{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.0, "global_step": 9632, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 2.99968853820598e-07, "loss": 0.1361, "step": 1 }, { "epoch": 0.0, "learning_rate": 2.99937707641196e-07, "loss": 0.1972, "step": 2 }, { "epoch": 0.0, "learning_rate": 2.99906561461794e-07, "loss": 0.2103, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.99875415282392e-07, "loss": 0.2061, "step": 4 }, { "epoch": 0.0, "learning_rate": 2.9984426910299004e-07, "loss": 0.1838, "step": 5 }, { "epoch": 0.0, "learning_rate": 2.9981312292358803e-07, "loss": 0.1893, "step": 6 }, { "epoch": 0.0, "learning_rate": 2.9978197674418607e-07, "loss": 0.2126, "step": 7 }, { "epoch": 0.0, "learning_rate": 2.99750830564784e-07, "loss": 0.1913, "step": 8 }, { "epoch": 0.0, "learning_rate": 2.9971968438538205e-07, "loss": 0.1985, "step": 9 }, { "epoch": 0.0, "learning_rate": 2.9968853820598004e-07, "loss": 0.2012, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.996573920265781e-07, "loss": 0.1443, "step": 11 }, { "epoch": 0.0, "learning_rate": 2.9962624584717607e-07, "loss": 0.1878, "step": 12 }, { "epoch": 0.0, "learning_rate": 2.9959509966777406e-07, "loss": 0.1572, "step": 13 }, { "epoch": 0.0, "learning_rate": 2.995639534883721e-07, "loss": 0.1738, "step": 14 }, { "epoch": 0.0, "learning_rate": 2.995328073089701e-07, "loss": 0.2092, "step": 15 }, { "epoch": 0.0, "learning_rate": 2.995016611295681e-07, "loss": 0.1552, "step": 16 }, { "epoch": 0.0, "learning_rate": 2.9947051495016607e-07, "loss": 0.177, "step": 17 }, { "epoch": 0.0, "learning_rate": 2.994393687707641e-07, "loss": 0.1584, "step": 18 }, { "epoch": 0.0, "learning_rate": 2.994082225913621e-07, "loss": 0.2361, "step": 19 }, { "epoch": 0.0, "learning_rate": 2.9937707641196014e-07, "loss": 0.1961, "step": 20 }, { "epoch": 0.0, "learning_rate": 2.9934593023255813e-07, "loss": 0.1828, "step": 21 }, { "epoch": 0.0, "learning_rate": 2.993147840531561e-07, "loss": 0.2031, "step": 22 }, { "epoch": 0.0, "learning_rate": 2.992836378737541e-07, "loss": 0.1842, "step": 23 }, { "epoch": 0.0, "learning_rate": 2.9925249169435215e-07, "loss": 0.2062, "step": 24 }, { "epoch": 0.01, "learning_rate": 2.9922134551495014e-07, "loss": 0.2117, "step": 25 }, { "epoch": 0.01, "learning_rate": 2.9919019933554813e-07, "loss": 0.2337, "step": 26 }, { "epoch": 0.01, "learning_rate": 2.991590531561462e-07, "loss": 0.2444, "step": 27 }, { "epoch": 0.01, "learning_rate": 2.9912790697674416e-07, "loss": 0.1802, "step": 28 }, { "epoch": 0.01, "learning_rate": 2.990967607973422e-07, "loss": 0.1402, "step": 29 }, { "epoch": 0.01, "learning_rate": 2.990656146179402e-07, "loss": 0.1803, "step": 30 }, { "epoch": 0.01, "learning_rate": 2.990344684385382e-07, "loss": 0.1838, "step": 31 }, { "epoch": 0.01, "learning_rate": 2.990033222591362e-07, "loss": 0.1684, "step": 32 }, { "epoch": 0.01, "learning_rate": 2.989721760797342e-07, "loss": 0.171, "step": 33 }, { "epoch": 0.01, "learning_rate": 2.989410299003322e-07, "loss": 0.1859, "step": 34 }, { "epoch": 0.01, "learning_rate": 2.9890988372093025e-07, "loss": 0.1981, "step": 35 }, { "epoch": 0.01, "learning_rate": 2.9887873754152824e-07, "loss": 0.1895, "step": 36 }, { "epoch": 0.01, "learning_rate": 2.9884759136212623e-07, "loss": 0.1899, "step": 37 }, { "epoch": 0.01, "learning_rate": 2.988164451827242e-07, "loss": 0.21, "step": 38 }, { "epoch": 0.01, "learning_rate": 2.9878529900332226e-07, "loss": 0.1965, "step": 39 }, { "epoch": 0.01, "learning_rate": 2.9875415282392025e-07, "loss": 0.1602, "step": 40 }, { "epoch": 0.01, "learning_rate": 2.9872300664451824e-07, "loss": 0.2062, "step": 41 }, { "epoch": 0.01, "learning_rate": 2.986918604651163e-07, "loss": 0.1504, "step": 42 }, { "epoch": 0.01, "learning_rate": 2.9866071428571427e-07, "loss": 0.1626, "step": 43 }, { "epoch": 0.01, "learning_rate": 2.986295681063123e-07, "loss": 0.1573, "step": 44 }, { "epoch": 0.01, "learning_rate": 2.9859842192691025e-07, "loss": 0.1913, "step": 45 }, { "epoch": 0.01, "learning_rate": 2.985672757475083e-07, "loss": 0.2285, "step": 46 }, { "epoch": 0.01, "learning_rate": 2.985361295681063e-07, "loss": 0.1347, "step": 47 }, { "epoch": 0.01, "learning_rate": 2.985049833887043e-07, "loss": 0.2189, "step": 48 }, { "epoch": 0.01, "learning_rate": 2.984738372093023e-07, "loss": 0.1608, "step": 49 }, { "epoch": 0.01, "learning_rate": 2.984426910299003e-07, "loss": 0.1764, "step": 50 }, { "epoch": 0.01, "learning_rate": 2.9841154485049834e-07, "loss": 0.1739, "step": 51 }, { "epoch": 0.01, "learning_rate": 2.9838039867109633e-07, "loss": 0.208, "step": 52 }, { "epoch": 0.01, "learning_rate": 2.983492524916943e-07, "loss": 0.1563, "step": 53 }, { "epoch": 0.01, "learning_rate": 2.9831810631229236e-07, "loss": 0.211, "step": 54 }, { "epoch": 0.01, "learning_rate": 2.9828696013289035e-07, "loss": 0.1963, "step": 55 }, { "epoch": 0.01, "learning_rate": 2.9825581395348834e-07, "loss": 0.1862, "step": 56 }, { "epoch": 0.01, "learning_rate": 2.982246677740864e-07, "loss": 0.2344, "step": 57 }, { "epoch": 0.01, "learning_rate": 2.981935215946844e-07, "loss": 0.1777, "step": 58 }, { "epoch": 0.01, "learning_rate": 2.9816237541528236e-07, "loss": 0.1525, "step": 59 }, { "epoch": 0.01, "learning_rate": 2.9813122923588035e-07, "loss": 0.1971, "step": 60 }, { "epoch": 0.01, "learning_rate": 2.981000830564784e-07, "loss": 0.1878, "step": 61 }, { "epoch": 0.01, "learning_rate": 2.980689368770764e-07, "loss": 0.1636, "step": 62 }, { "epoch": 0.01, "learning_rate": 2.9803779069767443e-07, "loss": 0.2065, "step": 63 }, { "epoch": 0.01, "learning_rate": 2.980066445182724e-07, "loss": 0.1686, "step": 64 }, { "epoch": 0.01, "learning_rate": 2.979754983388704e-07, "loss": 0.1696, "step": 65 }, { "epoch": 0.01, "learning_rate": 2.9794435215946845e-07, "loss": 0.1814, "step": 66 }, { "epoch": 0.01, "learning_rate": 2.9791320598006644e-07, "loss": 0.2087, "step": 67 }, { "epoch": 0.01, "learning_rate": 2.9788205980066443e-07, "loss": 0.2036, "step": 68 }, { "epoch": 0.01, "learning_rate": 2.978509136212624e-07, "loss": 0.2037, "step": 69 }, { "epoch": 0.01, "learning_rate": 2.9781976744186046e-07, "loss": 0.1614, "step": 70 }, { "epoch": 0.01, "learning_rate": 2.9778862126245845e-07, "loss": 0.1792, "step": 71 }, { "epoch": 0.01, "learning_rate": 2.977574750830565e-07, "loss": 0.2158, "step": 72 }, { "epoch": 0.02, "learning_rate": 2.977263289036545e-07, "loss": 0.1994, "step": 73 }, { "epoch": 0.02, "learning_rate": 2.9769518272425247e-07, "loss": 0.1786, "step": 74 }, { "epoch": 0.02, "learning_rate": 2.9766403654485046e-07, "loss": 0.2112, "step": 75 }, { "epoch": 0.02, "learning_rate": 2.976328903654485e-07, "loss": 0.161, "step": 76 }, { "epoch": 0.02, "learning_rate": 2.976017441860465e-07, "loss": 0.2171, "step": 77 }, { "epoch": 0.02, "learning_rate": 2.975705980066445e-07, "loss": 0.2409, "step": 78 }, { "epoch": 0.02, "learning_rate": 2.975394518272425e-07, "loss": 0.2281, "step": 79 }, { "epoch": 0.02, "learning_rate": 2.975083056478405e-07, "loss": 0.1536, "step": 80 }, { "epoch": 0.02, "learning_rate": 2.9747715946843855e-07, "loss": 0.1878, "step": 81 }, { "epoch": 0.02, "learning_rate": 2.9744601328903654e-07, "loss": 0.2334, "step": 82 }, { "epoch": 0.02, "learning_rate": 2.9741486710963453e-07, "loss": 0.1508, "step": 83 }, { "epoch": 0.02, "learning_rate": 2.973837209302325e-07, "loss": 0.1611, "step": 84 }, { "epoch": 0.02, "learning_rate": 2.9735257475083056e-07, "loss": 0.2322, "step": 85 }, { "epoch": 0.02, "learning_rate": 2.9732142857142855e-07, "loss": 0.1847, "step": 86 }, { "epoch": 0.02, "learning_rate": 2.972902823920266e-07, "loss": 0.2089, "step": 87 }, { "epoch": 0.02, "learning_rate": 2.972591362126246e-07, "loss": 0.2256, "step": 88 }, { "epoch": 0.02, "learning_rate": 2.972279900332226e-07, "loss": 0.218, "step": 89 }, { "epoch": 0.02, "learning_rate": 2.9719684385382056e-07, "loss": 0.1808, "step": 90 }, { "epoch": 0.02, "learning_rate": 2.971656976744186e-07, "loss": 0.1522, "step": 91 }, { "epoch": 0.02, "learning_rate": 2.971345514950166e-07, "loss": 0.2296, "step": 92 }, { "epoch": 0.02, "learning_rate": 2.971034053156146e-07, "loss": 0.2208, "step": 93 }, { "epoch": 0.02, "learning_rate": 2.9707225913621263e-07, "loss": 0.1752, "step": 94 }, { "epoch": 0.02, "learning_rate": 2.970411129568106e-07, "loss": 0.2167, "step": 95 }, { "epoch": 0.02, "learning_rate": 2.9700996677740866e-07, "loss": 0.266, "step": 96 }, { "epoch": 0.02, "learning_rate": 2.969788205980066e-07, "loss": 0.1891, "step": 97 }, { "epoch": 0.02, "learning_rate": 2.9694767441860464e-07, "loss": 0.1572, "step": 98 }, { "epoch": 0.02, "learning_rate": 2.9691652823920263e-07, "loss": 0.1493, "step": 99 }, { "epoch": 0.02, "learning_rate": 2.9688538205980067e-07, "loss": 0.2175, "step": 100 }, { "epoch": 0.02, "learning_rate": 2.9685423588039866e-07, "loss": 0.1529, "step": 101 }, { "epoch": 0.02, "learning_rate": 2.9682308970099665e-07, "loss": 0.1801, "step": 102 }, { "epoch": 0.02, "learning_rate": 2.967919435215947e-07, "loss": 0.1496, "step": 103 }, { "epoch": 0.02, "learning_rate": 2.967607973421927e-07, "loss": 0.2607, "step": 104 }, { "epoch": 0.02, "learning_rate": 2.9672965116279067e-07, "loss": 0.1926, "step": 105 }, { "epoch": 0.02, "learning_rate": 2.9669850498338866e-07, "loss": 0.2054, "step": 106 }, { "epoch": 0.02, "learning_rate": 2.966673588039867e-07, "loss": 0.1782, "step": 107 }, { "epoch": 0.02, "learning_rate": 2.966362126245847e-07, "loss": 0.2554, "step": 108 }, { "epoch": 0.02, "learning_rate": 2.9660506644518273e-07, "loss": 0.1733, "step": 109 }, { "epoch": 0.02, "learning_rate": 2.965739202657807e-07, "loss": 0.1927, "step": 110 }, { "epoch": 0.02, "learning_rate": 2.965427740863787e-07, "loss": 0.2675, "step": 111 }, { "epoch": 0.02, "learning_rate": 2.965116279069767e-07, "loss": 0.1755, "step": 112 }, { "epoch": 0.02, "learning_rate": 2.9648048172757474e-07, "loss": 0.1396, "step": 113 }, { "epoch": 0.02, "learning_rate": 2.9644933554817273e-07, "loss": 0.187, "step": 114 }, { "epoch": 0.02, "learning_rate": 2.964181893687708e-07, "loss": 0.1191, "step": 115 }, { "epoch": 0.02, "learning_rate": 2.9638704318936876e-07, "loss": 0.1501, "step": 116 }, { "epoch": 0.02, "learning_rate": 2.9635589700996675e-07, "loss": 0.1935, "step": 117 }, { "epoch": 0.02, "learning_rate": 2.963247508305648e-07, "loss": 0.1513, "step": 118 }, { "epoch": 0.02, "learning_rate": 2.962936046511628e-07, "loss": 0.1774, "step": 119 }, { "epoch": 0.02, "learning_rate": 2.962624584717608e-07, "loss": 0.1622, "step": 120 }, { "epoch": 0.03, "learning_rate": 2.9623131229235876e-07, "loss": 0.143, "step": 121 }, { "epoch": 0.03, "learning_rate": 2.962001661129568e-07, "loss": 0.1657, "step": 122 }, { "epoch": 0.03, "learning_rate": 2.961690199335548e-07, "loss": 0.1645, "step": 123 }, { "epoch": 0.03, "learning_rate": 2.9613787375415284e-07, "loss": 0.2004, "step": 124 }, { "epoch": 0.03, "learning_rate": 2.961067275747508e-07, "loss": 0.1994, "step": 125 }, { "epoch": 0.03, "learning_rate": 2.960755813953488e-07, "loss": 0.1685, "step": 126 }, { "epoch": 0.03, "learning_rate": 2.960444352159468e-07, "loss": 0.1541, "step": 127 }, { "epoch": 0.03, "learning_rate": 2.9601328903654485e-07, "loss": 0.2123, "step": 128 }, { "epoch": 0.03, "learning_rate": 2.9598214285714284e-07, "loss": 0.2003, "step": 129 }, { "epoch": 0.03, "learning_rate": 2.959509966777408e-07, "loss": 0.2224, "step": 130 }, { "epoch": 0.03, "learning_rate": 2.9591985049833887e-07, "loss": 0.2501, "step": 131 }, { "epoch": 0.03, "learning_rate": 2.9588870431893686e-07, "loss": 0.1913, "step": 132 }, { "epoch": 0.03, "learning_rate": 2.958575581395349e-07, "loss": 0.212, "step": 133 }, { "epoch": 0.03, "learning_rate": 2.958264119601329e-07, "loss": 0.1939, "step": 134 }, { "epoch": 0.03, "learning_rate": 2.957952657807309e-07, "loss": 0.1554, "step": 135 }, { "epoch": 0.03, "learning_rate": 2.9576411960132887e-07, "loss": 0.175, "step": 136 }, { "epoch": 0.03, "learning_rate": 2.957329734219269e-07, "loss": 0.2049, "step": 137 }, { "epoch": 0.03, "learning_rate": 2.957018272425249e-07, "loss": 0.2099, "step": 138 }, { "epoch": 0.03, "learning_rate": 2.956706810631229e-07, "loss": 0.1883, "step": 139 }, { "epoch": 0.03, "learning_rate": 2.9563953488372093e-07, "loss": 0.2081, "step": 140 }, { "epoch": 0.03, "learning_rate": 2.956083887043189e-07, "loss": 0.1339, "step": 141 }, { "epoch": 0.03, "learning_rate": 2.9557724252491696e-07, "loss": 0.1968, "step": 142 }, { "epoch": 0.03, "learning_rate": 2.9554609634551495e-07, "loss": 0.1507, "step": 143 }, { "epoch": 0.03, "learning_rate": 2.9551495016611294e-07, "loss": 0.1579, "step": 144 }, { "epoch": 0.03, "learning_rate": 2.9548380398671093e-07, "loss": 0.2467, "step": 145 }, { "epoch": 0.03, "learning_rate": 2.9545265780730897e-07, "loss": 0.2605, "step": 146 }, { "epoch": 0.03, "learning_rate": 2.9542151162790696e-07, "loss": 0.1937, "step": 147 }, { "epoch": 0.03, "learning_rate": 2.95390365448505e-07, "loss": 0.2137, "step": 148 }, { "epoch": 0.03, "learning_rate": 2.9535921926910294e-07, "loss": 0.1466, "step": 149 }, { "epoch": 0.03, "learning_rate": 2.95328073089701e-07, "loss": 0.2011, "step": 150 }, { "epoch": 0.03, "learning_rate": 2.9529692691029897e-07, "loss": 0.1522, "step": 151 }, { "epoch": 0.03, "learning_rate": 2.95265780730897e-07, "loss": 0.1583, "step": 152 }, { "epoch": 0.03, "learning_rate": 2.95234634551495e-07, "loss": 0.2724, "step": 153 }, { "epoch": 0.03, "learning_rate": 2.95203488372093e-07, "loss": 0.2721, "step": 154 }, { "epoch": 0.03, "learning_rate": 2.9517234219269104e-07, "loss": 0.2002, "step": 155 }, { "epoch": 0.03, "learning_rate": 2.95141196013289e-07, "loss": 0.1373, "step": 156 }, { "epoch": 0.03, "learning_rate": 2.9511004983388707e-07, "loss": 0.2267, "step": 157 }, { "epoch": 0.03, "learning_rate": 2.95078903654485e-07, "loss": 0.2276, "step": 158 }, { "epoch": 0.03, "learning_rate": 2.9504775747508305e-07, "loss": 0.1545, "step": 159 }, { "epoch": 0.03, "learning_rate": 2.9501661129568104e-07, "loss": 0.2029, "step": 160 }, { "epoch": 0.03, "learning_rate": 2.949854651162791e-07, "loss": 0.1621, "step": 161 }, { "epoch": 0.03, "learning_rate": 2.9495431893687707e-07, "loss": 0.1134, "step": 162 }, { "epoch": 0.03, "learning_rate": 2.9492317275747506e-07, "loss": 0.1683, "step": 163 }, { "epoch": 0.03, "learning_rate": 2.9489202657807305e-07, "loss": 0.164, "step": 164 }, { "epoch": 0.03, "learning_rate": 2.948608803986711e-07, "loss": 0.1247, "step": 165 }, { "epoch": 0.03, "learning_rate": 2.948297342192691e-07, "loss": 0.206, "step": 166 }, { "epoch": 0.03, "learning_rate": 2.9479858803986707e-07, "loss": 0.2303, "step": 167 }, { "epoch": 0.03, "learning_rate": 2.947674418604651e-07, "loss": 0.2173, "step": 168 }, { "epoch": 0.04, "learning_rate": 2.947362956810631e-07, "loss": 0.1611, "step": 169 }, { "epoch": 0.04, "learning_rate": 2.9470514950166114e-07, "loss": 0.1599, "step": 170 }, { "epoch": 0.04, "learning_rate": 2.9467400332225913e-07, "loss": 0.1988, "step": 171 }, { "epoch": 0.04, "learning_rate": 2.946428571428571e-07, "loss": 0.1822, "step": 172 }, { "epoch": 0.04, "learning_rate": 2.946117109634551e-07, "loss": 0.1847, "step": 173 }, { "epoch": 0.04, "learning_rate": 2.9458056478405315e-07, "loss": 0.1477, "step": 174 }, { "epoch": 0.04, "learning_rate": 2.9454941860465114e-07, "loss": 0.1718, "step": 175 }, { "epoch": 0.04, "learning_rate": 2.945182724252492e-07, "loss": 0.1716, "step": 176 }, { "epoch": 0.04, "learning_rate": 2.9448712624584717e-07, "loss": 0.1637, "step": 177 }, { "epoch": 0.04, "learning_rate": 2.9445598006644516e-07, "loss": 0.1321, "step": 178 }, { "epoch": 0.04, "learning_rate": 2.944248338870432e-07, "loss": 0.1888, "step": 179 }, { "epoch": 0.04, "learning_rate": 2.943936877076412e-07, "loss": 0.197, "step": 180 }, { "epoch": 0.04, "learning_rate": 2.943625415282392e-07, "loss": 0.2023, "step": 181 }, { "epoch": 0.04, "learning_rate": 2.9433139534883717e-07, "loss": 0.1651, "step": 182 }, { "epoch": 0.04, "learning_rate": 2.943002491694352e-07, "loss": 0.1783, "step": 183 }, { "epoch": 0.04, "learning_rate": 2.942691029900332e-07, "loss": 0.1562, "step": 184 }, { "epoch": 0.04, "learning_rate": 2.9423795681063125e-07, "loss": 0.1508, "step": 185 }, { "epoch": 0.04, "learning_rate": 2.942068106312292e-07, "loss": 0.1648, "step": 186 }, { "epoch": 0.04, "learning_rate": 2.941756644518272e-07, "loss": 0.1212, "step": 187 }, { "epoch": 0.04, "learning_rate": 2.941445182724252e-07, "loss": 0.1778, "step": 188 }, { "epoch": 0.04, "learning_rate": 2.9411337209302326e-07, "loss": 0.1862, "step": 189 }, { "epoch": 0.04, "learning_rate": 2.9408222591362125e-07, "loss": 0.1833, "step": 190 }, { "epoch": 0.04, "learning_rate": 2.9405107973421924e-07, "loss": 0.1668, "step": 191 }, { "epoch": 0.04, "learning_rate": 2.940199335548173e-07, "loss": 0.1467, "step": 192 }, { "epoch": 0.04, "learning_rate": 2.9398878737541527e-07, "loss": 0.2205, "step": 193 }, { "epoch": 0.04, "learning_rate": 2.939576411960133e-07, "loss": 0.1881, "step": 194 }, { "epoch": 0.04, "learning_rate": 2.939264950166113e-07, "loss": 0.1799, "step": 195 }, { "epoch": 0.04, "learning_rate": 2.938953488372093e-07, "loss": 0.1299, "step": 196 }, { "epoch": 0.04, "learning_rate": 2.938642026578073e-07, "loss": 0.1503, "step": 197 }, { "epoch": 0.04, "learning_rate": 2.938330564784053e-07, "loss": 0.2088, "step": 198 }, { "epoch": 0.04, "learning_rate": 2.938019102990033e-07, "loss": 0.1893, "step": 199 }, { "epoch": 0.04, "learning_rate": 2.937707641196013e-07, "loss": 0.1907, "step": 200 }, { "epoch": 0.04, "learning_rate": 2.937396179401993e-07, "loss": 0.1693, "step": 201 }, { "epoch": 0.04, "learning_rate": 2.9370847176079733e-07, "loss": 0.1735, "step": 202 }, { "epoch": 0.04, "learning_rate": 2.936773255813953e-07, "loss": 0.2171, "step": 203 }, { "epoch": 0.04, "learning_rate": 2.9364617940199336e-07, "loss": 0.2243, "step": 204 }, { "epoch": 0.04, "learning_rate": 2.9361503322259135e-07, "loss": 0.184, "step": 205 }, { "epoch": 0.04, "learning_rate": 2.9358388704318934e-07, "loss": 0.2409, "step": 206 }, { "epoch": 0.04, "learning_rate": 2.935527408637874e-07, "loss": 0.158, "step": 207 }, { "epoch": 0.04, "learning_rate": 2.9352159468438537e-07, "loss": 0.2106, "step": 208 }, { "epoch": 0.04, "learning_rate": 2.934904485049834e-07, "loss": 0.1585, "step": 209 }, { "epoch": 0.04, "learning_rate": 2.9345930232558135e-07, "loss": 0.1977, "step": 210 }, { "epoch": 0.04, "learning_rate": 2.934281561461794e-07, "loss": 0.1781, "step": 211 }, { "epoch": 0.04, "learning_rate": 2.933970099667774e-07, "loss": 0.2048, "step": 212 }, { "epoch": 0.04, "learning_rate": 2.933658637873754e-07, "loss": 0.1829, "step": 213 }, { "epoch": 0.04, "learning_rate": 2.933347176079734e-07, "loss": 0.256, "step": 214 }, { "epoch": 0.04, "learning_rate": 2.933035714285714e-07, "loss": 0.1665, "step": 215 }, { "epoch": 0.04, "learning_rate": 2.9327242524916945e-07, "loss": 0.1675, "step": 216 }, { "epoch": 0.05, "learning_rate": 2.9324127906976744e-07, "loss": 0.1534, "step": 217 }, { "epoch": 0.05, "learning_rate": 2.932101328903654e-07, "loss": 0.1388, "step": 218 }, { "epoch": 0.05, "learning_rate": 2.931789867109634e-07, "loss": 0.246, "step": 219 }, { "epoch": 0.05, "learning_rate": 2.9314784053156146e-07, "loss": 0.1581, "step": 220 }, { "epoch": 0.05, "learning_rate": 2.9311669435215945e-07, "loss": 0.1793, "step": 221 }, { "epoch": 0.05, "learning_rate": 2.930855481727575e-07, "loss": 0.1678, "step": 222 }, { "epoch": 0.05, "learning_rate": 2.930544019933555e-07, "loss": 0.298, "step": 223 }, { "epoch": 0.05, "learning_rate": 2.9302325581395347e-07, "loss": 0.1609, "step": 224 }, { "epoch": 0.05, "learning_rate": 2.9299210963455146e-07, "loss": 0.1803, "step": 225 }, { "epoch": 0.05, "learning_rate": 2.929609634551495e-07, "loss": 0.2065, "step": 226 }, { "epoch": 0.05, "learning_rate": 2.929298172757475e-07, "loss": 0.1911, "step": 227 }, { "epoch": 0.05, "learning_rate": 2.9289867109634553e-07, "loss": 0.1593, "step": 228 }, { "epoch": 0.05, "learning_rate": 2.928675249169435e-07, "loss": 0.1829, "step": 229 }, { "epoch": 0.05, "learning_rate": 2.928363787375415e-07, "loss": 0.186, "step": 230 }, { "epoch": 0.05, "learning_rate": 2.9280523255813955e-07, "loss": 0.2248, "step": 231 }, { "epoch": 0.05, "learning_rate": 2.9277408637873754e-07, "loss": 0.2751, "step": 232 }, { "epoch": 0.05, "learning_rate": 2.9274294019933553e-07, "loss": 0.2239, "step": 233 }, { "epoch": 0.05, "learning_rate": 2.927117940199335e-07, "loss": 0.2286, "step": 234 }, { "epoch": 0.05, "learning_rate": 2.9268064784053156e-07, "loss": 0.174, "step": 235 }, { "epoch": 0.05, "learning_rate": 2.9264950166112955e-07, "loss": 0.1855, "step": 236 }, { "epoch": 0.05, "learning_rate": 2.926183554817276e-07, "loss": 0.2321, "step": 237 }, { "epoch": 0.05, "learning_rate": 2.9258720930232553e-07, "loss": 0.2046, "step": 238 }, { "epoch": 0.05, "learning_rate": 2.9255606312292357e-07, "loss": 0.1771, "step": 239 }, { "epoch": 0.05, "learning_rate": 2.9252491694352156e-07, "loss": 0.2302, "step": 240 }, { "epoch": 0.05, "learning_rate": 2.924937707641196e-07, "loss": 0.1634, "step": 241 }, { "epoch": 0.05, "learning_rate": 2.924626245847176e-07, "loss": 0.1685, "step": 242 }, { "epoch": 0.05, "learning_rate": 2.924314784053156e-07, "loss": 0.1344, "step": 243 }, { "epoch": 0.05, "learning_rate": 2.924003322259136e-07, "loss": 0.1229, "step": 244 }, { "epoch": 0.05, "learning_rate": 2.923691860465116e-07, "loss": 0.1891, "step": 245 }, { "epoch": 0.05, "learning_rate": 2.9233803986710966e-07, "loss": 0.144, "step": 246 }, { "epoch": 0.05, "learning_rate": 2.923068936877076e-07, "loss": 0.2524, "step": 247 }, { "epoch": 0.05, "learning_rate": 2.9227574750830563e-07, "loss": 0.1707, "step": 248 }, { "epoch": 0.05, "learning_rate": 2.922446013289036e-07, "loss": 0.2026, "step": 249 }, { "epoch": 0.05, "learning_rate": 2.9221345514950167e-07, "loss": 0.1414, "step": 250 }, { "epoch": 0.05, "learning_rate": 2.9218230897009966e-07, "loss": 0.2226, "step": 251 }, { "epoch": 0.05, "learning_rate": 2.9215116279069764e-07, "loss": 0.1523, "step": 252 }, { "epoch": 0.05, "learning_rate": 2.921200166112957e-07, "loss": 0.2478, "step": 253 }, { "epoch": 0.05, "learning_rate": 2.920888704318937e-07, "loss": 0.1877, "step": 254 }, { "epoch": 0.05, "learning_rate": 2.9205772425249167e-07, "loss": 0.1823, "step": 255 }, { "epoch": 0.05, "learning_rate": 2.920265780730897e-07, "loss": 0.162, "step": 256 }, { "epoch": 0.05, "learning_rate": 2.919954318936877e-07, "loss": 0.1791, "step": 257 }, { "epoch": 0.05, "learning_rate": 2.919642857142857e-07, "loss": 0.2167, "step": 258 }, { "epoch": 0.05, "learning_rate": 2.9193313953488373e-07, "loss": 0.1626, "step": 259 }, { "epoch": 0.05, "learning_rate": 2.919019933554817e-07, "loss": 0.2191, "step": 260 }, { "epoch": 0.05, "learning_rate": 2.9187084717607976e-07, "loss": 0.185, "step": 261 }, { "epoch": 0.05, "learning_rate": 2.918397009966777e-07, "loss": 0.1993, "step": 262 }, { "epoch": 0.05, "learning_rate": 2.9180855481727574e-07, "loss": 0.1787, "step": 263 }, { "epoch": 0.05, "learning_rate": 2.9177740863787373e-07, "loss": 0.1897, "step": 264 }, { "epoch": 0.06, "learning_rate": 2.9174626245847177e-07, "loss": 0.1549, "step": 265 }, { "epoch": 0.06, "learning_rate": 2.9171511627906976e-07, "loss": 0.1792, "step": 266 }, { "epoch": 0.06, "learning_rate": 2.9168397009966775e-07, "loss": 0.1558, "step": 267 }, { "epoch": 0.06, "learning_rate": 2.916528239202658e-07, "loss": 0.1769, "step": 268 }, { "epoch": 0.06, "learning_rate": 2.916216777408638e-07, "loss": 0.1819, "step": 269 }, { "epoch": 0.06, "learning_rate": 2.9159053156146177e-07, "loss": 0.1587, "step": 270 }, { "epoch": 0.06, "learning_rate": 2.9155938538205976e-07, "loss": 0.1855, "step": 271 }, { "epoch": 0.06, "learning_rate": 2.915282392026578e-07, "loss": 0.2594, "step": 272 }, { "epoch": 0.06, "learning_rate": 2.914970930232558e-07, "loss": 0.2045, "step": 273 }, { "epoch": 0.06, "learning_rate": 2.9146594684385383e-07, "loss": 0.1703, "step": 274 }, { "epoch": 0.06, "learning_rate": 2.914348006644518e-07, "loss": 0.1881, "step": 275 }, { "epoch": 0.06, "learning_rate": 2.914036544850498e-07, "loss": 0.195, "step": 276 }, { "epoch": 0.06, "learning_rate": 2.913725083056478e-07, "loss": 0.1564, "step": 277 }, { "epoch": 0.06, "learning_rate": 2.9134136212624584e-07, "loss": 0.1629, "step": 278 }, { "epoch": 0.06, "learning_rate": 2.9131021594684383e-07, "loss": 0.2207, "step": 279 }, { "epoch": 0.06, "learning_rate": 2.912790697674418e-07, "loss": 0.2176, "step": 280 }, { "epoch": 0.06, "learning_rate": 2.9124792358803987e-07, "loss": 0.1993, "step": 281 }, { "epoch": 0.06, "learning_rate": 2.9121677740863785e-07, "loss": 0.1854, "step": 282 }, { "epoch": 0.06, "learning_rate": 2.911856312292359e-07, "loss": 0.1933, "step": 283 }, { "epoch": 0.06, "learning_rate": 2.911544850498339e-07, "loss": 0.1801, "step": 284 }, { "epoch": 0.06, "learning_rate": 2.911233388704319e-07, "loss": 0.1648, "step": 285 }, { "epoch": 0.06, "learning_rate": 2.9109219269102987e-07, "loss": 0.1993, "step": 286 }, { "epoch": 0.06, "learning_rate": 2.910610465116279e-07, "loss": 0.1788, "step": 287 }, { "epoch": 0.06, "learning_rate": 2.910299003322259e-07, "loss": 0.1966, "step": 288 }, { "epoch": 0.06, "learning_rate": 2.9099875415282394e-07, "loss": 0.1437, "step": 289 }, { "epoch": 0.06, "learning_rate": 2.9096760797342193e-07, "loss": 0.1687, "step": 290 }, { "epoch": 0.06, "learning_rate": 2.909364617940199e-07, "loss": 0.1745, "step": 291 }, { "epoch": 0.06, "learning_rate": 2.909053156146179e-07, "loss": 0.1802, "step": 292 }, { "epoch": 0.06, "learning_rate": 2.9087416943521595e-07, "loss": 0.1973, "step": 293 }, { "epoch": 0.06, "learning_rate": 2.9084302325581394e-07, "loss": 0.1518, "step": 294 }, { "epoch": 0.06, "learning_rate": 2.9081187707641193e-07, "loss": 0.1727, "step": 295 }, { "epoch": 0.06, "learning_rate": 2.9078073089700997e-07, "loss": 0.1834, "step": 296 }, { "epoch": 0.06, "learning_rate": 2.9074958471760796e-07, "loss": 0.1698, "step": 297 }, { "epoch": 0.06, "learning_rate": 2.90718438538206e-07, "loss": 0.1647, "step": 298 }, { "epoch": 0.06, "learning_rate": 2.9068729235880394e-07, "loss": 0.1485, "step": 299 }, { "epoch": 0.06, "learning_rate": 2.90656146179402e-07, "loss": 0.1986, "step": 300 }, { "epoch": 0.06, "learning_rate": 2.9062499999999997e-07, "loss": 0.1485, "step": 301 }, { "epoch": 0.06, "learning_rate": 2.90593853820598e-07, "loss": 0.2739, "step": 302 }, { "epoch": 0.06, "learning_rate": 2.90562707641196e-07, "loss": 0.1343, "step": 303 }, { "epoch": 0.06, "learning_rate": 2.90531561461794e-07, "loss": 0.1973, "step": 304 }, { "epoch": 0.06, "learning_rate": 2.9050041528239203e-07, "loss": 0.2152, "step": 305 }, { "epoch": 0.06, "learning_rate": 2.9046926910299e-07, "loss": 0.2001, "step": 306 }, { "epoch": 0.06, "learning_rate": 2.90438122923588e-07, "loss": 0.209, "step": 307 }, { "epoch": 0.06, "learning_rate": 2.9040697674418605e-07, "loss": 0.2177, "step": 308 }, { "epoch": 0.06, "learning_rate": 2.9037583056478404e-07, "loss": 0.195, "step": 309 }, { "epoch": 0.06, "learning_rate": 2.9034468438538203e-07, "loss": 0.2286, "step": 310 }, { "epoch": 0.06, "learning_rate": 2.903135382059801e-07, "loss": 0.1806, "step": 311 }, { "epoch": 0.06, "learning_rate": 2.9028239202657806e-07, "loss": 0.1411, "step": 312 }, { "epoch": 0.06, "learning_rate": 2.9025124584717605e-07, "loss": 0.1261, "step": 313 }, { "epoch": 0.07, "learning_rate": 2.9022009966777404e-07, "loss": 0.2034, "step": 314 }, { "epoch": 0.07, "learning_rate": 2.901889534883721e-07, "loss": 0.2308, "step": 315 }, { "epoch": 0.07, "learning_rate": 2.901578073089701e-07, "loss": 0.2662, "step": 316 }, { "epoch": 0.07, "learning_rate": 2.901266611295681e-07, "loss": 0.2645, "step": 317 }, { "epoch": 0.07, "learning_rate": 2.900955149501661e-07, "loss": 0.1832, "step": 318 }, { "epoch": 0.07, "learning_rate": 2.900643687707641e-07, "loss": 0.1447, "step": 319 }, { "epoch": 0.07, "learning_rate": 2.9003322259136214e-07, "loss": 0.215, "step": 320 }, { "epoch": 0.07, "learning_rate": 2.9000207641196013e-07, "loss": 0.1463, "step": 321 }, { "epoch": 0.07, "learning_rate": 2.899709302325581e-07, "loss": 0.1853, "step": 322 }, { "epoch": 0.07, "learning_rate": 2.899397840531561e-07, "loss": 0.1442, "step": 323 }, { "epoch": 0.07, "learning_rate": 2.8990863787375415e-07, "loss": 0.1828, "step": 324 }, { "epoch": 0.07, "learning_rate": 2.8987749169435214e-07, "loss": 0.2123, "step": 325 }, { "epoch": 0.07, "learning_rate": 2.898463455149502e-07, "loss": 0.1371, "step": 326 }, { "epoch": 0.07, "learning_rate": 2.8981519933554817e-07, "loss": 0.2703, "step": 327 }, { "epoch": 0.07, "learning_rate": 2.8978405315614616e-07, "loss": 0.1453, "step": 328 }, { "epoch": 0.07, "learning_rate": 2.8975290697674415e-07, "loss": 0.1749, "step": 329 }, { "epoch": 0.07, "learning_rate": 2.897217607973422e-07, "loss": 0.166, "step": 330 }, { "epoch": 0.07, "learning_rate": 2.896906146179402e-07, "loss": 0.1789, "step": 331 }, { "epoch": 0.07, "learning_rate": 2.8965946843853817e-07, "loss": 0.1487, "step": 332 }, { "epoch": 0.07, "learning_rate": 2.896283222591362e-07, "loss": 0.1195, "step": 333 }, { "epoch": 0.07, "learning_rate": 2.895971760797342e-07, "loss": 0.1512, "step": 334 }, { "epoch": 0.07, "learning_rate": 2.8956602990033224e-07, "loss": 0.1777, "step": 335 }, { "epoch": 0.07, "learning_rate": 2.8953488372093023e-07, "loss": 0.3041, "step": 336 }, { "epoch": 0.07, "learning_rate": 2.895037375415282e-07, "loss": 0.2901, "step": 337 }, { "epoch": 0.07, "learning_rate": 2.894725913621262e-07, "loss": 0.1788, "step": 338 }, { "epoch": 0.07, "learning_rate": 2.8944144518272425e-07, "loss": 0.2015, "step": 339 }, { "epoch": 0.07, "learning_rate": 2.8941029900332224e-07, "loss": 0.1966, "step": 340 }, { "epoch": 0.07, "learning_rate": 2.8937915282392023e-07, "loss": 0.1254, "step": 341 }, { "epoch": 0.07, "learning_rate": 2.893480066445183e-07, "loss": 0.2078, "step": 342 }, { "epoch": 0.07, "learning_rate": 2.8931686046511626e-07, "loss": 0.1571, "step": 343 }, { "epoch": 0.07, "learning_rate": 2.8928571428571425e-07, "loss": 0.1601, "step": 344 }, { "epoch": 0.07, "learning_rate": 2.892545681063123e-07, "loss": 0.2023, "step": 345 }, { "epoch": 0.07, "learning_rate": 2.892234219269103e-07, "loss": 0.1853, "step": 346 }, { "epoch": 0.07, "learning_rate": 2.891922757475083e-07, "loss": 0.13, "step": 347 }, { "epoch": 0.07, "learning_rate": 2.891611295681063e-07, "loss": 0.2148, "step": 348 }, { "epoch": 0.07, "learning_rate": 2.891299833887043e-07, "loss": 0.2, "step": 349 }, { "epoch": 0.07, "learning_rate": 2.8909883720930235e-07, "loss": 0.1985, "step": 350 }, { "epoch": 0.07, "learning_rate": 2.890676910299003e-07, "loss": 0.2238, "step": 351 }, { "epoch": 0.07, "learning_rate": 2.8903654485049833e-07, "loss": 0.1585, "step": 352 }, { "epoch": 0.07, "learning_rate": 2.890053986710963e-07, "loss": 0.1907, "step": 353 }, { "epoch": 0.07, "learning_rate": 2.8897425249169436e-07, "loss": 0.1862, "step": 354 }, { "epoch": 0.07, "learning_rate": 2.8894310631229235e-07, "loss": 0.1971, "step": 355 }, { "epoch": 0.07, "learning_rate": 2.8891196013289034e-07, "loss": 0.1627, "step": 356 }, { "epoch": 0.07, "learning_rate": 2.888808139534884e-07, "loss": 0.1952, "step": 357 }, { "epoch": 0.07, "learning_rate": 2.8884966777408637e-07, "loss": 0.19, "step": 358 }, { "epoch": 0.07, "learning_rate": 2.8881852159468436e-07, "loss": 0.1801, "step": 359 }, { "epoch": 0.07, "learning_rate": 2.8878737541528235e-07, "loss": 0.1866, "step": 360 }, { "epoch": 0.07, "learning_rate": 2.887562292358804e-07, "loss": 0.1566, "step": 361 }, { "epoch": 0.08, "learning_rate": 2.887250830564784e-07, "loss": 0.1661, "step": 362 }, { "epoch": 0.08, "learning_rate": 2.886939368770764e-07, "loss": 0.1867, "step": 363 }, { "epoch": 0.08, "learning_rate": 2.886627906976744e-07, "loss": 0.1612, "step": 364 }, { "epoch": 0.08, "learning_rate": 2.886316445182724e-07, "loss": 0.2069, "step": 365 }, { "epoch": 0.08, "learning_rate": 2.886004983388704e-07, "loss": 0.2334, "step": 366 }, { "epoch": 0.08, "learning_rate": 2.8856935215946843e-07, "loss": 0.1911, "step": 367 }, { "epoch": 0.08, "learning_rate": 2.885382059800664e-07, "loss": 0.1904, "step": 368 }, { "epoch": 0.08, "learning_rate": 2.8850705980066446e-07, "loss": 0.2275, "step": 369 }, { "epoch": 0.08, "learning_rate": 2.8847591362126245e-07, "loss": 0.2654, "step": 370 }, { "epoch": 0.08, "learning_rate": 2.8844476744186044e-07, "loss": 0.157, "step": 371 }, { "epoch": 0.08, "learning_rate": 2.884136212624585e-07, "loss": 0.1315, "step": 372 }, { "epoch": 0.08, "learning_rate": 2.883824750830565e-07, "loss": 0.1676, "step": 373 }, { "epoch": 0.08, "learning_rate": 2.8835132890365446e-07, "loss": 0.1772, "step": 374 }, { "epoch": 0.08, "learning_rate": 2.8832018272425245e-07, "loss": 0.183, "step": 375 }, { "epoch": 0.08, "learning_rate": 2.882890365448505e-07, "loss": 0.1884, "step": 376 }, { "epoch": 0.08, "learning_rate": 2.882578903654485e-07, "loss": 0.1051, "step": 377 }, { "epoch": 0.08, "learning_rate": 2.8822674418604653e-07, "loss": 0.2015, "step": 378 }, { "epoch": 0.08, "learning_rate": 2.881955980066445e-07, "loss": 0.2033, "step": 379 }, { "epoch": 0.08, "learning_rate": 2.881644518272425e-07, "loss": 0.2215, "step": 380 }, { "epoch": 0.08, "learning_rate": 2.881333056478405e-07, "loss": 0.1612, "step": 381 }, { "epoch": 0.08, "learning_rate": 2.8810215946843854e-07, "loss": 0.1437, "step": 382 }, { "epoch": 0.08, "learning_rate": 2.8807101328903653e-07, "loss": 0.1599, "step": 383 }, { "epoch": 0.08, "learning_rate": 2.880398671096345e-07, "loss": 0.2077, "step": 384 }, { "epoch": 0.08, "learning_rate": 2.8800872093023256e-07, "loss": 0.1946, "step": 385 }, { "epoch": 0.08, "learning_rate": 2.8797757475083055e-07, "loss": 0.2014, "step": 386 }, { "epoch": 0.08, "learning_rate": 2.879464285714286e-07, "loss": 0.2011, "step": 387 }, { "epoch": 0.08, "learning_rate": 2.879152823920266e-07, "loss": 0.1937, "step": 388 }, { "epoch": 0.08, "learning_rate": 2.8788413621262457e-07, "loss": 0.2259, "step": 389 }, { "epoch": 0.08, "learning_rate": 2.8785299003322256e-07, "loss": 0.2502, "step": 390 }, { "epoch": 0.08, "learning_rate": 2.878218438538206e-07, "loss": 0.166, "step": 391 }, { "epoch": 0.08, "learning_rate": 2.877906976744186e-07, "loss": 0.1877, "step": 392 }, { "epoch": 0.08, "learning_rate": 2.877595514950166e-07, "loss": 0.1546, "step": 393 }, { "epoch": 0.08, "learning_rate": 2.877284053156146e-07, "loss": 0.142, "step": 394 }, { "epoch": 0.08, "learning_rate": 2.876972591362126e-07, "loss": 0.186, "step": 395 }, { "epoch": 0.08, "learning_rate": 2.876661129568106e-07, "loss": 0.2379, "step": 396 }, { "epoch": 0.08, "learning_rate": 2.8763496677740864e-07, "loss": 0.152, "step": 397 }, { "epoch": 0.08, "learning_rate": 2.8760382059800663e-07, "loss": 0.2339, "step": 398 }, { "epoch": 0.08, "learning_rate": 2.875726744186046e-07, "loss": 0.1845, "step": 399 }, { "epoch": 0.08, "learning_rate": 2.8754152823920266e-07, "loss": 0.192, "step": 400 }, { "epoch": 0.08, "learning_rate": 2.8751038205980065e-07, "loss": 0.2054, "step": 401 }, { "epoch": 0.08, "learning_rate": 2.874792358803987e-07, "loss": 0.1437, "step": 402 }, { "epoch": 0.08, "learning_rate": 2.8744808970099663e-07, "loss": 0.1712, "step": 403 }, { "epoch": 0.08, "learning_rate": 2.874169435215947e-07, "loss": 0.2057, "step": 404 }, { "epoch": 0.08, "learning_rate": 2.8738579734219266e-07, "loss": 0.1726, "step": 405 }, { "epoch": 0.08, "learning_rate": 2.873546511627907e-07, "loss": 0.1802, "step": 406 }, { "epoch": 0.08, "learning_rate": 2.873235049833887e-07, "loss": 0.1543, "step": 407 }, { "epoch": 0.08, "learning_rate": 2.872923588039867e-07, "loss": 0.2168, "step": 408 }, { "epoch": 0.08, "learning_rate": 2.872612126245847e-07, "loss": 0.2083, "step": 409 }, { "epoch": 0.09, "learning_rate": 2.872300664451827e-07, "loss": 0.123, "step": 410 }, { "epoch": 0.09, "learning_rate": 2.871989202657807e-07, "loss": 0.2273, "step": 411 }, { "epoch": 0.09, "learning_rate": 2.871677740863787e-07, "loss": 0.2601, "step": 412 }, { "epoch": 0.09, "learning_rate": 2.8713662790697674e-07, "loss": 0.1639, "step": 413 }, { "epoch": 0.09, "learning_rate": 2.871054817275747e-07, "loss": 0.1759, "step": 414 }, { "epoch": 0.09, "learning_rate": 2.8707433554817277e-07, "loss": 0.1758, "step": 415 }, { "epoch": 0.09, "learning_rate": 2.8704318936877076e-07, "loss": 0.2001, "step": 416 }, { "epoch": 0.09, "learning_rate": 2.8701204318936875e-07, "loss": 0.134, "step": 417 }, { "epoch": 0.09, "learning_rate": 2.8698089700996674e-07, "loss": 0.1592, "step": 418 }, { "epoch": 0.09, "learning_rate": 2.869497508305648e-07, "loss": 0.1653, "step": 419 }, { "epoch": 0.09, "learning_rate": 2.8691860465116277e-07, "loss": 0.2538, "step": 420 }, { "epoch": 0.09, "learning_rate": 2.8688745847176076e-07, "loss": 0.1857, "step": 421 }, { "epoch": 0.09, "learning_rate": 2.868563122923588e-07, "loss": 0.2024, "step": 422 }, { "epoch": 0.09, "learning_rate": 2.868251661129568e-07, "loss": 0.1734, "step": 423 }, { "epoch": 0.09, "learning_rate": 2.8679401993355483e-07, "loss": 0.2132, "step": 424 }, { "epoch": 0.09, "learning_rate": 2.867628737541528e-07, "loss": 0.176, "step": 425 }, { "epoch": 0.09, "learning_rate": 2.867317275747508e-07, "loss": 0.1634, "step": 426 }, { "epoch": 0.09, "learning_rate": 2.867005813953488e-07, "loss": 0.1729, "step": 427 }, { "epoch": 0.09, "learning_rate": 2.8666943521594684e-07, "loss": 0.197, "step": 428 }, { "epoch": 0.09, "learning_rate": 2.8663828903654483e-07, "loss": 0.127, "step": 429 }, { "epoch": 0.09, "learning_rate": 2.8660714285714287e-07, "loss": 0.1581, "step": 430 }, { "epoch": 0.09, "learning_rate": 2.8657599667774086e-07, "loss": 0.1359, "step": 431 }, { "epoch": 0.09, "learning_rate": 2.8654485049833885e-07, "loss": 0.1908, "step": 432 }, { "epoch": 0.09, "learning_rate": 2.8651370431893684e-07, "loss": 0.1695, "step": 433 }, { "epoch": 0.09, "learning_rate": 2.864825581395349e-07, "loss": 0.239, "step": 434 }, { "epoch": 0.09, "learning_rate": 2.8645141196013287e-07, "loss": 0.2173, "step": 435 }, { "epoch": 0.09, "learning_rate": 2.8642026578073086e-07, "loss": 0.1446, "step": 436 }, { "epoch": 0.09, "learning_rate": 2.863891196013289e-07, "loss": 0.1258, "step": 437 }, { "epoch": 0.09, "learning_rate": 2.863579734219269e-07, "loss": 0.1792, "step": 438 }, { "epoch": 0.09, "learning_rate": 2.8632682724252494e-07, "loss": 0.1572, "step": 439 }, { "epoch": 0.09, "learning_rate": 2.8629568106312287e-07, "loss": 0.2286, "step": 440 }, { "epoch": 0.09, "learning_rate": 2.862645348837209e-07, "loss": 0.2688, "step": 441 }, { "epoch": 0.09, "learning_rate": 2.862333887043189e-07, "loss": 0.1666, "step": 442 }, { "epoch": 0.09, "learning_rate": 2.8620224252491695e-07, "loss": 0.1227, "step": 443 }, { "epoch": 0.09, "learning_rate": 2.8617109634551494e-07, "loss": 0.152, "step": 444 }, { "epoch": 0.09, "learning_rate": 2.861399501661129e-07, "loss": 0.1682, "step": 445 }, { "epoch": 0.09, "learning_rate": 2.8610880398671097e-07, "loss": 0.1798, "step": 446 }, { "epoch": 0.09, "learning_rate": 2.8607765780730896e-07, "loss": 0.1537, "step": 447 }, { "epoch": 0.09, "learning_rate": 2.8604651162790695e-07, "loss": 0.1955, "step": 448 }, { "epoch": 0.09, "learning_rate": 2.86015365448505e-07, "loss": 0.2223, "step": 449 }, { "epoch": 0.09, "learning_rate": 2.85984219269103e-07, "loss": 0.2022, "step": 450 }, { "epoch": 0.09, "learning_rate": 2.8595307308970097e-07, "loss": 0.1615, "step": 451 }, { "epoch": 0.09, "learning_rate": 2.85921926910299e-07, "loss": 0.1435, "step": 452 }, { "epoch": 0.09, "learning_rate": 2.85890780730897e-07, "loss": 0.1843, "step": 453 }, { "epoch": 0.09, "learning_rate": 2.85859634551495e-07, "loss": 0.1513, "step": 454 }, { "epoch": 0.09, "learning_rate": 2.85828488372093e-07, "loss": 0.19, "step": 455 }, { "epoch": 0.09, "learning_rate": 2.85797342192691e-07, "loss": 0.2265, "step": 456 }, { "epoch": 0.09, "learning_rate": 2.85766196013289e-07, "loss": 0.1384, "step": 457 }, { "epoch": 0.1, "learning_rate": 2.8573504983388705e-07, "loss": 0.1963, "step": 458 }, { "epoch": 0.1, "learning_rate": 2.8570390365448504e-07, "loss": 0.1958, "step": 459 }, { "epoch": 0.1, "learning_rate": 2.8567275747508303e-07, "loss": 0.207, "step": 460 }, { "epoch": 0.1, "learning_rate": 2.8564161129568107e-07, "loss": 0.2384, "step": 461 }, { "epoch": 0.1, "learning_rate": 2.8561046511627906e-07, "loss": 0.1455, "step": 462 }, { "epoch": 0.1, "learning_rate": 2.855793189368771e-07, "loss": 0.2082, "step": 463 }, { "epoch": 0.1, "learning_rate": 2.8554817275747504e-07, "loss": 0.2247, "step": 464 }, { "epoch": 0.1, "learning_rate": 2.855170265780731e-07, "loss": 0.204, "step": 465 }, { "epoch": 0.1, "learning_rate": 2.8548588039867107e-07, "loss": 0.1463, "step": 466 }, { "epoch": 0.1, "learning_rate": 2.854547342192691e-07, "loss": 0.1982, "step": 467 }, { "epoch": 0.1, "learning_rate": 2.854235880398671e-07, "loss": 0.2449, "step": 468 }, { "epoch": 0.1, "learning_rate": 2.853924418604651e-07, "loss": 0.1633, "step": 469 }, { "epoch": 0.1, "learning_rate": 2.853612956810631e-07, "loss": 0.1721, "step": 470 }, { "epoch": 0.1, "learning_rate": 2.853301495016611e-07, "loss": 0.1827, "step": 471 }, { "epoch": 0.1, "learning_rate": 2.852990033222591e-07, "loss": 0.1655, "step": 472 }, { "epoch": 0.1, "learning_rate": 2.852678571428571e-07, "loss": 0.1654, "step": 473 }, { "epoch": 0.1, "learning_rate": 2.8523671096345515e-07, "loss": 0.1964, "step": 474 }, { "epoch": 0.1, "learning_rate": 2.8520556478405314e-07, "loss": 0.187, "step": 475 }, { "epoch": 0.1, "learning_rate": 2.851744186046512e-07, "loss": 0.1365, "step": 476 }, { "epoch": 0.1, "learning_rate": 2.8514327242524917e-07, "loss": 0.1708, "step": 477 }, { "epoch": 0.1, "learning_rate": 2.8511212624584716e-07, "loss": 0.1152, "step": 478 }, { "epoch": 0.1, "learning_rate": 2.8508098006644515e-07, "loss": 0.1815, "step": 479 }, { "epoch": 0.1, "learning_rate": 2.850498338870432e-07, "loss": 0.1746, "step": 480 }, { "epoch": 0.1, "learning_rate": 2.850186877076412e-07, "loss": 0.1144, "step": 481 }, { "epoch": 0.1, "learning_rate": 2.849875415282392e-07, "loss": 0.1871, "step": 482 }, { "epoch": 0.1, "learning_rate": 2.849563953488372e-07, "loss": 0.2161, "step": 483 }, { "epoch": 0.1, "learning_rate": 2.849252491694352e-07, "loss": 0.1437, "step": 484 }, { "epoch": 0.1, "learning_rate": 2.848941029900332e-07, "loss": 0.155, "step": 485 }, { "epoch": 0.1, "learning_rate": 2.8486295681063123e-07, "loss": 0.1602, "step": 486 }, { "epoch": 0.1, "learning_rate": 2.848318106312292e-07, "loss": 0.1932, "step": 487 }, { "epoch": 0.1, "learning_rate": 2.848006644518272e-07, "loss": 0.1575, "step": 488 }, { "epoch": 0.1, "learning_rate": 2.8476951827242525e-07, "loss": 0.1775, "step": 489 }, { "epoch": 0.1, "learning_rate": 2.8473837209302324e-07, "loss": 0.166, "step": 490 }, { "epoch": 0.1, "learning_rate": 2.847072259136213e-07, "loss": 0.2082, "step": 491 }, { "epoch": 0.1, "learning_rate": 2.846760797342192e-07, "loss": 0.1362, "step": 492 }, { "epoch": 0.1, "learning_rate": 2.8464493355481726e-07, "loss": 0.1439, "step": 493 }, { "epoch": 0.1, "learning_rate": 2.8461378737541525e-07, "loss": 0.1562, "step": 494 }, { "epoch": 0.1, "learning_rate": 2.845826411960133e-07, "loss": 0.2182, "step": 495 }, { "epoch": 0.1, "learning_rate": 2.845514950166113e-07, "loss": 0.1367, "step": 496 }, { "epoch": 0.1, "learning_rate": 2.8452034883720927e-07, "loss": 0.2799, "step": 497 }, { "epoch": 0.1, "learning_rate": 2.844892026578073e-07, "loss": 0.1545, "step": 498 }, { "epoch": 0.1, "learning_rate": 2.844580564784053e-07, "loss": 0.1018, "step": 499 }, { "epoch": 0.1, "learning_rate": 2.8442691029900335e-07, "loss": 0.1826, "step": 500 }, { "epoch": 0.1, "learning_rate": 2.843957641196013e-07, "loss": 0.2225, "step": 501 }, { "epoch": 0.1, "learning_rate": 2.843646179401993e-07, "loss": 0.1845, "step": 502 }, { "epoch": 0.1, "learning_rate": 2.843334717607973e-07, "loss": 0.1724, "step": 503 }, { "epoch": 0.1, "learning_rate": 2.8430232558139536e-07, "loss": 0.1091, "step": 504 }, { "epoch": 0.1, "learning_rate": 2.8427117940199335e-07, "loss": 0.1734, "step": 505 }, { "epoch": 0.11, "learning_rate": 2.8424003322259133e-07, "loss": 0.1679, "step": 506 }, { "epoch": 0.11, "learning_rate": 2.842088870431893e-07, "loss": 0.1693, "step": 507 }, { "epoch": 0.11, "learning_rate": 2.8417774086378737e-07, "loss": 0.1962, "step": 508 }, { "epoch": 0.11, "learning_rate": 2.8414659468438536e-07, "loss": 0.1791, "step": 509 }, { "epoch": 0.11, "learning_rate": 2.841154485049834e-07, "loss": 0.1846, "step": 510 }, { "epoch": 0.11, "learning_rate": 2.840843023255814e-07, "loss": 0.1913, "step": 511 }, { "epoch": 0.11, "learning_rate": 2.840531561461794e-07, "loss": 0.2057, "step": 512 }, { "epoch": 0.11, "learning_rate": 2.840220099667774e-07, "loss": 0.1608, "step": 513 }, { "epoch": 0.11, "learning_rate": 2.839908637873754e-07, "loss": 0.1841, "step": 514 }, { "epoch": 0.11, "learning_rate": 2.839597176079734e-07, "loss": 0.1406, "step": 515 }, { "epoch": 0.11, "learning_rate": 2.839285714285714e-07, "loss": 0.1097, "step": 516 }, { "epoch": 0.11, "learning_rate": 2.8389742524916943e-07, "loss": 0.1769, "step": 517 }, { "epoch": 0.11, "learning_rate": 2.838662790697674e-07, "loss": 0.1664, "step": 518 }, { "epoch": 0.11, "learning_rate": 2.8383513289036546e-07, "loss": 0.1932, "step": 519 }, { "epoch": 0.11, "learning_rate": 2.8380398671096345e-07, "loss": 0.1454, "step": 520 }, { "epoch": 0.11, "learning_rate": 2.8377284053156144e-07, "loss": 0.1651, "step": 521 }, { "epoch": 0.11, "learning_rate": 2.8374169435215943e-07, "loss": 0.1269, "step": 522 }, { "epoch": 0.11, "learning_rate": 2.8371054817275747e-07, "loss": 0.1599, "step": 523 }, { "epoch": 0.11, "learning_rate": 2.8367940199335546e-07, "loss": 0.2095, "step": 524 }, { "epoch": 0.11, "learning_rate": 2.8364825581395345e-07, "loss": 0.1198, "step": 525 }, { "epoch": 0.11, "learning_rate": 2.836171096345515e-07, "loss": 0.1885, "step": 526 }, { "epoch": 0.11, "learning_rate": 2.835859634551495e-07, "loss": 0.184, "step": 527 }, { "epoch": 0.11, "learning_rate": 2.835548172757475e-07, "loss": 0.1796, "step": 528 }, { "epoch": 0.11, "learning_rate": 2.835236710963455e-07, "loss": 0.1437, "step": 529 }, { "epoch": 0.11, "learning_rate": 2.834925249169435e-07, "loss": 0.1675, "step": 530 }, { "epoch": 0.11, "learning_rate": 2.834613787375415e-07, "loss": 0.246, "step": 531 }, { "epoch": 0.11, "learning_rate": 2.8343023255813953e-07, "loss": 0.1302, "step": 532 }, { "epoch": 0.11, "learning_rate": 2.833990863787375e-07, "loss": 0.1914, "step": 533 }, { "epoch": 0.11, "learning_rate": 2.833679401993355e-07, "loss": 0.1559, "step": 534 }, { "epoch": 0.11, "learning_rate": 2.8333679401993356e-07, "loss": 0.1804, "step": 535 }, { "epoch": 0.11, "learning_rate": 2.8330564784053155e-07, "loss": 0.2436, "step": 536 }, { "epoch": 0.11, "learning_rate": 2.832745016611296e-07, "loss": 0.1785, "step": 537 }, { "epoch": 0.11, "learning_rate": 2.832433554817276e-07, "loss": 0.217, "step": 538 }, { "epoch": 0.11, "learning_rate": 2.8321220930232557e-07, "loss": 0.2102, "step": 539 }, { "epoch": 0.11, "learning_rate": 2.8318106312292356e-07, "loss": 0.1749, "step": 540 }, { "epoch": 0.11, "learning_rate": 2.831499169435216e-07, "loss": 0.1375, "step": 541 }, { "epoch": 0.11, "learning_rate": 2.831187707641196e-07, "loss": 0.1575, "step": 542 }, { "epoch": 0.11, "learning_rate": 2.8308762458471763e-07, "loss": 0.1252, "step": 543 }, { "epoch": 0.11, "learning_rate": 2.8305647840531557e-07, "loss": 0.1481, "step": 544 }, { "epoch": 0.11, "learning_rate": 2.830253322259136e-07, "loss": 0.2234, "step": 545 }, { "epoch": 0.11, "learning_rate": 2.829941860465116e-07, "loss": 0.1811, "step": 546 }, { "epoch": 0.11, "learning_rate": 2.8296303986710964e-07, "loss": 0.1531, "step": 547 }, { "epoch": 0.11, "learning_rate": 2.8293189368770763e-07, "loss": 0.1801, "step": 548 }, { "epoch": 0.11, "learning_rate": 2.829007475083056e-07, "loss": 0.1694, "step": 549 }, { "epoch": 0.11, "learning_rate": 2.8286960132890366e-07, "loss": 0.1697, "step": 550 }, { "epoch": 0.11, "learning_rate": 2.8283845514950165e-07, "loss": 0.1911, "step": 551 }, { "epoch": 0.11, "learning_rate": 2.828073089700997e-07, "loss": 0.1595, "step": 552 }, { "epoch": 0.11, "learning_rate": 2.8277616279069763e-07, "loss": 0.2127, "step": 553 }, { "epoch": 0.12, "learning_rate": 2.8274501661129567e-07, "loss": 0.2029, "step": 554 }, { "epoch": 0.12, "learning_rate": 2.8271387043189366e-07, "loss": 0.2082, "step": 555 }, { "epoch": 0.12, "learning_rate": 2.826827242524917e-07, "loss": 0.148, "step": 556 }, { "epoch": 0.12, "learning_rate": 2.826515780730897e-07, "loss": 0.1892, "step": 557 }, { "epoch": 0.12, "learning_rate": 2.826204318936877e-07, "loss": 0.2063, "step": 558 }, { "epoch": 0.12, "learning_rate": 2.8258928571428567e-07, "loss": 0.1648, "step": 559 }, { "epoch": 0.12, "learning_rate": 2.825581395348837e-07, "loss": 0.1605, "step": 560 }, { "epoch": 0.12, "learning_rate": 2.825269933554817e-07, "loss": 0.1616, "step": 561 }, { "epoch": 0.12, "learning_rate": 2.8249584717607974e-07, "loss": 0.2142, "step": 562 }, { "epoch": 0.12, "learning_rate": 2.8246470099667773e-07, "loss": 0.1512, "step": 563 }, { "epoch": 0.12, "learning_rate": 2.824335548172757e-07, "loss": 0.1383, "step": 564 }, { "epoch": 0.12, "learning_rate": 2.8240240863787377e-07, "loss": 0.1773, "step": 565 }, { "epoch": 0.12, "learning_rate": 2.8237126245847176e-07, "loss": 0.21, "step": 566 }, { "epoch": 0.12, "learning_rate": 2.8234011627906974e-07, "loss": 0.2213, "step": 567 }, { "epoch": 0.12, "learning_rate": 2.8230897009966773e-07, "loss": 0.1648, "step": 568 }, { "epoch": 0.12, "learning_rate": 2.822778239202658e-07, "loss": 0.1402, "step": 569 }, { "epoch": 0.12, "learning_rate": 2.8224667774086377e-07, "loss": 0.1311, "step": 570 }, { "epoch": 0.12, "learning_rate": 2.822155315614618e-07, "loss": 0.1461, "step": 571 }, { "epoch": 0.12, "learning_rate": 2.821843853820598e-07, "loss": 0.1717, "step": 572 }, { "epoch": 0.12, "learning_rate": 2.821532392026578e-07, "loss": 0.1532, "step": 573 }, { "epoch": 0.12, "learning_rate": 2.8212209302325583e-07, "loss": 0.1343, "step": 574 }, { "epoch": 0.12, "learning_rate": 2.820909468438538e-07, "loss": 0.1742, "step": 575 }, { "epoch": 0.12, "learning_rate": 2.820598006644518e-07, "loss": 0.1438, "step": 576 }, { "epoch": 0.12, "learning_rate": 2.820286544850498e-07, "loss": 0.2, "step": 577 }, { "epoch": 0.12, "learning_rate": 2.8199750830564784e-07, "loss": 0.2429, "step": 578 }, { "epoch": 0.12, "learning_rate": 2.8196636212624583e-07, "loss": 0.1711, "step": 579 }, { "epoch": 0.12, "learning_rate": 2.8193521594684387e-07, "loss": 0.1666, "step": 580 }, { "epoch": 0.12, "learning_rate": 2.819040697674418e-07, "loss": 0.1603, "step": 581 }, { "epoch": 0.12, "learning_rate": 2.8187292358803985e-07, "loss": 0.1592, "step": 582 }, { "epoch": 0.12, "learning_rate": 2.8184177740863784e-07, "loss": 0.1392, "step": 583 }, { "epoch": 0.12, "learning_rate": 2.818106312292359e-07, "loss": 0.2062, "step": 584 }, { "epoch": 0.12, "learning_rate": 2.8177948504983387e-07, "loss": 0.1602, "step": 585 }, { "epoch": 0.12, "learning_rate": 2.8174833887043186e-07, "loss": 0.1136, "step": 586 }, { "epoch": 0.12, "learning_rate": 2.817171926910299e-07, "loss": 0.1481, "step": 587 }, { "epoch": 0.12, "learning_rate": 2.816860465116279e-07, "loss": 0.1985, "step": 588 }, { "epoch": 0.12, "learning_rate": 2.8165490033222593e-07, "loss": 0.1734, "step": 589 }, { "epoch": 0.12, "learning_rate": 2.816237541528239e-07, "loss": 0.213, "step": 590 }, { "epoch": 0.12, "learning_rate": 2.815926079734219e-07, "loss": 0.1702, "step": 591 }, { "epoch": 0.12, "learning_rate": 2.815614617940199e-07, "loss": 0.2619, "step": 592 }, { "epoch": 0.12, "learning_rate": 2.8153031561461794e-07, "loss": 0.1575, "step": 593 }, { "epoch": 0.12, "learning_rate": 2.8149916943521593e-07, "loss": 0.181, "step": 594 }, { "epoch": 0.12, "learning_rate": 2.814680232558139e-07, "loss": 0.1772, "step": 595 }, { "epoch": 0.12, "learning_rate": 2.814368770764119e-07, "loss": 0.1634, "step": 596 }, { "epoch": 0.12, "learning_rate": 2.8140573089700995e-07, "loss": 0.197, "step": 597 }, { "epoch": 0.12, "learning_rate": 2.8137458471760794e-07, "loss": 0.1453, "step": 598 }, { "epoch": 0.12, "learning_rate": 2.81343438538206e-07, "loss": 0.2154, "step": 599 }, { "epoch": 0.12, "learning_rate": 2.81312292358804e-07, "loss": 0.2407, "step": 600 }, { "epoch": 0.12, "learning_rate": 2.8128114617940196e-07, "loss": 0.2004, "step": 601 }, { "epoch": 0.12, "learning_rate": 2.8125e-07, "loss": 0.1743, "step": 602 }, { "epoch": 0.13, "learning_rate": 2.81218853820598e-07, "loss": 0.1804, "step": 603 }, { "epoch": 0.13, "learning_rate": 2.8118770764119604e-07, "loss": 0.2512, "step": 604 }, { "epoch": 0.13, "learning_rate": 2.81156561461794e-07, "loss": 0.2213, "step": 605 }, { "epoch": 0.13, "learning_rate": 2.81125415282392e-07, "loss": 0.173, "step": 606 }, { "epoch": 0.13, "learning_rate": 2.8109426910299e-07, "loss": 0.1605, "step": 607 }, { "epoch": 0.13, "learning_rate": 2.8106312292358805e-07, "loss": 0.1915, "step": 608 }, { "epoch": 0.13, "learning_rate": 2.8103197674418604e-07, "loss": 0.1905, "step": 609 }, { "epoch": 0.13, "learning_rate": 2.8100083056478403e-07, "loss": 0.1711, "step": 610 }, { "epoch": 0.13, "learning_rate": 2.8096968438538207e-07, "loss": 0.1497, "step": 611 }, { "epoch": 0.13, "learning_rate": 2.8093853820598006e-07, "loss": 0.1147, "step": 612 }, { "epoch": 0.13, "learning_rate": 2.8090739202657805e-07, "loss": 0.2141, "step": 613 }, { "epoch": 0.13, "learning_rate": 2.8087624584717604e-07, "loss": 0.1224, "step": 614 }, { "epoch": 0.13, "learning_rate": 2.808450996677741e-07, "loss": 0.2912, "step": 615 }, { "epoch": 0.13, "learning_rate": 2.8081395348837207e-07, "loss": 0.1362, "step": 616 }, { "epoch": 0.13, "learning_rate": 2.807828073089701e-07, "loss": 0.1954, "step": 617 }, { "epoch": 0.13, "learning_rate": 2.807516611295681e-07, "loss": 0.2146, "step": 618 }, { "epoch": 0.13, "learning_rate": 2.807205149501661e-07, "loss": 0.2052, "step": 619 }, { "epoch": 0.13, "learning_rate": 2.806893687707641e-07, "loss": 0.1426, "step": 620 }, { "epoch": 0.13, "learning_rate": 2.806582225913621e-07, "loss": 0.1624, "step": 621 }, { "epoch": 0.13, "learning_rate": 2.806270764119601e-07, "loss": 0.2631, "step": 622 }, { "epoch": 0.13, "learning_rate": 2.8059593023255815e-07, "loss": 0.1749, "step": 623 }, { "epoch": 0.13, "learning_rate": 2.8056478405315614e-07, "loss": 0.1986, "step": 624 }, { "epoch": 0.13, "learning_rate": 2.8053363787375413e-07, "loss": 0.1956, "step": 625 }, { "epoch": 0.13, "learning_rate": 2.805024916943522e-07, "loss": 0.1535, "step": 626 }, { "epoch": 0.13, "learning_rate": 2.8047134551495016e-07, "loss": 0.2193, "step": 627 }, { "epoch": 0.13, "learning_rate": 2.8044019933554815e-07, "loss": 0.1634, "step": 628 }, { "epoch": 0.13, "learning_rate": 2.8040905315614614e-07, "loss": 0.1839, "step": 629 }, { "epoch": 0.13, "learning_rate": 2.803779069767442e-07, "loss": 0.1979, "step": 630 }, { "epoch": 0.13, "learning_rate": 2.803467607973422e-07, "loss": 0.2199, "step": 631 }, { "epoch": 0.13, "learning_rate": 2.803156146179402e-07, "loss": 0.1806, "step": 632 }, { "epoch": 0.13, "learning_rate": 2.8028446843853815e-07, "loss": 0.1928, "step": 633 }, { "epoch": 0.13, "learning_rate": 2.802533222591362e-07, "loss": 0.1457, "step": 634 }, { "epoch": 0.13, "learning_rate": 2.802221760797342e-07, "loss": 0.1839, "step": 635 }, { "epoch": 0.13, "learning_rate": 2.8019102990033223e-07, "loss": 0.1997, "step": 636 }, { "epoch": 0.13, "learning_rate": 2.801598837209302e-07, "loss": 0.1651, "step": 637 }, { "epoch": 0.13, "learning_rate": 2.801287375415282e-07, "loss": 0.258, "step": 638 }, { "epoch": 0.13, "learning_rate": 2.8009759136212625e-07, "loss": 0.2524, "step": 639 }, { "epoch": 0.13, "learning_rate": 2.8006644518272424e-07, "loss": 0.1923, "step": 640 }, { "epoch": 0.13, "learning_rate": 2.800352990033223e-07, "loss": 0.1457, "step": 641 }, { "epoch": 0.13, "learning_rate": 2.800041528239202e-07, "loss": 0.1306, "step": 642 }, { "epoch": 0.13, "learning_rate": 2.7997300664451826e-07, "loss": 0.1747, "step": 643 }, { "epoch": 0.13, "learning_rate": 2.7994186046511625e-07, "loss": 0.1161, "step": 644 }, { "epoch": 0.13, "learning_rate": 2.799107142857143e-07, "loss": 0.1664, "step": 645 }, { "epoch": 0.13, "learning_rate": 2.798795681063123e-07, "loss": 0.1713, "step": 646 }, { "epoch": 0.13, "learning_rate": 2.7984842192691027e-07, "loss": 0.1665, "step": 647 }, { "epoch": 0.13, "learning_rate": 2.798172757475083e-07, "loss": 0.1564, "step": 648 }, { "epoch": 0.13, "learning_rate": 2.797861295681063e-07, "loss": 0.1923, "step": 649 }, { "epoch": 0.13, "learning_rate": 2.797549833887043e-07, "loss": 0.1762, "step": 650 }, { "epoch": 0.14, "learning_rate": 2.7972383720930233e-07, "loss": 0.1591, "step": 651 }, { "epoch": 0.14, "learning_rate": 2.796926910299003e-07, "loss": 0.1718, "step": 652 }, { "epoch": 0.14, "learning_rate": 2.796615448504983e-07, "loss": 0.1626, "step": 653 }, { "epoch": 0.14, "learning_rate": 2.7963039867109635e-07, "loss": 0.1652, "step": 654 }, { "epoch": 0.14, "learning_rate": 2.7959925249169434e-07, "loss": 0.1088, "step": 655 }, { "epoch": 0.14, "learning_rate": 2.795681063122924e-07, "loss": 0.1551, "step": 656 }, { "epoch": 0.14, "learning_rate": 2.795369601328903e-07, "loss": 0.189, "step": 657 }, { "epoch": 0.14, "learning_rate": 2.7950581395348836e-07, "loss": 0.1838, "step": 658 }, { "epoch": 0.14, "learning_rate": 2.7947466777408635e-07, "loss": 0.1823, "step": 659 }, { "epoch": 0.14, "learning_rate": 2.794435215946844e-07, "loss": 0.1831, "step": 660 }, { "epoch": 0.14, "learning_rate": 2.794123754152824e-07, "loss": 0.2242, "step": 661 }, { "epoch": 0.14, "learning_rate": 2.793812292358804e-07, "loss": 0.2088, "step": 662 }, { "epoch": 0.14, "learning_rate": 2.793500830564784e-07, "loss": 0.2243, "step": 663 }, { "epoch": 0.14, "learning_rate": 2.793189368770764e-07, "loss": 0.2242, "step": 664 }, { "epoch": 0.14, "learning_rate": 2.792877906976744e-07, "loss": 0.1928, "step": 665 }, { "epoch": 0.14, "learning_rate": 2.792566445182724e-07, "loss": 0.1501, "step": 666 }, { "epoch": 0.14, "learning_rate": 2.7922549833887043e-07, "loss": 0.1628, "step": 667 }, { "epoch": 0.14, "learning_rate": 2.791943521594684e-07, "loss": 0.1854, "step": 668 }, { "epoch": 0.14, "learning_rate": 2.7916320598006646e-07, "loss": 0.1568, "step": 669 }, { "epoch": 0.14, "learning_rate": 2.7913205980066445e-07, "loss": 0.1418, "step": 670 }, { "epoch": 0.14, "learning_rate": 2.7910091362126244e-07, "loss": 0.1636, "step": 671 }, { "epoch": 0.14, "learning_rate": 2.7906976744186043e-07, "loss": 0.1827, "step": 672 }, { "epoch": 0.14, "learning_rate": 2.7903862126245847e-07, "loss": 0.1894, "step": 673 }, { "epoch": 0.14, "learning_rate": 2.7900747508305646e-07, "loss": 0.1969, "step": 674 }, { "epoch": 0.14, "learning_rate": 2.7897632890365445e-07, "loss": 0.1535, "step": 675 }, { "epoch": 0.14, "learning_rate": 2.789451827242525e-07, "loss": 0.1887, "step": 676 }, { "epoch": 0.14, "learning_rate": 2.789140365448505e-07, "loss": 0.175, "step": 677 }, { "epoch": 0.14, "learning_rate": 2.788828903654485e-07, "loss": 0.1796, "step": 678 }, { "epoch": 0.14, "learning_rate": 2.788517441860465e-07, "loss": 0.1625, "step": 679 }, { "epoch": 0.14, "learning_rate": 2.788205980066445e-07, "loss": 0.1518, "step": 680 }, { "epoch": 0.14, "learning_rate": 2.787894518272425e-07, "loss": 0.218, "step": 681 }, { "epoch": 0.14, "learning_rate": 2.7875830564784053e-07, "loss": 0.1559, "step": 682 }, { "epoch": 0.14, "learning_rate": 2.787271594684385e-07, "loss": 0.1494, "step": 683 }, { "epoch": 0.14, "learning_rate": 2.7869601328903656e-07, "loss": 0.1443, "step": 684 }, { "epoch": 0.14, "learning_rate": 2.7866486710963455e-07, "loss": 0.1805, "step": 685 }, { "epoch": 0.14, "learning_rate": 2.7863372093023254e-07, "loss": 0.1728, "step": 686 }, { "epoch": 0.14, "learning_rate": 2.7860257475083053e-07, "loss": 0.1447, "step": 687 }, { "epoch": 0.14, "learning_rate": 2.785714285714286e-07, "loss": 0.2093, "step": 688 }, { "epoch": 0.14, "learning_rate": 2.7854028239202656e-07, "loss": 0.24, "step": 689 }, { "epoch": 0.14, "learning_rate": 2.7850913621262455e-07, "loss": 0.1498, "step": 690 }, { "epoch": 0.14, "learning_rate": 2.784779900332226e-07, "loss": 0.1698, "step": 691 }, { "epoch": 0.14, "learning_rate": 2.784468438538206e-07, "loss": 0.186, "step": 692 }, { "epoch": 0.14, "learning_rate": 2.7841569767441863e-07, "loss": 0.1971, "step": 693 }, { "epoch": 0.14, "learning_rate": 2.7838455149501656e-07, "loss": 0.1641, "step": 694 }, { "epoch": 0.14, "learning_rate": 2.783534053156146e-07, "loss": 0.2706, "step": 695 }, { "epoch": 0.14, "learning_rate": 2.783222591362126e-07, "loss": 0.1358, "step": 696 }, { "epoch": 0.14, "learning_rate": 2.7829111295681064e-07, "loss": 0.1283, "step": 697 }, { "epoch": 0.14, "learning_rate": 2.782599667774086e-07, "loss": 0.184, "step": 698 }, { "epoch": 0.15, "learning_rate": 2.782288205980066e-07, "loss": 0.1926, "step": 699 }, { "epoch": 0.15, "learning_rate": 2.7819767441860466e-07, "loss": 0.1042, "step": 700 }, { "epoch": 0.15, "learning_rate": 2.7816652823920265e-07, "loss": 0.1646, "step": 701 }, { "epoch": 0.15, "learning_rate": 2.7813538205980064e-07, "loss": 0.1466, "step": 702 }, { "epoch": 0.15, "learning_rate": 2.781042358803987e-07, "loss": 0.2288, "step": 703 }, { "epoch": 0.15, "learning_rate": 2.7807308970099667e-07, "loss": 0.2149, "step": 704 }, { "epoch": 0.15, "learning_rate": 2.7804194352159466e-07, "loss": 0.1065, "step": 705 }, { "epoch": 0.15, "learning_rate": 2.780107973421927e-07, "loss": 0.2928, "step": 706 }, { "epoch": 0.15, "learning_rate": 2.779796511627907e-07, "loss": 0.1401, "step": 707 }, { "epoch": 0.15, "learning_rate": 2.779485049833887e-07, "loss": 0.143, "step": 708 }, { "epoch": 0.15, "learning_rate": 2.7791735880398667e-07, "loss": 0.1916, "step": 709 }, { "epoch": 0.15, "learning_rate": 2.778862126245847e-07, "loss": 0.1405, "step": 710 }, { "epoch": 0.15, "learning_rate": 2.778550664451827e-07, "loss": 0.1803, "step": 711 }, { "epoch": 0.15, "learning_rate": 2.7782392026578074e-07, "loss": 0.1824, "step": 712 }, { "epoch": 0.15, "learning_rate": 2.7779277408637873e-07, "loss": 0.2141, "step": 713 }, { "epoch": 0.15, "learning_rate": 2.777616279069767e-07, "loss": 0.1626, "step": 714 }, { "epoch": 0.15, "learning_rate": 2.7773048172757476e-07, "loss": 0.2095, "step": 715 }, { "epoch": 0.15, "learning_rate": 2.7769933554817275e-07, "loss": 0.2041, "step": 716 }, { "epoch": 0.15, "learning_rate": 2.7766818936877074e-07, "loss": 0.2089, "step": 717 }, { "epoch": 0.15, "learning_rate": 2.7763704318936873e-07, "loss": 0.1499, "step": 718 }, { "epoch": 0.15, "learning_rate": 2.7760589700996677e-07, "loss": 0.2059, "step": 719 }, { "epoch": 0.15, "learning_rate": 2.7757475083056476e-07, "loss": 0.1227, "step": 720 }, { "epoch": 0.15, "learning_rate": 2.775436046511628e-07, "loss": 0.1444, "step": 721 }, { "epoch": 0.15, "learning_rate": 2.775124584717608e-07, "loss": 0.1966, "step": 722 }, { "epoch": 0.15, "learning_rate": 2.774813122923588e-07, "loss": 0.1444, "step": 723 }, { "epoch": 0.15, "learning_rate": 2.7745016611295677e-07, "loss": 0.1807, "step": 724 }, { "epoch": 0.15, "learning_rate": 2.774190199335548e-07, "loss": 0.1931, "step": 725 }, { "epoch": 0.15, "learning_rate": 2.773878737541528e-07, "loss": 0.1986, "step": 726 }, { "epoch": 0.15, "learning_rate": 2.773567275747508e-07, "loss": 0.1858, "step": 727 }, { "epoch": 0.15, "learning_rate": 2.7732558139534884e-07, "loss": 0.2014, "step": 728 }, { "epoch": 0.15, "learning_rate": 2.772944352159468e-07, "loss": 0.1825, "step": 729 }, { "epoch": 0.15, "learning_rate": 2.7726328903654487e-07, "loss": 0.1284, "step": 730 }, { "epoch": 0.15, "learning_rate": 2.7723214285714286e-07, "loss": 0.1575, "step": 731 }, { "epoch": 0.15, "learning_rate": 2.7720099667774085e-07, "loss": 0.1679, "step": 732 }, { "epoch": 0.15, "learning_rate": 2.7716985049833884e-07, "loss": 0.1696, "step": 733 }, { "epoch": 0.15, "learning_rate": 2.771387043189369e-07, "loss": 0.1909, "step": 734 }, { "epoch": 0.15, "learning_rate": 2.7710755813953487e-07, "loss": 0.2503, "step": 735 }, { "epoch": 0.15, "learning_rate": 2.770764119601329e-07, "loss": 0.1458, "step": 736 }, { "epoch": 0.15, "learning_rate": 2.770452657807309e-07, "loss": 0.1263, "step": 737 }, { "epoch": 0.15, "learning_rate": 2.770141196013289e-07, "loss": 0.1808, "step": 738 }, { "epoch": 0.15, "learning_rate": 2.769829734219269e-07, "loss": 0.1692, "step": 739 }, { "epoch": 0.15, "learning_rate": 2.769518272425249e-07, "loss": 0.1587, "step": 740 }, { "epoch": 0.15, "learning_rate": 2.769206810631229e-07, "loss": 0.1614, "step": 741 }, { "epoch": 0.15, "learning_rate": 2.768895348837209e-07, "loss": 0.191, "step": 742 }, { "epoch": 0.15, "learning_rate": 2.7685838870431894e-07, "loss": 0.144, "step": 743 }, { "epoch": 0.15, "learning_rate": 2.7682724252491693e-07, "loss": 0.2159, "step": 744 }, { "epoch": 0.15, "learning_rate": 2.7679609634551497e-07, "loss": 0.1868, "step": 745 }, { "epoch": 0.15, "learning_rate": 2.767649501661129e-07, "loss": 0.191, "step": 746 }, { "epoch": 0.16, "learning_rate": 2.7673380398671095e-07, "loss": 0.151, "step": 747 }, { "epoch": 0.16, "learning_rate": 2.7670265780730894e-07, "loss": 0.1891, "step": 748 }, { "epoch": 0.16, "learning_rate": 2.76671511627907e-07, "loss": 0.1667, "step": 749 }, { "epoch": 0.16, "learning_rate": 2.7664036544850497e-07, "loss": 0.1433, "step": 750 }, { "epoch": 0.16, "learning_rate": 2.7660921926910296e-07, "loss": 0.1148, "step": 751 }, { "epoch": 0.16, "learning_rate": 2.76578073089701e-07, "loss": 0.2143, "step": 752 }, { "epoch": 0.16, "learning_rate": 2.76546926910299e-07, "loss": 0.1891, "step": 753 }, { "epoch": 0.16, "learning_rate": 2.76515780730897e-07, "loss": 0.1526, "step": 754 }, { "epoch": 0.16, "learning_rate": 2.7648463455149497e-07, "loss": 0.1289, "step": 755 }, { "epoch": 0.16, "learning_rate": 2.76453488372093e-07, "loss": 0.2119, "step": 756 }, { "epoch": 0.16, "learning_rate": 2.76422342192691e-07, "loss": 0.1706, "step": 757 }, { "epoch": 0.16, "learning_rate": 2.7639119601328905e-07, "loss": 0.1838, "step": 758 }, { "epoch": 0.16, "learning_rate": 2.7636004983388704e-07, "loss": 0.1552, "step": 759 }, { "epoch": 0.16, "learning_rate": 2.76328903654485e-07, "loss": 0.1167, "step": 760 }, { "epoch": 0.16, "learning_rate": 2.76297757475083e-07, "loss": 0.2545, "step": 761 }, { "epoch": 0.16, "learning_rate": 2.7626661129568106e-07, "loss": 0.1414, "step": 762 }, { "epoch": 0.16, "learning_rate": 2.7623546511627905e-07, "loss": 0.1455, "step": 763 }, { "epoch": 0.16, "learning_rate": 2.762043189368771e-07, "loss": 0.2005, "step": 764 }, { "epoch": 0.16, "learning_rate": 2.761731727574751e-07, "loss": 0.1374, "step": 765 }, { "epoch": 0.16, "learning_rate": 2.7614202657807307e-07, "loss": 0.2083, "step": 766 }, { "epoch": 0.16, "learning_rate": 2.761108803986711e-07, "loss": 0.1558, "step": 767 }, { "epoch": 0.16, "learning_rate": 2.760797342192691e-07, "loss": 0.1728, "step": 768 }, { "epoch": 0.16, "learning_rate": 2.760485880398671e-07, "loss": 0.1829, "step": 769 }, { "epoch": 0.16, "learning_rate": 2.760174418604651e-07, "loss": 0.1368, "step": 770 }, { "epoch": 0.16, "learning_rate": 2.759862956810631e-07, "loss": 0.2228, "step": 771 }, { "epoch": 0.16, "learning_rate": 2.759551495016611e-07, "loss": 0.1453, "step": 772 }, { "epoch": 0.16, "learning_rate": 2.7592400332225915e-07, "loss": 0.192, "step": 773 }, { "epoch": 0.16, "learning_rate": 2.7589285714285714e-07, "loss": 0.1657, "step": 774 }, { "epoch": 0.16, "learning_rate": 2.7586171096345513e-07, "loss": 0.1451, "step": 775 }, { "epoch": 0.16, "learning_rate": 2.758305647840531e-07, "loss": 0.1531, "step": 776 }, { "epoch": 0.16, "learning_rate": 2.7579941860465116e-07, "loss": 0.1616, "step": 777 }, { "epoch": 0.16, "learning_rate": 2.7576827242524915e-07, "loss": 0.1895, "step": 778 }, { "epoch": 0.16, "learning_rate": 2.7573712624584714e-07, "loss": 0.1609, "step": 779 }, { "epoch": 0.16, "learning_rate": 2.757059800664452e-07, "loss": 0.1953, "step": 780 }, { "epoch": 0.16, "learning_rate": 2.7567483388704317e-07, "loss": 0.149, "step": 781 }, { "epoch": 0.16, "learning_rate": 2.756436877076412e-07, "loss": 0.1298, "step": 782 }, { "epoch": 0.16, "learning_rate": 2.756125415282392e-07, "loss": 0.2103, "step": 783 }, { "epoch": 0.16, "learning_rate": 2.755813953488372e-07, "loss": 0.2062, "step": 784 }, { "epoch": 0.16, "learning_rate": 2.755502491694352e-07, "loss": 0.1831, "step": 785 }, { "epoch": 0.16, "learning_rate": 2.755191029900332e-07, "loss": 0.1857, "step": 786 }, { "epoch": 0.16, "learning_rate": 2.754879568106312e-07, "loss": 0.1494, "step": 787 }, { "epoch": 0.16, "learning_rate": 2.754568106312292e-07, "loss": 0.2047, "step": 788 }, { "epoch": 0.16, "learning_rate": 2.7542566445182725e-07, "loss": 0.2077, "step": 789 }, { "epoch": 0.16, "learning_rate": 2.7539451827242524e-07, "loss": 0.1583, "step": 790 }, { "epoch": 0.16, "learning_rate": 2.753633720930232e-07, "loss": 0.2394, "step": 791 }, { "epoch": 0.16, "learning_rate": 2.7533222591362127e-07, "loss": 0.168, "step": 792 }, { "epoch": 0.16, "learning_rate": 2.7530107973421926e-07, "loss": 0.2125, "step": 793 }, { "epoch": 0.16, "learning_rate": 2.7526993355481725e-07, "loss": 0.1611, "step": 794 }, { "epoch": 0.17, "learning_rate": 2.752387873754153e-07, "loss": 0.1725, "step": 795 }, { "epoch": 0.17, "learning_rate": 2.752076411960133e-07, "loss": 0.1591, "step": 796 }, { "epoch": 0.17, "learning_rate": 2.751764950166113e-07, "loss": 0.1472, "step": 797 }, { "epoch": 0.17, "learning_rate": 2.7514534883720926e-07, "loss": 0.1718, "step": 798 }, { "epoch": 0.17, "learning_rate": 2.751142026578073e-07, "loss": 0.1669, "step": 799 }, { "epoch": 0.17, "learning_rate": 2.750830564784053e-07, "loss": 0.1688, "step": 800 }, { "epoch": 0.17, "learning_rate": 2.7505191029900333e-07, "loss": 0.134, "step": 801 }, { "epoch": 0.17, "learning_rate": 2.750207641196013e-07, "loss": 0.1639, "step": 802 }, { "epoch": 0.17, "learning_rate": 2.749896179401993e-07, "loss": 0.1401, "step": 803 }, { "epoch": 0.17, "learning_rate": 2.7495847176079735e-07, "loss": 0.0903, "step": 804 }, { "epoch": 0.17, "learning_rate": 2.7492732558139534e-07, "loss": 0.1738, "step": 805 }, { "epoch": 0.17, "learning_rate": 2.7489617940199333e-07, "loss": 0.1696, "step": 806 }, { "epoch": 0.17, "learning_rate": 2.748650332225913e-07, "loss": 0.153, "step": 807 }, { "epoch": 0.17, "learning_rate": 2.7483388704318936e-07, "loss": 0.0962, "step": 808 }, { "epoch": 0.17, "learning_rate": 2.7480274086378735e-07, "loss": 0.1886, "step": 809 }, { "epoch": 0.17, "learning_rate": 2.747715946843854e-07, "loss": 0.1942, "step": 810 }, { "epoch": 0.17, "learning_rate": 2.747404485049834e-07, "loss": 0.1255, "step": 811 }, { "epoch": 0.17, "learning_rate": 2.7470930232558137e-07, "loss": 0.2046, "step": 812 }, { "epoch": 0.17, "learning_rate": 2.7467815614617936e-07, "loss": 0.1637, "step": 813 }, { "epoch": 0.17, "learning_rate": 2.746470099667774e-07, "loss": 0.1994, "step": 814 }, { "epoch": 0.17, "learning_rate": 2.746158637873754e-07, "loss": 0.1863, "step": 815 }, { "epoch": 0.17, "learning_rate": 2.745847176079734e-07, "loss": 0.2429, "step": 816 }, { "epoch": 0.17, "learning_rate": 2.745535714285714e-07, "loss": 0.1464, "step": 817 }, { "epoch": 0.17, "learning_rate": 2.745224252491694e-07, "loss": 0.1596, "step": 818 }, { "epoch": 0.17, "learning_rate": 2.7449127906976746e-07, "loss": 0.1895, "step": 819 }, { "epoch": 0.17, "learning_rate": 2.7446013289036545e-07, "loss": 0.1713, "step": 820 }, { "epoch": 0.17, "learning_rate": 2.7442898671096343e-07, "loss": 0.2277, "step": 821 }, { "epoch": 0.17, "learning_rate": 2.743978405315614e-07, "loss": 0.1777, "step": 822 }, { "epoch": 0.17, "learning_rate": 2.7436669435215947e-07, "loss": 0.1526, "step": 823 }, { "epoch": 0.17, "learning_rate": 2.7433554817275746e-07, "loss": 0.1751, "step": 824 }, { "epoch": 0.17, "learning_rate": 2.743044019933555e-07, "loss": 0.2005, "step": 825 }, { "epoch": 0.17, "learning_rate": 2.742732558139535e-07, "loss": 0.1648, "step": 826 }, { "epoch": 0.17, "learning_rate": 2.742421096345515e-07, "loss": 0.2114, "step": 827 }, { "epoch": 0.17, "learning_rate": 2.7421096345514947e-07, "loss": 0.1224, "step": 828 }, { "epoch": 0.17, "learning_rate": 2.741798172757475e-07, "loss": 0.1517, "step": 829 }, { "epoch": 0.17, "learning_rate": 2.741486710963455e-07, "loss": 0.1757, "step": 830 }, { "epoch": 0.17, "learning_rate": 2.741175249169435e-07, "loss": 0.1556, "step": 831 }, { "epoch": 0.17, "learning_rate": 2.7408637873754153e-07, "loss": 0.2084, "step": 832 }, { "epoch": 0.17, "learning_rate": 2.740552325581395e-07, "loss": 0.1421, "step": 833 }, { "epoch": 0.17, "learning_rate": 2.7402408637873756e-07, "loss": 0.1148, "step": 834 }, { "epoch": 0.17, "learning_rate": 2.739929401993355e-07, "loss": 0.1572, "step": 835 }, { "epoch": 0.17, "learning_rate": 2.7396179401993354e-07, "loss": 0.1808, "step": 836 }, { "epoch": 0.17, "learning_rate": 2.7393064784053153e-07, "loss": 0.2134, "step": 837 }, { "epoch": 0.17, "learning_rate": 2.7389950166112957e-07, "loss": 0.2281, "step": 838 }, { "epoch": 0.17, "learning_rate": 2.7386835548172756e-07, "loss": 0.159, "step": 839 }, { "epoch": 0.17, "learning_rate": 2.7383720930232555e-07, "loss": 0.1973, "step": 840 }, { "epoch": 0.17, "learning_rate": 2.738060631229236e-07, "loss": 0.2007, "step": 841 }, { "epoch": 0.17, "learning_rate": 2.737749169435216e-07, "loss": 0.1636, "step": 842 }, { "epoch": 0.18, "learning_rate": 2.7374377076411957e-07, "loss": 0.1397, "step": 843 }, { "epoch": 0.18, "learning_rate": 2.737126245847176e-07, "loss": 0.2228, "step": 844 }, { "epoch": 0.18, "learning_rate": 2.736814784053156e-07, "loss": 0.1146, "step": 845 }, { "epoch": 0.18, "learning_rate": 2.736503322259136e-07, "loss": 0.2355, "step": 846 }, { "epoch": 0.18, "learning_rate": 2.7361918604651163e-07, "loss": 0.1867, "step": 847 }, { "epoch": 0.18, "learning_rate": 2.735880398671096e-07, "loss": 0.1636, "step": 848 }, { "epoch": 0.18, "learning_rate": 2.735568936877076e-07, "loss": 0.2024, "step": 849 }, { "epoch": 0.18, "learning_rate": 2.735257475083056e-07, "loss": 0.247, "step": 850 }, { "epoch": 0.18, "learning_rate": 2.7349460132890364e-07, "loss": 0.1789, "step": 851 }, { "epoch": 0.18, "learning_rate": 2.7346345514950163e-07, "loss": 0.1726, "step": 852 }, { "epoch": 0.18, "learning_rate": 2.734323089700997e-07, "loss": 0.1695, "step": 853 }, { "epoch": 0.18, "learning_rate": 2.7340116279069767e-07, "loss": 0.1598, "step": 854 }, { "epoch": 0.18, "learning_rate": 2.7337001661129565e-07, "loss": 0.1408, "step": 855 }, { "epoch": 0.18, "learning_rate": 2.733388704318937e-07, "loss": 0.1487, "step": 856 }, { "epoch": 0.18, "learning_rate": 2.733077242524917e-07, "loss": 0.1529, "step": 857 }, { "epoch": 0.18, "learning_rate": 2.7327657807308973e-07, "loss": 0.1826, "step": 858 }, { "epoch": 0.18, "learning_rate": 2.7324543189368767e-07, "loss": 0.1622, "step": 859 }, { "epoch": 0.18, "learning_rate": 2.732142857142857e-07, "loss": 0.187, "step": 860 }, { "epoch": 0.18, "learning_rate": 2.731831395348837e-07, "loss": 0.2008, "step": 861 }, { "epoch": 0.18, "learning_rate": 2.7315199335548174e-07, "loss": 0.1266, "step": 862 }, { "epoch": 0.18, "learning_rate": 2.7312084717607973e-07, "loss": 0.2023, "step": 863 }, { "epoch": 0.18, "learning_rate": 2.730897009966777e-07, "loss": 0.1453, "step": 864 }, { "epoch": 0.18, "learning_rate": 2.730585548172757e-07, "loss": 0.1369, "step": 865 }, { "epoch": 0.18, "learning_rate": 2.7302740863787375e-07, "loss": 0.156, "step": 866 }, { "epoch": 0.18, "learning_rate": 2.7299626245847174e-07, "loss": 0.1662, "step": 867 }, { "epoch": 0.18, "learning_rate": 2.7296511627906973e-07, "loss": 0.2089, "step": 868 }, { "epoch": 0.18, "learning_rate": 2.7293397009966777e-07, "loss": 0.2111, "step": 869 }, { "epoch": 0.18, "learning_rate": 2.7290282392026576e-07, "loss": 0.2435, "step": 870 }, { "epoch": 0.18, "learning_rate": 2.728716777408638e-07, "loss": 0.2131, "step": 871 }, { "epoch": 0.18, "learning_rate": 2.728405315614618e-07, "loss": 0.1558, "step": 872 }, { "epoch": 0.18, "learning_rate": 2.728093853820598e-07, "loss": 0.1556, "step": 873 }, { "epoch": 0.18, "learning_rate": 2.7277823920265777e-07, "loss": 0.1277, "step": 874 }, { "epoch": 0.18, "learning_rate": 2.727470930232558e-07, "loss": 0.177, "step": 875 }, { "epoch": 0.18, "learning_rate": 2.727159468438538e-07, "loss": 0.227, "step": 876 }, { "epoch": 0.18, "learning_rate": 2.7268480066445184e-07, "loss": 0.211, "step": 877 }, { "epoch": 0.18, "learning_rate": 2.7265365448504983e-07, "loss": 0.1643, "step": 878 }, { "epoch": 0.18, "learning_rate": 2.726225083056478e-07, "loss": 0.2074, "step": 879 }, { "epoch": 0.18, "learning_rate": 2.725913621262458e-07, "loss": 0.1509, "step": 880 }, { "epoch": 0.18, "learning_rate": 2.7256021594684385e-07, "loss": 0.1632, "step": 881 }, { "epoch": 0.18, "learning_rate": 2.7252906976744184e-07, "loss": 0.2023, "step": 882 }, { "epoch": 0.18, "learning_rate": 2.7249792358803983e-07, "loss": 0.1683, "step": 883 }, { "epoch": 0.18, "learning_rate": 2.724667774086379e-07, "loss": 0.1815, "step": 884 }, { "epoch": 0.18, "learning_rate": 2.7243563122923586e-07, "loss": 0.1627, "step": 885 }, { "epoch": 0.18, "learning_rate": 2.724044850498339e-07, "loss": 0.3082, "step": 886 }, { "epoch": 0.18, "learning_rate": 2.7237333887043184e-07, "loss": 0.1652, "step": 887 }, { "epoch": 0.18, "learning_rate": 2.723421926910299e-07, "loss": 0.1592, "step": 888 }, { "epoch": 0.18, "learning_rate": 2.723110465116279e-07, "loss": 0.177, "step": 889 }, { "epoch": 0.18, "learning_rate": 2.722799003322259e-07, "loss": 0.1301, "step": 890 }, { "epoch": 0.19, "learning_rate": 2.722487541528239e-07, "loss": 0.1622, "step": 891 }, { "epoch": 0.19, "learning_rate": 2.722176079734219e-07, "loss": 0.1665, "step": 892 }, { "epoch": 0.19, "learning_rate": 2.7218646179401994e-07, "loss": 0.1669, "step": 893 }, { "epoch": 0.19, "learning_rate": 2.7215531561461793e-07, "loss": 0.1772, "step": 894 }, { "epoch": 0.19, "learning_rate": 2.7212416943521597e-07, "loss": 0.1584, "step": 895 }, { "epoch": 0.19, "learning_rate": 2.720930232558139e-07, "loss": 0.1184, "step": 896 }, { "epoch": 0.19, "learning_rate": 2.7206187707641195e-07, "loss": 0.1724, "step": 897 }, { "epoch": 0.19, "learning_rate": 2.7203073089700994e-07, "loss": 0.1886, "step": 898 }, { "epoch": 0.19, "learning_rate": 2.71999584717608e-07, "loss": 0.1478, "step": 899 }, { "epoch": 0.19, "learning_rate": 2.7196843853820597e-07, "loss": 0.2165, "step": 900 }, { "epoch": 0.19, "learning_rate": 2.7193729235880396e-07, "loss": 0.1699, "step": 901 }, { "epoch": 0.19, "learning_rate": 2.7190614617940195e-07, "loss": 0.2002, "step": 902 }, { "epoch": 0.19, "learning_rate": 2.71875e-07, "loss": 0.1793, "step": 903 }, { "epoch": 0.19, "learning_rate": 2.71843853820598e-07, "loss": 0.1611, "step": 904 }, { "epoch": 0.19, "learning_rate": 2.71812707641196e-07, "loss": 0.1469, "step": 905 }, { "epoch": 0.19, "learning_rate": 2.71781561461794e-07, "loss": 0.1603, "step": 906 }, { "epoch": 0.19, "learning_rate": 2.71750415282392e-07, "loss": 0.143, "step": 907 }, { "epoch": 0.19, "learning_rate": 2.7171926910299004e-07, "loss": 0.1961, "step": 908 }, { "epoch": 0.19, "learning_rate": 2.7168812292358803e-07, "loss": 0.1201, "step": 909 }, { "epoch": 0.19, "learning_rate": 2.716569767441861e-07, "loss": 0.1431, "step": 910 }, { "epoch": 0.19, "learning_rate": 2.71625830564784e-07, "loss": 0.1688, "step": 911 }, { "epoch": 0.19, "learning_rate": 2.7159468438538205e-07, "loss": 0.138, "step": 912 }, { "epoch": 0.19, "learning_rate": 2.7156353820598004e-07, "loss": 0.1169, "step": 913 }, { "epoch": 0.19, "learning_rate": 2.715323920265781e-07, "loss": 0.1679, "step": 914 }, { "epoch": 0.19, "learning_rate": 2.715012458471761e-07, "loss": 0.1561, "step": 915 }, { "epoch": 0.19, "learning_rate": 2.7147009966777406e-07, "loss": 0.2445, "step": 916 }, { "epoch": 0.19, "learning_rate": 2.7143895348837205e-07, "loss": 0.1733, "step": 917 }, { "epoch": 0.19, "learning_rate": 2.714078073089701e-07, "loss": 0.1593, "step": 918 }, { "epoch": 0.19, "learning_rate": 2.713766611295681e-07, "loss": 0.2061, "step": 919 }, { "epoch": 0.19, "learning_rate": 2.713455149501661e-07, "loss": 0.1115, "step": 920 }, { "epoch": 0.19, "learning_rate": 2.713143687707641e-07, "loss": 0.16, "step": 921 }, { "epoch": 0.19, "learning_rate": 2.712832225913621e-07, "loss": 0.2387, "step": 922 }, { "epoch": 0.19, "learning_rate": 2.7125207641196015e-07, "loss": 0.1947, "step": 923 }, { "epoch": 0.19, "learning_rate": 2.7122093023255814e-07, "loss": 0.1488, "step": 924 }, { "epoch": 0.19, "learning_rate": 2.7118978405315613e-07, "loss": 0.2025, "step": 925 }, { "epoch": 0.19, "learning_rate": 2.711586378737541e-07, "loss": 0.1667, "step": 926 }, { "epoch": 0.19, "learning_rate": 2.7112749169435216e-07, "loss": 0.1397, "step": 927 }, { "epoch": 0.19, "learning_rate": 2.7109634551495015e-07, "loss": 0.1403, "step": 928 }, { "epoch": 0.19, "learning_rate": 2.7106519933554814e-07, "loss": 0.158, "step": 929 }, { "epoch": 0.19, "learning_rate": 2.710340531561462e-07, "loss": 0.1323, "step": 930 }, { "epoch": 0.19, "learning_rate": 2.7100290697674417e-07, "loss": 0.1727, "step": 931 }, { "epoch": 0.19, "learning_rate": 2.709717607973422e-07, "loss": 0.1542, "step": 932 }, { "epoch": 0.19, "learning_rate": 2.709406146179402e-07, "loss": 0.1584, "step": 933 }, { "epoch": 0.19, "learning_rate": 2.709094684385382e-07, "loss": 0.1909, "step": 934 }, { "epoch": 0.19, "learning_rate": 2.708783222591362e-07, "loss": 0.2746, "step": 935 }, { "epoch": 0.19, "learning_rate": 2.708471760797342e-07, "loss": 0.1916, "step": 936 }, { "epoch": 0.19, "learning_rate": 2.708160299003322e-07, "loss": 0.1409, "step": 937 }, { "epoch": 0.19, "learning_rate": 2.7078488372093025e-07, "loss": 0.1943, "step": 938 }, { "epoch": 0.19, "learning_rate": 2.707537375415282e-07, "loss": 0.128, "step": 939 }, { "epoch": 0.2, "learning_rate": 2.7072259136212623e-07, "loss": 0.1576, "step": 940 }, { "epoch": 0.2, "learning_rate": 2.706914451827242e-07, "loss": 0.1836, "step": 941 }, { "epoch": 0.2, "learning_rate": 2.7066029900332226e-07, "loss": 0.1505, "step": 942 }, { "epoch": 0.2, "learning_rate": 2.7062915282392025e-07, "loss": 0.1279, "step": 943 }, { "epoch": 0.2, "learning_rate": 2.7059800664451824e-07, "loss": 0.1763, "step": 944 }, { "epoch": 0.2, "learning_rate": 2.705668604651163e-07, "loss": 0.2032, "step": 945 }, { "epoch": 0.2, "learning_rate": 2.705357142857143e-07, "loss": 0.1604, "step": 946 }, { "epoch": 0.2, "learning_rate": 2.705045681063123e-07, "loss": 0.1569, "step": 947 }, { "epoch": 0.2, "learning_rate": 2.7047342192691025e-07, "loss": 0.2062, "step": 948 }, { "epoch": 0.2, "learning_rate": 2.704422757475083e-07, "loss": 0.2068, "step": 949 }, { "epoch": 0.2, "learning_rate": 2.704111295681063e-07, "loss": 0.1989, "step": 950 }, { "epoch": 0.2, "learning_rate": 2.7037998338870433e-07, "loss": 0.1804, "step": 951 }, { "epoch": 0.2, "learning_rate": 2.703488372093023e-07, "loss": 0.2196, "step": 952 }, { "epoch": 0.2, "learning_rate": 2.703176910299003e-07, "loss": 0.1611, "step": 953 }, { "epoch": 0.2, "learning_rate": 2.702865448504983e-07, "loss": 0.1927, "step": 954 }, { "epoch": 0.2, "learning_rate": 2.7025539867109634e-07, "loss": 0.1446, "step": 955 }, { "epoch": 0.2, "learning_rate": 2.7022425249169433e-07, "loss": 0.1984, "step": 956 }, { "epoch": 0.2, "learning_rate": 2.7019310631229237e-07, "loss": 0.1645, "step": 957 }, { "epoch": 0.2, "learning_rate": 2.7016196013289036e-07, "loss": 0.2071, "step": 958 }, { "epoch": 0.2, "learning_rate": 2.7013081395348835e-07, "loss": 0.2512, "step": 959 }, { "epoch": 0.2, "learning_rate": 2.700996677740864e-07, "loss": 0.149, "step": 960 }, { "epoch": 0.2, "learning_rate": 2.700685215946844e-07, "loss": 0.1801, "step": 961 }, { "epoch": 0.2, "learning_rate": 2.7003737541528237e-07, "loss": 0.1485, "step": 962 }, { "epoch": 0.2, "learning_rate": 2.7000622923588036e-07, "loss": 0.2076, "step": 963 }, { "epoch": 0.2, "learning_rate": 2.699750830564784e-07, "loss": 0.1495, "step": 964 }, { "epoch": 0.2, "learning_rate": 2.699439368770764e-07, "loss": 0.1407, "step": 965 }, { "epoch": 0.2, "learning_rate": 2.6991279069767443e-07, "loss": 0.2778, "step": 966 }, { "epoch": 0.2, "learning_rate": 2.698816445182724e-07, "loss": 0.1229, "step": 967 }, { "epoch": 0.2, "learning_rate": 2.698504983388704e-07, "loss": 0.1517, "step": 968 }, { "epoch": 0.2, "learning_rate": 2.6981935215946845e-07, "loss": 0.2016, "step": 969 }, { "epoch": 0.2, "learning_rate": 2.6978820598006644e-07, "loss": 0.195, "step": 970 }, { "epoch": 0.2, "learning_rate": 2.6975705980066443e-07, "loss": 0.1285, "step": 971 }, { "epoch": 0.2, "learning_rate": 2.697259136212624e-07, "loss": 0.152, "step": 972 }, { "epoch": 0.2, "learning_rate": 2.6969476744186046e-07, "loss": 0.1326, "step": 973 }, { "epoch": 0.2, "learning_rate": 2.6966362126245845e-07, "loss": 0.1647, "step": 974 }, { "epoch": 0.2, "learning_rate": 2.696324750830565e-07, "loss": 0.2159, "step": 975 }, { "epoch": 0.2, "learning_rate": 2.6960132890365443e-07, "loss": 0.2107, "step": 976 }, { "epoch": 0.2, "learning_rate": 2.695701827242525e-07, "loss": 0.1877, "step": 977 }, { "epoch": 0.2, "learning_rate": 2.6953903654485046e-07, "loss": 0.0984, "step": 978 }, { "epoch": 0.2, "learning_rate": 2.695078903654485e-07, "loss": 0.1481, "step": 979 }, { "epoch": 0.2, "learning_rate": 2.694767441860465e-07, "loss": 0.1746, "step": 980 }, { "epoch": 0.2, "learning_rate": 2.694455980066445e-07, "loss": 0.1658, "step": 981 }, { "epoch": 0.2, "learning_rate": 2.694144518272425e-07, "loss": 0.1354, "step": 982 }, { "epoch": 0.2, "learning_rate": 2.693833056478405e-07, "loss": 0.1849, "step": 983 }, { "epoch": 0.2, "learning_rate": 2.6935215946843856e-07, "loss": 0.1443, "step": 984 }, { "epoch": 0.2, "learning_rate": 2.6932101328903655e-07, "loss": 0.1986, "step": 985 }, { "epoch": 0.2, "learning_rate": 2.6928986710963454e-07, "loss": 0.1903, "step": 986 }, { "epoch": 0.2, "learning_rate": 2.692587209302325e-07, "loss": 0.187, "step": 987 }, { "epoch": 0.21, "learning_rate": 2.6922757475083057e-07, "loss": 0.1416, "step": 988 }, { "epoch": 0.21, "learning_rate": 2.6919642857142856e-07, "loss": 0.1467, "step": 989 }, { "epoch": 0.21, "learning_rate": 2.691652823920266e-07, "loss": 0.2313, "step": 990 }, { "epoch": 0.21, "learning_rate": 2.6913413621262454e-07, "loss": 0.1748, "step": 991 }, { "epoch": 0.21, "learning_rate": 2.691029900332226e-07, "loss": 0.2094, "step": 992 }, { "epoch": 0.21, "learning_rate": 2.6907184385382057e-07, "loss": 0.1256, "step": 993 }, { "epoch": 0.21, "learning_rate": 2.690406976744186e-07, "loss": 0.248, "step": 994 }, { "epoch": 0.21, "learning_rate": 2.690095514950166e-07, "loss": 0.1909, "step": 995 }, { "epoch": 0.21, "learning_rate": 2.689784053156146e-07, "loss": 0.2302, "step": 996 }, { "epoch": 0.21, "learning_rate": 2.6894725913621263e-07, "loss": 0.1416, "step": 997 }, { "epoch": 0.21, "learning_rate": 2.689161129568106e-07, "loss": 0.1757, "step": 998 }, { "epoch": 0.21, "learning_rate": 2.6888496677740866e-07, "loss": 0.1157, "step": 999 }, { "epoch": 0.21, "learning_rate": 2.688538205980066e-07, "loss": 0.1727, "step": 1000 }, { "epoch": 0.21, "learning_rate": 2.6882267441860464e-07, "loss": 0.1914, "step": 1001 }, { "epoch": 0.21, "learning_rate": 2.6879152823920263e-07, "loss": 0.2152, "step": 1002 }, { "epoch": 0.21, "learning_rate": 2.6876038205980067e-07, "loss": 0.1909, "step": 1003 }, { "epoch": 0.21, "learning_rate": 2.6872923588039866e-07, "loss": 0.2407, "step": 1004 }, { "epoch": 0.21, "learning_rate": 2.6869808970099665e-07, "loss": 0.1612, "step": 1005 }, { "epoch": 0.21, "learning_rate": 2.686669435215947e-07, "loss": 0.1045, "step": 1006 }, { "epoch": 0.21, "learning_rate": 2.686357973421927e-07, "loss": 0.2254, "step": 1007 }, { "epoch": 0.21, "learning_rate": 2.6860465116279067e-07, "loss": 0.2031, "step": 1008 }, { "epoch": 0.21, "learning_rate": 2.6857350498338866e-07, "loss": 0.1717, "step": 1009 }, { "epoch": 0.21, "learning_rate": 2.685423588039867e-07, "loss": 0.1883, "step": 1010 }, { "epoch": 0.21, "learning_rate": 2.685112126245847e-07, "loss": 0.1818, "step": 1011 }, { "epoch": 0.21, "learning_rate": 2.6848006644518274e-07, "loss": 0.193, "step": 1012 }, { "epoch": 0.21, "learning_rate": 2.684489202657807e-07, "loss": 0.1401, "step": 1013 }, { "epoch": 0.21, "learning_rate": 2.684177740863787e-07, "loss": 0.1594, "step": 1014 }, { "epoch": 0.21, "learning_rate": 2.683866279069767e-07, "loss": 0.1744, "step": 1015 }, { "epoch": 0.21, "learning_rate": 2.6835548172757475e-07, "loss": 0.2386, "step": 1016 }, { "epoch": 0.21, "learning_rate": 2.6832433554817274e-07, "loss": 0.1987, "step": 1017 }, { "epoch": 0.21, "learning_rate": 2.682931893687708e-07, "loss": 0.1433, "step": 1018 }, { "epoch": 0.21, "learning_rate": 2.6826204318936877e-07, "loss": 0.1463, "step": 1019 }, { "epoch": 0.21, "learning_rate": 2.6823089700996676e-07, "loss": 0.1495, "step": 1020 }, { "epoch": 0.21, "learning_rate": 2.681997508305648e-07, "loss": 0.1478, "step": 1021 }, { "epoch": 0.21, "learning_rate": 2.681686046511628e-07, "loss": 0.1635, "step": 1022 }, { "epoch": 0.21, "learning_rate": 2.681374584717608e-07, "loss": 0.1196, "step": 1023 }, { "epoch": 0.21, "learning_rate": 2.6810631229235877e-07, "loss": 0.1596, "step": 1024 }, { "epoch": 0.21, "learning_rate": 2.680751661129568e-07, "loss": 0.1799, "step": 1025 }, { "epoch": 0.21, "learning_rate": 2.680440199335548e-07, "loss": 0.1398, "step": 1026 }, { "epoch": 0.21, "learning_rate": 2.6801287375415284e-07, "loss": 0.1486, "step": 1027 }, { "epoch": 0.21, "learning_rate": 2.679817275747508e-07, "loss": 0.1592, "step": 1028 }, { "epoch": 0.21, "learning_rate": 2.679505813953488e-07, "loss": 0.2052, "step": 1029 }, { "epoch": 0.21, "learning_rate": 2.679194352159468e-07, "loss": 0.1138, "step": 1030 }, { "epoch": 0.21, "learning_rate": 2.6788828903654485e-07, "loss": 0.2198, "step": 1031 }, { "epoch": 0.21, "learning_rate": 2.6785714285714284e-07, "loss": 0.1648, "step": 1032 }, { "epoch": 0.21, "learning_rate": 2.6782599667774083e-07, "loss": 0.1561, "step": 1033 }, { "epoch": 0.21, "learning_rate": 2.6779485049833887e-07, "loss": 0.1649, "step": 1034 }, { "epoch": 0.21, "learning_rate": 2.6776370431893686e-07, "loss": 0.1371, "step": 1035 }, { "epoch": 0.22, "learning_rate": 2.677325581395349e-07, "loss": 0.1386, "step": 1036 }, { "epoch": 0.22, "learning_rate": 2.677014119601329e-07, "loss": 0.1581, "step": 1037 }, { "epoch": 0.22, "learning_rate": 2.676702657807309e-07, "loss": 0.1833, "step": 1038 }, { "epoch": 0.22, "learning_rate": 2.6763911960132887e-07, "loss": 0.1391, "step": 1039 }, { "epoch": 0.22, "learning_rate": 2.676079734219269e-07, "loss": 0.1197, "step": 1040 }, { "epoch": 0.22, "learning_rate": 2.675768272425249e-07, "loss": 0.1354, "step": 1041 }, { "epoch": 0.22, "learning_rate": 2.675456810631229e-07, "loss": 0.1903, "step": 1042 }, { "epoch": 0.22, "learning_rate": 2.6751453488372094e-07, "loss": 0.1508, "step": 1043 }, { "epoch": 0.22, "learning_rate": 2.674833887043189e-07, "loss": 0.1542, "step": 1044 }, { "epoch": 0.22, "learning_rate": 2.674522425249169e-07, "loss": 0.1399, "step": 1045 }, { "epoch": 0.22, "learning_rate": 2.6742109634551496e-07, "loss": 0.1936, "step": 1046 }, { "epoch": 0.22, "learning_rate": 2.6738995016611295e-07, "loss": 0.1147, "step": 1047 }, { "epoch": 0.22, "learning_rate": 2.6735880398671094e-07, "loss": 0.1653, "step": 1048 }, { "epoch": 0.22, "learning_rate": 2.67327657807309e-07, "loss": 0.1419, "step": 1049 }, { "epoch": 0.22, "learning_rate": 2.6729651162790697e-07, "loss": 0.1979, "step": 1050 }, { "epoch": 0.22, "learning_rate": 2.67265365448505e-07, "loss": 0.1722, "step": 1051 }, { "epoch": 0.22, "learning_rate": 2.6723421926910295e-07, "loss": 0.1638, "step": 1052 }, { "epoch": 0.22, "learning_rate": 2.67203073089701e-07, "loss": 0.1842, "step": 1053 }, { "epoch": 0.22, "learning_rate": 2.67171926910299e-07, "loss": 0.1386, "step": 1054 }, { "epoch": 0.22, "learning_rate": 2.67140780730897e-07, "loss": 0.124, "step": 1055 }, { "epoch": 0.22, "learning_rate": 2.67109634551495e-07, "loss": 0.1386, "step": 1056 }, { "epoch": 0.22, "learning_rate": 2.67078488372093e-07, "loss": 0.1587, "step": 1057 }, { "epoch": 0.22, "learning_rate": 2.6704734219269104e-07, "loss": 0.1496, "step": 1058 }, { "epoch": 0.22, "learning_rate": 2.6701619601328903e-07, "loss": 0.1865, "step": 1059 }, { "epoch": 0.22, "learning_rate": 2.66985049833887e-07, "loss": 0.1405, "step": 1060 }, { "epoch": 0.22, "learning_rate": 2.66953903654485e-07, "loss": 0.1419, "step": 1061 }, { "epoch": 0.22, "learning_rate": 2.6692275747508305e-07, "loss": 0.2019, "step": 1062 }, { "epoch": 0.22, "learning_rate": 2.6689161129568104e-07, "loss": 0.1724, "step": 1063 }, { "epoch": 0.22, "learning_rate": 2.668604651162791e-07, "loss": 0.1866, "step": 1064 }, { "epoch": 0.22, "learning_rate": 2.6682931893687707e-07, "loss": 0.1409, "step": 1065 }, { "epoch": 0.22, "learning_rate": 2.6679817275747506e-07, "loss": 0.1764, "step": 1066 }, { "epoch": 0.22, "learning_rate": 2.6676702657807305e-07, "loss": 0.1793, "step": 1067 }, { "epoch": 0.22, "learning_rate": 2.667358803986711e-07, "loss": 0.1559, "step": 1068 }, { "epoch": 0.22, "learning_rate": 2.667047342192691e-07, "loss": 0.1767, "step": 1069 }, { "epoch": 0.22, "learning_rate": 2.6667358803986707e-07, "loss": 0.1469, "step": 1070 }, { "epoch": 0.22, "learning_rate": 2.666424418604651e-07, "loss": 0.148, "step": 1071 }, { "epoch": 0.22, "learning_rate": 2.666112956810631e-07, "loss": 0.1886, "step": 1072 }, { "epoch": 0.22, "learning_rate": 2.6658014950166115e-07, "loss": 0.1533, "step": 1073 }, { "epoch": 0.22, "learning_rate": 2.6654900332225914e-07, "loss": 0.1822, "step": 1074 }, { "epoch": 0.22, "learning_rate": 2.665178571428571e-07, "loss": 0.1885, "step": 1075 }, { "epoch": 0.22, "learning_rate": 2.664867109634551e-07, "loss": 0.1436, "step": 1076 }, { "epoch": 0.22, "learning_rate": 2.6645556478405316e-07, "loss": 0.1857, "step": 1077 }, { "epoch": 0.22, "learning_rate": 2.6642441860465115e-07, "loss": 0.164, "step": 1078 }, { "epoch": 0.22, "learning_rate": 2.663932724252492e-07, "loss": 0.1652, "step": 1079 }, { "epoch": 0.22, "learning_rate": 2.663621262458472e-07, "loss": 0.1841, "step": 1080 }, { "epoch": 0.22, "learning_rate": 2.6633098006644517e-07, "loss": 0.18, "step": 1081 }, { "epoch": 0.22, "learning_rate": 2.6629983388704316e-07, "loss": 0.159, "step": 1082 }, { "epoch": 0.22, "learning_rate": 2.662686877076412e-07, "loss": 0.1228, "step": 1083 }, { "epoch": 0.23, "learning_rate": 2.662375415282392e-07, "loss": 0.1827, "step": 1084 }, { "epoch": 0.23, "learning_rate": 2.662063953488372e-07, "loss": 0.1812, "step": 1085 }, { "epoch": 0.23, "learning_rate": 2.661752491694352e-07, "loss": 0.1698, "step": 1086 }, { "epoch": 0.23, "learning_rate": 2.661441029900332e-07, "loss": 0.1151, "step": 1087 }, { "epoch": 0.23, "learning_rate": 2.6611295681063125e-07, "loss": 0.0973, "step": 1088 }, { "epoch": 0.23, "learning_rate": 2.660818106312292e-07, "loss": 0.1632, "step": 1089 }, { "epoch": 0.23, "learning_rate": 2.6605066445182723e-07, "loss": 0.1627, "step": 1090 }, { "epoch": 0.23, "learning_rate": 2.660195182724252e-07, "loss": 0.1519, "step": 1091 }, { "epoch": 0.23, "learning_rate": 2.6598837209302326e-07, "loss": 0.194, "step": 1092 }, { "epoch": 0.23, "learning_rate": 2.6595722591362125e-07, "loss": 0.1918, "step": 1093 }, { "epoch": 0.23, "learning_rate": 2.6592607973421924e-07, "loss": 0.1966, "step": 1094 }, { "epoch": 0.23, "learning_rate": 2.658949335548173e-07, "loss": 0.1015, "step": 1095 }, { "epoch": 0.23, "learning_rate": 2.6586378737541527e-07, "loss": 0.1477, "step": 1096 }, { "epoch": 0.23, "learning_rate": 2.6583264119601326e-07, "loss": 0.1576, "step": 1097 }, { "epoch": 0.23, "learning_rate": 2.658014950166113e-07, "loss": 0.2683, "step": 1098 }, { "epoch": 0.23, "learning_rate": 2.657703488372093e-07, "loss": 0.1589, "step": 1099 }, { "epoch": 0.23, "learning_rate": 2.657392026578073e-07, "loss": 0.1536, "step": 1100 }, { "epoch": 0.23, "learning_rate": 2.657080564784053e-07, "loss": 0.185, "step": 1101 }, { "epoch": 0.23, "learning_rate": 2.656769102990033e-07, "loss": 0.13, "step": 1102 }, { "epoch": 0.23, "learning_rate": 2.656457641196013e-07, "loss": 0.1503, "step": 1103 }, { "epoch": 0.23, "learning_rate": 2.656146179401993e-07, "loss": 0.1768, "step": 1104 }, { "epoch": 0.23, "learning_rate": 2.6558347176079733e-07, "loss": 0.1524, "step": 1105 }, { "epoch": 0.23, "learning_rate": 2.655523255813953e-07, "loss": 0.1364, "step": 1106 }, { "epoch": 0.23, "learning_rate": 2.6552117940199337e-07, "loss": 0.1378, "step": 1107 }, { "epoch": 0.23, "learning_rate": 2.6549003322259136e-07, "loss": 0.1934, "step": 1108 }, { "epoch": 0.23, "learning_rate": 2.6545888704318935e-07, "loss": 0.206, "step": 1109 }, { "epoch": 0.23, "learning_rate": 2.654277408637874e-07, "loss": 0.1797, "step": 1110 }, { "epoch": 0.23, "learning_rate": 2.653965946843854e-07, "loss": 0.1329, "step": 1111 }, { "epoch": 0.23, "learning_rate": 2.6536544850498337e-07, "loss": 0.1525, "step": 1112 }, { "epoch": 0.23, "learning_rate": 2.6533430232558136e-07, "loss": 0.1774, "step": 1113 }, { "epoch": 0.23, "learning_rate": 2.653031561461794e-07, "loss": 0.1576, "step": 1114 }, { "epoch": 0.23, "learning_rate": 2.652720099667774e-07, "loss": 0.1337, "step": 1115 }, { "epoch": 0.23, "learning_rate": 2.6524086378737543e-07, "loss": 0.2629, "step": 1116 }, { "epoch": 0.23, "learning_rate": 2.652097176079734e-07, "loss": 0.1865, "step": 1117 }, { "epoch": 0.23, "learning_rate": 2.651785714285714e-07, "loss": 0.1647, "step": 1118 }, { "epoch": 0.23, "learning_rate": 2.651474252491694e-07, "loss": 0.1359, "step": 1119 }, { "epoch": 0.23, "learning_rate": 2.6511627906976744e-07, "loss": 0.1243, "step": 1120 }, { "epoch": 0.23, "learning_rate": 2.6508513289036543e-07, "loss": 0.1706, "step": 1121 }, { "epoch": 0.23, "learning_rate": 2.650539867109634e-07, "loss": 0.22, "step": 1122 }, { "epoch": 0.23, "learning_rate": 2.6502284053156146e-07, "loss": 0.2006, "step": 1123 }, { "epoch": 0.23, "learning_rate": 2.6499169435215945e-07, "loss": 0.1405, "step": 1124 }, { "epoch": 0.23, "learning_rate": 2.649605481727575e-07, "loss": 0.2096, "step": 1125 }, { "epoch": 0.23, "learning_rate": 2.649294019933555e-07, "loss": 0.1685, "step": 1126 }, { "epoch": 0.23, "learning_rate": 2.6489825581395347e-07, "loss": 0.165, "step": 1127 }, { "epoch": 0.23, "learning_rate": 2.6486710963455146e-07, "loss": 0.1165, "step": 1128 }, { "epoch": 0.23, "learning_rate": 2.648359634551495e-07, "loss": 0.1718, "step": 1129 }, { "epoch": 0.23, "learning_rate": 2.648048172757475e-07, "loss": 0.1234, "step": 1130 }, { "epoch": 0.23, "learning_rate": 2.6477367109634553e-07, "loss": 0.1947, "step": 1131 }, { "epoch": 0.24, "learning_rate": 2.647425249169435e-07, "loss": 0.1619, "step": 1132 }, { "epoch": 0.24, "learning_rate": 2.647113787375415e-07, "loss": 0.2106, "step": 1133 }, { "epoch": 0.24, "learning_rate": 2.646802325581395e-07, "loss": 0.178, "step": 1134 }, { "epoch": 0.24, "learning_rate": 2.6464908637873754e-07, "loss": 0.2264, "step": 1135 }, { "epoch": 0.24, "learning_rate": 2.6461794019933553e-07, "loss": 0.1085, "step": 1136 }, { "epoch": 0.24, "learning_rate": 2.645867940199335e-07, "loss": 0.1546, "step": 1137 }, { "epoch": 0.24, "learning_rate": 2.6455564784053157e-07, "loss": 0.133, "step": 1138 }, { "epoch": 0.24, "learning_rate": 2.6452450166112956e-07, "loss": 0.1561, "step": 1139 }, { "epoch": 0.24, "learning_rate": 2.644933554817276e-07, "loss": 0.1608, "step": 1140 }, { "epoch": 0.24, "learning_rate": 2.6446220930232553e-07, "loss": 0.2114, "step": 1141 }, { "epoch": 0.24, "learning_rate": 2.644310631229236e-07, "loss": 0.1521, "step": 1142 }, { "epoch": 0.24, "learning_rate": 2.6439991694352157e-07, "loss": 0.1166, "step": 1143 }, { "epoch": 0.24, "learning_rate": 2.643687707641196e-07, "loss": 0.1445, "step": 1144 }, { "epoch": 0.24, "learning_rate": 2.643376245847176e-07, "loss": 0.1585, "step": 1145 }, { "epoch": 0.24, "learning_rate": 2.643064784053156e-07, "loss": 0.1302, "step": 1146 }, { "epoch": 0.24, "learning_rate": 2.6427533222591363e-07, "loss": 0.1808, "step": 1147 }, { "epoch": 0.24, "learning_rate": 2.642441860465116e-07, "loss": 0.2229, "step": 1148 }, { "epoch": 0.24, "learning_rate": 2.642130398671096e-07, "loss": 0.1862, "step": 1149 }, { "epoch": 0.24, "learning_rate": 2.641818936877076e-07, "loss": 0.1589, "step": 1150 }, { "epoch": 0.24, "learning_rate": 2.6415074750830564e-07, "loss": 0.1788, "step": 1151 }, { "epoch": 0.24, "learning_rate": 2.6411960132890363e-07, "loss": 0.1729, "step": 1152 }, { "epoch": 0.24, "learning_rate": 2.6408845514950167e-07, "loss": 0.1181, "step": 1153 }, { "epoch": 0.24, "learning_rate": 2.6405730897009966e-07, "loss": 0.1441, "step": 1154 }, { "epoch": 0.24, "learning_rate": 2.6402616279069765e-07, "loss": 0.2054, "step": 1155 }, { "epoch": 0.24, "learning_rate": 2.6399501661129564e-07, "loss": 0.2357, "step": 1156 }, { "epoch": 0.24, "learning_rate": 2.639638704318937e-07, "loss": 0.168, "step": 1157 }, { "epoch": 0.24, "learning_rate": 2.6393272425249167e-07, "loss": 0.1656, "step": 1158 }, { "epoch": 0.24, "learning_rate": 2.639015780730897e-07, "loss": 0.1774, "step": 1159 }, { "epoch": 0.24, "learning_rate": 2.638704318936877e-07, "loss": 0.2083, "step": 1160 }, { "epoch": 0.24, "learning_rate": 2.638392857142857e-07, "loss": 0.2016, "step": 1161 }, { "epoch": 0.24, "learning_rate": 2.6380813953488373e-07, "loss": 0.158, "step": 1162 }, { "epoch": 0.24, "learning_rate": 2.637769933554817e-07, "loss": 0.1785, "step": 1163 }, { "epoch": 0.24, "learning_rate": 2.637458471760797e-07, "loss": 0.1755, "step": 1164 }, { "epoch": 0.24, "learning_rate": 2.637147009966777e-07, "loss": 0.145, "step": 1165 }, { "epoch": 0.24, "learning_rate": 2.6368355481727574e-07, "loss": 0.1079, "step": 1166 }, { "epoch": 0.24, "learning_rate": 2.6365240863787373e-07, "loss": 0.161, "step": 1167 }, { "epoch": 0.24, "learning_rate": 2.636212624584718e-07, "loss": 0.1834, "step": 1168 }, { "epoch": 0.24, "learning_rate": 2.6359011627906977e-07, "loss": 0.1498, "step": 1169 }, { "epoch": 0.24, "learning_rate": 2.6355897009966775e-07, "loss": 0.2144, "step": 1170 }, { "epoch": 0.24, "learning_rate": 2.6352782392026574e-07, "loss": 0.1096, "step": 1171 }, { "epoch": 0.24, "learning_rate": 2.634966777408638e-07, "loss": 0.1436, "step": 1172 }, { "epoch": 0.24, "learning_rate": 2.634655315614618e-07, "loss": 0.1637, "step": 1173 }, { "epoch": 0.24, "learning_rate": 2.6343438538205976e-07, "loss": 0.1691, "step": 1174 }, { "epoch": 0.24, "learning_rate": 2.634032392026578e-07, "loss": 0.1392, "step": 1175 }, { "epoch": 0.24, "learning_rate": 2.633720930232558e-07, "loss": 0.0959, "step": 1176 }, { "epoch": 0.24, "learning_rate": 2.6334094684385384e-07, "loss": 0.1303, "step": 1177 }, { "epoch": 0.24, "learning_rate": 2.6330980066445183e-07, "loss": 0.1318, "step": 1178 }, { "epoch": 0.24, "learning_rate": 2.632786544850498e-07, "loss": 0.1296, "step": 1179 }, { "epoch": 0.25, "learning_rate": 2.632475083056478e-07, "loss": 0.1377, "step": 1180 }, { "epoch": 0.25, "learning_rate": 2.6321636212624585e-07, "loss": 0.1821, "step": 1181 }, { "epoch": 0.25, "learning_rate": 2.6318521594684384e-07, "loss": 0.1713, "step": 1182 }, { "epoch": 0.25, "learning_rate": 2.6315406976744183e-07, "loss": 0.2545, "step": 1183 }, { "epoch": 0.25, "learning_rate": 2.6312292358803987e-07, "loss": 0.1422, "step": 1184 }, { "epoch": 0.25, "learning_rate": 2.6309177740863786e-07, "loss": 0.1256, "step": 1185 }, { "epoch": 0.25, "learning_rate": 2.6306063122923585e-07, "loss": 0.1303, "step": 1186 }, { "epoch": 0.25, "learning_rate": 2.630294850498339e-07, "loss": 0.1485, "step": 1187 }, { "epoch": 0.25, "learning_rate": 2.629983388704319e-07, "loss": 0.1198, "step": 1188 }, { "epoch": 0.25, "learning_rate": 2.6296719269102987e-07, "loss": 0.171, "step": 1189 }, { "epoch": 0.25, "learning_rate": 2.629360465116279e-07, "loss": 0.146, "step": 1190 }, { "epoch": 0.25, "learning_rate": 2.629049003322259e-07, "loss": 0.1781, "step": 1191 }, { "epoch": 0.25, "learning_rate": 2.6287375415282394e-07, "loss": 0.1829, "step": 1192 }, { "epoch": 0.25, "learning_rate": 2.628426079734219e-07, "loss": 0.2039, "step": 1193 }, { "epoch": 0.25, "learning_rate": 2.628114617940199e-07, "loss": 0.1732, "step": 1194 }, { "epoch": 0.25, "learning_rate": 2.627803156146179e-07, "loss": 0.1577, "step": 1195 }, { "epoch": 0.25, "learning_rate": 2.6274916943521595e-07, "loss": 0.1921, "step": 1196 }, { "epoch": 0.25, "learning_rate": 2.6271802325581394e-07, "loss": 0.1505, "step": 1197 }, { "epoch": 0.25, "learning_rate": 2.6268687707641193e-07, "loss": 0.146, "step": 1198 }, { "epoch": 0.25, "learning_rate": 2.6265573089701e-07, "loss": 0.1377, "step": 1199 }, { "epoch": 0.25, "learning_rate": 2.6262458471760796e-07, "loss": 0.1508, "step": 1200 }, { "epoch": 0.25, "learning_rate": 2.6259343853820595e-07, "loss": 0.1906, "step": 1201 }, { "epoch": 0.25, "learning_rate": 2.6256229235880394e-07, "loss": 0.1533, "step": 1202 }, { "epoch": 0.25, "learning_rate": 2.62531146179402e-07, "loss": 0.2297, "step": 1203 }, { "epoch": 0.25, "learning_rate": 2.625e-07, "loss": 0.1679, "step": 1204 }, { "epoch": 0.25, "learning_rate": 2.62468853820598e-07, "loss": 0.1617, "step": 1205 }, { "epoch": 0.25, "learning_rate": 2.62437707641196e-07, "loss": 0.1714, "step": 1206 }, { "epoch": 0.25, "learning_rate": 2.62406561461794e-07, "loss": 0.1649, "step": 1207 }, { "epoch": 0.25, "learning_rate": 2.62375415282392e-07, "loss": 0.1224, "step": 1208 }, { "epoch": 0.25, "learning_rate": 2.6234426910299003e-07, "loss": 0.1515, "step": 1209 }, { "epoch": 0.25, "learning_rate": 2.62313122923588e-07, "loss": 0.2165, "step": 1210 }, { "epoch": 0.25, "learning_rate": 2.6228197674418606e-07, "loss": 0.0909, "step": 1211 }, { "epoch": 0.25, "learning_rate": 2.6225083056478405e-07, "loss": 0.165, "step": 1212 }, { "epoch": 0.25, "learning_rate": 2.6221968438538204e-07, "loss": 0.1921, "step": 1213 }, { "epoch": 0.25, "learning_rate": 2.621885382059801e-07, "loss": 0.1567, "step": 1214 }, { "epoch": 0.25, "learning_rate": 2.6215739202657807e-07, "loss": 0.1251, "step": 1215 }, { "epoch": 0.25, "learning_rate": 