e-05, "loss": 2.6016, "step": 59848 }, { "epoch": 4.69, "learning_rate": 5.307065004312711e-05, "loss": 2.5968, "step": 59849 }, { "epoch": 4.69, "learning_rate": 5.3069865913902616e-05, "loss": 2.6173, "step": 59850 }, { "epoch": 4.69, "learning_rate": 5.306908178467812e-05, "loss": 2.612, "step": 59851 }, { "epoch": 4.69, "learning_rate": 5.3068297655453624e-05, "loss": 2.755, "step": 59852 }, { "epoch": 4.69, "learning_rate": 5.306751352622913e-05, "loss": 2.8077, "step": 59853 }, { "epoch": 4.69, "learning_rate": 5.306672939700463e-05, "loss": 2.47, "step": 59854 }, { "epoch": 4.69, "learning_rate": 5.3065945267780135e-05, "loss": 2.7144, "step": 59855 }, { "epoch": 4.69, "learning_rate": 5.306516113855564e-05, "loss": 2.8105, "step": 59856 }, { "epoch": 4.69, "learning_rate": 5.306437700933114e-05, "loss": 2.6307, "step": 59857 }, { "epoch": 4.69, "learning_rate": 5.306359288010665e-05, "loss": 2.6338, "step": 59858 }, { "epoch": 4.69, "learning_rate": 5.306280875088214e-05, "loss": 2.6877, "step": 59859 }, { "epoch": 4.69, "learning_rate": 5.3062024621657646e-05, "loss": 2.5402, "step": 59860 }, { "epoch": 4.69, "learning_rate": 5.306124049243315e-05, "loss": 2.578, "step": 59861 }, { "epoch": 4.69, "learning_rate": 5.3060456363208654e-05, "loss": 2.7807, "step": 59862 }, { "epoch": 4.69, "learning_rate": 5.305967223398416e-05, "loss": 2.8208, "step": 59863 }, { "epoch": 4.69, "learning_rate": 5.305888810475966e-05, "loss": 2.6264, "step": 59864 }, { "epoch": 4.69, "learning_rate": 5.3058103975535165e-05, "loss": 2.7809, "step": 59865 }, { "epoch": 4.69, "learning_rate": 5.305731984631067e-05, "loss": 2.5488, "step": 59866 }, { "epoch": 4.69, "learning_rate": 5.305653571708617e-05, "loss": 2.7389, "step": 59867 }, { "epoch": 4.69, "learning_rate": 5.305575158786168e-05, "loss": 2.828, "step": 59868 }, { "epoch": 4.69, "learning_rate": 5.3054967458637186e-05, "loss": 2.7319, "step": 59869 }, { "epoch": 4.69, "learning_rate": 5.305418332941269e-05, "loss": 2.7919, "step": 59870 }, { "epoch": 4.69, "learning_rate": 5.3053399200188194e-05, "loss": 2.4123, "step": 59871 }, { "epoch": 4.69, "learning_rate": 5.30526150709637e-05, "loss": 2.893, "step": 59872 }, { "epoch": 4.69, "learning_rate": 5.30518309417392e-05, "loss": 2.7397, "step": 59873 }, { "epoch": 4.69, "learning_rate": 5.3051046812514705e-05, "loss": 2.5471, "step": 59874 }, { "epoch": 4.69, "learning_rate": 5.305026268329021e-05, "loss": 2.7899, "step": 59875 }, { "epoch": 4.7, "learning_rate": 5.304947855406571e-05, "loss": 2.743, "step": 59876 }, { "epoch": 4.7, "learning_rate": 5.3048694424841216e-05, "loss": 2.6743, "step": 59877 }, { "epoch": 4.7, "learning_rate": 5.304791029561672e-05, "loss": 2.7807, "step": 59878 }, { "epoch": 4.7, "learning_rate": 5.304712616639222e-05, "loss": 2.5041, "step": 59879 }, { "epoch": 4.7, "learning_rate": 5.304634203716773e-05, "loss": 2.8106, "step": 59880 }, { "epoch": 4.7, "learning_rate": 5.304555790794323e-05, "loss": 2.5565, "step": 59881 }, { "epoch": 4.7, "learning_rate": 5.3044773778718734e-05, "loss": 2.7194, "step": 59882 }, { "epoch": 4.7, "learning_rate": 5.3043989649494244e-05, "loss": 2.5596, "step": 59883 }, { "epoch": 4.7, "learning_rate": 5.304320552026975e-05, "loss": 2.9203, "step": 59884 }, { "epoch": 4.7, "learning_rate": 5.304242139104525e-05, "loss": 2.4796, "step": 59885 }, { "epoch": 4.7, "learning_rate": 5.3041637261820755e-05, "loss": 2.6724, "step": 59886 }, { "epoch": 4.7, "learning_rate": 5.304085313259626e-05, "loss": 2.6985, "step": 59887 }, { "epoch": 4.7, "learning_rate": 5.304006900337176e-05, "loss": 3.0844, "step": 59888 }, { "epoch": 4.7, "learning_rate": 5.3039284874147267e-05, "loss": 2.5469, "step": 59889 }, { "epoch": 4.7, "learning_rate": 5.303850074492277e-05, "loss": 2.5609, "step": 59890 }, { "epoch": 4.7, "learning_rate": 5.3037716615698274e-05, "loss": 2.7156, "step": 59891 }, { "epoch": 4.7, "learning_rate": 5.303693248647378e-05, "loss": 2.8802, "step": 59892 }, { "epoch": 4.7, "learning_rate": 5.303614835724928e-05, "loss": 2.6288, "step": 59893 }, { "epoch": 4.7, "learning_rate": 5.3035364228024785e-05, "loss": 2.6296, "step": 59894 }, { "epoch": 4.7, "learning_rate": 5.303458009880028e-05, "loss": 2.6861, "step": 59895 }, { "epoch": 4.7, "learning_rate": 5.3033795969575785e-05, "loss": 2.4882, "step": 59896 }, { "epoch": 4.7, "learning_rate": 5.303301184035129e-05, "loss": 2.6318, "step": 59897 }, { "epoch": 4.7, "learning_rate": 5.303222771112679e-05, "loss": 2.7137, "step": 59898 }, { "epoch": 4.7, "learning_rate": 5.3031443581902297e-05, "loss": 2.7957, "step": 59899 }, { "epoch": 4.7, "learning_rate": 5.30306594526778e-05, "loss": 2.7452, "step": 59900 }, { "epoch": 4.7, "learning_rate": 5.3029875323453304e-05, "loss": 2.6514, "step": 59901 }, { "epoch": 4.7, "learning_rate": 5.302909119422881e-05, "loss": 2.5396, "step": 59902 }, { "epoch": 4.7, "learning_rate": 5.302830706500431e-05, "loss": 2.4903, "step": 59903 }, { "epoch": 4.7, "learning_rate": 5.3027522935779815e-05, "loss": 2.6076, "step": 59904 }, { "epoch": 4.7, "learning_rate": 5.302673880655532e-05, "loss": 2.6583, "step": 59905 }, { "epoch": 4.7, "learning_rate": 5.302595467733082e-05, "loss": 2.6461, "step": 59906 }, { "epoch": 4.7, "learning_rate": 5.3025170548106326e-05, "loss": 2.672, "step": 59907 }, { "epoch": 4.7, "learning_rate": 5.302438641888183e-05, "loss": 2.6069, "step": 59908 }, { "epoch": 4.7, "learning_rate": 5.302360228965734e-05, "loss": 2.6697, "step": 59909 }, { "epoch": 4.7, "learning_rate": 5.3022818160432844e-05, "loss": 2.5209, "step": 59910 }, { "epoch": 4.7, "learning_rate": 5.302203403120835e-05, "loss": 2.6225, "step": 59911 }, { "epoch": 4.7, "learning_rate": 5.302124990198385e-05, "loss": 2.5327, "step": 59912 }, { "epoch": 4.7, "learning_rate": 5.3020465772759355e-05, "loss": 2.6443, "step": 59913 }, { "epoch": 4.7, "learning_rate": 5.301968164353486e-05, "loss": 2.6297, "step": 59914 }, { "epoch": 4.7, "learning_rate": 5.301889751431036e-05, "loss": 2.5852, "step": 59915 }, { "epoch": 4.7, "learning_rate": 5.3018113385085866e-05, "loss": 2.5405, "step": 59916 }, { "epoch": 4.7, "learning_rate": 5.301732925586137e-05, "loss": 2.4875, "step": 59917 }, { "epoch": 4.7, "learning_rate": 5.301654512663687e-05, "loss": 2.6643, "step": 59918 }, { "epoch": 4.7, "learning_rate": 5.301576099741238e-05, "loss": 2.6598, "step": 59919 }, { "epoch": 4.7, "learning_rate": 5.301497686818788e-05, "loss": 2.5168, "step": 59920 }, { "epoch": 4.7, "learning_rate": 5.3014192738963384e-05, "loss": 2.5238, "step": 59921 }, { "epoch": 4.7, "learning_rate": 5.301340860973889e-05, "loss": 2.5191, "step": 59922 }, { "epoch": 4.7, "learning_rate": 5.301262448051439e-05, "loss": 2.5226, "step": 59923 }, { "epoch": 4.7, "learning_rate": 5.30118403512899e-05, "loss": 2.7605, "step": 59924 }, { "epoch": 4.7, "learning_rate": 5.3011056222065406e-05, "loss": 2.668, "step": 59925 }, { "epoch": 4.7, "learning_rate": 5.301027209284091e-05, "loss": 2.8029, "step": 59926 }, { "epoch": 4.7, "learning_rate": 5.300948796361641e-05, "loss": 2.6728, "step": 59927 }, { "epoch": 4.7, "learning_rate": 5.300870383439192e-05, "loss": 2.5513, "step": 59928 }, { "epoch": 4.7, "learning_rate": 5.300791970516742e-05, "loss": 2.7859, "step": 59929 }, { "epoch": 4.7, "learning_rate": 5.3007135575942924e-05, "loss": 2.7019, "step": 59930 }, { "epoch": 4.7, "learning_rate": 5.3006351446718414e-05, "loss": 2.5201, "step": 59931 }, { "epoch": 4.7, "learning_rate": 5.300556731749392e-05, "loss": 2.7901, "step": 59932 }, { "epoch": 4.7, "learning_rate": 5.300478318826942e-05, "loss": 2.627, "step": 59933 }, { "epoch": 4.7, "learning_rate": 5.300399905904493e-05, "loss": 2.6887, "step": 59934 }, { "epoch": 4.7, "learning_rate": 5.3003214929820436e-05, "loss": 2.5053, "step": 59935 }, { "epoch": 4.7, "learning_rate": 5.300243080059594e-05, "loss": 2.5975, "step": 59936 }, { "epoch": 4.7, "learning_rate": 5.300164667137144e-05, "loss": 2.7001, "step": 59937 }, { "epoch": 4.7, "learning_rate": 5.300086254214695e-05, "loss": 2.6971, "step": 59938 }, { "epoch": 4.7, "learning_rate": 5.300007841292245e-05, "loss": 2.5622, "step": 59939 }, { "epoch": 4.7, "learning_rate": 5.2999294283697954e-05, "loss": 2.6421, "step": 59940 }, { "epoch": 4.7, "learning_rate": 5.299851015447346e-05, "loss": 2.8299, "step": 59941 }, { "epoch": 4.7, "learning_rate": 5.299772602524896e-05, "loss": 2.6904, "step": 59942 }, { "epoch": 4.7, "learning_rate": 5.2996941896024465e-05, "loss": 2.6378, "step": 59943 }, { "epoch": 4.7, "learning_rate": 5.299615776679997e-05, "loss": 2.591, "step": 59944 }, { "epoch": 4.7, "learning_rate": 5.299537363757547e-05, "loss": 2.6382, "step": 59945 }, { "epoch": 4.7, "learning_rate": 5.2994589508350976e-05, "loss": 2.7285, "step": 59946 }, { "epoch": 4.7, "learning_rate": 5.299380537912648e-05, "loss": 2.6691, "step": 59947 }, { "epoch": 4.7, "learning_rate": 5.2993021249901984e-05, "loss": 2.8484, "step": 59948 }, { "epoch": 4.7, "learning_rate": 5.299223712067749e-05, "loss": 2.6569, "step": 59949 }, { "epoch": 4.7, "learning_rate": 5.2991452991453e-05, "loss": 2.4799, "step": 59950 }, { "epoch": 4.7, "learning_rate": 5.29906688622285e-05, "loss": 2.6934, "step": 59951 }, { "epoch": 4.7, "learning_rate": 5.2989884733004005e-05, "loss": 2.5737, "step": 59952 }, { "epoch": 4.7, "learning_rate": 5.298910060377951e-05, "loss": 2.7671, "step": 59953 }, { "epoch": 4.7, "learning_rate": 5.298831647455501e-05, "loss": 2.8504, "step": 59954 }, { "epoch": 4.7, "learning_rate": 5.2987532345330516e-05, "loss": 2.6311, "step": 59955 }, { "epoch": 4.7, "learning_rate": 5.298674821610602e-05, "loss": 2.698, "step": 59956 }, { "epoch": 4.7, "learning_rate": 5.298596408688152e-05, "loss": 2.6729, "step": 59957 }, { "epoch": 4.7, "learning_rate": 5.298517995765703e-05, "loss": 2.5222, "step": 59958 }, { "epoch": 4.7, "learning_rate": 5.298439582843253e-05, "loss": 2.6019, "step": 59959 }, { "epoch": 4.7, "learning_rate": 5.2983611699208034e-05, "loss": 2.6908, "step": 59960 }, { "epoch": 4.7, "learning_rate": 5.298282756998354e-05, "loss": 2.7313, "step": 59961 }, { "epoch": 4.7, "learning_rate": 5.298204344075904e-05, "loss": 2.6821, "step": 59962 }, { "epoch": 4.7, "learning_rate": 5.2981259311534545e-05, "loss": 2.4849, "step": 59963 }, { "epoch": 4.7, "learning_rate": 5.298047518231005e-05, "loss": 2.6619, "step": 59964 }, { "epoch": 4.7, "learning_rate": 5.297969105308555e-05, "loss": 2.7273, "step": 59965 }, { "epoch": 4.7, "learning_rate": 5.297890692386106e-05, "loss": 2.5663, "step": 59966 }, { "epoch": 4.7, "learning_rate": 5.297812279463655e-05, "loss": 2.6061, "step": 59967 }, { "epoch": 4.7, "learning_rate": 5.297733866541206e-05, "loss": 2.796, "step": 59968 }, { "epoch": 4.7, "learning_rate": 5.297655453618756e-05, "loss": 2.6646, "step": 59969 }, { "epoch": 4.7, "learning_rate": 5.2975770406963064e-05, "loss": 2.6699, "step": 59970 }, { "epoch": 4.7, "learning_rate": 5.297498627773857e-05, "loss": 2.6593, "step": 59971 }, { "epoch": 4.7, "learning_rate": 5.297420214851407e-05, "loss": 2.7867, "step": 59972 }, { "epoch": 4.7, "learning_rate": 5.2973418019289575e-05, "loss": 2.5861, "step": 59973 }, { "epoch": 4.7, "learning_rate": 5.297263389006508e-05, "loss": 2.7052, "step": 59974 }, { "epoch": 4.7, "learning_rate": 5.297184976084058e-05, "loss": 2.6421, "step": 59975 }, { "epoch": 4.7, "learning_rate": 5.297106563161609e-05, "loss": 2.8357, "step": 59976 }, { "epoch": 4.7, "learning_rate": 5.29702815023916e-05, "loss": 2.6681, "step": 59977 }, { "epoch": 4.7, "learning_rate": 5.29694973731671e-05, "loss": 2.6554, "step": 59978 }, { "epoch": 4.7, "learning_rate": 5.2968713243942604e-05, "loss": 2.6259, "step": 59979 }, { "epoch": 4.7, "learning_rate": 5.296792911471811e-05, "loss": 2.697, "step": 59980 }, { "epoch": 4.7, "learning_rate": 5.296714498549361e-05, "loss": 2.4256, "step": 59981 }, { "epoch": 4.7, "learning_rate": 5.2966360856269115e-05, "loss": 2.7531, "step": 59982 }, { "epoch": 4.7, "learning_rate": 5.296557672704462e-05, "loss": 2.5186, "step": 59983 }, { "epoch": 4.7, "learning_rate": 5.296479259782012e-05, "loss": 2.5887, "step": 59984 }, { "epoch": 4.7, "learning_rate": 5.2964008468595626e-05, "loss": 2.7138, "step": 59985 }, { "epoch": 4.7, "learning_rate": 5.296322433937113e-05, "loss": 2.7489, "step": 59986 }, { "epoch": 4.7, "learning_rate": 5.2962440210146634e-05, "loss": 2.8084, "step": 59987 }, { "epoch": 4.7, "learning_rate": 5.296165608092214e-05, "loss": 2.693, "step": 59988 }, { "epoch": 4.7, "learning_rate": 5.296087195169764e-05, "loss": 2.6119, "step": 59989 }, { "epoch": 4.7, "learning_rate": 5.2960087822473145e-05, "loss": 2.6195, "step": 59990 }, { "epoch": 4.7, "learning_rate": 5.2959303693248655e-05, "loss": 2.5773, "step": 59991 }, { "epoch": 4.7, "learning_rate": 5.295851956402416e-05, "loss": 2.7268, "step": 59992 }, { "epoch": 4.7, "learning_rate": 5.295773543479966e-05, "loss": 2.8001, "step": 59993 }, { "epoch": 4.7, "learning_rate": 5.2956951305575166e-05, "loss": 2.7271, "step": 59994 }, { "epoch": 4.7, "learning_rate": 5.295616717635067e-05, "loss": 2.5159, "step": 59995 }, { "epoch": 4.7, "learning_rate": 5.2955383047126174e-05, "loss": 2.5306, "step": 59996 }, { "epoch": 4.7, "learning_rate": 5.295459891790168e-05, "loss": 2.6216, "step": 59997 }, { "epoch": 4.7, "learning_rate": 5.295381478867718e-05, "loss": 2.6935, "step": 59998 }, { "epoch": 4.7, "learning_rate": 5.2953030659452685e-05, "loss": 2.7949, "step": 59999 }, { "epoch": 4.7, "learning_rate": 5.295224653022819e-05, "loss": 2.6438, "step": 60000 }, { "epoch": 4.7, "eval_loss": 2.8942551612854004, "eval_runtime": 217.6709, "eval_samples_per_second": 52.06, "eval_steps_per_second": 3.257, "step": 60000 }, { "epoch": 4.7, "learning_rate": 5.295146240100369e-05, "loss": 2.6265, "step": 60001 }, { "epoch": 4.7, "learning_rate": 5.2950678271779196e-05, "loss": 2.6488, "step": 60002 }, { "epoch": 4.71, "learning_rate": 5.294989414255469e-05, "loss": 2.8747, "step": 60003 }, { "epoch": 4.71, "learning_rate": 5.2949110013330196e-05, "loss": 2.5912, "step": 60004 }, { "epoch": 4.71, "learning_rate": 5.29483258841057e-05, "loss": 3.0971, "step": 60005 }, { "epoch": 4.71, "learning_rate": 5.2947541754881204e-05, "loss": 2.8002, "step": 60006 }, { "epoch": 4.71, "learning_rate": 5.294675762565671e-05, "loss": 2.4275, "step": 60007 }, { "epoch": 4.71, "learning_rate": 5.294597349643221e-05, "loss": 2.6564, "step": 60008 }, { "epoch": 4.71, "learning_rate": 5.2945189367207715e-05, "loss": 2.4893, "step": 60009 }, { "epoch": 4.71, "learning_rate": 5.294440523798322e-05, "loss": 2.7081, "step": 60010 }, { "epoch": 4.71, "learning_rate": 5.294362110875872e-05, "loss": 2.8753, "step": 60011 }, { "epoch": 4.71, "learning_rate": 5.2942836979534226e-05, "loss": 2.6376, "step": 60012 }, { "epoch": 4.71, "learning_rate": 5.294205285030973e-05, "loss": 2.618, "step": 60013 }, { "epoch": 4.71, "learning_rate": 5.294126872108523e-05, "loss": 2.7498, "step": 60014 }, { "epoch": 4.71, "learning_rate": 5.294048459186074e-05, "loss": 2.6135, "step": 60015 }, { "epoch": 4.71, "learning_rate": 5.293970046263624e-05, "loss": 2.5412, "step": 60016 }, { "epoch": 4.71, "learning_rate": 5.293891633341175e-05, "loss": 2.6305, "step": 60017 }, { "epoch": 4.71, "learning_rate": 5.2938132204187255e-05, "loss": 2.6003, "step": 60018 }, { "epoch": 4.71, "learning_rate": 5.293734807496276e-05, "loss": 2.6601, "step": 60019 }, { "epoch": 4.71, "learning_rate": 5.293656394573826e-05, "loss": 2.5978, "step": 60020 }, { "epoch": 4.71, "learning_rate": 5.2935779816513766e-05, "loss": 2.6471, "step": 60021 }, { "epoch": 4.71, "learning_rate": 5.293499568728927e-05, "loss": 2.7024, "step": 60022 }, { "epoch": 4.71, "learning_rate": 5.293421155806477e-05, "loss": 2.5582, "step": 60023 }, { "epoch": 4.71, "learning_rate": 5.2933427428840277e-05, "loss": 2.5895, "step": 60024 }, { "epoch": 4.71, "learning_rate": 5.293264329961578e-05, "loss": 2.8697, "step": 60025 }, { "epoch": 4.71, "learning_rate": 5.2931859170391284e-05, "loss": 2.7947, "step": 60026 }, { "epoch": 4.71, "learning_rate": 5.293107504116679e-05, "loss": 2.7049, "step": 60027 }, { "epoch": 4.71, "learning_rate": 5.293029091194229e-05, "loss": 2.8221, "step": 60028 }, { "epoch": 4.71, "learning_rate": 5.2929506782717795e-05, "loss": 2.5385, "step": 60029 }, { "epoch": 4.71, "learning_rate": 5.29287226534933e-05, "loss": 2.5278, "step": 60030 }, { "epoch": 4.71, "learning_rate": 5.29279385242688e-05, "loss": 2.8428, "step": 60031 }, { "epoch": 4.71, "learning_rate": 5.292715439504431e-05, "loss": 2.713, "step": 60032 }, { "epoch": 4.71, "learning_rate": 5.2926370265819816e-05, "loss": 2.68, "step": 60033 }, { "epoch": 4.71, "learning_rate": 5.292558613659532e-05, "loss": 2.807, "step": 60034 }, { "epoch": 4.71, "learning_rate": 5.2924802007370824e-05, "loss": 2.6131, "step": 60035 }, { "epoch": 4.71, "learning_rate": 5.292401787814633e-05, "loss": 2.5828, "step": 60036 }, { "epoch": 4.71, "learning_rate": 5.292323374892183e-05, "loss": 2.8037, "step": 60037 }, { "epoch": 4.71, "learning_rate": 5.2922449619697335e-05, "loss": 2.8974, "step": 60038 }, { "epoch": 4.71, "learning_rate": 5.2921665490472825e-05, "loss": 2.6322, "step": 60039 }, { "epoch": 4.71, "learning_rate": 5.292088136124833e-05, "loss": 2.59, "step": 60040 }, { "epoch": 4.71, "learning_rate": 5.292009723202383e-05, "loss": 2.6025, "step": 60041 }, { "epoch": 4.71, "learning_rate": 5.291931310279934e-05, "loss": 2.7109, "step": 60042 }, { "epoch": 4.71, "learning_rate": 5.2918528973574846e-05, "loss": 2.5797, "step": 60043 }, { "epoch": 4.71, "learning_rate": 5.291774484435035e-05, "loss": 2.8238, "step": 60044 }, { "epoch": 4.71, "learning_rate": 5.2916960715125854e-05, "loss": 2.5905, "step": 60045 }, { "epoch": 4.71, "learning_rate": 5.291617658590136e-05, "loss": 2.862, "step": 60046 }, { "epoch": 4.71, "learning_rate": 5.291539245667686e-05, "loss": 2.7934, "step": 60047 }, { "epoch": 4.71, "learning_rate": 5.2914608327452365e-05, "loss": 2.6653, "step": 60048 }, { "epoch": 4.71, "learning_rate": 5.291382419822787e-05, "loss": 2.7212, "step": 60049 }, { "epoch": 4.71, "learning_rate": 5.291304006900337e-05, "loss": 2.6323, "step": 60050 }, { "epoch": 4.71, "learning_rate": 5.2912255939778876e-05, "loss": 2.7151, "step": 60051 }, { "epoch": 4.71, "learning_rate": 5.291147181055438e-05, "loss": 2.6036, "step": 60052 }, { "epoch": 4.71, "learning_rate": 5.291068768132988e-05, "loss": 2.5594, "step": 60053 }, { "epoch": 4.71, "learning_rate": 5.290990355210539e-05, "loss": 2.7175, "step": 60054 }, { "epoch": 4.71, "learning_rate": 5.290911942288089e-05, "loss": 2.5916, "step": 60055 }, { "epoch": 4.71, "learning_rate": 5.2908335293656394e-05, "loss": 2.6379, "step": 60056 }, { "epoch": 4.71, "learning_rate": 5.29075511644319e-05, "loss": 2.6209, "step": 60057 }, { "epoch": 4.71, "learning_rate": 5.290676703520741e-05, "loss": 2.6205, "step": 60058 }, { "epoch": 4.71, "learning_rate": 5.290598290598291e-05, "loss": 2.6918, "step": 60059 }, { "epoch": 4.71, "learning_rate": 5.2905198776758416e-05, "loss": 2.8806, "step": 60060 }, { "epoch": 4.71, "learning_rate": 5.290441464753392e-05, "loss": 2.4956, "step": 60061 }, { "epoch": 4.71, "learning_rate": 5.290363051830942e-05, "loss": 2.8215, "step": 60062 }, { "epoch": 4.71, "learning_rate": 5.290284638908493e-05, "loss": 2.6292, "step": 60063 }, { "epoch": 4.71, "learning_rate": 5.290206225986043e-05, "loss": 2.5079, "step": 60064 }, { "epoch": 4.71, "learning_rate": 5.2901278130635934e-05, "loss": 2.7324, "step": 60065 }, { "epoch": 4.71, "learning_rate": 5.290049400141144e-05, "loss": 2.5197, "step": 60066 }, { "epoch": 4.71, "learning_rate": 5.289970987218694e-05, "loss": 2.6091, "step": 60067 }, { "epoch": 4.71, "learning_rate": 5.2898925742962445e-05, "loss": 2.7882, "step": 60068 }, { "epoch": 4.71, "learning_rate": 5.289814161373795e-05, "loss": 2.6648, "step": 60069 }, { "epoch": 4.71, "learning_rate": 5.289735748451345e-05, "loss": 2.6131, "step": 60070 }, { "epoch": 4.71, "learning_rate": 5.2896573355288956e-05, "loss": 2.5601, "step": 60071 }, { "epoch": 4.71, "learning_rate": 5.289578922606446e-05, "loss": 2.7395, "step": 60072 }, { "epoch": 4.71, "learning_rate": 5.2895005096839964e-05, "loss": 2.9792, "step": 60073 }, { "epoch": 4.71, "learning_rate": 5.2894220967615474e-05, "loss": 2.8459, "step": 60074 }, { "epoch": 4.71, "learning_rate": 5.2893436838390964e-05, "loss": 2.4884, "step": 60075 }, { "epoch": 4.71, "learning_rate": 5.289265270916647e-05, "loss": 2.8165, "step": 60076 }, { "epoch": 4.71, "learning_rate": 5.289186857994197e-05, "loss": 2.6951, "step": 60077 }, { "epoch": 4.71, "learning_rate": 5.2891084450717475e-05, "loss": 2.7789, "step": 60078 }, { "epoch": 4.71, "learning_rate": 5.289030032149298e-05, "loss": 2.671, "step": 60079 }, { "epoch": 4.71, "learning_rate": 5.288951619226848e-05, "loss": 2.6026, "step": 60080 }, { "epoch": 4.71, "learning_rate": 5.2888732063043986e-05, "loss": 2.8722, "step": 60081 }, { "epoch": 4.71, "learning_rate": 5.288794793381949e-05, "loss": 2.7054, "step": 60082 }, { "epoch": 4.71, "learning_rate": 5.2887163804594994e-05, "loss": 2.6333, "step": 60083 }, { "epoch": 4.71, "learning_rate": 5.2886379675370504e-05, "loss": 2.6142, "step": 60084 }, { "epoch": 4.71, "learning_rate": 5.288559554614601e-05, "loss": 2.8507, "step": 60085 }, { "epoch": 4.71, "learning_rate": 5.288481141692151e-05, "loss": 2.5708, "step": 60086 }, { "epoch": 4.71, "learning_rate": 5.2884027287697015e-05, "loss": 2.7987, "step": 60087 }, { "epoch": 4.71, "learning_rate": 5.288324315847252e-05, "loss": 2.8406, "step": 60088 }, { "epoch": 4.71, "learning_rate": 5.288245902924802e-05, "loss": 2.4454, "step": 60089 }, { "epoch": 4.71, "learning_rate": 5.2881674900023526e-05, "loss": 2.965, "step": 60090 }, { "epoch": 4.71, "learning_rate": 5.288089077079903e-05, "loss": 2.5899, "step": 60091 }, { "epoch": 4.71, "learning_rate": 5.2880106641574533e-05, "loss": 2.5669, "step": 60092 }, { "epoch": 4.71, "learning_rate": 5.287932251235004e-05, "loss": 2.6971, "step": 60093 }, { "epoch": 4.71, "learning_rate": 5.287853838312554e-05, "loss": 2.4925, "step": 60094 }, { "epoch": 4.71, "learning_rate": 5.2877754253901044e-05, "loss": 2.8433, "step": 60095 }, { "epoch": 4.71, "learning_rate": 5.287697012467655e-05, "loss": 2.6036, "step": 60096 }, { "epoch": 4.71, "learning_rate": 5.287618599545205e-05, "loss": 2.7194, "step": 60097 }, { "epoch": 4.71, "learning_rate": 5.2875401866227555e-05, "loss": 2.8583, "step": 60098 }, { "epoch": 4.71, "learning_rate": 5.2874617737003066e-05, "loss": 2.7689, "step": 60099 }, { "epoch": 4.71, "learning_rate": 5.287383360777857e-05, "loss": 2.4894, "step": 60100 }, { "epoch": 4.71, "learning_rate": 5.287304947855407e-05, "loss": 2.5608, "step": 60101 }, { "epoch": 4.71, "learning_rate": 5.287226534932958e-05, "loss": 2.7164, "step": 60102 }, { "epoch": 4.71, "learning_rate": 5.287148122010508e-05, "loss": 2.5711, "step": 60103 }, { "epoch": 4.71, "learning_rate": 5.2870697090880584e-05, "loss": 2.7331, "step": 60104 }, { "epoch": 4.71, "learning_rate": 5.286991296165609e-05, "loss": 2.6846, "step": 60105 }, { "epoch": 4.71, "learning_rate": 5.286912883243159e-05, "loss": 2.8458, "step": 60106 }, { "epoch": 4.71, "learning_rate": 5.2868344703207095e-05, "loss": 2.6928, "step": 60107 }, { "epoch": 4.71, "learning_rate": 5.28675605739826e-05, "loss": 2.6905, "step": 60108 }, { "epoch": 4.71, "learning_rate": 5.28667764447581e-05, "loss": 2.594, "step": 60109 }, { "epoch": 4.71, "learning_rate": 5.2865992315533606e-05, "loss": 2.7324, "step": 60110 }, { "epoch": 4.71, "learning_rate": 5.28652081863091e-05, "loss": 2.7298, "step": 60111 }, { "epoch": 4.71, "learning_rate": 5.286442405708461e-05, "loss": 2.6151, "step": 60112 }, { "epoch": 4.71, "learning_rate": 5.286363992786011e-05, "loss": 2.6585, "step": 60113 }, { "epoch": 4.71, "learning_rate": 5.2862855798635614e-05, "loss": 2.6915, "step": 60114 }, { "epoch": 4.71, "learning_rate": 5.286207166941112e-05, "loss": 2.4897, "step": 60115 }, { "epoch": 4.71, "learning_rate": 5.286128754018662e-05, "loss": 2.4919, "step": 60116 }, { "epoch": 4.71, "learning_rate": 5.2860503410962125e-05, "loss": 