{ "best_metric": null, "best_model_checkpoint": null, "epoch": 3.0, "global_step": 12981, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 4.99961482166243e-05, "loss": 0.0, "step": 1 }, { "epoch": 0.0, "learning_rate": 4.99922964332486e-05, "loss": 0.0, "step": 2 }, { "epoch": 0.0, "learning_rate": 4.9988444649872894e-05, "loss": 0.0, "step": 3 }, { "epoch": 0.0, "learning_rate": 4.9984592866497196e-05, "loss": 0.0, "step": 4 }, { "epoch": 0.0, "learning_rate": 4.9980741083121484e-05, "loss": 0.1085, "step": 5 }, { "epoch": 0.0, "learning_rate": 4.9976889299745786e-05, "loss": 0.0001, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.997303751637008e-05, "loss": 0.0, "step": 7 }, { "epoch": 0.0, "learning_rate": 4.9969185732994376e-05, "loss": 0.0, "step": 8 }, { "epoch": 0.0, "learning_rate": 4.996533394961868e-05, "loss": 0.0001, "step": 9 }, { "epoch": 0.0, "learning_rate": 4.996148216624297e-05, "loss": 0.0, "step": 10 }, { "epoch": 0.0, "learning_rate": 4.9957630382867274e-05, "loss": 0.0, "step": 11 }, { "epoch": 0.0, "learning_rate": 4.995377859949157e-05, "loss": 0.0, "step": 12 }, { "epoch": 0.0, "learning_rate": 4.9949926816115864e-05, "loss": 0.0, "step": 13 }, { "epoch": 0.0, "learning_rate": 4.994607503274016e-05, "loss": 0.0, "step": 14 }, { "epoch": 0.0, "learning_rate": 4.9942223249364454e-05, "loss": 0.0, "step": 15 }, { "epoch": 0.0, "learning_rate": 4.9938371465988756e-05, "loss": 0.0, "step": 16 }, { "epoch": 0.0, "learning_rate": 4.993451968261305e-05, "loss": 0.0, "step": 17 }, { "epoch": 0.0, "learning_rate": 4.993066789923735e-05, "loss": 0.0, "step": 18 }, { "epoch": 0.0, "learning_rate": 4.992681611586165e-05, "loss": 0.0, "step": 19 }, { "epoch": 0.0, "learning_rate": 4.992296433248594e-05, "loss": 0.0, "step": 20 }, { "epoch": 0.0, "learning_rate": 4.9919112549110244e-05, "loss": 0.2631, "step": 21 }, { "epoch": 0.01, "learning_rate": 4.991526076573454e-05, "loss": 0.0, "step": 22 }, { "epoch": 0.01, "learning_rate": 4.9911408982358834e-05, "loss": 0.0, "step": 23 }, { "epoch": 0.01, "learning_rate": 4.990755719898313e-05, "loss": 0.0, "step": 24 }, { "epoch": 0.01, "learning_rate": 4.9903705415607424e-05, "loss": 0.0, "step": 25 }, { "epoch": 0.01, "learning_rate": 4.9899853632231725e-05, "loss": 0.0, "step": 26 }, { "epoch": 0.01, "learning_rate": 4.989600184885602e-05, "loss": 0.0, "step": 27 }, { "epoch": 0.01, "learning_rate": 4.989215006548032e-05, "loss": 0.0, "step": 28 }, { "epoch": 0.01, "learning_rate": 4.988829828210462e-05, "loss": 0.0, "step": 29 }, { "epoch": 0.01, "learning_rate": 4.988444649872891e-05, "loss": 0.0, "step": 30 }, { "epoch": 0.01, "learning_rate": 4.9880594715353214e-05, "loss": 0.0, "step": 31 }, { "epoch": 0.01, "learning_rate": 4.987674293197751e-05, "loss": 0.0, "step": 32 }, { "epoch": 0.01, "learning_rate": 4.9872891148601804e-05, "loss": 0.0, "step": 33 }, { "epoch": 0.01, "learning_rate": 4.98690393652261e-05, "loss": 0.0, "step": 34 }, { "epoch": 0.01, "learning_rate": 4.98651875818504e-05, "loss": 0.0, "step": 35 }, { "epoch": 0.01, "learning_rate": 4.9861335798474695e-05, "loss": 0.1451, "step": 36 }, { "epoch": 0.01, "learning_rate": 4.985748401509899e-05, "loss": 0.001, "step": 37 }, { "epoch": 0.01, "learning_rate": 4.985363223172329e-05, "loss": 0.0, "step": 38 }, { "epoch": 0.01, "learning_rate": 4.984978044834759e-05, "loss": 0.0, "step": 39 }, { "epoch": 0.01, "learning_rate": 4.984592866497189e-05, "loss": 0.0, "step": 40 }, { "epoch": 0.01, "learning_rate": 4.9842076881596184e-05, "loss": 0.0, "step": 41 }, { "epoch": 0.01, "learning_rate": 4.983822509822048e-05, "loss": 0.0, "step": 42 }, { "epoch": 0.01, "learning_rate": 4.9834373314844774e-05, "loss": 0.166, "step": 43 }, { "epoch": 0.01, "learning_rate": 4.983052153146907e-05, "loss": 0.0, "step": 44 }, { "epoch": 0.01, "learning_rate": 4.982666974809337e-05, "loss": 0.0, "step": 45 }, { "epoch": 0.01, "learning_rate": 4.9822817964717665e-05, "loss": 0.0, "step": 46 }, { "epoch": 0.01, "learning_rate": 4.981896618134197e-05, "loss": 0.0, "step": 47 }, { "epoch": 0.01, "learning_rate": 4.981511439796626e-05, "loss": 0.0, "step": 48 }, { "epoch": 0.01, "learning_rate": 4.981126261459056e-05, "loss": 0.0001, "step": 49 }, { "epoch": 0.01, "learning_rate": 4.980741083121486e-05, "loss": 0.0, "step": 50 }, { "epoch": 0.01, "learning_rate": 4.9803559047839154e-05, "loss": 0.0, "step": 51 }, { "epoch": 0.01, "learning_rate": 4.979970726446345e-05, "loss": 0.0, "step": 52 }, { "epoch": 0.01, "learning_rate": 4.9795855481087743e-05, "loss": 0.0, "step": 53 }, { "epoch": 0.01, "learning_rate": 4.9792003697712045e-05, "loss": 0.1493, "step": 54 }, { "epoch": 0.01, "learning_rate": 4.978815191433634e-05, "loss": 0.0, "step": 55 }, { "epoch": 0.01, "learning_rate": 4.9784300130960635e-05, "loss": 0.0, "step": 56 }, { "epoch": 0.01, "learning_rate": 4.978044834758494e-05, "loss": 0.0, "step": 57 }, { "epoch": 0.01, "learning_rate": 4.977659656420923e-05, "loss": 0.0, "step": 58 }, { "epoch": 0.01, "learning_rate": 4.977274478083353e-05, "loss": 0.0, "step": 59 }, { "epoch": 0.01, "learning_rate": 4.976889299745783e-05, "loss": 0.0, "step": 60 }, { "epoch": 0.01, "learning_rate": 4.976504121408212e-05, "loss": 0.0, "step": 61 }, { "epoch": 0.01, "learning_rate": 4.976118943070642e-05, "loss": 0.0, "step": 62 }, { "epoch": 0.01, "learning_rate": 4.975733764733071e-05, "loss": 0.0, "step": 63 }, { "epoch": 0.01, "learning_rate": 4.9753485863955015e-05, "loss": 0.0, "step": 64 }, { "epoch": 0.02, "learning_rate": 4.974963408057931e-05, "loss": 0.079, "step": 65 }, { "epoch": 0.02, "learning_rate": 4.9745782297203605e-05, "loss": 0.0, "step": 66 }, { "epoch": 0.02, "learning_rate": 4.974193051382791e-05, "loss": 0.0, "step": 67 }, { "epoch": 0.02, "learning_rate": 4.97380787304522e-05, "loss": 0.0, "step": 68 }, { "epoch": 0.02, "learning_rate": 4.9734226947076503e-05, "loss": 0.0067, "step": 69 }, { "epoch": 0.02, "learning_rate": 4.973037516370079e-05, "loss": 0.0, "step": 70 }, { "epoch": 0.02, "learning_rate": 4.972652338032509e-05, "loss": 0.0569, "step": 71 }, { "epoch": 0.02, "learning_rate": 4.972267159694939e-05, "loss": 0.0, "step": 72 }, { "epoch": 0.02, "learning_rate": 4.971881981357368e-05, "loss": 0.0, "step": 73 }, { "epoch": 0.02, "learning_rate": 4.9714968030197985e-05, "loss": 0.0, "step": 74 }, { "epoch": 0.02, "learning_rate": 4.971111624682228e-05, "loss": 0.0, "step": 75 }, { "epoch": 0.02, "learning_rate": 4.970726446344658e-05, "loss": 0.0, "step": 76 }, { "epoch": 0.02, "learning_rate": 4.9703412680070877e-05, "loss": 0.0, "step": 77 }, { "epoch": 0.02, "learning_rate": 4.969956089669517e-05, "loss": 0.0, "step": 78 }, { "epoch": 0.02, "learning_rate": 4.969570911331947e-05, "loss": 0.0, "step": 79 }, { "epoch": 0.02, "learning_rate": 4.969185732994376e-05, "loss": 0.0, "step": 80 }, { "epoch": 0.02, "learning_rate": 4.968800554656806e-05, "loss": 0.0, "step": 81 }, { "epoch": 0.02, "learning_rate": 4.968415376319236e-05, "loss": 0.0002, "step": 82 }, { "epoch": 0.02, "learning_rate": 4.968030197981666e-05, "loss": 0.0, "step": 83 }, { "epoch": 0.02, "learning_rate": 4.9676450196440955e-05, "loss": 0.0, "step": 84 }, { "epoch": 0.02, "learning_rate": 4.967259841306525e-05, "loss": 0.0, "step": 85 }, { "epoch": 0.02, "learning_rate": 4.966874662968955e-05, "loss": 0.0, "step": 86 }, { "epoch": 0.02, "learning_rate": 4.9664894846313846e-05, "loss": 0.0001, "step": 87 }, { "epoch": 0.02, "learning_rate": 4.966104306293815e-05, "loss": 0.0, "step": 88 }, { "epoch": 0.02, "learning_rate": 4.9657191279562436e-05, "loss": 0.0, "step": 89 }, { "epoch": 0.02, "learning_rate": 4.965333949618674e-05, "loss": 0.0001, "step": 90 }, { "epoch": 0.02, "learning_rate": 4.964948771281103e-05, "loss": 0.0, "step": 91 }, { "epoch": 0.02, "learning_rate": 4.964563592943533e-05, "loss": 0.0, "step": 92 }, { "epoch": 0.02, "learning_rate": 4.964178414605963e-05, "loss": 0.0, "step": 93 }, { "epoch": 0.02, "learning_rate": 4.9637932362683925e-05, "loss": 0.0, "step": 94 }, { "epoch": 0.02, "learning_rate": 4.9634080579308226e-05, "loss": 0.0, "step": 95 }, { "epoch": 0.02, "learning_rate": 4.963022879593252e-05, "loss": 0.0, "step": 96 }, { "epoch": 0.02, "learning_rate": 4.9626377012556816e-05, "loss": 0.0, "step": 97 }, { "epoch": 0.02, "learning_rate": 4.962252522918111e-05, "loss": 0.0001, "step": 98 }, { "epoch": 0.02, "learning_rate": 4.9618673445805406e-05, "loss": 0.0001, "step": 99 }, { "epoch": 0.02, "learning_rate": 4.961482166242971e-05, "loss": 0.0, "step": 100 }, { "epoch": 0.02, "learning_rate": 4.9610969879054e-05, "loss": 0.0, "step": 101 }, { "epoch": 0.02, "learning_rate": 4.96071180956783e-05, "loss": 0.0001, "step": 102 }, { "epoch": 0.02, "learning_rate": 4.96032663123026e-05, "loss": 0.0, "step": 103 }, { "epoch": 0.02, "learning_rate": 4.9599414528926895e-05, "loss": 0.0002, "step": 104 }, { "epoch": 0.02, "learning_rate": 4.9595562745551196e-05, "loss": 0.0, "step": 105 }, { "epoch": 0.02, "learning_rate": 4.959171096217549e-05, "loss": 0.0, "step": 106 }, { "epoch": 0.02, "learning_rate": 4.9587859178799786e-05, "loss": 0.0, "step": 107 }, { "epoch": 0.02, "learning_rate": 4.958400739542408e-05, "loss": 0.0002, "step": 108 }, { "epoch": 0.03, "learning_rate": 4.9580155612048376e-05, "loss": 0.0, "step": 109 }, { "epoch": 0.03, "learning_rate": 4.957630382867268e-05, "loss": 0.0, "step": 110 }, { "epoch": 0.03, "learning_rate": 4.957245204529697e-05, "loss": 0.0, "step": 111 }, { "epoch": 0.03, "learning_rate": 4.9568600261921275e-05, "loss": 0.0, "step": 112 }, { "epoch": 0.03, "learning_rate": 4.956474847854557e-05, "loss": 0.0003, "step": 113 }, { "epoch": 0.03, "learning_rate": 4.9560896695169864e-05, "loss": 0.0, "step": 114 }, { "epoch": 0.03, "learning_rate": 4.9557044911794166e-05, "loss": 0.0, "step": 115 }, { "epoch": 0.03, "learning_rate": 4.955319312841846e-05, "loss": 0.0, "step": 116 }, { "epoch": 0.03, "learning_rate": 4.9549341345042756e-05, "loss": 0.0, "step": 117 }, { "epoch": 0.03, "learning_rate": 4.954548956166705e-05, "loss": 0.0, "step": 118 }, { "epoch": 0.03, "learning_rate": 4.954163777829135e-05, "loss": 0.0, "step": 119 }, { "epoch": 0.03, "learning_rate": 4.953778599491565e-05, "loss": 0.0, "step": 120 }, { "epoch": 0.03, "learning_rate": 4.953393421153994e-05, "loss": 0.0003, "step": 121 }, { "epoch": 0.03, "learning_rate": 4.9530082428164244e-05, "loss": 0.0, "step": 122 }, { "epoch": 0.03, "learning_rate": 4.952623064478854e-05, "loss": 0.0, "step": 123 }, { "epoch": 0.03, "learning_rate": 4.952237886141284e-05, "loss": 0.0, "step": 124 }, { "epoch": 0.03, "learning_rate": 4.9518527078037136e-05, "loss": 0.0, "step": 125 }, { "epoch": 0.03, "learning_rate": 4.951467529466143e-05, "loss": 0.0, "step": 126 }, { "epoch": 0.03, "learning_rate": 4.9510823511285726e-05, "loss": 0.0, "step": 127 }, { "epoch": 0.03, "learning_rate": 4.950697172791002e-05, "loss": 0.0002, "step": 128 }, { "epoch": 0.03, "learning_rate": 4.950311994453432e-05, "loss": 0.0, "step": 129 }, { "epoch": 0.03, "learning_rate": 4.949926816115862e-05, "loss": 0.0, "step": 130 }, { "epoch": 0.03, "learning_rate": 4.949541637778292e-05, "loss": 0.0, "step": 131 }, { "epoch": 0.03, "learning_rate": 4.9491564594407214e-05, "loss": 0.0, "step": 132 }, { "epoch": 0.03, "learning_rate": 4.948771281103151e-05, "loss": 0.0, "step": 133 }, { "epoch": 0.03, "learning_rate": 4.948386102765581e-05, "loss": 0.0, "step": 134 }, { "epoch": 0.03, "learning_rate": 4.9480009244280106e-05, "loss": 0.0001, "step": 135 }, { "epoch": 0.03, "learning_rate": 4.94761574609044e-05, "loss": 0.0, "step": 136 }, { "epoch": 0.03, "learning_rate": 4.9472305677528696e-05, "loss": 0.0, "step": 137 }, { "epoch": 0.03, "learning_rate": 4.946845389415299e-05, "loss": 0.0, "step": 138 }, { "epoch": 0.03, "learning_rate": 4.946460211077729e-05, "loss": 0.0, "step": 139 }, { "epoch": 0.03, "learning_rate": 4.946075032740159e-05, "loss": 0.0, "step": 140 }, { "epoch": 0.03, "learning_rate": 4.945689854402589e-05, "loss": 0.0, "step": 141 }, { "epoch": 0.03, "learning_rate": 4.9453046760650184e-05, "loss": 0.0008, "step": 142 }, { "epoch": 0.03, "learning_rate": 4.944919497727448e-05, "loss": 0.0, "step": 143 }, { "epoch": 0.03, "learning_rate": 4.944534319389878e-05, "loss": 0.0, "step": 144 }, { "epoch": 0.03, "learning_rate": 4.944149141052307e-05, "loss": 0.0, "step": 145 }, { "epoch": 0.03, "learning_rate": 4.943763962714737e-05, "loss": 0.0, "step": 146 }, { "epoch": 0.03, "learning_rate": 4.9433787843771666e-05, "loss": 0.0, "step": 147 }, { "epoch": 0.03, "learning_rate": 4.942993606039597e-05, "loss": 0.0, "step": 148 }, { "epoch": 0.03, "learning_rate": 4.942608427702026e-05, "loss": 0.0, "step": 149 }, { "epoch": 0.03, "learning_rate": 4.942223249364456e-05, "loss": 0.0, "step": 150 }, { "epoch": 0.03, "learning_rate": 4.941838071026886e-05, "loss": 0.0, "step": 151 }, { "epoch": 0.04, "learning_rate": 4.9414528926893154e-05, "loss": 0.0, "step": 152 }, { "epoch": 0.04, "learning_rate": 4.9410677143517456e-05, "loss": 0.0, "step": 153 }, { "epoch": 0.04, "learning_rate": 4.9406825360141744e-05, "loss": 0.0, "step": 154 }, { "epoch": 0.04, "learning_rate": 4.9402973576766046e-05, "loss": 0.0, "step": 155 }, { "epoch": 0.04, "learning_rate": 4.939912179339034e-05, "loss": 0.0001, "step": 156 }, { "epoch": 0.04, "learning_rate": 4.9395270010014636e-05, "loss": 0.0, "step": 157 }, { "epoch": 0.04, "learning_rate": 4.939141822663894e-05, "loss": 0.0, "step": 158 }, { "epoch": 0.04, "learning_rate": 4.938756644326323e-05, "loss": 0.0, "step": 159 }, { "epoch": 0.04, "learning_rate": 4.9383714659887534e-05, "loss": 0.0, "step": 160 }, { "epoch": 0.04, "learning_rate": 4.937986287651183e-05, "loss": 0.0, "step": 161 }, { "epoch": 0.04, "learning_rate": 4.9376011093136124e-05, "loss": 0.0, "step": 162 }, { "epoch": 0.04, "learning_rate": 4.9372159309760426e-05, "loss": 0.0, "step": 163 }, { "epoch": 0.04, "learning_rate": 4.9368307526384714e-05, "loss": 0.0016, "step": 164 }, { "epoch": 0.04, "learning_rate": 4.9364455743009016e-05, "loss": 0.0, "step": 165 }, { "epoch": 0.04, "learning_rate": 4.936060395963331e-05, "loss": 0.0, "step": 166 }, { "epoch": 0.04, "learning_rate": 4.935675217625761e-05, "loss": 0.0947, "step": 167 }, { "epoch": 0.04, "learning_rate": 4.935290039288191e-05, "loss": 0.0, "step": 168 }, { "epoch": 0.04, "learning_rate": 4.93490486095062e-05, "loss": 0.0, "step": 169 }, { "epoch": 0.04, "learning_rate": 4.9345196826130504e-05, "loss": 0.0005, "step": 170 }, { "epoch": 0.04, "learning_rate": 4.93413450427548e-05, "loss": 0.0, "step": 171 }, { "epoch": 0.04, "learning_rate": 4.9337493259379094e-05, "loss": 0.0009, "step": 172 }, { "epoch": 0.04, "learning_rate": 4.933364147600339e-05, "loss": 0.0, "step": 173 }, { "epoch": 0.04, "learning_rate": 4.9329789692627684e-05, "loss": 0.0, "step": 174 }, { "epoch": 0.04, "learning_rate": 4.9325937909251986e-05, "loss": 0.0001, "step": 175 }, { "epoch": 0.04, "learning_rate": 4.932208612587628e-05, "loss": 0.0, "step": 176 }, { "epoch": 0.04, "learning_rate": 4.931823434250058e-05, "loss": 0.0001, "step": 177 }, { "epoch": 0.04, "learning_rate": 4.931438255912488e-05, "loss": 0.0, "step": 178 }, { "epoch": 0.04, "learning_rate": 4.931053077574917e-05, "loss": 0.0, "step": 179 }, { "epoch": 0.04, "learning_rate": 4.9306678992373474e-05, "loss": 0.0, "step": 180 }, { "epoch": 0.04, "learning_rate": 4.930282720899777e-05, "loss": 0.0, "step": 181 }, { "epoch": 0.04, "learning_rate": 4.9298975425622064e-05, "loss": 0.0, "step": 182 }, { "epoch": 0.04, "learning_rate": 4.929512364224636e-05, "loss": 0.0, "step": 183 }, { "epoch": 0.04, "learning_rate": 4.929127185887066e-05, "loss": 0.0004, "step": 184 }, { "epoch": 0.04, "learning_rate": 4.9287420075494955e-05, "loss": 0.0, "step": 185 }, { "epoch": 0.04, "learning_rate": 4.928356829211925e-05, "loss": 0.0, "step": 186 }, { "epoch": 0.04, "learning_rate": 4.927971650874355e-05, "loss": 0.0, "step": 187 }, { "epoch": 0.04, "learning_rate": 4.927586472536785e-05, "loss": 0.0, "step": 188 }, { "epoch": 0.04, "learning_rate": 4.927201294199215e-05, "loss": 0.0, "step": 189 }, { "epoch": 0.04, "learning_rate": 4.9268161158616444e-05, "loss": 0.0, "step": 190 }, { "epoch": 0.04, "learning_rate": 4.926430937524074e-05, "loss": 0.0, "step": 191 }, { "epoch": 0.04, "learning_rate": 4.9260457591865034e-05, "loss": 0.0, "step": 192 }, { "epoch": 0.04, "learning_rate": 4.925660580848933e-05, "loss": 0.0, "step": 193 }, { "epoch": 0.04, "learning_rate": 4.925275402511363e-05, "loss": 0.0, "step": 194 }, { "epoch": 0.05, "learning_rate": 4.9248902241737925e-05, "loss": 0.0, "step": 195 }, { "epoch": 0.05, "learning_rate": 4.924505045836223e-05, "loss": 0.0001, "step": 196 }, { "epoch": 0.05, "learning_rate": 4.924119867498652e-05, "loss": 0.0, "step": 197 }, { "epoch": 0.05, "learning_rate": 4.923734689161082e-05, "loss": 0.0, "step": 198 }, { "epoch": 0.05, "learning_rate": 4.923349510823512e-05, "loss": 0.0, "step": 199 }, { "epoch": 0.05, "learning_rate": 4.9229643324859414e-05, "loss": 0.0041, "step": 200 }, { "epoch": 0.05, "learning_rate": 4.922579154148371e-05, "loss": 0.0, "step": 201 }, { "epoch": 0.05, "learning_rate": 4.9221939758108004e-05, "loss": 0.0, "step": 202 }, { "epoch": 0.05, "learning_rate": 4.9218087974732305e-05, "loss": 0.0, "step": 203 }, { "epoch": 0.05, "learning_rate": 4.92142361913566e-05, "loss": 0.0, "step": 204 }, { "epoch": 0.05, "learning_rate": 4.9210384407980895e-05, "loss": 0.0, "step": 205 }, { "epoch": 0.05, "learning_rate": 4.92065326246052e-05, "loss": 0.0, "step": 206 }, { "epoch": 0.05, "learning_rate": 4.920268084122949e-05, "loss": 0.0, "step": 207 }, { "epoch": 0.05, "learning_rate": 4.9198829057853794e-05, "loss": 0.0, "step": 208 }, { "epoch": 0.05, "learning_rate": 4.919497727447809e-05, "loss": 0.0, "step": 209 }, { "epoch": 0.05, "learning_rate": 4.919112549110238e-05, "loss": 0.0001, "step": 210 }, { "epoch": 0.05, "learning_rate": 4.918727370772668e-05, "loss": 0.0001, "step": 211 }, { "epoch": 0.05, "learning_rate": 4.9183421924350973e-05, "loss": 0.0, "step": 212 }, { "epoch": 0.05, "learning_rate": 4.9179570140975275e-05, "loss": 0.0, "step": 213 }, { "epoch": 0.05, "learning_rate": 4.917571835759957e-05, "loss": 0.0, "step": 214 }, { "epoch": 0.05, "learning_rate": 4.9171866574223865e-05, "loss": 0.0, "step": 215 }, { "epoch": 0.05, "learning_rate": 4.916801479084817e-05, "loss": 0.0, "step": 216 }, { "epoch": 0.05, "learning_rate": 4.916416300747246e-05, "loss": 0.0113, "step": 217 }, { "epoch": 0.05, "learning_rate": 4.9160311224096763e-05, "loss": 0.0, "step": 218 }, { "epoch": 0.05, "learning_rate": 4.915645944072106e-05, "loss": 0.0, "step": 219 }, { "epoch": 0.05, "learning_rate": 4.9152607657345353e-05, "loss": 0.0, "step": 220 }, { "epoch": 0.05, "learning_rate": 4.914875587396965e-05, "loss": 0.0, "step": 221 }, { "epoch": 0.05, "learning_rate": 4.914490409059394e-05, "loss": 0.0551, "step": 222 }, { "epoch": 0.05, "learning_rate": 4.9141052307218245e-05, "loss": 0.0001, "step": 223 }, { "epoch": 0.05, "learning_rate": 4.913720052384254e-05, "loss": 0.0, "step": 224 }, { "epoch": 0.05, "learning_rate": 4.913334874046684e-05, "loss": 0.0, "step": 225 }, { "epoch": 0.05, "learning_rate": 4.912949695709114e-05, "loss": 0.0, "step": 226 }, { "epoch": 0.05, "learning_rate": 4.912564517371543e-05, "loss": 0.0, "step": 227 }, { "epoch": 0.05, "learning_rate": 4.912179339033973e-05, "loss": 0.0, "step": 228 }, { "epoch": 0.05, "learning_rate": 4.911794160696402e-05, "loss": 0.0, "step": 229 }, { "epoch": 0.05, "learning_rate": 4.911408982358832e-05, "loss": 0.0, "step": 230 }, { "epoch": 0.05, "learning_rate": 4.911023804021262e-05, "loss": 0.0, "step": 231 }, { "epoch": 0.05, "learning_rate": 4.910638625683692e-05, "loss": 0.0, "step": 232 }, { "epoch": 0.05, "learning_rate": 4.9102534473461215e-05, "loss": 0.0, "step": 233 }, { "epoch": 0.05, "learning_rate": 4.909868269008551e-05, "loss": 0.0001, "step": 234 }, { "epoch": 0.05, "learning_rate": 4.909483090670981e-05, "loss": 0.0003, "step": 235 }, { "epoch": 0.05, "learning_rate": 4.9090979123334107e-05, "loss": 0.0, "step": 236 }, { "epoch": 0.05, "learning_rate": 4.908712733995841e-05, "loss": 0.0004, "step": 237 }, { "epoch": 0.06, "learning_rate": 4.9083275556582696e-05, "loss": 0.0, "step": 238 }, { "epoch": 0.06, "learning_rate": 4.9079423773207e-05, "loss": 0.0, "step": 239 }, { "epoch": 0.06, "learning_rate": 4.907557198983129e-05, "loss": 0.0, "step": 240 }, { "epoch": 0.06, "learning_rate": 4.907172020645559e-05, "loss": 0.0, "step": 241 }, { "epoch": 0.06, "learning_rate": 4.906786842307989e-05, "loss": 0.0, "step": 242 }, { "epoch": 0.06, "learning_rate": 4.9064016639704185e-05, "loss": 0.0, "step": 243 }, { "epoch": 0.06, "learning_rate": 4.9060164856328487e-05, "loss": 0.0, "step": 244 }, { "epoch": 0.06, "learning_rate": 4.905631307295278e-05, "loss": 0.0, "step": 245 }, { "epoch": 0.06, "learning_rate": 4.9052461289577076e-05, "loss": 0.0, "step": 246 }, { "epoch": 0.06, "learning_rate": 4.904860950620138e-05, "loss": 0.0, "step": 247 }, { "epoch": 0.06, "learning_rate": 4.9044757722825666e-05, "loss": 0.0, "step": 248 }, { "epoch": 0.06, "learning_rate": 4.904090593944997e-05, "loss": 0.0, "step": 249 }, { "epoch": 0.06, "learning_rate": 4.903705415607426e-05, "loss": 0.0, "step": 250 }, { "epoch": 0.06, "learning_rate": 4.903320237269856e-05, "loss": 0.0, "step": 251 }, { "epoch": 0.06, "learning_rate": 4.902935058932286e-05, "loss": 0.0, "step": 252 }, { "epoch": 0.06, "learning_rate": 4.9025498805947155e-05, "loss": 0.0, "step": 253 }, { "epoch": 0.06, "learning_rate": 4.9021647022571456e-05, "loss": 0.0001, "step": 254 }, { "epoch": 0.06, "learning_rate": 4.901779523919575e-05, "loss": 0.0, "step": 255 }, { "epoch": 0.06, "learning_rate": 4.9013943455820046e-05, "loss": 0.0, "step": 256 }, { "epoch": 0.06, "learning_rate": 4.901009167244434e-05, "loss": 0.0, "step": 257 }, { "epoch": 0.06, "learning_rate": 4.9006239889068636e-05, "loss": 0.0, "step": 258 }, { "epoch": 0.06, "learning_rate": 4.900238810569294e-05, "loss": 0.0, "step": 259 }, { "epoch": 0.06, "learning_rate": 4.899853632231723e-05, "loss": 0.0, "step": 260 }, { "epoch": 0.06, "learning_rate": 4.8994684538941535e-05, "loss": 0.0, "step": 261 }, { "epoch": 0.06, "learning_rate": 4.899083275556583e-05, "loss": 0.0, "step": 262 }, { "epoch": 0.06, "learning_rate": 4.8986980972190125e-05, "loss": 0.0, "step": 263 }, { "epoch": 0.06, "learning_rate": 4.8983129188814426e-05, "loss": 0.0, "step": 264 }, { "epoch": 0.06, "learning_rate": 4.897927740543872e-05, "loss": 0.0, "step": 265 }, { "epoch": 0.06, "learning_rate": 4.8975425622063016e-05, "loss": 0.0009, "step": 266 }, { "epoch": 0.06, "learning_rate": 4.897157383868731e-05, "loss": 0.0, "step": 267 }, { "epoch": 0.06, "learning_rate": 4.896772205531161e-05, "loss": 0.0054, "step": 268 }, { "epoch": 0.06, "learning_rate": 4.896387027193591e-05, "loss": 0.0, "step": 269 }, { "epoch": 0.06, "learning_rate": 4.89600184885602e-05, "loss": 0.0, "step": 270 }, { "epoch": 0.06, "learning_rate": 4.8956166705184505e-05, "loss": 0.0, "step": 271 }, { "epoch": 0.06, "learning_rate": 4.89523149218088e-05, "loss": 0.0, "step": 272 }, { "epoch": 0.06, "learning_rate": 4.89484631384331e-05, "loss": 0.0, "step": 273 }, { "epoch": 0.06, "learning_rate": 4.8944611355057396e-05, "loss": 0.0, "step": 274 }, { "epoch": 0.06, "learning_rate": 4.894075957168169e-05, "loss": 0.0, "step": 275 }, { "epoch": 0.06, "learning_rate": 4.8936907788305986e-05, "loss": 0.0, "step": 276 }, { "epoch": 0.06, "learning_rate": 4.893305600493028e-05, "loss": 0.0, "step": 277 }, { "epoch": 0.06, "learning_rate": 4.892920422155458e-05, "loss": 0.0, "step": 278 }, { "epoch": 0.06, "learning_rate": 4.892535243817888e-05, "loss": 0.0, "step": 279 }, { "epoch": 0.06, "learning_rate": 4.892150065480318e-05, "loss": 0.0, "step": 280 }, { "epoch": 0.06, "learning_rate": 4.8917648871427474e-05, "loss": 0.0, "step": 281 }, { "epoch": 0.07, "learning_rate": 4.891379708805177e-05, "loss": 0.0, "step": 282 }, { "epoch": 0.07, "learning_rate": 4.890994530467607e-05, "loss": 0.0, "step": 283 }, { "epoch": 0.07, "learning_rate": 4.8906093521300366e-05, "loss": 0.0, "step": 284 }, { "epoch": 0.07, "learning_rate": 4.890224173792466e-05, "loss": 0.0, "step": 285 }, { "epoch": 0.07, "learning_rate": 4.8898389954548956e-05, "loss": 0.0, "step": 286 }, { "epoch": 0.07, "learning_rate": 4.889453817117325e-05, "loss": 0.0, "step": 287 }, { "epoch": 0.07, "learning_rate": 4.889068638779755e-05, "loss": 0.0, "step": 288 }, { "epoch": 0.07, "learning_rate": 4.888683460442185e-05, "loss": 0.0003, "step": 289 }, { "epoch": 0.07, "learning_rate": 4.888298282104615e-05, "loss": 0.0, "step": 290 }, { "epoch": 0.07, "learning_rate": 4.8879131037670444e-05, "loss": 0.0, "step": 291 }, { "epoch": 0.07, "learning_rate": 4.887527925429474e-05, "loss": 0.0, "step": 292 }, { "epoch": 0.07, "learning_rate": 4.887142747091904e-05, "loss": 0.0, "step": 293 }, { "epoch": 0.07, "learning_rate": 4.886757568754333e-05, "loss": 0.0, "step": 294 }, { "epoch": 0.07, "learning_rate": 4.886372390416763e-05, "loss": 0.0, "step": 295 }, { "epoch": 0.07, "learning_rate": 4.8859872120791926e-05, "loss": 0.0, "step": 296 }, { "epoch": 0.07, "learning_rate": 4.885602033741623e-05, "loss": 0.0003, "step": 297 }, { "epoch": 0.07, "learning_rate": 4.885216855404052e-05, "loss": 0.0, "step": 298 }, { "epoch": 0.07, "learning_rate": 4.884831677066482e-05, "loss": 0.0081, "step": 299 }, { "epoch": 0.07, "learning_rate": 4.884446498728912e-05, "loss": 0.0, "step": 300 }, { "epoch": 0.07, "learning_rate": 4.8840613203913414e-05, "loss": 0.0, "step": 301 }, { "epoch": 0.07, "learning_rate": 4.8836761420537716e-05, "loss": 0.0, "step": 302 }, { "epoch": 0.07, "learning_rate": 4.883290963716201e-05, "loss": 0.0, "step": 303 }, { "epoch": 0.07, "learning_rate": 4.8829057853786306e-05, "loss": 0.0003, "step": 304 }, { "epoch": 0.07, "learning_rate": 4.88252060704106e-05, "loss": 0.0, "step": 305 }, { "epoch": 0.07, "learning_rate": 4.8821354287034896e-05, "loss": 0.0, "step": 306 }, { "epoch": 0.07, "learning_rate": 4.88175025036592e-05, "loss": 0.0, "step": 307 }, { "epoch": 0.07, "learning_rate": 4.881365072028349e-05, "loss": 0.0001, "step": 308 }, { "epoch": 0.07, "learning_rate": 4.8809798936907794e-05, "loss": 0.0001, "step": 309 }, { "epoch": 0.07, "learning_rate": 4.880594715353209e-05, "loss": 0.0, "step": 310 }, { "epoch": 0.07, "learning_rate": 4.8802095370156384e-05, "loss": 0.0, "step": 311 }, { "epoch": 0.07, "learning_rate": 4.8798243586780686e-05, "loss": 0.0, "step": 312 }, { "epoch": 0.07, "learning_rate": 4.8794391803404974e-05, "loss": 0.0001, "step": 313 }, { "epoch": 0.07, "learning_rate": 4.8790540020029276e-05, "loss": 0.0014, "step": 314 }, { "epoch": 0.07, "learning_rate": 4.878668823665357e-05, "loss": 0.0, "step": 315 }, { "epoch": 0.07, "learning_rate": 4.878283645327787e-05, "loss": 0.0, "step": 316 }, { "epoch": 0.07, "learning_rate": 4.877898466990217e-05, "loss": 0.0, "step": 317 }, { "epoch": 0.07, "learning_rate": 4.877513288652646e-05, "loss": 0.0, "step": 318 }, { "epoch": 0.07, "learning_rate": 4.8771281103150764e-05, "loss": 0.0, "step": 319 }, { "epoch": 0.07, "learning_rate": 4.876742931977506e-05, "loss": 0.0, "step": 320 }, { "epoch": 0.07, "learning_rate": 4.876357753639936e-05, "loss": 0.0001, "step": 321 }, { "epoch": 0.07, "learning_rate": 4.875972575302365e-05, "loss": 0.0, "step": 322 }, { "epoch": 0.07, "learning_rate": 4.8755873969647944e-05, "loss": 0.0, "step": 323 }, { "epoch": 0.07, "learning_rate": 4.8752022186272246e-05, "loss": 0.0, "step": 324 }, { "epoch": 0.08, "learning_rate": 4.874817040289654e-05, "loss": 0.0, "step": 325 }, { "epoch": 0.08, "learning_rate": 4.874431861952084e-05, "loss": 0.0, "step": 326 }, { "epoch": 0.08, "learning_rate": 4.874046683614514e-05, "loss": 0.0, "step": 327 }, { "epoch": 0.08, "learning_rate": 4.873661505276943e-05, "loss": 0.0, "step": 328 }, { "epoch": 0.08, "learning_rate": 4.8732763269393734e-05, "loss": 0.0, "step": 329 }, { "epoch": 0.08, "learning_rate": 4.872891148601803e-05, "loss": 0.003, "step": 330 }, { "epoch": 0.08, "learning_rate": 4.872505970264233e-05, "loss": 0.0, "step": 331 }, { "epoch": 0.08, "learning_rate": 4.872120791926662e-05, "loss": 0.0, "step": 332 }, { "epoch": 0.08, "learning_rate": 4.871735613589092e-05, "loss": 0.0, "step": 333 }, { "epoch": 0.08, "learning_rate": 4.8713504352515216e-05, "loss": 0.0, "step": 334 }, { "epoch": 0.08, "learning_rate": 4.870965256913951e-05, "loss": 0.0, "step": 335 }, { "epoch": 0.08, "learning_rate": 4.870580078576381e-05, "loss": 0.0034, "step": 336 }, { "epoch": 0.08, "learning_rate": 4.870194900238811e-05, "loss": 0.0, "step": 337 }, { "epoch": 0.08, "learning_rate": 4.869809721901241e-05, "loss": 0.0001, "step": 338 }, { "epoch": 0.08, "learning_rate": 4.8694245435636704e-05, "loss": 0.0, "step": 339 }, { "epoch": 0.08, "learning_rate": 4.8690393652261e-05, "loss": 0.0, "step": 340 }, { "epoch": 0.08, "learning_rate": 4.8686541868885294e-05, "loss": 0.0, "step": 341 }, { "epoch": 0.08, "learning_rate": 4.868269008550959e-05, "loss": 0.0001, "step": 342 }, { "epoch": 0.08, "learning_rate": 4.867883830213389e-05, "loss": 0.0, "step": 343 }, { "epoch": 0.08, "learning_rate": 4.8674986518758185e-05, "loss": 0.0, "step": 344 }, { "epoch": 0.08, "learning_rate": 4.867113473538249e-05, "loss": 0.0, "step": 345 }, { "epoch": 0.08, "learning_rate": 4.866728295200678e-05, "loss": 0.0, "step": 346 }, { "epoch": 0.08, "learning_rate": 4.866343116863108e-05, "loss": 0.0, "step": 347 }, { "epoch": 0.08, "learning_rate": 4.865957938525538e-05, "loss": 0.0, "step": 348 }, { "epoch": 0.08, "learning_rate": 4.8655727601879674e-05, "loss": 0.0, "step": 349 }, { "epoch": 0.08, "learning_rate": 4.865187581850397e-05, "loss": 0.0, "step": 350 }, { "epoch": 0.08, "learning_rate": 4.8648024035128264e-05, "loss": 0.0, "step": 351 }, { "epoch": 0.08, "learning_rate": 4.8644172251752565e-05, "loss": 0.0, "step": 352 }, { "epoch": 0.08, "learning_rate": 4.864032046837686e-05, "loss": 0.0, "step": 353 }, { "epoch": 0.08, "learning_rate": 4.8636468685001155e-05, "loss": 0.0, "step": 354 }, { "epoch": 0.08, "learning_rate": 4.863261690162546e-05, "loss": 0.0, "step": 355 }, { "epoch": 0.08, "learning_rate": 4.862876511824975e-05, "loss": 0.0, "step": 356 }, { "epoch": 0.08, "learning_rate": 4.8624913334874054e-05, "loss": 0.0, "step": 357 }, { "epoch": 0.08, "learning_rate": 4.862106155149835e-05, "loss": 0.0213, "step": 358 }, { "epoch": 0.08, "learning_rate": 4.8617209768122644e-05, "loss": 0.0001, "step": 359 }, { "epoch": 0.08, "learning_rate": 4.861335798474694e-05, "loss": 0.0, "step": 360 }, { "epoch": 0.08, "learning_rate": 4.8609506201371234e-05, "loss": 0.0, "step": 361 }, { "epoch": 0.08, "learning_rate": 4.8605654417995535e-05, "loss": 0.0, "step": 362 }, { "epoch": 0.08, "learning_rate": 4.860180263461983e-05, "loss": 0.0, "step": 363 }, { "epoch": 0.08, "learning_rate": 4.8597950851244125e-05, "loss": 0.0084, "step": 364 }, { "epoch": 0.08, "learning_rate": 4.859409906786843e-05, "loss": 0.0, "step": 365 }, { "epoch": 0.08, "learning_rate": 4.859024728449272e-05, "loss": 0.0, "step": 366 }, { "epoch": 0.08, "learning_rate": 4.8586395501117024e-05, "loss": 0.0, "step": 367 }, { "epoch": 0.09, "learning_rate": 4.858254371774132e-05, "loss": 0.0, "step": 368 }, { "epoch": 0.09, "learning_rate": 4.8578691934365613e-05, "loss": 0.0, "step": 369 }, { "epoch": 0.09, "learning_rate": 4.857484015098991e-05, "loss": 0.0, "step": 370 }, { "epoch": 0.09, "learning_rate": 4.8570988367614203e-05, "loss": 0.0, "step": 371 }, { "epoch": 0.09, "learning_rate": 4.8567136584238505e-05, "loss": 0.0, "step": 372 }, { "epoch": 0.09, "learning_rate": 4.85632848008628e-05, "loss": 0.0, "step": 373 }, { "epoch": 0.09, "learning_rate": 4.85594330174871e-05, "loss": 0.0, "step": 374 }, { "epoch": 0.09, "learning_rate": 4.85555812341114e-05, "loss": 0.0, "step": 375 }, { "epoch": 0.09, "learning_rate": 4.855172945073569e-05, "loss": 0.0, "step": 376 }, { "epoch": 0.09, "learning_rate": 4.8547877667359993e-05, "loss": 0.0, "step": 377 }, { "epoch": 0.09, "learning_rate": 4.854402588398428e-05, "loss": 0.0, "step": 378 }, { "epoch": 0.09, "learning_rate": 4.854017410060858e-05, "loss": 0.0, "step": 379 }, { "epoch": 0.09, "learning_rate": 4.853632231723288e-05, "loss": 0.0, "step": 380 }, { "epoch": 0.09, "learning_rate": 4.853247053385718e-05, "loss": 0.0, "step": 381 }, { "epoch": 0.09, "learning_rate": 4.8528618750481475e-05, "loss": 0.0, "step": 382 }, { "epoch": 0.09, "learning_rate": 4.852476696710577e-05, "loss": 0.0, "step": 383 }, { "epoch": 0.09, "learning_rate": 4.852091518373007e-05, "loss": 0.0006, "step": 384 }, { "epoch": 0.09, "learning_rate": 4.851706340035437e-05, "loss": 0.0, "step": 385 }, { "epoch": 0.09, "learning_rate": 4.851321161697867e-05, "loss": 0.0, "step": 386 }, { "epoch": 0.09, "learning_rate": 4.850935983360296e-05, "loss": 0.0, "step": 387 }, { "epoch": 0.09, "learning_rate": 4.850550805022726e-05, "loss": 0.0, "step": 388 }, { "epoch": 0.09, "learning_rate": 4.850165626685155e-05, "loss": 0.0, "step": 389 }, { "epoch": 0.09, "learning_rate": 4.849780448347585e-05, "loss": 0.0, "step": 390 }, { "epoch": 0.09, "learning_rate": 4.849395270010015e-05, "loss": 0.0, "step": 391 }, { "epoch": 0.09, "learning_rate": 4.8490100916724445e-05, "loss": 0.0001, "step": 392 }, { "epoch": 0.09, "learning_rate": 4.848624913334875e-05, "loss": 0.0, "step": 393 }, { "epoch": 0.09, "learning_rate": 4.848239734997304e-05, "loss": 0.0, "step": 394 }, { "epoch": 0.09, "learning_rate": 4.8478545566597337e-05, "loss": 0.0, "step": 395 }, { "epoch": 0.09, "learning_rate": 4.847469378322164e-05, "loss": 0.0, "step": 396 }, { "epoch": 0.09, "learning_rate": 4.8470841999845926e-05, "loss": 0.0, "step": 397 }, { "epoch": 0.09, "learning_rate": 4.846699021647023e-05, "loss": 0.0001, "step": 398 }, { "epoch": 0.09, "learning_rate": 4.846313843309452e-05, "loss": 0.0, "step": 399 }, { "epoch": 0.09, "learning_rate": 4.845928664971882e-05, "loss": 0.0, "step": 400 }, { "epoch": 0.09, "learning_rate": 4.845543486634312e-05, "loss": 0.0, "step": 401 }, { "epoch": 0.09, "learning_rate": 4.8451583082967415e-05, "loss": 0.0001, "step": 402 }, { "epoch": 0.09, "learning_rate": 4.8447731299591717e-05, "loss": 0.0, "step": 403 }, { "epoch": 0.09, "learning_rate": 4.844387951621601e-05, "loss": 0.0, "step": 404 }, { "epoch": 0.09, "learning_rate": 4.8440027732840306e-05, "loss": 0.0, "step": 405 }, { "epoch": 0.09, "learning_rate": 4.84361759494646e-05, "loss": 0.0, "step": 406 }, { "epoch": 0.09, "learning_rate": 4.8432324166088896e-05, "loss": 0.0, "step": 407 }, { "epoch": 0.09, "learning_rate": 4.84284723827132e-05, "loss": 0.0, "step": 408 }, { "epoch": 0.09, "learning_rate": 4.842462059933749e-05, "loss": 0.0, "step": 409 }, { "epoch": 0.09, "learning_rate": 4.8420768815961795e-05, "loss": 0.0, "step": 410 }, { "epoch": 0.09, "learning_rate": 4.841691703258609e-05, "loss": 0.0, "step": 411 }, { "epoch": 0.1, "learning_rate": 4.8413065249210385e-05, "loss": 0.0, "step": 412 }, { "epoch": 0.1, "learning_rate": 4.8409213465834686e-05, "loss": 0.0005, "step": 413 }, { "epoch": 0.1, "learning_rate": 4.840536168245898e-05, "loss": 0.0002, "step": 414 }, { "epoch": 0.1, "learning_rate": 4.840150989908328e-05, "loss": 0.0, "step": 415 }, { "epoch": 0.1, "learning_rate": 4.839765811570757e-05, "loss": 0.0, "step": 416 }, { "epoch": 0.1, "learning_rate": 4.839380633233187e-05, "loss": 0.0, "step": 417 }, { "epoch": 0.1, "learning_rate": 4.838995454895617e-05, "loss": 0.0002, "step": 418 }, { "epoch": 0.1, "learning_rate": 4.838610276558046e-05, "loss": 0.0, "step": 419 }, { "epoch": 0.1, "learning_rate": 4.8382250982204765e-05, "loss": 0.0, "step": 420 }, { "epoch": 0.1, "learning_rate": 4.837839919882906e-05, "loss": 0.0006, "step": 421 }, { "epoch": 0.1, "learning_rate": 4.837454741545336e-05, "loss": 0.0, "step": 422 }, { "epoch": 0.1, "learning_rate": 4.8370695632077656e-05, "loss": 0.0, "step": 423 }, { "epoch": 0.1, "learning_rate": 4.836684384870195e-05, "loss": 0.0, "step": 424 }, { "epoch": 0.1, "learning_rate": 4.8362992065326246e-05, "loss": 0.0, "step": 425 }, { "epoch": 0.1, "learning_rate": 4.835914028195054e-05, "loss": 0.0162, "step": 426 }, { "epoch": 0.1, "learning_rate": 4.835528849857484e-05, "loss": 0.0, "step": 427 }, { "epoch": 0.1, "learning_rate": 4.835143671519914e-05, "loss": 0.0, "step": 428 }, { "epoch": 0.1, "learning_rate": 4.834758493182344e-05, "loss": 0.0, "step": 429 }, { "epoch": 0.1, "learning_rate": 4.8343733148447735e-05, "loss": 0.0003, "step": 430 }, { "epoch": 0.1, "learning_rate": 4.833988136507203e-05, "loss": 0.0, "step": 431 }, { "epoch": 0.1, "learning_rate": 4.833602958169633e-05, "loss": 0.0, "step": 432 }, { "epoch": 0.1, "learning_rate": 4.8332177798320626e-05, "loss": 0.0, "step": 433 }, { "epoch": 0.1, "learning_rate": 4.832832601494492e-05, "loss": 0.0, "step": 434 }, { "epoch": 0.1, "learning_rate": 4.8324474231569216e-05, "loss": 0.0, "step": 435 }, { "epoch": 0.1, "learning_rate": 4.832062244819351e-05, "loss": 0.0016, "step": 436 }, { "epoch": 0.1, "learning_rate": 4.831677066481781e-05, "loss": 0.0094, "step": 437 }, { "epoch": 0.1, "learning_rate": 4.831291888144211e-05, "loss": 0.0, "step": 438 }, { "epoch": 0.1, "learning_rate": 4.830906709806641e-05, "loss": 0.0, "step": 439 }, { "epoch": 0.1, "learning_rate": 4.8305215314690704e-05, "loss": 0.0018, "step": 440 }, { "epoch": 0.1, "learning_rate": 4.8301363531315e-05, "loss": 0.0, "step": 441 }, { "epoch": 0.1, "learning_rate": 4.82975117479393e-05, "loss": 0.0, "step": 442 }, { "epoch": 0.1, "learning_rate": 4.8293659964563596e-05, "loss": 0.0, "step": 443 }, { "epoch": 0.1, "learning_rate": 4.828980818118789e-05, "loss": 0.0, "step": 444 }, { "epoch": 0.1, "learning_rate": 4.8285956397812186e-05, "loss": 0.0001, "step": 445 }, { "epoch": 0.1, "learning_rate": 4.828210461443649e-05, "loss": 0.0, "step": 446 }, { "epoch": 0.1, "learning_rate": 4.827825283106078e-05, "loss": 0.0022, "step": 447 }, { "epoch": 0.1, "learning_rate": 4.827440104768508e-05, "loss": 0.0, "step": 448 }, { "epoch": 0.1, "learning_rate": 4.827054926430938e-05, "loss": 0.0, "step": 449 }, { "epoch": 0.1, "learning_rate": 4.8266697480933674e-05, "loss": 0.0, "step": 450 }, { "epoch": 0.1, "learning_rate": 4.8262845697557976e-05, "loss": 0.0, "step": 451 }, { "epoch": 0.1, "learning_rate": 4.825899391418227e-05, "loss": 0.0002, "step": 452 }, { "epoch": 0.1, "learning_rate": 4.8255142130806566e-05, "loss": 0.0, "step": 453 }, { "epoch": 0.1, "learning_rate": 4.825129034743086e-05, "loss": 0.0, "step": 454 }, { "epoch": 0.11, "learning_rate": 4.8247438564055156e-05, "loss": 0.0, "step": 455 }, { "epoch": 0.11, "learning_rate": 4.824358678067946e-05, "loss": 0.0213, "step": 456 }, { "epoch": 0.11, "learning_rate": 4.823973499730375e-05, "loss": 0.0004, "step": 457 }, { "epoch": 0.11, "learning_rate": 4.8235883213928054e-05, "loss": 0.0002, "step": 458 }, { "epoch": 0.11, "learning_rate": 4.823203143055235e-05, "loss": 0.0, "step": 459 }, { "epoch": 0.11, "learning_rate": 4.8228179647176644e-05, "loss": 0.0, "step": 460 }, { "epoch": 0.11, "learning_rate": 4.8224327863800946e-05, "loss": 0.0, "step": 461 }, { "epoch": 0.11, "learning_rate": 4.8220476080425234e-05, "loss": 0.0, "step": 462 }, { "epoch": 0.11, "learning_rate": 4.8216624297049536e-05, "loss": 0.0001, "step": 463 }, { "epoch": 0.11, "learning_rate": 4.821277251367383e-05, "loss": 0.0, "step": 464 }, { "epoch": 0.11, "learning_rate": 4.820892073029813e-05, "loss": 0.0, "step": 465 }, { "epoch": 0.11, "learning_rate": 4.820506894692243e-05, "loss": 0.061, "step": 466 }, { "epoch": 0.11, "learning_rate": 4.820121716354672e-05, "loss": 0.0, "step": 467 }, { "epoch": 0.11, "learning_rate": 4.8197365380171024e-05, "loss": 0.0, "step": 468 }, { "epoch": 0.11, "learning_rate": 4.819351359679532e-05, "loss": 0.0074, "step": 469 }, { "epoch": 0.11, "learning_rate": 4.818966181341962e-05, "loss": 0.0, "step": 470 }, { "epoch": 0.11, "learning_rate": 4.8185810030043916e-05, "loss": 0.0, "step": 471 }, { "epoch": 0.11, "learning_rate": 4.8181958246668204e-05, "loss": 0.0001, "step": 472 }, { "epoch": 0.11, "learning_rate": 4.8178106463292506e-05, "loss": 0.0, "step": 473 }, { "epoch": 0.11, "learning_rate": 4.81742546799168e-05, "loss": 0.0, "step": 474 }, { "epoch": 0.11, "learning_rate": 4.81704028965411e-05, "loss": 0.0, "step": 475 }, { "epoch": 0.11, "learning_rate": 4.81665511131654e-05, "loss": 0.0026, "step": 476 }, { "epoch": 0.11, "learning_rate": 4.816269932978969e-05, "loss": 0.0, "step": 477 }, { "epoch": 0.11, "learning_rate": 4.8158847546413994e-05, "loss": 0.0, "step": 478 }, { "epoch": 0.11, "learning_rate": 4.815499576303829e-05, "loss": 0.0, "step": 479 }, { "epoch": 0.11, "learning_rate": 4.815114397966259e-05, "loss": 0.0002, "step": 480 }, { "epoch": 0.11, "learning_rate": 4.814729219628688e-05, "loss": 0.0, "step": 481 }, { "epoch": 0.11, "learning_rate": 4.814344041291118e-05, "loss": 0.0, "step": 482 }, { "epoch": 0.11, "learning_rate": 4.8139588629535476e-05, "loss": 0.0, "step": 483 }, { "epoch": 0.11, "learning_rate": 4.813573684615977e-05, "loss": 0.0006, "step": 484 }, { "epoch": 0.11, "learning_rate": 4.813188506278407e-05, "loss": 0.0, "step": 485 }, { "epoch": 0.11, "learning_rate": 4.812803327940837e-05, "loss": 0.0, "step": 486 }, { "epoch": 0.11, "learning_rate": 4.812418149603267e-05, "loss": 0.099, "step": 487 }, { "epoch": 0.11, "learning_rate": 4.8120329712656964e-05, "loss": 0.0, "step": 488 }, { "epoch": 0.11, "learning_rate": 4.811647792928126e-05, "loss": 0.0, "step": 489 }, { "epoch": 0.11, "learning_rate": 4.8112626145905554e-05, "loss": 0.0, "step": 490 }, { "epoch": 0.11, "learning_rate": 4.810877436252985e-05, "loss": 0.0, "step": 491 }, { "epoch": 0.11, "learning_rate": 4.810492257915415e-05, "loss": 0.0, "step": 492 }, { "epoch": 0.11, "learning_rate": 4.8101070795778445e-05, "loss": 0.0, "step": 493 }, { "epoch": 0.11, "learning_rate": 4.809721901240275e-05, "loss": 0.0018, "step": 494 }, { "epoch": 0.11, "learning_rate": 4.809336722902704e-05, "loss": 0.0, "step": 495 }, { "epoch": 0.11, "learning_rate": 4.808951544565134e-05, "loss": 0.0, "step": 496 }, { "epoch": 0.11, "learning_rate": 4.808566366227564e-05, "loss": 0.0, "step": 497 }, { "epoch": 0.12, "learning_rate": 4.8081811878899934e-05, "loss": 0.0, "step": 498 }, { "epoch": 0.12, "learning_rate": 4.8077960095524236e-05, "loss": 0.0041, "step": 499 }, { "epoch": 0.12, "learning_rate": 4.8074108312148524e-05, "loss": 0.0047, "step": 500 }, { "epoch": 0.12, "learning_rate": 4.8070256528772825e-05, "loss": 0.0, "step": 501 }, { "epoch": 0.12, "learning_rate": 4.806640474539712e-05, "loss": 0.0, "step": 502 }, { "epoch": 0.12, "learning_rate": 4.8062552962021415e-05, "loss": 0.0, "step": 503 }, { "epoch": 0.12, "learning_rate": 4.805870117864572e-05, "loss": 0.0, "step": 504 }, { "epoch": 0.12, "learning_rate": 4.805484939527001e-05, "loss": 0.0, "step": 505 }, { "epoch": 0.12, "learning_rate": 4.8050997611894314e-05, "loss": 0.0, "step": 506 }, { "epoch": 0.12, "learning_rate": 4.804714582851861e-05, "loss": 0.0, "step": 507 }, { "epoch": 0.12, "learning_rate": 4.8043294045142904e-05, "loss": 0.0, "step": 508 }, { "epoch": 0.12, "learning_rate": 4.80394422617672e-05, "loss": 0.0, "step": 509 }, { "epoch": 0.12, "learning_rate": 4.8035590478391494e-05, "loss": 0.0001, "step": 510 }, { "epoch": 0.12, "learning_rate": 4.8031738695015795e-05, "loss": 0.0014, "step": 511 }, { "epoch": 0.12, "learning_rate": 4.802788691164009e-05, "loss": 0.0, "step": 512 }, { "epoch": 0.12, "learning_rate": 4.8024035128264385e-05, "loss": 0.0, "step": 513 }, { "epoch": 0.12, "learning_rate": 4.802018334488869e-05, "loss": 0.0, "step": 514 }, { "epoch": 0.12, "learning_rate": 4.801633156151298e-05, "loss": 0.0, "step": 515 }, { "epoch": 0.12, "learning_rate": 4.8012479778137284e-05, "loss": 0.0, "step": 516 }, { "epoch": 0.12, "learning_rate": 4.800862799476158e-05, "loss": 0.0092, "step": 517 }, { "epoch": 0.12, "learning_rate": 4.8004776211385874e-05, "loss": 0.0, "step": 518 }, { "epoch": 0.12, "learning_rate": 4.800092442801017e-05, "loss": 0.0, "step": 519 }, { "epoch": 0.12, "learning_rate": 4.7997072644634463e-05, "loss": 0.0, "step": 520 }, { "epoch": 0.12, "learning_rate": 4.7993220861258765e-05, "loss": 0.0, "step": 521 }, { "epoch": 0.12, "learning_rate": 4.798936907788306e-05, "loss": 0.0, "step": 522 }, { "epoch": 0.12, "learning_rate": 4.798551729450736e-05, "loss": 0.0, "step": 523 }, { "epoch": 0.12, "learning_rate": 4.798166551113166e-05, "loss": 0.0001, "step": 524 }, { "epoch": 0.12, "learning_rate": 4.797781372775595e-05, "loss": 0.0, "step": 525 }, { "epoch": 0.12, "learning_rate": 4.7973961944380254e-05, "loss": 0.0, "step": 526 }, { "epoch": 0.12, "learning_rate": 4.797011016100455e-05, "loss": 0.0, "step": 527 }, { "epoch": 0.12, "learning_rate": 4.7966258377628843e-05, "loss": 0.0, "step": 528 }, { "epoch": 0.12, "learning_rate": 4.796240659425314e-05, "loss": 0.0, "step": 529 }, { "epoch": 0.12, "learning_rate": 4.795855481087744e-05, "loss": 0.0, "step": 530 }, { "epoch": 0.12, "learning_rate": 4.7954703027501735e-05, "loss": 0.0002, "step": 531 }, { "epoch": 0.12, "learning_rate": 4.795085124412603e-05, "loss": 0.0002, "step": 532 }, { "epoch": 0.12, "learning_rate": 4.794699946075033e-05, "loss": 0.0, "step": 533 }, { "epoch": 0.12, "learning_rate": 4.794314767737463e-05, "loss": 0.0004, "step": 534 }, { "epoch": 0.12, "learning_rate": 4.793929589399893e-05, "loss": 0.0, "step": 535 }, { "epoch": 0.12, "learning_rate": 4.7935444110623223e-05, "loss": 0.0004, "step": 536 }, { "epoch": 0.12, "learning_rate": 4.793159232724752e-05, "loss": 0.0, "step": 537 }, { "epoch": 0.12, "learning_rate": 4.792774054387181e-05, "loss": 0.0, "step": 538 }, { "epoch": 0.12, "learning_rate": 4.792388876049611e-05, "loss": 0.0, "step": 539 }, { "epoch": 0.12, "learning_rate": 4.792003697712041e-05, "loss": 0.0, "step": 540 }, { "epoch": 0.13, "learning_rate": 4.7916185193744705e-05, "loss": 0.0001, "step": 541 }, { "epoch": 0.13, "learning_rate": 4.791233341036901e-05, "loss": 0.0003, "step": 542 }, { "epoch": 0.13, "learning_rate": 4.79084816269933e-05, "loss": 0.0, "step": 543 }, { "epoch": 0.13, "learning_rate": 4.79046298436176e-05, "loss": 0.0, "step": 544 }, { "epoch": 0.13, "learning_rate": 4.79007780602419e-05, "loss": 0.0, "step": 545 }, { "epoch": 0.13, "learning_rate": 4.7896926276866187e-05, "loss": 0.0, "step": 546 }, { "epoch": 0.13, "learning_rate": 4.789307449349049e-05, "loss": 0.0, "step": 547 }, { "epoch": 0.13, "learning_rate": 4.788922271011478e-05, "loss": 0.0, "step": 548 }, { "epoch": 0.13, "learning_rate": 4.788537092673908e-05, "loss": 0.0, "step": 549 }, { "epoch": 0.13, "learning_rate": 4.788151914336338e-05, "loss": 0.0, "step": 550 }, { "epoch": 0.13, "learning_rate": 4.7877667359987675e-05, "loss": 0.0, "step": 551 }, { "epoch": 0.13, "learning_rate": 4.7873815576611977e-05, "loss": 0.0, "step": 552 }, { "epoch": 0.13, "learning_rate": 4.786996379323627e-05, "loss": 0.0, "step": 553 }, { "epoch": 0.13, "learning_rate": 4.7866112009860567e-05, "loss": 0.0, "step": 554 }, { "epoch": 0.13, "learning_rate": 4.786226022648487e-05, "loss": 0.0, "step": 555 }, { "epoch": 0.13, "learning_rate": 4.7858408443109156e-05, "loss": 0.0001, "step": 556 }, { "epoch": 0.13, "learning_rate": 4.785455665973346e-05, "loss": 0.0, "step": 557 }, { "epoch": 0.13, "learning_rate": 4.785070487635775e-05, "loss": 0.0, "step": 558 }, { "epoch": 0.13, "learning_rate": 4.7846853092982055e-05, "loss": 0.0, "step": 559 }, { "epoch": 0.13, "learning_rate": 4.784300130960635e-05, "loss": 0.0176, "step": 560 }, { "epoch": 0.13, "learning_rate": 4.7839149526230645e-05, "loss": 0.0006, "step": 561 }, { "epoch": 0.13, "learning_rate": 4.7835297742854946e-05, "loss": 0.0001, "step": 562 }, { "epoch": 0.13, "learning_rate": 4.783144595947924e-05, "loss": 0.0, "step": 563 }, { "epoch": 0.13, "learning_rate": 4.782759417610354e-05, "loss": 0.0, "step": 564 }, { "epoch": 0.13, "learning_rate": 4.782374239272783e-05, "loss": 0.0, "step": 565 }, { "epoch": 0.13, "learning_rate": 4.781989060935213e-05, "loss": 0.0, "step": 566 }, { "epoch": 0.13, "learning_rate": 4.781603882597643e-05, "loss": 0.0, "step": 567 }, { "epoch": 0.13, "learning_rate": 4.781218704260072e-05, "loss": 0.0, "step": 568 }, { "epoch": 0.13, "learning_rate": 4.7808335259225025e-05, "loss": 0.0, "step": 569 }, { "epoch": 0.13, "learning_rate": 4.780448347584932e-05, "loss": 0.0, "step": 570 }, { "epoch": 0.13, "learning_rate": 4.780063169247362e-05, "loss": 0.0343, "step": 571 }, { "epoch": 0.13, "learning_rate": 4.7796779909097916e-05, "loss": 0.0, "step": 572 }, { "epoch": 0.13, "learning_rate": 4.779292812572221e-05, "loss": 0.0, "step": 573 }, { "epoch": 0.13, "learning_rate": 4.7789076342346506e-05, "loss": 0.0, "step": 574 }, { "epoch": 0.13, "learning_rate": 4.77852245589708e-05, "loss": 0.0, "step": 575 }, { "epoch": 0.13, "learning_rate": 4.77813727755951e-05, "loss": 0.0001, "step": 576 }, { "epoch": 0.13, "learning_rate": 4.77775209922194e-05, "loss": 0.0, "step": 577 }, { "epoch": 0.13, "learning_rate": 4.77736692088437e-05, "loss": 0.0, "step": 578 }, { "epoch": 0.13, "learning_rate": 4.7769817425467995e-05, "loss": 0.0, "step": 579 }, { "epoch": 0.13, "learning_rate": 4.776596564209229e-05, "loss": 0.0, "step": 580 }, { "epoch": 0.13, "learning_rate": 4.776211385871659e-05, "loss": 0.0, "step": 581 }, { "epoch": 0.13, "learning_rate": 4.7758262075340886e-05, "loss": 0.0, "step": 582 }, { "epoch": 0.13, "learning_rate": 4.775441029196519e-05, "loss": 0.0, "step": 583 }, { "epoch": 0.13, "learning_rate": 4.7750558508589476e-05, "loss": 0.0, "step": 584 }, { "epoch": 0.14, "learning_rate": 4.774670672521378e-05, "loss": 0.0, "step": 585 }, { "epoch": 0.14, "learning_rate": 4.774285494183807e-05, "loss": 0.0, "step": 586 }, { "epoch": 0.14, "learning_rate": 4.773900315846237e-05, "loss": 0.0, "step": 587 }, { "epoch": 0.14, "learning_rate": 4.773515137508667e-05, "loss": 0.0, "step": 588 }, { "epoch": 0.14, "learning_rate": 4.7731299591710964e-05, "loss": 0.0, "step": 589 }, { "epoch": 0.14, "learning_rate": 4.772744780833526e-05, "loss": 0.0001, "step": 590 }, { "epoch": 0.14, "learning_rate": 4.772359602495956e-05, "loss": 0.0, "step": 591 }, { "epoch": 0.14, "learning_rate": 4.7719744241583856e-05, "loss": 0.0, "step": 592 }, { "epoch": 0.14, "learning_rate": 4.771589245820815e-05, "loss": 0.0, "step": 593 }, { "epoch": 0.14, "learning_rate": 4.7712040674832446e-05, "loss": 0.0, "step": 594 }, { "epoch": 0.14, "learning_rate": 4.770818889145675e-05, "loss": 0.0, "step": 595 }, { "epoch": 0.14, "learning_rate": 4.770433710808104e-05, "loss": 0.0, "step": 596 }, { "epoch": 0.14, "learning_rate": 4.770048532470534e-05, "loss": 0.0, "step": 597 }, { "epoch": 0.14, "learning_rate": 4.769663354132964e-05, "loss": 0.0014, "step": 598 }, { "epoch": 0.14, "learning_rate": 4.7692781757953934e-05, "loss": 0.0012, "step": 599 }, { "epoch": 0.14, "learning_rate": 4.7688929974578236e-05, "loss": 0.0, "step": 600 }, { "epoch": 0.14, "learning_rate": 4.768507819120253e-05, "loss": 0.0, "step": 601 }, { "epoch": 0.14, "learning_rate": 4.7681226407826826e-05, "loss": 0.0, "step": 602 }, { "epoch": 0.14, "learning_rate": 4.767737462445112e-05, "loss": 0.0, "step": 603 }, { "epoch": 0.14, "learning_rate": 4.7673522841075416e-05, "loss": 0.0, "step": 604 }, { "epoch": 0.14, "learning_rate": 4.766967105769972e-05, "loss": 0.0, "step": 605 }, { "epoch": 0.14, "learning_rate": 4.766581927432401e-05, "loss": 0.0, "step": 606 }, { "epoch": 0.14, "learning_rate": 4.7661967490948314e-05, "loss": 0.0011, "step": 607 }, { "epoch": 0.14, "learning_rate": 4.765811570757261e-05, "loss": 0.0, "step": 608 }, { "epoch": 0.14, "learning_rate": 4.7654263924196904e-05, "loss": 0.0, "step": 609 }, { "epoch": 0.14, "learning_rate": 4.7650412140821206e-05, "loss": 0.0, "step": 610 }, { "epoch": 0.14, "learning_rate": 4.76465603574455e-05, "loss": 0.0, "step": 611 }, { "epoch": 0.14, "learning_rate": 4.7642708574069796e-05, "loss": 0.0, "step": 612 }, { "epoch": 0.14, "learning_rate": 4.763885679069409e-05, "loss": 0.0, "step": 613 }, { "epoch": 0.14, "learning_rate": 4.763500500731839e-05, "loss": 0.0, "step": 614 }, { "epoch": 0.14, "learning_rate": 4.763115322394269e-05, "loss": 0.0, "step": 615 }, { "epoch": 0.14, "learning_rate": 4.762730144056698e-05, "loss": 0.0, "step": 616 }, { "epoch": 0.14, "learning_rate": 4.7623449657191284e-05, "loss": 0.0, "step": 617 }, { "epoch": 0.14, "learning_rate": 4.761959787381558e-05, "loss": 0.0, "step": 618 }, { "epoch": 0.14, "learning_rate": 4.761574609043988e-05, "loss": 0.0, "step": 619 }, { "epoch": 0.14, "learning_rate": 4.7611894307064176e-05, "loss": 0.0, "step": 620 }, { "epoch": 0.14, "learning_rate": 4.760804252368847e-05, "loss": 0.0, "step": 621 }, { "epoch": 0.14, "learning_rate": 4.7604190740312766e-05, "loss": 0.0, "step": 622 }, { "epoch": 0.14, "learning_rate": 4.760033895693706e-05, "loss": 0.0, "step": 623 }, { "epoch": 0.14, "learning_rate": 4.759648717356136e-05, "loss": 0.0001, "step": 624 }, { "epoch": 0.14, "learning_rate": 4.759263539018566e-05, "loss": 0.0, "step": 625 }, { "epoch": 0.14, "learning_rate": 4.758878360680995e-05, "loss": 0.0, "step": 626 }, { "epoch": 0.14, "learning_rate": 4.7584931823434254e-05, "loss": 0.0006, "step": 627 }, { "epoch": 0.15, "learning_rate": 4.758108004005855e-05, "loss": 0.0013, "step": 628 }, { "epoch": 0.15, "learning_rate": 4.757722825668285e-05, "loss": 0.0001, "step": 629 }, { "epoch": 0.15, "learning_rate": 4.757337647330714e-05, "loss": 0.0, "step": 630 }, { "epoch": 0.15, "learning_rate": 4.756952468993144e-05, "loss": 0.0, "step": 631 }, { "epoch": 0.15, "learning_rate": 4.7565672906555736e-05, "loss": 0.0, "step": 632 }, { "epoch": 0.15, "learning_rate": 4.756182112318003e-05, "loss": 0.0, "step": 633 }, { "epoch": 0.15, "learning_rate": 4.755796933980433e-05, "loss": 0.0, "step": 634 }, { "epoch": 0.15, "learning_rate": 4.755411755642863e-05, "loss": 0.0, "step": 635 }, { "epoch": 0.15, "learning_rate": 4.755026577305293e-05, "loss": 0.0, "step": 636 }, { "epoch": 0.15, "learning_rate": 4.7546413989677224e-05, "loss": 0.0, "step": 637 }, { "epoch": 0.15, "learning_rate": 4.754256220630152e-05, "loss": 0.0, "step": 638 }, { "epoch": 0.15, "learning_rate": 4.753871042292582e-05, "loss": 0.0, "step": 639 }, { "epoch": 0.15, "learning_rate": 4.753485863955011e-05, "loss": 0.0, "step": 640 }, { "epoch": 0.15, "learning_rate": 4.753100685617441e-05, "loss": 0.0008, "step": 641 }, { "epoch": 0.15, "learning_rate": 4.7527155072798706e-05, "loss": 0.0, "step": 642 }, { "epoch": 0.15, "learning_rate": 4.752330328942301e-05, "loss": 0.0, "step": 643 }, { "epoch": 0.15, "learning_rate": 4.75194515060473e-05, "loss": 0.0009, "step": 644 }, { "epoch": 0.15, "learning_rate": 4.75155997226716e-05, "loss": 0.0, "step": 645 }, { "epoch": 0.15, "learning_rate": 4.75117479392959e-05, "loss": 0.0001, "step": 646 }, { "epoch": 0.15, "learning_rate": 4.7507896155920194e-05, "loss": 0.0, "step": 647 }, { "epoch": 0.15, "learning_rate": 4.7504044372544496e-05, "loss": 0.0001, "step": 648 }, { "epoch": 0.15, "learning_rate": 4.7500192589168784e-05, "loss": 0.0, "step": 649 }, { "epoch": 0.15, "learning_rate": 4.7496340805793086e-05, "loss": 0.0, "step": 650 }, { "epoch": 0.15, "learning_rate": 4.749248902241738e-05, "loss": 0.0, "step": 651 }, { "epoch": 0.15, "learning_rate": 4.7488637239041675e-05, "loss": 0.0, "step": 652 }, { "epoch": 0.15, "learning_rate": 4.748478545566598e-05, "loss": 0.0, "step": 653 }, { "epoch": 0.15, "learning_rate": 4.748093367229027e-05, "loss": 0.0, "step": 654 }, { "epoch": 0.15, "learning_rate": 4.7477081888914574e-05, "loss": 0.0, "step": 655 }, { "epoch": 0.15, "learning_rate": 4.747323010553887e-05, "loss": 0.0001, "step": 656 }, { "epoch": 0.15, "learning_rate": 4.7469378322163164e-05, "loss": 0.0, "step": 657 }, { "epoch": 0.15, "learning_rate": 4.746552653878746e-05, "loss": 0.0, "step": 658 }, { "epoch": 0.15, "learning_rate": 4.7461674755411754e-05, "loss": 0.0, "step": 659 }, { "epoch": 0.15, "learning_rate": 4.7457822972036055e-05, "loss": 0.0, "step": 660 }, { "epoch": 0.15, "learning_rate": 4.745397118866035e-05, "loss": 0.0001, "step": 661 }, { "epoch": 0.15, "learning_rate": 4.745011940528465e-05, "loss": 0.0, "step": 662 }, { "epoch": 0.15, "learning_rate": 4.744626762190895e-05, "loss": 0.0, "step": 663 }, { "epoch": 0.15, "learning_rate": 4.744241583853324e-05, "loss": 0.0, "step": 664 }, { "epoch": 0.15, "learning_rate": 4.7438564055157544e-05, "loss": 0.0, "step": 665 }, { "epoch": 0.15, "learning_rate": 4.743471227178184e-05, "loss": 0.0, "step": 666 }, { "epoch": 0.15, "learning_rate": 4.7430860488406134e-05, "loss": 0.0, "step": 667 }, { "epoch": 0.15, "learning_rate": 4.742700870503043e-05, "loss": 0.0021, "step": 668 }, { "epoch": 0.15, "learning_rate": 4.7423156921654724e-05, "loss": 0.0, "step": 669 }, { "epoch": 0.15, "learning_rate": 4.7419305138279025e-05, "loss": 0.0, "step": 670 }, { "epoch": 0.16, "learning_rate": 4.741545335490332e-05, "loss": 0.0001, "step": 671 }, { "epoch": 0.16, "learning_rate": 4.741160157152762e-05, "loss": 0.0, "step": 672 }, { "epoch": 0.16, "learning_rate": 4.740774978815192e-05, "loss": 0.0, "step": 673 }, { "epoch": 0.16, "learning_rate": 4.740389800477621e-05, "loss": 0.0, "step": 674 }, { "epoch": 0.16, "learning_rate": 4.7400046221400514e-05, "loss": 0.0001, "step": 675 }, { "epoch": 0.16, "learning_rate": 4.739619443802481e-05, "loss": 0.0, "step": 676 }, { "epoch": 0.16, "learning_rate": 4.7392342654649104e-05, "loss": 0.0, "step": 677 }, { "epoch": 0.16, "learning_rate": 4.73884908712734e-05, "loss": 0.0, "step": 678 }, { "epoch": 0.16, "learning_rate": 4.73846390878977e-05, "loss": 0.0, "step": 679 }, { "epoch": 0.16, "learning_rate": 4.7380787304521995e-05, "loss": 0.0, "step": 680 }, { "epoch": 0.16, "learning_rate": 4.737693552114629e-05, "loss": 0.0007, "step": 681 }, { "epoch": 0.16, "learning_rate": 4.737308373777059e-05, "loss": 0.0, "step": 682 }, { "epoch": 0.16, "learning_rate": 4.736923195439489e-05, "loss": 0.0, "step": 683 }, { "epoch": 0.16, "learning_rate": 4.736538017101919e-05, "loss": 0.0017, "step": 684 }, { "epoch": 0.16, "learning_rate": 4.7361528387643484e-05, "loss": 0.0, "step": 685 }, { "epoch": 0.16, "learning_rate": 4.735767660426778e-05, "loss": 0.0, "step": 686 }, { "epoch": 0.16, "learning_rate": 4.7353824820892073e-05, "loss": 0.0, "step": 687 }, { "epoch": 0.16, "learning_rate": 4.734997303751637e-05, "loss": 0.0002, "step": 688 }, { "epoch": 0.16, "learning_rate": 4.734612125414067e-05, "loss": 0.0, "step": 689 }, { "epoch": 0.16, "learning_rate": 4.7342269470764965e-05, "loss": 0.0001, "step": 690 }, { "epoch": 0.16, "learning_rate": 4.733841768738927e-05, "loss": 0.0, "step": 691 }, { "epoch": 0.16, "learning_rate": 4.733456590401356e-05, "loss": 0.0, "step": 692 }, { "epoch": 0.16, "learning_rate": 4.733071412063786e-05, "loss": 0.0001, "step": 693 }, { "epoch": 0.16, "learning_rate": 4.732686233726216e-05, "loss": 0.0, "step": 694 }, { "epoch": 0.16, "learning_rate": 4.7323010553886453e-05, "loss": 0.0, "step": 695 }, { "epoch": 0.16, "learning_rate": 4.731915877051075e-05, "loss": 0.0, "step": 696 }, { "epoch": 0.16, "learning_rate": 4.731530698713504e-05, "loss": 0.0, "step": 697 }, { "epoch": 0.16, "learning_rate": 4.7311455203759345e-05, "loss": 0.0, "step": 698 }, { "epoch": 0.16, "learning_rate": 4.730760342038364e-05, "loss": 0.0, "step": 699 }, { "epoch": 0.16, "learning_rate": 4.7303751637007935e-05, "loss": 0.0008, "step": 700 }, { "epoch": 0.16, "learning_rate": 4.729989985363224e-05, "loss": 0.0, "step": 701 }, { "epoch": 0.16, "learning_rate": 4.729604807025653e-05, "loss": 0.0, "step": 702 }, { "epoch": 0.16, "learning_rate": 4.7292196286880827e-05, "loss": 0.0, "step": 703 }, { "epoch": 0.16, "learning_rate": 4.728834450350513e-05, "loss": 0.0, "step": 704 }, { "epoch": 0.16, "learning_rate": 4.7284492720129417e-05, "loss": 0.0, "step": 705 }, { "epoch": 0.16, "learning_rate": 4.728064093675372e-05, "loss": 0.0, "step": 706 }, { "epoch": 0.16, "learning_rate": 4.727678915337801e-05, "loss": 0.0, "step": 707 }, { "epoch": 0.16, "learning_rate": 4.7272937370002315e-05, "loss": 0.0, "step": 708 }, { "epoch": 0.16, "learning_rate": 4.726908558662661e-05, "loss": 0.0056, "step": 709 }, { "epoch": 0.16, "learning_rate": 4.7265233803250905e-05, "loss": 0.0, "step": 710 }, { "epoch": 0.16, "learning_rate": 4.7261382019875207e-05, "loss": 0.0, "step": 711 }, { "epoch": 0.16, "learning_rate": 4.72575302364995e-05, "loss": 0.0, "step": 712 }, { "epoch": 0.16, "learning_rate": 4.72536784531238e-05, "loss": 0.0, "step": 713 }, { "epoch": 0.17, "learning_rate": 4.724982666974809e-05, "loss": 0.0, "step": 714 }, { "epoch": 0.17, "learning_rate": 4.724597488637239e-05, "loss": 0.0001, "step": 715 }, { "epoch": 0.17, "learning_rate": 4.724212310299669e-05, "loss": 0.0, "step": 716 }, { "epoch": 0.17, "learning_rate": 4.723827131962098e-05, "loss": 0.0, "step": 717 }, { "epoch": 0.17, "learning_rate": 4.7234419536245285e-05, "loss": 0.0, "step": 718 }, { "epoch": 0.17, "learning_rate": 4.723056775286958e-05, "loss": 0.0, "step": 719 }, { "epoch": 0.17, "learning_rate": 4.722671596949388e-05, "loss": 0.0, "step": 720 }, { "epoch": 0.17, "learning_rate": 4.7222864186118176e-05, "loss": 0.0, "step": 721 }, { "epoch": 0.17, "learning_rate": 4.721901240274247e-05, "loss": 0.0008, "step": 722 }, { "epoch": 0.17, "learning_rate": 4.721516061936677e-05, "loss": 0.0002, "step": 723 }, { "epoch": 0.17, "learning_rate": 4.721130883599106e-05, "loss": 0.0, "step": 724 }, { "epoch": 0.17, "learning_rate": 4.720745705261536e-05, "loss": 0.001, "step": 725 }, { "epoch": 0.17, "learning_rate": 4.720360526923966e-05, "loss": 0.0033, "step": 726 }, { "epoch": 0.17, "learning_rate": 4.719975348586396e-05, "loss": 0.0, "step": 727 }, { "epoch": 0.17, "learning_rate": 4.7195901702488255e-05, "loss": 0.0, "step": 728 }, { "epoch": 0.17, "learning_rate": 4.719204991911255e-05, "loss": 0.0, "step": 729 }, { "epoch": 0.17, "learning_rate": 4.718819813573685e-05, "loss": 0.0, "step": 730 }, { "epoch": 0.17, "learning_rate": 4.7184346352361146e-05, "loss": 0.0, "step": 731 }, { "epoch": 0.17, "learning_rate": 4.718049456898545e-05, "loss": 0.0, "step": 732 }, { "epoch": 0.17, "learning_rate": 4.7176642785609736e-05, "loss": 0.0, "step": 733 }, { "epoch": 0.17, "learning_rate": 4.717279100223404e-05, "loss": 0.0, "step": 734 }, { "epoch": 0.17, "learning_rate": 4.716893921885833e-05, "loss": 0.0004, "step": 735 }, { "epoch": 0.17, "learning_rate": 4.716508743548263e-05, "loss": 0.0, "step": 736 }, { "epoch": 0.17, "learning_rate": 4.716123565210693e-05, "loss": 0.0, "step": 737 }, { "epoch": 0.17, "learning_rate": 4.7157383868731225e-05, "loss": 0.0, "step": 738 }, { "epoch": 0.17, "learning_rate": 4.715353208535552e-05, "loss": 0.0, "step": 739 }, { "epoch": 0.17, "learning_rate": 4.714968030197982e-05, "loss": 0.0038, "step": 740 }, { "epoch": 0.17, "learning_rate": 4.7145828518604116e-05, "loss": 0.0002, "step": 741 }, { "epoch": 0.17, "learning_rate": 4.714197673522841e-05, "loss": 0.0, "step": 742 }, { "epoch": 0.17, "learning_rate": 4.7138124951852706e-05, "loss": 0.0, "step": 743 }, { "epoch": 0.17, "learning_rate": 4.713427316847701e-05, "loss": 0.0, "step": 744 }, { "epoch": 0.17, "learning_rate": 4.71304213851013e-05, "loss": 0.0, "step": 745 }, { "epoch": 0.17, "learning_rate": 4.71265696017256e-05, "loss": 0.0, "step": 746 }, { "epoch": 0.17, "learning_rate": 4.71227178183499e-05, "loss": 0.0, "step": 747 }, { "epoch": 0.17, "learning_rate": 4.7118866034974194e-05, "loss": 0.0, "step": 748 }, { "epoch": 0.17, "learning_rate": 4.7115014251598496e-05, "loss": 0.0, "step": 749 }, { "epoch": 0.17, "learning_rate": 4.711116246822279e-05, "loss": 0.0, "step": 750 }, { "epoch": 0.17, "learning_rate": 4.7107310684847086e-05, "loss": 0.0, "step": 751 }, { "epoch": 0.17, "learning_rate": 4.710345890147138e-05, "loss": 0.0002, "step": 752 }, { "epoch": 0.17, "learning_rate": 4.7099607118095676e-05, "loss": 0.0, "step": 753 }, { "epoch": 0.17, "learning_rate": 4.709575533471998e-05, "loss": 0.0, "step": 754 }, { "epoch": 0.17, "learning_rate": 4.709190355134427e-05, "loss": 0.0, "step": 755 }, { "epoch": 0.17, "learning_rate": 4.7088051767968574e-05, "loss": 0.0, "step": 756 }, { "epoch": 0.17, "learning_rate": 4.708419998459287e-05, "loss": 0.0, "step": 757 }, { "epoch": 0.18, "learning_rate": 4.7080348201217164e-05, "loss": 0.0, "step": 758 }, { "epoch": 0.18, "learning_rate": 4.7076496417841466e-05, "loss": 0.0, "step": 759 }, { "epoch": 0.18, "learning_rate": 4.707264463446576e-05, "loss": 0.0, "step": 760 }, { "epoch": 0.18, "learning_rate": 4.7068792851090056e-05, "loss": 0.0, "step": 761 }, { "epoch": 0.18, "learning_rate": 4.706494106771435e-05, "loss": 0.0001, "step": 762 }, { "epoch": 0.18, "learning_rate": 4.706108928433865e-05, "loss": 0.0, "step": 763 }, { "epoch": 0.18, "learning_rate": 4.705723750096295e-05, "loss": 0.0, "step": 764 }, { "epoch": 0.18, "learning_rate": 4.705338571758724e-05, "loss": 0.0002, "step": 765 }, { "epoch": 0.18, "learning_rate": 4.7049533934211544e-05, "loss": 0.0, "step": 766 }, { "epoch": 0.18, "learning_rate": 4.704568215083584e-05, "loss": 0.0, "step": 767 }, { "epoch": 0.18, "learning_rate": 4.704183036746014e-05, "loss": 0.0, "step": 768 }, { "epoch": 0.18, "learning_rate": 4.7037978584084436e-05, "loss": 0.0143, "step": 769 }, { "epoch": 0.18, "learning_rate": 4.703412680070873e-05, "loss": 0.0001, "step": 770 }, { "epoch": 0.18, "learning_rate": 4.7030275017333026e-05, "loss": 0.0, "step": 771 }, { "epoch": 0.18, "learning_rate": 4.702642323395732e-05, "loss": 0.0002, "step": 772 }, { "epoch": 0.18, "learning_rate": 4.702257145058162e-05, "loss": 0.0001, "step": 773 }, { "epoch": 0.18, "learning_rate": 4.701871966720592e-05, "loss": 0.0, "step": 774 }, { "epoch": 0.18, "learning_rate": 4.701486788383022e-05, "loss": 0.0, "step": 775 }, { "epoch": 0.18, "learning_rate": 4.7011016100454514e-05, "loss": 0.0001, "step": 776 }, { "epoch": 0.18, "learning_rate": 4.700716431707881e-05, "loss": 0.0002, "step": 777 }, { "epoch": 0.18, "learning_rate": 4.700331253370311e-05, "loss": 0.0, "step": 778 }, { "epoch": 0.18, "learning_rate": 4.6999460750327406e-05, "loss": 0.0, "step": 779 }, { "epoch": 0.18, "learning_rate": 4.69956089669517e-05, "loss": 0.0, "step": 780 }, { "epoch": 0.18, "learning_rate": 4.6991757183575996e-05, "loss": 0.0, "step": 781 }, { "epoch": 0.18, "learning_rate": 4.698790540020029e-05, "loss": 0.0, "step": 782 }, { "epoch": 0.18, "learning_rate": 4.698405361682459e-05, "loss": 0.0, "step": 783 }, { "epoch": 0.18, "learning_rate": 4.698020183344889e-05, "loss": 0.0, "step": 784 }, { "epoch": 0.18, "learning_rate": 4.697635005007319e-05, "loss": 0.0, "step": 785 }, { "epoch": 0.18, "learning_rate": 4.6972498266697484e-05, "loss": 0.0, "step": 786 }, { "epoch": 0.18, "learning_rate": 4.696864648332178e-05, "loss": 0.0, "step": 787 }, { "epoch": 0.18, "learning_rate": 4.696479469994608e-05, "loss": 0.0, "step": 788 }, { "epoch": 0.18, "learning_rate": 4.696094291657037e-05, "loss": 0.0001, "step": 789 }, { "epoch": 0.18, "learning_rate": 4.695709113319467e-05, "loss": 0.0, "step": 790 }, { "epoch": 0.18, "learning_rate": 4.6953239349818966e-05, "loss": 0.0, "step": 791 }, { "epoch": 0.18, "learning_rate": 4.694938756644327e-05, "loss": 0.0, "step": 792 }, { "epoch": 0.18, "learning_rate": 4.694553578306756e-05, "loss": 0.0, "step": 793 }, { "epoch": 0.18, "learning_rate": 4.694168399969186e-05, "loss": 0.0, "step": 794 }, { "epoch": 0.18, "learning_rate": 4.693783221631616e-05, "loss": 0.0, "step": 795 }, { "epoch": 0.18, "learning_rate": 4.6933980432940454e-05, "loss": 0.0, "step": 796 }, { "epoch": 0.18, "learning_rate": 4.6930128649564756e-05, "loss": 0.0, "step": 797 }, { "epoch": 0.18, "learning_rate": 4.6926276866189044e-05, "loss": 0.0, "step": 798 }, { "epoch": 0.18, "learning_rate": 4.6922425082813346e-05, "loss": 0.0, "step": 799 }, { "epoch": 0.18, "learning_rate": 4.691857329943764e-05, "loss": 0.0001, "step": 800 }, { "epoch": 0.19, "learning_rate": 4.6914721516061936e-05, "loss": 0.0, "step": 801 }, { "epoch": 0.19, "learning_rate": 4.691086973268624e-05, "loss": 0.0, "step": 802 }, { "epoch": 0.19, "learning_rate": 4.690701794931053e-05, "loss": 0.003, "step": 803 }, { "epoch": 0.19, "learning_rate": 4.6903166165934834e-05, "loss": 0.0003, "step": 804 }, { "epoch": 0.19, "learning_rate": 4.689931438255913e-05, "loss": 0.0, "step": 805 }, { "epoch": 0.19, "learning_rate": 4.6895462599183424e-05, "loss": 0.0, "step": 806 }, { "epoch": 0.19, "learning_rate": 4.6891610815807726e-05, "loss": 0.0, "step": 807 }, { "epoch": 0.19, "learning_rate": 4.6887759032432014e-05, "loss": 0.0, "step": 808 }, { "epoch": 0.19, "learning_rate": 4.6883907249056316e-05, "loss": 0.0, "step": 809 }, { "epoch": 0.19, "learning_rate": 4.688005546568061e-05, "loss": 0.0, "step": 810 }, { "epoch": 0.19, "learning_rate": 4.687620368230491e-05, "loss": 0.0, "step": 811 }, { "epoch": 0.19, "learning_rate": 4.687235189892921e-05, "loss": 0.0002, "step": 812 }, { "epoch": 0.19, "learning_rate": 4.68685001155535e-05, "loss": 0.0002, "step": 813 }, { "epoch": 0.19, "learning_rate": 4.6864648332177804e-05, "loss": 0.001, "step": 814 }, { "epoch": 0.19, "learning_rate": 4.68607965488021e-05, "loss": 0.0, "step": 815 }, { "epoch": 0.19, "learning_rate": 4.6856944765426394e-05, "loss": 0.0, "step": 816 }, { "epoch": 0.19, "learning_rate": 4.685309298205069e-05, "loss": 0.0, "step": 817 }, { "epoch": 0.19, "learning_rate": 4.6849241198674984e-05, "loss": 0.0, "step": 818 }, { "epoch": 0.19, "learning_rate": 4.6845389415299285e-05, "loss": 0.0, "step": 819 }, { "epoch": 0.19, "learning_rate": 4.684153763192358e-05, "loss": 0.0, "step": 820 }, { "epoch": 0.19, "learning_rate": 4.683768584854788e-05, "loss": 0.0, "step": 821 }, { "epoch": 0.19, "learning_rate": 4.683383406517218e-05, "loss": 0.0, "step": 822 }, { "epoch": 0.19, "learning_rate": 4.682998228179647e-05, "loss": 0.0003, "step": 823 }, { "epoch": 0.19, "learning_rate": 4.6826130498420774e-05, "loss": 0.0, "step": 824 }, { "epoch": 0.19, "learning_rate": 4.682227871504507e-05, "loss": 0.0, "step": 825 }, { "epoch": 0.19, "learning_rate": 4.6818426931669364e-05, "loss": 0.0025, "step": 826 }, { "epoch": 0.19, "learning_rate": 4.681457514829366e-05, "loss": 0.0, "step": 827 }, { "epoch": 0.19, "learning_rate": 4.681072336491796e-05, "loss": 0.0, "step": 828 }, { "epoch": 0.19, "learning_rate": 4.6806871581542255e-05, "loss": 0.0, "step": 829 }, { "epoch": 0.19, "learning_rate": 4.680301979816655e-05, "loss": 0.0011, "step": 830 }, { "epoch": 0.19, "learning_rate": 4.679916801479085e-05, "loss": 0.0, "step": 831 }, { "epoch": 0.19, "learning_rate": 4.679531623141515e-05, "loss": 0.0, "step": 832 }, { "epoch": 0.19, "learning_rate": 4.679146444803945e-05, "loss": 0.0004, "step": 833 }, { "epoch": 0.19, "learning_rate": 4.6787612664663744e-05, "loss": 0.0, "step": 834 }, { "epoch": 0.19, "learning_rate": 4.678376088128804e-05, "loss": 0.0001, "step": 835 }, { "epoch": 0.19, "learning_rate": 4.6779909097912334e-05, "loss": 0.0, "step": 836 }, { "epoch": 0.19, "learning_rate": 4.677605731453663e-05, "loss": 0.0, "step": 837 }, { "epoch": 0.19, "learning_rate": 4.677220553116093e-05, "loss": 0.0, "step": 838 }, { "epoch": 0.19, "learning_rate": 4.6768353747785225e-05, "loss": 0.0, "step": 839 }, { "epoch": 0.19, "learning_rate": 4.676450196440953e-05, "loss": 0.0, "step": 840 }, { "epoch": 0.19, "learning_rate": 4.676065018103382e-05, "loss": 0.0, "step": 841 }, { "epoch": 0.19, "learning_rate": 4.675679839765812e-05, "loss": 0.0, "step": 842 }, { "epoch": 0.19, "learning_rate": 4.675294661428242e-05, "loss": 0.0011, "step": 843 }, { "epoch": 0.2, "learning_rate": 4.6749094830906713e-05, "loss": 0.0, "step": 844 }, { "epoch": 0.2, "learning_rate": 4.674524304753101e-05, "loss": 0.0, "step": 845 }, { "epoch": 0.2, "learning_rate": 4.6741391264155303e-05, "loss": 0.0, "step": 846 }, { "epoch": 0.2, "learning_rate": 4.6737539480779605e-05, "loss": 0.0002, "step": 847 }, { "epoch": 0.2, "learning_rate": 4.67336876974039e-05, "loss": 0.0, "step": 848 }, { "epoch": 0.2, "learning_rate": 4.6729835914028195e-05, "loss": 0.0, "step": 849 }, { "epoch": 0.2, "learning_rate": 4.67259841306525e-05, "loss": 0.0, "step": 850 }, { "epoch": 0.2, "learning_rate": 4.672213234727679e-05, "loss": 0.0, "step": 851 }, { "epoch": 0.2, "learning_rate": 4.6718280563901093e-05, "loss": 0.0, "step": 852 }, { "epoch": 0.2, "learning_rate": 4.671442878052539e-05, "loss": 0.0, "step": 853 }, { "epoch": 0.2, "learning_rate": 4.6710576997149677e-05, "loss": 0.0, "step": 854 }, { "epoch": 0.2, "learning_rate": 4.670672521377398e-05, "loss": 0.0, "step": 855 }, { "epoch": 0.2, "learning_rate": 4.670287343039827e-05, "loss": 0.0, "step": 856 }, { "epoch": 0.2, "learning_rate": 4.6699021647022575e-05, "loss": 0.0, "step": 857 }, { "epoch": 0.2, "learning_rate": 4.669516986364687e-05, "loss": 0.0, "step": 858 }, { "epoch": 0.2, "learning_rate": 4.6691318080271165e-05, "loss": 0.0, "step": 859 }, { "epoch": 0.2, "learning_rate": 4.668746629689547e-05, "loss": 0.0, "step": 860 }, { "epoch": 0.2, "learning_rate": 4.668361451351976e-05, "loss": 0.0004, "step": 861 }, { "epoch": 0.2, "learning_rate": 4.667976273014406e-05, "loss": 0.0, "step": 862 }, { "epoch": 0.2, "learning_rate": 4.667591094676836e-05, "loss": 0.0, "step": 863 }, { "epoch": 0.2, "learning_rate": 4.667205916339265e-05, "loss": 0.0002, "step": 864 }, { "epoch": 0.2, "learning_rate": 4.666820738001695e-05, "loss": 0.0, "step": 865 }, { "epoch": 0.2, "learning_rate": 4.666435559664124e-05, "loss": 0.0, "step": 866 }, { "epoch": 0.2, "learning_rate": 4.6660503813265545e-05, "loss": 0.0, "step": 867 }, { "epoch": 0.2, "learning_rate": 4.665665202988984e-05, "loss": 0.0001, "step": 868 }, { "epoch": 0.2, "learning_rate": 4.665280024651414e-05, "loss": 0.0, "step": 869 }, { "epoch": 0.2, "learning_rate": 4.6648948463138437e-05, "loss": 0.0, "step": 870 }, { "epoch": 0.2, "learning_rate": 4.664509667976273e-05, "loss": 0.0001, "step": 871 }, { "epoch": 0.2, "learning_rate": 4.664124489638703e-05, "loss": 0.0, "step": 872 }, { "epoch": 0.2, "learning_rate": 4.663739311301132e-05, "loss": 0.0002, "step": 873 }, { "epoch": 0.2, "learning_rate": 4.663354132963562e-05, "loss": 0.0, "step": 874 }, { "epoch": 0.2, "learning_rate": 4.662968954625992e-05, "loss": 0.0, "step": 875 }, { "epoch": 0.2, "learning_rate": 4.662583776288422e-05, "loss": 0.0, "step": 876 }, { "epoch": 0.2, "learning_rate": 4.6621985979508515e-05, "loss": 0.0, "step": 877 }, { "epoch": 0.2, "learning_rate": 4.661813419613281e-05, "loss": 0.0032, "step": 878 }, { "epoch": 0.2, "learning_rate": 4.661428241275711e-05, "loss": 0.0, "step": 879 }, { "epoch": 0.2, "learning_rate": 4.6610430629381406e-05, "loss": 0.0, "step": 880 }, { "epoch": 0.2, "learning_rate": 4.660657884600571e-05, "loss": 0.0, "step": 881 }, { "epoch": 0.2, "learning_rate": 4.6602727062629996e-05, "loss": 0.0, "step": 882 }, { "epoch": 0.2, "learning_rate": 4.65988752792543e-05, "loss": 0.0, "step": 883 }, { "epoch": 0.2, "learning_rate": 4.659502349587859e-05, "loss": 0.0, "step": 884 }, { "epoch": 0.2, "learning_rate": 4.659117171250289e-05, "loss": 0.0002, "step": 885 }, { "epoch": 0.2, "learning_rate": 4.658731992912719e-05, "loss": 0.0001, "step": 886 }, { "epoch": 0.2, "learning_rate": 4.6583468145751485e-05, "loss": 0.0, "step": 887 }, { "epoch": 0.21, "learning_rate": 4.6579616362375786e-05, "loss": 0.0, "step": 888 }, { "epoch": 0.21, "learning_rate": 4.657576457900008e-05, "loss": 0.0, "step": 889 }, { "epoch": 0.21, "learning_rate": 4.6571912795624376e-05, "loss": 0.0001, "step": 890 }, { "epoch": 0.21, "learning_rate": 4.656806101224868e-05, "loss": 0.0006, "step": 891 }, { "epoch": 0.21, "learning_rate": 4.6564209228872966e-05, "loss": 0.0, "step": 892 }, { "epoch": 0.21, "learning_rate": 4.656035744549727e-05, "loss": 0.0, "step": 893 }, { "epoch": 0.21, "learning_rate": 4.655650566212156e-05, "loss": 0.0, "step": 894 }, { "epoch": 0.21, "learning_rate": 4.655265387874586e-05, "loss": 0.0001, "step": 895 }, { "epoch": 0.21, "learning_rate": 4.654880209537016e-05, "loss": 0.0, "step": 896 }, { "epoch": 0.21, "learning_rate": 4.6544950311994455e-05, "loss": 0.0, "step": 897 }, { "epoch": 0.21, "learning_rate": 4.6541098528618756e-05, "loss": 0.0, "step": 898 }, { "epoch": 0.21, "learning_rate": 4.653724674524305e-05, "loss": 0.0, "step": 899 }, { "epoch": 0.21, "learning_rate": 4.6533394961867346e-05, "loss": 0.0, "step": 900 }, { "epoch": 0.21, "learning_rate": 4.652954317849164e-05, "loss": 0.0, "step": 901 }, { "epoch": 0.21, "learning_rate": 4.6525691395115936e-05, "loss": 0.0001, "step": 902 }, { "epoch": 0.21, "learning_rate": 4.652183961174024e-05, "loss": 0.0, "step": 903 }, { "epoch": 0.21, "learning_rate": 4.651798782836453e-05, "loss": 0.0, "step": 904 }, { "epoch": 0.21, "learning_rate": 4.6514136044988835e-05, "loss": 0.0, "step": 905 }, { "epoch": 0.21, "learning_rate": 4.651028426161313e-05, "loss": 0.0, "step": 906 }, { "epoch": 0.21, "learning_rate": 4.6506432478237424e-05, "loss": 0.0001, "step": 907 }, { "epoch": 0.21, "learning_rate": 4.6502580694861726e-05, "loss": 0.0, "step": 908 }, { "epoch": 0.21, "learning_rate": 4.649872891148602e-05, "loss": 0.0, "step": 909 }, { "epoch": 0.21, "learning_rate": 4.6494877128110316e-05, "loss": 0.0, "step": 910 }, { "epoch": 0.21, "learning_rate": 4.649102534473461e-05, "loss": 0.0, "step": 911 }, { "epoch": 0.21, "learning_rate": 4.648717356135891e-05, "loss": 0.0001, "step": 912 }, { "epoch": 0.21, "learning_rate": 4.648332177798321e-05, "loss": 0.0, "step": 913 }, { "epoch": 0.21, "learning_rate": 4.64794699946075e-05, "loss": 0.0, "step": 914 }, { "epoch": 0.21, "learning_rate": 4.6475618211231804e-05, "loss": 0.0, "step": 915 }, { "epoch": 0.21, "learning_rate": 4.64717664278561e-05, "loss": 0.0, "step": 916 }, { "epoch": 0.21, "learning_rate": 4.64679146444804e-05, "loss": 0.0, "step": 917 }, { "epoch": 0.21, "learning_rate": 4.6464062861104696e-05, "loss": 0.0, "step": 918 }, { "epoch": 0.21, "learning_rate": 4.646021107772899e-05, "loss": 0.0, "step": 919 }, { "epoch": 0.21, "learning_rate": 4.6456359294353286e-05, "loss": 0.0, "step": 920 }, { "epoch": 0.21, "learning_rate": 4.645250751097758e-05, "loss": 0.0009, "step": 921 }, { "epoch": 0.21, "learning_rate": 4.644865572760188e-05, "loss": 0.0, "step": 922 }, { "epoch": 0.21, "learning_rate": 4.644480394422618e-05, "loss": 0.0001, "step": 923 }, { "epoch": 0.21, "learning_rate": 4.644095216085048e-05, "loss": 0.0, "step": 924 }, { "epoch": 0.21, "learning_rate": 4.6437100377474774e-05, "loss": 0.0, "step": 925 }, { "epoch": 0.21, "learning_rate": 4.643324859409907e-05, "loss": 0.0, "step": 926 }, { "epoch": 0.21, "learning_rate": 4.642939681072337e-05, "loss": 0.0, "step": 927 }, { "epoch": 0.21, "learning_rate": 4.6425545027347666e-05, "loss": 0.0002, "step": 928 }, { "epoch": 0.21, "learning_rate": 4.642169324397196e-05, "loss": 0.0, "step": 929 }, { "epoch": 0.21, "learning_rate": 4.6417841460596256e-05, "loss": 0.0, "step": 930 }, { "epoch": 0.22, "learning_rate": 4.641398967722055e-05, "loss": 0.0, "step": 931 }, { "epoch": 0.22, "learning_rate": 4.641013789384485e-05, "loss": 0.0, "step": 932 }, { "epoch": 0.22, "learning_rate": 4.640628611046915e-05, "loss": 0.0, "step": 933 }, { "epoch": 0.22, "learning_rate": 4.640243432709345e-05, "loss": 0.0, "step": 934 }, { "epoch": 0.22, "learning_rate": 4.6398582543717744e-05, "loss": 0.0, "step": 935 }, { "epoch": 0.22, "learning_rate": 4.639473076034204e-05, "loss": 0.0, "step": 936 }, { "epoch": 0.22, "learning_rate": 4.639087897696634e-05, "loss": 0.0, "step": 937 }, { "epoch": 0.22, "learning_rate": 4.638702719359063e-05, "loss": 0.0004, "step": 938 }, { "epoch": 0.22, "learning_rate": 4.638317541021493e-05, "loss": 0.0, "step": 939 }, { "epoch": 0.22, "learning_rate": 4.6379323626839226e-05, "loss": 0.0001, "step": 940 }, { "epoch": 0.22, "learning_rate": 4.637547184346353e-05, "loss": 0.0, "step": 941 }, { "epoch": 0.22, "learning_rate": 4.637162006008782e-05, "loss": 0.0, "step": 942 }, { "epoch": 0.22, "learning_rate": 4.636776827671212e-05, "loss": 0.0, "step": 943 }, { "epoch": 0.22, "learning_rate": 4.636391649333642e-05, "loss": 0.0, "step": 944 }, { "epoch": 0.22, "learning_rate": 4.6360064709960714e-05, "loss": 0.0001, "step": 945 }, { "epoch": 0.22, "learning_rate": 4.6356212926585016e-05, "loss": 0.0, "step": 946 }, { "epoch": 0.22, "learning_rate": 4.635236114320931e-05, "loss": 0.0, "step": 947 }, { "epoch": 0.22, "learning_rate": 4.6348509359833606e-05, "loss": 0.0, "step": 948 }, { "epoch": 0.22, "learning_rate": 4.63446575764579e-05, "loss": 0.0, "step": 949 }, { "epoch": 0.22, "learning_rate": 4.6340805793082196e-05, "loss": 0.0, "step": 950 }, { "epoch": 0.22, "learning_rate": 4.63369540097065e-05, "loss": 0.0, "step": 951 }, { "epoch": 0.22, "learning_rate": 4.633310222633079e-05, "loss": 0.0, "step": 952 }, { "epoch": 0.22, "learning_rate": 4.6329250442955094e-05, "loss": 0.0, "step": 953 }, { "epoch": 0.22, "learning_rate": 4.632539865957939e-05, "loss": 0.0, "step": 954 }, { "epoch": 0.22, "learning_rate": 4.6321546876203684e-05, "loss": 0.0, "step": 955 }, { "epoch": 0.22, "learning_rate": 4.6317695092827986e-05, "loss": 0.0002, "step": 956 }, { "epoch": 0.22, "learning_rate": 4.6313843309452274e-05, "loss": 0.0, "step": 957 }, { "epoch": 0.22, "learning_rate": 4.6309991526076576e-05, "loss": 0.0, "step": 958 }, { "epoch": 0.22, "learning_rate": 4.630613974270087e-05, "loss": 0.0, "step": 959 }, { "epoch": 0.22, "learning_rate": 4.630228795932517e-05, "loss": 0.0033, "step": 960 }, { "epoch": 0.22, "learning_rate": 4.629843617594947e-05, "loss": 0.0005, "step": 961 }, { "epoch": 0.22, "learning_rate": 4.629458439257376e-05, "loss": 0.0, "step": 962 }, { "epoch": 0.22, "learning_rate": 4.6290732609198064e-05, "loss": 0.0, "step": 963 }, { "epoch": 0.22, "learning_rate": 4.628688082582236e-05, "loss": 0.0, "step": 964 }, { "epoch": 0.22, "learning_rate": 4.628302904244666e-05, "loss": 0.0, "step": 965 }, { "epoch": 0.22, "learning_rate": 4.627917725907095e-05, "loss": 0.0, "step": 966 }, { "epoch": 0.22, "learning_rate": 4.6275325475695244e-05, "loss": 0.0, "step": 967 }, { "epoch": 0.22, "learning_rate": 4.6271473692319545e-05, "loss": 0.0002, "step": 968 }, { "epoch": 0.22, "learning_rate": 4.626762190894384e-05, "loss": 0.0, "step": 969 }, { "epoch": 0.22, "learning_rate": 4.626377012556814e-05, "loss": 0.0, "step": 970 }, { "epoch": 0.22, "learning_rate": 4.625991834219244e-05, "loss": 0.0, "step": 971 }, { "epoch": 0.22, "learning_rate": 4.625606655881673e-05, "loss": 0.0, "step": 972 }, { "epoch": 0.22, "learning_rate": 4.6252214775441034e-05, "loss": 0.0, "step": 973 }, { "epoch": 0.23, "learning_rate": 4.624836299206533e-05, "loss": 0.0, "step": 974 }, { "epoch": 0.23, "learning_rate": 4.624451120868963e-05, "loss": 0.0, "step": 975 }, { "epoch": 0.23, "learning_rate": 4.624065942531392e-05, "loss": 0.0003, "step": 976 }, { "epoch": 0.23, "learning_rate": 4.623680764193822e-05, "loss": 0.0, "step": 977 }, { "epoch": 0.23, "learning_rate": 4.6232955858562515e-05, "loss": 0.0036, "step": 978 }, { "epoch": 0.23, "learning_rate": 4.622910407518681e-05, "loss": 0.0, "step": 979 }, { "epoch": 0.23, "learning_rate": 4.622525229181111e-05, "loss": 0.0003, "step": 980 }, { "epoch": 0.23, "learning_rate": 4.622140050843541e-05, "loss": 0.0, "step": 981 }, { "epoch": 0.23, "learning_rate": 4.621754872505971e-05, "loss": 0.0, "step": 982 }, { "epoch": 0.23, "learning_rate": 4.6213696941684004e-05, "loss": 0.0001, "step": 983 }, { "epoch": 0.23, "learning_rate": 4.62098451583083e-05, "loss": 0.0, "step": 984 }, { "epoch": 0.23, "learning_rate": 4.6205993374932594e-05, "loss": 0.0, "step": 985 }, { "epoch": 0.23, "learning_rate": 4.620214159155689e-05, "loss": 0.0, "step": 986 }, { "epoch": 0.23, "learning_rate": 4.619828980818119e-05, "loss": 0.0, "step": 987 }, { "epoch": 0.23, "learning_rate": 4.6194438024805485e-05, "loss": 0.0, "step": 988 }, { "epoch": 0.23, "learning_rate": 4.619058624142979e-05, "loss": 0.0, "step": 989 }, { "epoch": 0.23, "learning_rate": 4.618673445805408e-05, "loss": 0.0, "step": 990 }, { "epoch": 0.23, "learning_rate": 4.618288267467838e-05, "loss": 0.0, "step": 991 }, { "epoch": 0.23, "learning_rate": 4.617903089130268e-05, "loss": 0.0, "step": 992 }, { "epoch": 0.23, "learning_rate": 4.6175179107926974e-05, "loss": 0.0, "step": 993 }, { "epoch": 0.23, "learning_rate": 4.617132732455127e-05, "loss": 0.0, "step": 994 }, { "epoch": 0.23, "learning_rate": 4.6167475541175563e-05, "loss": 0.0, "step": 995 }, { "epoch": 0.23, "learning_rate": 4.6163623757799865e-05, "loss": 0.0, "step": 996 }, { "epoch": 0.23, "learning_rate": 4.615977197442416e-05, "loss": 0.0, "step": 997 }, { "epoch": 0.23, "learning_rate": 4.6155920191048455e-05, "loss": 0.0005, "step": 998 }, { "epoch": 0.23, "learning_rate": 4.615206840767276e-05, "loss": 0.0, "step": 999 }, { "epoch": 0.23, "learning_rate": 4.614821662429705e-05, "loss": 0.0, "step": 1000 }, { "epoch": 0.23, "learning_rate": 4.6144364840921354e-05, "loss": 0.0, "step": 1001 }, { "epoch": 0.23, "learning_rate": 4.614051305754565e-05, "loss": 0.0, "step": 1002 }, { "epoch": 0.23, "learning_rate": 4.6136661274169943e-05, "loss": 0.0, "step": 1003 }, { "epoch": 0.23, "learning_rate": 4.613280949079424e-05, "loss": 0.0, "step": 1004 }, { "epoch": 0.23, "learning_rate": 4.612895770741853e-05, "loss": 0.0001, "step": 1005 }, { "epoch": 0.23, "learning_rate": 4.6125105924042835e-05, "loss": 0.0, "step": 1006 }, { "epoch": 0.23, "learning_rate": 4.612125414066713e-05, "loss": 0.0, "step": 1007 }, { "epoch": 0.23, "learning_rate": 4.6117402357291425e-05, "loss": 0.0, "step": 1008 }, { "epoch": 0.23, "learning_rate": 4.611355057391573e-05, "loss": 0.0, "step": 1009 }, { "epoch": 0.23, "learning_rate": 4.610969879054002e-05, "loss": 0.0, "step": 1010 }, { "epoch": 0.23, "learning_rate": 4.6105847007164323e-05, "loss": 0.0, "step": 1011 }, { "epoch": 0.23, "learning_rate": 4.610199522378862e-05, "loss": 0.0, "step": 1012 }, { "epoch": 0.23, "learning_rate": 4.609814344041291e-05, "loss": 0.0, "step": 1013 }, { "epoch": 0.23, "learning_rate": 4.609429165703721e-05, "loss": 0.0, "step": 1014 }, { "epoch": 0.23, "learning_rate": 4.60904398736615e-05, "loss": 0.0, "step": 1015 }, { "epoch": 0.23, "learning_rate": 4.6086588090285805e-05, "loss": 0.0, "step": 1016 }, { "epoch": 0.24, "learning_rate": 4.60827363069101e-05, "loss": 0.0, "step": 1017 }, { "epoch": 0.24, "learning_rate": 4.60788845235344e-05, "loss": 0.0, "step": 1018 }, { "epoch": 0.24, "learning_rate": 4.60750327401587e-05, "loss": 0.0, "step": 1019 }, { "epoch": 0.24, "learning_rate": 4.607118095678299e-05, "loss": 0.0, "step": 1020 }, { "epoch": 0.24, "learning_rate": 4.606732917340729e-05, "loss": 0.0, "step": 1021 }, { "epoch": 0.24, "learning_rate": 4.606347739003158e-05, "loss": 0.0001, "step": 1022 }, { "epoch": 0.24, "learning_rate": 4.605962560665588e-05, "loss": 0.0, "step": 1023 }, { "epoch": 0.24, "learning_rate": 4.605577382328018e-05, "loss": 0.0, "step": 1024 }, { "epoch": 0.24, "learning_rate": 4.605192203990448e-05, "loss": 0.0, "step": 1025 }, { "epoch": 0.24, "learning_rate": 4.6048070256528775e-05, "loss": 0.0, "step": 1026 }, { "epoch": 0.24, "learning_rate": 4.604421847315307e-05, "loss": 0.0, "step": 1027 }, { "epoch": 0.24, "learning_rate": 4.604036668977737e-05, "loss": 0.0, "step": 1028 }, { "epoch": 0.24, "learning_rate": 4.6036514906401667e-05, "loss": 0.0, "step": 1029 }, { "epoch": 0.24, "learning_rate": 4.603266312302597e-05, "loss": 0.0, "step": 1030 }, { "epoch": 0.24, "learning_rate": 4.602881133965026e-05, "loss": 0.0, "step": 1031 }, { "epoch": 0.24, "learning_rate": 4.602495955627456e-05, "loss": 0.0, "step": 1032 }, { "epoch": 0.24, "learning_rate": 4.602110777289885e-05, "loss": 0.0, "step": 1033 }, { "epoch": 0.24, "learning_rate": 4.601725598952315e-05, "loss": 0.0, "step": 1034 }, { "epoch": 0.24, "learning_rate": 4.601340420614745e-05, "loss": 0.0, "step": 1035 }, { "epoch": 0.24, "learning_rate": 4.6009552422771745e-05, "loss": 0.0, "step": 1036 }, { "epoch": 0.24, "learning_rate": 4.6005700639396046e-05, "loss": 0.0, "step": 1037 }, { "epoch": 0.24, "learning_rate": 4.600184885602034e-05, "loss": 0.0, "step": 1038 }, { "epoch": 0.24, "learning_rate": 4.5997997072644636e-05, "loss": 0.0, "step": 1039 }, { "epoch": 0.24, "learning_rate": 4.599414528926894e-05, "loss": 0.0, "step": 1040 }, { "epoch": 0.24, "learning_rate": 4.5990293505893226e-05, "loss": 0.0001, "step": 1041 }, { "epoch": 0.24, "learning_rate": 4.598644172251753e-05, "loss": 0.0001, "step": 1042 }, { "epoch": 0.24, "learning_rate": 4.598258993914182e-05, "loss": 0.0, "step": 1043 }, { "epoch": 0.24, "learning_rate": 4.597873815576612e-05, "loss": 0.0, "step": 1044 }, { "epoch": 0.24, "learning_rate": 4.597488637239042e-05, "loss": 0.0, "step": 1045 }, { "epoch": 0.24, "learning_rate": 4.5971034589014715e-05, "loss": 0.0, "step": 1046 }, { "epoch": 0.24, "learning_rate": 4.5967182805639016e-05, "loss": 0.0, "step": 1047 }, { "epoch": 0.24, "learning_rate": 4.596333102226331e-05, "loss": 0.0, "step": 1048 }, { "epoch": 0.24, "learning_rate": 4.5959479238887606e-05, "loss": 0.0, "step": 1049 }, { "epoch": 0.24, "learning_rate": 4.59556274555119e-05, "loss": 0.0001, "step": 1050 }, { "epoch": 0.24, "learning_rate": 4.5951775672136196e-05, "loss": 0.0, "step": 1051 }, { "epoch": 0.24, "learning_rate": 4.59479238887605e-05, "loss": 0.0001, "step": 1052 }, { "epoch": 0.24, "learning_rate": 4.594407210538479e-05, "loss": 0.0001, "step": 1053 }, { "epoch": 0.24, "learning_rate": 4.5940220322009095e-05, "loss": 0.0, "step": 1054 }, { "epoch": 0.24, "learning_rate": 4.593636853863339e-05, "loss": 0.0, "step": 1055 }, { "epoch": 0.24, "learning_rate": 4.5932516755257685e-05, "loss": 0.0, "step": 1056 }, { "epoch": 0.24, "learning_rate": 4.5928664971881986e-05, "loss": 0.0, "step": 1057 }, { "epoch": 0.24, "learning_rate": 4.592481318850628e-05, "loss": 0.0001, "step": 1058 }, { "epoch": 0.24, "learning_rate": 4.592096140513058e-05, "loss": 0.0004, "step": 1059 }, { "epoch": 0.24, "learning_rate": 4.591710962175487e-05, "loss": 0.0, "step": 1060 }, { "epoch": 0.25, "learning_rate": 4.591325783837917e-05, "loss": 0.0, "step": 1061 }, { "epoch": 0.25, "learning_rate": 4.590940605500347e-05, "loss": 0.0, "step": 1062 }, { "epoch": 0.25, "learning_rate": 4.590555427162776e-05, "loss": 0.0, "step": 1063 }, { "epoch": 0.25, "learning_rate": 4.5901702488252064e-05, "loss": 0.0, "step": 1064 }, { "epoch": 0.25, "learning_rate": 4.589785070487636e-05, "loss": 0.0, "step": 1065 }, { "epoch": 0.25, "learning_rate": 4.589399892150066e-05, "loss": 0.0, "step": 1066 }, { "epoch": 0.25, "learning_rate": 4.5890147138124956e-05, "loss": 0.0001, "step": 1067 }, { "epoch": 0.25, "learning_rate": 4.588629535474925e-05, "loss": 0.0, "step": 1068 }, { "epoch": 0.25, "learning_rate": 4.5882443571373546e-05, "loss": 0.0, "step": 1069 }, { "epoch": 0.25, "learning_rate": 4.587859178799784e-05, "loss": 0.0001, "step": 1070 }, { "epoch": 0.25, "learning_rate": 4.587474000462214e-05, "loss": 0.0002, "step": 1071 }, { "epoch": 0.25, "learning_rate": 4.587088822124644e-05, "loss": 0.0, "step": 1072 }, { "epoch": 0.25, "learning_rate": 4.586703643787074e-05, "loss": 0.0019, "step": 1073 }, { "epoch": 0.25, "learning_rate": 4.5863184654495034e-05, "loss": 0.0, "step": 1074 }, { "epoch": 0.25, "learning_rate": 4.585933287111933e-05, "loss": 0.0002, "step": 1075 }, { "epoch": 0.25, "learning_rate": 4.585548108774363e-05, "loss": 0.0, "step": 1076 }, { "epoch": 0.25, "learning_rate": 4.5851629304367926e-05, "loss": 0.0, "step": 1077 }, { "epoch": 0.25, "learning_rate": 4.584777752099222e-05, "loss": 0.0028, "step": 1078 }, { "epoch": 0.25, "learning_rate": 4.5843925737616516e-05, "loss": 0.0, "step": 1079 }, { "epoch": 0.25, "learning_rate": 4.584007395424081e-05, "loss": 0.0, "step": 1080 }, { "epoch": 0.25, "learning_rate": 4.583622217086511e-05, "loss": 0.0, "step": 1081 }, { "epoch": 0.25, "learning_rate": 4.583237038748941e-05, "loss": 0.0, "step": 1082 }, { "epoch": 0.25, "learning_rate": 4.582851860411371e-05, "loss": 0.0, "step": 1083 }, { "epoch": 0.25, "learning_rate": 4.5824666820738004e-05, "loss": 0.002, "step": 1084 }, { "epoch": 0.25, "learning_rate": 4.58208150373623e-05, "loss": 0.0, "step": 1085 }, { "epoch": 0.25, "learning_rate": 4.58169632539866e-05, "loss": 0.0, "step": 1086 }, { "epoch": 0.25, "learning_rate": 4.5813111470610896e-05, "loss": 0.0, "step": 1087 }, { "epoch": 0.25, "learning_rate": 4.580925968723519e-05, "loss": 0.0, "step": 1088 }, { "epoch": 0.25, "learning_rate": 4.5805407903859486e-05, "loss": 0.0, "step": 1089 }, { "epoch": 0.25, "learning_rate": 4.580155612048379e-05, "loss": 0.0, "step": 1090 }, { "epoch": 0.25, "learning_rate": 4.579770433710808e-05, "loss": 0.0, "step": 1091 }, { "epoch": 0.25, "learning_rate": 4.579385255373238e-05, "loss": 0.0, "step": 1092 }, { "epoch": 0.25, "learning_rate": 4.579000077035668e-05, "loss": 0.0, "step": 1093 }, { "epoch": 0.25, "learning_rate": 4.5786148986980974e-05, "loss": 0.0, "step": 1094 }, { "epoch": 0.25, "learning_rate": 4.5782297203605276e-05, "loss": 0.0013, "step": 1095 }, { "epoch": 0.25, "learning_rate": 4.577844542022957e-05, "loss": 0.0, "step": 1096 }, { "epoch": 0.25, "learning_rate": 4.5774593636853866e-05, "loss": 0.0001, "step": 1097 }, { "epoch": 0.25, "learning_rate": 4.577074185347816e-05, "loss": 0.0, "step": 1098 }, { "epoch": 0.25, "learning_rate": 4.5766890070102456e-05, "loss": 0.0002, "step": 1099 }, { "epoch": 0.25, "learning_rate": 4.576303828672676e-05, "loss": 0.0, "step": 1100 }, { "epoch": 0.25, "learning_rate": 4.575918650335105e-05, "loss": 0.0001, "step": 1101 }, { "epoch": 0.25, "learning_rate": 4.5755334719975354e-05, "loss": 0.0, "step": 1102 }, { "epoch": 0.25, "learning_rate": 4.575148293659965e-05, "loss": 0.0, "step": 1103 }, { "epoch": 0.26, "learning_rate": 4.5747631153223944e-05, "loss": 0.0, "step": 1104 }, { "epoch": 0.26, "learning_rate": 4.5743779369848246e-05, "loss": 0.0, "step": 1105 }, { "epoch": 0.26, "learning_rate": 4.5739927586472534e-05, "loss": 0.0, "step": 1106 }, { "epoch": 0.26, "learning_rate": 4.5736075803096836e-05, "loss": 0.0, "step": 1107 }, { "epoch": 0.26, "learning_rate": 4.573222401972113e-05, "loss": 0.0001, "step": 1108 }, { "epoch": 0.26, "learning_rate": 4.572837223634543e-05, "loss": 0.0, "step": 1109 }, { "epoch": 0.26, "learning_rate": 4.572452045296973e-05, "loss": 0.0, "step": 1110 }, { "epoch": 0.26, "learning_rate": 4.572066866959402e-05, "loss": 0.0, "step": 1111 }, { "epoch": 0.26, "learning_rate": 4.5716816886218324e-05, "loss": 0.0001, "step": 1112 }, { "epoch": 0.26, "learning_rate": 4.571296510284262e-05, "loss": 0.0006, "step": 1113 }, { "epoch": 0.26, "learning_rate": 4.570911331946692e-05, "loss": 0.0, "step": 1114 }, { "epoch": 0.26, "learning_rate": 4.5705261536091216e-05, "loss": 0.0, "step": 1115 }, { "epoch": 0.26, "learning_rate": 4.5701409752715504e-05, "loss": 0.0, "step": 1116 }, { "epoch": 0.26, "learning_rate": 4.5697557969339806e-05, "loss": 0.0, "step": 1117 }, { "epoch": 0.26, "learning_rate": 4.56937061859641e-05, "loss": 0.0, "step": 1118 }, { "epoch": 0.26, "learning_rate": 4.56898544025884e-05, "loss": 0.0, "step": 1119 }, { "epoch": 0.26, "learning_rate": 4.56860026192127e-05, "loss": 0.0, "step": 1120 }, { "epoch": 0.26, "learning_rate": 4.568215083583699e-05, "loss": 0.0001, "step": 1121 }, { "epoch": 0.26, "learning_rate": 4.5678299052461294e-05, "loss": 0.0001, "step": 1122 }, { "epoch": 0.26, "learning_rate": 4.567444726908559e-05, "loss": 0.0, "step": 1123 }, { "epoch": 0.26, "learning_rate": 4.567059548570989e-05, "loss": 0.0, "step": 1124 }, { "epoch": 0.26, "learning_rate": 4.566674370233418e-05, "loss": 0.0, "step": 1125 }, { "epoch": 0.26, "learning_rate": 4.566289191895848e-05, "loss": 0.0, "step": 1126 }, { "epoch": 0.26, "learning_rate": 4.5659040135582775e-05, "loss": 0.0, "step": 1127 }, { "epoch": 0.26, "learning_rate": 4.565518835220707e-05, "loss": 0.0, "step": 1128 }, { "epoch": 0.26, "learning_rate": 4.565133656883137e-05, "loss": 0.0, "step": 1129 }, { "epoch": 0.26, "learning_rate": 4.564748478545567e-05, "loss": 0.0001, "step": 1130 }, { "epoch": 0.26, "learning_rate": 4.564363300207997e-05, "loss": 0.0002, "step": 1131 }, { "epoch": 0.26, "learning_rate": 4.5639781218704264e-05, "loss": 0.0, "step": 1132 }, { "epoch": 0.26, "learning_rate": 4.563592943532856e-05, "loss": 0.0, "step": 1133 }, { "epoch": 0.26, "learning_rate": 4.5632077651952854e-05, "loss": 0.0003, "step": 1134 }, { "epoch": 0.26, "learning_rate": 4.562822586857715e-05, "loss": 0.0, "step": 1135 }, { "epoch": 0.26, "learning_rate": 4.562437408520145e-05, "loss": 0.0, "step": 1136 }, { "epoch": 0.26, "learning_rate": 4.5620522301825745e-05, "loss": 0.0, "step": 1137 }, { "epoch": 0.26, "learning_rate": 4.561667051845005e-05, "loss": 0.0, "step": 1138 }, { "epoch": 0.26, "learning_rate": 4.561281873507434e-05, "loss": 0.0, "step": 1139 }, { "epoch": 0.26, "learning_rate": 4.560896695169864e-05, "loss": 0.0, "step": 1140 }, { "epoch": 0.26, "learning_rate": 4.560511516832294e-05, "loss": 0.0, "step": 1141 }, { "epoch": 0.26, "learning_rate": 4.5601263384947234e-05, "loss": 0.0, "step": 1142 }, { "epoch": 0.26, "learning_rate": 4.5597411601571535e-05, "loss": 0.0, "step": 1143 }, { "epoch": 0.26, "learning_rate": 4.5593559818195824e-05, "loss": 0.0, "step": 1144 }, { "epoch": 0.26, "learning_rate": 4.5589708034820125e-05, "loss": 0.0, "step": 1145 }, { "epoch": 0.26, "learning_rate": 4.558585625144442e-05, "loss": 0.0, "step": 1146 }, { "epoch": 0.27, "learning_rate": 4.5582004468068715e-05, "loss": 0.0, "step": 1147 }, { "epoch": 0.27, "learning_rate": 4.557815268469302e-05, "loss": 0.0, "step": 1148 }, { "epoch": 0.27, "learning_rate": 4.557430090131731e-05, "loss": 0.0005, "step": 1149 }, { "epoch": 0.27, "learning_rate": 4.5570449117941614e-05, "loss": 0.0, "step": 1150 }, { "epoch": 0.27, "learning_rate": 4.556659733456591e-05, "loss": 0.0, "step": 1151 }, { "epoch": 0.27, "learning_rate": 4.5562745551190204e-05, "loss": 0.0, "step": 1152 }, { "epoch": 0.27, "learning_rate": 4.55588937678145e-05, "loss": 0.0, "step": 1153 }, { "epoch": 0.27, "learning_rate": 4.5555041984438793e-05, "loss": 0.0, "step": 1154 }, { "epoch": 0.27, "learning_rate": 4.5551190201063095e-05, "loss": 0.0, "step": 1155 }, { "epoch": 0.27, "learning_rate": 4.554733841768739e-05, "loss": 0.0, "step": 1156 }, { "epoch": 0.27, "learning_rate": 4.5543486634311685e-05, "loss": 0.0, "step": 1157 }, { "epoch": 0.27, "learning_rate": 4.553963485093599e-05, "loss": 0.0, "step": 1158 }, { "epoch": 0.27, "learning_rate": 4.553578306756028e-05, "loss": 0.0001, "step": 1159 }, { "epoch": 0.27, "learning_rate": 4.5531931284184584e-05, "loss": 0.0, "step": 1160 }, { "epoch": 0.27, "learning_rate": 4.552807950080888e-05, "loss": 0.0007, "step": 1161 }, { "epoch": 0.27, "learning_rate": 4.5524227717433173e-05, "loss": 0.0, "step": 1162 }, { "epoch": 0.27, "learning_rate": 4.552037593405747e-05, "loss": 0.0, "step": 1163 }, { "epoch": 0.27, "learning_rate": 4.551652415068176e-05, "loss": 0.0, "step": 1164 }, { "epoch": 0.27, "learning_rate": 4.5512672367306065e-05, "loss": 0.0, "step": 1165 }, { "epoch": 0.27, "learning_rate": 4.550882058393036e-05, "loss": 0.0, "step": 1166 }, { "epoch": 0.27, "learning_rate": 4.550496880055466e-05, "loss": 0.0, "step": 1167 }, { "epoch": 0.27, "learning_rate": 4.550111701717896e-05, "loss": 0.0, "step": 1168 }, { "epoch": 0.27, "learning_rate": 4.549726523380325e-05, "loss": 0.0, "step": 1169 }, { "epoch": 0.27, "learning_rate": 4.5493413450427553e-05, "loss": 0.0, "step": 1170 }, { "epoch": 0.27, "learning_rate": 4.548956166705185e-05, "loss": 0.0, "step": 1171 }, { "epoch": 0.27, "learning_rate": 4.548570988367614e-05, "loss": 0.0, "step": 1172 }, { "epoch": 0.27, "learning_rate": 4.548185810030044e-05, "loss": 0.0002, "step": 1173 }, { "epoch": 0.27, "learning_rate": 4.547800631692474e-05, "loss": 0.0001, "step": 1174 }, { "epoch": 0.27, "learning_rate": 4.5474154533549035e-05, "loss": 0.0, "step": 1175 }, { "epoch": 0.27, "learning_rate": 4.547030275017333e-05, "loss": 0.0, "step": 1176 }, { "epoch": 0.27, "learning_rate": 4.546645096679763e-05, "loss": 0.0, "step": 1177 }, { "epoch": 0.27, "learning_rate": 4.5462599183421927e-05, "loss": 0.0, "step": 1178 }, { "epoch": 0.27, "learning_rate": 4.545874740004623e-05, "loss": 0.0, "step": 1179 }, { "epoch": 0.27, "learning_rate": 4.545489561667052e-05, "loss": 0.0, "step": 1180 }, { "epoch": 0.27, "learning_rate": 4.545104383329482e-05, "loss": 0.0, "step": 1181 }, { "epoch": 0.27, "learning_rate": 4.544719204991911e-05, "loss": 0.0, "step": 1182 }, { "epoch": 0.27, "learning_rate": 4.544334026654341e-05, "loss": 0.0, "step": 1183 }, { "epoch": 0.27, "learning_rate": 4.543948848316771e-05, "loss": 0.0, "step": 1184 }, { "epoch": 0.27, "learning_rate": 4.5435636699792005e-05, "loss": 0.0, "step": 1185 }, { "epoch": 0.27, "learning_rate": 4.5431784916416307e-05, "loss": 0.0093, "step": 1186 }, { "epoch": 0.27, "learning_rate": 4.54279331330406e-05, "loss": 0.0, "step": 1187 }, { "epoch": 0.27, "learning_rate": 4.5424081349664896e-05, "loss": 0.0, "step": 1188 }, { "epoch": 0.27, "learning_rate": 4.54202295662892e-05, "loss": 0.0, "step": 1189 }, { "epoch": 0.28, "learning_rate": 4.5416377782913486e-05, "loss": 0.0001, "step": 1190 }, { "epoch": 0.28, "learning_rate": 4.541252599953779e-05, "loss": 0.0, "step": 1191 }, { "epoch": 0.28, "learning_rate": 4.540867421616208e-05, "loss": 0.0, "step": 1192 }, { "epoch": 0.28, "learning_rate": 4.540482243278638e-05, "loss": 0.0, "step": 1193 }, { "epoch": 0.28, "learning_rate": 4.540097064941068e-05, "loss": 0.0, "step": 1194 }, { "epoch": 0.28, "learning_rate": 4.5397118866034975e-05, "loss": 0.0001, "step": 1195 }, { "epoch": 0.28, "learning_rate": 4.5393267082659276e-05, "loss": 0.0, "step": 1196 }, { "epoch": 0.28, "learning_rate": 4.538941529928357e-05, "loss": 0.0002, "step": 1197 }, { "epoch": 0.28, "learning_rate": 4.5385563515907866e-05, "loss": 0.0, "step": 1198 }, { "epoch": 0.28, "learning_rate": 4.538171173253217e-05, "loss": 0.0, "step": 1199 }, { "epoch": 0.28, "learning_rate": 4.5377859949156456e-05, "loss": 0.0, "step": 1200 }, { "epoch": 0.28, "learning_rate": 4.537400816578076e-05, "loss": 0.0, "step": 1201 }, { "epoch": 0.28, "learning_rate": 4.537015638240505e-05, "loss": 0.0057, "step": 1202 }, { "epoch": 0.28, "learning_rate": 4.5366304599029355e-05, "loss": 0.0, "step": 1203 }, { "epoch": 0.28, "learning_rate": 4.536245281565365e-05, "loss": 0.0, "step": 1204 }, { "epoch": 0.28, "learning_rate": 4.5358601032277945e-05, "loss": 0.0, "step": 1205 }, { "epoch": 0.28, "learning_rate": 4.5354749248902246e-05, "loss": 0.0, "step": 1206 }, { "epoch": 0.28, "learning_rate": 4.535089746552654e-05, "loss": 0.0, "step": 1207 }, { "epoch": 0.28, "learning_rate": 4.534704568215084e-05, "loss": 0.0002, "step": 1208 }, { "epoch": 0.28, "learning_rate": 4.534319389877513e-05, "loss": 0.0, "step": 1209 }, { "epoch": 0.28, "learning_rate": 4.533934211539943e-05, "loss": 0.0, "step": 1210 }, { "epoch": 0.28, "learning_rate": 4.533549033202373e-05, "loss": 0.0, "step": 1211 }, { "epoch": 0.28, "learning_rate": 4.533163854864802e-05, "loss": 0.0002, "step": 1212 }, { "epoch": 0.28, "learning_rate": 4.5327786765272325e-05, "loss": 0.0, "step": 1213 }, { "epoch": 0.28, "learning_rate": 4.532393498189662e-05, "loss": 0.0, "step": 1214 }, { "epoch": 0.28, "learning_rate": 4.532008319852092e-05, "loss": 0.0, "step": 1215 }, { "epoch": 0.28, "learning_rate": 4.5316231415145216e-05, "loss": 0.0001, "step": 1216 }, { "epoch": 0.28, "learning_rate": 4.531237963176951e-05, "loss": 0.0, "step": 1217 }, { "epoch": 0.28, "learning_rate": 4.5308527848393806e-05, "loss": 0.0, "step": 1218 }, { "epoch": 0.28, "learning_rate": 4.53046760650181e-05, "loss": 0.0, "step": 1219 }, { "epoch": 0.28, "learning_rate": 4.53008242816424e-05, "loss": 0.0, "step": 1220 }, { "epoch": 0.28, "learning_rate": 4.52969724982667e-05, "loss": 0.0, "step": 1221 }, { "epoch": 0.28, "learning_rate": 4.5293120714891e-05, "loss": 0.0006, "step": 1222 }, { "epoch": 0.28, "learning_rate": 4.5289268931515294e-05, "loss": 0.0, "step": 1223 }, { "epoch": 0.28, "learning_rate": 4.528541714813959e-05, "loss": 0.0, "step": 1224 }, { "epoch": 0.28, "learning_rate": 4.528156536476389e-05, "loss": 0.0, "step": 1225 }, { "epoch": 0.28, "learning_rate": 4.5277713581388186e-05, "loss": 0.0, "step": 1226 }, { "epoch": 0.28, "learning_rate": 4.527386179801249e-05, "loss": 0.0, "step": 1227 }, { "epoch": 0.28, "learning_rate": 4.5270010014636776e-05, "loss": 0.0, "step": 1228 }, { "epoch": 0.28, "learning_rate": 4.526615823126107e-05, "loss": 0.0, "step": 1229 }, { "epoch": 0.28, "learning_rate": 4.526230644788537e-05, "loss": 0.0003, "step": 1230 }, { "epoch": 0.28, "learning_rate": 4.525845466450967e-05, "loss": 0.0, "step": 1231 }, { "epoch": 0.28, "learning_rate": 4.525460288113397e-05, "loss": 0.0002, "step": 1232 }, { "epoch": 0.28, "learning_rate": 4.5250751097758264e-05, "loss": 0.0, "step": 1233 }, { "epoch": 0.29, "learning_rate": 4.524689931438256e-05, "loss": 0.0, "step": 1234 }, { "epoch": 0.29, "learning_rate": 4.524304753100686e-05, "loss": 0.0, "step": 1235 }, { "epoch": 0.29, "learning_rate": 4.5239195747631156e-05, "loss": 0.0, "step": 1236 }, { "epoch": 0.29, "learning_rate": 4.523534396425545e-05, "loss": 0.0, "step": 1237 }, { "epoch": 0.29, "learning_rate": 4.5231492180879746e-05, "loss": 0.0, "step": 1238 }, { "epoch": 0.29, "learning_rate": 4.522764039750405e-05, "loss": 0.0, "step": 1239 }, { "epoch": 0.29, "learning_rate": 4.522378861412834e-05, "loss": 0.0, "step": 1240 }, { "epoch": 0.29, "learning_rate": 4.521993683075264e-05, "loss": 0.0, "step": 1241 }, { "epoch": 0.29, "learning_rate": 4.521608504737694e-05, "loss": 0.0, "step": 1242 }, { "epoch": 0.29, "learning_rate": 4.5212233264001234e-05, "loss": 0.0, "step": 1243 }, { "epoch": 0.29, "learning_rate": 4.5208381480625536e-05, "loss": 0.0, "step": 1244 }, { "epoch": 0.29, "learning_rate": 4.520452969724983e-05, "loss": 0.0, "step": 1245 }, { "epoch": 0.29, "learning_rate": 4.5200677913874126e-05, "loss": 0.0, "step": 1246 }, { "epoch": 0.29, "learning_rate": 4.519682613049842e-05, "loss": 0.0, "step": 1247 }, { "epoch": 0.29, "learning_rate": 4.5192974347122716e-05, "loss": 0.0, "step": 1248 }, { "epoch": 0.29, "learning_rate": 4.518912256374702e-05, "loss": 0.0, "step": 1249 }, { "epoch": 0.29, "learning_rate": 4.518527078037131e-05, "loss": 0.0, "step": 1250 }, { "epoch": 0.29, "learning_rate": 4.5181418996995614e-05, "loss": 0.0001, "step": 1251 }, { "epoch": 0.29, "learning_rate": 4.517756721361991e-05, "loss": 0.0001, "step": 1252 }, { "epoch": 0.29, "learning_rate": 4.5173715430244204e-05, "loss": 0.0, "step": 1253 }, { "epoch": 0.29, "learning_rate": 4.5169863646868506e-05, "loss": 0.0, "step": 1254 }, { "epoch": 0.29, "learning_rate": 4.51660118634928e-05, "loss": 0.0, "step": 1255 }, { "epoch": 0.29, "learning_rate": 4.5162160080117096e-05, "loss": 0.0, "step": 1256 }, { "epoch": 0.29, "learning_rate": 4.515830829674139e-05, "loss": 0.0, "step": 1257 }, { "epoch": 0.29, "learning_rate": 4.515445651336569e-05, "loss": 0.0, "step": 1258 }, { "epoch": 0.29, "learning_rate": 4.515060472998999e-05, "loss": 0.0, "step": 1259 }, { "epoch": 0.29, "learning_rate": 4.514675294661428e-05, "loss": 0.0, "step": 1260 }, { "epoch": 0.29, "learning_rate": 4.5142901163238584e-05, "loss": 0.0, "step": 1261 }, { "epoch": 0.29, "learning_rate": 4.513904937986288e-05, "loss": 0.0, "step": 1262 }, { "epoch": 0.29, "learning_rate": 4.513519759648718e-05, "loss": 0.0, "step": 1263 }, { "epoch": 0.29, "learning_rate": 4.5131345813111476e-05, "loss": 0.0, "step": 1264 }, { "epoch": 0.29, "learning_rate": 4.512749402973577e-05, "loss": 0.0, "step": 1265 }, { "epoch": 0.29, "learning_rate": 4.5123642246360066e-05, "loss": 0.0, "step": 1266 }, { "epoch": 0.29, "learning_rate": 4.511979046298436e-05, "loss": 0.0009, "step": 1267 }, { "epoch": 0.29, "learning_rate": 4.511593867960866e-05, "loss": 0.0001, "step": 1268 }, { "epoch": 0.29, "learning_rate": 4.511208689623296e-05, "loss": 0.0, "step": 1269 }, { "epoch": 0.29, "learning_rate": 4.510823511285725e-05, "loss": 0.0001, "step": 1270 }, { "epoch": 0.29, "learning_rate": 4.5104383329481554e-05, "loss": 0.0, "step": 1271 }, { "epoch": 0.29, "learning_rate": 4.510053154610585e-05, "loss": 0.0, "step": 1272 }, { "epoch": 0.29, "learning_rate": 4.509667976273015e-05, "loss": 0.0, "step": 1273 }, { "epoch": 0.29, "learning_rate": 4.509282797935444e-05, "loss": 0.0, "step": 1274 }, { "epoch": 0.29, "learning_rate": 4.508897619597874e-05, "loss": 0.0001, "step": 1275 }, { "epoch": 0.29, "learning_rate": 4.5085124412603036e-05, "loss": 0.0, "step": 1276 }, { "epoch": 0.3, "learning_rate": 4.508127262922733e-05, "loss": 0.0, "step": 1277 }, { "epoch": 0.3, "learning_rate": 4.507742084585163e-05, "loss": 0.0, "step": 1278 }, { "epoch": 0.3, "learning_rate": 4.507356906247593e-05, "loss": 0.0, "step": 1279 }, { "epoch": 0.3, "learning_rate": 4.506971727910023e-05, "loss": 0.0, "step": 1280 }, { "epoch": 0.3, "learning_rate": 4.5065865495724524e-05, "loss": 0.0, "step": 1281 }, { "epoch": 0.3, "learning_rate": 4.506201371234882e-05, "loss": 0.0, "step": 1282 }, { "epoch": 0.3, "learning_rate": 4.505816192897312e-05, "loss": 0.0, "step": 1283 }, { "epoch": 0.3, "learning_rate": 4.505431014559741e-05, "loss": 0.0, "step": 1284 }, { "epoch": 0.3, "learning_rate": 4.505045836222171e-05, "loss": 0.0, "step": 1285 }, { "epoch": 0.3, "learning_rate": 4.5046606578846005e-05, "loss": 0.0, "step": 1286 }, { "epoch": 0.3, "learning_rate": 4.504275479547031e-05, "loss": 0.0, "step": 1287 }, { "epoch": 0.3, "learning_rate": 4.50389030120946e-05, "loss": 0.0, "step": 1288 }, { "epoch": 0.3, "learning_rate": 4.50350512287189e-05, "loss": 0.0, "step": 1289 }, { "epoch": 0.3, "learning_rate": 4.50311994453432e-05, "loss": 0.0004, "step": 1290 }, { "epoch": 0.3, "learning_rate": 4.5027347661967494e-05, "loss": 0.0, "step": 1291 }, { "epoch": 0.3, "learning_rate": 4.5023495878591795e-05, "loss": 0.0, "step": 1292 }, { "epoch": 0.3, "learning_rate": 4.5019644095216084e-05, "loss": 0.0, "step": 1293 }, { "epoch": 0.3, "learning_rate": 4.5015792311840385e-05, "loss": 0.0, "step": 1294 }, { "epoch": 0.3, "learning_rate": 4.501194052846468e-05, "loss": 0.0, "step": 1295 }, { "epoch": 0.3, "learning_rate": 4.5008088745088975e-05, "loss": 0.0, "step": 1296 }, { "epoch": 0.3, "learning_rate": 4.500423696171328e-05, "loss": 0.0, "step": 1297 }, { "epoch": 0.3, "learning_rate": 4.500038517833757e-05, "loss": 0.0, "step": 1298 }, { "epoch": 0.3, "learning_rate": 4.4996533394961874e-05, "loss": 0.0, "step": 1299 }, { "epoch": 0.3, "learning_rate": 4.499268161158617e-05, "loss": 0.0, "step": 1300 }, { "epoch": 0.3, "learning_rate": 4.4988829828210464e-05, "loss": 0.0004, "step": 1301 }, { "epoch": 0.3, "learning_rate": 4.498497804483476e-05, "loss": 0.0, "step": 1302 }, { "epoch": 0.3, "learning_rate": 4.4981126261459054e-05, "loss": 0.0, "step": 1303 }, { "epoch": 0.3, "learning_rate": 4.4977274478083355e-05, "loss": 0.0, "step": 1304 }, { "epoch": 0.3, "learning_rate": 4.497342269470765e-05, "loss": 0.0, "step": 1305 }, { "epoch": 0.3, "learning_rate": 4.4969570911331945e-05, "loss": 0.0, "step": 1306 }, { "epoch": 0.3, "learning_rate": 4.496571912795625e-05, "loss": 0.0, "step": 1307 }, { "epoch": 0.3, "learning_rate": 4.496186734458054e-05, "loss": 0.0, "step": 1308 }, { "epoch": 0.3, "learning_rate": 4.4958015561204844e-05, "loss": 0.0, "step": 1309 }, { "epoch": 0.3, "learning_rate": 4.495416377782914e-05, "loss": 0.0, "step": 1310 }, { "epoch": 0.3, "learning_rate": 4.4950311994453434e-05, "loss": 0.0, "step": 1311 }, { "epoch": 0.3, "learning_rate": 4.494646021107773e-05, "loss": 0.0, "step": 1312 }, { "epoch": 0.3, "learning_rate": 4.4942608427702023e-05, "loss": 0.0001, "step": 1313 }, { "epoch": 0.3, "learning_rate": 4.4938756644326325e-05, "loss": 0.0, "step": 1314 }, { "epoch": 0.3, "learning_rate": 4.493490486095062e-05, "loss": 0.0, "step": 1315 }, { "epoch": 0.3, "learning_rate": 4.493105307757492e-05, "loss": 0.0, "step": 1316 }, { "epoch": 0.3, "learning_rate": 4.492720129419922e-05, "loss": 0.0, "step": 1317 }, { "epoch": 0.3, "learning_rate": 4.492334951082351e-05, "loss": 0.0001, "step": 1318 }, { "epoch": 0.3, "learning_rate": 4.4919497727447813e-05, "loss": 0.0, "step": 1319 }, { "epoch": 0.31, "learning_rate": 4.491564594407211e-05, "loss": 0.0, "step": 1320 }, { "epoch": 0.31, "learning_rate": 4.4911794160696403e-05, "loss": 0.0, "step": 1321 }, { "epoch": 0.31, "learning_rate": 4.49079423773207e-05, "loss": 0.0, "step": 1322 }, { "epoch": 0.31, "learning_rate": 4.4904090593945e-05, "loss": 0.0, "step": 1323 }, { "epoch": 0.31, "learning_rate": 4.4900238810569295e-05, "loss": 0.0, "step": 1324 }, { "epoch": 0.31, "learning_rate": 4.489638702719359e-05, "loss": 0.0005, "step": 1325 }, { "epoch": 0.31, "learning_rate": 4.489253524381789e-05, "loss": 0.0001, "step": 1326 }, { "epoch": 0.31, "learning_rate": 4.488868346044219e-05, "loss": 0.0, "step": 1327 }, { "epoch": 0.31, "learning_rate": 4.488483167706649e-05, "loss": 0.0, "step": 1328 }, { "epoch": 0.31, "learning_rate": 4.488097989369078e-05, "loss": 0.0, "step": 1329 }, { "epoch": 0.31, "learning_rate": 4.487712811031508e-05, "loss": 0.0, "step": 1330 }, { "epoch": 0.31, "learning_rate": 4.487327632693937e-05, "loss": 0.0, "step": 1331 }, { "epoch": 0.31, "learning_rate": 4.486942454356367e-05, "loss": 0.0, "step": 1332 }, { "epoch": 0.31, "learning_rate": 4.486557276018797e-05, "loss": 0.0, "step": 1333 }, { "epoch": 0.31, "learning_rate": 4.4861720976812265e-05, "loss": 0.0, "step": 1334 }, { "epoch": 0.31, "learning_rate": 4.485786919343657e-05, "loss": 0.0045, "step": 1335 }, { "epoch": 0.31, "learning_rate": 4.485401741006086e-05, "loss": 0.0001, "step": 1336 }, { "epoch": 0.31, "learning_rate": 4.4850165626685157e-05, "loss": 0.0, "step": 1337 }, { "epoch": 0.31, "learning_rate": 4.484631384330946e-05, "loss": 0.0, "step": 1338 }, { "epoch": 0.31, "learning_rate": 4.484246205993375e-05, "loss": 0.0, "step": 1339 }, { "epoch": 0.31, "learning_rate": 4.483861027655805e-05, "loss": 0.0, "step": 1340 }, { "epoch": 0.31, "learning_rate": 4.483475849318234e-05, "loss": 0.0, "step": 1341 }, { "epoch": 0.31, "learning_rate": 4.4830906709806645e-05, "loss": 0.0, "step": 1342 }, { "epoch": 0.31, "learning_rate": 4.482705492643094e-05, "loss": 0.0, "step": 1343 }, { "epoch": 0.31, "learning_rate": 4.4823203143055235e-05, "loss": 0.0, "step": 1344 }, { "epoch": 0.31, "learning_rate": 4.4819351359679537e-05, "loss": 0.0, "step": 1345 }, { "epoch": 0.31, "learning_rate": 4.481549957630383e-05, "loss": 0.0, "step": 1346 }, { "epoch": 0.31, "learning_rate": 4.4811647792928126e-05, "loss": 0.0, "step": 1347 }, { "epoch": 0.31, "learning_rate": 4.480779600955243e-05, "loss": 0.0, "step": 1348 }, { "epoch": 0.31, "learning_rate": 4.4803944226176716e-05, "loss": 0.0, "step": 1349 }, { "epoch": 0.31, "learning_rate": 4.480009244280102e-05, "loss": 0.0, "step": 1350 }, { "epoch": 0.31, "learning_rate": 4.479624065942531e-05, "loss": 0.0, "step": 1351 }, { "epoch": 0.31, "learning_rate": 4.4792388876049615e-05, "loss": 0.0, "step": 1352 }, { "epoch": 0.31, "learning_rate": 4.478853709267391e-05, "loss": 0.0, "step": 1353 }, { "epoch": 0.31, "learning_rate": 4.4784685309298205e-05, "loss": 0.0, "step": 1354 }, { "epoch": 0.31, "learning_rate": 4.4780833525922506e-05, "loss": 0.0, "step": 1355 }, { "epoch": 0.31, "learning_rate": 4.47769817425468e-05, "loss": 0.0001, "step": 1356 }, { "epoch": 0.31, "learning_rate": 4.47731299591711e-05, "loss": 0.0, "step": 1357 }, { "epoch": 0.31, "learning_rate": 4.476927817579539e-05, "loss": 0.0, "step": 1358 }, { "epoch": 0.31, "learning_rate": 4.476542639241969e-05, "loss": 0.0, "step": 1359 }, { "epoch": 0.31, "learning_rate": 4.476157460904399e-05, "loss": 0.0, "step": 1360 }, { "epoch": 0.31, "learning_rate": 4.475772282566828e-05, "loss": 0.0, "step": 1361 }, { "epoch": 0.31, "learning_rate": 4.4753871042292585e-05, "loss": 0.0, "step": 1362 }, { "epoch": 0.31, "learning_rate": 4.475001925891688e-05, "loss": 0.0, "step": 1363 }, { "epoch": 0.32, "learning_rate": 4.474616747554118e-05, "loss": 0.0, "step": 1364 }, { "epoch": 0.32, "learning_rate": 4.4742315692165476e-05, "loss": 0.0, "step": 1365 }, { "epoch": 0.32, "learning_rate": 4.473846390878977e-05, "loss": 0.0, "step": 1366 }, { "epoch": 0.32, "learning_rate": 4.473461212541407e-05, "loss": 0.0, "step": 1367 }, { "epoch": 0.32, "learning_rate": 4.473076034203836e-05, "loss": 0.0, "step": 1368 }, { "epoch": 0.32, "learning_rate": 4.472690855866266e-05, "loss": 0.0, "step": 1369 }, { "epoch": 0.32, "learning_rate": 4.472305677528696e-05, "loss": 0.0001, "step": 1370 }, { "epoch": 0.32, "learning_rate": 4.471920499191126e-05, "loss": 0.0, "step": 1371 }, { "epoch": 0.32, "learning_rate": 4.4715353208535555e-05, "loss": 0.0, "step": 1372 }, { "epoch": 0.32, "learning_rate": 4.471150142515985e-05, "loss": 0.0, "step": 1373 }, { "epoch": 0.32, "learning_rate": 4.470764964178415e-05, "loss": 0.0, "step": 1374 }, { "epoch": 0.32, "learning_rate": 4.4703797858408446e-05, "loss": 0.0, "step": 1375 }, { "epoch": 0.32, "learning_rate": 4.469994607503275e-05, "loss": 0.0, "step": 1376 }, { "epoch": 0.32, "learning_rate": 4.4696094291657036e-05, "loss": 0.0002, "step": 1377 }, { "epoch": 0.32, "learning_rate": 4.469224250828134e-05, "loss": 0.0, "step": 1378 }, { "epoch": 0.32, "learning_rate": 4.468839072490563e-05, "loss": 0.0, "step": 1379 }, { "epoch": 0.32, "learning_rate": 4.468453894152993e-05, "loss": 0.0, "step": 1380 }, { "epoch": 0.32, "learning_rate": 4.468068715815423e-05, "loss": 0.0, "step": 1381 }, { "epoch": 0.32, "learning_rate": 4.4676835374778524e-05, "loss": 0.0, "step": 1382 }, { "epoch": 0.32, "learning_rate": 4.467298359140282e-05, "loss": 0.0, "step": 1383 }, { "epoch": 0.32, "learning_rate": 4.466913180802712e-05, "loss": 0.0, "step": 1384 }, { "epoch": 0.32, "learning_rate": 4.4665280024651416e-05, "loss": 0.0, "step": 1385 }, { "epoch": 0.32, "learning_rate": 4.466142824127571e-05, "loss": 0.0, "step": 1386 }, { "epoch": 0.32, "learning_rate": 4.4657576457900006e-05, "loss": 0.0, "step": 1387 }, { "epoch": 0.32, "learning_rate": 4.465372467452431e-05, "loss": 0.0, "step": 1388 }, { "epoch": 0.32, "learning_rate": 4.46498728911486e-05, "loss": 0.0, "step": 1389 }, { "epoch": 0.32, "learning_rate": 4.46460211077729e-05, "loss": 0.0, "step": 1390 }, { "epoch": 0.32, "learning_rate": 4.46421693243972e-05, "loss": 0.0, "step": 1391 }, { "epoch": 0.32, "learning_rate": 4.4638317541021494e-05, "loss": 0.0, "step": 1392 }, { "epoch": 0.32, "learning_rate": 4.4634465757645796e-05, "loss": 0.0, "step": 1393 }, { "epoch": 0.32, "learning_rate": 4.463061397427009e-05, "loss": 0.0, "step": 1394 }, { "epoch": 0.32, "learning_rate": 4.4626762190894386e-05, "loss": 0.0014, "step": 1395 }, { "epoch": 0.32, "learning_rate": 4.462291040751868e-05, "loss": 0.0, "step": 1396 }, { "epoch": 0.32, "learning_rate": 4.4619058624142976e-05, "loss": 0.0, "step": 1397 }, { "epoch": 0.32, "learning_rate": 4.461520684076728e-05, "loss": 0.0, "step": 1398 }, { "epoch": 0.32, "learning_rate": 4.461135505739157e-05, "loss": 0.0, "step": 1399 }, { "epoch": 0.32, "learning_rate": 4.4607503274015874e-05, "loss": 0.0, "step": 1400 }, { "epoch": 0.32, "learning_rate": 4.460365149064017e-05, "loss": 0.0001, "step": 1401 }, { "epoch": 0.32, "learning_rate": 4.4599799707264464e-05, "loss": 0.0, "step": 1402 }, { "epoch": 0.32, "learning_rate": 4.4595947923888766e-05, "loss": 0.0, "step": 1403 }, { "epoch": 0.32, "learning_rate": 4.459209614051306e-05, "loss": 0.0, "step": 1404 }, { "epoch": 0.32, "learning_rate": 4.4588244357137356e-05, "loss": 0.0, "step": 1405 }, { "epoch": 0.32, "learning_rate": 4.458439257376165e-05, "loss": 0.0, "step": 1406 }, { "epoch": 0.33, "learning_rate": 4.458054079038595e-05, "loss": 0.0, "step": 1407 }, { "epoch": 0.33, "learning_rate": 4.457668900701025e-05, "loss": 0.0, "step": 1408 }, { "epoch": 0.33, "learning_rate": 4.457283722363454e-05, "loss": 0.0, "step": 1409 }, { "epoch": 0.33, "learning_rate": 4.4568985440258844e-05, "loss": 0.0, "step": 1410 }, { "epoch": 0.33, "learning_rate": 4.456513365688314e-05, "loss": 0.0, "step": 1411 }, { "epoch": 0.33, "learning_rate": 4.456128187350744e-05, "loss": 0.0, "step": 1412 }, { "epoch": 0.33, "learning_rate": 4.4557430090131736e-05, "loss": 0.0, "step": 1413 }, { "epoch": 0.33, "learning_rate": 4.455357830675603e-05, "loss": 0.0, "step": 1414 }, { "epoch": 0.33, "learning_rate": 4.4549726523380326e-05, "loss": 0.0, "step": 1415 }, { "epoch": 0.33, "learning_rate": 4.454587474000462e-05, "loss": 0.0, "step": 1416 }, { "epoch": 0.33, "learning_rate": 4.454202295662892e-05, "loss": 0.0, "step": 1417 }, { "epoch": 0.33, "learning_rate": 4.453817117325322e-05, "loss": 0.0, "step": 1418 }, { "epoch": 0.33, "learning_rate": 4.453431938987752e-05, "loss": 0.0, "step": 1419 }, { "epoch": 0.33, "learning_rate": 4.4530467606501814e-05, "loss": 0.0, "step": 1420 }, { "epoch": 0.33, "learning_rate": 4.452661582312611e-05, "loss": 0.0, "step": 1421 }, { "epoch": 0.33, "learning_rate": 4.452276403975041e-05, "loss": 0.0, "step": 1422 }, { "epoch": 0.33, "learning_rate": 4.4518912256374706e-05, "loss": 0.0, "step": 1423 }, { "epoch": 0.33, "learning_rate": 4.4515060472999e-05, "loss": 0.0, "step": 1424 }, { "epoch": 0.33, "learning_rate": 4.4511208689623296e-05, "loss": 0.0, "step": 1425 }, { "epoch": 0.33, "learning_rate": 4.450735690624759e-05, "loss": 0.0, "step": 1426 }, { "epoch": 0.33, "learning_rate": 4.450350512287189e-05, "loss": 0.0, "step": 1427 }, { "epoch": 0.33, "learning_rate": 4.449965333949619e-05, "loss": 0.0, "step": 1428 }, { "epoch": 0.33, "learning_rate": 4.449580155612049e-05, "loss": 0.0, "step": 1429 }, { "epoch": 0.33, "learning_rate": 4.4491949772744784e-05, "loss": 0.0, "step": 1430 }, { "epoch": 0.33, "learning_rate": 4.448809798936908e-05, "loss": 0.0, "step": 1431 }, { "epoch": 0.33, "learning_rate": 4.448424620599338e-05, "loss": 0.0, "step": 1432 }, { "epoch": 0.33, "learning_rate": 4.448039442261767e-05, "loss": 0.0, "step": 1433 }, { "epoch": 0.33, "learning_rate": 4.447654263924197e-05, "loss": 0.0, "step": 1434 }, { "epoch": 0.33, "learning_rate": 4.4472690855866265e-05, "loss": 0.0, "step": 1435 }, { "epoch": 0.33, "learning_rate": 4.446883907249057e-05, "loss": 0.0001, "step": 1436 }, { "epoch": 0.33, "learning_rate": 4.446498728911486e-05, "loss": 0.0, "step": 1437 }, { "epoch": 0.33, "learning_rate": 4.446113550573916e-05, "loss": 0.0, "step": 1438 }, { "epoch": 0.33, "learning_rate": 4.445728372236346e-05, "loss": 0.0001, "step": 1439 }, { "epoch": 0.33, "learning_rate": 4.4453431938987754e-05, "loss": 0.0, "step": 1440 }, { "epoch": 0.33, "learning_rate": 4.4449580155612056e-05, "loss": 0.0003, "step": 1441 }, { "epoch": 0.33, "learning_rate": 4.4445728372236344e-05, "loss": 0.0, "step": 1442 }, { "epoch": 0.33, "learning_rate": 4.4441876588860645e-05, "loss": 0.0278, "step": 1443 }, { "epoch": 0.33, "learning_rate": 4.443802480548494e-05, "loss": 0.0, "step": 1444 }, { "epoch": 0.33, "learning_rate": 4.4434173022109235e-05, "loss": 0.0, "step": 1445 }, { "epoch": 0.33, "learning_rate": 4.443032123873354e-05, "loss": 0.0006, "step": 1446 }, { "epoch": 0.33, "learning_rate": 4.442646945535783e-05, "loss": 0.0, "step": 1447 }, { "epoch": 0.33, "learning_rate": 4.4422617671982134e-05, "loss": 0.0, "step": 1448 }, { "epoch": 0.33, "learning_rate": 4.441876588860643e-05, "loss": 0.0, "step": 1449 }, { "epoch": 0.34, "learning_rate": 4.4414914105230724e-05, "loss": 0.0, "step": 1450 }, { "epoch": 0.34, "learning_rate": 4.4411062321855025e-05, "loss": 0.0, "step": 1451 }, { "epoch": 0.34, "learning_rate": 4.4407210538479314e-05, "loss": 0.0, "step": 1452 }, { "epoch": 0.34, "learning_rate": 4.4403358755103615e-05, "loss": 0.0, "step": 1453 }, { "epoch": 0.34, "learning_rate": 4.439950697172791e-05, "loss": 0.0, "step": 1454 }, { "epoch": 0.34, "learning_rate": 4.439565518835221e-05, "loss": 0.0, "step": 1455 }, { "epoch": 0.34, "learning_rate": 4.439180340497651e-05, "loss": 0.0, "step": 1456 }, { "epoch": 0.34, "learning_rate": 4.43879516216008e-05, "loss": 0.0, "step": 1457 }, { "epoch": 0.34, "learning_rate": 4.4384099838225104e-05, "loss": 0.0, "step": 1458 }, { "epoch": 0.34, "learning_rate": 4.43802480548494e-05, "loss": 0.0, "step": 1459 }, { "epoch": 0.34, "learning_rate": 4.4376396271473694e-05, "loss": 0.0, "step": 1460 }, { "epoch": 0.34, "learning_rate": 4.437254448809799e-05, "loss": 0.0, "step": 1461 }, { "epoch": 0.34, "learning_rate": 4.4368692704722284e-05, "loss": 0.0, "step": 1462 }, { "epoch": 0.34, "learning_rate": 4.4364840921346585e-05, "loss": 0.0, "step": 1463 }, { "epoch": 0.34, "learning_rate": 4.436098913797088e-05, "loss": 0.0, "step": 1464 }, { "epoch": 0.34, "learning_rate": 4.435713735459518e-05, "loss": 0.0, "step": 1465 }, { "epoch": 0.34, "learning_rate": 4.435328557121948e-05, "loss": 0.0, "step": 1466 }, { "epoch": 0.34, "learning_rate": 4.434943378784377e-05, "loss": 0.0, "step": 1467 }, { "epoch": 0.34, "learning_rate": 4.4345582004468074e-05, "loss": 0.0, "step": 1468 }, { "epoch": 0.34, "learning_rate": 4.434173022109237e-05, "loss": 0.0, "step": 1469 }, { "epoch": 0.34, "learning_rate": 4.4337878437716663e-05, "loss": 0.0, "step": 1470 }, { "epoch": 0.34, "learning_rate": 4.433402665434096e-05, "loss": 0.0, "step": 1471 }, { "epoch": 0.34, "learning_rate": 4.433017487096526e-05, "loss": 0.0, "step": 1472 }, { "epoch": 0.34, "learning_rate": 4.4326323087589555e-05, "loss": 0.0, "step": 1473 }, { "epoch": 0.34, "learning_rate": 4.432247130421385e-05, "loss": 0.0, "step": 1474 }, { "epoch": 0.34, "learning_rate": 4.431861952083815e-05, "loss": 0.0, "step": 1475 }, { "epoch": 0.34, "learning_rate": 4.431476773746245e-05, "loss": 0.0, "step": 1476 }, { "epoch": 0.34, "learning_rate": 4.431091595408675e-05, "loss": 0.0, "step": 1477 }, { "epoch": 0.34, "learning_rate": 4.4307064170711043e-05, "loss": 0.0, "step": 1478 }, { "epoch": 0.34, "learning_rate": 4.430321238733534e-05, "loss": 0.0, "step": 1479 }, { "epoch": 0.34, "learning_rate": 4.429936060395963e-05, "loss": 0.0, "step": 1480 }, { "epoch": 0.34, "learning_rate": 4.429550882058393e-05, "loss": 0.0, "step": 1481 }, { "epoch": 0.34, "learning_rate": 4.429165703720823e-05, "loss": 0.0, "step": 1482 }, { "epoch": 0.34, "learning_rate": 4.4287805253832525e-05, "loss": 0.0305, "step": 1483 }, { "epoch": 0.34, "learning_rate": 4.428395347045683e-05, "loss": 0.0, "step": 1484 }, { "epoch": 0.34, "learning_rate": 4.428010168708112e-05, "loss": 0.0, "step": 1485 }, { "epoch": 0.34, "learning_rate": 4.427624990370542e-05, "loss": 0.0001, "step": 1486 }, { "epoch": 0.34, "learning_rate": 4.427239812032972e-05, "loss": 0.0001, "step": 1487 }, { "epoch": 0.34, "learning_rate": 4.426854633695401e-05, "loss": 0.0, "step": 1488 }, { "epoch": 0.34, "learning_rate": 4.426469455357831e-05, "loss": 0.0, "step": 1489 }, { "epoch": 0.34, "learning_rate": 4.42608427702026e-05, "loss": 0.0005, "step": 1490 }, { "epoch": 0.34, "learning_rate": 4.4256990986826905e-05, "loss": 0.0, "step": 1491 }, { "epoch": 0.34, "learning_rate": 4.42531392034512e-05, "loss": 0.0, "step": 1492 }, { "epoch": 0.35, "learning_rate": 4.4249287420075495e-05, "loss": 0.0, "step": 1493 }, { "epoch": 0.35, "learning_rate": 4.42454356366998e-05, "loss": 0.0, "step": 1494 }, { "epoch": 0.35, "learning_rate": 4.424158385332409e-05, "loss": 0.0005, "step": 1495 }, { "epoch": 0.35, "learning_rate": 4.423773206994839e-05, "loss": 0.0, "step": 1496 }, { "epoch": 0.35, "learning_rate": 4.423388028657269e-05, "loss": 0.0, "step": 1497 }, { "epoch": 0.35, "learning_rate": 4.4230028503196976e-05, "loss": 0.0, "step": 1498 }, { "epoch": 0.35, "learning_rate": 4.422617671982128e-05, "loss": 0.0, "step": 1499 }, { "epoch": 0.35, "learning_rate": 4.422232493644557e-05, "loss": 0.0, "step": 1500 }, { "epoch": 0.35, "learning_rate": 4.4218473153069875e-05, "loss": 0.0, "step": 1501 }, { "epoch": 0.35, "learning_rate": 4.421462136969417e-05, "loss": 0.0, "step": 1502 }, { "epoch": 0.35, "learning_rate": 4.4210769586318465e-05, "loss": 0.001, "step": 1503 }, { "epoch": 0.35, "learning_rate": 4.4206917802942767e-05, "loss": 0.0, "step": 1504 }, { "epoch": 0.35, "learning_rate": 4.420306601956706e-05, "loss": 0.0, "step": 1505 }, { "epoch": 0.35, "learning_rate": 4.419921423619136e-05, "loss": 0.0, "step": 1506 }, { "epoch": 0.35, "learning_rate": 4.419536245281566e-05, "loss": 0.0, "step": 1507 }, { "epoch": 0.35, "learning_rate": 4.419151066943995e-05, "loss": 0.0022, "step": 1508 }, { "epoch": 0.35, "learning_rate": 4.418765888606425e-05, "loss": 0.0004, "step": 1509 }, { "epoch": 0.35, "learning_rate": 4.418380710268854e-05, "loss": 0.0, "step": 1510 }, { "epoch": 0.35, "learning_rate": 4.4179955319312845e-05, "loss": 0.0, "step": 1511 }, { "epoch": 0.35, "learning_rate": 4.417610353593714e-05, "loss": 0.0, "step": 1512 }, { "epoch": 0.35, "learning_rate": 4.417225175256144e-05, "loss": 0.0, "step": 1513 }, { "epoch": 0.35, "learning_rate": 4.4168399969185736e-05, "loss": 0.0, "step": 1514 }, { "epoch": 0.35, "learning_rate": 4.416454818581003e-05, "loss": 0.0, "step": 1515 }, { "epoch": 0.35, "learning_rate": 4.416069640243433e-05, "loss": 0.0, "step": 1516 }, { "epoch": 0.35, "learning_rate": 4.415684461905862e-05, "loss": 0.0, "step": 1517 }, { "epoch": 0.35, "learning_rate": 4.415299283568292e-05, "loss": 0.0, "step": 1518 }, { "epoch": 0.35, "learning_rate": 4.414914105230722e-05, "loss": 0.0, "step": 1519 }, { "epoch": 0.35, "learning_rate": 4.414528926893152e-05, "loss": 0.0, "step": 1520 }, { "epoch": 0.35, "learning_rate": 4.4141437485555815e-05, "loss": 0.0, "step": 1521 }, { "epoch": 0.35, "learning_rate": 4.413758570218011e-05, "loss": 0.0001, "step": 1522 }, { "epoch": 0.35, "learning_rate": 4.413373391880441e-05, "loss": 0.0, "step": 1523 }, { "epoch": 0.35, "learning_rate": 4.4129882135428706e-05, "loss": 0.0, "step": 1524 }, { "epoch": 0.35, "learning_rate": 4.412603035205301e-05, "loss": 0.0003, "step": 1525 }, { "epoch": 0.35, "learning_rate": 4.4122178568677296e-05, "loss": 0.0, "step": 1526 }, { "epoch": 0.35, "learning_rate": 4.41183267853016e-05, "loss": 0.0, "step": 1527 }, { "epoch": 0.35, "learning_rate": 4.411447500192589e-05, "loss": 0.0, "step": 1528 }, { "epoch": 0.35, "learning_rate": 4.411062321855019e-05, "loss": 0.0, "step": 1529 }, { "epoch": 0.35, "learning_rate": 4.410677143517449e-05, "loss": 0.0002, "step": 1530 }, { "epoch": 0.35, "learning_rate": 4.4102919651798785e-05, "loss": 0.0, "step": 1531 }, { "epoch": 0.35, "learning_rate": 4.4099067868423086e-05, "loss": 0.0, "step": 1532 }, { "epoch": 0.35, "learning_rate": 4.409521608504738e-05, "loss": 0.0, "step": 1533 }, { "epoch": 0.35, "learning_rate": 4.4091364301671676e-05, "loss": 0.0, "step": 1534 }, { "epoch": 0.35, "learning_rate": 4.408751251829598e-05, "loss": 0.0, "step": 1535 }, { "epoch": 0.35, "learning_rate": 4.4083660734920266e-05, "loss": 0.0, "step": 1536 }, { "epoch": 0.36, "learning_rate": 4.407980895154457e-05, "loss": 0.0, "step": 1537 }, { "epoch": 0.36, "learning_rate": 4.407595716816886e-05, "loss": 0.0, "step": 1538 }, { "epoch": 0.36, "learning_rate": 4.407210538479316e-05, "loss": 0.0, "step": 1539 }, { "epoch": 0.36, "learning_rate": 4.406825360141746e-05, "loss": 0.0, "step": 1540 }, { "epoch": 0.36, "learning_rate": 4.4064401818041754e-05, "loss": 0.0, "step": 1541 }, { "epoch": 0.36, "learning_rate": 4.4060550034666056e-05, "loss": 0.0001, "step": 1542 }, { "epoch": 0.36, "learning_rate": 4.405669825129035e-05, "loss": 0.0001, "step": 1543 }, { "epoch": 0.36, "learning_rate": 4.4052846467914646e-05, "loss": 0.0, "step": 1544 }, { "epoch": 0.36, "learning_rate": 4.404899468453894e-05, "loss": 0.0001, "step": 1545 }, { "epoch": 0.36, "learning_rate": 4.4045142901163236e-05, "loss": 0.0, "step": 1546 }, { "epoch": 0.36, "learning_rate": 4.404129111778754e-05, "loss": 0.0, "step": 1547 }, { "epoch": 0.36, "learning_rate": 4.403743933441183e-05, "loss": 0.0, "step": 1548 }, { "epoch": 0.36, "learning_rate": 4.4033587551036134e-05, "loss": 0.0002, "step": 1549 }, { "epoch": 0.36, "learning_rate": 4.402973576766043e-05, "loss": 0.0, "step": 1550 }, { "epoch": 0.36, "learning_rate": 4.4025883984284724e-05, "loss": 0.0009, "step": 1551 }, { "epoch": 0.36, "learning_rate": 4.4022032200909026e-05, "loss": 0.0, "step": 1552 }, { "epoch": 0.36, "learning_rate": 4.401818041753332e-05, "loss": 0.0, "step": 1553 }, { "epoch": 0.36, "learning_rate": 4.4014328634157616e-05, "loss": 0.0, "step": 1554 }, { "epoch": 0.36, "learning_rate": 4.401047685078191e-05, "loss": 0.0001, "step": 1555 }, { "epoch": 0.36, "learning_rate": 4.400662506740621e-05, "loss": 0.0, "step": 1556 }, { "epoch": 0.36, "learning_rate": 4.400277328403051e-05, "loss": 0.0, "step": 1557 }, { "epoch": 0.36, "learning_rate": 4.39989215006548e-05, "loss": 0.0, "step": 1558 }, { "epoch": 0.36, "learning_rate": 4.3995069717279104e-05, "loss": 0.0, "step": 1559 }, { "epoch": 0.36, "learning_rate": 4.39912179339034e-05, "loss": 0.0, "step": 1560 }, { "epoch": 0.36, "learning_rate": 4.39873661505277e-05, "loss": 0.0, "step": 1561 }, { "epoch": 0.36, "learning_rate": 4.3983514367151996e-05, "loss": 0.0001, "step": 1562 }, { "epoch": 0.36, "learning_rate": 4.397966258377629e-05, "loss": 0.0, "step": 1563 }, { "epoch": 0.36, "learning_rate": 4.3975810800400586e-05, "loss": 0.0074, "step": 1564 }, { "epoch": 0.36, "learning_rate": 4.397195901702488e-05, "loss": 0.0, "step": 1565 }, { "epoch": 0.36, "learning_rate": 4.396810723364918e-05, "loss": 0.0, "step": 1566 }, { "epoch": 0.36, "learning_rate": 4.396425545027348e-05, "loss": 0.0, "step": 1567 }, { "epoch": 0.36, "learning_rate": 4.396040366689778e-05, "loss": 0.0, "step": 1568 }, { "epoch": 0.36, "learning_rate": 4.3956551883522074e-05, "loss": 0.0, "step": 1569 }, { "epoch": 0.36, "learning_rate": 4.395270010014637e-05, "loss": 0.0, "step": 1570 }, { "epoch": 0.36, "learning_rate": 4.394884831677067e-05, "loss": 0.0, "step": 1571 }, { "epoch": 0.36, "learning_rate": 4.3944996533394966e-05, "loss": 0.0, "step": 1572 }, { "epoch": 0.36, "learning_rate": 4.394114475001926e-05, "loss": 0.0, "step": 1573 }, { "epoch": 0.36, "learning_rate": 4.3937292966643556e-05, "loss": 0.0, "step": 1574 }, { "epoch": 0.36, "learning_rate": 4.393344118326785e-05, "loss": 0.0, "step": 1575 }, { "epoch": 0.36, "learning_rate": 4.392958939989215e-05, "loss": 0.0, "step": 1576 }, { "epoch": 0.36, "learning_rate": 4.392573761651645e-05, "loss": 0.0, "step": 1577 }, { "epoch": 0.36, "learning_rate": 4.392188583314075e-05, "loss": 0.0, "step": 1578 }, { "epoch": 0.36, "learning_rate": 4.3918034049765044e-05, "loss": 0.0, "step": 1579 }, { "epoch": 0.37, "learning_rate": 4.391418226638934e-05, "loss": 0.0, "step": 1580 }, { "epoch": 0.37, "learning_rate": 4.391033048301364e-05, "loss": 0.0, "step": 1581 }, { "epoch": 0.37, "learning_rate": 4.390647869963793e-05, "loss": 0.0, "step": 1582 }, { "epoch": 0.37, "learning_rate": 4.390262691626223e-05, "loss": 0.0, "step": 1583 }, { "epoch": 0.37, "learning_rate": 4.3898775132886526e-05, "loss": 0.0001, "step": 1584 }, { "epoch": 0.37, "learning_rate": 4.389492334951083e-05, "loss": 0.0, "step": 1585 }, { "epoch": 0.37, "learning_rate": 4.389107156613512e-05, "loss": 0.0, "step": 1586 }, { "epoch": 0.37, "learning_rate": 4.388721978275942e-05, "loss": 0.0, "step": 1587 }, { "epoch": 0.37, "learning_rate": 4.388336799938372e-05, "loss": 0.0, "step": 1588 }, { "epoch": 0.37, "learning_rate": 4.3879516216008014e-05, "loss": 0.0002, "step": 1589 }, { "epoch": 0.37, "learning_rate": 4.3875664432632316e-05, "loss": 0.0002, "step": 1590 }, { "epoch": 0.37, "learning_rate": 4.387181264925661e-05, "loss": 0.0, "step": 1591 }, { "epoch": 0.37, "learning_rate": 4.3867960865880906e-05, "loss": 0.0, "step": 1592 }, { "epoch": 0.37, "learning_rate": 4.38641090825052e-05, "loss": 0.0001, "step": 1593 }, { "epoch": 0.37, "learning_rate": 4.3860257299129495e-05, "loss": 0.0, "step": 1594 }, { "epoch": 0.37, "learning_rate": 4.38564055157538e-05, "loss": 0.0, "step": 1595 }, { "epoch": 0.37, "learning_rate": 4.385255373237809e-05, "loss": 0.0, "step": 1596 }, { "epoch": 0.37, "learning_rate": 4.3848701949002394e-05, "loss": 0.0, "step": 1597 }, { "epoch": 0.37, "learning_rate": 4.384485016562669e-05, "loss": 0.0001, "step": 1598 }, { "epoch": 0.37, "learning_rate": 4.3840998382250984e-05, "loss": 0.0, "step": 1599 }, { "epoch": 0.37, "learning_rate": 4.3837146598875286e-05, "loss": 0.0, "step": 1600 }, { "epoch": 0.37, "learning_rate": 4.3833294815499574e-05, "loss": 0.0, "step": 1601 }, { "epoch": 0.37, "learning_rate": 4.3829443032123875e-05, "loss": 0.0, "step": 1602 }, { "epoch": 0.37, "learning_rate": 4.382559124874817e-05, "loss": 0.0, "step": 1603 }, { "epoch": 0.37, "learning_rate": 4.382173946537247e-05, "loss": 0.0, "step": 1604 }, { "epoch": 0.37, "learning_rate": 4.381788768199677e-05, "loss": 0.0, "step": 1605 }, { "epoch": 0.37, "learning_rate": 4.381403589862106e-05, "loss": 0.0, "step": 1606 }, { "epoch": 0.37, "learning_rate": 4.3810184115245364e-05, "loss": 0.0, "step": 1607 }, { "epoch": 0.37, "learning_rate": 4.380633233186966e-05, "loss": 0.0, "step": 1608 }, { "epoch": 0.37, "learning_rate": 4.380248054849396e-05, "loss": 0.0, "step": 1609 }, { "epoch": 0.37, "learning_rate": 4.379862876511825e-05, "loss": 0.0, "step": 1610 }, { "epoch": 0.37, "learning_rate": 4.3794776981742544e-05, "loss": 0.0, "step": 1611 }, { "epoch": 0.37, "learning_rate": 4.3790925198366845e-05, "loss": 0.0001, "step": 1612 }, { "epoch": 0.37, "learning_rate": 4.378707341499114e-05, "loss": 0.0, "step": 1613 }, { "epoch": 0.37, "learning_rate": 4.378322163161544e-05, "loss": 0.0001, "step": 1614 }, { "epoch": 0.37, "learning_rate": 4.377936984823974e-05, "loss": 0.0, "step": 1615 }, { "epoch": 0.37, "learning_rate": 4.377551806486403e-05, "loss": 0.0, "step": 1616 }, { "epoch": 0.37, "learning_rate": 4.3771666281488334e-05, "loss": 0.0, "step": 1617 }, { "epoch": 0.37, "learning_rate": 4.376781449811263e-05, "loss": 0.0001, "step": 1618 }, { "epoch": 0.37, "learning_rate": 4.376396271473693e-05, "loss": 0.0, "step": 1619 }, { "epoch": 0.37, "learning_rate": 4.376011093136122e-05, "loss": 0.0002, "step": 1620 }, { "epoch": 0.37, "learning_rate": 4.375625914798552e-05, "loss": 0.0, "step": 1621 }, { "epoch": 0.37, "learning_rate": 4.3752407364609815e-05, "loss": 0.0, "step": 1622 }, { "epoch": 0.38, "learning_rate": 4.374855558123411e-05, "loss": 0.0, "step": 1623 }, { "epoch": 0.38, "learning_rate": 4.374470379785841e-05, "loss": 0.0, "step": 1624 }, { "epoch": 0.38, "learning_rate": 4.374085201448271e-05, "loss": 0.0, "step": 1625 }, { "epoch": 0.38, "learning_rate": 4.373700023110701e-05, "loss": 0.0, "step": 1626 }, { "epoch": 0.38, "learning_rate": 4.3733148447731304e-05, "loss": 0.0, "step": 1627 }, { "epoch": 0.38, "learning_rate": 4.37292966643556e-05, "loss": 0.0, "step": 1628 }, { "epoch": 0.38, "learning_rate": 4.3725444880979893e-05, "loss": 0.0, "step": 1629 }, { "epoch": 0.38, "learning_rate": 4.372159309760419e-05, "loss": 0.0, "step": 1630 }, { "epoch": 0.38, "learning_rate": 4.371774131422849e-05, "loss": 0.0, "step": 1631 }, { "epoch": 0.38, "learning_rate": 4.3713889530852785e-05, "loss": 0.0, "step": 1632 }, { "epoch": 0.38, "learning_rate": 4.371003774747709e-05, "loss": 0.0, "step": 1633 }, { "epoch": 0.38, "learning_rate": 4.370618596410138e-05, "loss": 0.0, "step": 1634 }, { "epoch": 0.38, "learning_rate": 4.370233418072568e-05, "loss": 0.0, "step": 