2.6212624584717606e-07, "loss": 0.1647, "step": 1216 }, { "epoch": 0.25, "learning_rate": 2.6209509966777405e-07, "loss": 0.1669, "step": 1217 }, { "epoch": 0.25, "learning_rate": 2.620639534883721e-07, "loss": 0.1782, "step": 1218 }, { "epoch": 0.25, "learning_rate": 2.620328073089701e-07, "loss": 0.2501, "step": 1219 }, { "epoch": 0.25, "learning_rate": 2.620016611295681e-07, "loss": 0.1332, "step": 1220 }, { "epoch": 0.25, "learning_rate": 2.619705149501661e-07, "loss": 0.127, "step": 1221 }, { "epoch": 0.25, "learning_rate": 2.619393687707641e-07, "loss": 0.1476, "step": 1222 }, { "epoch": 0.25, "learning_rate": 2.619082225913621e-07, "loss": 0.162, "step": 1223 }, { "epoch": 0.25, "learning_rate": 2.6187707641196013e-07, "loss": 0.1652, "step": 1224 }, { "epoch": 0.25, "learning_rate": 2.618459302325581e-07, "loss": 0.1784, "step": 1225 }, { "epoch": 0.25, "learning_rate": 2.618147840531561e-07, "loss": 0.1301, "step": 1226 }, { "epoch": 0.25, "learning_rate": 2.6178363787375415e-07, "loss": 0.1351, "step": 1227 }, { "epoch": 0.25, "learning_rate": 2.6175249169435214e-07, "loss": 0.1633, "step": 1228 }, { "epoch": 0.26, "learning_rate": 2.617213455149502e-07, "loss": 0.2021, "step": 1229 }, { "epoch": 0.26, "learning_rate": 2.616901993355481e-07, "loss": 0.1679, "step": 1230 }, { "epoch": 0.26, "learning_rate": 2.6165905315614616e-07, "loss": 0.1575, "step": 1231 }, { "epoch": 0.26, "learning_rate": 2.6162790697674415e-07, "loss": 0.1388, "step": 1232 }, { "epoch": 0.26, "learning_rate": 2.615967607973422e-07, "loss": 0.1428, "step": 1233 }, { "epoch": 0.26, "learning_rate": 2.615656146179402e-07, "loss": 0.1296, "step": 1234 }, { "epoch": 0.26, "learning_rate": 2.615344684385382e-07, "loss": 0.189, "step": 1235 }, { "epoch": 0.26, "learning_rate": 2.615033222591362e-07, "loss": 0.1237, "step": 1236 }, { "epoch": 0.26, "learning_rate": 2.614721760797342e-07, "loss": 0.1214, "step": 1237 }, { "epoch": 0.26, "learning_rate": 2.614410299003322e-07, "loss": 0.1303, "step": 1238 }, { "epoch": 0.26, "learning_rate": 2.6140988372093024e-07, "loss": 0.142, "step": 1239 }, { "epoch": 0.26, "learning_rate": 2.6137873754152823e-07, "loss": 0.1469, "step": 1240 }, { "epoch": 0.26, "learning_rate": 2.613475913621262e-07, "loss": 0.1412, "step": 1241 }, { "epoch": 0.26, "learning_rate": 2.6131644518272426e-07, "loss": 0.1191, "step": 1242 }, { "epoch": 0.26, "learning_rate": 2.6128529900332225e-07, "loss": 0.1516, "step": 1243 }, { "epoch": 0.26, "learning_rate": 2.6125415282392024e-07, "loss": 0.1978, "step": 1244 }, { "epoch": 0.26, "learning_rate": 2.6122300664451823e-07, "loss": 0.1624, "step": 1245 }, { "epoch": 0.26, "learning_rate": 2.6119186046511627e-07, "loss": 0.1329, "step": 1246 }, { "epoch": 0.26, "learning_rate": 2.6116071428571426e-07, "loss": 0.1757, "step": 1247 }, { "epoch": 0.26, "learning_rate": 2.611295681063123e-07, "loss": 0.2074, "step": 1248 }, { "epoch": 0.26, "learning_rate": 2.610984219269103e-07, "loss": 0.1502, "step": 1249 }, { "epoch": 0.26, "learning_rate": 2.610672757475083e-07, "loss": 0.2018, "step": 1250 }, { "epoch": 0.26, "learning_rate": 2.610361295681063e-07, "loss": 0.1961, "step": 1251 }, { "epoch": 0.26, "learning_rate": 2.610049833887043e-07, "loss": 0.2013, "step": 1252 }, { "epoch": 0.26, "learning_rate": 2.6097383720930235e-07, "loss": 0.1947, "step": 1253 }, { "epoch": 0.26, "learning_rate": 2.609426910299003e-07, "loss": 0.1771, "step": 1254 }, { "epoch": 0.26, "learning_rate": 2.6091154485049833e-07, "loss": 0.1134, "step": 1255 }, { "epoch": 0.26, "learning_rate": 2.608803986710963e-07, "loss": 0.1877, "step": 1256 }, { "epoch": 0.26, "learning_rate": 2.6084925249169436e-07, "loss": 0.1356, "step": 1257 }, { "epoch": 0.26, "learning_rate": 2.6081810631229235e-07, "loss": 0.1271, "step": 1258 }, { "epoch": 0.26, "learning_rate": 2.6078696013289034e-07, "loss": 0.1769, "step": 1259 }, { "epoch": 0.26, "learning_rate": 2.6075581395348833e-07, "loss": 0.1639, "step": 1260 }, { "epoch": 0.26, "learning_rate": 2.607246677740864e-07, "loss": 0.1435, "step": 1261 }, { "epoch": 0.26, "learning_rate": 2.6069352159468436e-07, "loss": 0.1604, "step": 1262 }, { "epoch": 0.26, "learning_rate": 2.6066237541528235e-07, "loss": 0.1617, "step": 1263 }, { "epoch": 0.26, "learning_rate": 2.606312292358804e-07, "loss": 0.2078, "step": 1264 }, { "epoch": 0.26, "learning_rate": 2.606000830564784e-07, "loss": 0.1706, "step": 1265 }, { "epoch": 0.26, "learning_rate": 2.6056893687707643e-07, "loss": 0.1348, "step": 1266 }, { "epoch": 0.26, "learning_rate": 2.605377906976744e-07, "loss": 0.1585, "step": 1267 }, { "epoch": 0.26, "learning_rate": 2.605066445182724e-07, "loss": 0.2068, "step": 1268 }, { "epoch": 0.26, "learning_rate": 2.604754983388704e-07, "loss": 0.1314, "step": 1269 }, { "epoch": 0.26, "learning_rate": 2.6044435215946844e-07, "loss": 0.1699, "step": 1270 }, { "epoch": 0.26, "learning_rate": 2.604132059800664e-07, "loss": 0.1877, "step": 1271 }, { "epoch": 0.26, "learning_rate": 2.6038205980066447e-07, "loss": 0.1996, "step": 1272 }, { "epoch": 0.26, "learning_rate": 2.6035091362126246e-07, "loss": 0.1374, "step": 1273 }, { "epoch": 0.26, "learning_rate": 2.6031976744186045e-07, "loss": 0.1256, "step": 1274 }, { "epoch": 0.26, "learning_rate": 2.602886212624585e-07, "loss": 0.1529, "step": 1275 }, { "epoch": 0.26, "learning_rate": 2.602574750830565e-07, "loss": 0.139, "step": 1276 }, { "epoch": 0.27, "learning_rate": 2.6022632890365447e-07, "loss": 0.1908, "step": 1277 }, { "epoch": 0.27, "learning_rate": 2.6019518272425246e-07, "loss": 0.2166, "step": 1278 }, { "epoch": 0.27, "learning_rate": 2.601640365448505e-07, "loss": 0.1296, "step": 1279 }, { "epoch": 0.27, "learning_rate": 2.601328903654485e-07, "loss": 0.155, "step": 1280 }, { "epoch": 0.27, "learning_rate": 2.6010174418604653e-07, "loss": 0.1316, "step": 1281 }, { "epoch": 0.27, "learning_rate": 2.6007059800664447e-07, "loss": 0.2331, "step": 1282 }, { "epoch": 0.27, "learning_rate": 2.600394518272425e-07, "loss": 0.1653, "step": 1283 }, { "epoch": 0.27, "learning_rate": 2.600083056478405e-07, "loss": 0.1089, "step": 1284 }, { "epoch": 0.27, "learning_rate": 2.5997715946843854e-07, "loss": 0.1395, "step": 1285 }, { "epoch": 0.27, "learning_rate": 2.5994601328903653e-07, "loss": 0.13, "step": 1286 }, { "epoch": 0.27, "learning_rate": 2.599148671096345e-07, "loss": 0.1595, "step": 1287 }, { "epoch": 0.27, "learning_rate": 2.5988372093023256e-07, "loss": 0.2027, "step": 1288 }, { "epoch": 0.27, "learning_rate": 2.5985257475083055e-07, "loss": 0.1991, "step": 1289 }, { "epoch": 0.27, "learning_rate": 2.598214285714286e-07, "loss": 0.1346, "step": 1290 }, { "epoch": 0.27, "learning_rate": 2.597902823920266e-07, "loss": 0.1808, "step": 1291 }, { "epoch": 0.27, "learning_rate": 2.5975913621262457e-07, "loss": 0.1564, "step": 1292 }, { "epoch": 0.27, "learning_rate": 2.5972799003322256e-07, "loss": 0.1694, "step": 1293 }, { "epoch": 0.27, "learning_rate": 2.596968438538206e-07, "loss": 0.1419, "step": 1294 }, { "epoch": 0.27, "learning_rate": 2.596656976744186e-07, "loss": 0.1628, "step": 1295 }, { "epoch": 0.27, "learning_rate": 2.596345514950166e-07, "loss": 0.1722, "step": 1296 }, { "epoch": 0.27, "learning_rate": 2.5960340531561457e-07, "loss": 0.1626, "step": 1297 }, { "epoch": 0.27, "learning_rate": 2.595722591362126e-07, "loss": 0.2045, "step": 1298 }, { "epoch": 0.27, "learning_rate": 2.595411129568106e-07, "loss": 0.0874, "step": 1299 }, { "epoch": 0.27, "learning_rate": 2.5950996677740865e-07, "loss": 0.1491, "step": 1300 }, { "epoch": 0.27, "learning_rate": 2.5947882059800664e-07, "loss": 0.183, "step": 1301 }, { "epoch": 0.27, "learning_rate": 2.594476744186046e-07, "loss": 0.1553, "step": 1302 }, { "epoch": 0.27, "learning_rate": 2.5941652823920267e-07, "loss": 0.15, "step": 1303 }, { "epoch": 0.27, "learning_rate": 2.5938538205980066e-07, "loss": 0.1514, "step": 1304 }, { "epoch": 0.27, "learning_rate": 2.593542358803987e-07, "loss": 0.1305, "step": 1305 }, { "epoch": 0.27, "learning_rate": 2.5932308970099664e-07, "loss": 0.1149, "step": 1306 }, { "epoch": 0.27, "learning_rate": 2.592919435215947e-07, "loss": 0.2199, "step": 1307 }, { "epoch": 0.27, "learning_rate": 2.5926079734219267e-07, "loss": 0.1256, "step": 1308 }, { "epoch": 0.27, "learning_rate": 2.592296511627907e-07, "loss": 0.1563, "step": 1309 }, { "epoch": 0.27, "learning_rate": 2.591985049833887e-07, "loss": 0.1624, "step": 1310 }, { "epoch": 0.27, "learning_rate": 2.591673588039867e-07, "loss": 0.1081, "step": 1311 }, { "epoch": 0.27, "learning_rate": 2.5913621262458473e-07, "loss": 0.1552, "step": 1312 }, { "epoch": 0.27, "learning_rate": 2.591050664451827e-07, "loss": 0.1358, "step": 1313 }, { "epoch": 0.27, "learning_rate": 2.590739202657807e-07, "loss": 0.1527, "step": 1314 }, { "epoch": 0.27, "learning_rate": 2.590427740863787e-07, "loss": 0.1933, "step": 1315 }, { "epoch": 0.27, "learning_rate": 2.5901162790697674e-07, "loss": 0.1496, "step": 1316 }, { "epoch": 0.27, "learning_rate": 2.5898048172757473e-07, "loss": 0.1245, "step": 1317 }, { "epoch": 0.27, "learning_rate": 2.5894933554817277e-07, "loss": 0.1489, "step": 1318 }, { "epoch": 0.27, "learning_rate": 2.5891818936877076e-07, "loss": 0.1588, "step": 1319 }, { "epoch": 0.27, "learning_rate": 2.5888704318936875e-07, "loss": 0.1595, "step": 1320 }, { "epoch": 0.27, "learning_rate": 2.5885589700996674e-07, "loss": 0.2057, "step": 1321 }, { "epoch": 0.27, "learning_rate": 2.588247508305648e-07, "loss": 0.1503, "step": 1322 }, { "epoch": 0.27, "learning_rate": 2.5879360465116277e-07, "loss": 0.1284, "step": 1323 }, { "epoch": 0.27, "learning_rate": 2.5876245847176076e-07, "loss": 0.1751, "step": 1324 }, { "epoch": 0.28, "learning_rate": 2.587313122923588e-07, "loss": 0.1877, "step": 1325 }, { "epoch": 0.28, "learning_rate": 2.587001661129568e-07, "loss": 0.1702, "step": 1326 }, { "epoch": 0.28, "learning_rate": 2.5866901993355484e-07, "loss": 0.1273, "step": 1327 }, { "epoch": 0.28, "learning_rate": 2.586378737541528e-07, "loss": 0.1959, "step": 1328 }, { "epoch": 0.28, "learning_rate": 2.586067275747508e-07, "loss": 0.1211, "step": 1329 }, { "epoch": 0.28, "learning_rate": 2.585755813953488e-07, "loss": 0.1489, "step": 1330 }, { "epoch": 0.28, "learning_rate": 2.5854443521594685e-07, "loss": 0.2164, "step": 1331 }, { "epoch": 0.28, "learning_rate": 2.5851328903654484e-07, "loss": 0.172, "step": 1332 }, { "epoch": 0.28, "learning_rate": 2.584821428571429e-07, "loss": 0.1707, "step": 1333 }, { "epoch": 0.28, "learning_rate": 2.584509966777408e-07, "loss": 0.1101, "step": 1334 }, { "epoch": 0.28, "learning_rate": 2.5841985049833886e-07, "loss": 0.1573, "step": 1335 }, { "epoch": 0.28, "learning_rate": 2.5838870431893685e-07, "loss": 0.1686, "step": 1336 }, { "epoch": 0.28, "learning_rate": 2.583575581395349e-07, "loss": 0.162, "step": 1337 }, { "epoch": 0.28, "learning_rate": 2.583264119601329e-07, "loss": 0.2087, "step": 1338 }, { "epoch": 0.28, "learning_rate": 2.5829526578073087e-07, "loss": 0.128, "step": 1339 }, { "epoch": 0.28, "learning_rate": 2.582641196013289e-07, "loss": 0.1498, "step": 1340 }, { "epoch": 0.28, "learning_rate": 2.582329734219269e-07, "loss": 0.1406, "step": 1341 }, { "epoch": 0.28, "learning_rate": 2.5820182724252494e-07, "loss": 0.193, "step": 1342 }, { "epoch": 0.28, "learning_rate": 2.581706810631229e-07, "loss": 0.1232, "step": 1343 }, { "epoch": 0.28, "learning_rate": 2.581395348837209e-07, "loss": 0.1569, "step": 1344 }, { "epoch": 0.28, "learning_rate": 2.581083887043189e-07, "loss": 0.167, "step": 1345 }, { "epoch": 0.28, "learning_rate": 2.5807724252491695e-07, "loss": 0.1482, "step": 1346 }, { "epoch": 0.28, "learning_rate": 2.5804609634551494e-07, "loss": 0.1331, "step": 1347 }, { "epoch": 0.28, "learning_rate": 2.5801495016611293e-07, "loss": 0.2004, "step": 1348 }, { "epoch": 0.28, "learning_rate": 2.5798380398671097e-07, "loss": 0.143, "step": 1349 }, { "epoch": 0.28, "learning_rate": 2.5795265780730896e-07, "loss": 0.171, "step": 1350 }, { "epoch": 0.28, "learning_rate": 2.5792151162790695e-07, "loss": 0.151, "step": 1351 }, { "epoch": 0.28, "learning_rate": 2.57890365448505e-07, "loss": 0.1226, "step": 1352 }, { "epoch": 0.28, "learning_rate": 2.57859219269103e-07, "loss": 0.1387, "step": 1353 }, { "epoch": 0.28, "learning_rate": 2.5782807308970097e-07, "loss": 0.1348, "step": 1354 }, { "epoch": 0.28, "learning_rate": 2.57796926910299e-07, "loss": 0.1706, "step": 1355 }, { "epoch": 0.28, "learning_rate": 2.57765780730897e-07, "loss": 0.1268, "step": 1356 }, { "epoch": 0.28, "learning_rate": 2.57734634551495e-07, "loss": 0.1103, "step": 1357 }, { "epoch": 0.28, "learning_rate": 2.57703488372093e-07, "loss": 0.1709, "step": 1358 }, { "epoch": 0.28, "learning_rate": 2.57672342192691e-07, "loss": 0.1443, "step": 1359 }, { "epoch": 0.28, "learning_rate": 2.57641196013289e-07, "loss": 0.163, "step": 1360 }, { "epoch": 0.28, "learning_rate": 2.5761004983388706e-07, "loss": 0.179, "step": 1361 }, { "epoch": 0.28, "learning_rate": 2.5757890365448505e-07, "loss": 0.1213, "step": 1362 }, { "epoch": 0.28, "learning_rate": 2.5754775747508304e-07, "loss": 0.136, "step": 1363 }, { "epoch": 0.28, "learning_rate": 2.575166112956811e-07, "loss": 0.1922, "step": 1364 }, { "epoch": 0.28, "learning_rate": 2.5748546511627907e-07, "loss": 0.156, "step": 1365 }, { "epoch": 0.28, "learning_rate": 2.5745431893687706e-07, "loss": 0.1322, "step": 1366 }, { "epoch": 0.28, "learning_rate": 2.5742317275747505e-07, "loss": 0.1053, "step": 1367 }, { "epoch": 0.28, "learning_rate": 2.573920265780731e-07, "loss": 0.1566, "step": 1368 }, { "epoch": 0.28, "learning_rate": 2.573608803986711e-07, "loss": 0.1529, "step": 1369 }, { "epoch": 0.28, "learning_rate": 2.573297342192691e-07, "loss": 0.1752, "step": 1370 }, { "epoch": 0.28, "learning_rate": 2.5729858803986706e-07, "loss": 0.138, "step": 1371 }, { "epoch": 0.28, "learning_rate": 2.572674418604651e-07, "loss": 0.1482, "step": 1372 }, { "epoch": 0.29, "learning_rate": 2.572362956810631e-07, "loss": 0.1169, "step": 1373 }, { "epoch": 0.29, "learning_rate": 2.5720514950166113e-07, "loss": 0.1458, "step": 1374 }, { "epoch": 0.29, "learning_rate": 2.571740033222591e-07, "loss": 0.1906, "step": 1375 }, { "epoch": 0.29, "learning_rate": 2.571428571428571e-07, "loss": 0.1079, "step": 1376 }, { "epoch": 0.29, "learning_rate": 2.5711171096345515e-07, "loss": 0.1733, "step": 1377 }, { "epoch": 0.29, "learning_rate": 2.5708056478405314e-07, "loss": 0.1277, "step": 1378 }, { "epoch": 0.29, "learning_rate": 2.570494186046512e-07, "loss": 0.13, "step": 1379 }, { "epoch": 0.29, "learning_rate": 2.5701827242524917e-07, "loss": 0.1753, "step": 1380 }, { "epoch": 0.29, "learning_rate": 2.5698712624584716e-07, "loss": 0.1096, "step": 1381 }, { "epoch": 0.29, "learning_rate": 2.5695598006644515e-07, "loss": 0.1652, "step": 1382 }, { "epoch": 0.29, "learning_rate": 2.569248338870432e-07, "loss": 0.1578, "step": 1383 }, { "epoch": 0.29, "learning_rate": 2.568936877076412e-07, "loss": 0.137, "step": 1384 }, { "epoch": 0.29, "learning_rate": 2.568625415282392e-07, "loss": 0.1421, "step": 1385 }, { "epoch": 0.29, "learning_rate": 2.568313953488372e-07, "loss": 0.1362, "step": 1386 }, { "epoch": 0.29, "learning_rate": 2.568002491694352e-07, "loss": 0.1604, "step": 1387 }, { "epoch": 0.29, "learning_rate": 2.567691029900332e-07, "loss": 0.1317, "step": 1388 }, { "epoch": 0.29, "learning_rate": 2.5673795681063123e-07, "loss": 0.1535, "step": 1389 }, { "epoch": 0.29, "learning_rate": 2.567068106312292e-07, "loss": 0.1667, "step": 1390 }, { "epoch": 0.29, "learning_rate": 2.566756644518272e-07, "loss": 0.1199, "step": 1391 }, { "epoch": 0.29, "learning_rate": 2.5664451827242526e-07, "loss": 0.1522, "step": 1392 }, { "epoch": 0.29, "learning_rate": 2.5661337209302325e-07, "loss": 0.1585, "step": 1393 }, { "epoch": 0.29, "learning_rate": 2.565822259136213e-07, "loss": 0.1634, "step": 1394 }, { "epoch": 0.29, "learning_rate": 2.565510797342192e-07, "loss": 0.2116, "step": 1395 }, { "epoch": 0.29, "learning_rate": 2.5651993355481727e-07, "loss": 0.2191, "step": 1396 }, { "epoch": 0.29, "learning_rate": 2.5648878737541526e-07, "loss": 0.1317, "step": 1397 }, { "epoch": 0.29, "learning_rate": 2.564576411960133e-07, "loss": 0.1372, "step": 1398 }, { "epoch": 0.29, "learning_rate": 2.564264950166113e-07, "loss": 0.1996, "step": 1399 }, { "epoch": 0.29, "learning_rate": 2.563953488372093e-07, "loss": 0.1474, "step": 1400 }, { "epoch": 0.29, "learning_rate": 2.563642026578073e-07, "loss": 0.1334, "step": 1401 }, { "epoch": 0.29, "learning_rate": 2.563330564784053e-07, "loss": 0.1628, "step": 1402 }, { "epoch": 0.29, "learning_rate": 2.563019102990033e-07, "loss": 0.1684, "step": 1403 }, { "epoch": 0.29, "learning_rate": 2.562707641196013e-07, "loss": 0.1899, "step": 1404 }, { "epoch": 0.29, "learning_rate": 2.5623961794019933e-07, "loss": 0.1515, "step": 1405 }, { "epoch": 0.29, "learning_rate": 2.562084717607973e-07, "loss": 0.14, "step": 1406 }, { "epoch": 0.29, "learning_rate": 2.5617732558139536e-07, "loss": 0.0996, "step": 1407 }, { "epoch": 0.29, "learning_rate": 2.5614617940199335e-07, "loss": 0.1243, "step": 1408 }, { "epoch": 0.29, "learning_rate": 2.5611503322259134e-07, "loss": 0.1472, "step": 1409 }, { "epoch": 0.29, "learning_rate": 2.5608388704318933e-07, "loss": 0.1716, "step": 1410 }, { "epoch": 0.29, "learning_rate": 2.5605274086378737e-07, "loss": 0.1898, "step": 1411 }, { "epoch": 0.29, "learning_rate": 2.5602159468438536e-07, "loss": 0.1345, "step": 1412 }, { "epoch": 0.29, "learning_rate": 2.559904485049834e-07, "loss": 0.1545, "step": 1413 }, { "epoch": 0.29, "learning_rate": 2.559593023255814e-07, "loss": 0.1328, "step": 1414 }, { "epoch": 0.29, "learning_rate": 2.559281561461794e-07, "loss": 0.1189, "step": 1415 }, { "epoch": 0.29, "learning_rate": 2.558970099667774e-07, "loss": 0.1064, "step": 1416 }, { "epoch": 0.29, "learning_rate": 2.558658637873754e-07, "loss": 0.122, "step": 1417 }, { "epoch": 0.29, "learning_rate": 2.558347176079734e-07, "loss": 0.2183, "step": 1418 }, { "epoch": 0.29, "learning_rate": 2.558035714285714e-07, "loss": 0.2059, "step": 1419 }, { "epoch": 0.29, "learning_rate": 2.5577242524916943e-07, "loss": 0.2169, "step": 1420 }, { "epoch": 0.3, "learning_rate": 2.557412790697674e-07, "loss": 0.093, "step": 1421 }, { "epoch": 0.3, "learning_rate": 2.5571013289036547e-07, "loss": 0.1053, "step": 1422 }, { "epoch": 0.3, "learning_rate": 2.5567898671096346e-07, "loss": 0.1692, "step": 1423 }, { "epoch": 0.3, "learning_rate": 2.5564784053156144e-07, "loss": 0.1826, "step": 1424 }, { "epoch": 0.3, "learning_rate": 2.5561669435215943e-07, "loss": 0.1862, "step": 1425 }, { "epoch": 0.3, "learning_rate": 2.555855481727575e-07, "loss": 0.1561, "step": 1426 }, { "epoch": 0.3, "learning_rate": 2.5555440199335547e-07, "loss": 0.135, "step": 1427 }, { "epoch": 0.3, "learning_rate": 2.5552325581395345e-07, "loss": 0.1197, "step": 1428 }, { "epoch": 0.3, "learning_rate": 2.554921096345515e-07, "loss": 0.1812, "step": 1429 }, { "epoch": 0.3, "learning_rate": 2.554609634551495e-07, "loss": 0.1213, "step": 1430 }, { "epoch": 0.3, "learning_rate": 2.5542981727574753e-07, "loss": 0.0939, "step": 1431 }, { "epoch": 0.3, "learning_rate": 2.553986710963455e-07, "loss": 0.1419, "step": 1432 }, { "epoch": 0.3, "learning_rate": 2.553675249169435e-07, "loss": 0.1198, "step": 1433 }, { "epoch": 0.3, "learning_rate": 2.553363787375415e-07, "loss": 0.0788, "step": 1434 }, { "epoch": 0.3, "learning_rate": 2.5530523255813954e-07, "loss": 0.1357, "step": 1435 }, { "epoch": 0.3, "learning_rate": 2.5527408637873753e-07, "loss": 0.1307, "step": 1436 }, { "epoch": 0.3, "learning_rate": 2.552429401993355e-07, "loss": 0.1494, "step": 1437 }, { "epoch": 0.3, "learning_rate": 2.5521179401993356e-07, "loss": 0.1204, "step": 1438 }, { "epoch": 0.3, "learning_rate": 2.5518064784053155e-07, "loss": 0.1776, "step": 1439 }, { "epoch": 0.3, "learning_rate": 2.5514950166112954e-07, "loss": 0.0839, "step": 1440 }, { "epoch": 0.3, "learning_rate": 2.551183554817276e-07, "loss": 0.1495, "step": 1441 }, { "epoch": 0.3, "learning_rate": 2.5508720930232557e-07, "loss": 0.1307, "step": 1442 }, { "epoch": 0.3, "learning_rate": 2.5505606312292356e-07, "loss": 0.2182, "step": 1443 }, { "epoch": 0.3, "learning_rate": 2.550249169435216e-07, "loss": 0.1862, "step": 1444 }, { "epoch": 0.3, "learning_rate": 2.549937707641196e-07, "loss": 0.1612, "step": 1445 }, { "epoch": 0.3, "learning_rate": 2.5496262458471763e-07, "loss": 0.1391, "step": 1446 }, { "epoch": 0.3, "learning_rate": 2.5493147840531557e-07, "loss": 0.147, "step": 1447 }, { "epoch": 0.3, "learning_rate": 2.549003322259136e-07, "loss": 0.1141, "step": 1448 }, { "epoch": 0.3, "learning_rate": 2.548691860465116e-07, "loss": 0.2195, "step": 1449 }, { "epoch": 0.3, "learning_rate": 2.5483803986710964e-07, "loss": 0.1776, "step": 1450 }, { "epoch": 0.3, "learning_rate": 2.5480689368770763e-07, "loss": 0.1481, "step": 1451 }, { "epoch": 0.3, "learning_rate": 2.547757475083056e-07, "loss": 0.1824, "step": 1452 }, { "epoch": 0.3, "learning_rate": 2.5474460132890367e-07, "loss": 0.0955, "step": 1453 }, { "epoch": 0.3, "learning_rate": 2.5471345514950165e-07, "loss": 0.1693, "step": 1454 }, { "epoch": 0.3, "learning_rate": 2.5468230897009964e-07, "loss": 0.1604, "step": 1455 }, { "epoch": 0.3, "learning_rate": 2.5465116279069763e-07, "loss": 0.1462, "step": 1456 }, { "epoch": 0.3, "learning_rate": 2.546200166112957e-07, "loss": 0.122, "step": 1457 }, { "epoch": 0.3, "learning_rate": 2.5458887043189366e-07, "loss": 0.1358, "step": 1458 }, { "epoch": 0.3, "learning_rate": 2.545577242524917e-07, "loss": 0.1149, "step": 1459 }, { "epoch": 0.3, "learning_rate": 2.545265780730897e-07, "loss": 0.175, "step": 1460 }, { "epoch": 0.3, "learning_rate": 2.544954318936877e-07, "loss": 0.1609, "step": 1461 }, { "epoch": 0.3, "learning_rate": 2.544642857142857e-07, "loss": 0.1103, "step": 1462 }, { "epoch": 0.3, "learning_rate": 2.544331395348837e-07, "loss": 0.0812, "step": 1463 }, { "epoch": 0.3, "learning_rate": 2.544019933554817e-07, "loss": 0.183, "step": 1464 }, { "epoch": 0.3, "learning_rate": 2.5437084717607975e-07, "loss": 0.1758, "step": 1465 }, { "epoch": 0.3, "learning_rate": 2.5433970099667774e-07, "loss": 0.1773, "step": 1466 }, { "epoch": 0.3, "learning_rate": 2.5430855481727573e-07, "loss": 0.1499, "step": 1467 }, { "epoch": 0.3, "learning_rate": 2.5427740863787377e-07, "loss": 0.1489, "step": 1468 }, { "epoch": 0.31, "learning_rate": 2.5424626245847176e-07, "loss": 0.2066, "step": 1469 }, { "epoch": 0.31, "learning_rate": 2.5421511627906975e-07, "loss": 0.1748, "step": 1470 }, { "epoch": 0.31, "learning_rate": 2.5418397009966774e-07, "loss": 0.1291, "step": 1471 }, { "epoch": 0.31, "learning_rate": 2.541528239202658e-07, "loss": 0.1614, "step": 1472 }, { "epoch": 0.31, "learning_rate": 2.5412167774086377e-07, "loss": 0.1504, "step": 1473 }, { "epoch": 0.31, "learning_rate": 2.540905315614618e-07, "loss": 0.1345, "step": 1474 }, { "epoch": 0.31, "learning_rate": 2.540593853820598e-07, "loss": 0.2023, "step": 1475 }, { "epoch": 0.31, "learning_rate": 2.540282392026578e-07, "loss": 0.1624, "step": 1476 }, { "epoch": 0.31, "learning_rate": 2.539970930232558e-07, "loss": 0.1615, "step": 1477 }, { "epoch": 0.31, "learning_rate": 2.539659468438538e-07, "loss": 0.1881, "step": 1478 }, { "epoch": 0.31, "learning_rate": 2.539348006644518e-07, "loss": 0.1444, "step": 1479 }, { "epoch": 0.31, "learning_rate": 2.539036544850498e-07, "loss": 0.1481, "step": 1480 }, { "epoch": 0.31, "learning_rate": 2.5387250830564784e-07, "loss": 0.1679, "step": 1481 }, { "epoch": 0.31, "learning_rate": 2.5384136212624583e-07, "loss": 0.1308, "step": 1482 }, { "epoch": 0.31, "learning_rate": 2.538102159468439e-07, "loss": 0.1249, "step": 1483 }, { "epoch": 0.31, "learning_rate": 2.537790697674418e-07, "loss": 0.2345, "step": 1484 }, { "epoch": 0.31, "learning_rate": 2.5374792358803985e-07, "loss": 0.1402, "step": 1485 }, { "epoch": 0.31, "learning_rate": 2.5371677740863784e-07, "loss": 0.2108, "step": 1486 }, { "epoch": 0.31, "learning_rate": 2.536856312292359e-07, "loss": 0.0991, "step": 1487 }, { "epoch": 0.31, "learning_rate": 2.536544850498339e-07, "loss": 0.1561, "step": 1488 }, { "epoch": 0.31, "learning_rate": 2.5362333887043186e-07, "loss": 0.1808, "step": 1489 }, { "epoch": 0.31, "learning_rate": 2.535921926910299e-07, "loss": 0.1127, "step": 1490 }, { "epoch": 0.31, "learning_rate": 2.535610465116279e-07, "loss": 0.1355, "step": 1491 }, { "epoch": 0.31, "learning_rate": 2.535299003322259e-07, "loss": 0.1804, "step": 1492 }, { "epoch": 0.31, "learning_rate": 2.5349875415282393e-07, "loss": 0.1445, "step": 1493 }, { "epoch": 0.31, "learning_rate": 2.534676079734219e-07, "loss": 0.1301, "step": 1494 }, { "epoch": 0.31, "learning_rate": 2.534364617940199e-07, "loss": 0.1176, "step": 1495 }, { "epoch": 0.31, "learning_rate": 2.5340531561461795e-07, "loss": 0.0989, "step": 1496 }, { "epoch": 0.31, "learning_rate": 2.5337416943521594e-07, "loss": 0.1688, "step": 1497 }, { "epoch": 0.31, "learning_rate": 2.5334302325581393e-07, "loss": 0.1016, "step": 1498 }, { "epoch": 0.31, "learning_rate": 2.533118770764119e-07, "loss": 0.17, "step": 1499 }, { "epoch": 0.31, "learning_rate": 2.5328073089700996e-07, "loss": 0.1319, "step": 1500 }, { "epoch": 0.31, "learning_rate": 2.5324958471760795e-07, "loss": 0.1316, "step": 1501 }, { "epoch": 0.31, "learning_rate": 2.53218438538206e-07, "loss": 0.1765, "step": 1502 }, { "epoch": 0.31, "learning_rate": 2.53187292358804e-07, "loss": 0.1804, "step": 1503 }, { "epoch": 0.31, "learning_rate": 2.5315614617940197e-07, "loss": 0.217, "step": 1504 }, { "epoch": 0.31, "learning_rate": 2.53125e-07, "loss": 0.11, "step": 1505 }, { "epoch": 0.31, "learning_rate": 2.53093853820598e-07, "loss": 0.1332, "step": 1506 }, { "epoch": 0.31, "learning_rate": 2.53062707641196e-07, "loss": 0.1591, "step": 1507 }, { "epoch": 0.31, "learning_rate": 2.53031561461794e-07, "loss": 0.1587, "step": 1508 }, { "epoch": 0.31, "learning_rate": 2.53000415282392e-07, "loss": 0.172, "step": 1509 }, { "epoch": 0.31, "learning_rate": 2.5296926910299e-07, "loss": 0.1551, "step": 1510 }, { "epoch": 0.31, "learning_rate": 2.5293812292358805e-07, "loss": 0.1539, "step": 1511 }, { "epoch": 0.31, "learning_rate": 2.5290697674418604e-07, "loss": 0.1937, "step": 1512 }, { "epoch": 0.31, "learning_rate": 2.5287583056478403e-07, "loss": 0.1444, "step": 1513 }, { "epoch": 0.31, "learning_rate": 2.52844684385382e-07, "loss": 0.1744, "step": 1514 }, { "epoch": 0.31, "learning_rate": 2.5281353820598006e-07, "loss": 0.1512, "step": 1515 }, { "epoch": 0.31, "learning_rate": 2.5278239202657805e-07, "loss": 0.1625, "step": 1516 }, { "epoch": 0.31, "learning_rate": 2.5275124584717604e-07, "loss": 0.1223, "step": 1517 }, { "epoch": 0.32, "learning_rate": 2.527200996677741e-07, "loss": 0.1387, "step": 1518 }, { "epoch": 0.32, "learning_rate": 2.526889534883721e-07, "loss": 0.1434, "step": 1519 }, { "epoch": 0.32, "learning_rate": 2.526578073089701e-07, "loss": 0.1745, "step": 1520 }, { "epoch": 0.32, "learning_rate": 2.526266611295681e-07, "loss": 0.1557, "step": 1521 }, { "epoch": 0.32, "learning_rate": 2.525955149501661e-07, "loss": 0.1717, "step": 1522 }, { "epoch": 0.32, "learning_rate": 2.525643687707641e-07, "loss": 0.1654, "step": 1523 }, { "epoch": 0.32, "learning_rate": 2.5253322259136213e-07, "loss": 0.1337, "step": 1524 }, { "epoch": 0.32, "learning_rate": 2.525020764119601e-07, "loss": 0.1515, "step": 1525 }, { "epoch": 0.32, "learning_rate": 2.5247093023255816e-07, "loss": 0.2177, "step": 1526 }, { "epoch": 0.32, "learning_rate": 2.5243978405315615e-07, "loss": 0.1146, "step": 1527 }, { "epoch": 0.32, "learning_rate": 2.5240863787375414e-07, "loss": 0.1521, "step": 1528 }, { "epoch": 0.32, "learning_rate": 2.5237749169435213e-07, "loss": 0.1235, "step": 1529 }, { "epoch": 0.32, "learning_rate": 2.5234634551495017e-07, "loss": 0.1642, "step": 1530 }, { "epoch": 0.32, "learning_rate": 2.5231519933554816e-07, "loss": 0.0918, "step": 1531 }, { "epoch": 0.32, "learning_rate": 2.5228405315614615e-07, "loss": 0.1311, "step": 1532 }, { "epoch": 0.32, "learning_rate": 2.522529069767442e-07, "loss": 0.0969, "step": 1533 }, { "epoch": 0.32, "learning_rate": 2.522217607973422e-07, "loss": 0.1232, "step": 1534 }, { "epoch": 0.32, "learning_rate": 2.521906146179402e-07, "loss": 0.1686, "step": 1535 }, { "epoch": 0.32, "learning_rate": 2.5215946843853816e-07, "loss": 0.1081, "step": 1536 }, { "epoch": 0.32, "learning_rate": 2.521283222591362e-07, "loss": 0.1056, "step": 1537 }, { "epoch": 0.32, "learning_rate": 2.520971760797342e-07, "loss": 0.1765, "step": 1538 }, { "epoch": 0.32, "learning_rate": 2.5206602990033223e-07, "loss": 0.162, "step": 1539 }, { "epoch": 0.32, "learning_rate": 2.520348837209302e-07, "loss": 0.1628, "step": 1540 }, { "epoch": 0.32, "learning_rate": 2.520037375415282e-07, "loss": 0.1264, "step": 1541 }, { "epoch": 0.32, "learning_rate": 2.5197259136212625e-07, "loss": 0.1506, "step": 1542 }, { "epoch": 0.32, "learning_rate": 2.5194144518272424e-07, "loss": 0.1415, "step": 1543 }, { "epoch": 0.32, "learning_rate": 2.5191029900332223e-07, "loss": 0.129, "step": 1544 }, { "epoch": 0.32, "learning_rate": 2.518791528239202e-07, "loss": 0.1356, "step": 1545 }, { "epoch": 0.32, "learning_rate": 2.5184800664451826e-07, "loss": 0.149, "step": 1546 }, { "epoch": 0.32, "learning_rate": 2.5181686046511625e-07, "loss": 0.1337, "step": 1547 }, { "epoch": 0.32, "learning_rate": 2.517857142857143e-07, "loss": 0.1929, "step": 1548 }, { "epoch": 0.32, "learning_rate": 2.517545681063123e-07, "loss": 0.1537, "step": 1549 }, { "epoch": 0.32, "learning_rate": 2.517234219269103e-07, "loss": 0.181, "step": 1550 }, { "epoch": 0.32, "learning_rate": 2.5169227574750826e-07, "loss": 0.1533, "step": 1551 }, { "epoch": 0.32, "learning_rate": 2.516611295681063e-07, "loss": 0.1918, "step": 1552 }, { "epoch": 0.32, "learning_rate": 2.516299833887043e-07, "loss": 0.0995, "step": 1553 }, { "epoch": 0.32, "learning_rate": 2.5159883720930234e-07, "loss": 0.1366, "step": 1554 }, { "epoch": 0.32, "learning_rate": 2.5156769102990033e-07, "loss": 0.1319, "step": 1555 }, { "epoch": 0.32, "learning_rate": 2.515365448504983e-07, "loss": 0.1231, "step": 1556 }, { "epoch": 0.32, "learning_rate": 2.5150539867109636e-07, "loss": 0.1207, "step": 1557 }, { "epoch": 0.32, "learning_rate": 2.5147425249169435e-07, "loss": 0.1289, "step": 1558 }, { "epoch": 0.32, "learning_rate": 2.514431063122924e-07, "loss": 0.1629, "step": 1559 }, { "epoch": 0.32, "learning_rate": 2.514119601328903e-07, "loss": 0.1236, "step": 1560 }, { "epoch": 0.32, "learning_rate": 2.5138081395348837e-07, "loss": 0.2192, "step": 1561 }, { "epoch": 0.32, "learning_rate": 2.5134966777408636e-07, "loss": 0.114, "step": 1562 }, { "epoch": 0.32, "learning_rate": 2.513185215946844e-07, "loss": 0.1718, "step": 1563 }, { "epoch": 0.32, "learning_rate": 2.512873754152824e-07, "loss": 0.1448, "step": 1564 }, { "epoch": 0.32, "learning_rate": 2.512562292358804e-07, "loss": 0.189, "step": 1565 }, { "epoch": 0.33, "learning_rate": 2.5122508305647837e-07, "loss": 0.1376, "step": 1566 }, { "epoch": 0.33, "learning_rate": 2.511939368770764e-07, "loss": 0.1469, "step": 1567 }, { "epoch": 0.33, "learning_rate": 2.511627906976744e-07, "loss": 0.1398, "step": 1568 }, { "epoch": 0.33, "learning_rate": 2.511316445182724e-07, "loss": 0.1355, "step": 1569 }, { "epoch": 0.33, "learning_rate": 2.5110049833887043e-07, "loss": 0.1179, "step": 1570 }, { "epoch": 0.33, "learning_rate": 2.510693521594684e-07, "loss": 0.1259, "step": 1571 }, { "epoch": 0.33, "learning_rate": 2.5103820598006646e-07, "loss": 0.1258, "step": 1572 }, { "epoch": 0.33, "learning_rate": 2.5100705980066445e-07, "loss": 0.1363, "step": 1573 }, { "epoch": 0.33, "learning_rate": 2.5097591362126244e-07, "loss": 0.1041, "step": 1574 }, { "epoch": 0.33, "learning_rate": 2.5094476744186043e-07, "loss": 0.1349, "step": 1575 }, { "epoch": 0.33, "learning_rate": 2.5091362126245847e-07, "loss": 0.1432, "step": 1576 }, { "epoch": 0.33, "learning_rate": 2.5088247508305646e-07, "loss": 0.1114, "step": 1577 }, { "epoch": 0.33, "learning_rate": 2.5085132890365445e-07, "loss": 0.0952, "step": 1578 }, { "epoch": 0.33, "learning_rate": 2.508201827242525e-07, "loss": 0.1599, "step": 1579 }, { "epoch": 0.33, "learning_rate": 2.507890365448505e-07, "loss": 0.1364, "step": 1580 }, { "epoch": 0.33, "learning_rate": 2.5075789036544847e-07, "loss": 0.1078, "step": 1581 }, { "epoch": 0.33, "learning_rate": 2.507267441860465e-07, "loss": 0.1552, "step": 1582 }, { "epoch": 0.33, "learning_rate": 2.506955980066445e-07, "loss": 0.1539, "step": 1583 }, { "epoch": 0.33, "learning_rate": 2.506644518272425e-07, "loss": 0.1149, "step": 1584 }, { "epoch": 0.33, "learning_rate": 2.5063330564784054e-07, "loss": 0.2042, "step": 1585 }, { "epoch": 0.33, "learning_rate": 2.506021594684385e-07, "loss": 0.0787, "step": 1586 }, { "epoch": 0.33, "learning_rate": 2.5057101328903657e-07, "loss": 0.1639, "step": 1587 }, { "epoch": 0.33, "learning_rate": 2.505398671096345e-07, "loss": 0.1834, "step": 1588 }, { "epoch": 0.33, "learning_rate": 2.5050872093023255e-07, "loss": 0.1046, "step": 1589 }, { "epoch": 0.33, "learning_rate": 2.5047757475083054e-07, "loss": 0.168, "step": 1590 }, { "epoch": 0.33, "learning_rate": 2.504464285714286e-07, "loss": 0.1303, "step": 1591 }, { "epoch": 0.33, "learning_rate": 2.5041528239202657e-07, "loss": 0.1371, "step": 1592 }, { "epoch": 0.33, "learning_rate": 2.5038413621262456e-07, "loss": 0.0795, "step": 1593 }, { "epoch": 0.33, "learning_rate": 2.503529900332226e-07, "loss": 0.1033, "step": 1594 }, { "epoch": 0.33, "learning_rate": 2.503218438538206e-07, "loss": 0.1504, "step": 1595 }, { "epoch": 0.33, "learning_rate": 2.5029069767441863e-07, "loss": 0.1399, "step": 1596 }, { "epoch": 0.33, "learning_rate": 2.5025955149501657e-07, "loss": 0.2113, "step": 1597 }, { "epoch": 0.33, "learning_rate": 2.502284053156146e-07, "loss": 0.216, "step": 1598 }, { "epoch": 0.33, "learning_rate": 2.501972591362126e-07, "loss": 0.1457, "step": 1599 }, { "epoch": 0.33, "learning_rate": 2.5016611295681064e-07, "loss": 0.1083, "step": 1600 }, { "epoch": 0.33, "learning_rate": 2.5013496677740863e-07, "loss": 0.1121, "step": 1601 }, { "epoch": 0.33, "learning_rate": 2.501038205980066e-07, "loss": 0.0977, "step": 1602 }, { "epoch": 0.33, "learning_rate": 2.500726744186046e-07, "loss": 0.1733, "step": 1603 }, { "epoch": 0.33, "learning_rate": 2.5004152823920265e-07, "loss": 0.1852, "step": 1604 }, { "epoch": 0.33, "learning_rate": 2.5001038205980064e-07, "loss": 0.1677, "step": 1605 }, { "epoch": 0.33, "learning_rate": 2.499792358803987e-07, "loss": 0.1031, "step": 1606 }, { "epoch": 0.33, "learning_rate": 2.4994808970099667e-07, "loss": 0.1597, "step": 1607 }, { "epoch": 0.33, "learning_rate": 2.4991694352159466e-07, "loss": 0.1368, "step": 1608 }, { "epoch": 0.33, "learning_rate": 2.498857973421927e-07, "loss": 0.1411, "step": 1609 }, { "epoch": 0.33, "learning_rate": 2.498546511627907e-07, "loss": 0.1361, "step": 1610 }, { "epoch": 0.33, "learning_rate": 2.498235049833887e-07, "loss": 0.1561, "step": 1611 }, { "epoch": 0.33, "learning_rate": 2.4979235880398667e-07, "loss": 0.1029, "step": 1612 }, { "epoch": 0.33, "learning_rate": 2.497612126245847e-07, "loss": 0.1191, "step": 1613 }, { "epoch": 0.34, "learning_rate": 2.497300664451827e-07, "loss": 0.1485, "step": 1614 }, { "epoch": 0.34, "learning_rate": 2.4969892026578075e-07, "loss": 0.1235, "step": 1615 }, { "epoch": 0.34, "learning_rate": 2.4966777408637874e-07, "loss": 0.1389, "step": 1616 }, { "epoch": 0.34, "learning_rate": 2.496366279069767e-07, "loss": 0.1183, "step": 1617 }, { "epoch": 0.34, "learning_rate": 2.496054817275747e-07, "loss": 0.155, "step": 1618 }, { "epoch": 0.34, "learning_rate": 2.4957433554817276e-07, "loss": 0.1421, "step": 1619 }, { "epoch": 0.34, "learning_rate": 2.4954318936877075e-07, "loss": 0.1595, "step": 1620 }, { "epoch": 0.34, "learning_rate": 2.4951204318936874e-07, "loss": 0.1674, "step": 1621 }, { "epoch": 0.34, "learning_rate": 2.494808970099668e-07, "loss": 0.1781, "step": 1622 }, { "epoch": 0.34, "learning_rate": 2.4944975083056477e-07, "loss": 0.1317, "step": 1623 }, { "epoch": 0.34, "learning_rate": 2.494186046511628e-07, "loss": 0.1352, "step": 1624 }, { "epoch": 0.34, "learning_rate": 2.4938745847176075e-07, "loss": 0.1094, "step": 1625 }, { "epoch": 0.34, "learning_rate": 2.493563122923588e-07, "loss": 0.127, "step": 1626 }, { "epoch": 0.34, "learning_rate": 2.493251661129568e-07, "loss": 0.1159, "step": 1627 }, { "epoch": 0.34, "learning_rate": 2.492940199335548e-07, "loss": 0.1876, "step": 1628 }, { "epoch": 0.34, "learning_rate": 2.492628737541528e-07, "loss": 0.1665, "step": 1629 }, { "epoch": 0.34, "learning_rate": 2.492317275747508e-07, "loss": 0.1209, "step": 1630 }, { "epoch": 0.34, "learning_rate": 2.4920058139534884e-07, "loss": 0.1028, "step": 1631 }, { "epoch": 0.34, "learning_rate": 2.4916943521594683e-07, "loss": 0.0899, "step": 1632 }, { "epoch": 0.34, "learning_rate": 2.4913828903654487e-07, "loss": 0.1436, "step": 1633 }, { "epoch": 0.34, "learning_rate": 2.4910714285714286e-07, "loss": 0.1414, "step": 1634 }, { "epoch": 0.34, "learning_rate": 2.4907599667774085e-07, "loss": 0.1516, "step": 1635 }, { "epoch": 0.34, "learning_rate": 2.4904485049833884e-07, "loss": 0.0916, "step": 1636 }, { "epoch": 0.34, "learning_rate": 2.490137043189369e-07, "loss": 0.2078, "step": 1637 }, { "epoch": 0.34, "learning_rate": 2.4898255813953487e-07, "loss": 0.15, "step": 1638 }, { "epoch": 0.34, "learning_rate": 2.489514119601329e-07, "loss": 0.1677, "step": 1639 }, { "epoch": 0.34, "learning_rate": 2.4892026578073085e-07, "loss": 0.1792, "step": 1640 }, { "epoch": 0.34, "learning_rate": 2.488891196013289e-07, "loss": 0.1253, "step": 1641 }, { "epoch": 0.34, "learning_rate": 2.488579734219269e-07, "loss": 0.1773, "step": 1642 }, { "epoch": 0.34, "learning_rate": 2.488268272425249e-07, "loss": 0.1009, "step": 1643 }, { "epoch": 0.34, "learning_rate": 2.487956810631229e-07, "loss": 0.1616, "step": 1644 }, { "epoch": 0.34, "learning_rate": 2.487645348837209e-07, "loss": 0.1119, "step": 1645 }, { "epoch": 0.34, "learning_rate": 2.4873338870431895e-07, "loss": 0.1438, "step": 1646 }, { "epoch": 0.34, "learning_rate": 2.4870224252491694e-07, "loss": 0.1251, "step": 1647 }, { "epoch": 0.34, "learning_rate": 2.48671096345515e-07, "loss": 0.13, "step": 1648 }, { "epoch": 0.34, "learning_rate": 2.486399501661129e-07, "loss": 0.122, "step": 1649 }, { "epoch": 0.34, "learning_rate": 2.4860880398671096e-07, "loss": 0.1237, "step": 1650 }, { "epoch": 0.34, "learning_rate": 2.4857765780730895e-07, "loss": 0.1165, "step": 1651 }, { "epoch": 0.34, "learning_rate": 2.48546511627907e-07, "loss": 0.1617, "step": 1652 }, { "epoch": 0.34, "learning_rate": 2.48515365448505e-07, "loss": 0.1207, "step": 1653 }, { "epoch": 0.34, "learning_rate": 2.4848421926910297e-07, "loss": 0.1456, "step": 1654 }, { "epoch": 0.34, "learning_rate": 2.4845307308970096e-07, "loss": 0.1511, "step": 1655 }, { "epoch": 0.34, "learning_rate": 2.48421926910299e-07, "loss": 0.1365, "step": 1656 }, { "epoch": 0.34, "learning_rate": 2.48390780730897e-07, "loss": 0.0829, "step": 1657 }, { "epoch": 0.34, "learning_rate": 2.48359634551495e-07, "loss": 0.1326, "step": 1658 }, { "epoch": 0.34, "learning_rate": 2.48328488372093e-07, "loss": 0.1107, "step": 1659 }, { "epoch": 0.34, "learning_rate": 2.48297342192691e-07, "loss": 0.0789, "step": 1660 }, { "epoch": 0.34, "learning_rate": 2.4826619601328905e-07, "loss": 0.115, "step": 1661 }, { "epoch": 0.35, "learning_rate": 2.4823504983388704e-07, "loss": 0.1361, "step": 1662 }, { "epoch": 0.35, "learning_rate": 2.4820390365448503e-07, "loss": 0.0979, "step": 1663 }, { "epoch": 0.35, "learning_rate": 2.48172757475083e-07, "loss": 0.1619, "step": 1664 }, { "epoch": 0.35, "learning_rate": 2.4814161129568106e-07, "loss": 0.1399, "step": 1665 }, { "epoch": 0.35, "learning_rate": 2.4811046511627905e-07, "loss": 0.1032, "step": 1666 }, { "epoch": 0.35, "learning_rate": 2.480793189368771e-07, "loss": 0.1985, "step": 1667 }, { "epoch": 0.35, "learning_rate": 2.480481727574751e-07, "loss": 0.201, "step": 1668 }, { "epoch": 0.35, "learning_rate": 2.4801702657807307e-07, "loss": 0.1729, "step": 1669 }, { "epoch": 0.35, "learning_rate": 2.479858803986711e-07, "loss": 0.2115, "step": 1670 }, { "epoch": 0.35, "learning_rate": 2.479547342192691e-07, "loss": 0.1545, "step": 1671 }, { "epoch": 0.35, "learning_rate": 2.479235880398671e-07, "loss": 0.2351, "step": 1672 }, { "epoch": 0.35, "learning_rate": 2.478924418604651e-07, "loss": 0.2031, "step": 1673 }, { "epoch": 0.35, "learning_rate": 2.478612956810631e-07, "loss": 0.191, "step": 1674 }, { "epoch": 0.35, "learning_rate": 2.478301495016611e-07, "loss": 0.2231, "step": 1675 }, { "epoch": 0.35, "learning_rate": 2.4779900332225916e-07, "loss": 0.1617, "step": 1676 }, { "epoch": 0.35, "learning_rate": 2.477678571428571e-07, "loss": 0.215, "step": 1677 }, { "epoch": 0.35, "learning_rate": 2.4773671096345513e-07, "loss": 0.234, "step": 1678 }, { "epoch": 0.35, "learning_rate": 2.477055647840531e-07, "loss": 0.1819, "step": 1679 }, { "epoch": 0.35, "learning_rate": 2.4767441860465117e-07, "loss": 0.2073, "step": 1680 }, { "epoch": 0.35, "learning_rate": 2.4764327242524916e-07, "loss": 0.2401, "step": 1681 }, { "epoch": 0.35, "learning_rate": 2.4761212624584715e-07, "loss": 0.2131, "step": 1682 }, { "epoch": 0.35, "learning_rate": 2.475809800664452e-07, "loss": 0.2114, "step": 1683 }, { "epoch": 0.35, "learning_rate": 2.475498338870432e-07, "loss": 0.2099, "step": 1684 }, { "epoch": 0.35, "learning_rate": 2.475186877076412e-07, "loss": 0.1815, "step": 1685 }, { "epoch": 0.35, "learning_rate": 2.474875415282392e-07, "loss": 0.2186, "step": 1686 }, { "epoch": 0.35, "learning_rate": 2.474875415282392e-07, "loss": 0.1655, "step": 1687 }, { "epoch": 0.35, "learning_rate": 2.474563953488372e-07, "loss": 0.1904, "step": 1688 }, { "epoch": 0.35, "learning_rate": 2.474252491694352e-07, "loss": 0.1931, "step": 1689 }, { "epoch": 0.35, "learning_rate": 2.4739410299003323e-07, "loss": 0.2384, "step": 1690 }, { "epoch": 0.35, "learning_rate": 2.473629568106312e-07, "loss": 0.2241, "step": 1691 }, { "epoch": 0.35, "learning_rate": 2.473318106312292e-07, "loss": 0.1626, "step": 1692 }, { "epoch": 0.35, "learning_rate": 2.473006644518272e-07, "loss": 0.2578, "step": 1693 }, { "epoch": 0.35, "learning_rate": 2.4726951827242524e-07, "loss": 0.1884, "step": 1694 }, { "epoch": 0.35, "learning_rate": 2.4723837209302323e-07, "loss": 0.1518, "step": 1695 }, { "epoch": 0.35, "learning_rate": 2.4720722591362127e-07, "loss": 0.1974, "step": 1696 }, { "epoch": 0.35, "learning_rate": 2.4717607973421926e-07, "loss": 0.209, "step": 1697 }, { "epoch": 0.35, "learning_rate": 2.4714493355481725e-07, "loss": 0.2234, "step": 1698 }, { "epoch": 0.35, "learning_rate": 2.471137873754153e-07, "loss": 0.2311, "step": 1699 }, { "epoch": 0.35, "learning_rate": 2.470826411960133e-07, "loss": 0.2238, "step": 1700 }, { "epoch": 0.35, "learning_rate": 2.470514950166113e-07, "loss": 0.2472, "step": 1701 }, { "epoch": 0.35, "learning_rate": 2.4702034883720926e-07, "loss": 0.24, "step": 1702 }, { "epoch": 0.35, "learning_rate": 2.469892026578073e-07, "loss": 0.2347, "step": 1703 }, { "epoch": 0.35, "learning_rate": 2.469580564784053e-07, "loss": 0.2032, "step": 1704 }, { "epoch": 0.35, "learning_rate": 2.4692691029900333e-07, "loss": 0.1749, "step": 1705 }, { "epoch": 0.35, "learning_rate": 2.468957641196013e-07, "loss": 0.3205, "step": 1706 }, { "epoch": 0.35, "learning_rate": 2.468646179401993e-07, "loss": 0.2164, "step": 1707 }, { "epoch": 0.35, "learning_rate": 2.4683347176079736e-07, "loss": 0.207, "step": 1708 }, { "epoch": 0.35, "learning_rate": 2.4680232558139534e-07, "loss": 0.1575, "step": 1709 }, { "epoch": 0.36, "learning_rate": 2.4677117940199333e-07, "loss": 0.172, "step": 1710 }, { "epoch": 0.36, "learning_rate": 2.467400332225913e-07, "loss": 0.2494, "step": 1711 }, { "epoch": 0.36, "learning_rate": 2.4670888704318937e-07, "loss": 0.148, "step": 1712 }, { "epoch": 0.36, "learning_rate": 2.4667774086378736e-07, "loss": 0.2221, "step": 1713 }, { "epoch": 0.36, "learning_rate": 2.466465946843854e-07, "loss": 0.1856, "step": 1714 }, { "epoch": 0.36, "learning_rate": 2.466154485049834e-07, "loss": 0.1677, "step": 1715 }, { "epoch": 0.36, "learning_rate": 2.465843023255814e-07, "loss": 0.1355, "step": 1716 }, { "epoch": 0.36, "learning_rate": 2.4655315614617937e-07, "loss": 0.1801, "step": 1717 }, { "epoch": 0.36, "learning_rate": 2.465220099667774e-07, "loss": 0.2019, "step": 1718 }, { "epoch": 0.36, "learning_rate": 2.464908637873754e-07, "loss": 0.2019, "step": 1719 }, { "epoch": 0.36, "learning_rate": 2.464597176079734e-07, "loss": 0.2194, "step": 1720 }, { "epoch": 0.36, "learning_rate": 2.4642857142857143e-07, "loss": 0.2096, "step": 1721 }, { "epoch": 0.36, "learning_rate": 2.463974252491694e-07, "loss": 0.2109, "step": 1722 }, { "epoch": 0.36, "learning_rate": 2.4636627906976746e-07, "loss": 0.196, "step": 1723 }, { "epoch": 0.36, "learning_rate": 2.4633513289036545e-07, "loss": 0.1887, "step": 1724 }, { "epoch": 0.36, "learning_rate": 2.4630398671096344e-07, "loss": 0.1843, "step": 1725 }, { "epoch": 0.36, "learning_rate": 2.4627284053156143e-07, "loss": 0.2091, "step": 1726 }, { "epoch": 0.36, "learning_rate": 2.4624169435215947e-07, "loss": 0.2065, "step": 1727 }, { "epoch": 0.36, "learning_rate": 2.4621054817275746e-07, "loss": 0.2401, "step": 1728 }, { "epoch": 0.36, "learning_rate": 2.461794019933555e-07, "loss": 0.2204, "step": 1729 }, { "epoch": 0.36, "learning_rate": 2.4614825581395344e-07, "loss": 0.1979, "step": 1730 }, { "epoch": 0.36, "learning_rate": 2.461171096345515e-07, "loss": 0.2534, "step": 1731 }, { "epoch": 0.36, "learning_rate": 2.4608596345514947e-07, "loss": 0.2015, "step": 1732 }, { "epoch": 0.36, "learning_rate": 2.460548172757475e-07, "loss": 0.214, "step": 1733 }, { "epoch": 0.36, "learning_rate": 2.460236710963455e-07, "loss": 0.1572, "step": 1734 }, { "epoch": 0.36, "learning_rate": 2.459925249169435e-07, "loss": 0.2237, "step": 1735 }, { "epoch": 0.36, "learning_rate": 2.4596137873754153e-07, "loss": 0.2359, "step": 1736 }, { "epoch": 0.36, "learning_rate": 2.459302325581395e-07, "loss": 0.2012, "step": 1737 }, { "epoch": 0.36, "learning_rate": 2.4589908637873757e-07, "loss": 0.2319, "step": 1738 }, { "epoch": 0.36, "learning_rate": 2.458679401993355e-07, "loss": 0.1559, "step": 1739 }, { "epoch": 0.36, "learning_rate": 2.4583679401993354e-07, "loss": 0.1991, "step": 1740 }, { "epoch": 0.36, "learning_rate": 2.4580564784053153e-07, "loss": 0.2004, "step": 1741 }, { "epoch": 0.36, "learning_rate": 2.457745016611296e-07, "loss": 0.2178, "step": 1742 }, { "epoch": 0.36, "learning_rate": 2.4574335548172757e-07, "loss": 0.2111, "step": 1743 }, { "epoch": 0.36, "learning_rate": 2.4571220930232555e-07, "loss": 0.1472, "step": 1744 }, { "epoch": 0.36, "learning_rate": 2.456810631229236e-07, "loss": 0.1423, "step": 1745 }, { "epoch": 0.36, "learning_rate": 2.456499169435216e-07, "loss": 0.2127, "step": 1746 }, { "epoch": 0.36, "learning_rate": 2.456187707641196e-07, "loss": 0.2013, "step": 1747 }, { "epoch": 0.36, "learning_rate": 2.455876245847176e-07, "loss": 0.1519, "step": 1748 }, { "epoch": 0.36, "learning_rate": 2.455564784053156e-07, "loss": 0.2059, "step": 1749 }, { "epoch": 0.36, "learning_rate": 2.455253322259136e-07, "loss": 0.2216, "step": 1750 }, { "epoch": 0.36, "learning_rate": 2.4549418604651164e-07, "loss": 0.1877, "step": 1751 }, { "epoch": 0.36, "learning_rate": 2.4546303986710963e-07, "loss": 0.172, "step": 1752 }, { "epoch": 0.36, "learning_rate": 2.454318936877076e-07, "loss": 0.2271, "step": 1753 }, { "epoch": 0.36, "learning_rate": 2.454007475083056e-07, "loss": 0.1813, "step": 1754 }, { "epoch": 0.36, "learning_rate": 2.4536960132890365e-07, "loss": 0.2669, "step": 1755 }, { "epoch": 0.36, "learning_rate": 2.4533845514950164e-07, "loss": 0.1946, "step": 1756 }, { "epoch": 0.36, "learning_rate": 2.453073089700997e-07, "loss": 0.1563, "step": 1757 }, { "epoch": 0.37, "learning_rate": 2.4527616279069767e-07, "loss": 0.2492, "step": 1758 }, { "epoch": 0.37, "learning_rate": 2.4524501661129566e-07, "loss": 0.1861, "step": 1759 }, { "epoch": 0.37, "learning_rate": 2.452138704318937e-07, "loss": 0.2326, "step": 1760 }, { "epoch": 0.37, "learning_rate": 2.451827242524917e-07, "loss": 0.2683, "step": 1761 }, { "epoch": 0.37, "learning_rate": 2.451515780730897e-07, "loss": 0.1763, "step": 1762 }, { "epoch": 0.37, "learning_rate": 2.4512043189368767e-07, "loss": 0.2047, "step": 1763 }, { "epoch": 0.37, "learning_rate": 2.450892857142857e-07, "loss": 0.2025, "step": 1764 }, { "epoch": 0.37, "learning_rate": 2.450581395348837e-07, "loss": 0.1916, "step": 1765 }, { "epoch": 0.37, "learning_rate": 2.4502699335548174e-07, "loss": 0.1882, "step": 1766 }, { "epoch": 0.37, "learning_rate": 2.4499584717607973e-07, "loss": 0.2278, "step": 1767 }, { "epoch": 0.37, "learning_rate": 2.449647009966777e-07, "loss": 0.2284, "step": 1768 }, { "epoch": 0.37, "learning_rate": 2.449335548172757e-07, "loss": 0.2668, "step": 1769 }, { "epoch": 0.37, "learning_rate": 2.4490240863787375e-07, "loss": 0.2797, "step": 1770 }, { "epoch": 0.37, "learning_rate": 2.4487126245847174e-07, "loss": 0.1407, "step": 1771 }, { "epoch": 0.37, "learning_rate": 2.4484011627906973e-07, "loss": 0.1877, "step": 1772 }, { "epoch": 0.37, "learning_rate": 2.448089700996678e-07, "loss": 0.1989, "step": 1773 }, { "epoch": 0.37, "learning_rate": 2.4477782392026576e-07, "loss": 0.2496, "step": 1774 }, { "epoch": 0.37, "learning_rate": 2.447466777408638e-07, "loss": 0.2266, "step": 1775 }, { "epoch": 0.37, "learning_rate": 2.447155315614618e-07, "loss": 0.1771, "step": 1776 }, { "epoch": 0.37, "learning_rate": 2.446843853820598e-07, "loss": 0.171, "step": 1777 }, { "epoch": 0.37, "learning_rate": 2.446532392026578e-07, "loss": 0.1545, "step": 1778 }, { "epoch": 0.37, "learning_rate": 2.446220930232558e-07, "loss": 0.2347, "step": 1779 }, { "epoch": 0.37, "learning_rate": 2.445909468438538e-07, "loss": 0.2395, "step": 1780 }, { "epoch": 0.37, "learning_rate": 2.4455980066445185e-07, "loss": 0.2446, "step": 1781 }, { "epoch": 0.37, "learning_rate": 2.4452865448504984e-07, "loss": 0.1994, "step": 1782 }, { "epoch": 0.37, "learning_rate": 2.4449750830564783e-07, "loss": 0.2219, "step": 1783 }, { "epoch": 0.37, "learning_rate": 2.444663621262458e-07, "loss": 0.1928, "step": 1784 }, { "epoch": 0.37, "learning_rate": 2.4443521594684386e-07, "loss": 0.2461, "step": 1785 }, { "epoch": 0.37, "learning_rate": 2.4440406976744185e-07, "loss": 0.1929, "step": 1786 }, { "epoch": 0.37, "learning_rate": 2.4437292358803984e-07, "loss": 0.1956, "step": 1787 }, { "epoch": 0.37, "learning_rate": 2.443417774086379e-07, "loss": 0.1904, "step": 1788 }, { "epoch": 0.37, "learning_rate": 2.4431063122923587e-07, "loss": 0.1961, "step": 1789 }, { "epoch": 0.37, "learning_rate": 2.442794850498339e-07, "loss": 0.1947, "step": 1790 }, { "epoch": 0.37, "learning_rate": 2.4424833887043185e-07, "loss": 0.1413, "step": 1791 }, { "epoch": 0.37, "learning_rate": 2.442171926910299e-07, "loss": 0.2395, "step": 1792 }, { "epoch": 0.37, "learning_rate": 2.441860465116279e-07, "loss": 0.2966, "step": 1793 }, { "epoch": 0.37, "learning_rate": 2.441549003322259e-07, "loss": 0.1797, "step": 1794 }, { "epoch": 0.37, "learning_rate": 2.441237541528239e-07, "loss": 0.1906, "step": 1795 }, { "epoch": 0.37, "learning_rate": 2.440926079734219e-07, "loss": 0.23, "step": 1796 }, { "epoch": 0.37, "learning_rate": 2.4406146179401994e-07, "loss": 0.2021, "step": 1797 }, { "epoch": 0.37, "learning_rate": 2.4403031561461793e-07, "loss": 0.1711, "step": 1798 }, { "epoch": 0.37, "learning_rate": 2.439991694352159e-07, "loss": 0.2342, "step": 1799 }, { "epoch": 0.37, "learning_rate": 2.439680232558139e-07, "loss": 0.2337, "step": 1800 }, { "epoch": 0.37, "learning_rate": 2.4393687707641195e-07, "loss": 0.1206, "step": 1801 }, { "epoch": 0.37, "learning_rate": 2.4390573089700994e-07, "loss": 0.2067, "step": 1802 }, { "epoch": 0.37, "learning_rate": 2.43874584717608e-07, "loss": 0.1706, "step": 1803 }, { "epoch": 0.37, "learning_rate": 2.43843438538206e-07, "loss": 0.2026, "step": 1804 }, { "epoch": 0.37, "learning_rate": 2.4381229235880396e-07, "loss": 0.2052, "step": 1805 }, { "epoch": 0.38, "learning_rate": 2.4378114617940195e-07, "loss": 0.2073, "step": 1806 }, { "epoch": 0.38, "learning_rate": 2.4375e-07, "loss": 0.1791, "step": 1807 }, { "epoch": 0.38, "learning_rate": 2.43718853820598e-07, "loss": 0.1974, "step": 1808 }, { "epoch": 0.38, "learning_rate": 2.4368770764119603e-07, "loss": 0.1709, "step": 1809 }, { "epoch": 0.38, "learning_rate": 2.43656561461794e-07, "loss": 0.2354, "step": 1810 }, { "epoch": 0.38, "learning_rate": 2.43625415282392e-07, "loss": 0.1802, "step": 1811 }, { "epoch": 0.38, "learning_rate": 2.4359426910299005e-07, "loss": 0.1985, "step": 1812 }, { "epoch": 0.38, "learning_rate": 2.4356312292358804e-07, "loss": 0.283, "step": 1813 }, { "epoch": 0.38, "learning_rate": 2.4353197674418603e-07, "loss": 0.2575, "step": 1814 }, { "epoch": 0.38, "learning_rate": 2.43500830564784e-07, "loss": 0.1736, "step": 1815 }, { "epoch": 0.38, "learning_rate": 2.4346968438538206e-07, "loss": 0.1621, "step": 1816 }, { "epoch": 0.38, "learning_rate": 2.4343853820598005e-07, "loss": 0.2173, "step": 1817 }, { "epoch": 0.38, "learning_rate": 2.434073920265781e-07, "loss": 0.2257, "step": 1818 }, { "epoch": 0.38, "learning_rate": 2.433762458471761e-07, "loss": 0.1206, "step": 1819 }, { "epoch": 0.38, "learning_rate": 2.4334509966777407e-07, "loss": 0.1888, "step": 1820 }, { "epoch": 0.38, "learning_rate": 2.4331395348837206e-07, "loss": 0.225, "step": 1821 }, { "epoch": 0.38, "learning_rate": 2.432828073089701e-07, "loss": 0.1486, "step": 1822 }, { "epoch": 0.38, "learning_rate": 2.432516611295681e-07, "loss": 0.1597, "step": 1823 }, { "epoch": 0.38, "learning_rate": 2.432205149501661e-07, "loss": 0.1885, "step": 1824 }, { "epoch": 0.38, "learning_rate": 2.431893687707641e-07, "loss": 0.1935, "step": 1825 }, { "epoch": 0.38, "learning_rate": 2.431582225913621e-07, "loss": 0.2024, "step": 1826 }, { "epoch": 0.38, "learning_rate": 2.4312707641196015e-07, "loss": 0.2389, "step": 1827 }, { "epoch": 0.38, "learning_rate": 2.4309593023255814e-07, "loss": 0.2533, "step": 1828 }, { "epoch": 0.38, "learning_rate": 2.4306478405315613e-07, "loss": 0.193, "step": 1829 }, { "epoch": 0.38, "learning_rate": 2.430336378737541e-07, "loss": 0.1762, "step": 1830 }, { "epoch": 0.38, "learning_rate": 2.4300249169435216e-07, "loss": 0.1941, "step": 1831 }, { "epoch": 0.38, "learning_rate": 2.4297134551495015e-07, "loss": 0.2111, "step": 1832 }, { "epoch": 0.38, "learning_rate": 2.4294019933554814e-07, "loss": 0.1869, "step": 1833 }, { "epoch": 0.38, "learning_rate": 2.429090531561462e-07, "loss": 0.2121, "step": 1834 }, { "epoch": 0.38, "learning_rate": 2.428779069767442e-07, "loss": 0.2145, "step": 1835 }, { "epoch": 0.38, "learning_rate": 2.4284676079734216e-07, "loss": 0.1901, "step": 1836 }, { "epoch": 0.38, "learning_rate": 2.428156146179402e-07, "loss": 0.1645, "step": 1837 }, { "epoch": 0.38, "learning_rate": 2.427844684385382e-07, "loss": 0.1478, "step": 1838 }, { "epoch": 0.38, "learning_rate": 2.427533222591362e-07, "loss": 0.1824, "step": 1839 }, { "epoch": 0.38, "learning_rate": 2.4272217607973423e-07, "loss": 0.187, "step": 1840 }, { "epoch": 0.38, "learning_rate": 2.426910299003322e-07, "loss": 0.2116, "step": 1841 }, { "epoch": 0.38, "learning_rate": 2.4265988372093026e-07, "loss": 0.1711, "step": 1842 }, { "epoch": 0.38, "learning_rate": 2.426287375415282e-07, "loss": 0.1749, "step": 1843 }, { "epoch": 0.38, "learning_rate": 2.4259759136212624e-07, "loss": 0.2406, "step": 1844 }, { "epoch": 0.38, "learning_rate": 2.425664451827242e-07, "loss": 0.1952, "step": 1845 }, { "epoch": 0.38, "learning_rate": 2.4253529900332227e-07, "loss": 0.1771, "step": 1846 }, { "epoch": 0.38, "learning_rate": 2.4250415282392026e-07, "loss": 0.1661, "step": 1847 }, { "epoch": 0.38, "learning_rate": 2.4247300664451825e-07, "loss": 0.2508, "step": 1848 }, { "epoch": 0.38, "learning_rate": 2.424418604651163e-07, "loss": 0.2143, "step": 1849 }, { "epoch": 0.38, "learning_rate": 2.424107142857143e-07, "loss": 0.1888, "step": 1850 }, { "epoch": 0.38, "learning_rate": 2.4237956810631227e-07, "loss": 0.2002, "step": 1851 }, { "epoch": 0.38, "learning_rate": 2.4234842192691026e-07, "loss": 0.1941, "step": 1852 }, { "epoch": 0.38, "learning_rate": 2.423172757475083e-07, "loss": 0.1804, "step": 1853 }, { "epoch": 0.38, "learning_rate": 2.422861295681063e-07, "loss": 0.1562, "step": 1854 }, { "epoch": 0.39, "learning_rate": 2.4225498338870433e-07, "loss": 0.1997, "step": 1855 }, { "epoch": 0.39, "learning_rate": 2.422238372093023e-07, "loss": 0.2222, "step": 1856 }, { "epoch": 0.39, "learning_rate": 2.421926910299003e-07, "loss": 0.1347, "step": 1857 }, { "epoch": 0.39, "learning_rate": 2.421615448504983e-07, "loss": 0.2469, "step": 1858 }, { "epoch": 0.39, "learning_rate": 2.4213039867109634e-07, "loss": 0.2212, "step": 1859 }, { "epoch": 0.39, "learning_rate": 2.4209925249169433e-07, "loss": 0.1589, "step": 1860 }, { "epoch": 0.39, "learning_rate": 2.420681063122924e-07, "loss": 0.2113, "step": 1861 }, { "epoch": 0.39, "learning_rate": 2.4203696013289036e-07, "loss": 0.17, "step": 1862 }, { "epoch": 0.39, "learning_rate": 2.4200581395348835e-07, "loss": 0.1822, "step": 1863 }, { "epoch": 0.39, "learning_rate": 2.419746677740864e-07, "loss": 0.2426, "step": 1864 }, { "epoch": 0.39, "learning_rate": 2.419435215946844e-07, "loss": 0.205, "step": 1865 }, { "epoch": 0.39, "learning_rate": 2.4191237541528237e-07, "loss": 0.2467, "step": 1866 }, { "epoch": 0.39, "learning_rate": 2.4188122923588036e-07, "loss": 0.1778, "step": 1867 }, { "epoch": 0.39, "learning_rate": 2.418500830564784e-07, "loss": 0.2048, "step": 1868 }, { "epoch": 0.39, "learning_rate": 2.418189368770764e-07, "loss": 0.2368, "step": 1869 }, { "epoch": 0.39, "learning_rate": 2.4178779069767444e-07, "loss": 0.2163, "step": 1870 }, { "epoch": 0.39, "learning_rate": 2.417566445182724e-07, "loss": 0.1527, "step": 1871 }, { "epoch": 0.39, "learning_rate": 2.417254983388704e-07, "loss": 0.176, "step": 1872 }, { "epoch": 0.39, "learning_rate": 2.416943521594684e-07, "loss": 0.195, "step": 1873 }, { "epoch": 0.39, "learning_rate": 2.4166320598006645e-07, "loss": 0.2203, "step": 1874 }, { "epoch": 0.39, "learning_rate": 2.4163205980066444e-07, "loss": 0.19, "step": 1875 }, { "epoch": 0.39, "learning_rate": 2.416009136212624e-07, "loss": 0.2407, "step": 1876 }, { "epoch": 0.39, "learning_rate": 2.4156976744186047e-07, "loss": 0.1888, "step": 1877 }, { "epoch": 0.39, "learning_rate": 2.4153862126245846e-07, "loss": 0.3025, "step": 1878 }, { "epoch": 0.39, "learning_rate": 2.415074750830565e-07, "loss": 0.2323, "step": 1879 }, { "epoch": 0.39, "learning_rate": 2.4147632890365444e-07, "loss": 0.2136, "step": 1880 }, { "epoch": 0.39, "learning_rate": 2.414451827242525e-07, "loss": 0.1355, "step": 1881 }, { "epoch": 0.39, "learning_rate": 2.4141403654485047e-07, "loss": 0.2298, "step": 1882 }, { "epoch": 0.39, "learning_rate": 2.413828903654485e-07, "loss": 0.174, "step": 1883 }, { "epoch": 0.39, "learning_rate": 2.413517441860465e-07, "loss": 0.2401, "step": 1884 }, { "epoch": 0.39, "learning_rate": 2.413205980066445e-07, "loss": 0.2982, "step": 1885 }, { "epoch": 0.39, "learning_rate": 2.4128945182724253e-07, "loss": 0.2123, "step": 1886 }, { "epoch": 0.39, "learning_rate": 2.412583056478405e-07, "loss": 0.2002, "step": 1887 }, { "epoch": 0.39, "learning_rate": 2.412271594684385e-07, "loss": 0.1601, "step": 1888 }, { "epoch": 0.39, "learning_rate": 2.4119601328903655e-07, "loss": 0.2054, "step": 1889 }, { "epoch": 0.39, "learning_rate": 2.4116486710963454e-07, "loss": 0.2109, "step": 1890 }, { "epoch": 0.39, "learning_rate": 2.4113372093023253e-07, "loss": 0.2175, "step": 1891 }, { "epoch": 0.39, "learning_rate": 2.4110257475083057e-07, "loss": 0.2376, "step": 1892 }, { "epoch": 0.39, "learning_rate": 2.4107142857142856e-07, "loss": 0.2633, "step": 1893 }, { "epoch": 0.39, "learning_rate": 2.410402823920266e-07, "loss": 0.1757, "step": 1894 }, { "epoch": 0.39, "learning_rate": 2.4100913621262454e-07, "loss": 0.1694, "step": 1895 }, { "epoch": 0.39, "learning_rate": 2.409779900332226e-07, "loss": 0.1485, "step": 1896 }, { "epoch": 0.39, "learning_rate": 2.4094684385382057e-07, "loss": 0.1568, "step": 1897 }, { "epoch": 0.39, "learning_rate": 2.409156976744186e-07, "loss": 0.2542, "step": 1898 }, { "epoch": 0.39, "learning_rate": 2.408845514950166e-07, "loss": 0.2865, "step": 1899 }, { "epoch": 0.39, "learning_rate": 2.408534053156146e-07, "loss": 0.2129, "step": 1900 }, { "epoch": 0.39, "learning_rate": 2.4082225913621264e-07, "loss": 0.1611, "step": 1901 }, { "epoch": 0.39, "learning_rate": 2.407911129568106e-07, "loss": 0.2239, "step": 1902 }, { "epoch": 0.4, "learning_rate": 2.407599667774086e-07, "loss": 0.213, "step": 1903 }, { "epoch": 0.4, "learning_rate": 2.407288205980066e-07, "loss": 0.1733, "step": 1904 }, { "epoch": 0.4, "learning_rate": 2.4069767441860465e-07, "loss": 0.2293, "step": 1905 }, { "epoch": 0.4, "learning_rate": 2.4066652823920264e-07, "loss": 0.1834, "step": 1906 }, { "epoch": 0.4, "learning_rate": 2.406353820598007e-07, "loss": 0.1547, "step": 1907 }, { "epoch": 0.4, "learning_rate": 2.4060423588039867e-07, "loss": 0.1412, "step": 1908 }, { "epoch": 0.4, "learning_rate": 2.4057308970099666e-07, "loss": 0.2047, "step": 1909 }, { "epoch": 0.4, "learning_rate": 2.4054194352159465e-07, "loss": 0.177, "step": 1910 }, { "epoch": 0.4, "learning_rate": 2.405107973421927e-07, "loss": 0.1639, "step": 1911 }, { "epoch": 0.4, "learning_rate": 2.404796511627907e-07, "loss": 0.2634, "step": 1912 }, { "epoch": 0.4, "learning_rate": 2.4044850498338867e-07, "loss": 0.2071, "step": 1913 }, { "epoch": 0.4, "learning_rate": 2.404173588039867e-07, "loss": 0.1966, "step": 1914 }, { "epoch": 0.4, "learning_rate": 2.403862126245847e-07, "loss": 0.2382, "step": 1915 }, { "epoch": 0.4, "learning_rate": 2.4035506644518274e-07, "loss": 0.1712, "step": 1916 }, { "epoch": 0.4, "learning_rate": 2.4032392026578073e-07, "loss": 0.1707, "step": 1917 }, { "epoch": 0.4, "learning_rate": 2.402927740863787e-07, "loss": 0.1442, "step": 1918 }, { "epoch": 0.4, "learning_rate": 2.402616279069767e-07, "loss": 0.2114, "step": 1919 }, { "epoch": 0.4, "learning_rate": 2.4023048172757475e-07, "loss": 0.1792, "step": 1920 }, { "epoch": 0.4, "learning_rate": 2.4019933554817274e-07, "loss": 0.1534, "step": 1921 }, { "epoch": 0.4, "learning_rate": 2.401681893687708e-07, "loss": 0.1422, "step": 1922 }, { "epoch": 0.4, "learning_rate": 2.4013704318936877e-07, "loss": 0.2078, "step": 1923 }, { "epoch": 0.4, "learning_rate": 2.4010589700996676e-07, "loss": 0.2859, "step": 1924 }, { "epoch": 0.4, "learning_rate": 2.4007475083056475e-07, "loss": 0.2379, "step": 1925 }, { "epoch": 0.4, "learning_rate": 2.400436046511628e-07, "loss": 0.1769, "step": 1926 }, { "epoch": 0.4, "learning_rate": 2.400124584717608e-07, "loss": 0.2528, "step": 1927 }, { "epoch": 0.4, "learning_rate": 2.3998131229235877e-07, "loss": 0.2466, "step": 1928 }, { "epoch": 0.4, "learning_rate": 2.399501661129568e-07, "loss": 0.2192, "step": 1929 }, { "epoch": 0.4, "learning_rate": 2.399190199335548e-07, "loss": 0.1619, "step": 1930 }, { "epoch": 0.4, "learning_rate": 2.3988787375415285e-07, "loss": 0.1987, "step": 1931 }, { "epoch": 0.4, "learning_rate": 2.398567275747508e-07, "loss": 0.2086, "step": 1932 }, { "epoch": 0.4, "learning_rate": 2.398255813953488e-07, "loss": 0.2164, "step": 1933 }, { "epoch": 0.4, "learning_rate": 2.397944352159468e-07, "loss": 0.161, "step": 1934 }, { "epoch": 0.4, "learning_rate": 2.3976328903654486e-07, "loss": 0.1954, "step": 1935 }, { "epoch": 0.4, "learning_rate": 2.3973214285714285e-07, "loss": 0.1887, "step": 1936 }, { "epoch": 0.4, "learning_rate": 2.3970099667774084e-07, "loss": 0.2466, "step": 1937 }, { "epoch": 0.4, "learning_rate": 2.396698504983389e-07, "loss": 0.1684, "step": 1938 }, { "epoch": 0.4, "learning_rate": 2.3963870431893687e-07, "loss": 0.2537, "step": 1939 }, { "epoch": 0.4, "learning_rate": 2.3960755813953486e-07, "loss": 0.2713, "step": 1940 }, { "epoch": 0.4, "learning_rate": 2.395764119601329e-07, "loss": 0.1521, "step": 1941 }, { "epoch": 0.4, "learning_rate": 2.395452657807309e-07, "loss": 0.1708, "step": 1942 }, { "epoch": 0.4, "learning_rate": 2.395141196013289e-07, "loss": 0.21, "step": 1943 }, { "epoch": 0.4, "learning_rate": 2.394829734219269e-07, "loss": 0.3064, "step": 1944 }, { "epoch": 0.4, "learning_rate": 2.394518272425249e-07, "loss": 0.2276, "step": 1945 }, { "epoch": 0.4, "learning_rate": 2.394206810631229e-07, "loss": 0.2358, "step": 1946 }, { "epoch": 0.4, "learning_rate": 2.393895348837209e-07, "loss": 0.2453, "step": 1947 }, { "epoch": 0.4, "learning_rate": 2.3935838870431893e-07, "loss": 0.1916, "step": 1948 }, { "epoch": 0.4, "learning_rate": 2.393272425249169e-07, "loss": 0.2177, "step": 1949 }, { "epoch": 0.4, "learning_rate": 2.3929609634551496e-07, "loss": 0.1944, "step": 1950 }, { "epoch": 0.41, "learning_rate": 2.3926495016611295e-07, "loss": 0.2337, "step": 1951 }, { "epoch": 0.41, "learning_rate": 2.3923380398671094e-07, "loss": 0.2222, "step": 1952 }, { "epoch": 0.41, "learning_rate": 2.39202657807309e-07, "loss": 0.2197, "step": 1953 }, { "epoch": 0.41, "learning_rate": 2.3917151162790697e-07, "loss": 0.1906, "step": 1954 }, { "epoch": 0.41, "learning_rate": 2.39140365448505e-07, "loss": 0.2808, "step": 1955 }, { "epoch": 0.41, "learning_rate": 2.3910921926910295e-07, "loss": 0.1877, "step": 1956 }, { "epoch": 0.41, "learning_rate": 2.39078073089701e-07, "loss": 0.2465, "step": 1957 }, { "epoch": 0.41, "learning_rate": 2.39046926910299e-07, "loss": 0.1965, "step": 1958 }, { "epoch": 0.41, "learning_rate": 2.39015780730897e-07, "loss": 0.1997, "step": 1959 }, { "epoch": 0.41, "learning_rate": 2.38984634551495e-07, "loss": 0.1644, "step": 1960 }, { "epoch": 0.41, "learning_rate": 2.38953488372093e-07, "loss": 0.169, "step": 1961 }, { "epoch": 0.41, "learning_rate": 2.38922342192691e-07, "loss": 0.2254, "step": 1962 }, { "epoch": 0.41, "learning_rate": 2.3889119601328903e-07, "loss": 0.224, "step": 1963 }, { "epoch": 0.41, "learning_rate": 2.38860049833887e-07, "loss": 0.18, "step": 1964 }, { "epoch": 0.41, "learning_rate": 2.38828903654485e-07, "loss": 0.2156, "step": 1965 }, { "epoch": 0.41, "learning_rate": 2.3879775747508306e-07, "loss": 0.1705, "step": 1966 }, { "epoch": 0.41, "learning_rate": 2.3876661129568105e-07, "loss": 0.2319, "step": 1967 }, { "epoch": 0.41, "learning_rate": 2.387354651162791e-07, "loss": 0.244, "step": 1968 }, { "epoch": 0.41, "learning_rate": 2.387043189368771e-07, "loss": 0.1905, "step": 1969 }, { "epoch": 0.41, "learning_rate": 2.3867317275747507e-07, "loss": 0.1982, "step": 1970 }, { "epoch": 0.41, "learning_rate": 2.3864202657807306e-07, "loss": 0.2399, "step": 1971 }, { "epoch": 0.41, "learning_rate": 2.386108803986711e-07, "loss": 0.1924, "step": 1972 }, { "epoch": 0.41, "learning_rate": 2.385797342192691e-07, "loss": 0.1914, "step": 1973 }, { "epoch": 0.41, "learning_rate": 2.385485880398671e-07, "loss": 0.2072, "step": 1974 }, { "epoch": 0.41, "learning_rate": 2.385174418604651e-07, "loss": 0.1615, "step": 1975 }, { "epoch": 0.41, "learning_rate": 2.384862956810631e-07, "loss": 0.1908, "step": 1976 }, { "epoch": 0.41, "learning_rate": 2.384551495016611e-07, "loss": 0.201, "step": 1977 }, { "epoch": 0.41, "learning_rate": 2.3842400332225914e-07, "loss": 0.1906, "step": 1978 }, { "epoch": 0.41, "learning_rate": 2.3839285714285713e-07, "loss": 0.1965, "step": 1979 }, { "epoch": 0.41, "learning_rate": 2.3836171096345512e-07, "loss": 0.2341, "step": 1980 }, { "epoch": 0.41, "learning_rate": 2.3833056478405313e-07, "loss": 0.2009, "step": 1981 }, { "epoch": 0.41, "learning_rate": 2.3829941860465115e-07, "loss": 0.18, "step": 1982 }, { "epoch": 0.41, "learning_rate": 2.3826827242524917e-07, "loss": 0.1354, "step": 1983 }, { "epoch": 0.41, "learning_rate": 2.3823712624584716e-07, "loss": 0.1954, "step": 1984 }, { "epoch": 0.41, "learning_rate": 2.3820598006644517e-07, "loss": 0.2654, "step": 1985 }, { "epoch": 0.41, "learning_rate": 2.381748338870432e-07, "loss": 0.1976, "step": 1986 }, { "epoch": 0.41, "learning_rate": 2.381436877076412e-07, "loss": 0.2476, "step": 1987 }, { "epoch": 0.41, "learning_rate": 2.3811254152823922e-07, "loss": 0.1828, "step": 1988 }, { "epoch": 0.41, "learning_rate": 2.3808139534883718e-07, "loss": 0.1406, "step": 1989 }, { "epoch": 0.41, "learning_rate": 2.380502491694352e-07, "loss": 0.1962, "step": 1990 }, { "epoch": 0.41, "learning_rate": 2.3801910299003321e-07, "loss": 0.2113, "step": 1991 }, { "epoch": 0.41, "learning_rate": 2.3798795681063123e-07, "loss": 0.2334, "step": 1992 }, { "epoch": 0.41, "learning_rate": 2.3795681063122922e-07, "loss": 0.2149, "step": 1993 }, { "epoch": 0.41, "learning_rate": 2.3792566445182723e-07, "loss": 0.208, "step": 1994 }, { "epoch": 0.41, "learning_rate": 2.3789451827242525e-07, "loss": 0.1949, "step": 1995 }, { "epoch": 0.41, "learning_rate": 2.3786337209302324e-07, "loss": 0.1768, "step": 1996 }, { "epoch": 0.41, "learning_rate": 2.3783222591362126e-07, "loss": 0.1888, "step": 1997 }, { "epoch": 0.41, "learning_rate": 2.3780107973421924e-07, "loss": 0.2506, "step": 1998 }, { "epoch": 0.42, "learning_rate": 2.3776993355481726e-07, "loss": 0.2224, "step": 1999 }, { "epoch": 0.42, "learning_rate": 2.3773878737541528e-07, "loss": 0.2497, "step": 2000 }, { "epoch": 0.42, "learning_rate": 2.377076411960133e-07, "loss": 0.1892, "step": 2001 }, { "epoch": 0.42, "learning_rate": 2.376764950166113e-07, "loss": 0.2103, "step": 2002 }, { "epoch": 0.42, "learning_rate": 2.3764534883720927e-07, "loss": 0.1597, "step": 2003 }, { "epoch": 0.42, "learning_rate": 2.3761420265780729e-07, "loss": 0.1364, "step": 2004 }, { "epoch": 0.42, "learning_rate": 2.375830564784053e-07, "loss": 0.2654, "step": 2005 }, { "epoch": 0.42, "learning_rate": 2.3755191029900332e-07, "loss": 0.1774, "step": 2006 }, { "epoch": 0.42, "learning_rate": 2.375207641196013e-07, "loss": 0.2019, "step": 2007 }, { "epoch": 0.42, "learning_rate": 2.3748961794019932e-07, "loss": 0.2244, "step": 2008 }, { "epoch": 0.42, "learning_rate": 2.3745847176079734e-07, "loss": 0.1942, "step": 2009 }, { "epoch": 0.42, "learning_rate": 2.3742732558139536e-07, "loss": 0.2164, "step": 2010 }, { "epoch": 0.42, "learning_rate": 2.3739617940199334e-07, "loss": 0.2322, "step": 2011 }, { "epoch": 0.42, "learning_rate": 2.3736503322259133e-07, "loss": 0.1894, "step": 2012 }, { "epoch": 0.42, "learning_rate": 2.3733388704318935e-07, "loss": 0.2118, "step": 2013 }, { "epoch": 0.42, "learning_rate": 2.3730274086378737e-07, "loss": 0.1699, "step": 2014 }, { "epoch": 0.42, "learning_rate": 2.3727159468438538e-07, "loss": 0.218, "step": 2015 }, { "epoch": 0.42, "learning_rate": 2.372404485049834e-07, "loss": 0.2, "step": 2016 }, { "epoch": 0.42, "learning_rate": 2.3720930232558136e-07, "loss": 0.163, "step": 2017 }, { "epoch": 0.42, "learning_rate": 2.3717815614617938e-07, "loss": 0.2042, "step": 2018 }, { "epoch": 0.42, "learning_rate": 2.371470099667774e-07, "loss": 0.2399, "step": 2019 }, { "epoch": 0.42, "learning_rate": 2.371158637873754e-07, "loss": 0.176, "step": 2020 }, { "epoch": 0.42, "learning_rate": 2.370847176079734e-07, "loss": 0.1836, "step": 2021 }, { "epoch": 0.42, "learning_rate": 2.370535714285714e-07, "loss": 0.1673, "step": 2022 }, { "epoch": 0.42, "learning_rate": 2.3702242524916943e-07, "loss": 0.1992, "step": 2023 }, { "epoch": 0.42, "learning_rate": 2.3699127906976744e-07, "loss": 0.2314, "step": 2024 }, { "epoch": 0.42, "learning_rate": 2.3696013289036546e-07, "loss": 0.1676, "step": 2025 }, { "epoch": 0.42, "learning_rate": 2.3692898671096342e-07, "loss": 0.187, "step": 2026 }, { "epoch": 0.42, "learning_rate": 2.3689784053156144e-07, "loss": 0.1587, "step": 2027 }, { "epoch": 0.42, "learning_rate": 2.3686669435215945e-07, "loss": 0.1615, "step": 2028 }, { "epoch": 0.42, "learning_rate": 2.3683554817275747e-07, "loss": 0.1557, "step": 2029 }, { "epoch": 0.42, "learning_rate": 2.3680440199335549e-07, "loss": 0.255, "step": 2030 }, { "epoch": 0.42, "learning_rate": 2.3677325581395348e-07, "loss": 0.1701, "step": 2031 }, { "epoch": 0.42, "learning_rate": 2.367421096345515e-07, "loss": 0.2269, "step": 2032 }, { "epoch": 0.42, "learning_rate": 2.3671096345514948e-07, "loss": 0.2609, "step": 2033 }, { "epoch": 0.42, "learning_rate": 2.366798172757475e-07, "loss": 0.193, "step": 2034 }, { "epoch": 0.42, "learning_rate": 2.366486710963455e-07, "loss": 0.2174, "step": 2035 }, { "epoch": 0.42, "learning_rate": 2.366175249169435e-07, "loss": 0.1841, "step": 2036 }, { "epoch": 0.42, "learning_rate": 2.3658637873754152e-07, "loss": 0.1647, "step": 2037 }, { "epoch": 0.42, "learning_rate": 2.3655523255813953e-07, "loss": 0.1979, "step": 2038 }, { "epoch": 0.42, "learning_rate": 2.3652408637873755e-07, "loss": 0.2032, "step": 2039 }, { "epoch": 0.42, "learning_rate": 2.364929401993355e-07, "loss": 0.2259, "step": 2040 }, { "epoch": 0.42, "learning_rate": 2.3646179401993353e-07, "loss": 0.1919, "step": 2041 }, { "epoch": 0.42, "learning_rate": 2.3643064784053154e-07, "loss": 0.1774, "step": 2042 }, { "epoch": 0.42, "learning_rate": 2.3639950166112956e-07, "loss": 0.221, "step": 2043 }, { "epoch": 0.42, "learning_rate": 2.3636835548172758e-07, "loss": 0.1894, "step": 2044 }, { "epoch": 0.42, "learning_rate": 2.3633720930232556e-07, "loss": 0.2109, "step": 2045 }, { "epoch": 0.42, "learning_rate": 2.3630606312292358e-07, "loss": 0.1781, "step": 2046 }, { "epoch": 0.43, "learning_rate": 2.362749169435216e-07, "loss": 0.2582, "step": 2047 }, { "epoch": 0.43, "learning_rate": 2.3624377076411959e-07, "loss": 0.1878, "step": 2048 }, { "epoch": 0.43, "learning_rate": 2.362126245847176e-07, "loss": 0.2298, "step": 2049 }, { "epoch": 0.43, "learning_rate": 2.361814784053156e-07, "loss": 0.2261, "step": 2050 }, { "epoch": 0.43, "learning_rate": 2.361503322259136e-07, "loss": 0.2014, "step": 2051 }, { "epoch": 0.43, "learning_rate": 2.3611918604651162e-07, "loss": 0.1846, "step": 2052 }, { "epoch": 0.43, "learning_rate": 2.3608803986710964e-07, "loss": 0.2551, "step": 2053 }, { "epoch": 0.43, "learning_rate": 2.360568936877076e-07, "loss": 0.2108, "step": 2054 }, { "epoch": 0.43, "learning_rate": 2.3602574750830562e-07, "loss": 0.1728, "step": 2055 }, { "epoch": 0.43, "learning_rate": 2.3599460132890363e-07, "loss": 0.2568, "step": 2056 }, { "epoch": 0.43, "learning_rate": 2.3596345514950165e-07, "loss": 0.2363, "step": 2057 }, { "epoch": 0.43, "learning_rate": 2.3593230897009966e-07, "loss": 0.178, "step": 2058 }, { "epoch": 0.43, "learning_rate": 2.3590116279069765e-07, "loss": 0.2483, "step": 2059 }, { "epoch": 0.43, "learning_rate": 2.3587001661129567e-07, "loss": 0.2057, "step": 2060 }, { "epoch": 0.43, "learning_rate": 2.3583887043189369e-07, "loss": 0.2131, "step": 2061 }, { "epoch": 0.43, "learning_rate": 2.358077242524917e-07, "loss": 0.2658, "step": 2062 }, { "epoch": 0.43, "learning_rate": 2.3577657807308972e-07, "loss": 0.159, "step": 2063 }, { "epoch": 0.43, "learning_rate": 2.3574543189368768e-07, "loss": 0.2022, "step": 2064 }, { "epoch": 0.43, "learning_rate": 2.357142857142857e-07, "loss": 0.1576, "step": 2065 }, { "epoch": 0.43, "learning_rate": 2.356831395348837e-07, "loss": 0.2515, "step": 2066 }, { "epoch": 0.43, "learning_rate": 2.3565199335548173e-07, "loss": 0.2206, "step": 2067 }, { "epoch": 0.43, "learning_rate": 2.3562084717607974e-07, "loss": 0.2629, "step": 2068 }, { "epoch": 0.43, "learning_rate": 2.3558970099667773e-07, "loss": 0.2029, "step": 2069 }, { "epoch": 0.43, "learning_rate": 2.3555855481727572e-07, "loss": 0.2366, "step": 2070 }, { "epoch": 0.43, "learning_rate": 2.3552740863787374e-07, "loss": 0.1531, "step": 2071 }, { "epoch": 0.43, "learning_rate": 2.3549626245847175e-07, "loss": 0.1752, "step": 2072 }, { "epoch": 0.43, "learning_rate": 2.3546511627906974e-07, "loss": 0.188, "step": 2073 }, { "epoch": 0.43, "learning_rate": 2.3543397009966776e-07, "loss": 0.1927, "step": 2074 }, { "epoch": 0.43, "learning_rate": 2.3540282392026577e-07, "loss": 0.2113, "step": 2075 }, { "epoch": 0.43, "learning_rate": 2.353716777408638e-07, "loss": 0.2034, "step": 2076 }, { "epoch": 0.43, "learning_rate": 2.353405315614618e-07, "loss": 0.1756, "step": 2077 }, { "epoch": 0.43, "learning_rate": 2.3530938538205977e-07, "loss": 0.1925, "step": 2078 }, { "epoch": 0.43, "learning_rate": 2.3527823920265779e-07, "loss": 0.1364, "step": 2079 }, { "epoch": 0.43, "learning_rate": 2.352470930232558e-07, "loss": 0.1684, "step": 2080 }, { "epoch": 0.43, "learning_rate": 2.3521594684385382e-07, "loss": 0.2241, "step": 2081 }, { "epoch": 0.43, "learning_rate": 2.3518480066445183e-07, "loss": 0.149, "step": 2082 }, { "epoch": 0.43, "learning_rate": 2.3515365448504982e-07, "loss": 0.1864, "step": 2083 }, { "epoch": 0.43, "learning_rate": 2.3512250830564784e-07, "loss": 0.1764, "step": 2084 }, { "epoch": 0.43, "learning_rate": 2.3509136212624583e-07, "loss": 0.2506, "step": 2085 }, { "epoch": 0.43, "learning_rate": 2.3506021594684384e-07, "loss": 0.2169, "step": 2086 }, { "epoch": 0.43, "learning_rate": 2.3502906976744183e-07, "loss": 0.1823, "step": 2087 }, { "epoch": 0.43, "learning_rate": 2.3499792358803985e-07, "loss": 0.2158, "step": 2088 }, { "epoch": 0.43, "learning_rate": 2.3496677740863786e-07, "loss": 0.2509, "step": 2089 }, { "epoch": 0.43, "learning_rate": 2.3493563122923588e-07, "loss": 0.1856, "step": 2090 }, { "epoch": 0.43, "learning_rate": 2.349044850498339e-07, "loss": 0.201, "step": 2091 }, { "epoch": 0.43, "learning_rate": 2.3487333887043186e-07, "loss": 0.1675, "step": 2092 }, { "epoch": 0.43, "learning_rate": 2.3484219269102987e-07, "loss": 0.2143, "step": 2093 }, { "epoch": 0.43, "learning_rate": 2.348110465116279e-07, "loss": 0.1589, "step": 2094 }, { "epoch": 0.44, "learning_rate": 2.347799003322259e-07, "loss": 0.2065, "step": 2095 }, { "epoch": 0.44, "learning_rate": 2.3474875415282392e-07, "loss": 0.2052, "step": 2096 }, { "epoch": 0.44, "learning_rate": 2.347176079734219e-07, "loss": 0.1676, "step": 2097 }, { "epoch": 0.44, "learning_rate": 2.3468646179401993e-07, "loss": 0.2297, "step": 2098 }, { "epoch": 0.44, "learning_rate": 2.3465531561461794e-07, "loss": 0.2502, "step": 2099 }, { "epoch": 0.44, "learning_rate": 2.3462416943521596e-07, "loss": 0.1985, "step": 2100 }, { "epoch": 0.44, "learning_rate": 2.3459302325581392e-07, "loss": 0.1935, "step": 2101 }, { "epoch": 0.44, "learning_rate": 2.3456187707641194e-07, "loss": 0.1625, "step": 2102 }, { "epoch": 0.44, "learning_rate": 2.3453073089700995e-07, "loss": 0.239, "step": 2103 }, { "epoch": 0.44, "learning_rate": 2.3449958471760797e-07, "loss": 0.2302, "step": 2104 }, { "epoch": 0.44, "learning_rate": 2.3446843853820598e-07, "loss": 0.2772, "step": 2105 }, { "epoch": 0.44, "learning_rate": 2.3443729235880397e-07, "loss": 0.1804, "step": 2106 }, { "epoch": 0.44, "learning_rate": 2.3440614617940196e-07, "loss": 0.248, "step": 2107 }, { "epoch": 0.44, "learning_rate": 2.3437499999999998e-07, "loss": 0.1951, "step": 2108 }, { "epoch": 0.44, "learning_rate": 2.34343853820598e-07, "loss": 0.2109, "step": 2109 }, { "epoch": 0.44, "learning_rate": 2.34312707641196e-07, "loss": 0.2352, "step": 2110 }, { "epoch": 0.44, "learning_rate": 2.34281561461794e-07, "loss": 0.2149, "step": 2111 }, { "epoch": 0.44, "learning_rate": 2.3425041528239202e-07, "loss": 0.2106, "step": 2112 }, { "epoch": 0.44, "learning_rate": 2.3421926910299003e-07, "loss": 0.1796, "step": 2113 }, { "epoch": 0.44, "learning_rate": 2.3418812292358805e-07, "loss": 0.2119, "step": 2114 }, { "epoch": 0.44, "learning_rate": 2.3415697674418606e-07, "loss": 0.18, "step": 2115 }, { "epoch": 0.44, "learning_rate": 2.3412583056478403e-07, "loss": 0.2229, "step": 2116 }, { "epoch": 0.44, "learning_rate": 2.3409468438538204e-07, "loss": 0.1969, "step": 2117 }, { "epoch": 0.44, "learning_rate": 2.3406353820598006e-07, "loss": 0.1924, "step": 2118 }, { "epoch": 0.44, "learning_rate": 2.3403239202657807e-07, "loss": 0.1834, "step": 2119 }, { "epoch": 0.44, "learning_rate": 2.3400124584717606e-07, "loss": 0.2039, "step": 2120 }, { "epoch": 0.44, "learning_rate": 2.3397009966777408e-07, "loss": 0.1999, "step": 2121 }, { "epoch": 0.44, "learning_rate": 2.3393895348837207e-07, "loss": 0.1756, "step": 2122 }, { "epoch": 0.44, "learning_rate": 2.3390780730897008e-07, "loss": 0.2115, "step": 2123 }, { "epoch": 0.44, "learning_rate": 2.338766611295681e-07, "loss": 0.2511, "step": 2124 }, { "epoch": 0.44, "learning_rate": 2.338455149501661e-07, "loss": 0.1845, "step": 2125 }, { "epoch": 0.44, "learning_rate": 2.338143687707641e-07, "loss": 0.1922, "step": 2126 }, { "epoch": 0.44, "learning_rate": 2.3378322259136212e-07, "loss": 0.154, "step": 2127 }, { "epoch": 0.44, "learning_rate": 2.3375207641196014e-07, "loss": 0.1807, "step": 2128 }, { "epoch": 0.44, "learning_rate": 2.3372093023255815e-07, "loss": 0.2073, "step": 2129 }, { "epoch": 0.44, "learning_rate": 2.3368978405315612e-07, "loss": 0.1879, "step": 2130 }, { "epoch": 0.44, "learning_rate": 2.3365863787375413e-07, "loss": 0.2376, "step": 2131 }, { "epoch": 0.44, "learning_rate": 2.3362749169435215e-07, "loss": 0.1475, "step": 2132 }, { "epoch": 0.44, "learning_rate": 2.3359634551495016e-07, "loss": 0.1883, "step": 2133 }, { "epoch": 0.44, "learning_rate": 2.3356519933554815e-07, "loss": 0.1724, "step": 2134 }, { "epoch": 0.44, "learning_rate": 2.3353405315614617e-07, "loss": 0.1765, "step": 2135 }, { "epoch": 0.44, "learning_rate": 2.3350290697674418e-07, "loss": 0.2315, "step": 2136 }, { "epoch": 0.44, "learning_rate": 2.334717607973422e-07, "loss": 0.1776, "step": 2137 }, { "epoch": 0.44, "learning_rate": 2.334406146179402e-07, "loss": 0.2067, "step": 2138 }, { "epoch": 0.44, "learning_rate": 2.3340946843853818e-07, "loss": 0.1619, "step": 2139 }, { "epoch": 0.44, "learning_rate": 2.333783222591362e-07, "loss": 0.2363, "step": 2140 }, { "epoch": 0.44, "learning_rate": 2.333471760797342e-07, "loss": 0.2279, "step": 2141 }, { "epoch": 0.44, "learning_rate": 2.3331602990033223e-07, "loss": 0.1982, "step": 2142 }, { "epoch": 0.44, "learning_rate": 2.3328488372093024e-07, "loss": 0.2053, "step": 2143 }, { "epoch": 0.45, "learning_rate": 2.332537375415282e-07, "loss": 0.1739, "step": 2144 }, { "epoch": 0.45, "learning_rate": 2.3322259136212622e-07, "loss": 0.1243, "step": 2145 }, { "epoch": 0.45, "learning_rate": 2.3319144518272424e-07, "loss": 0.1759, "step": 2146 }, { "epoch": 0.45, "learning_rate": 2.3316029900332225e-07, "loss": 0.2259, "step": 2147 }, { "epoch": 0.45, "learning_rate": 2.3312915282392024e-07, "loss": 0.1444, "step": 2148 }, { "epoch": 0.45, "learning_rate": 2.3309800664451826e-07, "loss": 0.1928, "step": 2149 }, { "epoch": 0.45, "learning_rate": 2.3306686046511627e-07, "loss": 0.2154, "step": 2150 }, { "epoch": 0.45, "learning_rate": 2.330357142857143e-07, "loss": 0.2028, "step": 2151 }, { "epoch": 0.45, "learning_rate": 2.330045681063123e-07, "loss": 0.1472, "step": 2152 }, { "epoch": 0.45, "learning_rate": 2.3297342192691027e-07, "loss": 0.1957, "step": 2153 }, { "epoch": 0.45, "learning_rate": 2.3294227574750828e-07, "loss": 0.1629, "step": 2154 }, { "epoch": 0.45, "learning_rate": 2.329111295681063e-07, "loss": 0.2175, "step": 2155 }, { "epoch": 0.45, "learning_rate": 2.3287998338870432e-07, "loss": 0.2012, "step": 2156 }, { "epoch": 0.45, "learning_rate": 2.3284883720930233e-07, "loss": 0.1779, "step": 2157 }, { "epoch": 0.45, "learning_rate": 2.3281769102990032e-07, "loss": 0.1482, "step": 2158 }, { "epoch": 0.45, "learning_rate": 2.327865448504983e-07, "loss": 0.1791, "step": 2159 }, { "epoch": 0.45, "learning_rate": 2.3275539867109633e-07, "loss": 0.2407, "step": 2160 }, { "epoch": 0.45, "learning_rate": 2.3272425249169434e-07, "loss": 0.1987, "step": 2161 }, { "epoch": 0.45, "learning_rate": 2.3269310631229236e-07, "loss": 0.2314, "step": 2162 }, { "epoch": 0.45, "learning_rate": 2.3266196013289035e-07, "loss": 0.1506, "step": 2163 }, { "epoch": 0.45, "learning_rate": 2.3263081395348836e-07, "loss": 0.1923, "step": 2164 }, { "epoch": 0.45, "learning_rate": 2.3259966777408638e-07, "loss": 0.2549, "step": 2165 }, { "epoch": 0.45, "learning_rate": 2.325685215946844e-07, "loss": 0.2013, "step": 2166 }, { "epoch": 0.45, "learning_rate": 2.3253737541528236e-07, "loss": 0.1614, "step": 2167 }, { "epoch": 0.45, "learning_rate": 2.3250622923588037e-07, "loss": 0.1719, "step": 2168 }, { "epoch": 0.45, "learning_rate": 2.324750830564784e-07, "loss": 0.2571, "step": 2169 }, { "epoch": 0.45, "learning_rate": 2.324439368770764e-07, "loss": 0.2062, "step": 2170 }, { "epoch": 0.45, "learning_rate": 2.3241279069767442e-07, "loss": 0.2333, "step": 2171 }, { "epoch": 0.45, "learning_rate": 2.323816445182724e-07, "loss": 0.2281, "step": 2172 }, { "epoch": 0.45, "learning_rate": 2.3235049833887043e-07, "loss": 0.1473, "step": 2173 }, { "epoch": 0.45, "learning_rate": 2.3231935215946844e-07, "loss": 0.1738, "step": 2174 }, { "epoch": 0.45, "learning_rate": 2.3228820598006643e-07, "loss": 0.24, "step": 2175 }, { "epoch": 0.45, "learning_rate": 2.3225705980066445e-07, "loss": 0.2391, "step": 2176 }, { "epoch": 0.45, "learning_rate": 2.3222591362126244e-07, "loss": 0.1854, "step": 2177 }, { "epoch": 0.45, "learning_rate": 2.3219476744186045e-07, "loss": 0.2391, "step": 2178 }, { "epoch": 0.45, "learning_rate": 2.3216362126245847e-07, "loss": 0.1964, "step": 2179 }, { "epoch": 0.45, "learning_rate": 2.3213247508305648e-07, "loss": 0.1743, "step": 2180 }, { "epoch": 0.45, "learning_rate": 2.3210132890365445e-07, "loss": 0.21, "step": 2181 }, { "epoch": 0.45, "learning_rate": 2.3207018272425246e-07, "loss": 0.2003, "step": 2182 }, { "epoch": 0.45, "learning_rate": 2.3203903654485048e-07, "loss": 0.1887, "step": 2183 }, { "epoch": 0.45, "learning_rate": 2.320078903654485e-07, "loss": 0.2428, "step": 2184 }, { "epoch": 0.45, "learning_rate": 2.319767441860465e-07, "loss": 0.1485, "step": 2185 }, { "epoch": 0.45, "learning_rate": 2.319455980066445e-07, "loss": 0.2296, "step": 2186 }, { "epoch": 0.45, "learning_rate": 2.3191445182724251e-07, "loss": 0.1838, "step": 2187 }, { "epoch": 0.45, "learning_rate": 2.3188330564784053e-07, "loss": 0.2235, "step": 2188 }, { "epoch": 0.45, "learning_rate": 2.3185215946843855e-07, "loss": 0.1883, "step": 2189 }, { "epoch": 0.45, "learning_rate": 2.3182101328903654e-07, "loss": 0.2221, "step": 2190 }, { "epoch": 0.45, "learning_rate": 2.3178986710963453e-07, "loss": 0.2316, "step": 2191 }, { "epoch": 0.46, "learning_rate": 2.3175872093023254e-07, "loss": 0.2125, "step": 2192 }, { "epoch": 0.46, "learning_rate": 2.3172757475083056e-07, "loss": 0.2583, "step": 2193 }, { "epoch": 0.46, "learning_rate": 2.3169642857142857e-07, "loss": 0.2003, "step": 2194 }, { "epoch": 0.46, "learning_rate": 2.316652823920266e-07, "loss": 0.1695, "step": 2195 }, { "epoch": 0.46, "learning_rate": 2.3163413621262455e-07, "loss": 0.2209, "step": 2196 }, { "epoch": 0.46, "learning_rate": 2.3160299003322257e-07, "loss": 0.1575, "step": 2197 }, { "epoch": 0.46, "learning_rate": 2.3157184385382058e-07, "loss": 0.1252, "step": 2198 }, { "epoch": 0.46, "learning_rate": 2.315406976744186e-07, "loss": 0.1557, "step": 2199 }, { "epoch": 0.46, "learning_rate": 2.315095514950166e-07, "loss": 0.1992, "step": 2200 }, { "epoch": 0.46, "learning_rate": 2.314784053156146e-07, "loss": 0.2535, "step": 2201 }, { "epoch": 0.46, "learning_rate": 2.3144725913621262e-07, "loss": 0.151, "step": 2202 }, { "epoch": 0.46, "learning_rate": 2.3141611295681064e-07, "loss": 0.1808, "step": 2203 }, { "epoch": 0.46, "learning_rate": 2.3138496677740865e-07, "loss": 0.204, "step": 2204 }, { "epoch": 0.46, "learning_rate": 2.3135382059800661e-07, "loss": 0.2501, "step": 2205 }, { "epoch": 0.46, "learning_rate": 2.3132267441860463e-07, "loss": 0.1964, "step": 2206 }, { "epoch": 0.46, "learning_rate": 2.3129152823920265e-07, "loss": 0.2597, "step": 2207 }, { "epoch": 0.46, "learning_rate": 2.3126038205980066e-07, "loss": 0.1991, "step": 2208 }, { "epoch": 0.46, "learning_rate": 2.3122923588039868e-07, "loss": 0.2824, "step": 2209 }, { "epoch": 0.46, "learning_rate": 2.3119808970099667e-07, "loss": 0.2179, "step": 2210 }, { "epoch": 0.46, "learning_rate": 2.3116694352159468e-07, "loss": 0.2097, "step": 2211 }, { "epoch": 0.46, "learning_rate": 2.3113579734219267e-07, "loss": 0.1531, "step": 2212 }, { "epoch": 0.46, "learning_rate": 2.311046511627907e-07, "loss": 0.1766, "step": 2213 }, { "epoch": 0.46, "learning_rate": 2.3107350498338868e-07, "loss": 0.1957, "step": 2214 }, { "epoch": 0.46, "learning_rate": 2.310423588039867e-07, "loss": 0.1953, "step": 2215 }, { "epoch": 0.46, "learning_rate": 2.310112126245847e-07, "loss": 0.1926, "step": 2216 }, { "epoch": 0.46, "learning_rate": 2.3098006644518272e-07, "loss": 0.1405, "step": 2217 }, { "epoch": 0.46, "learning_rate": 2.3094892026578074e-07, "loss": 0.2294, "step": 2218 }, { "epoch": 0.46, "learning_rate": 2.309177740863787e-07, "loss": 0.2054, "step": 2219 }, { "epoch": 0.46, "learning_rate": 2.3088662790697672e-07, "loss": 0.1998, "step": 2220 }, { "epoch": 0.46, "learning_rate": 2.3085548172757474e-07, "loss": 0.2052, "step": 2221 }, { "epoch": 0.46, "learning_rate": 2.3082433554817275e-07, "loss": 0.2067, "step": 2222 }, { "epoch": 0.46, "learning_rate": 2.3079318936877077e-07, "loss": 0.1551, "step": 2223 }, { "epoch": 0.46, "learning_rate": 2.3076204318936876e-07, "loss": 0.1638, "step": 2224 }, { "epoch": 0.46, "learning_rate": 2.3073089700996677e-07, "loss": 0.233, "step": 2225 }, { "epoch": 0.46, "learning_rate": 2.306997508305648e-07, "loss": 0.2507, "step": 2226 }, { "epoch": 0.46, "learning_rate": 2.3066860465116278e-07, "loss": 0.2469, "step": 2227 }, { "epoch": 0.46, "learning_rate": 2.3063745847176077e-07, "loss": 0.255, "step": 2228 }, { "epoch": 0.46, "learning_rate": 2.3060631229235878e-07, "loss": 0.1116, "step": 2229 }, { "epoch": 0.46, "learning_rate": 2.305751661129568e-07, "loss": 0.2301, "step": 2230 }, { "epoch": 0.46, "learning_rate": 2.3054401993355481e-07, "loss": 0.1812, "step": 2231 }, { "epoch": 0.46, "learning_rate": 2.3051287375415283e-07, "loss": 0.2338, "step": 2232 }, { "epoch": 0.46, "learning_rate": 2.304817275747508e-07, "loss": 0.1932, "step": 2233 }, { "epoch": 0.46, "learning_rate": 2.304505813953488e-07, "loss": 0.1557, "step": 2234 }, { "epoch": 0.46, "learning_rate": 2.3041943521594682e-07, "loss": 0.2287, "step": 2235 }, { "epoch": 0.46, "learning_rate": 2.3038828903654484e-07, "loss": 0.1979, "step": 2236 }, { "epoch": 0.46, "learning_rate": 2.3035714285714286e-07, "loss": 0.1762, "step": 2237 }, { "epoch": 0.46, "learning_rate": 2.3032599667774085e-07, "loss": 0.2068, "step": 2238 }, { "epoch": 0.46, "learning_rate": 2.3029485049833886e-07, "loss": 0.1832, "step": 2239 }, { "epoch": 0.47, "learning_rate": 2.3026370431893688e-07, "loss": 0.1688, "step": 2240 }, { "epoch": 0.47, "learning_rate": 2.302325581395349e-07, "loss": 0.1763, "step": 2241 }, { "epoch": 0.47, "learning_rate": 2.302014119601329e-07, "loss": 0.2487, "step": 2242 }, { "epoch": 0.47, "learning_rate": 2.3017026578073087e-07, "loss": 0.2394, "step": 2243 }, { "epoch": 0.47, "learning_rate": 2.301391196013289e-07, "loss": 0.1829, "step": 2244 }, { "epoch": 0.47, "learning_rate": 2.301079734219269e-07, "loss": 0.1954, "step": 2245 }, { "epoch": 0.47, "learning_rate": 2.3007682724252492e-07, "loss": 0.2163, "step": 2246 }, { "epoch": 0.47, "learning_rate": 2.300456810631229e-07, "loss": 0.1998, "step": 2247 }, { "epoch": 0.47, "learning_rate": 2.3001453488372092e-07, "loss": 0.1713, "step": 2248 }, { "epoch": 0.47, "learning_rate": 2.2998338870431891e-07, "loss": 0.2292, "step": 2249 }, { "epoch": 0.47, "learning_rate": 2.2995224252491693e-07, "loss": 0.2592, "step": 2250 }, { "epoch": 0.47, "learning_rate": 2.2992109634551495e-07, "loss": 0.2083, "step": 2251 }, { "epoch": 0.47, "learning_rate": 2.2988995016611293e-07, "loss": 0.1718, "step": 2252 }, { "epoch": 0.47, "learning_rate": 2.2985880398671095e-07, "loss": 0.2182, "step": 2253 }, { "epoch": 0.47, "learning_rate": 2.2982765780730897e-07, "loss": 0.1748, "step": 2254 }, { "epoch": 0.47, "learning_rate": 2.2979651162790698e-07, "loss": 0.2283, "step": 2255 }, { "epoch": 0.47, "learning_rate": 2.29765365448505e-07, "loss": 0.2265, "step": 2256 }, { "epoch": 0.47, "learning_rate": 2.2973421926910296e-07, "loss": 0.2006, "step": 2257 }, { "epoch": 0.47, "learning_rate": 2.2970307308970098e-07, "loss": 0.1727, "step": 2258 }, { "epoch": 0.47, "learning_rate": 2.29671926910299e-07, "loss": 0.2255, "step": 2259 }, { "epoch": 0.47, "learning_rate": 2.29640780730897e-07, "loss": 0.2347, "step": 2260 }, { "epoch": 0.47, "learning_rate": 2.29609634551495e-07, "loss": 0.1876, "step": 2261 }, { "epoch": 0.47, "learning_rate": 2.2957848837209301e-07, "loss": 0.2368, "step": 2262 }, { "epoch": 0.47, "learning_rate": 2.2954734219269103e-07, "loss": 0.2093, "step": 2263 }, { "epoch": 0.47, "learning_rate": 2.2951619601328902e-07, "loss": 0.1757, "step": 2264 }, { "epoch": 0.47, "learning_rate": 2.2948504983388703e-07, "loss": 0.221, "step": 2265 }, { "epoch": 0.47, "learning_rate": 2.2945390365448502e-07, "loss": 0.1778, "step": 2266 }, { "epoch": 0.47, "learning_rate": 2.2942275747508304e-07, "loss": 0.2319, "step": 2267 }, { "epoch": 0.47, "learning_rate": 2.2939161129568106e-07, "loss": 0.1681, "step": 2268 }, { "epoch": 0.47, "learning_rate": 2.2936046511627907e-07, "loss": 0.1944, "step": 2269 }, { "epoch": 0.47, "learning_rate": 2.293293189368771e-07, "loss": 0.2615, "step": 2270 }, { "epoch": 0.47, "learning_rate": 2.2929817275747505e-07, "loss": 0.1956, "step": 2271 }, { "epoch": 0.47, "learning_rate": 2.2926702657807307e-07, "loss": 0.2532, "step": 2272 }, { "epoch": 0.47, "learning_rate": 2.2923588039867108e-07, "loss": 0.2033, "step": 2273 }, { "epoch": 0.47, "learning_rate": 2.292047342192691e-07, "loss": 0.1949, "step": 2274 }, { "epoch": 0.47, "learning_rate": 2.291735880398671e-07, "loss": 0.1782, "step": 2275 }, { "epoch": 0.47, "learning_rate": 2.291424418604651e-07, "loss": 0.1811, "step": 2276 }, { "epoch": 0.47, "learning_rate": 2.2911129568106312e-07, "loss": 0.2463, "step": 2277 }, { "epoch": 0.47, "learning_rate": 2.2908014950166113e-07, "loss": 0.2134, "step": 2278 }, { "epoch": 0.47, "learning_rate": 2.2904900332225915e-07, "loss": 0.1354, "step": 2279 }, { "epoch": 0.47, "learning_rate": 2.290178571428571e-07, "loss": 0.179, "step": 2280 }, { "epoch": 0.47, "learning_rate": 2.2898671096345513e-07, "loss": 0.1775, "step": 2281 }, { "epoch": 0.47, "learning_rate": 2.2895556478405314e-07, "loss": 0.2439, "step": 2282 }, { "epoch": 0.47, "learning_rate": 2.2892441860465116e-07, "loss": 0.201, "step": 2283 }, { "epoch": 0.47, "learning_rate": 2.2889327242524918e-07, "loss": 0.1847, "step": 2284 }, { "epoch": 0.47, "learning_rate": 2.2886212624584717e-07, "loss": 0.2245, "step": 2285 }, { "epoch": 0.47, "learning_rate": 2.2883098006644516e-07, "loss": 0.1882, "step": 2286 }, { "epoch": 0.47, "learning_rate": 2.2879983388704317e-07, "loss": 0.1802, "step": 2287 }, { "epoch": 0.48, "learning_rate": 2.287686877076412e-07, "loss": 0.2459, "step": 2288 }, { "epoch": 0.48, "learning_rate": 2.287375415282392e-07, "loss": 0.1512, "step": 2289 }, { "epoch": 0.48, "learning_rate": 2.287063953488372e-07, "loss": 0.1791, "step": 2290 }, { "epoch": 0.48, "learning_rate": 2.286752491694352e-07, "loss": 0.1495, "step": 2291 }, { "epoch": 0.48, "learning_rate": 2.2864410299003322e-07, "loss": 0.2091, "step": 2292 }, { "epoch": 0.48, "learning_rate": 2.2861295681063124e-07, "loss": 0.1266, "step": 2293 }, { "epoch": 0.48, "learning_rate": 2.285818106312292e-07, "loss": 0.2603, "step": 2294 }, { "epoch": 0.48, "learning_rate": 2.2855066445182722e-07, "loss": 0.2238, "step": 2295 }, { "epoch": 0.48, "learning_rate": 2.2851951827242523e-07, "loss": 0.2032, "step": 2296 }, { "epoch": 0.48, "learning_rate": 2.2848837209302325e-07, "loss": 0.1631, "step": 2297 }, { "epoch": 0.48, "learning_rate": 2.2845722591362127e-07, "loss": 0.2356, "step": 2298 }, { "epoch": 0.48, "learning_rate": 2.2842607973421925e-07, "loss": 0.1895, "step": 2299 }, { "epoch": 0.48, "learning_rate": 2.2839493355481727e-07, "loss": 0.1992, "step": 2300 }, { "epoch": 0.48, "learning_rate": 2.2836378737541526e-07, "loss": 0.2307, "step": 2301 }, { "epoch": 0.48, "learning_rate": 2.2833264119601328e-07, "loss": 0.2449, "step": 2302 }, { "epoch": 0.48, "learning_rate": 2.283014950166113e-07, "loss": 0.1822, "step": 2303 }, { "epoch": 0.48, "learning_rate": 2.2827034883720928e-07, "loss": 0.2141, "step": 2304 }, { "epoch": 0.48, "learning_rate": 2.282392026578073e-07, "loss": 0.1692, "step": 2305 }, { "epoch": 0.48, "learning_rate": 2.282080564784053e-07, "loss": 0.1656, "step": 2306 }, { "epoch": 0.48, "learning_rate": 2.2817691029900333e-07, "loss": 0.2312, "step": 2307 }, { "epoch": 0.48, "learning_rate": 2.281457641196013e-07, "loss": 0.1793, "step": 2308 }, { "epoch": 0.48, "learning_rate": 2.281146179401993e-07, "loss": 0.1831, "step": 2309 }, { "epoch": 0.48, "learning_rate": 2.2808347176079732e-07, "loss": 0.1596, "step": 2310 }, { "epoch": 0.48, "learning_rate": 2.2805232558139534e-07, "loss": 0.2244, "step": 2311 }, { "epoch": 0.48, "learning_rate": 2.2802117940199335e-07, "loss": 0.1921, "step": 2312 }, { "epoch": 0.48, "learning_rate": 2.2799003322259134e-07, "loss": 0.1812, "step": 2313 }, { "epoch": 0.48, "learning_rate": 2.2795888704318936e-07, "loss": 0.18, "step": 2314 }, { "epoch": 0.48, "learning_rate": 2.2792774086378738e-07, "loss": 0.1819, "step": 2315 }, { "epoch": 0.48, "learning_rate": 2.278965946843854e-07, "loss": 0.1794, "step": 2316 }, { "epoch": 0.48, "learning_rate": 2.2786544850498338e-07, "loss": 0.2173, "step": 2317 }, { "epoch": 0.48, "learning_rate": 2.2783430232558137e-07, "loss": 0.2082, "step": 2318 }, { "epoch": 0.48, "learning_rate": 2.2780315614617939e-07, "loss": 0.1539, "step": 2319 }, { "epoch": 0.48, "learning_rate": 2.277720099667774e-07, "loss": 0.189, "step": 2320 }, { "epoch": 0.48, "learning_rate": 2.2774086378737542e-07, "loss": 0.2373, "step": 2321 }, { "epoch": 0.48, "learning_rate": 2.277097176079734e-07, "loss": 0.2035, "step": 2322 }, { "epoch": 0.48, "learning_rate": 2.276785714285714e-07, "loss": 0.1991, "step": 2323 }, { "epoch": 0.48, "learning_rate": 2.276474252491694e-07, "loss": 0.2089, "step": 2324 }, { "epoch": 0.48, "learning_rate": 2.2761627906976743e-07, "loss": 0.2121, "step": 2325 }, { "epoch": 0.48, "learning_rate": 2.2758513289036544e-07, "loss": 0.1954, "step": 2326 }, { "epoch": 0.48, "learning_rate": 2.2755398671096343e-07, "loss": 0.1722, "step": 2327 }, { "epoch": 0.48, "learning_rate": 2.2752284053156145e-07, "loss": 0.1589, "step": 2328 }, { "epoch": 0.48, "learning_rate": 2.2749169435215947e-07, "loss": 0.1926, "step": 2329 }, { "epoch": 0.48, "learning_rate": 2.2746054817275748e-07, "loss": 0.2301, "step": 2330 }, { "epoch": 0.48, "learning_rate": 2.274294019933555e-07, "loss": 0.1897, "step": 2331 }, { "epoch": 0.48, "learning_rate": 2.2739825581395346e-07, "loss": 0.185, "step": 2332 }, { "epoch": 0.48, "learning_rate": 2.2736710963455148e-07, "loss": 0.2292, "step": 2333 }, { "epoch": 0.48, "learning_rate": 2.273359634551495e-07, "loss": 0.2321, "step": 2334 }, { "epoch": 0.48, "learning_rate": 2.273048172757475e-07, "loss": 0.2442, "step": 2335 }, { "epoch": 0.49, "learning_rate": 2.2727367109634552e-07, "loss": 0.1881, "step": 2336 }, { "epoch": 0.49, "learning_rate": 2.272425249169435e-07, "loss": 0.1434, "step": 2337 }, { "epoch": 0.49, "learning_rate": 2.272113787375415e-07, "loss": 0.1729, "step": 2338 }, { "epoch": 0.49, "learning_rate": 2.2718023255813952e-07, "loss": 0.1711, "step": 2339 }, { "epoch": 0.49, "learning_rate": 2.2714908637873753e-07, "loss": 0.1897, "step": 2340 }, { "epoch": 0.49, "learning_rate": 2.2711794019933552e-07, "loss": 0.1985, "step": 2341 }, { "epoch": 0.49, "learning_rate": 2.2708679401993354e-07, "loss": 0.1946, "step": 2342 }, { "epoch": 0.49, "learning_rate": 2.2705564784053155e-07, "loss": 0.1736, "step": 2343 }, { "epoch": 0.49, "learning_rate": 2.2702450166112957e-07, "loss": 0.2054, "step": 2344 }, { "epoch": 0.49, "learning_rate": 2.2699335548172759e-07, "loss": 0.2058, "step": 2345 }, { "epoch": 0.49, "learning_rate": 2.2696220930232555e-07, "loss": 0.1394, "step": 2346 }, { "epoch": 0.49, "learning_rate": 2.2693106312292356e-07, "loss": 0.2059, "step": 2347 }, { "epoch": 0.49, "learning_rate": 2.2689991694352158e-07, "loss": 0.2028, "step": 2348 }, { "epoch": 0.49, "learning_rate": 2.268687707641196e-07, "loss": 0.2002, "step": 2349 }, { "epoch": 0.49, "learning_rate": 2.268376245847176e-07, "loss": 0.2319, "step": 2350 }, { "epoch": 0.49, "learning_rate": 2.268064784053156e-07, "loss": 0.1921, "step": 2351 }, { "epoch": 0.49, "learning_rate": 2.2677533222591362e-07, "loss": 0.1721, "step": 2352 }, { "epoch": 0.49, "learning_rate": 2.2674418604651163e-07, "loss": 0.1854, "step": 2353 }, { "epoch": 0.49, "learning_rate": 2.2671303986710962e-07, "loss": 0.1878, "step": 2354 }, { "epoch": 0.49, "learning_rate": 2.266818936877076e-07, "loss": 0.2207, "step": 2355 }, { "epoch": 0.49, "learning_rate": 2.2665074750830563e-07, "loss": 0.1175, "step": 2356 }, { "epoch": 0.49, "learning_rate": 2.2661960132890364e-07, "loss": 0.1357, "step": 2357 }, { "epoch": 0.49, "learning_rate": 2.2658845514950166e-07, "loss": 0.2581, "step": 2358 }, { "epoch": 0.49, "learning_rate": 2.2655730897009968e-07, "loss": 0.2218, "step": 2359 }, { "epoch": 0.49, "learning_rate": 2.2652616279069764e-07, "loss": 0.1741, "step": 2360 }, { "epoch": 0.49, "learning_rate": 2.2649501661129565e-07, "loss": 0.1328, "step": 2361 }, { "epoch": 0.49, "learning_rate": 2.2646387043189367e-07, "loss": 0.2096, "step": 2362 }, { "epoch": 0.49, "learning_rate": 2.2643272425249169e-07, "loss": 0.2279, "step": 2363 }, { "epoch": 0.49, "learning_rate": 2.264015780730897e-07, "loss": 0.2521, "step": 2364 }, { "epoch": 0.49, "learning_rate": 2.263704318936877e-07, "loss": 0.1975, "step": 2365 }, { "epoch": 0.49, "learning_rate": 2.263392857142857e-07, "loss": 0.1684, "step": 2366 }, { "epoch": 0.49, "learning_rate": 2.2630813953488372e-07, "loss": 0.132, "step": 2367 }, { "epoch": 0.49, "learning_rate": 2.2627699335548174e-07, "loss": 0.2645, "step": 2368 }, { "epoch": 0.49, "learning_rate": 2.2624584717607973e-07, "loss": 0.295, "step": 2369 }, { "epoch": 0.49, "learning_rate": 2.2621470099667772e-07, "loss": 0.1948, "step": 2370 }, { "epoch": 0.49, "learning_rate": 2.2618355481727573e-07, "loss": 0.2266, "step": 2371 }, { "epoch": 0.49, "learning_rate": 2.2615240863787375e-07, "loss": 0.1483, "step": 2372 }, { "epoch": 0.49, "learning_rate": 2.2612126245847176e-07, "loss": 0.1809, "step": 2373 }, { "epoch": 0.49, "learning_rate": 2.2609011627906975e-07, "loss": 0.1755, "step": 2374 }, { "epoch": 0.49, "learning_rate": 2.2605897009966777e-07, "loss": 0.1904, "step": 2375 }, { "epoch": 0.49, "learning_rate": 2.2602782392026576e-07, "loss": 0.2443, "step": 2376 }, { "epoch": 0.49, "learning_rate": 2.2599667774086377e-07, "loss": 0.1893, "step": 2377 }, { "epoch": 0.49, "learning_rate": 2.259655315614618e-07, "loss": 0.2286, "step": 2378 }, { "epoch": 0.49, "learning_rate": 2.2593438538205978e-07, "loss": 0.2578, "step": 2379 }, { "epoch": 0.49, "learning_rate": 2.259032392026578e-07, "loss": 0.1819, "step": 2380 }, { "epoch": 0.49, "learning_rate": 2.258720930232558e-07, "loss": 0.1981, "step": 2381 }, { "epoch": 0.49, "learning_rate": 2.2584094684385383e-07, "loss": 0.1437, "step": 2382 }, { "epoch": 0.49, "learning_rate": 2.2580980066445184e-07, "loss": 0.2168, "step": 2383 }, { "epoch": 0.5, "learning_rate": 2.257786544850498e-07, "loss": 0.2151, "step": 2384 }, { "epoch": 0.5, "learning_rate": 2.2574750830564782e-07, "loss": 0.189, "step": 2385 }, { "epoch": 0.5, "learning_rate": 2.2571636212624584e-07, "loss": 0.2104, "step": 2386 }, { "epoch": 0.5, "learning_rate": 2.2568521594684385e-07, "loss": 0.1985, "step": 2387 }, { "epoch": 0.5, "learning_rate": 2.2565406976744184e-07, "loss": 0.2216, "step": 2388 }, { "epoch": 0.5, "learning_rate": 2.2562292358803986e-07, "loss": 0.2009, "step": 2389 }, { "epoch": 0.5, "learning_rate": 2.2559177740863787e-07, "loss": 0.2307, "step": 2390 }, { "epoch": 0.5, "learning_rate": 2.2556063122923586e-07, "loss": 0.1537, "step": 2391 }, { "epoch": 0.5, "learning_rate": 2.2552948504983388e-07, "loss": 0.1723, "step": 2392 }, { "epoch": 0.5, "learning_rate": 2.2549833887043187e-07, "loss": 0.1998, "step": 2393 }, { "epoch": 0.5, "learning_rate": 2.2546719269102988e-07, "loss": 0.161, "step": 2394 }, { "epoch": 0.5, "learning_rate": 2.254360465116279e-07, "loss": 0.1656, "step": 2395 }, { "epoch": 0.5, "learning_rate": 2.2540490033222592e-07, "loss": 0.2001, "step": 2396 }, { "epoch": 0.5, "learning_rate": 2.2537375415282393e-07, "loss": 0.1312, "step": 2397 }, { "epoch": 0.5, "learning_rate": 2.253426079734219e-07, "loss": 0.2173, "step": 2398 }, { "epoch": 0.5, "learning_rate": 2.253114617940199e-07, "loss": 0.2209, "step": 2399 }, { "epoch": 0.5, "learning_rate": 2.2528031561461793e-07, "loss": 0.1778, "step": 2400 }, { "epoch": 0.5, "learning_rate": 2.2524916943521594e-07, "loss": 0.2389, "step": 2401 }, { "epoch": 0.5, "learning_rate": 2.2521802325581393e-07, "loss": 0.1896, "step": 2402 }, { "epoch": 0.5, "learning_rate": 2.2518687707641195e-07, "loss": 0.161, "step": 2403 }, { "epoch": 0.5, "learning_rate": 2.2515573089700996e-07, "loss": 0.2151, "step": 2404 }, { "epoch": 0.5, "learning_rate": 2.2512458471760798e-07, "loss": 0.1361, "step": 2405 }, { "epoch": 0.5, "learning_rate": 2.2509343853820597e-07, "loss": 0.2356, "step": 2406 }, { "epoch": 0.5, "learning_rate": 2.2506229235880396e-07, "loss": 0.2092, "step": 2407 }, { "epoch": 0.5, "learning_rate": 2.2503114617940197e-07, "loss": 0.2075, "step": 2408 }, { "epoch": 0.5, "learning_rate": 2.25e-07, "loss": 0.1634, "step": 2409 }, { "epoch": 0.5, "learning_rate": 2.24968853820598e-07, "loss": 0.2066, "step": 2410 }, { "epoch": 0.5, "learning_rate": 2.2493770764119602e-07, "loss": 0.1805, "step": 2411 }, { "epoch": 0.5, "learning_rate": 2.24906561461794e-07, "loss": 0.1895, "step": 2412 }, { "epoch": 0.5, "learning_rate": 2.24875415282392e-07, "loss": 0.1654, "step": 2413 }, { "epoch": 0.5, "learning_rate": 2.2484426910299002e-07, "loss": 0.1684, "step": 2414 }, { "epoch": 0.5, "learning_rate": 2.2481312292358803e-07, "loss": 0.2139, "step": 2415 }, { "epoch": 0.5, "learning_rate": 2.2478197674418605e-07, "loss": 0.2289, "step": 2416 }, { "epoch": 0.5, "learning_rate": 2.2475083056478404e-07, "loss": 0.1998, "step": 2417 }, { "epoch": 0.5, "learning_rate": 2.2471968438538205e-07, "loss": 0.1873, "step": 2418 }, { "epoch": 0.5, "learning_rate": 2.2468853820598007e-07, "loss": 