2.4262, "step": 60117 }, { "epoch": 4.71, "learning_rate": 5.285971928173763e-05, "loss": 2.6826, "step": 60118 }, { "epoch": 4.71, "learning_rate": 5.285893515251313e-05, "loss": 2.5501, "step": 60119 }, { "epoch": 4.71, "learning_rate": 5.2858151023288636e-05, "loss": 2.7316, "step": 60120 }, { "epoch": 4.71, "learning_rate": 5.285736689406414e-05, "loss": 2.5775, "step": 60121 }, { "epoch": 4.71, "learning_rate": 5.2856582764839644e-05, "loss": 2.5784, "step": 60122 }, { "epoch": 4.71, "learning_rate": 5.285579863561515e-05, "loss": 2.6122, "step": 60123 }, { "epoch": 4.71, "learning_rate": 5.285501450639065e-05, "loss": 2.6144, "step": 60124 }, { "epoch": 4.71, "learning_rate": 5.285423037716616e-05, "loss": 2.5723, "step": 60125 }, { "epoch": 4.71, "learning_rate": 5.2853446247941665e-05, "loss": 2.4903, "step": 60126 }, { "epoch": 4.71, "learning_rate": 5.285266211871717e-05, "loss": 2.7393, "step": 60127 }, { "epoch": 4.71, "learning_rate": 5.285187798949267e-05, "loss": 2.6193, "step": 60128 }, { "epoch": 4.71, "learning_rate": 5.2851093860268176e-05, "loss": 2.6718, "step": 60129 }, { "epoch": 4.71, "learning_rate": 5.285030973104368e-05, "loss": 2.524, "step": 60130 }, { "epoch": 4.72, "learning_rate": 5.2849525601819184e-05, "loss": 2.7906, "step": 60131 }, { "epoch": 4.72, "learning_rate": 5.284874147259469e-05, "loss": 2.6899, "step": 60132 }, { "epoch": 4.72, "learning_rate": 5.284795734337019e-05, "loss": 2.7229, "step": 60133 }, { "epoch": 4.72, "learning_rate": 5.2847173214145695e-05, "loss": 2.5371, "step": 60134 }, { "epoch": 4.72, "learning_rate": 5.28463890849212e-05, "loss": 2.7223, "step": 60135 }, { "epoch": 4.72, "learning_rate": 5.28456049556967e-05, "loss": 2.7905, "step": 60136 }, { "epoch": 4.72, "learning_rate": 5.2844820826472206e-05, "loss": 2.7638, "step": 60137 }, { "epoch": 4.72, "learning_rate": 5.284403669724771e-05, "loss": 2.535, "step": 60138 }, { "epoch": 4.72, "learning_rate": 5.284325256802321e-05, "loss": 2.6, "step": 60139 }, { "epoch": 4.72, "learning_rate": 5.284246843879872e-05, "loss": 2.5996, "step": 60140 }, { "epoch": 4.72, "learning_rate": 5.284168430957423e-05, "loss": 2.5385, "step": 60141 }, { "epoch": 4.72, "learning_rate": 5.284090018034973e-05, "loss": 2.745, "step": 60142 }, { "epoch": 4.72, "learning_rate": 5.2840116051125235e-05, "loss": 2.8544, "step": 60143 }, { "epoch": 4.72, "learning_rate": 5.283933192190074e-05, "loss": 2.5624, "step": 60144 }, { "epoch": 4.72, "learning_rate": 5.283854779267624e-05, "loss": 2.7733, "step": 60145 }, { "epoch": 4.72, "learning_rate": 5.2837763663451746e-05, "loss": 2.6467, "step": 60146 }, { "epoch": 4.72, "learning_rate": 5.2836979534227236e-05, "loss": 2.7118, "step": 60147 }, { "epoch": 4.72, "learning_rate": 5.283619540500274e-05, "loss": 2.4179, "step": 60148 }, { "epoch": 4.72, "learning_rate": 5.283541127577824e-05, "loss": 2.9853, "step": 60149 }, { "epoch": 4.72, "learning_rate": 5.2834627146553754e-05, "loss": 2.5724, "step": 60150 }, { "epoch": 4.72, "learning_rate": 5.283384301732926e-05, "loss": 2.5853, "step": 60151 }, { "epoch": 4.72, "learning_rate": 5.283305888810476e-05, "loss": 2.5399, "step": 60152 }, { "epoch": 4.72, "learning_rate": 5.2832274758880265e-05, "loss": 2.6726, "step": 60153 }, { "epoch": 4.72, "learning_rate": 5.283149062965577e-05, "loss": 2.8413, "step": 60154 }, { "epoch": 4.72, "learning_rate": 5.283070650043127e-05, "loss": 2.7473, "step": 60155 }, { "epoch": 4.72, "learning_rate": 5.2829922371206776e-05, "loss": 2.8939, "step": 60156 }, { "epoch": 4.72, "learning_rate": 5.282913824198228e-05, "loss": 2.6057, "step": 60157 }, { "epoch": 4.72, "learning_rate": 5.282835411275778e-05, "loss": 2.6001, "step": 60158 }, { "epoch": 4.72, "learning_rate": 5.2827569983533287e-05, "loss": 2.7027, "step": 60159 }, { "epoch": 4.72, "learning_rate": 5.282678585430879e-05, "loss": 2.5859, "step": 60160 }, { "epoch": 4.72, "learning_rate": 5.2826001725084294e-05, "loss": 2.6847, "step": 60161 }, { "epoch": 4.72, "learning_rate": 5.28252175958598e-05, "loss": 2.6205, "step": 60162 }, { "epoch": 4.72, "learning_rate": 5.28244334666353e-05, "loss": 2.737, "step": 60163 }, { "epoch": 4.72, "learning_rate": 5.2823649337410805e-05, "loss": 2.8071, "step": 60164 }, { "epoch": 4.72, "learning_rate": 5.282286520818631e-05, "loss": 2.7011, "step": 60165 }, { "epoch": 4.72, "learning_rate": 5.282208107896182e-05, "loss": 2.4704, "step": 60166 }, { "epoch": 4.72, "learning_rate": 5.282129694973732e-05, "loss": 2.6019, "step": 60167 }, { "epoch": 4.72, "learning_rate": 5.2820512820512826e-05, "loss": 2.7489, "step": 60168 }, { "epoch": 4.72, "learning_rate": 5.281972869128833e-05, "loss": 2.4482, "step": 60169 }, { "epoch": 4.72, "learning_rate": 5.2818944562063834e-05, "loss": 2.5296, "step": 60170 }, { "epoch": 4.72, "learning_rate": 5.281816043283934e-05, "loss": 2.6194, "step": 60171 }, { "epoch": 4.72, "learning_rate": 5.281737630361484e-05, "loss": 2.5844, "step": 60172 }, { "epoch": 4.72, "learning_rate": 5.2816592174390345e-05, "loss": 2.4893, "step": 60173 }, { "epoch": 4.72, "learning_rate": 5.281580804516585e-05, "loss": 2.7518, "step": 60174 }, { "epoch": 4.72, "learning_rate": 5.281502391594135e-05, "loss": 2.6126, "step": 60175 }, { "epoch": 4.72, "learning_rate": 5.2814239786716856e-05, "loss": 2.8262, "step": 60176 }, { "epoch": 4.72, "learning_rate": 5.281345565749236e-05, "loss": 2.6025, "step": 60177 }, { "epoch": 4.72, "learning_rate": 5.281267152826786e-05, "loss": 2.6139, "step": 60178 }, { "epoch": 4.72, "learning_rate": 5.281188739904337e-05, "loss": 2.5942, "step": 60179 }, { "epoch": 4.72, "learning_rate": 5.281110326981887e-05, "loss": 2.6737, "step": 60180 }, { "epoch": 4.72, "learning_rate": 5.2810319140594374e-05, "loss": 2.6208, "step": 60181 }, { "epoch": 4.72, "learning_rate": 5.2809535011369885e-05, "loss": 2.5685, "step": 60182 }, { "epoch": 4.72, "learning_rate": 5.2808750882145375e-05, "loss": 2.762, "step": 60183 }, { "epoch": 4.72, "learning_rate": 5.280796675292088e-05, "loss": 2.5509, "step": 60184 }, { "epoch": 4.72, "learning_rate": 5.280718262369638e-05, "loss": 2.5958, "step": 60185 }, { "epoch": 4.72, "learning_rate": 5.2806398494471886e-05, "loss": 2.7164, "step": 60186 }, { "epoch": 4.72, "learning_rate": 5.280561436524739e-05, "loss": 2.7401, "step": 60187 }, { "epoch": 4.72, "learning_rate": 5.280483023602289e-05, "loss": 2.6982, "step": 60188 }, { "epoch": 4.72, "learning_rate": 5.28040461067984e-05, "loss": 2.6524, "step": 60189 }, { "epoch": 4.72, "learning_rate": 5.28032619775739e-05, "loss": 2.7642, "step": 60190 }, { "epoch": 4.72, "learning_rate": 5.2802477848349404e-05, "loss": 2.6061, "step": 60191 }, { "epoch": 4.72, "learning_rate": 5.2801693719124915e-05, "loss": 2.6031, "step": 60192 }, { "epoch": 4.72, "learning_rate": 5.280090958990042e-05, "loss": 2.9044, "step": 60193 }, { "epoch": 4.72, "learning_rate": 5.280012546067592e-05, "loss": 2.5942, "step": 60194 }, { "epoch": 4.72, "learning_rate": 5.2799341331451426e-05, "loss": 2.717, "step": 60195 }, { "epoch": 4.72, "learning_rate": 5.279855720222693e-05, "loss": 2.5417, "step": 60196 }, { "epoch": 4.72, "learning_rate": 5.279777307300243e-05, "loss": 2.5621, "step": 60197 }, { "epoch": 4.72, "learning_rate": 5.279698894377794e-05, "loss": 2.8479, "step": 60198 }, { "epoch": 4.72, "learning_rate": 5.279620481455344e-05, "loss": 2.6578, "step": 60199 }, { "epoch": 4.72, "learning_rate": 5.2795420685328944e-05, "loss": 2.6824, "step": 60200 }, { "epoch": 4.72, "learning_rate": 5.279463655610445e-05, "loss": 2.6734, "step": 60201 }, { "epoch": 4.72, "learning_rate": 5.279385242687995e-05, "loss": 2.5964, "step": 60202 }, { "epoch": 4.72, "learning_rate": 5.2793068297655455e-05, "loss": 2.6747, "step": 60203 }, { "epoch": 4.72, "learning_rate": 5.279228416843096e-05, "loss": 2.7337, "step": 60204 }, { "epoch": 4.72, "learning_rate": 5.279150003920646e-05, "loss": 2.6358, "step": 60205 }, { "epoch": 4.72, "learning_rate": 5.2790715909981966e-05, "loss": 2.7203, "step": 60206 }, { "epoch": 4.72, "learning_rate": 5.278993178075748e-05, "loss": 2.5144, "step": 60207 }, { "epoch": 4.72, "learning_rate": 5.278914765153298e-05, "loss": 2.7813, "step": 60208 }, { "epoch": 4.72, "learning_rate": 5.2788363522308484e-05, "loss": 2.5856, "step": 60209 }, { "epoch": 4.72, "learning_rate": 5.278757939308399e-05, "loss": 2.5914, "step": 60210 }, { "epoch": 4.72, "learning_rate": 5.278679526385949e-05, "loss": 2.8906, "step": 60211 }, { "epoch": 4.72, "learning_rate": 5.2786011134634995e-05, "loss": 2.8588, "step": 60212 }, { "epoch": 4.72, "learning_rate": 5.27852270054105e-05, "loss": 2.5248, "step": 60213 }, { "epoch": 4.72, "learning_rate": 5.2784442876186e-05, "loss": 2.5689, "step": 60214 }, { "epoch": 4.72, "learning_rate": 5.2783658746961506e-05, "loss": 2.8639, "step": 60215 }, { "epoch": 4.72, "learning_rate": 5.278287461773701e-05, "loss": 2.7465, "step": 60216 }, { "epoch": 4.72, "learning_rate": 5.2782090488512513e-05, "loss": 2.6565, "step": 60217 }, { "epoch": 4.72, "learning_rate": 5.278130635928802e-05, "loss": 2.587, "step": 60218 }, { "epoch": 4.72, "learning_rate": 5.2780522230063514e-05, "loss": 2.7159, "step": 60219 }, { "epoch": 4.72, "learning_rate": 5.277973810083902e-05, "loss": 2.9988, "step": 60220 }, { "epoch": 4.72, "learning_rate": 5.277895397161452e-05, "loss": 2.5765, "step": 60221 }, { "epoch": 4.72, "learning_rate": 5.2778169842390025e-05, "loss": 2.7753, "step": 60222 }, { "epoch": 4.72, "learning_rate": 5.277738571316553e-05, "loss": 2.7507, "step": 60223 }, { "epoch": 4.72, "learning_rate": 5.277660158394103e-05, "loss": 2.7091, "step": 60224 }, { "epoch": 4.72, "learning_rate": 5.2775817454716536e-05, "loss": 2.6637, "step": 60225 }, { "epoch": 4.72, "learning_rate": 5.277503332549204e-05, "loss": 2.6805, "step": 60226 }, { "epoch": 4.72, "learning_rate": 5.2774249196267543e-05, "loss": 2.6784, "step": 60227 }, { "epoch": 4.72, "learning_rate": 5.277346506704305e-05, "loss": 2.8342, "step": 60228 }, { "epoch": 4.72, "learning_rate": 5.277268093781855e-05, "loss": 2.5756, "step": 60229 }, { "epoch": 4.72, "learning_rate": 5.2771896808594055e-05, "loss": 2.6626, "step": 60230 }, { "epoch": 4.72, "learning_rate": 5.277111267936956e-05, "loss": 2.6013, "step": 60231 }, { "epoch": 4.72, "learning_rate": 5.277032855014506e-05, "loss": 2.6452, "step": 60232 }, { "epoch": 4.72, "learning_rate": 5.276954442092057e-05, "loss": 2.4535, "step": 60233 }, { "epoch": 4.72, "learning_rate": 5.2768760291696076e-05, "loss": 2.7931, "step": 60234 }, { "epoch": 4.72, "learning_rate": 5.276797616247158e-05, "loss": 2.6456, "step": 60235 }, { "epoch": 4.72, "learning_rate": 5.276719203324708e-05, "loss": 2.6612, "step": 60236 }, { "epoch": 4.72, "learning_rate": 5.276640790402259e-05, "loss": 2.6972, "step": 60237 }, { "epoch": 4.72, "learning_rate": 5.276562377479809e-05, "loss": 2.4749, "step": 60238 }, { "epoch": 4.72, "learning_rate": 5.2764839645573594e-05, "loss": 2.7567, "step": 60239 }, { "epoch": 4.72, "learning_rate": 5.27640555163491e-05, "loss": 2.675, "step": 60240 }, { "epoch": 4.72, "learning_rate": 5.27632713871246e-05, "loss": 2.6427, "step": 60241 }, { "epoch": 4.72, "learning_rate": 5.2762487257900105e-05, "loss": 2.6409, "step": 60242 }, { "epoch": 4.72, "learning_rate": 5.276170312867561e-05, "loss": 2.6829, "step": 60243 }, { "epoch": 4.72, "learning_rate": 5.276091899945111e-05, "loss": 2.6385, "step": 60244 }, { "epoch": 4.72, "learning_rate": 5.2760134870226616e-05, "loss": 2.7001, "step": 60245 }, { "epoch": 4.72, "learning_rate": 5.275935074100212e-05, "loss": 2.7796, "step": 60246 }, { "epoch": 4.72, "learning_rate": 5.2758566611777624e-05, "loss": 2.752, "step": 60247 }, { "epoch": 4.72, "learning_rate": 5.275778248255313e-05, "loss": 2.8073, "step": 60248 }, { "epoch": 4.72, "learning_rate": 5.275699835332864e-05, "loss": 2.7664, "step": 60249 }, { "epoch": 4.72, "learning_rate": 5.275621422410414e-05, "loss": 2.6096, "step": 60250 }, { "epoch": 4.72, "learning_rate": 5.2755430094879645e-05, "loss": 2.7111, "step": 60251 }, { "epoch": 4.72, "learning_rate": 5.275464596565515e-05, "loss": 2.7997, "step": 60252 }, { "epoch": 4.72, "learning_rate": 5.275386183643065e-05, "loss": 2.7245, "step": 60253 }, { "epoch": 4.72, "learning_rate": 5.2753077707206156e-05, "loss": 2.7428, "step": 60254 }, { "epoch": 4.72, "learning_rate": 5.2752293577981646e-05, "loss": 2.6305, "step": 60255 }, { "epoch": 4.72, "learning_rate": 5.275150944875715e-05, "loss": 2.709, "step": 60256 }, { "epoch": 4.72, "learning_rate": 5.2750725319532654e-05, "loss": 2.6102, "step": 60257 }, { "epoch": 4.73, "learning_rate": 5.274994119030816e-05, "loss": 2.5086, "step": 60258 }, { "epoch": 4.73, "learning_rate": 5.274915706108367e-05, "loss": 2.6628, "step": 60259 }, { "epoch": 4.73, "learning_rate": 5.274837293185917e-05, "loss": 2.6119, "step": 60260 }, { "epoch": 4.73, "learning_rate": 5.2747588802634675e-05, "loss": 2.949, "step": 60261 }, { "epoch": 4.73, "learning_rate": 5.274680467341018e-05, "loss": 2.8446, "step": 60262 }, { "epoch": 4.73, "learning_rate": 5.274602054418568e-05, "loss": 2.4617, "step": 60263 }, { "epoch": 4.73, "learning_rate": 5.2745236414961186e-05, "loss": 2.6612, "step": 60264 }, { "epoch": 4.73, "learning_rate": 5.274445228573669e-05, "loss": 2.5879, "step": 60265 }, { "epoch": 4.73, "learning_rate": 5.2743668156512194e-05, "loss": 2.5157, "step": 60266 }, { "epoch": 4.73, "learning_rate": 5.27428840272877e-05, "loss": 3.0451, "step": 60267 }, { "epoch": 4.73, "learning_rate": 5.27420998980632e-05, "loss": 2.6064, "step": 60268 }, { "epoch": 4.73, "learning_rate": 5.2741315768838705e-05, "loss": 2.8692, "step": 60269 }, { "epoch": 4.73, "learning_rate": 5.274053163961421e-05, "loss": 2.6961, "step": 60270 }, { "epoch": 4.73, "learning_rate": 5.273974751038971e-05, "loss": 2.6557, "step": 60271 }, { "epoch": 4.73, "learning_rate": 5.2738963381165216e-05, "loss": 2.6627, "step": 60272 }, { "epoch": 4.73, "learning_rate": 5.273817925194072e-05, "loss": 2.7571, "step": 60273 }, { "epoch": 4.73, "learning_rate": 5.273739512271623e-05, "loss": 2.6777, "step": 60274 }, { "epoch": 4.73, "learning_rate": 5.2736610993491734e-05, "loss": 2.6269, "step": 60275 }, { "epoch": 4.73, "learning_rate": 5.273582686426724e-05, "loss": 2.5878, "step": 60276 }, { "epoch": 4.73, "learning_rate": 5.273504273504274e-05, "loss": 2.4865, "step": 60277 }, { "epoch": 4.73, "learning_rate": 5.2734258605818245e-05, "loss": 2.5445, "step": 60278 }, { "epoch": 4.73, "learning_rate": 5.273347447659375e-05, "loss": 2.8713, "step": 60279 }, { "epoch": 4.73, "learning_rate": 5.273269034736925e-05, "loss": 2.7162, "step": 60280 }, { "epoch": 4.73, "learning_rate": 5.2731906218144756e-05, "loss": 2.9477, "step": 60281 }, { "epoch": 4.73, "learning_rate": 5.273112208892026e-05, "loss": 2.7168, "step": 60282 }, { "epoch": 4.73, "learning_rate": 5.273033795969576e-05, "loss": 2.8132, "step": 60283 }, { "epoch": 4.73, "learning_rate": 5.272955383047127e-05, "loss": 2.6377, "step": 60284 }, { "epoch": 4.73, "learning_rate": 5.272876970124677e-05, "loss": 2.6794, "step": 60285 }, { "epoch": 4.73, "learning_rate": 5.2727985572022274e-05, "loss": 2.6915, "step": 60286 }, { "epoch": 4.73, "learning_rate": 5.272720144279778e-05, "loss": 2.4332, "step": 60287 }, { "epoch": 4.73, "learning_rate": 5.272641731357328e-05, "loss": 2.5573, "step": 60288 }, { "epoch": 4.73, "learning_rate": 5.2725633184348785e-05, "loss": 2.712, "step": 60289 }, { "epoch": 4.73, "learning_rate": 5.2724849055124295e-05, "loss": 2.6934, "step": 60290 }, { "epoch": 4.73, "learning_rate": 5.2724064925899786e-05, "loss": 2.6722, "step": 60291 }, { "epoch": 4.73, "learning_rate": 5.272328079667529e-05, "loss": 2.6373, "step": 60292 }, { "epoch": 4.73, "learning_rate": 5.272249666745079e-05, "loss": 2.858, "step": 60293 }, { "epoch": 4.73, "learning_rate": 5.27217125382263e-05, "loss": 2.7653, "step": 60294 }, { "epoch": 4.73, "learning_rate": 5.27209284090018e-05, "loss": 2.6067, "step": 60295 }, { "epoch": 4.73, "learning_rate": 5.2720144279777304e-05, "loss": 2.7548, "step": 60296 }, { "epoch": 4.73, "learning_rate": 5.271936015055281e-05, "loss": 2.7677, "step": 60297 }, { "epoch": 4.73, "learning_rate": 5.271857602132831e-05, "loss": 2.7629, "step": 60298 }, { "epoch": 4.73, "learning_rate": 5.2717791892103815e-05, "loss": 2.6627, "step": 60299 }, { "epoch": 4.73, "learning_rate": 5.2717007762879325e-05, "loss": 2.5629, "step": 60300 }, { "epoch": 4.73, "learning_rate": 5.271622363365483e-05, "loss": 2.6284, "step": 60301 }, { "epoch": 4.73, "learning_rate": 5.271543950443033e-05, "loss": 2.6533, "step": 60302 }, { "epoch": 4.73, "learning_rate": 5.2714655375205837e-05, "loss": 2.5866, "step": 60303 }, { "epoch": 4.73, "learning_rate": 5.271387124598134e-05, "loss": 2.6662, "step": 60304 }, { "epoch": 4.73, "learning_rate": 5.2713087116756844e-05, "loss": 2.5955, "step": 60305 }, { "epoch": 4.73, "learning_rate": 5.271230298753235e-05, "loss": 2.7367, "step": 60306 }, { "epoch": 4.73, "learning_rate": 5.271151885830785e-05, "loss": 2.6575, "step": 60307 }, { "epoch": 4.73, "learning_rate": 5.2710734729083355e-05, "loss": 2.9262, "step": 60308 }, { "epoch": 4.73, "learning_rate": 5.270995059985886e-05, "loss": 2.7201, "step": 60309 }, { "epoch": 4.73, "learning_rate": 5.270916647063436e-05, "loss": 2.5011, "step": 60310 }, { "epoch": 4.73, "learning_rate": 5.2708382341409866e-05, "loss": 2.4402, "step": 60311 }, { "epoch": 4.73, "learning_rate": 5.270759821218537e-05, "loss": 2.6102, "step": 60312 }, { "epoch": 4.73, "learning_rate": 5.270681408296087e-05, "loss": 2.8025, "step": 60313 }, { "epoch": 4.73, "learning_rate": 5.270602995373638e-05, "loss": 2.6718, "step": 60314 }, { "epoch": 4.73, "learning_rate": 5.270524582451189e-05, "loss": 2.597, "step": 60315 }, { "epoch": 4.73, "learning_rate": 5.270446169528739e-05, "loss": 2.6829, "step": 60316 }, { "epoch": 4.73, "learning_rate": 5.2703677566062895e-05, "loss": 2.6764, "step": 60317 }, { "epoch": 4.73, "learning_rate": 5.27028934368384e-05, "loss": 2.7388, "step": 60318 }, { "epoch": 4.73, "learning_rate": 5.27021093076139e-05, "loss": 2.7718, "step": 60319 }, { "epoch": 4.73, "learning_rate": 5.2701325178389406e-05, "loss": 2.9176, "step": 60320 }, { "epoch": 4.73, "learning_rate": 5.270054104916491e-05, "loss": 2.6308, "step": 60321 }, { "epoch": 4.73, "learning_rate": 5.269975691994041e-05, "loss": 2.6542, "step": 60322 }, { "epoch": 4.73, "learning_rate": 5.269897279071592e-05, "loss": 2.5125, "step": 60323 }, { "epoch": 4.73, "learning_rate": 5.269818866149142e-05, "loss": 2.7957, "step": 60324 }, { "epoch": 4.73, "learning_rate": 5.2697404532266924e-05, "loss": 2.7343, "step": 60325 }, { "epoch": 4.73, "learning_rate": 5.269662040304243e-05, "loss": 2.6441, "step": 60326 }, { "epoch": 4.73, "learning_rate": 5.2695836273817925e-05, "loss": 2.4706, "step": 60327 }, { "epoch": 4.73, "learning_rate": 5.269505214459343e-05, "loss": 2.5583, "step": 60328 }, { "epoch": 4.73, "learning_rate": 5.269426801536893e-05, "loss": 2.8347, "step": 60329 }, { "epoch": 4.73, "learning_rate": 5.2693483886144436e-05, "loss": 2.5226, "step": 60330 }, { "epoch": 4.73, "learning_rate": 5.269269975691994e-05, "loss": 2.5371, "step": 60331 }, { "epoch": 4.73, "learning_rate": 5.269191562769544e-05, "loss": 2.7821, "step": 60332 }, { "epoch": 4.73, "learning_rate": 5.269113149847095e-05, "loss": 2.665, "step": 60333 }, { "epoch": 4.73, "learning_rate": 5.269034736924645e-05, "loss": 2.7167, "step": 60334 }, { "epoch": 4.73, "learning_rate": 5.2689563240021954e-05, "loss": 2.6208, "step": 60335 }, { "epoch": 4.73, "learning_rate": 5.268877911079746e-05, "loss": 2.4971, "step": 60336 }, { "epoch": 4.73, "learning_rate": 5.268799498157296e-05, "loss": 2.5715, "step": 60337 }, { "epoch": 4.73, "learning_rate": 5.2687210852348465e-05, "loss": 2.9473, "step": 60338 }, { "epoch": 4.73, "learning_rate": 5.268642672312397e-05, "loss": 2.982, "step": 60339 }, { "epoch": 4.73, "learning_rate": 5.268564259389947e-05, "loss": 2.6963, "step": 60340 }, { "epoch": 4.73, "learning_rate": 5.268485846467498e-05, "loss": 2.6247, "step": 60341 }, { "epoch": 4.73, "learning_rate": 5.268407433545049e-05, "loss": 2.8879, "step": 60342 }, { "epoch": 4.73, "learning_rate": 5.268329020622599e-05, "loss": 2.5958, "step": 60343 }, { "epoch": 4.73, "learning_rate": 5.2682506077001494e-05, "loss": 2.6712, "step": 60344 }, { "epoch": 4.73, "learning_rate": 5.2681721947777e-05, "loss": 2.5312, "step": 60345 }, { "epoch": 4.73, "learning_rate": 5.26809378185525e-05, "loss": 2.7053, "step": 60346 }, { "epoch": 4.73, "learning_rate": 5.2680153689328005e-05, "loss": 2.6545, "step": 60347 }, { "epoch": 4.73, "learning_rate": 5.267936956010351e-05, "loss": 2.6899, "step": 60348 }, { "epoch": 4.73, "learning_rate": 5.267858543087901e-05, "loss": 2.4626, "step": 60349 }, { "epoch": 4.73, "learning_rate": 5.2677801301654516e-05, "loss": 2.7134, "step": 60350 }, { "epoch": 4.73, "learning_rate": 5.267701717243002e-05, "loss": 2.6365, "step": 60351 }, { "epoch": 4.73, "learning_rate": 5.2676233043205524e-05, "loss": 2.6656, "step": 60352 }, { "epoch": 4.73, "learning_rate": 5.267544891398103e-05, "loss": 2.6502, "step": 60353 }, { "epoch": 4.73, "learning_rate": 5.267466478475653e-05, "loss": 2.8589, "step": 60354 }, { "epoch": 4.73, "learning_rate": 5.2673880655532035e-05, "loss": 2.8123, "step": 60355 }, { "epoch": 4.73, "learning_rate": 5.267309652630754e-05, "loss": 2.56, "step": 60356 }, { "epoch": 4.73, "learning_rate": 5.267231239708305e-05, "loss": 2.6421, "step": 60357 }, { "epoch": 4.73, "learning_rate": 5.267152826785855e-05, "loss": 2.5146, "step": 60358 }, { "epoch": 4.73, "learning_rate": 5.2670744138634056e-05, "loss": 3.0574, "step": 60359 }, { "epoch": 4.73, "learning_rate": 5.266996000940956e-05, "loss": 2.7318, "step": 60360 }, { "epoch": 4.73, "learning_rate": 5.266917588018506e-05, "loss": 2.6175, "step": 60361 }, { "epoch": 4.73, "learning_rate": 5.266839175096057e-05, "loss": 2.5612, "step": 60362 }, { "epoch": 4.73, "learning_rate": 5.266760762173606e-05, "loss": 2.7301, "step": 60363 }, { "epoch": 4.73, "learning_rate": 5.266682349251156e-05, "loss": 2.6609, "step": 60364 }, { "epoch": 4.73, "learning_rate": 5.2666039363287065e-05, "loss": 2.5296, "step": 60365 }, { "epoch": 4.73, "learning_rate": 5.266525523406257e-05, "loss": 2.4813, "step": 60366 }, { "epoch": 4.73, "learning_rate": 5.266447110483808e-05, "loss": 2.6255, "step": 60367 }, { "epoch": 4.73, "learning_rate": 5.266368697561358e-05, "loss": 2.7029, "step": 60368 }, { "epoch": 4.73, "learning_rate": 5.2662902846389086e-05, "loss": 2.7035, "step": 60369 }, { "epoch": 4.73, "learning_rate": 5.266211871716459e-05, "loss": 2.5455, "step": 60370 }, { "epoch": 4.73, "learning_rate": 5.266133458794009e-05, "loss": 2.4662, "step": 60371 }, { "epoch": 4.73, "learning_rate": 5.26605504587156e-05, "loss": 2.6852, "step": 60372 }, { "epoch": 4.73, "learning_rate": 5.26597663294911e-05, "loss": 2.6734, "step": 60373 }, { "epoch": 4.73, "learning_rate": 5.2658982200266604e-05, "loss": 2.94, "step": 60374 }, { "epoch": 4.73, "learning_rate": 5.265819807104211e-05, "loss": 2.6581, "step": 60375 }, { "epoch": 4.73, "learning_rate": 5.265741394181761e-05, "loss": 2.545, "step": 60376 }, { "epoch": 4.73, "learning_rate": 5.2656629812593115e-05, "loss": 2.8591, "step": 60377 }, { "epoch": 4.73, "learning_rate": 5.265584568336862e-05, "loss": 2.5087, "step": 60378 }, { "epoch": 4.73, "learning_rate": 5.265506155414412e-05, "loss": 2.8416, "step": 60379 }, { "epoch": 4.73, "learning_rate": 5.2654277424919626e-05, "loss": 2.6845, "step": 60380 }, { "epoch": 4.73, "learning_rate": 5.265349329569513e-05, "loss": 2.5822, "step": 60381 }, { "epoch": 4.73, "learning_rate": 5.265270916647064e-05, "loss": 2.5935, "step": 60382 }, { "epoch": 4.73, "learning_rate": 5.2651925037246144e-05, "loss": 2.6568, "step": 60383 }, { "epoch": 4.73, "learning_rate": 5.265114090802165e-05, "loss": 2.4745, "step": 60384 }, { "epoch": 4.73, "learning_rate": 5.265035677879715e-05, "loss": 2.7044, "step": 60385 }, { "epoch": 4.74, "learning_rate": 5.2649572649572655e-05, "loss": 2.7628, "step": 60386 }, { "epoch": 