1635 }, { "epoch": 0.38, "learning_rate": 4.369848239734998e-05, "loss": 0.0, "step": 1636 }, { "epoch": 0.38, "learning_rate": 4.3694630613974273e-05, "loss": 0.0, "step": 1637 }, { "epoch": 0.38, "learning_rate": 4.369077883059857e-05, "loss": 0.0001, "step": 1638 }, { "epoch": 0.38, "learning_rate": 4.368692704722286e-05, "loss": 0.0, "step": 1639 }, { "epoch": 0.38, "learning_rate": 4.3683075263847165e-05, "loss": 0.0, "step": 1640 }, { "epoch": 0.38, "learning_rate": 4.367922348047146e-05, "loss": 0.0, "step": 1641 }, { "epoch": 0.38, "learning_rate": 4.3675371697095755e-05, "loss": 0.0, "step": 1642 }, { "epoch": 0.38, "learning_rate": 4.367151991372006e-05, "loss": 0.0, "step": 1643 }, { "epoch": 0.38, "learning_rate": 4.366766813034435e-05, "loss": 0.0, "step": 1644 }, { "epoch": 0.38, "learning_rate": 4.3663816346968653e-05, "loss": 0.0, "step": 1645 }, { "epoch": 0.38, "learning_rate": 4.365996456359295e-05, "loss": 0.0, "step": 1646 }, { "epoch": 0.38, "learning_rate": 4.365611278021724e-05, "loss": 0.0, "step": 1647 }, { "epoch": 0.38, "learning_rate": 4.365226099684154e-05, "loss": 0.0, "step": 1648 }, { "epoch": 0.38, "learning_rate": 4.364840921346583e-05, "loss": 0.0, "step": 1649 }, { "epoch": 0.38, "learning_rate": 4.3644557430090135e-05, "loss": 0.0, "step": 1650 }, { "epoch": 0.38, "learning_rate": 4.364070564671443e-05, "loss": 0.0, "step": 1651 }, { "epoch": 0.38, "learning_rate": 4.3636853863338725e-05, "loss": 0.0, "step": 1652 }, { "epoch": 0.38, "learning_rate": 4.3633002079963027e-05, "loss": 0.0, "step": 1653 }, { "epoch": 0.38, "learning_rate": 4.362915029658732e-05, "loss": 0.0, "step": 1654 }, { "epoch": 0.38, "learning_rate": 4.362529851321162e-05, "loss": 0.0, "step": 1655 }, { "epoch": 0.38, "learning_rate": 4.362144672983592e-05, "loss": 0.0001, "step": 1656 }, { "epoch": 0.38, "learning_rate": 4.361759494646021e-05, "loss": 0.0, "step": 1657 }, { "epoch": 0.38, "learning_rate": 4.361374316308451e-05, "loss": 0.0, "step": 1658 }, { "epoch": 0.38, "learning_rate": 4.36098913797088e-05, "loss": 0.0, "step": 1659 }, { "epoch": 0.38, "learning_rate": 4.3606039596333105e-05, "loss": 0.0002, "step": 1660 }, { "epoch": 0.38, "learning_rate": 4.36021878129574e-05, "loss": 0.0, "step": 1661 }, { "epoch": 0.38, "learning_rate": 4.35983360295817e-05, "loss": 0.0, "step": 1662 }, { "epoch": 0.38, "learning_rate": 4.3594484246205996e-05, "loss": 0.0, "step": 1663 }, { "epoch": 0.38, "learning_rate": 4.359063246283029e-05, "loss": 0.0, "step": 1664 }, { "epoch": 0.38, "learning_rate": 4.358678067945459e-05, "loss": 0.0, "step": 1665 }, { "epoch": 0.39, "learning_rate": 4.358292889607888e-05, "loss": 0.0002, "step": 1666 }, { "epoch": 0.39, "learning_rate": 4.357907711270318e-05, "loss": 0.0, "step": 1667 }, { "epoch": 0.39, "learning_rate": 4.357522532932748e-05, "loss": 0.0, "step": 1668 }, { "epoch": 0.39, "learning_rate": 4.357137354595178e-05, "loss": 0.0004, "step": 1669 }, { "epoch": 0.39, "learning_rate": 4.3567521762576075e-05, "loss": 0.0, "step": 1670 }, { "epoch": 0.39, "learning_rate": 4.356366997920037e-05, "loss": 0.0, "step": 1671 }, { "epoch": 0.39, "learning_rate": 4.355981819582467e-05, "loss": 0.0, "step": 1672 }, { "epoch": 0.39, "learning_rate": 4.3555966412448966e-05, "loss": 0.0, "step": 1673 }, { "epoch": 0.39, "learning_rate": 4.355211462907327e-05, "loss": 0.0, "step": 1674 }, { "epoch": 0.39, "learning_rate": 4.354826284569756e-05, "loss": 0.0, "step": 1675 }, { "epoch": 0.39, "learning_rate": 4.354441106232186e-05, "loss": 0.0, "step": 1676 }, { "epoch": 0.39, "learning_rate": 4.354055927894615e-05, "loss": 0.0, "step": 1677 }, { "epoch": 0.39, "learning_rate": 4.353670749557045e-05, "loss": 0.0001, "step": 1678 }, { "epoch": 0.39, "learning_rate": 4.353285571219475e-05, "loss": 0.0, "step": 1679 }, { "epoch": 0.39, "learning_rate": 4.3529003928819045e-05, "loss": 0.0, "step": 1680 }, { "epoch": 0.39, "learning_rate": 4.3525152145443346e-05, "loss": 0.0, "step": 1681 }, { "epoch": 0.39, "learning_rate": 4.352130036206764e-05, "loss": 0.0, "step": 1682 }, { "epoch": 0.39, "learning_rate": 4.3517448578691936e-05, "loss": 0.0, "step": 1683 }, { "epoch": 0.39, "learning_rate": 4.351359679531624e-05, "loss": 0.0, "step": 1684 }, { "epoch": 0.39, "learning_rate": 4.3509745011940526e-05, "loss": 0.0, "step": 1685 }, { "epoch": 0.39, "learning_rate": 4.350589322856483e-05, "loss": 0.0, "step": 1686 }, { "epoch": 0.39, "learning_rate": 4.350204144518912e-05, "loss": 0.0, "step": 1687 }, { "epoch": 0.39, "learning_rate": 4.349818966181342e-05, "loss": 0.0, "step": 1688 }, { "epoch": 0.39, "learning_rate": 4.349433787843772e-05, "loss": 0.0, "step": 1689 }, { "epoch": 0.39, "learning_rate": 4.3490486095062014e-05, "loss": 0.0001, "step": 1690 }, { "epoch": 0.39, "learning_rate": 4.3486634311686316e-05, "loss": 0.0, "step": 1691 }, { "epoch": 0.39, "learning_rate": 4.348278252831061e-05, "loss": 0.0, "step": 1692 }, { "epoch": 0.39, "learning_rate": 4.3478930744934906e-05, "loss": 0.0, "step": 1693 }, { "epoch": 0.39, "learning_rate": 4.34750789615592e-05, "loss": 0.0, "step": 1694 }, { "epoch": 0.39, "learning_rate": 4.3471227178183496e-05, "loss": 0.0, "step": 1695 }, { "epoch": 0.39, "learning_rate": 4.34673753948078e-05, "loss": 0.0, "step": 1696 }, { "epoch": 0.39, "learning_rate": 4.346352361143209e-05, "loss": 0.0, "step": 1697 }, { "epoch": 0.39, "learning_rate": 4.3459671828056394e-05, "loss": 0.0, "step": 1698 }, { "epoch": 0.39, "learning_rate": 4.345582004468069e-05, "loss": 0.0, "step": 1699 }, { "epoch": 0.39, "learning_rate": 4.3451968261304984e-05, "loss": 0.0, "step": 1700 }, { "epoch": 0.39, "learning_rate": 4.3448116477929286e-05, "loss": 0.0001, "step": 1701 }, { "epoch": 0.39, "learning_rate": 4.344426469455358e-05, "loss": 0.0, "step": 1702 }, { "epoch": 0.39, "learning_rate": 4.344041291117788e-05, "loss": 0.0, "step": 1703 }, { "epoch": 0.39, "learning_rate": 4.343656112780217e-05, "loss": 0.0, "step": 1704 }, { "epoch": 0.39, "learning_rate": 4.343270934442647e-05, "loss": 0.0, "step": 1705 }, { "epoch": 0.39, "learning_rate": 4.342885756105077e-05, "loss": 0.0001, "step": 1706 }, { "epoch": 0.39, "learning_rate": 4.342500577767506e-05, "loss": 0.0, "step": 1707 }, { "epoch": 0.39, "learning_rate": 4.3421153994299364e-05, "loss": 0.0001, "step": 1708 }, { "epoch": 0.39, "learning_rate": 4.341730221092366e-05, "loss": 0.0, "step": 1709 }, { "epoch": 0.4, "learning_rate": 4.341345042754796e-05, "loss": 0.0001, "step": 1710 }, { "epoch": 0.4, "learning_rate": 4.3409598644172256e-05, "loss": 0.0002, "step": 1711 }, { "epoch": 0.4, "learning_rate": 4.340574686079655e-05, "loss": 0.0, "step": 1712 }, { "epoch": 0.4, "learning_rate": 4.3401895077420846e-05, "loss": 0.0, "step": 1713 }, { "epoch": 0.4, "learning_rate": 4.339804329404514e-05, "loss": 0.0, "step": 1714 }, { "epoch": 0.4, "learning_rate": 4.339419151066944e-05, "loss": 0.0, "step": 1715 }, { "epoch": 0.4, "learning_rate": 4.339033972729374e-05, "loss": 0.0001, "step": 1716 }, { "epoch": 0.4, "learning_rate": 4.338648794391804e-05, "loss": 0.0, "step": 1717 }, { "epoch": 0.4, "learning_rate": 4.3382636160542334e-05, "loss": 0.0, "step": 1718 }, { "epoch": 0.4, "learning_rate": 4.337878437716663e-05, "loss": 0.0, "step": 1719 }, { "epoch": 0.4, "learning_rate": 4.337493259379093e-05, "loss": 0.0, "step": 1720 }, { "epoch": 0.4, "learning_rate": 4.3371080810415226e-05, "loss": 0.0, "step": 1721 }, { "epoch": 0.4, "learning_rate": 4.336722902703952e-05, "loss": 0.0, "step": 1722 }, { "epoch": 0.4, "learning_rate": 4.3363377243663816e-05, "loss": 0.0, "step": 1723 }, { "epoch": 0.4, "learning_rate": 4.335952546028811e-05, "loss": 0.0027, "step": 1724 }, { "epoch": 0.4, "learning_rate": 4.335567367691241e-05, "loss": 0.0, "step": 1725 }, { "epoch": 0.4, "learning_rate": 4.335182189353671e-05, "loss": 0.0, "step": 1726 }, { "epoch": 0.4, "learning_rate": 4.334797011016101e-05, "loss": 0.0, "step": 1727 }, { "epoch": 0.4, "learning_rate": 4.3344118326785304e-05, "loss": 0.0, "step": 1728 }, { "epoch": 0.4, "learning_rate": 4.33402665434096e-05, "loss": 0.0, "step": 1729 }, { "epoch": 0.4, "learning_rate": 4.33364147600339e-05, "loss": 0.0, "step": 1730 }, { "epoch": 0.4, "learning_rate": 4.3332562976658196e-05, "loss": 0.0, "step": 1731 }, { "epoch": 0.4, "learning_rate": 4.332871119328249e-05, "loss": 0.0, "step": 1732 }, { "epoch": 0.4, "learning_rate": 4.3324859409906786e-05, "loss": 0.0, "step": 1733 }, { "epoch": 0.4, "learning_rate": 4.332100762653109e-05, "loss": 0.0, "step": 1734 }, { "epoch": 0.4, "learning_rate": 4.331715584315538e-05, "loss": 0.0, "step": 1735 }, { "epoch": 0.4, "learning_rate": 4.331330405977968e-05, "loss": 0.0, "step": 1736 }, { "epoch": 0.4, "learning_rate": 4.330945227640398e-05, "loss": 0.0, "step": 1737 }, { "epoch": 0.4, "learning_rate": 4.3305600493028274e-05, "loss": 0.0, "step": 1738 }, { "epoch": 0.4, "learning_rate": 4.3301748709652576e-05, "loss": 0.0, "step": 1739 }, { "epoch": 0.4, "learning_rate": 4.329789692627687e-05, "loss": 0.0, "step": 1740 }, { "epoch": 0.4, "learning_rate": 4.3294045142901166e-05, "loss": 0.0, "step": 1741 }, { "epoch": 0.4, "learning_rate": 4.329019335952546e-05, "loss": 0.0, "step": 1742 }, { "epoch": 0.4, "learning_rate": 4.3286341576149756e-05, "loss": 0.0, "step": 1743 }, { "epoch": 0.4, "learning_rate": 4.328248979277406e-05, "loss": 0.0, "step": 1744 }, { "epoch": 0.4, "learning_rate": 4.327863800939835e-05, "loss": 0.0, "step": 1745 }, { "epoch": 0.4, "learning_rate": 4.3274786226022654e-05, "loss": 0.0, "step": 1746 }, { "epoch": 0.4, "learning_rate": 4.327093444264695e-05, "loss": 0.0, "step": 1747 }, { "epoch": 0.4, "learning_rate": 4.3267082659271244e-05, "loss": 0.0012, "step": 1748 }, { "epoch": 0.4, "learning_rate": 4.3263230875895546e-05, "loss": 0.0, "step": 1749 }, { "epoch": 0.4, "learning_rate": 4.3259379092519834e-05, "loss": 0.0003, "step": 1750 }, { "epoch": 0.4, "learning_rate": 4.3255527309144136e-05, "loss": 0.0, "step": 1751 }, { "epoch": 0.4, "learning_rate": 4.325167552576843e-05, "loss": 0.0013, "step": 1752 }, { "epoch": 0.41, "learning_rate": 4.324782374239273e-05, "loss": 0.0, "step": 1753 }, { "epoch": 0.41, "learning_rate": 4.324397195901703e-05, "loss": 0.0002, "step": 1754 }, { "epoch": 0.41, "learning_rate": 4.324012017564132e-05, "loss": 0.0, "step": 1755 }, { "epoch": 0.41, "learning_rate": 4.3236268392265624e-05, "loss": 0.0, "step": 1756 }, { "epoch": 0.41, "learning_rate": 4.323241660888992e-05, "loss": 0.0024, "step": 1757 }, { "epoch": 0.41, "learning_rate": 4.322856482551422e-05, "loss": 0.0001, "step": 1758 }, { "epoch": 0.41, "learning_rate": 4.3224713042138515e-05, "loss": 0.0, "step": 1759 }, { "epoch": 0.41, "learning_rate": 4.3220861258762804e-05, "loss": 0.0, "step": 1760 }, { "epoch": 0.41, "learning_rate": 4.3217009475387105e-05, "loss": 0.0, "step": 1761 }, { "epoch": 0.41, "learning_rate": 4.32131576920114e-05, "loss": 0.0, "step": 1762 }, { "epoch": 0.41, "learning_rate": 4.32093059086357e-05, "loss": 0.0001, "step": 1763 }, { "epoch": 0.41, "learning_rate": 4.320545412526e-05, "loss": 0.0, "step": 1764 }, { "epoch": 0.41, "learning_rate": 4.320160234188429e-05, "loss": 0.0, "step": 1765 }, { "epoch": 0.41, "learning_rate": 4.3197750558508594e-05, "loss": 0.0, "step": 1766 }, { "epoch": 0.41, "learning_rate": 4.319389877513289e-05, "loss": 0.0, "step": 1767 }, { "epoch": 0.41, "learning_rate": 4.319004699175719e-05, "loss": 0.0, "step": 1768 }, { "epoch": 0.41, "learning_rate": 4.318619520838148e-05, "loss": 0.0, "step": 1769 }, { "epoch": 0.41, "learning_rate": 4.318234342500578e-05, "loss": 0.0, "step": 1770 }, { "epoch": 0.41, "learning_rate": 4.3178491641630075e-05, "loss": 0.0, "step": 1771 }, { "epoch": 0.41, "learning_rate": 4.317463985825437e-05, "loss": 0.0, "step": 1772 }, { "epoch": 0.41, "learning_rate": 4.317078807487867e-05, "loss": 0.0, "step": 1773 }, { "epoch": 0.41, "learning_rate": 4.316693629150297e-05, "loss": 0.0, "step": 1774 }, { "epoch": 0.41, "learning_rate": 4.316308450812727e-05, "loss": 0.0, "step": 1775 }, { "epoch": 0.41, "learning_rate": 4.3159232724751564e-05, "loss": 0.0011, "step": 1776 }, { "epoch": 0.41, "learning_rate": 4.315538094137586e-05, "loss": 0.0003, "step": 1777 }, { "epoch": 0.41, "learning_rate": 4.3151529158000154e-05, "loss": 0.0001, "step": 1778 }, { "epoch": 0.41, "learning_rate": 4.314767737462445e-05, "loss": 0.0, "step": 1779 }, { "epoch": 0.41, "learning_rate": 4.314382559124875e-05, "loss": 0.0, "step": 1780 }, { "epoch": 0.41, "learning_rate": 4.3139973807873045e-05, "loss": 0.0, "step": 1781 }, { "epoch": 0.41, "learning_rate": 4.313612202449735e-05, "loss": 0.0, "step": 1782 }, { "epoch": 0.41, "learning_rate": 4.313227024112164e-05, "loss": 0.0015, "step": 1783 }, { "epoch": 0.41, "learning_rate": 4.312841845774594e-05, "loss": 0.0, "step": 1784 }, { "epoch": 0.41, "learning_rate": 4.312456667437024e-05, "loss": 0.0, "step": 1785 }, { "epoch": 0.41, "learning_rate": 4.3120714890994534e-05, "loss": 0.0, "step": 1786 }, { "epoch": 0.41, "learning_rate": 4.3116863107618835e-05, "loss": 0.0, "step": 1787 }, { "epoch": 0.41, "learning_rate": 4.3113011324243123e-05, "loss": 0.0, "step": 1788 }, { "epoch": 0.41, "learning_rate": 4.3109159540867425e-05, "loss": 0.0, "step": 1789 }, { "epoch": 0.41, "learning_rate": 4.310530775749172e-05, "loss": 0.0, "step": 1790 }, { "epoch": 0.41, "learning_rate": 4.3101455974116015e-05, "loss": 0.0, "step": 1791 }, { "epoch": 0.41, "learning_rate": 4.309760419074032e-05, "loss": 0.0, "step": 1792 }, { "epoch": 0.41, "learning_rate": 4.309375240736461e-05, "loss": 0.0002, "step": 1793 }, { "epoch": 0.41, "learning_rate": 4.3089900623988913e-05, "loss": 0.0, "step": 1794 }, { "epoch": 0.41, "learning_rate": 4.308604884061321e-05, "loss": 0.0, "step": 1795 }, { "epoch": 0.42, "learning_rate": 4.3082197057237503e-05, "loss": 0.0, "step": 1796 }, { "epoch": 0.42, "learning_rate": 4.30783452738618e-05, "loss": 0.0, "step": 1797 }, { "epoch": 0.42, "learning_rate": 4.307449349048609e-05, "loss": 0.0, "step": 1798 }, { "epoch": 0.42, "learning_rate": 4.3070641707110395e-05, "loss": 0.0, "step": 1799 }, { "epoch": 0.42, "learning_rate": 4.306678992373469e-05, "loss": 0.0, "step": 1800 }, { "epoch": 0.42, "learning_rate": 4.3062938140358985e-05, "loss": 0.0, "step": 1801 }, { "epoch": 0.42, "learning_rate": 4.305908635698329e-05, "loss": 0.0, "step": 1802 }, { "epoch": 0.42, "learning_rate": 4.305523457360758e-05, "loss": 0.0, "step": 1803 }, { "epoch": 0.42, "learning_rate": 4.305138279023188e-05, "loss": 0.0, "step": 1804 }, { "epoch": 0.42, "learning_rate": 4.304753100685618e-05, "loss": 0.0, "step": 1805 }, { "epoch": 0.42, "learning_rate": 4.304367922348047e-05, "loss": 0.0, "step": 1806 }, { "epoch": 0.42, "learning_rate": 4.303982744010477e-05, "loss": 0.0, "step": 1807 }, { "epoch": 0.42, "learning_rate": 4.303597565672906e-05, "loss": 0.0003, "step": 1808 }, { "epoch": 0.42, "learning_rate": 4.3032123873353365e-05, "loss": 0.0, "step": 1809 }, { "epoch": 0.42, "learning_rate": 4.302827208997766e-05, "loss": 0.0, "step": 1810 }, { "epoch": 0.42, "learning_rate": 4.302442030660196e-05, "loss": 0.0, "step": 1811 }, { "epoch": 0.42, "learning_rate": 4.3020568523226257e-05, "loss": 0.0, "step": 1812 }, { "epoch": 0.42, "learning_rate": 4.301671673985055e-05, "loss": 0.0, "step": 1813 }, { "epoch": 0.42, "learning_rate": 4.301286495647485e-05, "loss": 0.0, "step": 1814 }, { "epoch": 0.42, "learning_rate": 4.300901317309915e-05, "loss": 0.0, "step": 1815 }, { "epoch": 0.42, "learning_rate": 4.300516138972344e-05, "loss": 0.0, "step": 1816 }, { "epoch": 0.42, "learning_rate": 4.300130960634774e-05, "loss": 0.0, "step": 1817 }, { "epoch": 0.42, "learning_rate": 4.299745782297204e-05, "loss": 0.0, "step": 1818 }, { "epoch": 0.42, "learning_rate": 4.2993606039596335e-05, "loss": 0.0, "step": 1819 }, { "epoch": 0.42, "learning_rate": 4.298975425622063e-05, "loss": 0.0, "step": 1820 }, { "epoch": 0.42, "learning_rate": 4.298590247284493e-05, "loss": 0.0, "step": 1821 }, { "epoch": 0.42, "learning_rate": 4.2982050689469226e-05, "loss": 0.0, "step": 1822 }, { "epoch": 0.42, "learning_rate": 4.297819890609353e-05, "loss": 0.0001, "step": 1823 }, { "epoch": 0.42, "learning_rate": 4.297434712271782e-05, "loss": 0.0, "step": 1824 }, { "epoch": 0.42, "learning_rate": 4.297049533934212e-05, "loss": 0.0, "step": 1825 }, { "epoch": 0.42, "learning_rate": 4.296664355596641e-05, "loss": 0.0243, "step": 1826 }, { "epoch": 0.42, "learning_rate": 4.296279177259071e-05, "loss": 0.0, "step": 1827 }, { "epoch": 0.42, "learning_rate": 4.295893998921501e-05, "loss": 0.0, "step": 1828 }, { "epoch": 0.42, "learning_rate": 4.2955088205839305e-05, "loss": 0.0002, "step": 1829 }, { "epoch": 0.42, "learning_rate": 4.2951236422463606e-05, "loss": 0.0, "step": 1830 }, { "epoch": 0.42, "learning_rate": 4.29473846390879e-05, "loss": 0.0, "step": 1831 }, { "epoch": 0.42, "learning_rate": 4.2943532855712196e-05, "loss": 0.0, "step": 1832 }, { "epoch": 0.42, "learning_rate": 4.29396810723365e-05, "loss": 0.0003, "step": 1833 }, { "epoch": 0.42, "learning_rate": 4.2935829288960786e-05, "loss": 0.0, "step": 1834 }, { "epoch": 0.42, "learning_rate": 4.293197750558509e-05, "loss": 0.0, "step": 1835 }, { "epoch": 0.42, "learning_rate": 4.292812572220938e-05, "loss": 0.0, "step": 1836 }, { "epoch": 0.42, "learning_rate": 4.292427393883368e-05, "loss": 0.0, "step": 1837 }, { "epoch": 0.42, "learning_rate": 4.292042215545798e-05, "loss": 0.0, "step": 1838 }, { "epoch": 0.43, "learning_rate": 4.2916570372082275e-05, "loss": 0.0, "step": 1839 }, { "epoch": 0.43, "learning_rate": 4.2912718588706576e-05, "loss": 0.0, "step": 1840 }, { "epoch": 0.43, "learning_rate": 4.290886680533087e-05, "loss": 0.0001, "step": 1841 }, { "epoch": 0.43, "learning_rate": 4.2905015021955166e-05, "loss": 0.0, "step": 1842 }, { "epoch": 0.43, "learning_rate": 4.290116323857947e-05, "loss": 0.0, "step": 1843 }, { "epoch": 0.43, "learning_rate": 4.2897311455203756e-05, "loss": 0.0, "step": 1844 }, { "epoch": 0.43, "learning_rate": 4.289345967182806e-05, "loss": 0.0, "step": 1845 }, { "epoch": 0.43, "learning_rate": 4.288960788845235e-05, "loss": 0.0, "step": 1846 }, { "epoch": 0.43, "learning_rate": 4.2885756105076655e-05, "loss": 0.0, "step": 1847 }, { "epoch": 0.43, "learning_rate": 4.288190432170095e-05, "loss": 0.0, "step": 1848 }, { "epoch": 0.43, "learning_rate": 4.2878052538325244e-05, "loss": 0.0001, "step": 1849 }, { "epoch": 0.43, "learning_rate": 4.2874200754949546e-05, "loss": 0.0, "step": 1850 }, { "epoch": 0.43, "learning_rate": 4.287034897157384e-05, "loss": 0.0, "step": 1851 }, { "epoch": 0.43, "learning_rate": 4.286649718819814e-05, "loss": 0.0, "step": 1852 }, { "epoch": 0.43, "learning_rate": 4.286264540482243e-05, "loss": 0.0, "step": 1853 }, { "epoch": 0.43, "learning_rate": 4.285879362144673e-05, "loss": 0.0, "step": 1854 }, { "epoch": 0.43, "learning_rate": 4.285494183807103e-05, "loss": 0.0, "step": 1855 }, { "epoch": 0.43, "learning_rate": 4.285109005469532e-05, "loss": 0.0, "step": 1856 }, { "epoch": 0.43, "learning_rate": 4.2847238271319624e-05, "loss": 0.0, "step": 1857 }, { "epoch": 0.43, "learning_rate": 4.284338648794392e-05, "loss": 0.0, "step": 1858 }, { "epoch": 0.43, "learning_rate": 4.283953470456822e-05, "loss": 0.0, "step": 1859 }, { "epoch": 0.43, "learning_rate": 4.2835682921192516e-05, "loss": 0.0, "step": 1860 }, { "epoch": 0.43, "learning_rate": 4.283183113781681e-05, "loss": 0.0, "step": 1861 }, { "epoch": 0.43, "learning_rate": 4.2827979354441106e-05, "loss": 0.0, "step": 1862 }, { "epoch": 0.43, "learning_rate": 4.28241275710654e-05, "loss": 0.0021, "step": 1863 }, { "epoch": 0.43, "learning_rate": 4.28202757876897e-05, "loss": 0.0, "step": 1864 }, { "epoch": 0.43, "learning_rate": 4.2816424004314e-05, "loss": 0.0, "step": 1865 }, { "epoch": 0.43, "learning_rate": 4.28125722209383e-05, "loss": 0.0, "step": 1866 }, { "epoch": 0.43, "learning_rate": 4.2808720437562594e-05, "loss": 0.0, "step": 1867 }, { "epoch": 0.43, "learning_rate": 4.280486865418689e-05, "loss": 0.0, "step": 1868 }, { "epoch": 0.43, "learning_rate": 4.280101687081119e-05, "loss": 0.0, "step": 1869 }, { "epoch": 0.43, "learning_rate": 4.2797165087435486e-05, "loss": 0.0, "step": 1870 }, { "epoch": 0.43, "learning_rate": 4.279331330405979e-05, "loss": 0.0, "step": 1871 }, { "epoch": 0.43, "learning_rate": 4.2789461520684076e-05, "loss": 0.0, "step": 1872 }, { "epoch": 0.43, "learning_rate": 4.278560973730837e-05, "loss": 0.0, "step": 1873 }, { "epoch": 0.43, "learning_rate": 4.278175795393267e-05, "loss": 0.0, "step": 1874 }, { "epoch": 0.43, "learning_rate": 4.277790617055697e-05, "loss": 0.0, "step": 1875 }, { "epoch": 0.43, "learning_rate": 4.277405438718127e-05, "loss": 0.0, "step": 1876 }, { "epoch": 0.43, "learning_rate": 4.2770202603805564e-05, "loss": 0.0, "step": 1877 }, { "epoch": 0.43, "learning_rate": 4.276635082042986e-05, "loss": 0.0006, "step": 1878 }, { "epoch": 0.43, "learning_rate": 4.276249903705416e-05, "loss": 0.0, "step": 1879 }, { "epoch": 0.43, "learning_rate": 4.2758647253678456e-05, "loss": 0.0, "step": 1880 }, { "epoch": 0.43, "learning_rate": 4.275479547030275e-05, "loss": 0.0, "step": 1881 }, { "epoch": 0.43, "learning_rate": 4.2750943686927046e-05, "loss": 0.0, "step": 1882 }, { "epoch": 0.44, "learning_rate": 4.274709190355135e-05, "loss": 0.0, "step": 1883 }, { "epoch": 0.44, "learning_rate": 4.274324012017564e-05, "loss": 0.0, "step": 1884 }, { "epoch": 0.44, "learning_rate": 4.273938833679994e-05, "loss": 0.0, "step": 1885 }, { "epoch": 0.44, "learning_rate": 4.273553655342424e-05, "loss": 0.0, "step": 1886 }, { "epoch": 0.44, "learning_rate": 4.2731684770048534e-05, "loss": 0.0, "step": 1887 }, { "epoch": 0.44, "learning_rate": 4.2727832986672836e-05, "loss": 0.0, "step": 1888 }, { "epoch": 0.44, "learning_rate": 4.272398120329713e-05, "loss": 0.0, "step": 1889 }, { "epoch": 0.44, "learning_rate": 4.2720129419921426e-05, "loss": 0.0, "step": 1890 }, { "epoch": 0.44, "learning_rate": 4.271627763654572e-05, "loss": 0.0, "step": 1891 }, { "epoch": 0.44, "learning_rate": 4.2712425853170016e-05, "loss": 0.0, "step": 1892 }, { "epoch": 0.44, "learning_rate": 4.270857406979432e-05, "loss": 0.0, "step": 1893 }, { "epoch": 0.44, "learning_rate": 4.270472228641861e-05, "loss": 0.0, "step": 1894 }, { "epoch": 0.44, "learning_rate": 4.2700870503042914e-05, "loss": 0.0, "step": 1895 }, { "epoch": 0.44, "learning_rate": 4.269701871966721e-05, "loss": 0.0, "step": 1896 }, { "epoch": 0.44, "learning_rate": 4.2693166936291504e-05, "loss": 0.0, "step": 1897 }, { "epoch": 0.44, "learning_rate": 4.2689315152915806e-05, "loss": 0.0, "step": 1898 }, { "epoch": 0.44, "learning_rate": 4.26854633695401e-05, "loss": 0.0, "step": 1899 }, { "epoch": 0.44, "learning_rate": 4.2681611586164396e-05, "loss": 0.0, "step": 1900 }, { "epoch": 0.44, "learning_rate": 4.267775980278869e-05, "loss": 0.0, "step": 1901 }, { "epoch": 0.44, "learning_rate": 4.267390801941299e-05, "loss": 0.0, "step": 1902 }, { "epoch": 0.44, "learning_rate": 4.267005623603729e-05, "loss": 0.0, "step": 1903 }, { "epoch": 0.44, "learning_rate": 4.266620445266158e-05, "loss": 0.0, "step": 1904 }, { "epoch": 0.44, "learning_rate": 4.2662352669285884e-05, "loss": 0.0, "step": 1905 }, { "epoch": 0.44, "learning_rate": 4.265850088591018e-05, "loss": 0.0, "step": 1906 }, { "epoch": 0.44, "learning_rate": 4.265464910253448e-05, "loss": 0.0, "step": 1907 }, { "epoch": 0.44, "learning_rate": 4.2650797319158776e-05, "loss": 0.0001, "step": 1908 }, { "epoch": 0.44, "learning_rate": 4.264694553578307e-05, "loss": 0.0, "step": 1909 }, { "epoch": 0.44, "learning_rate": 4.2643093752407365e-05, "loss": 0.0001, "step": 1910 }, { "epoch": 0.44, "learning_rate": 4.263924196903166e-05, "loss": 0.0, "step": 1911 }, { "epoch": 0.44, "learning_rate": 4.263539018565596e-05, "loss": 0.0, "step": 1912 }, { "epoch": 0.44, "learning_rate": 4.263153840228026e-05, "loss": 0.0001, "step": 1913 }, { "epoch": 0.44, "learning_rate": 4.262768661890455e-05, "loss": 0.0, "step": 1914 }, { "epoch": 0.44, "learning_rate": 4.2623834835528854e-05, "loss": 0.0, "step": 1915 }, { "epoch": 0.44, "learning_rate": 4.261998305215315e-05, "loss": 0.0, "step": 1916 }, { "epoch": 0.44, "learning_rate": 4.261613126877745e-05, "loss": 0.0031, "step": 1917 }, { "epoch": 0.44, "learning_rate": 4.261227948540174e-05, "loss": 0.0, "step": 1918 }, { "epoch": 0.44, "learning_rate": 4.260842770202604e-05, "loss": 0.0, "step": 1919 }, { "epoch": 0.44, "learning_rate": 4.2604575918650335e-05, "loss": 0.0, "step": 1920 }, { "epoch": 0.44, "learning_rate": 4.260072413527463e-05, "loss": 0.0, "step": 1921 }, { "epoch": 0.44, "learning_rate": 4.259687235189893e-05, "loss": 0.0, "step": 1922 }, { "epoch": 0.44, "learning_rate": 4.259302056852323e-05, "loss": 0.0, "step": 1923 }, { "epoch": 0.44, "learning_rate": 4.258916878514753e-05, "loss": 0.0, "step": 1924 }, { "epoch": 0.44, "learning_rate": 4.2585317001771824e-05, "loss": 0.0001, "step": 1925 }, { "epoch": 0.45, "learning_rate": 4.258146521839612e-05, "loss": 0.0002, "step": 1926 }, { "epoch": 0.45, "learning_rate": 4.257761343502042e-05, "loss": 0.0, "step": 1927 }, { "epoch": 0.45, "learning_rate": 4.257376165164471e-05, "loss": 0.0, "step": 1928 }, { "epoch": 0.45, "learning_rate": 4.256990986826901e-05, "loss": 0.0, "step": 1929 }, { "epoch": 0.45, "learning_rate": 4.2566058084893305e-05, "loss": 0.0, "step": 1930 }, { "epoch": 0.45, "learning_rate": 4.256220630151761e-05, "loss": 0.0, "step": 1931 }, { "epoch": 0.45, "learning_rate": 4.25583545181419e-05, "loss": 0.0, "step": 1932 }, { "epoch": 0.45, "learning_rate": 4.25545027347662e-05, "loss": 0.0, "step": 1933 }, { "epoch": 0.45, "learning_rate": 4.25506509513905e-05, "loss": 0.0, "step": 1934 }, { "epoch": 0.45, "learning_rate": 4.2546799168014794e-05, "loss": 0.0, "step": 1935 }, { "epoch": 0.45, "learning_rate": 4.2542947384639095e-05, "loss": 0.0, "step": 1936 }, { "epoch": 0.45, "learning_rate": 4.2539095601263384e-05, "loss": 0.0001, "step": 1937 }, { "epoch": 0.45, "learning_rate": 4.2535243817887685e-05, "loss": 0.0, "step": 1938 }, { "epoch": 0.45, "learning_rate": 4.253139203451198e-05, "loss": 0.0, "step": 1939 }, { "epoch": 0.45, "learning_rate": 4.2527540251136275e-05, "loss": 0.0, "step": 1940 }, { "epoch": 0.45, "learning_rate": 4.252368846776058e-05, "loss": 0.0001, "step": 1941 }, { "epoch": 0.45, "learning_rate": 4.251983668438487e-05, "loss": 0.0001, "step": 1942 }, { "epoch": 0.45, "learning_rate": 4.2515984901009174e-05, "loss": 0.0, "step": 1943 }, { "epoch": 0.45, "learning_rate": 4.251213311763347e-05, "loss": 0.0, "step": 1944 }, { "epoch": 0.45, "learning_rate": 4.2508281334257763e-05, "loss": 0.0, "step": 1945 }, { "epoch": 0.45, "learning_rate": 4.250442955088206e-05, "loss": 0.0, "step": 1946 }, { "epoch": 0.45, "learning_rate": 4.2500577767506353e-05, "loss": 0.0, "step": 1947 }, { "epoch": 0.45, "learning_rate": 4.2496725984130655e-05, "loss": 0.0, "step": 1948 }, { "epoch": 0.45, "learning_rate": 4.249287420075495e-05, "loss": 0.0003, "step": 1949 }, { "epoch": 0.45, "learning_rate": 4.2489022417379245e-05, "loss": 0.0, "step": 1950 }, { "epoch": 0.45, "learning_rate": 4.248517063400355e-05, "loss": 0.0, "step": 1951 }, { "epoch": 0.45, "learning_rate": 4.248131885062784e-05, "loss": 0.0, "step": 1952 }, { "epoch": 0.45, "learning_rate": 4.2477467067252143e-05, "loss": 0.0, "step": 1953 }, { "epoch": 0.45, "learning_rate": 4.247361528387644e-05, "loss": 0.0, "step": 1954 }, { "epoch": 0.45, "learning_rate": 4.246976350050073e-05, "loss": 0.0, "step": 1955 }, { "epoch": 0.45, "learning_rate": 4.246591171712503e-05, "loss": 0.0, "step": 1956 }, { "epoch": 0.45, "learning_rate": 4.246205993374932e-05, "loss": 0.0, "step": 1957 }, { "epoch": 0.45, "learning_rate": 4.2458208150373625e-05, "loss": 0.0, "step": 1958 }, { "epoch": 0.45, "learning_rate": 4.245435636699792e-05, "loss": 0.0001, "step": 1959 }, { "epoch": 0.45, "learning_rate": 4.245050458362222e-05, "loss": 0.0, "step": 1960 }, { "epoch": 0.45, "learning_rate": 4.244665280024652e-05, "loss": 0.0, "step": 1961 }, { "epoch": 0.45, "learning_rate": 4.244280101687081e-05, "loss": 0.0, "step": 1962 }, { "epoch": 0.45, "learning_rate": 4.243894923349511e-05, "loss": 0.0, "step": 1963 }, { "epoch": 0.45, "learning_rate": 4.243509745011941e-05, "loss": 0.0, "step": 1964 }, { "epoch": 0.45, "learning_rate": 4.24312456667437e-05, "loss": 0.0, "step": 1965 }, { "epoch": 0.45, "learning_rate": 4.2427393883368e-05, "loss": 0.0, "step": 1966 }, { "epoch": 0.45, "learning_rate": 4.24235420999923e-05, "loss": 0.0, "step": 1967 }, { "epoch": 0.45, "learning_rate": 4.2419690316616595e-05, "loss": 0.0, "step": 1968 }, { "epoch": 0.46, "learning_rate": 4.241583853324089e-05, "loss": 0.0, "step": 1969 }, { "epoch": 0.46, "learning_rate": 4.241198674986519e-05, "loss": 0.0, "step": 1970 }, { "epoch": 0.46, "learning_rate": 4.2408134966489487e-05, "loss": 0.0004, "step": 1971 }, { "epoch": 0.46, "learning_rate": 4.240428318311379e-05, "loss": 0.0, "step": 1972 }, { "epoch": 0.46, "learning_rate": 4.240043139973808e-05, "loss": 0.0001, "step": 1973 }, { "epoch": 0.46, "learning_rate": 4.239657961636238e-05, "loss": 0.0, "step": 1974 }, { "epoch": 0.46, "learning_rate": 4.239272783298667e-05, "loss": 0.0, "step": 1975 }, { "epoch": 0.46, "learning_rate": 4.238887604961097e-05, "loss": 0.0002, "step": 1976 }, { "epoch": 0.46, "learning_rate": 4.238502426623527e-05, "loss": 0.0, "step": 1977 }, { "epoch": 0.46, "learning_rate": 4.2381172482859565e-05, "loss": 0.0, "step": 1978 }, { "epoch": 0.46, "learning_rate": 4.2377320699483867e-05, "loss": 0.0, "step": 1979 }, { "epoch": 0.46, "learning_rate": 4.237346891610816e-05, "loss": 0.0, "step": 1980 }, { "epoch": 0.46, "learning_rate": 4.2369617132732456e-05, "loss": 0.0, "step": 1981 }, { "epoch": 0.46, "learning_rate": 4.236576534935676e-05, "loss": 0.0, "step": 1982 }, { "epoch": 0.46, "learning_rate": 4.236191356598105e-05, "loss": 0.0, "step": 1983 }, { "epoch": 0.46, "learning_rate": 4.235806178260535e-05, "loss": 0.0, "step": 1984 }, { "epoch": 0.46, "learning_rate": 4.235420999922964e-05, "loss": 0.0, "step": 1985 }, { "epoch": 0.46, "learning_rate": 4.2350358215853945e-05, "loss": 0.0, "step": 1986 }, { "epoch": 0.46, "learning_rate": 4.234650643247824e-05, "loss": 0.0, "step": 1987 }, { "epoch": 0.46, "learning_rate": 4.2342654649102535e-05, "loss": 0.0, "step": 1988 }, { "epoch": 0.46, "learning_rate": 4.2338802865726836e-05, "loss": 0.0, "step": 1989 }, { "epoch": 0.46, "learning_rate": 4.233495108235113e-05, "loss": 0.0001, "step": 1990 }, { "epoch": 0.46, "learning_rate": 4.2331099298975426e-05, "loss": 0.0, "step": 1991 }, { "epoch": 0.46, "learning_rate": 4.232724751559973e-05, "loss": 0.0, "step": 1992 }, { "epoch": 0.46, "learning_rate": 4.2323395732224016e-05, "loss": 0.0, "step": 1993 }, { "epoch": 0.46, "learning_rate": 4.231954394884832e-05, "loss": 0.0, "step": 1994 }, { "epoch": 0.46, "learning_rate": 4.231569216547261e-05, "loss": 0.0, "step": 1995 }, { "epoch": 0.46, "learning_rate": 4.2311840382096915e-05, "loss": 0.0, "step": 1996 }, { "epoch": 0.46, "learning_rate": 4.230798859872121e-05, "loss": 0.0029, "step": 1997 }, { "epoch": 0.46, "learning_rate": 4.2304136815345505e-05, "loss": 0.0, "step": 1998 }, { "epoch": 0.46, "learning_rate": 4.2300285031969806e-05, "loss": 0.0, "step": 1999 }, { "epoch": 0.46, "learning_rate": 4.22964332485941e-05, "loss": 0.0, "step": 2000 }, { "epoch": 0.46, "learning_rate": 4.22925814652184e-05, "loss": 0.0, "step": 2001 }, { "epoch": 0.46, "learning_rate": 4.228872968184269e-05, "loss": 0.0, "step": 2002 }, { "epoch": 0.46, "learning_rate": 4.228487789846699e-05, "loss": 0.0, "step": 2003 }, { "epoch": 0.46, "learning_rate": 4.228102611509129e-05, "loss": 0.0, "step": 2004 }, { "epoch": 0.46, "learning_rate": 4.227717433171558e-05, "loss": 0.0, "step": 2005 }, { "epoch": 0.46, "learning_rate": 4.2273322548339885e-05, "loss": 0.0, "step": 2006 }, { "epoch": 0.46, "learning_rate": 4.226947076496418e-05, "loss": 0.0, "step": 2007 }, { "epoch": 0.46, "learning_rate": 4.226561898158848e-05, "loss": 0.0, "step": 2008 }, { "epoch": 0.46, "learning_rate": 4.2261767198212776e-05, "loss": 0.0, "step": 2009 }, { "epoch": 0.46, "learning_rate": 4.225791541483707e-05, "loss": 0.0013, "step": 2010 }, { "epoch": 0.46, "learning_rate": 4.225406363146137e-05, "loss": 0.0, "step": 2011 }, { "epoch": 0.46, "learning_rate": 4.225021184808566e-05, "loss": 0.0, "step": 2012 }, { "epoch": 0.47, "learning_rate": 4.224636006470996e-05, "loss": 0.0, "step": 2013 }, { "epoch": 0.47, "learning_rate": 4.224250828133426e-05, "loss": 0.0, "step": 2014 }, { "epoch": 0.47, "learning_rate": 4.223865649795856e-05, "loss": 0.0, "step": 2015 }, { "epoch": 0.47, "learning_rate": 4.2234804714582854e-05, "loss": 0.0, "step": 2016 }, { "epoch": 0.47, "learning_rate": 4.223095293120715e-05, "loss": 0.0, "step": 2017 }, { "epoch": 0.47, "learning_rate": 4.222710114783145e-05, "loss": 0.0, "step": 2018 }, { "epoch": 0.47, "learning_rate": 4.2223249364455746e-05, "loss": 0.0, "step": 2019 }, { "epoch": 0.47, "learning_rate": 4.221939758108005e-05, "loss": 0.0, "step": 2020 }, { "epoch": 0.47, "learning_rate": 4.2215545797704336e-05, "loss": 0.0, "step": 2021 }, { "epoch": 0.47, "learning_rate": 4.221169401432864e-05, "loss": 0.0, "step": 2022 }, { "epoch": 0.47, "learning_rate": 4.220784223095293e-05, "loss": 0.0, "step": 2023 }, { "epoch": 0.47, "learning_rate": 4.220399044757723e-05, "loss": 0.0003, "step": 2024 }, { "epoch": 0.47, "learning_rate": 4.220013866420153e-05, "loss": 0.0, "step": 2025 }, { "epoch": 0.47, "learning_rate": 4.2196286880825824e-05, "loss": 0.0, "step": 2026 }, { "epoch": 0.47, "learning_rate": 4.219243509745012e-05, "loss": 0.0, "step": 2027 }, { "epoch": 0.47, "learning_rate": 4.218858331407442e-05, "loss": 0.0, "step": 2028 }, { "epoch": 0.47, "learning_rate": 4.2184731530698716e-05, "loss": 0.0, "step": 2029 }, { "epoch": 0.47, "learning_rate": 4.218087974732301e-05, "loss": 0.0001, "step": 2030 }, { "epoch": 0.47, "learning_rate": 4.2177027963947306e-05, "loss": 0.0, "step": 2031 }, { "epoch": 0.47, "learning_rate": 4.217317618057161e-05, "loss": 0.0, "step": 2032 }, { "epoch": 0.47, "learning_rate": 4.21693243971959e-05, "loss": 0.0, "step": 2033 }, { "epoch": 0.47, "learning_rate": 4.21654726138202e-05, "loss": 0.0, "step": 2034 }, { "epoch": 0.47, "learning_rate": 4.21616208304445e-05, "loss": 0.0, "step": 2035 }, { "epoch": 0.47, "learning_rate": 4.2157769047068794e-05, "loss": 0.0, "step": 2036 }, { "epoch": 0.47, "learning_rate": 4.2153917263693096e-05, "loss": 0.0001, "step": 2037 }, { "epoch": 0.47, "learning_rate": 4.215006548031739e-05, "loss": 0.0, "step": 2038 }, { "epoch": 0.47, "learning_rate": 4.2146213696941686e-05, "loss": 0.0, "step": 2039 }, { "epoch": 0.47, "learning_rate": 4.214236191356598e-05, "loss": 0.0, "step": 2040 }, { "epoch": 0.47, "learning_rate": 4.2138510130190276e-05, "loss": 0.0005, "step": 2041 }, { "epoch": 0.47, "learning_rate": 4.213465834681458e-05, "loss": 0.0, "step": 2042 }, { "epoch": 0.47, "learning_rate": 4.213080656343887e-05, "loss": 0.0001, "step": 2043 }, { "epoch": 0.47, "learning_rate": 4.2126954780063174e-05, "loss": 0.0, "step": 2044 }, { "epoch": 0.47, "learning_rate": 4.212310299668747e-05, "loss": 0.0, "step": 2045 }, { "epoch": 0.47, "learning_rate": 4.2119251213311764e-05, "loss": 0.0, "step": 2046 }, { "epoch": 0.47, "learning_rate": 4.2115399429936066e-05, "loss": 0.0, "step": 2047 }, { "epoch": 0.47, "learning_rate": 4.211154764656036e-05, "loss": 0.0037, "step": 2048 }, { "epoch": 0.47, "learning_rate": 4.2107695863184656e-05, "loss": 0.0, "step": 2049 }, { "epoch": 0.47, "learning_rate": 4.210384407980895e-05, "loss": 0.0, "step": 2050 }, { "epoch": 0.47, "learning_rate": 4.209999229643325e-05, "loss": 0.0, "step": 2051 }, { "epoch": 0.47, "learning_rate": 4.209614051305755e-05, "loss": 0.0, "step": 2052 }, { "epoch": 0.47, "learning_rate": 4.209228872968184e-05, "loss": 0.0, "step": 2053 }, { "epoch": 0.47, "learning_rate": 4.2088436946306144e-05, "loss": 0.0, "step": 2054 }, { "epoch": 0.47, "learning_rate": 4.208458516293044e-05, "loss": 0.0, "step": 2055 }, { "epoch": 0.48, "learning_rate": 4.208073337955474e-05, "loss": 0.0001, "step": 2056 }, { "epoch": 0.48, "learning_rate": 4.2076881596179036e-05, "loss": 0.0, "step": 2057 }, { "epoch": 0.48, "learning_rate": 4.207302981280333e-05, "loss": 0.0, "step": 2058 }, { "epoch": 0.48, "learning_rate": 4.2069178029427626e-05, "loss": 0.0, "step": 2059 }, { "epoch": 0.48, "learning_rate": 4.206532624605192e-05, "loss": 0.0, "step": 2060 }, { "epoch": 0.48, "learning_rate": 4.206147446267622e-05, "loss": 0.0, "step": 2061 }, { "epoch": 0.48, "learning_rate": 4.205762267930052e-05, "loss": 0.0, "step": 2062 }, { "epoch": 0.48, "learning_rate": 4.205377089592481e-05, "loss": 0.0, "step": 2063 }, { "epoch": 0.48, "learning_rate": 4.2049919112549114e-05, "loss": 0.0, "step": 2064 }, { "epoch": 0.48, "learning_rate": 4.204606732917341e-05, "loss": 0.0, "step": 2065 }, { "epoch": 0.48, "learning_rate": 4.204221554579771e-05, "loss": 0.0, "step": 2066 }, { "epoch": 0.48, "learning_rate": 4.2038363762422006e-05, "loss": 0.0, "step": 2067 }, { "epoch": 0.48, "learning_rate": 4.20345119790463e-05, "loss": 0.0, "step": 2068 }, { "epoch": 0.48, "learning_rate": 4.2030660195670595e-05, "loss": 0.0, "step": 2069 }, { "epoch": 0.48, "learning_rate": 4.202680841229489e-05, "loss": 0.0, "step": 2070 }, { "epoch": 0.48, "learning_rate": 4.202295662891919e-05, "loss": 0.0, "step": 2071 }, { "epoch": 0.48, "learning_rate": 4.201910484554349e-05, "loss": 0.0, "step": 2072 }, { "epoch": 0.48, "learning_rate": 4.201525306216779e-05, "loss": 0.0, "step": 2073 }, { "epoch": 0.48, "learning_rate": 4.2011401278792084e-05, "loss": 0.0001, "step": 2074 }, { "epoch": 0.48, "learning_rate": 4.200754949541638e-05, "loss": 0.0001, "step": 2075 }, { "epoch": 0.48, "learning_rate": 4.200369771204068e-05, "loss": 0.0, "step": 2076 }, { "epoch": 0.48, "learning_rate": 4.199984592866497e-05, "loss": 0.0001, "step": 2077 }, { "epoch": 0.48, "learning_rate": 4.199599414528927e-05, "loss": 0.0, "step": 2078 }, { "epoch": 0.48, "learning_rate": 4.1992142361913565e-05, "loss": 0.0002, "step": 2079 }, { "epoch": 0.48, "learning_rate": 4.198829057853787e-05, "loss": 0.0, "step": 2080 }, { "epoch": 0.48, "learning_rate": 4.198443879516216e-05, "loss": 0.0, "step": 2081 }, { "epoch": 0.48, "learning_rate": 4.198058701178646e-05, "loss": 0.0, "step": 2082 }, { "epoch": 0.48, "learning_rate": 4.197673522841076e-05, "loss": 0.0, "step": 2083 }, { "epoch": 0.48, "learning_rate": 4.1972883445035054e-05, "loss": 0.0, "step": 2084 }, { "epoch": 0.48, "learning_rate": 4.1969031661659355e-05, "loss": 0.0, "step": 2085 }, { "epoch": 0.48, "learning_rate": 4.1965179878283644e-05, "loss": 0.0002, "step": 2086 }, { "epoch": 0.48, "learning_rate": 4.1961328094907945e-05, "loss": 0.0, "step": 2087 }, { "epoch": 0.48, "learning_rate": 4.195747631153224e-05, "loss": 0.0, "step": 2088 }, { "epoch": 0.48, "learning_rate": 4.1953624528156535e-05, "loss": 0.0, "step": 2089 }, { "epoch": 0.48, "learning_rate": 4.194977274478084e-05, "loss": 0.0, "step": 2090 }, { "epoch": 0.48, "learning_rate": 4.194592096140513e-05, "loss": 0.0, "step": 2091 }, { "epoch": 0.48, "learning_rate": 4.1942069178029434e-05, "loss": 0.0, "step": 2092 }, { "epoch": 0.48, "learning_rate": 4.193821739465373e-05, "loss": 0.0, "step": 2093 }, { "epoch": 0.48, "learning_rate": 4.1934365611278024e-05, "loss": 0.0, "step": 2094 }, { "epoch": 0.48, "learning_rate": 4.1930513827902325e-05, "loss": 0.0, "step": 2095 }, { "epoch": 0.48, "learning_rate": 4.1926662044526613e-05, "loss": 0.0, "step": 2096 }, { "epoch": 0.48, "learning_rate": 4.1922810261150915e-05, "loss": 0.0, "step": 2097 }, { "epoch": 0.48, "learning_rate": 4.191895847777521e-05, "loss": 0.0, "step": 2098 }, { "epoch": 0.49, "learning_rate": 4.191510669439951e-05, "loss": 0.0, "step": 2099 }, { "epoch": 0.49, "learning_rate": 4.191125491102381e-05, "loss": 0.0, "step": 2100 }, { "epoch": 0.49, "learning_rate": 4.19074031276481e-05, "loss": 0.0, "step": 2101 }, { "epoch": 0.49, "learning_rate": 4.1903551344272404e-05, "loss": 0.0001, "step": 2102 }, { "epoch": 0.49, "learning_rate": 4.18996995608967e-05, "loss": 0.0001, "step": 2103 }, { "epoch": 0.49, "learning_rate": 4.1895847777520993e-05, "loss": 0.0, "step": 2104 }, { "epoch": 0.49, "learning_rate": 4.189199599414529e-05, "loss": 0.0, "step": 2105 }, { "epoch": 0.49, "learning_rate": 4.188814421076958e-05, "loss": 0.0, "step": 2106 }, { "epoch": 0.49, "learning_rate": 4.1884292427393885e-05, "loss": 0.0, "step": 2107 }, { "epoch": 0.49, "learning_rate": 4.188044064401818e-05, "loss": 0.0, "step": 2108 }, { "epoch": 0.49, "learning_rate": 4.187658886064248e-05, "loss": 0.0, "step": 2109 }, { "epoch": 0.49, "learning_rate": 4.187273707726678e-05, "loss": 0.0001, "step": 2110 }, { "epoch": 0.49, "learning_rate": 4.186888529389107e-05, "loss": 0.0, "step": 2111 }, { "epoch": 0.49, "learning_rate": 4.1865033510515373e-05, "loss": 0.0, "step": 2112 }, { "epoch": 0.49, "learning_rate": 4.186118172713967e-05, "loss": 0.0, "step": 2113 }, { "epoch": 0.49, "learning_rate": 4.185732994376396e-05, "loss": 0.0, "step": 2114 }, { "epoch": 0.49, "learning_rate": 4.185347816038826e-05, "loss": 0.0, "step": 2115 }, { "epoch": 0.49, "learning_rate": 4.184962637701256e-05, "loss": 0.0, "step": 2116 }, { "epoch": 0.49, "learning_rate": 4.1845774593636855e-05, "loss": 0.0, "step": 2117 }, { "epoch": 0.49, "learning_rate": 4.184192281026115e-05, "loss": 0.0, "step": 2118 }, { "epoch": 0.49, "learning_rate": 4.183807102688545e-05, "loss": 0.0, "step": 2119 }, { "epoch": 0.49, "learning_rate": 4.183421924350975e-05, "loss": 0.0, "step": 2120 }, { "epoch": 0.49, "learning_rate": 4.183036746013405e-05, "loss": 0.0, "step": 2121 }, { "epoch": 0.49, "learning_rate": 4.182651567675834e-05, "loss": 0.0, "step": 2122 }, { "epoch": 0.49, "learning_rate": 4.182266389338264e-05, "loss": 0.0, "step": 2123 }, { "epoch": 0.49, "learning_rate": 4.181881211000693e-05, "loss": 0.0, "step": 2124 }, { "epoch": 0.49, "learning_rate": 4.181496032663123e-05, "loss": 0.0, "step": 2125 }, { "epoch": 0.49, "learning_rate": 4.181110854325553e-05, "loss": 0.0001, "step": 2126 }, { "epoch": 0.49, "learning_rate": 4.1807256759879825e-05, "loss": 0.0, "step": 2127 }, { "epoch": 0.49, "learning_rate": 4.1803404976504127e-05, "loss": 0.0001, "step": 2128 }, { "epoch": 0.49, "learning_rate": 4.179955319312842e-05, "loss": 0.0, "step": 2129 }, { "epoch": 0.49, "learning_rate": 4.1795701409752717e-05, "loss": 0.0, "step": 2130 }, { "epoch": 0.49, "learning_rate": 4.179184962637702e-05, "loss": 0.0001, "step": 2131 }, { "epoch": 0.49, "learning_rate": 4.178799784300131e-05, "loss": 0.0, "step": 2132 }, { "epoch": 0.49, "learning_rate": 4.178414605962561e-05, "loss": 0.0, "step": 2133 }, { "epoch": 0.49, "learning_rate": 4.17802942762499e-05, "loss": 0.0, "step": 2134 }, { "epoch": 0.49, "learning_rate": 4.1776442492874205e-05, "loss": 0.0, "step": 2135 }, { "epoch": 0.49, "learning_rate": 4.17725907094985e-05, "loss": 0.0, "step": 2136 }, { "epoch": 0.49, "learning_rate": 4.1768738926122795e-05, "loss": 0.0, "step": 2137 }, { "epoch": 0.49, "learning_rate": 4.1764887142747096e-05, "loss": 0.0, "step": 2138 }, { "epoch": 0.49, "learning_rate": 4.176103535937139e-05, "loss": 0.0, "step": 2139 }, { "epoch": 0.49, "learning_rate": 4.1757183575995686e-05, "loss": 0.0, "step": 2140 }, { "epoch": 0.49, "learning_rate": 4.175333179261999e-05, "loss": 0.0, "step": 2141 }, { "epoch": 0.5, "learning_rate": 4.1749480009244276e-05, "loss": 0.0, "step": 2142 }, { "epoch": 0.5, "learning_rate": 4.174562822586858e-05, "loss": 0.0, "step": 2143 }, { "epoch": 0.5, "learning_rate": 4.174177644249287e-05, "loss": 0.0, "step": 2144 }, { "epoch": 0.5, "learning_rate": 4.1737924659117175e-05, "loss": 0.0, "step": 2145 }, { "epoch": 0.5, "learning_rate": 4.173407287574147e-05, "loss": 0.0, "step": 2146 }, { "epoch": 0.5, "learning_rate": 4.1730221092365765e-05, "loss": 0.0, "step": 2147 }, { "epoch": 0.5, "learning_rate": 4.1726369308990066e-05, "loss": 0.0, "step": 2148 }, { "epoch": 0.5, "learning_rate": 4.172251752561436e-05, "loss": 0.0, "step": 2149 }, { "epoch": 0.5, "learning_rate": 4.171866574223866e-05, "loss": 0.0, "step": 2150 }, { "epoch": 0.5, "learning_rate": 4.171481395886296e-05, "loss": 0.0001, "step": 2151 }, { "epoch": 0.5, "learning_rate": 4.171096217548725e-05, "loss": 0.0002, "step": 2152 }, { "epoch": 0.5, "learning_rate": 4.170711039211155e-05, "loss": 0.0, "step": 2153 }, { "epoch": 0.5, "learning_rate": 4.170325860873584e-05, "loss": 0.0, "step": 2154 }, { "epoch": 0.5, "learning_rate": 4.1699406825360145e-05, "loss": 0.0, "step": 2155 }, { "epoch": 0.5, "learning_rate": 4.169555504198444e-05, "loss": 0.0, "step": 2156 }, { "epoch": 0.5, "learning_rate": 4.169170325860874e-05, "loss": 0.0, "step": 2157 }, { "epoch": 0.5, "learning_rate": 4.1687851475233036e-05, "loss": 0.0001, "step": 2158 }, { "epoch": 0.5, "learning_rate": 4.168399969185733e-05, "loss": 0.0, "step": 2159 }, { "epoch": 0.5, "learning_rate": 4.168014790848163e-05, "loss": 0.0, "step": 2160 }, { "epoch": 0.5, "learning_rate": 4.167629612510592e-05, "loss": 0.0, "step": 2161 }, { "epoch": 0.5, "learning_rate": 4.167244434173022e-05, "loss": 0.0, "step": 2162 }, { "epoch": 0.5, "learning_rate": 4.166859255835452e-05, "loss": 0.0, "step": 2163 }, { "epoch": 0.5, "learning_rate": 4.166474077497882e-05, "loss": 0.0, "step": 2164 }, { "epoch": 0.5, "learning_rate": 4.1660888991603114e-05, "loss": 0.0, "step": 2165 }, { "epoch": 0.5, "learning_rate": 4.165703720822741e-05, "loss": 0.0058, "step": 2166 }, { "epoch": 0.5, "learning_rate": 4.165318542485171e-05, "loss": 0.0, "step": 2167 }, { "epoch": 0.5, "learning_rate": 4.1649333641476006e-05, "loss": 0.0, "step": 2168 }, { "epoch": 0.5, "learning_rate": 4.164548185810031e-05, "loss": 0.0, "step": 2169 }, { "epoch": 0.5, "learning_rate": 4.1641630074724596e-05, "loss": 0.0, "step": 2170 }, { "epoch": 0.5, "learning_rate": 4.16377782913489e-05, "loss": 0.0, "step": 2171 }, { "epoch": 0.5, "learning_rate": 4.163392650797319e-05, "loss": 0.0, "step": 2172 }, { "epoch": 0.5, "learning_rate": 4.163007472459749e-05, "loss": 0.0002, "step": 2173 }, { "epoch": 0.5, "learning_rate": 4.162622294122179e-05, "loss": 0.0, "step": 2174 }, { "epoch": 0.5, "learning_rate": 4.1622371157846084e-05, "loss": 0.0, "step": 2175 }, { "epoch": 0.5, "learning_rate": 4.1618519374470386e-05, "loss": 0.0, "step": 2176 }, { "epoch": 0.5, "learning_rate": 4.161466759109468e-05, "loss": 0.0, "step": 2177 }, { "epoch": 0.5, "learning_rate": 4.1610815807718976e-05, "loss": 0.0, "step": 2178 }, { "epoch": 0.5, "learning_rate": 4.160696402434328e-05, "loss": 0.0, "step": 2179 }, { "epoch": 0.5, "learning_rate": 4.1603112240967566e-05, "loss": 0.0, "step": 2180 }, { "epoch": 0.5, "learning_rate": 4.159926045759187e-05, "loss": 0.0, "step": 2181 }, { "epoch": 0.5, "learning_rate": 4.159540867421616e-05, "loss": 0.0, "step": 2182 }, { "epoch": 0.5, "learning_rate": 4.159155689084046e-05, "loss": 0.0, "step": 2183 }, { "epoch": 0.5, "learning_rate": 4.158770510746476e-05, "loss": 0.0, "step": 2184 }, { "epoch": 0.5, "learning_rate": 4.1583853324089054e-05, "loss": 0.0002, "step": 2185 }, { "epoch": 0.51, "learning_rate": 4.1580001540713356e-05, "loss": 0.0, "step": 2186 }, { "epoch": 0.51, "learning_rate": 4.157614975733765e-05, "loss": 0.0, "step": 2187 }, { "epoch": 0.51, "learning_rate": 4.1572297973961946e-05, "loss": 0.0, "step": 2188 }, { "epoch": 0.51, "learning_rate": 4.156844619058624e-05, "loss": 0.0, "step": 2189 }, { "epoch": 0.51, "learning_rate": 4.1564594407210536e-05, "loss": 0.0, "step": 2190 }, { "epoch": 0.51, "learning_rate": 4.156074262383484e-05, "loss": 0.0, "step": 2191 }, { "epoch": 0.51, "learning_rate": 4.155689084045913e-05, "loss": 0.0, "step": 2192 }, { "epoch": 0.51, "learning_rate": 4.1553039057083434e-05, "loss": 0.0, "step": 2193 }, { "epoch": 0.51, "learning_rate": 4.154918727370773e-05, "loss": 0.0, "step": 2194 }, { "epoch": 0.51, "learning_rate": 4.1545335490332024e-05, "loss": 0.0, "step": 2195 }, { "epoch": 0.51, "learning_rate": 4.1541483706956326e-05, "loss": 0.0064, "step": 2196 }, { "epoch": 0.51, "learning_rate": 4.153763192358062e-05, "loss": 0.0, "step": 2197 }, { "epoch": 0.51, "learning_rate": 4.1533780140204916e-05, "loss": 0.0, "step": 2198 }, { "epoch": 0.51, "learning_rate": 4.152992835682921e-05, "loss": 0.0008, "step": 2199 }, { "epoch": 0.51, "learning_rate": 4.152607657345351e-05, "loss": 0.0, "step": 2200 }, { "epoch": 0.51, "learning_rate": 4.152222479007781e-05, "loss": 0.0001, "step": 2201 }, { "epoch": 0.51, "learning_rate": 4.15183730067021e-05, "loss": 0.0, "step": 2202 }, { "epoch": 0.51, "learning_rate": 4.1514521223326404e-05, "loss": 0.0, "step": 2203 }, { "epoch": 0.51, "learning_rate": 4.15106694399507e-05, "loss": 0.0, "step": 2204 }, { "epoch": 0.51, "learning_rate": 4.1506817656575e-05, "loss": 0.0, "step": 2205 }, { "epoch": 0.51, "learning_rate": 4.1502965873199296e-05, "loss": 0.0, "step": 2206 }, { "epoch": 0.51, "learning_rate": 4.149911408982359e-05, "loss": 0.0, "step": 2207 }, { "epoch": 0.51, "learning_rate": 4.1495262306447886e-05, "loss": 0.0, "step": 2208 }, { "epoch": 0.51, "learning_rate": 4.149141052307218e-05, "loss": 0.0, "step": 2209 }, { "epoch": 0.51, "learning_rate": 4.148755873969648e-05, "loss": 0.0, "step": 2210 }, { "epoch": 0.51, "learning_rate": 4.148370695632078e-05, "loss": 0.0001, "step": 2211 }, { "epoch": 0.51, "learning_rate": 4.147985517294508e-05, "loss": 0.0, "step": 2212 }, { "epoch": 0.51, "learning_rate": 4.1476003389569374e-05, "loss": 0.0, "step": 2213 }, { "epoch": 0.51, "learning_rate": 4.147215160619367e-05, "loss": 0.0, "step": 2214 }, { "epoch": 0.51, "learning_rate": 4.146829982281797e-05, "loss": 0.0001, "step": 2215 }, { "epoch": 0.51, "learning_rate": 4.1464448039442266e-05, "loss": 0.0, "step": 2216 }, { "epoch": 0.51, "learning_rate": 4.146059625606656e-05, "loss": 0.0, "step": 2217 }, { "epoch": 0.51, "learning_rate": 4.1456744472690856e-05, "loss": 0.0, "step": 2218 }, { "epoch": 0.51, "learning_rate": 4.145289268931515e-05, "loss": 0.0, "step": 2219 }, { "epoch": 0.51, "learning_rate": 4.144904090593945e-05, "loss": 0.0, "step": 2220 }, { "epoch": 0.51, "learning_rate": 4.144518912256375e-05, "loss": 0.0, "step": 2221 }, { "epoch": 0.51, "learning_rate": 4.144133733918805e-05, "loss": 0.0, "step": 2222 }, { "epoch": 0.51, "learning_rate": 4.1437485555812344e-05, "loss": 0.0, "step": 2223 }, { "epoch": 0.51, "learning_rate": 4.143363377243664e-05, "loss": 0.0, "step": 2224 }, { "epoch": 0.51, "learning_rate": 4.142978198906094e-05, "loss": 0.0001, "step": 2225 }, { "epoch": 0.51, "learning_rate": 4.142593020568523e-05, "loss": 0.0001, "step": 2226 }, { "epoch": 0.51, "learning_rate": 4.142207842230953e-05, "loss": 0.0, "step": 2227 }, { "epoch": 0.51, "learning_rate": 4.1418226638933825e-05, "loss": 0.0082, "step": 2228 }, { "epoch": 0.52, "learning_rate": 4.141437485555813e-05, "loss": 0.0001, "step": 2229 }, { "epoch": 0.52, "learning_rate": 4.141052307218242e-05, "loss": 0.0, "step": 2230 }, { "epoch": 0.52, "learning_rate": 4.140667128880672e-05, "loss": 0.0, "step": 2231 }, { "epoch": 0.52, "learning_rate": 4.140281950543102e-05, "loss": 0.0, "step": 2232 }, { "epoch": 0.52, "learning_rate": 4.1398967722055314e-05, "loss": 0.0, "step": 2233 }, { "epoch": 0.52, "learning_rate": 4.1395115938679615e-05, "loss": 0.0, "step": 2234 }, { "epoch": 0.52, "learning_rate": 4.139126415530391e-05, "loss": 0.0, "step": 2235 }, { "epoch": 0.52, "learning_rate": 4.1387412371928205e-05, "loss": 0.0, "step": 2236 }, { "epoch": 0.52, "learning_rate": 4.13835605885525e-05, "loss": 0.0, "step": 2237 }, { "epoch": 0.52, "learning_rate": 4.1379708805176795e-05, "loss": 0.0, "step": 2238 }, { "epoch": 0.52, "learning_rate": 4.13758570218011e-05, "loss": 0.0005, "step": 2239 }, { "epoch": 0.52, "learning_rate": 4.137200523842539e-05, "loss": 0.0, "step": 2240 }, { "epoch": 0.52, "learning_rate": 4.1368153455049694e-05, "loss": 0.0, "step": 2241 }, { "epoch": 0.52, "learning_rate": 4.136430167167399e-05, "loss": 0.0, "step": 2242 }, { "epoch": 0.52, "learning_rate": 4.1360449888298284e-05, "loss": 0.0, "step": 2243 }, { "epoch": 0.52, "learning_rate": 4.1356598104922585e-05, "loss": 0.0, "step": 2244 }, { "epoch": 0.52, "learning_rate": 4.1352746321546874e-05, "loss": 0.0, "step": 2245 }, { "epoch": 0.52, "learning_rate": 4.1348894538171175e-05, "loss": 0.0, "step": 2246 }, { "epoch": 0.52, "learning_rate": 4.134504275479547e-05, "loss": 0.0, "step": 2247 }, { "epoch": 0.52, "learning_rate": 4.134119097141977e-05, "loss": 0.0, "step": 2248 }, { "epoch": 0.52, "learning_rate": 4.133733918804407e-05, "loss": 0.0, "step": 2249 }, { "epoch": 0.52, "learning_rate": 4.133348740466836e-05, "loss": 0.0, "step": 2250 }, { "epoch": 0.52, "learning_rate": 4.1329635621292664e-05, "loss": 0.0, "step": 2251 }, { "epoch": 0.52, "learning_rate": 4.132578383791696e-05, "loss": 0.0, "step": 2252 }, { "epoch": 0.52, "learning_rate": 4.132193205454126e-05, "loss": 0.0, "step": 2253 }, { "epoch": 0.52, "learning_rate": 4.131808027116555e-05, "loss": 0.0, "step": 2254 }, { "epoch": 0.52, "learning_rate": 4.1314228487789843e-05, "loss": 0.0, "step": 2255 }, { "epoch": 0.52, "learning_rate": 4.1310376704414145e-05, "loss": 0.0, "step": 2256 }, { "epoch": 0.52, "learning_rate": 4.130652492103844e-05, "loss": 0.0144, "step": 2257 }, { "epoch": 0.52, "learning_rate": 4.130267313766274e-05, "loss": 0.0, "step": 2258 }, { "epoch": 0.52, "learning_rate": 4.129882135428704e-05, "loss": 0.0, "step": 2259 }, { "epoch": 0.52, "learning_rate": 4.129496957091133e-05, "loss": 0.0006, "step": 2260 }, { "epoch": 0.52, "learning_rate": 4.1291117787535634e-05, "loss": 0.0, "step": 2261 }, { "epoch": 0.52, "learning_rate": 4.128726600415993e-05, "loss": 0.0, "step": 2262 }, { "epoch": 0.52, "learning_rate": 4.128341422078423e-05, "loss": 0.0, "step": 2263 }, { "epoch": 0.52, "learning_rate": 4.127956243740852e-05, "loss": 0.0, "step": 2264 }, { "epoch": 0.52, "learning_rate": 4.127571065403282e-05, "loss": 0.0, "step": 2265 }, { "epoch": 0.52, "learning_rate": 4.1271858870657115e-05, "loss": 0.0, "step": 2266 }, { "epoch": 0.52, "learning_rate": 4.126800708728141e-05, "loss": 0.0, "step": 2267 }, { "epoch": 0.52, "learning_rate": 4.126415530390571e-05, "loss": 0.0, "step": 2268 }, { "epoch": 0.52, "learning_rate": 4.126030352053001e-05, "loss": 0.0, "step": 2269 }, { "epoch": 0.52, "learning_rate": 4.125645173715431e-05, "loss": 0.0, "step": 2270 }, { "epoch": 0.52, "learning_rate": 4.1252599953778603e-05, "loss": 0.0, "step": 2271 }, { "epoch": 0.53, "learning_rate": 4.12487481704029e-05, "loss": 0.0, "step": 2272 }, { "epoch": 0.53, "learning_rate": 4.124489638702719e-05, "loss": 0.0, "step": 2273 }, { "epoch": 0.53, "learning_rate": 4.124104460365149e-05, "loss": 0.0, "step": 2274 }, { "epoch": 0.53, "learning_rate": 4.123719282027579e-05, "loss": 0.0, "step": 2275 }, { "epoch": 0.53, "learning_rate": 4.1233341036900085e-05, "loss": 0.0, "step": 2276 }, { "epoch": 0.53, "learning_rate": 4.122948925352439e-05, "loss": 0.0, "step": 2277 }, { "epoch": 0.53, "learning_rate": 4.122563747014868e-05, "loss": 0.0, "step": 2278 }, { "epoch": 0.53, "learning_rate": 4.1221785686772977e-05, "loss": 0.0, "step": 2279 }, { "epoch": 0.53, "learning_rate": 4.121793390339728e-05, "loss": 0.0, "step": 2280 }, { "epoch": 0.53, "learning_rate": 4.121408212002157e-05, "loss": 0.0, "step": 2281 }, { "epoch": 0.53, "learning_rate": 4.121023033664587e-05, "loss": 0.0, "step": 2282 }, { "epoch": 0.53, "learning_rate": 4.120637855327016e-05, "loss": 0.0, "step": 2283 }, { "epoch": 0.53, "learning_rate": 4.1202526769894465e-05, "loss": 0.0002, "step": 2284 }, { "epoch": 0.53, "learning_rate": 4.119867498651876e-05, "loss": 0.0, "step": 2285 }, { "epoch": 0.53, "learning_rate": 4.1194823203143055e-05, "loss": 0.0, "step": 2286 }, { "epoch": 0.53, "learning_rate": 4.1190971419767357e-05, "loss": 0.0, "step": 2287 }, { "epoch": 0.53, "learning_rate": 4.118711963639165e-05, "loss": 0.0, "step": 2288 }, { "epoch": 0.53, "learning_rate": 4.118326785301595e-05, "loss": 0.0003, "step": 2289 }, { "epoch": 0.53, "learning_rate": 4.117941606964025e-05, "loss": 0.0, "step": 2290 }, { "epoch": 0.53, "learning_rate": 4.117556428626454e-05, "loss": 0.0003, "step": 2291 }, { "epoch": 0.53, "learning_rate": 4.117171250288884e-05, "loss": 0.0, "step": 2292 }, { "epoch": 0.53, "learning_rate": 4.116786071951313e-05, "loss": 0.0, "step": 2293 }, { "epoch": 0.53, "learning_rate": 4.1164008936137435e-05, "loss": 0.0, "step": 2294 }, { "epoch": 0.53, "learning_rate": 4.116015715276173e-05, "loss": 0.0, "step": 2295 }, { "epoch": 0.53, "learning_rate": 4.1156305369386025e-05, "loss": 0.0, "step": 2296 }, { "epoch": 0.53, "learning_rate": 4.1152453586010326e-05, "loss": 0.0, "step": 2297 }, { "epoch": 0.53, "learning_rate": 4.114860180263462e-05, "loss": 0.0, "step": 2298 }, { "epoch": 0.53, "learning_rate": 4.114475001925892e-05, "loss": 0.0, "step": 2299 }, { "epoch": 0.53, "learning_rate": 4.114089823588322e-05, "loss": 0.0, "step": 2300 }, { "epoch": 0.53, "learning_rate": 4.113704645250751e-05, "loss": 0.0, "step": 2301 }, { "epoch": 0.53, "learning_rate": 4.113319466913181e-05, "loss": 0.0, "step": 2302 }, { "epoch": 0.53, "learning_rate": 4.11293428857561e-05, "loss": 0.0, "step": 2303 }, { "epoch": 0.53, "learning_rate": 4.1125491102380405e-05, "loss": 0.0, "step": 2304 }, { "epoch": 0.53, "learning_rate": 4.11216393190047e-05, "loss": 0.0, "step": 2305 }, { "epoch": 0.53, "learning_rate": 4.1117787535629e-05, "loss": 0.0001, "step": 2306 }, { "epoch": 0.53, "learning_rate": 4.1113935752253296e-05, "loss": 0.0, "step": 2307 }, { "epoch": 0.53, "learning_rate": 4.111008396887759e-05, "loss": 0.0, "step": 2308 }, { "epoch": 0.53, "learning_rate": 4.110623218550189e-05, "loss": 0.0, "step": 2309 }, { "epoch": 0.53, "learning_rate": 4.110238040212618e-05, "loss": 0.0, "step": 2310 }, { "epoch": 0.53, "learning_rate": 4.109852861875048e-05, "loss": 0.0, "step": 2311 }, { "epoch": 0.53, "learning_rate": 4.109467683537478e-05, "loss": 0.0095, "step": 2312 }, { "epoch": 0.53, "learning_rate": 4.109082505199908e-05, "loss": 0.0, "step": 2313 }, { "epoch": 0.53, "learning_rate": 4.1086973268623375e-05, "loss": 0.0, "step": 2314 }, { "epoch": 0.54, "learning_rate": 4.108312148524767e-05, "loss": 0.0001, "step": 2315 }, { "epoch": 0.54, "learning_rate": 4.107926970187197e-05, "loss": 0.0, "step": 2316 }, { "epoch": 0.54, "learning_rate": 4.1075417918496266e-05, "loss": 0.0, "step": 2317 }, { "epoch": 0.54, "learning_rate": 4.107156613512057e-05, "loss": 0.0, "step": 2318 }, { "epoch": 0.54, "learning_rate": 4.106771435174486e-05, "loss": 0.0, "step": 2319 }, { "epoch": 0.54, "learning_rate": 4.106386256836916e-05, "loss": 0.0, "step": 2320 }, { "epoch": 0.54, "learning_rate": 4.106001078499345e-05, "loss": 0.0, "step": 2321 }, { "epoch": 0.54, "learning_rate": 4.105615900161775e-05, "loss": 0.0, "step": 2322 }, { "epoch": 0.54, "learning_rate": 4.105230721824205e-05, "loss": 0.0, "step": 2323 }, { "epoch": 0.54, "learning_rate": 4.1048455434866344e-05, "loss": 0.0, "step": 2324 }, { "epoch": 0.54, "learning_rate": 4.1044603651490646e-05, "loss": 0.0002, "step": 2325 }, { "epoch": 0.54, "learning_rate": 4.104075186811494e-05, "loss": 0.0, "step": 2326 }, { "epoch": 0.54, "learning_rate": 4.1036900084739236e-05, "loss": 0.0001, "step": 2327 }, { "epoch": 0.54, "learning_rate": 4.103304830136354e-05, "loss": 0.0, "step": 2328 }, { "epoch": 0.54, "learning_rate": 4.1029196517987826e-05, "loss": 0.0, "step": 2329 }, { "epoch": 0.54, "learning_rate": 4.102534473461213e-05, "loss": 0.0, "step": 2330 }, { "epoch": 0.54, "learning_rate": 4.102149295123642e-05, "loss": 0.0003, "step": 2331 }, { "epoch": 0.54, "learning_rate": 4.101764116786072e-05, "loss": 0.0, "step": 2332 }, { "epoch": 0.54, "learning_rate": 4.101378938448502e-05, "loss": 0.0002, "step": 2333 }, { "epoch": 0.54, "learning_rate": 4.1009937601109314e-05, "loss": 0.0, "step": 2334 }, { "epoch": 0.54, "learning_rate": 4.1006085817733616e-05, "loss": 0.0001, "step": 2335 }, { "epoch": 0.54, "learning_rate": 4.100223403435791e-05, "loss": 0.0, "step": 2336 }, { "epoch": 0.54, "learning_rate": 4.0998382250982206e-05, "loss": 0.0, "step": 2337 }, { "epoch": 0.54, "learning_rate": 4.09945304676065e-05, "loss": 0.0, "step": 2338 }, { "epoch": 0.54, "learning_rate": 4.0990678684230796e-05, "loss": 0.0024, "step": 2339 }, { "epoch": 0.54, "learning_rate": 4.09868269008551e-05, "loss": 0.0, "step": 2340 }, { "epoch": 0.54, "learning_rate": 4.098297511747939e-05, "loss": 0.0, "step": 2341 }, { "epoch": 0.54, "learning_rate": 4.0979123334103694e-05, "loss": 0.0008, "step": 2342 }, { "epoch": 0.54, "learning_rate": 4.097527155072799e-05, "loss": 0.0, "step": 2343 }, { "epoch": 0.54, "learning_rate": 4.0971419767352284e-05, "loss": 0.0, "step": 2344 }, { "epoch": 0.54, "learning_rate": 4.0967567983976586e-05, "loss": 0.0, "step": 2345 }, { "epoch": 0.54, "learning_rate": 4.096371620060088e-05, "loss": 0.0, "step": 2346 }, { "epoch": 0.54, "learning_rate": 4.095986441722518e-05, "loss": 0.0001, "step": 2347 }, { "epoch": 0.54, "learning_rate": 4.095601263384947e-05, "loss": 0.0, "step": 2348 }, { "epoch": 0.54, "learning_rate": 4.095216085047377e-05, "loss": 0.0, "step": 2349 }, { "epoch": 0.54, "learning_rate": 4.094830906709807e-05, "loss": 0.0, "step": 2350 }, { "epoch": 0.54, "learning_rate": 4.094445728372236e-05, "loss": 0.0, "step": 2351 }, { "epoch": 0.54, "learning_rate": 4.0940605500346664e-05, "loss": 0.0, "step": 2352 }, { "epoch": 0.54, "learning_rate": 4.093675371697096e-05, "loss": 0.0, "step": 2353 }, { "epoch": 0.54, "learning_rate": 4.093290193359526e-05, "loss": 0.0, "step": 2354 }, { "epoch": 0.54, "learning_rate": 4.0929050150219556e-05, "loss": 0.0, "step": 2355 }, { "epoch": 0.54, "learning_rate": 4.092519836684385e-05, "loss": 0.0, "step": 2356 }, { "epoch": 0.54, "learning_rate": 4.0921346583468146e-05, "loss": 0.0, "step": 2357 }, { "epoch": 0.54, "learning_rate": 4.091749480009244e-05, "loss": 0.0, "step": 2358 }, { "epoch": 0.55, "learning_rate": 4.091364301671674e-05, "loss": 0.0, "step": 2359 }, { "epoch": 0.55, "learning_rate": 4.090979123334104e-05, "loss": 0.0, "step": 2360 }, { "epoch": 0.55, "learning_rate": 4.090593944996534e-05, "loss": 0.0, "step": 2361 }, { "epoch": 0.55, "learning_rate": 4.0902087666589634e-05, "loss": 0.0, "step": 2362 }, { "epoch": 0.55, "learning_rate": 4.089823588321393e-05, "loss": 0.0005, "step": 2363 }, { "epoch": 0.55, "learning_rate": 4.089438409983823e-05, "loss": 0.0, "step": 2364 }, { "epoch": 0.55, "learning_rate": 4.0890532316462526e-05, "loss": 0.0, "step": 2365 }, { "epoch": 0.55, "learning_rate": 4.088668053308682e-05, "loss": 0.0, "step": 2366 }, { "epoch": 0.55, "learning_rate": 4.0882828749711116e-05, "loss": 0.0, "step": 2367 }, { "epoch": 0.55, "learning_rate": 4.087897696633541e-05, "loss": 0.0, "step": 2368 }, { "epoch": 0.55, "learning_rate": 4.087512518295971e-05, "loss": 0.0, "step": 2369 }, { "epoch": 0.55, "learning_rate": 4.087127339958401e-05, "loss": 0.0, "step": 2370 }, { "epoch": 0.55, "learning_rate": 4.086742161620831e-05, "loss": 0.0, "step": 2371 }, { "epoch": 0.55, "learning_rate": 4.0863569832832604e-05, "loss": 0.0, "step": 2372 }, { "epoch": 0.55, "learning_rate": 4.08597180494569e-05, "loss": 0.0, "step": 2373 }, { "epoch": 0.55, "learning_rate": 4.08558662660812e-05, "loss": 0.0003, "step": 2374 }, { "epoch": 0.55, "learning_rate": 4.0852014482705496e-05, "loss": 0.0001, "step": 2375 }, { "epoch": 0.55, "learning_rate": 4.084816269932979e-05, "loss": 0.0, "step": 2376 }, { "epoch": 0.55, "learning_rate": 4.0844310915954086e-05, "loss": 0.0, "step": 2377 }, { "epoch": 0.55, "learning_rate": 4.084045913257839e-05, "loss": 0.0, "step": 2378 }, { "epoch": 0.55, "learning_rate": 4.083660734920268e-05, "loss": 0.0, "step": 2379 }, { "epoch": 0.55, "learning_rate": 4.083275556582698e-05, "loss": 0.0, "step": 2380 }, { "epoch": 0.55, "learning_rate": 4.082890378245128e-05, "loss": 0.0, "step": 2381 }, { "epoch": 0.55, "learning_rate": 4.0825051999075574e-05, "loss": 0.0001, "step": 2382 }, { "epoch": 0.55, "learning_rate": 4.0821200215699876e-05, "loss": 0.0, "step": 2383 }, { "epoch": 0.55, "learning_rate": 4.081734843232417e-05, "loss": 0.0, "step": 2384 }, { "epoch": 0.55, "learning_rate": 4.0813496648948465e-05, "loss": 0.0, "step": 2385 }, { "epoch": 0.55, "learning_rate": 4.080964486557276e-05, "loss": 0.0, "step": 2386 }, { "epoch": 0.55, "learning_rate": 4.0805793082197055e-05, "loss": 0.0, "step": 2387 }, { "epoch": 0.55, "learning_rate": 4.080194129882136e-05, "loss": 0.0, "step": 2388 }, { "epoch": 0.55, "learning_rate": 4.079808951544565e-05, "loss": 0.0, "step": 2389 }, { "epoch": 0.55, "learning_rate": 4.0794237732069954e-05, "loss": 0.0, "step": 2390 }, { "epoch": 0.55, "learning_rate": 4.079038594869425e-05, "loss": 0.0, "step": 2391 }, { "epoch": 0.55, "learning_rate": 4.0786534165318544e-05, "loss": 0.0, "step": 2392 }, { "epoch": 0.55, "learning_rate": 4.0782682381942845e-05, "loss": 0.0, "step": 2393 }, { "epoch": 0.55, "learning_rate": 4.0778830598567134e-05, "loss": 0.0, "step": 2394 }, { "epoch": 0.55, "learning_rate": 4.0774978815191435e-05, "loss": 0.0, "step": 2395 }, { "epoch": 0.55, "learning_rate": 4.077112703181573e-05, "loss": 0.0, "step": 2396 }, { "epoch": 0.55, "learning_rate": 4.076727524844003e-05, "loss": 0.0, "step": 2397 }, { "epoch": 0.55, "learning_rate": 4.076342346506433e-05, "loss": 0.0, "step": 2398 }, { "epoch": 0.55, "learning_rate": 4.075957168168862e-05, "loss": 0.0001, "step": 2399 }, { "epoch": 0.55, "learning_rate": 4.0755719898312924e-05, "loss": 0.0001, "step": 2400 }, { "epoch": 0.55, "learning_rate": 4.075186811493722e-05, "loss": 0.0, "step": 2401 }, { "epoch": 0.56, "learning_rate": 4.074801633156152e-05, "loss": 0.0, "step": 2402 }, { "epoch": 0.56, "learning_rate": 4.0744164548185815e-05, "loss": 0.0, "step": 2403 }, { "epoch": 0.56, "learning_rate": 4.0740312764810104e-05, "loss": 0.0, "step": 2404 }, { "epoch": 0.56, "learning_rate": 4.0736460981434405e-05, "loss": 0.0, "step": 2405 }, { "epoch": 0.56, "learning_rate": 4.07326091980587e-05, "loss": 0.0, "step": 2406 }, { "epoch": 0.56, "learning_rate": 4.0728757414683e-05, "loss": 0.0, "step": 2407 }, { "epoch": 0.56, "learning_rate": 4.07249056313073e-05, "loss": 0.0, "step": 2408 }, { "epoch": 0.56, "learning_rate": 4.072105384793159e-05, "loss": 0.0, "step": 2409 }, { "epoch": 0.56, "learning_rate": 4.0717202064555894e-05, "loss": 0.0, "step": 2410 }, { "epoch": 0.56, "learning_rate": 4.071335028118019e-05, "loss": 0.0, "step": 2411 }, { "epoch": 0.56, "learning_rate": 4.070949849780449e-05, "loss": 0.0, "step": 2412 }, { "epoch": 0.56, "learning_rate": 4.070564671442878e-05, "loss": 0.0, "step": 2413 }, { "epoch": 0.56, "learning_rate": 4.070179493105308e-05, "loss": 0.0, "step": 2414 }, { "epoch": 0.56, "learning_rate": 4.0697943147677375e-05, "loss": 0.0, "step": 2415 }, { "epoch": 0.56, "learning_rate": 4.069409136430167e-05, "loss": 0.0, "step": 2416 }, { "epoch": 0.56, "learning_rate": 4.069023958092597e-05, "loss": 0.0, "step": 2417 }, { "epoch": 0.56, "learning_rate": 4.068638779755027e-05, "loss": 0.0, "step": 2418 }, { "epoch": 0.56, "learning_rate": 4.068253601417457e-05, "loss": 0.0, "step": 2419 }, { "epoch": 0.56, "learning_rate": 4.0678684230798863e-05, "loss": 0.0, "step": 2420 }, { "epoch": 0.56, "learning_rate": 4.067483244742316e-05, "loss": 0.0002, "step": 2421 }, { "epoch": 0.56, "learning_rate": 4.067098066404745e-05, "loss": 0.0, "step": 2422 }, { "epoch": 0.56, "learning_rate": 4.066712888067175e-05, "loss": 0.0, "step": 2423 }, { "epoch": 0.56, "learning_rate": 4.066327709729605e-05, "loss": 0.0, "step": 2424 }, { "epoch": 0.56, "learning_rate": 4.0659425313920345e-05, "loss": 0.0, "step": 2425 }, { "epoch": 0.56, "learning_rate": 4.065557353054465e-05, "loss": 0.0, "step": 2426 }, { "epoch": 0.56, "learning_rate": 4.065172174716894e-05, "loss": 0.0, "step": 2427 }, { "epoch": 0.56, "learning_rate": 4.064786996379324e-05, "loss": 0.0, "step": 2428 }, { "epoch": 0.56, "learning_rate": 4.064401818041754e-05, "loss": 0.0, "step": 2429 }, { "epoch": 0.56, "learning_rate": 4.064016639704183e-05, "loss": 0.0, "step": 2430 }, { "epoch": 0.56, "learning_rate": 4.0636314613666135e-05, "loss": 0.0, "step": 2431 }, { "epoch": 0.56, "learning_rate": 4.063246283029042e-05, "loss": 0.0, "step": 2432 }, { "epoch": 0.56, "learning_rate": 4.0628611046914725e-05, "loss": 0.0, "step": 2433 }, { "epoch": 0.56, "learning_rate": 4.062475926353902e-05, "loss": 0.0, "step": 2434 }, { "epoch": 0.56, "learning_rate": 4.0620907480163315e-05, "loss": 0.0, "step": 2435 }, { "epoch": 0.56, "learning_rate": 4.061705569678762e-05, "loss": 0.0, "step": 2436 }, { "epoch": 0.56, "learning_rate": 4.061320391341191e-05, "loss": 0.0, "step": 2437 }, { "epoch": 0.56, "learning_rate": 4.060935213003621e-05, "loss": 0.0, "step": 2438 }, { "epoch": 0.56, "learning_rate": 4.060550034666051e-05, "loss": 0.0, "step": 2439 }, { "epoch": 0.56, "learning_rate": 4.06016485632848e-05, "loss": 0.0, "step": 2440 }, { "epoch": 0.56, "learning_rate": 4.05977967799091e-05, "loss": 0.0, "step": 2441 }, { "epoch": 0.56, "learning_rate": 4.059394499653339e-05, "loss": 0.0, "step": 2442 }, { "epoch": 0.56, "learning_rate": 4.0590093213157695e-05, "loss": 0.0, "step": 2443 }, { "epoch": 0.56, "learning_rate": 4.058624142978199e-05, "loss": 0.0, "step": 2444 }, { "epoch": 0.57, "learning_rate": 4.0582389646406285e-05, "loss": 0.0, "step": 2445 }, { "epoch": 0.57, "learning_rate": 4.0578537863030587e-05, "loss": 0.0, "step": 2446 }, { "epoch": 0.57, "learning_rate": 4.057468607965488e-05, "loss": 0.0, "step": 2447 }, { "epoch": 0.57, "learning_rate": 4.057083429627918e-05, "loss": 0.0, "step": 2448 }, { "epoch": 0.57, "learning_rate": 4.056698251290348e-05, "loss": 0.0006, "step": 2449 }, { "epoch": 0.57, "learning_rate": 4.056313072952777e-05, "loss": 0.0, "step": 2450 }, { "epoch": 0.57, "learning_rate": 4.055927894615207e-05, "loss": 0.0, "step": 2451 }, { "epoch": 0.57, "learning_rate": 4.055542716277636e-05, "loss": 0.0, "step": 2452 }, { "epoch": 0.57, "learning_rate": 4.0551575379400665e-05, "loss": 0.0, "step": 2453 }, { "epoch": 0.57, "learning_rate": 4.054772359602496e-05, "loss": 0.0, "step": 2454 }, { "epoch": 0.57, "learning_rate": 4.054387181264926e-05, "loss": 0.0, "step": 2455 }, { "epoch": 0.57, "learning_rate": 4.0540020029273556e-05, "loss": 0.0, "step": 2456 }, { "epoch": 0.57, "learning_rate": 4.053616824589785e-05, "loss": 0.0, "step": 2457 }, { "epoch": 0.57, "learning_rate": 4.053231646252215e-05, "loss": 0.0, "step": 2458 }, { "epoch": 0.57, "learning_rate": 4.052846467914645e-05, "loss": 0.0, "step": 2459 }, { "epoch": 0.57, "learning_rate": 4.052461289577074e-05, "loss": 0.0001, "step": 2460 }, { "epoch": 0.57, "learning_rate": 4.052076111239504e-05, "loss": 0.0, "step": 2461 }, { "epoch": 0.57, "learning_rate": 4.051690932901934e-05, "loss": 0.0, "step": 2462 }, { "epoch": 0.57, "learning_rate": 4.0513057545643635e-05, "loss": 0.0, "step": 2463 }, { "epoch": 0.57, "learning_rate": 4.050920576226793e-05, "loss": 0.0, "step": 2464 }, { "epoch": 0.57, "learning_rate": 4.050535397889223e-05, "loss": 0.0, "step": 2465 }, { "epoch": 0.57, "learning_rate": 4.0501502195516526e-05, "loss": 0.0, "step": 2466 }, { "epoch": 0.57, "learning_rate": 4.049765041214083e-05, "loss": 0.0, "step": 2467 }, { "epoch": 0.57, "learning_rate": 4.049379862876512e-05, "loss": 0.0098, "step": 2468 }, { "epoch": 0.57, "learning_rate": 4.048994684538942e-05, "loss": 0.0, "step": 2469 }, { "epoch": 0.57, "learning_rate": 4.048609506201371e-05, "loss": 0.0, "step": 2470 }, { "epoch": 0.57, "learning_rate": 4.048224327863801e-05, "loss": 0.0, "step": 2471 }, { "epoch": 0.57, "learning_rate": 4.047839149526231e-05, "loss": 0.0, "step": 2472 }, { "epoch": 0.57, "learning_rate": 4.0474539711886605e-05, "loss": 0.0, "step": 2473 }, { "epoch": 0.57, "learning_rate": 4.0470687928510906e-05, "loss": 0.0, "step": 2474 }, { "epoch": 0.57, "learning_rate": 4.04668361451352e-05, "loss": 0.0, "step": 2475 }, { "epoch": 0.57, "learning_rate": 4.0462984361759496e-05, "loss": 0.0, "step": 2476 }, { "epoch": 0.57, "learning_rate": 4.04591325783838e-05, "loss": 0.0, "step": 2477 }, { "epoch": 0.57, "learning_rate": 4.0455280795008086e-05, "loss": 0.0, "step": 2478 }, { "epoch": 0.57, "learning_rate": 4.045142901163239e-05, "loss": 0.0, "step": 2479 }, { "epoch": 0.57, "learning_rate": 4.044757722825668e-05, "loss": 0.0, "step": 2480 }, { "epoch": 0.57, "learning_rate": 4.044372544488098e-05, "loss": 0.0, "step": 2481 }, { "epoch": 0.57, "learning_rate": 4.043987366150528e-05, "loss": 0.0002, "step": 2482 }, { "epoch": 0.57, "learning_rate": 4.0436021878129574e-05, "loss": 0.0, "step": 2483 }, { "epoch": 0.57, "learning_rate": 4.0432170094753876e-05, "loss": 0.0, "step": 2484 }, { "epoch": 0.57, "learning_rate": 4.042831831137817e-05, "loss": 0.0, "step": 2485 }, { "epoch": 0.57, "learning_rate": 4.0424466528002466e-05, "loss": 0.0, "step": 2486 }, { "epoch": 0.57, "learning_rate": 4.042061474462677e-05, "loss": 0.0, "step": 2487 }, { "epoch": 0.57, "learning_rate": 4.0416762961251056e-05, "loss": 0.0001, "step": 2488 }, { "epoch": 0.58, "learning_rate": 4.041291117787536e-05, "loss": 0.0, "step": 2489 }, { "epoch": 0.58, "learning_rate": 4.040905939449965e-05, "loss": 0.0, "step": 2490 }, { "epoch": 0.58, "learning_rate": 4.0405207611123954e-05, "loss": 0.0, "step": 2491 }, { "epoch": 0.58, "learning_rate": 4.040135582774825e-05, "loss": 0.0001, "step": 2492 }, { "epoch": 0.58, "learning_rate": 4.0397504044372544e-05, "loss": 0.0, "step": 2493 }, { "epoch": 0.58, "learning_rate": 4.0393652260996846e-05, "loss": 0.0007, "step": 2494 }, { "epoch": 0.58, "learning_rate": 4.038980047762114e-05, "loss": 0.0003, "step": 2495 }, { "epoch": 0.58, "learning_rate": 4.038594869424544e-05, "loss": 0.0, "step": 2496 }, { "epoch": 0.58, "learning_rate": 4.038209691086973e-05, "loss": 0.0001, "step": 2497 }, { "epoch": 0.58, "learning_rate": 4.037824512749403e-05, "loss": 0.0, "step": 2498 }, { "epoch": 0.58, "learning_rate": 4.037439334411833e-05, "loss": 0.0, "step": 2499 }, { "epoch": 0.58, "learning_rate": 4.037054156074262e-05, "loss": 0.0, "step": 2500 }, { "epoch": 0.58, "learning_rate": 4.0366689777366924e-05, "loss": 0.0, "step": 2501 }, { "epoch": 0.58, "learning_rate": 4.036283799399122e-05, "loss": 0.0, "step": 2502 }, { "epoch": 0.58, "learning_rate": 4.035898621061552e-05, "loss": 0.0, "step": 2503 }, { "epoch": 0.58, "learning_rate": 4.0355134427239816e-05, "loss": 0.0002, "step": 2504 }, { "epoch": 0.58, "learning_rate": 4.035128264386411e-05, "loss": 0.0, "step": 2505 }, { "epoch": 0.58, "learning_rate": 4.0347430860488406e-05, "loss": 0.0, "step": 2506 }, { "epoch": 0.58, "learning_rate": 4.03435790771127e-05, "loss": 0.0, "step": 2507 }, { "epoch": 0.58, "learning_rate": 4.0339727293737e-05, "loss": 0.0, "step": 2508 }, { "epoch": 0.58, "learning_rate": 4.03358755103613e-05, "loss": 0.0, "step": 2509 }, { "epoch": 0.58, "learning_rate": 4.03320237269856e-05, "loss": 0.0, "step": 2510 }, { "epoch": 0.58, "learning_rate": 4.0328171943609894e-05, "loss": 0.0, "step": 2511 }, { "epoch": 0.58, "learning_rate": 4.032432016023419e-05, "loss": 0.0, "step": 2512 }, { "epoch": 0.58, "learning_rate": 4.032046837685849e-05, "loss": 0.0, "step": 2513 }, { "epoch": 0.58, "learning_rate": 4.0316616593482786e-05, "loss": 0.0, "step": 2514 }, { "epoch": 0.58, "learning_rate": 4.031276481010709e-05, "loss": 0.0, "step": 2515 }, { "epoch": 0.58, "learning_rate": 4.0308913026731376e-05, "loss": 0.0, "step": 2516 }, { "epoch": 0.58, "learning_rate": 4.030506124335567e-05, "loss": 0.0, "step": 2517 }, { "epoch": 0.58, "learning_rate": 4.030120945997997e-05, "loss": 0.0, "step": 2518 }, { "epoch": 0.58, "learning_rate": 4.029735767660427e-05, "loss": 0.0001, "step": 2519 }, { "epoch": 0.58, "learning_rate": 4.029350589322857e-05, "loss": 0.0, "step": 2520 }, { "epoch": 0.58, "learning_rate": 4.0289654109852864e-05, "loss": 0.0, "step": 2521 }, { "epoch": 0.58, "learning_rate": 4.028580232647716e-05, "loss": 0.0, "step": 2522 }, { "epoch": 0.58, "learning_rate": 4.028195054310146e-05, "loss": 0.0, "step": 2523 }, { "epoch": 0.58, "learning_rate": 4.0278098759725756e-05, "loss": 0.0, "step": 2524 }, { "epoch": 0.58, "learning_rate": 4.027424697635005e-05, "loss": 0.0, "step": 2525 }, { "epoch": 0.58, "learning_rate": 4.0270395192974346e-05, "loss": 0.0, "step": 2526 }, { "epoch": 0.58, "learning_rate": 4.026654340959865e-05, "loss": 0.0, "step": 2527 }, { "epoch": 0.58, "learning_rate": 4.026269162622294e-05, "loss": 0.0, "step": 2528 }, { "epoch": 0.58, "learning_rate": 4.025883984284724e-05, "loss": 0.0, "step": 2529 }, { "epoch": 0.58, "learning_rate": 4.025498805947154e-05, "loss": 0.0, "step": 2530 }, { "epoch": 0.58, "learning_rate": 4.0251136276095834e-05, "loss": 0.0, "step": 2531 }, { "epoch": 0.59, "learning_rate": 4.0247284492720136e-05, "loss": 0.0, "step": 2532 }, { "epoch": 0.59, "learning_rate": 4.024343270934443e-05, "loss": 0.0, "step": 2533 }, { "epoch": 0.59, "learning_rate": 4.0239580925968726e-05, "loss": 0.0, "step": 2534 }, { "epoch": 0.59, "learning_rate": 4.023572914259302e-05, "loss": 0.0, "step": 2535 }, { "epoch": 0.59, "learning_rate": 4.0231877359217315e-05, "loss": 0.0, "step": 2536 }, { "epoch": 0.59, "learning_rate": 4.022802557584162e-05, "loss": 0.0, "step": 2537 }, { "epoch": 0.59, "learning_rate": 4.022417379246591e-05, "loss": 0.0, "step": 2538 }, { "epoch": 0.59, "learning_rate": 4.0220322009090214e-05, "loss": 0.0, "step": 2539 }, { "epoch": 0.59, "learning_rate": 4.021647022571451e-05, "loss": 0.0, "step": 2540 }, { "epoch": 0.59, "learning_rate": 4.0212618442338804e-05, "loss": 0.0, "step": 2541 }, { "epoch": 0.59, "learning_rate": 4.0208766658963106e-05, "loss": 0.0, "step": 2542 }, { "epoch": 0.59, "learning_rate": 4.02049148755874e-05, "loss": 0.0, "step": 2543 }, { "epoch": 0.59, "learning_rate": 4.0201063092211695e-05, "loss": 0.0, "step": 2544 }, { "epoch": 0.59, "learning_rate": 4.019721130883599e-05, "loss": 0.0, "step": 2545 }, { "epoch": 0.59, "learning_rate": 4.019335952546029e-05, "loss": 0.0, "step": 2546 }, { "epoch": 0.59, "learning_rate": 4.018950774208459e-05, "loss": 0.0001, "step": 2547 }, { "epoch": 0.59, "learning_rate": 4.018565595870888e-05, "loss": 0.0, "step": 2548 }, { "epoch": 0.59, "learning_rate": 4.0181804175333184e-05, "loss": 0.0, "step": 2549 }, { "epoch": 0.59, "learning_rate": 4.017795239195748e-05, "loss": 0.0003, "step": 2550 }, { "epoch": 0.59, "learning_rate": 4.017410060858178e-05, "loss": 0.0, "step": 2551 }, { "epoch": 0.59, "learning_rate": 4.0170248825206075e-05, "loss": 0.0001, "step": 2552 }, { "epoch": 0.59, "learning_rate": 4.016639704183037e-05, "loss": 0.0, "step": 2553 }, { "epoch": 0.59, "learning_rate": 4.0162545258454665e-05, "loss": 0.0, "step": 2554 }, { "epoch": 0.59, "learning_rate": 4.015869347507896e-05, "loss": 0.0, "step": 2555 }, { "epoch": 0.59, "learning_rate": 4.015484169170326e-05, "loss": 0.0, "step": 2556 }, { "epoch": 0.59, "learning_rate": 4.015098990832756e-05, "loss": 0.0, "step": 2557 }, { "epoch": 0.59, "learning_rate": 4.014713812495185e-05, "loss": 0.0, "step": 2558 }, { "epoch": 0.59, "learning_rate": 4.0143286341576154e-05, "loss": 0.0, "step": 2559 }, { "epoch": 0.59, "learning_rate": 4.013943455820045e-05, "loss": 0.0, "step": 2560 }, { "epoch": 0.59, "learning_rate": 4.013558277482475e-05, "loss": 0.0, "step": 2561 }, { "epoch": 0.59, "learning_rate": 4.013173099144904e-05, "loss": 0.0, "step": 2562 }, { "epoch": 0.59, "learning_rate": 4.012787920807334e-05, "loss": 0.0, "step": 2563 }, { "epoch": 0.59, "learning_rate": 4.0124027424697635e-05, "loss": 0.0, "step": 2564 }, { "epoch": 0.59, "learning_rate": 4.012017564132193e-05, "loss": 0.0, "step": 2565 }, { "epoch": 0.59, "learning_rate": 4.011632385794623e-05, "loss": 0.0, "step": 2566 }, { "epoch": 0.59, "learning_rate": 4.011247207457053e-05, "loss": 0.0, "step": 2567 }, { "epoch": 0.59, "learning_rate": 4.010862029119483e-05, "loss": 0.0, "step": 2568 }, { "epoch": 0.59, "learning_rate": 4.0104768507819124e-05, "loss": 0.0, "step": 2569 }, { "epoch": 0.59, "learning_rate": 4.010091672444342e-05, "loss": 0.0, "step": 2570 }, { "epoch": 0.59, "learning_rate": 4.009706494106772e-05, "loss": 0.0, "step": 2571 }, { "epoch": 0.59, "learning_rate": 4.009321315769201e-05, "loss": 0.0, "step": 2572 }, { "epoch": 0.59, "learning_rate": 4.008936137431631e-05, "loss": 0.0, "step": 2573 }, { "epoch": 0.59, "learning_rate": 4.0085509590940605e-05, "loss": 0.0, "step": 2574 }, { "epoch": 0.6, "learning_rate": 4.008165780756491e-05, "loss": 0.0, "step": 2575 }, { "epoch": 0.6, "learning_rate": 4.00778060241892e-05, "loss": 0.0, "step": 2576 }, { "epoch": 0.6, "learning_rate": 4.00739542408135e-05, "loss": 0.0, "step": 2577 }, { "epoch": 0.6, "learning_rate": 4.00701024574378e-05, "loss": 0.0, "step": 2578 }, { "epoch": 0.6, "learning_rate": 4.0066250674062093e-05, "loss": 0.0, "step": 2579 }, { "epoch": 0.6, "learning_rate": 4.0062398890686395e-05, "loss": 0.0, "step": 2580 }, { "epoch": 0.6, "learning_rate": 4.005854710731068e-05, "loss": 0.0, "step": 2581 }, { "epoch": 0.6, "learning_rate": 4.0054695323934985e-05, "loss": 0.0, "step": 2582 }, { "epoch": 0.6, "learning_rate": 4.005084354055928e-05, "loss": 0.0, "step": 2583 }, { "epoch": 0.6, "learning_rate": 4.0046991757183575e-05, "loss": 0.0076, "step": 2584 }, { "epoch": 0.6, "learning_rate": 4.004313997380788e-05, "loss": 0.0, "step": 2585 }, { "epoch": 0.6, "learning_rate": 4.003928819043217e-05, "loss": 0.0, "step": 2586 }, { "epoch": 0.6, "learning_rate": 4.0035436407056473e-05, "loss": 0.0, "step": 2587 }, { "epoch": 0.6, "learning_rate": 4.003158462368077e-05, "loss": 0.0, "step": 2588 }, { "epoch": 0.6, "learning_rate": 4.002773284030506e-05, "loss": 0.0, "step": 2589 }, { "epoch": 0.6, "learning_rate": 4.002388105692936e-05, "loss": 0.0, "step": 2590 }, { "epoch": 0.6, "learning_rate": 4.002002927355365e-05, "loss": 0.0, "step": 2591 }, { "epoch": 0.6, "learning_rate": 4.0016177490177955e-05, "loss": 0.0, "step": 2592 }, { "epoch": 0.6, "learning_rate": 4.001232570680225e-05, "loss": 0.0, "step": 2593 }, { "epoch": 0.6, "learning_rate": 4.0008473923426545e-05, "loss": 0.0, "step": 2594 }, { "epoch": 0.6, "learning_rate": 4.000462214005085e-05, "loss": 0.0, "step": 2595 }, { "epoch": 0.6, "learning_rate": 4.000077035667514e-05, "loss": 0.0, "step": 2596 }, { "epoch": 0.6, "learning_rate": 3.999691857329944e-05, "loss": 0.0, "step": 2597 }, { "epoch": 0.6, "learning_rate": 3.999306678992374e-05, "loss": 0.0, "step": 2598 }, { "epoch": 0.6, "learning_rate": 3.998921500654803e-05, "loss": 0.0, "step": 2599 }, { "epoch": 0.6, "learning_rate": 3.998536322317233e-05, "loss": 0.0, "step": 2600 }, { "epoch": 0.6, "learning_rate": 3.998151143979662e-05, "loss": 0.0001, "step": 2601 }, { "epoch": 0.6, "learning_rate": 3.9977659656420925e-05, "loss": 0.0, "step": 2602 }, { "epoch": 0.6, "learning_rate": 3.997380787304522e-05, "loss": 0.0, "step": 2603 }, { "epoch": 0.6, "learning_rate": 3.996995608966952e-05, "loss": 0.0, "step": 2604 }, { "epoch": 0.6, "learning_rate": 3.9966104306293816e-05, "loss": 0.0, "step": 2605 }, { "epoch": 0.6, "learning_rate": 3.996225252291811e-05, "loss": 0.0, "step": 2606 }, { "epoch": 0.6, "learning_rate": 3.995840073954241e-05, "loss": 0.0001, "step": 2607 }, { "epoch": 0.6, "learning_rate": 3.995454895616671e-05, "loss": 0.0, "step": 2608 }, { "epoch": 0.6, "learning_rate": 3.9950697172791e-05, "loss": 0.0001, "step": 2609 }, { "epoch": 0.6, "learning_rate": 3.99468453894153e-05, "loss": 0.0, "step": 2610 }, { "epoch": 0.6, "learning_rate": 3.99429936060396e-05, "loss": 0.0001, "step": 2611 }, { "epoch": 0.6, "learning_rate": 3.9939141822663895e-05, "loss": 0.0, "step": 2612 }, { "epoch": 0.6, "learning_rate": 3.993529003928819e-05, "loss": 0.0, "step": 2613 }, { "epoch": 0.6, "learning_rate": 3.993143825591249e-05, "loss": 0.0, "step": 2614 }, { "epoch": 0.6, "learning_rate": 3.9927586472536786e-05, "loss": 0.0, "step": 2615 }, { "epoch": 0.6, "learning_rate": 3.992373468916109e-05, "loss": 0.0001, "step": 2616 }, { "epoch": 0.6, "learning_rate": 3.991988290578538e-05, "loss": 0.0, "step": 2617 }, { "epoch": 0.61, "learning_rate": 3.991603112240968e-05, "loss": 0.0, "step": 2618 }, { "epoch": 0.61, "learning_rate": 3.991217933903397e-05, "loss": 0.0, "step": 2619 }, { "epoch": 0.61, "learning_rate": 3.990832755565827e-05, "loss": 0.0, "step": 2620 }, { "epoch": 0.61, "learning_rate": 3.990447577228257e-05, "loss": 0.0, "step": 2621 }, { "epoch": 0.61, "learning_rate": 3.9900623988906865e-05, "loss": 0.0, "step": 2622 }, { "epoch": 0.61, "learning_rate": 3.9896772205531166e-05, "loss": 0.0, "step": 2623 }, { "epoch": 0.61, "learning_rate": 3.989292042215546e-05, "loss": 0.0, "step": 2624 }, { "epoch": 0.61, "learning_rate": 3.9889068638779756e-05, "loss": 0.0, "step": 2625 }, { "epoch": 0.61, "learning_rate": 3.988521685540406e-05, "loss": 0.0, "step": 2626 }, { "epoch": 0.61, "learning_rate": 3.9881365072028346e-05, "loss": 0.0, "step": 2627 }, { "epoch": 0.61, "learning_rate": 3.987751328865265e-05, "loss": 0.0, "step": 2628 }, { "epoch": 0.61, "learning_rate": 3.987366150527694e-05, "loss": 0.0, "step": 2629 }, { "epoch": 0.61, "learning_rate": 3.986980972190124e-05, "loss": 0.0001, "step": 2630 }, { "epoch": 0.61, "learning_rate": 3.986595793852554e-05, "loss": 0.0, "step": 2631 }, { "epoch": 0.61, "learning_rate": 3.9862106155149835e-05, "loss": 0.0, "step": 2632 }, { "epoch": 0.61, "learning_rate": 3.9858254371774136e-05, "loss": 0.0, "step": 2633 }, { "epoch": 0.61, "learning_rate": 3.985440258839843e-05, "loss": 0.0, "step": 2634 }, { "epoch": 0.61, "learning_rate": 3.9850550805022726e-05, "loss": 0.0, "step": 2635 }, { "epoch": 0.61, "learning_rate": 3.984669902164703e-05, "loss": 0.0, "step": 2636 }, { "epoch": 0.61, "learning_rate": 3.9842847238271316e-05, "loss": 0.0, "step": 2637 }, { "epoch": 0.61, "learning_rate": 3.983899545489562e-05, "loss": 0.0, "step": 2638 }, { "epoch": 0.61, "learning_rate": 3.983514367151991e-05, "loss": 0.0, "step": 2639 }, { "epoch": 0.61, "learning_rate": 3.9831291888144214e-05, "loss": 0.0, "step": 2640 }, { "epoch": 0.61, "learning_rate": 3.982744010476851e-05, "loss": 0.0, "step": 2641 }, { "epoch": 0.61, "learning_rate": 3.9823588321392804e-05, "loss": 0.0, "step": 2642 }, { "epoch": 0.61, "learning_rate": 3.9819736538017106e-05, "loss": 0.0, "step": 2643 }, { "epoch": 0.61, "learning_rate": 3.98158847546414e-05, "loss": 0.0, "step": 2644 }, { "epoch": 0.61, "learning_rate": 3.98120329712657e-05, "loss": 0.0, "step": 2645 }, { "epoch": 0.61, "learning_rate": 3.980818118788999e-05, "loss": 0.0, "step": 2646 }, { "epoch": 0.61, "learning_rate": 3.980432940451429e-05, "loss": 0.0, "step": 2647 }, { "epoch": 0.61, "learning_rate": 3.980047762113859e-05, "loss": 0.0, "step": 2648 }, { "epoch": 0.61, "learning_rate": 3.979662583776288e-05, "loss": 0.0, "step": 2649 }, { "epoch": 0.61, "learning_rate": 3.9792774054387184e-05, "loss": 0.0, "step": 2650 }, { "epoch": 0.61, "learning_rate": 3.978892227101148e-05, "loss": 0.0, "step": 2651 }, { "epoch": 0.61, "learning_rate": 3.978507048763578e-05, "loss": 0.0, "step": 2652 }, { "epoch": 0.61, "learning_rate": 3.9781218704260076e-05, "loss": 0.0, "step": 2653 }, { "epoch": 0.61, "learning_rate": 3.977736692088437e-05, "loss": 0.0, "step": 2654 }, { "epoch": 0.61, "learning_rate": 3.9773515137508666e-05, "loss": 0.0, "step": 2655 }, { "epoch": 0.61, "learning_rate": 3.976966335413296e-05, "loss": 0.0, "step": 2656 }, { "epoch": 0.61, "learning_rate": 3.976581157075726e-05, "loss": 0.0, "step": 2657 }, { "epoch": 0.61, "learning_rate": 3.976195978738156e-05, "loss": 0.0, "step": 2658 }, { "epoch": 0.61, "learning_rate": 3.975810800400586e-05, "loss": 0.0, "step": 2659 }, { "epoch": 0.61, "learning_rate": 3.9754256220630154e-05, "loss": 0.0, "step": 2660 }, { "epoch": 0.61, "learning_rate": 3.975040443725445e-05, "loss": 0.0, "step": 2661 }, { "epoch": 0.62, "learning_rate": 3.974655265387875e-05, "loss": 0.0, "step": 2662 }, { "epoch": 0.62, "learning_rate": 3.9742700870503046e-05, "loss": 0.0, "step": 2663 }, { "epoch": 0.62, "learning_rate": 3.973884908712735e-05, "loss": 0.0, "step": 2664 }, { "epoch": 0.62, "learning_rate": 3.9734997303751636e-05, "loss": 0.0, "step": 2665 }, { "epoch": 0.62, "learning_rate": 3.973114552037594e-05, "loss": 0.0, "step": 2666 }, { "epoch": 0.62, "learning_rate": 3.972729373700023e-05, "loss": 0.0, "step": 2667 }, { "epoch": 0.62, "learning_rate": 3.972344195362453e-05, "loss": 0.0008, "step": 2668 }, { "epoch": 0.62, "learning_rate": 3.971959017024883e-05, "loss": 0.0, "step": 2669 }, { "epoch": 0.62, "learning_rate": 3.9715738386873124e-05, "loss": 0.0, "step": 2670 }, { "epoch": 0.62, "learning_rate": 3.971188660349742e-05, "loss": 0.0, "step": 2671 }, { "epoch": 0.62, "learning_rate": 3.970803482012172e-05, "loss": 0.0, "step": 2672 }, { "epoch": 0.62, "learning_rate": 3.9704183036746016e-05, "loss": 0.0, "step": 2673 }, { "epoch": 0.62, "learning_rate": 3.970033125337031e-05, "loss": 0.0001, "step": 2674 }, { "epoch": 0.62, "learning_rate": 3.9696479469994606e-05, "loss": 0.0, "step": 2675 }, { "epoch": 0.62, "learning_rate": 3.969262768661891e-05, "loss": 0.0, "step": 2676 }, { "epoch": 0.62, "learning_rate": 3.96887759032432e-05, "loss": 0.0, "step": 2677 }, { "epoch": 0.62, "learning_rate": 3.96849241198675e-05, "loss": 0.0, "step": 2678 }, { "epoch": 0.62, "learning_rate": 3.96810723364918e-05, "loss": 0.0, "step": 2679 }, { "epoch": 0.62, "learning_rate": 3.9677220553116094e-05, "loss": 0.0, "step": 2680 }, { "epoch": 0.62, "learning_rate": 3.9673368769740396e-05, "loss": 0.0, "step": 2681 }, { "epoch": 0.62, "learning_rate": 3.966951698636469e-05, "loss": 0.0, "step": 2682 }, { "epoch": 0.62, "learning_rate": 3.9665665202988986e-05, "loss": 0.0, "step": 2683 }, { "epoch": 0.62, "learning_rate": 3.966181341961328e-05, "loss": 0.0001, "step": 2684 }, { "epoch": 0.62, "learning_rate": 3.9657961636237576e-05, "loss": 0.0001, "step": 2685 }, { "epoch": 0.62, "learning_rate": 3.965410985286188e-05, "loss": 0.0, "step": 2686 }, { "epoch": 0.62, "learning_rate": 3.965025806948617e-05, "loss": 0.0, "step": 2687 }, { "epoch": 0.62, "learning_rate": 3.9646406286110474e-05, "loss": 0.0, "step": 2688 }, { "epoch": 0.62, "learning_rate": 3.964255450273477e-05, "loss": 0.0, "step": 2689 }, { "epoch": 0.62, "learning_rate": 3.9638702719359064e-05, "loss": 0.0, "step": 2690 }, { "epoch": 0.62, "learning_rate": 3.9634850935983366e-05, "loss": 0.0, "step": 2691 }, { "epoch": 0.62, "learning_rate": 3.963099915260766e-05, "loss": 0.0, "step": 2692 }, { "epoch": 0.62, "learning_rate": 3.9627147369231956e-05, "loss": 0.0, "step": 2693 }, { "epoch": 0.62, "learning_rate": 3.962329558585625e-05, "loss": 0.0, "step": 2694 }, { "epoch": 0.62, "learning_rate": 3.961944380248055e-05, "loss": 0.0, "step": 2695 }, { "epoch": 0.62, "learning_rate": 3.961559201910485e-05, "loss": 0.0, "step": 2696 }, { "epoch": 0.62, "learning_rate": 3.961174023572914e-05, "loss": 0.0, "step": 2697 }, { "epoch": 0.62, "learning_rate": 3.9607888452353444e-05, "loss": 0.0001, "step": 2698 }, { "epoch": 0.62, "learning_rate": 3.960403666897774e-05, "loss": 0.0, "step": 2699 }, { "epoch": 0.62, "learning_rate": 3.960018488560204e-05, "loss": 0.0001, "step": 2700 }, { "epoch": 0.62, "learning_rate": 3.9596333102226336e-05, "loss": 0.0001, "step": 2701 }, { "epoch": 0.62, "learning_rate": 3.959248131885063e-05, "loss": 0.0, "step": 2702 }, { "epoch": 0.62, "learning_rate": 3.9588629535474925e-05, "loss": 0.0, "step": 2703 }, { "epoch": 0.62, "learning_rate": 3.958477775209922e-05, "loss": 0.0, "step": 2704 }, { "epoch": 0.63, "learning_rate": 3.958092596872352e-05, "loss": 0.0, "step": 2705 }, { "epoch": 0.63, "learning_rate": 3.957707418534782e-05, "loss": 0.0001, "step": 2706 }, { "epoch": 0.63, "learning_rate": 3.957322240197211e-05, "loss": 0.0, "step": 2707 }, { "epoch": 0.63, "learning_rate": 3.9569370618596414e-05, "loss": 0.0, "step": 2708 }, { "epoch": 0.63, "learning_rate": 3.956551883522071e-05, "loss": 0.0, "step": 2709 }, { "epoch": 0.63, "learning_rate": 3.956166705184501e-05, "loss": 0.0, "step": 2710 }, { "epoch": 0.63, "learning_rate": 3.95578152684693e-05, "loss": 0.0, "step": 2711 }, { "epoch": 0.63, "learning_rate": 3.95539634850936e-05, "loss": 0.0, "step": 2712 }, { "epoch": 0.63, "learning_rate": 3.9550111701717895e-05, "loss": 0.0, "step": 2713 }, { "epoch": 0.63, "learning_rate": 3.954625991834219e-05, "loss": 0.0, "step": 2714 }, { "epoch": 0.63, "learning_rate": 3.954240813496649e-05, "loss": 0.0, "step": 2715 }, { "epoch": 0.63, "learning_rate": 3.953855635159079e-05, "loss": 0.0, "step": 2716 }, { "epoch": 0.63, "learning_rate": 3.953470456821509e-05, "loss": 0.0, "step": 2717 }, { "epoch": 0.63, "learning_rate": 3.9530852784839384e-05, "loss": 0.0, "step": 2718 }, { "epoch": 0.63, "learning_rate": 3.952700100146368e-05, "loss": 0.0001, "step": 2719 }, { "epoch": 0.63, "learning_rate": 3.952314921808798e-05, "loss": 0.0, "step": 2720 }, { "epoch": 0.63, "learning_rate": 3.951929743471227e-05, "loss": 0.0, "step": 2721 }, { "epoch": 0.63, "learning_rate": 3.951544565133657e-05, "loss": 0.0049, "step": 2722 }, { "epoch": 0.63, "learning_rate": 3.9511593867960865e-05, "loss": 0.0, "step": 2723 }, { "epoch": 0.63, "learning_rate": 3.950774208458517e-05, "loss": 0.0, "step": 2724 }, { "epoch": 0.63, "learning_rate": 3.950389030120946e-05, "loss": 0.0, "step": 2725 }, { "epoch": 0.63, "learning_rate": 3.950003851783376e-05, "loss": 0.0, "step": 2726 }, { "epoch": 0.63, "learning_rate": 3.949618673445806e-05, "loss": 0.0044, "step": 2727 }, { "epoch": 0.63, "learning_rate": 3.9492334951082354e-05, "loss": 0.0, "step": 2728 }, { "epoch": 0.63, "learning_rate": 3.9488483167706655e-05, "loss": 0.0001, "step": 2729 }, { "epoch": 0.63, "learning_rate": 3.9484631384330943e-05, "loss": 0.0, "step": 2730 }, { "epoch": 0.63, "learning_rate": 3.9480779600955245e-05, "loss": 0.0, "step": 2731 }, { "epoch": 0.63, "learning_rate": 3.947692781757954e-05, "loss": 0.0, "step": 2732 }, { "epoch": 0.63, "learning_rate": 3.9473076034203835e-05, "loss": 0.0, "step": 2733 }, { "epoch": 0.63, "learning_rate": 3.946922425082814e-05, "loss": 0.0, "step": 2734 }, { "epoch": 0.63, "learning_rate": 3.946537246745243e-05, "loss": 0.0, "step": 2735 }, { "epoch": 0.63, "learning_rate": 3.9461520684076734e-05, "loss": 0.0, "step": 2736 }, { "epoch": 0.63, "learning_rate": 3.945766890070103e-05, "loss": 0.0, "step": 2737 }, { "epoch": 0.63, "learning_rate": 3.9453817117325323e-05, "loss": 0.0, "step": 2738 }, { "epoch": 0.63, "learning_rate": 3.944996533394962e-05, "loss": 0.0, "step": 2739 }, { "epoch": 0.63, "learning_rate": 3.944611355057391e-05, "loss": 0.0, "step": 2740 }, { "epoch": 0.63, "learning_rate": 3.9442261767198215e-05, "loss": 0.0, "step": 2741 }, { "epoch": 0.63, "learning_rate": 3.943840998382251e-05, "loss": 0.0, "step": 2742 }, { "epoch": 0.63, "learning_rate": 3.943455820044681e-05, "loss": 0.0, "step": 2743 }, { "epoch": 0.63, "learning_rate": 3.943070641707111e-05, "loss": 0.0006, "step": 2744 }, { "epoch": 0.63, "learning_rate": 3.94268546336954e-05, "loss": 0.0, "step": 2745 }, { "epoch": 0.63, "learning_rate": 3.94230028503197e-05, "loss": 0.0, "step": 2746 }, { "epoch": 0.63, "learning_rate": 3.9419151066944e-05, "loss": 0.0, "step": 2747 }, { "epoch": 0.64, "learning_rate": 3.941529928356829e-05, "loss": 0.0, "step": 2748 }, { "epoch": 0.64, "learning_rate": 3.941144750019259e-05, "loss": 0.0, "step": 2749 }, { "epoch": 0.64, "learning_rate": 3.940759571681688e-05, "loss": 0.0, "step": 2750 }, { "epoch": 0.64, "learning_rate": 3.9403743933441185e-05, "loss": 0.0, "step": 2751 }, { "epoch": 0.64, "learning_rate": 3.939989215006548e-05, "loss": 0.0, "step": 2752 }, { "epoch": 0.64, "learning_rate": 3.939604036668978e-05, "loss": 0.0008, "step": 2753 }, { "epoch": 0.64, "learning_rate": 3.9392188583314077e-05, "loss": 0.0, "step": 2754 }, { "epoch": 0.64, "learning_rate": 3.938833679993837e-05, "loss": 0.0001, "step": 2755 }, { "epoch": 0.64, "learning_rate": 3.938448501656267e-05, "loss": 0.0, "step": 2756 }, { "epoch": 0.64, "learning_rate": 3.938063323318697e-05, "loss": 0.0, "step": 2757 }, { "epoch": 0.64, "learning_rate": 3.937678144981126e-05, "loss": 0.0002, "step": 2758 }, { "epoch": 0.64, "learning_rate": 3.937292966643556e-05, "loss": 0.0, "step": 2759 }, { "epoch": 0.64, "learning_rate": 3.936907788305986e-05, "loss": 0.0001, "step": 2760 }, { "epoch": 0.64, "learning_rate": 3.9365226099684155e-05, "loss": 0.0, "step": 2761 }, { "epoch": 0.64, "learning_rate": 3.936137431630845e-05, "loss": 0.0004, "step": 2762 }, { "epoch": 0.64, "learning_rate": 3.935752253293275e-05, "loss": 0.0, "step": 2763 }, { "epoch": 0.64, "learning_rate": 3.9353670749557046e-05, "loss": 0.0, "step": 2764 }, { "epoch": 0.64, "learning_rate": 3.934981896618135e-05, "loss": 0.0, "step": 2765 }, { "epoch": 0.64, "learning_rate": 3.934596718280564e-05, "loss": 0.0, "step": 2766 }, { "epoch": 0.64, "learning_rate": 3.934211539942994e-05, "loss": 0.0, "step": 2767 }, { "epoch": 0.64, "learning_rate": 3.933826361605423e-05, "loss": 0.0, "step": 2768 }, { "epoch": 0.64, "learning_rate": 3.933441183267853e-05, "loss": 0.0, "step": 2769 }, { "epoch": 0.64, "learning_rate": 3.933056004930283e-05, "loss": 0.0, "step": 2770 }, { "epoch": 0.64, "learning_rate": 3.9326708265927125e-05, "loss": 0.0, "step": 2771 }, { "epoch": 0.64, "learning_rate": 3.9322856482551426e-05, "loss": 0.0132, "step": 2772 }, { "epoch": 0.64, "learning_rate": 3.931900469917572e-05, "loss": 0.0, "step": 2773 }, { "epoch": 0.64, "learning_rate": 3.9315152915800016e-05, "loss": 0.0, "step": 2774 }, { "epoch": 0.64, "learning_rate": 3.931130113242432e-05, "loss": 0.0, "step": 2775 }, { "epoch": 0.64, "learning_rate": 3.930744934904861e-05, "loss": 0.0, "step": 2776 }, { "epoch": 0.64, "learning_rate": 3.930359756567291e-05, "loss": 0.0, "step": 2777 }, { "epoch": 0.64, "learning_rate": 3.92997457822972e-05, "loss": 0.0, "step": 2778 }, { "epoch": 0.64, "learning_rate": 3.9295893998921505e-05, "loss": 0.001, "step": 2779 }, { "epoch": 0.64, "learning_rate": 3.92920422155458e-05, "loss": 0.0, "step": 2780 }, { "epoch": 0.64, "learning_rate": 3.9288190432170095e-05, "loss": 0.0002, "step": 2781 }, { "epoch": 0.64, "learning_rate": 3.9284338648794396e-05, "loss": 0.0, "step": 2782 }, { "epoch": 0.64, "learning_rate": 3.928048686541869e-05, "loss": 0.0144, "step": 2783 }, { "epoch": 0.64, "learning_rate": 3.9276635082042986e-05, "loss": 0.0, "step": 2784 }, { "epoch": 0.64, "learning_rate": 3.927278329866729e-05, "loss": 0.0, "step": 2785 }, { "epoch": 0.64, "learning_rate": 3.9268931515291576e-05, "loss": 0.0, "step": 2786 }, { "epoch": 0.64, "learning_rate": 3.926507973191588e-05, "loss": 0.0, "step": 2787 }, { "epoch": 0.64, "learning_rate": 3.926122794854017e-05, "loss": 0.0, "step": 2788 }, { "epoch": 0.64, "learning_rate": 3.9257376165164475e-05, "loss": 0.0, "step": 2789 }, { "epoch": 0.64, "learning_rate": 3.925352438178877e-05, "loss": 0.0, "step": 2790 }, { "epoch": 0.65, "learning_rate": 3.9249672598413064e-05, "loss": 0.0, "step": 2791 }, { "epoch": 0.65, "learning_rate": 3.9245820815037366e-05, "loss": 0.0, "step": 2792 }, { "epoch": 0.65, "learning_rate": 3.924196903166166e-05, "loss": 0.0, "step": 2793 }, { "epoch": 0.65, "learning_rate": 3.923811724828596e-05, "loss": 0.0, "step": 2794 }, { "epoch": 0.65, "learning_rate": 3.923426546491025e-05, "loss": 0.0, "step": 2795 }, { "epoch": 0.65, "learning_rate": 3.923041368153455e-05, "loss": 0.0, "step": 2796 }, { "epoch": 0.65, "learning_rate": 3.922656189815885e-05, "loss": 0.0, "step": 2797 }, { "epoch": 0.65, "learning_rate": 3.922271011478314e-05, "loss": 0.0012, "step": 2798 }, { "epoch": 0.65, "learning_rate": 3.9218858331407444e-05, "loss": 0.0, "step": 2799 }, { "epoch": 0.65, "learning_rate": 3.921500654803174e-05, "loss": 0.0, "step": 2800 }, { "epoch": 0.65, "learning_rate": 3.921115476465604e-05, "loss": 0.0, "step": 2801 }, { "epoch": 0.65, "learning_rate": 3.9207302981280336e-05, "loss": 0.001, "step": 2802 }, { "epoch": 0.65, "learning_rate": 3.920345119790463e-05, "loss": 0.0, "step": 2803 }, { "epoch": 0.65, "learning_rate": 3.919959941452893e-05, "loss": 0.0, "step": 2804 }, { "epoch": 0.65, "learning_rate": 3.919574763115322e-05, "loss": 0.0, "step": 2805 }, { "epoch": 0.65, "learning_rate": 3.919189584777752e-05, "loss": 0.0, "step": 2806 }, { "epoch": 0.65, "learning_rate": 3.918804406440182e-05, "loss": 0.0, "step": 2807 }, { "epoch": 0.65, "learning_rate": 3.918419228102612e-05, "loss": 0.0, "step": 2808 }, { "epoch": 0.65, "learning_rate": 3.9180340497650414e-05, "loss": 0.0, "step": 2809 }, { "epoch": 0.65, "learning_rate": 3.917648871427471e-05, "loss": 0.0, "step": 2810 }, { "epoch": 0.65, "learning_rate": 3.917263693089901e-05, "loss": 0.0, "step": 2811 }, { "epoch": 0.65, "learning_rate": 3.9168785147523306e-05, "loss": 0.0, "step": 2812 }, { "epoch": 0.65, "learning_rate": 3.916493336414761e-05, "loss": 0.0, "step": 2813 }, { "epoch": 0.65, "learning_rate": 3.9161081580771896e-05, "loss": 0.0, "step": 2814 }, { "epoch": 0.65, "learning_rate": 3.91572297973962e-05, "loss": 0.0, "step": 2815 }, { "epoch": 0.65, "learning_rate": 3.915337801402049e-05, "loss": 0.0, "step": 2816 }, { "epoch": 0.65, "learning_rate": 3.914952623064479e-05, "loss": 0.0, "step": 2817 }, { "epoch": 0.65, "learning_rate": 3.914567444726909e-05, "loss": 0.0, "step": 2818 }, { "epoch": 0.65, "learning_rate": 3.9141822663893384e-05, "loss": 0.0, "step": 2819 }, { "epoch": 0.65, "learning_rate": 3.9137970880517686e-05, "loss": 0.0, "step": 2820 }, { "epoch": 0.65, "learning_rate": 3.913411909714198e-05, "loss": 0.0, "step": 2821 }, { "epoch": 0.65, "learning_rate": 3.9130267313766276e-05, "loss": 0.0, "step": 2822 }, { "epoch": 0.65, "learning_rate": 3.912641553039057e-05, "loss": 0.0, "step": 2823 }, { "epoch": 0.65, "learning_rate": 3.9122563747014866e-05, "loss": 0.0, "step": 2824 }, { "epoch": 0.65, "learning_rate": 3.911871196363917e-05, "loss": 0.0, "step": 2825 }, { "epoch": 0.65, "learning_rate": 3.911486018026346e-05, "loss": 0.0, "step": 2826 }, { "epoch": 0.65, "learning_rate": 3.911100839688776e-05, "loss": 0.0, "step": 2827 }, { "epoch": 0.65, "learning_rate": 3.910715661351206e-05, "loss": 0.0005, "step": 2828 }, { "epoch": 0.65, "learning_rate": 3.9103304830136354e-05, "loss": 0.0001, "step": 2829 }, { "epoch": 0.65, "learning_rate": 3.9099453046760656e-05, "loss": 0.0, "step": 2830 }, { "epoch": 0.65, "learning_rate": 3.909560126338495e-05, "loss": 0.0, "step": 2831 }, { "epoch": 0.65, "learning_rate": 3.9091749480009246e-05, "loss": 0.0, "step": 2832 }, { "epoch": 0.65, "learning_rate": 3.908789769663354e-05, "loss": 0.0, "step": 2833 }, { "epoch": 0.65, "learning_rate": 3.9084045913257836e-05, "loss": 0.0, "step": 2834 }, { "epoch": 0.66, "learning_rate": 3.908019412988214e-05, "loss": 0.0, "step": 2835 }, { "epoch": 0.66, "learning_rate": 3.907634234650643e-05, "loss": 0.0, "step": 2836 }, { "epoch": 0.66, "learning_rate": 3.9072490563130734e-05, "loss": 0.0, "step": 2837 }, { "epoch": 0.66, "learning_rate": 3.906863877975503e-05, "loss": 0.0, "step": 2838 }, { "epoch": 0.66, "learning_rate": 3.9064786996379324e-05, "loss": 0.0, "step": 2839 }, { "epoch": 0.66, "learning_rate": 3.9060935213003626e-05, "loss": 0.0001, "step": 2840 }, { "epoch": 0.66, "learning_rate": 3.905708342962792e-05, "loss": 0.0, "step": 2841 }, { "epoch": 0.66, "learning_rate": 3.9053231646252216e-05, "loss": 0.0, "step": 2842 }, { "epoch": 0.66, "learning_rate": 3.904937986287651e-05, "loss": 0.0, "step": 2843 }, { "epoch": 0.66, "learning_rate": 3.904552807950081e-05, "loss": 0.0, "step": 2844 }, { "epoch": 0.66, "learning_rate": 3.904167629612511e-05, "loss": 0.0, "step": 2845 }, { "epoch": 0.66, "learning_rate": 3.90378245127494e-05, "loss": 0.0, "step": 2846 }, { "epoch": 0.66, "learning_rate": 3.9033972729373704e-05, "loss": 0.0008, "step": 2847 }, { "epoch": 0.66, "learning_rate": 3.9030120945998e-05, "loss": 0.0, "step": 2848 }, { "epoch": 0.66, "learning_rate": 3.90262691626223e-05, "loss": 0.0, "step": 2849 }, { "epoch": 0.66, "learning_rate": 3.9022417379246596e-05, "loss": 0.0, "step": 2850 }, { "epoch": 0.66, "learning_rate": 3.901856559587089e-05, "loss": 0.0, "step": 2851 }, { "epoch": 0.66, "learning_rate": 3.9014713812495186e-05, "loss": 0.0, "step": 2852 }, { "epoch": 0.66, "learning_rate": 3.901086202911948e-05, "loss": 0.0, "step": 2853 }, { "epoch": 0.66, "learning_rate": 3.900701024574378e-05, "loss": 0.0, "step": 2854 }, { "epoch": 0.66, "learning_rate": 3.900315846236808e-05, "loss": 0.0, "step": 2855 }, { "epoch": 0.66, "learning_rate": 3.899930667899238e-05, "loss": 0.0, "step": 2856 }, { "epoch": 0.66, "learning_rate": 3.8995454895616674e-05, "loss": 0.0, "step": 2857 }, { "epoch": 0.66, "learning_rate": 3.899160311224097e-05, "loss": 0.0, "step": 2858 }, { "epoch": 0.66, "learning_rate": 3.898775132886527e-05, "loss": 0.0, "step": 2859 }, { "epoch": 0.66, "learning_rate": 3.8983899545489565e-05, "loss": 0.0, "step": 2860 }, { "epoch": 0.66, "learning_rate": 3.898004776211386e-05, "loss": 0.0, "step": 2861 }, { "epoch": 0.66, "learning_rate": 3.8976195978738155e-05, "loss": 0.0, "step": 2862 }, { "epoch": 0.66, "learning_rate": 3.897234419536245e-05, "loss": 0.0, "step": 2863 }, { "epoch": 0.66, "learning_rate": 3.896849241198675e-05, "loss": 0.0, "step": 2864 }, { "epoch": 0.66, "learning_rate": 3.896464062861105e-05, "loss": 0.0, "step": 2865 }, { "epoch": 0.66, "learning_rate": 3.896078884523535e-05, "loss": 0.0, "step": 2866 }, { "epoch": 0.66, "learning_rate": 3.8956937061859644e-05, "loss": 0.0, "step": 2867 }, { "epoch": 0.66, "learning_rate": 3.895308527848394e-05, "loss": 0.0002, "step": 2868 }, { "epoch": 0.66, "learning_rate": 3.894923349510824e-05, "loss": 0.0001, "step": 2869 }, { "epoch": 0.66, "learning_rate": 3.894538171173253e-05, "loss": 0.0001, "step": 2870 }, { "epoch": 0.66, "learning_rate": 3.894152992835683e-05, "loss": 0.0, "step": 2871 }, { "epoch": 0.66, "learning_rate": 3.8937678144981125e-05, "loss": 0.0, "step": 2872 }, { "epoch": 0.66, "learning_rate": 3.893382636160543e-05, "loss": 0.0, "step": 2873 }, { "epoch": 0.66, "learning_rate": 3.892997457822972e-05, "loss": 0.0001, "step": 2874 }, { "epoch": 0.66, "learning_rate": 3.892612279485402e-05, "loss": 0.0, "step": 2875 }, { "epoch": 0.66, "learning_rate": 3.892227101147832e-05, "loss": 0.0, "step": 2876 }, { "epoch": 0.66, "learning_rate": 3.8918419228102614e-05, "loss": 0.0001, "step": 2877 }, { "epoch": 0.67, "learning_rate": 3.8914567444726915e-05, "loss": 0.0, "step": 2878 }, { "epoch": 0.67, "learning_rate": 3.8910715661351204e-05, "loss": 0.0, "step": 2879 }, { "epoch": 0.67, "learning_rate": 3.8906863877975505e-05, "loss": 0.0, "step": 2880 }, { "epoch": 0.67, "learning_rate": 3.89030120945998e-05, "loss": 0.0, "step": 2881 }, { "epoch": 0.67, "learning_rate": 3.8899160311224095e-05, "loss": 0.0, "step": 2882 }, { "epoch": 0.67, "learning_rate": 3.88953085278484e-05, "loss": 0.0, "step": 2883 }, { "epoch": 0.67, "learning_rate": 3.889145674447269e-05, "loss": 0.0, "step": 2884 }, { "epoch": 0.67, "learning_rate": 3.8887604961096994e-05, "loss": 0.0001, "step": 2885 }, { "epoch": 0.67, "learning_rate": 3.888375317772129e-05, "loss": 0.0, "step": 2886 }, { "epoch": 0.67, "learning_rate": 3.8879901394345583e-05, "loss": 0.0, "step": 2887 }, { "epoch": 0.67, "learning_rate": 3.8876049610969885e-05, "loss": 0.0, "step": 2888 }, { "epoch": 0.67, "learning_rate": 3.8872197827594173e-05, "loss": 0.0, "step": 2889 }, { "epoch": 0.67, "learning_rate": 3.8868346044218475e-05, "loss": 0.0, "step": 2890 }, { "epoch": 0.67, "learning_rate": 3.886449426084277e-05, "loss": 0.0, "step": 2891 }, { "epoch": 0.67, "learning_rate": 3.886064247746707e-05, "loss": 0.0, "step": 2892 }, { "epoch": 0.67, "learning_rate": 3.885679069409137e-05, "loss": 0.0, "step": 2893 }, { "epoch": 0.67, "learning_rate": 3.885293891071566e-05, "loss": 0.0, "step": 2894 }, { "epoch": 0.67, "learning_rate": 3.8849087127339963e-05, "loss": 0.0, "step": 2895 }, { "epoch": 0.67, "learning_rate": 3.884523534396426e-05, "loss": 0.0, "step": 2896 }, { "epoch": 0.67, "learning_rate": 3.884138356058855e-05, "loss": 0.0, "step": 2897 }, { "epoch": 0.67, "learning_rate": 3.883753177721285e-05, "loss": 0.0, "step": 2898 }, { "epoch": 0.67, "learning_rate": 3.883367999383714e-05, "loss": 0.0, "step": 2899 }, { "epoch": 0.67, "learning_rate": 3.8829828210461445e-05, "loss": 0.0, "step": 2900 }, { "epoch": 0.67, "learning_rate": 3.882597642708574e-05, "loss": 0.0, "step": 2901 }, { "epoch": 0.67, "learning_rate": 3.882212464371004e-05, "loss": 0.0001, "step": 2902 }, { "epoch": 0.67, "learning_rate": 3.881827286033434e-05, "loss": 0.0, "step": 2903 }, { "epoch": 0.67, "learning_rate": 3.881442107695863e-05, "loss": 0.0001, "step": 2904 }, { "epoch": 0.67, "learning_rate": 3.881056929358293e-05, "loss": 0.0, "step": 2905 }, { "epoch": 0.67, "learning_rate": 3.880671751020723e-05, "loss": 0.0, "step": 2906 }, { "epoch": 0.67, "learning_rate": 3.880286572683152e-05, "loss": 0.0, "step": 2907 }, { "epoch": 0.67, "learning_rate": 3.879901394345582e-05, "loss": 0.0, "step": 2908 }, { "epoch": 0.67, "learning_rate": 3.879516216008012e-05, "loss": 0.0, "step": 2909 }, { "epoch": 0.67, "learning_rate": 3.8791310376704415e-05, "loss": 0.0, "step": 2910 }, { "epoch": 0.67, "learning_rate": 3.878745859332871e-05, "loss": 0.0, "step": 2911 }, { "epoch": 0.67, "learning_rate": 3.878360680995301e-05, "loss": 0.0, "step": 2912 }, { "epoch": 0.67, "learning_rate": 3.8779755026577307e-05, "loss": 0.0, "step": 2913 }, { "epoch": 0.67, "learning_rate": 3.877590324320161e-05, "loss": 0.0, "step": 2914 }, { "epoch": 0.67, "learning_rate": 3.87720514598259e-05, "loss": 0.0, "step": 2915 }, { "epoch": 0.67, "learning_rate": 3.87681996764502e-05, "loss": 0.0, "step": 2916 }, { "epoch": 0.67, "learning_rate": 3.876434789307449e-05, "loss": 0.0, "step": 2917 }, { "epoch": 0.67, "learning_rate": 3.876049610969879e-05, "loss": 0.0, "step": 2918 }, { "epoch": 0.67, "learning_rate": 3.875664432632309e-05, "loss": 0.0, "step": 2919 }, { "epoch": 0.67, "learning_rate": 3.8752792542947385e-05, "loss": 0.0, "step": 2920 }, { "epoch": 0.68, "learning_rate": 3.8748940759571687e-05, "loss": 0.0, "step": 2921 }, { "epoch": 0.68, "learning_rate": 3.874508897619598e-05, "loss": 0.0, "step": 2922 }, { "epoch": 0.68, "learning_rate": 3.8741237192820276e-05, "loss": 0.0, "step": 2923 }, { "epoch": 0.68, "learning_rate": 3.873738540944458e-05, "loss": 0.0, "step": 2924 }, { "epoch": 0.68, "learning_rate": 3.873353362606887e-05, "loss": 0.0, "step": 2925 }, { "epoch": 0.68, "learning_rate": 3.872968184269317e-05, "loss": 0.0, "step": 2926 }, { "epoch": 0.68, "learning_rate": 3.872583005931746e-05, "loss": 0.0, "step": 2927 }, { "epoch": 0.68, "learning_rate": 3.8721978275941765e-05, "loss": 0.0, "step": 2928 }, { "epoch": 0.68, "learning_rate": 3.871812649256606e-05, "loss": 0.0, "step": 2929 }, { "epoch": 0.68, "learning_rate": 3.8714274709190355e-05, "loss": 0.0, "step": 2930 }, { "epoch": 0.68, "learning_rate": 3.8710422925814656e-05, "loss": 0.0, "step": 2931 }, { "epoch": 0.68, "learning_rate": 3.870657114243895e-05, "loss": 0.0, "step": 2932 }, { "epoch": 0.68, "learning_rate": 3.870271935906325e-05, "loss": 0.0, "step": 2933 }, { "epoch": 0.68, "learning_rate": 3.869886757568755e-05, "loss": 0.0, "step": 2934 }, { "epoch": 0.68, "learning_rate": 3.8695015792311836e-05, "loss": 0.0, "step": 2935 }, { "epoch": 0.68, "learning_rate": 3.869116400893614e-05, "loss": 0.0, "step": 2936 }, { "epoch": 0.68, "learning_rate": 3.868731222556043e-05, "loss": 0.0, "step": 2937 }, { "epoch": 0.68, "learning_rate": 3.8683460442184735e-05, "loss": 0.0, "step": 2938 }, { "epoch": 0.68, "learning_rate": 3.867960865880903e-05, "loss": 0.0, "step": 2939 }, { "epoch": 0.68, "learning_rate": 3.8675756875433325e-05, "loss": 0.0, "step": 2940 }, { "epoch": 0.68, "learning_rate": 3.8671905092057626e-05, "loss": 0.0, "step": 2941 }, { "epoch": 0.68, "learning_rate": 3.866805330868192e-05, "loss": 0.0, "step": 2942 }, { "epoch": 0.68, "learning_rate": 3.866420152530622e-05, "loss": 0.0, "step": 2943 }, { "epoch": 0.68, "learning_rate": 3.866034974193052e-05, "loss": 0.0, "step": 2944 }, { "epoch": 0.68, "learning_rate": 3.865649795855481e-05, "loss": 0.0, "step": 2945 }, { "epoch": 0.68, "learning_rate": 3.865264617517911e-05, "loss": 0.0, "step": 2946 }, { "epoch": 0.68, "learning_rate": 3.86487943918034e-05, "loss": 0.0, "step": 2947 }, { "epoch": 0.68, "learning_rate": 3.8644942608427705e-05, "loss": 0.0, "step": 2948 }, { "epoch": 0.68, "learning_rate": 3.8641090825052e-05, "loss": 0.0, "step": 2949 }, { "epoch": 0.68, "learning_rate": 3.86372390416763e-05, "loss": 0.0, "step": 2950 }, { "epoch": 0.68, "learning_rate": 3.8633387258300596e-05, "loss": 0.0, "step": 2951 }, { "epoch": 0.68, "learning_rate": 3.862953547492489e-05, "loss": 0.0001, "step": 2952 }, { "epoch": 0.68, "learning_rate": 3.862568369154919e-05, "loss": 0.0001, "step": 2953 }, { "epoch": 0.68, "learning_rate": 3.862183190817348e-05, "loss": 0.0, "step": 2954 }, { "epoch": 0.68, "learning_rate": 3.861798012479778e-05, "loss": 0.0, "step": 2955 }, { "epoch": 0.68, "learning_rate": 3.861412834142208e-05, "loss": 0.0004, "step": 2956 }, { "epoch": 0.68, "learning_rate": 3.861027655804638e-05, "loss": 0.0019, "step": 2957 }, { "epoch": 0.68, "learning_rate": 3.8606424774670674e-05, "loss": 0.0, "step": 2958 }, { "epoch": 0.68, "learning_rate": 3.860257299129497e-05, "loss": 0.0, "step": 2959 }, { "epoch": 0.68, "learning_rate": 3.859872120791927e-05, "loss": 0.0, "step": 2960 }, { "epoch": 0.68, "learning_rate": 3.8594869424543566e-05, "loss": 0.0, "step": 2961 }, { "epoch": 0.68, "learning_rate": 3.859101764116787e-05, "loss": 0.0, "step": 2962 }, { "epoch": 0.68, "learning_rate": 3.8587165857792156e-05, "loss": 0.0, "step": 2963 }, { "epoch": 0.69, "learning_rate": 3.858331407441646e-05, "loss": 0.0, "step": 2964 }, { "epoch": 0.69, "learning_rate": 3.857946229104075e-05, "loss": 0.0, "step": 2965 }, { "epoch": 0.69, "learning_rate": 3.857561050766505e-05, "loss": 0.0, "step": 2966 }, { "epoch": 0.69, "learning_rate": 3.857175872428935e-05, "loss": 0.0, "step": 2967 }, { "epoch": 0.69, "learning_rate": 3.8567906940913644e-05, "loss": 0.0, "step": 2968 }, { "epoch": 0.69, "learning_rate": 3.8564055157537946e-05, "loss": 0.0, "step": 2969 }, { "epoch": 0.69, "learning_rate": 3.856020337416224e-05, "loss": 0.0, "step": 2970 }, { "epoch": 0.69, "learning_rate": 3.8556351590786536e-05, "loss": 0.0, "step": 2971 }, { "epoch": 0.69, "learning_rate": 3.855249980741084e-05, "loss": 0.0, "step": 2972 }, { "epoch": 0.69, "learning_rate": 3.8548648024035126e-05, "loss": 0.0, "step": 2973 }, { "epoch": 0.69, "learning_rate": 3.854479624065943e-05, "loss": 0.0001, "step": 2974 }, { "epoch": 0.69, "learning_rate": 3.854094445728372e-05, "loss": 0.0, "step": 2975 }, { "epoch": 0.69, "learning_rate": 3.853709267390802e-05, "loss": 0.0, "step": 2976 }, { "epoch": 0.69, "learning_rate": 3.853324089053232e-05, "loss": 0.0, "step": 2977 }, { "epoch": 0.69, "learning_rate": 3.8529389107156614e-05, "loss": 0.0, "step": 2978 }, { "epoch": 0.69, "learning_rate": 3.8525537323780916e-05, "loss": 0.0, "step": 2979 }, { "epoch": 0.69, "learning_rate": 3.852168554040521e-05, "loss": 0.0, "step": 2980 }, { "epoch": 0.69, "learning_rate": 3.8517833757029506e-05, "loss": 0.0, "step": 2981 }, { "epoch": 0.69, "learning_rate": 3.85139819736538e-05, "loss": 0.0, "step": 2982 }, { "epoch": 0.69, "learning_rate": 3.8510130190278096e-05, "loss": 0.0, "step": 2983 }, { "epoch": 0.69, "learning_rate": 3.85062784069024e-05, "loss": 0.0, "step": 2984 }, { "epoch": 0.69, "learning_rate": 3.850242662352669e-05, "loss": 0.0, "step": 2985 }, { "epoch": 0.69, "learning_rate": 3.8498574840150994e-05, "loss": 0.0, "step": 2986 }, { "epoch": 0.69, "learning_rate": 3.849472305677529e-05, "loss": 0.0, "step": 2987 }, { "epoch": 0.69, "learning_rate": 3.8490871273399584e-05, "loss": 0.0, "step": 2988 }, { "epoch": 0.69, "learning_rate": 3.8487019490023886e-05, "loss": 0.0, "step": 2989 }, { "epoch": 0.69, "learning_rate": 3.848316770664818e-05, "loss": 0.0, "step": 2990 }, { "epoch": 0.69, "learning_rate": 3.8479315923272476e-05, "loss": 0.0, "step": 2991 }, { "epoch": 0.69, "learning_rate": 3.847546413989677e-05, "loss": 0.0003, "step": 2992 }, { "epoch": 0.69, "learning_rate": 3.847161235652107e-05, "loss": 0.0, "step": 2993 }, { "epoch": 0.69, "learning_rate": 3.846776057314537e-05, "loss": 0.0001, "step": 2994 }, { "epoch": 0.69, "learning_rate": 3.846390878976966e-05, "loss": 0.0, "step": 2995 }, { "epoch": 0.69, "learning_rate": 3.8460057006393964e-05, "loss": 0.0, "step": 2996 }, { "epoch": 0.69, "learning_rate": 3.845620522301826e-05, "loss": 0.0, "step": 2997 }, { "epoch": 0.69, "learning_rate": 3.845235343964256e-05, "loss": 0.0, "step": 2998 }, { "epoch": 0.69, "learning_rate": 3.8448501656266856e-05, "loss": 0.0, "step": 2999 }, { "epoch": 0.69, "learning_rate": 3.844464987289115e-05, "loss": 0.0, "step": 3000 }, { "epoch": 0.69, "learning_rate": 3.8440798089515446e-05, "loss": 0.0001, "step": 3001 }, { "epoch": 0.69, "learning_rate": 3.843694630613974e-05, "loss": 0.0, "step": 3002 }, { "epoch": 0.69, "learning_rate": 3.843309452276404e-05, "loss": 0.0, "step": 3003 }, { "epoch": 0.69, "learning_rate": 3.842924273938834e-05, "loss": 0.0, "step": 3004 }, { "epoch": 0.69, "learning_rate": 3.842539095601264e-05, "loss": 0.0, "step": 3005 }, { "epoch": 0.69, "learning_rate": 3.8421539172636934e-05, "loss": 0.0, "step": 3006 }, { "epoch": 0.69, "learning_rate": 3.841768738926123e-05, "loss": 0.0, "step": 3007 }, { "epoch": 0.7, "learning_rate": 3.841383560588553e-05, "loss": 0.0, "step": 3008 }, { "epoch": 0.7, "learning_rate": 3.8409983822509826e-05, "loss": 0.0, "step": 3009 }, { "epoch": 0.7, "learning_rate": 3.840613203913412e-05, "loss": 0.0, "step": 3010 }, { "epoch": 0.7, "learning_rate": 3.8402280255758415e-05, "loss": 0.0, "step": 3011 }, { "epoch": 0.7, "learning_rate": 3.839842847238271e-05, "loss": 0.0, "step": 3012 }, { "epoch": 0.7, "learning_rate": 3.839457668900701e-05, "loss": 0.0, "step": 3013 }, { "epoch": 0.7, "learning_rate": 3.839072490563131e-05, "loss": 0.0, "step": 3014 }, { "epoch": 0.7, "learning_rate": 3.838687312225561e-05, "loss": 0.0, "step": 3015 }, { "epoch": 0.7, "learning_rate": 3.8383021338879904e-05, "loss": 0.0, "step": 3016 }, { "epoch": 0.7, "learning_rate": 3.83791695555042e-05, "loss": 0.0, "step": 3017 }, { "epoch": 0.7, "learning_rate": 3.83753177721285e-05, "loss": 0.0, "step": 3018 }, { "epoch": 0.7, "learning_rate": 3.837146598875279e-05, "loss": 0.0, "step": 3019 }, { "epoch": 0.7, "learning_rate": 3.836761420537709e-05, "loss": 0.0, "step": 3020 }, { "epoch": 0.7, "learning_rate": 3.8363762422001385e-05, "loss": 0.0, "step": 3021 }, { "epoch": 0.7, "learning_rate": 3.835991063862569e-05, "loss": 0.0, "step": 3022 }, { "epoch": 0.7, "learning_rate": 3.835605885524998e-05, "loss": 0.0, "step": 3023 }, { "epoch": 0.7, "learning_rate": 3.835220707187428e-05, "loss": 0.0, "step": 3024 }, { "epoch": 0.7, "learning_rate": 3.834835528849858e-05, "loss": 0.0, "step": 3025 }, { "epoch": 0.7, "learning_rate": 3.8344503505122874e-05, "loss": 0.0, "step": 3026 }, { "epoch": 0.7, "learning_rate": 3.8340651721747175e-05, "loss": 0.0, "step": 3027 }, { "epoch": 0.7, "learning_rate": 3.833679993837147e-05, "loss": 0.0, "step": 3028 }, { "epoch": 0.7, "learning_rate": 3.8332948154995765e-05, "loss": 0.0, "step": 3029 }, { "epoch": 0.7, "learning_rate": 3.832909637162006e-05, "loss": 0.0, "step": 3030 }, { "epoch": 0.7, "learning_rate": 3.8325244588244355e-05, "loss": 0.0, "step": 3031 }, { "epoch": 0.7, "learning_rate": 3.832139280486866e-05, "loss": 0.0, "step": 3032 }, { "epoch": 0.7, "learning_rate": 3.831754102149295e-05, "loss": 0.0, "step": 3033 }, { "epoch": 0.7, "learning_rate": 3.8313689238117254e-05, "loss": 0.0, "step": 3034 }, { "epoch": 0.7, "learning_rate": 3.830983745474155e-05, "loss": 0.0, "step": 3035 }, { "epoch": 0.7, "learning_rate": 3.8305985671365844e-05, "loss": 0.0, "step": 3036 }, { "epoch": 0.7, "learning_rate": 3.8302133887990145e-05, "loss": 0.0, "step": 3037 }, { "epoch": 0.7, "learning_rate": 3.8298282104614433e-05, "loss": 0.0, "step": 3038 }, { "epoch": 0.7, "learning_rate": 3.8294430321238735e-05, "loss": 0.0, "step": 3039 }, { "epoch": 0.7, "learning_rate": 3.829057853786303e-05, "loss": 0.0, "step": 3040 }, { "epoch": 0.7, "learning_rate": 3.828672675448733e-05, "loss": 0.0, "step": 3041 }, { "epoch": 0.7, "learning_rate": 3.828287497111163e-05, "loss": 0.0, "step": 3042 }, { "epoch": 0.7, "learning_rate": 3.827902318773592e-05, "loss": 0.0, "step": 3043 }, { "epoch": 0.7, "learning_rate": 3.8275171404360224e-05, "loss": 0.0, "step": 3044 }, { "epoch": 0.7, "learning_rate": 3.827131962098452e-05, "loss": 0.0001, "step": 3045 }, { "epoch": 0.7, "learning_rate": 3.826746783760882e-05, "loss": 0.0, "step": 3046 }, { "epoch": 0.7, "learning_rate": 3.826361605423311e-05, "loss": 0.0, "step": 3047 }, { "epoch": 0.7, "learning_rate": 3.82597642708574e-05, "loss": 0.0, "step": 3048 }, { "epoch": 0.7, "learning_rate": 3.8255912487481705e-05, "loss": 0.0, "step": 3049 }, { "epoch": 0.7, "learning_rate": 3.8252060704106e-05, "loss": 0.0, "step": 3050 }, { "epoch": 0.71, "learning_rate": 3.82482089207303e-05, "loss": 0.0, "step": 3051 }, { "epoch": 0.71, "learning_rate": 3.82443571373546e-05, "loss": 0.0001, "step": 3052 }, { "epoch": 0.71, "learning_rate": 3.824050535397889e-05, "loss": 0.0, "step": 3053 }, { "epoch": 0.71, "learning_rate": 3.8236653570603193e-05, "loss": 0.0, "step": 3054 }, { "epoch": 0.71, "learning_rate": 3.823280178722749e-05, "loss": 0.0, "step": 3055 }, { "epoch": 0.71, "learning_rate": 3.822895000385179e-05, "loss": 0.0, "step": 3056 }, { "epoch": 0.71, "learning_rate": 3.822509822047608e-05, "loss": 0.0, "step": 3057 }, { "epoch": 0.71, "learning_rate": 3.822124643710038e-05, "loss": 0.0, "step": 3058 }, { "epoch": 0.71, "learning_rate": 3.8217394653724675e-05, "loss": 0.0, "step": 3059 }, { "epoch": 0.71, "learning_rate": 3.821354287034897e-05, "loss": 0.0, "step": 3060 }, { "epoch": 0.71, "learning_rate": 3.820969108697327e-05, "loss": 0.0, "step": 3061 }, { "epoch": 0.71, "learning_rate": 3.820583930359757e-05, "loss": 0.0, "step": 3062 }, { "epoch": 0.71, "learning_rate": 3.820198752022187e-05, "loss": 0.0, "step": 3063 }, { "epoch": 0.71, "learning_rate": 3.819813573684616e-05, "loss": 0.0, "step": 3064 }, { "epoch": 0.71, "learning_rate": 3.819428395347046e-05, "loss": 0.0, "step": 3065 }, { "epoch": 0.71, "learning_rate": 3.819043217009475e-05, "loss": 0.0, "step": 3066 }, { "epoch": 0.71, "learning_rate": 3.818658038671905e-05, "loss": 0.0, "step": 3067 }, { "epoch": 0.71, "learning_rate": 3.818272860334335e-05, "loss": 0.0, "step": 3068 }, { "epoch": 0.71, "learning_rate": 3.8178876819967645e-05, "loss": 0.0, "step": 3069 }, { "epoch": 0.71, "learning_rate": 3.817502503659195e-05, "loss": 0.0, "step": 3070 }, { "epoch": 0.71, "learning_rate": 3.817117325321624e-05, "loss": 0.0, "step": 3071 }, { "epoch": 0.71, "learning_rate": 3.8167321469840537e-05, "loss": 0.0, "step": 3072 }, { "epoch": 0.71, "learning_rate": 3.816346968646484e-05, "loss": 0.0, "step": 3073 }, { "epoch": 0.71, "learning_rate": 3.815961790308913e-05, "loss": 0.0, "step": 3074 }, { "epoch": 0.71, "learning_rate": 3.815576611971343e-05, "loss": 0.0, "step": 3075 }, { "epoch": 0.71, "learning_rate": 3.815191433633772e-05, "loss": 0.0, "step": 3076 }, { "epoch": 0.71, "learning_rate": 3.8148062552962025e-05, "loss": 0.0, "step": 3077 }, { "epoch": 0.71, "learning_rate": 3.814421076958632e-05, "loss": 0.0, "step": 3078 }, { "epoch": 0.71, "learning_rate": 3.8140358986210615e-05, "loss": 0.0004, "step": 3079 }, { "epoch": 0.71, "learning_rate": 3.8136507202834916e-05, "loss": 0.0, "step": 3080 }, { "epoch": 0.71, "learning_rate": 3.813265541945921e-05, "loss": 0.0, "step": 3081 }, { "epoch": 0.71, "learning_rate": 3.812880363608351e-05, "loss": 0.0, "step": 3082 }, { "epoch": 0.71, "learning_rate": 3.812495185270781e-05, "loss": 0.0, "step": 3083 }, { "epoch": 0.71, "learning_rate": 3.81211000693321e-05, "loss": 0.0, "step": 3084 }, { "epoch": 0.71, "learning_rate": 3.81172482859564e-05, "loss": 0.0, "step": 3085 }, { "epoch": 0.71, "learning_rate": 3.811339650258069e-05, "loss": 0.0, "step": 3086 }, { "epoch": 0.71, "learning_rate": 3.8109544719204995e-05, "loss": 0.0, "step": 3087 }, { "epoch": 0.71, "learning_rate": 3.810569293582929e-05, "loss": 0.0, "step": 3088 }, { "epoch": 0.71, "learning_rate": 3.8101841152453585e-05, "loss": 0.0, "step": 3089 }, { "epoch": 0.71, "learning_rate": 3.8097989369077886e-05, "loss": 0.0, "step": 3090 }, { "epoch": 0.71, "learning_rate": 3.809413758570218e-05, "loss": 0.0001, "step": 3091 }, { "epoch": 0.71, "learning_rate": 3.809028580232648e-05, "loss": 0.0, "step": 3092 }, { "epoch": 0.71, "learning_rate": 3.808643401895078e-05, "loss": 0.0001, "step": 3093 }, { "epoch": 0.72, "learning_rate": 3.808258223557507e-05, "loss": 0.0, "step": 3094 }, { "epoch": 0.72, "learning_rate": 3.807873045219937e-05, "loss": 0.0, "step": 3095 }, { "epoch": 0.72, "learning_rate": 3.807487866882366e-05, "loss": 0.0, "step": 3096 }, { "epoch": 0.72, "learning_rate": 3.8071026885447965e-05, "loss": 0.0, "step": 3097 }, { "epoch": 0.72, "learning_rate": 3.806717510207226e-05, "loss": 0.0, "step": 3098 }, { "epoch": 0.72, "learning_rate": 3.806332331869656e-05, "loss": 0.0, "step": 3099 }, { "epoch": 0.72, "learning_rate": 3.8059471535320856e-05, "loss": 0.0, "step": 3100 }, { "epoch": 0.72, "learning_rate": 3.805561975194515e-05, "loss": 0.0, "step": 3101 }, { "epoch": 0.72, "learning_rate": 3.805176796856945e-05, "loss": 0.0, "step": 3102 }, { "epoch": 0.72, "learning_rate": 3.804791618519374e-05, "loss": 0.0, "step": 3103 }, { "epoch": 0.72, "learning_rate": 3.804406440181804e-05, "loss": 0.0, "step": 3104 }, { "epoch": 0.72, "learning_rate": 3.804021261844234e-05, "loss": 0.0, "step": 3105 }, { "epoch": 0.72, "learning_rate": 3.803636083506664e-05, "loss": 0.0002, "step": 3106 }, { "epoch": 0.72, "learning_rate": 3.8032509051690935e-05, "loss": 0.0, "step": 3107 }, { "epoch": 0.72, "learning_rate": 3.802865726831523e-05, "loss": 0.0, "step": 3108 }, { "epoch": 0.72, "learning_rate": 3.802480548493953e-05, "loss": 0.0, "step": 3109 }, { "epoch": 0.72, "learning_rate": 3.8020953701563826e-05, "loss": 0.0, "step": 3110 }, { "epoch": 0.72, "learning_rate": 3.801710191818813e-05, "loss": 0.0, "step": 3111 }, { "epoch": 0.72, "learning_rate": 3.801325013481242e-05, "loss": 0.0001, "step": 3112 }, { "epoch": 0.72, "learning_rate": 3.800939835143672e-05, "loss": 0.0, "step": 3113 }, { "epoch": 0.72, "learning_rate": 3.800554656806101e-05, "loss": 0.0, "step": 3114 }, { "epoch": 0.72, "learning_rate": 3.800169478468531e-05, "loss": 0.0, "step": 3115 }, { "epoch": 0.72, "learning_rate": 3.799784300130961e-05, "loss": 0.0, "step": 3116 }, { "epoch": 0.72, "learning_rate": 3.7993991217933904e-05, "loss": 0.0, "step": 3117 }, { "epoch": 0.72, "learning_rate": 3.7990139434558206e-05, "loss": 0.0, "step": 3118 }, { "epoch": 0.72, "learning_rate": 3.79862876511825e-05, "loss": 0.0, "step": 3119 }, { "epoch": 0.72, "learning_rate": 3.7982435867806796e-05, "loss": 0.0, "step": 3120 }, { "epoch": 0.72, "learning_rate": 3.79785840844311e-05, "loss": 0.0, "step": 3121 }, { "epoch": 0.72, "learning_rate": 3.7974732301055386e-05, "loss": 0.0, "step": 3122 }, { "epoch": 0.72, "learning_rate": 3.797088051767969e-05, "loss": 0.0, "step": 3123 }, { "epoch": 0.72, "learning_rate": 3.796702873430398e-05, "loss": 0.0, "step": 3124 }, { "epoch": 0.72, "learning_rate": 3.796317695092828e-05, "loss": 0.0, "step": 3125 }, { "epoch": 0.72, "learning_rate": 3.795932516755258e-05, "loss": 0.0, "step": 3126 }, { "epoch": 0.72, "learning_rate": 3.7955473384176874e-05, "loss": 0.0, "step": 3127 }, { "epoch": 0.72, "learning_rate": 3.7951621600801176e-05, "loss": 0.0, "step": 3128 }, { "epoch": 0.72, "learning_rate": 3.794776981742547e-05, "loss": 0.0, "step": 3129 }, { "epoch": 0.72, "learning_rate": 3.7943918034049766e-05, "loss": 0.0, "step": 3130 }, { "epoch": 0.72, "learning_rate": 3.794006625067406e-05, "loss": 0.0, "step": 3131 }, { "epoch": 0.72, "learning_rate": 3.7936214467298356e-05, "loss": 0.0001, "step": 3132 }, { "epoch": 0.72, "learning_rate": 3.793236268392266e-05, "loss": 0.0, "step": 3133 }, { "epoch": 0.72, "learning_rate": 3.792851090054695e-05, "loss": 0.0, "step": 3134 }, { "epoch": 0.72, "learning_rate": 3.7924659117171254e-05, "loss": 0.0001, "step": 3135 }, { "epoch": 0.72, "learning_rate": 3.792080733379555e-05, "loss": 0.0002, "step": 3136 }, { "epoch": 0.72, "learning_rate": 3.7916955550419844e-05, "loss": 0.0006, "step": 3137 }, { "epoch": 0.73, "learning_rate": 3.7913103767044146e-05, "loss": 0.0, "step": 3138 }, { "epoch": 0.73, "learning_rate": 3.790925198366844e-05, "loss": 0.0, "step": 3139 }, { "epoch": 0.73, "learning_rate": 3.790540020029274e-05, "loss": 0.0, "step": 3140 }, { "epoch": 0.73, "learning_rate": 3.790154841691703e-05, "loss": 0.0, "step": 3141 }, { "epoch": 0.73, "learning_rate": 3.789769663354133e-05, "loss": 0.0, "step": 3142 }, { "epoch": 0.73, "learning_rate": 3.789384485016563e-05, "loss": 0.0, "step": 3143 }, { "epoch": 0.73, "learning_rate": 3.788999306678992e-05, "loss": 0.0, "step": 3144 }, { "epoch": 0.73, "learning_rate": 3.7886141283414224e-05, "loss": 0.0001, "step": 3145 }, { "epoch": 0.73, "learning_rate": 3.788228950003852e-05, "loss": 0.0, "step": 3146 }, { "epoch": 0.73, "learning_rate": 3.787843771666282e-05, "loss": 0.0, "step": 3147 }, { "epoch": 0.73, "learning_rate": 3.7874585933287116e-05, "loss": 0.0, "step": 3148 }, { "epoch": 0.73, "learning_rate": 3.787073414991141e-05, "loss": 0.0, "step": 3149 }, { "epoch": 0.73, "learning_rate": 3.7866882366535706e-05, "loss": 0.0, "step": 3150 }, { "epoch": 0.73, "learning_rate": 3.786303058316e-05, "loss": 0.0, "step": 3151 }, { "epoch": 0.73, "learning_rate": 3.78591787997843e-05, "loss": 0.0, "step": 3152 }, { "epoch": 0.73, "learning_rate": 3.78553270164086e-05, "loss": 0.0002, "step": 3153 }, { "epoch": 0.73, "learning_rate": 3.78514752330329e-05, "loss": 0.0, "step": 3154 }, { "epoch": 0.73, "learning_rate": 3.7847623449657194e-05, "loss": 0.0, "step": 3155 }, { "epoch": 0.73, "learning_rate": 3.784377166628149e-05, "loss": 0.0, "step": 3156 }, { "epoch": 0.73, "learning_rate": 3.783991988290579e-05, "loss": 0.0001, "step": 3157 }, { "epoch": 0.73, "learning_rate": 3.7836068099530086e-05, "loss": 0.0, "step": 3158 }, { "epoch": 0.73, "learning_rate": 3.783221631615438e-05, "loss": 0.0, "step": 3159 }, { "epoch": 0.73, "learning_rate": 3.7828364532778676e-05, "loss": 0.0, "step": 3160 }, { "epoch": 0.73, "learning_rate": 3.782451274940297e-05, "loss": 0.0, "step": 3161 }, { "epoch": 0.73, "learning_rate": 3.782066096602727e-05, "loss": 0.0, "step": 3162 }, { "epoch": 0.73, "learning_rate": 3.781680918265157e-05, "loss": 0.0, "step": 3163 }, { "epoch": 0.73, "learning_rate": 3.781295739927587e-05, "loss": 0.0, "step": 3164 }, { "epoch": 0.73, "learning_rate": 3.7809105615900164e-05, "loss": 0.0, "step": 3165 }, { "epoch": 0.73, "learning_rate": 3.780525383252446e-05, "loss": 0.0, "step": 3166 }, { "epoch": 0.73, "learning_rate": 3.780140204914876e-05, "loss": 0.0, "step": 3167 }, { "epoch": 0.73, "learning_rate": 3.7797550265773056e-05, "loss": 0.0, "step": 3168 }, { "epoch": 0.73, "learning_rate": 3.779369848239735e-05, "loss": 0.0, "step": 3169 }, { "epoch": 0.73, "learning_rate": 3.7789846699021645e-05, "loss": 0.0, "step": 3170 }, { "epoch": 0.73, "learning_rate": 3.778599491564595e-05, "loss": 0.0, "step": 3171 }, { "epoch": 0.73, "learning_rate": 3.778214313227024e-05, "loss": 0.0001, "step": 3172 }, { "epoch": 0.73, "learning_rate": 3.777829134889454e-05, "loss": 0.0, "step": 3173 }, { "epoch": 0.73, "learning_rate": 3.777443956551884e-05, "loss": 0.0, "step": 3174 }, { "epoch": 0.73, "learning_rate": 3.7770587782143134e-05, "loss": 0.0, "step": 3175 }, { "epoch": 0.73, "learning_rate": 3.7766735998767436e-05, "loss": 0.0, "step": 3176 }, { "epoch": 0.73, "learning_rate": 3.776288421539173e-05, "loss": 0.0, "step": 3177 }, { "epoch": 0.73, "learning_rate": 3.7759032432016025e-05, "loss": 0.0, "step": 3178 }, { "epoch": 0.73, "learning_rate": 3.775518064864032e-05, "loss": 0.0001, "step": 3179 }, { "epoch": 0.73, "learning_rate": 3.7751328865264615e-05, "loss": 0.0, "step": 3180 }, { "epoch": 0.74, "learning_rate": 3.774747708188892e-05, "loss": 0.0, "step": 3181 }, { "epoch": 0.74, "learning_rate": 3.774362529851321e-05, "loss": 0.0, "step": 3182 }, { "epoch": 0.74, "learning_rate": 3.7739773515137514e-05, "loss": 0.0, "step": 3183 }, { "epoch": 0.74, "learning_rate": 3.773592173176181e-05, "loss": 0.0004, "step": 3184 }, { "epoch": 0.74, "learning_rate": 3.7732069948386104e-05, "loss": 0.0, "step": 3185 }, { "epoch": 0.74, "learning_rate": 3.7728218165010405e-05, "loss": 0.0, "step": 3186 }, { "epoch": 0.74, "learning_rate": 3.7724366381634694e-05, "loss": 0.0, "step": 3187 }, { "epoch": 0.74, "learning_rate": 3.7720514598258995e-05, "loss": 0.0, "step": 3188 }, { "epoch": 0.74, "learning_rate": 3.771666281488329e-05, "loss": 0.0, "step": 3189 }, { "epoch": 0.74, "learning_rate": 3.771281103150759e-05, "loss": 0.0, "step": 3190 }, { "epoch": 0.74, "learning_rate": 3.770895924813189e-05, "loss": 0.0, "step": 3191 }, { "epoch": 0.74, "learning_rate": 3.770510746475618e-05, "loss": 0.0, "step": 3192 }, { "epoch": 0.74, "learning_rate": 3.7701255681380484e-05, "loss": 0.0, "step": 3193 }, { "epoch": 0.74, "learning_rate": 3.769740389800478e-05, "loss": 0.0, "step": 3194 }, { "epoch": 0.74, "learning_rate": 3.769355211462908e-05, "loss": 0.0, "step": 3195 }, { "epoch": 0.74, "learning_rate": 3.7689700331253375e-05, "loss": 0.0001, "step": 3196 }, { "epoch": 0.74, "learning_rate": 3.7685848547877663e-05, "loss": 0.0, "step": 3197 }, { "epoch": 0.74, "learning_rate": 3.7681996764501965e-05, "loss": 0.0, "step": 3198 }, { "epoch": 0.74, "learning_rate": 3.767814498112626e-05, "loss": 0.0, "step": 3199 }, { "epoch": 0.74, "learning_rate": 3.767429319775056e-05, "loss": 0.0, "step": 3200 }, { "epoch": 0.74, "learning_rate": 3.767044141437486e-05, "loss": 0.0, "step": 3201 }, { "epoch": 0.74, "learning_rate": 3.766658963099915e-05, "loss": 0.0, "step": 3202 }, { "epoch": 0.74, "learning_rate": 3.7662737847623454e-05, "loss": 0.0, "step": 3203 }, { "epoch": 0.74, "learning_rate": 3.765888606424775e-05, "loss": 0.0, "step": 3204 }, { "epoch": 0.74, "learning_rate": 3.765503428087205e-05, "loss": 0.0, "step": 3205 }, { "epoch": 0.74, "learning_rate": 3.765118249749634e-05, "loss": 0.0, "step": 3206 }, { "epoch": 0.74, "learning_rate": 3.764733071412064e-05, "loss": 0.0, "step": 3207 }, { "epoch": 0.74, "learning_rate": 3.7643478930744935e-05, "loss": 0.0, "step": 3208 }, { "epoch": 0.74, "learning_rate": 3.763962714736923e-05, "loss": 0.0, "step": 3209 }, { "epoch": 0.74, "learning_rate": 3.763577536399353e-05, "loss": 0.0, "step": 3210 }, { "epoch": 0.74, "learning_rate": 3.763192358061783e-05, "loss": 0.0, "step": 3211 }, { "epoch": 0.74, "learning_rate": 3.762807179724213e-05, "loss": 0.0, "step": 3212 }, { "epoch": 0.74, "learning_rate": 3.7624220013866423e-05, "loss": 0.0, "step": 3213 }, { "epoch": 0.74, "learning_rate": 3.762036823049072e-05, "loss": 0.0, "step": 3214 }, { "epoch": 0.74, "learning_rate": 3.761651644711501e-05, "loss": 0.0, "step": 3215 }, { "epoch": 0.74, "learning_rate": 3.761266466373931e-05, "loss": 0.0, "step": 3216 }, { "epoch": 0.74, "learning_rate": 3.760881288036361e-05, "loss": 0.0, "step": 3217 }, { "epoch": 0.74, "learning_rate": 3.7604961096987905e-05, "loss": 0.0, "step": 3218 }, { "epoch": 0.74, "learning_rate": 3.760110931361221e-05, "loss": 0.0, "step": 3219 }, { "epoch": 0.74, "learning_rate": 3.75972575302365e-05, "loss": 0.0, "step": 3220 }, { "epoch": 0.74, "learning_rate": 3.7593405746860797e-05, "loss": 0.0, "step": 3221 }, { "epoch": 0.74, "learning_rate": 3.75895539634851e-05, "loss": 0.0, "step": 3222 }, { "epoch": 0.74, "learning_rate": 3.758570218010939e-05, "loss": 0.0, "step": 3223 }, { "epoch": 0.75, "learning_rate": 3.7581850396733695e-05, "loss": 0.0, "step": 3224 }, { "epoch": 0.75, "learning_rate": 3.757799861335798e-05, "loss": 0.0, "step": 3225 }, { "epoch": 0.75, "learning_rate": 3.7574146829982285e-05, "loss": 0.0, "step": 3226 }, { "epoch": 0.75, "learning_rate": 3.757029504660658e-05, "loss": 0.0, "step": 3227 }, { "epoch": 0.75, "learning_rate": 3.7566443263230875e-05, "loss": 0.0001, "step": 3228 }, { "epoch": 0.75, "learning_rate": 3.7562591479855177e-05, "loss": 0.0, "step": 3229 }, { "epoch": 0.75, "learning_rate": 3.755873969647947e-05, "loss": 0.0, "step": 3230 }, { "epoch": 0.75, "learning_rate": 3.755488791310377e-05, "loss": 0.0, "step": 3231 }, { "epoch": 0.75, "learning_rate": 3.755103612972807e-05, "loss": 0.0, "step": 3232 }, { "epoch": 0.75, "learning_rate": 3.754718434635236e-05, "loss": 0.0, "step": 3233 }, { "epoch": 0.75, "learning_rate": 3.754333256297666e-05, "loss": 0.0, "step": 3234 }, { "epoch": 0.75, "learning_rate": 3.753948077960095e-05, "loss": 0.0, "step": 3235 }, { "epoch": 0.75, "learning_rate": 3.7535628996225255e-05, "loss": 0.0, "step": 3236 }, { "epoch": 0.75, "learning_rate": 3.753177721284955e-05, "loss": 0.0, "step": 3237 }, { "epoch": 0.75, "learning_rate": 3.7527925429473845e-05, "loss": 0.0, "step": 3238 }, { "epoch": 0.75, "learning_rate": 3.7524073646098146e-05, "loss": 0.0, "step": 3239 }, { "epoch": 0.75, "learning_rate": 3.752022186272244e-05, "loss": 0.0, "step": 3240 }, { "epoch": 0.75, "learning_rate": 3.751637007934674e-05, "loss": 0.0001, "step": 3241 }, { "epoch": 0.75, "learning_rate": 3.751251829597104e-05, "loss": 0.0, "step": 3242 }, { "epoch": 0.75, "learning_rate": 3.750866651259533e-05, "loss": 0.0, "step": 3243 }, { "epoch": 0.75, "learning_rate": 3.750481472921963e-05, "loss": 0.0, "step": 3244 }, { "epoch": 0.75, "learning_rate": 3.750096294584392e-05, "loss": 0.0, "step": 3245 }, { "epoch": 0.75, "learning_rate": 3.7497111162468225e-05, "loss": 0.0004, "step": 3246 }, { "epoch": 0.75, "learning_rate": 3.749325937909252e-05, "loss": 0.0, "step": 3247 }, { "epoch": 0.75, "learning_rate": 3.748940759571682e-05, "loss": 0.0, "step": 3248 }, { "epoch": 0.75, "learning_rate": 3.7485555812341116e-05, "loss": 0.0, "step": 3249 }, { "epoch": 0.75, "learning_rate": 3.748170402896541e-05, "loss": 0.0, "step": 3250 }, { "epoch": 0.75, "learning_rate": 3.747785224558971e-05, "loss": 0.0, "step": 3251 }, { "epoch": 0.75, "learning_rate": 3.747400046221401e-05, "loss": 0.0, "step": 3252 }, { "epoch": 0.75, "learning_rate": 3.74701486788383e-05, "loss": 0.0, "step": 3253 }, { "epoch": 0.75, "learning_rate": 3.74662968954626e-05, "loss": 0.0, "step": 3254 }, { "epoch": 0.75, "learning_rate": 3.74624451120869e-05, "loss": 0.0, "step": 3255 }, { "epoch": 0.75, "learning_rate": 3.7458593328711195e-05, "loss": 0.0, "step": 3256 }, { "epoch": 0.75, "learning_rate": 3.745474154533549e-05, "loss": 0.0, "step": 3257 }, { "epoch": 0.75, "learning_rate": 3.745088976195979e-05, "loss": 0.0, "step": 3258 }, { "epoch": 0.75, "learning_rate": 3.7447037978584086e-05, "loss": 0.0, "step": 3259 }, { "epoch": 0.75, "learning_rate": 3.744318619520839e-05, "loss": 0.0, "step": 3260 }, { "epoch": 0.75, "learning_rate": 3.743933441183268e-05, "loss": 0.0, "step": 3261 }, { "epoch": 0.75, "learning_rate": 3.743548262845698e-05, "loss": 0.0, "step": 3262 }, { "epoch": 0.75, "learning_rate": 