0.1419, "step": 2419 }, { "epoch": 0.5, "learning_rate": 2.2465739202657808e-07, "loss": 0.1544, "step": 2420 }, { "epoch": 0.5, "learning_rate": 2.2462624584717605e-07, "loss": 0.2269, "step": 2421 }, { "epoch": 0.5, "learning_rate": 2.2459509966777406e-07, "loss": 0.2348, "step": 2422 }, { "epoch": 0.5, "learning_rate": 2.2456395348837208e-07, "loss": 0.1904, "step": 2423 }, { "epoch": 0.5, "learning_rate": 2.245328073089701e-07, "loss": 0.2308, "step": 2424 }, { "epoch": 0.5, "learning_rate": 2.245016611295681e-07, "loss": 0.3021, "step": 2425 }, { "epoch": 0.5, "learning_rate": 2.244705149501661e-07, "loss": 0.2411, "step": 2426 }, { "epoch": 0.5, "learning_rate": 2.2443936877076412e-07, "loss": 0.2044, "step": 2427 }, { "epoch": 0.5, "learning_rate": 2.244082225913621e-07, "loss": 0.2496, "step": 2428 }, { "epoch": 0.5, "learning_rate": 2.2437707641196012e-07, "loss": 0.2552, "step": 2429 }, { "epoch": 0.5, "learning_rate": 2.2434593023255814e-07, "loss": 0.1984, "step": 2430 }, { "epoch": 0.5, "learning_rate": 2.2431478405315613e-07, "loss": 0.2297, "step": 2431 }, { "epoch": 0.5, "learning_rate": 2.2428363787375414e-07, "loss": 0.1803, "step": 2432 }, { "epoch": 0.51, "learning_rate": 2.2425249169435216e-07, "loss": 0.2124, "step": 2433 }, { "epoch": 0.51, "learning_rate": 2.2422134551495017e-07, "loss": 0.1972, "step": 2434 }, { "epoch": 0.51, "learning_rate": 2.2419019933554814e-07, "loss": 0.186, "step": 2435 }, { "epoch": 0.51, "learning_rate": 2.2415905315614615e-07, "loss": 0.1708, "step": 2436 }, { "epoch": 0.51, "learning_rate": 2.2412790697674417e-07, "loss": 0.2498, "step": 2437 }, { "epoch": 0.51, "learning_rate": 2.2409676079734218e-07, "loss": 0.2295, "step": 2438 }, { "epoch": 0.51, "learning_rate": 2.240656146179402e-07, "loss": 0.1851, "step": 2439 }, { "epoch": 0.51, "learning_rate": 2.240344684385382e-07, "loss": 0.1619, "step": 2440 }, { "epoch": 0.51, "learning_rate": 2.240033222591362e-07, "loss": 0.1959, "step": 2441 }, { "epoch": 0.51, "learning_rate": 2.2397217607973422e-07, "loss": 0.2363, "step": 2442 }, { "epoch": 0.51, "learning_rate": 2.239410299003322e-07, "loss": 0.1491, "step": 2443 }, { "epoch": 0.51, "learning_rate": 2.2390988372093023e-07, "loss": 0.18, "step": 2444 }, { "epoch": 0.51, "learning_rate": 2.2387873754152822e-07, "loss": 0.2088, "step": 2445 }, { "epoch": 0.51, "learning_rate": 2.2384759136212623e-07, "loss": 0.1361, "step": 2446 }, { "epoch": 0.51, "learning_rate": 2.2381644518272425e-07, "loss": 0.1624, "step": 2447 }, { "epoch": 0.51, "learning_rate": 2.2378529900332226e-07, "loss": 0.1719, "step": 2448 }, { "epoch": 0.51, "learning_rate": 2.2375415282392025e-07, "loss": 0.1685, "step": 2449 }, { "epoch": 0.51, "learning_rate": 2.2372300664451824e-07, "loss": 0.198, "step": 2450 }, { "epoch": 0.51, "learning_rate": 2.2369186046511626e-07, "loss": 0.1751, "step": 2451 }, { "epoch": 0.51, "learning_rate": 2.2366071428571427e-07, "loss": 0.1888, "step": 2452 }, { "epoch": 0.51, "learning_rate": 2.236295681063123e-07, "loss": 0.1898, "step": 2453 }, { "epoch": 0.51, "learning_rate": 2.2359842192691028e-07, "loss": 0.1979, "step": 2454 }, { "epoch": 0.51, "learning_rate": 2.235672757475083e-07, "loss": 0.3167, "step": 2455 }, { "epoch": 0.51, "learning_rate": 2.235361295681063e-07, "loss": 0.16, "step": 2456 }, { "epoch": 0.51, "learning_rate": 2.2350498338870433e-07, "loss": 0.1758, "step": 2457 }, { "epoch": 0.51, "learning_rate": 2.2347383720930234e-07, "loss": 0.2226, "step": 2458 }, { "epoch": 0.51, "learning_rate": 2.234426910299003e-07, "loss": 0.2193, "step": 2459 }, { "epoch": 0.51, "learning_rate": 2.2341154485049832e-07, "loss": 0.1133, "step": 2460 }, { "epoch": 0.51, "learning_rate": 2.2338039867109634e-07, "loss": 0.2073, "step": 2461 }, { "epoch": 0.51, "learning_rate": 2.2334925249169435e-07, "loss": 0.1747, "step": 2462 }, { "epoch": 0.51, "learning_rate": 2.2331810631229237e-07, "loss": 0.2108, "step": 2463 }, { "epoch": 0.51, "learning_rate": 2.2328696013289036e-07, "loss": 0.198, "step": 2464 }, { "epoch": 0.51, "learning_rate": 2.2325581395348835e-07, "loss": 0.2329, "step": 2465 }, { "epoch": 0.51, "learning_rate": 2.2322466777408636e-07, "loss": 0.1661, "step": 2466 }, { "epoch": 0.51, "learning_rate": 2.2319352159468438e-07, "loss": 0.2002, "step": 2467 }, { "epoch": 0.51, "learning_rate": 2.2316237541528237e-07, "loss": 0.186, "step": 2468 }, { "epoch": 0.51, "learning_rate": 2.2313122923588038e-07, "loss": 0.1612, "step": 2469 }, { "epoch": 0.51, "learning_rate": 2.231000830564784e-07, "loss": 0.1807, "step": 2470 }, { "epoch": 0.51, "learning_rate": 2.2306893687707642e-07, "loss": 0.2011, "step": 2471 }, { "epoch": 0.51, "learning_rate": 2.2303779069767443e-07, "loss": 0.1618, "step": 2472 }, { "epoch": 0.51, "learning_rate": 2.230066445182724e-07, "loss": 0.1875, "step": 2473 }, { "epoch": 0.51, "learning_rate": 2.229754983388704e-07, "loss": 0.2266, "step": 2474 }, { "epoch": 0.51, "learning_rate": 2.2294435215946843e-07, "loss": 0.19, "step": 2475 }, { "epoch": 0.51, "learning_rate": 2.2291320598006644e-07, "loss": 0.1985, "step": 2476 }, { "epoch": 0.51, "learning_rate": 2.2288205980066446e-07, "loss": 0.1861, "step": 2477 }, { "epoch": 0.51, "learning_rate": 2.2285091362126245e-07, "loss": 0.1544, "step": 2478 }, { "epoch": 0.51, "learning_rate": 2.2281976744186046e-07, "loss": 0.1887, "step": 2479 }, { "epoch": 0.51, "learning_rate": 2.2278862126245848e-07, "loss": 0.2357, "step": 2480 }, { "epoch": 0.52, "learning_rate": 2.2275747508305647e-07, "loss": 0.1813, "step": 2481 }, { "epoch": 0.52, "learning_rate": 2.2272632890365446e-07, "loss": 0.179, "step": 2482 }, { "epoch": 0.52, "learning_rate": 2.2269518272425247e-07, "loss": 0.2649, "step": 2483 }, { "epoch": 0.52, "learning_rate": 2.226640365448505e-07, "loss": 0.1614, "step": 2484 }, { "epoch": 0.52, "learning_rate": 2.226328903654485e-07, "loss": 0.2005, "step": 2485 }, { "epoch": 0.52, "learning_rate": 2.2260174418604652e-07, "loss": 0.1731, "step": 2486 }, { "epoch": 0.52, "learning_rate": 2.2257059800664448e-07, "loss": 0.2047, "step": 2487 }, { "epoch": 0.52, "learning_rate": 2.225394518272425e-07, "loss": 0.1669, "step": 2488 }, { "epoch": 0.52, "learning_rate": 2.2250830564784051e-07, "loss": 0.1728, "step": 2489 }, { "epoch": 0.52, "learning_rate": 2.2247715946843853e-07, "loss": 0.2209, "step": 2490 }, { "epoch": 0.52, "learning_rate": 2.2244601328903655e-07, "loss": 0.1685, "step": 2491 }, { "epoch": 0.52, "learning_rate": 2.2241486710963454e-07, "loss": 0.2982, "step": 2492 }, { "epoch": 0.52, "learning_rate": 2.2238372093023255e-07, "loss": 0.2078, "step": 2493 }, { "epoch": 0.52, "learning_rate": 2.2235257475083057e-07, "loss": 0.1994, "step": 2494 }, { "epoch": 0.52, "learning_rate": 2.2232142857142858e-07, "loss": 0.1845, "step": 2495 }, { "epoch": 0.52, "learning_rate": 2.2229028239202657e-07, "loss": 0.1581, "step": 2496 }, { "epoch": 0.52, "learning_rate": 2.2225913621262456e-07, "loss": 0.1545, "step": 2497 }, { "epoch": 0.52, "learning_rate": 2.2222799003322258e-07, "loss": 0.2432, "step": 2498 }, { "epoch": 0.52, "learning_rate": 2.221968438538206e-07, "loss": 0.1915, "step": 2499 }, { "epoch": 0.52, "learning_rate": 2.221656976744186e-07, "loss": 0.1515, "step": 2500 }, { "epoch": 0.52, "learning_rate": 2.221345514950166e-07, "loss": 0.1971, "step": 2501 }, { "epoch": 0.52, "learning_rate": 2.221034053156146e-07, "loss": 0.2638, "step": 2502 }, { "epoch": 0.52, "learning_rate": 2.220722591362126e-07, "loss": 0.2126, "step": 2503 }, { "epoch": 0.52, "learning_rate": 2.2204111295681062e-07, "loss": 0.2071, "step": 2504 }, { "epoch": 0.52, "learning_rate": 2.2200996677740864e-07, "loss": 0.213, "step": 2505 }, { "epoch": 0.52, "learning_rate": 2.2197882059800662e-07, "loss": 0.1959, "step": 2506 }, { "epoch": 0.52, "learning_rate": 2.2194767441860464e-07, "loss": 0.2559, "step": 2507 }, { "epoch": 0.52, "learning_rate": 2.2191652823920266e-07, "loss": 0.2089, "step": 2508 }, { "epoch": 0.52, "learning_rate": 2.2188538205980067e-07, "loss": 0.1704, "step": 2509 }, { "epoch": 0.52, "learning_rate": 2.218542358803987e-07, "loss": 0.1784, "step": 2510 }, { "epoch": 0.52, "learning_rate": 2.2182308970099665e-07, "loss": 0.177, "step": 2511 }, { "epoch": 0.52, "learning_rate": 2.2179194352159467e-07, "loss": 0.2027, "step": 2512 }, { "epoch": 0.52, "learning_rate": 2.2176079734219268e-07, "loss": 0.1629, "step": 2513 }, { "epoch": 0.52, "learning_rate": 2.217296511627907e-07, "loss": 0.1892, "step": 2514 }, { "epoch": 0.52, "learning_rate": 2.216985049833887e-07, "loss": 0.1953, "step": 2515 }, { "epoch": 0.52, "learning_rate": 2.216673588039867e-07, "loss": 0.2023, "step": 2516 }, { "epoch": 0.52, "learning_rate": 2.2163621262458472e-07, "loss": 0.1934, "step": 2517 }, { "epoch": 0.52, "learning_rate": 2.216050664451827e-07, "loss": 0.1656, "step": 2518 }, { "epoch": 0.52, "learning_rate": 2.2157392026578072e-07, "loss": 0.2207, "step": 2519 }, { "epoch": 0.52, "learning_rate": 2.2154277408637871e-07, "loss": 0.2014, "step": 2520 }, { "epoch": 0.52, "learning_rate": 2.2151162790697673e-07, "loss": 0.2258, "step": 2521 }, { "epoch": 0.52, "learning_rate": 2.2148048172757475e-07, "loss": 0.1256, "step": 2522 }, { "epoch": 0.52, "learning_rate": 2.2144933554817276e-07, "loss": 0.1424, "step": 2523 }, { "epoch": 0.52, "learning_rate": 2.2141818936877078e-07, "loss": 0.2139, "step": 2524 }, { "epoch": 0.52, "learning_rate": 2.2138704318936874e-07, "loss": 0.2123, "step": 2525 }, { "epoch": 0.52, "learning_rate": 2.2135589700996676e-07, "loss": 0.1689, "step": 2526 }, { "epoch": 0.52, "learning_rate": 2.2132475083056477e-07, "loss": 0.162, "step": 2527 }, { "epoch": 0.52, "learning_rate": 2.212936046511628e-07, "loss": 0.1777, "step": 2528 }, { "epoch": 0.53, "learning_rate": 2.2126245847176078e-07, "loss": 0.2796, "step": 2529 }, { "epoch": 0.53, "learning_rate": 2.212313122923588e-07, "loss": 0.2008, "step": 2530 }, { "epoch": 0.53, "learning_rate": 2.212001661129568e-07, "loss": 0.2287, "step": 2531 }, { "epoch": 0.53, "learning_rate": 2.2116901993355482e-07, "loss": 0.1736, "step": 2532 }, { "epoch": 0.53, "learning_rate": 2.2113787375415281e-07, "loss": 0.2126, "step": 2533 }, { "epoch": 0.53, "learning_rate": 2.211067275747508e-07, "loss": 0.1849, "step": 2534 }, { "epoch": 0.53, "learning_rate": 2.2107558139534882e-07, "loss": 0.204, "step": 2535 }, { "epoch": 0.53, "learning_rate": 2.2104443521594683e-07, "loss": 0.2532, "step": 2536 }, { "epoch": 0.53, "learning_rate": 2.2101328903654485e-07, "loss": 0.1822, "step": 2537 }, { "epoch": 0.53, "learning_rate": 2.2098214285714287e-07, "loss": 0.2145, "step": 2538 }, { "epoch": 0.53, "learning_rate": 2.2095099667774083e-07, "loss": 0.1544, "step": 2539 }, { "epoch": 0.53, "learning_rate": 2.2091985049833885e-07, "loss": 0.1397, "step": 2540 }, { "epoch": 0.53, "learning_rate": 2.2088870431893686e-07, "loss": 0.2303, "step": 2541 }, { "epoch": 0.53, "learning_rate": 2.2085755813953488e-07, "loss": 0.2631, "step": 2542 }, { "epoch": 0.53, "learning_rate": 2.208264119601329e-07, "loss": 0.2571, "step": 2543 }, { "epoch": 0.53, "learning_rate": 2.2079526578073088e-07, "loss": 0.205, "step": 2544 }, { "epoch": 0.53, "learning_rate": 2.207641196013289e-07, "loss": 0.2031, "step": 2545 }, { "epoch": 0.53, "learning_rate": 2.2073297342192691e-07, "loss": 0.1403, "step": 2546 }, { "epoch": 0.53, "learning_rate": 2.2070182724252493e-07, "loss": 0.2396, "step": 2547 }, { "epoch": 0.53, "learning_rate": 2.206706810631229e-07, "loss": 0.2167, "step": 2548 }, { "epoch": 0.53, "learning_rate": 2.206395348837209e-07, "loss": 0.1754, "step": 2549 }, { "epoch": 0.53, "learning_rate": 2.2060838870431892e-07, "loss": 0.2193, "step": 2550 }, { "epoch": 0.53, "learning_rate": 2.2057724252491694e-07, "loss": 0.2043, "step": 2551 }, { "epoch": 0.53, "learning_rate": 2.2054609634551496e-07, "loss": 0.1824, "step": 2552 }, { "epoch": 0.53, "learning_rate": 2.2051495016611295e-07, "loss": 0.1927, "step": 2553 }, { "epoch": 0.53, "learning_rate": 2.2048380398671096e-07, "loss": 0.2477, "step": 2554 }, { "epoch": 0.53, "learning_rate": 2.2045265780730895e-07, "loss": 0.2462, "step": 2555 }, { "epoch": 0.53, "learning_rate": 2.2042151162790697e-07, "loss": 0.2105, "step": 2556 }, { "epoch": 0.53, "learning_rate": 2.2039036544850498e-07, "loss": 0.1948, "step": 2557 }, { "epoch": 0.53, "learning_rate": 2.2035921926910297e-07, "loss": 0.1912, "step": 2558 }, { "epoch": 0.53, "learning_rate": 2.20328073089701e-07, "loss": 0.2223, "step": 2559 }, { "epoch": 0.53, "learning_rate": 2.20296926910299e-07, "loss": 0.2313, "step": 2560 }, { "epoch": 0.53, "learning_rate": 2.2026578073089702e-07, "loss": 0.1929, "step": 2561 }, { "epoch": 0.53, "learning_rate": 2.2023463455149498e-07, "loss": 0.1742, "step": 2562 }, { "epoch": 0.53, "learning_rate": 2.20203488372093e-07, "loss": 0.2917, "step": 2563 }, { "epoch": 0.53, "learning_rate": 2.2017234219269101e-07, "loss": 0.1894, "step": 2564 }, { "epoch": 0.53, "learning_rate": 2.2014119601328903e-07, "loss": 0.1837, "step": 2565 }, { "epoch": 0.53, "learning_rate": 2.2011004983388704e-07, "loss": 0.1383, "step": 2566 }, { "epoch": 0.53, "learning_rate": 2.2007890365448503e-07, "loss": 0.2034, "step": 2567 }, { "epoch": 0.53, "learning_rate": 2.2004775747508305e-07, "loss": 0.2373, "step": 2568 }, { "epoch": 0.53, "learning_rate": 2.2001661129568107e-07, "loss": 0.2307, "step": 2569 }, { "epoch": 0.53, "learning_rate": 2.1998546511627906e-07, "loss": 0.2496, "step": 2570 }, { "epoch": 0.53, "learning_rate": 2.1995431893687707e-07, "loss": 0.2702, "step": 2571 }, { "epoch": 0.53, "learning_rate": 2.1992317275747506e-07, "loss": 0.2077, "step": 2572 }, { "epoch": 0.53, "learning_rate": 2.1989202657807308e-07, "loss": 0.2668, "step": 2573 }, { "epoch": 0.53, "learning_rate": 2.198608803986711e-07, "loss": 0.2471, "step": 2574 }, { "epoch": 0.53, "learning_rate": 2.198297342192691e-07, "loss": 0.2078, "step": 2575 }, { "epoch": 0.53, "learning_rate": 2.1979858803986707e-07, "loss": 0.1755, "step": 2576 }, { "epoch": 0.54, "learning_rate": 2.1976744186046509e-07, "loss": 0.2736, "step": 2577 }, { "epoch": 0.54, "learning_rate": 2.197362956810631e-07, "loss": 0.1757, "step": 2578 }, { "epoch": 0.54, "learning_rate": 2.1970514950166112e-07, "loss": 0.2087, "step": 2579 }, { "epoch": 0.54, "learning_rate": 2.1967400332225913e-07, "loss": 0.201, "step": 2580 }, { "epoch": 0.54, "learning_rate": 2.1964285714285712e-07, "loss": 0.1669, "step": 2581 }, { "epoch": 0.54, "learning_rate": 2.1961171096345514e-07, "loss": 0.2013, "step": 2582 }, { "epoch": 0.54, "learning_rate": 2.1958056478405316e-07, "loss": 0.1778, "step": 2583 }, { "epoch": 0.54, "learning_rate": 2.1954941860465117e-07, "loss": 0.2425, "step": 2584 }, { "epoch": 0.54, "learning_rate": 2.1951827242524919e-07, "loss": 0.1663, "step": 2585 }, { "epoch": 0.54, "learning_rate": 2.1948712624584715e-07, "loss": 0.1857, "step": 2586 }, { "epoch": 0.54, "learning_rate": 2.1945598006644517e-07, "loss": 0.1876, "step": 2587 }, { "epoch": 0.54, "learning_rate": 2.1942483388704318e-07, "loss": 0.2191, "step": 2588 }, { "epoch": 0.54, "learning_rate": 2.193936877076412e-07, "loss": 0.2188, "step": 2589 }, { "epoch": 0.54, "learning_rate": 2.193625415282392e-07, "loss": 0.238, "step": 2590 }, { "epoch": 0.54, "learning_rate": 2.193313953488372e-07, "loss": 0.193, "step": 2591 }, { "epoch": 0.54, "learning_rate": 2.193002491694352e-07, "loss": 0.2147, "step": 2592 }, { "epoch": 0.54, "learning_rate": 2.192691029900332e-07, "loss": 0.2266, "step": 2593 }, { "epoch": 0.54, "learning_rate": 2.1923795681063122e-07, "loss": 0.1684, "step": 2594 }, { "epoch": 0.54, "learning_rate": 2.192068106312292e-07, "loss": 0.225, "step": 2595 }, { "epoch": 0.54, "learning_rate": 2.1917566445182723e-07, "loss": 0.2334, "step": 2596 }, { "epoch": 0.54, "learning_rate": 2.1914451827242524e-07, "loss": 0.1881, "step": 2597 }, { "epoch": 0.54, "learning_rate": 2.1911337209302326e-07, "loss": 0.2282, "step": 2598 }, { "epoch": 0.54, "learning_rate": 2.1908222591362128e-07, "loss": 0.1651, "step": 2599 }, { "epoch": 0.54, "learning_rate": 2.1905107973421924e-07, "loss": 0.1844, "step": 2600 }, { "epoch": 0.54, "learning_rate": 2.1901993355481725e-07, "loss": 0.2217, "step": 2601 }, { "epoch": 0.54, "learning_rate": 2.1898878737541527e-07, "loss": 0.1922, "step": 2602 }, { "epoch": 0.54, "learning_rate": 2.1895764119601329e-07, "loss": 0.2301, "step": 2603 }, { "epoch": 0.54, "learning_rate": 2.189264950166113e-07, "loss": 0.3, "step": 2604 }, { "epoch": 0.54, "learning_rate": 2.188953488372093e-07, "loss": 0.1861, "step": 2605 }, { "epoch": 0.54, "learning_rate": 2.188642026578073e-07, "loss": 0.2036, "step": 2606 }, { "epoch": 0.54, "learning_rate": 2.188330564784053e-07, "loss": 0.1699, "step": 2607 }, { "epoch": 0.54, "learning_rate": 2.188019102990033e-07, "loss": 0.1983, "step": 2608 }, { "epoch": 0.54, "learning_rate": 2.187707641196013e-07, "loss": 0.1907, "step": 2609 }, { "epoch": 0.54, "learning_rate": 2.1873961794019932e-07, "loss": 0.17, "step": 2610 }, { "epoch": 0.54, "learning_rate": 2.1870847176079733e-07, "loss": 0.1887, "step": 2611 }, { "epoch": 0.54, "learning_rate": 2.1867732558139535e-07, "loss": 0.2085, "step": 2612 }, { "epoch": 0.54, "learning_rate": 2.1864617940199337e-07, "loss": 0.1845, "step": 2613 }, { "epoch": 0.54, "learning_rate": 2.1861503322259133e-07, "loss": 0.221, "step": 2614 }, { "epoch": 0.54, "learning_rate": 2.1858388704318934e-07, "loss": 0.2353, "step": 2615 }, { "epoch": 0.54, "learning_rate": 2.1855274086378736e-07, "loss": 0.1476, "step": 2616 }, { "epoch": 0.54, "learning_rate": 2.1852159468438538e-07, "loss": 0.2452, "step": 2617 }, { "epoch": 0.54, "learning_rate": 2.184904485049834e-07, "loss": 0.2034, "step": 2618 }, { "epoch": 0.54, "learning_rate": 2.1845930232558138e-07, "loss": 0.2172, "step": 2619 }, { "epoch": 0.54, "learning_rate": 2.184281561461794e-07, "loss": 0.1782, "step": 2620 }, { "epoch": 0.54, "learning_rate": 2.183970099667774e-07, "loss": 0.1867, "step": 2621 }, { "epoch": 0.54, "learning_rate": 2.1836586378737543e-07, "loss": 0.2041, "step": 2622 }, { "epoch": 0.54, "learning_rate": 2.183347176079734e-07, "loss": 0.1671, "step": 2623 }, { "epoch": 0.54, "learning_rate": 2.183035714285714e-07, "loss": 0.1943, "step": 2624 }, { "epoch": 0.55, "learning_rate": 2.1827242524916942e-07, "loss": 0.1921, "step": 2625 }, { "epoch": 0.55, "learning_rate": 2.1824127906976744e-07, "loss": 0.138, "step": 2626 }, { "epoch": 0.55, "learning_rate": 2.1821013289036545e-07, "loss": 0.1569, "step": 2627 }, { "epoch": 0.55, "learning_rate": 2.1817898671096344e-07, "loss": 0.2314, "step": 2628 }, { "epoch": 0.55, "learning_rate": 2.1814784053156143e-07, "loss": 0.1798, "step": 2629 }, { "epoch": 0.55, "learning_rate": 2.1811669435215945e-07, "loss": 0.1916, "step": 2630 }, { "epoch": 0.55, "learning_rate": 2.1808554817275746e-07, "loss": 0.1684, "step": 2631 }, { "epoch": 0.55, "learning_rate": 2.1805440199335548e-07, "loss": 0.1535, "step": 2632 }, { "epoch": 0.55, "learning_rate": 2.1802325581395347e-07, "loss": 0.191, "step": 2633 }, { "epoch": 0.55, "learning_rate": 2.1799210963455149e-07, "loss": 0.2204, "step": 2634 }, { "epoch": 0.55, "learning_rate": 2.179609634551495e-07, "loss": 0.1728, "step": 2635 }, { "epoch": 0.55, "learning_rate": 2.1792981727574752e-07, "loss": 0.1827, "step": 2636 }, { "epoch": 0.55, "learning_rate": 2.1789867109634553e-07, "loss": 0.1871, "step": 2637 }, { "epoch": 0.55, "learning_rate": 2.178675249169435e-07, "loss": 0.1765, "step": 2638 }, { "epoch": 0.55, "learning_rate": 2.178363787375415e-07, "loss": 0.1827, "step": 2639 }, { "epoch": 0.55, "learning_rate": 2.1780523255813953e-07, "loss": 0.187, "step": 2640 }, { "epoch": 0.55, "learning_rate": 2.1777408637873754e-07, "loss": 0.2, "step": 2641 }, { "epoch": 0.55, "learning_rate": 2.1774294019933553e-07, "loss": 0.1497, "step": 2642 }, { "epoch": 0.55, "learning_rate": 2.1771179401993355e-07, "loss": 0.1759, "step": 2643 }, { "epoch": 0.55, "learning_rate": 2.1768064784053154e-07, "loss": 0.2721, "step": 2644 }, { "epoch": 0.55, "learning_rate": 2.1764950166112955e-07, "loss": 0.163, "step": 2645 }, { "epoch": 0.55, "learning_rate": 2.1761835548172757e-07, "loss": 0.1916, "step": 2646 }, { "epoch": 0.55, "learning_rate": 2.1758720930232556e-07, "loss": 0.1696, "step": 2647 }, { "epoch": 0.55, "learning_rate": 2.1755606312292357e-07, "loss": 0.2359, "step": 2648 }, { "epoch": 0.55, "learning_rate": 2.175249169435216e-07, "loss": 0.1994, "step": 2649 }, { "epoch": 0.55, "learning_rate": 2.174937707641196e-07, "loss": 0.1843, "step": 2650 }, { "epoch": 0.55, "learning_rate": 2.1746262458471762e-07, "loss": 0.1889, "step": 2651 }, { "epoch": 0.55, "learning_rate": 2.1743147840531559e-07, "loss": 0.2248, "step": 2652 }, { "epoch": 0.55, "learning_rate": 2.174003322259136e-07, "loss": 0.1986, "step": 2653 }, { "epoch": 0.55, "learning_rate": 2.1736918604651162e-07, "loss": 0.2384, "step": 2654 }, { "epoch": 0.55, "learning_rate": 2.1733803986710963e-07, "loss": 0.2608, "step": 2655 }, { "epoch": 0.55, "learning_rate": 2.1730689368770762e-07, "loss": 0.2208, "step": 2656 }, { "epoch": 0.55, "learning_rate": 2.1727574750830564e-07, "loss": 0.1973, "step": 2657 }, { "epoch": 0.55, "learning_rate": 2.1724460132890365e-07, "loss": 0.2444, "step": 2658 }, { "epoch": 0.55, "learning_rate": 2.1721345514950167e-07, "loss": 0.2013, "step": 2659 }, { "epoch": 0.55, "learning_rate": 2.1718230897009966e-07, "loss": 0.2195, "step": 2660 }, { "epoch": 0.55, "learning_rate": 2.1715116279069765e-07, "loss": 0.1923, "step": 2661 }, { "epoch": 0.55, "learning_rate": 2.1712001661129566e-07, "loss": 0.2376, "step": 2662 }, { "epoch": 0.55, "learning_rate": 2.1708887043189368e-07, "loss": 0.1979, "step": 2663 }, { "epoch": 0.55, "learning_rate": 2.170577242524917e-07, "loss": 0.1929, "step": 2664 }, { "epoch": 0.55, "learning_rate": 2.170265780730897e-07, "loss": 0.1327, "step": 2665 }, { "epoch": 0.55, "learning_rate": 2.1699543189368767e-07, "loss": 0.1685, "step": 2666 }, { "epoch": 0.55, "learning_rate": 2.169642857142857e-07, "loss": 0.157, "step": 2667 }, { "epoch": 0.55, "learning_rate": 2.169331395348837e-07, "loss": 0.2076, "step": 2668 }, { "epoch": 0.55, "learning_rate": 2.1690199335548172e-07, "loss": 0.2026, "step": 2669 }, { "epoch": 0.55, "learning_rate": 2.1687084717607974e-07, "loss": 0.2089, "step": 2670 }, { "epoch": 0.55, "learning_rate": 2.1683970099667773e-07, "loss": 0.1825, "step": 2671 }, { "epoch": 0.55, "learning_rate": 2.1680855481727574e-07, "loss": 0.1805, "step": 2672 }, { "epoch": 0.56, "learning_rate": 2.1677740863787376e-07, "loss": 0.191, "step": 2673 }, { "epoch": 0.56, "learning_rate": 2.1674626245847177e-07, "loss": 0.1701, "step": 2674 }, { "epoch": 0.56, "learning_rate": 2.1671511627906974e-07, "loss": 0.2509, "step": 2675 }, { "epoch": 0.56, "learning_rate": 2.1668397009966775e-07, "loss": 0.2308, "step": 2676 }, { "epoch": 0.56, "learning_rate": 2.1665282392026577e-07, "loss": 0.1543, "step": 2677 }, { "epoch": 0.56, "learning_rate": 2.1662167774086378e-07, "loss": 0.1895, "step": 2678 }, { "epoch": 0.56, "learning_rate": 2.165905315614618e-07, "loss": 0.3071, "step": 2679 }, { "epoch": 0.56, "learning_rate": 2.165593853820598e-07, "loss": 0.1546, "step": 2680 }, { "epoch": 0.56, "learning_rate": 2.1652823920265778e-07, "loss": 0.2043, "step": 2681 }, { "epoch": 0.56, "learning_rate": 2.164970930232558e-07, "loss": 0.1837, "step": 2682 }, { "epoch": 0.56, "learning_rate": 2.164659468438538e-07, "loss": 0.1465, "step": 2683 }, { "epoch": 0.56, "learning_rate": 2.1643480066445183e-07, "loss": 0.1634, "step": 2684 }, { "epoch": 0.56, "learning_rate": 2.1640365448504982e-07, "loss": 0.1657, "step": 2685 }, { "epoch": 0.56, "learning_rate": 2.1637250830564783e-07, "loss": 0.2026, "step": 2686 }, { "epoch": 0.56, "learning_rate": 2.1634136212624585e-07, "loss": 0.2088, "step": 2687 }, { "epoch": 0.56, "learning_rate": 2.1631021594684386e-07, "loss": 0.2009, "step": 2688 }, { "epoch": 0.56, "learning_rate": 2.1627906976744183e-07, "loss": 0.183, "step": 2689 }, { "epoch": 0.56, "learning_rate": 2.1624792358803984e-07, "loss": 0.1763, "step": 2690 }, { "epoch": 0.56, "learning_rate": 2.1621677740863786e-07, "loss": 0.1881, "step": 2691 }, { "epoch": 0.56, "learning_rate": 2.1618563122923587e-07, "loss": 0.2075, "step": 2692 }, { "epoch": 0.56, "learning_rate": 2.161544850498339e-07, "loss": 0.2071, "step": 2693 }, { "epoch": 0.56, "learning_rate": 2.1612333887043188e-07, "loss": 0.1954, "step": 2694 }, { "epoch": 0.56, "learning_rate": 2.160921926910299e-07, "loss": 0.2356, "step": 2695 }, { "epoch": 0.56, "learning_rate": 2.160610465116279e-07, "loss": 0.1971, "step": 2696 }, { "epoch": 0.56, "learning_rate": 2.160299003322259e-07, "loss": 0.2252, "step": 2697 }, { "epoch": 0.56, "learning_rate": 2.1599875415282392e-07, "loss": 0.1964, "step": 2698 }, { "epoch": 0.56, "learning_rate": 2.159676079734219e-07, "loss": 0.1772, "step": 2699 }, { "epoch": 0.56, "learning_rate": 2.1593646179401992e-07, "loss": 0.2973, "step": 2700 }, { "epoch": 0.56, "learning_rate": 2.1590531561461794e-07, "loss": 0.1681, "step": 2701 }, { "epoch": 0.56, "learning_rate": 2.1587416943521595e-07, "loss": 0.2139, "step": 2702 }, { "epoch": 0.56, "learning_rate": 2.1584302325581392e-07, "loss": 0.1889, "step": 2703 }, { "epoch": 0.56, "learning_rate": 2.1581187707641193e-07, "loss": 0.1762, "step": 2704 }, { "epoch": 0.56, "learning_rate": 2.1578073089700995e-07, "loss": 0.2447, "step": 2705 }, { "epoch": 0.56, "learning_rate": 2.1574958471760796e-07, "loss": 0.182, "step": 2706 }, { "epoch": 0.56, "learning_rate": 2.1571843853820598e-07, "loss": 0.2147, "step": 2707 }, { "epoch": 0.56, "learning_rate": 2.1568729235880397e-07, "loss": 0.1679, "step": 2708 }, { "epoch": 0.56, "learning_rate": 2.1565614617940198e-07, "loss": 0.1658, "step": 2709 }, { "epoch": 0.56, "learning_rate": 2.15625e-07, "loss": 0.2058, "step": 2710 }, { "epoch": 0.56, "learning_rate": 2.1559385382059802e-07, "loss": 0.2098, "step": 2711 }, { "epoch": 0.56, "learning_rate": 2.15562707641196e-07, "loss": 0.1768, "step": 2712 }, { "epoch": 0.56, "learning_rate": 2.15531561461794e-07, "loss": 0.1699, "step": 2713 }, { "epoch": 0.56, "learning_rate": 2.15500415282392e-07, "loss": 0.1575, "step": 2714 }, { "epoch": 0.56, "learning_rate": 2.1546926910299003e-07, "loss": 0.1877, "step": 2715 }, { "epoch": 0.56, "learning_rate": 2.1543812292358804e-07, "loss": 0.1917, "step": 2716 }, { "epoch": 0.56, "learning_rate": 2.1540697674418606e-07, "loss": 0.1814, "step": 2717 }, { "epoch": 0.56, "learning_rate": 2.1537583056478402e-07, "loss": 0.1624, "step": 2718 }, { "epoch": 0.56, "learning_rate": 2.1534468438538204e-07, "loss": 0.1614, "step": 2719 }, { "epoch": 0.56, "learning_rate": 2.1531353820598005e-07, "loss": 0.212, "step": 2720 }, { "epoch": 0.56, "learning_rate": 2.1528239202657807e-07, "loss": 0.1954, "step": 2721 }, { "epoch": 0.57, "learning_rate": 2.1525124584717606e-07, "loss": 0.233, "step": 2722 }, { "epoch": 0.57, "learning_rate": 2.1522009966777407e-07, "loss": 0.1749, "step": 2723 }, { "epoch": 0.57, "learning_rate": 2.151889534883721e-07, "loss": 0.1842, "step": 2724 }, { "epoch": 0.57, "learning_rate": 2.151578073089701e-07, "loss": 0.3168, "step": 2725 }, { "epoch": 0.57, "learning_rate": 2.1512666112956812e-07, "loss": 0.1535, "step": 2726 }, { "epoch": 0.57, "learning_rate": 2.1509551495016608e-07, "loss": 0.1773, "step": 2727 }, { "epoch": 0.57, "learning_rate": 2.150643687707641e-07, "loss": 0.1866, "step": 2728 }, { "epoch": 0.57, "learning_rate": 2.1503322259136212e-07, "loss": 0.1846, "step": 2729 }, { "epoch": 0.57, "learning_rate": 2.1500207641196013e-07, "loss": 0.1503, "step": 2730 }, { "epoch": 0.57, "learning_rate": 2.1497093023255815e-07, "loss": 0.2046, "step": 2731 }, { "epoch": 0.57, "learning_rate": 2.1493978405315614e-07, "loss": 0.168, "step": 2732 }, { "epoch": 0.57, "learning_rate": 2.1490863787375415e-07, "loss": 0.1965, "step": 2733 }, { "epoch": 0.57, "learning_rate": 2.1487749169435214e-07, "loss": 0.1297, "step": 2734 }, { "epoch": 0.57, "learning_rate": 2.1484634551495016e-07, "loss": 0.1869, "step": 2735 }, { "epoch": 0.57, "learning_rate": 2.1481519933554815e-07, "loss": 0.1964, "step": 2736 }, { "epoch": 0.57, "learning_rate": 2.1478405315614616e-07, "loss": 0.1895, "step": 2737 }, { "epoch": 0.57, "learning_rate": 2.1475290697674418e-07, "loss": 0.2054, "step": 2738 }, { "epoch": 0.57, "learning_rate": 2.147217607973422e-07, "loss": 0.1779, "step": 2739 }, { "epoch": 0.57, "learning_rate": 2.146906146179402e-07, "loss": 0.2428, "step": 2740 }, { "epoch": 0.57, "learning_rate": 2.1465946843853817e-07, "loss": 0.2339, "step": 2741 }, { "epoch": 0.57, "learning_rate": 2.146283222591362e-07, "loss": 0.2128, "step": 2742 }, { "epoch": 0.57, "learning_rate": 2.145971760797342e-07, "loss": 0.1857, "step": 2743 }, { "epoch": 0.57, "learning_rate": 2.1456602990033222e-07, "loss": 0.5919, "step": 2744 }, { "epoch": 0.57, "learning_rate": 2.1453488372093024e-07, "loss": 0.1858, "step": 2745 }, { "epoch": 0.57, "learning_rate": 2.1450373754152823e-07, "loss": 0.1679, "step": 2746 }, { "epoch": 0.57, "learning_rate": 2.1447259136212624e-07, "loss": 0.2424, "step": 2747 }, { "epoch": 0.57, "learning_rate": 2.1444144518272426e-07, "loss": 0.1884, "step": 2748 }, { "epoch": 0.57, "learning_rate": 2.1441029900332225e-07, "loss": 0.2798, "step": 2749 }, { "epoch": 0.57, "learning_rate": 2.1437915282392024e-07, "loss": 0.2197, "step": 2750 }, { "epoch": 0.57, "learning_rate": 2.1434800664451825e-07, "loss": 0.1363, "step": 2751 }, { "epoch": 0.57, "learning_rate": 2.1431686046511627e-07, "loss": 0.1815, "step": 2752 }, { "epoch": 0.57, "learning_rate": 2.1428571428571428e-07, "loss": 0.225, "step": 2753 }, { "epoch": 0.57, "learning_rate": 2.142545681063123e-07, "loss": 0.1676, "step": 2754 }, { "epoch": 0.57, "learning_rate": 2.1422342192691026e-07, "loss": 0.2025, "step": 2755 }, { "epoch": 0.57, "learning_rate": 2.1419227574750828e-07, "loss": 0.19, "step": 2756 }, { "epoch": 0.57, "learning_rate": 2.141611295681063e-07, "loss": 0.2238, "step": 2757 }, { "epoch": 0.57, "learning_rate": 2.141299833887043e-07, "loss": 0.1594, "step": 2758 }, { "epoch": 0.57, "learning_rate": 2.1409883720930233e-07, "loss": 0.215, "step": 2759 }, { "epoch": 0.57, "learning_rate": 2.1406769102990031e-07, "loss": 0.1968, "step": 2760 }, { "epoch": 0.57, "learning_rate": 2.1403654485049833e-07, "loss": 0.1439, "step": 2761 }, { "epoch": 0.57, "learning_rate": 2.1400539867109635e-07, "loss": 0.2314, "step": 2762 }, { "epoch": 0.57, "learning_rate": 2.1397425249169436e-07, "loss": 0.1721, "step": 2763 }, { "epoch": 0.57, "learning_rate": 2.1394310631229238e-07, "loss": 0.2167, "step": 2764 }, { "epoch": 0.57, "learning_rate": 2.1391196013289034e-07, "loss": 0.2198, "step": 2765 }, { "epoch": 0.57, "learning_rate": 2.1388081395348836e-07, "loss": 0.1947, "step": 2766 }, { "epoch": 0.57, "learning_rate": 2.1384966777408637e-07, "loss": 0.1589, "step": 2767 }, { "epoch": 0.57, "learning_rate": 2.138185215946844e-07, "loss": 0.1688, "step": 2768 }, { "epoch": 0.57, "learning_rate": 2.1378737541528238e-07, "loss": 0.2322, "step": 2769 }, { "epoch": 0.58, "learning_rate": 2.137562292358804e-07, "loss": 0.1922, "step": 2770 }, { "epoch": 0.58, "learning_rate": 2.1372508305647838e-07, "loss": 0.1682, "step": 2771 }, { "epoch": 0.58, "learning_rate": 2.136939368770764e-07, "loss": 0.1591, "step": 2772 }, { "epoch": 0.58, "learning_rate": 2.1366279069767441e-07, "loss": 0.2042, "step": 2773 }, { "epoch": 0.58, "learning_rate": 2.136316445182724e-07, "loss": 0.2114, "step": 2774 }, { "epoch": 0.58, "learning_rate": 2.1360049833887042e-07, "loss": 0.2225, "step": 2775 }, { "epoch": 0.58, "learning_rate": 2.1356935215946844e-07, "loss": 0.1822, "step": 2776 }, { "epoch": 0.58, "learning_rate": 2.1353820598006645e-07, "loss": 0.2072, "step": 2777 }, { "epoch": 0.58, "learning_rate": 2.1350705980066447e-07, "loss": 0.1928, "step": 2778 }, { "epoch": 0.58, "learning_rate": 2.1347591362126243e-07, "loss": 0.2305, "step": 2779 }, { "epoch": 0.58, "learning_rate": 2.1344476744186045e-07, "loss": 0.1685, "step": 2780 }, { "epoch": 0.58, "learning_rate": 2.1341362126245846e-07, "loss": 0.2274, "step": 2781 }, { "epoch": 0.58, "learning_rate": 2.1338247508305648e-07, "loss": 0.2147, "step": 2782 }, { "epoch": 0.58, "learning_rate": 2.1335132890365447e-07, "loss": 0.1791, "step": 2783 }, { "epoch": 0.58, "learning_rate": 2.1332018272425248e-07, "loss": 0.1824, "step": 2784 }, { "epoch": 0.58, "learning_rate": 2.132890365448505e-07, "loss": 0.2072, "step": 2785 }, { "epoch": 0.58, "learning_rate": 2.132578903654485e-07, "loss": 0.1723, "step": 2786 }, { "epoch": 0.58, "learning_rate": 2.132267441860465e-07, "loss": 0.1931, "step": 2787 }, { "epoch": 0.58, "learning_rate": 2.131955980066445e-07, "loss": 0.2058, "step": 2788 }, { "epoch": 0.58, "learning_rate": 2.131644518272425e-07, "loss": 0.2157, "step": 2789 }, { "epoch": 0.58, "learning_rate": 2.1313330564784052e-07, "loss": 0.1457, "step": 2790 }, { "epoch": 0.58, "learning_rate": 2.1310215946843854e-07, "loss": 0.2266, "step": 2791 }, { "epoch": 0.58, "learning_rate": 2.1307101328903656e-07, "loss": 0.2669, "step": 2792 }, { "epoch": 0.58, "learning_rate": 2.1303986710963452e-07, "loss": 0.2086, "step": 2793 }, { "epoch": 0.58, "learning_rate": 2.1300872093023254e-07, "loss": 0.1807, "step": 2794 }, { "epoch": 0.58, "learning_rate": 2.1297757475083055e-07, "loss": 0.1826, "step": 2795 }, { "epoch": 0.58, "learning_rate": 2.1294642857142857e-07, "loss": 0.1834, "step": 2796 }, { "epoch": 0.58, "learning_rate": 2.1291528239202658e-07, "loss": 0.2609, "step": 2797 }, { "epoch": 0.58, "learning_rate": 2.1288413621262457e-07, "loss": 0.2027, "step": 2798 }, { "epoch": 0.58, "learning_rate": 2.128529900332226e-07, "loss": 0.2764, "step": 2799 }, { "epoch": 0.58, "learning_rate": 2.128218438538206e-07, "loss": 0.1107, "step": 2800 }, { "epoch": 0.58, "learning_rate": 2.1279069767441862e-07, "loss": 0.1891, "step": 2801 }, { "epoch": 0.58, "learning_rate": 2.1275955149501658e-07, "loss": 0.1798, "step": 2802 }, { "epoch": 0.58, "learning_rate": 2.127284053156146e-07, "loss": 0.1727, "step": 2803 }, { "epoch": 0.58, "learning_rate": 2.1269725913621261e-07, "loss": 0.1659, "step": 2804 }, { "epoch": 0.58, "learning_rate": 2.1266611295681063e-07, "loss": 0.1713, "step": 2805 }, { "epoch": 0.58, "learning_rate": 2.1263496677740865e-07, "loss": 0.203, "step": 2806 }, { "epoch": 0.58, "learning_rate": 2.1260382059800664e-07, "loss": 0.1501, "step": 2807 }, { "epoch": 0.58, "learning_rate": 2.1257267441860462e-07, "loss": 0.1658, "step": 2808 }, { "epoch": 0.58, "learning_rate": 2.1254152823920264e-07, "loss": 0.2032, "step": 2809 }, { "epoch": 0.58, "learning_rate": 2.1251038205980066e-07, "loss": 0.2349, "step": 2810 }, { "epoch": 0.58, "learning_rate": 2.1247923588039867e-07, "loss": 0.2134, "step": 2811 }, { "epoch": 0.58, "learning_rate": 2.1244808970099666e-07, "loss": 0.229, "step": 2812 }, { "epoch": 0.58, "learning_rate": 2.1241694352159468e-07, "loss": 0.1857, "step": 2813 }, { "epoch": 0.58, "learning_rate": 2.123857973421927e-07, "loss": 0.1777, "step": 2814 }, { "epoch": 0.58, "learning_rate": 2.123546511627907e-07, "loss": 0.1982, "step": 2815 }, { "epoch": 0.58, "learning_rate": 2.1232350498338867e-07, "loss": 0.1706, "step": 2816 }, { "epoch": 0.58, "learning_rate": 2.122923588039867e-07, "loss": 0.1845, "step": 2817 }, { "epoch": 0.59, "learning_rate": 2.122612126245847e-07, "loss": 0.2161, "step": 2818 }, { "epoch": 0.59, "learning_rate": 2.1223006644518272e-07, "loss": 0.1286, "step": 2819 }, { "epoch": 0.59, "learning_rate": 2.1219892026578074e-07, "loss": 0.2245, "step": 2820 }, { "epoch": 0.59, "learning_rate": 2.1216777408637872e-07, "loss": 0.1947, "step": 2821 }, { "epoch": 0.59, "learning_rate": 2.1213662790697674e-07, "loss": 0.2398, "step": 2822 }, { "epoch": 0.59, "learning_rate": 2.1210548172757473e-07, "loss": 0.1492, "step": 2823 }, { "epoch": 0.59, "learning_rate": 2.1207433554817275e-07, "loss": 0.1672, "step": 2824 }, { "epoch": 0.59, "learning_rate": 2.1204318936877076e-07, "loss": 0.161, "step": 2825 }, { "epoch": 0.59, "learning_rate": 2.1201204318936875e-07, "loss": 0.2309, "step": 2826 }, { "epoch": 0.59, "learning_rate": 2.1198089700996677e-07, "loss": 0.199, "step": 2827 }, { "epoch": 0.59, "learning_rate": 2.1194975083056478e-07, "loss": 0.2152, "step": 2828 }, { "epoch": 0.59, "learning_rate": 2.119186046511628e-07, "loss": 0.1667, "step": 2829 }, { "epoch": 0.59, "learning_rate": 2.1188745847176076e-07, "loss": 0.1814, "step": 2830 }, { "epoch": 0.59, "learning_rate": 2.1185631229235878e-07, "loss": 0.2551, "step": 2831 }, { "epoch": 0.59, "learning_rate": 2.118251661129568e-07, "loss": 0.2112, "step": 2832 }, { "epoch": 0.59, "learning_rate": 2.117940199335548e-07, "loss": 0.2081, "step": 2833 }, { "epoch": 0.59, "learning_rate": 2.1176287375415282e-07, "loss": 0.187, "step": 2834 }, { "epoch": 0.59, "learning_rate": 2.1173172757475081e-07, "loss": 0.232, "step": 2835 }, { "epoch": 0.59, "learning_rate": 2.1170058139534883e-07, "loss": 0.216, "step": 2836 }, { "epoch": 0.59, "learning_rate": 2.1166943521594685e-07, "loss": 0.2239, "step": 2837 }, { "epoch": 0.59, "learning_rate": 2.1163828903654486e-07, "loss": 0.1567, "step": 2838 }, { "epoch": 0.59, "learning_rate": 2.1160714285714285e-07, "loss": 0.2079, "step": 2839 }, { "epoch": 0.59, "learning_rate": 2.1157599667774084e-07, "loss": 0.2193, "step": 2840 }, { "epoch": 0.59, "learning_rate": 2.1154485049833886e-07, "loss": 0.1642, "step": 2841 }, { "epoch": 0.59, "learning_rate": 2.1151370431893687e-07, "loss": 0.1578, "step": 2842 }, { "epoch": 0.59, "learning_rate": 2.114825581395349e-07, "loss": 0.2644, "step": 2843 }, { "epoch": 0.59, "learning_rate": 2.114514119601329e-07, "loss": 0.2185, "step": 2844 }, { "epoch": 0.59, "learning_rate": 2.1142026578073087e-07, "loss": 0.2342, "step": 2845 }, { "epoch": 0.59, "learning_rate": 2.1138911960132888e-07, "loss": 0.1447, "step": 2846 }, { "epoch": 0.59, "learning_rate": 2.113579734219269e-07, "loss": 0.1403, "step": 2847 }, { "epoch": 0.59, "learning_rate": 2.1132682724252491e-07, "loss": 0.2458, "step": 2848 }, { "epoch": 0.59, "learning_rate": 2.112956810631229e-07, "loss": 0.1794, "step": 2849 }, { "epoch": 0.59, "learning_rate": 2.1126453488372092e-07, "loss": 0.1608, "step": 2850 }, { "epoch": 0.59, "learning_rate": 2.1123338870431893e-07, "loss": 0.2031, "step": 2851 }, { "epoch": 0.59, "learning_rate": 2.1120224252491695e-07, "loss": 0.1892, "step": 2852 }, { "epoch": 0.59, "learning_rate": 2.1117109634551497e-07, "loss": 0.1778, "step": 2853 }, { "epoch": 0.59, "learning_rate": 2.1113995016611293e-07, "loss": 0.2602, "step": 2854 }, { "epoch": 0.59, "learning_rate": 2.1110880398671094e-07, "loss": 0.2247, "step": 2855 }, { "epoch": 0.59, "learning_rate": 2.1107765780730896e-07, "loss": 0.2629, "step": 2856 }, { "epoch": 0.59, "learning_rate": 2.1104651162790698e-07, "loss": 0.2289, "step": 2857 }, { "epoch": 0.59, "learning_rate": 2.11015365448505e-07, "loss": 0.2517, "step": 2858 }, { "epoch": 0.59, "learning_rate": 2.1098421926910298e-07, "loss": 0.1952, "step": 2859 }, { "epoch": 0.59, "learning_rate": 2.1095307308970097e-07, "loss": 0.1952, "step": 2860 }, { "epoch": 0.59, "learning_rate": 2.10921926910299e-07, "loss": 0.1605, "step": 2861 }, { "epoch": 0.59, "learning_rate": 2.10890780730897e-07, "loss": 0.2248, "step": 2862 }, { "epoch": 0.59, "learning_rate": 2.10859634551495e-07, "loss": 0.177, "step": 2863 }, { "epoch": 0.59, "learning_rate": 2.10828488372093e-07, "loss": 0.1628, "step": 2864 }, { "epoch": 0.59, "learning_rate": 2.1079734219269102e-07, "loss": 0.2126, "step": 2865 }, { "epoch": 0.6, "learning_rate": 2.1076619601328904e-07, "loss": 0.1918, "step": 2866 }, { "epoch": 0.6, "learning_rate": 2.1073504983388706e-07, "loss": 0.2187, "step": 2867 }, { "epoch": 0.6, "learning_rate": 2.1070390365448502e-07, "loss": 0.2293, "step": 2868 }, { "epoch": 0.6, "learning_rate": 2.1067275747508303e-07, "loss": 0.251, "step": 2869 }, { "epoch": 0.6, "learning_rate": 2.1064161129568105e-07, "loss": 0.1681, "step": 2870 }, { "epoch": 0.6, "learning_rate": 2.1061046511627907e-07, "loss": 0.1468, "step": 2871 }, { "epoch": 0.6, "learning_rate": 2.1057931893687708e-07, "loss": 0.1948, "step": 2872 }, { "epoch": 0.6, "learning_rate": 2.1054817275747507e-07, "loss": 0.1797, "step": 2873 }, { "epoch": 0.6, "learning_rate": 2.1051702657807309e-07, "loss": 0.1615, "step": 2874 }, { "epoch": 0.6, "learning_rate": 2.104858803986711e-07, "loss": 0.1854, "step": 2875 }, { "epoch": 0.6, "learning_rate": 2.104547342192691e-07, "loss": 0.1965, "step": 2876 }, { "epoch": 0.6, "learning_rate": 2.1042358803986708e-07, "loss": 0.2207, "step": 2877 }, { "epoch": 0.6, "learning_rate": 2.103924418604651e-07, "loss": 0.2268, "step": 2878 }, { "epoch": 0.6, "learning_rate": 2.103612956810631e-07, "loss": 0.2231, "step": 2879 }, { "epoch": 0.6, "learning_rate": 2.1033014950166113e-07, "loss": 0.1675, "step": 2880 }, { "epoch": 0.6, "learning_rate": 2.1029900332225914e-07, "loss": 0.2725, "step": 2881 }, { "epoch": 0.6, "learning_rate": 2.102678571428571e-07, "loss": 0.2175, "step": 2882 }, { "epoch": 0.6, "learning_rate": 2.1023671096345512e-07, "loss": 0.141, "step": 2883 }, { "epoch": 0.6, "learning_rate": 2.1020556478405314e-07, "loss": 0.1943, "step": 2884 }, { "epoch": 0.6, "learning_rate": 2.1017441860465115e-07, "loss": 0.2394, "step": 2885 }, { "epoch": 0.6, "learning_rate": 2.1014327242524917e-07, "loss": 0.2382, "step": 2886 }, { "epoch": 0.6, "learning_rate": 2.1011212624584716e-07, "loss": 0.1972, "step": 2887 }, { "epoch": 0.6, "learning_rate": 2.1008098006644518e-07, "loss": 0.1758, "step": 2888 }, { "epoch": 0.6, "learning_rate": 2.100498338870432e-07, "loss": 0.2308, "step": 2889 }, { "epoch": 0.6, "learning_rate": 2.100186877076412e-07, "loss": 0.2042, "step": 2890 }, { "epoch": 0.6, "learning_rate": 2.099875415282392e-07, "loss": 0.2158, "step": 2891 }, { "epoch": 0.6, "learning_rate": 2.0995639534883719e-07, "loss": 0.2285, "step": 2892 }, { "epoch": 0.6, "learning_rate": 2.099252491694352e-07, "loss": 0.2364, "step": 2893 }, { "epoch": 0.6, "learning_rate": 2.0989410299003322e-07, "loss": 0.1876, "step": 2894 }, { "epoch": 0.6, "learning_rate": 2.0986295681063123e-07, "loss": 0.2143, "step": 2895 }, { "epoch": 0.6, "learning_rate": 2.0983181063122922e-07, "loss": 0.1828, "step": 2896 }, { "epoch": 0.6, "learning_rate": 2.098006644518272e-07, "loss": 0.2839, "step": 2897 }, { "epoch": 0.6, "learning_rate": 2.0976951827242523e-07, "loss": 0.1578, "step": 2898 }, { "epoch": 0.6, "learning_rate": 2.0973837209302324e-07, "loss": 0.1912, "step": 2899 }, { "epoch": 0.6, "learning_rate": 2.0970722591362126e-07, "loss": 0.2386, "step": 2900 }, { "epoch": 0.6, "learning_rate": 2.0967607973421925e-07, "loss": 0.2306, "step": 2901 }, { "epoch": 0.6, "learning_rate": 2.0964493355481727e-07, "loss": 0.2049, "step": 2902 }, { "epoch": 0.6, "learning_rate": 2.0961378737541528e-07, "loss": 0.1909, "step": 2903 }, { "epoch": 0.6, "learning_rate": 2.095826411960133e-07, "loss": 0.2235, "step": 2904 }, { "epoch": 0.6, "learning_rate": 2.095514950166113e-07, "loss": 0.2281, "step": 2905 }, { "epoch": 0.6, "learning_rate": 2.0952034883720928e-07, "loss": 0.1503, "step": 2906 }, { "epoch": 0.6, "learning_rate": 2.094892026578073e-07, "loss": 0.147, "step": 2907 }, { "epoch": 0.6, "learning_rate": 2.094580564784053e-07, "loss": 0.1199, "step": 2908 }, { "epoch": 0.6, "learning_rate": 2.0942691029900332e-07, "loss": 0.2273, "step": 2909 }, { "epoch": 0.6, "learning_rate": 2.093957641196013e-07, "loss": 0.2411, "step": 2910 }, { "epoch": 0.6, "learning_rate": 2.0936461794019933e-07, "loss": 0.1748, "step": 2911 }, { "epoch": 0.6, "learning_rate": 2.0933347176079734e-07, "loss": 0.2181, "step": 2912 }, { "epoch": 0.6, "learning_rate": 2.0930232558139533e-07, "loss": 0.1794, "step": 2913 }, { "epoch": 0.61, "learning_rate": 2.0927117940199335e-07, "loss": 0.1576, "step": 2914 }, { "epoch": 0.61, "learning_rate": 2.0924003322259134e-07, "loss": 0.2079, "step": 2915 }, { "epoch": 0.61, "learning_rate": 2.0920888704318935e-07, "loss": 0.2189, "step": 2916 }, { "epoch": 0.61, "learning_rate": 2.0917774086378737e-07, "loss": 0.2352, "step": 2917 }, { "epoch": 0.61, "learning_rate": 2.0914659468438539e-07, "loss": 0.1988, "step": 2918 }, { "epoch": 0.61, "learning_rate": 2.091154485049834e-07, "loss": 0.1956, "step": 2919 }, { "epoch": 0.61, "learning_rate": 2.0908430232558136e-07, "loss": 0.2153, "step": 2920 }, { "epoch": 0.61, "learning_rate": 2.0905315614617938e-07, "loss": 0.1824, "step": 2921 }, { "epoch": 0.61, "learning_rate": 2.090220099667774e-07, "loss": 0.1725, "step": 2922 }, { "epoch": 0.61, "learning_rate": 2.089908637873754e-07, "loss": 0.1739, "step": 2923 }, { "epoch": 0.61, "learning_rate": 2.089597176079734e-07, "loss": 0.395, "step": 2924 }, { "epoch": 0.61, "learning_rate": 2.0892857142857142e-07, "loss": 0.2145, "step": 2925 }, { "epoch": 0.61, "learning_rate": 2.0889742524916943e-07, "loss": 0.1892, "step": 2926 }, { "epoch": 0.61, "learning_rate": 2.0886627906976745e-07, "loss": 0.1802, "step": 2927 }, { "epoch": 0.61, "learning_rate": 2.0883513289036544e-07, "loss": 0.1932, "step": 2928 }, { "epoch": 0.61, "learning_rate": 2.0880398671096343e-07, "loss": 0.1818, "step": 2929 }, { "epoch": 0.61, "learning_rate": 2.0877284053156144e-07, "loss": 0.1982, "step": 2930 }, { "epoch": 0.61, "learning_rate": 2.0874169435215946e-07, "loss": 0.1382, "step": 2931 }, { "epoch": 0.61, "learning_rate": 2.0871054817275748e-07, "loss": 0.1796, "step": 2932 }, { "epoch": 0.61, "learning_rate": 2.086794019933555e-07, "loss": 0.2359, "step": 2933 }, { "epoch": 0.61, "learning_rate": 2.0864825581395345e-07, "loss": 0.247, "step": 2934 }, { "epoch": 0.61, "learning_rate": 2.0861710963455147e-07, "loss": 0.2521, "step": 2935 }, { "epoch": 0.61, "learning_rate": 2.0858596345514949e-07, "loss": 0.1751, "step": 2936 }, { "epoch": 0.61, "learning_rate": 2.085548172757475e-07, "loss": 0.212, "step": 2937 }, { "epoch": 0.61, "learning_rate": 2.0852367109634552e-07, "loss": 0.1457, "step": 2938 }, { "epoch": 0.61, "learning_rate": 2.084925249169435e-07, "loss": 0.177, "step": 2939 }, { "epoch": 0.61, "learning_rate": 2.0846137873754152e-07, "loss": 0.2095, "step": 2940 }, { "epoch": 0.61, "learning_rate": 2.0843023255813954e-07, "loss": 0.2155, "step": 2941 }, { "epoch": 0.61, "learning_rate": 2.0839908637873755e-07, "loss": 0.1857, "step": 2942 }, { "epoch": 0.61, "learning_rate": 2.0836794019933552e-07, "loss": 0.137, "step": 2943 }, { "epoch": 0.61, "learning_rate": 2.0833679401993353e-07, "loss": 0.1676, "step": 2944 }, { "epoch": 0.61, "learning_rate": 2.0830564784053155e-07, "loss": 0.1961, "step": 2945 }, { "epoch": 0.61, "learning_rate": 2.0827450166112956e-07, "loss": 0.1905, "step": 2946 }, { "epoch": 0.61, "learning_rate": 2.0824335548172758e-07, "loss": 0.1306, "step": 2947 }, { "epoch": 0.61, "learning_rate": 2.0821220930232557e-07, "loss": 0.1983, "step": 2948 }, { "epoch": 0.61, "learning_rate": 2.0818106312292359e-07, "loss": 0.164, "step": 2949 }, { "epoch": 0.61, "learning_rate": 2.0814991694352157e-07, "loss": 0.1573, "step": 2950 }, { "epoch": 0.61, "learning_rate": 2.081187707641196e-07, "loss": 0.2572, "step": 2951 }, { "epoch": 0.61, "learning_rate": 2.080876245847176e-07, "loss": 0.183, "step": 2952 }, { "epoch": 0.61, "learning_rate": 2.080564784053156e-07, "loss": 0.1515, "step": 2953 }, { "epoch": 0.61, "learning_rate": 2.080253322259136e-07, "loss": 0.1961, "step": 2954 }, { "epoch": 0.61, "learning_rate": 2.0799418604651163e-07, "loss": 0.2445, "step": 2955 }, { "epoch": 0.61, "learning_rate": 2.0796303986710964e-07, "loss": 0.1809, "step": 2956 }, { "epoch": 0.61, "learning_rate": 2.079318936877076e-07, "loss": 0.1822, "step": 2957 }, { "epoch": 0.61, "learning_rate": 2.0790074750830562e-07, "loss": 0.2348, "step": 2958 }, { "epoch": 0.61, "learning_rate": 2.0786960132890364e-07, "loss": 0.196, "step": 2959 }, { "epoch": 0.61, "learning_rate": 2.0783845514950165e-07, "loss": 0.1775, "step": 2960 }, { "epoch": 0.61, "learning_rate": 2.0780730897009967e-07, "loss": 0.2125, "step": 2961 }, { "epoch": 0.62, "learning_rate": 2.0777616279069766e-07, "loss": 0.3099, "step": 2962 }, { "epoch": 0.62, "learning_rate": 2.0774501661129567e-07, "loss": 0.1474, "step": 2963 }, { "epoch": 0.62, "learning_rate": 2.077138704318937e-07, "loss": 0.1679, "step": 2964 }, { "epoch": 0.62, "learning_rate": 2.0768272425249168e-07, "loss": 0.188, "step": 2965 }, { "epoch": 0.62, "learning_rate": 2.076515780730897e-07, "loss": 0.1563, "step": 2966 }, { "epoch": 0.62, "learning_rate": 2.0762043189368768e-07, "loss": 0.1274, "step": 2967 }, { "epoch": 0.62, "learning_rate": 2.075892857142857e-07, "loss": 0.1694, "step": 2968 }, { "epoch": 0.62, "learning_rate": 2.0755813953488372e-07, "loss": 0.2043, "step": 2969 }, { "epoch": 0.62, "learning_rate": 2.0752699335548173e-07, "loss": 0.1936, "step": 2970 }, { "epoch": 0.62, "learning_rate": 2.0749584717607975e-07, "loss": 0.1767, "step": 2971 }, { "epoch": 0.62, "learning_rate": 2.074647009966777e-07, "loss": 0.2127, "step": 2972 }, { "epoch": 0.62, "learning_rate": 2.0743355481727573e-07, "loss": 0.1961, "step": 2973 }, { "epoch": 0.62, "learning_rate": 2.0740240863787374e-07, "loss": 0.1563, "step": 2974 }, { "epoch": 0.62, "learning_rate": 2.0737126245847176e-07, "loss": 0.1567, "step": 2975 }, { "epoch": 0.62, "learning_rate": 2.0734011627906975e-07, "loss": 0.1801, "step": 2976 }, { "epoch": 0.62, "learning_rate": 2.0730897009966776e-07, "loss": 0.2137, "step": 2977 }, { "epoch": 0.62, "learning_rate": 2.0727782392026578e-07, "loss": 0.2532, "step": 2978 }, { "epoch": 0.62, "learning_rate": 2.072466777408638e-07, "loss": 0.1803, "step": 2979 }, { "epoch": 0.62, "learning_rate": 2.072155315614618e-07, "loss": 0.1647, "step": 2980 }, { "epoch": 0.62, "learning_rate": 2.0718438538205977e-07, "loss": 0.2078, "step": 2981 }, { "epoch": 0.62, "learning_rate": 2.071532392026578e-07, "loss": 0.2821, "step": 2982 }, { "epoch": 0.62, "learning_rate": 2.071220930232558e-07, "loss": 0.1966, "step": 2983 }, { "epoch": 0.62, "learning_rate": 2.0709094684385382e-07, "loss": 0.1921, "step": 2984 }, { "epoch": 0.62, "learning_rate": 2.0705980066445184e-07, "loss": 0.2161, "step": 2985 }, { "epoch": 0.62, "learning_rate": 2.0702865448504983e-07, "loss": 0.2024, "step": 2986 }, { "epoch": 0.62, "learning_rate": 2.0699750830564782e-07, "loss": 0.2011, "step": 2987 }, { "epoch": 0.62, "learning_rate": 2.0696636212624583e-07, "loss": 0.1765, "step": 2988 }, { "epoch": 0.62, "learning_rate": 2.0693521594684385e-07, "loss": 0.1732, "step": 2989 }, { "epoch": 0.62, "learning_rate": 2.0690406976744184e-07, "loss": 0.2416, "step": 2990 }, { "epoch": 0.62, "learning_rate": 2.0687292358803985e-07, "loss": 0.2018, "step": 2991 }, { "epoch": 0.62, "learning_rate": 2.0684177740863787e-07, "loss": 0.2408, "step": 2992 }, { "epoch": 0.62, "learning_rate": 2.0681063122923588e-07, "loss": 0.14, "step": 2993 }, { "epoch": 0.62, "learning_rate": 2.067794850498339e-07, "loss": 0.1949, "step": 2994 }, { "epoch": 0.62, "learning_rate": 2.0674833887043186e-07, "loss": 0.2267, "step": 2995 }, { "epoch": 0.62, "learning_rate": 2.0671719269102988e-07, "loss": 0.1605, "step": 2996 }, { "epoch": 0.62, "learning_rate": 2.066860465116279e-07, "loss": 0.1681, "step": 2997 }, { "epoch": 0.62, "learning_rate": 2.066549003322259e-07, "loss": 0.2508, "step": 2998 }, { "epoch": 0.62, "learning_rate": 2.0662375415282393e-07, "loss": 0.2198, "step": 2999 }, { "epoch": 0.62, "learning_rate": 2.0659260797342192e-07, "loss": 0.1633, "step": 3000 }, { "epoch": 0.62, "learning_rate": 2.0656146179401993e-07, "loss": 0.1786, "step": 3001 }, { "epoch": 0.62, "learning_rate": 2.0653031561461792e-07, "loss": 0.216, "step": 3002 }, { "epoch": 0.62, "learning_rate": 2.0649916943521594e-07, "loss": 0.2179, "step": 3003 }, { "epoch": 0.62, "learning_rate": 2.0646802325581393e-07, "loss": 0.1841, "step": 3004 }, { "epoch": 0.62, "learning_rate": 2.0643687707641194e-07, "loss": 0.1493, "step": 3005 }, { "epoch": 0.62, "learning_rate": 2.0640573089700996e-07, "loss": 0.2382, "step": 3006 }, { "epoch": 0.62, "learning_rate": 2.0637458471760797e-07, "loss": 0.2087, "step": 3007 }, { "epoch": 0.62, "learning_rate": 2.06343438538206e-07, "loss": 0.1742, "step": 3008 }, { "epoch": 0.62, "learning_rate": 2.0631229235880395e-07, "loss": 0.2201, "step": 3009 }, { "epoch": 0.62, "learning_rate": 2.0628114617940197e-07, "loss": 0.2204, "step": 3010 }, { "epoch": 0.63, "learning_rate": 2.0624999999999998e-07, "loss": 0.2075, "step": 3011 }, { "epoch": 0.63, "learning_rate": 2.06218853820598e-07, "loss": 0.2046, "step": 3012 }, { "epoch": 0.63, "learning_rate": 2.0618770764119602e-07, "loss": 0.2184, "step": 3013 }, { "epoch": 0.63, "learning_rate": 2.06156561461794e-07, "loss": 0.1691, "step": 3014 }, { "epoch": 0.63, "learning_rate": 2.0612541528239202e-07, "loss": 0.2057, "step": 3015 }, { "epoch": 0.63, "learning_rate": 2.0609426910299004e-07, "loss": 0.2034, "step": 3016 }, { "epoch": 0.63, "learning_rate": 2.0606312292358805e-07, "loss": 0.2208, "step": 3017 }, { "epoch": 0.63, "learning_rate": 2.0603197674418604e-07, "loss": 0.2153, "step": 3018 }, { "epoch": 0.63, "learning_rate": 2.0600083056478403e-07, "loss": 0.1876, "step": 3019 }, { "epoch": 0.63, "learning_rate": 2.0596968438538205e-07, "loss": 0.2566, "step": 3020 }, { "epoch": 0.63, "learning_rate": 2.0593853820598006e-07, "loss": 0.2787, "step": 3021 }, { "epoch": 0.63, "learning_rate": 2.0590739202657808e-07, "loss": 0.1933, "step": 3022 }, { "epoch": 0.63, "learning_rate": 2.0587624584717607e-07, "loss": 0.2108, "step": 3023 }, { "epoch": 0.63, "learning_rate": 2.0584509966777406e-07, "loss": 0.1885, "step": 3024 }, { "epoch": 0.63, "learning_rate": 2.0581395348837207e-07, "loss": 0.2554, "step": 3025 }, { "epoch": 0.63, "learning_rate": 2.057828073089701e-07, "loss": 0.2036, "step": 3026 }, { "epoch": 0.63, "learning_rate": 2.057516611295681e-07, "loss": 0.186, "step": 3027 }, { "epoch": 0.63, "learning_rate": 2.057205149501661e-07, "loss": 0.1727, "step": 3028 }, { "epoch": 0.63, "learning_rate": 2.056893687707641e-07, "loss": 0.2128, "step": 3029 }, { "epoch": 0.63, "learning_rate": 2.0565822259136213e-07, "loss": 0.1567, "step": 3030 }, { "epoch": 0.63, "learning_rate": 2.0562707641196014e-07, "loss": 0.2199, "step": 3031 }, { "epoch": 0.63, "learning_rate": 2.0559593023255816e-07, "loss": 0.1992, "step": 3032 }, { "epoch": 0.63, "learning_rate": 2.0556478405315612e-07, "loss": 0.1708, "step": 3033 }, { "epoch": 0.63, "learning_rate": 2.0553363787375414e-07, "loss": 0.2037, "step": 3034 }, { "epoch": 0.63, "learning_rate": 2.0550249169435215e-07, "loss": 0.1735, "step": 3035 }, { "epoch": 0.63, "learning_rate": 2.0547134551495017e-07, "loss": 0.189, "step": 3036 }, { "epoch": 0.63, "learning_rate": 2.0544019933554816e-07, "loss": 0.2112, "step": 3037 }, { "epoch": 0.63, "learning_rate": 2.0540905315614617e-07, "loss": 0.1781, "step": 3038 }, { "epoch": 0.63, "learning_rate": 2.0537790697674416e-07, "loss": 0.2039, "step": 3039 }, { "epoch": 0.63, "learning_rate": 2.0534676079734218e-07, "loss": 0.1829, "step": 3040 }, { "epoch": 0.63, "learning_rate": 2.053156146179402e-07, "loss": 0.146, "step": 3041 }, { "epoch": 0.63, "learning_rate": 2.0528446843853818e-07, "loss": 0.2008, "step": 3042 }, { "epoch": 0.63, "learning_rate": 2.052533222591362e-07, "loss": 0.1916, "step": 3043 }, { "epoch": 0.63, "learning_rate": 2.0522217607973422e-07, "loss": 0.1364, "step": 3044 }, { "epoch": 0.63, "learning_rate": 2.0519102990033223e-07, "loss": 0.2349, "step": 3045 }, { "epoch": 0.63, "learning_rate": 2.0515988372093025e-07, "loss": 0.1693, "step": 3046 }, { "epoch": 0.63, "learning_rate": 2.051287375415282e-07, "loss": 0.1994, "step": 3047 }, { "epoch": 0.63, "learning_rate": 2.0509759136212623e-07, "loss": 0.1844, "step": 3048 }, { "epoch": 0.63, "learning_rate": 2.0506644518272424e-07, "loss": 0.1994, "step": 3049 }, { "epoch": 0.63, "learning_rate": 2.0503529900332226e-07, "loss": 0.1805, "step": 3050 }, { "epoch": 0.63, "learning_rate": 2.0500415282392025e-07, "loss": 0.1857, "step": 3051 }, { "epoch": 0.63, "learning_rate": 2.0497300664451826e-07, "loss": 0.1232, "step": 3052 }, { "epoch": 0.63, "learning_rate": 2.0494186046511628e-07, "loss": 0.2038, "step": 3053 }, { "epoch": 0.63, "learning_rate": 2.049107142857143e-07, "loss": 0.2004, "step": 3054 }, { "epoch": 0.63, "learning_rate": 2.0487956810631228e-07, "loss": 0.1266, "step": 3055 }, { "epoch": 0.63, "learning_rate": 2.0484842192691027e-07, "loss": 0.1855, "step": 3056 }, { "epoch": 0.63, "learning_rate": 2.048172757475083e-07, "loss": 0.2012, "step": 3057 }, { "epoch": 0.63, "learning_rate": 2.047861295681063e-07, "loss": 0.2119, "step": 3058 }, { "epoch": 0.64, "learning_rate": 2.0475498338870432e-07, "loss": 0.2071, "step": 3059 }, { "epoch": 0.64, "learning_rate": 2.0472383720930234e-07, "loss": 0.1946, "step": 3060 }, { "epoch": 0.64, "learning_rate": 2.046926910299003e-07, "loss": 0.1678, "step": 3061 }, { "epoch": 0.64, "learning_rate": 2.0466154485049831e-07, "loss": 0.1852, "step": 3062 }, { "epoch": 0.64, "learning_rate": 2.0463039867109633e-07, "loss": 0.175, "step": 3063 }, { "epoch": 0.64, "learning_rate": 2.0459925249169435e-07, "loss": 0.1776, "step": 3064 }, { "epoch": 0.64, "learning_rate": 2.0456810631229236e-07, "loss": 0.1668, "step": 3065 }, { "epoch": 0.64, "learning_rate": 2.0453696013289035e-07, "loss": 0.2045, "step": 3066 }, { "epoch": 0.64, "learning_rate": 2.0450581395348837e-07, "loss": 0.1616, "step": 3067 }, { "epoch": 0.64, "learning_rate": 2.0447466777408638e-07, "loss": 0.2087, "step": 3068 }, { "epoch": 0.64, "learning_rate": 2.044435215946844e-07, "loss": 0.2158, "step": 3069 }, { "epoch": 0.64, "learning_rate": 2.0441237541528236e-07, "loss": 0.2115, "step": 3070 }, { "epoch": 0.64, "learning_rate": 2.0438122923588038e-07, "loss": 0.1633, "step": 3071 }, { "epoch": 0.64, "learning_rate": 2.043500830564784e-07, "loss": 0.155, "step": 3072 }, { "epoch": 0.64, "learning_rate": 2.043189368770764e-07, "loss": 0.1756, "step": 3073 }, { "epoch": 0.64, "learning_rate": 2.0428779069767443e-07, "loss": 0.1694, "step": 3074 }, { "epoch": 0.64, "learning_rate": 2.0425664451827241e-07, "loss": 0.1629, "step": 3075 }, { "epoch": 0.64, "learning_rate": 2.042254983388704e-07, "loss": 0.2216, "step": 3076 }, { "epoch": 0.64, "learning_rate": 2.0419435215946842e-07, "loss": 0.1993, "step": 3077 }, { "epoch": 0.64, "learning_rate": 2.0416320598006644e-07, "loss": 0.1993, "step": 3078 }, { "epoch": 0.64, "learning_rate": 2.0413205980066445e-07, "loss": 0.2025, "step": 3079 }, { "epoch": 0.64, "learning_rate": 2.0410091362126244e-07, "loss": 0.2227, "step": 3080 }, { "epoch": 0.64, "learning_rate": 2.0406976744186046e-07, "loss": 0.1979, "step": 3081 }, { "epoch": 0.64, "learning_rate": 2.0403862126245847e-07, "loss": 0.236, "step": 3082 }, { "epoch": 0.64, "learning_rate": 2.040074750830565e-07, "loss": 0.2296, "step": 3083 }, { "epoch": 0.64, "learning_rate": 2.0397632890365445e-07, "loss": 0.1524, "step": 3084 }, { "epoch": 0.64, "learning_rate": 2.0394518272425247e-07, "loss": 0.223, "step": 3085 }, { "epoch": 0.64, "learning_rate": 2.0391403654485048e-07, "loss": 0.1949, "step": 3086 }, { "epoch": 0.64, "learning_rate": 2.038828903654485e-07, "loss": 0.155, "step": 3087 }, { "epoch": 0.64, "learning_rate": 2.0385174418604651e-07, "loss": 0.1996, "step": 3088 }, { "epoch": 0.64, "learning_rate": 2.038205980066445e-07, "loss": 0.2025, "step": 3089 }, { "epoch": 0.64, "learning_rate": 2.0378945182724252e-07, "loss": 0.2347, "step": 3090 }, { "epoch": 0.64, "learning_rate": 2.0375830564784054e-07, "loss": 0.2016, "step": 3091 }, { "epoch": 0.64, "learning_rate": 2.0372715946843852e-07, "loss": 0.2037, "step": 3092 }, { "epoch": 0.64, "learning_rate": 2.0369601328903654e-07, "loss": 0.3015, "step": 3093 }, { "epoch": 0.64, "learning_rate": 2.0366486710963453e-07, "loss": 0.2032, "step": 3094 }, { "epoch": 0.64, "learning_rate": 2.0363372093023255e-07, "loss": 0.171, "step": 3095 }, { "epoch": 0.64, "learning_rate": 2.0360257475083056e-07, "loss": 0.1924, "step": 3096 }, { "epoch": 0.64, "learning_rate": 2.0357142857142858e-07, "loss": 0.2381, "step": 3097 }, { "epoch": 0.64, "learning_rate": 2.035402823920266e-07, "loss": 0.1923, "step": 3098 }, { "epoch": 0.64, "learning_rate": 2.0350913621262456e-07, "loss": 0.2206, "step": 3099 }, { "epoch": 0.64, "learning_rate": 2.0347799003322257e-07, "loss": 0.2016, "step": 3100 }, { "epoch": 0.64, "learning_rate": 2.034468438538206e-07, "loss": 0.2018, "step": 3101 }, { "epoch": 0.64, "learning_rate": 2.034156976744186e-07, "loss": 0.211, "step": 3102 }, { "epoch": 0.64, "learning_rate": 2.033845514950166e-07, "loss": 0.2448, "step": 3103 }, { "epoch": 0.64, "learning_rate": 2.033534053156146e-07, "loss": 0.1777, "step": 3104 }, { "epoch": 0.64, "learning_rate": 2.0332225913621262e-07, "loss": 0.1856, "step": 3105 }, { "epoch": 0.64, "learning_rate": 2.0329111295681064e-07, "loss": 0.2468, "step": 3106 }, { "epoch": 0.65, "learning_rate": 2.0325996677740863e-07, "loss": 0.2471, "step": 3107 }, { "epoch": 0.65, "learning_rate": 2.0322882059800662e-07, "loss": 0.1918, "step": 3108 }, { "epoch": 0.65, "learning_rate": 2.0319767441860463e-07, "loss": 0.2527, "step": 3109 }, { "epoch": 0.65, "learning_rate": 2.0316652823920265e-07, "loss": 0.2361, "step": 3110 }, { "epoch": 0.65, "learning_rate": 2.0313538205980067e-07, "loss": 0.2148, "step": 3111 }, { "epoch": 0.65, "learning_rate": 2.0310423588039868e-07, "loss": 0.202, "step": 3112 }, { "epoch": 0.65, "learning_rate": 2.0307308970099665e-07, "loss": 0.2332, "step": 3113 }, { "epoch": 0.65, "learning_rate": 2.0304194352159466e-07, "loss": 0.1786, "step": 3114 }, { "epoch": 0.65, "learning_rate": 2.0301079734219268e-07, "loss": 0.1659, "step": 3115 }, { "epoch": 0.65, "learning_rate": 2.029796511627907e-07, "loss": 0.2019, "step": 3116 }, { "epoch": 0.65, "learning_rate": 2.0294850498338868e-07, "loss": 0.1835, "step": 3117 }, { "epoch": 0.65, "learning_rate": 2.029173588039867e-07, "loss": 0.2249, "step": 3118 }, { "epoch": 0.65, "learning_rate": 2.0288621262458471e-07, "loss": 0.2241, "step": 3119 }, { "epoch": 0.65, "learning_rate": 2.0285506644518273e-07, "loss": 0.2243, "step": 3120 }, { "epoch": 0.65, "learning_rate": 2.0282392026578075e-07, "loss": 0.1785, "step": 3121 }, { "epoch": 0.65, "learning_rate": 2.027927740863787e-07, "loss": 0.2133, "step": 3122 }, { "epoch": 0.65, "learning_rate": 2.0276162790697672e-07, "loss": 0.205, "step": 3123 }, { "epoch": 0.65, "learning_rate": 2.0273048172757474e-07, "loss": 0.1433, "step": 3124 }, { "epoch": 0.65, "learning_rate": 2.0269933554817276e-07, "loss": 0.1755, "step": 3125 }, { "epoch": 0.65, "learning_rate": 2.0266818936877077e-07, "loss": 0.1522, "step": 3126 }, { "epoch": 0.65, "learning_rate": 2.0263704318936876e-07, "loss": 0.2056, "step": 3127 }, { "epoch": 0.65, "learning_rate": 2.0260589700996678e-07, "loss": 0.1855, "step": 3128 }, { "epoch": 0.65, "learning_rate": 2.0257475083056477e-07, "loss": 0.1786, "step": 3129 }, { "epoch": 0.65, "learning_rate": 2.0254360465116278e-07, "loss": 0.1841, "step": 3130 }, { "epoch": 0.65, "learning_rate": 2.0251245847176077e-07, "loss": 0.1783, "step": 3131 }, { "epoch": 0.65, "learning_rate": 2.024813122923588e-07, "loss": 0.2074, "step": 3132 }, { "epoch": 0.65, "learning_rate": 2.024501661129568e-07, "loss": 0.1254, "step": 3133 }, { "epoch": 0.65, "learning_rate": 2.0241901993355482e-07, "loss": 0.1626, "step": 3134 }, { "epoch": 0.65, "learning_rate": 2.0238787375415283e-07, "loss": 0.228, "step": 3135 }, { "epoch": 0.65, "learning_rate": 2.023567275747508e-07, "loss": 0.1926, "step": 3136 }, { "epoch": 0.65, "learning_rate": 2.0232558139534881e-07, "loss": 0.2111, "step": 3137 }, { "epoch": 0.65, "learning_rate": 2.0229443521594683e-07, "loss": 0.2419, "step": 3138 }, { "epoch": 0.65, "learning_rate": 2.0226328903654484e-07, "loss": 0.1691, "step": 3139 }, { "epoch": 0.65, "learning_rate": 2.0223214285714286e-07, "loss": 0.1538, "step": 3140 }, { "epoch": 0.65, "learning_rate": 2.0220099667774085e-07, "loss": 0.2604, "step": 3141 }, { "epoch": 0.65, "learning_rate": 2.0216985049833887e-07, "loss": 0.2637, "step": 3142 }, { "epoch": 0.65, "learning_rate": 2.0213870431893688e-07, "loss": 0.2149, "step": 3143 }, { "epoch": 0.65, "learning_rate": 2.0210755813953487e-07, "loss": 0.2189, "step": 3144 }, { "epoch": 0.65, "learning_rate": 2.020764119601329e-07, "loss": 0.2223, "step": 3145 }, { "epoch": 0.65, "learning_rate": 2.0204526578073088e-07, "loss": 0.2284, "step": 3146 }, { "epoch": 0.65, "learning_rate": 2.020141196013289e-07, "loss": 0.2034, "step": 3147 }, { "epoch": 0.65, "learning_rate": 2.019829734219269e-07, "loss": 0.1914, "step": 3148 }, { "epoch": 0.65, "learning_rate": 2.0195182724252492e-07, "loss": 0.2347, "step": 3149 }, { "epoch": 0.65, "learning_rate": 2.0192068106312289e-07, "loss": 0.2499, "step": 3150 }, { "epoch": 0.65, "learning_rate": 2.018895348837209e-07, "loss": 0.2202, "step": 3151 }, { "epoch": 0.65, "learning_rate": 2.0185838870431892e-07, "loss": 0.1642, "step": 3152 }, { "epoch": 0.65, "learning_rate": 2.0182724252491693e-07, "loss": 0.2045, "step": 3153 }, { "epoch": 0.65, "learning_rate": 2.0179609634551495e-07, "loss": 0.1708, "step": 3154 }, { "epoch": 0.66, "learning_rate": 2.0176495016611294e-07, "loss": 0.1318, "step": 3155 }, { "epoch": 0.66, "learning_rate": 2.0173380398671096e-07, "loss": 0.1569, "step": 3156 }, { "epoch": 0.66, "learning_rate": 2.0170265780730897e-07, "loss": 0.167, "step": 3157 }, { "epoch": 0.66, "learning_rate": 2.01671511627907e-07, "loss": 0.1752, "step": 3158 }, { "epoch": 0.66, "learning_rate": 2.01640365448505e-07, "loss": 0.1836, "step": 3159 }, { "epoch": 0.66, "learning_rate": 2.0160921926910297e-07, "loss": 0.1816, "step": 3160 }, { "epoch": 0.66, "learning_rate": 2.0157807308970098e-07, "loss": 0.1555, "step": 3161 }, { "epoch": 0.66, "learning_rate": 2.01546926910299e-07, "loss": 0.1787, "step": 3162 }, { "epoch": 0.66, "learning_rate": 2.01515780730897e-07, "loss": 0.1208, "step": 3163 }, { "epoch": 0.66, "learning_rate": 2.01484634551495e-07, "loss": 0.1978, "step": 3164 }, { "epoch": 0.66, "learning_rate": 2.0145348837209302e-07, "loss": 0.1294, "step": 3165 }, { "epoch": 0.66, "learning_rate": 2.01422342192691e-07, "loss": 0.2279, "step": 3166 }, { "epoch": 0.66, "learning_rate": 2.0139119601328902e-07, "loss": 0.1797, "step": 3167 }, { "epoch": 0.66, "learning_rate": 2.0136004983388704e-07, "loss": 0.1911, "step": 3168 }, { "epoch": 0.66, "learning_rate": 2.0132890365448503e-07, "loss": 0.2525, "step": 3169 }, { "epoch": 0.66, "learning_rate": 2.0129775747508304e-07, "loss": 0.2178, "step": 3170 }, { "epoch": 0.66, "learning_rate": 2.0126661129568106e-07, "loss": 0.1838, "step": 3171 }, { "epoch": 0.66, "learning_rate": 2.0123546511627908e-07, "loss": 0.2002, "step": 3172 }, { "epoch": 0.66, "learning_rate": 2.012043189368771e-07, "loss": 0.1597, "step": 3173 }, { "epoch": 0.66, "learning_rate": 2.0117317275747505e-07, "loss": 0.2134, "step": 3174 }, { "epoch": 0.66, "learning_rate": 2.0114202657807307e-07, "loss": 0.1572, "step": 3175 }, { "epoch": 0.66, "learning_rate": 2.0111088039867109e-07, "loss": 0.1999, "step": 3176 }, { "epoch": 0.66, "learning_rate": 2.010797342192691e-07, "loss": 0.1657, "step": 3177 }, { "epoch": 0.66, "learning_rate": 2.010485880398671e-07, "loss": 0.1773, "step": 3178 }, { "epoch": 0.66, "learning_rate": 2.010174418604651e-07, "loss": 0.2183, "step": 3179 }, { "epoch": 0.66, "learning_rate": 2.0098629568106312e-07, "loss": 0.1928, "step": 3180 }, { "epoch": 0.66, "learning_rate": 2.009551495016611e-07, "loss": 0.2128, "step": 3181 }, { "epoch": 0.66, "learning_rate": 2.0092400332225913e-07, "loss": 0.1944, "step": 3182 }, { "epoch": 0.66, "learning_rate": 2.0089285714285712e-07, "loss": 0.2008, "step": 3183 }, { "epoch": 0.66, "learning_rate": 2.0086171096345513e-07, "loss": 0.1482, "step": 3184 }, { "epoch": 0.66, "learning_rate": 2.0083056478405315e-07, "loss": 0.1623, "step": 3185 }, { "epoch": 0.66, "learning_rate": 2.0079941860465117e-07, "loss": 0.1628, "step": 3186 }, { "epoch": 0.66, "learning_rate": 2.0076827242524918e-07, "loss": 0.1618, "step": 3187 }, { "epoch": 0.66, "learning_rate": 2.0073712624584714e-07, "loss": 0.1564, "step": 3188 }, { "epoch": 0.66, "learning_rate": 2.0070598006644516e-07, "loss": 0.1948, "step": 3189 }, { "epoch": 0.66, "learning_rate": 2.0067483388704318e-07, "loss": 0.1998, "step": 3190 }, { "epoch": 0.66, "learning_rate": 2.006436877076412e-07, "loss": 0.197, "step": 3191 }, { "epoch": 0.66, "learning_rate": 2.006125415282392e-07, "loss": 0.3115, "step": 3192 }, { "epoch": 0.66, "learning_rate": 2.005813953488372e-07, "loss": 0.209, "step": 3193 }, { "epoch": 0.66, "learning_rate": 2.005502491694352e-07, "loss": 0.1957, "step": 3194 }, { "epoch": 0.66, "learning_rate": 2.0051910299003323e-07, "loss": 0.1973, "step": 3195 }, { "epoch": 0.66, "learning_rate": 2.0048795681063124e-07, "loss": 0.1667, "step": 3196 }, { "epoch": 0.66, "learning_rate": 2.004568106312292e-07, "loss": 0.1504, "step": 3197 }, { "epoch": 0.66, "learning_rate": 2.0042566445182722e-07, "loss": 0.1996, "step": 3198 }, { "epoch": 0.66, "learning_rate": 2.0039451827242524e-07, "loss": 0.235, "step": 3199 }, { "epoch": 0.66, "learning_rate": 2.0036337209302325e-07, "loss": 0.2211, "step": 3200 }, { "epoch": 0.66, "learning_rate": 2.0033222591362127e-07, "loss": 0.1975, "step": 3201 }, { "epoch": 0.66, "learning_rate": 2.0030107973421926e-07, "loss": 0.2001, "step": 3202 }, { "epoch": 0.67, "learning_rate": 2.0026993355481725e-07, "loss": 0.27, "step": 3203 }, { "epoch": 0.67, "learning_rate": 2.0023878737541526e-07, "loss": 0.2447, "step": 3204 }, { "epoch": 0.67, "learning_rate": 2.0020764119601328e-07, "loss": 0.2112, "step": 3205 }, { "epoch": 0.67, "learning_rate": 2.001764950166113e-07, "loss": 0.2137, "step": 3206 }, { "epoch": 0.67, "learning_rate": 2.0014534883720929e-07, "loss": 0.2197, "step": 3207 }, { "epoch": 0.67, "learning_rate": 2.001142026578073e-07, "loss": 0.1826, "step": 3208 }, { "epoch": 0.67, "learning_rate": 2.0008305647840532e-07, "loss": 0.2026, "step": 3209 }, { "epoch": 0.67, "learning_rate": 2.0005191029900333e-07, "loss": 0.1971, "step": 3210 }, { "epoch": 0.67, "learning_rate": 2.000207641196013e-07, "loss": 0.2244, "step": 3211 }, { "epoch": 0.67, "learning_rate": 1.999896179401993e-07, "loss": 0.1746, "step": 3212 }, { "epoch": 0.67, "learning_rate": 1.9995847176079733e-07, "loss": 0.2107, "step": 3213 }, { "epoch": 0.67, "learning_rate": 1.9992732558139534e-07, "loss": 0.2253, "step": 3214 }, { "epoch": 0.67, "learning_rate": 1.9989617940199336e-07, "loss": 0.2211, "step": 3215 }, { "epoch": 0.67, "learning_rate": 1.9986503322259135e-07, "loss": 0.2559, "step": 3216 }, { "epoch": 0.67, "learning_rate": 1.9983388704318936e-07, "loss": 0.1565, "step": 3217 }, { "epoch": 0.67, "learning_rate": 1.9980274086378735e-07, "loss": 0.1766, "step": 3218 }, { "epoch": 0.67, "learning_rate": 1.9977159468438537e-07, "loss": 0.2104, "step": 3219 }, { "epoch": 0.67, "learning_rate": 1.9974044850498339e-07, "loss": 0.2138, "step": 3220 }, { "epoch": 0.67, "learning_rate": 1.9970930232558137e-07, "loss": 0.2571, "step": 3221 }, { "epoch": 0.67, "learning_rate": 1.996781561461794e-07, "loss": 0.2916, "step": 3222 }, { "epoch": 0.67, "learning_rate": 1.996470099667774e-07, "loss": 0.1806, "step": 3223 }, { "epoch": 0.67, "learning_rate": 1.9961586378737542e-07, "loss": 0.1851, "step": 3224 }, { "epoch": 0.67, "learning_rate": 1.9958471760797339e-07, "loss": 0.2076, "step": 3225 }, { "epoch": 0.67, "learning_rate": 1.995535714285714e-07, "loss": 0.1507, "step": 3226 }, { "epoch": 0.67, "learning_rate": 1.9952242524916942e-07, "loss": 0.1474, "step": 3227 }, { "epoch": 0.67, "learning_rate": 1.9949127906976743e-07, "loss": 0.2003, "step": 3228 }, { "epoch": 0.67, "learning_rate": 1.9946013289036545e-07, "loss": 0.115, "step": 3229 }, { "epoch": 0.67, "learning_rate": 1.9942898671096344e-07, "loss": 0.1753, "step": 3230 }, { "epoch": 0.67, "learning_rate": 1.9939784053156145e-07, "loss": 0.1987, "step": 3231 }, { "epoch": 0.67, "learning_rate": 1.9936669435215947e-07, "loss": 0.1451, "step": 3232 }, { "epoch": 0.67, "learning_rate": 1.9933554817275749e-07, "loss": 0.1983, "step": 3233 }, { "epoch": 0.67, "learning_rate": 1.9930440199335547e-07, "loss": 0.1669, "step": 3234 }, { "epoch": 0.67, "learning_rate": 1.9927325581395346e-07, "loss": 0.1684, "step": 3235 }, { "epoch": 0.67, "learning_rate": 1.9924210963455148e-07, "loss": 0.1582, "step": 3236 }, { "epoch": 0.67, "learning_rate": 1.992109634551495e-07, "loss": 0.1276, "step": 3237 }, { "epoch": 0.67, "learning_rate": 1.991798172757475e-07, "loss": 0.2038, "step": 3238 }, { "epoch": 0.67, "learning_rate": 1.9914867109634553e-07, "loss": 0.1513, "step": 3239 }, { "epoch": 0.67, "learning_rate": 1.991175249169435e-07, "loss": 0.1184, "step": 3240 }, { "epoch": 0.67, "learning_rate": 1.990863787375415e-07, "loss": 0.2399, "step": 3241 }, { "epoch": 0.67, "learning_rate": 1.9905523255813952e-07, "loss": 0.1973, "step": 3242 }, { "epoch": 0.67, "learning_rate": 1.9902408637873754e-07, "loss": 0.1838, "step": 3243 }, { "epoch": 0.67, "learning_rate": 1.9899294019933553e-07, "loss": 0.222, "step": 3244 }, { "epoch": 0.67, "learning_rate": 1.9896179401993354e-07, "loss": 0.1816, "step": 3245 }, { "epoch": 0.67, "learning_rate": 1.9893064784053156e-07, "loss": 0.1801, "step": 3246 }, { "epoch": 0.67, "learning_rate": 1.9889950166112957e-07, "loss": 0.1606, "step": 3247 }, { "epoch": 0.67, "learning_rate": 1.988683554817276e-07, "loss": 0.1962, "step": 3248 }, { "epoch": 0.67, "learning_rate": 1.9883720930232555e-07, "loss": 0.1584, "step": 3249 }, { "epoch": 0.67, "learning_rate": 1.9880606312292357e-07, "loss": 0.1301, "step": 3250 }, { "epoch": 0.68, "learning_rate": 1.9877491694352158e-07, "loss": 0.2097, "step": 3251 }, { "epoch": 0.68, "learning_rate": 1.987437707641196e-07, "loss": 0.1497, "step": 3252 }, { "epoch": 0.68, "learning_rate": 1.9871262458471762e-07, "loss": 0.1839, "step": 3253 }, { "epoch": 0.68, "learning_rate": 1.986814784053156e-07, "loss": 0.184, "step": 3254 }, { "epoch": 0.68, "learning_rate": 1.986503322259136e-07, "loss": 0.1928, "step": 3255 }, { "epoch": 0.68, "learning_rate": 1.986191860465116e-07, "loss": 0.1798, "step": 3256 }, { "epoch": 0.68, "learning_rate": 1.9858803986710963e-07, "loss": 0.2313, "step": 3257 }, { "epoch": 0.68, "learning_rate": 1.9855689368770762e-07, "loss": 0.2404, "step": 3258 }, { "epoch": 0.68, "learning_rate": 1.9852574750830563e-07, "loss": 0.1685, "step": 3259 }, { "epoch": 0.68, "learning_rate": 1.9849460132890365e-07, "loss": 0.1566, "step": 3260 }, { "epoch": 0.68, "learning_rate": 1.9846345514950166e-07, "loss": 0.2189, "step": 3261 }, { "epoch": 0.68, "learning_rate": 1.9843230897009968e-07, "loss": 0.2093, "step": 3262 }, { "epoch": 0.68, "learning_rate": 1.9840116279069764e-07, "loss": 0.1756, "step": 3263 }, { "epoch": 0.68, "learning_rate": 1.9837001661129566e-07, "loss": 0.173, "step": 3264 }, { "epoch": 0.68, "learning_rate": 1.9833887043189367e-07, "loss": 0.1386, "step": 3265 }, { "epoch": 0.68, "learning_rate": 1.983077242524917e-07, "loss": 0.1594, "step": 3266 }, { "epoch": 0.68, "learning_rate": 1.982765780730897e-07, "loss": 0.2007, "step": 3267 }, { "epoch": 0.68, "learning_rate": 1.982454318936877e-07, "loss": 0.1921, "step": 3268 }, { "epoch": 0.68, "learning_rate": 1.982142857142857e-07, "loss": 0.1656, "step": 3269 }, { "epoch": 0.68, "learning_rate": 1.9818313953488373e-07, "loss": 0.3152, "step": 3270 }, { "epoch": 0.68, "learning_rate": 1.9815199335548172e-07, "loss": 0.1779, "step": 3271 }, { "epoch": 0.68, "learning_rate": 1.9812084717607973e-07, "loss": 0.2712, "step": 3272 }, { "epoch": 0.68, "learning_rate": 1.9808970099667772e-07, "loss": 0.1597, "step": 3273 }, { "epoch": 0.68, "learning_rate": 1.9805855481727574e-07, "loss": 0.2388, "step": 3274 }, { "epoch": 0.68, "learning_rate": 1.9802740863787375e-07, "loss": 0.1657, "step": 3275 }, { "epoch": 0.68, "learning_rate": 1.9799626245847177e-07, "loss": 0.169, "step": 3276 }, { "epoch": 0.68, "learning_rate": 1.9796511627906973e-07, "loss": 0.1598, "step": 3277 }, { "epoch": 0.68, "learning_rate": 1.9793397009966775e-07, "loss": 0.2226, "step": 3278 }, { "epoch": 0.68, "learning_rate": 1.9790282392026576e-07, "loss": 0.2108, "step": 3279 }, { "epoch": 0.68, "learning_rate": 1.9787167774086378e-07, "loss": 0.1544, "step": 3280 }, { "epoch": 0.68, "learning_rate": 1.978405315614618e-07, "loss": 0.1618, "step": 3281 }, { "epoch": 0.68, "learning_rate": 1.9780938538205978e-07, "loss": 0.1953, "step": 3282 }, { "epoch": 0.68, "learning_rate": 1.977782392026578e-07, "loss": 0.2459, "step": 3283 }, { "epoch": 0.68, "learning_rate": 1.9774709302325582e-07, "loss": 0.1896, "step": 3284 }, { "epoch": 0.68, "learning_rate": 1.9771594684385383e-07, "loss": 0.1974, "step": 3285 }, { "epoch": 0.68, "learning_rate": 1.9768480066445182e-07, "loss": 0.2211, "step": 3286 }, { "epoch": 0.68, "learning_rate": 1.976536544850498e-07, "loss": 0.2168, "step": 3287 }, { "epoch": 0.68, "learning_rate": 1.9762250830564783e-07, "loss": 0.2073, "step": 3288 }, { "epoch": 0.68, "learning_rate": 1.9759136212624584e-07, "loss": 0.2064, "step": 3289 }, { "epoch": 0.68, "learning_rate": 1.9756021594684386e-07, "loss": 0.1606, "step": 3290 }, { "epoch": 0.68, "learning_rate": 1.9752906976744185e-07, "loss": 0.1566, "step": 3291 }, { "epoch": 0.68, "learning_rate": 1.9749792358803984e-07, "loss": 0.2087, "step": 3292 }, { "epoch": 0.68, "learning_rate": 1.9746677740863785e-07, "loss": 0.191, "step": 3293 }, { "epoch": 0.68, "learning_rate": 1.9743563122923587e-07, "loss": 0.201, "step": 3294 }, { "epoch": 0.68, "learning_rate": 1.9740448504983388e-07, "loss": 0.2077, "step": 3295 }, { "epoch": 0.68, "learning_rate": 1.9737333887043187e-07, "loss": 0.2108, "step": 3296 }, { "epoch": 0.68, "learning_rate": 1.973421926910299e-07, "loss": 0.2118, "step": 3297 }, { "epoch": 0.68, "learning_rate": 1.973110465116279e-07, "loss": 0.1775, "step": 3298 }, { "epoch": 0.69, "learning_rate": 1.9727990033222592e-07, "loss": 0.2111, "step": 3299 }, { "epoch": 0.69, "learning_rate": 1.9724875415282394e-07, "loss": 0.1918, "step": 3300 }, { "epoch": 0.69, "learning_rate": 1.972176079734219e-07, "loss": 0.1774, "step": 3301 }, { "epoch": 0.69, "learning_rate": 1.9718646179401992e-07, "loss": 0.1978, "step": 3302 }, { "epoch": 0.69, "learning_rate": 1.9715531561461793e-07, "loss": 0.2103, "step": 3303 }, { "epoch": 0.69, "learning_rate": 1.9712416943521595e-07, "loss": 0.194, "step": 3304 }, { "epoch": 0.69, "learning_rate": 1.9709302325581394e-07, "loss": 0.212, "step": 3305 }, { "epoch": 0.69, "learning_rate": 1.9706187707641195e-07, "loss": 0.2069, "step": 3306 }, { "epoch": 0.69, "learning_rate": 1.9703073089700997e-07, "loss": 0.1813, "step": 3307 }, { "epoch": 0.69, "learning_rate": 1.9699958471760796e-07, "loss": 0.1951, "step": 3308 }, { "epoch": 0.69, "learning_rate": 1.9696843853820597e-07, "loss": 0.17, "step": 3309 }, { "epoch": 0.69, "learning_rate": 1.9693729235880396e-07, "loss": 0.2247, "step": 3310 }, { "epoch": 0.69, "learning_rate": 1.9690614617940198e-07, "loss": 0.2132, "step": 3311 }, { "epoch": 0.69, "learning_rate": 1.96875e-07, "loss": 0.1907, "step": 3312 }, { "epoch": 0.69, "learning_rate": 1.96843853820598e-07, "loss": 0.2055, "step": 3313 }, { "epoch": 0.69, "learning_rate": 1.9681270764119603e-07, "loss": 0.165, "step": 3314 }, { "epoch": 0.69, "learning_rate": 1.96781561461794e-07, "loss": 0.2101, "step": 3315 }, { "epoch": 0.69, "learning_rate": 1.96750415282392e-07, "loss": 0.2184, "step": 3316 }, { "epoch": 0.69, "learning_rate": 1.9671926910299002e-07, "loss": 0.2525, "step": 3317 }, { "epoch": 0.69, "learning_rate": 1.9668812292358804e-07, "loss": 0.2272, "step": 3318 }, { "epoch": 0.69, "learning_rate": 1.9665697674418605e-07, "loss": 0.139, "step": 3319 }, { "epoch": 0.69, "learning_rate": 1.9662583056478404e-07, "loss": 0.1727, "step": 3320 }, { "epoch": 0.69, "learning_rate": 1.9659468438538206e-07, "loss": 0.2328, "step": 3321 }, { "epoch": 0.69, "learning_rate": 1.9656353820598007e-07, "loss": 0.2052, "step": 3322 }, { "epoch": 0.69, "learning_rate": 1.9653239202657806e-07, "loss": 0.2094, "step": 3323 }, { "epoch": 0.69, "learning_rate": 1.9650124584717605e-07, "loss": 0.1745, "step": 3324 }, { "epoch": 0.69, "learning_rate": 1.9647009966777407e-07, "loss": 0.183, "step": 3325 }, { "epoch": 0.69, "learning_rate": 1.9643895348837208e-07, "loss": 0.144, "step": 3326 }, { "epoch": 0.69, "learning_rate": 1.964078073089701e-07, "loss": 0.2702, "step": 3327 }, { "epoch": 0.69, "learning_rate": 1.9637666112956812e-07, "loss": 0.2184, "step": 3328 }, { "epoch": 0.69, "learning_rate": 1.9634551495016608e-07, "loss": 0.2044, "step": 3329 }, { "epoch": 0.69, "learning_rate": 1.963143687707641e-07, "loss": 0.1675, "step": 3330 }, { "epoch": 0.69, "learning_rate": 1.962832225913621e-07, "loss": 0.1652, "step": 3331 }, { "epoch": 0.69, "learning_rate": 1.9625207641196013e-07, "loss": 0.1894, "step": 3332 }, { "epoch": 0.69, "learning_rate": 1.9622093023255814e-07, "loss": 0.178, "step": 3333 }, { "epoch": 0.69, "learning_rate": 1.9618978405315613e-07, "loss": 0.1325, "step": 3334 }, { "epoch": 0.69, "learning_rate": 1.9615863787375415e-07, "loss": 0.1959, "step": 3335 }, { "epoch": 0.69, "learning_rate": 1.9612749169435216e-07, "loss": 0.2284, "step": 3336 }, { "epoch": 0.69, "learning_rate": 1.9609634551495018e-07, "loss": 0.2164, "step": 3337 }, { "epoch": 0.69, "learning_rate": 1.9606519933554814e-07, "loss": 0.1842, "step": 3338 }, { "epoch": 0.69, "learning_rate": 1.9603405315614616e-07, "loss": 0.2037, "step": 3339 }, { "epoch": 0.69, "learning_rate": 1.9600290697674417e-07, "loss": 0.1492, "step": 3340 }, { "epoch": 0.69, "learning_rate": 1.959717607973422e-07, "loss": 0.1904, "step": 3341 }, { "epoch": 0.69, "learning_rate": 1.959406146179402e-07, "loss": 0.2648, "step": 3342 }, { "epoch": 0.69, "learning_rate": 1.959094684385382e-07, "loss": 0.1851, "step": 3343 }, { "epoch": 0.69, "learning_rate": 1.958783222591362e-07, "loss": 0.188, "step": 3344 }, { "epoch": 0.69, "learning_rate": 1.958471760797342e-07, "loss": 0.1805, "step": 3345 }, { "epoch": 0.69, "learning_rate": 1.9581602990033221e-07, "loss": 0.2551, "step": 3346 }, { "epoch": 0.69, "learning_rate": 1.9578488372093023e-07, "loss": 0.1614, "step": 3347 }, { "epoch": 0.7, "learning_rate": 1.9575373754152822e-07, "loss": 0.178, "step": 3348 }, { "epoch": 0.7, "learning_rate": 1.9572259136212624e-07, "loss": 0.2223, "step": 3349 }, { "epoch": 0.7, "learning_rate": 1.9569144518272425e-07, "loss": 0.2411, "step": 3350 }, { "epoch": 0.7, "learning_rate": 1.9566029900332227e-07, "loss": 0.229, "step": 3351 }, { "epoch": 0.7, "learning_rate": 1.9562915282392023e-07, "loss": 0.1829, "step": 3352 }, { "epoch": 0.7, "learning_rate": 1.9559800664451825e-07, "loss": 0.2512, "step": 3353 }, { "epoch": 0.7, "learning_rate": 1.9556686046511626e-07, "loss": 0.1782, "step": 3354 }, { "epoch": 0.7, "learning_rate": 1.9553571428571428e-07, "loss": 0.1725, "step": 3355 }, { "epoch": 0.7, "learning_rate": 1.955045681063123e-07, "loss": 0.1466, "step": 3356 }, { "epoch": 0.7, "learning_rate": 1.9547342192691028e-07, "loss": 0.2049, "step": 3357 }, { "epoch": 0.7, "learning_rate": 1.954422757475083e-07, "loss": 0.1803, "step": 3358 }, { "epoch": 0.7, "learning_rate": 1.9541112956810631e-07, "loss": 0.1685, "step": 3359 }, { "epoch": 0.7, "learning_rate": 1.953799833887043e-07, "loss": 0.1872, "step": 3360 }, { "epoch": 0.7, "learning_rate": 1.9534883720930232e-07, "loss": 0.2233, "step": 3361 }, { "epoch": 0.7, "learning_rate": 1.953176910299003e-07, "loss": 0.2252, "step": 3362 }, { "epoch": 0.7, "learning_rate": 1.9528654485049832e-07, "loss": 0.1664, "step": 3363 }, { "epoch": 0.7, "learning_rate": 1.9525539867109634e-07, "loss": 0.1901, "step": 3364 }, { "epoch": 0.7, "learning_rate": 1.9522425249169436e-07, "loss": 0.191, "step": 3365 }, { "epoch": 0.7, "learning_rate": 1.9519310631229237e-07, "loss": 0.196, "step": 3366 }, { "epoch": 0.7, "learning_rate": 1.9516196013289034e-07, "loss": 0.1913, "step": 3367 }, { "epoch": 0.7, "learning_rate": 1.9513081395348835e-07, "loss": 0.1629, "step": 3368 }, { "epoch": 0.7, "learning_rate": 1.9509966777408637e-07, "loss": 0.1748, "step": 3369 }, { "epoch": 0.7, "learning_rate": 1.9506852159468438e-07, "loss": 0.1948, "step": 3370 }, { "epoch": 0.7, "learning_rate": 1.9503737541528237e-07, "loss": 0.1824, "step": 3371 }, { "epoch": 0.7, "learning_rate": 1.950062292358804e-07, "loss": 0.209, "step": 3372 }, { "epoch": 0.7, "learning_rate": 1.949750830564784e-07, "loss": 0.1748, "step": 3373 }, { "epoch": 0.7, "learning_rate": 1.9494393687707642e-07, "loss": 0.2677, "step": 3374 }, { "epoch": 0.7, "learning_rate": 1.9491279069767444e-07, "loss": 0.2649, "step": 3375 }, { "epoch": 0.7, "learning_rate": 1.948816445182724e-07, "loss": 0.2177, "step": 3376 }, { "epoch": 0.7, "learning_rate": 1.9485049833887041e-07, "loss": 0.1983, "step": 3377 }, { "epoch": 0.7, "learning_rate": 1.9481935215946843e-07, "loss": 0.1528, "step": 3378 }, { "epoch": 0.7, "learning_rate": 1.9478820598006645e-07, "loss": 0.1675, "step": 3379 }, { "epoch": 0.7, "learning_rate": 1.9475705980066446e-07, "loss": 0.1788, "step": 3380 }, { "epoch": 0.7, "learning_rate": 1.9472591362126245e-07, "loss": 0.224, "step": 3381 }, { "epoch": 0.7, "learning_rate": 1.9469476744186044e-07, "loss": 0.1704, "step": 3382 }, { "epoch": 0.7, "learning_rate": 1.9466362126245846e-07, "loss": 0.1702, "step": 3383 }, { "epoch": 0.7, "learning_rate": 1.9463247508305647e-07, "loss": 0.223, "step": 3384 }, { "epoch": 0.7, "learning_rate": 1.9460132890365446e-07, "loss": 0.1895, "step": 3385 }, { "epoch": 0.7, "learning_rate": 1.9457018272425248e-07, "loss": 0.1516, "step": 3386 }, { "epoch": 0.7, "learning_rate": 1.945390365448505e-07, "loss": 0.2313, "step": 3387 }, { "epoch": 0.7, "learning_rate": 1.945078903654485e-07, "loss": 0.1869, "step": 3388 }, { "epoch": 0.7, "learning_rate": 1.9447674418604652e-07, "loss": 0.238, "step": 3389 }, { "epoch": 0.7, "learning_rate": 1.944455980066445e-07, "loss": 0.192, "step": 3390 }, { "epoch": 0.7, "learning_rate": 1.944144518272425e-07, "loss": 0.2374, "step": 3391 }, { "epoch": 0.7, "learning_rate": 1.9438330564784052e-07, "loss": 0.1618, "step": 3392 }, { "epoch": 0.7, "learning_rate": 1.9435215946843854e-07, "loss": 0.1872, "step": 3393 }, { "epoch": 0.7, "learning_rate": 1.9432101328903655e-07, "loss": 0.1848, "step": 3394 }, { "epoch": 0.7, "learning_rate": 1.9428986710963454e-07, "loss": 0.191, "step": 3395 }, { "epoch": 0.71, "learning_rate": 1.9425872093023256e-07, "loss": 0.2032, "step": 3396 }, { "epoch": 0.71, "learning_rate": 1.9422757475083055e-07, "loss": 0.1517, "step": 3397 }, { "epoch": 0.71, "learning_rate": 1.9419642857142856e-07, "loss": 0.1787, "step": 3398 }, { "epoch": 0.71, "learning_rate": 1.9416528239202658e-07, "loss": 0.1541, "step": 3399 }, { "epoch": 0.71, "learning_rate": 1.9413413621262457e-07, "loss": 0.2368, "step": 3400 }, { "epoch": 0.71, "learning_rate": 1.9410299003322258e-07, "loss": 0.2394, "step": 3401 }, { "epoch": 0.71, "learning_rate": 1.940718438538206e-07, "loss": 0.1536, "step": 3402 }, { "epoch": 0.71, "learning_rate": 1.9404069767441861e-07, "loss": 0.2218, "step": 3403 }, { "epoch": 0.71, "learning_rate": 1.9400955149501658e-07, "loss": 0.2179, "step": 3404 }, { "epoch": 0.71, "learning_rate": 1.939784053156146e-07, "loss": 0.2036, "step": 3405 }, { "epoch": 0.71, "learning_rate": 1.939472591362126e-07, "loss": 0.1539, "step": 3406 }, { "epoch": 0.71, "learning_rate": 1.9391611295681062e-07, "loss": 0.1861, "step": 3407 }, { "epoch": 0.71, "learning_rate": 1.9388496677740864e-07, "loss": 0.1666, "step": 3408 }, { "epoch": 0.71, "learning_rate": 1.9385382059800663e-07, "loss": 0.1736, "step": 3409 }, { "epoch": 0.71, "learning_rate": 1.9382267441860465e-07, "loss": 0.1269, "step": 3410 }, { "epoch": 0.71, "learning_rate": 1.9379152823920266e-07, "loss": 0.2183, "step": 3411 }, { "epoch": 0.71, "learning_rate": 1.9376038205980068e-07, "loss": 0.204, "step": 3412 }, { "epoch": 0.71, "learning_rate": 1.9372923588039867e-07, "loss": 0.2244, "step": 3413 }, { "epoch": 0.71, "learning_rate": 1.9369808970099666e-07, "loss": 0.1598, "step": 3414 }, { "epoch": 0.71, "learning_rate": 1.9366694352159467e-07, "loss": 0.1771, "step": 3415 }, { "epoch": 0.71, "learning_rate": 1.936357973421927e-07, "loss": 0.2133, "step": 3416 }, { "epoch": 0.71, "learning_rate": 1.936046511627907e-07, "loss": 0.1927, "step": 3417 }, { "epoch": 0.71, "learning_rate": 1.935735049833887e-07, "loss": 0.2154, "step": 3418 }, { "epoch": 0.71, "learning_rate": 1.9354235880398668e-07, "loss": 0.1589, "step": 3419 }, { "epoch": 0.71, "learning_rate": 1.935112126245847e-07, "loss": 0.1767, "step": 3420 }, { "epoch": 0.71, "learning_rate": 1.9348006644518271e-07, "loss": 0.1791, "step": 3421 }, { "epoch": 0.71, "learning_rate": 1.9344892026578073e-07, "loss": 0.1538, "step": 3422 }, { "epoch": 0.71, "learning_rate": 1.9341777408637872e-07, "loss": 0.2437, "step": 3423 }, { "epoch": 0.71, "learning_rate": 1.9338662790697673e-07, "loss": 0.2402, "step": 3424 }, { "epoch": 0.71, "learning_rate": 1.9335548172757475e-07, "loss": 0.2721, "step": 3425 }, { "epoch": 0.71, "learning_rate": 1.9332433554817277e-07, "loss": 0.1488, "step": 3426 }, { "epoch": 0.71, "learning_rate": 1.9329318936877078e-07, "loss": 0.1661, "step": 3427 }, { "epoch": 0.71, "learning_rate": 1.9326204318936874e-07, "loss": 0.1695, "step": 3428 }, { "epoch": 0.71, "learning_rate": 1.9323089700996676e-07, "loss": 0.1555, "step": 3429 }, { "epoch": 0.71, "learning_rate": 1.9319975083056478e-07, "loss": 0.2353, "step": 3430 }, { "epoch": 0.71, "learning_rate": 1.931686046511628e-07, "loss": 0.1498, "step": 3431 }, { "epoch": 0.71, "learning_rate": 1.9313745847176078e-07, "loss": 0.2265, "step": 3432 }, { "epoch": 0.71, "learning_rate": 1.931063122923588e-07, "loss": 0.1661, "step": 3433 }, { "epoch": 0.71, "learning_rate": 1.930751661129568e-07, "loss": 0.1682, "step": 3434 }, { "epoch": 0.71, "learning_rate": 1.930440199335548e-07, "loss": 0.2398, "step": 3435 }, { "epoch": 0.71, "learning_rate": 1.9301287375415282e-07, "loss": 0.2041, "step": 3436 }, { "epoch": 0.71, "learning_rate": 1.929817275747508e-07, "loss": 0.2033, "step": 3437 }, { "epoch": 0.71, "learning_rate": 1.9295058139534882e-07, "loss": 0.1572, "step": 3438 }, { "epoch": 0.71, "learning_rate": 1.9291943521594684e-07, "loss": 0.175, "step": 3439 }, { "epoch": 0.71, "learning_rate": 1.9288828903654486e-07, "loss": 0.2379, "step": 3440 }, { "epoch": 0.71, "learning_rate": 1.9285714285714287e-07, "loss": 0.1735, "step": 3441 }, { "epoch": 0.71, "learning_rate": 1.9282599667774083e-07, "loss": 0.2073, "step": 3442 }, { "epoch": 0.71, "learning_rate": 1.9279485049833885e-07, "loss": 0.1748, "step": 3443 }, { "epoch": 0.72, "learning_rate": 1.9276370431893687e-07, "loss": 0.2357, "step": 3444 }, { "epoch": 0.72, "learning_rate": 1.9273255813953488e-07, "loss": 0.1609, "step": 3445 }, { "epoch": 0.72, "learning_rate": 1.927014119601329e-07, "loss": 0.2188, "step": 3446 }, { "epoch": 0.72, "learning_rate": 1.9267026578073089e-07, "loss": 0.1492, "step": 3447 }, { "epoch": 0.72, "learning_rate": 1.926391196013289e-07, "loss": 0.1873, "step": 3448 }, { "epoch": 0.72, "learning_rate": 1.9260797342192692e-07, "loss": 0.1541, "step": 3449 }, { "epoch": 0.72, "learning_rate": 1.925768272425249e-07, "loss": 0.1876, "step": 3450 }, { "epoch": 0.72, "learning_rate": 1.925456810631229e-07, "loss": 0.1557, "step": 3451 }, { "epoch": 0.72, "learning_rate": 1.925145348837209e-07, "loss": 0.2053, "step": 3452 }, { "epoch": 0.72, "learning_rate": 1.9248338870431893e-07, "loss": 0.1982, "step": 3453 }, { "epoch": 0.72, "learning_rate": 1.9245224252491694e-07, "loss": 0.1872, "step": 3454 }, { "epoch": 0.72, "learning_rate": 1.9242109634551496e-07, "loss": 0.1815, "step": 3455 }, { "epoch": 0.72, "learning_rate": 1.9238995016611292e-07, "loss": 0.1795, "step": 3456 }, { "epoch": 0.72, "learning_rate": 1.9235880398671094e-07, "loss": 0.2151, "step": 3457 }, { "epoch": 0.72, "learning_rate": 1.9232765780730895e-07, "loss": 0.1353, "step": 3458 }, { "epoch": 0.72, "learning_rate": 1.9229651162790697e-07, "loss": 0.2364, "step": 3459 }, { "epoch": 0.72, "learning_rate": 1.9226536544850499e-07, "loss": 0.175, "step": 3460 }, { "epoch": 0.72, "learning_rate": 1.9223421926910298e-07, "loss": 0.1586, "step": 3461 }, { "epoch": 0.72, "learning_rate": 1.92203073089701e-07, "loss": 0.1771, "step": 3462 }, { "epoch": 0.72, "learning_rate": 1.92171926910299e-07, "loss": 0.2104, "step": 3463 }, { "epoch": 0.72, "learning_rate": 1.9214078073089702e-07, "loss": 0.1679, "step": 3464 }, { "epoch": 0.72, "learning_rate": 1.9210963455149499e-07, "loss": 0.177, "step": 3465 }, { "epoch": 0.72, "learning_rate": 1.92078488372093e-07, "loss": 0.1895, "step": 3466 }, { "epoch": 0.72, "learning_rate": 1.9204734219269102e-07, "loss": 0.25, "step": 3467 }, { "epoch": 0.72, "learning_rate": 1.9201619601328903e-07, "loss": 0.2274, "step": 3468 }, { "epoch": 0.72, "learning_rate": 1.9198504983388705e-07, "loss": 0.2251, "step": 3469 }, { "epoch": 0.72, "learning_rate": 1.9195390365448504e-07, "loss": 0.1949, "step": 3470 }, { "epoch": 0.72, "learning_rate": 1.9192275747508303e-07, "loss": 0.1591, "step": 3471 }, { "epoch": 0.72, "learning_rate": 1.9189161129568104e-07, "loss": 0.245, "step": 3472 }, { "epoch": 0.72, "learning_rate": 1.9186046511627906e-07, "loss": 0.1407, "step": 3473 }, { "epoch": 0.72, "learning_rate": 1.9182931893687708e-07, "loss": 0.2017, "step": 3474 }, { "epoch": 0.72, "learning_rate": 1.9179817275747507e-07, "loss": 0.1807, "step": 3475 }, { "epoch": 0.72, "learning_rate": 1.9176702657807308e-07, "loss": 0.1107, "step": 3476 }, { "epoch": 0.72, "learning_rate": 1.917358803986711e-07, "loss": 0.1732, "step": 3477 }, { "epoch": 0.72, "learning_rate": 1.917047342192691e-07, "loss": 0.2475, "step": 3478 }, { "epoch": 0.72, "learning_rate": 1.9167358803986708e-07, "loss": 0.1721, "step": 3479 }, { "epoch": 0.72, "learning_rate": 1.916424418604651e-07, "loss": 0.2423, "step": 3480 }, { "epoch": 0.72, "learning_rate": 1.916112956810631e-07, "loss": 0.1914, "step": 3481 }, { "epoch": 0.72, "learning_rate": 1.9158014950166112e-07, "loss": 0.1714, "step": 3482 }, { "epoch": 0.72, "learning_rate": 1.9154900332225914e-07, "loss": 0.1962, "step": 3483 }, { "epoch": 0.72, "learning_rate": 1.9151785714285713e-07, "loss": 0.1592, "step": 3484 }, { "epoch": 0.72, "learning_rate": 1.9148671096345514e-07, "loss": 0.1745, "step": 3485 }, { "epoch": 0.72, "learning_rate": 1.9145556478405316e-07, "loss": 0.2097, "step": 3486 }, { "epoch": 0.72, "learning_rate": 1.9142441860465115e-07, "loss": 0.1831, "step": 3487 }, { "epoch": 0.72, "learning_rate": 1.9139327242524916e-07, "loss": 0.2553, "step": 3488 }, { "epoch": 0.72, "learning_rate": 1.9136212624584715e-07, "loss": 0.1594, "step": 3489 }, { "epoch": 0.72, "learning_rate": 1.9133098006644517e-07, "loss": 0.2138, "step": 3490 }, { "epoch": 0.72, "learning_rate": 1.9129983388704319e-07, "loss": 0.2472, "step": 3491 }, { "epoch": 0.73, "learning_rate": 1.912686877076412e-07, "loss": 0.1893, "step": 3492 }, { "epoch": 0.73, "learning_rate": 1.9123754152823922e-07, "loss": 0.2258, "step": 3493 }, { "epoch": 0.73, "learning_rate": 1.9120639534883718e-07, "loss": 0.2212, "step": 3494 }, { "epoch": 0.73, "learning_rate": 1.911752491694352e-07, "loss": 0.1798, "step": 3495 }, { "epoch": 0.73, "learning_rate": 1.911441029900332e-07, "loss": 0.2135, "step": 3496 }, { "epoch": 0.73, "learning_rate": 1.9111295681063123e-07, "loss": 0.2362, "step": 3497 }, { "epoch": 0.73, "learning_rate": 1.9108181063122922e-07, "loss": 0.2029, "step": 3498 }, { "epoch": 0.73, "learning_rate": 1.9105066445182723e-07, "loss": 0.2005, "step": 3499 }, { "epoch": 0.73, "learning_rate": 1.9101951827242525e-07, "loss": 0.1803, "step": 3500 }, { "epoch": 0.73, "learning_rate": 1.9098837209302326e-07, "loss": 0.2191, "step": 3501 }, { "epoch": 0.73, "learning_rate": 1.9095722591362125e-07, "loss": 0.1748, "step": 3502 }, { "epoch": 0.73, "learning_rate": 1.9092607973421924e-07, "loss": 0.1958, "step": 3503 }, { "epoch": 0.73, "learning_rate": 1.9089493355481726e-07, "loss": 0.1844, "step": 3504 }, { "epoch": 0.73, "learning_rate": 1.9086378737541528e-07, "loss": 0.1649, "step": 3505 }, { "epoch": 0.73, "learning_rate": 1.908326411960133e-07, "loss": 0.155, "step": 3506 }, { "epoch": 0.73, "learning_rate": 1.908014950166113e-07, "loss": 0.2256, "step": 3507 }, { "epoch": 0.73, "learning_rate": 1.907703488372093e-07, "loss": 0.1839, "step": 3508 }, { "epoch": 0.73, "learning_rate": 1.9073920265780729e-07, "loss": 0.2203, "step": 3509 }, { "epoch": 0.73, "learning_rate": 1.907080564784053e-07, "loss": 0.2085, "step": 3510 }, { "epoch": 0.73, "learning_rate": 1.9067691029900332e-07, "loss": 0.2032, "step": 3511 }, { "epoch": 0.73, "learning_rate": 1.906457641196013e-07, "loss": 0.177, "step": 3512 }, { "epoch": 0.73, "learning_rate": 1.9061461794019932e-07, "loss": 0.2091, "step": 3513 }, { "epoch": 0.73, "learning_rate": 1.9058347176079734e-07, "loss": 0.1662, "step": 3514 }, { "epoch": 0.73, "learning_rate": 1.9055232558139535e-07, "loss": 0.1553, "step": 3515 }, { "epoch": 0.73, "learning_rate": 1.9052117940199337e-07, "loss": 0.2362, "step": 3516 }, { "epoch": 0.73, "learning_rate": 1.9049003322259133e-07, "loss": 0.161, "step": 3517 }, { "epoch": 0.73, "learning_rate": 1.9045888704318935e-07, "loss": 0.2429, "step": 3518 }, { "epoch": 0.73, "learning_rate": 1.9042774086378736e-07, "loss": 0.1985, "step": 3519 }, { "epoch": 0.73, "learning_rate": 1.9039659468438538e-07, "loss": 0.1537, "step": 3520 }, { "epoch": 0.73, "learning_rate": 1.903654485049834e-07, "loss": 0.1799, "step": 3521 }, { "epoch": 0.73, "learning_rate": 1.9033430232558139e-07, "loss": 0.1943, "step": 3522 }, { "epoch": 0.73, "learning_rate": 1.903031561461794e-07, "loss": 0.1706, "step": 3523 }, { "epoch": 0.73, "learning_rate": 1.902720099667774e-07, "loss": 0.1831, "step": 3524 }, { "epoch": 0.73, "learning_rate": 1.902408637873754e-07, "loss": 0.2017, "step": 3525 }, { "epoch": 0.73, "learning_rate": 1.902097176079734e-07, "loss": 0.1925, "step": 3526 }, { "epoch": 0.73, "learning_rate": 1.901785714285714e-07, "loss": 0.2046, "step": 3527 }, { "epoch": 0.73, "learning_rate": 1.9014742524916943e-07, "loss": 0.2889, "step": 3528 }, { "epoch": 0.73, "learning_rate": 1.9011627906976744e-07, "loss": 0.1958, "step": 3529 }, { "epoch": 0.73, "learning_rate": 1.9008513289036546e-07, "loss": 0.1612, "step": 3530 }, { "epoch": 0.73, "learning_rate": 1.9005398671096342e-07, "loss": 0.2281, "step": 3531 }, { "epoch": 0.73, "learning_rate": 1.9002284053156144e-07, "loss": 0.1438, "step": 3532 }, { "epoch": 0.73, "learning_rate": 1.8999169435215945e-07, "loss": 0.2084, "step": 3533 }, { "epoch": 0.73, "learning_rate": 1.8996054817275747e-07, "loss": 0.1595, "step": 3534 }, { "epoch": 0.73, "learning_rate": 1.8992940199335549e-07, "loss": 0.1892, "step": 3535 }, { "epoch": 0.73, "learning_rate": 1.8989825581395347e-07, "loss": 0.1524, "step": 3536 }, { "epoch": 0.73, "learning_rate": 1.898671096345515e-07, "loss": 0.1214, "step": 3537 }, { "epoch": 0.73, "learning_rate": 1.898359634551495e-07, "loss": 0.184, "step": 3538 }, { "epoch": 0.73, "learning_rate": 1.898048172757475e-07, "loss": 0.1477, "step": 3539 }, { "epoch": 0.74, "learning_rate": 1.897736710963455e-07, "loss": 0.2457, "step": 3540 }, { "epoch": 0.74, "learning_rate": 1.897425249169435e-07, "loss": 0.238, "step": 3541 }, { "epoch": 0.74, "learning_rate": 1.8971137873754152e-07, "loss": 0.1563, "step": 3542 }, { "epoch": 0.74, "learning_rate": 1.8968023255813953e-07, "loss": 0.1671, "step": 3543 }, { "epoch": 0.74, "learning_rate": 1.8964908637873755e-07, "loss": 0.2161, "step": 3544 }, { "epoch": 0.74, "learning_rate": 1.8961794019933554e-07, "loss": 0.1613, "step": 3545 }, { "epoch": 0.74, "learning_rate": 1.8958679401993353e-07, "loss": 0.195, "step": 3546 }, { "epoch": 0.74, "learning_rate": 1.8955564784053154e-07, "loss": 0.1944, "step": 3547 }, { "epoch": 0.74, "learning_rate": 1.8952450166112956e-07, "loss": 0.1855, "step": 3548 }, { "epoch": 0.74, "learning_rate": 1.8949335548172757e-07, "loss": 0.1396, "step": 3549 }, { "epoch": 0.74, "learning_rate": 1.8946220930232556e-07, "loss": 0.2662, "step": 3550 }, { "epoch": 0.74, "learning_rate": 1.8943106312292358e-07, "loss": 0.2177, "step": 3551 }, { "epoch": 0.74, "learning_rate": 1.893999169435216e-07, "loss": 0.1959, "step": 3552 }, { "epoch": 0.74, "learning_rate": 1.893687707641196e-07, "loss": 0.1633, "step": 3553 }, { "epoch": 0.74, "learning_rate": 1.8933762458471763e-07, "loss": 0.2058, "step": 3554 }, { "epoch": 0.74, "learning_rate": 1.893064784053156e-07, "loss": 0.1693, "step": 3555 }, { "epoch": 0.74, "learning_rate": 1.892753322259136e-07, "loss": 0.2616, "step": 3556 }, { "epoch": 0.74, "learning_rate": 1.8924418604651162e-07, "loss": 0.2175, "step": 3557 }, { "epoch": 0.74, "learning_rate": 1.8921303986710964e-07, "loss": 0.2192, "step": 3558 }, { "epoch": 0.74, "learning_rate": 1.8918189368770763e-07, "loss": 0.2523, "step": 3559 }, { "epoch": 0.74, "learning_rate": 1.8915074750830564e-07, "loss": 0.2048, "step": 3560 }, { "epoch": 0.74, "learning_rate": 1.8911960132890363e-07, "loss": 0.1666, "step": 3561 }, { "epoch": 0.74, "learning_rate": 1.8908845514950165e-07, "loss": 0.1955, "step": 3562 }, { "epoch": 0.74, "learning_rate": 1.8905730897009966e-07, "loss": 0.1687, "step": 3563 }, { "epoch": 0.74, "learning_rate": 1.8902616279069765e-07, "loss": 0.1397, "step": 3564 }, { "epoch": 0.74, "learning_rate": 1.8899501661129567e-07, "loss": 0.206, "step": 3565 }, { "epoch": 0.74, "learning_rate": 1.8896387043189368e-07, "loss": 0.1628, "step": 3566 }, { "epoch": 0.74, "learning_rate": 1.889327242524917e-07, "loss": 0.2403, "step": 3567 }, { "epoch": 0.74, "learning_rate": 1.8890157807308972e-07, "loss": 0.1568, "step": 3568 }, { "epoch": 0.74, "learning_rate": 1.8887043189368768e-07, "loss": 0.2212, "step": 3569 }, { "epoch": 0.74, "learning_rate": 1.888392857142857e-07, "loss": 0.2194, "step": 3570 }, { "epoch": 0.74, "learning_rate": 1.888081395348837e-07, "loss": 0.2611, "step": 3571 }, { "epoch": 0.74, "learning_rate": 1.8877699335548173e-07, "loss": 0.1696, "step": 3572 }, { "epoch": 0.74, "learning_rate": 1.8874584717607974e-07, "loss": 0.2102, "step": 3573 }, { "epoch": 0.74, "learning_rate": 1.8871470099667773e-07, "loss": 0.1857, "step": 3574 }, { "epoch": 0.74, "learning_rate": 1.8868355481727575e-07, "loss": 0.2058, "step": 3575 }, { "epoch": 0.74, "learning_rate": 1.8865240863787374e-07, "loss": 0.1988, "step": 3576 }, { "epoch": 0.74, "learning_rate": 1.8862126245847175e-07, "loss": 0.1358, "step": 3577 }, { "epoch": 0.74, "learning_rate": 1.8859011627906974e-07, "loss": 0.2813, "step": 3578 }, { "epoch": 0.74, "learning_rate": 1.8855897009966776e-07, "loss": 0.1433, "step": 3579 }, { "epoch": 0.74, "learning_rate": 1.8852782392026577e-07, "loss": 0.2799, "step": 3580 }, { "epoch": 0.74, "learning_rate": 1.884966777408638e-07, "loss": 0.2505, "step": 3581 }, { "epoch": 0.74, "learning_rate": 1.884655315614618e-07, "loss": 0.1424, "step": 3582 }, { "epoch": 0.74, "learning_rate": 1.8843438538205977e-07, "loss": 0.1911, "step": 3583 }, { "epoch": 0.74, "learning_rate": 1.8840323920265778e-07, "loss": 0.2081, "step": 3584 }, { "epoch": 0.74, "learning_rate": 1.883720930232558e-07, "loss": 0.1493, "step": 3585 }, { "epoch": 0.74, "learning_rate": 1.8834094684385382e-07, "loss": 0.2237, "step": 3586 }, { "epoch": 0.74, "learning_rate": 1.8830980066445183e-07, "loss": 0.2448, "step": 3587 }, { "epoch": 0.75, "learning_rate": 1.8827865448504982e-07, "loss": 0.1871, "step": 3588 }, { "epoch": 0.75, "learning_rate": 1.8824750830564784e-07, "loss": 0.2073, "step": 3589 }, { "epoch": 0.75, "learning_rate": 1.8821636212624585e-07, "loss": 0.2079, "step": 3590 }, { "epoch": 0.75, "learning_rate": 1.8818521594684387e-07, "loss": 0.245, "step": 3591 }, { "epoch": 0.75, "learning_rate": 1.8815406976744183e-07, "loss": 0.2183, "step": 3592 }, { "epoch": 0.75, "learning_rate": 1.8812292358803985e-07, "loss": 0.211, "step": 3593 }, { "epoch": 0.75, "learning_rate": 1.8809177740863786e-07, "loss": 0.187, "step": 3594 }, { "epoch": 0.75, "learning_rate": 1.8806063122923588e-07, "loss": 0.1762, "step": 3595 }, { "epoch": 0.75, "learning_rate": 1.880294850498339e-07, "loss": 0.1624, "step": 3596 }, { "epoch": 0.75, "learning_rate": 1.8799833887043188e-07, "loss": 0.1721, "step": 3597 }, { "epoch": 0.75, "learning_rate": 1.8796719269102987e-07, "loss": 0.2212, "step": 3598 }, { "epoch": 0.75, "learning_rate": 1.879360465116279e-07, "loss": 0.1467, "step": 3599 }, { "epoch": 0.75, "learning_rate": 1.879049003322259e-07, "loss": 0.3094, "step": 3600 }, { "epoch": 0.75, "learning_rate": 1.8787375415282392e-07, "loss": 0.1762, "step": 3601 }, { "epoch": 0.75, "learning_rate": 1.878426079734219e-07, "loss": 0.2376, "step": 3602 }, { "epoch": 0.75, "learning_rate": 1.8781146179401993e-07, "loss": 0.1584, "step": 3603 }, { "epoch": 0.75, "learning_rate": 1.8778031561461794e-07, "loss": 0.238, "step": 3604 }, { "epoch": 0.75, "learning_rate": 1.8774916943521596e-07, "loss": 0.2493, "step": 3605 }, { "epoch": 0.75, "learning_rate": 1.8771802325581392e-07, "loss": 0.2293, "step": 3606 }, { "epoch": 0.75, "learning_rate": 1.8768687707641194e-07, "loss": 0.1908, "step": 3607 }, { "epoch": 0.75, "learning_rate": 1.8765573089700995e-07, "loss": 0.1974, "step": 3608 }, { "epoch": 0.75, "learning_rate": 1.8762458471760797e-07, "loss": 0.2073, "step": 3609 }, { "epoch": 0.75, "learning_rate": 1.8759343853820598e-07, "loss": 0.2298, "step": 3610 }, { "epoch": 0.75, "learning_rate": 1.8756229235880397e-07, "loss": 0.1576, "step": 3611 }, { "epoch": 0.75, "learning_rate": 1.87531146179402e-07, "loss": 0.1857, "step": 3612 }, { "epoch": 0.75, "learning_rate": 1.875e-07, "loss": 0.1669, "step": 3613 }, { "epoch": 0.75, "learning_rate": 1.87468853820598e-07, "loss": 0.1887, "step": 3614 }, { "epoch": 0.75, "learning_rate": 1.87437707641196e-07, "loss": 0.2174, "step": 3615 }, { "epoch": 0.75, "learning_rate": 1.87406561461794e-07, "loss": 0.1736, "step": 3616 }, { "epoch": 0.75, "learning_rate": 1.8737541528239202e-07, "loss": 0.2309, "step": 3617 }, { "epoch": 0.75, "learning_rate": 1.8734426910299003e-07, "loss": 0.1817, "step": 3618 }, { "epoch": 0.75, "learning_rate": 1.8731312292358805e-07, "loss": 0.2168, "step": 3619 }, { "epoch": 