4.74, "learning_rate": 5.264878852034816e-05, "loss": 2.7661, "step": 60387 }, { "epoch": 4.74, "learning_rate": 5.264800439112366e-05, "loss": 2.5182, "step": 60388 }, { "epoch": 4.74, "learning_rate": 5.2647220261899166e-05, "loss": 2.3383, "step": 60389 }, { "epoch": 4.74, "learning_rate": 5.264643613267467e-05, "loss": 2.4952, "step": 60390 }, { "epoch": 4.74, "learning_rate": 5.2645652003450174e-05, "loss": 2.524, "step": 60391 }, { "epoch": 4.74, "learning_rate": 5.264486787422568e-05, "loss": 2.6501, "step": 60392 }, { "epoch": 4.74, "learning_rate": 5.264408374500118e-05, "loss": 2.5809, "step": 60393 }, { "epoch": 4.74, "learning_rate": 5.2643299615776685e-05, "loss": 2.5269, "step": 60394 }, { "epoch": 4.74, "learning_rate": 5.264251548655219e-05, "loss": 2.6541, "step": 60395 }, { "epoch": 4.74, "learning_rate": 5.264173135732769e-05, "loss": 2.5725, "step": 60396 }, { "epoch": 4.74, "learning_rate": 5.2640947228103196e-05, "loss": 2.3947, "step": 60397 }, { "epoch": 4.74, "learning_rate": 5.2640163098878706e-05, "loss": 2.732, "step": 60398 }, { "epoch": 4.74, "learning_rate": 5.2639378969654196e-05, "loss": 2.4945, "step": 60399 }, { "epoch": 4.74, "learning_rate": 5.26385948404297e-05, "loss": 2.6347, "step": 60400 }, { "epoch": 4.74, "learning_rate": 5.2637810711205204e-05, "loss": 2.7664, "step": 60401 }, { "epoch": 4.74, "learning_rate": 5.263702658198071e-05, "loss": 2.5095, "step": 60402 }, { "epoch": 4.74, "learning_rate": 5.263624245275621e-05, "loss": 2.6807, "step": 60403 }, { "epoch": 4.74, "learning_rate": 5.2635458323531715e-05, "loss": 2.9749, "step": 60404 }, { "epoch": 4.74, "learning_rate": 5.263467419430722e-05, "loss": 2.8781, "step": 60405 }, { "epoch": 4.74, "learning_rate": 5.263389006508272e-05, "loss": 2.7423, "step": 60406 }, { "epoch": 4.74, "learning_rate": 5.2633105935858226e-05, "loss": 2.7691, "step": 60407 }, { "epoch": 4.74, "learning_rate": 5.2632321806633736e-05, "loss": 2.5942, "step": 60408 }, { "epoch": 4.74, "learning_rate": 5.263153767740924e-05, "loss": 2.614, "step": 60409 }, { "epoch": 4.74, "learning_rate": 5.2630753548184744e-05, "loss": 2.4765, "step": 60410 }, { "epoch": 4.74, "learning_rate": 5.262996941896025e-05, "loss": 2.6673, "step": 60411 }, { "epoch": 4.74, "learning_rate": 5.262918528973575e-05, "loss": 2.6691, "step": 60412 }, { "epoch": 4.74, "learning_rate": 5.2628401160511255e-05, "loss": 2.6758, "step": 60413 }, { "epoch": 4.74, "learning_rate": 5.262761703128676e-05, "loss": 2.5073, "step": 60414 }, { "epoch": 4.74, "learning_rate": 5.262683290206226e-05, "loss": 2.5089, "step": 60415 }, { "epoch": 4.74, "learning_rate": 5.2626048772837766e-05, "loss": 2.5193, "step": 60416 }, { "epoch": 4.74, "learning_rate": 5.262526464361327e-05, "loss": 2.579, "step": 60417 }, { "epoch": 4.74, "learning_rate": 5.262448051438877e-05, "loss": 2.5885, "step": 60418 }, { "epoch": 4.74, "learning_rate": 5.262369638516428e-05, "loss": 2.7582, "step": 60419 }, { "epoch": 4.74, "learning_rate": 5.262291225593978e-05, "loss": 2.7267, "step": 60420 }, { "epoch": 4.74, "learning_rate": 5.2622128126715284e-05, "loss": 2.6979, "step": 60421 }, { "epoch": 4.74, "learning_rate": 5.262134399749079e-05, "loss": 2.7034, "step": 60422 }, { "epoch": 4.74, "learning_rate": 5.26205598682663e-05, "loss": 2.6336, "step": 60423 }, { "epoch": 4.74, "learning_rate": 5.26197757390418e-05, "loss": 2.6435, "step": 60424 }, { "epoch": 4.74, "learning_rate": 5.2618991609817306e-05, "loss": 2.6157, "step": 60425 }, { "epoch": 4.74, "learning_rate": 5.261820748059281e-05, "loss": 2.9763, "step": 60426 }, { "epoch": 4.74, "learning_rate": 5.261742335136831e-05, "loss": 2.6003, "step": 60427 }, { "epoch": 4.74, "learning_rate": 5.2616639222143817e-05, "loss": 2.7395, "step": 60428 }, { "epoch": 4.74, "learning_rate": 5.261585509291932e-05, "loss": 2.6977, "step": 60429 }, { "epoch": 4.74, "learning_rate": 5.2615070963694824e-05, "loss": 3.0205, "step": 60430 }, { "epoch": 4.74, "learning_rate": 5.261428683447033e-05, "loss": 2.876, "step": 60431 }, { "epoch": 4.74, "learning_rate": 5.261350270524583e-05, "loss": 2.7306, "step": 60432 }, { "epoch": 4.74, "learning_rate": 5.2612718576021335e-05, "loss": 2.7832, "step": 60433 }, { "epoch": 4.74, "learning_rate": 5.261193444679684e-05, "loss": 2.5169, "step": 60434 }, { "epoch": 4.74, "learning_rate": 5.2611150317572336e-05, "loss": 2.4506, "step": 60435 }, { "epoch": 4.74, "learning_rate": 5.261036618834784e-05, "loss": 2.5754, "step": 60436 }, { "epoch": 4.74, "learning_rate": 5.260958205912334e-05, "loss": 2.6528, "step": 60437 }, { "epoch": 4.74, "learning_rate": 5.2608797929898847e-05, "loss": 2.6879, "step": 60438 }, { "epoch": 4.74, "learning_rate": 5.260801380067435e-05, "loss": 2.5956, "step": 60439 }, { "epoch": 4.74, "learning_rate": 5.2607229671449854e-05, "loss": 2.8164, "step": 60440 }, { "epoch": 4.74, "learning_rate": 5.260644554222536e-05, "loss": 2.6904, "step": 60441 }, { "epoch": 4.74, "learning_rate": 5.260566141300086e-05, "loss": 2.813, "step": 60442 }, { "epoch": 4.74, "learning_rate": 5.2604877283776365e-05, "loss": 2.5379, "step": 60443 }, { "epoch": 4.74, "learning_rate": 5.260409315455187e-05, "loss": 2.5148, "step": 60444 }, { "epoch": 4.74, "learning_rate": 5.260330902532737e-05, "loss": 2.4488, "step": 60445 }, { "epoch": 4.74, "learning_rate": 5.2602524896102876e-05, "loss": 2.4124, "step": 60446 }, { "epoch": 4.74, "learning_rate": 5.260174076687838e-05, "loss": 2.7612, "step": 60447 }, { "epoch": 4.74, "learning_rate": 5.260095663765388e-05, "loss": 2.8244, "step": 60448 }, { "epoch": 4.74, "learning_rate": 5.2600172508429394e-05, "loss": 2.6937, "step": 60449 }, { "epoch": 4.74, "learning_rate": 5.25993883792049e-05, "loss": 2.7204, "step": 60450 }, { "epoch": 4.74, "learning_rate": 5.25986042499804e-05, "loss": 2.508, "step": 60451 }, { "epoch": 4.74, "learning_rate": 5.2597820120755905e-05, "loss": 2.5275, "step": 60452 }, { "epoch": 4.74, "learning_rate": 5.259703599153141e-05, "loss": 2.6211, "step": 60453 }, { "epoch": 4.74, "learning_rate": 5.259625186230691e-05, "loss": 2.5973, "step": 60454 }, { "epoch": 4.74, "learning_rate": 5.2595467733082416e-05, "loss": 2.8209, "step": 60455 }, { "epoch": 4.74, "learning_rate": 5.259468360385792e-05, "loss": 2.5821, "step": 60456 }, { "epoch": 4.74, "learning_rate": 5.259389947463342e-05, "loss": 2.8093, "step": 60457 }, { "epoch": 4.74, "learning_rate": 5.259311534540893e-05, "loss": 2.6901, "step": 60458 }, { "epoch": 4.74, "learning_rate": 5.259233121618443e-05, "loss": 2.5612, "step": 60459 }, { "epoch": 4.74, "learning_rate": 5.2591547086959934e-05, "loss": 2.6471, "step": 60460 }, { "epoch": 4.74, "learning_rate": 5.259076295773544e-05, "loss": 2.7056, "step": 60461 }, { "epoch": 4.74, "learning_rate": 5.258997882851094e-05, "loss": 2.7238, "step": 60462 }, { "epoch": 4.74, "learning_rate": 5.2589194699286445e-05, "loss": 2.6913, "step": 60463 }, { "epoch": 4.74, "learning_rate": 5.258841057006195e-05, "loss": 2.7229, "step": 60464 }, { "epoch": 4.74, "learning_rate": 5.258762644083746e-05, "loss": 2.4404, "step": 60465 }, { "epoch": 4.74, "learning_rate": 5.258684231161296e-05, "loss": 2.5938, "step": 60466 }, { "epoch": 4.74, "learning_rate": 5.258605818238847e-05, "loss": 2.5986, "step": 60467 }, { "epoch": 4.74, "learning_rate": 5.258527405316397e-05, "loss": 2.5518, "step": 60468 }, { "epoch": 4.74, "learning_rate": 5.2584489923939474e-05, "loss": 2.6422, "step": 60469 }, { "epoch": 4.74, "learning_rate": 5.258370579471498e-05, "loss": 2.7181, "step": 60470 }, { "epoch": 4.74, "learning_rate": 5.258292166549047e-05, "loss": 2.7278, "step": 60471 }, { "epoch": 4.74, "learning_rate": 5.258213753626597e-05, "loss": 2.6874, "step": 60472 }, { "epoch": 4.74, "learning_rate": 5.2581353407041475e-05, "loss": 2.5627, "step": 60473 }, { "epoch": 4.74, "learning_rate": 5.258056927781698e-05, "loss": 2.7543, "step": 60474 }, { "epoch": 4.74, "learning_rate": 5.257978514859249e-05, "loss": 2.6083, "step": 60475 }, { "epoch": 4.74, "learning_rate": 5.257900101936799e-05, "loss": 2.5936, "step": 60476 }, { "epoch": 4.74, "learning_rate": 5.25782168901435e-05, "loss": 2.6373, "step": 60477 }, { "epoch": 4.74, "learning_rate": 5.2577432760919e-05, "loss": 2.5108, "step": 60478 }, { "epoch": 4.74, "learning_rate": 5.2576648631694504e-05, "loss": 2.5871, "step": 60479 }, { "epoch": 4.74, "learning_rate": 5.257586450247001e-05, "loss": 2.5546, "step": 60480 }, { "epoch": 4.74, "learning_rate": 5.257508037324551e-05, "loss": 2.6173, "step": 60481 }, { "epoch": 4.74, "learning_rate": 5.2574296244021015e-05, "loss": 2.4842, "step": 60482 }, { "epoch": 4.74, "learning_rate": 5.257351211479652e-05, "loss": 2.7776, "step": 60483 }, { "epoch": 4.74, "learning_rate": 5.257272798557202e-05, "loss": 2.7965, "step": 60484 }, { "epoch": 4.74, "learning_rate": 5.2571943856347526e-05, "loss": 2.7711, "step": 60485 }, { "epoch": 4.74, "learning_rate": 5.257115972712303e-05, "loss": 2.7512, "step": 60486 }, { "epoch": 4.74, "learning_rate": 5.2570375597898534e-05, "loss": 2.5172, "step": 60487 }, { "epoch": 4.74, "learning_rate": 5.256959146867404e-05, "loss": 2.749, "step": 60488 }, { "epoch": 4.74, "learning_rate": 5.256880733944954e-05, "loss": 2.5727, "step": 60489 }, { "epoch": 4.74, "learning_rate": 5.256802321022505e-05, "loss": 2.6368, "step": 60490 }, { "epoch": 4.74, "learning_rate": 5.2567239081000555e-05, "loss": 2.7167, "step": 60491 }, { "epoch": 4.74, "learning_rate": 5.256645495177606e-05, "loss": 2.7848, "step": 60492 }, { "epoch": 4.74, "learning_rate": 5.256567082255156e-05, "loss": 2.9367, "step": 60493 }, { "epoch": 4.74, "learning_rate": 5.2564886693327066e-05, "loss": 2.7376, "step": 60494 }, { "epoch": 4.74, "learning_rate": 5.256410256410257e-05, "loss": 2.6537, "step": 60495 }, { "epoch": 4.74, "learning_rate": 5.2563318434878073e-05, "loss": 2.9333, "step": 60496 }, { "epoch": 4.74, "learning_rate": 5.256253430565358e-05, "loss": 2.5517, "step": 60497 }, { "epoch": 4.74, "learning_rate": 5.256175017642908e-05, "loss": 2.7588, "step": 60498 }, { "epoch": 4.74, "learning_rate": 5.2560966047204584e-05, "loss": 2.8245, "step": 60499 }, { "epoch": 4.74, "learning_rate": 5.256018191798009e-05, "loss": 2.7112, "step": 60500 }, { "epoch": 4.74, "learning_rate": 5.255939778875559e-05, "loss": 2.8054, "step": 60501 }, { "epoch": 4.74, "learning_rate": 5.2558613659531095e-05, "loss": 2.5427, "step": 60502 }, { "epoch": 4.74, "learning_rate": 5.25578295303066e-05, "loss": 2.5269, "step": 60503 }, { "epoch": 4.74, "learning_rate": 5.25570454010821e-05, "loss": 2.8144, "step": 60504 }, { "epoch": 4.74, "learning_rate": 5.2556261271857607e-05, "loss": 2.6816, "step": 60505 }, { "epoch": 4.74, "learning_rate": 5.255547714263312e-05, "loss": 2.5325, "step": 60506 }, { "epoch": 4.74, "learning_rate": 5.255469301340861e-05, "loss": 2.6102, "step": 60507 }, { "epoch": 4.74, "learning_rate": 5.255390888418411e-05, "loss": 2.6775, "step": 60508 }, { "epoch": 4.74, "learning_rate": 5.2553124754959614e-05, "loss": 2.6397, "step": 60509 }, { "epoch": 4.74, "learning_rate": 5.255234062573512e-05, "loss": 2.84, "step": 60510 }, { "epoch": 4.74, "learning_rate": 5.255155649651062e-05, "loss": 2.6904, "step": 60511 }, { "epoch": 4.74, "learning_rate": 5.2550772367286125e-05, "loss": 2.6674, "step": 60512 }, { "epoch": 4.75, "learning_rate": 5.254998823806163e-05, "loss": 2.6611, "step": 60513 }, { "epoch": 4.75, "learning_rate": 5.254920410883713e-05, "loss": 2.8428, "step": 60514 }, { "epoch": 4.75, "learning_rate": 5.2548419979612637e-05, "loss": 2.6618, "step": 60515 }, { "epoch": 4.75, "learning_rate": 5.254763585038815e-05, "loss": 2.5816, "step": 60516 }, { "epoch": 4.75, "learning_rate": 5.254685172116365e-05, "loss": 2.5777, "step": 60517 }, { "epoch": 4.75, "learning_rate": 5.2546067591939154e-05, "loss": 2.7905, "step": 60518 }, { "epoch": 4.75, "learning_rate": 5.254528346271466e-05, "loss": 2.7504, "step": 60519 }, { "epoch": 4.75, "learning_rate": 5.254449933349016e-05, "loss": 2.5652, "step": 60520 }, { "epoch": 4.75, "learning_rate": 5.2543715204265665e-05, "loss": 2.5319, "step": 60521 }, { "epoch": 4.75, "learning_rate": 5.254293107504117e-05, "loss": 2.6764, "step": 60522 }, { "epoch": 4.75, "learning_rate": 5.254214694581667e-05, "loss": 2.564, "step": 60523 }, { "epoch": 4.75, "learning_rate": 5.2541362816592176e-05, "loss": 2.7958, "step": 60524 }, { "epoch": 4.75, "learning_rate": 5.254057868736768e-05, "loss": 2.7289, "step": 60525 }, { "epoch": 4.75, "learning_rate": 5.2539794558143184e-05, "loss": 2.7308, "step": 60526 }, { "epoch": 4.75, "learning_rate": 5.253901042891869e-05, "loss": 2.7543, "step": 60527 }, { "epoch": 4.75, "learning_rate": 5.253822629969419e-05, "loss": 2.7196, "step": 60528 }, { "epoch": 4.75, "learning_rate": 5.2537442170469695e-05, "loss": 2.5509, "step": 60529 }, { "epoch": 4.75, "learning_rate": 5.25366580412452e-05, "loss": 2.5549, "step": 60530 }, { "epoch": 4.75, "learning_rate": 5.25358739120207e-05, "loss": 2.8135, "step": 60531 }, { "epoch": 4.75, "learning_rate": 5.253508978279621e-05, "loss": 2.8111, "step": 60532 }, { "epoch": 4.75, "learning_rate": 5.2534305653571716e-05, "loss": 2.6654, "step": 60533 }, { "epoch": 4.75, "learning_rate": 5.253352152434722e-05, "loss": 2.6442, "step": 60534 }, { "epoch": 4.75, "learning_rate": 5.2532737395122724e-05, "loss": 2.4682, "step": 60535 }, { "epoch": 4.75, "learning_rate": 5.253195326589823e-05, "loss": 2.8144, "step": 60536 }, { "epoch": 4.75, "learning_rate": 5.253116913667373e-05, "loss": 3.0066, "step": 60537 }, { "epoch": 4.75, "learning_rate": 5.2530385007449235e-05, "loss": 2.5575, "step": 60538 }, { "epoch": 4.75, "learning_rate": 5.252960087822474e-05, "loss": 2.6113, "step": 60539 }, { "epoch": 4.75, "learning_rate": 5.252881674900024e-05, "loss": 2.6856, "step": 60540 }, { "epoch": 4.75, "learning_rate": 5.2528032619775746e-05, "loss": 2.5623, "step": 60541 }, { "epoch": 4.75, "learning_rate": 5.252724849055125e-05, "loss": 2.5356, "step": 60542 }, { "epoch": 4.75, "learning_rate": 5.2526464361326746e-05, "loss": 2.5121, "step": 60543 }, { "epoch": 4.75, "learning_rate": 5.252568023210225e-05, "loss": 2.616, "step": 60544 }, { "epoch": 4.75, "learning_rate": 5.2524896102877754e-05, "loss": 2.7247, "step": 60545 }, { "epoch": 4.75, "learning_rate": 5.252411197365326e-05, "loss": 2.6393, "step": 60546 }, { "epoch": 4.75, "learning_rate": 5.252332784442876e-05, "loss": 2.808, "step": 60547 }, { "epoch": 4.75, "learning_rate": 5.2522543715204265e-05, "loss": 2.6749, "step": 60548 }, { "epoch": 4.75, "learning_rate": 5.252175958597977e-05, "loss": 2.603, "step": 60549 }, { "epoch": 4.75, "learning_rate": 5.252097545675527e-05, "loss": 2.904, "step": 60550 }, { "epoch": 4.75, "learning_rate": 5.2520191327530776e-05, "loss": 2.5588, "step": 60551 }, { "epoch": 4.75, "learning_rate": 5.251940719830628e-05, "loss": 2.8871, "step": 60552 }, { "epoch": 4.75, "learning_rate": 5.251862306908178e-05, "loss": 2.6231, "step": 60553 }, { "epoch": 4.75, "learning_rate": 5.251783893985729e-05, "loss": 3.0425, "step": 60554 }, { "epoch": 4.75, "learning_rate": 5.251705481063279e-05, "loss": 2.5906, "step": 60555 }, { "epoch": 4.75, "learning_rate": 5.2516270681408294e-05, "loss": 2.5768, "step": 60556 }, { "epoch": 4.75, "learning_rate": 5.2515486552183805e-05, "loss": 2.4848, "step": 60557 }, { "epoch": 4.75, "learning_rate": 5.251470242295931e-05, "loss": 2.5152, "step": 60558 }, { "epoch": 4.75, "learning_rate": 5.251391829373481e-05, "loss": 2.8031, "step": 60559 }, { "epoch": 4.75, "learning_rate": 5.2513134164510316e-05, "loss": 2.7449, "step": 60560 }, { "epoch": 4.75, "learning_rate": 5.251235003528582e-05, "loss": 2.8918, "step": 60561 }, { "epoch": 4.75, "learning_rate": 5.251156590606132e-05, "loss": 2.8472, "step": 60562 }, { "epoch": 4.75, "learning_rate": 5.2510781776836827e-05, "loss": 2.5759, "step": 60563 }, { "epoch": 4.75, "learning_rate": 5.250999764761233e-05, "loss": 2.6809, "step": 60564 }, { "epoch": 4.75, "learning_rate": 5.2509213518387834e-05, "loss": 2.5332, "step": 60565 }, { "epoch": 4.75, "learning_rate": 5.250842938916334e-05, "loss": 2.7691, "step": 60566 }, { "epoch": 4.75, "learning_rate": 5.250764525993884e-05, "loss": 2.6846, "step": 60567 }, { "epoch": 4.75, "learning_rate": 5.2506861130714345e-05, "loss": 2.6307, "step": 60568 }, { "epoch": 4.75, "learning_rate": 5.250607700148985e-05, "loss": 2.6744, "step": 60569 }, { "epoch": 4.75, "learning_rate": 5.250529287226535e-05, "loss": 2.8275, "step": 60570 }, { "epoch": 4.75, "learning_rate": 5.2504508743040856e-05, "loss": 2.5181, "step": 60571 }, { "epoch": 4.75, "learning_rate": 5.250372461381636e-05, "loss": 2.6467, "step": 60572 }, { "epoch": 4.75, "learning_rate": 5.250294048459187e-05, "loss": 2.5808, "step": 60573 }, { "epoch": 4.75, "learning_rate": 5.2502156355367374e-05, "loss": 2.6441, "step": 60574 }, { "epoch": 4.75, "learning_rate": 5.250137222614288e-05, "loss": 2.4867, "step": 60575 }, { "epoch": 4.75, "learning_rate": 5.250058809691838e-05, "loss": 2.5839, "step": 60576 }, { "epoch": 4.75, "learning_rate": 5.2499803967693885e-05, "loss": 2.7474, "step": 60577 }, { "epoch": 4.75, "learning_rate": 5.249901983846939e-05, "loss": 2.6647, "step": 60578 }, { "epoch": 4.75, "learning_rate": 5.249823570924488e-05, "loss": 2.7035, "step": 60579 }, { "epoch": 4.75, "learning_rate": 5.249745158002038e-05, "loss": 2.5185, "step": 60580 }, { "epoch": 4.75, "learning_rate": 5.2496667450795886e-05, "loss": 2.6777, "step": 60581 }, { "epoch": 4.75, "learning_rate": 5.249588332157139e-05, "loss": 2.565, "step": 60582 }, { "epoch": 4.75, "learning_rate": 5.24950991923469e-05, "loss": 2.7932, "step": 60583 }, { "epoch": 4.75, "learning_rate": 5.2494315063122404e-05, "loss": 2.7569, "step": 60584 }, { "epoch": 4.75, "learning_rate": 5.249353093389791e-05, "loss": 2.4926, "step": 60585 }, { "epoch": 4.75, "learning_rate": 5.249274680467341e-05, "loss": 2.5853, "step": 60586 }, { "epoch": 4.75, "learning_rate": 5.2491962675448915e-05, "loss": 2.6214, "step": 60587 }, { "epoch": 4.75, "learning_rate": 5.249117854622442e-05, "loss": 2.851, "step": 60588 }, { "epoch": 4.75, "learning_rate": 5.249039441699992e-05, "loss": 2.9629, "step": 60589 }, { "epoch": 4.75, "learning_rate": 5.2489610287775426e-05, "loss": 2.6191, "step": 60590 }, { "epoch": 4.75, "learning_rate": 5.248882615855093e-05, "loss": 2.8653, "step": 60591 }, { "epoch": 4.75, "learning_rate": 5.248804202932643e-05, "loss": 2.7213, "step": 60592 }, { "epoch": 4.75, "learning_rate": 5.248725790010194e-05, "loss": 2.7106, "step": 60593 }, { "epoch": 4.75, "learning_rate": 5.248647377087744e-05, "loss": 2.5763, "step": 60594 }, { "epoch": 4.75, "learning_rate": 5.2485689641652944e-05, "loss": 2.6954, "step": 60595 }, { "epoch": 4.75, "learning_rate": 5.248490551242845e-05, "loss": 2.7721, "step": 60596 }, { "epoch": 4.75, "learning_rate": 5.248412138320395e-05, "loss": 2.7919, "step": 60597 }, { "epoch": 4.75, "learning_rate": 5.248333725397946e-05, "loss": 2.7737, "step": 60598 }, { "epoch": 4.75, "learning_rate": 5.2482553124754966e-05, "loss": 2.6085, "step": 60599 }, { "epoch": 4.75, "learning_rate": 5.248176899553047e-05, "loss": 2.6808, "step": 60600 }, { "epoch": 4.75, "learning_rate": 5.248098486630597e-05, "loss": 2.7415, "step": 60601 }, { "epoch": 4.75, "learning_rate": 5.248020073708148e-05, "loss": 2.68, "step": 60602 }, { "epoch": 4.75, "learning_rate": 5.247941660785698e-05, "loss": 2.6252, "step": 60603 }, { "epoch": 4.75, "learning_rate": 5.2478632478632484e-05, "loss": 2.6545, "step": 60604 }, { "epoch": 4.75, "learning_rate": 5.247784834940799e-05, "loss": 2.4905, "step": 60605 }, { "epoch": 4.75, "learning_rate": 5.247706422018349e-05, "loss": 2.7281, "step": 60606 }, { "epoch": 4.75, "learning_rate": 5.2476280090958995e-05, "loss": 2.5679, "step": 60607 }, { "epoch": 4.75, "learning_rate": 5.24754959617345e-05, "loss": 2.4341, "step": 60608 }, { "epoch": 4.75, "learning_rate": 5.247471183251e-05, "loss": 2.686, "step": 60609 }, { "epoch": 4.75, "learning_rate": 5.2473927703285506e-05, "loss": 2.846, "step": 60610 }, { "epoch": 4.75, "learning_rate": 5.247314357406101e-05, "loss": 2.6704, "step": 60611 }, { "epoch": 4.75, "learning_rate": 5.2472359444836514e-05, "loss": 2.6901, "step": 60612 }, { "epoch": 4.75, "learning_rate": 5.247157531561202e-05, "loss": 2.5385, "step": 60613 }, { "epoch": 4.75, "learning_rate": 5.247079118638753e-05, "loss": 2.702, "step": 60614 }, { "epoch": 4.75, "learning_rate": 5.247000705716302e-05, "loss": 2.516, "step": 60615 }, { "epoch": 4.75, "learning_rate": 5.246922292793852e-05, "loss": 2.637, "step": 60616 }, { "epoch": 4.75, "learning_rate": 5.2468438798714025e-05, "loss": 2.6836, "step": 60617 }, { "epoch": 4.75, "learning_rate": 5.246765466948953e-05, "loss": 2.5807, "step": 60618 }, { "epoch": 4.75, "learning_rate": 5.246687054026503e-05, "loss": 2.7364, "step": 60619 }, { "epoch": 4.75, "learning_rate": 5.2466086411040536e-05, "loss": 2.8301, "step": 60620 }, { "epoch": 4.75, "learning_rate": 5.246530228181604e-05, "loss": 2.8084, "step": 60621 }, { "epoch": 4.75, "learning_rate": 5.2464518152591544e-05, "loss": 2.8685, "step": 60622 }, { "epoch": 4.75, "learning_rate": 5.246373402336705e-05, "loss": 2.7803, "step": 60623 }, { "epoch": 4.75, "learning_rate": 5.246294989414256e-05, "loss": 2.6914, "step": 60624 }, { "epoch": 4.75, "learning_rate": 5.246216576491806e-05, "loss": 2.6011, "step": 60625 }, { "epoch": 4.75, "learning_rate": 5.2461381635693565e-05, "loss": 2.4202, "step": 60626 }, { "epoch": 4.75, "learning_rate": 5.246059750646907e-05, "loss": 2.5484, "step": 60627 }, { "epoch": 4.75, "learning_rate": 5.245981337724457e-05, "loss": 2.7774, "step": 60628 }, { "epoch": 4.75, "learning_rate": 5.2459029248020076e-05, "loss": 2.653, "step": 60629 }, { "epoch": 4.75, "learning_rate": 5.245824511879558e-05, "loss": 2.7158, "step": 60630 }, { "epoch": 4.75, "learning_rate": 5.2457460989571083e-05, "loss": 2.7565, "step": 60631 }, { "epoch": 4.75, "learning_rate": 5.245667686034659e-05, "loss": 2.8321, "step": 60632 }, { "epoch": 4.75, "learning_rate": 5.245589273112209e-05, "loss": 2.827, "step": 60633 }, { "epoch": 4.75, "learning_rate": 5.2455108601897595e-05, "loss": 2.8085, "step": 60634 }, { "epoch": 4.75, "learning_rate": 5.24543244726731e-05, "loss": 2.6384, "step": 60635 }, { "epoch": 4.75, "learning_rate": 5.24535403434486e-05, "loss": 2.6965, "step": 60636 }, { "epoch": 4.75, "learning_rate": 5.2452756214224106e-05, "loss": 2.7203, "step": 60637 }, { "epoch": 4.75, "learning_rate": 5.245197208499961e-05, "loss": 2.8489, "step": 60638 }, { "epoch": 4.75, "learning_rate": 5.245118795577511e-05, "loss": 2.4993, "step": 60639 }, { "epoch": 4.75, "learning_rate": 5.245040382655062e-05, "loss": 2.6379, "step": 60640 }, { "epoch": 4.76, "learning_rate": 5.244961969732613e-05, "loss": 2.7301, "step": 60641 }, { "epoch": 4.76, "learning_rate": 5.244883556810163e-05, "loss": 2.518, "step": 60642 }, { "epoch": 4.76, "learning_rate": 5.2448051438877134e-05, "loss": 2.6197, "step": 60643 }, { "epoch": 4.76, "learning_rate": 5.244726730965264e-05, "loss": 2.6122, "step": 60644 }, { "epoch": 4.76, "learning_rate": 5.244648318042814e-05, "loss": 2.702, "step": 60645 }, { "epoch": 4.76, "learning_rate": 5.2445699051203645e-05, "loss": 2.6095, "step": 60646 }, { "epoch": 4.76, "learning_rate": 5.244491492197915e-05, "loss": 2.7394, "step": 60647 }, { "epoch": 4.76, "learning_rate": 5.244413079275465e-05, "loss": 2.6792, "step": 60648 }, { "epoch": 4.76, "learning_rate": 5.2443346663530156e-05, "loss": 2.6581, "step": 60649 }, { "epoch": 4.76, "learning_rate": 5.244256253430566e-05, "loss": 2.7171, "step": 60650 }, { "epoch": 4.76, "learning_rate": 5.244177840508116e-05, "loss": 2.5967, "step": 60651 }, { "epoch": 4.76, "learning_rate": 5.244099427585666e-05, "loss": 2.7178, "step": 60652 }, { "epoch": 4.76, "learning_rate": 5.2440210146632164e-05, "loss": 2.544, "step": 60653 }, { "epoch": 4.76, "learning_rate": 5.243942601740767e-05, "loss": 2.6769, "step": 60654 }, { "epoch": 4.76, "learning_rate": 5.243864188818317e-05, "loss": 2.8226, "step": 60655 }, { "epoch": 4.76, "learning_rate": 5.2437857758958675e-05, "loss": 