3.743163084508127e-05, "loss": 0.0, "step": 3263 }, { "epoch": 0.75, "learning_rate": 3.742777906170557e-05, "loss": 0.0, "step": 3264 }, { "epoch": 0.75, "learning_rate": 3.742392727832987e-05, "loss": 0.0001, "step": 3265 }, { "epoch": 0.75, "learning_rate": 3.7420075494954164e-05, "loss": 0.0, "step": 3266 }, { "epoch": 0.76, "learning_rate": 3.7416223711578466e-05, "loss": 0.0001, "step": 3267 }, { "epoch": 0.76, "learning_rate": 3.741237192820276e-05, "loss": 0.0, "step": 3268 }, { "epoch": 0.76, "learning_rate": 3.7408520144827056e-05, "loss": 0.0, "step": 3269 }, { "epoch": 0.76, "learning_rate": 3.740466836145136e-05, "loss": 0.0, "step": 3270 }, { "epoch": 0.76, "learning_rate": 3.7400816578075646e-05, "loss": 0.0, "step": 3271 }, { "epoch": 0.76, "learning_rate": 3.739696479469995e-05, "loss": 0.0, "step": 3272 }, { "epoch": 0.76, "learning_rate": 3.739311301132424e-05, "loss": 0.0, "step": 3273 }, { "epoch": 0.76, "learning_rate": 3.738926122794854e-05, "loss": 0.0, "step": 3274 }, { "epoch": 0.76, "learning_rate": 3.738540944457284e-05, "loss": 0.0244, "step": 3275 }, { "epoch": 0.76, "learning_rate": 3.7381557661197134e-05, "loss": 0.0, "step": 3276 }, { "epoch": 0.76, "learning_rate": 3.7377705877821436e-05, "loss": 0.0, "step": 3277 }, { "epoch": 0.76, "learning_rate": 3.737385409444573e-05, "loss": 0.0003, "step": 3278 }, { "epoch": 0.76, "learning_rate": 3.7370002311070026e-05, "loss": 0.0, "step": 3279 }, { "epoch": 0.76, "learning_rate": 3.736615052769433e-05, "loss": 0.0, "step": 3280 }, { "epoch": 0.76, "learning_rate": 3.7362298744318616e-05, "loss": 0.0, "step": 3281 }, { "epoch": 0.76, "learning_rate": 3.735844696094292e-05, "loss": 0.0, "step": 3282 }, { "epoch": 0.76, "learning_rate": 3.735459517756721e-05, "loss": 0.0, "step": 3283 }, { "epoch": 0.76, "learning_rate": 3.7350743394191514e-05, "loss": 0.0, "step": 3284 }, { "epoch": 0.76, "learning_rate": 3.734689161081581e-05, "loss": 0.0, "step": 3285 }, { "epoch": 0.76, "learning_rate": 3.7343039827440104e-05, "loss": 0.0, "step": 3286 }, { "epoch": 0.76, "learning_rate": 3.7339188044064406e-05, "loss": 0.0, "step": 3287 }, { "epoch": 0.76, "learning_rate": 3.73353362606887e-05, "loss": 0.0, "step": 3288 }, { "epoch": 0.76, "learning_rate": 3.7331484477313e-05, "loss": 0.0, "step": 3289 }, { "epoch": 0.76, "learning_rate": 3.732763269393729e-05, "loss": 0.0, "step": 3290 }, { "epoch": 0.76, "learning_rate": 3.732378091056159e-05, "loss": 0.0, "step": 3291 }, { "epoch": 0.76, "learning_rate": 3.731992912718589e-05, "loss": 0.0, "step": 3292 }, { "epoch": 0.76, "learning_rate": 3.731607734381018e-05, "loss": 0.0, "step": 3293 }, { "epoch": 0.76, "learning_rate": 3.7312225560434484e-05, "loss": 0.0, "step": 3294 }, { "epoch": 0.76, "learning_rate": 3.730837377705878e-05, "loss": 0.0, "step": 3295 }, { "epoch": 0.76, "learning_rate": 3.730452199368308e-05, "loss": 0.0, "step": 3296 }, { "epoch": 0.76, "learning_rate": 3.7300670210307376e-05, "loss": 0.0, "step": 3297 }, { "epoch": 0.76, "learning_rate": 3.729681842693167e-05, "loss": 0.0008, "step": 3298 }, { "epoch": 0.76, "learning_rate": 3.7292966643555966e-05, "loss": 0.0001, "step": 3299 }, { "epoch": 0.76, "learning_rate": 3.728911486018026e-05, "loss": 0.0, "step": 3300 }, { "epoch": 0.76, "learning_rate": 3.728526307680456e-05, "loss": 0.0, "step": 3301 }, { "epoch": 0.76, "learning_rate": 3.728141129342886e-05, "loss": 0.0, "step": 3302 }, { "epoch": 0.76, "learning_rate": 3.727755951005316e-05, "loss": 0.0, "step": 3303 }, { "epoch": 0.76, "learning_rate": 3.7273707726677454e-05, "loss": 0.0, "step": 3304 }, { "epoch": 0.76, "learning_rate": 3.726985594330175e-05, "loss": 0.0, "step": 3305 }, { "epoch": 0.76, "learning_rate": 3.726600415992605e-05, "loss": 0.0, "step": 3306 }, { "epoch": 0.76, "learning_rate": 3.7262152376550346e-05, "loss": 0.0, "step": 3307 }, { "epoch": 0.76, "learning_rate": 3.725830059317465e-05, "loss": 0.0, "step": 3308 }, { "epoch": 0.76, "learning_rate": 3.7254448809798936e-05, "loss": 0.0004, "step": 3309 }, { "epoch": 0.76, "learning_rate": 3.725059702642324e-05, "loss": 0.0, "step": 3310 }, { "epoch": 0.77, "learning_rate": 3.724674524304753e-05, "loss": 0.0001, "step": 3311 }, { "epoch": 0.77, "learning_rate": 3.724289345967183e-05, "loss": 0.0, "step": 3312 }, { "epoch": 0.77, "learning_rate": 3.723904167629613e-05, "loss": 0.0, "step": 3313 }, { "epoch": 0.77, "learning_rate": 3.7235189892920424e-05, "loss": 0.0, "step": 3314 }, { "epoch": 0.77, "learning_rate": 3.723133810954472e-05, "loss": 0.0001, "step": 3315 }, { "epoch": 0.77, "learning_rate": 3.722748632616902e-05, "loss": 0.0, "step": 3316 }, { "epoch": 0.77, "learning_rate": 3.7223634542793316e-05, "loss": 0.0, "step": 3317 }, { "epoch": 0.77, "learning_rate": 3.721978275941761e-05, "loss": 0.0, "step": 3318 }, { "epoch": 0.77, "learning_rate": 3.7215930976041906e-05, "loss": 0.0, "step": 3319 }, { "epoch": 0.77, "learning_rate": 3.721207919266621e-05, "loss": 0.0, "step": 3320 }, { "epoch": 0.77, "learning_rate": 3.72082274092905e-05, "loss": 0.0, "step": 3321 }, { "epoch": 0.77, "learning_rate": 3.72043756259148e-05, "loss": 0.0, "step": 3322 }, { "epoch": 0.77, "learning_rate": 3.72005238425391e-05, "loss": 0.0099, "step": 3323 }, { "epoch": 0.77, "learning_rate": 3.7196672059163394e-05, "loss": 0.0, "step": 3324 }, { "epoch": 0.77, "learning_rate": 3.7192820275787696e-05, "loss": 0.0, "step": 3325 }, { "epoch": 0.77, "learning_rate": 3.718896849241199e-05, "loss": 0.0, "step": 3326 }, { "epoch": 0.77, "learning_rate": 3.7185116709036286e-05, "loss": 0.0, "step": 3327 }, { "epoch": 0.77, "learning_rate": 3.718126492566058e-05, "loss": 0.0, "step": 3328 }, { "epoch": 0.77, "learning_rate": 3.7177413142284875e-05, "loss": 0.0, "step": 3329 }, { "epoch": 0.77, "learning_rate": 3.717356135890918e-05, "loss": 0.0, "step": 3330 }, { "epoch": 0.77, "learning_rate": 3.716970957553347e-05, "loss": 0.0, "step": 3331 }, { "epoch": 0.77, "learning_rate": 3.7165857792157774e-05, "loss": 0.0001, "step": 3332 }, { "epoch": 0.77, "learning_rate": 3.716200600878207e-05, "loss": 0.0, "step": 3333 }, { "epoch": 0.77, "learning_rate": 3.7158154225406364e-05, "loss": 0.0, "step": 3334 }, { "epoch": 0.77, "learning_rate": 3.7154302442030665e-05, "loss": 0.0, "step": 3335 }, { "epoch": 0.77, "learning_rate": 3.715045065865496e-05, "loss": 0.0, "step": 3336 }, { "epoch": 0.77, "learning_rate": 3.7146598875279255e-05, "loss": 0.0, "step": 3337 }, { "epoch": 0.77, "learning_rate": 3.714274709190355e-05, "loss": 0.0, "step": 3338 }, { "epoch": 0.77, "learning_rate": 3.713889530852785e-05, "loss": 0.0, "step": 3339 }, { "epoch": 0.77, "learning_rate": 3.713504352515215e-05, "loss": 0.0, "step": 3340 }, { "epoch": 0.77, "learning_rate": 3.713119174177644e-05, "loss": 0.0, "step": 3341 }, { "epoch": 0.77, "learning_rate": 3.7127339958400744e-05, "loss": 0.0, "step": 3342 }, { "epoch": 0.77, "learning_rate": 3.712348817502504e-05, "loss": 0.0, "step": 3343 }, { "epoch": 0.77, "learning_rate": 3.711963639164934e-05, "loss": 0.0, "step": 3344 }, { "epoch": 0.77, "learning_rate": 3.7115784608273635e-05, "loss": 0.0, "step": 3345 }, { "epoch": 0.77, "learning_rate": 3.711193282489793e-05, "loss": 0.0, "step": 3346 }, { "epoch": 0.77, "learning_rate": 3.7108081041522225e-05, "loss": 0.0, "step": 3347 }, { "epoch": 0.77, "learning_rate": 3.710422925814652e-05, "loss": 0.0, "step": 3348 }, { "epoch": 0.77, "learning_rate": 3.710037747477082e-05, "loss": 0.0001, "step": 3349 }, { "epoch": 0.77, "learning_rate": 3.709652569139512e-05, "loss": 0.0, "step": 3350 }, { "epoch": 0.77, "learning_rate": 3.709267390801941e-05, "loss": 0.0002, "step": 3351 }, { "epoch": 0.77, "learning_rate": 3.7088822124643714e-05, "loss": 0.0043, "step": 3352 }, { "epoch": 0.77, "learning_rate": 3.708497034126801e-05, "loss": 0.0, "step": 3353 }, { "epoch": 0.78, "learning_rate": 3.708111855789231e-05, "loss": 0.0, "step": 3354 }, { "epoch": 0.78, "learning_rate": 3.70772667745166e-05, "loss": 0.0, "step": 3355 }, { "epoch": 0.78, "learning_rate": 3.70734149911409e-05, "loss": 0.0, "step": 3356 }, { "epoch": 0.78, "learning_rate": 3.7069563207765195e-05, "loss": 0.0001, "step": 3357 }, { "epoch": 0.78, "learning_rate": 3.706571142438949e-05, "loss": 0.0, "step": 3358 }, { "epoch": 0.78, "learning_rate": 3.706185964101379e-05, "loss": 0.0002, "step": 3359 }, { "epoch": 0.78, "learning_rate": 3.705800785763809e-05, "loss": 0.0, "step": 3360 }, { "epoch": 0.78, "learning_rate": 3.705415607426239e-05, "loss": 0.0, "step": 3361 }, { "epoch": 0.78, "learning_rate": 3.7050304290886683e-05, "loss": 0.0, "step": 3362 }, { "epoch": 0.78, "learning_rate": 3.704645250751098e-05, "loss": 0.0, "step": 3363 }, { "epoch": 0.78, "learning_rate": 3.704260072413528e-05, "loss": 0.0, "step": 3364 }, { "epoch": 0.78, "learning_rate": 3.703874894075957e-05, "loss": 0.0, "step": 3365 }, { "epoch": 0.78, "learning_rate": 3.703489715738387e-05, "loss": 0.0, "step": 3366 }, { "epoch": 0.78, "learning_rate": 3.7031045374008165e-05, "loss": 0.0, "step": 3367 }, { "epoch": 0.78, "learning_rate": 3.702719359063247e-05, "loss": 0.0, "step": 3368 }, { "epoch": 0.78, "learning_rate": 3.702334180725676e-05, "loss": 0.0, "step": 3369 }, { "epoch": 0.78, "learning_rate": 3.701949002388106e-05, "loss": 0.0, "step": 3370 }, { "epoch": 0.78, "learning_rate": 3.701563824050536e-05, "loss": 0.0, "step": 3371 }, { "epoch": 0.78, "learning_rate": 3.701178645712965e-05, "loss": 0.0, "step": 3372 }, { "epoch": 0.78, "learning_rate": 3.7007934673753955e-05, "loss": 0.0, "step": 3373 }, { "epoch": 0.78, "learning_rate": 3.700408289037824e-05, "loss": 0.0, "step": 3374 }, { "epoch": 0.78, "learning_rate": 3.7000231107002545e-05, "loss": 0.0, "step": 3375 }, { "epoch": 0.78, "learning_rate": 3.699637932362684e-05, "loss": 0.0, "step": 3376 }, { "epoch": 0.78, "learning_rate": 3.6992527540251135e-05, "loss": 0.0, "step": 3377 }, { "epoch": 0.78, "learning_rate": 3.698867575687544e-05, "loss": 0.0, "step": 3378 }, { "epoch": 0.78, "learning_rate": 3.698482397349973e-05, "loss": 0.0, "step": 3379 }, { "epoch": 0.78, "learning_rate": 3.698097219012403e-05, "loss": 0.0, "step": 3380 }, { "epoch": 0.78, "learning_rate": 3.697712040674833e-05, "loss": 0.0019, "step": 3381 }, { "epoch": 0.78, "learning_rate": 3.697326862337262e-05, "loss": 0.0, "step": 3382 }, { "epoch": 0.78, "learning_rate": 3.696941683999692e-05, "loss": 0.0, "step": 3383 }, { "epoch": 0.78, "learning_rate": 3.696556505662121e-05, "loss": 0.0, "step": 3384 }, { "epoch": 0.78, "learning_rate": 3.6961713273245515e-05, "loss": 0.0, "step": 3385 }, { "epoch": 0.78, "learning_rate": 3.695786148986981e-05, "loss": 0.0, "step": 3386 }, { "epoch": 0.78, "learning_rate": 3.6954009706494105e-05, "loss": 0.0, "step": 3387 }, { "epoch": 0.78, "learning_rate": 3.6950157923118407e-05, "loss": 0.0, "step": 3388 }, { "epoch": 0.78, "learning_rate": 3.69463061397427e-05, "loss": 0.0, "step": 3389 }, { "epoch": 0.78, "learning_rate": 3.6942454356367e-05, "loss": 0.0, "step": 3390 }, { "epoch": 0.78, "learning_rate": 3.69386025729913e-05, "loss": 0.0, "step": 3391 }, { "epoch": 0.78, "learning_rate": 3.693475078961559e-05, "loss": 0.0048, "step": 3392 }, { "epoch": 0.78, "learning_rate": 3.693089900623989e-05, "loss": 0.0, "step": 3393 }, { "epoch": 0.78, "learning_rate": 3.692704722286418e-05, "loss": 0.0, "step": 3394 }, { "epoch": 0.78, "learning_rate": 3.6923195439488485e-05, "loss": 0.0, "step": 3395 }, { "epoch": 0.78, "learning_rate": 3.691934365611278e-05, "loss": 0.0, "step": 3396 }, { "epoch": 0.79, "learning_rate": 3.691549187273708e-05, "loss": 0.0001, "step": 3397 }, { "epoch": 0.79, "learning_rate": 3.6911640089361376e-05, "loss": 0.0, "step": 3398 }, { "epoch": 0.79, "learning_rate": 3.690778830598567e-05, "loss": 0.0, "step": 3399 }, { "epoch": 0.79, "learning_rate": 3.690393652260997e-05, "loss": 0.0, "step": 3400 }, { "epoch": 0.79, "learning_rate": 3.690008473923427e-05, "loss": 0.0001, "step": 3401 }, { "epoch": 0.79, "learning_rate": 3.689623295585856e-05, "loss": 0.0, "step": 3402 }, { "epoch": 0.79, "learning_rate": 3.689238117248286e-05, "loss": 0.0002, "step": 3403 }, { "epoch": 0.79, "learning_rate": 3.688852938910716e-05, "loss": 0.0, "step": 3404 }, { "epoch": 0.79, "learning_rate": 3.6884677605731455e-05, "loss": 0.0, "step": 3405 }, { "epoch": 0.79, "learning_rate": 3.688082582235575e-05, "loss": 0.0, "step": 3406 }, { "epoch": 0.79, "learning_rate": 3.687697403898005e-05, "loss": 0.0022, "step": 3407 }, { "epoch": 0.79, "learning_rate": 3.6873122255604346e-05, "loss": 0.0, "step": 3408 }, { "epoch": 0.79, "learning_rate": 3.686927047222865e-05, "loss": 0.0029, "step": 3409 }, { "epoch": 0.79, "learning_rate": 3.686541868885294e-05, "loss": 0.0, "step": 3410 }, { "epoch": 0.79, "learning_rate": 3.686156690547724e-05, "loss": 0.0001, "step": 3411 }, { "epoch": 0.79, "learning_rate": 3.685771512210153e-05, "loss": 0.0, "step": 3412 }, { "epoch": 0.79, "learning_rate": 3.685386333872583e-05, "loss": 0.0, "step": 3413 }, { "epoch": 0.79, "learning_rate": 3.685001155535013e-05, "loss": 0.0, "step": 3414 }, { "epoch": 0.79, "learning_rate": 3.6846159771974425e-05, "loss": 0.0, "step": 3415 }, { "epoch": 0.79, "learning_rate": 3.6842307988598726e-05, "loss": 0.0, "step": 3416 }, { "epoch": 0.79, "learning_rate": 3.683845620522302e-05, "loss": 0.0, "step": 3417 }, { "epoch": 0.79, "learning_rate": 3.6834604421847316e-05, "loss": 0.0, "step": 3418 }, { "epoch": 0.79, "learning_rate": 3.683075263847162e-05, "loss": 0.0, "step": 3419 }, { "epoch": 0.79, "learning_rate": 3.682690085509591e-05, "loss": 0.0, "step": 3420 }, { "epoch": 0.79, "learning_rate": 3.682304907172021e-05, "loss": 0.0001, "step": 3421 }, { "epoch": 0.79, "learning_rate": 3.68191972883445e-05, "loss": 0.0, "step": 3422 }, { "epoch": 0.79, "learning_rate": 3.6815345504968805e-05, "loss": 0.0001, "step": 3423 }, { "epoch": 0.79, "learning_rate": 3.68114937215931e-05, "loss": 0.0, "step": 3424 }, { "epoch": 0.79, "learning_rate": 3.6807641938217394e-05, "loss": 0.0002, "step": 3425 }, { "epoch": 0.79, "learning_rate": 3.6803790154841696e-05, "loss": 0.0, "step": 3426 }, { "epoch": 0.79, "learning_rate": 3.679993837146599e-05, "loss": 0.0001, "step": 3427 }, { "epoch": 0.79, "learning_rate": 3.6796086588090286e-05, "loss": 0.0, "step": 3428 }, { "epoch": 0.79, "learning_rate": 3.679223480471459e-05, "loss": 0.0, "step": 3429 }, { "epoch": 0.79, "learning_rate": 3.6788383021338876e-05, "loss": 0.0, "step": 3430 }, { "epoch": 0.79, "learning_rate": 3.678453123796318e-05, "loss": 0.0, "step": 3431 }, { "epoch": 0.79, "learning_rate": 3.678067945458747e-05, "loss": 0.0, "step": 3432 }, { "epoch": 0.79, "learning_rate": 3.6776827671211774e-05, "loss": 0.0, "step": 3433 }, { "epoch": 0.79, "learning_rate": 3.677297588783607e-05, "loss": 0.0, "step": 3434 }, { "epoch": 0.79, "learning_rate": 3.6769124104460364e-05, "loss": 0.0, "step": 3435 }, { "epoch": 0.79, "learning_rate": 3.6765272321084666e-05, "loss": 0.0, "step": 3436 }, { "epoch": 0.79, "learning_rate": 3.676142053770896e-05, "loss": 0.0, "step": 3437 }, { "epoch": 0.79, "learning_rate": 3.675756875433326e-05, "loss": 0.0, "step": 3438 }, { "epoch": 0.79, "learning_rate": 3.675371697095755e-05, "loss": 0.0, "step": 3439 }, { "epoch": 0.8, "learning_rate": 3.674986518758185e-05, "loss": 0.0, "step": 3440 }, { "epoch": 0.8, "learning_rate": 3.674601340420615e-05, "loss": 0.0, "step": 3441 }, { "epoch": 0.8, "learning_rate": 3.674216162083044e-05, "loss": 0.0, "step": 3442 }, { "epoch": 0.8, "learning_rate": 3.6738309837454744e-05, "loss": 0.0, "step": 3443 }, { "epoch": 0.8, "learning_rate": 3.673445805407904e-05, "loss": 0.0, "step": 3444 }, { "epoch": 0.8, "learning_rate": 3.673060627070334e-05, "loss": 0.0, "step": 3445 }, { "epoch": 0.8, "learning_rate": 3.6726754487327636e-05, "loss": 0.0, "step": 3446 }, { "epoch": 0.8, "learning_rate": 3.672290270395193e-05, "loss": 0.0, "step": 3447 }, { "epoch": 0.8, "learning_rate": 3.671905092057623e-05, "loss": 0.0, "step": 3448 }, { "epoch": 0.8, "learning_rate": 3.671519913720052e-05, "loss": 0.0001, "step": 3449 }, { "epoch": 0.8, "learning_rate": 3.671134735382482e-05, "loss": 0.0, "step": 3450 }, { "epoch": 0.8, "learning_rate": 3.670749557044912e-05, "loss": 0.0, "step": 3451 }, { "epoch": 0.8, "learning_rate": 3.670364378707342e-05, "loss": 0.0, "step": 3452 }, { "epoch": 0.8, "learning_rate": 3.6699792003697714e-05, "loss": 0.0, "step": 3453 }, { "epoch": 0.8, "learning_rate": 3.669594022032201e-05, "loss": 0.0003, "step": 3454 }, { "epoch": 0.8, "learning_rate": 3.669208843694631e-05, "loss": 0.0, "step": 3455 }, { "epoch": 0.8, "learning_rate": 3.6688236653570606e-05, "loss": 0.0, "step": 3456 }, { "epoch": 0.8, "learning_rate": 3.668438487019491e-05, "loss": 0.0, "step": 3457 }, { "epoch": 0.8, "learning_rate": 3.6680533086819196e-05, "loss": 0.0, "step": 3458 }, { "epoch": 0.8, "learning_rate": 3.66766813034435e-05, "loss": 0.0, "step": 3459 }, { "epoch": 0.8, "learning_rate": 3.667282952006779e-05, "loss": 0.0, "step": 3460 }, { "epoch": 0.8, "learning_rate": 3.666897773669209e-05, "loss": 0.0, "step": 3461 }, { "epoch": 0.8, "learning_rate": 3.666512595331639e-05, "loss": 0.0, "step": 3462 }, { "epoch": 0.8, "learning_rate": 3.6661274169940684e-05, "loss": 0.0, "step": 3463 }, { "epoch": 0.8, "learning_rate": 3.665742238656498e-05, "loss": 0.0, "step": 3464 }, { "epoch": 0.8, "learning_rate": 3.665357060318928e-05, "loss": 0.0001, "step": 3465 }, { "epoch": 0.8, "learning_rate": 3.6649718819813576e-05, "loss": 0.0001, "step": 3466 }, { "epoch": 0.8, "learning_rate": 3.664586703643787e-05, "loss": 0.0, "step": 3467 }, { "epoch": 0.8, "learning_rate": 3.6642015253062166e-05, "loss": 0.0, "step": 3468 }, { "epoch": 0.8, "learning_rate": 3.663816346968647e-05, "loss": 0.0, "step": 3469 }, { "epoch": 0.8, "learning_rate": 3.663431168631076e-05, "loss": 0.0, "step": 3470 }, { "epoch": 0.8, "learning_rate": 3.663045990293506e-05, "loss": 0.0, "step": 3471 }, { "epoch": 0.8, "learning_rate": 3.662660811955936e-05, "loss": 0.0001, "step": 3472 }, { "epoch": 0.8, "learning_rate": 3.6622756336183654e-05, "loss": 0.0, "step": 3473 }, { "epoch": 0.8, "learning_rate": 3.6618904552807956e-05, "loss": 0.0, "step": 3474 }, { "epoch": 0.8, "learning_rate": 3.661505276943225e-05, "loss": 0.0, "step": 3475 }, { "epoch": 0.8, "learning_rate": 3.6611200986056546e-05, "loss": 0.0, "step": 3476 }, { "epoch": 0.8, "learning_rate": 3.660734920268084e-05, "loss": 0.0, "step": 3477 }, { "epoch": 0.8, "learning_rate": 3.6603497419305136e-05, "loss": 0.0, "step": 3478 }, { "epoch": 0.8, "learning_rate": 3.659964563592944e-05, "loss": 0.0012, "step": 3479 }, { "epoch": 0.8, "learning_rate": 3.659579385255373e-05, "loss": 0.0, "step": 3480 }, { "epoch": 0.8, "learning_rate": 3.6591942069178034e-05, "loss": 0.0, "step": 3481 }, { "epoch": 0.8, "learning_rate": 3.658809028580233e-05, "loss": 0.0, "step": 3482 }, { "epoch": 0.8, "learning_rate": 3.6584238502426624e-05, "loss": 0.0, "step": 3483 }, { "epoch": 0.81, "learning_rate": 3.6580386719050926e-05, "loss": 0.0, "step": 3484 }, { "epoch": 0.81, "learning_rate": 3.657653493567522e-05, "loss": 0.0, "step": 3485 }, { "epoch": 0.81, "learning_rate": 3.6572683152299515e-05, "loss": 0.0, "step": 3486 }, { "epoch": 0.81, "learning_rate": 3.656883136892381e-05, "loss": 0.0, "step": 3487 }, { "epoch": 0.81, "learning_rate": 3.656497958554811e-05, "loss": 0.0, "step": 3488 }, { "epoch": 0.81, "learning_rate": 3.656112780217241e-05, "loss": 0.0, "step": 3489 }, { "epoch": 0.81, "learning_rate": 3.65572760187967e-05, "loss": 0.0, "step": 3490 }, { "epoch": 0.81, "learning_rate": 3.6553424235421004e-05, "loss": 0.0, "step": 3491 }, { "epoch": 0.81, "learning_rate": 3.65495724520453e-05, "loss": 0.0, "step": 3492 }, { "epoch": 0.81, "learning_rate": 3.65457206686696e-05, "loss": 0.0004, "step": 3493 }, { "epoch": 0.81, "learning_rate": 3.6541868885293895e-05, "loss": 0.0, "step": 3494 }, { "epoch": 0.81, "learning_rate": 3.653801710191819e-05, "loss": 0.0, "step": 3495 }, { "epoch": 0.81, "learning_rate": 3.6534165318542485e-05, "loss": 0.0, "step": 3496 }, { "epoch": 0.81, "learning_rate": 3.653031353516678e-05, "loss": 0.0, "step": 3497 }, { "epoch": 0.81, "learning_rate": 3.652646175179108e-05, "loss": 0.0, "step": 3498 }, { "epoch": 0.81, "learning_rate": 3.652260996841538e-05, "loss": 0.0001, "step": 3499 }, { "epoch": 0.81, "learning_rate": 3.651875818503968e-05, "loss": 0.0001, "step": 3500 }, { "epoch": 0.81, "learning_rate": 3.6514906401663974e-05, "loss": 0.0, "step": 3501 }, { "epoch": 0.81, "learning_rate": 3.651105461828827e-05, "loss": 0.0, "step": 3502 }, { "epoch": 0.81, "learning_rate": 3.650720283491257e-05, "loss": 0.0, "step": 3503 }, { "epoch": 0.81, "learning_rate": 3.6503351051536865e-05, "loss": 0.0, "step": 3504 }, { "epoch": 0.81, "learning_rate": 3.649949926816116e-05, "loss": 0.0, "step": 3505 }, { "epoch": 0.81, "learning_rate": 3.6495647484785455e-05, "loss": 0.0, "step": 3506 }, { "epoch": 0.81, "learning_rate": 3.649179570140975e-05, "loss": 0.0, "step": 3507 }, { "epoch": 0.81, "learning_rate": 3.648794391803405e-05, "loss": 0.0, "step": 3508 }, { "epoch": 0.81, "learning_rate": 3.648409213465835e-05, "loss": 0.0, "step": 3509 }, { "epoch": 0.81, "learning_rate": 3.648024035128265e-05, "loss": 0.0, "step": 3510 }, { "epoch": 0.81, "learning_rate": 3.6476388567906944e-05, "loss": 0.0, "step": 3511 }, { "epoch": 0.81, "learning_rate": 3.647253678453124e-05, "loss": 0.0, "step": 3512 }, { "epoch": 0.81, "learning_rate": 3.646868500115554e-05, "loss": 0.0, "step": 3513 }, { "epoch": 0.81, "learning_rate": 3.646483321777983e-05, "loss": 0.0, "step": 3514 }, { "epoch": 0.81, "learning_rate": 3.646098143440413e-05, "loss": 0.0, "step": 3515 }, { "epoch": 0.81, "learning_rate": 3.6457129651028425e-05, "loss": 0.0, "step": 3516 }, { "epoch": 0.81, "learning_rate": 3.645327786765273e-05, "loss": 0.0, "step": 3517 }, { "epoch": 0.81, "learning_rate": 3.644942608427702e-05, "loss": 0.0, "step": 3518 }, { "epoch": 0.81, "learning_rate": 3.644557430090132e-05, "loss": 0.0, "step": 3519 }, { "epoch": 0.81, "learning_rate": 3.644172251752562e-05, "loss": 0.0, "step": 3520 }, { "epoch": 0.81, "learning_rate": 3.6437870734149913e-05, "loss": 0.0, "step": 3521 }, { "epoch": 0.81, "learning_rate": 3.6434018950774215e-05, "loss": 0.0, "step": 3522 }, { "epoch": 0.81, "learning_rate": 3.64301671673985e-05, "loss": 0.0001, "step": 3523 }, { "epoch": 0.81, "learning_rate": 3.6426315384022805e-05, "loss": 0.0, "step": 3524 }, { "epoch": 0.81, "learning_rate": 3.64224636006471e-05, "loss": 0.0, "step": 3525 }, { "epoch": 0.81, "learning_rate": 3.6418611817271395e-05, "loss": 0.0, "step": 3526 }, { "epoch": 0.82, "learning_rate": 3.64147600338957e-05, "loss": 0.0, "step": 3527 }, { "epoch": 0.82, "learning_rate": 3.641090825051999e-05, "loss": 0.0084, "step": 3528 }, { "epoch": 0.82, "learning_rate": 3.6407056467144293e-05, "loss": 0.0, "step": 3529 }, { "epoch": 0.82, "learning_rate": 3.640320468376859e-05, "loss": 0.0, "step": 3530 }, { "epoch": 0.82, "learning_rate": 3.639935290039288e-05, "loss": 0.0, "step": 3531 }, { "epoch": 0.82, "learning_rate": 3.6395501117017185e-05, "loss": 0.0, "step": 3532 }, { "epoch": 0.82, "learning_rate": 3.639164933364147e-05, "loss": 0.0, "step": 3533 }, { "epoch": 0.82, "learning_rate": 3.6387797550265775e-05, "loss": 0.0, "step": 3534 }, { "epoch": 0.82, "learning_rate": 3.638394576689007e-05, "loss": 0.0, "step": 3535 }, { "epoch": 0.82, "learning_rate": 3.638009398351437e-05, "loss": 0.0, "step": 3536 }, { "epoch": 0.82, "learning_rate": 3.637624220013867e-05, "loss": 0.0, "step": 3537 }, { "epoch": 0.82, "learning_rate": 3.637239041676296e-05, "loss": 0.0, "step": 3538 }, { "epoch": 0.82, "learning_rate": 3.636853863338726e-05, "loss": 0.0, "step": 3539 }, { "epoch": 0.82, "learning_rate": 3.636468685001156e-05, "loss": 0.0001, "step": 3540 }, { "epoch": 0.82, "learning_rate": 3.636083506663585e-05, "loss": 0.0, "step": 3541 }, { "epoch": 0.82, "learning_rate": 3.635698328326015e-05, "loss": 0.0, "step": 3542 }, { "epoch": 0.82, "learning_rate": 3.635313149988444e-05, "loss": 0.0, "step": 3543 }, { "epoch": 0.82, "learning_rate": 3.6349279716508745e-05, "loss": 0.0, "step": 3544 }, { "epoch": 0.82, "learning_rate": 3.634542793313304e-05, "loss": 0.0, "step": 3545 }, { "epoch": 0.82, "learning_rate": 3.634157614975734e-05, "loss": 0.0, "step": 3546 }, { "epoch": 0.82, "learning_rate": 3.6337724366381637e-05, "loss": 0.0, "step": 3547 }, { "epoch": 0.82, "learning_rate": 3.633387258300593e-05, "loss": 0.0, "step": 3548 }, { "epoch": 0.82, "learning_rate": 3.633002079963023e-05, "loss": 0.0, "step": 3549 }, { "epoch": 0.82, "learning_rate": 3.632616901625453e-05, "loss": 0.0, "step": 3550 }, { "epoch": 0.82, "learning_rate": 3.632231723287882e-05, "loss": 0.0, "step": 3551 }, { "epoch": 0.82, "learning_rate": 3.631846544950312e-05, "loss": 0.0, "step": 3552 }, { "epoch": 0.82, "learning_rate": 3.631461366612742e-05, "loss": 0.0, "step": 3553 }, { "epoch": 0.82, "learning_rate": 3.6310761882751715e-05, "loss": 0.0, "step": 3554 }, { "epoch": 0.82, "learning_rate": 3.630691009937601e-05, "loss": 0.0, "step": 3555 }, { "epoch": 0.82, "learning_rate": 3.630305831600031e-05, "loss": 0.0, "step": 3556 }, { "epoch": 0.82, "learning_rate": 3.6299206532624606e-05, "loss": 0.0, "step": 3557 }, { "epoch": 0.82, "learning_rate": 3.629535474924891e-05, "loss": 0.0, "step": 3558 }, { "epoch": 0.82, "learning_rate": 3.62915029658732e-05, "loss": 0.0, "step": 3559 }, { "epoch": 0.82, "learning_rate": 3.62876511824975e-05, "loss": 0.0, "step": 3560 }, { "epoch": 0.82, "learning_rate": 3.628379939912179e-05, "loss": 0.0, "step": 3561 }, { "epoch": 0.82, "learning_rate": 3.627994761574609e-05, "loss": 0.0, "step": 3562 }, { "epoch": 0.82, "learning_rate": 3.627609583237039e-05, "loss": 0.0, "step": 3563 }, { "epoch": 0.82, "learning_rate": 3.6272244048994685e-05, "loss": 0.0004, "step": 3564 }, { "epoch": 0.82, "learning_rate": 3.6268392265618986e-05, "loss": 0.0, "step": 3565 }, { "epoch": 0.82, "learning_rate": 3.626454048224328e-05, "loss": 0.0, "step": 3566 }, { "epoch": 0.82, "learning_rate": 3.6260688698867576e-05, "loss": 0.0497, "step": 3567 }, { "epoch": 0.82, "learning_rate": 3.625683691549188e-05, "loss": 0.0, "step": 3568 }, { "epoch": 0.82, "learning_rate": 3.625298513211617e-05, "loss": 0.0, "step": 3569 }, { "epoch": 0.83, "learning_rate": 3.624913334874047e-05, "loss": 0.0001, "step": 3570 }, { "epoch": 0.83, "learning_rate": 3.624528156536476e-05, "loss": 0.0003, "step": 3571 }, { "epoch": 0.83, "learning_rate": 3.6241429781989065e-05, "loss": 0.0, "step": 3572 }, { "epoch": 0.83, "learning_rate": 3.623757799861336e-05, "loss": 0.0002, "step": 3573 }, { "epoch": 0.83, "learning_rate": 3.6233726215237655e-05, "loss": 0.0001, "step": 3574 }, { "epoch": 0.83, "learning_rate": 3.6229874431861956e-05, "loss": 0.0, "step": 3575 }, { "epoch": 0.83, "learning_rate": 3.622602264848625e-05, "loss": 0.0, "step": 3576 }, { "epoch": 0.83, "learning_rate": 3.622217086511055e-05, "loss": 0.0, "step": 3577 }, { "epoch": 0.83, "learning_rate": 3.621831908173485e-05, "loss": 0.0, "step": 3578 }, { "epoch": 0.83, "learning_rate": 3.6214467298359136e-05, "loss": 0.0, "step": 3579 }, { "epoch": 0.83, "learning_rate": 3.621061551498344e-05, "loss": 0.0, "step": 3580 }, { "epoch": 0.83, "learning_rate": 3.620676373160773e-05, "loss": 0.0, "step": 3581 }, { "epoch": 0.83, "learning_rate": 3.6202911948232035e-05, "loss": 0.0, "step": 3582 }, { "epoch": 0.83, "learning_rate": 3.619906016485633e-05, "loss": 0.0, "step": 3583 }, { "epoch": 0.83, "learning_rate": 3.6195208381480624e-05, "loss": 0.0001, "step": 3584 }, { "epoch": 0.83, "learning_rate": 3.6191356598104926e-05, "loss": 0.0, "step": 3585 }, { "epoch": 0.83, "learning_rate": 3.618750481472922e-05, "loss": 0.0, "step": 3586 }, { "epoch": 0.83, "learning_rate": 3.618365303135352e-05, "loss": 0.0, "step": 3587 }, { "epoch": 0.83, "learning_rate": 3.617980124797782e-05, "loss": 0.0, "step": 3588 }, { "epoch": 0.83, "learning_rate": 3.617594946460211e-05, "loss": 0.0, "step": 3589 }, { "epoch": 0.83, "learning_rate": 3.617209768122641e-05, "loss": 0.0, "step": 3590 }, { "epoch": 0.83, "learning_rate": 3.61682458978507e-05, "loss": 0.0004, "step": 3591 }, { "epoch": 0.83, "learning_rate": 3.6164394114475004e-05, "loss": 0.0, "step": 3592 }, { "epoch": 0.83, "learning_rate": 3.61605423310993e-05, "loss": 0.0, "step": 3593 }, { "epoch": 0.83, "learning_rate": 3.61566905477236e-05, "loss": 0.0, "step": 3594 }, { "epoch": 0.83, "learning_rate": 3.6152838764347896e-05, "loss": 0.0003, "step": 3595 }, { "epoch": 0.83, "learning_rate": 3.614898698097219e-05, "loss": 0.0, "step": 3596 }, { "epoch": 0.83, "learning_rate": 3.614513519759649e-05, "loss": 0.0002, "step": 3597 }, { "epoch": 0.83, "learning_rate": 3.614128341422078e-05, "loss": 0.0002, "step": 3598 }, { "epoch": 0.83, "learning_rate": 3.613743163084508e-05, "loss": 0.0, "step": 3599 }, { "epoch": 0.83, "learning_rate": 3.613357984746938e-05, "loss": 0.0, "step": 3600 }, { "epoch": 0.83, "learning_rate": 3.612972806409368e-05, "loss": 0.0, "step": 3601 }, { "epoch": 0.83, "learning_rate": 3.6125876280717974e-05, "loss": 0.0, "step": 3602 }, { "epoch": 0.83, "learning_rate": 3.612202449734227e-05, "loss": 0.0, "step": 3603 }, { "epoch": 0.83, "learning_rate": 3.611817271396657e-05, "loss": 0.0, "step": 3604 }, { "epoch": 0.83, "learning_rate": 3.6114320930590866e-05, "loss": 0.0, "step": 3605 }, { "epoch": 0.83, "learning_rate": 3.611046914721517e-05, "loss": 0.0929, "step": 3606 }, { "epoch": 0.83, "learning_rate": 3.6106617363839456e-05, "loss": 0.0, "step": 3607 }, { "epoch": 0.83, "learning_rate": 3.610276558046376e-05, "loss": 0.0, "step": 3608 }, { "epoch": 0.83, "learning_rate": 3.609891379708805e-05, "loss": 0.0, "step": 3609 }, { "epoch": 0.83, "learning_rate": 3.609506201371235e-05, "loss": 0.0, "step": 3610 }, { "epoch": 0.83, "learning_rate": 3.609121023033665e-05, "loss": 0.0, "step": 3611 }, { "epoch": 0.83, "learning_rate": 3.6087358446960944e-05, "loss": 0.0, "step": 3612 }, { "epoch": 0.83, "learning_rate": 3.6083506663585246e-05, "loss": 0.0, "step": 3613 }, { "epoch": 0.84, "learning_rate": 3.607965488020954e-05, "loss": 0.0, "step": 3614 }, { "epoch": 0.84, "learning_rate": 3.6075803096833836e-05, "loss": 0.0, "step": 3615 }, { "epoch": 0.84, "learning_rate": 3.607195131345814e-05, "loss": 0.0, "step": 3616 }, { "epoch": 0.84, "learning_rate": 3.6068099530082426e-05, "loss": 0.0, "step": 3617 }, { "epoch": 0.84, "learning_rate": 3.606424774670673e-05, "loss": 0.0, "step": 3618 }, { "epoch": 0.84, "learning_rate": 3.606039596333102e-05, "loss": 0.0001, "step": 3619 }, { "epoch": 0.84, "learning_rate": 3.605654417995532e-05, "loss": 0.0, "step": 3620 }, { "epoch": 0.84, "learning_rate": 3.605269239657962e-05, "loss": 0.0, "step": 3621 }, { "epoch": 0.84, "learning_rate": 3.6048840613203914e-05, "loss": 0.0, "step": 3622 }, { "epoch": 0.84, "learning_rate": 3.6044988829828216e-05, "loss": 0.0, "step": 3623 }, { "epoch": 0.84, "learning_rate": 3.604113704645251e-05, "loss": 0.0, "step": 3624 }, { "epoch": 0.84, "learning_rate": 3.6037285263076806e-05, "loss": 0.0, "step": 3625 }, { "epoch": 0.84, "learning_rate": 3.60334334797011e-05, "loss": 0.0001, "step": 3626 }, { "epoch": 0.84, "learning_rate": 3.6029581696325396e-05, "loss": 0.0, "step": 3627 }, { "epoch": 0.84, "learning_rate": 3.60257299129497e-05, "loss": 0.0, "step": 3628 }, { "epoch": 0.84, "learning_rate": 3.602187812957399e-05, "loss": 0.0, "step": 3629 }, { "epoch": 0.84, "learning_rate": 3.6018026346198294e-05, "loss": 0.0, "step": 3630 }, { "epoch": 0.84, "learning_rate": 3.601417456282259e-05, "loss": 0.0, "step": 3631 }, { "epoch": 0.84, "learning_rate": 3.6010322779446884e-05, "loss": 0.0, "step": 3632 }, { "epoch": 0.84, "learning_rate": 3.6006470996071186e-05, "loss": 0.0, "step": 3633 }, { "epoch": 0.84, "learning_rate": 3.600261921269548e-05, "loss": 0.0, "step": 3634 }, { "epoch": 0.84, "learning_rate": 3.5998767429319776e-05, "loss": 0.0, "step": 3635 }, { "epoch": 0.84, "learning_rate": 3.599491564594407e-05, "loss": 0.0002, "step": 3636 }, { "epoch": 0.84, "learning_rate": 3.599106386256837e-05, "loss": 0.0, "step": 3637 }, { "epoch": 0.84, "learning_rate": 3.598721207919267e-05, "loss": 0.0, "step": 3638 }, { "epoch": 0.84, "learning_rate": 3.598336029581696e-05, "loss": 0.0, "step": 3639 }, { "epoch": 0.84, "learning_rate": 3.5979508512441264e-05, "loss": 0.0, "step": 3640 }, { "epoch": 0.84, "learning_rate": 3.597565672906556e-05, "loss": 0.0, "step": 3641 }, { "epoch": 0.84, "learning_rate": 3.597180494568986e-05, "loss": 0.0001, "step": 3642 }, { "epoch": 0.84, "learning_rate": 3.5967953162314156e-05, "loss": 0.0, "step": 3643 }, { "epoch": 0.84, "learning_rate": 3.596410137893845e-05, "loss": 0.0, "step": 3644 }, { "epoch": 0.84, "learning_rate": 3.5960249595562745e-05, "loss": 0.0063, "step": 3645 }, { "epoch": 0.84, "learning_rate": 3.595639781218704e-05, "loss": 0.0, "step": 3646 }, { "epoch": 0.84, "learning_rate": 3.595254602881134e-05, "loss": 0.0, "step": 3647 }, { "epoch": 0.84, "learning_rate": 3.594869424543564e-05, "loss": 0.0, "step": 3648 }, { "epoch": 0.84, "learning_rate": 3.594484246205994e-05, "loss": 0.0, "step": 3649 }, { "epoch": 0.84, "learning_rate": 3.5940990678684234e-05, "loss": 0.0, "step": 3650 }, { "epoch": 0.84, "learning_rate": 3.593713889530853e-05, "loss": 0.0, "step": 3651 }, { "epoch": 0.84, "learning_rate": 3.593328711193283e-05, "loss": 0.0, "step": 3652 }, { "epoch": 0.84, "learning_rate": 3.5929435328557125e-05, "loss": 0.0, "step": 3653 }, { "epoch": 0.84, "learning_rate": 3.592558354518142e-05, "loss": 0.0, "step": 3654 }, { "epoch": 0.84, "learning_rate": 3.5921731761805715e-05, "loss": 0.0, "step": 3655 }, { "epoch": 0.84, "learning_rate": 3.591787997843001e-05, "loss": 0.0, "step": 3656 }, { "epoch": 0.85, "learning_rate": 3.591402819505431e-05, "loss": 0.0, "step": 3657 }, { "epoch": 0.85, "learning_rate": 3.591017641167861e-05, "loss": 0.0, "step": 3658 }, { "epoch": 0.85, "learning_rate": 3.590632462830291e-05, "loss": 0.0, "step": 3659 }, { "epoch": 0.85, "learning_rate": 3.5902472844927204e-05, "loss": 0.0, "step": 3660 }, { "epoch": 0.85, "learning_rate": 3.58986210615515e-05, "loss": 0.0001, "step": 3661 }, { "epoch": 0.85, "learning_rate": 3.58947692781758e-05, "loss": 0.0, "step": 3662 }, { "epoch": 0.85, "learning_rate": 3.589091749480009e-05, "loss": 0.0, "step": 3663 }, { "epoch": 0.85, "learning_rate": 3.588706571142439e-05, "loss": 0.0, "step": 3664 }, { "epoch": 0.85, "learning_rate": 3.5883213928048685e-05, "loss": 0.0, "step": 3665 }, { "epoch": 0.85, "learning_rate": 3.587936214467299e-05, "loss": 0.0002, "step": 3666 }, { "epoch": 0.85, "learning_rate": 3.587551036129728e-05, "loss": 0.0, "step": 3667 }, { "epoch": 0.85, "learning_rate": 3.587165857792158e-05, "loss": 0.0, "step": 3668 }, { "epoch": 0.85, "learning_rate": 3.586780679454588e-05, "loss": 0.0, "step": 3669 }, { "epoch": 0.85, "learning_rate": 3.5863955011170174e-05, "loss": 0.0, "step": 3670 }, { "epoch": 0.85, "learning_rate": 3.5860103227794475e-05, "loss": 0.0, "step": 3671 }, { "epoch": 0.85, "learning_rate": 3.585625144441877e-05, "loss": 0.0001, "step": 3672 }, { "epoch": 0.85, "learning_rate": 3.5852399661043065e-05, "loss": 0.0, "step": 3673 }, { "epoch": 0.85, "learning_rate": 3.584854787766736e-05, "loss": 0.0, "step": 3674 }, { "epoch": 0.85, "learning_rate": 3.5844696094291655e-05, "loss": 0.0, "step": 3675 }, { "epoch": 0.85, "learning_rate": 3.584084431091596e-05, "loss": 0.0, "step": 3676 }, { "epoch": 0.85, "learning_rate": 3.583699252754025e-05, "loss": 0.0, "step": 3677 }, { "epoch": 0.85, "learning_rate": 3.5833140744164554e-05, "loss": 0.0001, "step": 3678 }, { "epoch": 0.85, "learning_rate": 3.582928896078885e-05, "loss": 0.0, "step": 3679 }, { "epoch": 0.85, "learning_rate": 3.5825437177413143e-05, "loss": 0.0, "step": 3680 }, { "epoch": 0.85, "learning_rate": 3.5821585394037445e-05, "loss": 0.0, "step": 3681 }, { "epoch": 0.85, "learning_rate": 3.581773361066173e-05, "loss": 0.0, "step": 3682 }, { "epoch": 0.85, "learning_rate": 3.5813881827286035e-05, "loss": 0.0, "step": 3683 }, { "epoch": 0.85, "learning_rate": 3.581003004391033e-05, "loss": 0.0, "step": 3684 }, { "epoch": 0.85, "learning_rate": 3.580617826053463e-05, "loss": 0.0001, "step": 3685 }, { "epoch": 0.85, "learning_rate": 3.580232647715893e-05, "loss": 0.0, "step": 3686 }, { "epoch": 0.85, "learning_rate": 3.579847469378322e-05, "loss": 0.0, "step": 3687 }, { "epoch": 0.85, "learning_rate": 3.5794622910407523e-05, "loss": 0.0, "step": 3688 }, { "epoch": 0.85, "learning_rate": 3.579077112703182e-05, "loss": 0.0, "step": 3689 }, { "epoch": 0.85, "learning_rate": 3.578691934365612e-05, "loss": 0.0, "step": 3690 }, { "epoch": 0.85, "learning_rate": 3.578306756028041e-05, "loss": 0.0, "step": 3691 }, { "epoch": 0.85, "learning_rate": 3.57792157769047e-05, "loss": 0.0003, "step": 3692 }, { "epoch": 0.85, "learning_rate": 3.5775363993529005e-05, "loss": 0.0, "step": 3693 }, { "epoch": 0.85, "learning_rate": 3.57715122101533e-05, "loss": 0.0, "step": 3694 }, { "epoch": 0.85, "learning_rate": 3.57676604267776e-05, "loss": 0.0, "step": 3695 }, { "epoch": 0.85, "learning_rate": 3.5763808643401897e-05, "loss": 0.0, "step": 3696 }, { "epoch": 0.85, "learning_rate": 3.575995686002619e-05, "loss": 0.0, "step": 3697 }, { "epoch": 0.85, "learning_rate": 3.575610507665049e-05, "loss": 0.0, "step": 3698 }, { "epoch": 0.85, "learning_rate": 3.575225329327479e-05, "loss": 0.0, "step": 3699 }, { "epoch": 0.86, "learning_rate": 3.574840150989909e-05, "loss": 0.0, "step": 3700 }, { "epoch": 0.86, "learning_rate": 3.574454972652338e-05, "loss": 0.0, "step": 3701 }, { "epoch": 0.86, "learning_rate": 3.574069794314768e-05, "loss": 0.0, "step": 3702 }, { "epoch": 0.86, "learning_rate": 3.5736846159771975e-05, "loss": 0.0, "step": 3703 }, { "epoch": 0.86, "learning_rate": 3.573299437639627e-05, "loss": 0.0, "step": 3704 }, { "epoch": 0.86, "learning_rate": 3.572914259302057e-05, "loss": 0.0, "step": 3705 }, { "epoch": 0.86, "learning_rate": 3.5725290809644866e-05, "loss": 0.0, "step": 3706 }, { "epoch": 0.86, "learning_rate": 3.572143902626917e-05, "loss": 0.0, "step": 3707 }, { "epoch": 0.86, "learning_rate": 3.571758724289346e-05, "loss": 0.0, "step": 3708 }, { "epoch": 0.86, "learning_rate": 3.571373545951776e-05, "loss": 0.0007, "step": 3709 }, { "epoch": 0.86, "learning_rate": 3.570988367614205e-05, "loss": 0.0, "step": 3710 }, { "epoch": 0.86, "learning_rate": 3.570603189276635e-05, "loss": 0.0, "step": 3711 }, { "epoch": 0.86, "learning_rate": 3.570218010939065e-05, "loss": 0.0, "step": 3712 }, { "epoch": 0.86, "learning_rate": 3.5698328326014945e-05, "loss": 0.0, "step": 3713 }, { "epoch": 0.86, "learning_rate": 3.5694476542639246e-05, "loss": 0.0, "step": 3714 }, { "epoch": 0.86, "learning_rate": 3.569062475926354e-05, "loss": 0.0, "step": 3715 }, { "epoch": 0.86, "learning_rate": 3.5686772975887836e-05, "loss": 0.0, "step": 3716 }, { "epoch": 0.86, "learning_rate": 3.568292119251214e-05, "loss": 0.0, "step": 3717 }, { "epoch": 0.86, "learning_rate": 3.567906940913643e-05, "loss": 0.0, "step": 3718 }, { "epoch": 0.86, "learning_rate": 3.567521762576073e-05, "loss": 0.0, "step": 3719 }, { "epoch": 0.86, "learning_rate": 3.567136584238502e-05, "loss": 0.0002, "step": 3720 }, { "epoch": 0.86, "learning_rate": 3.5667514059009325e-05, "loss": 0.0, "step": 3721 }, { "epoch": 0.86, "learning_rate": 3.566366227563362e-05, "loss": 0.0, "step": 3722 }, { "epoch": 0.86, "learning_rate": 3.5659810492257915e-05, "loss": 0.0, "step": 3723 }, { "epoch": 0.86, "learning_rate": 3.5655958708882216e-05, "loss": 0.0, "step": 3724 }, { "epoch": 0.86, "learning_rate": 3.565210692550651e-05, "loss": 0.0, "step": 3725 }, { "epoch": 0.86, "learning_rate": 3.564825514213081e-05, "loss": 0.0, "step": 3726 }, { "epoch": 0.86, "learning_rate": 3.564440335875511e-05, "loss": 0.0084, "step": 3727 }, { "epoch": 0.86, "learning_rate": 3.56405515753794e-05, "loss": 0.0, "step": 3728 }, { "epoch": 0.86, "learning_rate": 3.56366997920037e-05, "loss": 0.0, "step": 3729 }, { "epoch": 0.86, "learning_rate": 3.563284800862799e-05, "loss": 0.0, "step": 3730 }, { "epoch": 0.86, "learning_rate": 3.5628996225252295e-05, "loss": 0.0, "step": 3731 }, { "epoch": 0.86, "learning_rate": 3.562514444187659e-05, "loss": 0.0001, "step": 3732 }, { "epoch": 0.86, "learning_rate": 3.5621292658500885e-05, "loss": 0.0, "step": 3733 }, { "epoch": 0.86, "learning_rate": 3.5617440875125186e-05, "loss": 0.0, "step": 3734 }, { "epoch": 0.86, "learning_rate": 3.561358909174948e-05, "loss": 0.0, "step": 3735 }, { "epoch": 0.86, "learning_rate": 3.560973730837378e-05, "loss": 0.0, "step": 3736 }, { "epoch": 0.86, "learning_rate": 3.560588552499808e-05, "loss": 0.0, "step": 3737 }, { "epoch": 0.86, "learning_rate": 3.560203374162237e-05, "loss": 0.0, "step": 3738 }, { "epoch": 0.86, "learning_rate": 3.559818195824667e-05, "loss": 0.0, "step": 3739 }, { "epoch": 0.86, "learning_rate": 3.559433017487096e-05, "loss": 0.0, "step": 3740 }, { "epoch": 0.86, "learning_rate": 3.5590478391495264e-05, "loss": 0.0, "step": 3741 }, { "epoch": 0.86, "learning_rate": 3.558662660811956e-05, "loss": 0.0, "step": 3742 }, { "epoch": 0.87, "learning_rate": 3.558277482474386e-05, "loss": 0.0, "step": 3743 }, { "epoch": 0.87, "learning_rate": 3.5578923041368156e-05, "loss": 0.0, "step": 3744 }, { "epoch": 0.87, "learning_rate": 3.557507125799245e-05, "loss": 0.0, "step": 3745 }, { "epoch": 0.87, "learning_rate": 3.557121947461675e-05, "loss": 0.0, "step": 3746 }, { "epoch": 0.87, "learning_rate": 3.556736769124104e-05, "loss": 0.0179, "step": 3747 }, { "epoch": 0.87, "learning_rate": 3.556351590786534e-05, "loss": 0.0, "step": 3748 }, { "epoch": 0.87, "learning_rate": 3.555966412448964e-05, "loss": 0.0, "step": 3749 }, { "epoch": 0.87, "learning_rate": 3.555581234111394e-05, "loss": 0.0, "step": 3750 }, { "epoch": 0.87, "learning_rate": 3.5551960557738234e-05, "loss": 0.0001, "step": 3751 }, { "epoch": 0.87, "learning_rate": 3.554810877436253e-05, "loss": 0.0, "step": 3752 }, { "epoch": 0.87, "learning_rate": 3.554425699098683e-05, "loss": 0.0001, "step": 3753 }, { "epoch": 0.87, "learning_rate": 3.5540405207611126e-05, "loss": 0.0, "step": 3754 }, { "epoch": 0.87, "learning_rate": 3.553655342423543e-05, "loss": 0.0, "step": 3755 }, { "epoch": 0.87, "learning_rate": 3.553270164085972e-05, "loss": 0.0, "step": 3756 }, { "epoch": 0.87, "learning_rate": 3.552884985748402e-05, "loss": 0.0, "step": 3757 }, { "epoch": 0.87, "learning_rate": 3.552499807410831e-05, "loss": 0.0, "step": 3758 }, { "epoch": 0.87, "learning_rate": 3.552114629073261e-05, "loss": 0.0002, "step": 3759 }, { "epoch": 0.87, "learning_rate": 3.551729450735691e-05, "loss": 0.0, "step": 3760 }, { "epoch": 0.87, "learning_rate": 3.5513442723981204e-05, "loss": 0.0, "step": 3761 }, { "epoch": 0.87, "learning_rate": 3.5509590940605506e-05, "loss": 0.0, "step": 3762 }, { "epoch": 0.87, "learning_rate": 3.55057391572298e-05, "loss": 0.0, "step": 3763 }, { "epoch": 0.87, "learning_rate": 3.5501887373854096e-05, "loss": 0.0, "step": 3764 }, { "epoch": 0.87, "learning_rate": 3.54980355904784e-05, "loss": 0.0, "step": 3765 }, { "epoch": 0.87, "learning_rate": 3.5494183807102686e-05, "loss": 0.0, "step": 3766 }, { "epoch": 0.87, "learning_rate": 3.549033202372699e-05, "loss": 0.0001, "step": 3767 }, { "epoch": 0.87, "learning_rate": 3.548648024035128e-05, "loss": 0.0001, "step": 3768 }, { "epoch": 0.87, "learning_rate": 3.548262845697558e-05, "loss": 0.0, "step": 3769 }, { "epoch": 0.87, "learning_rate": 3.547877667359988e-05, "loss": 0.0, "step": 3770 }, { "epoch": 0.87, "learning_rate": 3.5474924890224174e-05, "loss": 0.0, "step": 3771 }, { "epoch": 0.87, "learning_rate": 3.5471073106848476e-05, "loss": 0.0, "step": 3772 }, { "epoch": 0.87, "learning_rate": 3.546722132347277e-05, "loss": 0.0, "step": 3773 }, { "epoch": 0.87, "learning_rate": 3.5463369540097066e-05, "loss": 0.0, "step": 3774 }, { "epoch": 0.87, "learning_rate": 3.545951775672136e-05, "loss": 0.0, "step": 3775 }, { "epoch": 0.87, "learning_rate": 3.5455665973345656e-05, "loss": 0.0, "step": 3776 }, { "epoch": 0.87, "learning_rate": 3.545181418996996e-05, "loss": 0.0, "step": 3777 }, { "epoch": 0.87, "learning_rate": 3.544796240659425e-05, "loss": 0.0, "step": 3778 }, { "epoch": 0.87, "learning_rate": 3.5444110623218554e-05, "loss": 0.0, "step": 3779 }, { "epoch": 0.87, "learning_rate": 3.544025883984285e-05, "loss": 0.0, "step": 3780 }, { "epoch": 0.87, "learning_rate": 3.5436407056467144e-05, "loss": 0.0, "step": 3781 }, { "epoch": 0.87, "learning_rate": 3.5432555273091446e-05, "loss": 0.0, "step": 3782 }, { "epoch": 0.87, "learning_rate": 3.542870348971574e-05, "loss": 0.0019, "step": 3783 }, { "epoch": 0.87, "learning_rate": 3.542485170634004e-05, "loss": 0.0002, "step": 3784 }, { "epoch": 0.87, "learning_rate": 3.542099992296433e-05, "loss": 0.0, "step": 3785 }, { "epoch": 0.87, "learning_rate": 3.541714813958863e-05, "loss": 0.0004, "step": 3786 }, { "epoch": 0.88, "learning_rate": 3.541329635621293e-05, "loss": 0.0, "step": 3787 }, { "epoch": 0.88, "learning_rate": 3.540944457283722e-05, "loss": 0.0, "step": 3788 }, { "epoch": 0.88, "learning_rate": 3.5405592789461524e-05, "loss": 0.0, "step": 3789 }, { "epoch": 0.88, "learning_rate": 3.540174100608582e-05, "loss": 0.0, "step": 3790 }, { "epoch": 0.88, "learning_rate": 3.539788922271012e-05, "loss": 0.0, "step": 3791 }, { "epoch": 0.88, "learning_rate": 3.5394037439334416e-05, "loss": 0.0, "step": 3792 }, { "epoch": 0.88, "learning_rate": 3.539018565595871e-05, "loss": 0.0, "step": 3793 }, { "epoch": 0.88, "learning_rate": 3.5386333872583006e-05, "loss": 0.0, "step": 3794 }, { "epoch": 0.88, "learning_rate": 3.53824820892073e-05, "loss": 0.0001, "step": 3795 }, { "epoch": 0.88, "learning_rate": 3.53786303058316e-05, "loss": 0.0, "step": 3796 }, { "epoch": 0.88, "learning_rate": 3.53747785224559e-05, "loss": 0.0, "step": 3797 }, { "epoch": 0.88, "learning_rate": 3.53709267390802e-05, "loss": 0.0, "step": 3798 }, { "epoch": 0.88, "learning_rate": 3.5367074955704494e-05, "loss": 0.0004, "step": 3799 }, { "epoch": 0.88, "learning_rate": 3.536322317232879e-05, "loss": 0.0001, "step": 3800 }, { "epoch": 0.88, "learning_rate": 3.535937138895309e-05, "loss": 0.0, "step": 3801 }, { "epoch": 0.88, "learning_rate": 3.5355519605577386e-05, "loss": 0.0, "step": 3802 }, { "epoch": 0.88, "learning_rate": 3.535166782220168e-05, "loss": 0.0, "step": 3803 }, { "epoch": 0.88, "learning_rate": 3.5347816038825975e-05, "loss": 0.0001, "step": 3804 }, { "epoch": 0.88, "learning_rate": 3.534396425545027e-05, "loss": 0.0, "step": 3805 }, { "epoch": 0.88, "learning_rate": 3.534011247207457e-05, "loss": 0.0, "step": 3806 }, { "epoch": 0.88, "learning_rate": 3.533626068869887e-05, "loss": 0.0006, "step": 3807 }, { "epoch": 0.88, "learning_rate": 3.533240890532317e-05, "loss": 0.0, "step": 3808 }, { "epoch": 0.88, "learning_rate": 3.5328557121947464e-05, "loss": 0.0, "step": 3809 }, { "epoch": 0.88, "learning_rate": 3.532470533857176e-05, "loss": 0.0012, "step": 3810 }, { "epoch": 0.88, "learning_rate": 3.532085355519606e-05, "loss": 0.0001, "step": 3811 }, { "epoch": 0.88, "learning_rate": 3.5317001771820355e-05, "loss": 0.0, "step": 3812 }, { "epoch": 0.88, "learning_rate": 3.531314998844465e-05, "loss": 0.0001, "step": 3813 }, { "epoch": 0.88, "learning_rate": 3.5309298205068945e-05, "loss": 0.0, "step": 3814 }, { "epoch": 0.88, "learning_rate": 3.530544642169325e-05, "loss": 0.0, "step": 3815 }, { "epoch": 0.88, "learning_rate": 3.530159463831754e-05, "loss": 0.0, "step": 3816 }, { "epoch": 0.88, "learning_rate": 3.529774285494184e-05, "loss": 0.0, "step": 3817 }, { "epoch": 0.88, "learning_rate": 3.529389107156614e-05, "loss": 0.0001, "step": 3818 }, { "epoch": 0.88, "learning_rate": 3.5290039288190434e-05, "loss": 0.0, "step": 3819 }, { "epoch": 0.88, "learning_rate": 3.5286187504814735e-05, "loss": 0.0, "step": 3820 }, { "epoch": 0.88, "learning_rate": 3.528233572143903e-05, "loss": 0.0, "step": 3821 }, { "epoch": 0.88, "learning_rate": 3.5278483938063325e-05, "loss": 0.0, "step": 3822 }, { "epoch": 0.88, "learning_rate": 3.527463215468762e-05, "loss": 0.0, "step": 3823 }, { "epoch": 0.88, "learning_rate": 3.5270780371311915e-05, "loss": 0.0, "step": 3824 }, { "epoch": 0.88, "learning_rate": 3.526692858793622e-05, "loss": 0.0, "step": 3825 }, { "epoch": 0.88, "learning_rate": 3.526307680456051e-05, "loss": 0.0, "step": 3826 }, { "epoch": 0.88, "learning_rate": 3.5259225021184814e-05, "loss": 0.0, "step": 3827 }, { "epoch": 0.88, "learning_rate": 3.525537323780911e-05, "loss": 0.0, "step": 3828 }, { "epoch": 0.88, "learning_rate": 3.5251521454433404e-05, "loss": 0.0, "step": 3829 }, { "epoch": 0.89, "learning_rate": 3.5247669671057705e-05, "loss": 0.0006, "step": 3830 }, { "epoch": 0.89, "learning_rate": 3.5243817887681993e-05, "loss": 0.0016, "step": 3831 }, { "epoch": 0.89, "learning_rate": 3.5239966104306295e-05, "loss": 0.0, "step": 3832 }, { "epoch": 0.89, "learning_rate": 3.523611432093059e-05, "loss": 0.0, "step": 3833 }, { "epoch": 0.89, "learning_rate": 3.523226253755489e-05, "loss": 0.0, "step": 3834 }, { "epoch": 0.89, "learning_rate": 3.522841075417919e-05, "loss": 0.0, "step": 3835 }, { "epoch": 0.89, "learning_rate": 3.522455897080348e-05, "loss": 0.0, "step": 3836 }, { "epoch": 0.89, "learning_rate": 3.5220707187427783e-05, "loss": 0.0, "step": 3837 }, { "epoch": 0.89, "learning_rate": 3.521685540405208e-05, "loss": 0.0, "step": 3838 }, { "epoch": 0.89, "learning_rate": 3.521300362067638e-05, "loss": 0.0, "step": 3839 }, { "epoch": 0.89, "learning_rate": 3.5209151837300675e-05, "loss": 0.0001, "step": 3840 }, { "epoch": 0.89, "learning_rate": 3.520530005392496e-05, "loss": 0.0, "step": 3841 }, { "epoch": 0.89, "learning_rate": 3.5201448270549265e-05, "loss": 0.001, "step": 3842 }, { "epoch": 0.89, "learning_rate": 3.519759648717356e-05, "loss": 0.0, "step": 3843 }, { "epoch": 0.89, "learning_rate": 3.519374470379786e-05, "loss": 0.0, "step": 3844 }, { "epoch": 0.89, "learning_rate": 3.518989292042216e-05, "loss": 0.0, "step": 3845 }, { "epoch": 0.89, "learning_rate": 3.518604113704645e-05, "loss": 0.0, "step": 3846 }, { "epoch": 0.89, "learning_rate": 3.518218935367075e-05, "loss": 0.0, "step": 3847 }, { "epoch": 0.89, "learning_rate": 3.517833757029505e-05, "loss": 0.0011, "step": 3848 }, { "epoch": 0.89, "learning_rate": 3.517448578691935e-05, "loss": 0.0, "step": 3849 }, { "epoch": 0.89, "learning_rate": 3.517063400354364e-05, "loss": 0.0, "step": 3850 }, { "epoch": 0.89, "learning_rate": 3.516678222016794e-05, "loss": 0.0, "step": 3851 }, { "epoch": 0.89, "learning_rate": 3.5162930436792235e-05, "loss": 0.0001, "step": 3852 }, { "epoch": 0.89, "learning_rate": 3.515907865341653e-05, "loss": 0.0, "step": 3853 }, { "epoch": 0.89, "learning_rate": 3.515522687004083e-05, "loss": 0.0, "step": 3854 }, { "epoch": 0.89, "learning_rate": 3.5151375086665127e-05, "loss": 0.0, "step": 3855 }, { "epoch": 0.89, "learning_rate": 3.514752330328943e-05, "loss": 0.0, "step": 3856 }, { "epoch": 0.89, "learning_rate": 3.514367151991372e-05, "loss": 0.0, "step": 3857 }, { "epoch": 0.89, "learning_rate": 3.513981973653802e-05, "loss": 0.0001, "step": 3858 }, { "epoch": 0.89, "learning_rate": 3.513596795316231e-05, "loss": 0.0, "step": 3859 }, { "epoch": 0.89, "learning_rate": 3.513211616978661e-05, "loss": 0.0001, "step": 3860 }, { "epoch": 0.89, "learning_rate": 3.512826438641091e-05, "loss": 0.0001, "step": 3861 }, { "epoch": 0.89, "learning_rate": 3.5124412603035205e-05, "loss": 0.0, "step": 3862 }, { "epoch": 0.89, "learning_rate": 3.5120560819659507e-05, "loss": 0.0, "step": 3863 }, { "epoch": 0.89, "learning_rate": 3.51167090362838e-05, "loss": 0.0, "step": 3864 }, { "epoch": 0.89, "learning_rate": 3.5112857252908096e-05, "loss": 0.0, "step": 3865 }, { "epoch": 0.89, "learning_rate": 3.51090054695324e-05, "loss": 0.0, "step": 3866 }, { "epoch": 0.89, "learning_rate": 3.510515368615669e-05, "loss": 0.0, "step": 3867 }, { "epoch": 0.89, "learning_rate": 3.5101301902780995e-05, "loss": 0.0, "step": 3868 }, { "epoch": 0.89, "learning_rate": 3.509745011940528e-05, "loss": 0.0005, "step": 3869 }, { "epoch": 0.89, "learning_rate": 3.5093598336029585e-05, "loss": 0.0, "step": 3870 }, { "epoch": 0.89, "learning_rate": 3.508974655265388e-05, "loss": 0.0, "step": 3871 }, { "epoch": 0.89, "learning_rate": 3.5085894769278175e-05, "loss": 0.0, "step": 3872 }, { "epoch": 0.9, "learning_rate": 3.5082042985902476e-05, "loss": 0.0002, "step": 3873 }, { "epoch": 0.9, "learning_rate": 3.507819120252677e-05, "loss": 0.0, "step": 3874 }, { "epoch": 0.9, "learning_rate": 3.507433941915107e-05, "loss": 0.0, "step": 3875 }, { "epoch": 0.9, "learning_rate": 3.507048763577537e-05, "loss": 0.0, "step": 3876 }, { "epoch": 0.9, "learning_rate": 3.506663585239966e-05, "loss": 0.0, "step": 3877 }, { "epoch": 0.9, "learning_rate": 3.506278406902396e-05, "loss": 0.0, "step": 3878 }, { "epoch": 0.9, "learning_rate": 3.505893228564825e-05, "loss": 0.0, "step": 3879 }, { "epoch": 0.9, "learning_rate": 3.5055080502272555e-05, "loss": 0.0, "step": 3880 }, { "epoch": 0.9, "learning_rate": 3.505122871889685e-05, "loss": 0.0, "step": 3881 }, { "epoch": 0.9, "learning_rate": 3.5047376935521145e-05, "loss": 0.0, "step": 3882 }, { "epoch": 0.9, "learning_rate": 3.5043525152145446e-05, "loss": 0.0, "step": 3883 }, { "epoch": 0.9, "learning_rate": 3.503967336876974e-05, "loss": 0.0, "step": 3884 }, { "epoch": 0.9, "learning_rate": 3.503582158539404e-05, "loss": 0.0, "step": 3885 }, { "epoch": 0.9, "learning_rate": 3.503196980201834e-05, "loss": 0.0002, "step": 3886 }, { "epoch": 0.9, "learning_rate": 3.502811801864263e-05, "loss": 0.0, "step": 3887 }, { "epoch": 0.9, "learning_rate": 3.502426623526693e-05, "loss": 0.0, "step": 3888 }, { "epoch": 0.9, "learning_rate": 3.502041445189122e-05, "loss": 0.0, "step": 3889 }, { "epoch": 0.9, "learning_rate": 3.5016562668515525e-05, "loss": 0.0, "step": 3890 }, { "epoch": 0.9, "learning_rate": 3.501271088513982e-05, "loss": 0.0004, "step": 3891 }, { "epoch": 0.9, "learning_rate": 3.500885910176412e-05, "loss": 0.0, "step": 3892 }, { "epoch": 0.9, "learning_rate": 3.5005007318388416e-05, "loss": 0.0001, "step": 3893 }, { "epoch": 0.9, "learning_rate": 3.500115553501271e-05, "loss": 0.0, "step": 3894 }, { "epoch": 0.9, "learning_rate": 3.499730375163701e-05, "loss": 0.0, "step": 3895 }, { "epoch": 0.9, "learning_rate": 3.499345196826131e-05, "loss": 0.0, "step": 3896 }, { "epoch": 0.9, "learning_rate": 3.49896001848856e-05, "loss": 0.0, "step": 3897 }, { "epoch": 0.9, "learning_rate": 3.49857484015099e-05, "loss": 0.0, "step": 3898 }, { "epoch": 0.9, "learning_rate": 3.49818966181342e-05, "loss": 0.0, "step": 3899 }, { "epoch": 0.9, "learning_rate": 3.4978044834758494e-05, "loss": 0.0, "step": 3900 }, { "epoch": 0.9, "learning_rate": 3.497419305138279e-05, "loss": 0.0, "step": 3901 }, { "epoch": 0.9, "learning_rate": 3.497034126800709e-05, "loss": 0.0, "step": 3902 }, { "epoch": 0.9, "learning_rate": 3.4966489484631386e-05, "loss": 0.0, "step": 3903 }, { "epoch": 0.9, "learning_rate": 3.496263770125569e-05, "loss": 0.0, "step": 3904 }, { "epoch": 0.9, "learning_rate": 3.495878591787998e-05, "loss": 0.0, "step": 3905 }, { "epoch": 0.9, "learning_rate": 3.495493413450428e-05, "loss": 0.0006, "step": 3906 }, { "epoch": 0.9, "learning_rate": 3.495108235112857e-05, "loss": 0.0, "step": 3907 }, { "epoch": 0.9, "learning_rate": 3.494723056775287e-05, "loss": 0.0, "step": 3908 }, { "epoch": 0.9, "learning_rate": 3.494337878437717e-05, "loss": 0.0, "step": 3909 }, { "epoch": 0.9, "learning_rate": 3.4939527001001464e-05, "loss": 0.0003, "step": 3910 }, { "epoch": 0.9, "learning_rate": 3.4935675217625766e-05, "loss": 0.0, "step": 3911 }, { "epoch": 0.9, "learning_rate": 3.493182343425006e-05, "loss": 0.0, "step": 3912 }, { "epoch": 0.9, "learning_rate": 3.4927971650874356e-05, "loss": 0.0, "step": 3913 }, { "epoch": 0.9, "learning_rate": 3.492411986749866e-05, "loss": 0.0003, "step": 3914 }, { "epoch": 0.9, "learning_rate": 3.4920268084122946e-05, "loss": 0.0, "step": 3915 }, { "epoch": 0.91, "learning_rate": 3.491641630074725e-05, "loss": 0.0, "step": 3916 }, { "epoch": 0.91, "learning_rate": 3.491256451737154e-05, "loss": 0.0, "step": 3917 }, { "epoch": 0.91, "learning_rate": 3.490871273399584e-05, "loss": 0.0, "step": 3918 }, { "epoch": 0.91, "learning_rate": 3.490486095062014e-05, "loss": 0.0, "step": 3919 }, { "epoch": 0.91, "learning_rate": 3.4901009167244434e-05, "loss": 0.0, "step": 3920 }, { "epoch": 0.91, "learning_rate": 3.4897157383868736e-05, "loss": 0.0006, "step": 3921 }, { "epoch": 0.91, "learning_rate": 3.489330560049303e-05, "loss": 0.0, "step": 3922 }, { "epoch": 0.91, "learning_rate": 3.4889453817117326e-05, "loss": 0.0, "step": 3923 }, { "epoch": 0.91, "learning_rate": 3.488560203374163e-05, "loss": 0.0, "step": 3924 }, { "epoch": 0.91, "learning_rate": 3.4881750250365916e-05, "loss": 0.0, "step": 3925 }, { "epoch": 0.91, "learning_rate": 3.487789846699022e-05, "loss": 0.0002, "step": 3926 }, { "epoch": 0.91, "learning_rate": 3.487404668361451e-05, "loss": 0.0, "step": 3927 }, { "epoch": 0.91, "learning_rate": 3.4870194900238814e-05, "loss": 0.0, "step": 3928 }, { "epoch": 0.91, "learning_rate": 3.486634311686311e-05, "loss": 0.0, "step": 3929 }, { "epoch": 0.91, "learning_rate": 3.4862491333487404e-05, "loss": 0.0, "step": 3930 }, { "epoch": 0.91, "learning_rate": 3.4858639550111706e-05, "loss": 0.0002, "step": 3931 }, { "epoch": 0.91, "learning_rate": 3.4854787766736e-05, "loss": 0.0, "step": 3932 }, { "epoch": 0.91, "learning_rate": 3.48509359833603e-05, "loss": 0.0, "step": 3933 }, { "epoch": 0.91, "learning_rate": 3.484708419998459e-05, "loss": 0.0, "step": 3934 }, { "epoch": 0.91, "learning_rate": 3.484323241660889e-05, "loss": 0.0, "step": 3935 }, { "epoch": 0.91, "learning_rate": 3.483938063323319e-05, "loss": 0.0, "step": 3936 }, { "epoch": 0.91, "learning_rate": 3.483552884985748e-05, "loss": 0.0, "step": 3937 }, { "epoch": 0.91, "learning_rate": 3.4831677066481784e-05, "loss": 0.0, "step": 3938 }, { "epoch": 0.91, "learning_rate": 3.482782528310608e-05, "loss": 0.0, "step": 3939 }, { "epoch": 0.91, "learning_rate": 3.482397349973038e-05, "loss": 0.0, "step": 3940 }, { "epoch": 0.91, "learning_rate": 3.4820121716354676e-05, "loss": 0.0, "step": 3941 }, { "epoch": 0.91, "learning_rate": 3.481626993297897e-05, "loss": 0.0001, "step": 3942 }, { "epoch": 0.91, "learning_rate": 3.4812418149603266e-05, "loss": 0.0, "step": 3943 }, { "epoch": 0.91, "learning_rate": 3.480856636622756e-05, "loss": 0.0, "step": 3944 }, { "epoch": 0.91, "learning_rate": 3.480471458285186e-05, "loss": 0.0, "step": 3945 }, { "epoch": 0.91, "learning_rate": 3.480086279947616e-05, "loss": 0.0001, "step": 3946 }, { "epoch": 0.91, "learning_rate": 3.479701101610046e-05, "loss": 0.0, "step": 3947 }, { "epoch": 0.91, "learning_rate": 3.4793159232724754e-05, "loss": 0.0083, "step": 3948 }, { "epoch": 0.91, "learning_rate": 3.478930744934905e-05, "loss": 0.0, "step": 3949 }, { "epoch": 0.91, "learning_rate": 3.478545566597335e-05, "loss": 0.0, "step": 3950 }, { "epoch": 0.91, "learning_rate": 3.4781603882597646e-05, "loss": 0.0, "step": 3951 }, { "epoch": 0.91, "learning_rate": 3.477775209922195e-05, "loss": 0.0, "step": 3952 }, { "epoch": 0.91, "learning_rate": 3.4773900315846236e-05, "loss": 0.0001, "step": 3953 }, { "epoch": 0.91, "learning_rate": 3.477004853247053e-05, "loss": 0.0, "step": 3954 }, { "epoch": 0.91, "learning_rate": 3.476619674909483e-05, "loss": 0.0, "step": 3955 }, { "epoch": 0.91, "learning_rate": 3.476234496571913e-05, "loss": 0.0002, "step": 3956 }, { "epoch": 0.91, "learning_rate": 3.475849318234343e-05, "loss": 0.0, "step": 3957 }, { "epoch": 0.91, "learning_rate": 3.4754641398967724e-05, "loss": 0.0, "step": 3958 }, { "epoch": 0.91, "learning_rate": 3.475078961559202e-05, "loss": 0.0001, "step": 3959 }, { "epoch": 0.92, "learning_rate": 3.474693783221632e-05, "loss": 0.0, "step": 3960 }, { "epoch": 0.92, "learning_rate": 3.4743086048840615e-05, "loss": 0.0, "step": 3961 }, { "epoch": 0.92, "learning_rate": 3.473923426546491e-05, "loss": 0.0, "step": 3962 }, { "epoch": 0.92, "learning_rate": 3.4735382482089205e-05, "loss": 0.0, "step": 3963 }, { "epoch": 0.92, "learning_rate": 3.473153069871351e-05, "loss": 0.0, "step": 3964 }, { "epoch": 0.92, "learning_rate": 3.47276789153378e-05, "loss": 0.0, "step": 3965 }, { "epoch": 0.92, "learning_rate": 3.47238271319621e-05, "loss": 0.0, "step": 3966 }, { "epoch": 0.92, "learning_rate": 3.47199753485864e-05, "loss": 0.0, "step": 3967 }, { "epoch": 0.92, "learning_rate": 3.4716123565210694e-05, "loss": 0.0, "step": 3968 }, { "epoch": 0.92, "learning_rate": 3.4712271781834995e-05, "loss": 0.0, "step": 3969 }, { "epoch": 0.92, "learning_rate": 3.470841999845929e-05, "loss": 0.0009, "step": 3970 }, { "epoch": 0.92, "learning_rate": 3.4704568215083585e-05, "loss": 0.0, "step": 3971 }, { "epoch": 0.92, "learning_rate": 3.470071643170788e-05, "loss": 0.0001, "step": 3972 }, { "epoch": 0.92, "learning_rate": 3.4696864648332175e-05, "loss": 0.0002, "step": 3973 }, { "epoch": 0.92, "learning_rate": 3.469301286495648e-05, "loss": 0.0, "step": 3974 }, { "epoch": 0.92, "learning_rate": 3.468916108158077e-05, "loss": 0.0, "step": 3975 }, { "epoch": 0.92, "learning_rate": 3.4685309298205074e-05, "loss": 0.0, "step": 3976 }, { "epoch": 0.92, "learning_rate": 3.468145751482937e-05, "loss": 0.0, "step": 3977 }, { "epoch": 0.92, "learning_rate": 3.4677605731453664e-05, "loss": 0.0, "step": 3978 }, { "epoch": 0.92, "learning_rate": 3.4673753948077965e-05, "loss": 0.0, "step": 3979 }, { "epoch": 0.92, "learning_rate": 3.466990216470226e-05, "loss": 0.0, "step": 3980 }, { "epoch": 0.92, "learning_rate": 3.4666050381326555e-05, "loss": 0.0, "step": 3981 }, { "epoch": 0.92, "learning_rate": 3.466219859795085e-05, "loss": 0.0, "step": 3982 }, { "epoch": 0.92, "learning_rate": 3.465834681457515e-05, "loss": 0.0, "step": 3983 }, { "epoch": 0.92, "learning_rate": 3.465449503119945e-05, "loss": 0.0, "step": 3984 }, { "epoch": 0.92, "learning_rate": 3.465064324782374e-05, "loss": 0.0, "step": 3985 }, { "epoch": 0.92, "learning_rate": 3.4646791464448044e-05, "loss": 0.0, "step": 3986 }, { "epoch": 0.92, "learning_rate": 3.464293968107234e-05, "loss": 0.0002, "step": 3987 }, { "epoch": 0.92, "learning_rate": 3.463908789769664e-05, "loss": 0.0, "step": 3988 }, { "epoch": 0.92, "learning_rate": 3.4635236114320935e-05, "loss": 0.0, "step": 3989 }, { "epoch": 0.92, "learning_rate": 3.463138433094523e-05, "loss": 0.0, "step": 3990 }, { "epoch": 0.92, "learning_rate": 3.4627532547569525e-05, "loss": 0.0, "step": 3991 }, { "epoch": 0.92, "learning_rate": 3.462368076419382e-05, "loss": 0.0, "step": 3992 }, { "epoch": 0.92, "learning_rate": 3.461982898081812e-05, "loss": 0.0, "step": 3993 }, { "epoch": 0.92, "learning_rate": 3.461597719744242e-05, "loss": 0.0, "step": 3994 }, { "epoch": 0.92, "learning_rate": 3.461212541406671e-05, "loss": 0.0, "step": 3995 }, { "epoch": 0.92, "learning_rate": 3.4608273630691013e-05, "loss": 0.0001, "step": 3996 }, { "epoch": 0.92, "learning_rate": 3.460442184731531e-05, "loss": 0.0, "step": 3997 }, { "epoch": 0.92, "learning_rate": 3.460057006393961e-05, "loss": 0.0, "step": 3998 }, { "epoch": 0.92, "learning_rate": 3.45967182805639e-05, "loss": 0.0, "step": 3999 }, { "epoch": 0.92, "learning_rate": 3.45928664971882e-05, "loss": 0.0, "step": 4000 }, { "epoch": 0.92, "learning_rate": 3.4589014713812495e-05, "loss": 0.0, "step": 4001 }, { "epoch": 0.92, "learning_rate": 3.458516293043679e-05, "loss": 0.0001, "step": 4002 }, { "epoch": 0.93, "learning_rate": 3.458131114706109e-05, "loss": 0.0, "step": 4003 }, { "epoch": 0.93, "learning_rate": 3.457745936368539e-05, "loss": 0.0, "step": 4004 }, { "epoch": 0.93, "learning_rate": 3.457360758030969e-05, "loss": 0.0, "step": 4005 }, { "epoch": 0.93, "learning_rate": 3.456975579693398e-05, "loss": 0.0, "step": 4006 }, { "epoch": 0.93, "learning_rate": 3.456590401355828e-05, "loss": 0.0001, "step": 4007 }, { "epoch": 0.93, "learning_rate": 3.456205223018258e-05, "loss": 0.0003, "step": 4008 }, { "epoch": 0.93, "learning_rate": 3.455820044680687e-05, "loss": 0.0, "step": 4009 }, { "epoch": 0.93, "learning_rate": 3.455434866343117e-05, "loss": 0.0, "step": 4010 }, { "epoch": 0.93, "learning_rate": 3.4550496880055465e-05, "loss": 0.0, "step": 4011 }, { "epoch": 0.93, "learning_rate": 3.454664509667977e-05, "loss": 0.0001, "step": 4012 }, { "epoch": 0.93, "learning_rate": 3.454279331330406e-05, "loss": 0.0, "step": 4013 }, { "epoch": 0.93, "learning_rate": 3.4538941529928357e-05, "loss": 0.0, "step": 4014 }, { "epoch": 0.93, "learning_rate": 3.453508974655266e-05, "loss": 0.0, "step": 4015 }, { "epoch": 0.93, "learning_rate": 3.453123796317695e-05, "loss": 0.0001, "step": 4016 }, { "epoch": 0.93, "learning_rate": 3.4527386179801255e-05, "loss": 0.0, "step": 4017 }, { "epoch": 0.93, "learning_rate": 3.452353439642554e-05, "loss": 0.0, "step": 4018 }, { "epoch": 0.93, "learning_rate": 3.4519682613049845e-05, "loss": 0.0, "step": 4019 }, { "epoch": 0.93, "learning_rate": 3.451583082967414e-05, "loss": 0.0, "step": 4020 }, { "epoch": 0.93, "learning_rate": 3.4511979046298435e-05, "loss": 0.0, "step": 4021 }, { "epoch": 0.93, "learning_rate": 3.4508127262922737e-05, "loss": 0.0, "step": 4022 }, { "epoch": 0.93, "learning_rate": 3.450427547954703e-05, "loss": 0.0, "step": 4023 }, { "epoch": 0.93, "learning_rate": 3.450042369617133e-05, "loss": 0.0, "step": 4024 }, { "epoch": 0.93, "learning_rate": 3.449657191279563e-05, "loss": 0.0, "step": 4025 }, { "epoch": 0.93, "learning_rate": 3.449272012941992e-05, "loss": 0.0, "step": 4026 }, { "epoch": 0.93, "learning_rate": 3.448886834604422e-05, "loss": 0.0, "step": 4027 }, { "epoch": 0.93, "learning_rate": 3.448501656266851e-05, "loss": 0.0, "step": 4028 }, { "epoch": 0.93, "learning_rate": 3.4481164779292815e-05, "loss": 0.0, "step": 4029 }, { "epoch": 0.93, "learning_rate": 3.447731299591711e-05, "loss": 0.0, "step": 4030 }, { "epoch": 0.93, "learning_rate": 3.4473461212541405e-05, "loss": 0.0, "step": 4031 }, { "epoch": 0.93, "learning_rate": 3.4469609429165706e-05, "loss": 0.0, "step": 4032 }, { "epoch": 0.93, "learning_rate": 3.446575764579e-05, "loss": 0.0, "step": 4033 }, { "epoch": 0.93, "learning_rate": 3.44619058624143e-05, "loss": 0.0, "step": 4034 }, { "epoch": 0.93, "learning_rate": 3.44580540790386e-05, "loss": 0.0, "step": 4035 }, { "epoch": 0.93, "learning_rate": 3.445420229566289e-05, "loss": 0.0, "step": 4036 }, { "epoch": 0.93, "learning_rate": 3.445035051228719e-05, "loss": 0.0, "step": 4037 }, { "epoch": 0.93, "learning_rate": 3.444649872891148e-05, "loss": 0.0, "step": 4038 }, { "epoch": 0.93, "learning_rate": 3.4442646945535785e-05, "loss": 0.0, "step": 4039 }, { "epoch": 0.93, "learning_rate": 3.443879516216008e-05, "loss": 0.0001, "step": 4040 }, { "epoch": 0.93, "learning_rate": 3.443494337878438e-05, "loss": 0.0, "step": 4041 }, { "epoch": 0.93, "learning_rate": 3.4431091595408676e-05, "loss": 0.0001, "step": 4042 }, { "epoch": 0.93, "learning_rate": 3.442723981203297e-05, "loss": 0.0, "step": 4043 }, { "epoch": 0.93, "learning_rate": 3.442338802865727e-05, "loss": 0.0, "step": 4044 }, { "epoch": 0.93, "learning_rate": 3.441953624528157e-05, "loss": 0.0, "step": 4045 }, { "epoch": 0.94, "learning_rate": 3.441568446190586e-05, "loss": 0.0, "step": 4046 }, { "epoch": 0.94, "learning_rate": 3.441183267853016e-05, "loss": 0.0, "step": 4047 }, { "epoch": 0.94, "learning_rate": 3.440798089515446e-05, "loss": 0.0001, "step": 4048 }, { "epoch": 0.94, "learning_rate": 3.4404129111778755e-05, "loss": 0.0002, "step": 4049 }, { "epoch": 0.94, "learning_rate": 3.440027732840305e-05, "loss": 0.0009, "step": 4050 }, { "epoch": 0.94, "learning_rate": 3.439642554502735e-05, "loss": 0.0002, "step": 4051 }, { "epoch": 0.94, "learning_rate": 3.4392573761651646e-05, "loss": 0.0, "step": 4052 }, { "epoch": 0.94, "learning_rate": 3.438872197827595e-05, "loss": 0.0, "step": 4053 }, { "epoch": 0.94, "learning_rate": 3.438487019490024e-05, "loss": 0.0, "step": 4054 }, { "epoch": 0.94, "learning_rate": 3.438101841152454e-05, "loss": 0.0001, "step": 4055 }, { "epoch": 0.94, "learning_rate": 3.437716662814883e-05, "loss": 0.0, "step": 4056 }, { "epoch": 0.94, "learning_rate": 3.437331484477313e-05, "loss": 0.0, "step": 4057 }, { "epoch": 0.94, "learning_rate": 3.436946306139743e-05, "loss": 0.0, "step": 4058 }, { "epoch": 0.94, "learning_rate": 3.4365611278021724e-05, "loss": 0.0, "step": 4059 }, { "epoch": 0.94, "learning_rate": 3.4361759494646026e-05, "loss": 0.0076, "step": 4060 }, { "epoch": 0.94, "learning_rate": 3.435790771127032e-05, "loss": 0.0, "step": 4061 }, { "epoch": 0.94, "learning_rate": 3.4354055927894616e-05, "loss": 0.0, "step": 4062 }, { "epoch": 0.94, "learning_rate": 3.435020414451892e-05, "loss": 0.0, "step": 4063 }, { "epoch": 0.94, "learning_rate": 3.434635236114321e-05, "loss": 0.001, "step": 4064 }, { "epoch": 0.94, "learning_rate": 3.434250057776751e-05, "loss": 0.0, "step": 4065 }, { "epoch": 0.94, "learning_rate": 3.43386487943918e-05, "loss": 0.0, "step": 4066 }, { "epoch": 0.94, "learning_rate": 3.4334797011016104e-05, "loss": 0.0, "step": 4067 }, { "epoch": 0.94, "learning_rate": 3.43309452276404e-05, "loss": 0.0, "step": 4068 }, { "epoch": 0.94, "learning_rate": 3.4327093444264694e-05, "loss": 0.0, "step": 4069 }, { "epoch": 0.94, "learning_rate": 3.4323241660888996e-05, "loss": 0.0, "step": 4070 }, { "epoch": 0.94, "learning_rate": 3.431938987751329e-05, "loss": 0.0, "step": 4071 }, { "epoch": 0.94, "learning_rate": 3.4315538094137586e-05, "loss": 0.0, "step": 4072 }, { "epoch": 0.94, "learning_rate": 3.431168631076189e-05, "loss": 0.0, "step": 4073 }, { "epoch": 0.94, "learning_rate": 3.4307834527386176e-05, "loss": 0.0001, "step": 4074 }, { "epoch": 0.94, "learning_rate": 3.430398274401048e-05, "loss": 0.0, "step": 4075 }, { "epoch": 0.94, "learning_rate": 3.430013096063477e-05, "loss": 0.0, "step": 4076 }, { "epoch": 0.94, "learning_rate": 3.4296279177259074e-05, "loss": 0.0, "step": 4077 }, { "epoch": 0.94, "learning_rate": 3.429242739388337e-05, "loss": 0.0, "step": 4078 }, { "epoch": 0.94, "learning_rate": 3.4288575610507664e-05, "loss": 0.0, "step": 4079 }, { "epoch": 0.94, "learning_rate": 3.4284723827131966e-05, "loss": 0.0, "step": 4080 }, { "epoch": 0.94, "learning_rate": 3.428087204375626e-05, "loss": 0.0, "step": 4081 }, { "epoch": 0.94, "learning_rate": 3.427702026038056e-05, "loss": 0.0, "step": 4082 }, { "epoch": 0.94, "learning_rate": 3.427316847700485e-05, "loss": 0.0, "step": 4083 }, { "epoch": 0.94, "learning_rate": 3.426931669362915e-05, "loss": 0.0, "step": 4084 }, { "epoch": 0.94, "learning_rate": 3.426546491025345e-05, "loss": 0.0, "step": 4085 }, { "epoch": 0.94, "learning_rate": 3.426161312687774e-05, "loss": 0.0, "step": 4086 }, { "epoch": 0.94, "learning_rate": 3.4257761343502044e-05, "loss": 0.0, "step": 4087 }, { "epoch": 0.94, "learning_rate": 3.425390956012634e-05, "loss": 0.0001, "step": 4088 }, { "epoch": 0.94, "learning_rate": 3.425005777675064e-05, "loss": 0.0, "step": 4089 }, { "epoch": 0.95, "learning_rate": 3.4246205993374936e-05, "loss": 0.0, "step": 4090 }, { "epoch": 0.95, "learning_rate": 3.424235420999923e-05, "loss": 0.0, "step": 4091 }, { "epoch": 0.95, "learning_rate": 3.423850242662353e-05, "loss": 0.0, "step": 4092 }, { "epoch": 0.95, "learning_rate": 3.423465064324782e-05, "loss": 0.0, "step": 4093 }, { "epoch": 0.95, "learning_rate": 3.423079885987212e-05, "loss": 0.0, "step": 4094 }, { "epoch": 0.95, "learning_rate": 3.422694707649642e-05, "loss": 0.0, "step": 4095 }, { "epoch": 0.95, "learning_rate": 3.422309529312072e-05, "loss": 0.0, "step": 4096 }, { "epoch": 0.95, "learning_rate": 3.4219243509745014e-05, "loss": 0.0, "step": 4097 }, { "epoch": 0.95, "learning_rate": 3.421539172636931e-05, "loss": 0.0, "step": 4098 }, { "epoch": 0.95, "learning_rate": 3.421153994299361e-05, "loss": 0.0, "step": 4099 }, { "epoch": 0.95, "learning_rate": 3.4207688159617906e-05, "loss": 0.0004, "step": 4100 }, { "epoch": 0.95, "learning_rate": 3.420383637624221e-05, "loss": 0.0, "step": 4101 }, { "epoch": 0.95, "learning_rate": 3.4199984592866496e-05, "loss": 0.0, "step": 4102 }, { "epoch": 0.95, "learning_rate": 3.41961328094908e-05, "loss": 0.0, "step": 4103 }, { "epoch": 0.95, "learning_rate": 3.419228102611509e-05, "loss": 0.0001, "step": 4104 }, { "epoch": 0.95, "learning_rate": 3.418842924273939e-05, "loss": 0.0, "step": 4105 }, { "epoch": 0.95, "learning_rate": 3.418457745936369e-05, "loss": 0.0001, "step": 4106 }, { "epoch": 0.95, "learning_rate": 3.4180725675987984e-05, "loss": 0.0, "step": 4107 }, { "epoch": 0.95, "learning_rate": 3.417687389261228e-05, "loss": 0.0, "step": 4108 }, { "epoch": 0.95, "learning_rate": 3.417302210923658e-05, "loss": 0.0, "step": 4109 }, { "epoch": 0.95, "learning_rate": 3.4169170325860876e-05, "loss": 0.0, "step": 4110 }, { "epoch": 0.95, "learning_rate": 3.416531854248517e-05, "loss": 0.0, "step": 4111 }, { "epoch": 0.95, "learning_rate": 3.4161466759109465e-05, "loss": 0.0, "step": 4112 }, { "epoch": 0.95, "learning_rate": 3.415761497573377e-05, "loss": 0.0, "step": 4113 }, { "epoch": 0.95, "learning_rate": 3.415376319235806e-05, "loss": 0.0, "step": 4114 }, { "epoch": 0.95, "learning_rate": 3.414991140898236e-05, "loss": 0.0, "step": 4115 }, { "epoch": 0.95, "learning_rate": 3.414605962560666e-05, "loss": 0.0, "step": 4116 }, { "epoch": 0.95, "learning_rate": 3.4142207842230954e-05, "loss": 0.0, "step": 4117 }, { "epoch": 0.95, "learning_rate": 3.4138356058855256e-05, "loss": 0.0, "step": 4118 }, { "epoch": 0.95, "learning_rate": 3.413450427547955e-05, "loss": 0.0, "step": 4119 }, { "epoch": 0.95, "learning_rate": 3.4130652492103845e-05, "loss": 0.0, "step": 4120 }, { "epoch": 0.95, "learning_rate": 3.412680070872814e-05, "loss": 0.0, "step": 4121 }, { "epoch": 0.95, "learning_rate": 3.4122948925352435e-05, "loss": 0.0, "step": 4122 }, { "epoch": 0.95, "learning_rate": 3.411909714197674e-05, "loss": 0.0, "step": 4123 }, { "epoch": 0.95, "learning_rate": 3.411524535860103e-05, "loss": 0.0, "step": 4124 }, { "epoch": 0.95, "learning_rate": 3.4111393575225334e-05, "loss": 0.0, "step": 4125 }, { "epoch": 0.95, "learning_rate": 3.410754179184963e-05, "loss": 0.0, "step": 4126 }, { "epoch": 0.95, "learning_rate": 3.4103690008473924e-05, "loss": 0.0603, "step": 4127 }, { "epoch": 0.95, "learning_rate": 3.4099838225098225e-05, "loss": 0.0, "step": 4128 }, { "epoch": 0.95, "learning_rate": 3.409598644172252e-05, "loss": 0.0, "step": 4129 }, { "epoch": 0.95, "learning_rate": 3.4092134658346815e-05, "loss": 0.0, "step": 4130 }, { "epoch": 0.95, "learning_rate": 3.408828287497111e-05, "loss": 0.0, "step": 4131 }, { "epoch": 0.95, "learning_rate": 3.408443109159541e-05, "loss": 0.0, "step": 4132 }, { "epoch": 0.96, "learning_rate": 3.408057930821971e-05, "loss": 0.0, "step": 4133 }, { "epoch": 0.96, "learning_rate": 3.4076727524844e-05, "loss": 0.0001, "step": 4134 }, { "epoch": 0.96, "learning_rate": 3.4072875741468304e-05, "loss": 0.0, "step": 4135 }, { "epoch": 0.96, "learning_rate": 3.40690239580926e-05, "loss": 0.0, "step": 4136 }, { "epoch": 0.96, "learning_rate": 3.40651721747169e-05, "loss": 0.0, "step": 4137 }, { "epoch": 0.96, "learning_rate": 3.4061320391341195e-05, "loss": 0.0, "step": 4138 }, { "epoch": 0.96, "learning_rate": 3.405746860796549e-05, "loss": 0.0, "step": 4139 }, { "epoch": 0.96, "learning_rate": 3.4053616824589785e-05, "loss": 0.0, "step": 4140 }, { "epoch": 0.96, "learning_rate": 3.404976504121408e-05, "loss": 0.0001, "step": 4141 }, { "epoch": 0.96, "learning_rate": 3.404591325783838e-05, "loss": 0.0, "step": 4142 }, { "epoch": 0.96, "learning_rate": 3.404206147446268e-05, "loss": 0.0, "step": 4143 }, { "epoch": 0.96, "learning_rate": 3.403820969108698e-05, "loss": 0.0, "step": 4144 }, { "epoch": 0.96, "learning_rate": 3.4034357907711274e-05, "loss": 0.0, "step": 4145 }, { "epoch": 0.96, "learning_rate": 3.403050612433557e-05, "loss": 0.0001, "step": 4146 }, { "epoch": 0.96, "learning_rate": 3.402665434095987e-05, "loss": 0.0, "step": 4147 }, { "epoch": 0.96, "learning_rate": 3.4022802557584165e-05, "loss": 0.0, "step": 4148 }, { "epoch": 0.96, "learning_rate": 3.401895077420846e-05, "loss": 0.0, "step": 4149 }, { "epoch": 0.96, "learning_rate": 3.4015098990832755e-05, "loss": 0.0, "step": 4150 }, { "epoch": 0.96, "learning_rate": 3.401124720745705e-05, "loss": 0.0, "step": 4151 }, { "epoch": 0.96, "learning_rate": 3.400739542408135e-05, "loss": 0.0, "step": 4152 }, { "epoch": 0.96, "learning_rate": 3.400354364070565e-05, "loss": 0.0, "step": 4153 }, { "epoch": 0.96, "learning_rate": 3.399969185732995e-05, "loss": 0.0, "step": 4154 }, { "epoch": 0.96, "learning_rate": 3.3995840073954243e-05, "loss": 0.0006, "step": 4155 }, { "epoch": 0.96, "learning_rate": 3.399198829057854e-05, "loss": 0.0001, "step": 4156 }, { "epoch": 0.96, "learning_rate": 3.398813650720284e-05, "loss": 0.0, "step": 4157 }, { "epoch": 0.96, "learning_rate": 3.398428472382713e-05, "loss": 0.0, "step": 4158 }, { "epoch": 0.96, "learning_rate": 3.398043294045143e-05, "loss": 0.0, "step": 4159 }, { "epoch": 0.96, "learning_rate": 3.3976581157075725e-05, "loss": 0.0, "step": 4160 }, { "epoch": 0.96, "learning_rate": 3.397272937370003e-05, "loss": 0.0, "step": 4161 }, { "epoch": 0.96, "learning_rate": 3.396887759032432e-05, "loss": 0.0, "step": 4162 }, { "epoch": 0.96, "learning_rate": 3.396502580694862e-05, "loss": 0.0014, "step": 4163 }, { "epoch": 0.96, "learning_rate": 3.396117402357292e-05, "loss": 0.0, "step": 4164 }, { "epoch": 0.96, "learning_rate": 3.395732224019721e-05, "loss": 0.0, "step": 4165 }, { "epoch": 0.96, "learning_rate": 3.3953470456821515e-05, "loss": 0.0001, "step": 4166 }, { "epoch": 0.96, "learning_rate": 3.39496186734458e-05, "loss": 0.0, "step": 4167 }, { "epoch": 0.96, "learning_rate": 3.3945766890070105e-05, "loss": 0.0, "step": 4168 }, { "epoch": 0.96, "learning_rate": 3.39419151066944e-05, "loss": 0.0, "step": 4169 }, { "epoch": 0.96, "learning_rate": 3.3938063323318695e-05, "loss": 0.0, "step": 4170 }, { "epoch": 0.96, "learning_rate": 3.3934211539942997e-05, "loss": 0.0, "step": 4171 }, { "epoch": 0.96, "learning_rate": 3.393035975656729e-05, "loss": 0.0, "step": 4172 }, { "epoch": 0.96, "learning_rate": 3.392650797319159e-05, "loss": 0.0, "step": 4173 }, { "epoch": 0.96, "learning_rate": 3.392265618981589e-05, "loss": 0.0, "step": 4174 }, { "epoch": 0.96, "learning_rate": 3.391880440644018e-05, "loss": 0.0, "step": 4175 }, { "epoch": 0.97, "learning_rate": 3.3914952623064485e-05, "loss": 0.0, "step": 4176 }, { "epoch": 0.97, "learning_rate": 3.391110083968877e-05, "loss": 0.0, "step": 4177 }, { "epoch": 0.97, "learning_rate": 3.3907249056313075e-05, "loss": 0.0, "step": 4178 }, { "epoch": 0.97, "learning_rate": 3.390339727293737e-05, "loss": 0.0, "step": 4179 }, { "epoch": 0.97, "learning_rate": 3.389954548956167e-05, "loss": 0.0, "step": 4180 }, { "epoch": 0.97, "learning_rate": 3.3895693706185966e-05, "loss": 0.0, "step": 4181 }, { "epoch": 0.97, "learning_rate": 3.389184192281026e-05, "loss": 0.0, "step": 4182 }, { "epoch": 0.97, "learning_rate": 3.388799013943456e-05, "loss": 0.0, "step": 4183 }, { "epoch": 0.97, "learning_rate": 3.388413835605886e-05, "loss": 0.0, "step": 4184 }, { "epoch": 0.97, "learning_rate": 3.388028657268315e-05, "loss": 0.0, "step": 4185 }, { "epoch": 0.97, "learning_rate": 3.387643478930745e-05, "loss": 0.0, "step": 4186 }, { "epoch": 0.97, "learning_rate": 3.387258300593174e-05, "loss": 0.0, "step": 4187 }, { "epoch": 0.97, "learning_rate": 3.3868731222556045e-05, "loss": 0.0, "step": 4188 }, { "epoch": 0.97, "learning_rate": 3.386487943918034e-05, "loss": 0.0001, "step": 4189 }, { "epoch": 0.97, "learning_rate": 3.386102765580464e-05, "loss": 0.0, "step": 4190 }, { "epoch": 0.97, "learning_rate": 3.3857175872428936e-05, "loss": 0.0, "step": 4191 }, { "epoch": 0.97, "learning_rate": 3.385332408905323e-05, "loss": 0.0, "step": 4192 }, { "epoch": 0.97, "learning_rate": 3.384947230567753e-05, "loss": 0.0, "step": 4193 }, { "epoch": 0.97, "learning_rate": 3.384562052230183e-05, "loss": 0.0, "step": 4194 }, { "epoch": 0.97, "learning_rate": 3.384176873892612e-05, "loss": 0.0, "step": 4195 }, { "epoch": 0.97, "learning_rate": 3.383791695555042e-05, "loss": 0.0, "step": 4196 }, { "epoch": 0.97, "learning_rate": 3.383406517217472e-05, "loss": 0.0, "step": 4197 }, { "epoch": 0.97, "learning_rate": 3.3830213388799015e-05, "loss": 0.0, "step": 4198 }, { "epoch": 0.97, "learning_rate": 3.382636160542331e-05, "loss": 0.0, "step": 4199 }, { "epoch": 0.97, "learning_rate": 3.382250982204761e-05, "loss": 0.0, "step": 4200 }, { "epoch": 0.97, "learning_rate": 3.3818658038671906e-05, "loss": 0.0, "step": 4201 }, { "epoch": 0.97, "learning_rate": 3.381480625529621e-05, "loss": 0.0, "step": 4202 }, { "epoch": 0.97, "learning_rate": 3.38109544719205e-05, "loss": 0.0, "step": 4203 }, { "epoch": 0.97, "learning_rate": 3.38071026885448e-05, "loss": 0.0, "step": 4204 }, { "epoch": 0.97, "learning_rate": 3.380325090516909e-05, "loss": 0.0011, "step": 4205 }, { "epoch": 0.97, "learning_rate": 3.379939912179339e-05, "loss": 0.0, "step": 4206 }, { "epoch": 0.97, "learning_rate": 3.379554733841769e-05, "loss": 0.0, "step": 4207 }, { "epoch": 0.97, "learning_rate": 3.3791695555041984e-05, "loss": 0.0, "step": 4208 }, { "epoch": 0.97, "learning_rate": 3.3787843771666286e-05, "loss": 0.0, "step": 4209 }, { "epoch": 0.97, "learning_rate": 3.378399198829058e-05, "loss": 0.0, "step": 4210 }, { "epoch": 0.97, "learning_rate": 3.3780140204914876e-05, "loss": 0.0, "step": 4211 }, { "epoch": 0.97, "learning_rate": 3.377628842153918e-05, "loss": 0.0, "step": 4212 }, { "epoch": 0.97, "learning_rate": 3.377243663816347e-05, "loss": 0.0, "step": 4213 }, { "epoch": 0.97, "learning_rate": 3.376858485478777e-05, "loss": 0.0, "step": 4214 }, { "epoch": 0.97, "learning_rate": 3.376473307141206e-05, "loss": 0.0001, "step": 4215 }, { "epoch": 0.97, "learning_rate": 3.3760881288036364e-05, "loss": 0.0, "step": 4216 }, { "epoch": 0.97, "learning_rate": 3.375702950466066e-05, "loss": 0.0001, "step": 4217 }, { "epoch": 0.97, "learning_rate": 3.3753177721284954e-05, "loss": 0.0, "step": 4218 }, { "epoch": 0.98, "learning_rate": 3.3749325937909256e-05, "loss": 0.0, "step": 4219 }, { "epoch": 0.98, "learning_rate": 3.374547415453355e-05, "loss": 0.0, "step": 4220 }, { "epoch": 0.98, "learning_rate": 3.3741622371157846e-05, "loss": 0.0, "step": 4221 }, { "epoch": 0.98, "learning_rate": 3.373777058778215e-05, "loss": 0.0002, "step": 4222 }, { "epoch": 0.98, "learning_rate": 3.3733918804406436e-05, "loss": 0.0, "step": 4223 }, { "epoch": 0.98, "learning_rate": 3.373006702103074e-05, "loss": 0.0, "step": 4224 }, { "epoch": 0.98, "learning_rate": 3.372621523765503e-05, "loss": 0.0, "step": 4225 }, { "epoch": 0.98, "learning_rate": 3.3722363454279334e-05, "loss": 0.0, "step": 4226 }, { "epoch": 0.98, "learning_rate": 3.371851167090363e-05, "loss": 0.0002, "step": 4227 }, { "epoch": 0.98, "learning_rate": 3.3714659887527924e-05, "loss": 0.0, "step": 4228 }, { "epoch": 0.98, "learning_rate": 3.3710808104152226e-05, "loss": 0.0, "step": 4229 }, { "epoch": 0.98, "learning_rate": 3.370695632077652e-05, "loss": 0.0, "step": 4230 }, { "epoch": 0.98, "learning_rate": 3.370310453740082e-05, "loss": 0.0, "step": 4231 }, { "epoch": 0.98, "learning_rate": 3.369925275402512e-05, "loss": 0.0, "step": 4232 }, { "epoch": 0.98, "learning_rate": 3.369540097064941e-05, "loss": 0.0, "step": 4233 }, { "epoch": 0.98, "learning_rate": 3.369154918727371e-05, "loss": 0.0, "step": 4234 }, { "epoch": 0.98, "learning_rate": 3.3687697403898e-05, "loss": 0.0, "step": 4235 }, { "epoch": 0.98, "learning_rate": 3.3683845620522304e-05, "loss": 0.0001, "step": 4236 }, { "epoch": 0.98, "learning_rate": 3.36799938371466e-05, "loss": 0.0, "step": 4237 }, { "epoch": 0.98, "learning_rate": 3.36761420537709e-05, "loss": 0.0, "step": 4238 }, { "epoch": 0.98, "learning_rate": 3.3672290270395196e-05, "loss": 0.0, "step": 4239 }, { "epoch": 0.98, "learning_rate": 3.366843848701949e-05, "loss": 0.0, "step": 4240 }, { "epoch": 0.98, "learning_rate": 3.366458670364379e-05, "loss": 0.0, "step": 4241 }, { "epoch": 0.98, "learning_rate": 3.366073492026808e-05, "loss": 0.0, "step": 4242 }, { "epoch": 0.98, "learning_rate": 3.365688313689238e-05, "loss": 0.0, "step": 4243 }, { "epoch": 0.98, "learning_rate": 3.365303135351668e-05, "loss": 0.0001, "step": 4244 }, { "epoch": 0.98, "learning_rate": 3.364917957014098e-05, "loss": 0.0, "step": 4245 }, { "epoch": 0.98, "learning_rate": 3.3645327786765274e-05, "loss": 0.0005, "step": 4246 }, { "epoch": 0.98, "learning_rate": 3.364147600338957e-05, "loss": 0.0, "step": 4247 }, { "epoch": 0.98, "learning_rate": 3.363762422001387e-05, "loss": 0.0, "step": 4248 }, { "epoch": 0.98, "learning_rate": 3.3633772436638166e-05, "loss": 0.0, "step": 4249 }, { "epoch": 0.98, "learning_rate": 3.362992065326247e-05, "loss": 0.0, "step": 4250 }, { "epoch": 0.98, "learning_rate": 3.3626068869886756e-05, "loss": 0.002, "step": 4251 }, { "epoch": 0.98, "learning_rate": 3.362221708651106e-05, "loss": 0.0, "step": 4252 }, { "epoch": 0.98, "learning_rate": 3.361836530313535e-05, "loss": 0.0, "step": 4253 }, { "epoch": 0.98, "learning_rate": 3.361451351975965e-05, "loss": 0.0, "step": 4254 }, { "epoch": 0.98, "learning_rate": 3.361066173638395e-05, "loss": 0.0, "step": 4255 }, { "epoch": 0.98, "learning_rate": 3.3606809953008244e-05, "loss": 0.0, "step": 4256 }, { "epoch": 0.98, "learning_rate": 3.3602958169632546e-05, "loss": 0.0, "step": 4257 }, { "epoch": 0.98, "learning_rate": 3.359910638625684e-05, "loss": 0.0, "step": 4258 }, { "epoch": 0.98, "learning_rate": 3.3595254602881136e-05, "loss": 0.0002, "step": 4259 }, { "epoch": 0.98, "learning_rate": 3.359140281950544e-05, "loss": 0.0, "step": 4260 }, { "epoch": 0.98, "learning_rate": 3.3587551036129726e-05, "loss": 0.0, "step": 4261 }, { "epoch": 0.98, "learning_rate": 3.358369925275403e-05, "loss": 0.0, "step": 4262 }, { "epoch": 0.99, "learning_rate": 3.357984746937832e-05, "loss": 0.0, "step": 4263 }, { "epoch": 0.99, "learning_rate": 3.357599568600262e-05, "loss": 0.0, "step": 4264 }, { "epoch": 0.99, "learning_rate": 3.357214390262692e-05, "loss": 0.0, "step": 4265 }, { "epoch": 0.99, "learning_rate": 3.3568292119251214e-05, "loss": 0.0001, "step": 4266 }, { "epoch": 0.99, "learning_rate": 3.3564440335875516e-05, "loss": 0.0, "step": 4267 }, { "epoch": 0.99, "learning_rate": 3.356058855249981e-05, "loss": 0.0, "step": 4268 }, { "epoch": 0.99, "learning_rate": 3.3556736769124106e-05, "loss": 0.0, "step": 4269 }, { "epoch": 0.99, "learning_rate": 3.35528849857484e-05, "loss": 0.0, "step": 4270 }, { "epoch": 0.99, "learning_rate": 3.3549033202372695e-05, "loss": 0.0, "step": 4271 }, { "epoch": 0.99, "learning_rate": 3.3545181418997e-05, "loss": 0.0, "step": 4272 }, { "epoch": 0.99, "learning_rate": 3.354132963562129e-05, "loss": 0.0, "step": 4273 }, { "epoch": 0.99, "learning_rate": 3.3537477852245594e-05, "loss": 0.0, "step": 4274 }, { "epoch": 0.99, "learning_rate": 3.353362606886989e-05, "loss": 0.0, "step": 4275 }, { "epoch": 0.99, "learning_rate": 3.3529774285494184e-05, "loss": 0.0, "step": 4276 }, { "epoch": 0.99, "learning_rate": 3.3525922502118486e-05, "loss": 0.0002, "step": 4277 }, { "epoch": 0.99, "learning_rate": 3.352207071874278e-05, "loss": 0.0, "step": 4278 }, { "epoch": 0.99, "learning_rate": 3.3518218935367075e-05, "loss": 0.0, "step": 4279 }, { "epoch": 0.99, "learning_rate": 3.351436715199137e-05, "loss": 0.0, "step": 4280 }, { "epoch": 0.99, "learning_rate": 3.351051536861567e-05, "loss": 0.0, "step": 4281 }, { "epoch": 0.99, "learning_rate": 3.350666358523997e-05, "loss": 0.0, "step": 4282 }, { "epoch": 0.99, "learning_rate": 3.350281180186426e-05, "loss": 0.0, "step": 4283 }, { "epoch": 0.99, "learning_rate": 3.3498960018488564e-05, "loss": 0.0, "step": 4284 }, { "epoch": 0.99, "learning_rate": 3.349510823511286e-05, "loss": 0.0, "step": 4285 }, { "epoch": 0.99, "learning_rate": 3.349125645173716e-05, "loss": 0.0001, "step": 4286 }, { "epoch": 0.99, "learning_rate": 3.3487404668361455e-05, "loss": 0.0, "step": 4287 }, { "epoch": 0.99, "learning_rate": 3.348355288498575e-05, "loss": 0.0, "step": 4288 }, { "epoch": 0.99, "learning_rate": 3.3479701101610045e-05, "loss": 0.0, "step": 4289 }, { "epoch": 0.99, "learning_rate": 3.347584931823434e-05, "loss": 0.0, "step": 4290 }, { "epoch": 0.99, "learning_rate": 3.347199753485864e-05, "loss": 0.0, "step": 4291 }, { "epoch": 0.99, "learning_rate": 3.346814575148294e-05, "loss": 0.0, "step": 4292 }, { "epoch": 0.99, "learning_rate": 3.346429396810724e-05, "loss": 0.0, "step": 4293 }, { "epoch": 0.99, "learning_rate": 3.3460442184731534e-05, "loss": 0.0, "step": 4294 }, { "epoch": 0.99, "learning_rate": 3.345659040135583e-05, "loss": 0.0, "step": 4295 }, { "epoch": 0.99, "learning_rate": 3.345273861798013e-05, "loss": 0.0, "step": 4296 }, { "epoch": 0.99, "learning_rate": 3.3448886834604425e-05, "loss": 0.0001, "step": 4297 }, { "epoch": 0.99, "learning_rate": 3.344503505122872e-05, "loss": 0.0, "step": 4298 }, { "epoch": 0.99, "learning_rate": 3.3441183267853015e-05, "loss": 0.0, "step": 4299 }, { "epoch": 0.99, "learning_rate": 3.343733148447731e-05, "loss": 0.0, "step": 4300 }, { "epoch": 0.99, "learning_rate": 3.343347970110161e-05, "loss": 0.0, "step": 4301 }, { "epoch": 0.99, "learning_rate": 3.342962791772591e-05, "loss": 0.0, "step": 4302 }, { "epoch": 0.99, "learning_rate": 3.342577613435021e-05, "loss": 0.0, "step": 4303 }, { "epoch": 0.99, "learning_rate": 3.3421924350974504e-05, "loss": 0.0, "step": 4304 }, { "epoch": 0.99, "learning_rate": 3.34180725675988e-05, "loss": 0.0, "step": 4305 }, { "epoch": 1.0, "learning_rate": 3.34142207842231e-05, "loss": 0.0, "step": 4306 }, { "epoch": 1.0, "learning_rate": 3.341036900084739e-05, "loss": 0.0, "step": 4307 }, { "epoch": 1.0, "learning_rate": 3.340651721747169e-05, "loss": 0.0, "step": 4308 }, { "epoch": 1.0, "learning_rate": 3.3402665434095985e-05, "loss": 0.0, "step": 4309 }, { "epoch": 1.0, "learning_rate": 3.339881365072029e-05, "loss": 0.0, "step": 4310 }, { "epoch": 1.0, "learning_rate": 3.339496186734458e-05, "loss": 0.0001, "step": 4311 }, { "epoch": 1.0, "learning_rate": 3.339111008396888e-05, "loss": 0.0001, "step": 4312 }, { "epoch": 1.0, "learning_rate": 3.338725830059318e-05, "loss": 0.0, "step": 4313 }, { "epoch": 1.0, "learning_rate": 3.3383406517217473e-05, "loss": 0.0, "step": 4314 }, { "epoch": 1.0, "learning_rate": 3.3379554733841775e-05, "loss": 0.0, "step": 4315 }, { "epoch": 1.0, "learning_rate": 3.337570295046607e-05, "loss": 0.0001, "step": 4316 }, { "epoch": 1.0, "learning_rate": 3.3371851167090365e-05, "loss": 0.0, "step": 4317 }, { "epoch": 1.0, "learning_rate": 3.336799938371466e-05, "loss": 0.0, "step": 4318 }, { "epoch": 1.0, "learning_rate": 3.3364147600338955e-05, "loss": 0.0, "step": 4319 }, { "epoch": 1.0, "learning_rate": 3.336029581696326e-05, "loss": 0.0, "step": 4320 }, { "epoch": 1.0, "learning_rate": 3.335644403358755e-05, "loss": 0.0, "step": 4321 }, { "epoch": 1.0, "learning_rate": 3.335259225021185e-05, "loss": 0.0, "step": 4322 }, { "epoch": 1.0, "learning_rate": 3.334874046683615e-05, "loss": 0.0, "step": 4323 }, { "epoch": 1.0, "learning_rate": 3.334488868346044e-05, "loss": 0.0, "step": 4324 }, { "epoch": 1.0, "learning_rate": 3.3341036900084745e-05, "loss": 0.0, "step": 4325 }, { "epoch": 1.0, "learning_rate": 3.333718511670903e-05, "loss": 0.0, "step": 4326 }, { "epoch": 1.0, "learning_rate": 3.3333333333333335e-05, "loss": 0.0, "step": 4327 }, { "epoch": 1.0, "learning_rate": 3.332948154995763e-05, "loss": 0.0, "step": 4328 }, { "epoch": 1.0, "learning_rate": 3.332562976658193e-05, "loss": 0.0, "step": 4329 }, { "epoch": 1.0, "learning_rate": 3.3321777983206227e-05, "loss": 0.0, "step": 4330 }, { "epoch": 1.0, "learning_rate": 3.331792619983052e-05, "loss": 0.0, "step": 4331 }, { "epoch": 1.0, "learning_rate": 3.331407441645482e-05, "loss": 0.0, "step": 4332 }, { "epoch": 1.0, "learning_rate": 3.331022263307912e-05, "loss": 0.0, "step": 4333 }, { "epoch": 1.0, "learning_rate": 3.330637084970342e-05, "loss": 0.0, "step": 4334 }, { "epoch": 1.0, "learning_rate": 3.330251906632771e-05, "loss": 0.0, "step": 4335 }, { "epoch": 1.0, "learning_rate": 3.3298667282952e-05, "loss": 0.0, "step": 4336 }, { "epoch": 1.0, "learning_rate": 3.3294815499576305e-05, "loss": 0.0, "step": 4337 }, { "epoch": 1.0, "learning_rate": 3.32909637162006e-05, "loss": 0.0, "step": 4338 }, { "epoch": 1.0, "learning_rate": 3.32871119328249e-05, "loss": 0.0, "step": 4339 }, { "epoch": 1.0, "learning_rate": 3.3283260149449196e-05, "loss": 0.0, "step": 4340 }, { "epoch": 1.0, "learning_rate": 3.327940836607349e-05, "loss": 0.0, "step": 4341 }, { "epoch": 1.0, "learning_rate": 3.327555658269779e-05, "loss": 0.0, "step": 4342 }, { "epoch": 1.0, "learning_rate": 3.327170479932209e-05, "loss": 0.0001, "step": 4343 }, { "epoch": 1.0, "learning_rate": 3.326785301594639e-05, "loss": 0.0, "step": 4344 }, { "epoch": 1.0, "learning_rate": 3.326400123257068e-05, "loss": 0.0, "step": 4345 }, { "epoch": 1.0, "learning_rate": 3.326014944919498e-05, "loss": 0.0, "step": 4346 }, { "epoch": 1.0, "learning_rate": 3.3256297665819275e-05, "loss": 0.0, "step": 4347 }, { "epoch": 1.0, "learning_rate": 3.325244588244357e-05, "loss": 0.0002, "step": 4348 }, { "epoch": 1.01, "learning_rate": 3.324859409906787e-05, "loss": 0.0, "step": 4349 }, { "epoch": 1.01, "learning_rate": 3.3244742315692166e-05, "loss": 0.0, "step": 4350 }, { "epoch": 1.01, "learning_rate": 3.324089053231647e-05, "loss": 0.0, "step": 4351 }, { "epoch": 1.01, "learning_rate": 3.323703874894076e-05, "loss": 0.0, "step": 4352 }, { "epoch": 1.01, "learning_rate": 3.323318696556506e-05, "loss": 0.0, "step": 4353 }, { "epoch": 1.01, "learning_rate": 3.322933518218935e-05, "loss": 0.0, "step": 4354 }, { "epoch": 1.01, "learning_rate": 3.322548339881365e-05, "loss": 0.0003, "step": 4355 }, { "epoch": 1.01, "learning_rate": 3.322163161543795e-05, "loss": 0.0, "step": 4356 }, { "epoch": 1.01, "learning_rate": 3.3217779832062245e-05, "loss": 0.0, "step": 4357 }, { "epoch": 1.01, "learning_rate": 3.3213928048686546e-05, "loss": 0.0001, "step": 4358 }, { "epoch": 1.01, "learning_rate": 3.321007626531084e-05, "loss": 0.0, "step": 4359 }, { "epoch": 1.01, "learning_rate": 3.3206224481935136e-05, "loss": 0.0, "step": 4360 }, { "epoch": 1.01, "learning_rate": 3.320237269855944e-05, "loss": 0.0001, "step": 4361 }, { "epoch": 1.01, "learning_rate": 3.319852091518373e-05, "loss": 0.0, "step": 4362 }, { "epoch": 1.01, "learning_rate": 3.319466913180803e-05, "loss": 0.0, "step": 4363 }, { "epoch": 1.01, "learning_rate": 3.319081734843232e-05, "loss": 0.0, "step": 4364 }, { "epoch": 1.01, "learning_rate": 3.3186965565056625e-05, "loss": 0.0, "step": 4365 }, { "epoch": 1.01, "learning_rate": 3.318311378168092e-05, "loss": 0.0, "step": 4366 }, { "epoch": 1.01, "learning_rate": 3.3179261998305214e-05, "loss": 0.0001, "step": 4367 }, { "epoch": 1.01, "learning_rate": 3.3175410214929516e-05, "loss": 0.0, "step": 4368 }, { "epoch": 1.01, "learning_rate": 3.317155843155381e-05, "loss": 0.0, "step": 4369 }, { "epoch": 1.01, "learning_rate": 3.316770664817811e-05, "loss": 0.0, "step": 4370 }, { "epoch": 1.01, "learning_rate": 3.316385486480241e-05, "loss": 0.0, "step": 4371 }, { "epoch": 1.01, "learning_rate": 3.31600030814267e-05, "loss": 0.0, "step": 4372 }, { "epoch": 1.01, "learning_rate": 3.3156151298051e-05, "loss": 0.0, "step": 4373 }, { "epoch": 1.01, "learning_rate": 3.315229951467529e-05, "loss": 0.0, "step": 4374 }, { "epoch": 1.01, "learning_rate": 3.3148447731299594e-05, "loss": 0.0, "step": 4375 }, { "epoch": 1.01, "learning_rate": 3.314459594792389e-05, "loss": 0.0, "step": 4376 }, { "epoch": 1.01, "learning_rate": 3.3140744164548184e-05, "loss": 0.0, "step": 4377 }, { "epoch": 1.01, "learning_rate": 3.3136892381172486e-05, "loss": 0.0009, "step": 4378 }, { "epoch": 1.01, "learning_rate": 3.313304059779678e-05, "loss": 0.0, "step": 4379 }, { "epoch": 1.01, "learning_rate": 3.312918881442108e-05, "loss": 0.0, "step": 4380 }, { "epoch": 1.01, "learning_rate": 3.312533703104538e-05, "loss": 0.0027, "step": 4381 }, { "epoch": 1.01, "learning_rate": 3.312148524766967e-05, "loss": 0.0, "step": 4382 }, { "epoch": 1.01, "learning_rate": 3.311763346429397e-05, "loss": 0.0, "step": 4383 }, { "epoch": 1.01, "learning_rate": 3.311378168091826e-05, "loss": 0.0, "step": 4384 }, { "epoch": 1.01, "learning_rate": 3.3109929897542564e-05, "loss": 0.0, "step": 4385 }, { "epoch": 1.01, "learning_rate": 3.310607811416686e-05, "loss": 0.0, "step": 4386 }, { "epoch": 1.01, "learning_rate": 3.310222633079116e-05, "loss": 0.0, "step": 4387 }, { "epoch": 1.01, "learning_rate": 3.3098374547415456e-05, "loss": 0.0004, "step": 4388 }, { "epoch": 1.01, "learning_rate": 3.309452276403975e-05, "loss": 0.0, "step": 4389 }, { "epoch": 1.01, "learning_rate": 3.309067098066405e-05, "loss": 0.0, "step": 4390 }, { "epoch": 1.01, "learning_rate": 3.308681919728834e-05, "loss": 0.0, "step": 4391 }, { "epoch": 1.02, "learning_rate": 3.308296741391264e-05, "loss": 0.0, "step": 4392 }, { "epoch": 1.02, "learning_rate": 3.307911563053694e-05, "loss": 0.0, "step": 4393 }, { "epoch": 1.02, "learning_rate": 3.307526384716124e-05, "loss": 0.0, "step": 4394 }, { "epoch": 1.02, "learning_rate": 3.3071412063785534e-05, "loss": 0.0, "step": 4395 }, { "epoch": 1.02, "learning_rate": 3.306756028040983e-05, "loss": 0.0, "step": 4396 }, { "epoch": 1.02, "learning_rate": 3.306370849703413e-05, "loss": 0.0, "step": 4397 }, { "epoch": 1.02, "learning_rate": 3.3059856713658426e-05, "loss": 0.0, "step": 4398 }, { "epoch": 1.02, "learning_rate": 3.305600493028273e-05, "loss": 0.0, "step": 4399 }, { "epoch": 1.02, "learning_rate": 3.305215314690702e-05, "loss": 0.0003, "step": 4400 }, { "epoch": 1.02, "learning_rate": 3.304830136353132e-05, "loss": 0.0, "step": 4401 }, { "epoch": 1.02, "learning_rate": 3.304444958015561e-05, "loss": 0.0, "step": 4402 }, { "epoch": 1.02, "learning_rate": 3.304059779677991e-05, "loss": 0.0, "step": 4403 }, { "epoch": 1.02, "learning_rate": 3.303674601340421e-05, "loss": 0.0001, "step": 4404 }, { "epoch": 1.02, "learning_rate": 3.3032894230028504e-05, "loss": 0.0, "step": 4405 }, { "epoch": 1.02, "learning_rate": 3.3029042446652806e-05, "loss": 0.0, "step": 4406 }, { "epoch": 1.02, "learning_rate": 3.30251906632771e-05, "loss": 0.0, "step": 4407 }, { "epoch": 1.02, "learning_rate": 3.3021338879901396e-05, "loss": 0.0, "step": 4408 }, { "epoch": 1.02, "learning_rate": 3.30174870965257e-05, "loss": 0.0, "step": 4409 }, { "epoch": 1.02, "learning_rate": 3.3013635313149986e-05, "loss": 0.0, "step": 4410 }, { "epoch": 1.02, "learning_rate": 3.300978352977429e-05, "loss": 0.0, "step": 4411 }, { "epoch": 1.02, "learning_rate": 3.300593174639858e-05, "loss": 0.0002, "step": 4412 }, { "epoch": 1.02, "learning_rate": 3.300207996302288e-05, "loss": 0.0, "step": 4413 }, { "epoch": 1.02, "learning_rate": 3.299822817964718e-05, "loss": 0.0, "step": 4414 }, { "epoch": 1.02, "learning_rate": 3.2994376396271474e-05, "loss": 0.0, "step": 4415 }, { "epoch": 1.02, "learning_rate": 3.2990524612895776e-05, "loss": 0.0004, "step": 4416 }, { "epoch": 1.02, "learning_rate": 3.298667282952007e-05, "loss": 0.0001, "step": 4417 }, { "epoch": 1.02, "learning_rate": 3.2982821046144366e-05, "loss": 0.0, "step": 4418 }, { "epoch": 1.02, "learning_rate": 3.297896926276866e-05, "loss": 0.0, "step": 4419 }, { "epoch": 1.02, "learning_rate": 3.2975117479392956e-05, "loss": 0.0, "step": 4420 }, { "epoch": 1.02, "learning_rate": 3.297126569601726e-05, "loss": 0.0, "step": 4421 }, { "epoch": 1.02, "learning_rate": 3.296741391264155e-05, "loss": 0.0, "step": 4422 }, { "epoch": 1.02, "learning_rate": 3.2963562129265854e-05, "loss": 0.0, "step": 4423 }, { "epoch": 1.02, "learning_rate": 3.295971034589015e-05, "loss": 0.0, "step": 4424 }, { "epoch": 1.02, "learning_rate": 3.2955858562514444e-05, "loss": 0.0, "step": 4425 }, { "epoch": 1.02, "learning_rate": 3.2952006779138746e-05, "loss": 0.0, "step": 4426 }, { "epoch": 1.02, "learning_rate": 3.294815499576304e-05, "loss": 0.0002, "step": 4427 }, { "epoch": 1.02, "learning_rate": 3.294430321238734e-05, "loss": 0.0012, "step": 4428 }, { "epoch": 1.02, "learning_rate": 3.294045142901163e-05, "loss": 0.0, "step": 4429 }, { "epoch": 1.02, "learning_rate": 3.293659964563593e-05, "loss": 0.0, "step": 4430 }, { "epoch": 1.02, "learning_rate": 3.293274786226023e-05, "loss": 0.0, "step": 4431 }, { "epoch": 1.02, "learning_rate": 3.292889607888452e-05, "loss": 0.0, "step": 4432 }, { "epoch": 1.02, "learning_rate": 3.2925044295508824e-05, "loss": 0.0, "step": 4433 }, { "epoch": 1.02, "learning_rate": 3.292119251213312e-05, "loss": 0.0, "step": 4434 }, { "epoch": 1.02, "learning_rate": 3.291734072875742e-05, "loss": 0.0, "step": 4435 }, { "epoch": 1.03, "learning_rate": 3.2913488945381715e-05, "loss": 0.0, "step": 4436 }, { "epoch": 1.03, "learning_rate": 3.290963716200601e-05, "loss": 0.0, "step": 4437 }, { "epoch": 1.03, "learning_rate": 3.2905785378630305e-05, "loss": 0.0, "step": 4438 }, { "epoch": 1.03, "learning_rate": 3.29019335952546e-05, "loss": 0.0, "step": 4439 }, { "epoch": 1.03, "learning_rate": 3.28980818118789e-05, "loss": 0.0, "step": 4440 }, { "epoch": 1.03, "learning_rate": 3.28942300285032e-05, "loss": 0.0, "step": 4441 }, { "epoch": 1.03, "learning_rate": 3.28903782451275e-05, "loss": 0.0002, "step": 4442 }, { "epoch": 1.03, "learning_rate": 3.2886526461751794e-05, "loss": 0.0, "step": 4443 }, { "epoch": 1.03, "learning_rate": 3.288267467837609e-05, "loss": 0.0, "step": 4444 }, { "epoch": 1.03, "learning_rate": 3.287882289500039e-05, "loss": 0.0, "step": 4445 }, { "epoch": 1.03, "learning_rate": 3.2874971111624685e-05, "loss": 0.0, "step": 4446 }, { "epoch": 1.03, "learning_rate": 3.287111932824898e-05, "loss": 0.0, "step": 4447 }, { "epoch": 1.03, "learning_rate": 3.2867267544873275e-05, "loss": 0.0, "step": 4448 }, { "epoch": 1.03, "learning_rate": 3.286341576149757e-05, "loss": 0.0, "step": 4449 }, { "epoch": 1.03, "learning_rate": 3.285956397812187e-05, "loss": 0.0001, "step": 4450 }, { "epoch": 1.03, "learning_rate": 3.285571219474617e-05, "loss": 0.0, "step": 4451 }, { "epoch": 1.03, "learning_rate": 3.285186041137047e-05, "loss": 0.0002, "step": 4452 }, { "epoch": 1.03, "learning_rate": 3.2848008627994764e-05, "loss": 0.0, "step": 4453 }, { "epoch": 1.03, "learning_rate": 3.284415684461906e-05, "loss": 0.0, "step": 4454 }, { "epoch": 1.03, "learning_rate": 3.284030506124336e-05, "loss": 0.0, "step": 4455 }, { "epoch": 1.03, "learning_rate": 3.2836453277867655e-05, "loss": 0.0, "step": 4456 }, { "epoch": 1.03, "learning_rate": 3.283260149449195e-05, "loss": 0.0, "step": 4457 }, { "epoch": 1.03, "learning_rate": 3.2828749711116245e-05, "loss": 0.0, "step": 4458 }, { "epoch": 1.03, "learning_rate": 3.282489792774055e-05, "loss": 0.0, "step": 4459 }, { "epoch": 1.03, "learning_rate": 3.282104614436484e-05, "loss": 0.0, "step": 4460 }, { "epoch": 1.03, "learning_rate": 3.281719436098914e-05, "loss": 0.0, "step": 4461 }, { "epoch": 1.03, "learning_rate": 3.281334257761344e-05, "loss": 0.0001, "step": 4462 }, { "epoch": 1.03, "learning_rate": 3.2809490794237733e-05, "loss": 0.0, "step": 4463 }, { "epoch": 1.03, "learning_rate": 3.2805639010862035e-05, "loss": 0.0, "step": 4464 }, { "epoch": 1.03, "learning_rate": 3.280178722748633e-05, "loss": 0.0, "step": 4465 }, { "epoch": 1.03, "learning_rate": 3.2797935444110625e-05, "loss": 0.0, "step": 4466 }, { "epoch": 1.03, "learning_rate": 3.279408366073492e-05, "loss": 0.0, "step": 4467 }, { "epoch": 1.03, "learning_rate": 3.2790231877359215e-05, "loss": 0.0, "step": 4468 }, { "epoch": 1.03, "learning_rate": 3.278638009398352e-05, "loss": 0.0, "step": 4469 }, { "epoch": 1.03, "learning_rate": 3.278252831060781e-05, "loss": 0.0001, "step": 4470 }, { "epoch": 1.03, "learning_rate": 3.2778676527232113e-05, "loss": 0.0001, "step": 4471 }, { "epoch": 1.03, "learning_rate": 3.277482474385641e-05, "loss": 0.0, "step": 4472 }, { "epoch": 1.03, "learning_rate": 3.27709729604807e-05, "loss": 0.0, "step": 4473 }, { "epoch": 1.03, "learning_rate": 3.2767121177105005e-05, "loss": 0.0002, "step": 4474 }, { "epoch": 1.03, "learning_rate": 3.276326939372929e-05, "loss": 0.0, "step": 4475 }, { "epoch": 1.03, "learning_rate": 3.2759417610353595e-05, "loss": 0.0, "step": 4476 }, { "epoch": 1.03, "learning_rate": 3.275556582697789e-05, "loss": 0.0, "step": 4477 }, { "epoch": 1.03, "learning_rate": 3.275171404360219e-05, "loss": 0.0, "step": 4478 }, { "epoch": 1.04, "learning_rate": 3.274786226022649e-05, "loss": 0.0, "step": 4479 }, { "epoch": 1.04, "learning_rate": 3.274401047685078e-05, "loss": 0.0, "step": 4480 }, { "epoch": 1.04, "learning_rate": 3.274015869347508e-05, "loss": 0.0, "step": 4481 }, { "epoch": 1.04, "learning_rate": 3.273630691009938e-05, "loss": 0.0, "step": 4482 }, { "epoch": 1.04, "learning_rate": 3.273245512672368e-05, "loss": 0.0, "step": 4483 }, { "epoch": 1.04, "learning_rate": 3.2728603343347975e-05, "loss": 0.0, "step": 4484 }, { "epoch": 1.04, "learning_rate": 3.272475155997226e-05, "loss": 0.0, "step": 4485 }, { "epoch": 1.04, "learning_rate": 3.2720899776596565e-05, "loss": 0.0, "step": 4486 }, { "epoch": 1.04, "learning_rate": 3.271704799322086e-05, "loss": 0.0, "step": 4487 }, { "epoch": 1.04, "learning_rate": 3.271319620984516e-05, "loss": 0.0, "step": 4488 }, { "epoch": 1.04, "learning_rate": 3.2709344426469457e-05, "loss": 0.0, "step": 4489 }, { "epoch": 1.04, "learning_rate": 3.270549264309375e-05, "loss": 0.0006, "step": 4490 }, { "epoch": 1.04, "learning_rate": 3.270164085971805e-05, "loss": 0.0, "step": 4491 }, { "epoch": 1.04, "learning_rate": 3.269778907634235e-05, "loss": 0.0, "step": 4492 }, { "epoch": 1.04, "learning_rate": 3.269393729296665e-05, "loss": 0.0, "step": 4493 }, { "epoch": 1.04, "learning_rate": 3.269008550959094e-05, "loss": 0.0, "step": 4494 }, { "epoch": 1.04, "learning_rate": 3.268623372621524e-05, "loss": 0.0, "step": 4495 }, { "epoch": 1.04, "learning_rate": 3.2682381942839535e-05, "loss": 0.0, "step": 4496 }, { "epoch": 1.04, "learning_rate": 3.267853015946383e-05, "loss": 0.0, "step": 4497 }, { "epoch": 1.04, "learning_rate": 3.267467837608813e-05, "loss": 0.0, "step": 4498 }, { "epoch": 1.04, "learning_rate": 3.2670826592712426e-05, "loss": 0.0, "step": 4499 }, { "epoch": 1.04, "learning_rate": 3.266697480933673e-05, "loss": 0.0, "step": 4500 }, { "epoch": 1.04, "learning_rate": 3.266312302596102e-05, "loss": 0.0001, "step": 4501 }, { "epoch": 1.04, "learning_rate": 3.265927124258532e-05, "loss": 0.0, "step": 4502 }, { "epoch": 1.04, "learning_rate": 3.265541945920961e-05, "loss": 0.0, "step": 4503 }, { "epoch": 1.04, "learning_rate": 3.265156767583391e-05, "loss": 0.0, "step": 4504 }, { "epoch": 1.04, "learning_rate": 3.264771589245821e-05, "loss": 0.0, "step": 4505 }, { "epoch": 1.04, "learning_rate": 3.2643864109082505e-05, "loss": 0.0, "step": 4506 }, { "epoch": 1.04, "learning_rate": 3.2640012325706806e-05, "loss": 0.0, "step": 4507 }, { "epoch": 1.04, "learning_rate": 3.26361605423311e-05, "loss": 0.0, "step": 4508 }, { "epoch": 1.04, "learning_rate": 3.2632308758955396e-05, "loss": 0.0, "step": 4509 }, { "epoch": 1.04, "learning_rate": 3.26284569755797e-05, "loss": 0.0, "step": 4510 }, { "epoch": 1.04, "learning_rate": 3.262460519220399e-05, "loss": 0.0, "step": 4511 }, { "epoch": 1.04, "learning_rate": 3.2620753408828295e-05, "loss": 0.0, "step": 4512 }, { "epoch": 1.04, "learning_rate": 3.261690162545258e-05, "loss": 0.0, "step": 4513 }, { "epoch": 1.04, "learning_rate": 3.2613049842076885e-05, "loss": 0.0, "step": 4514 }, { "epoch": 1.04, "learning_rate": 3.260919805870118e-05, "loss": 0.0, "step": 4515 }, { "epoch": 1.04, "learning_rate": 3.2605346275325475e-05, "loss": 0.0, "step": 4516 }, { "epoch": 1.04, "learning_rate": 3.2601494491949776e-05, "loss": 0.0, "step": 4517 }, { "epoch": 1.04, "learning_rate": 3.259764270857407e-05, "loss": 0.0, "step": 4518 }, { "epoch": 1.04, "learning_rate": 3.259379092519837e-05, "loss": 0.0, "step": 4519 }, { "epoch": 1.04, "learning_rate": 3.258993914182267e-05, "loss": 0.0001, "step": 4520 }, { "epoch": 1.04, "learning_rate": 3.258608735844696e-05, "loss": 0.0, "step": 4521 }, { "epoch": 1.05, "learning_rate": 3.258223557507126e-05, "loss": 0.0, "step": 4522 }, { "epoch": 1.05, "learning_rate": 3.257838379169555e-05, "loss": 0.0, "step": 4523 }, { "epoch": 1.05, "learning_rate": 3.2574532008319855e-05, "loss": 0.0, "step": 4524 }, { "epoch": 1.05, "learning_rate": 3.257068022494415e-05, "loss": 0.0, "step": 4525 }, { "epoch": 1.05, "learning_rate": 3.2566828441568444e-05, "loss": 0.0, "step": 4526 }, { "epoch": 1.05, "learning_rate": 3.2562976658192746e-05, "loss": 0.0, "step": 4527 }, { "epoch": 1.05, "learning_rate": 3.255912487481704e-05, "loss": 0.0001, "step": 4528 }, { "epoch": 1.05, "learning_rate": 3.255527309144134e-05, "loss": 0.0, "step": 4529 }, { "epoch": 1.05, "learning_rate": 3.255142130806564e-05, "loss": 0.0, "step": 4530 }, { "epoch": 1.05, "learning_rate": 3.254756952468993e-05, "loss": 0.0003, "step": 4531 }, { "epoch": 1.05, "learning_rate": 3.254371774131423e-05, "loss": 0.0, "step": 4532 }, { "epoch": 1.05, "learning_rate": 3.253986595793852e-05, "loss": 0.0001, "step": 4533 }, { "epoch": 1.05, "learning_rate": 3.2536014174562824e-05, "loss": 0.0, "step": 4534 }, { "epoch": 1.05, "learning_rate": 3.253216239118712e-05, "loss": 0.0, "step": 4535 }, { "epoch": 1.05, "learning_rate": 3.252831060781142e-05, "loss": 0.0, "step": 4536 }, { "epoch": 1.05, "learning_rate": 3.2524458824435716e-05, "loss": 0.0, "step": 4537 }, { "epoch": 1.05, "learning_rate": 3.252060704106001e-05, "loss": 0.0, "step": 4538 }, { "epoch": 1.05, "learning_rate": 3.251675525768431e-05, "loss": 0.0, "step": 4539 }, { "epoch": 1.05, "learning_rate": 3.251290347430861e-05, "loss": 0.0, "step": 4540 }, { "epoch": 1.05, "learning_rate": 3.25090516909329e-05, "loss": 0.0, "step": 4541 }, { "epoch": 1.05, "learning_rate": 3.25051999075572e-05, "loss": 0.0, "step": 4542 }, { "epoch": 1.05, "learning_rate": 3.25013481241815e-05, "loss": 0.0, "step": 4543 }, { "epoch": 1.05, "learning_rate": 3.2497496340805794e-05, "loss": 0.0002, "step": 4544 }, { "epoch": 1.05, "learning_rate": 3.249364455743009e-05, "loss": 0.0, "step": 4545 }, { "epoch": 1.05, "learning_rate": 3.248979277405439e-05, "loss": 0.0001, "step": 4546 }, { "epoch": 1.05, "learning_rate": 3.2485940990678686e-05, "loss": 0.0, "step": 4547 }, { "epoch": 1.05, "learning_rate": 3.248208920730299e-05, "loss": 0.0, "step": 4548 }, { "epoch": 1.05, "learning_rate": 3.247823742392728e-05, "loss": 0.0, "step": 4549 }, { "epoch": 1.05, "learning_rate": 3.247438564055158e-05, "loss": 0.0, "step": 4550 }, { "epoch": 1.05, "learning_rate": 3.247053385717587e-05, "loss": 0.0, "step": 4551 }, { "epoch": 1.05, "learning_rate": 3.246668207380017e-05, "loss": 0.0, "step": 4552 }, { "epoch": 1.05, "learning_rate": 3.246283029042447e-05, "loss": 0.0, "step": 4553 }, { "epoch": 1.05, "learning_rate": 3.2458978507048764e-05, "loss": 0.0, "step": 4554 }, { "epoch": 1.05, "learning_rate": 3.2455126723673066e-05, "loss": 0.0, "step": 4555 }, { "epoch": 1.05, "learning_rate": 3.245127494029736e-05, "loss": 0.0, "step": 4556 }, { "epoch": 1.05, "learning_rate": 3.2447423156921656e-05, "loss": 0.0, "step": 4557 }, { "epoch": 1.05, "learning_rate": 3.244357137354596e-05, "loss": 0.0, "step": 4558 }, { "epoch": 1.05, "learning_rate": 3.2439719590170246e-05, "loss": 0.0, "step": 4559 }, { "epoch": 1.05, "learning_rate": 3.243586780679455e-05, "loss": 0.0, "step": 4560 }, { "epoch": 1.05, "learning_rate": 3.243201602341884e-05, "loss": 0.0, "step": 4561 }, { "epoch": 1.05, "learning_rate": 3.242816424004314e-05, "loss": 0.0002, "step": 4562 }, { "epoch": 1.05, "learning_rate": 3.242431245666744e-05, "loss": 0.0, "step": 4563 }, { "epoch": 1.05, "learning_rate": 3.2420460673291734e-05, "loss": 0.0, "step": 4564 }, { "epoch": 1.06, "learning_rate": 3.2416608889916036e-05, "loss": 0.0, "step": 4565 }, { "epoch": 1.06, "learning_rate": 3.241275710654033e-05, "loss": 0.0, "step": 4566 }, { "epoch": 1.06, "learning_rate": 3.2408905323164626e-05, "loss": 0.0, "step": 4567 }, { "epoch": 1.06, "learning_rate": 3.240505353978893e-05, "loss": 0.0001, "step": 4568 }, { "epoch": 1.06, "learning_rate": 3.2401201756413216e-05, "loss": 0.0, "step": 4569 }, { "epoch": 1.06, "learning_rate": 3.239734997303752e-05, "loss": 0.0, "step": 4570 }, { "epoch": 1.06, "learning_rate": 3.239349818966181e-05, "loss": 0.0, "step": 4571 }, { "epoch": 1.06, "learning_rate": 3.2389646406286114e-05, "loss": 0.0, "step": 4572 }, { "epoch": 1.06, "learning_rate": 3.238579462291041e-05, "loss": 0.0, "step": 4573 }, { "epoch": 1.06, "learning_rate": 3.2381942839534704e-05, "loss": 0.0, "step": 4574 }, { "epoch": 1.06, "learning_rate": 3.2378091056159006e-05, "loss": 0.0, "step": 4575 }, { "epoch": 1.06, "learning_rate": 3.23742392727833e-05, "loss": 0.0, "step": 4576 }, { "epoch": 1.06, "learning_rate": 3.23703874894076e-05, "loss": 0.0, "step": 4577 }, { "epoch": 1.06, "learning_rate": 3.236653570603189e-05, "loss": 0.0, "step": 4578 }, { "epoch": 1.06, "learning_rate": 3.236268392265619e-05, "loss": 0.0, "step": 4579 }, { "epoch": 1.06, "learning_rate": 3.235883213928049e-05, "loss": 0.0007, "step": 4580 }, { "epoch": 1.06, "learning_rate": 3.235498035590478e-05, "loss": 0.0, "step": 4581 }, { "epoch": 1.06, "learning_rate": 3.2351128572529084e-05, "loss": 0.0, "step": 4582 }, { "epoch": 1.06, "learning_rate": 3.234727678915338e-05, "loss": 0.0, "step": 4583 }, { "epoch": 1.06, "learning_rate": 3.234342500577768e-05, "loss": 0.0, "step": 4584 }, { "epoch": 1.06, "learning_rate": 3.2339573222401976e-05, "loss": 0.0, "step": 4585 }, { "epoch": 1.06, "learning_rate": 3.233572143902627e-05, "loss": 0.0, "step": 4586 }, { "epoch": 1.06, "learning_rate": 3.2331869655650565e-05, "loss": 0.0, "step": 4587 }, { "epoch": 1.06, "learning_rate": 3.232801787227486e-05, "loss": 0.0, "step": 4588 }, { "epoch": 1.06, "learning_rate": 3.232416608889916e-05, "loss": 0.0, "step": 4589 }, { "epoch": 1.06, "learning_rate": 3.232031430552346e-05, "loss": 0.0, "step": 4590 }, { "epoch": 1.06, "learning_rate": 3.231646252214776e-05, "loss": 0.0, "step": 4591 }, { "epoch": 1.06, "learning_rate": 3.2312610738772054e-05, "loss": 0.0, "step": 4592 }, { "epoch": 1.06, "learning_rate": 3.230875895539635e-05, "loss": 0.0003, "step": 4593 }, { "epoch": 1.06, "learning_rate": 3.230490717202065e-05, "loss": 0.0, "step": 4594 }, { "epoch": 1.06, "learning_rate": 3.2301055388644945e-05, "loss": 0.0, "step": 4595 }, { "epoch": 1.06, "learning_rate": 3.229720360526925e-05, "loss": 0.0, "step": 4596 }, { "epoch": 1.06, "learning_rate": 3.2293351821893535e-05, "loss": 0.0, "step": 4597 }, { "epoch": 1.06, "learning_rate": 3.228950003851783e-05, "loss": 0.0, "step": 4598 }, { "epoch": 1.06, "learning_rate": 3.228564825514213e-05, "loss": 0.0, "step": 4599 }, { "epoch": 1.06, "learning_rate": 3.228179647176643e-05, "loss": 0.0, "step": 4600 }, { "epoch": 1.06, "learning_rate": 3.227794468839073e-05, "loss": 0.0, "step": 4601 }, { "epoch": 1.06, "learning_rate": 3.2274092905015024e-05, "loss": 0.0, "step": 4602 }, { "epoch": 1.06, "learning_rate": 3.227024112163932e-05, "loss": 0.0, "step": 4603 }, { "epoch": 1.06, "learning_rate": 3.226638933826362e-05, "loss": 0.0, "step": 4604 }, { "epoch": 1.06, "learning_rate": 3.2262537554887915e-05, "loss": 0.0, "step": 4605 }, { "epoch": 1.06, "learning_rate": 3.225868577151221e-05, "loss": 0.0, "step": 4606 }, { "epoch": 1.06, "learning_rate": 3.2254833988136505e-05, "loss": 0.0, "step": 4607 }, { "epoch": 1.06, "learning_rate": 3.225098220476081e-05, "loss": 0.0, "step": 4608 }, { "epoch": 1.07, "learning_rate": 3.22471304213851e-05, "loss": 0.0, "step": 4609 }, { "epoch": 1.07, "learning_rate": 3.22432786380094e-05, "loss": 0.0, "step": 4610 }, { "epoch": 1.07, "learning_rate": 3.22394268546337e-05, "loss": 0.0, "step": 4611 }, { "epoch": 1.07, "learning_rate": 3.2235575071257994e-05, "loss": 0.0, "step": 4612 }, { "epoch": 1.07, "learning_rate": 3.2231723287882295e-05, "loss": 0.0, "step": 4613 }, { "epoch": 1.07, "learning_rate": 3.222787150450659e-05, "loss": 0.0, "step": 4614 }, { "epoch": 1.07, "learning_rate": 3.2224019721130885e-05, "loss": 0.0, "step": 4615 }, { "epoch": 1.07, "learning_rate": 3.222016793775518e-05, "loss": 0.0, "step": 4616 }, { "epoch": 1.07, "learning_rate": 3.2216316154379475e-05, "loss": 0.0, "step": 4617 }, { "epoch": 1.07, "learning_rate": 3.221246437100378e-05, "loss": 0.0001, "step": 4618 }, { "epoch": 1.07, "learning_rate": 3.220861258762807e-05, "loss": 0.0, "step": 4619 }, { "epoch": 1.07, "learning_rate": 3.2204760804252374e-05, "loss": 0.0, "step": 4620 }, { "epoch": 1.07, "learning_rate": 3.220090902087667e-05, "loss": 0.0, "step": 4621 }, { "epoch": 1.07, "learning_rate": 3.2197057237500963e-05, "loss": 0.0, "step": 4622 }, { "epoch": 1.07, "learning_rate": 3.2193205454125265e-05, "loss": 0.0, "step": 4623 }, { "epoch": 1.07, "learning_rate": 3.218935367074956e-05, "loss": 0.0, "step": 4624 }, { "epoch": 1.07, "learning_rate": 3.2185501887373855e-05, "loss": 0.0, "step": 4625 }, { "epoch": 1.07, "learning_rate": 3.218165010399815e-05, "loss": 0.0, "step": 4626 }, { "epoch": 1.07, "learning_rate": 3.217779832062245e-05, "loss": 0.0, "step": 4627 }, { "epoch": 1.07, "learning_rate": 3.217394653724675e-05, "loss": 0.0, "step": 4628 }, { "epoch": 1.07, "learning_rate": 3.217009475387104e-05, "loss": 0.0, "step": 4629 }, { "epoch": 1.07, "learning_rate": 3.2166242970495343e-05, "loss": 0.0, "step": 4630 }, { "epoch": 1.07, "learning_rate": 3.216239118711964e-05, "loss": 0.0002, "step": 4631 }, { "epoch": 1.07, "learning_rate": 3.215853940374394e-05, "loss": 0.0, "step": 4632 }, { "epoch": 1.07, "learning_rate": 3.2154687620368235e-05, "loss": 0.0, "step": 4633 }, { "epoch": 1.07, "learning_rate": 3.215083583699253e-05, "loss": 0.0, "step": 4634 }, { "epoch": 1.07, "learning_rate": 3.2146984053616825e-05, "loss": 0.0, "step": 4635 }, { "epoch": 1.07, "learning_rate": 3.214313227024112e-05, "loss": 0.0, "step": 4636 }, { "epoch": 1.07, "learning_rate": 3.213928048686542e-05, "loss": 0.0, "step": 4637 }, { "epoch": 1.07, "learning_rate": 3.213542870348972e-05, "loss": 0.0, "step": 4638 }, { "epoch": 1.07, "learning_rate": 3.213157692011401e-05, "loss": 0.0, "step": 4639 }, { "epoch": 1.07, "learning_rate": 3.212772513673831e-05, "loss": 0.0002, "step": 4640 }, { "epoch": 1.07, "learning_rate": 3.212387335336261e-05, "loss": 0.0, "step": 4641 }, { "epoch": 1.07, "learning_rate": 3.212002156998691e-05, "loss": 0.0, "step": 4642 }, { "epoch": 1.07, "learning_rate": 3.21161697866112e-05, "loss": 0.0, "step": 4643 }, { "epoch": 1.07, "learning_rate": 3.21123180032355e-05, "loss": 0.0, "step": 4644 }, { "epoch": 1.07, "learning_rate": 3.2108466219859795e-05, "loss": 0.0, "step": 4645 }, { "epoch": 1.07, "learning_rate": 3.210461443648409e-05, "loss": 0.0, "step": 4646 }, { "epoch": 1.07, "learning_rate": 3.210076265310839e-05, "loss": 0.0, "step": 4647 }, { "epoch": 1.07, "learning_rate": 3.2096910869732687e-05, "loss": 0.0, "step": 4648 }, { "epoch": 1.07, "learning_rate": 3.209305908635699e-05, "loss": 0.0, "step": 4649 }, { "epoch": 1.07, "learning_rate": 3.208920730298128e-05, "loss": 0.0, "step": 4650 }, { "epoch": 1.07, "learning_rate": 3.208535551960558e-05, "loss": 0.0, "step": 4651 }, { "epoch": 1.08, "learning_rate": 3.208150373622988e-05, "loss": 0.0, "step": 4652 }, { "epoch": 1.08, "learning_rate": 3.207765195285417e-05, "loss": 0.0, "step": 4653 }, { "epoch": 1.08, "learning_rate": 3.207380016947847e-05, "loss": 0.0, "step": 4654 }, { "epoch": 1.08, "learning_rate": 3.2069948386102765e-05, "loss": 0.0, "step": 4655 }, { "epoch": 1.08, "learning_rate": 3.2066096602727066e-05, "loss": 0.0, "step": 4656 }, { "epoch": 1.08, "learning_rate": 3.206224481935136e-05, "loss": 0.0, "step": 4657 }, { "epoch": 1.08, "learning_rate": 3.2058393035975656e-05, "loss": 0.0, "step": 4658 }, { "epoch": 1.08, "learning_rate": 3.205454125259996e-05, "loss": 0.0, "step": 4659 }, { "epoch": 1.08, "learning_rate": 3.205068946922425e-05, "loss": 0.0, "step": 4660 }, { "epoch": 1.08, "learning_rate": 3.2046837685848555e-05, "loss": 0.0, "step": 4661 }, { "epoch": 1.08, "learning_rate": 3.204298590247284e-05, "loss": 0.0, "step": 4662 }, { "epoch": 1.08, "learning_rate": 3.2039134119097145e-05, "loss": 0.0, "step": 4663 }, { "epoch": 1.08, "learning_rate": 3.203528233572144e-05, "loss": 0.0, "step": 4664 }, { "epoch": 1.08, "learning_rate": 3.2031430552345735e-05, "loss": 0.0, "step": 4665 }, { "epoch": 1.08, "learning_rate": 3.2027578768970036e-05, "loss": 0.0, "step": 4666 }, { "epoch": 1.08, "learning_rate": 3.202372698559433e-05, "loss": 0.0001, "step": 4667 }, { "epoch": 1.08, "learning_rate": 3.201987520221863e-05, "loss": 0.0, "step": 4668 }, { "epoch": 1.08, "learning_rate": 3.201602341884293e-05, "loss": 0.0, "step": 4669 }, { "epoch": 1.08, "learning_rate": 3.201217163546722e-05, "loss": 0.0, "step": 4670 }, { "epoch": 1.08, "learning_rate": 3.200831985209152e-05, "loss": 0.0, "step": 4671 }, { "epoch": 1.08, "learning_rate": 3.200446806871581e-05, "loss": 0.0, "step": 4672 }, { "epoch": 1.08, "learning_rate": 3.2000616285340115e-05, "loss": 0.0, "step": 4673 }, { "epoch": 1.08, "learning_rate": 3.199676450196441e-05, "loss": 0.0, "step": 4674 }, { "epoch": 1.08, "learning_rate": 3.1992912718588705e-05, "loss": 0.0, "step": 4675 }, { "epoch": 1.08, "learning_rate": 3.1989060935213006e-05, "loss": 0.0, "step": 4676 }, { "epoch": 1.08, "learning_rate": 3.19852091518373e-05, "loss": 0.0, "step": 4677 }, { "epoch": 1.08, "learning_rate": 3.19813573684616e-05, "loss": 0.0, "step": 4678 }, { "epoch": 1.08, "learning_rate": 3.19775055850859e-05, "loss": 0.0, "step": 4679 }, { "epoch": 1.08, "learning_rate": 3.197365380171019e-05, "loss": 0.0, "step": 4680 }, { "epoch": 1.08, "learning_rate": 3.196980201833449e-05, "loss": 0.0, "step": 4681 }, { "epoch": 1.08, "learning_rate": 3.196595023495878e-05, "loss": 0.0, "step": 4682 }, { "epoch": 1.08, "learning_rate": 3.1962098451583084e-05, "loss": 0.0, "step": 4683 }, { "epoch": 1.08, "learning_rate": 3.195824666820738e-05, "loss": 0.0, "step": 4684 }, { "epoch": 1.08, "learning_rate": 3.195439488483168e-05, "loss": 0.0, "step": 4685 }, { "epoch": 1.08, "learning_rate": 3.1950543101455976e-05, "loss": 0.0, "step": 4686 }, { "epoch": 1.08, "learning_rate": 3.194669131808027e-05, "loss": 0.0, "step": 4687 }, { "epoch": 1.08, "learning_rate": 3.194283953470457e-05, "loss": 0.0, "step": 4688 }, { "epoch": 1.08, "learning_rate": 3.193898775132887e-05, "loss": 0.0, "step": 4689 }, { "epoch": 1.08, "learning_rate": 3.193513596795316e-05, "loss": 0.0, "step": 4690 }, { "epoch": 1.08, "learning_rate": 3.193128418457746e-05, "loss": 0.0001, "step": 4691 }, { "epoch": 1.08, "learning_rate": 3.192743240120176e-05, "loss": 0.0, "step": 4692 }, { "epoch": 1.08, "learning_rate": 3.1923580617826054e-05, "loss": 0.0, "step": 4693 }, { "epoch": 1.08, "learning_rate": 3.191972883445035e-05, "loss": 0.0001, "step": 4694 }, { "epoch": 1.09, "learning_rate": 3.191587705107465e-05, "loss": 0.0, "step": 4695 }, { "epoch": 1.09, "learning_rate": 3.1912025267698946e-05, "loss": 0.0, "step": 4696 }, { "epoch": 1.09, "learning_rate": 3.190817348432325e-05, "loss": 0.0, "step": 4697 }, { "epoch": 1.09, "learning_rate": 3.190432170094754e-05, "loss": 0.0, "step": 4698 }, { "epoch": 1.09, "learning_rate": 3.190046991757184e-05, "loss": 0.0, "step": 4699 }, { "epoch": 1.09, "learning_rate": 3.189661813419613e-05, "loss": 0.0, "step": 4700 }, { "epoch": 1.09, "learning_rate": 3.189276635082043e-05, "loss": 0.0, "step": 4701 }, { "epoch": 1.09, "learning_rate": 3.188891456744473e-05, "loss": 0.0, "step": 4702 }, { "epoch": 1.09, "learning_rate": 3.1885062784069024e-05, "loss": 0.0, "step": 4703 }, { "epoch": 1.09, "learning_rate": 3.1881211000693326e-05, "loss": 0.0001, "step": 4704 }, { "epoch": 1.09, "learning_rate": 3.187735921731762e-05, "loss": 0.0, "step": 4705 }, { "epoch": 1.09, "learning_rate": 3.1873507433941916e-05, "loss": 0.0, "step": 4706 }, { "epoch": 1.09, "learning_rate": 3.186965565056622e-05, "loss": 0.0, "step": 4707 }, { "epoch": 1.09, "learning_rate": 3.186580386719051e-05, "loss": 0.0, "step": 4708 }, { "epoch": 1.09, "learning_rate": 3.186195208381481e-05, "loss": 0.0, "step": 4709 }, { "epoch": 1.09, "learning_rate": 3.18581003004391e-05, "loss": 0.0, "step": 4710 }, { "epoch": 1.09, "learning_rate": 3.1854248517063404e-05, "loss": 0.0, "step": 4711 }, { "epoch": 1.09, "learning_rate": 3.18503967336877e-05, "loss": 0.0, "step": 4712 }, { "epoch": 1.09, "learning_rate": 3.1846544950311994e-05, "loss": 0.0, "step": 4713 }, { "epoch": 1.09, "learning_rate": 3.1842693166936296e-05, "loss": 0.0, "step": 4714 }, { "epoch": 1.09, "learning_rate": 3.183884138356059e-05, "loss": 0.0, "step": 4715 }, { "epoch": 1.09, "learning_rate": 3.1834989600184886e-05, "loss": 0.0, "step": 4716 }, { "epoch": 1.09, "learning_rate": 3.183113781680919e-05, "loss": 0.0, "step": 4717 }, { "epoch": 1.09, "learning_rate": 3.1827286033433476e-05, "loss": 0.0, "step": 4718 }, { "epoch": 1.09, "learning_rate": 3.182343425005778e-05, "loss": 0.0, "step": 4719 }, { "epoch": 1.09, "learning_rate": 3.181958246668207e-05, "loss": 0.0, "step": 4720 }, { "epoch": 1.09, "learning_rate": 3.1815730683306374e-05, "loss": 0.0, "step": 4721 }, { "epoch": 1.09, "learning_rate": 3.181187889993067e-05, "loss": 0.0, "step": 4722 }, { "epoch": 1.09, "learning_rate": 3.1808027116554964e-05, "loss": 0.0, "step": 4723 }, { "epoch": 1.09, "learning_rate": 3.1804175333179266e-05, "loss": 0.0, "step": 4724 }, { "epoch": 1.09, "learning_rate": 3.180032354980356e-05, "loss": 0.0, "step": 4725 }, { "epoch": 1.09, "learning_rate": 3.179647176642786e-05, "loss": 0.0, "step": 4726 }, { "epoch": 1.09, "learning_rate": 3.179261998305215e-05, "loss": 0.0, "step": 4727 }, { "epoch": 1.09, "learning_rate": 3.178876819967645e-05, "loss": 0.0, "step": 4728 }, { "epoch": 1.09, "learning_rate": 3.178491641630075e-05, "loss": 0.0, "step": 4729 }, { "epoch": 1.09, "learning_rate": 3.178106463292504e-05, "loss": 0.0, "step": 4730 }, { "epoch": 1.09, "learning_rate": 3.1777212849549344e-05, "loss": 0.0, "step": 4731 }, { "epoch": 1.09, "learning_rate": 3.177336106617364e-05, "loss": 0.0, "step": 4732 }, { "epoch": 1.09, "learning_rate": 3.176950928279794e-05, "loss": 0.0, "step": 4733 }, { "epoch": 1.09, "learning_rate": 3.1765657499422236e-05, "loss": 0.0, "step": 4734 }, { "epoch": 1.09, "learning_rate": 3.176180571604653e-05, "loss": 0.0001, "step": 4735 }, { "epoch": 1.09, "learning_rate": 3.175795393267083e-05, "loss": 0.0, "step": 4736 }, { "epoch": 1.09, "learning_rate": 3.175410214929512e-05, "loss": 0.0, "step": 4737 }, { "epoch": 1.09, "learning_rate": 3.175025036591942e-05, "loss": 0.0, "step": 4738 }, { "epoch": 1.1, "learning_rate": 3.174639858254372e-05, "loss": 0.0, "step": 4739 }, { "epoch": 1.1, "learning_rate": 3.174254679916802e-05, "loss": 0.0, "step": 4740 }, { "epoch": 1.1, "learning_rate": 3.1738695015792314e-05, "loss": 0.0, "step": 4741 }, { "epoch": 1.1, "learning_rate": 3.173484323241661e-05, "loss": 0.0, "step": 4742 }, { "epoch": 1.1, "learning_rate": 3.173099144904091e-05, "loss": 0.0, "step": 4743 }, { "epoch": 1.1, "learning_rate": 3.1727139665665206e-05, "loss": 0.0, "step": 4744 }, { "epoch": 1.1, "learning_rate": 3.172328788228951e-05, "loss": 0.0001, "step": 4745 }, { "epoch": 1.1, "learning_rate": 3.1719436098913795e-05, "loss": 0.0, "step": 4746 }, { "epoch": 1.1, "learning_rate": 3.17155843155381e-05, "loss": 0.0, "step": 4747 }, { "epoch": 1.1, "learning_rate": 3.171173253216239e-05, "loss": 0.0, "step": 4748 }, { "epoch": 1.1, "learning_rate": 3.170788074878669e-05, "loss": 0.0, "step": 4749 }, { "epoch": 1.1, "learning_rate": 3.170402896541099e-05, "loss": 0.0, "step": 4750 }, { "epoch": 1.1, "learning_rate": 3.1700177182035284e-05, "loss": 0.0, "step": 4751 }, { "epoch": 1.1, "learning_rate": 3.169632539865958e-05, "loss": 0.0, "step": 4752 }, { "epoch": 1.1, "learning_rate": 3.169247361528388e-05, "loss": 0.0, "step": 4753 }, { "epoch": 1.1, "learning_rate": 3.1688621831908175e-05, "loss": 0.0, "step": 4754 }, { "epoch": 1.1, "learning_rate": 3.168477004853247e-05, "loss": 0.0, "step": 4755 }, { "epoch": 1.1, "learning_rate": 3.1680918265156765e-05, "loss": 0.0, "step": 4756 }, { "epoch": 1.1, "learning_rate": 3.167706648178107e-05, "loss": 0.0, "step": 4757 }, { "epoch": 1.1, "learning_rate": 3.167321469840536e-05, "loss": 0.0001, "step": 4758 }, { "epoch": 1.1, "learning_rate": 3.166936291502966e-05, "loss": 0.0, "step": 4759 }, { "epoch": 1.1, "learning_rate": 3.166551113165396e-05, "loss": 0.0, "step": 4760 }, { "epoch": 1.1, "learning_rate": 3.1661659348278254e-05, "loss": 0.0, "step": 4761 }, { "epoch": 1.1, "learning_rate": 3.1657807564902555e-05, "loss": 0.0, "step": 4762 }, { "epoch": 1.1, "learning_rate": 3.165395578152685e-05, "loss": 0.0, "step": 4763 }, { "epoch": 1.1, "learning_rate": 3.1650103998151145e-05, "loss": 0.0, "step": 4764 }, { "epoch": 1.1, "learning_rate": 3.164625221477544e-05, "loss": 0.0, "step": 4765 }, { "epoch": 1.1, "learning_rate": 3.1642400431399735e-05, "loss": 0.0, "step": 4766 }, { "epoch": 1.1, "learning_rate": 3.163854864802404e-05, "loss": 0.0, "step": 4767 }, { "epoch": 1.1, "learning_rate": 3.163469686464833e-05, "loss": 0.0007, "step": 4768 }, { "epoch": 1.1, "learning_rate": 3.1630845081272634e-05, "loss": 0.0, "step": 4769 }, { "epoch": 1.1, "learning_rate": 3.162699329789693e-05, "loss": 0.0, "step": 4770 }, { "epoch": 1.1, "learning_rate": 3.1623141514521224e-05, "loss": 0.0, "step": 4771 }, { "epoch": 1.1, "learning_rate": 3.1619289731145525e-05, "loss": 0.0, "step": 4772 }, { "epoch": 1.1, "learning_rate": 3.161543794776982e-05, "loss": 0.0, "step": 4773 }, { "epoch": 1.1, "learning_rate": 3.1611586164394115e-05, "loss": 0.0, "step": 4774 }, { "epoch": 1.1, "learning_rate": 3.160773438101841e-05, "loss": 0.0, "step": 4775 }, { "epoch": 1.1, "learning_rate": 3.160388259764271e-05, "loss": 0.0, "step": 4776 }, { "epoch": 1.1, "learning_rate": 3.160003081426701e-05, "loss": 0.0, "step": 4777 }, { "epoch": 1.1, "learning_rate": 3.15961790308913e-05, "loss": 0.0, "step": 4778 }, { "epoch": 1.1, "learning_rate": 3.1592327247515604e-05, "loss": 0.0, "step": 4779 }, { "epoch": 1.1, "learning_rate": 3.15884754641399e-05, "loss": 0.0, "step": 4780 }, { "epoch": 1.1, "learning_rate": 3.15846236807642e-05, "loss": 0.0001, "step": 4781 }, { "epoch": 1.11, "learning_rate": 3.1580771897388495e-05, "loss": 0.0, "step": 4782 }, { "epoch": 1.11, "learning_rate": 3.157692011401279e-05, "loss": 0.0, "step": 4783 }, { "epoch": 1.11, "learning_rate": 3.1573068330637085e-05, "loss": 0.0, "step": 4784 }, { "epoch": 1.11, "learning_rate": 3.156921654726138e-05, "loss": 0.0, "step": 4785 }, { "epoch": 1.11, "learning_rate": 3.156536476388568e-05, "loss": 0.0, "step": 4786 }, { "epoch": 1.11, "learning_rate": 3.156151298050998e-05, "loss": 0.0, "step": 4787 }, { "epoch": 1.11, "learning_rate": 3.155766119713427e-05, "loss": 0.0, "step": 4788 }, { "epoch": 1.11, "learning_rate": 3.1553809413758573e-05, "loss": 0.0, "step": 4789 }, { "epoch": 1.11, "learning_rate": 3.154995763038287e-05, "loss": 0.0, "step": 4790 }, { "epoch": 1.11, "learning_rate": 3.154610584700717e-05, "loss": 0.0, "step": 4791 }, { "epoch": 1.11, "learning_rate": 3.1542254063631465e-05, "loss": 0.0, "step": 4792 }, { "epoch": 1.11, "learning_rate": 3.153840228025576e-05, "loss": 0.0, "step": 4793 }, { "epoch": 1.11, "learning_rate": 3.1534550496880055e-05, "loss": 0.0, "step": 4794 }, { "epoch": 1.11, "learning_rate": 3.153069871350435e-05, "loss": 0.0, "step": 4795 }, { "epoch": 1.11, "learning_rate": 3.152684693012865e-05, "loss": 0.0, "step": 4796 }, { "epoch": 1.11, "learning_rate": 3.1522995146752947e-05, "loss": 0.0, "step": 4797 }, { "epoch": 1.11, "learning_rate": 3.151914336337725e-05, "loss": 0.0, "step": 4798 }, { "epoch": 1.11, "learning_rate": 3.151529158000154e-05, "loss": 0.0, "step": 4799 }, { "epoch": 1.11, "learning_rate": 3.151143979662584e-05, "loss": 0.0, "step": 4800 }, { "epoch": 1.11, "learning_rate": 3.150758801325014e-05, "loss": 0.0, "step": 4801 }, { "epoch": 1.11, "learning_rate": 3.150373622987443e-05, "loss": 0.0, "step": 4802 }, { "epoch": 1.11, "learning_rate": 3.149988444649873e-05, "loss": 0.0, "step": 4803 }, { "epoch": 1.11, "learning_rate": 3.1496032663123025e-05, "loss": 0.0, "step": 4804 }, { "epoch": 1.11, "learning_rate": 3.1492180879747327e-05, "loss": 0.0, "step": 4805 }, { "epoch": 1.11, "learning_rate": 3.148832909637162e-05, "loss": 0.0, "step": 4806 }, { "epoch": 1.11, "learning_rate": 3.1484477312995916e-05, "loss": 0.0, "step": 4807 }, { "epoch": 1.11, "learning_rate": 3.148062552962022e-05, "loss": 0.0, "step": 4808 }, { "epoch": 1.11, "learning_rate": 3.147677374624451e-05, "loss": 0.0, "step": 4809 }, { "epoch": 1.11, "learning_rate": 3.1472921962868815e-05, "loss": 0.0002, "step": 4810 }, { "epoch": 1.11, "learning_rate": 3.14690701794931e-05, "loss": 0.0007, "step": 4811 }, { "epoch": 1.11, "learning_rate": 3.1465218396117405e-05, "loss": 0.0, "step": 4812 }, { "epoch": 1.11, "learning_rate": 3.14613666127417e-05, "loss": 0.0, "step": 4813 }, { "epoch": 1.11, "learning_rate": 3.1457514829365995e-05, "loss": 0.0, "step": 4814 }, { "epoch": 1.11, "learning_rate": 3.1453663045990296e-05, "loss": 0.0, "step": 4815 }, { "epoch": 1.11, "learning_rate": 3.144981126261459e-05, "loss": 0.0, "step": 4816 }, { "epoch": 1.11, "learning_rate": 3.144595947923889e-05, "loss": 0.0, "step": 4817 }, { "epoch": 1.11, "learning_rate": 3.144210769586319e-05, "loss": 0.0, "step": 4818 }, { "epoch": 1.11, "learning_rate": 3.143825591248748e-05, "loss": 0.0, "step": 4819 }, { "epoch": 1.11, "learning_rate": 3.1434404129111785e-05, "loss": 0.0, "step": 4820 }, { "epoch": 1.11, "learning_rate": 3.143055234573607e-05, "loss": 0.0, "step": 4821 }, { "epoch": 1.11, "learning_rate": 3.1426700562360375e-05, "loss": 0.0, "step": 4822 }, { "epoch": 1.11, "learning_rate": 3.142284877898467e-05, "loss": 0.0, "step": 4823 }, { "epoch": 1.11, "learning_rate": 3.141899699560897e-05, "loss": 0.0, "step": 4824 }, { "epoch": 1.12, "learning_rate": 3.1415145212233266e-05, "loss": 0.0004, "step": 4825 }, { "epoch": 1.12, "learning_rate": 3.141129342885756e-05, "loss": 0.0, "step": 4826 }, { "epoch": 1.12, "learning_rate": 3.140744164548186e-05, "loss": 0.0, "step": 4827 }, { "epoch": 1.12, "learning_rate": 3.140358986210616e-05, "loss": 0.0, "step": 4828 }, { "epoch": 1.12, "learning_rate": 3.139973807873045e-05, "loss": 0.0, "step": 4829 }, { "epoch": 1.12, "learning_rate": 3.139588629535475e-05, "loss": 0.0, "step": 4830 }, { "epoch": 1.12, "learning_rate": 3.139203451197904e-05, "loss": 0.0, "step": 4831 }, { "epoch": 1.12, "learning_rate": 3.1388182728603345e-05, "loss": 0.0, "step": 4832 }, { "epoch": 1.12, "learning_rate": 3.138433094522764e-05, "loss": 0.0, "step": 4833 }, { "epoch": 1.12, "learning_rate": 3.138047916185194e-05, "loss": 0.0, "step": 4834 }, { "epoch": 1.12, "learning_rate": 3.1376627378476236e-05, "loss": 0.0, "step": 4835 }, { "epoch": 1.12, "learning_rate": 3.137277559510053e-05, "loss": 0.0, "step": 4836 }, { "epoch": 1.12, "learning_rate": 3.136892381172483e-05, "loss": 0.0, "step": 4837 }, { "epoch": 1.12, "learning_rate": 3.136507202834913e-05, "loss": 0.0, "step": 4838 }, { "epoch": 1.12, "learning_rate": 3.136122024497342e-05, "loss": 0.0, "step": 4839 }, { "epoch": 1.12, "learning_rate": 3.135736846159772e-05, "loss": 0.0, "step": 4840 }, { "epoch": 1.12, "learning_rate": 3.135351667822202e-05, "loss": 0.0, "step": 4841 }, { "epoch": 1.12, "learning_rate": 3.1349664894846314e-05, "loss": 0.0, "step": 4842 }, { "epoch": 1.12, "learning_rate": 3.134581311147061e-05, "loss": 0.0, "step": 4843 }, { "epoch": 1.12, "learning_rate": 3.134196132809491e-05, "loss": 0.0, "step": 4844 }, { "epoch": 1.12, "learning_rate": 3.1338109544719206e-05, "loss": 0.0009, "step": 4845 }, { "epoch": 1.12, "learning_rate": 3.133425776134351e-05, "loss": 0.0, "step": 4846 }, { "epoch": 1.12, "learning_rate": 3.13304059779678e-05, "loss": 0.0, "step": 4847 }, { "epoch": 1.12, "learning_rate": 3.13265541945921e-05, "loss": 0.0001, "step": 4848 }, { "epoch": 1.12, "learning_rate": 3.132270241121639e-05, "loss": 0.0, "step": 4849 }, { "epoch": 1.12, "learning_rate": 3.131885062784069e-05, "loss": 0.0, "step": 4850 }, { "epoch": 1.12, "learning_rate": 3.131499884446499e-05, "loss": 0.0, "step": 4851 }, { "epoch": 1.12, "learning_rate": 3.1311147061089284e-05, "loss": 0.0, "step": 4852 }, { "epoch": 1.12, "learning_rate": 3.1307295277713586e-05, "loss": 0.0, "step": 4853 }, { "epoch": 1.12, "learning_rate": 3.130344349433788e-05, "loss": 0.0, "step": 4854 }, { "epoch": 1.12, "learning_rate": 3.1299591710962176e-05, "loss": 0.0, "step": 4855 }, { "epoch": 1.12, "learning_rate": 3.129573992758648e-05, "loss": 0.0, "step": 4856 }, { "epoch": 1.12, "learning_rate": 3.129188814421077e-05, "loss": 0.0, "step": 4857 }, { "epoch": 1.12, "learning_rate": 3.128803636083507e-05, "loss": 0.0, "step": 4858 }, { "epoch": 1.12, "learning_rate": 3.128418457745936e-05, "loss": 0.0, "step": 4859 }, { "epoch": 1.12, "learning_rate": 3.1280332794083664e-05, "loss": 0.0, "step": 4860 }, { "epoch": 1.12, "learning_rate": 3.127648101070796e-05, "loss": 0.0, "step": 4861 }, { "epoch": 1.12, "learning_rate": 3.1272629227332254e-05, "loss": 0.0, "step": 4862 }, { "epoch": 1.12, "learning_rate": 3.1268777443956556e-05, "loss": 0.0, "step": 4863 }, { "epoch": 1.12, "learning_rate": 3.126492566058085e-05, "loss": 0.0, "step": 4864 }, { "epoch": 1.12, "learning_rate": 3.1261073877205146e-05, "loss": 0.0, "step": 4865 }, { "epoch": 1.12, "learning_rate": 3.125722209382945e-05, "loss": 0.0, "step": 4866 }, { "epoch": 1.12, "learning_rate": 3.1253370310453736e-05, "loss": 0.0, "step": 4867 }, { "epoch": 1.13, "learning_rate": 3.124951852707804e-05, "loss": 0.0, "step": 4868 }, { "epoch": 1.13, "learning_rate": 3.124566674370233e-05, "loss": 0.0, "step": 4869 }, { "epoch": 1.13, "learning_rate": 3.1241814960326634e-05, "loss": 0.0, "step": 4870 }, { "epoch": 1.13, "learning_rate": 3.123796317695093e-05, "loss": 0.0, "step": 4871 }, { "epoch": 1.13, "learning_rate": 3.1234111393575224e-05, "loss": 0.0002, "step": 4872 }, { "epoch": 1.13, "learning_rate": 3.1230259610199526e-05, "loss": 0.0, "step": 4873 }, { "epoch": 1.13, "learning_rate": 3.122640782682382e-05, "loss": 0.0, "step": 4874 }, { "epoch": 1.13, "learning_rate": 3.122255604344812e-05, "loss": 0.0, "step": 4875 }, { "epoch": 1.13, "learning_rate": 3.121870426007242e-05, "loss": 0.0, "step": 4876 }, { "epoch": 1.13, "learning_rate": 3.121485247669671e-05, "loss": 0.0, "step": 4877 }, { "epoch": 1.13, "learning_rate": 3.121100069332101e-05, "loss": 0.0, "step": 4878 }, { "epoch": 1.13, "learning_rate": 3.12071489099453e-05, "loss": 0.0, "step": 4879 }, { "epoch": 1.13, "learning_rate": 3.1203297126569604e-05, "loss": 0.0002, "step": 4880 }, { "epoch": 1.13, "learning_rate": 3.11994453431939e-05, "loss": 0.0, "step": 4881 }, { "epoch": 1.13, "learning_rate": 3.11955935598182e-05, "loss": 0.0, "step": 4882 }, { "epoch": 1.13, "learning_rate": 3.1191741776442496e-05, "loss": 0.0, "step": 4883 }, { "epoch": 1.13, "learning_rate": 3.118788999306679e-05, "loss": 0.0, "step": 4884 }, { "epoch": 1.13, "learning_rate": 3.118403820969109e-05, "loss": 0.0, "step": 4885 }, { "epoch": 1.13, "learning_rate": 3.118018642631538e-05, "loss": 0.0, "step": 4886 }, { "epoch": 1.13, "learning_rate": 3.117633464293968e-05, "loss": 0.0, "step": 4887 }, { "epoch": 1.13, "learning_rate": 3.117248285956398e-05, "loss": 0.0, "step": 4888 }, { "epoch": 1.13, "learning_rate": 3.116863107618828e-05, "loss": 0.0042, "step": 4889 }, { "epoch": 