0.75, "learning_rate": 1.8728197674418606e-07, "loss": 0.1792, "step": 3620 }, { "epoch": 0.75, "learning_rate": 1.8725083056478403e-07, "loss": 0.1336, "step": 3621 }, { "epoch": 0.75, "learning_rate": 1.8721968438538204e-07, "loss": 0.2341, "step": 3622 }, { "epoch": 0.75, "learning_rate": 1.8718853820598006e-07, "loss": 0.2294, "step": 3623 }, { "epoch": 0.75, "learning_rate": 1.8715739202657807e-07, "loss": 0.2387, "step": 3624 }, { "epoch": 0.75, "learning_rate": 1.8712624584717606e-07, "loss": 0.1599, "step": 3625 }, { "epoch": 0.75, "learning_rate": 1.8709509966777408e-07, "loss": 0.1439, "step": 3626 }, { "epoch": 0.75, "learning_rate": 1.870639534883721e-07, "loss": 0.1798, "step": 3627 }, { "epoch": 0.75, "learning_rate": 1.870328073089701e-07, "loss": 0.1972, "step": 3628 }, { "epoch": 0.75, "learning_rate": 1.870016611295681e-07, "loss": 0.1559, "step": 3629 }, { "epoch": 0.75, "learning_rate": 1.869705149501661e-07, "loss": 0.1325, "step": 3630 }, { "epoch": 0.75, "learning_rate": 1.869393687707641e-07, "loss": 0.1719, "step": 3631 }, { "epoch": 0.75, "learning_rate": 1.8690822259136212e-07, "loss": 0.1882, "step": 3632 }, { "epoch": 0.75, "learning_rate": 1.8687707641196014e-07, "loss": 0.1935, "step": 3633 }, { "epoch": 0.75, "learning_rate": 1.8684593023255815e-07, "loss": 0.1939, "step": 3634 }, { "epoch": 0.75, "learning_rate": 1.8681478405315611e-07, "loss": 0.172, "step": 3635 }, { "epoch": 0.75, "learning_rate": 1.8678363787375413e-07, "loss": 0.1717, "step": 3636 }, { "epoch": 0.76, "learning_rate": 1.8675249169435215e-07, "loss": 0.1951, "step": 3637 }, { "epoch": 0.76, "learning_rate": 1.8672134551495016e-07, "loss": 0.174, "step": 3638 }, { "epoch": 0.76, "learning_rate": 1.8669019933554815e-07, "loss": 0.1486, "step": 3639 }, { "epoch": 0.76, "learning_rate": 1.8665905315614617e-07, "loss": 0.2095, "step": 3640 }, { "epoch": 0.76, "learning_rate": 1.8662790697674418e-07, "loss": 0.1996, "step": 3641 }, { "epoch": 0.76, "learning_rate": 1.865967607973422e-07, "loss": 0.1387, "step": 3642 }, { "epoch": 0.76, "learning_rate": 1.8656561461794021e-07, "loss": 0.1793, "step": 3643 }, { "epoch": 0.76, "learning_rate": 1.8653446843853818e-07, "loss": 0.1595, "step": 3644 }, { "epoch": 0.76, "learning_rate": 1.865033222591362e-07, "loss": 0.2333, "step": 3645 }, { "epoch": 0.76, "learning_rate": 1.864721760797342e-07, "loss": 0.1689, "step": 3646 }, { "epoch": 0.76, "learning_rate": 1.8644102990033223e-07, "loss": 0.2269, "step": 3647 }, { "epoch": 0.76, "learning_rate": 1.8640988372093024e-07, "loss": 0.2361, "step": 3648 }, { "epoch": 0.76, "learning_rate": 1.8637873754152823e-07, "loss": 0.1877, "step": 3649 }, { "epoch": 0.76, "learning_rate": 1.8634759136212625e-07, "loss": 0.21, "step": 3650 }, { "epoch": 0.76, "learning_rate": 1.8631644518272424e-07, "loss": 0.1695, "step": 3651 }, { "epoch": 0.76, "learning_rate": 1.8628529900332225e-07, "loss": 0.2034, "step": 3652 }, { "epoch": 0.76, "learning_rate": 1.8625415282392024e-07, "loss": 0.1657, "step": 3653 }, { "epoch": 0.76, "learning_rate": 1.8622300664451826e-07, "loss": 0.2353, "step": 3654 }, { "epoch": 0.76, "learning_rate": 1.8619186046511627e-07, "loss": 0.2576, "step": 3655 }, { "epoch": 0.76, "learning_rate": 1.861607142857143e-07, "loss": 0.1818, "step": 3656 }, { "epoch": 0.76, "learning_rate": 1.861295681063123e-07, "loss": 0.1854, "step": 3657 }, { "epoch": 0.76, "learning_rate": 1.8609842192691027e-07, "loss": 0.1938, "step": 3658 }, { "epoch": 0.76, "learning_rate": 1.8606727574750828e-07, "loss": 0.2087, "step": 3659 }, { "epoch": 0.76, "learning_rate": 1.860361295681063e-07, "loss": 0.1792, "step": 3660 }, { "epoch": 0.76, "learning_rate": 1.8600498338870431e-07, "loss": 0.1509, "step": 3661 }, { "epoch": 0.76, "learning_rate": 1.8597383720930233e-07, "loss": 0.2082, "step": 3662 }, { "epoch": 0.76, "learning_rate": 1.8594269102990032e-07, "loss": 0.2363, "step": 3663 }, { "epoch": 0.76, "learning_rate": 1.8591154485049834e-07, "loss": 0.2149, "step": 3664 }, { "epoch": 0.76, "learning_rate": 1.8588039867109635e-07, "loss": 0.1725, "step": 3665 }, { "epoch": 0.76, "learning_rate": 1.8584925249169434e-07, "loss": 0.1797, "step": 3666 }, { "epoch": 0.76, "learning_rate": 1.8581810631229236e-07, "loss": 0.278, "step": 3667 }, { "epoch": 0.76, "learning_rate": 1.8578696013289035e-07, "loss": 0.1893, "step": 3668 }, { "epoch": 0.76, "learning_rate": 1.8575581395348836e-07, "loss": 0.2055, "step": 3669 }, { "epoch": 0.76, "learning_rate": 1.8572466777408638e-07, "loss": 0.2237, "step": 3670 }, { "epoch": 0.76, "learning_rate": 1.856935215946844e-07, "loss": 0.1753, "step": 3671 }, { "epoch": 0.76, "learning_rate": 1.8566237541528236e-07, "loss": 0.1945, "step": 3672 }, { "epoch": 0.76, "learning_rate": 1.8563122923588037e-07, "loss": 0.1461, "step": 3673 }, { "epoch": 0.76, "learning_rate": 1.856000830564784e-07, "loss": 0.1883, "step": 3674 }, { "epoch": 0.76, "learning_rate": 1.855689368770764e-07, "loss": 0.1534, "step": 3675 }, { "epoch": 0.76, "learning_rate": 1.8553779069767442e-07, "loss": 0.1815, "step": 3676 }, { "epoch": 0.76, "learning_rate": 1.855066445182724e-07, "loss": 0.1886, "step": 3677 }, { "epoch": 0.76, "learning_rate": 1.8547549833887042e-07, "loss": 0.1604, "step": 3678 }, { "epoch": 0.76, "learning_rate": 1.8544435215946844e-07, "loss": 0.1856, "step": 3679 }, { "epoch": 0.76, "learning_rate": 1.8541320598006646e-07, "loss": 0.2687, "step": 3680 }, { "epoch": 0.76, "learning_rate": 1.8538205980066445e-07, "loss": 0.1894, "step": 3681 }, { "epoch": 0.76, "learning_rate": 1.8535091362126243e-07, "loss": 0.1702, "step": 3682 }, { "epoch": 0.76, "learning_rate": 1.8531976744186045e-07, "loss": 0.2441, "step": 3683 }, { "epoch": 0.76, "learning_rate": 1.8528862126245847e-07, "loss": 0.2199, "step": 3684 }, { "epoch": 0.77, "learning_rate": 1.8525747508305648e-07, "loss": 0.2242, "step": 3685 }, { "epoch": 0.77, "learning_rate": 1.8522632890365447e-07, "loss": 0.1942, "step": 3686 }, { "epoch": 0.77, "learning_rate": 1.851951827242525e-07, "loss": 0.2321, "step": 3687 }, { "epoch": 0.77, "learning_rate": 1.8516403654485048e-07, "loss": 0.2262, "step": 3688 }, { "epoch": 0.77, "learning_rate": 1.851328903654485e-07, "loss": 0.2625, "step": 3689 }, { "epoch": 0.77, "learning_rate": 1.851017441860465e-07, "loss": 0.1419, "step": 3690 }, { "epoch": 0.77, "learning_rate": 1.850705980066445e-07, "loss": 0.2396, "step": 3691 }, { "epoch": 0.77, "learning_rate": 1.8503945182724251e-07, "loss": 0.1743, "step": 3692 }, { "epoch": 0.77, "learning_rate": 1.8500830564784053e-07, "loss": 0.1427, "step": 3693 }, { "epoch": 0.77, "learning_rate": 1.8497715946843855e-07, "loss": 0.199, "step": 3694 }, { "epoch": 0.77, "learning_rate": 1.8494601328903656e-07, "loss": 0.2533, "step": 3695 }, { "epoch": 0.77, "learning_rate": 1.8491486710963452e-07, "loss": 0.2767, "step": 3696 }, { "epoch": 0.77, "learning_rate": 1.8488372093023254e-07, "loss": 0.234, "step": 3697 }, { "epoch": 0.77, "learning_rate": 1.8485257475083056e-07, "loss": 0.3038, "step": 3698 }, { "epoch": 0.77, "learning_rate": 1.8482142857142857e-07, "loss": 0.1716, "step": 3699 }, { "epoch": 0.77, "learning_rate": 1.847902823920266e-07, "loss": 0.1936, "step": 3700 }, { "epoch": 0.77, "learning_rate": 1.8475913621262458e-07, "loss": 0.2044, "step": 3701 }, { "epoch": 0.77, "learning_rate": 1.847279900332226e-07, "loss": 0.2232, "step": 3702 }, { "epoch": 0.77, "learning_rate": 1.8469684385382058e-07, "loss": 0.1661, "step": 3703 }, { "epoch": 0.77, "learning_rate": 1.846656976744186e-07, "loss": 0.1719, "step": 3704 }, { "epoch": 0.77, "learning_rate": 1.846345514950166e-07, "loss": 0.1903, "step": 3705 }, { "epoch": 0.77, "learning_rate": 1.846034053156146e-07, "loss": 0.2421, "step": 3706 }, { "epoch": 0.77, "learning_rate": 1.8457225913621262e-07, "loss": 0.2125, "step": 3707 }, { "epoch": 0.77, "learning_rate": 1.8454111295681063e-07, "loss": 0.2097, "step": 3708 }, { "epoch": 0.77, "learning_rate": 1.8450996677740865e-07, "loss": 0.1586, "step": 3709 }, { "epoch": 0.77, "learning_rate": 1.8447882059800661e-07, "loss": 0.1641, "step": 3710 }, { "epoch": 0.77, "learning_rate": 1.8444767441860463e-07, "loss": 0.2777, "step": 3711 }, { "epoch": 0.77, "learning_rate": 1.8441652823920264e-07, "loss": 0.1692, "step": 3712 }, { "epoch": 0.77, "learning_rate": 1.8438538205980066e-07, "loss": 0.1795, "step": 3713 }, { "epoch": 0.77, "learning_rate": 1.8435423588039868e-07, "loss": 0.1761, "step": 3714 }, { "epoch": 0.77, "learning_rate": 1.8432308970099667e-07, "loss": 0.2666, "step": 3715 }, { "epoch": 0.77, "learning_rate": 1.8429194352159468e-07, "loss": 0.2331, "step": 3716 }, { "epoch": 0.77, "learning_rate": 1.842607973421927e-07, "loss": 0.1988, "step": 3717 }, { "epoch": 0.77, "learning_rate": 1.842296511627907e-07, "loss": 0.1363, "step": 3718 }, { "epoch": 0.77, "learning_rate": 1.8419850498338868e-07, "loss": 0.2038, "step": 3719 }, { "epoch": 0.77, "learning_rate": 1.841673588039867e-07, "loss": 0.2042, "step": 3720 }, { "epoch": 0.77, "learning_rate": 1.841362126245847e-07, "loss": 0.2112, "step": 3721 }, { "epoch": 0.77, "learning_rate": 1.8410506644518272e-07, "loss": 0.1872, "step": 3722 }, { "epoch": 0.77, "learning_rate": 1.8407392026578074e-07, "loss": 0.2244, "step": 3723 }, { "epoch": 0.77, "learning_rate": 1.8404277408637873e-07, "loss": 0.2137, "step": 3724 }, { "epoch": 0.77, "learning_rate": 1.8401162790697672e-07, "loss": 0.1709, "step": 3725 }, { "epoch": 0.77, "learning_rate": 1.8398048172757473e-07, "loss": 0.1733, "step": 3726 }, { "epoch": 0.77, "learning_rate": 1.8394933554817275e-07, "loss": 0.182, "step": 3727 }, { "epoch": 0.77, "learning_rate": 1.8391818936877077e-07, "loss": 0.2767, "step": 3728 }, { "epoch": 0.77, "learning_rate": 1.8388704318936876e-07, "loss": 0.2122, "step": 3729 }, { "epoch": 0.77, "learning_rate": 1.8385589700996677e-07, "loss": 0.1715, "step": 3730 }, { "epoch": 0.77, "learning_rate": 1.838247508305648e-07, "loss": 0.2143, "step": 3731 }, { "epoch": 0.77, "learning_rate": 1.837936046511628e-07, "loss": 0.1987, "step": 3732 }, { "epoch": 0.78, "learning_rate": 1.8376245847176077e-07, "loss": 0.1562, "step": 3733 }, { "epoch": 0.78, "learning_rate": 1.8373131229235878e-07, "loss": 0.2326, "step": 3734 }, { "epoch": 0.78, "learning_rate": 1.837001661129568e-07, "loss": 0.2015, "step": 3735 }, { "epoch": 0.78, "learning_rate": 1.836690199335548e-07, "loss": 0.1684, "step": 3736 }, { "epoch": 0.78, "learning_rate": 1.8363787375415283e-07, "loss": 0.1699, "step": 3737 }, { "epoch": 0.78, "learning_rate": 1.8360672757475082e-07, "loss": 0.2447, "step": 3738 }, { "epoch": 0.78, "learning_rate": 1.8357558139534883e-07, "loss": 0.2347, "step": 3739 }, { "epoch": 0.78, "learning_rate": 1.8354443521594682e-07, "loss": 0.1986, "step": 3740 }, { "epoch": 0.78, "learning_rate": 1.8351328903654484e-07, "loss": 0.1576, "step": 3741 }, { "epoch": 0.78, "learning_rate": 1.8348214285714285e-07, "loss": 0.1753, "step": 3742 }, { "epoch": 0.78, "learning_rate": 1.8345099667774084e-07, "loss": 0.18, "step": 3743 }, { "epoch": 0.78, "learning_rate": 1.8341985049833886e-07, "loss": 0.2017, "step": 3744 }, { "epoch": 0.78, "learning_rate": 1.8338870431893688e-07, "loss": 0.1905, "step": 3745 }, { "epoch": 0.78, "learning_rate": 1.833575581395349e-07, "loss": 0.1757, "step": 3746 }, { "epoch": 0.78, "learning_rate": 1.833264119601329e-07, "loss": 0.2148, "step": 3747 }, { "epoch": 0.78, "learning_rate": 1.8329526578073087e-07, "loss": 0.1724, "step": 3748 }, { "epoch": 0.78, "learning_rate": 1.8326411960132889e-07, "loss": 0.1795, "step": 3749 }, { "epoch": 0.78, "learning_rate": 1.832329734219269e-07, "loss": 0.1791, "step": 3750 }, { "epoch": 0.78, "learning_rate": 1.8320182724252492e-07, "loss": 0.1946, "step": 3751 }, { "epoch": 0.78, "learning_rate": 1.831706810631229e-07, "loss": 0.1816, "step": 3752 }, { "epoch": 0.78, "learning_rate": 1.8313953488372092e-07, "loss": 0.1959, "step": 3753 }, { "epoch": 0.78, "learning_rate": 1.8310838870431894e-07, "loss": 0.2027, "step": 3754 }, { "epoch": 0.78, "learning_rate": 1.8307724252491695e-07, "loss": 0.2104, "step": 3755 }, { "epoch": 0.78, "learning_rate": 1.8304609634551494e-07, "loss": 0.1829, "step": 3756 }, { "epoch": 0.78, "learning_rate": 1.8301495016611293e-07, "loss": 0.1672, "step": 3757 }, { "epoch": 0.78, "learning_rate": 1.8298380398671095e-07, "loss": 0.1937, "step": 3758 }, { "epoch": 0.78, "learning_rate": 1.8295265780730897e-07, "loss": 0.2215, "step": 3759 }, { "epoch": 0.78, "learning_rate": 1.8292151162790698e-07, "loss": 0.157, "step": 3760 }, { "epoch": 0.78, "learning_rate": 1.82890365448505e-07, "loss": 0.1766, "step": 3761 }, { "epoch": 0.78, "learning_rate": 1.8285921926910296e-07, "loss": 0.214, "step": 3762 }, { "epoch": 0.78, "learning_rate": 1.8282807308970098e-07, "loss": 0.1564, "step": 3763 }, { "epoch": 0.78, "learning_rate": 1.82796926910299e-07, "loss": 0.1423, "step": 3764 }, { "epoch": 0.78, "learning_rate": 1.82765780730897e-07, "loss": 0.2619, "step": 3765 }, { "epoch": 0.78, "learning_rate": 1.82734634551495e-07, "loss": 0.154, "step": 3766 }, { "epoch": 0.78, "learning_rate": 1.82703488372093e-07, "loss": 0.2034, "step": 3767 }, { "epoch": 0.78, "learning_rate": 1.8267234219269103e-07, "loss": 0.2064, "step": 3768 }, { "epoch": 0.78, "learning_rate": 1.8264119601328904e-07, "loss": 0.145, "step": 3769 }, { "epoch": 0.78, "learning_rate": 1.8261004983388706e-07, "loss": 0.2639, "step": 3770 }, { "epoch": 0.78, "learning_rate": 1.8257890365448502e-07, "loss": 0.1665, "step": 3771 }, { "epoch": 0.78, "learning_rate": 1.8254775747508304e-07, "loss": 0.2505, "step": 3772 }, { "epoch": 0.78, "learning_rate": 1.8251661129568105e-07, "loss": 0.1796, "step": 3773 }, { "epoch": 0.78, "learning_rate": 1.8248546511627907e-07, "loss": 0.1593, "step": 3774 }, { "epoch": 0.78, "learning_rate": 1.8245431893687709e-07, "loss": 0.1838, "step": 3775 }, { "epoch": 0.78, "learning_rate": 1.8242317275747508e-07, "loss": 0.1764, "step": 3776 }, { "epoch": 0.78, "learning_rate": 1.8239202657807306e-07, "loss": 0.1844, "step": 3777 }, { "epoch": 0.78, "learning_rate": 1.8236088039867108e-07, "loss": 0.1761, "step": 3778 }, { "epoch": 0.78, "learning_rate": 1.823297342192691e-07, "loss": 0.1481, "step": 3779 }, { "epoch": 0.78, "learning_rate": 1.8229858803986709e-07, "loss": 0.1841, "step": 3780 }, { "epoch": 0.79, "learning_rate": 1.822674418604651e-07, "loss": 0.197, "step": 3781 }, { "epoch": 0.79, "learning_rate": 1.8223629568106312e-07, "loss": 0.1442, "step": 3782 }, { "epoch": 0.79, "learning_rate": 1.8220514950166113e-07, "loss": 0.2141, "step": 3783 }, { "epoch": 0.79, "learning_rate": 1.8217400332225915e-07, "loss": 0.1938, "step": 3784 }, { "epoch": 0.79, "learning_rate": 1.821428571428571e-07, "loss": 0.2224, "step": 3785 }, { "epoch": 0.79, "learning_rate": 1.8211171096345513e-07, "loss": 0.2276, "step": 3786 }, { "epoch": 0.79, "learning_rate": 1.8208056478405314e-07, "loss": 0.2399, "step": 3787 }, { "epoch": 0.79, "learning_rate": 1.8204941860465116e-07, "loss": 0.1846, "step": 3788 }, { "epoch": 0.79, "learning_rate": 1.8201827242524918e-07, "loss": 0.1621, "step": 3789 }, { "epoch": 0.79, "learning_rate": 1.8198712624584716e-07, "loss": 0.2294, "step": 3790 }, { "epoch": 0.79, "learning_rate": 1.8195598006644518e-07, "loss": 0.2093, "step": 3791 }, { "epoch": 0.79, "learning_rate": 1.819248338870432e-07, "loss": 0.1608, "step": 3792 }, { "epoch": 0.79, "learning_rate": 1.8189368770764119e-07, "loss": 0.1833, "step": 3793 }, { "epoch": 0.79, "learning_rate": 1.818625415282392e-07, "loss": 0.1778, "step": 3794 }, { "epoch": 0.79, "learning_rate": 1.818313953488372e-07, "loss": 0.1999, "step": 3795 }, { "epoch": 0.79, "learning_rate": 1.818002491694352e-07, "loss": 0.2127, "step": 3796 }, { "epoch": 0.79, "learning_rate": 1.8176910299003322e-07, "loss": 0.221, "step": 3797 }, { "epoch": 0.79, "learning_rate": 1.8173795681063124e-07, "loss": 0.2056, "step": 3798 }, { "epoch": 0.79, "learning_rate": 1.817068106312292e-07, "loss": 0.1654, "step": 3799 }, { "epoch": 0.79, "learning_rate": 1.8167566445182722e-07, "loss": 0.1984, "step": 3800 }, { "epoch": 0.79, "learning_rate": 1.8164451827242523e-07, "loss": 0.2228, "step": 3801 }, { "epoch": 0.79, "learning_rate": 1.8161337209302325e-07, "loss": 0.1725, "step": 3802 }, { "epoch": 0.79, "learning_rate": 1.8158222591362126e-07, "loss": 0.2148, "step": 3803 }, { "epoch": 0.79, "learning_rate": 1.8155107973421925e-07, "loss": 0.2267, "step": 3804 }, { "epoch": 0.79, "learning_rate": 1.8151993355481727e-07, "loss": 0.2234, "step": 3805 }, { "epoch": 0.79, "learning_rate": 1.8148878737541529e-07, "loss": 0.1837, "step": 3806 }, { "epoch": 0.79, "learning_rate": 1.814576411960133e-07, "loss": 0.2365, "step": 3807 }, { "epoch": 0.79, "learning_rate": 1.814264950166113e-07, "loss": 0.1989, "step": 3808 }, { "epoch": 0.79, "learning_rate": 1.8139534883720928e-07, "loss": 0.1827, "step": 3809 }, { "epoch": 0.79, "learning_rate": 1.813642026578073e-07, "loss": 0.193, "step": 3810 }, { "epoch": 0.79, "learning_rate": 1.813330564784053e-07, "loss": 0.1961, "step": 3811 }, { "epoch": 0.79, "learning_rate": 1.8130191029900333e-07, "loss": 0.1572, "step": 3812 }, { "epoch": 0.79, "learning_rate": 1.8127076411960132e-07, "loss": 0.2148, "step": 3813 }, { "epoch": 0.79, "learning_rate": 1.812396179401993e-07, "loss": 0.2394, "step": 3814 }, { "epoch": 0.79, "learning_rate": 1.8120847176079732e-07, "loss": 0.2149, "step": 3815 }, { "epoch": 0.79, "learning_rate": 1.8117732558139534e-07, "loss": 0.1667, "step": 3816 }, { "epoch": 0.79, "learning_rate": 1.8114617940199335e-07, "loss": 0.1487, "step": 3817 }, { "epoch": 0.79, "learning_rate": 1.8111503322259134e-07, "loss": 0.1945, "step": 3818 }, { "epoch": 0.79, "learning_rate": 1.8108388704318936e-07, "loss": 0.238, "step": 3819 }, { "epoch": 0.79, "learning_rate": 1.8105274086378737e-07, "loss": 0.1846, "step": 3820 }, { "epoch": 0.79, "learning_rate": 1.810215946843854e-07, "loss": 0.2214, "step": 3821 }, { "epoch": 0.79, "learning_rate": 1.809904485049834e-07, "loss": 0.1571, "step": 3822 }, { "epoch": 0.79, "learning_rate": 1.8095930232558137e-07, "loss": 0.1771, "step": 3823 }, { "epoch": 0.79, "learning_rate": 1.8092815614617938e-07, "loss": 0.1829, "step": 3824 }, { "epoch": 0.79, "learning_rate": 1.808970099667774e-07, "loss": 0.2332, "step": 3825 }, { "epoch": 0.79, "learning_rate": 1.8086586378737542e-07, "loss": 0.2675, "step": 3826 }, { "epoch": 0.79, "learning_rate": 1.808347176079734e-07, "loss": 0.1626, "step": 3827 }, { "epoch": 0.79, "learning_rate": 1.8080357142857142e-07, "loss": 0.1247, "step": 3828 }, { "epoch": 0.8, "learning_rate": 1.8077242524916944e-07, "loss": 0.1421, "step": 3829 }, { "epoch": 0.8, "learning_rate": 1.8074127906976743e-07, "loss": 0.2422, "step": 3830 }, { "epoch": 0.8, "learning_rate": 1.8071013289036544e-07, "loss": 0.1985, "step": 3831 }, { "epoch": 0.8, "learning_rate": 1.8067898671096343e-07, "loss": 0.2108, "step": 3832 }, { "epoch": 0.8, "learning_rate": 1.8064784053156145e-07, "loss": 0.227, "step": 3833 }, { "epoch": 0.8, "learning_rate": 1.8061669435215946e-07, "loss": 0.2142, "step": 3834 }, { "epoch": 0.8, "learning_rate": 1.8058554817275748e-07, "loss": 0.1873, "step": 3835 }, { "epoch": 0.8, "learning_rate": 1.805544019933555e-07, "loss": 0.1682, "step": 3836 }, { "epoch": 0.8, "learning_rate": 1.8052325581395346e-07, "loss": 0.2174, "step": 3837 }, { "epoch": 0.8, "learning_rate": 1.8049210963455147e-07, "loss": 0.2015, "step": 3838 }, { "epoch": 0.8, "learning_rate": 1.804609634551495e-07, "loss": 0.2165, "step": 3839 }, { "epoch": 0.8, "learning_rate": 1.804298172757475e-07, "loss": 0.2001, "step": 3840 }, { "epoch": 0.8, "learning_rate": 1.8039867109634552e-07, "loss": 0.1961, "step": 3841 }, { "epoch": 0.8, "learning_rate": 1.803675249169435e-07, "loss": 0.216, "step": 3842 }, { "epoch": 0.8, "learning_rate": 1.8033637873754153e-07, "loss": 0.1954, "step": 3843 }, { "epoch": 0.8, "learning_rate": 1.8030523255813954e-07, "loss": 0.1911, "step": 3844 }, { "epoch": 0.8, "learning_rate": 1.8027408637873753e-07, "loss": 0.2382, "step": 3845 }, { "epoch": 0.8, "learning_rate": 1.8024294019933552e-07, "loss": 0.1782, "step": 3846 }, { "epoch": 0.8, "learning_rate": 1.8021179401993354e-07, "loss": 0.1915, "step": 3847 }, { "epoch": 0.8, "learning_rate": 1.8018064784053155e-07, "loss": 0.1905, "step": 3848 }, { "epoch": 0.8, "learning_rate": 1.8014950166112957e-07, "loss": 0.1791, "step": 3849 }, { "epoch": 0.8, "learning_rate": 1.8011835548172758e-07, "loss": 0.3007, "step": 3850 }, { "epoch": 0.8, "learning_rate": 1.8008720930232555e-07, "loss": 0.2022, "step": 3851 }, { "epoch": 0.8, "learning_rate": 1.8005606312292356e-07, "loss": 0.1918, "step": 3852 }, { "epoch": 0.8, "learning_rate": 1.8002491694352158e-07, "loss": 0.1782, "step": 3853 }, { "epoch": 0.8, "learning_rate": 1.799937707641196e-07, "loss": 0.1412, "step": 3854 }, { "epoch": 0.8, "learning_rate": 1.799626245847176e-07, "loss": 0.1886, "step": 3855 }, { "epoch": 0.8, "learning_rate": 1.799314784053156e-07, "loss": 0.2265, "step": 3856 }, { "epoch": 0.8, "learning_rate": 1.7990033222591362e-07, "loss": 0.1599, "step": 3857 }, { "epoch": 0.8, "learning_rate": 1.7986918604651163e-07, "loss": 0.1663, "step": 3858 }, { "epoch": 0.8, "learning_rate": 1.7983803986710965e-07, "loss": 0.2303, "step": 3859 }, { "epoch": 0.8, "learning_rate": 1.798068936877076e-07, "loss": 0.1569, "step": 3860 }, { "epoch": 0.8, "learning_rate": 1.7977574750830563e-07, "loss": 0.1901, "step": 3861 }, { "epoch": 0.8, "learning_rate": 1.7974460132890364e-07, "loss": 0.19, "step": 3862 }, { "epoch": 0.8, "learning_rate": 1.7971345514950166e-07, "loss": 0.1404, "step": 3863 }, { "epoch": 0.8, "learning_rate": 1.7968230897009967e-07, "loss": 0.1489, "step": 3864 }, { "epoch": 0.8, "learning_rate": 1.7965116279069766e-07, "loss": 0.1734, "step": 3865 }, { "epoch": 0.8, "learning_rate": 1.7962001661129568e-07, "loss": 0.2184, "step": 3866 }, { "epoch": 0.8, "learning_rate": 1.7958887043189367e-07, "loss": 0.1355, "step": 3867 }, { "epoch": 0.8, "learning_rate": 1.7955772425249168e-07, "loss": 0.2069, "step": 3868 }, { "epoch": 0.8, "learning_rate": 1.795265780730897e-07, "loss": 0.1816, "step": 3869 }, { "epoch": 0.8, "learning_rate": 1.794954318936877e-07, "loss": 0.1765, "step": 3870 }, { "epoch": 0.8, "learning_rate": 1.794642857142857e-07, "loss": 0.2772, "step": 3871 }, { "epoch": 0.8, "learning_rate": 1.7943313953488372e-07, "loss": 0.1964, "step": 3872 }, { "epoch": 0.8, "learning_rate": 1.7940199335548174e-07, "loss": 0.1787, "step": 3873 }, { "epoch": 0.8, "learning_rate": 1.7937084717607975e-07, "loss": 0.1849, "step": 3874 }, { "epoch": 0.8, "learning_rate": 1.7933970099667772e-07, "loss": 0.1616, "step": 3875 }, { "epoch": 0.8, "learning_rate": 1.7930855481727573e-07, "loss": 0.178, "step": 3876 }, { "epoch": 0.81, "learning_rate": 1.7927740863787375e-07, "loss": 0.2361, "step": 3877 }, { "epoch": 0.81, "learning_rate": 1.7924626245847176e-07, "loss": 0.2173, "step": 3878 }, { "epoch": 0.81, "learning_rate": 1.7921511627906975e-07, "loss": 0.2994, "step": 3879 }, { "epoch": 0.81, "learning_rate": 1.7918397009966777e-07, "loss": 0.1972, "step": 3880 }, { "epoch": 0.81, "learning_rate": 1.7915282392026578e-07, "loss": 0.1189, "step": 3881 }, { "epoch": 0.81, "learning_rate": 1.7912167774086377e-07, "loss": 0.1259, "step": 3882 }, { "epoch": 0.81, "learning_rate": 1.790905315614618e-07, "loss": 0.247, "step": 3883 }, { "epoch": 0.81, "learning_rate": 1.7905938538205978e-07, "loss": 0.2331, "step": 3884 }, { "epoch": 0.81, "learning_rate": 1.790282392026578e-07, "loss": 0.1532, "step": 3885 }, { "epoch": 0.81, "learning_rate": 1.789970930232558e-07, "loss": 0.1834, "step": 3886 }, { "epoch": 0.81, "learning_rate": 1.7896594684385383e-07, "loss": 0.1921, "step": 3887 }, { "epoch": 0.81, "learning_rate": 1.7893480066445184e-07, "loss": 0.2107, "step": 3888 }, { "epoch": 0.81, "learning_rate": 1.789036544850498e-07, "loss": 0.1496, "step": 3889 }, { "epoch": 0.81, "learning_rate": 1.7887250830564782e-07, "loss": 0.2279, "step": 3890 }, { "epoch": 0.81, "learning_rate": 1.7884136212624584e-07, "loss": 0.1984, "step": 3891 }, { "epoch": 0.81, "learning_rate": 1.7881021594684385e-07, "loss": 0.2329, "step": 3892 }, { "epoch": 0.81, "learning_rate": 1.7877906976744184e-07, "loss": 0.1776, "step": 3893 }, { "epoch": 0.81, "learning_rate": 1.7874792358803986e-07, "loss": 0.2491, "step": 3894 }, { "epoch": 0.81, "learning_rate": 1.7871677740863787e-07, "loss": 0.163, "step": 3895 }, { "epoch": 0.81, "learning_rate": 1.786856312292359e-07, "loss": 0.2316, "step": 3896 }, { "epoch": 0.81, "learning_rate": 1.786544850498339e-07, "loss": 0.1968, "step": 3897 }, { "epoch": 0.81, "learning_rate": 1.7862333887043187e-07, "loss": 0.2152, "step": 3898 }, { "epoch": 0.81, "learning_rate": 1.7859219269102988e-07, "loss": 0.1999, "step": 3899 }, { "epoch": 0.81, "learning_rate": 1.785610465116279e-07, "loss": 0.2383, "step": 3900 }, { "epoch": 0.81, "learning_rate": 1.7852990033222592e-07, "loss": 0.1763, "step": 3901 }, { "epoch": 0.81, "learning_rate": 1.7849875415282393e-07, "loss": 0.1542, "step": 3902 }, { "epoch": 0.81, "learning_rate": 1.7846760797342192e-07, "loss": 0.2045, "step": 3903 }, { "epoch": 0.81, "learning_rate": 1.784364617940199e-07, "loss": 0.1649, "step": 3904 }, { "epoch": 0.81, "learning_rate": 1.7840531561461793e-07, "loss": 0.2513, "step": 3905 }, { "epoch": 0.81, "learning_rate": 1.7837416943521594e-07, "loss": 0.1975, "step": 3906 }, { "epoch": 0.81, "learning_rate": 1.7834302325581393e-07, "loss": 0.1729, "step": 3907 }, { "epoch": 0.81, "learning_rate": 1.7831187707641195e-07, "loss": 0.2357, "step": 3908 }, { "epoch": 0.81, "learning_rate": 1.7828073089700996e-07, "loss": 0.2078, "step": 3909 }, { "epoch": 0.81, "learning_rate": 1.7824958471760798e-07, "loss": 0.1841, "step": 3910 }, { "epoch": 0.81, "learning_rate": 1.78218438538206e-07, "loss": 0.2256, "step": 3911 }, { "epoch": 0.81, "learning_rate": 1.7818729235880396e-07, "loss": 0.1703, "step": 3912 }, { "epoch": 0.81, "learning_rate": 1.7815614617940197e-07, "loss": 0.1591, "step": 3913 }, { "epoch": 0.81, "learning_rate": 1.78125e-07, "loss": 0.1765, "step": 3914 }, { "epoch": 0.81, "learning_rate": 1.78093853820598e-07, "loss": 0.2458, "step": 3915 }, { "epoch": 0.81, "learning_rate": 1.7806270764119602e-07, "loss": 0.1637, "step": 3916 }, { "epoch": 0.81, "learning_rate": 1.78031561461794e-07, "loss": 0.1643, "step": 3917 }, { "epoch": 0.81, "learning_rate": 1.7800041528239203e-07, "loss": 0.2535, "step": 3918 }, { "epoch": 0.81, "learning_rate": 1.7796926910299001e-07, "loss": 0.1687, "step": 3919 }, { "epoch": 0.81, "learning_rate": 1.7793812292358803e-07, "loss": 0.1582, "step": 3920 }, { "epoch": 0.81, "learning_rate": 1.7790697674418605e-07, "loss": 0.1932, "step": 3921 }, { "epoch": 0.81, "learning_rate": 1.7787583056478404e-07, "loss": 0.1665, "step": 3922 }, { "epoch": 0.81, "learning_rate": 1.7784468438538205e-07, "loss": 0.2077, "step": 3923 }, { "epoch": 0.81, "learning_rate": 1.7781353820598007e-07, "loss": 0.25, "step": 3924 }, { "epoch": 0.81, "learning_rate": 1.7778239202657808e-07, "loss": 0.1579, "step": 3925 }, { "epoch": 0.82, "learning_rate": 1.7775124584717605e-07, "loss": 0.1391, "step": 3926 }, { "epoch": 0.82, "learning_rate": 1.7772009966777406e-07, "loss": 0.2298, "step": 3927 }, { "epoch": 0.82, "learning_rate": 1.7768895348837208e-07, "loss": 0.2248, "step": 3928 }, { "epoch": 0.82, "learning_rate": 1.776578073089701e-07, "loss": 0.1664, "step": 3929 }, { "epoch": 0.82, "learning_rate": 1.776266611295681e-07, "loss": 0.2905, "step": 3930 }, { "epoch": 0.82, "learning_rate": 1.775955149501661e-07, "loss": 0.2269, "step": 3931 }, { "epoch": 0.82, "learning_rate": 1.7756436877076411e-07, "loss": 0.1853, "step": 3932 }, { "epoch": 0.82, "learning_rate": 1.7753322259136213e-07, "loss": 0.1532, "step": 3933 }, { "epoch": 0.82, "learning_rate": 1.7750207641196015e-07, "loss": 0.248, "step": 3934 }, { "epoch": 0.82, "learning_rate": 1.7747093023255814e-07, "loss": 0.1529, "step": 3935 }, { "epoch": 0.82, "learning_rate": 1.7743978405315613e-07, "loss": 0.2002, "step": 3936 }, { "epoch": 0.82, "learning_rate": 1.7740863787375414e-07, "loss": 0.1649, "step": 3937 }, { "epoch": 0.82, "learning_rate": 1.7737749169435216e-07, "loss": 0.1691, "step": 3938 }, { "epoch": 0.82, "learning_rate": 1.7734634551495017e-07, "loss": 0.1964, "step": 3939 }, { "epoch": 0.82, "learning_rate": 1.7731519933554816e-07, "loss": 0.2032, "step": 3940 }, { "epoch": 0.82, "learning_rate": 1.7728405315614615e-07, "loss": 0.2303, "step": 3941 }, { "epoch": 0.82, "learning_rate": 1.7725290697674417e-07, "loss": 0.2189, "step": 3942 }, { "epoch": 0.82, "learning_rate": 1.7722176079734218e-07, "loss": 0.2124, "step": 3943 }, { "epoch": 0.82, "learning_rate": 1.771906146179402e-07, "loss": 0.1711, "step": 3944 }, { "epoch": 0.82, "learning_rate": 1.771594684385382e-07, "loss": 0.2023, "step": 3945 }, { "epoch": 0.82, "learning_rate": 1.771283222591362e-07, "loss": 0.2366, "step": 3946 }, { "epoch": 0.82, "learning_rate": 1.7709717607973422e-07, "loss": 0.1879, "step": 3947 }, { "epoch": 0.82, "learning_rate": 1.7706602990033224e-07, "loss": 0.1789, "step": 3948 }, { "epoch": 0.82, "learning_rate": 1.7703488372093025e-07, "loss": 0.2288, "step": 3949 }, { "epoch": 0.82, "learning_rate": 1.7700373754152821e-07, "loss": 0.1893, "step": 3950 }, { "epoch": 0.82, "learning_rate": 1.7697259136212623e-07, "loss": 0.1673, "step": 3951 }, { "epoch": 0.82, "learning_rate": 1.7694144518272425e-07, "loss": 0.1913, "step": 3952 }, { "epoch": 0.82, "learning_rate": 1.7691029900332226e-07, "loss": 0.2093, "step": 3953 }, { "epoch": 0.82, "learning_rate": 1.7687915282392025e-07, "loss": 0.1727, "step": 3954 }, { "epoch": 0.82, "learning_rate": 1.7684800664451827e-07, "loss": 0.1709, "step": 3955 }, { "epoch": 0.82, "learning_rate": 1.7681686046511626e-07, "loss": 0.1285, "step": 3956 }, { "epoch": 0.82, "learning_rate": 1.7678571428571427e-07, "loss": 0.1382, "step": 3957 }, { "epoch": 0.82, "learning_rate": 1.767545681063123e-07, "loss": 0.2066, "step": 3958 }, { "epoch": 0.82, "learning_rate": 1.7672342192691028e-07, "loss": 0.1927, "step": 3959 }, { "epoch": 0.82, "learning_rate": 1.766922757475083e-07, "loss": 0.24, "step": 3960 }, { "epoch": 0.82, "learning_rate": 1.766611295681063e-07, "loss": 0.2678, "step": 3961 }, { "epoch": 0.82, "learning_rate": 1.7662998338870432e-07, "loss": 0.1719, "step": 3962 }, { "epoch": 0.82, "learning_rate": 1.7659883720930234e-07, "loss": 0.2013, "step": 3963 }, { "epoch": 0.82, "learning_rate": 1.765676910299003e-07, "loss": 0.1959, "step": 3964 }, { "epoch": 0.82, "learning_rate": 1.7653654485049832e-07, "loss": 0.1918, "step": 3965 }, { "epoch": 0.82, "learning_rate": 1.7650539867109634e-07, "loss": 0.2608, "step": 3966 }, { "epoch": 0.82, "learning_rate": 1.7647425249169435e-07, "loss": 0.1761, "step": 3967 }, { "epoch": 0.82, "learning_rate": 1.7644310631229237e-07, "loss": 0.1601, "step": 3968 }, { "epoch": 0.82, "learning_rate": 1.7641196013289036e-07, "loss": 0.1971, "step": 3969 }, { "epoch": 0.82, "learning_rate": 1.7638081395348837e-07, "loss": 0.1825, "step": 3970 }, { "epoch": 0.82, "learning_rate": 1.763496677740864e-07, "loss": 0.1395, "step": 3971 }, { "epoch": 0.82, "learning_rate": 1.7631852159468438e-07, "loss": 0.1605, "step": 3972 }, { "epoch": 0.82, "learning_rate": 1.7628737541528237e-07, "loss": 0.2007, "step": 3973 }, { "epoch": 0.83, "learning_rate": 1.7625622923588038e-07, "loss": 0.1306, "step": 3974 }, { "epoch": 0.83, "learning_rate": 1.762250830564784e-07, "loss": 0.1852, "step": 3975 }, { "epoch": 0.83, "learning_rate": 1.7619393687707641e-07, "loss": 0.1879, "step": 3976 }, { "epoch": 0.83, "learning_rate": 1.7616279069767443e-07, "loss": 0.234, "step": 3977 }, { "epoch": 0.83, "learning_rate": 1.761316445182724e-07, "loss": 0.1688, "step": 3978 }, { "epoch": 0.83, "learning_rate": 1.761004983388704e-07, "loss": 0.1643, "step": 3979 }, { "epoch": 0.83, "learning_rate": 1.7606935215946842e-07, "loss": 0.2512, "step": 3980 }, { "epoch": 0.83, "learning_rate": 1.7603820598006644e-07, "loss": 0.1999, "step": 3981 }, { "epoch": 0.83, "learning_rate": 1.7600705980066446e-07, "loss": 0.2125, "step": 3982 }, { "epoch": 0.83, "learning_rate": 1.7597591362126245e-07, "loss": 0.2485, "step": 3983 }, { "epoch": 0.83, "learning_rate": 1.7594476744186046e-07, "loss": 0.2618, "step": 3984 }, { "epoch": 0.83, "learning_rate": 1.7591362126245848e-07, "loss": 0.186, "step": 3985 }, { "epoch": 0.83, "learning_rate": 1.758824750830565e-07, "loss": 0.2313, "step": 3986 }, { "epoch": 0.83, "learning_rate": 1.7585132890365446e-07, "loss": 0.2209, "step": 3987 }, { "epoch": 0.83, "learning_rate": 1.7582018272425247e-07, "loss": 0.1899, "step": 3988 }, { "epoch": 0.83, "learning_rate": 1.757890365448505e-07, "loss": 0.1869, "step": 3989 }, { "epoch": 0.83, "learning_rate": 1.757578903654485e-07, "loss": 0.1545, "step": 3990 }, { "epoch": 0.83, "learning_rate": 1.7572674418604652e-07, "loss": 0.2636, "step": 3991 }, { "epoch": 0.83, "learning_rate": 1.756955980066445e-07, "loss": 0.245, "step": 3992 }, { "epoch": 0.83, "learning_rate": 1.756644518272425e-07, "loss": 0.2215, "step": 3993 }, { "epoch": 0.83, "learning_rate": 1.7563330564784051e-07, "loss": 0.1995, "step": 3994 }, { "epoch": 0.83, "learning_rate": 1.7560215946843853e-07, "loss": 0.1594, "step": 3995 }, { "epoch": 0.83, "learning_rate": 1.7557101328903655e-07, "loss": 0.2025, "step": 3996 }, { "epoch": 0.83, "learning_rate": 1.7553986710963453e-07, "loss": 0.1768, "step": 3997 }, { "epoch": 0.83, "learning_rate": 1.7550872093023255e-07, "loss": 0.2049, "step": 3998 }, { "epoch": 0.83, "learning_rate": 1.7547757475083057e-07, "loss": 0.2225, "step": 3999 }, { "epoch": 0.83, "learning_rate": 1.7544642857142858e-07, "loss": 0.1452, "step": 4000 }, { "epoch": 0.83, "learning_rate": 1.754152823920266e-07, "loss": 0.1633, "step": 4001 }, { "epoch": 0.83, "learning_rate": 1.7538413621262456e-07, "loss": 0.207, "step": 4002 }, { "epoch": 0.83, "learning_rate": 1.7535299003322258e-07, "loss": 0.155, "step": 4003 }, { "epoch": 0.83, "learning_rate": 1.753218438538206e-07, "loss": 0.2373, "step": 4004 }, { "epoch": 0.83, "learning_rate": 1.752906976744186e-07, "loss": 0.1662, "step": 4005 }, { "epoch": 0.83, "learning_rate": 1.752595514950166e-07, "loss": 0.2356, "step": 4006 }, { "epoch": 0.83, "learning_rate": 1.7522840531561461e-07, "loss": 0.201, "step": 4007 }, { "epoch": 0.83, "learning_rate": 1.7519725913621263e-07, "loss": 0.2497, "step": 4008 }, { "epoch": 0.83, "learning_rate": 1.7516611295681062e-07, "loss": 0.1643, "step": 4009 }, { "epoch": 0.83, "learning_rate": 1.7513496677740863e-07, "loss": 0.2149, "step": 4010 }, { "epoch": 0.83, "learning_rate": 1.7510382059800662e-07, "loss": 0.238, "step": 4011 }, { "epoch": 0.83, "learning_rate": 1.7507267441860464e-07, "loss": 0.183, "step": 4012 }, { "epoch": 0.83, "learning_rate": 1.7504152823920266e-07, "loss": 0.304, "step": 4013 }, { "epoch": 0.83, "learning_rate": 1.7501038205980067e-07, "loss": 0.1591, "step": 4014 }, { "epoch": 0.83, "learning_rate": 1.749792358803987e-07, "loss": 0.2368, "step": 4015 }, { "epoch": 0.83, "learning_rate": 1.7494808970099665e-07, "loss": 0.1853, "step": 4016 }, { "epoch": 0.83, "learning_rate": 1.7491694352159467e-07, "loss": 0.1919, "step": 4017 }, { "epoch": 0.83, "learning_rate": 1.7488579734219268e-07, "loss": 0.1997, "step": 4018 }, { "epoch": 0.83, "learning_rate": 1.748546511627907e-07, "loss": 0.2533, "step": 4019 }, { "epoch": 0.83, "learning_rate": 1.7482350498338869e-07, "loss": 0.1641, "step": 4020 }, { "epoch": 0.83, "learning_rate": 1.747923588039867e-07, "loss": 0.1455, "step": 4021 }, { "epoch": 0.84, "learning_rate": 1.7476121262458472e-07, "loss": 0.2029, "step": 4022 }, { "epoch": 0.84, "learning_rate": 1.7473006644518273e-07, "loss": 0.2045, "step": 4023 }, { "epoch": 0.84, "learning_rate": 1.7469892026578072e-07, "loss": 0.1579, "step": 4024 }, { "epoch": 0.84, "learning_rate": 1.746677740863787e-07, "loss": 0.2165, "step": 4025 }, { "epoch": 0.84, "learning_rate": 1.7463662790697673e-07, "loss": 0.1754, "step": 4026 }, { "epoch": 0.84, "learning_rate": 1.7460548172757474e-07, "loss": 0.2677, "step": 4027 }, { "epoch": 0.84, "learning_rate": 1.7457433554817276e-07, "loss": 0.1477, "step": 4028 }, { "epoch": 0.84, "learning_rate": 1.7454318936877078e-07, "loss": 0.152, "step": 4029 }, { "epoch": 0.84, "learning_rate": 1.7451204318936874e-07, "loss": 0.1895, "step": 4030 }, { "epoch": 0.84, "learning_rate": 1.7448089700996675e-07, "loss": 0.1462, "step": 4031 }, { "epoch": 0.84, "learning_rate": 1.7444975083056477e-07, "loss": 0.2642, "step": 4032 }, { "epoch": 0.84, "learning_rate": 1.7441860465116279e-07, "loss": 0.1946, "step": 4033 }, { "epoch": 0.84, "learning_rate": 1.7438745847176078e-07, "loss": 0.2031, "step": 4034 }, { "epoch": 0.84, "learning_rate": 1.743563122923588e-07, "loss": 0.2501, "step": 4035 }, { "epoch": 0.84, "learning_rate": 1.743251661129568e-07, "loss": 0.1544, "step": 4036 }, { "epoch": 0.84, "learning_rate": 1.7429401993355482e-07, "loss": 0.1704, "step": 4037 }, { "epoch": 0.84, "learning_rate": 1.7426287375415284e-07, "loss": 0.1552, "step": 4038 }, { "epoch": 0.84, "learning_rate": 1.742317275747508e-07, "loss": 0.1658, "step": 4039 }, { "epoch": 0.84, "learning_rate": 1.7420058139534882e-07, "loss": 0.1603, "step": 4040 }, { "epoch": 0.84, "learning_rate": 1.7416943521594683e-07, "loss": 0.1392, "step": 4041 }, { "epoch": 0.84, "learning_rate": 1.7413828903654485e-07, "loss": 0.1466, "step": 4042 }, { "epoch": 0.84, "learning_rate": 1.7410714285714287e-07, "loss": 0.2297, "step": 4043 }, { "epoch": 0.84, "learning_rate": 1.7407599667774085e-07, "loss": 0.1746, "step": 4044 }, { "epoch": 0.84, "learning_rate": 1.7404485049833887e-07, "loss": 0.1597, "step": 4045 }, { "epoch": 0.84, "learning_rate": 1.7401370431893686e-07, "loss": 0.1917, "step": 4046 }, { "epoch": 0.84, "learning_rate": 1.7398255813953488e-07, "loss": 0.2745, "step": 4047 }, { "epoch": 0.84, "learning_rate": 1.739514119601329e-07, "loss": 0.1727, "step": 4048 }, { "epoch": 0.84, "learning_rate": 1.7392026578073088e-07, "loss": 0.2019, "step": 4049 }, { "epoch": 0.84, "learning_rate": 1.738891196013289e-07, "loss": 0.226, "step": 4050 }, { "epoch": 0.84, "learning_rate": 1.738579734219269e-07, "loss": 0.1427, "step": 4051 }, { "epoch": 0.84, "learning_rate": 1.7382682724252493e-07, "loss": 0.1357, "step": 4052 }, { "epoch": 0.84, "learning_rate": 1.737956810631229e-07, "loss": 0.1978, "step": 4053 }, { "epoch": 0.84, "learning_rate": 1.737645348837209e-07, "loss": 0.2002, "step": 4054 }, { "epoch": 0.84, "learning_rate": 1.7373338870431892e-07, "loss": 0.1985, "step": 4055 }, { "epoch": 0.84, "learning_rate": 1.7370224252491694e-07, "loss": 0.166, "step": 4056 }, { "epoch": 0.84, "learning_rate": 1.7367109634551495e-07, "loss": 0.1916, "step": 4057 }, { "epoch": 0.84, "learning_rate": 1.7363995016611294e-07, "loss": 0.1653, "step": 4058 }, { "epoch": 0.84, "learning_rate": 1.7360880398671096e-07, "loss": 0.1922, "step": 4059 }, { "epoch": 0.84, "learning_rate": 1.7357765780730898e-07, "loss": 0.1697, "step": 4060 }, { "epoch": 0.84, "learning_rate": 1.7354651162790696e-07, "loss": 0.1941, "step": 4061 }, { "epoch": 0.84, "learning_rate": 1.7351536544850498e-07, "loss": 0.1944, "step": 4062 }, { "epoch": 0.84, "learning_rate": 1.7348421926910297e-07, "loss": 0.1795, "step": 4063 }, { "epoch": 0.84, "learning_rate": 1.7345307308970099e-07, "loss": 0.2296, "step": 4064 }, { "epoch": 0.84, "learning_rate": 1.73421926910299e-07, "loss": 0.278, "step": 4065 }, { "epoch": 0.84, "learning_rate": 1.7339078073089702e-07, "loss": 0.1559, "step": 4066 }, { "epoch": 0.84, "learning_rate": 1.7335963455149498e-07, "loss": 0.2084, "step": 4067 }, { "epoch": 0.84, "learning_rate": 1.73328488372093e-07, "loss": 0.232, "step": 4068 }, { "epoch": 0.84, "learning_rate": 1.73297342192691e-07, "loss": 0.1821, "step": 4069 }, { "epoch": 0.85, "learning_rate": 1.7326619601328903e-07, "loss": 0.1402, "step": 4070 }, { "epoch": 0.85, "learning_rate": 1.7323504983388704e-07, "loss": 0.184, "step": 4071 }, { "epoch": 0.85, "learning_rate": 1.7320390365448503e-07, "loss": 0.2039, "step": 4072 }, { "epoch": 0.85, "learning_rate": 1.7317275747508305e-07, "loss": 0.2031, "step": 4073 }, { "epoch": 0.85, "learning_rate": 1.7314161129568106e-07, "loss": 0.1665, "step": 4074 }, { "epoch": 0.85, "learning_rate": 1.7311046511627908e-07, "loss": 0.1909, "step": 4075 }, { "epoch": 0.85, "learning_rate": 1.730793189368771e-07, "loss": 0.2112, "step": 4076 }, { "epoch": 0.85, "learning_rate": 1.7304817275747506e-07, "loss": 0.1528, "step": 4077 }, { "epoch": 0.85, "learning_rate": 1.7301702657807308e-07, "loss": 0.2464, "step": 4078 }, { "epoch": 0.85, "learning_rate": 1.729858803986711e-07, "loss": 0.2474, "step": 4079 }, { "epoch": 0.85, "learning_rate": 1.729547342192691e-07, "loss": 0.2406, "step": 4080 }, { "epoch": 0.85, "learning_rate": 1.729235880398671e-07, "loss": 0.2428, "step": 4081 }, { "epoch": 0.85, "learning_rate": 1.728924418604651e-07, "loss": 0.1939, "step": 4082 }, { "epoch": 0.85, "learning_rate": 1.728612956810631e-07, "loss": 0.1872, "step": 4083 }, { "epoch": 0.85, "learning_rate": 1.7283014950166112e-07, "loss": 0.1989, "step": 4084 }, { "epoch": 0.85, "learning_rate": 1.7279900332225913e-07, "loss": 0.2636, "step": 4085 }, { "epoch": 0.85, "learning_rate": 1.7276785714285712e-07, "loss": 0.1882, "step": 4086 }, { "epoch": 0.85, "learning_rate": 1.7273671096345514e-07, "loss": 0.1858, "step": 4087 }, { "epoch": 0.85, "learning_rate": 1.7270556478405315e-07, "loss": 0.184, "step": 4088 }, { "epoch": 0.85, "learning_rate": 1.7267441860465117e-07, "loss": 0.2241, "step": 4089 }, { "epoch": 0.85, "learning_rate": 1.7264327242524919e-07, "loss": 0.233, "step": 4090 }, { "epoch": 0.85, "learning_rate": 1.7261212624584715e-07, "loss": 0.1969, "step": 4091 }, { "epoch": 0.85, "learning_rate": 1.7258098006644516e-07, "loss": 0.2021, "step": 4092 }, { "epoch": 0.85, "learning_rate": 1.7254983388704318e-07, "loss": 0.169, "step": 4093 }, { "epoch": 0.85, "learning_rate": 1.725186877076412e-07, "loss": 0.1783, "step": 4094 }, { "epoch": 0.85, "learning_rate": 1.724875415282392e-07, "loss": 0.1643, "step": 4095 }, { "epoch": 0.85, "learning_rate": 1.724563953488372e-07, "loss": 0.2166, "step": 4096 }, { "epoch": 0.85, "learning_rate": 1.7242524916943522e-07, "loss": 0.2142, "step": 4097 }, { "epoch": 0.85, "learning_rate": 1.723941029900332e-07, "loss": 0.183, "step": 4098 }, { "epoch": 0.85, "learning_rate": 1.7236295681063122e-07, "loss": 0.171, "step": 4099 }, { "epoch": 0.85, "learning_rate": 1.723318106312292e-07, "loss": 0.1743, "step": 4100 }, { "epoch": 0.85, "learning_rate": 1.7230066445182723e-07, "loss": 0.1437, "step": 4101 }, { "epoch": 0.85, "learning_rate": 1.7226951827242524e-07, "loss": 0.1497, "step": 4102 }, { "epoch": 0.85, "learning_rate": 1.7223837209302326e-07, "loss": 0.2041, "step": 4103 }, { "epoch": 0.85, "learning_rate": 1.7220722591362127e-07, "loss": 0.2, "step": 4104 }, { "epoch": 0.85, "learning_rate": 1.7217607973421924e-07, "loss": 0.1883, "step": 4105 }, { "epoch": 0.85, "learning_rate": 1.7214493355481725e-07, "loss": 0.163, "step": 4106 }, { "epoch": 0.85, "learning_rate": 1.7211378737541527e-07, "loss": 0.213, "step": 4107 }, { "epoch": 0.85, "learning_rate": 1.7208264119601329e-07, "loss": 0.203, "step": 4108 }, { "epoch": 0.85, "learning_rate": 1.720514950166113e-07, "loss": 0.2446, "step": 4109 }, { "epoch": 0.85, "learning_rate": 1.720203488372093e-07, "loss": 0.2112, "step": 4110 }, { "epoch": 0.85, "learning_rate": 1.719892026578073e-07, "loss": 0.2247, "step": 4111 }, { "epoch": 0.85, "learning_rate": 1.7195805647840532e-07, "loss": 0.2097, "step": 4112 }, { "epoch": 0.85, "learning_rate": 1.7192691029900334e-07, "loss": 0.2193, "step": 4113 }, { "epoch": 0.85, "learning_rate": 1.718957641196013e-07, "loss": 0.2093, "step": 4114 }, { "epoch": 0.85, "learning_rate": 1.7186461794019932e-07, "loss": 0.2147, "step": 4115 }, { "epoch": 0.85, "learning_rate": 1.7183347176079733e-07, "loss": 0.1788, "step": 4116 }, { "epoch": 0.85, "learning_rate": 1.7180232558139535e-07, "loss": 0.1805, "step": 4117 }, { "epoch": 0.86, "learning_rate": 1.7177117940199336e-07, "loss": 0.193, "step": 4118 }, { "epoch": 0.86, "learning_rate": 1.7174003322259135e-07, "loss": 0.2366, "step": 4119 }, { "epoch": 0.86, "learning_rate": 1.7170888704318934e-07, "loss": 0.2816, "step": 4120 }, { "epoch": 0.86, "learning_rate": 1.7167774086378736e-07, "loss": 0.2507, "step": 4121 }, { "epoch": 0.86, "learning_rate": 1.7164659468438537e-07, "loss": 0.2181, "step": 4122 }, { "epoch": 0.86, "learning_rate": 1.716154485049834e-07, "loss": 0.1774, "step": 4123 }, { "epoch": 0.86, "learning_rate": 1.7158430232558138e-07, "loss": 0.1485, "step": 4124 }, { "epoch": 0.86, "learning_rate": 1.715531561461794e-07, "loss": 0.1569, "step": 4125 }, { "epoch": 0.86, "learning_rate": 1.715220099667774e-07, "loss": 0.266, "step": 4126 }, { "epoch": 0.86, "learning_rate": 1.7149086378737543e-07, "loss": 0.169, "step": 4127 }, { "epoch": 0.86, "learning_rate": 1.714597176079734e-07, "loss": 0.1669, "step": 4128 }, { "epoch": 0.86, "learning_rate": 1.714285714285714e-07, "loss": 0.1992, "step": 4129 }, { "epoch": 0.86, "learning_rate": 1.7139742524916942e-07, "loss": 0.1894, "step": 4130 }, { "epoch": 0.86, "learning_rate": 1.7136627906976744e-07, "loss": 0.1754, "step": 4131 }, { "epoch": 0.86, "learning_rate": 1.7133513289036545e-07, "loss": 0.2486, "step": 4132 }, { "epoch": 0.86, "learning_rate": 1.7130398671096344e-07, "loss": 0.1947, "step": 4133 }, { "epoch": 0.86, "learning_rate": 1.7127284053156146e-07, "loss": 0.2042, "step": 4134 }, { "epoch": 0.86, "learning_rate": 1.7124169435215945e-07, "loss": 0.1484, "step": 4135 }, { "epoch": 0.86, "learning_rate": 1.7121054817275746e-07, "loss": 0.2006, "step": 4136 }, { "epoch": 0.86, "learning_rate": 1.7117940199335548e-07, "loss": 0.1704, "step": 4137 }, { "epoch": 0.86, "learning_rate": 1.7114825581395347e-07, "loss": 0.1982, "step": 4138 }, { "epoch": 0.86, "learning_rate": 1.7111710963455148e-07, "loss": 0.1369, "step": 4139 }, { "epoch": 0.86, "learning_rate": 1.710859634551495e-07, "loss": 0.2382, "step": 4140 }, { "epoch": 0.86, "learning_rate": 1.7105481727574752e-07, "loss": 0.2055, "step": 4141 }, { "epoch": 0.86, "learning_rate": 1.7102367109634553e-07, "loss": 0.2523, "step": 4142 }, { "epoch": 0.86, "learning_rate": 1.709925249169435e-07, "loss": 0.167, "step": 4143 }, { "epoch": 0.86, "learning_rate": 1.709613787375415e-07, "loss": 0.1735, "step": 4144 }, { "epoch": 0.86, "learning_rate": 1.7093023255813953e-07, "loss": 0.2118, "step": 4145 }, { "epoch": 0.86, "learning_rate": 1.7089908637873754e-07, "loss": 0.1868, "step": 4146 }, { "epoch": 0.86, "learning_rate": 1.7086794019933553e-07, "loss": 0.164, "step": 4147 }, { "epoch": 0.86, "learning_rate": 1.7083679401993355e-07, "loss": 0.1697, "step": 4148 }, { "epoch": 0.86, "learning_rate": 1.7080564784053156e-07, "loss": 0.2039, "step": 4149 }, { "epoch": 0.86, "learning_rate": 1.7077450166112958e-07, "loss": 0.2302, "step": 4150 }, { "epoch": 0.86, "learning_rate": 1.7074335548172757e-07, "loss": 0.1879, "step": 4151 }, { "epoch": 0.86, "learning_rate": 1.7071220930232556e-07, "loss": 0.1774, "step": 4152 }, { "epoch": 0.86, "learning_rate": 1.7068106312292357e-07, "loss": 0.1681, "step": 4153 }, { "epoch": 0.86, "learning_rate": 1.706499169435216e-07, "loss": 0.1982, "step": 4154 }, { "epoch": 0.86, "learning_rate": 1.706187707641196e-07, "loss": 0.1847, "step": 4155 }, { "epoch": 0.86, "learning_rate": 1.7058762458471762e-07, "loss": 0.1484, "step": 4156 }, { "epoch": 0.86, "learning_rate": 1.7055647840531558e-07, "loss": 0.2145, "step": 4157 }, { "epoch": 0.86, "learning_rate": 1.705253322259136e-07, "loss": 0.2171, "step": 4158 }, { "epoch": 0.86, "learning_rate": 1.7049418604651162e-07, "loss": 0.1772, "step": 4159 }, { "epoch": 0.86, "learning_rate": 1.7046303986710963e-07, "loss": 0.2575, "step": 4160 }, { "epoch": 0.86, "learning_rate": 1.7043189368770762e-07, "loss": 0.1969, "step": 4161 }, { "epoch": 0.86, "learning_rate": 1.7040074750830564e-07, "loss": 0.2053, "step": 4162 }, { "epoch": 0.86, "learning_rate": 1.7036960132890365e-07, "loss": 0.1984, "step": 4163 }, { "epoch": 0.86, "learning_rate": 1.7033845514950167e-07, "loss": 0.1791, "step": 4164 }, { "epoch": 0.86, "learning_rate": 1.7030730897009968e-07, "loss": 0.1785, "step": 4165 }, { "epoch": 0.87, "learning_rate": 1.7027616279069765e-07, "loss": 0.2014, "step": 4166 }, { "epoch": 0.87, "learning_rate": 1.7024501661129566e-07, "loss": 0.1588, "step": 4167 }, { "epoch": 0.87, "learning_rate": 1.7021387043189368e-07, "loss": 0.1409, "step": 4168 }, { "epoch": 0.87, "learning_rate": 1.701827242524917e-07, "loss": 0.1831, "step": 4169 }, { "epoch": 0.87, "learning_rate": 1.701515780730897e-07, "loss": 0.171, "step": 4170 }, { "epoch": 0.87, "learning_rate": 1.701204318936877e-07, "loss": 0.1682, "step": 4171 }, { "epoch": 0.87, "learning_rate": 1.700892857142857e-07, "loss": 0.2321, "step": 4172 }, { "epoch": 0.87, "learning_rate": 1.700581395348837e-07, "loss": 0.2332, "step": 4173 }, { "epoch": 0.87, "learning_rate": 1.7002699335548172e-07, "loss": 0.1415, "step": 4174 }, { "epoch": 0.87, "learning_rate": 1.6999584717607974e-07, "loss": 0.2393, "step": 4175 }, { "epoch": 0.87, "learning_rate": 1.6996470099667773e-07, "loss": 0.2583, "step": 4176 }, { "epoch": 0.87, "learning_rate": 1.6993355481727574e-07, "loss": 0.2583, "step": 4177 }, { "epoch": 0.87, "learning_rate": 1.6990240863787376e-07, "loss": 0.1783, "step": 4178 }, { "epoch": 0.87, "learning_rate": 1.6987126245847177e-07, "loss": 0.2071, "step": 4179 }, { "epoch": 0.87, "learning_rate": 1.6984011627906974e-07, "loss": 0.1616, "step": 4180 }, { "epoch": 0.87, "learning_rate": 1.6980897009966775e-07, "loss": 0.17, "step": 4181 }, { "epoch": 0.87, "learning_rate": 1.6977782392026577e-07, "loss": 0.1815, "step": 4182 }, { "epoch": 0.87, "learning_rate": 1.6974667774086378e-07, "loss": 0.1495, "step": 4183 }, { "epoch": 0.87, "learning_rate": 1.697155315614618e-07, "loss": 0.1903, "step": 4184 }, { "epoch": 0.87, "learning_rate": 1.696843853820598e-07, "loss": 0.1762, "step": 4185 }, { "epoch": 0.87, "learning_rate": 1.696532392026578e-07, "loss": 0.173, "step": 4186 }, { "epoch": 0.87, "learning_rate": 1.6962209302325582e-07, "loss": 0.224, "step": 4187 }, { "epoch": 0.87, "learning_rate": 1.695909468438538e-07, "loss": 0.2581, "step": 4188 }, { "epoch": 0.87, "learning_rate": 1.6955980066445183e-07, "loss": 0.167, "step": 4189 }, { "epoch": 0.87, "learning_rate": 1.6952865448504982e-07, "loss": 0.1772, "step": 4190 }, { "epoch": 0.87, "learning_rate": 1.6949750830564783e-07, "loss": 0.2341, "step": 4191 }, { "epoch": 0.87, "learning_rate": 1.6946636212624585e-07, "loss": 0.1619, "step": 4192 }, { "epoch": 0.87, "learning_rate": 1.6943521594684386e-07, "loss": 0.2414, "step": 4193 }, { "epoch": 0.87, "learning_rate": 1.6940406976744183e-07, "loss": 0.2061, "step": 4194 }, { "epoch": 0.87, "learning_rate": 1.6937292358803984e-07, "loss": 0.2122, "step": 4195 }, { "epoch": 0.87, "learning_rate": 1.6934177740863786e-07, "loss": 0.1462, "step": 4196 }, { "epoch": 0.87, "learning_rate": 1.6931063122923587e-07, "loss": 0.2162, "step": 4197 }, { "epoch": 0.87, "learning_rate": 1.692794850498339e-07, "loss": 0.2057, "step": 4198 }, { "epoch": 0.87, "learning_rate": 1.6924833887043188e-07, "loss": 0.1768, "step": 4199 }, { "epoch": 0.87, "learning_rate": 1.692171926910299e-07, "loss": 0.2135, "step": 4200 }, { "epoch": 0.87, "learning_rate": 1.691860465116279e-07, "loss": 0.1611, "step": 4201 }, { "epoch": 0.87, "learning_rate": 1.6915490033222593e-07, "loss": 0.1687, "step": 4202 }, { "epoch": 0.87, "learning_rate": 1.6912375415282391e-07, "loss": 0.145, "step": 4203 }, { "epoch": 0.87, "learning_rate": 1.690926079734219e-07, "loss": 0.1705, "step": 4204 }, { "epoch": 0.87, "learning_rate": 1.6906146179401992e-07, "loss": 0.2018, "step": 4205 }, { "epoch": 0.87, "learning_rate": 1.6903031561461794e-07, "loss": 0.1798, "step": 4206 }, { "epoch": 0.87, "learning_rate": 1.6899916943521595e-07, "loss": 0.2131, "step": 4207 }, { "epoch": 0.87, "learning_rate": 1.6896802325581394e-07, "loss": 0.2217, "step": 4208 }, { "epoch": 0.87, "learning_rate": 1.6893687707641193e-07, "loss": 0.1762, "step": 4209 }, { "epoch": 0.87, "learning_rate": 1.6890573089700995e-07, "loss": 0.2122, "step": 4210 }, { "epoch": 0.87, "learning_rate": 1.6887458471760796e-07, "loss": 0.1499, "step": 4211 }, { "epoch": 0.87, "learning_rate": 1.6884343853820598e-07, "loss": 0.2316, "step": 4212 }, { "epoch": 0.87, "learning_rate": 1.6881229235880397e-07, "loss": 0.2286, "step": 4213 }, { "epoch": 0.88, "learning_rate": 1.6878114617940198e-07, "loss": 0.2205, "step": 4214 }, { "epoch": 0.88, "learning_rate": 1.6875e-07, "loss": 0.1591, "step": 4215 }, { "epoch": 0.88, "learning_rate": 1.6871885382059801e-07, "loss": 0.1533, "step": 4216 }, { "epoch": 0.88, "learning_rate": 1.6868770764119603e-07, "loss": 0.1708, "step": 4217 }, { "epoch": 0.88, "learning_rate": 1.68656561461794e-07, "loss": 0.1534, "step": 4218 }, { "epoch": 0.88, "learning_rate": 1.68625415282392e-07, "loss": 0.1928, "step": 4219 }, { "epoch": 0.88, "learning_rate": 1.6859426910299003e-07, "loss": 0.157, "step": 4220 }, { "epoch": 0.88, "learning_rate": 1.6856312292358804e-07, "loss": 0.2019, "step": 4221 }, { "epoch": 0.88, "learning_rate": 1.6853197674418606e-07, "loss": 0.2724, "step": 4222 }, { "epoch": 0.88, "learning_rate": 1.6850083056478405e-07, "loss": 0.1768, "step": 4223 }, { "epoch": 0.88, "learning_rate": 1.6846968438538206e-07, "loss": 0.1906, "step": 4224 }, { "epoch": 0.88, "learning_rate": 1.6843853820598005e-07, "loss": 0.2092, "step": 4225 }, { "epoch": 0.88, "learning_rate": 1.6840739202657807e-07, "loss": 0.2169, "step": 4226 }, { "epoch": 0.88, "learning_rate": 1.6837624584717606e-07, "loss": 0.1266, "step": 4227 }, { "epoch": 0.88, "learning_rate": 1.6834509966777407e-07, "loss": 0.22, "step": 4228 }, { "epoch": 0.88, "learning_rate": 1.683139534883721e-07, "loss": 0.2299, "step": 4229 }, { "epoch": 0.88, "learning_rate": 1.682828073089701e-07, "loss": 0.2002, "step": 4230 }, { "epoch": 0.88, "learning_rate": 1.6825166112956812e-07, "loss": 0.1938, "step": 4231 }, { "epoch": 0.88, "learning_rate": 1.6822051495016608e-07, "loss": 0.2077, "step": 4232 }, { "epoch": 0.88, "learning_rate": 1.681893687707641e-07, "loss": 0.2275, "step": 4233 }, { "epoch": 0.88, "learning_rate": 1.6815822259136211e-07, "loss": 0.1504, "step": 4234 }, { "epoch": 0.88, "learning_rate": 1.6812707641196013e-07, "loss": 0.247, "step": 4235 }, { "epoch": 0.88, "learning_rate": 1.6809593023255815e-07, "loss": 0.1757, "step": 4236 }, { "epoch": 0.88, "learning_rate": 1.6806478405315614e-07, "loss": 0.2436, "step": 4237 }, { "epoch": 0.88, "learning_rate": 1.6803363787375415e-07, "loss": 0.164, "step": 4238 }, { "epoch": 0.88, "learning_rate": 1.6800249169435217e-07, "loss": 0.1889, "step": 4239 }, { "epoch": 0.88, "learning_rate": 1.6797134551495016e-07, "loss": 0.1583, "step": 4240 }, { "epoch": 0.88, "learning_rate": 1.6794019933554815e-07, "loss": 0.1919, "step": 4241 }, { "epoch": 0.88, "learning_rate": 1.6790905315614616e-07, "loss": 0.2122, "step": 4242 }, { "epoch": 0.88, "learning_rate": 1.6787790697674418e-07, "loss": 0.2402, "step": 4243 }, { "epoch": 0.88, "learning_rate": 1.678467607973422e-07, "loss": 0.1899, "step": 4244 }, { "epoch": 0.88, "learning_rate": 1.678156146179402e-07, "loss": 0.1902, "step": 4245 }, { "epoch": 0.88, "learning_rate": 1.6778446843853817e-07, "loss": 0.2297, "step": 4246 }, { "epoch": 0.88, "learning_rate": 1.677533222591362e-07, "loss": 0.2128, "step": 4247 }, { "epoch": 0.88, "learning_rate": 1.677221760797342e-07, "loss": 0.1904, "step": 4248 }, { "epoch": 0.88, "learning_rate": 1.6769102990033222e-07, "loss": 0.2383, "step": 4249 }, { "epoch": 0.88, "learning_rate": 1.6765988372093024e-07, "loss": 0.2375, "step": 4250 }, { "epoch": 0.88, "learning_rate": 1.6762873754152822e-07, "loss": 0.1946, "step": 4251 }, { "epoch": 0.88, "learning_rate": 1.6759759136212624e-07, "loss": 0.2289, "step": 4252 }, { "epoch": 0.88, "learning_rate": 1.6756644518272426e-07, "loss": 0.2051, "step": 4253 }, { "epoch": 0.88, "learning_rate": 1.6753529900332227e-07, "loss": 0.2039, "step": 4254 }, { "epoch": 0.88, "learning_rate": 1.6750415282392023e-07, "loss": 0.2547, "step": 4255 }, { "epoch": 0.88, "learning_rate": 1.6747300664451825e-07, "loss": 0.19, "step": 4256 }, { "epoch": 0.88, "learning_rate": 1.6744186046511627e-07, "loss": 0.2079, "step": 4257 }, { "epoch": 0.88, "learning_rate": 1.6741071428571428e-07, "loss": 0.1793, "step": 4258 }, { "epoch": 0.88, "learning_rate": 1.673795681063123e-07, "loss": 0.2806, "step": 4259 }, { "epoch": 0.88, "learning_rate": 1.673484219269103e-07, "loss": 0.2406, "step": 4260 }, { "epoch": 0.88, "learning_rate": 1.673172757475083e-07, "loss": 0.23, "step": 4261 }, { "epoch": 0.88, "learning_rate": 1.672861295681063e-07, "loss": 0.162, "step": 4262 }, { "epoch": 0.89, "learning_rate": 1.672549833887043e-07, "loss": 0.2134, "step": 4263 }, { "epoch": 0.89, "learning_rate": 1.6722383720930232e-07, "loss": 0.2262, "step": 4264 }, { "epoch": 0.89, "learning_rate": 1.6719269102990031e-07, "loss": 0.2299, "step": 4265 }, { "epoch": 0.89, "learning_rate": 1.6716154485049833e-07, "loss": 0.1597, "step": 4266 }, { "epoch": 0.89, "learning_rate": 1.6713039867109635e-07, "loss": 0.1577, "step": 4267 }, { "epoch": 0.89, "learning_rate": 1.6709925249169436e-07, "loss": 0.1739, "step": 4268 }, { "epoch": 0.89, "learning_rate": 1.6706810631229238e-07, "loss": 0.1517, "step": 4269 }, { "epoch": 0.89, "learning_rate": 1.6703696013289034e-07, "loss": 0.1753, "step": 4270 }, { "epoch": 0.89, "learning_rate": 1.6700581395348836e-07, "loss": 0.2103, "step": 4271 }, { "epoch": 0.89, "learning_rate": 1.6697466777408637e-07, "loss": 0.1938, "step": 4272 }, { "epoch": 0.89, "learning_rate": 1.669435215946844e-07, "loss": 0.1746, "step": 4273 }, { "epoch": 0.89, "learning_rate": 1.6691237541528238e-07, "loss": 0.2175, "step": 4274 }, { "epoch": 0.89, "learning_rate": 1.668812292358804e-07, "loss": 0.1741, "step": 4275 }, { "epoch": 0.89, "learning_rate": 1.668500830564784e-07, "loss": 0.1355, "step": 4276 }, { "epoch": 0.89, "learning_rate": 1.668189368770764e-07, "loss": 0.1432, "step": 4277 }, { "epoch": 0.89, "learning_rate": 1.6678779069767441e-07, "loss": 0.1765, "step": 4278 }, { "epoch": 0.89, "learning_rate": 1.667566445182724e-07, "loss": 0.179, "step": 4279 }, { "epoch": 0.89, "learning_rate": 1.6672549833887042e-07, "loss": 0.1649, "step": 4280 }, { "epoch": 0.89, "learning_rate": 1.6669435215946843e-07, "loss": 0.1467, "step": 4281 }, { "epoch": 0.89, "learning_rate": 1.6666320598006645e-07, "loss": 0.2199, "step": 4282 }, { "epoch": 0.89, "learning_rate": 1.6663205980066447e-07, "loss": 0.1509, "step": 4283 }, { "epoch": 0.89, "learning_rate": 1.6660091362126243e-07, "loss": 0.1815, "step": 4284 }, { "epoch": 0.89, "learning_rate": 1.6656976744186044e-07, "loss": 0.1888, "step": 4285 }, { "epoch": 0.89, "learning_rate": 1.6653862126245846e-07, "loss": 0.1453, "step": 4286 }, { "epoch": 0.89, "learning_rate": 1.6650747508305648e-07, "loss": 0.1962, "step": 4287 }, { "epoch": 0.89, "learning_rate": 1.6647632890365447e-07, "loss": 0.1704, "step": 4288 }, { "epoch": 0.89, "learning_rate": 1.6644518272425248e-07, "loss": 0.2117, "step": 4289 }, { "epoch": 0.89, "learning_rate": 1.664140365448505e-07, "loss": 0.1756, "step": 4290 }, { "epoch": 0.89, "learning_rate": 1.6638289036544851e-07, "loss": 0.1764, "step": 4291 }, { "epoch": 0.89, "learning_rate": 1.6635174418604653e-07, "loss": 0.2005, "step": 4292 }, { "epoch": 0.89, "learning_rate": 1.663205980066445e-07, "loss": 0.2373, "step": 4293 }, { "epoch": 0.89, "learning_rate": 1.662894518272425e-07, "loss": 0.2122, "step": 4294 }, { "epoch": 0.89, "learning_rate": 1.6625830564784052e-07, "loss": 0.128, "step": 4295 }, { "epoch": 0.89, "learning_rate": 1.6622715946843854e-07, "loss": 0.2621, "step": 4296 }, { "epoch": 0.89, "learning_rate": 1.6619601328903656e-07, "loss": 0.1982, "step": 4297 }, { "epoch": 0.89, "learning_rate": 1.6616486710963454e-07, "loss": 0.2268, "step": 4298 }, { "epoch": 0.89, "learning_rate": 1.6613372093023253e-07, "loss": 0.1656, "step": 4299 }, { "epoch": 0.89, "learning_rate": 1.6610257475083055e-07, "loss": 0.1826, "step": 4300 }, { "epoch": 0.89, "learning_rate": 1.6607142857142857e-07, "loss": 0.1929, "step": 4301 }, { "epoch": 0.89, "learning_rate": 1.6604028239202658e-07, "loss": 0.2123, "step": 4302 }, { "epoch": 0.89, "learning_rate": 1.6600913621262457e-07, "loss": 0.1617, "step": 4303 }, { "epoch": 0.89, "learning_rate": 1.659779900332226e-07, "loss": 0.2149, "step": 4304 }, { "epoch": 0.89, "learning_rate": 1.659468438538206e-07, "loss": 0.1433, "step": 4305 }, { "epoch": 0.89, "learning_rate": 1.6591569767441862e-07, "loss": 0.1657, "step": 4306 }, { "epoch": 0.89, "learning_rate": 1.6588455149501658e-07, "loss": 0.1896, "step": 4307 }, { "epoch": 0.89, "learning_rate": 1.658534053156146e-07, "loss": 0.1763, "step": 4308 }, { "epoch": 0.89, "learning_rate": 1.658222591362126e-07, "loss": 0.1968, "step": 4309 }, { "epoch": 0.89, "learning_rate": 1.6579111295681063e-07, "loss": 0.176, "step": 4310 }, { "epoch": 0.9, "learning_rate": 1.6575996677740864e-07, "loss": 0.1463, "step": 4311 }, { "epoch": 0.9, "learning_rate": 1.6572882059800663e-07, "loss": 0.2233, "step": 4312 }, { "epoch": 0.9, "learning_rate": 1.6569767441860465e-07, "loss": 0.235, "step": 4313 }, { "epoch": 0.9, "learning_rate": 1.6566652823920264e-07, "loss": 0.249, "step": 4314 }, { "epoch": 0.9, "learning_rate": 1.6563538205980065e-07, "loss": 0.1744, "step": 4315 }, { "epoch": 0.9, "learning_rate": 1.6560423588039867e-07, "loss": 0.1701, "step": 4316 }, { "epoch": 0.9, "learning_rate": 1.6557308970099666e-07, "loss": 0.1631, "step": 4317 }, { "epoch": 0.9, "learning_rate": 1.6554194352159468e-07, "loss": 0.1747, "step": 4318 }, { "epoch": 0.9, "learning_rate": 1.655107973421927e-07, "loss": 0.1375, "step": 4319 }, { "epoch": 0.9, "learning_rate": 1.654796511627907e-07, "loss": 0.2389, "step": 4320 }, { "epoch": 0.9, "learning_rate": 1.6544850498338867e-07, "loss": 0.1439, "step": 4321 }, { "epoch": 0.9, "learning_rate": 1.6541735880398669e-07, "loss": 0.2178, "step": 4322 }, { "epoch": 0.9, "learning_rate": 1.653862126245847e-07, "loss": 0.2, "step": 4323 }, { "epoch": 0.9, "learning_rate": 1.6535506644518272e-07, "loss": 0.1674, "step": 4324 }, { "epoch": 0.9, "learning_rate": 1.6532392026578073e-07, "loss": 0.218, "step": 4325 }, { "epoch": 0.9, "learning_rate": 1.6529277408637872e-07, "loss": 0.1512, "step": 4326 }, { "epoch": 0.9, "learning_rate": 1.6526162790697674e-07, "loss": 0.2032, "step": 4327 }, { "epoch": 0.9, "learning_rate": 1.6523048172757475e-07, "loss": 0.2047, "step": 4328 }, { "epoch": 0.9, "learning_rate": 1.6519933554817277e-07, "loss": 0.1989, "step": 4329 }, { "epoch": 0.9, "learning_rate": 1.6516818936877076e-07, "loss": 0.2258, "step": 4330 }, { "epoch": 0.9, "learning_rate": 1.6513704318936875e-07, "loss": 0.1898, "step": 4331 }, { "epoch": 0.9, "learning_rate": 1.6510589700996677e-07, "loss": 0.1588, "step": 4332 }, { "epoch": 0.9, "learning_rate": 1.6507475083056478e-07, "loss": 0.1875, "step": 4333 }, { "epoch": 0.9, "learning_rate": 1.650436046511628e-07, "loss": 0.1614, "step": 4334 }, { "epoch": 0.9, "learning_rate": 1.6501245847176079e-07, "loss": 0.2548, "step": 4335 }, { "epoch": 0.9, "learning_rate": 1.6498131229235878e-07, "loss": 0.2254, "step": 4336 }, { "epoch": 0.9, "learning_rate": 1.649501661129568e-07, "loss": 0.19, "step": 4337 }, { "epoch": 0.9, "learning_rate": 1.649190199335548e-07, "loss": 0.2168, "step": 4338 }, { "epoch": 0.9, "learning_rate": 1.6488787375415282e-07, "loss": 0.1896, "step": 4339 }, { "epoch": 0.9, "learning_rate": 1.648567275747508e-07, "loss": 0.2464, "step": 4340 }, { "epoch": 0.9, "learning_rate": 1.6482558139534883e-07, "loss": 0.2246, "step": 4341 }, { "epoch": 0.9, "learning_rate": 1.6479443521594684e-07, "loss": 0.171, "step": 4342 }, { "epoch": 0.9, "learning_rate": 1.6476328903654486e-07, "loss": 0.2286, "step": 4343 }, { "epoch": 0.9, "learning_rate": 1.6473214285714288e-07, "loss": 0.1973, "step": 4344 }, { "epoch": 0.9, "learning_rate": 1.6470099667774084e-07, "loss": 0.1728, "step": 4345 }, { "epoch": 0.9, "learning_rate": 1.6466985049833885e-07, "loss": 0.1849, "step": 4346 }, { "epoch": 0.9, "learning_rate": 1.6463870431893687e-07, "loss": 0.1949, "step": 4347 }, { "epoch": 0.9, "learning_rate": 1.6460755813953489e-07, "loss": 0.1797, "step": 4348 }, { "epoch": 0.9, "learning_rate": 1.645764119601329e-07, "loss": 0.2127, "step": 4349 }, { "epoch": 0.9, "learning_rate": 1.645452657807309e-07, "loss": 0.1797, "step": 4350 }, { "epoch": 0.9, "learning_rate": 1.6451411960132888e-07, "loss": 0.1569, "step": 4351 }, { "epoch": 0.9, "learning_rate": 1.644829734219269e-07, "loss": 0.1829, "step": 4352 }, { "epoch": 0.9, "learning_rate": 1.644518272425249e-07, "loss": 0.2037, "step": 4353 }, { "epoch": 0.9, "learning_rate": 1.644206810631229e-07, "loss": 0.1966, "step": 4354 }, { "epoch": 0.9, "learning_rate": 1.6438953488372092e-07, "loss": 0.2142, "step": 4355 }, { "epoch": 0.9, "learning_rate": 1.6435838870431893e-07, "loss": 0.1834, "step": 4356 }, { "epoch": 0.9, "learning_rate": 1.6432724252491695e-07, "loss": 0.2004, "step": 4357 }, { "epoch": 0.9, "learning_rate": 1.6429609634551496e-07, "loss": 0.1924, "step": 4358 }, { "epoch": 0.91, "learning_rate": 1.6426495016611293e-07, "loss": 0.194, "step": 4359 }, { "epoch": 0.91, "learning_rate": 1.6423380398671094e-07, "loss": 0.1656, "step": 4360 }, { "epoch": 0.91, "learning_rate": 1.6420265780730896e-07, "loss": 0.187, "step": 4361 }, { "epoch": 0.91, "learning_rate": 1.6417151162790698e-07, "loss": 0.1917, "step": 4362 }, { "epoch": 0.91, "learning_rate": 1.64140365448505e-07, "loss": 0.2501, "step": 4363 }, { "epoch": 0.91, "learning_rate": 1.6410921926910298e-07, "loss": 0.1964, "step": 4364 }, { "epoch": 0.91, "learning_rate": 1.64078073089701e-07, "loss": 0.1797, "step": 4365 }, { "epoch": 0.91, "learning_rate": 1.64046926910299e-07, "loss": 0.1616, "step": 4366 }, { "epoch": 0.91, "learning_rate": 1.64015780730897e-07, "loss": 0.2327, "step": 4367 }, { "epoch": 0.91, "learning_rate": 1.63984634551495e-07, "loss": 0.2071, "step": 4368 }, { "epoch": 0.91, "learning_rate": 1.63953488372093e-07, "loss": 0.2282, "step": 4369 }, { "epoch": 0.91, "learning_rate": 1.6392234219269102e-07, "loss": 0.1863, "step": 4370 }, { "epoch": 0.91, "learning_rate": 1.6389119601328904e-07, "loss": 0.201, "step": 4371 }, { "epoch": 0.91, "learning_rate": 1.6386004983388705e-07, "loss": 0.2089, "step": 4372 }, { "epoch": 0.91, "learning_rate": 1.6382890365448502e-07, "loss": 0.2309, "step": 4373 }, { "epoch": 0.91, "learning_rate": 1.6379775747508303e-07, "loss": 0.2102, "step": 4374 }, { "epoch": 0.91, "learning_rate": 1.6376661129568105e-07, "loss": 0.2424, "step": 4375 }, { "epoch": 0.91, "learning_rate": 1.6373546511627906e-07, "loss": 0.1641, "step": 4376 }, { "epoch": 0.91, "learning_rate": 1.6370431893687708e-07, "loss": 0.1709, "step": 4377 }, { "epoch": 0.91, "learning_rate": 1.6367317275747507e-07, "loss": 0.2142, "step": 4378 }, { "epoch": 0.91, "learning_rate": 1.6364202657807309e-07, "loss": 0.1816, "step": 4379 }, { "epoch": 0.91, "learning_rate": 1.636108803986711e-07, "loss": 0.2943, "step": 4380 }, { "epoch": 0.91, "learning_rate": 1.6357973421926912e-07, "loss": 0.2267, "step": 4381 }, { "epoch": 0.91, "learning_rate": 1.6354858803986708e-07, "loss": 0.1852, "step": 4382 }, { "epoch": 0.91, "learning_rate": 1.635174418604651e-07, "loss": 0.2472, "step": 4383 }, { "epoch": 0.91, "learning_rate": 1.634862956810631e-07, "loss": 0.2196, "step": 4384 }, { "epoch": 0.91, "learning_rate": 1.6345514950166113e-07, "loss": 0.1495, "step": 4385 }, { "epoch": 0.91, "learning_rate": 1.6342400332225914e-07, "loss": 0.2123, "step": 4386 }, { "epoch": 0.91, "learning_rate": 1.6339285714285713e-07, "loss": 0.2073, "step": 4387 }, { "epoch": 0.91, "learning_rate": 1.6336171096345512e-07, "loss": 0.1857, "step": 4388 }, { "epoch": 0.91, "learning_rate": 1.6333056478405314e-07, "loss": 0.1693, "step": 4389 }, { "epoch": 0.91, "learning_rate": 1.6329941860465115e-07, "loss": 0.222, "step": 4390 }, { "epoch": 0.91, "learning_rate": 1.6326827242524917e-07, "loss": 0.1985, "step": 4391 }, { "epoch": 0.91, "learning_rate": 1.6323712624584716e-07, "loss": 0.2298, "step": 4392 }, { "epoch": 0.91, "learning_rate": 1.6320598006644517e-07, "loss": 0.1938, "step": 4393 }, { "epoch": 0.91, "learning_rate": 1.631748338870432e-07, "loss": 0.1693, "step": 4394 }, { "epoch": 0.91, "learning_rate": 1.631436877076412e-07, "loss": 0.1525, "step": 4395 }, { "epoch": 0.91, "learning_rate": 1.6311254152823922e-07, "loss": 0.1978, "step": 4396 }, { "epoch": 0.91, "learning_rate": 1.6308139534883718e-07, "loss": 0.1699, "step": 4397 }, { "epoch": 0.91, "learning_rate": 1.630502491694352e-07, "loss": 0.177, "step": 4398 }, { "epoch": 0.91, "learning_rate": 1.6301910299003322e-07, "loss": 0.2307, "step": 4399 }, { "epoch": 0.91, "learning_rate": 1.6298795681063123e-07, "loss": 0.2178, "step": 4400 }, { "epoch": 0.91, "learning_rate": 1.6295681063122922e-07, "loss": 0.2647, "step": 4401 }, { "epoch": 0.91, "learning_rate": 1.6292566445182724e-07, "loss": 0.2166, "step": 4402 }, { "epoch": 0.91, "learning_rate": 1.6289451827242525e-07, "loss": 0.2118, "step": 4403 }, { "epoch": 0.91, "learning_rate": 1.6286337209302324e-07, "loss": 0.1981, "step": 4404 }, { "epoch": 0.91, "learning_rate": 1.6283222591362126e-07, "loss": 0.1854, "step": 4405 }, { "epoch": 0.91, "learning_rate": 1.6280107973421925e-07, "loss": 0.1972, "step": 4406 }, { "epoch": 0.92, "learning_rate": 1.6276993355481726e-07, "loss": 0.1612, "step": 4407 }, { "epoch": 0.92, "learning_rate": 1.6273878737541528e-07, "loss": 0.1805, "step": 4408 }, { "epoch": 0.92, "learning_rate": 1.627076411960133e-07, "loss": 0.1911, "step": 4409 }, { "epoch": 0.92, "learning_rate": 1.626764950166113e-07, "loss": 0.2136, "step": 4410 }, { "epoch": 0.92, "learning_rate": 1.6264534883720927e-07, "loss": 0.1939, "step": 4411 }, { "epoch": 0.92, "learning_rate": 1.626142026578073e-07, "loss": 0.1909, "step": 4412 }, { "epoch": 0.92, "learning_rate": 1.625830564784053e-07, "loss": 0.2302, "step": 4413 }, { "epoch": 0.92, "learning_rate": 1.6255191029900332e-07, "loss": 0.1664, "step": 4414 }, { "epoch": 0.92, "learning_rate": 1.625207641196013e-07, "loss": 0.1715, "step": 4415 }, { "epoch": 0.92, "learning_rate": 1.6248961794019933e-07, "loss": 0.2054, "step": 4416 }, { "epoch": 0.92, "learning_rate": 1.6245847176079734e-07, "loss": 0.2255, "step": 4417 }, { "epoch": 0.92, "learning_rate": 1.6242732558139536e-07, "loss": 0.2031, "step": 4418 }, { "epoch": 0.92, "learning_rate": 1.6239617940199335e-07, "loss": 0.2643, "step": 4419 }, { "epoch": 0.92, "learning_rate": 1.6236503322259134e-07, "loss": 0.2063, "step": 4420 }, { "epoch": 0.92, "learning_rate": 1.6233388704318935e-07, "loss": 0.135, "step": 4421 }, { "epoch": 0.92, "learning_rate": 1.6230274086378737e-07, "loss": 0.1841, "step": 4422 }, { "epoch": 0.92, "learning_rate": 1.6227159468438538e-07, "loss": 0.1767, "step": 4423 }, { "epoch": 0.92, "learning_rate": 1.622404485049834e-07, "loss": 0.1966, "step": 4424 }, { "epoch": 0.92, "learning_rate": 1.6220930232558136e-07, "loss": 0.1617, "step": 4425 }, { "epoch": 0.92, "learning_rate": 1.6217815614617938e-07, "loss": 0.2098, "step": 4426 }, { "epoch": 0.92, "learning_rate": 1.621470099667774e-07, "loss": 0.2044, "step": 4427 }, { "epoch": 0.92, "learning_rate": 1.621158637873754e-07, "loss": 0.1819, "step": 4428 }, { "epoch": 0.92, "learning_rate": 1.620847176079734e-07, "loss": 0.226, "step": 4429 }, { "epoch": 0.92, "learning_rate": 1.6205357142857142e-07, "loss": 0.1644, "step": 4430 }, { "epoch": 0.92, "learning_rate": 1.6202242524916943e-07, "loss": 0.2235, "step": 4431 }, { "epoch": 0.92, "learning_rate": 1.6199127906976745e-07, "loss": 0.1801, "step": 4432 }, { "epoch": 0.92, "learning_rate": 1.6196013289036546e-07, "loss": 0.1575, "step": 4433 }, { "epoch": 0.92, "learning_rate": 1.6192898671096343e-07, "loss": 0.1966, "step": 4434 }, { "epoch": 0.92, "learning_rate": 1.6189784053156144e-07, "loss": 0.24, "step": 4435 }, { "epoch": 0.92, "learning_rate": 1.6186669435215946e-07, "loss": 0.1968, "step": 4436 }, { "epoch": 0.92, "learning_rate": 1.6183554817275747e-07, "loss": 0.2322, "step": 4437 }, { "epoch": 0.92, "learning_rate": 1.618044019933555e-07, "loss": 0.2148, "step": 4438 }, { "epoch": 0.92, "learning_rate": 1.6177325581395348e-07, "loss": 0.2379, "step": 4439 }, { "epoch": 0.92, "learning_rate": 1.617421096345515e-07, "loss": 0.2505, "step": 4440 }, { "epoch": 0.92, "learning_rate": 1.6171096345514948e-07, "loss": 0.1874, "step": 4441 }, { "epoch": 0.92, "learning_rate": 1.616798172757475e-07, "loss": 0.2237, "step": 4442 }, { "epoch": 0.92, "learning_rate": 1.6164867109634552e-07, "loss": 0.2363, "step": 4443 }, { "epoch": 0.92, "learning_rate": 1.616175249169435e-07, "loss": 0.2372, "step": 4444 }, { "epoch": 0.92, "learning_rate": 1.6158637873754152e-07, "loss": 0.1791, "step": 4445 }, { "epoch": 0.92, "learning_rate": 1.6155523255813954e-07, "loss": 0.2128, "step": 4446 }, { "epoch": 0.92, "learning_rate": 1.6152408637873755e-07, "loss": 0.1675, "step": 4447 }, { "epoch": 0.92, "learning_rate": 1.6149294019933552e-07, "loss": 0.2143, "step": 4448 }, { "epoch": 0.92, "learning_rate": 1.6146179401993353e-07, "loss": 0.2, "step": 4449 }, { "epoch": 0.92, "learning_rate": 1.6143064784053155e-07, "loss": 0.1492, "step": 4450 }, { "epoch": 0.92, "learning_rate": 1.6139950166112956e-07, "loss": 0.1891, "step": 4451 }, { "epoch": 0.92, "learning_rate": 1.6136835548172758e-07, "loss": 0.1456, "step": 4452 }, { "epoch": 0.92, "learning_rate": 1.6133720930232557e-07, "loss": 0.1495, "step": 4453 }, { "epoch": 0.92, "learning_rate": 1.6130606312292358e-07, "loss": 0.1994, "step": 4454 }, { "epoch": 0.93, "learning_rate": 1.612749169435216e-07, "loss": 0.2463, "step": 4455 }, { "epoch": 0.93, "learning_rate": 1.612437707641196e-07, "loss": 0.1618, "step": 4456 }, { "epoch": 0.93, "learning_rate": 1.612126245847176e-07, "loss": 0.1819, "step": 4457 }, { "epoch": 0.93, "learning_rate": 1.611814784053156e-07, "loss": 0.1714, "step": 4458 }, { "epoch": 0.93, "learning_rate": 1.611503322259136e-07, "loss": 0.2108, "step": 4459 }, { "epoch": 0.93, "learning_rate": 1.6111918604651163e-07, "loss": 0.1688, "step": 4460 }, { "epoch": 0.93, "learning_rate": 1.6108803986710964e-07, "loss": 0.1807, "step": 4461 }, { "epoch": 0.93, "learning_rate": 1.610568936877076e-07, "loss": 0.2153, "step": 4462 }, { "epoch": 0.93, "learning_rate": 1.6102574750830562e-07, "loss": 0.15, "step": 4463 }, { "epoch": 0.93, "learning_rate": 1.6099460132890364e-07, "loss": 0.1864, "step": 4464 }, { "epoch": 0.93, "learning_rate": 1.6096345514950165e-07, "loss": 0.1637, "step": 4465 }, { "epoch": 0.93, "learning_rate": 1.6093230897009967e-07, "loss": 0.1847, "step": 4466 }, { "epoch": 0.93, "learning_rate": 1.6090116279069766e-07, "loss": 0.1836, "step": 4467 }, { "epoch": 0.93, "learning_rate": 1.6087001661129567e-07, "loss": 0.2143, "step": 4468 }, { "epoch": 0.93, "learning_rate": 1.608388704318937e-07, "loss": 0.2122, "step": 4469 }, { "epoch": 0.93, "learning_rate": 1.608077242524917e-07, "loss": 0.1856, "step": 4470 }, { "epoch": 0.93, "learning_rate": 1.6077657807308972e-07, "loss": 0.1577, "step": 4471 }, { "epoch": 0.93, "learning_rate": 1.6074543189368768e-07, "loss": 0.2213, "step": 4472 }, { "epoch": 0.93, "learning_rate": 1.607142857142857e-07, "loss": 0.1384, "step": 4473 }, { "epoch": 0.93, "learning_rate": 1.6068313953488372e-07, "loss": 0.2154, "step": 4474 }, { "epoch": 0.93, "learning_rate": 1.6065199335548173e-07, "loss": 0.1865, "step": 4475 }, { "epoch": 0.93, "learning_rate": 1.6062084717607975e-07, "loss": 0.1747, "step": 4476 }, { "epoch": 0.93, "learning_rate": 1.6058970099667774e-07, "loss": 0.1842, "step": 4477 }, { "epoch": 0.93, "learning_rate": 1.6055855481727573e-07, "loss": 0.1751, "step": 4478 }, { "epoch": 0.93, "learning_rate": 1.6052740863787374e-07, "loss": 0.208, "step": 4479 }, { "epoch": 0.93, "learning_rate": 1.6049626245847176e-07, "loss": 0.22, "step": 4480 }, { "epoch": 0.93, "learning_rate": 1.6046511627906975e-07, "loss": 0.1823, "step": 4481 }, { "epoch": 0.93, "learning_rate": 1.6043397009966776e-07, "loss": 0.1996, "step": 4482 }, { "epoch": 0.93, "learning_rate": 1.6040282392026578e-07, "loss": 0.204, "step": 4483 }, { "epoch": 0.93, "learning_rate": 1.603716777408638e-07, "loss": 0.207, "step": 4484 }, { "epoch": 0.93, "learning_rate": 1.603405315614618e-07, "loss": 0.1639, "step": 4485 }, { "epoch": 0.93, "learning_rate": 1.6030938538205977e-07, "loss": 0.1771, "step": 4486 }, { "epoch": 0.93, "learning_rate": 1.602782392026578e-07, "loss": 0.2061, "step": 4487 }, { "epoch": 0.93, "learning_rate": 1.602470930232558e-07, "loss": 0.1897, "step": 4488 }, { "epoch": 0.93, "learning_rate": 1.6021594684385382e-07, "loss": 0.1974, "step": 4489 }, { "epoch": 0.93, "learning_rate": 1.6018480066445184e-07, "loss": 0.2389, "step": 4490 }, { "epoch": 0.93, "learning_rate": 1.6015365448504983e-07, "loss": 0.175, "step": 4491 }, { "epoch": 0.93, "learning_rate": 1.6012250830564784e-07, "loss": 0.2125, "step": 4492 }, { "epoch": 0.93, "learning_rate": 1.6009136212624583e-07, "loss": 0.182, "step": 4493 }, { "epoch": 0.93, "learning_rate": 1.6006021594684385e-07, "loss": 0.2077, "step": 4494 }, { "epoch": 0.93, "learning_rate": 1.6002906976744184e-07, "loss": 0.1734, "step": 4495 }, { "epoch": 0.93, "learning_rate": 1.5999792358803985e-07, "loss": 0.1391, "step": 4496 }, { "epoch": 0.93, "learning_rate": 1.5996677740863787e-07, "loss": 0.1214, "step": 4497 }, { "epoch": 0.93, "learning_rate": 1.5993563122923588e-07, "loss": 0.1603, "step": 4498 }, { "epoch": 0.93, "learning_rate": 1.599044850498339e-07, "loss": 0.1882, "step": 4499 }, { "epoch": 0.93, "learning_rate": 1.5987333887043186e-07, "loss": 0.1784, "step": 4500 }, { "epoch": 0.93, "learning_rate": 1.5984219269102988e-07, "loss": 0.1749, "step": 4501 }, { "epoch": 0.93, "learning_rate": 1.598110465116279e-07, "loss": 0.1598, "step": 4502 }, { "epoch": 0.94, "learning_rate": 1.597799003322259e-07, "loss": 0.2587, "step": 4503 }, { "epoch": 0.94, "learning_rate": 1.5974875415282393e-07, "loss": 0.1854, "step": 4504 }, { "epoch": 0.94, "learning_rate": 1.5971760797342191e-07, "loss": 0.17, "step": 4505 }, { "epoch": 0.94, "learning_rate": 1.5968646179401993e-07, "loss": 0.1875, "step": 4506 }, { "epoch": 0.94, "learning_rate": 1.5965531561461795e-07, "loss": 0.1995, "step": 4507 }, { "epoch": 0.94, "learning_rate": 1.5962416943521596e-07, "loss": 0.2628, "step": 4508 }, { "epoch": 0.94, "learning_rate": 1.5959302325581393e-07, "loss": 0.2317, "step": 4509 }, { "epoch": 0.94, "learning_rate": 1.5956187707641194e-07, "loss": 0.2017, "step": 4510 }, { "epoch": 0.94, "learning_rate": 1.5953073089700996e-07, "loss": 0.1414, "step": 4511 }, { "epoch": 0.94, "learning_rate": 1.5949958471760797e-07, "loss": 0.156, "step": 4512 }, { "epoch": 0.94, "learning_rate": 1.59468438538206e-07, "loss": 0.1913, "step": 4513 }, { "epoch": 0.94, "learning_rate": 1.5943729235880398e-07, "loss": 0.2687, "step": 4514 }, { "epoch": 0.94, "learning_rate": 1.5940614617940197e-07, "loss": 0.2661, "step": 4515 }, { "epoch": 0.94, "learning_rate": 1.5937499999999998e-07, "loss": 0.1709, "step": 4516 }, { "epoch": 0.94, "learning_rate": 1.59343853820598e-07, "loss": 0.1331, "step": 4517 }, { "epoch": 0.94, "learning_rate": 1.5931270764119601e-07, "loss": 0.233, "step": 4518 }, { "epoch": 0.94, "learning_rate": 1.59281561461794e-07, "loss": 0.2003, "step": 4519 }, { "epoch": 0.94, "learning_rate": 1.5925041528239202e-07, "loss": 0.183, "step": 4520 }, { "epoch": 0.94, "learning_rate": 1.5921926910299004e-07, "loss": 0.2185, "step": 4521 }, { "epoch": 0.94, "learning_rate": 1.5918812292358805e-07, "loss": 0.1782, "step": 4522 }, { "epoch": 0.94, "learning_rate": 1.5915697674418607e-07, "loss": 0.2143, "step": 4523 }, { "epoch": 0.94, "learning_rate": 1.5912583056478403e-07, "loss": 0.2118, "step": 4524 }, { "epoch": 0.94, "learning_rate": 1.5909468438538205e-07, "loss": 0.1871, "step": 4525 }, { "epoch": 0.94, "learning_rate": 1.5906353820598006e-07, "loss": 0.2276, "step": 4526 }, { "epoch": 0.94, "learning_rate": 1.5903239202657808e-07, "loss": 0.1848, "step": 4527 }, { "epoch": 0.94, "learning_rate": 1.5900124584717607e-07, "loss": 0.1589, "step": 4528 }, { "epoch": 0.94, "learning_rate": 1.5897009966777408e-07, "loss": 0.1971, "step": 4529 }, { "epoch": 0.94, "learning_rate": 1.5893895348837207e-07, "loss": 0.1479, "step": 4530 }, { "epoch": 0.94, "learning_rate": 1.589078073089701e-07, "loss": 0.2067, "step": 4531 }, { "epoch": 0.94, "learning_rate": 1.588766611295681e-07, "loss": 0.195, "step": 4532 }, { "epoch": 0.94, "learning_rate": 1.588455149501661e-07, "loss": 0.2605, "step": 4533 }, { "epoch": 0.94, "learning_rate": 1.588143687707641e-07, "loss": 0.1457, "step": 4534 }, { "epoch": 0.94, "learning_rate": 1.5878322259136212e-07, "loss": 0.1698, "step": 4535 }, { "epoch": 0.94, "learning_rate": 1.5875207641196014e-07, "loss": 0.1789, "step": 4536 }, { "epoch": 0.94, "learning_rate": 1.5872093023255816e-07, "loss": 0.1541, "step": 4537 }, { "epoch": 0.94, "learning_rate": 1.5868978405315612e-07, "loss": 0.1724, "step": 4538 }, { "epoch": 0.94, "learning_rate": 1.5865863787375414e-07, "loss": 0.2069, "step": 4539 }, { "epoch": 0.94, "learning_rate": 1.5862749169435215e-07, "loss": 0.1948, "step": 4540 }, { "epoch": 0.94, "learning_rate": 1.5859634551495017e-07, "loss": 0.2091, "step": 4541 }, { "epoch": 0.94, "learning_rate": 1.5856519933554816e-07, "loss": 0.1638, "step": 4542 }, { "epoch": 0.94, "learning_rate": 1.5853405315614617e-07, "loss": 0.1532, "step": 4543 }, { "epoch": 0.94, "learning_rate": 1.585029069767442e-07, "loss": 0.1683, "step": 4544 }, { "epoch": 0.94, "learning_rate": 1.584717607973422e-07, "loss": 0.2136, "step": 4545 }, { "epoch": 0.94, "learning_rate": 1.584406146179402e-07, "loss": 0.2136, "step": 4546 }, { "epoch": 0.94, "learning_rate": 1.5840946843853818e-07, "loss": 0.231, "step": 4547 }, { "epoch": 0.94, "learning_rate": 1.583783222591362e-07, "loss": 0.2329, "step": 4548 }, { "epoch": 0.94, "learning_rate": 1.5834717607973421e-07, "loss": 0.1711, "step": 4549 }, { "epoch": 0.94, "learning_rate": 1.5831602990033223e-07, "loss": 0.1214, "step": 4550 }, { "epoch": 0.94, "learning_rate": 1.5828488372093025e-07, "loss": 0.1218, "step": 4551 }, { "epoch": 0.95, "learning_rate": 1.582537375415282e-07, "loss": 0.1847, "step": 4552 }, { "epoch": 0.95, "learning_rate": 1.5822259136212622e-07, "loss": 0.2475, "step": 4553 }, { "epoch": 0.95, "learning_rate": 1.5819144518272424e-07, "loss": 0.2046, "step": 4554 }, { "epoch": 0.95, "learning_rate": 1.5816029900332226e-07, "loss": 0.2487, "step": 4555 }, { "epoch": 0.95, "learning_rate": 1.5812915282392025e-07, "loss": 0.1317, "step": 4556 }, { "epoch": 0.95, "learning_rate": 1.5809800664451826e-07, "loss": 0.1969, "step": 4557 }, { "epoch": 0.95, "learning_rate": 1.5806686046511628e-07, "loss": 0.201, "step": 4558 }, { "epoch": 0.95, "learning_rate": 1.580357142857143e-07, "loss": 0.149, "step": 4559 }, { "epoch": 0.95, "learning_rate": 1.580045681063123e-07, "loss": 0.1872, "step": 4560 }, { "epoch": 0.95, "learning_rate": 1.5797342192691027e-07, "loss": 0.2184, "step": 4561 }, { "epoch": 0.95, "learning_rate": 1.579422757475083e-07, "loss": 0.2114, "step": 4562 }, { "epoch": 0.95, "learning_rate": 1.579111295681063e-07, "loss": 0.1493, "step": 4563 }, { "epoch": 0.95, "learning_rate": 1.5787998338870432e-07, "loss": 0.1793, "step": 4564 }, { "epoch": 0.95, "learning_rate": 1.5784883720930233e-07, "loss": 0.1626, "step": 4565 }, { "epoch": 0.95, "learning_rate": 1.5781769102990032e-07, "loss": 0.2206, "step": 4566 }, { "epoch": 0.95, "learning_rate": 1.5778654485049831e-07, "loss": 0.1298, "step": 4567 }, { "epoch": 0.95, "learning_rate": 1.5775539867109633e-07, "loss": 0.1624, "step": 4568 }, { "epoch": 0.95, "learning_rate": 1.5772425249169435e-07, "loss": 0.1935, "step": 4569 }, { "epoch": 0.95, "learning_rate": 1.5769310631229236e-07, "loss": 0.1028, "step": 4570 }, { "epoch": 0.95, "learning_rate": 1.5766196013289035e-07, "loss": 0.1867, "step": 4571 }, { "epoch": 0.95, "learning_rate": 1.5763081395348837e-07, "loss": 0.1852, "step": 4572 }, { "epoch": 0.95, "learning_rate": 1.5759966777408638e-07, "loss": 0.1583, "step": 4573 }, { "epoch": 0.95, "learning_rate": 1.575685215946844e-07, "loss": 0.2019, "step": 4574 }, { "epoch": 0.95, "learning_rate": 1.5753737541528236e-07, "loss": 0.2336, "step": 4575 }, { "epoch": 0.95, "learning_rate": 1.5750622923588038e-07, "loss": 0.1462, "step": 4576 }, { "epoch": 0.95, "learning_rate": 1.574750830564784e-07, "loss": 0.1762, "step": 4577 }, { "epoch": 0.95, "learning_rate": 1.574439368770764e-07, "loss": 0.2149, "step": 4578 }, { "epoch": 0.95, "learning_rate": 1.5741279069767442e-07, "loss": 0.2156, "step": 4579 }, { "epoch": 0.95, "learning_rate": 1.5738164451827241e-07, "loss": 0.2043, "step": 4580 }, { "epoch": 0.95, "learning_rate": 1.5735049833887043e-07, "loss": 0.1435, "step": 4581 }, { "epoch": 0.95, "learning_rate": 1.5731935215946844e-07, "loss": 0.2121, "step": 4582 }, { "epoch": 0.95, "learning_rate": 1.5728820598006643e-07, "loss": 0.1914, "step": 4583 }, { "epoch": 0.95, "learning_rate": 1.5725705980066445e-07, "loss": 0.2298, "step": 4584 }, { "epoch": 0.95, "learning_rate": 1.5722591362126244e-07, "loss": 0.2023, "step": 4585 }, { "epoch": 0.95, "learning_rate": 1.5719476744186046e-07, "loss": 0.1777, "step": 4586 }, { "epoch": 0.95, "learning_rate": 1.5716362126245847e-07, "loss": 0.2176, "step": 4587 }, { "epoch": 0.95, "learning_rate": 1.571324750830565e-07, "loss": 0.1888, "step": 4588 }, { "epoch": 0.95, "learning_rate": 1.5710132890365445e-07, "loss": 0.199, "step": 4589 }, { "epoch": 0.95, "learning_rate": 1.5707018272425247e-07, "loss": 0.3023, "step": 4590 }, { "epoch": 0.95, "learning_rate": 1.5703903654485048e-07, "loss": 0.2047, "step": 4591 }, { "epoch": 0.95, "learning_rate": 1.570078903654485e-07, "loss": 0.1831, "step": 4592 }, { "epoch": 0.95, "learning_rate": 1.569767441860465e-07, "loss": 0.1628, "step": 4593 }, { "epoch": 0.95, "learning_rate": 1.569455980066445e-07, "loss": 0.2049, "step": 4594 }, { "epoch": 0.95, "learning_rate": 1.5691445182724252e-07, "loss": 0.2022, "step": 4595 }, { "epoch": 0.95, "learning_rate": 1.5688330564784053e-07, "loss": 0.1931, "step": 4596 }, { "epoch": 0.95, "learning_rate": 1.5685215946843855e-07, "loss": 0.1699, "step": 4597 }, { "epoch": 0.95, "learning_rate": 1.5682101328903654e-07, "loss": 0.2421, "step": 4598 }, { "epoch": 0.95, "learning_rate": 1.5678986710963453e-07, "loss": 0.1558, "step": 4599 }, { "epoch": 0.96, "learning_rate": 1.5675872093023254e-07, "loss": 0.2645, "step": 4600 }, { "epoch": 0.96, "learning_rate": 1.5672757475083056e-07, "loss": 0.1669, "step": 4601 }, { "epoch": 0.96, "learning_rate": 1.5669642857142858e-07, "loss": 0.1436, "step": 4602 }, { "epoch": 0.96, "learning_rate": 1.566652823920266e-07, "loss": 0.1492, "step": 4603 }, { "epoch": 0.96, "learning_rate": 1.5663413621262455e-07, "loss": 0.1891, "step": 4604 }, { "epoch": 0.96, "learning_rate": 1.5660299003322257e-07, "loss": 0.1867, "step": 4605 }, { "epoch": 0.96, "learning_rate": 1.5657184385382059e-07, "loss": 0.1986, "step": 4606 }, { "epoch": 0.96, "learning_rate": 1.565406976744186e-07, "loss": 0.1848, "step": 4607 }, { "epoch": 0.96, "learning_rate": 1.565095514950166e-07, "loss": 0.19, "step": 4608 }, { "epoch": 0.96, "learning_rate": 1.564784053156146e-07, "loss": 0.2004, "step": 4609 }, { "epoch": 0.96, "learning_rate": 1.5644725913621262e-07, "loss": 0.2007, "step": 4610 }, { "epoch": 0.96, "learning_rate": 1.5641611295681064e-07, "loss": 0.226, "step": 4611 }, { "epoch": 0.96, "learning_rate": 1.5638496677740865e-07, "loss": 0.2302, "step": 4612 }, { "epoch": 0.96, "learning_rate": 1.5635382059800662e-07, "loss": 0.2052, "step": 4613 }, { "epoch": 0.96, "learning_rate": 1.5632267441860463e-07, "loss": 0.1871, "step": 4614 }, { "epoch": 0.96, "learning_rate": 1.5629152823920265e-07, "loss": 0.1943, "step": 4615 }, { "epoch": 0.96, "learning_rate": 1.5626038205980067e-07, "loss": 0.2751, "step": 4616 }, { "epoch": 0.96, "learning_rate": 1.5622923588039868e-07, "loss": 0.1663, "step": 4617 }, { "epoch": 0.96, "learning_rate": 1.5619808970099667e-07, "loss": 0.1942, "step": 4618 }, { "epoch": 0.96, "learning_rate": 1.5616694352159469e-07, "loss": 0.1963, "step": 4619 }, { "epoch": 0.96, "learning_rate": 1.5613579734219268e-07, "loss": 0.1601, "step": 4620 }, { "epoch": 0.96, "learning_rate": 1.561046511627907e-07, "loss": 0.1891, "step": 4621 }, { "epoch": 0.96, "learning_rate": 1.5607350498338868e-07, "loss": 0.1577, "step": 4622 }, { "epoch": 0.96, "learning_rate": 1.560423588039867e-07, "loss": 0.1877, "step": 4623 }, { "epoch": 0.96, "learning_rate": 1.560112126245847e-07, "loss": 0.221, "step": 4624 }, { "epoch": 0.96, "learning_rate": 1.5598006644518273e-07, "loss": 0.2224, "step": 4625 }, { "epoch": 0.96, "learning_rate": 1.5594892026578074e-07, "loss": 0.2247, "step": 4626 }, { "epoch": 0.96, "learning_rate": 1.559177740863787e-07, "loss": 0.2039, "step": 4627 }, { "epoch": 0.96, "learning_rate": 1.5588662790697672e-07, "loss": 0.1502, "step": 4628 }, { "epoch": 0.96, "learning_rate": 1.5585548172757474e-07, "loss": 0.1869, "step": 4629 }, { "epoch": 0.96, "learning_rate": 1.5582433554817275e-07, "loss": 0.2002, "step": 4630 }, { "epoch": 0.96, "learning_rate": 1.5579318936877077e-07, "loss": 0.2031, "step": 4631 }, { "epoch": 0.96, "learning_rate": 1.5576204318936876e-07, "loss": 0.2295, "step": 4632 }, { "epoch": 0.96, "learning_rate": 1.5573089700996678e-07, "loss": 0.2004, "step": 4633 }, { "epoch": 0.96, "learning_rate": 1.556997508305648e-07, "loss": 0.2011, "step": 4634 }, { "epoch": 0.96, "learning_rate": 1.5566860465116278e-07, "loss": 0.187, "step": 4635 }, { "epoch": 0.96, "learning_rate": 1.5563745847176077e-07, "loss": 0.3241, "step": 4636 }, { "epoch": 0.96, "learning_rate": 1.5560631229235879e-07, "loss": 0.1886, "step": 4637 }, { "epoch": 0.96, "learning_rate": 1.555751661129568e-07, "loss": 0.1528, "step": 4638 }, { "epoch": 0.96, "learning_rate": 1.5554401993355482e-07, "loss": 0.3153, "step": 4639 }, { "epoch": 0.96, "learning_rate": 1.5551287375415283e-07, "loss": 0.1478, "step": 4640 }, { "epoch": 0.96, "learning_rate": 1.554817275747508e-07, "loss": 0.1597, "step": 4641 }, { "epoch": 0.96, "learning_rate": 1.554505813953488e-07, "loss": 0.2025, "step": 4642 }, { "epoch": 0.96, "learning_rate": 1.5541943521594683e-07, "loss": 0.1686, "step": 4643 }, { "epoch": 0.96, "learning_rate": 1.5538828903654484e-07, "loss": 0.1943, "step": 4644 }, { "epoch": 0.96, "learning_rate": 1.5535714285714286e-07, "loss": 0.2451, "step": 4645 }, { "epoch": 0.96, "learning_rate": 1.5532599667774085e-07, "loss": 0.2082, "step": 4646 }, { "epoch": 0.96, "learning_rate": 1.5529485049833886e-07, "loss": 0.2615, "step": 4647 }, { "epoch": 0.97, "learning_rate": 1.5526370431893688e-07, "loss": 0.2199, "step": 4648 }, { "epoch": 0.97, "learning_rate": 1.552325581395349e-07, "loss": 0.1845, "step": 4649 }, { "epoch": 0.97, "learning_rate": 1.552014119601329e-07, "loss": 0.2078, "step": 4650 }, { "epoch": 0.97, "learning_rate": 1.5517026578073088e-07, "loss": 0.1951, "step": 4651 }, { "epoch": 0.97, "learning_rate": 1.551391196013289e-07, "loss": 0.1917, "step": 4652 }, { "epoch": 0.97, "learning_rate": 1.551079734219269e-07, "loss": 0.2154, "step": 4653 }, { "epoch": 0.97, "learning_rate": 1.5507682724252492e-07, "loss": 0.2269, "step": 4654 }, { "epoch": 0.97, "learning_rate": 1.550456810631229e-07, "loss": 0.1563, "step": 4655 }, { "epoch": 0.97, "learning_rate": 1.5501453488372093e-07, "loss": 0.1529, "step": 4656 }, { "epoch": 0.97, "learning_rate": 1.5498338870431892e-07, "loss": 0.168, "step": 4657 }, { "epoch": 0.97, "learning_rate": 1.5495224252491693e-07, "loss": 0.2283, "step": 4658 }, { "epoch": 0.97, "learning_rate": 1.5492109634551495e-07, "loss": 0.2148, "step": 4659 }, { "epoch": 0.97, "learning_rate": 1.5488995016611294e-07, "loss": 0.1572, "step": 4660 }, { "epoch": 0.97, "learning_rate": 1.5485880398671095e-07, "loss": 0.1602, "step": 4661 }, { "epoch": 0.97, "learning_rate": 1.5482765780730897e-07, "loss": 0.2014, "step": 4662 }, { "epoch": 0.97, "learning_rate": 1.5479651162790699e-07, "loss": 0.1336, "step": 4663 }, { "epoch": 0.97, "learning_rate": 1.54765365448505e-07, "loss": 0.1594, "step": 4664 }, { "epoch": 0.97, "learning_rate": 1.5473421926910296e-07, "loss": 0.2061, "step": 4665 }, { "epoch": 0.97, "learning_rate": 1.5470307308970098e-07, "loss": 0.1783, "step": 4666 }, { "epoch": 0.97, "learning_rate": 1.54671926910299e-07, "loss": 0.1677, "step": 4667 }, { "epoch": 0.97, "learning_rate": 1.54640780730897e-07, "loss": 0.2373, "step": 4668 }, { "epoch": 0.97, "learning_rate": 1.54609634551495e-07, "loss": 0.2282, "step": 4669 }, { "epoch": 0.97, "learning_rate": 1.5457848837209302e-07, "loss": 0.2584, "step": 4670 }, { "epoch": 0.97, "learning_rate": 1.5454734219269103e-07, "loss": 0.1453, "step": 4671 }, { "epoch": 0.97, "learning_rate": 1.5451619601328902e-07, "loss": 0.252, "step": 4672 }, { "epoch": 0.97, "learning_rate": 1.5448504983388704e-07, "loss": 0.1971, "step": 4673 }, { "epoch": 0.97, "learning_rate": 1.5445390365448503e-07, "loss": 0.1528, "step": 4674 }, { "epoch": 0.97, "learning_rate": 1.5442275747508304e-07, "loss": 0.2336, "step": 4675 }, { "epoch": 0.97, "learning_rate": 1.5439161129568106e-07, "loss": 0.1929, "step": 4676 }, { "epoch": 0.97, "learning_rate": 1.5436046511627907e-07, "loss": 0.2276, "step": 4677 }, { "epoch": 0.97, "learning_rate": 1.543293189368771e-07, "loss": 0.2032, "step": 4678 }, { "epoch": 0.97, "learning_rate": 1.5429817275747505e-07, "loss": 0.1816, "step": 4679 }, { "epoch": 0.97, "learning_rate": 1.5426702657807307e-07, "loss": 0.2574, "step": 4680 }, { "epoch": 0.97, "learning_rate": 1.5423588039867109e-07, "loss": 0.1629, "step": 4681 }, { "epoch": 0.97, "learning_rate": 1.542047342192691e-07, "loss": 0.2283, "step": 4682 }, { "epoch": 0.97, "learning_rate": 1.541735880398671e-07, "loss": 0.227, "step": 4683 }, { "epoch": 0.97, "learning_rate": 1.541424418604651e-07, "loss": 0.2203, "step": 4684 }, { "epoch": 0.97, "learning_rate": 1.5411129568106312e-07, "loss": 0.233, "step": 4685 }, { "epoch": 0.97, "learning_rate": 1.5408014950166114e-07, "loss": 0.1859, "step": 4686 }, { "epoch": 0.97, "learning_rate": 1.5404900332225915e-07, "loss": 0.2839, "step": 4687 }, { "epoch": 0.97, "learning_rate": 1.5401785714285712e-07, "loss": 0.165, "step": 4688 }, { "epoch": 0.97, "learning_rate": 1.5398671096345513e-07, "loss": 0.2199, "step": 4689 }, { "epoch": 0.97, "learning_rate": 1.5395556478405315e-07, "loss": 0.2237, "step": 4690 }, { "epoch": 0.97, "learning_rate": 1.5392441860465116e-07, "loss": 0.199, "step": 4691 }, { "epoch": 0.97, "learning_rate": 1.5389327242524918e-07, "loss": 0.1731, "step": 4692 }, { "epoch": 0.97, "learning_rate": 1.5386212624584717e-07, "loss": 0.1925, "step": 4693 }, { "epoch": 0.97, "learning_rate": 1.5383098006644516e-07, "loss": 0.1717, "step": 4694 }, { "epoch": 0.97, "learning_rate": 1.5379983388704317e-07, "loss": 0.2186, "step": 4695 }, { "epoch": 0.98, "learning_rate": 1.537686877076412e-07, "loss": 0.2205, "step": 4696 }, { "epoch": 0.98, "learning_rate": 1.537375415282392e-07, "loss": 0.1558, "step": 4697 }, { "epoch": 0.98, "learning_rate": 1.537063953488372e-07, "loss": 0.1594, "step": 4698 }, { "epoch": 0.98, "learning_rate": 1.536752491694352e-07, "loss": 0.1826, "step": 4699 }, { "epoch": 0.98, "learning_rate": 1.5364410299003323e-07, "loss": 0.2381, "step": 4700 }, { "epoch": 0.98, "learning_rate": 1.5361295681063124e-07, "loss": 0.2034, "step": 4701 }, { "epoch": 0.98, "learning_rate": 1.535818106312292e-07, "loss": 0.1635, "step": 4702 }, { "epoch": 0.98, "learning_rate": 1.5355066445182722e-07, "loss": 0.2167, "step": 4703 }, { "epoch": 0.98, "learning_rate": 1.5351951827242524e-07, "loss": 0.2321, "step": 4704 }, { "epoch": 0.98, "learning_rate": 1.5348837209302325e-07, "loss": 0.149, "step": 4705 }, { "epoch": 0.98, "learning_rate": 1.5345722591362127e-07, "loss": 0.2613, "step": 4706 }, { "epoch": 0.98, "learning_rate": 1.5342607973421926e-07, "loss": 0.2364, "step": 4707 }, { "epoch": 0.98, "learning_rate": 1.5339493355481727e-07, "loss": 0.1653, "step": 4708 }, { "epoch": 0.98, "learning_rate": 1.5336378737541526e-07, "loss": 0.164, "step": 4709 }, { "epoch": 0.98, "learning_rate": 1.5333264119601328e-07, "loss": 0.234, "step": 4710 }, { "epoch": 0.98, "learning_rate": 1.533014950166113e-07, "loss": 0.2037, "step": 4711 }, { "epoch": 0.98, "learning_rate": 1.5327034883720928e-07, "loss": 0.2108, "step": 4712 }, { "epoch": 0.98, "learning_rate": 1.532392026578073e-07, "loss": 0.2648, "step": 4713 }, { "epoch": 0.98, "learning_rate": 1.5320805647840532e-07, "loss": 0.1595, "step": 4714 }, { "epoch": 0.98, "learning_rate": 1.5317691029900333e-07, "loss": 0.2154, "step": 4715 }, { "epoch": 0.98, "learning_rate": 1.531457641196013e-07, "loss": 0.2486, "step": 4716 }, { "epoch": 0.98, "learning_rate": 1.531146179401993e-07, "loss": 0.166, "step": 4717 }, { "epoch": 0.98, "learning_rate": 1.5308347176079733e-07, "loss": 0.1977, "step": 4718 }, { "epoch": 0.98, "learning_rate": 1.5305232558139534e-07, "loss": 0.2028, "step": 4719 }, { "epoch": 0.98, "learning_rate": 1.5302117940199336e-07, "loss": 0.1856, "step": 4720 }, { "epoch": 0.98, "learning_rate": 1.5299003322259135e-07, "loss": 0.2504, "step": 4721 }, { "epoch": 0.98, "learning_rate": 1.5295888704318936e-07, "loss": 0.2196, "step": 4722 }, { "epoch": 0.98, "learning_rate": 1.5292774086378738e-07, "loss": 0.1634, "step": 4723 }, { "epoch": 0.98, "learning_rate": 1.528965946843854e-07, "loss": 0.2618, "step": 4724 }, { "epoch": 0.98, "learning_rate": 1.5286544850498338e-07, "loss": 0.2583, "step": 4725 }, { "epoch": 0.98, "learning_rate": 1.5283430232558137e-07, "loss": 0.1932, "step": 4726 }, { "epoch": 0.98, "learning_rate": 1.528031561461794e-07, "loss": 0.2201, "step": 4727 }, { "epoch": 0.98, "learning_rate": 1.527720099667774e-07, "loss": 0.203, "step": 4728 }, { "epoch": 0.98, "learning_rate": 1.5274086378737542e-07, "loss": 0.1623, "step": 4729 }, { "epoch": 0.98, "learning_rate": 1.527097176079734e-07, "loss": 0.2276, "step": 4730 }, { "epoch": 0.98, "learning_rate": 1.526785714285714e-07, "loss": 0.2261, "step": 4731 }, { "epoch": 0.98, "learning_rate": 1.5264742524916942e-07, "loss": 0.166, "step": 4732 }, { "epoch": 0.98, "learning_rate": 1.5261627906976743e-07, "loss": 0.2053, "step": 4733 }, { "epoch": 0.98, "learning_rate": 1.5258513289036545e-07, "loss": 0.13, "step": 4734 }, { "epoch": 0.98, "learning_rate": 1.5255398671096344e-07, "loss": 0.2177, "step": 4735 }, { "epoch": 0.98, "learning_rate": 1.5252284053156145e-07, "loss": 0.187, "step": 4736 }, { "epoch": 0.98, "learning_rate": 1.5249169435215947e-07, "loss": 0.1766, "step": 4737 }, { "epoch": 0.98, "learning_rate": 1.5246054817275748e-07, "loss": 0.2188, "step": 4738 }, { "epoch": 0.98, "learning_rate": 1.524294019933555e-07, "loss": 0.1966, "step": 4739 }, { "epoch": 0.98, "learning_rate": 1.5239825581395346e-07, "loss": 0.1897, "step": 4740 }, { "epoch": 0.98, "learning_rate": 1.5236710963455148e-07, "loss": 0.1955, "step": 4741 }, { "epoch": 0.98, "learning_rate": 1.523359634551495e-07, "loss": 0.2086, "step": 4742 }, { "epoch": 0.98, "learning_rate": 1.523048172757475e-07, "loss": 0.2536, "step": 4743 }, { "epoch": 0.99, "learning_rate": 1.5227367109634553e-07, "loss": 0.1879, "step": 4744 }, { "epoch": 0.99, "learning_rate": 1.5224252491694352e-07, "loss": 0.1799, "step": 4745 }, { "epoch": 0.99, "learning_rate": 1.522113787375415e-07, "loss": 0.2238, "step": 4746 }, { "epoch": 0.99, "learning_rate": 1.5218023255813952e-07, "loss": 0.1514, "step": 4747 }, { "epoch": 0.99, "learning_rate": 1.5214908637873754e-07, "loss": 0.1739, "step": 4748 }, { "epoch": 0.99, "learning_rate": 1.5211794019933553e-07, "loss": 0.2092, "step": 4749 }, { "epoch": 0.99, "learning_rate": 1.5208679401993354e-07, "loss": 0.1429, "step": 4750 }, { "epoch": 0.99, "learning_rate": 1.5205564784053156e-07, "loss": 0.2015, "step": 4751 }, { "epoch": 0.99, "learning_rate": 1.5202450166112957e-07, "loss": 0.1404, "step": 4752 }, { "epoch": 0.99, "learning_rate": 1.519933554817276e-07, "loss": 0.1967, "step": 4753 }, { "epoch": 0.99, "learning_rate": 1.5196220930232555e-07, "loss": 0.2218, "step": 4754 }, { "epoch": 0.99, "learning_rate": 1.5193106312292357e-07, "loss": 0.2195, "step": 4755 }, { "epoch": 0.99, "learning_rate": 1.5189991694352158e-07, "loss": 0.1738, "step": 4756 }, { "epoch": 0.99, "learning_rate": 1.518687707641196e-07, "loss": 0.1888, "step": 4757 }, { "epoch": 0.99, "learning_rate": 1.5183762458471762e-07, "loss": 0.2115, "step": 4758 }, { "epoch": 0.99, "learning_rate": 1.518064784053156e-07, "loss": 0.1928, "step": 4759 }, { "epoch": 0.99, "learning_rate": 1.5177533222591362e-07, "loss": 0.1551, "step": 4760 }, { "epoch": 0.99, "learning_rate": 1.5174418604651164e-07, "loss": 0.2022, "step": 4761 }, { "epoch": 0.99, "learning_rate": 1.5171303986710963e-07, "loss": 0.1725, "step": 4762 }, { "epoch": 0.99, "learning_rate": 1.5168189368770762e-07, "loss": 0.1484, "step": 4763 }, { "epoch": 0.99, "learning_rate": 1.5165074750830563e-07, "loss": 0.1989, "step": 4764 }, { "epoch": 0.99, "learning_rate": 1.5161960132890365e-07, "loss": 0.1852, "step": 4765 }, { "epoch": 0.99, "learning_rate": 1.5158845514950166e-07, "loss": 0.1452, "step": 4766 }, { "epoch": 0.99, "learning_rate": 1.5155730897009968e-07, "loss": 0.2117, "step": 4767 }, { "epoch": 0.99, "learning_rate": 1.5152616279069764e-07, "loss": 0.1895, "step": 4768 }, { "epoch": 0.99, "learning_rate": 1.5149501661129566e-07, "loss": 0.1998, "step": 4769 }, { "epoch": 0.99, "learning_rate": 1.5146387043189367e-07, "loss": 0.168, "step": 4770 }, { "epoch": 0.99, "learning_rate": 1.514327242524917e-07, "loss": 0.1494, "step": 4771 }, { "epoch": 0.99, "learning_rate": 1.514015780730897e-07, "loss": 0.1558, "step": 4772 }, { "epoch": 0.99, "learning_rate": 1.513704318936877e-07, "loss": 0.1512, "step": 4773 }, { "epoch": 0.99, "learning_rate": 1.513392857142857e-07, "loss": 0.2367, "step": 4774 }, { "epoch": 0.99, "learning_rate": 1.5130813953488373e-07, "loss": 0.1912, "step": 4775 }, { "epoch": 0.99, "learning_rate": 1.5127699335548174e-07, "loss": 0.1795, "step": 4776 }, { "epoch": 0.99, "learning_rate": 1.5124584717607973e-07, "loss": 0.2096, "step": 4777 }, { "epoch": 0.99, "learning_rate": 1.5121470099667772e-07, "loss": 0.1812, "step": 4778 }, { "epoch": 0.99, "learning_rate": 1.5118355481727574e-07, "loss": 0.2396, "step": 4779 }, { "epoch": 0.99, "learning_rate": 1.5115240863787375e-07, "loss": 0.1655, "step": 4780 }, { "epoch": 0.99, "learning_rate": 1.5112126245847177e-07, "loss": 0.1624, "step": 4781 }, { "epoch": 0.99, "learning_rate": 1.5109011627906976e-07, "loss": 0.1705, "step": 4782 }, { "epoch": 0.99, "learning_rate": 1.5105897009966777e-07, "loss": 0.2731, "step": 4783 }, { "epoch": 0.99, "learning_rate": 1.5102782392026576e-07, "loss": 0.1335, "step": 4784 }, { "epoch": 0.99, "learning_rate": 1.5099667774086378e-07, "loss": 0.1325, "step": 4785 }, { "epoch": 0.99, "learning_rate": 1.509655315614618e-07, "loss": 0.1688, "step": 4786 }, { "epoch": 0.99, "learning_rate": 1.5093438538205978e-07, "loss": 0.1816, "step": 4787 }, { "epoch": 0.99, "learning_rate": 1.509032392026578e-07, "loss": 0.1392, "step": 4788 }, { "epoch": 0.99, "learning_rate": 1.5087209302325581e-07, "loss": 0.1897, "step": 4789 }, { "epoch": 0.99, "learning_rate": 1.5084094684385383e-07, "loss": 0.2053, "step": 4790 }, { "epoch": 0.99, "learning_rate": 1.5080980066445185e-07, "loss": 0.164, "step": 4791 }, { "epoch": 1.0, "learning_rate": 1.507786544850498e-07, "loss": 0.2152, "step": 4792 }, { "epoch": 1.0, "learning_rate": 1.5074750830564783e-07, "loss": 0.2073, "step": 4793 }, { "epoch": 1.0, "learning_rate": 1.5071636212624584e-07, "loss": 0.193, "step": 4794 }, { "epoch": 1.0, "learning_rate": 1.5068521594684386e-07, "loss": 0.2004, "step": 4795 }, { "epoch": 1.0, "learning_rate": 1.5065406976744185e-07, "loss": 0.2452, "step": 4796 }, { "epoch": 1.0, "learning_rate": 1.5062292358803986e-07, "loss": 0.2007, "step": 4797 }, { "epoch": 1.0, "learning_rate": 1.5059177740863788e-07, "loss": 0.2019, "step": 4798 }, { "epoch": 1.0, "learning_rate": 1.5056063122923587e-07, "loss": 0.2132, "step": 4799 }, { "epoch": 1.0, "learning_rate": 1.5052948504983388e-07, "loss": 0.1927, "step": 4800 }, { "epoch": 1.0, "learning_rate": 1.5049833887043187e-07, "loss": 0.1865, "step": 4801 }, { "epoch": 1.0, "learning_rate": 1.504671926910299e-07, "loss": 0.1753, "step": 4802 }, { "epoch": 1.0, "learning_rate": 1.504360465116279e-07, "loss": 0.2319, "step": 4803 }, { "epoch": 1.0, "learning_rate": 1.5040490033222592e-07, "loss": 0.1415, "step": 4804 }, { "epoch": 1.0, "learning_rate": 1.5037375415282394e-07, "loss": 0.2711, "step": 4805 }, { "epoch": 1.0, "learning_rate": 1.503426079734219e-07, "loss": 0.1726, "step": 4806 }, { "epoch": 1.0, "learning_rate": 1.5031146179401991e-07, "loss": 0.2751, "step": 4807 }, { "epoch": 1.0, "learning_rate": 1.5028031561461793e-07, "loss": 0.1974, "step": 4808 }, { "epoch": 1.0, "learning_rate": 1.5024916943521595e-07, "loss": 0.2648, "step": 4809 }, { "epoch": 1.0, "learning_rate": 1.5021802325581394e-07, "loss": 0.2307, "step": 4810 }, { "epoch": 1.0, "learning_rate": 1.5018687707641195e-07, "loss": 0.1855, "step": 4811 }, { "epoch": 1.0, "learning_rate": 1.5015573089700997e-07, "loss": 0.2155, "step": 4812 }, { "epoch": 1.0, "learning_rate": 1.5012458471760798e-07, "loss": 0.2153, "step": 4813 }, { "epoch": 1.0, "learning_rate": 1.5009343853820597e-07, "loss": 0.1926, "step": 4814 }, { "epoch": 1.0, "learning_rate": 1.5006229235880396e-07, "loss": 0.1852, "step": 4815 }, { "epoch": 1.0, "learning_rate": 1.5003114617940198e-07, "loss": 0.2009, "step": 4816 }, { "epoch": 1.0, "eval_loss": 0.3485559821128845, "eval_runtime": 3318.7586, "eval_samples_per_second": 4.803, "eval_steps_per_second": 0.1, "eval_wer": 0.2940012490088624, "step": 4816 }, { "epoch": 1.0, "learning_rate": 1.5e-07, "loss": 0.2198, "step": 4817 }, { "epoch": 1.0, "learning_rate": 1.49968853820598e-07, "loss": 