2.9226, "step": 60656 }, { "epoch": 4.76, "learning_rate": 5.243707362973418e-05, "loss": 2.8471, "step": 60657 }, { "epoch": 4.76, "learning_rate": 5.243628950050968e-05, "loss": 2.8086, "step": 60658 }, { "epoch": 4.76, "learning_rate": 5.2435505371285186e-05, "loss": 2.5127, "step": 60659 }, { "epoch": 4.76, "learning_rate": 5.243472124206069e-05, "loss": 2.6624, "step": 60660 }, { "epoch": 4.76, "learning_rate": 5.2433937112836194e-05, "loss": 2.7429, "step": 60661 }, { "epoch": 4.76, "learning_rate": 5.24331529836117e-05, "loss": 2.6532, "step": 60662 }, { "epoch": 4.76, "learning_rate": 5.24323688543872e-05, "loss": 2.5516, "step": 60663 }, { "epoch": 4.76, "learning_rate": 5.2431584725162705e-05, "loss": 2.8838, "step": 60664 }, { "epoch": 4.76, "learning_rate": 5.2430800595938215e-05, "loss": 2.6687, "step": 60665 }, { "epoch": 4.76, "learning_rate": 5.243001646671372e-05, "loss": 2.5045, "step": 60666 }, { "epoch": 4.76, "learning_rate": 5.242923233748922e-05, "loss": 2.6763, "step": 60667 }, { "epoch": 4.76, "learning_rate": 5.2428448208264726e-05, "loss": 2.8194, "step": 60668 }, { "epoch": 4.76, "learning_rate": 5.242766407904023e-05, "loss": 2.6871, "step": 60669 }, { "epoch": 4.76, "learning_rate": 5.2426879949815734e-05, "loss": 2.4683, "step": 60670 }, { "epoch": 4.76, "learning_rate": 5.242609582059124e-05, "loss": 2.8532, "step": 60671 }, { "epoch": 4.76, "learning_rate": 5.242531169136674e-05, "loss": 2.7392, "step": 60672 }, { "epoch": 4.76, "learning_rate": 5.2424527562142245e-05, "loss": 2.7776, "step": 60673 }, { "epoch": 4.76, "learning_rate": 5.242374343291775e-05, "loss": 2.6507, "step": 60674 }, { "epoch": 4.76, "learning_rate": 5.242295930369325e-05, "loss": 2.5339, "step": 60675 }, { "epoch": 4.76, "learning_rate": 5.2422175174468756e-05, "loss": 2.6591, "step": 60676 }, { "epoch": 4.76, "learning_rate": 5.242139104524426e-05, "loss": 2.6669, "step": 60677 }, { "epoch": 4.76, "learning_rate": 5.242060691601976e-05, "loss": 2.7328, "step": 60678 }, { "epoch": 4.76, "learning_rate": 5.241982278679527e-05, "loss": 2.8526, "step": 60679 }, { "epoch": 4.76, "learning_rate": 5.241903865757077e-05, "loss": 2.6266, "step": 60680 }, { "epoch": 4.76, "learning_rate": 5.241825452834628e-05, "loss": 2.7122, "step": 60681 }, { "epoch": 4.76, "learning_rate": 5.2417470399121785e-05, "loss": 2.7306, "step": 60682 }, { "epoch": 4.76, "learning_rate": 5.241668626989729e-05, "loss": 2.7821, "step": 60683 }, { "epoch": 4.76, "learning_rate": 5.241590214067279e-05, "loss": 2.4783, "step": 60684 }, { "epoch": 4.76, "learning_rate": 5.2415118011448296e-05, "loss": 2.7165, "step": 60685 }, { "epoch": 4.76, "learning_rate": 5.24143338822238e-05, "loss": 2.7079, "step": 60686 }, { "epoch": 4.76, "learning_rate": 5.241354975299929e-05, "loss": 2.7939, "step": 60687 }, { "epoch": 4.76, "learning_rate": 5.241276562377479e-05, "loss": 2.6715, "step": 60688 }, { "epoch": 4.76, "learning_rate": 5.24119814945503e-05, "loss": 2.6227, "step": 60689 }, { "epoch": 4.76, "learning_rate": 5.24111973653258e-05, "loss": 2.8193, "step": 60690 }, { "epoch": 4.76, "learning_rate": 5.241041323610131e-05, "loss": 2.7178, "step": 60691 }, { "epoch": 4.76, "learning_rate": 5.2409629106876815e-05, "loss": 2.5559, "step": 60692 }, { "epoch": 4.76, "learning_rate": 5.240884497765232e-05, "loss": 2.6788, "step": 60693 }, { "epoch": 4.76, "learning_rate": 5.240806084842782e-05, "loss": 2.5424, "step": 60694 }, { "epoch": 4.76, "learning_rate": 5.2407276719203326e-05, "loss": 2.4987, "step": 60695 }, { "epoch": 4.76, "learning_rate": 5.240649258997883e-05, "loss": 2.6359, "step": 60696 }, { "epoch": 4.76, "learning_rate": 5.240570846075433e-05, "loss": 2.9508, "step": 60697 }, { "epoch": 4.76, "learning_rate": 5.240492433152984e-05, "loss": 2.4944, "step": 60698 }, { "epoch": 4.76, "learning_rate": 5.240414020230534e-05, "loss": 2.8374, "step": 60699 }, { "epoch": 4.76, "learning_rate": 5.2403356073080844e-05, "loss": 2.7806, "step": 60700 }, { "epoch": 4.76, "learning_rate": 5.240257194385635e-05, "loss": 2.6499, "step": 60701 }, { "epoch": 4.76, "learning_rate": 5.240178781463185e-05, "loss": 2.6506, "step": 60702 }, { "epoch": 4.76, "learning_rate": 5.2401003685407355e-05, "loss": 2.4662, "step": 60703 }, { "epoch": 4.76, "learning_rate": 5.240021955618286e-05, "loss": 2.517, "step": 60704 }, { "epoch": 4.76, "learning_rate": 5.239943542695836e-05, "loss": 2.6908, "step": 60705 }, { "epoch": 4.76, "learning_rate": 5.239865129773387e-05, "loss": 2.8468, "step": 60706 }, { "epoch": 4.76, "learning_rate": 5.2397867168509377e-05, "loss": 2.5724, "step": 60707 }, { "epoch": 4.76, "learning_rate": 5.239708303928488e-05, "loss": 2.7444, "step": 60708 }, { "epoch": 4.76, "learning_rate": 5.2396298910060384e-05, "loss": 2.6992, "step": 60709 }, { "epoch": 4.76, "learning_rate": 5.239551478083589e-05, "loss": 2.4913, "step": 60710 }, { "epoch": 4.76, "learning_rate": 5.239473065161139e-05, "loss": 2.5849, "step": 60711 }, { "epoch": 4.76, "learning_rate": 5.2393946522386895e-05, "loss": 2.6155, "step": 60712 }, { "epoch": 4.76, "learning_rate": 5.23931623931624e-05, "loss": 2.8871, "step": 60713 }, { "epoch": 4.76, "learning_rate": 5.23923782639379e-05, "loss": 2.8024, "step": 60714 }, { "epoch": 4.76, "learning_rate": 5.2391594134713406e-05, "loss": 2.6229, "step": 60715 }, { "epoch": 4.76, "learning_rate": 5.239081000548891e-05, "loss": 2.7288, "step": 60716 }, { "epoch": 4.76, "learning_rate": 5.239002587626441e-05, "loss": 2.4719, "step": 60717 }, { "epoch": 4.76, "learning_rate": 5.238924174703992e-05, "loss": 2.6546, "step": 60718 }, { "epoch": 4.76, "learning_rate": 5.238845761781542e-05, "loss": 2.583, "step": 60719 }, { "epoch": 4.76, "learning_rate": 5.2387673488590924e-05, "loss": 2.8702, "step": 60720 }, { "epoch": 4.76, "learning_rate": 5.238688935936643e-05, "loss": 2.5332, "step": 60721 }, { "epoch": 4.76, "learning_rate": 5.238610523014194e-05, "loss": 2.4906, "step": 60722 }, { "epoch": 4.76, "learning_rate": 5.238532110091743e-05, "loss": 2.5906, "step": 60723 }, { "epoch": 4.76, "learning_rate": 5.238453697169293e-05, "loss": 2.85, "step": 60724 }, { "epoch": 4.76, "learning_rate": 5.2383752842468436e-05, "loss": 2.5245, "step": 60725 }, { "epoch": 4.76, "learning_rate": 5.238296871324394e-05, "loss": 2.6314, "step": 60726 }, { "epoch": 4.76, "learning_rate": 5.238218458401944e-05, "loss": 2.5362, "step": 60727 }, { "epoch": 4.76, "learning_rate": 5.238140045479495e-05, "loss": 2.4657, "step": 60728 }, { "epoch": 4.76, "learning_rate": 5.238061632557045e-05, "loss": 2.5827, "step": 60729 }, { "epoch": 4.76, "learning_rate": 5.2379832196345954e-05, "loss": 2.6527, "step": 60730 }, { "epoch": 4.76, "learning_rate": 5.237904806712146e-05, "loss": 2.6371, "step": 60731 }, { "epoch": 4.76, "learning_rate": 5.237826393789697e-05, "loss": 2.8602, "step": 60732 }, { "epoch": 4.76, "learning_rate": 5.237747980867247e-05, "loss": 2.7811, "step": 60733 }, { "epoch": 4.76, "learning_rate": 5.2376695679447976e-05, "loss": 2.5421, "step": 60734 }, { "epoch": 4.76, "learning_rate": 5.237591155022348e-05, "loss": 2.6868, "step": 60735 }, { "epoch": 4.76, "learning_rate": 5.237512742099898e-05, "loss": 2.5006, "step": 60736 }, { "epoch": 4.76, "learning_rate": 5.237434329177449e-05, "loss": 2.5385, "step": 60737 }, { "epoch": 4.76, "learning_rate": 5.237355916254999e-05, "loss": 2.5865, "step": 60738 }, { "epoch": 4.76, "learning_rate": 5.2372775033325494e-05, "loss": 2.5949, "step": 60739 }, { "epoch": 4.76, "learning_rate": 5.2371990904101e-05, "loss": 2.6368, "step": 60740 }, { "epoch": 4.76, "learning_rate": 5.23712067748765e-05, "loss": 2.5548, "step": 60741 }, { "epoch": 4.76, "learning_rate": 5.2370422645652005e-05, "loss": 2.7069, "step": 60742 }, { "epoch": 4.76, "learning_rate": 5.236963851642751e-05, "loss": 2.7538, "step": 60743 }, { "epoch": 4.76, "learning_rate": 5.236885438720301e-05, "loss": 2.5703, "step": 60744 }, { "epoch": 4.76, "learning_rate": 5.2368070257978516e-05, "loss": 2.6958, "step": 60745 }, { "epoch": 4.76, "learning_rate": 5.236728612875402e-05, "loss": 2.5779, "step": 60746 }, { "epoch": 4.76, "learning_rate": 5.2366501999529524e-05, "loss": 2.7309, "step": 60747 }, { "epoch": 4.76, "learning_rate": 5.2365717870305034e-05, "loss": 2.6235, "step": 60748 }, { "epoch": 4.76, "learning_rate": 5.236493374108054e-05, "loss": 2.7109, "step": 60749 }, { "epoch": 4.76, "learning_rate": 5.236414961185604e-05, "loss": 2.7943, "step": 60750 }, { "epoch": 4.76, "learning_rate": 5.2363365482631545e-05, "loss": 2.7099, "step": 60751 }, { "epoch": 4.76, "learning_rate": 5.236258135340705e-05, "loss": 2.5437, "step": 60752 }, { "epoch": 4.76, "learning_rate": 5.236179722418255e-05, "loss": 2.872, "step": 60753 }, { "epoch": 4.76, "learning_rate": 5.2361013094958056e-05, "loss": 2.5693, "step": 60754 }, { "epoch": 4.76, "learning_rate": 5.236022896573356e-05, "loss": 2.7726, "step": 60755 }, { "epoch": 4.76, "learning_rate": 5.2359444836509064e-05, "loss": 2.7032, "step": 60756 }, { "epoch": 4.76, "learning_rate": 5.235866070728457e-05, "loss": 2.6635, "step": 60757 }, { "epoch": 4.76, "learning_rate": 5.235787657806007e-05, "loss": 2.4207, "step": 60758 }, { "epoch": 4.76, "learning_rate": 5.235709244883557e-05, "loss": 2.655, "step": 60759 }, { "epoch": 4.76, "learning_rate": 5.235630831961107e-05, "loss": 2.6593, "step": 60760 }, { "epoch": 4.76, "learning_rate": 5.2355524190386575e-05, "loss": 2.5385, "step": 60761 }, { "epoch": 4.76, "learning_rate": 5.235474006116208e-05, "loss": 2.8254, "step": 60762 }, { "epoch": 4.76, "learning_rate": 5.235395593193758e-05, "loss": 2.4793, "step": 60763 }, { "epoch": 4.76, "learning_rate": 5.2353171802713086e-05, "loss": 2.5756, "step": 60764 }, { "epoch": 4.76, "learning_rate": 5.235238767348859e-05, "loss": 2.5708, "step": 60765 }, { "epoch": 4.76, "learning_rate": 5.2351603544264094e-05, "loss": 2.6742, "step": 60766 }, { "epoch": 4.76, "learning_rate": 5.23508194150396e-05, "loss": 2.536, "step": 60767 }, { "epoch": 4.76, "learning_rate": 5.23500352858151e-05, "loss": 2.8677, "step": 60768 }, { "epoch": 4.77, "learning_rate": 5.2349251156590605e-05, "loss": 2.5869, "step": 60769 }, { "epoch": 4.77, "learning_rate": 5.234846702736611e-05, "loss": 2.5165, "step": 60770 }, { "epoch": 4.77, "learning_rate": 5.234768289814161e-05, "loss": 2.845, "step": 60771 }, { "epoch": 4.77, "learning_rate": 5.2346898768917116e-05, "loss": 2.5727, "step": 60772 }, { "epoch": 4.77, "learning_rate": 5.2346114639692626e-05, "loss": 2.5587, "step": 60773 }, { "epoch": 4.77, "learning_rate": 5.234533051046813e-05, "loss": 2.6951, "step": 60774 }, { "epoch": 4.77, "learning_rate": 5.234454638124363e-05, "loss": 2.6095, "step": 60775 }, { "epoch": 4.77, "learning_rate": 5.234376225201914e-05, "loss": 2.5778, "step": 60776 }, { "epoch": 4.77, "learning_rate": 5.234297812279464e-05, "loss": 2.6183, "step": 60777 }, { "epoch": 4.77, "learning_rate": 5.2342193993570144e-05, "loss": 2.6245, "step": 60778 }, { "epoch": 4.77, "learning_rate": 5.234140986434565e-05, "loss": 2.6142, "step": 60779 }, { "epoch": 4.77, "learning_rate": 5.234062573512115e-05, "loss": 2.8856, "step": 60780 }, { "epoch": 4.77, "learning_rate": 5.2339841605896655e-05, "loss": 2.7414, "step": 60781 }, { "epoch": 4.77, "learning_rate": 5.233905747667216e-05, "loss": 2.5232, "step": 60782 }, { "epoch": 4.77, "learning_rate": 5.233827334744766e-05, "loss": 2.6727, "step": 60783 }, { "epoch": 4.77, "learning_rate": 5.2337489218223166e-05, "loss": 2.5982, "step": 60784 }, { "epoch": 4.77, "learning_rate": 5.233670508899867e-05, "loss": 2.7721, "step": 60785 }, { "epoch": 4.77, "learning_rate": 5.2335920959774174e-05, "loss": 2.7406, "step": 60786 }, { "epoch": 4.77, "learning_rate": 5.233513683054968e-05, "loss": 2.733, "step": 60787 }, { "epoch": 4.77, "learning_rate": 5.233435270132518e-05, "loss": 2.5298, "step": 60788 }, { "epoch": 4.77, "learning_rate": 5.233356857210069e-05, "loss": 2.6856, "step": 60789 }, { "epoch": 4.77, "learning_rate": 5.2332784442876195e-05, "loss": 2.6763, "step": 60790 }, { "epoch": 4.77, "learning_rate": 5.23320003136517e-05, "loss": 2.5872, "step": 60791 }, { "epoch": 4.77, "learning_rate": 5.23312161844272e-05, "loss": 2.7539, "step": 60792 }, { "epoch": 4.77, "learning_rate": 5.2330432055202706e-05, "loss": 2.7156, "step": 60793 }, { "epoch": 4.77, "learning_rate": 5.232964792597821e-05, "loss": 2.5243, "step": 60794 }, { "epoch": 4.77, "learning_rate": 5.23288637967537e-05, "loss": 2.7576, "step": 60795 }, { "epoch": 4.77, "learning_rate": 5.2328079667529204e-05, "loss": 2.5908, "step": 60796 }, { "epoch": 4.77, "learning_rate": 5.232729553830471e-05, "loss": 2.6665, "step": 60797 }, { "epoch": 4.77, "learning_rate": 5.232651140908021e-05, "loss": 2.5966, "step": 60798 }, { "epoch": 4.77, "learning_rate": 5.232572727985572e-05, "loss": 2.6849, "step": 60799 }, { "epoch": 4.77, "learning_rate": 5.2324943150631225e-05, "loss": 2.8766, "step": 60800 }, { "epoch": 4.77, "learning_rate": 5.232415902140673e-05, "loss": 2.7096, "step": 60801 }, { "epoch": 4.77, "learning_rate": 5.232337489218223e-05, "loss": 2.6584, "step": 60802 }, { "epoch": 4.77, "learning_rate": 5.2322590762957736e-05, "loss": 3.1437, "step": 60803 }, { "epoch": 4.77, "learning_rate": 5.232180663373324e-05, "loss": 2.5334, "step": 60804 }, { "epoch": 4.77, "learning_rate": 5.2321022504508744e-05, "loss": 2.7697, "step": 60805 }, { "epoch": 4.77, "learning_rate": 5.232023837528425e-05, "loss": 2.6583, "step": 60806 }, { "epoch": 4.77, "learning_rate": 5.231945424605975e-05, "loss": 2.9552, "step": 60807 }, { "epoch": 4.77, "learning_rate": 5.2318670116835255e-05, "loss": 2.7599, "step": 60808 }, { "epoch": 4.77, "learning_rate": 5.231788598761076e-05, "loss": 2.6489, "step": 60809 }, { "epoch": 4.77, "learning_rate": 5.231710185838626e-05, "loss": 2.6497, "step": 60810 }, { "epoch": 4.77, "learning_rate": 5.2316317729161766e-05, "loss": 2.6725, "step": 60811 }, { "epoch": 4.77, "learning_rate": 5.231553359993727e-05, "loss": 2.8946, "step": 60812 }, { "epoch": 4.77, "learning_rate": 5.231474947071277e-05, "loss": 2.7, "step": 60813 }, { "epoch": 4.77, "learning_rate": 5.2313965341488284e-05, "loss": 2.869, "step": 60814 }, { "epoch": 4.77, "learning_rate": 5.231318121226379e-05, "loss": 2.4543, "step": 60815 }, { "epoch": 4.77, "learning_rate": 5.231239708303929e-05, "loss": 2.7248, "step": 60816 }, { "epoch": 4.77, "learning_rate": 5.2311612953814795e-05, "loss": 2.59, "step": 60817 }, { "epoch": 4.77, "learning_rate": 5.23108288245903e-05, "loss": 2.6858, "step": 60818 }, { "epoch": 4.77, "learning_rate": 5.23100446953658e-05, "loss": 2.7161, "step": 60819 }, { "epoch": 4.77, "learning_rate": 5.2309260566141306e-05, "loss": 2.717, "step": 60820 }, { "epoch": 4.77, "learning_rate": 5.230847643691681e-05, "loss": 2.6591, "step": 60821 }, { "epoch": 4.77, "learning_rate": 5.230769230769231e-05, "loss": 2.8339, "step": 60822 }, { "epoch": 4.77, "learning_rate": 5.230690817846782e-05, "loss": 2.6352, "step": 60823 }, { "epoch": 4.77, "learning_rate": 5.230612404924332e-05, "loss": 2.5973, "step": 60824 }, { "epoch": 4.77, "learning_rate": 5.2305339920018824e-05, "loss": 2.6661, "step": 60825 }, { "epoch": 4.77, "learning_rate": 5.230455579079433e-05, "loss": 2.7251, "step": 60826 }, { "epoch": 4.77, "learning_rate": 5.230377166156983e-05, "loss": 2.5814, "step": 60827 }, { "epoch": 4.77, "learning_rate": 5.2302987532345335e-05, "loss": 2.6782, "step": 60828 }, { "epoch": 4.77, "learning_rate": 5.230220340312084e-05, "loss": 2.6757, "step": 60829 }, { "epoch": 4.77, "learning_rate": 5.230141927389635e-05, "loss": 2.7023, "step": 60830 }, { "epoch": 4.77, "learning_rate": 5.230063514467184e-05, "loss": 2.6238, "step": 60831 }, { "epoch": 4.77, "learning_rate": 5.229985101544734e-05, "loss": 2.4617, "step": 60832 }, { "epoch": 4.77, "learning_rate": 5.229906688622285e-05, "loss": 2.5847, "step": 60833 }, { "epoch": 4.77, "learning_rate": 5.229828275699835e-05, "loss": 2.7579, "step": 60834 }, { "epoch": 4.77, "learning_rate": 5.2297498627773854e-05, "loss": 2.904, "step": 60835 }, { "epoch": 4.77, "learning_rate": 5.229671449854936e-05, "loss": 2.7373, "step": 60836 }, { "epoch": 4.77, "learning_rate": 5.229593036932486e-05, "loss": 2.6299, "step": 60837 }, { "epoch": 4.77, "learning_rate": 5.2295146240100365e-05, "loss": 2.4407, "step": 60838 }, { "epoch": 4.77, "learning_rate": 5.229436211087587e-05, "loss": 2.5466, "step": 60839 }, { "epoch": 4.77, "learning_rate": 5.229357798165138e-05, "loss": 2.6926, "step": 60840 }, { "epoch": 4.77, "learning_rate": 5.229279385242688e-05, "loss": 2.7835, "step": 60841 }, { "epoch": 4.77, "learning_rate": 5.2292009723202387e-05, "loss": 2.6644, "step": 60842 }, { "epoch": 4.77, "learning_rate": 5.229122559397789e-05, "loss": 2.5761, "step": 60843 }, { "epoch": 4.77, "learning_rate": 5.2290441464753394e-05, "loss": 2.4514, "step": 60844 }, { "epoch": 4.77, "learning_rate": 5.22896573355289e-05, "loss": 2.8697, "step": 60845 }, { "epoch": 4.77, "learning_rate": 5.22888732063044e-05, "loss": 2.6171, "step": 60846 }, { "epoch": 4.77, "learning_rate": 5.2288089077079905e-05, "loss": 2.53, "step": 60847 }, { "epoch": 4.77, "learning_rate": 5.228730494785541e-05, "loss": 2.7494, "step": 60848 }, { "epoch": 4.77, "learning_rate": 5.228652081863091e-05, "loss": 2.778, "step": 60849 }, { "epoch": 4.77, "learning_rate": 5.2285736689406416e-05, "loss": 2.5685, "step": 60850 }, { "epoch": 4.77, "learning_rate": 5.228495256018192e-05, "loss": 2.6094, "step": 60851 }, { "epoch": 4.77, "learning_rate": 5.228416843095742e-05, "loss": 2.6447, "step": 60852 }, { "epoch": 4.77, "learning_rate": 5.228338430173293e-05, "loss": 2.7535, "step": 60853 }, { "epoch": 4.77, "learning_rate": 5.228260017250843e-05, "loss": 2.6573, "step": 60854 }, { "epoch": 4.77, "learning_rate": 5.2281816043283934e-05, "loss": 2.5214, "step": 60855 }, { "epoch": 4.77, "learning_rate": 5.2281031914059445e-05, "loss": 2.6219, "step": 60856 }, { "epoch": 4.77, "learning_rate": 5.228024778483495e-05, "loss": 2.7414, "step": 60857 }, { "epoch": 4.77, "learning_rate": 5.227946365561045e-05, "loss": 2.7185, "step": 60858 }, { "epoch": 4.77, "learning_rate": 5.2278679526385956e-05, "loss": 2.7426, "step": 60859 }, { "epoch": 4.77, "learning_rate": 5.227789539716146e-05, "loss": 2.6583, "step": 60860 }, { "epoch": 4.77, "learning_rate": 5.227711126793696e-05, "loss": 2.6546, "step": 60861 }, { "epoch": 4.77, "learning_rate": 5.227632713871247e-05, "loss": 2.6961, "step": 60862 }, { "epoch": 4.77, "learning_rate": 5.227554300948797e-05, "loss": 2.7326, "step": 60863 }, { "epoch": 4.77, "learning_rate": 5.2274758880263474e-05, "loss": 2.6535, "step": 60864 }, { "epoch": 4.77, "learning_rate": 5.227397475103898e-05, "loss": 2.6534, "step": 60865 }, { "epoch": 4.77, "learning_rate": 5.227319062181448e-05, "loss": 2.6129, "step": 60866 }, { "epoch": 4.77, "learning_rate": 5.227240649258998e-05, "loss": 2.543, "step": 60867 }, { "epoch": 4.77, "learning_rate": 5.227162236336548e-05, "loss": 2.8161, "step": 60868 }, { "epoch": 4.77, "learning_rate": 5.2270838234140986e-05, "loss": 2.6238, "step": 60869 }, { "epoch": 4.77, "learning_rate": 5.227005410491649e-05, "loss": 2.6498, "step": 60870 }, { "epoch": 4.77, "learning_rate": 5.226926997569199e-05, "loss": 2.8608, "step": 60871 }, { "epoch": 4.77, "learning_rate": 5.22684858464675e-05, "loss": 2.8205, "step": 60872 }, { "epoch": 4.77, "learning_rate": 5.2267701717243e-05, "loss": 2.8181, "step": 60873 }, { "epoch": 4.77, "learning_rate": 5.2266917588018504e-05, "loss": 2.5425, "step": 60874 }, { "epoch": 4.77, "learning_rate": 5.226613345879401e-05, "loss": 2.6665, "step": 60875 }, { "epoch": 4.77, "learning_rate": 5.226534932956951e-05, "loss": 2.7468, "step": 60876 }, { "epoch": 4.77, "learning_rate": 5.2264565200345015e-05, "loss": 2.6572, "step": 60877 }, { "epoch": 4.77, "learning_rate": 5.226378107112052e-05, "loss": 2.6438, "step": 60878 }, { "epoch": 4.77, "learning_rate": 5.226299694189602e-05, "loss": 2.7594, "step": 60879 }, { "epoch": 4.77, "learning_rate": 5.2262212812671526e-05, "loss": 2.6398, "step": 60880 }, { "epoch": 4.77, "learning_rate": 5.226142868344704e-05, "loss": 2.8148, "step": 60881 }, { "epoch": 4.77, "learning_rate": 5.226064455422254e-05, "loss": 2.685, "step": 60882 }, { "epoch": 4.77, "learning_rate": 5.2259860424998044e-05, "loss": 2.8081, "step": 60883 }, { "epoch": 4.77, "learning_rate": 5.225907629577355e-05, "loss": 2.7433, "step": 60884 }, { "epoch": 4.77, "learning_rate": 5.225829216654905e-05, "loss": 2.6414, "step": 60885 }, { "epoch": 4.77, "learning_rate": 5.2257508037324555e-05, "loss": 2.7354, "step": 60886 }, { "epoch": 4.77, "learning_rate": 5.225672390810006e-05, "loss": 2.7788, "step": 60887 }, { "epoch": 4.77, "learning_rate": 5.225593977887556e-05, "loss": 2.8797, "step": 60888 }, { "epoch": 4.77, "learning_rate": 5.2255155649651066e-05, "loss": 2.7376, "step": 60889 }, { "epoch": 4.77, "learning_rate": 5.225437152042657e-05, "loss": 2.5668, "step": 60890 }, { "epoch": 4.77, "learning_rate": 5.2253587391202074e-05, "loss": 2.6707, "step": 60891 }, { "epoch": 4.77, "learning_rate": 5.225280326197758e-05, "loss": 2.5102, "step": 60892 }, { "epoch": 4.77, "learning_rate": 5.225201913275308e-05, "loss": 2.6859, "step": 60893 }, { "epoch": 4.77, "learning_rate": 5.2251235003528585e-05, "loss": 2.8124, "step": 60894 }, { "epoch": 4.77, "learning_rate": 5.225045087430409e-05, "loss": 2.5039, "step": 60895 }, { "epoch": 4.78, "learning_rate": 5.224966674507959e-05, "loss": 2.8517, "step": 60896 }, { "epoch": 4.78, "learning_rate": 5.22488826158551e-05, "loss": 2.6607, "step": 60897 }, { "epoch": 4.78, "learning_rate": 5.2248098486630606e-05, "loss": 2.6603, "step": 60898 }, { "epoch": 4.78, "learning_rate": 5.224731435740611e-05, "loss": 2.5699, "step": 60899 }, { "epoch": 4.78, "learning_rate": 5.2246530228181613e-05, "loss": 2.501, "step": 60900 }, { "epoch": 4.78, "learning_rate": 5.224574609895712e-05, "loss": 2.6715, "step": 60901 }, { "epoch": 4.78, "learning_rate": 5.224496196973262e-05, "loss": 2.7577, "step": 60902 }, { "epoch": 4.78, "learning_rate": 5.224417784050811e-05, "loss": 2.587, "step": 60903 }, { "epoch": 4.78, "learning_rate": 5.2243393711283615e-05, "loss": 2.5968, "step": 60904 }, { "epoch": 4.78, "learning_rate": 5.224260958205912e-05, "loss": 2.5587, "step": 60905 }, { "epoch": 4.78, "learning_rate": 5.224182545283462e-05, "loss": 2.6973, "step": 60906 }, { "epoch": 4.78, "learning_rate": 5.224104132361013e-05, "loss": 2.6132, "step": 60907 }, { "epoch": 4.78, "learning_rate": 5.2240257194385636e-05, "loss": 2.5954, "step": 60908 }, { "epoch": 4.78, "learning_rate": 5.223947306516114e-05, "loss": 2.5313, "step": 60909 }, { "epoch": 4.78, "learning_rate": 5.2238688935936643e-05, "loss": 2.6674, "step": 60910 }, { "epoch": 4.78, "learning_rate": 5.223790480671215e-05, "loss": 2.6866, "step": 60911 }, { "epoch": 4.78, "learning_rate": 5.223712067748765e-05, "loss": 2.7341, "step": 60912 }, { "epoch": 4.78, "learning_rate": 5.2236336548263154e-05, "loss": 2.6499, "step": 60913 }, { "epoch": 4.78, "learning_rate": 5.223555241903866e-05, "loss": 2.6147, "step": 60914 }, { "epoch": 4.78, "learning_rate": 5.223476828981416e-05, "loss": 2.7512, "step": 60915 }, { "epoch": 4.78, "learning_rate": 5.2233984160589665e-05, "loss": 2.6146, "step": 60916 }, { "epoch": 4.78, "learning_rate": 5.223320003136517e-05, "loss": 2.5315, "step": 60917 }, { "epoch": 4.78, "learning_rate": 5.223241590214067e-05, "loss": 2.8391, "step": 60918 }, { "epoch": 4.78, "learning_rate": 5.2231631772916177e-05, "loss": 2.5544, "step": 60919 }, { "epoch": 4.78, "learning_rate": 5.223084764369168e-05, "loss": 2.6312, "step": 60920 }, { "epoch": 4.78, "learning_rate": 5.2230063514467184e-05, "loss": 2.6068, "step": 60921 }, { "epoch": 4.78, "learning_rate": 5.222927938524269e-05, "loss": 2.4659, "step": 60922 }, { "epoch": 4.78, "learning_rate": 5.22284952560182e-05, "loss": 2.6639, "step": 60923 }, { "epoch": 4.78, "learning_rate": 5.22277111267937e-05, "loss": 2.623, "step": 60924 }, { "epoch": 4.78, "learning_rate": 5.2226926997569205e-05, "loss": 2.7569, "step": 60925 }, { "epoch": 4.78, "learning_rate": 