1.13, "learning_rate": 3.1164779292812574e-05, "loss": 0.0, "step": 4890 }, { "epoch": 1.13, "learning_rate": 3.116092750943687e-05, "loss": 0.0, "step": 4891 }, { "epoch": 1.13, "learning_rate": 3.115707572606117e-05, "loss": 0.0, "step": 4892 }, { "epoch": 1.13, "learning_rate": 3.1153223942685466e-05, "loss": 0.0, "step": 4893 }, { "epoch": 1.13, "learning_rate": 3.114937215930977e-05, "loss": 0.0, "step": 4894 }, { "epoch": 1.13, "learning_rate": 3.1145520375934056e-05, "loss": 0.0, "step": 4895 }, { "epoch": 1.13, "learning_rate": 3.114166859255836e-05, "loss": 0.0, "step": 4896 }, { "epoch": 1.13, "learning_rate": 3.113781680918265e-05, "loss": 0.0, "step": 4897 }, { "epoch": 1.13, "learning_rate": 3.113396502580695e-05, "loss": 0.0, "step": 4898 }, { "epoch": 1.13, "learning_rate": 3.113011324243125e-05, "loss": 0.0, "step": 4899 }, { "epoch": 1.13, "learning_rate": 3.1126261459055544e-05, "loss": 0.0, "step": 4900 }, { "epoch": 1.13, "learning_rate": 3.1122409675679846e-05, "loss": 0.0, "step": 4901 }, { "epoch": 1.13, "learning_rate": 3.111855789230414e-05, "loss": 0.0, "step": 4902 }, { "epoch": 1.13, "learning_rate": 3.1114706108928436e-05, "loss": 0.0, "step": 4903 }, { "epoch": 1.13, "learning_rate": 3.111085432555274e-05, "loss": 0.0, "step": 4904 }, { "epoch": 1.13, "learning_rate": 3.1107002542177025e-05, "loss": 0.0, "step": 4905 }, { "epoch": 1.13, "learning_rate": 3.110315075880133e-05, "loss": 0.0, "step": 4906 }, { "epoch": 1.13, "learning_rate": 3.109929897542562e-05, "loss": 0.0, "step": 4907 }, { "epoch": 1.13, "learning_rate": 3.109544719204992e-05, "loss": 0.0, "step": 4908 }, { "epoch": 1.13, "learning_rate": 3.109159540867422e-05, "loss": 0.0, "step": 4909 }, { "epoch": 1.13, "learning_rate": 3.1087743625298514e-05, "loss": 0.0, "step": 4910 }, { "epoch": 1.13, "learning_rate": 3.1083891841922815e-05, "loss": 0.0, "step": 4911 }, { "epoch": 1.14, "learning_rate": 3.108004005854711e-05, "loss": 0.0, "step": 4912 }, { "epoch": 1.14, "learning_rate": 3.1076188275171405e-05, "loss": 0.0, "step": 4913 }, { "epoch": 1.14, "learning_rate": 3.10723364917957e-05, "loss": 0.0, "step": 4914 }, { "epoch": 1.14, "learning_rate": 3.1068484708419995e-05, "loss": 0.0, "step": 4915 }, { "epoch": 1.14, "learning_rate": 3.10646329250443e-05, "loss": 0.0, "step": 4916 }, { "epoch": 1.14, "learning_rate": 3.106078114166859e-05, "loss": 0.0002, "step": 4917 }, { "epoch": 1.14, "learning_rate": 3.1056929358292894e-05, "loss": 0.0, "step": 4918 }, { "epoch": 1.14, "learning_rate": 3.105307757491719e-05, "loss": 0.0, "step": 4919 }, { "epoch": 1.14, "learning_rate": 3.1049225791541484e-05, "loss": 0.0, "step": 4920 }, { "epoch": 1.14, "learning_rate": 3.1045374008165785e-05, "loss": 0.0, "step": 4921 }, { "epoch": 1.14, "learning_rate": 3.104152222479008e-05, "loss": 0.0, "step": 4922 }, { "epoch": 1.14, "learning_rate": 3.1037670441414375e-05, "loss": 0.0, "step": 4923 }, { "epoch": 1.14, "learning_rate": 3.103381865803867e-05, "loss": 0.0, "step": 4924 }, { "epoch": 1.14, "learning_rate": 3.102996687466297e-05, "loss": 0.0, "step": 4925 }, { "epoch": 1.14, "learning_rate": 3.102611509128727e-05, "loss": 0.0002, "step": 4926 }, { "epoch": 1.14, "learning_rate": 3.102226330791156e-05, "loss": 0.0, "step": 4927 }, { "epoch": 1.14, "learning_rate": 3.1018411524535864e-05, "loss": 0.0, "step": 4928 }, { "epoch": 1.14, "learning_rate": 3.101455974116016e-05, "loss": 0.0, "step": 4929 }, { "epoch": 1.14, "learning_rate": 3.101070795778446e-05, "loss": 0.0, "step": 4930 }, { "epoch": 1.14, "learning_rate": 3.1006856174408755e-05, "loss": 0.0, "step": 4931 }, { "epoch": 1.14, "learning_rate": 3.100300439103305e-05, "loss": 0.0, "step": 4932 }, { "epoch": 1.14, "learning_rate": 3.0999152607657345e-05, "loss": 0.0, "step": 4933 }, { "epoch": 1.14, "learning_rate": 3.099530082428164e-05, "loss": 0.0, "step": 4934 }, { "epoch": 1.14, "learning_rate": 3.099144904090594e-05, "loss": 0.0, "step": 4935 }, { "epoch": 1.14, "learning_rate": 3.098759725753024e-05, "loss": 0.0, "step": 4936 }, { "epoch": 1.14, "learning_rate": 3.098374547415454e-05, "loss": 0.0, "step": 4937 }, { "epoch": 1.14, "learning_rate": 3.0979893690778833e-05, "loss": 0.0, "step": 4938 }, { "epoch": 1.14, "learning_rate": 3.097604190740313e-05, "loss": 0.0, "step": 4939 }, { "epoch": 1.14, "learning_rate": 3.097219012402743e-05, "loss": 0.0, "step": 4940 }, { "epoch": 1.14, "learning_rate": 3.0968338340651725e-05, "loss": 0.0, "step": 4941 }, { "epoch": 1.14, "learning_rate": 3.096448655727602e-05, "loss": 0.0, "step": 4942 }, { "epoch": 1.14, "learning_rate": 3.0960634773900315e-05, "loss": 0.0, "step": 4943 }, { "epoch": 1.14, "learning_rate": 3.095678299052461e-05, "loss": 0.0, "step": 4944 }, { "epoch": 1.14, "learning_rate": 3.095293120714891e-05, "loss": 0.0, "step": 4945 }, { "epoch": 1.14, "learning_rate": 3.094907942377321e-05, "loss": 0.0, "step": 4946 }, { "epoch": 1.14, "learning_rate": 3.094522764039751e-05, "loss": 0.0, "step": 4947 }, { "epoch": 1.14, "learning_rate": 3.09413758570218e-05, "loss": 0.0, "step": 4948 }, { "epoch": 1.14, "learning_rate": 3.09375240736461e-05, "loss": 0.0, "step": 4949 }, { "epoch": 1.14, "learning_rate": 3.09336722902704e-05, "loss": 0.0, "step": 4950 }, { "epoch": 1.14, "learning_rate": 3.092982050689469e-05, "loss": 0.0, "step": 4951 }, { "epoch": 1.14, "learning_rate": 3.092596872351899e-05, "loss": 0.0, "step": 4952 }, { "epoch": 1.14, "learning_rate": 3.0922116940143285e-05, "loss": 0.0, "step": 4953 }, { "epoch": 1.14, "learning_rate": 3.091826515676759e-05, "loss": 0.0, "step": 4954 }, { "epoch": 1.15, "learning_rate": 3.091441337339188e-05, "loss": 0.0, "step": 4955 }, { "epoch": 1.15, "learning_rate": 3.0910561590016177e-05, "loss": 0.0, "step": 4956 }, { "epoch": 1.15, "learning_rate": 3.090670980664048e-05, "loss": 0.0, "step": 4957 }, { "epoch": 1.15, "learning_rate": 3.090285802326477e-05, "loss": 0.0, "step": 4958 }, { "epoch": 1.15, "learning_rate": 3.0899006239889075e-05, "loss": 0.0, "step": 4959 }, { "epoch": 1.15, "learning_rate": 3.089515445651337e-05, "loss": 0.0, "step": 4960 }, { "epoch": 1.15, "learning_rate": 3.0891302673137665e-05, "loss": 0.0, "step": 4961 }, { "epoch": 1.15, "learning_rate": 3.088745088976196e-05, "loss": 0.0, "step": 4962 }, { "epoch": 1.15, "learning_rate": 3.0883599106386255e-05, "loss": 0.0, "step": 4963 }, { "epoch": 1.15, "learning_rate": 3.0879747323010557e-05, "loss": 0.0, "step": 4964 }, { "epoch": 1.15, "learning_rate": 3.087589553963485e-05, "loss": 0.0, "step": 4965 }, { "epoch": 1.15, "learning_rate": 3.087204375625915e-05, "loss": 0.0, "step": 4966 }, { "epoch": 1.15, "learning_rate": 3.086819197288345e-05, "loss": 0.0, "step": 4967 }, { "epoch": 1.15, "learning_rate": 3.086434018950774e-05, "loss": 0.0, "step": 4968 }, { "epoch": 1.15, "learning_rate": 3.0860488406132045e-05, "loss": 0.0, "step": 4969 }, { "epoch": 1.15, "learning_rate": 3.085663662275633e-05, "loss": 0.0, "step": 4970 }, { "epoch": 1.15, "learning_rate": 3.0852784839380635e-05, "loss": 0.0, "step": 4971 }, { "epoch": 1.15, "learning_rate": 3.084893305600493e-05, "loss": 0.0, "step": 4972 }, { "epoch": 1.15, "learning_rate": 3.084508127262923e-05, "loss": 0.0, "step": 4973 }, { "epoch": 1.15, "learning_rate": 3.0841229489253526e-05, "loss": 0.0, "step": 4974 }, { "epoch": 1.15, "learning_rate": 3.083737770587782e-05, "loss": 0.0, "step": 4975 }, { "epoch": 1.15, "learning_rate": 3.083352592250212e-05, "loss": 0.0, "step": 4976 }, { "epoch": 1.15, "learning_rate": 3.082967413912642e-05, "loss": 0.0, "step": 4977 }, { "epoch": 1.15, "learning_rate": 3.082582235575072e-05, "loss": 0.0, "step": 4978 }, { "epoch": 1.15, "learning_rate": 3.082197057237501e-05, "loss": 0.0, "step": 4979 }, { "epoch": 1.15, "learning_rate": 3.08181187889993e-05, "loss": 0.0, "step": 4980 }, { "epoch": 1.15, "learning_rate": 3.0814267005623605e-05, "loss": 0.0, "step": 4981 }, { "epoch": 1.15, "learning_rate": 3.08104152222479e-05, "loss": 0.0, "step": 4982 }, { "epoch": 1.15, "learning_rate": 3.08065634388722e-05, "loss": 0.0, "step": 4983 }, { "epoch": 1.15, "learning_rate": 3.0802711655496496e-05, "loss": 0.0, "step": 4984 }, { "epoch": 1.15, "learning_rate": 3.079885987212079e-05, "loss": 0.0, "step": 4985 }, { "epoch": 1.15, "learning_rate": 3.079500808874509e-05, "loss": 0.0, "step": 4986 }, { "epoch": 1.15, "learning_rate": 3.079115630536939e-05, "loss": 0.0, "step": 4987 }, { "epoch": 1.15, "learning_rate": 3.078730452199369e-05, "loss": 0.0, "step": 4988 }, { "epoch": 1.15, "learning_rate": 3.078345273861798e-05, "loss": 0.0, "step": 4989 }, { "epoch": 1.15, "learning_rate": 3.077960095524228e-05, "loss": 0.0001, "step": 4990 }, { "epoch": 1.15, "learning_rate": 3.0775749171866575e-05, "loss": 0.0, "step": 4991 }, { "epoch": 1.15, "learning_rate": 3.077189738849087e-05, "loss": 0.0, "step": 4992 }, { "epoch": 1.15, "learning_rate": 3.076804560511517e-05, "loss": 0.0, "step": 4993 }, { "epoch": 1.15, "learning_rate": 3.0764193821739466e-05, "loss": 0.0, "step": 4994 }, { "epoch": 1.15, "learning_rate": 3.076034203836377e-05, "loss": 0.0, "step": 4995 }, { "epoch": 1.15, "learning_rate": 3.075649025498806e-05, "loss": 0.0, "step": 4996 }, { "epoch": 1.15, "learning_rate": 3.075263847161236e-05, "loss": 0.0, "step": 4997 }, { "epoch": 1.16, "learning_rate": 3.074878668823665e-05, "loss": 0.0, "step": 4998 }, { "epoch": 1.16, "learning_rate": 3.074493490486095e-05, "loss": 0.0, "step": 4999 }, { "epoch": 1.16, "learning_rate": 3.074108312148525e-05, "loss": 0.0224, "step": 5000 }, { "epoch": 1.16, "learning_rate": 3.0737231338109544e-05, "loss": 0.0, "step": 5001 }, { "epoch": 1.16, "learning_rate": 3.0733379554733846e-05, "loss": 0.0, "step": 5002 }, { "epoch": 1.16, "learning_rate": 3.072952777135814e-05, "loss": 0.0, "step": 5003 }, { "epoch": 1.16, "learning_rate": 3.0725675987982436e-05, "loss": 0.0, "step": 5004 }, { "epoch": 1.16, "learning_rate": 3.072182420460674e-05, "loss": 0.0, "step": 5005 }, { "epoch": 1.16, "learning_rate": 3.071797242123103e-05, "loss": 0.0, "step": 5006 }, { "epoch": 1.16, "learning_rate": 3.071412063785533e-05, "loss": 0.0, "step": 5007 }, { "epoch": 1.16, "learning_rate": 3.071026885447962e-05, "loss": 0.0001, "step": 5008 }, { "epoch": 1.16, "learning_rate": 3.0706417071103924e-05, "loss": 0.0, "step": 5009 }, { "epoch": 1.16, "learning_rate": 3.070256528772822e-05, "loss": 0.0, "step": 5010 }, { "epoch": 1.16, "learning_rate": 3.0698713504352514e-05, "loss": 0.0, "step": 5011 }, { "epoch": 1.16, "learning_rate": 3.0694861720976816e-05, "loss": 0.0, "step": 5012 }, { "epoch": 1.16, "learning_rate": 3.069100993760111e-05, "loss": 0.0, "step": 5013 }, { "epoch": 1.16, "learning_rate": 3.068715815422541e-05, "loss": 0.0, "step": 5014 }, { "epoch": 1.16, "learning_rate": 3.068330637084971e-05, "loss": 0.0, "step": 5015 }, { "epoch": 1.16, "learning_rate": 3.0679454587474e-05, "loss": 0.0, "step": 5016 }, { "epoch": 1.16, "learning_rate": 3.06756028040983e-05, "loss": 0.0, "step": 5017 }, { "epoch": 1.16, "learning_rate": 3.067175102072259e-05, "loss": 0.0, "step": 5018 }, { "epoch": 1.16, "learning_rate": 3.0667899237346894e-05, "loss": 0.0001, "step": 5019 }, { "epoch": 1.16, "learning_rate": 3.066404745397119e-05, "loss": 0.0, "step": 5020 }, { "epoch": 1.16, "learning_rate": 3.0660195670595484e-05, "loss": 0.0, "step": 5021 }, { "epoch": 1.16, "learning_rate": 3.0656343887219786e-05, "loss": 0.0, "step": 5022 }, { "epoch": 1.16, "learning_rate": 3.065249210384408e-05, "loss": 0.0, "step": 5023 }, { "epoch": 1.16, "learning_rate": 3.064864032046838e-05, "loss": 0.0, "step": 5024 }, { "epoch": 1.16, "learning_rate": 3.064478853709268e-05, "loss": 0.0003, "step": 5025 }, { "epoch": 1.16, "learning_rate": 3.064093675371697e-05, "loss": 0.0, "step": 5026 }, { "epoch": 1.16, "learning_rate": 3.063708497034127e-05, "loss": 0.0, "step": 5027 }, { "epoch": 1.16, "learning_rate": 3.063323318696556e-05, "loss": 0.0001, "step": 5028 }, { "epoch": 1.16, "learning_rate": 3.0629381403589864e-05, "loss": 0.0, "step": 5029 }, { "epoch": 1.16, "learning_rate": 3.062552962021416e-05, "loss": 0.0, "step": 5030 }, { "epoch": 1.16, "learning_rate": 3.062167783683846e-05, "loss": 0.0, "step": 5031 }, { "epoch": 1.16, "learning_rate": 3.0617826053462756e-05, "loss": 0.0, "step": 5032 }, { "epoch": 1.16, "learning_rate": 3.061397427008705e-05, "loss": 0.0011, "step": 5033 }, { "epoch": 1.16, "learning_rate": 3.061012248671135e-05, "loss": 0.0, "step": 5034 }, { "epoch": 1.16, "learning_rate": 3.060627070333564e-05, "loss": 0.0, "step": 5035 }, { "epoch": 1.16, "learning_rate": 3.060241891995994e-05, "loss": 0.0, "step": 5036 }, { "epoch": 1.16, "learning_rate": 3.059856713658424e-05, "loss": 0.0001, "step": 5037 }, { "epoch": 1.16, "learning_rate": 3.059471535320854e-05, "loss": 0.0, "step": 5038 }, { "epoch": 1.16, "learning_rate": 3.0590863569832834e-05, "loss": 0.0, "step": 5039 }, { "epoch": 1.16, "learning_rate": 3.058701178645713e-05, "loss": 0.0, "step": 5040 }, { "epoch": 1.17, "learning_rate": 3.058316000308143e-05, "loss": 0.0001, "step": 5041 }, { "epoch": 1.17, "learning_rate": 3.0579308219705726e-05, "loss": 0.0, "step": 5042 }, { "epoch": 1.17, "learning_rate": 3.057545643633003e-05, "loss": 0.0, "step": 5043 }, { "epoch": 1.17, "learning_rate": 3.057160465295432e-05, "loss": 0.0, "step": 5044 }, { "epoch": 1.17, "learning_rate": 3.056775286957862e-05, "loss": 0.0, "step": 5045 }, { "epoch": 1.17, "learning_rate": 3.056390108620291e-05, "loss": 0.0, "step": 5046 }, { "epoch": 1.17, "learning_rate": 3.056004930282721e-05, "loss": 0.0, "step": 5047 }, { "epoch": 1.17, "learning_rate": 3.055619751945151e-05, "loss": 0.0, "step": 5048 }, { "epoch": 1.17, "learning_rate": 3.0552345736075804e-05, "loss": 0.0, "step": 5049 }, { "epoch": 1.17, "learning_rate": 3.0548493952700106e-05, "loss": 0.0, "step": 5050 }, { "epoch": 1.17, "learning_rate": 3.05446421693244e-05, "loss": 0.0002, "step": 5051 }, { "epoch": 1.17, "learning_rate": 3.0540790385948696e-05, "loss": 0.0, "step": 5052 }, { "epoch": 1.17, "learning_rate": 3.0536938602573e-05, "loss": 0.0, "step": 5053 }, { "epoch": 1.17, "learning_rate": 3.0533086819197285e-05, "loss": 0.0, "step": 5054 }, { "epoch": 1.17, "learning_rate": 3.052923503582159e-05, "loss": 0.0, "step": 5055 }, { "epoch": 1.17, "learning_rate": 3.052538325244588e-05, "loss": 0.0, "step": 5056 }, { "epoch": 1.17, "learning_rate": 3.052153146907018e-05, "loss": 0.0, "step": 5057 }, { "epoch": 1.17, "learning_rate": 3.051767968569448e-05, "loss": 0.0, "step": 5058 }, { "epoch": 1.17, "learning_rate": 3.0513827902318777e-05, "loss": 0.0, "step": 5059 }, { "epoch": 1.17, "learning_rate": 3.0509976118943072e-05, "loss": 0.0, "step": 5060 }, { "epoch": 1.17, "learning_rate": 3.050612433556737e-05, "loss": 0.0001, "step": 5061 }, { "epoch": 1.17, "learning_rate": 3.050227255219167e-05, "loss": 0.0001, "step": 5062 }, { "epoch": 1.17, "learning_rate": 3.049842076881596e-05, "loss": 0.0, "step": 5063 }, { "epoch": 1.17, "learning_rate": 3.049456898544026e-05, "loss": 0.0, "step": 5064 }, { "epoch": 1.17, "learning_rate": 3.0490717202064557e-05, "loss": 0.0, "step": 5065 }, { "epoch": 1.17, "learning_rate": 3.0486865418688852e-05, "loss": 0.0598, "step": 5066 }, { "epoch": 1.17, "learning_rate": 3.048301363531315e-05, "loss": 0.0, "step": 5067 }, { "epoch": 1.17, "learning_rate": 3.047916185193745e-05, "loss": 0.0, "step": 5068 }, { "epoch": 1.17, "learning_rate": 3.0475310068561747e-05, "loss": 0.0, "step": 5069 }, { "epoch": 1.17, "learning_rate": 3.0471458285186045e-05, "loss": 0.0, "step": 5070 }, { "epoch": 1.17, "learning_rate": 3.046760650181034e-05, "loss": 0.0, "step": 5071 }, { "epoch": 1.17, "learning_rate": 3.046375471843464e-05, "loss": 0.0, "step": 5072 }, { "epoch": 1.17, "learning_rate": 3.045990293505893e-05, "loss": 0.0, "step": 5073 }, { "epoch": 1.17, "learning_rate": 3.045605115168323e-05, "loss": 0.0007, "step": 5074 }, { "epoch": 1.17, "learning_rate": 3.0452199368307527e-05, "loss": 0.0, "step": 5075 }, { "epoch": 1.17, "learning_rate": 3.0448347584931825e-05, "loss": 0.0, "step": 5076 }, { "epoch": 1.17, "learning_rate": 3.0444495801556124e-05, "loss": 0.0, "step": 5077 }, { "epoch": 1.17, "learning_rate": 3.044064401818042e-05, "loss": 0.0, "step": 5078 }, { "epoch": 1.17, "learning_rate": 3.0436792234804717e-05, "loss": 0.0, "step": 5079 }, { "epoch": 1.17, "learning_rate": 3.0432940451429015e-05, "loss": 0.0, "step": 5080 }, { "epoch": 1.17, "learning_rate": 3.0429088668053314e-05, "loss": 0.0, "step": 5081 }, { "epoch": 1.17, "learning_rate": 3.0425236884677605e-05, "loss": 0.0, "step": 5082 }, { "epoch": 1.17, "learning_rate": 3.0421385101301904e-05, "loss": 0.0001, "step": 5083 }, { "epoch": 1.17, "learning_rate": 3.04175333179262e-05, "loss": 0.0, "step": 5084 }, { "epoch": 1.18, "learning_rate": 3.0413681534550497e-05, "loss": 0.0, "step": 5085 }, { "epoch": 1.18, "learning_rate": 3.0409829751174795e-05, "loss": 0.0, "step": 5086 }, { "epoch": 1.18, "learning_rate": 3.0405977967799094e-05, "loss": 0.0002, "step": 5087 }, { "epoch": 1.18, "learning_rate": 3.0402126184423392e-05, "loss": 0.0, "step": 5088 }, { "epoch": 1.18, "learning_rate": 3.0398274401047687e-05, "loss": 0.0, "step": 5089 }, { "epoch": 1.18, "learning_rate": 3.0394422617671985e-05, "loss": 0.0001, "step": 5090 }, { "epoch": 1.18, "learning_rate": 3.0390570834296277e-05, "loss": 0.0, "step": 5091 }, { "epoch": 1.18, "learning_rate": 3.0386719050920575e-05, "loss": 0.0, "step": 5092 }, { "epoch": 1.18, "learning_rate": 3.0382867267544873e-05, "loss": 0.0, "step": 5093 }, { "epoch": 1.18, "learning_rate": 3.0379015484169172e-05, "loss": 0.0, "step": 5094 }, { "epoch": 1.18, "learning_rate": 3.037516370079347e-05, "loss": 0.0004, "step": 5095 }, { "epoch": 1.18, "learning_rate": 3.0371311917417765e-05, "loss": 0.0, "step": 5096 }, { "epoch": 1.18, "learning_rate": 3.0367460134042063e-05, "loss": 0.0001, "step": 5097 }, { "epoch": 1.18, "learning_rate": 3.0363608350666362e-05, "loss": 0.0, "step": 5098 }, { "epoch": 1.18, "learning_rate": 3.035975656729066e-05, "loss": 0.0, "step": 5099 }, { "epoch": 1.18, "learning_rate": 3.035590478391496e-05, "loss": 0.0, "step": 5100 }, { "epoch": 1.18, "learning_rate": 3.035205300053925e-05, "loss": 0.0, "step": 5101 }, { "epoch": 1.18, "learning_rate": 3.0348201217163545e-05, "loss": 0.0, "step": 5102 }, { "epoch": 1.18, "learning_rate": 3.0344349433787843e-05, "loss": 0.0001, "step": 5103 }, { "epoch": 1.18, "learning_rate": 3.0340497650412142e-05, "loss": 0.0, "step": 5104 }, { "epoch": 1.18, "learning_rate": 3.033664586703644e-05, "loss": 0.0, "step": 5105 }, { "epoch": 1.18, "learning_rate": 3.033279408366074e-05, "loss": 0.0, "step": 5106 }, { "epoch": 1.18, "learning_rate": 3.0328942300285033e-05, "loss": 0.0, "step": 5107 }, { "epoch": 1.18, "learning_rate": 3.032509051690933e-05, "loss": 0.0, "step": 5108 }, { "epoch": 1.18, "learning_rate": 3.032123873353363e-05, "loss": 0.0, "step": 5109 }, { "epoch": 1.18, "learning_rate": 3.031738695015792e-05, "loss": 0.0, "step": 5110 }, { "epoch": 1.18, "learning_rate": 3.031353516678222e-05, "loss": 0.0, "step": 5111 }, { "epoch": 1.18, "learning_rate": 3.0309683383406518e-05, "loss": 0.0, "step": 5112 }, { "epoch": 1.18, "learning_rate": 3.0305831600030817e-05, "loss": 0.0, "step": 5113 }, { "epoch": 1.18, "learning_rate": 3.030197981665511e-05, "loss": 0.0, "step": 5114 }, { "epoch": 1.18, "learning_rate": 3.029812803327941e-05, "loss": 0.0001, "step": 5115 }, { "epoch": 1.18, "learning_rate": 3.0294276249903708e-05, "loss": 0.0, "step": 5116 }, { "epoch": 1.18, "learning_rate": 3.0290424466528007e-05, "loss": 0.0, "step": 5117 }, { "epoch": 1.18, "learning_rate": 3.0286572683152305e-05, "loss": 0.0, "step": 5118 }, { "epoch": 1.18, "learning_rate": 3.0282720899776597e-05, "loss": 0.0, "step": 5119 }, { "epoch": 1.18, "learning_rate": 3.027886911640089e-05, "loss": 0.0, "step": 5120 }, { "epoch": 1.18, "learning_rate": 3.027501733302519e-05, "loss": 0.0, "step": 5121 }, { "epoch": 1.18, "learning_rate": 3.0271165549649488e-05, "loss": 0.0, "step": 5122 }, { "epoch": 1.18, "learning_rate": 3.0267313766273787e-05, "loss": 0.0, "step": 5123 }, { "epoch": 1.18, "learning_rate": 3.0263461982898085e-05, "loss": 0.0, "step": 5124 }, { "epoch": 1.18, "learning_rate": 3.025961019952238e-05, "loss": 0.0, "step": 5125 }, { "epoch": 1.18, "learning_rate": 3.0255758416146678e-05, "loss": 0.0, "step": 5126 }, { "epoch": 1.18, "learning_rate": 3.0251906632770976e-05, "loss": 0.0, "step": 5127 }, { "epoch": 1.19, "learning_rate": 3.0248054849395275e-05, "loss": 0.0, "step": 5128 }, { "epoch": 1.19, "learning_rate": 3.0244203066019566e-05, "loss": 0.0, "step": 5129 }, { "epoch": 1.19, "learning_rate": 3.0240351282643865e-05, "loss": 0.0, "step": 5130 }, { "epoch": 1.19, "learning_rate": 3.0236499499268163e-05, "loss": 0.0, "step": 5131 }, { "epoch": 1.19, "learning_rate": 3.0232647715892458e-05, "loss": 0.0, "step": 5132 }, { "epoch": 1.19, "learning_rate": 3.0228795932516756e-05, "loss": 0.0, "step": 5133 }, { "epoch": 1.19, "learning_rate": 3.0224944149141055e-05, "loss": 0.0, "step": 5134 }, { "epoch": 1.19, "learning_rate": 3.0221092365765353e-05, "loss": 0.0001, "step": 5135 }, { "epoch": 1.19, "learning_rate": 3.021724058238965e-05, "loss": 0.0, "step": 5136 }, { "epoch": 1.19, "learning_rate": 3.0213388799013946e-05, "loss": 0.0, "step": 5137 }, { "epoch": 1.19, "learning_rate": 3.0209537015638238e-05, "loss": 0.0, "step": 5138 }, { "epoch": 1.19, "learning_rate": 3.0205685232262536e-05, "loss": 0.0, "step": 5139 }, { "epoch": 1.19, "learning_rate": 3.0201833448886835e-05, "loss": 0.0, "step": 5140 }, { "epoch": 1.19, "learning_rate": 3.0197981665511133e-05, "loss": 0.0, "step": 5141 }, { "epoch": 1.19, "learning_rate": 3.019412988213543e-05, "loss": 0.0, "step": 5142 }, { "epoch": 1.19, "learning_rate": 3.0190278098759726e-05, "loss": 0.0, "step": 5143 }, { "epoch": 1.19, "learning_rate": 3.0186426315384025e-05, "loss": 0.0, "step": 5144 }, { "epoch": 1.19, "learning_rate": 3.0182574532008323e-05, "loss": 0.0, "step": 5145 }, { "epoch": 1.19, "learning_rate": 3.017872274863262e-05, "loss": 0.0, "step": 5146 }, { "epoch": 1.19, "learning_rate": 3.0174870965256913e-05, "loss": 0.0, "step": 5147 }, { "epoch": 1.19, "learning_rate": 3.017101918188121e-05, "loss": 0.0, "step": 5148 }, { "epoch": 1.19, "learning_rate": 3.016716739850551e-05, "loss": 0.0001, "step": 5149 }, { "epoch": 1.19, "learning_rate": 3.0163315615129805e-05, "loss": 0.0, "step": 5150 }, { "epoch": 1.19, "learning_rate": 3.0159463831754103e-05, "loss": 0.0, "step": 5151 }, { "epoch": 1.19, "learning_rate": 3.01556120483784e-05, "loss": 0.0, "step": 5152 }, { "epoch": 1.19, "learning_rate": 3.01517602650027e-05, "loss": 0.0, "step": 5153 }, { "epoch": 1.19, "learning_rate": 3.0147908481626998e-05, "loss": 0.0, "step": 5154 }, { "epoch": 1.19, "learning_rate": 3.0144056698251293e-05, "loss": 0.0, "step": 5155 }, { "epoch": 1.19, "learning_rate": 3.014020491487559e-05, "loss": 0.0002, "step": 5156 }, { "epoch": 1.19, "learning_rate": 3.0136353131499883e-05, "loss": 0.0, "step": 5157 }, { "epoch": 1.19, "learning_rate": 3.013250134812418e-05, "loss": 0.0, "step": 5158 }, { "epoch": 1.19, "learning_rate": 3.012864956474848e-05, "loss": 0.0, "step": 5159 }, { "epoch": 1.19, "learning_rate": 3.0124797781372778e-05, "loss": 0.0, "step": 5160 }, { "epoch": 1.19, "learning_rate": 3.0120945997997073e-05, "loss": 0.0002, "step": 5161 }, { "epoch": 1.19, "learning_rate": 3.011709421462137e-05, "loss": 0.0, "step": 5162 }, { "epoch": 1.19, "learning_rate": 3.011324243124567e-05, "loss": 0.0, "step": 5163 }, { "epoch": 1.19, "learning_rate": 3.0109390647869968e-05, "loss": 0.0, "step": 5164 }, { "epoch": 1.19, "learning_rate": 3.0105538864494266e-05, "loss": 0.0, "step": 5165 }, { "epoch": 1.19, "learning_rate": 3.0101687081118558e-05, "loss": 0.0, "step": 5166 }, { "epoch": 1.19, "learning_rate": 3.0097835297742856e-05, "loss": 0.0, "step": 5167 }, { "epoch": 1.19, "learning_rate": 3.009398351436715e-05, "loss": 0.0, "step": 5168 }, { "epoch": 1.19, "learning_rate": 3.009013173099145e-05, "loss": 0.0, "step": 5169 }, { "epoch": 1.19, "learning_rate": 3.0086279947615748e-05, "loss": 0.0, "step": 5170 }, { "epoch": 1.2, "learning_rate": 3.0082428164240046e-05, "loss": 0.0, "step": 5171 }, { "epoch": 1.2, "learning_rate": 3.0078576380864344e-05, "loss": 0.0, "step": 5172 }, { "epoch": 1.2, "learning_rate": 3.007472459748864e-05, "loss": 0.0, "step": 5173 }, { "epoch": 1.2, "learning_rate": 3.0070872814112938e-05, "loss": 0.0, "step": 5174 }, { "epoch": 1.2, "learning_rate": 3.006702103073723e-05, "loss": 0.0, "step": 5175 }, { "epoch": 1.2, "learning_rate": 3.0063169247361528e-05, "loss": 0.0, "step": 5176 }, { "epoch": 1.2, "learning_rate": 3.0059317463985826e-05, "loss": 0.0, "step": 5177 }, { "epoch": 1.2, "learning_rate": 3.0055465680610124e-05, "loss": 0.0, "step": 5178 }, { "epoch": 1.2, "learning_rate": 3.005161389723442e-05, "loss": 0.0, "step": 5179 }, { "epoch": 1.2, "learning_rate": 3.0047762113858718e-05, "loss": 0.0, "step": 5180 }, { "epoch": 1.2, "learning_rate": 3.0043910330483016e-05, "loss": 0.0, "step": 5181 }, { "epoch": 1.2, "learning_rate": 3.0040058547107314e-05, "loss": 0.0, "step": 5182 }, { "epoch": 1.2, "learning_rate": 3.0036206763731613e-05, "loss": 0.0, "step": 5183 }, { "epoch": 1.2, "learning_rate": 3.0032354980355908e-05, "loss": 0.0, "step": 5184 }, { "epoch": 1.2, "learning_rate": 3.0028503196980203e-05, "loss": 0.0, "step": 5185 }, { "epoch": 1.2, "learning_rate": 3.0024651413604497e-05, "loss": 0.0, "step": 5186 }, { "epoch": 1.2, "learning_rate": 3.0020799630228796e-05, "loss": 0.0, "step": 5187 }, { "epoch": 1.2, "learning_rate": 3.0016947846853094e-05, "loss": 0.0, "step": 5188 }, { "epoch": 1.2, "learning_rate": 3.0013096063477392e-05, "loss": 0.0, "step": 5189 }, { "epoch": 1.2, "learning_rate": 3.000924428010169e-05, "loss": 0.0, "step": 5190 }, { "epoch": 1.2, "learning_rate": 3.0005392496725986e-05, "loss": 0.0, "step": 5191 }, { "epoch": 1.2, "learning_rate": 3.0001540713350284e-05, "loss": 0.0, "step": 5192 }, { "epoch": 1.2, "learning_rate": 2.9997688929974582e-05, "loss": 0.0, "step": 5193 }, { "epoch": 1.2, "learning_rate": 2.9993837146598874e-05, "loss": 0.0, "step": 5194 }, { "epoch": 1.2, "learning_rate": 2.9989985363223172e-05, "loss": 0.0, "step": 5195 }, { "epoch": 1.2, "learning_rate": 2.998613357984747e-05, "loss": 0.0, "step": 5196 }, { "epoch": 1.2, "learning_rate": 2.9982281796471766e-05, "loss": 0.0, "step": 5197 }, { "epoch": 1.2, "learning_rate": 2.9978430013096064e-05, "loss": 0.0, "step": 5198 }, { "epoch": 1.2, "learning_rate": 2.9974578229720362e-05, "loss": 0.0, "step": 5199 }, { "epoch": 1.2, "learning_rate": 2.997072644634466e-05, "loss": 0.0001, "step": 5200 }, { "epoch": 1.2, "learning_rate": 2.996687466296896e-05, "loss": 0.0, "step": 5201 }, { "epoch": 1.2, "learning_rate": 2.9963022879593254e-05, "loss": 0.0, "step": 5202 }, { "epoch": 1.2, "learning_rate": 2.995917109621755e-05, "loss": 0.0, "step": 5203 }, { "epoch": 1.2, "learning_rate": 2.9955319312841844e-05, "loss": 0.0, "step": 5204 }, { "epoch": 1.2, "learning_rate": 2.9951467529466142e-05, "loss": 0.0, "step": 5205 }, { "epoch": 1.2, "learning_rate": 2.994761574609044e-05, "loss": 0.0, "step": 5206 }, { "epoch": 1.2, "learning_rate": 2.994376396271474e-05, "loss": 0.0, "step": 5207 }, { "epoch": 1.2, "learning_rate": 2.9939912179339037e-05, "loss": 0.0, "step": 5208 }, { "epoch": 1.2, "learning_rate": 2.9936060395963332e-05, "loss": 0.0, "step": 5209 }, { "epoch": 1.2, "learning_rate": 2.993220861258763e-05, "loss": 0.0, "step": 5210 }, { "epoch": 1.2, "learning_rate": 2.992835682921193e-05, "loss": 0.0, "step": 5211 }, { "epoch": 1.2, "learning_rate": 2.992450504583622e-05, "loss": 0.0, "step": 5212 }, { "epoch": 1.2, "learning_rate": 2.992065326246052e-05, "loss": 0.0, "step": 5213 }, { "epoch": 1.2, "learning_rate": 2.9916801479084817e-05, "loss": 0.0003, "step": 5214 }, { "epoch": 1.21, "learning_rate": 2.9912949695709112e-05, "loss": 0.0, "step": 5215 }, { "epoch": 1.21, "learning_rate": 2.990909791233341e-05, "loss": 0.0, "step": 5216 }, { "epoch": 1.21, "learning_rate": 2.990524612895771e-05, "loss": 0.0, "step": 5217 }, { "epoch": 1.21, "learning_rate": 2.9901394345582007e-05, "loss": 0.0, "step": 5218 }, { "epoch": 1.21, "learning_rate": 2.9897542562206306e-05, "loss": 0.0, "step": 5219 }, { "epoch": 1.21, "learning_rate": 2.98936907788306e-05, "loss": 0.0, "step": 5220 }, { "epoch": 1.21, "learning_rate": 2.98898389954549e-05, "loss": 0.0, "step": 5221 }, { "epoch": 1.21, "learning_rate": 2.988598721207919e-05, "loss": 0.0, "step": 5222 }, { "epoch": 1.21, "learning_rate": 2.988213542870349e-05, "loss": 0.0, "step": 5223 }, { "epoch": 1.21, "learning_rate": 2.9878283645327787e-05, "loss": 0.0, "step": 5224 }, { "epoch": 1.21, "learning_rate": 2.9874431861952085e-05, "loss": 0.0, "step": 5225 }, { "epoch": 1.21, "learning_rate": 2.9870580078576384e-05, "loss": 0.0, "step": 5226 }, { "epoch": 1.21, "learning_rate": 2.986672829520068e-05, "loss": 0.0, "step": 5227 }, { "epoch": 1.21, "learning_rate": 2.9862876511824977e-05, "loss": 0.0, "step": 5228 }, { "epoch": 1.21, "learning_rate": 2.9859024728449275e-05, "loss": 0.0, "step": 5229 }, { "epoch": 1.21, "learning_rate": 2.9855172945073574e-05, "loss": 0.0, "step": 5230 }, { "epoch": 1.21, "learning_rate": 2.9851321161697865e-05, "loss": 0.0, "step": 5231 }, { "epoch": 1.21, "learning_rate": 2.9847469378322164e-05, "loss": 0.0, "step": 5232 }, { "epoch": 1.21, "learning_rate": 2.984361759494646e-05, "loss": 0.0, "step": 5233 }, { "epoch": 1.21, "learning_rate": 2.9839765811570757e-05, "loss": 0.0, "step": 5234 }, { "epoch": 1.21, "learning_rate": 2.9835914028195055e-05, "loss": 0.0, "step": 5235 }, { "epoch": 1.21, "learning_rate": 2.9832062244819354e-05, "loss": 0.0, "step": 5236 }, { "epoch": 1.21, "learning_rate": 2.9828210461443652e-05, "loss": 0.0, "step": 5237 }, { "epoch": 1.21, "learning_rate": 2.9824358678067947e-05, "loss": 0.0, "step": 5238 }, { "epoch": 1.21, "learning_rate": 2.9820506894692245e-05, "loss": 0.0, "step": 5239 }, { "epoch": 1.21, "learning_rate": 2.9816655111316537e-05, "loss": 0.0, "step": 5240 }, { "epoch": 1.21, "learning_rate": 2.9812803327940835e-05, "loss": 0.0, "step": 5241 }, { "epoch": 1.21, "learning_rate": 2.9808951544565134e-05, "loss": 0.0, "step": 5242 }, { "epoch": 1.21, "learning_rate": 2.9805099761189432e-05, "loss": 0.0, "step": 5243 }, { "epoch": 1.21, "learning_rate": 2.980124797781373e-05, "loss": 0.0, "step": 5244 }, { "epoch": 1.21, "learning_rate": 2.9797396194438025e-05, "loss": 0.0, "step": 5245 }, { "epoch": 1.21, "learning_rate": 2.9793544411062324e-05, "loss": 0.0, "step": 5246 }, { "epoch": 1.21, "learning_rate": 2.9789692627686622e-05, "loss": 0.0, "step": 5247 }, { "epoch": 1.21, "learning_rate": 2.978584084431092e-05, "loss": 0.0, "step": 5248 }, { "epoch": 1.21, "learning_rate": 2.978198906093522e-05, "loss": 0.0, "step": 5249 }, { "epoch": 1.21, "learning_rate": 2.977813727755951e-05, "loss": 0.0, "step": 5250 }, { "epoch": 1.21, "learning_rate": 2.9774285494183805e-05, "loss": 0.0, "step": 5251 }, { "epoch": 1.21, "learning_rate": 2.9770433710808103e-05, "loss": 0.0, "step": 5252 }, { "epoch": 1.21, "learning_rate": 2.9766581927432402e-05, "loss": 0.0, "step": 5253 }, { "epoch": 1.21, "learning_rate": 2.97627301440567e-05, "loss": 0.0, "step": 5254 }, { "epoch": 1.21, "learning_rate": 2.9758878360681e-05, "loss": 0.0, "step": 5255 }, { "epoch": 1.21, "learning_rate": 2.9755026577305293e-05, "loss": 0.0, "step": 5256 }, { "epoch": 1.21, "learning_rate": 2.9751174793929592e-05, "loss": 0.0, "step": 5257 }, { "epoch": 1.22, "learning_rate": 2.974732301055389e-05, "loss": 0.0, "step": 5258 }, { "epoch": 1.22, "learning_rate": 2.974347122717818e-05, "loss": 0.0, "step": 5259 }, { "epoch": 1.22, "learning_rate": 2.973961944380248e-05, "loss": 0.0, "step": 5260 }, { "epoch": 1.22, "learning_rate": 2.973576766042678e-05, "loss": 0.0, "step": 5261 }, { "epoch": 1.22, "learning_rate": 2.9731915877051077e-05, "loss": 0.0, "step": 5262 }, { "epoch": 1.22, "learning_rate": 2.972806409367537e-05, "loss": 0.0, "step": 5263 }, { "epoch": 1.22, "learning_rate": 2.972421231029967e-05, "loss": 0.0, "step": 5264 }, { "epoch": 1.22, "learning_rate": 2.972036052692397e-05, "loss": 0.0, "step": 5265 }, { "epoch": 1.22, "learning_rate": 2.9716508743548267e-05, "loss": 0.0, "step": 5266 }, { "epoch": 1.22, "learning_rate": 2.9712656960172565e-05, "loss": 0.0, "step": 5267 }, { "epoch": 1.22, "learning_rate": 2.9708805176796857e-05, "loss": 0.0, "step": 5268 }, { "epoch": 1.22, "learning_rate": 2.970495339342115e-05, "loss": 0.0, "step": 5269 }, { "epoch": 1.22, "learning_rate": 2.970110161004545e-05, "loss": 0.0, "step": 5270 }, { "epoch": 1.22, "learning_rate": 2.9697249826669748e-05, "loss": 0.0, "step": 5271 }, { "epoch": 1.22, "learning_rate": 2.9693398043294047e-05, "loss": 0.0, "step": 5272 }, { "epoch": 1.22, "learning_rate": 2.9689546259918345e-05, "loss": 0.0, "step": 5273 }, { "epoch": 1.22, "learning_rate": 2.968569447654264e-05, "loss": 0.0, "step": 5274 }, { "epoch": 1.22, "learning_rate": 2.9681842693166938e-05, "loss": 0.0, "step": 5275 }, { "epoch": 1.22, "learning_rate": 2.9677990909791237e-05, "loss": 0.0, "step": 5276 }, { "epoch": 1.22, "learning_rate": 2.9674139126415535e-05, "loss": 0.0, "step": 5277 }, { "epoch": 1.22, "learning_rate": 2.9670287343039826e-05, "loss": 0.0, "step": 5278 }, { "epoch": 1.22, "learning_rate": 2.9666435559664125e-05, "loss": 0.0, "step": 5279 }, { "epoch": 1.22, "learning_rate": 2.9662583776288423e-05, "loss": 0.0, "step": 5280 }, { "epoch": 1.22, "learning_rate": 2.9658731992912718e-05, "loss": 0.0, "step": 5281 }, { "epoch": 1.22, "learning_rate": 2.9654880209537016e-05, "loss": 0.0, "step": 5282 }, { "epoch": 1.22, "learning_rate": 2.9651028426161315e-05, "loss": 0.0, "step": 5283 }, { "epoch": 1.22, "learning_rate": 2.9647176642785613e-05, "loss": 0.0, "step": 5284 }, { "epoch": 1.22, "learning_rate": 2.964332485940991e-05, "loss": 0.0, "step": 5285 }, { "epoch": 1.22, "learning_rate": 2.9639473076034206e-05, "loss": 0.0, "step": 5286 }, { "epoch": 1.22, "learning_rate": 2.96356212926585e-05, "loss": 0.0, "step": 5287 }, { "epoch": 1.22, "learning_rate": 2.9631769509282796e-05, "loss": 0.0, "step": 5288 }, { "epoch": 1.22, "learning_rate": 2.9627917725907095e-05, "loss": 0.0, "step": 5289 }, { "epoch": 1.22, "learning_rate": 2.9624065942531393e-05, "loss": 0.0, "step": 5290 }, { "epoch": 1.22, "learning_rate": 2.962021415915569e-05, "loss": 0.0, "step": 5291 }, { "epoch": 1.22, "learning_rate": 2.9616362375779986e-05, "loss": 0.0, "step": 5292 }, { "epoch": 1.22, "learning_rate": 2.9612510592404285e-05, "loss": 0.0003, "step": 5293 }, { "epoch": 1.22, "learning_rate": 2.9608658809028583e-05, "loss": 0.0, "step": 5294 }, { "epoch": 1.22, "learning_rate": 2.960480702565288e-05, "loss": 0.0, "step": 5295 }, { "epoch": 1.22, "learning_rate": 2.9600955242277173e-05, "loss": 0.0, "step": 5296 }, { "epoch": 1.22, "learning_rate": 2.959710345890147e-05, "loss": 0.0, "step": 5297 }, { "epoch": 1.22, "learning_rate": 2.959325167552577e-05, "loss": 0.0, "step": 5298 }, { "epoch": 1.22, "learning_rate": 2.9589399892150065e-05, "loss": 0.0, "step": 5299 }, { "epoch": 1.22, "learning_rate": 2.9585548108774363e-05, "loss": 0.0, "step": 5300 }, { "epoch": 1.23, "learning_rate": 2.958169632539866e-05, "loss": 0.0, "step": 5301 }, { "epoch": 1.23, "learning_rate": 2.957784454202296e-05, "loss": 0.0001, "step": 5302 }, { "epoch": 1.23, "learning_rate": 2.9573992758647258e-05, "loss": 0.0, "step": 5303 }, { "epoch": 1.23, "learning_rate": 2.9570140975271553e-05, "loss": 0.0, "step": 5304 }, { "epoch": 1.23, "learning_rate": 2.956628919189585e-05, "loss": 0.0, "step": 5305 }, { "epoch": 1.23, "learning_rate": 2.9562437408520143e-05, "loss": 0.0, "step": 5306 }, { "epoch": 1.23, "learning_rate": 2.955858562514444e-05, "loss": 0.0, "step": 5307 }, { "epoch": 1.23, "learning_rate": 2.955473384176874e-05, "loss": 0.0, "step": 5308 }, { "epoch": 1.23, "learning_rate": 2.9550882058393038e-05, "loss": 0.0, "step": 5309 }, { "epoch": 1.23, "learning_rate": 2.9547030275017333e-05, "loss": 0.0, "step": 5310 }, { "epoch": 1.23, "learning_rate": 2.954317849164163e-05, "loss": 0.0, "step": 5311 }, { "epoch": 1.23, "learning_rate": 2.953932670826593e-05, "loss": 0.0, "step": 5312 }, { "epoch": 1.23, "learning_rate": 2.9535474924890228e-05, "loss": 0.0, "step": 5313 }, { "epoch": 1.23, "learning_rate": 2.9531623141514526e-05, "loss": 0.0, "step": 5314 }, { "epoch": 1.23, "learning_rate": 2.9527771358138818e-05, "loss": 0.0, "step": 5315 }, { "epoch": 1.23, "learning_rate": 2.9523919574763116e-05, "loss": 0.0, "step": 5316 }, { "epoch": 1.23, "learning_rate": 2.952006779138741e-05, "loss": 0.0002, "step": 5317 }, { "epoch": 1.23, "learning_rate": 2.951621600801171e-05, "loss": 0.0, "step": 5318 }, { "epoch": 1.23, "learning_rate": 2.9512364224636008e-05, "loss": 0.0, "step": 5319 }, { "epoch": 1.23, "learning_rate": 2.9508512441260306e-05, "loss": 0.0, "step": 5320 }, { "epoch": 1.23, "learning_rate": 2.9504660657884604e-05, "loss": 0.0001, "step": 5321 }, { "epoch": 1.23, "learning_rate": 2.95008088745089e-05, "loss": 0.0, "step": 5322 }, { "epoch": 1.23, "learning_rate": 2.9496957091133198e-05, "loss": 0.0, "step": 5323 }, { "epoch": 1.23, "learning_rate": 2.949310530775749e-05, "loss": 0.0, "step": 5324 }, { "epoch": 1.23, "learning_rate": 2.9489253524381788e-05, "loss": 0.0, "step": 5325 }, { "epoch": 1.23, "learning_rate": 2.9485401741006086e-05, "loss": 0.0, "step": 5326 }, { "epoch": 1.23, "learning_rate": 2.9481549957630384e-05, "loss": 0.0, "step": 5327 }, { "epoch": 1.23, "learning_rate": 2.947769817425468e-05, "loss": 0.0, "step": 5328 }, { "epoch": 1.23, "learning_rate": 2.9473846390878978e-05, "loss": 0.0, "step": 5329 }, { "epoch": 1.23, "learning_rate": 2.9469994607503276e-05, "loss": 0.0005, "step": 5330 }, { "epoch": 1.23, "learning_rate": 2.9466142824127574e-05, "loss": 0.0001, "step": 5331 }, { "epoch": 1.23, "learning_rate": 2.9462291040751873e-05, "loss": 0.0, "step": 5332 }, { "epoch": 1.23, "learning_rate": 2.9458439257376168e-05, "loss": 0.0, "step": 5333 }, { "epoch": 1.23, "learning_rate": 2.9454587474000463e-05, "loss": 0.0, "step": 5334 }, { "epoch": 1.23, "learning_rate": 2.9450735690624758e-05, "loss": 0.0, "step": 5335 }, { "epoch": 1.23, "learning_rate": 2.9446883907249056e-05, "loss": 0.0, "step": 5336 }, { "epoch": 1.23, "learning_rate": 2.9443032123873354e-05, "loss": 0.0, "step": 5337 }, { "epoch": 1.23, "learning_rate": 2.9439180340497653e-05, "loss": 0.0, "step": 5338 }, { "epoch": 1.23, "learning_rate": 2.943532855712195e-05, "loss": 0.0, "step": 5339 }, { "epoch": 1.23, "learning_rate": 2.9431476773746246e-05, "loss": 0.0, "step": 5340 }, { "epoch": 1.23, "learning_rate": 2.9427624990370544e-05, "loss": 0.0, "step": 5341 }, { "epoch": 1.23, "learning_rate": 2.9423773206994843e-05, "loss": 0.0, "step": 5342 }, { "epoch": 1.23, "learning_rate": 2.9419921423619134e-05, "loss": 0.0, "step": 5343 }, { "epoch": 1.24, "learning_rate": 2.9416069640243432e-05, "loss": 0.0, "step": 5344 }, { "epoch": 1.24, "learning_rate": 2.941221785686773e-05, "loss": 0.0, "step": 5345 }, { "epoch": 1.24, "learning_rate": 2.9408366073492026e-05, "loss": 0.0, "step": 5346 }, { "epoch": 1.24, "learning_rate": 2.9404514290116324e-05, "loss": 0.0, "step": 5347 }, { "epoch": 1.24, "learning_rate": 2.9400662506740622e-05, "loss": 0.0, "step": 5348 }, { "epoch": 1.24, "learning_rate": 2.939681072336492e-05, "loss": 0.0, "step": 5349 }, { "epoch": 1.24, "learning_rate": 2.939295893998922e-05, "loss": 0.0, "step": 5350 }, { "epoch": 1.24, "learning_rate": 2.9389107156613514e-05, "loss": 0.0, "step": 5351 }, { "epoch": 1.24, "learning_rate": 2.938525537323781e-05, "loss": 0.0, "step": 5352 }, { "epoch": 1.24, "learning_rate": 2.9381403589862104e-05, "loss": 0.0001, "step": 5353 }, { "epoch": 1.24, "learning_rate": 2.9377551806486402e-05, "loss": 0.0, "step": 5354 }, { "epoch": 1.24, "learning_rate": 2.93737000231107e-05, "loss": 0.0, "step": 5355 }, { "epoch": 1.24, "learning_rate": 2.9369848239735e-05, "loss": 0.0, "step": 5356 }, { "epoch": 1.24, "learning_rate": 2.9365996456359297e-05, "loss": 0.0, "step": 5357 }, { "epoch": 1.24, "learning_rate": 2.9362144672983592e-05, "loss": 0.0, "step": 5358 }, { "epoch": 1.24, "learning_rate": 2.935829288960789e-05, "loss": 0.0, "step": 5359 }, { "epoch": 1.24, "learning_rate": 2.935444110623219e-05, "loss": 0.0, "step": 5360 }, { "epoch": 1.24, "learning_rate": 2.9350589322856487e-05, "loss": 0.0, "step": 5361 }, { "epoch": 1.24, "learning_rate": 2.934673753948078e-05, "loss": 0.0, "step": 5362 }, { "epoch": 1.24, "learning_rate": 2.9342885756105077e-05, "loss": 0.0, "step": 5363 }, { "epoch": 1.24, "learning_rate": 2.9339033972729372e-05, "loss": 0.0, "step": 5364 }, { "epoch": 1.24, "learning_rate": 2.933518218935367e-05, "loss": 0.0001, "step": 5365 }, { "epoch": 1.24, "learning_rate": 2.933133040597797e-05, "loss": 0.0, "step": 5366 }, { "epoch": 1.24, "learning_rate": 2.9327478622602267e-05, "loss": 0.0, "step": 5367 }, { "epoch": 1.24, "learning_rate": 2.9323626839226566e-05, "loss": 0.0, "step": 5368 }, { "epoch": 1.24, "learning_rate": 2.931977505585086e-05, "loss": 0.0, "step": 5369 }, { "epoch": 1.24, "learning_rate": 2.931592327247516e-05, "loss": 0.0, "step": 5370 }, { "epoch": 1.24, "learning_rate": 2.931207148909945e-05, "loss": 0.0, "step": 5371 }, { "epoch": 1.24, "learning_rate": 2.930821970572375e-05, "loss": 0.0, "step": 5372 }, { "epoch": 1.24, "learning_rate": 2.9304367922348047e-05, "loss": 0.0, "step": 5373 }, { "epoch": 1.24, "learning_rate": 2.9300516138972346e-05, "loss": 0.0, "step": 5374 }, { "epoch": 1.24, "learning_rate": 2.9296664355596644e-05, "loss": 0.0, "step": 5375 }, { "epoch": 1.24, "learning_rate": 2.929281257222094e-05, "loss": 0.0, "step": 5376 }, { "epoch": 1.24, "learning_rate": 2.9288960788845237e-05, "loss": 0.0, "step": 5377 }, { "epoch": 1.24, "learning_rate": 2.9285109005469536e-05, "loss": 0.0, "step": 5378 }, { "epoch": 1.24, "learning_rate": 2.9281257222093834e-05, "loss": 0.0, "step": 5379 }, { "epoch": 1.24, "learning_rate": 2.9277405438718125e-05, "loss": 0.0, "step": 5380 }, { "epoch": 1.24, "learning_rate": 2.9273553655342424e-05, "loss": 0.0, "step": 5381 }, { "epoch": 1.24, "learning_rate": 2.9269701871966722e-05, "loss": 0.0, "step": 5382 }, { "epoch": 1.24, "learning_rate": 2.9265850088591017e-05, "loss": 0.0, "step": 5383 }, { "epoch": 1.24, "learning_rate": 2.9261998305215315e-05, "loss": 0.0011, "step": 5384 }, { "epoch": 1.24, "learning_rate": 2.9258146521839614e-05, "loss": 0.0, "step": 5385 }, { "epoch": 1.24, "learning_rate": 2.9254294738463912e-05, "loss": 0.0, "step": 5386 }, { "epoch": 1.24, "learning_rate": 2.9250442955088207e-05, "loss": 0.0, "step": 5387 }, { "epoch": 1.25, "learning_rate": 2.9246591171712505e-05, "loss": 0.0, "step": 5388 }, { "epoch": 1.25, "learning_rate": 2.9242739388336804e-05, "loss": 0.0, "step": 5389 }, { "epoch": 1.25, "learning_rate": 2.9238887604961095e-05, "loss": 0.0, "step": 5390 }, { "epoch": 1.25, "learning_rate": 2.9235035821585394e-05, "loss": 0.0, "step": 5391 }, { "epoch": 1.25, "learning_rate": 2.9231184038209692e-05, "loss": 0.0, "step": 5392 }, { "epoch": 1.25, "learning_rate": 2.922733225483399e-05, "loss": 0.0, "step": 5393 }, { "epoch": 1.25, "learning_rate": 2.9223480471458285e-05, "loss": 0.0, "step": 5394 }, { "epoch": 1.25, "learning_rate": 2.9219628688082584e-05, "loss": 0.0, "step": 5395 }, { "epoch": 1.25, "learning_rate": 2.9215776904706882e-05, "loss": 0.0, "step": 5396 }, { "epoch": 1.25, "learning_rate": 2.921192512133118e-05, "loss": 0.0, "step": 5397 }, { "epoch": 1.25, "learning_rate": 2.920807333795548e-05, "loss": 0.0, "step": 5398 }, { "epoch": 1.25, "learning_rate": 2.920422155457977e-05, "loss": 0.0, "step": 5399 }, { "epoch": 1.25, "learning_rate": 2.920036977120407e-05, "loss": 0.0, "step": 5400 }, { "epoch": 1.25, "learning_rate": 2.9196517987828364e-05, "loss": 0.0, "step": 5401 }, { "epoch": 1.25, "learning_rate": 2.9192666204452662e-05, "loss": 0.0, "step": 5402 }, { "epoch": 1.25, "learning_rate": 2.918881442107696e-05, "loss": 0.0, "step": 5403 }, { "epoch": 1.25, "learning_rate": 2.918496263770126e-05, "loss": 0.0, "step": 5404 }, { "epoch": 1.25, "learning_rate": 2.9181110854325554e-05, "loss": 0.0, "step": 5405 }, { "epoch": 1.25, "learning_rate": 2.9177259070949852e-05, "loss": 0.0, "step": 5406 }, { "epoch": 1.25, "learning_rate": 2.917340728757415e-05, "loss": 0.0, "step": 5407 }, { "epoch": 1.25, "learning_rate": 2.9169555504198442e-05, "loss": 0.0, "step": 5408 }, { "epoch": 1.25, "learning_rate": 2.916570372082274e-05, "loss": 0.0, "step": 5409 }, { "epoch": 1.25, "learning_rate": 2.916185193744704e-05, "loss": 0.0, "step": 5410 }, { "epoch": 1.25, "learning_rate": 2.9158000154071337e-05, "loss": 0.0, "step": 5411 }, { "epoch": 1.25, "learning_rate": 2.9154148370695632e-05, "loss": 0.0, "step": 5412 }, { "epoch": 1.25, "learning_rate": 2.915029658731993e-05, "loss": 0.0, "step": 5413 }, { "epoch": 1.25, "learning_rate": 2.914644480394423e-05, "loss": 0.0, "step": 5414 }, { "epoch": 1.25, "learning_rate": 2.9142593020568527e-05, "loss": 0.0, "step": 5415 }, { "epoch": 1.25, "learning_rate": 2.9138741237192825e-05, "loss": 0.0, "step": 5416 }, { "epoch": 1.25, "learning_rate": 2.913488945381712e-05, "loss": 0.0, "step": 5417 }, { "epoch": 1.25, "learning_rate": 2.9131037670441415e-05, "loss": 0.0, "step": 5418 }, { "epoch": 1.25, "learning_rate": 2.912718588706571e-05, "loss": 0.0, "step": 5419 }, { "epoch": 1.25, "learning_rate": 2.912333410369001e-05, "loss": 0.0, "step": 5420 }, { "epoch": 1.25, "learning_rate": 2.9119482320314307e-05, "loss": 0.0, "step": 5421 }, { "epoch": 1.25, "learning_rate": 2.9115630536938605e-05, "loss": 0.0, "step": 5422 }, { "epoch": 1.25, "learning_rate": 2.91117787535629e-05, "loss": 0.0, "step": 5423 }, { "epoch": 1.25, "learning_rate": 2.91079269701872e-05, "loss": 0.0, "step": 5424 }, { "epoch": 1.25, "learning_rate": 2.9104075186811497e-05, "loss": 0.0, "step": 5425 }, { "epoch": 1.25, "learning_rate": 2.9100223403435795e-05, "loss": 0.0, "step": 5426 }, { "epoch": 1.25, "learning_rate": 2.9096371620060087e-05, "loss": 0.0, "step": 5427 }, { "epoch": 1.25, "learning_rate": 2.9092519836684385e-05, "loss": 0.0, "step": 5428 }, { "epoch": 1.25, "learning_rate": 2.9088668053308683e-05, "loss": 0.0, "step": 5429 }, { "epoch": 1.25, "learning_rate": 2.9084816269932978e-05, "loss": 0.0, "step": 5430 }, { "epoch": 1.26, "learning_rate": 2.9080964486557277e-05, "loss": 0.0, "step": 5431 }, { "epoch": 1.26, "learning_rate": 2.9077112703181575e-05, "loss": 0.0, "step": 5432 }, { "epoch": 1.26, "learning_rate": 2.9073260919805873e-05, "loss": 0.0, "step": 5433 }, { "epoch": 1.26, "learning_rate": 2.906940913643017e-05, "loss": 0.0, "step": 5434 }, { "epoch": 1.26, "learning_rate": 2.9065557353054467e-05, "loss": 0.0, "step": 5435 }, { "epoch": 1.26, "learning_rate": 2.906170556967876e-05, "loss": 0.0, "step": 5436 }, { "epoch": 1.26, "learning_rate": 2.9057853786303056e-05, "loss": 0.0, "step": 5437 }, { "epoch": 1.26, "learning_rate": 2.9054002002927355e-05, "loss": 0.0002, "step": 5438 }, { "epoch": 1.26, "learning_rate": 2.9050150219551653e-05, "loss": 0.0, "step": 5439 }, { "epoch": 1.26, "learning_rate": 2.904629843617595e-05, "loss": 0.0, "step": 5440 }, { "epoch": 1.26, "learning_rate": 2.9042446652800246e-05, "loss": 0.0, "step": 5441 }, { "epoch": 1.26, "learning_rate": 2.9038594869424545e-05, "loss": 0.0, "step": 5442 }, { "epoch": 1.26, "learning_rate": 2.9034743086048843e-05, "loss": 0.0, "step": 5443 }, { "epoch": 1.26, "learning_rate": 2.903089130267314e-05, "loss": 0.0, "step": 5444 }, { "epoch": 1.26, "learning_rate": 2.902703951929744e-05, "loss": 0.0, "step": 5445 }, { "epoch": 1.26, "learning_rate": 2.902318773592173e-05, "loss": 0.0, "step": 5446 }, { "epoch": 1.26, "learning_rate": 2.901933595254603e-05, "loss": 0.0, "step": 5447 }, { "epoch": 1.26, "learning_rate": 2.9015484169170325e-05, "loss": 0.0, "step": 5448 }, { "epoch": 1.26, "learning_rate": 2.9011632385794623e-05, "loss": 0.0, "step": 5449 }, { "epoch": 1.26, "learning_rate": 2.900778060241892e-05, "loss": 0.0, "step": 5450 }, { "epoch": 1.26, "learning_rate": 2.900392881904322e-05, "loss": 0.0, "step": 5451 }, { "epoch": 1.26, "learning_rate": 2.9000077035667518e-05, "loss": 0.0, "step": 5452 }, { "epoch": 1.26, "learning_rate": 2.8996225252291813e-05, "loss": 0.0, "step": 5453 }, { "epoch": 1.26, "learning_rate": 2.899237346891611e-05, "loss": 0.0, "step": 5454 }, { "epoch": 1.26, "learning_rate": 2.8988521685540403e-05, "loss": 0.0, "step": 5455 }, { "epoch": 1.26, "learning_rate": 2.89846699021647e-05, "loss": 0.0, "step": 5456 }, { "epoch": 1.26, "learning_rate": 2.8980818118789e-05, "loss": 0.0, "step": 5457 }, { "epoch": 1.26, "learning_rate": 2.8976966335413298e-05, "loss": 0.0001, "step": 5458 }, { "epoch": 1.26, "learning_rate": 2.8973114552037596e-05, "loss": 0.0, "step": 5459 }, { "epoch": 1.26, "learning_rate": 2.896926276866189e-05, "loss": 0.0, "step": 5460 }, { "epoch": 1.26, "learning_rate": 2.896541098528619e-05, "loss": 0.0, "step": 5461 }, { "epoch": 1.26, "learning_rate": 2.8961559201910488e-05, "loss": 0.0, "step": 5462 }, { "epoch": 1.26, "learning_rate": 2.8957707418534786e-05, "loss": 0.0, "step": 5463 }, { "epoch": 1.26, "learning_rate": 2.8953855635159078e-05, "loss": 0.0, "step": 5464 }, { "epoch": 1.26, "learning_rate": 2.8950003851783376e-05, "loss": 0.0, "step": 5465 }, { "epoch": 1.26, "learning_rate": 2.894615206840767e-05, "loss": 0.0, "step": 5466 }, { "epoch": 1.26, "learning_rate": 2.894230028503197e-05, "loss": 0.0, "step": 5467 }, { "epoch": 1.26, "learning_rate": 2.8938448501656268e-05, "loss": 0.0, "step": 5468 }, { "epoch": 1.26, "learning_rate": 2.8934596718280566e-05, "loss": 0.0, "step": 5469 }, { "epoch": 1.26, "learning_rate": 2.8930744934904865e-05, "loss": 0.0, "step": 5470 }, { "epoch": 1.26, "learning_rate": 2.892689315152916e-05, "loss": 0.0, "step": 5471 }, { "epoch": 1.26, "learning_rate": 2.8923041368153458e-05, "loss": 0.0, "step": 5472 }, { "epoch": 1.26, "learning_rate": 2.8919189584777756e-05, "loss": 0.0, "step": 5473 }, { "epoch": 1.27, "learning_rate": 2.8915337801402048e-05, "loss": 0.0, "step": 5474 }, { "epoch": 1.27, "learning_rate": 2.8911486018026346e-05, "loss": 0.0, "step": 5475 }, { "epoch": 1.27, "learning_rate": 2.8907634234650644e-05, "loss": 0.0, "step": 5476 }, { "epoch": 1.27, "learning_rate": 2.8903782451274943e-05, "loss": 0.0, "step": 5477 }, { "epoch": 1.27, "learning_rate": 2.8899930667899238e-05, "loss": 0.0, "step": 5478 }, { "epoch": 1.27, "learning_rate": 2.8896078884523536e-05, "loss": 0.0, "step": 5479 }, { "epoch": 1.27, "learning_rate": 2.8892227101147834e-05, "loss": 0.0, "step": 5480 }, { "epoch": 1.27, "learning_rate": 2.8888375317772133e-05, "loss": 0.0, "step": 5481 }, { "epoch": 1.27, "learning_rate": 2.8884523534396428e-05, "loss": 0.0001, "step": 5482 }, { "epoch": 1.27, "learning_rate": 2.8880671751020723e-05, "loss": 0.0, "step": 5483 }, { "epoch": 1.27, "learning_rate": 2.8876819967645018e-05, "loss": 0.0, "step": 5484 }, { "epoch": 1.27, "learning_rate": 2.8872968184269316e-05, "loss": 0.0, "step": 5485 }, { "epoch": 1.27, "learning_rate": 2.8869116400893614e-05, "loss": 0.0, "step": 5486 }, { "epoch": 1.27, "learning_rate": 2.8865264617517913e-05, "loss": 0.0, "step": 5487 }, { "epoch": 1.27, "learning_rate": 2.886141283414221e-05, "loss": 0.0, "step": 5488 }, { "epoch": 1.27, "learning_rate": 2.8857561050766506e-05, "loss": 0.0, "step": 5489 }, { "epoch": 1.27, "learning_rate": 2.8853709267390804e-05, "loss": 0.0, "step": 5490 }, { "epoch": 1.27, "learning_rate": 2.8849857484015103e-05, "loss": 0.0, "step": 5491 }, { "epoch": 1.27, "learning_rate": 2.8846005700639394e-05, "loss": 0.0, "step": 5492 }, { "epoch": 1.27, "learning_rate": 2.8842153917263693e-05, "loss": 0.0018, "step": 5493 }, { "epoch": 1.27, "learning_rate": 2.883830213388799e-05, "loss": 0.0, "step": 5494 }, { "epoch": 1.27, "learning_rate": 2.883445035051229e-05, "loss": 0.0, "step": 5495 }, { "epoch": 1.27, "learning_rate": 2.8830598567136584e-05, "loss": 0.0, "step": 5496 }, { "epoch": 1.27, "learning_rate": 2.8826746783760883e-05, "loss": 0.0, "step": 5497 }, { "epoch": 1.27, "learning_rate": 2.882289500038518e-05, "loss": 0.0, "step": 5498 }, { "epoch": 1.27, "learning_rate": 2.881904321700948e-05, "loss": 0.0, "step": 5499 }, { "epoch": 1.27, "learning_rate": 2.8815191433633774e-05, "loss": 0.0, "step": 5500 }, { "epoch": 1.27, "learning_rate": 2.8811339650258073e-05, "loss": 0.0, "step": 5501 }, { "epoch": 1.27, "learning_rate": 2.8807487866882364e-05, "loss": 0.0, "step": 5502 }, { "epoch": 1.27, "learning_rate": 2.8803636083506662e-05, "loss": 0.0, "step": 5503 }, { "epoch": 1.27, "learning_rate": 2.879978430013096e-05, "loss": 0.0, "step": 5504 }, { "epoch": 1.27, "learning_rate": 2.879593251675526e-05, "loss": 0.0, "step": 5505 }, { "epoch": 1.27, "learning_rate": 2.8792080733379557e-05, "loss": 0.0, "step": 5506 }, { "epoch": 1.27, "learning_rate": 2.8788228950003852e-05, "loss": 0.0, "step": 5507 }, { "epoch": 1.27, "learning_rate": 2.878437716662815e-05, "loss": 0.0, "step": 5508 }, { "epoch": 1.27, "learning_rate": 2.878052538325245e-05, "loss": 0.0, "step": 5509 }, { "epoch": 1.27, "learning_rate": 2.8776673599876747e-05, "loss": 0.0, "step": 5510 }, { "epoch": 1.27, "learning_rate": 2.877282181650104e-05, "loss": 0.0001, "step": 5511 }, { "epoch": 1.27, "learning_rate": 2.8768970033125337e-05, "loss": 0.0, "step": 5512 }, { "epoch": 1.27, "learning_rate": 2.8765118249749636e-05, "loss": 0.0, "step": 5513 }, { "epoch": 1.27, "learning_rate": 2.876126646637393e-05, "loss": 0.0, "step": 5514 }, { "epoch": 1.27, "learning_rate": 2.875741468299823e-05, "loss": 0.0, "step": 5515 }, { "epoch": 1.27, "learning_rate": 2.8753562899622527e-05, "loss": 0.0, "step": 5516 }, { "epoch": 1.28, "learning_rate": 2.8749711116246826e-05, "loss": 0.0001, "step": 5517 }, { "epoch": 1.28, "learning_rate": 2.874585933287112e-05, "loss": 0.0, "step": 5518 }, { "epoch": 1.28, "learning_rate": 2.874200754949542e-05, "loss": 0.0, "step": 5519 }, { "epoch": 1.28, "learning_rate": 2.873815576611971e-05, "loss": 0.0, "step": 5520 }, { "epoch": 1.28, "learning_rate": 2.873430398274401e-05, "loss": 0.0, "step": 5521 }, { "epoch": 1.28, "learning_rate": 2.8730452199368307e-05, "loss": 0.0, "step": 5522 }, { "epoch": 1.28, "learning_rate": 2.8726600415992606e-05, "loss": 0.0, "step": 5523 }, { "epoch": 1.28, "learning_rate": 2.8722748632616904e-05, "loss": 0.0, "step": 5524 }, { "epoch": 1.28, "learning_rate": 2.87188968492412e-05, "loss": 0.0, "step": 5525 }, { "epoch": 1.28, "learning_rate": 2.8715045065865497e-05, "loss": 0.0, "step": 5526 }, { "epoch": 1.28, "learning_rate": 2.8711193282489796e-05, "loss": 0.0, "step": 5527 }, { "epoch": 1.28, "learning_rate": 2.8707341499114094e-05, "loss": 0.0, "step": 5528 }, { "epoch": 1.28, "learning_rate": 2.8703489715738392e-05, "loss": 0.0, "step": 5529 }, { "epoch": 1.28, "learning_rate": 2.8699637932362684e-05, "loss": 0.0, "step": 5530 }, { "epoch": 1.28, "learning_rate": 2.8695786148986982e-05, "loss": 0.0, "step": 5531 }, { "epoch": 1.28, "learning_rate": 2.8691934365611277e-05, "loss": 0.0, "step": 5532 }, { "epoch": 1.28, "learning_rate": 2.8688082582235575e-05, "loss": 0.0, "step": 5533 }, { "epoch": 1.28, "learning_rate": 2.8684230798859874e-05, "loss": 0.0, "step": 5534 }, { "epoch": 1.28, "learning_rate": 2.8680379015484172e-05, "loss": 0.0, "step": 5535 }, { "epoch": 1.28, "learning_rate": 2.8676527232108467e-05, "loss": 0.0, "step": 5536 }, { "epoch": 1.28, "learning_rate": 2.8672675448732765e-05, "loss": 0.0, "step": 5537 }, { "epoch": 1.28, "learning_rate": 2.8668823665357064e-05, "loss": 0.0, "step": 5538 }, { "epoch": 1.28, "learning_rate": 2.8664971881981355e-05, "loss": 0.0, "step": 5539 }, { "epoch": 1.28, "learning_rate": 2.8661120098605654e-05, "loss": 0.0, "step": 5540 }, { "epoch": 1.28, "learning_rate": 2.8657268315229952e-05, "loss": 0.0, "step": 5541 }, { "epoch": 1.28, "learning_rate": 2.865341653185425e-05, "loss": 0.0, "step": 5542 }, { "epoch": 1.28, "learning_rate": 2.8649564748478545e-05, "loss": 0.0, "step": 5543 }, { "epoch": 1.28, "learning_rate": 2.8645712965102844e-05, "loss": 0.0, "step": 5544 }, { "epoch": 1.28, "learning_rate": 2.8641861181727142e-05, "loss": 0.0001, "step": 5545 }, { "epoch": 1.28, "learning_rate": 2.863800939835144e-05, "loss": 0.0, "step": 5546 }, { "epoch": 1.28, "learning_rate": 2.863415761497574e-05, "loss": 0.0, "step": 5547 }, { "epoch": 1.28, "learning_rate": 2.863030583160003e-05, "loss": 0.0, "step": 5548 }, { "epoch": 1.28, "learning_rate": 2.862645404822433e-05, "loss": 0.0, "step": 5549 }, { "epoch": 1.28, "learning_rate": 2.8622602264848624e-05, "loss": 0.0, "step": 5550 }, { "epoch": 1.28, "learning_rate": 2.8618750481472922e-05, "loss": 0.0004, "step": 5551 }, { "epoch": 1.28, "learning_rate": 2.861489869809722e-05, "loss": 0.0, "step": 5552 }, { "epoch": 1.28, "learning_rate": 2.861104691472152e-05, "loss": 0.0001, "step": 5553 }, { "epoch": 1.28, "learning_rate": 2.8607195131345817e-05, "loss": 0.0, "step": 5554 }, { "epoch": 1.28, "learning_rate": 2.8603343347970112e-05, "loss": 0.0, "step": 5555 }, { "epoch": 1.28, "learning_rate": 2.859949156459441e-05, "loss": 0.0, "step": 5556 }, { "epoch": 1.28, "learning_rate": 2.859563978121871e-05, "loss": 0.0, "step": 5557 }, { "epoch": 1.28, "learning_rate": 2.8591787997843e-05, "loss": 0.0, "step": 5558 }, { "epoch": 1.28, "learning_rate": 2.85879362144673e-05, "loss": 0.0, "step": 5559 }, { "epoch": 1.28, "learning_rate": 2.8584084431091597e-05, "loss": 0.0, "step": 5560 }, { "epoch": 1.29, "learning_rate": 2.8580232647715892e-05, "loss": 0.0, "step": 5561 }, { "epoch": 1.29, "learning_rate": 2.857638086434019e-05, "loss": 0.0001, "step": 5562 }, { "epoch": 1.29, "learning_rate": 2.857252908096449e-05, "loss": 0.0, "step": 5563 }, { "epoch": 1.29, "learning_rate": 2.8568677297588787e-05, "loss": 0.0, "step": 5564 }, { "epoch": 1.29, "learning_rate": 2.8564825514213085e-05, "loss": 0.0, "step": 5565 }, { "epoch": 1.29, "learning_rate": 2.856097373083738e-05, "loss": 0.0, "step": 5566 }, { "epoch": 1.29, "learning_rate": 2.8557121947461675e-05, "loss": 0.0, "step": 5567 }, { "epoch": 1.29, "learning_rate": 2.855327016408597e-05, "loss": 0.0, "step": 5568 }, { "epoch": 1.29, "learning_rate": 2.854941838071027e-05, "loss": 0.0, "step": 5569 }, { "epoch": 1.29, "learning_rate": 2.8545566597334567e-05, "loss": 0.0, "step": 5570 }, { "epoch": 1.29, "learning_rate": 2.8541714813958865e-05, "loss": 0.0, "step": 5571 }, { "epoch": 1.29, "learning_rate": 2.8537863030583163e-05, "loss": 0.0, "step": 5572 }, { "epoch": 1.29, "learning_rate": 2.853401124720746e-05, "loss": 0.0, "step": 5573 }, { "epoch": 1.29, "learning_rate": 2.8530159463831757e-05, "loss": 0.0, "step": 5574 }, { "epoch": 1.29, "learning_rate": 2.8526307680456055e-05, "loss": 0.0, "step": 5575 }, { "epoch": 1.29, "learning_rate": 2.8522455897080347e-05, "loss": 0.0, "step": 5576 }, { "epoch": 1.29, "learning_rate": 2.8518604113704645e-05, "loss": 0.0002, "step": 5577 }, { "epoch": 1.29, "learning_rate": 2.8514752330328943e-05, "loss": 0.0, "step": 5578 }, { "epoch": 1.29, "learning_rate": 2.851090054695324e-05, "loss": 0.0, "step": 5579 }, { "epoch": 1.29, "learning_rate": 2.8507048763577537e-05, "loss": 0.0, "step": 5580 }, { "epoch": 1.29, "learning_rate": 2.8503196980201835e-05, "loss": 0.0, "step": 5581 }, { "epoch": 1.29, "learning_rate": 2.8499345196826133e-05, "loss": 0.0, "step": 5582 }, { "epoch": 1.29, "learning_rate": 2.849549341345043e-05, "loss": 0.0, "step": 5583 }, { "epoch": 1.29, "learning_rate": 2.8491641630074727e-05, "loss": 0.0, "step": 5584 }, { "epoch": 1.29, "learning_rate": 2.8487789846699025e-05, "loss": 0.0, "step": 5585 }, { "epoch": 1.29, "learning_rate": 2.8483938063323317e-05, "loss": 0.0, "step": 5586 }, { "epoch": 1.29, "learning_rate": 2.8480086279947615e-05, "loss": 0.0, "step": 5587 }, { "epoch": 1.29, "learning_rate": 2.8476234496571913e-05, "loss": 0.0, "step": 5588 }, { "epoch": 1.29, "learning_rate": 2.847238271319621e-05, "loss": 0.0, "step": 5589 }, { "epoch": 1.29, "learning_rate": 2.846853092982051e-05, "loss": 0.0, "step": 5590 }, { "epoch": 1.29, "learning_rate": 2.8464679146444805e-05, "loss": 0.0, "step": 5591 }, { "epoch": 1.29, "learning_rate": 2.8460827363069103e-05, "loss": 0.0002, "step": 5592 }, { "epoch": 1.29, "learning_rate": 2.84569755796934e-05, "loss": 0.0, "step": 5593 }, { "epoch": 1.29, "learning_rate": 2.84531237963177e-05, "loss": 0.0, "step": 5594 }, { "epoch": 1.29, "learning_rate": 2.844927201294199e-05, "loss": 0.0001, "step": 5595 }, { "epoch": 1.29, "learning_rate": 2.844542022956629e-05, "loss": 0.0, "step": 5596 }, { "epoch": 1.29, "learning_rate": 2.8441568446190585e-05, "loss": 0.0001, "step": 5597 }, { "epoch": 1.29, "learning_rate": 2.8437716662814883e-05, "loss": 0.0, "step": 5598 }, { "epoch": 1.29, "learning_rate": 2.843386487943918e-05, "loss": 0.0, "step": 5599 }, { "epoch": 1.29, "learning_rate": 2.843001309606348e-05, "loss": 0.0, "step": 5600 }, { "epoch": 1.29, "learning_rate": 2.8426161312687778e-05, "loss": 0.0, "step": 5601 }, { "epoch": 1.29, "learning_rate": 2.8422309529312073e-05, "loss": 0.0, "step": 5602 }, { "epoch": 1.29, "learning_rate": 2.841845774593637e-05, "loss": 0.0, "step": 5603 }, { "epoch": 1.3, "learning_rate": 2.8414605962560663e-05, "loss": 0.0, "step": 5604 }, { "epoch": 1.3, "learning_rate": 2.841075417918496e-05, "loss": 0.0, "step": 5605 }, { "epoch": 1.3, "learning_rate": 2.840690239580926e-05, "loss": 0.0, "step": 5606 }, { "epoch": 1.3, "learning_rate": 2.8403050612433558e-05, "loss": 0.0, "step": 5607 }, { "epoch": 1.3, "learning_rate": 2.8399198829057856e-05, "loss": 0.0, "step": 5608 }, { "epoch": 1.3, "learning_rate": 2.839534704568215e-05, "loss": 0.0, "step": 5609 }, { "epoch": 1.3, "learning_rate": 2.839149526230645e-05, "loss": 0.0, "step": 5610 }, { "epoch": 1.3, "learning_rate": 2.8387643478930748e-05, "loss": 0.0, "step": 5611 }, { "epoch": 1.3, "learning_rate": 2.8383791695555046e-05, "loss": 0.0, "step": 5612 }, { "epoch": 1.3, "learning_rate": 2.837993991217934e-05, "loss": 0.0, "step": 5613 }, { "epoch": 1.3, "learning_rate": 2.8376088128803636e-05, "loss": 0.0, "step": 5614 }, { "epoch": 1.3, "learning_rate": 2.837223634542793e-05, "loss": 0.0, "step": 5615 }, { "epoch": 1.3, "learning_rate": 2.836838456205223e-05, "loss": 0.0001, "step": 5616 }, { "epoch": 1.3, "learning_rate": 2.8364532778676528e-05, "loss": 0.0, "step": 5617 }, { "epoch": 1.3, "learning_rate": 2.8360680995300826e-05, "loss": 0.0, "step": 5618 }, { "epoch": 1.3, "learning_rate": 2.8356829211925125e-05, "loss": 0.0, "step": 5619 }, { "epoch": 1.3, "learning_rate": 2.835297742854942e-05, "loss": 0.0, "step": 5620 }, { "epoch": 1.3, "learning_rate": 2.8349125645173718e-05, "loss": 0.0, "step": 5621 }, { "epoch": 1.3, "learning_rate": 2.8345273861798016e-05, "loss": 0.0, "step": 5622 }, { "epoch": 1.3, "learning_rate": 2.8341422078422308e-05, "loss": 0.0, "step": 5623 }, { "epoch": 1.3, "learning_rate": 2.8337570295046606e-05, "loss": 0.0, "step": 5624 }, { "epoch": 1.3, "learning_rate": 2.8333718511670905e-05, "loss": 0.0, "step": 5625 }, { "epoch": 1.3, "learning_rate": 2.8329866728295203e-05, "loss": 0.0, "step": 5626 }, { "epoch": 1.3, "learning_rate": 2.8326014944919498e-05, "loss": 0.0, "step": 5627 }, { "epoch": 1.3, "learning_rate": 2.8322163161543796e-05, "loss": 0.0, "step": 5628 }, { "epoch": 1.3, "learning_rate": 2.8318311378168095e-05, "loss": 0.0, "step": 5629 }, { "epoch": 1.3, "learning_rate": 2.8314459594792393e-05, "loss": 0.0, "step": 5630 }, { "epoch": 1.3, "learning_rate": 2.8310607811416688e-05, "loss": 0.0, "step": 5631 }, { "epoch": 1.3, "learning_rate": 2.8306756028040983e-05, "loss": 0.0, "step": 5632 }, { "epoch": 1.3, "learning_rate": 2.8302904244665278e-05, "loss": 0.0, "step": 5633 }, { "epoch": 1.3, "learning_rate": 2.8299052461289576e-05, "loss": 0.0, "step": 5634 }, { "epoch": 1.3, "learning_rate": 2.8295200677913874e-05, "loss": 0.0, "step": 5635 }, { "epoch": 1.3, "learning_rate": 2.8291348894538173e-05, "loss": 0.0, "step": 5636 }, { "epoch": 1.3, "learning_rate": 2.828749711116247e-05, "loss": 0.0, "step": 5637 }, { "epoch": 1.3, "learning_rate": 2.8283645327786766e-05, "loss": 0.0001, "step": 5638 }, { "epoch": 1.3, "learning_rate": 2.8279793544411064e-05, "loss": 0.0, "step": 5639 }, { "epoch": 1.3, "learning_rate": 2.8275941761035363e-05, "loss": 0.0, "step": 5640 }, { "epoch": 1.3, "learning_rate": 2.827208997765966e-05, "loss": 0.0, "step": 5641 }, { "epoch": 1.3, "learning_rate": 2.8268238194283953e-05, "loss": 0.0, "step": 5642 }, { "epoch": 1.3, "learning_rate": 2.826438641090825e-05, "loss": 0.0, "step": 5643 }, { "epoch": 1.3, "learning_rate": 2.826053462753255e-05, "loss": 0.0, "step": 5644 }, { "epoch": 1.3, "learning_rate": 2.8256682844156844e-05, "loss": 0.0, "step": 5645 }, { "epoch": 1.3, "learning_rate": 2.8252831060781143e-05, "loss": 0.0, "step": 5646 }, { "epoch": 1.31, "learning_rate": 2.824897927740544e-05, "loss": 0.0, "step": 5647 }, { "epoch": 1.31, "learning_rate": 2.824512749402974e-05, "loss": 0.0, "step": 5648 }, { "epoch": 1.31, "learning_rate": 2.8241275710654038e-05, "loss": 0.0, "step": 5649 }, { "epoch": 1.31, "learning_rate": 2.8237423927278333e-05, "loss": 0.0, "step": 5650 }, { "epoch": 1.31, "learning_rate": 2.8233572143902624e-05, "loss": 0.0, "step": 5651 }, { "epoch": 1.31, "learning_rate": 2.8229720360526923e-05, "loss": 0.0, "step": 5652 }, { "epoch": 1.31, "learning_rate": 2.822586857715122e-05, "loss": 0.0005, "step": 5653 }, { "epoch": 1.31, "learning_rate": 2.822201679377552e-05, "loss": 0.0, "step": 5654 }, { "epoch": 1.31, "learning_rate": 2.8218165010399818e-05, "loss": 0.0002, "step": 5655 }, { "epoch": 1.31, "learning_rate": 2.8214313227024113e-05, "loss": 0.0, "step": 5656 }, { "epoch": 1.31, "learning_rate": 2.821046144364841e-05, "loss": 0.0, "step": 5657 }, { "epoch": 1.31, "learning_rate": 2.820660966027271e-05, "loss": 0.0, "step": 5658 }, { "epoch": 1.31, "learning_rate": 2.8202757876897008e-05, "loss": 0.0, "step": 5659 }, { "epoch": 1.31, "learning_rate": 2.81989060935213e-05, "loss": 0.0, "step": 5660 }, { "epoch": 1.31, "learning_rate": 2.8195054310145597e-05, "loss": 0.0, "step": 5661 }, { "epoch": 1.31, "learning_rate": 2.8191202526769896e-05, "loss": 0.0, "step": 5662 }, { "epoch": 1.31, "learning_rate": 2.818735074339419e-05, "loss": 0.0, "step": 5663 }, { "epoch": 1.31, "learning_rate": 2.818349896001849e-05, "loss": 0.0, "step": 5664 }, { "epoch": 1.31, "learning_rate": 2.8179647176642787e-05, "loss": 0.0, "step": 5665 }, { "epoch": 1.31, "learning_rate": 2.8175795393267086e-05, "loss": 0.0, "step": 5666 }, { "epoch": 1.31, "learning_rate": 2.8171943609891384e-05, "loss": 0.0, "step": 5667 }, { "epoch": 1.31, "learning_rate": 2.816809182651568e-05, "loss": 0.0, "step": 5668 }, { "epoch": 1.31, "learning_rate": 2.8164240043139977e-05, "loss": 0.0, "step": 5669 }, { "epoch": 1.31, "learning_rate": 2.816038825976427e-05, "loss": 0.0, "step": 5670 }, { "epoch": 1.31, "learning_rate": 2.8156536476388567e-05, "loss": 0.0, "step": 5671 }, { "epoch": 1.31, "learning_rate": 2.8152684693012866e-05, "loss": 0.0, "step": 5672 }, { "epoch": 1.31, "learning_rate": 2.8148832909637164e-05, "loss": 0.0, "step": 5673 }, { "epoch": 1.31, "learning_rate": 2.814498112626146e-05, "loss": 0.0001, "step": 5674 }, { "epoch": 1.31, "learning_rate": 2.8141129342885757e-05, "loss": 0.0, "step": 5675 }, { "epoch": 1.31, "learning_rate": 2.8137277559510056e-05, "loss": 0.0, "step": 5676 }, { "epoch": 1.31, "learning_rate": 2.8133425776134354e-05, "loss": 0.0001, "step": 5677 }, { "epoch": 1.31, "learning_rate": 2.8129573992758652e-05, "loss": 0.0, "step": 5678 }, { "epoch": 1.31, "learning_rate": 2.8125722209382944e-05, "loss": 0.0004, "step": 5679 }, { "epoch": 1.31, "learning_rate": 2.8121870426007242e-05, "loss": 0.0, "step": 5680 }, { "epoch": 1.31, "learning_rate": 2.8118018642631537e-05, "loss": 0.0, "step": 5681 }, { "epoch": 1.31, "learning_rate": 2.8114166859255836e-05, "loss": 0.0, "step": 5682 }, { "epoch": 1.31, "learning_rate": 2.8110315075880134e-05, "loss": 0.0, "step": 5683 }, { "epoch": 1.31, "learning_rate": 2.8106463292504432e-05, "loss": 0.0, "step": 5684 }, { "epoch": 1.31, "learning_rate": 2.810261150912873e-05, "loss": 0.0, "step": 5685 }, { "epoch": 1.31, "learning_rate": 2.8098759725753026e-05, "loss": 0.0, "step": 5686 }, { "epoch": 1.31, "learning_rate": 2.8094907942377324e-05, "loss": 0.0, "step": 5687 }, { "epoch": 1.31, "learning_rate": 2.8091056159001615e-05, "loss": 0.0, "step": 5688 }, { "epoch": 1.31, "learning_rate": 2.8087204375625914e-05, "loss": 0.0, "step": 5689 }, { "epoch": 1.31, "learning_rate": 2.8083352592250212e-05, "loss": 0.0, "step": 5690 }, { "epoch": 1.32, "learning_rate": 2.807950080887451e-05, "loss": 0.0, "step": 5691 }, { "epoch": 1.32, "learning_rate": 2.8075649025498805e-05, "loss": 0.0001, "step": 5692 }, { "epoch": 1.32, "learning_rate": 2.8071797242123104e-05, "loss": 0.0, "step": 5693 }, { "epoch": 1.32, "learning_rate": 2.8067945458747402e-05, "loss": 0.0, "step": 5694 }, { "epoch": 1.32, "learning_rate": 2.80640936753717e-05, "loss": 0.0, "step": 5695 }, { "epoch": 1.32, "learning_rate": 2.8060241891996e-05, "loss": 0.0, "step": 5696 }, { "epoch": 1.32, "learning_rate": 2.8056390108620294e-05, "loss": 0.0, "step": 5697 }, { "epoch": 1.32, "learning_rate": 2.805253832524459e-05, "loss": 0.0, "step": 5698 }, { "epoch": 1.32, "learning_rate": 2.8048686541868884e-05, "loss": 0.0, "step": 5699 }, { "epoch": 1.32, "learning_rate": 2.8044834758493182e-05, "loss": 0.0, "step": 5700 }, { "epoch": 1.32, "learning_rate": 2.804098297511748e-05, "loss": 0.0, "step": 5701 }, { "epoch": 1.32, "learning_rate": 2.803713119174178e-05, "loss": 0.0, "step": 5702 }, { "epoch": 1.32, "learning_rate": 2.8033279408366077e-05, "loss": 0.0, "step": 5703 }, { "epoch": 1.32, "learning_rate": 2.8029427624990372e-05, "loss": 0.0, "step": 5704 }, { "epoch": 1.32, "learning_rate": 2.802557584161467e-05, "loss": 0.0, "step": 5705 }, { "epoch": 1.32, "learning_rate": 2.802172405823897e-05, "loss": 0.0, "step": 5706 }, { "epoch": 1.32, "learning_rate": 2.801787227486326e-05, "loss": 0.0, "step": 5707 }, { "epoch": 1.32, "learning_rate": 2.801402049148756e-05, "loss": 0.0, "step": 5708 }, { "epoch": 1.32, "learning_rate": 2.8010168708111857e-05, "loss": 0.0, "step": 5709 }, { "epoch": 1.32, "learning_rate": 2.8006316924736152e-05, "loss": 0.0002, "step": 5710 }, { "epoch": 1.32, "learning_rate": 2.800246514136045e-05, "loss": 0.0, "step": 5711 }, { "epoch": 1.32, "learning_rate": 2.799861335798475e-05, "loss": 0.0, "step": 5712 }, { "epoch": 1.32, "learning_rate": 2.7994761574609047e-05, "loss": 0.0, "step": 5713 }, { "epoch": 1.32, "learning_rate": 2.7990909791233345e-05, "loss": 0.0, "step": 5714 }, { "epoch": 1.32, "learning_rate": 2.798705800785764e-05, "loss": 0.0, "step": 5715 }, { "epoch": 1.32, "learning_rate": 2.7983206224481935e-05, "loss": 0.0, "step": 5716 }, { "epoch": 1.32, "learning_rate": 2.797935444110623e-05, "loss": 0.0, "step": 5717 }, { "epoch": 1.32, "learning_rate": 2.797550265773053e-05, "loss": 0.0, "step": 5718 }, { "epoch": 1.32, "learning_rate": 2.7971650874354827e-05, "loss": 0.0, "step": 5719 }, { "epoch": 1.32, "learning_rate": 2.7967799090979125e-05, "loss": 0.0, "step": 5720 }, { "epoch": 1.32, "learning_rate": 2.7963947307603424e-05, "loss": 0.0, "step": 5721 }, { "epoch": 1.32, "learning_rate": 2.796009552422772e-05, "loss": 0.0, "step": 5722 }, { "epoch": 1.32, "learning_rate": 2.7956243740852017e-05, "loss": 0.0, "step": 5723 }, { "epoch": 1.32, "learning_rate": 2.7952391957476315e-05, "loss": 0.0, "step": 5724 }, { "epoch": 1.32, "learning_rate": 2.7948540174100614e-05, "loss": 0.0, "step": 5725 }, { "epoch": 1.32, "learning_rate": 2.7944688390724905e-05, "loss": 0.0, "step": 5726 }, { "epoch": 1.32, "learning_rate": 2.7940836607349203e-05, "loss": 0.0, "step": 5727 }, { "epoch": 1.32, "learning_rate": 2.79369848239735e-05, "loss": 0.0, "step": 5728 }, { "epoch": 1.32, "learning_rate": 2.7933133040597797e-05, "loss": 0.0, "step": 5729 }, { "epoch": 1.32, "learning_rate": 2.7929281257222095e-05, "loss": 0.0, "step": 5730 }, { "epoch": 1.32, "learning_rate": 2.7925429473846393e-05, "loss": 0.0, "step": 5731 }, { "epoch": 1.32, "learning_rate": 2.7921577690470692e-05, "loss": 0.0, "step": 5732 }, { "epoch": 1.32, "learning_rate": 2.7917725907094987e-05, "loss": 0.0, "step": 5733 }, { "epoch": 1.33, "learning_rate": 2.7913874123719285e-05, "loss": 0.0, "step": 5734 }, { "epoch": 1.33, "learning_rate": 2.7910022340343577e-05, "loss": 0.0, "step": 5735 }, { "epoch": 1.33, "learning_rate": 2.7906170556967875e-05, "loss": 0.0, "step": 5736 }, { "epoch": 1.33, "learning_rate": 2.7902318773592173e-05, "loss": 0.0, "step": 5737 }, { "epoch": 1.33, "learning_rate": 2.789846699021647e-05, "loss": 0.0, "step": 5738 }, { "epoch": 1.33, "learning_rate": 2.789461520684077e-05, "loss": 0.0, "step": 5739 }, { "epoch": 1.33, "learning_rate": 2.7890763423465065e-05, "loss": 0.0, "step": 5740 }, { "epoch": 1.33, "learning_rate": 2.7886911640089363e-05, "loss": 0.0001, "step": 5741 }, { "epoch": 1.33, "learning_rate": 2.788305985671366e-05, "loss": 0.0, "step": 5742 }, { "epoch": 1.33, "learning_rate": 2.787920807333796e-05, "loss": 0.0001, "step": 5743 }, { "epoch": 1.33, "learning_rate": 2.787535628996225e-05, "loss": 0.0, "step": 5744 }, { "epoch": 1.33, "learning_rate": 2.787150450658655e-05, "loss": 0.0, "step": 5745 }, { "epoch": 1.33, "learning_rate": 2.7867652723210845e-05, "loss": 0.0, "step": 5746 }, { "epoch": 1.33, "learning_rate": 2.7863800939835143e-05, "loss": 0.0, "step": 5747 }, { "epoch": 1.33, "learning_rate": 2.785994915645944e-05, "loss": 0.0, "step": 5748 }, { "epoch": 1.33, "learning_rate": 2.785609737308374e-05, "loss": 0.0, "step": 5749 }, { "epoch": 1.33, "learning_rate": 2.7852245589708038e-05, "loss": 0.0, "step": 5750 }, { "epoch": 1.33, "learning_rate": 2.7848393806332333e-05, "loss": 0.0, "step": 5751 }, { "epoch": 1.33, "learning_rate": 2.784454202295663e-05, "loss": 0.0, "step": 5752 }, { "epoch": 1.33, "learning_rate": 2.784069023958093e-05, "loss": 0.0, "step": 5753 }, { "epoch": 1.33, "learning_rate": 2.783683845620522e-05, "loss": 0.0, "step": 5754 }, { "epoch": 1.33, "learning_rate": 2.783298667282952e-05, "loss": 0.0, "step": 5755 }, { "epoch": 1.33, "learning_rate": 2.7829134889453818e-05, "loss": 0.0001, "step": 5756 }, { "epoch": 1.33, "learning_rate": 2.7825283106078116e-05, "loss": 0.0, "step": 5757 }, { "epoch": 1.33, "learning_rate": 2.782143132270241e-05, "loss": 0.0, "step": 5758 }, { "epoch": 1.33, "learning_rate": 2.781757953932671e-05, "loss": 0.0, "step": 5759 }, { "epoch": 1.33, "learning_rate": 2.7813727755951008e-05, "loss": 0.0, "step": 5760 }, { "epoch": 1.33, "learning_rate": 2.7809875972575306e-05, "loss": 0.0, "step": 5761 }, { "epoch": 1.33, "learning_rate": 2.7806024189199605e-05, "loss": 0.0, "step": 5762 }, { "epoch": 1.33, "learning_rate": 2.7802172405823896e-05, "loss": 0.0, "step": 5763 }, { "epoch": 1.33, "learning_rate": 2.779832062244819e-05, "loss": 0.0, "step": 5764 }, { "epoch": 1.33, "learning_rate": 2.779446883907249e-05, "loss": 0.0, "step": 5765 }, { "epoch": 1.33, "learning_rate": 2.7790617055696788e-05, "loss": 0.0, "step": 5766 }, { "epoch": 1.33, "learning_rate": 2.7786765272321086e-05, "loss": 0.0, "step": 5767 }, { "epoch": 1.33, "learning_rate": 2.7782913488945385e-05, "loss": 0.0, "step": 5768 }, { "epoch": 1.33, "learning_rate": 2.777906170556968e-05, "loss": 0.0, "step": 5769 }, { "epoch": 1.33, "learning_rate": 2.7775209922193978e-05, "loss": 0.0, "step": 5770 }, { "epoch": 1.33, "learning_rate": 2.7771358138818276e-05, "loss": 0.0, "step": 5771 }, { "epoch": 1.33, "learning_rate": 2.7767506355442568e-05, "loss": 0.0, "step": 5772 }, { "epoch": 1.33, "learning_rate": 2.7763654572066866e-05, "loss": 0.0, "step": 5773 }, { "epoch": 1.33, "learning_rate": 2.7759802788691165e-05, "loss": 0.0001, "step": 5774 }, { "epoch": 1.33, "learning_rate": 2.7755951005315463e-05, "loss": 0.0, "step": 5775 }, { "epoch": 1.33, "learning_rate": 2.7752099221939758e-05, "loss": 0.0, "step": 5776 }, { "epoch": 1.34, "learning_rate": 2.7748247438564056e-05, "loss": 0.0, "step": 5777 }, { "epoch": 1.34, "learning_rate": 2.7744395655188355e-05, "loss": 0.0, "step": 5778 }, { "epoch": 1.34, "learning_rate": 2.7740543871812653e-05, "loss": 0.0, "step": 5779 }, { "epoch": 1.34, "learning_rate": 2.773669208843695e-05, "loss": 0.0, "step": 5780 }, { "epoch": 1.34, "learning_rate": 2.7732840305061246e-05, "loss": 0.0, "step": 5781 }, { "epoch": 1.34, "learning_rate": 2.7728988521685538e-05, "loss": 0.0, "step": 5782 }, { "epoch": 1.34, "learning_rate": 2.7725136738309836e-05, "loss": 0.0, "step": 5783 }, { "epoch": 1.34, "learning_rate": 2.7721284954934134e-05, "loss": 0.0, "step": 5784 }, { "epoch": 1.34, "learning_rate": 2.7717433171558433e-05, "loss": 0.0, "step": 5785 }, { "epoch": 1.34, "learning_rate": 2.771358138818273e-05, "loss": 0.0, "step": 5786 }, { "epoch": 1.34, "learning_rate": 2.7709729604807026e-05, "loss": 0.0, "step": 5787 }, { "epoch": 1.34, "learning_rate": 2.7705877821431324e-05, "loss": 0.0, "step": 5788 }, { "epoch": 1.34, "learning_rate": 2.7702026038055623e-05, "loss": 0.0, "step": 5789 }, { "epoch": 1.34, "learning_rate": 2.769817425467992e-05, "loss": 0.0, "step": 5790 }, { "epoch": 1.34, "learning_rate": 2.7694322471304213e-05, "loss": 0.0, "step": 5791 }, { "epoch": 1.34, "learning_rate": 2.769047068792851e-05, "loss": 0.0, "step": 5792 }, { "epoch": 1.34, "learning_rate": 2.768661890455281e-05, "loss": 0.0, "step": 5793 }, { "epoch": 1.34, "learning_rate": 2.7682767121177104e-05, "loss": 0.0, "step": 5794 }, { "epoch": 1.34, "learning_rate": 2.7678915337801403e-05, "loss": 0.0, "step": 5795 }, { "epoch": 1.34, "learning_rate": 2.76750635544257e-05, "loss": 0.0, "step": 5796 }, { "epoch": 1.34, "learning_rate": 2.767121177105e-05, "loss": 0.0, "step": 5797 }, { "epoch": 1.34, "learning_rate": 2.7667359987674298e-05, "loss": 0.0, "step": 5798 }, { "epoch": 1.34, "learning_rate": 2.7663508204298593e-05, "loss": 0.0, "step": 5799 }, { "epoch": 1.34, "learning_rate": 2.7659656420922884e-05, "loss": 0.0, "step": 5800 }, { "epoch": 1.34, "learning_rate": 2.7655804637547183e-05, "loss": 0.0, "step": 5801 }, { "epoch": 1.34, "learning_rate": 2.765195285417148e-05, "loss": 0.0, "step": 5802 }, { "epoch": 1.34, "learning_rate": 2.764810107079578e-05, "loss": 0.0, "step": 5803 }, { "epoch": 1.34, "learning_rate": 2.7644249287420078e-05, "loss": 0.0, "step": 5804 }, { "epoch": 1.34, "learning_rate": 2.7640397504044373e-05, "loss": 0.0, "step": 5805 }, { "epoch": 1.34, "learning_rate": 2.763654572066867e-05, "loss": 0.0, "step": 5806 }, { "epoch": 1.34, "learning_rate": 2.763269393729297e-05, "loss": 0.0, "step": 5807 }, { "epoch": 1.34, "learning_rate": 2.7628842153917268e-05, "loss": 0.0, "step": 5808 }, { "epoch": 1.34, "learning_rate": 2.7624990370541566e-05, "loss": 0.0, "step": 5809 }, { "epoch": 1.34, "learning_rate": 2.7621138587165858e-05, "loss": 0.0, "step": 5810 }, { "epoch": 1.34, "learning_rate": 2.7617286803790156e-05, "loss": 0.0, "step": 5811 }, { "epoch": 1.34, "learning_rate": 2.761343502041445e-05, "loss": 0.0, "step": 5812 }, { "epoch": 1.34, "learning_rate": 2.760958323703875e-05, "loss": 0.0, "step": 5813 }, { "epoch": 1.34, "learning_rate": 2.7605731453663048e-05, "loss": 0.0, "step": 5814 }, { "epoch": 1.34, "learning_rate": 2.7601879670287346e-05, "loss": 0.0, "step": 5815 }, { "epoch": 1.34, "learning_rate": 2.7598027886911644e-05, "loss": 0.0, "step": 5816 }, { "epoch": 1.34, "learning_rate": 2.759417610353594e-05, "loss": 0.0, "step": 5817 }, { "epoch": 1.34, "learning_rate": 2.7590324320160238e-05, "loss": 0.0, "step": 5818 }, { "epoch": 1.34, "learning_rate": 2.758647253678453e-05, "loss": 0.0, "step": 5819 }, { "epoch": 1.35, "learning_rate": 2.7582620753408827e-05, "loss": 0.0, "step": 5820 }, { "epoch": 1.35, "learning_rate": 2.7578768970033126e-05, "loss": 0.0, "step": 5821 }, { "epoch": 1.35, "learning_rate": 2.7574917186657424e-05, "loss": 0.0, "step": 5822 }, { "epoch": 1.35, "learning_rate": 2.757106540328172e-05, "loss": 0.0, "step": 5823 }, { "epoch": 1.35, "learning_rate": 2.7567213619906017e-05, "loss": 0.0, "step": 5824 }, { "epoch": 1.35, "learning_rate": 2.7563361836530316e-05, "loss": 0.0, "step": 5825 }, { "epoch": 1.35, "learning_rate": 2.7559510053154614e-05, "loss": 0.0, "step": 5826 }, { "epoch": 1.35, "learning_rate": 2.7555658269778912e-05, "loss": 0.0, "step": 5827 }, { "epoch": 1.35, "learning_rate": 2.7551806486403204e-05, "loss": 0.0, "step": 5828 }, { "epoch": 1.35, "learning_rate": 2.7547954703027502e-05, "loss": 0.0, "step": 5829 }, { "epoch": 1.35, "learning_rate": 2.7544102919651797e-05, "loss": 0.0, "step": 5830 }, { "epoch": 1.35, "learning_rate": 2.7540251136276096e-05, "loss": 0.0, "step": 5831 }, { "epoch": 1.35, "learning_rate": 2.7536399352900394e-05, "loss": 0.0, "step": 5832 }, { "epoch": 1.35, "learning_rate": 2.7532547569524692e-05, "loss": 0.0, "step": 5833 }, { "epoch": 1.35, "learning_rate": 2.752869578614899e-05, "loss": 0.0, "step": 5834 }, { "epoch": 1.35, "learning_rate": 2.7524844002773286e-05, "loss": 0.0, "step": 5835 }, { "epoch": 1.35, "learning_rate": 2.7520992219397584e-05, "loss": 0.0, "step": 5836 }, { "epoch": 1.35, "learning_rate": 2.7517140436021882e-05, "loss": 0.0, "step": 5837 }, { "epoch": 1.35, "learning_rate": 2.7513288652646174e-05, "loss": 0.0, "step": 5838 }, { "epoch": 1.35, "learning_rate": 2.7509436869270472e-05, "loss": 0.0, "step": 5839 }, { "epoch": 1.35, "learning_rate": 2.750558508589477e-05, "loss": 0.0, "step": 5840 }, { "epoch": 1.35, "learning_rate": 2.7501733302519066e-05, "loss": 0.0, "step": 5841 }, { "epoch": 1.35, "learning_rate": 2.7497881519143364e-05, "loss": 0.0, "step": 5842 }, { "epoch": 1.35, "learning_rate": 2.7494029735767662e-05, "loss": 0.0, "step": 5843 }, { "epoch": 1.35, "learning_rate": 2.749017795239196e-05, "loss": 0.0, "step": 5844 }, { "epoch": 1.35, "learning_rate": 2.748632616901626e-05, "loss": 0.0, "step": 5845 }, { "epoch": 1.35, "learning_rate": 2.7482474385640554e-05, "loss": 0.0, "step": 5846 }, { "epoch": 1.35, "learning_rate": 2.747862260226485e-05, "loss": 0.0, "step": 5847 }, { "epoch": 1.35, "learning_rate": 2.7474770818889144e-05, "loss": 0.0005, "step": 5848 }, { "epoch": 1.35, "learning_rate": 2.7470919035513442e-05, "loss": 0.0, "step": 5849 }, { "epoch": 1.35, "learning_rate": 2.746706725213774e-05, "loss": 0.0, "step": 5850 }, { "epoch": 1.35, "learning_rate": 2.746321546876204e-05, "loss": 0.0, "step": 5851 }, { "epoch": 1.35, "learning_rate": 2.7459363685386337e-05, "loss": 0.0, "step": 5852 }, { "epoch": 1.35, "learning_rate": 2.7455511902010632e-05, "loss": 0.0, "step": 5853 }, { "epoch": 1.35, "learning_rate": 2.745166011863493e-05, "loss": 0.0, "step": 5854 }, { "epoch": 1.35, "learning_rate": 2.744780833525923e-05, "loss": 0.0, "step": 5855 }, { "epoch": 1.35, "learning_rate": 2.744395655188352e-05, "loss": 0.0, "step": 5856 }, { "epoch": 1.35, "learning_rate": 2.744010476850782e-05, "loss": 0.0, "step": 5857 }, { "epoch": 1.35, "learning_rate": 2.7436252985132117e-05, "loss": 0.0, "step": 5858 }, { "epoch": 1.35, "learning_rate": 2.7432401201756412e-05, "loss": 0.0072, "step": 5859 }, { "epoch": 1.35, "learning_rate": 2.742854941838071e-05, "loss": 0.0, "step": 5860 }, { "epoch": 1.35, "learning_rate": 2.742469763500501e-05, "loss": 0.0, "step": 5861 }, { "epoch": 1.35, "learning_rate": 2.7420845851629307e-05, "loss": 0.0, "step": 5862 }, { "epoch": 1.35, "learning_rate": 2.7416994068253605e-05, "loss": 0.0, "step": 5863 }, { "epoch": 1.36, "learning_rate": 2.74131422848779e-05, "loss": 0.0, "step": 5864 }, { "epoch": 1.36, "learning_rate": 2.74092905015022e-05, "loss": 0.0, "step": 5865 }, { "epoch": 1.36, "learning_rate": 2.740543871812649e-05, "loss": 0.0, "step": 5866 }, { "epoch": 1.36, "learning_rate": 2.740158693475079e-05, "loss": 0.0, "step": 5867 }, { "epoch": 1.36, "learning_rate": 2.7397735151375087e-05, "loss": 0.0, "step": 5868 }, { "epoch": 1.36, "learning_rate": 2.7393883367999385e-05, "loss": 0.0, "step": 5869 }, { "epoch": 1.36, "learning_rate": 2.7390031584623684e-05, "loss": 0.0, "step": 5870 }, { "epoch": 1.36, "learning_rate": 2.738617980124798e-05, "loss": 0.0, "step": 5871 }, { "epoch": 1.36, "learning_rate": 2.7382328017872277e-05, "loss": 0.0, "step": 5872 }, { "epoch": 1.36, "learning_rate": 2.7378476234496575e-05, "loss": 0.0, "step": 5873 }, { "epoch": 1.36, "learning_rate": 2.7374624451120874e-05, "loss": 0.0, "step": 5874 }, { "epoch": 1.36, "learning_rate": 2.7370772667745165e-05, "loss": 0.0, "step": 5875 }, { "epoch": 1.36, "learning_rate": 2.7366920884369464e-05, "loss": 0.0, "step": 5876 }, { "epoch": 1.36, "learning_rate": 2.736306910099376e-05, "loss": 0.0, "step": 5877 }, { "epoch": 1.36, "learning_rate": 2.7359217317618057e-05, "loss": 0.0, "step": 5878 }, { "epoch": 1.36, "learning_rate": 2.7355365534242355e-05, "loss": 0.0, "step": 5879 }, { "epoch": 1.36, "learning_rate": 2.7351513750866654e-05, "loss": 0.0, "step": 5880 }, { "epoch": 1.36, "learning_rate": 2.7347661967490952e-05, "loss": 0.0, "step": 5881 }, { "epoch": 1.36, "learning_rate": 2.7343810184115247e-05, "loss": 0.0, "step": 5882 }, { "epoch": 1.36, "learning_rate": 2.7339958400739545e-05, "loss": 0.0, "step": 5883 }, { "epoch": 1.36, "learning_rate": 2.7336106617363837e-05, "loss": 0.0002, "step": 5884 }, { "epoch": 1.36, "learning_rate": 2.7332254833988135e-05, "loss": 0.0, "step": 5885 }, { "epoch": 1.36, "learning_rate": 2.7328403050612433e-05, "loss": 0.0, "step": 5886 }, { "epoch": 1.36, "learning_rate": 2.7324551267236732e-05, "loss": 0.0, "step": 5887 }, { "epoch": 1.36, "learning_rate": 2.732069948386103e-05, "loss": 0.0, "step": 5888 }, { "epoch": 1.36, "learning_rate": 2.7316847700485325e-05, "loss": 0.0003, "step": 5889 }, { "epoch": 1.36, "learning_rate": 2.7312995917109623e-05, "loss": 0.0, "step": 5890 }, { "epoch": 1.36, "learning_rate": 2.7309144133733922e-05, "loss": 0.0, "step": 5891 }, { "epoch": 1.36, "learning_rate": 2.730529235035822e-05, "loss": 0.0, "step": 5892 }, { "epoch": 1.36, "learning_rate": 2.730144056698252e-05, "loss": 0.0, "step": 5893 }, { "epoch": 1.36, "learning_rate": 2.729758878360681e-05, "loss": 0.0, "step": 5894 }, { "epoch": 1.36, "learning_rate": 2.7293737000231105e-05, "loss": 0.0, "step": 5895 }, { "epoch": 1.36, "learning_rate": 2.7289885216855403e-05, "loss": 0.0001, "step": 5896 }, { "epoch": 1.36, "learning_rate": 2.72860334334797e-05, "loss": 0.0, "step": 5897 }, { "epoch": 1.36, "learning_rate": 2.7282181650104e-05, "loss": 0.0098, "step": 5898 }, { "epoch": 1.36, "learning_rate": 2.72783298667283e-05, "loss": 0.0, "step": 5899 }, { "epoch": 1.36, "learning_rate": 2.7274478083352593e-05, "loss": 0.0, "step": 5900 }, { "epoch": 1.36, "learning_rate": 2.727062629997689e-05, "loss": 0.0, "step": 5901 }, { "epoch": 1.36, "learning_rate": 2.726677451660119e-05, "loss": 0.0, "step": 5902 }, { "epoch": 1.36, "learning_rate": 2.726292273322548e-05, "loss": 0.0, "step": 5903 }, { "epoch": 1.36, "learning_rate": 2.725907094984978e-05, "loss": 0.0, "step": 5904 }, { "epoch": 1.36, "learning_rate": 2.7255219166474078e-05, "loss": 0.0, "step": 5905 }, { "epoch": 1.36, "learning_rate": 2.7251367383098377e-05, "loss": 0.0291, "step": 5906 }, { "epoch": 1.37, "learning_rate": 2.724751559972267e-05, "loss": 0.0, "step": 5907 }, { "epoch": 1.37, "learning_rate": 2.724366381634697e-05, "loss": 0.0, "step": 5908 }, { "epoch": 1.37, "learning_rate": 2.7239812032971268e-05, "loss": 0.0, "step": 5909 }, { "epoch": 1.37, "learning_rate": 2.7235960249595567e-05, "loss": 0.0, "step": 5910 }, { "epoch": 1.37, "learning_rate": 2.7232108466219865e-05, "loss": 0.0, "step": 5911 }, { "epoch": 1.37, "learning_rate": 2.7228256682844156e-05, "loss": 0.0, "step": 5912 }, { "epoch": 1.37, "learning_rate": 2.722440489946845e-05, "loss": 0.0, "step": 5913 }, { "epoch": 1.37, "learning_rate": 2.722055311609275e-05, "loss": 0.0, "step": 5914 }, { "epoch": 1.37, "learning_rate": 2.7216701332717048e-05, "loss": 0.0, "step": 5915 }, { "epoch": 1.37, "learning_rate": 2.7212849549341346e-05, "loss": 0.0, "step": 5916 }, { "epoch": 1.37, "learning_rate": 2.7208997765965645e-05, "loss": 0.0001, "step": 5917 }, { "epoch": 1.37, "learning_rate": 2.720514598258994e-05, "loss": 0.0, "step": 5918 }, { "epoch": 1.37, "learning_rate": 2.7201294199214238e-05, "loss": 0.0001, "step": 5919 }, { "epoch": 1.37, "learning_rate": 2.7197442415838536e-05, "loss": 0.0, "step": 5920 }, { "epoch": 1.37, "learning_rate": 2.7193590632462835e-05, "loss": 0.0, "step": 5921 }, { "epoch": 1.37, "learning_rate": 2.7189738849087126e-05, "loss": 0.0, "step": 5922 }, { "epoch": 1.37, "learning_rate": 2.7185887065711425e-05, "loss": 0.0, "step": 5923 }, { "epoch": 1.37, "learning_rate": 2.7182035282335723e-05, "loss": 0.0, "step": 5924 }, { "epoch": 1.37, "learning_rate": 2.7178183498960018e-05, "loss": 0.0, "step": 5925 }, { "epoch": 1.37, "learning_rate": 2.7174331715584316e-05, "loss": 0.0001, "step": 5926 }, { "epoch": 1.37, "learning_rate": 2.7170479932208615e-05, "loss": 0.0, "step": 5927 }, { "epoch": 1.37, "learning_rate": 2.7166628148832913e-05, "loss": 0.0, "step": 5928 }, { "epoch": 1.37, "learning_rate": 2.716277636545721e-05, "loss": 0.0, "step": 5929 }, { "epoch": 1.37, "learning_rate": 2.7158924582081506e-05, "loss": 0.0, "step": 5930 }, { "epoch": 1.37, "learning_rate": 2.7155072798705798e-05, "loss": 0.0, "step": 5931 }, { "epoch": 1.37, "learning_rate": 2.7151221015330096e-05, "loss": 0.0, "step": 5932 }, { "epoch": 1.37, "learning_rate": 2.7147369231954395e-05, "loss": 0.0, "step": 5933 }, { "epoch": 1.37, "learning_rate": 2.7143517448578693e-05, "loss": 0.0, "step": 5934 }, { "epoch": 1.37, "learning_rate": 2.713966566520299e-05, "loss": 0.0, "step": 5935 }, { "epoch": 1.37, "learning_rate": 2.7135813881827286e-05, "loss": 0.0, "step": 5936 }, { "epoch": 1.37, "learning_rate": 2.7131962098451585e-05, "loss": 0.0003, "step": 5937 }, { "epoch": 1.37, "learning_rate": 2.7128110315075883e-05, "loss": 0.0, "step": 5938 }, { "epoch": 1.37, "learning_rate": 2.712425853170018e-05, "loss": 0.0, "step": 5939 }, { "epoch": 1.37, "learning_rate": 2.7120406748324473e-05, "loss": 0.0, "step": 5940 }, { "epoch": 1.37, "learning_rate": 2.711655496494877e-05, "loss": 0.0, "step": 5941 }, { "epoch": 1.37, "learning_rate": 2.711270318157307e-05, "loss": 0.0, "step": 5942 }, { "epoch": 1.37, "learning_rate": 2.7108851398197364e-05, "loss": 0.0, "step": 5943 }, { "epoch": 1.37, "learning_rate": 2.7104999614821663e-05, "loss": 0.0, "step": 5944 }, { "epoch": 1.37, "learning_rate": 2.710114783144596e-05, "loss": 0.0, "step": 5945 }, { "epoch": 1.37, "learning_rate": 2.709729604807026e-05, "loss": 0.0, "step": 5946 }, { "epoch": 1.37, "learning_rate": 2.7093444264694558e-05, "loss": 0.0, "step": 5947 }, { "epoch": 1.37, "learning_rate": 2.7089592481318853e-05, "loss": 0.0, "step": 5948 }, { "epoch": 1.37, "learning_rate": 2.708574069794315e-05, "loss": 0.0, "step": 5949 }, { "epoch": 1.38, "learning_rate": 2.7081888914567443e-05, "loss": 0.0, "step": 5950 }, { "epoch": 1.38, "learning_rate": 2.707803713119174e-05, "loss": 0.0002, "step": 5951 }, { "epoch": 1.38, "learning_rate": 2.707418534781604e-05, "loss": 0.0, "step": 5952 }, { "epoch": 1.38, "learning_rate": 2.7070333564440338e-05, "loss": 0.0001, "step": 5953 }, { "epoch": 1.38, "learning_rate": 2.7066481781064633e-05, "loss": 0.0, "step": 5954 }, { "epoch": 1.38, "learning_rate": 2.706262999768893e-05, "loss": 0.0001, "step": 5955 }, { "epoch": 1.38, "learning_rate": 2.705877821431323e-05, "loss": 0.0001, "step": 5956 }, { "epoch": 1.38, "learning_rate": 2.7054926430937528e-05, "loss": 0.0, "step": 5957 }, { "epoch": 1.38, "learning_rate": 2.7051074647561826e-05, "loss": 0.0, "step": 5958 }, { "epoch": 1.38, "learning_rate": 2.7047222864186118e-05, "loss": 0.0, "step": 5959 }, { "epoch": 1.38, "learning_rate": 2.7043371080810416e-05, "loss": 0.0, "step": 5960 }, { "epoch": 1.38, "learning_rate": 2.703951929743471e-05, "loss": 0.0, "step": 5961 }, { "epoch": 1.38, "learning_rate": 2.703566751405901e-05, "loss": 0.0, "step": 5962 }, { "epoch": 1.38, "learning_rate": 2.7031815730683308e-05, "loss": 0.0, "step": 5963 }, { "epoch": 1.38, "learning_rate": 2.7027963947307606e-05, "loss": 0.0, "step": 5964 }, { "epoch": 1.38, "learning_rate": 2.7024112163931904e-05, "loss": 0.0, "step": 5965 }, { "epoch": 1.38, "learning_rate": 2.70202603805562e-05, "loss": 0.0, "step": 5966 }, { "epoch": 1.38, "learning_rate": 2.7016408597180498e-05, "loss": 0.0, "step": 5967 }, { "epoch": 1.38, "learning_rate": 2.701255681380479e-05, "loss": 0.0, "step": 5968 }, { "epoch": 1.38, "learning_rate": 2.7008705030429088e-05, "loss": 0.0, "step": 5969 }, { "epoch": 1.38, "learning_rate": 2.7004853247053386e-05, "loss": 0.0, "step": 5970 }, { "epoch": 1.38, "learning_rate": 2.7001001463677684e-05, "loss": 0.0, "step": 5971 }, { "epoch": 1.38, "learning_rate": 2.699714968030198e-05, "loss": 0.0, "step": 5972 }, { "epoch": 1.38, "learning_rate": 2.6993297896926277e-05, "loss": 0.0001, "step": 5973 }, { "epoch": 1.38, "learning_rate": 2.6989446113550576e-05, "loss": 0.0, "step": 5974 }, { "epoch": 1.38, "learning_rate": 2.6985594330174874e-05, "loss": 0.0, "step": 5975 }, { "epoch": 1.38, "learning_rate": 2.6981742546799173e-05, "loss": 0.0, "step": 5976 }, { "epoch": 1.38, "learning_rate": 2.6977890763423467e-05, "loss": 0.0, "step": 5977 }, { "epoch": 1.38, "learning_rate": 2.6974038980047762e-05, "loss": 0.0001, "step": 5978 }, { "epoch": 1.38, "learning_rate": 2.6970187196672057e-05, "loss": 0.0, "step": 5979 }, { "epoch": 1.38, "learning_rate": 2.6966335413296356e-05, "loss": 0.0, "step": 5980 }, { "epoch": 1.38, "learning_rate": 2.6962483629920654e-05, "loss": 0.0, "step": 5981 }, { "epoch": 1.38, "learning_rate": 2.6958631846544952e-05, "loss": 0.0, "step": 5982 }, { "epoch": 1.38, "learning_rate": 2.695478006316925e-05, "loss": 0.0, "step": 5983 }, { "epoch": 1.38, "learning_rate": 2.6950928279793546e-05, "loss": 0.0, "step": 5984 }, { "epoch": 1.38, "learning_rate": 2.6947076496417844e-05, "loss": 0.0, "step": 5985 }, { "epoch": 1.38, "learning_rate": 2.6943224713042142e-05, "loss": 0.0, "step": 5986 }, { "epoch": 1.38, "learning_rate": 2.6939372929666434e-05, "loss": 0.0, "step": 5987 }, { "epoch": 1.38, "learning_rate": 2.6935521146290732e-05, "loss": 0.0, "step": 5988 }, { "epoch": 1.38, "learning_rate": 2.693166936291503e-05, "loss": 0.0, "step": 5989 }, { "epoch": 1.38, "learning_rate": 2.6927817579539326e-05, "loss": 0.0, "step": 5990 }, { "epoch": 1.38, "learning_rate": 2.6923965796163624e-05, "loss": 0.0, "step": 5991 }, { "epoch": 1.38, "learning_rate": 2.6920114012787922e-05, "loss": 0.0, "step": 5992 }, { "epoch": 1.39, "learning_rate": 2.691626222941222e-05, "loss": 0.0, "step": 5993 }, { "epoch": 1.39, "learning_rate": 2.691241044603652e-05, "loss": 0.0, "step": 5994 }, { "epoch": 1.39, "learning_rate": 2.6908558662660814e-05, "loss": 0.0, "step": 5995 }, { "epoch": 1.39, "learning_rate": 2.690470687928511e-05, "loss": 0.0, "step": 5996 }, { "epoch": 1.39, "learning_rate": 2.6900855095909404e-05, "loss": 0.0, "step": 5997 }, { "epoch": 1.39, "learning_rate": 2.6897003312533702e-05, "loss": 0.0, "step": 5998 }, { "epoch": 1.39, "learning_rate": 2.6893151529158e-05, "loss": 0.0, "step": 5999 }, { "epoch": 1.39, "learning_rate": 2.68892997457823e-05, "loss": 0.0, "step": 6000 }, { "epoch": 1.39, "learning_rate": 2.6885447962406597e-05, "loss": 0.0, "step": 6001 }, { "epoch": 1.39, "learning_rate": 2.6881596179030892e-05, "loss": 0.0001, "step": 6002 }, { "epoch": 1.39, "learning_rate": 2.687774439565519e-05, "loss": 0.0, "step": 6003 }, { "epoch": 1.39, "learning_rate": 2.687389261227949e-05, "loss": 0.0, "step": 6004 }, { "epoch": 1.39, "learning_rate": 2.6870040828903787e-05, "loss": 0.0, "step": 6005 }, { "epoch": 1.39, "learning_rate": 2.686618904552808e-05, "loss": 0.0, "step": 6006 }, { "epoch": 1.39, "learning_rate": 2.6862337262152377e-05, "loss": 0.0, "step": 6007 }, { "epoch": 1.39, "learning_rate": 2.6858485478776672e-05, "loss": 0.0, "step": 6008 }, { "epoch": 1.39, "learning_rate": 2.685463369540097e-05, "loss": 0.0, "step": 6009 }, { "epoch": 1.39, "learning_rate": 2.685078191202527e-05, "loss": 0.0, "step": 6010 }, { "epoch": 1.39, "learning_rate": 2.6846930128649567e-05, "loss": 0.0, "step": 6011 }, { "epoch": 1.39, "learning_rate": 2.6843078345273865e-05, "loss": 0.0, "step": 6012 }, { "epoch": 1.39, "learning_rate": 2.683922656189816e-05, "loss": 0.0, "step": 6013 }, { "epoch": 1.39, "learning_rate": 2.683537477852246e-05, "loss": 0.0, "step": 6014 }, { "epoch": 1.39, "learning_rate": 2.683152299514675e-05, "loss": 0.0, "step": 6015 }, { "epoch": 1.39, "learning_rate": 2.682767121177105e-05, "loss": 0.0, "step": 6016 }, { "epoch": 1.39, "learning_rate": 2.6823819428395347e-05, "loss": 0.0, "step": 6017 }, { "epoch": 1.39, "learning_rate": 2.6819967645019645e-05, "loss": 0.0, "step": 6018 }, { "epoch": 1.39, "learning_rate": 2.6816115861643944e-05, "loss": 0.0, "step": 6019 }, { "epoch": 1.39, "learning_rate": 2.681226407826824e-05, "loss": 0.0002, "step": 6020 }, { "epoch": 1.39, "learning_rate": 2.6808412294892537e-05, "loss": 0.0, "step": 6021 }, { "epoch": 1.39, "learning_rate": 2.6804560511516835e-05, "loss": 0.0, "step": 6022 }, { "epoch": 1.39, "learning_rate": 2.6800708728141134e-05, "loss": 0.0003, "step": 6023 }, { "epoch": 1.39, "learning_rate": 2.6796856944765425e-05, "loss": 0.0, "step": 6024 }, { "epoch": 1.39, "learning_rate": 2.6793005161389724e-05, "loss": 0.0, "step": 6025 }, { "epoch": 1.39, "learning_rate": 2.678915337801402e-05, "loss": 0.0, "step": 6026 }, { "epoch": 1.39, "learning_rate": 2.6785301594638317e-05, "loss": 0.0, "step": 6027 }, { "epoch": 1.39, "learning_rate": 2.6781449811262615e-05, "loss": 0.0, "step": 6028 }, { "epoch": 1.39, "learning_rate": 2.6777598027886914e-05, "loss": 0.0, "step": 6029 }, { "epoch": 1.39, "learning_rate": 2.6773746244511212e-05, "loss": 0.0, "step": 6030 }, { "epoch": 1.39, "learning_rate": 2.6769894461135507e-05, "loss": 0.0, "step": 6031 }, { "epoch": 1.39, "learning_rate": 2.6766042677759805e-05, "loss": 0.0, "step": 6032 }, { "epoch": 1.39, "learning_rate": 2.6762190894384104e-05, "loss": 0.0, "step": 6033 }, { "epoch": 1.39, "learning_rate": 2.6758339111008395e-05, "loss": 0.0, "step": 6034 }, { "epoch": 1.39, "learning_rate": 2.6754487327632693e-05, "loss": 0.0001, "step": 6035 }, { "epoch": 1.39, "learning_rate": 2.6750635544256992e-05, "loss": 0.0, "step": 6036 }, { "epoch": 1.4, "learning_rate": 2.674678376088129e-05, "loss": 0.0, "step": 6037 }, { "epoch": 1.4, "learning_rate": 2.6742931977505585e-05, "loss": 0.0, "step": 6038 }, { "epoch": 1.4, "learning_rate": 2.6739080194129883e-05, "loss": 0.0, "step": 6039 }, { "epoch": 1.4, "learning_rate": 2.6735228410754182e-05, "loss": 0.0, "step": 6040 }, { "epoch": 1.4, "learning_rate": 2.673137662737848e-05, "loss": 0.0, "step": 6041 }, { "epoch": 1.4, "learning_rate": 2.672752484400278e-05, "loss": 0.0, "step": 6042 }, { "epoch": 1.4, "learning_rate": 2.672367306062707e-05, "loss": 0.0, "step": 6043 }, { "epoch": 1.4, "learning_rate": 2.671982127725137e-05, "loss": 0.0, "step": 6044 }, { "epoch": 1.4, "learning_rate": 2.6715969493875663e-05, "loss": 0.0, "step": 6045 }, { "epoch": 1.4, "learning_rate": 2.6712117710499962e-05, "loss": 0.0, "step": 6046 }, { "epoch": 1.4, "learning_rate": 2.670826592712426e-05, "loss": 0.0, "step": 6047 }, { "epoch": 1.4, "learning_rate": 2.670441414374856e-05, "loss": 0.0, "step": 6048 }, { "epoch": 1.4, "learning_rate": 2.6700562360372853e-05, "loss": 0.0, "step": 6049 }, { "epoch": 1.4, "learning_rate": 2.6696710576997152e-05, "loss": 0.0, "step": 6050 }, { "epoch": 1.4, "learning_rate": 2.669285879362145e-05, "loss": 0.0, "step": 6051 }, { "epoch": 1.4, "learning_rate": 2.668900701024574e-05, "loss": 0.0, "step": 6052 }, { "epoch": 1.4, "learning_rate": 2.668515522687004e-05, "loss": 0.0, "step": 6053 }, { "epoch": 1.4, "learning_rate": 2.6681303443494338e-05, "loss": 0.0, "step": 6054 }, { "epoch": 1.4, "learning_rate": 2.6677451660118637e-05, "loss": 0.0, "step": 6055 }, { "epoch": 1.4, "learning_rate": 2.667359987674293e-05, "loss": 0.0001, "step": 6056 }, { "epoch": 1.4, "learning_rate": 2.666974809336723e-05, "loss": 0.0, "step": 6057 }, { "epoch": 1.4, "learning_rate": 2.6665896309991528e-05, "loss": 0.0, "step": 6058 }, { "epoch": 1.4, "learning_rate": 2.6662044526615827e-05, "loss": 0.0, "step": 6059 }, { "epoch": 1.4, "learning_rate": 2.6658192743240125e-05, "loss": 0.0, "step": 6060 }, { "epoch": 1.4, "learning_rate": 2.665434095986442e-05, "loss": 0.0, "step": 6061 }, { "epoch": 1.4, "learning_rate": 2.6650489176488715e-05, "loss": 0.0, "step": 6062 }, { "epoch": 1.4, "learning_rate": 2.664663739311301e-05, "loss": 0.0, "step": 6063 }, { "epoch": 1.4, "learning_rate": 2.6642785609737308e-05, "loss": 0.0, "step": 6064 }, { "epoch": 1.4, "learning_rate": 2.6638933826361607e-05, "loss": 0.0, "step": 6065 }, { "epoch": 1.4, "learning_rate": 2.6635082042985905e-05, "loss": 0.0001, "step": 6066 }, { "epoch": 1.4, "learning_rate": 2.66312302596102e-05, "loss": 0.0001, "step": 6067 }, { "epoch": 1.4, "learning_rate": 2.6627378476234498e-05, "loss": 0.0, "step": 6068 }, { "epoch": 1.4, "learning_rate": 2.6623526692858797e-05, "loss": 0.0, "step": 6069 }, { "epoch": 1.4, "learning_rate": 2.6619674909483095e-05, "loss": 0.0, "step": 6070 }, { "epoch": 1.4, "learning_rate": 2.6615823126107386e-05, "loss": 0.0, "step": 6071 }, { "epoch": 1.4, "learning_rate": 2.6611971342731685e-05, "loss": 0.0, "step": 6072 }, { "epoch": 1.4, "learning_rate": 2.6608119559355983e-05, "loss": 0.0, "step": 6073 }, { "epoch": 1.4, "learning_rate": 2.6604267775980278e-05, "loss": 0.0, "step": 6074 }, { "epoch": 1.4, "learning_rate": 2.6600415992604576e-05, "loss": 0.0001, "step": 6075 }, { "epoch": 1.4, "learning_rate": 2.6596564209228875e-05, "loss": 0.0, "step": 6076 }, { "epoch": 1.4, "learning_rate": 2.6592712425853173e-05, "loss": 0.0, "step": 6077 }, { "epoch": 1.4, "learning_rate": 2.658886064247747e-05, "loss": 0.0, "step": 6078 }, { "epoch": 1.4, "learning_rate": 2.6585008859101766e-05, "loss": 0.0, "step": 6079 }, { "epoch": 1.41, "learning_rate": 2.658115707572606e-05, "loss": 0.0, "step": 6080 }, { "epoch": 1.41, "learning_rate": 2.6577305292350356e-05, "loss": 0.0, "step": 6081 }, { "epoch": 1.41, "learning_rate": 2.6573453508974655e-05, "loss": 0.0, "step": 6082 }, { "epoch": 1.41, "learning_rate": 2.6569601725598953e-05, "loss": 0.0, "step": 6083 }, { "epoch": 1.41, "learning_rate": 2.656574994222325e-05, "loss": 0.0, "step": 6084 }, { "epoch": 1.41, "learning_rate": 2.6561898158847546e-05, "loss": 0.0, "step": 6085 }, { "epoch": 1.41, "learning_rate": 2.6558046375471845e-05, "loss": 0.0, "step": 6086 }, { "epoch": 1.41, "learning_rate": 2.6554194592096143e-05, "loss": 0.0, "step": 6087 }, { "epoch": 1.41, "learning_rate": 2.655034280872044e-05, "loss": 0.0, "step": 6088 }, { "epoch": 1.41, "learning_rate": 2.654649102534474e-05, "loss": 0.0, "step": 6089 }, { "epoch": 1.41, "learning_rate": 2.654263924196903e-05, "loss": 0.0, "step": 6090 }, { "epoch": 1.41, "learning_rate": 2.653878745859333e-05, "loss": 0.0, "step": 6091 }, { "epoch": 1.41, "learning_rate": 2.6534935675217625e-05, "loss": 0.0, "step": 6092 }, { "epoch": 1.41, "learning_rate": 2.6531083891841923e-05, "loss": 0.0, "step": 6093 }, { "epoch": 1.41, "learning_rate": 2.652723210846622e-05, "loss": 0.0, "step": 6094 }, { "epoch": 1.41, "learning_rate": 2.652338032509052e-05, "loss": 0.0, "step": 6095 }, { "epoch": 1.41, "learning_rate": 2.6519528541714818e-05, "loss": 0.0, "step": 6096 }, { "epoch": 1.41, "learning_rate": 2.6515676758339113e-05, "loss": 0.0041, "step": 6097 }, { "epoch": 1.41, "learning_rate": 2.651182497496341e-05, "loss": 0.0, "step": 6098 }, { "epoch": 1.41, "learning_rate": 2.6507973191587703e-05, "loss": 0.0, "step": 6099 }, { "epoch": 1.41, "learning_rate": 2.6504121408212e-05, "loss": 0.0, "step": 6100 }, { "epoch": 1.41, "learning_rate": 2.65002696248363e-05, "loss": 0.0, "step": 6101 }, { "epoch": 1.41, "learning_rate": 2.6496417841460598e-05, "loss": 0.0, "step": 6102 }, { "epoch": 1.41, "learning_rate": 2.6492566058084893e-05, "loss": 0.0, "step": 6103 }, { "epoch": 1.41, "learning_rate": 2.648871427470919e-05, "loss": 0.0, "step": 6104 }, { "epoch": 1.41, "learning_rate": 2.648486249133349e-05, "loss": 0.0, "step": 6105 }, { "epoch": 1.41, "learning_rate": 2.6481010707957788e-05, "loss": 0.0, "step": 6106 }, { "epoch": 1.41, "learning_rate": 2.6477158924582086e-05, "loss": 0.0, "step": 6107 }, { "epoch": 1.41, "learning_rate": 2.6473307141206378e-05, "loss": 0.0, "step": 6108 }, { "epoch": 1.41, "learning_rate": 2.6469455357830676e-05, "loss": 0.0, "step": 6109 }, { "epoch": 1.41, "learning_rate": 2.646560357445497e-05, "loss": 0.0, "step": 6110 }, { "epoch": 1.41, "learning_rate": 2.646175179107927e-05, "loss": 0.0, "step": 6111 }, { "epoch": 1.41, "learning_rate": 2.6457900007703568e-05, "loss": 0.0, "step": 6112 }, { "epoch": 1.41, "learning_rate": 2.6454048224327866e-05, "loss": 0.0, "step": 6113 }, { "epoch": 1.41, "learning_rate": 2.6450196440952164e-05, "loss": 0.0, "step": 6114 }, { "epoch": 1.41, "learning_rate": 2.644634465757646e-05, "loss": 0.0001, "step": 6115 }, { "epoch": 1.41, "learning_rate": 2.6442492874200758e-05, "loss": 0.0, "step": 6116 }, { "epoch": 1.41, "learning_rate": 2.6438641090825056e-05, "loss": 0.0, "step": 6117 }, { "epoch": 1.41, "learning_rate": 2.6434789307449348e-05, "loss": 0.0, "step": 6118 }, { "epoch": 1.41, "learning_rate": 2.6430937524073646e-05, "loss": 0.0, "step": 6119 }, { "epoch": 1.41, "learning_rate": 2.6427085740697944e-05, "loss": 0.0, "step": 6120 }, { "epoch": 1.41, "learning_rate": 2.6423233957322243e-05, "loss": 0.0, "step": 6121 }, { "epoch": 1.41, "learning_rate": 2.6419382173946538e-05, "loss": 0.0, "step": 6122 }, { "epoch": 1.42, "learning_rate": 2.6415530390570836e-05, "loss": 0.0, "step": 6123 }, { "epoch": 1.42, "learning_rate": 2.6411678607195134e-05, "loss": 0.0, "step": 6124 }, { "epoch": 1.42, "learning_rate": 2.6407826823819433e-05, "loss": 0.0, "step": 6125 }, { "epoch": 1.42, "learning_rate": 2.6403975040443728e-05, "loss": 0.0001, "step": 6126 }, { "epoch": 1.42, "learning_rate": 2.6400123257068023e-05, "loss": 0.0, "step": 6127 }, { "epoch": 1.42, "learning_rate": 2.6396271473692317e-05, "loss": 0.0, "step": 6128 }, { "epoch": 1.42, "learning_rate": 2.6392419690316616e-05, "loss": 0.0, "step": 6129 }, { "epoch": 1.42, "learning_rate": 2.6388567906940914e-05, "loss": 0.0, "step": 6130 }, { "epoch": 1.42, "learning_rate": 2.6384716123565213e-05, "loss": 0.0, "step": 6131 }, { "epoch": 1.42, "learning_rate": 2.638086434018951e-05, "loss": 0.0, "step": 6132 }, { "epoch": 1.42, "learning_rate": 2.6377012556813806e-05, "loss": 0.0, "step": 6133 }, { "epoch": 1.42, "learning_rate": 2.6373160773438104e-05, "loss": 0.0, "step": 6134 }, { "epoch": 1.42, "learning_rate": 2.6369308990062402e-05, "loss": 0.0, "step": 6135 }, { "epoch": 1.42, "learning_rate": 2.6365457206686694e-05, "loss": 0.0, "step": 6136 }, { "epoch": 1.42, "learning_rate": 2.6361605423310992e-05, "loss": 0.0, "step": 6137 }, { "epoch": 1.42, "learning_rate": 2.635775363993529e-05, "loss": 0.0, "step": 6138 }, { "epoch": 1.42, "learning_rate": 2.635390185655959e-05, "loss": 0.0, "step": 6139 }, { "epoch": 1.42, "learning_rate": 2.6350050073183884e-05, "loss": 0.0, "step": 6140 }, { "epoch": 1.42, "learning_rate": 2.6346198289808182e-05, "loss": 0.0, "step": 6141 }, { "epoch": 1.42, "learning_rate": 2.634234650643248e-05, "loss": 0.0, "step": 6142 }, { "epoch": 1.42, "learning_rate": 2.633849472305678e-05, "loss": 0.0, "step": 6143 }, { "epoch": 1.42, "learning_rate": 2.6334642939681074e-05, "loss": 0.0, "step": 6144 }, { "epoch": 1.42, "learning_rate": 2.6330791156305372e-05, "loss": 0.0, "step": 6145 }, { "epoch": 1.42, "learning_rate": 2.6326939372929664e-05, "loss": 0.0, "step": 6146 }, { "epoch": 1.42, "learning_rate": 2.6323087589553962e-05, "loss": 0.0, "step": 6147 }, { "epoch": 1.42, "learning_rate": 2.631923580617826e-05, "loss": 0.0, "step": 6148 }, { "epoch": 1.42, "learning_rate": 2.631538402280256e-05, "loss": 0.0, "step": 6149 }, { "epoch": 1.42, "learning_rate": 2.6311532239426857e-05, "loss": 0.0, "step": 6150 }, { "epoch": 1.42, "learning_rate": 2.6307680456051152e-05, "loss": 0.0001, "step": 6151 }, { "epoch": 1.42, "learning_rate": 2.630382867267545e-05, "loss": 0.0, "step": 6152 }, { "epoch": 1.42, "learning_rate": 2.629997688929975e-05, "loss": 0.0, "step": 6153 }, { "epoch": 1.42, "learning_rate": 2.6296125105924047e-05, "loss": 0.0, "step": 6154 }, { "epoch": 1.42, "learning_rate": 2.629227332254834e-05, "loss": 0.0, "step": 6155 }, { "epoch": 1.42, "learning_rate": 2.6288421539172637e-05, "loss": 0.0, "step": 6156 }, { "epoch": 1.42, "learning_rate": 2.6284569755796936e-05, "loss": 0.0, "step": 6157 }, { "epoch": 1.42, "learning_rate": 2.628071797242123e-05, "loss": 0.0, "step": 6158 }, { "epoch": 1.42, "learning_rate": 2.627686618904553e-05, "loss": 0.0, "step": 6159 }, { "epoch": 1.42, "learning_rate": 2.6273014405669827e-05, "loss": 0.0, "step": 6160 }, { "epoch": 1.42, "learning_rate": 2.6269162622294126e-05, "loss": 0.0, "step": 6161 }, { "epoch": 1.42, "learning_rate": 2.626531083891842e-05, "loss": 0.0, "step": 6162 }, { "epoch": 1.42, "learning_rate": 2.626145905554272e-05, "loss": 0.0, "step": 6163 }, { "epoch": 1.42, "learning_rate": 2.625760727216701e-05, "loss": 0.0, "step": 6164 }, { "epoch": 1.42, "learning_rate": 2.625375548879131e-05, "loss": 0.0, "step": 6165 }, { "epoch": 1.43, "learning_rate": 2.6249903705415607e-05, "loss": 0.0, "step": 6166 }, { "epoch": 1.43, "learning_rate": 2.6246051922039905e-05, "loss": 0.0, "step": 6167 }, { "epoch": 1.43, "learning_rate": 2.6242200138664204e-05, "loss": 0.0, "step": 6168 }, { "epoch": 1.43, "learning_rate": 2.62383483552885e-05, "loss": 0.0, "step": 6169 }, { "epoch": 1.43, "learning_rate": 2.6234496571912797e-05, "loss": 0.0, "step": 6170 }, { "epoch": 1.43, "learning_rate": 2.6230644788537095e-05, "loss": 0.0, "step": 6171 }, { "epoch": 1.43, "learning_rate": 2.6226793005161394e-05, "loss": 0.0, "step": 6172 }, { "epoch": 1.43, "learning_rate": 2.6222941221785692e-05, "loss": 0.0, "step": 6173 }, { "epoch": 1.43, "learning_rate": 2.6219089438409984e-05, "loss": 0.0, "step": 6174 }, { "epoch": 1.43, "learning_rate": 2.6215237655034282e-05, "loss": 0.0, "step": 6175 }, { "epoch": 1.43, "learning_rate": 2.6211385871658577e-05, "loss": 0.0, "step": 6176 }, { "epoch": 1.43, "learning_rate": 2.6207534088282875e-05, "loss": 0.0, "step": 6177 }, { "epoch": 1.43, "learning_rate": 2.6203682304907174e-05, "loss": 0.0, "step": 6178 }, { "epoch": 1.43, "learning_rate": 2.6199830521531472e-05, "loss": 0.0, "step": 6179 }, { "epoch": 1.43, "learning_rate": 2.6195978738155767e-05, "loss": 0.0001, "step": 6180 }, { "epoch": 1.43, "learning_rate": 2.6192126954780065e-05, "loss": 0.0, "step": 6181 }, { "epoch": 1.43, "learning_rate": 2.6188275171404364e-05, "loss": 0.0, "step": 6182 }, { "epoch": 1.43, "learning_rate": 2.6184423388028655e-05, "loss": 0.0, "step": 6183 }, { "epoch": 1.43, "learning_rate": 2.6180571604652954e-05, "loss": 0.0, "step": 6184 }, { "epoch": 1.43, "learning_rate": 2.6176719821277252e-05, "loss": 0.0, "step": 6185 }, { "epoch": 1.43, "learning_rate": 2.617286803790155e-05, "loss": 0.0, "step": 6186 }, { "epoch": 1.43, "learning_rate": 2.6169016254525845e-05, "loss": 0.0, "step": 6187 }, { "epoch": 1.43, "learning_rate": 2.6165164471150144e-05, "loss": 0.0, "step": 6188 }, { "epoch": 1.43, "learning_rate": 2.6161312687774442e-05, "loss": 0.0, "step": 6189 }, { "epoch": 1.43, "learning_rate": 2.615746090439874e-05, "loss": 0.0, "step": 6190 }, { "epoch": 1.43, "learning_rate": 2.615360912102304e-05, "loss": 0.0, "step": 6191 }, { "epoch": 1.43, "learning_rate": 2.614975733764733e-05, "loss": 0.0, "step": 6192 }, { "epoch": 1.43, "learning_rate": 2.614590555427163e-05, "loss": 0.0, "step": 6193 }, { "epoch": 1.43, "learning_rate": 2.6142053770895923e-05, "loss": 0.0, "step": 6194 }, { "epoch": 1.43, "learning_rate": 2.6138201987520222e-05, "loss": 0.0, "step": 6195 }, { "epoch": 1.43, "learning_rate": 2.613435020414452e-05, "loss": 0.0, "step": 6196 }, { "epoch": 1.43, "learning_rate": 2.613049842076882e-05, "loss": 0.0, "step": 6197 }, { "epoch": 1.43, "learning_rate": 2.6126646637393113e-05, "loss": 0.0, "step": 6198 }, { "epoch": 1.43, "learning_rate": 2.6122794854017412e-05, "loss": 0.0, "step": 6199 }, { "epoch": 1.43, "learning_rate": 2.611894307064171e-05, "loss": 0.0, "step": 6200 }, { "epoch": 1.43, "learning_rate": 2.611509128726601e-05, "loss": 0.0, "step": 6201 }, { "epoch": 1.43, "learning_rate": 2.61112395038903e-05, "loss": 0.0, "step": 6202 }, { "epoch": 1.43, "learning_rate": 2.61073877205146e-05, "loss": 0.0001, "step": 6203 }, { "epoch": 1.43, "learning_rate": 2.6103535937138897e-05, "loss": 0.0, "step": 6204 }, { "epoch": 1.43, "learning_rate": 2.609968415376319e-05, "loss": 0.1005, "step": 6205 }, { "epoch": 1.43, "learning_rate": 2.609583237038749e-05, "loss": 0.0, "step": 6206 }, { "epoch": 1.43, "learning_rate": 2.609198058701179e-05, "loss": 0.0, "step": 6207 }, { "epoch": 1.43, "learning_rate": 2.6088128803636087e-05, "loss": 0.0, "step": 6208 }, { "epoch": 1.43, "learning_rate": 2.6084277020260385e-05, "loss": 0.0001, "step": 6209 }, { "epoch": 1.44, "learning_rate": 2.608042523688468e-05, "loss": 0.0, "step": 6210 }, { "epoch": 1.44, "learning_rate": 2.6076573453508975e-05, "loss": 0.0, "step": 6211 }, { "epoch": 1.44, "learning_rate": 2.607272167013327e-05, "loss": 0.0, "step": 6212 }, { "epoch": 1.44, "learning_rate": 2.6068869886757568e-05, "loss": 0.0, "step": 6213 }, { "epoch": 1.44, "learning_rate": 2.6065018103381867e-05, "loss": 0.0, "step": 6214 }, { "epoch": 1.44, "learning_rate": 2.6061166320006165e-05, "loss": 0.0, "step": 6215 }, { "epoch": 1.44, "learning_rate": 2.6057314536630463e-05, "loss": 0.0001, "step": 6216 }, { "epoch": 1.44, "learning_rate": 2.6053462753254758e-05, "loss": 0.0, "step": 6217 }, { "epoch": 1.44, "learning_rate": 2.6049610969879057e-05, "loss": 0.0, "step": 6218 }, { "epoch": 1.44, "learning_rate": 2.6045759186503355e-05, "loss": 0.0, "step": 6219 }, { "epoch": 1.44, "learning_rate": 2.6041907403127647e-05, "loss": 0.0, "step": 6220 }, { "epoch": 1.44, "learning_rate": 2.6038055619751945e-05, "loss": 0.0, "step": 6221 }, { "epoch": 1.44, "learning_rate": 2.6034203836376243e-05, "loss": 0.0, "step": 6222 }, { "epoch": 1.44, "learning_rate": 2.6030352053000538e-05, "loss": 0.0, "step": 6223 }, { "epoch": 1.44, "learning_rate": 2.6026500269624837e-05, "loss": 0.0, "step": 6224 }, { "epoch": 1.44, "learning_rate": 2.6022648486249135e-05, "loss": 0.0, "step": 6225 }, { "epoch": 1.44, "learning_rate": 2.6018796702873433e-05, "loss": 0.0, "step": 6226 }, { "epoch": 1.44, "learning_rate": 2.601494491949773e-05, "loss": 0.0, "step": 6227 }, { "epoch": 1.44, "learning_rate": 2.6011093136122026e-05, "loss": 0.0, "step": 6228 }, { "epoch": 1.44, "learning_rate": 2.6007241352746325e-05, "loss": 0.0, "step": 6229 }, { "epoch": 1.44, "learning_rate": 2.6003389569370616e-05, "loss": 0.0, "step": 6230 }, { "epoch": 1.44, "learning_rate": 2.5999537785994915e-05, "loss": 0.0, "step": 6231 }, { "epoch": 1.44, "learning_rate": 2.5995686002619213e-05, "loss": 0.0, "step": 6232 }, { "epoch": 1.44, "learning_rate": 2.599183421924351e-05, "loss": 0.0002, "step": 6233 }, { "epoch": 1.44, "learning_rate": 2.598798243586781e-05, "loss": 0.0, "step": 6234 }, { "epoch": 1.44, "learning_rate": 2.5984130652492105e-05, "loss": 0.0, "step": 6235 }, { "epoch": 1.44, "learning_rate": 2.5980278869116403e-05, "loss": 0.0, "step": 6236 }, { "epoch": 1.44, "learning_rate": 2.59764270857407e-05, "loss": 0.0, "step": 6237 }, { "epoch": 1.44, "learning_rate": 2.5972575302365e-05, "loss": 0.0, "step": 6238 }, { "epoch": 1.44, "learning_rate": 2.596872351898929e-05, "loss": 0.0, "step": 6239 }, { "epoch": 1.44, "learning_rate": 2.596487173561359e-05, "loss": 0.0, "step": 6240 }, { "epoch": 1.44, "learning_rate": 2.5961019952237885e-05, "loss": 0.0, "step": 6241 }, { "epoch": 1.44, "learning_rate": 2.5957168168862183e-05, "loss": 0.0, "step": 6242 }, { "epoch": 1.44, "learning_rate": 2.595331638548648e-05, "loss": 0.0, "step": 6243 }, { "epoch": 1.44, "learning_rate": 2.594946460211078e-05, "loss": 0.0, "step": 6244 }, { "epoch": 1.44, "learning_rate": 2.5945612818735078e-05, "loss": 0.0, "step": 6245 }, { "epoch": 1.44, "learning_rate": 2.5941761035359373e-05, "loss": 0.0, "step": 6246 }, { "epoch": 1.44, "learning_rate": 2.593790925198367e-05, "loss": 0.0, "step": 6247 }, { "epoch": 1.44, "learning_rate": 2.5934057468607963e-05, "loss": 0.0, "step": 6248 }, { "epoch": 1.44, "learning_rate": 2.593020568523226e-05, "loss": 0.0001, "step": 6249 }, { "epoch": 1.44, "learning_rate": 2.592635390185656e-05, "loss": 0.0006, "step": 6250 }, { "epoch": 1.44, "learning_rate": 2.5922502118480858e-05, "loss": 0.0, "step": 6251 }, { "epoch": 1.44, "learning_rate": 2.5918650335105156e-05, "loss": 0.0, "step": 6252 }, { "epoch": 1.45, "learning_rate": 2.591479855172945e-05, "loss": 0.0, "step": 6253 }, { "epoch": 1.45, "learning_rate": 2.591094676835375e-05, "loss": 0.0, "step": 6254 }, { "epoch": 1.45, "learning_rate": 2.5907094984978048e-05, "loss": 0.0001, "step": 6255 }, { "epoch": 1.45, "learning_rate": 2.5903243201602346e-05, "loss": 0.0002, "step": 6256 }, { "epoch": 1.45, "learning_rate": 2.589939141822664e-05, "loss": 0.0, "step": 6257 }, { "epoch": 1.45, "learning_rate": 2.5895539634850936e-05, "loss": 0.0001, "step": 6258 }, { "epoch": 1.45, "learning_rate": 2.589168785147523e-05, "loss": 0.0, "step": 6259 }, { "epoch": 1.45, "learning_rate": 2.588783606809953e-05, "loss": 0.0, "step": 6260 }, { "epoch": 1.45, "learning_rate": 2.5883984284723828e-05, "loss": 0.0, "step": 6261 }, { "epoch": 1.45, "learning_rate": 2.5880132501348126e-05, "loss": 0.0, "step": 6262 }, { "epoch": 1.45, "learning_rate": 2.5876280717972424e-05, "loss": 0.0005, "step": 6263 }, { "epoch": 1.45, "learning_rate": 2.587242893459672e-05, "loss": 0.0, "step": 6264 }, { "epoch": 1.45, "learning_rate": 2.5868577151221018e-05, "loss": 0.0, "step": 6265 }, { "epoch": 1.45, "learning_rate": 2.5864725367845316e-05, "loss": 0.0001, "step": 6266 }, { "epoch": 1.45, "learning_rate": 2.5860873584469608e-05, "loss": 0.0, "step": 6267 }, { "epoch": 1.45, "learning_rate": 2.5857021801093906e-05, "loss": 0.0, "step": 6268 }, { "epoch": 1.45, "learning_rate": 2.5853170017718204e-05, "loss": 0.0, "step": 6269 }, { "epoch": 1.45, "learning_rate": 2.5849318234342503e-05, "loss": 0.0, "step": 6270 }, { "epoch": 1.45, "learning_rate": 2.5845466450966798e-05, "loss": 0.0, "step": 6271 }, { "epoch": 1.45, "learning_rate": 2.5841614667591096e-05, "loss": 0.0, "step": 6272 }, { "epoch": 1.45, "learning_rate": 2.5837762884215394e-05, "loss": 0.0001, "step": 6273 }, { "epoch": 1.45, "learning_rate": 2.5833911100839693e-05, "loss": 0.0, "step": 6274 }, { "epoch": 1.45, "learning_rate": 2.5830059317463988e-05, "loss": 0.0, "step": 6275 }, { "epoch": 1.45, "learning_rate": 2.5826207534088283e-05, "loss": 0.0, "step": 6276 }, { "epoch": 1.45, "learning_rate": 2.5822355750712578e-05, "loss": 0.0, "step": 6277 }, { "epoch": 1.45, "learning_rate": 2.5818503967336876e-05, "loss": 0.0, "step": 6278 }, { "epoch": 1.45, "learning_rate": 2.5814652183961174e-05, "loss": 0.0, "step": 6279 }, { "epoch": 1.45, "learning_rate": 2.5810800400585473e-05, "loss": 0.0, "step": 6280 }, { "epoch": 1.45, "learning_rate": 2.580694861720977e-05, "loss": 0.0, "step": 6281 }, { "epoch": 1.45, "learning_rate": 2.5803096833834066e-05, "loss": 0.0, "step": 6282 }, { "epoch": 1.45, "learning_rate": 2.5799245050458364e-05, "loss": 0.0, "step": 6283 }, { "epoch": 1.45, "learning_rate": 2.5795393267082663e-05, "loss": 0.0001, "step": 6284 }, { "epoch": 1.45, "learning_rate": 2.579154148370696e-05, "loss": 0.0, "step": 6285 }, { "epoch": 1.45, "learning_rate": 2.5787689700331252e-05, "loss": 0.0001, "step": 6286 }, { "epoch": 1.45, "learning_rate": 2.578383791695555e-05, "loss": 0.0, "step": 6287 }, { "epoch": 1.45, "learning_rate": 2.577998613357985e-05, "loss": 0.0, "step": 6288 }, { "epoch": 1.45, "learning_rate": 2.5776134350204144e-05, "loss": 0.0, "step": 6289 }, { "epoch": 1.45, "learning_rate": 2.5772282566828442e-05, "loss": 0.0, "step": 6290 }, { "epoch": 1.45, "learning_rate": 2.576843078345274e-05, "loss": 0.0, "step": 6291 }, { "epoch": 1.45, "learning_rate": 2.576457900007704e-05, "loss": 0.0004, "step": 6292 }, { "epoch": 1.45, "learning_rate": 2.5760727216701338e-05, "loss": 0.0001, "step": 6293 }, { "epoch": 1.45, "learning_rate": 2.5756875433325632e-05, "loss": 0.0, "step": 6294 }, { "epoch": 1.45, "learning_rate": 2.5753023649949924e-05, "loss": 0.0, "step": 6295 }, { "epoch": 1.46, "learning_rate": 2.5749171866574222e-05, "loss": 0.0, "step": 6296 }, { "epoch": 1.46, "learning_rate": 2.574532008319852e-05, "loss": 0.0, "step": 6297 }, { "epoch": 1.46, "learning_rate": 2.574146829982282e-05, "loss": 0.0, "step": 6298 }, { "epoch": 1.46, "learning_rate": 2.5737616516447117e-05, "loss": 0.0, "step": 6299 }, { "epoch": 1.46, "learning_rate": 2.5733764733071412e-05, "loss": 0.0, "step": 6300 }, { "epoch": 1.46, "learning_rate": 2.572991294969571e-05, "loss": 0.0001, "step": 6301 }, { "epoch": 1.46, "learning_rate": 2.572606116632001e-05, "loss": 0.0, "step": 6302 }, { "epoch": 1.46, "learning_rate": 2.5722209382944307e-05, "loss": 0.0, "step": 6303 }, { "epoch": 1.46, "learning_rate": 2.57183575995686e-05, "loss": 0.0, "step": 6304 }, { "epoch": 1.46, "learning_rate": 2.5714505816192897e-05, "loss": 0.0, "step": 6305 }, { "epoch": 1.46, "learning_rate": 2.5710654032817196e-05, "loss": 0.0, "step": 6306 }, { "epoch": 1.46, "learning_rate": 2.570680224944149e-05, "loss": 0.0, "step": 6307 }, { "epoch": 1.46, "learning_rate": 2.570295046606579e-05, "loss": 0.0, "step": 6308 }, { "epoch": 1.46, "learning_rate": 2.5699098682690087e-05, "loss": 0.0, "step": 6309 }, { "epoch": 1.46, "learning_rate": 2.5695246899314386e-05, "loss": 0.0, "step": 6310 }, { "epoch": 1.46, "learning_rate": 2.5691395115938684e-05, "loss": 0.0, "step": 6311 }, { "epoch": 1.46, "learning_rate": 2.568754333256298e-05, "loss": 0.0, "step": 6312 }, { "epoch": 1.46, "learning_rate": 2.5683691549187277e-05, "loss": 0.0, "step": 6313 }, { "epoch": 1.46, "learning_rate": 2.567983976581157e-05, "loss": 0.0, "step": 6314 }, { "epoch": 1.46, "learning_rate": 2.5675987982435867e-05, "loss": 0.0, "step": 6315 }, { "epoch": 1.46, "learning_rate": 2.5672136199060166e-05, "loss": 0.0, "step": 6316 }, { "epoch": 1.46, "learning_rate": 2.5668284415684464e-05, "loss": 0.0, "step": 6317 }, { "epoch": 1.46, "learning_rate": 2.566443263230876e-05, "loss": 0.0086, "step": 6318 }, { "epoch": 1.46, "learning_rate": 2.5660580848933057e-05, "loss": 0.0, "step": 6319 }, { "epoch": 1.46, "learning_rate": 2.5656729065557356e-05, "loss": 0.0, "step": 6320 }, { "epoch": 1.46, "learning_rate": 2.5652877282181654e-05, "loss": 0.0, "step": 6321 }, { "epoch": 1.46, "learning_rate": 2.5649025498805952e-05, "loss": 0.0, "step": 6322 }, { "epoch": 1.46, "learning_rate": 2.5645173715430244e-05, "loss": 0.0, "step": 6323 }, { "epoch": 1.46, "learning_rate": 2.5641321932054542e-05, "loss": 0.0, "step": 6324 }, { "epoch": 1.46, "learning_rate": 2.5637470148678837e-05, "loss": 0.0, "step": 6325 }, { "epoch": 1.46, "learning_rate": 2.5633618365303135e-05, "loss": 0.0, "step": 6326 }, { "epoch": 1.46, "learning_rate": 2.5629766581927434e-05, "loss": 0.0, "step": 6327 }, { "epoch": 1.46, "learning_rate": 2.5625914798551732e-05, "loss": 0.0, "step": 6328 }, { "epoch": 1.46, "learning_rate": 2.562206301517603e-05, "loss": 0.0, "step": 6329 }, { "epoch": 1.46, "learning_rate": 2.5618211231800325e-05, "loss": 0.0, "step": 6330 }, { "epoch": 1.46, "learning_rate": 2.5614359448424624e-05, "loss": 0.0, "step": 6331 }, { "epoch": 1.46, "learning_rate": 2.5610507665048915e-05, "loss": 0.0, "step": 6332 }, { "epoch": 1.46, "learning_rate": 2.5606655881673214e-05, "loss": 0.0, "step": 6333 }, { "epoch": 1.46, "learning_rate": 2.5602804098297512e-05, "loss": 0.0, "step": 6334 }, { "epoch": 1.46, "learning_rate": 2.559895231492181e-05, "loss": 0.0, "step": 6335 }, { "epoch": 1.46, "learning_rate": 2.5595100531546105e-05, "loss": 0.0, "step": 6336 }, { "epoch": 1.46, "learning_rate": 2.5591248748170404e-05, "loss": 0.0, "step": 6337 }, { "epoch": 1.46, "learning_rate": 2.5587396964794702e-05, "loss": 0.0, "step": 6338 }, { "epoch": 1.46, "learning_rate": 2.5583545181419e-05, "loss": 0.0, "step": 6339 }, { "epoch": 1.47, "learning_rate": 2.55796933980433e-05, "loss": 0.0, "step": 6340 }, { "epoch": 1.47, "learning_rate": 2.5575841614667594e-05, "loss": 0.0, "step": 6341 }, { "epoch": 1.47, "learning_rate": 2.557198983129189e-05, "loss": 0.0, "step": 6342 }, { "epoch": 1.47, "learning_rate": 2.5568138047916184e-05, "loss": 0.0, "step": 6343 }, { "epoch": 1.47, "learning_rate": 2.5564286264540482e-05, "loss": 0.0, "step": 6344 }, { "epoch": 1.47, "learning_rate": 2.556043448116478e-05, "loss": 0.0, "step": 6345 }, { "epoch": 1.47, "learning_rate": 2.555658269778908e-05, "loss": 0.0, "step": 6346 }, { "epoch": 1.47, "learning_rate": 2.5552730914413377e-05, "loss": 0.0, "step": 6347 }, { "epoch": 1.47, "learning_rate": 2.5548879131037672e-05, "loss": 0.0, "step": 6348 }, { "epoch": 1.47, "learning_rate": 2.554502734766197e-05, "loss": 0.0001, "step": 6349 }, { "epoch": 1.47, "learning_rate": 2.554117556428627e-05, "loss": 0.0, "step": 6350 }, { "epoch": 1.47, "learning_rate": 2.553732378091056e-05, "loss": 0.0004, "step": 6351 }, { "epoch": 1.47, "learning_rate": 2.553347199753486e-05, "loss": 0.0, "step": 6352 }, { "epoch": 1.47, "learning_rate": 2.5529620214159157e-05, "loss": 0.0, "step": 6353 }, { "epoch": 1.47, "learning_rate": 2.5525768430783452e-05, "loss": 0.0, "step": 6354 }, { "epoch": 1.47, "learning_rate": 2.552191664740775e-05, "loss": 0.0, "step": 6355 }, { "epoch": 1.47, "learning_rate": 2.551806486403205e-05, "loss": 0.0, "step": 6356 }, { "epoch": 1.47, "learning_rate": 2.5514213080656347e-05, "loss": 0.0, "step": 6357 }, { "epoch": 1.47, "learning_rate": 2.5510361297280645e-05, "loss": 0.0, "step": 6358 }, { "epoch": 1.47, "learning_rate": 2.550650951390494e-05, "loss": 0.0, "step": 6359 }, { "epoch": 1.47, "learning_rate": 2.5502657730529235e-05, "loss": 0.0, "step": 6360 }, { "epoch": 1.47, "learning_rate": 2.549880594715353e-05, "loss": 0.0, "step": 6361 }, { "epoch": 1.47, "learning_rate": 2.549495416377783e-05, "loss": 0.0, "step": 6362 }, { "epoch": 1.47, "learning_rate": 2.5491102380402127e-05, "loss": 0.0, "step": 6363 }, { "epoch": 1.47, "learning_rate": 2.5487250597026425e-05, "loss": 0.0, "step": 6364 }, { "epoch": 1.47, "learning_rate": 2.5483398813650723e-05, "loss": 0.0, "step": 6365 }, { "epoch": 1.47, "learning_rate": 2.547954703027502e-05, "loss": 0.0, "step": 6366 }, { "epoch": 1.47, "learning_rate": 2.5475695246899317e-05, "loss": 0.0, "step": 6367 }, { "epoch": 1.47, "learning_rate": 2.5471843463523615e-05, "loss": 0.0, "step": 6368 }, { "epoch": 1.47, "learning_rate": 2.5467991680147913e-05, "loss": 0.0, "step": 6369 }, { "epoch": 1.47, "learning_rate": 2.5464139896772205e-05, "loss": 0.0001, "step": 6370 }, { "epoch": 1.47, "learning_rate": 2.5460288113396503e-05, "loss": 0.0005, "step": 6371 }, { "epoch": 1.47, "learning_rate": 2.5456436330020798e-05, "loss": 0.0, "step": 6372 }, { "epoch": 1.47, "learning_rate": 2.5452584546645097e-05, "loss": 0.0, "step": 6373 }, { "epoch": 1.47, "learning_rate": 2.5448732763269395e-05, "loss": 0.0, "step": 6374 }, { "epoch": 1.47, "learning_rate": 2.5444880979893693e-05, "loss": 0.0, "step": 6375 }, { "epoch": 1.47, "learning_rate": 2.544102919651799e-05, "loss": 0.0, "step": 6376 }, { "epoch": 1.47, "learning_rate": 2.5437177413142287e-05, "loss": 0.0, "step": 6377 }, { "epoch": 1.47, "learning_rate": 2.5433325629766585e-05, "loss": 0.0, "step": 6378 }, { "epoch": 1.47, "learning_rate": 2.5429473846390876e-05, "loss": 0.0, "step": 6379 }, { "epoch": 1.47, "learning_rate": 2.5425622063015175e-05, "loss": 0.0, "step": 6380 }, { "epoch": 1.47, "learning_rate": 2.5421770279639473e-05, "loss": 0.0, "step": 6381 }, { "epoch": 1.47, "learning_rate": 2.541791849626377e-05, "loss": 0.0, "step": 6382 }, { "epoch": 1.48, "learning_rate": 2.541406671288807e-05, "loss": 0.0, "step": 6383 }, { "epoch": 1.48, "learning_rate": 2.5410214929512365e-05, "loss": 0.0, "step": 6384 }, { "epoch": 1.48, "learning_rate": 2.5406363146136663e-05, "loss": 0.0, "step": 6385 }, { "epoch": 1.48, "learning_rate": 2.540251136276096e-05, "loss": 0.0, "step": 6386 }, { "epoch": 1.48, "learning_rate": 2.539865957938526e-05, "loss": 0.0, "step": 6387 }, { "epoch": 1.48, "learning_rate": 2.539480779600955e-05, "loss": 0.0, "step": 6388 }, { "epoch": 1.48, "learning_rate": 2.539095601263385e-05, "loss": 0.0, "step": 6389 }, { "epoch": 1.48, "learning_rate": 2.5387104229258145e-05, "loss": 0.0, "step": 6390 }, { "epoch": 1.48, "learning_rate": 2.5383252445882443e-05, "loss": 0.0, "step": 6391 }, { "epoch": 1.48, "learning_rate": 2.537940066250674e-05, "loss": 0.0001, "step": 6392 }, { "epoch": 1.48, "learning_rate": 2.537554887913104e-05, "loss": 0.0, "step": 6393 }, { "epoch": 1.48, "learning_rate": 2.5371697095755338e-05, "loss": 0.0, "step": 6394 }, { "epoch": 1.48, "learning_rate": 2.5367845312379633e-05, "loss": 0.0, "step": 6395 }, { "epoch": 1.48, "learning_rate": 2.536399352900393e-05, "loss": 0.0, "step": 6396 }, { "epoch": 1.48, "learning_rate": 2.536014174562823e-05, "loss": 0.0, "step": 6397 }, { "epoch": 1.48, "learning_rate": 2.535628996225252e-05, "loss": 0.0016, "step": 6398 }, { "epoch": 1.48, "learning_rate": 2.535243817887682e-05, "loss": 0.0, "step": 6399 }, { "epoch": 1.48, "learning_rate": 2.5348586395501118e-05, "loss": 0.0, "step": 6400 }, { "epoch": 1.48, "learning_rate": 2.5344734612125416e-05, "loss": 0.0, "step": 6401 }, { "epoch": 1.48, "learning_rate": 2.534088282874971e-05, "loss": 0.0, "step": 6402 }, { "epoch": 1.48, "learning_rate": 2.533703104537401e-05, "loss": 0.0, "step": 6403 }, { "epoch": 1.48, "learning_rate": 2.5333179261998308e-05, "loss": 0.0, "step": 6404 }, { "epoch": 1.48, "learning_rate": 2.5329327478622606e-05, "loss": 0.0, "step": 6405 }, { "epoch": 1.48, "learning_rate": 2.5325475695246905e-05, "loss": 0.0, "step": 6406 }, { "epoch": 1.48, "learning_rate": 2.5321623911871196e-05, "loss": 0.0, "step": 6407 }, { "epoch": 1.48, "learning_rate": 2.531777212849549e-05, "loss": 0.0, "step": 6408 }, { "epoch": 1.48, "learning_rate": 2.531392034511979e-05, "loss": 0.0, "step": 6409 }, { "epoch": 1.48, "learning_rate": 2.5310068561744088e-05, "loss": 0.0, "step": 6410 }, { "epoch": 1.48, "learning_rate": 2.5306216778368386e-05, "loss": 0.0, "step": 6411 }, { "epoch": 1.48, "learning_rate": 2.5302364994992685e-05, "loss": 0.0, "step": 6412 }, { "epoch": 1.48, "learning_rate": 2.529851321161698e-05, "loss": 0.0, "step": 6413 }, { "epoch": 1.48, "learning_rate": 2.5294661428241278e-05, "loss": 0.0001, "step": 6414 }, { "epoch": 1.48, "learning_rate": 2.5290809644865576e-05, "loss": 0.0, "step": 6415 }, { "epoch": 1.48, "learning_rate": 2.5286957861489868e-05, "loss": 0.0, "step": 6416 }, { "epoch": 1.48, "learning_rate": 2.5283106078114166e-05, "loss": 0.0, "step": 6417 }, { "epoch": 1.48, "learning_rate": 2.5279254294738464e-05, "loss": 0.0, "step": 6418 }, { "epoch": 1.48, "learning_rate": 2.5275402511362763e-05, "loss": 0.0, "step": 6419 }, { "epoch": 1.48, "learning_rate": 2.5271550727987058e-05, "loss": 0.0, "step": 6420 }, { "epoch": 1.48, "learning_rate": 2.5267698944611356e-05, "loss": 0.0, "step": 6421 }, { "epoch": 1.48, "learning_rate": 2.5263847161235654e-05, "loss": 0.0, "step": 6422 }, { "epoch": 1.48, "learning_rate": 2.5259995377859953e-05, "loss": 0.0, "step": 6423 }, { "epoch": 1.48, "learning_rate": 2.525614359448425e-05, "loss": 0.0, "step": 6424 }, { "epoch": 1.48, "learning_rate": 2.5252291811108546e-05, "loss": 0.0, "step": 6425 }, { "epoch": 1.49, "learning_rate": 2.5248440027732838e-05, "loss": 0.0, "step": 6426 }, { "epoch": 1.49, "learning_rate": 2.5244588244357136e-05, "loss": 0.0, "step": 6427 }, { "epoch": 1.49, "learning_rate": 2.5240736460981434e-05, "loss": 0.0, "step": 6428 }, { "epoch": 1.49, "learning_rate": 2.5236884677605733e-05, "loss": 0.0, "step": 6429 }, { "epoch": 1.49, "learning_rate": 2.523303289423003e-05, "loss": 0.0, "step": 6430 }, { "epoch": 1.49, "learning_rate": 2.5229181110854326e-05, "loss": 0.0, "step": 6431 }, { "epoch": 1.49, "learning_rate": 2.5225329327478624e-05, "loss": 0.0, "step": 6432 }, { "epoch": 1.49, "learning_rate": 2.5221477544102923e-05, "loss": 0.0, "step": 6433 }, { "epoch": 1.49, "learning_rate": 2.521762576072722e-05, "loss": 0.0, "step": 6434 }, { "epoch": 1.49, "learning_rate": 2.5213773977351513e-05, "loss": 0.0, "step": 6435 }, { "epoch": 1.49, "learning_rate": 2.520992219397581e-05, "loss": 0.0, "step": 6436 }, { "epoch": 1.49, "learning_rate": 2.520607041060011e-05, "loss": 0.0, "step": 6437 }, { "epoch": 1.49, "learning_rate": 2.5202218627224404e-05, "loss": 0.0, "step": 6438 }, { "epoch": 1.49, "learning_rate": 2.5198366843848703e-05, "loss": 0.0, "step": 6439 }, { "epoch": 1.49, "learning_rate": 2.5194515060473e-05, "loss": 0.0, "step": 6440 }, { "epoch": 1.49, "learning_rate": 2.51906632770973e-05, "loss": 0.0001, "step": 6441 }, { "epoch": 1.49, "learning_rate": 2.5186811493721598e-05, "loss": 0.0, "step": 6442 }, { "epoch": 1.49, "learning_rate": 2.5182959710345893e-05, "loss": 0.0, "step": 6443 }, { "epoch": 1.49, "learning_rate": 2.5179107926970184e-05, "loss": 0.0, "step": 6444 }, { "epoch": 1.49, "learning_rate": 2.5175256143594482e-05, "loss": 0.0, "step": 6445 }, { "epoch": 1.49, "learning_rate": 2.517140436021878e-05, "loss": 0.0, "step": 6446 }, { "epoch": 1.49, "learning_rate": 2.516755257684308e-05, "loss": 0.0, "step": 6447 }, { "epoch": 1.49, "learning_rate": 2.5163700793467377e-05, "loss": 0.0, "step": 6448 }, { "epoch": 1.49, "learning_rate": 2.5159849010091672e-05, "loss": 0.0, "step": 6449 }, { "epoch": 1.49, "learning_rate": 2.515599722671597e-05, "loss": 0.0001, "step": 6450 }, { "epoch": 1.49, "learning_rate": 2.515214544334027e-05, "loss": 0.0, "step": 6451 }, { "epoch": 1.49, "learning_rate": 2.5148293659964567e-05, "loss": 0.0, "step": 6452 }, { "epoch": 1.49, "learning_rate": 2.5144441876588866e-05, "loss": 0.0, "step": 6453 }, { "epoch": 1.49, "learning_rate": 2.5140590093213157e-05, "loss": 0.0, "step": 6454 }, { "epoch": 1.49, "learning_rate": 2.5136738309837456e-05, "loss": 0.0, "step": 6455 }, { "epoch": 1.49, "learning_rate": 2.513288652646175e-05, "loss": 0.0, "step": 6456 }, { "epoch": 1.49, "learning_rate": 2.512903474308605e-05, "loss": 0.0, "step": 6457 }, { "epoch": 1.49, "learning_rate": 2.5125182959710347e-05, "loss": 0.0, "step": 6458 }, { "epoch": 1.49, "learning_rate": 2.5121331176334646e-05, "loss": 0.0, "step": 6459 }, { "epoch": 1.49, "learning_rate": 2.5117479392958944e-05, "loss": 0.0, "step": 6460 }, { "epoch": 1.49, "learning_rate": 2.511362760958324e-05, "loss": 0.0, "step": 6461 }, { "epoch": 1.49, "learning_rate": 2.5109775826207537e-05, "loss": 0.0, "step": 6462 }, { "epoch": 1.49, "learning_rate": 2.510592404283183e-05, "loss": 0.0, "step": 6463 }, { "epoch": 1.49, "learning_rate": 2.5102072259456127e-05, "loss": 0.0, "step": 6464 }, { "epoch": 1.49, "learning_rate": 2.5098220476080426e-05, "loss": 0.0001, "step": 6465 }, { "epoch": 1.49, "learning_rate": 2.5094368692704724e-05, "loss": 0.0, "step": 6466 }, { "epoch": 1.49, "learning_rate": 2.509051690932902e-05, "loss": 0.0, "step": 6467 }, { "epoch": 1.49, "learning_rate": 2.5086665125953317e-05, "loss": 0.0, "step": 6468 }, { "epoch": 1.5, "learning_rate": 2.5082813342577616e-05, "loss": 0.0066, "step": 6469 }, { "epoch": 1.5, "learning_rate": 2.5078961559201914e-05, "loss": 0.0, "step": 6470 }, { "epoch": 1.5, "learning_rate": 2.5075109775826212e-05, "loss": 0.0, "step": 6471 }, { "epoch": 1.5, "learning_rate": 2.5071257992450504e-05, "loss": 0.0001, "step": 6472 }, { "epoch": 1.5, "learning_rate": 2.5067406209074802e-05, "loss": 0.0, "step": 6473 }, { "epoch": 1.5, "learning_rate": 2.5063554425699097e-05, "loss": 0.0002, "step": 6474 }, { "epoch": 1.5, "learning_rate": 2.5059702642323396e-05, "loss": 0.0, "step": 6475 }, { "epoch": 1.5, "learning_rate": 2.5055850858947694e-05, "loss": 0.0, "step": 6476 }, { "epoch": 1.5, "learning_rate": 2.5051999075571992e-05, "loss": 0.0, "step": 6477 }, { "epoch": 1.5, "learning_rate": 2.504814729219629e-05, "loss": 0.0001, "step": 6478 }, { "epoch": 1.5, "learning_rate": 2.5044295508820585e-05, "loss": 0.0, "step": 6479 }, { "epoch": 1.5, "learning_rate": 2.5040443725444884e-05, "loss": 0.0, "step": 6480 }, { "epoch": 1.5, "learning_rate": 2.5036591942069182e-05, "loss": 0.0, "step": 6481 }, { "epoch": 1.5, "learning_rate": 2.5032740158693474e-05, "loss": 0.0, "step": 6482 }, { "epoch": 1.5, "learning_rate": 2.5028888375317772e-05, "loss": 0.0, "step": 6483 }, { "epoch": 1.5, "learning_rate": 2.502503659194207e-05, "loss": 0.0, "step": 6484 }, { "epoch": 1.5, "learning_rate": 2.5021184808566365e-05, "loss": 0.0, "step": 6485 }, { "epoch": 1.5, "learning_rate": 2.5017333025190664e-05, "loss": 0.0, "step": 6486 }, { "epoch": 1.5, "learning_rate": 2.5013481241814962e-05, "loss": 0.0, "step": 6487 }, { "epoch": 1.5, "learning_rate": 2.500962945843926e-05, "loss": 0.0007, "step": 6488 }, { "epoch": 1.5, "learning_rate": 2.500577767506356e-05, "loss": 0.0, "step": 6489 }, { "epoch": 1.5, "learning_rate": 2.5001925891687854e-05, "loss": 0.0023, "step": 6490 }, { "epoch": 1.5, "learning_rate": 2.499807410831215e-05, "loss": 0.0, "step": 6491 }, { "epoch": 1.5, "learning_rate": 2.4994222324936447e-05, "loss": 0.0, "step": 6492 }, { "epoch": 1.5, "learning_rate": 2.4990370541560742e-05, "loss": 0.0001, "step": 6493 }, { "epoch": 1.5, "learning_rate": 2.498651875818504e-05, "loss": 0.0, "step": 6494 }, { "epoch": 1.5, "learning_rate": 2.498266697480934e-05, "loss": 0.0, "step": 6495 }, { "epoch": 1.5, "learning_rate": 2.4978815191433637e-05, "loss": 0.0, "step": 6496 }, { "epoch": 1.5, "learning_rate": 2.4974963408057932e-05, "loss": 0.0, "step": 6497 }, { "epoch": 1.5, "learning_rate": 2.4971111624682227e-05, "loss": 0.0, "step": 6498 }, { "epoch": 1.5, "learning_rate": 2.4967259841306525e-05, "loss": 0.0001, "step": 6499 }, { "epoch": 1.5, "learning_rate": 2.4963408057930824e-05, "loss": 0.0, "step": 6500 }, { "epoch": 1.5, "learning_rate": 2.4959556274555122e-05, "loss": 0.0, "step": 6501 }, { "epoch": 1.5, "learning_rate": 2.4955704491179417e-05, "loss": 0.0, "step": 6502 }, { "epoch": 1.5, "learning_rate": 2.4951852707803712e-05, "loss": 0.0, "step": 6503 }, { "epoch": 1.5, "learning_rate": 2.494800092442801e-05, "loss": 0.0, "step": 6504 }, { "epoch": 1.5, "learning_rate": 2.494414914105231e-05, "loss": 0.0, "step": 6505 }, { "epoch": 1.5, "learning_rate": 2.4940297357676607e-05, "loss": 0.0, "step": 6506 }, { "epoch": 1.5, "learning_rate": 2.4936445574300902e-05, "loss": 0.0, "step": 6507 }, { "epoch": 1.5, "learning_rate": 2.49325937909252e-05, "loss": 0.0, "step": 6508 }, { "epoch": 1.5, "learning_rate": 2.4928742007549495e-05, "loss": 0.0, "step": 6509 }, { "epoch": 1.5, "learning_rate": 2.4924890224173793e-05, "loss": 0.0, "step": 6510 }, { "epoch": 1.5, "learning_rate": 2.4921038440798092e-05, "loss": 0.0, "step": 6511 }, { "epoch": 1.5, "learning_rate": 2.4917186657422387e-05, "loss": 0.0, "step": 6512 }, { "epoch": 1.51, "learning_rate": 