0.2095, "step": 4818 }, { "epoch": 1.0, "learning_rate": 1.49937707641196e-07, "loss": 0.1922, "step": 4819 }, { "epoch": 1.0, "learning_rate": 1.4990656146179401e-07, "loss": 0.2167, "step": 4820 }, { "epoch": 1.0, "learning_rate": 1.49875415282392e-07, "loss": 0.2095, "step": 4821 }, { "epoch": 1.0, "learning_rate": 1.4984426910299002e-07, "loss": 0.22, "step": 4822 }, { "epoch": 1.0, "learning_rate": 1.4981312292358804e-07, "loss": 0.1607, "step": 4823 }, { "epoch": 1.0, "learning_rate": 1.4978197674418605e-07, "loss": 0.1622, "step": 4824 }, { "epoch": 1.0, "learning_rate": 1.4975083056478404e-07, "loss": 0.2052, "step": 4825 }, { "epoch": 1.0, "learning_rate": 1.4971968438538206e-07, "loss": 0.1885, "step": 4826 }, { "epoch": 1.0, "learning_rate": 1.4968853820598007e-07, "loss": 0.143, "step": 4827 }, { "epoch": 1.0, "learning_rate": 1.4965739202657806e-07, "loss": 0.1951, "step": 4828 }, { "epoch": 1.0, "learning_rate": 1.4962624584717608e-07, "loss": 0.1613, "step": 4829 }, { "epoch": 1.0, "learning_rate": 1.4959509966777407e-07, "loss": 0.2333, "step": 4830 }, { "epoch": 1.0, "learning_rate": 1.4956395348837208e-07, "loss": 0.2493, "step": 4831 }, { "epoch": 1.0, "learning_rate": 1.495328073089701e-07, "loss": 0.1941, "step": 4832 }, { "epoch": 1.0, "learning_rate": 1.495016611295681e-07, "loss": 0.1671, "step": 4833 }, { "epoch": 1.0, "learning_rate": 1.494705149501661e-07, "loss": 0.1432, "step": 4834 }, { "epoch": 1.0, "learning_rate": 1.4943936877076412e-07, "loss": 0.2038, "step": 4835 }, { "epoch": 1.0, "learning_rate": 1.494082225913621e-07, "loss": 0.2251, "step": 4836 }, { "epoch": 1.0, "learning_rate": 1.4937707641196012e-07, "loss": 0.1876, "step": 4837 }, { "epoch": 1.0, "learning_rate": 1.4934593023255814e-07, "loss": 0.1947, "step": 4838 }, { "epoch": 1.0, "learning_rate": 1.4931478405315616e-07, "loss": 0.1431, "step": 4839 }, { "epoch": 1.0, "learning_rate": 1.4928363787375415e-07, "loss": 0.1424, "step": 4840 }, { "epoch": 1.01, "learning_rate": 1.4925249169435216e-07, "loss": 0.1317, "step": 4841 }, { "epoch": 1.01, "learning_rate": 1.4922134551495015e-07, "loss": 0.1744, "step": 4842 }, { "epoch": 1.01, "learning_rate": 1.4919019933554817e-07, "loss": 0.1412, "step": 4843 }, { "epoch": 1.01, "learning_rate": 1.4915905315614618e-07, "loss": 0.2088, "step": 4844 }, { "epoch": 1.01, "learning_rate": 1.4912790697674417e-07, "loss": 0.1577, "step": 4845 }, { "epoch": 1.01, "learning_rate": 1.490967607973422e-07, "loss": 0.1922, "step": 4846 }, { "epoch": 1.01, "learning_rate": 1.4906561461794018e-07, "loss": 0.151, "step": 4847 }, { "epoch": 1.01, "learning_rate": 1.490344684385382e-07, "loss": 0.1661, "step": 4848 }, { "epoch": 1.01, "learning_rate": 1.490033222591362e-07, "loss": 0.2665, "step": 4849 }, { "epoch": 1.01, "learning_rate": 1.4897217607973422e-07, "loss": 0.1923, "step": 4850 }, { "epoch": 1.01, "learning_rate": 1.4894102990033221e-07, "loss": 0.1538, "step": 4851 }, { "epoch": 1.01, "learning_rate": 1.4890988372093023e-07, "loss": 0.1807, "step": 4852 }, { "epoch": 1.01, "learning_rate": 1.4887873754152825e-07, "loss": 0.1408, "step": 4853 }, { "epoch": 1.01, "learning_rate": 1.4884759136212623e-07, "loss": 0.158, "step": 4854 }, { "epoch": 1.01, "learning_rate": 1.4881644518272425e-07, "loss": 0.1546, "step": 4855 }, { "epoch": 1.01, "learning_rate": 1.4878529900332224e-07, "loss": 0.1847, "step": 4856 }, { "epoch": 1.01, "learning_rate": 1.4875415282392026e-07, "loss": 0.1676, "step": 4857 }, { "epoch": 1.01, "learning_rate": 1.4872300664451827e-07, "loss": 0.2425, "step": 4858 }, { "epoch": 1.01, "learning_rate": 1.4869186046511626e-07, "loss": 0.2001, "step": 4859 }, { "epoch": 1.01, "learning_rate": 1.4866071428571428e-07, "loss": 0.1479, "step": 4860 }, { "epoch": 1.01, "learning_rate": 1.486295681063123e-07, "loss": 0.2746, "step": 4861 }, { "epoch": 1.01, "learning_rate": 1.4859842192691028e-07, "loss": 0.2338, "step": 4862 }, { "epoch": 1.01, "learning_rate": 1.485672757475083e-07, "loss": 0.2167, "step": 4863 }, { "epoch": 1.01, "learning_rate": 1.4853612956810631e-07, "loss": 0.186, "step": 4864 }, { "epoch": 1.01, "learning_rate": 1.4850498338870433e-07, "loss": 0.1562, "step": 4865 }, { "epoch": 1.01, "learning_rate": 1.4847383720930232e-07, "loss": 0.2767, "step": 4866 }, { "epoch": 1.01, "learning_rate": 1.4844269102990033e-07, "loss": 0.1761, "step": 4867 }, { "epoch": 1.01, "learning_rate": 1.4841154485049832e-07, "loss": 0.1687, "step": 4868 }, { "epoch": 1.01, "learning_rate": 1.4838039867109634e-07, "loss": 0.1786, "step": 4869 }, { "epoch": 1.01, "learning_rate": 1.4834925249169433e-07, "loss": 0.1595, "step": 4870 }, { "epoch": 1.01, "learning_rate": 1.4831810631229234e-07, "loss": 0.1, "step": 4871 }, { "epoch": 1.01, "learning_rate": 1.4828696013289036e-07, "loss": 0.1335, "step": 4872 }, { "epoch": 1.01, "learning_rate": 1.4825581395348835e-07, "loss": 0.2255, "step": 4873 }, { "epoch": 1.01, "learning_rate": 1.4822466777408637e-07, "loss": 0.1621, "step": 4874 }, { "epoch": 1.01, "learning_rate": 1.4819352159468438e-07, "loss": 0.2552, "step": 4875 }, { "epoch": 1.01, "learning_rate": 1.481623754152824e-07, "loss": 0.2008, "step": 4876 }, { "epoch": 1.01, "learning_rate": 1.481312292358804e-07, "loss": 0.1695, "step": 4877 }, { "epoch": 1.01, "learning_rate": 1.481000830564784e-07, "loss": 0.1399, "step": 4878 }, { "epoch": 1.01, "learning_rate": 1.4806893687707642e-07, "loss": 0.2069, "step": 4879 }, { "epoch": 1.01, "learning_rate": 1.480377906976744e-07, "loss": 0.1447, "step": 4880 }, { "epoch": 1.01, "learning_rate": 1.4800664451827242e-07, "loss": 0.1842, "step": 4881 }, { "epoch": 1.01, "learning_rate": 1.479754983388704e-07, "loss": 0.1416, "step": 4882 }, { "epoch": 1.01, "learning_rate": 1.4794435215946843e-07, "loss": 0.1909, "step": 4883 }, { "epoch": 1.01, "learning_rate": 1.4791320598006644e-07, "loss": 0.204, "step": 4884 }, { "epoch": 1.01, "learning_rate": 1.4788205980066443e-07, "loss": 0.2076, "step": 4885 }, { "epoch": 1.01, "learning_rate": 1.4785091362126245e-07, "loss": 0.1463, "step": 4886 }, { "epoch": 1.01, "learning_rate": 1.4781976744186047e-07, "loss": 0.1876, "step": 4887 }, { "epoch": 1.01, "learning_rate": 1.4778862126245848e-07, "loss": 0.1902, "step": 4888 }, { "epoch": 1.02, "learning_rate": 1.4775747508305647e-07, "loss": 0.2065, "step": 4889 }, { "epoch": 1.02, "learning_rate": 1.4772632890365449e-07, "loss": 0.1573, "step": 4890 }, { "epoch": 1.02, "learning_rate": 1.476951827242525e-07, "loss": 0.1975, "step": 4891 }, { "epoch": 1.02, "learning_rate": 1.476640365448505e-07, "loss": 0.1528, "step": 4892 }, { "epoch": 1.02, "learning_rate": 1.476328903654485e-07, "loss": 0.1567, "step": 4893 }, { "epoch": 1.02, "learning_rate": 1.476017441860465e-07, "loss": 0.1638, "step": 4894 }, { "epoch": 1.02, "learning_rate": 1.475705980066445e-07, "loss": 0.2041, "step": 4895 }, { "epoch": 1.02, "learning_rate": 1.475394518272425e-07, "loss": 0.1581, "step": 4896 }, { "epoch": 1.02, "learning_rate": 1.4750830564784052e-07, "loss": 0.1784, "step": 4897 }, { "epoch": 1.02, "learning_rate": 1.4747715946843853e-07, "loss": 0.1362, "step": 4898 }, { "epoch": 1.02, "learning_rate": 1.4744601328903652e-07, "loss": 0.1906, "step": 4899 }, { "epoch": 1.02, "learning_rate": 1.4741486710963454e-07, "loss": 0.1567, "step": 4900 }, { "epoch": 1.02, "learning_rate": 1.4738372093023255e-07, "loss": 0.123, "step": 4901 }, { "epoch": 1.02, "learning_rate": 1.4735257475083057e-07, "loss": 0.2321, "step": 4902 }, { "epoch": 1.02, "learning_rate": 1.4732142857142856e-07, "loss": 0.1904, "step": 4903 }, { "epoch": 1.02, "learning_rate": 1.4729028239202658e-07, "loss": 0.1656, "step": 4904 }, { "epoch": 1.02, "learning_rate": 1.472591362126246e-07, "loss": 0.1935, "step": 4905 }, { "epoch": 1.02, "learning_rate": 1.4722799003322258e-07, "loss": 0.2086, "step": 4906 }, { "epoch": 1.02, "learning_rate": 1.471968438538206e-07, "loss": 0.2151, "step": 4907 }, { "epoch": 1.02, "learning_rate": 1.4716569767441859e-07, "loss": 0.1975, "step": 4908 }, { "epoch": 1.02, "learning_rate": 1.471345514950166e-07, "loss": 0.2031, "step": 4909 }, { "epoch": 1.02, "learning_rate": 1.471034053156146e-07, "loss": 0.1738, "step": 4910 }, { "epoch": 1.02, "learning_rate": 1.470722591362126e-07, "loss": 0.1784, "step": 4911 }, { "epoch": 1.02, "learning_rate": 1.4704111295681062e-07, "loss": 0.1903, "step": 4912 }, { "epoch": 1.02, "learning_rate": 1.4700996677740864e-07, "loss": 0.1929, "step": 4913 }, { "epoch": 1.02, "learning_rate": 1.4697882059800665e-07, "loss": 0.1771, "step": 4914 }, { "epoch": 1.02, "learning_rate": 1.4694767441860464e-07, "loss": 0.2042, "step": 4915 }, { "epoch": 1.02, "learning_rate": 1.4691652823920266e-07, "loss": 0.1489, "step": 4916 }, { "epoch": 1.02, "learning_rate": 1.4688538205980065e-07, "loss": 0.1661, "step": 4917 }, { "epoch": 1.02, "learning_rate": 1.4685423588039867e-07, "loss": 0.234, "step": 4918 }, { "epoch": 1.02, "learning_rate": 1.4682308970099668e-07, "loss": 0.1606, "step": 4919 }, { "epoch": 1.02, "learning_rate": 1.4679194352159467e-07, "loss": 0.1631, "step": 4920 }, { "epoch": 1.02, "learning_rate": 1.4676079734219269e-07, "loss": 0.1867, "step": 4921 }, { "epoch": 1.02, "learning_rate": 1.4672965116279068e-07, "loss": 0.1619, "step": 4922 }, { "epoch": 1.02, "learning_rate": 1.466985049833887e-07, "loss": 0.1325, "step": 4923 }, { "epoch": 1.02, "learning_rate": 1.466673588039867e-07, "loss": 0.1735, "step": 4924 }, { "epoch": 1.02, "learning_rate": 1.4663621262458472e-07, "loss": 0.1401, "step": 4925 }, { "epoch": 1.02, "learning_rate": 1.466050664451827e-07, "loss": 0.1393, "step": 4926 }, { "epoch": 1.02, "learning_rate": 1.4657392026578073e-07, "loss": 0.1866, "step": 4927 }, { "epoch": 1.02, "learning_rate": 1.4654277408637874e-07, "loss": 0.1991, "step": 4928 }, { "epoch": 1.02, "learning_rate": 1.4651162790697673e-07, "loss": 0.1764, "step": 4929 }, { "epoch": 1.02, "learning_rate": 1.4648048172757475e-07, "loss": 0.1283, "step": 4930 }, { "epoch": 1.02, "learning_rate": 1.4644933554817276e-07, "loss": 0.1499, "step": 4931 }, { "epoch": 1.02, "learning_rate": 1.4641818936877075e-07, "loss": 0.1677, "step": 4932 }, { "epoch": 1.02, "learning_rate": 1.4638704318936877e-07, "loss": 0.2008, "step": 4933 }, { "epoch": 1.02, "learning_rate": 1.4635589700996676e-07, "loss": 0.1422, "step": 4934 }, { "epoch": 1.02, "learning_rate": 1.4632475083056478e-07, "loss": 0.1933, "step": 4935 }, { "epoch": 1.02, "learning_rate": 1.4629360465116276e-07, "loss": 0.1641, "step": 4936 }, { "epoch": 1.03, "learning_rate": 1.4626245847176078e-07, "loss": 0.2095, "step": 4937 }, { "epoch": 1.03, "learning_rate": 1.462313122923588e-07, "loss": 0.1868, "step": 4938 }, { "epoch": 1.03, "learning_rate": 1.462001661129568e-07, "loss": 0.2342, "step": 4939 }, { "epoch": 1.03, "learning_rate": 1.4616901993355483e-07, "loss": 0.2362, "step": 4940 }, { "epoch": 1.03, "learning_rate": 1.4613787375415282e-07, "loss": 0.1531, "step": 4941 }, { "epoch": 1.03, "learning_rate": 1.4610672757475083e-07, "loss": 0.1765, "step": 4942 }, { "epoch": 1.03, "learning_rate": 1.4607558139534882e-07, "loss": 0.1604, "step": 4943 }, { "epoch": 1.03, "learning_rate": 1.4604443521594684e-07, "loss": 0.2163, "step": 4944 }, { "epoch": 1.03, "learning_rate": 1.4601328903654485e-07, "loss": 0.1585, "step": 4945 }, { "epoch": 1.03, "learning_rate": 1.4598214285714284e-07, "loss": 0.1569, "step": 4946 }, { "epoch": 1.03, "learning_rate": 1.4595099667774086e-07, "loss": 0.1815, "step": 4947 }, { "epoch": 1.03, "learning_rate": 1.4591985049833885e-07, "loss": 0.2449, "step": 4948 }, { "epoch": 1.03, "learning_rate": 1.4588870431893686e-07, "loss": 0.1759, "step": 4949 }, { "epoch": 1.03, "learning_rate": 1.4585755813953488e-07, "loss": 0.1486, "step": 4950 }, { "epoch": 1.03, "learning_rate": 1.458264119601329e-07, "loss": 0.1652, "step": 4951 }, { "epoch": 1.03, "learning_rate": 1.4579526578073089e-07, "loss": 0.1808, "step": 4952 }, { "epoch": 1.03, "learning_rate": 1.457641196013289e-07, "loss": 0.1455, "step": 4953 }, { "epoch": 1.03, "learning_rate": 1.4573297342192692e-07, "loss": 0.1936, "step": 4954 }, { "epoch": 1.03, "learning_rate": 1.457018272425249e-07, "loss": 0.1869, "step": 4955 }, { "epoch": 1.03, "learning_rate": 1.4567068106312292e-07, "loss": 0.2224, "step": 4956 }, { "epoch": 1.03, "learning_rate": 1.456395348837209e-07, "loss": 0.1732, "step": 4957 }, { "epoch": 1.03, "learning_rate": 1.4560838870431893e-07, "loss": 0.1924, "step": 4958 }, { "epoch": 1.03, "learning_rate": 1.4557724252491694e-07, "loss": 0.1458, "step": 4959 }, { "epoch": 1.03, "learning_rate": 1.4554609634551493e-07, "loss": 0.1498, "step": 4960 }, { "epoch": 1.03, "learning_rate": 1.4551495016611295e-07, "loss": 0.2647, "step": 4961 }, { "epoch": 1.03, "learning_rate": 1.4548380398671096e-07, "loss": 0.1895, "step": 4962 }, { "epoch": 1.03, "learning_rate": 1.4545265780730895e-07, "loss": 0.1563, "step": 4963 }, { "epoch": 1.03, "learning_rate": 1.4542151162790697e-07, "loss": 0.2592, "step": 4964 }, { "epoch": 1.03, "learning_rate": 1.4539036544850499e-07, "loss": 0.1956, "step": 4965 }, { "epoch": 1.03, "learning_rate": 1.45359219269103e-07, "loss": 0.1688, "step": 4966 }, { "epoch": 1.03, "learning_rate": 1.45328073089701e-07, "loss": 0.2138, "step": 4967 }, { "epoch": 1.03, "learning_rate": 1.45296926910299e-07, "loss": 0.1809, "step": 4968 }, { "epoch": 1.03, "learning_rate": 1.45265780730897e-07, "loss": 0.2203, "step": 4969 }, { "epoch": 1.03, "learning_rate": 1.45234634551495e-07, "loss": 0.2236, "step": 4970 }, { "epoch": 1.03, "learning_rate": 1.4520348837209303e-07, "loss": 0.2223, "step": 4971 }, { "epoch": 1.03, "learning_rate": 1.4517234219269102e-07, "loss": 0.2081, "step": 4972 }, { "epoch": 1.03, "learning_rate": 1.4514119601328903e-07, "loss": 0.1928, "step": 4973 }, { "epoch": 1.03, "learning_rate": 1.4511004983388702e-07, "loss": 0.1635, "step": 4974 }, { "epoch": 1.03, "learning_rate": 1.4507890365448504e-07, "loss": 0.1487, "step": 4975 }, { "epoch": 1.03, "learning_rate": 1.4504775747508305e-07, "loss": 0.1765, "step": 4976 }, { "epoch": 1.03, "learning_rate": 1.4501661129568107e-07, "loss": 0.1376, "step": 4977 }, { "epoch": 1.03, "learning_rate": 1.4498546511627906e-07, "loss": 0.1733, "step": 4978 }, { "epoch": 1.03, "learning_rate": 1.4495431893687707e-07, "loss": 0.2042, "step": 4979 }, { "epoch": 1.03, "learning_rate": 1.449231727574751e-07, "loss": 0.2014, "step": 4980 }, { "epoch": 1.03, "learning_rate": 1.4489202657807308e-07, "loss": 0.1461, "step": 4981 }, { "epoch": 1.03, "learning_rate": 1.448608803986711e-07, "loss": 0.1496, "step": 4982 }, { "epoch": 1.03, "learning_rate": 1.4482973421926908e-07, "loss": 0.1505, "step": 4983 }, { "epoch": 1.03, "learning_rate": 1.447985880398671e-07, "loss": 0.1503, "step": 4984 }, { "epoch": 1.04, "learning_rate": 1.4476744186046512e-07, "loss": 0.1652, "step": 4985 }, { "epoch": 1.04, "learning_rate": 1.447362956810631e-07, "loss": 0.1969, "step": 4986 }, { "epoch": 1.04, "learning_rate": 1.4470514950166112e-07, "loss": 0.2544, "step": 4987 }, { "epoch": 1.04, "learning_rate": 1.4467400332225914e-07, "loss": 0.1604, "step": 4988 }, { "epoch": 1.04, "learning_rate": 1.4464285714285713e-07, "loss": 0.1499, "step": 4989 }, { "epoch": 1.04, "learning_rate": 1.4461171096345514e-07, "loss": 0.1283, "step": 4990 }, { "epoch": 1.04, "learning_rate": 1.4458056478405316e-07, "loss": 0.2568, "step": 4991 }, { "epoch": 1.04, "learning_rate": 1.4454941860465117e-07, "loss": 0.1886, "step": 4992 }, { "epoch": 1.04, "learning_rate": 1.4451827242524916e-07, "loss": 0.1827, "step": 4993 }, { "epoch": 1.04, "learning_rate": 1.4448712624584718e-07, "loss": 0.165, "step": 4994 }, { "epoch": 1.04, "learning_rate": 1.4445598006644517e-07, "loss": 0.1514, "step": 4995 }, { "epoch": 1.04, "learning_rate": 1.4442483388704318e-07, "loss": 0.1541, "step": 4996 }, { "epoch": 1.04, "learning_rate": 1.4439368770764117e-07, "loss": 0.1489, "step": 4997 }, { "epoch": 1.04, "learning_rate": 1.443625415282392e-07, "loss": 0.2009, "step": 4998 }, { "epoch": 1.04, "learning_rate": 1.443313953488372e-07, "loss": 0.1928, "step": 4999 }, { "epoch": 1.04, "learning_rate": 1.443002491694352e-07, "loss": 0.2234, "step": 5000 }, { "epoch": 1.04, "learning_rate": 1.442691029900332e-07, "loss": 0.2199, "step": 5001 }, { "epoch": 1.04, "learning_rate": 1.4423795681063123e-07, "loss": 0.193, "step": 5002 }, { "epoch": 1.04, "learning_rate": 1.4420681063122924e-07, "loss": 0.2105, "step": 5003 }, { "epoch": 1.04, "learning_rate": 1.4417566445182723e-07, "loss": 0.14, "step": 5004 }, { "epoch": 1.04, "learning_rate": 1.4414451827242525e-07, "loss": 0.2325, "step": 5005 }, { "epoch": 1.04, "learning_rate": 1.4411337209302326e-07, "loss": 0.2534, "step": 5006 }, { "epoch": 1.04, "learning_rate": 1.4408222591362125e-07, "loss": 0.2045, "step": 5007 }, { "epoch": 1.04, "learning_rate": 1.4405107973421927e-07, "loss": 0.1676, "step": 5008 }, { "epoch": 1.04, "learning_rate": 1.4401993355481726e-07, "loss": 0.2257, "step": 5009 }, { "epoch": 1.04, "learning_rate": 1.4398878737541527e-07, "loss": 0.182, "step": 5010 }, { "epoch": 1.04, "learning_rate": 1.439576411960133e-07, "loss": 0.2221, "step": 5011 }, { "epoch": 1.04, "learning_rate": 1.4392649501661128e-07, "loss": 0.1438, "step": 5012 }, { "epoch": 1.04, "learning_rate": 1.438953488372093e-07, "loss": 0.2234, "step": 5013 }, { "epoch": 1.04, "learning_rate": 1.438642026578073e-07, "loss": 0.1959, "step": 5014 }, { "epoch": 1.04, "learning_rate": 1.438330564784053e-07, "loss": 0.2248, "step": 5015 }, { "epoch": 1.04, "learning_rate": 1.4380191029900332e-07, "loss": 0.1998, "step": 5016 }, { "epoch": 1.04, "learning_rate": 1.4377076411960133e-07, "loss": 0.1728, "step": 5017 }, { "epoch": 1.04, "learning_rate": 1.4373961794019935e-07, "loss": 0.1936, "step": 5018 }, { "epoch": 1.04, "learning_rate": 1.4370847176079734e-07, "loss": 0.2054, "step": 5019 }, { "epoch": 1.04, "learning_rate": 1.4367732558139535e-07, "loss": 0.1863, "step": 5020 }, { "epoch": 1.04, "learning_rate": 1.4364617940199334e-07, "loss": 0.1766, "step": 5021 }, { "epoch": 1.04, "learning_rate": 1.4361503322259136e-07, "loss": 0.1905, "step": 5022 }, { "epoch": 1.04, "learning_rate": 1.4358388704318935e-07, "loss": 0.1992, "step": 5023 }, { "epoch": 1.04, "learning_rate": 1.4355274086378736e-07, "loss": 0.1978, "step": 5024 }, { "epoch": 1.04, "learning_rate": 1.4352159468438538e-07, "loss": 0.1496, "step": 5025 }, { "epoch": 1.04, "learning_rate": 1.4349044850498337e-07, "loss": 0.197, "step": 5026 }, { "epoch": 1.04, "learning_rate": 1.4345930232558138e-07, "loss": 0.2047, "step": 5027 }, { "epoch": 1.04, "learning_rate": 1.434281561461794e-07, "loss": 0.1985, "step": 5028 }, { "epoch": 1.04, "learning_rate": 1.4339700996677742e-07, "loss": 0.1778, "step": 5029 }, { "epoch": 1.04, "learning_rate": 1.433658637873754e-07, "loss": 0.1774, "step": 5030 }, { "epoch": 1.04, "learning_rate": 1.4333471760797342e-07, "loss": 0.1942, "step": 5031 }, { "epoch": 1.04, "learning_rate": 1.4330357142857144e-07, "loss": 0.222, "step": 5032 }, { "epoch": 1.05, "learning_rate": 1.4327242524916943e-07, "loss": 0.1824, "step": 5033 }, { "epoch": 1.05, "learning_rate": 1.4324127906976744e-07, "loss": 0.1765, "step": 5034 }, { "epoch": 1.05, "learning_rate": 1.4321013289036543e-07, "loss": 0.2041, "step": 5035 }, { "epoch": 1.05, "learning_rate": 1.4317898671096345e-07, "loss": 0.1931, "step": 5036 }, { "epoch": 1.05, "learning_rate": 1.4314784053156144e-07, "loss": 0.1834, "step": 5037 }, { "epoch": 1.05, "learning_rate": 1.4311669435215945e-07, "loss": 0.2141, "step": 5038 }, { "epoch": 1.05, "learning_rate": 1.4308554817275747e-07, "loss": 0.136, "step": 5039 }, { "epoch": 1.05, "learning_rate": 1.4305440199335548e-07, "loss": 0.2175, "step": 5040 }, { "epoch": 1.05, "learning_rate": 1.4302325581395347e-07, "loss": 0.184, "step": 5041 }, { "epoch": 1.05, "learning_rate": 1.429921096345515e-07, "loss": 0.1775, "step": 5042 }, { "epoch": 1.05, "learning_rate": 1.429609634551495e-07, "loss": 0.22, "step": 5043 }, { "epoch": 1.05, "learning_rate": 1.429298172757475e-07, "loss": 0.1991, "step": 5044 }, { "epoch": 1.05, "learning_rate": 1.428986710963455e-07, "loss": 0.1645, "step": 5045 }, { "epoch": 1.05, "learning_rate": 1.4286752491694353e-07, "loss": 0.1393, "step": 5046 }, { "epoch": 1.05, "learning_rate": 1.4283637873754152e-07, "loss": 0.2175, "step": 5047 }, { "epoch": 1.05, "learning_rate": 1.4280523255813953e-07, "loss": 0.2091, "step": 5048 }, { "epoch": 1.05, "learning_rate": 1.4277408637873752e-07, "loss": 0.2368, "step": 5049 }, { "epoch": 1.05, "learning_rate": 1.4274294019933554e-07, "loss": 0.1623, "step": 5050 }, { "epoch": 1.05, "learning_rate": 1.4271179401993355e-07, "loss": 0.2048, "step": 5051 }, { "epoch": 1.05, "learning_rate": 1.4268064784053154e-07, "loss": 0.1815, "step": 5052 }, { "epoch": 1.05, "learning_rate": 1.4264950166112956e-07, "loss": 0.1833, "step": 5053 }, { "epoch": 1.05, "learning_rate": 1.4261835548172757e-07, "loss": 0.179, "step": 5054 }, { "epoch": 1.05, "learning_rate": 1.425872093023256e-07, "loss": 0.1927, "step": 5055 }, { "epoch": 1.05, "learning_rate": 1.4255606312292358e-07, "loss": 0.1571, "step": 5056 }, { "epoch": 1.05, "learning_rate": 1.425249169435216e-07, "loss": 0.2038, "step": 5057 }, { "epoch": 1.05, "learning_rate": 1.424937707641196e-07, "loss": 0.2095, "step": 5058 }, { "epoch": 1.05, "learning_rate": 1.424626245847176e-07, "loss": 0.1784, "step": 5059 }, { "epoch": 1.05, "learning_rate": 1.4243147840531562e-07, "loss": 0.2062, "step": 5060 }, { "epoch": 1.05, "learning_rate": 1.424003322259136e-07, "loss": 0.1529, "step": 5061 }, { "epoch": 1.05, "learning_rate": 1.4236918604651162e-07, "loss": 0.1934, "step": 5062 }, { "epoch": 1.05, "learning_rate": 1.423380398671096e-07, "loss": 0.1339, "step": 5063 }, { "epoch": 1.05, "learning_rate": 1.4230689368770763e-07, "loss": 0.2277, "step": 5064 }, { "epoch": 1.05, "learning_rate": 1.4227574750830564e-07, "loss": 0.216, "step": 5065 }, { "epoch": 1.05, "learning_rate": 1.4224460132890366e-07, "loss": 0.1139, "step": 5066 }, { "epoch": 1.05, "learning_rate": 1.4221345514950167e-07, "loss": 0.1465, "step": 5067 }, { "epoch": 1.05, "learning_rate": 1.4218230897009966e-07, "loss": 0.1809, "step": 5068 }, { "epoch": 1.05, "learning_rate": 1.4215116279069768e-07, "loss": 0.1586, "step": 5069 }, { "epoch": 1.05, "learning_rate": 1.4212001661129567e-07, "loss": 0.1846, "step": 5070 }, { "epoch": 1.05, "learning_rate": 1.4208887043189368e-07, "loss": 0.1382, "step": 5071 }, { "epoch": 1.05, "learning_rate": 1.420577242524917e-07, "loss": 0.1928, "step": 5072 }, { "epoch": 1.05, "learning_rate": 1.420265780730897e-07, "loss": 0.1463, "step": 5073 }, { "epoch": 1.05, "learning_rate": 1.419954318936877e-07, "loss": 0.1443, "step": 5074 }, { "epoch": 1.05, "learning_rate": 1.419642857142857e-07, "loss": 0.1498, "step": 5075 }, { "epoch": 1.05, "learning_rate": 1.419331395348837e-07, "loss": 0.1525, "step": 5076 }, { "epoch": 1.05, "learning_rate": 1.4190199335548173e-07, "loss": 0.1656, "step": 5077 }, { "epoch": 1.05, "learning_rate": 1.4187084717607971e-07, "loss": 0.1775, "step": 5078 }, { "epoch": 1.05, "learning_rate": 1.4183970099667773e-07, "loss": 0.1495, "step": 5079 }, { "epoch": 1.05, "learning_rate": 1.4180855481727575e-07, "loss": 0.1777, "step": 5080 }, { "epoch": 1.06, "learning_rate": 1.4177740863787376e-07, "loss": 0.1739, "step": 5081 }, { "epoch": 1.06, "learning_rate": 1.4174626245847175e-07, "loss": 0.1657, "step": 5082 }, { "epoch": 1.06, "learning_rate": 1.4171511627906977e-07, "loss": 0.1353, "step": 5083 }, { "epoch": 1.06, "learning_rate": 1.4168397009966776e-07, "loss": 0.2163, "step": 5084 }, { "epoch": 1.06, "learning_rate": 1.4165282392026577e-07, "loss": 0.2248, "step": 5085 }, { "epoch": 1.06, "learning_rate": 1.416216777408638e-07, "loss": 0.2051, "step": 5086 }, { "epoch": 1.06, "learning_rate": 1.4159053156146178e-07, "loss": 0.1422, "step": 5087 }, { "epoch": 1.06, "learning_rate": 1.415593853820598e-07, "loss": 0.24, "step": 5088 }, { "epoch": 1.06, "learning_rate": 1.4152823920265778e-07, "loss": 0.1413, "step": 5089 }, { "epoch": 1.06, "learning_rate": 1.414970930232558e-07, "loss": 0.1903, "step": 5090 }, { "epoch": 1.06, "learning_rate": 1.4146594684385381e-07, "loss": 0.1748, "step": 5091 }, { "epoch": 1.06, "learning_rate": 1.4143480066445183e-07, "loss": 0.2361, "step": 5092 }, { "epoch": 1.06, "learning_rate": 1.4140365448504985e-07, "loss": 0.1943, "step": 5093 }, { "epoch": 1.06, "learning_rate": 1.4137250830564784e-07, "loss": 0.2298, "step": 5094 }, { "epoch": 1.06, "learning_rate": 1.4134136212624585e-07, "loss": 0.2089, "step": 5095 }, { "epoch": 1.06, "learning_rate": 1.4131021594684384e-07, "loss": 0.1539, "step": 5096 }, { "epoch": 1.06, "learning_rate": 1.4127906976744186e-07, "loss": 0.1313, "step": 5097 }, { "epoch": 1.06, "learning_rate": 1.4124792358803987e-07, "loss": 0.2119, "step": 5098 }, { "epoch": 1.06, "learning_rate": 1.4121677740863786e-07, "loss": 0.2211, "step": 5099 }, { "epoch": 1.06, "learning_rate": 1.4118563122923588e-07, "loss": 0.2161, "step": 5100 }, { "epoch": 1.06, "learning_rate": 1.4115448504983387e-07, "loss": 0.1754, "step": 5101 }, { "epoch": 1.06, "learning_rate": 1.4112333887043188e-07, "loss": 0.2266, "step": 5102 }, { "epoch": 1.06, "learning_rate": 1.410921926910299e-07, "loss": 0.2178, "step": 5103 }, { "epoch": 1.06, "learning_rate": 1.4106104651162791e-07, "loss": 0.2084, "step": 5104 }, { "epoch": 1.06, "learning_rate": 1.410299003322259e-07, "loss": 0.2062, "step": 5105 }, { "epoch": 1.06, "learning_rate": 1.4099875415282392e-07, "loss": 0.1825, "step": 5106 }, { "epoch": 1.06, "learning_rate": 1.4096760797342194e-07, "loss": 0.184, "step": 5107 }, { "epoch": 1.06, "learning_rate": 1.4093646179401992e-07, "loss": 0.174, "step": 5108 }, { "epoch": 1.06, "learning_rate": 1.4090531561461794e-07, "loss": 0.1582, "step": 5109 }, { "epoch": 1.06, "learning_rate": 1.4087416943521593e-07, "loss": 0.147, "step": 5110 }, { "epoch": 1.06, "learning_rate": 1.4084302325581395e-07, "loss": 0.1751, "step": 5111 }, { "epoch": 1.06, "learning_rate": 1.4081187707641196e-07, "loss": 0.1887, "step": 5112 }, { "epoch": 1.06, "learning_rate": 1.4078073089700995e-07, "loss": 0.205, "step": 5113 }, { "epoch": 1.06, "learning_rate": 1.4074958471760797e-07, "loss": 0.1899, "step": 5114 }, { "epoch": 1.06, "learning_rate": 1.4071843853820596e-07, "loss": 0.1796, "step": 5115 }, { "epoch": 1.06, "learning_rate": 1.4068729235880397e-07, "loss": 0.2039, "step": 5116 }, { "epoch": 1.06, "learning_rate": 1.40656146179402e-07, "loss": 0.144, "step": 5117 }, { "epoch": 1.06, "learning_rate": 1.40625e-07, "loss": 0.2316, "step": 5118 }, { "epoch": 1.06, "learning_rate": 1.4059385382059802e-07, "loss": 0.173, "step": 5119 }, { "epoch": 1.06, "learning_rate": 1.40562707641196e-07, "loss": 0.1627, "step": 5120 }, { "epoch": 1.06, "learning_rate": 1.4053156146179402e-07, "loss": 0.1431, "step": 5121 }, { "epoch": 1.06, "learning_rate": 1.4050041528239201e-07, "loss": 0.2357, "step": 5122 }, { "epoch": 1.06, "learning_rate": 1.4046926910299003e-07, "loss": 0.1884, "step": 5123 }, { "epoch": 1.06, "learning_rate": 1.4043812292358802e-07, "loss": 0.1724, "step": 5124 }, { "epoch": 1.06, "learning_rate": 1.4040697674418603e-07, "loss": 0.2552, "step": 5125 }, { "epoch": 1.06, "learning_rate": 1.4037583056478405e-07, "loss": 0.2471, "step": 5126 }, { "epoch": 1.06, "learning_rate": 1.4034468438538204e-07, "loss": 0.1508, "step": 5127 }, { "epoch": 1.06, "learning_rate": 1.4031353820598006e-07, "loss": 0.1552, "step": 5128 }, { "epoch": 1.06, "learning_rate": 1.4028239202657807e-07, "loss": 0.1421, "step": 5129 }, { "epoch": 1.07, "learning_rate": 1.402512458471761e-07, "loss": 0.1415, "step": 5130 }, { "epoch": 1.07, "learning_rate": 1.4022009966777408e-07, "loss": 0.2078, "step": 5131 }, { "epoch": 1.07, "learning_rate": 1.401889534883721e-07, "loss": 0.1723, "step": 5132 }, { "epoch": 1.07, "learning_rate": 1.401578073089701e-07, "loss": 0.1423, "step": 5133 }, { "epoch": 1.07, "learning_rate": 1.401266611295681e-07, "loss": 0.1309, "step": 5134 }, { "epoch": 1.07, "learning_rate": 1.4009551495016611e-07, "loss": 0.213, "step": 5135 }, { "epoch": 1.07, "learning_rate": 1.400643687707641e-07, "loss": 0.1295, "step": 5136 }, { "epoch": 1.07, "learning_rate": 1.4003322259136212e-07, "loss": 0.1645, "step": 5137 }, { "epoch": 1.07, "learning_rate": 1.400020764119601e-07, "loss": 0.1587, "step": 5138 }, { "epoch": 1.07, "learning_rate": 1.3997093023255812e-07, "loss": 0.2198, "step": 5139 }, { "epoch": 1.07, "learning_rate": 1.3993978405315614e-07, "loss": 0.1466, "step": 5140 }, { "epoch": 1.07, "learning_rate": 1.3990863787375416e-07, "loss": 0.1715, "step": 5141 }, { "epoch": 1.07, "learning_rate": 1.3987749169435215e-07, "loss": 0.1493, "step": 5142 }, { "epoch": 1.07, "learning_rate": 1.3984634551495016e-07, "loss": 0.1425, "step": 5143 }, { "epoch": 1.07, "learning_rate": 1.3981519933554818e-07, "loss": 0.1978, "step": 5144 }, { "epoch": 1.07, "learning_rate": 1.397840531561462e-07, "loss": 0.1917, "step": 5145 }, { "epoch": 1.07, "learning_rate": 1.3975290697674418e-07, "loss": 0.2319, "step": 5146 }, { "epoch": 1.07, "learning_rate": 1.397217607973422e-07, "loss": 0.2118, "step": 5147 }, { "epoch": 1.07, "learning_rate": 1.396906146179402e-07, "loss": 0.1949, "step": 5148 }, { "epoch": 1.07, "learning_rate": 1.396594684385382e-07, "loss": 0.1929, "step": 5149 }, { "epoch": 1.07, "learning_rate": 1.396283222591362e-07, "loss": 0.1571, "step": 5150 }, { "epoch": 1.07, "learning_rate": 1.395971760797342e-07, "loss": 0.1661, "step": 5151 }, { "epoch": 1.07, "learning_rate": 1.3956602990033222e-07, "loss": 0.2053, "step": 5152 }, { "epoch": 1.07, "learning_rate": 1.3953488372093021e-07, "loss": 0.1533, "step": 5153 }, { "epoch": 1.07, "learning_rate": 1.3950373754152823e-07, "loss": 0.2044, "step": 5154 }, { "epoch": 1.07, "learning_rate": 1.3947259136212624e-07, "loss": 0.1413, "step": 5155 }, { "epoch": 1.07, "learning_rate": 1.3944144518272426e-07, "loss": 0.2042, "step": 5156 }, { "epoch": 1.07, "learning_rate": 1.3941029900332225e-07, "loss": 0.1563, "step": 5157 }, { "epoch": 1.07, "learning_rate": 1.3937915282392027e-07, "loss": 0.1729, "step": 5158 }, { "epoch": 1.07, "learning_rate": 1.3934800664451828e-07, "loss": 0.204, "step": 5159 }, { "epoch": 1.07, "learning_rate": 1.3931686046511627e-07, "loss": 0.1854, "step": 5160 }, { "epoch": 1.07, "learning_rate": 1.392857142857143e-07, "loss": 0.1601, "step": 5161 }, { "epoch": 1.07, "learning_rate": 1.3925456810631228e-07, "loss": 0.1858, "step": 5162 }, { "epoch": 1.07, "learning_rate": 1.392234219269103e-07, "loss": 0.1705, "step": 5163 }, { "epoch": 1.07, "learning_rate": 1.3919227574750828e-07, "loss": 0.2434, "step": 5164 }, { "epoch": 1.07, "learning_rate": 1.391611295681063e-07, "loss": 0.2011, "step": 5165 }, { "epoch": 1.07, "learning_rate": 1.391299833887043e-07, "loss": 0.1285, "step": 5166 }, { "epoch": 1.07, "learning_rate": 1.3909883720930233e-07, "loss": 0.145, "step": 5167 }, { "epoch": 1.07, "learning_rate": 1.3906769102990032e-07, "loss": 0.1527, "step": 5168 }, { "epoch": 1.07, "learning_rate": 1.3903654485049833e-07, "loss": 0.1307, "step": 5169 }, { "epoch": 1.07, "learning_rate": 1.3900539867109635e-07, "loss": 0.1575, "step": 5170 }, { "epoch": 1.07, "learning_rate": 1.3897425249169434e-07, "loss": 0.2232, "step": 5171 }, { "epoch": 1.07, "learning_rate": 1.3894310631229236e-07, "loss": 0.1641, "step": 5172 }, { "epoch": 1.07, "learning_rate": 1.3891196013289037e-07, "loss": 0.1493, "step": 5173 }, { "epoch": 1.07, "learning_rate": 1.3888081395348836e-07, "loss": 0.1387, "step": 5174 }, { "epoch": 1.07, "learning_rate": 1.3884966777408638e-07, "loss": 0.2049, "step": 5175 }, { "epoch": 1.07, "learning_rate": 1.3881852159468437e-07, "loss": 0.1519, "step": 5176 }, { "epoch": 1.07, "learning_rate": 1.3878737541528238e-07, "loss": 0.1666, "step": 5177 }, { "epoch": 1.08, "learning_rate": 1.387562292358804e-07, "loss": 0.1914, "step": 5178 }, { "epoch": 1.08, "learning_rate": 1.3872508305647839e-07, "loss": 0.216, "step": 5179 }, { "epoch": 1.08, "learning_rate": 1.386939368770764e-07, "loss": 0.1611, "step": 5180 }, { "epoch": 1.08, "learning_rate": 1.3866279069767442e-07, "loss": 0.1694, "step": 5181 }, { "epoch": 1.08, "learning_rate": 1.3863164451827243e-07, "loss": 0.199, "step": 5182 }, { "epoch": 1.08, "learning_rate": 1.3860049833887042e-07, "loss": 0.1485, "step": 5183 }, { "epoch": 1.08, "learning_rate": 1.3856935215946844e-07, "loss": 0.2047, "step": 5184 }, { "epoch": 1.08, "learning_rate": 1.3853820598006646e-07, "loss": 0.1853, "step": 5185 }, { "epoch": 1.08, "learning_rate": 1.3850705980066444e-07, "loss": 0.1358, "step": 5186 }, { "epoch": 1.08, "learning_rate": 1.3847591362126246e-07, "loss": 0.1512, "step": 5187 }, { "epoch": 1.08, "learning_rate": 1.3844476744186045e-07, "loss": 0.1449, "step": 5188 }, { "epoch": 1.08, "learning_rate": 1.3841362126245847e-07, "loss": 0.175, "step": 5189 }, { "epoch": 1.08, "learning_rate": 1.3838247508305645e-07, "loss": 0.2119, "step": 5190 }, { "epoch": 1.08, "learning_rate": 1.3835132890365447e-07, "loss": 0.1815, "step": 5191 }, { "epoch": 1.08, "learning_rate": 1.3832018272425249e-07, "loss": 0.19, "step": 5192 }, { "epoch": 1.08, "learning_rate": 1.382890365448505e-07, "loss": 0.1991, "step": 5193 }, { "epoch": 1.08, "learning_rate": 1.382578903654485e-07, "loss": 0.1603, "step": 5194 }, { "epoch": 1.08, "learning_rate": 1.382267441860465e-07, "loss": 0.1591, "step": 5195 }, { "epoch": 1.08, "learning_rate": 1.3819559800664452e-07, "loss": 0.1558, "step": 5196 }, { "epoch": 1.08, "learning_rate": 1.381644518272425e-07, "loss": 0.1511, "step": 5197 }, { "epoch": 1.08, "learning_rate": 1.3813330564784053e-07, "loss": 0.1497, "step": 5198 }, { "epoch": 1.08, "learning_rate": 1.3810215946843854e-07, "loss": 0.1754, "step": 5199 }, { "epoch": 1.08, "learning_rate": 1.3807101328903653e-07, "loss": 0.1504, "step": 5200 }, { "epoch": 1.08, "learning_rate": 1.3803986710963455e-07, "loss": 0.1658, "step": 5201 }, { "epoch": 1.08, "learning_rate": 1.3800872093023254e-07, "loss": 0.1824, "step": 5202 }, { "epoch": 1.08, "learning_rate": 1.3797757475083055e-07, "loss": 0.1538, "step": 5203 }, { "epoch": 1.08, "learning_rate": 1.3794642857142857e-07, "loss": 0.1896, "step": 5204 }, { "epoch": 1.08, "learning_rate": 1.3791528239202656e-07, "loss": 0.1722, "step": 5205 }, { "epoch": 1.08, "learning_rate": 1.3788413621262458e-07, "loss": 0.226, "step": 5206 }, { "epoch": 1.08, "learning_rate": 1.378529900332226e-07, "loss": 0.1647, "step": 5207 }, { "epoch": 1.08, "learning_rate": 1.378218438538206e-07, "loss": 0.172, "step": 5208 }, { "epoch": 1.08, "learning_rate": 1.377906976744186e-07, "loss": 0.1378, "step": 5209 }, { "epoch": 1.08, "learning_rate": 1.377595514950166e-07, "loss": 0.2325, "step": 5210 }, { "epoch": 1.08, "learning_rate": 1.377284053156146e-07, "loss": 0.1976, "step": 5211 }, { "epoch": 1.08, "learning_rate": 1.3769725913621262e-07, "loss": 0.1899, "step": 5212 }, { "epoch": 1.08, "learning_rate": 1.3766611295681063e-07, "loss": 0.1657, "step": 5213 }, { "epoch": 1.08, "learning_rate": 1.3763496677740862e-07, "loss": 0.2172, "step": 5214 }, { "epoch": 1.08, "learning_rate": 1.3760382059800664e-07, "loss": 0.202, "step": 5215 }, { "epoch": 1.08, "learning_rate": 1.3757267441860463e-07, "loss": 0.1433, "step": 5216 }, { "epoch": 1.08, "learning_rate": 1.3754152823920264e-07, "loss": 0.1318, "step": 5217 }, { "epoch": 1.08, "learning_rate": 1.3751038205980066e-07, "loss": 0.1931, "step": 5218 }, { "epoch": 1.08, "learning_rate": 1.3747923588039868e-07, "loss": 0.1764, "step": 5219 }, { "epoch": 1.08, "learning_rate": 1.3744808970099666e-07, "loss": 0.1883, "step": 5220 }, { "epoch": 1.08, "learning_rate": 1.3741694352159468e-07, "loss": 0.1723, "step": 5221 }, { "epoch": 1.08, "learning_rate": 1.373857973421927e-07, "loss": 0.1743, "step": 5222 }, { "epoch": 1.08, "learning_rate": 1.3735465116279069e-07, "loss": 0.2826, "step": 5223 }, { "epoch": 1.08, "learning_rate": 1.373235049833887e-07, "loss": 0.2115, "step": 5224 }, { "epoch": 1.08, "learning_rate": 1.372923588039867e-07, "loss": 0.1865, "step": 5225 }, { "epoch": 1.09, "learning_rate": 1.372612126245847e-07, "loss": 0.2618, "step": 5226 }, { "epoch": 1.09, "learning_rate": 1.3723006644518272e-07, "loss": 0.2008, "step": 5227 }, { "epoch": 1.09, "learning_rate": 1.371989202657807e-07, "loss": 0.1362, "step": 5228 }, { "epoch": 1.09, "learning_rate": 1.3716777408637873e-07, "loss": 0.2278, "step": 5229 }, { "epoch": 1.09, "learning_rate": 1.3713662790697674e-07, "loss": 0.2102, "step": 5230 }, { "epoch": 1.09, "learning_rate": 1.3710548172757473e-07, "loss": 0.1925, "step": 5231 }, { "epoch": 1.09, "learning_rate": 1.3707433554817275e-07, "loss": 0.1796, "step": 5232 }, { "epoch": 1.09, "learning_rate": 1.3704318936877076e-07, "loss": 0.1581, "step": 5233 }, { "epoch": 1.09, "learning_rate": 1.3701204318936878e-07, "loss": 0.1622, "step": 5234 }, { "epoch": 1.09, "learning_rate": 1.3698089700996677e-07, "loss": 0.162, "step": 5235 }, { "epoch": 1.09, "learning_rate": 1.3694975083056479e-07, "loss": 0.1737, "step": 5236 }, { "epoch": 1.09, "learning_rate": 1.3691860465116277e-07, "loss": 0.1701, "step": 5237 }, { "epoch": 1.09, "learning_rate": 1.368874584717608e-07, "loss": 0.1705, "step": 5238 }, { "epoch": 1.09, "learning_rate": 1.368563122923588e-07, "loss": 0.1797, "step": 5239 }, { "epoch": 1.09, "learning_rate": 1.368251661129568e-07, "loss": 0.1852, "step": 5240 }, { "epoch": 1.09, "learning_rate": 1.367940199335548e-07, "loss": 0.1312, "step": 5241 }, { "epoch": 1.09, "learning_rate": 1.367628737541528e-07, "loss": 0.2034, "step": 5242 }, { "epoch": 1.09, "learning_rate": 1.3673172757475082e-07, "loss": 0.2046, "step": 5243 }, { "epoch": 1.09, "learning_rate": 1.3670058139534883e-07, "loss": 0.2157, "step": 5244 }, { "epoch": 1.09, "learning_rate": 1.3666943521594685e-07, "loss": 0.2218, "step": 5245 }, { "epoch": 1.09, "learning_rate": 1.3663828903654486e-07, "loss": 0.2367, "step": 5246 }, { "epoch": 1.09, "learning_rate": 1.3660714285714285e-07, "loss": 0.1765, "step": 5247 }, { "epoch": 1.09, "learning_rate": 1.3657599667774087e-07, "loss": 0.1864, "step": 5248 }, { "epoch": 1.09, "learning_rate": 1.3654485049833886e-07, "loss": 0.1568, "step": 5249 }, { "epoch": 1.09, "learning_rate": 1.3651370431893687e-07, "loss": 0.1932, "step": 5250 }, { "epoch": 1.09, "learning_rate": 1.3648255813953486e-07, "loss": 0.1679, "step": 5251 }, { "epoch": 1.09, "learning_rate": 1.3645141196013288e-07, "loss": 0.1725, "step": 5252 }, { "epoch": 1.09, "learning_rate": 1.364202657807309e-07, "loss": 0.1616, "step": 5253 }, { "epoch": 1.09, "learning_rate": 1.3638911960132889e-07, "loss": 0.1611, "step": 5254 }, { "epoch": 1.09, "learning_rate": 1.363579734219269e-07, "loss": 0.1679, "step": 5255 }, { "epoch": 1.09, "learning_rate": 1.3632682724252492e-07, "loss": 0.1481, "step": 5256 }, { "epoch": 1.09, "learning_rate": 1.362956810631229e-07, "loss": 0.2309, "step": 5257 }, { "epoch": 1.09, "learning_rate": 1.3626453488372092e-07, "loss": 0.152, "step": 5258 }, { "epoch": 1.09, "learning_rate": 1.3623338870431894e-07, "loss": 0.1763, "step": 5259 }, { "epoch": 1.09, "learning_rate": 1.3620224252491695e-07, "loss": 0.1365, "step": 5260 }, { "epoch": 1.09, "learning_rate": 1.3617109634551494e-07, "loss": 0.1864, "step": 5261 }, { "epoch": 1.09, "learning_rate": 1.3613995016611296e-07, "loss": 0.1838, "step": 5262 }, { "epoch": 1.09, "learning_rate": 1.3610880398671095e-07, "loss": 0.2006, "step": 5263 }, { "epoch": 1.09, "learning_rate": 1.3607765780730896e-07, "loss": 0.1903, "step": 5264 }, { "epoch": 1.09, "learning_rate": 1.3604651162790695e-07, "loss": 0.2013, "step": 5265 }, { "epoch": 1.09, "learning_rate": 1.3601536544850497e-07, "loss": 0.1328, "step": 5266 }, { "epoch": 1.09, "learning_rate": 1.3598421926910299e-07, "loss": 0.1487, "step": 5267 }, { "epoch": 1.09, "learning_rate": 1.3595307308970097e-07, "loss": 0.1673, "step": 5268 }, { "epoch": 1.09, "learning_rate": 1.35921926910299e-07, "loss": 0.189, "step": 5269 }, { "epoch": 1.09, "learning_rate": 1.35890780730897e-07, "loss": 0.1877, "step": 5270 }, { "epoch": 1.09, "learning_rate": 1.3585963455149502e-07, "loss": 0.1606, "step": 5271 }, { "epoch": 1.09, "learning_rate": 1.3582848837209304e-07, "loss": 0.1477, "step": 5272 }, { "epoch": 1.09, "learning_rate": 1.3579734219269103e-07, "loss": 0.159, "step": 5273 }, { "epoch": 1.1, "learning_rate": 1.3576619601328904e-07, "loss": 0.1543, "step": 5274 }, { "epoch": 1.1, "learning_rate": 1.3573504983388703e-07, "loss": 0.2157, "step": 5275 }, { "epoch": 1.1, "learning_rate": 1.3570390365448505e-07, "loss": 0.1776, "step": 5276 }, { "epoch": 1.1, "learning_rate": 1.3567275747508304e-07, "loss": 0.186, "step": 5277 }, { "epoch": 1.1, "learning_rate": 1.3564161129568105e-07, "loss": 0.1464, "step": 5278 }, { "epoch": 1.1, "learning_rate": 1.3561046511627907e-07, "loss": 0.2417, "step": 5279 }, { "epoch": 1.1, "learning_rate": 1.3557931893687706e-07, "loss": 0.2534, "step": 5280 }, { "epoch": 1.1, "learning_rate": 1.3554817275747507e-07, "loss": 0.1532, "step": 5281 }, { "epoch": 1.1, "learning_rate": 1.355170265780731e-07, "loss": 0.1331, "step": 5282 }, { "epoch": 1.1, "learning_rate": 1.354858803986711e-07, "loss": 0.1908, "step": 5283 }, { "epoch": 1.1, "learning_rate": 1.354547342192691e-07, "loss": 0.2029, "step": 5284 }, { "epoch": 1.1, "learning_rate": 1.354235880398671e-07, "loss": 0.1807, "step": 5285 }, { "epoch": 1.1, "learning_rate": 1.3539244186046513e-07, "loss": 0.2796, "step": 5286 }, { "epoch": 1.1, "learning_rate": 1.3536129568106312e-07, "loss": 0.1788, "step": 5287 }, { "epoch": 1.1, "learning_rate": 1.3533014950166113e-07, "loss": 0.1685, "step": 5288 }, { "epoch": 1.1, "learning_rate": 1.3529900332225912e-07, "loss": 0.1722, "step": 5289 }, { "epoch": 1.1, "learning_rate": 1.3526785714285714e-07, "loss": 0.1725, "step": 5290 }, { "epoch": 1.1, "learning_rate": 1.3523671096345513e-07, "loss": 0.2111, "step": 5291 }, { "epoch": 1.1, "learning_rate": 1.3520556478405314e-07, "loss": 0.2321, "step": 5292 }, { "epoch": 1.1, "learning_rate": 1.3517441860465116e-07, "loss": 0.1413, "step": 5293 }, { "epoch": 1.1, "learning_rate": 1.3514327242524915e-07, "loss": 0.1888, "step": 5294 }, { "epoch": 1.1, "learning_rate": 1.3511212624584716e-07, "loss": 0.1421, "step": 5295 }, { "epoch": 1.1, "learning_rate": 1.3508098006644518e-07, "loss": 0.1411, "step": 5296 }, { "epoch": 1.1, "learning_rate": 1.350498338870432e-07, "loss": 0.1156, "step": 5297 }, { "epoch": 1.1, "learning_rate": 1.3501868770764118e-07, "loss": 0.125, "step": 5298 }, { "epoch": 1.1, "learning_rate": 1.349875415282392e-07, "loss": 0.1756, "step": 5299 }, { "epoch": 1.1, "learning_rate": 1.3495639534883722e-07, "loss": 0.1465, "step": 5300 }, { "epoch": 1.1, "learning_rate": 1.349252491694352e-07, "loss": 0.2137, "step": 5301 }, { "epoch": 1.1, "learning_rate": 1.3489410299003322e-07, "loss": 0.1636, "step": 5302 }, { "epoch": 1.1, "learning_rate": 1.348629568106312e-07, "loss": 0.2193, "step": 5303 }, { "epoch": 1.1, "learning_rate": 1.3483181063122923e-07, "loss": 0.2128, "step": 5304 }, { "epoch": 1.1, "learning_rate": 1.3480066445182722e-07, "loss": 0.21, "step": 5305 }, { "epoch": 1.1, "learning_rate": 1.3476951827242523e-07, "loss": 0.176, "step": 5306 }, { "epoch": 1.1, "learning_rate": 1.3473837209302325e-07, "loss": 0.1396, "step": 5307 }, { "epoch": 1.1, "learning_rate": 1.3470722591362126e-07, "loss": 0.1688, "step": 5308 }, { "epoch": 1.1, "learning_rate": 1.3467607973421928e-07, "loss": 0.2193, "step": 5309 }, { "epoch": 1.1, "learning_rate": 1.3464493355481727e-07, "loss": 0.2047, "step": 5310 }, { "epoch": 1.1, "learning_rate": 1.3461378737541528e-07, "loss": 0.2411, "step": 5311 }, { "epoch": 1.1, "learning_rate": 1.345826411960133e-07, "loss": 0.2328, "step": 5312 }, { "epoch": 1.1, "learning_rate": 1.345514950166113e-07, "loss": 0.1762, "step": 5313 }, { "epoch": 1.1, "learning_rate": 1.345203488372093e-07, "loss": 0.1432, "step": 5314 }, { "epoch": 1.1, "learning_rate": 1.344892026578073e-07, "loss": 0.2182, "step": 5315 }, { "epoch": 1.1, "learning_rate": 1.344580564784053e-07, "loss": 0.1399, "step": 5316 }, { "epoch": 1.1, "learning_rate": 1.344269102990033e-07, "loss": 0.2003, "step": 5317 }, { "epoch": 1.1, "learning_rate": 1.3439576411960132e-07, "loss": 0.2544, "step": 5318 }, { "epoch": 1.1, "learning_rate": 1.3436461794019933e-07, "loss": 0.1718, "step": 5319 }, { "epoch": 1.1, "learning_rate": 1.3433347176079735e-07, "loss": 0.1524, "step": 5320 }, { "epoch": 1.1, "learning_rate": 1.3430232558139534e-07, "loss": 0.1646, "step": 5321 }, { "epoch": 1.11, "learning_rate": 1.3427117940199335e-07, "loss": 0.1848, "step": 5322 }, { "epoch": 1.11, "learning_rate": 1.3424003322259137e-07, "loss": 0.2306, "step": 5323 }, { "epoch": 1.11, "learning_rate": 1.3420888704318936e-07, "loss": 0.1828, "step": 5324 }, { "epoch": 1.11, "learning_rate": 1.3417774086378737e-07, "loss": 0.2296, "step": 5325 }, { "epoch": 1.11, "learning_rate": 1.341465946843854e-07, "loss": 0.1611, "step": 5326 }, { "epoch": 1.11, "learning_rate": 1.3411544850498338e-07, "loss": 0.1715, "step": 5327 }, { "epoch": 1.11, "learning_rate": 1.340843023255814e-07, "loss": 0.2157, "step": 5328 }, { "epoch": 1.11, "learning_rate": 1.3405315614617938e-07, "loss": 0.175, "step": 5329 }, { "epoch": 1.11, "learning_rate": 1.340220099667774e-07, "loss": 0.1843, "step": 5330 }, { "epoch": 1.11, "learning_rate": 1.339908637873754e-07, "loss": 0.1858, "step": 5331 }, { "epoch": 1.11, "learning_rate": 1.339597176079734e-07, "loss": 0.2144, "step": 5332 }, { "epoch": 1.11, "learning_rate": 1.3392857142857142e-07, "loss": 0.1385, "step": 5333 }, { "epoch": 1.11, "learning_rate": 1.3389742524916944e-07, "loss": 0.1681, "step": 5334 }, { "epoch": 1.11, "learning_rate": 1.3386627906976745e-07, "loss": 0.1628, "step": 5335 }, { "epoch": 1.11, "learning_rate": 1.3383513289036544e-07, "loss": 0.1695, "step": 5336 }, { "epoch": 1.11, "learning_rate": 1.3380398671096346e-07, "loss": 0.1727, "step": 5337 }, { "epoch": 1.11, "learning_rate": 1.3377284053156145e-07, "loss": 0.1883, "step": 5338 }, { "epoch": 1.11, "learning_rate": 1.3374169435215946e-07, "loss": 0.238, "step": 5339 }, { "epoch": 1.11, "learning_rate": 1.3371054817275748e-07, "loss": 0.2417, "step": 5340 }, { "epoch": 1.11, "learning_rate": 1.3367940199335547e-07, "loss": 0.1291, "step": 5341 }, { "epoch": 1.11, "learning_rate": 1.3364825581395348e-07, "loss": 0.2188, "step": 5342 }, { "epoch": 1.11, "learning_rate": 1.3361710963455147e-07, "loss": 0.1743, "step": 5343 }, { "epoch": 1.11, "learning_rate": 1.335859634551495e-07, "loss": 0.1662, "step": 5344 }, { "epoch": 1.11, "learning_rate": 1.335548172757475e-07, "loss": 0.1584, "step": 5345 }, { "epoch": 1.11, "learning_rate": 1.3352367109634552e-07, "loss": 0.1342, "step": 5346 }, { "epoch": 1.11, "learning_rate": 1.334925249169435e-07, "loss": 0.2372, "step": 5347 }, { "epoch": 1.11, "learning_rate": 1.3346137873754153e-07, "loss": 0.1406, "step": 5348 }, { "epoch": 1.11, "learning_rate": 1.3343023255813954e-07, "loss": 0.2062, "step": 5349 }, { "epoch": 1.11, "learning_rate": 1.3339908637873753e-07, "loss": 0.1661, "step": 5350 }, { "epoch": 1.11, "learning_rate": 1.3336794019933555e-07, "loss": 0.1758, "step": 5351 }, { "epoch": 1.11, "learning_rate": 1.3333679401993354e-07, "loss": 0.1212, "step": 5352 }, { "epoch": 1.11, "learning_rate": 1.3330564784053155e-07, "loss": 0.2303, "step": 5353 }, { "epoch": 1.11, "learning_rate": 1.3327450166112957e-07, "loss": 0.1604, "step": 5354 }, { "epoch": 1.11, "learning_rate": 1.3324335548172756e-07, "loss": 0.1176, "step": 5355 }, { "epoch": 1.11, "learning_rate": 1.3321220930232557e-07, "loss": 0.1465, "step": 5356 }, { "epoch": 1.11, "learning_rate": 1.331810631229236e-07, "loss": 0.2232, "step": 5357 }, { "epoch": 1.11, "learning_rate": 1.3314991694352158e-07, "loss": 0.209, "step": 5358 }, { "epoch": 1.11, "learning_rate": 1.331187707641196e-07, "loss": 0.1391, "step": 5359 }, { "epoch": 1.11, "learning_rate": 1.330876245847176e-07, "loss": 0.2633, "step": 5360 }, { "epoch": 1.11, "learning_rate": 1.3305647840531563e-07, "loss": 0.1931, "step": 5361 }, { "epoch": 1.11, "learning_rate": 1.3302533222591361e-07, "loss": 0.1645, "step": 5362 }, { "epoch": 1.11, "learning_rate": 1.3299418604651163e-07, "loss": 0.1438, "step": 5363 }, { "epoch": 1.11, "learning_rate": 1.3296303986710962e-07, "loss": 0.1989, "step": 5364 }, { "epoch": 1.11, "learning_rate": 1.3293189368770764e-07, "loss": 0.1632, "step": 5365 }, { "epoch": 1.11, "learning_rate": 1.3290074750830565e-07, "loss": 0.1961, "step": 5366 }, { "epoch": 1.11, "learning_rate": 1.3286960132890364e-07, "loss": 0.1423, "step": 5367 }, { "epoch": 1.11, "learning_rate": 1.3283845514950166e-07, "loss": 0.2857, "step": 5368 }, { "epoch": 1.11, "learning_rate": 1.3280730897009965e-07, "loss": 0.1825, "step": 5369 }, { "epoch": 1.12, "learning_rate": 1.3277616279069766e-07, "loss": 0.2039, "step": 5370 }, { "epoch": 1.12, "learning_rate": 1.3274501661129568e-07, "loss": 0.13, "step": 5371 }, { "epoch": 1.12, "learning_rate": 1.327138704318937e-07, "loss": 0.1601, "step": 5372 }, { "epoch": 1.12, "learning_rate": 1.3268272425249168e-07, "loss": 0.13, "step": 5373 }, { "epoch": 1.12, "learning_rate": 1.326515780730897e-07, "loss": 0.131, "step": 5374 }, { "epoch": 1.12, "learning_rate": 1.3262043189368771e-07, "loss": 0.159, "step": 5375 }, { "epoch": 1.12, "learning_rate": 1.325892857142857e-07, "loss": 0.1861, "step": 5376 }, { "epoch": 1.12, "learning_rate": 1.3255813953488372e-07, "loss": 0.1501, "step": 5377 }, { "epoch": 1.12, "learning_rate": 1.325269933554817e-07, "loss": 0.1676, "step": 5378 }, { "epoch": 1.12, "learning_rate": 1.3249584717607973e-07, "loss": 0.1798, "step": 5379 }, { "epoch": 1.12, "learning_rate": 1.3246470099667774e-07, "loss": 0.1724, "step": 5380 }, { "epoch": 1.12, "learning_rate": 1.3243355481727573e-07, "loss": 0.1489, "step": 5381 }, { "epoch": 1.12, "learning_rate": 1.3240240863787375e-07, "loss": 0.2324, "step": 5382 }, { "epoch": 1.12, "learning_rate": 1.3237126245847176e-07, "loss": 0.2337, "step": 5383 }, { "epoch": 1.12, "learning_rate": 1.3234011627906975e-07, "loss": 0.1488, "step": 5384 }, { "epoch": 1.12, "learning_rate": 1.3230897009966777e-07, "loss": 0.185, "step": 5385 }, { "epoch": 1.12, "learning_rate": 1.3227782392026578e-07, "loss": 0.1913, "step": 5386 }, { "epoch": 1.12, "learning_rate": 1.322466777408638e-07, "loss": 0.1561, "step": 5387 }, { "epoch": 1.12, "learning_rate": 1.322155315614618e-07, "loss": 0.1134, "step": 5388 }, { "epoch": 1.12, "learning_rate": 1.321843853820598e-07, "loss": 0.2205, "step": 5389 }, { "epoch": 1.12, "learning_rate": 1.321532392026578e-07, "loss": 0.1767, "step": 5390 }, { "epoch": 1.12, "learning_rate": 1.321220930232558e-07, "loss": 0.1844, "step": 5391 }, { "epoch": 1.12, "learning_rate": 1.320909468438538e-07, "loss": 0.2202, "step": 5392 }, { "epoch": 1.12, "learning_rate": 1.3205980066445181e-07, "loss": 0.1545, "step": 5393 }, { "epoch": 1.12, "learning_rate": 1.3202865448504983e-07, "loss": 0.1412, "step": 5394 }, { "epoch": 1.12, "learning_rate": 1.3199750830564782e-07, "loss": 0.2444, "step": 5395 }, { "epoch": 1.12, "learning_rate": 1.3196636212624584e-07, "loss": 0.1822, "step": 5396 }, { "epoch": 1.12, "learning_rate": 1.3193521594684385e-07, "loss": 0.2439, "step": 5397 }, { "epoch": 1.12, "learning_rate": 1.3190406976744187e-07, "loss": 0.1853, "step": 5398 }, { "epoch": 1.12, "learning_rate": 1.3187292358803986e-07, "loss": 0.2325, "step": 5399 }, { "epoch": 1.12, "learning_rate": 1.3184177740863787e-07, "loss": 0.3075, "step": 5400 }, { "epoch": 1.12, "learning_rate": 1.318106312292359e-07, "loss": 0.2228, "step": 5401 }, { "epoch": 1.12, "learning_rate": 1.3177948504983388e-07, "loss": 0.1982, "step": 5402 }, { "epoch": 1.12, "learning_rate": 1.317483388704319e-07, "loss": 0.1754, "step": 5403 }, { "epoch": 1.12, "learning_rate": 1.3171719269102988e-07, "loss": 0.2003, "step": 5404 }, { "epoch": 1.12, "learning_rate": 1.316860465116279e-07, "loss": 0.1553, "step": 5405 }, { "epoch": 1.12, "learning_rate": 1.3165490033222591e-07, "loss": 0.1451, "step": 5406 }, { "epoch": 1.12, "learning_rate": 1.316237541528239e-07, "loss": 0.1491, "step": 5407 }, { "epoch": 1.12, "learning_rate": 1.3159260797342192e-07, "loss": 0.2205, "step": 5408 }, { "epoch": 1.12, "learning_rate": 1.3156146179401994e-07, "loss": 0.1891, "step": 5409 }, { "epoch": 1.12, "learning_rate": 1.3153031561461792e-07, "loss": 0.1365, "step": 5410 }, { "epoch": 1.12, "learning_rate": 1.3149916943521594e-07, "loss": 0.1772, "step": 5411 }, { "epoch": 1.12, "learning_rate": 1.3146802325581396e-07, "loss": 0.1506, "step": 5412 }, { "epoch": 1.12, "learning_rate": 1.3143687707641197e-07, "loss": 0.188, "step": 5413 }, { "epoch": 1.12, "learning_rate": 1.3140573089700996e-07, "loss": 0.1329, "step": 5414 }, { "epoch": 1.12, "learning_rate": 1.3137458471760798e-07, "loss": 0.1906, "step": 5415 }, { "epoch": 1.12, "learning_rate": 1.3134343853820597e-07, "loss": 0.1638, "step": 5416 }, { "epoch": 1.12, "learning_rate": 1.3131229235880398e-07, "loss": 0.192, "step": 5417 }, { "epoch": 1.12, "learning_rate": 1.3128114617940197e-07, "loss": 0.1427, "step": 5418 }, { "epoch": 1.13, "learning_rate": 1.3125e-07, "loss": 0.1439, "step": 5419 }, { "epoch": 1.13, "learning_rate": 1.31218853820598e-07, "loss": 0.1444, "step": 5420 }, { "epoch": 1.13, "learning_rate": 1.31187707641196e-07, "loss": 0.2707, "step": 5421 }, { "epoch": 1.13, "learning_rate": 1.31156561461794e-07, "loss": 0.1858, "step": 5422 }, { "epoch": 1.13, "learning_rate": 1.3112541528239202e-07, "loss": 0.1392, "step": 5423 }, { "epoch": 1.13, "learning_rate": 1.3109426910299004e-07, "loss": 0.1455, "step": 5424 }, { "epoch": 1.13, "learning_rate": 1.3106312292358803e-07, "loss": 0.2153, "step": 5425 }, { "epoch": 1.13, "learning_rate": 1.3103197674418605e-07, "loss": 0.1439, "step": 5426 }, { "epoch": 1.13, "learning_rate": 1.3100083056478406e-07, "loss": 0.1776, "step": 5427 }, { "epoch": 1.13, "learning_rate": 1.3096968438538205e-07, "loss": 0.1519, "step": 5428 }, { "epoch": 1.13, "learning_rate": 1.3093853820598007e-07, "loss": 0.1513, "step": 5429 }, { "epoch": 1.13, "learning_rate": 1.3090739202657806e-07, "loss": 0.1512, "step": 5430 }, { "epoch": 1.13, "learning_rate": 1.3087624584717607e-07, "loss": 0.1761, "step": 5431 }, { "epoch": 1.13, "learning_rate": 1.3084509966777406e-07, "loss": 0.1917, "step": 5432 }, { "epoch": 1.13, "learning_rate": 1.3081395348837208e-07, "loss": 0.1945, "step": 5433 }, { "epoch": 1.13, "learning_rate": 1.307828073089701e-07, "loss": 0.1687, "step": 5434 }, { "epoch": 1.13, "learning_rate": 1.307516611295681e-07, "loss": 0.1448, "step": 5435 }, { "epoch": 1.13, "learning_rate": 1.307205149501661e-07, "loss": 0.1687, "step": 5436 }, { "epoch": 1.13, "learning_rate": 1.3068936877076411e-07, "loss": 0.2011, "step": 5437 }, { "epoch": 1.13, "learning_rate": 1.3065822259136213e-07, "loss": 0.1999, "step": 5438 }, { "epoch": 1.13, "learning_rate": 1.3062707641196012e-07, "loss": 0.1682, "step": 5439 }, { "epoch": 1.13, "learning_rate": 1.3059593023255813e-07, "loss": 0.1733, "step": 5440 }, { "epoch": 1.13, "learning_rate": 1.3056478405315615e-07, "loss": 0.1671, "step": 5441 }, { "epoch": 1.13, "learning_rate": 1.3053363787375414e-07, "loss": 0.1378, "step": 5442 }, { "epoch": 1.13, "learning_rate": 1.3050249169435216e-07, "loss": 0.2179, "step": 5443 }, { "epoch": 1.13, "learning_rate": 1.3047134551495014e-07, "loss": 0.1286, "step": 5444 }, { "epoch": 1.13, "learning_rate": 1.3044019933554816e-07, "loss": 0.1981, "step": 5445 }, { "epoch": 1.13, "learning_rate": 1.3040905315614618e-07, "loss": 0.1732, "step": 5446 }, { "epoch": 1.13, "learning_rate": 1.3037790697674417e-07, "loss": 0.1487, "step": 5447 }, { "epoch": 1.13, "learning_rate": 1.3034676079734218e-07, "loss": 0.1763, "step": 5448 }, { "epoch": 1.13, "learning_rate": 1.303156146179402e-07, "loss": 0.2403, "step": 5449 }, { "epoch": 1.13, "learning_rate": 1.3028446843853821e-07, "loss": 0.1423, "step": 5450 }, { "epoch": 1.13, "learning_rate": 1.302533222591362e-07, "loss": 0.1746, "step": 5451 }, { "epoch": 1.13, "learning_rate": 1.3022217607973422e-07, "loss": 0.1306, "step": 5452 }, { "epoch": 1.13, "learning_rate": 1.3019102990033223e-07, "loss": 0.138, "step": 5453 }, { "epoch": 1.13, "learning_rate": 1.3015988372093022e-07, "loss": 0.2112, "step": 5454 }, { "epoch": 1.13, "learning_rate": 1.3012873754152824e-07, "loss": 0.2023, "step": 5455 }, { "epoch": 1.13, "learning_rate": 1.3009759136212623e-07, "loss": 0.1522, "step": 5456 }, { "epoch": 1.13, "learning_rate": 1.3006644518272424e-07, "loss": 0.1606, "step": 5457 }, { "epoch": 1.13, "learning_rate": 1.3003529900332223e-07, "loss": 0.1826, "step": 5458 }, { "epoch": 1.13, "learning_rate": 1.3000415282392025e-07, "loss": 0.2155, "step": 5459 }, { "epoch": 1.13, "learning_rate": 1.2997300664451827e-07, "loss": 0.1743, "step": 5460 }, { "epoch": 1.13, "learning_rate": 1.2994186046511628e-07, "loss": 0.1373, "step": 5461 }, { "epoch": 1.13, "learning_rate": 1.299107142857143e-07, "loss": 0.1514, "step": 5462 }, { "epoch": 1.13, "learning_rate": 1.2987956810631229e-07, "loss": 0.1445, "step": 5463 }, { "epoch": 1.13, "learning_rate": 1.298484219269103e-07, "loss": 0.1636, "step": 5464 }, { "epoch": 1.13, "learning_rate": 1.298172757475083e-07, "loss": 0.1327, "step": 5465 }, { "epoch": 1.13, "learning_rate": 1.297861295681063e-07, "loss": 0.1773, "step": 5466 }, { "epoch": 1.14, "learning_rate": 1.2975498338870432e-07, "loss": 0.1615, "step": 5467 }, { "epoch": 1.14, "learning_rate": 1.297238372093023e-07, "loss": 0.1679, "step": 5468 }, { "epoch": 1.14, "learning_rate": 1.2969269102990033e-07, "loss": 0.1511, "step": 5469 }, { "epoch": 1.14, "learning_rate": 1.2966154485049832e-07, "loss": 0.1807, "step": 5470 }, { "epoch": 1.14, "learning_rate": 1.2963039867109633e-07, "loss": 0.1336, "step": 5471 }, { "epoch": 1.14, "learning_rate": 1.2959925249169435e-07, "loss": 0.1727, "step": 5472 }, { "epoch": 1.14, "learning_rate": 1.2956810631229237e-07, "loss": 0.2307, "step": 5473 }, { "epoch": 1.14, "learning_rate": 1.2953696013289035e-07, "loss": 0.2224, "step": 5474 }, { "epoch": 1.14, "learning_rate": 1.2950581395348837e-07, "loss": 0.2022, "step": 5475 }, { "epoch": 1.14, "learning_rate": 1.2947466777408639e-07, "loss": 0.2099, "step": 5476 }, { "epoch": 1.14, "learning_rate": 1.2944352159468438e-07, "loss": 0.218, "step": 5477 }, { "epoch": 1.14, "learning_rate": 1.294123754152824e-07, "loss": 0.2363, "step": 5478 }, { "epoch": 1.14, "learning_rate": 1.2938122923588038e-07, "loss": 0.2045, "step": 5479 }, { "epoch": 1.14, "learning_rate": 1.293500830564784e-07, "loss": 0.1557, "step": 5480 }, { "epoch": 1.14, "learning_rate": 1.293189368770764e-07, "loss": 0.2488, "step": 5481 }, { "epoch": 1.14, "learning_rate": 1.292877906976744e-07, "loss": 0.1712, "step": 5482 }, { "epoch": 1.14, "learning_rate": 1.2925664451827242e-07, "loss": 0.1418, "step": 5483 }, { "epoch": 1.14, "learning_rate": 1.292254983388704e-07, "loss": 0.1413, "step": 5484 }, { "epoch": 1.14, "learning_rate": 1.2919435215946842e-07, "loss": 0.1899, "step": 5485 }, { "epoch": 1.14, "learning_rate": 1.2916320598006644e-07, "loss": 0.1902, "step": 5486 }, { "epoch": 1.14, "learning_rate": 1.2913205980066445e-07, "loss": 0.2304, "step": 5487 }, { "epoch": 1.14, "learning_rate": 1.2910091362126247e-07, "loss": 0.1517, "step": 5488 }, { "epoch": 1.14, "learning_rate": 1.2906976744186046e-07, "loss": 0.1622, "step": 5489 }, { "epoch": 1.14, "learning_rate": 1.2903862126245848e-07, "loss": 0.1799, "step": 5490 }, { "epoch": 1.14, "learning_rate": 1.2900747508305647e-07, "loss": 0.1483, "step": 5491 }, { "epoch": 1.14, "learning_rate": 1.2897632890365448e-07, "loss": 0.1605, "step": 5492 }, { "epoch": 1.14, "learning_rate": 1.289451827242525e-07, "loss": 0.2235, "step": 5493 }, { "epoch": 1.14, "learning_rate": 1.2891403654485049e-07, "loss": 0.2025, "step": 5494 }, { "epoch": 1.14, "learning_rate": 1.288828903654485e-07, "loss": 0.1716, "step": 5495 }, { "epoch": 1.14, "learning_rate": 1.288517441860465e-07, "loss": 0.1621, "step": 5496 }, { "epoch": 1.14, "learning_rate": 1.288205980066445e-07, "loss": 0.1536, "step": 5497 }, { "epoch": 1.14, "learning_rate": 1.2878945182724252e-07, "loss": 0.2403, "step": 5498 }, { "epoch": 1.14, "learning_rate": 1.2875830564784054e-07, "loss": 0.2246, "step": 5499 }, { "epoch": 1.14, "learning_rate": 1.2872715946843853e-07, "loss": 0.1793, "step": 5500 }, { "epoch": 1.14, "learning_rate": 1.2869601328903654e-07, "loss": 0.1349, "step": 5501 }, { "epoch": 1.14, "learning_rate": 1.2866486710963456e-07, "loss": 0.2328, "step": 5502 }, { "epoch": 1.14, "learning_rate": 1.2863372093023255e-07, "loss": 0.213, "step": 5503 }, { "epoch": 1.14, "learning_rate": 1.2860257475083056e-07, "loss": 0.1914, "step": 5504 }, { "epoch": 1.14, "learning_rate": 1.2857142857142855e-07, "loss": 0.1442, "step": 5505 }, { "epoch": 1.14, "learning_rate": 1.2854028239202657e-07, "loss": 0.1842, "step": 5506 }, { "epoch": 1.14, "learning_rate": 1.2850913621262459e-07, "loss": 0.1705, "step": 5507 }, { "epoch": 1.14, "learning_rate": 1.2847799003322258e-07, "loss": 0.1572, "step": 5508 }, { "epoch": 1.14, "learning_rate": 1.284468438538206e-07, "loss": 0.16, "step": 5509 }, { "epoch": 1.14, "learning_rate": 1.284156976744186e-07, "loss": 0.1161, "step": 5510 }, { "epoch": 1.14, "learning_rate": 1.283845514950166e-07, "loss": 0.1888, "step": 5511 }, { "epoch": 1.14, "learning_rate": 1.283534053156146e-07, "loss": 0.1681, "step": 5512 }, { "epoch": 1.14, "learning_rate": 1.2832225913621263e-07, "loss": 0.1558, "step": 5513 }, { "epoch": 1.14, "learning_rate": 1.2829111295681064e-07, "loss": 0.1994, "step": 5514 }, { "epoch": 1.15, "learning_rate": 1.2825996677740863e-07, "loss": 0.2003, "step": 5515 }, { "epoch": 1.15, "learning_rate": 1.2822882059800665e-07, "loss": 0.2188, "step": 5516 }, { "epoch": 1.15, "learning_rate": 1.2819767441860464e-07, "loss": 0.1746, "step": 5517 }, { "epoch": 1.15, "learning_rate": 1.2816652823920265e-07, "loss": 0.1585, "step": 5518 }, { "epoch": 1.15, "learning_rate": 1.2813538205980064e-07, "loss": 0.1909, "step": 5519 }, { "epoch": 1.15, "learning_rate": 1.2810423588039866e-07, "loss": 0.2004, "step": 5520 }, { "epoch": 1.15, "learning_rate": 1.2807308970099668e-07, "loss": 0.2023, "step": 5521 }, { "epoch": 1.15, "learning_rate": 1.2804194352159466e-07, "loss": 0.1682, "step": 5522 }, { "epoch": 1.15, "learning_rate": 1.2801079734219268e-07, "loss": 0.1358, "step": 5523 }, { "epoch": 1.15, "learning_rate": 1.279796511627907e-07, "loss": 0.1414, "step": 5524 }, { "epoch": 1.15, "learning_rate": 1.279485049833887e-07, "loss": 0.1755, "step": 5525 }, { "epoch": 1.15, "learning_rate": 1.279173588039867e-07, "loss": 0.2185, "step": 5526 }, { "epoch": 1.15, "learning_rate": 1.2788621262458472e-07, "loss": 0.1814, "step": 5527 }, { "epoch": 1.15, "learning_rate": 1.2785506644518273e-07, "loss": 0.2031, "step": 5528 }, { "epoch": 1.15, "learning_rate": 1.2782392026578072e-07, "loss": 0.1662, "step": 5529 }, { "epoch": 1.15, "learning_rate": 1.2779277408637874e-07, "loss": 0.1859, "step": 5530 }, { "epoch": 1.15, "learning_rate": 1.2776162790697673e-07, "loss": 0.2049, "step": 5531 }, { "epoch": 1.15, "learning_rate": 1.2773048172757474e-07, "loss": 0.2053, "step": 5532 }, { "epoch": 1.15, "learning_rate": 1.2769933554817276e-07, "loss": 0.2268, "step": 5533 }, { "epoch": 1.15, "learning_rate": 1.2766818936877075e-07, "loss": 0.19, "step": 5534 }, { "epoch": 1.15, "learning_rate": 1.2763704318936876e-07, "loss": 0.1584, "step": 5535 }, { "epoch": 1.15, "learning_rate": 1.2760589700996678e-07, "loss": 0.1419, "step": 5536 }, { "epoch": 1.15, "learning_rate": 1.2757475083056477e-07, "loss": 0.1597, "step": 5537 }, { "epoch": 1.15, "learning_rate": 1.2754360465116279e-07, "loss": 0.1976, "step": 5538 }, { "epoch": 1.15, "learning_rate": 1.275124584717608e-07, "loss": 0.169, "step": 5539 }, { "epoch": 1.15, "learning_rate": 1.2748131229235882e-07, "loss": 0.1758, "step": 5540 }, { "epoch": 1.15, "learning_rate": 1.274501661129568e-07, "loss": 0.1162, "step": 5541 }, { "epoch": 1.15, "learning_rate": 1.2741901993355482e-07, "loss": 0.1884, "step": 5542 }, { "epoch": 1.15, "learning_rate": 1.273878737541528e-07, "loss": 0.1518, "step": 5543 }, { "epoch": 1.15, "learning_rate": 1.2735672757475083e-07, "loss": 0.1691, "step": 5544 }, { "epoch": 1.15, "learning_rate": 1.2732558139534882e-07, "loss": 0.2031, "step": 5545 }, { "epoch": 1.15, "learning_rate": 1.2729443521594683e-07, "loss": 0.1712, "step": 5546 }, { "epoch": 1.15, "learning_rate": 1.2726328903654485e-07, "loss": 0.1906, "step": 5547 }, { "epoch": 1.15, "learning_rate": 1.2723214285714284e-07, "loss": 0.2215, "step": 5548 }, { "epoch": 1.15, "learning_rate": 1.2720099667774085e-07, "loss": 0.206, "step": 5549 }, { "epoch": 1.15, "learning_rate": 1.2716985049833887e-07, "loss": 0.1949, "step": 5550 }, { "epoch": 1.15, "learning_rate": 1.2713870431893689e-07, "loss": 0.1678, "step": 5551 }, { "epoch": 1.15, "learning_rate": 1.2710755813953487e-07, "loss": 0.2188, "step": 5552 }, { "epoch": 1.15, "learning_rate": 1.270764119601329e-07, "loss": 0.1584, "step": 5553 }, { "epoch": 1.15, "learning_rate": 1.270452657807309e-07, "loss": 0.114, "step": 5554 }, { "epoch": 1.15, "learning_rate": 1.270141196013289e-07, "loss": 0.2098, "step": 5555 }, { "epoch": 1.15, "learning_rate": 1.269829734219269e-07, "loss": 0.1499, "step": 5556 }, { "epoch": 1.15, "learning_rate": 1.269518272425249e-07, "loss": 0.1394, "step": 5557 }, { "epoch": 1.15, "learning_rate": 1.2692068106312292e-07, "loss": 0.2352, "step": 5558 }, { "epoch": 1.15, "learning_rate": 1.268895348837209e-07, "loss": 0.1678, "step": 5559 }, { "epoch": 1.15, "learning_rate": 1.2685838870431892e-07, "loss": 0.2163, "step": 5560 }, { "epoch": 1.15, "learning_rate": 1.2682724252491694e-07, "loss": 0.2218, "step": 5561 }, { "epoch": 1.15, "learning_rate": 1.2679609634551495e-07, "loss": 0.1685, "step": 5562 }, { "epoch": 1.16, "learning_rate": 1.2676495016611294e-07, "loss": 0.1818, "step": 5563 }, { "epoch": 1.16, "learning_rate": 1.2673380398671096e-07, "loss": 0.1984, "step": 5564 }, { "epoch": 1.16, "learning_rate": 1.2670265780730897e-07, "loss": 0.2285, "step": 5565 }, { "epoch": 1.16, "learning_rate": 1.2667151162790696e-07, "loss": 0.1544, "step": 5566 }, { "epoch": 1.16, "learning_rate": 1.2664036544850498e-07, "loss": 0.1747, "step": 5567 }, { "epoch": 1.16, "learning_rate": 1.26609219269103e-07, "loss": 0.21, "step": 5568 }, { "epoch": 1.16, "learning_rate": 1.2657807308970098e-07, "loss": 0.129, "step": 5569 }, { "epoch": 1.16, "learning_rate": 1.26546926910299e-07, "loss": 0.1461, "step": 5570 }, { "epoch": 1.16, "learning_rate": 1.26515780730897e-07, "loss": 0.1197, "step": 5571 }, { "epoch": 1.16, "learning_rate": 1.26484634551495e-07, "loss": 0.2221, "step": 5572 }, { "epoch": 1.16, "learning_rate": 1.2645348837209302e-07, "loss": 0.181, "step": 5573 }, { "epoch": 1.16, "learning_rate": 1.26422342192691e-07, "loss": 0.1643, "step": 5574 }, { "epoch": 1.16, "learning_rate": 1.2639119601328903e-07, "loss": 0.1913, "step": 5575 }, { "epoch": 1.16, "learning_rate": 1.2636004983388704e-07, "loss": 0.2347, "step": 5576 }, { "epoch": 1.16, "learning_rate": 1.2632890365448506e-07, "loss": 0.1542, "step": 5577 }, { "epoch": 1.16, "learning_rate": 1.2629775747508305e-07, "loss": 0.2171, "step": 5578 }, { "epoch": 1.16, "learning_rate": 1.2626661129568106e-07, "loss": 0.1404, "step": 5579 }, { "epoch": 1.16, "learning_rate": 1.2623546511627908e-07, "loss": 0.2123, "step": 5580 }, { "epoch": 1.16, "learning_rate": 1.2620431893687707e-07, "loss": 0.2052, "step": 5581 }, { "epoch": 1.16, "learning_rate": 1.2617317275747508e-07, "loss": 0.241, "step": 5582 }, { "epoch": 1.16, "learning_rate": 1.2614202657807307e-07, "loss": 0.1146, "step": 5583 }, { "epoch": 1.16, "learning_rate": 1.261108803986711e-07, "loss": 0.1571, "step": 5584 }, { "epoch": 1.16, "learning_rate": 1.2607973421926908e-07, "loss": 0.1625, "step": 5585 }, { "epoch": 1.16, "learning_rate": 1.260485880398671e-07, "loss": 0.1772, "step": 5586 }, { "epoch": 1.16, "learning_rate": 1.260174418604651e-07, "loss": 0.1482, "step": 5587 }, { "epoch": 1.16, "learning_rate": 1.2598629568106313e-07, "loss": 0.248, "step": 5588 }, { "epoch": 1.16, "learning_rate": 1.2595514950166112e-07, "loss": 0.1519, "step": 5589 }, { "epoch": 1.16, "learning_rate": 1.2592400332225913e-07, "loss": 0.234, "step": 5590 }, { "epoch": 1.16, "learning_rate": 1.2589285714285715e-07, "loss": 0.179, "step": 5591 }, { "epoch": 1.16, "learning_rate": 1.2586171096345514e-07, "loss": 0.2145, "step": 5592 }, { "epoch": 1.16, "learning_rate": 1.2583056478405315e-07, "loss": 0.1759, "step": 5593 }, { "epoch": 1.16, "learning_rate": 1.2579941860465117e-07, "loss": 0.1887, "step": 5594 }, { "epoch": 1.16, "learning_rate": 1.2576827242524916e-07, "loss": 0.19, "step": 5595 }, { "epoch": 1.16, "learning_rate": 1.2573712624584717e-07, "loss": 0.1504, "step": 5596 }, { "epoch": 1.16, "learning_rate": 1.2570598006644516e-07, "loss": 0.1117, "step": 5597 }, { "epoch": 1.16, "learning_rate": 1.2567483388704318e-07, "loss": 0.121, "step": 5598 }, { "epoch": 1.16, "learning_rate": 1.256436877076412e-07, "loss": 0.1927, "step": 5599 }, { "epoch": 1.16, "learning_rate": 1.2561254152823918e-07, "loss": 0.1579, "step": 5600 }, { "epoch": 1.16, "learning_rate": 1.255813953488372e-07, "loss": 0.1833, "step": 5601 }, { "epoch": 1.16, "learning_rate": 1.2555024916943522e-07, "loss": 0.1806, "step": 5602 }, { "epoch": 1.16, "learning_rate": 1.2551910299003323e-07, "loss": 0.1774, "step": 5603 }, { "epoch": 1.16, "learning_rate": 1.2548795681063122e-07, "loss": 0.1695, "step": 5604 }, { "epoch": 1.16, "learning_rate": 1.2545681063122924e-07, "loss": 0.1638, "step": 5605 }, { "epoch": 1.16, "learning_rate": 1.2542566445182723e-07, "loss": 0.1442, "step": 5606 }, { "epoch": 1.16, "learning_rate": 1.2539451827242524e-07, "loss": 0.1933, "step": 5607 }, { "epoch": 1.16, "learning_rate": 1.2536337209302326e-07, "loss": 0.2062, "step": 5608 }, { "epoch": 1.16, "learning_rate": 1.2533222591362125e-07, "loss": 0.2304, "step": 5609 }, { "epoch": 1.16, "learning_rate": 1.2530107973421926e-07, "loss": 0.2285, "step": 5610 }, { "epoch": 1.17, "learning_rate": 1.2526993355481725e-07, "loss": 0.1955, "step": 5611 }, { "epoch": 1.17, "learning_rate": 1.2523878737541527e-07, "loss": 0.1849, "step": 5612 }, { "epoch": 1.17, "learning_rate": 1.2520764119601328e-07, "loss": 0.1255, "step": 5613 }, { "epoch": 1.17, "learning_rate": 1.251764950166113e-07, "loss": 0.1764, "step": 5614 }, { "epoch": 1.17, "learning_rate": 1.2514534883720932e-07, "loss": 0.1704, "step": 5615 }, { "epoch": 1.17, "learning_rate": 1.251142026578073e-07, "loss": 0.2154, "step": 5616 }, { "epoch": 1.17, "learning_rate": 1.2508305647840532e-07, "loss": 0.1945, "step": 5617 }, { "epoch": 1.17, "learning_rate": 1.250519102990033e-07, "loss": 0.2016, "step": 5618 }, { "epoch": 1.17, "learning_rate": 1.2502076411960133e-07, "loss": 0.1821, "step": 5619 }, { "epoch": 1.17, "learning_rate": 1.2498961794019934e-07, "loss": 0.1932, "step": 5620 }, { "epoch": 1.17, "learning_rate": 1.2495847176079733e-07, "loss": 0.1801, "step": 5621 }, { "epoch": 1.17, "learning_rate": 1.2492732558139535e-07, "loss": 0.1983, "step": 5622 }, { "epoch": 1.17, "learning_rate": 1.2489617940199334e-07, "loss": 0.1964, "step": 5623 }, { "epoch": 1.17, "learning_rate": 1.2486503322259135e-07, "loss": 0.2497, "step": 5624 }, { "epoch": 1.17, "learning_rate": 1.2483388704318937e-07, "loss": 0.1951, "step": 5625 }, { "epoch": 1.17, "learning_rate": 1.2480274086378736e-07, "loss": 0.1781, "step": 5626 }, { "epoch": 1.17, "learning_rate": 1.2477159468438537e-07, "loss": 0.2067, "step": 5627 }, { "epoch": 1.17, "learning_rate": 1.247404485049834e-07, "loss": 0.2004, "step": 5628 }, { "epoch": 1.17, "learning_rate": 1.247093023255814e-07, "loss": 0.1474, "step": 5629 }, { "epoch": 1.17, "learning_rate": 1.246781561461794e-07, "loss": 0.1621, "step": 5630 }, { "epoch": 1.17, "learning_rate": 1.246470099667774e-07, "loss": 0.1878, "step": 5631 }, { "epoch": 1.17, "learning_rate": 1.246158637873754e-07, "loss": 0.231, "step": 5632 }, { "epoch": 1.17, "learning_rate": 1.2458471760797342e-07, "loss": 0.2028, "step": 5633 }, { "epoch": 1.17, "learning_rate": 1.2455357142857143e-07, "loss": 0.253, "step": 5634 }, { "epoch": 1.17, "learning_rate": 1.2452242524916942e-07, "loss": 0.1477, "step": 5635 }, { "epoch": 1.17, "learning_rate": 1.2449127906976744e-07, "loss": 0.2178, "step": 5636 }, { "epoch": 1.17, "learning_rate": 1.2446013289036543e-07, "loss": 0.1475, "step": 5637 }, { "epoch": 1.17, "learning_rate": 1.2442898671096344e-07, "loss": 0.1696, "step": 5638 }, { "epoch": 1.17, "learning_rate": 1.2439784053156146e-07, "loss": 0.14, "step": 5639 }, { "epoch": 1.17, "learning_rate": 1.2436669435215947e-07, "loss": 0.1414, "step": 5640 }, { "epoch": 1.17, "learning_rate": 1.243355481727575e-07, "loss": 0.151, "step": 5641 }, { "epoch": 1.17, "learning_rate": 1.2430440199335548e-07, "loss": 0.2003, "step": 5642 }, { "epoch": 1.17, "learning_rate": 1.242732558139535e-07, "loss": 0.1365, "step": 5643 }, { "epoch": 1.17, "learning_rate": 1.2424210963455148e-07, "loss": 0.2071, "step": 5644 }, { "epoch": 1.17, "learning_rate": 1.242109634551495e-07, "loss": 0.1357, "step": 5645 }, { "epoch": 1.17, "learning_rate": 1.241798172757475e-07, "loss": 0.1839, "step": 5646 }, { "epoch": 1.17, "learning_rate": 1.241486710963455e-07, "loss": 0.2179, "step": 5647 }, { "epoch": 1.17, "learning_rate": 1.2411752491694352e-07, "loss": 0.1959, "step": 5648 }, { "epoch": 1.17, "learning_rate": 1.240863787375415e-07, "loss": 0.1927, "step": 5649 }, { "epoch": 1.17, "learning_rate": 1.2405523255813953e-07, "loss": 0.1637, "step": 5650 }, { "epoch": 1.17, "learning_rate": 1.2402408637873754e-07, "loss": 0.1556, "step": 5651 }, { "epoch": 1.17, "learning_rate": 1.2399294019933556e-07, "loss": 0.1664, "step": 5652 }, { "epoch": 1.17, "learning_rate": 1.2396179401993355e-07, "loss": 0.1365, "step": 5653 }, { "epoch": 1.17, "learning_rate": 1.2393064784053156e-07, "loss": 0.1637, "step": 5654 }, { "epoch": 1.17, "learning_rate": 1.2389950166112958e-07, "loss": 0.1721, "step": 5655 }, { "epoch": 1.17, "learning_rate": 1.2386835548172757e-07, "loss": 0.1948, "step": 5656 }, { "epoch": 1.17, "learning_rate": 1.2383720930232558e-07, "loss": 0.168, "step": 5657 }, { "epoch": 1.17, "learning_rate": 1.2380606312292357e-07, "loss": 0.1393, "step": 5658 }, { "epoch": 1.18, "learning_rate": 1.237749169435216e-07, "loss": 0.1574, "step": 5659 }, { "epoch": 1.18, "learning_rate": 1.237437707641196e-07, "loss": 0.1989, "step": 5660 }, { "epoch": 1.18, "learning_rate": 1.237126245847176e-07, "loss": 0.1631, "step": 5661 }, { "epoch": 1.18, "learning_rate": 1.236814784053156e-07, "loss": 0.2262, "step": 5662 }, { "epoch": 1.18, "learning_rate": 1.236503322259136e-07, "loss": 0.2007, "step": 5663 }, { "epoch": 1.18, "learning_rate": 1.2361918604651161e-07, "loss": 0.1782, "step": 5664 }, { "epoch": 1.18, "learning_rate": 1.2358803986710963e-07, "loss": 0.1624, "step": 5665 }, { "epoch": 1.18, "learning_rate": 1.2355689368770765e-07, "loss": 0.1536, "step": 5666 }, { "epoch": 1.18, "learning_rate": 1.2352574750830566e-07, "loss": 0.1863, "step": 5667 }, { "epoch": 1.18, "learning_rate": 1.2349460132890365e-07, "loss": 0.1146, "step": 5668 }, { "epoch": 1.18, "learning_rate": 1.2346345514950167e-07, "loss": 0.122, "step": 5669 }, { "epoch": 1.18, "learning_rate": 1.2343230897009966e-07, "loss": 0.1921, "step": 5670 }, { "epoch": 1.18, "learning_rate": 1.2340116279069767e-07, "loss": 0.1935, "step": 5671 }, { "epoch": 1.18, "learning_rate": 1.2337001661129566e-07, "loss": 0.1995, "step": 5672 }, { "epoch": 1.18, "learning_rate": 1.2333887043189368e-07, "loss": 0.1921, "step": 5673 }, { "epoch": 1.18, "learning_rate": 1.233077242524917e-07, "loss": 0.1523, "step": 5674 }, { "epoch": 1.18, "learning_rate": 1.2327657807308968e-07, "loss": 0.1787, "step": 5675 }, { "epoch": 1.18, "learning_rate": 1.232454318936877e-07, "loss": 0.2199, "step": 5676 }, { "epoch": 1.18, "learning_rate": 1.2321428571428571e-07, "loss": 0.1891, "step": 5677 }, { "epoch": 1.18, "learning_rate": 1.2318313953488373e-07, "loss": 0.2156, "step": 5678 }, { "epoch": 1.18, "learning_rate": 1.2315199335548172e-07, "loss": 0.1617, "step": 5679 }, { "epoch": 1.18, "learning_rate": 1.2312084717607974e-07, "loss": 0.1236, "step": 5680 }, { "epoch": 1.18, "learning_rate": 1.2308970099667775e-07, "loss": 0.1615, "step": 5681 }, { "epoch": 1.18, "learning_rate": 1.2305855481727574e-07, "loss": 0.1478, "step": 5682 }, { "epoch": 1.18, "learning_rate": 1.2302740863787376e-07, "loss": 0.1502, "step": 5683 }, { "epoch": 1.18, "learning_rate": 1.2299626245847175e-07, "loss": 0.1653, "step": 5684 }, { "epoch": 1.18, "learning_rate": 1.2296511627906976e-07, "loss": 0.1954, "step": 5685 }, { "epoch": 1.18, "learning_rate": 1.2293397009966775e-07, "loss": 0.1551, "step": 5686 }, { "epoch": 1.18, "learning_rate": 1.2290282392026577e-07, "loss": 0.2182, "step": 5687 }, { "epoch": 1.18, "learning_rate": 1.2287167774086378e-07, "loss": 0.2076, "step": 5688 }, { "epoch": 1.18, "learning_rate": 1.228405315614618e-07, "loss": 0.1493, "step": 5689 }, { "epoch": 1.18, "learning_rate": 1.228093853820598e-07, "loss": 0.2001, "step": 5690 }, { "epoch": 1.18, "learning_rate": 1.227782392026578e-07, "loss": 0.1849, "step": 5691 }, { "epoch": 1.18, "learning_rate": 1.2274709302325582e-07, "loss": 0.212, "step": 5692 }, { "epoch": 1.18, "learning_rate": 1.227159468438538e-07, "loss": 0.1155, "step": 5693 }, { "epoch": 1.18, "learning_rate": 1.2268480066445182e-07, "loss": 0.1907, "step": 5694 }, { "epoch": 1.18, "learning_rate": 1.2265365448504984e-07, "loss": 0.1327, "step": 5695 }, { "epoch": 1.18, "learning_rate": 1.2262250830564783e-07, "loss": 0.2367, "step": 5696 }, { "epoch": 1.18, "learning_rate": 1.2259136212624585e-07, "loss": 0.2226, "step": 5697 }, { "epoch": 1.18, "learning_rate": 1.2256021594684383e-07, "loss": 0.1242, "step": 5698 }, { "epoch": 1.18, "learning_rate": 1.2252906976744185e-07, "loss": 0.2505, "step": 5699 }, { "epoch": 1.18, "learning_rate": 1.2249792358803987e-07, "loss": 0.1888, "step": 5700 }, { "epoch": 1.18, "learning_rate": 1.2246677740863786e-07, "loss": 0.1805, "step": 5701 }, { "epoch": 1.18, "learning_rate": 1.2243563122923587e-07, "loss": 0.233, "step": 5702 }, { "epoch": 1.18, "learning_rate": 1.224044850498339e-07, "loss": 0.1595, "step": 5703 }, { "epoch": 1.18, "learning_rate": 1.223733388704319e-07, "loss": 0.1653, "step": 5704 }, { "epoch": 1.18, "learning_rate": 1.223421926910299e-07, "loss": 0.1711, "step": 5705 }, { "epoch": 1.18, "learning_rate": 1.223110465116279e-07, "loss": 0.1919, "step": 5706 }, { "epoch": 1.19, "learning_rate": 1.2227990033222592e-07, "loss": 0.1615, "step": 5707 }, { "epoch": 1.19, "learning_rate": 1.2224875415282391e-07, "loss": 0.2126, "step": 5708 }, { "epoch": 1.19, "learning_rate": 1.2221760797342193e-07, "loss": 0.1962, "step": 5709 }, { "epoch": 1.19, "learning_rate": 1.2218646179401992e-07, "loss": 0.1352, "step": 5710 }, { "epoch": 1.19, "learning_rate": 1.2215531561461793e-07, "loss": 0.1718, "step": 5711 }, { "epoch": 1.19, "learning_rate": 1.2212416943521592e-07, "loss": 0.1594, "step": 5712 }, { "epoch": 1.19, "learning_rate": 1.2209302325581394e-07, "loss": 0.1907, "step": 5713 }, { "epoch": 1.19, "learning_rate": 1.2206187707641196e-07, "loss": 0.1941, "step": 5714 }, { "epoch": 1.19, "learning_rate": 1.2203073089700997e-07, "loss": 0.0976, "step": 5715 }, { "epoch": 1.19, "learning_rate": 1.2199958471760796e-07, "loss": 0.1612, "step": 5716 }, { "epoch": 1.19, "learning_rate": 1.2196843853820598e-07, "loss": 0.2328, "step": 5717 }, { "epoch": 1.19, "learning_rate": 1.21937292358804e-07, "loss": 0.1679, "step": 5718 }, { "epoch": 1.19, "learning_rate": 1.2190614617940198e-07, "loss": 0.1992, "step": 5719 }, { "epoch": 1.19, "learning_rate": 1.21875e-07, "loss": 0.1765, "step": 5720 }, { "epoch": 1.19, "learning_rate": 1.2184385382059801e-07, "loss": 0.2257, "step": 5721 }, { "epoch": 1.19, "learning_rate": 1.21812707641196e-07, "loss": 0.2353, "step": 5722 }, { "epoch": 1.19, "learning_rate": 1.2178156146179402e-07, "loss": 0.1545, "step": 5723 }, { "epoch": 1.19, "learning_rate": 1.21750415282392e-07, "loss": 0.2054, "step": 5724 }, { "epoch": 1.19, "learning_rate": 1.2171926910299002e-07, "loss": 0.2074, "step": 5725 }, { "epoch": 1.19, "learning_rate": 1.2168812292358804e-07, "loss": 0.1962, "step": 5726 }, { "epoch": 1.19, "learning_rate": 1.2165697674418603e-07, "loss": 0.1781, "step": 5727 }, { "epoch": 1.19, "learning_rate": 1.2162583056478404e-07, "loss": 0.1788, "step": 5728 }, { "epoch": 1.19, "learning_rate": 1.2159468438538206e-07, "loss": 0.1766, "step": 5729 }, { "epoch": 1.19, "learning_rate": 1.2156353820598008e-07, "loss": 0.2348, "step": 5730 }, { "epoch": 1.19, "learning_rate": 1.2153239202657807e-07, "loss": 0.1951, "step": 5731 }, { "epoch": 1.19, "learning_rate": 1.2150124584717608e-07, "loss": 0.213, "step": 5732 }, { "epoch": 1.19, "learning_rate": 1.2147009966777407e-07, "loss": 0.1657, "step": 5733 }, { "epoch": 1.19, "learning_rate": 1.214389534883721e-07, "loss": 0.2073, "step": 5734 }, { "epoch": 1.19, "learning_rate": 1.214078073089701e-07, "loss": 0.1714, "step": 5735 }, { "epoch": 1.19, "learning_rate": 1.213766611295681e-07, "loss": 0.1469, "step": 5736 }, { "epoch": 1.19, "learning_rate": 1.213455149501661e-07, "loss": 0.1745, "step": 5737 }, { "epoch": 1.19, "learning_rate": 1.213143687707641e-07, "loss": 0.1804, "step": 5738 }, { "epoch": 1.19, "learning_rate": 1.212832225913621e-07, "loss": 0.2044, "step": 5739 }, { "epoch": 1.19, "learning_rate": 1.2125207641196013e-07, "loss": 0.1451, "step": 5740 }, { "epoch": 1.19, "learning_rate": 1.2122093023255814e-07, "loss": 0.2144, "step": 5741 }, { "epoch": 1.19, "learning_rate": 1.2118978405315613e-07, "loss": 0.2135, "step": 5742 }, { "epoch": 1.19, "learning_rate": 1.2115863787375415e-07, "loss": 0.1249, "step": 5743 }, { "epoch": 1.19, "learning_rate": 1.2112749169435217e-07, "loss": 0.2097, "step": 5744 }, { "epoch": 1.19, "learning_rate": 1.2109634551495016e-07, "loss": 0.1119, "step": 5745 }, { "epoch": 1.19, "learning_rate": 1.2106519933554817e-07, "loss": 0.2649, "step": 5746 }, { "epoch": 1.19, "learning_rate": 1.210340531561462e-07, "loss": 0.2324, "step": 5747 }, { "epoch": 1.19, "learning_rate": 1.2100290697674418e-07, "loss": 0.2315, "step": 5748 }, { "epoch": 1.19, "learning_rate": 1.209717607973422e-07, "loss": 0.2405, "step": 5749 }, { "epoch": 1.19, "learning_rate": 1.2094061461794018e-07, "loss": 0.1363, "step": 5750 }, { "epoch": 1.19, "learning_rate": 1.209094684385382e-07, "loss": 0.1663, "step": 5751 }, { "epoch": 1.19, "learning_rate": 1.208783222591362e-07, "loss": 0.1856, "step": 5752 }, { "epoch": 1.19, "learning_rate": 1.208471760797342e-07, "loss": 0.1897, "step": 5753 }, { "epoch": 1.19, "learning_rate": 1.2081602990033222e-07, "loss": 0.1986, "step": 5754 }, { "epoch": 1.19, "learning_rate": 1.2078488372093023e-07, "loss": 0.159, "step": 5755 }, { "epoch": 1.2, "learning_rate": 1.2075373754152825e-07, "loss": 0.1883, "step": 5756 }, { "epoch": 1.2, "learning_rate": 1.2072259136212624e-07, "loss": 0.1528, "step": 5757 }, { "epoch": 1.2, "learning_rate": 1.2069144518272426e-07, "loss": 0.1749, "step": 5758 }, { "epoch": 1.2, "learning_rate": 1.2066029900332224e-07, "loss": 0.1741, "step": 5759 }, { "epoch": 1.2, "learning_rate": 1.2062915282392026e-07, "loss": 0.1542, "step": 5760 }, { "epoch": 1.2, "learning_rate": 1.2059800664451828e-07, "loss": 0.1517, "step": 5761 }, { "epoch": 1.2, "learning_rate": 1.2056686046511627e-07, "loss": 0.1608, "step": 5762 }, { "epoch": 1.2, "learning_rate": 1.2053571428571428e-07, "loss": 0.1777, "step": 5763 }, { "epoch": 1.2, "learning_rate": 1.2050456810631227e-07, "loss": 0.2293, "step": 5764 }, { "epoch": 1.2, "learning_rate": 1.2047342192691029e-07, "loss": 0.1747, "step": 5765 }, { "epoch": 1.2, "learning_rate": 1.2047342192691029e-07, "loss": 0.1522, "step": 5766 }, { "epoch": 1.2, "learning_rate": 1.204422757475083e-07, "loss": 0.2336, "step": 5767 }, { "epoch": 1.2, "learning_rate": 1.2041112956810632e-07, "loss": 0.1894, "step": 5768 }, { "epoch": 1.2, "learning_rate": 1.203799833887043e-07, "loss": 0.1855, "step": 5769 }, { "epoch": 1.2, "learning_rate": 1.2034883720930232e-07, "loss": 0.1939, "step": 5770 }, { "epoch": 1.2, "learning_rate": 1.2031769102990034e-07, "loss": 0.15, "step": 5771 }, { "epoch": 1.2, "learning_rate": 1.2028654485049833e-07, "loss": 0.1792, "step": 5772 }, { "epoch": 1.2, "learning_rate": 1.2025539867109634e-07, "loss": 0.1516, "step": 5773 }, { "epoch": 1.2, "learning_rate": 1.2022425249169433e-07, "loss": 0.1992, "step": 5774 }, { "epoch": 1.2, "learning_rate": 1.2019310631229235e-07, "loss": 0.1633, "step": 5775 }, { "epoch": 1.2, "learning_rate": 1.2016196013289037e-07, "loss": 0.2019, "step": 5776 }, { "epoch": 1.2, "learning_rate": 1.2013081395348835e-07, "loss": 0.1584, "step": 5777 }, { "epoch": 1.2, "learning_rate": 1.2009966777408637e-07, "loss": 0.1403, "step": 5778 }, { "epoch": 1.2, "learning_rate": 1.2006852159468439e-07, "loss": 0.1401, "step": 5779 }, { "epoch": 1.2, "learning_rate": 1.2003737541528238e-07, "loss": 0.1496, "step": 5780 }, { "epoch": 1.2, "learning_rate": 1.200062292358804e-07, "loss": 0.1466, "step": 5781 }, { "epoch": 1.2, "learning_rate": 1.199750830564784e-07, "loss": 0.1994, "step": 5782 }, { "epoch": 1.2, "learning_rate": 1.1994393687707642e-07, "loss": 0.1615, "step": 5783 }, { "epoch": 1.2, "learning_rate": 1.199127906976744e-07, "loss": 0.1951, "step": 5784 }, { "epoch": 1.2, "learning_rate": 1.1988164451827243e-07, "loss": 0.2382, "step": 5785 }, { "epoch": 1.2, "learning_rate": 1.1985049833887042e-07, "loss": 0.1569, "step": 5786 }, { "epoch": 1.2, "learning_rate": 1.1981935215946843e-07, "loss": 0.1088, "step": 5787 }, { "epoch": 1.2, "learning_rate": 1.1978820598006645e-07, "loss": 0.161, "step": 5788 }, { "epoch": 1.2, "learning_rate": 1.1975705980066444e-07, "loss": 0.1623, "step": 5789 }, { "epoch": 1.2, "learning_rate": 1.1972591362126245e-07, "loss": 0.2079, "step": 5790 }, { "epoch": 1.2, "learning_rate": 1.1969476744186044e-07, "loss": 0.173, "step": 5791 }, { "epoch": 1.2, "learning_rate": 1.1966362126245846e-07, "loss": 0.16, "step": 5792 }, { "epoch": 1.2, "learning_rate": 1.1963247508305648e-07, "loss": 0.2055, "step": 5793 }, { "epoch": 1.2, "learning_rate": 1.196013289036545e-07, "loss": 0.1941, "step": 5794 }, { "epoch": 1.2, "learning_rate": 1.195701827242525e-07, "loss": 0.1743, "step": 5795 }, { "epoch": 1.2, "learning_rate": 1.195390365448505e-07, "loss": 0.1177, "step": 5796 }, { "epoch": 1.2, "learning_rate": 1.195078903654485e-07, "loss": 0.215, "step": 5797 }, { "epoch": 1.2, "learning_rate": 1.194767441860465e-07, "loss": 0.1812, "step": 5798 }, { "epoch": 1.2, "learning_rate": 1.1944559800664452e-07, "loss": 0.145, "step": 5799 }, { "epoch": 1.2, "learning_rate": 1.194144518272425e-07, "loss": 0.1384, "step": 5800 }, { "epoch": 1.2, "learning_rate": 1.1938330564784052e-07, "loss": 0.2065, "step": 5801 }, { "epoch": 1.2, "learning_rate": 1.1935215946843854e-07, "loss": 0.166, "step": 5802 }, { "epoch": 1.2, "learning_rate": 1.1932101328903653e-07, "loss": 0.2194, "step": 5803 }, { "epoch": 1.21, "learning_rate": 1.1928986710963454e-07, "loss": 0.1737, "step": 5804 }, { "epoch": 1.21, "learning_rate": 1.1925872093023256e-07, "loss": 0.1938, "step": 5805 }, { "epoch": 1.21, "learning_rate": 1.1922757475083055e-07, "loss": 0.152, "step": 5806 }, { "epoch": 1.21, "learning_rate": 1.1919642857142856e-07, "loss": 0.1331, "step": 5807 }, { "epoch": 1.21, "learning_rate": 1.1916528239202657e-07, "loss": 0.2561, "step": 5808 }, { "epoch": 1.21, "learning_rate": 1.1913413621262458e-07, "loss": 0.1832, "step": 5809 }, { "epoch": 1.21, "learning_rate": 1.1910299003322259e-07, "loss": 0.1408, "step": 5810 }, { "epoch": 1.21, "learning_rate": 1.190718438538206e-07, "loss": 0.1589, "step": 5811 }, { "epoch": 1.21, "learning_rate": 1.1904069767441859e-07, "loss": 0.2159, "step": 5812 }, { "epoch": 1.21, "learning_rate": 1.1900955149501661e-07, "loss": 0.2005, "step": 5813 }, { "epoch": 1.21, "learning_rate": 1.1897840531561461e-07, "loss": 0.2106, "step": 5814 }, { "epoch": 1.21, "learning_rate": 1.1894725913621263e-07, "loss": 0.2005, "step": 5815 }, { "epoch": 1.21, "learning_rate": 1.1891611295681063e-07, "loss": 0.1429, "step": 5816 }, { "epoch": 1.21, "learning_rate": 1.1888496677740863e-07, "loss": 0.2485, "step": 5817 }, { "epoch": 1.21, "learning_rate": 1.1885382059800665e-07, "loss": 0.2411, "step": 5818 }, { "epoch": 1.21, "learning_rate": 1.1882267441860464e-07, "loss": 0.2826, "step": 5819 }, { "epoch": 1.21, "learning_rate": 1.1879152823920265e-07, "loss": 0.1669, "step": 5820 }, { "epoch": 1.21, "learning_rate": 1.1876038205980065e-07, "loss": 0.1753, "step": 5821 }, { "epoch": 1.21, "learning_rate": 1.1872923588039867e-07, "loss": 0.1924, "step": 5822 }, { "epoch": 1.21, "learning_rate": 1.1869808970099667e-07, "loss": 0.1713, "step": 5823 }, { "epoch": 1.21, "learning_rate": 1.1866694352159467e-07, "loss": 0.18, "step": 5824 }, { "epoch": 1.21, "learning_rate": 1.1863579734219269e-07, "loss": 0.1927, "step": 5825 }, { "epoch": 1.21, "learning_rate": 1.1860465116279068e-07, "loss": 0.1699, "step": 5826 }, { "epoch": 1.21, "learning_rate": 1.185735049833887e-07, "loss": 0.148, "step": 5827 }, { "epoch": 1.21, "learning_rate": 1.185423588039867e-07, "loss": 0.199, "step": 5828 }, { "epoch": 1.21, "learning_rate": 1.1851121262458471e-07, "loss": 0.2213, "step": 5829 }, { "epoch": 1.21, "learning_rate": 1.1848006644518273e-07, "loss": 0.2072, "step": 5830 }, { "epoch": 1.21, "learning_rate": 1.1844892026578072e-07, "loss": 0.1541, "step": 5831 }, { "epoch": 1.21, "learning_rate": 1.1841777408637874e-07, "loss": 0.1414, "step": 5832 }, { "epoch": 1.21, "learning_rate": 1.1838662790697674e-07, "loss": 0.2053, "step": 5833 }, { "epoch": 1.21, "learning_rate": 1.1835548172757474e-07, "loss": 0.2179, "step": 5834 }, { "epoch": 1.21, "learning_rate": 1.1832433554817276e-07, "loss": 0.1878, "step": 5835 }, { "epoch": 1.21, "learning_rate": 1.1829318936877076e-07, "loss": 0.1654, "step": 5836 }, { "epoch": 1.21, "learning_rate": 1.1826204318936877e-07, "loss": 0.2376, "step": 5837 }, { "epoch": 1.21, "learning_rate": 1.1823089700996676e-07, "loss": 0.1724, "step": 5838 }, { "epoch": 1.21, "learning_rate": 1.1819975083056478e-07, "loss": 0.1573, "step": 5839 }, { "epoch": 1.21, "learning_rate": 1.1816860465116278e-07, "loss": 0.1625, "step": 5840 }, { "epoch": 1.21, "learning_rate": 1.181374584717608e-07, "loss": 0.1616, "step": 5841 }, { "epoch": 1.21, "learning_rate": 1.181063122923588e-07, "loss": 0.1784, "step": 5842 }, { "epoch": 1.21, "learning_rate": 1.180751661129568e-07, "loss": 0.1853, "step": 5843 }, { "epoch": 1.21, "learning_rate": 1.1804401993355482e-07, "loss": 0.1627, "step": 5844 }, { "epoch": 1.21, "learning_rate": 1.1801287375415281e-07, "loss": 0.1671, "step": 5845 }, { "epoch": 1.21, "learning_rate": 1.1798172757475082e-07, "loss": 0.195, "step": 5846 }, { "epoch": 1.21, "learning_rate": 1.1795058139534883e-07, "loss": 0.1892, "step": 5847 }, { "epoch": 1.21, "learning_rate": 1.1791943521594684e-07, "loss": 0.238, "step": 5848 }, { "epoch": 1.21, "learning_rate": 1.1788828903654486e-07, "loss": 0.1202, "step": 5849 }, { "epoch": 1.21, "learning_rate": 1.1785714285714285e-07, "loss": 0.2251, "step": 5850 }, { "epoch": 1.21, "learning_rate": 1.1782599667774086e-07, "loss": 0.1867, "step": 5851 }, { "epoch": 1.22, "learning_rate": 1.1779485049833887e-07, "loss": 0.2224, "step": 5852 }, { "epoch": 1.22, "learning_rate": 1.1776370431893687e-07, "loss": 0.2972, "step": 5853 }, { "epoch": 1.22, "learning_rate": 1.1773255813953487e-07, "loss": 0.1989, "step": 5854 }, { "epoch": 1.22, "learning_rate": 1.1770141196013289e-07, "loss": 0.1658, "step": 5855 }, { "epoch": 1.22, "learning_rate": 1.176702657807309e-07, "loss": 0.1909, "step": 5856 }, { "epoch": 1.22, "learning_rate": 1.1763911960132889e-07, "loss": 0.1866, "step": 5857 }, { "epoch": 1.22, "learning_rate": 1.1760797342192691e-07, "loss": 0.2221, "step": 5858 }, { "epoch": 1.22, "learning_rate": 1.1757682724252491e-07, "loss": 0.1893, "step": 5859 }, { "epoch": 1.22, "learning_rate": 1.1754568106312291e-07, "loss": 0.1932, "step": 5860 }, { "epoch": 1.22, "learning_rate": 1.1751453488372092e-07, "loss": 0.12, "step": 5861 }, { "epoch": 1.22, "learning_rate": 1.1748338870431893e-07, "loss": 0.2094, "step": 5862 }, { "epoch": 1.22, "learning_rate": 1.1745224252491695e-07, "loss": 0.1648, "step": 5863 }, { "epoch": 1.22, "learning_rate": 1.1742109634551494e-07, "loss": 0.182, "step": 5864 }, { "epoch": 1.22, "learning_rate": 1.1738995016611295e-07, "loss": 0.1605, "step": 5865 }, { "epoch": 1.22, "learning_rate": 1.1735880398671096e-07, "loss": 0.2165, "step": 5866 }, { "epoch": 1.22, "learning_rate": 1.1732765780730897e-07, "loss": 0.2319, "step": 5867 }, { "epoch": 1.22, "learning_rate": 1.1729651162790696e-07, "loss": 0.1706, "step": 5868 }, { "epoch": 1.22, "learning_rate": 1.1726536544850498e-07, "loss": 0.144, "step": 5869 }, { "epoch": 1.22, "learning_rate": 1.1723421926910299e-07, "loss": 0.1715, "step": 5870 }, { "epoch": 1.22, "learning_rate": 1.1720307308970098e-07, "loss": 0.1828, "step": 5871 }, { "epoch": 1.22, "learning_rate": 1.17171926910299e-07, "loss": 0.1782, "step": 5872 }, { "epoch": 1.22, "learning_rate": 1.17140780730897e-07, "loss": 0.1786, "step": 5873 }, { "epoch": 1.22, "learning_rate": 1.1710963455149502e-07, "loss": 0.1535, "step": 5874 }, { "epoch": 1.22, "learning_rate": 1.1707848837209303e-07, "loss": 0.1201, "step": 5875 }, { "epoch": 1.22, "learning_rate": 1.1704734219269102e-07, "loss": 0.1656, "step": 5876 }, { "epoch": 1.22, "learning_rate": 1.1701619601328904e-07, "loss": 0.1677, "step": 5877 }, { "epoch": 1.22, "learning_rate": 1.1698504983388704e-07, "loss": 0.1721, "step": 5878 }, { "epoch": 1.22, "learning_rate": 1.1695390365448504e-07, "loss": 0.1164, "step": 5879 }, { "epoch": 1.22, "learning_rate": 1.1692275747508304e-07, "loss": 0.1945, "step": 5880 }, { "epoch": 1.22, "learning_rate": 1.1689161129568106e-07, "loss": 0.1546, "step": 5881 }, { "epoch": 1.22, "learning_rate": 1.1686046511627908e-07, "loss": 0.1755, "step": 5882 }, { "epoch": 1.22, "learning_rate": 1.1682931893687707e-07, "loss": 0.219, "step": 5883 }, { "epoch": 1.22, "learning_rate": 1.1679817275747508e-07, "loss": 0.1202, "step": 5884 }, { "epoch": 1.22, "learning_rate": 1.1676702657807308e-07, "loss": 0.1853, "step": 5885 }, { "epoch": 1.22, "learning_rate": 1.167358803986711e-07, "loss": 0.2102, "step": 5886 }, { "epoch": 1.22, "learning_rate": 1.1670473421926909e-07, "loss": 0.1773, "step": 5887 }, { "epoch": 1.22, "learning_rate": 1.166735880398671e-07, "loss": 0.1888, "step": 5888 }, { "epoch": 1.22, "learning_rate": 1.1664244186046512e-07, "loss": 0.1753, "step": 5889 }, { "epoch": 1.22, "learning_rate": 1.1661129568106311e-07, "loss": 0.1328, "step": 5890 }, { "epoch": 1.22, "learning_rate": 1.1658014950166113e-07, "loss": 0.1613, "step": 5891 }, { "epoch": 1.22, "learning_rate": 1.1654900332225913e-07, "loss": 0.1382, "step": 5892 }, { "epoch": 1.22, "learning_rate": 1.1651785714285714e-07, "loss": 0.1523, "step": 5893 }, { "epoch": 1.22, "learning_rate": 1.1648671096345513e-07, "loss": 0.1802, "step": 5894 }, { "epoch": 1.22, "learning_rate": 1.1645556478405315e-07, "loss": 0.2215, "step": 5895 }, { "epoch": 1.22, "learning_rate": 1.1642441860465117e-07, "loss": 0.2049, "step": 5896 }, { "epoch": 1.22, "learning_rate": 1.1639327242524916e-07, "loss": 0.2242, "step": 5897 }, { "epoch": 1.22, "learning_rate": 1.1636212624584717e-07, "loss": 0.1927, "step": 5898 }, { "epoch": 1.22, "learning_rate": 1.1633098006644517e-07, "loss": 0.1896, "step": 5899 }, { "epoch": 1.23, "learning_rate": 1.1629983388704319e-07, "loss": 0.186, "step": 5900 }, { "epoch": 1.23, "learning_rate": 1.1626868770764118e-07, "loss": 0.2147, "step": 5901 }, { "epoch": 1.23, "learning_rate": 1.162375415282392e-07, "loss": 0.1644, "step": 5902 }, { "epoch": 1.23, "learning_rate": 1.1620639534883721e-07, "loss": 0.1265, "step": 5903 }, { "epoch": 1.23, "learning_rate": 1.1617524916943521e-07, "loss": 0.1922, "step": 5904 }, { "epoch": 1.23, "learning_rate": 1.1614410299003322e-07, "loss": 0.2435, "step": 5905 }, { "epoch": 1.23, "learning_rate": 1.1611295681063122e-07, "loss": 0.14, "step": 5906 }, { "epoch": 1.23, "learning_rate": 1.1608181063122923e-07, "loss": 0.1374, "step": 5907 }, { "epoch": 1.23, "learning_rate": 1.1605066445182722e-07, "loss": 0.1988, "step": 5908 }, { "epoch": 1.23, "learning_rate": 1.1601951827242524e-07, "loss": 0.2243, "step": 5909 }, { "epoch": 1.23, "learning_rate": 1.1598837209302325e-07, "loss": 0.2012, "step": 5910 }, { "epoch": 1.23, "learning_rate": 1.1595722591362126e-07, "loss": 0.1617, "step": 5911 }, { "epoch": 1.23, "learning_rate": 1.1592607973421927e-07, "loss": 0.1539, "step": 5912 }, { "epoch": 1.23, "learning_rate": 1.1589493355481726e-07, "loss": 0.152, "step": 5913 }, { "epoch": 1.23, "learning_rate": 1.1586378737541528e-07, "loss": 0.1739, "step": 5914 }, { "epoch": 1.23, "learning_rate": 1.158326411960133e-07, "loss": 0.2011, "step": 5915 }, { "epoch": 1.23, "learning_rate": 1.1580149501661128e-07, "loss": 0.1909, "step": 5916 }, { "epoch": 1.23, "learning_rate": 1.157703488372093e-07, "loss": 0.1854, "step": 5917 }, { "epoch": 1.23, "learning_rate": 1.157392026578073e-07, "loss": 0.1372, "step": 5918 }, { "epoch": 1.23, "learning_rate": 1.1570805647840532e-07, "loss": 0.1618, "step": 5919 }, { "epoch": 1.23, "learning_rate": 1.1567691029900331e-07, "loss": 0.1511, "step": 5920 }, { "epoch": 1.23, "learning_rate": 1.1564576411960132e-07, "loss": 0.1374, "step": 5921 }, { "epoch": 1.23, "learning_rate": 1.1561461794019934e-07, "loss": 0.1788, "step": 5922 }, { "epoch": 1.23, "learning_rate": 1.1558347176079734e-07, "loss": 0.1439, "step": 5923 }, { "epoch": 1.23, "learning_rate": 1.1555232558139534e-07, "loss": 0.2127, "step": 5924 }, { "epoch": 1.23, "learning_rate": 1.1552117940199335e-07, "loss": 0.1787, "step": 5925 }, { "epoch": 1.23, "learning_rate": 1.1549003322259136e-07, "loss": 0.203, "step": 5926 }, { "epoch": 1.23, "learning_rate": 1.1545888704318935e-07, "loss": 0.1682, "step": 5927 }, { "epoch": 1.23, "learning_rate": 1.1542774086378737e-07, "loss": 0.1611, "step": 5928 }, { "epoch": 1.23, "learning_rate": 1.1539659468438538e-07, "loss": 0.1511, "step": 5929 }, { "epoch": 1.23, "learning_rate": 1.1536544850498339e-07, "loss": 0.1607, "step": 5930 }, { "epoch": 1.23, "learning_rate": 1.1533430232558139e-07, "loss": 0.1426, "step": 5931 }, { "epoch": 1.23, "learning_rate": 1.1530315614617939e-07, "loss": 0.2321, "step": 5932 }, { "epoch": 1.23, "learning_rate": 1.1527200996677741e-07, "loss": 0.1488, "step": 5933 }, { "epoch": 1.23, "learning_rate": 1.152408637873754e-07, "loss": 0.2079, "step": 5934 }, { "epoch": 1.23, "learning_rate": 1.1520971760797341e-07, "loss": 0.2156, "step": 5935 }, { "epoch": 1.23, "learning_rate": 1.1517857142857143e-07, "loss": 0.1712, "step": 5936 }, { "epoch": 1.23, "learning_rate": 1.1514742524916943e-07, "loss": 0.1885, "step": 5937 }, { "epoch": 1.23, "learning_rate": 1.1511627906976745e-07, "loss": 0.2709, "step": 5938 }, { "epoch": 1.23, "learning_rate": 1.1508513289036544e-07, "loss": 0.2001, "step": 5939 }, { "epoch": 1.23, "learning_rate": 1.1505398671096345e-07, "loss": 0.1735, "step": 5940 }, { "epoch": 1.23, "learning_rate": 1.1502284053156145e-07, "loss": 0.1705, "step": 5941 }, { "epoch": 1.23, "learning_rate": 1.1499169435215946e-07, "loss": 0.189, "step": 5942 }, { "epoch": 1.23, "learning_rate": 1.1496054817275747e-07, "loss": 0.1618, "step": 5943 }, { "epoch": 1.23, "learning_rate": 1.1492940199335548e-07, "loss": 0.171, "step": 5944 }, { "epoch": 1.23, "learning_rate": 1.1489825581395349e-07, "loss": 0.1975, "step": 5945 }, { "epoch": 1.23, "learning_rate": 1.1486710963455148e-07, "loss": 0.2308, "step": 5946 }, { "epoch": 1.23, "learning_rate": 1.148359634551495e-07, "loss": 0.2146, "step": 5947 }, { "epoch": 1.24, "learning_rate": 1.148048172757475e-07, "loss": 0.1526, "step": 5948 }, { "epoch": 1.24, "learning_rate": 1.1477367109634551e-07, "loss": 0.21, "step": 5949 }, { "epoch": 1.24, "learning_rate": 1.1474252491694352e-07, "loss": 0.1549, "step": 5950 }, { "epoch": 1.24, "learning_rate": 1.1471137873754152e-07, "loss": 0.1352, "step": 5951 }, { "epoch": 1.24, "learning_rate": 1.1468023255813954e-07, "loss": 0.1859, "step": 5952 }, { "epoch": 1.24, "learning_rate": 1.1464908637873753e-07, "loss": 0.2378, "step": 5953 }, { "epoch": 1.24, "learning_rate": 1.1461794019933554e-07, "loss": 0.1637, "step": 5954 }, { "epoch": 1.24, "learning_rate": 1.1458679401993354e-07, "loss": 0.1807, "step": 5955 }, { "epoch": 1.24, "learning_rate": 1.1455564784053156e-07, "loss": 0.1471, "step": 5956 }, { "epoch": 1.24, "learning_rate": 1.1452450166112958e-07, "loss": 0.1948, "step": 5957 }, { "epoch": 1.24, "learning_rate": 1.1449335548172756e-07, "loss": 0.2157, "step": 5958 }, { "epoch": 1.24, "learning_rate": 1.1446220930232558e-07, "loss": 0.1513, "step": 5959 }, { "epoch": 1.24, "learning_rate": 1.1443106312292358e-07, "loss": 0.1981, "step": 5960 }, { "epoch": 1.24, "learning_rate": 1.1439991694352159e-07, "loss": 0.2243, "step": 5961 }, { "epoch": 1.24, "learning_rate": 1.143687707641196e-07, "loss": 0.2045, "step": 5962 }, { "epoch": 1.24, "learning_rate": 1.143376245847176e-07, "loss": 0.1835, "step": 5963 }, { "epoch": 1.24, "learning_rate": 1.1430647840531562e-07, "loss": 0.2121, "step": 5964 }, { "epoch": 1.24, "learning_rate": 1.1427533222591361e-07, "loss": 0.2019, "step": 5965 }, { "epoch": 1.24, "learning_rate": 1.1424418604651162e-07, "loss": 0.1672, "step": 5966 }, { "epoch": 1.24, "learning_rate": 1.1421303986710963e-07, "loss": 0.1762, "step": 5967 }, { "epoch": 1.24, "learning_rate": 1.1418189368770763e-07, "loss": 0.1982, "step": 5968 }, { "epoch": 1.24, "learning_rate": 1.1415074750830565e-07, "loss": 0.1814, "step": 5969 }, { "epoch": 1.24, "learning_rate": 1.1411960132890365e-07, "loss": 0.173, "step": 5970 }, { "epoch": 1.24, "learning_rate": 1.1408845514950166e-07, "loss": 0.1911, "step": 5971 }, { "epoch": 1.24, "learning_rate": 1.1405730897009965e-07, "loss": 0.2149, "step": 5972 }, { "epoch": 1.24, "learning_rate": 1.1402616279069767e-07, "loss": 0.1491, "step": 5973 }, { "epoch": 1.24, "learning_rate": 1.1399501661129567e-07, "loss": 0.244, "step": 5974 }, { "epoch": 1.24, "learning_rate": 1.1396387043189369e-07, "loss": 0.1592, "step": 5975 }, { "epoch": 1.24, "learning_rate": 1.1393272425249169e-07, "loss": 0.2081, "step": 5976 }, { "epoch": 1.24, "learning_rate": 1.1390157807308969e-07, "loss": 0.1317, "step": 5977 }, { "epoch": 1.24, "learning_rate": 1.1387043189368771e-07, "loss": 0.1618, "step": 5978 }, { "epoch": 1.24, "learning_rate": 1.138392857142857e-07, "loss": 0.1661, "step": 5979 }, { "epoch": 1.24, "learning_rate": 1.1380813953488371e-07, "loss": 0.1633, "step": 5980 }, { "epoch": 1.24, "learning_rate": 1.1377699335548172e-07, "loss": 0.2033, "step": 5981 }, { "epoch": 1.24, "learning_rate": 1.1374584717607973e-07, "loss": 0.2028, "step": 5982 }, { "epoch": 1.24, "learning_rate": 1.1371470099667775e-07, "loss": 0.2083, "step": 5983 }, { "epoch": 1.24, "learning_rate": 1.1368355481727574e-07, "loss": 0.146, "step": 5984 }, { "epoch": 1.24, "learning_rate": 1.1365240863787375e-07, "loss": 0.1413, "step": 5985 }, { "epoch": 1.24, "learning_rate": 1.1362126245847176e-07, "loss": 0.1109, "step": 5986 }, { "epoch": 1.24, "learning_rate": 1.1359011627906976e-07, "loss": 0.1725, "step": 5987 }, { "epoch": 1.24, "learning_rate": 1.1355897009966776e-07, "loss": 0.2099, "step": 5988 }, { "epoch": 1.24, "learning_rate": 1.1352782392026578e-07, "loss": 0.1327, "step": 5989 }, { "epoch": 1.24, "learning_rate": 1.1349667774086379e-07, "loss": 0.1896, "step": 5990 }, { "epoch": 1.24, "learning_rate": 1.1346553156146178e-07, "loss": 0.2075, "step": 5991 }, { "epoch": 1.24, "learning_rate": 1.134343853820598e-07, "loss": 0.1067, "step": 5992 }, { "epoch": 1.24, "learning_rate": 1.134032392026578e-07, "loss": 0.198, "step": 5993 }, { "epoch": 1.24, "learning_rate": 1.1337209302325582e-07, "loss": 0.2005, "step": 5994 }, { "epoch": 1.24, "learning_rate": 1.133409468438538e-07, "loss": 0.1582, "step": 5995 }, { "epoch": 1.25, "learning_rate": 1.1330980066445182e-07, "loss": 0.1719, "step": 5996 }, { "epoch": 1.25, "learning_rate": 1.1327865448504984e-07, "loss": 0.1878, "step": 5997 }, { "epoch": 1.25, "learning_rate": 1.1324750830564783e-07, "loss": 0.1676, "step": 5998 }, { "epoch": 1.25, "learning_rate": 1.1321636212624584e-07, "loss": 0.182, "step": 5999 }, { "epoch": 1.25, "learning_rate": 1.1318521594684385e-07, "loss": 0.1836, "step": 6000 }, { "epoch": 1.25, "learning_rate": 1.1315406976744186e-07, "loss": 0.1897, "step": 6001 }, { "epoch": 1.25, "learning_rate": 1.1312292358803986e-07, "loss": 0.1318, "step": 6002 }, { "epoch": 1.25, "learning_rate": 1.1309177740863787e-07, "loss": 0.1587, "step": 6003 }, { "epoch": 1.25, "learning_rate": 1.1306063122923588e-07, "loss": 0.1979, "step": 6004 }, { "epoch": 1.25, "learning_rate": 1.1302948504983388e-07, "loss": 0.2212, "step": 6005 }, { "epoch": 1.25, "learning_rate": 1.1299833887043189e-07, "loss": 0.2087, "step": 6006 }, { "epoch": 1.25, "learning_rate": 1.1296719269102989e-07, "loss": 0.211, "step": 6007 }, { "epoch": 1.25, "learning_rate": 1.129360465116279e-07, "loss": 0.1226, "step": 6008 }, { "epoch": 1.25, "learning_rate": 1.1290490033222592e-07, "loss": 0.1872, "step": 6009 }, { "epoch": 1.25, "learning_rate": 1.1287375415282391e-07, "loss": 0.2216, "step": 6010 }, { "epoch": 1.25, "learning_rate": 1.1284260797342193e-07, "loss": 0.2701, "step": 6011 }, { "epoch": 1.25, "learning_rate": 1.1281146179401993e-07, "loss": 0.1849, "step": 6012 }, { "epoch": 1.25, "learning_rate": 1.1278031561461793e-07, "loss": 0.2012, "step": 6013 }, { "epoch": 1.25, "learning_rate": 1.1274916943521593e-07, "loss": 0.1938, "step": 6014 }, { "epoch": 1.25, "learning_rate": 1.1271802325581395e-07, "loss": 0.1959, "step": 6015 }, { "epoch": 1.25, "learning_rate": 1.1268687707641197e-07, "loss": 0.177, "step": 6016 }, { "epoch": 1.25, "learning_rate": 1.1265573089700996e-07, "loss": 0.1852, "step": 6017 }, { "epoch": 1.25, "learning_rate": 1.1262458471760797e-07, "loss": 