5.222614286834471e-05, "loss": 2.5719, "step": 60926 }, { "epoch": 4.78, "learning_rate": 5.222535873912021e-05, "loss": 2.6771, "step": 60927 }, { "epoch": 4.78, "learning_rate": 5.2224574609895716e-05, "loss": 2.7393, "step": 60928 }, { "epoch": 4.78, "learning_rate": 5.222379048067122e-05, "loss": 2.7131, "step": 60929 }, { "epoch": 4.78, "learning_rate": 5.2223006351446724e-05, "loss": 2.718, "step": 60930 }, { "epoch": 4.78, "learning_rate": 5.222222222222223e-05, "loss": 2.6275, "step": 60931 }, { "epoch": 4.78, "learning_rate": 5.222143809299773e-05, "loss": 2.5463, "step": 60932 }, { "epoch": 4.78, "learning_rate": 5.2220653963773235e-05, "loss": 2.5665, "step": 60933 }, { "epoch": 4.78, "learning_rate": 5.221986983454874e-05, "loss": 2.6847, "step": 60934 }, { "epoch": 4.78, "learning_rate": 5.221908570532424e-05, "loss": 2.7864, "step": 60935 }, { "epoch": 4.78, "learning_rate": 5.2218301576099746e-05, "loss": 2.5638, "step": 60936 }, { "epoch": 4.78, "learning_rate": 5.221751744687525e-05, "loss": 2.7884, "step": 60937 }, { "epoch": 4.78, "learning_rate": 5.221673331765076e-05, "loss": 2.6191, "step": 60938 }, { "epoch": 4.78, "learning_rate": 5.221594918842625e-05, "loss": 2.5154, "step": 60939 }, { "epoch": 4.78, "learning_rate": 5.2215165059201754e-05, "loss": 2.6118, "step": 60940 }, { "epoch": 4.78, "learning_rate": 5.221438092997726e-05, "loss": 2.6757, "step": 60941 }, { "epoch": 4.78, "learning_rate": 5.221359680075276e-05, "loss": 2.999, "step": 60942 }, { "epoch": 4.78, "learning_rate": 5.2212812671528265e-05, "loss": 2.6938, "step": 60943 }, { "epoch": 4.78, "learning_rate": 5.221202854230377e-05, "loss": 2.6099, "step": 60944 }, { "epoch": 4.78, "learning_rate": 5.221124441307927e-05, "loss": 2.6368, "step": 60945 }, { "epoch": 4.78, "learning_rate": 5.2210460283854776e-05, "loss": 2.6135, "step": 60946 }, { "epoch": 4.78, "learning_rate": 5.220967615463028e-05, "loss": 2.5513, "step": 60947 }, { "epoch": 4.78, "learning_rate": 5.220889202540579e-05, "loss": 2.6268, "step": 60948 }, { "epoch": 4.78, "learning_rate": 5.2208107896181294e-05, "loss": 2.5377, "step": 60949 }, { "epoch": 4.78, "learning_rate": 5.22073237669568e-05, "loss": 2.6174, "step": 60950 }, { "epoch": 4.78, "learning_rate": 5.22065396377323e-05, "loss": 2.6265, "step": 60951 }, { "epoch": 4.78, "learning_rate": 5.2205755508507805e-05, "loss": 2.8819, "step": 60952 }, { "epoch": 4.78, "learning_rate": 5.220497137928331e-05, "loss": 2.5853, "step": 60953 }, { "epoch": 4.78, "learning_rate": 5.220418725005881e-05, "loss": 2.8007, "step": 60954 }, { "epoch": 4.78, "learning_rate": 5.2203403120834316e-05, "loss": 2.6073, "step": 60955 }, { "epoch": 4.78, "learning_rate": 5.220261899160982e-05, "loss": 2.7055, "step": 60956 }, { "epoch": 4.78, "learning_rate": 5.220183486238532e-05, "loss": 2.5952, "step": 60957 }, { "epoch": 4.78, "learning_rate": 5.220105073316083e-05, "loss": 2.5209, "step": 60958 }, { "epoch": 4.78, "learning_rate": 5.220026660393633e-05, "loss": 2.5637, "step": 60959 }, { "epoch": 4.78, "learning_rate": 5.2199482474711834e-05, "loss": 3.0705, "step": 60960 }, { "epoch": 4.78, "learning_rate": 5.219869834548734e-05, "loss": 2.321, "step": 60961 }, { "epoch": 4.78, "learning_rate": 5.219791421626284e-05, "loss": 2.7106, "step": 60962 }, { "epoch": 4.78, "learning_rate": 5.2197130087038345e-05, "loss": 2.5983, "step": 60963 }, { "epoch": 4.78, "learning_rate": 5.2196345957813856e-05, "loss": 2.7145, "step": 60964 }, { "epoch": 4.78, "learning_rate": 5.219556182858936e-05, "loss": 2.7586, "step": 60965 }, { "epoch": 4.78, "learning_rate": 5.219477769936486e-05, "loss": 2.4421, "step": 60966 }, { "epoch": 4.78, "learning_rate": 5.2193993570140367e-05, "loss": 2.6302, "step": 60967 }, { "epoch": 4.78, "learning_rate": 5.219320944091587e-05, "loss": 2.7709, "step": 60968 }, { "epoch": 4.78, "learning_rate": 5.2192425311691374e-05, "loss": 2.7677, "step": 60969 }, { "epoch": 4.78, "learning_rate": 5.219164118246688e-05, "loss": 2.8825, "step": 60970 }, { "epoch": 4.78, "learning_rate": 5.219085705324238e-05, "loss": 2.6172, "step": 60971 }, { "epoch": 4.78, "learning_rate": 5.2190072924017885e-05, "loss": 2.7084, "step": 60972 }, { "epoch": 4.78, "learning_rate": 5.218928879479339e-05, "loss": 2.7682, "step": 60973 }, { "epoch": 4.78, "learning_rate": 5.218850466556889e-05, "loss": 2.5448, "step": 60974 }, { "epoch": 4.78, "learning_rate": 5.218772053634439e-05, "loss": 2.6386, "step": 60975 }, { "epoch": 4.78, "learning_rate": 5.218693640711989e-05, "loss": 2.609, "step": 60976 }, { "epoch": 4.78, "learning_rate": 5.2186152277895397e-05, "loss": 2.6705, "step": 60977 }, { "epoch": 4.78, "learning_rate": 5.21853681486709e-05, "loss": 2.479, "step": 60978 }, { "epoch": 4.78, "learning_rate": 5.2184584019446404e-05, "loss": 2.6934, "step": 60979 }, { "epoch": 4.78, "learning_rate": 5.218379989022191e-05, "loss": 2.5882, "step": 60980 }, { "epoch": 4.78, "learning_rate": 5.218301576099741e-05, "loss": 2.6888, "step": 60981 }, { "epoch": 4.78, "learning_rate": 5.2182231631772915e-05, "loss": 2.8094, "step": 60982 }, { "epoch": 4.78, "learning_rate": 5.218144750254842e-05, "loss": 2.803, "step": 60983 }, { "epoch": 4.78, "learning_rate": 5.218066337332392e-05, "loss": 2.8235, "step": 60984 }, { "epoch": 4.78, "learning_rate": 5.2179879244099426e-05, "loss": 2.6498, "step": 60985 }, { "epoch": 4.78, "learning_rate": 5.217909511487493e-05, "loss": 2.7718, "step": 60986 }, { "epoch": 4.78, "learning_rate": 5.217831098565043e-05, "loss": 2.7159, "step": 60987 }, { "epoch": 4.78, "learning_rate": 5.217752685642594e-05, "loss": 2.6414, "step": 60988 }, { "epoch": 4.78, "learning_rate": 5.217674272720145e-05, "loss": 2.8882, "step": 60989 }, { "epoch": 4.78, "learning_rate": 5.217595859797695e-05, "loss": 2.6015, "step": 60990 }, { "epoch": 4.78, "learning_rate": 5.2175174468752455e-05, "loss": 2.556, "step": 60991 }, { "epoch": 4.78, "learning_rate": 5.217439033952796e-05, "loss": 2.5176, "step": 60992 }, { "epoch": 4.78, "learning_rate": 5.217360621030346e-05, "loss": 2.6422, "step": 60993 }, { "epoch": 4.78, "learning_rate": 5.2172822081078966e-05, "loss": 2.6014, "step": 60994 }, { "epoch": 4.78, "learning_rate": 5.217203795185447e-05, "loss": 2.8435, "step": 60995 }, { "epoch": 4.78, "learning_rate": 5.217125382262997e-05, "loss": 2.5195, "step": 60996 }, { "epoch": 4.78, "learning_rate": 5.217046969340548e-05, "loss": 2.431, "step": 60997 }, { "epoch": 4.78, "learning_rate": 5.216968556418098e-05, "loss": 2.707, "step": 60998 }, { "epoch": 4.78, "learning_rate": 5.2168901434956484e-05, "loss": 2.8292, "step": 60999 }, { "epoch": 4.78, "learning_rate": 5.216811730573199e-05, "loss": 2.7423, "step": 61000 }, { "epoch": 4.78, "eval_loss": 2.8938300609588623, "eval_runtime": 217.7254, "eval_samples_per_second": 52.047, "eval_steps_per_second": 3.256, "step": 61000 }, { "epoch": 4.78, "learning_rate": 5.216733317650749e-05, "loss": 2.7812, "step": 61001 }, { "epoch": 4.78, "learning_rate": 5.2166549047282995e-05, "loss": 2.5768, "step": 61002 }, { "epoch": 4.78, "learning_rate": 5.21657649180585e-05, "loss": 2.6103, "step": 61003 }, { "epoch": 4.78, "learning_rate": 5.2164980788834e-05, "loss": 2.5714, "step": 61004 }, { "epoch": 4.78, "learning_rate": 5.216419665960951e-05, "loss": 2.617, "step": 61005 }, { "epoch": 4.78, "learning_rate": 5.216341253038502e-05, "loss": 2.6003, "step": 61006 }, { "epoch": 4.78, "learning_rate": 5.216262840116052e-05, "loss": 2.7594, "step": 61007 }, { "epoch": 4.78, "learning_rate": 5.2161844271936024e-05, "loss": 2.6997, "step": 61008 }, { "epoch": 4.78, "learning_rate": 5.216106014271153e-05, "loss": 2.7204, "step": 61009 }, { "epoch": 4.78, "learning_rate": 5.216027601348703e-05, "loss": 2.7704, "step": 61010 }, { "epoch": 4.78, "learning_rate": 5.215949188426252e-05, "loss": 2.8691, "step": 61011 }, { "epoch": 4.78, "learning_rate": 5.2158707755038025e-05, "loss": 2.7472, "step": 61012 }, { "epoch": 4.78, "learning_rate": 5.215792362581353e-05, "loss": 2.7054, "step": 61013 }, { "epoch": 4.78, "learning_rate": 5.215713949658903e-05, "loss": 2.6896, "step": 61014 }, { "epoch": 4.78, "learning_rate": 5.215635536736454e-05, "loss": 2.743, "step": 61015 }, { "epoch": 4.78, "learning_rate": 5.215557123814005e-05, "loss": 2.7201, "step": 61016 }, { "epoch": 4.78, "learning_rate": 5.215478710891555e-05, "loss": 2.6219, "step": 61017 }, { "epoch": 4.78, "learning_rate": 5.2154002979691054e-05, "loss": 2.6255, "step": 61018 }, { "epoch": 4.78, "learning_rate": 5.215321885046656e-05, "loss": 2.5479, "step": 61019 }, { "epoch": 4.78, "learning_rate": 5.215243472124206e-05, "loss": 2.4978, "step": 61020 }, { "epoch": 4.78, "learning_rate": 5.2151650592017565e-05, "loss": 2.5345, "step": 61021 }, { "epoch": 4.78, "learning_rate": 5.215086646279307e-05, "loss": 2.6194, "step": 61022 }, { "epoch": 4.78, "learning_rate": 5.215008233356857e-05, "loss": 2.7596, "step": 61023 }, { "epoch": 4.79, "learning_rate": 5.2149298204344076e-05, "loss": 2.6679, "step": 61024 }, { "epoch": 4.79, "learning_rate": 5.214851407511958e-05, "loss": 2.6022, "step": 61025 }, { "epoch": 4.79, "learning_rate": 5.2147729945895084e-05, "loss": 2.7443, "step": 61026 }, { "epoch": 4.79, "learning_rate": 5.214694581667059e-05, "loss": 2.6999, "step": 61027 }, { "epoch": 4.79, "learning_rate": 5.214616168744609e-05, "loss": 2.7061, "step": 61028 }, { "epoch": 4.79, "learning_rate": 5.2145377558221595e-05, "loss": 2.5482, "step": 61029 }, { "epoch": 4.79, "learning_rate": 5.21445934289971e-05, "loss": 2.7492, "step": 61030 }, { "epoch": 4.79, "learning_rate": 5.214380929977261e-05, "loss": 2.7167, "step": 61031 }, { "epoch": 4.79, "learning_rate": 5.214302517054811e-05, "loss": 2.8698, "step": 61032 }, { "epoch": 4.79, "learning_rate": 5.2142241041323616e-05, "loss": 2.5543, "step": 61033 }, { "epoch": 4.79, "learning_rate": 5.214145691209912e-05, "loss": 2.6057, "step": 61034 }, { "epoch": 4.79, "learning_rate": 5.2140672782874623e-05, "loss": 2.6627, "step": 61035 }, { "epoch": 4.79, "learning_rate": 5.213988865365013e-05, "loss": 2.4451, "step": 61036 }, { "epoch": 4.79, "learning_rate": 5.213910452442563e-05, "loss": 2.7685, "step": 61037 }, { "epoch": 4.79, "learning_rate": 5.2138320395201135e-05, "loss": 2.6938, "step": 61038 }, { "epoch": 4.79, "learning_rate": 5.213753626597664e-05, "loss": 2.6147, "step": 61039 }, { "epoch": 4.79, "learning_rate": 5.213675213675214e-05, "loss": 2.6993, "step": 61040 }, { "epoch": 4.79, "learning_rate": 5.2135968007527646e-05, "loss": 2.6141, "step": 61041 }, { "epoch": 4.79, "learning_rate": 5.213518387830315e-05, "loss": 2.6198, "step": 61042 }, { "epoch": 4.79, "learning_rate": 5.213439974907865e-05, "loss": 2.5632, "step": 61043 }, { "epoch": 4.79, "learning_rate": 5.2133615619854157e-05, "loss": 2.6786, "step": 61044 }, { "epoch": 4.79, "learning_rate": 5.213283149062966e-05, "loss": 2.5177, "step": 61045 }, { "epoch": 4.79, "learning_rate": 5.213204736140517e-05, "loss": 2.7804, "step": 61046 }, { "epoch": 4.79, "learning_rate": 5.213126323218066e-05, "loss": 2.7433, "step": 61047 }, { "epoch": 4.79, "learning_rate": 5.2130479102956165e-05, "loss": 2.7109, "step": 61048 }, { "epoch": 4.79, "learning_rate": 5.212969497373167e-05, "loss": 2.6294, "step": 61049 }, { "epoch": 4.79, "learning_rate": 5.212891084450717e-05, "loss": 2.8323, "step": 61050 }, { "epoch": 4.79, "learning_rate": 5.2128126715282676e-05, "loss": 2.6389, "step": 61051 }, { "epoch": 4.79, "learning_rate": 5.212734258605818e-05, "loss": 2.7802, "step": 61052 }, { "epoch": 4.79, "learning_rate": 5.212655845683368e-05, "loss": 2.5709, "step": 61053 }, { "epoch": 4.79, "learning_rate": 5.2125774327609187e-05, "loss": 2.6847, "step": 61054 }, { "epoch": 4.79, "learning_rate": 5.212499019838469e-05, "loss": 2.7732, "step": 61055 }, { "epoch": 4.79, "learning_rate": 5.21242060691602e-05, "loss": 2.4741, "step": 61056 }, { "epoch": 4.79, "learning_rate": 5.2123421939935704e-05, "loss": 2.8086, "step": 61057 }, { "epoch": 4.79, "learning_rate": 5.212263781071121e-05, "loss": 2.7371, "step": 61058 }, { "epoch": 4.79, "learning_rate": 5.212185368148671e-05, "loss": 2.6001, "step": 61059 }, { "epoch": 4.79, "learning_rate": 5.2121069552262215e-05, "loss": 2.6206, "step": 61060 }, { "epoch": 4.79, "learning_rate": 5.212028542303772e-05, "loss": 2.658, "step": 61061 }, { "epoch": 4.79, "learning_rate": 5.211950129381322e-05, "loss": 2.7063, "step": 61062 }, { "epoch": 4.79, "learning_rate": 5.2118717164588726e-05, "loss": 2.7484, "step": 61063 }, { "epoch": 4.79, "learning_rate": 5.211793303536423e-05, "loss": 2.758, "step": 61064 }, { "epoch": 4.79, "learning_rate": 5.2117148906139734e-05, "loss": 2.6546, "step": 61065 }, { "epoch": 4.79, "learning_rate": 5.211636477691524e-05, "loss": 2.772, "step": 61066 }, { "epoch": 4.79, "learning_rate": 5.211558064769074e-05, "loss": 2.6018, "step": 61067 }, { "epoch": 4.79, "learning_rate": 5.2114796518466245e-05, "loss": 2.6943, "step": 61068 }, { "epoch": 4.79, "learning_rate": 5.211401238924175e-05, "loss": 2.7515, "step": 61069 }, { "epoch": 4.79, "learning_rate": 5.211322826001725e-05, "loss": 2.6393, "step": 61070 }, { "epoch": 4.79, "learning_rate": 5.2112444130792756e-05, "loss": 2.9413, "step": 61071 }, { "epoch": 4.79, "learning_rate": 5.2111660001568266e-05, "loss": 2.6608, "step": 61072 }, { "epoch": 4.79, "learning_rate": 5.211087587234377e-05, "loss": 2.654, "step": 61073 }, { "epoch": 4.79, "learning_rate": 5.2110091743119274e-05, "loss": 2.7053, "step": 61074 }, { "epoch": 4.79, "learning_rate": 5.210930761389478e-05, "loss": 2.5371, "step": 61075 }, { "epoch": 4.79, "learning_rate": 5.210852348467028e-05, "loss": 2.7733, "step": 61076 }, { "epoch": 4.79, "learning_rate": 5.2107739355445785e-05, "loss": 2.6919, "step": 61077 }, { "epoch": 4.79, "learning_rate": 5.210695522622129e-05, "loss": 2.7563, "step": 61078 }, { "epoch": 4.79, "learning_rate": 5.210617109699679e-05, "loss": 2.7555, "step": 61079 }, { "epoch": 4.79, "learning_rate": 5.2105386967772296e-05, "loss": 2.5902, "step": 61080 }, { "epoch": 4.79, "learning_rate": 5.21046028385478e-05, "loss": 2.7429, "step": 61081 }, { "epoch": 4.79, "learning_rate": 5.21038187093233e-05, "loss": 2.62, "step": 61082 }, { "epoch": 4.79, "learning_rate": 5.21030345800988e-05, "loss": 2.6303, "step": 61083 }, { "epoch": 4.79, "learning_rate": 5.2102250450874304e-05, "loss": 2.7065, "step": 61084 }, { "epoch": 4.79, "learning_rate": 5.210146632164981e-05, "loss": 2.5952, "step": 61085 }, { "epoch": 4.79, "learning_rate": 5.210068219242531e-05, "loss": 2.5864, "step": 61086 }, { "epoch": 4.79, "learning_rate": 5.2099898063200815e-05, "loss": 2.7677, "step": 61087 }, { "epoch": 4.79, "learning_rate": 5.209911393397632e-05, "loss": 2.5286, "step": 61088 }, { "epoch": 4.79, "learning_rate": 5.209832980475182e-05, "loss": 2.5323, "step": 61089 }, { "epoch": 4.79, "learning_rate": 5.2097545675527326e-05, "loss": 2.6141, "step": 61090 }, { "epoch": 4.79, "learning_rate": 5.209676154630283e-05, "loss": 2.5905, "step": 61091 }, { "epoch": 4.79, "learning_rate": 5.209597741707833e-05, "loss": 2.7169, "step": 61092 }, { "epoch": 4.79, "learning_rate": 5.209519328785384e-05, "loss": 2.6455, "step": 61093 }, { "epoch": 4.79, "learning_rate": 5.209440915862934e-05, "loss": 2.6446, "step": 61094 }, { "epoch": 4.79, "learning_rate": 5.2093625029404844e-05, "loss": 2.6951, "step": 61095 }, { "epoch": 4.79, "learning_rate": 5.209284090018035e-05, "loss": 2.7192, "step": 61096 }, { "epoch": 4.79, "learning_rate": 5.209205677095586e-05, "loss": 2.5509, "step": 61097 }, { "epoch": 4.79, "learning_rate": 5.209127264173136e-05, "loss": 2.6793, "step": 61098 }, { "epoch": 4.79, "learning_rate": 5.2090488512506866e-05, "loss": 2.66, "step": 61099 }, { "epoch": 4.79, "learning_rate": 5.208970438328237e-05, "loss": 2.6359, "step": 61100 }, { "epoch": 4.79, "learning_rate": 5.208892025405787e-05, "loss": 2.7753, "step": 61101 }, { "epoch": 4.79, "learning_rate": 5.208813612483338e-05, "loss": 2.5518, "step": 61102 }, { "epoch": 4.79, "learning_rate": 5.208735199560888e-05, "loss": 2.6245, "step": 61103 }, { "epoch": 4.79, "learning_rate": 5.2086567866384384e-05, "loss": 2.968, "step": 61104 }, { "epoch": 4.79, "learning_rate": 5.208578373715989e-05, "loss": 2.9307, "step": 61105 }, { "epoch": 4.79, "learning_rate": 5.208499960793539e-05, "loss": 2.778, "step": 61106 }, { "epoch": 4.79, "learning_rate": 5.2084215478710895e-05, "loss": 2.6402, "step": 61107 }, { "epoch": 4.79, "learning_rate": 5.20834313494864e-05, "loss": 2.5649, "step": 61108 }, { "epoch": 4.79, "learning_rate": 5.20826472202619e-05, "loss": 2.718, "step": 61109 }, { "epoch": 4.79, "learning_rate": 5.2081863091037406e-05, "loss": 2.8237, "step": 61110 }, { "epoch": 4.79, "learning_rate": 5.208107896181291e-05, "loss": 2.8694, "step": 61111 }, { "epoch": 4.79, "learning_rate": 5.2080294832588413e-05, "loss": 2.5301, "step": 61112 }, { "epoch": 4.79, "learning_rate": 5.2079510703363924e-05, "loss": 2.718, "step": 61113 }, { "epoch": 4.79, "learning_rate": 5.207872657413943e-05, "loss": 2.7547, "step": 61114 }, { "epoch": 4.79, "learning_rate": 5.207794244491493e-05, "loss": 2.8197, "step": 61115 }, { "epoch": 4.79, "learning_rate": 5.2077158315690435e-05, "loss": 2.6245, "step": 61116 }, { "epoch": 4.79, "learning_rate": 5.207637418646594e-05, "loss": 2.7161, "step": 61117 }, { "epoch": 4.79, "learning_rate": 5.207559005724144e-05, "loss": 2.4771, "step": 61118 }, { "epoch": 4.79, "learning_rate": 5.207480592801693e-05, "loss": 2.8549, "step": 61119 }, { "epoch": 4.79, "learning_rate": 5.2074021798792436e-05, "loss": 2.5367, "step": 61120 }, { "epoch": 4.79, "learning_rate": 5.207323766956794e-05, "loss": 2.6585, "step": 61121 }, { "epoch": 4.79, "learning_rate": 5.2072453540343443e-05, "loss": 2.6262, "step": 61122 }, { "epoch": 4.79, "learning_rate": 5.2071669411118954e-05, "loss": 2.6133, "step": 61123 }, { "epoch": 4.79, "learning_rate": 5.207088528189446e-05, "loss": 2.4673, "step": 61124 }, { "epoch": 4.79, "learning_rate": 5.207010115266996e-05, "loss": 2.7659, "step": 61125 }, { "epoch": 4.79, "learning_rate": 5.2069317023445465e-05, "loss": 2.8604, "step": 61126 }, { "epoch": 4.79, "learning_rate": 5.206853289422097e-05, "loss": 2.747, "step": 61127 }, { "epoch": 4.79, "learning_rate": 5.206774876499647e-05, "loss": 2.6121, "step": 61128 }, { "epoch": 4.79, "learning_rate": 5.2066964635771976e-05, "loss": 2.7536, "step": 61129 }, { "epoch": 4.79, "learning_rate": 5.206618050654748e-05, "loss": 2.5579, "step": 61130 }, { "epoch": 4.79, "learning_rate": 5.206539637732298e-05, "loss": 2.6712, "step": 61131 }, { "epoch": 4.79, "learning_rate": 5.206461224809849e-05, "loss": 2.5828, "step": 61132 }, { "epoch": 4.79, "learning_rate": 5.206382811887399e-05, "loss": 2.7126, "step": 61133 }, { "epoch": 4.79, "learning_rate": 5.2063043989649494e-05, "loss": 2.7917, "step": 61134 }, { "epoch": 4.79, "learning_rate": 5.2062259860425e-05, "loss": 2.5558, "step": 61135 }, { "epoch": 4.79, "learning_rate": 5.20614757312005e-05, "loss": 2.6599, "step": 61136 }, { "epoch": 4.79, "learning_rate": 5.2060691601976005e-05, "loss": 2.7741, "step": 61137 }, { "epoch": 4.79, "learning_rate": 5.205990747275151e-05, "loss": 2.9695, "step": 61138 }, { "epoch": 4.79, "learning_rate": 5.205912334352702e-05, "loss": 2.7211, "step": 61139 }, { "epoch": 4.79, "learning_rate": 5.205833921430252e-05, "loss": 3.0563, "step": 61140 }, { "epoch": 4.79, "learning_rate": 5.205755508507803e-05, "loss": 2.7209, "step": 61141 }, { "epoch": 4.79, "learning_rate": 5.205677095585353e-05, "loss": 2.4192, "step": 61142 }, { "epoch": 4.79, "learning_rate": 5.2055986826629034e-05, "loss": 2.6213, "step": 61143 }, { "epoch": 4.79, "learning_rate": 5.205520269740454e-05, "loss": 2.4581, "step": 61144 }, { "epoch": 4.79, "learning_rate": 5.205441856818004e-05, "loss": 2.7028, "step": 61145 }, { "epoch": 4.79, "learning_rate": 5.2053634438955545e-05, "loss": 2.5207, "step": 61146 }, { "epoch": 4.79, "learning_rate": 5.205285030973105e-05, "loss": 2.5975, "step": 61147 }, { "epoch": 4.79, "learning_rate": 5.205206618050655e-05, "loss": 2.7088, "step": 61148 }, { "epoch": 4.79, "learning_rate": 5.2051282051282056e-05, "loss": 2.7866, "step": 61149 }, { "epoch": 4.79, "learning_rate": 5.205049792205756e-05, "loss": 2.5183, "step": 61150 }, { "epoch": 4.8, "learning_rate": 5.2049713792833064e-05, "loss": 2.6035, "step": 61151 }, { "epoch": 4.8, "learning_rate": 5.204892966360857e-05, "loss": 2.8003, "step": 61152 }, { "epoch": 4.8, "learning_rate": 5.204814553438407e-05, "loss": 2.7446, "step": 61153 }, { "epoch": 4.8, "learning_rate": 5.204736140515958e-05, "loss": 3.0594, "step": 61154 }, { "epoch": 4.8, "learning_rate": 5.204657727593507e-05, "loss": 2.6087, "step": 61155 }, { "epoch": 4.8, "learning_rate": 5.2045793146710575e-05, "loss": 2.488, "step": 61156 }, { "epoch": 4.8, "learning_rate": 5.204500901748608e-05, "loss": 2.4678, "step": 61157 }, { "epoch": 4.8, "learning_rate": 5.204422488826158e-05, "loss": 2.6702, "step": 61158 }, { "epoch": 4.8, "learning_rate": 5.2043440759037086e-05, "loss": 2.6395, "step": 61159 }, { "epoch": 4.8, "learning_rate": 5.204265662981259e-05, "loss": 2.7035, "step": 61160 }, { "epoch": 4.8, "learning_rate": 5.2041872500588094e-05, "loss": 2.5354, "step": 61161 }, { "epoch": 4.8, "learning_rate": 5.20410883713636e-05, "loss": 2.6193, "step": 61162 }, { "epoch": 4.8, "learning_rate": 5.20403042421391e-05, "loss": 2.6281, "step": 61163 }, { "epoch": 4.8, "learning_rate": 5.203952011291461e-05, "loss": 2.7506, "step": 61164 }, { "epoch": 4.8, "learning_rate": 5.2038735983690115e-05, "loss": 2.6735, "step": 61165 }, { "epoch": 4.8, "learning_rate": 5.203795185446562e-05, "loss": 2.7428, "step": 61166 }, { "epoch": 4.8, "learning_rate": 5.203716772524112e-05, "loss": 2.6487, "step": 61167 }, { "epoch": 4.8, "learning_rate": 5.2036383596016626e-05, "loss": 2.645, "step": 61168 }, { "epoch": 4.8, "learning_rate": 5.203559946679213e-05, "loss": 2.6941, "step": 61169 }, { "epoch": 4.8, "learning_rate": 5.2034815337567634e-05, "loss": 2.5587, "step": 61170 }, { "epoch": 4.8, "learning_rate": 5.203403120834314e-05, "loss": 2.8306, "step": 61171 }, { "epoch": 4.8, "learning_rate": 5.203324707911864e-05, "loss": 2.8856, "step": 61172 }, { "epoch": 4.8, "learning_rate": 5.2032462949894145e-05, "loss": 2.7106, "step": 61173 }, { "epoch": 4.8, "learning_rate": 5.203167882066965e-05, "loss": 2.877, "step": 61174 }, { "epoch": 4.8, "learning_rate": 5.203089469144515e-05, "loss": 2.5698, "step": 61175 }, { "epoch": 4.8, "learning_rate": 5.2030110562220656e-05, "loss": 2.6223, "step": 61176 }, { "epoch": 4.8, "learning_rate": 5.202932643299616e-05, "loss": 2.3906, "step": 61177 }, { "epoch": 4.8, "learning_rate": 5.202854230377166e-05, "loss": 2.5245, "step": 61178 }, { "epoch": 4.8, "learning_rate": 5.2027758174547167e-05, "loss": 2.761, "step": 61179 }, { "epoch": 4.8, "learning_rate": 5.202697404532268e-05, "loss": 2.6791, "step": 61180 }, { "epoch": 4.8, "learning_rate": 5.202618991609818e-05, "loss": 2.7535, "step": 61181 }, { "epoch": 4.8, "learning_rate": 5.2025405786873684e-05, "loss": 2.5563, "step": 61182 }, { "epoch": 4.8, "learning_rate": 5.202462165764919e-05, "loss": 2.6979, "step": 61183 }, { "epoch": 4.8, "learning_rate": 5.202383752842469e-05, "loss": 2.511, "step": 61184 }, { "epoch": 4.8, "learning_rate": 5.2023053399200195e-05, "loss": 2.5634, "step": 61185 }, { "epoch": 4.8, "learning_rate": 5.20222692699757e-05, "loss": 2.4798, "step": 61186 }, { "epoch": 4.8, "learning_rate": 5.20214851407512e-05, "loss": 2.5398, "step": 61187 }, { "epoch": 4.8, "learning_rate": 5.2020701011526706e-05, "loss": 2.767, "step": 61188 }, { "epoch": 4.8, "learning_rate": 5.201991688230221e-05, "loss": 2.7319, "step": 61189 }, { "epoch": 4.8, "learning_rate": 5.2019132753077714e-05, "loss": 2.7586, "step": 61190 }, { "epoch": 4.8, "learning_rate": 5.201834862385321e-05, "loss": 2.5805, "step": 61191 }, { "epoch": 4.8, "learning_rate": 5.2017564494628714e-05, "loss": 2.6884, "step": 61192 }, { "epoch": 4.8, "learning_rate": 5.201678036540422e-05, "loss": 2.5778, "step": 61193 }, { "epoch": 4.8, "learning_rate": 5.201599623617972e-05, "loss": 