2.4913334874046685e-05, "loss": 0.0, "step": 6513 }, { "epoch": 1.51, "learning_rate": 2.4909483090670983e-05, "loss": 0.0, "step": 6514 }, { "epoch": 1.51, "learning_rate": 2.490563130729528e-05, "loss": 0.0, "step": 6515 }, { "epoch": 1.51, "learning_rate": 2.4901779523919577e-05, "loss": 0.0, "step": 6516 }, { "epoch": 1.51, "learning_rate": 2.4897927740543872e-05, "loss": 0.0, "step": 6517 }, { "epoch": 1.51, "learning_rate": 2.489407595716817e-05, "loss": 0.0, "step": 6518 }, { "epoch": 1.51, "learning_rate": 2.489022417379247e-05, "loss": 0.0, "step": 6519 }, { "epoch": 1.51, "learning_rate": 2.4886372390416763e-05, "loss": 0.0, "step": 6520 }, { "epoch": 1.51, "learning_rate": 2.488252060704106e-05, "loss": 0.0, "step": 6521 }, { "epoch": 1.51, "learning_rate": 2.4878668823665357e-05, "loss": 0.0, "step": 6522 }, { "epoch": 1.51, "learning_rate": 2.4874817040289655e-05, "loss": 0.0, "step": 6523 }, { "epoch": 1.51, "learning_rate": 2.4870965256913953e-05, "loss": 0.0, "step": 6524 }, { "epoch": 1.51, "learning_rate": 2.4867113473538252e-05, "loss": 0.0, "step": 6525 }, { "epoch": 1.51, "learning_rate": 2.4863261690162547e-05, "loss": 0.0, "step": 6526 }, { "epoch": 1.51, "learning_rate": 2.485940990678684e-05, "loss": 0.0, "step": 6527 }, { "epoch": 1.51, "learning_rate": 2.485555812341114e-05, "loss": 0.0, "step": 6528 }, { "epoch": 1.51, "learning_rate": 2.4851706340035438e-05, "loss": 0.0, "step": 6529 }, { "epoch": 1.51, "learning_rate": 2.4847854556659737e-05, "loss": 0.0, "step": 6530 }, { "epoch": 1.51, "learning_rate": 2.484400277328403e-05, "loss": 0.0, "step": 6531 }, { "epoch": 1.51, "learning_rate": 2.484015098990833e-05, "loss": 0.0, "step": 6532 }, { "epoch": 1.51, "learning_rate": 2.4836299206532625e-05, "loss": 0.0, "step": 6533 }, { "epoch": 1.51, "learning_rate": 2.4832447423156923e-05, "loss": 0.0, "step": 6534 }, { "epoch": 1.51, "learning_rate": 2.4828595639781218e-05, "loss": 0.0, "step": 6535 }, { "epoch": 1.51, "learning_rate": 2.4824743856405517e-05, "loss": 0.0, "step": 6536 }, { "epoch": 1.51, "learning_rate": 2.4820892073029815e-05, "loss": 0.0, "step": 6537 }, { "epoch": 1.51, "learning_rate": 2.4817040289654113e-05, "loss": 0.0, "step": 6538 }, { "epoch": 1.51, "learning_rate": 2.4813188506278408e-05, "loss": 0.0, "step": 6539 }, { "epoch": 1.51, "learning_rate": 2.4809336722902703e-05, "loss": 0.0, "step": 6540 }, { "epoch": 1.51, "learning_rate": 2.4805484939527e-05, "loss": 0.0, "step": 6541 }, { "epoch": 1.51, "learning_rate": 2.48016331561513e-05, "loss": 0.0, "step": 6542 }, { "epoch": 1.51, "learning_rate": 2.4797781372775598e-05, "loss": 0.0, "step": 6543 }, { "epoch": 1.51, "learning_rate": 2.4793929589399893e-05, "loss": 0.0, "step": 6544 }, { "epoch": 1.51, "learning_rate": 2.4790077806024188e-05, "loss": 0.0, "step": 6545 }, { "epoch": 1.51, "learning_rate": 2.4786226022648486e-05, "loss": 0.0, "step": 6546 }, { "epoch": 1.51, "learning_rate": 2.4782374239272785e-05, "loss": 0.0001, "step": 6547 }, { "epoch": 1.51, "learning_rate": 2.4778522455897083e-05, "loss": 0.0, "step": 6548 }, { "epoch": 1.51, "learning_rate": 2.4774670672521378e-05, "loss": 0.0, "step": 6549 }, { "epoch": 1.51, "learning_rate": 2.4770818889145676e-05, "loss": 0.0, "step": 6550 }, { "epoch": 1.51, "learning_rate": 2.476696710576997e-05, "loss": 0.0, "step": 6551 }, { "epoch": 1.51, "learning_rate": 2.476311532239427e-05, "loss": 0.0, "step": 6552 }, { "epoch": 1.51, "learning_rate": 2.4759263539018568e-05, "loss": 0.0, "step": 6553 }, { "epoch": 1.51, "learning_rate": 2.4755411755642863e-05, "loss": 0.0, "step": 6554 }, { "epoch": 1.51, "learning_rate": 2.475155997226716e-05, "loss": 0.0, "step": 6555 }, { "epoch": 1.52, "learning_rate": 2.474770818889146e-05, "loss": 0.0, "step": 6556 }, { "epoch": 1.52, "learning_rate": 2.4743856405515755e-05, "loss": 0.0, "step": 6557 }, { "epoch": 1.52, "learning_rate": 2.4740004622140053e-05, "loss": 0.0, "step": 6558 }, { "epoch": 1.52, "learning_rate": 2.4736152838764348e-05, "loss": 0.0, "step": 6559 }, { "epoch": 1.52, "learning_rate": 2.4732301055388646e-05, "loss": 0.0, "step": 6560 }, { "epoch": 1.52, "learning_rate": 2.4728449272012945e-05, "loss": 0.0, "step": 6561 }, { "epoch": 1.52, "learning_rate": 2.472459748863724e-05, "loss": 0.0, "step": 6562 }, { "epoch": 1.52, "learning_rate": 2.4720745705261535e-05, "loss": 0.0, "step": 6563 }, { "epoch": 1.52, "learning_rate": 2.4716893921885833e-05, "loss": 0.0, "step": 6564 }, { "epoch": 1.52, "learning_rate": 2.471304213851013e-05, "loss": 0.0, "step": 6565 }, { "epoch": 1.52, "learning_rate": 2.470919035513443e-05, "loss": 0.0, "step": 6566 }, { "epoch": 1.52, "learning_rate": 2.4705338571758728e-05, "loss": 0.0, "step": 6567 }, { "epoch": 1.52, "learning_rate": 2.4701486788383023e-05, "loss": 0.0, "step": 6568 }, { "epoch": 1.52, "learning_rate": 2.4697635005007318e-05, "loss": 0.0, "step": 6569 }, { "epoch": 1.52, "learning_rate": 2.4693783221631616e-05, "loss": 0.0003, "step": 6570 }, { "epoch": 1.52, "learning_rate": 2.4689931438255915e-05, "loss": 0.0, "step": 6571 }, { "epoch": 1.52, "learning_rate": 2.4686079654880213e-05, "loss": 0.0, "step": 6572 }, { "epoch": 1.52, "learning_rate": 2.4682227871504508e-05, "loss": 0.0, "step": 6573 }, { "epoch": 1.52, "learning_rate": 2.4678376088128806e-05, "loss": 0.0, "step": 6574 }, { "epoch": 1.52, "learning_rate": 2.46745243047531e-05, "loss": 0.0, "step": 6575 }, { "epoch": 1.52, "learning_rate": 2.46706725213774e-05, "loss": 0.0, "step": 6576 }, { "epoch": 1.52, "learning_rate": 2.4666820738001694e-05, "loss": 0.0, "step": 6577 }, { "epoch": 1.52, "learning_rate": 2.4662968954625993e-05, "loss": 0.0004, "step": 6578 }, { "epoch": 1.52, "learning_rate": 2.465911717125029e-05, "loss": 0.0, "step": 6579 }, { "epoch": 1.52, "learning_rate": 2.4655265387874586e-05, "loss": 0.0, "step": 6580 }, { "epoch": 1.52, "learning_rate": 2.4651413604498884e-05, "loss": 0.0, "step": 6581 }, { "epoch": 1.52, "learning_rate": 2.464756182112318e-05, "loss": 0.0, "step": 6582 }, { "epoch": 1.52, "learning_rate": 2.4643710037747478e-05, "loss": 0.0, "step": 6583 }, { "epoch": 1.52, "learning_rate": 2.4639858254371776e-05, "loss": 0.0, "step": 6584 }, { "epoch": 1.52, "learning_rate": 2.4636006470996074e-05, "loss": 0.0, "step": 6585 }, { "epoch": 1.52, "learning_rate": 2.463215468762037e-05, "loss": 0.0002, "step": 6586 }, { "epoch": 1.52, "learning_rate": 2.4628302904244664e-05, "loss": 0.0, "step": 6587 }, { "epoch": 1.52, "learning_rate": 2.4624451120868963e-05, "loss": 0.0, "step": 6588 }, { "epoch": 1.52, "learning_rate": 2.462059933749326e-05, "loss": 0.0, "step": 6589 }, { "epoch": 1.52, "learning_rate": 2.461674755411756e-05, "loss": 0.0, "step": 6590 }, { "epoch": 1.52, "learning_rate": 2.4612895770741854e-05, "loss": 0.0, "step": 6591 }, { "epoch": 1.52, "learning_rate": 2.4609043987366153e-05, "loss": 0.0, "step": 6592 }, { "epoch": 1.52, "learning_rate": 2.4605192203990448e-05, "loss": 0.0, "step": 6593 }, { "epoch": 1.52, "learning_rate": 2.4601340420614746e-05, "loss": 0.0005, "step": 6594 }, { "epoch": 1.52, "learning_rate": 2.4597488637239044e-05, "loss": 0.0002, "step": 6595 }, { "epoch": 1.52, "learning_rate": 2.459363685386334e-05, "loss": 0.0004, "step": 6596 }, { "epoch": 1.52, "learning_rate": 2.4589785070487638e-05, "loss": 0.0, "step": 6597 }, { "epoch": 1.52, "learning_rate": 2.4585933287111933e-05, "loss": 0.0, "step": 6598 }, { "epoch": 1.53, "learning_rate": 2.458208150373623e-05, "loss": 0.0, "step": 6599 }, { "epoch": 1.53, "learning_rate": 2.457822972036053e-05, "loss": 0.0, "step": 6600 }, { "epoch": 1.53, "learning_rate": 2.4574377936984824e-05, "loss": 0.0, "step": 6601 }, { "epoch": 1.53, "learning_rate": 2.4570526153609123e-05, "loss": 0.0, "step": 6602 }, { "epoch": 1.53, "learning_rate": 2.456667437023342e-05, "loss": 0.0, "step": 6603 }, { "epoch": 1.53, "learning_rate": 2.4562822586857716e-05, "loss": 0.0, "step": 6604 }, { "epoch": 1.53, "learning_rate": 2.455897080348201e-05, "loss": 0.0, "step": 6605 }, { "epoch": 1.53, "learning_rate": 2.455511902010631e-05, "loss": 0.0, "step": 6606 }, { "epoch": 1.53, "learning_rate": 2.4551267236730607e-05, "loss": 0.0, "step": 6607 }, { "epoch": 1.53, "learning_rate": 2.4547415453354906e-05, "loss": 0.0, "step": 6608 }, { "epoch": 1.53, "learning_rate": 2.4543563669979204e-05, "loss": 0.0, "step": 6609 }, { "epoch": 1.53, "learning_rate": 2.45397118866035e-05, "loss": 0.0001, "step": 6610 }, { "epoch": 1.53, "learning_rate": 2.4535860103227794e-05, "loss": 0.0, "step": 6611 }, { "epoch": 1.53, "learning_rate": 2.4532008319852092e-05, "loss": 0.0, "step": 6612 }, { "epoch": 1.53, "learning_rate": 2.452815653647639e-05, "loss": 0.0, "step": 6613 }, { "epoch": 1.53, "learning_rate": 2.452430475310069e-05, "loss": 0.0, "step": 6614 }, { "epoch": 1.53, "learning_rate": 2.4520452969724984e-05, "loss": 0.0, "step": 6615 }, { "epoch": 1.53, "learning_rate": 2.451660118634928e-05, "loss": 0.0005, "step": 6616 }, { "epoch": 1.53, "learning_rate": 2.4512749402973577e-05, "loss": 0.0, "step": 6617 }, { "epoch": 1.53, "learning_rate": 2.4508897619597876e-05, "loss": 0.0001, "step": 6618 }, { "epoch": 1.53, "learning_rate": 2.450504583622217e-05, "loss": 0.0, "step": 6619 }, { "epoch": 1.53, "learning_rate": 2.450119405284647e-05, "loss": 0.0, "step": 6620 }, { "epoch": 1.53, "learning_rate": 2.4497342269470767e-05, "loss": 0.0, "step": 6621 }, { "epoch": 1.53, "learning_rate": 2.4493490486095062e-05, "loss": 0.0004, "step": 6622 }, { "epoch": 1.53, "learning_rate": 2.448963870271936e-05, "loss": 0.0, "step": 6623 }, { "epoch": 1.53, "learning_rate": 2.4485786919343656e-05, "loss": 0.0, "step": 6624 }, { "epoch": 1.53, "learning_rate": 2.4481935135967954e-05, "loss": 0.0, "step": 6625 }, { "epoch": 1.53, "learning_rate": 2.4478083352592252e-05, "loss": 0.0, "step": 6626 }, { "epoch": 1.53, "learning_rate": 2.447423156921655e-05, "loss": 0.0, "step": 6627 }, { "epoch": 1.53, "learning_rate": 2.4470379785840846e-05, "loss": 0.0, "step": 6628 }, { "epoch": 1.53, "learning_rate": 2.446652800246514e-05, "loss": 0.0, "step": 6629 }, { "epoch": 1.53, "learning_rate": 2.446267621908944e-05, "loss": 0.0, "step": 6630 }, { "epoch": 1.53, "learning_rate": 2.4458824435713737e-05, "loss": 0.0, "step": 6631 }, { "epoch": 1.53, "learning_rate": 2.4454972652338036e-05, "loss": 0.0, "step": 6632 }, { "epoch": 1.53, "learning_rate": 2.445112086896233e-05, "loss": 0.0, "step": 6633 }, { "epoch": 1.53, "learning_rate": 2.4447269085586625e-05, "loss": 0.0, "step": 6634 }, { "epoch": 1.53, "learning_rate": 2.4443417302210924e-05, "loss": 0.0, "step": 6635 }, { "epoch": 1.53, "learning_rate": 2.4439565518835222e-05, "loss": 0.0, "step": 6636 }, { "epoch": 1.53, "learning_rate": 2.443571373545952e-05, "loss": 0.0, "step": 6637 }, { "epoch": 1.53, "learning_rate": 2.4431861952083815e-05, "loss": 0.0, "step": 6638 }, { "epoch": 1.53, "learning_rate": 2.4428010168708114e-05, "loss": 0.0, "step": 6639 }, { "epoch": 1.53, "learning_rate": 2.442415838533241e-05, "loss": 0.0001, "step": 6640 }, { "epoch": 1.53, "learning_rate": 2.4420306601956707e-05, "loss": 0.0, "step": 6641 }, { "epoch": 1.54, "learning_rate": 2.4416454818581005e-05, "loss": 0.0, "step": 6642 }, { "epoch": 1.54, "learning_rate": 2.44126030352053e-05, "loss": 0.0, "step": 6643 }, { "epoch": 1.54, "learning_rate": 2.44087512518296e-05, "loss": 0.0, "step": 6644 }, { "epoch": 1.54, "learning_rate": 2.4404899468453897e-05, "loss": 0.0, "step": 6645 }, { "epoch": 1.54, "learning_rate": 2.4401047685078192e-05, "loss": 0.0, "step": 6646 }, { "epoch": 1.54, "learning_rate": 2.4397195901702487e-05, "loss": 0.0, "step": 6647 }, { "epoch": 1.54, "learning_rate": 2.4393344118326785e-05, "loss": 0.0, "step": 6648 }, { "epoch": 1.54, "learning_rate": 2.4389492334951084e-05, "loss": 0.0, "step": 6649 }, { "epoch": 1.54, "learning_rate": 2.4385640551575382e-05, "loss": 0.0, "step": 6650 }, { "epoch": 1.54, "learning_rate": 2.438178876819968e-05, "loss": 0.0, "step": 6651 }, { "epoch": 1.54, "learning_rate": 2.4377936984823972e-05, "loss": 0.0, "step": 6652 }, { "epoch": 1.54, "learning_rate": 2.437408520144827e-05, "loss": 0.0, "step": 6653 }, { "epoch": 1.54, "learning_rate": 2.437023341807257e-05, "loss": 0.0, "step": 6654 }, { "epoch": 1.54, "learning_rate": 2.4366381634696867e-05, "loss": 0.0, "step": 6655 }, { "epoch": 1.54, "learning_rate": 2.4362529851321165e-05, "loss": 0.0, "step": 6656 }, { "epoch": 1.54, "learning_rate": 2.435867806794546e-05, "loss": 0.0, "step": 6657 }, { "epoch": 1.54, "learning_rate": 2.4354826284569755e-05, "loss": 0.0, "step": 6658 }, { "epoch": 1.54, "learning_rate": 2.4350974501194054e-05, "loss": 0.0, "step": 6659 }, { "epoch": 1.54, "learning_rate": 2.4347122717818352e-05, "loss": 0.0, "step": 6660 }, { "epoch": 1.54, "learning_rate": 2.4343270934442647e-05, "loss": 0.0, "step": 6661 }, { "epoch": 1.54, "learning_rate": 2.4339419151066945e-05, "loss": 0.0, "step": 6662 }, { "epoch": 1.54, "learning_rate": 2.4335567367691244e-05, "loss": 0.0, "step": 6663 }, { "epoch": 1.54, "learning_rate": 2.433171558431554e-05, "loss": 0.0, "step": 6664 }, { "epoch": 1.54, "learning_rate": 2.4327863800939837e-05, "loss": 0.0, "step": 6665 }, { "epoch": 1.54, "learning_rate": 2.4324012017564132e-05, "loss": 0.0, "step": 6666 }, { "epoch": 1.54, "learning_rate": 2.432016023418843e-05, "loss": 0.0, "step": 6667 }, { "epoch": 1.54, "learning_rate": 2.431630845081273e-05, "loss": 0.0, "step": 6668 }, { "epoch": 1.54, "learning_rate": 2.4312456667437027e-05, "loss": 0.0, "step": 6669 }, { "epoch": 1.54, "learning_rate": 2.4308604884061322e-05, "loss": 0.0028, "step": 6670 }, { "epoch": 1.54, "learning_rate": 2.4304753100685617e-05, "loss": 0.0, "step": 6671 }, { "epoch": 1.54, "learning_rate": 2.4300901317309915e-05, "loss": 0.0, "step": 6672 }, { "epoch": 1.54, "learning_rate": 2.4297049533934213e-05, "loss": 0.0, "step": 6673 }, { "epoch": 1.54, "learning_rate": 2.4293197750558512e-05, "loss": 0.0, "step": 6674 }, { "epoch": 1.54, "learning_rate": 2.4289345967182807e-05, "loss": 0.0, "step": 6675 }, { "epoch": 1.54, "learning_rate": 2.4285494183807102e-05, "loss": 0.0, "step": 6676 }, { "epoch": 1.54, "learning_rate": 2.42816424004314e-05, "loss": 0.0, "step": 6677 }, { "epoch": 1.54, "learning_rate": 2.42777906170557e-05, "loss": 0.0, "step": 6678 }, { "epoch": 1.54, "learning_rate": 2.4273938833679997e-05, "loss": 0.0, "step": 6679 }, { "epoch": 1.54, "learning_rate": 2.427008705030429e-05, "loss": 0.0, "step": 6680 }, { "epoch": 1.54, "learning_rate": 2.426623526692859e-05, "loss": 0.0, "step": 6681 }, { "epoch": 1.54, "learning_rate": 2.4262383483552885e-05, "loss": 0.0, "step": 6682 }, { "epoch": 1.54, "learning_rate": 2.4258531700177183e-05, "loss": 0.0, "step": 6683 }, { "epoch": 1.54, "learning_rate": 2.425467991680148e-05, "loss": 0.0, "step": 6684 }, { "epoch": 1.54, "learning_rate": 2.4250828133425777e-05, "loss": 0.0, "step": 6685 }, { "epoch": 1.55, "learning_rate": 2.4246976350050075e-05, "loss": 0.0, "step": 6686 }, { "epoch": 1.55, "learning_rate": 2.4243124566674373e-05, "loss": 0.0, "step": 6687 }, { "epoch": 1.55, "learning_rate": 2.4239272783298668e-05, "loss": 0.0, "step": 6688 }, { "epoch": 1.55, "learning_rate": 2.4235420999922963e-05, "loss": 0.0, "step": 6689 }, { "epoch": 1.55, "learning_rate": 2.423156921654726e-05, "loss": 0.0, "step": 6690 }, { "epoch": 1.55, "learning_rate": 2.422771743317156e-05, "loss": 0.0, "step": 6691 }, { "epoch": 1.55, "learning_rate": 2.4223865649795858e-05, "loss": 0.0, "step": 6692 }, { "epoch": 1.55, "learning_rate": 2.4220013866420153e-05, "loss": 0.0, "step": 6693 }, { "epoch": 1.55, "learning_rate": 2.4216162083044448e-05, "loss": 0.0, "step": 6694 }, { "epoch": 1.55, "learning_rate": 2.4212310299668747e-05, "loss": 0.0001, "step": 6695 }, { "epoch": 1.55, "learning_rate": 2.4208458516293045e-05, "loss": 0.0001, "step": 6696 }, { "epoch": 1.55, "learning_rate": 2.4204606732917343e-05, "loss": 0.0, "step": 6697 }, { "epoch": 1.55, "learning_rate": 2.420075494954164e-05, "loss": 0.0, "step": 6698 }, { "epoch": 1.55, "learning_rate": 2.4196903166165936e-05, "loss": 0.0, "step": 6699 }, { "epoch": 1.55, "learning_rate": 2.419305138279023e-05, "loss": 0.0, "step": 6700 }, { "epoch": 1.55, "learning_rate": 2.418919959941453e-05, "loss": 0.0, "step": 6701 }, { "epoch": 1.55, "learning_rate": 2.4185347816038828e-05, "loss": 0.0, "step": 6702 }, { "epoch": 1.55, "learning_rate": 2.4181496032663123e-05, "loss": 0.0, "step": 6703 }, { "epoch": 1.55, "learning_rate": 2.417764424928742e-05, "loss": 0.0, "step": 6704 }, { "epoch": 1.55, "learning_rate": 2.417379246591172e-05, "loss": 0.0, "step": 6705 }, { "epoch": 1.55, "learning_rate": 2.4169940682536015e-05, "loss": 0.0, "step": 6706 }, { "epoch": 1.55, "learning_rate": 2.4166088899160313e-05, "loss": 0.0002, "step": 6707 }, { "epoch": 1.55, "learning_rate": 2.4162237115784608e-05, "loss": 0.0, "step": 6708 }, { "epoch": 1.55, "learning_rate": 2.4158385332408906e-05, "loss": 0.0, "step": 6709 }, { "epoch": 1.55, "learning_rate": 2.4154533549033205e-05, "loss": 0.0, "step": 6710 }, { "epoch": 1.55, "learning_rate": 2.41506817656575e-05, "loss": 0.0, "step": 6711 }, { "epoch": 1.55, "learning_rate": 2.4146829982281798e-05, "loss": 0.0, "step": 6712 }, { "epoch": 1.55, "learning_rate": 2.4142978198906093e-05, "loss": 0.0, "step": 6713 }, { "epoch": 1.55, "learning_rate": 2.413912641553039e-05, "loss": 0.0, "step": 6714 }, { "epoch": 1.55, "learning_rate": 2.413527463215469e-05, "loss": 0.0, "step": 6715 }, { "epoch": 1.55, "learning_rate": 2.4131422848778988e-05, "loss": 0.0, "step": 6716 }, { "epoch": 1.55, "learning_rate": 2.4127571065403283e-05, "loss": 0.0001, "step": 6717 }, { "epoch": 1.55, "learning_rate": 2.4123719282027578e-05, "loss": 0.0, "step": 6718 }, { "epoch": 1.55, "learning_rate": 2.4119867498651876e-05, "loss": 0.0001, "step": 6719 }, { "epoch": 1.55, "learning_rate": 2.4116015715276175e-05, "loss": 0.0001, "step": 6720 }, { "epoch": 1.55, "learning_rate": 2.4112163931900473e-05, "loss": 0.0, "step": 6721 }, { "epoch": 1.55, "learning_rate": 2.4108312148524768e-05, "loss": 0.0, "step": 6722 }, { "epoch": 1.55, "learning_rate": 2.4104460365149066e-05, "loss": 0.0, "step": 6723 }, { "epoch": 1.55, "learning_rate": 2.410060858177336e-05, "loss": 0.0001, "step": 6724 }, { "epoch": 1.55, "learning_rate": 2.409675679839766e-05, "loss": 0.0001, "step": 6725 }, { "epoch": 1.55, "learning_rate": 2.4092905015021958e-05, "loss": 0.0, "step": 6726 }, { "epoch": 1.55, "learning_rate": 2.4089053231646253e-05, "loss": 0.0, "step": 6727 }, { "epoch": 1.55, "learning_rate": 2.408520144827055e-05, "loss": 0.0, "step": 6728 }, { "epoch": 1.56, "learning_rate": 2.4081349664894846e-05, "loss": 0.0, "step": 6729 }, { "epoch": 1.56, "learning_rate": 2.4077497881519144e-05, "loss": 0.0, "step": 6730 }, { "epoch": 1.56, "learning_rate": 2.407364609814344e-05, "loss": 0.0, "step": 6731 }, { "epoch": 1.56, "learning_rate": 2.4069794314767738e-05, "loss": 0.0, "step": 6732 }, { "epoch": 1.56, "learning_rate": 2.4065942531392036e-05, "loss": 0.0, "step": 6733 }, { "epoch": 1.56, "learning_rate": 2.4062090748016334e-05, "loss": 0.0, "step": 6734 }, { "epoch": 1.56, "learning_rate": 2.405823896464063e-05, "loss": 0.0, "step": 6735 }, { "epoch": 1.56, "learning_rate": 2.4054387181264924e-05, "loss": 0.0, "step": 6736 }, { "epoch": 1.56, "learning_rate": 2.4050535397889223e-05, "loss": 0.0, "step": 6737 }, { "epoch": 1.56, "learning_rate": 2.404668361451352e-05, "loss": 0.0, "step": 6738 }, { "epoch": 1.56, "learning_rate": 2.404283183113782e-05, "loss": 0.0012, "step": 6739 }, { "epoch": 1.56, "learning_rate": 2.4038980047762118e-05, "loss": 0.0, "step": 6740 }, { "epoch": 1.56, "learning_rate": 2.4035128264386413e-05, "loss": 0.0, "step": 6741 }, { "epoch": 1.56, "learning_rate": 2.4031276481010708e-05, "loss": 0.0, "step": 6742 }, { "epoch": 1.56, "learning_rate": 2.4027424697635006e-05, "loss": 0.0, "step": 6743 }, { "epoch": 1.56, "learning_rate": 2.4023572914259304e-05, "loss": 0.0, "step": 6744 }, { "epoch": 1.56, "learning_rate": 2.40197211308836e-05, "loss": 0.0, "step": 6745 }, { "epoch": 1.56, "learning_rate": 2.4015869347507898e-05, "loss": 0.0, "step": 6746 }, { "epoch": 1.56, "learning_rate": 2.4012017564132193e-05, "loss": 0.0, "step": 6747 }, { "epoch": 1.56, "learning_rate": 2.400816578075649e-05, "loss": 0.0, "step": 6748 }, { "epoch": 1.56, "learning_rate": 2.400431399738079e-05, "loss": 0.0, "step": 6749 }, { "epoch": 1.56, "learning_rate": 2.4000462214005084e-05, "loss": 0.0, "step": 6750 }, { "epoch": 1.56, "learning_rate": 2.3996610430629383e-05, "loss": 0.0, "step": 6751 }, { "epoch": 1.56, "learning_rate": 2.399275864725368e-05, "loss": 0.0, "step": 6752 }, { "epoch": 1.56, "learning_rate": 2.3988906863877976e-05, "loss": 0.0, "step": 6753 }, { "epoch": 1.56, "learning_rate": 2.3985055080502274e-05, "loss": 0.0, "step": 6754 }, { "epoch": 1.56, "learning_rate": 2.398120329712657e-05, "loss": 0.0, "step": 6755 }, { "epoch": 1.56, "learning_rate": 2.3977351513750868e-05, "loss": 0.0, "step": 6756 }, { "epoch": 1.56, "learning_rate": 2.3973499730375166e-05, "loss": 0.0, "step": 6757 }, { "epoch": 1.56, "learning_rate": 2.3969647946999464e-05, "loss": 0.0, "step": 6758 }, { "epoch": 1.56, "learning_rate": 2.396579616362376e-05, "loss": 0.0, "step": 6759 }, { "epoch": 1.56, "learning_rate": 2.3961944380248054e-05, "loss": 0.0, "step": 6760 }, { "epoch": 1.56, "learning_rate": 2.3958092596872352e-05, "loss": 0.0, "step": 6761 }, { "epoch": 1.56, "learning_rate": 2.395424081349665e-05, "loss": 0.0, "step": 6762 }, { "epoch": 1.56, "learning_rate": 2.395038903012095e-05, "loss": 0.0, "step": 6763 }, { "epoch": 1.56, "learning_rate": 2.3946537246745244e-05, "loss": 0.0, "step": 6764 }, { "epoch": 1.56, "learning_rate": 2.394268546336954e-05, "loss": 0.0, "step": 6765 }, { "epoch": 1.56, "learning_rate": 2.3938833679993837e-05, "loss": 0.0, "step": 6766 }, { "epoch": 1.56, "learning_rate": 2.3934981896618136e-05, "loss": 0.0, "step": 6767 }, { "epoch": 1.56, "learning_rate": 2.3931130113242434e-05, "loss": 0.0, "step": 6768 }, { "epoch": 1.56, "learning_rate": 2.392727832986673e-05, "loss": 0.0004, "step": 6769 }, { "epoch": 1.56, "learning_rate": 2.3923426546491027e-05, "loss": 0.0, "step": 6770 }, { "epoch": 1.56, "learning_rate": 2.3919574763115322e-05, "loss": 0.0, "step": 6771 }, { "epoch": 1.57, "learning_rate": 2.391572297973962e-05, "loss": 0.0, "step": 6772 }, { "epoch": 1.57, "learning_rate": 2.3911871196363916e-05, "loss": 0.0, "step": 6773 }, { "epoch": 1.57, "learning_rate": 2.3908019412988214e-05, "loss": 0.0001, "step": 6774 }, { "epoch": 1.57, "learning_rate": 2.3904167629612512e-05, "loss": 0.0, "step": 6775 }, { "epoch": 1.57, "learning_rate": 2.390031584623681e-05, "loss": 0.0, "step": 6776 }, { "epoch": 1.57, "learning_rate": 2.3896464062861106e-05, "loss": 0.0, "step": 6777 }, { "epoch": 1.57, "learning_rate": 2.38926122794854e-05, "loss": 0.0, "step": 6778 }, { "epoch": 1.57, "learning_rate": 2.38887604961097e-05, "loss": 0.0007, "step": 6779 }, { "epoch": 1.57, "learning_rate": 2.3884908712733997e-05, "loss": 0.0, "step": 6780 }, { "epoch": 1.57, "learning_rate": 2.3881056929358296e-05, "loss": 0.0, "step": 6781 }, { "epoch": 1.57, "learning_rate": 2.3877205145982594e-05, "loss": 0.0, "step": 6782 }, { "epoch": 1.57, "learning_rate": 2.387335336260689e-05, "loss": 0.0, "step": 6783 }, { "epoch": 1.57, "learning_rate": 2.3869501579231184e-05, "loss": 0.0, "step": 6784 }, { "epoch": 1.57, "learning_rate": 2.3865649795855482e-05, "loss": 0.0, "step": 6785 }, { "epoch": 1.57, "learning_rate": 2.386179801247978e-05, "loss": 0.0, "step": 6786 }, { "epoch": 1.57, "learning_rate": 2.3857946229104076e-05, "loss": 0.0, "step": 6787 }, { "epoch": 1.57, "learning_rate": 2.3854094445728374e-05, "loss": 0.0, "step": 6788 }, { "epoch": 1.57, "learning_rate": 2.385024266235267e-05, "loss": 0.0, "step": 6789 }, { "epoch": 1.57, "learning_rate": 2.3846390878976967e-05, "loss": 0.0, "step": 6790 }, { "epoch": 1.57, "learning_rate": 2.3842539095601266e-05, "loss": 0.0, "step": 6791 }, { "epoch": 1.57, "learning_rate": 2.383868731222556e-05, "loss": 0.0, "step": 6792 }, { "epoch": 1.57, "learning_rate": 2.383483552884986e-05, "loss": 0.0, "step": 6793 }, { "epoch": 1.57, "learning_rate": 2.3830983745474157e-05, "loss": 0.0, "step": 6794 }, { "epoch": 1.57, "learning_rate": 2.3827131962098452e-05, "loss": 0.0, "step": 6795 }, { "epoch": 1.57, "learning_rate": 2.382328017872275e-05, "loss": 0.0, "step": 6796 }, { "epoch": 1.57, "learning_rate": 2.3819428395347045e-05, "loss": 0.0, "step": 6797 }, { "epoch": 1.57, "learning_rate": 2.3815576611971344e-05, "loss": 0.0, "step": 6798 }, { "epoch": 1.57, "learning_rate": 2.3811724828595642e-05, "loss": 0.0, "step": 6799 }, { "epoch": 1.57, "learning_rate": 2.380787304521994e-05, "loss": 0.0, "step": 6800 }, { "epoch": 1.57, "learning_rate": 2.3804021261844235e-05, "loss": 0.0001, "step": 6801 }, { "epoch": 1.57, "learning_rate": 2.380016947846853e-05, "loss": 0.0, "step": 6802 }, { "epoch": 1.57, "learning_rate": 2.379631769509283e-05, "loss": 0.0, "step": 6803 }, { "epoch": 1.57, "learning_rate": 2.3792465911717127e-05, "loss": 0.0, "step": 6804 }, { "epoch": 1.57, "learning_rate": 2.3788614128341425e-05, "loss": 0.0, "step": 6805 }, { "epoch": 1.57, "learning_rate": 2.378476234496572e-05, "loss": 0.0, "step": 6806 }, { "epoch": 1.57, "learning_rate": 2.3780910561590015e-05, "loss": 0.0, "step": 6807 }, { "epoch": 1.57, "learning_rate": 2.3777058778214314e-05, "loss": 0.0, "step": 6808 }, { "epoch": 1.57, "learning_rate": 2.3773206994838612e-05, "loss": 0.0, "step": 6809 }, { "epoch": 1.57, "learning_rate": 2.376935521146291e-05, "loss": 0.0, "step": 6810 }, { "epoch": 1.57, "learning_rate": 2.3765503428087205e-05, "loss": 0.0, "step": 6811 }, { "epoch": 1.57, "learning_rate": 2.3761651644711504e-05, "loss": 0.0, "step": 6812 }, { "epoch": 1.57, "learning_rate": 2.37577998613358e-05, "loss": 0.0, "step": 6813 }, { "epoch": 1.57, "learning_rate": 2.3753948077960097e-05, "loss": 0.0, "step": 6814 }, { "epoch": 1.57, "learning_rate": 2.3750096294584392e-05, "loss": 0.0, "step": 6815 }, { "epoch": 1.58, "learning_rate": 2.374624451120869e-05, "loss": 0.0, "step": 6816 }, { "epoch": 1.58, "learning_rate": 2.374239272783299e-05, "loss": 0.0, "step": 6817 }, { "epoch": 1.58, "learning_rate": 2.3738540944457287e-05, "loss": 0.0, "step": 6818 }, { "epoch": 1.58, "learning_rate": 2.3734689161081582e-05, "loss": 0.0, "step": 6819 }, { "epoch": 1.58, "learning_rate": 2.3730837377705877e-05, "loss": 0.0002, "step": 6820 }, { "epoch": 1.58, "learning_rate": 2.3726985594330175e-05, "loss": 0.0, "step": 6821 }, { "epoch": 1.58, "learning_rate": 2.3723133810954474e-05, "loss": 0.0, "step": 6822 }, { "epoch": 1.58, "learning_rate": 2.3719282027578772e-05, "loss": 0.0, "step": 6823 }, { "epoch": 1.58, "learning_rate": 2.3715430244203067e-05, "loss": 0.0, "step": 6824 }, { "epoch": 1.58, "learning_rate": 2.3711578460827362e-05, "loss": 0.0002, "step": 6825 }, { "epoch": 1.58, "learning_rate": 2.370772667745166e-05, "loss": 0.0, "step": 6826 }, { "epoch": 1.58, "learning_rate": 2.370387489407596e-05, "loss": 0.0, "step": 6827 }, { "epoch": 1.58, "learning_rate": 2.3700023110700257e-05, "loss": 0.0, "step": 6828 }, { "epoch": 1.58, "learning_rate": 2.3696171327324552e-05, "loss": 0.0, "step": 6829 }, { "epoch": 1.58, "learning_rate": 2.369231954394885e-05, "loss": 0.0, "step": 6830 }, { "epoch": 1.58, "learning_rate": 2.3688467760573145e-05, "loss": 0.0, "step": 6831 }, { "epoch": 1.58, "learning_rate": 2.3684615977197443e-05, "loss": 0.0, "step": 6832 }, { "epoch": 1.58, "learning_rate": 2.3680764193821742e-05, "loss": 0.0, "step": 6833 }, { "epoch": 1.58, "learning_rate": 2.3676912410446037e-05, "loss": 0.0, "step": 6834 }, { "epoch": 1.58, "learning_rate": 2.3673060627070335e-05, "loss": 0.0, "step": 6835 }, { "epoch": 1.58, "learning_rate": 2.3669208843694633e-05, "loss": 0.0, "step": 6836 }, { "epoch": 1.58, "learning_rate": 2.366535706031893e-05, "loss": 0.0001, "step": 6837 }, { "epoch": 1.58, "learning_rate": 2.3661505276943227e-05, "loss": 0.0, "step": 6838 }, { "epoch": 1.58, "learning_rate": 2.365765349356752e-05, "loss": 0.0, "step": 6839 }, { "epoch": 1.58, "learning_rate": 2.365380171019182e-05, "loss": 0.0, "step": 6840 }, { "epoch": 1.58, "learning_rate": 2.364994992681612e-05, "loss": 0.0001, "step": 6841 }, { "epoch": 1.58, "learning_rate": 2.3646098143440413e-05, "loss": 0.0, "step": 6842 }, { "epoch": 1.58, "learning_rate": 2.3642246360064708e-05, "loss": 0.0, "step": 6843 }, { "epoch": 1.58, "learning_rate": 2.3638394576689007e-05, "loss": 0.0, "step": 6844 }, { "epoch": 1.58, "learning_rate": 2.3634542793313305e-05, "loss": 0.0, "step": 6845 }, { "epoch": 1.58, "learning_rate": 2.3630691009937603e-05, "loss": 0.0, "step": 6846 }, { "epoch": 1.58, "learning_rate": 2.36268392265619e-05, "loss": 0.0, "step": 6847 }, { "epoch": 1.58, "learning_rate": 2.3622987443186197e-05, "loss": 0.0, "step": 6848 }, { "epoch": 1.58, "learning_rate": 2.361913565981049e-05, "loss": 0.0002, "step": 6849 }, { "epoch": 1.58, "learning_rate": 2.361528387643479e-05, "loss": 0.0, "step": 6850 }, { "epoch": 1.58, "learning_rate": 2.3611432093059088e-05, "loss": 0.0, "step": 6851 }, { "epoch": 1.58, "learning_rate": 2.3607580309683387e-05, "loss": 0.0, "step": 6852 }, { "epoch": 1.58, "learning_rate": 2.360372852630768e-05, "loss": 0.0, "step": 6853 }, { "epoch": 1.58, "learning_rate": 2.359987674293198e-05, "loss": 0.0, "step": 6854 }, { "epoch": 1.58, "learning_rate": 2.3596024959556275e-05, "loss": 0.0, "step": 6855 }, { "epoch": 1.58, "learning_rate": 2.3592173176180573e-05, "loss": 0.0, "step": 6856 }, { "epoch": 1.58, "learning_rate": 2.3588321392804868e-05, "loss": 0.0, "step": 6857 }, { "epoch": 1.58, "learning_rate": 2.3584469609429166e-05, "loss": 0.0, "step": 6858 }, { "epoch": 1.59, "learning_rate": 2.3580617826053465e-05, "loss": 0.0001, "step": 6859 }, { "epoch": 1.59, "learning_rate": 2.357676604267776e-05, "loss": 0.0, "step": 6860 }, { "epoch": 1.59, "learning_rate": 2.3572914259302058e-05, "loss": 0.0, "step": 6861 }, { "epoch": 1.59, "learning_rate": 2.3569062475926353e-05, "loss": 0.0, "step": 6862 }, { "epoch": 1.59, "learning_rate": 2.356521069255065e-05, "loss": 0.0, "step": 6863 }, { "epoch": 1.59, "learning_rate": 2.356135890917495e-05, "loss": 0.0, "step": 6864 }, { "epoch": 1.59, "learning_rate": 2.3557507125799248e-05, "loss": 0.0, "step": 6865 }, { "epoch": 1.59, "learning_rate": 2.3553655342423543e-05, "loss": 0.0, "step": 6866 }, { "epoch": 1.59, "learning_rate": 2.3549803559047838e-05, "loss": 0.0, "step": 6867 }, { "epoch": 1.59, "learning_rate": 2.3545951775672136e-05, "loss": 0.0, "step": 6868 }, { "epoch": 1.59, "learning_rate": 2.3542099992296435e-05, "loss": 0.0, "step": 6869 }, { "epoch": 1.59, "learning_rate": 2.3538248208920733e-05, "loss": 0.0, "step": 6870 }, { "epoch": 1.59, "learning_rate": 2.3534396425545028e-05, "loss": 0.0, "step": 6871 }, { "epoch": 1.59, "learning_rate": 2.3530544642169326e-05, "loss": 0.0001, "step": 6872 }, { "epoch": 1.59, "learning_rate": 2.352669285879362e-05, "loss": 0.0, "step": 6873 }, { "epoch": 1.59, "learning_rate": 2.352284107541792e-05, "loss": 0.0, "step": 6874 }, { "epoch": 1.59, "learning_rate": 2.3518989292042218e-05, "loss": 0.0, "step": 6875 }, { "epoch": 1.59, "learning_rate": 2.3515137508666513e-05, "loss": 0.0, "step": 6876 }, { "epoch": 1.59, "learning_rate": 2.351128572529081e-05, "loss": 0.0, "step": 6877 }, { "epoch": 1.59, "learning_rate": 2.350743394191511e-05, "loss": 0.0, "step": 6878 }, { "epoch": 1.59, "learning_rate": 2.3503582158539405e-05, "loss": 0.0, "step": 6879 }, { "epoch": 1.59, "learning_rate": 2.3499730375163703e-05, "loss": 0.0, "step": 6880 }, { "epoch": 1.59, "learning_rate": 2.3495878591787998e-05, "loss": 0.0, "step": 6881 }, { "epoch": 1.59, "learning_rate": 2.3492026808412296e-05, "loss": 0.0, "step": 6882 }, { "epoch": 1.59, "learning_rate": 2.3488175025036595e-05, "loss": 0.0, "step": 6883 }, { "epoch": 1.59, "learning_rate": 2.348432324166089e-05, "loss": 0.0002, "step": 6884 }, { "epoch": 1.59, "learning_rate": 2.3480471458285184e-05, "loss": 0.0, "step": 6885 }, { "epoch": 1.59, "learning_rate": 2.3476619674909483e-05, "loss": 0.0, "step": 6886 }, { "epoch": 1.59, "learning_rate": 2.347276789153378e-05, "loss": 0.0, "step": 6887 }, { "epoch": 1.59, "learning_rate": 2.346891610815808e-05, "loss": 0.0, "step": 6888 }, { "epoch": 1.59, "learning_rate": 2.3465064324782378e-05, "loss": 0.0, "step": 6889 }, { "epoch": 1.59, "learning_rate": 2.3461212541406673e-05, "loss": 0.0, "step": 6890 }, { "epoch": 1.59, "learning_rate": 2.3457360758030968e-05, "loss": 0.0, "step": 6891 }, { "epoch": 1.59, "learning_rate": 2.3453508974655266e-05, "loss": 0.0, "step": 6892 }, { "epoch": 1.59, "learning_rate": 2.3449657191279564e-05, "loss": 0.0005, "step": 6893 }, { "epoch": 1.59, "learning_rate": 2.3445805407903863e-05, "loss": 0.0, "step": 6894 }, { "epoch": 1.59, "learning_rate": 2.3441953624528158e-05, "loss": 0.0001, "step": 6895 }, { "epoch": 1.59, "learning_rate": 2.3438101841152456e-05, "loss": 0.0001, "step": 6896 }, { "epoch": 1.59, "learning_rate": 2.343425005777675e-05, "loss": 0.0, "step": 6897 }, { "epoch": 1.59, "learning_rate": 2.343039827440105e-05, "loss": 0.0, "step": 6898 }, { "epoch": 1.59, "learning_rate": 2.3426546491025344e-05, "loss": 0.0, "step": 6899 }, { "epoch": 1.59, "learning_rate": 2.3422694707649643e-05, "loss": 0.0, "step": 6900 }, { "epoch": 1.59, "learning_rate": 2.341884292427394e-05, "loss": 0.0, "step": 6901 }, { "epoch": 1.6, "learning_rate": 2.3414991140898236e-05, "loss": 0.0, "step": 6902 }, { "epoch": 1.6, "learning_rate": 2.3411139357522534e-05, "loss": 0.0, "step": 6903 }, { "epoch": 1.6, "learning_rate": 2.340728757414683e-05, "loss": 0.0, "step": 6904 }, { "epoch": 1.6, "learning_rate": 2.3403435790771128e-05, "loss": 0.0, "step": 6905 }, { "epoch": 1.6, "learning_rate": 2.3399584007395426e-05, "loss": 0.0, "step": 6906 }, { "epoch": 1.6, "learning_rate": 2.3395732224019724e-05, "loss": 0.0, "step": 6907 }, { "epoch": 1.6, "learning_rate": 2.339188044064402e-05, "loss": 0.0, "step": 6908 }, { "epoch": 1.6, "learning_rate": 2.3388028657268314e-05, "loss": 0.0, "step": 6909 }, { "epoch": 1.6, "learning_rate": 2.3384176873892613e-05, "loss": 0.0, "step": 6910 }, { "epoch": 1.6, "learning_rate": 2.338032509051691e-05, "loss": 0.0, "step": 6911 }, { "epoch": 1.6, "learning_rate": 2.337647330714121e-05, "loss": 0.0, "step": 6912 }, { "epoch": 1.6, "learning_rate": 2.3372621523765504e-05, "loss": 0.0, "step": 6913 }, { "epoch": 1.6, "learning_rate": 2.3368769740389803e-05, "loss": 0.0, "step": 6914 }, { "epoch": 1.6, "learning_rate": 2.3364917957014098e-05, "loss": 0.0, "step": 6915 }, { "epoch": 1.6, "learning_rate": 2.3361066173638396e-05, "loss": 0.0, "step": 6916 }, { "epoch": 1.6, "learning_rate": 2.3357214390262694e-05, "loss": 0.0, "step": 6917 }, { "epoch": 1.6, "learning_rate": 2.335336260688699e-05, "loss": 0.0, "step": 6918 }, { "epoch": 1.6, "learning_rate": 2.3349510823511288e-05, "loss": 0.0001, "step": 6919 }, { "epoch": 1.6, "learning_rate": 2.3345659040135582e-05, "loss": 0.0, "step": 6920 }, { "epoch": 1.6, "learning_rate": 2.334180725675988e-05, "loss": 0.0, "step": 6921 }, { "epoch": 1.6, "learning_rate": 2.333795547338418e-05, "loss": 0.0, "step": 6922 }, { "epoch": 1.6, "learning_rate": 2.3334103690008474e-05, "loss": 0.0, "step": 6923 }, { "epoch": 1.6, "learning_rate": 2.3330251906632772e-05, "loss": 0.0, "step": 6924 }, { "epoch": 1.6, "learning_rate": 2.332640012325707e-05, "loss": 0.0, "step": 6925 }, { "epoch": 1.6, "learning_rate": 2.3322548339881366e-05, "loss": 0.0, "step": 6926 }, { "epoch": 1.6, "learning_rate": 2.331869655650566e-05, "loss": 0.0, "step": 6927 }, { "epoch": 1.6, "learning_rate": 2.331484477312996e-05, "loss": 0.0, "step": 6928 }, { "epoch": 1.6, "learning_rate": 2.3310992989754257e-05, "loss": 0.0, "step": 6929 }, { "epoch": 1.6, "learning_rate": 2.3307141206378556e-05, "loss": 0.0, "step": 6930 }, { "epoch": 1.6, "learning_rate": 2.3303289423002854e-05, "loss": 0.0, "step": 6931 }, { "epoch": 1.6, "learning_rate": 2.329943763962715e-05, "loss": 0.0, "step": 6932 }, { "epoch": 1.6, "learning_rate": 2.3295585856251444e-05, "loss": 0.0, "step": 6933 }, { "epoch": 1.6, "learning_rate": 2.3291734072875742e-05, "loss": 0.0, "step": 6934 }, { "epoch": 1.6, "learning_rate": 2.328788228950004e-05, "loss": 0.0001, "step": 6935 }, { "epoch": 1.6, "learning_rate": 2.328403050612434e-05, "loss": 0.0, "step": 6936 }, { "epoch": 1.6, "learning_rate": 2.3280178722748634e-05, "loss": 0.0, "step": 6937 }, { "epoch": 1.6, "learning_rate": 2.327632693937293e-05, "loss": 0.0, "step": 6938 }, { "epoch": 1.6, "learning_rate": 2.3272475155997227e-05, "loss": 0.0, "step": 6939 }, { "epoch": 1.6, "learning_rate": 2.3268623372621526e-05, "loss": 0.0, "step": 6940 }, { "epoch": 1.6, "learning_rate": 2.326477158924582e-05, "loss": 0.0, "step": 6941 }, { "epoch": 1.6, "learning_rate": 2.326091980587012e-05, "loss": 0.0, "step": 6942 }, { "epoch": 1.6, "learning_rate": 2.3257068022494417e-05, "loss": 0.0, "step": 6943 }, { "epoch": 1.6, "learning_rate": 2.3253216239118712e-05, "loss": 0.0305, "step": 6944 }, { "epoch": 1.61, "learning_rate": 2.324936445574301e-05, "loss": 0.0, "step": 6945 }, { "epoch": 1.61, "learning_rate": 2.3245512672367306e-05, "loss": 0.0, "step": 6946 }, { "epoch": 1.61, "learning_rate": 2.3241660888991604e-05, "loss": 0.0, "step": 6947 }, { "epoch": 1.61, "learning_rate": 2.3237809105615902e-05, "loss": 0.0001, "step": 6948 }, { "epoch": 1.61, "learning_rate": 2.32339573222402e-05, "loss": 0.0, "step": 6949 }, { "epoch": 1.61, "learning_rate": 2.3230105538864496e-05, "loss": 0.0, "step": 6950 }, { "epoch": 1.61, "learning_rate": 2.322625375548879e-05, "loss": 0.0, "step": 6951 }, { "epoch": 1.61, "learning_rate": 2.322240197211309e-05, "loss": 0.0, "step": 6952 }, { "epoch": 1.61, "learning_rate": 2.3218550188737387e-05, "loss": 0.0, "step": 6953 }, { "epoch": 1.61, "learning_rate": 2.3214698405361685e-05, "loss": 0.0, "step": 6954 }, { "epoch": 1.61, "learning_rate": 2.321084662198598e-05, "loss": 0.0, "step": 6955 }, { "epoch": 1.61, "learning_rate": 2.3206994838610275e-05, "loss": 0.0, "step": 6956 }, { "epoch": 1.61, "learning_rate": 2.3203143055234574e-05, "loss": 0.0, "step": 6957 }, { "epoch": 1.61, "learning_rate": 2.3199291271858872e-05, "loss": 0.0, "step": 6958 }, { "epoch": 1.61, "learning_rate": 2.319543948848317e-05, "loss": 0.0, "step": 6959 }, { "epoch": 1.61, "learning_rate": 2.3191587705107465e-05, "loss": 0.0, "step": 6960 }, { "epoch": 1.61, "learning_rate": 2.3187735921731764e-05, "loss": 0.016, "step": 6961 }, { "epoch": 1.61, "learning_rate": 2.318388413835606e-05, "loss": 0.0, "step": 6962 }, { "epoch": 1.61, "learning_rate": 2.3180032354980357e-05, "loss": 0.0, "step": 6963 }, { "epoch": 1.61, "learning_rate": 2.3176180571604655e-05, "loss": 0.0, "step": 6964 }, { "epoch": 1.61, "learning_rate": 2.317232878822895e-05, "loss": 0.0, "step": 6965 }, { "epoch": 1.61, "learning_rate": 2.316847700485325e-05, "loss": 0.0, "step": 6966 }, { "epoch": 1.61, "learning_rate": 2.3164625221477547e-05, "loss": 0.0, "step": 6967 }, { "epoch": 1.61, "learning_rate": 2.3160773438101842e-05, "loss": 0.0, "step": 6968 }, { "epoch": 1.61, "learning_rate": 2.3156921654726137e-05, "loss": 0.0, "step": 6969 }, { "epoch": 1.61, "learning_rate": 2.3153069871350435e-05, "loss": 0.0, "step": 6970 }, { "epoch": 1.61, "learning_rate": 2.3149218087974734e-05, "loss": 0.0, "step": 6971 }, { "epoch": 1.61, "learning_rate": 2.3145366304599032e-05, "loss": 0.0, "step": 6972 }, { "epoch": 1.61, "learning_rate": 2.314151452122333e-05, "loss": 0.0, "step": 6973 }, { "epoch": 1.61, "learning_rate": 2.3137662737847622e-05, "loss": 0.0, "step": 6974 }, { "epoch": 1.61, "learning_rate": 2.313381095447192e-05, "loss": 0.0, "step": 6975 }, { "epoch": 1.61, "learning_rate": 2.312995917109622e-05, "loss": 0.0, "step": 6976 }, { "epoch": 1.61, "learning_rate": 2.3126107387720517e-05, "loss": 0.0, "step": 6977 }, { "epoch": 1.61, "learning_rate": 2.3122255604344815e-05, "loss": 0.0, "step": 6978 }, { "epoch": 1.61, "learning_rate": 2.311840382096911e-05, "loss": 0.0, "step": 6979 }, { "epoch": 1.61, "learning_rate": 2.3114552037593405e-05, "loss": 0.0, "step": 6980 }, { "epoch": 1.61, "learning_rate": 2.3110700254217703e-05, "loss": 0.0, "step": 6981 }, { "epoch": 1.61, "learning_rate": 2.3106848470842002e-05, "loss": 0.0, "step": 6982 }, { "epoch": 1.61, "learning_rate": 2.3102996687466297e-05, "loss": 0.0, "step": 6983 }, { "epoch": 1.61, "learning_rate": 2.3099144904090595e-05, "loss": 0.0, "step": 6984 }, { "epoch": 1.61, "learning_rate": 2.3095293120714893e-05, "loss": 0.0, "step": 6985 }, { "epoch": 1.61, "learning_rate": 2.309144133733919e-05, "loss": 0.0, "step": 6986 }, { "epoch": 1.61, "learning_rate": 2.3087589553963487e-05, "loss": 0.0, "step": 6987 }, { "epoch": 1.61, "learning_rate": 2.3083737770587782e-05, "loss": 0.0, "step": 6988 }, { "epoch": 1.62, "learning_rate": 2.307988598721208e-05, "loss": 0.0, "step": 6989 }, { "epoch": 1.62, "learning_rate": 2.307603420383638e-05, "loss": 0.0, "step": 6990 }, { "epoch": 1.62, "learning_rate": 2.3072182420460677e-05, "loss": 0.0, "step": 6991 }, { "epoch": 1.62, "learning_rate": 2.3068330637084972e-05, "loss": 0.0, "step": 6992 }, { "epoch": 1.62, "learning_rate": 2.3064478853709267e-05, "loss": 0.0, "step": 6993 }, { "epoch": 1.62, "learning_rate": 2.3060627070333565e-05, "loss": 0.0, "step": 6994 }, { "epoch": 1.62, "learning_rate": 2.3056775286957863e-05, "loss": 0.0, "step": 6995 }, { "epoch": 1.62, "learning_rate": 2.3052923503582162e-05, "loss": 0.0, "step": 6996 }, { "epoch": 1.62, "learning_rate": 2.3049071720206457e-05, "loss": 0.0, "step": 6997 }, { "epoch": 1.62, "learning_rate": 2.304521993683075e-05, "loss": 0.0, "step": 6998 }, { "epoch": 1.62, "learning_rate": 2.304136815345505e-05, "loss": 0.009, "step": 6999 }, { "epoch": 1.62, "learning_rate": 2.303751637007935e-05, "loss": 0.0, "step": 7000 }, { "epoch": 1.62, "learning_rate": 2.3033664586703647e-05, "loss": 0.0, "step": 7001 }, { "epoch": 1.62, "learning_rate": 2.302981280332794e-05, "loss": 0.0001, "step": 7002 }, { "epoch": 1.62, "learning_rate": 2.302596101995224e-05, "loss": 0.0, "step": 7003 }, { "epoch": 1.62, "learning_rate": 2.3022109236576535e-05, "loss": 0.0, "step": 7004 }, { "epoch": 1.62, "learning_rate": 2.3018257453200833e-05, "loss": 0.0, "step": 7005 }, { "epoch": 1.62, "learning_rate": 2.301440566982513e-05, "loss": 0.0001, "step": 7006 }, { "epoch": 1.62, "learning_rate": 2.3010553886449427e-05, "loss": 0.0001, "step": 7007 }, { "epoch": 1.62, "learning_rate": 2.3006702103073725e-05, "loss": 0.0, "step": 7008 }, { "epoch": 1.62, "learning_rate": 2.3002850319698023e-05, "loss": 0.0, "step": 7009 }, { "epoch": 1.62, "learning_rate": 2.2998998536322318e-05, "loss": 0.0, "step": 7010 }, { "epoch": 1.62, "learning_rate": 2.2995146752946613e-05, "loss": 0.0, "step": 7011 }, { "epoch": 1.62, "learning_rate": 2.299129496957091e-05, "loss": 0.0, "step": 7012 }, { "epoch": 1.62, "learning_rate": 2.298744318619521e-05, "loss": 0.0, "step": 7013 }, { "epoch": 1.62, "learning_rate": 2.2983591402819508e-05, "loss": 0.0, "step": 7014 }, { "epoch": 1.62, "learning_rate": 2.2979739619443803e-05, "loss": 0.0, "step": 7015 }, { "epoch": 1.62, "learning_rate": 2.2975887836068098e-05, "loss": 0.0, "step": 7016 }, { "epoch": 1.62, "learning_rate": 2.2972036052692396e-05, "loss": 0.0, "step": 7017 }, { "epoch": 1.62, "learning_rate": 2.2968184269316695e-05, "loss": 0.0005, "step": 7018 }, { "epoch": 1.62, "learning_rate": 2.2964332485940993e-05, "loss": 0.0, "step": 7019 }, { "epoch": 1.62, "learning_rate": 2.296048070256529e-05, "loss": 0.0, "step": 7020 }, { "epoch": 1.62, "learning_rate": 2.2956628919189586e-05, "loss": 0.0, "step": 7021 }, { "epoch": 1.62, "learning_rate": 2.295277713581388e-05, "loss": 0.0, "step": 7022 }, { "epoch": 1.62, "learning_rate": 2.294892535243818e-05, "loss": 0.0, "step": 7023 }, { "epoch": 1.62, "learning_rate": 2.2945073569062478e-05, "loss": 0.0, "step": 7024 }, { "epoch": 1.62, "learning_rate": 2.2941221785686773e-05, "loss": 0.0, "step": 7025 }, { "epoch": 1.62, "learning_rate": 2.293737000231107e-05, "loss": 0.0, "step": 7026 }, { "epoch": 1.62, "learning_rate": 2.293351821893537e-05, "loss": 0.0, "step": 7027 }, { "epoch": 1.62, "learning_rate": 2.2929666435559665e-05, "loss": 0.0, "step": 7028 }, { "epoch": 1.62, "learning_rate": 2.2925814652183963e-05, "loss": 0.0, "step": 7029 }, { "epoch": 1.62, "learning_rate": 2.2921962868808258e-05, "loss": 0.0, "step": 7030 }, { "epoch": 1.62, "learning_rate": 2.2918111085432556e-05, "loss": 0.0, "step": 7031 }, { "epoch": 1.63, "learning_rate": 2.2914259302056855e-05, "loss": 0.0, "step": 7032 }, { "epoch": 1.63, "learning_rate": 2.291040751868115e-05, "loss": 0.0, "step": 7033 }, { "epoch": 1.63, "learning_rate": 2.2906555735305448e-05, "loss": 0.0, "step": 7034 }, { "epoch": 1.63, "learning_rate": 2.2902703951929743e-05, "loss": 0.0, "step": 7035 }, { "epoch": 1.63, "learning_rate": 2.289885216855404e-05, "loss": 0.0003, "step": 7036 }, { "epoch": 1.63, "learning_rate": 2.289500038517834e-05, "loss": 0.0, "step": 7037 }, { "epoch": 1.63, "learning_rate": 2.2891148601802638e-05, "loss": 0.0, "step": 7038 }, { "epoch": 1.63, "learning_rate": 2.2887296818426933e-05, "loss": 0.0, "step": 7039 }, { "epoch": 1.63, "learning_rate": 2.2883445035051228e-05, "loss": 0.0, "step": 7040 }, { "epoch": 1.63, "learning_rate": 2.2879593251675526e-05, "loss": 0.0, "step": 7041 }, { "epoch": 1.63, "learning_rate": 2.2875741468299825e-05, "loss": 0.0, "step": 7042 }, { "epoch": 1.63, "learning_rate": 2.2871889684924123e-05, "loss": 0.0, "step": 7043 }, { "epoch": 1.63, "learning_rate": 2.2868037901548418e-05, "loss": 0.0, "step": 7044 }, { "epoch": 1.63, "learning_rate": 2.2864186118172716e-05, "loss": 0.0, "step": 7045 }, { "epoch": 1.63, "learning_rate": 2.286033433479701e-05, "loss": 0.0, "step": 7046 }, { "epoch": 1.63, "learning_rate": 2.285648255142131e-05, "loss": 0.0, "step": 7047 }, { "epoch": 1.63, "learning_rate": 2.2852630768045608e-05, "loss": 0.0, "step": 7048 }, { "epoch": 1.63, "learning_rate": 2.2848778984669903e-05, "loss": 0.0, "step": 7049 }, { "epoch": 1.63, "learning_rate": 2.28449272012942e-05, "loss": 0.0003, "step": 7050 }, { "epoch": 1.63, "learning_rate": 2.2841075417918496e-05, "loss": 0.0, "step": 7051 }, { "epoch": 1.63, "learning_rate": 2.2837223634542794e-05, "loss": 0.0, "step": 7052 }, { "epoch": 1.63, "learning_rate": 2.283337185116709e-05, "loss": 0.0, "step": 7053 }, { "epoch": 1.63, "learning_rate": 2.2829520067791388e-05, "loss": 0.0, "step": 7054 }, { "epoch": 1.63, "learning_rate": 2.2825668284415686e-05, "loss": 0.0, "step": 7055 }, { "epoch": 1.63, "learning_rate": 2.2821816501039984e-05, "loss": 0.0, "step": 7056 }, { "epoch": 1.63, "learning_rate": 2.281796471766428e-05, "loss": 0.0, "step": 7057 }, { "epoch": 1.63, "learning_rate": 2.2814112934288574e-05, "loss": 0.0, "step": 7058 }, { "epoch": 1.63, "learning_rate": 2.2810261150912873e-05, "loss": 0.0, "step": 7059 }, { "epoch": 1.63, "learning_rate": 2.280640936753717e-05, "loss": 0.0, "step": 7060 }, { "epoch": 1.63, "learning_rate": 2.280255758416147e-05, "loss": 0.0001, "step": 7061 }, { "epoch": 1.63, "learning_rate": 2.2798705800785768e-05, "loss": 0.0, "step": 7062 }, { "epoch": 1.63, "learning_rate": 2.2794854017410063e-05, "loss": 0.0, "step": 7063 }, { "epoch": 1.63, "learning_rate": 2.2791002234034358e-05, "loss": 0.0, "step": 7064 }, { "epoch": 1.63, "learning_rate": 2.2787150450658656e-05, "loss": 0.0, "step": 7065 }, { "epoch": 1.63, "learning_rate": 2.2783298667282954e-05, "loss": 0.0, "step": 7066 }, { "epoch": 1.63, "learning_rate": 2.277944688390725e-05, "loss": 0.0, "step": 7067 }, { "epoch": 1.63, "learning_rate": 2.2775595100531548e-05, "loss": 0.0, "step": 7068 }, { "epoch": 1.63, "learning_rate": 2.2771743317155843e-05, "loss": 0.0, "step": 7069 }, { "epoch": 1.63, "learning_rate": 2.276789153378014e-05, "loss": 0.0, "step": 7070 }, { "epoch": 1.63, "learning_rate": 2.276403975040444e-05, "loss": 0.0, "step": 7071 }, { "epoch": 1.63, "learning_rate": 2.2760187967028734e-05, "loss": 0.0001, "step": 7072 }, { "epoch": 1.63, "learning_rate": 2.2756336183653033e-05, "loss": 0.0, "step": 7073 }, { "epoch": 1.63, "learning_rate": 2.275248440027733e-05, "loss": 0.0, "step": 7074 }, { "epoch": 1.64, "learning_rate": 2.2748632616901626e-05, "loss": 0.0, "step": 7075 }, { "epoch": 1.64, "learning_rate": 2.2744780833525924e-05, "loss": 0.0, "step": 7076 }, { "epoch": 1.64, "learning_rate": 2.274092905015022e-05, "loss": 0.0, "step": 7077 }, { "epoch": 1.64, "learning_rate": 2.2737077266774517e-05, "loss": 0.0, "step": 7078 }, { "epoch": 1.64, "learning_rate": 2.2733225483398816e-05, "loss": 0.0, "step": 7079 }, { "epoch": 1.64, "learning_rate": 2.2729373700023114e-05, "loss": 0.0, "step": 7080 }, { "epoch": 1.64, "learning_rate": 2.272552191664741e-05, "loss": 0.0, "step": 7081 }, { "epoch": 1.64, "learning_rate": 2.2721670133271704e-05, "loss": 0.0, "step": 7082 }, { "epoch": 1.64, "learning_rate": 2.2717818349896002e-05, "loss": 0.0, "step": 7083 }, { "epoch": 1.64, "learning_rate": 2.27139665665203e-05, "loss": 0.0, "step": 7084 }, { "epoch": 1.64, "learning_rate": 2.27101147831446e-05, "loss": 0.0, "step": 7085 }, { "epoch": 1.64, "learning_rate": 2.2706262999768894e-05, "loss": 0.0, "step": 7086 }, { "epoch": 1.64, "learning_rate": 2.270241121639319e-05, "loss": 0.0, "step": 7087 }, { "epoch": 1.64, "learning_rate": 2.2698559433017487e-05, "loss": 0.0, "step": 7088 }, { "epoch": 1.64, "learning_rate": 2.2694707649641786e-05, "loss": 0.0, "step": 7089 }, { "epoch": 1.64, "learning_rate": 2.2690855866266084e-05, "loss": 0.0, "step": 7090 }, { "epoch": 1.64, "learning_rate": 2.268700408289038e-05, "loss": 0.0001, "step": 7091 }, { "epoch": 1.64, "learning_rate": 2.2683152299514677e-05, "loss": 0.0, "step": 7092 }, { "epoch": 1.64, "learning_rate": 2.2679300516138972e-05, "loss": 0.0, "step": 7093 }, { "epoch": 1.64, "learning_rate": 2.267544873276327e-05, "loss": 0.0, "step": 7094 }, { "epoch": 1.64, "learning_rate": 2.2671596949387566e-05, "loss": 0.0, "step": 7095 }, { "epoch": 1.64, "learning_rate": 2.2667745166011864e-05, "loss": 0.0, "step": 7096 }, { "epoch": 1.64, "learning_rate": 2.2663893382636162e-05, "loss": 0.0, "step": 7097 }, { "epoch": 1.64, "learning_rate": 2.266004159926046e-05, "loss": 0.0, "step": 7098 }, { "epoch": 1.64, "learning_rate": 2.2656189815884756e-05, "loss": 0.0, "step": 7099 }, { "epoch": 1.64, "learning_rate": 2.265233803250905e-05, "loss": 0.0, "step": 7100 }, { "epoch": 1.64, "learning_rate": 2.264848624913335e-05, "loss": 0.0, "step": 7101 }, { "epoch": 1.64, "learning_rate": 2.2644634465757647e-05, "loss": 0.0, "step": 7102 }, { "epoch": 1.64, "learning_rate": 2.2640782682381946e-05, "loss": 0.0, "step": 7103 }, { "epoch": 1.64, "learning_rate": 2.2636930899006244e-05, "loss": 0.0019, "step": 7104 }, { "epoch": 1.64, "learning_rate": 2.2633079115630535e-05, "loss": 0.0, "step": 7105 }, { "epoch": 1.64, "learning_rate": 2.2629227332254834e-05, "loss": 0.0, "step": 7106 }, { "epoch": 1.64, "learning_rate": 2.2625375548879132e-05, "loss": 0.0, "step": 7107 }, { "epoch": 1.64, "learning_rate": 2.262152376550343e-05, "loss": 0.0, "step": 7108 }, { "epoch": 1.64, "learning_rate": 2.2617671982127725e-05, "loss": 0.0, "step": 7109 }, { "epoch": 1.64, "learning_rate": 2.2613820198752024e-05, "loss": 0.0, "step": 7110 }, { "epoch": 1.64, "learning_rate": 2.260996841537632e-05, "loss": 0.0, "step": 7111 }, { "epoch": 1.64, "learning_rate": 2.2606116632000617e-05, "loss": 0.0, "step": 7112 }, { "epoch": 1.64, "learning_rate": 2.2602264848624915e-05, "loss": 0.0001, "step": 7113 }, { "epoch": 1.64, "learning_rate": 2.259841306524921e-05, "loss": 0.0, "step": 7114 }, { "epoch": 1.64, "learning_rate": 2.259456128187351e-05, "loss": 0.0, "step": 7115 }, { "epoch": 1.64, "learning_rate": 2.2590709498497807e-05, "loss": 0.0, "step": 7116 }, { "epoch": 1.64, "learning_rate": 2.2586857715122102e-05, "loss": 0.0, "step": 7117 }, { "epoch": 1.65, "learning_rate": 2.25830059317464e-05, "loss": 0.0, "step": 7118 }, { "epoch": 1.65, "learning_rate": 2.2579154148370695e-05, "loss": 0.0, "step": 7119 }, { "epoch": 1.65, "learning_rate": 2.2575302364994994e-05, "loss": 0.0, "step": 7120 }, { "epoch": 1.65, "learning_rate": 2.2571450581619292e-05, "loss": 0.0, "step": 7121 }, { "epoch": 1.65, "learning_rate": 2.256759879824359e-05, "loss": 0.0, "step": 7122 }, { "epoch": 1.65, "learning_rate": 2.2563747014867885e-05, "loss": 0.0, "step": 7123 }, { "epoch": 1.65, "learning_rate": 2.255989523149218e-05, "loss": 0.0, "step": 7124 }, { "epoch": 1.65, "learning_rate": 2.255604344811648e-05, "loss": 0.0, "step": 7125 }, { "epoch": 1.65, "learning_rate": 2.2552191664740777e-05, "loss": 0.0, "step": 7126 }, { "epoch": 1.65, "learning_rate": 2.2548339881365075e-05, "loss": 0.0, "step": 7127 }, { "epoch": 1.65, "learning_rate": 2.254448809798937e-05, "loss": 0.0, "step": 7128 }, { "epoch": 1.65, "learning_rate": 2.2540636314613665e-05, "loss": 0.0, "step": 7129 }, { "epoch": 1.65, "learning_rate": 2.2536784531237964e-05, "loss": 0.0, "step": 7130 }, { "epoch": 1.65, "learning_rate": 2.2532932747862262e-05, "loss": 0.0, "step": 7131 }, { "epoch": 1.65, "learning_rate": 2.252908096448656e-05, "loss": 0.0, "step": 7132 }, { "epoch": 1.65, "learning_rate": 2.2525229181110855e-05, "loss": 0.0, "step": 7133 }, { "epoch": 1.65, "learning_rate": 2.2521377397735154e-05, "loss": 0.0, "step": 7134 }, { "epoch": 1.65, "learning_rate": 2.251752561435945e-05, "loss": 0.0, "step": 7135 }, { "epoch": 1.65, "learning_rate": 2.2513673830983747e-05, "loss": 0.0, "step": 7136 }, { "epoch": 1.65, "learning_rate": 2.2509822047608042e-05, "loss": 0.0, "step": 7137 }, { "epoch": 1.65, "learning_rate": 2.250597026423234e-05, "loss": 0.0, "step": 7138 }, { "epoch": 1.65, "learning_rate": 2.250211848085664e-05, "loss": 0.0001, "step": 7139 }, { "epoch": 1.65, "learning_rate": 2.2498266697480937e-05, "loss": 0.0, "step": 7140 }, { "epoch": 1.65, "learning_rate": 2.2494414914105232e-05, "loss": 0.0001, "step": 7141 }, { "epoch": 1.65, "learning_rate": 2.2490563130729527e-05, "loss": 0.0, "step": 7142 }, { "epoch": 1.65, "learning_rate": 2.2486711347353825e-05, "loss": 0.0, "step": 7143 }, { "epoch": 1.65, "learning_rate": 2.2482859563978123e-05, "loss": 0.0, "step": 7144 }, { "epoch": 1.65, "learning_rate": 2.2479007780602422e-05, "loss": 0.0, "step": 7145 }, { "epoch": 1.65, "learning_rate": 2.2475155997226717e-05, "loss": 0.0, "step": 7146 }, { "epoch": 1.65, "learning_rate": 2.2471304213851012e-05, "loss": 0.0, "step": 7147 }, { "epoch": 1.65, "learning_rate": 2.246745243047531e-05, "loss": 0.0, "step": 7148 }, { "epoch": 1.65, "learning_rate": 2.246360064709961e-05, "loss": 0.0001, "step": 7149 }, { "epoch": 1.65, "learning_rate": 2.2459748863723907e-05, "loss": 0.0001, "step": 7150 }, { "epoch": 1.65, "learning_rate": 2.2455897080348202e-05, "loss": 0.0, "step": 7151 }, { "epoch": 1.65, "learning_rate": 2.24520452969725e-05, "loss": 0.0, "step": 7152 }, { "epoch": 1.65, "learning_rate": 2.2448193513596795e-05, "loss": 0.0, "step": 7153 }, { "epoch": 1.65, "learning_rate": 2.2444341730221093e-05, "loss": 0.0, "step": 7154 }, { "epoch": 1.65, "learning_rate": 2.244048994684539e-05, "loss": 0.0, "step": 7155 }, { "epoch": 1.65, "learning_rate": 2.2436638163469687e-05, "loss": 0.0, "step": 7156 }, { "epoch": 1.65, "learning_rate": 2.2432786380093985e-05, "loss": 0.0001, "step": 7157 }, { "epoch": 1.65, "learning_rate": 2.2428934596718283e-05, "loss": 0.0, "step": 7158 }, { "epoch": 1.65, "learning_rate": 2.2425082813342578e-05, "loss": 0.0, "step": 7159 }, { "epoch": 1.65, "learning_rate": 2.2421231029966877e-05, "loss": 0.0, "step": 7160 }, { "epoch": 1.65, "learning_rate": 2.241737924659117e-05, "loss": 0.0, "step": 7161 }, { "epoch": 1.66, "learning_rate": 2.241352746321547e-05, "loss": 0.0, "step": 7162 }, { "epoch": 1.66, "learning_rate": 2.2409675679839768e-05, "loss": 0.0, "step": 7163 }, { "epoch": 1.66, "learning_rate": 2.2405823896464063e-05, "loss": 0.0, "step": 7164 }, { "epoch": 1.66, "learning_rate": 2.2401972113088358e-05, "loss": 0.0, "step": 7165 }, { "epoch": 1.66, "learning_rate": 2.2398120329712657e-05, "loss": 0.0, "step": 7166 }, { "epoch": 1.66, "learning_rate": 2.2394268546336955e-05, "loss": 0.0, "step": 7167 }, { "epoch": 1.66, "learning_rate": 2.2390416762961253e-05, "loss": 0.0, "step": 7168 }, { "epoch": 1.66, "learning_rate": 2.238656497958555e-05, "loss": 0.0, "step": 7169 }, { "epoch": 1.66, "learning_rate": 2.2382713196209847e-05, "loss": 0.0, "step": 7170 }, { "epoch": 1.66, "learning_rate": 2.237886141283414e-05, "loss": 0.0, "step": 7171 }, { "epoch": 1.66, "learning_rate": 2.237500962945844e-05, "loss": 0.0, "step": 7172 }, { "epoch": 1.66, "learning_rate": 2.2371157846082738e-05, "loss": 0.0, "step": 7173 }, { "epoch": 1.66, "learning_rate": 2.2367306062707036e-05, "loss": 0.0, "step": 7174 }, { "epoch": 1.66, "learning_rate": 2.236345427933133e-05, "loss": 0.0, "step": 7175 }, { "epoch": 1.66, "learning_rate": 2.235960249595563e-05, "loss": 0.0, "step": 7176 }, { "epoch": 1.66, "learning_rate": 2.2355750712579925e-05, "loss": 0.0, "step": 7177 }, { "epoch": 1.66, "learning_rate": 2.2351898929204223e-05, "loss": 0.0, "step": 7178 }, { "epoch": 1.66, "learning_rate": 2.2348047145828518e-05, "loss": 0.0, "step": 7179 }, { "epoch": 1.66, "learning_rate": 2.2344195362452816e-05, "loss": 0.0, "step": 7180 }, { "epoch": 1.66, "learning_rate": 2.2340343579077115e-05, "loss": 0.0, "step": 7181 }, { "epoch": 1.66, "learning_rate": 2.233649179570141e-05, "loss": 0.0, "step": 7182 }, { "epoch": 1.66, "learning_rate": 2.2332640012325708e-05, "loss": 0.0, "step": 7183 }, { "epoch": 1.66, "learning_rate": 2.2328788228950003e-05, "loss": 0.0, "step": 7184 }, { "epoch": 1.66, "learning_rate": 2.23249364455743e-05, "loss": 0.0, "step": 7185 }, { "epoch": 1.66, "learning_rate": 2.23210846621986e-05, "loss": 0.0, "step": 7186 }, { "epoch": 1.66, "learning_rate": 2.2317232878822898e-05, "loss": 0.0, "step": 7187 }, { "epoch": 1.66, "learning_rate": 2.2313381095447193e-05, "loss": 0.0001, "step": 7188 }, { "epoch": 1.66, "learning_rate": 2.2309529312071488e-05, "loss": 0.0, "step": 7189 }, { "epoch": 1.66, "learning_rate": 2.2305677528695786e-05, "loss": 0.0, "step": 7190 }, { "epoch": 1.66, "learning_rate": 2.2301825745320085e-05, "loss": 0.0, "step": 7191 }, { "epoch": 1.66, "learning_rate": 2.2297973961944383e-05, "loss": 0.0, "step": 7192 }, { "epoch": 1.66, "learning_rate": 2.2294122178568678e-05, "loss": 0.0, "step": 7193 }, { "epoch": 1.66, "learning_rate": 2.2290270395192976e-05, "loss": 0.0, "step": 7194 }, { "epoch": 1.66, "learning_rate": 2.228641861181727e-05, "loss": 0.0, "step": 7195 }, { "epoch": 1.66, "learning_rate": 2.228256682844157e-05, "loss": 0.0, "step": 7196 }, { "epoch": 1.66, "learning_rate": 2.2278715045065868e-05, "loss": 0.0, "step": 7197 }, { "epoch": 1.66, "learning_rate": 2.2274863261690163e-05, "loss": 0.0, "step": 7198 }, { "epoch": 1.66, "learning_rate": 2.227101147831446e-05, "loss": 0.0, "step": 7199 }, { "epoch": 1.66, "learning_rate": 2.226715969493876e-05, "loss": 0.0, "step": 7200 }, { "epoch": 1.66, "learning_rate": 2.2263307911563055e-05, "loss": 0.0, "step": 7201 }, { "epoch": 1.66, "learning_rate": 2.2259456128187353e-05, "loss": 0.0, "step": 7202 }, { "epoch": 1.66, "learning_rate": 2.2255604344811648e-05, "loss": 0.0, "step": 7203 }, { "epoch": 1.66, "learning_rate": 2.2251752561435946e-05, "loss": 0.0, "step": 7204 }, { "epoch": 1.67, "learning_rate": 2.2247900778060244e-05, "loss": 0.0, "step": 7205 }, { "epoch": 1.67, "learning_rate": 2.224404899468454e-05, "loss": 0.0, "step": 7206 }, { "epoch": 1.67, "learning_rate": 2.2240197211308834e-05, "loss": 0.0, "step": 7207 }, { "epoch": 1.67, "learning_rate": 2.2236345427933133e-05, "loss": 0.0, "step": 7208 }, { "epoch": 1.67, "learning_rate": 2.223249364455743e-05, "loss": 0.0, "step": 7209 }, { "epoch": 1.67, "learning_rate": 2.222864186118173e-05, "loss": 0.0, "step": 7210 }, { "epoch": 1.67, "learning_rate": 2.2224790077806028e-05, "loss": 0.0, "step": 7211 }, { "epoch": 1.67, "learning_rate": 2.2220938294430323e-05, "loss": 0.0, "step": 7212 }, { "epoch": 1.67, "learning_rate": 2.2217086511054618e-05, "loss": 0.0, "step": 7213 }, { "epoch": 1.67, "learning_rate": 2.2213234727678916e-05, "loss": 0.0, "step": 7214 }, { "epoch": 1.67, "learning_rate": 2.2209382944303214e-05, "loss": 0.0, "step": 7215 }, { "epoch": 1.67, "learning_rate": 2.2205531160927513e-05, "loss": 0.0, "step": 7216 }, { "epoch": 1.67, "learning_rate": 2.2201679377551808e-05, "loss": 0.0, "step": 7217 }, { "epoch": 1.67, "learning_rate": 2.2197827594176106e-05, "loss": 0.0, "step": 7218 }, { "epoch": 1.67, "learning_rate": 2.21939758108004e-05, "loss": 0.0, "step": 7219 }, { "epoch": 1.67, "learning_rate": 2.21901240274247e-05, "loss": 0.0, "step": 7220 }, { "epoch": 1.67, "learning_rate": 2.2186272244048994e-05, "loss": 0.0, "step": 7221 }, { "epoch": 1.67, "learning_rate": 2.2182420460673293e-05, "loss": 0.0, "step": 7222 }, { "epoch": 1.67, "learning_rate": 2.217856867729759e-05, "loss": 0.0, "step": 7223 }, { "epoch": 1.67, "learning_rate": 2.2174716893921886e-05, "loss": 0.0, "step": 7224 }, { "epoch": 1.67, "learning_rate": 2.2170865110546184e-05, "loss": 0.0, "step": 7225 }, { "epoch": 1.67, "learning_rate": 2.216701332717048e-05, "loss": 0.0001, "step": 7226 }, { "epoch": 1.67, "learning_rate": 2.2163161543794778e-05, "loss": 0.0, "step": 7227 }, { "epoch": 1.67, "learning_rate": 2.2159309760419076e-05, "loss": 0.0, "step": 7228 }, { "epoch": 1.67, "learning_rate": 2.2155457977043374e-05, "loss": 0.0, "step": 7229 }, { "epoch": 1.67, "learning_rate": 2.215160619366767e-05, "loss": 0.0, "step": 7230 }, { "epoch": 1.67, "learning_rate": 2.2147754410291964e-05, "loss": 0.0, "step": 7231 }, { "epoch": 1.67, "learning_rate": 2.2143902626916263e-05, "loss": 0.0006, "step": 7232 }, { "epoch": 1.67, "learning_rate": 2.214005084354056e-05, "loss": 0.0, "step": 7233 }, { "epoch": 1.67, "learning_rate": 2.213619906016486e-05, "loss": 0.0, "step": 7234 }, { "epoch": 1.67, "learning_rate": 2.2132347276789154e-05, "loss": 0.0, "step": 7235 }, { "epoch": 1.67, "learning_rate": 2.2128495493413452e-05, "loss": 0.0, "step": 7236 }, { "epoch": 1.67, "learning_rate": 2.2124643710037747e-05, "loss": 0.0, "step": 7237 }, { "epoch": 1.67, "learning_rate": 2.2120791926662046e-05, "loss": 0.0, "step": 7238 }, { "epoch": 1.67, "learning_rate": 2.2116940143286344e-05, "loss": 0.0, "step": 7239 }, { "epoch": 1.67, "learning_rate": 2.211308835991064e-05, "loss": 0.0, "step": 7240 }, { "epoch": 1.67, "learning_rate": 2.2109236576534937e-05, "loss": 0.0, "step": 7241 }, { "epoch": 1.67, "learning_rate": 2.2105384793159232e-05, "loss": 0.0, "step": 7242 }, { "epoch": 1.67, "learning_rate": 2.210153300978353e-05, "loss": 0.0, "step": 7243 }, { "epoch": 1.67, "learning_rate": 2.209768122640783e-05, "loss": 0.0, "step": 7244 }, { "epoch": 1.67, "learning_rate": 2.2093829443032124e-05, "loss": 0.0, "step": 7245 }, { "epoch": 1.67, "learning_rate": 2.2089977659656422e-05, "loss": 0.0, "step": 7246 }, { "epoch": 1.67, "learning_rate": 2.208612587628072e-05, "loss": 0.0, "step": 7247 }, { "epoch": 1.68, "learning_rate": 2.2082274092905016e-05, "loss": 0.0, "step": 7248 }, { "epoch": 1.68, "learning_rate": 2.207842230952931e-05, "loss": 0.0, "step": 7249 }, { "epoch": 1.68, "learning_rate": 2.207457052615361e-05, "loss": 0.0, "step": 7250 }, { "epoch": 1.68, "learning_rate": 2.2070718742777907e-05, "loss": 0.0, "step": 7251 }, { "epoch": 1.68, "learning_rate": 2.2066866959402206e-05, "loss": 0.0, "step": 7252 }, { "epoch": 1.68, "learning_rate": 2.2063015176026504e-05, "loss": 0.0, "step": 7253 }, { "epoch": 1.68, "learning_rate": 2.20591633926508e-05, "loss": 0.0, "step": 7254 }, { "epoch": 1.68, "learning_rate": 2.2055311609275094e-05, "loss": 0.0, "step": 7255 }, { "epoch": 1.68, "learning_rate": 2.2051459825899392e-05, "loss": 0.0, "step": 7256 }, { "epoch": 1.68, "learning_rate": 2.204760804252369e-05, "loss": 0.0, "step": 7257 }, { "epoch": 1.68, "learning_rate": 2.204375625914799e-05, "loss": 0.0, "step": 7258 }, { "epoch": 1.68, "learning_rate": 2.2039904475772284e-05, "loss": 0.0, "step": 7259 }, { "epoch": 1.68, "learning_rate": 2.203605269239658e-05, "loss": 0.0, "step": 7260 }, { "epoch": 1.68, "learning_rate": 2.2032200909020877e-05, "loss": 0.0, "step": 7261 }, { "epoch": 1.68, "learning_rate": 2.2028349125645176e-05, "loss": 0.0, "step": 7262 }, { "epoch": 1.68, "learning_rate": 2.202449734226947e-05, "loss": 0.0, "step": 7263 }, { "epoch": 1.68, "learning_rate": 2.202064555889377e-05, "loss": 0.0, "step": 7264 }, { "epoch": 1.68, "learning_rate": 2.2016793775518067e-05, "loss": 0.0, "step": 7265 }, { "epoch": 1.68, "learning_rate": 2.2012941992142362e-05, "loss": 0.0, "step": 7266 }, { "epoch": 1.68, "learning_rate": 2.200909020876666e-05, "loss": 0.0, "step": 7267 }, { "epoch": 1.68, "learning_rate": 2.2005238425390955e-05, "loss": 0.0, "step": 7268 }, { "epoch": 1.68, "learning_rate": 2.2001386642015254e-05, "loss": 0.0, "step": 7269 }, { "epoch": 1.68, "learning_rate": 2.1997534858639552e-05, "loss": 0.0, "step": 7270 }, { "epoch": 1.68, "learning_rate": 2.199368307526385e-05, "loss": 0.0, "step": 7271 }, { "epoch": 1.68, "learning_rate": 2.1989831291888145e-05, "loss": 0.0, "step": 7272 }, { "epoch": 1.68, "learning_rate": 2.198597950851244e-05, "loss": 0.0, "step": 7273 }, { "epoch": 1.68, "learning_rate": 2.198212772513674e-05, "loss": 0.0, "step": 7274 }, { "epoch": 1.68, "learning_rate": 2.1978275941761037e-05, "loss": 0.0, "step": 7275 }, { "epoch": 1.68, "learning_rate": 2.1974424158385335e-05, "loss": 0.0035, "step": 7276 }, { "epoch": 1.68, "learning_rate": 2.197057237500963e-05, "loss": 0.0, "step": 7277 }, { "epoch": 1.68, "learning_rate": 2.1966720591633925e-05, "loss": 0.0, "step": 7278 }, { "epoch": 1.68, "learning_rate": 2.1962868808258224e-05, "loss": 0.0, "step": 7279 }, { "epoch": 1.68, "learning_rate": 2.1959017024882522e-05, "loss": 0.0, "step": 7280 }, { "epoch": 1.68, "learning_rate": 2.195516524150682e-05, "loss": 0.0, "step": 7281 }, { "epoch": 1.68, "learning_rate": 2.1951313458131115e-05, "loss": 0.0, "step": 7282 }, { "epoch": 1.68, "learning_rate": 2.1947461674755414e-05, "loss": 0.0, "step": 7283 }, { "epoch": 1.68, "learning_rate": 2.194360989137971e-05, "loss": 0.0, "step": 7284 }, { "epoch": 1.68, "learning_rate": 2.1939758108004007e-05, "loss": 0.0, "step": 7285 }, { "epoch": 1.68, "learning_rate": 2.1935906324628305e-05, "loss": 0.0001, "step": 7286 }, { "epoch": 1.68, "learning_rate": 2.19320545412526e-05, "loss": 0.0, "step": 7287 }, { "epoch": 1.68, "learning_rate": 2.19282027578769e-05, "loss": 0.0, "step": 7288 }, { "epoch": 1.68, "learning_rate": 2.1924350974501197e-05, "loss": 0.0, "step": 7289 }, { "epoch": 1.68, "learning_rate": 2.1920499191125492e-05, "loss": 0.0, "step": 7290 }, { "epoch": 1.69, "learning_rate": 2.1916647407749787e-05, "loss": 0.0, "step": 7291 }, { "epoch": 1.69, "learning_rate": 2.1912795624374085e-05, "loss": 0.0, "step": 7292 }, { "epoch": 1.69, "learning_rate": 2.1908943840998384e-05, "loss": 0.0001, "step": 7293 }, { "epoch": 1.69, "learning_rate": 2.1905092057622682e-05, "loss": 0.0, "step": 7294 }, { "epoch": 1.69, "learning_rate": 2.190124027424698e-05, "loss": 0.0, "step": 7295 }, { "epoch": 1.69, "learning_rate": 2.1897388490871272e-05, "loss": 0.0, "step": 7296 }, { "epoch": 1.69, "learning_rate": 2.189353670749557e-05, "loss": 0.0006, "step": 7297 }, { "epoch": 1.69, "learning_rate": 2.188968492411987e-05, "loss": 0.0, "step": 7298 }, { "epoch": 1.69, "learning_rate": 2.1885833140744167e-05, "loss": 0.0, "step": 7299 }, { "epoch": 1.69, "learning_rate": 2.1881981357368465e-05, "loss": 0.0, "step": 7300 }, { "epoch": 1.69, "learning_rate": 2.187812957399276e-05, "loss": 0.0, "step": 7301 }, { "epoch": 1.69, "learning_rate": 2.1874277790617055e-05, "loss": 0.0003, "step": 7302 }, { "epoch": 1.69, "learning_rate": 2.1870426007241353e-05, "loss": 0.0, "step": 7303 }, { "epoch": 1.69, "learning_rate": 2.1866574223865652e-05, "loss": 0.0, "step": 7304 }, { "epoch": 1.69, "learning_rate": 2.1862722440489947e-05, "loss": 0.0, "step": 7305 }, { "epoch": 1.69, "learning_rate": 2.1858870657114245e-05, "loss": 0.0, "step": 7306 }, { "epoch": 1.69, "learning_rate": 2.1855018873738543e-05, "loss": 0.0, "step": 7307 }, { "epoch": 1.69, "learning_rate": 2.185116709036284e-05, "loss": 0.0, "step": 7308 }, { "epoch": 1.69, "learning_rate": 2.1847315306987137e-05, "loss": 0.0, "step": 7309 }, { "epoch": 1.69, "learning_rate": 2.184346352361143e-05, "loss": 0.0, "step": 7310 }, { "epoch": 1.69, "learning_rate": 2.183961174023573e-05, "loss": 0.0, "step": 7311 }, { "epoch": 1.69, "learning_rate": 2.183575995686003e-05, "loss": 0.0, "step": 7312 }, { "epoch": 1.69, "learning_rate": 2.1831908173484327e-05, "loss": 0.0, "step": 7313 }, { "epoch": 1.69, "learning_rate": 2.182805639010862e-05, "loss": 0.0001, "step": 7314 }, { "epoch": 1.69, "learning_rate": 2.1824204606732917e-05, "loss": 0.0, "step": 7315 }, { "epoch": 1.69, "learning_rate": 2.1820352823357215e-05, "loss": 0.0, "step": 7316 }, { "epoch": 1.69, "learning_rate": 2.1816501039981513e-05, "loss": 0.0, "step": 7317 }, { "epoch": 1.69, "learning_rate": 2.181264925660581e-05, "loss": 0.0, "step": 7318 }, { "epoch": 1.69, "learning_rate": 2.1808797473230107e-05, "loss": 0.0, "step": 7319 }, { "epoch": 1.69, "learning_rate": 2.18049456898544e-05, "loss": 0.0, "step": 7320 }, { "epoch": 1.69, "learning_rate": 2.18010939064787e-05, "loss": 0.0, "step": 7321 }, { "epoch": 1.69, "learning_rate": 2.1797242123102998e-05, "loss": 0.0, "step": 7322 }, { "epoch": 1.69, "learning_rate": 2.1793390339727297e-05, "loss": 0.0, "step": 7323 }, { "epoch": 1.69, "learning_rate": 2.178953855635159e-05, "loss": 0.0, "step": 7324 }, { "epoch": 1.69, "learning_rate": 2.178568677297589e-05, "loss": 0.0001, "step": 7325 }, { "epoch": 1.69, "learning_rate": 2.1781834989600185e-05, "loss": 0.0, "step": 7326 }, { "epoch": 1.69, "learning_rate": 2.1777983206224483e-05, "loss": 0.0, "step": 7327 }, { "epoch": 1.69, "learning_rate": 2.177413142284878e-05, "loss": 0.0, "step": 7328 }, { "epoch": 1.69, "learning_rate": 2.1770279639473076e-05, "loss": 0.0, "step": 7329 }, { "epoch": 1.69, "learning_rate": 2.1766427856097375e-05, "loss": 0.0, "step": 7330 }, { "epoch": 1.69, "learning_rate": 2.1762576072721673e-05, "loss": 0.0, "step": 7331 }, { "epoch": 1.69, "learning_rate": 2.1758724289345968e-05, "loss": 0.0, "step": 7332 }, { "epoch": 1.69, "learning_rate": 2.1754872505970263e-05, "loss": 0.0, "step": 7333 }, { "epoch": 1.69, "learning_rate": 2.175102072259456e-05, "loss": 0.0, "step": 7334 }, { "epoch": 1.7, "learning_rate": 2.174716893921886e-05, "loss": 0.0, "step": 7335 }, { "epoch": 1.7, "learning_rate": 2.1743317155843158e-05, "loss": 0.0, "step": 7336 }, { "epoch": 1.7, "learning_rate": 2.1739465372467453e-05, "loss": 0.0, "step": 7337 }, { "epoch": 1.7, "learning_rate": 2.1735613589091748e-05, "loss": 0.0, "step": 7338 }, { "epoch": 1.7, "learning_rate": 2.1731761805716046e-05, "loss": 0.0, "step": 7339 }, { "epoch": 1.7, "learning_rate": 2.1727910022340345e-05, "loss": 0.0, "step": 7340 }, { "epoch": 1.7, "learning_rate": 2.1724058238964643e-05, "loss": 0.0, "step": 7341 }, { "epoch": 1.7, "learning_rate": 2.172020645558894e-05, "loss": 0.0, "step": 7342 }, { "epoch": 1.7, "learning_rate": 2.1716354672213236e-05, "loss": 0.0, "step": 7343 }, { "epoch": 1.7, "learning_rate": 2.171250288883753e-05, "loss": 0.0, "step": 7344 }, { "epoch": 1.7, "learning_rate": 2.170865110546183e-05, "loss": 0.0, "step": 7345 }, { "epoch": 1.7, "learning_rate": 2.1704799322086128e-05, "loss": 0.0, "step": 7346 }, { "epoch": 1.7, "learning_rate": 2.1700947538710423e-05, "loss": 0.0, "step": 7347 }, { "epoch": 1.7, "learning_rate": 2.169709575533472e-05, "loss": 0.0, "step": 7348 }, { "epoch": 1.7, "learning_rate": 2.169324397195902e-05, "loss": 0.0008, "step": 7349 }, { "epoch": 1.7, "learning_rate": 2.1689392188583315e-05, "loss": 0.0, "step": 7350 }, { "epoch": 1.7, "learning_rate": 2.1685540405207613e-05, "loss": 0.0, "step": 7351 }, { "epoch": 1.7, "learning_rate": 2.1681688621831908e-05, "loss": 0.0, "step": 7352 }, { "epoch": 1.7, "learning_rate": 2.1677836838456206e-05, "loss": 0.0, "step": 7353 }, { "epoch": 1.7, "learning_rate": 2.1673985055080505e-05, "loss": 0.0, "step": 7354 }, { "epoch": 1.7, "learning_rate": 2.16701332717048e-05, "loss": 0.0, "step": 7355 }, { "epoch": 1.7, "learning_rate": 2.1666281488329098e-05, "loss": 0.0, "step": 7356 }, { "epoch": 1.7, "learning_rate": 2.1662429704953393e-05, "loss": 0.0, "step": 7357 }, { "epoch": 1.7, "learning_rate": 2.165857792157769e-05, "loss": 0.0, "step": 7358 }, { "epoch": 1.7, "learning_rate": 2.165472613820199e-05, "loss": 0.0, "step": 7359 }, { "epoch": 1.7, "learning_rate": 2.1650874354826288e-05, "loss": 0.0, "step": 7360 }, { "epoch": 1.7, "learning_rate": 2.1647022571450583e-05, "loss": 0.0001, "step": 7361 }, { "epoch": 1.7, "learning_rate": 2.1643170788074878e-05, "loss": 0.0, "step": 7362 }, { "epoch": 1.7, "learning_rate": 2.1639319004699176e-05, "loss": 0.0, "step": 7363 }, { "epoch": 1.7, "learning_rate": 2.1635467221323474e-05, "loss": 0.0, "step": 7364 }, { "epoch": 1.7, "learning_rate": 2.1631615437947773e-05, "loss": 0.0, "step": 7365 }, { "epoch": 1.7, "learning_rate": 2.1627763654572068e-05, "loss": 0.0, "step": 7366 }, { "epoch": 1.7, "learning_rate": 2.1623911871196366e-05, "loss": 0.0, "step": 7367 }, { "epoch": 1.7, "learning_rate": 2.162006008782066e-05, "loss": 0.0, "step": 7368 }, { "epoch": 1.7, "learning_rate": 2.161620830444496e-05, "loss": 0.0, "step": 7369 }, { "epoch": 1.7, "learning_rate": 2.1612356521069258e-05, "loss": 0.0, "step": 7370 }, { "epoch": 1.7, "learning_rate": 2.1608504737693553e-05, "loss": 0.0006, "step": 7371 }, { "epoch": 1.7, "learning_rate": 2.160465295431785e-05, "loss": 0.0, "step": 7372 }, { "epoch": 1.7, "learning_rate": 2.1600801170942146e-05, "loss": 0.0, "step": 7373 }, { "epoch": 1.7, "learning_rate": 2.1596949387566444e-05, "loss": 0.0, "step": 7374 }, { "epoch": 1.7, "learning_rate": 2.159309760419074e-05, "loss": 0.0, "step": 7375 }, { "epoch": 1.7, "learning_rate": 2.1589245820815038e-05, "loss": 0.0, "step": 7376 }, { "epoch": 1.7, "learning_rate": 2.1585394037439336e-05, "loss": 0.0, "step": 7377 }, { "epoch": 1.71, "learning_rate": 2.1581542254063634e-05, "loss": 0.0, "step": 7378 }, { "epoch": 1.71, "learning_rate": 2.157769047068793e-05, "loss": 0.0, "step": 7379 }, { "epoch": 1.71, "learning_rate": 2.1573838687312224e-05, "loss": 0.0, "step": 7380 }, { "epoch": 1.71, "learning_rate": 2.1569986903936523e-05, "loss": 0.0, "step": 7381 }, { "epoch": 1.71, "learning_rate": 2.156613512056082e-05, "loss": 0.0, "step": 7382 }, { "epoch": 1.71, "learning_rate": 2.156228333718512e-05, "loss": 0.0, "step": 7383 }, { "epoch": 1.71, "learning_rate": 2.1558431553809418e-05, "loss": 0.0, "step": 7384 }, { "epoch": 1.71, "learning_rate": 2.1554579770433713e-05, "loss": 0.0, "step": 7385 }, { "epoch": 1.71, "learning_rate": 2.1550727987058008e-05, "loss": 0.0, "step": 7386 }, { "epoch": 1.71, "learning_rate": 2.1546876203682306e-05, "loss": 0.0, "step": 7387 }, { "epoch": 1.71, "learning_rate": 2.1543024420306604e-05, "loss": 0.0, "step": 7388 }, { "epoch": 1.71, "learning_rate": 2.15391726369309e-05, "loss": 0.0, "step": 7389 }, { "epoch": 1.71, "learning_rate": 2.1535320853555198e-05, "loss": 0.0, "step": 7390 }, { "epoch": 1.71, "learning_rate": 2.1531469070179492e-05, "loss": 0.0, "step": 7391 }, { "epoch": 1.71, "learning_rate": 2.152761728680379e-05, "loss": 0.0, "step": 7392 }, { "epoch": 1.71, "learning_rate": 2.152376550342809e-05, "loss": 0.0, "step": 7393 }, { "epoch": 1.71, "learning_rate": 2.1519913720052384e-05, "loss": 0.0, "step": 7394 }, { "epoch": 1.71, "learning_rate": 2.1516061936676682e-05, "loss": 0.0, "step": 7395 }, { "epoch": 1.71, "learning_rate": 2.151221015330098e-05, "loss": 0.0038, "step": 7396 }, { "epoch": 1.71, "learning_rate": 2.1508358369925276e-05, "loss": 0.0, "step": 7397 }, { "epoch": 1.71, "learning_rate": 2.1504506586549574e-05, "loss": 0.0, "step": 7398 }, { "epoch": 1.71, "learning_rate": 2.150065480317387e-05, "loss": 0.0001, "step": 7399 }, { "epoch": 1.71, "learning_rate": 2.1496803019798167e-05, "loss": 0.0, "step": 7400 }, { "epoch": 1.71, "learning_rate": 2.1492951236422466e-05, "loss": 0.0, "step": 7401 }, { "epoch": 1.71, "learning_rate": 2.1489099453046764e-05, "loss": 0.0, "step": 7402 }, { "epoch": 1.71, "learning_rate": 2.148524766967106e-05, "loss": 0.0, "step": 7403 }, { "epoch": 1.71, "learning_rate": 2.1481395886295354e-05, "loss": 0.0, "step": 7404 }, { "epoch": 1.71, "learning_rate": 2.1477544102919652e-05, "loss": 0.0, "step": 7405 }, { "epoch": 1.71, "learning_rate": 2.147369231954395e-05, "loss": 0.0, "step": 7406 }, { "epoch": 1.71, "learning_rate": 2.146984053616825e-05, "loss": 0.0, "step": 7407 }, { "epoch": 1.71, "learning_rate": 2.1465988752792544e-05, "loss": 0.0, "step": 7408 }, { "epoch": 1.71, "learning_rate": 2.146213696941684e-05, "loss": 0.0, "step": 7409 }, { "epoch": 1.71, "learning_rate": 2.1458285186041137e-05, "loss": 0.0, "step": 7410 }, { "epoch": 1.71, "learning_rate": 2.1454433402665436e-05, "loss": 0.0, "step": 7411 }, { "epoch": 1.71, "learning_rate": 2.1450581619289734e-05, "loss": 0.0, "step": 7412 }, { "epoch": 1.71, "learning_rate": 2.144672983591403e-05, "loss": 0.0, "step": 7413 }, { "epoch": 1.71, "learning_rate": 2.1442878052538327e-05, "loss": 0.0, "step": 7414 }, { "epoch": 1.71, "learning_rate": 2.1439026269162622e-05, "loss": 0.0, "step": 7415 }, { "epoch": 1.71, "learning_rate": 2.143517448578692e-05, "loss": 0.0, "step": 7416 }, { "epoch": 1.71, "learning_rate": 2.1431322702411216e-05, "loss": 0.0, "step": 7417 }, { "epoch": 1.71, "learning_rate": 2.1427470919035514e-05, "loss": 0.0, "step": 7418 }, { "epoch": 1.71, "learning_rate": 2.1423619135659812e-05, "loss": 0.0, "step": 7419 }, { "epoch": 1.71, "learning_rate": 2.141976735228411e-05, "loss": 0.0, "step": 7420 }, { "epoch": 1.72, "learning_rate": 2.1415915568908406e-05, "loss": 0.0, "step": 7421 }, { "epoch": 1.72, "learning_rate": 2.14120637855327e-05, "loss": 0.0, "step": 7422 }, { "epoch": 1.72, "learning_rate": 2.1408212002157e-05, "loss": 0.0, "step": 7423 }, { "epoch": 1.72, "learning_rate": 2.1404360218781297e-05, "loss": 0.0, "step": 7424 }, { "epoch": 1.72, "learning_rate": 2.1400508435405596e-05, "loss": 0.0, "step": 7425 }, { "epoch": 1.72, "learning_rate": 2.1396656652029894e-05, "loss": 0.0, "step": 7426 }, { "epoch": 1.72, "learning_rate": 2.1392804868654185e-05, "loss": 0.0, "step": 7427 }, { "epoch": 1.72, "learning_rate": 2.1388953085278484e-05, "loss": 0.0, "step": 7428 }, { "epoch": 1.72, "learning_rate": 2.1385101301902782e-05, "loss": 0.0, "step": 7429 }, { "epoch": 1.72, "learning_rate": 2.138124951852708e-05, "loss": 0.0, "step": 7430 }, { "epoch": 1.72, "learning_rate": 2.1377397735151375e-05, "loss": 0.0, "step": 7431 }, { "epoch": 1.72, "learning_rate": 2.1373545951775674e-05, "loss": 0.0, "step": 7432 }, { "epoch": 1.72, "learning_rate": 2.136969416839997e-05, "loss": 0.0, "step": 7433 }, { "epoch": 1.72, "learning_rate": 2.1365842385024267e-05, "loss": 0.0, "step": 7434 }, { "epoch": 1.72, "learning_rate": 2.1361990601648565e-05, "loss": 0.0, "step": 7435 }, { "epoch": 1.72, "learning_rate": 2.135813881827286e-05, "loss": 0.0002, "step": 7436 }, { "epoch": 1.72, "learning_rate": 2.135428703489716e-05, "loss": 0.0001, "step": 7437 }, { "epoch": 1.72, "learning_rate": 2.1350435251521457e-05, "loss": 0.0, "step": 7438 }, { "epoch": 1.72, "learning_rate": 2.1346583468145752e-05, "loss": 0.0, "step": 7439 }, { "epoch": 1.72, "learning_rate": 2.134273168477005e-05, "loss": 0.0001, "step": 7440 }, { "epoch": 1.72, "learning_rate": 2.1338879901394345e-05, "loss": 0.0, "step": 7441 }, { "epoch": 1.72, "learning_rate": 2.1335028118018644e-05, "loss": 0.0, "step": 7442 }, { "epoch": 1.72, "learning_rate": 2.1331176334642942e-05, "loss": 0.0, "step": 7443 }, { "epoch": 1.72, "learning_rate": 2.132732455126724e-05, "loss": 0.0, "step": 7444 }, { "epoch": 1.72, "learning_rate": 2.1323472767891535e-05, "loss": 0.0, "step": 7445 }, { "epoch": 1.72, "learning_rate": 2.131962098451583e-05, "loss": 0.0, "step": 7446 }, { "epoch": 1.72, "learning_rate": 2.131576920114013e-05, "loss": 0.0, "step": 7447 }, { "epoch": 1.72, "learning_rate": 2.1311917417764427e-05, "loss": 0.0, "step": 7448 }, { "epoch": 1.72, "learning_rate": 2.1308065634388725e-05, "loss": 0.0, "step": 7449 }, { "epoch": 1.72, "learning_rate": 2.130421385101302e-05, "loss": 0.0, "step": 7450 }, { "epoch": 1.72, "learning_rate": 2.1300362067637315e-05, "loss": 0.0, "step": 7451 }, { "epoch": 1.72, "learning_rate": 2.1296510284261614e-05, "loss": 0.0, "step": 7452 }, { "epoch": 1.72, "learning_rate": 2.1292658500885912e-05, "loss": 0.0, "step": 7453 }, { "epoch": 1.72, "learning_rate": 2.128880671751021e-05, "loss": 0.0, "step": 7454 }, { "epoch": 1.72, "learning_rate": 2.1284954934134505e-05, "loss": 0.0002, "step": 7455 }, { "epoch": 1.72, "learning_rate": 2.1281103150758803e-05, "loss": 0.0003, "step": 7456 }, { "epoch": 1.72, "learning_rate": 2.12772513673831e-05, "loss": 0.0, "step": 7457 }, { "epoch": 1.72, "learning_rate": 2.1273399584007397e-05, "loss": 0.0, "step": 7458 }, { "epoch": 1.72, "learning_rate": 2.1269547800631692e-05, "loss": 0.0, "step": 7459 }, { "epoch": 1.72, "learning_rate": 2.126569601725599e-05, "loss": 0.0, "step": 7460 }, { "epoch": 1.72, "learning_rate": 2.126184423388029e-05, "loss": 0.0, "step": 7461 }, { "epoch": 1.72, "learning_rate": 2.1257992450504587e-05, "loss": 0.0, "step": 7462 }, { "epoch": 1.72, "learning_rate": 2.1254140667128882e-05, "loss": 0.0, "step": 7463 }, { "epoch": 1.72, "learning_rate": 2.1250288883753177e-05, "loss": 0.0, "step": 7464 }, { "epoch": 1.73, "learning_rate": 2.1246437100377475e-05, "loss": 0.0, "step": 7465 }, { "epoch": 1.73, "learning_rate": 2.1242585317001773e-05, "loss": 0.0, "step": 7466 }, { "epoch": 1.73, "learning_rate": 2.1238733533626072e-05, "loss": 0.0, "step": 7467 }, { "epoch": 1.73, "learning_rate": 2.1234881750250367e-05, "loss": 0.0, "step": 7468 }, { "epoch": 1.73, "learning_rate": 2.123102996687466e-05, "loss": 0.0001, "step": 7469 }, { "epoch": 1.73, "learning_rate": 2.122717818349896e-05, "loss": 0.0, "step": 7470 }, { "epoch": 1.73, "learning_rate": 2.122332640012326e-05, "loss": 0.0, "step": 7471 }, { "epoch": 1.73, "learning_rate": 2.1219474616747557e-05, "loss": 0.0, "step": 7472 }, { "epoch": 1.73, "learning_rate": 2.121562283337185e-05, "loss": 0.0, "step": 7473 }, { "epoch": 1.73, "learning_rate": 2.121177104999615e-05, "loss": 0.0, "step": 7474 }, { "epoch": 1.73, "learning_rate": 2.1207919266620445e-05, "loss": 0.0, "step": 7475 }, { "epoch": 1.73, "learning_rate": 2.1204067483244743e-05, "loss": 0.0001, "step": 7476 }, { "epoch": 1.73, "learning_rate": 2.120021569986904e-05, "loss": 0.0, "step": 7477 }, { "epoch": 1.73, "learning_rate": 2.1196363916493337e-05, "loss": 0.0, "step": 7478 }, { "epoch": 1.73, "learning_rate": 2.1192512133117635e-05, "loss": 0.0, "step": 7479 }, { "epoch": 1.73, "learning_rate": 2.1188660349741933e-05, "loss": 0.0, "step": 7480 }, { "epoch": 1.73, "learning_rate": 2.1184808566366228e-05, "loss": 0.0, "step": 7481 }, { "epoch": 1.73, "learning_rate": 2.1180956782990527e-05, "loss": 0.0, "step": 7482 }, { "epoch": 1.73, "learning_rate": 2.117710499961482e-05, "loss": 0.0, "step": 7483 }, { "epoch": 1.73, "learning_rate": 2.117325321623912e-05, "loss": 0.0, "step": 7484 }, { "epoch": 1.73, "learning_rate": 2.1169401432863418e-05, "loss": 0.0, "step": 7485 }, { "epoch": 1.73, "learning_rate": 2.1165549649487713e-05, "loss": 0.0, "step": 7486 }, { "epoch": 1.73, "learning_rate": 2.1161697866112008e-05, "loss": 0.0, "step": 7487 }, { "epoch": 1.73, "learning_rate": 2.1157846082736306e-05, "loss": 0.0, "step": 7488 }, { "epoch": 1.73, "learning_rate": 2.1153994299360605e-05, "loss": 0.0, "step": 7489 }, { "epoch": 1.73, "learning_rate": 2.1150142515984903e-05, "loss": 0.0, "step": 7490 }, { "epoch": 1.73, "learning_rate": 2.11462907326092e-05, "loss": 0.0, "step": 7491 }, { "epoch": 1.73, "learning_rate": 2.1142438949233496e-05, "loss": 0.0, "step": 7492 }, { "epoch": 1.73, "learning_rate": 2.113858716585779e-05, "loss": 0.0, "step": 7493 }, { "epoch": 1.73, "learning_rate": 2.113473538248209e-05, "loss": 0.0, "step": 7494 }, { "epoch": 1.73, "learning_rate": 2.1130883599106388e-05, "loss": 0.0, "step": 7495 }, { "epoch": 1.73, "learning_rate": 2.1127031815730686e-05, "loss": 0.0001, "step": 7496 }, { "epoch": 1.73, "learning_rate": 2.112318003235498e-05, "loss": 0.0, "step": 7497 }, { "epoch": 1.73, "learning_rate": 2.111932824897928e-05, "loss": 0.0, "step": 7498 }, { "epoch": 1.73, "learning_rate": 2.1115476465603575e-05, "loss": 0.0, "step": 7499 }, { "epoch": 1.73, "learning_rate": 2.1111624682227873e-05, "loss": 0.0, "step": 7500 }, { "epoch": 1.73, "learning_rate": 2.1107772898852168e-05, "loss": 0.0, "step": 7501 }, { "epoch": 1.73, "learning_rate": 2.1103921115476466e-05, "loss": 0.0, "step": 7502 }, { "epoch": 1.73, "learning_rate": 2.1100069332100765e-05, "loss": 0.0, "step": 7503 }, { "epoch": 1.73, "learning_rate": 2.109621754872506e-05, "loss": 0.0, "step": 7504 }, { "epoch": 1.73, "learning_rate": 2.1092365765349358e-05, "loss": 0.0, "step": 7505 }, { "epoch": 1.73, "learning_rate": 2.1088513981973653e-05, "loss": 0.0, "step": 7506 }, { "epoch": 1.73, "learning_rate": 2.108466219859795e-05, "loss": 0.0, "step": 7507 }, { "epoch": 1.74, "learning_rate": 2.108081041522225e-05, "loss": 0.0, "step": 7508 }, { "epoch": 1.74, "learning_rate": 2.1076958631846548e-05, "loss": 0.0003, "step": 7509 }, { "epoch": 1.74, "learning_rate": 2.1073106848470843e-05, "loss": 0.0, "step": 7510 }, { "epoch": 1.74, "learning_rate": 2.1069255065095138e-05, "loss": 0.0007, "step": 7511 }, { "epoch": 1.74, "learning_rate": 2.1065403281719436e-05, "loss": 0.0, "step": 7512 }, { "epoch": 1.74, "learning_rate": 2.1061551498343735e-05, "loss": 0.0, "step": 7513 }, { "epoch": 1.74, "learning_rate": 2.1057699714968033e-05, "loss": 0.0, "step": 7514 }, { "epoch": 1.74, "learning_rate": 2.1053847931592328e-05, "loss": 0.0, "step": 7515 }, { "epoch": 1.74, "learning_rate": 2.1049996148216626e-05, "loss": 0.0, "step": 7516 }, { "epoch": 1.74, "learning_rate": 2.104614436484092e-05, "loss": 0.0, "step": 7517 }, { "epoch": 1.74, "learning_rate": 2.104229258146522e-05, "loss": 0.0, "step": 7518 }, { "epoch": 1.74, "learning_rate": 2.1038440798089518e-05, "loss": 0.0, "step": 7519 }, { "epoch": 1.74, "learning_rate": 2.1034589014713813e-05, "loss": 0.0003, "step": 7520 }, { "epoch": 1.74, "learning_rate": 2.103073723133811e-05, "loss": 0.0, "step": 7521 }, { "epoch": 1.74, "learning_rate": 2.1026885447962406e-05, "loss": 0.0, "step": 7522 }, { "epoch": 1.74, "learning_rate": 2.1023033664586704e-05, "loss": 0.0, "step": 7523 }, { "epoch": 1.74, "learning_rate": 2.1019181881211003e-05, "loss": 0.0001, "step": 7524 }, { "epoch": 1.74, "learning_rate": 2.1015330097835298e-05, "loss": 0.0, "step": 7525 }, { "epoch": 1.74, "learning_rate": 2.1011478314459596e-05, "loss": 0.0, "step": 7526 }, { "epoch": 1.74, "learning_rate": 2.1007626531083894e-05, "loss": 0.0, "step": 7527 }, { "epoch": 1.74, "learning_rate": 2.100377474770819e-05, "loss": 0.0, "step": 7528 }, { "epoch": 1.74, "learning_rate": 2.0999922964332484e-05, "loss": 0.0, "step": 7529 }, { "epoch": 1.74, "learning_rate": 2.0996071180956783e-05, "loss": 0.0, "step": 7530 }, { "epoch": 1.74, "learning_rate": 2.099221939758108e-05, "loss": 0.0, "step": 7531 }, { "epoch": 1.74, "learning_rate": 2.098836761420538e-05, "loss": 0.0, "step": 7532 }, { "epoch": 1.74, "learning_rate": 2.0984515830829678e-05, "loss": 0.0, "step": 7533 }, { "epoch": 1.74, "learning_rate": 2.0980664047453973e-05, "loss": 0.0, "step": 7534 }, { "epoch": 1.74, "learning_rate": 2.0976812264078268e-05, "loss": 0.0, "step": 7535 }, { "epoch": 1.74, "learning_rate": 2.0972960480702566e-05, "loss": 0.0, "step": 7536 }, { "epoch": 1.74, "learning_rate": 2.0969108697326864e-05, "loss": 0.0, "step": 7537 }, { "epoch": 1.74, "learning_rate": 2.0965256913951163e-05, "loss": 0.0, "step": 7538 }, { "epoch": 1.74, "learning_rate": 2.0961405130575458e-05, "loss": 0.0, "step": 7539 }, { "epoch": 1.74, "learning_rate": 2.0957553347199756e-05, "loss": 0.0, "step": 7540 }, { "epoch": 1.74, "learning_rate": 2.095370156382405e-05, "loss": 0.0, "step": 7541 }, { "epoch": 1.74, "learning_rate": 2.094984978044835e-05, "loss": 0.0002, "step": 7542 }, { "epoch": 1.74, "learning_rate": 2.0945997997072644e-05, "loss": 0.0002, "step": 7543 }, { "epoch": 1.74, "learning_rate": 2.0942146213696943e-05, "loss": 0.0, "step": 7544 }, { "epoch": 1.74, "learning_rate": 2.093829443032124e-05, "loss": 0.0, "step": 7545 }, { "epoch": 1.74, "learning_rate": 2.0934442646945536e-05, "loss": 0.0, "step": 7546 }, { "epoch": 1.74, "learning_rate": 2.0930590863569834e-05, "loss": 0.0067, "step": 7547 }, { "epoch": 1.74, "learning_rate": 2.092673908019413e-05, "loss": 0.0, "step": 7548 }, { "epoch": 1.74, "learning_rate": 2.0922887296818427e-05, "loss": 0.0, "step": 7549 }, { "epoch": 1.74, "learning_rate": 2.0919035513442726e-05, "loss": 0.0, "step": 7550 }, { "epoch": 1.75, "learning_rate": 2.0915183730067024e-05, "loss": 0.0, "step": 7551 }, { "epoch": 1.75, "learning_rate": 2.091133194669132e-05, "loss": 0.0, "step": 7552 }, { "epoch": 1.75, "learning_rate": 2.0907480163315614e-05, "loss": 0.0, "step": 7553 }, { "epoch": 1.75, "learning_rate": 2.0903628379939912e-05, "loss": 0.0, "step": 7554 }, { "epoch": 1.75, "learning_rate": 2.089977659656421e-05, "loss": 0.0, "step": 7555 }, { "epoch": 1.75, "learning_rate": 2.089592481318851e-05, "loss": 0.0, "step": 7556 }, { "epoch": 1.75, "learning_rate": 2.0892073029812804e-05, "loss": 0.0012, "step": 7557 }, { "epoch": 1.75, "learning_rate": 2.0888221246437102e-05, "loss": 0.0, "step": 7558 }, { "epoch": 1.75, "learning_rate": 2.0884369463061397e-05, "loss": 0.0, "step": 7559 }, { "epoch": 1.75, "learning_rate": 2.0880517679685696e-05, "loss": 0.0, "step": 7560 }, { "epoch": 1.75, "learning_rate": 2.0876665896309994e-05, "loss": 0.0, "step": 7561 }, { "epoch": 1.75, "learning_rate": 2.087281411293429e-05, "loss": 0.0, "step": 7562 }, { "epoch": 1.75, "learning_rate": 2.0868962329558587e-05, "loss": 0.0, "step": 7563 }, { "epoch": 1.75, "learning_rate": 2.0865110546182882e-05, "loss": 0.0, "step": 7564 }, { "epoch": 1.75, "learning_rate": 2.086125876280718e-05, "loss": 0.0, "step": 7565 }, { "epoch": 1.75, "learning_rate": 2.085740697943148e-05, "loss": 0.0, "step": 7566 }, { "epoch": 1.75, "learning_rate": 2.0853555196055774e-05, "loss": 0.0, "step": 7567 }, { "epoch": 1.75, "learning_rate": 2.0849703412680072e-05, "loss": 0.0, "step": 7568 }, { "epoch": 1.75, "learning_rate": 2.084585162930437e-05, "loss": 0.0002, "step": 7569 }, { "epoch": 1.75, "learning_rate": 2.0841999845928666e-05, "loss": 0.0, "step": 7570 }, { "epoch": 1.75, "learning_rate": 2.083814806255296e-05, "loss": 0.0, "step": 7571 }, { "epoch": 1.75, "learning_rate": 2.083429627917726e-05, "loss": 0.0, "step": 7572 }, { "epoch": 1.75, "learning_rate": 2.0830444495801557e-05, "loss": 0.0, "step": 7573 }, { "epoch": 1.75, "learning_rate": 2.0826592712425856e-05, "loss": 0.0, "step": 7574 }, { "epoch": 1.75, "learning_rate": 2.0822740929050154e-05, "loss": 0.0, "step": 7575 }, { "epoch": 1.75, "learning_rate": 2.081888914567445e-05, "loss": 0.0, "step": 7576 }, { "epoch": 1.75, "learning_rate": 2.0815037362298744e-05, "loss": 0.0, "step": 7577 }, { "epoch": 1.75, "learning_rate": 2.0811185578923042e-05, "loss": 0.0, "step": 7578 }, { "epoch": 1.75, "learning_rate": 2.080733379554734e-05, "loss": 0.0, "step": 7579 }, { "epoch": 1.75, "learning_rate": 2.080348201217164e-05, "loss": 0.0, "step": 7580 }, { "epoch": 1.75, "learning_rate": 2.0799630228795934e-05, "loss": 0.0, "step": 7581 }, { "epoch": 1.75, "learning_rate": 2.079577844542023e-05, "loss": 0.0, "step": 7582 }, { "epoch": 1.75, "learning_rate": 2.0791926662044527e-05, "loss": 0.0, "step": 7583 }, { "epoch": 1.75, "learning_rate": 2.0788074878668825e-05, "loss": 0.0, "step": 7584 }, { "epoch": 1.75, "learning_rate": 2.078422309529312e-05, "loss": 0.0, "step": 7585 }, { "epoch": 1.75, "learning_rate": 2.078037131191742e-05, "loss": 0.0, "step": 7586 }, { "epoch": 1.75, "learning_rate": 2.0776519528541717e-05, "loss": 0.0, "step": 7587 }, { "epoch": 1.75, "learning_rate": 2.0772667745166012e-05, "loss": 0.0, "step": 7588 }, { "epoch": 1.75, "learning_rate": 2.076881596179031e-05, "loss": 0.0, "step": 7589 }, { "epoch": 1.75, "learning_rate": 2.0764964178414605e-05, "loss": 0.0, "step": 7590 }, { "epoch": 1.75, "learning_rate": 2.0761112395038904e-05, "loss": 0.0, "step": 7591 }, { "epoch": 1.75, "learning_rate": 2.0757260611663202e-05, "loss": 0.0, "step": 7592 }, { "epoch": 1.75, "learning_rate": 2.07534088282875e-05, "loss": 0.0, "step": 7593 }, { "epoch": 1.76, "learning_rate": 2.0749557044911795e-05, "loss": 0.0, "step": 7594 }, { "epoch": 1.76, "learning_rate": 2.074570526153609e-05, "loss": 0.0, "step": 7595 }, { "epoch": 1.76, "learning_rate": 2.074185347816039e-05, "loss": 0.0, "step": 7596 }, { "epoch": 1.76, "learning_rate": 2.0738001694784687e-05, "loss": 0.0, "step": 7597 }, { "epoch": 1.76, "learning_rate": 2.0734149911408985e-05, "loss": 0.0, "step": 7598 }, { "epoch": 1.76, "learning_rate": 2.073029812803328e-05, "loss": 0.0, "step": 7599 }, { "epoch": 1.76, "learning_rate": 2.0726446344657575e-05, "loss": 0.0, "step": 7600 }, { "epoch": 1.76, "learning_rate": 2.0722594561281874e-05, "loss": 0.0, "step": 7601 }, { "epoch": 1.76, "learning_rate": 2.0718742777906172e-05, "loss": 0.0, "step": 7602 }, { "epoch": 1.76, "learning_rate": 2.071489099453047e-05, "loss": 0.0, "step": 7603 }, { "epoch": 1.76, "learning_rate": 2.0711039211154765e-05, "loss": 0.0, "step": 7604 }, { "epoch": 1.76, "learning_rate": 2.0707187427779064e-05, "loss": 0.0, "step": 7605 }, { "epoch": 1.76, "learning_rate": 2.070333564440336e-05, "loss": 0.0, "step": 7606 }, { "epoch": 1.76, "learning_rate": 2.0699483861027657e-05, "loss": 0.0, "step": 7607 }, { "epoch": 1.76, "learning_rate": 2.0695632077651955e-05, "loss": 0.0, "step": 7608 }, { "epoch": 1.76, "learning_rate": 2.069178029427625e-05, "loss": 0.0, "step": 7609 }, { "epoch": 1.76, "learning_rate": 2.068792851090055e-05, "loss": 0.0, "step": 7610 }, { "epoch": 1.76, "learning_rate": 2.0684076727524847e-05, "loss": 0.0, "step": 7611 }, { "epoch": 1.76, "learning_rate": 2.0680224944149142e-05, "loss": 0.0, "step": 7612 }, { "epoch": 1.76, "learning_rate": 2.0676373160773437e-05, "loss": 0.0002, "step": 7613 }, { "epoch": 1.76, "learning_rate": 2.0672521377397735e-05, "loss": 0.0, "step": 7614 }, { "epoch": 1.76, "learning_rate": 2.0668669594022033e-05, "loss": 0.0, "step": 7615 }, { "epoch": 1.76, "learning_rate": 2.0664817810646332e-05, "loss": 0.0, "step": 7616 }, { "epoch": 1.76, "learning_rate": 2.066096602727063e-05, "loss": 0.0, "step": 7617 }, { "epoch": 1.76, "learning_rate": 2.0657114243894922e-05, "loss": 0.0, "step": 7618 }, { "epoch": 1.76, "learning_rate": 2.065326246051922e-05, "loss": 0.0, "step": 7619 }, { "epoch": 1.76, "learning_rate": 2.064941067714352e-05, "loss": 0.0, "step": 7620 }, { "epoch": 1.76, "learning_rate": 2.0645558893767817e-05, "loss": 0.0, "step": 7621 }, { "epoch": 1.76, "learning_rate": 2.0641707110392115e-05, "loss": 0.0, "step": 7622 }, { "epoch": 1.76, "learning_rate": 2.063785532701641e-05, "loss": 0.0, "step": 7623 }, { "epoch": 1.76, "learning_rate": 2.0634003543640705e-05, "loss": 0.0, "step": 7624 }, { "epoch": 1.76, "learning_rate": 2.0630151760265003e-05, "loss": 0.0, "step": 7625 }, { "epoch": 1.76, "learning_rate": 2.0626299976889302e-05, "loss": 0.0, "step": 7626 }, { "epoch": 1.76, "learning_rate": 2.0622448193513597e-05, "loss": 0.0, "step": 7627 }, { "epoch": 1.76, "learning_rate": 2.0618596410137895e-05, "loss": 0.0, "step": 7628 }, { "epoch": 1.76, "learning_rate": 2.0614744626762193e-05, "loss": 0.0, "step": 7629 }, { "epoch": 1.76, "learning_rate": 2.0610892843386488e-05, "loss": 0.0, "step": 7630 }, { "epoch": 1.76, "learning_rate": 2.0607041060010787e-05, "loss": 0.0, "step": 7631 }, { "epoch": 1.76, "learning_rate": 2.060318927663508e-05, "loss": 0.0001, "step": 7632 }, { "epoch": 1.76, "learning_rate": 2.059933749325938e-05, "loss": 0.0, "step": 7633 }, { "epoch": 1.76, "learning_rate": 2.0595485709883678e-05, "loss": 0.0, "step": 7634 }, { "epoch": 1.76, "learning_rate": 2.0591633926507977e-05, "loss": 0.0, "step": 7635 }, { "epoch": 1.76, "learning_rate": 2.058778214313227e-05, "loss": 0.0, "step": 7636 }, { "epoch": 1.76, "learning_rate": 2.0583930359756567e-05, "loss": 0.0, "step": 7637 }, { "epoch": 1.77, "learning_rate": 2.0580078576380865e-05, "loss": 0.0, "step": 7638 }, { "epoch": 1.77, "learning_rate": 2.0576226793005163e-05, "loss": 0.0, "step": 7639 }, { "epoch": 1.77, "learning_rate": 2.057237500962946e-05, "loss": 0.0, "step": 7640 }, { "epoch": 1.77, "learning_rate": 2.0568523226253757e-05, "loss": 0.0, "step": 7641 }, { "epoch": 1.77, "learning_rate": 2.056467144287805e-05, "loss": 0.0, "step": 7642 }, { "epoch": 1.77, "learning_rate": 2.056081965950235e-05, "loss": 0.0, "step": 7643 }, { "epoch": 1.77, "learning_rate": 2.0556967876126648e-05, "loss": 0.0, "step": 7644 }, { "epoch": 1.77, "learning_rate": 2.0553116092750947e-05, "loss": 0.0, "step": 7645 }, { "epoch": 1.77, "learning_rate": 2.054926430937524e-05, "loss": 0.0, "step": 7646 }, { "epoch": 1.77, "learning_rate": 2.054541252599954e-05, "loss": 0.0, "step": 7647 }, { "epoch": 1.77, "learning_rate": 2.0541560742623835e-05, "loss": 0.0, "step": 7648 }, { "epoch": 1.77, "learning_rate": 2.0537708959248133e-05, "loss": 0.0, "step": 7649 }, { "epoch": 1.77, "learning_rate": 2.053385717587243e-05, "loss": 0.0, "step": 7650 }, { "epoch": 1.77, "learning_rate": 2.0530005392496726e-05, "loss": 0.0, "step": 7651 }, { "epoch": 1.77, "learning_rate": 2.0526153609121025e-05, "loss": 0.0, "step": 7652 }, { "epoch": 1.77, "learning_rate": 2.0522301825745323e-05, "loss": 0.0, "step": 7653 }, { "epoch": 1.77, "learning_rate": 2.0518450042369618e-05, "loss": 0.0, "step": 7654 }, { "epoch": 1.77, "learning_rate": 2.0514598258993913e-05, "loss": 0.0, "step": 7655 }, { "epoch": 1.77, "learning_rate": 2.051074647561821e-05, "loss": 0.0, "step": 7656 }, { "epoch": 1.77, "learning_rate": 2.050689469224251e-05, "loss": 0.0, "step": 7657 }, { "epoch": 1.77, "learning_rate": 2.0503042908866808e-05, "loss": 0.0, "step": 7658 }, { "epoch": 1.77, "learning_rate": 2.0499191125491103e-05, "loss": 0.0, "step": 7659 }, { "epoch": 1.77, "learning_rate": 2.0495339342115398e-05, "loss": 0.0, "step": 7660 }, { "epoch": 1.77, "learning_rate": 2.0491487558739696e-05, "loss": 0.0, "step": 7661 }, { "epoch": 1.77, "learning_rate": 2.0487635775363995e-05, "loss": 0.0, "step": 7662 }, { "epoch": 1.77, "learning_rate": 2.0483783991988293e-05, "loss": 0.0, "step": 7663 }, { "epoch": 1.77, "learning_rate": 2.047993220861259e-05, "loss": 0.0, "step": 7664 }, { "epoch": 1.77, "learning_rate": 2.0476080425236886e-05, "loss": 0.0, "step": 7665 }, { "epoch": 1.77, "learning_rate": 2.047222864186118e-05, "loss": 0.0, "step": 7666 }, { "epoch": 1.77, "learning_rate": 2.046837685848548e-05, "loss": 0.0, "step": 7667 }, { "epoch": 1.77, "learning_rate": 2.0464525075109778e-05, "loss": 0.0, "step": 7668 }, { "epoch": 1.77, "learning_rate": 2.0460673291734073e-05, "loss": 0.0, "step": 7669 }, { "epoch": 1.77, "learning_rate": 2.045682150835837e-05, "loss": 0.0, "step": 7670 }, { "epoch": 1.77, "learning_rate": 2.045296972498267e-05, "loss": 0.0, "step": 7671 }, { "epoch": 1.77, "learning_rate": 2.0449117941606965e-05, "loss": 0.0, "step": 7672 }, { "epoch": 1.77, "learning_rate": 2.0445266158231263e-05, "loss": 0.0, "step": 7673 }, { "epoch": 1.77, "learning_rate": 2.0441414374855558e-05, "loss": 0.0, "step": 7674 }, { "epoch": 1.77, "learning_rate": 2.0437562591479856e-05, "loss": 0.0, "step": 7675 }, { "epoch": 1.77, "learning_rate": 2.0433710808104155e-05, "loss": 0.0, "step": 7676 }, { "epoch": 1.77, "learning_rate": 2.042985902472845e-05, "loss": 0.0, "step": 7677 }, { "epoch": 1.77, "learning_rate": 2.0426007241352748e-05, "loss": 0.0, "step": 7678 }, { "epoch": 1.77, "learning_rate": 2.0422155457977043e-05, "loss": 0.0, "step": 7679 }, { "epoch": 1.77, "learning_rate": 2.041830367460134e-05, "loss": 0.0, "step": 7680 }, { "epoch": 1.78, "learning_rate": 2.041445189122564e-05, "loss": 0.0, "step": 7681 }, { "epoch": 1.78, "learning_rate": 2.0410600107849938e-05, "loss": 0.0, "step": 7682 }, { "epoch": 1.78, "learning_rate": 2.0406748324474233e-05, "loss": 0.0, "step": 7683 }, { "epoch": 1.78, "learning_rate": 2.0402896541098528e-05, "loss": 0.0, "step": 7684 }, { "epoch": 1.78, "learning_rate": 2.0399044757722826e-05, "loss": 0.0002, "step": 7685 }, { "epoch": 1.78, "learning_rate": 2.0395192974347124e-05, "loss": 0.0, "step": 7686 }, { "epoch": 1.78, "learning_rate": 2.0391341190971423e-05, "loss": 0.0, "step": 7687 }, { "epoch": 1.78, "learning_rate": 2.0387489407595718e-05, "loss": 0.0, "step": 7688 }, { "epoch": 1.78, "learning_rate": 2.0383637624220016e-05, "loss": 0.0, "step": 7689 }, { "epoch": 1.78, "learning_rate": 2.037978584084431e-05, "loss": 0.0, "step": 7690 }, { "epoch": 1.78, "learning_rate": 2.037593405746861e-05, "loss": 0.0, "step": 7691 }, { "epoch": 1.78, "learning_rate": 2.0372082274092908e-05, "loss": 0.0, "step": 7692 }, { "epoch": 1.78, "learning_rate": 2.0368230490717203e-05, "loss": 0.0, "step": 7693 }, { "epoch": 1.78, "learning_rate": 2.03643787073415e-05, "loss": 0.0, "step": 7694 }, { "epoch": 1.78, "learning_rate": 2.0360526923965796e-05, "loss": 0.0, "step": 7695 }, { "epoch": 1.78, "learning_rate": 2.0356675140590094e-05, "loss": 0.0, "step": 7696 }, { "epoch": 1.78, "learning_rate": 2.035282335721439e-05, "loss": 0.0, "step": 7697 }, { "epoch": 1.78, "learning_rate": 2.0348971573838688e-05, "loss": 0.0, "step": 7698 }, { "epoch": 1.78, "learning_rate": 2.0345119790462986e-05, "loss": 0.0, "step": 7699 }, { "epoch": 1.78, "learning_rate": 2.0341268007087284e-05, "loss": 0.0, "step": 7700 }, { "epoch": 1.78, "learning_rate": 2.033741622371158e-05, "loss": 0.0, "step": 7701 }, { "epoch": 1.78, "learning_rate": 2.0333564440335874e-05, "loss": 0.0, "step": 7702 }, { "epoch": 1.78, "learning_rate": 2.0329712656960173e-05, "loss": 0.0, "step": 7703 }, { "epoch": 1.78, "learning_rate": 2.032586087358447e-05, "loss": 0.0, "step": 7704 }, { "epoch": 1.78, "learning_rate": 2.032200909020877e-05, "loss": 0.0, "step": 7705 }, { "epoch": 1.78, "learning_rate": 2.0318157306833068e-05, "loss": 0.0, "step": 7706 }, { "epoch": 1.78, "learning_rate": 2.0314305523457363e-05, "loss": 0.0, "step": 7707 }, { "epoch": 1.78, "learning_rate": 2.0310453740081657e-05, "loss": 0.0, "step": 7708 }, { "epoch": 1.78, "learning_rate": 2.0306601956705956e-05, "loss": 0.0, "step": 7709 }, { "epoch": 1.78, "learning_rate": 2.0302750173330254e-05, "loss": 0.0, "step": 7710 }, { "epoch": 1.78, "learning_rate": 2.029889838995455e-05, "loss": 0.0, "step": 7711 }, { "epoch": 1.78, "learning_rate": 2.0295046606578847e-05, "loss": 0.0, "step": 7712 }, { "epoch": 1.78, "learning_rate": 2.0291194823203142e-05, "loss": 0.0, "step": 7713 }, { "epoch": 1.78, "learning_rate": 2.028734303982744e-05, "loss": 0.0, "step": 7714 }, { "epoch": 1.78, "learning_rate": 2.028349125645174e-05, "loss": 0.0, "step": 7715 }, { "epoch": 1.78, "learning_rate": 2.0279639473076034e-05, "loss": 0.0, "step": 7716 }, { "epoch": 1.78, "learning_rate": 2.0275787689700332e-05, "loss": 0.0, "step": 7717 }, { "epoch": 1.78, "learning_rate": 2.027193590632463e-05, "loss": 0.0, "step": 7718 }, { "epoch": 1.78, "learning_rate": 2.0268084122948926e-05, "loss": 0.0, "step": 7719 }, { "epoch": 1.78, "learning_rate": 2.0264232339573224e-05, "loss": 0.0, "step": 7720 }, { "epoch": 1.78, "learning_rate": 2.026038055619752e-05, "loss": 0.0, "step": 7721 }, { "epoch": 1.78, "learning_rate": 2.0256528772821817e-05, "loss": 0.0, "step": 7722 }, { "epoch": 1.78, "learning_rate": 2.0252676989446116e-05, "loss": 0.0001, "step": 7723 }, { "epoch": 1.79, "learning_rate": 2.0248825206070414e-05, "loss": 0.0, "step": 7724 }, { "epoch": 1.79, "learning_rate": 2.024497342269471e-05, "loss": 0.0, "step": 7725 }, { "epoch": 1.79, "learning_rate": 2.0241121639319004e-05, "loss": 0.0001, "step": 7726 }, { "epoch": 1.79, "learning_rate": 2.0237269855943302e-05, "loss": 0.0, "step": 7727 }, { "epoch": 1.79, "learning_rate": 2.02334180725676e-05, "loss": 0.0, "step": 7728 }, { "epoch": 1.79, "learning_rate": 2.02295662891919e-05, "loss": 0.0, "step": 7729 }, { "epoch": 1.79, "learning_rate": 2.0225714505816194e-05, "loss": 0.0, "step": 7730 }, { "epoch": 1.79, "learning_rate": 2.022186272244049e-05, "loss": 0.0, "step": 7731 }, { "epoch": 1.79, "learning_rate": 2.0218010939064787e-05, "loss": 0.0, "step": 7732 }, { "epoch": 1.79, "learning_rate": 2.0214159155689086e-05, "loss": 0.0, "step": 7733 }, { "epoch": 1.79, "learning_rate": 2.0210307372313384e-05, "loss": 0.0, "step": 7734 }, { "epoch": 1.79, "learning_rate": 2.020645558893768e-05, "loss": 0.0, "step": 7735 }, { "epoch": 1.79, "learning_rate": 2.0202603805561977e-05, "loss": 0.0, "step": 7736 }, { "epoch": 1.79, "learning_rate": 2.0198752022186272e-05, "loss": 0.0, "step": 7737 }, { "epoch": 1.79, "learning_rate": 2.019490023881057e-05, "loss": 0.0, "step": 7738 }, { "epoch": 1.79, "learning_rate": 2.0191048455434865e-05, "loss": 0.0, "step": 7739 }, { "epoch": 1.79, "learning_rate": 2.0187196672059164e-05, "loss": 0.0, "step": 7740 }, { "epoch": 1.79, "learning_rate": 2.0183344888683462e-05, "loss": 0.0, "step": 7741 }, { "epoch": 1.79, "learning_rate": 2.017949310530776e-05, "loss": 0.0, "step": 7742 }, { "epoch": 1.79, "learning_rate": 2.0175641321932055e-05, "loss": 0.0, "step": 7743 }, { "epoch": 1.79, "learning_rate": 2.017178953855635e-05, "loss": 0.0, "step": 7744 }, { "epoch": 1.79, "learning_rate": 2.016793775518065e-05, "loss": 0.0, "step": 7745 }, { "epoch": 1.79, "learning_rate": 2.0164085971804947e-05, "loss": 0.0002, "step": 7746 }, { "epoch": 1.79, "learning_rate": 2.0160234188429245e-05, "loss": 0.0, "step": 7747 }, { "epoch": 1.79, "learning_rate": 2.0156382405053544e-05, "loss": 0.0, "step": 7748 }, { "epoch": 1.79, "learning_rate": 2.0152530621677835e-05, "loss": 0.0, "step": 7749 }, { "epoch": 1.79, "learning_rate": 2.0148678838302134e-05, "loss": 0.0, "step": 7750 }, { "epoch": 1.79, "learning_rate": 2.0144827054926432e-05, "loss": 0.0003, "step": 7751 }, { "epoch": 1.79, "learning_rate": 2.014097527155073e-05, "loss": 0.0, "step": 7752 }, { "epoch": 1.79, "learning_rate": 2.0137123488175025e-05, "loss": 0.0, "step": 7753 }, { "epoch": 1.79, "learning_rate": 2.0133271704799324e-05, "loss": 0.0, "step": 7754 }, { "epoch": 1.79, "learning_rate": 2.012941992142362e-05, "loss": 0.0, "step": 7755 }, { "epoch": 1.79, "learning_rate": 2.0125568138047917e-05, "loss": 0.0, "step": 7756 }, { "epoch": 1.79, "learning_rate": 2.0121716354672215e-05, "loss": 0.0, "step": 7757 }, { "epoch": 1.79, "learning_rate": 2.011786457129651e-05, "loss": 0.0, "step": 7758 }, { "epoch": 1.79, "learning_rate": 2.011401278792081e-05, "loss": 0.0, "step": 7759 }, { "epoch": 1.79, "learning_rate": 2.0110161004545107e-05, "loss": 0.0, "step": 7760 }, { "epoch": 1.79, "learning_rate": 2.0106309221169402e-05, "loss": 0.0, "step": 7761 }, { "epoch": 1.79, "learning_rate": 2.01024574377937e-05, "loss": 0.0, "step": 7762 }, { "epoch": 1.79, "learning_rate": 2.0098605654417995e-05, "loss": 0.0, "step": 7763 }, { "epoch": 1.79, "learning_rate": 2.0094753871042294e-05, "loss": 0.0, "step": 7764 }, { "epoch": 1.79, "learning_rate": 2.0090902087666592e-05, "loss": 0.0, "step": 7765 }, { "epoch": 1.79, "learning_rate": 2.008705030429089e-05, "loss": 0.0, "step": 7766 }, { "epoch": 1.8, "learning_rate": 2.0083198520915185e-05, "loss": 0.0, "step": 7767 }, { "epoch": 1.8, "learning_rate": 2.007934673753948e-05, "loss": 0.0, "step": 7768 }, { "epoch": 1.8, "learning_rate": 2.007549495416378e-05, "loss": 0.0001, "step": 7769 }, { "epoch": 1.8, "learning_rate": 2.0071643170788077e-05, "loss": 0.0, "step": 7770 }, { "epoch": 1.8, "learning_rate": 2.0067791387412375e-05, "loss": 0.0, "step": 7771 }, { "epoch": 1.8, "learning_rate": 2.006393960403667e-05, "loss": 0.0, "step": 7772 }, { "epoch": 1.8, "learning_rate": 2.0060087820660965e-05, "loss": 0.0, "step": 7773 }, { "epoch": 1.8, "learning_rate": 2.0056236037285263e-05, "loss": 0.0, "step": 7774 }, { "epoch": 1.8, "learning_rate": 2.0052384253909562e-05, "loss": 0.0, "step": 7775 }, { "epoch": 1.8, "learning_rate": 2.004853247053386e-05, "loss": 0.0, "step": 7776 }, { "epoch": 1.8, "learning_rate": 2.0044680687158155e-05, "loss": 0.0, "step": 7777 }, { "epoch": 1.8, "learning_rate": 2.0040828903782453e-05, "loss": 0.0, "step": 7778 }, { "epoch": 1.8, "learning_rate": 2.003697712040675e-05, "loss": 0.0, "step": 7779 }, { "epoch": 1.8, "learning_rate": 2.0033125337031047e-05, "loss": 0.0, "step": 7780 }, { "epoch": 1.8, "learning_rate": 2.002927355365534e-05, "loss": 0.0, "step": 7781 }, { "epoch": 1.8, "learning_rate": 2.002542177027964e-05, "loss": 0.0, "step": 7782 }, { "epoch": 1.8, "learning_rate": 2.002156998690394e-05, "loss": 0.0, "step": 7783 }, { "epoch": 1.8, "learning_rate": 2.0017718203528237e-05, "loss": 0.0, "step": 7784 }, { "epoch": 1.8, "learning_rate": 2.001386642015253e-05, "loss": 0.0, "step": 7785 }, { "epoch": 1.8, "learning_rate": 2.0010014636776827e-05, "loss": 0.0, "step": 7786 }, { "epoch": 1.8, "learning_rate": 2.0006162853401125e-05, "loss": 0.0, "step": 7787 }, { "epoch": 1.8, "learning_rate": 2.0002311070025423e-05, "loss": 0.0, "step": 7788 }, { "epoch": 1.8, "learning_rate": 1.999845928664972e-05, "loss": 0.0, "step": 7789 }, { "epoch": 1.8, "learning_rate": 1.9994607503274017e-05, "loss": 0.0, "step": 7790 }, { "epoch": 1.8, "learning_rate": 1.999075571989831e-05, "loss": 0.0, "step": 7791 }, { "epoch": 1.8, "learning_rate": 1.998690393652261e-05, "loss": 0.0, "step": 7792 }, { "epoch": 1.8, "learning_rate": 1.9983052153146908e-05, "loss": 0.0, "step": 7793 }, { "epoch": 1.8, "learning_rate": 1.9979200369771207e-05, "loss": 0.0, "step": 7794 }, { "epoch": 1.8, "learning_rate": 1.99753485863955e-05, "loss": 0.0, "step": 7795 }, { "epoch": 1.8, "learning_rate": 1.99714968030198e-05, "loss": 0.0, "step": 7796 }, { "epoch": 1.8, "learning_rate": 1.9967645019644095e-05, "loss": 0.0, "step": 7797 }, { "epoch": 1.8, "learning_rate": 1.9963793236268393e-05, "loss": 0.001, "step": 7798 }, { "epoch": 1.8, "learning_rate": 1.995994145289269e-05, "loss": 0.0, "step": 7799 }, { "epoch": 1.8, "learning_rate": 1.9956089669516986e-05, "loss": 0.0, "step": 7800 }, { "epoch": 1.8, "learning_rate": 1.9952237886141285e-05, "loss": 0.0, "step": 7801 }, { "epoch": 1.8, "learning_rate": 1.9948386102765583e-05, "loss": 0.0, "step": 7802 }, { "epoch": 1.8, "learning_rate": 1.9944534319389878e-05, "loss": 0.0, "step": 7803 }, { "epoch": 1.8, "learning_rate": 1.9940682536014173e-05, "loss": 0.0, "step": 7804 }, { "epoch": 1.8, "learning_rate": 1.993683075263847e-05, "loss": 0.0, "step": 7805 }, { "epoch": 1.8, "learning_rate": 1.993297896926277e-05, "loss": 0.0001, "step": 7806 }, { "epoch": 1.8, "learning_rate": 1.9929127185887068e-05, "loss": 0.0, "step": 7807 }, { "epoch": 1.8, "learning_rate": 1.9925275402511363e-05, "loss": 0.0, "step": 7808 }, { "epoch": 1.8, "learning_rate": 1.9921423619135658e-05, "loss": 0.0, "step": 7809 }, { "epoch": 1.8, "learning_rate": 1.9917571835759956e-05, "loss": 0.0, "step": 7810 }, { "epoch": 1.81, "learning_rate": 1.9913720052384255e-05, "loss": 0.0, "step": 7811 }, { "epoch": 1.81, "learning_rate": 1.9909868269008553e-05, "loss": 0.0, "step": 7812 }, { "epoch": 1.81, "learning_rate": 1.990601648563285e-05, "loss": 0.0, "step": 7813 }, { "epoch": 1.81, "learning_rate": 1.9902164702257146e-05, "loss": 0.0, "step": 7814 }, { "epoch": 1.81, "learning_rate": 1.989831291888144e-05, "loss": 0.0, "step": 7815 }, { "epoch": 1.81, "learning_rate": 1.989446113550574e-05, "loss": 0.0, "step": 7816 }, { "epoch": 1.81, "learning_rate": 1.9890609352130038e-05, "loss": 0.0, "step": 7817 }, { "epoch": 1.81, "learning_rate": 1.9886757568754333e-05, "loss": 0.0, "step": 7818 }, { "epoch": 1.81, "learning_rate": 1.988290578537863e-05, "loss": 0.0, "step": 7819 }, { "epoch": 1.81, "learning_rate": 1.987905400200293e-05, "loss": 0.0, "step": 7820 }, { "epoch": 1.81, "learning_rate": 1.9875202218627225e-05, "loss": 0.0, "step": 7821 }, { "epoch": 1.81, "learning_rate": 1.9871350435251523e-05, "loss": 0.0, "step": 7822 }, { "epoch": 1.81, "learning_rate": 1.9867498651875818e-05, "loss": 0.0, "step": 7823 }, { "epoch": 1.81, "learning_rate": 1.9863646868500116e-05, "loss": 0.0, "step": 7824 }, { "epoch": 1.81, "learning_rate": 1.9859795085124415e-05, "loss": 0.0, "step": 7825 }, { "epoch": 1.81, "learning_rate": 1.985594330174871e-05, "loss": 0.0, "step": 7826 }, { "epoch": 1.81, "learning_rate": 1.9852091518373008e-05, "loss": 0.0, "step": 7827 }, { "epoch": 1.81, "learning_rate": 1.9848239734997303e-05, "loss": 0.0, "step": 7828 }, { "epoch": 1.81, "learning_rate": 1.98443879516216e-05, "loss": 0.0, "step": 7829 }, { "epoch": 1.81, "learning_rate": 1.98405361682459e-05, "loss": 0.0, "step": 7830 }, { "epoch": 1.81, "learning_rate": 1.9836684384870198e-05, "loss": 0.0, "step": 7831 }, { "epoch": 1.81, "learning_rate": 1.9832832601494493e-05, "loss": 0.0, "step": 7832 }, { "epoch": 1.81, "learning_rate": 1.9828980818118788e-05, "loss": 0.0, "step": 7833 }, { "epoch": 1.81, "learning_rate": 1.9825129034743086e-05, "loss": 0.0, "step": 7834 }, { "epoch": 1.81, "learning_rate": 1.9821277251367384e-05, "loss": 0.0, "step": 7835 }, { "epoch": 1.81, "learning_rate": 1.9817425467991683e-05, "loss": 0.0, "step": 7836 }, { "epoch": 1.81, "learning_rate": 1.9813573684615978e-05, "loss": 0.0, "step": 7837 }, { "epoch": 1.81, "learning_rate": 1.9809721901240276e-05, "loss": 0.0, "step": 7838 }, { "epoch": 1.81, "learning_rate": 1.980587011786457e-05, "loss": 0.0, "step": 7839 }, { "epoch": 1.81, "learning_rate": 1.980201833448887e-05, "loss": 0.0, "step": 7840 }, { "epoch": 1.81, "learning_rate": 1.9798166551113168e-05, "loss": 0.0, "step": 7841 }, { "epoch": 1.81, "learning_rate": 1.9794314767737463e-05, "loss": 0.0, "step": 7842 }, { "epoch": 1.81, "learning_rate": 1.979046298436176e-05, "loss": 0.0, "step": 7843 }, { "epoch": 1.81, "learning_rate": 1.9786611200986056e-05, "loss": 0.0, "step": 7844 }, { "epoch": 1.81, "learning_rate": 1.9782759417610354e-05, "loss": 0.0, "step": 7845 }, { "epoch": 1.81, "learning_rate": 1.977890763423465e-05, "loss": 0.0, "step": 7846 }, { "epoch": 1.81, "learning_rate": 1.9775055850858948e-05, "loss": 0.0, "step": 7847 }, { "epoch": 1.81, "learning_rate": 1.9771204067483246e-05, "loss": 0.0, "step": 7848 }, { "epoch": 1.81, "learning_rate": 1.9767352284107544e-05, "loss": 0.0, "step": 7849 }, { "epoch": 1.81, "learning_rate": 1.976350050073184e-05, "loss": 0.0, "step": 7850 }, { "epoch": 1.81, "learning_rate": 1.9759648717356134e-05, "loss": 0.0, "step": 7851 }, { "epoch": 1.81, "learning_rate": 1.9755796933980433e-05, "loss": 0.0, "step": 7852 }, { "epoch": 1.81, "learning_rate": 1.975194515060473e-05, "loss": 0.0, "step": 7853 }, { "epoch": 1.82, "learning_rate": 1.974809336722903e-05, "loss": 0.0, "step": 7854 }, { "epoch": 1.82, "learning_rate": 1.9744241583853328e-05, "loss": 0.0, "step": 7855 }, { "epoch": 1.82, "learning_rate": 1.9740389800477623e-05, "loss": 0.0, "step": 7856 }, { "epoch": 1.82, "learning_rate": 1.9736538017101918e-05, "loss": 0.0, "step": 7857 }, { "epoch": 1.82, "learning_rate": 1.9732686233726216e-05, "loss": 0.0, "step": 7858 }, { "epoch": 1.82, "learning_rate": 1.9728834450350514e-05, "loss": 0.0, "step": 7859 }, { "epoch": 1.82, "learning_rate": 1.972498266697481e-05, "loss": 0.0, "step": 7860 }, { "epoch": 1.82, "learning_rate": 1.9721130883599108e-05, "loss": 0.0, "step": 7861 }, { "epoch": 1.82, "learning_rate": 1.9717279100223406e-05, "loss": 0.0, "step": 7862 }, { "epoch": 1.82, "learning_rate": 1.97134273168477e-05, "loss": 0.0294, "step": 7863 }, { "epoch": 1.82, "learning_rate": 1.9709575533472e-05, "loss": 0.0, "step": 7864 }, { "epoch": 1.82, "learning_rate": 1.9705723750096294e-05, "loss": 0.0, "step": 7865 }, { "epoch": 1.82, "learning_rate": 1.9701871966720592e-05, "loss": 0.0, "step": 7866 }, { "epoch": 1.82, "learning_rate": 1.969802018334489e-05, "loss": 0.0, "step": 7867 }, { "epoch": 1.82, "learning_rate": 1.9694168399969186e-05, "loss": 0.0, "step": 7868 }, { "epoch": 1.82, "learning_rate": 1.9690316616593484e-05, "loss": 0.0, "step": 7869 }, { "epoch": 1.82, "learning_rate": 1.968646483321778e-05, "loss": 0.0, "step": 7870 }, { "epoch": 1.82, "learning_rate": 1.9682613049842077e-05, "loss": 0.0, "step": 7871 }, { "epoch": 1.82, "learning_rate": 1.9678761266466376e-05, "loss": 0.0, "step": 7872 }, { "epoch": 1.82, "learning_rate": 1.9674909483090674e-05, "loss": 0.0, "step": 7873 }, { "epoch": 1.82, "learning_rate": 1.967105769971497e-05, "loss": 0.0, "step": 7874 }, { "epoch": 1.82, "learning_rate": 1.9667205916339264e-05, "loss": 0.0, "step": 7875 }, { "epoch": 1.82, "learning_rate": 1.9663354132963562e-05, "loss": 0.0, "step": 7876 }, { "epoch": 1.82, "learning_rate": 1.965950234958786e-05, "loss": 0.0, "step": 7877 }, { "epoch": 1.82, "learning_rate": 1.965565056621216e-05, "loss": 0.0, "step": 7878 }, { "epoch": 1.82, "learning_rate": 1.9651798782836454e-05, "loss": 0.0, "step": 7879 }, { "epoch": 1.82, "learning_rate": 1.9647946999460752e-05, "loss": 0.0, "step": 7880 }, { "epoch": 1.82, "learning_rate": 1.9644095216085047e-05, "loss": 0.0, "step": 7881 }, { "epoch": 1.82, "learning_rate": 1.9640243432709346e-05, "loss": 0.0, "step": 7882 }, { "epoch": 1.82, "learning_rate": 1.9636391649333644e-05, "loss": 0.0, "step": 7883 }, { "epoch": 1.82, "learning_rate": 1.963253986595794e-05, "loss": 0.0, "step": 7884 }, { "epoch": 1.82, "learning_rate": 1.9628688082582237e-05, "loss": 0.0, "step": 7885 }, { "epoch": 1.82, "learning_rate": 1.9624836299206532e-05, "loss": 0.0, "step": 7886 }, { "epoch": 1.82, "learning_rate": 1.962098451583083e-05, "loss": 0.0, "step": 7887 }, { "epoch": 1.82, "learning_rate": 1.9617132732455126e-05, "loss": 0.0, "step": 7888 }, { "epoch": 1.82, "learning_rate": 1.9613280949079424e-05, "loss": 0.0, "step": 7889 }, { "epoch": 1.82, "learning_rate": 1.9609429165703722e-05, "loss": 0.0, "step": 7890 }, { "epoch": 1.82, "learning_rate": 1.960557738232802e-05, "loss": 0.0, "step": 7891 }, { "epoch": 1.82, "learning_rate": 1.9601725598952316e-05, "loss": 0.0, "step": 7892 }, { "epoch": 1.82, "learning_rate": 1.959787381557661e-05, "loss": 0.0, "step": 7893 }, { "epoch": 1.82, "learning_rate": 1.959402203220091e-05, "loss": 0.0, "step": 7894 }, { "epoch": 1.82, "learning_rate": 1.9590170248825207e-05, "loss": 0.0019, "step": 7895 }, { "epoch": 1.82, "learning_rate": 1.9586318465449506e-05, "loss": 0.0, "step": 7896 }, { "epoch": 1.83, "learning_rate": 1.9582466682073804e-05, "loss": 0.0, "step": 7897 }, { "epoch": 1.83, "learning_rate": 1.95786148986981e-05, "loss": 0.0, "step": 7898 }, { "epoch": 1.83, "learning_rate": 1.9574763115322394e-05, "loss": 0.0, "step": 7899 }, { "epoch": 1.83, "learning_rate": 1.9570911331946692e-05, "loss": 0.0, "step": 7900 }, { "epoch": 1.83, "learning_rate": 1.956705954857099e-05, "loss": 0.0, "step": 7901 }, { "epoch": 1.83, "learning_rate": 1.9563207765195285e-05, "loss": 0.0, "step": 7902 }, { "epoch": 1.83, "learning_rate": 1.9559355981819584e-05, "loss": 0.0, "step": 7903 }, { "epoch": 1.83, "learning_rate": 1.955550419844388e-05, "loss": 0.0, "step": 7904 }, { "epoch": 1.83, "learning_rate": 1.9551652415068177e-05, "loss": 0.0, "step": 7905 }, { "epoch": 1.83, "learning_rate": 1.9547800631692475e-05, "loss": 0.0, "step": 7906 }, { "epoch": 1.83, "learning_rate": 1.954394884831677e-05, "loss": 0.0, "step": 7907 }, { "epoch": 1.83, "learning_rate": 1.954009706494107e-05, "loss": 0.0, "step": 7908 }, { "epoch": 1.83, "learning_rate": 1.9536245281565367e-05, "loss": 0.0, "step": 7909 }, { "epoch": 1.83, "learning_rate": 1.9532393498189662e-05, "loss": 0.0, "step": 7910 }, { "epoch": 1.83, "learning_rate": 1.952854171481396e-05, "loss": 0.0, "step": 7911 }, { "epoch": 1.83, "learning_rate": 1.9524689931438255e-05, "loss": 0.0, "step": 7912 }, { "epoch": 1.83, "learning_rate": 1.9520838148062554e-05, "loss": 0.0001, "step": 7913 }, { "epoch": 1.83, "learning_rate": 1.9516986364686852e-05, "loss": 0.0, "step": 7914 }, { "epoch": 1.83, "learning_rate": 1.951313458131115e-05, "loss": 0.0, "step": 7915 }, { "epoch": 1.83, "learning_rate": 1.9509282797935445e-05, "loss": 0.0, "step": 7916 }, { "epoch": 1.83, "learning_rate": 1.950543101455974e-05, "loss": 0.0, "step": 7917 }, { "epoch": 1.83, "learning_rate": 1.950157923118404e-05, "loss": 0.0002, "step": 7918 }, { "epoch": 1.83, "learning_rate": 1.9497727447808337e-05, "loss": 0.0, "step": 7919 }, { "epoch": 1.83, "learning_rate": 1.9493875664432635e-05, "loss": 0.0, "step": 7920 }, { "epoch": 1.83, "learning_rate": 1.949002388105693e-05, "loss": 0.0, "step": 7921 }, { "epoch": 1.83, "learning_rate": 1.9486172097681225e-05, "loss": 0.0, "step": 7922 }, { "epoch": 1.83, "learning_rate": 1.9482320314305524e-05, "loss": 0.0, "step": 7923 }, { "epoch": 1.83, "learning_rate": 1.9478468530929822e-05, "loss": 0.0, "step": 7924 }, { "epoch": 1.83, "learning_rate": 1.947461674755412e-05, "loss": 0.0, "step": 7925 }, { "epoch": 1.83, "learning_rate": 1.9470764964178415e-05, "loss": 0.0, "step": 7926 }, { "epoch": 1.83, "learning_rate": 1.9466913180802714e-05, "loss": 0.0, "step": 7927 }, { "epoch": 1.83, "learning_rate": 1.946306139742701e-05, "loss": 0.0, "step": 7928 }, { "epoch": 1.83, "learning_rate": 1.9459209614051307e-05, "loss": 0.0, "step": 7929 }, { "epoch": 1.83, "learning_rate": 1.9455357830675602e-05, "loss": 0.0, "step": 7930 }, { "epoch": 1.83, "learning_rate": 1.94515060472999e-05, "loss": 0.0, "step": 7931 }, { "epoch": 1.83, "learning_rate": 1.94476542639242e-05, "loss": 0.0, "step": 7932 }, { "epoch": 1.83, "learning_rate": 1.9443802480548497e-05, "loss": 0.0, "step": 7933 }, { "epoch": 1.83, "learning_rate": 1.9439950697172792e-05, "loss": 0.0, "step": 7934 }, { "epoch": 1.83, "learning_rate": 1.9436098913797087e-05, "loss": 0.0, "step": 7935 }, { "epoch": 1.83, "learning_rate": 1.9432247130421385e-05, "loss": 0.0, "step": 7936 }, { "epoch": 1.83, "learning_rate": 1.9428395347045683e-05, "loss": 0.0, "step": 7937 }, { "epoch": 1.83, "learning_rate": 1.9424543563669982e-05, "loss": 0.0, "step": 7938 }, { "epoch": 1.83, "learning_rate": 1.9420691780294277e-05, "loss": 0.0, "step": 7939 }, { "epoch": 1.83, "learning_rate": 1.941683999691857e-05, "loss": 0.0, "step": 7940 }, { "epoch": 1.84, "learning_rate": 1.941298821354287e-05, "loss": 0.0, "step": 7941 }, { "epoch": 1.84, "learning_rate": 1.940913643016717e-05, "loss": 0.0, "step": 7942 }, { "epoch": 1.84, "learning_rate": 1.9405284646791467e-05, "loss": 0.0, "step": 7943 }, { "epoch": 1.84, "learning_rate": 1.940143286341576e-05, "loss": 0.0, "step": 7944 }, { "epoch": 1.84, "learning_rate": 1.939758108004006e-05, "loss": 0.0, "step": 7945 }, { "epoch": 1.84, "learning_rate": 1.9393729296664355e-05, "loss": 0.0, "step": 7946 }, { "epoch": 1.84, "learning_rate": 1.9389877513288653e-05, "loss": 0.0, "step": 7947 }, { "epoch": 1.84, "learning_rate": 1.938602572991295e-05, "loss": 0.0, "step": 7948 }, { "epoch": 1.84, "learning_rate": 1.9382173946537247e-05, "loss": 0.0, "step": 7949 }, { "epoch": 1.84, "learning_rate": 1.9378322163161545e-05, "loss": 0.0, "step": 7950 }, { "epoch": 1.84, "learning_rate": 1.9374470379785843e-05, "loss": 0.0, "step": 7951 }, { "epoch": 1.84, "learning_rate": 1.9370618596410138e-05, "loss": 0.0, "step": 7952 }, { "epoch": 1.84, "learning_rate": 1.9366766813034437e-05, "loss": 0.0, "step": 7953 }, { "epoch": 1.84, "learning_rate": 1.936291502965873e-05, "loss": 0.0, "step": 7954 }, { "epoch": 1.84, "learning_rate": 1.935906324628303e-05, "loss": 0.0, "step": 7955 }, { "epoch": 1.84, "learning_rate": 1.9355211462907328e-05, "loss": 0.0, "step": 7956 }, { "epoch": 1.84, "learning_rate": 1.9351359679531627e-05, "loss": 0.0, "step": 7957 }, { "epoch": 1.84, "learning_rate": 1.9347507896155918e-05, "loss": 0.0, "step": 7958 }, { "epoch": 1.84, "learning_rate": 1.9343656112780216e-05, "loss": 0.0, "step": 7959 }, { "epoch": 1.84, "learning_rate": 1.9339804329404515e-05, "loss": 0.0, "step": 7960 }, { "epoch": 1.84, "learning_rate": 1.9335952546028813e-05, "loss": 0.0, "step": 7961 }, { "epoch": 1.84, "learning_rate": 1.933210076265311e-05, "loss": 0.0, "step": 7962 }, { "epoch": 1.84, "learning_rate": 1.9328248979277406e-05, "loss": 0.0, "step": 7963 }, { "epoch": 1.84, "learning_rate": 1.93243971959017e-05, "loss": 0.0001, "step": 7964 }, { "epoch": 1.84, "learning_rate": 1.9320545412526e-05, "loss": 0.0, "step": 7965 }, { "epoch": 1.84, "learning_rate": 1.9316693629150298e-05, "loss": 0.0, "step": 7966 }, { "epoch": 1.84, "learning_rate": 1.9312841845774596e-05, "loss": 0.0, "step": 7967 }, { "epoch": 1.84, "learning_rate": 1.930899006239889e-05, "loss": 0.0, "step": 7968 }, { "epoch": 1.84, "learning_rate": 1.930513827902319e-05, "loss": 0.0, "step": 7969 }, { "epoch": 1.84, "learning_rate": 1.9301286495647485e-05, "loss": 0.0, "step": 7970 }, { "epoch": 1.84, "learning_rate": 1.9297434712271783e-05, "loss": 0.0, "step": 7971 }, { "epoch": 1.84, "learning_rate": 1.9293582928896078e-05, "loss": 0.0, "step": 7972 }, { "epoch": 1.84, "learning_rate": 1.9289731145520376e-05, "loss": 0.0, "step": 7973 }, { "epoch": 1.84, "learning_rate": 1.9285879362144675e-05, "loss": 0.0, "step": 7974 }, { "epoch": 1.84, "learning_rate": 1.9282027578768973e-05, "loss": 0.0004, "step": 7975 }, { "epoch": 1.84, "learning_rate": 1.9278175795393268e-05, "loss": 0.0, "step": 7976 }, { "epoch": 1.84, "learning_rate": 1.9274324012017563e-05, "loss": 0.0, "step": 7977 }, { "epoch": 1.84, "learning_rate": 1.927047222864186e-05, "loss": 0.0, "step": 7978 }, { "epoch": 1.84, "learning_rate": 1.926662044526616e-05, "loss": 0.0, "step": 7979 }, { "epoch": 1.84, "learning_rate": 1.9262768661890458e-05, "loss": 0.0, "step": 7980 }, { "epoch": 1.84, "learning_rate": 1.9258916878514753e-05, "loss": 0.0, "step": 7981 }, { "epoch": 1.84, "learning_rate": 1.9255065095139048e-05, "loss": 0.0, "step": 7982 }, { "epoch": 1.84, "learning_rate": 1.9251213311763346e-05, "loss": 0.0, "step": 7983 }, { "epoch": 1.85, "learning_rate": 1.9247361528387645e-05, "loss": 0.0, "step": 7984 }, { "epoch": 1.85, "learning_rate": 1.9243509745011943e-05, "loss": 0.0, "step": 7985 }, { "epoch": 1.85, "learning_rate": 1.9239657961636238e-05, "loss": 0.0, "step": 7986 }, { "epoch": 1.85, "learning_rate": 1.9235806178260536e-05, "loss": 0.0, "step": 7987 }, { "epoch": 1.85, "learning_rate": 1.923195439488483e-05, "loss": 0.0, "step": 7988 }, { "epoch": 1.85, "learning_rate": 1.922810261150913e-05, "loss": 0.0, "step": 7989 }, { "epoch": 1.85, "learning_rate": 1.9224250828133428e-05, "loss": 0.0, "step": 7990 }, { "epoch": 1.85, "learning_rate": 1.9220399044757723e-05, "loss": 0.0, "step": 7991 }, { "epoch": 1.85, "learning_rate": 1.921654726138202e-05, "loss": 0.0, "step": 7992 }, { "epoch": 1.85, "learning_rate": 1.921269547800632e-05, "loss": 0.0, "step": 7993 }, { "epoch": 1.85, "learning_rate": 1.9208843694630614e-05, "loss": 0.0, "step": 7994 }, { "epoch": 1.85, "learning_rate": 1.9204991911254913e-05, "loss": 0.0, "step": 7995 }, { "epoch": 1.85, "learning_rate": 1.9201140127879208e-05, "loss": 0.0, "step": 7996 }, { "epoch": 1.85, "learning_rate": 1.9197288344503506e-05, "loss": 0.0001, "step": 7997 }, { "epoch": 1.85, "learning_rate": 1.9193436561127804e-05, "loss": 0.0, "step": 7998 }, { "epoch": 1.85, "learning_rate": 1.91895847777521e-05, "loss": 0.0, "step": 7999 }, { "epoch": 1.85, "learning_rate": 1.9185732994376394e-05, "loss": 0.0, "step": 8000 }, { "epoch": 1.85, "learning_rate": 1.9181881211000693e-05, "loss": 0.0, "step": 8001 }, { "epoch": 1.85, "learning_rate": 1.917802942762499e-05, "loss": 0.0, "step": 8002 }, { "epoch": 1.85, "learning_rate": 1.917417764424929e-05, "loss": 0.0, "step": 8003 }, { "epoch": 1.85, "learning_rate": 1.9170325860873588e-05, "loss": 0.0, "step": 8004 }, { "epoch": 1.85, "learning_rate": 1.9166474077497883e-05, "loss": 0.0, "step": 8005 }, { "epoch": 1.85, "learning_rate": 1.9162622294122178e-05, "loss": 0.0018, "step": 8006 }, { "epoch": 1.85, "learning_rate": 1.9158770510746476e-05, "loss": 0.0, "step": 8007 }, { "epoch": 1.85, "learning_rate": 1.9154918727370774e-05, "loss": 0.0, "step": 8008 }, { "epoch": 1.85, "learning_rate": 1.9151066943995073e-05, "loss": 0.0, "step": 8009 }, { "epoch": 1.85, "learning_rate": 1.9147215160619368e-05, "loss": 0.0007, "step": 8010 }, { "epoch": 1.85, "learning_rate": 1.9143363377243666e-05, "loss": 0.0, "step": 8011 }, { "epoch": 1.85, "learning_rate": 1.913951159386796e-05, "loss": 0.0, "step": 8012 }, { "epoch": 1.85, "learning_rate": 1.913565981049226e-05, "loss": 0.0, "step": 8013 }, { "epoch": 1.85, "learning_rate": 1.9131808027116554e-05, "loss": 0.0, "step": 8014 }, { "epoch": 1.85, "learning_rate": 1.9127956243740853e-05, "loss": 0.0, "step": 8015 }, { "epoch": 1.85, "learning_rate": 1.912410446036515e-05, "loss": 0.0, "step": 8016 }, { "epoch": 1.85, "learning_rate": 1.9120252676989446e-05, "loss": 0.0, "step": 8017 }, { "epoch": 1.85, "learning_rate": 1.9116400893613744e-05, "loss": 0.0, "step": 8018 }, { "epoch": 1.85, "learning_rate": 1.911254911023804e-05, "loss": 0.0, "step": 8019 }, { "epoch": 1.85, "learning_rate": 1.9108697326862337e-05, "loss": 0.0, "step": 8020 }, { "epoch": 1.85, "learning_rate": 1.9104845543486636e-05, "loss": 0.0, "step": 8021 }, { "epoch": 1.85, "learning_rate": 1.9100993760110934e-05, "loss": 0.0, "step": 8022 }, { "epoch": 1.85, "learning_rate": 1.909714197673523e-05, "loss": 0.0, "step": 8023 }, { "epoch": 1.85, "learning_rate": 1.9093290193359524e-05, "loss": 0.0, "step": 8024 }, { "epoch": 1.85, "learning_rate": 1.9089438409983822e-05, "loss": 0.0, "step": 8025 }, { "epoch": 1.85, "learning_rate": 1.908558662660812e-05, "loss": 0.0, "step": 8026 }, { "epoch": 1.86, "learning_rate": 1.908173484323242e-05, "loss": 0.0, "step": 8027 }, { "epoch": 1.86, "learning_rate": 1.9077883059856714e-05, "loss": 0.0, "step": 8028 }, { "epoch": 1.86, "learning_rate": 1.9074031276481012e-05, "loss": 0.0, "step": 8029 }, { "epoch": 1.86, "learning_rate": 1.9070179493105307e-05, "loss": 0.0, "step": 8030 }, { "epoch": 1.86, "learning_rate": 1.9066327709729606e-05, "loss": 0.0, "step": 8031 }, { "epoch": 1.86, "learning_rate": 1.9062475926353904e-05, "loss": 0.0, "step": 8032 }, { "epoch": 1.86, "learning_rate": 1.90586241429782e-05, "loss": 0.0, "step": 8033 }, { "epoch": 1.86, "learning_rate": 1.9054772359602497e-05, "loss": 0.0, "step": 8034 }, { "epoch": 1.86, "learning_rate": 1.9050920576226792e-05, "loss": 0.0, "step": 8035 }, { "epoch": 1.86, "learning_rate": 1.904706879285109e-05, "loss": 0.0, "step": 8036 }, { "epoch": 1.86, "learning_rate": 1.904321700947539e-05, "loss": 0.0, "step": 8037 }, { "epoch": 1.86, "learning_rate": 1.9039365226099684e-05, "loss": 0.0, "step": 8038 }, { "epoch": 1.86, "learning_rate": 1.9035513442723982e-05, "loss": 0.0, "step": 8039 }, { "epoch": 1.86, "learning_rate": 1.903166165934828e-05, "loss": 0.0, "step": 8040 }, { "epoch": 1.86, "learning_rate": 1.9027809875972576e-05, "loss": 0.0, "step": 8041 }, { "epoch": 1.86, "learning_rate": 1.902395809259687e-05, "loss": 0.0, "step": 8042 }, { "epoch": 1.86, "learning_rate": 1.902010630922117e-05, "loss": 0.0, "step": 8043 }, { "epoch": 1.86, "learning_rate": 1.9016254525845467e-05, "loss": 0.0, "step": 8044 }, { "epoch": 1.86, "learning_rate": 1.9012402742469766e-05, "loss": 0.0, "step": 8045 }, { "epoch": 1.86, "learning_rate": 1.9008550959094064e-05, "loss": 0.0, "step": 8046 }, { "epoch": 1.86, "learning_rate": 1.900469917571836e-05, "loss": 0.0, "step": 8047 }, { "epoch": 1.86, "learning_rate": 1.9000847392342654e-05, "loss": 0.0, "step": 8048 }, { "epoch": 1.86, "learning_rate": 1.8996995608966952e-05, "loss": 0.0, "step": 8049 }, { "epoch": 1.86, "learning_rate": 1.899314382559125e-05, "loss": 0.0, "step": 8050 }, { "epoch": 1.86, "learning_rate": 1.898929204221555e-05, "loss": 0.0, "step": 8051 }, { "epoch": 1.86, "learning_rate": 1.8985440258839844e-05, "loss": 0.0, "step": 8052 }, { "epoch": 1.86, "learning_rate": 1.898158847546414e-05, "loss": 0.0, "step": 8053 }, { "epoch": 1.86, "learning_rate": 1.8977736692088437e-05, "loss": 0.0, "step": 8054 }, { "epoch": 1.86, "learning_rate": 1.8973884908712735e-05, "loss": 0.0, "step": 8055 }, { "epoch": 1.86, "learning_rate": 1.897003312533703e-05, "loss": 0.0, "step": 8056 }, { "epoch": 1.86, "learning_rate": 1.896618134196133e-05, "loss": 0.0, "step": 8057 }, { "epoch": 1.86, "learning_rate": 1.8962329558585627e-05, "loss": 0.0, "step": 8058 }, { "epoch": 1.86, "learning_rate": 1.8958477775209922e-05, "loss": 0.0, "step": 8059 }, { "epoch": 1.86, "learning_rate": 1.895462599183422e-05, "loss": 0.0, "step": 8060 }, { "epoch": 1.86, "learning_rate": 1.8950774208458515e-05, "loss": 0.0, "step": 8061 }, { "epoch": 1.86, "learning_rate": 1.8946922425082814e-05, "loss": 0.0, "step": 8062 }, { "epoch": 1.86, "learning_rate": 1.8943070641707112e-05, "loss": 0.0, "step": 8063 }, { "epoch": 1.86, "learning_rate": 1.893921885833141e-05, "loss": 0.0, "step": 8064 }, { "epoch": 1.86, "learning_rate": 1.8935367074955705e-05, "loss": 0.0, "step": 8065 }, { "epoch": 1.86, "learning_rate": 1.893151529158e-05, "loss": 0.0, "step": 8066 }, { "epoch": 1.86, "learning_rate": 1.89276635082043e-05, "loss": 0.0, "step": 8067 }, { "epoch": 1.86, "learning_rate": 1.8923811724828597e-05, "loss": 0.0, "step": 8068 }, { "epoch": 1.86, "learning_rate": 1.8919959941452895e-05, "loss": 0.0, "step": 8069 }, { "epoch": 1.87, "learning_rate": 1.891610815807719e-05, "loss": 0.0, "step": 8070 }, { "epoch": 1.87, "learning_rate": 1.8912256374701485e-05, "loss": 0.0, "step": 8071 }, { "epoch": 1.87, "learning_rate": 1.8908404591325784e-05, "loss": 0.0, "step": 8072 }, { "epoch": 1.87, "learning_rate": 1.8904552807950082e-05, "loss": 0.0, "step": 8073 }, { "epoch": 1.87, "learning_rate": 1.890070102457438e-05, "loss": 0.0, "step": 8074 }, { "epoch": 1.87, "learning_rate": 1.8896849241198675e-05, "loss": 0.0, "step": 8075 }, { "epoch": 1.87, "learning_rate": 1.8892997457822974e-05, "loss": 0.0, "step": 8076 }, { "epoch": 1.87, "learning_rate": 1.888914567444727e-05, "loss": 0.0, "step": 8077 }, { "epoch": 1.87, "learning_rate": 1.8885293891071567e-05, "loss": 0.0, "step": 8078 }, { "epoch": 1.87, "learning_rate": 1.8881442107695865e-05, "loss": 0.0, "step": 8079 }, { "epoch": 1.87, "learning_rate": 1.887759032432016e-05, "loss": 0.0, "step": 8080 }, { "epoch": 1.87, "learning_rate": 1.887373854094446e-05, "loss": 0.0, "step": 8081 }, { "epoch": 1.87, "learning_rate": 1.8869886757568757e-05, "loss": 0.0, "step": 8082 }, { "epoch": 1.87, "learning_rate": 1.8866034974193052e-05, "loss": 0.0, "step": 8083 }, { "epoch": 1.87, "learning_rate": 1.8862183190817347e-05, "loss": 0.0, "step": 8084 }, { "epoch": 1.87, "learning_rate": 1.8858331407441645e-05, "loss": 0.0, "step": 8085 }, { "epoch": 1.87, "learning_rate": 1.8854479624065943e-05, "loss": 0.0, "step": 8086 }, { "epoch": 1.87, "learning_rate": 1.8850627840690242e-05, "loss": 0.0, "step": 8087 }, { "epoch": 1.87, "learning_rate": 1.884677605731454e-05, "loss": 0.0, "step": 8088 }, { "epoch": 1.87, "learning_rate": 1.8842924273938832e-05, "loss": 0.0, "step": 8089 }, { "epoch": 1.87, "learning_rate": 1.883907249056313e-05, "loss": 0.0, "step": 8090 }, { "epoch": 1.87, "learning_rate": 1.883522070718743e-05, "loss": 0.0, "step": 8091 }, { "epoch": 1.87, "learning_rate": 1.8831368923811727e-05, "loss": 0.0, "step": 8092 }, { "epoch": 1.87, "learning_rate": 1.8827517140436025e-05, "loss": 0.0, "step": 8093 }, { "epoch": 1.87, "learning_rate": 1.882366535706032e-05, "loss": 0.0, "step": 8094 }, { "epoch": 1.87, "learning_rate": 1.8819813573684615e-05, "loss": 0.0, "step": 8095 }, { "epoch": 1.87, "learning_rate": 1.8815961790308913e-05, "loss": 0.0, "step": 8096 }, { "epoch": 1.87, "learning_rate": 1.8812110006933212e-05, "loss": 0.0, "step": 8097 }, { "epoch": 1.87, "learning_rate": 1.8808258223557507e-05, "loss": 0.0, "step": 8098 }, { "epoch": 1.87, "learning_rate": 1.8804406440181805e-05, "loss": 0.0, "step": 8099 }, { "epoch": 1.87, "learning_rate": 1.8800554656806103e-05, "loss": 0.0, "step": 8100 }, { "epoch": 1.87, "learning_rate": 1.8796702873430398e-05, "loss": 0.0, "step": 8101 }, { "epoch": 1.87, "learning_rate": 1.8792851090054697e-05, "loss": 0.0, "step": 8102 }, { "epoch": 1.87, "learning_rate": 1.878899930667899e-05, "loss": 0.0, "step": 8103 }, { "epoch": 1.87, "learning_rate": 1.878514752330329e-05, "loss": 0.0, "step": 8104 }, { "epoch": 1.87, "learning_rate": 1.8781295739927588e-05, "loss": 0.0, "step": 8105 }, { "epoch": 1.87, "learning_rate": 1.8777443956551887e-05, "loss": 0.0, "step": 8106 }, { "epoch": 1.87, "learning_rate": 1.877359217317618e-05, "loss": 0.0, "step": 8107 }, { "epoch": 1.87, "learning_rate": 1.8769740389800477e-05, "loss": 0.0, "step": 8108 }, { "epoch": 1.87, "learning_rate": 1.8765888606424775e-05, "loss": 0.0, "step": 8109 }, { "epoch": 1.87, "learning_rate": 1.8762036823049073e-05, "loss": 0.0035, "step": 8110 }, { "epoch": 1.87, "learning_rate": 1.875818503967337e-05, "loss": 0.0, "step": 8111 }, { "epoch": 1.87, "learning_rate": 1.8754333256297667e-05, "loss": 0.0, "step": 8112 }, { "epoch": 1.87, "learning_rate": 1.875048147292196e-05, "loss": 0.0, "step": 8113 }, { "epoch": 1.88, "learning_rate": 1.874662968954626e-05, "loss": 0.0, "step": 8114 }, { "epoch": 1.88, "learning_rate": 1.8742777906170558e-05, "loss": 0.0, "step": 8115 }, { "epoch": 1.88, "learning_rate": 1.8738926122794857e-05, "loss": 0.0, "step": 8116 }, { "epoch": 1.88, "learning_rate": 1.873507433941915e-05, "loss": 0.0, "step": 8117 }, { "epoch": 1.88, "learning_rate": 1.873122255604345e-05, "loss": 0.0, "step": 8118 }, { "epoch": 1.88, "learning_rate": 1.8727370772667745e-05, "loss": 0.0, "step": 8119 }, { "epoch": 1.88, "learning_rate": 1.8723518989292043e-05, "loss": 0.0, "step": 8120 }, { "epoch": 1.88, "learning_rate": 1.871966720591634e-05, "loss": 0.0, "step": 8121 }, { "epoch": 1.88, "learning_rate": 1.8715815422540636e-05, "loss": 0.0, "step": 8122 }, { "epoch": 1.88, "learning_rate": 1.8711963639164935e-05, "loss": 0.0, "step": 8123 }, { "epoch": 1.88, "learning_rate": 1.8708111855789233e-05, "loss": 0.0, "step": 8124 }, { "epoch": 1.88, "learning_rate": 1.8704260072413528e-05, "loss": 0.0, "step": 8125 }, { "epoch": 1.88, "learning_rate": 1.8700408289037823e-05, "loss": 0.0, "step": 8126 }, { "epoch": 1.88, "learning_rate": 1.869655650566212e-05, "loss": 0.0, "step": 8127 }, { "epoch": 1.88, "learning_rate": 1.869270472228642e-05, "loss": 0.0, "step": 8128 }, { "epoch": 1.88, "learning_rate": 1.8688852938910718e-05, "loss": 0.0, "step": 8129 }, { "epoch": 1.88, "learning_rate": 1.8685001155535013e-05, "loss": 0.0, "step": 8130 }, { "epoch": 1.88, "learning_rate": 1.8681149372159308e-05, "loss": 0.0, "step": 8131 }, { "epoch": 1.88, "learning_rate": 1.8677297588783606e-05, "loss": 0.0, "step": 8132 }, { "epoch": 1.88, "learning_rate": 1.8673445805407905e-05, "loss": 0.0, "step": 8133 }, { "epoch": 1.88, "learning_rate": 1.8669594022032203e-05, "loss": 0.0, "step": 8134 }, { "epoch": 1.88, "learning_rate": 1.86657422386565e-05, "loss": 0.0, "step": 8135 }, { "epoch": 