0.184, "step": 6018 }, { "epoch": 1.25, "learning_rate": 1.1259343853820597e-07, "loss": 0.1687, "step": 6019 }, { "epoch": 1.25, "learning_rate": 1.1256229235880399e-07, "loss": 0.1692, "step": 6020 }, { "epoch": 1.25, "learning_rate": 1.1253114617940198e-07, "loss": 0.2, "step": 6021 }, { "epoch": 1.25, "learning_rate": 1.125e-07, "loss": 0.1626, "step": 6022 }, { "epoch": 1.25, "learning_rate": 1.1246885382059801e-07, "loss": 0.1726, "step": 6023 }, { "epoch": 1.25, "learning_rate": 1.12437707641196e-07, "loss": 0.2329, "step": 6024 }, { "epoch": 1.25, "learning_rate": 1.1240656146179402e-07, "loss": 0.1376, "step": 6025 }, { "epoch": 1.25, "learning_rate": 1.1237541528239202e-07, "loss": 0.199, "step": 6026 }, { "epoch": 1.25, "learning_rate": 1.1234426910299003e-07, "loss": 0.2383, "step": 6027 }, { "epoch": 1.25, "learning_rate": 1.1231312292358802e-07, "loss": 0.1829, "step": 6028 }, { "epoch": 1.25, "learning_rate": 1.1228197674418604e-07, "loss": 0.1675, "step": 6029 }, { "epoch": 1.25, "learning_rate": 1.1225083056478406e-07, "loss": 0.182, "step": 6030 }, { "epoch": 1.25, "learning_rate": 1.1221968438538206e-07, "loss": 0.202, "step": 6031 }, { "epoch": 1.25, "learning_rate": 1.1218853820598006e-07, "loss": 0.1425, "step": 6032 }, { "epoch": 1.25, "learning_rate": 1.1215739202657806e-07, "loss": 0.2627, "step": 6033 }, { "epoch": 1.25, "learning_rate": 1.1212624584717608e-07, "loss": 0.1262, "step": 6034 }, { "epoch": 1.25, "learning_rate": 1.1209509966777407e-07, "loss": 0.1871, "step": 6035 }, { "epoch": 1.25, "learning_rate": 1.1206395348837208e-07, "loss": 0.1618, "step": 6036 }, { "epoch": 1.25, "learning_rate": 1.120328073089701e-07, "loss": 0.1695, "step": 6037 }, { "epoch": 1.25, "learning_rate": 1.120016611295681e-07, "loss": 0.1869, "step": 6038 }, { "epoch": 1.25, "learning_rate": 1.119705149501661e-07, "loss": 0.182, "step": 6039 }, { "epoch": 1.25, "learning_rate": 1.1193936877076411e-07, "loss": 0.1685, "step": 6040 }, { "epoch": 1.25, "learning_rate": 1.1190822259136212e-07, "loss": 0.142, "step": 6041 }, { "epoch": 1.25, "learning_rate": 1.1187707641196013e-07, "loss": 0.1671, "step": 6042 }, { "epoch": 1.25, "learning_rate": 1.1184593023255813e-07, "loss": 0.1758, "step": 6043 }, { "epoch": 1.25, "learning_rate": 1.1181478405315614e-07, "loss": 0.1633, "step": 6044 }, { "epoch": 1.26, "learning_rate": 1.1178363787375415e-07, "loss": 0.1749, "step": 6045 }, { "epoch": 1.26, "learning_rate": 1.1175249169435216e-07, "loss": 0.1231, "step": 6046 }, { "epoch": 1.26, "learning_rate": 1.1172134551495015e-07, "loss": 0.2305, "step": 6047 }, { "epoch": 1.26, "learning_rate": 1.1169019933554817e-07, "loss": 0.1498, "step": 6048 }, { "epoch": 1.26, "learning_rate": 1.1165905315614618e-07, "loss": 0.2271, "step": 6049 }, { "epoch": 1.26, "learning_rate": 1.1162790697674417e-07, "loss": 0.1791, "step": 6050 }, { "epoch": 1.26, "learning_rate": 1.1159676079734219e-07, "loss": 0.1782, "step": 6051 }, { "epoch": 1.26, "learning_rate": 1.1156561461794019e-07, "loss": 0.1829, "step": 6052 }, { "epoch": 1.26, "learning_rate": 1.1153446843853821e-07, "loss": 0.1705, "step": 6053 }, { "epoch": 1.26, "learning_rate": 1.115033222591362e-07, "loss": 0.1409, "step": 6054 }, { "epoch": 1.26, "learning_rate": 1.1147217607973421e-07, "loss": 0.1597, "step": 6055 }, { "epoch": 1.26, "learning_rate": 1.1144102990033223e-07, "loss": 0.1934, "step": 6056 }, { "epoch": 1.26, "learning_rate": 1.1140988372093023e-07, "loss": 0.1621, "step": 6057 }, { "epoch": 1.26, "learning_rate": 1.1137873754152823e-07, "loss": 0.1717, "step": 6058 }, { "epoch": 1.26, "learning_rate": 1.1134759136212624e-07, "loss": 0.1598, "step": 6059 }, { "epoch": 1.26, "learning_rate": 1.1131644518272425e-07, "loss": 0.1598, "step": 6060 }, { "epoch": 1.26, "learning_rate": 1.1128529900332224e-07, "loss": 0.2007, "step": 6061 }, { "epoch": 1.26, "learning_rate": 1.1125415282392026e-07, "loss": 0.1868, "step": 6062 }, { "epoch": 1.26, "learning_rate": 1.1122300664451827e-07, "loss": 0.1695, "step": 6063 }, { "epoch": 1.26, "learning_rate": 1.1119186046511628e-07, "loss": 0.2615, "step": 6064 }, { "epoch": 1.26, "learning_rate": 1.1116071428571429e-07, "loss": 0.1514, "step": 6065 }, { "epoch": 1.26, "learning_rate": 1.1112956810631228e-07, "loss": 0.1755, "step": 6066 }, { "epoch": 1.26, "learning_rate": 1.110984219269103e-07, "loss": 0.1462, "step": 6067 }, { "epoch": 1.26, "learning_rate": 1.110672757475083e-07, "loss": 0.1752, "step": 6068 }, { "epoch": 1.26, "learning_rate": 1.110361295681063e-07, "loss": 0.2722, "step": 6069 }, { "epoch": 1.26, "learning_rate": 1.1100498338870432e-07, "loss": 0.1731, "step": 6070 }, { "epoch": 1.26, "learning_rate": 1.1097383720930232e-07, "loss": 0.2695, "step": 6071 }, { "epoch": 1.26, "learning_rate": 1.1094269102990034e-07, "loss": 0.2174, "step": 6072 }, { "epoch": 1.26, "learning_rate": 1.1091154485049833e-07, "loss": 0.2036, "step": 6073 }, { "epoch": 1.26, "learning_rate": 1.1088039867109634e-07, "loss": 0.2051, "step": 6074 }, { "epoch": 1.26, "learning_rate": 1.1084925249169434e-07, "loss": 0.1772, "step": 6075 }, { "epoch": 1.26, "learning_rate": 1.1081810631229236e-07, "loss": 0.1836, "step": 6076 }, { "epoch": 1.26, "learning_rate": 1.1078696013289036e-07, "loss": 0.1942, "step": 6077 }, { "epoch": 1.26, "learning_rate": 1.1075581395348836e-07, "loss": 0.1499, "step": 6078 }, { "epoch": 1.26, "learning_rate": 1.1072466777408638e-07, "loss": 0.2012, "step": 6079 }, { "epoch": 1.26, "learning_rate": 1.1069352159468437e-07, "loss": 0.1736, "step": 6080 }, { "epoch": 1.26, "learning_rate": 1.1066237541528239e-07, "loss": 0.2162, "step": 6081 }, { "epoch": 1.26, "learning_rate": 1.1063122923588039e-07, "loss": 0.1325, "step": 6082 }, { "epoch": 1.26, "learning_rate": 1.106000830564784e-07, "loss": 0.149, "step": 6083 }, { "epoch": 1.26, "learning_rate": 1.1056893687707641e-07, "loss": 0.191, "step": 6084 }, { "epoch": 1.26, "learning_rate": 1.1053779069767441e-07, "loss": 0.1574, "step": 6085 }, { "epoch": 1.26, "learning_rate": 1.1050664451827243e-07, "loss": 0.2035, "step": 6086 }, { "epoch": 1.26, "learning_rate": 1.1047549833887041e-07, "loss": 0.1899, "step": 6087 }, { "epoch": 1.26, "learning_rate": 1.1044435215946843e-07, "loss": 0.22, "step": 6088 }, { "epoch": 1.26, "learning_rate": 1.1041320598006645e-07, "loss": 0.203, "step": 6089 }, { "epoch": 1.26, "learning_rate": 1.1038205980066445e-07, "loss": 0.1897, "step": 6090 }, { "epoch": 1.26, "learning_rate": 1.1035091362126246e-07, "loss": 0.1947, "step": 6091 }, { "epoch": 1.26, "learning_rate": 1.1031976744186045e-07, "loss": 0.1385, "step": 6092 }, { "epoch": 1.27, "learning_rate": 1.1028862126245847e-07, "loss": 0.1768, "step": 6093 }, { "epoch": 1.27, "learning_rate": 1.1025747508305647e-07, "loss": 0.1977, "step": 6094 }, { "epoch": 1.27, "learning_rate": 1.1022632890365448e-07, "loss": 0.227, "step": 6095 }, { "epoch": 1.27, "learning_rate": 1.1019518272425249e-07, "loss": 0.1459, "step": 6096 }, { "epoch": 1.27, "learning_rate": 1.101640365448505e-07, "loss": 0.1565, "step": 6097 }, { "epoch": 1.27, "learning_rate": 1.1013289036544851e-07, "loss": 0.2037, "step": 6098 }, { "epoch": 1.27, "learning_rate": 1.101017441860465e-07, "loss": 0.2391, "step": 6099 }, { "epoch": 1.27, "learning_rate": 1.1007059800664451e-07, "loss": 0.1755, "step": 6100 }, { "epoch": 1.27, "learning_rate": 1.1003945182724252e-07, "loss": 0.1558, "step": 6101 }, { "epoch": 1.27, "learning_rate": 1.1000830564784053e-07, "loss": 0.1788, "step": 6102 }, { "epoch": 1.27, "learning_rate": 1.0997715946843854e-07, "loss": 0.1994, "step": 6103 }, { "epoch": 1.27, "learning_rate": 1.0994601328903654e-07, "loss": 0.1881, "step": 6104 }, { "epoch": 1.27, "learning_rate": 1.0991486710963455e-07, "loss": 0.1811, "step": 6105 }, { "epoch": 1.27, "learning_rate": 1.0988372093023254e-07, "loss": 0.1196, "step": 6106 }, { "epoch": 1.27, "learning_rate": 1.0985257475083056e-07, "loss": 0.2087, "step": 6107 }, { "epoch": 1.27, "learning_rate": 1.0982142857142856e-07, "loss": 0.1388, "step": 6108 }, { "epoch": 1.27, "learning_rate": 1.0979028239202658e-07, "loss": 0.196, "step": 6109 }, { "epoch": 1.27, "learning_rate": 1.0975913621262459e-07, "loss": 0.1939, "step": 6110 }, { "epoch": 1.27, "learning_rate": 1.0972799003322258e-07, "loss": 0.1582, "step": 6111 }, { "epoch": 1.27, "learning_rate": 1.096968438538206e-07, "loss": 0.1695, "step": 6112 }, { "epoch": 1.27, "learning_rate": 1.096656976744186e-07, "loss": 0.1746, "step": 6113 }, { "epoch": 1.27, "learning_rate": 1.096345514950166e-07, "loss": 0.1884, "step": 6114 }, { "epoch": 1.27, "learning_rate": 1.096034053156146e-07, "loss": 0.1924, "step": 6115 }, { "epoch": 1.27, "learning_rate": 1.0957225913621262e-07, "loss": 0.163, "step": 6116 }, { "epoch": 1.27, "learning_rate": 1.0954111295681064e-07, "loss": 0.1528, "step": 6117 }, { "epoch": 1.27, "learning_rate": 1.0950996677740863e-07, "loss": 0.2411, "step": 6118 }, { "epoch": 1.27, "learning_rate": 1.0947882059800664e-07, "loss": 0.2161, "step": 6119 }, { "epoch": 1.27, "learning_rate": 1.0944767441860465e-07, "loss": 0.1747, "step": 6120 }, { "epoch": 1.27, "learning_rate": 1.0941652823920265e-07, "loss": 0.148, "step": 6121 }, { "epoch": 1.27, "learning_rate": 1.0938538205980065e-07, "loss": 0.1596, "step": 6122 }, { "epoch": 1.27, "learning_rate": 1.0935423588039867e-07, "loss": 0.1954, "step": 6123 }, { "epoch": 1.27, "learning_rate": 1.0932308970099668e-07, "loss": 0.1831, "step": 6124 }, { "epoch": 1.27, "learning_rate": 1.0929194352159467e-07, "loss": 0.1514, "step": 6125 }, { "epoch": 1.27, "learning_rate": 1.0926079734219269e-07, "loss": 0.1414, "step": 6126 }, { "epoch": 1.27, "learning_rate": 1.0922965116279069e-07, "loss": 0.2011, "step": 6127 }, { "epoch": 1.27, "learning_rate": 1.091985049833887e-07, "loss": 0.1919, "step": 6128 }, { "epoch": 1.27, "learning_rate": 1.091673588039867e-07, "loss": 0.1642, "step": 6129 }, { "epoch": 1.27, "learning_rate": 1.0913621262458471e-07, "loss": 0.1592, "step": 6130 }, { "epoch": 1.27, "learning_rate": 1.0910506644518273e-07, "loss": 0.2196, "step": 6131 }, { "epoch": 1.27, "learning_rate": 1.0907392026578072e-07, "loss": 0.2251, "step": 6132 }, { "epoch": 1.27, "learning_rate": 1.0904277408637873e-07, "loss": 0.2087, "step": 6133 }, { "epoch": 1.27, "learning_rate": 1.0901162790697673e-07, "loss": 0.1547, "step": 6134 }, { "epoch": 1.27, "learning_rate": 1.0898048172757475e-07, "loss": 0.1886, "step": 6135 }, { "epoch": 1.27, "learning_rate": 1.0894933554817277e-07, "loss": 0.1687, "step": 6136 }, { "epoch": 1.27, "learning_rate": 1.0891818936877076e-07, "loss": 0.1636, "step": 6137 }, { "epoch": 1.27, "learning_rate": 1.0888704318936877e-07, "loss": 0.1324, "step": 6138 }, { "epoch": 1.27, "learning_rate": 1.0885589700996677e-07, "loss": 0.1736, "step": 6139 }, { "epoch": 1.27, "learning_rate": 1.0882475083056478e-07, "loss": 0.1814, "step": 6140 }, { "epoch": 1.28, "learning_rate": 1.0879360465116278e-07, "loss": 0.2506, "step": 6141 }, { "epoch": 1.28, "learning_rate": 1.087624584717608e-07, "loss": 0.1717, "step": 6142 }, { "epoch": 1.28, "learning_rate": 1.0873131229235881e-07, "loss": 0.1443, "step": 6143 }, { "epoch": 1.28, "learning_rate": 1.087001661129568e-07, "loss": 0.2012, "step": 6144 }, { "epoch": 1.28, "learning_rate": 1.0866901993355482e-07, "loss": 0.1502, "step": 6145 }, { "epoch": 1.28, "learning_rate": 1.0863787375415282e-07, "loss": 0.2214, "step": 6146 }, { "epoch": 1.28, "learning_rate": 1.0860672757475083e-07, "loss": 0.1372, "step": 6147 }, { "epoch": 1.28, "learning_rate": 1.0857558139534882e-07, "loss": 0.2014, "step": 6148 }, { "epoch": 1.28, "learning_rate": 1.0854443521594684e-07, "loss": 0.1598, "step": 6149 }, { "epoch": 1.28, "learning_rate": 1.0851328903654486e-07, "loss": 0.1528, "step": 6150 }, { "epoch": 1.28, "learning_rate": 1.0848214285714285e-07, "loss": 0.204, "step": 6151 }, { "epoch": 1.28, "learning_rate": 1.0845099667774086e-07, "loss": 0.1821, "step": 6152 }, { "epoch": 1.28, "learning_rate": 1.0841985049833886e-07, "loss": 0.1941, "step": 6153 }, { "epoch": 1.28, "learning_rate": 1.0838870431893688e-07, "loss": 0.2414, "step": 6154 }, { "epoch": 1.28, "learning_rate": 1.0835755813953487e-07, "loss": 0.1941, "step": 6155 }, { "epoch": 1.28, "learning_rate": 1.0832641196013288e-07, "loss": 0.2017, "step": 6156 }, { "epoch": 1.28, "learning_rate": 1.082952657807309e-07, "loss": 0.2485, "step": 6157 }, { "epoch": 1.28, "learning_rate": 1.0826411960132889e-07, "loss": 0.2095, "step": 6158 }, { "epoch": 1.28, "learning_rate": 1.082329734219269e-07, "loss": 0.2112, "step": 6159 }, { "epoch": 1.28, "learning_rate": 1.0820182724252491e-07, "loss": 0.1736, "step": 6160 }, { "epoch": 1.28, "learning_rate": 1.0817068106312292e-07, "loss": 0.1736, "step": 6161 }, { "epoch": 1.28, "learning_rate": 1.0813953488372091e-07, "loss": 0.1807, "step": 6162 }, { "epoch": 1.28, "learning_rate": 1.0810838870431893e-07, "loss": 0.1755, "step": 6163 }, { "epoch": 1.28, "learning_rate": 1.0807724252491694e-07, "loss": 0.1849, "step": 6164 }, { "epoch": 1.28, "learning_rate": 1.0804609634551495e-07, "loss": 0.2086, "step": 6165 }, { "epoch": 1.28, "learning_rate": 1.0801495016611295e-07, "loss": 0.2361, "step": 6166 }, { "epoch": 1.28, "learning_rate": 1.0798380398671095e-07, "loss": 0.1226, "step": 6167 }, { "epoch": 1.28, "learning_rate": 1.0795265780730897e-07, "loss": 0.2417, "step": 6168 }, { "epoch": 1.28, "learning_rate": 1.0792151162790696e-07, "loss": 0.1429, "step": 6169 }, { "epoch": 1.28, "learning_rate": 1.0789036544850497e-07, "loss": 0.2189, "step": 6170 }, { "epoch": 1.28, "learning_rate": 1.0785921926910299e-07, "loss": 0.1653, "step": 6171 }, { "epoch": 1.28, "learning_rate": 1.0782807308970099e-07, "loss": 0.1893, "step": 6172 }, { "epoch": 1.28, "learning_rate": 1.0779692691029901e-07, "loss": 0.1777, "step": 6173 }, { "epoch": 1.28, "learning_rate": 1.07765780730897e-07, "loss": 0.1325, "step": 6174 }, { "epoch": 1.28, "learning_rate": 1.0773463455149501e-07, "loss": 0.1939, "step": 6175 }, { "epoch": 1.28, "learning_rate": 1.0770348837209303e-07, "loss": 0.1579, "step": 6176 }, { "epoch": 1.28, "learning_rate": 1.0767234219269102e-07, "loss": 0.1357, "step": 6177 }, { "epoch": 1.28, "learning_rate": 1.0764119601328903e-07, "loss": 0.2378, "step": 6178 }, { "epoch": 1.28, "learning_rate": 1.0761004983388704e-07, "loss": 0.1847, "step": 6179 }, { "epoch": 1.28, "learning_rate": 1.0757890365448505e-07, "loss": 0.2103, "step": 6180 }, { "epoch": 1.28, "learning_rate": 1.0754775747508304e-07, "loss": 0.1395, "step": 6181 }, { "epoch": 1.28, "learning_rate": 1.0751661129568106e-07, "loss": 0.2088, "step": 6182 }, { "epoch": 1.28, "learning_rate": 1.0748546511627907e-07, "loss": 0.1722, "step": 6183 }, { "epoch": 1.28, "learning_rate": 1.0745431893687708e-07, "loss": 0.1977, "step": 6184 }, { "epoch": 1.28, "learning_rate": 1.0742317275747508e-07, "loss": 0.172, "step": 6185 }, { "epoch": 1.28, "learning_rate": 1.0739202657807308e-07, "loss": 0.1879, "step": 6186 }, { "epoch": 1.28, "learning_rate": 1.073608803986711e-07, "loss": 0.2254, "step": 6187 }, { "epoch": 1.28, "learning_rate": 1.0732973421926909e-07, "loss": 0.1346, "step": 6188 }, { "epoch": 1.29, "learning_rate": 1.072985880398671e-07, "loss": 0.166, "step": 6189 }, { "epoch": 1.29, "learning_rate": 1.0726744186046512e-07, "loss": 0.1421, "step": 6190 }, { "epoch": 1.29, "learning_rate": 1.0723629568106312e-07, "loss": 0.1685, "step": 6191 }, { "epoch": 1.29, "learning_rate": 1.0720514950166112e-07, "loss": 0.1917, "step": 6192 }, { "epoch": 1.29, "learning_rate": 1.0717400332225913e-07, "loss": 0.1928, "step": 6193 }, { "epoch": 1.29, "learning_rate": 1.0714285714285714e-07, "loss": 0.1209, "step": 6194 }, { "epoch": 1.29, "learning_rate": 1.0711171096345513e-07, "loss": 0.1643, "step": 6195 }, { "epoch": 1.29, "learning_rate": 1.0708056478405315e-07, "loss": 0.166, "step": 6196 }, { "epoch": 1.29, "learning_rate": 1.0704941860465116e-07, "loss": 0.1249, "step": 6197 }, { "epoch": 1.29, "learning_rate": 1.0701827242524917e-07, "loss": 0.2151, "step": 6198 }, { "epoch": 1.29, "learning_rate": 1.0698712624584718e-07, "loss": 0.2036, "step": 6199 }, { "epoch": 1.29, "learning_rate": 1.0695598006644517e-07, "loss": 0.1815, "step": 6200 }, { "epoch": 1.29, "learning_rate": 1.0692483388704319e-07, "loss": 0.2029, "step": 6201 }, { "epoch": 1.29, "learning_rate": 1.0689368770764119e-07, "loss": 0.1478, "step": 6202 }, { "epoch": 1.29, "learning_rate": 1.0686254152823919e-07, "loss": 0.1965, "step": 6203 }, { "epoch": 1.29, "learning_rate": 1.0683139534883721e-07, "loss": 0.1568, "step": 6204 }, { "epoch": 1.29, "learning_rate": 1.0680024916943521e-07, "loss": 0.189, "step": 6205 }, { "epoch": 1.29, "learning_rate": 1.0676910299003323e-07, "loss": 0.2362, "step": 6206 }, { "epoch": 1.29, "learning_rate": 1.0673795681063122e-07, "loss": 0.1837, "step": 6207 }, { "epoch": 1.29, "learning_rate": 1.0670681063122923e-07, "loss": 0.1502, "step": 6208 }, { "epoch": 1.29, "learning_rate": 1.0667566445182723e-07, "loss": 0.2014, "step": 6209 }, { "epoch": 1.29, "learning_rate": 1.0664451827242525e-07, "loss": 0.1515, "step": 6210 }, { "epoch": 1.29, "learning_rate": 1.0661337209302325e-07, "loss": 0.195, "step": 6211 }, { "epoch": 1.29, "learning_rate": 1.0658222591362125e-07, "loss": 0.1821, "step": 6212 }, { "epoch": 1.29, "learning_rate": 1.0655107973421927e-07, "loss": 0.1403, "step": 6213 }, { "epoch": 1.29, "learning_rate": 1.0651993355481726e-07, "loss": 0.2168, "step": 6214 }, { "epoch": 1.29, "learning_rate": 1.0648878737541528e-07, "loss": 0.1925, "step": 6215 }, { "epoch": 1.29, "learning_rate": 1.0645764119601329e-07, "loss": 0.1585, "step": 6216 }, { "epoch": 1.29, "learning_rate": 1.064264950166113e-07, "loss": 0.2149, "step": 6217 }, { "epoch": 1.29, "learning_rate": 1.0639534883720931e-07, "loss": 0.185, "step": 6218 }, { "epoch": 1.29, "learning_rate": 1.063642026578073e-07, "loss": 0.1686, "step": 6219 }, { "epoch": 1.29, "learning_rate": 1.0633305647840532e-07, "loss": 0.2268, "step": 6220 }, { "epoch": 1.29, "learning_rate": 1.0630191029900332e-07, "loss": 0.1699, "step": 6221 }, { "epoch": 1.29, "learning_rate": 1.0627076411960132e-07, "loss": 0.1341, "step": 6222 }, { "epoch": 1.29, "learning_rate": 1.0623961794019934e-07, "loss": 0.145, "step": 6223 }, { "epoch": 1.29, "learning_rate": 1.0620847176079734e-07, "loss": 0.1818, "step": 6224 }, { "epoch": 1.29, "learning_rate": 1.0617732558139535e-07, "loss": 0.1478, "step": 6225 }, { "epoch": 1.29, "learning_rate": 1.0614617940199334e-07, "loss": 0.1466, "step": 6226 }, { "epoch": 1.29, "learning_rate": 1.0611503322259136e-07, "loss": 0.2016, "step": 6227 }, { "epoch": 1.29, "learning_rate": 1.0608388704318936e-07, "loss": 0.1862, "step": 6228 }, { "epoch": 1.29, "learning_rate": 1.0605274086378736e-07, "loss": 0.137, "step": 6229 }, { "epoch": 1.29, "learning_rate": 1.0602159468438538e-07, "loss": 0.1793, "step": 6230 }, { "epoch": 1.29, "learning_rate": 1.0599044850498338e-07, "loss": 0.1584, "step": 6231 }, { "epoch": 1.29, "learning_rate": 1.059593023255814e-07, "loss": 0.1522, "step": 6232 }, { "epoch": 1.29, "learning_rate": 1.0592815614617939e-07, "loss": 0.1162, "step": 6233 }, { "epoch": 1.29, "learning_rate": 1.058970099667774e-07, "loss": 0.156, "step": 6234 }, { "epoch": 1.29, "learning_rate": 1.0586586378737541e-07, "loss": 0.1443, "step": 6235 }, { "epoch": 1.29, "learning_rate": 1.0583471760797342e-07, "loss": 0.1903, "step": 6236 }, { "epoch": 1.3, "learning_rate": 1.0580357142857143e-07, "loss": 0.1839, "step": 6237 }, { "epoch": 1.3, "learning_rate": 1.0577242524916943e-07, "loss": 0.2138, "step": 6238 }, { "epoch": 1.3, "learning_rate": 1.0574127906976744e-07, "loss": 0.1842, "step": 6239 }, { "epoch": 1.3, "learning_rate": 1.0571013289036543e-07, "loss": 0.1425, "step": 6240 }, { "epoch": 1.3, "learning_rate": 1.0567898671096345e-07, "loss": 0.178, "step": 6241 }, { "epoch": 1.3, "learning_rate": 1.0564784053156145e-07, "loss": 0.182, "step": 6242 }, { "epoch": 1.3, "learning_rate": 1.0561669435215947e-07, "loss": 0.1328, "step": 6243 }, { "epoch": 1.3, "learning_rate": 1.0558554817275748e-07, "loss": 0.1799, "step": 6244 }, { "epoch": 1.3, "learning_rate": 1.0555440199335547e-07, "loss": 0.1332, "step": 6245 }, { "epoch": 1.3, "learning_rate": 1.0552325581395349e-07, "loss": 0.1855, "step": 6246 }, { "epoch": 1.3, "learning_rate": 1.0549210963455149e-07, "loss": 0.1742, "step": 6247 }, { "epoch": 1.3, "learning_rate": 1.054609634551495e-07, "loss": 0.1656, "step": 6248 }, { "epoch": 1.3, "learning_rate": 1.054298172757475e-07, "loss": 0.1777, "step": 6249 }, { "epoch": 1.3, "learning_rate": 1.0539867109634551e-07, "loss": 0.2327, "step": 6250 }, { "epoch": 1.3, "learning_rate": 1.0536752491694353e-07, "loss": 0.153, "step": 6251 }, { "epoch": 1.3, "learning_rate": 1.0533637873754152e-07, "loss": 0.2055, "step": 6252 }, { "epoch": 1.3, "learning_rate": 1.0530523255813953e-07, "loss": 0.1917, "step": 6253 }, { "epoch": 1.3, "learning_rate": 1.0527408637873754e-07, "loss": 0.167, "step": 6254 }, { "epoch": 1.3, "learning_rate": 1.0524294019933555e-07, "loss": 0.2044, "step": 6255 }, { "epoch": 1.3, "learning_rate": 1.0521179401993354e-07, "loss": 0.232, "step": 6256 }, { "epoch": 1.3, "learning_rate": 1.0518064784053156e-07, "loss": 0.2277, "step": 6257 }, { "epoch": 1.3, "learning_rate": 1.0514950166112957e-07, "loss": 0.1882, "step": 6258 }, { "epoch": 1.3, "learning_rate": 1.0511835548172756e-07, "loss": 0.172, "step": 6259 }, { "epoch": 1.3, "learning_rate": 1.0508720930232558e-07, "loss": 0.2347, "step": 6260 }, { "epoch": 1.3, "learning_rate": 1.0505606312292358e-07, "loss": 0.1982, "step": 6261 }, { "epoch": 1.3, "learning_rate": 1.050249169435216e-07, "loss": 0.1542, "step": 6262 }, { "epoch": 1.3, "learning_rate": 1.049937707641196e-07, "loss": 0.1552, "step": 6263 }, { "epoch": 1.3, "learning_rate": 1.049626245847176e-07, "loss": 0.1813, "step": 6264 }, { "epoch": 1.3, "learning_rate": 1.0493147840531562e-07, "loss": 0.1259, "step": 6265 }, { "epoch": 1.3, "learning_rate": 1.049003322259136e-07, "loss": 0.1749, "step": 6266 }, { "epoch": 1.3, "learning_rate": 1.0486918604651162e-07, "loss": 0.2059, "step": 6267 }, { "epoch": 1.3, "learning_rate": 1.0483803986710962e-07, "loss": 0.2361, "step": 6268 }, { "epoch": 1.3, "learning_rate": 1.0480689368770764e-07, "loss": 0.2194, "step": 6269 }, { "epoch": 1.3, "learning_rate": 1.0477574750830566e-07, "loss": 0.1605, "step": 6270 }, { "epoch": 1.3, "learning_rate": 1.0474460132890365e-07, "loss": 0.2057, "step": 6271 }, { "epoch": 1.3, "learning_rate": 1.0471345514950166e-07, "loss": 0.1496, "step": 6272 }, { "epoch": 1.3, "learning_rate": 1.0468230897009966e-07, "loss": 0.2284, "step": 6273 }, { "epoch": 1.3, "learning_rate": 1.0465116279069767e-07, "loss": 0.1839, "step": 6274 }, { "epoch": 1.3, "learning_rate": 1.0462001661129567e-07, "loss": 0.1563, "step": 6275 }, { "epoch": 1.3, "learning_rate": 1.0458887043189369e-07, "loss": 0.1817, "step": 6276 }, { "epoch": 1.3, "learning_rate": 1.045577242524917e-07, "loss": 0.2294, "step": 6277 }, { "epoch": 1.3, "learning_rate": 1.0452657807308969e-07, "loss": 0.1653, "step": 6278 }, { "epoch": 1.3, "learning_rate": 1.044954318936877e-07, "loss": 0.2112, "step": 6279 }, { "epoch": 1.3, "learning_rate": 1.0446428571428571e-07, "loss": 0.1836, "step": 6280 }, { "epoch": 1.3, "learning_rate": 1.0443313953488372e-07, "loss": 0.1127, "step": 6281 }, { "epoch": 1.3, "learning_rate": 1.0440199335548171e-07, "loss": 0.1318, "step": 6282 }, { "epoch": 1.3, "learning_rate": 1.0437084717607973e-07, "loss": 0.1376, "step": 6283 }, { "epoch": 1.3, "learning_rate": 1.0433970099667775e-07, "loss": 0.1654, "step": 6284 }, { "epoch": 1.31, "learning_rate": 1.0430855481727573e-07, "loss": 0.162, "step": 6285 }, { "epoch": 1.31, "learning_rate": 1.0427740863787375e-07, "loss": 0.1614, "step": 6286 }, { "epoch": 1.31, "learning_rate": 1.0424626245847175e-07, "loss": 0.2991, "step": 6287 }, { "epoch": 1.31, "learning_rate": 1.0421511627906977e-07, "loss": 0.1906, "step": 6288 }, { "epoch": 1.31, "learning_rate": 1.0418397009966776e-07, "loss": 0.162, "step": 6289 }, { "epoch": 1.31, "learning_rate": 1.0415282392026577e-07, "loss": 0.1883, "step": 6290 }, { "epoch": 1.31, "learning_rate": 1.0412167774086379e-07, "loss": 0.2157, "step": 6291 }, { "epoch": 1.31, "learning_rate": 1.0409053156146179e-07, "loss": 0.207, "step": 6292 }, { "epoch": 1.31, "learning_rate": 1.040593853820598e-07, "loss": 0.1887, "step": 6293 }, { "epoch": 1.31, "learning_rate": 1.040282392026578e-07, "loss": 0.1003, "step": 6294 }, { "epoch": 1.31, "learning_rate": 1.0399709302325581e-07, "loss": 0.1412, "step": 6295 }, { "epoch": 1.31, "learning_rate": 1.039659468438538e-07, "loss": 0.197, "step": 6296 }, { "epoch": 1.31, "learning_rate": 1.0393480066445182e-07, "loss": 0.1467, "step": 6297 }, { "epoch": 1.31, "learning_rate": 1.0390365448504983e-07, "loss": 0.1519, "step": 6298 }, { "epoch": 1.31, "learning_rate": 1.0387250830564784e-07, "loss": 0.1186, "step": 6299 }, { "epoch": 1.31, "learning_rate": 1.0384136212624584e-07, "loss": 0.2291, "step": 6300 }, { "epoch": 1.31, "learning_rate": 1.0381021594684384e-07, "loss": 0.1556, "step": 6301 }, { "epoch": 1.31, "learning_rate": 1.0377906976744186e-07, "loss": 0.224, "step": 6302 }, { "epoch": 1.31, "learning_rate": 1.0374792358803987e-07, "loss": 0.1601, "step": 6303 }, { "epoch": 1.31, "learning_rate": 1.0371677740863786e-07, "loss": 0.2249, "step": 6304 }, { "epoch": 1.31, "learning_rate": 1.0368563122923588e-07, "loss": 0.1502, "step": 6305 }, { "epoch": 1.31, "learning_rate": 1.0365448504983388e-07, "loss": 0.149, "step": 6306 }, { "epoch": 1.31, "learning_rate": 1.036233388704319e-07, "loss": 0.1629, "step": 6307 }, { "epoch": 1.31, "learning_rate": 1.0359219269102989e-07, "loss": 0.1893, "step": 6308 }, { "epoch": 1.31, "learning_rate": 1.035610465116279e-07, "loss": 0.1536, "step": 6309 }, { "epoch": 1.31, "learning_rate": 1.0352990033222592e-07, "loss": 0.1903, "step": 6310 }, { "epoch": 1.31, "learning_rate": 1.0349875415282391e-07, "loss": 0.1958, "step": 6311 }, { "epoch": 1.31, "learning_rate": 1.0346760797342192e-07, "loss": 0.2257, "step": 6312 }, { "epoch": 1.31, "learning_rate": 1.0343646179401993e-07, "loss": 0.1732, "step": 6313 }, { "epoch": 1.31, "learning_rate": 1.0340531561461794e-07, "loss": 0.2047, "step": 6314 }, { "epoch": 1.31, "learning_rate": 1.0337416943521593e-07, "loss": 0.1493, "step": 6315 }, { "epoch": 1.31, "learning_rate": 1.0334302325581395e-07, "loss": 0.1824, "step": 6316 }, { "epoch": 1.31, "learning_rate": 1.0331187707641196e-07, "loss": 0.1769, "step": 6317 }, { "epoch": 1.31, "learning_rate": 1.0328073089700997e-07, "loss": 0.1999, "step": 6318 }, { "epoch": 1.31, "learning_rate": 1.0324958471760797e-07, "loss": 0.1532, "step": 6319 }, { "epoch": 1.31, "learning_rate": 1.0321843853820597e-07, "loss": 0.1575, "step": 6320 }, { "epoch": 1.31, "learning_rate": 1.0318729235880399e-07, "loss": 0.1313, "step": 6321 }, { "epoch": 1.31, "learning_rate": 1.0315614617940198e-07, "loss": 0.1632, "step": 6322 }, { "epoch": 1.31, "learning_rate": 1.0312499999999999e-07, "loss": 0.1645, "step": 6323 }, { "epoch": 1.31, "learning_rate": 1.0309385382059801e-07, "loss": 0.2082, "step": 6324 }, { "epoch": 1.31, "learning_rate": 1.0306270764119601e-07, "loss": 0.1852, "step": 6325 }, { "epoch": 1.31, "learning_rate": 1.0303156146179403e-07, "loss": 0.1545, "step": 6326 }, { "epoch": 1.31, "learning_rate": 1.0300041528239202e-07, "loss": 0.1804, "step": 6327 }, { "epoch": 1.31, "learning_rate": 1.0296926910299003e-07, "loss": 0.1357, "step": 6328 }, { "epoch": 1.31, "learning_rate": 1.0293812292358803e-07, "loss": 0.1535, "step": 6329 }, { "epoch": 1.31, "learning_rate": 1.0290697674418604e-07, "loss": 0.1732, "step": 6330 }, { "epoch": 1.31, "learning_rate": 1.0287583056478405e-07, "loss": 0.1719, "step": 6331 }, { "epoch": 1.31, "learning_rate": 1.0284468438538206e-07, "loss": 0.2612, "step": 6332 }, { "epoch": 1.31, "learning_rate": 1.0281353820598007e-07, "loss": 0.1672, "step": 6333 }, { "epoch": 1.32, "learning_rate": 1.0278239202657806e-07, "loss": 0.1979, "step": 6334 }, { "epoch": 1.32, "learning_rate": 1.0275124584717608e-07, "loss": 0.1679, "step": 6335 }, { "epoch": 1.32, "learning_rate": 1.0272009966777408e-07, "loss": 0.2084, "step": 6336 }, { "epoch": 1.32, "learning_rate": 1.0268895348837208e-07, "loss": 0.1568, "step": 6337 }, { "epoch": 1.32, "learning_rate": 1.026578073089701e-07, "loss": 0.1568, "step": 6338 }, { "epoch": 1.32, "learning_rate": 1.026266611295681e-07, "loss": 0.1944, "step": 6339 }, { "epoch": 1.32, "learning_rate": 1.0259551495016612e-07, "loss": 0.21, "step": 6340 }, { "epoch": 1.32, "learning_rate": 1.025643687707641e-07, "loss": 0.2239, "step": 6341 }, { "epoch": 1.32, "learning_rate": 1.0253322259136212e-07, "loss": 0.1513, "step": 6342 }, { "epoch": 1.32, "learning_rate": 1.0250207641196012e-07, "loss": 0.2563, "step": 6343 }, { "epoch": 1.32, "learning_rate": 1.0247093023255814e-07, "loss": 0.1683, "step": 6344 }, { "epoch": 1.32, "learning_rate": 1.0243978405315614e-07, "loss": 0.1708, "step": 6345 }, { "epoch": 1.32, "learning_rate": 1.0240863787375414e-07, "loss": 0.2066, "step": 6346 }, { "epoch": 1.32, "learning_rate": 1.0237749169435216e-07, "loss": 0.1895, "step": 6347 }, { "epoch": 1.32, "learning_rate": 1.0234634551495015e-07, "loss": 0.2125, "step": 6348 }, { "epoch": 1.32, "learning_rate": 1.0231519933554817e-07, "loss": 0.1998, "step": 6349 }, { "epoch": 1.32, "learning_rate": 1.0228405315614618e-07, "loss": 0.1565, "step": 6350 }, { "epoch": 1.32, "learning_rate": 1.0225290697674418e-07, "loss": 0.138, "step": 6351 }, { "epoch": 1.32, "learning_rate": 1.022217607973422e-07, "loss": 0.1621, "step": 6352 }, { "epoch": 1.32, "learning_rate": 1.0219061461794019e-07, "loss": 0.1878, "step": 6353 }, { "epoch": 1.32, "learning_rate": 1.021594684385382e-07, "loss": 0.1686, "step": 6354 }, { "epoch": 1.32, "learning_rate": 1.0212832225913621e-07, "loss": 0.1886, "step": 6355 }, { "epoch": 1.32, "learning_rate": 1.0209717607973421e-07, "loss": 0.2031, "step": 6356 }, { "epoch": 1.32, "learning_rate": 1.0206602990033223e-07, "loss": 0.1824, "step": 6357 }, { "epoch": 1.32, "learning_rate": 1.0203488372093023e-07, "loss": 0.208, "step": 6358 }, { "epoch": 1.32, "learning_rate": 1.0200373754152824e-07, "loss": 0.1166, "step": 6359 }, { "epoch": 1.32, "learning_rate": 1.0197259136212623e-07, "loss": 0.2283, "step": 6360 }, { "epoch": 1.32, "learning_rate": 1.0194144518272425e-07, "loss": 0.1208, "step": 6361 }, { "epoch": 1.32, "learning_rate": 1.0191029900332225e-07, "loss": 0.2036, "step": 6362 }, { "epoch": 1.32, "learning_rate": 1.0187915282392027e-07, "loss": 0.1359, "step": 6363 }, { "epoch": 1.32, "learning_rate": 1.0184800664451827e-07, "loss": 0.213, "step": 6364 }, { "epoch": 1.32, "learning_rate": 1.0181686046511627e-07, "loss": 0.1869, "step": 6365 }, { "epoch": 1.32, "learning_rate": 1.0178571428571429e-07, "loss": 0.2186, "step": 6366 }, { "epoch": 1.32, "learning_rate": 1.0175456810631228e-07, "loss": 0.1527, "step": 6367 }, { "epoch": 1.32, "learning_rate": 1.017234219269103e-07, "loss": 0.2044, "step": 6368 }, { "epoch": 1.32, "learning_rate": 1.016922757475083e-07, "loss": 0.2185, "step": 6369 }, { "epoch": 1.32, "learning_rate": 1.0166112956810631e-07, "loss": 0.1785, "step": 6370 }, { "epoch": 1.32, "learning_rate": 1.0162998338870431e-07, "loss": 0.2001, "step": 6371 }, { "epoch": 1.32, "learning_rate": 1.0159883720930232e-07, "loss": 0.1933, "step": 6372 }, { "epoch": 1.32, "learning_rate": 1.0156769102990033e-07, "loss": 0.1851, "step": 6373 }, { "epoch": 1.32, "learning_rate": 1.0153654485049832e-07, "loss": 0.2283, "step": 6374 }, { "epoch": 1.32, "learning_rate": 1.0150539867109634e-07, "loss": 0.16, "step": 6375 }, { "epoch": 1.32, "learning_rate": 1.0147425249169434e-07, "loss": 0.143, "step": 6376 }, { "epoch": 1.32, "learning_rate": 1.0144310631229236e-07, "loss": 0.1446, "step": 6377 }, { "epoch": 1.32, "learning_rate": 1.0141196013289037e-07, "loss": 0.2068, "step": 6378 }, { "epoch": 1.32, "learning_rate": 1.0138081395348836e-07, "loss": 0.2105, "step": 6379 }, { "epoch": 1.32, "learning_rate": 1.0134966777408638e-07, "loss": 0.1518, "step": 6380 }, { "epoch": 1.32, "learning_rate": 1.0131852159468438e-07, "loss": 0.1524, "step": 6381 }, { "epoch": 1.33, "learning_rate": 1.0128737541528238e-07, "loss": 0.2402, "step": 6382 }, { "epoch": 1.33, "learning_rate": 1.0125622923588039e-07, "loss": 0.1481, "step": 6383 }, { "epoch": 1.33, "learning_rate": 1.012250830564784e-07, "loss": 0.2267, "step": 6384 }, { "epoch": 1.33, "learning_rate": 1.0119393687707642e-07, "loss": 0.2023, "step": 6385 }, { "epoch": 1.33, "learning_rate": 1.0116279069767441e-07, "loss": 0.2024, "step": 6386 }, { "epoch": 1.33, "learning_rate": 1.0113164451827242e-07, "loss": 0.1387, "step": 6387 }, { "epoch": 1.33, "learning_rate": 1.0110049833887043e-07, "loss": 0.145, "step": 6388 }, { "epoch": 1.33, "learning_rate": 1.0106935215946844e-07, "loss": 0.1969, "step": 6389 }, { "epoch": 1.33, "learning_rate": 1.0103820598006644e-07, "loss": 0.1661, "step": 6390 }, { "epoch": 1.33, "learning_rate": 1.0100705980066445e-07, "loss": 0.2177, "step": 6391 }, { "epoch": 1.33, "learning_rate": 1.0097591362126246e-07, "loss": 0.1965, "step": 6392 }, { "epoch": 1.33, "learning_rate": 1.0094476744186045e-07, "loss": 0.2271, "step": 6393 }, { "epoch": 1.33, "learning_rate": 1.0091362126245847e-07, "loss": 0.127, "step": 6394 }, { "epoch": 1.33, "learning_rate": 1.0088247508305647e-07, "loss": 0.1536, "step": 6395 }, { "epoch": 1.33, "learning_rate": 1.0085132890365449e-07, "loss": 0.2241, "step": 6396 }, { "epoch": 1.33, "learning_rate": 1.008201827242525e-07, "loss": 0.1664, "step": 6397 }, { "epoch": 1.33, "learning_rate": 1.0078903654485049e-07, "loss": 0.1754, "step": 6398 }, { "epoch": 1.33, "learning_rate": 1.007578903654485e-07, "loss": 0.1366, "step": 6399 }, { "epoch": 1.33, "learning_rate": 1.0072674418604651e-07, "loss": 0.1408, "step": 6400 }, { "epoch": 1.33, "learning_rate": 1.0069559800664451e-07, "loss": 0.156, "step": 6401 }, { "epoch": 1.33, "learning_rate": 1.0066445182724251e-07, "loss": 0.1085, "step": 6402 }, { "epoch": 1.33, "learning_rate": 1.0063330564784053e-07, "loss": 0.1527, "step": 6403 }, { "epoch": 1.33, "learning_rate": 1.0060215946843855e-07, "loss": 0.1393, "step": 6404 }, { "epoch": 1.33, "learning_rate": 1.0057101328903654e-07, "loss": 0.1245, "step": 6405 }, { "epoch": 1.33, "learning_rate": 1.0053986710963455e-07, "loss": 0.2462, "step": 6406 }, { "epoch": 1.33, "learning_rate": 1.0050872093023255e-07, "loss": 0.1389, "step": 6407 }, { "epoch": 1.33, "learning_rate": 1.0047757475083056e-07, "loss": 0.1474, "step": 6408 }, { "epoch": 1.33, "learning_rate": 1.0044642857142856e-07, "loss": 0.2097, "step": 6409 }, { "epoch": 1.33, "learning_rate": 1.0041528239202657e-07, "loss": 0.1261, "step": 6410 }, { "epoch": 1.33, "learning_rate": 1.0038413621262459e-07, "loss": 0.1649, "step": 6411 }, { "epoch": 1.33, "learning_rate": 1.0035299003322258e-07, "loss": 0.2141, "step": 6412 }, { "epoch": 1.33, "learning_rate": 1.003218438538206e-07, "loss": 0.1712, "step": 6413 }, { "epoch": 1.33, "learning_rate": 1.002906976744186e-07, "loss": 0.2173, "step": 6414 }, { "epoch": 1.33, "learning_rate": 1.0025955149501661e-07, "loss": 0.186, "step": 6415 }, { "epoch": 1.33, "learning_rate": 1.002284053156146e-07, "loss": 0.2118, "step": 6416 }, { "epoch": 1.33, "learning_rate": 1.0019725913621262e-07, "loss": 0.1455, "step": 6417 }, { "epoch": 1.33, "learning_rate": 1.0016611295681064e-07, "loss": 0.137, "step": 6418 }, { "epoch": 1.33, "learning_rate": 1.0013496677740862e-07, "loss": 0.1752, "step": 6419 }, { "epoch": 1.33, "learning_rate": 1.0010382059800664e-07, "loss": 0.1851, "step": 6420 }, { "epoch": 1.33, "learning_rate": 1.0007267441860464e-07, "loss": 0.2014, "step": 6421 }, { "epoch": 1.33, "learning_rate": 1.0004152823920266e-07, "loss": 0.1634, "step": 6422 }, { "epoch": 1.33, "learning_rate": 1.0001038205980065e-07, "loss": 0.2345, "step": 6423 }, { "epoch": 1.33, "learning_rate": 9.997923588039866e-08, "loss": 0.1886, "step": 6424 }, { "epoch": 1.33, "learning_rate": 9.994808970099668e-08, "loss": 0.165, "step": 6425 }, { "epoch": 1.33, "learning_rate": 9.991694352159468e-08, "loss": 0.1768, "step": 6426 }, { "epoch": 1.33, "learning_rate": 9.988579734219268e-08, "loss": 0.1784, "step": 6427 }, { "epoch": 1.33, "learning_rate": 9.985465116279069e-08, "loss": 0.1437, "step": 6428 }, { "epoch": 1.33, "learning_rate": 9.98235049833887e-08, "loss": 0.1772, "step": 6429 }, { "epoch": 1.34, "learning_rate": 9.979235880398669e-08, "loss": 0.1293, "step": 6430 }, { "epoch": 1.34, "learning_rate": 9.976121262458471e-08, "loss": 0.2212, "step": 6431 }, { "epoch": 1.34, "learning_rate": 9.973006644518272e-08, "loss": 0.1951, "step": 6432 }, { "epoch": 1.34, "learning_rate": 9.969892026578073e-08, "loss": 0.1414, "step": 6433 }, { "epoch": 1.34, "learning_rate": 9.966777408637874e-08, "loss": 0.1861, "step": 6434 }, { "epoch": 1.34, "learning_rate": 9.963662790697673e-08, "loss": 0.1748, "step": 6435 }, { "epoch": 1.34, "learning_rate": 9.960548172757475e-08, "loss": 0.2023, "step": 6436 }, { "epoch": 1.34, "learning_rate": 9.957433554817276e-08, "loss": 0.1385, "step": 6437 }, { "epoch": 1.34, "learning_rate": 9.954318936877075e-08, "loss": 0.1482, "step": 6438 }, { "epoch": 1.34, "learning_rate": 9.951204318936877e-08, "loss": 0.1634, "step": 6439 }, { "epoch": 1.34, "learning_rate": 9.948089700996677e-08, "loss": 0.1335, "step": 6440 }, { "epoch": 1.34, "learning_rate": 9.944975083056479e-08, "loss": 0.1387, "step": 6441 }, { "epoch": 1.34, "learning_rate": 9.941860465116278e-08, "loss": 0.1745, "step": 6442 }, { "epoch": 1.34, "learning_rate": 9.938745847176079e-08, "loss": 0.1081, "step": 6443 }, { "epoch": 1.34, "learning_rate": 9.935631229235881e-08, "loss": 0.2224, "step": 6444 }, { "epoch": 1.34, "learning_rate": 9.93251661129568e-08, "loss": 0.159, "step": 6445 }, { "epoch": 1.34, "learning_rate": 9.929401993355481e-08, "loss": 0.1711, "step": 6446 }, { "epoch": 1.34, "learning_rate": 9.926287375415282e-08, "loss": 0.2329, "step": 6447 }, { "epoch": 1.34, "learning_rate": 9.923172757475083e-08, "loss": 0.1752, "step": 6448 }, { "epoch": 1.34, "learning_rate": 9.920058139534882e-08, "loss": 0.1969, "step": 6449 }, { "epoch": 1.34, "learning_rate": 9.916943521594684e-08, "loss": 0.2279, "step": 6450 }, { "epoch": 1.34, "learning_rate": 9.913828903654485e-08, "loss": 0.1899, "step": 6451 }, { "epoch": 1.34, "learning_rate": 9.910714285714286e-08, "loss": 0.1452, "step": 6452 }, { "epoch": 1.34, "learning_rate": 9.907599667774086e-08, "loss": 0.1151, "step": 6453 }, { "epoch": 1.34, "learning_rate": 9.904485049833886e-08, "loss": 0.1557, "step": 6454 }, { "epoch": 1.34, "learning_rate": 9.901370431893688e-08, "loss": 0.1742, "step": 6455 }, { "epoch": 1.34, "learning_rate": 9.898255813953487e-08, "loss": 0.1627, "step": 6456 }, { "epoch": 1.34, "learning_rate": 9.895141196013288e-08, "loss": 0.203, "step": 6457 }, { "epoch": 1.34, "learning_rate": 9.89202657807309e-08, "loss": 0.1982, "step": 6458 }, { "epoch": 1.34, "learning_rate": 9.88891196013289e-08, "loss": 0.1726, "step": 6459 }, { "epoch": 1.34, "learning_rate": 9.885797342192692e-08, "loss": 0.1789, "step": 6460 }, { "epoch": 1.34, "learning_rate": 9.88268272425249e-08, "loss": 0.1524, "step": 6461 }, { "epoch": 1.34, "learning_rate": 9.879568106312292e-08, "loss": 0.2066, "step": 6462 }, { "epoch": 1.34, "learning_rate": 9.876453488372092e-08, "loss": 0.1344, "step": 6463 }, { "epoch": 1.34, "learning_rate": 9.873338870431893e-08, "loss": 0.2099, "step": 6464 }, { "epoch": 1.34, "learning_rate": 9.870224252491694e-08, "loss": 0.1824, "step": 6465 }, { "epoch": 1.34, "learning_rate": 9.867109634551494e-08, "loss": 0.1573, "step": 6466 }, { "epoch": 1.34, "learning_rate": 9.863995016611296e-08, "loss": 0.2236, "step": 6467 }, { "epoch": 1.34, "learning_rate": 9.860880398671095e-08, "loss": 0.1708, "step": 6468 }, { "epoch": 1.34, "learning_rate": 9.857765780730897e-08, "loss": 0.2226, "step": 6469 }, { "epoch": 1.34, "learning_rate": 9.854651162790697e-08, "loss": 0.1539, "step": 6470 }, { "epoch": 1.34, "learning_rate": 9.851536544850498e-08, "loss": 0.2139, "step": 6471 }, { "epoch": 1.34, "learning_rate": 9.848421926910299e-08, "loss": 0.164, "step": 6472 }, { "epoch": 1.34, "learning_rate": 9.845307308970099e-08, "loss": 0.1421, "step": 6473 }, { "epoch": 1.34, "learning_rate": 9.8421926910299e-08, "loss": 0.1766, "step": 6474 }, { "epoch": 1.34, "learning_rate": 9.8390780730897e-08, "loss": 0.1785, "step": 6475 }, { "epoch": 1.34, "learning_rate": 9.835963455149501e-08, "loss": 0.1585, "step": 6476 }, { "epoch": 1.34, "learning_rate": 9.832848837209303e-08, "loss": 0.2163, "step": 6477 }, { "epoch": 1.35, "learning_rate": 9.829734219269103e-08, "loss": 0.1785, "step": 6478 }, { "epoch": 1.35, "learning_rate": 9.826619601328903e-08, "loss": 0.1563, "step": 6479 }, { "epoch": 1.35, "learning_rate": 9.823504983388703e-08, "loss": 0.1487, "step": 6480 }, { "epoch": 1.35, "learning_rate": 9.820390365448505e-08, "loss": 0.1765, "step": 6481 }, { "epoch": 1.35, "learning_rate": 9.817275747508304e-08, "loss": 0.2255, "step": 6482 }, { "epoch": 1.35, "learning_rate": 9.814161129568105e-08, "loss": 0.1969, "step": 6483 }, { "epoch": 1.35, "learning_rate": 9.811046511627907e-08, "loss": 0.2002, "step": 6484 }, { "epoch": 1.35, "learning_rate": 9.807931893687707e-08, "loss": 0.1743, "step": 6485 }, { "epoch": 1.35, "learning_rate": 9.804817275747509e-08, "loss": 0.214, "step": 6486 }, { "epoch": 1.35, "learning_rate": 9.801702657807308e-08, "loss": 0.1688, "step": 6487 }, { "epoch": 1.35, "learning_rate": 9.79858803986711e-08, "loss": 0.1812, "step": 6488 }, { "epoch": 1.35, "learning_rate": 9.79547342192691e-08, "loss": 0.1792, "step": 6489 }, { "epoch": 1.35, "learning_rate": 9.79235880398671e-08, "loss": 0.212, "step": 6490 }, { "epoch": 1.35, "learning_rate": 9.789244186046512e-08, "loss": 0.1931, "step": 6491 }, { "epoch": 1.35, "learning_rate": 9.786129568106312e-08, "loss": 0.1783, "step": 6492 }, { "epoch": 1.35, "learning_rate": 9.783014950166113e-08, "loss": 0.1705, "step": 6493 }, { "epoch": 1.35, "learning_rate": 9.779900332225912e-08, "loss": 0.1948, "step": 6494 }, { "epoch": 1.35, "learning_rate": 9.776785714285714e-08, "loss": 0.1634, "step": 6495 }, { "epoch": 1.35, "learning_rate": 9.773671096345514e-08, "loss": 0.1197, "step": 6496 }, { "epoch": 1.35, "learning_rate": 9.770556478405316e-08, "loss": 0.2207, "step": 6497 }, { "epoch": 1.35, "learning_rate": 9.767441860465116e-08, "loss": 0.188, "step": 6498 }, { "epoch": 1.35, "learning_rate": 9.764327242524916e-08, "loss": 0.1777, "step": 6499 }, { "epoch": 1.35, "learning_rate": 9.761212624584718e-08, "loss": 0.2117, "step": 6500 }, { "epoch": 1.35, "learning_rate": 9.758098006644517e-08, "loss": 0.129, "step": 6501 }, { "epoch": 1.35, "learning_rate": 9.754983388704318e-08, "loss": 0.1686, "step": 6502 }, { "epoch": 1.35, "learning_rate": 9.751868770764119e-08, "loss": 0.1551, "step": 6503 }, { "epoch": 1.35, "learning_rate": 9.74875415282392e-08, "loss": 0.2033, "step": 6504 }, { "epoch": 1.35, "learning_rate": 9.745639534883722e-08, "loss": 0.1895, "step": 6505 }, { "epoch": 1.35, "learning_rate": 9.742524916943521e-08, "loss": 0.1653, "step": 6506 }, { "epoch": 1.35, "learning_rate": 9.739410299003322e-08, "loss": 0.1713, "step": 6507 }, { "epoch": 1.35, "learning_rate": 9.736295681063123e-08, "loss": 0.1726, "step": 6508 }, { "epoch": 1.35, "learning_rate": 9.733181063122923e-08, "loss": 0.1505, "step": 6509 }, { "epoch": 1.35, "learning_rate": 9.730066445182723e-08, "loss": 0.1594, "step": 6510 }, { "epoch": 1.35, "learning_rate": 9.726951827242525e-08, "loss": 0.1874, "step": 6511 }, { "epoch": 1.35, "learning_rate": 9.723837209302326e-08, "loss": 0.184, "step": 6512 }, { "epoch": 1.35, "learning_rate": 9.720722591362125e-08, "loss": 0.1792, "step": 6513 }, { "epoch": 1.35, "learning_rate": 9.717607973421927e-08, "loss": 0.1584, "step": 6514 }, { "epoch": 1.35, "learning_rate": 9.714493355481727e-08, "loss": 0.1787, "step": 6515 }, { "epoch": 1.35, "learning_rate": 9.711378737541527e-08, "loss": 0.2666, "step": 6516 }, { "epoch": 1.35, "learning_rate": 9.708264119601329e-08, "loss": 0.1535, "step": 6517 }, { "epoch": 1.35, "learning_rate": 9.705149501661129e-08, "loss": 0.215, "step": 6518 }, { "epoch": 1.35, "learning_rate": 9.702034883720931e-08, "loss": 0.129, "step": 6519 }, { "epoch": 1.35, "learning_rate": 9.69892026578073e-08, "loss": 0.1506, "step": 6520 }, { "epoch": 1.35, "learning_rate": 9.695805647840531e-08, "loss": 0.1572, "step": 6521 }, { "epoch": 1.35, "learning_rate": 9.692691029900331e-08, "loss": 0.158, "step": 6522 }, { "epoch": 1.35, "learning_rate": 9.689576411960133e-08, "loss": 0.2096, "step": 6523 }, { "epoch": 1.35, "learning_rate": 9.686461794019933e-08, "loss": 0.1574, "step": 6524 }, { "epoch": 1.35, "learning_rate": 9.683347176079734e-08, "loss": 0.2162, "step": 6525 }, { "epoch": 1.36, "learning_rate": 9.680232558139535e-08, "loss": 0.1452, "step": 6526 }, { "epoch": 1.36, "learning_rate": 9.677117940199334e-08, "loss": 0.3019, "step": 6527 }, { "epoch": 1.36, "learning_rate": 9.674003322259136e-08, "loss": 0.1393, "step": 6528 }, { "epoch": 1.36, "learning_rate": 9.670888704318936e-08, "loss": 0.2133, "step": 6529 }, { "epoch": 1.36, "learning_rate": 9.667774086378738e-08, "loss": 0.219, "step": 6530 }, { "epoch": 1.36, "learning_rate": 9.664659468438539e-08, "loss": 0.1219, "step": 6531 }, { "epoch": 1.36, "learning_rate": 9.661544850498338e-08, "loss": 0.1704, "step": 6532 }, { "epoch": 1.36, "learning_rate": 9.65843023255814e-08, "loss": 0.1725, "step": 6533 }, { "epoch": 1.36, "learning_rate": 9.65531561461794e-08, "loss": 0.2211, "step": 6534 }, { "epoch": 1.36, "learning_rate": 9.65220099667774e-08, "loss": 0.1545, "step": 6535 }, { "epoch": 1.36, "learning_rate": 9.64908637873754e-08, "loss": 0.2139, "step": 6536 }, { "epoch": 1.36, "learning_rate": 9.645971760797342e-08, "loss": 0.2151, "step": 6537 }, { "epoch": 1.36, "learning_rate": 9.642857142857144e-08, "loss": 0.1847, "step": 6538 }, { "epoch": 1.36, "learning_rate": 9.639742524916942e-08, "loss": 0.1217, "step": 6539 }, { "epoch": 1.36, "learning_rate": 9.636627906976744e-08, "loss": 0.2062, "step": 6540 }, { "epoch": 1.36, "learning_rate": 9.633513289036544e-08, "loss": 0.1466, "step": 6541 }, { "epoch": 1.36, "learning_rate": 9.630398671096346e-08, "loss": 0.1308, "step": 6542 }, { "epoch": 1.36, "learning_rate": 9.627284053156145e-08, "loss": 0.224, "step": 6543 }, { "epoch": 1.36, "learning_rate": 9.624169435215946e-08, "loss": 0.1642, "step": 6544 }, { "epoch": 1.36, "learning_rate": 9.621054817275748e-08, "loss": 0.1665, "step": 6545 }, { "epoch": 1.36, "learning_rate": 9.617940199335547e-08, "loss": 0.2208, "step": 6546 }, { "epoch": 1.36, "learning_rate": 9.614825581395349e-08, "loss": 0.1907, "step": 6547 }, { "epoch": 1.36, "learning_rate": 9.611710963455149e-08, "loss": 0.3126, "step": 6548 }, { "epoch": 1.36, "learning_rate": 9.60859634551495e-08, "loss": 0.145, "step": 6549 }, { "epoch": 1.36, "learning_rate": 9.605481727574749e-08, "loss": 0.1975, "step": 6550 }, { "epoch": 1.36, "learning_rate": 9.602367109634551e-08, "loss": 0.1958, "step": 6551 }, { "epoch": 1.36, "learning_rate": 9.599252491694352e-08, "loss": 0.2112, "step": 6552 }, { "epoch": 1.36, "learning_rate": 9.596137873754151e-08, "loss": 0.1745, "step": 6553 }, { "epoch": 1.36, "learning_rate": 9.593023255813953e-08, "loss": 0.2103, "step": 6554 }, { "epoch": 1.36, "learning_rate": 9.589908637873753e-08, "loss": 0.1754, "step": 6555 }, { "epoch": 1.36, "learning_rate": 9.586794019933555e-08, "loss": 0.2074, "step": 6556 }, { "epoch": 1.36, "learning_rate": 9.583679401993354e-08, "loss": 0.134, "step": 6557 }, { "epoch": 1.36, "learning_rate": 9.580564784053155e-08, "loss": 0.1734, "step": 6558 }, { "epoch": 1.36, "learning_rate": 9.577450166112957e-08, "loss": 0.2111, "step": 6559 }, { "epoch": 1.36, "learning_rate": 9.574335548172757e-08, "loss": 0.1842, "step": 6560 }, { "epoch": 1.36, "learning_rate": 9.571220930232557e-08, "loss": 0.2004, "step": 6561 }, { "epoch": 1.36, "learning_rate": 9.568106312292358e-08, "loss": 0.1959, "step": 6562 }, { "epoch": 1.36, "learning_rate": 9.564991694352159e-08, "loss": 0.2307, "step": 6563 }, { "epoch": 1.36, "learning_rate": 9.561877076411961e-08, "loss": 0.195, "step": 6564 }, { "epoch": 1.36, "learning_rate": 9.55876245847176e-08, "loss": 0.1803, "step": 6565 }, { "epoch": 1.36, "learning_rate": 9.555647840531561e-08, "loss": 0.1319, "step": 6566 }, { "epoch": 1.36, "learning_rate": 9.552533222591362e-08, "loss": 0.2226, "step": 6567 }, { "epoch": 1.36, "learning_rate": 9.549418604651163e-08, "loss": 0.1825, "step": 6568 }, { "epoch": 1.36, "learning_rate": 9.546303986710962e-08, "loss": 0.1657, "step": 6569 }, { "epoch": 1.36, "learning_rate": 9.543189368770764e-08, "loss": 0.1816, "step": 6570 }, { "epoch": 1.36, "learning_rate": 9.540074750830565e-08, "loss": 0.2152, "step": 6571 }, { "epoch": 1.36, "learning_rate": 9.536960132890364e-08, "loss": 0.1903, "step": 6572 }, { "epoch": 1.36, "learning_rate": 9.533845514950166e-08, "loss": 0.1985, "step": 6573 }, { "epoch": 1.37, "learning_rate": 9.530730897009966e-08, "loss": 0.1992, "step": 6574 }, { "epoch": 1.37, "learning_rate": 9.527616279069768e-08, "loss": 0.2122, "step": 6575 }, { "epoch": 1.37, "learning_rate": 9.524501661129567e-08, "loss": 0.1958, "step": 6576 }, { "epoch": 1.37, "learning_rate": 9.521387043189368e-08, "loss": 0.2271, "step": 6577 }, { "epoch": 1.37, "learning_rate": 9.51827242524917e-08, "loss": 0.1727, "step": 6578 }, { "epoch": 1.37, "learning_rate": 9.51515780730897e-08, "loss": 0.1308, "step": 6579 }, { "epoch": 1.37, "learning_rate": 9.51204318936877e-08, "loss": 0.1614, "step": 6580 }, { "epoch": 1.37, "learning_rate": 9.50892857142857e-08, "loss": 0.2057, "step": 6581 }, { "epoch": 1.37, "learning_rate": 9.505813953488372e-08, "loss": 0.1893, "step": 6582 }, { "epoch": 1.37, "learning_rate": 9.502699335548171e-08, "loss": 0.18, "step": 6583 }, { "epoch": 1.37, "learning_rate": 9.499584717607973e-08, "loss": 0.1893, "step": 6584 }, { "epoch": 1.37, "learning_rate": 9.496470099667774e-08, "loss": 0.1573, "step": 6585 }, { "epoch": 1.37, "learning_rate": 9.493355481727575e-08, "loss": 0.1754, "step": 6586 }, { "epoch": 1.37, "learning_rate": 9.490240863787375e-08, "loss": 0.2711, "step": 6587 }, { "epoch": 1.37, "learning_rate": 9.487126245847175e-08, "loss": 0.2084, "step": 6588 }, { "epoch": 1.37, "learning_rate": 9.484011627906977e-08, "loss": 0.2129, "step": 6589 }, { "epoch": 1.37, "learning_rate": 9.480897009966777e-08, "loss": 0.1421, "step": 6590 }, { "epoch": 1.37, "learning_rate": 9.477782392026577e-08, "loss": 0.1476, "step": 6591 }, { "epoch": 1.37, "learning_rate": 9.474667774086379e-08, "loss": 0.163, "step": 6592 }, { "epoch": 1.37, "learning_rate": 9.471553156146179e-08, "loss": 0.1681, "step": 6593 }, { "epoch": 1.37, "learning_rate": 9.46843853820598e-08, "loss": 0.2152, "step": 6594 }, { "epoch": 1.37, "learning_rate": 9.46532392026578e-08, "loss": 0.1857, "step": 6595 }, { "epoch": 1.37, "learning_rate": 9.462209302325581e-08, "loss": 0.2245, "step": 6596 }, { "epoch": 1.37, "learning_rate": 9.459094684385381e-08, "loss": 0.2353, "step": 6597 }, { "epoch": 1.37, "learning_rate": 9.455980066445182e-08, "loss": 0.1573, "step": 6598 }, { "epoch": 1.37, "learning_rate": 9.452865448504983e-08, "loss": 0.1519, "step": 6599 }, { "epoch": 1.37, "learning_rate": 9.449750830564783e-08, "loss": 0.1711, "step": 6600 }, { "epoch": 1.37, "learning_rate": 9.446636212624585e-08, "loss": 0.1554, "step": 6601 }, { "epoch": 1.37, "learning_rate": 9.443521594684384e-08, "loss": 0.167, "step": 6602 }, { "epoch": 1.37, "learning_rate": 9.440406976744186e-08, "loss": 0.2062, "step": 6603 }, { "epoch": 1.37, "learning_rate": 9.437292358803987e-08, "loss": 0.1745, "step": 6604 }, { "epoch": 1.37, "learning_rate": 9.434177740863787e-08, "loss": 0.1733, "step": 6605 }, { "epoch": 1.37, "learning_rate": 9.431063122923588e-08, "loss": 0.1444, "step": 6606 }, { "epoch": 1.37, "learning_rate": 9.427948504983388e-08, "loss": 0.2293, "step": 6607 }, { "epoch": 1.37, "learning_rate": 9.42483388704319e-08, "loss": 0.2488, "step": 6608 }, { "epoch": 1.37, "learning_rate": 9.421719269102988e-08, "loss": 0.1763, "step": 6609 }, { "epoch": 1.37, "learning_rate": 9.41860465116279e-08, "loss": 0.171, "step": 6610 }, { "epoch": 1.37, "learning_rate": 9.415490033222592e-08, "loss": 0.2073, "step": 6611 }, { "epoch": 1.37, "learning_rate": 9.412375415282392e-08, "loss": 0.1503, "step": 6612 }, { "epoch": 1.37, "learning_rate": 9.409260797342193e-08, "loss": 0.2344, "step": 6613 }, { "epoch": 1.37, "learning_rate": 9.406146179401992e-08, "loss": 0.1405, "step": 6614 }, { "epoch": 1.37, "learning_rate": 9.403031561461794e-08, "loss": 0.1514, "step": 6615 }, { "epoch": 1.37, "learning_rate": 9.399916943521594e-08, "loss": 0.2302, "step": 6616 }, { "epoch": 1.37, "learning_rate": 9.396802325581394e-08, "loss": 0.1624, "step": 6617 }, { "epoch": 1.37, "learning_rate": 9.393687707641196e-08, "loss": 0.1651, "step": 6618 }, { "epoch": 1.37, "learning_rate": 9.390573089700996e-08, "loss": 0.1643, "step": 6619 }, { "epoch": 1.37, "learning_rate": 9.387458471760798e-08, "loss": 0.1657, "step": 6620 }, { "epoch": 1.37, "learning_rate": 9.384343853820597e-08, "loss": 0.2077, "step": 6621 }, { "epoch": 1.38, "learning_rate": 9.381229235880398e-08, "loss": 0.1693, "step": 6622 }, { "epoch": 1.38, "learning_rate": 9.378114617940199e-08, "loss": 0.1808, "step": 6623 }, { "epoch": 1.38, "learning_rate": 9.375e-08, "loss": 0.2503, "step": 6624 }, { "epoch": 1.38, "learning_rate": 9.3718853820598e-08, "loss": 0.2343, "step": 6625 }, { "epoch": 1.38, "learning_rate": 9.368770764119601e-08, "loss": 0.1561, "step": 6626 }, { "epoch": 1.38, "learning_rate": 9.365656146179402e-08, "loss": 0.1893, "step": 6627 }, { "epoch": 1.38, "learning_rate": 9.362541528239201e-08, "loss": 0.1731, "step": 6628 }, { "epoch": 1.38, "learning_rate": 9.359426910299003e-08, "loss": 0.2088, "step": 6629 }, { "epoch": 1.38, "learning_rate": 9.356312292358803e-08, "loss": 0.2002, "step": 6630 }, { "epoch": 1.38, "learning_rate": 9.353197674418605e-08, "loss": 0.1833, "step": 6631 }, { "epoch": 1.38, "learning_rate": 9.350083056478405e-08, "loss": 0.1765, "step": 6632 }, { "epoch": 1.38, "learning_rate": 9.346968438538205e-08, "loss": 0.2084, "step": 6633 }, { "epoch": 1.38, "learning_rate": 9.343853820598007e-08, "loss": 0.1579, "step": 6634 }, { "epoch": 1.38, "learning_rate": 9.340739202657806e-08, "loss": 0.1742, "step": 6635 }, { "epoch": 1.38, "learning_rate": 9.337624584717607e-08, "loss": 0.1484, "step": 6636 }, { "epoch": 1.38, "learning_rate": 9.334509966777408e-08, "loss": 0.1644, "step": 6637 }, { "epoch": 1.38, "learning_rate": 9.331395348837209e-08, "loss": 0.236, "step": 6638 }, { "epoch": 1.38, "learning_rate": 9.328280730897011e-08, "loss": 0.1372, "step": 6639 }, { "epoch": 1.38, "learning_rate": 9.32516611295681e-08, "loss": 0.1657, "step": 6640 }, { "epoch": 1.38, "learning_rate": 9.322051495016611e-08, "loss": 0.2114, "step": 6641 }, { "epoch": 1.38, "learning_rate": 9.318936877076412e-08, "loss": 0.2207, "step": 6642 }, { "epoch": 1.38, "learning_rate": 9.315822259136212e-08, "loss": 0.2108, "step": 6643 }, { "epoch": 1.38, "learning_rate": 9.312707641196012e-08, "loss": 0.2026, "step": 6644 }, { "epoch": 1.38, "learning_rate": 9.309593023255814e-08, "loss": 0.1865, "step": 6645 }, { "epoch": 1.38, "learning_rate": 9.306478405315615e-08, "loss": 0.157, "step": 6646 }, { "epoch": 1.38, "learning_rate": 9.303363787375414e-08, "loss": 0.2598, "step": 6647 }, { "epoch": 1.38, "learning_rate": 9.300249169435216e-08, "loss": 0.1823, "step": 6648 }, { "epoch": 1.38, "learning_rate": 9.297134551495016e-08, "loss": 0.1216, "step": 6649 }, { "epoch": 1.38, "learning_rate": 9.294019933554818e-08, "loss": 0.1783, "step": 6650 }, { "epoch": 1.38, "learning_rate": 9.290905315614618e-08, "loss": 0.1484, "step": 6651 }, { "epoch": 1.38, "learning_rate": 9.287790697674418e-08, "loss": 0.1568, "step": 6652 }, { "epoch": 1.38, "learning_rate": 9.28467607973422e-08, "loss": 0.1878, "step": 6653 }, { "epoch": 1.38, "learning_rate": 9.281561461794019e-08, "loss": 0.1942, "step": 6654 }, { "epoch": 1.38, "learning_rate": 9.27844684385382e-08, "loss": 0.1803, "step": 6655 }, { "epoch": 1.38, "learning_rate": 9.27533222591362e-08, "loss": 0.1536, "step": 6656 }, { "epoch": 1.38, "learning_rate": 9.272217607973422e-08, "loss": 0.1501, "step": 6657 }, { "epoch": 1.38, "learning_rate": 9.269102990033222e-08, "loss": 0.1386, "step": 6658 }, { "epoch": 1.38, "learning_rate": 9.265988372093023e-08, "loss": 0.177, "step": 6659 }, { "epoch": 1.38, "learning_rate": 9.262873754152824e-08, "loss": 0.1545, "step": 6660 }, { "epoch": 1.38, "learning_rate": 9.259759136212624e-08, "loss": 0.1097, "step": 6661 }, { "epoch": 1.38, "learning_rate": 9.256644518272425e-08, "loss": 0.2311, "step": 6662 }, { "epoch": 1.38, "learning_rate": 9.253529900332225e-08, "loss": 0.2186, "step": 6663 }, { "epoch": 1.38, "learning_rate": 9.250415282392026e-08, "loss": 0.1837, "step": 6664 }, { "epoch": 1.38, "learning_rate": 9.247300664451828e-08, "loss": 0.1671, "step": 6665 }, { "epoch": 1.38, "learning_rate": 9.244186046511627e-08, "loss": 0.1587, "step": 6666 }, { "epoch": 1.38, "learning_rate": 9.241071428571429e-08, "loss": 0.1945, "step": 6667 }, { "epoch": 1.38, "learning_rate": 9.237956810631229e-08, "loss": 0.1468, "step": 6668 }, { "epoch": 1.38, "learning_rate": 9.234842192691029e-08, "loss": 0.2116, "step": 6669 }, { "epoch": 1.38, "learning_rate": 9.23172757475083e-08, "loss": 0.1909, "step": 6670 }, { "epoch": 1.39, "learning_rate": 9.228612956810631e-08, "loss": 0.2212, "step": 6671 }, { "epoch": 1.39, "learning_rate": 9.225498338870433e-08, "loss": 0.174, "step": 6672 }, { "epoch": 1.39, "learning_rate": 9.222383720930231e-08, "loss": 0.1944, "step": 6673 }, { "epoch": 1.39, "learning_rate": 9.219269102990033e-08, "loss": 0.1821, "step": 6674 }, { "epoch": 1.39, "learning_rate": 9.216154485049833e-08, "loss": 0.2465, "step": 6675 }, { "epoch": 1.39, "learning_rate": 9.213039867109635e-08, "loss": 0.1705, "step": 6676 }, { "epoch": 1.39, "learning_rate": 9.209925249169434e-08, "loss": 0.1461, "step": 6677 }, { "epoch": 1.39, "learning_rate": 9.206810631229235e-08, "loss": 0.1487, "step": 6678 }, { "epoch": 1.39, "learning_rate": 9.203696013289037e-08, "loss": 0.1891, "step": 6679 }, { "epoch": 1.39, "learning_rate": 9.200581395348836e-08, "loss": 0.1742, "step": 6680 }, { "epoch": 1.39, "learning_rate": 9.197466777408637e-08, "loss": 0.1726, "step": 6681 }, { "epoch": 1.39, "learning_rate": 9.194352159468438e-08, "loss": 0.1878, "step": 6682 }, { "epoch": 1.39, "learning_rate": 9.19123754152824e-08, "loss": 0.1197, "step": 6683 }, { "epoch": 1.39, "learning_rate": 9.188122923588038e-08, "loss": 0.2111, "step": 6684 }, { "epoch": 1.39, "learning_rate": 9.18500830564784e-08, "loss": 0.1534, "step": 6685 }, { "epoch": 1.39, "learning_rate": 9.181893687707641e-08, "loss": 0.1236, "step": 6686 }, { "epoch": 1.39, "learning_rate": 9.178779069767442e-08, "loss": 0.2035, "step": 6687 }, { "epoch": 1.39, "learning_rate": 9.175664451827242e-08, "loss": 0.1651, "step": 6688 }, { "epoch": 1.39, "learning_rate": 9.172549833887042e-08, "loss": 0.1925, "step": 6689 }, { "epoch": 1.39, "learning_rate": 9.169435215946844e-08, "loss": 0.1906, "step": 6690 }, { "epoch": 1.39, "learning_rate": 9.166320598006645e-08, "loss": 0.1552, "step": 6691 }, { "epoch": 1.39, "learning_rate": 9.163205980066444e-08, "loss": 0.2262, "step": 6692 }, { "epoch": 1.39, "learning_rate": 9.160091362126246e-08, "loss": 0.2704, "step": 6693 }, { "epoch": 1.39, "learning_rate": 9.156976744186046e-08, "loss": 0.1747, "step": 6694 }, { "epoch": 1.39, "learning_rate": 9.153862126245848e-08, "loss": 0.1849, "step": 6695 }, { "epoch": 1.39, "learning_rate": 9.150747508305647e-08, "loss": 0.2275, "step": 6696 }, { "epoch": 1.39, "learning_rate": 9.147632890365448e-08, "loss": 0.192, "step": 6697 }, { "epoch": 1.39, "learning_rate": 9.14451827242525e-08, "loss": 0.1674, "step": 6698 }, { "epoch": 1.39, "learning_rate": 9.141403654485049e-08, "loss": 0.1475, "step": 6699 }, { "epoch": 1.39, "learning_rate": 9.13828903654485e-08, "loss": 0.2218, "step": 6700 }, { "epoch": 1.39, "learning_rate": 9.13517441860465e-08, "loss": 0.1611, "step": 6701 }, { "epoch": 1.39, "learning_rate": 9.132059800664452e-08, "loss": 0.1861, "step": 6702 }, { "epoch": 1.39, "learning_rate": 9.128945182724251e-08, "loss": 0.1691, "step": 6703 }, { "epoch": 1.39, "learning_rate": 9.125830564784053e-08, "loss": 0.2117, "step": 6704 }, { "epoch": 1.39, "learning_rate": 9.122715946843854e-08, "loss": 0.164, "step": 6705 }, { "epoch": 1.39, "learning_rate": 9.119601328903653e-08, "loss": 0.1673, "step": 6706 }, { "epoch": 1.39, "learning_rate": 9.116486710963455e-08, "loss": 0.189, "step": 6707 }, { "epoch": 1.39, "learning_rate": 9.113372093023255e-08, "loss": 0.1492, "step": 6708 }, { "epoch": 1.39, "learning_rate": 9.110257475083057e-08, "loss": 0.1506, "step": 6709 }, { "epoch": 1.39, "learning_rate": 9.107142857142856e-08, "loss": 0.1767, "step": 6710 }, { "epoch": 1.39, "learning_rate": 9.104028239202657e-08, "loss": 0.2047, "step": 6711 }, { "epoch": 1.39, "learning_rate": 9.100913621262459e-08, "loss": 0.1929, "step": 6712 }, { "epoch": 1.39, "learning_rate": 9.097799003322259e-08, "loss": 0.2056, "step": 6713 }, { "epoch": 1.39, "learning_rate": 9.094684385382059e-08, "loss": 0.1726, "step": 6714 }, { "epoch": 1.39, "learning_rate": 9.09156976744186e-08, "loss": 0.2639, "step": 6715 }, { "epoch": 1.39, "learning_rate": 9.088455149501661e-08, "loss": 0.1782, "step": 6716 }, { "epoch": 1.39, "learning_rate": 9.08534053156146e-08, "loss": 0.2439, "step": 6717 }, { "epoch": 1.39, "learning_rate": 9.082225913621262e-08, "loss": 0.2233, "step": 6718 }, { "epoch": 1.4, "learning_rate": 9.079111295681063e-08, "loss": 0.2088, "step": 6719 }, { "epoch": 1.4, "learning_rate": 9.075996677740863e-08, "loss": 0.1327, "step": 6720 }, { "epoch": 1.4, "learning_rate": 9.072882059800665e-08, "loss": 0.1796, "step": 6721 }, { "epoch": 1.4, "learning_rate": 9.069767441860464e-08, "loss": 0.1533, "step": 6722 }, { "epoch": 1.4, "learning_rate": 9.066652823920266e-08, "loss": 0.198, "step": 6723 }, { "epoch": 1.4, "learning_rate": 9.063538205980066e-08, "loss": 0.1559, "step": 6724 }, { "epoch": 1.4, "learning_rate": 9.060423588039866e-08, "loss": 0.2105, "step": 6725 }, { "epoch": 1.4, "learning_rate": 9.057308970099668e-08, "loss": 0.1998, "step": 6726 }, { "epoch": 1.4, "learning_rate": 9.054194352159468e-08, "loss": 0.1305, "step": 6727 }, { "epoch": 1.4, "learning_rate": 9.05107973421927e-08, "loss": 0.2092, "step": 6728 }, { "epoch": 1.4, "learning_rate": 9.047965116279068e-08, "loss": 0.1592, "step": 6729 }, { "epoch": 1.4, "learning_rate": 9.04485049833887e-08, "loss": 0.2046, "step": 6730 }, { "epoch": 1.4, "learning_rate": 9.04173588039867e-08, "loss": 0.1624, "step": 6731 }, { "epoch": 1.4, "learning_rate": 9.038621262458472e-08, "loss": 0.1856, "step": 6732 }, { "epoch": 1.4, "learning_rate": 9.035506644518272e-08, "loss": 0.2206, "step": 6733 }, { "epoch": 1.4, "learning_rate": 9.032392026578072e-08, "loss": 0.1774, "step": 6734 }, { "epoch": 1.4, "learning_rate": 9.029277408637874e-08, "loss": 0.1894, "step": 6735 }, { "epoch": 1.4, "learning_rate": 9.026162790697673e-08, "loss": 0.1972, "step": 6736 }, { "epoch": 1.4, "learning_rate": 9.023048172757474e-08, "loss": 0.2149, "step": 6737 }, { "epoch": 1.4, "learning_rate": 9.019933554817276e-08, "loss": 0.1253, "step": 6738 }, { "epoch": 1.4, "learning_rate": 9.016818936877076e-08, "loss": 0.195, "step": 6739 }, { "epoch": 1.4, "learning_rate": 9.013704318936877e-08, "loss": 0.1986, "step": 6740 }, { "epoch": 1.4, "learning_rate": 9.010589700996677e-08, "loss": 0.229, "step": 6741 }, { "epoch": 1.4, "learning_rate": 9.007475083056478e-08, "loss": 0.1431, "step": 6742 }, { "epoch": 1.4, "learning_rate": 9.004360465116277e-08, "loss": 0.1204, "step": 6743 }, { "epoch": 1.4, "learning_rate": 9.001245847176079e-08, "loss": 0.1587, "step": 6744 }, { "epoch": 1.4, "learning_rate": 8.99813122923588e-08, "loss": 0.304, "step": 6745 }, { "epoch": 1.4, "learning_rate": 8.995016611295681e-08, "loss": 0.1837, "step": 6746 }, { "epoch": 1.4, "learning_rate": 8.991901993355482e-08, "loss": 0.2281, "step": 6747 }, { "epoch": 1.4, "learning_rate": 8.988787375415281e-08, "loss": 0.1737, "step": 6748 }, { "epoch": 1.4, "learning_rate": 8.985672757475083e-08, "loss": 0.1765, "step": 6749 }, { "epoch": 1.4, "learning_rate": 8.982558139534883e-08, "loss": 0.1696, "step": 6750 }, { "epoch": 1.4, "learning_rate": 8.979443521594683e-08, "loss": 0.2096, "step": 6751 }, { "epoch": 1.4, "learning_rate": 8.976328903654485e-08, "loss": 0.2465, "step": 6752 }, { "epoch": 1.4, "learning_rate": 8.973214285714285e-08, "loss": 0.1738, "step": 6753 }, { "epoch": 1.4, "learning_rate": 8.970099667774087e-08, "loss": 0.1953, "step": 6754 }, { "epoch": 1.4, "learning_rate": 8.966985049833886e-08, "loss": 0.1907, "step": 6755 }, { "epoch": 1.4, "learning_rate": 8.963870431893687e-08, "loss": 0.2232, "step": 6756 }, { "epoch": 1.4, "learning_rate": 8.960755813953488e-08, "loss": 0.1767, "step": 6757 }, { "epoch": 1.4, "learning_rate": 8.957641196013289e-08, "loss": 0.154, "step": 6758 }, { "epoch": 1.4, "learning_rate": 8.95452657807309e-08, "loss": 0.1798, "step": 6759 }, { "epoch": 1.4, "learning_rate": 8.95141196013289e-08, "loss": 0.221, "step": 6760 }, { "epoch": 1.4, "learning_rate": 8.948297342192691e-08, "loss": 0.1224, "step": 6761 }, { "epoch": 1.4, "learning_rate": 8.94518272425249e-08, "loss": 0.1741, "step": 6762 }, { "epoch": 1.4, "learning_rate": 8.942068106312292e-08, "loss": 0.1622, "step": 6763 }, { "epoch": 1.4, "learning_rate": 8.938953488372092e-08, "loss": 0.2275, "step": 6764 }, { "epoch": 1.4, "learning_rate": 8.935838870431894e-08, "loss": 0.1181, "step": 6765 }, { "epoch": 1.4, "learning_rate": 8.932724252491695e-08, "loss": 0.1833, "step": 6766 }, { "epoch": 1.41, "learning_rate": 8.929609634551494e-08, "loss": 0.1364, "step": 6767 }, { "epoch": 1.41, "learning_rate": 8.926495016611296e-08, "loss": 0.1536, "step": 6768 }, { "epoch": 1.41, "learning_rate": 8.923380398671096e-08, "loss": 0.1905, "step": 6769 }, { "epoch": 1.41, "learning_rate": 8.920265780730896e-08, "loss": 0.1437, "step": 6770 }, { "epoch": 1.41, "learning_rate": 8.917151162790697e-08, "loss": 0.174, "step": 6771 }, { "epoch": 1.41, "learning_rate": 8.914036544850498e-08, "loss": 0.1592, "step": 6772 }, { "epoch": 1.41, "learning_rate": 8.9109219269103e-08, "loss": 0.2147, "step": 6773 }, { "epoch": 1.41, "learning_rate": 8.907807308970099e-08, "loss": 0.1481, "step": 6774 }, { "epoch": 1.41, "learning_rate": 8.9046926910299e-08, "loss": 0.2044, "step": 6775 }, { "epoch": 1.41, "learning_rate": 8.9015780730897e-08, "loss": 0.1312, "step": 6776 }, { "epoch": 1.41, "learning_rate": 8.898463455149501e-08, "loss": 0.1395, "step": 6777 }, { "epoch": 1.41, "learning_rate": 8.895348837209302e-08, "loss": 0.2343, "step": 6778 }, { "epoch": 1.41, "learning_rate": 8.892234219269103e-08, "loss": 0.1727, "step": 6779 }, { "epoch": 1.41, "learning_rate": 8.889119601328904e-08, "loss": 0.1583, "step": 6780 }, { "epoch": 1.41, "learning_rate": 8.886004983388703e-08, "loss": 0.2116, "step": 6781 }, { "epoch": 1.41, "learning_rate": 8.882890365448505e-08, "loss": 0.2011, "step": 6782 }, { "epoch": 1.41, "learning_rate": 8.879775747508305e-08, "loss": 0.1653, "step": 6783 }, { "epoch": 1.41, "learning_rate": 8.876661129568107e-08, "loss": 0.1539, "step": 6784 }, { "epoch": 1.41, "learning_rate": 8.873546511627907e-08, "loss": 0.187, "step": 6785 }, { "epoch": 1.41, "learning_rate": 8.870431893687707e-08, "loss": 0.1269, "step": 6786 }, { "epoch": 1.41, "learning_rate": 8.867317275747509e-08, "loss": 0.2515, "step": 6787 }, { "epoch": 1.41, "learning_rate": 8.864202657807308e-08, "loss": 0.1825, "step": 6788 }, { "epoch": 1.41, "learning_rate": 8.861088039867109e-08, "loss": 0.2519, "step": 6789 }, { "epoch": 1.41, "learning_rate": 8.85797342192691e-08, "loss": 0.1975, "step": 6790 }, { "epoch": 1.41, "learning_rate": 8.854858803986711e-08, "loss": 0.1829, "step": 6791 }, { "epoch": 1.41, "learning_rate": 8.851744186046513e-08, "loss": 0.1521, "step": 6792 }, { "epoch": 1.41, "learning_rate": 8.848629568106312e-08, "loss": 0.2111, "step": 6793 }, { "epoch": 1.41, "learning_rate": 8.845514950166113e-08, "loss": 0.1811, "step": 6794 }, { "epoch": 1.41, "learning_rate": 8.842400332225913e-08, "loss": 0.1695, "step": 6795 }, { "epoch": 1.41, "learning_rate": 8.839285714285714e-08, "loss": 0.1796, "step": 6796 }, { "epoch": 1.41, "learning_rate": 8.836171096345514e-08, "loss": 0.1843, "step": 6797 }, { "epoch": 1.41, "learning_rate": 8.833056478405315e-08, "loss": 0.1724, "step": 6798 }, { "epoch": 1.41, "learning_rate": 8.829941860465117e-08, "loss": 0.1944, "step": 6799 }, { "epoch": 1.41, "learning_rate": 8.826827242524916e-08, "loss": 0.1856, "step": 6800 }, { "epoch": 1.41, "learning_rate": 8.823712624584718e-08, "loss": 0.1928, "step": 6801 }, { "epoch": 1.41, "learning_rate": 8.820598006644518e-08, "loss": 0.1842, "step": 6802 }, { "epoch": 1.41, "learning_rate": 8.81748338870432e-08, "loss": 0.1501, "step": 6803 }, { "epoch": 1.41, "learning_rate": 8.814368770764118e-08, "loss": 0.1729, "step": 6804 }, { "epoch": 1.41, "learning_rate": 8.81125415282392e-08, "loss": 0.1948, "step": 6805 }, { "epoch": 1.41, "learning_rate": 8.808139534883721e-08, "loss": 0.1473, "step": 6806 }, { "epoch": 1.41, "learning_rate": 8.80502491694352e-08, "loss": 0.1583, "step": 6807 }, { "epoch": 1.41, "learning_rate": 8.801910299003322e-08, "loss": 0.1882, "step": 6808 }, { "epoch": 1.41, "learning_rate": 8.798795681063122e-08, "loss": 0.1949, "step": 6809 }, { "epoch": 1.41, "learning_rate": 8.795681063122924e-08, "loss": 0.1506, "step": 6810 }, { "epoch": 1.41, "learning_rate": 8.792566445182723e-08, "loss": 0.1773, "step": 6811 }, { "epoch": 1.41, "learning_rate": 8.789451827242524e-08, "loss": 0.2064, "step": 6812 }, { "epoch": 1.41, "learning_rate": 8.786337209302326e-08, "loss": 0.1422, "step": 6813 }, { "epoch": 1.41, "learning_rate": 8.783222591362125e-08, "loss": 0.1069, "step": 6814 }, { "epoch": 1.42, "learning_rate": 8.780107973421926e-08, "loss": 0.1547, "step": 6815 }, { "epoch": 1.42, "learning_rate": 8.776993355481727e-08, "loss": 0.2038, "step": 6816 }, { "epoch": 1.42, "learning_rate": 8.773878737541528e-08, "loss": 0.1833, "step": 6817 }, { "epoch": 1.42, "learning_rate": 8.77076411960133e-08, "loss": 0.1869, "step": 6818 }, { "epoch": 1.42, "learning_rate": 8.767649501661129e-08, "loss": 0.1977, "step": 6819 }, { "epoch": 1.42, "learning_rate": 8.76453488372093e-08, "loss": 0.2146, "step": 6820 }, { "epoch": 1.42, "learning_rate": 8.761420265780731e-08, "loss": 0.1986, "step": 6821 }, { "epoch": 1.42, "learning_rate": 8.758305647840531e-08, "loss": 0.2183, "step": 6822 }, { "epoch": 1.42, "learning_rate": 8.755191029900331e-08, "loss": 0.1408, "step": 6823 }, { "epoch": 1.42, "learning_rate": 8.752076411960133e-08, "loss": 0.1858, "step": 6824 }, { "epoch": 1.42, "learning_rate": 8.748961794019934e-08, "loss": 0.2002, "step": 6825 }, { "epoch": 1.42, "learning_rate": 8.745847176079733e-08, "loss": 0.1358, "step": 6826 }, { "epoch": 1.42, "learning_rate": 8.742732558139535e-08, "loss": 0.1769, "step": 6827 }, { "epoch": 1.42, "learning_rate": 8.739617940199335e-08, "loss": 0.1597, "step": 6828 }, { "epoch": 1.42, "learning_rate": 8.736503322259137e-08, "loss": 0.184, "step": 6829 }, { "epoch": 1.42, "learning_rate": 8.733388704318936e-08, "loss": 0.209, "step": 6830 }, { "epoch": 1.42, "learning_rate": 8.730274086378737e-08, "loss": 0.1414, "step": 6831 }, { "epoch": 1.42, "learning_rate": 8.727159468438539e-08, "loss": 0.206, "step": 6832 }, { "epoch": 1.42, "learning_rate": 8.724044850498338e-08, "loss": 0.1364, "step": 6833 }, { "epoch": 1.42, "learning_rate": 8.720930232558139e-08, "loss": 0.2608, "step": 6834 }, { "epoch": 1.42, "learning_rate": 8.71781561461794e-08, "loss": 0.1666, "step": 6835 }, { "epoch": 1.42, "learning_rate": 8.714700996677741e-08, "loss": 0.1795, "step": 6836 }, { "epoch": 1.42, "learning_rate": 8.71158637873754e-08, "loss": 0.2026, "step": 6837 }, { "epoch": 1.42, "learning_rate": 8.708471760797342e-08, "loss": 0.2111, "step": 6838 }, { "epoch": 1.42, "learning_rate": 8.705357142857143e-08, "loss": 0.1958, "step": 6839 }, { "epoch": 1.42, "learning_rate": 8.702242524916944e-08, "loss": 0.2098, "step": 6840 }, { "epoch": 1.42, "learning_rate": 8.699127906976744e-08, "loss": 0.2627, "step": 6841 }, { "epoch": 1.42, "learning_rate": 8.696013289036544e-08, "loss": 0.1891, "step": 6842 }, { "epoch": 1.42, "learning_rate": 8.692898671096346e-08, "loss": 0.2582, "step": 6843 }, { "epoch": 1.42, "learning_rate": 8.689784053156145e-08, "loss": 0.1537, "step": 6844 }, { "epoch": 1.42, "learning_rate": 8.686669435215946e-08, "loss": 0.241, "step": 6845 }, { "epoch": 1.42, "learning_rate": 8.683554817275748e-08, "loss": 0.1371, "step": 6846 }, { "epoch": 1.42, "learning_rate": 8.680440199335548e-08, "loss": 0.1747, "step": 6847 }, { "epoch": 1.42, "learning_rate": 8.677325581395348e-08, "loss": 0.172, "step": 6848 }, { "epoch": 1.42, "learning_rate": 8.674210963455149e-08, "loss": 0.1373, "step": 6849 }, { "epoch": 1.42, "learning_rate": 8.67109634551495e-08, "loss": 0.2108, "step": 6850 }, { "epoch": 1.42, "learning_rate": 8.667981727574749e-08, "loss": 0.2766, "step": 6851 }, { "epoch": 1.42, "learning_rate": 8.66486710963455e-08, "loss": 0.1931, "step": 6852 }, { "epoch": 1.42, "learning_rate": 8.661752491694352e-08, "loss": 0.1978, "step": 6853 }, { "epoch": 1.42, "learning_rate": 8.658637873754152e-08, "loss": 0.1748, "step": 6854 }, { "epoch": 1.42, "learning_rate": 8.655523255813954e-08, "loss": 0.2258, "step": 6855 }, { "epoch": 1.42, "learning_rate": 8.652408637873753e-08, "loss": 0.1871, "step": 6856 }, { "epoch": 1.42, "learning_rate": 8.649294019933555e-08, "loss": 0.188, "step": 6857 }, { "epoch": 1.42, "learning_rate": 8.646179401993355e-08, "loss": 0.1614, "step": 6858 }, { "epoch": 1.42, "learning_rate": 8.643064784053155e-08, "loss": 0.1998, "step": 6859 }, { "epoch": 1.42, "learning_rate": 8.639950166112957e-08, "loss": 0.1691, "step": 6860 }, { "epoch": 1.42, "learning_rate": 8.636835548172757e-08, "loss": 0.2236, "step": 6861 }, { "epoch": 1.42, "learning_rate": 8.633720930232558e-08, "loss": 0.1948, "step": 6862 }, { "epoch": 1.43, "learning_rate": 8.630606312292357e-08, "loss": 0.2694, "step": 6863 }, { "epoch": 1.43, "learning_rate": 8.627491694352159e-08, "loss": 0.1716, "step": 6864 }, { "epoch": 1.43, "learning_rate": 8.62437707641196e-08, "loss": 0.175, "step": 6865 }, { "epoch": 1.43, "learning_rate": 8.621262458471761e-08, "loss": 0.2058, "step": 6866 }, { "epoch": 1.43, "learning_rate": 8.618147840531561e-08, "loss": 0.2005, "step": 6867 }, { "epoch": 1.43, "learning_rate": 8.615033222591361e-08, "loss": 0.1964, "step": 6868 }, { "epoch": 1.43, "learning_rate": 8.611918604651163e-08, "loss": 0.2283, "step": 6869 }, { "epoch": 1.43, "learning_rate": 8.608803986710962e-08, "loss": 0.1622, "step": 6870 }, { "epoch": 1.43, "learning_rate": 8.605689368770763e-08, "loss": 0.1889, "step": 6871 }, { "epoch": 1.43, "learning_rate": 8.602574750830565e-08, "loss": 0.173, "step": 6872 }, { "epoch": 1.43, "learning_rate": 8.599460132890365e-08, "loss": 0.1331, "step": 6873 }, { "epoch": 1.43, "learning_rate": 8.596345514950167e-08, "loss": 0.1228, "step": 6874 }, { "epoch": 1.43, "learning_rate": 8.593230897009966e-08, "loss": 0.1421, "step": 6875 }, { "epoch": 1.43, "learning_rate": 8.590116279069767e-08, "loss": 0.1538, "step": 6876 }, { "epoch": 1.43, "learning_rate": 8.587001661129568e-08, "loss": 0.1796, "step": 6877 }, { "epoch": 1.43, "learning_rate": 8.583887043189368e-08, "loss": 0.2041, "step": 6878 }, { "epoch": 1.43, "learning_rate": 8.58077242524917e-08, "loss": 0.1899, "step": 6879 }, { "epoch": 1.43, "learning_rate": 8.57765780730897e-08, "loss": 0.159, "step": 6880 }, { "epoch": 1.43, "learning_rate": 8.574543189368771e-08, "loss": 0.2022, "step": 6881 }, { "epoch": 1.43, "learning_rate": 8.57142857142857e-08, "loss": 0.1816, "step": 6882 }, { "epoch": 1.43, "learning_rate": 8.568313953488372e-08, "loss": 0.1701, "step": 6883 }, { "epoch": 1.43, "learning_rate": 8.565199335548172e-08, "loss": 0.1639, "step": 6884 }, { "epoch": 1.43, "learning_rate": 8.562084717607972e-08, "loss": 0.1748, "step": 6885 }, { "epoch": 1.43, "learning_rate": 8.558970099667774e-08, "loss": 0.1421, "step": 6886 }, { "epoch": 1.43, "learning_rate": 8.555855481727574e-08, "loss": 0.1857, "step": 6887 }, { "epoch": 1.43, "learning_rate": 8.552740863787376e-08, "loss": 0.1696, "step": 6888 }, { "epoch": 1.43, "learning_rate": 8.549626245847175e-08, "loss": 0.2027, "step": 6889 }, { "epoch": 1.43, "learning_rate": 8.546511627906976e-08, "loss": 0.2442, "step": 6890 }, { "epoch": 1.43, "learning_rate": 8.543397009966777e-08, "loss": 0.1265, "step": 6891 }, { "epoch": 1.43, "learning_rate": 8.540282392026578e-08, "loss": 0.1766, "step": 6892 }, { "epoch": 1.43, "learning_rate": 8.537167774086378e-08, "loss": 0.1702, "step": 6893 }, { "epoch": 1.43, "learning_rate": 8.534053156146179e-08, "loss": 0.1907, "step": 6894 }, { "epoch": 1.43, "learning_rate": 8.53093853820598e-08, "loss": 0.1464, "step": 6895 }, { "epoch": 1.43, "learning_rate": 8.527823920265779e-08, "loss": 0.2811, "step": 6896 }, { "epoch": 1.43, "learning_rate": 8.524709302325581e-08, "loss": 0.15, "step": 6897 }, { "epoch": 1.43, "learning_rate": 8.521594684385381e-08, "loss": 0.2181, "step": 6898 }, { "epoch": 1.43, "learning_rate": 8.518480066445183e-08, "loss": 0.1491, "step": 6899 }, { "epoch": 1.43, "learning_rate": 8.515365448504984e-08, "loss": 0.172, "step": 6900 }, { "epoch": 1.43, "learning_rate": 8.512250830564783e-08, "loss": 0.1999, "step": 6901 }, { "epoch": 1.43, "learning_rate": 8.509136212624585e-08, "loss": 0.1911, "step": 6902 }, { "epoch": 1.43, "learning_rate": 8.506021594684385e-08, "loss": 0.1257, "step": 6903 }, { "epoch": 1.43, "learning_rate": 8.502906976744185e-08, "loss": 0.1516, "step": 6904 }, { "epoch": 1.43, "learning_rate": 8.499792358803987e-08, "loss": 0.2526, "step": 6905 }, { "epoch": 1.43, "learning_rate": 8.496677740863787e-08, "loss": 0.2003, "step": 6906 }, { "epoch": 1.43, "learning_rate": 8.493563122923589e-08, "loss": 0.184, "step": 6907 }, { "epoch": 1.43, "learning_rate": 8.490448504983388e-08, "loss": 0.184, "step": 6908 }, { "epoch": 1.43, "learning_rate": 8.487333887043189e-08, "loss": 0.2309, "step": 6909 }, { "epoch": 1.43, "learning_rate": 8.48421926910299e-08, "loss": 0.1694, "step": 6910 }, { "epoch": 1.44, "learning_rate": 8.481104651162791e-08, "loss": 0.1698, "step": 6911 }, { "epoch": 1.44, "learning_rate": 8.477990033222591e-08, "loss": 0.1293, "step": 6912 }, { "epoch": 1.44, "learning_rate": 8.474875415282392e-08, "loss": 0.1648, "step": 6913 }, { "epoch": 1.44, "learning_rate": 8.471760797342193e-08, "loss": 0.1952, "step": 6914 }, { "epoch": 1.44, "learning_rate": 8.468646179401992e-08, "loss": 0.1459, "step": 6915 }, { "epoch": 1.44, "learning_rate": 8.465531561461794e-08, "loss": 0.1493, "step": 6916 }, { "epoch": 1.44, "learning_rate": 8.462416943521594e-08, "loss": 0.1455, "step": 6917 }, { "epoch": 1.44, "learning_rate": 8.459302325581395e-08, "loss": 0.1923, "step": 6918 }, { "epoch": 1.44, "learning_rate": 8.456187707641196e-08, "loss": 0.2121, "step": 6919 }, { "epoch": 1.44, "learning_rate": 8.453073089700996e-08, "loss": 0.1597, "step": 6920 }, { "epoch": 1.44, "learning_rate": 8.449958471760798e-08, "loss": 0.2473, "step": 6921 }, { "epoch": 1.44, "learning_rate": 8.446843853820597e-08, "loss": 0.1885, "step": 6922 }, { "epoch": 1.44, "learning_rate": 8.443729235880398e-08, "loss": 0.157, "step": 6923 }, { "epoch": 1.44, "learning_rate": 8.440614617940198e-08, "loss": 0.1679, "step": 6924 }, { "epoch": 1.44, "learning_rate": 8.4375e-08, "loss": 0.161, "step": 6925 }, { "epoch": 1.44, "learning_rate": 8.434385382059802e-08, "loss": 0.2143, "step": 6926 }, { "epoch": 1.44, "learning_rate": 8.4312707641196e-08, "loss": 0.1924, "step": 6927 }, { "epoch": 1.44, "learning_rate": 8.428156146179402e-08, "loss": 0.2341, "step": 6928 }, { "epoch": 1.44, "learning_rate": 8.425041528239202e-08, "loss": 0.1591, "step": 6929 }, { "epoch": 1.44, "learning_rate": 8.421926910299003e-08, "loss": 0.1653, "step": 6930 }, { "epoch": 1.44, "learning_rate": 8.418812292358803e-08, "loss": 0.1458, "step": 6931 }, { "epoch": 1.44, "learning_rate": 8.415697674418604e-08, "loss": 0.2573, "step": 6932 }, { "epoch": 1.44, "learning_rate": 8.412583056478406e-08, "loss": 0.1635, "step": 6933 }, { "epoch": 1.44, "learning_rate": 8.409468438538205e-08, "loss": 0.1408, "step": 6934 }, { "epoch": 1.44, "learning_rate": 8.406353820598007e-08, "loss": 0.1761, "step": 6935 }, { "epoch": 1.44, "learning_rate": 8.403239202657807e-08, "loss": 0.1758, "step": 6936 }, { "epoch": 1.44, "learning_rate": 8.400124584717608e-08, "loss": 0.1941, "step": 6937 }, { "epoch": 1.44, "learning_rate": 8.397009966777407e-08, "loss": 0.1775, "step": 6938 }, { "epoch": 1.44, "learning_rate": 8.393895348837209e-08, "loss": 0.2342, "step": 6939 }, { "epoch": 1.44, "learning_rate": 8.39078073089701e-08, "loss": 0.1582, "step": 6940 }, { "epoch": 1.44, "learning_rate": 8.38766611295681e-08, "loss": 0.1774, "step": 6941 }, { "epoch": 1.44, "learning_rate": 8.384551495016611e-08, "loss": 0.1517, "step": 6942 }, { "epoch": 1.44, "learning_rate": 8.381436877076411e-08, "loss": 0.1334, "step": 6943 }, { "epoch": 1.44, "learning_rate": 8.378322259136213e-08, "loss": 0.1711, "step": 6944 }, { "epoch": 1.44, "learning_rate": 8.375207641196012e-08, "loss": 0.1685, "step": 6945 }, { "epoch": 1.44, "learning_rate": 8.372093023255813e-08, "loss": 0.1886, "step": 6946 }, { "epoch": 1.44, "learning_rate": 8.368978405315615e-08, "loss": 0.1356, "step": 6947 }, { "epoch": 1.44, "learning_rate": 8.365863787375415e-08, "loss": 0.1112, "step": 6948 }, { "epoch": 1.44, "learning_rate": 8.362749169435215e-08, "loss": 0.2038, "step": 6949 }, { "epoch": 1.44, "learning_rate": 8.359634551495016e-08, "loss": 0.2088, "step": 6950 }, { "epoch": 1.44, "learning_rate": 8.356519933554817e-08, "loss": 0.1448, "step": 6951 }, { "epoch": 1.44, "learning_rate": 8.353405315614619e-08, "loss": 0.1528, "step": 6952 }, { "epoch": 1.44, "learning_rate": 8.350290697674418e-08, "loss": 0.1922, "step": 6953 }, { "epoch": 1.44, "learning_rate": 8.34717607973422e-08, "loss": 0.2612, "step": 6954 }, { "epoch": 1.44, "learning_rate": 8.34406146179402e-08, "loss": 0.1448, "step": 6955 }, { "epoch": 1.44, "learning_rate": 8.34094684385382e-08, "loss": 0.1838, "step": 6956 }, { "epoch": 1.44, "learning_rate": 8.33783222591362e-08, "loss": 0.2024, "step": 6957 }, { "epoch": 1.44, "learning_rate": 8.334717607973422e-08, "loss": 0.1477, "step": 6958 }, { "epoch": 1.44, "learning_rate": 8.331602990033223e-08, "loss": 0.1758, "step": 6959 }, { "epoch": 1.45, "learning_rate": 8.328488372093022e-08, "loss": 0.1936, "step": 6960 }, { "epoch": 1.45, "learning_rate": 8.325373754152824e-08, "loss": 0.1593, "step": 6961 }, { "epoch": 1.45, "learning_rate": 8.322259136212624e-08, "loss": 0.1936, "step": 6962 }, { "epoch": 1.45, "learning_rate": 8.319144518272426e-08, "loss": 0.1832, "step": 6963 }, { "epoch": 1.45, "learning_rate": 8.316029900332225e-08, "loss": 0.1967, "step": 6964 }, { "epoch": 1.45, "learning_rate": 8.312915282392026e-08, "loss": 0.1596, "step": 6965 }, { "epoch": 1.45, "learning_rate": 8.309800664451828e-08, "loss": 0.1608, "step": 6966 }, { "epoch": 1.45, "learning_rate": 8.306686046511627e-08, "loss": 0.1627, "step": 6967 }, { "epoch": 1.45, "learning_rate": 8.303571428571428e-08, "loss": 0.1758, "step": 6968 }, { "epoch": 1.45, "learning_rate": 8.300456810631229e-08, "loss": 0.1635, "step": 6969 }, { "epoch": 1.45, "learning_rate": 8.29734219269103e-08, "loss": 0.1797, "step": 6970 }, { "epoch": 1.45, "learning_rate": 8.294227574750829e-08, "loss": 0.1574, "step": 6971 }, { "epoch": 1.45, "learning_rate": 8.29111295681063e-08, "loss": 0.1436, "step": 6972 }, { "epoch": 1.45, "learning_rate": 8.287998338870432e-08, "loss": 0.212, "step": 6973 }, { "epoch": 1.45, "learning_rate": 8.284883720930232e-08, "loss": 0.1791, "step": 6974 }, { "epoch": 1.45, "learning_rate": 8.281769102990033e-08, "loss": 0.2641, "step": 6975 }, { "epoch": 1.45, "learning_rate": 8.278654485049833e-08, "loss": 0.1095, "step": 6976 }, { "epoch": 1.45, "learning_rate": 8.275539867109635e-08, "loss": 0.2117, "step": 6977 }, { "epoch": 1.45, "learning_rate": 8.272425249169434e-08, "loss": 0.168, "step": 6978 }, { "epoch": 1.45, "learning_rate": 8.269310631229235e-08, "loss": 0.2351, "step": 6979 }, { "epoch": 1.45, "learning_rate": 8.266196013289037e-08, "loss": 0.1887, "step": 6980 }, { "epoch": 1.45, "learning_rate": 8.263081395348837e-08, "loss": 0.1554, "step": 6981 }, { "epoch": 1.45, "learning_rate": 8.259966777408639e-08, "loss": 0.1722, "step": 6982 }, { "epoch": 1.45, "learning_rate": 8.256852159468437e-08, "loss": 0.175, "step": 6983 }, { "epoch": 1.45, "learning_rate": 8.253737541528239e-08, "loss": 0.1565, "step": 6984 }, { "epoch": 1.45, "learning_rate": 8.250622923588039e-08, "loss": 0.1941, "step": 6985 }, { "epoch": 1.45, "learning_rate": 8.24750830564784e-08, "loss": 0.1459, "step": 6986 }, { "epoch": 1.45, "learning_rate": 8.244393687707641e-08, "loss": 0.2187, "step": 6987 }, { "epoch": 1.45, "learning_rate": 8.241279069767441e-08, "loss": 0.1649, "step": 6988 }, { "epoch": 1.45, "learning_rate": 8.238164451827243e-08, "loss": 0.199, "step": 6989 }, { "epoch": 1.45, "learning_rate": 8.235049833887042e-08, "loss": 0.2101, "step": 6990 }, { "epoch": 1.45, "learning_rate": 8.231935215946844e-08, "loss": 0.1975, "step": 6991 }, { "epoch": 1.45, "learning_rate": 8.228820598006645e-08, "loss": 0.2037, "step": 6992 }, { "epoch": 1.45, "learning_rate": 8.225705980066444e-08, "loss": 0.2005, "step": 6993 }, { "epoch": 1.45, "learning_rate": 8.222591362126246e-08, "loss": 0.1742, "step": 6994 }, { "epoch": 1.45, "learning_rate": 8.219476744186046e-08, "loss": 0.2223, "step": 6995 }, { "epoch": 1.45, "learning_rate": 8.216362126245847e-08, "loss": 0.2174, "step": 6996 }, { "epoch": 1.45, "learning_rate": 8.213247508305646e-08, "loss": 0.1966, "step": 6997 }, { "epoch": 1.45, "learning_rate": 8.210132890365448e-08, "loss": 0.1591, "step": 6998 }, { "epoch": 1.45, "learning_rate": 8.20701827242525e-08, "loss": 0.1699, "step": 6999 }, { "epoch": 1.45, "learning_rate": 8.20390365448505e-08, "loss": 0.1628, "step": 7000 }, { "epoch": 1.45, "learning_rate": 8.20078903654485e-08, "loss": 0.1905, "step": 7001 }, { "epoch": 1.45, "learning_rate": 8.19767441860465e-08, "loss": 0.1814, "step": 7002 }, { "epoch": 1.45, "learning_rate": 8.194559800664452e-08, "loss": 0.1963, "step": 7003 }, { "epoch": 1.45, "learning_rate": 8.191445182724251e-08, "loss": 0.1655, "step": 7004 }, { "epoch": 1.45, "learning_rate": 8.188330564784052e-08, "loss": 0.1073, "step": 7005 }, { "epoch": 1.45, "learning_rate": 8.185215946843854e-08, "loss": 0.1495, "step": 7006 }, { "epoch": 1.45, "learning_rate": 8.182101328903654e-08, "loss": 0.2019, "step": 7007 }, { "epoch": 1.46, "learning_rate": 8.178986710963456e-08, "loss": 0.1479, "step": 7008 }, { "epoch": 1.46, "learning_rate": 8.175872093023255e-08, "loss": 0.1782, "step": 7009 }, { "epoch": 1.46, "learning_rate": 8.172757475083056e-08, "loss": 0.1658, "step": 7010 }, { "epoch": 1.46, "learning_rate": 8.169642857142857e-08, "loss": 0.1643, "step": 7011 }, { "epoch": 1.46, "learning_rate": 8.166528239202657e-08, "loss": 0.1849, "step": 7012 }, { "epoch": 1.46, "learning_rate": 8.163413621262458e-08, "loss": 0.1702, "step": 7013 }, { "epoch": 1.46, "learning_rate": 8.160299003322259e-08, "loss": 0.1594, "step": 7014 }, { "epoch": 1.46, "learning_rate": 8.15718438538206e-08, "loss": 0.1877, "step": 7015 }, { "epoch": 1.46, "learning_rate": 8.154069767441859e-08, "loss": 0.205, "step": 7016 }, { "epoch": 1.46, "learning_rate": 8.150955149501661e-08, "loss": 0.2433, "step": 7017 }, { "epoch": 1.46, "learning_rate": 8.147840531561461e-08, "loss": 0.1349, "step": 7018 }, { "epoch": 1.46, "learning_rate": 8.144725913621263e-08, "loss": 0.2188, "step": 7019 }, { "epoch": 1.46, "learning_rate": 8.141611295681063e-08, "loss": 0.1937, "step": 7020 }, { "epoch": 1.46, "learning_rate": 8.138496677740863e-08, "loss": 0.1218, "step": 7021 }, { "epoch": 1.46, "learning_rate": 8.135382059800665e-08, "loss": 0.1516, "step": 7022 }, { "epoch": 1.46, "learning_rate": 8.132267441860464e-08, "loss": 0.1813, "step": 7023 }, { "epoch": 1.46, "learning_rate": 8.129152823920265e-08, "loss": 0.188, "step": 7024 }, { "epoch": 1.46, "learning_rate": 8.126038205980066e-08, "loss": 0.1839, "step": 7025 }, { "epoch": 1.46, "learning_rate": 8.122923588039867e-08, "loss": 0.1786, "step": 7026 }, { "epoch": 1.46, "learning_rate": 8.119808970099667e-08, "loss": 0.1389, "step": 7027 }, { "epoch": 1.46, "learning_rate": 8.116694352159468e-08, "loss": 0.25, "step": 7028 }, { "epoch": 1.46, "learning_rate": 8.113579734219269e-08, "loss": 0.1833, "step": 7029 }, { "epoch": 1.46, "learning_rate": 8.110465116279068e-08, "loss": 0.1552, "step": 7030 }, { "epoch": 1.46, "learning_rate": 8.10735049833887e-08, "loss": 0.1479, "step": 7031 }, { "epoch": 1.46, "learning_rate": 8.10423588039867e-08, "loss": 0.239, "step": 7032 }, { "epoch": 1.46, "learning_rate": 8.101121262458472e-08, "loss": 0.1339, "step": 7033 }, { "epoch": 1.46, "learning_rate": 8.098006644518273e-08, "loss": 0.201, "step": 7034 }, { "epoch": 1.46, "learning_rate": 8.094892026578072e-08, "loss": 0.2157, "step": 7035 }, { "epoch": 1.46, "learning_rate": 8.091777408637874e-08, "loss": 0.1316, "step": 7036 }, { "epoch": 1.46, "learning_rate": 8.088662790697674e-08, "loss": 0.1968, "step": 7037 }, { "epoch": 1.46, "learning_rate": 8.085548172757474e-08, "loss": 0.1466, "step": 7038 }, { "epoch": 1.46, "learning_rate": 8.082433554817276e-08, "loss": 0.1712, "step": 7039 }, { "epoch": 1.46, "learning_rate": 8.079318936877076e-08, "loss": 0.1158, "step": 7040 }, { "epoch": 1.46, "learning_rate": 8.076204318936878e-08, "loss": 0.1723, "step": 7041 }, { "epoch": 1.46, "learning_rate": 8.073089700996677e-08, "loss": 0.1859, "step": 7042 }, { "epoch": 1.46, "learning_rate": 8.069975083056478e-08, "loss": 0.1902, "step": 7043 }, { "epoch": 1.46, "learning_rate": 8.066860465116278e-08, "loss": 0.2107, "step": 7044 }, { "epoch": 1.46, "learning_rate": 8.06374584717608e-08, "loss": 0.174, "step": 7045 }, { "epoch": 1.46, "learning_rate": 8.06063122923588e-08, "loss": 0.2042, "step": 7046 }, { "epoch": 1.46, "learning_rate": 8.05751661129568e-08, "loss": 0.2107, "step": 7047 }, { "epoch": 1.46, "learning_rate": 8.054401993355482e-08, "loss": 0.1711, "step": 7048 }, { "epoch": 1.46, "learning_rate": 8.051287375415281e-08, "loss": 0.1269, "step": 7049 }, { "epoch": 1.46, "learning_rate": 8.048172757475083e-08, "loss": 0.2135, "step": 7050 }, { "epoch": 1.46, "learning_rate": 8.045058139534883e-08, "loss": 0.1573, "step": 7051 }, { "epoch": 1.46, "learning_rate": 8.041943521594684e-08, "loss": 0.2029, "step": 7052 }, { "epoch": 1.46, "learning_rate": 8.038828903654486e-08, "loss": 0.1647, "step": 7053 }, { "epoch": 1.46, "learning_rate": 8.035714285714285e-08, "loss": 0.1569, "step": 7054 }, { "epoch": 1.46, "learning_rate": 8.032599667774087e-08, "loss": 0.1991, "step": 7055 }, { "epoch": 1.47, "learning_rate": 8.029485049833887e-08, "loss": 0.137, "step": 7056 }, { "epoch": 1.47, "learning_rate": 8.026370431893687e-08, "loss": 0.1616, "step": 7057 }, { "epoch": 1.47, "learning_rate": 8.023255813953487e-08, "loss": 0.1455, "step": 7058 }, { "epoch": 1.47, "learning_rate": 8.020141196013289e-08, "loss": 0.1916, "step": 7059 }, { "epoch": 1.47, "learning_rate": 8.01702657807309e-08, "loss": 0.1688, "step": 7060 }, { "epoch": 1.47, "learning_rate": 8.01391196013289e-08, "loss": 0.1679, "step": 7061 }, { "epoch": 1.47, "learning_rate": 8.010797342192691e-08, "loss": 0.1651, "step": 7062 }, { "epoch": 1.47, "learning_rate": 8.007682724252491e-08, "loss": 0.1633, "step": 7063 }, { "epoch": 1.47, "learning_rate": 8.004568106312292e-08, "loss": 0.2093, "step": 7064 }, { "epoch": 1.47, "learning_rate": 8.001453488372092e-08, "loss": 0.1637, "step": 7065 }, { "epoch": 1.47, "learning_rate": 7.998338870431893e-08, "loss": 0.174, "step": 7066 }, { "epoch": 1.47, "learning_rate": 7.995224252491695e-08, "loss": 0.1556, "step": 7067 }, { "epoch": 1.47, "learning_rate": 7.992109634551494e-08, "loss": 0.2084, "step": 7068 }, { "epoch": 1.47, "learning_rate": 7.988995016611295e-08, "loss": 0.1584, "step": 7069 }, { "epoch": 1.47, "learning_rate": 7.985880398671096e-08, "loss": 0.169, "step": 7070 }, { "epoch": 1.47, "learning_rate": 7.982765780730897e-08, "loss": 0.2019, "step": 7071 }, { "epoch": 1.47, "learning_rate": 7.979651162790696e-08, "loss": 0.1497, "step": 7072 }, { "epoch": 1.47, "learning_rate": 7.976536544850498e-08, "loss": 0.1768, "step": 7073 }, { "epoch": 1.47, "learning_rate": 7.9734219269103e-08, "loss": 0.1807, "step": 7074 }, { "epoch": 1.47, "learning_rate": 7.970307308970098e-08, "loss": 0.1973, "step": 7075 }, { "epoch": 1.47, "learning_rate": 7.9671926910299e-08, "loss": 0.2018, "step": 7076 }, { "epoch": 1.47, "learning_rate": 7.9640780730897e-08, "loss": 0.185, "step": 7077 }, { "epoch": 1.47, "learning_rate": 7.960963455149502e-08, "loss": 0.175, "step": 7078 }, { "epoch": 1.47, "learning_rate": 7.957848837209303e-08, "loss": 0.1986, "step": 7079 }, { "epoch": 1.47, "learning_rate": 7.954734219269102e-08, "loss": 0.1603, "step": 7080 }, { "epoch": 1.47, "learning_rate": 7.951619601328904e-08, "loss": 0.1972, "step": 7081 }, { "epoch": 1.47, "learning_rate": 7.948504983388704e-08, "loss": 0.1709, "step": 7082 }, { "epoch": 1.47, "learning_rate": 7.945390365448504e-08, "loss": 0.1851, "step": 7083 }, { "epoch": 1.47, "learning_rate": 7.942275747508305e-08, "loss": 0.1748, "step": 7084 }, { "epoch": 1.47, "learning_rate": 7.939161129568106e-08, "loss": 0.1655, "step": 7085 }, { "epoch": 1.47, "learning_rate": 7.936046511627908e-08, "loss": 0.175, "step": 7086 }, { "epoch": 1.47, "learning_rate": 7.932931893687707e-08, "loss": 0.1576, "step": 7087 }, { "epoch": 1.47, "learning_rate": 7.929817275747508e-08, "loss": 0.1842, "step": 7088 }, { "epoch": 1.47, "learning_rate": 7.926702657807309e-08, "loss": 0.1765, "step": 7089 }, { "epoch": 1.47, "learning_rate": 7.92358803986711e-08, "loss": 0.1876, "step": 7090 }, { "epoch": 1.47, "learning_rate": 7.920473421926909e-08, "loss": 0.1855, "step": 7091 }, { "epoch": 1.47, "learning_rate": 7.917358803986711e-08, "loss": 0.203, "step": 7092 }, { "epoch": 1.47, "learning_rate": 7.914244186046512e-08, "loss": 0.1397, "step": 7093 }, { "epoch": 1.47, "learning_rate": 7.911129568106311e-08, "loss": 0.1205, "step": 7094 }, { "epoch": 1.47, "learning_rate": 7.908014950166113e-08, "loss": 0.1525, "step": 7095 }, { "epoch": 1.47, "learning_rate": 7.904900332225913e-08, "loss": 0.1427, "step": 7096 }, { "epoch": 1.47, "learning_rate": 7.901785714285715e-08, "loss": 0.154, "step": 7097 }, { "epoch": 1.47, "learning_rate": 7.898671096345514e-08, "loss": 0.1657, "step": 7098 }, { "epoch": 1.47, "learning_rate": 7.895556478405315e-08, "loss": 0.2225, "step": 7099 }, { "epoch": 1.47, "learning_rate": 7.892441860465117e-08, "loss": 0.186, "step": 7100 }, { "epoch": 1.47, "learning_rate": 7.889327242524916e-08, "loss": 0.1617, "step": 7101 }, { "epoch": 1.47, "learning_rate": 7.886212624584717e-08, "loss": 0.154, "step": 7102 }, { "epoch": 1.47, "learning_rate": 7.883098006644518e-08, "loss": 0.1808, "step": 7103 }, { "epoch": 1.48, "learning_rate": 7.879983388704319e-08, "loss": 0.1589, "step": 7104 }, { "epoch": 1.48, "learning_rate": 7.876868770764118e-08, "loss": 0.2074, "step": 7105 }, { "epoch": 1.48, "learning_rate": 7.87375415282392e-08, "loss": 0.1476, "step": 7106 }, { "epoch": 1.48, "learning_rate": 7.870639534883721e-08, "loss": 0.1822, "step": 7107 }, { "epoch": 1.48, "learning_rate": 7.867524916943521e-08, "loss": 0.2287, "step": 7108 }, { "epoch": 1.48, "learning_rate": 7.864410299003322e-08, "loss": 0.1694, "step": 7109 }, { "epoch": 1.48, "learning_rate": 7.861295681063122e-08, "loss": 0.1641, "step": 7110 }, { "epoch": 1.48, "learning_rate": 7.858181063122924e-08, "loss": 0.2263, "step": 7111 }, { "epoch": 1.48, "learning_rate": 7.855066445182722e-08, "loss": 0.1777, "step": 7112 }, { "epoch": 1.48, "learning_rate": 7.851951827242524e-08, "loss": 0.1263, "step": 7113 }, { "epoch": 1.48, "learning_rate": 7.848837209302326e-08, "loss": 0.1484, "step": 7114 }, { "epoch": 1.48, "learning_rate": 7.845722591362126e-08, "loss": 0.1673, "step": 7115 }, { "epoch": 1.48, "learning_rate": 7.842607973421927e-08, "loss": 0.1442, "step": 7116 }, { "epoch": 1.48, "learning_rate": 7.839493355481726e-08, "loss": 0.1594, "step": 7117 }, { "epoch": 1.48, "learning_rate": 7.836378737541528e-08, "loss": 0.1563, "step": 7118 }, { "epoch": 1.48, "learning_rate": 7.83326411960133e-08, "loss": 0.1654, "step": 7119 }, { "epoch": 1.48, "learning_rate": 7.830149501661129e-08, "loss": 0.1642, "step": 7120 }, { "epoch": 1.48, "learning_rate": 7.82703488372093e-08, "loss": 0.1393, "step": 7121 }, { "epoch": 1.48, "learning_rate": 7.82392026578073e-08, "loss": 0.1857, "step": 7122 }, { "epoch": 1.48, "learning_rate": 7.820805647840532e-08, "loss": 0.1441, "step": 7123 }, { "epoch": 1.48, "learning_rate": 7.817691029900331e-08, "loss": 0.2101, "step": 7124 }, { "epoch": 1.48, "learning_rate": 7.814576411960132e-08, "loss": 0.1849, "step": 7125 }, { "epoch": 1.48, "learning_rate": 7.811461794019934e-08, "loss": 0.1405, "step": 7126 }, { "epoch": 1.48, "learning_rate": 7.808347176079734e-08, "loss": 0.1573, "step": 7127 }, { "epoch": 1.48, "learning_rate": 7.805232558139535e-08, "loss": 0.182, "step": 7128 }, { "epoch": 1.48, "learning_rate": 7.802117940199335e-08, "loss": 0.1436, "step": 7129 }, { "epoch": 1.48, "learning_rate": 7.799003322259136e-08, "loss": 0.1635, "step": 7130 }, { "epoch": 1.48, "learning_rate": 7.795888704318935e-08, "loss": 0.1741, "step": 7131 }, { "epoch": 1.48, "learning_rate": 7.792774086378737e-08, "loss": 0.2322, "step": 7132 }, { "epoch": 1.48, "learning_rate": 7.789659468438539e-08, "loss": 0.1803, "step": 7133 }, { "epoch": 1.48, "learning_rate": 7.786544850498339e-08, "loss": 0.1916, "step": 7134 }, { "epoch": 1.48, "learning_rate": 7.783430232558139e-08, "loss": 0.1783, "step": 7135 }, { "epoch": 1.48, "learning_rate": 7.780315614617939e-08, "loss": 0.1341, "step": 7136 }, { "epoch": 1.48, "learning_rate": 7.777200996677741e-08, "loss": 0.1792, "step": 7137 }, { "epoch": 1.48, "learning_rate": 7.77408637873754e-08, "loss": 0.2208, "step": 7138 }, { "epoch": 1.48, "learning_rate": 7.770971760797341e-08, "loss": 0.1818, "step": 7139 }, { "epoch": 1.48, "learning_rate": 7.767857142857143e-08, "loss": 0.2267, "step": 7140 }, { "epoch": 1.48, "learning_rate": 7.764742524916943e-08, "loss": 0.208, "step": 7141 }, { "epoch": 1.48, "learning_rate": 7.761627906976745e-08, "loss": 0.1749, "step": 7142 }, { "epoch": 1.48, "learning_rate": 7.758513289036544e-08, "loss": 0.1664, "step": 7143 }, { "epoch": 1.48, "learning_rate": 7.755398671096345e-08, "loss": 0.1564, "step": 7144 }, { "epoch": 1.48, "learning_rate": 7.752284053156146e-08, "loss": 0.1793, "step": 7145 }, { "epoch": 1.48, "learning_rate": 7.749169435215946e-08, "loss": 0.1661, "step": 7146 }, { "epoch": 1.48, "learning_rate": 7.746054817275747e-08, "loss": 0.218, "step": 7147 }, { "epoch": 1.48, "learning_rate": 7.742940199335548e-08, "loss": 0.1954, "step": 7148 }, { "epoch": 1.48, "learning_rate": 7.739825581395349e-08, "loss": 0.1848, "step": 7149 }, { "epoch": 1.48, "learning_rate": 7.736710963455148e-08, "loss": 0.1596, "step": 7150 }, { "epoch": 1.48, "learning_rate": 7.73359634551495e-08, "loss": 0.16, "step": 7151 }, { "epoch": 1.49, "learning_rate": 7.73048172757475e-08, "loss": 0.1416, "step": 7152 }, { "epoch": 1.49, "learning_rate": 7.727367109634552e-08, "loss": 0.1652, "step": 7153 }, { "epoch": 1.49, "learning_rate": 7.724252491694352e-08, "loss": 0.1637, "step": 7154 }, { "epoch": 1.49, "learning_rate": 7.721137873754152e-08, "loss": 0.1693, "step": 7155 }, { "epoch": 1.49, "learning_rate": 7.718023255813954e-08, "loss": 0.1573, "step": 7156 }, { "epoch": 1.49, "learning_rate": 7.714908637873753e-08, "loss": 0.1895, "step": 7157 }, { "epoch": 1.49, "learning_rate": 7.711794019933554e-08, "loss": 0.2116, "step": 7158 }, { "epoch": 1.49, "learning_rate": 7.708679401993355e-08, "loss": 0.1787, "step": 7159 }, { "epoch": 1.49, "learning_rate": 7.705564784053156e-08, "loss": 0.1089, "step": 7160 }, { "epoch": 1.49, "learning_rate": 7.702450166112958e-08, "loss": 0.1863, "step": 7161 }, { "epoch": 1.49, "learning_rate": 7.699335548172757e-08, "loss": 0.1493, "step": 7162 }, { "epoch": 1.49, "learning_rate": 7.696220930232558e-08, "loss": 0.1631, "step": 7163 }, { "epoch": 1.49, "learning_rate": 7.693106312292358e-08, "loss": 0.2331, "step": 7164 }, { "epoch": 1.49, "learning_rate": 7.689991694352159e-08, "loss": 0.1701, "step": 7165 }, { "epoch": 1.49, "learning_rate": 7.68687707641196e-08, "loss": 0.2213, "step": 7166 }, { "epoch": 1.49, "learning_rate": 7.68376245847176e-08, "loss": 0.2134, "step": 7167 }, { "epoch": 1.49, "learning_rate": 7.680647840531562e-08, "loss": 0.1928, "step": 7168 }, { "epoch": 1.49, "learning_rate": 7.677533222591361e-08, "loss": 0.1962, "step": 7169 }, { "epoch": 1.49, "learning_rate": 7.674418604651163e-08, "loss": 0.1298, "step": 7170 }, { "epoch": 1.49, "learning_rate": 7.671303986710963e-08, "loss": 0.2322, "step": 7171 }, { "epoch": 1.49, "learning_rate": 7.668189368770763e-08, "loss": 0.1984, "step": 7172 }, { "epoch": 1.49, "learning_rate": 7.665074750830565e-08, "loss": 0.24, "step": 7173 }, { "epoch": 1.49, "learning_rate": 7.661960132890365e-08, "loss": 0.1781, "step": 7174 }, { "epoch": 1.49, "learning_rate": 7.658845514950167e-08, "loss": 0.1687, "step": 7175 }, { "epoch": 1.49, "learning_rate": 7.655730897009966e-08, "loss": 0.2153, "step": 7176 }, { "epoch": 1.49, "learning_rate": 7.652616279069767e-08, "loss": 0.1723, "step": 7177 }, { "epoch": 1.49, "learning_rate": 7.649501661129567e-08, "loss": 0.2191, "step": 7178 }, { "epoch": 1.49, "learning_rate": 7.646387043189369e-08, "loss": 0.1805, "step": 7179 }, { "epoch": 1.49, "learning_rate": 7.643272425249169e-08, "loss": 0.1721, "step": 7180 }, { "epoch": 1.49, "learning_rate": 7.64015780730897e-08, "loss": 0.1919, "step": 7181 }, { "epoch": 1.49, "learning_rate": 7.637043189368771e-08, "loss": 0.1911, "step": 7182 }, { "epoch": 1.49, "learning_rate": 7.63392857142857e-08, "loss": 0.1661, "step": 7183 }, { "epoch": 1.49, "learning_rate": 7.630813953488372e-08, "loss": 0.1769, "step": 7184 }, { "epoch": 1.49, "learning_rate": 7.627699335548172e-08, "loss": 0.1762, "step": 7185 }, { "epoch": 1.49, "learning_rate": 7.624584717607973e-08, "loss": 0.1176, "step": 7186 }, { "epoch": 1.49, "learning_rate": 7.621470099667775e-08, "loss": 0.1656, "step": 7187 }, { "epoch": 1.49, "learning_rate": 7.618355481727574e-08, "loss": 0.1549, "step": 7188 }, { "epoch": 1.49, "learning_rate": 7.615240863787376e-08, "loss": 0.1592, "step": 7189 }, { "epoch": 1.49, "learning_rate": 7.612126245847176e-08, "loss": 0.1392, "step": 7190 }, { "epoch": 1.49, "learning_rate": 7.609011627906976e-08, "loss": 0.1931, "step": 7191 }, { "epoch": 1.49, "learning_rate": 7.605897009966776e-08, "loss": 0.2441, "step": 7192 }, { "epoch": 1.49, "learning_rate": 7.602782392026578e-08, "loss": 0.1731, "step": 7193 }, { "epoch": 1.49, "learning_rate": 7.59966777408638e-08, "loss": 0.2457, "step": 7194 }, { "epoch": 1.49, "learning_rate": 7.596553156146178e-08, "loss": 0.1856, "step": 7195 }, { "epoch": 1.49, "learning_rate": 7.59343853820598e-08, "loss": 0.1242, "step": 7196 }, { "epoch": 1.49, "learning_rate": 7.59032392026578e-08, "loss": 0.1713, "step": 7197 }, { "epoch": 1.49, "learning_rate": 7.587209302325582e-08, "loss": 0.1779, "step": 7198 }, { "epoch": 1.49, "learning_rate": 7.584094684385381e-08, "loss": 0.2074, "step": 7199 }, { "epoch": 1.5, "learning_rate": 7.580980066445182e-08, "loss": 0.1371, "step": 7200 }, { "epoch": 1.5, "learning_rate": 7.577865448504984e-08, "loss": 0.1615, "step": 7201 }, { "epoch": 1.5, "learning_rate": 7.574750830564783e-08, "loss": 0.1496, "step": 7202 }, { "epoch": 1.5, "learning_rate": 7.571636212624584e-08, "loss": 0.1849, "step": 7203 }, { "epoch": 1.5, "learning_rate": 7.568521594684385e-08, "loss": 0.195, "step": 7204 }, { "epoch": 1.5, "learning_rate": 7.565406976744186e-08, "loss": 0.206, "step": 7205 }, { "epoch": 1.5, "learning_rate": 7.562292358803987e-08, "loss": 0.203, "step": 7206 }, { "epoch": 1.5, "learning_rate": 7.559177740863787e-08, "loss": 0.2216, "step": 7207 }, { "epoch": 1.5, "learning_rate": 7.556063122923588e-08, "loss": 0.1924, "step": 7208 }, { "epoch": 1.5, "learning_rate": 7.552948504983389e-08, "loss": 0.2288, "step": 7209 }, { "epoch": 1.5, "learning_rate": 7.549833887043189e-08, "loss": 0.1788, "step": 7210 }, { "epoch": 1.5, "learning_rate": 7.546719269102989e-08, "loss": 0.1714, "step": 7211 }, { "epoch": 1.5, "learning_rate": 7.543604651162791e-08, "loss": 0.1914, "step": 7212 }, { "epoch": 1.5, "learning_rate": 7.540490033222592e-08, "loss": 0.21, "step": 7213 }, { "epoch": 1.5, "learning_rate": 7.537375415282391e-08, "loss": 0.1966, "step": 7214 }, { "epoch": 1.5, "learning_rate": 7.534260797342193e-08, "loss": 0.1482, "step": 7215 }, { "epoch": 1.5, "learning_rate": 7.531146179401993e-08, "loss": 0.1697, "step": 7216 }, { "epoch": 1.5, "learning_rate": 7.528031561461793e-08, "loss": 0.1652, "step": 7217 }, { "epoch": 1.5, "learning_rate": 7.524916943521594e-08, "loss": 0.1733, "step": 7218 }, { "epoch": 1.5, "learning_rate": 7.521802325581395e-08, "loss": 0.2144, "step": 7219 }, { "epoch": 1.5, "learning_rate": 7.518687707641197e-08, "loss": 0.1478, "step": 7220 }, { "epoch": 1.5, "learning_rate": 7.515573089700996e-08, "loss": 0.1944, "step": 7221 }, { "epoch": 1.5, "learning_rate": 7.512458471760797e-08, "loss": 0.1791, "step": 7222 }, { "epoch": 1.5, "learning_rate": 7.509343853820598e-08, "loss": 0.2422, "step": 7223 }, { "epoch": 1.5, "learning_rate": 7.506229235880399e-08, "loss": 0.1206, "step": 7224 }, { "epoch": 1.5, "learning_rate": 7.503114617940198e-08, "loss": 0.1712, "step": 7225 }, { "epoch": 