2.8215, "step": 61194 }, { "epoch": 4.8, "learning_rate": 5.2015212106955225e-05, "loss": 2.6986, "step": 61195 }, { "epoch": 4.8, "learning_rate": 5.201442797773073e-05, "loss": 2.7377, "step": 61196 }, { "epoch": 4.8, "learning_rate": 5.201364384850623e-05, "loss": 2.6821, "step": 61197 }, { "epoch": 4.8, "learning_rate": 5.2012859719281736e-05, "loss": 2.742, "step": 61198 }, { "epoch": 4.8, "learning_rate": 5.201207559005724e-05, "loss": 2.7847, "step": 61199 }, { "epoch": 4.8, "learning_rate": 5.2011291460832744e-05, "loss": 2.7466, "step": 61200 }, { "epoch": 4.8, "learning_rate": 5.201050733160825e-05, "loss": 2.5562, "step": 61201 }, { "epoch": 4.8, "learning_rate": 5.200972320238375e-05, "loss": 2.5118, "step": 61202 }, { "epoch": 4.8, "learning_rate": 5.2008939073159255e-05, "loss": 2.5394, "step": 61203 }, { "epoch": 4.8, "learning_rate": 5.200815494393476e-05, "loss": 2.6858, "step": 61204 }, { "epoch": 4.8, "learning_rate": 5.200737081471027e-05, "loss": 2.6548, "step": 61205 }, { "epoch": 4.8, "learning_rate": 5.200658668548577e-05, "loss": 2.8276, "step": 61206 }, { "epoch": 4.8, "learning_rate": 5.2005802556261276e-05, "loss": 2.6278, "step": 61207 }, { "epoch": 4.8, "learning_rate": 5.200501842703678e-05, "loss": 2.6897, "step": 61208 }, { "epoch": 4.8, "learning_rate": 5.2004234297812284e-05, "loss": 2.7974, "step": 61209 }, { "epoch": 4.8, "learning_rate": 5.200345016858779e-05, "loss": 2.7098, "step": 61210 }, { "epoch": 4.8, "learning_rate": 5.200266603936329e-05, "loss": 2.7963, "step": 61211 }, { "epoch": 4.8, "learning_rate": 5.2001881910138795e-05, "loss": 2.8415, "step": 61212 }, { "epoch": 4.8, "learning_rate": 5.20010977809143e-05, "loss": 2.5216, "step": 61213 }, { "epoch": 4.8, "learning_rate": 5.20003136516898e-05, "loss": 2.7068, "step": 61214 }, { "epoch": 4.8, "learning_rate": 5.1999529522465306e-05, "loss": 2.5706, "step": 61215 }, { "epoch": 4.8, "learning_rate": 5.199874539324081e-05, "loss": 2.6582, "step": 61216 }, { "epoch": 4.8, "learning_rate": 5.199796126401631e-05, "loss": 2.5694, "step": 61217 }, { "epoch": 4.8, "learning_rate": 5.199717713479182e-05, "loss": 2.5833, "step": 61218 }, { "epoch": 4.8, "learning_rate": 5.199639300556732e-05, "loss": 2.6927, "step": 61219 }, { "epoch": 4.8, "learning_rate": 5.1995608876342824e-05, "loss": 2.7665, "step": 61220 }, { "epoch": 4.8, "learning_rate": 5.1994824747118335e-05, "loss": 2.7464, "step": 61221 }, { "epoch": 4.8, "learning_rate": 5.199404061789384e-05, "loss": 2.833, "step": 61222 }, { "epoch": 4.8, "learning_rate": 5.199325648866934e-05, "loss": 2.6577, "step": 61223 }, { "epoch": 4.8, "learning_rate": 5.1992472359444846e-05, "loss": 2.6756, "step": 61224 }, { "epoch": 4.8, "learning_rate": 5.199168823022035e-05, "loss": 2.5897, "step": 61225 }, { "epoch": 4.8, "learning_rate": 5.199090410099585e-05, "loss": 2.5259, "step": 61226 }, { "epoch": 4.8, "learning_rate": 5.199011997177134e-05, "loss": 2.409, "step": 61227 }, { "epoch": 4.8, "learning_rate": 5.198933584254685e-05, "loss": 2.7658, "step": 61228 }, { "epoch": 4.8, "learning_rate": 5.198855171332235e-05, "loss": 2.7228, "step": 61229 }, { "epoch": 4.8, "learning_rate": 5.1987767584097854e-05, "loss": 2.6471, "step": 61230 }, { "epoch": 4.8, "learning_rate": 5.1986983454873365e-05, "loss": 2.5589, "step": 61231 }, { "epoch": 4.8, "learning_rate": 5.198619932564887e-05, "loss": 2.5886, "step": 61232 }, { "epoch": 4.8, "learning_rate": 5.198541519642437e-05, "loss": 2.5179, "step": 61233 }, { "epoch": 4.8, "learning_rate": 5.1984631067199876e-05, "loss": 2.7655, "step": 61234 }, { "epoch": 4.8, "learning_rate": 5.198384693797538e-05, "loss": 2.7502, "step": 61235 }, { "epoch": 4.8, "learning_rate": 5.198306280875088e-05, "loss": 2.6067, "step": 61236 }, { "epoch": 4.8, "learning_rate": 5.198227867952639e-05, "loss": 2.7593, "step": 61237 }, { "epoch": 4.8, "learning_rate": 5.198149455030189e-05, "loss": 2.7676, "step": 61238 }, { "epoch": 4.8, "learning_rate": 5.1980710421077394e-05, "loss": 2.6632, "step": 61239 }, { "epoch": 4.8, "learning_rate": 5.19799262918529e-05, "loss": 2.7671, "step": 61240 }, { "epoch": 4.8, "learning_rate": 5.19791421626284e-05, "loss": 2.7016, "step": 61241 }, { "epoch": 4.8, "learning_rate": 5.1978358033403905e-05, "loss": 2.7223, "step": 61242 }, { "epoch": 4.8, "learning_rate": 5.197757390417941e-05, "loss": 2.507, "step": 61243 }, { "epoch": 4.8, "learning_rate": 5.197678977495491e-05, "loss": 2.7897, "step": 61244 }, { "epoch": 4.8, "learning_rate": 5.1976005645730416e-05, "loss": 2.612, "step": 61245 }, { "epoch": 4.8, "learning_rate": 5.197522151650592e-05, "loss": 3.0396, "step": 61246 }, { "epoch": 4.8, "learning_rate": 5.197443738728143e-05, "loss": 2.5587, "step": 61247 }, { "epoch": 4.8, "learning_rate": 5.1973653258056934e-05, "loss": 2.6684, "step": 61248 }, { "epoch": 4.8, "learning_rate": 5.197286912883244e-05, "loss": 2.6878, "step": 61249 }, { "epoch": 4.8, "learning_rate": 5.197208499960794e-05, "loss": 2.579, "step": 61250 }, { "epoch": 4.8, "learning_rate": 5.1971300870383445e-05, "loss": 2.8736, "step": 61251 }, { "epoch": 4.8, "learning_rate": 5.197051674115895e-05, "loss": 2.7983, "step": 61252 }, { "epoch": 4.8, "learning_rate": 5.196973261193445e-05, "loss": 2.624, "step": 61253 }, { "epoch": 4.8, "learning_rate": 5.1968948482709956e-05, "loss": 3.0332, "step": 61254 }, { "epoch": 4.8, "learning_rate": 5.196816435348546e-05, "loss": 2.7548, "step": 61255 }, { "epoch": 4.8, "learning_rate": 5.196738022426096e-05, "loss": 2.5893, "step": 61256 }, { "epoch": 4.8, "learning_rate": 5.196659609503647e-05, "loss": 2.4201, "step": 61257 }, { "epoch": 4.8, "learning_rate": 5.196581196581197e-05, "loss": 2.6615, "step": 61258 }, { "epoch": 4.8, "learning_rate": 5.1965027836587474e-05, "loss": 2.7039, "step": 61259 }, { "epoch": 4.8, "learning_rate": 5.196424370736298e-05, "loss": 2.549, "step": 61260 }, { "epoch": 4.8, "learning_rate": 5.196345957813848e-05, "loss": 2.7783, "step": 61261 }, { "epoch": 4.8, "learning_rate": 5.196267544891399e-05, "loss": 2.75, "step": 61262 }, { "epoch": 4.8, "learning_rate": 5.196189131968948e-05, "loss": 2.558, "step": 61263 }, { "epoch": 4.8, "learning_rate": 5.1961107190464986e-05, "loss": 2.6508, "step": 61264 }, { "epoch": 4.8, "learning_rate": 5.196032306124049e-05, "loss": 2.6672, "step": 61265 }, { "epoch": 4.8, "learning_rate": 5.195953893201599e-05, "loss": 2.5208, "step": 61266 }, { "epoch": 4.8, "learning_rate": 5.19587548027915e-05, "loss": 3.073, "step": 61267 }, { "epoch": 4.8, "learning_rate": 5.1957970673567e-05, "loss": 2.7552, "step": 61268 }, { "epoch": 4.8, "learning_rate": 5.1957186544342504e-05, "loss": 2.6972, "step": 61269 }, { "epoch": 4.8, "learning_rate": 5.195640241511801e-05, "loss": 2.6784, "step": 61270 }, { "epoch": 4.8, "learning_rate": 5.195561828589351e-05, "loss": 2.7013, "step": 61271 }, { "epoch": 4.8, "learning_rate": 5.195483415666902e-05, "loss": 2.4648, "step": 61272 }, { "epoch": 4.8, "learning_rate": 5.1954050027444526e-05, "loss": 2.7209, "step": 61273 }, { "epoch": 4.8, "learning_rate": 5.195326589822003e-05, "loss": 2.5659, "step": 61274 }, { "epoch": 4.8, "learning_rate": 5.195248176899553e-05, "loss": 2.715, "step": 61275 }, { "epoch": 4.8, "learning_rate": 5.195169763977104e-05, "loss": 2.7146, "step": 61276 }, { "epoch": 4.8, "learning_rate": 5.195091351054654e-05, "loss": 2.6603, "step": 61277 }, { "epoch": 4.8, "learning_rate": 5.1950129381322044e-05, "loss": 2.6878, "step": 61278 }, { "epoch": 4.81, "learning_rate": 5.194934525209755e-05, "loss": 2.7006, "step": 61279 }, { "epoch": 4.81, "learning_rate": 5.194856112287305e-05, "loss": 2.7587, "step": 61280 }, { "epoch": 4.81, "learning_rate": 5.1947776993648555e-05, "loss": 2.7329, "step": 61281 }, { "epoch": 4.81, "learning_rate": 5.194699286442406e-05, "loss": 2.9111, "step": 61282 }, { "epoch": 4.81, "learning_rate": 5.194620873519956e-05, "loss": 2.7244, "step": 61283 }, { "epoch": 4.81, "learning_rate": 5.1945424605975066e-05, "loss": 2.7257, "step": 61284 }, { "epoch": 4.81, "learning_rate": 5.194464047675057e-05, "loss": 2.9445, "step": 61285 }, { "epoch": 4.81, "learning_rate": 5.1943856347526074e-05, "loss": 2.6977, "step": 61286 }, { "epoch": 4.81, "learning_rate": 5.194307221830158e-05, "loss": 2.7543, "step": 61287 }, { "epoch": 4.81, "learning_rate": 5.194228808907709e-05, "loss": 2.5563, "step": 61288 }, { "epoch": 4.81, "learning_rate": 5.194150395985259e-05, "loss": 2.7235, "step": 61289 }, { "epoch": 4.81, "learning_rate": 5.1940719830628095e-05, "loss": 2.716, "step": 61290 }, { "epoch": 4.81, "learning_rate": 5.19399357014036e-05, "loss": 2.7269, "step": 61291 }, { "epoch": 4.81, "learning_rate": 5.19391515721791e-05, "loss": 2.8247, "step": 61292 }, { "epoch": 4.81, "learning_rate": 5.1938367442954606e-05, "loss": 2.8077, "step": 61293 }, { "epoch": 4.81, "learning_rate": 5.193758331373011e-05, "loss": 2.5475, "step": 61294 }, { "epoch": 4.81, "learning_rate": 5.1936799184505614e-05, "loss": 2.8128, "step": 61295 }, { "epoch": 4.81, "learning_rate": 5.193601505528112e-05, "loss": 2.5645, "step": 61296 }, { "epoch": 4.81, "learning_rate": 5.193523092605662e-05, "loss": 2.6837, "step": 61297 }, { "epoch": 4.81, "learning_rate": 5.1934446796832125e-05, "loss": 2.6699, "step": 61298 }, { "epoch": 4.81, "learning_rate": 5.193366266760762e-05, "loss": 2.9468, "step": 61299 }, { "epoch": 4.81, "learning_rate": 5.1932878538383125e-05, "loss": 2.8381, "step": 61300 }, { "epoch": 4.81, "learning_rate": 5.193209440915863e-05, "loss": 2.6442, "step": 61301 }, { "epoch": 4.81, "learning_rate": 5.193131027993413e-05, "loss": 2.5859, "step": 61302 }, { "epoch": 4.81, "learning_rate": 5.1930526150709636e-05, "loss": 2.7114, "step": 61303 }, { "epoch": 4.81, "learning_rate": 5.192974202148514e-05, "loss": 2.8587, "step": 61304 }, { "epoch": 4.81, "learning_rate": 5.1928957892260644e-05, "loss": 2.4049, "step": 61305 }, { "epoch": 4.81, "learning_rate": 5.192817376303615e-05, "loss": 2.6449, "step": 61306 }, { "epoch": 4.81, "learning_rate": 5.192738963381165e-05, "loss": 2.8196, "step": 61307 }, { "epoch": 4.81, "learning_rate": 5.1926605504587155e-05, "loss": 2.5822, "step": 61308 }, { "epoch": 4.81, "learning_rate": 5.192582137536266e-05, "loss": 2.6223, "step": 61309 }, { "epoch": 4.81, "learning_rate": 5.192503724613816e-05, "loss": 2.7738, "step": 61310 }, { "epoch": 4.81, "learning_rate": 5.1924253116913666e-05, "loss": 2.5509, "step": 61311 }, { "epoch": 4.81, "learning_rate": 5.192346898768917e-05, "loss": 2.8508, "step": 61312 }, { "epoch": 4.81, "learning_rate": 5.192268485846467e-05, "loss": 2.6817, "step": 61313 }, { "epoch": 4.81, "learning_rate": 5.1921900729240183e-05, "loss": 2.9745, "step": 61314 }, { "epoch": 4.81, "learning_rate": 5.192111660001569e-05, "loss": 2.6951, "step": 61315 }, { "epoch": 4.81, "learning_rate": 5.192033247079119e-05, "loss": 2.7133, "step": 61316 }, { "epoch": 4.81, "learning_rate": 5.1919548341566694e-05, "loss": 2.4909, "step": 61317 }, { "epoch": 4.81, "learning_rate": 5.19187642123422e-05, "loss": 2.7746, "step": 61318 }, { "epoch": 4.81, "learning_rate": 5.19179800831177e-05, "loss": 2.7005, "step": 61319 }, { "epoch": 4.81, "learning_rate": 5.1917195953893205e-05, "loss": 2.6814, "step": 61320 }, { "epoch": 4.81, "learning_rate": 5.191641182466871e-05, "loss": 2.6299, "step": 61321 }, { "epoch": 4.81, "learning_rate": 5.191562769544421e-05, "loss": 2.6247, "step": 61322 }, { "epoch": 4.81, "learning_rate": 5.1914843566219717e-05, "loss": 2.5467, "step": 61323 }, { "epoch": 4.81, "learning_rate": 5.191405943699522e-05, "loss": 2.4982, "step": 61324 }, { "epoch": 4.81, "learning_rate": 5.1913275307770724e-05, "loss": 2.7687, "step": 61325 }, { "epoch": 4.81, "learning_rate": 5.191249117854623e-05, "loss": 2.7518, "step": 61326 }, { "epoch": 4.81, "learning_rate": 5.191170704932173e-05, "loss": 2.5711, "step": 61327 }, { "epoch": 4.81, "learning_rate": 5.1910922920097235e-05, "loss": 2.5262, "step": 61328 }, { "epoch": 4.81, "learning_rate": 5.1910138790872745e-05, "loss": 2.8139, "step": 61329 }, { "epoch": 4.81, "learning_rate": 5.190935466164825e-05, "loss": 2.5382, "step": 61330 }, { "epoch": 4.81, "learning_rate": 5.190857053242375e-05, "loss": 2.7722, "step": 61331 }, { "epoch": 4.81, "learning_rate": 5.1907786403199256e-05, "loss": 2.7512, "step": 61332 }, { "epoch": 4.81, "learning_rate": 5.190700227397476e-05, "loss": 2.6285, "step": 61333 }, { "epoch": 4.81, "learning_rate": 5.1906218144750264e-05, "loss": 2.7311, "step": 61334 }, { "epoch": 4.81, "learning_rate": 5.1905434015525754e-05, "loss": 2.6424, "step": 61335 }, { "epoch": 4.81, "learning_rate": 5.190464988630126e-05, "loss": 2.598, "step": 61336 }, { "epoch": 4.81, "learning_rate": 5.190386575707676e-05, "loss": 2.7345, "step": 61337 }, { "epoch": 4.81, "learning_rate": 5.1903081627852265e-05, "loss": 2.7576, "step": 61338 }, { "epoch": 4.81, "learning_rate": 5.1902297498627775e-05, "loss": 2.5854, "step": 61339 }, { "epoch": 4.81, "learning_rate": 5.190151336940328e-05, "loss": 2.5831, "step": 61340 }, { "epoch": 4.81, "learning_rate": 5.190072924017878e-05, "loss": 2.693, "step": 61341 }, { "epoch": 4.81, "learning_rate": 5.1899945110954286e-05, "loss": 2.8941, "step": 61342 }, { "epoch": 4.81, "learning_rate": 5.189916098172979e-05, "loss": 2.5571, "step": 61343 }, { "epoch": 4.81, "learning_rate": 5.1898376852505294e-05, "loss": 2.8105, "step": 61344 }, { "epoch": 4.81, "learning_rate": 5.18975927232808e-05, "loss": 2.5294, "step": 61345 }, { "epoch": 4.81, "learning_rate": 5.18968085940563e-05, "loss": 2.7239, "step": 61346 }, { "epoch": 4.81, "learning_rate": 5.1896024464831805e-05, "loss": 2.7392, "step": 61347 }, { "epoch": 4.81, "learning_rate": 5.189524033560731e-05, "loss": 2.6356, "step": 61348 }, { "epoch": 4.81, "learning_rate": 5.189445620638281e-05, "loss": 2.549, "step": 61349 }, { "epoch": 4.81, "learning_rate": 5.1893672077158316e-05, "loss": 2.7081, "step": 61350 }, { "epoch": 4.81, "learning_rate": 5.189288794793382e-05, "loss": 2.6616, "step": 61351 }, { "epoch": 4.81, "learning_rate": 5.189210381870932e-05, "loss": 2.7574, "step": 61352 }, { "epoch": 4.81, "learning_rate": 5.189131968948483e-05, "loss": 2.7889, "step": 61353 }, { "epoch": 4.81, "learning_rate": 5.189053556026033e-05, "loss": 2.5651, "step": 61354 }, { "epoch": 4.81, "learning_rate": 5.188975143103584e-05, "loss": 2.6128, "step": 61355 }, { "epoch": 4.81, "learning_rate": 5.1888967301811345e-05, "loss": 2.6342, "step": 61356 }, { "epoch": 4.81, "learning_rate": 5.188818317258685e-05, "loss": 2.8151, "step": 61357 }, { "epoch": 4.81, "learning_rate": 5.188739904336235e-05, "loss": 2.7736, "step": 61358 }, { "epoch": 4.81, "learning_rate": 5.1886614914137856e-05, "loss": 2.7095, "step": 61359 }, { "epoch": 4.81, "learning_rate": 5.188583078491336e-05, "loss": 2.7133, "step": 61360 }, { "epoch": 4.81, "learning_rate": 5.188504665568886e-05, "loss": 2.4749, "step": 61361 }, { "epoch": 4.81, "learning_rate": 5.188426252646437e-05, "loss": 2.6791, "step": 61362 }, { "epoch": 4.81, "learning_rate": 5.188347839723987e-05, "loss": 2.8029, "step": 61363 }, { "epoch": 4.81, "learning_rate": 5.1882694268015374e-05, "loss": 2.5025, "step": 61364 }, { "epoch": 4.81, "learning_rate": 5.188191013879088e-05, "loss": 2.7366, "step": 61365 }, { "epoch": 4.81, "learning_rate": 5.188112600956638e-05, "loss": 2.682, "step": 61366 }, { "epoch": 4.81, "learning_rate": 5.1880341880341885e-05, "loss": 2.6533, "step": 61367 }, { "epoch": 4.81, "learning_rate": 5.187955775111739e-05, "loss": 2.6625, "step": 61368 }, { "epoch": 4.81, "learning_rate": 5.187877362189289e-05, "loss": 2.6479, "step": 61369 }, { "epoch": 4.81, "learning_rate": 5.18779894926684e-05, "loss": 2.6672, "step": 61370 }, { "epoch": 4.81, "learning_rate": 5.187720536344389e-05, "loss": 2.6474, "step": 61371 }, { "epoch": 4.81, "learning_rate": 5.18764212342194e-05, "loss": 2.7378, "step": 61372 }, { "epoch": 4.81, "learning_rate": 5.18756371049949e-05, "loss": 2.7259, "step": 61373 }, { "epoch": 4.81, "learning_rate": 5.1874852975770404e-05, "loss": 2.7497, "step": 61374 }, { "epoch": 4.81, "learning_rate": 5.187406884654591e-05, "loss": 2.5377, "step": 61375 }, { "epoch": 4.81, "learning_rate": 5.187328471732141e-05, "loss": 2.8654, "step": 61376 }, { "epoch": 4.81, "learning_rate": 5.1872500588096915e-05, "loss": 2.8086, "step": 61377 }, { "epoch": 4.81, "learning_rate": 5.187171645887242e-05, "loss": 2.6262, "step": 61378 }, { "epoch": 4.81, "learning_rate": 5.187093232964792e-05, "loss": 2.6293, "step": 61379 }, { "epoch": 4.81, "learning_rate": 5.187014820042343e-05, "loss": 2.5975, "step": 61380 }, { "epoch": 4.81, "learning_rate": 5.1869364071198937e-05, "loss": 2.5988, "step": 61381 }, { "epoch": 4.81, "learning_rate": 5.186857994197444e-05, "loss": 2.5254, "step": 61382 }, { "epoch": 4.81, "learning_rate": 5.1867795812749944e-05, "loss": 2.4861, "step": 61383 }, { "epoch": 4.81, "learning_rate": 5.186701168352545e-05, "loss": 2.6032, "step": 61384 }, { "epoch": 4.81, "learning_rate": 5.186622755430095e-05, "loss": 2.5085, "step": 61385 }, { "epoch": 4.81, "learning_rate": 5.1865443425076455e-05, "loss": 2.646, "step": 61386 }, { "epoch": 4.81, "learning_rate": 5.186465929585196e-05, "loss": 2.6404, "step": 61387 }, { "epoch": 4.81, "learning_rate": 5.186387516662746e-05, "loss": 3.1012, "step": 61388 }, { "epoch": 4.81, "learning_rate": 5.1863091037402966e-05, "loss": 2.6112, "step": 61389 }, { "epoch": 4.81, "learning_rate": 5.186230690817847e-05, "loss": 2.6001, "step": 61390 }, { "epoch": 4.81, "learning_rate": 5.186152277895397e-05, "loss": 2.6233, "step": 61391 }, { "epoch": 4.81, "learning_rate": 5.186073864972948e-05, "loss": 2.6825, "step": 61392 }, { "epoch": 4.81, "learning_rate": 5.185995452050498e-05, "loss": 2.6053, "step": 61393 }, { "epoch": 4.81, "learning_rate": 5.1859170391280484e-05, "loss": 2.6839, "step": 61394 }, { "epoch": 4.81, "learning_rate": 5.185838626205599e-05, "loss": 2.6688, "step": 61395 }, { "epoch": 4.81, "learning_rate": 5.18576021328315e-05, "loss": 2.5517, "step": 61396 }, { "epoch": 4.81, "learning_rate": 5.1856818003607e-05, "loss": 2.8065, "step": 61397 }, { "epoch": 4.81, "learning_rate": 5.1856033874382506e-05, "loss": 2.6201, "step": 61398 }, { "epoch": 4.81, "learning_rate": 5.185524974515801e-05, "loss": 2.7949, "step": 61399 }, { "epoch": 4.81, "learning_rate": 5.185446561593351e-05, "loss": 2.6069, "step": 61400 }, { "epoch": 4.81, "learning_rate": 5.185368148670902e-05, "loss": 2.7412, "step": 61401 }, { "epoch": 4.81, "learning_rate": 5.185289735748452e-05, "loss": 2.9405, "step": 61402 }, { "epoch": 4.81, "learning_rate": 5.1852113228260024e-05, "loss": 2.663, "step": 61403 }, { "epoch": 4.81, "learning_rate": 5.185132909903553e-05, "loss": 2.7185, "step": 61404 }, { "epoch": 4.81, "learning_rate": 5.185054496981103e-05, "loss": 2.9315, "step": 61405 }, { "epoch": 4.82, "learning_rate": 5.1849760840586535e-05, "loss": 2.555, "step": 61406 }, { "epoch": 4.82, "learning_rate": 5.184897671136203e-05, "loss": 2.578, "step": 61407 }, { "epoch": 4.82, "learning_rate": 5.1848192582137536e-05, "loss": 2.7561, "step": 61408 }, { "epoch": 4.82, "learning_rate": 5.184740845291304e-05, "loss": 2.8602, "step": 61409 }, { "epoch": 4.82, "learning_rate": 5.184662432368854e-05, "loss": 2.6737, "step": 61410 }, { "epoch": 4.82, "learning_rate": 5.184584019446405e-05, "loss": 2.5164, "step": 61411 }, { "epoch": 4.82, "learning_rate": 5.184505606523955e-05, "loss": 2.8565, "step": 61412 }, { "epoch": 4.82, "learning_rate": 5.1844271936015054e-05, "loss": 2.6097, "step": 61413 }, { "epoch": 4.82, "learning_rate": 5.184348780679056e-05, "loss": 2.7953, "step": 61414 }, { "epoch": 4.82, "learning_rate": 5.184270367756606e-05, "loss": 2.6844, "step": 61415 }, { "epoch": 4.82, "learning_rate": 5.1841919548341565e-05, "loss": 2.5401, "step": 61416 }, { "epoch": 4.82, "learning_rate": 5.184113541911707e-05, "loss": 2.7444, "step": 61417 }, { "epoch": 4.82, "learning_rate": 5.184035128989257e-05, "loss": 2.6903, "step": 61418 }, { "epoch": 4.82, "learning_rate": 5.1839567160668076e-05, "loss": 2.7919, "step": 61419 }, { "epoch": 4.82, "learning_rate": 5.183878303144358e-05, "loss": 2.6962, "step": 61420 }, { "epoch": 4.82, "learning_rate": 5.1837998902219084e-05, "loss": 2.6584, "step": 61421 }, { "epoch": 4.82, "learning_rate": 5.1837214772994594e-05, "loss": 2.8737, "step": 61422 }, { "epoch": 4.82, "learning_rate": 5.18364306437701e-05, "loss": 2.7347, "step": 61423 }, { "epoch": 4.82, "learning_rate": 5.18356465145456e-05, "loss": 2.8252, "step": 61424 }, { "epoch": 4.82, "learning_rate": 5.1834862385321105e-05, "loss": 2.7152, "step": 61425 }, { "epoch": 4.82, "learning_rate": 5.183407825609661e-05, "loss": 2.6692, "step": 61426 }, { "epoch": 4.82, "learning_rate": 5.183329412687211e-05, "loss": 2.8613, "step": 61427 }, { "epoch": 4.82, "learning_rate": 5.1832509997647616e-05, "loss": 2.7822, "step": 61428 }, { "epoch": 4.82, "learning_rate": 5.183172586842312e-05, "loss": 2.6167, "step": 61429 }, { "epoch": 4.82, "learning_rate": 5.1830941739198624e-05, "loss": 2.8216, "step": 61430 }, { "epoch": 4.82, "learning_rate": 5.183015760997413e-05, "loss": 2.6943, "step": 61431 }, { "epoch": 4.82, "learning_rate": 5.182937348074963e-05, "loss": 2.6964, "step": 61432 }, { "epoch": 4.82, "learning_rate": 5.1828589351525135e-05, "loss": 2.8329, "step": 61433 }, { "epoch": 4.82, "learning_rate": 5.182780522230064e-05, "loss": 2.7374, "step": 61434 }, { "epoch": 4.82, "learning_rate": 5.182702109307614e-05, "loss": 2.5465, "step": 61435 }, { "epoch": 4.82, "learning_rate": 5.1826236963851646e-05, "loss": 2.7537, "step": 61436 }, { "epoch": 4.82, "learning_rate": 5.1825452834627156e-05, "loss": 2.7358, "step": 61437 }, { "epoch": 4.82, "learning_rate": 5.182466870540266e-05, "loss": 2.5962, "step": 61438 }, { "epoch": 4.82, "learning_rate": 5.1823884576178163e-05, "loss": 2.6454, "step": 61439 }, { "epoch": 4.82, "learning_rate": 5.182310044695367e-05, "loss": 2.6753, "step": 61440 }, { "epoch": 4.82, "learning_rate": 5.182231631772917e-05, "loss": 2.5578, "step": 61441 }, { "epoch": 4.82, "learning_rate": 5.1821532188504675e-05, "loss": 2.5286, "step": 61442 }, { "epoch": 4.82, "learning_rate": 5.1820748059280165e-05, "loss": 2.5457, "step": 61443 }, { "epoch": 4.82, "learning_rate": 5.181996393005567e-05, "loss": 2.7475, "step": 61444 }, { "epoch": 4.82, "learning_rate": 5.181917980083117e-05, "loss": 2.5747, "step": 61445 }, { "epoch": 4.82, "learning_rate": 5.1818395671606676e-05, "loss": 2.8328, "step": 61446 }, { "epoch": 4.82, "learning_rate": 5.1817611542382186e-05, "loss": 2.6171, "step": 61447 }, { "epoch": 4.82, "learning_rate": 5.181682741315769e-05, "loss": 2.8576, "step": 61448 }, { "epoch": 4.82, "learning_rate": 5.1816043283933193e-05, "loss": 2.6822, "step": 61449 }, { "epoch": 4.82, "learning_rate": 5.18152591547087e-05, "loss": 2.76, "step": 61450 }, { "epoch": 4.82, "learning_rate": 5.18144750254842e-05, "loss": 2.8727, "step": 61451 }, { "epoch": 4.82, "learning_rate": 5.1813690896259705e-05, "loss": 2.57, "step": 61452 }, { "epoch": 4.82, "learning_rate": 5.181290676703521e-05, "loss": 2.569, "step": 61453 }, { "epoch": 4.82, "learning_rate": 5.181212263781071e-05, "loss": 2.6156, "step": 61454 }, { "epoch": 4.82, "learning_rate": 5.1811338508586216e-05, "loss": 2.5671, "step": 61455 }, { "epoch": 4.82, "learning_rate": 5.181055437936172e-05, "loss": 2.9296, "step": 61456 }, { "epoch": 4.82, "learning_rate": 5.180977025013722e-05, "loss": 2.7042, "step": 61457 }, { "epoch": 4.82, "learning_rate": 5.1808986120912727e-05, "loss": 2.8084, "step": 61458 }, { "epoch": 4.82, "learning_rate": 5.180820199168823e-05, "loss": 2.5433, "step": 61459 }, { "epoch": 4.82, "learning_rate": 5.1807417862463734e-05, "loss": 2.6696, "step": 61460 }, { "epoch": 4.82, "learning_rate": 5.180663373323924e-05, "loss": 2.5834, "step": 61461 }, { "epoch": 4.82, "learning_rate": 5.180584960401474e-05, "loss": 2.6945, "step": 61462 }, { "epoch": 4.82, "learning_rate": 5.180506547479025e-05, "loss": 2.6797, "step": 61463 }, { "epoch": 4.82, "learning_rate": 5.1804281345565755e-05, "loss": 2.7923, "step": 61464 }, { "epoch": 