1.88, "learning_rate": 1.8661890455280796e-05, "loss": 0.0, "step": 8136 }, { "epoch": 1.88, "learning_rate": 1.865803867190509e-05, "loss": 0.0, "step": 8137 }, { "epoch": 1.88, "learning_rate": 1.865418688852939e-05, "loss": 0.0, "step": 8138 }, { "epoch": 1.88, "learning_rate": 1.8650335105153688e-05, "loss": 0.0, "step": 8139 }, { "epoch": 1.88, "learning_rate": 1.8646483321777983e-05, "loss": 0.0, "step": 8140 }, { "epoch": 1.88, "learning_rate": 1.864263153840228e-05, "loss": 0.0, "step": 8141 }, { "epoch": 1.88, "learning_rate": 1.863877975502658e-05, "loss": 0.0, "step": 8142 }, { "epoch": 1.88, "learning_rate": 1.8634927971650875e-05, "loss": 0.0, "step": 8143 }, { "epoch": 1.88, "learning_rate": 1.8631076188275173e-05, "loss": 0.0, "step": 8144 }, { "epoch": 1.88, "learning_rate": 1.8627224404899468e-05, "loss": 0.0001, "step": 8145 }, { "epoch": 1.88, "learning_rate": 1.8623372621523766e-05, "loss": 0.0, "step": 8146 }, { "epoch": 1.88, "learning_rate": 1.8619520838148065e-05, "loss": 0.0, "step": 8147 }, { "epoch": 1.88, "learning_rate": 1.861566905477236e-05, "loss": 0.0, "step": 8148 }, { "epoch": 1.88, "learning_rate": 1.8611817271396658e-05, "loss": 0.0, "step": 8149 }, { "epoch": 1.88, "learning_rate": 1.8607965488020953e-05, "loss": 0.0, "step": 8150 }, { "epoch": 1.88, "learning_rate": 1.860411370464525e-05, "loss": 0.0, "step": 8151 }, { "epoch": 1.88, "learning_rate": 1.860026192126955e-05, "loss": 0.0, "step": 8152 }, { "epoch": 1.88, "learning_rate": 1.8596410137893848e-05, "loss": 0.0, "step": 8153 }, { "epoch": 1.88, "learning_rate": 1.8592558354518143e-05, "loss": 0.0, "step": 8154 }, { "epoch": 1.88, "learning_rate": 1.8588706571142438e-05, "loss": 0.0, "step": 8155 }, { "epoch": 1.88, "learning_rate": 1.8584854787766736e-05, "loss": 0.0, "step": 8156 }, { "epoch": 1.89, "learning_rate": 1.8581003004391034e-05, "loss": 0.0, "step": 8157 }, { "epoch": 1.89, "learning_rate": 1.8577151221015333e-05, "loss": 0.0, "step": 8158 }, { "epoch": 1.89, "learning_rate": 1.8573299437639628e-05, "loss": 0.0, "step": 8159 }, { "epoch": 1.89, "learning_rate": 1.8569447654263926e-05, "loss": 0.0001, "step": 8160 }, { "epoch": 1.89, "learning_rate": 1.856559587088822e-05, "loss": 0.0, "step": 8161 }, { "epoch": 1.89, "learning_rate": 1.856174408751252e-05, "loss": 0.0, "step": 8162 }, { "epoch": 1.89, "learning_rate": 1.8557892304136818e-05, "loss": 0.0, "step": 8163 }, { "epoch": 1.89, "learning_rate": 1.8554040520761113e-05, "loss": 0.0, "step": 8164 }, { "epoch": 1.89, "learning_rate": 1.855018873738541e-05, "loss": 0.0, "step": 8165 }, { "epoch": 1.89, "learning_rate": 1.8546336954009706e-05, "loss": 0.0, "step": 8166 }, { "epoch": 1.89, "learning_rate": 1.8542485170634004e-05, "loss": 0.0, "step": 8167 }, { "epoch": 1.89, "learning_rate": 1.85386333872583e-05, "loss": 0.0, "step": 8168 }, { "epoch": 1.89, "learning_rate": 1.8534781603882598e-05, "loss": 0.0, "step": 8169 }, { "epoch": 1.89, "learning_rate": 1.8530929820506896e-05, "loss": 0.0, "step": 8170 }, { "epoch": 1.89, "learning_rate": 1.8527078037131194e-05, "loss": 0.0, "step": 8171 }, { "epoch": 1.89, "learning_rate": 1.852322625375549e-05, "loss": 0.0, "step": 8172 }, { "epoch": 1.89, "learning_rate": 1.8519374470379784e-05, "loss": 0.0, "step": 8173 }, { "epoch": 1.89, "learning_rate": 1.8515522687004083e-05, "loss": 0.0, "step": 8174 }, { "epoch": 1.89, "learning_rate": 1.851167090362838e-05, "loss": 0.0, "step": 8175 }, { "epoch": 1.89, "learning_rate": 1.850781912025268e-05, "loss": 0.0, "step": 8176 }, { "epoch": 1.89, "learning_rate": 1.8503967336876978e-05, "loss": 0.0, "step": 8177 }, { "epoch": 1.89, "learning_rate": 1.8500115553501273e-05, "loss": 0.0, "step": 8178 }, { "epoch": 1.89, "learning_rate": 1.8496263770125567e-05, "loss": 0.0, "step": 8179 }, { "epoch": 1.89, "learning_rate": 1.8492411986749866e-05, "loss": 0.0, "step": 8180 }, { "epoch": 1.89, "learning_rate": 1.8488560203374164e-05, "loss": 0.0, "step": 8181 }, { "epoch": 1.89, "learning_rate": 1.848470841999846e-05, "loss": 0.0, "step": 8182 }, { "epoch": 1.89, "learning_rate": 1.8480856636622757e-05, "loss": 0.0, "step": 8183 }, { "epoch": 1.89, "learning_rate": 1.8477004853247052e-05, "loss": 0.0, "step": 8184 }, { "epoch": 1.89, "learning_rate": 1.847315306987135e-05, "loss": 0.0, "step": 8185 }, { "epoch": 1.89, "learning_rate": 1.846930128649565e-05, "loss": 0.0001, "step": 8186 }, { "epoch": 1.89, "learning_rate": 1.8465449503119944e-05, "loss": 0.0, "step": 8187 }, { "epoch": 1.89, "learning_rate": 1.8461597719744242e-05, "loss": 0.0011, "step": 8188 }, { "epoch": 1.89, "learning_rate": 1.845774593636854e-05, "loss": 0.0, "step": 8189 }, { "epoch": 1.89, "learning_rate": 1.8453894152992836e-05, "loss": 0.0, "step": 8190 }, { "epoch": 1.89, "learning_rate": 1.8450042369617134e-05, "loss": 0.0, "step": 8191 }, { "epoch": 1.89, "learning_rate": 1.844619058624143e-05, "loss": 0.0, "step": 8192 }, { "epoch": 1.89, "learning_rate": 1.8442338802865727e-05, "loss": 0.0, "step": 8193 }, { "epoch": 1.89, "learning_rate": 1.8438487019490026e-05, "loss": 0.0, "step": 8194 }, { "epoch": 1.89, "learning_rate": 1.8434635236114324e-05, "loss": 0.0, "step": 8195 }, { "epoch": 1.89, "learning_rate": 1.843078345273862e-05, "loss": 0.0, "step": 8196 }, { "epoch": 1.89, "learning_rate": 1.8426931669362914e-05, "loss": 0.0, "step": 8197 }, { "epoch": 1.89, "learning_rate": 1.8423079885987212e-05, "loss": 0.0, "step": 8198 }, { "epoch": 1.89, "learning_rate": 1.841922810261151e-05, "loss": 0.0, "step": 8199 }, { "epoch": 1.9, "learning_rate": 1.841537631923581e-05, "loss": 0.0, "step": 8200 }, { "epoch": 1.9, "learning_rate": 1.8411524535860104e-05, "loss": 0.0001, "step": 8201 }, { "epoch": 1.9, "learning_rate": 1.8407672752484402e-05, "loss": 0.0, "step": 8202 }, { "epoch": 1.9, "learning_rate": 1.8403820969108697e-05, "loss": 0.0, "step": 8203 }, { "epoch": 1.9, "learning_rate": 1.8399969185732996e-05, "loss": 0.0, "step": 8204 }, { "epoch": 1.9, "learning_rate": 1.8396117402357294e-05, "loss": 0.0, "step": 8205 }, { "epoch": 1.9, "learning_rate": 1.839226561898159e-05, "loss": 0.0, "step": 8206 }, { "epoch": 1.9, "learning_rate": 1.8388413835605887e-05, "loss": 0.0, "step": 8207 }, { "epoch": 1.9, "learning_rate": 1.8384562052230182e-05, "loss": 0.0, "step": 8208 }, { "epoch": 1.9, "learning_rate": 1.838071026885448e-05, "loss": 0.0, "step": 8209 }, { "epoch": 1.9, "learning_rate": 1.8376858485478775e-05, "loss": 0.0, "step": 8210 }, { "epoch": 1.9, "learning_rate": 1.8373006702103074e-05, "loss": 0.0, "step": 8211 }, { "epoch": 1.9, "learning_rate": 1.8369154918727372e-05, "loss": 0.0, "step": 8212 }, { "epoch": 1.9, "learning_rate": 1.836530313535167e-05, "loss": 0.0, "step": 8213 }, { "epoch": 1.9, "learning_rate": 1.8361451351975965e-05, "loss": 0.0, "step": 8214 }, { "epoch": 1.9, "learning_rate": 1.835759956860026e-05, "loss": 0.0, "step": 8215 }, { "epoch": 1.9, "learning_rate": 1.835374778522456e-05, "loss": 0.0, "step": 8216 }, { "epoch": 1.9, "learning_rate": 1.8349896001848857e-05, "loss": 0.0, "step": 8217 }, { "epoch": 1.9, "learning_rate": 1.8346044218473155e-05, "loss": 0.0, "step": 8218 }, { "epoch": 1.9, "learning_rate": 1.8342192435097454e-05, "loss": 0.021, "step": 8219 }, { "epoch": 1.9, "learning_rate": 1.833834065172175e-05, "loss": 0.0, "step": 8220 }, { "epoch": 1.9, "learning_rate": 1.8334488868346044e-05, "loss": 0.0, "step": 8221 }, { "epoch": 1.9, "learning_rate": 1.8330637084970342e-05, "loss": 0.0, "step": 8222 }, { "epoch": 1.9, "learning_rate": 1.832678530159464e-05, "loss": 0.0, "step": 8223 }, { "epoch": 1.9, "learning_rate": 1.8322933518218935e-05, "loss": 0.0, "step": 8224 }, { "epoch": 1.9, "learning_rate": 1.8319081734843234e-05, "loss": 0.0, "step": 8225 }, { "epoch": 1.9, "learning_rate": 1.831522995146753e-05, "loss": 0.0, "step": 8226 }, { "epoch": 1.9, "learning_rate": 1.8311378168091827e-05, "loss": 0.0, "step": 8227 }, { "epoch": 1.9, "learning_rate": 1.8307526384716125e-05, "loss": 0.0002, "step": 8228 }, { "epoch": 1.9, "learning_rate": 1.830367460134042e-05, "loss": 0.0, "step": 8229 }, { "epoch": 1.9, "learning_rate": 1.829982281796472e-05, "loss": 0.0, "step": 8230 }, { "epoch": 1.9, "learning_rate": 1.8295971034589017e-05, "loss": 0.0, "step": 8231 }, { "epoch": 1.9, "learning_rate": 1.8292119251213312e-05, "loss": 0.0, "step": 8232 }, { "epoch": 1.9, "learning_rate": 1.828826746783761e-05, "loss": 0.0, "step": 8233 }, { "epoch": 1.9, "learning_rate": 1.8284415684461905e-05, "loss": 0.0, "step": 8234 }, { "epoch": 1.9, "learning_rate": 1.8280563901086204e-05, "loss": 0.0, "step": 8235 }, { "epoch": 1.9, "learning_rate": 1.8276712117710502e-05, "loss": 0.0, "step": 8236 }, { "epoch": 1.9, "learning_rate": 1.82728603343348e-05, "loss": 0.0, "step": 8237 }, { "epoch": 1.9, "learning_rate": 1.8269008550959095e-05, "loss": 0.0, "step": 8238 }, { "epoch": 1.9, "learning_rate": 1.826515676758339e-05, "loss": 0.0, "step": 8239 }, { "epoch": 1.9, "learning_rate": 1.826130498420769e-05, "loss": 0.0001, "step": 8240 }, { "epoch": 1.9, "learning_rate": 1.8257453200831987e-05, "loss": 0.0, "step": 8241 }, { "epoch": 1.9, "learning_rate": 1.8253601417456285e-05, "loss": 0.0, "step": 8242 }, { "epoch": 1.91, "learning_rate": 1.824974963408058e-05, "loss": 0.0, "step": 8243 }, { "epoch": 1.91, "learning_rate": 1.8245897850704875e-05, "loss": 0.0, "step": 8244 }, { "epoch": 1.91, "learning_rate": 1.8242046067329173e-05, "loss": 0.0, "step": 8245 }, { "epoch": 1.91, "learning_rate": 1.8238194283953472e-05, "loss": 0.0, "step": 8246 }, { "epoch": 1.91, "learning_rate": 1.823434250057777e-05, "loss": 0.0, "step": 8247 }, { "epoch": 1.91, "learning_rate": 1.8230490717202065e-05, "loss": 0.0, "step": 8248 }, { "epoch": 1.91, "learning_rate": 1.8226638933826363e-05, "loss": 0.0, "step": 8249 }, { "epoch": 1.91, "learning_rate": 1.822278715045066e-05, "loss": 0.0, "step": 8250 }, { "epoch": 1.91, "learning_rate": 1.8218935367074957e-05, "loss": 0.0, "step": 8251 }, { "epoch": 1.91, "learning_rate": 1.821508358369925e-05, "loss": 0.0, "step": 8252 }, { "epoch": 1.91, "learning_rate": 1.821123180032355e-05, "loss": 0.0, "step": 8253 }, { "epoch": 1.91, "learning_rate": 1.820738001694785e-05, "loss": 0.0, "step": 8254 }, { "epoch": 1.91, "learning_rate": 1.8203528233572147e-05, "loss": 0.0, "step": 8255 }, { "epoch": 1.91, "learning_rate": 1.819967645019644e-05, "loss": 0.0, "step": 8256 }, { "epoch": 1.91, "learning_rate": 1.8195824666820737e-05, "loss": 0.0, "step": 8257 }, { "epoch": 1.91, "learning_rate": 1.8191972883445035e-05, "loss": 0.0, "step": 8258 }, { "epoch": 1.91, "learning_rate": 1.8188121100069333e-05, "loss": 0.0, "step": 8259 }, { "epoch": 1.91, "learning_rate": 1.818426931669363e-05, "loss": 0.0, "step": 8260 }, { "epoch": 1.91, "learning_rate": 1.8180417533317927e-05, "loss": 0.0, "step": 8261 }, { "epoch": 1.91, "learning_rate": 1.817656574994222e-05, "loss": 0.0, "step": 8262 }, { "epoch": 1.91, "learning_rate": 1.817271396656652e-05, "loss": 0.0, "step": 8263 }, { "epoch": 1.91, "learning_rate": 1.8168862183190818e-05, "loss": 0.0, "step": 8264 }, { "epoch": 1.91, "learning_rate": 1.8165010399815117e-05, "loss": 0.0, "step": 8265 }, { "epoch": 1.91, "learning_rate": 1.816115861643941e-05, "loss": 0.0, "step": 8266 }, { "epoch": 1.91, "learning_rate": 1.815730683306371e-05, "loss": 0.0, "step": 8267 }, { "epoch": 1.91, "learning_rate": 1.8153455049688005e-05, "loss": 0.0, "step": 8268 }, { "epoch": 1.91, "learning_rate": 1.8149603266312303e-05, "loss": 0.0, "step": 8269 }, { "epoch": 1.91, "learning_rate": 1.81457514829366e-05, "loss": 0.0, "step": 8270 }, { "epoch": 1.91, "learning_rate": 1.8141899699560897e-05, "loss": 0.0, "step": 8271 }, { "epoch": 1.91, "learning_rate": 1.8138047916185195e-05, "loss": 0.0, "step": 8272 }, { "epoch": 1.91, "learning_rate": 1.8134196132809493e-05, "loss": 0.0, "step": 8273 }, { "epoch": 1.91, "learning_rate": 1.8130344349433788e-05, "loss": 0.0, "step": 8274 }, { "epoch": 1.91, "learning_rate": 1.8126492566058086e-05, "loss": 0.0, "step": 8275 }, { "epoch": 1.91, "learning_rate": 1.812264078268238e-05, "loss": 0.0, "step": 8276 }, { "epoch": 1.91, "learning_rate": 1.811878899930668e-05, "loss": 0.0, "step": 8277 }, { "epoch": 1.91, "learning_rate": 1.8114937215930978e-05, "loss": 0.0, "step": 8278 }, { "epoch": 1.91, "learning_rate": 1.8111085432555276e-05, "loss": 0.0, "step": 8279 }, { "epoch": 1.91, "learning_rate": 1.8107233649179568e-05, "loss": 0.0, "step": 8280 }, { "epoch": 1.91, "learning_rate": 1.8103381865803866e-05, "loss": 0.0, "step": 8281 }, { "epoch": 1.91, "learning_rate": 1.8099530082428165e-05, "loss": 0.0, "step": 8282 }, { "epoch": 1.91, "learning_rate": 1.8095678299052463e-05, "loss": 0.0, "step": 8283 }, { "epoch": 1.91, "learning_rate": 1.809182651567676e-05, "loss": 0.0, "step": 8284 }, { "epoch": 1.91, "learning_rate": 1.8087974732301056e-05, "loss": 0.0, "step": 8285 }, { "epoch": 1.91, "learning_rate": 1.808412294892535e-05, "loss": 0.0, "step": 8286 }, { "epoch": 1.92, "learning_rate": 1.808027116554965e-05, "loss": 0.0, "step": 8287 }, { "epoch": 1.92, "learning_rate": 1.8076419382173948e-05, "loss": 0.0, "step": 8288 }, { "epoch": 1.92, "learning_rate": 1.8072567598798246e-05, "loss": 0.0, "step": 8289 }, { "epoch": 1.92, "learning_rate": 1.806871581542254e-05, "loss": 0.0, "step": 8290 }, { "epoch": 1.92, "learning_rate": 1.806486403204684e-05, "loss": 0.0, "step": 8291 }, { "epoch": 1.92, "learning_rate": 1.8061012248671135e-05, "loss": 0.0, "step": 8292 }, { "epoch": 1.92, "learning_rate": 1.8057160465295433e-05, "loss": 0.0, "step": 8293 }, { "epoch": 1.92, "learning_rate": 1.8053308681919728e-05, "loss": 0.0, "step": 8294 }, { "epoch": 1.92, "learning_rate": 1.8049456898544026e-05, "loss": 0.0, "step": 8295 }, { "epoch": 1.92, "learning_rate": 1.8045605115168325e-05, "loss": 0.0, "step": 8296 }, { "epoch": 1.92, "learning_rate": 1.8041753331792623e-05, "loss": 0.0, "step": 8297 }, { "epoch": 1.92, "learning_rate": 1.8037901548416918e-05, "loss": 0.0, "step": 8298 }, { "epoch": 1.92, "learning_rate": 1.8034049765041213e-05, "loss": 0.0, "step": 8299 }, { "epoch": 1.92, "learning_rate": 1.803019798166551e-05, "loss": 0.0, "step": 8300 }, { "epoch": 1.92, "learning_rate": 1.802634619828981e-05, "loss": 0.0, "step": 8301 }, { "epoch": 1.92, "learning_rate": 1.8022494414914108e-05, "loss": 0.0, "step": 8302 }, { "epoch": 1.92, "learning_rate": 1.8018642631538403e-05, "loss": 0.0, "step": 8303 }, { "epoch": 1.92, "learning_rate": 1.8014790848162698e-05, "loss": 0.0, "step": 8304 }, { "epoch": 1.92, "learning_rate": 1.8010939064786996e-05, "loss": 0.0, "step": 8305 }, { "epoch": 1.92, "learning_rate": 1.8007087281411294e-05, "loss": 0.0, "step": 8306 }, { "epoch": 1.92, "learning_rate": 1.8003235498035593e-05, "loss": 0.0, "step": 8307 }, { "epoch": 1.92, "learning_rate": 1.7999383714659888e-05, "loss": 0.0, "step": 8308 }, { "epoch": 1.92, "learning_rate": 1.7995531931284186e-05, "loss": 0.0001, "step": 8309 }, { "epoch": 1.92, "learning_rate": 1.799168014790848e-05, "loss": 0.0, "step": 8310 }, { "epoch": 1.92, "learning_rate": 1.798782836453278e-05, "loss": 0.0, "step": 8311 }, { "epoch": 1.92, "learning_rate": 1.7983976581157078e-05, "loss": 0.0, "step": 8312 }, { "epoch": 1.92, "learning_rate": 1.7980124797781373e-05, "loss": 0.0, "step": 8313 }, { "epoch": 1.92, "learning_rate": 1.797627301440567e-05, "loss": 0.0, "step": 8314 }, { "epoch": 1.92, "learning_rate": 1.797242123102997e-05, "loss": 0.0, "step": 8315 }, { "epoch": 1.92, "learning_rate": 1.7968569447654264e-05, "loss": 0.0, "step": 8316 }, { "epoch": 1.92, "learning_rate": 1.7964717664278563e-05, "loss": 0.0, "step": 8317 }, { "epoch": 1.92, "learning_rate": 1.7960865880902858e-05, "loss": 0.0002, "step": 8318 }, { "epoch": 1.92, "learning_rate": 1.7957014097527156e-05, "loss": 0.0, "step": 8319 }, { "epoch": 1.92, "learning_rate": 1.7953162314151454e-05, "loss": 0.0, "step": 8320 }, { "epoch": 1.92, "learning_rate": 1.794931053077575e-05, "loss": 0.0, "step": 8321 }, { "epoch": 1.92, "learning_rate": 1.7945458747400044e-05, "loss": 0.0, "step": 8322 }, { "epoch": 1.92, "learning_rate": 1.7941606964024343e-05, "loss": 0.0, "step": 8323 }, { "epoch": 1.92, "learning_rate": 1.793775518064864e-05, "loss": 0.0, "step": 8324 }, { "epoch": 1.92, "learning_rate": 1.793390339727294e-05, "loss": 0.0, "step": 8325 }, { "epoch": 1.92, "learning_rate": 1.7930051613897238e-05, "loss": 0.0, "step": 8326 }, { "epoch": 1.92, "learning_rate": 1.7926199830521533e-05, "loss": 0.0, "step": 8327 }, { "epoch": 1.92, "learning_rate": 1.7922348047145828e-05, "loss": 0.0, "step": 8328 }, { "epoch": 1.92, "learning_rate": 1.7918496263770126e-05, "loss": 0.0, "step": 8329 }, { "epoch": 1.93, "learning_rate": 1.7914644480394424e-05, "loss": 0.0, "step": 8330 }, { "epoch": 1.93, "learning_rate": 1.7910792697018723e-05, "loss": 0.0, "step": 8331 }, { "epoch": 1.93, "learning_rate": 1.7906940913643018e-05, "loss": 0.0, "step": 8332 }, { "epoch": 1.93, "learning_rate": 1.7903089130267316e-05, "loss": 0.0, "step": 8333 }, { "epoch": 1.93, "learning_rate": 1.789923734689161e-05, "loss": 0.0, "step": 8334 }, { "epoch": 1.93, "learning_rate": 1.789538556351591e-05, "loss": 0.0, "step": 8335 }, { "epoch": 1.93, "learning_rate": 1.7891533780140204e-05, "loss": 0.0, "step": 8336 }, { "epoch": 1.93, "learning_rate": 1.7887681996764502e-05, "loss": 0.0, "step": 8337 }, { "epoch": 1.93, "learning_rate": 1.78838302133888e-05, "loss": 0.0, "step": 8338 }, { "epoch": 1.93, "learning_rate": 1.7879978430013096e-05, "loss": 0.0, "step": 8339 }, { "epoch": 1.93, "learning_rate": 1.7876126646637394e-05, "loss": 0.0, "step": 8340 }, { "epoch": 1.93, "learning_rate": 1.787227486326169e-05, "loss": 0.0, "step": 8341 }, { "epoch": 1.93, "learning_rate": 1.7868423079885987e-05, "loss": 0.0, "step": 8342 }, { "epoch": 1.93, "learning_rate": 1.7864571296510286e-05, "loss": 0.0, "step": 8343 }, { "epoch": 1.93, "learning_rate": 1.7860719513134584e-05, "loss": 0.0, "step": 8344 }, { "epoch": 1.93, "learning_rate": 1.785686772975888e-05, "loss": 0.0, "step": 8345 }, { "epoch": 1.93, "learning_rate": 1.7853015946383174e-05, "loss": 0.0, "step": 8346 }, { "epoch": 1.93, "learning_rate": 1.7849164163007472e-05, "loss": 0.0001, "step": 8347 }, { "epoch": 1.93, "learning_rate": 1.784531237963177e-05, "loss": 0.0, "step": 8348 }, { "epoch": 1.93, "learning_rate": 1.784146059625607e-05, "loss": 0.0, "step": 8349 }, { "epoch": 1.93, "learning_rate": 1.7837608812880364e-05, "loss": 0.0, "step": 8350 }, { "epoch": 1.93, "learning_rate": 1.7833757029504662e-05, "loss": 0.0, "step": 8351 }, { "epoch": 1.93, "learning_rate": 1.7829905246128957e-05, "loss": 0.0, "step": 8352 }, { "epoch": 1.93, "learning_rate": 1.7826053462753256e-05, "loss": 0.0, "step": 8353 }, { "epoch": 1.93, "learning_rate": 1.7822201679377554e-05, "loss": 0.0001, "step": 8354 }, { "epoch": 1.93, "learning_rate": 1.781834989600185e-05, "loss": 0.0, "step": 8355 }, { "epoch": 1.93, "learning_rate": 1.7814498112626147e-05, "loss": 0.0, "step": 8356 }, { "epoch": 1.93, "learning_rate": 1.7810646329250442e-05, "loss": 0.0002, "step": 8357 }, { "epoch": 1.93, "learning_rate": 1.780679454587474e-05, "loss": 0.0, "step": 8358 }, { "epoch": 1.93, "learning_rate": 1.780294276249904e-05, "loss": 0.0, "step": 8359 }, { "epoch": 1.93, "learning_rate": 1.7799090979123334e-05, "loss": 0.0, "step": 8360 }, { "epoch": 1.93, "learning_rate": 1.7795239195747632e-05, "loss": 0.0, "step": 8361 }, { "epoch": 1.93, "learning_rate": 1.779138741237193e-05, "loss": 0.0, "step": 8362 }, { "epoch": 1.93, "learning_rate": 1.7787535628996226e-05, "loss": 0.0, "step": 8363 }, { "epoch": 1.93, "learning_rate": 1.778368384562052e-05, "loss": 0.0, "step": 8364 }, { "epoch": 1.93, "learning_rate": 1.777983206224482e-05, "loss": 0.0, "step": 8365 }, { "epoch": 1.93, "learning_rate": 1.7775980278869117e-05, "loss": 0.0, "step": 8366 }, { "epoch": 1.93, "learning_rate": 1.7772128495493416e-05, "loss": 0.0, "step": 8367 }, { "epoch": 1.93, "learning_rate": 1.7768276712117714e-05, "loss": 0.0, "step": 8368 }, { "epoch": 1.93, "learning_rate": 1.776442492874201e-05, "loss": 0.0, "step": 8369 }, { "epoch": 1.93, "learning_rate": 1.7760573145366304e-05, "loss": 0.0, "step": 8370 }, { "epoch": 1.93, "learning_rate": 1.7756721361990602e-05, "loss": 0.0, "step": 8371 }, { "epoch": 1.93, "learning_rate": 1.77528695786149e-05, "loss": 0.0, "step": 8372 }, { "epoch": 1.94, "learning_rate": 1.77490177952392e-05, "loss": 0.0, "step": 8373 }, { "epoch": 1.94, "learning_rate": 1.7745166011863494e-05, "loss": 0.0, "step": 8374 }, { "epoch": 1.94, "learning_rate": 1.774131422848779e-05, "loss": 0.0, "step": 8375 }, { "epoch": 1.94, "learning_rate": 1.7737462445112087e-05, "loss": 0.0, "step": 8376 }, { "epoch": 1.94, "learning_rate": 1.7733610661736385e-05, "loss": 0.0, "step": 8377 }, { "epoch": 1.94, "learning_rate": 1.772975887836068e-05, "loss": 0.0, "step": 8378 }, { "epoch": 1.94, "learning_rate": 1.772590709498498e-05, "loss": 0.0, "step": 8379 }, { "epoch": 1.94, "learning_rate": 1.7722055311609277e-05, "loss": 0.0, "step": 8380 }, { "epoch": 1.94, "learning_rate": 1.7718203528233572e-05, "loss": 0.0005, "step": 8381 }, { "epoch": 1.94, "learning_rate": 1.771435174485787e-05, "loss": 0.0, "step": 8382 }, { "epoch": 1.94, "learning_rate": 1.7710499961482165e-05, "loss": 0.0, "step": 8383 }, { "epoch": 1.94, "learning_rate": 1.7706648178106464e-05, "loss": 0.0, "step": 8384 }, { "epoch": 1.94, "learning_rate": 1.7702796394730762e-05, "loss": 0.0, "step": 8385 }, { "epoch": 1.94, "learning_rate": 1.769894461135506e-05, "loss": 0.0, "step": 8386 }, { "epoch": 1.94, "learning_rate": 1.7695092827979355e-05, "loss": 0.0, "step": 8387 }, { "epoch": 1.94, "learning_rate": 1.769124104460365e-05, "loss": 0.0, "step": 8388 }, { "epoch": 1.94, "learning_rate": 1.768738926122795e-05, "loss": 0.0, "step": 8389 }, { "epoch": 1.94, "learning_rate": 1.7683537477852247e-05, "loss": 0.0, "step": 8390 }, { "epoch": 1.94, "learning_rate": 1.7679685694476545e-05, "loss": 0.0, "step": 8391 }, { "epoch": 1.94, "learning_rate": 1.767583391110084e-05, "loss": 0.0, "step": 8392 }, { "epoch": 1.94, "learning_rate": 1.7671982127725135e-05, "loss": 0.0, "step": 8393 }, { "epoch": 1.94, "learning_rate": 1.7668130344349434e-05, "loss": 0.0001, "step": 8394 }, { "epoch": 1.94, "learning_rate": 1.7664278560973732e-05, "loss": 0.0, "step": 8395 }, { "epoch": 1.94, "learning_rate": 1.766042677759803e-05, "loss": 0.0, "step": 8396 }, { "epoch": 1.94, "learning_rate": 1.7656574994222325e-05, "loss": 0.0, "step": 8397 }, { "epoch": 1.94, "learning_rate": 1.7652723210846624e-05, "loss": 0.0, "step": 8398 }, { "epoch": 1.94, "learning_rate": 1.764887142747092e-05, "loss": 0.0, "step": 8399 }, { "epoch": 1.94, "learning_rate": 1.7645019644095217e-05, "loss": 0.0, "step": 8400 }, { "epoch": 1.94, "learning_rate": 1.7641167860719515e-05, "loss": 0.0, "step": 8401 }, { "epoch": 1.94, "learning_rate": 1.763731607734381e-05, "loss": 0.0, "step": 8402 }, { "epoch": 1.94, "learning_rate": 1.763346429396811e-05, "loss": 0.0, "step": 8403 }, { "epoch": 1.94, "learning_rate": 1.7629612510592407e-05, "loss": 0.0, "step": 8404 }, { "epoch": 1.94, "learning_rate": 1.7625760727216702e-05, "loss": 0.0, "step": 8405 }, { "epoch": 1.94, "learning_rate": 1.7621908943840997e-05, "loss": 0.0, "step": 8406 }, { "epoch": 1.94, "learning_rate": 1.7618057160465295e-05, "loss": 0.0, "step": 8407 }, { "epoch": 1.94, "learning_rate": 1.7614205377089593e-05, "loss": 0.0, "step": 8408 }, { "epoch": 1.94, "learning_rate": 1.7610353593713892e-05, "loss": 0.0, "step": 8409 }, { "epoch": 1.94, "learning_rate": 1.760650181033819e-05, "loss": 0.0, "step": 8410 }, { "epoch": 1.94, "learning_rate": 1.760265002696248e-05, "loss": 0.0, "step": 8411 }, { "epoch": 1.94, "learning_rate": 1.759879824358678e-05, "loss": 0.0, "step": 8412 }, { "epoch": 1.94, "learning_rate": 1.759494646021108e-05, "loss": 0.0, "step": 8413 }, { "epoch": 1.94, "learning_rate": 1.7591094676835377e-05, "loss": 0.0, "step": 8414 }, { "epoch": 1.94, "learning_rate": 1.7587242893459675e-05, "loss": 0.0, "step": 8415 }, { "epoch": 1.94, "learning_rate": 1.758339111008397e-05, "loss": 0.0, "step": 8416 }, { "epoch": 1.95, "learning_rate": 1.7579539326708265e-05, "loss": 0.0, "step": 8417 }, { "epoch": 1.95, "learning_rate": 1.7575687543332563e-05, "loss": 0.0, "step": 8418 }, { "epoch": 1.95, "learning_rate": 1.757183575995686e-05, "loss": 0.0, "step": 8419 }, { "epoch": 1.95, "learning_rate": 1.7567983976581157e-05, "loss": 0.0, "step": 8420 }, { "epoch": 1.95, "learning_rate": 1.7564132193205455e-05, "loss": 0.0, "step": 8421 }, { "epoch": 1.95, "learning_rate": 1.7560280409829753e-05, "loss": 0.0, "step": 8422 }, { "epoch": 1.95, "learning_rate": 1.7556428626454048e-05, "loss": 0.0, "step": 8423 }, { "epoch": 1.95, "learning_rate": 1.7552576843078347e-05, "loss": 0.0, "step": 8424 }, { "epoch": 1.95, "learning_rate": 1.754872505970264e-05, "loss": 0.0, "step": 8425 }, { "epoch": 1.95, "learning_rate": 1.754487327632694e-05, "loss": 0.0, "step": 8426 }, { "epoch": 1.95, "learning_rate": 1.7541021492951238e-05, "loss": 0.0, "step": 8427 }, { "epoch": 1.95, "learning_rate": 1.7537169709575537e-05, "loss": 0.0, "step": 8428 }, { "epoch": 1.95, "learning_rate": 1.753331792619983e-05, "loss": 0.0, "step": 8429 }, { "epoch": 1.95, "learning_rate": 1.7529466142824126e-05, "loss": 0.0001, "step": 8430 }, { "epoch": 1.95, "learning_rate": 1.7525614359448425e-05, "loss": 0.0, "step": 8431 }, { "epoch": 1.95, "learning_rate": 1.7521762576072723e-05, "loss": 0.0, "step": 8432 }, { "epoch": 1.95, "learning_rate": 1.751791079269702e-05, "loss": 0.0, "step": 8433 }, { "epoch": 1.95, "learning_rate": 1.7514059009321316e-05, "loss": 0.0, "step": 8434 }, { "epoch": 1.95, "learning_rate": 1.751020722594561e-05, "loss": 0.0, "step": 8435 }, { "epoch": 1.95, "learning_rate": 1.750635544256991e-05, "loss": 0.0, "step": 8436 }, { "epoch": 1.95, "learning_rate": 1.7502503659194208e-05, "loss": 0.0, "step": 8437 }, { "epoch": 1.95, "learning_rate": 1.7498651875818506e-05, "loss": 0.0, "step": 8438 }, { "epoch": 1.95, "learning_rate": 1.74948000924428e-05, "loss": 0.0, "step": 8439 }, { "epoch": 1.95, "learning_rate": 1.74909483090671e-05, "loss": 0.0, "step": 8440 }, { "epoch": 1.95, "learning_rate": 1.7487096525691395e-05, "loss": 0.0, "step": 8441 }, { "epoch": 1.95, "learning_rate": 1.7483244742315693e-05, "loss": 0.0, "step": 8442 }, { "epoch": 1.95, "learning_rate": 1.747939295893999e-05, "loss": 0.0, "step": 8443 }, { "epoch": 1.95, "learning_rate": 1.7475541175564286e-05, "loss": 0.0, "step": 8444 }, { "epoch": 1.95, "learning_rate": 1.7471689392188585e-05, "loss": 0.0, "step": 8445 }, { "epoch": 1.95, "learning_rate": 1.7467837608812883e-05, "loss": 0.0, "step": 8446 }, { "epoch": 1.95, "learning_rate": 1.7463985825437178e-05, "loss": 0.0, "step": 8447 }, { "epoch": 1.95, "learning_rate": 1.7460134042061473e-05, "loss": 0.0, "step": 8448 }, { "epoch": 1.95, "learning_rate": 1.745628225868577e-05, "loss": 0.0, "step": 8449 }, { "epoch": 1.95, "learning_rate": 1.745243047531007e-05, "loss": 0.0, "step": 8450 }, { "epoch": 1.95, "learning_rate": 1.7448578691934368e-05, "loss": 0.0, "step": 8451 }, { "epoch": 1.95, "learning_rate": 1.7444726908558663e-05, "loss": 0.0008, "step": 8452 }, { "epoch": 1.95, "learning_rate": 1.7440875125182958e-05, "loss": 0.001, "step": 8453 }, { "epoch": 1.95, "learning_rate": 1.7437023341807256e-05, "loss": 0.0, "step": 8454 }, { "epoch": 1.95, "learning_rate": 1.7433171558431555e-05, "loss": 0.0, "step": 8455 }, { "epoch": 1.95, "learning_rate": 1.7429319775055853e-05, "loss": 0.0, "step": 8456 }, { "epoch": 1.95, "learning_rate": 1.742546799168015e-05, "loss": 0.0, "step": 8457 }, { "epoch": 1.95, "learning_rate": 1.7421616208304446e-05, "loss": 0.0, "step": 8458 }, { "epoch": 1.95, "learning_rate": 1.741776442492874e-05, "loss": 0.0, "step": 8459 }, { "epoch": 1.96, "learning_rate": 1.741391264155304e-05, "loss": 0.0, "step": 8460 }, { "epoch": 1.96, "learning_rate": 1.7410060858177338e-05, "loss": 0.0, "step": 8461 }, { "epoch": 1.96, "learning_rate": 1.7406209074801633e-05, "loss": 0.0, "step": 8462 }, { "epoch": 1.96, "learning_rate": 1.740235729142593e-05, "loss": 0.0, "step": 8463 }, { "epoch": 1.96, "learning_rate": 1.739850550805023e-05, "loss": 0.0, "step": 8464 }, { "epoch": 1.96, "learning_rate": 1.7394653724674524e-05, "loss": 0.0, "step": 8465 }, { "epoch": 1.96, "learning_rate": 1.7390801941298823e-05, "loss": 0.0, "step": 8466 }, { "epoch": 1.96, "learning_rate": 1.7386950157923118e-05, "loss": 0.0, "step": 8467 }, { "epoch": 1.96, "learning_rate": 1.7383098374547416e-05, "loss": 0.0, "step": 8468 }, { "epoch": 1.96, "learning_rate": 1.7379246591171714e-05, "loss": 0.0, "step": 8469 }, { "epoch": 1.96, "learning_rate": 1.737539480779601e-05, "loss": 0.0, "step": 8470 }, { "epoch": 1.96, "learning_rate": 1.7371543024420308e-05, "loss": 0.0, "step": 8471 }, { "epoch": 1.96, "learning_rate": 1.7367691241044603e-05, "loss": 0.0, "step": 8472 }, { "epoch": 1.96, "learning_rate": 1.73638394576689e-05, "loss": 0.0, "step": 8473 }, { "epoch": 1.96, "learning_rate": 1.73599876742932e-05, "loss": 0.0, "step": 8474 }, { "epoch": 1.96, "learning_rate": 1.7356135890917498e-05, "loss": 0.0, "step": 8475 }, { "epoch": 1.96, "learning_rate": 1.7352284107541793e-05, "loss": 0.0, "step": 8476 }, { "epoch": 1.96, "learning_rate": 1.7348432324166088e-05, "loss": 0.0, "step": 8477 }, { "epoch": 1.96, "learning_rate": 1.7344580540790386e-05, "loss": 0.0, "step": 8478 }, { "epoch": 1.96, "learning_rate": 1.7340728757414684e-05, "loss": 0.0, "step": 8479 }, { "epoch": 1.96, "learning_rate": 1.7336876974038983e-05, "loss": 0.0, "step": 8480 }, { "epoch": 1.96, "learning_rate": 1.7333025190663278e-05, "loss": 0.0, "step": 8481 }, { "epoch": 1.96, "learning_rate": 1.7329173407287576e-05, "loss": 0.0, "step": 8482 }, { "epoch": 1.96, "learning_rate": 1.732532162391187e-05, "loss": 0.0, "step": 8483 }, { "epoch": 1.96, "learning_rate": 1.732146984053617e-05, "loss": 0.0, "step": 8484 }, { "epoch": 1.96, "learning_rate": 1.7317618057160468e-05, "loss": 0.0, "step": 8485 }, { "epoch": 1.96, "learning_rate": 1.7313766273784763e-05, "loss": 0.0003, "step": 8486 }, { "epoch": 1.96, "learning_rate": 1.730991449040906e-05, "loss": 0.0, "step": 8487 }, { "epoch": 1.96, "learning_rate": 1.7306062707033356e-05, "loss": 0.0, "step": 8488 }, { "epoch": 1.96, "learning_rate": 1.7302210923657654e-05, "loss": 0.0, "step": 8489 }, { "epoch": 1.96, "learning_rate": 1.729835914028195e-05, "loss": 0.0001, "step": 8490 }, { "epoch": 1.96, "learning_rate": 1.7294507356906248e-05, "loss": 0.0, "step": 8491 }, { "epoch": 1.96, "learning_rate": 1.7290655573530546e-05, "loss": 0.0, "step": 8492 }, { "epoch": 1.96, "learning_rate": 1.7286803790154844e-05, "loss": 0.0, "step": 8493 }, { "epoch": 1.96, "learning_rate": 1.728295200677914e-05, "loss": 0.0, "step": 8494 }, { "epoch": 1.96, "learning_rate": 1.7279100223403434e-05, "loss": 0.0, "step": 8495 }, { "epoch": 1.96, "learning_rate": 1.7275248440027732e-05, "loss": 0.0, "step": 8496 }, { "epoch": 1.96, "learning_rate": 1.727139665665203e-05, "loss": 0.0, "step": 8497 }, { "epoch": 1.96, "learning_rate": 1.726754487327633e-05, "loss": 0.0, "step": 8498 }, { "epoch": 1.96, "learning_rate": 1.7263693089900627e-05, "loss": 0.0001, "step": 8499 }, { "epoch": 1.96, "learning_rate": 1.7259841306524922e-05, "loss": 0.0, "step": 8500 }, { "epoch": 1.96, "learning_rate": 1.7255989523149217e-05, "loss": 0.0, "step": 8501 }, { "epoch": 1.96, "learning_rate": 1.7252137739773516e-05, "loss": 0.0001, "step": 8502 }, { "epoch": 1.97, "learning_rate": 1.7248285956397814e-05, "loss": 0.0, "step": 8503 }, { "epoch": 1.97, "learning_rate": 1.724443417302211e-05, "loss": 0.0, "step": 8504 }, { "epoch": 1.97, "learning_rate": 1.7240582389646407e-05, "loss": 0.0, "step": 8505 }, { "epoch": 1.97, "learning_rate": 1.7236730606270702e-05, "loss": 0.0, "step": 8506 }, { "epoch": 1.97, "learning_rate": 1.7232878822895e-05, "loss": 0.0, "step": 8507 }, { "epoch": 1.97, "learning_rate": 1.72290270395193e-05, "loss": 0.0, "step": 8508 }, { "epoch": 1.97, "learning_rate": 1.7225175256143594e-05, "loss": 0.0, "step": 8509 }, { "epoch": 1.97, "learning_rate": 1.7221323472767892e-05, "loss": 0.0, "step": 8510 }, { "epoch": 1.97, "learning_rate": 1.721747168939219e-05, "loss": 0.0, "step": 8511 }, { "epoch": 1.97, "learning_rate": 1.7213619906016486e-05, "loss": 0.0, "step": 8512 }, { "epoch": 1.97, "learning_rate": 1.7209768122640784e-05, "loss": 0.0, "step": 8513 }, { "epoch": 1.97, "learning_rate": 1.720591633926508e-05, "loss": 0.0, "step": 8514 }, { "epoch": 1.97, "learning_rate": 1.7202064555889377e-05, "loss": 0.0, "step": 8515 }, { "epoch": 1.97, "learning_rate": 1.7198212772513676e-05, "loss": 0.0, "step": 8516 }, { "epoch": 1.97, "learning_rate": 1.7194360989137974e-05, "loss": 0.0, "step": 8517 }, { "epoch": 1.97, "learning_rate": 1.719050920576227e-05, "loss": 0.0, "step": 8518 }, { "epoch": 1.97, "learning_rate": 1.7186657422386564e-05, "loss": 0.0, "step": 8519 }, { "epoch": 1.97, "learning_rate": 1.7182805639010862e-05, "loss": 0.0, "step": 8520 }, { "epoch": 1.97, "learning_rate": 1.717895385563516e-05, "loss": 0.0, "step": 8521 }, { "epoch": 1.97, "learning_rate": 1.717510207225946e-05, "loss": 0.0, "step": 8522 }, { "epoch": 1.97, "learning_rate": 1.7171250288883754e-05, "loss": 0.0, "step": 8523 }, { "epoch": 1.97, "learning_rate": 1.7167398505508052e-05, "loss": 0.0, "step": 8524 }, { "epoch": 1.97, "learning_rate": 1.7163546722132347e-05, "loss": 0.0, "step": 8525 }, { "epoch": 1.97, "learning_rate": 1.7159694938756645e-05, "loss": 0.0, "step": 8526 }, { "epoch": 1.97, "learning_rate": 1.7155843155380944e-05, "loss": 0.0, "step": 8527 }, { "epoch": 1.97, "learning_rate": 1.715199137200524e-05, "loss": 0.0, "step": 8528 }, { "epoch": 1.97, "learning_rate": 1.7148139588629537e-05, "loss": 0.0, "step": 8529 }, { "epoch": 1.97, "learning_rate": 1.7144287805253832e-05, "loss": 0.0, "step": 8530 }, { "epoch": 1.97, "learning_rate": 1.714043602187813e-05, "loss": 0.0, "step": 8531 }, { "epoch": 1.97, "learning_rate": 1.7136584238502425e-05, "loss": 0.0, "step": 8532 }, { "epoch": 1.97, "learning_rate": 1.7132732455126724e-05, "loss": 0.0, "step": 8533 }, { "epoch": 1.97, "learning_rate": 1.7128880671751022e-05, "loss": 0.0, "step": 8534 }, { "epoch": 1.97, "learning_rate": 1.712502888837532e-05, "loss": 0.0, "step": 8535 }, { "epoch": 1.97, "learning_rate": 1.7121177104999615e-05, "loss": 0.0001, "step": 8536 }, { "epoch": 1.97, "learning_rate": 1.711732532162391e-05, "loss": 0.0, "step": 8537 }, { "epoch": 1.97, "learning_rate": 1.711347353824821e-05, "loss": 0.0, "step": 8538 }, { "epoch": 1.97, "learning_rate": 1.7109621754872507e-05, "loss": 0.0, "step": 8539 }, { "epoch": 1.97, "learning_rate": 1.7105769971496805e-05, "loss": 0.0, "step": 8540 }, { "epoch": 1.97, "learning_rate": 1.7101918188121104e-05, "loss": 0.0, "step": 8541 }, { "epoch": 1.97, "learning_rate": 1.70980664047454e-05, "loss": 0.0, "step": 8542 }, { "epoch": 1.97, "learning_rate": 1.7094214621369694e-05, "loss": 0.0, "step": 8543 }, { "epoch": 1.97, "learning_rate": 1.7090362837993992e-05, "loss": 0.0, "step": 8544 }, { "epoch": 1.97, "learning_rate": 1.708651105461829e-05, "loss": 0.0, "step": 8545 }, { "epoch": 1.98, "learning_rate": 1.7082659271242585e-05, "loss": 0.0, "step": 8546 }, { "epoch": 1.98, "learning_rate": 1.7078807487866884e-05, "loss": 0.0, "step": 8547 }, { "epoch": 1.98, "learning_rate": 1.707495570449118e-05, "loss": 0.0, "step": 8548 }, { "epoch": 1.98, "learning_rate": 1.7071103921115477e-05, "loss": 0.0, "step": 8549 }, { "epoch": 1.98, "learning_rate": 1.7067252137739775e-05, "loss": 0.0, "step": 8550 }, { "epoch": 1.98, "learning_rate": 1.706340035436407e-05, "loss": 0.0, "step": 8551 }, { "epoch": 1.98, "learning_rate": 1.705954857098837e-05, "loss": 0.0, "step": 8552 }, { "epoch": 1.98, "learning_rate": 1.7055696787612667e-05, "loss": 0.0, "step": 8553 }, { "epoch": 1.98, "learning_rate": 1.7051845004236962e-05, "loss": 0.0, "step": 8554 }, { "epoch": 1.98, "learning_rate": 1.704799322086126e-05, "loss": 0.0, "step": 8555 }, { "epoch": 1.98, "learning_rate": 1.7044141437485555e-05, "loss": 0.0, "step": 8556 }, { "epoch": 1.98, "learning_rate": 1.7040289654109853e-05, "loss": 0.0, "step": 8557 }, { "epoch": 1.98, "learning_rate": 1.7036437870734152e-05, "loss": 0.0, "step": 8558 }, { "epoch": 1.98, "learning_rate": 1.703258608735845e-05, "loss": 0.0, "step": 8559 }, { "epoch": 1.98, "learning_rate": 1.7028734303982745e-05, "loss": 0.0, "step": 8560 }, { "epoch": 1.98, "learning_rate": 1.702488252060704e-05, "loss": 0.0, "step": 8561 }, { "epoch": 1.98, "learning_rate": 1.702103073723134e-05, "loss": 0.0, "step": 8562 }, { "epoch": 1.98, "learning_rate": 1.7017178953855637e-05, "loss": 0.0, "step": 8563 }, { "epoch": 1.98, "learning_rate": 1.7013327170479935e-05, "loss": 0.0, "step": 8564 }, { "epoch": 1.98, "learning_rate": 1.700947538710423e-05, "loss": 0.0, "step": 8565 }, { "epoch": 1.98, "learning_rate": 1.7005623603728525e-05, "loss": 0.0, "step": 8566 }, { "epoch": 1.98, "learning_rate": 1.7001771820352823e-05, "loss": 0.0, "step": 8567 }, { "epoch": 1.98, "learning_rate": 1.6997920036977122e-05, "loss": 0.0, "step": 8568 }, { "epoch": 1.98, "learning_rate": 1.699406825360142e-05, "loss": 0.0, "step": 8569 }, { "epoch": 1.98, "learning_rate": 1.6990216470225715e-05, "loss": 0.0, "step": 8570 }, { "epoch": 1.98, "learning_rate": 1.6986364686850013e-05, "loss": 0.0, "step": 8571 }, { "epoch": 1.98, "learning_rate": 1.698251290347431e-05, "loss": 0.0, "step": 8572 }, { "epoch": 1.98, "learning_rate": 1.6978661120098607e-05, "loss": 0.0, "step": 8573 }, { "epoch": 1.98, "learning_rate": 1.69748093367229e-05, "loss": 0.0, "step": 8574 }, { "epoch": 1.98, "learning_rate": 1.69709575533472e-05, "loss": 0.0, "step": 8575 }, { "epoch": 1.98, "learning_rate": 1.6967105769971498e-05, "loss": 0.0, "step": 8576 }, { "epoch": 1.98, "learning_rate": 1.6963253986595797e-05, "loss": 0.0, "step": 8577 }, { "epoch": 1.98, "learning_rate": 1.695940220322009e-05, "loss": 0.0001, "step": 8578 }, { "epoch": 1.98, "learning_rate": 1.6955550419844387e-05, "loss": 0.0, "step": 8579 }, { "epoch": 1.98, "learning_rate": 1.6951698636468685e-05, "loss": 0.0, "step": 8580 }, { "epoch": 1.98, "learning_rate": 1.6947846853092983e-05, "loss": 0.0, "step": 8581 }, { "epoch": 1.98, "learning_rate": 1.694399506971728e-05, "loss": 0.0, "step": 8582 }, { "epoch": 1.98, "learning_rate": 1.6940143286341577e-05, "loss": 0.0, "step": 8583 }, { "epoch": 1.98, "learning_rate": 1.693629150296587e-05, "loss": 0.0, "step": 8584 }, { "epoch": 1.98, "learning_rate": 1.693243971959017e-05, "loss": 0.0002, "step": 8585 }, { "epoch": 1.98, "learning_rate": 1.6928587936214468e-05, "loss": 0.0, "step": 8586 }, { "epoch": 1.98, "learning_rate": 1.6924736152838767e-05, "loss": 0.0, "step": 8587 }, { "epoch": 1.98, "learning_rate": 1.692088436946306e-05, "loss": 0.0, "step": 8588 }, { "epoch": 1.98, "learning_rate": 1.691703258608736e-05, "loss": 0.0, "step": 8589 }, { "epoch": 1.99, "learning_rate": 1.6913180802711655e-05, "loss": 0.0001, "step": 8590 }, { "epoch": 1.99, "learning_rate": 1.6909329019335953e-05, "loss": 0.0, "step": 8591 }, { "epoch": 1.99, "learning_rate": 1.690547723596025e-05, "loss": 0.0, "step": 8592 }, { "epoch": 1.99, "learning_rate": 1.6901625452584546e-05, "loss": 0.0, "step": 8593 }, { "epoch": 1.99, "learning_rate": 1.6897773669208845e-05, "loss": 0.0, "step": 8594 }, { "epoch": 1.99, "learning_rate": 1.6893921885833143e-05, "loss": 0.0, "step": 8595 }, { "epoch": 1.99, "learning_rate": 1.6890070102457438e-05, "loss": 0.0, "step": 8596 }, { "epoch": 1.99, "learning_rate": 1.6886218319081736e-05, "loss": 0.0, "step": 8597 }, { "epoch": 1.99, "learning_rate": 1.688236653570603e-05, "loss": 0.0, "step": 8598 }, { "epoch": 1.99, "learning_rate": 1.687851475233033e-05, "loss": 0.0, "step": 8599 }, { "epoch": 1.99, "learning_rate": 1.6874662968954628e-05, "loss": 0.0, "step": 8600 }, { "epoch": 1.99, "learning_rate": 1.6870811185578923e-05, "loss": 0.0, "step": 8601 }, { "epoch": 1.99, "learning_rate": 1.6866959402203218e-05, "loss": 0.0, "step": 8602 }, { "epoch": 1.99, "learning_rate": 1.6863107618827516e-05, "loss": 0.0, "step": 8603 }, { "epoch": 1.99, "learning_rate": 1.6859255835451815e-05, "loss": 0.0, "step": 8604 }, { "epoch": 1.99, "learning_rate": 1.6855404052076113e-05, "loss": 0.0, "step": 8605 }, { "epoch": 1.99, "learning_rate": 1.685155226870041e-05, "loss": 0.0, "step": 8606 }, { "epoch": 1.99, "learning_rate": 1.6847700485324706e-05, "loss": 0.0, "step": 8607 }, { "epoch": 1.99, "learning_rate": 1.6843848701949e-05, "loss": 0.0, "step": 8608 }, { "epoch": 1.99, "learning_rate": 1.68399969185733e-05, "loss": 0.0, "step": 8609 }, { "epoch": 1.99, "learning_rate": 1.6836145135197598e-05, "loss": 0.0, "step": 8610 }, { "epoch": 1.99, "learning_rate": 1.6832293351821896e-05, "loss": 0.0002, "step": 8611 }, { "epoch": 1.99, "learning_rate": 1.682844156844619e-05, "loss": 0.0, "step": 8612 }, { "epoch": 1.99, "learning_rate": 1.682458978507049e-05, "loss": 0.0, "step": 8613 }, { "epoch": 1.99, "learning_rate": 1.6820738001694785e-05, "loss": 0.0, "step": 8614 }, { "epoch": 1.99, "learning_rate": 1.6816886218319083e-05, "loss": 0.0, "step": 8615 }, { "epoch": 1.99, "learning_rate": 1.6813034434943378e-05, "loss": 0.0, "step": 8616 }, { "epoch": 1.99, "learning_rate": 1.6809182651567676e-05, "loss": 0.0, "step": 8617 }, { "epoch": 1.99, "learning_rate": 1.6805330868191975e-05, "loss": 0.0, "step": 8618 }, { "epoch": 1.99, "learning_rate": 1.6801479084816273e-05, "loss": 0.0, "step": 8619 }, { "epoch": 1.99, "learning_rate": 1.6797627301440568e-05, "loss": 0.0, "step": 8620 }, { "epoch": 1.99, "learning_rate": 1.6793775518064863e-05, "loss": 0.0, "step": 8621 }, { "epoch": 1.99, "learning_rate": 1.678992373468916e-05, "loss": 0.0, "step": 8622 }, { "epoch": 1.99, "learning_rate": 1.678607195131346e-05, "loss": 0.0, "step": 8623 }, { "epoch": 1.99, "learning_rate": 1.6782220167937758e-05, "loss": 0.0, "step": 8624 }, { "epoch": 1.99, "learning_rate": 1.6778368384562053e-05, "loss": 0.0, "step": 8625 }, { "epoch": 1.99, "learning_rate": 1.6774516601186348e-05, "loss": 0.0, "step": 8626 }, { "epoch": 1.99, "learning_rate": 1.6770664817810646e-05, "loss": 0.0001, "step": 8627 }, { "epoch": 1.99, "learning_rate": 1.6766813034434944e-05, "loss": 0.0, "step": 8628 }, { "epoch": 1.99, "learning_rate": 1.6762961251059243e-05, "loss": 0.0, "step": 8629 }, { "epoch": 1.99, "learning_rate": 1.6759109467683538e-05, "loss": 0.0, "step": 8630 }, { "epoch": 1.99, "learning_rate": 1.6755257684307836e-05, "loss": 0.0, "step": 8631 }, { "epoch": 1.99, "learning_rate": 1.675140590093213e-05, "loss": 0.0, "step": 8632 }, { "epoch": 2.0, "learning_rate": 1.674755411755643e-05, "loss": 0.0, "step": 8633 }, { "epoch": 2.0, "learning_rate": 1.6743702334180728e-05, "loss": 0.0, "step": 8634 }, { "epoch": 2.0, "learning_rate": 1.6739850550805023e-05, "loss": 0.0, "step": 8635 }, { "epoch": 2.0, "learning_rate": 1.673599876742932e-05, "loss": 0.0, "step": 8636 }, { "epoch": 2.0, "learning_rate": 1.673214698405362e-05, "loss": 0.0001, "step": 8637 }, { "epoch": 2.0, "learning_rate": 1.6728295200677914e-05, "loss": 0.0, "step": 8638 }, { "epoch": 2.0, "learning_rate": 1.6724443417302213e-05, "loss": 0.0, "step": 8639 }, { "epoch": 2.0, "learning_rate": 1.6720591633926508e-05, "loss": 0.0, "step": 8640 }, { "epoch": 2.0, "learning_rate": 1.6716739850550806e-05, "loss": 0.0, "step": 8641 }, { "epoch": 2.0, "learning_rate": 1.6712888067175104e-05, "loss": 0.0, "step": 8642 }, { "epoch": 2.0, "learning_rate": 1.67090362837994e-05, "loss": 0.0, "step": 8643 }, { "epoch": 2.0, "learning_rate": 1.6705184500423694e-05, "loss": 0.0, "step": 8644 }, { "epoch": 2.0, "learning_rate": 1.6701332717047993e-05, "loss": 0.0, "step": 8645 }, { "epoch": 2.0, "learning_rate": 1.669748093367229e-05, "loss": 0.0, "step": 8646 }, { "epoch": 2.0, "learning_rate": 1.669362915029659e-05, "loss": 0.0, "step": 8647 }, { "epoch": 2.0, "learning_rate": 1.6689777366920888e-05, "loss": 0.0, "step": 8648 }, { "epoch": 2.0, "learning_rate": 1.6685925583545183e-05, "loss": 0.0, "step": 8649 }, { "epoch": 2.0, "learning_rate": 1.6682073800169477e-05, "loss": 0.0, "step": 8650 }, { "epoch": 2.0, "learning_rate": 1.6678222016793776e-05, "loss": 0.0, "step": 8651 }, { "epoch": 2.0, "learning_rate": 1.6674370233418074e-05, "loss": 0.0, "step": 8652 }, { "epoch": 2.0, "learning_rate": 1.6670518450042373e-05, "loss": 0.0, "step": 8653 }, { "epoch": 2.0, "learning_rate": 1.6666666666666667e-05, "loss": 0.0, "step": 8654 }, { "epoch": 2.0, "learning_rate": 1.6662814883290966e-05, "loss": 0.0, "step": 8655 }, { "epoch": 2.0, "learning_rate": 1.665896309991526e-05, "loss": 0.0, "step": 8656 }, { "epoch": 2.0, "learning_rate": 1.665511131653956e-05, "loss": 0.0009, "step": 8657 }, { "epoch": 2.0, "learning_rate": 1.6651259533163854e-05, "loss": 0.0, "step": 8658 }, { "epoch": 2.0, "learning_rate": 1.6647407749788152e-05, "loss": 0.0, "step": 8659 }, { "epoch": 2.0, "learning_rate": 1.664355596641245e-05, "loss": 0.0, "step": 8660 }, { "epoch": 2.0, "learning_rate": 1.6639704183036746e-05, "loss": 0.0, "step": 8661 }, { "epoch": 2.0, "learning_rate": 1.6635852399661044e-05, "loss": 0.0, "step": 8662 }, { "epoch": 2.0, "learning_rate": 1.663200061628534e-05, "loss": 0.0, "step": 8663 }, { "epoch": 2.0, "learning_rate": 1.6628148832909637e-05, "loss": 0.0, "step": 8664 }, { "epoch": 2.0, "learning_rate": 1.6624297049533936e-05, "loss": 0.0, "step": 8665 }, { "epoch": 2.0, "learning_rate": 1.6620445266158234e-05, "loss": 0.0, "step": 8666 }, { "epoch": 2.0, "learning_rate": 1.661659348278253e-05, "loss": 0.0, "step": 8667 }, { "epoch": 2.0, "learning_rate": 1.6612741699406824e-05, "loss": 0.0, "step": 8668 }, { "epoch": 2.0, "learning_rate": 1.6608889916031122e-05, "loss": 0.0, "step": 8669 }, { "epoch": 2.0, "learning_rate": 1.660503813265542e-05, "loss": 0.0, "step": 8670 }, { "epoch": 2.0, "learning_rate": 1.660118634927972e-05, "loss": 0.0, "step": 8671 }, { "epoch": 2.0, "learning_rate": 1.6597334565904014e-05, "loss": 0.0, "step": 8672 }, { "epoch": 2.0, "learning_rate": 1.6593482782528312e-05, "loss": 0.0, "step": 8673 }, { "epoch": 2.0, "learning_rate": 1.6589630999152607e-05, "loss": 0.0, "step": 8674 }, { "epoch": 2.0, "learning_rate": 1.6585779215776906e-05, "loss": 0.0, "step": 8675 }, { "epoch": 2.01, "learning_rate": 1.6581927432401204e-05, "loss": 0.0, "step": 8676 }, { "epoch": 2.01, "learning_rate": 1.65780756490255e-05, "loss": 0.0, "step": 8677 }, { "epoch": 2.01, "learning_rate": 1.6574223865649797e-05, "loss": 0.0, "step": 8678 }, { "epoch": 2.01, "learning_rate": 1.6570372082274092e-05, "loss": 0.0, "step": 8679 }, { "epoch": 2.01, "learning_rate": 1.656652029889839e-05, "loss": 0.0, "step": 8680 }, { "epoch": 2.01, "learning_rate": 1.656266851552269e-05, "loss": 0.0, "step": 8681 }, { "epoch": 2.01, "learning_rate": 1.6558816732146984e-05, "loss": 0.0, "step": 8682 }, { "epoch": 2.01, "learning_rate": 1.6554964948771282e-05, "loss": 0.0, "step": 8683 }, { "epoch": 2.01, "learning_rate": 1.655111316539558e-05, "loss": 0.0, "step": 8684 }, { "epoch": 2.01, "learning_rate": 1.6547261382019875e-05, "loss": 0.0, "step": 8685 }, { "epoch": 2.01, "learning_rate": 1.654340959864417e-05, "loss": 0.0, "step": 8686 }, { "epoch": 2.01, "learning_rate": 1.653955781526847e-05, "loss": 0.0, "step": 8687 }, { "epoch": 2.01, "learning_rate": 1.6535706031892767e-05, "loss": 0.0, "step": 8688 }, { "epoch": 2.01, "learning_rate": 1.6531854248517065e-05, "loss": 0.0, "step": 8689 }, { "epoch": 2.01, "learning_rate": 1.6528002465141364e-05, "loss": 0.0001, "step": 8690 }, { "epoch": 2.01, "learning_rate": 1.652415068176566e-05, "loss": 0.0, "step": 8691 }, { "epoch": 2.01, "learning_rate": 1.6520298898389954e-05, "loss": 0.0, "step": 8692 }, { "epoch": 2.01, "learning_rate": 1.6516447115014252e-05, "loss": 0.0002, "step": 8693 }, { "epoch": 2.01, "learning_rate": 1.651259533163855e-05, "loss": 0.0, "step": 8694 }, { "epoch": 2.01, "learning_rate": 1.650874354826285e-05, "loss": 0.0, "step": 8695 }, { "epoch": 2.01, "learning_rate": 1.6504891764887144e-05, "loss": 0.0, "step": 8696 }, { "epoch": 2.01, "learning_rate": 1.650103998151144e-05, "loss": 0.0, "step": 8697 }, { "epoch": 2.01, "learning_rate": 1.6497188198135737e-05, "loss": 0.0, "step": 8698 }, { "epoch": 2.01, "learning_rate": 1.6493336414760035e-05, "loss": 0.0, "step": 8699 }, { "epoch": 2.01, "learning_rate": 1.648948463138433e-05, "loss": 0.0, "step": 8700 }, { "epoch": 2.01, "learning_rate": 1.648563284800863e-05, "loss": 0.0, "step": 8701 }, { "epoch": 2.01, "learning_rate": 1.6481781064632927e-05, "loss": 0.0, "step": 8702 }, { "epoch": 2.01, "learning_rate": 1.6477929281257222e-05, "loss": 0.0, "step": 8703 }, { "epoch": 2.01, "learning_rate": 1.647407749788152e-05, "loss": 0.0, "step": 8704 }, { "epoch": 2.01, "learning_rate": 1.6470225714505815e-05, "loss": 0.0, "step": 8705 }, { "epoch": 2.01, "learning_rate": 1.6466373931130114e-05, "loss": 0.0, "step": 8706 }, { "epoch": 2.01, "learning_rate": 1.6462522147754412e-05, "loss": 0.0, "step": 8707 }, { "epoch": 2.01, "learning_rate": 1.645867036437871e-05, "loss": 0.0, "step": 8708 }, { "epoch": 2.01, "learning_rate": 1.6454818581003005e-05, "loss": 0.0, "step": 8709 }, { "epoch": 2.01, "learning_rate": 1.64509667976273e-05, "loss": 0.0, "step": 8710 }, { "epoch": 2.01, "learning_rate": 1.64471150142516e-05, "loss": 0.0, "step": 8711 }, { "epoch": 2.01, "learning_rate": 1.6443263230875897e-05, "loss": 0.0, "step": 8712 }, { "epoch": 2.01, "learning_rate": 1.6439411447500195e-05, "loss": 0.0, "step": 8713 }, { "epoch": 2.01, "learning_rate": 1.643555966412449e-05, "loss": 0.0, "step": 8714 }, { "epoch": 2.01, "learning_rate": 1.6431707880748785e-05, "loss": 0.0, "step": 8715 }, { "epoch": 2.01, "learning_rate": 1.6427856097373083e-05, "loss": 0.0, "step": 8716 }, { "epoch": 2.01, "learning_rate": 1.6424004313997382e-05, "loss": 0.0, "step": 8717 }, { "epoch": 2.01, "learning_rate": 1.642015253062168e-05, "loss": 0.0, "step": 8718 }, { "epoch": 2.02, "learning_rate": 1.6416300747245975e-05, "loss": 0.0, "step": 8719 }, { "epoch": 2.02, "learning_rate": 1.6412448963870273e-05, "loss": 0.0, "step": 8720 }, { "epoch": 2.02, "learning_rate": 1.640859718049457e-05, "loss": 0.0, "step": 8721 }, { "epoch": 2.02, "learning_rate": 1.6404745397118867e-05, "loss": 0.0, "step": 8722 }, { "epoch": 2.02, "learning_rate": 1.6400893613743165e-05, "loss": 0.0, "step": 8723 }, { "epoch": 2.02, "learning_rate": 1.639704183036746e-05, "loss": 0.0, "step": 8724 }, { "epoch": 2.02, "learning_rate": 1.639319004699176e-05, "loss": 0.0, "step": 8725 }, { "epoch": 2.02, "learning_rate": 1.6389338263616057e-05, "loss": 0.0, "step": 8726 }, { "epoch": 2.02, "learning_rate": 1.638548648024035e-05, "loss": 0.0, "step": 8727 }, { "epoch": 2.02, "learning_rate": 1.6381634696864647e-05, "loss": 0.0, "step": 8728 }, { "epoch": 2.02, "learning_rate": 1.6377782913488945e-05, "loss": 0.0, "step": 8729 }, { "epoch": 2.02, "learning_rate": 1.6373931130113243e-05, "loss": 0.0, "step": 8730 }, { "epoch": 2.02, "learning_rate": 1.637007934673754e-05, "loss": 0.0, "step": 8731 }, { "epoch": 2.02, "learning_rate": 1.636622756336184e-05, "loss": 0.0, "step": 8732 }, { "epoch": 2.02, "learning_rate": 1.636237577998613e-05, "loss": 0.0, "step": 8733 }, { "epoch": 2.02, "learning_rate": 1.635852399661043e-05, "loss": 0.0, "step": 8734 }, { "epoch": 2.02, "learning_rate": 1.6354672213234728e-05, "loss": 0.0, "step": 8735 }, { "epoch": 2.02, "learning_rate": 1.6350820429859027e-05, "loss": 0.0, "step": 8736 }, { "epoch": 2.02, "learning_rate": 1.6346968646483325e-05, "loss": 0.0, "step": 8737 }, { "epoch": 2.02, "learning_rate": 1.634311686310762e-05, "loss": 0.0, "step": 8738 }, { "epoch": 2.02, "learning_rate": 1.6339265079731915e-05, "loss": 0.0, "step": 8739 }, { "epoch": 2.02, "learning_rate": 1.6335413296356213e-05, "loss": 0.0, "step": 8740 }, { "epoch": 2.02, "learning_rate": 1.633156151298051e-05, "loss": 0.0, "step": 8741 }, { "epoch": 2.02, "learning_rate": 1.6327709729604807e-05, "loss": 0.0, "step": 8742 }, { "epoch": 2.02, "learning_rate": 1.6323857946229105e-05, "loss": 0.0, "step": 8743 }, { "epoch": 2.02, "learning_rate": 1.6320006162853403e-05, "loss": 0.0, "step": 8744 }, { "epoch": 2.02, "learning_rate": 1.6316154379477698e-05, "loss": 0.0, "step": 8745 }, { "epoch": 2.02, "learning_rate": 1.6312302596101996e-05, "loss": 0.0, "step": 8746 }, { "epoch": 2.02, "learning_rate": 1.630845081272629e-05, "loss": 0.0, "step": 8747 }, { "epoch": 2.02, "learning_rate": 1.630459902935059e-05, "loss": 0.0, "step": 8748 }, { "epoch": 2.02, "learning_rate": 1.6300747245974888e-05, "loss": 0.0, "step": 8749 }, { "epoch": 2.02, "learning_rate": 1.6296895462599186e-05, "loss": 0.0, "step": 8750 }, { "epoch": 2.02, "learning_rate": 1.629304367922348e-05, "loss": 0.0, "step": 8751 }, { "epoch": 2.02, "learning_rate": 1.6289191895847776e-05, "loss": 0.0, "step": 8752 }, { "epoch": 2.02, "learning_rate": 1.6285340112472075e-05, "loss": 0.0, "step": 8753 }, { "epoch": 2.02, "learning_rate": 1.6281488329096373e-05, "loss": 0.0, "step": 8754 }, { "epoch": 2.02, "learning_rate": 1.627763654572067e-05, "loss": 0.0, "step": 8755 }, { "epoch": 2.02, "learning_rate": 1.6273784762344966e-05, "loss": 0.0, "step": 8756 }, { "epoch": 2.02, "learning_rate": 1.626993297896926e-05, "loss": 0.0, "step": 8757 }, { "epoch": 2.02, "learning_rate": 1.626608119559356e-05, "loss": 0.0, "step": 8758 }, { "epoch": 2.02, "learning_rate": 1.6262229412217858e-05, "loss": 0.0, "step": 8759 }, { "epoch": 2.02, "learning_rate": 1.6258377628842156e-05, "loss": 0.0, "step": 8760 }, { "epoch": 2.02, "learning_rate": 1.625452584546645e-05, "loss": 0.0, "step": 8761 }, { "epoch": 2.02, "learning_rate": 1.625067406209075e-05, "loss": 0.0, "step": 8762 }, { "epoch": 2.03, "learning_rate": 1.6246822278715045e-05, "loss": 0.0, "step": 8763 }, { "epoch": 2.03, "learning_rate": 1.6242970495339343e-05, "loss": 0.0, "step": 8764 }, { "epoch": 2.03, "learning_rate": 1.623911871196364e-05, "loss": 0.0, "step": 8765 }, { "epoch": 2.03, "learning_rate": 1.6235266928587936e-05, "loss": 0.0, "step": 8766 }, { "epoch": 2.03, "learning_rate": 1.6231415145212235e-05, "loss": 0.0, "step": 8767 }, { "epoch": 2.03, "learning_rate": 1.6227563361836533e-05, "loss": 0.0, "step": 8768 }, { "epoch": 2.03, "learning_rate": 1.6223711578460828e-05, "loss": 0.0, "step": 8769 }, { "epoch": 2.03, "learning_rate": 1.6219859795085123e-05, "loss": 0.0, "step": 8770 }, { "epoch": 2.03, "learning_rate": 1.621600801170942e-05, "loss": 0.0, "step": 8771 }, { "epoch": 2.03, "learning_rate": 1.621215622833372e-05, "loss": 0.0, "step": 8772 }, { "epoch": 2.03, "learning_rate": 1.6208304444958018e-05, "loss": 0.0, "step": 8773 }, { "epoch": 2.03, "learning_rate": 1.6204452661582313e-05, "loss": 0.0003, "step": 8774 }, { "epoch": 2.03, "learning_rate": 1.6200600878206608e-05, "loss": 0.0, "step": 8775 }, { "epoch": 2.03, "learning_rate": 1.6196749094830906e-05, "loss": 0.0, "step": 8776 }, { "epoch": 2.03, "learning_rate": 1.6192897311455204e-05, "loss": 0.0, "step": 8777 }, { "epoch": 2.03, "learning_rate": 1.6189045528079503e-05, "loss": 0.0, "step": 8778 }, { "epoch": 2.03, "learning_rate": 1.61851937447038e-05, "loss": 0.0, "step": 8779 }, { "epoch": 2.03, "learning_rate": 1.6181341961328096e-05, "loss": 0.0, "step": 8780 }, { "epoch": 2.03, "learning_rate": 1.617749017795239e-05, "loss": 0.0, "step": 8781 }, { "epoch": 2.03, "learning_rate": 1.617363839457669e-05, "loss": 0.0, "step": 8782 }, { "epoch": 2.03, "learning_rate": 1.6169786611200988e-05, "loss": 0.0, "step": 8783 }, { "epoch": 2.03, "learning_rate": 1.6165934827825283e-05, "loss": 0.0, "step": 8784 }, { "epoch": 2.03, "learning_rate": 1.616208304444958e-05, "loss": 0.0, "step": 8785 }, { "epoch": 2.03, "learning_rate": 1.615823126107388e-05, "loss": 0.0, "step": 8786 }, { "epoch": 2.03, "learning_rate": 1.6154379477698174e-05, "loss": 0.0, "step": 8787 }, { "epoch": 2.03, "learning_rate": 1.6150527694322473e-05, "loss": 0.0, "step": 8788 }, { "epoch": 2.03, "learning_rate": 1.6146675910946768e-05, "loss": 0.0, "step": 8789 }, { "epoch": 2.03, "learning_rate": 1.6142824127571066e-05, "loss": 0.0, "step": 8790 }, { "epoch": 2.03, "learning_rate": 1.6138972344195364e-05, "loss": 0.0, "step": 8791 }, { "epoch": 2.03, "learning_rate": 1.613512056081966e-05, "loss": 0.0, "step": 8792 }, { "epoch": 2.03, "learning_rate": 1.6131268777443958e-05, "loss": 0.0, "step": 8793 }, { "epoch": 2.03, "learning_rate": 1.6127416994068253e-05, "loss": 0.0, "step": 8794 }, { "epoch": 2.03, "learning_rate": 1.612356521069255e-05, "loss": 0.0, "step": 8795 }, { "epoch": 2.03, "learning_rate": 1.611971342731685e-05, "loss": 0.0, "step": 8796 }, { "epoch": 2.03, "learning_rate": 1.6115861643941148e-05, "loss": 0.0, "step": 8797 }, { "epoch": 2.03, "learning_rate": 1.6112009860565443e-05, "loss": 0.0, "step": 8798 }, { "epoch": 2.03, "learning_rate": 1.6108158077189738e-05, "loss": 0.0001, "step": 8799 }, { "epoch": 2.03, "learning_rate": 1.6104306293814036e-05, "loss": 0.0, "step": 8800 }, { "epoch": 2.03, "learning_rate": 1.6100454510438334e-05, "loss": 0.0, "step": 8801 }, { "epoch": 2.03, "learning_rate": 1.6096602727062633e-05, "loss": 0.0, "step": 8802 }, { "epoch": 2.03, "learning_rate": 1.6092750943686928e-05, "loss": 0.0, "step": 8803 }, { "epoch": 2.03, "learning_rate": 1.6088899160311226e-05, "loss": 0.0, "step": 8804 }, { "epoch": 2.03, "learning_rate": 1.608504737693552e-05, "loss": 0.0001, "step": 8805 }, { "epoch": 2.04, "learning_rate": 1.608119559355982e-05, "loss": 0.0, "step": 8806 }, { "epoch": 2.04, "learning_rate": 1.6077343810184118e-05, "loss": 0.0, "step": 8807 }, { "epoch": 2.04, "learning_rate": 1.6073492026808412e-05, "loss": 0.0, "step": 8808 }, { "epoch": 2.04, "learning_rate": 1.606964024343271e-05, "loss": 0.0, "step": 8809 }, { "epoch": 2.04, "learning_rate": 1.6065788460057006e-05, "loss": 0.0, "step": 8810 }, { "epoch": 2.04, "learning_rate": 1.6061936676681304e-05, "loss": 0.0, "step": 8811 }, { "epoch": 2.04, "learning_rate": 1.60580848933056e-05, "loss": 0.0, "step": 8812 }, { "epoch": 2.04, "learning_rate": 1.6054233109929897e-05, "loss": 0.0, "step": 8813 }, { "epoch": 2.04, "learning_rate": 1.6050381326554196e-05, "loss": 0.0, "step": 8814 }, { "epoch": 2.04, "learning_rate": 1.6046529543178494e-05, "loss": 0.0, "step": 8815 }, { "epoch": 2.04, "learning_rate": 1.604267775980279e-05, "loss": 0.0, "step": 8816 }, { "epoch": 2.04, "learning_rate": 1.6038825976427084e-05, "loss": 0.0, "step": 8817 }, { "epoch": 2.04, "learning_rate": 1.6034974193051382e-05, "loss": 0.0, "step": 8818 }, { "epoch": 2.04, "learning_rate": 1.603112240967568e-05, "loss": 0.0, "step": 8819 }, { "epoch": 2.04, "learning_rate": 1.602727062629998e-05, "loss": 0.0, "step": 8820 }, { "epoch": 2.04, "learning_rate": 1.6023418842924277e-05, "loss": 0.0, "step": 8821 }, { "epoch": 2.04, "learning_rate": 1.6019567059548572e-05, "loss": 0.0001, "step": 8822 }, { "epoch": 2.04, "learning_rate": 1.6015715276172867e-05, "loss": 0.0, "step": 8823 }, { "epoch": 2.04, "learning_rate": 1.6011863492797166e-05, "loss": 0.0, "step": 8824 }, { "epoch": 2.04, "learning_rate": 1.6008011709421464e-05, "loss": 0.0, "step": 8825 }, { "epoch": 2.04, "learning_rate": 1.600415992604576e-05, "loss": 0.0, "step": 8826 }, { "epoch": 2.04, "learning_rate": 1.6000308142670057e-05, "loss": 0.0, "step": 8827 }, { "epoch": 2.04, "learning_rate": 1.5996456359294352e-05, "loss": 0.0, "step": 8828 }, { "epoch": 2.04, "learning_rate": 1.599260457591865e-05, "loss": 0.0001, "step": 8829 }, { "epoch": 2.04, "learning_rate": 1.598875279254295e-05, "loss": 0.0, "step": 8830 }, { "epoch": 2.04, "learning_rate": 1.5984901009167244e-05, "loss": 0.0, "step": 8831 }, { "epoch": 2.04, "learning_rate": 1.5981049225791542e-05, "loss": 0.0, "step": 8832 }, { "epoch": 2.04, "learning_rate": 1.597719744241584e-05, "loss": 0.0, "step": 8833 }, { "epoch": 2.04, "learning_rate": 1.5973345659040136e-05, "loss": 0.0, "step": 8834 }, { "epoch": 2.04, "learning_rate": 1.5969493875664434e-05, "loss": 0.0, "step": 8835 }, { "epoch": 2.04, "learning_rate": 1.596564209228873e-05, "loss": 0.0, "step": 8836 }, { "epoch": 2.04, "learning_rate": 1.5961790308913027e-05, "loss": 0.0, "step": 8837 }, { "epoch": 2.04, "learning_rate": 1.5957938525537326e-05, "loss": 0.0001, "step": 8838 }, { "epoch": 2.04, "learning_rate": 1.5954086742161624e-05, "loss": 0.0, "step": 8839 }, { "epoch": 2.04, "learning_rate": 1.595023495878592e-05, "loss": 0.0, "step": 8840 }, { "epoch": 2.04, "learning_rate": 1.5946383175410214e-05, "loss": 0.0, "step": 8841 }, { "epoch": 2.04, "learning_rate": 1.5942531392034512e-05, "loss": 0.0001, "step": 8842 }, { "epoch": 2.04, "learning_rate": 1.593867960865881e-05, "loss": 0.0, "step": 8843 }, { "epoch": 2.04, "learning_rate": 1.593482782528311e-05, "loss": 0.0, "step": 8844 }, { "epoch": 2.04, "learning_rate": 1.5930976041907404e-05, "loss": 0.0, "step": 8845 }, { "epoch": 2.04, "learning_rate": 1.5927124258531702e-05, "loss": 0.0, "step": 8846 }, { "epoch": 2.04, "learning_rate": 1.5923272475155997e-05, "loss": 0.0, "step": 8847 }, { "epoch": 2.04, "learning_rate": 1.5919420691780295e-05, "loss": 0.0, "step": 8848 }, { "epoch": 2.05, "learning_rate": 1.5915568908404594e-05, "loss": 0.0, "step": 8849 }, { "epoch": 2.05, "learning_rate": 1.591171712502889e-05, "loss": 0.0, "step": 8850 }, { "epoch": 2.05, "learning_rate": 1.5907865341653187e-05, "loss": 0.0, "step": 8851 }, { "epoch": 2.05, "learning_rate": 1.5904013558277482e-05, "loss": 0.0, "step": 8852 }, { "epoch": 2.05, "learning_rate": 1.590016177490178e-05, "loss": 0.0, "step": 8853 }, { "epoch": 2.05, "learning_rate": 1.5896309991526075e-05, "loss": 0.0, "step": 8854 }, { "epoch": 2.05, "learning_rate": 1.5892458208150374e-05, "loss": 0.0, "step": 8855 }, { "epoch": 2.05, "learning_rate": 1.5888606424774672e-05, "loss": 0.0, "step": 8856 }, { "epoch": 2.05, "learning_rate": 1.588475464139897e-05, "loss": 0.0, "step": 8857 }, { "epoch": 2.05, "learning_rate": 1.5880902858023265e-05, "loss": 0.0, "step": 8858 }, { "epoch": 2.05, "learning_rate": 1.587705107464756e-05, "loss": 0.0, "step": 8859 }, { "epoch": 2.05, "learning_rate": 1.587319929127186e-05, "loss": 0.0, "step": 8860 }, { "epoch": 2.05, "learning_rate": 1.5869347507896157e-05, "loss": 0.0, "step": 8861 }, { "epoch": 2.05, "learning_rate": 1.5865495724520455e-05, "loss": 0.0, "step": 8862 }, { "epoch": 2.05, "learning_rate": 1.5861643941144754e-05, "loss": 0.0, "step": 8863 }, { "epoch": 2.05, "learning_rate": 1.585779215776905e-05, "loss": 0.0, "step": 8864 }, { "epoch": 2.05, "learning_rate": 1.5853940374393344e-05, "loss": 0.0, "step": 8865 }, { "epoch": 2.05, "learning_rate": 1.5850088591017642e-05, "loss": 0.0, "step": 8866 }, { "epoch": 2.05, "learning_rate": 1.584623680764194e-05, "loss": 0.0, "step": 8867 }, { "epoch": 2.05, "learning_rate": 1.5842385024266235e-05, "loss": 0.0, "step": 8868 }, { "epoch": 2.05, "learning_rate": 1.5838533240890534e-05, "loss": 0.0, "step": 8869 }, { "epoch": 2.05, "learning_rate": 1.583468145751483e-05, "loss": 0.0, "step": 8870 }, { "epoch": 2.05, "learning_rate": 1.5830829674139127e-05, "loss": 0.0, "step": 8871 }, { "epoch": 2.05, "learning_rate": 1.5826977890763425e-05, "loss": 0.0, "step": 8872 }, { "epoch": 2.05, "learning_rate": 1.582312610738772e-05, "loss": 0.0, "step": 8873 }, { "epoch": 2.05, "learning_rate": 1.581927432401202e-05, "loss": 0.0, "step": 8874 }, { "epoch": 2.05, "learning_rate": 1.5815422540636317e-05, "loss": 0.0, "step": 8875 }, { "epoch": 2.05, "learning_rate": 1.5811570757260612e-05, "loss": 0.0, "step": 8876 }, { "epoch": 2.05, "learning_rate": 1.580771897388491e-05, "loss": 0.0, "step": 8877 }, { "epoch": 2.05, "learning_rate": 1.5803867190509205e-05, "loss": 0.0, "step": 8878 }, { "epoch": 2.05, "learning_rate": 1.5800015407133503e-05, "loss": 0.0, "step": 8879 }, { "epoch": 2.05, "learning_rate": 1.5796163623757802e-05, "loss": 0.0, "step": 8880 }, { "epoch": 2.05, "learning_rate": 1.57923118403821e-05, "loss": 0.0, "step": 8881 }, { "epoch": 2.05, "learning_rate": 1.5788460057006395e-05, "loss": 0.0, "step": 8882 }, { "epoch": 2.05, "learning_rate": 1.578460827363069e-05, "loss": 0.0, "step": 8883 }, { "epoch": 2.05, "learning_rate": 1.578075649025499e-05, "loss": 0.0, "step": 8884 }, { "epoch": 2.05, "learning_rate": 1.5776904706879287e-05, "loss": 0.0, "step": 8885 }, { "epoch": 2.05, "learning_rate": 1.5773052923503585e-05, "loss": 0.0, "step": 8886 }, { "epoch": 2.05, "learning_rate": 1.576920114012788e-05, "loss": 0.0, "step": 8887 }, { "epoch": 2.05, "learning_rate": 1.5765349356752175e-05, "loss": 0.0, "step": 8888 }, { "epoch": 2.05, "learning_rate": 1.5761497573376473e-05, "loss": 0.0, "step": 8889 }, { "epoch": 2.05, "learning_rate": 1.575764579000077e-05, "loss": 0.0, "step": 8890 }, { "epoch": 2.05, "learning_rate": 1.575379400662507e-05, "loss": 0.0, "step": 8891 }, { "epoch": 2.06, "learning_rate": 1.5749942223249365e-05, "loss": 0.0, "step": 8892 }, { "epoch": 2.06, "learning_rate": 1.5746090439873663e-05, "loss": 0.0, "step": 8893 }, { "epoch": 2.06, "learning_rate": 1.5742238656497958e-05, "loss": 0.0, "step": 8894 }, { "epoch": 2.06, "learning_rate": 1.5738386873122257e-05, "loss": 0.0, "step": 8895 }, { "epoch": 2.06, "learning_rate": 1.573453508974655e-05, "loss": 0.0, "step": 8896 }, { "epoch": 2.06, "learning_rate": 1.573068330637085e-05, "loss": 0.0, "step": 8897 }, { "epoch": 2.06, "learning_rate": 1.5726831522995148e-05, "loss": 0.0, "step": 8898 }, { "epoch": 2.06, "learning_rate": 1.5722979739619447e-05, "loss": 0.0, "step": 8899 }, { "epoch": 2.06, "learning_rate": 1.571912795624374e-05, "loss": 0.0, "step": 8900 }, { "epoch": 2.06, "learning_rate": 1.5715276172868036e-05, "loss": 0.0, "step": 8901 }, { "epoch": 2.06, "learning_rate": 1.5711424389492335e-05, "loss": 0.0, "step": 8902 }, { "epoch": 2.06, "learning_rate": 1.5707572606116633e-05, "loss": 0.0, "step": 8903 }, { "epoch": 2.06, "learning_rate": 1.570372082274093e-05, "loss": 0.0, "step": 8904 }, { "epoch": 2.06, "learning_rate": 1.5699869039365226e-05, "loss": 0.0, "step": 8905 }, { "epoch": 2.06, "learning_rate": 1.569601725598952e-05, "loss": 0.0, "step": 8906 }, { "epoch": 2.06, "learning_rate": 1.569216547261382e-05, "loss": 0.0, "step": 8907 }, { "epoch": 2.06, "learning_rate": 1.5688313689238118e-05, "loss": 0.0, "step": 8908 }, { "epoch": 2.06, "learning_rate": 1.5684461905862416e-05, "loss": 0.0001, "step": 8909 }, { "epoch": 2.06, "learning_rate": 1.568061012248671e-05, "loss": 0.0, "step": 8910 }, { "epoch": 2.06, "learning_rate": 1.567675833911101e-05, "loss": 0.0, "step": 8911 }, { "epoch": 2.06, "learning_rate": 1.5672906555735305e-05, "loss": 0.0001, "step": 8912 }, { "epoch": 2.06, "learning_rate": 1.5669054772359603e-05, "loss": 0.0, "step": 8913 }, { "epoch": 2.06, "learning_rate": 1.56652029889839e-05, "loss": 0.0, "step": 8914 }, { "epoch": 2.06, "learning_rate": 1.5661351205608196e-05, "loss": 0.0001, "step": 8915 }, { "epoch": 2.06, "learning_rate": 1.5657499422232495e-05, "loss": 0.0, "step": 8916 }, { "epoch": 2.06, "learning_rate": 1.5653647638856793e-05, "loss": 0.0, "step": 8917 }, { "epoch": 2.06, "learning_rate": 1.5649795855481088e-05, "loss": 0.0, "step": 8918 }, { "epoch": 2.06, "learning_rate": 1.5645944072105386e-05, "loss": 0.0, "step": 8919 }, { "epoch": 2.06, "learning_rate": 1.564209228872968e-05, "loss": 0.0, "step": 8920 }, { "epoch": 2.06, "learning_rate": 1.563824050535398e-05, "loss": 0.0, "step": 8921 }, { "epoch": 2.06, "learning_rate": 1.5634388721978278e-05, "loss": 0.0, "step": 8922 }, { "epoch": 2.06, "learning_rate": 1.5630536938602573e-05, "loss": 0.0, "step": 8923 }, { "epoch": 2.06, "learning_rate": 1.5626685155226868e-05, "loss": 0.0, "step": 8924 }, { "epoch": 2.06, "learning_rate": 1.5622833371851166e-05, "loss": 0.0, "step": 8925 }, { "epoch": 2.06, "learning_rate": 1.5618981588475465e-05, "loss": 0.0, "step": 8926 }, { "epoch": 2.06, "learning_rate": 1.5615129805099763e-05, "loss": 0.0, "step": 8927 }, { "epoch": 2.06, "learning_rate": 1.561127802172406e-05, "loss": 0.0, "step": 8928 }, { "epoch": 2.06, "learning_rate": 1.5607426238348356e-05, "loss": 0.0, "step": 8929 }, { "epoch": 2.06, "learning_rate": 1.560357445497265e-05, "loss": 0.0, "step": 8930 }, { "epoch": 2.06, "learning_rate": 1.559972267159695e-05, "loss": 0.0, "step": 8931 }, { "epoch": 2.06, "learning_rate": 1.5595870888221248e-05, "loss": 0.0, "step": 8932 }, { "epoch": 2.06, "learning_rate": 1.5592019104845546e-05, "loss": 0.0, "step": 8933 }, { "epoch": 2.06, "learning_rate": 1.558816732146984e-05, "loss": 0.0, "step": 8934 }, { "epoch": 2.06, "learning_rate": 1.558431553809414e-05, "loss": 0.0, "step": 8935 }, { "epoch": 2.07, "learning_rate": 1.5580463754718434e-05, "loss": 0.0, "step": 8936 }, { "epoch": 2.07, "learning_rate": 1.5576611971342733e-05, "loss": 0.0, "step": 8937 }, { "epoch": 2.07, "learning_rate": 1.5572760187967028e-05, "loss": 0.0, "step": 8938 }, { "epoch": 2.07, "learning_rate": 1.5568908404591326e-05, "loss": 0.0, "step": 8939 }, { "epoch": 2.07, "learning_rate": 1.5565056621215624e-05, "loss": 0.0, "step": 8940 }, { "epoch": 2.07, "learning_rate": 1.5561204837839923e-05, "loss": 0.0, "step": 8941 }, { "epoch": 2.07, "learning_rate": 1.5557353054464218e-05, "loss": 0.0, "step": 8942 }, { "epoch": 2.07, "learning_rate": 1.5553501271088513e-05, "loss": 0.0, "step": 8943 }, { "epoch": 2.07, "learning_rate": 1.554964948771281e-05, "loss": 0.0, "step": 8944 }, { "epoch": 2.07, "learning_rate": 1.554579770433711e-05, "loss": 0.0001, "step": 8945 }, { "epoch": 2.07, "learning_rate": 1.5541945920961408e-05, "loss": 0.0, "step": 8946 }, { "epoch": 2.07, "learning_rate": 1.5538094137585703e-05, "loss": 0.0, "step": 8947 }, { "epoch": 2.07, "learning_rate": 1.5534242354209998e-05, "loss": 0.0, "step": 8948 }, { "epoch": 2.07, "learning_rate": 1.5530390570834296e-05, "loss": 0.0, "step": 8949 }, { "epoch": 2.07, "learning_rate": 1.5526538787458594e-05, "loss": 0.0, "step": 8950 }, { "epoch": 2.07, "learning_rate": 1.5522687004082893e-05, "loss": 0.0, "step": 8951 }, { "epoch": 2.07, "learning_rate": 1.5518835220707188e-05, "loss": 0.0, "step": 8952 }, { "epoch": 2.07, "learning_rate": 1.5514983437331486e-05, "loss": 0.0, "step": 8953 }, { "epoch": 2.07, "learning_rate": 1.551113165395578e-05, "loss": 0.0, "step": 8954 }, { "epoch": 2.07, "learning_rate": 1.550727987058008e-05, "loss": 0.0, "step": 8955 }, { "epoch": 2.07, "learning_rate": 1.5503428087204378e-05, "loss": 0.0, "step": 8956 }, { "epoch": 2.07, "learning_rate": 1.5499576303828673e-05, "loss": 0.0, "step": 8957 }, { "epoch": 2.07, "learning_rate": 1.549572452045297e-05, "loss": 0.0, "step": 8958 }, { "epoch": 2.07, "learning_rate": 1.549187273707727e-05, "loss": 0.0, "step": 8959 }, { "epoch": 2.07, "learning_rate": 1.5488020953701564e-05, "loss": 0.0, "step": 8960 }, { "epoch": 2.07, "learning_rate": 1.5484169170325863e-05, "loss": 0.0, "step": 8961 }, { "epoch": 2.07, "learning_rate": 1.5480317386950158e-05, "loss": 0.0, "step": 8962 }, { "epoch": 2.07, "learning_rate": 1.5476465603574456e-05, "loss": 0.0, "step": 8963 }, { "epoch": 2.07, "learning_rate": 1.5472613820198754e-05, "loss": 0.0, "step": 8964 }, { "epoch": 2.07, "learning_rate": 1.546876203682305e-05, "loss": 0.0, "step": 8965 }, { "epoch": 2.07, "learning_rate": 1.5464910253447344e-05, "loss": 0.0, "step": 8966 }, { "epoch": 2.07, "learning_rate": 1.5461058470071642e-05, "loss": 0.0, "step": 8967 }, { "epoch": 2.07, "learning_rate": 1.545720668669594e-05, "loss": 0.0, "step": 8968 }, { "epoch": 2.07, "learning_rate": 1.545335490332024e-05, "loss": 0.0, "step": 8969 }, { "epoch": 2.07, "learning_rate": 1.5449503119944537e-05, "loss": 0.0, "step": 8970 }, { "epoch": 2.07, "learning_rate": 1.5445651336568832e-05, "loss": 0.0, "step": 8971 }, { "epoch": 2.07, "learning_rate": 1.5441799553193127e-05, "loss": 0.0, "step": 8972 }, { "epoch": 2.07, "learning_rate": 1.5437947769817426e-05, "loss": 0.0, "step": 8973 }, { "epoch": 2.07, "learning_rate": 1.5434095986441724e-05, "loss": 0.0, "step": 8974 }, { "epoch": 2.07, "learning_rate": 1.5430244203066022e-05, "loss": 0.0, "step": 8975 }, { "epoch": 2.07, "learning_rate": 1.5426392419690317e-05, "loss": 0.0, "step": 8976 }, { "epoch": 2.07, "learning_rate": 1.5422540636314616e-05, "loss": 0.0, "step": 8977 }, { "epoch": 2.07, "learning_rate": 1.541868885293891e-05, "loss": 0.0, "step": 8978 }, { "epoch": 2.08, "learning_rate": 1.541483706956321e-05, "loss": 0.0, "step": 8979 }, { "epoch": 2.08, "learning_rate": 1.5410985286187504e-05, "loss": 0.0, "step": 8980 }, { "epoch": 2.08, "learning_rate": 1.5407133502811802e-05, "loss": 0.0, "step": 8981 }, { "epoch": 2.08, "learning_rate": 1.54032817194361e-05, "loss": 0.0, "step": 8982 }, { "epoch": 2.08, "learning_rate": 1.5399429936060396e-05, "loss": 0.0, "step": 8983 }, { "epoch": 2.08, "learning_rate": 1.5395578152684694e-05, "loss": 0.0, "step": 8984 }, { "epoch": 2.08, "learning_rate": 1.539172636930899e-05, "loss": 0.0, "step": 8985 }, { "epoch": 2.08, "learning_rate": 1.5387874585933287e-05, "loss": 0.0, "step": 8986 }, { "epoch": 2.08, "learning_rate": 1.5384022802557586e-05, "loss": 0.0, "step": 8987 }, { "epoch": 2.08, "learning_rate": 1.5380171019181884e-05, "loss": 0.0, "step": 8988 }, { "epoch": 2.08, "learning_rate": 1.537631923580618e-05, "loss": 0.0, "step": 8989 }, { "epoch": 2.08, "learning_rate": 1.5372467452430474e-05, "loss": 0.0, "step": 8990 }, { "epoch": 2.08, "learning_rate": 1.5368615669054772e-05, "loss": 0.0, "step": 8991 }, { "epoch": 2.08, "learning_rate": 1.536476388567907e-05, "loss": 0.0, "step": 8992 }, { "epoch": 2.08, "learning_rate": 1.536091210230337e-05, "loss": 0.0, "step": 8993 }, { "epoch": 2.08, "learning_rate": 1.5357060318927664e-05, "loss": 0.0, "step": 8994 }, { "epoch": 2.08, "learning_rate": 1.5353208535551962e-05, "loss": 0.0, "step": 8995 }, { "epoch": 2.08, "learning_rate": 1.5349356752176257e-05, "loss": 0.0, "step": 8996 }, { "epoch": 2.08, "learning_rate": 1.5345504968800556e-05, "loss": 0.0, "step": 8997 }, { "epoch": 2.08, "learning_rate": 1.5341653185424854e-05, "loss": 0.0, "step": 8998 }, { "epoch": 2.08, "learning_rate": 1.533780140204915e-05, "loss": 0.0, "step": 8999 }, { "epoch": 2.08, "learning_rate": 1.5333949618673447e-05, "loss": 0.0, "step": 9000 }, { "epoch": 2.08, "learning_rate": 1.5330097835297742e-05, "loss": 0.0, "step": 9001 }, { "epoch": 2.08, "learning_rate": 1.532624605192204e-05, "loss": 0.0, "step": 9002 }, { "epoch": 2.08, "learning_rate": 1.532239426854634e-05, "loss": 0.0, "step": 9003 }, { "epoch": 2.08, "learning_rate": 1.5318542485170634e-05, "loss": 0.0, "step": 9004 }, { "epoch": 2.08, "learning_rate": 1.5314690701794932e-05, "loss": 0.0, "step": 9005 }, { "epoch": 2.08, "learning_rate": 1.531083891841923e-05, "loss": 0.0, "step": 9006 }, { "epoch": 2.08, "learning_rate": 1.5306987135043525e-05, "loss": 0.0, "step": 9007 }, { "epoch": 2.08, "learning_rate": 1.530313535166782e-05, "loss": 0.0, "step": 9008 }, { "epoch": 2.08, "learning_rate": 1.529928356829212e-05, "loss": 0.0, "step": 9009 }, { "epoch": 2.08, "learning_rate": 1.5295431784916417e-05, "loss": 0.0, "step": 9010 }, { "epoch": 2.08, "learning_rate": 1.5291580001540715e-05, "loss": 0.0, "step": 9011 }, { "epoch": 2.08, "learning_rate": 1.5287728218165014e-05, "loss": 0.0, "step": 9012 }, { "epoch": 2.08, "learning_rate": 1.528387643478931e-05, "loss": 0.0, "step": 9013 }, { "epoch": 2.08, "learning_rate": 1.5280024651413604e-05, "loss": 0.0, "step": 9014 }, { "epoch": 2.08, "learning_rate": 1.5276172868037902e-05, "loss": 0.0, "step": 9015 }, { "epoch": 2.08, "learning_rate": 1.52723210846622e-05, "loss": 0.0, "step": 9016 }, { "epoch": 2.08, "learning_rate": 1.52684693012865e-05, "loss": 0.0, "step": 9017 }, { "epoch": 2.08, "learning_rate": 1.5264617517910794e-05, "loss": 0.0, "step": 9018 }, { "epoch": 2.08, "learning_rate": 1.526076573453509e-05, "loss": 0.0, "step": 9019 }, { "epoch": 2.08, "learning_rate": 1.5256913951159389e-05, "loss": 0.0, "step": 9020 }, { "epoch": 2.08, "learning_rate": 1.5253062167783685e-05, "loss": 0.0, "step": 9021 }, { "epoch": 2.09, "learning_rate": 1.524921038440798e-05, "loss": 0.0, "step": 9022 }, { "epoch": 2.09, "learning_rate": 1.5245358601032279e-05, "loss": 0.0, "step": 9023 }, { "epoch": 2.09, "learning_rate": 1.5241506817656575e-05, "loss": 0.0001, "step": 9024 }, { "epoch": 2.09, "learning_rate": 1.5237655034280874e-05, "loss": 0.0, "step": 9025 }, { "epoch": 2.09, "learning_rate": 1.523380325090517e-05, "loss": 0.0, "step": 9026 }, { "epoch": 2.09, "learning_rate": 1.5229951467529465e-05, "loss": 0.0, "step": 9027 }, { "epoch": 2.09, "learning_rate": 1.5226099684153763e-05, "loss": 0.0, "step": 9028 }, { "epoch": 2.09, "learning_rate": 1.5222247900778062e-05, "loss": 0.0, "step": 9029 }, { "epoch": 2.09, "learning_rate": 1.5218396117402358e-05, "loss": 0.0, "step": 9030 }, { "epoch": 2.09, "learning_rate": 1.5214544334026657e-05, "loss": 0.0, "step": 9031 }, { "epoch": 2.09, "learning_rate": 1.5210692550650952e-05, "loss": 0.0, "step": 9032 }, { "epoch": 2.09, "learning_rate": 1.5206840767275248e-05, "loss": 0.0, "step": 9033 }, { "epoch": 2.09, "learning_rate": 1.5202988983899547e-05, "loss": 0.0, "step": 9034 }, { "epoch": 2.09, "learning_rate": 1.5199137200523843e-05, "loss": 0.0, "step": 9035 }, { "epoch": 2.09, "learning_rate": 1.5195285417148138e-05, "loss": 0.0, "step": 9036 }, { "epoch": 2.09, "learning_rate": 1.5191433633772437e-05, "loss": 0.0, "step": 9037 }, { "epoch": 2.09, "learning_rate": 1.5187581850396735e-05, "loss": 0.0, "step": 9038 }, { "epoch": 2.09, "learning_rate": 1.5183730067021032e-05, "loss": 0.0, "step": 9039 }, { "epoch": 2.09, "learning_rate": 1.517987828364533e-05, "loss": 0.0, "step": 9040 }, { "epoch": 2.09, "learning_rate": 1.5176026500269625e-05, "loss": 0.0, "step": 9041 }, { "epoch": 2.09, "learning_rate": 1.5172174716893922e-05, "loss": 0.0, "step": 9042 }, { "epoch": 2.09, "learning_rate": 1.516832293351822e-05, "loss": 0.0, "step": 9043 }, { "epoch": 2.09, "learning_rate": 1.5164471150142517e-05, "loss": 0.0, "step": 9044 }, { "epoch": 2.09, "learning_rate": 1.5160619366766815e-05, "loss": 0.0, "step": 9045 }, { "epoch": 2.09, "learning_rate": 1.515676758339111e-05, "loss": 0.0, "step": 9046 }, { "epoch": 2.09, "learning_rate": 1.5152915800015408e-05, "loss": 0.0, "step": 9047 }, { "epoch": 2.09, "learning_rate": 1.5149064016639705e-05, "loss": 0.0, "step": 9048 }, { "epoch": 2.09, "learning_rate": 1.5145212233264003e-05, "loss": 0.0, "step": 9049 }, { "epoch": 2.09, "learning_rate": 1.5141360449888298e-05, "loss": 0.0, "step": 9050 }, { "epoch": 2.09, "learning_rate": 1.5137508666512595e-05, "loss": 0.0, "step": 9051 }, { "epoch": 2.09, "learning_rate": 1.5133656883136893e-05, "loss": 0.0, "step": 9052 }, { "epoch": 2.09, "learning_rate": 1.512980509976119e-05, "loss": 0.0, "step": 9053 }, { "epoch": 2.09, "learning_rate": 1.5125953316385488e-05, "loss": 0.0, "step": 9054 }, { "epoch": 2.09, "learning_rate": 1.5122101533009783e-05, "loss": 0.0, "step": 9055 }, { "epoch": 2.09, "learning_rate": 1.5118249749634082e-05, "loss": 0.0, "step": 9056 }, { "epoch": 2.09, "learning_rate": 1.5114397966258378e-05, "loss": 0.0, "step": 9057 }, { "epoch": 2.09, "learning_rate": 1.5110546182882677e-05, "loss": 0.0, "step": 9058 }, { "epoch": 2.09, "learning_rate": 1.5106694399506973e-05, "loss": 0.0, "step": 9059 }, { "epoch": 2.09, "learning_rate": 1.5102842616131268e-05, "loss": 0.0, "step": 9060 }, { "epoch": 2.09, "learning_rate": 1.5098990832755566e-05, "loss": 0.0, "step": 9061 }, { "epoch": 2.09, "learning_rate": 1.5095139049379863e-05, "loss": 0.0, "step": 9062 }, { "epoch": 2.09, "learning_rate": 1.5091287266004161e-05, "loss": 0.0, "step": 9063 }, { "epoch": 2.09, "learning_rate": 1.5087435482628456e-05, "loss": 0.0, "step": 9064 }, { "epoch": 2.09, "learning_rate": 1.5083583699252755e-05, "loss": 0.0, "step": 9065 }, { "epoch": 2.1, "learning_rate": 1.5079731915877051e-05, "loss": 0.0, "step": 9066 }, { "epoch": 2.1, "learning_rate": 1.507588013250135e-05, "loss": 0.0, "step": 9067 }, { "epoch": 2.1, "learning_rate": 1.5072028349125646e-05, "loss": 0.0, "step": 9068 }, { "epoch": 2.1, "learning_rate": 1.5068176565749941e-05, "loss": 0.0, "step": 9069 }, { "epoch": 2.1, "learning_rate": 1.506432478237424e-05, "loss": 0.0, "step": 9070 }, { "epoch": 2.1, "learning_rate": 1.5060472998998536e-05, "loss": 0.0, "step": 9071 }, { "epoch": 2.1, "learning_rate": 1.5056621215622835e-05, "loss": 0.0, "step": 9072 }, { "epoch": 2.1, "learning_rate": 1.5052769432247133e-05, "loss": 0.0, "step": 9073 }, { "epoch": 2.1, "learning_rate": 1.5048917648871428e-05, "loss": 0.0, "step": 9074 }, { "epoch": 2.1, "learning_rate": 1.5045065865495725e-05, "loss": 0.0, "step": 9075 }, { "epoch": 2.1, "learning_rate": 1.5041214082120023e-05, "loss": 0.0, "step": 9076 }, { "epoch": 2.1, "learning_rate": 1.503736229874432e-05, "loss": 0.0, "step": 9077 }, { "epoch": 2.1, "learning_rate": 1.5033510515368615e-05, "loss": 0.0, "step": 9078 }, { "epoch": 2.1, "learning_rate": 1.5029658731992913e-05, "loss": 0.0, "step": 9079 }, { "epoch": 2.1, "learning_rate": 1.502580694861721e-05, "loss": 0.0, "step": 9080 }, { "epoch": 2.1, "learning_rate": 1.5021955165241508e-05, "loss": 0.0, "step": 9081 }, { "epoch": 2.1, "learning_rate": 1.5018103381865806e-05, "loss": 0.0, "step": 9082 }, { "epoch": 2.1, "learning_rate": 1.5014251598490101e-05, "loss": 0.0, "step": 9083 }, { "epoch": 2.1, "learning_rate": 1.5010399815114398e-05, "loss": 0.0, "step": 9084 }, { "epoch": 2.1, "learning_rate": 1.5006548031738696e-05, "loss": 0.0002, "step": 9085 }, { "epoch": 2.1, "learning_rate": 1.5002696248362993e-05, "loss": 0.0, "step": 9086 }, { "epoch": 2.1, "learning_rate": 1.4998844464987291e-05, "loss": 0.0, "step": 9087 }, { "epoch": 2.1, "learning_rate": 1.4994992681611586e-05, "loss": 0.0001, "step": 9088 }, { "epoch": 2.1, "learning_rate": 1.4991140898235883e-05, "loss": 0.0, "step": 9089 }, { "epoch": 2.1, "learning_rate": 1.4987289114860181e-05, "loss": 0.0, "step": 9090 }, { "epoch": 2.1, "learning_rate": 1.498343733148448e-05, "loss": 0.0, "step": 9091 }, { "epoch": 2.1, "learning_rate": 1.4979585548108774e-05, "loss": 0.0, "step": 9092 }, { "epoch": 2.1, "learning_rate": 1.4975733764733071e-05, "loss": 0.0, "step": 9093 }, { "epoch": 2.1, "learning_rate": 1.497188198135737e-05, "loss": 0.0, "step": 9094 }, { "epoch": 2.1, "learning_rate": 1.4968030197981666e-05, "loss": 0.0002, "step": 9095 }, { "epoch": 2.1, "learning_rate": 1.4964178414605964e-05, "loss": 0.0, "step": 9096 }, { "epoch": 2.1, "learning_rate": 1.496032663123026e-05, "loss": 0.0, "step": 9097 }, { "epoch": 2.1, "learning_rate": 1.4956474847854556e-05, "loss": 0.0, "step": 9098 }, { "epoch": 2.1, "learning_rate": 1.4952623064478854e-05, "loss": 0.0, "step": 9099 }, { "epoch": 2.1, "learning_rate": 1.4948771281103153e-05, "loss": 0.0, "step": 9100 }, { "epoch": 2.1, "learning_rate": 1.494491949772745e-05, "loss": 0.0, "step": 9101 }, { "epoch": 2.1, "learning_rate": 1.4941067714351744e-05, "loss": 0.0, "step": 9102 }, { "epoch": 2.1, "learning_rate": 1.4937215930976043e-05, "loss": 0.0, "step": 9103 }, { "epoch": 2.1, "learning_rate": 1.493336414760034e-05, "loss": 0.0, "step": 9104 }, { "epoch": 2.1, "learning_rate": 1.4929512364224638e-05, "loss": 0.0, "step": 9105 }, { "epoch": 2.1, "learning_rate": 1.4925660580848933e-05, "loss": 0.0, "step": 9106 }, { "epoch": 2.1, "learning_rate": 1.492180879747323e-05, "loss": 0.0, "step": 9107 }, { "epoch": 2.1, "learning_rate": 1.4917957014097528e-05, "loss": 0.0, "step": 9108 }, { "epoch": 2.11, "learning_rate": 1.4914105230721826e-05, "loss": 0.0, "step": 9109 }, { "epoch": 2.11, "learning_rate": 1.4910253447346123e-05, "loss": 0.0, "step": 9110 }, { "epoch": 2.11, "learning_rate": 1.4906401663970418e-05, "loss": 0.0, "step": 9111 }, { "epoch": 2.11, "learning_rate": 1.4902549880594716e-05, "loss": 0.0, "step": 9112 }, { "epoch": 2.11, "learning_rate": 1.4898698097219013e-05, "loss": 0.0, "step": 9113 }, { "epoch": 2.11, "learning_rate": 1.4894846313843311e-05, "loss": 0.0, "step": 9114 }, { "epoch": 2.11, "learning_rate": 1.489099453046761e-05, "loss": 0.0, "step": 9115 }, { "epoch": 2.11, "learning_rate": 1.4887142747091903e-05, "loss": 0.0, "step": 9116 }, { "epoch": 2.11, "learning_rate": 1.4883290963716201e-05, "loss": 0.0, "step": 9117 }, { "epoch": 2.11, "learning_rate": 1.48794391803405e-05, "loss": 0.0, "step": 9118 }, { "epoch": 2.11, "learning_rate": 1.4875587396964796e-05, "loss": 0.0, "step": 9119 }, { "epoch": 2.11, "learning_rate": 1.487173561358909e-05, "loss": 0.0, "step": 9120 }, { "epoch": 2.11, "learning_rate": 1.486788383021339e-05, "loss": 0.0, "step": 9121 }, { "epoch": 2.11, "learning_rate": 1.4864032046837686e-05, "loss": 0.0, "step": 9122 }, { "epoch": 2.11, "learning_rate": 1.4860180263461984e-05, "loss": 0.0, "step": 9123 }, { "epoch": 2.11, "learning_rate": 1.4856328480086283e-05, "loss": 0.0, "step": 9124 }, { "epoch": 2.11, "learning_rate": 1.4852476696710576e-05, "loss": 0.0, "step": 9125 }, { "epoch": 2.11, "learning_rate": 1.4848624913334874e-05, "loss": 0.0, "step": 9126 }, { "epoch": 2.11, "learning_rate": 1.4844773129959172e-05, "loss": 0.0, "step": 9127 }, { "epoch": 2.11, "learning_rate": 1.4840921346583469e-05, "loss": 0.0, "step": 9128 }, { "epoch": 2.11, "learning_rate": 1.4837069563207767e-05, "loss": 0.0001, "step": 9129 }, { "epoch": 2.11, "learning_rate": 1.4833217779832062e-05, "loss": 0.0, "step": 9130 }, { "epoch": 2.11, "learning_rate": 1.4829365996456359e-05, "loss": 0.0, "step": 9131 }, { "epoch": 2.11, "learning_rate": 1.4825514213080657e-05, "loss": 0.0, "step": 9132 }, { "epoch": 2.11, "learning_rate": 1.4821662429704956e-05, "loss": 0.0, "step": 9133 }, { "epoch": 2.11, "learning_rate": 1.481781064632925e-05, "loss": 0.0, "step": 9134 }, { "epoch": 2.11, "learning_rate": 1.4813958862953547e-05, "loss": 0.0, "step": 9135 }, { "epoch": 2.11, "learning_rate": 1.4810107079577846e-05, "loss": 0.0, "step": 9136 }, { "epoch": 2.11, "learning_rate": 1.4806255296202142e-05, "loss": 0.0, "step": 9137 }, { "epoch": 2.11, "learning_rate": 1.480240351282644e-05, "loss": 0.0, "step": 9138 }, { "epoch": 2.11, "learning_rate": 1.4798551729450736e-05, "loss": 0.0, "step": 9139 }, { "epoch": 2.11, "learning_rate": 1.4794699946075032e-05, "loss": 0.0, "step": 9140 }, { "epoch": 2.11, "learning_rate": 1.479084816269933e-05, "loss": 0.0, "step": 9141 }, { "epoch": 2.11, "learning_rate": 1.4786996379323629e-05, "loss": 0.0, "step": 9142 }, { "epoch": 2.11, "learning_rate": 1.4783144595947926e-05, "loss": 0.0, "step": 9143 }, { "epoch": 2.11, "learning_rate": 1.477929281257222e-05, "loss": 0.0, "step": 9144 }, { "epoch": 2.11, "learning_rate": 1.4775441029196519e-05, "loss": 0.0, "step": 9145 }, { "epoch": 2.11, "learning_rate": 1.4771589245820816e-05, "loss": 0.0, "step": 9146 }, { "epoch": 2.11, "learning_rate": 1.4767737462445114e-05, "loss": 0.0001, "step": 9147 }, { "epoch": 2.11, "learning_rate": 1.4763885679069409e-05, "loss": 0.0, "step": 9148 }, { "epoch": 2.11, "learning_rate": 1.4760033895693706e-05, "loss": 0.0, "step": 9149 }, { "epoch": 2.11, "learning_rate": 1.4756182112318004e-05, "loss": 0.0, "step": 9150 }, { "epoch": 2.11, "learning_rate": 1.4752330328942302e-05, "loss": 0.0, "step": 9151 }, { "epoch": 2.12, "learning_rate": 1.4748478545566599e-05, "loss": 0.0, "step": 9152 }, { "epoch": 2.12, "learning_rate": 1.4744626762190894e-05, "loss": 0.0, "step": 9153 }, { "epoch": 2.12, "learning_rate": 1.4740774978815192e-05, "loss": 0.024, "step": 9154 }, { "epoch": 2.12, "learning_rate": 1.4736923195439489e-05, "loss": 0.0, "step": 9155 }, { "epoch": 2.12, "learning_rate": 1.4733071412063787e-05, "loss": 0.0, "step": 9156 }, { "epoch": 2.12, "learning_rate": 1.4729219628688084e-05, "loss": 0.0, "step": 9157 }, { "epoch": 2.12, "learning_rate": 1.4725367845312379e-05, "loss": 0.0, "step": 9158 }, { "epoch": 2.12, "learning_rate": 1.4721516061936677e-05, "loss": 0.0, "step": 9159 }, { "epoch": 2.12, "learning_rate": 1.4717664278560975e-05, "loss": 0.0, "step": 9160 }, { "epoch": 2.12, "learning_rate": 1.4713812495185272e-05, "loss": 0.0, "step": 9161 }, { "epoch": 2.12, "learning_rate": 1.4709960711809567e-05, "loss": 0.0, "step": 9162 }, { "epoch": 2.12, "learning_rate": 1.4706108928433865e-05, "loss": 0.0, "step": 9163 }, { "epoch": 2.12, "learning_rate": 1.4702257145058162e-05, "loss": 0.0, "step": 9164 }, { "epoch": 2.12, "learning_rate": 1.469840536168246e-05, "loss": 0.0, "step": 9165 }, { "epoch": 2.12, "learning_rate": 1.4694553578306757e-05, "loss": 0.0, "step": 9166 }, { "epoch": 2.12, "learning_rate": 1.4690701794931052e-05, "loss": 0.0, "step": 9167 }, { "epoch": 2.12, "learning_rate": 1.468685001155535e-05, "loss": 0.0, "step": 9168 }, { "epoch": 2.12, "learning_rate": 1.4682998228179649e-05, "loss": 0.0, "step": 9169 }, { "epoch": 2.12, "learning_rate": 1.4679146444803945e-05, "loss": 0.0, "step": 9170 }, { "epoch": 2.12, "learning_rate": 1.4675294661428244e-05, "loss": 0.0, "step": 9171 }, { "epoch": 2.12, "learning_rate": 1.4671442878052539e-05, "loss": 0.0, "step": 9172 }, { "epoch": 2.12, "learning_rate": 1.4667591094676835e-05, "loss": 0.0, "step": 9173 }, { "epoch": 2.12, "learning_rate": 1.4663739311301134e-05, "loss": 0.0, "step": 9174 }, { "epoch": 2.12, "learning_rate": 1.465988752792543e-05, "loss": 0.0, "step": 9175 }, { "epoch": 2.12, "learning_rate": 1.4656035744549725e-05, "loss": 0.0, "step": 9176 }, { "epoch": 2.12, "learning_rate": 1.4652183961174024e-05, "loss": 0.0, "step": 9177 }, { "epoch": 2.12, "learning_rate": 1.4648332177798322e-05, "loss": 0.0, "step": 9178 }, { "epoch": 2.12, "learning_rate": 1.4644480394422619e-05, "loss": 0.0, "step": 9179 }, { "epoch": 2.12, "learning_rate": 1.4640628611046917e-05, "loss": 0.0, "step": 9180 }, { "epoch": 2.12, "learning_rate": 1.4636776827671212e-05, "loss": 0.0, "step": 9181 }, { "epoch": 2.12, "learning_rate": 1.4632925044295509e-05, "loss": 0.0, "step": 9182 }, { "epoch": 2.12, "learning_rate": 1.4629073260919807e-05, "loss": 0.0, "step": 9183 }, { "epoch": 2.12, "learning_rate": 1.4625221477544104e-05, "loss": 0.0, "step": 9184 }, { "epoch": 2.12, "learning_rate": 1.4621369694168402e-05, "loss": 0.0, "step": 9185 }, { "epoch": 2.12, "learning_rate": 1.4617517910792697e-05, "loss": 0.0, "step": 9186 }, { "epoch": 2.12, "learning_rate": 1.4613666127416995e-05, "loss": 0.0, "step": 9187 }, { "epoch": 2.12, "learning_rate": 1.4609814344041292e-05, "loss": 0.0, "step": 9188 }, { "epoch": 2.12, "learning_rate": 1.460596256066559e-05, "loss": 0.0, "step": 9189 }, { "epoch": 2.12, "learning_rate": 1.4602110777289885e-05, "loss": 0.0, "step": 9190 }, { "epoch": 2.12, "learning_rate": 1.4598258993914182e-05, "loss": 0.0, "step": 9191 }, { "epoch": 2.12, "learning_rate": 1.459440721053848e-05, "loss": 0.0, "step": 9192 }, { "epoch": 2.12, "learning_rate": 1.4590555427162777e-05, "loss": 0.0, "step": 9193 }, { "epoch": 2.12, "learning_rate": 1.4586703643787075e-05, "loss": 0.0, "step": 9194 }, { "epoch": 2.13, "learning_rate": 1.458285186041137e-05, "loss": 0.0, "step": 9195 }, { "epoch": 2.13, "learning_rate": 1.4579000077035668e-05, "loss": 0.0, "step": 9196 }, { "epoch": 2.13, "learning_rate": 1.4575148293659965e-05, "loss": 0.0, "step": 9197 }, { "epoch": 2.13, "learning_rate": 1.4571296510284263e-05, "loss": 0.0, "step": 9198 }, { "epoch": 2.13, "learning_rate": 1.456744472690856e-05, "loss": 0.0, "step": 9199 }, { "epoch": 2.13, "learning_rate": 1.4563592943532855e-05, "loss": 0.0001, "step": 9200 }, { "epoch": 2.13, "learning_rate": 1.4559741160157153e-05, "loss": 0.0, "step": 9201 }, { "epoch": 2.13, "learning_rate": 1.455588937678145e-05, "loss": 0.0001, "step": 9202 }, { "epoch": 2.13, "learning_rate": 1.4552037593405748e-05, "loss": 0.0, "step": 9203 }, { "epoch": 2.13, "learning_rate": 1.4548185810030043e-05, "loss": 0.0, "step": 9204 }, { "epoch": 2.13, "learning_rate": 1.4544334026654342e-05, "loss": 0.0001, "step": 9205 }, { "epoch": 2.13, "learning_rate": 1.4540482243278638e-05, "loss": 0.0, "step": 9206 }, { "epoch": 2.13, "learning_rate": 1.4536630459902937e-05, "loss": 0.0, "step": 9207 }, { "epoch": 2.13, "learning_rate": 1.4532778676527233e-05, "loss": 0.0, "step": 9208 }, { "epoch": 2.13, "learning_rate": 1.4528926893151528e-05, "loss": 0.0, "step": 9209 }, { "epoch": 2.13, "learning_rate": 1.4525075109775827e-05, "loss": 0.0001, "step": 9210 }, { "epoch": 2.13, "learning_rate": 1.4521223326400123e-05, "loss": 0.0, "step": 9211 }, { "epoch": 2.13, "learning_rate": 1.4517371543024422e-05, "loss": 0.0001, "step": 9212 }, { "epoch": 2.13, "learning_rate": 1.451351975964872e-05, "loss": 0.0, "step": 9213 }, { "epoch": 2.13, "learning_rate": 1.4509667976273015e-05, "loss": 0.0, "step": 9214 }, { "epoch": 2.13, "learning_rate": 1.4505816192897312e-05, "loss": 0.0, "step": 9215 }, { "epoch": 2.13, "learning_rate": 1.450196440952161e-05, "loss": 0.0, "step": 9216 }, { "epoch": 2.13, "learning_rate": 1.4498112626145907e-05, "loss": 0.0, "step": 9217 }, { "epoch": 2.13, "learning_rate": 1.4494260842770201e-05, "loss": 0.0, "step": 9218 }, { "epoch": 2.13, "learning_rate": 1.44904090593945e-05, "loss": 0.0, "step": 9219 }, { "epoch": 2.13, "learning_rate": 1.4486557276018798e-05, "loss": 0.0, "step": 9220 }, { "epoch": 2.13, "learning_rate": 1.4482705492643095e-05, "loss": 0.0, "step": 9221 }, { "epoch": 2.13, "learning_rate": 1.4478853709267393e-05, "loss": 0.0, "step": 9222 }, { "epoch": 2.13, "learning_rate": 1.4475001925891688e-05, "loss": 0.0, "step": 9223 }, { "epoch": 2.13, "learning_rate": 1.4471150142515985e-05, "loss": 0.0, "step": 9224 }, { "epoch": 2.13, "learning_rate": 1.4467298359140283e-05, "loss": 0.0, "step": 9225 }, { "epoch": 2.13, "learning_rate": 1.446344657576458e-05, "loss": 0.0, "step": 9226 }, { "epoch": 2.13, "learning_rate": 1.4459594792388878e-05, "loss": 0.0, "step": 9227 }, { "epoch": 2.13, "learning_rate": 1.4455743009013173e-05, "loss": 0.0, "step": 9228 }, { "epoch": 2.13, "learning_rate": 1.4451891225637471e-05, "loss": 0.0, "step": 9229 }, { "epoch": 2.13, "learning_rate": 1.4448039442261768e-05, "loss": 0.0001, "step": 9230 }, { "epoch": 2.13, "learning_rate": 1.4444187658886066e-05, "loss": 0.0, "step": 9231 }, { "epoch": 2.13, "learning_rate": 1.4440335875510361e-05, "loss": 0.0, "step": 9232 }, { "epoch": 2.13, "learning_rate": 1.4436484092134658e-05, "loss": 0.0, "step": 9233 }, { "epoch": 2.13, "learning_rate": 1.4432632308758956e-05, "loss": 0.0, "step": 9234 }, { "epoch": 2.13, "learning_rate": 1.4428780525383253e-05, "loss": 0.0001, "step": 9235 }, { "epoch": 2.13, "learning_rate": 1.4424928742007551e-05, "loss": 0.0, "step": 9236 }, { "epoch": 2.13, "learning_rate": 1.4421076958631846e-05, "loss": 0.0, "step": 9237 }, { "epoch": 2.13, "learning_rate": 1.4417225175256145e-05, "loss": 0.0001, "step": 9238 }, { "epoch": 2.14, "learning_rate": 1.4413373391880441e-05, "loss": 0.0, "step": 9239 }, { "epoch": 2.14, "learning_rate": 1.440952160850474e-05, "loss": 0.0, "step": 9240 }, { "epoch": 2.14, "learning_rate": 1.4405669825129036e-05, "loss": 0.0, "step": 9241 }, { "epoch": 2.14, "learning_rate": 1.4401818041753331e-05, "loss": 0.0, "step": 9242 }, { "epoch": 2.14, "learning_rate": 1.439796625837763e-05, "loss": 0.0, "step": 9243 }, { "epoch": 2.14, "learning_rate": 1.4394114475001926e-05, "loss": 0.0, "step": 9244 }, { "epoch": 2.14, "learning_rate": 1.4390262691626225e-05, "loss": 0.0, "step": 9245 }, { "epoch": 2.14, "learning_rate": 1.438641090825052e-05, "loss": 0.0, "step": 9246 }, { "epoch": 2.14, "learning_rate": 1.4382559124874818e-05, "loss": 0.0, "step": 9247 }, { "epoch": 2.14, "learning_rate": 1.4378707341499115e-05, "loss": 0.0, "step": 9248 }, { "epoch": 2.14, "learning_rate": 1.4374855558123413e-05, "loss": 0.0, "step": 9249 }, { "epoch": 2.14, "learning_rate": 1.437100377474771e-05, "loss": 0.0, "step": 9250 }, { "epoch": 2.14, "learning_rate": 1.4367151991372004e-05, "loss": 0.0, "step": 9251 }, { "epoch": 2.14, "learning_rate": 1.4363300207996303e-05, "loss": 0.0, "step": 9252 }, { "epoch": 2.14, "learning_rate": 1.43594484246206e-05, "loss": 0.0, "step": 9253 }, { "epoch": 2.14, "learning_rate": 1.4355596641244898e-05, "loss": 0.0, "step": 9254 }, { "epoch": 2.14, "learning_rate": 1.4351744857869196e-05, "loss": 0.0, "step": 9255 }, { "epoch": 2.14, "learning_rate": 1.4347893074493491e-05, "loss": 0.0173, "step": 9256 }, { "epoch": 2.14, "learning_rate": 1.4344041291117788e-05, "loss": 0.0, "step": 9257 }, { "epoch": 2.14, "learning_rate": 1.4340189507742086e-05, "loss": 0.0, "step": 9258 }, { "epoch": 2.14, "learning_rate": 1.4336337724366383e-05, "loss": 0.0, "step": 9259 }, { "epoch": 2.14, "learning_rate": 1.4332485940990678e-05, "loss": 0.0, "step": 9260 }, { "epoch": 2.14, "learning_rate": 1.4328634157614976e-05, "loss": 0.0, "step": 9261 }, { "epoch": 2.14, "learning_rate": 1.4324782374239273e-05, "loss": 0.0, "step": 9262 }, { "epoch": 2.14, "learning_rate": 1.4320930590863571e-05, "loss": 0.0, "step": 9263 }, { "epoch": 2.14, "learning_rate": 1.431707880748787e-05, "loss": 0.0, "step": 9264 }, { "epoch": 2.14, "learning_rate": 1.4313227024112164e-05, "loss": 0.0001, "step": 9265 }, { "epoch": 2.14, "learning_rate": 1.4309375240736461e-05, "loss": 0.0, "step": 9266 }, { "epoch": 2.14, "learning_rate": 1.430552345736076e-05, "loss": 0.0, "step": 9267 }, { "epoch": 2.14, "learning_rate": 1.4301671673985056e-05, "loss": 0.0001, "step": 9268 }, { "epoch": 2.14, "learning_rate": 1.4297819890609354e-05, "loss": 0.0, "step": 9269 }, { "epoch": 2.14, "learning_rate": 1.429396810723365e-05, "loss": 0.0, "step": 9270 }, { "epoch": 2.14, "learning_rate": 1.4290116323857946e-05, "loss": 0.0, "step": 9271 }, { "epoch": 2.14, "learning_rate": 1.4286264540482244e-05, "loss": 0.0, "step": 9272 }, { "epoch": 2.14, "learning_rate": 1.4282412757106543e-05, "loss": 0.0, "step": 9273 }, { "epoch": 2.14, "learning_rate": 1.4278560973730838e-05, "loss": 0.0, "step": 9274 }, { "epoch": 2.14, "learning_rate": 1.4274709190355134e-05, "loss": 0.0, "step": 9275 }, { "epoch": 2.14, "learning_rate": 1.4270857406979433e-05, "loss": 0.0, "step": 9276 }, { "epoch": 2.14, "learning_rate": 1.426700562360373e-05, "loss": 0.0, "step": 9277 }, { "epoch": 2.14, "learning_rate": 1.4263153840228028e-05, "loss": 0.0, "step": 9278 }, { "epoch": 2.14, "learning_rate": 1.4259302056852323e-05, "loss": 0.0, "step": 9279 }, { "epoch": 2.14, "learning_rate": 1.425545027347662e-05, "loss": 0.0, "step": 9280 }, { "epoch": 2.14, "learning_rate": 1.4251598490100917e-05, "loss": 0.0, "step": 9281 }, { "epoch": 2.15, "learning_rate": 1.4247746706725216e-05, "loss": 0.0, "step": 9282 }, { "epoch": 2.15, "learning_rate": 1.4243894923349512e-05, "loss": 0.0, "step": 9283 }, { "epoch": 2.15, "learning_rate": 1.4240043139973807e-05, "loss": 0.0, "step": 9284 }, { "epoch": 2.15, "learning_rate": 1.4236191356598106e-05, "loss": 0.0, "step": 9285 }, { "epoch": 2.15, "learning_rate": 1.4232339573222402e-05, "loss": 0.0002, "step": 9286 }, { "epoch": 2.15, "learning_rate": 1.42284877898467e-05, "loss": 0.0, "step": 9287 }, { "epoch": 2.15, "learning_rate": 1.4224636006470996e-05, "loss": 0.0, "step": 9288 }, { "epoch": 2.15, "learning_rate": 1.4220784223095292e-05, "loss": 0.0, "step": 9289 }, { "epoch": 2.15, "learning_rate": 1.421693243971959e-05, "loss": 0.0, "step": 9290 }, { "epoch": 2.15, "learning_rate": 1.4213080656343889e-05, "loss": 0.0, "step": 9291 }, { "epoch": 2.15, "learning_rate": 1.4209228872968186e-05, "loss": 0.0, "step": 9292 }, { "epoch": 2.15, "learning_rate": 1.420537708959248e-05, "loss": 0.0, "step": 9293 }, { "epoch": 2.15, "learning_rate": 1.4201525306216779e-05, "loss": 0.0, "step": 9294 }, { "epoch": 2.15, "learning_rate": 1.4197673522841076e-05, "loss": 0.0, "step": 9295 }, { "epoch": 2.15, "learning_rate": 1.4193821739465374e-05, "loss": 0.0, "step": 9296 }, { "epoch": 2.15, "learning_rate": 1.418996995608967e-05, "loss": 0.0, "step": 9297 }, { "epoch": 2.15, "learning_rate": 1.4186118172713966e-05, "loss": 0.0, "step": 9298 }, { "epoch": 2.15, "learning_rate": 1.4182266389338264e-05, "loss": 0.0, "step": 9299 }, { "epoch": 2.15, "learning_rate": 1.4178414605962562e-05, "loss": 0.0, "step": 9300 }, { "epoch": 2.15, "learning_rate": 1.4174562822586859e-05, "loss": 0.0, "step": 9301 }, { "epoch": 2.15, "learning_rate": 1.4170711039211154e-05, "loss": 0.0, "step": 9302 }, { "epoch": 2.15, "learning_rate": 1.4166859255835452e-05, "loss": 0.0001, "step": 9303 }, { "epoch": 2.15, "learning_rate": 1.4163007472459749e-05, "loss": 0.0, "step": 9304 }, { "epoch": 2.15, "learning_rate": 1.4159155689084047e-05, "loss": 0.0, "step": 9305 }, { "epoch": 2.15, "learning_rate": 1.4155303905708344e-05, "loss": 0.0, "step": 9306 }, { "epoch": 2.15, "learning_rate": 1.4151452122332639e-05, "loss": 0.0, "step": 9307 }, { "epoch": 2.15, "learning_rate": 1.4147600338956937e-05, "loss": 0.0, "step": 9308 }, { "epoch": 2.15, "learning_rate": 1.4143748555581236e-05, "loss": 0.0, "step": 9309 }, { "epoch": 2.15, "learning_rate": 1.4139896772205532e-05, "loss": 0.0, "step": 9310 }, { "epoch": 2.15, "learning_rate": 1.413604498882983e-05, "loss": 0.0, "step": 9311 }, { "epoch": 2.15, "learning_rate": 1.4132193205454125e-05, "loss": 0.0, "step": 9312 }, { "epoch": 2.15, "learning_rate": 1.4128341422078422e-05, "loss": 0.0, "step": 9313 }, { "epoch": 2.15, "learning_rate": 1.412448963870272e-05, "loss": 0.0, "step": 9314 }, { "epoch": 2.15, "learning_rate": 1.4120637855327019e-05, "loss": 0.0, "step": 9315 }, { "epoch": 2.15, "learning_rate": 1.4116786071951312e-05, "loss": 0.0, "step": 9316 }, { "epoch": 2.15, "learning_rate": 1.411293428857561e-05, "loss": 0.0, "step": 9317 }, { "epoch": 2.15, "learning_rate": 1.4109082505199909e-05, "loss": 0.0, "step": 9318 }, { "epoch": 2.15, "learning_rate": 1.4105230721824205e-05, "loss": 0.0, "step": 9319 }, { "epoch": 2.15, "learning_rate": 1.4101378938448504e-05, "loss": 0.0, "step": 9320 }, { "epoch": 2.15, "learning_rate": 1.4097527155072799e-05, "loss": 0.0, "step": 9321 }, { "epoch": 2.15, "learning_rate": 1.4093675371697095e-05, "loss": 0.0, "step": 9322 }, { "epoch": 2.15, "learning_rate": 1.4089823588321394e-05, "loss": 0.0, "step": 9323 }, { "epoch": 2.15, "learning_rate": 1.4085971804945692e-05, "loss": 0.0, "step": 9324 }, { "epoch": 2.16, "learning_rate": 1.4082120021569989e-05, "loss": 0.0, "step": 9325 }, { "epoch": 2.16, "learning_rate": 1.4078268238194284e-05, "loss": 0.0, "step": 9326 }, { "epoch": 2.16, "learning_rate": 1.4074416454818582e-05, "loss": 0.0, "step": 9327 }, { "epoch": 2.16, "learning_rate": 1.4070564671442879e-05, "loss": 0.0, "step": 9328 }, { "epoch": 2.16, "learning_rate": 1.4066712888067177e-05, "loss": 0.0, "step": 9329 }, { "epoch": 2.16, "learning_rate": 1.4062861104691472e-05, "loss": 0.0, "step": 9330 }, { "epoch": 2.16, "learning_rate": 1.4059009321315769e-05, "loss": 0.0, "step": 9331 }, { "epoch": 2.16, "learning_rate": 1.4055157537940067e-05, "loss": 0.0, "step": 9332 }, { "epoch": 2.16, "learning_rate": 1.4051305754564365e-05, "loss": 0.0, "step": 9333 }, { "epoch": 2.16, "learning_rate": 1.4047453971188662e-05, "loss": 0.0, "step": 9334 }, { "epoch": 2.16, "learning_rate": 1.4043602187812957e-05, "loss": 0.0, "step": 9335 }, { "epoch": 2.16, "learning_rate": 1.4039750404437255e-05, "loss": 0.0, "step": 9336 }, { "epoch": 2.16, "learning_rate": 1.4035898621061552e-05, "loss": 0.0, "step": 9337 }, { "epoch": 2.16, "learning_rate": 1.403204683768585e-05, "loss": 0.0, "step": 9338 }, { "epoch": 2.16, "learning_rate": 1.4028195054310147e-05, "loss": 0.0, "step": 9339 }, { "epoch": 2.16, "learning_rate": 1.4024343270934442e-05, "loss": 0.0, "step": 9340 }, { "epoch": 2.16, "learning_rate": 1.402049148755874e-05, "loss": 0.0, "step": 9341 }, { "epoch": 2.16, "learning_rate": 1.4016639704183039e-05, "loss": 0.0, "step": 9342 }, { "epoch": 2.16, "learning_rate": 1.4012787920807335e-05, "loss": 0.0, "step": 9343 }, { "epoch": 2.16, "learning_rate": 1.400893613743163e-05, "loss": 0.0, "step": 9344 }, { "epoch": 2.16, "learning_rate": 1.4005084354055928e-05, "loss": 0.0, "step": 9345 }, { "epoch": 2.16, "learning_rate": 1.4001232570680225e-05, "loss": 0.0, "step": 9346 }, { "epoch": 2.16, "learning_rate": 1.3997380787304523e-05, "loss": 0.0, "step": 9347 }, { "epoch": 2.16, "learning_rate": 1.399352900392882e-05, "loss": 0.0, "step": 9348 }, { "epoch": 2.16, "learning_rate": 1.3989677220553115e-05, "loss": 0.0, "step": 9349 }, { "epoch": 2.16, "learning_rate": 1.3985825437177413e-05, "loss": 0.0, "step": 9350 }, { "epoch": 2.16, "learning_rate": 1.3981973653801712e-05, "loss": 0.0, "step": 9351 }, { "epoch": 2.16, "learning_rate": 1.3978121870426008e-05, "loss": 0.0005, "step": 9352 }, { "epoch": 2.16, "learning_rate": 1.3974270087050307e-05, "loss": 0.0, "step": 9353 }, { "epoch": 2.16, "learning_rate": 1.3970418303674602e-05, "loss": 0.0, "step": 9354 }, { "epoch": 2.16, "learning_rate": 1.3966566520298898e-05, "loss": 0.0, "step": 9355 }, { "epoch": 2.16, "learning_rate": 1.3962714736923197e-05, "loss": 0.0, "step": 9356 }, { "epoch": 2.16, "learning_rate": 1.3958862953547493e-05, "loss": 0.0, "step": 9357 }, { "epoch": 2.16, "learning_rate": 1.3955011170171788e-05, "loss": 0.0, "step": 9358 }, { "epoch": 2.16, "learning_rate": 1.3951159386796087e-05, "loss": 0.0, "step": 9359 }, { "epoch": 2.16, "learning_rate": 1.3947307603420385e-05, "loss": 0.0, "step": 9360 }, { "epoch": 2.16, "learning_rate": 1.3943455820044682e-05, "loss": 0.0, "step": 9361 }, { "epoch": 2.16, "learning_rate": 1.393960403666898e-05, "loss": 0.0, "step": 9362 }, { "epoch": 2.16, "learning_rate": 1.3935752253293275e-05, "loss": 0.0, "step": 9363 }, { "epoch": 2.16, "learning_rate": 1.3931900469917572e-05, "loss": 0.0, "step": 9364 }, { "epoch": 2.16, "learning_rate": 1.392804868654187e-05, "loss": 0.0001, "step": 9365 }, { "epoch": 2.16, "learning_rate": 1.3924196903166167e-05, "loss": 0.0, "step": 9366 }, { "epoch": 2.16, "learning_rate": 1.3920345119790465e-05, "loss": 0.0, "step": 9367 }, { "epoch": 2.17, "learning_rate": 1.391649333641476e-05, "loss": 0.0, "step": 9368 }, { "epoch": 2.17, "learning_rate": 1.3912641553039058e-05, "loss": 0.0, "step": 9369 }, { "epoch": 2.17, "learning_rate": 1.3908789769663355e-05, "loss": 0.0, "step": 9370 }, { "epoch": 2.17, "learning_rate": 1.3904937986287653e-05, "loss": 0.0, "step": 9371 }, { "epoch": 2.17, "learning_rate": 1.3901086202911948e-05, "loss": 0.0, "step": 9372 }, { "epoch": 2.17, "learning_rate": 1.3897234419536245e-05, "loss": 0.0, "step": 9373 }, { "epoch": 2.17, "learning_rate": 1.3893382636160543e-05, "loss": 0.0, "step": 9374 }, { "epoch": 2.17, "learning_rate": 1.388953085278484e-05, "loss": 0.0, "step": 9375 }, { "epoch": 2.17, "learning_rate": 1.3885679069409138e-05, "loss": 0.0, "step": 9376 }, { "epoch": 2.17, "learning_rate": 1.3881827286033433e-05, "loss": 0.0001, "step": 9377 }, { "epoch": 2.17, "learning_rate": 1.3877975502657731e-05, "loss": 0.0, "step": 9378 }, { "epoch": 2.17, "learning_rate": 1.3874123719282028e-05, "loss": 0.0, "step": 9379 }, { "epoch": 2.17, "learning_rate": 1.3870271935906326e-05, "loss": 0.0, "step": 9380 }, { "epoch": 2.17, "learning_rate": 1.3866420152530623e-05, "loss": 0.0, "step": 9381 }, { "epoch": 2.17, "learning_rate": 1.3862568369154918e-05, "loss": 0.0, "step": 9382 }, { "epoch": 2.17, "learning_rate": 1.3858716585779216e-05, "loss": 0.0, "step": 9383 }, { "epoch": 2.17, "learning_rate": 1.3854864802403513e-05, "loss": 0.0009, "step": 9384 }, { "epoch": 2.17, "learning_rate": 1.3851013019027811e-05, "loss": 0.0, "step": 9385 }, { "epoch": 2.17, "learning_rate": 1.3847161235652106e-05, "loss": 0.0, "step": 9386 }, { "epoch": 2.17, "learning_rate": 1.3843309452276405e-05, "loss": 0.0, "step": 9387 }, { "epoch": 2.17, "learning_rate": 1.3839457668900701e-05, "loss": 0.0, "step": 9388 }, { "epoch": 2.17, "learning_rate": 1.3835605885525e-05, "loss": 0.0, "step": 9389 }, { "epoch": 2.17, "learning_rate": 1.3831754102149296e-05, "loss": 0.0, "step": 9390 }, { "epoch": 2.17, "learning_rate": 1.3827902318773591e-05, "loss": 0.0, "step": 9391 }, { "epoch": 2.17, "learning_rate": 1.382405053539789e-05, "loss": 0.0, "step": 9392 }, { "epoch": 2.17, "learning_rate": 1.3820198752022186e-05, "loss": 0.0, "step": 9393 }, { "epoch": 2.17, "learning_rate": 1.3816346968646485e-05, "loss": 0.0, "step": 9394 }, { "epoch": 2.17, "learning_rate": 1.3812495185270783e-05, "loss": 0.0, "step": 9395 }, { "epoch": 2.17, "learning_rate": 1.3808643401895078e-05, "loss": 0.0, "step": 9396 }, { "epoch": 2.17, "learning_rate": 1.3804791618519375e-05, "loss": 0.0, "step": 9397 }, { "epoch": 2.17, "learning_rate": 1.3800939835143673e-05, "loss": 0.0, "step": 9398 }, { "epoch": 2.17, "learning_rate": 1.379708805176797e-05, "loss": 0.0, "step": 9399 }, { "epoch": 2.17, "learning_rate": 1.3793236268392265e-05, "loss": 0.0, "step": 9400 }, { "epoch": 2.17, "learning_rate": 1.3789384485016563e-05, "loss": 0.0, "step": 9401 }, { "epoch": 2.17, "learning_rate": 1.378553270164086e-05, "loss": 0.0074, "step": 9402 }, { "epoch": 2.17, "learning_rate": 1.3781680918265158e-05, "loss": 0.0, "step": 9403 }, { "epoch": 2.17, "learning_rate": 1.3777829134889456e-05, "loss": 0.0, "step": 9404 }, { "epoch": 2.17, "learning_rate": 1.3773977351513751e-05, "loss": 0.0, "step": 9405 }, { "epoch": 2.17, "learning_rate": 1.3770125568138048e-05, "loss": 0.0, "step": 9406 }, { "epoch": 2.17, "learning_rate": 1.3766273784762346e-05, "loss": 0.0, "step": 9407 }, { "epoch": 2.17, "learning_rate": 1.3762422001386643e-05, "loss": 0.0, "step": 9408 }, { "epoch": 2.17, "learning_rate": 1.3758570218010941e-05, "loss": 0.0, "step": 9409 }, { "epoch": 2.17, "learning_rate": 1.3754718434635236e-05, "loss": 0.0, "step": 9410 }, { "epoch": 2.17, "learning_rate": 1.3750866651259533e-05, "loss": 0.0, "step": 9411 }, { "epoch": 2.18, "learning_rate": 1.3747014867883831e-05, "loss": 0.0, "step": 9412 }, { "epoch": 2.18, "learning_rate": 1.374316308450813e-05, "loss": 0.0001, "step": 9413 }, { "epoch": 2.18, "learning_rate": 1.3739311301132424e-05, "loss": 0.0, "step": 9414 }, { "epoch": 2.18, "learning_rate": 1.3735459517756721e-05, "loss": 0.0, "step": 9415 }, { "epoch": 2.18, "learning_rate": 1.373160773438102e-05, "loss": 0.0, "step": 9416 }, { "epoch": 2.18, "learning_rate": 1.3727755951005316e-05, "loss": 0.0, "step": 9417 }, { "epoch": 2.18, "learning_rate": 1.3723904167629614e-05, "loss": 0.0, "step": 9418 }, { "epoch": 2.18, "learning_rate": 1.372005238425391e-05, "loss": 0.0, "step": 9419 }, { "epoch": 2.18, "learning_rate": 1.3716200600878206e-05, "loss": 0.0, "step": 9420 }, { "epoch": 2.18, "learning_rate": 1.3712348817502504e-05, "loss": 0.0, "step": 9421 }, { "epoch": 2.18, "learning_rate": 1.3708497034126803e-05, "loss": 0.0, "step": 9422 }, { "epoch": 2.18, "learning_rate": 1.37046452507511e-05, "loss": 0.0, "step": 9423 }, { "epoch": 2.18, "learning_rate": 1.3700793467375394e-05, "loss": 0.0, "step": 9424 }, { "epoch": 2.18, "learning_rate": 1.3696941683999693e-05, "loss": 0.0, "step": 9425 }, { "epoch": 2.18, "learning_rate": 1.369308990062399e-05, "loss": 0.0, "step": 9426 }, { "epoch": 2.18, "learning_rate": 1.3689238117248288e-05, "loss": 0.0, "step": 9427 }, { "epoch": 2.18, "learning_rate": 1.3685386333872583e-05, "loss": 0.0, "step": 9428 }, { "epoch": 2.18, "learning_rate": 1.368153455049688e-05, "loss": 0.0, "step": 9429 }, { "epoch": 2.18, "learning_rate": 1.3677682767121178e-05, "loss": 0.0, "step": 9430 }, { "epoch": 2.18, "learning_rate": 1.3673830983745476e-05, "loss": 0.0, "step": 9431 }, { "epoch": 2.18, "learning_rate": 1.3669979200369773e-05, "loss": 0.0004, "step": 9432 }, { "epoch": 2.18, "learning_rate": 1.3666127416994068e-05, "loss": 0.0, "step": 9433 }, { "epoch": 2.18, "learning_rate": 1.3662275633618366e-05, "loss": 0.0, "step": 9434 }, { "epoch": 2.18, "learning_rate": 1.3658423850242663e-05, "loss": 0.0, "step": 9435 }, { "epoch": 2.18, "learning_rate": 1.3654572066866961e-05, "loss": 0.0, "step": 9436 }, { "epoch": 2.18, "learning_rate": 1.365072028349126e-05, "loss": 0.0, "step": 9437 }, { "epoch": 2.18, "learning_rate": 1.3646868500115552e-05, "loss": 0.0, "step": 9438 }, { "epoch": 2.18, "learning_rate": 1.364301671673985e-05, "loss": 0.0, "step": 9439 }, { "epoch": 2.18, "learning_rate": 1.363916493336415e-05, "loss": 0.0, "step": 9440 }, { "epoch": 2.18, "learning_rate": 1.3635313149988446e-05, "loss": 0.0, "step": 9441 }, { "epoch": 2.18, "learning_rate": 1.363146136661274e-05, "loss": 0.0001, "step": 9442 }, { "epoch": 2.18, "learning_rate": 1.3627609583237039e-05, "loss": 0.0, "step": 9443 }, { "epoch": 2.18, "learning_rate": 1.3623757799861336e-05, "loss": 0.0, "step": 9444 }, { "epoch": 2.18, "learning_rate": 1.3619906016485634e-05, "loss": 0.0, "step": 9445 }, { "epoch": 2.18, "learning_rate": 1.3616054233109932e-05, "loss": 0.0, "step": 9446 }, { "epoch": 2.18, "learning_rate": 1.3612202449734226e-05, "loss": 0.0, "step": 9447 }, { "epoch": 2.18, "learning_rate": 1.3608350666358524e-05, "loss": 0.0, "step": 9448 }, { "epoch": 2.18, "learning_rate": 1.3604498882982822e-05, "loss": 0.0, "step": 9449 }, { "epoch": 2.18, "learning_rate": 1.3600647099607119e-05, "loss": 0.0, "step": 9450 }, { "epoch": 2.18, "learning_rate": 1.3596795316231417e-05, "loss": 0.0, "step": 9451 }, { "epoch": 2.18, "learning_rate": 1.3592943532855712e-05, "loss": 0.0002, "step": 9452 }, { "epoch": 2.18, "learning_rate": 1.3589091749480009e-05, "loss": 0.0, "step": 9453 }, { "epoch": 2.18, "learning_rate": 1.3585239966104307e-05, "loss": 0.0, "step": 9454 }, { "epoch": 2.19, "learning_rate": 1.3581388182728606e-05, "loss": 0.0, "step": 9455 }, { "epoch": 2.19, "learning_rate": 1.3577536399352899e-05, "loss": 0.0, "step": 9456 }, { "epoch": 2.19, "learning_rate": 1.3573684615977197e-05, "loss": 0.0, "step": 9457 }, { "epoch": 2.19, "learning_rate": 1.3569832832601496e-05, "loss": 0.0, "step": 9458 }, { "epoch": 2.19, "learning_rate": 1.3565981049225792e-05, "loss": 0.0, "step": 9459 }, { "epoch": 2.19, "learning_rate": 1.356212926585009e-05, "loss": 0.0, "step": 9460 }, { "epoch": 2.19, "learning_rate": 1.3558277482474386e-05, "loss": 0.0, "step": 9461 }, { "epoch": 2.19, "learning_rate": 1.3554425699098682e-05, "loss": 0.0, "step": 9462 }, { "epoch": 2.19, "learning_rate": 1.355057391572298e-05, "loss": 0.0, "step": 9463 }, { "epoch": 2.19, "learning_rate": 1.3546722132347279e-05, "loss": 0.0, "step": 9464 }, { "epoch": 2.19, "learning_rate": 1.3542870348971576e-05, "loss": 0.0, "step": 9465 }, { "epoch": 2.19, "learning_rate": 1.353901856559587e-05, "loss": 0.0, "step": 9466 }, { "epoch": 2.19, "learning_rate": 1.3535166782220169e-05, "loss": 0.0001, "step": 9467 }, { "epoch": 2.19, "learning_rate": 1.3531314998844466e-05, "loss": 0.0, "step": 9468 }, { "epoch": 2.19, "learning_rate": 1.3527463215468764e-05, "loss": 0.0001, "step": 9469 }, { "epoch": 2.19, "learning_rate": 1.3523611432093059e-05, "loss": 0.0, "step": 9470 }, { "epoch": 2.19, "learning_rate": 1.3519759648717355e-05, "loss": 0.0, "step": 9471 }, { "epoch": 2.19, "learning_rate": 1.3515907865341654e-05, "loss": 0.0, "step": 9472 }, { "epoch": 2.19, "learning_rate": 1.3512056081965952e-05, "loss": 0.0, "step": 9473 }, { "epoch": 2.19, "learning_rate": 1.3508204298590249e-05, "loss": 0.0, "step": 9474 }, { "epoch": 2.19, "learning_rate": 1.3504352515214544e-05, "loss": 0.0, "step": 9475 }, { "epoch": 2.19, "learning_rate": 1.3500500731838842e-05, "loss": 0.0, "step": 9476 }, { "epoch": 2.19, "learning_rate": 1.3496648948463139e-05, "loss": 0.0, "step": 9477 }, { "epoch": 2.19, "learning_rate": 1.3492797165087437e-05, "loss": 0.0, "step": 9478 }, { "epoch": 2.19, "learning_rate": 1.3488945381711734e-05, "loss": 0.0, "step": 9479 }, { "epoch": 2.19, "learning_rate": 1.3485093598336029e-05, "loss": 0.0, "step": 9480 }, { "epoch": 2.19, "learning_rate": 1.3481241814960327e-05, "loss": 0.0, "step": 9481 }, { "epoch": 2.19, "learning_rate": 1.3477390031584625e-05, "loss": 0.0, "step": 9482 }, { "epoch": 2.19, "learning_rate": 1.3473538248208922e-05, "loss": 0.0, "step": 9483 }, { "epoch": 2.19, "learning_rate": 1.3469686464833217e-05, "loss": 0.0, "step": 9484 }, { "epoch": 2.19, "learning_rate": 1.3465834681457515e-05, "loss": 0.0, "step": 9485 }, { "epoch": 2.19, "learning_rate": 1.3461982898081812e-05, "loss": 0.0, "step": 9486 }, { "epoch": 2.19, "learning_rate": 1.345813111470611e-05, "loss": 0.0, "step": 9487 }, { "epoch": 2.19, "learning_rate": 1.3454279331330407e-05, "loss": 0.0, "step": 9488 }, { "epoch": 2.19, "learning_rate": 1.3450427547954702e-05, "loss": 0.0, "step": 9489 }, { "epoch": 2.19, "learning_rate": 1.3446575764579e-05, "loss": 0.0, "step": 9490 }, { "epoch": 2.19, "learning_rate": 1.3442723981203299e-05, "loss": 0.0, "step": 9491 }, { "epoch": 2.19, "learning_rate": 1.3438872197827595e-05, "loss": 0.0, "step": 9492 }, { "epoch": 2.19, "learning_rate": 1.3435020414451894e-05, "loss": 0.0, "step": 9493 }, { "epoch": 2.19, "learning_rate": 1.3431168631076189e-05, "loss": 0.0, "step": 9494 }, { "epoch": 2.19, "learning_rate": 1.3427316847700485e-05, "loss": 0.0, "step": 9495 }, { "epoch": 2.19, "learning_rate": 1.3423465064324784e-05, "loss": 0.0, "step": 9496 }, { "epoch": 2.19, "learning_rate": 1.341961328094908e-05, "loss": 0.0, "step": 9497 }, { "epoch": 2.2, "learning_rate": 1.3415761497573375e-05, "loss": 0.0, "step": 9498 }, { "epoch": 2.2, "learning_rate": 1.3411909714197674e-05, "loss": 0.0, "step": 9499 }, { "epoch": 2.2, "learning_rate": 1.3408057930821972e-05, "loss": 0.0, "step": 9500 }, { "epoch": 2.2, "learning_rate": 1.3404206147446269e-05, "loss": 0.0, "step": 9501 }, { "epoch": 2.2, "learning_rate": 1.3400354364070567e-05, "loss": 0.0, "step": 9502 }, { "epoch": 2.2, "learning_rate": 1.3396502580694862e-05, "loss": 0.0, "step": 9503 }, { "epoch": 2.2, "learning_rate": 1.3392650797319158e-05, "loss": 0.0002, "step": 9504 }, { "epoch": 2.2, "learning_rate": 1.3388799013943457e-05, "loss": 0.0, "step": 9505 }, { "epoch": 2.2, "learning_rate": 1.3384947230567753e-05, "loss": 0.0, "step": 9506 }, { "epoch": 2.2, "learning_rate": 1.3381095447192052e-05, "loss": 0.0003, "step": 9507 }, { "epoch": 2.2, "learning_rate": 1.3377243663816347e-05, "loss": 0.0, "step": 9508 }, { "epoch": 2.2, "learning_rate": 1.3373391880440645e-05, "loss": 0.0, "step": 9509 }, { "epoch": 2.2, "learning_rate": 1.3369540097064942e-05, "loss": 0.0, "step": 9510 }, { "epoch": 2.2, "learning_rate": 1.336568831368924e-05, "loss": 0.0, "step": 9511 }, { "epoch": 2.2, "learning_rate": 1.3361836530313535e-05, "loss": 0.0001, "step": 9512 }, { "epoch": 2.2, "learning_rate": 1.3357984746937832e-05, "loss": 0.0, "step": 9513 }, { "epoch": 2.2, "learning_rate": 1.335413296356213e-05, "loss": 0.0, "step": 9514 }, { "epoch": 2.2, "learning_rate": 1.3350281180186427e-05, "loss": 0.0, "step": 9515 }, { "epoch": 2.2, "learning_rate": 1.3346429396810725e-05, "loss": 0.0, "step": 9516 }, { "epoch": 2.2, "learning_rate": 1.334257761343502e-05, "loss": 0.0, "step": 9517 }, { "epoch": 2.2, "learning_rate": 1.3338725830059318e-05, "loss": 0.0, "step": 9518 }, { "epoch": 2.2, "learning_rate": 1.3334874046683615e-05, "loss": 0.0, "step": 9519 }, { "epoch": 2.2, "learning_rate": 1.3331022263307913e-05, "loss": 0.0, "step": 9520 }, { "epoch": 2.2, "learning_rate": 1.332717047993221e-05, "loss": 0.0, "step": 9521 }, { "epoch": 2.2, "learning_rate": 1.3323318696556505e-05, "loss": 0.0, "step": 9522 }, { "epoch": 2.2, "learning_rate": 1.3319466913180803e-05, "loss": 0.0, "step": 9523 }, { "epoch": 2.2, "learning_rate": 1.33156151298051e-05, "loss": 0.0, "step": 9524 }, { "epoch": 2.2, "learning_rate": 1.3311763346429398e-05, "loss": 0.0, "step": 9525 }, { "epoch": 2.2, "learning_rate": 1.3307911563053693e-05, "loss": 0.0, "step": 9526 }, { "epoch": 2.2, "learning_rate": 1.3304059779677992e-05, "loss": 0.0, "step": 9527 }, { "epoch": 2.2, "learning_rate": 1.3300207996302288e-05, "loss": 0.0, "step": 9528 }, { "epoch": 2.2, "learning_rate": 1.3296356212926587e-05, "loss": 0.0001, "step": 9529 }, { "epoch": 2.2, "learning_rate": 1.3292504429550883e-05, "loss": 0.0, "step": 9530 }, { "epoch": 2.2, "learning_rate": 1.3288652646175178e-05, "loss": 0.0, "step": 9531 }, { "epoch": 2.2, "learning_rate": 1.3284800862799477e-05, "loss": 0.0001, "step": 9532 }, { "epoch": 2.2, "learning_rate": 1.3280949079423773e-05, "loss": 0.0, "step": 9533 }, { "epoch": 2.2, "learning_rate": 1.3277097296048071e-05, "loss": 0.0, "step": 9534 }, { "epoch": 2.2, "learning_rate": 1.327324551267237e-05, "loss": 0.0, "step": 9535 }, { "epoch": 2.2, "learning_rate": 1.3269393729296665e-05, "loss": 0.0, "step": 9536 }, { "epoch": 2.2, "learning_rate": 1.3265541945920961e-05, "loss": 0.0, "step": 9537 }, { "epoch": 2.2, "learning_rate": 1.326169016254526e-05, "loss": 0.0, "step": 9538 }, { "epoch": 2.2, "learning_rate": 1.3257838379169556e-05, "loss": 0.0, "step": 9539 }, { "epoch": 2.2, "learning_rate": 1.3253986595793851e-05, "loss": 0.0, "step": 9540 }, { "epoch": 2.2, "learning_rate": 1.325013481241815e-05, "loss": 0.0, "step": 9541 }, { "epoch": 2.21, "learning_rate": 1.3246283029042446e-05, "loss": 0.0001, "step": 9542 }, { "epoch": 2.21, "learning_rate": 1.3242431245666745e-05, "loss": 0.0, "step": 9543 }, { "epoch": 2.21, "learning_rate": 1.3238579462291043e-05, "loss": 0.0, "step": 9544 }, { "epoch": 2.21, "learning_rate": 1.3234727678915338e-05, "loss": 0.0, "step": 9545 }, { "epoch": 2.21, "learning_rate": 1.3230875895539635e-05, "loss": 0.0, "step": 9546 }, { "epoch": 2.21, "learning_rate": 1.3227024112163933e-05, "loss": 0.0, "step": 9547 }, { "epoch": 2.21, "learning_rate": 1.322317232878823e-05, "loss": 0.0, "step": 9548 }, { "epoch": 2.21, "learning_rate": 1.3219320545412528e-05, "loss": 0.0, "step": 9549 }, { "epoch": 2.21, "learning_rate": 1.3215468762036823e-05, "loss": 0.0, "step": 9550 }, { "epoch": 2.21, "learning_rate": 1.3211616978661121e-05, "loss": 0.0, "step": 9551 }, { "epoch": 2.21, "learning_rate": 1.3207765195285418e-05, "loss": 0.0, "step": 9552 }, { "epoch": 2.21, "learning_rate": 1.3203913411909716e-05, "loss": 0.0, "step": 9553 }, { "epoch": 2.21, "learning_rate": 1.3200061628534011e-05, "loss": 0.0, "step": 9554 }, { "epoch": 2.21, "learning_rate": 1.3196209845158308e-05, "loss": 0.0, "step": 9555 }, { "epoch": 2.21, "learning_rate": 1.3192358061782606e-05, "loss": 0.0, "step": 9556 }, { "epoch": 2.21, "learning_rate": 1.3188506278406903e-05, "loss": 0.0, "step": 9557 }, { "epoch": 2.21, "learning_rate": 1.3184654495031201e-05, "loss": 0.0, "step": 9558 }, { "epoch": 2.21, "learning_rate": 1.3180802711655496e-05, "loss": 0.0, "step": 9559 }, { "epoch": 2.21, "learning_rate": 1.3176950928279795e-05, "loss": 0.0, "step": 9560 }, { "epoch": 2.21, "learning_rate": 1.3173099144904091e-05, "loss": 0.0, "step": 9561 }, { "epoch": 2.21, "learning_rate": 1.316924736152839e-05, "loss": 0.0, "step": 9562 }, { "epoch": 2.21, "learning_rate": 1.3165395578152686e-05, "loss": 0.0, "step": 9563 }, { "epoch": 2.21, "learning_rate": 1.3161543794776981e-05, "loss": 0.0, "step": 9564 }, { "epoch": 2.21, "learning_rate": 1.315769201140128e-05, "loss": 0.0, "step": 9565 }, { "epoch": 2.21, "learning_rate": 1.3153840228025576e-05, "loss": 0.0, "step": 9566 }, { "epoch": 2.21, "learning_rate": 1.3149988444649874e-05, "loss": 0.0, "step": 9567 }, { "epoch": 2.21, "learning_rate": 1.314613666127417e-05, "loss": 0.0, "step": 9568 }, { "epoch": 2.21, "learning_rate": 1.3142284877898468e-05, "loss": 0.0, "step": 9569 }, { "epoch": 2.21, "learning_rate": 1.3138433094522764e-05, "loss": 0.0, "step": 9570 }, { "epoch": 2.21, "learning_rate": 1.3134581311147063e-05, "loss": 0.0, "step": 9571 }, { "epoch": 2.21, "learning_rate": 1.313072952777136e-05, "loss": 0.0, "step": 9572 }, { "epoch": 2.21, "learning_rate": 1.3126877744395654e-05, "loss": 0.0, "step": 9573 }, { "epoch": 2.21, "learning_rate": 1.3123025961019953e-05, "loss": 0.0, "step": 9574 }, { "epoch": 2.21, "learning_rate": 1.311917417764425e-05, "loss": 0.0, "step": 9575 }, { "epoch": 2.21, "learning_rate": 1.3115322394268548e-05, "loss": 0.0, "step": 9576 }, { "epoch": 2.21, "learning_rate": 1.3111470610892846e-05, "loss": 0.0, "step": 9577 }, { "epoch": 2.21, "learning_rate": 1.3107618827517141e-05, "loss": 0.0, "step": 9578 }, { "epoch": 2.21, "learning_rate": 1.3103767044141438e-05, "loss": 0.0, "step": 9579 }, { "epoch": 2.21, "learning_rate": 1.3099915260765736e-05, "loss": 0.0, "step": 9580 }, { "epoch": 2.21, "learning_rate": 1.3096063477390033e-05, "loss": 0.0, "step": 9581 }, { "epoch": 2.21, "learning_rate": 1.3092211694014328e-05, "loss": 0.0, "step": 9582 }, { "epoch": 2.21, "learning_rate": 1.3088359910638626e-05, "loss": 0.0, "step": 9583 }, { "epoch": 2.21, "learning_rate": 1.3084508127262923e-05, "loss": 0.0, "step": 9584 }, { "epoch": 2.22, "learning_rate": 1.3080656343887221e-05, "loss": 0.0, "step": 9585 }, { "epoch": 2.22, "learning_rate": 1.307680456051152e-05, "loss": 0.0, "step": 9586 }, { "epoch": 2.22, "learning_rate": 1.3072952777135814e-05, "loss": 0.0, "step": 9587 }, { "epoch": 2.22, "learning_rate": 1.3069100993760111e-05, "loss": 0.0, "step": 9588 }, { "epoch": 2.22, "learning_rate": 1.306524921038441e-05, "loss": 0.0, "step": 9589 }, { "epoch": 2.22, "learning_rate": 1.3061397427008706e-05, "loss": 0.0, "step": 9590 }, { "epoch": 2.22, "learning_rate": 1.3057545643633004e-05, "loss": 0.0, "step": 9591 }, { "epoch": 2.22, "learning_rate": 1.30536938602573e-05, "loss": 0.0, "step": 9592 }, { "epoch": 2.22, "learning_rate": 1.3049842076881596e-05, "loss": 0.0, "step": 9593 }, { "epoch": 2.22, "learning_rate": 1.3045990293505894e-05, "loss": 0.0, "step": 9594 }, { "epoch": 2.22, "learning_rate": 1.3042138510130193e-05, "loss": 0.0002, "step": 9595 }, { "epoch": 2.22, "learning_rate": 1.3038286726754487e-05, "loss": 0.0, "step": 9596 }, { "epoch": 2.22, "learning_rate": 1.3034434943378784e-05, "loss": 0.0, "step": 9597 }, { "epoch": 2.22, "learning_rate": 1.3030583160003082e-05, "loss": 0.0, "step": 9598 }, { "epoch": 2.22, "learning_rate": 1.3026731376627379e-05, "loss": 0.0, "step": 9599 }, { "epoch": 2.22, "learning_rate": 1.3022879593251677e-05, "loss": 0.0, "step": 9600 }, { "epoch": 2.22, "learning_rate": 1.3019027809875972e-05, "loss": 0.0001, "step": 9601 }, { "epoch": 2.22, "learning_rate": 1.3015176026500269e-05, "loss": 0.0, "step": 9602 }, { "epoch": 2.22, "learning_rate": 1.3011324243124567e-05, "loss": 0.0, "step": 9603 }, { "epoch": 2.22, "learning_rate": 1.3007472459748866e-05, "loss": 0.0, "step": 9604 }, { "epoch": 2.22, "learning_rate": 1.3003620676373162e-05, "loss": 0.0, "step": 9605 }, { "epoch": 2.22, "learning_rate": 1.2999768892997457e-05, "loss": 0.0, "step": 9606 }, { "epoch": 2.22, "learning_rate": 1.2995917109621756e-05, "loss": 0.0, "step": 9607 }, { "epoch": 2.22, "learning_rate": 1.2992065326246052e-05, "loss": 0.0, "step": 9608 }, { "epoch": 2.22, "learning_rate": 1.298821354287035e-05, "loss": 0.0, "step": 9609 }, { "epoch": 2.22, "learning_rate": 1.2984361759494646e-05, "loss": 0.0, "step": 9610 }, { "epoch": 2.22, "learning_rate": 1.2980509976118942e-05, "loss": 0.0, "step": 9611 }, { "epoch": 2.22, "learning_rate": 1.297665819274324e-05, "loss": 0.0, "step": 9612 }, { "epoch": 2.22, "learning_rate": 1.2972806409367539e-05, "loss": 0.0, "step": 9613 }, { "epoch": 2.22, "learning_rate": 1.2968954625991836e-05, "loss": 0.0, "step": 9614 }, { "epoch": 2.22, "learning_rate": 1.296510284261613e-05, "loss": 0.0, "step": 9615 }, { "epoch": 2.22, "learning_rate": 1.2961251059240429e-05, "loss": 0.0, "step": 9616 }, { "epoch": 2.22, "learning_rate": 1.2957399275864726e-05, "loss": 0.0, "step": 9617 }, { "epoch": 2.22, "learning_rate": 1.2953547492489024e-05, "loss": 0.0, "step": 9618 }, { "epoch": 2.22, "learning_rate": 1.294969570911332e-05, "loss": 0.0, "step": 9619 }, { "epoch": 2.22, "learning_rate": 1.2945843925737616e-05, "loss": 0.0, "step": 9620 }, { "epoch": 2.22, "learning_rate": 1.2941992142361914e-05, "loss": 0.0, "step": 9621 }, { "epoch": 2.22, "learning_rate": 1.2938140358986212e-05, "loss": 0.0, "step": 9622 }, { "epoch": 2.22, "learning_rate": 1.2934288575610509e-05, "loss": 0.0, "step": 9623 }, { "epoch": 2.22, "learning_rate": 1.2930436792234804e-05, "loss": 0.0, "step": 9624 }, { "epoch": 2.22, "learning_rate": 1.2926585008859102e-05, "loss": 0.0, "step": 9625 }, { "epoch": 2.22, "learning_rate": 1.2922733225483399e-05, "loss": 0.0, "step": 9626 }, { "epoch": 2.22, "learning_rate": 1.2918881442107697e-05, "loss": 0.0, "step": 9627 }, { "epoch": 2.23, "learning_rate": 1.2915029658731994e-05, "loss": 0.0, "step": 9628 }, { "epoch": 2.23, "learning_rate": 1.2911177875356289e-05, "loss": 0.0, "step": 9629 }, { "epoch": 2.23, "learning_rate": 1.2907326091980587e-05, "loss": 0.0, "step": 9630 }, { "epoch": 2.23, "learning_rate": 1.2903474308604885e-05, "loss": 0.0, "step": 9631 }, { "epoch": 2.23, "learning_rate": 1.2899622525229182e-05, "loss": 0.0, "step": 9632 }, { "epoch": 2.23, "learning_rate": 1.289577074185348e-05, "loss": 0.0, "step": 9633 }, { "epoch": 2.23, "learning_rate": 1.2891918958477775e-05, "loss": 0.0, "step": 9634 }, { "epoch": 2.23, "learning_rate": 1.2888067175102072e-05, "loss": 0.0, "step": 9635 }, { "epoch": 2.23, "learning_rate": 1.288421539172637e-05, "loss": 0.0, "step": 9636 }, { "epoch": 2.23, "learning_rate": 1.2880363608350669e-05, "loss": 0.0, "step": 9637 }, { "epoch": 2.23, "learning_rate": 1.2876511824974962e-05, "loss": 0.0, "step": 9638 }, { "epoch": 2.23, "learning_rate": 1.287266004159926e-05, "loss": 0.0, "step": 9639 }, { "epoch": 2.23, "learning_rate": 1.2868808258223559e-05, "loss": 0.0011, "step": 9640 }, { "epoch": 2.23, "learning_rate": 1.2864956474847855e-05, "loss": 0.0, "step": 9641 }, { "epoch": 2.23, "learning_rate": 1.2861104691472154e-05, "loss": 0.0, "step": 9642 }, { "epoch": 2.23, "learning_rate": 1.2857252908096449e-05, "loss": 0.0, "step": 9643 }, { "epoch": 2.23, "learning_rate": 1.2853401124720745e-05, "loss": 0.0, "step": 9644 }, { "epoch": 2.23, "learning_rate": 1.2849549341345044e-05, "loss": 0.0, "step": 9645 }, { "epoch": 2.23, "learning_rate": 1.2845697557969342e-05, "loss": 0.0, "step": 9646 }, { "epoch": 2.23, "learning_rate": 1.2841845774593639e-05, "loss": 0.0, "step": 9647 }, { "epoch": 2.23, "learning_rate": 1.2837993991217934e-05, "loss": 0.0, "step": 9648 }, { "epoch": 2.23, "learning_rate": 1.2834142207842232e-05, "loss": 0.0, "step": 9649 }, { "epoch": 2.23, "learning_rate": 1.2830290424466529e-05, "loss": 0.0, "step": 9650 }, { "epoch": 2.23, "learning_rate": 1.2826438641090827e-05, "loss": 0.0, "step": 9651 }, { "epoch": 2.23, "learning_rate": 1.2822586857715122e-05, "loss": 0.0, "step": 9652 }, { "epoch": 2.23, "learning_rate": 1.2818735074339419e-05, "loss": 0.0, "step": 9653 }, { "epoch": 2.23, "learning_rate": 1.2814883290963717e-05, "loss": 0.0, "step": 9654 }, { "epoch": 2.23, "learning_rate": 1.2811031507588015e-05, "loss": 0.0005, "step": 9655 }, { "epoch": 2.23, "learning_rate": 1.2807179724212312e-05, "loss": 0.0, "step": 9656 }, { "epoch": 2.23, "learning_rate": 1.2803327940836607e-05, "loss": 0.0, "step": 9657 }, { "epoch": 2.23, "learning_rate": 1.2799476157460905e-05, "loss": 0.0, "step": 9658 }, { "epoch": 2.23, "learning_rate": 1.2795624374085202e-05, "loss": 0.0, "step": 9659 }, { "epoch": 2.23, "learning_rate": 1.27917725907095e-05, "loss": 0.0, "step": 9660 }, { "epoch": 2.23, "learning_rate": 1.2787920807333797e-05, "loss": 0.0, "step": 9661 }, { "epoch": 2.23, "learning_rate": 1.2784069023958092e-05, "loss": 0.0, "step": 9662 }, { "epoch": 2.23, "learning_rate": 1.278021724058239e-05, "loss": 0.0, "step": 9663 }, { "epoch": 2.23, "learning_rate": 1.2776365457206688e-05, "loss": 0.0011, "step": 9664 }, { "epoch": 2.23, "learning_rate": 1.2772513673830985e-05, "loss": 0.0, "step": 9665 }, { "epoch": 2.23, "learning_rate": 1.276866189045528e-05, "loss": 0.0, "step": 9666 }, { "epoch": 2.23, "learning_rate": 1.2764810107079578e-05, "loss": 0.0, "step": 9667 }, { "epoch": 2.23, "learning_rate": 1.2760958323703875e-05, "loss": 0.0001, "step": 9668 }, { "epoch": 2.23, "learning_rate": 1.2757106540328173e-05, "loss": 0.0, "step": 9669 }, { "epoch": 2.23, "learning_rate": 1.275325475695247e-05, "loss": 0.0, "step": 9670 }, { "epoch": 2.24, "learning_rate": 1.2749402973576765e-05, "loss": 0.0, "step": 9671 }, { "epoch": 2.24, "learning_rate": 1.2745551190201063e-05, "loss": 0.0, "step": 9672 }, { "epoch": 2.24, "learning_rate": 1.2741699406825362e-05, "loss": 0.0002, "step": 9673 }, { "epoch": 2.24, "learning_rate": 1.2737847623449658e-05, "loss": 0.0, "step": 9674 }, { "epoch": 2.24, "learning_rate": 1.2733995840073957e-05, "loss": 0.0, "step": 9675 }, { "epoch": 2.24, "learning_rate": 1.2730144056698252e-05, "loss": 0.0, "step": 9676 }, { "epoch": 2.24, "learning_rate": 1.2726292273322548e-05, "loss": 0.0, "step": 9677 }, { "epoch": 2.24, "learning_rate": 1.2722440489946847e-05, "loss": 0.0, "step": 9678 }, { "epoch": 2.24, "learning_rate": 1.2718588706571143e-05, "loss": 0.0057, "step": 9679 }, { "epoch": 2.24, "learning_rate": 1.2714736923195438e-05, "loss": 0.0, "step": 9680 }, { "epoch": 2.24, "learning_rate": 1.2710885139819737e-05, "loss": 0.0, "step": 9681 }, { "epoch": 2.24, "learning_rate": 1.2707033356444035e-05, "loss": 0.0, "step": 9682 }, { "epoch": 2.24, "learning_rate": 1.2703181573068332e-05, "loss": 0.0002, "step": 9683 }, { "epoch": 2.24, "learning_rate": 1.269932978969263e-05, "loss": 0.0, "step": 9684 }, { "epoch": 2.24, "learning_rate": 1.2695478006316925e-05, "loss": 0.0, "step": 9685 }, { "epoch": 2.24, "learning_rate": 1.2691626222941222e-05, "loss": 0.0, "step": 9686 }, { "epoch": 2.24, "learning_rate": 1.268777443956552e-05, "loss": 0.0, "step": 9687 }, { "epoch": 2.24, "learning_rate": 1.2683922656189817e-05, "loss": 0.0, "step": 9688 }, { "epoch": 2.24, "learning_rate": 1.2680070872814115e-05, "loss": 0.0, "step": 9689 }, { "epoch": 2.24, "learning_rate": 1.267621908943841e-05, "loss": 0.0, "step": 9690 }, { "epoch": 2.24, "learning_rate": 1.2672367306062708e-05, "loss": 0.0, "step": 9691 }, { "epoch": 2.24, "learning_rate": 1.2668515522687005e-05, "loss": 0.0, "step": 9692 }, { "epoch": 2.24, "learning_rate": 1.2664663739311303e-05, "loss": 0.0, "step": 9693 }, { "epoch": 2.24, "learning_rate": 1.2660811955935598e-05, "loss": 0.0, "step": 9694 }, { "epoch": 2.24, "learning_rate": 1.2656960172559895e-05, "loss": 0.0, "step": 9695 }, { "epoch": 2.24, "learning_rate": 1.2653108389184193e-05, "loss": 0.0, "step": 9696 }, { "epoch": 2.24, "learning_rate": 1.264925660580849e-05, "loss": 0.0, "step": 9697 }, { "epoch": 2.24, "learning_rate": 1.2645404822432788e-05, "loss": 0.0, "step": 9698 }, { "epoch": 2.24, "learning_rate": 1.2641553039057083e-05, "loss": 0.0, "step": 9699 }, { "epoch": 2.24, "learning_rate": 1.2637701255681381e-05, "loss": 0.0, "step": 9700 }, { "epoch": 2.24, "learning_rate": 1.2633849472305678e-05, "loss": 0.0, "step": 9701 }, { "epoch": 2.24, "learning_rate": 1.2629997688929976e-05, "loss": 0.0008, "step": 9702 }, { "epoch": 2.24, "learning_rate": 1.2626145905554273e-05, "loss": 0.0, "step": 9703 }, { "epoch": 2.24, "learning_rate": 1.2622294122178568e-05, "loss": 0.0, "step": 9704 }, { "epoch": 2.24, "learning_rate": 1.2618442338802866e-05, "loss": 0.0, "step": 9705 }, { "epoch": 2.24, "learning_rate": 1.2614590555427163e-05, "loss": 0.0, "step": 9706 }, { "epoch": 2.24, "learning_rate": 1.2610738772051461e-05, "loss": 0.0, "step": 9707 }, { "epoch": 2.24, "learning_rate": 1.2606886988675756e-05, "loss": 0.0, "step": 9708 }, { "epoch": 2.24, "learning_rate": 1.2603035205300055e-05, "loss": 0.0, "step": 9709 }, { "epoch": 2.24, "learning_rate": 1.2599183421924351e-05, "loss": 0.0, "step": 9710 }, { "epoch": 2.24, "learning_rate": 1.259533163854865e-05, "loss": 0.0, "step": 9711 }, { "epoch": 2.24, "learning_rate": 1.2591479855172946e-05, "loss": 0.0, "step": 9712 }, { "epoch": 2.24, "learning_rate": 1.2587628071797241e-05, "loss": 0.0, "step": 9713 }, { "epoch": 2.24, "learning_rate": 1.258377628842154e-05, "loss": 0.0, "step": 9714 }, { "epoch": 2.25, "learning_rate": 1.2579924505045836e-05, "loss": 0.0, "step": 9715 }, { "epoch": 2.25, "learning_rate": 1.2576072721670135e-05, "loss": 0.0, "step": 9716 }, { "epoch": 2.25, "learning_rate": 1.2572220938294433e-05, "loss": 0.0, "step": 9717 }, { "epoch": 2.25, "learning_rate": 1.2568369154918728e-05, "loss": 0.0, "step": 9718 }, { "epoch": 2.25, "learning_rate": 1.2564517371543025e-05, "loss": 0.0, "step": 9719 }, { "epoch": 2.25, "learning_rate": 1.2560665588167323e-05, "loss": 0.0, "step": 9720 }, { "epoch": 2.25, "learning_rate": 1.255681380479162e-05, "loss": 0.0, "step": 9721 }, { "epoch": 2.25, "learning_rate": 1.2552962021415914e-05, "loss": 0.0, "step": 9722 }, { "epoch": 2.25, "learning_rate": 1.2549110238040213e-05, "loss": 0.0, "step": 9723 }, { "epoch": 2.25, "learning_rate": 1.254525845466451e-05, "loss": 0.0001, "step": 9724 }, { "epoch": 2.25, "learning_rate": 1.2541406671288808e-05, "loss": 0.0, "step": 9725 }, { "epoch": 2.25, "learning_rate": 1.2537554887913106e-05, "loss": 0.0, "step": 9726 }, { "epoch": 2.25, "learning_rate": 1.2533703104537401e-05, "loss": 0.0, "step": 9727 }, { "epoch": 2.25, "learning_rate": 1.2529851321161698e-05, "loss": 0.0, "step": 9728 }, { "epoch": 2.25, "learning_rate": 1.2525999537785996e-05, "loss": 0.0, "step": 9729 }, { "epoch": 2.25, "learning_rate": 1.2522147754410293e-05, "loss": 0.0, "step": 9730 }, { "epoch": 2.25, "learning_rate": 1.2518295971034591e-05, "loss": 0.0, "step": 9731 }, { "epoch": 2.25, "learning_rate": 1.2514444187658886e-05, "loss": 0.0002, "step": 9732 }, { "epoch": 2.25, "learning_rate": 1.2510592404283183e-05, "loss": 0.0, "step": 9733 }, { "epoch": 2.25, "learning_rate": 1.2506740620907481e-05, "loss": 0.0, "step": 9734 }, { "epoch": 2.25, "learning_rate": 1.250288883753178e-05, "loss": 0.0, "step": 9735 }, { "epoch": 2.25, "learning_rate": 1.2499037054156074e-05, "loss": 0.0, "step": 9736 }, { "epoch": 2.25, "learning_rate": 1.2495185270780371e-05, "loss": 0.0002, "step": 9737 }, { "epoch": 2.25, "learning_rate": 1.249133348740467e-05, "loss": 0.0, "step": 9738 }, { "epoch": 2.25, "learning_rate": 1.2487481704028966e-05, "loss": 0.0, "step": 9739 }, { "epoch": 2.25, "learning_rate": 1.2483629920653263e-05, "loss": 0.0, "step": 9740 }, { "epoch": 2.25, "learning_rate": 1.2479778137277561e-05, "loss": 0.0, "step": 9741 }, { "epoch": 2.25, "learning_rate": 1.2475926353901856e-05, "loss": 0.0, "step": 9742 }, { "epoch": 2.25, "learning_rate": 1.2472074570526154e-05, "loss": 0.0, "step": 9743 }, { "epoch": 2.25, "learning_rate": 1.2468222787150451e-05, "loss": 0.0, "step": 9744 }, { "epoch": 2.25, "learning_rate": 1.2464371003774748e-05, "loss": 0.0, "step": 9745 }, { "epoch": 2.25, "learning_rate": 1.2460519220399046e-05, "loss": 0.0, "step": 9746 }, { "epoch": 2.25, "learning_rate": 1.2456667437023343e-05, "loss": 0.0, "step": 9747 }, { "epoch": 2.25, "learning_rate": 1.245281565364764e-05, "loss": 0.0, "step": 9748 }, { "epoch": 2.25, "learning_rate": 1.2448963870271936e-05, "loss": 0.0, "step": 9749 }, { "epoch": 2.25, "learning_rate": 1.2445112086896234e-05, "loss": 0.0, "step": 9750 }, { "epoch": 2.25, "learning_rate": 1.244126030352053e-05, "loss": 0.0, "step": 9751 }, { "epoch": 2.25, "learning_rate": 1.2437408520144828e-05, "loss": 0.0, "step": 9752 }, { "epoch": 2.25, "learning_rate": 1.2433556736769126e-05, "loss": 0.0, "step": 9753 }, { "epoch": 2.25, "learning_rate": 1.242970495339342e-05, "loss": 0.0, "step": 9754 }, { "epoch": 2.25, "learning_rate": 1.2425853170017719e-05, "loss": 0.0, "step": 9755 }, { "epoch": 2.25, "learning_rate": 1.2422001386642016e-05, "loss": 0.0, "step": 9756 }, { "epoch": 2.25, "learning_rate": 1.2418149603266312e-05, "loss": 0.0, "step": 9757 }, { "epoch": 2.26, "learning_rate": 