1.5, "learning_rate": 7.5e-08, "loss": 0.2012, "step": 7226 }, { "epoch": 1.5, "learning_rate": 7.4968853820598e-08, "loss": 0.2133, "step": 7227 }, { "epoch": 1.5, "learning_rate": 7.4937707641196e-08, "loss": 0.1988, "step": 7228 }, { "epoch": 1.5, "learning_rate": 7.490656146179402e-08, "loss": 0.1447, "step": 7229 }, { "epoch": 1.5, "learning_rate": 7.487541528239202e-08, "loss": 0.2235, "step": 7230 }, { "epoch": 1.5, "learning_rate": 7.484426910299004e-08, "loss": 0.1342, "step": 7231 }, { "epoch": 1.5, "learning_rate": 7.481312292358804e-08, "loss": 0.1196, "step": 7232 }, { "epoch": 1.5, "learning_rate": 7.478197674418604e-08, "loss": 0.1631, "step": 7233 }, { "epoch": 1.5, "learning_rate": 7.475083056478404e-08, "loss": 0.1339, "step": 7234 }, { "epoch": 1.5, "learning_rate": 7.471968438538206e-08, "loss": 0.1428, "step": 7235 }, { "epoch": 1.5, "learning_rate": 7.468853820598006e-08, "loss": 0.1624, "step": 7236 }, { "epoch": 1.5, "learning_rate": 7.465739202657808e-08, "loss": 0.18, "step": 7237 }, { "epoch": 1.5, "learning_rate": 7.462624584717608e-08, "loss": 0.1904, "step": 7238 }, { "epoch": 1.5, "learning_rate": 7.459509966777408e-08, "loss": 0.2269, "step": 7239 }, { "epoch": 1.5, "learning_rate": 7.456395348837209e-08, "loss": 0.1452, "step": 7240 }, { "epoch": 1.5, "learning_rate": 7.453280730897009e-08, "loss": 0.1488, "step": 7241 }, { "epoch": 1.5, "learning_rate": 7.45016611295681e-08, "loss": 0.1791, "step": 7242 }, { "epoch": 1.5, "learning_rate": 7.447051495016611e-08, "loss": 0.1494, "step": 7243 }, { "epoch": 1.5, "learning_rate": 7.443936877076412e-08, "loss": 0.1748, "step": 7244 }, { "epoch": 1.5, "learning_rate": 7.440822259136213e-08, "loss": 0.1434, "step": 7245 }, { "epoch": 1.5, "learning_rate": 7.437707641196013e-08, "loss": 0.1313, "step": 7246 }, { "epoch": 1.5, "learning_rate": 7.434593023255813e-08, "loss": 0.2433, "step": 7247 }, { "epoch": 1.5, "learning_rate": 7.431478405315615e-08, "loss": 0.1559, "step": 7248 }, { "epoch": 1.51, "learning_rate": 7.428363787375415e-08, "loss": 0.181, "step": 7249 }, { "epoch": 1.51, "learning_rate": 7.425249169435216e-08, "loss": 0.1676, "step": 7250 }, { "epoch": 1.51, "learning_rate": 7.422134551495017e-08, "loss": 0.1501, "step": 7251 }, { "epoch": 1.51, "learning_rate": 7.419019933554817e-08, "loss": 0.1596, "step": 7252 }, { "epoch": 1.51, "learning_rate": 7.415905315614617e-08, "loss": 0.1677, "step": 7253 }, { "epoch": 1.51, "learning_rate": 7.412790697674417e-08, "loss": 0.2383, "step": 7254 }, { "epoch": 1.51, "learning_rate": 7.409676079734219e-08, "loss": 0.2873, "step": 7255 }, { "epoch": 1.51, "learning_rate": 7.40656146179402e-08, "loss": 0.1721, "step": 7256 }, { "epoch": 1.51, "learning_rate": 7.403446843853821e-08, "loss": 0.2037, "step": 7257 }, { "epoch": 1.51, "learning_rate": 7.400332225913621e-08, "loss": 0.1366, "step": 7258 }, { "epoch": 1.51, "learning_rate": 7.397217607973421e-08, "loss": 0.2011, "step": 7259 }, { "epoch": 1.51, "learning_rate": 7.394102990033222e-08, "loss": 0.1341, "step": 7260 }, { "epoch": 1.51, "learning_rate": 7.390988372093023e-08, "loss": 0.1962, "step": 7261 }, { "epoch": 1.51, "learning_rate": 7.387873754152824e-08, "loss": 0.1834, "step": 7262 }, { "epoch": 1.51, "learning_rate": 7.384759136212625e-08, "loss": 0.2502, "step": 7263 }, { "epoch": 1.51, "learning_rate": 7.381644518272425e-08, "loss": 0.2262, "step": 7264 }, { "epoch": 1.51, "learning_rate": 7.378529900332226e-08, "loss": 0.1816, "step": 7265 }, { "epoch": 1.51, "learning_rate": 7.375415282392026e-08, "loss": 0.1235, "step": 7266 }, { "epoch": 1.51, "learning_rate": 7.372300664451826e-08, "loss": 0.1395, "step": 7267 }, { "epoch": 1.51, "learning_rate": 7.369186046511628e-08, "loss": 0.1846, "step": 7268 }, { "epoch": 1.51, "learning_rate": 7.366071428571428e-08, "loss": 0.2002, "step": 7269 }, { "epoch": 1.51, "learning_rate": 7.36295681063123e-08, "loss": 0.1786, "step": 7270 }, { "epoch": 1.51, "learning_rate": 7.35984219269103e-08, "loss": 0.1627, "step": 7271 }, { "epoch": 1.51, "learning_rate": 7.35672757475083e-08, "loss": 0.2009, "step": 7272 }, { "epoch": 1.51, "learning_rate": 7.35361295681063e-08, "loss": 0.1672, "step": 7273 }, { "epoch": 1.51, "learning_rate": 7.350498338870432e-08, "loss": 0.2237, "step": 7274 }, { "epoch": 1.51, "learning_rate": 7.347383720930232e-08, "loss": 0.2202, "step": 7275 }, { "epoch": 1.51, "learning_rate": 7.344269102990032e-08, "loss": 0.2303, "step": 7276 }, { "epoch": 1.51, "learning_rate": 7.341154485049834e-08, "loss": 0.1784, "step": 7277 }, { "epoch": 1.51, "learning_rate": 7.338039867109634e-08, "loss": 0.1613, "step": 7278 }, { "epoch": 1.51, "learning_rate": 7.334925249169435e-08, "loss": 0.1511, "step": 7279 }, { "epoch": 1.51, "learning_rate": 7.331810631229236e-08, "loss": 0.1496, "step": 7280 }, { "epoch": 1.51, "learning_rate": 7.328696013289036e-08, "loss": 0.1728, "step": 7281 }, { "epoch": 1.51, "learning_rate": 7.325581395348837e-08, "loss": 0.1723, "step": 7282 }, { "epoch": 1.51, "learning_rate": 7.322466777408638e-08, "loss": 0.1071, "step": 7283 }, { "epoch": 1.51, "learning_rate": 7.319352159468439e-08, "loss": 0.2125, "step": 7284 }, { "epoch": 1.51, "learning_rate": 7.316237541528239e-08, "loss": 0.1602, "step": 7285 }, { "epoch": 1.51, "learning_rate": 7.313122923588039e-08, "loss": 0.1703, "step": 7286 }, { "epoch": 1.51, "learning_rate": 7.31000830564784e-08, "loss": 0.2187, "step": 7287 }, { "epoch": 1.51, "learning_rate": 7.306893687707641e-08, "loss": 0.1267, "step": 7288 }, { "epoch": 1.51, "learning_rate": 7.303779069767441e-08, "loss": 0.1606, "step": 7289 }, { "epoch": 1.51, "learning_rate": 7.300664451827243e-08, "loss": 0.1615, "step": 7290 }, { "epoch": 1.51, "learning_rate": 7.297549833887043e-08, "loss": 0.162, "step": 7291 }, { "epoch": 1.51, "learning_rate": 7.294435215946843e-08, "loss": 0.1568, "step": 7292 }, { "epoch": 1.51, "learning_rate": 7.291320598006645e-08, "loss": 0.2251, "step": 7293 }, { "epoch": 1.51, "learning_rate": 7.288205980066445e-08, "loss": 0.1721, "step": 7294 }, { "epoch": 1.51, "learning_rate": 7.285091362126245e-08, "loss": 0.1898, "step": 7295 }, { "epoch": 1.51, "learning_rate": 7.281976744186046e-08, "loss": 0.1966, "step": 7296 }, { "epoch": 1.52, "learning_rate": 7.278862126245847e-08, "loss": 0.15, "step": 7297 }, { "epoch": 1.52, "learning_rate": 7.275747508305647e-08, "loss": 0.1736, "step": 7298 }, { "epoch": 1.52, "learning_rate": 7.272632890365448e-08, "loss": 0.1796, "step": 7299 }, { "epoch": 1.52, "learning_rate": 7.269518272425249e-08, "loss": 0.1711, "step": 7300 }, { "epoch": 1.52, "learning_rate": 7.26640365448505e-08, "loss": 0.2169, "step": 7301 }, { "epoch": 1.52, "learning_rate": 7.26328903654485e-08, "loss": 0.1509, "step": 7302 }, { "epoch": 1.52, "learning_rate": 7.260174418604651e-08, "loss": 0.2025, "step": 7303 }, { "epoch": 1.52, "learning_rate": 7.257059800664452e-08, "loss": 0.2195, "step": 7304 }, { "epoch": 1.52, "learning_rate": 7.253945182724252e-08, "loss": 0.1634, "step": 7305 }, { "epoch": 1.52, "learning_rate": 7.250830564784053e-08, "loss": 0.2112, "step": 7306 }, { "epoch": 1.52, "learning_rate": 7.247715946843854e-08, "loss": 0.1689, "step": 7307 }, { "epoch": 1.52, "learning_rate": 7.244601328903654e-08, "loss": 0.1788, "step": 7308 }, { "epoch": 1.52, "learning_rate": 7.241486710963454e-08, "loss": 0.1311, "step": 7309 }, { "epoch": 1.52, "learning_rate": 7.238372093023256e-08, "loss": 0.1539, "step": 7310 }, { "epoch": 1.52, "learning_rate": 7.235257475083056e-08, "loss": 0.1516, "step": 7311 }, { "epoch": 1.52, "learning_rate": 7.232142857142856e-08, "loss": 0.2344, "step": 7312 }, { "epoch": 1.52, "learning_rate": 7.229028239202658e-08, "loss": 0.191, "step": 7313 }, { "epoch": 1.52, "learning_rate": 7.225913621262458e-08, "loss": 0.1316, "step": 7314 }, { "epoch": 1.52, "learning_rate": 7.222799003322258e-08, "loss": 0.1431, "step": 7315 }, { "epoch": 1.52, "learning_rate": 7.219684385382059e-08, "loss": 0.1858, "step": 7316 }, { "epoch": 1.52, "learning_rate": 7.21656976744186e-08, "loss": 0.1701, "step": 7317 }, { "epoch": 1.52, "learning_rate": 7.21345514950166e-08, "loss": 0.1765, "step": 7318 }, { "epoch": 1.52, "learning_rate": 7.210340531561462e-08, "loss": 0.2383, "step": 7319 }, { "epoch": 1.52, "learning_rate": 7.207225913621262e-08, "loss": 0.2464, "step": 7320 }, { "epoch": 1.52, "learning_rate": 7.204111295681063e-08, "loss": 0.2237, "step": 7321 }, { "epoch": 1.52, "learning_rate": 7.200996677740863e-08, "loss": 0.1902, "step": 7322 }, { "epoch": 1.52, "learning_rate": 7.197882059800664e-08, "loss": 0.1356, "step": 7323 }, { "epoch": 1.52, "learning_rate": 7.194767441860465e-08, "loss": 0.2434, "step": 7324 }, { "epoch": 1.52, "learning_rate": 7.191652823920265e-08, "loss": 0.2282, "step": 7325 }, { "epoch": 1.52, "learning_rate": 7.188538205980067e-08, "loss": 0.1808, "step": 7326 }, { "epoch": 1.52, "learning_rate": 7.185423588039867e-08, "loss": 0.132, "step": 7327 }, { "epoch": 1.52, "learning_rate": 7.182308970099667e-08, "loss": 0.2067, "step": 7328 }, { "epoch": 1.52, "learning_rate": 7.179194352159467e-08, "loss": 0.2278, "step": 7329 }, { "epoch": 1.52, "learning_rate": 7.176079734219269e-08, "loss": 0.1596, "step": 7330 }, { "epoch": 1.52, "learning_rate": 7.172965116279069e-08, "loss": 0.2479, "step": 7331 }, { "epoch": 1.52, "learning_rate": 7.169850498338871e-08, "loss": 0.1721, "step": 7332 }, { "epoch": 1.52, "learning_rate": 7.166735880398671e-08, "loss": 0.2506, "step": 7333 }, { "epoch": 1.52, "learning_rate": 7.163621262458471e-08, "loss": 0.117, "step": 7334 }, { "epoch": 1.52, "learning_rate": 7.160506644518272e-08, "loss": 0.1937, "step": 7335 }, { "epoch": 1.52, "learning_rate": 7.157392026578072e-08, "loss": 0.1115, "step": 7336 }, { "epoch": 1.52, "learning_rate": 7.154277408637873e-08, "loss": 0.2085, "step": 7337 }, { "epoch": 1.52, "learning_rate": 7.151162790697674e-08, "loss": 0.1769, "step": 7338 }, { "epoch": 1.52, "learning_rate": 7.148048172757475e-08, "loss": 0.1512, "step": 7339 }, { "epoch": 1.52, "learning_rate": 7.144933554817276e-08, "loss": 0.1998, "step": 7340 }, { "epoch": 1.52, "learning_rate": 7.141818936877076e-08, "loss": 0.1715, "step": 7341 }, { "epoch": 1.52, "learning_rate": 7.138704318936876e-08, "loss": 0.1565, "step": 7342 }, { "epoch": 1.52, "learning_rate": 7.135589700996678e-08, "loss": 0.1371, "step": 7343 }, { "epoch": 1.52, "learning_rate": 7.132475083056478e-08, "loss": 0.1741, "step": 7344 }, { "epoch": 1.53, "learning_rate": 7.12936046511628e-08, "loss": 0.2251, "step": 7345 }, { "epoch": 1.53, "learning_rate": 7.12624584717608e-08, "loss": 0.1719, "step": 7346 }, { "epoch": 1.53, "learning_rate": 7.12313122923588e-08, "loss": 0.2071, "step": 7347 }, { "epoch": 1.53, "learning_rate": 7.12001661129568e-08, "loss": 0.1977, "step": 7348 }, { "epoch": 1.53, "learning_rate": 7.11690199335548e-08, "loss": 0.1686, "step": 7349 }, { "epoch": 1.53, "learning_rate": 7.113787375415282e-08, "loss": 0.1826, "step": 7350 }, { "epoch": 1.53, "learning_rate": 7.110672757475084e-08, "loss": 0.1711, "step": 7351 }, { "epoch": 1.53, "learning_rate": 7.107558139534884e-08, "loss": 0.1629, "step": 7352 }, { "epoch": 1.53, "learning_rate": 7.104443521594684e-08, "loss": 0.2024, "step": 7353 }, { "epoch": 1.53, "learning_rate": 7.101328903654484e-08, "loss": 0.2077, "step": 7354 }, { "epoch": 1.53, "learning_rate": 7.098214285714285e-08, "loss": 0.2082, "step": 7355 }, { "epoch": 1.53, "learning_rate": 7.095099667774086e-08, "loss": 0.1832, "step": 7356 }, { "epoch": 1.53, "learning_rate": 7.091985049833887e-08, "loss": 0.1156, "step": 7357 }, { "epoch": 1.53, "learning_rate": 7.088870431893688e-08, "loss": 0.1491, "step": 7358 }, { "epoch": 1.53, "learning_rate": 7.085755813953488e-08, "loss": 0.1962, "step": 7359 }, { "epoch": 1.53, "learning_rate": 7.082641196013289e-08, "loss": 0.2168, "step": 7360 }, { "epoch": 1.53, "learning_rate": 7.079526578073089e-08, "loss": 0.1548, "step": 7361 }, { "epoch": 1.53, "learning_rate": 7.076411960132889e-08, "loss": 0.163, "step": 7362 }, { "epoch": 1.53, "learning_rate": 7.073297342192691e-08, "loss": 0.1952, "step": 7363 }, { "epoch": 1.53, "learning_rate": 7.070182724252492e-08, "loss": 0.2182, "step": 7364 }, { "epoch": 1.53, "learning_rate": 7.067068106312293e-08, "loss": 0.1865, "step": 7365 }, { "epoch": 1.53, "learning_rate": 7.063953488372093e-08, "loss": 0.1951, "step": 7366 }, { "epoch": 1.53, "learning_rate": 7.060838870431893e-08, "loss": 0.1975, "step": 7367 }, { "epoch": 1.53, "learning_rate": 7.057724252491693e-08, "loss": 0.1431, "step": 7368 }, { "epoch": 1.53, "learning_rate": 7.054609634551495e-08, "loss": 0.1588, "step": 7369 }, { "epoch": 1.53, "learning_rate": 7.051495016611295e-08, "loss": 0.1729, "step": 7370 }, { "epoch": 1.53, "learning_rate": 7.048380398671097e-08, "loss": 0.2111, "step": 7371 }, { "epoch": 1.53, "learning_rate": 7.045265780730897e-08, "loss": 0.1679, "step": 7372 }, { "epoch": 1.53, "learning_rate": 7.042151162790697e-08, "loss": 0.2114, "step": 7373 }, { "epoch": 1.53, "learning_rate": 7.039036544850498e-08, "loss": 0.2365, "step": 7374 }, { "epoch": 1.53, "learning_rate": 7.035921926910298e-08, "loss": 0.2048, "step": 7375 }, { "epoch": 1.53, "learning_rate": 7.0328073089701e-08, "loss": 0.1691, "step": 7376 }, { "epoch": 1.53, "learning_rate": 7.029692691029901e-08, "loss": 0.1986, "step": 7377 }, { "epoch": 1.53, "learning_rate": 7.026578073089701e-08, "loss": 0.2148, "step": 7378 }, { "epoch": 1.53, "learning_rate": 7.023463455149501e-08, "loss": 0.2034, "step": 7379 }, { "epoch": 1.53, "learning_rate": 7.020348837209302e-08, "loss": 0.2011, "step": 7380 }, { "epoch": 1.53, "learning_rate": 7.017234219269102e-08, "loss": 0.1585, "step": 7381 }, { "epoch": 1.53, "learning_rate": 7.014119601328904e-08, "loss": 0.2075, "step": 7382 }, { "epoch": 1.53, "learning_rate": 7.011004983388704e-08, "loss": 0.1897, "step": 7383 }, { "epoch": 1.53, "learning_rate": 7.007890365448505e-08, "loss": 0.1889, "step": 7384 }, { "epoch": 1.53, "learning_rate": 7.004775747508306e-08, "loss": 0.1897, "step": 7385 }, { "epoch": 1.53, "learning_rate": 7.001661129568106e-08, "loss": 0.122, "step": 7386 }, { "epoch": 1.53, "learning_rate": 6.998546511627906e-08, "loss": 0.1578, "step": 7387 }, { "epoch": 1.53, "learning_rate": 6.995431893687708e-08, "loss": 0.1891, "step": 7388 }, { "epoch": 1.53, "learning_rate": 6.992317275747508e-08, "loss": 0.202, "step": 7389 }, { "epoch": 1.53, "learning_rate": 6.98920265780731e-08, "loss": 0.1581, "step": 7390 }, { "epoch": 1.53, "learning_rate": 6.98608803986711e-08, "loss": 0.1796, "step": 7391 }, { "epoch": 1.53, "learning_rate": 6.98297342192691e-08, "loss": 0.1281, "step": 7392 }, { "epoch": 1.54, "learning_rate": 6.97985880398671e-08, "loss": 0.1744, "step": 7393 }, { "epoch": 1.54, "learning_rate": 6.976744186046511e-08, "loss": 0.1431, "step": 7394 }, { "epoch": 1.54, "learning_rate": 6.973629568106312e-08, "loss": 0.142, "step": 7395 }, { "epoch": 1.54, "learning_rate": 6.970514950166113e-08, "loss": 0.2187, "step": 7396 }, { "epoch": 1.54, "learning_rate": 6.967400332225914e-08, "loss": 0.1451, "step": 7397 }, { "epoch": 1.54, "learning_rate": 6.964285714285714e-08, "loss": 0.1926, "step": 7398 }, { "epoch": 1.54, "learning_rate": 6.961171096345515e-08, "loss": 0.2114, "step": 7399 }, { "epoch": 1.54, "learning_rate": 6.958056478405315e-08, "loss": 0.1868, "step": 7400 }, { "epoch": 1.54, "learning_rate": 6.954941860465116e-08, "loss": 0.1666, "step": 7401 }, { "epoch": 1.54, "learning_rate": 6.951827242524917e-08, "loss": 0.1669, "step": 7402 }, { "epoch": 1.54, "learning_rate": 6.948712624584717e-08, "loss": 0.1793, "step": 7403 }, { "epoch": 1.54, "learning_rate": 6.945598006644519e-08, "loss": 0.1712, "step": 7404 }, { "epoch": 1.54, "learning_rate": 6.942483388704319e-08, "loss": 0.1937, "step": 7405 }, { "epoch": 1.54, "learning_rate": 6.939368770764119e-08, "loss": 0.181, "step": 7406 }, { "epoch": 1.54, "learning_rate": 6.936254152823919e-08, "loss": 0.1797, "step": 7407 }, { "epoch": 1.54, "learning_rate": 6.933139534883721e-08, "loss": 0.2078, "step": 7408 }, { "epoch": 1.54, "learning_rate": 6.930024916943521e-08, "loss": 0.1745, "step": 7409 }, { "epoch": 1.54, "learning_rate": 6.926910299003323e-08, "loss": 0.2464, "step": 7410 }, { "epoch": 1.54, "learning_rate": 6.923795681063123e-08, "loss": 0.1759, "step": 7411 }, { "epoch": 1.54, "learning_rate": 6.920681063122923e-08, "loss": 0.1541, "step": 7412 }, { "epoch": 1.54, "learning_rate": 6.917566445182724e-08, "loss": 0.1507, "step": 7413 }, { "epoch": 1.54, "learning_rate": 6.914451827242525e-08, "loss": 0.1521, "step": 7414 }, { "epoch": 1.54, "learning_rate": 6.911337209302325e-08, "loss": 0.192, "step": 7415 }, { "epoch": 1.54, "learning_rate": 6.908222591362126e-08, "loss": 0.1935, "step": 7416 }, { "epoch": 1.54, "learning_rate": 6.905107973421927e-08, "loss": 0.1439, "step": 7417 }, { "epoch": 1.54, "learning_rate": 6.901993355481727e-08, "loss": 0.1838, "step": 7418 }, { "epoch": 1.54, "learning_rate": 6.898878737541528e-08, "loss": 0.1427, "step": 7419 }, { "epoch": 1.54, "learning_rate": 6.895764119601328e-08, "loss": 0.1862, "step": 7420 }, { "epoch": 1.54, "learning_rate": 6.89264950166113e-08, "loss": 0.1368, "step": 7421 }, { "epoch": 1.54, "learning_rate": 6.88953488372093e-08, "loss": 0.1568, "step": 7422 }, { "epoch": 1.54, "learning_rate": 6.88642026578073e-08, "loss": 0.1376, "step": 7423 }, { "epoch": 1.54, "learning_rate": 6.883305647840532e-08, "loss": 0.2068, "step": 7424 }, { "epoch": 1.54, "learning_rate": 6.880191029900332e-08, "loss": 0.1517, "step": 7425 }, { "epoch": 1.54, "learning_rate": 6.877076411960132e-08, "loss": 0.167, "step": 7426 }, { "epoch": 1.54, "learning_rate": 6.873961794019934e-08, "loss": 0.1573, "step": 7427 }, { "epoch": 1.54, "learning_rate": 6.870847176079734e-08, "loss": 0.1803, "step": 7428 }, { "epoch": 1.54, "learning_rate": 6.867732558139534e-08, "loss": 0.1921, "step": 7429 }, { "epoch": 1.54, "learning_rate": 6.864617940199335e-08, "loss": 0.2038, "step": 7430 }, { "epoch": 1.54, "learning_rate": 6.861503322259136e-08, "loss": 0.2189, "step": 7431 }, { "epoch": 1.54, "learning_rate": 6.858388704318936e-08, "loss": 0.1215, "step": 7432 }, { "epoch": 1.54, "learning_rate": 6.855274086378737e-08, "loss": 0.1464, "step": 7433 }, { "epoch": 1.54, "learning_rate": 6.852159468438538e-08, "loss": 0.2513, "step": 7434 }, { "epoch": 1.54, "learning_rate": 6.849044850498338e-08, "loss": 0.202, "step": 7435 }, { "epoch": 1.54, "learning_rate": 6.845930232558139e-08, "loss": 0.2009, "step": 7436 }, { "epoch": 1.54, "learning_rate": 6.84281561461794e-08, "loss": 0.1651, "step": 7437 }, { "epoch": 1.54, "learning_rate": 6.83970099667774e-08, "loss": 0.2238, "step": 7438 }, { "epoch": 1.54, "learning_rate": 6.836586378737541e-08, "loss": 0.1662, "step": 7439 }, { "epoch": 1.54, "learning_rate": 6.833471760797342e-08, "loss": 0.2556, "step": 7440 }, { "epoch": 1.55, "learning_rate": 6.830357142857143e-08, "loss": 0.1813, "step": 7441 }, { "epoch": 1.55, "learning_rate": 6.827242524916943e-08, "loss": 0.1597, "step": 7442 }, { "epoch": 1.55, "learning_rate": 6.824127906976743e-08, "loss": 0.2058, "step": 7443 }, { "epoch": 1.55, "learning_rate": 6.821013289036545e-08, "loss": 0.1675, "step": 7444 }, { "epoch": 1.55, "learning_rate": 6.817898671096345e-08, "loss": 0.2093, "step": 7445 }, { "epoch": 1.55, "learning_rate": 6.814784053156145e-08, "loss": 0.1292, "step": 7446 }, { "epoch": 1.55, "learning_rate": 6.811669435215947e-08, "loss": 0.1513, "step": 7447 }, { "epoch": 1.55, "learning_rate": 6.808554817275747e-08, "loss": 0.223, "step": 7448 }, { "epoch": 1.55, "learning_rate": 6.805440199335547e-08, "loss": 0.1786, "step": 7449 }, { "epoch": 1.55, "learning_rate": 6.802325581395348e-08, "loss": 0.2052, "step": 7450 }, { "epoch": 1.55, "learning_rate": 6.799210963455149e-08, "loss": 0.1677, "step": 7451 }, { "epoch": 1.55, "learning_rate": 6.79609634551495e-08, "loss": 0.1616, "step": 7452 }, { "epoch": 1.55, "learning_rate": 6.792981727574751e-08, "loss": 0.1935, "step": 7453 }, { "epoch": 1.55, "learning_rate": 6.789867109634551e-08, "loss": 0.1628, "step": 7454 }, { "epoch": 1.55, "learning_rate": 6.786752491694352e-08, "loss": 0.2196, "step": 7455 }, { "epoch": 1.55, "learning_rate": 6.783637873754152e-08, "loss": 0.2379, "step": 7456 }, { "epoch": 1.55, "learning_rate": 6.780523255813953e-08, "loss": 0.157, "step": 7457 }, { "epoch": 1.55, "learning_rate": 6.777408637873754e-08, "loss": 0.231, "step": 7458 }, { "epoch": 1.55, "learning_rate": 6.774294019933555e-08, "loss": 0.1811, "step": 7459 }, { "epoch": 1.55, "learning_rate": 6.771179401993356e-08, "loss": 0.2543, "step": 7460 }, { "epoch": 1.55, "learning_rate": 6.768064784053156e-08, "loss": 0.1653, "step": 7461 }, { "epoch": 1.55, "learning_rate": 6.764950166112956e-08, "loss": 0.1904, "step": 7462 }, { "epoch": 1.55, "learning_rate": 6.761835548172756e-08, "loss": 0.1836, "step": 7463 }, { "epoch": 1.55, "learning_rate": 6.758720930232558e-08, "loss": 0.1561, "step": 7464 }, { "epoch": 1.55, "learning_rate": 6.755606312292358e-08, "loss": 0.2312, "step": 7465 }, { "epoch": 1.55, "learning_rate": 6.75249169435216e-08, "loss": 0.1482, "step": 7466 }, { "epoch": 1.55, "learning_rate": 6.74937707641196e-08, "loss": 0.1455, "step": 7467 }, { "epoch": 1.55, "learning_rate": 6.74626245847176e-08, "loss": 0.1324, "step": 7468 }, { "epoch": 1.55, "learning_rate": 6.74314784053156e-08, "loss": 0.1577, "step": 7469 }, { "epoch": 1.55, "learning_rate": 6.740033222591361e-08, "loss": 0.1543, "step": 7470 }, { "epoch": 1.55, "learning_rate": 6.736918604651162e-08, "loss": 0.2268, "step": 7471 }, { "epoch": 1.55, "learning_rate": 6.733803986710964e-08, "loss": 0.1635, "step": 7472 }, { "epoch": 1.55, "learning_rate": 6.730689368770764e-08, "loss": 0.2103, "step": 7473 }, { "epoch": 1.55, "learning_rate": 6.727574750830564e-08, "loss": 0.1302, "step": 7474 }, { "epoch": 1.55, "learning_rate": 6.724460132890365e-08, "loss": 0.1516, "step": 7475 }, { "epoch": 1.55, "learning_rate": 6.721345514950165e-08, "loss": 0.169, "step": 7476 }, { "epoch": 1.55, "learning_rate": 6.718230897009967e-08, "loss": 0.2041, "step": 7477 }, { "epoch": 1.55, "learning_rate": 6.715116279069767e-08, "loss": 0.147, "step": 7478 }, { "epoch": 1.55, "learning_rate": 6.712001661129568e-08, "loss": 0.2279, "step": 7479 }, { "epoch": 1.55, "learning_rate": 6.708887043189369e-08, "loss": 0.1615, "step": 7480 }, { "epoch": 1.55, "learning_rate": 6.705772425249169e-08, "loss": 0.2333, "step": 7481 }, { "epoch": 1.55, "learning_rate": 6.702657807308969e-08, "loss": 0.2134, "step": 7482 }, { "epoch": 1.55, "learning_rate": 6.69954318936877e-08, "loss": 0.2134, "step": 7483 }, { "epoch": 1.55, "learning_rate": 6.696428571428571e-08, "loss": 0.1589, "step": 7484 }, { "epoch": 1.55, "learning_rate": 6.693313953488373e-08, "loss": 0.1952, "step": 7485 }, { "epoch": 1.55, "learning_rate": 6.690199335548173e-08, "loss": 0.2092, "step": 7486 }, { "epoch": 1.55, "learning_rate": 6.687084717607973e-08, "loss": 0.1453, "step": 7487 }, { "epoch": 1.55, "learning_rate": 6.683970099667773e-08, "loss": 0.2227, "step": 7488 }, { "epoch": 1.56, "learning_rate": 6.680855481727574e-08, "loss": 0.1898, "step": 7489 }, { "epoch": 1.56, "learning_rate": 6.677740863787375e-08, "loss": 0.1691, "step": 7490 }, { "epoch": 1.56, "learning_rate": 6.674626245847175e-08, "loss": 0.1547, "step": 7491 }, { "epoch": 1.56, "learning_rate": 6.671511627906977e-08, "loss": 0.122, "step": 7492 }, { "epoch": 1.56, "learning_rate": 6.668397009966777e-08, "loss": 0.2117, "step": 7493 }, { "epoch": 1.56, "learning_rate": 6.665282392026578e-08, "loss": 0.2053, "step": 7494 }, { "epoch": 1.56, "learning_rate": 6.662167774086378e-08, "loss": 0.1731, "step": 7495 }, { "epoch": 1.56, "learning_rate": 6.65905315614618e-08, "loss": 0.1249, "step": 7496 }, { "epoch": 1.56, "learning_rate": 6.65593853820598e-08, "loss": 0.2008, "step": 7497 }, { "epoch": 1.56, "learning_rate": 6.652823920265781e-08, "loss": 0.2389, "step": 7498 }, { "epoch": 1.56, "learning_rate": 6.649709302325582e-08, "loss": 0.1524, "step": 7499 }, { "epoch": 1.56, "learning_rate": 6.646594684385382e-08, "loss": 0.2002, "step": 7500 }, { "epoch": 1.56, "learning_rate": 6.643480066445182e-08, "loss": 0.1312, "step": 7501 }, { "epoch": 1.56, "learning_rate": 6.640365448504982e-08, "loss": 0.2098, "step": 7502 }, { "epoch": 1.56, "learning_rate": 6.637250830564784e-08, "loss": 0.134, "step": 7503 }, { "epoch": 1.56, "learning_rate": 6.634136212624584e-08, "loss": 0.2012, "step": 7504 }, { "epoch": 1.56, "learning_rate": 6.631021594684386e-08, "loss": 0.2013, "step": 7505 }, { "epoch": 1.56, "learning_rate": 6.627906976744186e-08, "loss": 0.1617, "step": 7506 }, { "epoch": 1.56, "learning_rate": 6.624792358803986e-08, "loss": 0.1451, "step": 7507 }, { "epoch": 1.56, "learning_rate": 6.621677740863787e-08, "loss": 0.1589, "step": 7508 }, { "epoch": 1.56, "learning_rate": 6.618563122923588e-08, "loss": 0.1698, "step": 7509 }, { "epoch": 1.56, "learning_rate": 6.615448504983388e-08, "loss": 0.1241, "step": 7510 }, { "epoch": 1.56, "learning_rate": 6.61233388704319e-08, "loss": 0.1519, "step": 7511 }, { "epoch": 1.56, "learning_rate": 6.60921926910299e-08, "loss": 0.2018, "step": 7512 }, { "epoch": 1.56, "learning_rate": 6.60610465116279e-08, "loss": 0.2752, "step": 7513 }, { "epoch": 1.56, "learning_rate": 6.602990033222591e-08, "loss": 0.1362, "step": 7514 }, { "epoch": 1.56, "learning_rate": 6.599875415282391e-08, "loss": 0.1407, "step": 7515 }, { "epoch": 1.56, "learning_rate": 6.596760797342193e-08, "loss": 0.2069, "step": 7516 }, { "epoch": 1.56, "learning_rate": 6.593646179401993e-08, "loss": 0.1659, "step": 7517 }, { "epoch": 1.56, "learning_rate": 6.590531561461794e-08, "loss": 0.2, "step": 7518 }, { "epoch": 1.56, "learning_rate": 6.587416943521595e-08, "loss": 0.1926, "step": 7519 }, { "epoch": 1.56, "learning_rate": 6.584302325581395e-08, "loss": 0.1648, "step": 7520 }, { "epoch": 1.56, "learning_rate": 6.581187707641195e-08, "loss": 0.1505, "step": 7521 }, { "epoch": 1.56, "learning_rate": 6.578073089700997e-08, "loss": 0.2133, "step": 7522 }, { "epoch": 1.56, "learning_rate": 6.574958471760797e-08, "loss": 0.1368, "step": 7523 }, { "epoch": 1.56, "learning_rate": 6.571843853820599e-08, "loss": 0.2261, "step": 7524 }, { "epoch": 1.56, "learning_rate": 6.568729235880399e-08, "loss": 0.1871, "step": 7525 }, { "epoch": 1.56, "learning_rate": 6.565614617940199e-08, "loss": 0.171, "step": 7526 }, { "epoch": 1.56, "learning_rate": 6.5625e-08, "loss": 0.1767, "step": 7527 }, { "epoch": 1.56, "learning_rate": 6.5593853820598e-08, "loss": 0.2045, "step": 7528 }, { "epoch": 1.56, "learning_rate": 6.556270764119601e-08, "loss": 0.1373, "step": 7529 }, { "epoch": 1.56, "learning_rate": 6.553156146179401e-08, "loss": 0.1865, "step": 7530 }, { "epoch": 1.56, "learning_rate": 6.550041528239203e-08, "loss": 0.2429, "step": 7531 }, { "epoch": 1.56, "learning_rate": 6.546926910299003e-08, "loss": 0.17, "step": 7532 }, { "epoch": 1.56, "learning_rate": 6.543812292358804e-08, "loss": 0.1841, "step": 7533 }, { "epoch": 1.56, "learning_rate": 6.540697674418604e-08, "loss": 0.1716, "step": 7534 }, { "epoch": 1.56, "learning_rate": 6.537583056478405e-08, "loss": 0.1676, "step": 7535 }, { "epoch": 1.56, "learning_rate": 6.534468438538206e-08, "loss": 0.1743, "step": 7536 }, { "epoch": 1.56, "learning_rate": 6.531353820598006e-08, "loss": 0.1613, "step": 7537 }, { "epoch": 1.57, "learning_rate": 6.528239202657808e-08, "loss": 0.2014, "step": 7538 }, { "epoch": 1.57, "learning_rate": 6.525124584717608e-08, "loss": 0.1719, "step": 7539 }, { "epoch": 1.57, "learning_rate": 6.522009966777408e-08, "loss": 0.1784, "step": 7540 }, { "epoch": 1.57, "learning_rate": 6.518895348837208e-08, "loss": 0.1628, "step": 7541 }, { "epoch": 1.57, "learning_rate": 6.51578073089701e-08, "loss": 0.2215, "step": 7542 }, { "epoch": 1.57, "learning_rate": 6.51266611295681e-08, "loss": 0.1685, "step": 7543 }, { "epoch": 1.57, "learning_rate": 6.509551495016612e-08, "loss": 0.1617, "step": 7544 }, { "epoch": 1.57, "learning_rate": 6.506436877076412e-08, "loss": 0.2178, "step": 7545 }, { "epoch": 1.57, "learning_rate": 6.503322259136212e-08, "loss": 0.1689, "step": 7546 }, { "epoch": 1.57, "learning_rate": 6.500207641196012e-08, "loss": 0.2191, "step": 7547 }, { "epoch": 1.57, "learning_rate": 6.497093023255814e-08, "loss": 0.2352, "step": 7548 }, { "epoch": 1.57, "learning_rate": 6.493978405315614e-08, "loss": 0.1379, "step": 7549 }, { "epoch": 1.57, "learning_rate": 6.490863787375415e-08, "loss": 0.1553, "step": 7550 }, { "epoch": 1.57, "learning_rate": 6.487749169435216e-08, "loss": 0.175, "step": 7551 }, { "epoch": 1.57, "learning_rate": 6.484634551495016e-08, "loss": 0.1333, "step": 7552 }, { "epoch": 1.57, "learning_rate": 6.481519933554817e-08, "loss": 0.1608, "step": 7553 }, { "epoch": 1.57, "learning_rate": 6.478405315614618e-08, "loss": 0.2432, "step": 7554 }, { "epoch": 1.57, "learning_rate": 6.475290697674419e-08, "loss": 0.1463, "step": 7555 }, { "epoch": 1.57, "learning_rate": 6.472176079734219e-08, "loss": 0.1832, "step": 7556 }, { "epoch": 1.57, "learning_rate": 6.469061461794019e-08, "loss": 0.179, "step": 7557 }, { "epoch": 1.57, "learning_rate": 6.46594684385382e-08, "loss": 0.1691, "step": 7558 }, { "epoch": 1.57, "learning_rate": 6.462832225913621e-08, "loss": 0.1906, "step": 7559 }, { "epoch": 1.57, "learning_rate": 6.459717607973421e-08, "loss": 0.1676, "step": 7560 }, { "epoch": 1.57, "learning_rate": 6.456602990033223e-08, "loss": 0.1477, "step": 7561 }, { "epoch": 1.57, "learning_rate": 6.453488372093023e-08, "loss": 0.1511, "step": 7562 }, { "epoch": 1.57, "learning_rate": 6.450373754152823e-08, "loss": 0.1869, "step": 7563 }, { "epoch": 1.57, "learning_rate": 6.447259136212625e-08, "loss": 0.1419, "step": 7564 }, { "epoch": 1.57, "learning_rate": 6.444144518272425e-08, "loss": 0.1438, "step": 7565 }, { "epoch": 1.57, "learning_rate": 6.441029900332225e-08, "loss": 0.1882, "step": 7566 }, { "epoch": 1.57, "learning_rate": 6.437915282392027e-08, "loss": 0.198, "step": 7567 }, { "epoch": 1.57, "learning_rate": 6.434800664451827e-08, "loss": 0.157, "step": 7568 }, { "epoch": 1.57, "learning_rate": 6.431686046511627e-08, "loss": 0.1317, "step": 7569 }, { "epoch": 1.57, "learning_rate": 6.428571428571428e-08, "loss": 0.1825, "step": 7570 }, { "epoch": 1.57, "learning_rate": 6.425456810631229e-08, "loss": 0.1716, "step": 7571 }, { "epoch": 1.57, "learning_rate": 6.42234219269103e-08, "loss": 0.2071, "step": 7572 }, { "epoch": 1.57, "learning_rate": 6.41922757475083e-08, "loss": 0.1472, "step": 7573 }, { "epoch": 1.57, "learning_rate": 6.416112956810631e-08, "loss": 0.1817, "step": 7574 }, { "epoch": 1.57, "learning_rate": 6.412998338870432e-08, "loss": 0.1526, "step": 7575 }, { "epoch": 1.57, "learning_rate": 6.409883720930232e-08, "loss": 0.1326, "step": 7576 }, { "epoch": 1.57, "learning_rate": 6.406769102990032e-08, "loss": 0.159, "step": 7577 }, { "epoch": 1.57, "learning_rate": 6.403654485049834e-08, "loss": 0.1422, "step": 7578 }, { "epoch": 1.57, "learning_rate": 6.400539867109634e-08, "loss": 0.1589, "step": 7579 }, { "epoch": 1.57, "learning_rate": 6.397425249169436e-08, "loss": 0.1631, "step": 7580 }, { "epoch": 1.57, "learning_rate": 6.394310631229236e-08, "loss": 0.1845, "step": 7581 }, { "epoch": 1.57, "learning_rate": 6.391196013289036e-08, "loss": 0.1689, "step": 7582 }, { "epoch": 1.57, "learning_rate": 6.388081395348836e-08, "loss": 0.1611, "step": 7583 }, { "epoch": 1.57, "learning_rate": 6.384966777408638e-08, "loss": 0.2526, "step": 7584 }, { "epoch": 1.57, "learning_rate": 6.381852159468438e-08, "loss": 0.1937, "step": 7585 }, { "epoch": 1.58, "learning_rate": 6.378737541528238e-08, "loss": 0.1687, "step": 7586 }, { "epoch": 1.58, "learning_rate": 6.37562292358804e-08, "loss": 0.1696, "step": 7587 }, { "epoch": 1.58, "learning_rate": 6.37250830564784e-08, "loss": 0.1865, "step": 7588 }, { "epoch": 1.58, "learning_rate": 6.36939368770764e-08, "loss": 0.1739, "step": 7589 }, { "epoch": 1.58, "learning_rate": 6.366279069767441e-08, "loss": 0.1751, "step": 7590 }, { "epoch": 1.58, "learning_rate": 6.363164451827242e-08, "loss": 0.1717, "step": 7591 }, { "epoch": 1.58, "learning_rate": 6.360049833887043e-08, "loss": 0.1985, "step": 7592 }, { "epoch": 1.58, "learning_rate": 6.356935215946844e-08, "loss": 0.1262, "step": 7593 }, { "epoch": 1.58, "learning_rate": 6.353820598006645e-08, "loss": 0.1865, "step": 7594 }, { "epoch": 1.58, "learning_rate": 6.350705980066445e-08, "loss": 0.1811, "step": 7595 }, { "epoch": 1.58, "learning_rate": 6.347591362126245e-08, "loss": 0.1449, "step": 7596 }, { "epoch": 1.58, "learning_rate": 6.344476744186045e-08, "loss": 0.2136, "step": 7597 }, { "epoch": 1.58, "learning_rate": 6.341362126245847e-08, "loss": 0.1321, "step": 7598 }, { "epoch": 1.58, "learning_rate": 6.338247508305647e-08, "loss": 0.1794, "step": 7599 }, { "epoch": 1.58, "learning_rate": 6.335132890365449e-08, "loss": 0.19, "step": 7600 }, { "epoch": 1.58, "learning_rate": 6.332018272425249e-08, "loss": 0.2327, "step": 7601 }, { "epoch": 1.58, "learning_rate": 6.328903654485049e-08, "loss": 0.1815, "step": 7602 }, { "epoch": 1.58, "learning_rate": 6.32578903654485e-08, "loss": 0.208, "step": 7603 }, { "epoch": 1.58, "learning_rate": 6.322674418604651e-08, "loss": 0.1496, "step": 7604 }, { "epoch": 1.58, "learning_rate": 6.319559800664451e-08, "loss": 0.1444, "step": 7605 }, { "epoch": 1.58, "learning_rate": 6.316445182724253e-08, "loss": 0.1811, "step": 7606 }, { "epoch": 1.58, "learning_rate": 6.313330564784053e-08, "loss": 0.1627, "step": 7607 }, { "epoch": 1.58, "learning_rate": 6.310215946843853e-08, "loss": 0.151, "step": 7608 }, { "epoch": 1.58, "learning_rate": 6.307101328903654e-08, "loss": 0.1421, "step": 7609 }, { "epoch": 1.58, "learning_rate": 6.303986710963454e-08, "loss": 0.195, "step": 7610 }, { "epoch": 1.58, "learning_rate": 6.300872093023256e-08, "loss": 0.2108, "step": 7611 }, { "epoch": 1.58, "learning_rate": 6.297757475083056e-08, "loss": 0.1794, "step": 7612 }, { "epoch": 1.58, "learning_rate": 6.294642857142857e-08, "loss": 0.1777, "step": 7613 }, { "epoch": 1.58, "learning_rate": 6.291528239202658e-08, "loss": 0.1875, "step": 7614 }, { "epoch": 1.58, "learning_rate": 6.288413621262458e-08, "loss": 0.2305, "step": 7615 }, { "epoch": 1.58, "learning_rate": 6.285299003322258e-08, "loss": 0.1767, "step": 7616 }, { "epoch": 1.58, "learning_rate": 6.28218438538206e-08, "loss": 0.2235, "step": 7617 }, { "epoch": 1.58, "learning_rate": 6.27906976744186e-08, "loss": 0.2209, "step": 7618 }, { "epoch": 1.58, "learning_rate": 6.275955149501662e-08, "loss": 0.1558, "step": 7619 }, { "epoch": 1.58, "learning_rate": 6.272840531561462e-08, "loss": 0.1785, "step": 7620 }, { "epoch": 1.58, "learning_rate": 6.269725913621262e-08, "loss": 0.157, "step": 7621 }, { "epoch": 1.58, "learning_rate": 6.266611295681062e-08, "loss": 0.1462, "step": 7622 }, { "epoch": 1.58, "learning_rate": 6.263496677740863e-08, "loss": 0.1628, "step": 7623 }, { "epoch": 1.58, "learning_rate": 6.260382059800664e-08, "loss": 0.1948, "step": 7624 }, { "epoch": 1.58, "learning_rate": 6.257267441860466e-08, "loss": 0.1796, "step": 7625 }, { "epoch": 1.58, "learning_rate": 6.254152823920266e-08, "loss": 0.1762, "step": 7626 }, { "epoch": 1.58, "learning_rate": 6.251038205980066e-08, "loss": 0.1636, "step": 7627 }, { "epoch": 1.58, "learning_rate": 6.247923588039867e-08, "loss": 0.1817, "step": 7628 }, { "epoch": 1.58, "learning_rate": 6.244808970099667e-08, "loss": 0.1755, "step": 7629 }, { "epoch": 1.58, "learning_rate": 6.241694352159468e-08, "loss": 0.1627, "step": 7630 }, { "epoch": 1.58, "learning_rate": 6.238579734219269e-08, "loss": 0.1583, "step": 7631 }, { "epoch": 1.58, "learning_rate": 6.23546511627907e-08, "loss": 0.2055, "step": 7632 }, { "epoch": 1.58, "learning_rate": 6.23235049833887e-08, "loss": 0.2316, "step": 7633 }, { "epoch": 1.59, "learning_rate": 6.229235880398671e-08, "loss": 0.2041, "step": 7634 }, { "epoch": 1.59, "learning_rate": 6.226121262458471e-08, "loss": 0.144, "step": 7635 }, { "epoch": 1.59, "learning_rate": 6.223006644518271e-08, "loss": 0.1642, "step": 7636 }, { "epoch": 1.59, "learning_rate": 6.219892026578073e-08, "loss": 0.1676, "step": 7637 }, { "epoch": 1.59, "learning_rate": 6.216777408637874e-08, "loss": 0.1598, "step": 7638 }, { "epoch": 1.59, "learning_rate": 6.213662790697675e-08, "loss": 0.1498, "step": 7639 }, { "epoch": 1.59, "learning_rate": 6.210548172757475e-08, "loss": 0.1395, "step": 7640 }, { "epoch": 1.59, "learning_rate": 6.207433554817275e-08, "loss": 0.1756, "step": 7641 }, { "epoch": 1.59, "learning_rate": 6.204318936877075e-08, "loss": 0.145, "step": 7642 }, { "epoch": 1.59, "learning_rate": 6.201204318936877e-08, "loss": 0.1897, "step": 7643 }, { "epoch": 1.59, "learning_rate": 6.198089700996677e-08, "loss": 0.1724, "step": 7644 }, { "epoch": 1.59, "learning_rate": 6.194975083056479e-08, "loss": 0.2174, "step": 7645 }, { "epoch": 1.59, "learning_rate": 6.191860465116279e-08, "loss": 0.2019, "step": 7646 }, { "epoch": 1.59, "learning_rate": 6.18874584717608e-08, "loss": 0.1488, "step": 7647 }, { "epoch": 1.59, "learning_rate": 6.18563122923588e-08, "loss": 0.1711, "step": 7648 }, { "epoch": 1.59, "learning_rate": 6.18251661129568e-08, "loss": 0.1924, "step": 7649 }, { "epoch": 1.59, "learning_rate": 6.179401993355482e-08, "loss": 0.1861, "step": 7650 }, { "epoch": 1.59, "learning_rate": 6.176287375415283e-08, "loss": 0.2019, "step": 7651 }, { "epoch": 1.59, "learning_rate": 6.173172757475083e-08, "loss": 0.1488, "step": 7652 }, { "epoch": 1.59, "learning_rate": 6.170058139534884e-08, "loss": 0.1545, "step": 7653 }, { "epoch": 1.59, "learning_rate": 6.166943521594684e-08, "loss": 0.1479, "step": 7654 }, { "epoch": 1.59, "learning_rate": 6.163828903654484e-08, "loss": 0.1757, "step": 7655 }, { "epoch": 1.59, "learning_rate": 6.160714285714286e-08, "loss": 0.1689, "step": 7656 }, { "epoch": 1.59, "learning_rate": 6.157599667774086e-08, "loss": 0.1489, "step": 7657 }, { "epoch": 1.59, "learning_rate": 6.154485049833888e-08, "loss": 0.1999, "step": 7658 }, { "epoch": 1.59, "learning_rate": 6.151370431893688e-08, "loss": 0.1941, "step": 7659 }, { "epoch": 1.59, "learning_rate": 6.148255813953488e-08, "loss": 0.1522, "step": 7660 }, { "epoch": 1.59, "learning_rate": 6.145141196013288e-08, "loss": 0.199, "step": 7661 }, { "epoch": 1.59, "learning_rate": 6.14202657807309e-08, "loss": 0.1936, "step": 7662 }, { "epoch": 1.59, "learning_rate": 6.13891196013289e-08, "loss": 0.1593, "step": 7663 }, { "epoch": 1.59, "learning_rate": 6.13579734219269e-08, "loss": 0.1792, "step": 7664 }, { "epoch": 1.59, "learning_rate": 6.132682724252492e-08, "loss": 0.1149, "step": 7665 }, { "epoch": 1.59, "learning_rate": 6.129568106312292e-08, "loss": 0.2224, "step": 7666 }, { "epoch": 1.59, "learning_rate": 6.126453488372093e-08, "loss": 0.158, "step": 7667 }, { "epoch": 1.59, "learning_rate": 6.123338870431893e-08, "loss": 0.2113, "step": 7668 }, { "epoch": 1.59, "learning_rate": 6.120224252491694e-08, "loss": 0.2046, "step": 7669 }, { "epoch": 1.59, "learning_rate": 6.117109634551495e-08, "loss": 0.1373, "step": 7670 }, { "epoch": 1.59, "learning_rate": 6.113995016611296e-08, "loss": 0.1701, "step": 7671 }, { "epoch": 1.59, "learning_rate": 6.110880398671096e-08, "loss": 0.1521, "step": 7672 }, { "epoch": 1.59, "learning_rate": 6.107765780730897e-08, "loss": 0.1699, "step": 7673 }, { "epoch": 1.59, "learning_rate": 6.104651162790697e-08, "loss": 0.1688, "step": 7674 }, { "epoch": 1.59, "learning_rate": 6.101536544850499e-08, "loss": 0.1613, "step": 7675 }, { "epoch": 1.59, "learning_rate": 6.098421926910299e-08, "loss": 0.1282, "step": 7676 }, { "epoch": 1.59, "learning_rate": 6.095307308970099e-08, "loss": 0.1912, "step": 7677 }, { "epoch": 1.59, "learning_rate": 6.092192691029901e-08, "loss": 0.1772, "step": 7678 }, { "epoch": 1.59, "learning_rate": 6.089078073089701e-08, "loss": 0.1351, "step": 7679 }, { "epoch": 1.59, "learning_rate": 6.085963455149501e-08, "loss": 0.1635, "step": 7680 }, { "epoch": 1.59, "learning_rate": 6.082848837209301e-08, "loss": 0.1741, "step": 7681 }, { "epoch": 1.6, "learning_rate": 6.079734219269103e-08, "loss": 0.1969, "step": 7682 }, { "epoch": 1.6, "learning_rate": 6.076619601328903e-08, "loss": 0.1714, "step": 7683 }, { "epoch": 1.6, "learning_rate": 6.073504983388704e-08, "loss": 0.2181, "step": 7684 }, { "epoch": 1.6, "learning_rate": 6.070390365448505e-08, "loss": 0.151, "step": 7685 }, { "epoch": 1.6, "learning_rate": 6.067275747508305e-08, "loss": 0.1786, "step": 7686 }, { "epoch": 1.6, "learning_rate": 6.064161129568106e-08, "loss": 0.1896, "step": 7687 }, { "epoch": 1.6, "learning_rate": 6.061046511627907e-08, "loss": 0.2337, "step": 7688 }, { "epoch": 1.6, "learning_rate": 6.057931893687708e-08, "loss": 0.2184, "step": 7689 }, { "epoch": 1.6, "learning_rate": 6.054817275747508e-08, "loss": 0.2221, "step": 7690 }, { "epoch": 1.6, "learning_rate": 6.05170265780731e-08, "loss": 0.1553, "step": 7691 }, { "epoch": 1.6, "learning_rate": 6.04858803986711e-08, "loss": 0.183, "step": 7692 }, { "epoch": 1.6, "learning_rate": 6.04547342192691e-08, "loss": 0.2362, "step": 7693 }, { "epoch": 1.6, "learning_rate": 6.04235880398671e-08, "loss": 0.1691, "step": 7694 }, { "epoch": 1.6, "learning_rate": 6.039244186046512e-08, "loss": 0.1484, "step": 7695 }, { "epoch": 1.6, "learning_rate": 6.036129568106312e-08, "loss": 0.1453, "step": 7696 }, { "epoch": 1.6, "learning_rate": 6.033014950166112e-08, "loss": 0.1804, "step": 7697 }, { "epoch": 1.6, "learning_rate": 6.029900332225914e-08, "loss": 0.1734, "step": 7698 }, { "epoch": 1.6, "learning_rate": 6.026785714285714e-08, "loss": 0.2209, "step": 7699 }, { "epoch": 1.6, "learning_rate": 6.023671096345514e-08, "loss": 0.1701, "step": 7700 }, { "epoch": 1.6, "learning_rate": 6.020556478405316e-08, "loss": 0.1699, "step": 7701 }, { "epoch": 1.6, "learning_rate": 6.017441860465116e-08, "loss": 0.1368, "step": 7702 }, { "epoch": 1.6, "learning_rate": 6.014327242524916e-08, "loss": 0.1415, "step": 7703 }, { "epoch": 1.6, "learning_rate": 6.011212624584717e-08, "loss": 0.1644, "step": 7704 }, { "epoch": 1.6, "learning_rate": 6.008098006644518e-08, "loss": 0.1689, "step": 7705 }, { "epoch": 1.6, "learning_rate": 6.004983388704319e-08, "loss": 0.1468, "step": 7706 }, { "epoch": 1.6, "learning_rate": 6.001868770764119e-08, "loss": 0.2142, "step": 7707 }, { "epoch": 1.6, "learning_rate": 5.99875415282392e-08, "loss": 0.2098, "step": 7708 }, { "epoch": 1.6, "learning_rate": 5.99563953488372e-08, "loss": 0.1842, "step": 7709 }, { "epoch": 1.6, "learning_rate": 5.992524916943521e-08, "loss": 0.1643, "step": 7710 }, { "epoch": 1.6, "learning_rate": 5.989410299003322e-08, "loss": 0.1342, "step": 7711 }, { "epoch": 1.6, "learning_rate": 5.986295681063123e-08, "loss": 0.1822, "step": 7712 }, { "epoch": 1.6, "learning_rate": 5.983181063122923e-08, "loss": 0.1404, "step": 7713 }, { "epoch": 1.6, "learning_rate": 5.980066445182725e-08, "loss": 0.2279, "step": 7714 }, { "epoch": 1.6, "learning_rate": 5.976951827242525e-08, "loss": 0.1113, "step": 7715 }, { "epoch": 1.6, "learning_rate": 5.973837209302325e-08, "loss": 0.1927, "step": 7716 }, { "epoch": 1.6, "learning_rate": 5.970722591362125e-08, "loss": 0.2095, "step": 7717 }, { "epoch": 1.6, "learning_rate": 5.967607973421927e-08, "loss": 0.1701, "step": 7718 }, { "epoch": 1.6, "learning_rate": 5.964493355481727e-08, "loss": 0.2039, "step": 7719 }, { "epoch": 1.6, "learning_rate": 5.961378737541527e-08, "loss": 0.1446, "step": 7720 }, { "epoch": 1.6, "learning_rate": 5.9582641196013284e-08, "loss": 0.1527, "step": 7721 }, { "epoch": 1.6, "learning_rate": 5.955149501661129e-08, "loss": 0.1675, "step": 7722 }, { "epoch": 1.6, "learning_rate": 5.9520348837209295e-08, "loss": 0.1324, "step": 7723 }, { "epoch": 1.6, "learning_rate": 5.9489202657807305e-08, "loss": 0.2499, "step": 7724 }, { "epoch": 1.6, "learning_rate": 5.9458056478405314e-08, "loss": 0.2466, "step": 7725 }, { "epoch": 1.6, "learning_rate": 5.942691029900332e-08, "loss": 0.1291, "step": 7726 }, { "epoch": 1.6, "learning_rate": 5.9395764119601326e-08, "loss": 0.2712, "step": 7727 }, { "epoch": 1.6, "learning_rate": 5.9364617940199335e-08, "loss": 0.1869, "step": 7728 }, { "epoch": 1.6, "learning_rate": 5.933347176079734e-08, "loss": 0.1616, "step": 7729 }, { "epoch": 1.61, "learning_rate": 5.930232558139534e-08, "loss": 0.2113, "step": 7730 }, { "epoch": 1.61, "learning_rate": 5.927117940199335e-08, "loss": 0.221, "step": 7731 }, { "epoch": 1.61, "learning_rate": 5.9240033222591365e-08, "loss": 0.213, "step": 7732 }, { "epoch": 1.61, "learning_rate": 5.920888704318937e-08, "loss": 0.2117, "step": 7733 }, { "epoch": 1.61, "learning_rate": 5.917774086378737e-08, "loss": 0.1512, "step": 7734 }, { "epoch": 1.61, "learning_rate": 5.914659468438538e-08, "loss": 0.151, "step": 7735 }, { "epoch": 1.61, "learning_rate": 5.911544850498338e-08, "loss": 0.1237, "step": 7736 }, { "epoch": 1.61, "learning_rate": 5.908430232558139e-08, "loss": 0.1699, "step": 7737 }, { "epoch": 1.61, "learning_rate": 5.90531561461794e-08, "loss": 0.1944, "step": 7738 }, { "epoch": 1.61, "learning_rate": 5.902200996677741e-08, "loss": 0.1967, "step": 7739 }, { "epoch": 1.61, "learning_rate": 5.899086378737541e-08, "loss": 0.1843, "step": 7740 }, { "epoch": 1.61, "learning_rate": 5.895971760797342e-08, "loss": 0.2017, "step": 7741 }, { "epoch": 1.61, "learning_rate": 5.8928571428571424e-08, "loss": 0.1588, "step": 7742 }, { "epoch": 1.61, "learning_rate": 5.889742524916943e-08, "loss": 0.1371, "step": 7743 }, { "epoch": 1.61, "learning_rate": 5.8866279069767436e-08, "loss": 0.1927, "step": 7744 }, { "epoch": 1.61, "learning_rate": 5.883513289036545e-08, "loss": 0.1582, "step": 7745 }, { "epoch": 1.61, "learning_rate": 5.8803986710963454e-08, "loss": 0.1759, "step": 7746 }, { "epoch": 1.61, "learning_rate": 5.877284053156146e-08, "loss": 0.1526, "step": 7747 }, { "epoch": 1.61, "learning_rate": 5.8741694352159466e-08, "loss": 0.1605, "step": 7748 }, { "epoch": 1.61, "learning_rate": 5.871054817275747e-08, "loss": 0.1309, "step": 7749 }, { "epoch": 1.61, "learning_rate": 5.867940199335548e-08, "loss": 0.199, "step": 7750 }, { "epoch": 1.61, "learning_rate": 5.864825581395348e-08, "loss": 0.1956, "step": 7751 }, { "epoch": 1.61, "learning_rate": 5.8617109634551496e-08, "loss": 0.2511, "step": 7752 }, { "epoch": 1.61, "learning_rate": 5.85859634551495e-08, "loss": 0.2031, "step": 7753 }, { "epoch": 1.61, "learning_rate": 5.855481727574751e-08, "loss": 0.2669, "step": 7754 }, { "epoch": 1.61, "learning_rate": 5.852367109634551e-08, "loss": 0.1753, "step": 7755 }, { "epoch": 1.61, "learning_rate": 5.849252491694352e-08, "loss": 0.154, "step": 7756 }, { "epoch": 1.61, "learning_rate": 5.846137873754152e-08, "loss": 0.234, "step": 7757 }, { "epoch": 1.61, "learning_rate": 5.843023255813954e-08, "loss": 0.1647, "step": 7758 }, { "epoch": 1.61, "learning_rate": 5.839908637873754e-08, "loss": 0.1494, "step": 7759 }, { "epoch": 1.61, "learning_rate": 5.836794019933555e-08, "loss": 0.191, "step": 7760 }, { "epoch": 1.61, "learning_rate": 5.833679401993355e-08, "loss": 0.1791, "step": 7761 }, { "epoch": 1.61, "learning_rate": 5.8305647840531555e-08, "loss": 0.2201, "step": 7762 }, { "epoch": 1.61, "learning_rate": 5.8274501661129564e-08, "loss": 0.1984, "step": 7763 }, { "epoch": 1.61, "learning_rate": 5.824335548172757e-08, "loss": 0.1619, "step": 7764 }, { "epoch": 1.61, "learning_rate": 5.821220930232558e-08, "loss": 0.1796, "step": 7765 }, { "epoch": 1.61, "learning_rate": 5.8181063122923585e-08, "loss": 0.1476, "step": 7766 }, { "epoch": 1.61, "learning_rate": 5.8149916943521595e-08, "loss": 0.135, "step": 7767 }, { "epoch": 1.61, "learning_rate": 5.81187707641196e-08, "loss": 0.1325, "step": 7768 }, { "epoch": 1.61, "learning_rate": 5.8087624584717606e-08, "loss": 0.1941, "step": 7769 }, { "epoch": 1.61, "learning_rate": 5.805647840531561e-08, "loss": 0.1983, "step": 7770 }, { "epoch": 1.61, "learning_rate": 5.802533222591361e-08, "loss": 0.1217, "step": 7771 }, { "epoch": 1.61, "learning_rate": 5.799418604651163e-08, "loss": 0.2286, "step": 7772 }, { "epoch": 1.61, "learning_rate": 5.7963039867109637e-08, "loss": 0.1572, "step": 7773 }, { "epoch": 1.61, "learning_rate": 5.793189368770764e-08, "loss": 0.1589, "step": 7774 }, { "epoch": 1.61, "learning_rate": 5.790074750830564e-08, "loss": 0.166, "step": 7775 }, { "epoch": 1.61, "learning_rate": 5.786960132890365e-08, "loss": 0.1839, "step": 7776 }, { "epoch": 1.61, "learning_rate": 5.7838455149501654e-08, "loss": 0.2457, "step": 7777 }, { "epoch": 1.62, "learning_rate": 5.780730897009967e-08, "loss": 0.1738, "step": 7778 }, { "epoch": 1.62, "learning_rate": 5.777616279069767e-08, "loss": 0.1848, "step": 7779 }, { "epoch": 1.62, "learning_rate": 5.774501661129568e-08, "loss": 0.1473, "step": 7780 }, { "epoch": 1.62, "learning_rate": 5.7713870431893684e-08, "loss": 0.2035, "step": 7781 }, { "epoch": 1.62, "learning_rate": 5.768272425249169e-08, "loss": 0.1205, "step": 7782 }, { "epoch": 1.62, "learning_rate": 5.7651578073089696e-08, "loss": 0.1375, "step": 7783 }, { "epoch": 1.62, "learning_rate": 5.76204318936877e-08, "loss": 0.1575, "step": 7784 }, { "epoch": 1.62, "learning_rate": 5.7589285714285714e-08, "loss": 0.1843, "step": 7785 }, { "epoch": 1.62, "learning_rate": 5.755813953488372e-08, "loss": 0.1985, "step": 7786 }, { "epoch": 1.62, "learning_rate": 5.7526993355481726e-08, "loss": 0.1793, "step": 7787 }, { "epoch": 1.62, "learning_rate": 5.749584717607973e-08, "loss": 0.2315, "step": 7788 }, { "epoch": 1.62, "learning_rate": 5.746470099667774e-08, "loss": 0.2265, "step": 7789 }, { "epoch": 1.62, "learning_rate": 5.743355481727574e-08, "loss": 0.1979, "step": 7790 }, { "epoch": 1.62, "learning_rate": 5.740240863787375e-08, "loss": 0.179, "step": 7791 }, { "epoch": 1.62, "learning_rate": 5.737126245847176e-08, "loss": 0.2461, "step": 7792 }, { "epoch": 1.62, "learning_rate": 5.734011627906977e-08, "loss": 0.1906, "step": 7793 }, { "epoch": 1.62, "learning_rate": 5.730897009966777e-08, "loss": 0.1543, "step": 7794 }, { "epoch": 1.62, "learning_rate": 5.727782392026578e-08, "loss": 0.171, "step": 7795 }, { "epoch": 1.62, "learning_rate": 5.724667774086378e-08, "loss": 0.1891, "step": 7796 }, { "epoch": 1.62, "learning_rate": 5.721553156146179e-08, "loss": 0.1397, "step": 7797 }, { "epoch": 1.62, "learning_rate": 5.71843853820598e-08, "loss": 0.203, "step": 7798 }, { "epoch": 1.62, "learning_rate": 5.715323920265781e-08, "loss": 0.1845, "step": 7799 }, { "epoch": 1.62, "learning_rate": 5.712209302325581e-08, "loss": 0.1458, "step": 7800 }, { "epoch": 1.62, "learning_rate": 5.7090946843853815e-08, "loss": 0.1802, "step": 7801 }, { "epoch": 1.62, "learning_rate": 5.7059800664451824e-08, "loss": 0.1812, "step": 7802 }, { "epoch": 1.62, "learning_rate": 5.702865448504983e-08, "loss": 0.1595, "step": 7803 }, { "epoch": 1.62, "learning_rate": 5.6997508305647836e-08, "loss": 0.1977, "step": 7804 }, { "epoch": 1.62, "learning_rate": 5.6966362126245845e-08, "loss": 0.1948, "step": 7805 }, { "epoch": 1.62, "learning_rate": 5.6935215946843854e-08, "loss": 0.1784, "step": 7806 }, { "epoch": 1.62, "learning_rate": 5.690406976744186e-08, "loss": 0.1634, "step": 7807 }, { "epoch": 1.62, "learning_rate": 5.6872923588039866e-08, "loss": 0.1721, "step": 7808 }, { "epoch": 1.62, "learning_rate": 5.684177740863787e-08, "loss": 0.1389, "step": 7809 }, { "epoch": 1.62, "learning_rate": 5.681063122923588e-08, "loss": 0.1378, "step": 7810 }, { "epoch": 1.62, "learning_rate": 5.677948504983388e-08, "loss": 0.1426, "step": 7811 }, { "epoch": 1.62, "learning_rate": 5.6748338870431896e-08, "loss": 0.1653, "step": 7812 }, { "epoch": 1.62, "learning_rate": 5.67171926910299e-08, "loss": 0.1709, "step": 7813 }, { "epoch": 1.62, "learning_rate": 5.668604651162791e-08, "loss": 0.1616, "step": 7814 }, { "epoch": 1.62, "learning_rate": 5.665490033222591e-08, "loss": 0.1439, "step": 7815 }, { "epoch": 1.62, "learning_rate": 5.6623754152823913e-08, "loss": 0.1919, "step": 7816 }, { "epoch": 1.62, "learning_rate": 5.659260797342192e-08, "loss": 0.1711, "step": 7817 }, { "epoch": 1.62, "learning_rate": 5.656146179401993e-08, "loss": 0.1293, "step": 7818 }, { "epoch": 1.62, "learning_rate": 5.653031561461794e-08, "loss": 0.1412, "step": 7819 }, { "epoch": 1.62, "learning_rate": 5.6499169435215944e-08, "loss": 0.2099, "step": 7820 }, { "epoch": 1.62, "learning_rate": 5.646802325581395e-08, "loss": 0.1474, "step": 7821 }, { "epoch": 1.62, "learning_rate": 5.6436877076411955e-08, "loss": 0.215, "step": 7822 }, { "epoch": 1.62, "learning_rate": 5.6405730897009965e-08, "loss": 0.1284, "step": 7823 }, { "epoch": 1.62, "learning_rate": 5.637458471760797e-08, "loss": 0.1939, "step": 7824 }, { "epoch": 1.62, "learning_rate": 5.634343853820598e-08, "loss": 0.1815, "step": 7825 }, { "epoch": 1.62, "learning_rate": 5.6312292358803986e-08, "loss": 0.1751, "step": 7826 }, { "epoch": 1.63, "learning_rate": 5.6281146179401995e-08, "loss": 0.2081, "step": 7827 }, { "epoch": 1.63, "learning_rate": 5.625e-08, "loss": 0.1899, "step": 7828 }, { "epoch": 1.63, "learning_rate": 5.6218853820598e-08, "loss": 0.1716, "step": 7829 }, { "epoch": 1.63, "learning_rate": 5.618770764119601e-08, "loss": 0.1501, "step": 7830 }, { "epoch": 1.63, "learning_rate": 5.615656146179401e-08, "loss": 0.2204, "step": 7831 }, { "epoch": 1.63, "learning_rate": 5.612541528239203e-08, "loss": 0.1494, "step": 7832 }, { "epoch": 1.63, "learning_rate": 5.609426910299003e-08, "loss": 0.1657, "step": 7833 }, { "epoch": 1.63, "learning_rate": 5.606312292358804e-08, "loss": 0.1868, "step": 7834 }, { "epoch": 1.63, "learning_rate": 5.603197674418604e-08, "loss": 0.1836, "step": 7835 }, { "epoch": 1.63, "learning_rate": 5.600083056478405e-08, "loss": 0.1459, "step": 7836 }, { "epoch": 1.63, "learning_rate": 5.5969684385382054e-08, "loss": 0.1691, "step": 7837 }, { "epoch": 1.63, "learning_rate": 5.593853820598006e-08, "loss": 0.2117, "step": 7838 }, { "epoch": 1.63, "learning_rate": 5.590739202657807e-08, "loss": 0.1952, "step": 7839 }, { "epoch": 1.63, "learning_rate": 5.587624584717608e-08, "loss": 0.1703, "step": 7840 }, { "epoch": 1.63, "learning_rate": 5.5845099667774084e-08, "loss": 0.2374, "step": 7841 }, { "epoch": 1.63, "learning_rate": 5.5813953488372087e-08, "loss": 0.1257, "step": 7842 }, { "epoch": 1.63, "learning_rate": 5.5782807308970096e-08, "loss": 0.2349, "step": 7843 }, { "epoch": 1.63, "learning_rate": 5.57516611295681e-08, "loss": 0.2022, "step": 7844 }, { "epoch": 1.63, "learning_rate": 5.5720514950166114e-08, "loss": 0.2177, "step": 7845 }, { "epoch": 1.63, "learning_rate": 5.568936877076412e-08, "loss": 0.2028, "step": 7846 }, { "epoch": 1.63, "learning_rate": 5.5658222591362126e-08, "loss": 0.223, "step": 7847 }, { "epoch": 1.63, "learning_rate": 5.562707641196013e-08, "loss": 0.1373, "step": 7848 }, { "epoch": 1.63, "learning_rate": 5.559593023255814e-08, "loss": 0.1686, "step": 7849 }, { "epoch": 1.63, "learning_rate": 5.556478405315614e-08, "loss": 0.2013, "step": 7850 }, { "epoch": 1.63, "learning_rate": 5.553363787375415e-08, "loss": 0.1644, "step": 7851 }, { "epoch": 1.63, "learning_rate": 5.550249169435216e-08, "loss": 0.1477, "step": 7852 }, { "epoch": 1.63, "learning_rate": 5.547134551495017e-08, "loss": 0.1241, "step": 7853 }, { "epoch": 1.63, "learning_rate": 5.544019933554817e-08, "loss": 0.1219, "step": 7854 }, { "epoch": 1.63, "learning_rate": 5.540905315614618e-08, "loss": 0.2182, "step": 7855 }, { "epoch": 1.63, "learning_rate": 5.537790697674418e-08, "loss": 0.2314, "step": 7856 }, { "epoch": 1.63, "learning_rate": 5.5346760797342185e-08, "loss": 0.1593, "step": 7857 }, { "epoch": 1.63, "learning_rate": 5.5315614617940194e-08, "loss": 0.1684, "step": 7858 }, { "epoch": 1.63, "learning_rate": 5.5284468438538203e-08, "loss": 0.1482, "step": 7859 }, { "epoch": 1.63, "learning_rate": 5.525332225913621e-08, "loss": 0.1677, "step": 7860 }, { "epoch": 1.63, "learning_rate": 5.5222176079734215e-08, "loss": 0.1585, "step": 7861 }, { "epoch": 1.63, "learning_rate": 5.5191029900332224e-08, "loss": 0.2318, "step": 7862 }, { "epoch": 1.63, "learning_rate": 5.515988372093023e-08, "loss": 0.2329, "step": 7863 }, { "epoch": 1.63, "learning_rate": 5.5128737541528236e-08, "loss": 0.1786, "step": 7864 }, { "epoch": 1.63, "learning_rate": 5.5097591362126245e-08, "loss": 0.2117, "step": 7865 }, { "epoch": 1.63, "learning_rate": 5.5066445182724255e-08, "loss": 0.1527, "step": 7866 }, { "epoch": 1.63, "learning_rate": 5.503529900332226e-08, "loss": 0.1905, "step": 7867 }, { "epoch": 1.63, "learning_rate": 5.5004152823920266e-08, "loss": 0.1734, "step": 7868 }, { "epoch": 1.63, "learning_rate": 5.497300664451827e-08, "loss": 0.1499, "step": 7869 }, { "epoch": 1.63, "learning_rate": 5.494186046511627e-08, "loss": 0.1381, "step": 7870 }, { "epoch": 1.63, "learning_rate": 5.491071428571428e-08, "loss": 0.1883, "step": 7871 }, { "epoch": 1.63, "learning_rate": 5.4879568106312297e-08, "loss": 0.2458, "step": 7872 }, { "epoch": 1.63, "learning_rate": 5.48484219269103e-08, "loss": 0.1578, "step": 7873 }, { "epoch": 1.63, "learning_rate": 5.48172757475083e-08, "loss": 0.1308, "step": 7874 }, { "epoch": 1.64, "learning_rate": 5.478612956810631e-08, "loss": 0.1261, "step": 7875 }, { "epoch": 1.64, "learning_rate": 5.4754983388704314e-08, "loss": 0.2146, "step": 7876 }, { "epoch": 1.64, "learning_rate": 5.472383720930232e-08, "loss": 0.1928, "step": 7877 }, { "epoch": 1.64, "learning_rate": 5.4692691029900325e-08, "loss": 0.1493, "step": 7878 }, { "epoch": 1.64, "learning_rate": 5.466154485049834e-08, "loss": 0.1679, "step": 7879 }, { "epoch": 1.64, "learning_rate": 5.4630398671096344e-08, "loss": 0.226, "step": 7880 }, { "epoch": 1.64, "learning_rate": 5.459925249169435e-08, "loss": 0.205, "step": 7881 }, { "epoch": 1.64, "learning_rate": 5.4568106312292356e-08, "loss": 0.1323, "step": 7882 }, { "epoch": 1.64, "learning_rate": 5.453696013289036e-08, "loss": 0.1692, "step": 7883 }, { "epoch": 1.64, "learning_rate": 5.450581395348837e-08, "loss": 0.1611, "step": 7884 }, { "epoch": 1.64, "learning_rate": 5.447466777408638e-08, "loss": 0.1744, "step": 7885 }, { "epoch": 1.64, "learning_rate": 5.4443521594684386e-08, "loss": 0.1418, "step": 7886 }, { "epoch": 1.64, "learning_rate": 5.441237541528239e-08, "loss": 0.1309, "step": 7887 }, { "epoch": 1.64, "learning_rate": 5.43812292358804e-08, "loss": 0.1698, "step": 7888 }, { "epoch": 1.64, "learning_rate": 5.43500830564784e-08, "loss": 0.1847, "step": 7889 }, { "epoch": 1.64, "learning_rate": 5.431893687707641e-08, "loss": 0.2363, "step": 7890 }, { "epoch": 1.64, "learning_rate": 5.428779069767441e-08, "loss": 0.1455, "step": 7891 }, { "epoch": 1.64, "learning_rate": 5.425664451827243e-08, "loss": 0.1427, "step": 7892 }, { "epoch": 1.64, "learning_rate": 5.422549833887043e-08, "loss": 0.1917, "step": 7893 }, { "epoch": 1.64, "learning_rate": 5.419435215946844e-08, "loss": 0.2246, "step": 7894 }, { "epoch": 1.64, "learning_rate": 5.416320598006644e-08, "loss": 0.2157, "step": 7895 }, { "epoch": 1.64, "learning_rate": 5.4132059800664445e-08, "loss": 0.2237, "step": 7896 }, { "epoch": 1.64, "learning_rate": 5.4100913621262454e-08, "loss": 0.1822, "step": 7897 }, { "epoch": 1.64, "learning_rate": 5.4069767441860457e-08, "loss": 0.2423, "step": 7898 }, { "epoch": 1.64, "learning_rate": 5.403862126245847e-08, "loss": 0.1757, "step": 7899 }, { "epoch": 1.64, "learning_rate": 5.4007475083056475e-08, "loss": 0.1813, "step": 7900 }, { "epoch": 1.64, "learning_rate": 5.3976328903654484e-08, "loss": 0.2292, "step": 7901 }, { "epoch": 1.64, "learning_rate": 5.394518272425249e-08, "loss": 0.1706, "step": 7902 }, { "epoch": 1.64, "learning_rate": 5.3914036544850496e-08, "loss": 0.1124, "step": 7903 }, { "epoch": 1.64, "learning_rate": 5.38828903654485e-08, "loss": 0.1506, "step": 7904 }, { "epoch": 1.64, "learning_rate": 5.3851744186046515e-08, "loss": 0.175, "step": 7905 }, { "epoch": 1.64, "learning_rate": 5.382059800664452e-08, "loss": 0.1846, "step": 7906 }, { "epoch": 1.64, "learning_rate": 5.3789451827242526e-08, "loss": 0.1896, "step": 7907 }, { "epoch": 1.64, "learning_rate": 5.375830564784053e-08, "loss": 0.1759, "step": 7908 }, { "epoch": 1.64, "learning_rate": 5.372715946843854e-08, "loss": 0.1936, "step": 7909 }, { "epoch": 1.64, "learning_rate": 5.369601328903654e-08, "loss": 0.1605, "step": 7910 }, { "epoch": 1.64, "learning_rate": 5.366486710963454e-08, "loss": 0.1659, "step": 7911 }, { "epoch": 1.64, "learning_rate": 5.363372093023256e-08, "loss": 0.1793, "step": 7912 }, { "epoch": 1.64, "learning_rate": 5.360257475083056e-08, "loss": 0.2096, "step": 7913 }, { "epoch": 1.64, "learning_rate": 5.357142857142857e-08, "loss": 0.205, "step": 7914 }, { "epoch": 1.64, "learning_rate": 5.3540282392026573e-08, "loss": 0.1851, "step": 7915 }, { "epoch": 1.64, "learning_rate": 5.350913621262458e-08, "loss": 0.1789, "step": 7916 }, { "epoch": 1.64, "learning_rate": 5.3477990033222585e-08, "loss": 0.1819, "step": 7917 }, { "epoch": 1.64, "learning_rate": 5.3446843853820594e-08, "loss": 0.1423, "step": 7918 }, { "epoch": 1.64, "learning_rate": 5.3415697674418604e-08, "loss": 0.1872, "step": 7919 }, { "epoch": 1.64, "learning_rate": 5.338455149501661e-08, "loss": 0.1548, "step": 7920 }, { "epoch": 1.64, "learning_rate": 5.3353405315614615e-08, "loss": 0.1417, "step": 7921 }, { "epoch": 1.64, "learning_rate": 5.3322259136212625e-08, "loss": 0.167, "step": 7922 }, { "epoch": 1.65, "learning_rate": 5.329111295681063e-08, "loss": 0.1871, "step": 7923 }, { "epoch": 1.65, "learning_rate": 5.325996677740863e-08, "loss": 0.1779, "step": 7924 }, { "epoch": 1.65, "learning_rate": 5.3228820598006646e-08, "loss": 0.239, "step": 7925 }, { "epoch": 1.65, "learning_rate": 5.3197674418604655e-08, "loss": 0.1272, "step": 7926 }, { "epoch": 1.65, "learning_rate": 5.316652823920266e-08, "loss": 0.1366, "step": 7927 }, { "epoch": 1.65, "learning_rate": 5.313538205980066e-08, "loss": 0.2153, "step": 7928 }, { "epoch": 1.65, "learning_rate": 5.310423588039867e-08, "loss": 0.1803, "step": 7929 }, { "epoch": 1.65, "learning_rate": 5.307308970099667e-08, "loss": 0.1966, "step": 7930 }, { "epoch": 1.65, "learning_rate": 5.304194352159468e-08, "loss": 0.2072, "step": 7931 }, { "epoch": 1.65, "learning_rate": 5.301079734219269e-08, "loss": 0.2402, "step": 7932 }, { "epoch": 1.65, "learning_rate": 5.29796511627907e-08, "loss": 0.171, "step": 7933 }, { "epoch": 1.65, "learning_rate": 5.29485049833887e-08, "loss": 0.1416, "step": 7934 }, { "epoch": 1.65, "learning_rate": 5.291735880398671e-08, "loss": 0.1075, "step": 7935 }, { "epoch": 1.65, "learning_rate": 5.2886212624584714e-08, "loss": 0.1821, "step": 7936 }, { "epoch": 1.65, "learning_rate": 5.2855066445182716e-08, "loss": 0.1443, "step": 7937 }, { "epoch": 1.65, "learning_rate": 5.2823920265780726e-08, "loss": 0.161, "step": 7938 }, { "epoch": 1.65, "learning_rate": 5.279277408637874e-08, "loss": 0.1818, "step": 7939 }, { "epoch": 1.65, "learning_rate": 5.2761627906976744e-08, "loss": 0.1706, "step": 7940 }, { "epoch": 1.65, "learning_rate": 5.273048172757475e-08, "loss": 0.1748, "step": 7941 }, { "epoch": 1.65, "learning_rate": 5.2699335548172756e-08, "loss": 0.205, "step": 7942 }, { "epoch": 1.65, "learning_rate": 5.266818936877076e-08, "loss": 0.202, "step": 7943 }, { "epoch": 1.65, "learning_rate": 5.263704318936877e-08, "loss": 0.1714, "step": 7944 }, { "epoch": 1.65, "learning_rate": 5.260589700996677e-08, "loss": 0.1516, "step": 7945 }, { "epoch": 1.65, "learning_rate": 5.2574750830564786e-08, "loss": 0.1896, "step": 7946 }, { "epoch": 1.65, "learning_rate": 5.254360465116279e-08, "loss": 0.2046, "step": 7947 }, { "epoch": 1.65, "learning_rate": 5.25124584717608e-08, "loss": 0.1653, "step": 7948 }, { "epoch": 1.65, "learning_rate": 5.24813122923588e-08, "loss": 0.1766, "step": 7949 }, { "epoch": 1.65, "learning_rate": 5.24501661129568e-08, "loss": 0.1841, "step": 7950 }, { "epoch": 1.65, "learning_rate": 5.241901993355481e-08, "loss": 0.2366, "step": 7951 }, { "epoch": 1.65, "learning_rate": 5.238787375415283e-08, "loss": 0.1742, "step": 7952 }, { "epoch": 1.65, "learning_rate": 5.235672757475083e-08, "loss": 0.1766, "step": 7953 }, { "epoch": 1.65, "learning_rate": 5.232558139534883e-08, "loss": 0.2302, "step": 7954 }, { "epoch": 1.65, "learning_rate": 5.229443521594684e-08, "loss": 0.1756, "step": 7955 }, { "epoch": 1.65, "learning_rate": 5.2263289036544845e-08, "loss": 0.2436, "step": 7956 }, { "epoch": 1.65, "learning_rate": 5.2232142857142854e-08, "loss": 0.174, "step": 7957 }, { "epoch": 1.65, "learning_rate": 5.220099667774086e-08, "loss": 0.1808, "step": 7958 }, { "epoch": 1.65, "learning_rate": 5.216985049833887e-08, "loss": 0.196, "step": 7959 }, { "epoch": 1.65, "learning_rate": 5.2138704318936875e-08, "loss": 0.1986, "step": 7960 }, { "epoch": 1.65, "learning_rate": 5.2107558139534885e-08, "loss": 0.1569, "step": 7961 }, { "epoch": 1.65, "learning_rate": 5.207641196013289e-08, "loss": 0.2, "step": 7962 }, { "epoch": 1.65, "learning_rate": 5.2045265780730896e-08, "loss": 0.1925, "step": 7963 }, { "epoch": 1.65, "learning_rate": 5.20141196013289e-08, "loss": 0.1306, "step": 7964 }, { "epoch": 1.65, "learning_rate": 5.19829734219269e-08, "loss": 0.1949, "step": 7965 }, { "epoch": 1.65, "learning_rate": 5.195182724252492e-08, "loss": 0.2294, "step": 7966 }, { "epoch": 1.65, "learning_rate": 5.192068106312292e-08, "loss": 0.199, "step": 7967 }, { "epoch": 1.65, "learning_rate": 5.188953488372093e-08, "loss": 0.2339, "step": 7968 }, { "epoch": 1.65, "learning_rate": 5.185838870431893e-08, "loss": 0.2108, "step": 7969 }, { "epoch": 1.65, "learning_rate": 5.182724252491694e-08, "loss": 0.1716, "step": 7970 }, { "epoch": 1.66, "learning_rate": 5.1796096345514944e-08, "loss": 0.1692, "step": 7971 }, { "epoch": 1.66, "learning_rate": 5.176495016611296e-08, "loss": 0.1578, "step": 7972 }, { "epoch": 1.66, "learning_rate": 5.173380398671096e-08, "loss": 0.1899, "step": 7973 }, { "epoch": 1.66, "learning_rate": 5.170265780730897e-08, "loss": 0.2738, "step": 7974 }, { "epoch": 1.66, "learning_rate": 5.1671511627906974e-08, "loss": 0.1745, "step": 7975 }, { "epoch": 1.66, "learning_rate": 5.164036544850498e-08, "loss": 0.1968, "step": 7976 }, { "epoch": 1.66, "learning_rate": 5.1609219269102986e-08, "loss": 0.2065, "step": 7977 }, { "epoch": 1.66, "learning_rate": 5.157807308970099e-08, "loss": 0.1812, "step": 7978 }, { "epoch": 1.66, "learning_rate": 5.1546926910299004e-08, "loss": 0.169, "step": 7979 }, { "epoch": 1.66, "learning_rate": 5.151578073089701e-08, "loss": 0.1689, "step": 7980 }, { "epoch": 1.66, "learning_rate": 5.1484634551495016e-08, "loss": 0.1833, "step": 7981 }, { "epoch": 1.66, "learning_rate": 5.145348837209302e-08, "loss": 0.183, "step": 7982 }, { "epoch": 1.66, "learning_rate": 5.142234219269103e-08, "loss": 0.1908, "step": 7983 }, { "epoch": 1.66, "learning_rate": 5.139119601328903e-08, "loss": 0.1484, "step": 7984 }, { "epoch": 1.66, "learning_rate": 5.136004983388704e-08, "loss": 0.1607, "step": 7985 }, { "epoch": 1.66, "learning_rate": 5.132890365448505e-08, "loss": 0.2056, "step": 7986 }, { "epoch": 1.66, "learning_rate": 5.129775747508306e-08, "loss": 0.1987, "step": 7987 }, { "epoch": 1.66, "learning_rate": 5.126661129568106e-08, "loss": 0.1765, "step": 7988 }, { "epoch": 1.66, "learning_rate": 5.123546511627907e-08, "loss": 0.1663, "step": 7989 }, { "epoch": 1.66, "learning_rate": 5.120431893687707e-08, "loss": 0.1837, "step": 7990 }, { "epoch": 1.66, "learning_rate": 5.1173172757475075e-08, "loss": 0.199, "step": 7991 }, { "epoch": 1.66, "learning_rate": 5.114202657807309e-08, "loss": 0.1577, "step": 7992 }, { "epoch": 1.66, "learning_rate": 5.11108803986711e-08, "loss": 0.1273, "step": 7993 }, { "epoch": 1.66, "learning_rate": 5.10797342192691e-08, "loss": 0.1739, "step": 7994 }, { "epoch": 1.66, "learning_rate": 5.1048588039867105e-08, "loss": 0.2192, "step": 7995 }, { "epoch": 1.66, "learning_rate": 5.1017441860465114e-08, "loss": 0.148, "step": 7996 }, { "epoch": 1.66, "learning_rate": 5.098629568106312e-08, "loss": 0.1533, "step": 7997 }, { "epoch": 1.66, "learning_rate": 5.0955149501661126e-08, "loss": 0.1778, "step": 7998 }, { "epoch": 1.66, "learning_rate": 5.0924003322259135e-08, "loss": 0.1807, "step": 7999 }, { "epoch": 1.66, "learning_rate": 5.0892857142857144e-08, "loss": 0.2095, "step": 8000 }, { "epoch": 1.66, "learning_rate": 5.086171096345515e-08, "loss": 0.1123, "step": 8001 }, { "epoch": 1.66, "learning_rate": 5.0830564784053156e-08, "loss": 0.2008, "step": 8002 }, { "epoch": 1.66, "learning_rate": 5.079941860465116e-08, "loss": 0.2381, "step": 8003 }, { "epoch": 1.66, "learning_rate": 5.076827242524916e-08, "loss": 0.169, "step": 8004 }, { "epoch": 1.66, "learning_rate": 5.073712624584717e-08, "loss": 0.2162, "step": 8005 }, { "epoch": 1.66, "learning_rate": 5.0705980066445186e-08, "loss": 0.1641, "step": 8006 }, { "epoch": 1.66, "learning_rate": 5.067483388704319e-08, "loss": 0.2465, "step": 8007 }, { "epoch": 1.66, "learning_rate": 5.064368770764119e-08, "loss": 0.1731, "step": 8008 }, { "epoch": 1.66, "learning_rate": 5.06125415282392e-08, "loss": 0.1931, "step": 8009 }, { "epoch": 1.66, "learning_rate": 5.0581395348837203e-08, "loss": 0.1856, "step": 8010 }, { "epoch": 1.66, "learning_rate": 5.055024916943521e-08, "loss": 0.1802, "step": 8011 }, { "epoch": 1.66, "learning_rate": 5.051910299003322e-08, "loss": 0.1944, "step": 8012 }, { "epoch": 1.66, "learning_rate": 5.048795681063123e-08, "loss": 0.1153, "step": 8013 }, { "epoch": 1.66, "learning_rate": 5.0456810631229234e-08, "loss": 0.167, "step": 8014 }, { "epoch": 1.66, "learning_rate": 5.042566445182724e-08, "loss": 0.169, "step": 8015 }, { "epoch": 1.66, "learning_rate": 5.0394518272425245e-08, "loss": 0.1667, "step": 8016 }, { "epoch": 1.66, "learning_rate": 5.0363372093023255e-08, "loss": 0.1551, "step": 8017 }, { "epoch": 1.66, "learning_rate": 5.033222591362126e-08, "loss": 0.1868, "step": 8018 }, { "epoch": 1.67, "learning_rate": 5.030107973421927e-08, "loss": 0.1923, "step": 8019 }, { "epoch": 1.67, "learning_rate": 5.0269933554817276e-08, "loss": 0.1878, "step": 8020 }, { "epoch": 1.67, "learning_rate": 5.023878737541528e-08, "loss": 0.1674, "step": 8021 }, { "epoch": 1.67, "learning_rate": 5.020764119601329e-08, "loss": 0.1624, "step": 8022 }, { "epoch": 1.67, "learning_rate": 5.017649501661129e-08, "loss": 0.1737, "step": 8023 }, { "epoch": 1.67, "learning_rate": 5.01453488372093e-08, "loss": 0.1456, "step": 8024 }, { "epoch": 1.67, "learning_rate": 5.01142026578073e-08, "loss": 0.1737, "step": 8025 }, { "epoch": 1.67, "learning_rate": 5.008305647840532e-08, "loss": 0.1576, "step": 8026 }, { "epoch": 1.67, "learning_rate": 5.005191029900332e-08, "loss": 0.1919, "step": 8027 }, { "epoch": 1.67, "learning_rate": 5.002076411960133e-08, "loss": 0.2014, "step": 8028 }, { "epoch": 1.67, "learning_rate": 4.998961794019933e-08, "loss": 0.2259, "step": 8029 }, { "epoch": 1.67, "learning_rate": 4.995847176079734e-08, "loss": 0.1678, "step": 8030 }, { "epoch": 1.67, "learning_rate": 4.9927325581395344e-08, "loss": 0.169, "step": 8031 }, { "epoch": 1.67, "learning_rate": 4.9896179401993346e-08, "loss": 0.2228, "step": 8032 }, { "epoch": 1.67, "learning_rate": 4.986503322259136e-08, "loss": 0.203, "step": 8033 }, { "epoch": 1.67, "learning_rate": 4.983388704318937e-08, "loss": 0.2269, "step": 8034 }, { "epoch": 1.67, "learning_rate": 4.9802740863787374e-08, "loss": 0.1823, "step": 8035 }, { "epoch": 1.67, "learning_rate": 4.9771594684385377e-08, "loss": 0.2189, "step": 8036 }, { "epoch": 1.67, "learning_rate": 4.9740448504983386e-08, "loss": 0.1652, "step": 8037 }, { "epoch": 1.67, "learning_rate": 4.970930232558139e-08, "loss": 0.1199, "step": 8038 }, { "epoch": 1.67, "learning_rate": 4.9678156146179404e-08, "loss": 0.1514, "step": 8039 }, { "epoch": 1.67, "learning_rate": 4.964700996677741e-08, "loss": 0.1529, "step": 8040 }, { "epoch": 1.67, "learning_rate": 4.9615863787375416e-08, "loss": 0.1437, "step": 8041 }, { "epoch": 1.67, "learning_rate": 4.958471760797342e-08, "loss": 0.1766, "step": 8042 }, { "epoch": 1.67, "learning_rate": 4.955357142857143e-08, "loss": 0.176, "step": 8043 }, { "epoch": 1.67, "learning_rate": 4.952242524916943e-08, "loss": 0.1669, "step": 8044 }, { "epoch": 1.67, "learning_rate": 4.949127906976743e-08, "loss": 0.1578, "step": 8045 }, { "epoch": 1.67, "learning_rate": 4.946013289036545e-08, "loss": 0.2194, "step": 8046 }, { "epoch": 1.67, "learning_rate": 4.942898671096346e-08, "loss": 0.1605, "step": 8047 }, { "epoch": 1.67, "learning_rate": 4.939784053156146e-08, "loss": 0.1757, "step": 8048 }, { "epoch": 1.67, "learning_rate": 4.936669435215946e-08, "loss": 0.1794, "step": 8049 }, { "epoch": 1.67, "learning_rate": 4.933554817275747e-08, "loss": 0.1548, "step": 8050 }, { "epoch": 1.67, "learning_rate": 4.9304401993355475e-08, "loss": 0.1636, "step": 8051 }, { "epoch": 1.67, "learning_rate": 4.9273255813953484e-08, "loss": 0.2643, "step": 8052 }, { "epoch": 1.67, "learning_rate": 4.9242109634551493e-08, "loss": 0.144, "step": 8053 }, { "epoch": 1.67, "learning_rate": 4.92109634551495e-08, "loss": 0.1867, "step": 8054 }, { "epoch": 1.67, "learning_rate": 4.9179817275747505e-08, "loss": 0.1425, "step": 8055 }, { "epoch": 1.67, "learning_rate": 4.9148671096345514e-08, "loss": 0.2193, "step": 8056 }, { "epoch": 1.67, "learning_rate": 4.911752491694352e-08, "loss": 0.164, "step": 8057 }, { "epoch": 1.67, "learning_rate": 4.908637873754152e-08, "loss": 0.1771, "step": 8058 }, { "epoch": 1.67, "learning_rate": 4.9055232558139535e-08, "loss": 0.1686, "step": 8059 }, { "epoch": 1.67, "learning_rate": 4.9024086378737545e-08, "loss": 0.2088, "step": 8060 }, { "epoch": 1.67, "learning_rate": 4.899294019933555e-08, "loss": 0.1404, "step": 8061 }, { "epoch": 1.67, "learning_rate": 4.896179401993355e-08, "loss": 0.2096, "step": 8062 }, { "epoch": 1.67, "learning_rate": 4.893064784053156e-08, "loss": 0.148, "step": 8063 }, { "epoch": 1.67, "learning_rate": 4.889950166112956e-08, "loss": 0.2045, "step": 8064 }, { "epoch": 1.67, "learning_rate": 4.886835548172757e-08, "loss": 0.1535, "step": 8065 }, { "epoch": 1.67, "learning_rate": 4.883720930232558e-08, "loss": 0.2096, "step": 8066 }, { "epoch": 1.68, "learning_rate": 4.880606312292359e-08, "loss": 0.1299, "step": 8067 }, { "epoch": 1.68, "learning_rate": 4.877491694352159e-08, "loss": 0.2488, "step": 8068 }, { "epoch": 1.68, "learning_rate": 4.87437707641196e-08, "loss": 0.2027, "step": 8069 }, { "epoch": 1.68, "learning_rate": 4.8712624584717604e-08, "loss": 0.1366, "step": 8070 }, { "epoch": 1.68, "learning_rate": 4.868147840531561e-08, "loss": 0.1491, "step": 8071 }, { "epoch": 1.68, "learning_rate": 4.8650332225913615e-08, "loss": 0.1959, "step": 8072 }, { "epoch": 1.68, "learning_rate": 4.861918604651163e-08, "loss": 0.1695, "step": 8073 }, { "epoch": 1.68, "learning_rate": 4.8588039867109634e-08, "loss": 0.173, "step": 8074 }, { "epoch": 1.68, "learning_rate": 4.8556893687707636e-08, "loss": 0.1814, "step": 8075 }, { "epoch": 1.68, "learning_rate": 4.8525747508305646e-08, "loss": 0.1751, "step": 8076 }, { "epoch": 1.68, "learning_rate": 4.849460132890365e-08, "loss": 0.1578, "step": 8077 }, { "epoch": 1.68, "learning_rate": 4.846345514950166e-08, "loss": 0.1813, "step": 8078 }, { "epoch": 1.68, "learning_rate": 4.8432308970099667e-08, "loss": 0.1839, "step": 8079 }, { "epoch": 1.68, "learning_rate": 4.8401162790697676e-08, "loss": 0.2107, "step": 8080 }, { "epoch": 1.68, "learning_rate": 4.837001661129568e-08, "loss": 0.1755, "step": 8081 }, { "epoch": 1.68, "learning_rate": 4.833887043189369e-08, "loss": 0.201, "step": 8082 }, { "epoch": 1.68, "learning_rate": 4.830772425249169e-08, "loss": 0.2697, "step": 8083 }, { "epoch": 1.68, "learning_rate": 4.82765780730897e-08, "loss": 0.1923, "step": 8084 }, { "epoch": 1.68, "learning_rate": 4.82454318936877e-08, "loss": 0.198, "step": 8085 }, { "epoch": 1.68, "learning_rate": 4.821428571428572e-08, "loss": 0.2096, "step": 8086 }, { "epoch": 1.68, "learning_rate": 4.818313953488372e-08, "loss": 0.1241, "step": 8087 }, { "epoch": 1.68, "learning_rate": 4.815199335548173e-08, "loss": 0.1697, "step": 8088 }, { "epoch": 1.68, "learning_rate": 4.812084717607973e-08, "loss": 0.1296, "step": 8089 }, { "epoch": 1.68, "learning_rate": 4.8089700996677735e-08, "loss": 0.2049, "step": 8090 }, { "epoch": 1.68, "learning_rate": 4.8058554817275744e-08, "loss": 0.1811, "step": 8091 }, { "epoch": 1.68, "learning_rate": 4.8027408637873747e-08, "loss": 0.1758, "step": 8092 }, { "epoch": 1.68, "learning_rate": 4.799626245847176e-08, "loss": 0.175, "step": 8093 }, { "epoch": 1.68, "learning_rate": 4.7965116279069765e-08, "loss": 0.2429, "step": 8094 }, { "epoch": 1.68, "learning_rate": 4.7933970099667774e-08, "loss": 0.2069, "step": 8095 }, { "epoch": 1.68, "learning_rate": 4.790282392026578e-08, "loss": 0.1538, "step": 8096 }, { "epoch": 1.68, "learning_rate": 4.7871677740863786e-08, "loss": 0.1783, "step": 8097 }, { "epoch": 1.68, "learning_rate": 4.784053156146179e-08, "loss": 0.1487, "step": 8098 }, { "epoch": 1.68, "learning_rate": 4.7809385382059804e-08, "loss": 0.1624, "step": 8099 }, { "epoch": 1.68, "learning_rate": 4.777823920265781e-08, "loss": 0.1829, "step": 8100 }, { "epoch": 1.68, "learning_rate": 4.7747093023255816e-08, "loss": 0.2188, "step": 8101 }, { "epoch": 1.68, "learning_rate": 4.771594684385382e-08, "loss": 0.1802, "step": 8102 }, { "epoch": 1.68, "learning_rate": 4.768480066445182e-08, "loss": 0.1657, "step": 8103 }, { "epoch": 1.68, "learning_rate": 4.765365448504983e-08, "loss": 0.2011, "step": 8104 }, { "epoch": 1.68, "learning_rate": 4.762250830564783e-08, "loss": 0.1497, "step": 8105 }, { "epoch": 1.68, "learning_rate": 4.759136212624585e-08, "loss": 0.1672, "step": 8106 }, { "epoch": 1.68, "learning_rate": 4.756021594684385e-08, "loss": 0.1883, "step": 8107 }, { "epoch": 1.68, "learning_rate": 4.752906976744186e-08, "loss": 0.1719, "step": 8108 }, { "epoch": 1.68, "learning_rate": 4.7497923588039863e-08, "loss": 0.1059, "step": 8109 }, { "epoch": 1.68, "learning_rate": 4.746677740863787e-08, "loss": 0.2165, "step": 8110 }, { "epoch": 1.68, "learning_rate": 4.7435631229235875e-08, "loss": 0.1729, "step": 8111 }, { "epoch": 1.68, "learning_rate": 4.7404485049833884e-08, "loss": 0.1849, "step": 8112 }, { "epoch": 1.68, "learning_rate": 4.7373338870431894e-08, "loss": 0.1391, "step": 8113 }, { "epoch": 1.68, "learning_rate": 4.73421926910299e-08, "loss": 0.2098, "step": 8114 }, { "epoch": 1.69, "learning_rate": 4.7311046511627905e-08, "loss": 0.2511, "step": 8115 }, { "epoch": 1.69, "learning_rate": 4.727990033222591e-08, "loss": 0.1133, "step": 8116 }, { "epoch": 1.69, "learning_rate": 4.724875415282392e-08, "loss": 0.1786, "step": 8117 }, { "epoch": 1.69, "learning_rate": 4.721760797342192e-08, "loss": 0.1666, "step": 8118 }, { "epoch": 1.69, "learning_rate": 4.7186461794019936e-08, "loss": 0.1428, "step": 8119 }, { "epoch": 1.69, "learning_rate": 4.715531561461794e-08, "loss": 0.159, "step": 8120 }, { "epoch": 1.69, "learning_rate": 4.712416943521595e-08, "loss": 0.1604, "step": 8121 }, { "epoch": 1.69, "learning_rate": 4.709302325581395e-08, "loss": 0.1735, "step": 8122 }, { "epoch": 1.69, "learning_rate": 4.706187707641196e-08, "loss": 0.1783, "step": 8123 }, { "epoch": 1.69, "learning_rate": 4.703073089700996e-08, "loss": 0.1401, "step": 8124 }, { "epoch": 1.69, "learning_rate": 4.699958471760797e-08, "loss": 0.1548, "step": 8125 }, { "epoch": 1.69, "learning_rate": 4.696843853820598e-08, "loss": 0.1892, "step": 8126 }, { "epoch": 1.69, "learning_rate": 4.693729235880399e-08, "loss": 0.1865, "step": 8127 }, { "epoch": 1.69, "learning_rate": 4.690614617940199e-08, "loss": 0.1482, "step": 8128 }, { "epoch": 1.69, "learning_rate": 4.6875e-08, "loss": 0.162, "step": 8129 }, { "epoch": 1.69, "learning_rate": 4.6843853820598004e-08, "loss": 0.1425, "step": 8130 }, { "epoch": 1.69, "learning_rate": 4.6812707641196006e-08, "loss": 0.1722, "step": 8131 }, { "epoch": 1.69, "learning_rate": 4.6781561461794016e-08, "loss": 0.1983, "step": 8132 }, { "epoch": 1.69, "learning_rate": 4.6750415282392025e-08, "loss": 0.14, "step": 8133 }, { "epoch": 1.69, "learning_rate": 4.6719269102990034e-08, "loss": 0.1613, "step": 8134 }, { "epoch": 1.69, "learning_rate": 4.6688122923588037e-08, "loss": 0.2206, "step": 8135 }, { "epoch": 1.69, "learning_rate": 4.6656976744186046e-08, "loss": 0.1749, "step": 8136 }, { "epoch": 1.69, "learning_rate": 4.662583056478405e-08, "loss": 0.159, "step": 8137 }, { "epoch": 1.69, "learning_rate": 4.659468438538206e-08, "loss": 0.1678, "step": 8138 }, { "epoch": 1.69, "learning_rate": 4.656353820598006e-08, "loss": 0.1829, "step": 8139 }, { "epoch": 1.69, "learning_rate": 4.6532392026578076e-08, "loss": 0.1572, "step": 8140 }, { "epoch": 1.69, "learning_rate": 4.650124584717608e-08, "loss": 0.2545, "step": 8141 }, { "epoch": 1.69, "learning_rate": 4.647009966777409e-08, "loss": 0.1297, "step": 8142 }, { "epoch": 1.69, "learning_rate": 4.643895348837209e-08, "loss": 0.2063, "step": 8143 }, { "epoch": 1.69, "learning_rate": 4.640780730897009e-08, "loss": 0.1921, "step": 8144 }, { "epoch": 1.69, "learning_rate": 4.63766611295681e-08, "loss": 0.1794, "step": 8145 }, { "epoch": 1.69, "learning_rate": 4.634551495016611e-08, "loss": 0.2046, "step": 8146 }, { "epoch": 1.69, "learning_rate": 4.631436877076412e-08, "loss": 0.141, "step": 8147 }, { "epoch": 1.69, "learning_rate": 4.628322259136212e-08, "loss": 0.1877, "step": 8148 }, { "epoch": 1.69, "learning_rate": 4.625207641196013e-08, "loss": 0.1257, "step": 8149 }, { "epoch": 1.69, "learning_rate": 4.6220930232558135e-08, "loss": 0.1769, "step": 8150 }, { "epoch": 1.69, "learning_rate": 4.6189784053156144e-08, "loss": 0.2211, "step": 8151 }, { "epoch": 1.69, "learning_rate": 4.615863787375415e-08, "loss": 0.2194, "step": 8152 }, { "epoch": 1.69, "learning_rate": 4.612749169435216e-08, "loss": 0.1066, "step": 8153 }, { "epoch": 1.69, "learning_rate": 4.6096345514950165e-08, "loss": 0.1977, "step": 8154 }, { "epoch": 1.69, "learning_rate": 4.6065199335548174e-08, "loss": 0.2037, "step": 8155 }, { "epoch": 1.69, "learning_rate": 4.603405315614618e-08, "loss": 0.2431, "step": 8156 }, { "epoch": 1.69, "learning_rate": 4.600290697674418e-08, "loss": 0.2025, "step": 8157 }, { "epoch": 1.69, "learning_rate": 4.597176079734219e-08, "loss": 0.2063, "step": 8158 }, { "epoch": 1.69, "learning_rate": 4.594061461794019e-08, "loss": 0.2001, "step": 8159 }, { "epoch": 1.69, "learning_rate": 4.590946843853821e-08, "loss": 0.193, "step": 8160 }, { "epoch": 1.69, "learning_rate": 4.587832225913621e-08, "loss": 0.1667, "step": 8161 }, { "epoch": 1.69, "learning_rate": 4.584717607973422e-08, "loss": 0.2432, "step": 8162 }, { "epoch": 1.69, "learning_rate": 4.581602990033222e-08, "loss": 0.1827, "step": 8163 }, { "epoch": 1.7, "learning_rate": 4.578488372093023e-08, "loss": 0.1497, "step": 8164 }, { "epoch": 1.7, "learning_rate": 4.5753737541528233e-08, "loss": 0.1874, "step": 8165 }, { "epoch": 1.7, "learning_rate": 4.572259136212625e-08, "loss": 0.2342, "step": 8166 }, { "epoch": 1.7, "learning_rate": 4.569144518272425e-08, "loss": 0.206, "step": 8167 }, { "epoch": 1.7, "learning_rate": 4.566029900332226e-08, "loss": 0.1694, "step": 8168 }, { "epoch": 1.7, "learning_rate": 4.5629152823920264e-08, "loss": 0.1872, "step": 8169 }, { "epoch": 1.7, "learning_rate": 4.5598006644518266e-08, "loss": 0.1459, "step": 8170 }, { "epoch": 1.7, "learning_rate": 4.5566860465116275e-08, "loss": 0.1437, "step": 8171 }, { "epoch": 1.7, "learning_rate": 4.553571428571428e-08, "loss": 0.1278, "step": 8172 }, { "epoch": 1.7, "learning_rate": 4.5504568106312294e-08, "loss": 0.1556, "step": 8173 }, { "epoch": 1.7, "learning_rate": 4.5473421926910296e-08, "loss": 0.1768, "step": 8174 }, { "epoch": 1.7, "learning_rate": 4.5442275747508306e-08, "loss": 0.155, "step": 8175 }, { "epoch": 1.7, "learning_rate": 4.541112956810631e-08, "loss": 0.2296, "step": 8176 }, { "epoch": 1.7, "learning_rate": 4.537998338870432e-08, "loss": 0.1274, "step": 8177 }, { "epoch": 1.7, "learning_rate": 4.534883720930232e-08, "loss": 0.1857, "step": 8178 }, { "epoch": 1.7, "learning_rate": 4.531769102990033e-08, "loss": 0.1116, "step": 8179 }, { "epoch": 1.7, "learning_rate": 4.528654485049834e-08, "loss": 0.1302, "step": 8180 }, { "epoch": 1.7, "learning_rate": 4.525539867109635e-08, "loss": 0.1754, "step": 8181 }, { "epoch": 1.7, "learning_rate": 4.522425249169435e-08, "loss": 0.1716, "step": 8182 }, { "epoch": 1.7, "learning_rate": 4.519310631229236e-08, "loss": 0.1494, "step": 8183 }, { "epoch": 1.7, "learning_rate": 4.516196013289036e-08, "loss": 0.1764, "step": 8184 }, { "epoch": 1.7, "learning_rate": 4.5130813953488365e-08, "loss": 0.1649, "step": 8185 }, { "epoch": 1.7, "learning_rate": 4.509966777408638e-08, "loss": 0.2763, "step": 8186 }, { "epoch": 1.7, "learning_rate": 4.506852159468438e-08, "loss": 0.217, "step": 8187 }, { "epoch": 1.7, "learning_rate": 4.503737541528239e-08, "loss": 0.1722, "step": 8188 }, { "epoch": 1.7, "learning_rate": 4.5006229235880395e-08, "loss": 0.174, "step": 8189 }, { "epoch": 1.7, "learning_rate": 4.4975083056478404e-08, "loss": 0.2066, "step": 8190 }, { "epoch": 1.7, "learning_rate": 4.4943936877076407e-08, "loss": 0.1715, "step": 8191 }, { "epoch": 1.7, "learning_rate": 4.4912790697674416e-08, "loss": 0.1668, "step": 8192 }, { "epoch": 1.7, "learning_rate": 4.4881644518272425e-08, "loss": 0.2343, "step": 8193 }, { "epoch": 1.7, "learning_rate": 4.4850498338870434e-08, "loss": 0.2276, "step": 8194 }, { "epoch": 1.7, "learning_rate": 4.481935215946844e-08, "loss": 0.1636, "step": 8195 }, { "epoch": 1.7, "learning_rate": 4.4788205980066446e-08, "loss": 0.1234, "step": 8196 }, { "epoch": 1.7, "learning_rate": 4.475705980066445e-08, "loss": 0.2469, "step": 8197 }, { "epoch": 1.7, "learning_rate": 4.472591362126245e-08, "loss": 0.1845, "step": 8198 }, { "epoch": 1.7, "learning_rate": 4.469476744186046e-08, "loss": 0.1819, "step": 8199 }, { "epoch": 1.7, "learning_rate": 4.4663621262458476e-08, "loss": 0.1857, "step": 8200 }, { "epoch": 1.7, "learning_rate": 4.463247508305648e-08, "loss": 0.172, "step": 8201 }, { "epoch": 1.7, "learning_rate": 4.460132890365448e-08, "loss": 0.2152, "step": 8202 }, { "epoch": 1.7, "learning_rate": 4.457018272425249e-08, "loss": 0.143, "step": 8203 }, { "epoch": 1.7, "learning_rate": 4.453903654485049e-08, "loss": 0.1343, "step": 8204 }, { "epoch": 1.7, "learning_rate": 4.45078903654485e-08, "loss": 0.1495, "step": 8205 }, { "epoch": 1.7, "learning_rate": 4.447674418604651e-08, "loss": 0.1315, "step": 8206 }, { "epoch": 1.7, "learning_rate": 4.444559800664452e-08, "loss": 0.2154, "step": 8207 }, { "epoch": 1.7, "learning_rate": 4.4414451827242523e-08, "loss": 0.1749, "step": 8208 }, { "epoch": 1.7, "learning_rate": 4.438330564784053e-08, "loss": 0.1528, "step": 8209 }, { "epoch": 1.7, "learning_rate": 4.4352159468438535e-08, "loss": 0.1854, "step": 8210 }, { "epoch": 1.7, "learning_rate": 4.432101328903654e-08, "loss": 0.1654, "step": 8211 }, { "epoch": 1.71, "learning_rate": 4.428986710963455e-08, "loss": 0.1523, "step": 8212 }, { "epoch": 1.71, "learning_rate": 4.425872093023256e-08, "loss": 0.2019, "step": 8213 }, { "epoch": 1.71, "learning_rate": 4.4227574750830565e-08, "loss": 0.1778, "step": 8214 }, { "epoch": 1.71, "learning_rate": 4.419642857142857e-08, "loss": 0.2183, "step": 8215 }, { "epoch": 1.71, "learning_rate": 4.416528239202658e-08, "loss": 0.1931, "step": 8216 }, { "epoch": 1.71, "learning_rate": 4.413413621262458e-08, "loss": 0.1527, "step": 8217 }, { "epoch": 1.71, "learning_rate": 4.410299003322259e-08, "loss": 0.195, "step": 8218 }, { "epoch": 1.71, "learning_rate": 4.407184385382059e-08, "loss": 0.1542, "step": 8219 }, { "epoch": 1.71, "learning_rate": 4.404069767441861e-08, "loss": 0.2183, "step": 8220 }, { "epoch": 1.71, "learning_rate": 4.400955149501661e-08, "loss": 0.1873, "step": 8221 }, { "epoch": 1.71, "learning_rate": 4.397840531561462e-08, "loss": 0.1529, "step": 8222 }, { "epoch": 1.71, "learning_rate": 4.394725913621262e-08, "loss": 0.2132, "step": 8223 }, { "epoch": 1.71, "learning_rate": 4.3916112956810624e-08, "loss": 0.1708, "step": 8224 }, { "epoch": 1.71, "learning_rate": 4.3884966777408634e-08, "loss": 0.1695, "step": 8225 }, { "epoch": 1.71, "learning_rate": 4.385382059800665e-08, "loss": 0.1693, "step": 8226 }, { "epoch": 1.71, "learning_rate": 4.382267441860465e-08, "loss": 0.1759, "step": 8227 }, { "epoch": 1.71, "learning_rate": 4.3791528239202655e-08, "loss": 0.1923, "step": 8228 }, { "epoch": 1.71, "learning_rate": 4.3760382059800664e-08, "loss": 0.2318, "step": 8229 }, { "epoch": 1.71, "learning_rate": 4.3729235880398666e-08, "loss": 0.1995, "step": 8230 }, { "epoch": 1.71, "learning_rate": 4.3698089700996676e-08, "loss": 0.1309, "step": 8231 }, { "epoch": 1.71, "learning_rate": 4.366694352159468e-08, "loss": 0.204, "step": 8232 }, { "epoch": 1.71, "learning_rate": 4.3635797342192694e-08, "loss": 0.183, "step": 8233 }, { "epoch": 1.71, "learning_rate": 4.3604651162790697e-08, "loss": 0.1739, "step": 8234 }, { "epoch": 1.71, "learning_rate": 4.3573504983388706e-08, "loss": 0.1445, "step": 8235 }, { "epoch": 1.71, "learning_rate": 4.354235880398671e-08, "loss": 0.2203, "step": 8236 }, { "epoch": 1.71, "learning_rate": 4.351121262458472e-08, "loss": 0.1721, "step": 8237 }, { "epoch": 1.71, "learning_rate": 4.348006644518272e-08, "loss": 0.1406, "step": 8238 }, { "epoch": 1.71, "learning_rate": 4.344892026578072e-08, "loss": 0.1923, "step": 8239 }, { "epoch": 1.71, "learning_rate": 4.341777408637874e-08, "loss": 0.1766, "step": 8240 }, { "epoch": 1.71, "learning_rate": 4.338662790697674e-08, "loss": 0.1155, "step": 8241 }, { "epoch": 1.71, "learning_rate": 4.335548172757475e-08, "loss": 0.1994, "step": 8242 }, { "epoch": 1.71, "learning_rate": 4.332433554817275e-08, "loss": 0.1807, "step": 8243 }, { "epoch": 1.71, "learning_rate": 4.329318936877076e-08, "loss": 0.1499, "step": 8244 }, { "epoch": 1.71, "learning_rate": 4.3262043189368765e-08, "loss": 0.1663, "step": 8245 }, { "epoch": 1.71, "learning_rate": 4.3230897009966774e-08, "loss": 0.1799, "step": 8246 }, { "epoch": 1.71, "learning_rate": 4.319975083056478e-08, "loss": 0.1103, "step": 8247 }, { "epoch": 1.71, "learning_rate": 4.316860465116279e-08, "loss": 0.1967, "step": 8248 }, { "epoch": 1.71, "learning_rate": 4.3137458471760795e-08, "loss": 0.1471, "step": 8249 }, { "epoch": 1.71, "learning_rate": 4.3106312292358804e-08, "loss": 0.1907, "step": 8250 }, { "epoch": 1.71, "learning_rate": 4.307516611295681e-08, "loss": 0.1659, "step": 8251 }, { "epoch": 1.71, "learning_rate": 4.304401993355481e-08, "loss": 0.1612, "step": 8252 }, { "epoch": 1.71, "learning_rate": 4.3012873754152825e-08, "loss": 0.1512, "step": 8253 }, { "epoch": 1.71, "learning_rate": 4.2981727574750834e-08, "loss": 0.169, "step": 8254 }, { "epoch": 1.71, "learning_rate": 4.295058139534884e-08, "loss": 0.1449, "step": 8255 }, { "epoch": 1.71, "learning_rate": 4.291943521594684e-08, "loss": 0.1455, "step": 8256 }, { "epoch": 1.71, "learning_rate": 4.288828903654485e-08, "loss": 0.2029, "step": 8257 }, { "epoch": 1.71, "learning_rate": 4.285714285714285e-08, "loss": 0.1823, "step": 8258 }, { "epoch": 1.71, "learning_rate": 4.282599667774086e-08, "loss": 0.2017, "step": 8259 }, { "epoch": 1.72, "learning_rate": 4.279485049833887e-08, "loss": 0.1577, "step": 8260 }, { "epoch": 1.72, "learning_rate": 4.276370431893688e-08, "loss": 0.2145, "step": 8261 }, { "epoch": 1.72, "learning_rate": 4.273255813953488e-08, "loss": 0.195, "step": 8262 }, { "epoch": 1.72, "learning_rate": 4.270141196013289e-08, "loss": 0.1588, "step": 8263 }, { "epoch": 1.72, "learning_rate": 4.2670265780730893e-08, "loss": 0.1913, "step": 8264 }, { "epoch": 1.72, "learning_rate": 4.2639119601328896e-08, "loss": 0.122, "step": 8265 }, { "epoch": 1.72, "learning_rate": 4.2607973421926905e-08, "loss": 0.2005, "step": 8266 }, { "epoch": 1.72, "learning_rate": 4.257682724252492e-08, "loss": 0.1706, "step": 8267 }, { "epoch": 1.72, "learning_rate": 4.2545681063122924e-08, "loss": 0.1769, "step": 8268 }, { "epoch": 1.72, "learning_rate": 4.2514534883720926e-08, "loss": 0.1858, "step": 8269 }, { "epoch": 1.72, "learning_rate": 4.2483388704318935e-08, "loss": 0.1448, "step": 8270 }, { "epoch": 1.72, "learning_rate": 4.245224252491694e-08, "loss": 0.1695, "step": 8271 }, { "epoch": 1.72, "learning_rate": 4.242109634551495e-08, "loss": 0.1231, "step": 8272 }, { "epoch": 1.72, "learning_rate": 4.2389950166112956e-08, "loss": 0.181, "step": 8273 }, { "epoch": 1.72, "learning_rate": 4.2358803986710966e-08, "loss": 0.193, "step": 8274 }, { "epoch": 1.72, "learning_rate": 4.232765780730897e-08, "loss": 0.1166, "step": 8275 }, { "epoch": 1.72, "learning_rate": 4.229651162790698e-08, "loss": 0.1444, "step": 8276 }, { "epoch": 1.72, "learning_rate": 4.226536544850498e-08, "loss": 0.256, "step": 8277 }, { "epoch": 1.72, "learning_rate": 4.223421926910298e-08, "loss": 0.159, "step": 8278 }, { "epoch": 1.72, "learning_rate": 4.220307308970099e-08, "loss": 0.2044, "step": 8279 }, { "epoch": 1.72, "learning_rate": 4.217192691029901e-08, "loss": 0.2471, "step": 8280 }, { "epoch": 1.72, "learning_rate": 4.214078073089701e-08, "loss": 0.1887, "step": 8281 }, { "epoch": 1.72, "learning_rate": 4.210963455149501e-08, "loss": 0.2105, "step": 8282 }, { "epoch": 1.72, "learning_rate": 4.207848837209302e-08, "loss": 0.1937, "step": 8283 }, { "epoch": 1.72, "learning_rate": 4.2047342192691025e-08, "loss": 0.2071, "step": 8284 }, { "epoch": 1.72, "learning_rate": 4.2016196013289034e-08, "loss": 0.1626, "step": 8285 }, { "epoch": 1.72, "learning_rate": 4.1985049833887036e-08, "loss": 0.1915, "step": 8286 }, { "epoch": 1.72, "learning_rate": 4.195390365448505e-08, "loss": 0.152, "step": 8287 }, { "epoch": 1.72, "learning_rate": 4.1922757475083055e-08, "loss": 0.193, "step": 8288 }, { "epoch": 1.72, "learning_rate": 4.1891611295681064e-08, "loss": 0.2288, "step": 8289 }, { "epoch": 1.72, "learning_rate": 4.1860465116279067e-08, "loss": 0.1754, "step": 8290 }, { "epoch": 1.72, "learning_rate": 4.1829318936877076e-08, "loss": 0.2135, "step": 8291 }, { "epoch": 1.72, "learning_rate": 4.179817275747508e-08, "loss": 0.1893, "step": 8292 }, { "epoch": 1.72, "learning_rate": 4.1767026578073094e-08, "loss": 0.2478, "step": 8293 }, { "epoch": 1.72, "learning_rate": 4.17358803986711e-08, "loss": 0.1389, "step": 8294 }, { "epoch": 1.72, "learning_rate": 4.17047342192691e-08, "loss": 0.2073, "step": 8295 }, { "epoch": 1.72, "learning_rate": 4.167358803986711e-08, "loss": 0.2295, "step": 8296 }, { "epoch": 1.72, "learning_rate": 4.164244186046511e-08, "loss": 0.1266, "step": 8297 }, { "epoch": 1.72, "learning_rate": 4.161129568106312e-08, "loss": 0.2017, "step": 8298 }, { "epoch": 1.72, "learning_rate": 4.158014950166112e-08, "loss": 0.1765, "step": 8299 }, { "epoch": 1.72, "learning_rate": 4.154900332225914e-08, "loss": 0.2406, "step": 8300 }, { "epoch": 1.72, "learning_rate": 4.151785714285714e-08, "loss": 0.1964, "step": 8301 }, { "epoch": 1.72, "learning_rate": 4.148671096345515e-08, "loss": 0.2132, "step": 8302 }, { "epoch": 1.72, "learning_rate": 4.145556478405315e-08, "loss": 0.1893, "step": 8303 }, { "epoch": 1.72, "learning_rate": 4.142441860465116e-08, "loss": 0.1937, "step": 8304 }, { "epoch": 1.72, "learning_rate": 4.1393272425249165e-08, "loss": 0.2332, "step": 8305 }, { "epoch": 1.72, "learning_rate": 4.136212624584717e-08, "loss": 0.168, "step": 8306 }, { "epoch": 1.72, "learning_rate": 4.1330980066445183e-08, "loss": 0.2114, "step": 8307 }, { "epoch": 1.73, "learning_rate": 4.129983388704319e-08, "loss": 0.1809, "step": 8308 }, { "epoch": 1.73, "learning_rate": 4.1268687707641195e-08, "loss": 0.1486, "step": 8309 }, { "epoch": 1.73, "learning_rate": 4.12375415282392e-08, "loss": 0.1854, "step": 8310 }, { "epoch": 1.73, "learning_rate": 4.120639534883721e-08, "loss": 0.1804, "step": 8311 }, { "epoch": 1.73, "learning_rate": 4.117524916943521e-08, "loss": 0.1658, "step": 8312 }, { "epoch": 1.73, "learning_rate": 4.1144102990033225e-08, "loss": 0.1961, "step": 8313 }, { "epoch": 1.73, "learning_rate": 4.111295681063123e-08, "loss": 0.162, "step": 8314 }, { "epoch": 1.73, "learning_rate": 4.108181063122924e-08, "loss": 0.1811, "step": 8315 }, { "epoch": 1.73, "learning_rate": 4.105066445182724e-08, "loss": 0.1345, "step": 8316 }, { "epoch": 1.73, "learning_rate": 4.101951827242525e-08, "loss": 0.1592, "step": 8317 }, { "epoch": 1.73, "learning_rate": 4.098837209302325e-08, "loss": 0.1897, "step": 8318 }, { "epoch": 1.73, "learning_rate": 4.0957225913621254e-08, "loss": 0.2046, "step": 8319 }, { "epoch": 1.73, "learning_rate": 4.092607973421927e-08, "loss": 0.2073, "step": 8320 }, { "epoch": 1.73, "learning_rate": 4.089493355481728e-08, "loss": 0.1877, "step": 8321 }, { "epoch": 1.73, "learning_rate": 4.086378737541528e-08, "loss": 0.1389, "step": 8322 }, { "epoch": 1.73, "learning_rate": 4.0832641196013284e-08, "loss": 0.1657, "step": 8323 }, { "epoch": 1.73, "learning_rate": 4.0801495016611294e-08, "loss": 0.1831, "step": 8324 }, { "epoch": 1.73, "learning_rate": 4.0770348837209296e-08, "loss": 0.1707, "step": 8325 }, { "epoch": 1.73, "learning_rate": 4.0739202657807305e-08, "loss": 0.2126, "step": 8326 }, { "epoch": 1.73, "learning_rate": 4.0708056478405315e-08, "loss": 0.1835, "step": 8327 }, { "epoch": 1.73, "learning_rate": 4.0676910299003324e-08, "loss": 0.1658, "step": 8328 }, { "epoch": 1.73, "learning_rate": 4.0645764119601326e-08, "loss": 0.1318, "step": 8329 }, { "epoch": 1.73, "learning_rate": 4.0614617940199336e-08, "loss": 0.1979, "step": 8330 }, { "epoch": 1.73, "learning_rate": 4.058347176079734e-08, "loss": 0.2174, "step": 8331 }, { "epoch": 1.73, "learning_rate": 4.055232558139534e-08, "loss": 0.2428, "step": 8332 }, { "epoch": 1.73, "learning_rate": 4.052117940199335e-08, "loss": 0.1423, "step": 8333 }, { "epoch": 1.73, "learning_rate": 4.0490033222591366e-08, "loss": 0.1456, "step": 8334 }, { "epoch": 1.73, "learning_rate": 4.045888704318937e-08, "loss": 0.168, "step": 8335 }, { "epoch": 1.73, "learning_rate": 4.042774086378737e-08, "loss": 0.1979, "step": 8336 }, { "epoch": 1.73, "learning_rate": 4.039659468438538e-08, "loss": 0.1884, "step": 8337 }, { "epoch": 1.73, "learning_rate": 4.036544850498338e-08, "loss": 0.193, "step": 8338 }, { "epoch": 1.73, "learning_rate": 4.033430232558139e-08, "loss": 0.1727, "step": 8339 }, { "epoch": 1.73, "learning_rate": 4.03031561461794e-08, "loss": 0.177, "step": 8340 }, { "epoch": 1.73, "learning_rate": 4.027200996677741e-08, "loss": 0.2224, "step": 8341 }, { "epoch": 1.73, "learning_rate": 4.024086378737541e-08, "loss": 0.204, "step": 8342 }, { "epoch": 1.73, "learning_rate": 4.020971760797342e-08, "loss": 0.2411, "step": 8343 }, { "epoch": 1.73, "learning_rate": 4.0178571428571425e-08, "loss": 0.1564, "step": 8344 }, { "epoch": 1.73, "learning_rate": 4.0147425249169434e-08, "loss": 0.2161, "step": 8345 }, { "epoch": 1.73, "learning_rate": 4.0116279069767437e-08, "loss": 0.2345, "step": 8346 }, { "epoch": 1.73, "learning_rate": 4.008513289036545e-08, "loss": 0.1531, "step": 8347 }, { "epoch": 1.73, "learning_rate": 4.0053986710963455e-08, "loss": 0.1836, "step": 8348 }, { "epoch": 1.73, "learning_rate": 4.002284053156146e-08, "loss": 0.1928, "step": 8349 }, { "epoch": 1.73, "learning_rate": 3.999169435215947e-08, "loss": 0.1554, "step": 8350 }, { "epoch": 1.73, "learning_rate": 3.996054817275747e-08, "loss": 0.2497, "step": 8351 }, { "epoch": 1.73, "learning_rate": 3.992940199335548e-08, "loss": 0.1652, "step": 8352 }, { "epoch": 1.73, "learning_rate": 3.989825581395348e-08, "loss": 0.1557, "step": 8353 }, { "epoch": 1.73, "learning_rate": 3.98671096345515e-08, "loss": 0.1496, "step": 8354 }, { "epoch": 1.73, "learning_rate": 3.98359634551495e-08, "loss": 0.1661, "step": 8355 }, { "epoch": 1.74, "learning_rate": 3.980481727574751e-08, "loss": 0.1455, "step": 8356 }, { "epoch": 1.74, "learning_rate": 3.977367109634551e-08, "loss": 0.1905, "step": 8357 }, { "epoch": 1.74, "learning_rate": 3.974252491694352e-08, "loss": 0.2075, "step": 8358 }, { "epoch": 1.74, "learning_rate": 3.971137873754152e-08, "loss": 0.2134, "step": 8359 }, { "epoch": 1.74, "learning_rate": 3.968023255813954e-08, "loss": 0.188, "step": 8360 }, { "epoch": 1.74, "learning_rate": 3.964908637873754e-08, "loss": 0.1785, "step": 8361 }, { "epoch": 1.74, "learning_rate": 3.961794019933555e-08, "loss": 0.1694, "step": 8362 }, { "epoch": 1.74, "learning_rate": 3.9586794019933553e-08, "loss": 0.2167, "step": 8363 }, { "epoch": 1.74, "learning_rate": 3.9555647840531556e-08, "loss": 0.1086, "step": 8364 }, { "epoch": 1.74, "learning_rate": 3.9524501661129565e-08, "loss": 0.1783, "step": 8365 }, { "epoch": 1.74, "learning_rate": 3.949335548172757e-08, "loss": 0.1707, "step": 8366 }, { "epoch": 1.74, "learning_rate": 3.9462209302325584e-08, "loss": 0.1814, "step": 8367 }, { "epoch": 1.74, "learning_rate": 3.9431063122923586e-08, "loss": 0.1726, "step": 8368 }, { "epoch": 1.74, "learning_rate": 3.9399916943521595e-08, "loss": 0.1646, "step": 8369 }, { "epoch": 1.74, "learning_rate": 3.93687707641196e-08, "loss": 0.2335, "step": 8370 }, { "epoch": 1.74, "learning_rate": 3.933762458471761e-08, "loss": 0.207, "step": 8371 }, { "epoch": 1.74, "learning_rate": 3.930647840531561e-08, "loss": 0.1647, "step": 8372 }, { "epoch": 1.74, "learning_rate": 3.927533222591361e-08, "loss": 0.1689, "step": 8373 }, { "epoch": 1.74, "learning_rate": 3.924418604651163e-08, "loss": 0.1898, "step": 8374 }, { "epoch": 1.74, "learning_rate": 3.921303986710964e-08, "loss": 0.2612, "step": 8375 }, { "epoch": 1.74, "learning_rate": 3.918189368770764e-08, "loss": 0.1661, "step": 8376 }, { "epoch": 1.74, "learning_rate": 3.915074750830564e-08, "loss": 0.2112, "step": 8377 }, { "epoch": 1.74, "learning_rate": 3.911960132890365e-08, "loss": 0.2352, "step": 8378 }, { "epoch": 1.74, "learning_rate": 3.9088455149501654e-08, "loss": 0.1788, "step": 8379 }, { "epoch": 1.74, "learning_rate": 3.905730897009967e-08, "loss": 0.1432, "step": 8380 }, { "epoch": 1.74, "learning_rate": 3.902616279069767e-08, "loss": 0.1928, "step": 8381 }, { "epoch": 1.74, "learning_rate": 3.899501661129568e-08, "loss": 0.2114, "step": 8382 }, { "epoch": 1.74, "learning_rate": 3.8963870431893685e-08, "loss": 0.1776, "step": 8383 }, { "epoch": 1.74, "learning_rate": 3.8932724252491694e-08, "loss": 0.2256, "step": 8384 }, { "epoch": 1.74, "learning_rate": 3.8901578073089696e-08, "loss": 0.1687, "step": 8385 }, { "epoch": 1.74, "learning_rate": 3.88704318936877e-08, "loss": 0.1904, "step": 8386 }, { "epoch": 1.74, "learning_rate": 3.8839285714285715e-08, "loss": 0.1982, "step": 8387 }, { "epoch": 1.74, "learning_rate": 3.8808139534883724e-08, "loss": 0.1752, "step": 8388 }, { "epoch": 1.74, "learning_rate": 3.8776993355481727e-08, "loss": 0.1898, "step": 8389 }, { "epoch": 1.74, "learning_rate": 3.874584717607973e-08, "loss": 0.1116, "step": 8390 }, { "epoch": 1.74, "learning_rate": 3.871470099667774e-08, "loss": 0.1594, "step": 8391 }, { "epoch": 1.74, "learning_rate": 3.868355481727574e-08, "loss": 0.1899, "step": 8392 }, { "epoch": 1.74, "learning_rate": 3.865240863787375e-08, "loss": 0.2564, "step": 8393 }, { "epoch": 1.74, "learning_rate": 3.862126245847176e-08, "loss": 0.1605, "step": 8394 }, { "epoch": 1.74, "learning_rate": 3.859011627906977e-08, "loss": 0.2301, "step": 8395 }, { "epoch": 1.74, "learning_rate": 3.855897009966777e-08, "loss": 0.2144, "step": 8396 }, { "epoch": 1.74, "learning_rate": 3.852782392026578e-08, "loss": 0.1686, "step": 8397 }, { "epoch": 1.74, "learning_rate": 3.849667774086378e-08, "loss": 0.2029, "step": 8398 }, { "epoch": 1.74, "learning_rate": 3.846553156146179e-08, "loss": 0.1647, "step": 8399 }, { "epoch": 1.74, "learning_rate": 3.84343853820598e-08, "loss": 0.1975, "step": 8400 }, { "epoch": 1.74, "learning_rate": 3.840323920265781e-08, "loss": 0.1601, "step": 8401 }, { "epoch": 1.74, "learning_rate": 3.837209302325581e-08, "loss": 0.1871, "step": 8402 }, { "epoch": 1.74, "learning_rate": 3.8340946843853816e-08, "loss": 0.1754, "step": 8403 }, { "epoch": 1.75, "learning_rate": 3.8309800664451825e-08, "loss": 0.1732, "step": 8404 }, { "epoch": 1.75, "learning_rate": 3.827865448504983e-08, "loss": 0.1526, "step": 8405 }, { "epoch": 1.75, "learning_rate": 3.824750830564784e-08, "loss": 0.1975, "step": 8406 }, { "epoch": 1.75, "learning_rate": 3.8216362126245846e-08, "loss": 0.1495, "step": 8407 }, { "epoch": 1.75, "learning_rate": 3.8185215946843855e-08, "loss": 0.1737, "step": 8408 }, { "epoch": 1.75, "learning_rate": 3.815406976744186e-08, "loss": 0.3138, "step": 8409 }, { "epoch": 1.75, "learning_rate": 3.812292358803987e-08, "loss": 0.1613, "step": 8410 }, { "epoch": 1.75, "learning_rate": 3.809177740863787e-08, "loss": 0.1831, "step": 8411 }, { "epoch": 1.75, "learning_rate": 3.806063122923588e-08, "loss": 0.2799, "step": 8412 }, { "epoch": 1.75, "learning_rate": 3.802948504983388e-08, "loss": 0.2009, "step": 8413 }, { "epoch": 1.75, "learning_rate": 3.79983388704319e-08, "loss": 0.239, "step": 8414 }, { "epoch": 1.75, "learning_rate": 3.79671926910299e-08, "loss": 0.1723, "step": 8415 }, { "epoch": 1.75, "learning_rate": 3.793604651162791e-08, "loss": 0.2469, "step": 8416 }, { "epoch": 1.75, "learning_rate": 3.790490033222591e-08, "loss": 0.1739, "step": 8417 }, { "epoch": 1.75, "learning_rate": 3.7873754152823914e-08, "loss": 0.1746, "step": 8418 }, { "epoch": 1.75, "learning_rate": 3.7842607973421923e-08, "loss": 0.1963, "step": 8419 }, { "epoch": 1.75, "learning_rate": 3.781146179401993e-08, "loss": 0.17, "step": 8420 }, { "epoch": 1.75, "learning_rate": 3.778031561461794e-08, "loss": 0.1629, "step": 8421 }, { "epoch": 1.75, "learning_rate": 3.7749169435215944e-08, "loss": 0.1356, "step": 8422 }, { "epoch": 1.75, "learning_rate": 3.7718023255813954e-08, "loss": 0.178, "step": 8423 }, { "epoch": 1.75, "learning_rate": 3.7686877076411956e-08, "loss": 0.1248, "step": 8424 }, { "epoch": 1.75, "learning_rate": 3.7655730897009966e-08, "loss": 0.1866, "step": 8425 }, { "epoch": 1.75, "learning_rate": 3.762458471760797e-08, "loss": 0.2344, "step": 8426 }, { "epoch": 1.75, "learning_rate": 3.7593438538205984e-08, "loss": 0.2206, "step": 8427 }, { "epoch": 1.75, "learning_rate": 3.7562292358803987e-08, "loss": 0.28, "step": 8428 }, { "epoch": 1.75, "learning_rate": 3.7531146179401996e-08, "loss": 0.1751, "step": 8429 }, { "epoch": 1.75, "learning_rate": 3.75e-08, "loss": 0.1522, "step": 8430 }, { "epoch": 1.75, "learning_rate": 3.7468853820598e-08, "loss": 0.1816, "step": 8431 }, { "epoch": 1.75, "learning_rate": 3.743770764119601e-08, "loss": 0.1656, "step": 8432 }, { "epoch": 1.75, "learning_rate": 3.740656146179402e-08, "loss": 0.1706, "step": 8433 }, { "epoch": 1.75, "learning_rate": 3.737541528239202e-08, "loss": 0.22, "step": 8434 }, { "epoch": 