4.82, "learning_rate": 5.180349721634126e-05, "loss": 2.6016, "step": 61465 }, { "epoch": 4.82, "learning_rate": 5.180271308711676e-05, "loss": 2.6532, "step": 61466 }, { "epoch": 4.82, "learning_rate": 5.1801928957892266e-05, "loss": 2.6767, "step": 61467 }, { "epoch": 4.82, "learning_rate": 5.180114482866777e-05, "loss": 2.6442, "step": 61468 }, { "epoch": 4.82, "learning_rate": 5.1800360699443274e-05, "loss": 2.6792, "step": 61469 }, { "epoch": 4.82, "learning_rate": 5.179957657021878e-05, "loss": 2.5824, "step": 61470 }, { "epoch": 4.82, "learning_rate": 5.179879244099428e-05, "loss": 2.7845, "step": 61471 }, { "epoch": 4.82, "learning_rate": 5.1798008311769785e-05, "loss": 2.6875, "step": 61472 }, { "epoch": 4.82, "learning_rate": 5.179722418254529e-05, "loss": 2.7897, "step": 61473 }, { "epoch": 4.82, "learning_rate": 5.179644005332079e-05, "loss": 2.5149, "step": 61474 }, { "epoch": 4.82, "learning_rate": 5.1795655924096296e-05, "loss": 2.6766, "step": 61475 }, { "epoch": 4.82, "learning_rate": 5.17948717948718e-05, "loss": 2.7261, "step": 61476 }, { "epoch": 4.82, "learning_rate": 5.17940876656473e-05, "loss": 2.5256, "step": 61477 }, { "epoch": 4.82, "learning_rate": 5.1793303536422814e-05, "loss": 2.7866, "step": 61478 }, { "epoch": 4.82, "learning_rate": 5.1792519407198304e-05, "loss": 2.6613, "step": 61479 }, { "epoch": 4.82, "learning_rate": 5.179173527797381e-05, "loss": 2.6912, "step": 61480 }, { "epoch": 4.82, "learning_rate": 5.179095114874931e-05, "loss": 2.5764, "step": 61481 }, { "epoch": 4.82, "learning_rate": 5.1790167019524815e-05, "loss": 2.6888, "step": 61482 }, { "epoch": 4.82, "learning_rate": 5.178938289030032e-05, "loss": 2.7137, "step": 61483 }, { "epoch": 4.82, "learning_rate": 5.178859876107582e-05, "loss": 2.5415, "step": 61484 }, { "epoch": 4.82, "learning_rate": 5.1787814631851326e-05, "loss": 2.5657, "step": 61485 }, { "epoch": 4.82, "learning_rate": 5.178703050262683e-05, "loss": 2.861, "step": 61486 }, { "epoch": 4.82, "learning_rate": 5.178624637340233e-05, "loss": 2.6802, "step": 61487 }, { "epoch": 4.82, "learning_rate": 5.1785462244177844e-05, "loss": 2.5426, "step": 61488 }, { "epoch": 4.82, "learning_rate": 5.178467811495335e-05, "loss": 2.836, "step": 61489 }, { "epoch": 4.82, "learning_rate": 5.178389398572885e-05, "loss": 2.7308, "step": 61490 }, { "epoch": 4.82, "learning_rate": 5.1783109856504355e-05, "loss": 2.7202, "step": 61491 }, { "epoch": 4.82, "learning_rate": 5.178232572727986e-05, "loss": 2.6539, "step": 61492 }, { "epoch": 4.82, "learning_rate": 5.178154159805536e-05, "loss": 2.9912, "step": 61493 }, { "epoch": 4.82, "learning_rate": 5.1780757468830866e-05, "loss": 2.8381, "step": 61494 }, { "epoch": 4.82, "learning_rate": 5.177997333960637e-05, "loss": 2.8319, "step": 61495 }, { "epoch": 4.82, "learning_rate": 5.177918921038187e-05, "loss": 2.6022, "step": 61496 }, { "epoch": 4.82, "learning_rate": 5.177840508115738e-05, "loss": 2.6544, "step": 61497 }, { "epoch": 4.82, "learning_rate": 5.177762095193288e-05, "loss": 2.7408, "step": 61498 }, { "epoch": 4.82, "learning_rate": 5.1776836822708384e-05, "loss": 2.8187, "step": 61499 }, { "epoch": 4.82, "learning_rate": 5.177605269348389e-05, "loss": 2.5956, "step": 61500 }, { "epoch": 4.82, "learning_rate": 5.177526856425939e-05, "loss": 2.7648, "step": 61501 }, { "epoch": 4.82, "learning_rate": 5.1774484435034895e-05, "loss": 2.5849, "step": 61502 }, { "epoch": 4.82, "learning_rate": 5.17737003058104e-05, "loss": 2.8293, "step": 61503 }, { "epoch": 4.82, "learning_rate": 5.177291617658591e-05, "loss": 2.6112, "step": 61504 }, { "epoch": 4.82, "learning_rate": 5.177213204736141e-05, "loss": 2.6214, "step": 61505 }, { "epoch": 4.82, "learning_rate": 5.177134791813692e-05, "loss": 2.5387, "step": 61506 }, { "epoch": 4.82, "learning_rate": 5.177056378891242e-05, "loss": 2.7724, "step": 61507 }, { "epoch": 4.82, "learning_rate": 5.1769779659687924e-05, "loss": 2.6422, "step": 61508 }, { "epoch": 4.82, "learning_rate": 5.176899553046343e-05, "loss": 2.5656, "step": 61509 }, { "epoch": 4.82, "learning_rate": 5.176821140123893e-05, "loss": 2.6266, "step": 61510 }, { "epoch": 4.82, "learning_rate": 5.1767427272014435e-05, "loss": 2.714, "step": 61511 }, { "epoch": 4.82, "learning_rate": 5.176664314278994e-05, "loss": 2.6433, "step": 61512 }, { "epoch": 4.82, "learning_rate": 5.176585901356544e-05, "loss": 2.4986, "step": 61513 }, { "epoch": 4.82, "learning_rate": 5.1765074884340946e-05, "loss": 2.7601, "step": 61514 }, { "epoch": 4.82, "learning_rate": 5.176429075511644e-05, "loss": 2.6031, "step": 61515 }, { "epoch": 4.82, "learning_rate": 5.176350662589195e-05, "loss": 2.7289, "step": 61516 }, { "epoch": 4.82, "learning_rate": 5.176272249666745e-05, "loss": 2.6085, "step": 61517 }, { "epoch": 4.82, "learning_rate": 5.1761938367442954e-05, "loss": 2.6655, "step": 61518 }, { "epoch": 4.82, "learning_rate": 5.176115423821846e-05, "loss": 2.6007, "step": 61519 }, { "epoch": 4.82, "learning_rate": 5.176037010899396e-05, "loss": 2.5964, "step": 61520 }, { "epoch": 4.82, "learning_rate": 5.1759585979769465e-05, "loss": 2.6137, "step": 61521 }, { "epoch": 4.82, "learning_rate": 5.175880185054497e-05, "loss": 2.6501, "step": 61522 }, { "epoch": 4.82, "learning_rate": 5.175801772132047e-05, "loss": 2.7889, "step": 61523 }, { "epoch": 4.82, "learning_rate": 5.1757233592095976e-05, "loss": 2.6179, "step": 61524 }, { "epoch": 4.82, "learning_rate": 5.175644946287148e-05, "loss": 2.5575, "step": 61525 }, { "epoch": 4.82, "learning_rate": 5.1755665333646983e-05, "loss": 2.7456, "step": 61526 }, { "epoch": 4.82, "learning_rate": 5.175488120442249e-05, "loss": 2.6068, "step": 61527 }, { "epoch": 4.82, "learning_rate": 5.175409707519799e-05, "loss": 2.5869, "step": 61528 }, { "epoch": 4.82, "learning_rate": 5.1753312945973494e-05, "loss": 2.5905, "step": 61529 }, { "epoch": 4.82, "learning_rate": 5.1752528816749005e-05, "loss": 2.8499, "step": 61530 }, { "epoch": 4.82, "learning_rate": 5.175174468752451e-05, "loss": 2.5615, "step": 61531 }, { "epoch": 4.82, "learning_rate": 5.175096055830001e-05, "loss": 2.4953, "step": 61532 }, { "epoch": 4.82, "learning_rate": 5.1750176429075516e-05, "loss": 2.3939, "step": 61533 }, { "epoch": 4.83, "learning_rate": 5.174939229985102e-05, "loss": 2.8274, "step": 61534 }, { "epoch": 4.83, "learning_rate": 5.174860817062652e-05, "loss": 2.549, "step": 61535 }, { "epoch": 4.83, "learning_rate": 5.174782404140203e-05, "loss": 2.838, "step": 61536 }, { "epoch": 4.83, "learning_rate": 5.174703991217753e-05, "loss": 2.7887, "step": 61537 }, { "epoch": 4.83, "learning_rate": 5.1746255782953034e-05, "loss": 2.4949, "step": 61538 }, { "epoch": 4.83, "learning_rate": 5.174547165372854e-05, "loss": 2.6418, "step": 61539 }, { "epoch": 4.83, "learning_rate": 5.174468752450404e-05, "loss": 2.5569, "step": 61540 }, { "epoch": 4.83, "learning_rate": 5.1743903395279545e-05, "loss": 2.765, "step": 61541 }, { "epoch": 4.83, "learning_rate": 5.174311926605505e-05, "loss": 2.5079, "step": 61542 }, { "epoch": 4.83, "learning_rate": 5.174233513683055e-05, "loss": 2.7033, "step": 61543 }, { "epoch": 4.83, "learning_rate": 5.1741551007606056e-05, "loss": 2.9052, "step": 61544 }, { "epoch": 4.83, "learning_rate": 5.174076687838157e-05, "loss": 2.7479, "step": 61545 }, { "epoch": 4.83, "learning_rate": 5.173998274915707e-05, "loss": 2.5104, "step": 61546 }, { "epoch": 4.83, "learning_rate": 5.1739198619932574e-05, "loss": 2.5842, "step": 61547 }, { "epoch": 4.83, "learning_rate": 5.173841449070808e-05, "loss": 2.9698, "step": 61548 }, { "epoch": 4.83, "learning_rate": 5.173763036148358e-05, "loss": 2.5881, "step": 61549 }, { "epoch": 4.83, "learning_rate": 5.1736846232259085e-05, "loss": 2.5376, "step": 61550 }, { "epoch": 4.83, "learning_rate": 5.1736062103034575e-05, "loss": 2.888, "step": 61551 }, { "epoch": 4.83, "learning_rate": 5.173527797381008e-05, "loss": 2.629, "step": 61552 }, { "epoch": 4.83, "learning_rate": 5.173449384458558e-05, "loss": 2.789, "step": 61553 }, { "epoch": 4.83, "learning_rate": 5.1733709715361086e-05, "loss": 2.9836, "step": 61554 }, { "epoch": 4.83, "learning_rate": 5.17329255861366e-05, "loss": 2.9173, "step": 61555 }, { "epoch": 4.83, "learning_rate": 5.17321414569121e-05, "loss": 2.7361, "step": 61556 }, { "epoch": 4.83, "learning_rate": 5.1731357327687604e-05, "loss": 2.7396, "step": 61557 }, { "epoch": 4.83, "learning_rate": 5.173057319846311e-05, "loss": 2.917, "step": 61558 }, { "epoch": 4.83, "learning_rate": 5.172978906923861e-05, "loss": 2.7909, "step": 61559 }, { "epoch": 4.83, "learning_rate": 5.1729004940014115e-05, "loss": 2.6027, "step": 61560 }, { "epoch": 4.83, "learning_rate": 5.172822081078962e-05, "loss": 2.8822, "step": 61561 }, { "epoch": 4.83, "learning_rate": 5.172743668156512e-05, "loss": 2.676, "step": 61562 }, { "epoch": 4.83, "learning_rate": 5.1726652552340626e-05, "loss": 2.7631, "step": 61563 }, { "epoch": 4.83, "learning_rate": 5.172586842311613e-05, "loss": 2.6001, "step": 61564 }, { "epoch": 4.83, "learning_rate": 5.1725084293891634e-05, "loss": 2.6499, "step": 61565 }, { "epoch": 4.83, "learning_rate": 5.172430016466714e-05, "loss": 2.9033, "step": 61566 }, { "epoch": 4.83, "learning_rate": 5.172351603544264e-05, "loss": 2.7876, "step": 61567 }, { "epoch": 4.83, "learning_rate": 5.1722731906218145e-05, "loss": 2.6825, "step": 61568 }, { "epoch": 4.83, "learning_rate": 5.172194777699365e-05, "loss": 2.8621, "step": 61569 }, { "epoch": 4.83, "learning_rate": 5.172116364776915e-05, "loss": 2.7392, "step": 61570 }, { "epoch": 4.83, "learning_rate": 5.172037951854466e-05, "loss": 2.6134, "step": 61571 }, { "epoch": 4.83, "learning_rate": 5.1719595389320166e-05, "loss": 2.7096, "step": 61572 }, { "epoch": 4.83, "learning_rate": 5.171881126009567e-05, "loss": 2.6975, "step": 61573 }, { "epoch": 4.83, "learning_rate": 5.1718027130871174e-05, "loss": 2.8204, "step": 61574 }, { "epoch": 4.83, "learning_rate": 5.171724300164668e-05, "loss": 2.5612, "step": 61575 }, { "epoch": 4.83, "learning_rate": 5.171645887242218e-05, "loss": 2.4536, "step": 61576 }, { "epoch": 4.83, "learning_rate": 5.1715674743197685e-05, "loss": 2.7594, "step": 61577 }, { "epoch": 4.83, "learning_rate": 5.171489061397319e-05, "loss": 2.6708, "step": 61578 }, { "epoch": 4.83, "learning_rate": 5.171410648474869e-05, "loss": 2.7256, "step": 61579 }, { "epoch": 4.83, "learning_rate": 5.1713322355524196e-05, "loss": 2.6432, "step": 61580 }, { "epoch": 4.83, "learning_rate": 5.17125382262997e-05, "loss": 2.617, "step": 61581 }, { "epoch": 4.83, "learning_rate": 5.17117540970752e-05, "loss": 2.746, "step": 61582 }, { "epoch": 4.83, "learning_rate": 5.1710969967850707e-05, "loss": 2.5431, "step": 61583 }, { "epoch": 4.83, "learning_rate": 5.171018583862621e-05, "loss": 2.8058, "step": 61584 }, { "epoch": 4.83, "learning_rate": 5.1709401709401714e-05, "loss": 2.7305, "step": 61585 }, { "epoch": 4.83, "learning_rate": 5.170861758017722e-05, "loss": 2.7102, "step": 61586 }, { "epoch": 4.83, "learning_rate": 5.1707833450952715e-05, "loss": 2.7646, "step": 61587 }, { "epoch": 4.83, "learning_rate": 5.170704932172822e-05, "loss": 2.8018, "step": 61588 }, { "epoch": 4.83, "learning_rate": 5.170626519250372e-05, "loss": 2.627, "step": 61589 }, { "epoch": 4.83, "learning_rate": 5.1705481063279226e-05, "loss": 2.7275, "step": 61590 }, { "epoch": 4.83, "learning_rate": 5.170469693405473e-05, "loss": 2.4677, "step": 61591 }, { "epoch": 4.83, "learning_rate": 5.170391280483023e-05, "loss": 2.7553, "step": 61592 }, { "epoch": 4.83, "learning_rate": 5.1703128675605737e-05, "loss": 2.6348, "step": 61593 }, { "epoch": 4.83, "learning_rate": 5.170234454638124e-05, "loss": 2.5979, "step": 61594 }, { "epoch": 4.83, "learning_rate": 5.1701560417156744e-05, "loss": 2.7119, "step": 61595 }, { "epoch": 4.83, "learning_rate": 5.1700776287932254e-05, "loss": 2.6503, "step": 61596 }, { "epoch": 4.83, "learning_rate": 5.169999215870776e-05, "loss": 2.7675, "step": 61597 }, { "epoch": 4.83, "learning_rate": 5.169920802948326e-05, "loss": 2.6062, "step": 61598 }, { "epoch": 4.83, "learning_rate": 5.1698423900258765e-05, "loss": 2.6798, "step": 61599 }, { "epoch": 4.83, "learning_rate": 5.169763977103427e-05, "loss": 2.693, "step": 61600 }, { "epoch": 4.83, "learning_rate": 5.169685564180977e-05, "loss": 2.7717, "step": 61601 }, { "epoch": 4.83, "learning_rate": 5.1696071512585276e-05, "loss": 2.6407, "step": 61602 }, { "epoch": 4.83, "learning_rate": 5.169528738336078e-05, "loss": 2.7638, "step": 61603 }, { "epoch": 4.83, "learning_rate": 5.1694503254136284e-05, "loss": 2.5329, "step": 61604 }, { "epoch": 4.83, "learning_rate": 5.169371912491179e-05, "loss": 2.8627, "step": 61605 }, { "epoch": 4.83, "learning_rate": 5.169293499568729e-05, "loss": 2.7503, "step": 61606 }, { "epoch": 4.83, "learning_rate": 5.1692150866462795e-05, "loss": 2.7674, "step": 61607 }, { "epoch": 4.83, "learning_rate": 5.16913667372383e-05, "loss": 2.8149, "step": 61608 }, { "epoch": 4.83, "learning_rate": 5.16905826080138e-05, "loss": 2.581, "step": 61609 }, { "epoch": 4.83, "learning_rate": 5.1689798478789306e-05, "loss": 2.659, "step": 61610 }, { "epoch": 4.83, "learning_rate": 5.168901434956481e-05, "loss": 2.6305, "step": 61611 }, { "epoch": 4.83, "learning_rate": 5.168823022034032e-05, "loss": 2.7189, "step": 61612 }, { "epoch": 4.83, "learning_rate": 5.1687446091115824e-05, "loss": 2.5217, "step": 61613 }, { "epoch": 4.83, "learning_rate": 5.168666196189133e-05, "loss": 2.693, "step": 61614 }, { "epoch": 4.83, "learning_rate": 5.168587783266683e-05, "loss": 2.9918, "step": 61615 }, { "epoch": 4.83, "learning_rate": 5.1685093703442335e-05, "loss": 2.6412, "step": 61616 }, { "epoch": 4.83, "learning_rate": 5.168430957421784e-05, "loss": 2.4787, "step": 61617 }, { "epoch": 4.83, "learning_rate": 5.168352544499334e-05, "loss": 2.8206, "step": 61618 }, { "epoch": 4.83, "learning_rate": 5.1682741315768846e-05, "loss": 2.6582, "step": 61619 }, { "epoch": 4.83, "learning_rate": 5.168195718654435e-05, "loss": 2.666, "step": 61620 }, { "epoch": 4.83, "learning_rate": 5.168117305731985e-05, "loss": 2.617, "step": 61621 }, { "epoch": 4.83, "learning_rate": 5.168038892809536e-05, "loss": 2.6357, "step": 61622 }, { "epoch": 4.83, "learning_rate": 5.1679604798870854e-05, "loss": 2.9226, "step": 61623 }, { "epoch": 4.83, "learning_rate": 5.167882066964636e-05, "loss": 2.5721, "step": 61624 }, { "epoch": 4.83, "learning_rate": 5.167803654042186e-05, "loss": 2.528, "step": 61625 }, { "epoch": 4.83, "learning_rate": 5.1677252411197365e-05, "loss": 2.658, "step": 61626 }, { "epoch": 4.83, "learning_rate": 5.167646828197287e-05, "loss": 2.6891, "step": 61627 }, { "epoch": 4.83, "learning_rate": 5.167568415274837e-05, "loss": 2.7204, "step": 61628 }, { "epoch": 4.83, "learning_rate": 5.1674900023523876e-05, "loss": 2.5917, "step": 61629 }, { "epoch": 4.83, "learning_rate": 5.167411589429938e-05, "loss": 2.5553, "step": 61630 }, { "epoch": 4.83, "learning_rate": 5.167333176507488e-05, "loss": 2.7727, "step": 61631 }, { "epoch": 4.83, "learning_rate": 5.167254763585039e-05, "loss": 2.7035, "step": 61632 }, { "epoch": 4.83, "learning_rate": 5.167176350662589e-05, "loss": 2.866, "step": 61633 }, { "epoch": 4.83, "learning_rate": 5.1670979377401394e-05, "loss": 2.6754, "step": 61634 }, { "epoch": 4.83, "learning_rate": 5.16701952481769e-05, "loss": 2.6314, "step": 61635 }, { "epoch": 4.83, "learning_rate": 5.16694111189524e-05, "loss": 2.7893, "step": 61636 }, { "epoch": 4.83, "learning_rate": 5.1668626989727905e-05, "loss": 2.7405, "step": 61637 }, { "epoch": 4.83, "learning_rate": 5.1667842860503416e-05, "loss": 2.4856, "step": 61638 }, { "epoch": 4.83, "learning_rate": 5.166705873127892e-05, "loss": 2.695, "step": 61639 }, { "epoch": 4.83, "learning_rate": 5.166627460205442e-05, "loss": 2.773, "step": 61640 }, { "epoch": 4.83, "learning_rate": 5.166549047282993e-05, "loss": 2.5397, "step": 61641 }, { "epoch": 4.83, "learning_rate": 5.166470634360543e-05, "loss": 2.6825, "step": 61642 }, { "epoch": 4.83, "learning_rate": 5.1663922214380934e-05, "loss": 2.6269, "step": 61643 }, { "epoch": 4.83, "learning_rate": 5.166313808515644e-05, "loss": 2.873, "step": 61644 }, { "epoch": 4.83, "learning_rate": 5.166235395593194e-05, "loss": 2.7366, "step": 61645 }, { "epoch": 4.83, "learning_rate": 5.1661569826707445e-05, "loss": 2.6427, "step": 61646 }, { "epoch": 4.83, "learning_rate": 5.166078569748295e-05, "loss": 2.8007, "step": 61647 }, { "epoch": 4.83, "learning_rate": 5.166000156825845e-05, "loss": 2.9423, "step": 61648 }, { "epoch": 4.83, "learning_rate": 5.1659217439033956e-05, "loss": 2.6497, "step": 61649 }, { "epoch": 4.83, "learning_rate": 5.165843330980946e-05, "loss": 2.5769, "step": 61650 }, { "epoch": 4.83, "learning_rate": 5.1657649180584963e-05, "loss": 2.6441, "step": 61651 }, { "epoch": 4.83, "learning_rate": 5.165686505136047e-05, "loss": 2.5864, "step": 61652 }, { "epoch": 4.83, "learning_rate": 5.165608092213598e-05, "loss": 2.6559, "step": 61653 }, { "epoch": 4.83, "learning_rate": 5.165529679291148e-05, "loss": 2.614, "step": 61654 }, { "epoch": 4.83, "learning_rate": 5.1654512663686985e-05, "loss": 2.5596, "step": 61655 }, { "epoch": 4.83, "learning_rate": 5.165372853446249e-05, "loss": 2.7449, "step": 61656 }, { "epoch": 4.83, "learning_rate": 5.165294440523799e-05, "loss": 2.6387, "step": 61657 }, { "epoch": 4.83, "learning_rate": 5.1652160276013496e-05, "loss": 2.7285, "step": 61658 }, { "epoch": 4.83, "learning_rate": 5.1651376146788986e-05, "loss": 2.6923, "step": 61659 }, { "epoch": 4.83, "learning_rate": 5.165059201756449e-05, "loss": 2.7961, "step": 61660 }, { "epoch": 4.84, "learning_rate": 5.1649807888339993e-05, "loss": 2.5229, "step": 61661 }, { "epoch": 4.84, "learning_rate": 5.16490237591155e-05, "loss": 3.0267, "step": 61662 }, { "epoch": 4.84, "learning_rate": 5.164823962989101e-05, "loss": 2.5603, "step": 61663 }, { "epoch": 4.84, "learning_rate": 5.164745550066651e-05, "loss": 2.5077, "step": 61664 }, { "epoch": 4.84, "learning_rate": 5.1646671371442015e-05, "loss": 2.5211, "step": 61665 }, { "epoch": 4.84, "learning_rate": 5.164588724221752e-05, "loss": 2.719, "step": 61666 }, { "epoch": 4.84, "learning_rate": 5.164510311299302e-05, "loss": 2.7475, "step": 61667 }, { "epoch": 4.84, "learning_rate": 5.1644318983768526e-05, "loss": 2.7127, "step": 61668 }, { "epoch": 4.84, "learning_rate": 5.164353485454403e-05, "loss": 2.5684, "step": 61669 }, { "epoch": 4.84, "learning_rate": 5.164275072531953e-05, "loss": 2.545, "step": 61670 }, { "epoch": 4.84, "learning_rate": 5.164196659609504e-05, "loss": 2.6302, "step": 61671 }, { "epoch": 4.84, "learning_rate": 5.164118246687054e-05, "loss": 2.667, "step": 61672 }, { "epoch": 4.84, "learning_rate": 5.1640398337646044e-05, "loss": 2.5677, "step": 61673 }, { "epoch": 4.84, "learning_rate": 5.163961420842155e-05, "loss": 2.5668, "step": 61674 }, { "epoch": 4.84, "learning_rate": 5.163883007919705e-05, "loss": 2.8516, "step": 61675 }, { "epoch": 4.84, "learning_rate": 5.1638045949972555e-05, "loss": 2.3281, "step": 61676 }, { "epoch": 4.84, "learning_rate": 5.163726182074806e-05, "loss": 2.7556, "step": 61677 }, { "epoch": 4.84, "learning_rate": 5.163647769152356e-05, "loss": 2.6322, "step": 61678 }, { "epoch": 4.84, "learning_rate": 5.163569356229907e-05, "loss": 2.8752, "step": 61679 }, { "epoch": 4.84, "learning_rate": 5.163490943307458e-05, "loss": 2.6773, "step": 61680 }, { "epoch": 4.84, "learning_rate": 5.163412530385008e-05, "loss": 2.8135, "step": 61681 }, { "epoch": 4.84, "learning_rate": 5.1633341174625584e-05, "loss": 2.5702, "step": 61682 }, { "epoch": 4.84, "learning_rate": 5.163255704540109e-05, "loss": 2.6234, "step": 61683 }, { "epoch": 4.84, "learning_rate": 5.163177291617659e-05, "loss": 2.8522, "step": 61684 }, { "epoch": 4.84, "learning_rate": 5.1630988786952095e-05, "loss": 2.5738, "step": 61685 }, { "epoch": 4.84, "learning_rate": 5.16302046577276e-05, "loss": 2.6268, "step": 61686 }, { "epoch": 4.84, "learning_rate": 5.16294205285031e-05, "loss": 2.7208, "step": 61687 }, { "epoch": 4.84, "learning_rate": 5.1628636399278606e-05, "loss": 2.5826, "step": 61688 }, { "epoch": 4.84, "learning_rate": 5.162785227005411e-05, "loss": 2.5294, "step": 61689 }, { "epoch": 4.84, "learning_rate": 5.1627068140829614e-05, "loss": 2.4743, "step": 61690 }, { "epoch": 4.84, "learning_rate": 5.162628401160512e-05, "loss": 2.6796, "step": 61691 }, { "epoch": 4.84, "learning_rate": 5.162549988238062e-05, "loss": 2.8324, "step": 61692 }, { "epoch": 4.84, "learning_rate": 5.1624715753156125e-05, "loss": 2.5515, "step": 61693 }, { "epoch": 4.84, "learning_rate": 5.162393162393163e-05, "loss": 2.6435, "step": 61694 }, { "epoch": 4.84, "learning_rate": 5.1623147494707125e-05, "loss": 2.9744, "step": 61695 }, { "epoch": 4.84, "learning_rate": 5.162236336548263e-05, "loss": 2.5838, "step": 61696 }, { "epoch": 4.84, "learning_rate": 5.162157923625813e-05, "loss": 2.7978, "step": 61697 }, { "epoch": 4.84, "learning_rate": 5.1620795107033636e-05, "loss": 2.6528, "step": 61698 }, { "epoch": 4.84, "learning_rate": 5.162001097780914e-05, "loss": 2.8964, "step": 61699 }, { "epoch": 4.84, "learning_rate": 5.1619226848584644e-05, "loss": 2.7715, "step": 61700 }, { "epoch": 4.84, "learning_rate": 5.161844271936015e-05, "loss": 2.7005, "step": 61701 }, { "epoch": 4.84, "learning_rate": 5.161765859013565e-05, "loss": 2.867, "step": 61702 }, { "epoch": 4.84, "learning_rate": 5.1616874460911155e-05, "loss": 2.9031, "step": 61703 }, { "epoch": 4.84, "learning_rate": 5.161609033168666e-05, "loss": 2.5996, "step": 61704 }, { "epoch": 4.84, "learning_rate": 5.161530620246217e-05, "loss": 2.7111, "step": 61705 }, { "epoch": 4.84, "learning_rate": 5.161452207323767e-05, "loss": 2.8716, "step": 61706 }, { "epoch": 4.84, "learning_rate": 5.1613737944013176e-05, "loss": 2.6077, "step": 61707 }, { "epoch": 4.84, "learning_rate": 5.161295381478868e-05, "loss": 2.7321, "step": 61708 }, { "epoch": 4.84, "learning_rate": 5.1612169685564184e-05, "loss": 2.5083, "step": 61709 }, { "epoch": 4.84, "learning_rate": 5.161138555633969e-05, "loss": 2.581, "step": 61710 }, { "epoch": 4.84, "learning_rate": 5.161060142711519e-05, "loss": 2.8048, "step": 61711 }, { "epoch": 4.84, "learning_rate": 5.1609817297890695e-05, "loss": 2.5954, "step": 61712 }, { "epoch": 4.84, "learning_rate": 5.16090331686662e-05, "loss": 2.8071, "step": 61713 }, { "epoch": 4.84, "learning_rate": 5.16082490394417e-05, "loss": 2.4896, "step": 61714 }, { "epoch": 4.84, "learning_rate": 5.1607464910217206e-05, "loss": 2.5688, "step": 61715 }, { "epoch": 4.84, "learning_rate": 5.160668078099271e-05, "loss": 2.4924, "step": 61716 }, { "epoch": 4.84, "learning_rate": 5.160589665176821e-05, "loss": 2.7197, "step": 61717 }, { "epoch": 4.84, "learning_rate": 5.160511252254372e-05, "loss": 2.7912, "step": 61718 }, { "epoch": 4.84, "learning_rate": 5.160432839331922e-05, "loss": 2.6533, "step": 61719 }, { "epoch": 4.84, "learning_rate": 5.160354426409473e-05, "loss": 2.5594, "step": 61720 }, { "epoch": 4.84, "learning_rate": 5.1602760134870234e-05, "loss": 2.5942, "step": 61721 }, { "epoch": 4.84, "learning_rate": 5.160197600564574e-05, "loss": 2.5364, "step": 61722 }, { "epoch": 4.84, "learning_rate": 5.160119187642124e-05, "loss": 2.6924, "step": 61723 }, { "epoch": 4.84, "learning_rate": 5.1600407747196745e-05, "loss": 2.601, "step": 61724 }, { "epoch": 4.84, "learning_rate": 5.159962361797225e-05, "loss": 2.6354, "step": 61725 }, { "epoch": 4.84, "learning_rate": 5.159883948874775e-05, "loss": 2.8107, "step": 61726 }, { "epoch": 4.84, "learning_rate": 5.1598055359523257e-05, "loss": 2.6056, "step": 61727 }, { "epoch": 4.84, "learning_rate": 5.159727123029876e-05, "loss": 2.8903, "step": 61728 }, { "epoch": 4.84, "learning_rate": 5.1596487101074264e-05, "loss": 3.0318, "step": 61729 }, { "epoch": 4.84, "learning_rate": 5.159570297184977e-05, "loss": 2.4658, "step": 61730 }, { "epoch": 4.84, "learning_rate": 5.1594918842625264e-05, "loss": 2.63, "step": 61731 }, { "epoch": 4.84, "learning_rate": 5.159413471340077e-05, "loss": 2.6539, "step": 61732 }, { "epoch": 4.84, "learning_rate": 5.159335058417627e-05, "loss": 2.7607, "step": 61733 }, { "epoch": 4.84, "learning_rate": 5.1592566454951775e-05, "loss": 