1.2414297819890609e-05, "loss": 0.0, "step": 9758 }, { "epoch": 2.26, "learning_rate": 1.2410446036514907e-05, "loss": 0.0, "step": 9759 }, { "epoch": 2.26, "learning_rate": 1.2406594253139204e-05, "loss": 0.0, "step": 9760 }, { "epoch": 2.26, "learning_rate": 1.24027424697635e-05, "loss": 0.0001, "step": 9761 }, { "epoch": 2.26, "learning_rate": 1.2398890686387799e-05, "loss": 0.0, "step": 9762 }, { "epoch": 2.26, "learning_rate": 1.2395038903012094e-05, "loss": 0.0, "step": 9763 }, { "epoch": 2.26, "learning_rate": 1.2391187119636392e-05, "loss": 0.0, "step": 9764 }, { "epoch": 2.26, "learning_rate": 1.2387335336260689e-05, "loss": 0.0, "step": 9765 }, { "epoch": 2.26, "learning_rate": 1.2383483552884986e-05, "loss": 0.0, "step": 9766 }, { "epoch": 2.26, "learning_rate": 1.2379631769509284e-05, "loss": 0.0, "step": 9767 }, { "epoch": 2.26, "learning_rate": 1.237577998613358e-05, "loss": 0.0, "step": 9768 }, { "epoch": 2.26, "learning_rate": 1.2371928202757877e-05, "loss": 0.0, "step": 9769 }, { "epoch": 2.26, "learning_rate": 1.2368076419382174e-05, "loss": 0.0, "step": 9770 }, { "epoch": 2.26, "learning_rate": 1.2364224636006472e-05, "loss": 0.0, "step": 9771 }, { "epoch": 2.26, "learning_rate": 1.2360372852630767e-05, "loss": 0.0, "step": 9772 }, { "epoch": 2.26, "learning_rate": 1.2356521069255066e-05, "loss": 0.0, "step": 9773 }, { "epoch": 2.26, "learning_rate": 1.2352669285879364e-05, "loss": 0.0, "step": 9774 }, { "epoch": 2.26, "learning_rate": 1.2348817502503659e-05, "loss": 0.0, "step": 9775 }, { "epoch": 2.26, "learning_rate": 1.2344965719127957e-05, "loss": 0.0, "step": 9776 }, { "epoch": 2.26, "learning_rate": 1.2341113935752254e-05, "loss": 0.0, "step": 9777 }, { "epoch": 2.26, "learning_rate": 1.233726215237655e-05, "loss": 0.0, "step": 9778 }, { "epoch": 2.26, "learning_rate": 1.2333410369000847e-05, "loss": 0.0, "step": 9779 }, { "epoch": 2.26, "learning_rate": 1.2329558585625146e-05, "loss": 0.0, "step": 9780 }, { "epoch": 2.26, "learning_rate": 1.2325706802249442e-05, "loss": 0.0, "step": 9781 }, { "epoch": 2.26, "learning_rate": 1.2321855018873739e-05, "loss": 0.0, "step": 9782 }, { "epoch": 2.26, "learning_rate": 1.2318003235498037e-05, "loss": 0.0, "step": 9783 }, { "epoch": 2.26, "learning_rate": 1.2314151452122332e-05, "loss": 0.0, "step": 9784 }, { "epoch": 2.26, "learning_rate": 1.231029966874663e-05, "loss": 0.0, "step": 9785 }, { "epoch": 2.26, "learning_rate": 1.2306447885370927e-05, "loss": 0.0, "step": 9786 }, { "epoch": 2.26, "learning_rate": 1.2302596101995224e-05, "loss": 0.0, "step": 9787 }, { "epoch": 2.26, "learning_rate": 1.2298744318619522e-05, "loss": 0.0, "step": 9788 }, { "epoch": 2.26, "learning_rate": 1.2294892535243819e-05, "loss": 0.0, "step": 9789 }, { "epoch": 2.26, "learning_rate": 1.2291040751868115e-05, "loss": 0.0, "step": 9790 }, { "epoch": 2.26, "learning_rate": 1.2287188968492412e-05, "loss": 0.0, "step": 9791 }, { "epoch": 2.26, "learning_rate": 1.228333718511671e-05, "loss": 0.0, "step": 9792 }, { "epoch": 2.26, "learning_rate": 1.2279485401741005e-05, "loss": 0.0, "step": 9793 }, { "epoch": 2.26, "learning_rate": 1.2275633618365304e-05, "loss": 0.0, "step": 9794 }, { "epoch": 2.26, "learning_rate": 1.2271781834989602e-05, "loss": 0.0, "step": 9795 }, { "epoch": 2.26, "learning_rate": 1.2267930051613897e-05, "loss": 0.0, "step": 9796 }, { "epoch": 2.26, "learning_rate": 1.2264078268238195e-05, "loss": 0.0, "step": 9797 }, { "epoch": 2.26, "learning_rate": 1.2260226484862492e-05, "loss": 0.0, "step": 9798 }, { "epoch": 2.26, "learning_rate": 1.2256374701486789e-05, "loss": 0.0, "step": 9799 }, { "epoch": 2.26, "learning_rate": 1.2252522918111085e-05, "loss": 0.0, "step": 9800 }, { "epoch": 2.27, "learning_rate": 1.2248671134735384e-05, "loss": 0.0, "step": 9801 }, { "epoch": 2.27, "learning_rate": 1.224481935135968e-05, "loss": 0.0, "step": 9802 }, { "epoch": 2.27, "learning_rate": 1.2240967567983977e-05, "loss": 0.0, "step": 9803 }, { "epoch": 2.27, "learning_rate": 1.2237115784608275e-05, "loss": 0.0, "step": 9804 }, { "epoch": 2.27, "learning_rate": 1.223326400123257e-05, "loss": 0.0, "step": 9805 }, { "epoch": 2.27, "learning_rate": 1.2229412217856869e-05, "loss": 0.0, "step": 9806 }, { "epoch": 2.27, "learning_rate": 1.2225560434481165e-05, "loss": 0.0, "step": 9807 }, { "epoch": 2.27, "learning_rate": 1.2221708651105462e-05, "loss": 0.0, "step": 9808 }, { "epoch": 2.27, "learning_rate": 1.221785686772976e-05, "loss": 0.0, "step": 9809 }, { "epoch": 2.27, "learning_rate": 1.2214005084354057e-05, "loss": 0.0, "step": 9810 }, { "epoch": 2.27, "learning_rate": 1.2210153300978354e-05, "loss": 0.0, "step": 9811 }, { "epoch": 2.27, "learning_rate": 1.220630151760265e-05, "loss": 0.0, "step": 9812 }, { "epoch": 2.27, "learning_rate": 1.2202449734226949e-05, "loss": 0.0, "step": 9813 }, { "epoch": 2.27, "learning_rate": 1.2198597950851244e-05, "loss": 0.0, "step": 9814 }, { "epoch": 2.27, "learning_rate": 1.2194746167475542e-05, "loss": 0.0, "step": 9815 }, { "epoch": 2.27, "learning_rate": 1.219089438409984e-05, "loss": 0.0, "step": 9816 }, { "epoch": 2.27, "learning_rate": 1.2187042600724135e-05, "loss": 0.0, "step": 9817 }, { "epoch": 2.27, "learning_rate": 1.2183190817348433e-05, "loss": 0.0, "step": 9818 }, { "epoch": 2.27, "learning_rate": 1.217933903397273e-05, "loss": 0.0, "step": 9819 }, { "epoch": 2.27, "learning_rate": 1.2175487250597027e-05, "loss": 0.0, "step": 9820 }, { "epoch": 2.27, "learning_rate": 1.2171635467221323e-05, "loss": 0.0, "step": 9821 }, { "epoch": 2.27, "learning_rate": 1.2167783683845622e-05, "loss": 0.0, "step": 9822 }, { "epoch": 2.27, "learning_rate": 1.2163931900469918e-05, "loss": 0.0, "step": 9823 }, { "epoch": 2.27, "learning_rate": 1.2160080117094215e-05, "loss": 0.0, "step": 9824 }, { "epoch": 2.27, "learning_rate": 1.2156228333718513e-05, "loss": 0.0, "step": 9825 }, { "epoch": 2.27, "learning_rate": 1.2152376550342808e-05, "loss": 0.0, "step": 9826 }, { "epoch": 2.27, "learning_rate": 1.2148524766967107e-05, "loss": 0.0, "step": 9827 }, { "epoch": 2.27, "learning_rate": 1.2144672983591403e-05, "loss": 0.0, "step": 9828 }, { "epoch": 2.27, "learning_rate": 1.21408212002157e-05, "loss": 0.0001, "step": 9829 }, { "epoch": 2.27, "learning_rate": 1.2136969416839998e-05, "loss": 0.0, "step": 9830 }, { "epoch": 2.27, "learning_rate": 1.2133117633464295e-05, "loss": 0.0, "step": 9831 }, { "epoch": 2.27, "learning_rate": 1.2129265850088592e-05, "loss": 0.0, "step": 9832 }, { "epoch": 2.27, "learning_rate": 1.2125414066712888e-05, "loss": 0.0, "step": 9833 }, { "epoch": 2.27, "learning_rate": 1.2121562283337187e-05, "loss": 0.0, "step": 9834 }, { "epoch": 2.27, "learning_rate": 1.2117710499961482e-05, "loss": 0.0, "step": 9835 }, { "epoch": 2.27, "learning_rate": 1.211385871658578e-05, "loss": 0.0, "step": 9836 }, { "epoch": 2.27, "learning_rate": 1.2110006933210077e-05, "loss": 0.0, "step": 9837 }, { "epoch": 2.27, "learning_rate": 1.2106155149834373e-05, "loss": 0.0, "step": 9838 }, { "epoch": 2.27, "learning_rate": 1.2102303366458672e-05, "loss": 0.0, "step": 9839 }, { "epoch": 2.27, "learning_rate": 1.2098451583082968e-05, "loss": 0.0, "step": 9840 }, { "epoch": 2.27, "learning_rate": 1.2094599799707265e-05, "loss": 0.0, "step": 9841 }, { "epoch": 2.27, "learning_rate": 1.2090748016331562e-05, "loss": 0.0, "step": 9842 }, { "epoch": 2.27, "learning_rate": 1.208689623295586e-05, "loss": 0.0, "step": 9843 }, { "epoch": 2.28, "learning_rate": 1.2083044449580157e-05, "loss": 0.0, "step": 9844 }, { "epoch": 2.28, "learning_rate": 1.2079192666204453e-05, "loss": 0.0, "step": 9845 }, { "epoch": 2.28, "learning_rate": 1.207534088282875e-05, "loss": 0.0, "step": 9846 }, { "epoch": 2.28, "learning_rate": 1.2071489099453046e-05, "loss": 0.0, "step": 9847 }, { "epoch": 2.28, "learning_rate": 1.2067637316077345e-05, "loss": 0.0, "step": 9848 }, { "epoch": 2.28, "learning_rate": 1.2063785532701641e-05, "loss": 0.0, "step": 9849 }, { "epoch": 2.28, "learning_rate": 1.2059933749325938e-05, "loss": 0.0, "step": 9850 }, { "epoch": 2.28, "learning_rate": 1.2056081965950236e-05, "loss": 0.0, "step": 9851 }, { "epoch": 2.28, "learning_rate": 1.2052230182574533e-05, "loss": 0.0, "step": 9852 }, { "epoch": 2.28, "learning_rate": 1.204837839919883e-05, "loss": 0.0, "step": 9853 }, { "epoch": 2.28, "learning_rate": 1.2044526615823126e-05, "loss": 0.0, "step": 9854 }, { "epoch": 2.28, "learning_rate": 1.2040674832447423e-05, "loss": 0.0, "step": 9855 }, { "epoch": 2.28, "learning_rate": 1.203682304907172e-05, "loss": 0.0, "step": 9856 }, { "epoch": 2.28, "learning_rate": 1.2032971265696018e-05, "loss": 0.0, "step": 9857 }, { "epoch": 2.28, "learning_rate": 1.2029119482320315e-05, "loss": 0.0, "step": 9858 }, { "epoch": 2.28, "learning_rate": 1.2025267698944611e-05, "loss": 0.0, "step": 9859 }, { "epoch": 2.28, "learning_rate": 1.202141591556891e-05, "loss": 0.0, "step": 9860 }, { "epoch": 2.28, "learning_rate": 1.2017564132193206e-05, "loss": 0.0, "step": 9861 }, { "epoch": 2.28, "learning_rate": 1.2013712348817503e-05, "loss": 0.0, "step": 9862 }, { "epoch": 2.28, "learning_rate": 1.20098605654418e-05, "loss": 0.0, "step": 9863 }, { "epoch": 2.28, "learning_rate": 1.2006008782066096e-05, "loss": 0.0, "step": 9864 }, { "epoch": 2.28, "learning_rate": 1.2002156998690395e-05, "loss": 0.0, "step": 9865 }, { "epoch": 2.28, "learning_rate": 1.1998305215314691e-05, "loss": 0.0, "step": 9866 }, { "epoch": 2.28, "learning_rate": 1.1994453431938988e-05, "loss": 0.0, "step": 9867 }, { "epoch": 2.28, "learning_rate": 1.1990601648563285e-05, "loss": 0.0, "step": 9868 }, { "epoch": 2.28, "learning_rate": 1.1986749865187583e-05, "loss": 0.0, "step": 9869 }, { "epoch": 2.28, "learning_rate": 1.198289808181188e-05, "loss": 0.0, "step": 9870 }, { "epoch": 2.28, "learning_rate": 1.1979046298436176e-05, "loss": 0.0, "step": 9871 }, { "epoch": 2.28, "learning_rate": 1.1975194515060475e-05, "loss": 0.0, "step": 9872 }, { "epoch": 2.28, "learning_rate": 1.197134273168477e-05, "loss": 0.0, "step": 9873 }, { "epoch": 2.28, "learning_rate": 1.1967490948309068e-05, "loss": 0.0, "step": 9874 }, { "epoch": 2.28, "learning_rate": 1.1963639164933365e-05, "loss": 0.0, "step": 9875 }, { "epoch": 2.28, "learning_rate": 1.1959787381557661e-05, "loss": 0.0, "step": 9876 }, { "epoch": 2.28, "learning_rate": 1.1955935598181958e-05, "loss": 0.0, "step": 9877 }, { "epoch": 2.28, "learning_rate": 1.1952083814806256e-05, "loss": 0.0, "step": 9878 }, { "epoch": 2.28, "learning_rate": 1.1948232031430553e-05, "loss": 0.0, "step": 9879 }, { "epoch": 2.28, "learning_rate": 1.194438024805485e-05, "loss": 0.0, "step": 9880 }, { "epoch": 2.28, "learning_rate": 1.1940528464679148e-05, "loss": 0.0, "step": 9881 }, { "epoch": 2.28, "learning_rate": 1.1936676681303444e-05, "loss": 0.0001, "step": 9882 }, { "epoch": 2.28, "learning_rate": 1.1932824897927741e-05, "loss": 0.0, "step": 9883 }, { "epoch": 2.28, "learning_rate": 1.1928973114552038e-05, "loss": 0.0, "step": 9884 }, { "epoch": 2.28, "learning_rate": 1.1925121331176334e-05, "loss": 0.0001, "step": 9885 }, { "epoch": 2.28, "learning_rate": 1.1921269547800633e-05, "loss": 0.0, "step": 9886 }, { "epoch": 2.28, "learning_rate": 1.191741776442493e-05, "loss": 0.0, "step": 9887 }, { "epoch": 2.29, "learning_rate": 1.1913565981049226e-05, "loss": 0.0, "step": 9888 }, { "epoch": 2.29, "learning_rate": 1.1909714197673523e-05, "loss": 0.0, "step": 9889 }, { "epoch": 2.29, "learning_rate": 1.1905862414297821e-05, "loss": 0.0003, "step": 9890 }, { "epoch": 2.29, "learning_rate": 1.1902010630922118e-05, "loss": 0.0, "step": 9891 }, { "epoch": 2.29, "learning_rate": 1.1898158847546414e-05, "loss": 0.0, "step": 9892 }, { "epoch": 2.29, "learning_rate": 1.1894307064170713e-05, "loss": 0.0, "step": 9893 }, { "epoch": 2.29, "learning_rate": 1.1890455280795008e-05, "loss": 0.0, "step": 9894 }, { "epoch": 2.29, "learning_rate": 1.1886603497419306e-05, "loss": 0.0, "step": 9895 }, { "epoch": 2.29, "learning_rate": 1.1882751714043603e-05, "loss": 0.0, "step": 9896 }, { "epoch": 2.29, "learning_rate": 1.18788999306679e-05, "loss": 0.0, "step": 9897 }, { "epoch": 2.29, "learning_rate": 1.1875048147292196e-05, "loss": 0.0, "step": 9898 }, { "epoch": 2.29, "learning_rate": 1.1871196363916494e-05, "loss": 0.0, "step": 9899 }, { "epoch": 2.29, "learning_rate": 1.1867344580540791e-05, "loss": 0.0, "step": 9900 }, { "epoch": 2.29, "learning_rate": 1.1863492797165088e-05, "loss": 0.0, "step": 9901 }, { "epoch": 2.29, "learning_rate": 1.1859641013789386e-05, "loss": 0.0, "step": 9902 }, { "epoch": 2.29, "learning_rate": 1.1855789230413681e-05, "loss": 0.0, "step": 9903 }, { "epoch": 2.29, "learning_rate": 1.185193744703798e-05, "loss": 0.0, "step": 9904 }, { "epoch": 2.29, "learning_rate": 1.1848085663662276e-05, "loss": 0.0, "step": 9905 }, { "epoch": 2.29, "learning_rate": 1.1844233880286573e-05, "loss": 0.0, "step": 9906 }, { "epoch": 2.29, "learning_rate": 1.1840382096910871e-05, "loss": 0.0, "step": 9907 }, { "epoch": 2.29, "learning_rate": 1.1836530313535168e-05, "loss": 0.0, "step": 9908 }, { "epoch": 2.29, "learning_rate": 1.1832678530159464e-05, "loss": 0.0, "step": 9909 }, { "epoch": 2.29, "learning_rate": 1.182882674678376e-05, "loss": 0.0, "step": 9910 }, { "epoch": 2.29, "learning_rate": 1.182497496340806e-05, "loss": 0.0, "step": 9911 }, { "epoch": 2.29, "learning_rate": 1.1821123180032354e-05, "loss": 0.0, "step": 9912 }, { "epoch": 2.29, "learning_rate": 1.1817271396656652e-05, "loss": 0.0, "step": 9913 }, { "epoch": 2.29, "learning_rate": 1.181341961328095e-05, "loss": 0.0, "step": 9914 }, { "epoch": 2.29, "learning_rate": 1.1809567829905246e-05, "loss": 0.0, "step": 9915 }, { "epoch": 2.29, "learning_rate": 1.1805716046529544e-05, "loss": 0.0, "step": 9916 }, { "epoch": 2.29, "learning_rate": 1.180186426315384e-05, "loss": 0.0001, "step": 9917 }, { "epoch": 2.29, "learning_rate": 1.1798012479778137e-05, "loss": 0.0, "step": 9918 }, { "epoch": 2.29, "learning_rate": 1.1794160696402434e-05, "loss": 0.0, "step": 9919 }, { "epoch": 2.29, "learning_rate": 1.1790308913026732e-05, "loss": 0.0, "step": 9920 }, { "epoch": 2.29, "learning_rate": 1.1786457129651029e-05, "loss": 0.0, "step": 9921 }, { "epoch": 2.29, "learning_rate": 1.1782605346275326e-05, "loss": 0.0, "step": 9922 }, { "epoch": 2.29, "learning_rate": 1.1778753562899624e-05, "loss": 0.0, "step": 9923 }, { "epoch": 2.29, "learning_rate": 1.1774901779523919e-05, "loss": 0.0, "step": 9924 }, { "epoch": 2.29, "learning_rate": 1.1771049996148217e-05, "loss": 0.0001, "step": 9925 }, { "epoch": 2.29, "learning_rate": 1.1767198212772514e-05, "loss": 0.0, "step": 9926 }, { "epoch": 2.29, "learning_rate": 1.176334642939681e-05, "loss": 0.0, "step": 9927 }, { "epoch": 2.29, "learning_rate": 1.1759494646021109e-05, "loss": 0.0, "step": 9928 }, { "epoch": 2.29, "learning_rate": 1.1755642862645406e-05, "loss": 0.0, "step": 9929 }, { "epoch": 2.29, "learning_rate": 1.1751791079269702e-05, "loss": 0.0, "step": 9930 }, { "epoch": 2.3, "learning_rate": 1.1747939295893999e-05, "loss": 0.0, "step": 9931 }, { "epoch": 2.3, "learning_rate": 1.1744087512518297e-05, "loss": 0.0, "step": 9932 }, { "epoch": 2.3, "learning_rate": 1.1740235729142592e-05, "loss": 0.0, "step": 9933 }, { "epoch": 2.3, "learning_rate": 1.173638394576689e-05, "loss": 0.0, "step": 9934 }, { "epoch": 2.3, "learning_rate": 1.1732532162391189e-05, "loss": 0.0021, "step": 9935 }, { "epoch": 2.3, "learning_rate": 1.1728680379015484e-05, "loss": 0.0, "step": 9936 }, { "epoch": 2.3, "learning_rate": 1.1724828595639782e-05, "loss": 0.0, "step": 9937 }, { "epoch": 2.3, "learning_rate": 1.1720976812264079e-05, "loss": 0.0, "step": 9938 }, { "epoch": 2.3, "learning_rate": 1.1717125028888376e-05, "loss": 0.0, "step": 9939 }, { "epoch": 2.3, "learning_rate": 1.1713273245512672e-05, "loss": 0.0, "step": 9940 }, { "epoch": 2.3, "learning_rate": 1.170942146213697e-05, "loss": 0.0, "step": 9941 }, { "epoch": 2.3, "learning_rate": 1.1705569678761267e-05, "loss": 0.0001, "step": 9942 }, { "epoch": 2.3, "learning_rate": 1.1701717895385564e-05, "loss": 0.0, "step": 9943 }, { "epoch": 2.3, "learning_rate": 1.1697866112009862e-05, "loss": 0.0, "step": 9944 }, { "epoch": 2.3, "learning_rate": 1.1694014328634157e-05, "loss": 0.0, "step": 9945 }, { "epoch": 2.3, "learning_rate": 1.1690162545258455e-05, "loss": 0.0, "step": 9946 }, { "epoch": 2.3, "learning_rate": 1.1686310761882752e-05, "loss": 0.0, "step": 9947 }, { "epoch": 2.3, "learning_rate": 1.1682458978507049e-05, "loss": 0.0082, "step": 9948 }, { "epoch": 2.3, "learning_rate": 1.1678607195131347e-05, "loss": 0.0, "step": 9949 }, { "epoch": 2.3, "learning_rate": 1.1674755411755644e-05, "loss": 0.0, "step": 9950 }, { "epoch": 2.3, "learning_rate": 1.167090362837994e-05, "loss": 0.0, "step": 9951 }, { "epoch": 2.3, "learning_rate": 1.1667051845004237e-05, "loss": 0.0, "step": 9952 }, { "epoch": 2.3, "learning_rate": 1.1663200061628535e-05, "loss": 0.0, "step": 9953 }, { "epoch": 2.3, "learning_rate": 1.165934827825283e-05, "loss": 0.0, "step": 9954 }, { "epoch": 2.3, "learning_rate": 1.1655496494877129e-05, "loss": 0.0, "step": 9955 }, { "epoch": 2.3, "learning_rate": 1.1651644711501427e-05, "loss": 0.0, "step": 9956 }, { "epoch": 2.3, "learning_rate": 1.1647792928125722e-05, "loss": 0.0, "step": 9957 }, { "epoch": 2.3, "learning_rate": 1.164394114475002e-05, "loss": 0.0, "step": 9958 }, { "epoch": 2.3, "learning_rate": 1.1640089361374317e-05, "loss": 0.0, "step": 9959 }, { "epoch": 2.3, "learning_rate": 1.1636237577998614e-05, "loss": 0.0, "step": 9960 }, { "epoch": 2.3, "learning_rate": 1.163238579462291e-05, "loss": 0.0, "step": 9961 }, { "epoch": 2.3, "learning_rate": 1.1628534011247209e-05, "loss": 0.0, "step": 9962 }, { "epoch": 2.3, "learning_rate": 1.1624682227871505e-05, "loss": 0.0, "step": 9963 }, { "epoch": 2.3, "learning_rate": 1.1620830444495802e-05, "loss": 0.0001, "step": 9964 }, { "epoch": 2.3, "learning_rate": 1.16169786611201e-05, "loss": 0.0, "step": 9965 }, { "epoch": 2.3, "learning_rate": 1.1613126877744395e-05, "loss": 0.0, "step": 9966 }, { "epoch": 2.3, "learning_rate": 1.1609275094368694e-05, "loss": 0.0, "step": 9967 }, { "epoch": 2.3, "learning_rate": 1.160542331099299e-05, "loss": 0.0, "step": 9968 }, { "epoch": 2.3, "learning_rate": 1.1601571527617287e-05, "loss": 0.0, "step": 9969 }, { "epoch": 2.3, "learning_rate": 1.1597719744241585e-05, "loss": 0.0, "step": 9970 }, { "epoch": 2.3, "learning_rate": 1.1593867960865882e-05, "loss": 0.0, "step": 9971 }, { "epoch": 2.3, "learning_rate": 1.1590016177490179e-05, "loss": 0.0, "step": 9972 }, { "epoch": 2.3, "learning_rate": 1.1586164394114475e-05, "loss": 0.0, "step": 9973 }, { "epoch": 2.31, "learning_rate": 1.1582312610738774e-05, "loss": 0.0, "step": 9974 }, { "epoch": 2.31, "learning_rate": 1.1578460827363068e-05, "loss": 0.0, "step": 9975 }, { "epoch": 2.31, "learning_rate": 1.1574609043987367e-05, "loss": 0.0, "step": 9976 }, { "epoch": 2.31, "learning_rate": 1.1570757260611665e-05, "loss": 0.0, "step": 9977 }, { "epoch": 2.31, "learning_rate": 1.156690547723596e-05, "loss": 0.0001, "step": 9978 }, { "epoch": 2.31, "learning_rate": 1.1563053693860258e-05, "loss": 0.0, "step": 9979 }, { "epoch": 2.31, "learning_rate": 1.1559201910484555e-05, "loss": 0.0001, "step": 9980 }, { "epoch": 2.31, "learning_rate": 1.1555350127108852e-05, "loss": 0.0, "step": 9981 }, { "epoch": 2.31, "learning_rate": 1.1551498343733148e-05, "loss": 0.0002, "step": 9982 }, { "epoch": 2.31, "learning_rate": 1.1547646560357447e-05, "loss": 0.0, "step": 9983 }, { "epoch": 2.31, "learning_rate": 1.1543794776981743e-05, "loss": 0.0, "step": 9984 }, { "epoch": 2.31, "learning_rate": 1.153994299360604e-05, "loss": 0.0, "step": 9985 }, { "epoch": 2.31, "learning_rate": 1.1536091210230338e-05, "loss": 0.0, "step": 9986 }, { "epoch": 2.31, "learning_rate": 1.1532239426854633e-05, "loss": 0.0, "step": 9987 }, { "epoch": 2.31, "learning_rate": 1.1528387643478932e-05, "loss": 0.0, "step": 9988 }, { "epoch": 2.31, "learning_rate": 1.1524535860103228e-05, "loss": 0.0, "step": 9989 }, { "epoch": 2.31, "learning_rate": 1.1520684076727525e-05, "loss": 0.0, "step": 9990 }, { "epoch": 2.31, "learning_rate": 1.1516832293351823e-05, "loss": 0.0, "step": 9991 }, { "epoch": 2.31, "learning_rate": 1.151298050997612e-05, "loss": 0.0, "step": 9992 }, { "epoch": 2.31, "learning_rate": 1.1509128726600417e-05, "loss": 0.0, "step": 9993 }, { "epoch": 2.31, "learning_rate": 1.1505276943224713e-05, "loss": 0.0, "step": 9994 }, { "epoch": 2.31, "learning_rate": 1.1501425159849012e-05, "loss": 0.0, "step": 9995 }, { "epoch": 2.31, "learning_rate": 1.1497573376473307e-05, "loss": 0.0, "step": 9996 }, { "epoch": 2.31, "learning_rate": 1.1493721593097605e-05, "loss": 0.0, "step": 9997 }, { "epoch": 2.31, "learning_rate": 1.1489869809721902e-05, "loss": 0.0, "step": 9998 }, { "epoch": 2.31, "learning_rate": 1.1486018026346198e-05, "loss": 0.0, "step": 9999 }, { "epoch": 2.31, "learning_rate": 1.1482166242970497e-05, "loss": 0.0, "step": 10000 }, { "epoch": 2.31, "learning_rate": 1.1478314459594793e-05, "loss": 0.0, "step": 10001 }, { "epoch": 2.31, "learning_rate": 1.147446267621909e-05, "loss": 0.0, "step": 10002 }, { "epoch": 2.31, "learning_rate": 1.1470610892843387e-05, "loss": 0.0, "step": 10003 }, { "epoch": 2.31, "learning_rate": 1.1466759109467685e-05, "loss": 0.0, "step": 10004 }, { "epoch": 2.31, "learning_rate": 1.1462907326091982e-05, "loss": 0.0, "step": 10005 }, { "epoch": 2.31, "learning_rate": 1.1459055542716278e-05, "loss": 0.0, "step": 10006 }, { "epoch": 2.31, "learning_rate": 1.1455203759340575e-05, "loss": 0.0, "step": 10007 }, { "epoch": 2.31, "learning_rate": 1.1451351975964871e-05, "loss": 0.0001, "step": 10008 }, { "epoch": 2.31, "learning_rate": 1.144750019258917e-05, "loss": 0.0, "step": 10009 }, { "epoch": 2.31, "learning_rate": 1.1443648409213466e-05, "loss": 0.0, "step": 10010 }, { "epoch": 2.31, "learning_rate": 1.1439796625837763e-05, "loss": 0.0, "step": 10011 }, { "epoch": 2.31, "learning_rate": 1.1435944842462061e-05, "loss": 0.0, "step": 10012 }, { "epoch": 2.31, "learning_rate": 1.1432093059086358e-05, "loss": 0.0, "step": 10013 }, { "epoch": 2.31, "learning_rate": 1.1428241275710655e-05, "loss": 0.0, "step": 10014 }, { "epoch": 2.31, "learning_rate": 1.1424389492334951e-05, "loss": 0.0, "step": 10015 }, { "epoch": 2.31, "learning_rate": 1.1420537708959248e-05, "loss": 0.0, "step": 10016 }, { "epoch": 2.31, "learning_rate": 1.1416685925583545e-05, "loss": 0.0001, "step": 10017 }, { "epoch": 2.32, "learning_rate": 1.1412834142207843e-05, "loss": 0.0, "step": 10018 }, { "epoch": 2.32, "learning_rate": 1.140898235883214e-05, "loss": 0.0, "step": 10019 }, { "epoch": 2.32, "learning_rate": 1.1405130575456436e-05, "loss": 0.0, "step": 10020 }, { "epoch": 2.32, "learning_rate": 1.1401278792080735e-05, "loss": 0.0, "step": 10021 }, { "epoch": 2.32, "learning_rate": 1.1397427008705031e-05, "loss": 0.0, "step": 10022 }, { "epoch": 2.32, "learning_rate": 1.1393575225329328e-05, "loss": 0.0002, "step": 10023 }, { "epoch": 2.32, "learning_rate": 1.1389723441953625e-05, "loss": 0.0, "step": 10024 }, { "epoch": 2.32, "learning_rate": 1.1385871658577921e-05, "loss": 0.0, "step": 10025 }, { "epoch": 2.32, "learning_rate": 1.138201987520222e-05, "loss": 0.0, "step": 10026 }, { "epoch": 2.32, "learning_rate": 1.1378168091826516e-05, "loss": 0.0, "step": 10027 }, { "epoch": 2.32, "learning_rate": 1.1374316308450813e-05, "loss": 0.0, "step": 10028 }, { "epoch": 2.32, "learning_rate": 1.137046452507511e-05, "loss": 0.0, "step": 10029 }, { "epoch": 2.32, "learning_rate": 1.1366612741699408e-05, "loss": 0.0, "step": 10030 }, { "epoch": 2.32, "learning_rate": 1.1362760958323705e-05, "loss": 0.0, "step": 10031 }, { "epoch": 2.32, "learning_rate": 1.1358909174948001e-05, "loss": 0.0, "step": 10032 }, { "epoch": 2.32, "learning_rate": 1.13550573915723e-05, "loss": 0.0, "step": 10033 }, { "epoch": 2.32, "learning_rate": 1.1351205608196595e-05, "loss": 0.0, "step": 10034 }, { "epoch": 2.32, "learning_rate": 1.1347353824820893e-05, "loss": 0.0, "step": 10035 }, { "epoch": 2.32, "learning_rate": 1.134350204144519e-05, "loss": 0.0, "step": 10036 }, { "epoch": 2.32, "learning_rate": 1.1339650258069486e-05, "loss": 0.0, "step": 10037 }, { "epoch": 2.32, "learning_rate": 1.1335798474693783e-05, "loss": 0.0, "step": 10038 }, { "epoch": 2.32, "learning_rate": 1.1331946691318081e-05, "loss": 0.0, "step": 10039 }, { "epoch": 2.32, "learning_rate": 1.1328094907942378e-05, "loss": 0.0, "step": 10040 }, { "epoch": 2.32, "learning_rate": 1.1324243124566674e-05, "loss": 0.0, "step": 10041 }, { "epoch": 2.32, "learning_rate": 1.1320391341190973e-05, "loss": 0.0, "step": 10042 }, { "epoch": 2.32, "learning_rate": 1.1316539557815268e-05, "loss": 0.0, "step": 10043 }, { "epoch": 2.32, "learning_rate": 1.1312687774439566e-05, "loss": 0.0002, "step": 10044 }, { "epoch": 2.32, "learning_rate": 1.1308835991063863e-05, "loss": 0.0, "step": 10045 }, { "epoch": 2.32, "learning_rate": 1.130498420768816e-05, "loss": 0.0, "step": 10046 }, { "epoch": 2.32, "learning_rate": 1.1301132424312458e-05, "loss": 0.0, "step": 10047 }, { "epoch": 2.32, "learning_rate": 1.1297280640936754e-05, "loss": 0.0, "step": 10048 }, { "epoch": 2.32, "learning_rate": 1.1293428857561051e-05, "loss": 0.0, "step": 10049 }, { "epoch": 2.32, "learning_rate": 1.1289577074185348e-05, "loss": 0.0001, "step": 10050 }, { "epoch": 2.32, "learning_rate": 1.1285725290809646e-05, "loss": 0.0, "step": 10051 }, { "epoch": 2.32, "learning_rate": 1.1281873507433943e-05, "loss": 0.0, "step": 10052 }, { "epoch": 2.32, "learning_rate": 1.127802172405824e-05, "loss": 0.0, "step": 10053 }, { "epoch": 2.32, "learning_rate": 1.1274169940682538e-05, "loss": 0.0, "step": 10054 }, { "epoch": 2.32, "learning_rate": 1.1270318157306833e-05, "loss": 0.0, "step": 10055 }, { "epoch": 2.32, "learning_rate": 1.1266466373931131e-05, "loss": 0.0, "step": 10056 }, { "epoch": 2.32, "learning_rate": 1.1262614590555428e-05, "loss": 0.0, "step": 10057 }, { "epoch": 2.32, "learning_rate": 1.1258762807179724e-05, "loss": 0.0, "step": 10058 }, { "epoch": 2.32, "learning_rate": 1.1254911023804021e-05, "loss": 0.0, "step": 10059 }, { "epoch": 2.32, "learning_rate": 1.125105924042832e-05, "loss": 0.0, "step": 10060 }, { "epoch": 2.33, "learning_rate": 1.1247207457052616e-05, "loss": 0.0, "step": 10061 }, { "epoch": 2.33, "learning_rate": 1.1243355673676913e-05, "loss": 0.0, "step": 10062 }, { "epoch": 2.33, "learning_rate": 1.1239503890301211e-05, "loss": 0.0, "step": 10063 }, { "epoch": 2.33, "learning_rate": 1.1235652106925506e-05, "loss": 0.0, "step": 10064 }, { "epoch": 2.33, "learning_rate": 1.1231800323549804e-05, "loss": 0.0, "step": 10065 }, { "epoch": 2.33, "learning_rate": 1.1227948540174101e-05, "loss": 0.0, "step": 10066 }, { "epoch": 2.33, "learning_rate": 1.1224096756798397e-05, "loss": 0.0, "step": 10067 }, { "epoch": 2.33, "learning_rate": 1.1220244973422696e-05, "loss": 0.0, "step": 10068 }, { "epoch": 2.33, "learning_rate": 1.1216393190046992e-05, "loss": 0.0, "step": 10069 }, { "epoch": 2.33, "learning_rate": 1.1212541406671289e-05, "loss": 0.0, "step": 10070 }, { "epoch": 2.33, "learning_rate": 1.1208689623295586e-05, "loss": 0.0, "step": 10071 }, { "epoch": 2.33, "learning_rate": 1.1204837839919884e-05, "loss": 0.0, "step": 10072 }, { "epoch": 2.33, "learning_rate": 1.1200986056544179e-05, "loss": 0.0, "step": 10073 }, { "epoch": 2.33, "learning_rate": 1.1197134273168477e-05, "loss": 0.0, "step": 10074 }, { "epoch": 2.33, "learning_rate": 1.1193282489792776e-05, "loss": 0.0, "step": 10075 }, { "epoch": 2.33, "learning_rate": 1.118943070641707e-05, "loss": 0.0, "step": 10076 }, { "epoch": 2.33, "learning_rate": 1.1185578923041369e-05, "loss": 0.0, "step": 10077 }, { "epoch": 2.33, "learning_rate": 1.1181727139665666e-05, "loss": 0.0, "step": 10078 }, { "epoch": 2.33, "learning_rate": 1.1177875356289962e-05, "loss": 0.0001, "step": 10079 }, { "epoch": 2.33, "learning_rate": 1.1174023572914259e-05, "loss": 0.0, "step": 10080 }, { "epoch": 2.33, "learning_rate": 1.1170171789538557e-05, "loss": 0.0, "step": 10081 }, { "epoch": 2.33, "learning_rate": 1.1166320006162854e-05, "loss": 0.0, "step": 10082 }, { "epoch": 2.33, "learning_rate": 1.116246822278715e-05, "loss": 0.0, "step": 10083 }, { "epoch": 2.33, "learning_rate": 1.1158616439411449e-05, "loss": 0.0, "step": 10084 }, { "epoch": 2.33, "learning_rate": 1.1154764656035744e-05, "loss": 0.0, "step": 10085 }, { "epoch": 2.33, "learning_rate": 1.1150912872660042e-05, "loss": 0.0, "step": 10086 }, { "epoch": 2.33, "learning_rate": 1.1147061089284339e-05, "loss": 0.0, "step": 10087 }, { "epoch": 2.33, "learning_rate": 1.1143209305908636e-05, "loss": 0.0, "step": 10088 }, { "epoch": 2.33, "learning_rate": 1.1139357522532934e-05, "loss": 0.0, "step": 10089 }, { "epoch": 2.33, "learning_rate": 1.113550573915723e-05, "loss": 0.0, "step": 10090 }, { "epoch": 2.33, "learning_rate": 1.1131653955781527e-05, "loss": 0.0, "step": 10091 }, { "epoch": 2.33, "learning_rate": 1.1127802172405824e-05, "loss": 0.0002, "step": 10092 }, { "epoch": 2.33, "learning_rate": 1.1123950389030122e-05, "loss": 0.0, "step": 10093 }, { "epoch": 2.33, "learning_rate": 1.1120098605654417e-05, "loss": 0.0002, "step": 10094 }, { "epoch": 2.33, "learning_rate": 1.1116246822278716e-05, "loss": 0.0, "step": 10095 }, { "epoch": 2.33, "learning_rate": 1.1112395038903014e-05, "loss": 0.0, "step": 10096 }, { "epoch": 2.33, "learning_rate": 1.1108543255527309e-05, "loss": 0.0, "step": 10097 }, { "epoch": 2.33, "learning_rate": 1.1104691472151607e-05, "loss": 0.0, "step": 10098 }, { "epoch": 2.33, "learning_rate": 1.1100839688775904e-05, "loss": 0.0, "step": 10099 }, { "epoch": 2.33, "learning_rate": 1.10969879054002e-05, "loss": 0.0, "step": 10100 }, { "epoch": 2.33, "learning_rate": 1.1093136122024497e-05, "loss": 0.0, "step": 10101 }, { "epoch": 2.33, "learning_rate": 1.1089284338648795e-05, "loss": 0.0, "step": 10102 }, { "epoch": 2.33, "learning_rate": 1.1085432555273092e-05, "loss": 0.0, "step": 10103 }, { "epoch": 2.34, "learning_rate": 1.1081580771897389e-05, "loss": 0.0, "step": 10104 }, { "epoch": 2.34, "learning_rate": 1.1077728988521687e-05, "loss": 0.0, "step": 10105 }, { "epoch": 2.34, "learning_rate": 1.1073877205145982e-05, "loss": 0.0, "step": 10106 }, { "epoch": 2.34, "learning_rate": 1.107002542177028e-05, "loss": 0.0, "step": 10107 }, { "epoch": 2.34, "learning_rate": 1.1066173638394577e-05, "loss": 0.0, "step": 10108 }, { "epoch": 2.34, "learning_rate": 1.1062321855018874e-05, "loss": 0.0, "step": 10109 }, { "epoch": 2.34, "learning_rate": 1.1058470071643172e-05, "loss": 0.0, "step": 10110 }, { "epoch": 2.34, "learning_rate": 1.1054618288267469e-05, "loss": 0.0, "step": 10111 }, { "epoch": 2.34, "learning_rate": 1.1050766504891765e-05, "loss": 0.0, "step": 10112 }, { "epoch": 2.34, "learning_rate": 1.1046914721516062e-05, "loss": 0.0, "step": 10113 }, { "epoch": 2.34, "learning_rate": 1.104306293814036e-05, "loss": 0.0, "step": 10114 }, { "epoch": 2.34, "learning_rate": 1.1039211154764655e-05, "loss": 0.0, "step": 10115 }, { "epoch": 2.34, "learning_rate": 1.1035359371388954e-05, "loss": 0.0, "step": 10116 }, { "epoch": 2.34, "learning_rate": 1.1031507588013252e-05, "loss": 0.0, "step": 10117 }, { "epoch": 2.34, "learning_rate": 1.1027655804637547e-05, "loss": 0.0, "step": 10118 }, { "epoch": 2.34, "learning_rate": 1.1023804021261845e-05, "loss": 0.0, "step": 10119 }, { "epoch": 2.34, "learning_rate": 1.1019952237886142e-05, "loss": 0.0004, "step": 10120 }, { "epoch": 2.34, "learning_rate": 1.1016100454510439e-05, "loss": 0.0001, "step": 10121 }, { "epoch": 2.34, "learning_rate": 1.1012248671134735e-05, "loss": 0.0, "step": 10122 }, { "epoch": 2.34, "learning_rate": 1.1008396887759034e-05, "loss": 0.0, "step": 10123 }, { "epoch": 2.34, "learning_rate": 1.100454510438333e-05, "loss": 0.0, "step": 10124 }, { "epoch": 2.34, "learning_rate": 1.1000693321007627e-05, "loss": 0.0, "step": 10125 }, { "epoch": 2.34, "learning_rate": 1.0996841537631925e-05, "loss": 0.0, "step": 10126 }, { "epoch": 2.34, "learning_rate": 1.099298975425622e-05, "loss": 0.0, "step": 10127 }, { "epoch": 2.34, "learning_rate": 1.0989137970880519e-05, "loss": 0.0, "step": 10128 }, { "epoch": 2.34, "learning_rate": 1.0985286187504815e-05, "loss": 0.0, "step": 10129 }, { "epoch": 2.34, "learning_rate": 1.0981434404129112e-05, "loss": 0.0, "step": 10130 }, { "epoch": 2.34, "learning_rate": 1.097758262075341e-05, "loss": 0.0, "step": 10131 }, { "epoch": 2.34, "learning_rate": 1.0973730837377707e-05, "loss": 0.0, "step": 10132 }, { "epoch": 2.34, "learning_rate": 1.0969879054002003e-05, "loss": 0.0, "step": 10133 }, { "epoch": 2.34, "learning_rate": 1.09660272706263e-05, "loss": 0.0, "step": 10134 }, { "epoch": 2.34, "learning_rate": 1.0962175487250598e-05, "loss": 0.0, "step": 10135 }, { "epoch": 2.34, "learning_rate": 1.0958323703874893e-05, "loss": 0.0, "step": 10136 }, { "epoch": 2.34, "learning_rate": 1.0954471920499192e-05, "loss": 0.0, "step": 10137 }, { "epoch": 2.34, "learning_rate": 1.095062013712349e-05, "loss": 0.0, "step": 10138 }, { "epoch": 2.34, "learning_rate": 1.0946768353747785e-05, "loss": 0.0, "step": 10139 }, { "epoch": 2.34, "learning_rate": 1.0942916570372083e-05, "loss": 0.0, "step": 10140 }, { "epoch": 2.34, "learning_rate": 1.093906478699638e-05, "loss": 0.0, "step": 10141 }, { "epoch": 2.34, "learning_rate": 1.0935213003620677e-05, "loss": 0.0, "step": 10142 }, { "epoch": 2.34, "learning_rate": 1.0931361220244973e-05, "loss": 0.0, "step": 10143 }, { "epoch": 2.34, "learning_rate": 1.0927509436869272e-05, "loss": 0.0, "step": 10144 }, { "epoch": 2.34, "learning_rate": 1.0923657653493568e-05, "loss": 0.0, "step": 10145 }, { "epoch": 2.34, "learning_rate": 1.0919805870117865e-05, "loss": 0.0, "step": 10146 }, { "epoch": 2.35, "learning_rate": 1.0915954086742163e-05, "loss": 0.0, "step": 10147 }, { "epoch": 2.35, "learning_rate": 1.0912102303366458e-05, "loss": 0.0, "step": 10148 }, { "epoch": 2.35, "learning_rate": 1.0908250519990757e-05, "loss": 0.0075, "step": 10149 }, { "epoch": 2.35, "learning_rate": 1.0904398736615053e-05, "loss": 0.0, "step": 10150 }, { "epoch": 2.35, "learning_rate": 1.090054695323935e-05, "loss": 0.0, "step": 10151 }, { "epoch": 2.35, "learning_rate": 1.0896695169863648e-05, "loss": 0.0, "step": 10152 }, { "epoch": 2.35, "learning_rate": 1.0892843386487945e-05, "loss": 0.0, "step": 10153 }, { "epoch": 2.35, "learning_rate": 1.0888991603112242e-05, "loss": 0.0, "step": 10154 }, { "epoch": 2.35, "learning_rate": 1.0885139819736538e-05, "loss": 0.0, "step": 10155 }, { "epoch": 2.35, "learning_rate": 1.0881288036360837e-05, "loss": 0.0, "step": 10156 }, { "epoch": 2.35, "learning_rate": 1.0877436252985132e-05, "loss": 0.0, "step": 10157 }, { "epoch": 2.35, "learning_rate": 1.087358446960943e-05, "loss": 0.0, "step": 10158 }, { "epoch": 2.35, "learning_rate": 1.0869732686233727e-05, "loss": 0.0, "step": 10159 }, { "epoch": 2.35, "learning_rate": 1.0865880902858023e-05, "loss": 0.0, "step": 10160 }, { "epoch": 2.35, "learning_rate": 1.0862029119482322e-05, "loss": 0.0, "step": 10161 }, { "epoch": 2.35, "learning_rate": 1.0858177336106618e-05, "loss": 0.0, "step": 10162 }, { "epoch": 2.35, "learning_rate": 1.0854325552730915e-05, "loss": 0.0, "step": 10163 }, { "epoch": 2.35, "learning_rate": 1.0850473769355211e-05, "loss": 0.0, "step": 10164 }, { "epoch": 2.35, "learning_rate": 1.084662198597951e-05, "loss": 0.0, "step": 10165 }, { "epoch": 2.35, "learning_rate": 1.0842770202603806e-05, "loss": 0.0, "step": 10166 }, { "epoch": 2.35, "learning_rate": 1.0838918419228103e-05, "loss": 0.0, "step": 10167 }, { "epoch": 2.35, "learning_rate": 1.08350666358524e-05, "loss": 0.0, "step": 10168 }, { "epoch": 2.35, "learning_rate": 1.0831214852476696e-05, "loss": 0.0, "step": 10169 }, { "epoch": 2.35, "learning_rate": 1.0827363069100995e-05, "loss": 0.0, "step": 10170 }, { "epoch": 2.35, "learning_rate": 1.0823511285725291e-05, "loss": 0.0, "step": 10171 }, { "epoch": 2.35, "learning_rate": 1.0819659502349588e-05, "loss": 0.0, "step": 10172 }, { "epoch": 2.35, "learning_rate": 1.0815807718973886e-05, "loss": 0.0, "step": 10173 }, { "epoch": 2.35, "learning_rate": 1.0811955935598183e-05, "loss": 0.0, "step": 10174 }, { "epoch": 2.35, "learning_rate": 1.080810415222248e-05, "loss": 0.0, "step": 10175 }, { "epoch": 2.35, "learning_rate": 1.0804252368846776e-05, "loss": 0.0, "step": 10176 }, { "epoch": 2.35, "learning_rate": 1.0800400585471073e-05, "loss": 0.0, "step": 10177 }, { "epoch": 2.35, "learning_rate": 1.079654880209537e-05, "loss": 0.0, "step": 10178 }, { "epoch": 2.35, "learning_rate": 1.0792697018719668e-05, "loss": 0.0001, "step": 10179 }, { "epoch": 2.35, "learning_rate": 1.0788845235343965e-05, "loss": 0.0, "step": 10180 }, { "epoch": 2.35, "learning_rate": 1.0784993451968261e-05, "loss": 0.0, "step": 10181 }, { "epoch": 2.35, "learning_rate": 1.078114166859256e-05, "loss": 0.0, "step": 10182 }, { "epoch": 2.35, "learning_rate": 1.0777289885216856e-05, "loss": 0.0, "step": 10183 }, { "epoch": 2.35, "learning_rate": 1.0773438101841153e-05, "loss": 0.0, "step": 10184 }, { "epoch": 2.35, "learning_rate": 1.076958631846545e-05, "loss": 0.0, "step": 10185 }, { "epoch": 2.35, "learning_rate": 1.0765734535089746e-05, "loss": 0.0, "step": 10186 }, { "epoch": 2.35, "learning_rate": 1.0761882751714045e-05, "loss": 0.0, "step": 10187 }, { "epoch": 2.35, "learning_rate": 1.0758030968338341e-05, "loss": 0.0, "step": 10188 }, { "epoch": 2.35, "learning_rate": 1.0754179184962638e-05, "loss": 0.0, "step": 10189 }, { "epoch": 2.35, "learning_rate": 1.0750327401586935e-05, "loss": 0.0, "step": 10190 }, { "epoch": 2.36, "learning_rate": 1.0746475618211233e-05, "loss": 0.0, "step": 10191 }, { "epoch": 2.36, "learning_rate": 1.074262383483553e-05, "loss": 0.0, "step": 10192 }, { "epoch": 2.36, "learning_rate": 1.0738772051459826e-05, "loss": 0.0, "step": 10193 }, { "epoch": 2.36, "learning_rate": 1.0734920268084125e-05, "loss": 0.0, "step": 10194 }, { "epoch": 2.36, "learning_rate": 1.073106848470842e-05, "loss": 0.0, "step": 10195 }, { "epoch": 2.36, "learning_rate": 1.0727216701332718e-05, "loss": 0.0, "step": 10196 }, { "epoch": 2.36, "learning_rate": 1.0723364917957014e-05, "loss": 0.0, "step": 10197 }, { "epoch": 2.36, "learning_rate": 1.0719513134581311e-05, "loss": 0.0, "step": 10198 }, { "epoch": 2.36, "learning_rate": 1.0715661351205608e-05, "loss": 0.0, "step": 10199 }, { "epoch": 2.36, "learning_rate": 1.0711809567829906e-05, "loss": 0.0, "step": 10200 }, { "epoch": 2.36, "learning_rate": 1.0707957784454203e-05, "loss": 0.0, "step": 10201 }, { "epoch": 2.36, "learning_rate": 1.07041060010785e-05, "loss": 0.0, "step": 10202 }, { "epoch": 2.36, "learning_rate": 1.0700254217702798e-05, "loss": 0.0, "step": 10203 }, { "epoch": 2.36, "learning_rate": 1.0696402434327093e-05, "loss": 0.0, "step": 10204 }, { "epoch": 2.36, "learning_rate": 1.0692550650951391e-05, "loss": 0.0, "step": 10205 }, { "epoch": 2.36, "learning_rate": 1.0688698867575688e-05, "loss": 0.0, "step": 10206 }, { "epoch": 2.36, "learning_rate": 1.0684847084199984e-05, "loss": 0.0, "step": 10207 }, { "epoch": 2.36, "learning_rate": 1.0680995300824283e-05, "loss": 0.0, "step": 10208 }, { "epoch": 2.36, "learning_rate": 1.067714351744858e-05, "loss": 0.0, "step": 10209 }, { "epoch": 2.36, "learning_rate": 1.0673291734072876e-05, "loss": 0.0, "step": 10210 }, { "epoch": 2.36, "learning_rate": 1.0669439950697173e-05, "loss": 0.0, "step": 10211 }, { "epoch": 2.36, "learning_rate": 1.0665588167321471e-05, "loss": 0.0, "step": 10212 }, { "epoch": 2.36, "learning_rate": 1.0661736383945768e-05, "loss": 0.0, "step": 10213 }, { "epoch": 2.36, "learning_rate": 1.0657884600570064e-05, "loss": 0.0, "step": 10214 }, { "epoch": 2.36, "learning_rate": 1.0654032817194363e-05, "loss": 0.0, "step": 10215 }, { "epoch": 2.36, "learning_rate": 1.0650181033818658e-05, "loss": 0.0004, "step": 10216 }, { "epoch": 2.36, "learning_rate": 1.0646329250442956e-05, "loss": 0.0, "step": 10217 }, { "epoch": 2.36, "learning_rate": 1.0642477467067253e-05, "loss": 0.0, "step": 10218 }, { "epoch": 2.36, "learning_rate": 1.063862568369155e-05, "loss": 0.0, "step": 10219 }, { "epoch": 2.36, "learning_rate": 1.0634773900315846e-05, "loss": 0.0, "step": 10220 }, { "epoch": 2.36, "learning_rate": 1.0630922116940144e-05, "loss": 0.0, "step": 10221 }, { "epoch": 2.36, "learning_rate": 1.0627070333564441e-05, "loss": 0.0, "step": 10222 }, { "epoch": 2.36, "learning_rate": 1.0623218550188738e-05, "loss": 0.0, "step": 10223 }, { "epoch": 2.36, "learning_rate": 1.0619366766813036e-05, "loss": 0.0, "step": 10224 }, { "epoch": 2.36, "learning_rate": 1.061551498343733e-05, "loss": 0.0, "step": 10225 }, { "epoch": 2.36, "learning_rate": 1.061166320006163e-05, "loss": 0.0, "step": 10226 }, { "epoch": 2.36, "learning_rate": 1.0607811416685926e-05, "loss": 0.0, "step": 10227 }, { "epoch": 2.36, "learning_rate": 1.0603959633310222e-05, "loss": 0.0, "step": 10228 }, { "epoch": 2.36, "learning_rate": 1.060010784993452e-05, "loss": 0.0, "step": 10229 }, { "epoch": 2.36, "learning_rate": 1.0596256066558817e-05, "loss": 0.0, "step": 10230 }, { "epoch": 2.36, "learning_rate": 1.0592404283183114e-05, "loss": 0.0, "step": 10231 }, { "epoch": 2.36, "learning_rate": 1.058855249980741e-05, "loss": 0.0, "step": 10232 }, { "epoch": 2.36, "learning_rate": 1.0584700716431709e-05, "loss": 0.0, "step": 10233 }, { "epoch": 2.37, "learning_rate": 1.0580848933056004e-05, "loss": 0.0, "step": 10234 }, { "epoch": 2.37, "learning_rate": 1.0576997149680302e-05, "loss": 0.0, "step": 10235 }, { "epoch": 2.37, "learning_rate": 1.05731453663046e-05, "loss": 0.0, "step": 10236 }, { "epoch": 2.37, "learning_rate": 1.0569293582928896e-05, "loss": 0.0, "step": 10237 }, { "epoch": 2.37, "learning_rate": 1.0565441799553194e-05, "loss": 0.0, "step": 10238 }, { "epoch": 2.37, "learning_rate": 1.056159001617749e-05, "loss": 0.0015, "step": 10239 }, { "epoch": 2.37, "learning_rate": 1.0557738232801787e-05, "loss": 0.0, "step": 10240 }, { "epoch": 2.37, "learning_rate": 1.0553886449426084e-05, "loss": 0.0, "step": 10241 }, { "epoch": 2.37, "learning_rate": 1.0550034666050382e-05, "loss": 0.0, "step": 10242 }, { "epoch": 2.37, "learning_rate": 1.0546182882674679e-05, "loss": 0.0, "step": 10243 }, { "epoch": 2.37, "learning_rate": 1.0542331099298976e-05, "loss": 0.0, "step": 10244 }, { "epoch": 2.37, "learning_rate": 1.0538479315923274e-05, "loss": 0.0, "step": 10245 }, { "epoch": 2.37, "learning_rate": 1.0534627532547569e-05, "loss": 0.0, "step": 10246 }, { "epoch": 2.37, "learning_rate": 1.0530775749171867e-05, "loss": 0.0, "step": 10247 }, { "epoch": 2.37, "learning_rate": 1.0526923965796164e-05, "loss": 0.0, "step": 10248 }, { "epoch": 2.37, "learning_rate": 1.052307218242046e-05, "loss": 0.0, "step": 10249 }, { "epoch": 2.37, "learning_rate": 1.0519220399044759e-05, "loss": 0.0, "step": 10250 }, { "epoch": 2.37, "learning_rate": 1.0515368615669056e-05, "loss": 0.0, "step": 10251 }, { "epoch": 2.37, "learning_rate": 1.0511516832293352e-05, "loss": 0.0, "step": 10252 }, { "epoch": 2.37, "learning_rate": 1.0507665048917649e-05, "loss": 0.0, "step": 10253 }, { "epoch": 2.37, "learning_rate": 1.0503813265541947e-05, "loss": 0.0, "step": 10254 }, { "epoch": 2.37, "learning_rate": 1.0499961482166242e-05, "loss": 0.0, "step": 10255 }, { "epoch": 2.37, "learning_rate": 1.049610969879054e-05, "loss": 0.0003, "step": 10256 }, { "epoch": 2.37, "learning_rate": 1.0492257915414839e-05, "loss": 0.0, "step": 10257 }, { "epoch": 2.37, "learning_rate": 1.0488406132039134e-05, "loss": 0.0, "step": 10258 }, { "epoch": 2.37, "learning_rate": 1.0484554348663432e-05, "loss": 0.0, "step": 10259 }, { "epoch": 2.37, "learning_rate": 1.0480702565287729e-05, "loss": 0.0, "step": 10260 }, { "epoch": 2.37, "learning_rate": 1.0476850781912025e-05, "loss": 0.0, "step": 10261 }, { "epoch": 2.37, "learning_rate": 1.0472998998536322e-05, "loss": 0.0, "step": 10262 }, { "epoch": 2.37, "learning_rate": 1.046914721516062e-05, "loss": 0.0, "step": 10263 }, { "epoch": 2.37, "learning_rate": 1.0465295431784917e-05, "loss": 0.0, "step": 10264 }, { "epoch": 2.37, "learning_rate": 1.0461443648409214e-05, "loss": 0.0, "step": 10265 }, { "epoch": 2.37, "learning_rate": 1.0457591865033512e-05, "loss": 0.0, "step": 10266 }, { "epoch": 2.37, "learning_rate": 1.0453740081657807e-05, "loss": 0.0, "step": 10267 }, { "epoch": 2.37, "learning_rate": 1.0449888298282105e-05, "loss": 0.0, "step": 10268 }, { "epoch": 2.37, "learning_rate": 1.0446036514906402e-05, "loss": 0.0, "step": 10269 }, { "epoch": 2.37, "learning_rate": 1.0442184731530699e-05, "loss": 0.0, "step": 10270 }, { "epoch": 2.37, "learning_rate": 1.0438332948154997e-05, "loss": 0.0, "step": 10271 }, { "epoch": 2.37, "learning_rate": 1.0434481164779294e-05, "loss": 0.0, "step": 10272 }, { "epoch": 2.37, "learning_rate": 1.043062938140359e-05, "loss": 0.0, "step": 10273 }, { "epoch": 2.37, "learning_rate": 1.0426777598027887e-05, "loss": 0.0, "step": 10274 }, { "epoch": 2.37, "learning_rate": 1.0422925814652185e-05, "loss": 0.0, "step": 10275 }, { "epoch": 2.37, "learning_rate": 1.041907403127648e-05, "loss": 0.0, "step": 10276 }, { "epoch": 2.38, "learning_rate": 1.0415222247900779e-05, "loss": 0.0, "step": 10277 }, { "epoch": 2.38, "learning_rate": 1.0411370464525077e-05, "loss": 0.0, "step": 10278 }, { "epoch": 2.38, "learning_rate": 1.0407518681149372e-05, "loss": 0.0, "step": 10279 }, { "epoch": 2.38, "learning_rate": 1.040366689777367e-05, "loss": 0.0, "step": 10280 }, { "epoch": 2.38, "learning_rate": 1.0399815114397967e-05, "loss": 0.0, "step": 10281 }, { "epoch": 2.38, "learning_rate": 1.0395963331022264e-05, "loss": 0.0, "step": 10282 }, { "epoch": 2.38, "learning_rate": 1.039211154764656e-05, "loss": 0.0, "step": 10283 }, { "epoch": 2.38, "learning_rate": 1.0388259764270859e-05, "loss": 0.0, "step": 10284 }, { "epoch": 2.38, "learning_rate": 1.0384407980895155e-05, "loss": 0.0, "step": 10285 }, { "epoch": 2.38, "learning_rate": 1.0380556197519452e-05, "loss": 0.0, "step": 10286 }, { "epoch": 2.38, "learning_rate": 1.037670441414375e-05, "loss": 0.0, "step": 10287 }, { "epoch": 2.38, "learning_rate": 1.0372852630768045e-05, "loss": 0.0, "step": 10288 }, { "epoch": 2.38, "learning_rate": 1.0369000847392343e-05, "loss": 0.0, "step": 10289 }, { "epoch": 2.38, "learning_rate": 1.036514906401664e-05, "loss": 0.0026, "step": 10290 }, { "epoch": 2.38, "learning_rate": 1.0361297280640937e-05, "loss": 0.0, "step": 10291 }, { "epoch": 2.38, "learning_rate": 1.0357445497265235e-05, "loss": 0.0, "step": 10292 }, { "epoch": 2.38, "learning_rate": 1.0353593713889532e-05, "loss": 0.0001, "step": 10293 }, { "epoch": 2.38, "learning_rate": 1.0349741930513828e-05, "loss": 0.0, "step": 10294 }, { "epoch": 2.38, "learning_rate": 1.0345890147138125e-05, "loss": 0.0, "step": 10295 }, { "epoch": 2.38, "learning_rate": 1.0342038363762423e-05, "loss": 0.0, "step": 10296 }, { "epoch": 2.38, "learning_rate": 1.0338186580386718e-05, "loss": 0.0, "step": 10297 }, { "epoch": 2.38, "learning_rate": 1.0334334797011017e-05, "loss": 0.0, "step": 10298 }, { "epoch": 2.38, "learning_rate": 1.0330483013635315e-05, "loss": 0.0, "step": 10299 }, { "epoch": 2.38, "learning_rate": 1.032663123025961e-05, "loss": 0.0, "step": 10300 }, { "epoch": 2.38, "learning_rate": 1.0322779446883908e-05, "loss": 0.0, "step": 10301 }, { "epoch": 2.38, "learning_rate": 1.0318927663508205e-05, "loss": 0.0, "step": 10302 }, { "epoch": 2.38, "learning_rate": 1.0315075880132502e-05, "loss": 0.0, "step": 10303 }, { "epoch": 2.38, "learning_rate": 1.0311224096756798e-05, "loss": 0.0, "step": 10304 }, { "epoch": 2.38, "learning_rate": 1.0307372313381097e-05, "loss": 0.0001, "step": 10305 }, { "epoch": 2.38, "learning_rate": 1.0303520530005393e-05, "loss": 0.0, "step": 10306 }, { "epoch": 2.38, "learning_rate": 1.029966874662969e-05, "loss": 0.0, "step": 10307 }, { "epoch": 2.38, "learning_rate": 1.0295816963253988e-05, "loss": 0.0, "step": 10308 }, { "epoch": 2.38, "learning_rate": 1.0291965179878283e-05, "loss": 0.0, "step": 10309 }, { "epoch": 2.38, "learning_rate": 1.0288113396502582e-05, "loss": 0.0, "step": 10310 }, { "epoch": 2.38, "learning_rate": 1.0284261613126878e-05, "loss": 0.0, "step": 10311 }, { "epoch": 2.38, "learning_rate": 1.0280409829751175e-05, "loss": 0.0, "step": 10312 }, { "epoch": 2.38, "learning_rate": 1.0276558046375473e-05, "loss": 0.0, "step": 10313 }, { "epoch": 2.38, "learning_rate": 1.027270626299977e-05, "loss": 0.0, "step": 10314 }, { "epoch": 2.38, "learning_rate": 1.0268854479624067e-05, "loss": 0.0, "step": 10315 }, { "epoch": 2.38, "learning_rate": 1.0265002696248363e-05, "loss": 0.0, "step": 10316 }, { "epoch": 2.38, "learning_rate": 1.0261150912872662e-05, "loss": 0.0, "step": 10317 }, { "epoch": 2.38, "learning_rate": 1.0257299129496957e-05, "loss": 0.0, "step": 10318 }, { "epoch": 2.38, "learning_rate": 1.0253447346121255e-05, "loss": 0.0, "step": 10319 }, { "epoch": 2.39, "learning_rate": 1.0249595562745551e-05, "loss": 0.0, "step": 10320 }, { "epoch": 2.39, "learning_rate": 1.0245743779369848e-05, "loss": 0.0, "step": 10321 }, { "epoch": 2.39, "learning_rate": 1.0241891995994146e-05, "loss": 0.0, "step": 10322 }, { "epoch": 2.39, "learning_rate": 1.0238040212618443e-05, "loss": 0.0, "step": 10323 }, { "epoch": 2.39, "learning_rate": 1.023418842924274e-05, "loss": 0.0, "step": 10324 }, { "epoch": 2.39, "learning_rate": 1.0230336645867036e-05, "loss": 0.0, "step": 10325 }, { "epoch": 2.39, "learning_rate": 1.0226484862491335e-05, "loss": 0.0, "step": 10326 }, { "epoch": 2.39, "learning_rate": 1.0222633079115631e-05, "loss": 0.0, "step": 10327 }, { "epoch": 2.39, "learning_rate": 1.0218781295739928e-05, "loss": 0.0, "step": 10328 }, { "epoch": 2.39, "learning_rate": 1.0214929512364225e-05, "loss": 0.0, "step": 10329 }, { "epoch": 2.39, "learning_rate": 1.0211077728988521e-05, "loss": 0.0, "step": 10330 }, { "epoch": 2.39, "learning_rate": 1.020722594561282e-05, "loss": 0.0, "step": 10331 }, { "epoch": 2.39, "learning_rate": 1.0203374162237116e-05, "loss": 0.0, "step": 10332 }, { "epoch": 2.39, "learning_rate": 1.0199522378861413e-05, "loss": 0.0, "step": 10333 }, { "epoch": 2.39, "learning_rate": 1.0195670595485711e-05, "loss": 0.0, "step": 10334 }, { "epoch": 2.39, "learning_rate": 1.0191818812110008e-05, "loss": 0.0, "step": 10335 }, { "epoch": 2.39, "learning_rate": 1.0187967028734305e-05, "loss": 0.0, "step": 10336 }, { "epoch": 2.39, "learning_rate": 1.0184115245358601e-05, "loss": 0.0, "step": 10337 }, { "epoch": 2.39, "learning_rate": 1.0180263461982898e-05, "loss": 0.0, "step": 10338 }, { "epoch": 2.39, "learning_rate": 1.0176411678607195e-05, "loss": 0.0, "step": 10339 }, { "epoch": 2.39, "learning_rate": 1.0172559895231493e-05, "loss": 0.0, "step": 10340 }, { "epoch": 2.39, "learning_rate": 1.016870811185579e-05, "loss": 0.0, "step": 10341 }, { "epoch": 2.39, "learning_rate": 1.0164856328480086e-05, "loss": 0.0, "step": 10342 }, { "epoch": 2.39, "learning_rate": 1.0161004545104385e-05, "loss": 0.0, "step": 10343 }, { "epoch": 2.39, "learning_rate": 1.0157152761728681e-05, "loss": 0.0, "step": 10344 }, { "epoch": 2.39, "learning_rate": 1.0153300978352978e-05, "loss": 0.0, "step": 10345 }, { "epoch": 2.39, "learning_rate": 1.0149449194977275e-05, "loss": 0.0, "step": 10346 }, { "epoch": 2.39, "learning_rate": 1.0145597411601571e-05, "loss": 0.0, "step": 10347 }, { "epoch": 2.39, "learning_rate": 1.014174562822587e-05, "loss": 0.0, "step": 10348 }, { "epoch": 2.39, "learning_rate": 1.0137893844850166e-05, "loss": 0.0, "step": 10349 }, { "epoch": 2.39, "learning_rate": 1.0134042061474463e-05, "loss": 0.0, "step": 10350 }, { "epoch": 2.39, "learning_rate": 1.013019027809876e-05, "loss": 0.0, "step": 10351 }, { "epoch": 2.39, "learning_rate": 1.0126338494723058e-05, "loss": 0.0, "step": 10352 }, { "epoch": 2.39, "learning_rate": 1.0122486711347354e-05, "loss": 0.0, "step": 10353 }, { "epoch": 2.39, "learning_rate": 1.0118634927971651e-05, "loss": 0.0, "step": 10354 }, { "epoch": 2.39, "learning_rate": 1.011478314459595e-05, "loss": 0.0, "step": 10355 }, { "epoch": 2.39, "learning_rate": 1.0110931361220244e-05, "loss": 0.0, "step": 10356 }, { "epoch": 2.39, "learning_rate": 1.0107079577844543e-05, "loss": 0.0, "step": 10357 }, { "epoch": 2.39, "learning_rate": 1.010322779446884e-05, "loss": 0.0, "step": 10358 }, { "epoch": 2.39, "learning_rate": 1.0099376011093136e-05, "loss": 0.0, "step": 10359 }, { "epoch": 2.39, "learning_rate": 1.0095524227717433e-05, "loss": 0.0001, "step": 10360 }, { "epoch": 2.39, "learning_rate": 1.0091672444341731e-05, "loss": 0.0, "step": 10361 }, { "epoch": 2.39, "learning_rate": 1.0087820660966028e-05, "loss": 0.0, "step": 10362 }, { "epoch": 2.39, "learning_rate": 1.0083968877590324e-05, "loss": 0.0, "step": 10363 }, { "epoch": 2.4, "learning_rate": 1.0080117094214623e-05, "loss": 0.0, "step": 10364 }, { "epoch": 2.4, "learning_rate": 1.0076265310838918e-05, "loss": 0.0, "step": 10365 }, { "epoch": 2.4, "learning_rate": 1.0072413527463216e-05, "loss": 0.0, "step": 10366 }, { "epoch": 2.4, "learning_rate": 1.0068561744087513e-05, "loss": 0.0, "step": 10367 }, { "epoch": 2.4, "learning_rate": 1.006470996071181e-05, "loss": 0.0, "step": 10368 }, { "epoch": 2.4, "learning_rate": 1.0060858177336108e-05, "loss": 0.0, "step": 10369 }, { "epoch": 2.4, "learning_rate": 1.0057006393960404e-05, "loss": 0.0, "step": 10370 }, { "epoch": 2.4, "learning_rate": 1.0053154610584701e-05, "loss": 0.0, "step": 10371 }, { "epoch": 2.4, "learning_rate": 1.0049302827208998e-05, "loss": 0.0, "step": 10372 }, { "epoch": 2.4, "learning_rate": 1.0045451043833296e-05, "loss": 0.0, "step": 10373 }, { "epoch": 2.4, "learning_rate": 1.0041599260457593e-05, "loss": 0.0, "step": 10374 }, { "epoch": 2.4, "learning_rate": 1.003774747708189e-05, "loss": 0.0, "step": 10375 }, { "epoch": 2.4, "learning_rate": 1.0033895693706188e-05, "loss": 0.0, "step": 10376 }, { "epoch": 2.4, "learning_rate": 1.0030043910330483e-05, "loss": 0.0, "step": 10377 }, { "epoch": 2.4, "learning_rate": 1.0026192126954781e-05, "loss": 0.0, "step": 10378 }, { "epoch": 2.4, "learning_rate": 1.0022340343579078e-05, "loss": 0.0, "step": 10379 }, { "epoch": 2.4, "learning_rate": 1.0018488560203374e-05, "loss": 0.0, "step": 10380 }, { "epoch": 2.4, "learning_rate": 1.001463677682767e-05, "loss": 0.0, "step": 10381 }, { "epoch": 2.4, "learning_rate": 1.001078499345197e-05, "loss": 0.0, "step": 10382 }, { "epoch": 2.4, "learning_rate": 1.0006933210076266e-05, "loss": 0.0, "step": 10383 }, { "epoch": 2.4, "learning_rate": 1.0003081426700562e-05, "loss": 0.0, "step": 10384 }, { "epoch": 2.4, "learning_rate": 9.99922964332486e-06, "loss": 0.0, "step": 10385 }, { "epoch": 2.4, "learning_rate": 9.995377859949156e-06, "loss": 0.0, "step": 10386 }, { "epoch": 2.4, "learning_rate": 9.991526076573454e-06, "loss": 0.0, "step": 10387 }, { "epoch": 2.4, "learning_rate": 9.98767429319775e-06, "loss": 0.0, "step": 10388 }, { "epoch": 2.4, "learning_rate": 9.983822509822047e-06, "loss": 0.0, "step": 10389 }, { "epoch": 2.4, "learning_rate": 9.979970726446346e-06, "loss": 0.0, "step": 10390 }, { "epoch": 2.4, "learning_rate": 9.976118943070642e-06, "loss": 0.0, "step": 10391 }, { "epoch": 2.4, "learning_rate": 9.972267159694939e-06, "loss": 0.0, "step": 10392 }, { "epoch": 2.4, "learning_rate": 9.968415376319236e-06, "loss": 0.0001, "step": 10393 }, { "epoch": 2.4, "learning_rate": 9.964563592943534e-06, "loss": 0.0, "step": 10394 }, { "epoch": 2.4, "learning_rate": 9.960711809567829e-06, "loss": 0.0, "step": 10395 }, { "epoch": 2.4, "learning_rate": 9.956860026192127e-06, "loss": 0.0, "step": 10396 }, { "epoch": 2.4, "learning_rate": 9.953008242816426e-06, "loss": 0.0, "step": 10397 }, { "epoch": 2.4, "learning_rate": 9.94915645944072e-06, "loss": 0.0, "step": 10398 }, { "epoch": 2.4, "learning_rate": 9.945304676065019e-06, "loss": 0.0, "step": 10399 }, { "epoch": 2.4, "learning_rate": 9.941452892689316e-06, "loss": 0.0, "step": 10400 }, { "epoch": 2.4, "learning_rate": 9.937601109313612e-06, "loss": 0.0, "step": 10401 }, { "epoch": 2.4, "learning_rate": 9.933749325937909e-06, "loss": 0.0, "step": 10402 }, { "epoch": 2.4, "learning_rate": 9.929897542562207e-06, "loss": 0.0, "step": 10403 }, { "epoch": 2.4, "learning_rate": 9.926045759186504e-06, "loss": 0.0, "step": 10404 }, { "epoch": 2.4, "learning_rate": 9.9221939758108e-06, "loss": 0.0, "step": 10405 }, { "epoch": 2.4, "learning_rate": 9.918342192435099e-06, "loss": 0.0, "step": 10406 }, { "epoch": 2.41, "learning_rate": 9.914490409059394e-06, "loss": 0.0, "step": 10407 }, { "epoch": 2.41, "learning_rate": 9.910638625683692e-06, "loss": 0.0, "step": 10408 }, { "epoch": 2.41, "learning_rate": 9.906786842307989e-06, "loss": 0.0, "step": 10409 }, { "epoch": 2.41, "learning_rate": 9.902935058932286e-06, "loss": 0.0, "step": 10410 }, { "epoch": 2.41, "learning_rate": 9.899083275556584e-06, "loss": 0.0, "step": 10411 }, { "epoch": 2.41, "learning_rate": 9.89523149218088e-06, "loss": 0.0, "step": 10412 }, { "epoch": 2.41, "learning_rate": 9.891379708805177e-06, "loss": 0.0, "step": 10413 }, { "epoch": 2.41, "learning_rate": 9.887527925429474e-06, "loss": 0.0001, "step": 10414 }, { "epoch": 2.41, "learning_rate": 9.883676142053772e-06, "loss": 0.0, "step": 10415 }, { "epoch": 2.41, "learning_rate": 9.879824358678067e-06, "loss": 0.0, "step": 10416 }, { "epoch": 2.41, "learning_rate": 9.875972575302365e-06, "loss": 0.0, "step": 10417 }, { "epoch": 2.41, "learning_rate": 9.872120791926664e-06, "loss": 0.0, "step": 10418 }, { "epoch": 2.41, "learning_rate": 9.868269008550959e-06, "loss": 0.0, "step": 10419 }, { "epoch": 2.41, "learning_rate": 9.864417225175257e-06, "loss": 0.0, "step": 10420 }, { "epoch": 2.41, "learning_rate": 9.860565441799554e-06, "loss": 0.0, "step": 10421 }, { "epoch": 2.41, "learning_rate": 9.85671365842385e-06, "loss": 0.0, "step": 10422 }, { "epoch": 2.41, "learning_rate": 9.852861875048147e-06, "loss": 0.0, "step": 10423 }, { "epoch": 2.41, "learning_rate": 9.849010091672445e-06, "loss": 0.0, "step": 10424 }, { "epoch": 2.41, "learning_rate": 9.845158308296742e-06, "loss": 0.0, "step": 10425 }, { "epoch": 2.41, "learning_rate": 9.841306524921039e-06, "loss": 0.0, "step": 10426 }, { "epoch": 2.41, "learning_rate": 9.837454741545337e-06, "loss": 0.0, "step": 10427 }, { "epoch": 2.41, "learning_rate": 9.833602958169632e-06, "loss": 0.0, "step": 10428 }, { "epoch": 2.41, "learning_rate": 9.82975117479393e-06, "loss": 0.0, "step": 10429 }, { "epoch": 2.41, "learning_rate": 9.825899391418227e-06, "loss": 0.0, "step": 10430 }, { "epoch": 2.41, "learning_rate": 9.822047608042524e-06, "loss": 0.0, "step": 10431 }, { "epoch": 2.41, "learning_rate": 9.818195824666822e-06, "loss": 0.0, "step": 10432 }, { "epoch": 2.41, "learning_rate": 9.814344041291119e-06, "loss": 0.0, "step": 10433 }, { "epoch": 2.41, "learning_rate": 9.810492257915415e-06, "loss": 0.0, "step": 10434 }, { "epoch": 2.41, "learning_rate": 9.806640474539712e-06, "loss": 0.0, "step": 10435 }, { "epoch": 2.41, "learning_rate": 9.80278869116401e-06, "loss": 0.0, "step": 10436 }, { "epoch": 2.41, "learning_rate": 9.798936907788305e-06, "loss": 0.0, "step": 10437 }, { "epoch": 2.41, "learning_rate": 9.795085124412604e-06, "loss": 0.0, "step": 10438 }, { "epoch": 2.41, "learning_rate": 9.791233341036902e-06, "loss": 0.0, "step": 10439 }, { "epoch": 2.41, "learning_rate": 9.787381557661197e-06, "loss": 0.0, "step": 10440 }, { "epoch": 2.41, "learning_rate": 9.783529774285495e-06, "loss": 0.0, "step": 10441 }, { "epoch": 2.41, "learning_rate": 9.779677990909792e-06, "loss": 0.0082, "step": 10442 }, { "epoch": 2.41, "learning_rate": 9.775826207534089e-06, "loss": 0.0, "step": 10443 }, { "epoch": 2.41, "learning_rate": 9.771974424158385e-06, "loss": 0.0, "step": 10444 }, { "epoch": 2.41, "learning_rate": 9.768122640782684e-06, "loss": 0.0, "step": 10445 }, { "epoch": 2.41, "learning_rate": 9.76427085740698e-06, "loss": 0.0, "step": 10446 }, { "epoch": 2.41, "learning_rate": 9.760419074031277e-06, "loss": 0.0, "step": 10447 }, { "epoch": 2.41, "learning_rate": 9.756567290655575e-06, "loss": 0.0, "step": 10448 }, { "epoch": 2.41, "learning_rate": 9.75271550727987e-06, "loss": 0.0, "step": 10449 }, { "epoch": 2.42, "learning_rate": 9.748863723904168e-06, "loss": 0.0, "step": 10450 }, { "epoch": 2.42, "learning_rate": 9.745011940528465e-06, "loss": 0.0, "step": 10451 }, { "epoch": 2.42, "learning_rate": 9.741160157152762e-06, "loss": 0.0, "step": 10452 }, { "epoch": 2.42, "learning_rate": 9.73730837377706e-06, "loss": 0.0, "step": 10453 }, { "epoch": 2.42, "learning_rate": 9.733456590401357e-06, "loss": 0.0, "step": 10454 }, { "epoch": 2.42, "learning_rate": 9.729604807025653e-06, "loss": 0.0, "step": 10455 }, { "epoch": 2.42, "learning_rate": 9.72575302364995e-06, "loss": 0.0, "step": 10456 }, { "epoch": 2.42, "learning_rate": 9.721901240274248e-06, "loss": 0.0, "step": 10457 }, { "epoch": 2.42, "learning_rate": 9.718049456898543e-06, "loss": 0.0, "step": 10458 }, { "epoch": 2.42, "learning_rate": 9.714197673522842e-06, "loss": 0.0, "step": 10459 }, { "epoch": 2.42, "learning_rate": 9.710345890147138e-06, "loss": 0.0, "step": 10460 }, { "epoch": 2.42, "learning_rate": 9.706494106771435e-06, "loss": 0.0, "step": 10461 }, { "epoch": 2.42, "learning_rate": 9.702642323395733e-06, "loss": 0.0, "step": 10462 }, { "epoch": 2.42, "learning_rate": 9.69879054002003e-06, "loss": 0.0, "step": 10463 }, { "epoch": 2.42, "learning_rate": 9.694938756644327e-06, "loss": 0.0, "step": 10464 }, { "epoch": 2.42, "learning_rate": 9.691086973268623e-06, "loss": 0.0, "step": 10465 }, { "epoch": 2.42, "learning_rate": 9.687235189892922e-06, "loss": 0.0001, "step": 10466 }, { "epoch": 2.42, "learning_rate": 9.683383406517218e-06, "loss": 0.0, "step": 10467 }, { "epoch": 2.42, "learning_rate": 9.679531623141515e-06, "loss": 0.0, "step": 10468 }, { "epoch": 2.42, "learning_rate": 9.675679839765813e-06, "loss": 0.0, "step": 10469 }, { "epoch": 2.42, "learning_rate": 9.671828056390108e-06, "loss": 0.0, "step": 10470 }, { "epoch": 2.42, "learning_rate": 9.667976273014407e-06, "loss": 0.0, "step": 10471 }, { "epoch": 2.42, "learning_rate": 9.664124489638703e-06, "loss": 0.0, "step": 10472 }, { "epoch": 2.42, "learning_rate": 9.660272706263e-06, "loss": 0.0, "step": 10473 }, { "epoch": 2.42, "learning_rate": 9.656420922887298e-06, "loss": 0.0, "step": 10474 }, { "epoch": 2.42, "learning_rate": 9.652569139511595e-06, "loss": 0.0, "step": 10475 }, { "epoch": 2.42, "learning_rate": 9.648717356135892e-06, "loss": 0.0, "step": 10476 }, { "epoch": 2.42, "learning_rate": 9.644865572760188e-06, "loss": 0.0, "step": 10477 }, { "epoch": 2.42, "learning_rate": 9.641013789384487e-06, "loss": 0.0, "step": 10478 }, { "epoch": 2.42, "learning_rate": 9.637162006008781e-06, "loss": 0.0, "step": 10479 }, { "epoch": 2.42, "learning_rate": 9.63331022263308e-06, "loss": 0.0, "step": 10480 }, { "epoch": 2.42, "learning_rate": 9.629458439257376e-06, "loss": 0.0, "step": 10481 }, { "epoch": 2.42, "learning_rate": 9.625606655881673e-06, "loss": 0.0, "step": 10482 }, { "epoch": 2.42, "learning_rate": 9.621754872505971e-06, "loss": 0.0, "step": 10483 }, { "epoch": 2.42, "learning_rate": 9.617903089130268e-06, "loss": 0.0, "step": 10484 }, { "epoch": 2.42, "learning_rate": 9.614051305754565e-06, "loss": 0.0, "step": 10485 }, { "epoch": 2.42, "learning_rate": 9.610199522378861e-06, "loss": 0.0, "step": 10486 }, { "epoch": 2.42, "learning_rate": 9.60634773900316e-06, "loss": 0.0, "step": 10487 }, { "epoch": 2.42, "learning_rate": 9.602495955627456e-06, "loss": 0.0, "step": 10488 }, { "epoch": 2.42, "learning_rate": 9.598644172251753e-06, "loss": 0.0, "step": 10489 }, { "epoch": 2.42, "learning_rate": 9.59479238887605e-06, "loss": 0.0, "step": 10490 }, { "epoch": 2.42, "learning_rate": 9.590940605500346e-06, "loss": 0.0, "step": 10491 }, { "epoch": 2.42, "learning_rate": 9.587088822124645e-06, "loss": 0.0, "step": 10492 }, { "epoch": 2.43, "learning_rate": 9.583237038748941e-06, "loss": 0.0, "step": 10493 }, { "epoch": 2.43, "learning_rate": 9.579385255373238e-06, "loss": 0.0, "step": 10494 }, { "epoch": 2.43, "learning_rate": 9.575533471997536e-06, "loss": 0.0, "step": 10495 }, { "epoch": 2.43, "learning_rate": 9.571681688621833e-06, "loss": 0.0, "step": 10496 }, { "epoch": 2.43, "learning_rate": 9.56782990524613e-06, "loss": 0.0, "step": 10497 }, { "epoch": 2.43, "learning_rate": 9.563978121870426e-06, "loss": 0.0, "step": 10498 }, { "epoch": 2.43, "learning_rate": 9.560126338494723e-06, "loss": 0.0, "step": 10499 }, { "epoch": 2.43, "learning_rate": 9.55627455511902e-06, "loss": 0.0, "step": 10500 }, { "epoch": 2.43, "learning_rate": 9.552422771743318e-06, "loss": 0.0, "step": 10501 }, { "epoch": 2.43, "learning_rate": 9.548570988367615e-06, "loss": 0.0, "step": 10502 }, { "epoch": 2.43, "learning_rate": 9.544719204991911e-06, "loss": 0.0, "step": 10503 }, { "epoch": 2.43, "learning_rate": 9.54086742161621e-06, "loss": 0.0, "step": 10504 }, { "epoch": 2.43, "learning_rate": 9.537015638240506e-06, "loss": 0.0, "step": 10505 }, { "epoch": 2.43, "learning_rate": 9.533163854864803e-06, "loss": 0.0, "step": 10506 }, { "epoch": 2.43, "learning_rate": 9.5293120714891e-06, "loss": 0.0, "step": 10507 }, { "epoch": 2.43, "learning_rate": 9.525460288113396e-06, "loss": 0.0, "step": 10508 }, { "epoch": 2.43, "learning_rate": 9.521608504737695e-06, "loss": 0.0, "step": 10509 }, { "epoch": 2.43, "learning_rate": 9.517756721361991e-06, "loss": 0.0, "step": 10510 }, { "epoch": 2.43, "learning_rate": 9.513904937986288e-06, "loss": 0.0, "step": 10511 }, { "epoch": 2.43, "learning_rate": 9.510053154610584e-06, "loss": 0.0, "step": 10512 }, { "epoch": 2.43, "learning_rate": 9.506201371234883e-06, "loss": 0.0, "step": 10513 }, { "epoch": 2.43, "learning_rate": 9.50234958785918e-06, "loss": 0.0, "step": 10514 }, { "epoch": 2.43, "learning_rate": 9.498497804483476e-06, "loss": 0.0, "step": 10515 }, { "epoch": 2.43, "learning_rate": 9.494646021107774e-06, "loss": 0.0, "step": 10516 }, { "epoch": 2.43, "learning_rate": 9.49079423773207e-06, "loss": 0.0, "step": 10517 }, { "epoch": 2.43, "learning_rate": 9.486942454356368e-06, "loss": 0.0, "step": 10518 }, { "epoch": 2.43, "learning_rate": 9.483090670980664e-06, "loss": 0.0, "step": 10519 }, { "epoch": 2.43, "learning_rate": 9.479238887604961e-06, "loss": 0.0, "step": 10520 }, { "epoch": 2.43, "learning_rate": 9.475387104229258e-06, "loss": 0.0, "step": 10521 }, { "epoch": 2.43, "learning_rate": 9.471535320853556e-06, "loss": 0.0, "step": 10522 }, { "epoch": 2.43, "learning_rate": 9.467683537477853e-06, "loss": 0.0, "step": 10523 }, { "epoch": 2.43, "learning_rate": 9.46383175410215e-06, "loss": 0.0, "step": 10524 }, { "epoch": 2.43, "learning_rate": 9.459979970726448e-06, "loss": 0.0, "step": 10525 }, { "epoch": 2.43, "learning_rate": 9.456128187350743e-06, "loss": 0.0, "step": 10526 }, { "epoch": 2.43, "learning_rate": 9.452276403975041e-06, "loss": 0.0, "step": 10527 }, { "epoch": 2.43, "learning_rate": 9.448424620599338e-06, "loss": 0.0, "step": 10528 }, { "epoch": 2.43, "learning_rate": 9.444572837223634e-06, "loss": 0.0, "step": 10529 }, { "epoch": 2.43, "learning_rate": 9.440721053847933e-06, "loss": 0.0, "step": 10530 }, { "epoch": 2.43, "learning_rate": 9.43686927047223e-06, "loss": 0.0, "step": 10531 }, { "epoch": 2.43, "learning_rate": 9.433017487096526e-06, "loss": 0.0, "step": 10532 }, { "epoch": 2.43, "learning_rate": 9.429165703720823e-06, "loss": 0.0, "step": 10533 }, { "epoch": 2.43, "learning_rate": 9.425313920345121e-06, "loss": 0.0, "step": 10534 }, { "epoch": 2.43, "learning_rate": 9.421462136969416e-06, "loss": 0.0, "step": 10535 }, { "epoch": 2.43, "learning_rate": 9.417610353593714e-06, "loss": 0.0, "step": 10536 }, { "epoch": 2.44, "learning_rate": 9.413758570218013e-06, "loss": 0.0, "step": 10537 }, { "epoch": 2.44, "learning_rate": 9.409906786842308e-06, "loss": 0.0, "step": 10538 }, { "epoch": 2.44, "learning_rate": 9.406055003466606e-06, "loss": 0.0002, "step": 10539 }, { "epoch": 2.44, "learning_rate": 9.402203220090903e-06, "loss": 0.0, "step": 10540 }, { "epoch": 2.44, "learning_rate": 9.398351436715199e-06, "loss": 0.0, "step": 10541 }, { "epoch": 2.44, "learning_rate": 9.394499653339496e-06, "loss": 0.0, "step": 10542 }, { "epoch": 2.44, "learning_rate": 9.390647869963794e-06, "loss": 0.0, "step": 10543 }, { "epoch": 2.44, "learning_rate": 9.38679608658809e-06, "loss": 0.0, "step": 10544 }, { "epoch": 2.44, "learning_rate": 9.382944303212387e-06, "loss": 0.0, "step": 10545 }, { "epoch": 2.44, "learning_rate": 9.379092519836686e-06, "loss": 0.0, "step": 10546 }, { "epoch": 2.44, "learning_rate": 9.37524073646098e-06, "loss": 0.0, "step": 10547 }, { "epoch": 2.44, "learning_rate": 9.371388953085279e-06, "loss": 0.0, "step": 10548 }, { "epoch": 2.44, "learning_rate": 9.367537169709576e-06, "loss": 0.0, "step": 10549 }, { "epoch": 2.44, "learning_rate": 9.363685386333872e-06, "loss": 0.0, "step": 10550 }, { "epoch": 2.44, "learning_rate": 9.35983360295817e-06, "loss": 0.0, "step": 10551 }, { "epoch": 2.44, "learning_rate": 9.355981819582467e-06, "loss": 0.0, "step": 10552 }, { "epoch": 2.44, "learning_rate": 9.352130036206764e-06, "loss": 0.0, "step": 10553 }, { "epoch": 2.44, "learning_rate": 9.34827825283106e-06, "loss": 0.0, "step": 10554 }, { "epoch": 2.44, "learning_rate": 9.344426469455359e-06, "loss": 0.0, "step": 10555 }, { "epoch": 2.44, "learning_rate": 9.340574686079654e-06, "loss": 0.0, "step": 10556 }, { "epoch": 2.44, "learning_rate": 9.336722902703952e-06, "loss": 0.0, "step": 10557 }, { "epoch": 2.44, "learning_rate": 9.33287111932825e-06, "loss": 0.0, "step": 10558 }, { "epoch": 2.44, "learning_rate": 9.329019335952546e-06, "loss": 0.0, "step": 10559 }, { "epoch": 2.44, "learning_rate": 9.325167552576844e-06, "loss": 0.0, "step": 10560 }, { "epoch": 2.44, "learning_rate": 9.32131576920114e-06, "loss": 0.0, "step": 10561 }, { "epoch": 2.44, "learning_rate": 9.317463985825437e-06, "loss": 0.0, "step": 10562 }, { "epoch": 2.44, "learning_rate": 9.313612202449734e-06, "loss": 0.0, "step": 10563 }, { "epoch": 2.44, "learning_rate": 9.309760419074032e-06, "loss": 0.0, "step": 10564 }, { "epoch": 2.44, "learning_rate": 9.305908635698329e-06, "loss": 0.0, "step": 10565 }, { "epoch": 2.44, "learning_rate": 9.302056852322626e-06, "loss": 0.0, "step": 10566 }, { "epoch": 2.44, "learning_rate": 9.298205068946924e-06, "loss": 0.0, "step": 10567 }, { "epoch": 2.44, "learning_rate": 9.294353285571219e-06, "loss": 0.0001, "step": 10568 }, { "epoch": 2.44, "learning_rate": 9.290501502195517e-06, "loss": 0.0, "step": 10569 }, { "epoch": 2.44, "learning_rate": 9.286649718819814e-06, "loss": 0.0, "step": 10570 }, { "epoch": 2.44, "learning_rate": 9.28279793544411e-06, "loss": 0.0, "step": 10571 }, { "epoch": 2.44, "learning_rate": 9.278946152068409e-06, "loss": 0.0, "step": 10572 }, { "epoch": 2.44, "learning_rate": 9.275094368692705e-06, "loss": 0.0, "step": 10573 }, { "epoch": 2.44, "learning_rate": 9.271242585317002e-06, "loss": 0.0, "step": 10574 }, { "epoch": 2.44, "learning_rate": 9.267390801941299e-06, "loss": 0.0, "step": 10575 }, { "epoch": 2.44, "learning_rate": 9.263539018565597e-06, "loss": 0.0, "step": 10576 }, { "epoch": 2.44, "learning_rate": 9.259687235189892e-06, "loss": 0.0, "step": 10577 }, { "epoch": 2.44, "learning_rate": 9.25583545181419e-06, "loss": 0.0, "step": 10578 }, { "epoch": 2.44, "learning_rate": 9.251983668438489e-06, "loss": 0.0, "step": 10579 }, { "epoch": 2.45, "learning_rate": 9.248131885062784e-06, "loss": 0.0, "step": 10580 }, { "epoch": 2.45, "learning_rate": 9.244280101687082e-06, "loss": 0.0, "step": 10581 }, { "epoch": 2.45, "learning_rate": 9.240428318311379e-06, "loss": 0.0, "step": 10582 }, { "epoch": 2.45, "learning_rate": 9.236576534935675e-06, "loss": 0.0, "step": 10583 }, { "epoch": 2.45, "learning_rate": 9.232724751559972e-06, "loss": 0.0, "step": 10584 }, { "epoch": 2.45, "learning_rate": 9.22887296818427e-06, "loss": 0.0, "step": 10585 }, { "epoch": 2.45, "learning_rate": 9.225021184808567e-06, "loss": 0.0, "step": 10586 }, { "epoch": 2.45, "learning_rate": 9.221169401432864e-06, "loss": 0.0, "step": 10587 }, { "epoch": 2.45, "learning_rate": 9.217317618057162e-06, "loss": 0.0, "step": 10588 }, { "epoch": 2.45, "learning_rate": 9.213465834681457e-06, "loss": 0.0, "step": 10589 }, { "epoch": 2.45, "learning_rate": 9.209614051305755e-06, "loss": 0.0, "step": 10590 }, { "epoch": 2.45, "learning_rate": 9.205762267930052e-06, "loss": 0.0, "step": 10591 }, { "epoch": 2.45, "learning_rate": 9.201910484554349e-06, "loss": 0.0, "step": 10592 }, { "epoch": 2.45, "learning_rate": 9.198058701178647e-06, "loss": 0.0, "step": 10593 }, { "epoch": 2.45, "learning_rate": 9.194206917802944e-06, "loss": 0.0, "step": 10594 }, { "epoch": 2.45, "learning_rate": 9.19035513442724e-06, "loss": 0.0, "step": 10595 }, { "epoch": 2.45, "learning_rate": 9.186503351051537e-06, "loss": 0.0, "step": 10596 }, { "epoch": 2.45, "learning_rate": 9.182651567675835e-06, "loss": 0.0, "step": 10597 }, { "epoch": 2.45, "learning_rate": 9.17879978430013e-06, "loss": 0.0, "step": 10598 }, { "epoch": 2.45, "learning_rate": 9.174948000924429e-06, "loss": 0.0, "step": 10599 }, { "epoch": 2.45, "learning_rate": 9.171096217548727e-06, "loss": 0.0, "step": 10600 }, { "epoch": 2.45, "learning_rate": 9.167244434173022e-06, "loss": 0.0, "step": 10601 }, { "epoch": 2.45, "learning_rate": 9.16339265079732e-06, "loss": 0.0, "step": 10602 }, { "epoch": 2.45, "learning_rate": 9.159540867421617e-06, "loss": 0.0, "step": 10603 }, { "epoch": 2.45, "learning_rate": 9.155689084045913e-06, "loss": 0.0001, "step": 10604 }, { "epoch": 2.45, "learning_rate": 9.15183730067021e-06, "loss": 0.0, "step": 10605 }, { "epoch": 2.45, "learning_rate": 9.147985517294508e-06, "loss": 0.0, "step": 10606 }, { "epoch": 2.45, "learning_rate": 9.144133733918805e-06, "loss": 0.0, "step": 10607 }, { "epoch": 2.45, "learning_rate": 9.140281950543102e-06, "loss": 0.0, "step": 10608 }, { "epoch": 2.45, "learning_rate": 9.1364301671674e-06, "loss": 0.0, "step": 10609 }, { "epoch": 2.45, "learning_rate": 9.132578383791695e-06, "loss": 0.0, "step": 10610 }, { "epoch": 2.45, "learning_rate": 9.128726600415993e-06, "loss": 0.0, "step": 10611 }, { "epoch": 2.45, "learning_rate": 9.12487481704029e-06, "loss": 0.0, "step": 10612 }, { "epoch": 2.45, "learning_rate": 9.121023033664587e-06, "loss": 0.0, "step": 10613 }, { "epoch": 2.45, "learning_rate": 9.117171250288885e-06, "loss": 0.0, "step": 10614 }, { "epoch": 2.45, "learning_rate": 9.113319466913182e-06, "loss": 0.0, "step": 10615 }, { "epoch": 2.45, "learning_rate": 9.109467683537478e-06, "loss": 0.0, "step": 10616 }, { "epoch": 2.45, "learning_rate": 9.105615900161775e-06, "loss": 0.0, "step": 10617 }, { "epoch": 2.45, "learning_rate": 9.101764116786073e-06, "loss": 0.0, "step": 10618 }, { "epoch": 2.45, "learning_rate": 9.097912333410368e-06, "loss": 0.0, "step": 10619 }, { "epoch": 2.45, "learning_rate": 9.094060550034667e-06, "loss": 0.0, "step": 10620 }, { "epoch": 2.45, "learning_rate": 9.090208766658963e-06, "loss": 0.0, "step": 10621 }, { "epoch": 2.45, "learning_rate": 9.08635698328326e-06, "loss": 0.0, "step": 10622 }, { "epoch": 2.46, "learning_rate": 9.082505199907558e-06, "loss": 0.0, "step": 10623 }, { "epoch": 2.46, "learning_rate": 9.078653416531855e-06, "loss": 0.0, "step": 10624 }, { "epoch": 2.46, "learning_rate": 9.074801633156152e-06, "loss": 0.0, "step": 10625 }, { "epoch": 2.46, "learning_rate": 9.070949849780448e-06, "loss": 0.0002, "step": 10626 }, { "epoch": 2.46, "learning_rate": 9.067098066404747e-06, "loss": 0.0, "step": 10627 }, { "epoch": 2.46, "learning_rate": 9.063246283029043e-06, "loss": 0.0003, "step": 10628 }, { "epoch": 2.46, "learning_rate": 9.05939449965334e-06, "loss": 0.0, "step": 10629 }, { "epoch": 2.46, "learning_rate": 9.055542716277638e-06, "loss": 0.0, "step": 10630 }, { "epoch": 2.46, "learning_rate": 9.051690932901933e-06, "loss": 0.0, "step": 10631 }, { "epoch": 2.46, "learning_rate": 9.047839149526232e-06, "loss": 0.0, "step": 10632 }, { "epoch": 2.46, "learning_rate": 9.043987366150528e-06, "loss": 0.0, "step": 10633 }, { "epoch": 2.46, "learning_rate": 9.040135582774825e-06, "loss": 0.0, "step": 10634 }, { "epoch": 2.46, "learning_rate": 9.036283799399123e-06, "loss": 0.0, "step": 10635 }, { "epoch": 2.46, "learning_rate": 9.03243201602342e-06, "loss": 0.0, "step": 10636 }, { "epoch": 2.46, "learning_rate": 9.028580232647716e-06, "loss": 0.0, "step": 10637 }, { "epoch": 2.46, "learning_rate": 9.024728449272013e-06, "loss": 0.0, "step": 10638 }, { "epoch": 2.46, "learning_rate": 9.020876665896311e-06, "loss": 0.0, "step": 10639 }, { "epoch": 2.46, "learning_rate": 9.017024882520606e-06, "loss": 0.0, "step": 10640 }, { "epoch": 2.46, "learning_rate": 9.013173099144905e-06, "loss": 0.0, "step": 10641 }, { "epoch": 2.46, "learning_rate": 9.009321315769201e-06, "loss": 0.0, "step": 10642 }, { "epoch": 2.46, "learning_rate": 9.005469532393498e-06, "loss": 0.0, "step": 10643 }, { "epoch": 2.46, "learning_rate": 9.001617749017796e-06, "loss": 0.0, "step": 10644 }, { "epoch": 2.46, "learning_rate": 8.997765965642093e-06, "loss": 0.0, "step": 10645 }, { "epoch": 2.46, "learning_rate": 8.99391418226639e-06, "loss": 0.0, "step": 10646 }, { "epoch": 2.46, "learning_rate": 8.990062398890686e-06, "loss": 0.0, "step": 10647 }, { "epoch": 2.46, "learning_rate": 8.986210615514985e-06, "loss": 0.0, "step": 10648 }, { "epoch": 2.46, "learning_rate": 8.982358832139281e-06, "loss": 0.0, "step": 10649 }, { "epoch": 2.46, "learning_rate": 8.978507048763578e-06, "loss": 0.0, "step": 10650 }, { "epoch": 2.46, "learning_rate": 8.974655265387875e-06, "loss": 0.0, "step": 10651 }, { "epoch": 2.46, "learning_rate": 8.970803482012171e-06, "loss": 0.0, "step": 10652 }, { "epoch": 2.46, "learning_rate": 8.96695169863647e-06, "loss": 0.0, "step": 10653 }, { "epoch": 2.46, "learning_rate": 8.963099915260766e-06, "loss": 0.0, "step": 10654 }, { "epoch": 2.46, "learning_rate": 8.959248131885063e-06, "loss": 0.0, "step": 10655 }, { "epoch": 2.46, "learning_rate": 8.955396348509361e-06, "loss": 0.0, "step": 10656 }, { "epoch": 2.46, "learning_rate": 8.951544565133658e-06, "loss": 0.0, "step": 10657 }, { "epoch": 2.46, "learning_rate": 8.947692781757955e-06, "loss": 0.0, "step": 10658 }, { "epoch": 2.46, "learning_rate": 8.943840998382251e-06, "loss": 0.0, "step": 10659 }, { "epoch": 2.46, "learning_rate": 8.939989215006548e-06, "loss": 0.0, "step": 10660 }, { "epoch": 2.46, "learning_rate": 8.936137431630845e-06, "loss": 0.0, "step": 10661 }, { "epoch": 2.46, "learning_rate": 8.932285648255143e-06, "loss": 0.0, "step": 10662 }, { "epoch": 2.46, "learning_rate": 8.92843386487944e-06, "loss": 0.0, "step": 10663 }, { "epoch": 2.46, "learning_rate": 8.924582081503736e-06, "loss": 0.0, "step": 10664 }, { "epoch": 2.46, "learning_rate": 8.920730298128035e-06, "loss": 0.0, "step": 10665 }, { "epoch": 2.46, "learning_rate": 8.916878514752331e-06, "loss": 0.0, "step": 10666 }, { "epoch": 2.47, "learning_rate": 8.913026731376628e-06, "loss": 0.0, "step": 10667 }, { "epoch": 2.47, "learning_rate": 8.909174948000924e-06, "loss": 0.0, "step": 10668 }, { "epoch": 2.47, "learning_rate": 8.905323164625221e-06, "loss": 0.0, "step": 10669 }, { "epoch": 2.47, "learning_rate": 8.90147138124952e-06, "loss": 0.0, "step": 10670 }, { "epoch": 2.47, "learning_rate": 8.897619597873816e-06, "loss": 0.0001, "step": 10671 }, { "epoch": 2.47, "learning_rate": 8.893767814498113e-06, "loss": 0.0, "step": 10672 }, { "epoch": 2.47, "learning_rate": 8.88991603112241e-06, "loss": 0.0, "step": 10673 }, { "epoch": 2.47, "learning_rate": 8.886064247746708e-06, "loss": 0.0, "step": 10674 }, { "epoch": 2.47, "learning_rate": 8.882212464371004e-06, "loss": 0.0, "step": 10675 }, { "epoch": 2.47, "learning_rate": 8.878360680995301e-06, "loss": 0.0, "step": 10676 }, { "epoch": 2.47, "learning_rate": 8.8745088976196e-06, "loss": 0.0, "step": 10677 }, { "epoch": 2.47, "learning_rate": 8.870657114243894e-06, "loss": 0.0, "step": 10678 }, { "epoch": 2.47, "learning_rate": 8.866805330868193e-06, "loss": 0.0002, "step": 10679 }, { "epoch": 2.47, "learning_rate": 8.86295354749249e-06, "loss": 0.0, "step": 10680 }, { "epoch": 2.47, "learning_rate": 8.859101764116786e-06, "loss": 0.0, "step": 10681 }, { "epoch": 2.47, "learning_rate": 8.855249980741083e-06, "loss": 0.0, "step": 10682 }, { "epoch": 2.47, "learning_rate": 8.851398197365381e-06, "loss": 0.0, "step": 10683 }, { "epoch": 2.47, "learning_rate": 8.847546413989678e-06, "loss": 0.0, "step": 10684 }, { "epoch": 2.47, "learning_rate": 8.843694630613974e-06, "loss": 0.0004, "step": 10685 }, { "epoch": 2.47, "learning_rate": 8.839842847238273e-06, "loss": 0.0, "step": 10686 }, { "epoch": 2.47, "learning_rate": 8.835991063862568e-06, "loss": 0.0, "step": 10687 }, { "epoch": 2.47, "learning_rate": 8.832139280486866e-06, "loss": 0.0, "step": 10688 }, { "epoch": 2.47, "learning_rate": 8.828287497111163e-06, "loss": 0.0, "step": 10689 }, { "epoch": 2.47, "learning_rate": 8.82443571373546e-06, "loss": 0.0, "step": 10690 }, { "epoch": 2.47, "learning_rate": 8.820583930359758e-06, "loss": 0.0, "step": 10691 }, { "epoch": 2.47, "learning_rate": 8.816732146984054e-06, "loss": 0.0001, "step": 10692 }, { "epoch": 2.47, "learning_rate": 8.812880363608351e-06, "loss": 0.0, "step": 10693 }, { "epoch": 2.47, "learning_rate": 8.809028580232648e-06, "loss": 0.0, "step": 10694 }, { "epoch": 2.47, "learning_rate": 8.805176796856946e-06, "loss": 0.0, "step": 10695 }, { "epoch": 2.47, "learning_rate": 8.80132501348124e-06, "loss": 0.0, "step": 10696 }, { "epoch": 2.47, "learning_rate": 8.79747323010554e-06, "loss": 0.0, "step": 10697 }, { "epoch": 2.47, "learning_rate": 8.793621446729838e-06, "loss": 0.0, "step": 10698 }, { "epoch": 2.47, "learning_rate": 8.789769663354132e-06, "loss": 0.0, "step": 10699 }, { "epoch": 2.47, "learning_rate": 8.78591787997843e-06, "loss": 0.0, "step": 10700 }, { "epoch": 2.47, "learning_rate": 8.782066096602727e-06, "loss": 0.0, "step": 10701 }, { "epoch": 2.47, "learning_rate": 8.778214313227024e-06, "loss": 0.0, "step": 10702 }, { "epoch": 2.47, "learning_rate": 8.77436252985132e-06, "loss": 0.0, "step": 10703 }, { "epoch": 2.47, "learning_rate": 8.770510746475619e-06, "loss": 0.0, "step": 10704 }, { "epoch": 2.47, "learning_rate": 8.766658963099916e-06, "loss": 0.0, "step": 10705 }, { "epoch": 2.47, "learning_rate": 8.762807179724212e-06, "loss": 0.0, "step": 10706 }, { "epoch": 2.47, "learning_rate": 8.75895539634851e-06, "loss": 0.0, "step": 10707 }, { "epoch": 2.47, "learning_rate": 8.755103612972806e-06, "loss": 0.0, "step": 10708 }, { "epoch": 2.47, "learning_rate": 8.751251829597104e-06, "loss": 0.0, "step": 10709 }, { "epoch": 2.48, "learning_rate": 8.7474000462214e-06, "loss": 0.0, "step": 10710 }, { "epoch": 2.48, "learning_rate": 8.743548262845697e-06, "loss": 0.0, "step": 10711 }, { "epoch": 2.48, "learning_rate": 8.739696479469996e-06, "loss": 0.0, "step": 10712 }, { "epoch": 2.48, "learning_rate": 8.735844696094292e-06, "loss": 0.0, "step": 10713 }, { "epoch": 2.48, "learning_rate": 8.731992912718589e-06, "loss": 0.0, "step": 10714 }, { "epoch": 2.48, "learning_rate": 8.728141129342886e-06, "loss": 0.0, "step": 10715 }, { "epoch": 2.48, "learning_rate": 8.724289345967184e-06, "loss": 0.0, "step": 10716 }, { "epoch": 2.48, "learning_rate": 8.720437562591479e-06, "loss": 0.0, "step": 10717 }, { "epoch": 2.48, "learning_rate": 8.716585779215777e-06, "loss": 0.0, "step": 10718 }, { "epoch": 2.48, "learning_rate": 8.712733995840076e-06, "loss": 0.0, "step": 10719 }, { "epoch": 2.48, "learning_rate": 8.70888221246437e-06, "loss": 0.0, "step": 10720 }, { "epoch": 2.48, "learning_rate": 8.705030429088669e-06, "loss": 0.0, "step": 10721 }, { "epoch": 2.48, "learning_rate": 8.701178645712966e-06, "loss": 0.0, "step": 10722 }, { "epoch": 2.48, "learning_rate": 8.697326862337262e-06, "loss": 0.0, "step": 10723 }, { "epoch": 2.48, "learning_rate": 8.693475078961559e-06, "loss": 0.0, "step": 10724 }, { "epoch": 2.48, "learning_rate": 8.689623295585857e-06, "loss": 0.0, "step": 10725 }, { "epoch": 2.48, "learning_rate": 8.685771512210154e-06, "loss": 0.0, "step": 10726 }, { "epoch": 2.48, "learning_rate": 8.68191972883445e-06, "loss": 0.0, "step": 10727 }, { "epoch": 2.48, "learning_rate": 8.678067945458749e-06, "loss": 0.0, "step": 10728 }, { "epoch": 2.48, "learning_rate": 8.674216162083044e-06, "loss": 0.0, "step": 10729 }, { "epoch": 2.48, "learning_rate": 8.670364378707342e-06, "loss": 0.0, "step": 10730 }, { "epoch": 2.48, "learning_rate": 8.666512595331639e-06, "loss": 0.0, "step": 10731 }, { "epoch": 2.48, "learning_rate": 8.662660811955935e-06, "loss": 0.0, "step": 10732 }, { "epoch": 2.48, "learning_rate": 8.658809028580234e-06, "loss": 0.0, "step": 10733 }, { "epoch": 2.48, "learning_rate": 8.65495724520453e-06, "loss": 0.0, "step": 10734 }, { "epoch": 2.48, "learning_rate": 8.651105461828827e-06, "loss": 0.0, "step": 10735 }, { "epoch": 2.48, "learning_rate": 8.647253678453124e-06, "loss": 0.0, "step": 10736 }, { "epoch": 2.48, "learning_rate": 8.643401895077422e-06, "loss": 0.0, "step": 10737 }, { "epoch": 2.48, "learning_rate": 8.639550111701717e-06, "loss": 0.0, "step": 10738 }, { "epoch": 2.48, "learning_rate": 8.635698328326015e-06, "loss": 0.0, "step": 10739 }, { "epoch": 2.48, "learning_rate": 8.631846544950314e-06, "loss": 0.0, "step": 10740 }, { "epoch": 2.48, "learning_rate": 8.627994761574609e-06, "loss": 0.0, "step": 10741 }, { "epoch": 2.48, "learning_rate": 8.624142978198907e-06, "loss": 0.0, "step": 10742 }, { "epoch": 2.48, "learning_rate": 8.620291194823204e-06, "loss": 0.0, "step": 10743 }, { "epoch": 2.48, "learning_rate": 8.6164394114475e-06, "loss": 0.0001, "step": 10744 }, { "epoch": 2.48, "learning_rate": 8.612587628071797e-06, "loss": 0.0, "step": 10745 }, { "epoch": 2.48, "learning_rate": 8.608735844696095e-06, "loss": 0.0, "step": 10746 }, { "epoch": 2.48, "learning_rate": 8.604884061320392e-06, "loss": 0.0001, "step": 10747 }, { "epoch": 2.48, "learning_rate": 8.601032277944689e-06, "loss": 0.0, "step": 10748 }, { "epoch": 2.48, "learning_rate": 8.597180494568987e-06, "loss": 0.0, "step": 10749 }, { "epoch": 2.48, "learning_rate": 8.593328711193282e-06, "loss": 0.0, "step": 10750 }, { "epoch": 2.48, "learning_rate": 8.58947692781758e-06, "loss": 0.0, "step": 10751 }, { "epoch": 2.48, "learning_rate": 8.585625144441877e-06, "loss": 0.0, "step": 10752 }, { "epoch": 2.49, "learning_rate": 8.581773361066174e-06, "loss": 0.0, "step": 10753 }, { "epoch": 2.49, "learning_rate": 8.577921577690472e-06, "loss": 0.0, "step": 10754 }, { "epoch": 2.49, "learning_rate": 8.574069794314769e-06, "loss": 0.0, "step": 10755 }, { "epoch": 2.49, "learning_rate": 8.570218010939065e-06, "loss": 0.0, "step": 10756 }, { "epoch": 2.49, "learning_rate": 8.566366227563362e-06, "loss": 0.0, "step": 10757 }, { "epoch": 2.49, "learning_rate": 8.56251444418766e-06, "loss": 0.0, "step": 10758 }, { "epoch": 2.49, "learning_rate": 8.558662660811955e-06, "loss": 0.0, "step": 10759 }, { "epoch": 2.49, "learning_rate": 8.554810877436254e-06, "loss": 0.0, "step": 10760 }, { "epoch": 2.49, "learning_rate": 8.550959094060552e-06, "loss": 0.0, "step": 10761 }, { "epoch": 2.49, "learning_rate": 8.547107310684847e-06, "loss": 0.0, "step": 10762 }, { "epoch": 2.49, "learning_rate": 8.543255527309145e-06, "loss": 0.0, "step": 10763 }, { "epoch": 2.49, "learning_rate": 8.539403743933442e-06, "loss": 0.0, "step": 10764 }, { "epoch": 2.49, "learning_rate": 8.535551960557738e-06, "loss": 0.0, "step": 10765 }, { "epoch": 2.49, "learning_rate": 8.531700177182035e-06, "loss": 0.0, "step": 10766 }, { "epoch": 2.49, "learning_rate": 8.527848393806333e-06, "loss": 0.0, "step": 10767 }, { "epoch": 2.49, "learning_rate": 8.52399661043063e-06, "loss": 0.0, "step": 10768 }, { "epoch": 2.49, "learning_rate": 8.520144827054927e-06, "loss": 0.0, "step": 10769 }, { "epoch": 2.49, "learning_rate": 8.516293043679225e-06, "loss": 0.0, "step": 10770 }, { "epoch": 2.49, "learning_rate": 8.51244126030352e-06, "loss": 0.0, "step": 10771 }, { "epoch": 2.49, "learning_rate": 8.508589476927818e-06, "loss": 0.0, "step": 10772 }, { "epoch": 2.49, "learning_rate": 8.504737693552115e-06, "loss": 0.0, "step": 10773 }, { "epoch": 2.49, "learning_rate": 8.500885910176412e-06, "loss": 0.0, "step": 10774 }, { "epoch": 2.49, "learning_rate": 8.49703412680071e-06, "loss": 0.0, "step": 10775 }, { "epoch": 2.49, "learning_rate": 8.493182343425007e-06, "loss": 0.0, "step": 10776 }, { "epoch": 2.49, "learning_rate": 8.489330560049303e-06, "loss": 0.0, "step": 10777 }, { "epoch": 2.49, "learning_rate": 8.4854787766736e-06, "loss": 0.0002, "step": 10778 }, { "epoch": 2.49, "learning_rate": 8.481626993297898e-06, "loss": 0.0, "step": 10779 }, { "epoch": 2.49, "learning_rate": 8.477775209922193e-06, "loss": 0.0, "step": 10780 }, { "epoch": 2.49, "learning_rate": 8.473923426546492e-06, "loss": 0.0, "step": 10781 }, { "epoch": 2.49, "learning_rate": 8.470071643170788e-06, "loss": 0.0, "step": 10782 }, { "epoch": 2.49, "learning_rate": 8.466219859795085e-06, "loss": 0.0, "step": 10783 }, { "epoch": 2.49, "learning_rate": 8.462368076419383e-06, "loss": 0.0, "step": 10784 }, { "epoch": 2.49, "learning_rate": 8.45851629304368e-06, "loss": 0.0, "step": 10785 }, { "epoch": 2.49, "learning_rate": 8.454664509667977e-06, "loss": 0.0, "step": 10786 }, { "epoch": 2.49, "learning_rate": 8.450812726292273e-06, "loss": 0.0, "step": 10787 }, { "epoch": 2.49, "learning_rate": 8.446960942916572e-06, "loss": 0.0, "step": 10788 }, { "epoch": 2.49, "learning_rate": 8.443109159540868e-06, "loss": 0.0, "step": 10789 }, { "epoch": 2.49, "learning_rate": 8.439257376165165e-06, "loss": 0.0, "step": 10790 }, { "epoch": 2.49, "learning_rate": 8.435405592789462e-06, "loss": 0.0, "step": 10791 }, { "epoch": 2.49, "learning_rate": 8.431553809413758e-06, "loss": 0.0, "step": 10792 }, { "epoch": 2.49, "learning_rate": 8.427702026038056e-06, "loss": 0.0, "step": 10793 }, { "epoch": 2.49, "learning_rate": 8.423850242662353e-06, "loss": 0.0, "step": 10794 }, { "epoch": 2.49, "learning_rate": 8.41999845928665e-06, "loss": 0.0, "step": 10795 }, { "epoch": 2.5, "learning_rate": 8.416146675910948e-06, "loss": 0.0, "step": 10796 }, { "epoch": 2.5, "learning_rate": 8.412294892535245e-06, "loss": 0.0, "step": 10797 }, { "epoch": 2.5, "learning_rate": 8.408443109159541e-06, "loss": 0.0, "step": 10798 }, { "epoch": 2.5, "learning_rate": 8.404591325783838e-06, "loss": 0.0, "step": 10799 }, { "epoch": 2.5, "learning_rate": 8.400739542408136e-06, "loss": 0.0, "step": 10800 }, { "epoch": 2.5, "learning_rate": 8.396887759032431e-06, "loss": 0.0, "step": 10801 }, { "epoch": 2.5, "learning_rate": 8.39303597565673e-06, "loss": 0.0, "step": 10802 }, { "epoch": 2.5, "learning_rate": 8.389184192281026e-06, "loss": 0.0016, "step": 10803 }, { "epoch": 2.5, "learning_rate": 8.385332408905323e-06, "loss": 0.0, "step": 10804 }, { "epoch": 2.5, "learning_rate": 8.381480625529621e-06, "loss": 0.0, "step": 10805 }, { "epoch": 2.5, "learning_rate": 8.377628842153918e-06, "loss": 0.0, "step": 10806 }, { "epoch": 2.5, "learning_rate": 8.373777058778215e-06, "loss": 0.0, "step": 10807 }, { "epoch": 2.5, "learning_rate": 8.369925275402511e-06, "loss": 0.0, "step": 10808 }, { "epoch": 2.5, "learning_rate": 8.36607349202681e-06, "loss": 0.0, "step": 10809 }, { "epoch": 2.5, "learning_rate": 8.362221708651106e-06, "loss": 0.0, "step": 10810 }, { "epoch": 2.5, "learning_rate": 8.358369925275403e-06, "loss": 0.0, "step": 10811 }, { "epoch": 2.5, "learning_rate": 8.3545181418997e-06, "loss": 0.0, "step": 10812 }, { "epoch": 2.5, "learning_rate": 8.350666358523996e-06, "loss": 0.0, "step": 10813 }, { "epoch": 2.5, "learning_rate": 8.346814575148295e-06, "loss": 0.0, "step": 10814 }, { "epoch": 2.5, "learning_rate": 8.342962791772591e-06, "loss": 0.0, "step": 10815 }, { "epoch": 2.5, "learning_rate": 8.339111008396888e-06, "loss": 0.0, "step": 10816 }, { "epoch": 2.5, "learning_rate": 8.335259225021186e-06, "loss": 0.0, "step": 10817 }, { "epoch": 2.5, "learning_rate": 8.331407441645483e-06, "loss": 0.0, "step": 10818 }, { "epoch": 2.5, "learning_rate": 8.32755565826978e-06, "loss": 0.0, "step": 10819 }, { "epoch": 2.5, "learning_rate": 8.323703874894076e-06, "loss": 0.0, "step": 10820 }, { "epoch": 2.5, "learning_rate": 8.319852091518373e-06, "loss": 0.0, "step": 10821 }, { "epoch": 2.5, "learning_rate": 8.31600030814267e-06, "loss": 0.0, "step": 10822 }, { "epoch": 2.5, "learning_rate": 8.312148524766968e-06, "loss": 0.0, "step": 10823 }, { "epoch": 2.5, "learning_rate": 8.308296741391264e-06, "loss": 0.0, "step": 10824 }, { "epoch": 2.5, "learning_rate": 8.304444958015561e-06, "loss": 0.0, "step": 10825 }, { "epoch": 2.5, "learning_rate": 8.30059317463986e-06, "loss": 0.0, "step": 10826 }, { "epoch": 2.5, "learning_rate": 8.296741391264156e-06, "loss": 0.0, "step": 10827 }, { "epoch": 2.5, "learning_rate": 8.292889607888453e-06, "loss": 0.0, "step": 10828 }, { "epoch": 2.5, "learning_rate": 8.28903782451275e-06, "loss": 0.0, "step": 10829 }, { "epoch": 2.5, "learning_rate": 8.285186041137046e-06, "loss": 0.0, "step": 10830 }, { "epoch": 2.5, "learning_rate": 8.281334257761344e-06, "loss": 0.0, "step": 10831 }, { "epoch": 2.5, "learning_rate": 8.277482474385641e-06, "loss": 0.0, "step": 10832 }, { "epoch": 2.5, "learning_rate": 8.273630691009938e-06, "loss": 0.0, "step": 10833 }, { "epoch": 2.5, "learning_rate": 8.269778907634234e-06, "loss": 0.0001, "step": 10834 }, { "epoch": 2.5, "learning_rate": 8.265927124258533e-06, "loss": 0.0, "step": 10835 }, { "epoch": 2.5, "learning_rate": 8.26207534088283e-06, "loss": 0.0, "step": 10836 }, { "epoch": 2.5, "learning_rate": 8.258223557507126e-06, "loss": 0.0001, "step": 10837 }, { "epoch": 2.5, "learning_rate": 8.254371774131424e-06, "loss": 0.0, "step": 10838 }, { "epoch": 2.5, "learning_rate": 8.25051999075572e-06, "loss": 0.0, "step": 10839 }, { "epoch": 2.51, "learning_rate": 8.246668207380018e-06, "loss": 0.0, "step": 10840 }, { "epoch": 2.51, "learning_rate": 8.242816424004314e-06, "loss": 0.0007, "step": 10841 }, { "epoch": 2.51, "learning_rate": 8.238964640628611e-06, "loss": 0.0, "step": 10842 }, { "epoch": 2.51, "learning_rate": 8.235112857252908e-06, "loss": 0.0001, "step": 10843 }, { "epoch": 2.51, "learning_rate": 8.231261073877206e-06, "loss": 0.0, "step": 10844 }, { "epoch": 2.51, "learning_rate": 8.227409290501503e-06, "loss": 0.0, "step": 10845 }, { "epoch": 2.51, "learning_rate": 8.2235575071258e-06, "loss": 0.0, "step": 10846 }, { "epoch": 2.51, "learning_rate": 8.219705723750098e-06, "loss": 0.0, "step": 10847 }, { "epoch": 2.51, "learning_rate": 8.215853940374393e-06, "loss": 0.0, "step": 10848 }, { "epoch": 2.51, "learning_rate": 8.212002156998691e-06, "loss": 0.0, "step": 10849 }, { "epoch": 2.51, "learning_rate": 8.208150373622988e-06, "loss": 0.0, "step": 10850 }, { "epoch": 2.51, "learning_rate": 8.204298590247284e-06, "loss": 0.0, "step": 10851 }, { "epoch": 2.51, "learning_rate": 8.200446806871583e-06, "loss": 0.0, "step": 10852 }, { "epoch": 2.51, "learning_rate": 8.19659502349588e-06, "loss": 0.0, "step": 10853 }, { "epoch": 2.51, "learning_rate": 8.192743240120176e-06, "loss": 0.0, "step": 10854 }, { "epoch": 2.51, "learning_rate": 8.188891456744472e-06, "loss": 0.0, "step": 10855 }, { "epoch": 2.51, "learning_rate": 8.18503967336877e-06, "loss": 0.0, "step": 10856 }, { "epoch": 2.51, "learning_rate": 8.181187889993066e-06, "loss": 0.0, "step": 10857 }, { "epoch": 2.51, "learning_rate": 8.177336106617364e-06, "loss": 0.0, "step": 10858 }, { "epoch": 2.51, "learning_rate": 8.173484323241662e-06, "loss": 0.0, "step": 10859 }, { "epoch": 2.51, "learning_rate": 8.169632539865957e-06, "loss": 0.0, "step": 10860 }, { "epoch": 2.51, "learning_rate": 8.165780756490256e-06, "loss": 0.0, "step": 10861 }, { "epoch": 2.51, "learning_rate": 8.161928973114552e-06, "loss": 0.0, "step": 10862 }, { "epoch": 2.51, "learning_rate": 8.158077189738849e-06, "loss": 0.0, "step": 10863 }, { "epoch": 2.51, "learning_rate": 8.154225406363146e-06, "loss": 0.0, "step": 10864 }, { "epoch": 2.51, "learning_rate": 8.150373622987444e-06, "loss": 0.0, "step": 10865 }, { "epoch": 2.51, "learning_rate": 8.14652183961174e-06, "loss": 0.0, "step": 10866 }, { "epoch": 2.51, "learning_rate": 8.142670056236037e-06, "loss": 0.0, "step": 10867 }, { "epoch": 2.51, "learning_rate": 8.138818272860336e-06, "loss": 0.0, "step": 10868 }, { "epoch": 2.51, "learning_rate": 8.13496648948463e-06, "loss": 0.0, "step": 10869 }, { "epoch": 2.51, "learning_rate": 8.131114706108929e-06, "loss": 0.0, "step": 10870 }, { "epoch": 2.51, "learning_rate": 8.127262922733226e-06, "loss": 0.0, "step": 10871 }, { "epoch": 2.51, "learning_rate": 8.123411139357522e-06, "loss": 0.0, "step": 10872 }, { "epoch": 2.51, "learning_rate": 8.11955935598182e-06, "loss": 0.0, "step": 10873 }, { "epoch": 2.51, "learning_rate": 8.115707572606117e-06, "loss": 0.0, "step": 10874 }, { "epoch": 2.51, "learning_rate": 8.111855789230414e-06, "loss": 0.0, "step": 10875 }, { "epoch": 2.51, "learning_rate": 8.10800400585471e-06, "loss": 0.0, "step": 10876 }, { "epoch": 2.51, "learning_rate": 8.104152222479009e-06, "loss": 0.0001, "step": 10877 }, { "epoch": 2.51, "learning_rate": 8.100300439103304e-06, "loss": 0.0, "step": 10878 }, { "epoch": 2.51, "learning_rate": 8.096448655727602e-06, "loss": 0.0, "step": 10879 }, { "epoch": 2.51, "learning_rate": 8.0925968723519e-06, "loss": 0.0, "step": 10880 }, { "epoch": 2.51, "learning_rate": 8.088745088976196e-06, "loss": 0.0, "step": 10881 }, { "epoch": 2.51, "learning_rate": 8.084893305600494e-06, "loss": 0.0, "step": 10882 }, { "epoch": 2.52, "learning_rate": 8.08104152222479e-06, "loss": 0.0, "step": 10883 }, { "epoch": 2.52, "learning_rate": 8.077189738849087e-06, "loss": 0.0, "step": 10884 }, { "epoch": 2.52, "learning_rate": 8.073337955473384e-06, "loss": 0.0, "step": 10885 }, { "epoch": 2.52, "learning_rate": 8.069486172097682e-06, "loss": 0.0, "step": 10886 }, { "epoch": 2.52, "learning_rate": 8.065634388721979e-06, "loss": 0.0, "step": 10887 }, { "epoch": 2.52, "learning_rate": 8.061782605346275e-06, "loss": 0.0, "step": 10888 }, { "epoch": 2.52, "learning_rate": 8.057930821970574e-06, "loss": 0.0, "step": 10889 }, { "epoch": 2.52, "learning_rate": 8.054079038594869e-06, "loss": 0.0, "step": 10890 }, { "epoch": 2.52, "learning_rate": 8.050227255219167e-06, "loss": 0.0, "step": 10891 }, { "epoch": 2.52, "learning_rate": 8.046375471843464e-06, "loss": 0.0, "step": 10892 }, { "epoch": 2.52, "learning_rate": 8.04252368846776e-06, "loss": 0.0, "step": 10893 }, { "epoch": 2.52, "learning_rate": 8.038671905092059e-06, "loss": 0.0, "step": 10894 }, { "epoch": 2.52, "learning_rate": 8.034820121716355e-06, "loss": 0.0, "step": 10895 }, { "epoch": 2.52, "learning_rate": 8.030968338340652e-06, "loss": 0.0, "step": 10896 }, { "epoch": 2.52, "learning_rate": 8.027116554964949e-06, "loss": 0.0, "step": 10897 }, { "epoch": 2.52, "learning_rate": 8.023264771589247e-06, "loss": 0.0, "step": 10898 }, { "epoch": 2.52, "learning_rate": 8.019412988213542e-06, "loss": 0.0, "step": 10899 }, { "epoch": 2.52, "learning_rate": 8.01556120483784e-06, "loss": 0.0, "step": 10900 }, { "epoch": 2.52, "learning_rate": 8.011709421462139e-06, "loss": 0.0, "step": 10901 }, { "epoch": 2.52, "learning_rate": 8.007857638086434e-06, "loss": 0.0, "step": 10902 }, { "epoch": 2.52, "learning_rate": 8.004005854710732e-06, "loss": 0.0, "step": 10903 }, { "epoch": 2.52, "learning_rate": 8.000154071335029e-06, "loss": 0.0, "step": 10904 }, { "epoch": 2.52, "learning_rate": 7.996302287959325e-06, "loss": 0.0001, "step": 10905 }, { "epoch": 2.52, "learning_rate": 7.992450504583622e-06, "loss": 0.0, "step": 10906 }, { "epoch": 2.52, "learning_rate": 7.98859872120792e-06, "loss": 0.0, "step": 10907 }, { "epoch": 2.52, "learning_rate": 7.984746937832217e-06, "loss": 0.0, "step": 10908 }, { "epoch": 2.52, "learning_rate": 7.980895154456514e-06, "loss": 0.0, "step": 10909 }, { "epoch": 2.52, "learning_rate": 7.977043371080812e-06, "loss": 0.0, "step": 10910 }, { "epoch": 2.52, "learning_rate": 7.973191587705107e-06, "loss": 0.0, "step": 10911 }, { "epoch": 2.52, "learning_rate": 7.969339804329405e-06, "loss": 0.0, "step": 10912 }, { "epoch": 2.52, "learning_rate": 7.965488020953702e-06, "loss": 0.0, "step": 10913 }, { "epoch": 2.52, "learning_rate": 7.961636237577999e-06, "loss": 0.0, "step": 10914 }, { "epoch": 2.52, "learning_rate": 7.957784454202297e-06, "loss": 0.0, "step": 10915 }, { "epoch": 2.52, "learning_rate": 7.953932670826594e-06, "loss": 0.0, "step": 10916 }, { "epoch": 2.52, "learning_rate": 7.95008088745089e-06, "loss": 0.0, "step": 10917 }, { "epoch": 2.52, "learning_rate": 7.946229104075187e-06, "loss": 0.0, "step": 10918 }, { "epoch": 2.52, "learning_rate": 7.942377320699485e-06, "loss": 0.0, "step": 10919 }, { "epoch": 2.52, "learning_rate": 7.93852553732378e-06, "loss": 0.0, "step": 10920 }, { "epoch": 2.52, "learning_rate": 7.934673753948078e-06, "loss": 0.0, "step": 10921 }, { "epoch": 2.52, "learning_rate": 7.930821970572377e-06, "loss": 0.0, "step": 10922 }, { "epoch": 2.52, "learning_rate": 7.926970187196672e-06, "loss": 0.0, "step": 10923 }, { "epoch": 2.52, "learning_rate": 7.92311840382097e-06, "loss": 0.0, "step": 10924 }, { "epoch": 2.52, "learning_rate": 7.919266620445267e-06, "loss": 0.0, "step": 10925 }, { "epoch": 2.53, "learning_rate": 7.915414837069563e-06, "loss": 0.0, "step": 10926 }, { "epoch": 2.53, "learning_rate": 7.91156305369386e-06, "loss": 0.0, "step": 10927 }, { "epoch": 2.53, "learning_rate": 7.907711270318158e-06, "loss": 0.0, "step": 10928 }, { "epoch": 2.53, "learning_rate": 7.903859486942455e-06, "loss": 0.0, "step": 10929 }, { "epoch": 2.53, "learning_rate": 7.900007703566752e-06, "loss": 0.0, "step": 10930 }, { "epoch": 2.53, "learning_rate": 7.89615592019105e-06, "loss": 0.0001, "step": 10931 }, { "epoch": 2.53, "learning_rate": 7.892304136815345e-06, "loss": 0.0, "step": 10932 }, { "epoch": 2.53, "learning_rate": 7.888452353439643e-06, "loss": 0.0, "step": 10933 }, { "epoch": 2.53, "learning_rate": 7.88460057006394e-06, "loss": 0.0, "step": 10934 }, { "epoch": 2.53, "learning_rate": 7.880748786688237e-06, "loss": 0.0, "step": 10935 }, { "epoch": 2.53, "learning_rate": 7.876897003312535e-06, "loss": 0.0, "step": 10936 }, { "epoch": 2.53, "learning_rate": 7.873045219936832e-06, "loss": 0.0, "step": 10937 }, { "epoch": 2.53, "learning_rate": 7.869193436561128e-06, "loss": 0.0, "step": 10938 }, { "epoch": 2.53, "learning_rate": 7.865341653185425e-06, "loss": 0.0, "step": 10939 }, { "epoch": 2.53, "learning_rate": 7.861489869809723e-06, "loss": 0.0, "step": 10940 }, { "epoch": 2.53, "learning_rate": 7.857638086434018e-06, "loss": 0.0, "step": 10941 }, { "epoch": 2.53, "learning_rate": 7.853786303058317e-06, "loss": 0.0, "step": 10942 }, { "epoch": 2.53, "learning_rate": 7.849934519682613e-06, "loss": 0.0, "step": 10943 }, { "epoch": 2.53, "learning_rate": 7.84608273630691e-06, "loss": 0.0, "step": 10944 }, { "epoch": 2.53, "learning_rate": 7.842230952931208e-06, "loss": 0.0, "step": 10945 }, { "epoch": 2.53, "learning_rate": 7.838379169555505e-06, "loss": 0.0, "step": 10946 }, { "epoch": 2.53, "learning_rate": 7.834527386179802e-06, "loss": 0.0, "step": 10947 }, { "epoch": 2.53, "learning_rate": 7.830675602804098e-06, "loss": 0.0, "step": 10948 }, { "epoch": 2.53, "learning_rate": 7.826823819428397e-06, "loss": 0.0, "step": 10949 }, { "epoch": 2.53, "learning_rate": 7.822972036052693e-06, "loss": 0.0, "step": 10950 }, { "epoch": 2.53, "learning_rate": 7.81912025267699e-06, "loss": 0.0, "step": 10951 }, { "epoch": 2.53, "learning_rate": 7.815268469301286e-06, "loss": 0.0, "step": 10952 }, { "epoch": 2.53, "learning_rate": 7.811416685925583e-06, "loss": 0.0, "step": 10953 }, { "epoch": 2.53, "learning_rate": 7.807564902549881e-06, "loss": 0.0001, "step": 10954 }, { "epoch": 2.53, "learning_rate": 7.803713119174178e-06, "loss": 0.0, "step": 10955 }, { "epoch": 2.53, "learning_rate": 7.799861335798475e-06, "loss": 0.0, "step": 10956 }, { "epoch": 2.53, "learning_rate": 7.796009552422773e-06, "loss": 0.0, "step": 10957 }, { "epoch": 2.53, "learning_rate": 7.79215776904707e-06, "loss": 0.0, "step": 10958 }, { "epoch": 2.53, "learning_rate": 7.788305985671366e-06, "loss": 0.0, "step": 10959 }, { "epoch": 2.53, "learning_rate": 7.784454202295663e-06, "loss": 0.0, "step": 10960 }, { "epoch": 2.53, "learning_rate": 7.780602418919961e-06, "loss": 0.0, "step": 10961 }, { "epoch": 2.53, "learning_rate": 7.776750635544256e-06, "loss": 0.0, "step": 10962 }, { "epoch": 2.53, "learning_rate": 7.772898852168555e-06, "loss": 0.0, "step": 10963 }, { "epoch": 2.53, "learning_rate": 7.769047068792851e-06, "loss": 0.0, "step": 10964 }, { "epoch": 2.53, "learning_rate": 7.765195285417148e-06, "loss": 0.0, "step": 10965 }, { "epoch": 2.53, "learning_rate": 7.761343502041446e-06, "loss": 0.0, "step": 10966 }, { "epoch": 2.53, "learning_rate": 7.757491718665743e-06, "loss": 0.0, "step": 10967 }, { "epoch": 2.53, "learning_rate": 7.75363993529004e-06, "loss": 0.0, "step": 10968 }, { "epoch": 2.54, "learning_rate": 7.749788151914336e-06, "loss": 0.0, "step": 10969 }, { "epoch": 2.54, "learning_rate": 7.745936368538635e-06, "loss": 0.0, "step": 10970 }, { "epoch": 2.54, "learning_rate": 7.742084585162931e-06, "loss": 0.0, "step": 10971 }, { "epoch": 2.54, "learning_rate": 7.738232801787228e-06, "loss": 0.0, "step": 10972 }, { "epoch": 2.54, "learning_rate": 7.734381018411525e-06, "loss": 0.0, "step": 10973 }, { "epoch": 2.54, "learning_rate": 7.730529235035821e-06, "loss": 0.0, "step": 10974 }, { "epoch": 2.54, "learning_rate": 7.72667745166012e-06, "loss": 0.0, "step": 10975 }, { "epoch": 2.54, "learning_rate": 7.722825668284416e-06, "loss": 0.0, "step": 10976 }, { "epoch": 2.54, "learning_rate": 7.718973884908713e-06, "loss": 0.0, "step": 10977 }, { "epoch": 2.54, "learning_rate": 7.715122101533011e-06, "loss": 0.0, "step": 10978 }, { "epoch": 2.54, "learning_rate": 7.711270318157308e-06, "loss": 0.0, "step": 10979 }, { "epoch": 2.54, "learning_rate": 7.707418534781605e-06, "loss": 0.0, "step": 10980 }, { "epoch": 2.54, "learning_rate": 7.703566751405901e-06, "loss": 0.0001, "step": 10981 }, { "epoch": 2.54, "learning_rate": 7.699714968030198e-06, "loss": 0.0, "step": 10982 }, { "epoch": 2.54, "learning_rate": 7.695863184654494e-06, "loss": 0.0, "step": 10983 }, { "epoch": 2.54, "learning_rate": 7.692011401278793e-06, "loss": 0.0, "step": 10984 }, { "epoch": 2.54, "learning_rate": 7.68815961790309e-06, "loss": 0.0, "step": 10985 }, { "epoch": 2.54, "learning_rate": 7.684307834527386e-06, "loss": 0.0, "step": 10986 }, { "epoch": 2.54, "learning_rate": 7.680456051151684e-06, "loss": 0.0, "step": 10987 }, { "epoch": 2.54, "learning_rate": 7.676604267775981e-06, "loss": 0.0, "step": 10988 }, { "epoch": 2.54, "learning_rate": 7.672752484400278e-06, "loss": 0.0, "step": 10989 }, { "epoch": 2.54, "learning_rate": 7.668900701024574e-06, "loss": 0.0, "step": 10990 }, { "epoch": 2.54, "learning_rate": 7.665048917648871e-06, "loss": 0.0, "step": 10991 }, { "epoch": 2.54, "learning_rate": 7.66119713427317e-06, "loss": 0.0, "step": 10992 }, { "epoch": 2.54, "learning_rate": 7.657345350897466e-06, "loss": 0.0, "step": 10993 }, { "epoch": 2.54, "learning_rate": 7.653493567521763e-06, "loss": 0.0, "step": 10994 }, { "epoch": 2.54, "learning_rate": 7.64964178414606e-06, "loss": 0.0, "step": 10995 }, { "epoch": 2.54, "learning_rate": 7.645790000770358e-06, "loss": 0.0, "step": 10996 }, { "epoch": 2.54, "learning_rate": 7.641938217394654e-06, "loss": 0.0, "step": 10997 }, { "epoch": 2.54, "learning_rate": 7.638086434018951e-06, "loss": 0.0, "step": 10998 }, { "epoch": 2.54, "learning_rate": 7.63423465064325e-06, "loss": 0.0, "step": 10999 }, { "epoch": 2.54, "learning_rate": 7.630382867267544e-06, "loss": 0.0, "step": 11000 }, { "epoch": 2.54, "learning_rate": 7.626531083891843e-06, "loss": 0.0, "step": 11001 }, { "epoch": 2.54, "learning_rate": 7.622679300516139e-06, "loss": 0.0, "step": 11002 }, { "epoch": 2.54, "learning_rate": 7.618827517140437e-06, "loss": 0.0, "step": 11003 }, { "epoch": 2.54, "learning_rate": 7.614975733764733e-06, "loss": 0.0, "step": 11004 }, { "epoch": 2.54, "learning_rate": 7.611123950389031e-06, "loss": 0.0, "step": 11005 }, { "epoch": 2.54, "learning_rate": 7.607272167013328e-06, "loss": 0.0, "step": 11006 }, { "epoch": 2.54, "learning_rate": 7.603420383637624e-06, "loss": 0.0, "step": 11007 }, { "epoch": 2.54, "learning_rate": 7.599568600261922e-06, "loss": 0.0, "step": 11008 }, { "epoch": 2.54, "learning_rate": 7.595716816886218e-06, "loss": 0.0, "step": 11009 }, { "epoch": 2.54, "learning_rate": 7.591865033510516e-06, "loss": 0.0, "step": 11010 }, { "epoch": 2.54, "learning_rate": 7.5880132501348125e-06, "loss": 0.0, "step": 11011 }, { "epoch": 2.54, "learning_rate": 7.58416146675911e-06, "loss": 0.0, "step": 11012 }, { "epoch": 2.55, "learning_rate": 7.5803096833834075e-06, "loss": 0.0, "step": 11013 }, { "epoch": 2.55, "learning_rate": 7.576457900007704e-06, "loss": 0.0001, "step": 11014 }, { "epoch": 2.55, "learning_rate": 7.572606116632002e-06, "loss": 0.0, "step": 11015 }, { "epoch": 2.55, "learning_rate": 7.5687543332562975e-06, "loss": 0.0, "step": 11016 }, { "epoch": 2.55, "learning_rate": 7.564902549880595e-06, "loss": 0.0, "step": 11017 }, { "epoch": 2.55, "learning_rate": 7.561050766504892e-06, "loss": 0.0, "step": 11018 }, { "epoch": 2.55, "learning_rate": 7.557198983129189e-06, "loss": 0.0, "step": 11019 }, { "epoch": 2.55, "learning_rate": 7.553347199753487e-06, "loss": 0.0, "step": 11020 }, { "epoch": 2.55, "learning_rate": 7.549495416377783e-06, "loss": 0.0, "step": 11021 }, { "epoch": 2.55, "learning_rate": 7.545643633002081e-06, "loss": 0.0, "step": 11022 }, { "epoch": 2.55, "learning_rate": 7.541791849626377e-06, "loss": 0.0, "step": 11023 }, { "epoch": 2.55, "learning_rate": 7.537940066250675e-06, "loss": 0.0, "step": 11024 }, { "epoch": 2.55, "learning_rate": 7.534088282874971e-06, "loss": 0.0, "step": 11025 }, { "epoch": 2.55, "learning_rate": 7.530236499499268e-06, "loss": 0.0, "step": 11026 }, { "epoch": 2.55, "learning_rate": 7.5263847161235665e-06, "loss": 0.0, "step": 11027 }, { "epoch": 2.55, "learning_rate": 7.522532932747862e-06, "loss": 0.0001, "step": 11028 }, { "epoch": 2.55, "learning_rate": 7.51868114937216e-06, "loss": 0.0, "step": 11029 }, { "epoch": 2.55, "learning_rate": 7.5148293659964565e-06, "loss": 0.0, "step": 11030 }, { "epoch": 2.55, "learning_rate": 7.510977582620754e-06, "loss": 0.0, "step": 11031 }, { "epoch": 2.55, "learning_rate": 7.507125799245051e-06, "loss": 0.0001, "step": 11032 }, { "epoch": 2.55, "learning_rate": 7.503274015869348e-06, "loss": 0.0, "step": 11033 }, { "epoch": 2.55, "learning_rate": 7.499422232493646e-06, "loss": 0.0, "step": 11034 }, { "epoch": 2.55, "learning_rate": 7.495570449117941e-06, "loss": 0.0, "step": 11035 }, { "epoch": 2.55, "learning_rate": 7.49171866574224e-06, "loss": 0.0, "step": 11036 }, { "epoch": 2.55, "learning_rate": 7.4878668823665356e-06, "loss": 0.0, "step": 11037 }, { "epoch": 2.55, "learning_rate": 7.484015098990833e-06, "loss": 0.0, "step": 11038 }, { "epoch": 2.55, "learning_rate": 7.48016331561513e-06, "loss": 0.0, "step": 11039 }, { "epoch": 2.55, "learning_rate": 7.476311532239427e-06, "loss": 0.0, "step": 11040 }, { "epoch": 2.55, "learning_rate": 7.472459748863725e-06, "loss": 0.0, "step": 11041 }, { "epoch": 2.55, "learning_rate": 7.468607965488021e-06, "loss": 0.0, "step": 11042 }, { "epoch": 2.55, "learning_rate": 7.464756182112319e-06, "loss": 0.0, "step": 11043 }, { "epoch": 2.55, "learning_rate": 7.460904398736615e-06, "loss": 0.0, "step": 11044 }, { "epoch": 2.55, "learning_rate": 7.457052615360913e-06, "loss": 0.0, "step": 11045 }, { "epoch": 2.55, "learning_rate": 7.453200831985209e-06, "loss": 0.0, "step": 11046 }, { "epoch": 2.55, "learning_rate": 7.449349048609506e-06, "loss": 0.0, "step": 11047 }, { "epoch": 2.55, "learning_rate": 7.445497265233805e-06, "loss": 0.0, "step": 11048 }, { "epoch": 2.55, "learning_rate": 7.4416454818581004e-06, "loss": 0.0, "step": 11049 }, { "epoch": 2.55, "learning_rate": 7.437793698482398e-06, "loss": 0.0, "step": 11050 }, { "epoch": 2.55, "learning_rate": 7.433941915106695e-06, "loss": 0.0, "step": 11051 }, { "epoch": 2.55, "learning_rate": 7.430090131730992e-06, "loss": 0.0, "step": 11052 }, { "epoch": 2.55, "learning_rate": 7.426238348355288e-06, "loss": 0.0, "step": 11053 }, { "epoch": 2.55, "learning_rate": 7.422386564979586e-06, "loss": 0.0, "step": 11054 }, { "epoch": 2.55, "learning_rate": 7.418534781603884e-06, "loss": 0.0, "step": 11055 }, { "epoch": 2.56, "learning_rate": 7.4146829982281795e-06, "loss": 0.0, "step": 11056 }, { "epoch": 2.56, "learning_rate": 7.410831214852478e-06, "loss": 0.0, "step": 11057 }, { "epoch": 2.56, "learning_rate": 7.406979431476774e-06, "loss": 0.0, "step": 11058 }, { "epoch": 2.56, "learning_rate": 7.403127648101071e-06, "loss": 0.0, "step": 11059 }, { "epoch": 2.56, "learning_rate": 7.399275864725368e-06, "loss": 0.0, "step": 11060 }, { "epoch": 2.56, "learning_rate": 7.395424081349665e-06, "loss": 0.0, "step": 11061 }, { "epoch": 2.56, "learning_rate": 7.391572297973963e-06, "loss": 0.0, "step": 11062 }, { "epoch": 2.56, "learning_rate": 7.3877205145982595e-06, "loss": 0.0, "step": 11063 }, { "epoch": 2.56, "learning_rate": 7.383868731222557e-06, "loss": 0.0, "step": 11064 }, { "epoch": 2.56, "learning_rate": 7.380016947846853e-06, "loss": 0.0, "step": 11065 }, { "epoch": 2.56, "learning_rate": 7.376165164471151e-06, "loss": 0.0001, "step": 11066 }, { "epoch": 2.56, "learning_rate": 7.372313381095447e-06, "loss": 0.0, "step": 11067 }, { "epoch": 2.56, "learning_rate": 7.368461597719744e-06, "loss": 0.0, "step": 11068 }, { "epoch": 2.56, "learning_rate": 7.364609814344042e-06, "loss": 0.0, "step": 11069 }, { "epoch": 2.56, "learning_rate": 7.3607580309683386e-06, "loss": 0.0, "step": 11070 }, { "epoch": 2.56, "learning_rate": 7.356906247592636e-06, "loss": 0.0016, "step": 11071 }, { "epoch": 2.56, "learning_rate": 7.353054464216933e-06, "loss": 0.0, "step": 11072 }, { "epoch": 2.56, "learning_rate": 7.34920268084123e-06, "loss": 0.0, "step": 11073 }, { "epoch": 2.56, "learning_rate": 7.345350897465526e-06, "loss": 0.0, "step": 11074 }, { "epoch": 2.56, "learning_rate": 7.341499114089824e-06, "loss": 0.0, "step": 11075 }, { "epoch": 2.56, "learning_rate": 7.337647330714122e-06, "loss": 0.0, "step": 11076 }, { "epoch": 2.56, "learning_rate": 7.333795547338418e-06, "loss": 0.0, "step": 11077 }, { "epoch": 2.56, "learning_rate": 7.329943763962715e-06, "loss": 0.0, "step": 11078 }, { "epoch": 2.56, "learning_rate": 7.326091980587012e-06, "loss": 0.0, "step": 11079 }, { "epoch": 2.56, "learning_rate": 7.322240197211309e-06, "loss": 0.0, "step": 11080 }, { "epoch": 2.56, "learning_rate": 7.318388413835606e-06, "loss": 0.0, "step": 11081 }, { "epoch": 2.56, "learning_rate": 7.3145366304599034e-06, "loss": 0.0, "step": 11082 }, { "epoch": 2.56, "learning_rate": 7.310684847084201e-06, "loss": 0.0, "step": 11083 }, { "epoch": 2.56, "learning_rate": 7.306833063708498e-06, "loss": 0.0, "step": 11084 }, { "epoch": 2.56, "learning_rate": 7.302981280332795e-06, "loss": 0.0, "step": 11085 }, { "epoch": 2.56, "learning_rate": 7.299129496957091e-06, "loss": 0.0, "step": 11086 }, { "epoch": 2.56, "learning_rate": 7.295277713581388e-06, "loss": 0.0, "step": 11087 }, { "epoch": 2.56, "learning_rate": 7.291425930205685e-06, "loss": 0.0, "step": 11088 }, { "epoch": 2.56, "learning_rate": 7.2875741468299825e-06, "loss": 0.0, "step": 11089 }, { "epoch": 2.56, "learning_rate": 7.28372236345428e-06, "loss": 0.0, "step": 11090 }, { "epoch": 2.56, "learning_rate": 7.279870580078577e-06, "loss": 0.0, "step": 11091 }, { "epoch": 2.56, "learning_rate": 7.276018796702874e-06, "loss": 0.0, "step": 11092 }, { "epoch": 2.56, "learning_rate": 7.272167013327171e-06, "loss": 0.0, "step": 11093 }, { "epoch": 2.56, "learning_rate": 7.268315229951468e-06, "loss": 0.0, "step": 11094 }, { "epoch": 2.56, "learning_rate": 7.264463446575764e-06, "loss": 0.0, "step": 11095 }, { "epoch": 2.56, "learning_rate": 7.260611663200062e-06, "loss": 0.0, "step": 11096 }, { "epoch": 2.56, "learning_rate": 7.25675987982436e-06, "loss": 0.0, "step": 11097 }, { "epoch": 2.56, "learning_rate": 7.252908096448656e-06, "loss": 0.0, "step": 11098 }, { "epoch": 2.57, "learning_rate": 7.249056313072953e-06, "loss": 0.0, "step": 11099 }, { "epoch": 2.57, "learning_rate": 7.24520452969725e-06, "loss": 0.0, "step": 11100 }, { "epoch": 2.57, "learning_rate": 7.241352746321547e-06, "loss": 0.0, "step": 11101 }, { "epoch": 2.57, "learning_rate": 7.237500962945844e-06, "loss": 0.0, "step": 11102 }, { "epoch": 2.57, "learning_rate": 7.2336491795701415e-06, "loss": 0.0, "step": 11103 }, { "epoch": 2.57, "learning_rate": 7.229797396194439e-06, "loss": 0.0, "step": 11104 }, { "epoch": 2.57, "learning_rate": 7.225945612818736e-06, "loss": 0.0, "step": 11105 }, { "epoch": 2.57, "learning_rate": 7.222093829443033e-06, "loss": 0.0, "step": 11106 }, { "epoch": 2.57, "learning_rate": 7.218242046067329e-06, "loss": 0.0, "step": 11107 }, { "epoch": 2.57, "learning_rate": 7.2143902626916265e-06, "loss": 0.0, "step": 11108 }, { "epoch": 2.57, "learning_rate": 7.210538479315923e-06, "loss": 0.0, "step": 11109 }, { "epoch": 2.57, "learning_rate": 7.206686695940221e-06, "loss": 0.0, "step": 11110 }, { "epoch": 2.57, "learning_rate": 7.202834912564518e-06, "loss": 0.0, "step": 11111 }, { "epoch": 2.57, "learning_rate": 7.198983129188815e-06, "loss": 0.0, "step": 11112 }, { "epoch": 2.57, "learning_rate": 7.195131345813112e-06, "loss": 0.0, "step": 11113 }, { "epoch": 2.57, "learning_rate": 7.191279562437409e-06, "loss": 0.0, "step": 11114 }, { "epoch": 2.57, "learning_rate": 7.1874277790617064e-06, "loss": 0.0, "step": 11115 }, { "epoch": 2.57, "learning_rate": 7.183575995686002e-06, "loss": 0.0, "step": 11116 }, { "epoch": 2.57, "learning_rate": 7.1797242123103e-06, "loss": 0.0, "step": 11117 }, { "epoch": 2.57, "learning_rate": 7.175872428934598e-06, "loss": 0.0, "step": 11118 }, { "epoch": 2.57, "learning_rate": 7.172020645558894e-06, "loss": 0.0, "step": 11119 }, { "epoch": 2.57, "learning_rate": 7.168168862183191e-06, "loss": 0.0, "step": 11120 }, { "epoch": 2.57, "learning_rate": 7.164317078807488e-06, "loss": 0.0, "step": 11121 }, { "epoch": 2.57, "learning_rate": 7.1604652954317855e-06, "loss": 0.0, "step": 11122 }, { "epoch": 2.57, "learning_rate": 7.156613512056082e-06, "loss": 0.0001, "step": 11123 }, { "epoch": 2.57, "learning_rate": 7.15276172868038e-06, "loss": 0.0, "step": 11124 }, { "epoch": 2.57, "learning_rate": 7.148909945304677e-06, "loss": 0.0, "step": 11125 }, { "epoch": 2.57, "learning_rate": 7.145058161928973e-06, "loss": 0.0, "step": 11126 }, { "epoch": 2.57, "learning_rate": 7.141206378553271e-06, "loss": 0.0, "step": 11127 }, { "epoch": 2.57, "learning_rate": 7.137354595177567e-06, "loss": 0.0, "step": 11128 }, { "epoch": 2.57, "learning_rate": 7.133502811801865e-06, "loss": 0.0, "step": 11129 }, { "epoch": 2.57, "learning_rate": 7.129651028426161e-06, "loss": 0.0, "step": 11130 }, { "epoch": 2.57, "learning_rate": 7.125799245050459e-06, "loss": 0.0, "step": 11131 }, { "epoch": 2.57, "learning_rate": 7.121947461674756e-06, "loss": 0.0, "step": 11132 }, { "epoch": 2.57, "learning_rate": 7.118095678299053e-06, "loss": 0.0, "step": 11133 }, { "epoch": 2.57, "learning_rate": 7.11424389492335e-06, "loss": 0.0, "step": 11134 }, { "epoch": 2.57, "learning_rate": 7.110392111547646e-06, "loss": 0.0, "step": 11135 }, { "epoch": 2.57, "learning_rate": 7.1065403281719445e-06, "loss": 0.0, "step": 11136 }, { "epoch": 2.57, "learning_rate": 7.10268854479624e-06, "loss": 0.0, "step": 11137 }, { "epoch": 2.57, "learning_rate": 7.098836761420538e-06, "loss": 0.0, "step": 11138 }, { "epoch": 2.57, "learning_rate": 7.094984978044835e-06, "loss": 0.0, "step": 11139 }, { "epoch": 2.57, "learning_rate": 7.091133194669132e-06, "loss": 0.0, "step": 11140 }, { "epoch": 2.57, "learning_rate": 7.0872814112934295e-06, "loss": 0.0, "step": 11141 }, { "epoch": 2.57, "learning_rate": 7.083429627917726e-06, "loss": 0.0, "step": 11142 }, { "epoch": 2.58, "learning_rate": 7.079577844542024e-06, "loss": 0.0, "step": 11143 }, { "epoch": 2.58, "learning_rate": 7.0757260611663194e-06, "loss": 0.0, "step": 11144 }, { "epoch": 2.58, "learning_rate": 7.071874277790618e-06, "loss": 0.0, "step": 11145 }, { "epoch": 2.58, "learning_rate": 7.068022494414915e-06, "loss": 0.0, "step": 11146 }, { "epoch": 2.58, "learning_rate": 7.064170711039211e-06, "loss": 0.0, "step": 11147 }, { "epoch": 2.58, "learning_rate": 7.060318927663509e-06, "loss": 0.0, "step": 11148 }, { "epoch": 2.58, "learning_rate": 7.056467144287805e-06, "loss": 0.0, "step": 11149 }, { "epoch": 2.58, "learning_rate": 7.052615360912103e-06, "loss": 0.0, "step": 11150 }, { "epoch": 2.58, "learning_rate": 7.048763577536399e-06, "loss": 0.0, "step": 11151 }, { "epoch": 2.58, "learning_rate": 7.044911794160697e-06, "loss": 0.0, "step": 11152 }, { "epoch": 2.58, "learning_rate": 7.041060010784994e-06, "loss": 0.0, "step": 11153 }, { "epoch": 2.58, "learning_rate": 7.037208227409291e-06, "loss": 0.0, "step": 11154 }, { "epoch": 2.58, "learning_rate": 7.0333564440335885e-06, "loss": 0.0, "step": 11155 }, { "epoch": 2.58, "learning_rate": 7.029504660657884e-06, "loss": 0.0, "step": 11156 }, { "epoch": 2.58, "learning_rate": 7.025652877282183e-06, "loss": 0.0, "step": 11157 }, { "epoch": 2.58, "learning_rate": 7.0218010939064785e-06, "loss": 0.0, "step": 11158 }, { "epoch": 2.58, "learning_rate": 7.017949310530776e-06, "loss": 0.0, "step": 11159 }, { "epoch": 2.58, "learning_rate": 7.0140975271550734e-06, "loss": 0.0, "step": 11160 }, { "epoch": 2.58, "learning_rate": 7.01024574377937e-06, "loss": 0.0, "step": 11161 }, { "epoch": 2.58, "learning_rate": 7.006393960403668e-06, "loss": 0.0, "step": 11162 }, { "epoch": 2.58, "learning_rate": 7.002542177027964e-06, "loss": 0.0, "step": 11163 }, { "epoch": 2.58, "learning_rate": 6.998690393652262e-06, "loss": 0.0, "step": 11164 }, { "epoch": 2.58, "learning_rate": 6.9948386102765575e-06, "loss": 0.0, "step": 11165 }, { "epoch": 2.58, "learning_rate": 6.990986826900856e-06, "loss": 0.0, "step": 11166 }, { "epoch": 2.58, "learning_rate": 6.987135043525153e-06, "loss": 0.0, "step": 11167 }, { "epoch": 2.58, "learning_rate": 6.983283260149449e-06, "loss": 0.0, "step": 11168 }, { "epoch": 2.58, "learning_rate": 6.979431476773747e-06, "loss": 0.0, "step": 11169 }, { "epoch": 2.58, "learning_rate": 6.975579693398043e-06, "loss": 0.0, "step": 11170 }, { "epoch": 2.58, "learning_rate": 6.971727910022341e-06, "loss": 0.0, "step": 11171 }, { "epoch": 2.58, "learning_rate": 6.9678761266466375e-06, "loss": 0.0, "step": 11172 }, { "epoch": 2.58, "learning_rate": 6.964024343270935e-06, "loss": 0.0, "step": 11173 }, { "epoch": 2.58, "learning_rate": 6.9601725598952325e-06, "loss": 0.0, "step": 11174 }, { "epoch": 2.58, "learning_rate": 6.956320776519529e-06, "loss": 0.0, "step": 11175 }, { "epoch": 2.58, "learning_rate": 6.952468993143827e-06, "loss": 0.0, "step": 11176 }, { "epoch": 2.58, "learning_rate": 6.948617209768122e-06, "loss": 0.0, "step": 11177 }, { "epoch": 2.58, "learning_rate": 6.94476542639242e-06, "loss": 0.0, "step": 11178 }, { "epoch": 2.58, "learning_rate": 6.9409136430167166e-06, "loss": 0.0, "step": 11179 }, { "epoch": 2.58, "learning_rate": 6.937061859641014e-06, "loss": 0.0, "step": 11180 }, { "epoch": 2.58, "learning_rate": 6.9332100762653116e-06, "loss": 0.0001, "step": 11181 }, { "epoch": 2.58, "learning_rate": 6.929358292889608e-06, "loss": 0.0, "step": 11182 }, { "epoch": 2.58, "learning_rate": 6.925506509513906e-06, "loss": 0.0, "step": 11183 }, { "epoch": 2.58, "learning_rate": 6.921654726138202e-06, "loss": 0.0, "step": 11184 }, { "epoch": 2.58, "learning_rate": 6.9178029427625e-06, "loss": 0.0, "step": 11185 }, { "epoch": 2.59, "learning_rate": 6.913951159386796e-06, "loss": 0.0, "step": 11186 }, { "epoch": 2.59, "learning_rate": 6.910099376011093e-06, "loss": 0.0, "step": 11187 }, { "epoch": 2.59, "learning_rate": 6.9062475926353915e-06, "loss": 0.0, "step": 11188 }, { "epoch": 2.59, "learning_rate": 6.902395809259687e-06, "loss": 0.0, "step": 11189 }, { "epoch": 2.59, "learning_rate": 6.898544025883985e-06, "loss": 0.0, "step": 11190 }, { "epoch": 2.59, "learning_rate": 6.8946922425082814e-06, "loss": 0.0, "step": 11191 }, { "epoch": 2.59, "learning_rate": 6.890840459132579e-06, "loss": 0.0, "step": 11192 }, { "epoch": 2.59, "learning_rate": 6.886988675756876e-06, "loss": 0.0, "step": 11193 }, { "epoch": 2.59, "learning_rate": 6.883136892381173e-06, "loss": 0.0, "step": 11194 }, { "epoch": 2.59, "learning_rate": 6.879285109005471e-06, "loss": 0.0, "step": 11195 }, { "epoch": 2.59, "learning_rate": 6.875433325629766e-06, "loss": 0.0, "step": 11196 }, { "epoch": 2.59, "learning_rate": 6.871581542254065e-06, "loss": 0.0, "step": 11197 }, { "epoch": 2.59, "learning_rate": 6.8677297588783605e-06, "loss": 0.0, "step": 11198 }, { "epoch": 2.59, "learning_rate": 6.863877975502658e-06, "loss": 0.0, "step": 11199 }, { "epoch": 2.59, "learning_rate": 6.860026192126955e-06, "loss": 0.0, "step": 11200 }, { "epoch": 2.59, "learning_rate": 6.856174408751252e-06, "loss": 0.0, "step": 11201 }, { "epoch": 2.59, "learning_rate": 6.85232262537555e-06, "loss": 0.0, "step": 11202 }, { "epoch": 2.59, "learning_rate": 6.848470841999846e-06, "loss": 0.0, "step": 11203 }, { "epoch": 2.59, "learning_rate": 6.844619058624144e-06, "loss": 0.0, "step": 11204 }, { "epoch": 2.59, "learning_rate": 6.84076727524844e-06, "loss": 0.0, "step": 11205 }, { "epoch": 2.59, "learning_rate": 6.836915491872738e-06, "loss": 0.0, "step": 11206 }, { "epoch": 2.59, "learning_rate": 6.833063708497034e-06, "loss": 0.0, "step": 11207 }, { "epoch": 2.59, "learning_rate": 6.829211925121331e-06, "loss": 0.0, "step": 11208 }, { "epoch": 2.59, "learning_rate": 6.82536014174563e-06, "loss": 0.0001, "step": 11209 }, { "epoch": 2.59, "learning_rate": 6.821508358369925e-06, "loss": 0.0, "step": 11210 }, { "epoch": 2.59, "learning_rate": 6.817656574994223e-06, "loss": 0.0001, "step": 11211 }, { "epoch": 2.59, "learning_rate": 6.8138047916185196e-06, "loss": 0.0, "step": 11212 }, { "epoch": 2.59, "learning_rate": 6.809953008242817e-06, "loss": 0.0, "step": 11213 }, { "epoch": 2.59, "learning_rate": 6.806101224867113e-06, "loss": 0.0, "step": 11214 }, { "epoch": 2.59, "learning_rate": 6.802249441491411e-06, "loss": 0.0, "step": 11215 }, { "epoch": 2.59, "learning_rate": 6.798397658115709e-06, "loss": 0.0, "step": 11216 }, { "epoch": 2.59, "learning_rate": 6.7945458747400045e-06, "loss": 0.0, "step": 11217 }, { "epoch": 2.59, "learning_rate": 6.790694091364303e-06, "loss": 0.0, "step": 11218 }, { "epoch": 2.59, "learning_rate": 6.786842307988599e-06, "loss": 0.0, "step": 11219 }, { "epoch": 2.59, "learning_rate": 6.782990524612896e-06, "loss": 0.0, "step": 11220 }, { "epoch": 2.59, "learning_rate": 6.779138741237193e-06, "loss": 0.0, "step": 11221 }, { "epoch": 2.59, "learning_rate": 6.77528695786149e-06, "loss": 0.0, "step": 11222 }, { "epoch": 2.59, "learning_rate": 6.771435174485788e-06, "loss": 0.0, "step": 11223 }, { "epoch": 2.59, "learning_rate": 6.7675833911100844e-06, "loss": 0.0, "step": 11224 }, { "epoch": 2.59, "learning_rate": 6.763731607734382e-06, "loss": 0.0, "step": 11225 }, { "epoch": 2.59, "learning_rate": 6.759879824358678e-06, "loss": 0.0, "step": 11226 }, { "epoch": 2.59, "learning_rate": 6.756028040982976e-06, "loss": 0.0002, "step": 11227 }, { "epoch": 2.59, "learning_rate": 6.752176257607272e-06, "loss": 0.0, "step": 11228 }, { "epoch": 2.6, "learning_rate": 6.748324474231569e-06, "loss": 0.0, "step": 11229 }, { "epoch": 2.6, "learning_rate": 6.744472690855867e-06, "loss": 0.0, "step": 11230 }, { "epoch": 2.6, "learning_rate": 6.7406209074801635e-06, "loss": 0.0, "step": 11231 }, { "epoch": 2.6, "learning_rate": 6.736769124104461e-06, "loss": 0.0, "step": 11232 }, { "epoch": 2.6, "learning_rate": 6.732917340728758e-06, "loss": 0.0, "step": 11233 }, { "epoch": 2.6, "learning_rate": 6.729065557353055e-06, "loss": 0.0, "step": 11234 }, { "epoch": 2.6, "learning_rate": 6.725213773977351e-06, "loss": 0.0, "step": 11235 }, { "epoch": 2.6, "learning_rate": 6.721361990601649e-06, "loss": 0.0, "step": 11236 }, { "epoch": 2.6, "learning_rate": 6.717510207225947e-06, "loss": 0.0, "step": 11237 }, { "epoch": 2.6, "learning_rate": 6.713658423850243e-06, "loss": 0.0, "step": 11238 }, { "epoch": 2.6, "learning_rate": 6.70980664047454e-06, "loss": 0.0, "step": 11239 }, { "epoch": 2.6, "learning_rate": 6.705954857098837e-06, "loss": 0.0, "step": 11240 }, { "epoch": 2.6, "learning_rate": 6.702103073723134e-06, "loss": 0.0, "step": 11241 }, { "epoch": 2.6, "learning_rate": 6.698251290347431e-06, "loss": 0.0, "step": 11242 }, { "epoch": 2.6, "learning_rate": 6.694399506971728e-06, "loss": 0.0, "step": 11243 }, { "epoch": 2.6, "learning_rate": 6.690547723596026e-06, "loss": 0.0, "step": 11244 }, { "epoch": 2.6, "learning_rate": 6.6866959402203225e-06, "loss": 0.0, "step": 11245 }, { "epoch": 2.6, "learning_rate": 6.68284415684462e-06, "loss": 0.0, "step": 11246 }, { "epoch": 2.6, "learning_rate": 6.678992373468916e-06, "loss": 0.0, "step": 11247 }, { "epoch": 2.6, "learning_rate": 6.675140590093213e-06, "loss": 0.0, "step": 11248 }, { "epoch": 2.6, "learning_rate": 6.67128880671751e-06, "loss": 0.0, "step": 11249 }, { "epoch": 2.6, "learning_rate": 6.6674370233418075e-06, "loss": 0.0, "step": 11250 }, { "epoch": 2.6, "learning_rate": 6.663585239966105e-06, "loss": 0.0, "step": 11251 }, { "epoch": 2.6, "learning_rate": 6.659733456590402e-06, "loss": 0.0, "step": 11252 }, { "epoch": 2.6, "learning_rate": 6.655881673214699e-06, "loss": 0.0, "step": 11253 }, { "epoch": 2.6, "learning_rate": 6.652029889838996e-06, "loss": 0.0, "step": 11254 }, { "epoch": 2.6, "learning_rate": 6.648178106463293e-06, "loss": 0.0, "step": 11255 }, { "epoch": 2.6, "learning_rate": 6.644326323087589e-06, "loss": 0.0, "step": 11256 }, { "epoch": 2.6, "learning_rate": 6.6404745397118866e-06, "loss": 0.0, "step": 11257 }, { "epoch": 2.6, "learning_rate": 6.636622756336185e-06, "loss": 0.0, "step": 11258 }, { "epoch": 2.6, "learning_rate": 6.632770972960481e-06, "loss": 0.0, "step": 11259 }, { "epoch": 2.6, "learning_rate": 6.628919189584778e-06, "loss": 0.0, "step": 11260 }, { "epoch": 2.6, "learning_rate": 6.625067406209075e-06, "loss": 0.0, "step": 11261 }, { "epoch": 2.6, "learning_rate": 6.621215622833372e-06, "loss": 0.0, "step": 11262 }, { "epoch": 2.6, "learning_rate": 6.617363839457669e-06, "loss": 0.0, "step": 11263 }, { "epoch": 2.6, "learning_rate": 6.6135120560819665e-06, "loss": 0.0, "step": 11264 }, { "epoch": 2.6, "learning_rate": 6.609660272706264e-06, "loss": 0.0, "step": 11265 }, { "epoch": 2.6, "learning_rate": 6.605808489330561e-06, "loss": 0.0, "step": 11266 }, { "epoch": 2.6, "learning_rate": 6.601956705954858e-06, "loss": 0.0, "step": 11267 }, { "epoch": 2.6, "learning_rate": 6.598104922579154e-06, "loss": 0.0, "step": 11268 }, { "epoch": 2.6, "learning_rate": 6.5942531392034515e-06, "loss": 0.0, "step": 11269 }, { "epoch": 2.6, "learning_rate": 6.590401355827748e-06, "loss": 0.0, "step": 11270 }, { "epoch": 2.6, "learning_rate": 6.586549572452046e-06, "loss": 0.0, "step": 11271 }, { "epoch": 2.61, "learning_rate": 6.582697789076343e-06, "loss": 0.0, "step": 11272 }, { "epoch": 2.61, "learning_rate": 6.57884600570064e-06, "loss": 0.0, "step": 11273 }, { "epoch": 2.61, "learning_rate": 6.574994222324937e-06, "loss": 0.0, "step": 11274 }, { "epoch": 2.61, "learning_rate": 6.571142438949234e-06, "loss": 0.0, "step": 11275 }, { "epoch": 2.61, "learning_rate": 6.567290655573531e-06, "loss": 0.0, "step": 11276 }, { "epoch": 2.61, "learning_rate": 6.563438872197827e-06, "loss": 0.0, "step": 11277 }, { "epoch": 2.61, "learning_rate": 6.559587088822125e-06, "loss": 0.0, "step": 11278 }, { "epoch": 2.61, "learning_rate": 6.555735305446423e-06, "loss": 0.0, "step": 11279 }, { "epoch": 2.61, "learning_rate": 6.551883522070719e-06, "loss": 0.0, "step": 11280 }, { "epoch": 2.61, "learning_rate": 6.548031738695016e-06, "loss": 0.0, "step": 11281 }, { "epoch": 2.61, "learning_rate": 6.544179955319313e-06, "loss": 0.0, "step": 11282 }, { "epoch": 2.61, "learning_rate": 6.5403281719436105e-06, "loss": 0.0, "step": 11283 }, { "epoch": 2.61, "learning_rate": 6.536476388567907e-06, "loss": 0.0, "step": 11284 }, { "epoch": 2.61, "learning_rate": 6.532624605192205e-06, "loss": 0.0, "step": 11285 }, { "epoch": 2.61, "learning_rate": 6.528772821816502e-06, "loss": 0.0, "step": 11286 }, { "epoch": 2.61, "learning_rate": 6.524921038440798e-06, "loss": 0.0, "step": 11287 }, { "epoch": 2.61, "learning_rate": 6.521069255065096e-06, "loss": 0.0, "step": 11288 }, { "epoch": 2.61, "learning_rate": 6.517217471689392e-06, "loss": 0.0, "step": 11289 }, { "epoch": 2.61, "learning_rate": 6.5133656883136896e-06, "loss": 0.0, "step": 11290 }, { "epoch": 2.61, "learning_rate": 6.509513904937986e-06, "loss": 0.0, "step": 11291 }, { "epoch": 2.61, "learning_rate": 6.505662121562284e-06, "loss": 0.0, "step": 11292 }, { "epoch": 2.61, "learning_rate": 6.501810338186581e-06, "loss": 0.0, "step": 11293 }, { "epoch": 2.61, "learning_rate": 6.497958554810878e-06, "loss": 0.0, "step": 11294 }, { "epoch": 2.61, "learning_rate": 6.494106771435175e-06, "loss": 0.0, "step": 11295 }, { "epoch": 2.61, "learning_rate": 6.490254988059471e-06, "loss": 0.0, "step": 11296 }, { "epoch": 2.61, "learning_rate": 6.4864032046837695e-06, "loss": 0.0, "step": 11297 }, { "epoch": 2.61, "learning_rate": 6.482551421308065e-06, "loss": 0.0, "step": 11298 }, { "epoch": 2.61, "learning_rate": 6.478699637932363e-06, "loss": 0.0, "step": 11299 }, { "epoch": 2.61, "learning_rate": 6.47484785455666e-06, "loss": 0.0, "step": 11300 }, { "epoch": 2.61, "learning_rate": 6.470996071180957e-06, "loss": 0.0, "step": 11301 }, { "epoch": 2.61, "learning_rate": 6.4671442878052544e-06, "loss": 0.0, "step": 11302 }, { "epoch": 2.61, "learning_rate": 6.463292504429551e-06, "loss": 0.0, "step": 11303 }, { "epoch": 2.61, "learning_rate": 6.459440721053849e-06, "loss": 0.0, "step": 11304 }, { "epoch": 2.61, "learning_rate": 6.455588937678144e-06, "loss": 0.0, "step": 11305 }, { "epoch": 2.61, "learning_rate": 6.451737154302443e-06, "loss": 0.0, "step": 11306 }, { "epoch": 2.61, "learning_rate": 6.44788537092674e-06, "loss": 0.0, "step": 11307 }, { "epoch": 2.61, "learning_rate": 6.444033587551036e-06, "loss": 0.0, "step": 11308 }, { "epoch": 2.61, "learning_rate": 6.440181804175334e-06, "loss": 0.0, "step": 11309 }, { "epoch": 2.61, "learning_rate": 6.43633002079963e-06, "loss": 0.0, "step": 11310 }, { "epoch": 2.61, "learning_rate": 6.432478237423928e-06, "loss": 0.0, "step": 11311 }, { "epoch": 2.61, "learning_rate": 6.428626454048224e-06, "loss": 0.0, "step": 11312 }, { "epoch": 2.61, "learning_rate": 6.424774670672522e-06, "loss": 0.0, "step": 11313 }, { "epoch": 2.61, "learning_rate": 6.420922887296819e-06, "loss": 0.0, "step": 11314 }, { "epoch": 2.61, "learning_rate": 6.417071103921116e-06, "loss": 0.0, "step": 11315 }, { "epoch": 2.62, "learning_rate": 6.4132193205454135e-06, "loss": 0.0, "step": 11316 }, { "epoch": 2.62, "learning_rate": 6.409367537169709e-06, "loss": 0.0, "step": 11317 }, { "epoch": 2.62, "learning_rate": 6.405515753794008e-06, "loss": 0.0, "step": 11318 }, { "epoch": 2.62, "learning_rate": 6.401663970418303e-06, "loss": 0.0, "step": 11319 }, { "epoch": 2.62, "learning_rate": 6.397812187042601e-06, "loss": 0.0, "step": 11320 }, { "epoch": 2.62, "learning_rate": 6.393960403666898e-06, "loss": 0.0, "step": 11321 }, { "epoch": 2.62, "learning_rate": 6.390108620291195e-06, "loss": 0.0, "step": 11322 }, { "epoch": 2.62, "learning_rate": 6.3862568369154926e-06, "loss": 0.0, "step": 11323 }, { "epoch": 2.62, "learning_rate": 6.382405053539789e-06, "loss": 0.0023, "step": 11324 }, { "epoch": 2.62, "learning_rate": 6.378553270164087e-06, "loss": 0.0, "step": 11325 }, { "epoch": 2.62, "learning_rate": 6.3747014867883825e-06, "loss": 0.0, "step": 11326 }, { "epoch": 2.62, "learning_rate": 6.370849703412681e-06, "loss": 0.0, "step": 11327 }, { "epoch": 2.62, "learning_rate": 6.366997920036978e-06, "loss": 0.0, "step": 11328 }, { "epoch": 2.62, "learning_rate": 6.363146136661274e-06, "loss": 0.0, "step": 11329 }, { "epoch": 2.62, "learning_rate": 6.359294353285572e-06, "loss": 0.0, "step": 11330 }, { "epoch": 2.62, "learning_rate": 6.355442569909868e-06, "loss": 0.0, "step": 11331 }, { "epoch": 2.62, "learning_rate": 6.351590786534166e-06, "loss": 0.0, "step": 11332 }, { "epoch": 2.62, "learning_rate": 6.3477390031584624e-06, "loss": 0.0, "step": 11333 }, { "epoch": 2.62, "learning_rate": 6.34388721978276e-06, "loss": 0.0, "step": 11334 }, { "epoch": 2.62, "learning_rate": 6.3400354364070574e-06, "loss": 0.0002, "step": 11335 }, { "epoch": 2.62, "learning_rate": 6.336183653031354e-06, "loss": 0.0, "step": 11336 }, { "epoch": 2.62, "learning_rate": 6.332331869655652e-06, "loss": 0.0, "step": 11337 }, { "epoch": 2.62, "learning_rate": 6.328480086279947e-06, "loss": 0.0, "step": 11338 }, { "epoch": 2.62, "learning_rate": 6.324628302904245e-06, "loss": 0.0, "step": 11339 }, { "epoch": 2.62, "learning_rate": 6.3207765195285415e-06, "loss": 0.0, "step": 11340 }, { "epoch": 2.62, "learning_rate": 6.316924736152839e-06, "loss": 0.0, "step": 11341 }, { "epoch": 2.62, "learning_rate": 6.3130729527771365e-06, "loss": 0.0, "step": 11342 }, { "epoch": 2.62, "learning_rate": 6.309221169401433e-06, "loss": 0.0, "step": 11343 }, { "epoch": 2.62, "learning_rate": 6.305369386025731e-06, "loss": 0.0, "step": 11344 }, { "epoch": 2.62, "learning_rate": 6.301517602650027e-06, "loss": 0.0, "step": 11345 }, { "epoch": 2.62, "learning_rate": 6.297665819274325e-06, "loss": 0.0, "step": 11346 }, { "epoch": 2.62, "learning_rate": 6.293814035898621e-06, "loss": 0.0, "step": 11347 }, { "epoch": 2.62, "learning_rate": 6.289962252522918e-06, "loss": 0.0, "step": 11348 }, { "epoch": 2.62, "learning_rate": 6.2861104691472165e-06, "loss": 0.0, "step": 11349 }, { "epoch": 2.62, "learning_rate": 6.282258685771512e-06, "loss": 0.0, "step": 11350 }, { "epoch": 2.62, "learning_rate": 6.27840690239581e-06, "loss": 0.0001, "step": 11351 }, { "epoch": 2.62, "learning_rate": 6.274555119020106e-06, "loss": 0.0, "step": 11352 }, { "epoch": 2.62, "learning_rate": 6.270703335644404e-06, "loss": 0.0, "step": 11353 }, { "epoch": 2.62, "learning_rate": 6.2668515522687006e-06, "loss": 0.0, "step": 11354 }, { "epoch": 2.62, "learning_rate": 6.262999768892998e-06, "loss": 0.0, "step": 11355 }, { "epoch": 2.62, "learning_rate": 6.2591479855172955e-06, "loss": 0.0, "step": 11356 }, { "epoch": 2.62, "learning_rate": 6.255296202141591e-06, "loss": 0.0001, "step": 11357 }, { "epoch": 2.62, "learning_rate": 6.25144441876589e-06, "loss": 0.0, "step": 11358 }, { "epoch": 2.63, "learning_rate": 6.2475926353901855e-06, "loss": 0.0, "step": 11359 }, { "epoch": 2.63, "learning_rate": 6.243740852014483e-06, "loss": 0.0, "step": 11360 }, { "epoch": 2.63, "learning_rate": 6.2398890686387805e-06, "loss": 0.0, "step": 11361 }, { "epoch": 2.63, "learning_rate": 6.236037285263077e-06, "loss": 0.0, "step": 11362 }, { "epoch": 2.63, "learning_rate": 6.232185501887374e-06, "loss": 0.0, "step": 11363 }, { "epoch": 2.63, "learning_rate": 6.228333718511671e-06, "loss": 0.0, "step": 11364 }, { "epoch": 2.63, "learning_rate": 6.224481935135968e-06, "loss": 0.0, "step": 11365 }, { "epoch": 2.63, "learning_rate": 6.220630151760265e-06, "loss": 0.0, "step": 11366 }, { "epoch": 2.63, "learning_rate": 6.216778368384563e-06, "loss": 0.0, "step": 11367 }, { "epoch": 2.63, "learning_rate": 6.2129265850088596e-06, "loss": 0.0, "step": 11368 }, { "epoch": 2.63, "learning_rate": 6.209074801633156e-06, "loss": 0.0, "step": 11369 }, { "epoch": 2.63, "learning_rate": 6.205223018257454e-06, "loss": 0.0, "step": 11370 }, { "epoch": 2.63, "learning_rate": 6.20137123488175e-06, "loss": 0.0, "step": 11371 }, { "epoch": 2.63, "learning_rate": 6.197519451506047e-06, "loss": 0.0, "step": 11372 }, { "epoch": 2.63, "learning_rate": 6.1936676681303445e-06, "loss": 