1.75, "learning_rate": 3.734426910299003e-08, "loss": 0.1959, "step": 8435 }, { "epoch": 1.75, "learning_rate": 3.731312292358804e-08, "loss": 0.1423, "step": 8436 }, { "epoch": 1.75, "learning_rate": 3.728197674418604e-08, "loss": 0.2462, "step": 8437 }, { "epoch": 1.75, "learning_rate": 3.725083056478405e-08, "loss": 0.1484, "step": 8438 }, { "epoch": 1.75, "learning_rate": 3.721968438538206e-08, "loss": 0.1366, "step": 8439 }, { "epoch": 1.75, "learning_rate": 3.7188538205980064e-08, "loss": 0.159, "step": 8440 }, { "epoch": 1.75, "learning_rate": 3.715739202657807e-08, "loss": 0.1735, "step": 8441 }, { "epoch": 1.75, "learning_rate": 3.712624584717608e-08, "loss": 0.1545, "step": 8442 }, { "epoch": 1.75, "learning_rate": 3.7095099667774085e-08, "loss": 0.2049, "step": 8443 }, { "epoch": 1.75, "learning_rate": 3.706395348837209e-08, "loss": 0.1465, "step": 8444 }, { "epoch": 1.75, "learning_rate": 3.70328073089701e-08, "loss": 0.1892, "step": 8445 }, { "epoch": 1.75, "learning_rate": 3.7001661129568106e-08, "loss": 0.1949, "step": 8446 }, { "epoch": 1.75, "learning_rate": 3.697051495016611e-08, "loss": 0.1589, "step": 8447 }, { "epoch": 1.75, "learning_rate": 3.693936877076412e-08, "loss": 0.1344, "step": 8448 }, { "epoch": 1.75, "learning_rate": 3.690822259136213e-08, "loss": 0.254, "step": 8449 }, { "epoch": 1.75, "learning_rate": 3.687707641196013e-08, "loss": 0.2471, "step": 8450 }, { "epoch": 1.75, "learning_rate": 3.684593023255814e-08, "loss": 0.123, "step": 8451 }, { "epoch": 1.75, "learning_rate": 3.681478405315615e-08, "loss": 0.2086, "step": 8452 }, { "epoch": 1.76, "learning_rate": 3.678363787375415e-08, "loss": 0.1843, "step": 8453 }, { "epoch": 1.76, "learning_rate": 3.675249169435216e-08, "loss": 0.1832, "step": 8454 }, { "epoch": 1.76, "learning_rate": 3.672134551495016e-08, "loss": 0.1501, "step": 8455 }, { "epoch": 1.76, "learning_rate": 3.669019933554817e-08, "loss": 0.2011, "step": 8456 }, { "epoch": 1.76, "learning_rate": 3.665905315614618e-08, "loss": 0.1878, "step": 8457 }, { "epoch": 1.76, "learning_rate": 3.6627906976744183e-08, "loss": 0.1918, "step": 8458 }, { "epoch": 1.76, "learning_rate": 3.659676079734219e-08, "loss": 0.1576, "step": 8459 }, { "epoch": 1.76, "learning_rate": 3.6565614617940195e-08, "loss": 0.1862, "step": 8460 }, { "epoch": 1.76, "learning_rate": 3.6534468438538204e-08, "loss": 0.2492, "step": 8461 }, { "epoch": 1.76, "learning_rate": 3.6503322259136214e-08, "loss": 0.1574, "step": 8462 }, { "epoch": 1.76, "learning_rate": 3.6472176079734216e-08, "loss": 0.209, "step": 8463 }, { "epoch": 1.76, "learning_rate": 3.6441029900332225e-08, "loss": 0.1315, "step": 8464 }, { "epoch": 1.76, "learning_rate": 3.640988372093023e-08, "loss": 0.2436, "step": 8465 }, { "epoch": 1.76, "learning_rate": 3.637873754152824e-08, "loss": 0.1406, "step": 8466 }, { "epoch": 1.76, "learning_rate": 3.6347591362126246e-08, "loss": 0.2037, "step": 8467 }, { "epoch": 1.76, "learning_rate": 3.631644518272425e-08, "loss": 0.1356, "step": 8468 }, { "epoch": 1.76, "learning_rate": 3.628529900332226e-08, "loss": 0.1829, "step": 8469 }, { "epoch": 1.76, "learning_rate": 3.625415282392027e-08, "loss": 0.2253, "step": 8470 }, { "epoch": 1.76, "learning_rate": 3.622300664451827e-08, "loss": 0.1718, "step": 8471 }, { "epoch": 1.76, "learning_rate": 3.619186046511628e-08, "loss": 0.1802, "step": 8472 }, { "epoch": 1.76, "learning_rate": 3.616071428571428e-08, "loss": 0.1768, "step": 8473 }, { "epoch": 1.76, "learning_rate": 3.612956810631229e-08, "loss": 0.1483, "step": 8474 }, { "epoch": 1.76, "learning_rate": 3.6098421926910294e-08, "loss": 0.1365, "step": 8475 }, { "epoch": 1.76, "learning_rate": 3.60672757475083e-08, "loss": 0.1814, "step": 8476 }, { "epoch": 1.76, "learning_rate": 3.603612956810631e-08, "loss": 0.1556, "step": 8477 }, { "epoch": 1.76, "learning_rate": 3.6004983388704315e-08, "loss": 0.1074, "step": 8478 }, { "epoch": 1.76, "learning_rate": 3.5973837209302324e-08, "loss": 0.1655, "step": 8479 }, { "epoch": 1.76, "learning_rate": 3.594269102990033e-08, "loss": 0.1308, "step": 8480 }, { "epoch": 1.76, "learning_rate": 3.5911544850498336e-08, "loss": 0.1273, "step": 8481 }, { "epoch": 1.76, "learning_rate": 3.5880398671096345e-08, "loss": 0.2048, "step": 8482 }, { "epoch": 1.76, "learning_rate": 3.5849252491694354e-08, "loss": 0.1873, "step": 8483 }, { "epoch": 1.76, "learning_rate": 3.5818106312292357e-08, "loss": 0.1508, "step": 8484 }, { "epoch": 1.76, "learning_rate": 3.578696013289036e-08, "loss": 0.1658, "step": 8485 }, { "epoch": 1.76, "learning_rate": 3.575581395348837e-08, "loss": 0.1882, "step": 8486 }, { "epoch": 1.76, "learning_rate": 3.572466777408638e-08, "loss": 0.1685, "step": 8487 }, { "epoch": 1.76, "learning_rate": 3.569352159468438e-08, "loss": 0.2165, "step": 8488 }, { "epoch": 1.76, "learning_rate": 3.566237541528239e-08, "loss": 0.1628, "step": 8489 }, { "epoch": 1.76, "learning_rate": 3.56312292358804e-08, "loss": 0.1516, "step": 8490 }, { "epoch": 1.76, "learning_rate": 3.56000830564784e-08, "loss": 0.1614, "step": 8491 }, { "epoch": 1.76, "learning_rate": 3.556893687707641e-08, "loss": 0.1645, "step": 8492 }, { "epoch": 1.76, "learning_rate": 3.553779069767442e-08, "loss": 0.1513, "step": 8493 }, { "epoch": 1.76, "learning_rate": 3.550664451827242e-08, "loss": 0.1797, "step": 8494 }, { "epoch": 1.76, "learning_rate": 3.547549833887043e-08, "loss": 0.1785, "step": 8495 }, { "epoch": 1.76, "learning_rate": 3.544435215946844e-08, "loss": 0.1593, "step": 8496 }, { "epoch": 1.76, "learning_rate": 3.541320598006644e-08, "loss": 0.1727, "step": 8497 }, { "epoch": 1.76, "learning_rate": 3.5382059800664446e-08, "loss": 0.246, "step": 8498 }, { "epoch": 1.76, "learning_rate": 3.535091362126246e-08, "loss": 0.238, "step": 8499 }, { "epoch": 1.76, "learning_rate": 3.5319767441860464e-08, "loss": 0.1681, "step": 8500 }, { "epoch": 1.77, "learning_rate": 3.528862126245847e-08, "loss": 0.1618, "step": 8501 }, { "epoch": 1.77, "learning_rate": 3.5257475083056476e-08, "loss": 0.1508, "step": 8502 }, { "epoch": 1.77, "learning_rate": 3.5226328903654485e-08, "loss": 0.2109, "step": 8503 }, { "epoch": 1.77, "learning_rate": 3.519518272425249e-08, "loss": 0.1838, "step": 8504 }, { "epoch": 1.77, "learning_rate": 3.51640365448505e-08, "loss": 0.1651, "step": 8505 }, { "epoch": 1.77, "learning_rate": 3.5132890365448506e-08, "loss": 0.2292, "step": 8506 }, { "epoch": 1.77, "learning_rate": 3.510174418604651e-08, "loss": 0.1385, "step": 8507 }, { "epoch": 1.77, "learning_rate": 3.507059800664452e-08, "loss": 0.1544, "step": 8508 }, { "epoch": 1.77, "learning_rate": 3.503945182724253e-08, "loss": 0.223, "step": 8509 }, { "epoch": 1.77, "learning_rate": 3.500830564784053e-08, "loss": 0.2259, "step": 8510 }, { "epoch": 1.77, "learning_rate": 3.497715946843854e-08, "loss": 0.151, "step": 8511 }, { "epoch": 1.77, "learning_rate": 3.494601328903655e-08, "loss": 0.1447, "step": 8512 }, { "epoch": 1.77, "learning_rate": 3.491486710963455e-08, "loss": 0.2439, "step": 8513 }, { "epoch": 1.77, "learning_rate": 3.4883720930232553e-08, "loss": 0.2145, "step": 8514 }, { "epoch": 1.77, "learning_rate": 3.485257475083056e-08, "loss": 0.2106, "step": 8515 }, { "epoch": 1.77, "learning_rate": 3.482142857142857e-08, "loss": 0.1737, "step": 8516 }, { "epoch": 1.77, "learning_rate": 3.4790282392026574e-08, "loss": 0.1784, "step": 8517 }, { "epoch": 1.77, "learning_rate": 3.4759136212624584e-08, "loss": 0.2167, "step": 8518 }, { "epoch": 1.77, "learning_rate": 3.472799003322259e-08, "loss": 0.2088, "step": 8519 }, { "epoch": 1.77, "learning_rate": 3.4696843853820595e-08, "loss": 0.1893, "step": 8520 }, { "epoch": 1.77, "learning_rate": 3.4665697674418605e-08, "loss": 0.1216, "step": 8521 }, { "epoch": 1.77, "learning_rate": 3.4634551495016614e-08, "loss": 0.1376, "step": 8522 }, { "epoch": 1.77, "learning_rate": 3.4603405315614616e-08, "loss": 0.1903, "step": 8523 }, { "epoch": 1.77, "learning_rate": 3.4572259136212626e-08, "loss": 0.1417, "step": 8524 }, { "epoch": 1.77, "learning_rate": 3.454111295681063e-08, "loss": 0.1929, "step": 8525 }, { "epoch": 1.77, "learning_rate": 3.450996677740864e-08, "loss": 0.1582, "step": 8526 }, { "epoch": 1.77, "learning_rate": 3.447882059800664e-08, "loss": 0.2314, "step": 8527 }, { "epoch": 1.77, "learning_rate": 3.444767441860465e-08, "loss": 0.1337, "step": 8528 }, { "epoch": 1.77, "learning_rate": 3.441652823920266e-08, "loss": 0.15, "step": 8529 }, { "epoch": 1.77, "learning_rate": 3.438538205980066e-08, "loss": 0.1848, "step": 8530 }, { "epoch": 1.77, "learning_rate": 3.435423588039867e-08, "loss": 0.1551, "step": 8531 }, { "epoch": 1.77, "learning_rate": 3.432308970099667e-08, "loss": 0.1789, "step": 8532 }, { "epoch": 1.77, "learning_rate": 3.429194352159468e-08, "loss": 0.1972, "step": 8533 }, { "epoch": 1.77, "learning_rate": 3.426079734219269e-08, "loss": 0.1842, "step": 8534 }, { "epoch": 1.77, "learning_rate": 3.4229651162790694e-08, "loss": 0.1642, "step": 8535 }, { "epoch": 1.77, "learning_rate": 3.41985049833887e-08, "loss": 0.2057, "step": 8536 }, { "epoch": 1.77, "learning_rate": 3.416735880398671e-08, "loss": 0.2287, "step": 8537 }, { "epoch": 1.77, "learning_rate": 3.4136212624584715e-08, "loss": 0.1887, "step": 8538 }, { "epoch": 1.77, "learning_rate": 3.4105066445182724e-08, "loss": 0.204, "step": 8539 }, { "epoch": 1.77, "learning_rate": 3.4073920265780727e-08, "loss": 0.1696, "step": 8540 }, { "epoch": 1.77, "learning_rate": 3.4042774086378736e-08, "loss": 0.1818, "step": 8541 }, { "epoch": 1.77, "learning_rate": 3.401162790697674e-08, "loss": 0.1747, "step": 8542 }, { "epoch": 1.77, "learning_rate": 3.398048172757475e-08, "loss": 0.1974, "step": 8543 }, { "epoch": 1.77, "learning_rate": 3.394933554817276e-08, "loss": 0.2029, "step": 8544 }, { "epoch": 1.77, "learning_rate": 3.391818936877076e-08, "loss": 0.2245, "step": 8545 }, { "epoch": 1.77, "learning_rate": 3.388704318936877e-08, "loss": 0.254, "step": 8546 }, { "epoch": 1.77, "learning_rate": 3.385589700996678e-08, "loss": 0.1534, "step": 8547 }, { "epoch": 1.77, "learning_rate": 3.382475083056478e-08, "loss": 0.1674, "step": 8548 }, { "epoch": 1.78, "learning_rate": 3.379360465116279e-08, "loss": 0.1792, "step": 8549 }, { "epoch": 1.78, "learning_rate": 3.37624584717608e-08, "loss": 0.187, "step": 8550 }, { "epoch": 1.78, "learning_rate": 3.37313122923588e-08, "loss": 0.194, "step": 8551 }, { "epoch": 1.78, "learning_rate": 3.3700166112956804e-08, "loss": 0.1468, "step": 8552 }, { "epoch": 1.78, "learning_rate": 3.366901993355482e-08, "loss": 0.2202, "step": 8553 }, { "epoch": 1.78, "learning_rate": 3.363787375415282e-08, "loss": 0.1804, "step": 8554 }, { "epoch": 1.78, "learning_rate": 3.3606727574750825e-08, "loss": 0.1618, "step": 8555 }, { "epoch": 1.78, "learning_rate": 3.3575581395348834e-08, "loss": 0.2594, "step": 8556 }, { "epoch": 1.78, "learning_rate": 3.3544435215946843e-08, "loss": 0.1528, "step": 8557 }, { "epoch": 1.78, "learning_rate": 3.3513289036544846e-08, "loss": 0.1681, "step": 8558 }, { "epoch": 1.78, "learning_rate": 3.3482142857142855e-08, "loss": 0.2757, "step": 8559 }, { "epoch": 1.78, "learning_rate": 3.3450996677740864e-08, "loss": 0.1872, "step": 8560 }, { "epoch": 1.78, "learning_rate": 3.341985049833887e-08, "loss": 0.1422, "step": 8561 }, { "epoch": 1.78, "learning_rate": 3.3388704318936876e-08, "loss": 0.1163, "step": 8562 }, { "epoch": 1.78, "learning_rate": 3.3357558139534885e-08, "loss": 0.1869, "step": 8563 }, { "epoch": 1.78, "learning_rate": 3.332641196013289e-08, "loss": 0.2389, "step": 8564 }, { "epoch": 1.78, "learning_rate": 3.32952657807309e-08, "loss": 0.1417, "step": 8565 }, { "epoch": 1.78, "learning_rate": 3.3264119601328906e-08, "loss": 0.1343, "step": 8566 }, { "epoch": 1.78, "learning_rate": 3.323297342192691e-08, "loss": 0.1683, "step": 8567 }, { "epoch": 1.78, "learning_rate": 3.320182724252491e-08, "loss": 0.1952, "step": 8568 }, { "epoch": 1.78, "learning_rate": 3.317068106312292e-08, "loss": 0.1742, "step": 8569 }, { "epoch": 1.78, "learning_rate": 3.313953488372093e-08, "loss": 0.1813, "step": 8570 }, { "epoch": 1.78, "learning_rate": 3.310838870431893e-08, "loss": 0.1633, "step": 8571 }, { "epoch": 1.78, "learning_rate": 3.307724252491694e-08, "loss": 0.1647, "step": 8572 }, { "epoch": 1.78, "learning_rate": 3.304609634551495e-08, "loss": 0.1898, "step": 8573 }, { "epoch": 1.78, "learning_rate": 3.3014950166112954e-08, "loss": 0.1651, "step": 8574 }, { "epoch": 1.78, "learning_rate": 3.298380398671096e-08, "loss": 0.1327, "step": 8575 }, { "epoch": 1.78, "learning_rate": 3.295265780730897e-08, "loss": 0.1837, "step": 8576 }, { "epoch": 1.78, "learning_rate": 3.2921511627906975e-08, "loss": 0.1711, "step": 8577 }, { "epoch": 1.78, "learning_rate": 3.2890365448504984e-08, "loss": 0.2532, "step": 8578 }, { "epoch": 1.78, "learning_rate": 3.285921926910299e-08, "loss": 0.1606, "step": 8579 }, { "epoch": 1.78, "learning_rate": 3.2828073089700996e-08, "loss": 0.1599, "step": 8580 }, { "epoch": 1.78, "learning_rate": 3.2796926910299e-08, "loss": 0.1549, "step": 8581 }, { "epoch": 1.78, "learning_rate": 3.276578073089701e-08, "loss": 0.1468, "step": 8582 }, { "epoch": 1.78, "learning_rate": 3.2734634551495017e-08, "loss": 0.2724, "step": 8583 }, { "epoch": 1.78, "learning_rate": 3.270348837209302e-08, "loss": 0.2385, "step": 8584 }, { "epoch": 1.78, "learning_rate": 3.267234219269103e-08, "loss": 0.2063, "step": 8585 }, { "epoch": 1.78, "learning_rate": 3.264119601328904e-08, "loss": 0.1424, "step": 8586 }, { "epoch": 1.78, "learning_rate": 3.261004983388704e-08, "loss": 0.1332, "step": 8587 }, { "epoch": 1.78, "learning_rate": 3.257890365448505e-08, "loss": 0.1609, "step": 8588 }, { "epoch": 1.78, "learning_rate": 3.254775747508306e-08, "loss": 0.1418, "step": 8589 }, { "epoch": 1.78, "learning_rate": 3.251661129568106e-08, "loss": 0.2207, "step": 8590 }, { "epoch": 1.78, "learning_rate": 3.248546511627907e-08, "loss": 0.1875, "step": 8591 }, { "epoch": 1.78, "learning_rate": 3.245431893687707e-08, "loss": 0.2059, "step": 8592 }, { "epoch": 1.78, "learning_rate": 3.242317275747508e-08, "loss": 0.2109, "step": 8593 }, { "epoch": 1.78, "learning_rate": 3.239202657807309e-08, "loss": 0.2279, "step": 8594 }, { "epoch": 1.78, "learning_rate": 3.2360880398671094e-08, "loss": 0.1884, "step": 8595 }, { "epoch": 1.78, "learning_rate": 3.23297342192691e-08, "loss": 0.2126, "step": 8596 }, { "epoch": 1.79, "learning_rate": 3.2298588039867106e-08, "loss": 0.1388, "step": 8597 }, { "epoch": 1.79, "learning_rate": 3.2267441860465115e-08, "loss": 0.2337, "step": 8598 }, { "epoch": 1.79, "learning_rate": 3.2236295681063124e-08, "loss": 0.1158, "step": 8599 }, { "epoch": 1.79, "learning_rate": 3.220514950166113e-08, "loss": 0.1653, "step": 8600 }, { "epoch": 1.79, "learning_rate": 3.2174003322259136e-08, "loss": 0.2349, "step": 8601 }, { "epoch": 1.79, "learning_rate": 3.214285714285714e-08, "loss": 0.1482, "step": 8602 }, { "epoch": 1.79, "learning_rate": 3.211171096345515e-08, "loss": 0.18, "step": 8603 }, { "epoch": 1.79, "learning_rate": 3.208056478405316e-08, "loss": 0.1821, "step": 8604 }, { "epoch": 1.79, "learning_rate": 3.204941860465116e-08, "loss": 0.1522, "step": 8605 }, { "epoch": 1.79, "learning_rate": 3.201827242524917e-08, "loss": 0.2146, "step": 8606 }, { "epoch": 1.79, "learning_rate": 3.198712624584718e-08, "loss": 0.1443, "step": 8607 }, { "epoch": 1.79, "learning_rate": 3.195598006644518e-08, "loss": 0.1791, "step": 8608 }, { "epoch": 1.79, "learning_rate": 3.192483388704319e-08, "loss": 0.2183, "step": 8609 }, { "epoch": 1.79, "learning_rate": 3.189368770764119e-08, "loss": 0.1432, "step": 8610 }, { "epoch": 1.79, "learning_rate": 3.18625415282392e-08, "loss": 0.1858, "step": 8611 }, { "epoch": 1.79, "learning_rate": 3.1831395348837204e-08, "loss": 0.1267, "step": 8612 }, { "epoch": 1.79, "learning_rate": 3.1800249169435213e-08, "loss": 0.2266, "step": 8613 }, { "epoch": 1.79, "learning_rate": 3.176910299003322e-08, "loss": 0.1963, "step": 8614 }, { "epoch": 1.79, "learning_rate": 3.1737956810631225e-08, "loss": 0.1848, "step": 8615 }, { "epoch": 1.79, "learning_rate": 3.1706810631229234e-08, "loss": 0.153, "step": 8616 }, { "epoch": 1.79, "learning_rate": 3.1675664451827244e-08, "loss": 0.2437, "step": 8617 }, { "epoch": 1.79, "learning_rate": 3.1644518272425246e-08, "loss": 0.1365, "step": 8618 }, { "epoch": 1.79, "learning_rate": 3.1613372093023255e-08, "loss": 0.1557, "step": 8619 }, { "epoch": 1.79, "learning_rate": 3.1582225913621265e-08, "loss": 0.1912, "step": 8620 }, { "epoch": 1.79, "learning_rate": 3.155107973421927e-08, "loss": 0.1618, "step": 8621 }, { "epoch": 1.79, "learning_rate": 3.151993355481727e-08, "loss": 0.2176, "step": 8622 }, { "epoch": 1.79, "learning_rate": 3.148878737541528e-08, "loss": 0.1652, "step": 8623 }, { "epoch": 1.79, "learning_rate": 3.145764119601329e-08, "loss": 0.173, "step": 8624 }, { "epoch": 1.79, "learning_rate": 3.142649501661129e-08, "loss": 0.1658, "step": 8625 }, { "epoch": 1.79, "learning_rate": 3.13953488372093e-08, "loss": 0.1962, "step": 8626 }, { "epoch": 1.79, "learning_rate": 3.136420265780731e-08, "loss": 0.1617, "step": 8627 }, { "epoch": 1.79, "learning_rate": 3.133305647840531e-08, "loss": 0.2033, "step": 8628 }, { "epoch": 1.79, "learning_rate": 3.130191029900332e-08, "loss": 0.2016, "step": 8629 }, { "epoch": 1.79, "learning_rate": 3.127076411960133e-08, "loss": 0.2153, "step": 8630 }, { "epoch": 1.79, "learning_rate": 3.123961794019933e-08, "loss": 0.1691, "step": 8631 }, { "epoch": 1.79, "learning_rate": 3.120847176079734e-08, "loss": 0.1785, "step": 8632 }, { "epoch": 1.79, "learning_rate": 3.117732558139535e-08, "loss": 0.124, "step": 8633 }, { "epoch": 1.79, "learning_rate": 3.1146179401993354e-08, "loss": 0.199, "step": 8634 }, { "epoch": 1.79, "learning_rate": 3.1115033222591356e-08, "loss": 0.2115, "step": 8635 }, { "epoch": 1.79, "learning_rate": 3.108388704318937e-08, "loss": 0.2435, "step": 8636 }, { "epoch": 1.79, "learning_rate": 3.1052740863787375e-08, "loss": 0.1311, "step": 8637 }, { "epoch": 1.79, "learning_rate": 3.102159468438538e-08, "loss": 0.2002, "step": 8638 }, { "epoch": 1.79, "learning_rate": 3.0990448504983387e-08, "loss": 0.2267, "step": 8639 }, { "epoch": 1.79, "learning_rate": 3.0959302325581396e-08, "loss": 0.1768, "step": 8640 }, { "epoch": 1.79, "learning_rate": 3.09281561461794e-08, "loss": 0.1855, "step": 8641 }, { "epoch": 1.79, "learning_rate": 3.089700996677741e-08, "loss": 0.1716, "step": 8642 }, { "epoch": 1.79, "learning_rate": 3.086586378737542e-08, "loss": 0.1604, "step": 8643 }, { "epoch": 1.79, "learning_rate": 3.083471760797342e-08, "loss": 0.1373, "step": 8644 }, { "epoch": 1.8, "learning_rate": 3.080357142857143e-08, "loss": 0.181, "step": 8645 }, { "epoch": 1.8, "learning_rate": 3.077242524916944e-08, "loss": 0.1982, "step": 8646 }, { "epoch": 1.8, "learning_rate": 3.074127906976744e-08, "loss": 0.1606, "step": 8647 }, { "epoch": 1.8, "learning_rate": 3.071013289036545e-08, "loss": 0.1872, "step": 8648 }, { "epoch": 1.8, "learning_rate": 3.067898671096345e-08, "loss": 0.1541, "step": 8649 }, { "epoch": 1.8, "learning_rate": 3.064784053156146e-08, "loss": 0.2058, "step": 8650 }, { "epoch": 1.8, "learning_rate": 3.0616694352159464e-08, "loss": 0.2095, "step": 8651 }, { "epoch": 1.8, "learning_rate": 3.058554817275747e-08, "loss": 0.1324, "step": 8652 }, { "epoch": 1.8, "learning_rate": 3.055440199335548e-08, "loss": 0.23, "step": 8653 }, { "epoch": 1.8, "learning_rate": 3.0523255813953485e-08, "loss": 0.1933, "step": 8654 }, { "epoch": 1.8, "learning_rate": 3.0492109634551494e-08, "loss": 0.1397, "step": 8655 }, { "epoch": 1.8, "learning_rate": 3.0460963455149503e-08, "loss": 0.1655, "step": 8656 }, { "epoch": 1.8, "learning_rate": 3.0429817275747506e-08, "loss": 0.1739, "step": 8657 }, { "epoch": 1.8, "learning_rate": 3.0398671096345515e-08, "loss": 0.1953, "step": 8658 }, { "epoch": 1.8, "learning_rate": 3.036752491694352e-08, "loss": 0.1487, "step": 8659 }, { "epoch": 1.8, "learning_rate": 3.033637873754153e-08, "loss": 0.1754, "step": 8660 }, { "epoch": 1.8, "learning_rate": 3.0305232558139536e-08, "loss": 0.2066, "step": 8661 }, { "epoch": 1.8, "learning_rate": 3.027408637873754e-08, "loss": 0.1886, "step": 8662 }, { "epoch": 1.8, "learning_rate": 3.024294019933555e-08, "loss": 0.1642, "step": 8663 }, { "epoch": 1.8, "learning_rate": 3.021179401993355e-08, "loss": 0.255, "step": 8664 }, { "epoch": 1.8, "learning_rate": 3.018064784053156e-08, "loss": 0.1616, "step": 8665 }, { "epoch": 1.8, "learning_rate": 3.014950166112957e-08, "loss": 0.2411, "step": 8666 }, { "epoch": 1.8, "learning_rate": 3.011835548172757e-08, "loss": 0.2167, "step": 8667 }, { "epoch": 1.8, "learning_rate": 3.008720930232558e-08, "loss": 0.2074, "step": 8668 }, { "epoch": 1.8, "learning_rate": 3.0056063122923583e-08, "loss": 0.1251, "step": 8669 }, { "epoch": 1.8, "learning_rate": 3.002491694352159e-08, "loss": 0.1883, "step": 8670 }, { "epoch": 1.8, "learning_rate": 2.99937707641196e-08, "loss": 0.1693, "step": 8671 }, { "epoch": 1.8, "learning_rate": 2.9962624584717604e-08, "loss": 0.2196, "step": 8672 }, { "epoch": 1.8, "learning_rate": 2.9931478405315614e-08, "loss": 0.2012, "step": 8673 }, { "epoch": 1.8, "learning_rate": 2.990033222591362e-08, "loss": 0.179, "step": 8674 }, { "epoch": 1.8, "learning_rate": 2.9869186046511625e-08, "loss": 0.1611, "step": 8675 }, { "epoch": 1.8, "learning_rate": 2.9838039867109635e-08, "loss": 0.2145, "step": 8676 }, { "epoch": 1.8, "learning_rate": 2.980689368770764e-08, "loss": 0.1387, "step": 8677 }, { "epoch": 1.8, "learning_rate": 2.9775747508305646e-08, "loss": 0.1493, "step": 8678 }, { "epoch": 1.8, "learning_rate": 2.9744601328903652e-08, "loss": 0.1569, "step": 8679 }, { "epoch": 1.8, "learning_rate": 2.971345514950166e-08, "loss": 0.1592, "step": 8680 }, { "epoch": 1.8, "learning_rate": 2.9682308970099667e-08, "loss": 0.217, "step": 8681 }, { "epoch": 1.8, "learning_rate": 2.965116279069767e-08, "loss": 0.1868, "step": 8682 }, { "epoch": 1.8, "learning_rate": 2.9620016611295683e-08, "loss": 0.162, "step": 8683 }, { "epoch": 1.8, "learning_rate": 2.9588870431893685e-08, "loss": 0.1552, "step": 8684 }, { "epoch": 1.8, "learning_rate": 2.955772425249169e-08, "loss": 0.2055, "step": 8685 }, { "epoch": 1.8, "learning_rate": 2.95265780730897e-08, "loss": 0.1633, "step": 8686 }, { "epoch": 1.8, "learning_rate": 2.9495431893687706e-08, "loss": 0.1408, "step": 8687 }, { "epoch": 1.8, "learning_rate": 2.9464285714285712e-08, "loss": 0.1923, "step": 8688 }, { "epoch": 1.8, "learning_rate": 2.9433139534883718e-08, "loss": 0.1891, "step": 8689 }, { "epoch": 1.8, "learning_rate": 2.9401993355481727e-08, "loss": 0.1573, "step": 8690 }, { "epoch": 1.8, "learning_rate": 2.9401993355481727e-08, "loss": 0.1653, "step": 8691 }, { "epoch": 1.8, "learning_rate": 2.9370847176079733e-08, "loss": 0.1976, "step": 8692 }, { "epoch": 1.81, "learning_rate": 2.933970099667774e-08, "loss": 0.1574, "step": 8693 }, { "epoch": 1.81, "learning_rate": 2.9308554817275748e-08, "loss": 0.1659, "step": 8694 }, { "epoch": 1.81, "learning_rate": 2.9277408637873754e-08, "loss": 0.1804, "step": 8695 }, { "epoch": 1.81, "learning_rate": 2.924626245847176e-08, "loss": 0.1416, "step": 8696 }, { "epoch": 1.81, "learning_rate": 2.921511627906977e-08, "loss": 0.1396, "step": 8697 }, { "epoch": 1.81, "learning_rate": 2.9183970099667775e-08, "loss": 0.191, "step": 8698 }, { "epoch": 1.81, "learning_rate": 2.9152823920265778e-08, "loss": 0.1461, "step": 8699 }, { "epoch": 1.81, "learning_rate": 2.9121677740863784e-08, "loss": 0.2239, "step": 8700 }, { "epoch": 1.81, "learning_rate": 2.9090531561461793e-08, "loss": 0.1909, "step": 8701 }, { "epoch": 1.81, "learning_rate": 2.90593853820598e-08, "loss": 0.1481, "step": 8702 }, { "epoch": 1.81, "learning_rate": 2.9028239202657805e-08, "loss": 0.2161, "step": 8703 }, { "epoch": 1.81, "learning_rate": 2.8997093023255814e-08, "loss": 0.1603, "step": 8704 }, { "epoch": 1.81, "learning_rate": 2.896594684385382e-08, "loss": 0.2117, "step": 8705 }, { "epoch": 1.81, "learning_rate": 2.8934800664451826e-08, "loss": 0.2215, "step": 8706 }, { "epoch": 1.81, "learning_rate": 2.8903654485049835e-08, "loss": 0.2125, "step": 8707 }, { "epoch": 1.81, "learning_rate": 2.887250830564784e-08, "loss": 0.2008, "step": 8708 }, { "epoch": 1.81, "learning_rate": 2.8841362126245847e-08, "loss": 0.187, "step": 8709 }, { "epoch": 1.81, "learning_rate": 2.881021594684385e-08, "loss": 0.1392, "step": 8710 }, { "epoch": 1.81, "learning_rate": 2.877906976744186e-08, "loss": 0.1464, "step": 8711 }, { "epoch": 1.81, "learning_rate": 2.8747923588039864e-08, "loss": 0.1692, "step": 8712 }, { "epoch": 1.81, "learning_rate": 2.871677740863787e-08, "loss": 0.1431, "step": 8713 }, { "epoch": 1.81, "learning_rate": 2.868563122923588e-08, "loss": 0.1765, "step": 8714 }, { "epoch": 1.81, "learning_rate": 2.8654485049833885e-08, "loss": 0.1633, "step": 8715 }, { "epoch": 1.81, "learning_rate": 2.862333887043189e-08, "loss": 0.2431, "step": 8716 }, { "epoch": 1.81, "learning_rate": 2.85921926910299e-08, "loss": 0.1932, "step": 8717 }, { "epoch": 1.81, "learning_rate": 2.8561046511627906e-08, "loss": 0.151, "step": 8718 }, { "epoch": 1.81, "learning_rate": 2.8529900332225912e-08, "loss": 0.2011, "step": 8719 }, { "epoch": 1.81, "learning_rate": 2.8498754152823918e-08, "loss": 0.1613, "step": 8720 }, { "epoch": 1.81, "learning_rate": 2.8467607973421927e-08, "loss": 0.1607, "step": 8721 }, { "epoch": 1.81, "learning_rate": 2.8436461794019933e-08, "loss": 0.1599, "step": 8722 }, { "epoch": 1.81, "learning_rate": 2.840531561461794e-08, "loss": 0.1914, "step": 8723 }, { "epoch": 1.81, "learning_rate": 2.8374169435215948e-08, "loss": 0.2284, "step": 8724 }, { "epoch": 1.81, "learning_rate": 2.8343023255813954e-08, "loss": 0.2191, "step": 8725 }, { "epoch": 1.81, "learning_rate": 2.8311877076411957e-08, "loss": 0.2604, "step": 8726 }, { "epoch": 1.81, "learning_rate": 2.8280730897009966e-08, "loss": 0.158, "step": 8727 }, { "epoch": 1.81, "learning_rate": 2.8249584717607972e-08, "loss": 0.1969, "step": 8728 }, { "epoch": 1.81, "learning_rate": 2.8218438538205978e-08, "loss": 0.1959, "step": 8729 }, { "epoch": 1.81, "learning_rate": 2.8187292358803984e-08, "loss": 0.1952, "step": 8730 }, { "epoch": 1.81, "learning_rate": 2.8156146179401993e-08, "loss": 0.1736, "step": 8731 }, { "epoch": 1.81, "learning_rate": 2.8125e-08, "loss": 0.2011, "step": 8732 }, { "epoch": 1.81, "learning_rate": 2.8093853820598005e-08, "loss": 0.1356, "step": 8733 }, { "epoch": 1.81, "learning_rate": 2.8062707641196014e-08, "loss": 0.216, "step": 8734 }, { "epoch": 1.81, "learning_rate": 2.803156146179402e-08, "loss": 0.1892, "step": 8735 }, { "epoch": 1.81, "learning_rate": 2.8000415282392026e-08, "loss": 0.1674, "step": 8736 }, { "epoch": 1.81, "learning_rate": 2.796926910299003e-08, "loss": 0.1665, "step": 8737 }, { "epoch": 1.81, "learning_rate": 2.793812292358804e-08, "loss": 0.1509, "step": 8738 }, { "epoch": 1.81, "learning_rate": 2.7906976744186043e-08, "loss": 0.1267, "step": 8739 }, { "epoch": 1.81, "learning_rate": 2.787583056478405e-08, "loss": 0.1904, "step": 8740 }, { "epoch": 1.81, "learning_rate": 2.784468438538206e-08, "loss": 0.2065, "step": 8741 }, { "epoch": 1.82, "learning_rate": 2.7813538205980064e-08, "loss": 0.1901, "step": 8742 }, { "epoch": 1.82, "learning_rate": 2.778239202657807e-08, "loss": 0.1574, "step": 8743 }, { "epoch": 1.82, "learning_rate": 2.775124584717608e-08, "loss": 0.1829, "step": 8744 }, { "epoch": 1.82, "learning_rate": 2.7720099667774085e-08, "loss": 0.147, "step": 8745 }, { "epoch": 1.82, "learning_rate": 2.768895348837209e-08, "loss": 0.1591, "step": 8746 }, { "epoch": 1.82, "learning_rate": 2.7657807308970097e-08, "loss": 0.1428, "step": 8747 }, { "epoch": 1.82, "learning_rate": 2.7626661129568106e-08, "loss": 0.2063, "step": 8748 }, { "epoch": 1.82, "learning_rate": 2.7595514950166112e-08, "loss": 0.1367, "step": 8749 }, { "epoch": 1.82, "learning_rate": 2.7564368770764118e-08, "loss": 0.1928, "step": 8750 }, { "epoch": 1.82, "learning_rate": 2.7533222591362127e-08, "loss": 0.1673, "step": 8751 }, { "epoch": 1.82, "learning_rate": 2.7502076411960133e-08, "loss": 0.1365, "step": 8752 }, { "epoch": 1.82, "learning_rate": 2.7470930232558136e-08, "loss": 0.2608, "step": 8753 }, { "epoch": 1.82, "learning_rate": 2.7439784053156148e-08, "loss": 0.1326, "step": 8754 }, { "epoch": 1.82, "learning_rate": 2.740863787375415e-08, "loss": 0.1678, "step": 8755 }, { "epoch": 1.82, "learning_rate": 2.7377491694352157e-08, "loss": 0.1907, "step": 8756 }, { "epoch": 1.82, "learning_rate": 2.7346345514950163e-08, "loss": 0.1355, "step": 8757 }, { "epoch": 1.82, "learning_rate": 2.7315199335548172e-08, "loss": 0.2214, "step": 8758 }, { "epoch": 1.82, "learning_rate": 2.7284053156146178e-08, "loss": 0.1417, "step": 8759 }, { "epoch": 1.82, "learning_rate": 2.7252906976744184e-08, "loss": 0.1532, "step": 8760 }, { "epoch": 1.82, "learning_rate": 2.7221760797342193e-08, "loss": 0.2041, "step": 8761 }, { "epoch": 1.82, "learning_rate": 2.71906146179402e-08, "loss": 0.2139, "step": 8762 }, { "epoch": 1.82, "learning_rate": 2.7159468438538205e-08, "loss": 0.1703, "step": 8763 }, { "epoch": 1.82, "learning_rate": 2.7128322259136214e-08, "loss": 0.1471, "step": 8764 }, { "epoch": 1.82, "learning_rate": 2.709717607973422e-08, "loss": 0.1696, "step": 8765 }, { "epoch": 1.82, "learning_rate": 2.7066029900332222e-08, "loss": 0.2893, "step": 8766 }, { "epoch": 1.82, "learning_rate": 2.7034883720930228e-08, "loss": 0.1702, "step": 8767 }, { "epoch": 1.82, "learning_rate": 2.7003737541528238e-08, "loss": 0.179, "step": 8768 }, { "epoch": 1.82, "learning_rate": 2.6972591362126243e-08, "loss": 0.1732, "step": 8769 }, { "epoch": 1.82, "learning_rate": 2.694144518272425e-08, "loss": 0.1702, "step": 8770 }, { "epoch": 1.82, "learning_rate": 2.691029900332226e-08, "loss": 0.1935, "step": 8771 }, { "epoch": 1.82, "learning_rate": 2.6879152823920264e-08, "loss": 0.1927, "step": 8772 }, { "epoch": 1.82, "learning_rate": 2.684800664451827e-08, "loss": 0.1419, "step": 8773 }, { "epoch": 1.82, "learning_rate": 2.681686046511628e-08, "loss": 0.1707, "step": 8774 }, { "epoch": 1.82, "learning_rate": 2.6785714285714285e-08, "loss": 0.1773, "step": 8775 }, { "epoch": 1.82, "learning_rate": 2.675456810631229e-08, "loss": 0.1684, "step": 8776 }, { "epoch": 1.82, "learning_rate": 2.6723421926910297e-08, "loss": 0.19, "step": 8777 }, { "epoch": 1.82, "learning_rate": 2.6692275747508306e-08, "loss": 0.1744, "step": 8778 }, { "epoch": 1.82, "learning_rate": 2.6661129568106312e-08, "loss": 0.1505, "step": 8779 }, { "epoch": 1.82, "learning_rate": 2.6629983388704315e-08, "loss": 0.1957, "step": 8780 }, { "epoch": 1.82, "learning_rate": 2.6598837209302327e-08, "loss": 0.162, "step": 8781 }, { "epoch": 1.82, "learning_rate": 2.656769102990033e-08, "loss": 0.2066, "step": 8782 }, { "epoch": 1.82, "learning_rate": 2.6536544850498336e-08, "loss": 0.1947, "step": 8783 }, { "epoch": 1.82, "learning_rate": 2.6505398671096345e-08, "loss": 0.2054, "step": 8784 }, { "epoch": 1.82, "learning_rate": 2.647425249169435e-08, "loss": 0.1304, "step": 8785 }, { "epoch": 1.82, "learning_rate": 2.6443106312292357e-08, "loss": 0.1632, "step": 8786 }, { "epoch": 1.82, "learning_rate": 2.6411960132890363e-08, "loss": 0.258, "step": 8787 }, { "epoch": 1.82, "learning_rate": 2.6380813953488372e-08, "loss": 0.2026, "step": 8788 }, { "epoch": 1.82, "learning_rate": 2.6349667774086378e-08, "loss": 0.1438, "step": 8789 }, { "epoch": 1.83, "learning_rate": 2.6318521594684384e-08, "loss": 0.159, "step": 8790 }, { "epoch": 1.83, "learning_rate": 2.6287375415282393e-08, "loss": 0.1798, "step": 8791 }, { "epoch": 1.83, "learning_rate": 2.62562292358804e-08, "loss": 0.1575, "step": 8792 }, { "epoch": 1.83, "learning_rate": 2.62250830564784e-08, "loss": 0.1908, "step": 8793 }, { "epoch": 1.83, "learning_rate": 2.6193936877076414e-08, "loss": 0.18, "step": 8794 }, { "epoch": 1.83, "learning_rate": 2.6162790697674417e-08, "loss": 0.171, "step": 8795 }, { "epoch": 1.83, "learning_rate": 2.6131644518272423e-08, "loss": 0.1727, "step": 8796 }, { "epoch": 1.83, "learning_rate": 2.610049833887043e-08, "loss": 0.1841, "step": 8797 }, { "epoch": 1.83, "learning_rate": 2.6069352159468438e-08, "loss": 0.163, "step": 8798 }, { "epoch": 1.83, "learning_rate": 2.6038205980066444e-08, "loss": 0.1938, "step": 8799 }, { "epoch": 1.83, "learning_rate": 2.600705980066445e-08, "loss": 0.2075, "step": 8800 }, { "epoch": 1.83, "learning_rate": 2.597591362126246e-08, "loss": 0.2318, "step": 8801 }, { "epoch": 1.83, "learning_rate": 2.5944767441860465e-08, "loss": 0.2043, "step": 8802 }, { "epoch": 1.83, "learning_rate": 2.591362126245847e-08, "loss": 0.1105, "step": 8803 }, { "epoch": 1.83, "learning_rate": 2.588247508305648e-08, "loss": 0.1674, "step": 8804 }, { "epoch": 1.83, "learning_rate": 2.5851328903654486e-08, "loss": 0.1603, "step": 8805 }, { "epoch": 1.83, "learning_rate": 2.582018272425249e-08, "loss": 0.2435, "step": 8806 }, { "epoch": 1.83, "learning_rate": 2.5789036544850494e-08, "loss": 0.1341, "step": 8807 }, { "epoch": 1.83, "learning_rate": 2.5757890365448507e-08, "loss": 0.1751, "step": 8808 }, { "epoch": 1.83, "learning_rate": 2.572674418604651e-08, "loss": 0.1761, "step": 8809 }, { "epoch": 1.83, "learning_rate": 2.5695598006644515e-08, "loss": 0.1747, "step": 8810 }, { "epoch": 1.83, "learning_rate": 2.5664451827242524e-08, "loss": 0.1607, "step": 8811 }, { "epoch": 1.83, "learning_rate": 2.563330564784053e-08, "loss": 0.1684, "step": 8812 }, { "epoch": 1.83, "learning_rate": 2.5602159468438536e-08, "loss": 0.1708, "step": 8813 }, { "epoch": 1.83, "learning_rate": 2.5571013289036545e-08, "loss": 0.1582, "step": 8814 }, { "epoch": 1.83, "learning_rate": 2.553986710963455e-08, "loss": 0.1923, "step": 8815 }, { "epoch": 1.83, "learning_rate": 2.5508720930232557e-08, "loss": 0.2821, "step": 8816 }, { "epoch": 1.83, "learning_rate": 2.5477574750830563e-08, "loss": 0.1981, "step": 8817 }, { "epoch": 1.83, "learning_rate": 2.5446428571428572e-08, "loss": 0.1974, "step": 8818 }, { "epoch": 1.83, "learning_rate": 2.5415282392026578e-08, "loss": 0.1888, "step": 8819 }, { "epoch": 1.83, "learning_rate": 2.538413621262458e-08, "loss": 0.1427, "step": 8820 }, { "epoch": 1.83, "learning_rate": 2.5352990033222593e-08, "loss": 0.15, "step": 8821 }, { "epoch": 1.83, "learning_rate": 2.5321843853820596e-08, "loss": 0.2482, "step": 8822 }, { "epoch": 1.83, "learning_rate": 2.5290697674418602e-08, "loss": 0.1435, "step": 8823 }, { "epoch": 1.83, "learning_rate": 2.525955149501661e-08, "loss": 0.1891, "step": 8824 }, { "epoch": 1.83, "learning_rate": 2.5228405315614617e-08, "loss": 0.1873, "step": 8825 }, { "epoch": 1.83, "learning_rate": 2.5197259136212623e-08, "loss": 0.227, "step": 8826 }, { "epoch": 1.83, "learning_rate": 2.516611295681063e-08, "loss": 0.1564, "step": 8827 }, { "epoch": 1.83, "learning_rate": 2.5134966777408638e-08, "loss": 0.16, "step": 8828 }, { "epoch": 1.83, "learning_rate": 2.5103820598006644e-08, "loss": 0.1737, "step": 8829 }, { "epoch": 1.83, "learning_rate": 2.507267441860465e-08, "loss": 0.1912, "step": 8830 }, { "epoch": 1.83, "learning_rate": 2.504152823920266e-08, "loss": 0.1483, "step": 8831 }, { "epoch": 1.83, "learning_rate": 2.5010382059800665e-08, "loss": 0.1492, "step": 8832 }, { "epoch": 1.83, "learning_rate": 2.497923588039867e-08, "loss": 0.1441, "step": 8833 }, { "epoch": 1.83, "learning_rate": 2.4948089700996673e-08, "loss": 0.1936, "step": 8834 }, { "epoch": 1.83, "learning_rate": 2.4916943521594686e-08, "loss": 0.1641, "step": 8835 }, { "epoch": 1.83, "learning_rate": 2.4885797342192688e-08, "loss": 0.1827, "step": 8836 }, { "epoch": 1.83, "learning_rate": 2.4854651162790694e-08, "loss": 0.1669, "step": 8837 }, { "epoch": 1.84, "learning_rate": 2.4823504983388703e-08, "loss": 0.2462, "step": 8838 }, { "epoch": 1.84, "learning_rate": 2.479235880398671e-08, "loss": 0.1823, "step": 8839 }, { "epoch": 1.84, "learning_rate": 2.4761212624584715e-08, "loss": 0.1842, "step": 8840 }, { "epoch": 1.84, "learning_rate": 2.4730066445182724e-08, "loss": 0.1867, "step": 8841 }, { "epoch": 1.84, "learning_rate": 2.469892026578073e-08, "loss": 0.2101, "step": 8842 }, { "epoch": 1.84, "learning_rate": 2.4667774086378736e-08, "loss": 0.1966, "step": 8843 }, { "epoch": 1.84, "learning_rate": 2.4636627906976742e-08, "loss": 0.1996, "step": 8844 }, { "epoch": 1.84, "learning_rate": 2.460548172757475e-08, "loss": 0.1129, "step": 8845 }, { "epoch": 1.84, "learning_rate": 2.4574335548172757e-08, "loss": 0.2202, "step": 8846 }, { "epoch": 1.84, "learning_rate": 2.454318936877076e-08, "loss": 0.1982, "step": 8847 }, { "epoch": 1.84, "learning_rate": 2.4512043189368772e-08, "loss": 0.2175, "step": 8848 }, { "epoch": 1.84, "learning_rate": 2.4480897009966775e-08, "loss": 0.1239, "step": 8849 }, { "epoch": 1.84, "learning_rate": 2.444975083056478e-08, "loss": 0.1918, "step": 8850 }, { "epoch": 1.84, "learning_rate": 2.441860465116279e-08, "loss": 0.1517, "step": 8851 }, { "epoch": 1.84, "learning_rate": 2.4387458471760796e-08, "loss": 0.158, "step": 8852 }, { "epoch": 1.84, "learning_rate": 2.4356312292358802e-08, "loss": 0.1686, "step": 8853 }, { "epoch": 1.84, "learning_rate": 2.4325166112956808e-08, "loss": 0.1495, "step": 8854 }, { "epoch": 1.84, "learning_rate": 2.4294019933554817e-08, "loss": 0.1695, "step": 8855 }, { "epoch": 1.84, "learning_rate": 2.4262873754152823e-08, "loss": 0.1922, "step": 8856 }, { "epoch": 1.84, "learning_rate": 2.423172757475083e-08, "loss": 0.185, "step": 8857 }, { "epoch": 1.84, "learning_rate": 2.4200581395348838e-08, "loss": 0.1911, "step": 8858 }, { "epoch": 1.84, "learning_rate": 2.4169435215946844e-08, "loss": 0.1309, "step": 8859 }, { "epoch": 1.84, "learning_rate": 2.413828903654485e-08, "loss": 0.1503, "step": 8860 }, { "epoch": 1.84, "learning_rate": 2.410714285714286e-08, "loss": 0.1886, "step": 8861 }, { "epoch": 1.84, "learning_rate": 2.4075996677740865e-08, "loss": 0.1479, "step": 8862 }, { "epoch": 1.84, "learning_rate": 2.4044850498338867e-08, "loss": 0.5918, "step": 8863 }, { "epoch": 1.84, "learning_rate": 2.4013704318936873e-08, "loss": 0.1531, "step": 8864 }, { "epoch": 1.84, "learning_rate": 2.3982558139534882e-08, "loss": 0.1722, "step": 8865 }, { "epoch": 1.84, "learning_rate": 2.395141196013289e-08, "loss": 0.2238, "step": 8866 }, { "epoch": 1.84, "learning_rate": 2.3920265780730894e-08, "loss": 0.1539, "step": 8867 }, { "epoch": 1.84, "learning_rate": 2.3889119601328903e-08, "loss": 0.2008, "step": 8868 }, { "epoch": 1.84, "learning_rate": 2.385797342192691e-08, "loss": 0.1273, "step": 8869 }, { "epoch": 1.84, "learning_rate": 2.3826827242524915e-08, "loss": 0.1728, "step": 8870 }, { "epoch": 1.84, "learning_rate": 2.3795681063122924e-08, "loss": 0.159, "step": 8871 }, { "epoch": 1.84, "learning_rate": 2.376453488372093e-08, "loss": 0.202, "step": 8872 }, { "epoch": 1.84, "learning_rate": 2.3733388704318936e-08, "loss": 0.2209, "step": 8873 }, { "epoch": 1.84, "learning_rate": 2.3702242524916942e-08, "loss": 0.221, "step": 8874 }, { "epoch": 1.84, "learning_rate": 2.367109634551495e-08, "loss": 0.1561, "step": 8875 }, { "epoch": 1.84, "learning_rate": 2.3639950166112954e-08, "loss": 0.1893, "step": 8876 }, { "epoch": 1.84, "learning_rate": 2.360880398671096e-08, "loss": 0.1756, "step": 8877 }, { "epoch": 1.84, "learning_rate": 2.357765780730897e-08, "loss": 0.1576, "step": 8878 }, { "epoch": 1.84, "learning_rate": 2.3546511627906975e-08, "loss": 0.1422, "step": 8879 }, { "epoch": 1.84, "learning_rate": 2.351536544850498e-08, "loss": 0.1403, "step": 8880 }, { "epoch": 1.84, "learning_rate": 2.348421926910299e-08, "loss": 0.1602, "step": 8881 }, { "epoch": 1.84, "learning_rate": 2.3453073089700996e-08, "loss": 0.2278, "step": 8882 }, { "epoch": 1.84, "learning_rate": 2.3421926910299002e-08, "loss": 0.129, "step": 8883 }, { "epoch": 1.84, "learning_rate": 2.3390780730897008e-08, "loss": 0.1535, "step": 8884 }, { "epoch": 1.84, "learning_rate": 2.3359634551495017e-08, "loss": 0.1511, "step": 8885 }, { "epoch": 1.85, "learning_rate": 2.3328488372093023e-08, "loss": 0.156, "step": 8886 }, { "epoch": 1.85, "learning_rate": 2.329734219269103e-08, "loss": 0.1743, "step": 8887 }, { "epoch": 1.85, "learning_rate": 2.3266196013289038e-08, "loss": 0.2054, "step": 8888 }, { "epoch": 1.85, "learning_rate": 2.3235049833887044e-08, "loss": 0.2293, "step": 8889 }, { "epoch": 1.85, "learning_rate": 2.3203903654485046e-08, "loss": 0.1314, "step": 8890 }, { "epoch": 1.85, "learning_rate": 2.3172757475083056e-08, "loss": 0.1759, "step": 8891 }, { "epoch": 1.85, "learning_rate": 2.314161129568106e-08, "loss": 0.147, "step": 8892 }, { "epoch": 1.85, "learning_rate": 2.3110465116279067e-08, "loss": 0.195, "step": 8893 }, { "epoch": 1.85, "learning_rate": 2.3079318936877073e-08, "loss": 0.1216, "step": 8894 }, { "epoch": 1.85, "learning_rate": 2.3048172757475083e-08, "loss": 0.1901, "step": 8895 }, { "epoch": 1.85, "learning_rate": 2.301702657807309e-08, "loss": 0.1844, "step": 8896 }, { "epoch": 1.85, "learning_rate": 2.2985880398671094e-08, "loss": 0.2287, "step": 8897 }, { "epoch": 1.85, "learning_rate": 2.2954734219269104e-08, "loss": 0.1543, "step": 8898 }, { "epoch": 1.85, "learning_rate": 2.292358803986711e-08, "loss": 0.1974, "step": 8899 }, { "epoch": 1.85, "learning_rate": 2.2892441860465115e-08, "loss": 0.1701, "step": 8900 }, { "epoch": 1.85, "learning_rate": 2.2861295681063125e-08, "loss": 0.2025, "step": 8901 }, { "epoch": 1.85, "learning_rate": 2.283014950166113e-08, "loss": 0.1834, "step": 8902 }, { "epoch": 1.85, "learning_rate": 2.2799003322259133e-08, "loss": 0.1759, "step": 8903 }, { "epoch": 1.85, "learning_rate": 2.276785714285714e-08, "loss": 0.2144, "step": 8904 }, { "epoch": 1.85, "learning_rate": 2.2736710963455148e-08, "loss": 0.1858, "step": 8905 }, { "epoch": 1.85, "learning_rate": 2.2705564784053154e-08, "loss": 0.208, "step": 8906 }, { "epoch": 1.85, "learning_rate": 2.267441860465116e-08, "loss": 0.1935, "step": 8907 }, { "epoch": 1.85, "learning_rate": 2.264327242524917e-08, "loss": 0.2064, "step": 8908 }, { "epoch": 1.85, "learning_rate": 2.2612126245847175e-08, "loss": 0.2071, "step": 8909 }, { "epoch": 1.85, "learning_rate": 2.258098006644518e-08, "loss": 0.1567, "step": 8910 }, { "epoch": 1.85, "learning_rate": 2.254983388704319e-08, "loss": 0.1714, "step": 8911 }, { "epoch": 1.85, "learning_rate": 2.2518687707641196e-08, "loss": 0.1464, "step": 8912 }, { "epoch": 1.85, "learning_rate": 2.2487541528239202e-08, "loss": 0.1766, "step": 8913 }, { "epoch": 1.85, "learning_rate": 2.2456395348837208e-08, "loss": 0.1713, "step": 8914 }, { "epoch": 1.85, "learning_rate": 2.2425249169435217e-08, "loss": 0.1985, "step": 8915 }, { "epoch": 1.85, "learning_rate": 2.2394102990033223e-08, "loss": 0.1983, "step": 8916 }, { "epoch": 1.85, "learning_rate": 2.2362956810631226e-08, "loss": 0.1794, "step": 8917 }, { "epoch": 1.85, "learning_rate": 2.2331810631229238e-08, "loss": 0.187, "step": 8918 }, { "epoch": 1.85, "learning_rate": 2.230066445182724e-08, "loss": 0.1738, "step": 8919 }, { "epoch": 1.85, "learning_rate": 2.2269518272425247e-08, "loss": 0.1788, "step": 8920 }, { "epoch": 1.85, "learning_rate": 2.2238372093023256e-08, "loss": 0.1475, "step": 8921 }, { "epoch": 1.85, "learning_rate": 2.2207225913621262e-08, "loss": 0.1967, "step": 8922 }, { "epoch": 1.85, "learning_rate": 2.2176079734219268e-08, "loss": 0.1644, "step": 8923 }, { "epoch": 1.85, "learning_rate": 2.2144933554817274e-08, "loss": 0.1836, "step": 8924 }, { "epoch": 1.85, "learning_rate": 2.2113787375415283e-08, "loss": 0.238, "step": 8925 }, { "epoch": 1.85, "learning_rate": 2.208264119601329e-08, "loss": 0.2027, "step": 8926 }, { "epoch": 1.85, "learning_rate": 2.2051495016611295e-08, "loss": 0.1478, "step": 8927 }, { "epoch": 1.85, "learning_rate": 2.2020348837209304e-08, "loss": 0.1577, "step": 8928 }, { "epoch": 1.85, "learning_rate": 2.198920265780731e-08, "loss": 0.1753, "step": 8929 }, { "epoch": 1.85, "learning_rate": 2.1958056478405312e-08, "loss": 0.1427, "step": 8930 }, { "epoch": 1.85, "learning_rate": 2.1926910299003325e-08, "loss": 0.1656, "step": 8931 }, { "epoch": 1.85, "learning_rate": 2.1895764119601327e-08, "loss": 0.1487, "step": 8932 }, { "epoch": 1.85, "learning_rate": 2.1864617940199333e-08, "loss": 0.1747, "step": 8933 }, { "epoch": 1.86, "learning_rate": 2.183347176079734e-08, "loss": 0.1806, "step": 8934 }, { "epoch": 1.86, "learning_rate": 2.1802325581395348e-08, "loss": 0.1869, "step": 8935 }, { "epoch": 1.86, "learning_rate": 2.1771179401993354e-08, "loss": 0.1477, "step": 8936 }, { "epoch": 1.86, "learning_rate": 2.174003322259136e-08, "loss": 0.2243, "step": 8937 }, { "epoch": 1.86, "learning_rate": 2.170888704318937e-08, "loss": 0.148, "step": 8938 }, { "epoch": 1.86, "learning_rate": 2.1677740863787375e-08, "loss": 0.1622, "step": 8939 }, { "epoch": 1.86, "learning_rate": 2.164659468438538e-08, "loss": 0.1646, "step": 8940 }, { "epoch": 1.86, "learning_rate": 2.1615448504983387e-08, "loss": 0.1523, "step": 8941 }, { "epoch": 1.86, "learning_rate": 2.1584302325581396e-08, "loss": 0.179, "step": 8942 }, { "epoch": 1.86, "learning_rate": 2.1553156146179402e-08, "loss": 0.1598, "step": 8943 }, { "epoch": 1.86, "learning_rate": 2.1522009966777405e-08, "loss": 0.2355, "step": 8944 }, { "epoch": 1.86, "learning_rate": 2.1490863787375417e-08, "loss": 0.1871, "step": 8945 }, { "epoch": 1.86, "learning_rate": 2.145971760797342e-08, "loss": 0.186, "step": 8946 }, { "epoch": 1.86, "learning_rate": 2.1428571428571426e-08, "loss": 0.1861, "step": 8947 }, { "epoch": 1.86, "learning_rate": 2.1397425249169435e-08, "loss": 0.1123, "step": 8948 }, { "epoch": 1.86, "learning_rate": 2.136627906976744e-08, "loss": 0.1507, "step": 8949 }, { "epoch": 1.86, "learning_rate": 2.1335132890365447e-08, "loss": 0.1485, "step": 8950 }, { "epoch": 1.86, "learning_rate": 2.1303986710963453e-08, "loss": 0.214, "step": 8951 }, { "epoch": 1.86, "learning_rate": 2.1272840531561462e-08, "loss": 0.227, "step": 8952 }, { "epoch": 1.86, "learning_rate": 2.1241694352159468e-08, "loss": 0.1498, "step": 8953 }, { "epoch": 1.86, "learning_rate": 2.1210548172757474e-08, "loss": 0.1798, "step": 8954 }, { "epoch": 1.86, "learning_rate": 2.1179401993355483e-08, "loss": 0.1606, "step": 8955 }, { "epoch": 1.86, "learning_rate": 2.114825581395349e-08, "loss": 0.1857, "step": 8956 }, { "epoch": 1.86, "learning_rate": 2.111710963455149e-08, "loss": 0.1512, "step": 8957 }, { "epoch": 1.86, "learning_rate": 2.1085963455149504e-08, "loss": 0.1901, "step": 8958 }, { "epoch": 1.86, "learning_rate": 2.1054817275747506e-08, "loss": 0.1761, "step": 8959 }, { "epoch": 1.86, "learning_rate": 2.1023671096345512e-08, "loss": 0.1469, "step": 8960 }, { "epoch": 1.86, "learning_rate": 2.0992524916943518e-08, "loss": 0.1776, "step": 8961 }, { "epoch": 1.86, "learning_rate": 2.0961378737541527e-08, "loss": 0.1729, "step": 8962 }, { "epoch": 1.86, "learning_rate": 2.0930232558139533e-08, "loss": 0.1873, "step": 8963 }, { "epoch": 1.86, "learning_rate": 2.089908637873754e-08, "loss": 0.1738, "step": 8964 }, { "epoch": 1.86, "learning_rate": 2.086794019933555e-08, "loss": 0.1996, "step": 8965 }, { "epoch": 1.86, "learning_rate": 2.0836794019933554e-08, "loss": 0.1876, "step": 8966 }, { "epoch": 1.86, "learning_rate": 2.080564784053156e-08, "loss": 0.1704, "step": 8967 }, { "epoch": 1.86, "learning_rate": 2.077450166112957e-08, "loss": 0.1893, "step": 8968 }, { "epoch": 1.86, "learning_rate": 2.0743355481727575e-08, "loss": 0.1584, "step": 8969 }, { "epoch": 1.86, "learning_rate": 2.071220930232558e-08, "loss": 0.1479, "step": 8970 }, { "epoch": 1.86, "learning_rate": 2.0681063122923584e-08, "loss": 0.161, "step": 8971 }, { "epoch": 1.86, "learning_rate": 2.0649916943521596e-08, "loss": 0.1612, "step": 8972 }, { "epoch": 1.86, "learning_rate": 2.06187707641196e-08, "loss": 0.1755, "step": 8973 }, { "epoch": 1.86, "learning_rate": 2.0587624584717605e-08, "loss": 0.1543, "step": 8974 }, { "epoch": 1.86, "learning_rate": 2.0556478405315614e-08, "loss": 0.1373, "step": 8975 }, { "epoch": 1.86, "learning_rate": 2.052533222591362e-08, "loss": 0.1361, "step": 8976 }, { "epoch": 1.86, "learning_rate": 2.0494186046511626e-08, "loss": 0.1732, "step": 8977 }, { "epoch": 1.86, "learning_rate": 2.0463039867109635e-08, "loss": 0.1762, "step": 8978 }, { "epoch": 1.86, "learning_rate": 2.043189368770764e-08, "loss": 0.2223, "step": 8979 }, { "epoch": 1.86, "learning_rate": 2.0400747508305647e-08, "loss": 0.2023, "step": 8980 }, { "epoch": 1.86, "learning_rate": 2.0369601328903653e-08, "loss": 0.171, "step": 8981 }, { "epoch": 1.87, "learning_rate": 2.0338455149501662e-08, "loss": 0.2041, "step": 8982 }, { "epoch": 1.87, "learning_rate": 2.0307308970099668e-08, "loss": 0.1572, "step": 8983 }, { "epoch": 1.87, "learning_rate": 2.027616279069767e-08, "loss": 0.1976, "step": 8984 }, { "epoch": 1.87, "learning_rate": 2.0245016611295683e-08, "loss": 0.1242, "step": 8985 }, { "epoch": 1.87, "learning_rate": 2.0213870431893686e-08, "loss": 0.1562, "step": 8986 }, { "epoch": 1.87, "learning_rate": 2.018272425249169e-08, "loss": 0.1905, "step": 8987 }, { "epoch": 1.87, "learning_rate": 2.01515780730897e-08, "loss": 0.1918, "step": 8988 }, { "epoch": 1.87, "learning_rate": 2.0120431893687707e-08, "loss": 0.2024, "step": 8989 }, { "epoch": 1.87, "learning_rate": 2.0089285714285712e-08, "loss": 0.1757, "step": 8990 }, { "epoch": 1.87, "learning_rate": 2.0058139534883718e-08, "loss": 0.2127, "step": 8991 }, { "epoch": 1.87, "learning_rate": 2.0026993355481728e-08, "loss": 0.2149, "step": 8992 }, { "epoch": 1.87, "learning_rate": 1.9995847176079733e-08, "loss": 0.185, "step": 8993 }, { "epoch": 1.87, "learning_rate": 1.996470099667774e-08, "loss": 0.1602, "step": 8994 }, { "epoch": 1.87, "learning_rate": 1.993355481727575e-08, "loss": 0.1958, "step": 8995 }, { "epoch": 1.87, "learning_rate": 1.9902408637873754e-08, "loss": 0.1792, "step": 8996 }, { "epoch": 1.87, "learning_rate": 1.987126245847176e-08, "loss": 0.1436, "step": 8997 }, { "epoch": 1.87, "learning_rate": 1.984011627906977e-08, "loss": 0.177, "step": 8998 }, { "epoch": 1.87, "learning_rate": 1.9808970099667775e-08, "loss": 0.1521, "step": 8999 }, { "epoch": 1.87, "learning_rate": 1.9777823920265778e-08, "loss": 0.1635, "step": 9000 }, { "epoch": 1.87, "learning_rate": 1.9746677740863784e-08, "loss": 0.2117, "step": 9001 }, { "epoch": 1.87, "learning_rate": 1.9715531561461793e-08, "loss": 0.1909, "step": 9002 }, { "epoch": 1.87, "learning_rate": 1.96843853820598e-08, "loss": 0.1309, "step": 9003 }, { "epoch": 1.87, "learning_rate": 1.9653239202657805e-08, "loss": 0.2257, "step": 9004 }, { "epoch": 1.87, "learning_rate": 1.9622093023255814e-08, "loss": 0.1473, "step": 9005 }, { "epoch": 1.87, "learning_rate": 1.959094684385382e-08, "loss": 0.1678, "step": 9006 }, { "epoch": 1.87, "learning_rate": 1.9559800664451826e-08, "loss": 0.2838, "step": 9007 }, { "epoch": 1.87, "learning_rate": 1.9528654485049835e-08, "loss": 0.2334, "step": 9008 }, { "epoch": 1.87, "learning_rate": 1.949750830564784e-08, "loss": 0.1643, "step": 9009 }, { "epoch": 1.87, "learning_rate": 1.9466362126245847e-08, "loss": 0.2068, "step": 9010 }, { "epoch": 1.87, "learning_rate": 1.943521594684385e-08, "loss": 0.1532, "step": 9011 }, { "epoch": 1.87, "learning_rate": 1.9404069767441862e-08, "loss": 0.1443, "step": 9012 }, { "epoch": 1.87, "learning_rate": 1.9372923588039865e-08, "loss": 0.1734, "step": 9013 }, { "epoch": 1.87, "learning_rate": 1.934177740863787e-08, "loss": 0.2344, "step": 9014 }, { "epoch": 1.87, "learning_rate": 1.931063122923588e-08, "loss": 0.1231, "step": 9015 }, { "epoch": 1.87, "learning_rate": 1.9279485049833886e-08, "loss": 0.1568, "step": 9016 }, { "epoch": 1.87, "learning_rate": 1.924833887043189e-08, "loss": 0.1852, "step": 9017 }, { "epoch": 1.87, "learning_rate": 1.92171926910299e-08, "loss": 0.1944, "step": 9018 }, { "epoch": 1.87, "learning_rate": 1.9186046511627907e-08, "loss": 0.2048, "step": 9019 }, { "epoch": 1.87, "learning_rate": 1.9154900332225913e-08, "loss": 0.1334, "step": 9020 }, { "epoch": 1.87, "learning_rate": 1.912375415282392e-08, "loss": 0.2322, "step": 9021 }, { "epoch": 1.87, "learning_rate": 1.9092607973421928e-08, "loss": 0.1686, "step": 9022 }, { "epoch": 1.87, "learning_rate": 1.9061461794019934e-08, "loss": 0.1454, "step": 9023 }, { "epoch": 1.87, "learning_rate": 1.903031561461794e-08, "loss": 0.1249, "step": 9024 }, { "epoch": 1.87, "learning_rate": 1.899916943521595e-08, "loss": 0.1849, "step": 9025 }, { "epoch": 1.87, "learning_rate": 1.8968023255813955e-08, "loss": 0.1574, "step": 9026 }, { "epoch": 1.87, "learning_rate": 1.8936877076411957e-08, "loss": 0.1547, "step": 9027 }, { "epoch": 1.87, "learning_rate": 1.8905730897009966e-08, "loss": 0.2008, "step": 9028 }, { "epoch": 1.87, "learning_rate": 1.8874584717607972e-08, "loss": 0.1328, "step": 9029 }, { "epoch": 1.88, "learning_rate": 1.8843438538205978e-08, "loss": 0.1651, "step": 9030 }, { "epoch": 1.88, "learning_rate": 1.8812292358803984e-08, "loss": 0.1837, "step": 9031 }, { "epoch": 1.88, "learning_rate": 1.8781146179401993e-08, "loss": 0.1901, "step": 9032 }, { "epoch": 1.88, "learning_rate": 1.875e-08, "loss": 0.1777, "step": 9033 }, { "epoch": 1.88, "learning_rate": 1.8718853820598005e-08, "loss": 0.1536, "step": 9034 }, { "epoch": 1.88, "learning_rate": 1.868770764119601e-08, "loss": 0.1074, "step": 9035 }, { "epoch": 1.88, "learning_rate": 1.865656146179402e-08, "loss": 0.1694, "step": 9036 }, { "epoch": 1.88, "learning_rate": 1.8625415282392026e-08, "loss": 0.0894, "step": 9037 }, { "epoch": 1.88, "learning_rate": 1.8594269102990032e-08, "loss": 0.2292, "step": 9038 }, { "epoch": 1.88, "learning_rate": 1.856312292358804e-08, "loss": 0.2048, "step": 9039 }, { "epoch": 1.88, "learning_rate": 1.8531976744186044e-08, "loss": 0.1575, "step": 9040 }, { "epoch": 1.88, "learning_rate": 1.8500830564784053e-08, "loss": 0.1704, "step": 9041 }, { "epoch": 1.88, "learning_rate": 1.846968438538206e-08, "loss": 0.173, "step": 9042 }, { "epoch": 1.88, "learning_rate": 1.8438538205980065e-08, "loss": 0.2029, "step": 9043 }, { "epoch": 1.88, "learning_rate": 1.8407392026578074e-08, "loss": 0.1686, "step": 9044 }, { "epoch": 1.88, "learning_rate": 1.837624584717608e-08, "loss": 0.1448, "step": 9045 }, { "epoch": 1.88, "learning_rate": 1.8345099667774086e-08, "loss": 0.1438, "step": 9046 }, { "epoch": 1.88, "learning_rate": 1.8313953488372092e-08, "loss": 0.2026, "step": 9047 }, { "epoch": 1.88, "learning_rate": 1.8282807308970098e-08, "loss": 0.1515, "step": 9048 }, { "epoch": 1.88, "learning_rate": 1.8251661129568107e-08, "loss": 0.1898, "step": 9049 }, { "epoch": 1.88, "learning_rate": 1.8220514950166113e-08, "loss": 0.2009, "step": 9050 }, { "epoch": 1.88, "learning_rate": 1.818936877076412e-08, "loss": 0.203, "step": 9051 }, { "epoch": 1.88, "learning_rate": 1.8158222591362124e-08, "loss": 0.2264, "step": 9052 }, { "epoch": 1.88, "learning_rate": 1.8127076411960134e-08, "loss": 0.1425, "step": 9053 }, { "epoch": 1.88, "learning_rate": 1.809593023255814e-08, "loss": 0.1585, "step": 9054 }, { "epoch": 1.88, "learning_rate": 1.8064784053156145e-08, "loss": 0.2035, "step": 9055 }, { "epoch": 1.88, "learning_rate": 1.803363787375415e-08, "loss": 0.1772, "step": 9056 }, { "epoch": 1.88, "learning_rate": 1.8002491694352157e-08, "loss": 0.1646, "step": 9057 }, { "epoch": 1.88, "learning_rate": 1.7971345514950166e-08, "loss": 0.1186, "step": 9058 }, { "epoch": 1.88, "learning_rate": 1.7940199335548172e-08, "loss": 0.1771, "step": 9059 }, { "epoch": 1.88, "learning_rate": 1.7909053156146178e-08, "loss": 0.233, "step": 9060 }, { "epoch": 1.88, "learning_rate": 1.7877906976744184e-08, "loss": 0.1508, "step": 9061 }, { "epoch": 1.88, "learning_rate": 1.784676079734219e-08, "loss": 0.1732, "step": 9062 }, { "epoch": 1.88, "learning_rate": 1.78156146179402e-08, "loss": 0.1186, "step": 9063 }, { "epoch": 1.88, "learning_rate": 1.7784468438538205e-08, "loss": 0.156, "step": 9064 }, { "epoch": 1.88, "learning_rate": 1.775332225913621e-08, "loss": 0.2399, "step": 9065 }, { "epoch": 1.88, "learning_rate": 1.772217607973422e-08, "loss": 0.1395, "step": 9066 }, { "epoch": 1.88, "learning_rate": 1.7691029900332223e-08, "loss": 0.1827, "step": 9067 }, { "epoch": 1.88, "learning_rate": 1.7659883720930232e-08, "loss": 0.1562, "step": 9068 }, { "epoch": 1.88, "learning_rate": 1.7628737541528238e-08, "loss": 0.2222, "step": 9069 }, { "epoch": 1.88, "learning_rate": 1.7597591362126244e-08, "loss": 0.2063, "step": 9070 }, { "epoch": 1.88, "learning_rate": 1.7566445182724253e-08, "loss": 0.1614, "step": 9071 }, { "epoch": 1.88, "learning_rate": 1.753529900332226e-08, "loss": 0.1894, "step": 9072 }, { "epoch": 1.88, "learning_rate": 1.7504152823920265e-08, "loss": 0.1836, "step": 9073 }, { "epoch": 1.88, "learning_rate": 1.7473006644518274e-08, "loss": 0.1792, "step": 9074 }, { "epoch": 1.88, "learning_rate": 1.7441860465116277e-08, "loss": 0.1775, "step": 9075 }, { "epoch": 1.88, "learning_rate": 1.7410714285714286e-08, "loss": 0.1285, "step": 9076 }, { "epoch": 1.88, "learning_rate": 1.7379568106312292e-08, "loss": 0.1396, "step": 9077 }, { "epoch": 1.88, "learning_rate": 1.7348421926910298e-08, "loss": 0.1916, "step": 9078 }, { "epoch": 1.89, "learning_rate": 1.7317275747508307e-08, "loss": 0.1326, "step": 9079 }, { "epoch": 1.89, "learning_rate": 1.7286129568106313e-08, "loss": 0.2342, "step": 9080 }, { "epoch": 1.89, "learning_rate": 1.725498338870432e-08, "loss": 0.15, "step": 9081 }, { "epoch": 1.89, "learning_rate": 1.7223837209302325e-08, "loss": 0.189, "step": 9082 }, { "epoch": 1.89, "learning_rate": 1.719269102990033e-08, "loss": 0.1899, "step": 9083 }, { "epoch": 1.89, "learning_rate": 1.7161544850498336e-08, "loss": 0.2184, "step": 9084 }, { "epoch": 1.89, "learning_rate": 1.7130398671096346e-08, "loss": 0.2068, "step": 9085 }, { "epoch": 1.89, "learning_rate": 1.709925249169435e-08, "loss": 0.177, "step": 9086 }, { "epoch": 1.89, "learning_rate": 1.7068106312292357e-08, "loss": 0.171, "step": 9087 }, { "epoch": 1.89, "learning_rate": 1.7036960132890363e-08, "loss": 0.153, "step": 9088 }, { "epoch": 1.89, "learning_rate": 1.700581395348837e-08, "loss": 0.1847, "step": 9089 }, { "epoch": 1.89, "learning_rate": 1.697466777408638e-08, "loss": 0.1741, "step": 9090 }, { "epoch": 1.89, "learning_rate": 1.6943521594684384e-08, "loss": 0.1451, "step": 9091 }, { "epoch": 1.89, "learning_rate": 1.691237541528239e-08, "loss": 0.1881, "step": 9092 }, { "epoch": 1.89, "learning_rate": 1.68812292358804e-08, "loss": 0.2053, "step": 9093 }, { "epoch": 1.89, "learning_rate": 1.6850083056478402e-08, "loss": 0.1671, "step": 9094 }, { "epoch": 1.89, "learning_rate": 1.681893687707641e-08, "loss": 0.2001, "step": 9095 }, { "epoch": 1.89, "learning_rate": 1.6787790697674417e-08, "loss": 0.1739, "step": 9096 }, { "epoch": 1.89, "learning_rate": 1.6756644518272423e-08, "loss": 0.1917, "step": 9097 }, { "epoch": 1.89, "learning_rate": 1.6725498338870432e-08, "loss": 0.1751, "step": 9098 }, { "epoch": 1.89, "learning_rate": 1.6694352159468438e-08, "loss": 0.2045, "step": 9099 }, { "epoch": 1.89, "learning_rate": 1.6663205980066444e-08, "loss": 0.1415, "step": 9100 }, { "epoch": 1.89, "learning_rate": 1.6632059800664453e-08, "loss": 0.1911, "step": 9101 }, { "epoch": 1.89, "learning_rate": 1.6600913621262456e-08, "loss": 0.1959, "step": 9102 }, { "epoch": 1.89, "learning_rate": 1.6569767441860465e-08, "loss": 0.1814, "step": 9103 }, { "epoch": 1.89, "learning_rate": 1.653862126245847e-08, "loss": 0.161, "step": 9104 }, { "epoch": 1.89, "learning_rate": 1.6507475083056477e-08, "loss": 0.1579, "step": 9105 }, { "epoch": 1.89, "learning_rate": 1.6476328903654486e-08, "loss": 0.1573, "step": 9106 }, { "epoch": 1.89, "learning_rate": 1.6445182724252492e-08, "loss": 0.1226, "step": 9107 }, { "epoch": 1.89, "learning_rate": 1.6414036544850498e-08, "loss": 0.1302, "step": 9108 }, { "epoch": 1.89, "learning_rate": 1.6382890365448504e-08, "loss": 0.1346, "step": 9109 }, { "epoch": 1.89, "learning_rate": 1.635174418604651e-08, "loss": 0.1939, "step": 9110 }, { "epoch": 1.89, "learning_rate": 1.632059800664452e-08, "loss": 0.2287, "step": 9111 }, { "epoch": 1.89, "learning_rate": 1.6289451827242525e-08, "loss": 0.1696, "step": 9112 }, { "epoch": 1.89, "learning_rate": 1.625830564784053e-08, "loss": 0.1927, "step": 9113 }, { "epoch": 1.89, "learning_rate": 1.6227159468438536e-08, "loss": 0.2266, "step": 9114 }, { "epoch": 1.89, "learning_rate": 1.6196013289036546e-08, "loss": 0.1868, "step": 9115 }, { "epoch": 1.89, "learning_rate": 1.616486710963455e-08, "loss": 0.1492, "step": 9116 }, { "epoch": 1.89, "learning_rate": 1.6133720930232557e-08, "loss": 0.1979, "step": 9117 }, { "epoch": 1.89, "learning_rate": 1.6102574750830563e-08, "loss": 0.2045, "step": 9118 }, { "epoch": 1.89, "learning_rate": 1.607142857142857e-08, "loss": 0.1513, "step": 9119 }, { "epoch": 1.89, "learning_rate": 1.604028239202658e-08, "loss": 0.2016, "step": 9120 }, { "epoch": 1.89, "learning_rate": 1.6009136212624584e-08, "loss": 0.1692, "step": 9121 }, { "epoch": 1.89, "learning_rate": 1.597799003322259e-08, "loss": 0.1922, "step": 9122 }, { "epoch": 1.89, "learning_rate": 1.5946843853820596e-08, "loss": 0.1411, "step": 9123 }, { "epoch": 1.89, "learning_rate": 1.5915697674418602e-08, "loss": 0.1803, "step": 9124 }, { "epoch": 1.89, "learning_rate": 1.588455149501661e-08, "loss": 0.1456, "step": 9125 }, { "epoch": 1.89, "learning_rate": 1.5853405315614617e-08, "loss": 0.1794, "step": 9126 }, { "epoch": 1.9, "learning_rate": 1.5822259136212623e-08, "loss": 0.1756, "step": 9127 }, { "epoch": 1.9, "learning_rate": 1.5791112956810632e-08, "loss": 0.1699, "step": 9128 }, { "epoch": 1.9, "learning_rate": 1.5759966777408635e-08, "loss": 0.2076, "step": 9129 }, { "epoch": 1.9, "learning_rate": 1.5728820598006644e-08, "loss": 0.1574, "step": 9130 }, { "epoch": 1.9, "learning_rate": 1.569767441860465e-08, "loss": 0.1841, "step": 9131 }, { "epoch": 1.9, "learning_rate": 1.5666528239202656e-08, "loss": 0.1381, "step": 9132 }, { "epoch": 1.9, "learning_rate": 1.5635382059800665e-08, "loss": 0.1907, "step": 9133 }, { "epoch": 1.9, "learning_rate": 1.560423588039867e-08, "loss": 0.1758, "step": 9134 }, { "epoch": 1.9, "learning_rate": 1.5573089700996677e-08, "loss": 0.1378, "step": 9135 }, { "epoch": 1.9, "learning_rate": 1.5541943521594686e-08, "loss": 0.1981, "step": 9136 }, { "epoch": 1.9, "learning_rate": 1.551079734219269e-08, "loss": 0.1832, "step": 9137 }, { "epoch": 1.9, "learning_rate": 1.5479651162790698e-08, "loss": 0.1856, "step": 9138 }, { "epoch": 1.9, "learning_rate": 1.5448504983388704e-08, "loss": 0.2015, "step": 9139 }, { "epoch": 1.9, "learning_rate": 1.541735880398671e-08, "loss": 0.1647, "step": 9140 }, { "epoch": 1.9, "learning_rate": 1.538621262458472e-08, "loss": 0.1211, "step": 9141 }, { "epoch": 1.9, "learning_rate": 1.5355066445182725e-08, "loss": 0.1843, "step": 9142 }, { "epoch": 1.9, "learning_rate": 1.532392026578073e-08, "loss": 0.2398, "step": 9143 }, { "epoch": 1.9, "learning_rate": 1.5292774086378737e-08, "loss": 0.1781, "step": 9144 }, { "epoch": 1.9, "learning_rate": 1.5261627906976742e-08, "loss": 0.1862, "step": 9145 }, { "epoch": 1.9, "learning_rate": 1.5230481727574752e-08, "loss": 0.1855, "step": 9146 }, { "epoch": 1.9, "learning_rate": 1.5199335548172758e-08, "loss": 0.1891, "step": 9147 }, { "epoch": 1.9, "learning_rate": 1.5168189368770763e-08, "loss": 0.1854, "step": 9148 }, { "epoch": 1.9, "learning_rate": 1.513704318936877e-08, "loss": 0.1691, "step": 9149 }, { "epoch": 1.9, "learning_rate": 1.5105897009966775e-08, "loss": 0.1451, "step": 9150 }, { "epoch": 1.9, "learning_rate": 1.5074750830564785e-08, "loss": 0.1916, "step": 9151 }, { "epoch": 1.9, "learning_rate": 1.504360465116279e-08, "loss": 0.1384, "step": 9152 }, { "epoch": 1.9, "learning_rate": 1.5012458471760796e-08, "loss": 0.2025, "step": 9153 }, { "epoch": 1.9, "learning_rate": 1.4981312292358802e-08, "loss": 0.236, "step": 9154 }, { "epoch": 1.9, "learning_rate": 1.495016611295681e-08, "loss": 0.2188, "step": 9155 }, { "epoch": 1.9, "learning_rate": 1.4919019933554817e-08, "loss": 0.1249, "step": 9156 }, { "epoch": 1.9, "learning_rate": 1.4887873754152823e-08, "loss": 0.1338, "step": 9157 }, { "epoch": 1.9, "learning_rate": 1.485672757475083e-08, "loss": 0.1924, "step": 9158 }, { "epoch": 1.9, "learning_rate": 1.4825581395348835e-08, "loss": 0.1801, "step": 9159 }, { "epoch": 1.9, "learning_rate": 1.4794435215946843e-08, "loss": 0.1872, "step": 9160 }, { "epoch": 1.9, "learning_rate": 1.476328903654485e-08, "loss": 0.1811, "step": 9161 }, { "epoch": 1.9, "learning_rate": 1.4732142857142856e-08, "loss": 0.1889, "step": 9162 }, { "epoch": 1.9, "learning_rate": 1.4700996677740864e-08, "loss": 0.1383, "step": 9163 }, { "epoch": 1.9, "learning_rate": 1.466985049833887e-08, "loss": 0.1875, "step": 9164 }, { "epoch": 1.9, "learning_rate": 1.4638704318936877e-08, "loss": 0.164, "step": 9165 }, { "epoch": 1.9, "learning_rate": 1.4607558139534885e-08, "loss": 0.1341, "step": 9166 }, { "epoch": 1.9, "learning_rate": 1.4576411960132889e-08, "loss": 0.2644, "step": 9167 }, { "epoch": 1.9, "learning_rate": 1.4545265780730896e-08, "loss": 0.1777, "step": 9168 }, { "epoch": 1.9, "learning_rate": 1.4514119601328902e-08, "loss": 0.1595, "step": 9169 }, { "epoch": 1.9, "learning_rate": 1.448297342192691e-08, "loss": 0.1378, "step": 9170 }, { "epoch": 1.9, "learning_rate": 1.4451827242524917e-08, "loss": 0.2341, "step": 9171 }, { "epoch": 1.9, "learning_rate": 1.4420681063122923e-08, "loss": 0.1389, "step": 9172 }, { "epoch": 1.9, "learning_rate": 1.438953488372093e-08, "loss": 0.1621, "step": 9173 }, { "epoch": 1.9, "learning_rate": 1.4358388704318935e-08, "loss": 0.1533, "step": 9174 }, { "epoch": 1.91, "learning_rate": 1.4327242524916943e-08, "loss": 0.1851, "step": 9175 }, { "epoch": 1.91, "learning_rate": 1.429609634551495e-08, "loss": 0.1237, "step": 9176 }, { "epoch": 1.91, "learning_rate": 1.4264950166112956e-08, "loss": 0.162, "step": 9177 }, { "epoch": 1.91, "learning_rate": 1.4233803986710964e-08, "loss": 0.1358, "step": 9178 }, { "epoch": 1.91, "learning_rate": 1.420265780730897e-08, "loss": 0.1534, "step": 9179 }, { "epoch": 1.91, "learning_rate": 1.4171511627906977e-08, "loss": 0.1559, "step": 9180 }, { "epoch": 1.91, "learning_rate": 1.4140365448504983e-08, "loss": 0.1271, "step": 9181 }, { "epoch": 1.91, "learning_rate": 1.4109219269102989e-08, "loss": 0.2359, "step": 9182 }, { "epoch": 1.91, "learning_rate": 1.4078073089700996e-08, "loss": 0.1883, "step": 9183 }, { "epoch": 1.91, "learning_rate": 1.4046926910299002e-08, "loss": 0.1616, "step": 9184 }, { "epoch": 1.91, "learning_rate": 1.401578073089701e-08, "loss": 0.177, "step": 9185 }, { "epoch": 1.91, "learning_rate": 1.3984634551495016e-08, "loss": 0.151, "step": 9186 }, { "epoch": 1.91, "learning_rate": 1.3953488372093022e-08, "loss": 0.1611, "step": 9187 }, { "epoch": 1.91, "learning_rate": 1.392234219269103e-08, "loss": 0.1505, "step": 9188 }, { "epoch": 1.91, "learning_rate": 1.3891196013289035e-08, "loss": 0.1996, "step": 9189 }, { "epoch": 1.91, "learning_rate": 1.3860049833887043e-08, "loss": 0.1645, "step": 9190 }, { "epoch": 1.91, "learning_rate": 1.3828903654485049e-08, "loss": 0.2075, "step": 9191 }, { "epoch": 1.91, "learning_rate": 1.3797757475083056e-08, "loss": 0.1627, "step": 9192 }, { "epoch": 1.91, "learning_rate": 1.3766611295681064e-08, "loss": 0.2068, "step": 9193 }, { "epoch": 1.91, "learning_rate": 1.3735465116279068e-08, "loss": 0.1727, "step": 9194 }, { "epoch": 1.91, "learning_rate": 1.3704318936877075e-08, "loss": 0.1666, "step": 9195 }, { "epoch": 1.91, "learning_rate": 1.3673172757475081e-08, "loss": 0.2073, "step": 9196 }, { "epoch": 1.91, "learning_rate": 1.3642026578073089e-08, "loss": 0.1706, "step": 9197 }, { "epoch": 1.91, "learning_rate": 1.3610880398671096e-08, "loss": 0.2219, "step": 9198 }, { "epoch": 1.91, "learning_rate": 1.3579734219269102e-08, "loss": 0.1907, "step": 9199 }, { "epoch": 1.91, "learning_rate": 1.354858803986711e-08, "loss": 0.2573, "step": 9200 }, { "epoch": 1.91, "learning_rate": 1.3517441860465114e-08, "loss": 0.1715, "step": 9201 }, { "epoch": 1.91, "learning_rate": 1.3486295681063122e-08, "loss": 0.2275, "step": 9202 }, { "epoch": 1.91, "learning_rate": 1.345514950166113e-08, "loss": 0.1624, "step": 9203 }, { "epoch": 1.91, "learning_rate": 1.3424003322259135e-08, "loss": 0.2095, "step": 9204 }, { "epoch": 1.91, "learning_rate": 1.3392857142857143e-08, "loss": 0.1444, "step": 9205 }, { "epoch": 1.91, "learning_rate": 1.3361710963455149e-08, "loss": 0.181, "step": 9206 }, { "epoch": 1.91, "learning_rate": 1.3330564784053156e-08, "loss": 0.1891, "step": 9207 }, { "epoch": 1.91, "learning_rate": 1.3299418604651164e-08, "loss": 0.1409, "step": 9208 }, { "epoch": 1.91, "learning_rate": 1.3268272425249168e-08, "loss": 0.1541, "step": 9209 }, { "epoch": 1.91, "learning_rate": 1.3237126245847176e-08, "loss": 0.1395, "step": 9210 }, { "epoch": 1.91, "learning_rate": 1.3205980066445181e-08, "loss": 0.2212, "step": 9211 }, { "epoch": 1.91, "learning_rate": 1.3174833887043189e-08, "loss": 0.1628, "step": 9212 }, { "epoch": 1.91, "learning_rate": 1.3143687707641197e-08, "loss": 0.2031, "step": 9213 }, { "epoch": 1.91, "learning_rate": 1.31125415282392e-08, "loss": 0.2104, "step": 9214 }, { "epoch": 1.91, "learning_rate": 1.3081395348837208e-08, "loss": 0.2105, "step": 9215 }, { "epoch": 1.91, "learning_rate": 1.3050249169435214e-08, "loss": 0.1912, "step": 9216 }, { "epoch": 1.91, "learning_rate": 1.3019102990033222e-08, "loss": 0.2161, "step": 9217 }, { "epoch": 1.91, "learning_rate": 1.298795681063123e-08, "loss": 0.1707, "step": 9218 }, { "epoch": 1.91, "learning_rate": 1.2956810631229235e-08, "loss": 0.1534, "step": 9219 }, { "epoch": 1.91, "learning_rate": 1.2925664451827243e-08, "loss": 0.1822, "step": 9220 }, { "epoch": 1.91, "learning_rate": 1.2894518272425247e-08, "loss": 0.1523, "step": 9221 }, { "epoch": 1.91, "learning_rate": 1.2863372093023255e-08, "loss": 0.2279, "step": 9222 }, { "epoch": 1.92, "learning_rate": 1.2832225913621262e-08, "loss": 0.175, "step": 9223 }, { "epoch": 1.92, "learning_rate": 1.2801079734219268e-08, "loss": 0.173, "step": 9224 }, { "epoch": 1.92, "learning_rate": 1.2769933554817276e-08, "loss": 0.157, "step": 9225 }, { "epoch": 1.92, "learning_rate": 1.2738787375415281e-08, "loss": 0.1632, "step": 9226 }, { "epoch": 1.92, "learning_rate": 1.2707641196013289e-08, "loss": 0.1898, "step": 9227 }, { "epoch": 1.92, "learning_rate": 1.2676495016611297e-08, "loss": 0.2064, "step": 9228 }, { "epoch": 1.92, "learning_rate": 1.2645348837209301e-08, "loss": 0.1833, "step": 9229 }, { "epoch": 1.92, "learning_rate": 1.2614202657807308e-08, "loss": 0.1633, "step": 9230 }, { "epoch": 1.92, "learning_rate": 1.2583056478405314e-08, "loss": 0.2065, "step": 9231 }, { "epoch": 1.92, "learning_rate": 1.2551910299003322e-08, "loss": 0.1737, "step": 9232 }, { "epoch": 1.92, "learning_rate": 1.252076411960133e-08, "loss": 0.1365, "step": 9233 }, { "epoch": 1.92, "learning_rate": 1.2489617940199335e-08, "loss": 0.1383, "step": 9234 }, { "epoch": 1.92, "learning_rate": 1.2458471760797343e-08, "loss": 0.2223, "step": 9235 }, { "epoch": 1.92, "learning_rate": 1.2427325581395347e-08, "loss": 0.1544, "step": 9236 }, { "epoch": 1.92, "learning_rate": 1.2396179401993355e-08, "loss": 0.2242, "step": 9237 }, { "epoch": 1.92, "learning_rate": 1.2365033222591362e-08, "loss": 0.2128, "step": 9238 }, { "epoch": 1.92, "learning_rate": 1.2333887043189368e-08, "loss": 0.1576, "step": 9239 }, { "epoch": 1.92, "learning_rate": 1.2302740863787376e-08, "loss": 0.1475, "step": 9240 }, { "epoch": 1.92, "learning_rate": 1.227159468438538e-08, "loss": 0.1391, "step": 9241 }, { "epoch": 1.92, "learning_rate": 1.2240448504983387e-08, "loss": 0.1307, "step": 9242 }, { "epoch": 1.92, "learning_rate": 1.2209302325581395e-08, "loss": 0.1652, "step": 9243 }, { "epoch": 1.92, "learning_rate": 1.2178156146179401e-08, "loss": 0.1517, "step": 9244 }, { "epoch": 1.92, "learning_rate": 1.2147009966777408e-08, "loss": 0.2056, "step": 9245 }, { "epoch": 1.92, "learning_rate": 1.2115863787375414e-08, "loss": 0.2057, "step": 9246 }, { "epoch": 1.92, "learning_rate": 1.2084717607973422e-08, "loss": 0.2306, "step": 9247 }, { "epoch": 1.92, "learning_rate": 1.205357142857143e-08, "loss": 0.1697, "step": 9248 }, { "epoch": 1.92, "learning_rate": 1.2022425249169434e-08, "loss": 0.1915, "step": 9249 }, { "epoch": 1.92, "learning_rate": 1.1991279069767441e-08, "loss": 0.1497, "step": 9250 }, { "epoch": 1.92, "learning_rate": 1.1960132890365447e-08, "loss": 0.2742, "step": 9251 }, { "epoch": 1.92, "learning_rate": 1.1928986710963455e-08, "loss": 0.1314, "step": 9252 }, { "epoch": 1.92, "learning_rate": 1.1897840531561462e-08, "loss": 0.1807, "step": 9253 }, { "epoch": 1.92, "learning_rate": 1.1866694352159468e-08, "loss": 0.1884, "step": 9254 }, { "epoch": 1.92, "learning_rate": 1.1835548172757476e-08, "loss": 0.2033, "step": 9255 }, { "epoch": 1.92, "learning_rate": 1.180440199335548e-08, "loss": 0.15, "step": 9256 }, { "epoch": 1.92, "learning_rate": 1.1773255813953487e-08, "loss": 0.1997, "step": 9257 }, { "epoch": 1.92, "learning_rate": 1.1742109634551495e-08, "loss": 0.1339, "step": 9258 }, { "epoch": 1.92, "learning_rate": 1.1710963455149501e-08, "loss": 0.1593, "step": 9259 }, { "epoch": 1.92, "learning_rate": 1.1679817275747508e-08, "loss": 0.1897, "step": 9260 }, { "epoch": 1.92, "learning_rate": 1.1648671096345514e-08, "loss": 0.161, "step": 9261 }, { "epoch": 1.92, "learning_rate": 1.1617524916943522e-08, "loss": 0.1459, "step": 9262 }, { "epoch": 1.92, "learning_rate": 1.1586378737541528e-08, "loss": 0.2344, "step": 9263 }, { "epoch": 1.92, "learning_rate": 1.1555232558139534e-08, "loss": 0.1477, "step": 9264 }, { "epoch": 1.92, "learning_rate": 1.1524086378737541e-08, "loss": 0.1597, "step": 9265 }, { "epoch": 1.92, "learning_rate": 1.1492940199335547e-08, "loss": 0.1628, "step": 9266 }, { "epoch": 1.92, "learning_rate": 1.1461794019933555e-08, "loss": 0.1743, "step": 9267 }, { "epoch": 1.92, "learning_rate": 1.1430647840531562e-08, "loss": 0.1871, "step": 9268 }, { "epoch": 1.92, "learning_rate": 1.1399501661129567e-08, "loss": 0.2157, "step": 9269 }, { "epoch": 1.92, "learning_rate": 1.1368355481727574e-08, "loss": 0.1928, "step": 9270 }, { "epoch": 1.93, "learning_rate": 1.133720930232558e-08, "loss": 0.1992, "step": 9271 }, { "epoch": 1.93, "learning_rate": 1.1306063122923588e-08, "loss": 0.1549, "step": 9272 }, { "epoch": 1.93, "learning_rate": 1.1274916943521595e-08, "loss": 0.1822, "step": 9273 }, { "epoch": 1.93, "learning_rate": 1.1243770764119601e-08, "loss": 0.1966, "step": 9274 }, { "epoch": 1.93, "learning_rate": 1.1212624584717609e-08, "loss": 0.1561, "step": 9275 }, { "epoch": 1.93, "learning_rate": 1.1181478405315613e-08, "loss": 0.2122, "step": 9276 }, { "epoch": 1.93, "learning_rate": 1.115033222591362e-08, "loss": 0.1549, "step": 9277 }, { "epoch": 1.93, "learning_rate": 1.1119186046511628e-08, "loss": 0.1936, "step": 9278 }, { "epoch": 1.93, "learning_rate": 1.1088039867109634e-08, "loss": 0.1553, "step": 9279 }, { "epoch": 1.93, "learning_rate": 1.1056893687707641e-08, "loss": 0.2028, "step": 9280 }, { "epoch": 1.93, "learning_rate": 1.1025747508305647e-08, "loss": 0.2173, "step": 9281 }, { "epoch": 1.93, "learning_rate": 1.0994601328903655e-08, "loss": 0.1948, "step": 9282 }, { "epoch": 1.93, "learning_rate": 1.0963455149501662e-08, "loss": 0.1617, "step": 9283 }, { "epoch": 1.93, "learning_rate": 1.0932308970099667e-08, "loss": 0.1785, "step": 9284 }, { "epoch": 1.93, "learning_rate": 1.0901162790697674e-08, "loss": 0.18, "step": 9285 }, { "epoch": 1.93, "learning_rate": 1.087001661129568e-08, "loss": 0.1828, "step": 9286 }, { "epoch": 1.93, "learning_rate": 1.0838870431893688e-08, "loss": 0.1823, "step": 9287 }, { "epoch": 1.93, "learning_rate": 1.0807724252491694e-08, "loss": 0.1894, "step": 9288 }, { "epoch": 1.93, "learning_rate": 1.0776578073089701e-08, "loss": 0.1952, "step": 9289 }, { "epoch": 1.93, "learning_rate": 1.0745431893687709e-08, "loss": 0.1703, "step": 9290 }, { "epoch": 1.93, "learning_rate": 1.0714285714285713e-08, "loss": 0.1404, "step": 9291 }, { "epoch": 1.93, "learning_rate": 1.068313953488372e-08, "loss": 0.1914, "step": 9292 }, { "epoch": 1.93, "learning_rate": 1.0651993355481726e-08, "loss": 0.169, "step": 9293 }, { "epoch": 1.93, "learning_rate": 1.0620847176079734e-08, "loss": 0.1757, "step": 9294 }, { "epoch": 1.93, "learning_rate": 1.0589700996677741e-08, "loss": 0.152, "step": 9295 }, { "epoch": 1.93, "learning_rate": 1.0558554817275746e-08, "loss": 0.1857, "step": 9296 }, { "epoch": 1.93, "learning_rate": 1.0527408637873753e-08, "loss": 0.174, "step": 9297 }, { "epoch": 1.93, "learning_rate": 1.0496262458471759e-08, "loss": 0.1856, "step": 9298 }, { "epoch": 1.93, "learning_rate": 1.0465116279069767e-08, "loss": 0.1437, "step": 9299 }, { "epoch": 1.93, "learning_rate": 1.0433970099667774e-08, "loss": 0.1785, "step": 9300 }, { "epoch": 1.93, "learning_rate": 1.040282392026578e-08, "loss": 0.1153, "step": 9301 }, { "epoch": 1.93, "learning_rate": 1.0371677740863788e-08, "loss": 0.1448, "step": 9302 }, { "epoch": 1.93, "learning_rate": 1.0340531561461792e-08, "loss": 0.1995, "step": 9303 }, { "epoch": 1.93, "learning_rate": 1.03093853820598e-08, "loss": 0.2008, "step": 9304 }, { "epoch": 1.93, "learning_rate": 1.0278239202657807e-08, "loss": 0.2073, "step": 9305 }, { "epoch": 1.93, "learning_rate": 1.0247093023255813e-08, "loss": 0.174, "step": 9306 }, { "epoch": 1.93, "learning_rate": 1.021594684385382e-08, "loss": 0.1593, "step": 9307 }, { "epoch": 1.93, "learning_rate": 1.0184800664451826e-08, "loss": 0.2243, "step": 9308 }, { "epoch": 1.93, "learning_rate": 1.0153654485049834e-08, "loss": 0.2257, "step": 9309 }, { "epoch": 1.93, "learning_rate": 1.0122508305647841e-08, "loss": 0.1971, "step": 9310 }, { "epoch": 1.93, "learning_rate": 1.0091362126245846e-08, "loss": 0.155, "step": 9311 }, { "epoch": 1.93, "learning_rate": 1.0060215946843853e-08, "loss": 0.1934, "step": 9312 }, { "epoch": 1.93, "learning_rate": 1.0029069767441859e-08, "loss": 0.1615, "step": 9313 }, { "epoch": 1.93, "learning_rate": 9.997923588039867e-09, "loss": 0.1406, "step": 9314 }, { "epoch": 1.93, "learning_rate": 9.966777408637874e-09, "loss": 0.2021, "step": 9315 }, { "epoch": 1.93, "learning_rate": 9.93563122923588e-09, "loss": 0.1909, "step": 9316 }, { "epoch": 1.93, "learning_rate": 9.904485049833888e-09, "loss": 0.1987, "step": 9317 }, { "epoch": 1.93, "learning_rate": 9.873338870431892e-09, "loss": 0.1731, "step": 9318 }, { "epoch": 1.94, "learning_rate": 9.8421926910299e-09, "loss": 0.1808, "step": 9319 }, { "epoch": 1.94, "learning_rate": 9.811046511627907e-09, "loss": 0.1668, "step": 9320 }, { "epoch": 1.94, "learning_rate": 9.779900332225913e-09, "loss": 0.1586, "step": 9321 }, { "epoch": 1.94, "learning_rate": 9.74875415282392e-09, "loss": 0.1719, "step": 9322 }, { "epoch": 1.94, "learning_rate": 9.717607973421925e-09, "loss": 0.1512, "step": 9323 }, { "epoch": 1.94, "learning_rate": 9.686461794019932e-09, "loss": 0.2417, "step": 9324 }, { "epoch": 1.94, "learning_rate": 9.65531561461794e-09, "loss": 0.1907, "step": 9325 }, { "epoch": 1.94, "learning_rate": 9.624169435215946e-09, "loss": 0.1991, "step": 9326 }, { "epoch": 1.94, "learning_rate": 9.593023255813953e-09, "loss": 0.2481, "step": 9327 }, { "epoch": 1.94, "learning_rate": 9.56187707641196e-09, "loss": 0.1591, "step": 9328 }, { "epoch": 1.94, "learning_rate": 9.530730897009967e-09, "loss": 0.2305, "step": 9329 }, { "epoch": 1.94, "learning_rate": 9.499584717607974e-09, "loss": 0.186, "step": 9330 }, { "epoch": 1.94, "learning_rate": 9.468438538205979e-09, "loss": 0.1892, "step": 9331 }, { "epoch": 1.94, "learning_rate": 9.437292358803986e-09, "loss": 0.2046, "step": 9332 }, { "epoch": 1.94, "learning_rate": 9.406146179401992e-09, "loss": 0.1742, "step": 9333 }, { "epoch": 1.94, "learning_rate": 9.375e-09, "loss": 0.1369, "step": 9334 }, { "epoch": 1.94, "learning_rate": 9.343853820598005e-09, "loss": 0.2167, "step": 9335 }, { "epoch": 1.94, "learning_rate": 9.312707641196013e-09, "loss": 0.1826, "step": 9336 }, { "epoch": 1.94, "learning_rate": 9.28156146179402e-09, "loss": 0.2057, "step": 9337 }, { "epoch": 1.94, "learning_rate": 9.250415282392026e-09, "loss": 0.1807, "step": 9338 }, { "epoch": 1.94, "learning_rate": 9.219269102990032e-09, "loss": 0.1724, "step": 9339 }, { "epoch": 1.94, "learning_rate": 9.18812292358804e-09, "loss": 0.1543, "step": 9340 }, { "epoch": 1.94, "learning_rate": 9.156976744186046e-09, "loss": 0.1505, "step": 9341 }, { "epoch": 1.94, "learning_rate": 9.125830564784053e-09, "loss": 0.1304, "step": 9342 }, { "epoch": 1.94, "learning_rate": 9.09468438538206e-09, "loss": 0.1599, "step": 9343 }, { "epoch": 1.94, "learning_rate": 9.063538205980067e-09, "loss": 0.1415, "step": 9344 }, { "epoch": 1.94, "learning_rate": 9.032392026578073e-09, "loss": 0.1842, "step": 9345 }, { "epoch": 1.94, "learning_rate": 9.001245847176079e-09, "loss": 0.1746, "step": 9346 }, { "epoch": 1.94, "learning_rate": 8.970099667774086e-09, "loss": 0.1676, "step": 9347 }, { "epoch": 1.94, "learning_rate": 8.938953488372092e-09, "loss": 0.1784, "step": 9348 }, { "epoch": 1.94, "learning_rate": 8.9078073089701e-09, "loss": 0.1573, "step": 9349 }, { "epoch": 1.94, "learning_rate": 8.876661129568106e-09, "loss": 0.1934, "step": 9350 }, { "epoch": 1.94, "learning_rate": 8.845514950166111e-09, "loss": 0.1578, "step": 9351 }, { "epoch": 1.94, "learning_rate": 8.814368770764119e-09, "loss": 0.182, "step": 9352 }, { "epoch": 1.94, "learning_rate": 8.783222591362127e-09, "loss": 0.2127, "step": 9353 }, { "epoch": 1.94, "learning_rate": 8.752076411960132e-09, "loss": 0.1263, "step": 9354 }, { "epoch": 1.94, "learning_rate": 8.720930232558138e-09, "loss": 0.1904, "step": 9355 }, { "epoch": 1.94, "learning_rate": 8.689784053156146e-09, "loss": 0.1213, "step": 9356 }, { "epoch": 1.94, "learning_rate": 8.658637873754153e-09, "loss": 0.1764, "step": 9357 }, { "epoch": 1.94, "learning_rate": 8.62749169435216e-09, "loss": 0.1897, "step": 9358 }, { "epoch": 1.94, "learning_rate": 8.596345514950165e-09, "loss": 0.1294, "step": 9359 }, { "epoch": 1.94, "learning_rate": 8.565199335548173e-09, "loss": 0.1689, "step": 9360 }, { "epoch": 1.94, "learning_rate": 8.534053156146179e-09, "loss": 0.159, "step": 9361 }, { "epoch": 1.94, "learning_rate": 8.502906976744185e-09, "loss": 0.2037, "step": 9362 }, { "epoch": 1.94, "learning_rate": 8.471760797342192e-09, "loss": 0.1698, "step": 9363 }, { "epoch": 1.94, "learning_rate": 8.4406146179402e-09, "loss": 0.1865, "step": 9364 }, { "epoch": 1.94, "learning_rate": 8.409468438538206e-09, "loss": 0.1173, "step": 9365 }, { "epoch": 1.94, "learning_rate": 8.378322259136211e-09, "loss": 0.1345, "step": 9366 }, { "epoch": 1.94, "learning_rate": 8.347176079734219e-09, "loss": 0.1405, "step": 9367 }, { "epoch": 1.95, "learning_rate": 8.316029900332227e-09, "loss": 0.165, "step": 9368 }, { "epoch": 1.95, "learning_rate": 8.284883720930232e-09, "loss": 0.2079, "step": 9369 }, { "epoch": 1.95, "learning_rate": 8.253737541528238e-09, "loss": 0.1373, "step": 9370 }, { "epoch": 1.95, "learning_rate": 8.222591362126246e-09, "loss": 0.1713, "step": 9371 }, { "epoch": 1.95, "learning_rate": 8.191445182724252e-09, "loss": 0.1934, "step": 9372 }, { "epoch": 1.95, "learning_rate": 8.16029900332226e-09, "loss": 0.1706, "step": 9373 }, { "epoch": 1.95, "learning_rate": 8.129152823920265e-09, "loss": 0.2158, "step": 9374 }, { "epoch": 1.95, "learning_rate": 8.098006644518273e-09, "loss": 0.1907, "step": 9375 }, { "epoch": 1.95, "learning_rate": 8.066860465116279e-09, "loss": 0.1763, "step": 9376 }, { "epoch": 1.95, "learning_rate": 8.035714285714285e-09, "loss": 0.167, "step": 9377 }, { "epoch": 1.95, "learning_rate": 8.004568106312292e-09, "loss": 0.1783, "step": 9378 }, { "epoch": 1.95, "learning_rate": 7.973421926910298e-09, "loss": 0.1501, "step": 9379 }, { "epoch": 1.95, "learning_rate": 7.942275747508306e-09, "loss": 0.1797, "step": 9380 }, { "epoch": 1.95, "learning_rate": 7.911129568106312e-09, "loss": 0.168, "step": 9381 }, { "epoch": 1.95, "learning_rate": 7.879983388704317e-09, "loss": 0.1319, "step": 9382 }, { "epoch": 1.95, "learning_rate": 7.848837209302325e-09, "loss": 0.1707, "step": 9383 }, { "epoch": 1.95, "learning_rate": 7.817691029900333e-09, "loss": 0.1783, "step": 9384 }, { "epoch": 1.95, "learning_rate": 7.786544850498338e-09, "loss": 0.1802, "step": 9385 }, { "epoch": 1.95, "learning_rate": 7.755398671096344e-09, "loss": 0.1208, "step": 9386 }, { "epoch": 1.95, "learning_rate": 7.724252491694352e-09, "loss": 0.1402, "step": 9387 }, { "epoch": 1.95, "learning_rate": 7.69310631229236e-09, "loss": 0.2076, "step": 9388 }, { "epoch": 1.95, "learning_rate": 7.661960132890365e-09, "loss": 0.2812, "step": 9389 }, { "epoch": 1.95, "learning_rate": 7.630813953488371e-09, "loss": 0.1808, "step": 9390 }, { "epoch": 1.95, "learning_rate": 7.599667774086379e-09, "loss": 0.2358, "step": 9391 }, { "epoch": 1.95, "learning_rate": 7.568521594684385e-09, "loss": 0.1787, "step": 9392 }, { "epoch": 1.95, "learning_rate": 7.537375415282392e-09, "loss": 0.2115, "step": 9393 }, { "epoch": 1.95, "learning_rate": 7.506229235880398e-09, "loss": 0.1784, "step": 9394 }, { "epoch": 1.95, "learning_rate": 7.475083056478406e-09, "loss": 0.1549, "step": 9395 }, { "epoch": 1.95, "learning_rate": 7.443936877076412e-09, "loss": 0.2165, "step": 9396 }, { "epoch": 1.95, "learning_rate": 7.4127906976744175e-09, "loss": 0.1857, "step": 9397 }, { "epoch": 1.95, "learning_rate": 7.381644518272425e-09, "loss": 0.1551, "step": 9398 }, { "epoch": 1.95, "learning_rate": 7.350498338870432e-09, "loss": 0.1658, "step": 9399 }, { "epoch": 1.95, "learning_rate": 7.3193521594684385e-09, "loss": 0.1248, "step": 9400 }, { "epoch": 1.95, "learning_rate": 7.288205980066444e-09, "loss": 0.1702, "step": 9401 }, { "epoch": 1.95, "learning_rate": 7.257059800664451e-09, "loss": 0.2376, "step": 9402 }, { "epoch": 1.95, "learning_rate": 7.225913621262459e-09, "loss": 0.2086, "step": 9403 }, { "epoch": 1.95, "learning_rate": 7.194767441860465e-09, "loss": 0.2327, "step": 9404 }, { "epoch": 1.95, "learning_rate": 7.163621262458471e-09, "loss": 0.1514, "step": 9405 }, { "epoch": 1.95, "learning_rate": 7.132475083056478e-09, "loss": 0.155, "step": 9406 }, { "epoch": 1.95, "learning_rate": 7.101328903654485e-09, "loss": 0.1949, "step": 9407 }, { "epoch": 1.95, "learning_rate": 7.0701827242524915e-09, "loss": 0.1506, "step": 9408 }, { "epoch": 1.95, "learning_rate": 7.039036544850498e-09, "loss": 0.1764, "step": 9409 }, { "epoch": 1.95, "learning_rate": 7.007890365448505e-09, "loss": 0.1643, "step": 9410 }, { "epoch": 1.95, "learning_rate": 6.976744186046511e-09, "loss": 0.2213, "step": 9411 }, { "epoch": 1.95, "learning_rate": 6.9455980066445176e-09, "loss": 0.1721, "step": 9412 }, { "epoch": 1.95, "learning_rate": 6.914451827242524e-09, "loss": 0.264, "step": 9413 }, { "epoch": 1.95, "learning_rate": 6.883305647840532e-09, "loss": 0.2195, "step": 9414 }, { "epoch": 1.95, "learning_rate": 6.852159468438538e-09, "loss": 0.1894, "step": 9415 }, { "epoch": 1.96, "learning_rate": 6.8210132890365445e-09, "loss": 0.1603, "step": 9416 }, { "epoch": 1.96, "learning_rate": 6.789867109634551e-09, "loss": 0.1793, "step": 9417 }, { "epoch": 1.96, "learning_rate": 6.758720930232557e-09, "loss": 0.1685, "step": 9418 }, { "epoch": 1.96, "learning_rate": 6.727574750830565e-09, "loss": 0.1395, "step": 9419 }, { "epoch": 1.96, "learning_rate": 6.696428571428571e-09, "loss": 0.1857, "step": 9420 }, { "epoch": 1.96, "learning_rate": 6.665282392026578e-09, "loss": 0.1792, "step": 9421 }, { "epoch": 1.96, "learning_rate": 6.634136212624584e-09, "loss": 0.1523, "step": 9422 }, { "epoch": 1.96, "learning_rate": 6.602990033222591e-09, "loss": 0.2071, "step": 9423 }, { "epoch": 1.96, "learning_rate": 6.571843853820598e-09, "loss": 0.1417, "step": 9424 }, { "epoch": 1.96, "learning_rate": 6.540697674418604e-09, "loss": 0.1711, "step": 9425 }, { "epoch": 1.96, "learning_rate": 6.509551495016611e-09, "loss": 0.1356, "step": 9426 }, { "epoch": 1.96, "learning_rate": 6.478405315614618e-09, "loss": 0.1558, "step": 9427 }, { "epoch": 1.96, "learning_rate": 6.4472591362126235e-09, "loss": 0.2264, "step": 9428 }, { "epoch": 1.96, "learning_rate": 6.416112956810631e-09, "loss": 0.1992, "step": 9429 }, { "epoch": 1.96, "learning_rate": 6.384966777408638e-09, "loss": 0.1506, "step": 9430 }, { "epoch": 1.96, "learning_rate": 6.3538205980066445e-09, "loss": 0.1481, "step": 9431 }, { "epoch": 1.96, "learning_rate": 6.3226744186046504e-09, "loss": 0.131, "step": 9432 }, { "epoch": 1.96, "learning_rate": 6.291528239202657e-09, "loss": 0.148, "step": 9433 }, { "epoch": 1.96, "learning_rate": 6.260382059800665e-09, "loss": 0.1462, "step": 9434 }, { "epoch": 1.96, "learning_rate": 6.2292358803986714e-09, "loss": 0.1304, "step": 9435 }, { "epoch": 1.96, "learning_rate": 6.198089700996677e-09, "loss": 0.1783, "step": 9436 }, { "epoch": 1.96, "learning_rate": 6.166943521594684e-09, "loss": 0.1938, "step": 9437 }, { "epoch": 1.96, "learning_rate": 6.13579734219269e-09, "loss": 0.2108, "step": 9438 }, { "epoch": 1.96, "learning_rate": 6.1046511627906975e-09, "loss": 0.1311, "step": 9439 }, { "epoch": 1.96, "learning_rate": 6.073504983388704e-09, "loss": 0.2393, "step": 9440 }, { "epoch": 1.96, "learning_rate": 6.042358803986711e-09, "loss": 0.1633, "step": 9441 }, { "epoch": 1.96, "learning_rate": 6.011212624584717e-09, "loss": 0.1938, "step": 9442 }, { "epoch": 1.96, "learning_rate": 5.9800664451827236e-09, "loss": 0.1555, "step": 9443 }, { "epoch": 1.96, "learning_rate": 5.948920265780731e-09, "loss": 0.1178, "step": 9444 }, { "epoch": 1.96, "learning_rate": 5.917774086378738e-09, "loss": 0.1564, "step": 9445 }, { "epoch": 1.96, "learning_rate": 5.886627906976744e-09, "loss": 0.155, "step": 9446 }, { "epoch": 1.96, "learning_rate": 5.8554817275747505e-09, "loss": 0.2052, "step": 9447 }, { "epoch": 1.96, "learning_rate": 5.824335548172757e-09, "loss": 0.1836, "step": 9448 }, { "epoch": 1.96, "learning_rate": 5.793189368770764e-09, "loss": 0.1665, "step": 9449 }, { "epoch": 1.96, "learning_rate": 5.762043189368771e-09, "loss": 0.2223, "step": 9450 }, { "epoch": 1.96, "learning_rate": 5.730897009966777e-09, "loss": 0.1935, "step": 9451 }, { "epoch": 1.96, "learning_rate": 5.699750830564783e-09, "loss": 0.1869, "step": 9452 }, { "epoch": 1.96, "learning_rate": 5.66860465116279e-09, "loss": 0.1978, "step": 9453 }, { "epoch": 1.96, "learning_rate": 5.6374584717607976e-09, "loss": 0.1967, "step": 9454 }, { "epoch": 1.96, "learning_rate": 5.606312292358804e-09, "loss": 0.1644, "step": 9455 }, { "epoch": 1.96, "learning_rate": 5.57516611295681e-09, "loss": 0.1903, "step": 9456 }, { "epoch": 1.96, "learning_rate": 5.544019933554817e-09, "loss": 0.1759, "step": 9457 }, { "epoch": 1.96, "learning_rate": 5.512873754152824e-09, "loss": 0.1887, "step": 9458 }, { "epoch": 1.96, "learning_rate": 5.481727574750831e-09, "loss": 0.1659, "step": 9459 }, { "epoch": 1.96, "learning_rate": 5.450581395348837e-09, "loss": 0.1943, "step": 9460 }, { "epoch": 1.96, "learning_rate": 5.419435215946844e-09, "loss": 0.1415, "step": 9461 }, { "epoch": 1.96, "learning_rate": 5.3882890365448505e-09, "loss": 0.1651, "step": 9462 }, { "epoch": 1.96, "learning_rate": 5.3571428571428564e-09, "loss": 0.166, "step": 9463 }, { "epoch": 1.97, "learning_rate": 5.325996677740863e-09, "loss": 0.1793, "step": 9464 }, { "epoch": 1.97, "learning_rate": 5.294850498338871e-09, "loss": 0.1288, "step": 9465 }, { "epoch": 1.97, "learning_rate": 5.263704318936877e-09, "loss": 0.1163, "step": 9466 }, { "epoch": 1.97, "learning_rate": 5.232558139534883e-09, "loss": 0.1366, "step": 9467 }, { "epoch": 1.97, "learning_rate": 5.20141196013289e-09, "loss": 0.1805, "step": 9468 }, { "epoch": 1.97, "learning_rate": 5.170265780730896e-09, "loss": 0.1584, "step": 9469 }, { "epoch": 1.97, "learning_rate": 5.1391196013289035e-09, "loss": 0.1489, "step": 9470 }, { "epoch": 1.97, "learning_rate": 5.10797342192691e-09, "loss": 0.145, "step": 9471 }, { "epoch": 1.97, "learning_rate": 5.076827242524917e-09, "loss": 0.2012, "step": 9472 }, { "epoch": 1.97, "learning_rate": 5.045681063122923e-09, "loss": 0.1433, "step": 9473 }, { "epoch": 1.97, "learning_rate": 5.0145348837209296e-09, "loss": 0.2195, "step": 9474 }, { "epoch": 1.97, "learning_rate": 4.983388704318937e-09, "loss": 0.1696, "step": 9475 }, { "epoch": 1.97, "learning_rate": 4.952242524916944e-09, "loss": 0.1525, "step": 9476 }, { "epoch": 1.97, "learning_rate": 4.92109634551495e-09, "loss": 0.1683, "step": 9477 }, { "epoch": 1.97, "learning_rate": 4.8899501661129565e-09, "loss": 0.1704, "step": 9478 }, { "epoch": 1.97, "learning_rate": 4.858803986710962e-09, "loss": 0.1411, "step": 9479 }, { "epoch": 1.97, "learning_rate": 4.82765780730897e-09, "loss": 0.1808, "step": 9480 }, { "epoch": 1.97, "learning_rate": 4.796511627906977e-09, "loss": 0.1614, "step": 9481 }, { "epoch": 1.97, "learning_rate": 4.765365448504983e-09, "loss": 0.1684, "step": 9482 }, { "epoch": 1.97, "learning_rate": 4.734219269102989e-09, "loss": 0.186, "step": 9483 }, { "epoch": 1.97, "learning_rate": 4.703073089700996e-09, "loss": 0.2478, "step": 9484 }, { "epoch": 1.97, "learning_rate": 4.671926910299003e-09, "loss": 0.1998, "step": 9485 }, { "epoch": 1.97, "learning_rate": 4.64078073089701e-09, "loss": 0.1958, "step": 9486 }, { "epoch": 1.97, "learning_rate": 4.609634551495016e-09, "loss": 0.1597, "step": 9487 }, { "epoch": 1.97, "learning_rate": 4.578488372093023e-09, "loss": 0.1591, "step": 9488 }, { "epoch": 1.97, "learning_rate": 4.54734219269103e-09, "loss": 0.1438, "step": 9489 }, { "epoch": 1.97, "learning_rate": 4.516196013289036e-09, "loss": 0.1683, "step": 9490 }, { "epoch": 1.97, "learning_rate": 4.485049833887043e-09, "loss": 0.1452, "step": 9491 }, { "epoch": 1.97, "learning_rate": 4.45390365448505e-09, "loss": 0.2474, "step": 9492 }, { "epoch": 1.97, "learning_rate": 4.422757475083056e-09, "loss": 0.1411, "step": 9493 }, { "epoch": 1.97, "learning_rate": 4.391611295681063e-09, "loss": 0.1239, "step": 9494 }, { "epoch": 1.97, "learning_rate": 4.360465116279069e-09, "loss": 0.1489, "step": 9495 }, { "epoch": 1.97, "learning_rate": 4.329318936877077e-09, "loss": 0.2125, "step": 9496 }, { "epoch": 1.97, "learning_rate": 4.298172757475083e-09, "loss": 0.2063, "step": 9497 }, { "epoch": 1.97, "learning_rate": 4.267026578073089e-09, "loss": 0.1733, "step": 9498 }, { "epoch": 1.97, "learning_rate": 4.235880398671096e-09, "loss": 0.2386, "step": 9499 }, { "epoch": 1.97, "learning_rate": 4.204734219269103e-09, "loss": 0.1944, "step": 9500 }, { "epoch": 1.97, "learning_rate": 4.1735880398671095e-09, "loss": 0.199, "step": 9501 }, { "epoch": 1.97, "learning_rate": 4.142441860465116e-09, "loss": 0.2028, "step": 9502 }, { "epoch": 1.97, "learning_rate": 4.111295681063123e-09, "loss": 0.2018, "step": 9503 }, { "epoch": 1.97, "learning_rate": 4.08014950166113e-09, "loss": 0.1775, "step": 9504 }, { "epoch": 1.97, "learning_rate": 4.0490033222591364e-09, "loss": 0.2108, "step": 9505 }, { "epoch": 1.97, "learning_rate": 4.017857142857142e-09, "loss": 0.201, "step": 9506 }, { "epoch": 1.97, "learning_rate": 3.986710963455149e-09, "loss": 0.1848, "step": 9507 }, { "epoch": 1.97, "learning_rate": 3.955564784053156e-09, "loss": 0.1444, "step": 9508 }, { "epoch": 1.97, "learning_rate": 3.9244186046511625e-09, "loss": 0.1809, "step": 9509 }, { "epoch": 1.97, "learning_rate": 3.893272425249169e-09, "loss": 0.1713, "step": 9510 }, { "epoch": 1.97, "learning_rate": 3.862126245847176e-09, "loss": 0.2055, "step": 9511 }, { "epoch": 1.98, "learning_rate": 3.830980066445183e-09, "loss": 0.2784, "step": 9512 }, { "epoch": 1.98, "learning_rate": 3.799833887043189e-09, "loss": 0.2089, "step": 9513 }, { "epoch": 1.98, "learning_rate": 3.768687707641196e-09, "loss": 0.2056, "step": 9514 }, { "epoch": 1.98, "learning_rate": 3.737541528239203e-09, "loss": 0.1667, "step": 9515 }, { "epoch": 1.98, "learning_rate": 3.7063953488372087e-09, "loss": 0.1292, "step": 9516 }, { "epoch": 1.98, "learning_rate": 3.675249169435216e-09, "loss": 0.1362, "step": 9517 }, { "epoch": 1.98, "learning_rate": 3.644102990033222e-09, "loss": 0.1795, "step": 9518 }, { "epoch": 1.98, "learning_rate": 3.6129568106312293e-09, "loss": 0.1975, "step": 9519 }, { "epoch": 1.98, "learning_rate": 3.5818106312292357e-09, "loss": 0.1405, "step": 9520 }, { "epoch": 1.98, "learning_rate": 3.5506644518272424e-09, "loss": 0.177, "step": 9521 }, { "epoch": 1.98, "learning_rate": 3.519518272425249e-09, "loss": 0.1179, "step": 9522 }, { "epoch": 1.98, "learning_rate": 3.4883720930232554e-09, "loss": 0.1682, "step": 9523 }, { "epoch": 1.98, "learning_rate": 3.457225913621262e-09, "loss": 0.1886, "step": 9524 }, { "epoch": 1.98, "learning_rate": 3.426079734219269e-09, "loss": 0.1298, "step": 9525 }, { "epoch": 1.98, "learning_rate": 3.3949335548172756e-09, "loss": 0.2017, "step": 9526 }, { "epoch": 1.98, "learning_rate": 3.3637873754152823e-09, "loss": 0.1666, "step": 9527 }, { "epoch": 1.98, "learning_rate": 3.332641196013289e-09, "loss": 0.1822, "step": 9528 }, { "epoch": 1.98, "learning_rate": 3.3014950166112954e-09, "loss": 0.1397, "step": 9529 }, { "epoch": 1.98, "learning_rate": 3.270348837209302e-09, "loss": 0.1756, "step": 9530 }, { "epoch": 1.98, "learning_rate": 3.239202657807309e-09, "loss": 0.146, "step": 9531 }, { "epoch": 1.98, "learning_rate": 3.2080564784053155e-09, "loss": 0.1777, "step": 9532 }, { "epoch": 1.98, "learning_rate": 3.1769102990033223e-09, "loss": 0.1665, "step": 9533 }, { "epoch": 1.98, "learning_rate": 3.1457641196013286e-09, "loss": 0.1699, "step": 9534 }, { "epoch": 1.98, "learning_rate": 3.1146179401993357e-09, "loss": 0.291, "step": 9535 }, { "epoch": 1.98, "learning_rate": 3.083471760797342e-09, "loss": 0.1836, "step": 9536 }, { "epoch": 1.98, "learning_rate": 3.0523255813953487e-09, "loss": 0.1575, "step": 9537 }, { "epoch": 1.98, "learning_rate": 3.0211794019933555e-09, "loss": 0.2023, "step": 9538 }, { "epoch": 1.98, "learning_rate": 2.9900332225913618e-09, "loss": 0.1554, "step": 9539 }, { "epoch": 1.98, "learning_rate": 2.958887043189369e-09, "loss": 0.1743, "step": 9540 }, { "epoch": 1.98, "learning_rate": 2.9277408637873752e-09, "loss": 0.136, "step": 9541 }, { "epoch": 1.98, "learning_rate": 2.896594684385382e-09, "loss": 0.2223, "step": 9542 }, { "epoch": 1.98, "learning_rate": 2.8654485049833887e-09, "loss": 0.206, "step": 9543 }, { "epoch": 1.98, "learning_rate": 2.834302325581395e-09, "loss": 0.1315, "step": 9544 }, { "epoch": 1.98, "learning_rate": 2.803156146179402e-09, "loss": 0.1621, "step": 9545 }, { "epoch": 1.98, "learning_rate": 2.7720099667774085e-09, "loss": 0.1239, "step": 9546 }, { "epoch": 1.98, "learning_rate": 2.7408637873754156e-09, "loss": 0.2196, "step": 9547 }, { "epoch": 1.98, "learning_rate": 2.709717607973422e-09, "loss": 0.1835, "step": 9548 }, { "epoch": 1.98, "learning_rate": 2.6785714285714282e-09, "loss": 0.2132, "step": 9549 }, { "epoch": 1.98, "learning_rate": 2.6474252491694354e-09, "loss": 0.2247, "step": 9550 }, { "epoch": 1.98, "learning_rate": 2.6162790697674417e-09, "loss": 0.1376, "step": 9551 }, { "epoch": 1.98, "learning_rate": 2.585132890365448e-09, "loss": 0.183, "step": 9552 }, { "epoch": 1.98, "learning_rate": 2.553986710963455e-09, "loss": 0.1935, "step": 9553 }, { "epoch": 1.98, "learning_rate": 2.5228405315614614e-09, "loss": 0.1945, "step": 9554 }, { "epoch": 1.98, "learning_rate": 2.4916943521594686e-09, "loss": 0.1403, "step": 9555 }, { "epoch": 1.98, "learning_rate": 2.460548172757475e-09, "loss": 0.1762, "step": 9556 }, { "epoch": 1.98, "learning_rate": 2.429401993355481e-09, "loss": 0.1575, "step": 9557 }, { "epoch": 1.98, "learning_rate": 2.3982558139534883e-09, "loss": 0.1804, "step": 9558 }, { "epoch": 1.98, "learning_rate": 2.3671096345514946e-09, "loss": 0.1356, "step": 9559 }, { "epoch": 1.99, "learning_rate": 2.3359634551495014e-09, "loss": 0.1856, "step": 9560 }, { "epoch": 1.99, "learning_rate": 2.304817275747508e-09, "loss": 0.1801, "step": 9561 }, { "epoch": 1.99, "learning_rate": 2.273671096345515e-09, "loss": 0.1404, "step": 9562 }, { "epoch": 1.99, "learning_rate": 2.2425249169435215e-09, "loss": 0.216, "step": 9563 }, { "epoch": 1.99, "learning_rate": 2.211378737541528e-09, "loss": 0.1411, "step": 9564 }, { "epoch": 1.99, "learning_rate": 2.1802325581395346e-09, "loss": 0.1398, "step": 9565 }, { "epoch": 1.99, "learning_rate": 2.1490863787375413e-09, "loss": 0.1505, "step": 9566 }, { "epoch": 1.99, "learning_rate": 2.117940199335548e-09, "loss": 0.2489, "step": 9567 }, { "epoch": 1.99, "learning_rate": 2.0867940199335548e-09, "loss": 0.1641, "step": 9568 }, { "epoch": 1.99, "learning_rate": 2.0556478405315615e-09, "loss": 0.2132, "step": 9569 }, { "epoch": 1.99, "learning_rate": 2.0245016611295682e-09, "loss": 0.1494, "step": 9570 }, { "epoch": 1.99, "learning_rate": 1.9933554817275745e-09, "loss": 0.2196, "step": 9571 }, { "epoch": 1.99, "learning_rate": 1.9622093023255812e-09, "loss": 0.2445, "step": 9572 }, { "epoch": 1.99, "learning_rate": 1.931063122923588e-09, "loss": 0.1913, "step": 9573 }, { "epoch": 1.99, "learning_rate": 1.8999169435215947e-09, "loss": 0.1871, "step": 9574 }, { "epoch": 1.99, "learning_rate": 1.8687707641196014e-09, "loss": 0.1453, "step": 9575 }, { "epoch": 1.99, "learning_rate": 1.837624584717608e-09, "loss": 0.1797, "step": 9576 }, { "epoch": 1.99, "learning_rate": 1.8064784053156147e-09, "loss": 0.3132, "step": 9577 }, { "epoch": 1.99, "learning_rate": 1.7753322259136212e-09, "loss": 0.1893, "step": 9578 }, { "epoch": 1.99, "learning_rate": 1.7441860465116277e-09, "loss": 0.1602, "step": 9579 }, { "epoch": 1.99, "learning_rate": 1.7130398671096344e-09, "loss": 0.1982, "step": 9580 }, { "epoch": 1.99, "learning_rate": 1.6818936877076412e-09, "loss": 0.2657, "step": 9581 }, { "epoch": 1.99, "learning_rate": 1.6507475083056477e-09, "loss": 0.2231, "step": 9582 }, { "epoch": 1.99, "learning_rate": 1.6196013289036544e-09, "loss": 0.1666, "step": 9583 }, { "epoch": 1.99, "learning_rate": 1.5884551495016611e-09, "loss": 0.2486, "step": 9584 }, { "epoch": 1.99, "learning_rate": 1.5573089700996679e-09, "loss": 0.1224, "step": 9585 }, { "epoch": 1.99, "learning_rate": 1.5261627906976744e-09, "loss": 0.2282, "step": 9586 }, { "epoch": 1.99, "learning_rate": 1.4950166112956809e-09, "loss": 0.159, "step": 9587 }, { "epoch": 1.99, "learning_rate": 1.4638704318936876e-09, "loss": 0.1425, "step": 9588 }, { "epoch": 1.99, "learning_rate": 1.4327242524916943e-09, "loss": 0.2023, "step": 9589 }, { "epoch": 1.99, "learning_rate": 1.401578073089701e-09, "loss": 0.1868, "step": 9590 }, { "epoch": 1.99, "learning_rate": 1.3704318936877078e-09, "loss": 0.1501, "step": 9591 }, { "epoch": 1.99, "learning_rate": 1.3392857142857141e-09, "loss": 0.1728, "step": 9592 }, { "epoch": 1.99, "learning_rate": 1.3081395348837208e-09, "loss": 0.1572, "step": 9593 }, { "epoch": 1.99, "learning_rate": 1.2769933554817276e-09, "loss": 0.2053, "step": 9594 }, { "epoch": 1.99, "learning_rate": 1.2458471760797343e-09, "loss": 0.1708, "step": 9595 }, { "epoch": 1.99, "learning_rate": 1.2147009966777406e-09, "loss": 0.2857, "step": 9596 }, { "epoch": 1.99, "learning_rate": 1.1835548172757473e-09, "loss": 0.1527, "step": 9597 }, { "epoch": 1.99, "learning_rate": 1.152408637873754e-09, "loss": 0.1585, "step": 9598 }, { "epoch": 1.99, "learning_rate": 1.1212624584717608e-09, "loss": 0.1632, "step": 9599 }, { "epoch": 1.99, "learning_rate": 1.0901162790697673e-09, "loss": 0.236, "step": 9600 }, { "epoch": 1.99, "learning_rate": 1.058970099667774e-09, "loss": 0.1966, "step": 9601 }, { "epoch": 1.99, "learning_rate": 1.0278239202657807e-09, "loss": 0.1323, "step": 9602 }, { "epoch": 1.99, "learning_rate": 9.966777408637873e-10, "loss": 0.173, "step": 9603 }, { "epoch": 1.99, "learning_rate": 9.65531561461794e-10, "loss": 0.1853, "step": 9604 }, { "epoch": 1.99, "learning_rate": 9.343853820598007e-10, "loss": 0.2135, "step": 9605 }, { "epoch": 1.99, "learning_rate": 9.032392026578073e-10, "loss": 0.1948, "step": 9606 }, { "epoch": 1.99, "learning_rate": 8.720930232558139e-10, "loss": 0.2382, "step": 9607 }, { "epoch": 2.0, "learning_rate": 8.409468438538206e-10, "loss": 0.1807, "step": 9608 }, { "epoch": 2.0, "learning_rate": 8.098006644518272e-10, "loss": 0.1397, "step": 9609 }, { "epoch": 2.0, "learning_rate": 7.786544850498339e-10, "loss": 0.1682, "step": 9610 }, { "epoch": 2.0, "learning_rate": 7.475083056478404e-10, "loss": 0.1112, "step": 9611 }, { "epoch": 2.0, "learning_rate": 7.163621262458472e-10, "loss": 0.2198, "step": 9612 }, { "epoch": 2.0, "learning_rate": 6.852159468438539e-10, "loss": 0.2179, "step": 9613 }, { "epoch": 2.0, "learning_rate": 6.540697674418604e-10, "loss": 0.1864, "step": 9614 }, { "epoch": 2.0, "learning_rate": 6.229235880398671e-10, "loss": 0.1894, "step": 9615 }, { "epoch": 2.0, "learning_rate": 5.917774086378737e-10, "loss": 0.1743, "step": 9616 }, { "epoch": 2.0, "learning_rate": 5.606312292358804e-10, "loss": 0.1691, "step": 9617 }, { "epoch": 2.0, "learning_rate": 5.29485049833887e-10, "loss": 0.1873, "step": 9618 }, { "epoch": 2.0, "learning_rate": 4.983388704318936e-10, "loss": 0.1208, "step": 9619 }, { "epoch": 2.0, "learning_rate": 4.671926910299004e-10, "loss": 0.1574, "step": 9620 }, { "epoch": 2.0, "learning_rate": 4.3604651162790693e-10, "loss": 0.2028, "step": 9621 }, { "epoch": 2.0, "learning_rate": 4.049003322259136e-10, "loss": 0.2317, "step": 9622 }, { "epoch": 2.0, "learning_rate": 3.737541528239202e-10, "loss": 0.2201, "step": 9623 }, { "epoch": 2.0, "learning_rate": 3.4260797342192695e-10, "loss": 0.2125, "step": 9624 }, { "epoch": 2.0, "learning_rate": 3.1146179401993357e-10, "loss": 0.1757, "step": 9625 }, { "epoch": 2.0, "learning_rate": 2.803156146179402e-10, "loss": 0.1922, "step": 9626 }, { "epoch": 2.0, "learning_rate": 2.491694352159468e-10, "loss": 0.1844, "step": 9627 }, { "epoch": 2.0, "learning_rate": 2.1802325581395346e-10, "loss": 0.1481, "step": 9628 }, { "epoch": 2.0, "learning_rate": 1.868770764119601e-10, "loss": 0.1672, "step": 9629 }, { "epoch": 2.0, "learning_rate": 1.5573089700996679e-10, "loss": 0.1933, "step": 9630 }, { "epoch": 2.0, "learning_rate": 1.245847176079734e-10, "loss": 0.1387, "step": 9631 }, { "epoch": 2.0, "learning_rate": 9.343853820598006e-11, "loss": 0.1637, "step": 9632 }, { "epoch": 2.0, "eval_loss": 0.3458508849143982, "eval_runtime": 3319.4214, "eval_samples_per_second": 4.802, "eval_steps_per_second": 0.1, "eval_wer": 0.2915172650916057, "step": 9632 }, { "epoch": 2.0, "step": 9632, "total_flos": 4.232677693574722e+20, "train_loss": 0.1840909074456663, "train_runtime": 239847.4984, "train_samples_per_second": 3.855, "train_steps_per_second": 0.04 } ], "max_steps": 9632, "num_train_epochs": 2, "total_flos": 4.232677693574722e+20, "trial_name": null, "trial_params": null }