2.6628, "step": 61734 }, { "epoch": 4.84, "learning_rate": 5.159178232572728e-05, "loss": 2.4968, "step": 61735 }, { "epoch": 4.84, "learning_rate": 5.159099819650278e-05, "loss": 2.5678, "step": 61736 }, { "epoch": 4.84, "learning_rate": 5.1590214067278287e-05, "loss": 2.6559, "step": 61737 }, { "epoch": 4.84, "learning_rate": 5.158942993805379e-05, "loss": 2.5936, "step": 61738 }, { "epoch": 4.84, "learning_rate": 5.1588645808829294e-05, "loss": 2.6324, "step": 61739 }, { "epoch": 4.84, "learning_rate": 5.15878616796048e-05, "loss": 2.4904, "step": 61740 }, { "epoch": 4.84, "learning_rate": 5.15870775503803e-05, "loss": 2.9316, "step": 61741 }, { "epoch": 4.84, "learning_rate": 5.1586293421155805e-05, "loss": 2.4971, "step": 61742 }, { "epoch": 4.84, "learning_rate": 5.158550929193131e-05, "loss": 2.5186, "step": 61743 }, { "epoch": 4.84, "learning_rate": 5.158472516270681e-05, "loss": 2.5147, "step": 61744 }, { "epoch": 4.84, "learning_rate": 5.1583941033482316e-05, "loss": 2.7019, "step": 61745 }, { "epoch": 4.84, "learning_rate": 5.1583156904257826e-05, "loss": 2.7098, "step": 61746 }, { "epoch": 4.84, "learning_rate": 5.158237277503333e-05, "loss": 2.624, "step": 61747 }, { "epoch": 4.84, "learning_rate": 5.1581588645808834e-05, "loss": 2.5013, "step": 61748 }, { "epoch": 4.84, "learning_rate": 5.158080451658434e-05, "loss": 2.7721, "step": 61749 }, { "epoch": 4.84, "learning_rate": 5.158002038735984e-05, "loss": 2.7999, "step": 61750 }, { "epoch": 4.84, "learning_rate": 5.1579236258135345e-05, "loss": 2.6051, "step": 61751 }, { "epoch": 4.84, "learning_rate": 5.157845212891085e-05, "loss": 2.5746, "step": 61752 }, { "epoch": 4.84, "learning_rate": 5.157766799968635e-05, "loss": 2.7126, "step": 61753 }, { "epoch": 4.84, "learning_rate": 5.1576883870461856e-05, "loss": 2.833, "step": 61754 }, { "epoch": 4.84, "learning_rate": 5.157609974123736e-05, "loss": 2.9679, "step": 61755 }, { "epoch": 4.84, "learning_rate": 5.157531561201286e-05, "loss": 2.7119, "step": 61756 }, { "epoch": 4.84, "learning_rate": 5.157453148278837e-05, "loss": 2.6565, "step": 61757 }, { "epoch": 4.84, "learning_rate": 5.157374735356387e-05, "loss": 2.5695, "step": 61758 }, { "epoch": 4.84, "learning_rate": 5.1572963224339374e-05, "loss": 2.5258, "step": 61759 }, { "epoch": 4.84, "learning_rate": 5.157217909511488e-05, "loss": 2.6608, "step": 61760 }, { "epoch": 4.84, "learning_rate": 5.157139496589039e-05, "loss": 2.6646, "step": 61761 }, { "epoch": 4.84, "learning_rate": 5.157061083666589e-05, "loss": 2.5781, "step": 61762 }, { "epoch": 4.84, "learning_rate": 5.1569826707441396e-05, "loss": 2.5501, "step": 61763 }, { "epoch": 4.84, "learning_rate": 5.15690425782169e-05, "loss": 2.6197, "step": 61764 }, { "epoch": 4.84, "learning_rate": 5.15682584489924e-05, "loss": 2.6565, "step": 61765 }, { "epoch": 4.84, "learning_rate": 5.156747431976791e-05, "loss": 2.5976, "step": 61766 }, { "epoch": 4.84, "learning_rate": 5.15666901905434e-05, "loss": 2.8121, "step": 61767 }, { "epoch": 4.84, "learning_rate": 5.15659060613189e-05, "loss": 2.624, "step": 61768 }, { "epoch": 4.84, "learning_rate": 5.1565121932094404e-05, "loss": 2.4941, "step": 61769 }, { "epoch": 4.84, "learning_rate": 5.156433780286991e-05, "loss": 2.7159, "step": 61770 }, { "epoch": 4.84, "learning_rate": 5.156355367364542e-05, "loss": 2.5834, "step": 61771 }, { "epoch": 4.84, "learning_rate": 5.156276954442092e-05, "loss": 2.6139, "step": 61772 }, { "epoch": 4.84, "learning_rate": 5.1561985415196426e-05, "loss": 2.6859, "step": 61773 }, { "epoch": 4.84, "learning_rate": 5.156120128597193e-05, "loss": 2.7411, "step": 61774 }, { "epoch": 4.84, "learning_rate": 5.156041715674743e-05, "loss": 2.7333, "step": 61775 }, { "epoch": 4.84, "learning_rate": 5.155963302752294e-05, "loss": 2.6049, "step": 61776 }, { "epoch": 4.84, "learning_rate": 5.155884889829844e-05, "loss": 2.7538, "step": 61777 }, { "epoch": 4.84, "learning_rate": 5.1558064769073944e-05, "loss": 2.5794, "step": 61778 }, { "epoch": 4.84, "learning_rate": 5.155728063984945e-05, "loss": 2.6098, "step": 61779 }, { "epoch": 4.84, "learning_rate": 5.155649651062495e-05, "loss": 2.7154, "step": 61780 }, { "epoch": 4.84, "learning_rate": 5.1555712381400455e-05, "loss": 2.6749, "step": 61781 }, { "epoch": 4.84, "learning_rate": 5.155492825217596e-05, "loss": 2.6619, "step": 61782 }, { "epoch": 4.84, "learning_rate": 5.155414412295146e-05, "loss": 2.6432, "step": 61783 }, { "epoch": 4.84, "learning_rate": 5.1553359993726966e-05, "loss": 2.7738, "step": 61784 }, { "epoch": 4.84, "learning_rate": 5.155257586450247e-05, "loss": 2.6123, "step": 61785 }, { "epoch": 4.84, "learning_rate": 5.1551791735277974e-05, "loss": 2.6588, "step": 61786 }, { "epoch": 4.84, "learning_rate": 5.1551007606053484e-05, "loss": 2.7702, "step": 61787 }, { "epoch": 4.84, "learning_rate": 5.155022347682899e-05, "loss": 2.709, "step": 61788 }, { "epoch": 4.85, "learning_rate": 5.154943934760449e-05, "loss": 2.6309, "step": 61789 }, { "epoch": 4.85, "learning_rate": 5.1548655218379995e-05, "loss": 2.7742, "step": 61790 }, { "epoch": 4.85, "learning_rate": 5.15478710891555e-05, "loss": 2.5573, "step": 61791 }, { "epoch": 4.85, "learning_rate": 5.1547086959931e-05, "loss": 2.4791, "step": 61792 }, { "epoch": 4.85, "learning_rate": 5.1546302830706506e-05, "loss": 2.7198, "step": 61793 }, { "epoch": 4.85, "learning_rate": 5.154551870148201e-05, "loss": 2.6211, "step": 61794 }, { "epoch": 4.85, "learning_rate": 5.154473457225751e-05, "loss": 2.8429, "step": 61795 }, { "epoch": 4.85, "learning_rate": 5.154395044303302e-05, "loss": 2.91, "step": 61796 }, { "epoch": 4.85, "learning_rate": 5.154316631380852e-05, "loss": 2.6437, "step": 61797 }, { "epoch": 4.85, "learning_rate": 5.1542382184584024e-05, "loss": 2.5398, "step": 61798 }, { "epoch": 4.85, "learning_rate": 5.154159805535953e-05, "loss": 2.6816, "step": 61799 }, { "epoch": 4.85, "learning_rate": 5.154081392613503e-05, "loss": 2.5841, "step": 61800 }, { "epoch": 4.85, "learning_rate": 5.1540029796910535e-05, "loss": 2.7076, "step": 61801 }, { "epoch": 4.85, "learning_rate": 5.153924566768604e-05, "loss": 2.532, "step": 61802 }, { "epoch": 4.85, "learning_rate": 5.1538461538461536e-05, "loss": 2.7731, "step": 61803 }, { "epoch": 4.85, "learning_rate": 5.153767740923704e-05, "loss": 2.8068, "step": 61804 }, { "epoch": 4.85, "learning_rate": 5.153689328001254e-05, "loss": 2.7118, "step": 61805 }, { "epoch": 4.85, "learning_rate": 5.153610915078805e-05, "loss": 2.7916, "step": 61806 }, { "epoch": 4.85, "learning_rate": 5.153532502156355e-05, "loss": 2.6427, "step": 61807 }, { "epoch": 4.85, "learning_rate": 5.1534540892339054e-05, "loss": 2.6137, "step": 61808 }, { "epoch": 4.85, "learning_rate": 5.153375676311456e-05, "loss": 2.7905, "step": 61809 }, { "epoch": 4.85, "learning_rate": 5.153297263389006e-05, "loss": 2.9106, "step": 61810 }, { "epoch": 4.85, "learning_rate": 5.1532188504665565e-05, "loss": 2.5926, "step": 61811 }, { "epoch": 4.85, "learning_rate": 5.153140437544107e-05, "loss": 2.8676, "step": 61812 }, { "epoch": 4.85, "learning_rate": 5.153062024621658e-05, "loss": 2.7989, "step": 61813 }, { "epoch": 4.85, "learning_rate": 5.152983611699208e-05, "loss": 2.6529, "step": 61814 }, { "epoch": 4.85, "learning_rate": 5.152905198776759e-05, "loss": 2.6708, "step": 61815 }, { "epoch": 4.85, "learning_rate": 5.152826785854309e-05, "loss": 2.5848, "step": 61816 }, { "epoch": 4.85, "learning_rate": 5.1527483729318594e-05, "loss": 2.7723, "step": 61817 }, { "epoch": 4.85, "learning_rate": 5.15266996000941e-05, "loss": 2.5148, "step": 61818 }, { "epoch": 4.85, "learning_rate": 5.15259154708696e-05, "loss": 2.7723, "step": 61819 }, { "epoch": 4.85, "learning_rate": 5.1525131341645105e-05, "loss": 2.7521, "step": 61820 }, { "epoch": 4.85, "learning_rate": 5.152434721242061e-05, "loss": 2.6543, "step": 61821 }, { "epoch": 4.85, "learning_rate": 5.152356308319611e-05, "loss": 2.695, "step": 61822 }, { "epoch": 4.85, "learning_rate": 5.1522778953971616e-05, "loss": 2.6347, "step": 61823 }, { "epoch": 4.85, "learning_rate": 5.152199482474712e-05, "loss": 2.6784, "step": 61824 }, { "epoch": 4.85, "learning_rate": 5.1521210695522624e-05, "loss": 2.5931, "step": 61825 }, { "epoch": 4.85, "learning_rate": 5.152042656629813e-05, "loss": 2.6055, "step": 61826 }, { "epoch": 4.85, "learning_rate": 5.151964243707363e-05, "loss": 2.4747, "step": 61827 }, { "epoch": 4.85, "learning_rate": 5.151885830784914e-05, "loss": 2.6933, "step": 61828 }, { "epoch": 4.85, "learning_rate": 5.1518074178624645e-05, "loss": 2.5465, "step": 61829 }, { "epoch": 4.85, "learning_rate": 5.151729004940015e-05, "loss": 2.8858, "step": 61830 }, { "epoch": 4.85, "learning_rate": 5.151650592017565e-05, "loss": 2.6978, "step": 61831 }, { "epoch": 4.85, "learning_rate": 5.1515721790951156e-05, "loss": 2.7627, "step": 61832 }, { "epoch": 4.85, "learning_rate": 5.151493766172666e-05, "loss": 2.6655, "step": 61833 }, { "epoch": 4.85, "learning_rate": 5.1514153532502164e-05, "loss": 2.8828, "step": 61834 }, { "epoch": 4.85, "learning_rate": 5.151336940327767e-05, "loss": 2.7117, "step": 61835 }, { "epoch": 4.85, "learning_rate": 5.151258527405317e-05, "loss": 2.6254, "step": 61836 }, { "epoch": 4.85, "learning_rate": 5.1511801144828675e-05, "loss": 2.5646, "step": 61837 }, { "epoch": 4.85, "learning_rate": 5.151101701560418e-05, "loss": 2.7356, "step": 61838 }, { "epoch": 4.85, "learning_rate": 5.1510232886379675e-05, "loss": 2.6062, "step": 61839 }, { "epoch": 4.85, "learning_rate": 5.150944875715518e-05, "loss": 2.7408, "step": 61840 }, { "epoch": 4.85, "learning_rate": 5.150866462793068e-05, "loss": 2.3697, "step": 61841 }, { "epoch": 4.85, "learning_rate": 5.1507880498706186e-05, "loss": 2.6752, "step": 61842 }, { "epoch": 4.85, "learning_rate": 5.150709636948169e-05, "loss": 2.8216, "step": 61843 }, { "epoch": 4.85, "learning_rate": 5.1506312240257194e-05, "loss": 3.0383, "step": 61844 }, { "epoch": 4.85, "learning_rate": 5.15055281110327e-05, "loss": 3.0453, "step": 61845 }, { "epoch": 4.85, "learning_rate": 5.15047439818082e-05, "loss": 2.6612, "step": 61846 }, { "epoch": 4.85, "learning_rate": 5.1503959852583705e-05, "loss": 2.6164, "step": 61847 }, { "epoch": 4.85, "learning_rate": 5.150317572335921e-05, "loss": 2.4734, "step": 61848 }, { "epoch": 4.85, "learning_rate": 5.150239159413471e-05, "loss": 2.5108, "step": 61849 }, { "epoch": 4.85, "learning_rate": 5.1501607464910216e-05, "loss": 2.8332, "step": 61850 }, { "epoch": 4.85, "learning_rate": 5.150082333568572e-05, "loss": 2.6006, "step": 61851 }, { "epoch": 4.85, "learning_rate": 5.150003920646122e-05, "loss": 2.6322, "step": 61852 }, { "epoch": 4.85, "learning_rate": 5.149925507723673e-05, "loss": 2.6616, "step": 61853 }, { "epoch": 4.85, "learning_rate": 5.149847094801224e-05, "loss": 2.494, "step": 61854 }, { "epoch": 4.85, "learning_rate": 5.149768681878774e-05, "loss": 2.7277, "step": 61855 }, { "epoch": 4.85, "learning_rate": 5.1496902689563245e-05, "loss": 2.6029, "step": 61856 }, { "epoch": 4.85, "learning_rate": 5.149611856033875e-05, "loss": 2.4912, "step": 61857 }, { "epoch": 4.85, "learning_rate": 5.149533443111425e-05, "loss": 2.634, "step": 61858 }, { "epoch": 4.85, "learning_rate": 5.1494550301889756e-05, "loss": 2.6216, "step": 61859 }, { "epoch": 4.85, "learning_rate": 5.149376617266526e-05, "loss": 2.5609, "step": 61860 }, { "epoch": 4.85, "learning_rate": 5.149298204344076e-05, "loss": 2.8165, "step": 61861 }, { "epoch": 4.85, "learning_rate": 5.1492197914216267e-05, "loss": 2.5852, "step": 61862 }, { "epoch": 4.85, "learning_rate": 5.149141378499177e-05, "loss": 2.5228, "step": 61863 }, { "epoch": 4.85, "learning_rate": 5.1490629655767274e-05, "loss": 2.6113, "step": 61864 }, { "epoch": 4.85, "learning_rate": 5.148984552654278e-05, "loss": 2.8829, "step": 61865 }, { "epoch": 4.85, "learning_rate": 5.148906139731828e-05, "loss": 2.589, "step": 61866 }, { "epoch": 4.85, "learning_rate": 5.1488277268093785e-05, "loss": 2.5568, "step": 61867 }, { "epoch": 4.85, "learning_rate": 5.148749313886929e-05, "loss": 2.7629, "step": 61868 }, { "epoch": 4.85, "learning_rate": 5.14867090096448e-05, "loss": 2.6456, "step": 61869 }, { "epoch": 4.85, "learning_rate": 5.14859248804203e-05, "loss": 2.7578, "step": 61870 }, { "epoch": 4.85, "learning_rate": 5.1485140751195806e-05, "loss": 2.7263, "step": 61871 }, { "epoch": 4.85, "learning_rate": 5.148435662197131e-05, "loss": 2.6995, "step": 61872 }, { "epoch": 4.85, "learning_rate": 5.1483572492746814e-05, "loss": 2.5768, "step": 61873 }, { "epoch": 4.85, "learning_rate": 5.148278836352232e-05, "loss": 2.6628, "step": 61874 }, { "epoch": 4.85, "learning_rate": 5.148200423429781e-05, "loss": 2.6418, "step": 61875 }, { "epoch": 4.85, "learning_rate": 5.148122010507331e-05, "loss": 2.653, "step": 61876 }, { "epoch": 4.85, "learning_rate": 5.1480435975848815e-05, "loss": 2.5329, "step": 61877 }, { "epoch": 4.85, "learning_rate": 5.147965184662432e-05, "loss": 2.5893, "step": 61878 }, { "epoch": 4.85, "learning_rate": 5.147886771739983e-05, "loss": 2.8034, "step": 61879 }, { "epoch": 4.85, "learning_rate": 5.147808358817533e-05, "loss": 2.5258, "step": 61880 }, { "epoch": 4.85, "learning_rate": 5.1477299458950836e-05, "loss": 2.756, "step": 61881 }, { "epoch": 4.85, "learning_rate": 5.147651532972634e-05, "loss": 2.6139, "step": 61882 }, { "epoch": 4.85, "learning_rate": 5.1475731200501844e-05, "loss": 2.5075, "step": 61883 }, { "epoch": 4.85, "learning_rate": 5.147494707127735e-05, "loss": 2.77, "step": 61884 }, { "epoch": 4.85, "learning_rate": 5.147416294205285e-05, "loss": 2.709, "step": 61885 }, { "epoch": 4.85, "learning_rate": 5.1473378812828355e-05, "loss": 2.7309, "step": 61886 }, { "epoch": 4.85, "learning_rate": 5.147259468360386e-05, "loss": 2.8116, "step": 61887 }, { "epoch": 4.85, "learning_rate": 5.147181055437936e-05, "loss": 2.7534, "step": 61888 }, { "epoch": 4.85, "learning_rate": 5.1471026425154866e-05, "loss": 2.7402, "step": 61889 }, { "epoch": 4.85, "learning_rate": 5.147024229593037e-05, "loss": 2.8311, "step": 61890 }, { "epoch": 4.85, "learning_rate": 5.146945816670587e-05, "loss": 2.6392, "step": 61891 }, { "epoch": 4.85, "learning_rate": 5.146867403748138e-05, "loss": 2.6115, "step": 61892 }, { "epoch": 4.85, "learning_rate": 5.146788990825688e-05, "loss": 2.6942, "step": 61893 }, { "epoch": 4.85, "learning_rate": 5.1467105779032384e-05, "loss": 2.6593, "step": 61894 }, { "epoch": 4.85, "learning_rate": 5.1466321649807895e-05, "loss": 2.6899, "step": 61895 }, { "epoch": 4.85, "learning_rate": 5.14655375205834e-05, "loss": 2.7292, "step": 61896 }, { "epoch": 4.85, "learning_rate": 5.14647533913589e-05, "loss": 2.6732, "step": 61897 }, { "epoch": 4.85, "learning_rate": 5.1463969262134406e-05, "loss": 2.7558, "step": 61898 }, { "epoch": 4.85, "learning_rate": 5.146318513290991e-05, "loss": 2.5607, "step": 61899 }, { "epoch": 4.85, "learning_rate": 5.146240100368541e-05, "loss": 2.6745, "step": 61900 }, { "epoch": 4.85, "learning_rate": 5.146161687446092e-05, "loss": 2.5585, "step": 61901 }, { "epoch": 4.85, "learning_rate": 5.146083274523642e-05, "loss": 2.5797, "step": 61902 }, { "epoch": 4.85, "learning_rate": 5.1460048616011924e-05, "loss": 2.6899, "step": 61903 }, { "epoch": 4.85, "learning_rate": 5.145926448678743e-05, "loss": 2.7071, "step": 61904 }, { "epoch": 4.85, "learning_rate": 5.145848035756293e-05, "loss": 2.9954, "step": 61905 }, { "epoch": 4.85, "learning_rate": 5.1457696228338435e-05, "loss": 2.5491, "step": 61906 }, { "epoch": 4.85, "learning_rate": 5.145691209911394e-05, "loss": 2.7954, "step": 61907 }, { "epoch": 4.85, "learning_rate": 5.145612796988944e-05, "loss": 2.6603, "step": 61908 }, { "epoch": 4.85, "learning_rate": 5.1455343840664946e-05, "loss": 2.7458, "step": 61909 }, { "epoch": 4.85, "learning_rate": 5.145455971144045e-05, "loss": 2.47, "step": 61910 }, { "epoch": 4.85, "learning_rate": 5.145377558221595e-05, "loss": 2.8182, "step": 61911 }, { "epoch": 4.85, "learning_rate": 5.145299145299145e-05, "loss": 2.6244, "step": 61912 }, { "epoch": 4.85, "learning_rate": 5.1452207323766954e-05, "loss": 2.4768, "step": 61913 }, { "epoch": 4.85, "learning_rate": 5.145142319454246e-05, "loss": 2.6322, "step": 61914 }, { "epoch": 4.85, "learning_rate": 5.145063906531796e-05, "loss": 2.7859, "step": 61915 }, { "epoch": 4.86, "learning_rate": 5.1449854936093465e-05, "loss": 2.6438, "step": 61916 }, { "epoch": 4.86, "learning_rate": 5.144907080686897e-05, "loss": 2.9555, "step": 61917 }, { "epoch": 4.86, "learning_rate": 5.144828667764447e-05, "loss": 2.4131, "step": 61918 }, { "epoch": 4.86, "learning_rate": 5.1447502548419976e-05, "loss": 2.6585, "step": 61919 }, { "epoch": 4.86, "learning_rate": 5.144671841919548e-05, "loss": 2.6112, "step": 61920 }, { "epoch": 4.86, "learning_rate": 5.144593428997099e-05, "loss": 2.8307, "step": 61921 }, { "epoch": 4.86, "learning_rate": 5.1445150160746494e-05, "loss": 2.682, "step": 61922 }, { "epoch": 4.86, "learning_rate": 5.1444366031522e-05, "loss": 2.6673, "step": 61923 }, { "epoch": 4.86, "learning_rate": 5.14435819022975e-05, "loss": 2.5824, "step": 61924 }, { "epoch": 4.86, "learning_rate": 5.1442797773073005e-05, "loss": 2.5563, "step": 61925 }, { "epoch": 4.86, "learning_rate": 5.144201364384851e-05, "loss": 2.6502, "step": 61926 }, { "epoch": 4.86, "learning_rate": 5.144122951462401e-05, "loss": 2.5458, "step": 61927 }, { "epoch": 4.86, "learning_rate": 5.1440445385399516e-05, "loss": 2.6714, "step": 61928 }, { "epoch": 4.86, "learning_rate": 5.143966125617502e-05, "loss": 2.7551, "step": 61929 }, { "epoch": 4.86, "learning_rate": 5.1438877126950523e-05, "loss": 2.7784, "step": 61930 }, { "epoch": 4.86, "learning_rate": 5.143809299772603e-05, "loss": 2.7057, "step": 61931 }, { "epoch": 4.86, "learning_rate": 5.143730886850153e-05, "loss": 2.676, "step": 61932 }, { "epoch": 4.86, "learning_rate": 5.1436524739277034e-05, "loss": 2.6192, "step": 61933 }, { "epoch": 4.86, "learning_rate": 5.143574061005254e-05, "loss": 2.6391, "step": 61934 }, { "epoch": 4.86, "learning_rate": 5.143495648082804e-05, "loss": 2.7176, "step": 61935 }, { "epoch": 4.86, "learning_rate": 5.143417235160355e-05, "loss": 2.642, "step": 61936 }, { "epoch": 4.86, "learning_rate": 5.1433388222379056e-05, "loss": 2.5508, "step": 61937 }, { "epoch": 4.86, "learning_rate": 5.143260409315456e-05, "loss": 2.6789, "step": 61938 }, { "epoch": 4.86, "learning_rate": 5.143181996393006e-05, "loss": 2.5255, "step": 61939 }, { "epoch": 4.86, "learning_rate": 5.143103583470557e-05, "loss": 2.8813, "step": 61940 }, { "epoch": 4.86, "learning_rate": 5.143025170548107e-05, "loss": 2.7272, "step": 61941 }, { "epoch": 4.86, "learning_rate": 5.1429467576256574e-05, "loss": 2.8024, "step": 61942 }, { "epoch": 4.86, "learning_rate": 5.142868344703208e-05, "loss": 2.7008, "step": 61943 }, { "epoch": 4.86, "learning_rate": 5.142789931780758e-05, "loss": 2.6768, "step": 61944 }, { "epoch": 4.86, "learning_rate": 5.1427115188583085e-05, "loss": 2.602, "step": 61945 }, { "epoch": 4.86, "learning_rate": 5.142633105935859e-05, "loss": 2.6026, "step": 61946 }, { "epoch": 4.86, "learning_rate": 5.1425546930134086e-05, "loss": 2.5418, "step": 61947 }, { "epoch": 4.86, "learning_rate": 5.142476280090959e-05, "loss": 2.9515, "step": 61948 }, { "epoch": 4.86, "learning_rate": 5.142397867168509e-05, "loss": 2.5883, "step": 61949 }, { "epoch": 4.86, "learning_rate": 5.14231945424606e-05, "loss": 2.6622, "step": 61950 }, { "epoch": 4.86, "learning_rate": 5.14224104132361e-05, "loss": 2.6573, "step": 61951 }, { "epoch": 4.86, "learning_rate": 5.1421626284011604e-05, "loss": 2.5403, "step": 61952 }, { "epoch": 4.86, "learning_rate": 5.142084215478711e-05, "loss": 2.6075, "step": 61953 }, { "epoch": 4.86, "learning_rate": 5.142005802556261e-05, "loss": 2.7, "step": 61954 }, { "epoch": 4.86, "learning_rate": 5.1419273896338115e-05, "loss": 2.5961, "step": 61955 }, { "epoch": 4.86, "learning_rate": 5.141848976711362e-05, "loss": 2.6276, "step": 61956 }, { "epoch": 4.86, "learning_rate": 5.141770563788912e-05, "loss": 2.8716, "step": 61957 }, { "epoch": 4.86, "learning_rate": 5.1416921508664626e-05, "loss": 2.715, "step": 61958 }, { "epoch": 4.86, "learning_rate": 5.141613737944013e-05, "loss": 2.6507, "step": 61959 }, { "epoch": 4.86, "learning_rate": 5.1415353250215634e-05, "loss": 2.5461, "step": 61960 }, { "epoch": 4.86, "learning_rate": 5.141456912099114e-05, "loss": 2.5834, "step": 61961 }, { "epoch": 4.86, "learning_rate": 5.141378499176665e-05, "loss": 2.7135, "step": 61962 }, { "epoch": 4.86, "learning_rate": 5.141300086254215e-05, "loss": 2.6265, "step": 61963 }, { "epoch": 4.86, "learning_rate": 5.1412216733317655e-05, "loss": 2.7548, "step": 61964 }, { "epoch": 4.86, "learning_rate": 5.141143260409316e-05, "loss": 2.5661, "step": 61965 }, { "epoch": 4.86, "learning_rate": 5.141064847486866e-05, "loss": 2.5706, "step": 61966 }, { "epoch": 4.86, "learning_rate": 5.1409864345644166e-05, "loss": 2.7833, "step": 61967 }, { "epoch": 4.86, "learning_rate": 5.140908021641967e-05, "loss": 2.4754, "step": 61968 }, { "epoch": 4.86, "learning_rate": 5.1408296087195174e-05, "loss": 2.7698, "step": 61969 }, { "epoch": 4.86, "learning_rate": 5.140751195797068e-05, "loss": 2.7197, "step": 61970 }, { "epoch": 4.86, "learning_rate": 5.140672782874618e-05, "loss": 2.5143, "step": 61971 }, { "epoch": 4.86, "learning_rate": 5.1405943699521685e-05, "loss": 2.8155, "step": 61972 }, { "epoch": 4.86, "learning_rate": 5.140515957029719e-05, "loss": 2.5313, "step": 61973 }, { "epoch": 4.86, "learning_rate": 5.140437544107269e-05, "loss": 2.7487, "step": 61974 }, { "epoch": 4.86, "learning_rate": 5.1403591311848196e-05, "loss": 2.5899, "step": 61975 }, { "epoch": 4.86, "learning_rate": 5.14028071826237e-05, "loss": 2.7177, "step": 61976 }, { "epoch": 4.86, "learning_rate": 5.14020230533992e-05, "loss": 2.5552, "step": 61977 }, { "epoch": 4.86, "learning_rate": 5.1401238924174714e-05, "loss": 2.7572, "step": 61978 }, { "epoch": 4.86, "learning_rate": 5.140045479495022e-05, "loss": 2.5826, "step": 61979 }, { "epoch": 4.86, "learning_rate": 5.139967066572572e-05, "loss": 2.7058, "step": 61980 }, { "epoch": 4.86, "learning_rate": 5.1398886536501225e-05, "loss": 2.7043, "step": 61981 }, { "epoch": 4.86, "learning_rate": 5.139810240727673e-05, "loss": 2.7325, "step": 61982 }, { "epoch": 4.86, "learning_rate": 5.139731827805222e-05, "loss": 2.8177, "step": 61983 }, { "epoch": 4.86, "learning_rate": 5.139653414882772e-05, "loss": 2.7786, "step": 61984 }, { "epoch": 4.86, "learning_rate": 5.1395750019603226e-05, "loss": 2.5438, "step": 61985 }, { "epoch": 4.86, "learning_rate": 5.139496589037873e-05, "loss": 2.8458, "step": 61986 }, { "epoch": 4.86, "learning_rate": 5.139418176115423e-05, "loss": 2.6387, "step": 61987 }, { "epoch": 4.86, "learning_rate": 5.1393397631929744e-05, "loss": 2.7612, "step": 61988 }, { "epoch": 4.86, "learning_rate": 5.139261350270525e-05, "loss": 2.6163, "step": 61989 }, { "epoch": 4.86, "learning_rate": 5.139182937348075e-05, "loss": 2.622, "step": 61990 }, { "epoch": 4.86, "learning_rate": 5.1391045244256255e-05, "loss": 2.6779, "step": 61991 }, { "epoch": 4.86, "learning_rate": 5.139026111503176e-05, "loss": 2.6089, "step": 61992 }, { "epoch": 4.86, "learning_rate": 5.138947698580726e-05, "loss": 2.7108, "step": 61993 }, { "epoch": 4.86, "learning_rate": 5.1388692856582766e-05, "loss": 2.6934, "step": 61994 }, { "epoch": 4.86, "learning_rate": 5.138790872735827e-05, "loss": 2.5822, "step": 61995 }, { "epoch": 4.86, "learning_rate": 5.138712459813377e-05, "loss": 2.7658, "step": 61996 }, { "epoch": 4.86, "learning_rate": 5.1386340468909277e-05, "loss": 2.6121, "step": 61997 }, { "epoch": 4.86, "learning_rate": 5.138555633968478e-05, "loss": 2.8288, "step": 61998 }, { "epoch": 4.86, "learning_rate": 5.1384772210460284e-05, "loss": 2.7288, "step": 61999 }, { "epoch": 4.86, "learning_rate": 5.138398808123579e-05, "loss": 2.5856, "step": 62000 }, { "epoch": 4.86, "eval_loss": 2.892299175262451, "eval_runtime": 217.6937, "eval_samples_per_second": 52.055, "eval_steps_per_second": 3.257, "step": 62000 } ], "max_steps": 127530, "num_train_epochs": 10, "total_flos": 1.542317451051221e+18, "trial_name": null, "trial_params": null }