0.0, "step": 11373 }, { "epoch": 2.63, "learning_rate": 6.189815884754642e-06, "loss": 0.0, "step": 11374 }, { "epoch": 2.63, "learning_rate": 6.185964101378939e-06, "loss": 0.0, "step": 11375 }, { "epoch": 2.63, "learning_rate": 6.182112318003236e-06, "loss": 0.0, "step": 11376 }, { "epoch": 2.63, "learning_rate": 6.178260534627533e-06, "loss": 0.0, "step": 11377 }, { "epoch": 2.63, "learning_rate": 6.1744087512518295e-06, "loss": 0.0, "step": 11378 }, { "epoch": 2.63, "learning_rate": 6.170556967876127e-06, "loss": 0.0, "step": 11379 }, { "epoch": 2.63, "learning_rate": 6.166705184500424e-06, "loss": 0.0, "step": 11380 }, { "epoch": 2.63, "learning_rate": 6.162853401124721e-06, "loss": 0.0, "step": 11381 }, { "epoch": 2.63, "learning_rate": 6.159001617749019e-06, "loss": 0.0, "step": 11382 }, { "epoch": 2.63, "learning_rate": 6.155149834373315e-06, "loss": 0.0, "step": 11383 }, { "epoch": 2.63, "learning_rate": 6.151298050997612e-06, "loss": 0.0001, "step": 11384 }, { "epoch": 2.63, "learning_rate": 6.147446267621909e-06, "loss": 0.0, "step": 11385 }, { "epoch": 2.63, "learning_rate": 6.143594484246206e-06, "loss": 0.0, "step": 11386 }, { "epoch": 2.63, "learning_rate": 6.139742700870503e-06, "loss": 0.0, "step": 11387 }, { "epoch": 2.63, "learning_rate": 6.135890917494801e-06, "loss": 0.0, "step": 11388 }, { "epoch": 2.63, "learning_rate": 6.132039134119098e-06, "loss": 0.0, "step": 11389 }, { "epoch": 2.63, "learning_rate": 6.128187350743394e-06, "loss": 0.0, "step": 11390 }, { "epoch": 2.63, "learning_rate": 6.124335567367692e-06, "loss": 0.0, "step": 11391 }, { "epoch": 2.63, "learning_rate": 6.1204837839919885e-06, "loss": 0.0, "step": 11392 }, { "epoch": 2.63, "learning_rate": 6.116632000616285e-06, "loss": 0.0, "step": 11393 }, { "epoch": 2.63, "learning_rate": 6.112780217240583e-06, "loss": 0.0, "step": 11394 }, { "epoch": 2.63, "learning_rate": 6.10892843386488e-06, "loss": 0.0, "step": 11395 }, { "epoch": 2.63, "learning_rate": 6.105076650489177e-06, "loss": 0.0, "step": 11396 }, { "epoch": 2.63, "learning_rate": 6.101224867113474e-06, "loss": 0.0, "step": 11397 }, { "epoch": 2.63, "learning_rate": 6.097373083737771e-06, "loss": 0.0, "step": 11398 }, { "epoch": 2.63, "learning_rate": 6.0935213003620676e-06, "loss": 0.0, "step": 11399 }, { "epoch": 2.63, "learning_rate": 6.089669516986365e-06, "loss": 0.0, "step": 11400 }, { "epoch": 2.63, "learning_rate": 6.085817733610662e-06, "loss": 0.0, "step": 11401 }, { "epoch": 2.64, "learning_rate": 6.081965950234959e-06, "loss": 0.0, "step": 11402 }, { "epoch": 2.64, "learning_rate": 6.078114166859257e-06, "loss": 0.0, "step": 11403 }, { "epoch": 2.64, "learning_rate": 6.074262383483553e-06, "loss": 0.0, "step": 11404 }, { "epoch": 2.64, "learning_rate": 6.07041060010785e-06, "loss": 0.0, "step": 11405 }, { "epoch": 2.64, "learning_rate": 6.0665588167321475e-06, "loss": 0.0001, "step": 11406 }, { "epoch": 2.64, "learning_rate": 6.062707033356444e-06, "loss": 0.0, "step": 11407 }, { "epoch": 2.64, "learning_rate": 6.058855249980741e-06, "loss": 0.0, "step": 11408 }, { "epoch": 2.64, "learning_rate": 6.055003466605038e-06, "loss": 0.0, "step": 11409 }, { "epoch": 2.64, "learning_rate": 6.051151683229336e-06, "loss": 0.0, "step": 11410 }, { "epoch": 2.64, "learning_rate": 6.0472998998536324e-06, "loss": 0.0, "step": 11411 }, { "epoch": 2.64, "learning_rate": 6.04344811647793e-06, "loss": 0.0, "step": 11412 }, { "epoch": 2.64, "learning_rate": 6.039596333102227e-06, "loss": 0.0, "step": 11413 }, { "epoch": 2.64, "learning_rate": 6.035744549726523e-06, "loss": 0.0, "step": 11414 }, { "epoch": 2.64, "learning_rate": 6.031892766350821e-06, "loss": 0.0, "step": 11415 }, { "epoch": 2.64, "learning_rate": 6.028040982975118e-06, "loss": 0.0, "step": 11416 }, { "epoch": 2.64, "learning_rate": 6.024189199599415e-06, "loss": 0.0, "step": 11417 }, { "epoch": 2.64, "learning_rate": 6.0203374162237115e-06, "loss": 0.0, "step": 11418 }, { "epoch": 2.64, "learning_rate": 6.016485632848009e-06, "loss": 0.0, "step": 11419 }, { "epoch": 2.64, "learning_rate": 6.012633849472306e-06, "loss": 0.0, "step": 11420 }, { "epoch": 2.64, "learning_rate": 6.008782066096603e-06, "loss": 0.0, "step": 11421 }, { "epoch": 2.64, "learning_rate": 6.0049302827209e-06, "loss": 0.0, "step": 11422 }, { "epoch": 2.64, "learning_rate": 6.001078499345197e-06, "loss": 0.0023, "step": 11423 }, { "epoch": 2.64, "learning_rate": 5.997226715969494e-06, "loss": 0.0, "step": 11424 }, { "epoch": 2.64, "learning_rate": 5.9933749325937915e-06, "loss": 0.0, "step": 11425 }, { "epoch": 2.64, "learning_rate": 5.989523149218088e-06, "loss": 0.0, "step": 11426 }, { "epoch": 2.64, "learning_rate": 5.985671365842385e-06, "loss": 0.0001, "step": 11427 }, { "epoch": 2.64, "learning_rate": 5.981819582466682e-06, "loss": 0.0, "step": 11428 }, { "epoch": 2.64, "learning_rate": 5.977967799090979e-06, "loss": 0.0, "step": 11429 }, { "epoch": 2.64, "learning_rate": 5.974116015715276e-06, "loss": 0.0, "step": 11430 }, { "epoch": 2.64, "learning_rate": 5.970264232339574e-06, "loss": 0.0, "step": 11431 }, { "epoch": 2.64, "learning_rate": 5.9664124489638706e-06, "loss": 0.0, "step": 11432 }, { "epoch": 2.64, "learning_rate": 5.962560665588167e-06, "loss": 0.0, "step": 11433 }, { "epoch": 2.64, "learning_rate": 5.958708882212465e-06, "loss": 0.0, "step": 11434 }, { "epoch": 2.64, "learning_rate": 5.954857098836761e-06, "loss": 0.0, "step": 11435 }, { "epoch": 2.64, "learning_rate": 5.951005315461059e-06, "loss": 0.0, "step": 11436 }, { "epoch": 2.64, "learning_rate": 5.947153532085356e-06, "loss": 0.0, "step": 11437 }, { "epoch": 2.64, "learning_rate": 5.943301748709653e-06, "loss": 0.0, "step": 11438 }, { "epoch": 2.64, "learning_rate": 5.93944996533395e-06, "loss": 0.0, "step": 11439 }, { "epoch": 2.64, "learning_rate": 5.935598181958247e-06, "loss": 0.0, "step": 11440 }, { "epoch": 2.64, "learning_rate": 5.931746398582544e-06, "loss": 0.0, "step": 11441 }, { "epoch": 2.64, "learning_rate": 5.9278946152068404e-06, "loss": 0.0, "step": 11442 }, { "epoch": 2.64, "learning_rate": 5.924042831831138e-06, "loss": 0.0, "step": 11443 }, { "epoch": 2.64, "learning_rate": 5.9201910484554354e-06, "loss": 0.0, "step": 11444 }, { "epoch": 2.65, "learning_rate": 5.916339265079732e-06, "loss": 0.0, "step": 11445 }, { "epoch": 2.65, "learning_rate": 5.91248748170403e-06, "loss": 0.0, "step": 11446 }, { "epoch": 2.65, "learning_rate": 5.908635698328326e-06, "loss": 0.0, "step": 11447 }, { "epoch": 2.65, "learning_rate": 5.904783914952623e-06, "loss": 0.0, "step": 11448 }, { "epoch": 2.65, "learning_rate": 5.90093213157692e-06, "loss": 0.0, "step": 11449 }, { "epoch": 2.65, "learning_rate": 5.897080348201217e-06, "loss": 0.0, "step": 11450 }, { "epoch": 2.65, "learning_rate": 5.8932285648255145e-06, "loss": 0.0, "step": 11451 }, { "epoch": 2.65, "learning_rate": 5.889376781449812e-06, "loss": 0.0, "step": 11452 }, { "epoch": 2.65, "learning_rate": 5.885524998074109e-06, "loss": 0.0, "step": 11453 }, { "epoch": 2.65, "learning_rate": 5.881673214698405e-06, "loss": 0.0, "step": 11454 }, { "epoch": 2.65, "learning_rate": 5.877821431322703e-06, "loss": 0.0, "step": 11455 }, { "epoch": 2.65, "learning_rate": 5.8739696479469995e-06, "loss": 0.0, "step": 11456 }, { "epoch": 2.65, "learning_rate": 5.870117864571296e-06, "loss": 0.0, "step": 11457 }, { "epoch": 2.65, "learning_rate": 5.8662660811955945e-06, "loss": 0.0, "step": 11458 }, { "epoch": 2.65, "learning_rate": 5.862414297819891e-06, "loss": 0.0, "step": 11459 }, { "epoch": 2.65, "learning_rate": 5.858562514444188e-06, "loss": 0.0, "step": 11460 }, { "epoch": 2.65, "learning_rate": 5.854710731068485e-06, "loss": 0.0, "step": 11461 }, { "epoch": 2.65, "learning_rate": 5.850858947692782e-06, "loss": 0.0, "step": 11462 }, { "epoch": 2.65, "learning_rate": 5.8470071643170786e-06, "loss": 0.0, "step": 11463 }, { "epoch": 2.65, "learning_rate": 5.843155380941376e-06, "loss": 0.0, "step": 11464 }, { "epoch": 2.65, "learning_rate": 5.8393035975656736e-06, "loss": 0.0, "step": 11465 }, { "epoch": 2.65, "learning_rate": 5.83545181418997e-06, "loss": 0.0, "step": 11466 }, { "epoch": 2.65, "learning_rate": 5.831600030814268e-06, "loss": 0.0, "step": 11467 }, { "epoch": 2.65, "learning_rate": 5.827748247438564e-06, "loss": 0.0, "step": 11468 }, { "epoch": 2.65, "learning_rate": 5.823896464062861e-06, "loss": 0.0, "step": 11469 }, { "epoch": 2.65, "learning_rate": 5.8200446806871585e-06, "loss": 0.0, "step": 11470 }, { "epoch": 2.65, "learning_rate": 5.816192897311455e-06, "loss": 0.0, "step": 11471 }, { "epoch": 2.65, "learning_rate": 5.812341113935753e-06, "loss": 0.0, "step": 11472 }, { "epoch": 2.65, "learning_rate": 5.80848933056005e-06, "loss": 0.0, "step": 11473 }, { "epoch": 2.65, "learning_rate": 5.804637547184347e-06, "loss": 0.0, "step": 11474 }, { "epoch": 2.65, "learning_rate": 5.8007857638086434e-06, "loss": 0.0, "step": 11475 }, { "epoch": 2.65, "learning_rate": 5.796933980432941e-06, "loss": 0.0, "step": 11476 }, { "epoch": 2.65, "learning_rate": 5.793082197057238e-06, "loss": 0.0, "step": 11477 }, { "epoch": 2.65, "learning_rate": 5.789230413681534e-06, "loss": 0.0, "step": 11478 }, { "epoch": 2.65, "learning_rate": 5.7853786303058326e-06, "loss": 0.0, "step": 11479 }, { "epoch": 2.65, "learning_rate": 5.781526846930129e-06, "loss": 0.0, "step": 11480 }, { "epoch": 2.65, "learning_rate": 5.777675063554426e-06, "loss": 0.0, "step": 11481 }, { "epoch": 2.65, "learning_rate": 5.773823280178723e-06, "loss": 0.0, "step": 11482 }, { "epoch": 2.65, "learning_rate": 5.76997149680302e-06, "loss": 0.0, "step": 11483 }, { "epoch": 2.65, "learning_rate": 5.766119713427317e-06, "loss": 0.0, "step": 11484 }, { "epoch": 2.65, "learning_rate": 5.762267930051614e-06, "loss": 0.0, "step": 11485 }, { "epoch": 2.65, "learning_rate": 5.758416146675912e-06, "loss": 0.0, "step": 11486 }, { "epoch": 2.65, "learning_rate": 5.754564363300208e-06, "loss": 0.0, "step": 11487 }, { "epoch": 2.65, "learning_rate": 5.750712579924506e-06, "loss": 0.0, "step": 11488 }, { "epoch": 2.66, "learning_rate": 5.7468607965488025e-06, "loss": 0.0, "step": 11489 }, { "epoch": 2.66, "learning_rate": 5.743009013173099e-06, "loss": 0.0, "step": 11490 }, { "epoch": 2.66, "learning_rate": 5.739157229797397e-06, "loss": 0.0, "step": 11491 }, { "epoch": 2.66, "learning_rate": 5.735305446421693e-06, "loss": 0.0, "step": 11492 }, { "epoch": 2.66, "learning_rate": 5.731453663045991e-06, "loss": 0.0, "step": 11493 }, { "epoch": 2.66, "learning_rate": 5.727601879670287e-06, "loss": 0.0016, "step": 11494 }, { "epoch": 2.66, "learning_rate": 5.723750096294585e-06, "loss": 0.0, "step": 11495 }, { "epoch": 2.66, "learning_rate": 5.7198983129188815e-06, "loss": 0.0, "step": 11496 }, { "epoch": 2.66, "learning_rate": 5.716046529543179e-06, "loss": 0.0, "step": 11497 }, { "epoch": 2.66, "learning_rate": 5.712194746167476e-06, "loss": 0.0002, "step": 11498 }, { "epoch": 2.66, "learning_rate": 5.708342962791772e-06, "loss": 0.0, "step": 11499 }, { "epoch": 2.66, "learning_rate": 5.70449117941607e-06, "loss": 0.0, "step": 11500 }, { "epoch": 2.66, "learning_rate": 5.700639396040367e-06, "loss": 0.0, "step": 11501 }, { "epoch": 2.66, "learning_rate": 5.696787612664664e-06, "loss": 0.0, "step": 11502 }, { "epoch": 2.66, "learning_rate": 5.692935829288961e-06, "loss": 0.0, "step": 11503 }, { "epoch": 2.66, "learning_rate": 5.689084045913258e-06, "loss": 0.0, "step": 11504 }, { "epoch": 2.66, "learning_rate": 5.685232262537555e-06, "loss": 0.0, "step": 11505 }, { "epoch": 2.66, "learning_rate": 5.681380479161852e-06, "loss": 0.0, "step": 11506 }, { "epoch": 2.66, "learning_rate": 5.67752869578615e-06, "loss": 0.0, "step": 11507 }, { "epoch": 2.66, "learning_rate": 5.6736769124104464e-06, "loss": 0.0, "step": 11508 }, { "epoch": 2.66, "learning_rate": 5.669825129034743e-06, "loss": 0.0, "step": 11509 }, { "epoch": 2.66, "learning_rate": 5.6659733456590406e-06, "loss": 0.0, "step": 11510 }, { "epoch": 2.66, "learning_rate": 5.662121562283337e-06, "loss": 0.0, "step": 11511 }, { "epoch": 2.66, "learning_rate": 5.658269778907634e-06, "loss": 0.0, "step": 11512 }, { "epoch": 2.66, "learning_rate": 5.654417995531931e-06, "loss": 0.0, "step": 11513 }, { "epoch": 2.66, "learning_rate": 5.650566212156229e-06, "loss": 0.0, "step": 11514 }, { "epoch": 2.66, "learning_rate": 5.6467144287805255e-06, "loss": 0.0, "step": 11515 }, { "epoch": 2.66, "learning_rate": 5.642862645404823e-06, "loss": 0.0, "step": 11516 }, { "epoch": 2.66, "learning_rate": 5.63901086202912e-06, "loss": 0.0, "step": 11517 }, { "epoch": 2.66, "learning_rate": 5.635159078653416e-06, "loss": 0.0, "step": 11518 }, { "epoch": 2.66, "learning_rate": 5.631307295277714e-06, "loss": 0.0, "step": 11519 }, { "epoch": 2.66, "learning_rate": 5.6274555119020105e-06, "loss": 0.0, "step": 11520 }, { "epoch": 2.66, "learning_rate": 5.623603728526308e-06, "loss": 0.0, "step": 11521 }, { "epoch": 2.66, "learning_rate": 5.6197519451506054e-06, "loss": 0.0, "step": 11522 }, { "epoch": 2.66, "learning_rate": 5.615900161774902e-06, "loss": 0.0, "step": 11523 }, { "epoch": 2.66, "learning_rate": 5.612048378399199e-06, "loss": 0.0, "step": 11524 }, { "epoch": 2.66, "learning_rate": 5.608196595023496e-06, "loss": 0.0, "step": 11525 }, { "epoch": 2.66, "learning_rate": 5.604344811647793e-06, "loss": 0.0, "step": 11526 }, { "epoch": 2.66, "learning_rate": 5.6004930282720895e-06, "loss": 0.0, "step": 11527 }, { "epoch": 2.66, "learning_rate": 5.596641244896388e-06, "loss": 0.0077, "step": 11528 }, { "epoch": 2.66, "learning_rate": 5.5927894615206845e-06, "loss": 0.0, "step": 11529 }, { "epoch": 2.66, "learning_rate": 5.588937678144981e-06, "loss": 0.0, "step": 11530 }, { "epoch": 2.66, "learning_rate": 5.585085894769279e-06, "loss": 0.0, "step": 11531 }, { "epoch": 2.67, "learning_rate": 5.581234111393575e-06, "loss": 0.0, "step": 11532 }, { "epoch": 2.67, "learning_rate": 5.577382328017872e-06, "loss": 0.0, "step": 11533 }, { "epoch": 2.67, "learning_rate": 5.5735305446421695e-06, "loss": 0.0, "step": 11534 }, { "epoch": 2.67, "learning_rate": 5.569678761266467e-06, "loss": 0.0, "step": 11535 }, { "epoch": 2.67, "learning_rate": 5.565826977890764e-06, "loss": 0.0, "step": 11536 }, { "epoch": 2.67, "learning_rate": 5.561975194515061e-06, "loss": 0.0, "step": 11537 }, { "epoch": 2.67, "learning_rate": 5.558123411139358e-06, "loss": 0.0, "step": 11538 }, { "epoch": 2.67, "learning_rate": 5.554271627763654e-06, "loss": 0.0, "step": 11539 }, { "epoch": 2.67, "learning_rate": 5.550419844387952e-06, "loss": 0.0, "step": 11540 }, { "epoch": 2.67, "learning_rate": 5.5465680610122486e-06, "loss": 0.0, "step": 11541 }, { "epoch": 2.67, "learning_rate": 5.542716277636546e-06, "loss": 0.0, "step": 11542 }, { "epoch": 2.67, "learning_rate": 5.5388644942608436e-06, "loss": 0.0, "step": 11543 }, { "epoch": 2.67, "learning_rate": 5.53501271088514e-06, "loss": 0.0, "step": 11544 }, { "epoch": 2.67, "learning_rate": 5.531160927509437e-06, "loss": 0.0, "step": 11545 }, { "epoch": 2.67, "learning_rate": 5.527309144133734e-06, "loss": 0.0, "step": 11546 }, { "epoch": 2.67, "learning_rate": 5.523457360758031e-06, "loss": 0.0, "step": 11547 }, { "epoch": 2.67, "learning_rate": 5.519605577382328e-06, "loss": 0.0, "step": 11548 }, { "epoch": 2.67, "learning_rate": 5.515753794006626e-06, "loss": 0.0, "step": 11549 }, { "epoch": 2.67, "learning_rate": 5.511902010630923e-06, "loss": 0.0, "step": 11550 }, { "epoch": 2.67, "learning_rate": 5.508050227255219e-06, "loss": 0.0, "step": 11551 }, { "epoch": 2.67, "learning_rate": 5.504198443879517e-06, "loss": 0.0, "step": 11552 }, { "epoch": 2.67, "learning_rate": 5.5003466605038134e-06, "loss": 0.0, "step": 11553 }, { "epoch": 2.67, "learning_rate": 5.49649487712811e-06, "loss": 0.0, "step": 11554 }, { "epoch": 2.67, "learning_rate": 5.492643093752408e-06, "loss": 0.0, "step": 11555 }, { "epoch": 2.67, "learning_rate": 5.488791310376705e-06, "loss": 0.0, "step": 11556 }, { "epoch": 2.67, "learning_rate": 5.484939527001002e-06, "loss": 0.0, "step": 11557 }, { "epoch": 2.67, "learning_rate": 5.481087743625299e-06, "loss": 0.0, "step": 11558 }, { "epoch": 2.67, "learning_rate": 5.477235960249596e-06, "loss": 0.0, "step": 11559 }, { "epoch": 2.67, "learning_rate": 5.4733841768738925e-06, "loss": 0.0, "step": 11560 }, { "epoch": 2.67, "learning_rate": 5.46953239349819e-06, "loss": 0.0, "step": 11561 }, { "epoch": 2.67, "learning_rate": 5.465680610122487e-06, "loss": 0.0, "step": 11562 }, { "epoch": 2.67, "learning_rate": 5.461828826746784e-06, "loss": 0.0, "step": 11563 }, { "epoch": 2.67, "learning_rate": 5.457977043371082e-06, "loss": 0.0, "step": 11564 }, { "epoch": 2.67, "learning_rate": 5.454125259995378e-06, "loss": 0.0, "step": 11565 }, { "epoch": 2.67, "learning_rate": 5.450273476619675e-06, "loss": 0.0, "step": 11566 }, { "epoch": 2.67, "learning_rate": 5.4464216932439725e-06, "loss": 0.0, "step": 11567 }, { "epoch": 2.67, "learning_rate": 5.442569909868269e-06, "loss": 0.0, "step": 11568 }, { "epoch": 2.67, "learning_rate": 5.438718126492566e-06, "loss": 0.0, "step": 11569 }, { "epoch": 2.67, "learning_rate": 5.434866343116863e-06, "loss": 0.0002, "step": 11570 }, { "epoch": 2.67, "learning_rate": 5.431014559741161e-06, "loss": 0.0, "step": 11571 }, { "epoch": 2.67, "learning_rate": 5.427162776365457e-06, "loss": 0.0, "step": 11572 }, { "epoch": 2.67, "learning_rate": 5.423310992989755e-06, "loss": 0.0, "step": 11573 }, { "epoch": 2.67, "learning_rate": 5.4194592096140516e-06, "loss": 0.0, "step": 11574 }, { "epoch": 2.68, "learning_rate": 5.415607426238348e-06, "loss": 0.0, "step": 11575 }, { "epoch": 2.68, "learning_rate": 5.411755642862646e-06, "loss": 0.0, "step": 11576 }, { "epoch": 2.68, "learning_rate": 5.407903859486943e-06, "loss": 0.0, "step": 11577 }, { "epoch": 2.68, "learning_rate": 5.40405207611124e-06, "loss": 0.0, "step": 11578 }, { "epoch": 2.68, "learning_rate": 5.4002002927355365e-06, "loss": 0.0, "step": 11579 }, { "epoch": 2.68, "learning_rate": 5.396348509359834e-06, "loss": 0.0, "step": 11580 }, { "epoch": 2.68, "learning_rate": 5.392496725984131e-06, "loss": 0.0, "step": 11581 }, { "epoch": 2.68, "learning_rate": 5.388644942608428e-06, "loss": 0.0, "step": 11582 }, { "epoch": 2.68, "learning_rate": 5.384793159232725e-06, "loss": 0.0, "step": 11583 }, { "epoch": 2.68, "learning_rate": 5.380941375857022e-06, "loss": 0.0, "step": 11584 }, { "epoch": 2.68, "learning_rate": 5.377089592481319e-06, "loss": 0.0, "step": 11585 }, { "epoch": 2.68, "learning_rate": 5.3732378091056164e-06, "loss": 0.0, "step": 11586 }, { "epoch": 2.68, "learning_rate": 5.369386025729913e-06, "loss": 0.0, "step": 11587 }, { "epoch": 2.68, "learning_rate": 5.36553424235421e-06, "loss": 0.0, "step": 11588 }, { "epoch": 2.68, "learning_rate": 5.361682458978507e-06, "loss": 0.0, "step": 11589 }, { "epoch": 2.68, "learning_rate": 5.357830675602804e-06, "loss": 0.0, "step": 11590 }, { "epoch": 2.68, "learning_rate": 5.353978892227101e-06, "loss": 0.0, "step": 11591 }, { "epoch": 2.68, "learning_rate": 5.350127108851399e-06, "loss": 0.0, "step": 11592 }, { "epoch": 2.68, "learning_rate": 5.3462753254756955e-06, "loss": 0.0, "step": 11593 }, { "epoch": 2.68, "learning_rate": 5.342423542099992e-06, "loss": 0.0, "step": 11594 }, { "epoch": 2.68, "learning_rate": 5.33857175872429e-06, "loss": 0.0001, "step": 11595 }, { "epoch": 2.68, "learning_rate": 5.334719975348586e-06, "loss": 0.0, "step": 11596 }, { "epoch": 2.68, "learning_rate": 5.330868191972884e-06, "loss": 0.0, "step": 11597 }, { "epoch": 2.68, "learning_rate": 5.327016408597181e-06, "loss": 0.0, "step": 11598 }, { "epoch": 2.68, "learning_rate": 5.323164625221478e-06, "loss": 0.0, "step": 11599 }, { "epoch": 2.68, "learning_rate": 5.319312841845775e-06, "loss": 0.0, "step": 11600 }, { "epoch": 2.68, "learning_rate": 5.315461058470072e-06, "loss": 0.0, "step": 11601 }, { "epoch": 2.68, "learning_rate": 5.311609275094369e-06, "loss": 0.0, "step": 11602 }, { "epoch": 2.68, "learning_rate": 5.307757491718665e-06, "loss": 0.0, "step": 11603 }, { "epoch": 2.68, "learning_rate": 5.303905708342963e-06, "loss": 0.0, "step": 11604 }, { "epoch": 2.68, "learning_rate": 5.30005392496726e-06, "loss": 0.0, "step": 11605 }, { "epoch": 2.68, "learning_rate": 5.296202141591557e-06, "loss": 0.0, "step": 11606 }, { "epoch": 2.68, "learning_rate": 5.2923503582158545e-06, "loss": 0.0, "step": 11607 }, { "epoch": 2.68, "learning_rate": 5.288498574840151e-06, "loss": 0.0, "step": 11608 }, { "epoch": 2.68, "learning_rate": 5.284646791464448e-06, "loss": 0.0, "step": 11609 }, { "epoch": 2.68, "learning_rate": 5.280795008088745e-06, "loss": 0.0, "step": 11610 }, { "epoch": 2.68, "learning_rate": 5.276943224713042e-06, "loss": 0.0, "step": 11611 }, { "epoch": 2.68, "learning_rate": 5.2730914413373395e-06, "loss": 0.0, "step": 11612 }, { "epoch": 2.68, "learning_rate": 5.269239657961637e-06, "loss": 0.0, "step": 11613 }, { "epoch": 2.68, "learning_rate": 5.265387874585934e-06, "loss": 0.0, "step": 11614 }, { "epoch": 2.68, "learning_rate": 5.26153609121023e-06, "loss": 0.0, "step": 11615 }, { "epoch": 2.68, "learning_rate": 5.257684307834528e-06, "loss": 0.0, "step": 11616 }, { "epoch": 2.68, "learning_rate": 5.2538325244588244e-06, "loss": 0.0, "step": 11617 }, { "epoch": 2.69, "learning_rate": 5.249980741083121e-06, "loss": 0.0, "step": 11618 }, { "epoch": 2.69, "learning_rate": 5.2461289577074194e-06, "loss": 0.0, "step": 11619 }, { "epoch": 2.69, "learning_rate": 5.242277174331716e-06, "loss": 0.0, "step": 11620 }, { "epoch": 2.69, "learning_rate": 5.238425390956013e-06, "loss": 0.0, "step": 11621 }, { "epoch": 2.69, "learning_rate": 5.23457360758031e-06, "loss": 0.0, "step": 11622 }, { "epoch": 2.69, "learning_rate": 5.230721824204607e-06, "loss": 0.0, "step": 11623 }, { "epoch": 2.69, "learning_rate": 5.2268700408289035e-06, "loss": 0.0, "step": 11624 }, { "epoch": 2.69, "learning_rate": 5.223018257453201e-06, "loss": 0.0, "step": 11625 }, { "epoch": 2.69, "learning_rate": 5.2191664740774985e-06, "loss": 0.0, "step": 11626 }, { "epoch": 2.69, "learning_rate": 5.215314690701795e-06, "loss": 0.0, "step": 11627 }, { "epoch": 2.69, "learning_rate": 5.211462907326093e-06, "loss": 0.0, "step": 11628 }, { "epoch": 2.69, "learning_rate": 5.207611123950389e-06, "loss": 0.0, "step": 11629 }, { "epoch": 2.69, "learning_rate": 5.203759340574686e-06, "loss": 0.0, "step": 11630 }, { "epoch": 2.69, "learning_rate": 5.1999075571989835e-06, "loss": 0.0, "step": 11631 }, { "epoch": 2.69, "learning_rate": 5.19605577382328e-06, "loss": 0.0, "step": 11632 }, { "epoch": 2.69, "learning_rate": 5.192203990447578e-06, "loss": 0.0, "step": 11633 }, { "epoch": 2.69, "learning_rate": 5.188352207071875e-06, "loss": 0.0, "step": 11634 }, { "epoch": 2.69, "learning_rate": 5.184500423696172e-06, "loss": 0.0, "step": 11635 }, { "epoch": 2.69, "learning_rate": 5.180648640320468e-06, "loss": 0.0, "step": 11636 }, { "epoch": 2.69, "learning_rate": 5.176796856944766e-06, "loss": 0.0, "step": 11637 }, { "epoch": 2.69, "learning_rate": 5.1729450735690625e-06, "loss": 0.0, "step": 11638 }, { "epoch": 2.69, "learning_rate": 5.169093290193359e-06, "loss": 0.0, "step": 11639 }, { "epoch": 2.69, "learning_rate": 5.1652415068176575e-06, "loss": 0.0, "step": 11640 }, { "epoch": 2.69, "learning_rate": 5.161389723441954e-06, "loss": 0.0, "step": 11641 }, { "epoch": 2.69, "learning_rate": 5.157537940066251e-06, "loss": 0.0, "step": 11642 }, { "epoch": 2.69, "learning_rate": 5.153686156690548e-06, "loss": 0.0, "step": 11643 }, { "epoch": 2.69, "learning_rate": 5.149834373314845e-06, "loss": 0.0, "step": 11644 }, { "epoch": 2.69, "learning_rate": 5.145982589939142e-06, "loss": 0.0, "step": 11645 }, { "epoch": 2.69, "learning_rate": 5.142130806563439e-06, "loss": 0.0, "step": 11646 }, { "epoch": 2.69, "learning_rate": 5.138279023187737e-06, "loss": 0.0, "step": 11647 }, { "epoch": 2.69, "learning_rate": 5.134427239812033e-06, "loss": 0.0, "step": 11648 }, { "epoch": 2.69, "learning_rate": 5.130575456436331e-06, "loss": 0.0, "step": 11649 }, { "epoch": 2.69, "learning_rate": 5.126723673060627e-06, "loss": 0.0, "step": 11650 }, { "epoch": 2.69, "learning_rate": 5.122871889684924e-06, "loss": 0.0, "step": 11651 }, { "epoch": 2.69, "learning_rate": 5.1190201063092216e-06, "loss": 0.0, "step": 11652 }, { "epoch": 2.69, "learning_rate": 5.115168322933518e-06, "loss": 0.0, "step": 11653 }, { "epoch": 2.69, "learning_rate": 5.111316539557816e-06, "loss": 0.0, "step": 11654 }, { "epoch": 2.69, "learning_rate": 5.107464756182112e-06, "loss": 0.0, "step": 11655 }, { "epoch": 2.69, "learning_rate": 5.10361297280641e-06, "loss": 0.0, "step": 11656 }, { "epoch": 2.69, "learning_rate": 5.0997611894307065e-06, "loss": 0.0, "step": 11657 }, { "epoch": 2.69, "learning_rate": 5.095909406055004e-06, "loss": 0.0, "step": 11658 }, { "epoch": 2.69, "learning_rate": 5.092057622679301e-06, "loss": 0.0, "step": 11659 }, { "epoch": 2.69, "learning_rate": 5.088205839303597e-06, "loss": 0.0, "step": 11660 }, { "epoch": 2.69, "learning_rate": 5.084354055927895e-06, "loss": 0.0, "step": 11661 }, { "epoch": 2.7, "learning_rate": 5.080502272552192e-06, "loss": 0.0, "step": 11662 }, { "epoch": 2.7, "learning_rate": 5.076650489176489e-06, "loss": 0.0, "step": 11663 }, { "epoch": 2.7, "learning_rate": 5.072798705800786e-06, "loss": 0.0, "step": 11664 }, { "epoch": 2.7, "learning_rate": 5.068946922425083e-06, "loss": 0.0, "step": 11665 }, { "epoch": 2.7, "learning_rate": 5.06509513904938e-06, "loss": 0.0, "step": 11666 }, { "epoch": 2.7, "learning_rate": 5.061243355673677e-06, "loss": 0.0, "step": 11667 }, { "epoch": 2.7, "learning_rate": 5.057391572297975e-06, "loss": 0.0001, "step": 11668 }, { "epoch": 2.7, "learning_rate": 5.053539788922271e-06, "loss": 0.0, "step": 11669 }, { "epoch": 2.7, "learning_rate": 5.049688005546568e-06, "loss": 0.0, "step": 11670 }, { "epoch": 2.7, "learning_rate": 5.0458362221708655e-06, "loss": 0.0001, "step": 11671 }, { "epoch": 2.7, "learning_rate": 5.041984438795162e-06, "loss": 0.0, "step": 11672 }, { "epoch": 2.7, "learning_rate": 5.038132655419459e-06, "loss": 0.0, "step": 11673 }, { "epoch": 2.7, "learning_rate": 5.034280872043756e-06, "loss": 0.0, "step": 11674 }, { "epoch": 2.7, "learning_rate": 5.030429088668054e-06, "loss": 0.0, "step": 11675 }, { "epoch": 2.7, "learning_rate": 5.0265773052923505e-06, "loss": 0.0, "step": 11676 }, { "epoch": 2.7, "learning_rate": 5.022725521916648e-06, "loss": 0.0, "step": 11677 }, { "epoch": 2.7, "learning_rate": 5.018873738540945e-06, "loss": 0.0, "step": 11678 }, { "epoch": 2.7, "learning_rate": 5.015021955165241e-06, "loss": 0.0, "step": 11679 }, { "epoch": 2.7, "learning_rate": 5.011170171789539e-06, "loss": 0.0, "step": 11680 }, { "epoch": 2.7, "learning_rate": 5.007318388413835e-06, "loss": 0.0, "step": 11681 }, { "epoch": 2.7, "learning_rate": 5.003466605038133e-06, "loss": 0.0, "step": 11682 }, { "epoch": 2.7, "learning_rate": 4.99961482166243e-06, "loss": 0.0, "step": 11683 }, { "epoch": 2.7, "learning_rate": 4.995763038286727e-06, "loss": 0.0, "step": 11684 }, { "epoch": 2.7, "learning_rate": 4.991911254911024e-06, "loss": 0.0, "step": 11685 }, { "epoch": 2.7, "learning_rate": 4.988059471535321e-06, "loss": 0.0, "step": 11686 }, { "epoch": 2.7, "learning_rate": 4.984207688159618e-06, "loss": 0.0, "step": 11687 }, { "epoch": 2.7, "learning_rate": 4.9803559047839145e-06, "loss": 0.0, "step": 11688 }, { "epoch": 2.7, "learning_rate": 4.976504121408213e-06, "loss": 0.0, "step": 11689 }, { "epoch": 2.7, "learning_rate": 4.9726523380325095e-06, "loss": 0.0, "step": 11690 }, { "epoch": 2.7, "learning_rate": 4.968800554656806e-06, "loss": 0.0, "step": 11691 }, { "epoch": 2.7, "learning_rate": 4.964948771281104e-06, "loss": 0.0, "step": 11692 }, { "epoch": 2.7, "learning_rate": 4.9610969879054e-06, "loss": 0.0, "step": 11693 }, { "epoch": 2.7, "learning_rate": 4.957245204529697e-06, "loss": 0.0, "step": 11694 }, { "epoch": 2.7, "learning_rate": 4.9533934211539944e-06, "loss": 0.0, "step": 11695 }, { "epoch": 2.7, "learning_rate": 4.949541637778292e-06, "loss": 0.0, "step": 11696 }, { "epoch": 2.7, "learning_rate": 4.945689854402589e-06, "loss": 0.0, "step": 11697 }, { "epoch": 2.7, "learning_rate": 4.941838071026886e-06, "loss": 0.0, "step": 11698 }, { "epoch": 2.7, "learning_rate": 4.937986287651183e-06, "loss": 0.0, "step": 11699 }, { "epoch": 2.7, "learning_rate": 4.934134504275479e-06, "loss": 0.0, "step": 11700 }, { "epoch": 2.7, "learning_rate": 4.930282720899777e-06, "loss": 0.0, "step": 11701 }, { "epoch": 2.7, "learning_rate": 4.9264309375240735e-06, "loss": 0.0, "step": 11702 }, { "epoch": 2.7, "learning_rate": 4.922579154148371e-06, "loss": 0.0, "step": 11703 }, { "epoch": 2.7, "learning_rate": 4.9187273707726685e-06, "loss": 0.0, "step": 11704 }, { "epoch": 2.71, "learning_rate": 4.914875587396965e-06, "loss": 0.0, "step": 11705 }, { "epoch": 2.71, "learning_rate": 4.911023804021262e-06, "loss": 0.0, "step": 11706 }, { "epoch": 2.71, "learning_rate": 4.907172020645559e-06, "loss": 0.0, "step": 11707 }, { "epoch": 2.71, "learning_rate": 4.903320237269856e-06, "loss": 0.0, "step": 11708 }, { "epoch": 2.71, "learning_rate": 4.899468453894153e-06, "loss": 0.0, "step": 11709 }, { "epoch": 2.71, "learning_rate": 4.895616670518451e-06, "loss": 0.0, "step": 11710 }, { "epoch": 2.71, "learning_rate": 4.891764887142748e-06, "loss": 0.0, "step": 11711 }, { "epoch": 2.71, "learning_rate": 4.887913103767044e-06, "loss": 0.0, "step": 11712 }, { "epoch": 2.71, "learning_rate": 4.884061320391342e-06, "loss": 0.0, "step": 11713 }, { "epoch": 2.71, "learning_rate": 4.880209537015638e-06, "loss": 0.0, "step": 11714 }, { "epoch": 2.71, "learning_rate": 4.876357753639935e-06, "loss": 0.0, "step": 11715 }, { "epoch": 2.71, "learning_rate": 4.8725059702642326e-06, "loss": 0.0, "step": 11716 }, { "epoch": 2.71, "learning_rate": 4.86865418688853e-06, "loss": 0.0, "step": 11717 }, { "epoch": 2.71, "learning_rate": 4.864802403512827e-06, "loss": 0.0, "step": 11718 }, { "epoch": 2.71, "learning_rate": 4.860950620137124e-06, "loss": 0.0, "step": 11719 }, { "epoch": 2.71, "learning_rate": 4.857098836761421e-06, "loss": 0.0, "step": 11720 }, { "epoch": 2.71, "learning_rate": 4.8532470533857175e-06, "loss": 0.0, "step": 11721 }, { "epoch": 2.71, "learning_rate": 4.849395270010015e-06, "loss": 0.0, "step": 11722 }, { "epoch": 2.71, "learning_rate": 4.845543486634312e-06, "loss": 0.0, "step": 11723 }, { "epoch": 2.71, "learning_rate": 4.841691703258609e-06, "loss": 0.0, "step": 11724 }, { "epoch": 2.71, "learning_rate": 4.837839919882907e-06, "loss": 0.0, "step": 11725 }, { "epoch": 2.71, "learning_rate": 4.833988136507203e-06, "loss": 0.0, "step": 11726 }, { "epoch": 2.71, "learning_rate": 4.8301363531315e-06, "loss": 0.0, "step": 11727 }, { "epoch": 2.71, "learning_rate": 4.8262845697557974e-06, "loss": 0.0, "step": 11728 }, { "epoch": 2.71, "learning_rate": 4.822432786380094e-06, "loss": 0.0, "step": 11729 }, { "epoch": 2.71, "learning_rate": 4.818581003004391e-06, "loss": 0.0, "step": 11730 }, { "epoch": 2.71, "learning_rate": 4.814729219628688e-06, "loss": 0.0, "step": 11731 }, { "epoch": 2.71, "learning_rate": 4.810877436252986e-06, "loss": 0.0, "step": 11732 }, { "epoch": 2.71, "learning_rate": 4.807025652877282e-06, "loss": 0.0, "step": 11733 }, { "epoch": 2.71, "learning_rate": 4.80317386950158e-06, "loss": 0.0, "step": 11734 }, { "epoch": 2.71, "learning_rate": 4.7993220861258765e-06, "loss": 0.0, "step": 11735 }, { "epoch": 2.71, "learning_rate": 4.795470302750173e-06, "loss": 0.0, "step": 11736 }, { "epoch": 2.71, "learning_rate": 4.791618519374471e-06, "loss": 0.0, "step": 11737 }, { "epoch": 2.71, "learning_rate": 4.787766735998768e-06, "loss": 0.0, "step": 11738 }, { "epoch": 2.71, "learning_rate": 4.783914952623065e-06, "loss": 0.0, "step": 11739 }, { "epoch": 2.71, "learning_rate": 4.7800631692473615e-06, "loss": 0.0, "step": 11740 }, { "epoch": 2.71, "learning_rate": 4.776211385871659e-06, "loss": 0.0, "step": 11741 }, { "epoch": 2.71, "learning_rate": 4.772359602495956e-06, "loss": 0.0001, "step": 11742 }, { "epoch": 2.71, "learning_rate": 4.768507819120253e-06, "loss": 0.0, "step": 11743 }, { "epoch": 2.71, "learning_rate": 4.76465603574455e-06, "loss": 0.0, "step": 11744 }, { "epoch": 2.71, "learning_rate": 4.760804252368847e-06, "loss": 0.0, "step": 11745 }, { "epoch": 2.71, "learning_rate": 4.756952468993144e-06, "loss": 0.0, "step": 11746 }, { "epoch": 2.71, "learning_rate": 4.753100685617441e-06, "loss": 0.0, "step": 11747 }, { "epoch": 2.72, "learning_rate": 4.749248902241738e-06, "loss": 0.0, "step": 11748 }, { "epoch": 2.72, "learning_rate": 4.745397118866035e-06, "loss": 0.0, "step": 11749 }, { "epoch": 2.72, "learning_rate": 4.741545335490332e-06, "loss": 0.0, "step": 11750 }, { "epoch": 2.72, "learning_rate": 4.737693552114629e-06, "loss": 0.0, "step": 11751 }, { "epoch": 2.72, "learning_rate": 4.733841768738926e-06, "loss": 0.0, "step": 11752 }, { "epoch": 2.72, "learning_rate": 4.729989985363224e-06, "loss": 0.0, "step": 11753 }, { "epoch": 2.72, "learning_rate": 4.7261382019875205e-06, "loss": 0.0, "step": 11754 }, { "epoch": 2.72, "learning_rate": 4.722286418611817e-06, "loss": 0.0, "step": 11755 }, { "epoch": 2.72, "learning_rate": 4.718434635236115e-06, "loss": 0.0, "step": 11756 }, { "epoch": 2.72, "learning_rate": 4.714582851860411e-06, "loss": 0.0, "step": 11757 }, { "epoch": 2.72, "learning_rate": 4.710731068484708e-06, "loss": 0.0, "step": 11758 }, { "epoch": 2.72, "learning_rate": 4.706879285109006e-06, "loss": 0.0, "step": 11759 }, { "epoch": 2.72, "learning_rate": 4.703027501733303e-06, "loss": 0.0, "step": 11760 }, { "epoch": 2.72, "learning_rate": 4.6991757183575996e-06, "loss": 0.0, "step": 11761 }, { "epoch": 2.72, "learning_rate": 4.695323934981897e-06, "loss": 0.0, "step": 11762 }, { "epoch": 2.72, "learning_rate": 4.691472151606194e-06, "loss": 0.0, "step": 11763 }, { "epoch": 2.72, "learning_rate": 4.68762036823049e-06, "loss": 0.0, "step": 11764 }, { "epoch": 2.72, "learning_rate": 4.683768584854788e-06, "loss": 0.0, "step": 11765 }, { "epoch": 2.72, "learning_rate": 4.679916801479085e-06, "loss": 0.0, "step": 11766 }, { "epoch": 2.72, "learning_rate": 4.676065018103382e-06, "loss": 0.0, "step": 11767 }, { "epoch": 2.72, "learning_rate": 4.6722132347276795e-06, "loss": 0.0, "step": 11768 }, { "epoch": 2.72, "learning_rate": 4.668361451351976e-06, "loss": 0.0, "step": 11769 }, { "epoch": 2.72, "learning_rate": 4.664509667976273e-06, "loss": 0.0, "step": 11770 }, { "epoch": 2.72, "learning_rate": 4.66065788460057e-06, "loss": 0.0, "step": 11771 }, { "epoch": 2.72, "learning_rate": 4.656806101224867e-06, "loss": 0.0, "step": 11772 }, { "epoch": 2.72, "learning_rate": 4.6529543178491645e-06, "loss": 0.0, "step": 11773 }, { "epoch": 2.72, "learning_rate": 4.649102534473462e-06, "loss": 0.0, "step": 11774 }, { "epoch": 2.72, "learning_rate": 4.645250751097759e-06, "loss": 0.0, "step": 11775 }, { "epoch": 2.72, "learning_rate": 4.641398967722055e-06, "loss": 0.0, "step": 11776 }, { "epoch": 2.72, "learning_rate": 4.637547184346353e-06, "loss": 0.0, "step": 11777 }, { "epoch": 2.72, "learning_rate": 4.633695400970649e-06, "loss": 0.0, "step": 11778 }, { "epoch": 2.72, "learning_rate": 4.629843617594946e-06, "loss": 0.0, "step": 11779 }, { "epoch": 2.72, "learning_rate": 4.625991834219244e-06, "loss": 0.0, "step": 11780 }, { "epoch": 2.72, "learning_rate": 4.622140050843541e-06, "loss": 0.0, "step": 11781 }, { "epoch": 2.72, "learning_rate": 4.618288267467838e-06, "loss": 0.0, "step": 11782 }, { "epoch": 2.72, "learning_rate": 4.614436484092135e-06, "loss": 0.0, "step": 11783 }, { "epoch": 2.72, "learning_rate": 4.610584700716432e-06, "loss": 0.0, "step": 11784 }, { "epoch": 2.72, "learning_rate": 4.6067329173407285e-06, "loss": 0.0, "step": 11785 }, { "epoch": 2.72, "learning_rate": 4.602881133965026e-06, "loss": 0.0, "step": 11786 }, { "epoch": 2.72, "learning_rate": 4.5990293505893235e-06, "loss": 0.0, "step": 11787 }, { "epoch": 2.72, "learning_rate": 4.59517756721362e-06, "loss": 0.0, "step": 11788 }, { "epoch": 2.72, "learning_rate": 4.591325783837918e-06, "loss": 0.0, "step": 11789 }, { "epoch": 2.72, "learning_rate": 4.587474000462214e-06, "loss": 0.0, "step": 11790 }, { "epoch": 2.72, "learning_rate": 4.583622217086511e-06, "loss": 0.0, "step": 11791 }, { "epoch": 2.73, "learning_rate": 4.579770433710808e-06, "loss": 0.0001, "step": 11792 }, { "epoch": 2.73, "learning_rate": 4.575918650335105e-06, "loss": 0.0, "step": 11793 }, { "epoch": 2.73, "learning_rate": 4.5720668669594026e-06, "loss": 0.0, "step": 11794 }, { "epoch": 2.73, "learning_rate": 4.5682150835837e-06, "loss": 0.0, "step": 11795 }, { "epoch": 2.73, "learning_rate": 4.564363300207997e-06, "loss": 0.0, "step": 11796 }, { "epoch": 2.73, "learning_rate": 4.560511516832293e-06, "loss": 0.0, "step": 11797 }, { "epoch": 2.73, "learning_rate": 4.556659733456591e-06, "loss": 0.0, "step": 11798 }, { "epoch": 2.73, "learning_rate": 4.5528079500808875e-06, "loss": 0.0, "step": 11799 }, { "epoch": 2.73, "learning_rate": 4.548956166705184e-06, "loss": 0.0, "step": 11800 }, { "epoch": 2.73, "learning_rate": 4.545104383329482e-06, "loss": 0.0, "step": 11801 }, { "epoch": 2.73, "learning_rate": 4.541252599953779e-06, "loss": 0.0, "step": 11802 }, { "epoch": 2.73, "learning_rate": 4.537400816578076e-06, "loss": 0.0, "step": 11803 }, { "epoch": 2.73, "learning_rate": 4.533549033202373e-06, "loss": 0.0001, "step": 11804 }, { "epoch": 2.73, "learning_rate": 4.52969724982667e-06, "loss": 0.0, "step": 11805 }, { "epoch": 2.73, "learning_rate": 4.525845466450967e-06, "loss": 0.0, "step": 11806 }, { "epoch": 2.73, "learning_rate": 4.521993683075264e-06, "loss": 0.0, "step": 11807 }, { "epoch": 2.73, "learning_rate": 4.518141899699562e-06, "loss": 0.0, "step": 11808 }, { "epoch": 2.73, "learning_rate": 4.514290116323858e-06, "loss": 0.0, "step": 11809 }, { "epoch": 2.73, "learning_rate": 4.510438332948156e-06, "loss": 0.0, "step": 11810 }, { "epoch": 2.73, "learning_rate": 4.506586549572452e-06, "loss": 0.0, "step": 11811 }, { "epoch": 2.73, "learning_rate": 4.502734766196749e-06, "loss": 0.0, "step": 11812 }, { "epoch": 2.73, "learning_rate": 4.4988829828210465e-06, "loss": 0.0, "step": 11813 }, { "epoch": 2.73, "learning_rate": 4.495031199445343e-06, "loss": 0.0, "step": 11814 }, { "epoch": 2.73, "learning_rate": 4.491179416069641e-06, "loss": 0.0, "step": 11815 }, { "epoch": 2.73, "learning_rate": 4.487327632693937e-06, "loss": 0.0011, "step": 11816 }, { "epoch": 2.73, "learning_rate": 4.483475849318235e-06, "loss": 0.0, "step": 11817 }, { "epoch": 2.73, "learning_rate": 4.4796240659425315e-06, "loss": 0.0, "step": 11818 }, { "epoch": 2.73, "learning_rate": 4.475772282566829e-06, "loss": 0.0001, "step": 11819 }, { "epoch": 2.73, "learning_rate": 4.471920499191126e-06, "loss": 0.0, "step": 11820 }, { "epoch": 2.73, "learning_rate": 4.468068715815422e-06, "loss": 0.0, "step": 11821 }, { "epoch": 2.73, "learning_rate": 4.46421693243972e-06, "loss": 0.0, "step": 11822 }, { "epoch": 2.73, "learning_rate": 4.460365149064017e-06, "loss": 0.0, "step": 11823 }, { "epoch": 2.73, "learning_rate": 4.456513365688314e-06, "loss": 0.0, "step": 11824 }, { "epoch": 2.73, "learning_rate": 4.4526615823126106e-06, "loss": 0.0, "step": 11825 }, { "epoch": 2.73, "learning_rate": 4.448809798936908e-06, "loss": 0.0, "step": 11826 }, { "epoch": 2.73, "learning_rate": 4.444958015561205e-06, "loss": 0.0, "step": 11827 }, { "epoch": 2.73, "learning_rate": 4.441106232185502e-06, "loss": 0.0, "step": 11828 }, { "epoch": 2.73, "learning_rate": 4.4372544488098e-06, "loss": 0.0, "step": 11829 }, { "epoch": 2.73, "learning_rate": 4.433402665434096e-06, "loss": 0.0, "step": 11830 }, { "epoch": 2.73, "learning_rate": 4.429550882058393e-06, "loss": 0.0, "step": 11831 }, { "epoch": 2.73, "learning_rate": 4.4256990986826905e-06, "loss": 0.0, "step": 11832 }, { "epoch": 2.73, "learning_rate": 4.421847315306987e-06, "loss": 0.0, "step": 11833 }, { "epoch": 2.73, "learning_rate": 4.417995531931284e-06, "loss": 0.0, "step": 11834 }, { "epoch": 2.74, "learning_rate": 4.414143748555581e-06, "loss": 0.0, "step": 11835 }, { "epoch": 2.74, "learning_rate": 4.410291965179879e-06, "loss": 0.0, "step": 11836 }, { "epoch": 2.74, "learning_rate": 4.4064401818041754e-06, "loss": 0.0, "step": 11837 }, { "epoch": 2.74, "learning_rate": 4.402588398428473e-06, "loss": 0.0, "step": 11838 }, { "epoch": 2.74, "learning_rate": 4.39873661505277e-06, "loss": 0.0, "step": 11839 }, { "epoch": 2.74, "learning_rate": 4.394884831677066e-06, "loss": 0.0, "step": 11840 }, { "epoch": 2.74, "learning_rate": 4.391033048301364e-06, "loss": 0.0, "step": 11841 }, { "epoch": 2.74, "learning_rate": 4.38718126492566e-06, "loss": 0.0, "step": 11842 }, { "epoch": 2.74, "learning_rate": 4.383329481549958e-06, "loss": 0.0, "step": 11843 }, { "epoch": 2.74, "learning_rate": 4.379477698174255e-06, "loss": 0.0, "step": 11844 }, { "epoch": 2.74, "learning_rate": 4.375625914798552e-06, "loss": 0.0, "step": 11845 }, { "epoch": 2.74, "learning_rate": 4.371774131422849e-06, "loss": 0.0, "step": 11846 }, { "epoch": 2.74, "learning_rate": 4.367922348047146e-06, "loss": 0.0, "step": 11847 }, { "epoch": 2.74, "learning_rate": 4.364070564671443e-06, "loss": 0.0, "step": 11848 }, { "epoch": 2.74, "learning_rate": 4.3602187812957395e-06, "loss": 0.0, "step": 11849 }, { "epoch": 2.74, "learning_rate": 4.356366997920038e-06, "loss": 0.0, "step": 11850 }, { "epoch": 2.74, "learning_rate": 4.3525152145443345e-06, "loss": 0.0, "step": 11851 }, { "epoch": 2.74, "learning_rate": 4.348663431168631e-06, "loss": 0.0, "step": 11852 }, { "epoch": 2.74, "learning_rate": 4.344811647792929e-06, "loss": 0.0, "step": 11853 }, { "epoch": 2.74, "learning_rate": 4.340959864417225e-06, "loss": 0.0, "step": 11854 }, { "epoch": 2.74, "learning_rate": 4.337108081041522e-06, "loss": 0.0, "step": 11855 }, { "epoch": 2.74, "learning_rate": 4.333256297665819e-06, "loss": 0.0, "step": 11856 }, { "epoch": 2.74, "learning_rate": 4.329404514290117e-06, "loss": 0.0, "step": 11857 }, { "epoch": 2.74, "learning_rate": 4.3255527309144136e-06, "loss": 0.0, "step": 11858 }, { "epoch": 2.74, "learning_rate": 4.321700947538711e-06, "loss": 0.0, "step": 11859 }, { "epoch": 2.74, "learning_rate": 4.317849164163008e-06, "loss": 0.0, "step": 11860 }, { "epoch": 2.74, "learning_rate": 4.313997380787304e-06, "loss": 0.0, "step": 11861 }, { "epoch": 2.74, "learning_rate": 4.310145597411602e-06, "loss": 0.0, "step": 11862 }, { "epoch": 2.74, "learning_rate": 4.3062938140358985e-06, "loss": 0.0, "step": 11863 }, { "epoch": 2.74, "learning_rate": 4.302442030660196e-06, "loss": 0.0, "step": 11864 }, { "epoch": 2.74, "learning_rate": 4.2985902472844935e-06, "loss": 0.0, "step": 11865 }, { "epoch": 2.74, "learning_rate": 4.29473846390879e-06, "loss": 0.0, "step": 11866 }, { "epoch": 2.74, "learning_rate": 4.290886680533087e-06, "loss": 0.0, "step": 11867 }, { "epoch": 2.74, "learning_rate": 4.287034897157384e-06, "loss": 0.0, "step": 11868 }, { "epoch": 2.74, "learning_rate": 4.283183113781681e-06, "loss": 0.0, "step": 11869 }, { "epoch": 2.74, "learning_rate": 4.279331330405978e-06, "loss": 0.0, "step": 11870 }, { "epoch": 2.74, "learning_rate": 4.275479547030276e-06, "loss": 0.0, "step": 11871 }, { "epoch": 2.74, "learning_rate": 4.2716277636545726e-06, "loss": 0.0, "step": 11872 }, { "epoch": 2.74, "learning_rate": 4.267775980278869e-06, "loss": 0.0, "step": 11873 }, { "epoch": 2.74, "learning_rate": 4.263924196903167e-06, "loss": 0.0, "step": 11874 }, { "epoch": 2.74, "learning_rate": 4.260072413527463e-06, "loss": 0.0, "step": 11875 }, { "epoch": 2.74, "learning_rate": 4.25622063015176e-06, "loss": 0.0001, "step": 11876 }, { "epoch": 2.74, "learning_rate": 4.2523688467760575e-06, "loss": 0.0, "step": 11877 }, { "epoch": 2.75, "learning_rate": 4.248517063400355e-06, "loss": 0.0, "step": 11878 }, { "epoch": 2.75, "learning_rate": 4.244665280024652e-06, "loss": 0.0, "step": 11879 }, { "epoch": 2.75, "learning_rate": 4.240813496648949e-06, "loss": 0.0, "step": 11880 }, { "epoch": 2.75, "learning_rate": 4.236961713273246e-06, "loss": 0.0, "step": 11881 }, { "epoch": 2.75, "learning_rate": 4.2331099298975425e-06, "loss": 0.0, "step": 11882 }, { "epoch": 2.75, "learning_rate": 4.22925814652184e-06, "loss": 0.0, "step": 11883 }, { "epoch": 2.75, "learning_rate": 4.225406363146137e-06, "loss": 0.0, "step": 11884 }, { "epoch": 2.75, "learning_rate": 4.221554579770434e-06, "loss": 0.0, "step": 11885 }, { "epoch": 2.75, "learning_rate": 4.217702796394731e-06, "loss": 0.0, "step": 11886 }, { "epoch": 2.75, "learning_rate": 4.213851013019028e-06, "loss": 0.0, "step": 11887 }, { "epoch": 2.75, "learning_rate": 4.209999229643325e-06, "loss": 0.0, "step": 11888 }, { "epoch": 2.75, "learning_rate": 4.206147446267622e-06, "loss": 0.0, "step": 11889 }, { "epoch": 2.75, "learning_rate": 4.202295662891919e-06, "loss": 0.0, "step": 11890 }, { "epoch": 2.75, "learning_rate": 4.198443879516216e-06, "loss": 0.0, "step": 11891 }, { "epoch": 2.75, "learning_rate": 4.194592096140513e-06, "loss": 0.0, "step": 11892 }, { "epoch": 2.75, "learning_rate": 4.190740312764811e-06, "loss": 0.0, "step": 11893 }, { "epoch": 2.75, "learning_rate": 4.186888529389107e-06, "loss": 0.0, "step": 11894 }, { "epoch": 2.75, "learning_rate": 4.183036746013405e-06, "loss": 0.0, "step": 11895 }, { "epoch": 2.75, "learning_rate": 4.1791849626377015e-06, "loss": 0.0, "step": 11896 }, { "epoch": 2.75, "learning_rate": 4.175333179261998e-06, "loss": 0.0, "step": 11897 }, { "epoch": 2.75, "learning_rate": 4.171481395886296e-06, "loss": 0.0, "step": 11898 }, { "epoch": 2.75, "learning_rate": 4.167629612510593e-06, "loss": 0.0, "step": 11899 }, { "epoch": 2.75, "learning_rate": 4.16377782913489e-06, "loss": 0.0, "step": 11900 }, { "epoch": 2.75, "learning_rate": 4.1599260457591864e-06, "loss": 0.0, "step": 11901 }, { "epoch": 2.75, "learning_rate": 4.156074262383484e-06, "loss": 0.0, "step": 11902 }, { "epoch": 2.75, "learning_rate": 4.1522224790077806e-06, "loss": 0.0001, "step": 11903 }, { "epoch": 2.75, "learning_rate": 4.148370695632078e-06, "loss": 0.0, "step": 11904 }, { "epoch": 2.75, "learning_rate": 4.144518912256375e-06, "loss": 0.0, "step": 11905 }, { "epoch": 2.75, "learning_rate": 4.140667128880672e-06, "loss": 0.0, "step": 11906 }, { "epoch": 2.75, "learning_rate": 4.136815345504969e-06, "loss": 0.0, "step": 11907 }, { "epoch": 2.75, "learning_rate": 4.132963562129266e-06, "loss": 0.0, "step": 11908 }, { "epoch": 2.75, "learning_rate": 4.129111778753563e-06, "loss": 0.0, "step": 11909 }, { "epoch": 2.75, "learning_rate": 4.12525999537786e-06, "loss": 0.0, "step": 11910 }, { "epoch": 2.75, "learning_rate": 4.121408212002157e-06, "loss": 0.0, "step": 11911 }, { "epoch": 2.75, "learning_rate": 4.117556428626454e-06, "loss": 0.0, "step": 11912 }, { "epoch": 2.75, "learning_rate": 4.113704645250751e-06, "loss": 0.0, "step": 11913 }, { "epoch": 2.75, "learning_rate": 4.109852861875049e-06, "loss": 0.0, "step": 11914 }, { "epoch": 2.75, "learning_rate": 4.1060010784993455e-06, "loss": 0.0, "step": 11915 }, { "epoch": 2.75, "learning_rate": 4.102149295123642e-06, "loss": 0.0, "step": 11916 }, { "epoch": 2.75, "learning_rate": 4.09829751174794e-06, "loss": 0.0, "step": 11917 }, { "epoch": 2.75, "learning_rate": 4.094445728372236e-06, "loss": 0.0, "step": 11918 }, { "epoch": 2.75, "learning_rate": 4.090593944996533e-06, "loss": 0.0, "step": 11919 }, { "epoch": 2.75, "learning_rate": 4.086742161620831e-06, "loss": 0.0, "step": 11920 }, { "epoch": 2.76, "learning_rate": 4.082890378245128e-06, "loss": 0.0, "step": 11921 }, { "epoch": 2.76, "learning_rate": 4.0790385948694245e-06, "loss": 0.0, "step": 11922 }, { "epoch": 2.76, "learning_rate": 4.075186811493722e-06, "loss": 0.0, "step": 11923 }, { "epoch": 2.76, "learning_rate": 4.071335028118019e-06, "loss": 0.0, "step": 11924 }, { "epoch": 2.76, "learning_rate": 4.067483244742315e-06, "loss": 0.0, "step": 11925 }, { "epoch": 2.76, "learning_rate": 4.063631461366613e-06, "loss": 0.0, "step": 11926 }, { "epoch": 2.76, "learning_rate": 4.05977967799091e-06, "loss": 0.0, "step": 11927 }, { "epoch": 2.76, "learning_rate": 4.055927894615207e-06, "loss": 0.0, "step": 11928 }, { "epoch": 2.76, "learning_rate": 4.0520761112395045e-06, "loss": 0.0, "step": 11929 }, { "epoch": 2.76, "learning_rate": 4.048224327863801e-06, "loss": 0.0, "step": 11930 }, { "epoch": 2.76, "learning_rate": 4.044372544488098e-06, "loss": 0.0, "step": 11931 }, { "epoch": 2.76, "learning_rate": 4.040520761112395e-06, "loss": 0.0, "step": 11932 }, { "epoch": 2.76, "learning_rate": 4.036668977736692e-06, "loss": 0.0, "step": 11933 }, { "epoch": 2.76, "learning_rate": 4.032817194360989e-06, "loss": 0.0, "step": 11934 }, { "epoch": 2.76, "learning_rate": 4.028965410985287e-06, "loss": 0.0, "step": 11935 }, { "epoch": 2.76, "learning_rate": 4.0251136276095836e-06, "loss": 0.0, "step": 11936 }, { "epoch": 2.76, "learning_rate": 4.02126184423388e-06, "loss": 0.0, "step": 11937 }, { "epoch": 2.76, "learning_rate": 4.017410060858178e-06, "loss": 0.0, "step": 11938 }, { "epoch": 2.76, "learning_rate": 4.013558277482474e-06, "loss": 0.0, "step": 11939 }, { "epoch": 2.76, "learning_rate": 4.009706494106771e-06, "loss": 0.0, "step": 11940 }, { "epoch": 2.76, "learning_rate": 4.005854710731069e-06, "loss": 0.0, "step": 11941 }, { "epoch": 2.76, "learning_rate": 4.002002927355366e-06, "loss": 0.0, "step": 11942 }, { "epoch": 2.76, "learning_rate": 3.998151143979663e-06, "loss": 0.0, "step": 11943 }, { "epoch": 2.76, "learning_rate": 3.99429936060396e-06, "loss": 0.0, "step": 11944 }, { "epoch": 2.76, "learning_rate": 3.990447577228257e-06, "loss": 0.0, "step": 11945 }, { "epoch": 2.76, "learning_rate": 3.9865957938525534e-06, "loss": 0.0, "step": 11946 }, { "epoch": 2.76, "learning_rate": 3.982744010476851e-06, "loss": 0.0, "step": 11947 }, { "epoch": 2.76, "learning_rate": 3.9788922271011484e-06, "loss": 0.0, "step": 11948 }, { "epoch": 2.76, "learning_rate": 3.975040443725445e-06, "loss": 0.0, "step": 11949 }, { "epoch": 2.76, "learning_rate": 3.971188660349743e-06, "loss": 0.0, "step": 11950 }, { "epoch": 2.76, "learning_rate": 3.967336876974039e-06, "loss": 0.0, "step": 11951 }, { "epoch": 2.76, "learning_rate": 3.963485093598336e-06, "loss": 0.0, "step": 11952 }, { "epoch": 2.76, "learning_rate": 3.959633310222633e-06, "loss": 0.0, "step": 11953 }, { "epoch": 2.76, "learning_rate": 3.95578152684693e-06, "loss": 0.0, "step": 11954 }, { "epoch": 2.76, "learning_rate": 3.9519297434712275e-06, "loss": 0.0, "step": 11955 }, { "epoch": 2.76, "learning_rate": 3.948077960095525e-06, "loss": 0.0, "step": 11956 }, { "epoch": 2.76, "learning_rate": 3.944226176719822e-06, "loss": 0.0, "step": 11957 }, { "epoch": 2.76, "learning_rate": 3.940374393344118e-06, "loss": 0.0, "step": 11958 }, { "epoch": 2.76, "learning_rate": 3.936522609968416e-06, "loss": 0.0, "step": 11959 }, { "epoch": 2.76, "learning_rate": 3.9326708265927125e-06, "loss": 0.0, "step": 11960 }, { "epoch": 2.76, "learning_rate": 3.928819043217009e-06, "loss": 0.0, "step": 11961 }, { "epoch": 2.76, "learning_rate": 3.924967259841307e-06, "loss": 0.0, "step": 11962 }, { "epoch": 2.76, "learning_rate": 3.921115476465604e-06, "loss": 0.0, "step": 11963 }, { "epoch": 2.76, "learning_rate": 3.917263693089901e-06, "loss": 0.0, "step": 11964 }, { "epoch": 2.77, "learning_rate": 3.913411909714198e-06, "loss": 0.0, "step": 11965 }, { "epoch": 2.77, "learning_rate": 3.909560126338495e-06, "loss": 0.0, "step": 11966 }, { "epoch": 2.77, "learning_rate": 3.9057083429627916e-06, "loss": 0.0, "step": 11967 }, { "epoch": 2.77, "learning_rate": 3.901856559587089e-06, "loss": 0.0, "step": 11968 }, { "epoch": 2.77, "learning_rate": 3.8980047762113866e-06, "loss": 0.0, "step": 11969 }, { "epoch": 2.77, "learning_rate": 3.894152992835683e-06, "loss": 0.0, "step": 11970 }, { "epoch": 2.77, "learning_rate": 3.890301209459981e-06, "loss": 0.0, "step": 11971 }, { "epoch": 2.77, "learning_rate": 3.886449426084277e-06, "loss": 0.0, "step": 11972 }, { "epoch": 2.77, "learning_rate": 3.882597642708574e-06, "loss": 0.0, "step": 11973 }, { "epoch": 2.77, "learning_rate": 3.8787458593328715e-06, "loss": 0.0, "step": 11974 }, { "epoch": 2.77, "learning_rate": 3.874894075957168e-06, "loss": 0.0, "step": 11975 }, { "epoch": 2.77, "learning_rate": 3.871042292581466e-06, "loss": 0.0, "step": 11976 }, { "epoch": 2.77, "learning_rate": 3.867190509205762e-06, "loss": 0.0, "step": 11977 }, { "epoch": 2.77, "learning_rate": 3.86333872583006e-06, "loss": 0.0, "step": 11978 }, { "epoch": 2.77, "learning_rate": 3.8594869424543564e-06, "loss": 0.0, "step": 11979 }, { "epoch": 2.77, "learning_rate": 3.855635159078654e-06, "loss": 0.0, "step": 11980 }, { "epoch": 2.77, "learning_rate": 3.851783375702951e-06, "loss": 0.0, "step": 11981 }, { "epoch": 2.77, "learning_rate": 3.847931592327247e-06, "loss": 0.0, "step": 11982 }, { "epoch": 2.77, "learning_rate": 3.844079808951545e-06, "loss": 0.0, "step": 11983 }, { "epoch": 2.77, "learning_rate": 3.840228025575842e-06, "loss": 0.0, "step": 11984 }, { "epoch": 2.77, "learning_rate": 3.836376242200139e-06, "loss": 0.0, "step": 11985 }, { "epoch": 2.77, "learning_rate": 3.8325244588244355e-06, "loss": 0.0, "step": 11986 }, { "epoch": 2.77, "learning_rate": 3.828672675448733e-06, "loss": 0.0, "step": 11987 }, { "epoch": 2.77, "learning_rate": 3.82482089207303e-06, "loss": 0.0, "step": 11988 }, { "epoch": 2.77, "learning_rate": 3.820969108697327e-06, "loss": 0.0, "step": 11989 }, { "epoch": 2.77, "learning_rate": 3.817117325321625e-06, "loss": 0.0, "step": 11990 }, { "epoch": 2.77, "learning_rate": 3.8132655419459213e-06, "loss": 0.0, "step": 11991 }, { "epoch": 2.77, "learning_rate": 3.8094137585702184e-06, "loss": 0.0, "step": 11992 }, { "epoch": 2.77, "learning_rate": 3.8055619751945155e-06, "loss": 0.0, "step": 11993 }, { "epoch": 2.77, "learning_rate": 3.801710191818812e-06, "loss": 0.0, "step": 11994 }, { "epoch": 2.77, "learning_rate": 3.797858408443109e-06, "loss": 0.0, "step": 11995 }, { "epoch": 2.77, "learning_rate": 3.7940066250674063e-06, "loss": 0.0, "step": 11996 }, { "epoch": 2.77, "learning_rate": 3.7901548416917038e-06, "loss": 0.0, "step": 11997 }, { "epoch": 2.77, "learning_rate": 3.786303058316001e-06, "loss": 0.0, "step": 11998 }, { "epoch": 2.77, "learning_rate": 3.7824512749402975e-06, "loss": 0.0, "step": 11999 }, { "epoch": 2.77, "learning_rate": 3.7785994915645945e-06, "loss": 0.0, "step": 12000 }, { "epoch": 2.77, "learning_rate": 3.7747477081888916e-06, "loss": 0.0, "step": 12001 }, { "epoch": 2.77, "learning_rate": 3.7708959248131887e-06, "loss": 0.0, "step": 12002 }, { "epoch": 2.77, "learning_rate": 3.7670441414374853e-06, "loss": 0.0, "step": 12003 }, { "epoch": 2.77, "learning_rate": 3.7631923580617833e-06, "loss": 0.0, "step": 12004 }, { "epoch": 2.77, "learning_rate": 3.75934057468608e-06, "loss": 0.0001, "step": 12005 }, { "epoch": 2.77, "learning_rate": 3.755488791310377e-06, "loss": 0.0, "step": 12006 }, { "epoch": 2.77, "learning_rate": 3.751637007934674e-06, "loss": 0.0, "step": 12007 }, { "epoch": 2.78, "learning_rate": 3.7477852245589707e-06, "loss": 0.0001, "step": 12008 }, { "epoch": 2.78, "learning_rate": 3.7439334411832678e-06, "loss": 0.0, "step": 12009 }, { "epoch": 2.78, "learning_rate": 3.740081657807565e-06, "loss": 0.0, "step": 12010 }, { "epoch": 2.78, "learning_rate": 3.7362298744318624e-06, "loss": 0.0, "step": 12011 }, { "epoch": 2.78, "learning_rate": 3.7323780910561594e-06, "loss": 0.0, "step": 12012 }, { "epoch": 2.78, "learning_rate": 3.7285263076804565e-06, "loss": 0.0, "step": 12013 }, { "epoch": 2.78, "learning_rate": 3.724674524304753e-06, "loss": 0.0, "step": 12014 }, { "epoch": 2.78, "learning_rate": 3.7208227409290502e-06, "loss": 0.0, "step": 12015 }, { "epoch": 2.78, "learning_rate": 3.7169709575533473e-06, "loss": 0.0, "step": 12016 }, { "epoch": 2.78, "learning_rate": 3.713119174177644e-06, "loss": 0.0, "step": 12017 }, { "epoch": 2.78, "learning_rate": 3.709267390801942e-06, "loss": 0.0, "step": 12018 }, { "epoch": 2.78, "learning_rate": 3.705415607426239e-06, "loss": 0.0, "step": 12019 }, { "epoch": 2.78, "learning_rate": 3.7015638240505356e-06, "loss": 0.0, "step": 12020 }, { "epoch": 2.78, "learning_rate": 3.6977120406748327e-06, "loss": 0.0, "step": 12021 }, { "epoch": 2.78, "learning_rate": 3.6938602572991297e-06, "loss": 0.0, "step": 12022 }, { "epoch": 2.78, "learning_rate": 3.6900084739234264e-06, "loss": 0.0, "step": 12023 }, { "epoch": 2.78, "learning_rate": 3.6861566905477235e-06, "loss": 0.0, "step": 12024 }, { "epoch": 2.78, "learning_rate": 3.682304907172021e-06, "loss": 0.0, "step": 12025 }, { "epoch": 2.78, "learning_rate": 3.678453123796318e-06, "loss": 0.0, "step": 12026 }, { "epoch": 2.78, "learning_rate": 3.674601340420615e-06, "loss": 0.0, "step": 12027 }, { "epoch": 2.78, "learning_rate": 3.670749557044912e-06, "loss": 0.0, "step": 12028 }, { "epoch": 2.78, "learning_rate": 3.666897773669209e-06, "loss": 0.0, "step": 12029 }, { "epoch": 2.78, "learning_rate": 3.663045990293506e-06, "loss": 0.0, "step": 12030 }, { "epoch": 2.78, "learning_rate": 3.659194206917803e-06, "loss": 0.0, "step": 12031 }, { "epoch": 2.78, "learning_rate": 3.6553424235421005e-06, "loss": 0.0, "step": 12032 }, { "epoch": 2.78, "learning_rate": 3.6514906401663975e-06, "loss": 0.0, "step": 12033 }, { "epoch": 2.78, "learning_rate": 3.647638856790694e-06, "loss": 0.0, "step": 12034 }, { "epoch": 2.78, "learning_rate": 3.6437870734149913e-06, "loss": 0.0, "step": 12035 }, { "epoch": 2.78, "learning_rate": 3.6399352900392883e-06, "loss": 0.0, "step": 12036 }, { "epoch": 2.78, "learning_rate": 3.6360835066635854e-06, "loss": 0.0, "step": 12037 }, { "epoch": 2.78, "learning_rate": 3.632231723287882e-06, "loss": 0.0, "step": 12038 }, { "epoch": 2.78, "learning_rate": 3.62837993991218e-06, "loss": 0.0, "step": 12039 }, { "epoch": 2.78, "learning_rate": 3.6245281565364766e-06, "loss": 0.0, "step": 12040 }, { "epoch": 2.78, "learning_rate": 3.6206763731607737e-06, "loss": 0.0, "step": 12041 }, { "epoch": 2.78, "learning_rate": 3.6168245897850708e-06, "loss": 0.0, "step": 12042 }, { "epoch": 2.78, "learning_rate": 3.612972806409368e-06, "loss": 0.0, "step": 12043 }, { "epoch": 2.78, "learning_rate": 3.6091210230336645e-06, "loss": 0.0, "step": 12044 }, { "epoch": 2.78, "learning_rate": 3.6052692396579616e-06, "loss": 0.0, "step": 12045 }, { "epoch": 2.78, "learning_rate": 3.601417456282259e-06, "loss": 0.0, "step": 12046 }, { "epoch": 2.78, "learning_rate": 3.597565672906556e-06, "loss": 0.0, "step": 12047 }, { "epoch": 2.78, "learning_rate": 3.5937138895308532e-06, "loss": 0.0, "step": 12048 }, { "epoch": 2.78, "learning_rate": 3.58986210615515e-06, "loss": 0.0, "step": 12049 }, { "epoch": 2.78, "learning_rate": 3.586010322779447e-06, "loss": 0.0, "step": 12050 }, { "epoch": 2.79, "learning_rate": 3.582158539403744e-06, "loss": 0.0, "step": 12051 }, { "epoch": 2.79, "learning_rate": 3.578306756028041e-06, "loss": 0.0, "step": 12052 }, { "epoch": 2.79, "learning_rate": 3.5744549726523386e-06, "loss": 0.0, "step": 12053 }, { "epoch": 2.79, "learning_rate": 3.5706031892766357e-06, "loss": 0.0, "step": 12054 }, { "epoch": 2.79, "learning_rate": 3.5667514059009323e-06, "loss": 0.0, "step": 12055 }, { "epoch": 2.79, "learning_rate": 3.5628996225252294e-06, "loss": 0.0, "step": 12056 }, { "epoch": 2.79, "learning_rate": 3.5590478391495264e-06, "loss": 0.0, "step": 12057 }, { "epoch": 2.79, "learning_rate": 3.555196055773823e-06, "loss": 0.0, "step": 12058 }, { "epoch": 2.79, "learning_rate": 3.55134427239812e-06, "loss": 0.0, "step": 12059 }, { "epoch": 2.79, "learning_rate": 3.5474924890224177e-06, "loss": 0.0, "step": 12060 }, { "epoch": 2.79, "learning_rate": 3.5436407056467147e-06, "loss": 0.0001, "step": 12061 }, { "epoch": 2.79, "learning_rate": 3.539788922271012e-06, "loss": 0.0, "step": 12062 }, { "epoch": 2.79, "learning_rate": 3.535937138895309e-06, "loss": 0.0, "step": 12063 }, { "epoch": 2.79, "learning_rate": 3.5320853555196055e-06, "loss": 0.0, "step": 12064 }, { "epoch": 2.79, "learning_rate": 3.5282335721439026e-06, "loss": 0.0, "step": 12065 }, { "epoch": 2.79, "learning_rate": 3.5243817887681997e-06, "loss": 0.0, "step": 12066 }, { "epoch": 2.79, "learning_rate": 3.520530005392497e-06, "loss": 0.0, "step": 12067 }, { "epoch": 2.79, "learning_rate": 3.5166782220167943e-06, "loss": 0.0, "step": 12068 }, { "epoch": 2.79, "learning_rate": 3.5128264386410913e-06, "loss": 0.0, "step": 12069 }, { "epoch": 2.79, "learning_rate": 3.508974655265388e-06, "loss": 0.0, "step": 12070 }, { "epoch": 2.79, "learning_rate": 3.505122871889685e-06, "loss": 0.0, "step": 12071 }, { "epoch": 2.79, "learning_rate": 3.501271088513982e-06, "loss": 0.0, "step": 12072 }, { "epoch": 2.79, "learning_rate": 3.4974193051382788e-06, "loss": 0.0, "step": 12073 }, { "epoch": 2.79, "learning_rate": 3.4935675217625767e-06, "loss": 0.0, "step": 12074 }, { "epoch": 2.79, "learning_rate": 3.4897157383868733e-06, "loss": 0.0, "step": 12075 }, { "epoch": 2.79, "learning_rate": 3.4858639550111704e-06, "loss": 0.0, "step": 12076 }, { "epoch": 2.79, "learning_rate": 3.4820121716354675e-06, "loss": 0.0, "step": 12077 }, { "epoch": 2.79, "learning_rate": 3.4781603882597646e-06, "loss": 0.0, "step": 12078 }, { "epoch": 2.79, "learning_rate": 3.474308604884061e-06, "loss": 0.0, "step": 12079 }, { "epoch": 2.79, "learning_rate": 3.4704568215083583e-06, "loss": 0.0, "step": 12080 }, { "epoch": 2.79, "learning_rate": 3.4666050381326558e-06, "loss": 0.0, "step": 12081 }, { "epoch": 2.79, "learning_rate": 3.462753254756953e-06, "loss": 0.0, "step": 12082 }, { "epoch": 2.79, "learning_rate": 3.45890147138125e-06, "loss": 0.0, "step": 12083 }, { "epoch": 2.79, "learning_rate": 3.4550496880055466e-06, "loss": 0.0, "step": 12084 }, { "epoch": 2.79, "learning_rate": 3.4511979046298436e-06, "loss": 0.0001, "step": 12085 }, { "epoch": 2.79, "learning_rate": 3.4473461212541407e-06, "loss": 0.0, "step": 12086 }, { "epoch": 2.79, "learning_rate": 3.443494337878438e-06, "loss": 0.0, "step": 12087 }, { "epoch": 2.79, "learning_rate": 3.4396425545027353e-06, "loss": 0.0, "step": 12088 }, { "epoch": 2.79, "learning_rate": 3.4357907711270324e-06, "loss": 0.0, "step": 12089 }, { "epoch": 2.79, "learning_rate": 3.431938987751329e-06, "loss": 0.0, "step": 12090 }, { "epoch": 2.79, "learning_rate": 3.428087204375626e-06, "loss": 0.0, "step": 12091 }, { "epoch": 2.79, "learning_rate": 3.424235420999923e-06, "loss": 0.0, "step": 12092 }, { "epoch": 2.79, "learning_rate": 3.42038363762422e-06, "loss": 0.0, "step": 12093 }, { "epoch": 2.8, "learning_rate": 3.416531854248517e-06, "loss": 0.0, "step": 12094 }, { "epoch": 2.8, "learning_rate": 3.412680070872815e-06, "loss": 0.0, "step": 12095 }, { "epoch": 2.8, "learning_rate": 3.4088282874971115e-06, "loss": 0.0, "step": 12096 }, { "epoch": 2.8, "learning_rate": 3.4049765041214085e-06, "loss": 0.0, "step": 12097 }, { "epoch": 2.8, "learning_rate": 3.4011247207457056e-06, "loss": 0.0, "step": 12098 }, { "epoch": 2.8, "learning_rate": 3.3972729373700022e-06, "loss": 0.0, "step": 12099 }, { "epoch": 2.8, "learning_rate": 3.3934211539942993e-06, "loss": 0.0, "step": 12100 }, { "epoch": 2.8, "learning_rate": 3.3895693706185964e-06, "loss": 0.0, "step": 12101 }, { "epoch": 2.8, "learning_rate": 3.385717587242894e-06, "loss": 0.0, "step": 12102 }, { "epoch": 2.8, "learning_rate": 3.381865803867191e-06, "loss": 0.0, "step": 12103 }, { "epoch": 2.8, "learning_rate": 3.378014020491488e-06, "loss": 0.0002, "step": 12104 }, { "epoch": 2.8, "learning_rate": 3.3741622371157847e-06, "loss": 0.0, "step": 12105 }, { "epoch": 2.8, "learning_rate": 3.3703104537400818e-06, "loss": 0.0, "step": 12106 }, { "epoch": 2.8, "learning_rate": 3.366458670364379e-06, "loss": 0.0, "step": 12107 }, { "epoch": 2.8, "learning_rate": 3.3626068869886755e-06, "loss": 0.0, "step": 12108 }, { "epoch": 2.8, "learning_rate": 3.3587551036129734e-06, "loss": 0.0, "step": 12109 }, { "epoch": 2.8, "learning_rate": 3.35490332023727e-06, "loss": 0.0, "step": 12110 }, { "epoch": 2.8, "learning_rate": 3.351051536861567e-06, "loss": 0.0, "step": 12111 }, { "epoch": 2.8, "learning_rate": 3.347199753485864e-06, "loss": 0.0, "step": 12112 }, { "epoch": 2.8, "learning_rate": 3.3433479701101613e-06, "loss": 0.0, "step": 12113 }, { "epoch": 2.8, "learning_rate": 3.339496186734458e-06, "loss": 0.0, "step": 12114 }, { "epoch": 2.8, "learning_rate": 3.335644403358755e-06, "loss": 0.0, "step": 12115 }, { "epoch": 2.8, "learning_rate": 3.3317926199830525e-06, "loss": 0.0, "step": 12116 }, { "epoch": 2.8, "learning_rate": 3.3279408366073496e-06, "loss": 0.0, "step": 12117 }, { "epoch": 2.8, "learning_rate": 3.3240890532316466e-06, "loss": 0.0, "step": 12118 }, { "epoch": 2.8, "learning_rate": 3.3202372698559433e-06, "loss": 0.0, "step": 12119 }, { "epoch": 2.8, "learning_rate": 3.3163854864802404e-06, "loss": 0.0, "step": 12120 }, { "epoch": 2.8, "learning_rate": 3.3125337031045374e-06, "loss": 0.0, "step": 12121 }, { "epoch": 2.8, "learning_rate": 3.3086819197288345e-06, "loss": 0.0, "step": 12122 }, { "epoch": 2.8, "learning_rate": 3.304830136353132e-06, "loss": 0.0, "step": 12123 }, { "epoch": 2.8, "learning_rate": 3.300978352977429e-06, "loss": 0.0, "step": 12124 }, { "epoch": 2.8, "learning_rate": 3.2971265696017257e-06, "loss": 0.0, "step": 12125 }, { "epoch": 2.8, "learning_rate": 3.293274786226023e-06, "loss": 0.0, "step": 12126 }, { "epoch": 2.8, "learning_rate": 3.28942300285032e-06, "loss": 0.0, "step": 12127 }, { "epoch": 2.8, "learning_rate": 3.285571219474617e-06, "loss": 0.0, "step": 12128 }, { "epoch": 2.8, "learning_rate": 3.2817194360989136e-06, "loss": 0.0, "step": 12129 }, { "epoch": 2.8, "learning_rate": 3.2778676527232115e-06, "loss": 0.0, "step": 12130 }, { "epoch": 2.8, "learning_rate": 3.274015869347508e-06, "loss": 0.0, "step": 12131 }, { "epoch": 2.8, "learning_rate": 3.2701640859718052e-06, "loss": 0.0, "step": 12132 }, { "epoch": 2.8, "learning_rate": 3.2663123025961023e-06, "loss": 0.0, "step": 12133 }, { "epoch": 2.8, "learning_rate": 3.262460519220399e-06, "loss": 0.0, "step": 12134 }, { "epoch": 2.8, "learning_rate": 3.258608735844696e-06, "loss": 0.0, "step": 12135 }, { "epoch": 2.8, "learning_rate": 3.254756952468993e-06, "loss": 0.0, "step": 12136 }, { "epoch": 2.8, "learning_rate": 3.2509051690932906e-06, "loss": 0.0, "step": 12137 }, { "epoch": 2.81, "learning_rate": 3.2470533857175877e-06, "loss": 0.0, "step": 12138 }, { "epoch": 2.81, "learning_rate": 3.2432016023418848e-06, "loss": 0.0, "step": 12139 }, { "epoch": 2.81, "learning_rate": 3.2393498189661814e-06, "loss": 0.0, "step": 12140 }, { "epoch": 2.81, "learning_rate": 3.2354980355904785e-06, "loss": 0.0, "step": 12141 }, { "epoch": 2.81, "learning_rate": 3.2316462522147755e-06, "loss": 0.0, "step": 12142 }, { "epoch": 2.81, "learning_rate": 3.227794468839072e-06, "loss": 0.0, "step": 12143 }, { "epoch": 2.81, "learning_rate": 3.22394268546337e-06, "loss": 0.0, "step": 12144 }, { "epoch": 2.81, "learning_rate": 3.220090902087667e-06, "loss": 0.0, "step": 12145 }, { "epoch": 2.81, "learning_rate": 3.216239118711964e-06, "loss": 0.0, "step": 12146 }, { "epoch": 2.81, "learning_rate": 3.212387335336261e-06, "loss": 0.0, "step": 12147 }, { "epoch": 2.81, "learning_rate": 3.208535551960558e-06, "loss": 0.0, "step": 12148 }, { "epoch": 2.81, "learning_rate": 3.2046837685848546e-06, "loss": 0.0, "step": 12149 }, { "epoch": 2.81, "learning_rate": 3.2008319852091517e-06, "loss": 0.0, "step": 12150 }, { "epoch": 2.81, "learning_rate": 3.196980201833449e-06, "loss": 0.0, "step": 12151 }, { "epoch": 2.81, "learning_rate": 3.1931284184577463e-06, "loss": 0.0, "step": 12152 }, { "epoch": 2.81, "learning_rate": 3.1892766350820434e-06, "loss": 0.0, "step": 12153 }, { "epoch": 2.81, "learning_rate": 3.1854248517063404e-06, "loss": 0.0, "step": 12154 }, { "epoch": 2.81, "learning_rate": 3.181573068330637e-06, "loss": 0.0, "step": 12155 }, { "epoch": 2.81, "learning_rate": 3.177721284954934e-06, "loss": 0.0, "step": 12156 }, { "epoch": 2.81, "learning_rate": 3.1738695015792312e-06, "loss": 0.0, "step": 12157 }, { "epoch": 2.81, "learning_rate": 3.1700177182035287e-06, "loss": 0.0, "step": 12158 }, { "epoch": 2.81, "learning_rate": 3.166165934827826e-06, "loss": 0.0, "step": 12159 }, { "epoch": 2.81, "learning_rate": 3.1623141514521224e-06, "loss": 0.0, "step": 12160 }, { "epoch": 2.81, "learning_rate": 3.1584623680764195e-06, "loss": 0.0, "step": 12161 }, { "epoch": 2.81, "learning_rate": 3.1546105847007166e-06, "loss": 0.0, "step": 12162 }, { "epoch": 2.81, "learning_rate": 3.1507588013250137e-06, "loss": 0.0, "step": 12163 }, { "epoch": 2.81, "learning_rate": 3.1469070179493103e-06, "loss": 0.0, "step": 12164 }, { "epoch": 2.81, "learning_rate": 3.1430552345736082e-06, "loss": 0.0002, "step": 12165 }, { "epoch": 2.81, "learning_rate": 3.139203451197905e-06, "loss": 0.0, "step": 12166 }, { "epoch": 2.81, "learning_rate": 3.135351667822202e-06, "loss": 0.0, "step": 12167 }, { "epoch": 2.81, "learning_rate": 3.131499884446499e-06, "loss": 0.0, "step": 12168 }, { "epoch": 2.81, "learning_rate": 3.1276481010707957e-06, "loss": 0.0, "step": 12169 }, { "epoch": 2.81, "learning_rate": 3.1237963176950927e-06, "loss": 0.0, "step": 12170 }, { "epoch": 2.81, "learning_rate": 3.1199445343193902e-06, "loss": 0.0, "step": 12171 }, { "epoch": 2.81, "learning_rate": 3.116092750943687e-06, "loss": 0.0, "step": 12172 }, { "epoch": 2.81, "learning_rate": 3.112240967567984e-06, "loss": 0.0, "step": 12173 }, { "epoch": 2.81, "learning_rate": 3.1083891841922815e-06, "loss": 0.0, "step": 12174 }, { "epoch": 2.81, "learning_rate": 3.104537400816578e-06, "loss": 0.0, "step": 12175 }, { "epoch": 2.81, "learning_rate": 3.100685617440875e-06, "loss": 0.0, "step": 12176 }, { "epoch": 2.81, "learning_rate": 3.0968338340651723e-06, "loss": 0.0, "step": 12177 }, { "epoch": 2.81, "learning_rate": 3.0929820506894693e-06, "loss": 0.0, "step": 12178 }, { "epoch": 2.81, "learning_rate": 3.0891302673137664e-06, "loss": 0.0, "step": 12179 }, { "epoch": 2.81, "learning_rate": 3.0852784839380635e-06, "loss": 0.0, "step": 12180 }, { "epoch": 2.82, "learning_rate": 3.0814267005623606e-06, "loss": 0.0, "step": 12181 }, { "epoch": 2.82, "learning_rate": 3.0775749171866576e-06, "loss": 0.0, "step": 12182 }, { "epoch": 2.82, "learning_rate": 3.0737231338109547e-06, "loss": 0.0, "step": 12183 }, { "epoch": 2.82, "learning_rate": 3.0698713504352513e-06, "loss": 0.0, "step": 12184 }, { "epoch": 2.82, "learning_rate": 3.066019567059549e-06, "loss": 0.0, "step": 12185 }, { "epoch": 2.82, "learning_rate": 3.062167783683846e-06, "loss": 0.0, "step": 12186 }, { "epoch": 2.82, "learning_rate": 3.0583160003081426e-06, "loss": 0.0, "step": 12187 }, { "epoch": 2.82, "learning_rate": 3.05446421693244e-06, "loss": 0.0, "step": 12188 }, { "epoch": 2.82, "learning_rate": 3.050612433556737e-06, "loss": 0.0, "step": 12189 }, { "epoch": 2.82, "learning_rate": 3.0467606501810338e-06, "loss": 0.0, "step": 12190 }, { "epoch": 2.82, "learning_rate": 3.042908866805331e-06, "loss": 0.0, "step": 12191 }, { "epoch": 2.82, "learning_rate": 3.0390570834296284e-06, "loss": 0.0, "step": 12192 }, { "epoch": 2.82, "learning_rate": 3.035205300053925e-06, "loss": 0.0, "step": 12193 }, { "epoch": 2.82, "learning_rate": 3.031353516678222e-06, "loss": 0.0, "step": 12194 }, { "epoch": 2.82, "learning_rate": 3.027501733302519e-06, "loss": 0.0, "step": 12195 }, { "epoch": 2.82, "learning_rate": 3.0236499499268162e-06, "loss": 0.0, "step": 12196 }, { "epoch": 2.82, "learning_rate": 3.0197981665511133e-06, "loss": 0.0, "step": 12197 }, { "epoch": 2.82, "learning_rate": 3.0159463831754104e-06, "loss": 0.0, "step": 12198 }, { "epoch": 2.82, "learning_rate": 3.0120945997997074e-06, "loss": 0.0, "step": 12199 }, { "epoch": 2.82, "learning_rate": 3.0082428164240045e-06, "loss": 0.0, "step": 12200 }, { "epoch": 2.82, "learning_rate": 3.0043910330483016e-06, "loss": 0.0005, "step": 12201 }, { "epoch": 2.82, "learning_rate": 3.0005392496725987e-06, "loss": 0.0, "step": 12202 }, { "epoch": 2.82, "learning_rate": 2.9966874662968957e-06, "loss": 0.0, "step": 12203 }, { "epoch": 2.82, "learning_rate": 2.9928356829211924e-06, "loss": 0.0, "step": 12204 }, { "epoch": 2.82, "learning_rate": 2.9889838995454895e-06, "loss": 0.0, "step": 12205 }, { "epoch": 2.82, "learning_rate": 2.985132116169787e-06, "loss": 0.0, "step": 12206 }, { "epoch": 2.82, "learning_rate": 2.9812803327940836e-06, "loss": 0.0, "step": 12207 }, { "epoch": 2.82, "learning_rate": 2.9774285494183807e-06, "loss": 0.0, "step": 12208 }, { "epoch": 2.82, "learning_rate": 2.973576766042678e-06, "loss": 0.0, "step": 12209 }, { "epoch": 2.82, "learning_rate": 2.969724982666975e-06, "loss": 0.0, "step": 12210 }, { "epoch": 2.82, "learning_rate": 2.965873199291272e-06, "loss": 0.0, "step": 12211 }, { "epoch": 2.82, "learning_rate": 2.962021415915569e-06, "loss": 0.0, "step": 12212 }, { "epoch": 2.82, "learning_rate": 2.958169632539866e-06, "loss": 0.0, "step": 12213 }, { "epoch": 2.82, "learning_rate": 2.954317849164163e-06, "loss": 0.0, "step": 12214 }, { "epoch": 2.82, "learning_rate": 2.95046606578846e-06, "loss": 0.0, "step": 12215 }, { "epoch": 2.82, "learning_rate": 2.9466142824127573e-06, "loss": 0.0, "step": 12216 }, { "epoch": 2.82, "learning_rate": 2.9427624990370543e-06, "loss": 0.0, "step": 12217 }, { "epoch": 2.82, "learning_rate": 2.9389107156613514e-06, "loss": 0.0, "step": 12218 }, { "epoch": 2.82, "learning_rate": 2.935058932285648e-06, "loss": 0.0, "step": 12219 }, { "epoch": 2.82, "learning_rate": 2.9312071489099456e-06, "loss": 0.0, "step": 12220 }, { "epoch": 2.82, "learning_rate": 2.9273553655342426e-06, "loss": 0.0, "step": 12221 }, { "epoch": 2.82, "learning_rate": 2.9235035821585393e-06, "loss": 0.0, "step": 12222 }, { "epoch": 2.82, "learning_rate": 2.9196517987828368e-06, "loss": 0.0, "step": 12223 }, { "epoch": 2.83, "learning_rate": 2.915800015407134e-06, "loss": 0.0, "step": 12224 }, { "epoch": 2.83, "learning_rate": 2.9119482320314305e-06, "loss": 0.0, "step": 12225 }, { "epoch": 2.83, "learning_rate": 2.9080964486557276e-06, "loss": 0.0, "step": 12226 }, { "epoch": 2.83, "learning_rate": 2.904244665280025e-06, "loss": 0.0, "step": 12227 }, { "epoch": 2.83, "learning_rate": 2.9003928819043217e-06, "loss": 0.0, "step": 12228 }, { "epoch": 2.83, "learning_rate": 2.896541098528619e-06, "loss": 0.0, "step": 12229 }, { "epoch": 2.83, "learning_rate": 2.8926893151529163e-06, "loss": 0.0, "step": 12230 }, { "epoch": 2.83, "learning_rate": 2.888837531777213e-06, "loss": 0.0, "step": 12231 }, { "epoch": 2.83, "learning_rate": 2.88498574840151e-06, "loss": 0.0, "step": 12232 }, { "epoch": 2.83, "learning_rate": 2.881133965025807e-06, "loss": 0.0, "step": 12233 }, { "epoch": 2.83, "learning_rate": 2.877282181650104e-06, "loss": 0.0, "step": 12234 }, { "epoch": 2.83, "learning_rate": 2.8734303982744012e-06, "loss": 0.0, "step": 12235 }, { "epoch": 2.83, "learning_rate": 2.8695786148986983e-06, "loss": 0.0, "step": 12236 }, { "epoch": 2.83, "learning_rate": 2.8657268315229954e-06, "loss": 0.0, "step": 12237 }, { "epoch": 2.83, "learning_rate": 2.8618750481472924e-06, "loss": 0.0, "step": 12238 }, { "epoch": 2.83, "learning_rate": 2.8580232647715895e-06, "loss": 0.0, "step": 12239 }, { "epoch": 2.83, "learning_rate": 2.854171481395886e-06, "loss": 0.0, "step": 12240 }, { "epoch": 2.83, "learning_rate": 2.8503196980201837e-06, "loss": 0.0, "step": 12241 }, { "epoch": 2.83, "learning_rate": 2.8464679146444803e-06, "loss": 0.0, "step": 12242 }, { "epoch": 2.83, "learning_rate": 2.8426161312687774e-06, "loss": 0.0, "step": 12243 }, { "epoch": 2.83, "learning_rate": 2.838764347893075e-06, "loss": 0.0, "step": 12244 }, { "epoch": 2.83, "learning_rate": 2.8349125645173715e-06, "loss": 0.0001, "step": 12245 }, { "epoch": 2.83, "learning_rate": 2.8310607811416686e-06, "loss": 0.0, "step": 12246 }, { "epoch": 2.83, "learning_rate": 2.8272089977659657e-06, "loss": 0.0, "step": 12247 }, { "epoch": 2.83, "learning_rate": 2.8233572143902628e-06, "loss": 0.0, "step": 12248 }, { "epoch": 2.83, "learning_rate": 2.81950543101456e-06, "loss": 0.0, "step": 12249 }, { "epoch": 2.83, "learning_rate": 2.815653647638857e-06, "loss": 0.0, "step": 12250 }, { "epoch": 2.83, "learning_rate": 2.811801864263154e-06, "loss": 0.0, "step": 12251 }, { "epoch": 2.83, "learning_rate": 2.807950080887451e-06, "loss": 0.0, "step": 12252 }, { "epoch": 2.83, "learning_rate": 2.804098297511748e-06, "loss": 0.0, "step": 12253 }, { "epoch": 2.83, "learning_rate": 2.8002465141360448e-06, "loss": 0.0, "step": 12254 }, { "epoch": 2.83, "learning_rate": 2.7963947307603423e-06, "loss": 0.0, "step": 12255 }, { "epoch": 2.83, "learning_rate": 2.7925429473846393e-06, "loss": 0.0, "step": 12256 }, { "epoch": 2.83, "learning_rate": 2.788691164008936e-06, "loss": 0.0, "step": 12257 }, { "epoch": 2.83, "learning_rate": 2.7848393806332335e-06, "loss": 0.0001, "step": 12258 }, { "epoch": 2.83, "learning_rate": 2.7809875972575306e-06, "loss": 0.0, "step": 12259 }, { "epoch": 2.83, "learning_rate": 2.777135813881827e-06, "loss": 0.0, "step": 12260 }, { "epoch": 2.83, "learning_rate": 2.7732840305061243e-06, "loss": 0.0, "step": 12261 }, { "epoch": 2.83, "learning_rate": 2.7694322471304218e-06, "loss": 0.0, "step": 12262 }, { "epoch": 2.83, "learning_rate": 2.7655804637547184e-06, "loss": 0.0, "step": 12263 }, { "epoch": 2.83, "learning_rate": 2.7617286803790155e-06, "loss": 0.0, "step": 12264 }, { "epoch": 2.83, "learning_rate": 2.757876897003313e-06, "loss": 0.0, "step": 12265 }, { "epoch": 2.83, "learning_rate": 2.7540251136276097e-06, "loss": 0.0, "step": 12266 }, { "epoch": 2.83, "learning_rate": 2.7501733302519067e-06, "loss": 0.0, "step": 12267 }, { "epoch": 2.84, "learning_rate": 2.746321546876204e-06, "loss": 0.0, "step": 12268 }, { "epoch": 2.84, "learning_rate": 2.742469763500501e-06, "loss": 0.0, "step": 12269 }, { "epoch": 2.84, "learning_rate": 2.738617980124798e-06, "loss": 0.0, "step": 12270 }, { "epoch": 2.84, "learning_rate": 2.734766196749095e-06, "loss": 0.0, "step": 12271 }, { "epoch": 2.84, "learning_rate": 2.730914413373392e-06, "loss": 0.0, "step": 12272 }, { "epoch": 2.84, "learning_rate": 2.727062629997689e-06, "loss": 0.0, "step": 12273 }, { "epoch": 2.84, "learning_rate": 2.7232108466219862e-06, "loss": 0.0, "step": 12274 }, { "epoch": 2.84, "learning_rate": 2.719359063246283e-06, "loss": 0.0, "step": 12275 }, { "epoch": 2.84, "learning_rate": 2.7155072798705804e-06, "loss": 0.0, "step": 12276 }, { "epoch": 2.84, "learning_rate": 2.7116554964948775e-06, "loss": 0.0, "step": 12277 }, { "epoch": 2.84, "learning_rate": 2.707803713119174e-06, "loss": 0.0, "step": 12278 }, { "epoch": 2.84, "learning_rate": 2.7039519297434716e-06, "loss": 0.0, "step": 12279 }, { "epoch": 2.84, "learning_rate": 2.7001001463677683e-06, "loss": 0.0, "step": 12280 }, { "epoch": 2.84, "learning_rate": 2.6962483629920653e-06, "loss": 0.0, "step": 12281 }, { "epoch": 2.84, "learning_rate": 2.6923965796163624e-06, "loss": 0.0, "step": 12282 }, { "epoch": 2.84, "learning_rate": 2.6885447962406595e-06, "loss": 0.0, "step": 12283 }, { "epoch": 2.84, "learning_rate": 2.6846930128649565e-06, "loss": 0.0, "step": 12284 }, { "epoch": 2.84, "learning_rate": 2.6808412294892536e-06, "loss": 0.0, "step": 12285 }, { "epoch": 2.84, "learning_rate": 2.6769894461135507e-06, "loss": 0.0, "step": 12286 }, { "epoch": 2.84, "learning_rate": 2.6731376627378478e-06, "loss": 0.0, "step": 12287 }, { "epoch": 2.84, "learning_rate": 2.669285879362145e-06, "loss": 0.0, "step": 12288 }, { "epoch": 2.84, "learning_rate": 2.665434095986442e-06, "loss": 0.0, "step": 12289 }, { "epoch": 2.84, "learning_rate": 2.661582312610739e-06, "loss": 0.0, "step": 12290 }, { "epoch": 2.84, "learning_rate": 2.657730529235036e-06, "loss": 0.0, "step": 12291 }, { "epoch": 2.84, "learning_rate": 2.6538787458593327e-06, "loss": 0.0, "step": 12292 }, { "epoch": 2.84, "learning_rate": 2.65002696248363e-06, "loss": 0.0, "step": 12293 }, { "epoch": 2.84, "learning_rate": 2.6461751791079273e-06, "loss": 0.0, "step": 12294 }, { "epoch": 2.84, "learning_rate": 2.642323395732224e-06, "loss": 0.0, "step": 12295 }, { "epoch": 2.84, "learning_rate": 2.638471612356521e-06, "loss": 0.0, "step": 12296 }, { "epoch": 2.84, "learning_rate": 2.6346198289808185e-06, "loss": 0.0, "step": 12297 }, { "epoch": 2.84, "learning_rate": 2.630768045605115e-06, "loss": 0.0, "step": 12298 }, { "epoch": 2.84, "learning_rate": 2.6269162622294122e-06, "loss": 0.0, "step": 12299 }, { "epoch": 2.84, "learning_rate": 2.6230644788537097e-06, "loss": 0.0, "step": 12300 }, { "epoch": 2.84, "learning_rate": 2.6192126954780064e-06, "loss": 0.0, "step": 12301 }, { "epoch": 2.84, "learning_rate": 2.6153609121023034e-06, "loss": 0.0, "step": 12302 }, { "epoch": 2.84, "learning_rate": 2.6115091287266005e-06, "loss": 0.0, "step": 12303 }, { "epoch": 2.84, "learning_rate": 2.6076573453508976e-06, "loss": 0.0, "step": 12304 }, { "epoch": 2.84, "learning_rate": 2.6038055619751947e-06, "loss": 0.0, "step": 12305 }, { "epoch": 2.84, "learning_rate": 2.5999537785994917e-06, "loss": 0.0, "step": 12306 }, { "epoch": 2.84, "learning_rate": 2.596101995223789e-06, "loss": 0.0, "step": 12307 }, { "epoch": 2.84, "learning_rate": 2.592250211848086e-06, "loss": 0.0, "step": 12308 }, { "epoch": 2.84, "learning_rate": 2.588398428472383e-06, "loss": 0.0, "step": 12309 }, { "epoch": 2.84, "learning_rate": 2.5845466450966796e-06, "loss": 0.0, "step": 12310 }, { "epoch": 2.85, "learning_rate": 2.580694861720977e-06, "loss": 0.0, "step": 12311 }, { "epoch": 2.85, "learning_rate": 2.576843078345274e-06, "loss": 0.0, "step": 12312 }, { "epoch": 2.85, "learning_rate": 2.572991294969571e-06, "loss": 0.0, "step": 12313 }, { "epoch": 2.85, "learning_rate": 2.5691395115938683e-06, "loss": 0.0, "step": 12314 }, { "epoch": 2.85, "learning_rate": 2.5652877282181654e-06, "loss": 0.0, "step": 12315 }, { "epoch": 2.85, "learning_rate": 2.561435944842462e-06, "loss": 0.0, "step": 12316 }, { "epoch": 2.85, "learning_rate": 2.557584161466759e-06, "loss": 0.0, "step": 12317 }, { "epoch": 2.85, "learning_rate": 2.553732378091056e-06, "loss": 0.0, "step": 12318 }, { "epoch": 2.85, "learning_rate": 2.5498805947153533e-06, "loss": 0.0, "step": 12319 }, { "epoch": 2.85, "learning_rate": 2.5460288113396503e-06, "loss": 0.0, "step": 12320 }, { "epoch": 2.85, "learning_rate": 2.5421770279639474e-06, "loss": 0.002, "step": 12321 }, { "epoch": 2.85, "learning_rate": 2.5383252445882445e-06, "loss": 0.0, "step": 12322 }, { "epoch": 2.85, "learning_rate": 2.5344734612125415e-06, "loss": 0.0, "step": 12323 }, { "epoch": 2.85, "learning_rate": 2.5306216778368386e-06, "loss": 0.0, "step": 12324 }, { "epoch": 2.85, "learning_rate": 2.5267698944611357e-06, "loss": 0.0, "step": 12325 }, { "epoch": 2.85, "learning_rate": 2.5229181110854328e-06, "loss": 0.0, "step": 12326 }, { "epoch": 2.85, "learning_rate": 2.5190663277097294e-06, "loss": 0.0, "step": 12327 }, { "epoch": 2.85, "learning_rate": 2.515214544334027e-06, "loss": 0.0, "step": 12328 }, { "epoch": 2.85, "learning_rate": 2.511362760958324e-06, "loss": 0.0, "step": 12329 }, { "epoch": 2.85, "learning_rate": 2.5075109775826206e-06, "loss": 0.0, "step": 12330 }, { "epoch": 2.85, "learning_rate": 2.5036591942069177e-06, "loss": 0.0, "step": 12331 }, { "epoch": 2.85, "learning_rate": 2.499807410831215e-06, "loss": 0.0, "step": 12332 }, { "epoch": 2.85, "learning_rate": 2.495955627455512e-06, "loss": 0.0, "step": 12333 }, { "epoch": 2.85, "learning_rate": 2.492103844079809e-06, "loss": 0.0, "step": 12334 }, { "epoch": 2.85, "learning_rate": 2.4882520607041064e-06, "loss": 0.0, "step": 12335 }, { "epoch": 2.85, "learning_rate": 2.484400277328403e-06, "loss": 0.0, "step": 12336 }, { "epoch": 2.85, "learning_rate": 2.4805484939527e-06, "loss": 0.0, "step": 12337 }, { "epoch": 2.85, "learning_rate": 2.4766967105769972e-06, "loss": 0.0, "step": 12338 }, { "epoch": 2.85, "learning_rate": 2.4728449272012943e-06, "loss": 0.0001, "step": 12339 }, { "epoch": 2.85, "learning_rate": 2.4689931438255914e-06, "loss": 0.0, "step": 12340 }, { "epoch": 2.85, "learning_rate": 2.4651413604498884e-06, "loss": 0.0, "step": 12341 }, { "epoch": 2.85, "learning_rate": 2.4612895770741855e-06, "loss": 0.0, "step": 12342 }, { "epoch": 2.85, "learning_rate": 2.4574377936984826e-06, "loss": 0.0, "step": 12343 }, { "epoch": 2.85, "learning_rate": 2.4535860103227797e-06, "loss": 0.0, "step": 12344 }, { "epoch": 2.85, "learning_rate": 2.4497342269470763e-06, "loss": 0.0, "step": 12345 }, { "epoch": 2.85, "learning_rate": 2.445882443571374e-06, "loss": 0.0, "step": 12346 }, { "epoch": 2.85, "learning_rate": 2.442030660195671e-06, "loss": 0.0, "step": 12347 }, { "epoch": 2.85, "learning_rate": 2.4381788768199675e-06, "loss": 0.0, "step": 12348 }, { "epoch": 2.85, "learning_rate": 2.434327093444265e-06, "loss": 0.0, "step": 12349 }, { "epoch": 2.85, "learning_rate": 2.430475310068562e-06, "loss": 0.0, "step": 12350 }, { "epoch": 2.85, "learning_rate": 2.4266235266928587e-06, "loss": 0.0, "step": 12351 }, { "epoch": 2.85, "learning_rate": 2.422771743317156e-06, "loss": 0.0, "step": 12352 }, { "epoch": 2.85, "learning_rate": 2.4189199599414533e-06, "loss": 0.0, "step": 12353 }, { "epoch": 2.86, "learning_rate": 2.41506817656575e-06, "loss": 0.0, "step": 12354 }, { "epoch": 2.86, "learning_rate": 2.411216393190047e-06, "loss": 0.0, "step": 12355 }, { "epoch": 2.86, "learning_rate": 2.407364609814344e-06, "loss": 0.0, "step": 12356 }, { "epoch": 2.86, "learning_rate": 2.403512826438641e-06, "loss": 0.0, "step": 12357 }, { "epoch": 2.86, "learning_rate": 2.3996610430629383e-06, "loss": 0.0, "step": 12358 }, { "epoch": 2.86, "learning_rate": 2.3958092596872353e-06, "loss": 0.0, "step": 12359 }, { "epoch": 2.86, "learning_rate": 2.3919574763115324e-06, "loss": 0.0, "step": 12360 }, { "epoch": 2.86, "learning_rate": 2.3881056929358295e-06, "loss": 0.0, "step": 12361 }, { "epoch": 2.86, "learning_rate": 2.3842539095601266e-06, "loss": 0.0, "step": 12362 }, { "epoch": 2.86, "learning_rate": 2.3804021261844236e-06, "loss": 0.0, "step": 12363 }, { "epoch": 2.86, "learning_rate": 2.3765503428087207e-06, "loss": 0.0, "step": 12364 }, { "epoch": 2.86, "learning_rate": 2.3726985594330173e-06, "loss": 0.0, "step": 12365 }, { "epoch": 2.86, "learning_rate": 2.3688467760573144e-06, "loss": 0.0, "step": 12366 }, { "epoch": 2.86, "learning_rate": 2.364994992681612e-06, "loss": 0.0, "step": 12367 }, { "epoch": 2.86, "learning_rate": 2.3611432093059086e-06, "loss": 0.0, "step": 12368 }, { "epoch": 2.86, "learning_rate": 2.3572914259302056e-06, "loss": 0.0, "step": 12369 }, { "epoch": 2.86, "learning_rate": 2.353439642554503e-06, "loss": 0.0, "step": 12370 }, { "epoch": 2.86, "learning_rate": 2.3495878591787998e-06, "loss": 0.0, "step": 12371 }, { "epoch": 2.86, "learning_rate": 2.345736075803097e-06, "loss": 0.0, "step": 12372 }, { "epoch": 2.86, "learning_rate": 2.341884292427394e-06, "loss": 0.0, "step": 12373 }, { "epoch": 2.86, "learning_rate": 2.338032509051691e-06, "loss": 0.0, "step": 12374 }, { "epoch": 2.86, "learning_rate": 2.334180725675988e-06, "loss": 0.0, "step": 12375 }, { "epoch": 2.86, "learning_rate": 2.330328942300285e-06, "loss": 0.0, "step": 12376 }, { "epoch": 2.86, "learning_rate": 2.3264771589245822e-06, "loss": 0.0, "step": 12377 }, { "epoch": 2.86, "learning_rate": 2.3226253755488793e-06, "loss": 0.0, "step": 12378 }, { "epoch": 2.86, "learning_rate": 2.3187735921731764e-06, "loss": 0.0, "step": 12379 }, { "epoch": 2.86, "learning_rate": 2.314921808797473e-06, "loss": 0.0, "step": 12380 }, { "epoch": 2.86, "learning_rate": 2.3110700254217705e-06, "loss": 0.0, "step": 12381 }, { "epoch": 2.86, "learning_rate": 2.3072182420460676e-06, "loss": 0.0, "step": 12382 }, { "epoch": 2.86, "learning_rate": 2.3033664586703642e-06, "loss": 0.0, "step": 12383 }, { "epoch": 2.86, "learning_rate": 2.2995146752946617e-06, "loss": 0.0, "step": 12384 }, { "epoch": 2.86, "learning_rate": 2.295662891918959e-06, "loss": 0.0, "step": 12385 }, { "epoch": 2.86, "learning_rate": 2.2918111085432555e-06, "loss": 0.0, "step": 12386 }, { "epoch": 2.86, "learning_rate": 2.2879593251675525e-06, "loss": 0.0, "step": 12387 }, { "epoch": 2.86, "learning_rate": 2.28410754179185e-06, "loss": 0.0, "step": 12388 }, { "epoch": 2.86, "learning_rate": 2.2802557584161467e-06, "loss": 0.0004, "step": 12389 }, { "epoch": 2.86, "learning_rate": 2.2764039750404438e-06, "loss": 0.0, "step": 12390 }, { "epoch": 2.86, "learning_rate": 2.272552191664741e-06, "loss": 0.0, "step": 12391 }, { "epoch": 2.86, "learning_rate": 2.268700408289038e-06, "loss": 0.0, "step": 12392 }, { "epoch": 2.86, "learning_rate": 2.264848624913335e-06, "loss": 0.0, "step": 12393 }, { "epoch": 2.86, "learning_rate": 2.260996841537632e-06, "loss": 0.0, "step": 12394 }, { "epoch": 2.86, "learning_rate": 2.257145058161929e-06, "loss": 0.0, "step": 12395 }, { "epoch": 2.86, "learning_rate": 2.253293274786226e-06, "loss": 0.0, "step": 12396 }, { "epoch": 2.87, "learning_rate": 2.2494414914105233e-06, "loss": 0.0, "step": 12397 }, { "epoch": 2.87, "learning_rate": 2.2455897080348203e-06, "loss": 0.0, "step": 12398 }, { "epoch": 2.87, "learning_rate": 2.2417379246591174e-06, "loss": 0.0, "step": 12399 }, { "epoch": 2.87, "learning_rate": 2.2378861412834145e-06, "loss": 0.0, "step": 12400 }, { "epoch": 2.87, "learning_rate": 2.234034357907711e-06, "loss": 0.0, "step": 12401 }, { "epoch": 2.87, "learning_rate": 2.2301825745320086e-06, "loss": 0.0002, "step": 12402 }, { "epoch": 2.87, "learning_rate": 2.2263307911563053e-06, "loss": 0.0, "step": 12403 }, { "epoch": 2.87, "learning_rate": 2.2224790077806024e-06, "loss": 0.0, "step": 12404 }, { "epoch": 2.87, "learning_rate": 2.2186272244049e-06, "loss": 0.0, "step": 12405 }, { "epoch": 2.87, "learning_rate": 2.2147754410291965e-06, "loss": 0.0, "step": 12406 }, { "epoch": 2.87, "learning_rate": 2.2109236576534936e-06, "loss": 0.0, "step": 12407 }, { "epoch": 2.87, "learning_rate": 2.2070718742777906e-06, "loss": 0.0, "step": 12408 }, { "epoch": 2.87, "learning_rate": 2.2032200909020877e-06, "loss": 0.0, "step": 12409 }, { "epoch": 2.87, "learning_rate": 2.199368307526385e-06, "loss": 0.0, "step": 12410 }, { "epoch": 2.87, "learning_rate": 2.195516524150682e-06, "loss": 0.0, "step": 12411 }, { "epoch": 2.87, "learning_rate": 2.191664740774979e-06, "loss": 0.0, "step": 12412 }, { "epoch": 2.87, "learning_rate": 2.187812957399276e-06, "loss": 0.0, "step": 12413 }, { "epoch": 2.87, "learning_rate": 2.183961174023573e-06, "loss": 0.0, "step": 12414 }, { "epoch": 2.87, "learning_rate": 2.1801093906478697e-06, "loss": 0.0001, "step": 12415 }, { "epoch": 2.87, "learning_rate": 2.1762576072721672e-06, "loss": 0.0, "step": 12416 }, { "epoch": 2.87, "learning_rate": 2.1724058238964643e-06, "loss": 0.0, "step": 12417 }, { "epoch": 2.87, "learning_rate": 2.168554040520761e-06, "loss": 0.0, "step": 12418 }, { "epoch": 2.87, "learning_rate": 2.1647022571450585e-06, "loss": 0.0, "step": 12419 }, { "epoch": 2.87, "learning_rate": 2.1608504737693555e-06, "loss": 0.0, "step": 12420 }, { "epoch": 2.87, "learning_rate": 2.156998690393652e-06, "loss": 0.0, "step": 12421 }, { "epoch": 2.87, "learning_rate": 2.1531469070179492e-06, "loss": 0.0, "step": 12422 }, { "epoch": 2.87, "learning_rate": 2.1492951236422467e-06, "loss": 0.0001, "step": 12423 }, { "epoch": 2.87, "learning_rate": 2.1454433402665434e-06, "loss": 0.0, "step": 12424 }, { "epoch": 2.87, "learning_rate": 2.1415915568908405e-06, "loss": 0.0, "step": 12425 }, { "epoch": 2.87, "learning_rate": 2.137739773515138e-06, "loss": 0.0, "step": 12426 }, { "epoch": 2.87, "learning_rate": 2.1338879901394346e-06, "loss": 0.0, "step": 12427 }, { "epoch": 2.87, "learning_rate": 2.1300362067637317e-06, "loss": 0.0, "step": 12428 }, { "epoch": 2.87, "learning_rate": 2.1261844233880288e-06, "loss": 0.0, "step": 12429 }, { "epoch": 2.87, "learning_rate": 2.122332640012326e-06, "loss": 0.0, "step": 12430 }, { "epoch": 2.87, "learning_rate": 2.118480856636623e-06, "loss": 0.0, "step": 12431 }, { "epoch": 2.87, "learning_rate": 2.11462907326092e-06, "loss": 0.0, "step": 12432 }, { "epoch": 2.87, "learning_rate": 2.110777289885217e-06, "loss": 0.0, "step": 12433 }, { "epoch": 2.87, "learning_rate": 2.106925506509514e-06, "loss": 0.0, "step": 12434 }, { "epoch": 2.87, "learning_rate": 2.103073723133811e-06, "loss": 0.0, "step": 12435 }, { "epoch": 2.87, "learning_rate": 2.099221939758108e-06, "loss": 0.0, "step": 12436 }, { "epoch": 2.87, "learning_rate": 2.0953701563824053e-06, "loss": 0.0, "step": 12437 }, { "epoch": 2.87, "learning_rate": 2.0915183730067024e-06, "loss": 0.0, "step": 12438 }, { "epoch": 2.87, "learning_rate": 2.087666589630999e-06, "loss": 0.0001, "step": 12439 }, { "epoch": 2.87, "learning_rate": 2.0838148062552966e-06, "loss": 0.0, "step": 12440 }, { "epoch": 2.88, "learning_rate": 2.0799630228795932e-06, "loss": 0.0, "step": 12441 }, { "epoch": 2.88, "learning_rate": 2.0761112395038903e-06, "loss": 0.0, "step": 12442 }, { "epoch": 2.88, "learning_rate": 2.0722594561281874e-06, "loss": 0.0, "step": 12443 }, { "epoch": 2.88, "learning_rate": 2.0684076727524844e-06, "loss": 0.0, "step": 12444 }, { "epoch": 2.88, "learning_rate": 2.0645558893767815e-06, "loss": 0.0, "step": 12445 }, { "epoch": 2.88, "learning_rate": 2.0607041060010786e-06, "loss": 0.0, "step": 12446 }, { "epoch": 2.88, "learning_rate": 2.0568523226253757e-06, "loss": 0.0, "step": 12447 }, { "epoch": 2.88, "learning_rate": 2.0530005392496727e-06, "loss": 0.0, "step": 12448 }, { "epoch": 2.88, "learning_rate": 2.04914875587397e-06, "loss": 0.0, "step": 12449 }, { "epoch": 2.88, "learning_rate": 2.0452969724982664e-06, "loss": 0.0, "step": 12450 }, { "epoch": 2.88, "learning_rate": 2.041445189122564e-06, "loss": 0.0, "step": 12451 }, { "epoch": 2.88, "learning_rate": 2.037593405746861e-06, "loss": 0.0, "step": 12452 }, { "epoch": 2.88, "learning_rate": 2.0337416223711577e-06, "loss": 0.0, "step": 12453 }, { "epoch": 2.88, "learning_rate": 2.029889838995455e-06, "loss": 0.0, "step": 12454 }, { "epoch": 2.88, "learning_rate": 2.0260380556197522e-06, "loss": 0.0, "step": 12455 }, { "epoch": 2.88, "learning_rate": 2.022186272244049e-06, "loss": 0.0, "step": 12456 }, { "epoch": 2.88, "learning_rate": 2.018334488868346e-06, "loss": 0.0, "step": 12457 }, { "epoch": 2.88, "learning_rate": 2.0144827054926435e-06, "loss": 0.0, "step": 12458 }, { "epoch": 2.88, "learning_rate": 2.01063092211694e-06, "loss": 0.0, "step": 12459 }, { "epoch": 2.88, "learning_rate": 2.006779138741237e-06, "loss": 0.0, "step": 12460 }, { "epoch": 2.88, "learning_rate": 2.0029273553655347e-06, "loss": 0.0, "step": 12461 }, { "epoch": 2.88, "learning_rate": 1.9990755719898313e-06, "loss": 0.0, "step": 12462 }, { "epoch": 2.88, "learning_rate": 1.9952237886141284e-06, "loss": 0.0, "step": 12463 }, { "epoch": 2.88, "learning_rate": 1.9913720052384255e-06, "loss": 0.0, "step": 12464 }, { "epoch": 2.88, "learning_rate": 1.9875202218627225e-06, "loss": 0.0, "step": 12465 }, { "epoch": 2.88, "learning_rate": 1.9836684384870196e-06, "loss": 0.0, "step": 12466 }, { "epoch": 2.88, "learning_rate": 1.9798166551113167e-06, "loss": 0.0, "step": 12467 }, { "epoch": 2.88, "learning_rate": 1.9759648717356138e-06, "loss": 0.0, "step": 12468 }, { "epoch": 2.88, "learning_rate": 1.972113088359911e-06, "loss": 0.0, "step": 12469 }, { "epoch": 2.88, "learning_rate": 1.968261304984208e-06, "loss": 0.0, "step": 12470 }, { "epoch": 2.88, "learning_rate": 1.9644095216085046e-06, "loss": 0.0, "step": 12471 }, { "epoch": 2.88, "learning_rate": 1.960557738232802e-06, "loss": 0.0, "step": 12472 }, { "epoch": 2.88, "learning_rate": 1.956705954857099e-06, "loss": 0.0, "step": 12473 }, { "epoch": 2.88, "learning_rate": 1.9528541714813958e-06, "loss": 0.0, "step": 12474 }, { "epoch": 2.88, "learning_rate": 1.9490023881056933e-06, "loss": 0.0, "step": 12475 }, { "epoch": 2.88, "learning_rate": 1.9451506047299903e-06, "loss": 0.0, "step": 12476 }, { "epoch": 2.88, "learning_rate": 1.941298821354287e-06, "loss": 0.0, "step": 12477 }, { "epoch": 2.88, "learning_rate": 1.937447037978584e-06, "loss": 0.0, "step": 12478 }, { "epoch": 2.88, "learning_rate": 1.933595254602881e-06, "loss": 0.0, "step": 12479 }, { "epoch": 2.88, "learning_rate": 1.9297434712271782e-06, "loss": 0.0, "step": 12480 }, { "epoch": 2.88, "learning_rate": 1.9258916878514753e-06, "loss": 0.0, "step": 12481 }, { "epoch": 2.88, "learning_rate": 1.9220399044757724e-06, "loss": 0.0, "step": 12482 }, { "epoch": 2.88, "learning_rate": 1.9181881211000694e-06, "loss": 0.0, "step": 12483 }, { "epoch": 2.89, "learning_rate": 1.9143363377243665e-06, "loss": 0.0, "step": 12484 }, { "epoch": 2.89, "learning_rate": 1.9104845543486636e-06, "loss": 0.0, "step": 12485 }, { "epoch": 2.89, "learning_rate": 1.9066327709729607e-06, "loss": 0.0, "step": 12486 }, { "epoch": 2.89, "learning_rate": 1.9027809875972577e-06, "loss": 0.0, "step": 12487 }, { "epoch": 2.89, "learning_rate": 1.8989292042215546e-06, "loss": 0.0, "step": 12488 }, { "epoch": 2.89, "learning_rate": 1.8950774208458519e-06, "loss": 0.0, "step": 12489 }, { "epoch": 2.89, "learning_rate": 1.8912256374701487e-06, "loss": 0.0, "step": 12490 }, { "epoch": 2.89, "learning_rate": 1.8873738540944458e-06, "loss": 0.0, "step": 12491 }, { "epoch": 2.89, "learning_rate": 1.8835220707187427e-06, "loss": 0.0, "step": 12492 }, { "epoch": 2.89, "learning_rate": 1.87967028734304e-06, "loss": 0.0, "step": 12493 }, { "epoch": 2.89, "learning_rate": 1.875818503967337e-06, "loss": 0.0, "step": 12494 }, { "epoch": 2.89, "learning_rate": 1.8719667205916339e-06, "loss": 0.0, "step": 12495 }, { "epoch": 2.89, "learning_rate": 1.8681149372159312e-06, "loss": 0.0, "step": 12496 }, { "epoch": 2.89, "learning_rate": 1.8642631538402283e-06, "loss": 0.0, "step": 12497 }, { "epoch": 2.89, "learning_rate": 1.8604113704645251e-06, "loss": 0.0, "step": 12498 }, { "epoch": 2.89, "learning_rate": 1.856559587088822e-06, "loss": 0.0, "step": 12499 }, { "epoch": 2.89, "learning_rate": 1.8527078037131195e-06, "loss": 0.0, "step": 12500 }, { "epoch": 2.89, "learning_rate": 1.8488560203374163e-06, "loss": 0.0, "step": 12501 }, { "epoch": 2.89, "learning_rate": 1.8450042369617132e-06, "loss": 0.0, "step": 12502 }, { "epoch": 2.89, "learning_rate": 1.8411524535860105e-06, "loss": 0.0, "step": 12503 }, { "epoch": 2.89, "learning_rate": 1.8373006702103076e-06, "loss": 0.0, "step": 12504 }, { "epoch": 2.89, "learning_rate": 1.8334488868346044e-06, "loss": 0.0, "step": 12505 }, { "epoch": 2.89, "learning_rate": 1.8295971034589015e-06, "loss": 0.0, "step": 12506 }, { "epoch": 2.89, "learning_rate": 1.8257453200831988e-06, "loss": 0.0, "step": 12507 }, { "epoch": 2.89, "learning_rate": 1.8218935367074956e-06, "loss": 0.0, "step": 12508 }, { "epoch": 2.89, "learning_rate": 1.8180417533317927e-06, "loss": 0.0, "step": 12509 }, { "epoch": 2.89, "learning_rate": 1.81418996995609e-06, "loss": 0.0, "step": 12510 }, { "epoch": 2.89, "learning_rate": 1.8103381865803869e-06, "loss": 0.0001, "step": 12511 }, { "epoch": 2.89, "learning_rate": 1.806486403204684e-06, "loss": 0.0, "step": 12512 }, { "epoch": 2.89, "learning_rate": 1.8026346198289808e-06, "loss": 0.0, "step": 12513 }, { "epoch": 2.89, "learning_rate": 1.798782836453278e-06, "loss": 0.0, "step": 12514 }, { "epoch": 2.89, "learning_rate": 1.794931053077575e-06, "loss": 0.0, "step": 12515 }, { "epoch": 2.89, "learning_rate": 1.791079269701872e-06, "loss": 0.0, "step": 12516 }, { "epoch": 2.89, "learning_rate": 1.7872274863261693e-06, "loss": 0.0, "step": 12517 }, { "epoch": 2.89, "learning_rate": 1.7833757029504662e-06, "loss": 0.0, "step": 12518 }, { "epoch": 2.89, "learning_rate": 1.7795239195747632e-06, "loss": 0.0, "step": 12519 }, { "epoch": 2.89, "learning_rate": 1.77567213619906e-06, "loss": 0.0, "step": 12520 }, { "epoch": 2.89, "learning_rate": 1.7718203528233574e-06, "loss": 0.0, "step": 12521 }, { "epoch": 2.89, "learning_rate": 1.7679685694476544e-06, "loss": 0.0, "step": 12522 }, { "epoch": 2.89, "learning_rate": 1.7641167860719513e-06, "loss": 0.0001, "step": 12523 }, { "epoch": 2.89, "learning_rate": 1.7602650026962486e-06, "loss": 0.0, "step": 12524 }, { "epoch": 2.89, "learning_rate": 1.7564132193205457e-06, "loss": 0.0, "step": 12525 }, { "epoch": 2.89, "learning_rate": 1.7525614359448425e-06, "loss": 0.0, "step": 12526 }, { "epoch": 2.9, "learning_rate": 1.7487096525691394e-06, "loss": 0.0, "step": 12527 }, { "epoch": 2.9, "learning_rate": 1.7448578691934367e-06, "loss": 0.0, "step": 12528 }, { "epoch": 2.9, "learning_rate": 1.7410060858177337e-06, "loss": 0.0, "step": 12529 }, { "epoch": 2.9, "learning_rate": 1.7371543024420306e-06, "loss": 0.0, "step": 12530 }, { "epoch": 2.9, "learning_rate": 1.7333025190663279e-06, "loss": 0.0, "step": 12531 }, { "epoch": 2.9, "learning_rate": 1.729450735690625e-06, "loss": 0.0, "step": 12532 }, { "epoch": 2.9, "learning_rate": 1.7255989523149218e-06, "loss": 0.0, "step": 12533 }, { "epoch": 2.9, "learning_rate": 1.721747168939219e-06, "loss": 0.0, "step": 12534 }, { "epoch": 2.9, "learning_rate": 1.7178953855635162e-06, "loss": 0.0001, "step": 12535 }, { "epoch": 2.9, "learning_rate": 1.714043602187813e-06, "loss": 0.0, "step": 12536 }, { "epoch": 2.9, "learning_rate": 1.71019181881211e-06, "loss": 0.0, "step": 12537 }, { "epoch": 2.9, "learning_rate": 1.7063400354364074e-06, "loss": 0.0, "step": 12538 }, { "epoch": 2.9, "learning_rate": 1.7024882520607043e-06, "loss": 0.0, "step": 12539 }, { "epoch": 2.9, "learning_rate": 1.6986364686850011e-06, "loss": 0.0, "step": 12540 }, { "epoch": 2.9, "learning_rate": 1.6947846853092982e-06, "loss": 0.0, "step": 12541 }, { "epoch": 2.9, "learning_rate": 1.6909329019335955e-06, "loss": 0.0, "step": 12542 }, { "epoch": 2.9, "learning_rate": 1.6870811185578923e-06, "loss": 0.0, "step": 12543 }, { "epoch": 2.9, "learning_rate": 1.6832293351821894e-06, "loss": 0.0, "step": 12544 }, { "epoch": 2.9, "learning_rate": 1.6793775518064867e-06, "loss": 0.0, "step": 12545 }, { "epoch": 2.9, "learning_rate": 1.6755257684307836e-06, "loss": 0.0, "step": 12546 }, { "epoch": 2.9, "learning_rate": 1.6716739850550806e-06, "loss": 0.0, "step": 12547 }, { "epoch": 2.9, "learning_rate": 1.6678222016793775e-06, "loss": 0.0, "step": 12548 }, { "epoch": 2.9, "learning_rate": 1.6639704183036748e-06, "loss": 0.0, "step": 12549 }, { "epoch": 2.9, "learning_rate": 1.6601186349279716e-06, "loss": 0.0, "step": 12550 }, { "epoch": 2.9, "learning_rate": 1.6562668515522687e-06, "loss": 0.0002, "step": 12551 }, { "epoch": 2.9, "learning_rate": 1.652415068176566e-06, "loss": 0.0, "step": 12552 }, { "epoch": 2.9, "learning_rate": 1.6485632848008629e-06, "loss": 0.0, "step": 12553 }, { "epoch": 2.9, "learning_rate": 1.64471150142516e-06, "loss": 0.0, "step": 12554 }, { "epoch": 2.9, "learning_rate": 1.6408597180494568e-06, "loss": 0.0, "step": 12555 }, { "epoch": 2.9, "learning_rate": 1.637007934673754e-06, "loss": 0.0, "step": 12556 }, { "epoch": 2.9, "learning_rate": 1.6331561512980512e-06, "loss": 0.0, "step": 12557 }, { "epoch": 2.9, "learning_rate": 1.629304367922348e-06, "loss": 0.0, "step": 12558 }, { "epoch": 2.9, "learning_rate": 1.6254525845466453e-06, "loss": 0.0, "step": 12559 }, { "epoch": 2.9, "learning_rate": 1.6216008011709424e-06, "loss": 0.0, "step": 12560 }, { "epoch": 2.9, "learning_rate": 1.6177490177952392e-06, "loss": 0.0, "step": 12561 }, { "epoch": 2.9, "learning_rate": 1.613897234419536e-06, "loss": 0.0, "step": 12562 }, { "epoch": 2.9, "learning_rate": 1.6100454510438336e-06, "loss": 0.0, "step": 12563 }, { "epoch": 2.9, "learning_rate": 1.6061936676681305e-06, "loss": 0.0, "step": 12564 }, { "epoch": 2.9, "learning_rate": 1.6023418842924273e-06, "loss": 0.0, "step": 12565 }, { "epoch": 2.9, "learning_rate": 1.5984901009167246e-06, "loss": 0.0, "step": 12566 }, { "epoch": 2.9, "learning_rate": 1.5946383175410217e-06, "loss": 0.0, "step": 12567 }, { "epoch": 2.9, "learning_rate": 1.5907865341653185e-06, "loss": 0.0, "step": 12568 }, { "epoch": 2.9, "learning_rate": 1.5869347507896156e-06, "loss": 0.0, "step": 12569 }, { "epoch": 2.91, "learning_rate": 1.583082967413913e-06, "loss": 0.0, "step": 12570 }, { "epoch": 2.91, "learning_rate": 1.5792311840382098e-06, "loss": 0.0, "step": 12571 }, { "epoch": 2.91, "learning_rate": 1.5753794006625068e-06, "loss": 0.0, "step": 12572 }, { "epoch": 2.91, "learning_rate": 1.5715276172868041e-06, "loss": 0.0, "step": 12573 }, { "epoch": 2.91, "learning_rate": 1.567675833911101e-06, "loss": 0.0001, "step": 12574 }, { "epoch": 2.91, "learning_rate": 1.5638240505353978e-06, "loss": 0.0, "step": 12575 }, { "epoch": 2.91, "learning_rate": 1.5599722671596951e-06, "loss": 0.0, "step": 12576 }, { "epoch": 2.91, "learning_rate": 1.556120483783992e-06, "loss": 0.0, "step": 12577 }, { "epoch": 2.91, "learning_rate": 1.552268700408289e-06, "loss": 0.0, "step": 12578 }, { "epoch": 2.91, "learning_rate": 1.5484169170325861e-06, "loss": 0.0, "step": 12579 }, { "epoch": 2.91, "learning_rate": 1.5445651336568832e-06, "loss": 0.0001, "step": 12580 }, { "epoch": 2.91, "learning_rate": 1.5407133502811803e-06, "loss": 0.0, "step": 12581 }, { "epoch": 2.91, "learning_rate": 1.5368615669054773e-06, "loss": 0.0, "step": 12582 }, { "epoch": 2.91, "learning_rate": 1.5330097835297744e-06, "loss": 0.0, "step": 12583 }, { "epoch": 2.91, "learning_rate": 1.5291580001540713e-06, "loss": 0.0, "step": 12584 }, { "epoch": 2.91, "learning_rate": 1.5253062167783686e-06, "loss": 0.0, "step": 12585 }, { "epoch": 2.91, "learning_rate": 1.5214544334026654e-06, "loss": 0.0, "step": 12586 }, { "epoch": 2.91, "learning_rate": 1.5176026500269625e-06, "loss": 0.0, "step": 12587 }, { "epoch": 2.91, "learning_rate": 1.5137508666512596e-06, "loss": 0.0, "step": 12588 }, { "epoch": 2.91, "learning_rate": 1.5098990832755566e-06, "loss": 0.0, "step": 12589 }, { "epoch": 2.91, "learning_rate": 1.5060472998998537e-06, "loss": 0.0, "step": 12590 }, { "epoch": 2.91, "learning_rate": 1.5021955165241508e-06, "loss": 0.0, "step": 12591 }, { "epoch": 2.91, "learning_rate": 1.4983437331484479e-06, "loss": 0.0, "step": 12592 }, { "epoch": 2.91, "learning_rate": 1.4944919497727447e-06, "loss": 0.0011, "step": 12593 }, { "epoch": 2.91, "learning_rate": 1.4906401663970418e-06, "loss": 0.0, "step": 12594 }, { "epoch": 2.91, "learning_rate": 1.486788383021339e-06, "loss": 0.0, "step": 12595 }, { "epoch": 2.91, "learning_rate": 1.482936599645636e-06, "loss": 0.0, "step": 12596 }, { "epoch": 2.91, "learning_rate": 1.479084816269933e-06, "loss": 0.0, "step": 12597 }, { "epoch": 2.91, "learning_rate": 1.47523303289423e-06, "loss": 0.0001, "step": 12598 }, { "epoch": 2.91, "learning_rate": 1.4713812495185272e-06, "loss": 0.0, "step": 12599 }, { "epoch": 2.91, "learning_rate": 1.467529466142824e-06, "loss": 0.0, "step": 12600 }, { "epoch": 2.91, "learning_rate": 1.4636776827671213e-06, "loss": 0.0, "step": 12601 }, { "epoch": 2.91, "learning_rate": 1.4598258993914184e-06, "loss": 0.0, "step": 12602 }, { "epoch": 2.91, "learning_rate": 1.4559741160157152e-06, "loss": 0.0, "step": 12603 }, { "epoch": 2.91, "learning_rate": 1.4521223326400125e-06, "loss": 0.0, "step": 12604 }, { "epoch": 2.91, "learning_rate": 1.4482705492643094e-06, "loss": 0.0, "step": 12605 }, { "epoch": 2.91, "learning_rate": 1.4444187658886065e-06, "loss": 0.0, "step": 12606 }, { "epoch": 2.91, "learning_rate": 1.4405669825129035e-06, "loss": 0.0, "step": 12607 }, { "epoch": 2.91, "learning_rate": 1.4367151991372006e-06, "loss": 0.0, "step": 12608 }, { "epoch": 2.91, "learning_rate": 1.4328634157614977e-06, "loss": 0.0, "step": 12609 }, { "epoch": 2.91, "learning_rate": 1.4290116323857948e-06, "loss": 0.0, "step": 12610 }, { "epoch": 2.91, "learning_rate": 1.4251598490100918e-06, "loss": 0.0, "step": 12611 }, { "epoch": 2.91, "learning_rate": 1.4213080656343887e-06, "loss": 0.0, "step": 12612 }, { "epoch": 2.91, "learning_rate": 1.4174562822586858e-06, "loss": 0.0001, "step": 12613 }, { "epoch": 2.92, "learning_rate": 1.4136044988829828e-06, "loss": 0.0, "step": 12614 }, { "epoch": 2.92, "learning_rate": 1.40975271550728e-06, "loss": 0.0006, "step": 12615 }, { "epoch": 2.92, "learning_rate": 1.405900932131577e-06, "loss": 0.0, "step": 12616 }, { "epoch": 2.92, "learning_rate": 1.402049148755874e-06, "loss": 0.0, "step": 12617 }, { "epoch": 2.92, "learning_rate": 1.3981973653801711e-06, "loss": 0.0, "step": 12618 }, { "epoch": 2.92, "learning_rate": 1.394345582004468e-06, "loss": 0.0, "step": 12619 }, { "epoch": 2.92, "learning_rate": 1.3904937986287653e-06, "loss": 0.0, "step": 12620 }, { "epoch": 2.92, "learning_rate": 1.3866420152530621e-06, "loss": 0.0, "step": 12621 }, { "epoch": 2.92, "learning_rate": 1.3827902318773592e-06, "loss": 0.0, "step": 12622 }, { "epoch": 2.92, "learning_rate": 1.3789384485016565e-06, "loss": 0.0, "step": 12623 }, { "epoch": 2.92, "learning_rate": 1.3750866651259534e-06, "loss": 0.0, "step": 12624 }, { "epoch": 2.92, "learning_rate": 1.3712348817502504e-06, "loss": 0.0, "step": 12625 }, { "epoch": 2.92, "learning_rate": 1.3673830983745475e-06, "loss": 0.0, "step": 12626 }, { "epoch": 2.92, "learning_rate": 1.3635313149988446e-06, "loss": 0.0, "step": 12627 }, { "epoch": 2.92, "learning_rate": 1.3596795316231414e-06, "loss": 0.0, "step": 12628 }, { "epoch": 2.92, "learning_rate": 1.3558277482474387e-06, "loss": 0.0, "step": 12629 }, { "epoch": 2.92, "learning_rate": 1.3519759648717358e-06, "loss": 0.0, "step": 12630 }, { "epoch": 2.92, "learning_rate": 1.3481241814960327e-06, "loss": 0.0, "step": 12631 }, { "epoch": 2.92, "learning_rate": 1.3442723981203297e-06, "loss": 0.0, "step": 12632 }, { "epoch": 2.92, "learning_rate": 1.3404206147446268e-06, "loss": 0.0, "step": 12633 }, { "epoch": 2.92, "learning_rate": 1.3365688313689239e-06, "loss": 0.0, "step": 12634 }, { "epoch": 2.92, "learning_rate": 1.332717047993221e-06, "loss": 0.0, "step": 12635 }, { "epoch": 2.92, "learning_rate": 1.328865264617518e-06, "loss": 0.0, "step": 12636 }, { "epoch": 2.92, "learning_rate": 1.325013481241815e-06, "loss": 0.0, "step": 12637 }, { "epoch": 2.92, "learning_rate": 1.321161697866112e-06, "loss": 0.0, "step": 12638 }, { "epoch": 2.92, "learning_rate": 1.3173099144904092e-06, "loss": 0.0002, "step": 12639 }, { "epoch": 2.92, "learning_rate": 1.3134581311147061e-06, "loss": 0.0, "step": 12640 }, { "epoch": 2.92, "learning_rate": 1.3096063477390032e-06, "loss": 0.0, "step": 12641 }, { "epoch": 2.92, "learning_rate": 1.3057545643633003e-06, "loss": 0.0, "step": 12642 }, { "epoch": 2.92, "learning_rate": 1.3019027809875973e-06, "loss": 0.0, "step": 12643 }, { "epoch": 2.92, "learning_rate": 1.2980509976118944e-06, "loss": 0.0141, "step": 12644 }, { "epoch": 2.92, "learning_rate": 1.2941992142361915e-06, "loss": 0.0, "step": 12645 }, { "epoch": 2.92, "learning_rate": 1.2903474308604885e-06, "loss": 0.0, "step": 12646 }, { "epoch": 2.92, "learning_rate": 1.2864956474847854e-06, "loss": 0.0, "step": 12647 }, { "epoch": 2.92, "learning_rate": 1.2826438641090827e-06, "loss": 0.0, "step": 12648 }, { "epoch": 2.92, "learning_rate": 1.2787920807333796e-06, "loss": 0.0, "step": 12649 }, { "epoch": 2.92, "learning_rate": 1.2749402973576766e-06, "loss": 0.0, "step": 12650 }, { "epoch": 2.92, "learning_rate": 1.2710885139819737e-06, "loss": 0.0, "step": 12651 }, { "epoch": 2.92, "learning_rate": 1.2672367306062708e-06, "loss": 0.0, "step": 12652 }, { "epoch": 2.92, "learning_rate": 1.2633849472305678e-06, "loss": 0.0, "step": 12653 }, { "epoch": 2.92, "learning_rate": 1.2595331638548647e-06, "loss": 0.0, "step": 12654 }, { "epoch": 2.92, "learning_rate": 1.255681380479162e-06, "loss": 0.0, "step": 12655 }, { "epoch": 2.92, "learning_rate": 1.2518295971034589e-06, "loss": 0.0, "step": 12656 }, { "epoch": 2.93, "learning_rate": 1.247977813727756e-06, "loss": 0.0, "step": 12657 }, { "epoch": 2.93, "learning_rate": 1.2441260303520532e-06, "loss": 0.0, "step": 12658 }, { "epoch": 2.93, "learning_rate": 1.24027424697635e-06, "loss": 0.0, "step": 12659 }, { "epoch": 2.93, "learning_rate": 1.2364224636006471e-06, "loss": 0.0, "step": 12660 }, { "epoch": 2.93, "learning_rate": 1.2325706802249442e-06, "loss": 0.0, "step": 12661 }, { "epoch": 2.93, "learning_rate": 1.2287188968492413e-06, "loss": 0.0, "step": 12662 }, { "epoch": 2.93, "learning_rate": 1.2248671134735382e-06, "loss": 0.0, "step": 12663 }, { "epoch": 2.93, "learning_rate": 1.2210153300978354e-06, "loss": 0.0, "step": 12664 }, { "epoch": 2.93, "learning_rate": 1.2171635467221325e-06, "loss": 0.0, "step": 12665 }, { "epoch": 2.93, "learning_rate": 1.2133117633464294e-06, "loss": 0.0, "step": 12666 }, { "epoch": 2.93, "learning_rate": 1.2094599799707267e-06, "loss": 0.0, "step": 12667 }, { "epoch": 2.93, "learning_rate": 1.2056081965950235e-06, "loss": 0.0, "step": 12668 }, { "epoch": 2.93, "learning_rate": 1.2017564132193206e-06, "loss": 0.0, "step": 12669 }, { "epoch": 2.93, "learning_rate": 1.1979046298436177e-06, "loss": 0.0, "step": 12670 }, { "epoch": 2.93, "learning_rate": 1.1940528464679147e-06, "loss": 0.0, "step": 12671 }, { "epoch": 2.93, "learning_rate": 1.1902010630922118e-06, "loss": 0.0, "step": 12672 }, { "epoch": 2.93, "learning_rate": 1.1863492797165087e-06, "loss": 0.0, "step": 12673 }, { "epoch": 2.93, "learning_rate": 1.182497496340806e-06, "loss": 0.0, "step": 12674 }, { "epoch": 2.93, "learning_rate": 1.1786457129651028e-06, "loss": 0.0, "step": 12675 }, { "epoch": 2.93, "learning_rate": 1.1747939295893999e-06, "loss": 0.0, "step": 12676 }, { "epoch": 2.93, "learning_rate": 1.170942146213697e-06, "loss": 0.0, "step": 12677 }, { "epoch": 2.93, "learning_rate": 1.167090362837994e-06, "loss": 0.0, "step": 12678 }, { "epoch": 2.93, "learning_rate": 1.1632385794622911e-06, "loss": 0.0, "step": 12679 }, { "epoch": 2.93, "learning_rate": 1.1593867960865882e-06, "loss": 0.0, "step": 12680 }, { "epoch": 2.93, "learning_rate": 1.1555350127108853e-06, "loss": 0.0, "step": 12681 }, { "epoch": 2.93, "learning_rate": 1.1516832293351821e-06, "loss": 0.0, "step": 12682 }, { "epoch": 2.93, "learning_rate": 1.1478314459594794e-06, "loss": 0.0, "step": 12683 }, { "epoch": 2.93, "learning_rate": 1.1439796625837763e-06, "loss": 0.0, "step": 12684 }, { "epoch": 2.93, "learning_rate": 1.1401278792080733e-06, "loss": 0.0, "step": 12685 }, { "epoch": 2.93, "learning_rate": 1.1362760958323704e-06, "loss": 0.0, "step": 12686 }, { "epoch": 2.93, "learning_rate": 1.1324243124566675e-06, "loss": 0.0, "step": 12687 }, { "epoch": 2.93, "learning_rate": 1.1285725290809646e-06, "loss": 0.0, "step": 12688 }, { "epoch": 2.93, "learning_rate": 1.1247207457052616e-06, "loss": 0.0, "step": 12689 }, { "epoch": 2.93, "learning_rate": 1.1208689623295587e-06, "loss": 0.0, "step": 12690 }, { "epoch": 2.93, "learning_rate": 1.1170171789538556e-06, "loss": 0.0, "step": 12691 }, { "epoch": 2.93, "learning_rate": 1.1131653955781526e-06, "loss": 0.0, "step": 12692 }, { "epoch": 2.93, "learning_rate": 1.10931361220245e-06, "loss": 0.0, "step": 12693 }, { "epoch": 2.93, "learning_rate": 1.1054618288267468e-06, "loss": 0.0, "step": 12694 }, { "epoch": 2.93, "learning_rate": 1.1016100454510439e-06, "loss": 0.0, "step": 12695 }, { "epoch": 2.93, "learning_rate": 1.097758262075341e-06, "loss": 0.0, "step": 12696 }, { "epoch": 2.93, "learning_rate": 1.093906478699638e-06, "loss": 0.0001, "step": 12697 }, { "epoch": 2.93, "learning_rate": 1.0900546953239349e-06, "loss": 0.0, "step": 12698 }, { "epoch": 2.93, "learning_rate": 1.0862029119482322e-06, "loss": 0.0, "step": 12699 }, { "epoch": 2.94, "learning_rate": 1.0823511285725292e-06, "loss": 0.0, "step": 12700 }, { "epoch": 2.94, "learning_rate": 1.078499345196826e-06, "loss": 0.0, "step": 12701 }, { "epoch": 2.94, "learning_rate": 1.0746475618211234e-06, "loss": 0.0, "step": 12702 }, { "epoch": 2.94, "learning_rate": 1.0707957784454202e-06, "loss": 0.0, "step": 12703 }, { "epoch": 2.94, "learning_rate": 1.0669439950697173e-06, "loss": 0.0, "step": 12704 }, { "epoch": 2.94, "learning_rate": 1.0630922116940144e-06, "loss": 0.0, "step": 12705 }, { "epoch": 2.94, "learning_rate": 1.0592404283183115e-06, "loss": 0.0, "step": 12706 }, { "epoch": 2.94, "learning_rate": 1.0553886449426085e-06, "loss": 0.0, "step": 12707 }, { "epoch": 2.94, "learning_rate": 1.0515368615669056e-06, "loss": 0.0, "step": 12708 }, { "epoch": 2.94, "learning_rate": 1.0476850781912027e-06, "loss": 0.0, "step": 12709 }, { "epoch": 2.94, "learning_rate": 1.0438332948154995e-06, "loss": 0.0, "step": 12710 }, { "epoch": 2.94, "learning_rate": 1.0399815114397966e-06, "loss": 0.0, "step": 12711 }, { "epoch": 2.94, "learning_rate": 1.0361297280640937e-06, "loss": 0.0, "step": 12712 }, { "epoch": 2.94, "learning_rate": 1.0322779446883908e-06, "loss": 0.0, "step": 12713 }, { "epoch": 2.94, "learning_rate": 1.0284261613126878e-06, "loss": 0.0, "step": 12714 }, { "epoch": 2.94, "learning_rate": 1.024574377936985e-06, "loss": 0.0, "step": 12715 }, { "epoch": 2.94, "learning_rate": 1.020722594561282e-06, "loss": 0.0, "step": 12716 }, { "epoch": 2.94, "learning_rate": 1.0168708111855788e-06, "loss": 0.0, "step": 12717 }, { "epoch": 2.94, "learning_rate": 1.0130190278098761e-06, "loss": 0.0, "step": 12718 }, { "epoch": 2.94, "learning_rate": 1.009167244434173e-06, "loss": 0.0, "step": 12719 }, { "epoch": 2.94, "learning_rate": 1.00531546105847e-06, "loss": 0.0, "step": 12720 }, { "epoch": 2.94, "learning_rate": 1.0014636776827673e-06, "loss": 0.0, "step": 12721 }, { "epoch": 2.94, "learning_rate": 9.976118943070642e-07, "loss": 0.0, "step": 12722 }, { "epoch": 2.94, "learning_rate": 9.937601109313613e-07, "loss": 0.0, "step": 12723 }, { "epoch": 2.94, "learning_rate": 9.899083275556583e-07, "loss": 0.0, "step": 12724 }, { "epoch": 2.94, "learning_rate": 9.860565441799554e-07, "loss": 0.0, "step": 12725 }, { "epoch": 2.94, "learning_rate": 9.822047608042523e-07, "loss": 0.0, "step": 12726 }, { "epoch": 2.94, "learning_rate": 9.783529774285496e-07, "loss": 0.0, "step": 12727 }, { "epoch": 2.94, "learning_rate": 9.745011940528466e-07, "loss": 0.0, "step": 12728 }, { "epoch": 2.94, "learning_rate": 9.706494106771435e-07, "loss": 0.0, "step": 12729 }, { "epoch": 2.94, "learning_rate": 9.667976273014406e-07, "loss": 0.0, "step": 12730 }, { "epoch": 2.94, "learning_rate": 9.629458439257376e-07, "loss": 0.0, "step": 12731 }, { "epoch": 2.94, "learning_rate": 9.590940605500347e-07, "loss": 0.0, "step": 12732 }, { "epoch": 2.94, "learning_rate": 9.552422771743318e-07, "loss": 0.0, "step": 12733 }, { "epoch": 2.94, "learning_rate": 9.513904937986289e-07, "loss": 0.0, "step": 12734 }, { "epoch": 2.94, "learning_rate": 9.475387104229259e-07, "loss": 0.0, "step": 12735 }, { "epoch": 2.94, "learning_rate": 9.436869270472229e-07, "loss": 0.0, "step": 12736 }, { "epoch": 2.94, "learning_rate": 9.3983514367152e-07, "loss": 0.0, "step": 12737 }, { "epoch": 2.94, "learning_rate": 9.359833602958169e-07, "loss": 0.0, "step": 12738 }, { "epoch": 2.94, "learning_rate": 9.321315769201141e-07, "loss": 0.0, "step": 12739 }, { "epoch": 2.94, "learning_rate": 9.28279793544411e-07, "loss": 0.0, "step": 12740 }, { "epoch": 2.94, "learning_rate": 9.244280101687082e-07, "loss": 0.0, "step": 12741 }, { "epoch": 2.94, "learning_rate": 9.205762267930052e-07, "loss": 0.0, "step": 12742 }, { "epoch": 2.94, "learning_rate": 9.167244434173022e-07, "loss": 0.0, "step": 12743 }, { "epoch": 2.95, "learning_rate": 9.128726600415994e-07, "loss": 0.0, "step": 12744 }, { "epoch": 2.95, "learning_rate": 9.090208766658964e-07, "loss": 0.0, "step": 12745 }, { "epoch": 2.95, "learning_rate": 9.051690932901934e-07, "loss": 0.0, "step": 12746 }, { "epoch": 2.95, "learning_rate": 9.013173099144904e-07, "loss": 0.0, "step": 12747 }, { "epoch": 2.95, "learning_rate": 8.974655265387875e-07, "loss": 0.0002, "step": 12748 }, { "epoch": 2.95, "learning_rate": 8.936137431630846e-07, "loss": 0.0, "step": 12749 }, { "epoch": 2.95, "learning_rate": 8.897619597873816e-07, "loss": 0.0, "step": 12750 }, { "epoch": 2.95, "learning_rate": 8.859101764116787e-07, "loss": 0.0, "step": 12751 }, { "epoch": 2.95, "learning_rate": 8.820583930359757e-07, "loss": 0.0, "step": 12752 }, { "epoch": 2.95, "learning_rate": 8.782066096602728e-07, "loss": 0.0, "step": 12753 }, { "epoch": 2.95, "learning_rate": 8.743548262845697e-07, "loss": 0.0, "step": 12754 }, { "epoch": 2.95, "learning_rate": 8.705030429088669e-07, "loss": 0.0, "step": 12755 }, { "epoch": 2.95, "learning_rate": 8.666512595331639e-07, "loss": 0.0, "step": 12756 }, { "epoch": 2.95, "learning_rate": 8.627994761574609e-07, "loss": 0.0, "step": 12757 }, { "epoch": 2.95, "learning_rate": 8.589476927817581e-07, "loss": 0.0, "step": 12758 }, { "epoch": 2.95, "learning_rate": 8.55095909406055e-07, "loss": 0.0, "step": 12759 }, { "epoch": 2.95, "learning_rate": 8.512441260303521e-07, "loss": 0.0, "step": 12760 }, { "epoch": 2.95, "learning_rate": 8.473923426546491e-07, "loss": 0.0, "step": 12761 }, { "epoch": 2.95, "learning_rate": 8.435405592789462e-07, "loss": 0.0, "step": 12762 }, { "epoch": 2.95, "learning_rate": 8.396887759032434e-07, "loss": 0.0, "step": 12763 }, { "epoch": 2.95, "learning_rate": 8.358369925275403e-07, "loss": 0.0, "step": 12764 }, { "epoch": 2.95, "learning_rate": 8.319852091518374e-07, "loss": 0.0, "step": 12765 }, { "epoch": 2.95, "learning_rate": 8.281334257761344e-07, "loss": 0.0, "step": 12766 }, { "epoch": 2.95, "learning_rate": 8.242816424004314e-07, "loss": 0.0, "step": 12767 }, { "epoch": 2.95, "learning_rate": 8.204298590247284e-07, "loss": 0.0, "step": 12768 }, { "epoch": 2.95, "learning_rate": 8.165780756490256e-07, "loss": 0.0, "step": 12769 }, { "epoch": 2.95, "learning_rate": 8.127262922733227e-07, "loss": 0.0, "step": 12770 }, { "epoch": 2.95, "learning_rate": 8.088745088976196e-07, "loss": 0.0, "step": 12771 }, { "epoch": 2.95, "learning_rate": 8.050227255219168e-07, "loss": 0.0, "step": 12772 }, { "epoch": 2.95, "learning_rate": 8.011709421462137e-07, "loss": 0.0, "step": 12773 }, { "epoch": 2.95, "learning_rate": 7.973191587705108e-07, "loss": 0.0, "step": 12774 }, { "epoch": 2.95, "learning_rate": 7.934673753948078e-07, "loss": 0.0, "step": 12775 }, { "epoch": 2.95, "learning_rate": 7.896155920191049e-07, "loss": 0.0, "step": 12776 }, { "epoch": 2.95, "learning_rate": 7.857638086434021e-07, "loss": 0.0, "step": 12777 }, { "epoch": 2.95, "learning_rate": 7.819120252676989e-07, "loss": 0.0, "step": 12778 }, { "epoch": 2.95, "learning_rate": 7.78060241891996e-07, "loss": 0.0, "step": 12779 }, { "epoch": 2.95, "learning_rate": 7.742084585162931e-07, "loss": 0.0, "step": 12780 }, { "epoch": 2.95, "learning_rate": 7.703566751405901e-07, "loss": 0.0, "step": 12781 }, { "epoch": 2.95, "learning_rate": 7.665048917648872e-07, "loss": 0.0, "step": 12782 }, { "epoch": 2.95, "learning_rate": 7.626531083891843e-07, "loss": 0.0, "step": 12783 }, { "epoch": 2.95, "learning_rate": 7.588013250134813e-07, "loss": 0.0, "step": 12784 }, { "epoch": 2.95, "learning_rate": 7.549495416377783e-07, "loss": 0.0, "step": 12785 }, { "epoch": 2.95, "learning_rate": 7.510977582620754e-07, "loss": 0.0, "step": 12786 }, { "epoch": 2.96, "learning_rate": 7.472459748863724e-07, "loss": 0.0, "step": 12787 }, { "epoch": 2.96, "learning_rate": 7.433941915106695e-07, "loss": 0.0001, "step": 12788 }, { "epoch": 2.96, "learning_rate": 7.395424081349665e-07, "loss": 0.0, "step": 12789 }, { "epoch": 2.96, "learning_rate": 7.356906247592636e-07, "loss": 0.0, "step": 12790 }, { "epoch": 2.96, "learning_rate": 7.318388413835607e-07, "loss": 0.0, "step": 12791 }, { "epoch": 2.96, "learning_rate": 7.279870580078576e-07, "loss": 0.0, "step": 12792 }, { "epoch": 2.96, "learning_rate": 7.241352746321547e-07, "loss": 0.0, "step": 12793 }, { "epoch": 2.96, "learning_rate": 7.202834912564518e-07, "loss": 0.0, "step": 12794 }, { "epoch": 2.96, "learning_rate": 7.164317078807488e-07, "loss": 0.0, "step": 12795 }, { "epoch": 2.96, "learning_rate": 7.125799245050459e-07, "loss": 0.0, "step": 12796 }, { "epoch": 2.96, "learning_rate": 7.087281411293429e-07, "loss": 0.0, "step": 12797 }, { "epoch": 2.96, "learning_rate": 7.0487635775364e-07, "loss": 0.0, "step": 12798 }, { "epoch": 2.96, "learning_rate": 7.01024574377937e-07, "loss": 0.0, "step": 12799 }, { "epoch": 2.96, "learning_rate": 6.97172791002234e-07, "loss": 0.0, "step": 12800 }, { "epoch": 2.96, "learning_rate": 6.933210076265311e-07, "loss": 0.0, "step": 12801 }, { "epoch": 2.96, "learning_rate": 6.894692242508283e-07, "loss": 0.0, "step": 12802 }, { "epoch": 2.96, "learning_rate": 6.856174408751252e-07, "loss": 0.0, "step": 12803 }, { "epoch": 2.96, "learning_rate": 6.817656574994223e-07, "loss": 0.0, "step": 12804 }, { "epoch": 2.96, "learning_rate": 6.779138741237194e-07, "loss": 0.0001, "step": 12805 }, { "epoch": 2.96, "learning_rate": 6.740620907480163e-07, "loss": 0.0, "step": 12806 }, { "epoch": 2.96, "learning_rate": 6.702103073723134e-07, "loss": 0.0001, "step": 12807 }, { "epoch": 2.96, "learning_rate": 6.663585239966105e-07, "loss": 0.0001, "step": 12808 }, { "epoch": 2.96, "learning_rate": 6.625067406209076e-07, "loss": 0.0, "step": 12809 }, { "epoch": 2.96, "learning_rate": 6.586549572452046e-07, "loss": 0.0, "step": 12810 }, { "epoch": 2.96, "learning_rate": 6.548031738695016e-07, "loss": 0.0, "step": 12811 }, { "epoch": 2.96, "learning_rate": 6.509513904937987e-07, "loss": 0.0, "step": 12812 }, { "epoch": 2.96, "learning_rate": 6.470996071180957e-07, "loss": 0.0, "step": 12813 }, { "epoch": 2.96, "learning_rate": 6.432478237423927e-07, "loss": 0.0, "step": 12814 }, { "epoch": 2.96, "learning_rate": 6.393960403666898e-07, "loss": 0.0, "step": 12815 }, { "epoch": 2.96, "learning_rate": 6.355442569909869e-07, "loss": 0.0, "step": 12816 }, { "epoch": 2.96, "learning_rate": 6.316924736152839e-07, "loss": 0.0, "step": 12817 }, { "epoch": 2.96, "learning_rate": 6.27840690239581e-07, "loss": 0.0, "step": 12818 }, { "epoch": 2.96, "learning_rate": 6.23988906863878e-07, "loss": 0.0, "step": 12819 }, { "epoch": 2.96, "learning_rate": 6.20137123488175e-07, "loss": 0.0, "step": 12820 }, { "epoch": 2.96, "learning_rate": 6.162853401124721e-07, "loss": 0.0, "step": 12821 }, { "epoch": 2.96, "learning_rate": 6.124335567367691e-07, "loss": 0.0, "step": 12822 }, { "epoch": 2.96, "learning_rate": 6.085817733610663e-07, "loss": 0.0, "step": 12823 }, { "epoch": 2.96, "learning_rate": 6.047299899853633e-07, "loss": 0.0, "step": 12824 }, { "epoch": 2.96, "learning_rate": 6.008782066096603e-07, "loss": 0.0, "step": 12825 }, { "epoch": 2.96, "learning_rate": 5.970264232339574e-07, "loss": 0.0, "step": 12826 }, { "epoch": 2.96, "learning_rate": 5.931746398582543e-07, "loss": 0.0, "step": 12827 }, { "epoch": 2.96, "learning_rate": 5.893228564825514e-07, "loss": 0.0, "step": 12828 }, { "epoch": 2.96, "learning_rate": 5.854710731068485e-07, "loss": 0.0, "step": 12829 }, { "epoch": 2.97, "learning_rate": 5.816192897311456e-07, "loss": 0.0, "step": 12830 }, { "epoch": 2.97, "learning_rate": 5.777675063554426e-07, "loss": 0.0, "step": 12831 }, { "epoch": 2.97, "learning_rate": 5.739157229797397e-07, "loss": 0.0, "step": 12832 }, { "epoch": 2.97, "learning_rate": 5.700639396040367e-07, "loss": 0.0, "step": 12833 }, { "epoch": 2.97, "learning_rate": 5.662121562283337e-07, "loss": 0.0, "step": 12834 }, { "epoch": 2.97, "learning_rate": 5.623603728526308e-07, "loss": 0.0, "step": 12835 }, { "epoch": 2.97, "learning_rate": 5.585085894769278e-07, "loss": 0.0001, "step": 12836 }, { "epoch": 2.97, "learning_rate": 5.54656806101225e-07, "loss": 0.0, "step": 12837 }, { "epoch": 2.97, "learning_rate": 5.508050227255219e-07, "loss": 0.0, "step": 12838 }, { "epoch": 2.97, "learning_rate": 5.46953239349819e-07, "loss": 0.0, "step": 12839 }, { "epoch": 2.97, "learning_rate": 5.431014559741161e-07, "loss": 0.0, "step": 12840 }, { "epoch": 2.97, "learning_rate": 5.39249672598413e-07, "loss": 0.0, "step": 12841 }, { "epoch": 2.97, "learning_rate": 5.353978892227101e-07, "loss": 0.0, "step": 12842 }, { "epoch": 2.97, "learning_rate": 5.315461058470072e-07, "loss": 0.0, "step": 12843 }, { "epoch": 2.97, "learning_rate": 5.276943224713043e-07, "loss": 0.0, "step": 12844 }, { "epoch": 2.97, "learning_rate": 5.238425390956013e-07, "loss": 0.0, "step": 12845 }, { "epoch": 2.97, "learning_rate": 5.199907557198983e-07, "loss": 0.0, "step": 12846 }, { "epoch": 2.97, "learning_rate": 5.161389723441954e-07, "loss": 0.0, "step": 12847 }, { "epoch": 2.97, "learning_rate": 5.122871889684924e-07, "loss": 0.0, "step": 12848 }, { "epoch": 2.97, "learning_rate": 5.084354055927894e-07, "loss": 0.0, "step": 12849 }, { "epoch": 2.97, "learning_rate": 5.045836222170865e-07, "loss": 0.0, "step": 12850 }, { "epoch": 2.97, "learning_rate": 5.007318388413837e-07, "loss": 0.0, "step": 12851 }, { "epoch": 2.97, "learning_rate": 4.968800554656806e-07, "loss": 0.0, "step": 12852 }, { "epoch": 2.97, "learning_rate": 4.930282720899777e-07, "loss": 0.0, "step": 12853 }, { "epoch": 2.97, "learning_rate": 4.891764887142748e-07, "loss": 0.0001, "step": 12854 }, { "epoch": 2.97, "learning_rate": 4.853247053385717e-07, "loss": 0.0, "step": 12855 }, { "epoch": 2.97, "learning_rate": 4.814729219628688e-07, "loss": 0.0, "step": 12856 }, { "epoch": 2.97, "learning_rate": 4.776211385871659e-07, "loss": 0.0, "step": 12857 }, { "epoch": 2.97, "learning_rate": 4.7376935521146297e-07, "loss": 0.0, "step": 12858 }, { "epoch": 2.97, "learning_rate": 4.6991757183576e-07, "loss": 0.0, "step": 12859 }, { "epoch": 2.97, "learning_rate": 4.6606578846005706e-07, "loss": 0.0, "step": 12860 }, { "epoch": 2.97, "learning_rate": 4.622140050843541e-07, "loss": 0.0, "step": 12861 }, { "epoch": 2.97, "learning_rate": 4.583622217086511e-07, "loss": 0.0, "step": 12862 }, { "epoch": 2.97, "learning_rate": 4.545104383329482e-07, "loss": 0.0, "step": 12863 }, { "epoch": 2.97, "learning_rate": 4.506586549572452e-07, "loss": 0.0, "step": 12864 }, { "epoch": 2.97, "learning_rate": 4.468068715815423e-07, "loss": 0.0, "step": 12865 }, { "epoch": 2.97, "learning_rate": 4.4295508820583934e-07, "loss": 0.0, "step": 12866 }, { "epoch": 2.97, "learning_rate": 4.391033048301364e-07, "loss": 0.0, "step": 12867 }, { "epoch": 2.97, "learning_rate": 4.3525152145443344e-07, "loss": 0.0, "step": 12868 }, { "epoch": 2.97, "learning_rate": 4.3139973807873046e-07, "loss": 0.0, "step": 12869 }, { "epoch": 2.97, "learning_rate": 4.275479547030275e-07, "loss": 0.0, "step": 12870 }, { "epoch": 2.97, "learning_rate": 4.2369617132732455e-07, "loss": 0.0, "step": 12871 }, { "epoch": 2.97, "learning_rate": 4.198443879516217e-07, "loss": 0.0002, "step": 12872 }, { "epoch": 2.98, "learning_rate": 4.159926045759187e-07, "loss": 0.0, "step": 12873 }, { "epoch": 2.98, "learning_rate": 4.121408212002157e-07, "loss": 0.0, "step": 12874 }, { "epoch": 2.98, "learning_rate": 4.082890378245128e-07, "loss": 0.0, "step": 12875 }, { "epoch": 2.98, "learning_rate": 4.044372544488098e-07, "loss": 0.0, "step": 12876 }, { "epoch": 2.98, "learning_rate": 4.0058547107310683e-07, "loss": 0.0, "step": 12877 }, { "epoch": 2.98, "learning_rate": 3.967336876974039e-07, "loss": 0.0, "step": 12878 }, { "epoch": 2.98, "learning_rate": 3.9288190432170103e-07, "loss": 0.0, "step": 12879 }, { "epoch": 2.98, "learning_rate": 3.89030120945998e-07, "loss": 0.0, "step": 12880 }, { "epoch": 2.98, "learning_rate": 3.8517833757029507e-07, "loss": 0.0, "step": 12881 }, { "epoch": 2.98, "learning_rate": 3.8132655419459214e-07, "loss": 0.0, "step": 12882 }, { "epoch": 2.98, "learning_rate": 3.7747477081888916e-07, "loss": 0.0, "step": 12883 }, { "epoch": 2.98, "learning_rate": 3.736229874431862e-07, "loss": 0.0, "step": 12884 }, { "epoch": 2.98, "learning_rate": 3.6977120406748326e-07, "loss": 0.0, "step": 12885 }, { "epoch": 2.98, "learning_rate": 3.6591942069178033e-07, "loss": 0.0, "step": 12886 }, { "epoch": 2.98, "learning_rate": 3.6206763731607735e-07, "loss": 0.0, "step": 12887 }, { "epoch": 2.98, "learning_rate": 3.582158539403744e-07, "loss": 0.0, "step": 12888 }, { "epoch": 2.98, "learning_rate": 3.5436407056467144e-07, "loss": 0.0, "step": 12889 }, { "epoch": 2.98, "learning_rate": 3.505122871889685e-07, "loss": 0.0, "step": 12890 }, { "epoch": 2.98, "learning_rate": 3.4666050381326554e-07, "loss": 0.0, "step": 12891 }, { "epoch": 2.98, "learning_rate": 3.428087204375626e-07, "loss": 0.0, "step": 12892 }, { "epoch": 2.98, "learning_rate": 3.389569370618597e-07, "loss": 0.0, "step": 12893 }, { "epoch": 2.98, "learning_rate": 3.351051536861567e-07, "loss": 0.0, "step": 12894 }, { "epoch": 2.98, "learning_rate": 3.312533703104538e-07, "loss": 0.0, "step": 12895 }, { "epoch": 2.98, "learning_rate": 3.274015869347508e-07, "loss": 0.0, "step": 12896 }, { "epoch": 2.98, "learning_rate": 3.2354980355904787e-07, "loss": 0.0, "step": 12897 }, { "epoch": 2.98, "learning_rate": 3.196980201833449e-07, "loss": 0.0, "step": 12898 }, { "epoch": 2.98, "learning_rate": 3.1584623680764196e-07, "loss": 0.0, "step": 12899 }, { "epoch": 2.98, "learning_rate": 3.11994453431939e-07, "loss": 0.0, "step": 12900 }, { "epoch": 2.98, "learning_rate": 3.0814267005623606e-07, "loss": 0.0, "step": 12901 }, { "epoch": 2.98, "learning_rate": 3.0429088668053313e-07, "loss": 0.0, "step": 12902 }, { "epoch": 2.98, "learning_rate": 3.0043910330483015e-07, "loss": 0.0, "step": 12903 }, { "epoch": 2.98, "learning_rate": 2.9658731992912717e-07, "loss": 0.0, "step": 12904 }, { "epoch": 2.98, "learning_rate": 2.9273553655342424e-07, "loss": 0.0, "step": 12905 }, { "epoch": 2.98, "learning_rate": 2.888837531777213e-07, "loss": 0.0, "step": 12906 }, { "epoch": 2.98, "learning_rate": 2.8503196980201834e-07, "loss": 0.0, "step": 12907 }, { "epoch": 2.98, "learning_rate": 2.811801864263154e-07, "loss": 0.0, "step": 12908 }, { "epoch": 2.98, "learning_rate": 2.773284030506125e-07, "loss": 0.0, "step": 12909 }, { "epoch": 2.98, "learning_rate": 2.734766196749095e-07, "loss": 0.0, "step": 12910 }, { "epoch": 2.98, "learning_rate": 2.696248362992065e-07, "loss": 0.0, "step": 12911 }, { "epoch": 2.98, "learning_rate": 2.657730529235036e-07, "loss": 0.0, "step": 12912 }, { "epoch": 2.98, "learning_rate": 2.6192126954780067e-07, "loss": 0.0, "step": 12913 }, { "epoch": 2.98, "learning_rate": 2.580694861720977e-07, "loss": 0.0, "step": 12914 }, { "epoch": 2.98, "learning_rate": 2.542177027963947e-07, "loss": 0.0, "step": 12915 }, { "epoch": 2.98, "learning_rate": 2.5036591942069183e-07, "loss": 0.0, "step": 12916 }, { "epoch": 2.99, "learning_rate": 2.4651413604498885e-07, "loss": 0.0, "step": 12917 }, { "epoch": 2.99, "learning_rate": 2.426623526692859e-07, "loss": 0.0, "step": 12918 }, { "epoch": 2.99, "learning_rate": 2.3881056929358295e-07, "loss": 0.0, "step": 12919 }, { "epoch": 2.99, "learning_rate": 2.3495878591788e-07, "loss": 0.0, "step": 12920 }, { "epoch": 2.99, "learning_rate": 2.3110700254217704e-07, "loss": 0.0, "step": 12921 }, { "epoch": 2.99, "learning_rate": 2.272552191664741e-07, "loss": 0.0, "step": 12922 }, { "epoch": 2.99, "learning_rate": 2.2340343579077116e-07, "loss": 0.0, "step": 12923 }, { "epoch": 2.99, "learning_rate": 2.195516524150682e-07, "loss": 0.0, "step": 12924 }, { "epoch": 2.99, "learning_rate": 2.1569986903936523e-07, "loss": 0.0, "step": 12925 }, { "epoch": 2.99, "learning_rate": 2.1184808566366227e-07, "loss": 0.0, "step": 12926 }, { "epoch": 2.99, "learning_rate": 2.0799630228795935e-07, "loss": 0.0, "step": 12927 }, { "epoch": 2.99, "learning_rate": 2.041445189122564e-07, "loss": 0.0, "step": 12928 }, { "epoch": 2.99, "learning_rate": 2.0029273553655341e-07, "loss": 0.0, "step": 12929 }, { "epoch": 2.99, "learning_rate": 1.9644095216085051e-07, "loss": 0.0, "step": 12930 }, { "epoch": 2.99, "learning_rate": 1.9258916878514753e-07, "loss": 0.0, "step": 12931 }, { "epoch": 2.99, "learning_rate": 1.8873738540944458e-07, "loss": 0.0, "step": 12932 }, { "epoch": 2.99, "learning_rate": 1.8488560203374163e-07, "loss": 0.0, "step": 12933 }, { "epoch": 2.99, "learning_rate": 1.8103381865803867e-07, "loss": 0.0, "step": 12934 }, { "epoch": 2.99, "learning_rate": 1.7718203528233572e-07, "loss": 0.0, "step": 12935 }, { "epoch": 2.99, "learning_rate": 1.7333025190663277e-07, "loss": 0.0, "step": 12936 }, { "epoch": 2.99, "learning_rate": 1.6947846853092984e-07, "loss": 0.0, "step": 12937 }, { "epoch": 2.99, "learning_rate": 1.656266851552269e-07, "loss": 0.0, "step": 12938 }, { "epoch": 2.99, "learning_rate": 1.6177490177952393e-07, "loss": 0.0, "step": 12939 }, { "epoch": 2.99, "learning_rate": 1.5792311840382098e-07, "loss": 0.0, "step": 12940 }, { "epoch": 2.99, "learning_rate": 1.5407133502811803e-07, "loss": 0.0, "step": 12941 }, { "epoch": 2.99, "learning_rate": 1.5021955165241507e-07, "loss": 0.0, "step": 12942 }, { "epoch": 2.99, "learning_rate": 1.4636776827671212e-07, "loss": 0.0, "step": 12943 }, { "epoch": 2.99, "learning_rate": 1.4251598490100917e-07, "loss": 0.0, "step": 12944 }, { "epoch": 2.99, "learning_rate": 1.3866420152530624e-07, "loss": 0.0, "step": 12945 }, { "epoch": 2.99, "learning_rate": 1.3481241814960326e-07, "loss": 0.0, "step": 12946 }, { "epoch": 2.99, "learning_rate": 1.3096063477390033e-07, "loss": 0.0, "step": 12947 }, { "epoch": 2.99, "learning_rate": 1.2710885139819735e-07, "loss": 0.0, "step": 12948 }, { "epoch": 2.99, "learning_rate": 1.2325706802249443e-07, "loss": 0.0, "step": 12949 }, { "epoch": 2.99, "learning_rate": 1.1940528464679147e-07, "loss": 0.0, "step": 12950 }, { "epoch": 2.99, "learning_rate": 1.1555350127108852e-07, "loss": 0.0, "step": 12951 }, { "epoch": 2.99, "learning_rate": 1.1170171789538558e-07, "loss": 0.0, "step": 12952 }, { "epoch": 2.99, "learning_rate": 1.0784993451968261e-07, "loss": 0.0, "step": 12953 }, { "epoch": 2.99, "learning_rate": 1.0399815114397967e-07, "loss": 0.0, "step": 12954 }, { "epoch": 2.99, "learning_rate": 1.0014636776827671e-07, "loss": 0.0, "step": 12955 }, { "epoch": 2.99, "learning_rate": 9.629458439257377e-08, "loss": 0.0, "step": 12956 }, { "epoch": 2.99, "learning_rate": 9.244280101687081e-08, "loss": 0.0, "step": 12957 }, { "epoch": 2.99, "learning_rate": 8.859101764116786e-08, "loss": 0.0, "step": 12958 }, { "epoch": 2.99, "learning_rate": 8.473923426546492e-08, "loss": 0.0, "step": 12959 }, { "epoch": 3.0, "learning_rate": 8.088745088976197e-08, "loss": 0.0, "step": 12960 }, { "epoch": 3.0, "learning_rate": 7.703566751405901e-08, "loss": 0.0, "step": 12961 }, { "epoch": 3.0, "learning_rate": 7.318388413835606e-08, "loss": 0.0, "step": 12962 }, { "epoch": 3.0, "learning_rate": 6.933210076265312e-08, "loss": 0.0, "step": 12963 }, { "epoch": 3.0, "learning_rate": 6.548031738695017e-08, "loss": 0.0, "step": 12964 }, { "epoch": 3.0, "learning_rate": 6.162853401124721e-08, "loss": 0.0, "step": 12965 }, { "epoch": 3.0, "learning_rate": 5.777675063554426e-08, "loss": 0.0, "step": 12966 }, { "epoch": 3.0, "learning_rate": 5.392496725984131e-08, "loss": 0.0, "step": 12967 }, { "epoch": 3.0, "learning_rate": 5.0073183884138354e-08, "loss": 0.0, "step": 12968 }, { "epoch": 3.0, "learning_rate": 4.622140050843541e-08, "loss": 0.0, "step": 12969 }, { "epoch": 3.0, "learning_rate": 4.236961713273246e-08, "loss": 0.0, "step": 12970 }, { "epoch": 3.0, "learning_rate": 3.851783375702951e-08, "loss": 0.0, "step": 12971 }, { "epoch": 3.0, "learning_rate": 3.466605038132656e-08, "loss": 0.0, "step": 12972 }, { "epoch": 3.0, "learning_rate": 3.081426700562361e-08, "loss": 0.0001, "step": 12973 }, { "epoch": 3.0, "learning_rate": 2.6962483629920654e-08, "loss": 0.0, "step": 12974 }, { "epoch": 3.0, "learning_rate": 2.3110700254217703e-08, "loss": 0.0, "step": 12975 }, { "epoch": 3.0, "learning_rate": 1.9258916878514753e-08, "loss": 0.0, "step": 12976 }, { "epoch": 3.0, "learning_rate": 1.5407133502811803e-08, "loss": 0.0, "step": 12977 }, { "epoch": 3.0, "learning_rate": 1.1555350127108852e-08, "loss": 0.0, "step": 12978 }, { "epoch": 3.0, "learning_rate": 7.703566751405902e-09, "loss": 0.0, "step": 12979 }, { "epoch": 3.0, "learning_rate": 3.851783375702951e-09, "loss": 0.0, "step": 12980 }, { "epoch": 3.0, "learning_rate": 0.0, "loss": 0.0, "step": 12981 }, { "epoch": 3.0, "step": 12981, "total_flos": 3.306560111709696e+16, "train_loss": 0.00022401868770512018, "train_runtime": 3245.5993, "train_samples_per_second": 127.965, "train_steps_per_second": 4.0 } ], "max_steps": 12981, "num_train_epochs": 3, "total_flos": 3.306560111709696e+16, "trial_name": null, "trial_params": null }